* [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 17:16 ` Jeffrey E Altman
2023-11-09 22:06 ` David Howells
2023-11-09 15:39 ` [PATCH 02/41] rxrpc: Fix two connection reaping bugs David Howells
` (39 subsequent siblings)
40 siblings, 2 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Fix RTT determination to use a PING ACK that is in marked as being in
response to a DATA packet as the response from which RTT can be calculated
from in lieu of a REQUESTED ACK. The server may send the latter instead of
the former.
Fixes: 4700c4d80b7b ("rxrpc: Fix loss of RTT samples due to interposed ACK")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
net/rxrpc/input.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c
index 030d64f282f3..fc0d404f3b91 100644
--- a/net/rxrpc/input.c
+++ b/net/rxrpc/input.c
@@ -806,6 +806,10 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb)
rxrpc_complete_rtt_probe(call, skb->tstamp, acked_serial, ack_serial,
rxrpc_rtt_rx_ping_response);
break;
+ case RXRPC_ACK_PING:
+ if (acked_serial == 0)
+ break;
+ fallthrough;
case RXRPC_ACK_REQUESTED:
rxrpc_complete_rtt_probe(call, skb->tstamp, acked_serial, ack_serial,
rxrpc_rtt_rx_requested_ack);
^ permalink raw reply related [flat|nested] 57+ messages in thread* Re: [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-09 15:39 ` [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source David Howells
@ 2023-11-09 17:16 ` Jeffrey E Altman
2023-11-09 22:06 ` David Howells
1 sibling, 0 replies; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-09 17:16 UTC (permalink / raw)
To: David Howells, Marc Dionne; +Cc: linux-afs, linux-fsdevel, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 2454 bytes --]
On 11/9/2023 10:39 AM, David Howells wrote:
> Fix RTT determination to use a PING ACK that is in marked as being in
> response to a DATA packet as the response from which RTT can be calculated
> from in lieu of a REQUESTED ACK. The server may send the latter instead of
> the former.
>
> Fixes: 4700c4d80b7b ("rxrpc: Fix loss of RTT samples due to interposed ACK")
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Marc Dionne <marc.dionne@auristor.com>
> cc: linux-afs@lists.infradead.org
> ---
> net/rxrpc/input.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c
> index 030d64f282f3..fc0d404f3b91 100644
> --- a/net/rxrpc/input.c
> +++ b/net/rxrpc/input.c
> @@ -806,6 +806,10 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb)
> rxrpc_complete_rtt_probe(call, skb->tstamp, acked_serial, ack_serial,
> rxrpc_rtt_rx_ping_response);
> break;
> + case RXRPC_ACK_PING:
> + if (acked_serial == 0)
> + break;
> + fallthrough;
> case RXRPC_ACK_REQUESTED:
> rxrpc_complete_rtt_probe(call, skb->tstamp, acked_serial, ack_serial,
> rxrpc_rtt_rx_requested_ack);
David,
I do not believe the ack_reason matters within rxrpc_input_ack(). As
long as the acked_serial is non-zero,
rxrpc_complete_rtt_probe() can be called to attempt to compute an RTT.
If there is an exact match for the
acked_serial then an RTT can be computed and if acked_serial is later
than the pending rtt probe, the probe
can be abandoned with the following caveats.
1. Receiving an acked_serial that is later than the serial of the
transmitted probe indicates that a packet
transmitted after the probe was received first. Or that reordering
of the transmitted packets occurred.
Or that the probe was never received by the peer; or that the peer's
response to the probe was lost in
transit.
2. The serial number namespace is unsigned 32-bit shared across all of
the call channels of the associated
rx connection. As the serial numbers will wrap the use of after()
within rxrpc_complete_rtt_probe to
compare their values is questionable. If serial numbers will be
compared in this manner then they
need to be locally tracked and compared as unsigned 64-bit values
where only the low 32-bits are
transmitted on the wire and any wire serial number equal to zero is
ignored.
Jeffrey Altman
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread* Re: [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-09 15:39 ` [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source David Howells
2023-11-09 17:16 ` Jeffrey E Altman
@ 2023-11-09 22:06 ` David Howells
2023-11-10 14:15 ` Jeffrey E Altman
2023-11-10 17:25 ` David Howells
1 sibling, 2 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 22:06 UTC (permalink / raw)
To: Jeffrey E Altman
Cc: dhowells, Marc Dionne, linux-afs, linux-fsdevel, linux-kernel
I'm going to drop this patch from this series for now and send it separately.
> I do not believe the ack_reason matters within rxrpc_input_ack(). As long as
> the acked_serial is non-zero,
> rxrpc_complete_rtt_probe() can be called to attempt to compute an RTT. If
> there is an exact match for the
> acked_serial then an RTT can be computed and if acked_serial is later than the
> pending rtt probe, the probe
> can be abandoned with the following caveats.
>
> 1. Receiving an acked_serial that is later than the serial of the
> transmitted probe indicates that a packet
> transmitted after the probe was received first. Or that reordering
> of the transmitted packets occurred.
> Or that the probe was never received by the peer; or that the peer's
> response to the probe was lost in
> transit.
> 2. The serial number namespace is unsigned 32-bit shared across all of
> the call channels of the associated
> rx connection. As the serial numbers will wrap the use of after()
> within rxrpc_complete_rtt_probe to
> compare their values is questionable. If serial numbers will be
> compared in this manner then they
> need to be locally tracked and compared as unsigned 64-bit values
> where only the low 32-bits are
> transmitted on the wire and any wire serial number equal to zero is
> ignored.
I do ignore ack.serial == 0 for this purpose.
I'm not sure how expanding it internally to 64-bits actually helps since the
upper 32 bits is not visible to the peer.
David
^ permalink raw reply [flat|nested] 57+ messages in thread
* Re: [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-09 22:06 ` David Howells
@ 2023-11-10 14:15 ` Jeffrey E Altman
2023-11-10 16:12 ` Jeffrey E Altman
2023-11-10 17:25 ` David Howells
1 sibling, 1 reply; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-10 14:15 UTC (permalink / raw)
To: David Howells; +Cc: Marc Dionne, linux-afs, linux-fsdevel, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 2874 bytes --]
On 11/9/2023 5:06 PM, David Howells wrote:
> I do not believe the ack_reason matters within rxrpc_input_ack(). As
> long as
>> the acked_serial is non-zero,
>> rxrpc_complete_rtt_probe() can be called to attempt to compute an RTT. If
>> there is an exact match for the
>> acked_serial then an RTT can be computed and if acked_serial is later than the
>> pending rtt probe, the probe
>> can be abandoned with the following caveats.
>>
>> 1. Receiving an acked_serial that is later than the serial of the
>> transmitted probe indicates that a packet
>> transmitted after the probe was received first. Or that reordering
>> of the transmitted packets occurred.
>> Or that the probe was never received by the peer; or that the peer's
>> response to the probe was lost in
>> transit.
>> 2. The serial number namespace is unsigned 32-bit shared across all of
>> the call channels of the associated
>> rx connection. As the serial numbers will wrap the use of after()
>> within rxrpc_complete_rtt_probe to
>> compare their values is questionable. If serial numbers will be
>> compared in this manner then they
>> need to be locally tracked and compared as unsigned 64-bit values
>> where only the low 32-bits are
>> transmitted on the wire and any wire serial number equal to zero is
>> ignored.
> I do ignore ack.serial == 0 for this purpose.
Zero has the special meaning - this ACK is not explicitly in response to
a received packet.
However, as mentioned, the serial number counter wraps frequently and
most RxRPC implementations
do not transition from serial 0xffffffff -> 0x00000001 when wrapping.
> I'm not sure how expanding it internally to 64-bits actually helps since the
> upper 32 bits is not visible to the peer.
rxrpc_complete_rtt_probe contains the following logic that relies on
after() being able to detect
a serial number wrap.
/* If a later serial is being acked, then mark this slot as
* being available.
*/
if (after(acked_serial, orig_serial)) {
trace_rxrpc_rtt_rx(call, rxrpc_rtt_rx_obsolete, i,
orig_serial, acked_serial, 0, 0);
clear_bit(i + RXRPC_CALL_RTT_PEND_SHIFT, &call->rtt_avail);
smp_wmb();
set_bit(i, &call->rtt_avail);
}
Otherwise, acked_serial = 0x01 will be considered smaller than
orig_serial = 0xfffffffe and the slot will not be marked available.
I will note that there is a similar problem with rxrpc_seq_t values
which are u32 on the wire but which will wrap for calls that transmit
more than approximately 5.5TB of data. Calls of this size are unlikely
for a cache manager but are common for any service transmitting volume
dumps.
Jeffrey Altman
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread* Re: [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-10 14:15 ` Jeffrey E Altman
@ 2023-11-10 16:12 ` Jeffrey E Altman
0 siblings, 0 replies; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-10 16:12 UTC (permalink / raw)
To: David Howells; +Cc: Marc Dionne, linux-afs, linux-fsdevel, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 1783 bytes --]
On 11/10/2023 9:15 AM, Jeffrey E Altman wrote:
>
>> I'm not sure how expanding it internally to 64-bits actually helps
>> since the
>> upper 32 bits is not visible to the peer.
> rxrpc_complete_rtt_probe contains the following logic that relies on
> after() being able to detect
> a serial number wrap.
>
> /* If a later serial is being acked, then mark this slot as
> * being available.
> */
>
> if (after(acked_serial, orig_serial)) {
> trace_rxrpc_rtt_rx(call, rxrpc_rtt_rx_obsolete, i,
> orig_serial, acked_serial, 0, 0);
> clear_bit(i + RXRPC_CALL_RTT_PEND_SHIFT, &call->rtt_avail);
> smp_wmb();
> set_bit(i, &call->rtt_avail);
> }
>
> Otherwise, acked_serial = 0x01 will be considered smaller than
> orig_serial = 0xfffffffe and the slot will not be marked available.
>
> I will note that there is a similar problem with rxrpc_seq_t values
> which are u32 on the wire but which will wrap for calls that transmit
> more than approximately 5.5TB of data. Calls of this size are unlikely
> for a cache manager but are common for any service transmitting volume
> dumps.
I misread the definition of after() which is
static inline bool after(u32 seq1, u32 seq2)
{
return (s32)(seq1 - seq2) > 0;
}
This is sufficient to detect the serial number and sequence number wrapping.
What it doesn't provide is a method for rxrpc tracing and /proc file to
report on the total number of packets that have been sent or received on
a call or connection which might or might not be important depending
upon the use case.
Jeffrey Altman
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread
* Re: [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-09 22:06 ` David Howells
2023-11-10 14:15 ` Jeffrey E Altman
@ 2023-11-10 17:25 ` David Howells
2023-11-10 21:52 ` Jeffrey E Altman
2023-11-10 21:54 ` Jeffrey E Altman
1 sibling, 2 replies; 57+ messages in thread
From: David Howells @ 2023-11-10 17:25 UTC (permalink / raw)
To: Jeffrey E Altman
Cc: dhowells, Marc Dionne, linux-afs, linux-fsdevel, linux-kernel
Jeffrey E Altman <jaltman@auristor.com> wrote:
> > I do ignore ack.serial == 0 for this purpose.
>
> Zero has the special meaning - this ACK is not explicitly in response to a
> received packet.
>
> However, as mentioned, the serial number counter wraps frequently and most
> RxRPC implementations
> do not transition from serial 0xffffffff -> 0x00000001 when wrapping.
I don't skip zero serial numbers either. I'm not sure whether it would be
better to do so.
> Otherwise, acked_serial = 0x01 will be considered smaller than orig_serial =
> 0xfffffffe and the slot will not be marked available.
As you mentioned in your follow up email, after() deals with that by casting
to signed, subtracting and then examining the result.
David
^ permalink raw reply [flat|nested] 57+ messages in thread
* Re: [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-10 17:25 ` David Howells
@ 2023-11-10 21:52 ` Jeffrey E Altman
2023-11-10 21:54 ` Jeffrey E Altman
1 sibling, 0 replies; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-10 21:52 UTC (permalink / raw)
To: David Howells; +Cc: Marc Dionne, linux-afs, linux-fsdevel, linux-kernel
[-- Attachment #1.1: Type: text/plain, Size: 827 bytes --]
On 11/10/2023 12:25 PM, David Howells wrote:
> Jeffrey E Altman<jaltman@auristor.com> wrote:
>
>>> I do ignore ack.serial == 0 for this purpose.
>> Zero has the special meaning - this ACK is not explicitly in response to a
>> received packet.
>>
>> However, as mentioned, the serial number counter wraps frequently and most
>> RxRPC implementations
>> do not transition from serial 0xffffffff -> 0x00000001 when wrapping.
> I don't skip zero serial numbers either. I'm not sure whether it would be
> better to do so.
If a DATA packet is sent with serial number zero and an ACK packet is
sent in response to it
with the ack.serial field set to the DATA packet serial number (zero),
then the receiver of the
ACK will be unable to compute an RTT from that DATA packet. It will
happen rarely but it
will happen.
Jeffrey
[-- Attachment #1.2: Type: text/html, Size: 1596 bytes --]
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread
* Re: [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source
2023-11-10 17:25 ` David Howells
2023-11-10 21:52 ` Jeffrey E Altman
@ 2023-11-10 21:54 ` Jeffrey E Altman
1 sibling, 0 replies; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-10 21:54 UTC (permalink / raw)
To: David Howells; +Cc: Marc Dionne, linux-afs, linux-fsdevel, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 828 bytes --]
On 11/10/2023 12:25 PM, David Howells wrote:
> Jeffrey E Altman<jaltman@auristor.com> wrote:
>
>>> I do ignore ack.serial == 0 for this purpose.
>> Zero has the special meaning - this ACK is not explicitly in response to a
>> received packet.
>>
>> However, as mentioned, the serial number counter wraps frequently and most
>> RxRPC implementations
>> do not transition from serial 0xffffffff -> 0x00000001 when wrapping.
> I don't skip zero serial numbers either. I'm not sure whether it would be
> better to do so.
If a DATA packet is sent with serial number zero and an ACK packet is
sent in response to it
with the ack.serial field set to the DATA packet serial number (zero),
then the receiver of the
ACK will be unable to compute an RTT from that DATA packet. It will
happen rarely but it
will happen.
Jeffrey
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread
* [PATCH 02/41] rxrpc: Fix two connection reaping bugs
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
2023-11-09 15:39 ` [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 17:27 ` Jeffrey E Altman
2023-11-09 17:50 ` patchwork-bot+netdevbpf
2023-11-09 15:39 ` [PATCH 03/41] rxrpc: Fix some minor issues with bundle tracing David Howells
` (38 subsequent siblings)
40 siblings, 2 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne
Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel,
David S. Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
netdev
Fix two connection reaping bugs:
(1) rxrpc_connection_expiry is in units of seconds, so
rxrpc_disconnect_call() needs to multiply it by HZ when adding it to
jiffies.
(2) rxrpc_client_conn_reap_timeout() should set RXRPC_CLIENT_REAP_TIMER if
local->kill_all_client_conns is clear, not if it is set (in which case
we don't need the timer). Without this, old client connections don't
get cleaned up until the local endpoint is cleaned up.
Fixes: 5040011d073d ("rxrpc: Make the local endpoint hold a ref on a connected call")
Fixes: 0d6bf319bc5a ("rxrpc: Move the client conn cache management to the I/O thread")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: netdev@vger.kernel.org
cc: linux-afs@lists.infradead.org
---
net/rxrpc/conn_object.c | 2 +-
net/rxrpc/local_object.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/rxrpc/conn_object.c b/net/rxrpc/conn_object.c
index ac85d4644a3c..df8a271948a1 100644
--- a/net/rxrpc/conn_object.c
+++ b/net/rxrpc/conn_object.c
@@ -212,7 +212,7 @@ void rxrpc_disconnect_call(struct rxrpc_call *call)
conn->idle_timestamp = jiffies;
if (atomic_dec_and_test(&conn->active))
rxrpc_set_service_reap_timer(conn->rxnet,
- jiffies + rxrpc_connection_expiry);
+ jiffies + rxrpc_connection_expiry * HZ);
}
rxrpc_put_call(call, rxrpc_call_put_io_thread);
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 7d910aee4f8c..c553a30e9c83 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -87,7 +87,7 @@ static void rxrpc_client_conn_reap_timeout(struct timer_list *timer)
struct rxrpc_local *local =
container_of(timer, struct rxrpc_local, client_conn_reap_timer);
- if (local->kill_all_client_conns &&
+ if (!local->kill_all_client_conns &&
test_and_set_bit(RXRPC_CLIENT_CONN_REAP_TIMER, &local->client_conn_flags))
rxrpc_wake_up_io_thread(local);
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* Re: [PATCH 02/41] rxrpc: Fix two connection reaping bugs
2023-11-09 15:39 ` [PATCH 02/41] rxrpc: Fix two connection reaping bugs David Howells
@ 2023-11-09 17:27 ` Jeffrey E Altman
2023-11-09 17:50 ` patchwork-bot+netdevbpf
1 sibling, 0 replies; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-09 17:27 UTC (permalink / raw)
To: David Howells, Marc Dionne
Cc: linux-afs, linux-fsdevel, linux-kernel, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, netdev
[-- Attachment #1: Type: text/plain, Size: 1098 bytes --]
On 11/9/2023 10:39 AM, David Howells wrote:
> Fix two connection reaping bugs:
>
> (1) rxrpc_connection_expiry is in units of seconds, so
> rxrpc_disconnect_call() needs to multiply it by HZ when adding it to
> jiffies.
>
> (2) rxrpc_client_conn_reap_timeout() should set RXRPC_CLIENT_REAP_TIMER if
> local->kill_all_client_conns is clear, not if it is set (in which case
> we don't need the timer). Without this, old client connections don't
> get cleaned up until the local endpoint is cleaned up.
>
> Fixes: 5040011d073d ("rxrpc: Make the local endpoint hold a ref on a connected call")
> Fixes: 0d6bf319bc5a ("rxrpc: Move the client conn cache management to the I/O thread")
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Marc Dionne <marc.dionne@auristor.com>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: netdev@vger.kernel.org
> cc: linux-afs@lists.infradead.org
> ---
Reviewed-by: Jeffrey Altman <jaltman@auristor.com>
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread* Re: [PATCH 02/41] rxrpc: Fix two connection reaping bugs
2023-11-09 15:39 ` [PATCH 02/41] rxrpc: Fix two connection reaping bugs David Howells
2023-11-09 17:27 ` Jeffrey E Altman
@ 2023-11-09 17:50 ` patchwork-bot+netdevbpf
1 sibling, 0 replies; 57+ messages in thread
From: patchwork-bot+netdevbpf @ 2023-11-09 17:50 UTC (permalink / raw)
To: David Howells
Cc: marc.dionne, linux-afs, linux-fsdevel, linux-kernel, davem,
edumazet, kuba, pabeni, netdev
Hello:
This patch was applied to bpf/bpf.git (master)
by Jakub Kicinski <kuba@kernel.org>:
On Thu, 9 Nov 2023 15:39:25 +0000 you wrote:
> Fix two connection reaping bugs:
>
> (1) rxrpc_connection_expiry is in units of seconds, so
> rxrpc_disconnect_call() needs to multiply it by HZ when adding it to
> jiffies.
>
> (2) rxrpc_client_conn_reap_timeout() should set RXRPC_CLIENT_REAP_TIMER if
> local->kill_all_client_conns is clear, not if it is set (in which case
> we don't need the timer). Without this, old client connections don't
> get cleaned up until the local endpoint is cleaned up.
>
> [...]
Here is the summary with links:
- [02/41] rxrpc: Fix two connection reaping bugs
https://git.kernel.org/bpf/bpf/c/61e4a8660002
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
^ permalink raw reply [flat|nested] 57+ messages in thread
* [PATCH 03/41] rxrpc: Fix some minor issues with bundle tracing
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
2023-11-09 15:39 ` [PATCH 01/41] rxrpc: Fix RTT determination to use PING ACKs as a source David Howells
2023-11-09 15:39 ` [PATCH 02/41] rxrpc: Fix two connection reaping bugs David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 04/41] afs: Fix afs_server_list to be cleaned up with RCU David Howells
` (37 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Fix some superficial issues with the tracing of rxrpc_bundle structs,
including:
(1) Set the debug_id when the bundle is allocated rather than when it is
set up so that the "NEW" trace line displays the correct bundle ID.
(2) Show the refcount when emitting the "FREE" traceline.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
net/rxrpc/conn_client.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/net/rxrpc/conn_client.c b/net/rxrpc/conn_client.c
index 981ca5b98bcb..1d95f8bc769f 100644
--- a/net/rxrpc/conn_client.c
+++ b/net/rxrpc/conn_client.c
@@ -73,6 +73,7 @@ static void rxrpc_destroy_client_conn_ids(struct rxrpc_local *local)
static struct rxrpc_bundle *rxrpc_alloc_bundle(struct rxrpc_call *call,
gfp_t gfp)
{
+ static atomic_t rxrpc_bundle_id;
struct rxrpc_bundle *bundle;
bundle = kzalloc(sizeof(*bundle), gfp);
@@ -85,6 +86,7 @@ static struct rxrpc_bundle *rxrpc_alloc_bundle(struct rxrpc_call *call,
bundle->upgrade = test_bit(RXRPC_CALL_UPGRADE, &call->flags);
bundle->service_id = call->dest_srx.srx_service;
bundle->security_level = call->security_level;
+ bundle->debug_id = atomic_inc_return(&rxrpc_bundle_id);
refcount_set(&bundle->ref, 1);
atomic_set(&bundle->active, 1);
INIT_LIST_HEAD(&bundle->waiting_calls);
@@ -105,7 +107,8 @@ struct rxrpc_bundle *rxrpc_get_bundle(struct rxrpc_bundle *bundle,
static void rxrpc_free_bundle(struct rxrpc_bundle *bundle)
{
- trace_rxrpc_bundle(bundle->debug_id, 1, rxrpc_bundle_free);
+ trace_rxrpc_bundle(bundle->debug_id, refcount_read(&bundle->ref),
+ rxrpc_bundle_free);
rxrpc_put_peer(bundle->peer, rxrpc_peer_put_bundle);
key_put(bundle->key);
kfree(bundle);
@@ -239,7 +242,6 @@ static bool rxrpc_may_reuse_conn(struct rxrpc_connection *conn)
*/
int rxrpc_look_up_bundle(struct rxrpc_call *call, gfp_t gfp)
{
- static atomic_t rxrpc_bundle_id;
struct rxrpc_bundle *bundle, *candidate;
struct rxrpc_local *local = call->local;
struct rb_node *p, **pp, *parent;
@@ -306,7 +308,6 @@ int rxrpc_look_up_bundle(struct rxrpc_call *call, gfp_t gfp)
}
_debug("new bundle");
- candidate->debug_id = atomic_inc_return(&rxrpc_bundle_id);
rb_link_node(&candidate->local_node, parent, pp);
rb_insert_color(&candidate->local_node, &local->client_bundles);
call->bundle = rxrpc_get_bundle(candidate, rxrpc_bundle_get_client_call);
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 04/41] afs: Fix afs_server_list to be cleaned up with RCU
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (2 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 03/41] rxrpc: Fix some minor issues with bundle tracing David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 05/41] afs: Make error on cell lookup failure consistent with OpenAFS David Howells
` (36 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
afs_server_list is accessed with the rcu_read_lock() held from
volume->servers, so it needs to be cleaned up correctly.
Fix this by using kfree_rcu() instead of kfree().
Fixes: 8a070a964877 ("afs: Detect cell aliases 1 - Cells with root volumes")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/internal.h | 1 +
fs/afs/server_list.c | 2 +-
2 files changed, 2 insertions(+), 1 deletion(-)
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index da73b97e19a9..5041eae64423 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -553,6 +553,7 @@ struct afs_server_entry {
};
struct afs_server_list {
+ struct rcu_head rcu;
afs_volid_t vids[AFS_MAXTYPES]; /* Volume IDs */
refcount_t usage;
unsigned char nr_servers;
diff --git a/fs/afs/server_list.c b/fs/afs/server_list.c
index ed9056703505..b59896b1de0a 100644
--- a/fs/afs/server_list.c
+++ b/fs/afs/server_list.c
@@ -17,7 +17,7 @@ void afs_put_serverlist(struct afs_net *net, struct afs_server_list *slist)
for (i = 0; i < slist->nr_servers; i++)
afs_unuse_server(net, slist->servers[i].server,
afs_server_trace_put_slist);
- kfree(slist);
+ kfree_rcu(slist, rcu);
}
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 05/41] afs: Make error on cell lookup failure consistent with OpenAFS
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (3 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 04/41] afs: Fix afs_server_list to be cleaned up with RCU David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 06/41] afs: Remove whitespace before most ')' from the trace header David Howells
` (35 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne
Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel,
Markus Suvanto
When kafs tries to look up a cell in the DNS or the local config, it will
translate a lookup failure into EDESTADDRREQ whereas OpenAFS translates it
into ENOENT. Applications such as West expect the latter behaviour and
fail if they see the former.
This can be seen by trying to mount an unknown cell:
# mount -t afs %example.com:cell.root /mnt
mount: /mnt: mount(2) system call failed: Destination address required.
Reported-by: Markus Suvanto <markus.suvanto@gmail.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216637
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/dynroot.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/fs/afs/dynroot.c b/fs/afs/dynroot.c
index 95bcbd7654d1..8081d68004d0 100644
--- a/fs/afs/dynroot.c
+++ b/fs/afs/dynroot.c
@@ -132,8 +132,8 @@ static int afs_probe_cell_name(struct dentry *dentry)
ret = dns_query(net->net, "afsdb", name, len, "srv=1",
NULL, NULL, false);
- if (ret == -ENODATA)
- ret = -EDESTADDRREQ;
+ if (ret == -ENODATA || ret == -ENOKEY)
+ ret = -ENOENT;
return ret;
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 06/41] afs: Remove whitespace before most ')' from the trace header
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (4 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 05/41] afs: Make error on cell lookup failure consistent with OpenAFS David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 07/41] afs: Automatically generate trace tag enums David Howells
` (34 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
checkpatch objects to whitespace before ')', so remove most of it from the
afs trace header.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
include/trace/events/afs.h | 242 ++++++++++++++++++-------------------
1 file changed, 121 insertions(+), 121 deletions(-)
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index e9d412d19dbb..cfcd6452c156 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -654,12 +654,12 @@ TRACE_EVENT(afs_receive_data,
TP_ARGS(call, iter, want_more, ret),
TP_STRUCT__entry(
- __field(loff_t, remain )
- __field(unsigned int, call )
- __field(enum afs_call_state, state )
- __field(unsigned short, unmarshall )
- __field(bool, want_more )
- __field(int, ret )
+ __field(loff_t, remain)
+ __field(unsigned int, call)
+ __field(enum afs_call_state, state)
+ __field(unsigned short, unmarshall)
+ __field(bool, want_more)
+ __field(int, ret)
),
TP_fast_assign(
@@ -686,9 +686,9 @@ TRACE_EVENT(afs_notify_call,
TP_ARGS(rxcall, call),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(enum afs_call_state, state )
- __field(unsigned short, unmarshall )
+ __field(unsigned int, call)
+ __field(enum afs_call_state, state)
+ __field(unsigned short, unmarshall)
),
TP_fast_assign(
@@ -708,9 +708,9 @@ TRACE_EVENT(afs_cb_call,
TP_ARGS(call),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(u32, op )
- __field(u16, service_id )
+ __field(unsigned int, call)
+ __field(u32, op)
+ __field(u16, service_id)
),
TP_fast_assign(
@@ -733,11 +733,11 @@ TRACE_EVENT(afs_call,
TP_ARGS(call_debug_id, op, ref, outstanding, where),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(int, op )
- __field(int, ref )
- __field(int, outstanding )
- __field(const void *, where )
+ __field(unsigned int, call)
+ __field(int, op)
+ __field(int, ref)
+ __field(int, outstanding)
+ __field(const void *, where)
),
TP_fast_assign(
@@ -762,9 +762,9 @@ TRACE_EVENT(afs_make_fs_call,
TP_ARGS(call, fid),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(enum afs_fs_operation, op )
- __field_struct(struct afs_fid, fid )
+ __field(unsigned int, call)
+ __field(enum afs_fs_operation, op)
+ __field_struct(struct afs_fid, fid)
),
TP_fast_assign(
@@ -794,10 +794,10 @@ TRACE_EVENT(afs_make_fs_calli,
TP_ARGS(call, fid, i),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(unsigned int, i )
- __field(enum afs_fs_operation, op )
- __field_struct(struct afs_fid, fid )
+ __field(unsigned int, call)
+ __field(unsigned int, i)
+ __field(enum afs_fs_operation, op)
+ __field_struct(struct afs_fid, fid)
),
TP_fast_assign(
@@ -829,10 +829,10 @@ TRACE_EVENT(afs_make_fs_call1,
TP_ARGS(call, fid, name),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(enum afs_fs_operation, op )
- __field_struct(struct afs_fid, fid )
- __array(char, name, 24 )
+ __field(unsigned int, call)
+ __field(enum afs_fs_operation, op)
+ __field_struct(struct afs_fid, fid)
+ __array(char, name, 24)
),
TP_fast_assign(
@@ -866,11 +866,11 @@ TRACE_EVENT(afs_make_fs_call2,
TP_ARGS(call, fid, name, name2),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(enum afs_fs_operation, op )
- __field_struct(struct afs_fid, fid )
- __array(char, name, 24 )
- __array(char, name2, 24 )
+ __field(unsigned int, call)
+ __field(enum afs_fs_operation, op)
+ __field_struct(struct afs_fid, fid)
+ __array(char, name, 24)
+ __array(char, name2, 24)
),
TP_fast_assign(
@@ -907,8 +907,8 @@ TRACE_EVENT(afs_make_vl_call,
TP_ARGS(call),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(enum afs_vl_operation, op )
+ __field(unsigned int, call)
+ __field(enum afs_vl_operation, op)
),
TP_fast_assign(
@@ -927,10 +927,10 @@ TRACE_EVENT(afs_call_done,
TP_ARGS(call),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(struct rxrpc_call *, rx_call )
- __field(int, ret )
- __field(u32, abort_code )
+ __field(unsigned int, call)
+ __field(struct rxrpc_call *, rx_call)
+ __field(int, ret)
+ __field(u32, abort_code)
),
TP_fast_assign(
@@ -953,10 +953,10 @@ TRACE_EVENT(afs_send_data,
TP_ARGS(call, msg),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(unsigned int, flags )
- __field(loff_t, offset )
- __field(loff_t, count )
+ __field(unsigned int, call)
+ __field(unsigned int, flags)
+ __field(loff_t, offset)
+ __field(loff_t, count)
),
TP_fast_assign(
@@ -977,10 +977,10 @@ TRACE_EVENT(afs_sent_data,
TP_ARGS(call, msg, ret),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(int, ret )
- __field(loff_t, offset )
- __field(loff_t, count )
+ __field(unsigned int, call)
+ __field(int, ret)
+ __field(loff_t, offset)
+ __field(loff_t, count)
),
TP_fast_assign(
@@ -1001,9 +1001,9 @@ TRACE_EVENT(afs_dir_check_failed,
TP_ARGS(vnode, off, i_size),
TP_STRUCT__entry(
- __field(struct afs_vnode *, vnode )
- __field(loff_t, off )
- __field(loff_t, i_size )
+ __field(struct afs_vnode *, vnode)
+ __field(loff_t, off)
+ __field(loff_t, i_size)
),
TP_fast_assign(
@@ -1022,11 +1022,11 @@ TRACE_EVENT(afs_folio_dirty,
TP_ARGS(vnode, where, folio),
TP_STRUCT__entry(
- __field(struct afs_vnode *, vnode )
- __field(const char *, where )
- __field(pgoff_t, index )
- __field(unsigned long, from )
- __field(unsigned long, to )
+ __field(struct afs_vnode *, vnode)
+ __field(const char *, where)
+ __field(pgoff_t, index)
+ __field(unsigned long, from)
+ __field(unsigned long, to)
),
TP_fast_assign(
@@ -1056,11 +1056,11 @@ TRACE_EVENT(afs_call_state,
TP_ARGS(call, from, to, ret, remote_abort),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(enum afs_call_state, from )
- __field(enum afs_call_state, to )
- __field(int, ret )
- __field(u32, abort )
+ __field(unsigned int, call)
+ __field(enum afs_call_state, from)
+ __field(enum afs_call_state, to)
+ __field(int, ret)
+ __field(u32, abort)
),
TP_fast_assign(
@@ -1084,9 +1084,9 @@ TRACE_EVENT(afs_lookup,
TP_ARGS(dvnode, name, fid),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, dfid )
- __field_struct(struct afs_fid, fid )
- __array(char, name, 24 )
+ __field_struct(struct afs_fid, dfid)
+ __field_struct(struct afs_fid, fid)
+ __array(char, name, 24)
),
TP_fast_assign(
@@ -1116,15 +1116,15 @@ TRACE_EVENT(afs_edit_dir,
TP_ARGS(dvnode, why, op, block, slot, f_vnode, f_unique, name),
TP_STRUCT__entry(
- __field(unsigned int, vnode )
- __field(unsigned int, unique )
- __field(enum afs_edit_dir_reason, why )
- __field(enum afs_edit_dir_op, op )
- __field(unsigned int, block )
- __field(unsigned short, slot )
- __field(unsigned int, f_vnode )
- __field(unsigned int, f_unique )
- __array(char, name, 24 )
+ __field(unsigned int, vnode)
+ __field(unsigned int, unique)
+ __field(enum afs_edit_dir_reason, why)
+ __field(enum afs_edit_dir_op, op)
+ __field(unsigned int, block)
+ __field(unsigned short, slot)
+ __field(unsigned int, f_vnode)
+ __field(unsigned int, f_unique)
+ __array(char, name, 24)
),
TP_fast_assign(
@@ -1157,8 +1157,8 @@ TRACE_EVENT(afs_protocol_error,
TP_ARGS(call, cause),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(enum afs_eproto_cause, cause )
+ __field(unsigned int, call)
+ __field(enum afs_eproto_cause, cause)
),
TP_fast_assign(
@@ -1177,9 +1177,9 @@ TRACE_EVENT(afs_io_error,
TP_ARGS(call, error, where),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(int, error )
- __field(enum afs_io_error, where )
+ __field(unsigned int, call)
+ __field(int, error)
+ __field(enum afs_io_error, where)
),
TP_fast_assign(
@@ -1199,9 +1199,9 @@ TRACE_EVENT(afs_file_error,
TP_ARGS(vnode, error, where),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, fid )
- __field(int, error )
- __field(enum afs_file_error, where )
+ __field_struct(struct afs_fid, fid)
+ __field(int, error)
+ __field(enum afs_file_error, where)
),
TP_fast_assign(
@@ -1222,9 +1222,9 @@ TRACE_EVENT(afs_cm_no_server,
TP_ARGS(call, srx),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(unsigned int, op_id )
- __field_struct(struct sockaddr_rxrpc, srx )
+ __field(unsigned int, call)
+ __field(unsigned int, op_id)
+ __field_struct(struct sockaddr_rxrpc, srx)
),
TP_fast_assign(
@@ -1243,9 +1243,9 @@ TRACE_EVENT(afs_cm_no_server_u,
TP_ARGS(call, uuid),
TP_STRUCT__entry(
- __field(unsigned int, call )
- __field(unsigned int, op_id )
- __field_struct(uuid_t, uuid )
+ __field(unsigned int, call)
+ __field(unsigned int, op_id)
+ __field_struct(uuid_t, uuid)
),
TP_fast_assign(
@@ -1265,11 +1265,11 @@ TRACE_EVENT(afs_flock_ev,
TP_ARGS(vnode, fl, event, error),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, fid )
- __field(enum afs_flock_event, event )
- __field(enum afs_lock_state, state )
- __field(int, error )
- __field(unsigned int, debug_id )
+ __field_struct(struct afs_fid, fid)
+ __field(enum afs_flock_event, event)
+ __field(enum afs_lock_state, state)
+ __field(int, error)
+ __field(unsigned int, debug_id)
),
TP_fast_assign(
@@ -1295,13 +1295,13 @@ TRACE_EVENT(afs_flock_op,
TP_ARGS(vnode, fl, op),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, fid )
- __field(loff_t, from )
- __field(loff_t, len )
- __field(enum afs_flock_operation, op )
- __field(unsigned char, type )
- __field(unsigned int, flags )
- __field(unsigned int, debug_id )
+ __field_struct(struct afs_fid, fid)
+ __field(loff_t, from)
+ __field(loff_t, len)
+ __field(enum afs_flock_operation, op)
+ __field(unsigned char, type)
+ __field(unsigned int, flags)
+ __field(unsigned int, debug_id)
),
TP_fast_assign(
@@ -1328,7 +1328,7 @@ TRACE_EVENT(afs_reload_dir,
TP_ARGS(vnode),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, fid )
+ __field_struct(struct afs_fid, fid)
),
TP_fast_assign(
@@ -1345,8 +1345,8 @@ TRACE_EVENT(afs_silly_rename,
TP_ARGS(vnode, done),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, fid )
- __field(bool, done )
+ __field_struct(struct afs_fid, fid)
+ __field(bool, done)
),
TP_fast_assign(
@@ -1365,9 +1365,9 @@ TRACE_EVENT(afs_get_tree,
TP_ARGS(cell, volume),
TP_STRUCT__entry(
- __field(u64, vid )
- __array(char, cell, 24 )
- __array(char, volume, 24 )
+ __field(u64, vid)
+ __array(char, cell, 24)
+ __array(char, volume, 24)
),
TP_fast_assign(
@@ -1392,10 +1392,10 @@ TRACE_EVENT(afs_cb_break,
TP_ARGS(fid, cb_break, reason, skipped),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, fid )
- __field(unsigned int, cb_break )
- __field(enum afs_cb_break_reason, reason )
- __field(bool, skipped )
+ __field_struct(struct afs_fid, fid)
+ __field(unsigned int, cb_break)
+ __field(enum afs_cb_break_reason, reason)
+ __field(bool, skipped)
),
TP_fast_assign(
@@ -1418,8 +1418,8 @@ TRACE_EVENT(afs_cb_miss,
TP_ARGS(fid, reason),
TP_STRUCT__entry(
- __field_struct(struct afs_fid, fid )
- __field(enum afs_cb_break_reason, reason )
+ __field_struct(struct afs_fid, fid)
+ __field(enum afs_cb_break_reason, reason)
),
TP_fast_assign(
@@ -1439,10 +1439,10 @@ TRACE_EVENT(afs_server,
TP_ARGS(server_debug_id, ref, active, reason),
TP_STRUCT__entry(
- __field(unsigned int, server )
- __field(int, ref )
- __field(int, active )
- __field(int, reason )
+ __field(unsigned int, server)
+ __field(int, ref)
+ __field(int, active)
+ __field(int, reason)
),
TP_fast_assign(
@@ -1465,9 +1465,9 @@ TRACE_EVENT(afs_volume,
TP_ARGS(vid, ref, reason),
TP_STRUCT__entry(
- __field(afs_volid_t, vid )
- __field(int, ref )
- __field(enum afs_volume_trace, reason )
+ __field(afs_volid_t, vid)
+ __field(int, ref)
+ __field(enum afs_volume_trace, reason)
),
TP_fast_assign(
@@ -1489,10 +1489,10 @@ TRACE_EVENT(afs_cell,
TP_ARGS(cell_debug_id, ref, active, reason),
TP_STRUCT__entry(
- __field(unsigned int, cell )
- __field(int, ref )
- __field(int, active )
- __field(int, reason )
+ __field(unsigned int, cell)
+ __field(int, ref)
+ __field(int, active)
+ __field(int, reason)
),
TP_fast_assign(
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 07/41] afs: Automatically generate trace tag enums
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (5 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 06/41] afs: Remove whitespace before most ')' from the trace header David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 08/41] afs: Add comments on abort handling David Howells
` (33 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Automatically generate trace tag enums from the symbol -> string mapping
tables rather than having the enums as well, thereby reducing duplicated
data.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
include/trace/events/afs.h | 233 +++++--------------------------------
1 file changed, 27 insertions(+), 206 deletions(-)
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index cfcd6452c156..597677acc6b1 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -18,97 +18,6 @@
#ifndef __AFS_DECLARE_TRACE_ENUMS_ONCE_ONLY
#define __AFS_DECLARE_TRACE_ENUMS_ONCE_ONLY
-enum afs_call_trace {
- afs_call_trace_alloc,
- afs_call_trace_free,
- afs_call_trace_get,
- afs_call_trace_put,
- afs_call_trace_wake,
- afs_call_trace_work,
-};
-
-enum afs_server_trace {
- afs_server_trace_alloc,
- afs_server_trace_callback,
- afs_server_trace_destroy,
- afs_server_trace_free,
- afs_server_trace_gc,
- afs_server_trace_get_by_addr,
- afs_server_trace_get_by_uuid,
- afs_server_trace_get_caps,
- afs_server_trace_get_install,
- afs_server_trace_get_new_cbi,
- afs_server_trace_get_probe,
- afs_server_trace_give_up_cb,
- afs_server_trace_purging,
- afs_server_trace_put_call,
- afs_server_trace_put_cbi,
- afs_server_trace_put_find_rsq,
- afs_server_trace_put_probe,
- afs_server_trace_put_slist,
- afs_server_trace_put_slist_isort,
- afs_server_trace_put_uuid_rsq,
- afs_server_trace_update,
-};
-
-
-enum afs_volume_trace {
- afs_volume_trace_alloc,
- afs_volume_trace_free,
- afs_volume_trace_get_alloc_sbi,
- afs_volume_trace_get_cell_insert,
- afs_volume_trace_get_new_op,
- afs_volume_trace_get_query_alias,
- afs_volume_trace_put_cell_dup,
- afs_volume_trace_put_cell_root,
- afs_volume_trace_put_destroy_sbi,
- afs_volume_trace_put_free_fc,
- afs_volume_trace_put_put_op,
- afs_volume_trace_put_query_alias,
- afs_volume_trace_put_validate_fc,
- afs_volume_trace_remove,
-};
-
-enum afs_cell_trace {
- afs_cell_trace_alloc,
- afs_cell_trace_free,
- afs_cell_trace_get_queue_dns,
- afs_cell_trace_get_queue_manage,
- afs_cell_trace_get_queue_new,
- afs_cell_trace_get_vol,
- afs_cell_trace_insert,
- afs_cell_trace_manage,
- afs_cell_trace_put_candidate,
- afs_cell_trace_put_destroy,
- afs_cell_trace_put_queue_fail,
- afs_cell_trace_put_queue_work,
- afs_cell_trace_put_vol,
- afs_cell_trace_see_source,
- afs_cell_trace_see_ws,
- afs_cell_trace_unuse_alias,
- afs_cell_trace_unuse_check_alias,
- afs_cell_trace_unuse_delete,
- afs_cell_trace_unuse_fc,
- afs_cell_trace_unuse_lookup,
- afs_cell_trace_unuse_mntpt,
- afs_cell_trace_unuse_no_pin,
- afs_cell_trace_unuse_parse,
- afs_cell_trace_unuse_pin,
- afs_cell_trace_unuse_probe,
- afs_cell_trace_unuse_sbi,
- afs_cell_trace_unuse_ws,
- afs_cell_trace_use_alias,
- afs_cell_trace_use_check_alias,
- afs_cell_trace_use_fc,
- afs_cell_trace_use_fc_alias,
- afs_cell_trace_use_lookup,
- afs_cell_trace_use_mntpt,
- afs_cell_trace_use_pin,
- afs_cell_trace_use_probe,
- afs_cell_trace_use_sbi,
- afs_cell_trace_wait,
-};
-
enum afs_fs_operation {
afs_FS_FetchData = 130, /* AFS Fetch file data */
afs_FS_FetchACL = 131, /* AFS Fetch file ACL */
@@ -202,121 +111,6 @@ enum yfs_cm_operation {
yfs_CB_CallBack = 64204,
};
-enum afs_edit_dir_op {
- afs_edit_dir_create,
- afs_edit_dir_create_error,
- afs_edit_dir_create_inval,
- afs_edit_dir_create_nospc,
- afs_edit_dir_delete,
- afs_edit_dir_delete_error,
- afs_edit_dir_delete_inval,
- afs_edit_dir_delete_noent,
-};
-
-enum afs_edit_dir_reason {
- afs_edit_dir_for_create,
- afs_edit_dir_for_link,
- afs_edit_dir_for_mkdir,
- afs_edit_dir_for_rename_0,
- afs_edit_dir_for_rename_1,
- afs_edit_dir_for_rename_2,
- afs_edit_dir_for_rmdir,
- afs_edit_dir_for_silly_0,
- afs_edit_dir_for_silly_1,
- afs_edit_dir_for_symlink,
- afs_edit_dir_for_unlink,
-};
-
-enum afs_eproto_cause {
- afs_eproto_bad_status,
- afs_eproto_cb_count,
- afs_eproto_cb_fid_count,
- afs_eproto_cellname_len,
- afs_eproto_file_type,
- afs_eproto_ibulkst_cb_count,
- afs_eproto_ibulkst_count,
- afs_eproto_motd_len,
- afs_eproto_offline_msg_len,
- afs_eproto_volname_len,
- afs_eproto_yvl_fsendpt4_len,
- afs_eproto_yvl_fsendpt6_len,
- afs_eproto_yvl_fsendpt_num,
- afs_eproto_yvl_fsendpt_type,
- afs_eproto_yvl_vlendpt4_len,
- afs_eproto_yvl_vlendpt6_len,
- afs_eproto_yvl_vlendpt_type,
-};
-
-enum afs_io_error {
- afs_io_error_cm_reply,
- afs_io_error_extract,
- afs_io_error_fs_probe_fail,
- afs_io_error_vl_lookup_fail,
- afs_io_error_vl_probe_fail,
-};
-
-enum afs_file_error {
- afs_file_error_dir_bad_magic,
- afs_file_error_dir_big,
- afs_file_error_dir_missing_page,
- afs_file_error_dir_name_too_long,
- afs_file_error_dir_over_end,
- afs_file_error_dir_small,
- afs_file_error_dir_unmarked_ext,
- afs_file_error_mntpt,
- afs_file_error_writeback_fail,
-};
-
-enum afs_flock_event {
- afs_flock_acquired,
- afs_flock_callback_break,
- afs_flock_defer_unlock,
- afs_flock_extend_fail,
- afs_flock_fail_other,
- afs_flock_fail_perm,
- afs_flock_no_lockers,
- afs_flock_release_fail,
- afs_flock_silly_delete,
- afs_flock_timestamp,
- afs_flock_try_to_lock,
- afs_flock_vfs_lock,
- afs_flock_vfs_locking,
- afs_flock_waited,
- afs_flock_waiting,
- afs_flock_work_extending,
- afs_flock_work_retry,
- afs_flock_work_unlocking,
- afs_flock_would_block,
-};
-
-enum afs_flock_operation {
- afs_flock_op_copy_lock,
- afs_flock_op_flock,
- afs_flock_op_grant,
- afs_flock_op_lock,
- afs_flock_op_release_lock,
- afs_flock_op_return_ok,
- afs_flock_op_return_eagain,
- afs_flock_op_return_edeadlk,
- afs_flock_op_return_error,
- afs_flock_op_set_lock,
- afs_flock_op_unlock,
- afs_flock_op_wake,
-};
-
-enum afs_cb_break_reason {
- afs_cb_break_no_break,
- afs_cb_break_no_promise,
- afs_cb_break_for_callback,
- afs_cb_break_for_deleted,
- afs_cb_break_for_lapsed,
- afs_cb_break_for_s_reinit,
- afs_cb_break_for_unlink,
- afs_cb_break_for_v_break,
- afs_cb_break_for_volume_callback,
- afs_cb_break_for_zap,
-};
-
#endif /* end __AFS_DECLARE_TRACE_ENUMS_ONCE_ONLY */
/*
@@ -391,6 +185,7 @@ enum afs_cb_break_reason {
EM(afs_cell_trace_unuse_fc, "UNU fc ") \
EM(afs_cell_trace_unuse_lookup, "UNU lookup") \
EM(afs_cell_trace_unuse_mntpt, "UNU mntpt ") \
+ EM(afs_cell_trace_unuse_no_pin, "UNU no-pin") \
EM(afs_cell_trace_unuse_parse, "UNU parse ") \
EM(afs_cell_trace_unuse_pin, "UNU pin ") \
EM(afs_cell_trace_unuse_probe, "UNU probe ") \
@@ -614,6 +409,32 @@ enum afs_cb_break_reason {
EM(afs_cb_break_for_volume_callback, "break-v-cb") \
E_(afs_cb_break_for_zap, "break-zap")
+/*
+ * Generate enums for tracing information.
+ */
+#ifndef __AFS_GENERATE_TRACE_ENUMS_ONCE_ONLY
+#define __AFS_GENERATE_TRACE_ENUMS_ONCE_ONLY
+
+#undef EM
+#undef E_
+#define EM(a, b) a,
+#define E_(a, b) a
+
+enum afs_call_trace { afs_call_traces } __mode(byte);
+enum afs_cb_break_reason { afs_cb_break_reasons } __mode(byte);
+enum afs_cell_trace { afs_cell_traces } __mode(byte);
+enum afs_edit_dir_op { afs_edit_dir_ops } __mode(byte);
+enum afs_edit_dir_reason { afs_edit_dir_reasons } __mode(byte);
+enum afs_eproto_cause { afs_eproto_causes } __mode(byte);
+enum afs_file_error { afs_file_errors } __mode(byte);
+enum afs_flock_event { afs_flock_events } __mode(byte);
+enum afs_flock_operation { afs_flock_operations } __mode(byte);
+enum afs_io_error { afs_io_errors } __mode(byte);
+enum afs_server_trace { afs_server_traces } __mode(byte);
+enum afs_volume_trace { afs_volume_traces } __mode(byte);
+
+#endif /* end __AFS_GENERATE_TRACE_ENUMS_ONCE_ONLY */
+
/*
* Export enum symbols via userspace.
*/
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 08/41] afs: Add comments on abort handling
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (6 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 07/41] afs: Automatically generate trace tag enums David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 17:41 ` Jeffrey E Altman
2023-11-09 15:39 ` [PATCH 09/41] afs: Turn the afs_addr_list address array into an array of structs David Howells
` (32 subsequent siblings)
40 siblings, 1 reply; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne
Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel,
Jeffrey E Altman
Add some comments on AFS abort code handling in the rotation algorithm and
adjust the errors produced to match.
Reported-by: Jeffrey E Altman <jaltman@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/rotate.c | 100 ++++++++++++++++++++++++++++++++++++++++++------
1 file changed, 89 insertions(+), 11 deletions(-)
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index a840c3588ebb..180bcad081dd 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -13,6 +13,7 @@
#include <linux/sched/signal.h>
#include "internal.h"
#include "afs_fs.h"
+#include "protocol_uae.h"
/*
* Begin iteration through a server list, starting with the vnode's last used
@@ -143,6 +144,11 @@ bool afs_select_fileserver(struct afs_operation *op)
case -ECONNABORTED:
/* The far side rejected the operation on some grounds. This
* might involve the server being busy or the volume having been moved.
+ *
+ * Note that various V* errors should not be sent to a cache manager
+ * by a fileserver as they should be translated to more modern UAE*
+ * errors instead. IBM AFS and OpenAFS fileservers, however, do leak
+ * these abort codes.
*/
switch (op->ac.abort_code) {
case VNOVOL:
@@ -150,6 +156,11 @@ bool afs_select_fileserver(struct afs_operation *op)
* - May indicate that the VL is wrong - retry once and compare
* the results.
* - May indicate that the fileserver couldn't attach to the vol.
+ * - The volume might have been temporarily removed so that it can
+ * be replaced by a volume restore. "vos" might have ended one
+ * transaction and has yet to create the next.
+ * - The volume might not be blessed or might not be in-service
+ * (administrative action).
*/
if (op->flags & AFS_OPERATION_VNOVOL) {
op->error = -EREMOTEIO;
@@ -183,16 +194,56 @@ bool afs_select_fileserver(struct afs_operation *op)
_leave(" = t [vnovol]");
return true;
- case VSALVAGE: /* TODO: Should this return an error or iterate? */
case VVOLEXISTS:
- case VNOSERVICE:
case VONLINE:
- case VDISKFULL:
- case VOVERQUOTA:
- op->error = afs_abort_to_error(op->ac.abort_code);
+ /* These should not be returned from the fileserver. */
+ pr_warn("Fileserver returned unexpected abort %d\n",
+ op->ac.abort_code);
+ op->error = -EREMOTEIO;
+ goto next_server;
+
+ case VNOSERVICE:
+ /* Prior to AFS 3.2 VNOSERVICE was returned from the fileserver
+ * if the volume was neither in-service nor administratively
+ * blessed. All usage was replaced by VNOVOL because AFS 3.1 and
+ * earlier cache managers did not handle VNOSERVICE and assumed
+ * it was the client OSes errno 105.
+ *
+ * Starting with OpenAFS 1.4.8 VNOSERVICE was repurposed as the
+ * fileserver idle dead time error which was sent in place of
+ * RX_CALL_TIMEOUT (-3). The error was intended to be sent if the
+ * fileserver took too long to send a reply to the client.
+ * RX_CALL_TIMEOUT would have caused the cache manager to mark the
+ * server down whereas VNOSERVICE since AFS 3.2 would cause cache
+ * manager to temporarily (up to 15 minutes) mark the volume
+ * instance as unusable.
+ *
+ * The idle dead logic resulted in cache inconsistency since a
+ * state changing call that the cache manager assumed was dead
+ * could still be processed to completion by the fileserver. This
+ * logic was removed in OpenAFS 1.8.0 and VNOSERVICE is no longer
+ * returned. However, many 1.4.8 through 1.6.24 fileservers are
+ * still in existence.
+ *
+ * AuriStorFS fileservers have never returned VNOSERVICE.
+ *
+ * VNOSERVICE should be treated as an alias for RX_CALL_TIMEOUT..
+ */
+ case RX_CALL_TIMEOUT:
+ op->error = -ETIMEDOUT;
goto next_server;
+ case VSALVAGING: /* This error should not be leaked to cache managers
+ * but is from OpenAFS demand attach fileservers.
+ * It should be treated as an alias for VOFFLINE.
+ */
+ case VSALVAGE: /* VSALVAGE should be treated as a synonym of VOFFLINE */
case VOFFLINE:
+ /* The volume is in use by the volserver or another volume utility
+ * for an operation that might alter the contents. The volume is
+ * expected to come back but it might take a long time (could be
+ * days).
+ */
if (!test_and_set_bit(AFS_VOLUME_OFFLINE, &op->volume->flags)) {
afs_busy(op->volume, op->ac.abort_code);
clear_bit(AFS_VOLUME_BUSY, &op->volume->flags);
@@ -207,11 +258,19 @@ bool afs_select_fileserver(struct afs_operation *op)
}
goto busy;
- case VSALVAGING:
- case VRESTARTING:
+ case VRESTARTING: /* The fileserver is either shutting down or starting up. */
case VBUSY:
- /* Retry after going round all the servers unless we
- * have a file lock we need to maintain.
+ /* The volume is in use by the volserver or another volume utility
+ * for an operation that is not expected to alter the contents of
+ * the volume. VBUSY should not be returned for a ROVOL or
+ * BACKVOL (but many OpenAFS fileserver versions are broken). The
+ * fileserver is supposed to continue serving content from ROVOLs
+ * and BACKVOLs during an ITBusy transaction because the content
+ * cannot change. The volume is expected to come back but it
+ * might take awhile.
+ *
+ * Retry after going round all the servers unless we have a file
+ * lock we need to maintain.
*/
if (op->flags & AFS_OPERATION_NO_VSLEEP) {
op->error = -EBUSY;
@@ -226,7 +285,7 @@ bool afs_select_fileserver(struct afs_operation *op)
if (!afs_sleep_and_retry(op))
goto failed;
- /* Retry with same server & address */
+ /* Retry with same server & address */
_leave(" = t [vbusy]");
return true;
}
@@ -270,10 +329,29 @@ bool afs_select_fileserver(struct afs_operation *op)
goto restart_from_beginning;
+ case VDISKFULL:
+ case UAENOSPC:
+ /* The partition is full. Only applies to RWVOLs.
+ * Translate locally and return ENOSPC.
+ * No replicas to failover to.
+ */
+ op->error = -ENOSPC;
+ goto failed_but_online;
+
+ case VOVERQUOTA:
+ case UAEDQUOT:
+ /* Volume is full. Only applies to RWVOLs.
+ * Translate locally and return EDQUOT.
+ * No replicas to failover to.
+ */
+ op->error = -EDQUOT;
+ goto failed_but_online;
+
default:
+ op->error = afs_abort_to_error(op->ac.abort_code);
+ failed_but_online:
clear_bit(AFS_VOLUME_OFFLINE, &op->volume->flags);
clear_bit(AFS_VOLUME_BUSY, &op->volume->flags);
- op->error = afs_abort_to_error(op->ac.abort_code);
goto failed;
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 09/41] afs: Turn the afs_addr_list address array into an array of structs
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (7 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 08/41] afs: Add comments on abort handling David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 10/41] rxrpc, afs: Allow afs to pin rxrpc_peer objects David Howells
` (31 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Turn the afs_addr_list address array into an array of structs, thereby
allowing per-address (such as RTT) info to be added.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 10 +++++-----
fs/afs/fs_probe.c | 6 +++---
fs/afs/internal.h | 6 +++++-
fs/afs/proc.c | 4 ++--
fs/afs/rotate.c | 2 +-
fs/afs/rxrpc.c | 4 ++--
fs/afs/server.c | 4 ++--
fs/afs/vl_alias.c | 4 ++--
fs/afs/vl_probe.c | 6 +++---
fs/afs/vl_rotate.c | 2 +-
10 files changed, 26 insertions(+), 22 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index de1ae0bead3b..ac05a59e9d46 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -45,7 +45,7 @@ struct afs_addr_list *afs_alloc_addrlist(unsigned int nr,
alist->max_addrs = nr;
for (i = 0; i < nr; i++) {
- struct sockaddr_rxrpc *srx = &alist->addrs[i];
+ struct sockaddr_rxrpc *srx = &alist->addrs[i].srx;
srx->srx_family = AF_RXRPC;
srx->srx_service = service;
srx->transport_type = SOCK_DGRAM;
@@ -281,7 +281,7 @@ void afs_merge_fs_addr4(struct afs_addr_list *alist, __be32 xdr, u16 port)
return;
for (i = 0; i < alist->nr_ipv4; i++) {
- struct sockaddr_in *a = &alist->addrs[i].transport.sin;
+ struct sockaddr_in *a = &alist->addrs[i].srx.transport.sin;
u32 a_addr = ntohl(a->sin_addr.s_addr);
u16 a_port = ntohs(a->sin_port);
@@ -298,7 +298,7 @@ void afs_merge_fs_addr4(struct afs_addr_list *alist, __be32 xdr, u16 port)
alist->addrs + i,
sizeof(alist->addrs[0]) * (alist->nr_addrs - i));
- srx = &alist->addrs[i];
+ srx = &alist->addrs[i].srx;
srx->srx_family = AF_RXRPC;
srx->transport_type = SOCK_DGRAM;
srx->transport_len = sizeof(srx->transport.sin);
@@ -321,7 +321,7 @@ void afs_merge_fs_addr6(struct afs_addr_list *alist, __be32 *xdr, u16 port)
return;
for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
- struct sockaddr_in6 *a = &alist->addrs[i].transport.sin6;
+ struct sockaddr_in6 *a = &alist->addrs[i].srx.transport.sin6;
u16 a_port = ntohs(a->sin6_port);
diff = memcmp(xdr, &a->sin6_addr, 16);
@@ -338,7 +338,7 @@ void afs_merge_fs_addr6(struct afs_addr_list *alist, __be32 *xdr, u16 port)
alist->addrs + i,
sizeof(alist->addrs[0]) * (alist->nr_addrs - i));
- srx = &alist->addrs[i];
+ srx = &alist->addrs[i].srx;
srx->srx_family = AF_RXRPC;
srx->transport_type = SOCK_DGRAM;
srx->transport_len = sizeof(srx->transport.sin6);
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index daaf3810cc92..3dd24842f277 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -153,12 +153,12 @@ void afs_fileserver_probe_result(struct afs_call *call)
if (call->service_id == YFS_FS_SERVICE) {
server->probe.is_yfs = true;
set_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx_service = call->service_id;
+ alist->addrs[index].srx.srx_service = call->service_id;
} else {
server->probe.not_yfs = true;
if (!server->probe.is_yfs) {
clear_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx_service = call->service_id;
+ alist->addrs[index].srx.srx_service = call->service_id;
}
cap0 = ntohl(call->tmp);
if (cap0 & AFS3_VICED_CAPABILITY_64BITFILES)
@@ -182,7 +182,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
spin_unlock(&server->probe_lock);
_debug("probe %pU [%u] %pISpc rtt=%u ret=%d",
- &server->uuid, index, &alist->addrs[index].transport,
+ &server->uuid, index, &alist->addrs[index].srx.transport,
rtt_us, ret);
return afs_done_one_fs_probe(call->net, server);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 5041eae64423..ae874baee249 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -87,7 +87,9 @@ struct afs_addr_list {
enum dns_lookup_status status:8;
unsigned long failed; /* Mask of addrs that failed locally/ICMP */
unsigned long responded; /* Mask of addrs that responded */
- struct sockaddr_rxrpc addrs[];
+ struct {
+ struct sockaddr_rxrpc srx;
+ } addrs[] __counted_by(max_addrs);
#define AFS_MAX_ADDRESSES ((unsigned int)(sizeof(unsigned long) * 8))
};
@@ -968,6 +970,8 @@ extern void afs_put_addrlist(struct afs_addr_list *);
extern struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *,
const char *, size_t, char,
unsigned short, unsigned short);
+bool afs_addr_list_same(const struct afs_addr_list *a,
+ const struct afs_addr_list *b);
extern struct afs_vlserver_list *afs_dns_query(struct afs_cell *, time64_t *);
extern bool afs_iterate_addresses(struct afs_addr_cursor *);
extern int afs_end_cursor(struct afs_addr_cursor *);
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index 2a0c83d71565..ab9cd986cfd9 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -307,7 +307,7 @@ static int afs_proc_cell_vlservers_show(struct seq_file *m, void *v)
for (i = 0; i < alist->nr_addrs; i++)
seq_printf(m, " %c %pISpc\n",
alist->preferred == i ? '>' : '-',
- &alist->addrs[i].transport);
+ &alist->addrs[i].srx.transport);
}
seq_printf(m, " info: fl=%lx rtt=%d\n", vlserver->flags, vlserver->rtt);
seq_printf(m, " probe: fl=%x e=%d ac=%d out=%d\n",
@@ -399,7 +399,7 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
alist->version, alist->responded, alist->failed);
for (i = 0; i < alist->nr_addrs; i++)
seq_printf(m, " [%x] %pISpc%s\n",
- i, &alist->addrs[i].transport,
+ i, &alist->addrs[i].srx.transport,
alist->preferred == i ? "*" : "");
return 0;
}
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 180bcad081dd..993e20d752d9 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -487,7 +487,7 @@ bool afs_select_fileserver(struct afs_operation *op)
_debug("address [%u] %u/%u %pISp",
op->index, op->ac.index, op->ac.alist->nr_addrs,
- &op->ac.alist->addrs[op->ac.index].transport);
+ &op->ac.alist->addrs[op->ac.index].srx.transport);
_leave(" = t");
return true;
diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
index ed1644e7683f..2b1a31b4249c 100644
--- a/fs/afs/rxrpc.c
+++ b/fs/afs/rxrpc.c
@@ -296,7 +296,7 @@ static void afs_notify_end_request_tx(struct sock *sock,
*/
void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
{
- struct sockaddr_rxrpc *srx = &ac->alist->addrs[ac->index];
+ struct sockaddr_rxrpc *srx = &ac->alist->addrs[ac->index].srx;
struct rxrpc_call *rxcall;
struct msghdr msg;
struct kvec iov[1];
@@ -461,7 +461,7 @@ static void afs_log_error(struct afs_call *call, s32 remote_abort)
max = m + 1;
pr_notice("kAFS: Peer reported %s failure on %s [%pISp]\n",
msg, call->type->name,
- &call->alist->addrs[call->addr_ix].transport);
+ &call->alist->addrs[call->addr_ix].srx.transport);
}
}
diff --git a/fs/afs/server.c b/fs/afs/server.c
index b5237206eac3..4b98788c7b12 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -42,7 +42,7 @@ struct afs_server *afs_find_server(struct afs_net *net,
hlist_for_each_entry_rcu(server, &net->fs_addresses6, addr6_link) {
alist = rcu_dereference(server->addresses);
for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
- b = &alist->addrs[i].transport.sin6;
+ b = &alist->addrs[i].srx.transport.sin6;
diff = ((u16 __force)a->sin6_port -
(u16 __force)b->sin6_port);
if (diff == 0)
@@ -58,7 +58,7 @@ struct afs_server *afs_find_server(struct afs_net *net,
hlist_for_each_entry_rcu(server, &net->fs_addresses4, addr4_link) {
alist = rcu_dereference(server->addresses);
for (i = 0; i < alist->nr_ipv4; i++) {
- b = &alist->addrs[i].transport.sin;
+ b = &alist->addrs[i].srx.transport.sin;
diff = ((u16 __force)a->sin_port -
(u16 __force)b->sin_port);
if (diff == 0)
diff --git a/fs/afs/vl_alias.c b/fs/afs/vl_alias.c
index f04a80e4f5c3..d3c0df70a1a5 100644
--- a/fs/afs/vl_alias.c
+++ b/fs/afs/vl_alias.c
@@ -94,8 +94,8 @@ static int afs_compare_fs_alists(const struct afs_server *server_a,
lb = rcu_dereference(server_b->addresses);
while (a < la->nr_addrs && b < lb->nr_addrs) {
- const struct sockaddr_rxrpc *srx_a = &la->addrs[a];
- const struct sockaddr_rxrpc *srx_b = &lb->addrs[b];
+ const struct sockaddr_rxrpc *srx_a = &la->addrs[a].srx;
+ const struct sockaddr_rxrpc *srx_b = &lb->addrs[b].srx;
int diff = afs_compare_addrs(srx_a, srx_b);
if (diff < 0) {
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index 58452b86e672..bdd9372e3fb2 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -106,12 +106,12 @@ void afs_vlserver_probe_result(struct afs_call *call)
if (call->service_id == YFS_VL_SERVICE) {
server->probe.flags |= AFS_VLSERVER_PROBE_IS_YFS;
set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx_service = call->service_id;
+ alist->addrs[index].srx.srx_service = call->service_id;
} else {
server->probe.flags |= AFS_VLSERVER_PROBE_NOT_YFS;
if (!(server->probe.flags & AFS_VLSERVER_PROBE_IS_YFS)) {
clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx_service = call->service_id;
+ alist->addrs[index].srx.srx_service = call->service_id;
}
}
@@ -131,7 +131,7 @@ void afs_vlserver_probe_result(struct afs_call *call)
spin_unlock(&server->probe_lock);
_debug("probe [%u][%u] %pISpc rtt=%u ret=%d",
- server_index, index, &alist->addrs[index].transport, rtt_us, ret);
+ server_index, index, &alist->addrs[index].srx.transport, rtt_us, ret);
afs_done_one_vl_probe(server, have_result);
}
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index 488e58490b16..0d0b54819128 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -243,7 +243,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
_debug("VL address %d/%d", vc->ac.index, vc->ac.alist->nr_addrs);
- _leave(" = t %pISpc", &vc->ac.alist->addrs[vc->ac.index].transport);
+ _leave(" = t %pISpc", &vc->ac.alist->addrs[vc->ac.index].srx.transport);
return true;
next_server:
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 10/41] rxrpc, afs: Allow afs to pin rxrpc_peer objects
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (8 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 09/41] afs: Turn the afs_addr_list address array into an array of structs David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 17:48 ` Marc Dionne
2023-11-09 15:39 ` [PATCH 11/41] afs: Don't skip server addresses for which we didn't get an RTT reading David Howells
` (30 subsequent siblings)
40 siblings, 1 reply; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Change rxrpc's API such that:
(1) A new function, rxrpc_kernel_lookup_peer(), is provided to look up an
rxrpc_peer record for a remote address and a corresponding function,
rxrpc_kernel_put_peer(), is provided to dispose of it again.
(2) When setting up a call, the rxrpc_peer object used during a call is
now passed in rather than being set up by rxrpc_connect_call(). For
afs, this meenat passing it to rxrpc_kernel_begin_call() rather than
the full address (the service ID then has to be passed in as a
separate parameter).
(3) A new function, rxrpc_kernel_remote_addr(), is added so that afs can
get a pointer to the transport address for display purposed, and
another, rxrpc_kernel_remote_srx(), to gain a pointer to the full
rxrpc address.
(4) The function to retrieve the RTT from a call, rxrpc_kernel_get_srtt(),
is then altered to take a peer. This now returns the RTT or -1 if
there are insufficient samples.
(5) Rename rxrpc_kernel_get_peer() to rxrpc_kernel_call_get_peer().
(6) Provide a new function, rxrpc_kernel_get_peer(), to get a ref on a
peer the caller already has.
This allows the afs filesystem to pin the rxrpc_peer records that it is
using, allowing faster lookups and pointer comparisons rather than
comparing sockaddr_rxrpc contents. It also makes it easier to get hold of
the RTT. The following changes are made to afs:
(1) The addr_list struct's addrs[] elements now hold a peer struct pointer
and a service ID rather than a sockaddr_rxrpc.
(2) When displaying the transport address, rxrpc_kernel_remote_addr() is
used.
(3) The port arg is removed from afs_alloc_addrlist() since it's always
overridden.
(4) afs_merge_fs_addr4() and afs_merge_fs_addr6() do peer lookup and may
now return an error that must be handled.
(5) afs_find_server() now takes a peer pointer to specify the address.
(6) afs_find_server(), afs_compare_fs_alists() and afs_merge_fs_addr[46]{}
now do peer pointer comparison rather than address comparison.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 125 ++++++++++++++++++-----------------
fs/afs/cmservice.c | 5 +-
fs/afs/fs_probe.c | 11 +--
fs/afs/internal.h | 26 ++++----
fs/afs/proc.c | 9 +--
fs/afs/rotate.c | 6 +-
fs/afs/rxrpc.c | 10 +--
fs/afs/server.c | 41 ++----------
fs/afs/vl_alias.c | 55 +--------------
fs/afs/vl_list.c | 15 +++--
fs/afs/vl_probe.c | 12 ++--
fs/afs/vl_rotate.c | 6 +-
fs/afs/vlclient.c | 22 ++++--
include/net/af_rxrpc.h | 15 +++--
include/trace/events/rxrpc.h | 3 +
net/rxrpc/af_rxrpc.c | 62 ++++++++++++++---
net/rxrpc/ar-internal.h | 2 +-
net/rxrpc/call_object.c | 17 ++---
net/rxrpc/peer_object.c | 56 ++++++++++------
net/rxrpc/sendmsg.c | 11 ++-
20 files changed, 271 insertions(+), 238 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index ac05a59e9d46..519821f5aedc 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -13,26 +13,33 @@
#include "internal.h"
#include "afs_fs.h"
+static void afs_free_addrlist(struct rcu_head *rcu)
+{
+ struct afs_addr_list *alist = container_of(rcu, struct afs_addr_list, rcu);
+ unsigned int i;
+
+ for (i = 0; i < alist->nr_addrs; i++)
+ rxrpc_kernel_put_peer(alist->addrs[i].peer);
+}
+
/*
* Release an address list.
*/
void afs_put_addrlist(struct afs_addr_list *alist)
{
if (alist && refcount_dec_and_test(&alist->usage))
- kfree_rcu(alist, rcu);
+ call_rcu(&alist->rcu, afs_free_addrlist);
}
/*
* Allocate an address list.
*/
-struct afs_addr_list *afs_alloc_addrlist(unsigned int nr,
- unsigned short service,
- unsigned short port)
+struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id)
{
struct afs_addr_list *alist;
unsigned int i;
- _enter("%u,%u,%u", nr, service, port);
+ _enter("%u,%u", nr, service_id);
if (nr > AFS_MAX_ADDRESSES)
nr = AFS_MAX_ADDRESSES;
@@ -44,16 +51,8 @@ struct afs_addr_list *afs_alloc_addrlist(unsigned int nr,
refcount_set(&alist->usage, 1);
alist->max_addrs = nr;
- for (i = 0; i < nr; i++) {
- struct sockaddr_rxrpc *srx = &alist->addrs[i].srx;
- srx->srx_family = AF_RXRPC;
- srx->srx_service = service;
- srx->transport_type = SOCK_DGRAM;
- srx->transport_len = sizeof(srx->transport.sin6);
- srx->transport.sin6.sin6_family = AF_INET6;
- srx->transport.sin6.sin6_port = htons(port);
- }
-
+ for (i = 0; i < nr; i++)
+ alist->addrs[i].service_id = service_id;
return alist;
}
@@ -126,7 +125,7 @@ struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *net,
if (!vllist->servers[0].server)
goto error_vl;
- alist = afs_alloc_addrlist(nr, service, AFS_VL_PORT);
+ alist = afs_alloc_addrlist(nr, service);
if (!alist)
goto error;
@@ -197,9 +196,11 @@ struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *net,
}
if (family == AF_INET)
- afs_merge_fs_addr4(alist, x[0], xport);
+ ret = afs_merge_fs_addr4(net, alist, x[0], xport);
else
- afs_merge_fs_addr6(alist, x, xport);
+ ret = afs_merge_fs_addr6(net, alist, x, xport);
+ if (ret < 0)
+ goto error;
} while (p < end);
@@ -271,25 +272,33 @@ struct afs_vlserver_list *afs_dns_query(struct afs_cell *cell, time64_t *_expiry
/*
* Merge an IPv4 entry into a fileserver address list.
*/
-void afs_merge_fs_addr4(struct afs_addr_list *alist, __be32 xdr, u16 port)
+int afs_merge_fs_addr4(struct afs_net *net, struct afs_addr_list *alist,
+ __be32 xdr, u16 port)
{
- struct sockaddr_rxrpc *srx;
- u32 addr = ntohl(xdr);
+ struct sockaddr_rxrpc srx;
+ struct rxrpc_peer *peer;
int i;
if (alist->nr_addrs >= alist->max_addrs)
- return;
+ return 0;
- for (i = 0; i < alist->nr_ipv4; i++) {
- struct sockaddr_in *a = &alist->addrs[i].srx.transport.sin;
- u32 a_addr = ntohl(a->sin_addr.s_addr);
- u16 a_port = ntohs(a->sin_port);
+ srx.srx_family = AF_RXRPC;
+ srx.transport_type = SOCK_DGRAM;
+ srx.transport_len = sizeof(srx.transport.sin);
+ srx.transport.sin.sin_family = AF_INET;
+ srx.transport.sin.sin_port = htons(port);
+ srx.transport.sin.sin_addr.s_addr = xdr;
- if (addr == a_addr && port == a_port)
- return;
- if (addr == a_addr && port < a_port)
- break;
- if (addr < a_addr)
+ peer = rxrpc_kernel_lookup_peer(net->socket, &srx, GFP_KERNEL);
+ if (!peer)
+ return -ENOMEM;
+
+ for (i = 0; i < alist->nr_ipv4; i++) {
+ if (peer == alist->addrs[i].peer) {
+ rxrpc_kernel_put_peer(peer);
+ return 0;
+ }
+ if (peer <= alist->addrs[i].peer)
break;
}
@@ -298,38 +307,42 @@ void afs_merge_fs_addr4(struct afs_addr_list *alist, __be32 xdr, u16 port)
alist->addrs + i,
sizeof(alist->addrs[0]) * (alist->nr_addrs - i));
- srx = &alist->addrs[i].srx;
- srx->srx_family = AF_RXRPC;
- srx->transport_type = SOCK_DGRAM;
- srx->transport_len = sizeof(srx->transport.sin);
- srx->transport.sin.sin_family = AF_INET;
- srx->transport.sin.sin_port = htons(port);
- srx->transport.sin.sin_addr.s_addr = xdr;
+ alist->addrs[i].peer = peer;
alist->nr_ipv4++;
alist->nr_addrs++;
+ return 0;
}
/*
* Merge an IPv6 entry into a fileserver address list.
*/
-void afs_merge_fs_addr6(struct afs_addr_list *alist, __be32 *xdr, u16 port)
+int afs_merge_fs_addr6(struct afs_net *net, struct afs_addr_list *alist,
+ __be32 *xdr, u16 port)
{
- struct sockaddr_rxrpc *srx;
- int i, diff;
+ struct sockaddr_rxrpc srx;
+ struct rxrpc_peer *peer;
+ int i;
if (alist->nr_addrs >= alist->max_addrs)
- return;
+ return 0;
- for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
- struct sockaddr_in6 *a = &alist->addrs[i].srx.transport.sin6;
- u16 a_port = ntohs(a->sin6_port);
+ srx.srx_family = AF_RXRPC;
+ srx.transport_type = SOCK_DGRAM;
+ srx.transport_len = sizeof(srx.transport.sin6);
+ srx.transport.sin6.sin6_family = AF_INET6;
+ srx.transport.sin6.sin6_port = htons(port);
+ memcpy(&srx.transport.sin6.sin6_addr, xdr, 16);
- diff = memcmp(xdr, &a->sin6_addr, 16);
- if (diff == 0 && port == a_port)
- return;
- if (diff == 0 && port < a_port)
- break;
- if (diff < 0)
+ peer = rxrpc_kernel_lookup_peer(net->socket, &srx, GFP_KERNEL);
+ if (!peer)
+ return -ENOMEM;
+
+ for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
+ if (peer == alist->addrs[i].peer) {
+ rxrpc_kernel_put_peer(peer);
+ return 0;
+ }
+ if (peer <= alist->addrs[i].peer)
break;
}
@@ -337,15 +350,9 @@ void afs_merge_fs_addr6(struct afs_addr_list *alist, __be32 *xdr, u16 port)
memmove(alist->addrs + i + 1,
alist->addrs + i,
sizeof(alist->addrs[0]) * (alist->nr_addrs - i));
-
- srx = &alist->addrs[i].srx;
- srx->srx_family = AF_RXRPC;
- srx->transport_type = SOCK_DGRAM;
- srx->transport_len = sizeof(srx->transport.sin6);
- srx->transport.sin6.sin6_family = AF_INET6;
- srx->transport.sin6.sin6_port = htons(port);
- memcpy(&srx->transport.sin6.sin6_addr, xdr, 16);
+ alist->addrs[i].peer = peer;
alist->nr_addrs++;
+ return 0;
}
/*
diff --git a/fs/afs/cmservice.c b/fs/afs/cmservice.c
index d4ddb20d6732..99a3f20bc786 100644
--- a/fs/afs/cmservice.c
+++ b/fs/afs/cmservice.c
@@ -146,10 +146,11 @@ static int afs_find_cm_server_by_peer(struct afs_call *call)
{
struct sockaddr_rxrpc srx;
struct afs_server *server;
+ struct rxrpc_peer *peer;
- rxrpc_kernel_get_peer(call->net->socket, call->rxcall, &srx);
+ peer = rxrpc_kernel_get_call_peer(call->net->socket, call->rxcall);
- server = afs_find_server(call->net, &srx);
+ server = afs_find_server(call->net, peer);
if (!server) {
trace_afs_cm_no_server(call, &srx);
return 0;
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index 3dd24842f277..58d28b82571e 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -101,6 +101,7 @@ static void afs_fs_probe_not_done(struct afs_net *net,
void afs_fileserver_probe_result(struct afs_call *call)
{
struct afs_addr_list *alist = call->alist;
+ struct afs_address *addr = &alist->addrs[call->addr_ix];
struct afs_server *server = call->server;
unsigned int index = call->addr_ix;
unsigned int rtt_us = 0, cap0;
@@ -153,12 +154,12 @@ void afs_fileserver_probe_result(struct afs_call *call)
if (call->service_id == YFS_FS_SERVICE) {
server->probe.is_yfs = true;
set_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx.srx_service = call->service_id;
+ addr->service_id = call->service_id;
} else {
server->probe.not_yfs = true;
if (!server->probe.is_yfs) {
clear_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx.srx_service = call->service_id;
+ addr->service_id = call->service_id;
}
cap0 = ntohl(call->tmp);
if (cap0 & AFS3_VICED_CAPABILITY_64BITFILES)
@@ -167,7 +168,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
clear_bit(AFS_SERVER_FL_HAS_FS64, &server->flags);
}
- rxrpc_kernel_get_srtt(call->net->socket, call->rxcall, &rtt_us);
+ rtt_us = rxrpc_kernel_get_srtt(addr->peer);
if (rtt_us < server->probe.rtt) {
server->probe.rtt = rtt_us;
server->rtt = rtt_us;
@@ -181,8 +182,8 @@ void afs_fileserver_probe_result(struct afs_call *call)
out:
spin_unlock(&server->probe_lock);
- _debug("probe %pU [%u] %pISpc rtt=%u ret=%d",
- &server->uuid, index, &alist->addrs[index].srx.transport,
+ _debug("probe %pU [%u] %pISpc rtt=%d ret=%d",
+ &server->uuid, index, rxrpc_kernel_remote_addr(alist->addrs[index].peer),
rtt_us, ret);
return afs_done_one_fs_probe(call->net, server);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index ae874baee249..caf89edc0644 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -72,6 +72,11 @@ enum afs_call_state {
AFS_CALL_COMPLETE, /* Completed or failed */
};
+struct afs_address {
+ struct rxrpc_peer *peer;
+ u16 service_id;
+};
+
/*
* List of server addresses.
*/
@@ -87,9 +92,7 @@ struct afs_addr_list {
enum dns_lookup_status status:8;
unsigned long failed; /* Mask of addrs that failed locally/ICMP */
unsigned long responded; /* Mask of addrs that responded */
- struct {
- struct sockaddr_rxrpc srx;
- } addrs[] __counted_by(max_addrs);
+ struct afs_address addrs[] __counted_by(max_addrs);
#define AFS_MAX_ADDRESSES ((unsigned int)(sizeof(unsigned long) * 8))
};
@@ -420,7 +423,7 @@ struct afs_vlserver {
atomic_t probe_outstanding;
spinlock_t probe_lock;
struct {
- unsigned int rtt; /* RTT in uS */
+ unsigned int rtt; /* Best RTT in uS (or UINT_MAX) */
u32 abort_code;
short error;
unsigned short flags;
@@ -537,7 +540,7 @@ struct afs_server {
atomic_t probe_outstanding;
spinlock_t probe_lock;
struct {
- unsigned int rtt; /* RTT in uS */
+ unsigned int rtt; /* Best RTT in uS (or UINT_MAX) */
u32 abort_code;
short error;
bool responded:1;
@@ -963,9 +966,7 @@ static inline struct afs_addr_list *afs_get_addrlist(struct afs_addr_list *alist
refcount_inc(&alist->usage);
return alist;
}
-extern struct afs_addr_list *afs_alloc_addrlist(unsigned int,
- unsigned short,
- unsigned short);
+extern struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id);
extern void afs_put_addrlist(struct afs_addr_list *);
extern struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *,
const char *, size_t, char,
@@ -976,8 +977,10 @@ extern struct afs_vlserver_list *afs_dns_query(struct afs_cell *, time64_t *);
extern bool afs_iterate_addresses(struct afs_addr_cursor *);
extern int afs_end_cursor(struct afs_addr_cursor *);
-extern void afs_merge_fs_addr4(struct afs_addr_list *, __be32, u16);
-extern void afs_merge_fs_addr6(struct afs_addr_list *, __be32 *, u16);
+extern int afs_merge_fs_addr4(struct afs_net *net, struct afs_addr_list *addr,
+ __be32 xdr, u16 port);
+extern int afs_merge_fs_addr6(struct afs_net *net, struct afs_addr_list *addr,
+ __be32 *xdr, u16 port);
/*
* callback.c
@@ -1404,8 +1407,7 @@ extern void __exit afs_clean_up_permit_cache(void);
*/
extern spinlock_t afs_server_peer_lock;
-extern struct afs_server *afs_find_server(struct afs_net *,
- const struct sockaddr_rxrpc *);
+extern struct afs_server *afs_find_server(struct afs_net *, const struct rxrpc_peer *);
extern struct afs_server *afs_find_server_by_uuid(struct afs_net *, const uuid_t *);
extern struct afs_server *afs_lookup_server(struct afs_cell *, struct key *, const uuid_t *, u32);
extern struct afs_server *afs_get_server(struct afs_server *, enum afs_server_trace);
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index ab9cd986cfd9..8a65a06908d2 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -307,7 +307,7 @@ static int afs_proc_cell_vlservers_show(struct seq_file *m, void *v)
for (i = 0; i < alist->nr_addrs; i++)
seq_printf(m, " %c %pISpc\n",
alist->preferred == i ? '>' : '-',
- &alist->addrs[i].srx.transport);
+ rxrpc_kernel_remote_addr(alist->addrs[i].peer));
}
seq_printf(m, " info: fl=%lx rtt=%d\n", vlserver->flags, vlserver->rtt);
seq_printf(m, " probe: fl=%x e=%d ac=%d out=%d\n",
@@ -398,9 +398,10 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
seq_printf(m, " - ALIST v=%u rsp=%lx f=%lx\n",
alist->version, alist->responded, alist->failed);
for (i = 0; i < alist->nr_addrs; i++)
- seq_printf(m, " [%x] %pISpc%s\n",
- i, &alist->addrs[i].srx.transport,
- alist->preferred == i ? "*" : "");
+ seq_printf(m, " [%x] %pISpc%s rtt=%d\n",
+ i, rxrpc_kernel_remote_addr(alist->addrs[i].peer),
+ alist->preferred == i ? "*" : "",
+ rxrpc_kernel_get_srtt(alist->addrs[i].peer));
return 0;
}
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 993e20d752d9..1c8f26a7f128 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -113,7 +113,7 @@ bool afs_select_fileserver(struct afs_operation *op)
struct afs_server *server;
struct afs_vnode *vnode = op->file[0].vnode;
struct afs_error e;
- u32 rtt;
+ unsigned int rtt;
int error = op->ac.error, i;
_enter("%lx[%d],%lx[%d],%d,%d",
@@ -419,7 +419,7 @@ bool afs_select_fileserver(struct afs_operation *op)
}
op->index = -1;
- rtt = U32_MAX;
+ rtt = UINT_MAX;
for (i = 0; i < op->server_list->nr_servers; i++) {
struct afs_server *s = op->server_list->servers[i].server;
@@ -487,7 +487,7 @@ bool afs_select_fileserver(struct afs_operation *op)
_debug("address [%u] %u/%u %pISp",
op->index, op->ac.index, op->ac.alist->nr_addrs,
- &op->ac.alist->addrs[op->ac.index].srx.transport);
+ rxrpc_kernel_remote_addr(op->ac.alist->addrs[op->ac.index].peer));
_leave(" = t");
return true;
diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
index 2b1a31b4249c..1a18fcbdec80 100644
--- a/fs/afs/rxrpc.c
+++ b/fs/afs/rxrpc.c
@@ -296,7 +296,8 @@ static void afs_notify_end_request_tx(struct sock *sock,
*/
void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
{
- struct sockaddr_rxrpc *srx = &ac->alist->addrs[ac->index].srx;
+ struct afs_address *addr = &ac->alist->addrs[ac->index];
+ struct rxrpc_peer *peer = addr->peer;
struct rxrpc_call *rxcall;
struct msghdr msg;
struct kvec iov[1];
@@ -304,7 +305,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
s64 tx_total_len;
int ret;
- _enter(",{%pISp},", &srx->transport);
+ _enter(",{%pISp},", rxrpc_kernel_remote_addr(addr->peer));
ASSERT(call->type != NULL);
ASSERT(call->type->name != NULL);
@@ -333,7 +334,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
}
/* create a call */
- rxcall = rxrpc_kernel_begin_call(call->net->socket, srx, call->key,
+ rxcall = rxrpc_kernel_begin_call(call->net->socket, peer, call->key,
(unsigned long)call,
tx_total_len,
call->max_lifespan,
@@ -341,6 +342,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
(call->async ?
afs_wake_up_async_call :
afs_wake_up_call_waiter),
+ addr->service_id,
call->upgrade,
(call->intr ? RXRPC_PREINTERRUPTIBLE :
RXRPC_UNINTERRUPTIBLE),
@@ -461,7 +463,7 @@ static void afs_log_error(struct afs_call *call, s32 remote_abort)
max = m + 1;
pr_notice("kAFS: Peer reported %s failure on %s [%pISp]\n",
msg, call->type->name,
- &call->alist->addrs[call->addr_ix].srx.transport);
+ rxrpc_kernel_remote_addr(call->alist->addrs[call->addr_ix].peer));
}
}
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 4b98788c7b12..831254d8ef9c 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -21,13 +21,12 @@ static void __afs_put_server(struct afs_net *, struct afs_server *);
/*
* Find a server by one of its addresses.
*/
-struct afs_server *afs_find_server(struct afs_net *net,
- const struct sockaddr_rxrpc *srx)
+struct afs_server *afs_find_server(struct afs_net *net, const struct rxrpc_peer *peer)
{
const struct afs_addr_list *alist;
struct afs_server *server = NULL;
unsigned int i;
- int seq = 0, diff;
+ int seq = 0;
rcu_read_lock();
@@ -37,37 +36,11 @@ struct afs_server *afs_find_server(struct afs_net *net,
server = NULL;
read_seqbegin_or_lock(&net->fs_addr_lock, &seq);
- if (srx->transport.family == AF_INET6) {
- const struct sockaddr_in6 *a = &srx->transport.sin6, *b;
- hlist_for_each_entry_rcu(server, &net->fs_addresses6, addr6_link) {
- alist = rcu_dereference(server->addresses);
- for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
- b = &alist->addrs[i].srx.transport.sin6;
- diff = ((u16 __force)a->sin6_port -
- (u16 __force)b->sin6_port);
- if (diff == 0)
- diff = memcmp(&a->sin6_addr,
- &b->sin6_addr,
- sizeof(struct in6_addr));
- if (diff == 0)
- goto found;
- }
- }
- } else {
- const struct sockaddr_in *a = &srx->transport.sin, *b;
- hlist_for_each_entry_rcu(server, &net->fs_addresses4, addr4_link) {
- alist = rcu_dereference(server->addresses);
- for (i = 0; i < alist->nr_ipv4; i++) {
- b = &alist->addrs[i].srx.transport.sin;
- diff = ((u16 __force)a->sin_port -
- (u16 __force)b->sin_port);
- if (diff == 0)
- diff = ((u32 __force)a->sin_addr.s_addr -
- (u32 __force)b->sin_addr.s_addr);
- if (diff == 0)
- goto found;
- }
- }
+ hlist_for_each_entry_rcu(server, &net->fs_addresses6, addr6_link) {
+ alist = rcu_dereference(server->addresses);
+ for (i = 0; i < alist->nr_addrs; i++)
+ if (alist->addrs[i].peer == peer)
+ goto found;
}
server = NULL;
diff --git a/fs/afs/vl_alias.c b/fs/afs/vl_alias.c
index d3c0df70a1a5..6fdf9f1bedc0 100644
--- a/fs/afs/vl_alias.c
+++ b/fs/afs/vl_alias.c
@@ -32,55 +32,6 @@ static struct afs_volume *afs_sample_volume(struct afs_cell *cell, struct key *k
return volume;
}
-/*
- * Compare two addresses.
- */
-static int afs_compare_addrs(const struct sockaddr_rxrpc *srx_a,
- const struct sockaddr_rxrpc *srx_b)
-{
- short port_a, port_b;
- int addr_a, addr_b, diff;
-
- diff = (short)srx_a->transport_type - (short)srx_b->transport_type;
- if (diff)
- goto out;
-
- switch (srx_a->transport_type) {
- case AF_INET: {
- const struct sockaddr_in *a = &srx_a->transport.sin;
- const struct sockaddr_in *b = &srx_b->transport.sin;
- addr_a = ntohl(a->sin_addr.s_addr);
- addr_b = ntohl(b->sin_addr.s_addr);
- diff = addr_a - addr_b;
- if (diff == 0) {
- port_a = ntohs(a->sin_port);
- port_b = ntohs(b->sin_port);
- diff = port_a - port_b;
- }
- break;
- }
-
- case AF_INET6: {
- const struct sockaddr_in6 *a = &srx_a->transport.sin6;
- const struct sockaddr_in6 *b = &srx_b->transport.sin6;
- diff = memcmp(&a->sin6_addr, &b->sin6_addr, 16);
- if (diff == 0) {
- port_a = ntohs(a->sin6_port);
- port_b = ntohs(b->sin6_port);
- diff = port_a - port_b;
- }
- break;
- }
-
- default:
- WARN_ON(1);
- diff = 1;
- }
-
-out:
- return diff;
-}
-
/*
* Compare the address lists of a pair of fileservers.
*/
@@ -94,9 +45,9 @@ static int afs_compare_fs_alists(const struct afs_server *server_a,
lb = rcu_dereference(server_b->addresses);
while (a < la->nr_addrs && b < lb->nr_addrs) {
- const struct sockaddr_rxrpc *srx_a = &la->addrs[a].srx;
- const struct sockaddr_rxrpc *srx_b = &lb->addrs[b].srx;
- int diff = afs_compare_addrs(srx_a, srx_b);
+ unsigned long pa = (unsigned long)la->addrs[a].peer;
+ unsigned long pb = (unsigned long)lb->addrs[b].peer;
+ long diff = pa - pb;
if (diff < 0) {
a++;
diff --git a/fs/afs/vl_list.c b/fs/afs/vl_list.c
index acc48216136a..ba89140eee9e 100644
--- a/fs/afs/vl_list.c
+++ b/fs/afs/vl_list.c
@@ -83,14 +83,15 @@ static u16 afs_extract_le16(const u8 **_b)
/*
* Build a VL server address list from a DNS queried server list.
*/
-static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
+static struct afs_addr_list *afs_extract_vl_addrs(struct afs_net *net,
+ const u8 **_b, const u8 *end,
u8 nr_addrs, u16 port)
{
struct afs_addr_list *alist;
const u8 *b = *_b;
int ret = -EINVAL;
- alist = afs_alloc_addrlist(nr_addrs, VL_SERVICE, port);
+ alist = afs_alloc_addrlist(nr_addrs, VL_SERVICE);
if (!alist)
return ERR_PTR(-ENOMEM);
if (nr_addrs == 0)
@@ -109,7 +110,9 @@ static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
goto error;
}
memcpy(x, b, 4);
- afs_merge_fs_addr4(alist, x[0], port);
+ ret = afs_merge_fs_addr4(net, alist, x[0], port);
+ if (ret < 0)
+ goto error;
b += 4;
break;
@@ -119,7 +122,9 @@ static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
goto error;
}
memcpy(x, b, 16);
- afs_merge_fs_addr6(alist, x, port);
+ ret = afs_merge_fs_addr6(net, alist, x, port);
+ if (ret < 0)
+ goto error;
b += 16;
break;
@@ -247,7 +252,7 @@ struct afs_vlserver_list *afs_extract_vlserver_list(struct afs_cell *cell,
/* Extract the addresses - note that we can't skip this as we
* have to advance the payload pointer.
*/
- addrs = afs_extract_vl_addrs(&b, end, bs.nr_addrs, bs.port);
+ addrs = afs_extract_vl_addrs(cell->net, &b, end, bs.nr_addrs, bs.port);
if (IS_ERR(addrs)) {
ret = PTR_ERR(addrs);
goto error_2;
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index bdd9372e3fb2..9551aef07cee 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -48,6 +48,7 @@ void afs_vlserver_probe_result(struct afs_call *call)
{
struct afs_addr_list *alist = call->alist;
struct afs_vlserver *server = call->vlserver;
+ struct afs_address *addr = &alist->addrs[call->addr_ix];
unsigned int server_index = call->server_index;
unsigned int rtt_us = 0;
unsigned int index = call->addr_ix;
@@ -106,16 +107,16 @@ void afs_vlserver_probe_result(struct afs_call *call)
if (call->service_id == YFS_VL_SERVICE) {
server->probe.flags |= AFS_VLSERVER_PROBE_IS_YFS;
set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx.srx_service = call->service_id;
+ addr->service_id = call->service_id;
} else {
server->probe.flags |= AFS_VLSERVER_PROBE_NOT_YFS;
if (!(server->probe.flags & AFS_VLSERVER_PROBE_IS_YFS)) {
clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
- alist->addrs[index].srx.srx_service = call->service_id;
+ addr->service_id = call->service_id;
}
}
- rxrpc_kernel_get_srtt(call->net->socket, call->rxcall, &rtt_us);
+ rtt_us = rxrpc_kernel_get_srtt(addr->peer);
if (rtt_us < server->probe.rtt) {
server->probe.rtt = rtt_us;
server->rtt = rtt_us;
@@ -130,8 +131,9 @@ void afs_vlserver_probe_result(struct afs_call *call)
out:
spin_unlock(&server->probe_lock);
- _debug("probe [%u][%u] %pISpc rtt=%u ret=%d",
- server_index, index, &alist->addrs[index].srx.transport, rtt_us, ret);
+ _debug("probe [%u][%u] %pISpc rtt=%d ret=%d",
+ server_index, index, rxrpc_kernel_remote_addr(addr->peer),
+ rtt_us, ret);
afs_done_one_vl_probe(server, have_result);
}
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index 0d0b54819128..af445e7d3a12 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -86,7 +86,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
struct afs_addr_list *alist;
struct afs_vlserver *vlserver;
struct afs_error e;
- u32 rtt;
+ unsigned int rtt;
int error = vc->ac.error, i;
_enter("%lx[%d],%lx[%d],%d,%d",
@@ -188,7 +188,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
goto selected_server;
vc->index = -1;
- rtt = U32_MAX;
+ rtt = UINT_MAX;
for (i = 0; i < vc->server_list->nr_servers; i++) {
struct afs_vlserver *s = vc->server_list->servers[i].server;
@@ -243,7 +243,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
_debug("VL address %d/%d", vc->ac.index, vc->ac.alist->nr_addrs);
- _leave(" = t %pISpc", &vc->ac.alist->addrs[vc->ac.index].srx.transport);
+ _leave(" = t %pISpc", rxrpc_kernel_remote_addr(vc->ac.alist->addrs[vc->ac.index].peer));
return true;
next_server:
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index 00fca3c66ba6..41e7932d75c6 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -208,7 +208,7 @@ static int afs_deliver_vl_get_addrs_u(struct afs_call *call)
count = ntohl(*bp);
nentries = min(nentries, count);
- alist = afs_alloc_addrlist(nentries, FS_SERVICE, AFS_FS_PORT);
+ alist = afs_alloc_addrlist(nentries, FS_SERVICE);
if (!alist)
return -ENOMEM;
alist->version = uniquifier;
@@ -230,9 +230,13 @@ static int afs_deliver_vl_get_addrs_u(struct afs_call *call)
alist = call->ret_alist;
bp = call->buffer;
count = min(call->count, 4U);
- for (i = 0; i < count; i++)
- if (alist->nr_addrs < call->count2)
- afs_merge_fs_addr4(alist, *bp++, AFS_FS_PORT);
+ for (i = 0; i < count; i++) {
+ if (alist->nr_addrs < call->count2) {
+ ret = afs_merge_fs_addr4(call->net, alist, *bp++, AFS_FS_PORT);
+ if (ret < 0)
+ return ret;
+ }
+ }
call->count -= count;
if (call->count > 0)
@@ -450,7 +454,7 @@ static int afs_deliver_yfsvl_get_endpoints(struct afs_call *call)
if (call->count > YFS_MAXENDPOINTS)
return afs_protocol_error(call, afs_eproto_yvl_fsendpt_num);
- alist = afs_alloc_addrlist(call->count, FS_SERVICE, AFS_FS_PORT);
+ alist = afs_alloc_addrlist(call->count, FS_SERVICE);
if (!alist)
return -ENOMEM;
alist->version = uniquifier;
@@ -488,14 +492,18 @@ static int afs_deliver_yfsvl_get_endpoints(struct afs_call *call)
if (ntohl(bp[0]) != sizeof(__be32) * 2)
return afs_protocol_error(
call, afs_eproto_yvl_fsendpt4_len);
- afs_merge_fs_addr4(alist, bp[1], ntohl(bp[2]));
+ ret = afs_merge_fs_addr4(call->net, alist, bp[1], ntohl(bp[2]));
+ if (ret < 0)
+ return ret;
bp += 3;
break;
case YFS_ENDPOINT_IPV6:
if (ntohl(bp[0]) != sizeof(__be32) * 5)
return afs_protocol_error(
call, afs_eproto_yvl_fsendpt6_len);
- afs_merge_fs_addr6(alist, bp + 1, ntohl(bp[5]));
+ ret = afs_merge_fs_addr6(call->net, alist, bp + 1, ntohl(bp[5]));
+ if (ret < 0)
+ return ret;
bp += 6;
break;
default:
diff --git a/include/net/af_rxrpc.h b/include/net/af_rxrpc.h
index 5531dd08061e..0754c463224a 100644
--- a/include/net/af_rxrpc.h
+++ b/include/net/af_rxrpc.h
@@ -15,6 +15,7 @@ struct key;
struct sock;
struct socket;
struct rxrpc_call;
+struct rxrpc_peer;
enum rxrpc_abort_reason;
enum rxrpc_interruptibility {
@@ -41,13 +42,14 @@ void rxrpc_kernel_new_call_notification(struct socket *,
rxrpc_notify_new_call_t,
rxrpc_discard_new_call_t);
struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
- struct sockaddr_rxrpc *srx,
+ struct rxrpc_peer *peer,
struct key *key,
unsigned long user_call_ID,
s64 tx_total_len,
u32 hard_timeout,
gfp_t gfp,
rxrpc_notify_rx_t notify_rx,
+ u16 service_id,
bool upgrade,
enum rxrpc_interruptibility interruptibility,
unsigned int debug_id);
@@ -60,9 +62,14 @@ bool rxrpc_kernel_abort_call(struct socket *, struct rxrpc_call *,
u32, int, enum rxrpc_abort_reason);
void rxrpc_kernel_shutdown_call(struct socket *sock, struct rxrpc_call *call);
void rxrpc_kernel_put_call(struct socket *sock, struct rxrpc_call *call);
-void rxrpc_kernel_get_peer(struct socket *, struct rxrpc_call *,
- struct sockaddr_rxrpc *);
-bool rxrpc_kernel_get_srtt(struct socket *, struct rxrpc_call *, u32 *);
+struct rxrpc_peer *rxrpc_kernel_lookup_peer(struct socket *sock,
+ struct sockaddr_rxrpc *srx, gfp_t gfp);
+void rxrpc_kernel_put_peer(struct rxrpc_peer *peer);
+struct rxrpc_peer *rxrpc_kernel_get_peer(struct rxrpc_peer *peer);
+struct rxrpc_peer *rxrpc_kernel_get_call_peer(struct socket *sock, struct rxrpc_call *call);
+const struct sockaddr_rxrpc *rxrpc_kernel_remote_srx(const struct rxrpc_peer *peer);
+const struct sockaddr *rxrpc_kernel_remote_addr(const struct rxrpc_peer *peer);
+unsigned int rxrpc_kernel_get_srtt(const struct rxrpc_peer *);
int rxrpc_kernel_charge_accept(struct socket *, rxrpc_notify_rx_t,
rxrpc_user_attach_call_t, unsigned long, gfp_t,
unsigned int);
diff --git a/include/trace/events/rxrpc.h b/include/trace/events/rxrpc.h
index 4c53a5ef6257..90a1e39d620e 100644
--- a/include/trace/events/rxrpc.h
+++ b/include/trace/events/rxrpc.h
@@ -178,7 +178,9 @@
#define rxrpc_peer_traces \
EM(rxrpc_peer_free, "FREE ") \
EM(rxrpc_peer_get_accept, "GET accept ") \
+ EM(rxrpc_peer_get_application, "GET app ") \
EM(rxrpc_peer_get_bundle, "GET bundle ") \
+ EM(rxrpc_peer_get_call, "GET call ") \
EM(rxrpc_peer_get_client_conn, "GET cln-conn") \
EM(rxrpc_peer_get_input, "GET input ") \
EM(rxrpc_peer_get_input_error, "GET inpt-err") \
@@ -187,6 +189,7 @@
EM(rxrpc_peer_get_service_conn, "GET srv-conn") \
EM(rxrpc_peer_new_client, "NEW client ") \
EM(rxrpc_peer_new_prealloc, "NEW prealloc") \
+ EM(rxrpc_peer_put_application, "PUT app ") \
EM(rxrpc_peer_put_bundle, "PUT bundle ") \
EM(rxrpc_peer_put_call, "PUT call ") \
EM(rxrpc_peer_put_conn, "PUT conn ") \
diff --git a/net/rxrpc/af_rxrpc.c b/net/rxrpc/af_rxrpc.c
index fa8aec78f63d..465bfe5eb061 100644
--- a/net/rxrpc/af_rxrpc.c
+++ b/net/rxrpc/af_rxrpc.c
@@ -258,16 +258,62 @@ static int rxrpc_listen(struct socket *sock, int backlog)
return ret;
}
+/**
+ * rxrpc_kernel_lookup_peer - Obtain remote transport endpoint for an address
+ * @sock: The socket through which it will be accessed
+ * @srx: The network address
+ * @gfp: Allocation flags
+ *
+ * Lookup or create a remote transport endpoint record for the specified
+ * address and return it with a ref held.
+ */
+struct rxrpc_peer *rxrpc_kernel_lookup_peer(struct socket *sock,
+ struct sockaddr_rxrpc *srx, gfp_t gfp)
+{
+ struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
+ int ret;
+
+ ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
+ if (ret < 0)
+ return ERR_PTR(ret);
+
+ return rxrpc_lookup_peer(rx->local, srx, gfp);
+}
+EXPORT_SYMBOL(rxrpc_kernel_lookup_peer);
+
+/**
+ * rxrpc_kernel_get_peer - Get a reference on a peer
+ * @peer: The peer to get a reference on.
+ *
+ * Get a record for the remote peer in a call.
+ */
+struct rxrpc_peer *rxrpc_kernel_get_peer(struct rxrpc_peer *peer)
+{
+ return peer ? rxrpc_get_peer(peer, rxrpc_peer_get_application) : NULL;
+}
+EXPORT_SYMBOL(rxrpc_kernel_get_peer);
+
+/**
+ * rxrpc_kernel_put_peer - Allow a kernel app to drop a peer reference
+ * @peer: The peer to drop a ref on
+ */
+void rxrpc_kernel_put_peer(struct rxrpc_peer *peer)
+{
+ rxrpc_put_peer(peer, rxrpc_peer_put_application);
+}
+EXPORT_SYMBOL(rxrpc_kernel_put_peer);
+
/**
* rxrpc_kernel_begin_call - Allow a kernel service to begin a call
* @sock: The socket on which to make the call
- * @srx: The address of the peer to contact
+ * @peer: The peer to contact
* @key: The security context to use (defaults to socket setting)
* @user_call_ID: The ID to use
* @tx_total_len: Total length of data to transmit during the call (or -1)
* @hard_timeout: The maximum lifespan of the call in sec
* @gfp: The allocation constraints
* @notify_rx: Where to send notifications instead of socket queue
+ * @service_id: The ID of the service to contact
* @upgrade: Request service upgrade for call
* @interruptibility: The call is interruptible, or can be canceled.
* @debug_id: The debug ID for tracing to be assigned to the call
@@ -280,13 +326,14 @@ static int rxrpc_listen(struct socket *sock, int backlog)
* supplying @srx and @key.
*/
struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
- struct sockaddr_rxrpc *srx,
+ struct rxrpc_peer *peer,
struct key *key,
unsigned long user_call_ID,
s64 tx_total_len,
u32 hard_timeout,
gfp_t gfp,
rxrpc_notify_rx_t notify_rx,
+ u16 service_id,
bool upgrade,
enum rxrpc_interruptibility interruptibility,
unsigned int debug_id)
@@ -295,13 +342,11 @@ struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
struct rxrpc_call_params p;
struct rxrpc_call *call;
struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
- int ret;
_enter(",,%x,%lx", key_serial(key), user_call_ID);
- ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
- if (ret < 0)
- return ERR_PTR(ret);
+ if (WARN_ON_ONCE(peer->local != rx->local))
+ return ERR_PTR(-EIO);
lock_sock(&rx->sk);
@@ -319,12 +364,13 @@ struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
memset(&cp, 0, sizeof(cp));
cp.local = rx->local;
+ cp.peer = peer;
cp.key = key;
cp.security_level = rx->min_sec_level;
cp.exclusive = false;
cp.upgrade = upgrade;
- cp.service_id = srx->srx_service;
- call = rxrpc_new_client_call(rx, &cp, srx, &p, gfp, debug_id);
+ cp.service_id = service_id;
+ call = rxrpc_new_client_call(rx, &cp, &p, gfp, debug_id);
/* The socket has been unlocked. */
if (!IS_ERR(call)) {
call->notify_rx = notify_rx;
diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
index e8e14c6f904d..8eea7a487380 100644
--- a/net/rxrpc/ar-internal.h
+++ b/net/rxrpc/ar-internal.h
@@ -364,6 +364,7 @@ struct rxrpc_conn_proto {
struct rxrpc_conn_parameters {
struct rxrpc_local *local; /* Representation of local endpoint */
+ struct rxrpc_peer *peer; /* Representation of remote endpoint */
struct key *key; /* Security details */
bool exclusive; /* T if conn is exclusive */
bool upgrade; /* T if service ID can be upgraded */
@@ -867,7 +868,6 @@ struct rxrpc_call *rxrpc_find_call_by_user_ID(struct rxrpc_sock *, unsigned long
struct rxrpc_call *rxrpc_alloc_call(struct rxrpc_sock *, gfp_t, unsigned int);
struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *,
struct rxrpc_conn_parameters *,
- struct sockaddr_rxrpc *,
struct rxrpc_call_params *, gfp_t,
unsigned int);
void rxrpc_start_call_timer(struct rxrpc_call *call);
diff --git a/net/rxrpc/call_object.c b/net/rxrpc/call_object.c
index 773eecd1e979..beea25ac88f5 100644
--- a/net/rxrpc/call_object.c
+++ b/net/rxrpc/call_object.c
@@ -193,7 +193,6 @@ struct rxrpc_call *rxrpc_alloc_call(struct rxrpc_sock *rx, gfp_t gfp,
* Allocate a new client call.
*/
static struct rxrpc_call *rxrpc_alloc_client_call(struct rxrpc_sock *rx,
- struct sockaddr_rxrpc *srx,
struct rxrpc_conn_parameters *cp,
struct rxrpc_call_params *p,
gfp_t gfp,
@@ -211,10 +210,12 @@ static struct rxrpc_call *rxrpc_alloc_client_call(struct rxrpc_sock *rx,
now = ktime_get_real();
call->acks_latest_ts = now;
call->cong_tstamp = now;
- call->dest_srx = *srx;
+ call->dest_srx = cp->peer->srx;
+ call->dest_srx.srx_service = cp->service_id;
call->interruptibility = p->interruptibility;
call->tx_total_len = p->tx_total_len;
call->key = key_get(cp->key);
+ call->peer = rxrpc_get_peer(cp->peer, rxrpc_peer_get_call);
call->local = rxrpc_get_local(cp->local, rxrpc_local_get_call);
call->security_level = cp->security_level;
if (p->kernel)
@@ -306,10 +307,6 @@ static int rxrpc_connect_call(struct rxrpc_call *call, gfp_t gfp)
_enter("{%d,%lx},", call->debug_id, call->user_call_ID);
- call->peer = rxrpc_lookup_peer(local, &call->dest_srx, gfp);
- if (!call->peer)
- goto error;
-
ret = rxrpc_look_up_bundle(call, gfp);
if (ret < 0)
goto error;
@@ -334,7 +331,6 @@ static int rxrpc_connect_call(struct rxrpc_call *call, gfp_t gfp)
*/
struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *rx,
struct rxrpc_conn_parameters *cp,
- struct sockaddr_rxrpc *srx,
struct rxrpc_call_params *p,
gfp_t gfp,
unsigned int debug_id)
@@ -349,13 +345,18 @@ struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *rx,
_enter("%p,%lx", rx, p->user_call_ID);
+ if (WARN_ON_ONCE(!cp->peer)) {
+ release_sock(&rx->sk);
+ return ERR_PTR(-EIO);
+ }
+
limiter = rxrpc_get_call_slot(p, gfp);
if (!limiter) {
release_sock(&rx->sk);
return ERR_PTR(-ERESTARTSYS);
}
- call = rxrpc_alloc_client_call(rx, srx, cp, p, gfp, debug_id);
+ call = rxrpc_alloc_client_call(rx, cp, p, gfp, debug_id);
if (IS_ERR(call)) {
release_sock(&rx->sk);
up(limiter);
diff --git a/net/rxrpc/peer_object.c b/net/rxrpc/peer_object.c
index 8d7a715a0bb1..65ea57b427a1 100644
--- a/net/rxrpc/peer_object.c
+++ b/net/rxrpc/peer_object.c
@@ -22,6 +22,8 @@
#include <net/ip6_route.h>
#include "ar-internal.h"
+static const struct sockaddr_rxrpc rxrpc_null_addr;
+
/*
* Hash a peer key.
*/
@@ -457,39 +459,51 @@ void rxrpc_destroy_all_peers(struct rxrpc_net *rxnet)
}
/**
- * rxrpc_kernel_get_peer - Get the peer address of a call
+ * rxrpc_kernel_get_call_peer - Get the peer address of a call
* @sock: The socket on which the call is in progress.
* @call: The call to query
- * @_srx: Where to place the result
*
- * Get the address of the remote peer in a call.
+ * Get a record for the remote peer in a call.
*/
-void rxrpc_kernel_get_peer(struct socket *sock, struct rxrpc_call *call,
- struct sockaddr_rxrpc *_srx)
+struct rxrpc_peer *rxrpc_kernel_get_call_peer(struct socket *sock, struct rxrpc_call *call)
{
- *_srx = call->peer->srx;
+ return call->peer;
}
-EXPORT_SYMBOL(rxrpc_kernel_get_peer);
+EXPORT_SYMBOL(rxrpc_kernel_get_call_peer);
/**
* rxrpc_kernel_get_srtt - Get a call's peer smoothed RTT
- * @sock: The socket on which the call is in progress.
- * @call: The call to query
- * @_srtt: Where to store the SRTT value.
+ * @peer: The peer to query
*
- * Get the call's peer smoothed RTT in uS.
+ * Get the call's peer smoothed RTT in uS or UINT_MAX if we have no samples.
*/
-bool rxrpc_kernel_get_srtt(struct socket *sock, struct rxrpc_call *call,
- u32 *_srtt)
+unsigned int rxrpc_kernel_get_srtt(const struct rxrpc_peer *peer)
{
- struct rxrpc_peer *peer = call->peer;
+ return peer->rtt_count > 0 ? peer->srtt_us >> 3 : UINT_MAX;
+}
+EXPORT_SYMBOL(rxrpc_kernel_get_srtt);
- if (peer->rtt_count == 0) {
- *_srtt = 1000000; /* 1S */
- return false;
- }
+/**
+ * rxrpc_kernel_remote_srx - Get the address of a peer
+ * @peer: The peer to query
+ *
+ * Get a pointer to the address from a peer record. The caller is responsible
+ * for making sure that the address is not deallocated.
+ */
+const struct sockaddr_rxrpc *rxrpc_kernel_remote_srx(const struct rxrpc_peer *peer)
+{
+ return peer ? &peer->srx : &rxrpc_null_addr;
+}
- *_srtt = call->peer->srtt_us >> 3;
- return true;
+/**
+ * rxrpc_kernel_remote_addr - Get the peer transport address of a call
+ * @peer: The peer to query
+ *
+ * Get a pointer to the transport address from a peer record. The caller is
+ * responsible for making sure that the address is not deallocated.
+ */
+const struct sockaddr *rxrpc_kernel_remote_addr(const struct rxrpc_peer *peer)
+{
+ return (const struct sockaddr *)
+ (peer ? &peer->srx.transport : &rxrpc_null_addr.transport);
}
-EXPORT_SYMBOL(rxrpc_kernel_get_srtt);
diff --git a/net/rxrpc/sendmsg.c b/net/rxrpc/sendmsg.c
index 8e0b94714e84..5677d5690a02 100644
--- a/net/rxrpc/sendmsg.c
+++ b/net/rxrpc/sendmsg.c
@@ -572,6 +572,7 @@ rxrpc_new_client_call_for_sendmsg(struct rxrpc_sock *rx, struct msghdr *msg,
__acquires(&call->user_mutex)
{
struct rxrpc_conn_parameters cp;
+ struct rxrpc_peer *peer;
struct rxrpc_call *call;
struct key *key;
@@ -584,21 +585,29 @@ rxrpc_new_client_call_for_sendmsg(struct rxrpc_sock *rx, struct msghdr *msg,
return ERR_PTR(-EDESTADDRREQ);
}
+ peer = rxrpc_lookup_peer(rx->local, srx, GFP_KERNEL);
+ if (!peer) {
+ release_sock(&rx->sk);
+ return ERR_PTR(-ENOMEM);
+ }
+
key = rx->key;
if (key && !rx->key->payload.data[0])
key = NULL;
memset(&cp, 0, sizeof(cp));
cp.local = rx->local;
+ cp.peer = peer;
cp.key = rx->key;
cp.security_level = rx->min_sec_level;
cp.exclusive = rx->exclusive | p->exclusive;
cp.upgrade = p->upgrade;
cp.service_id = srx->srx_service;
- call = rxrpc_new_client_call(rx, &cp, srx, &p->call, GFP_KERNEL,
+ call = rxrpc_new_client_call(rx, &cp, &p->call, GFP_KERNEL,
atomic_inc_return(&rxrpc_debug_id));
/* The socket is now unlocked */
+ rxrpc_put_peer(peer, rxrpc_peer_put_application);
_leave(" = %p\n", call);
return call;
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* Re: [PATCH 10/41] rxrpc, afs: Allow afs to pin rxrpc_peer objects
2023-11-09 15:39 ` [PATCH 10/41] rxrpc, afs: Allow afs to pin rxrpc_peer objects David Howells
@ 2023-11-09 17:48 ` Marc Dionne
0 siblings, 0 replies; 57+ messages in thread
From: Marc Dionne @ 2023-11-09 17:48 UTC (permalink / raw)
To: David Howells; +Cc: linux-afs, linux-fsdevel, linux-kernel
On Thu, Nov 9, 2023 at 11:40 AM David Howells <dhowells@redhat.com> wrote:
>
> Change rxrpc's API such that:
>
> (1) A new function, rxrpc_kernel_lookup_peer(), is provided to look up an
> rxrpc_peer record for a remote address and a corresponding function,
> rxrpc_kernel_put_peer(), is provided to dispose of it again.
>
> (2) When setting up a call, the rxrpc_peer object used during a call is
> now passed in rather than being set up by rxrpc_connect_call(). For
> afs, this meenat passing it to rxrpc_kernel_begin_call() rather than
> the full address (the service ID then has to be passed in as a
> separate parameter).
>
> (3) A new function, rxrpc_kernel_remote_addr(), is added so that afs can
> get a pointer to the transport address for display purposed, and
> another, rxrpc_kernel_remote_srx(), to gain a pointer to the full
> rxrpc address.
>
> (4) The function to retrieve the RTT from a call, rxrpc_kernel_get_srtt(),
> is then altered to take a peer. This now returns the RTT or -1 if
> there are insufficient samples.
>
> (5) Rename rxrpc_kernel_get_peer() to rxrpc_kernel_call_get_peer().
>
> (6) Provide a new function, rxrpc_kernel_get_peer(), to get a ref on a
> peer the caller already has.
>
> This allows the afs filesystem to pin the rxrpc_peer records that it is
> using, allowing faster lookups and pointer comparisons rather than
> comparing sockaddr_rxrpc contents. It also makes it easier to get hold of
> the RTT. The following changes are made to afs:
>
> (1) The addr_list struct's addrs[] elements now hold a peer struct pointer
> and a service ID rather than a sockaddr_rxrpc.
>
> (2) When displaying the transport address, rxrpc_kernel_remote_addr() is
> used.
>
> (3) The port arg is removed from afs_alloc_addrlist() since it's always
> overridden.
>
> (4) afs_merge_fs_addr4() and afs_merge_fs_addr6() do peer lookup and may
> now return an error that must be handled.
>
> (5) afs_find_server() now takes a peer pointer to specify the address.
>
> (6) afs_find_server(), afs_compare_fs_alists() and afs_merge_fs_addr[46]{}
> now do peer pointer comparison rather than address comparison.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Marc Dionne <marc.dionne@auristor.com>
> cc: linux-afs@lists.infradead.org
> ---
> fs/afs/addr_list.c | 125 ++++++++++++++++++-----------------
> fs/afs/cmservice.c | 5 +-
> fs/afs/fs_probe.c | 11 +--
> fs/afs/internal.h | 26 ++++----
> fs/afs/proc.c | 9 +--
> fs/afs/rotate.c | 6 +-
> fs/afs/rxrpc.c | 10 +--
> fs/afs/server.c | 41 ++----------
> fs/afs/vl_alias.c | 55 +--------------
> fs/afs/vl_list.c | 15 +++--
> fs/afs/vl_probe.c | 12 ++--
> fs/afs/vl_rotate.c | 6 +-
> fs/afs/vlclient.c | 22 ++++--
> include/net/af_rxrpc.h | 15 +++--
> include/trace/events/rxrpc.h | 3 +
> net/rxrpc/af_rxrpc.c | 62 ++++++++++++++---
> net/rxrpc/ar-internal.h | 2 +-
> net/rxrpc/call_object.c | 17 ++---
> net/rxrpc/peer_object.c | 56 ++++++++++------
> net/rxrpc/sendmsg.c | 11 ++-
> 20 files changed, 271 insertions(+), 238 deletions(-)
>
> diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
> index ac05a59e9d46..519821f5aedc 100644
> --- a/fs/afs/addr_list.c
> +++ b/fs/afs/addr_list.c
> @@ -13,26 +13,33 @@
> #include "internal.h"
> #include "afs_fs.h"
>
> +static void afs_free_addrlist(struct rcu_head *rcu)
> +{
> + struct afs_addr_list *alist = container_of(rcu, struct afs_addr_list, rcu);
> + unsigned int i;
> +
> + for (i = 0; i < alist->nr_addrs; i++)
> + rxrpc_kernel_put_peer(alist->addrs[i].peer);
> +}
> +
> /*
> * Release an address list.
> */
> void afs_put_addrlist(struct afs_addr_list *alist)
> {
> if (alist && refcount_dec_and_test(&alist->usage))
> - kfree_rcu(alist, rcu);
> + call_rcu(&alist->rcu, afs_free_addrlist);
> }
>
> /*
> * Allocate an address list.
> */
> -struct afs_addr_list *afs_alloc_addrlist(unsigned int nr,
> - unsigned short service,
> - unsigned short port)
> +struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id)
> {
> struct afs_addr_list *alist;
> unsigned int i;
>
> - _enter("%u,%u,%u", nr, service, port);
> + _enter("%u,%u", nr, service_id);
>
> if (nr > AFS_MAX_ADDRESSES)
> nr = AFS_MAX_ADDRESSES;
> @@ -44,16 +51,8 @@ struct afs_addr_list *afs_alloc_addrlist(unsigned int nr,
> refcount_set(&alist->usage, 1);
> alist->max_addrs = nr;
>
> - for (i = 0; i < nr; i++) {
> - struct sockaddr_rxrpc *srx = &alist->addrs[i].srx;
> - srx->srx_family = AF_RXRPC;
> - srx->srx_service = service;
> - srx->transport_type = SOCK_DGRAM;
> - srx->transport_len = sizeof(srx->transport.sin6);
> - srx->transport.sin6.sin6_family = AF_INET6;
> - srx->transport.sin6.sin6_port = htons(port);
> - }
> -
> + for (i = 0; i < nr; i++)
> + alist->addrs[i].service_id = service_id;
> return alist;
> }
>
> @@ -126,7 +125,7 @@ struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *net,
> if (!vllist->servers[0].server)
> goto error_vl;
>
> - alist = afs_alloc_addrlist(nr, service, AFS_VL_PORT);
> + alist = afs_alloc_addrlist(nr, service);
> if (!alist)
> goto error;
>
> @@ -197,9 +196,11 @@ struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *net,
> }
>
> if (family == AF_INET)
> - afs_merge_fs_addr4(alist, x[0], xport);
> + ret = afs_merge_fs_addr4(net, alist, x[0], xport);
> else
> - afs_merge_fs_addr6(alist, x, xport);
> + ret = afs_merge_fs_addr6(net, alist, x, xport);
> + if (ret < 0)
> + goto error;
>
> } while (p < end);
>
> @@ -271,25 +272,33 @@ struct afs_vlserver_list *afs_dns_query(struct afs_cell *cell, time64_t *_expiry
> /*
> * Merge an IPv4 entry into a fileserver address list.
> */
> -void afs_merge_fs_addr4(struct afs_addr_list *alist, __be32 xdr, u16 port)
> +int afs_merge_fs_addr4(struct afs_net *net, struct afs_addr_list *alist,
> + __be32 xdr, u16 port)
> {
> - struct sockaddr_rxrpc *srx;
> - u32 addr = ntohl(xdr);
> + struct sockaddr_rxrpc srx;
> + struct rxrpc_peer *peer;
> int i;
>
> if (alist->nr_addrs >= alist->max_addrs)
> - return;
> + return 0;
>
> - for (i = 0; i < alist->nr_ipv4; i++) {
> - struct sockaddr_in *a = &alist->addrs[i].srx.transport.sin;
> - u32 a_addr = ntohl(a->sin_addr.s_addr);
> - u16 a_port = ntohs(a->sin_port);
> + srx.srx_family = AF_RXRPC;
> + srx.transport_type = SOCK_DGRAM;
> + srx.transport_len = sizeof(srx.transport.sin);
> + srx.transport.sin.sin_family = AF_INET;
> + srx.transport.sin.sin_port = htons(port);
> + srx.transport.sin.sin_addr.s_addr = xdr;
>
> - if (addr == a_addr && port == a_port)
> - return;
> - if (addr == a_addr && port < a_port)
> - break;
> - if (addr < a_addr)
> + peer = rxrpc_kernel_lookup_peer(net->socket, &srx, GFP_KERNEL);
> + if (!peer)
> + return -ENOMEM;
> +
> + for (i = 0; i < alist->nr_ipv4; i++) {
> + if (peer == alist->addrs[i].peer) {
> + rxrpc_kernel_put_peer(peer);
> + return 0;
> + }
> + if (peer <= alist->addrs[i].peer)
> break;
> }
>
> @@ -298,38 +307,42 @@ void afs_merge_fs_addr4(struct afs_addr_list *alist, __be32 xdr, u16 port)
> alist->addrs + i,
> sizeof(alist->addrs[0]) * (alist->nr_addrs - i));
>
> - srx = &alist->addrs[i].srx;
> - srx->srx_family = AF_RXRPC;
> - srx->transport_type = SOCK_DGRAM;
> - srx->transport_len = sizeof(srx->transport.sin);
> - srx->transport.sin.sin_family = AF_INET;
> - srx->transport.sin.sin_port = htons(port);
> - srx->transport.sin.sin_addr.s_addr = xdr;
> + alist->addrs[i].peer = peer;
> alist->nr_ipv4++;
> alist->nr_addrs++;
> + return 0;
> }
>
> /*
> * Merge an IPv6 entry into a fileserver address list.
> */
> -void afs_merge_fs_addr6(struct afs_addr_list *alist, __be32 *xdr, u16 port)
> +int afs_merge_fs_addr6(struct afs_net *net, struct afs_addr_list *alist,
> + __be32 *xdr, u16 port)
> {
> - struct sockaddr_rxrpc *srx;
> - int i, diff;
> + struct sockaddr_rxrpc srx;
> + struct rxrpc_peer *peer;
> + int i;
>
> if (alist->nr_addrs >= alist->max_addrs)
> - return;
> + return 0;
>
> - for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
> - struct sockaddr_in6 *a = &alist->addrs[i].srx.transport.sin6;
> - u16 a_port = ntohs(a->sin6_port);
> + srx.srx_family = AF_RXRPC;
> + srx.transport_type = SOCK_DGRAM;
> + srx.transport_len = sizeof(srx.transport.sin6);
> + srx.transport.sin6.sin6_family = AF_INET6;
> + srx.transport.sin6.sin6_port = htons(port);
> + memcpy(&srx.transport.sin6.sin6_addr, xdr, 16);
>
> - diff = memcmp(xdr, &a->sin6_addr, 16);
> - if (diff == 0 && port == a_port)
> - return;
> - if (diff == 0 && port < a_port)
> - break;
> - if (diff < 0)
> + peer = rxrpc_kernel_lookup_peer(net->socket, &srx, GFP_KERNEL);
> + if (!peer)
> + return -ENOMEM;
> +
> + for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
> + if (peer == alist->addrs[i].peer) {
> + rxrpc_kernel_put_peer(peer);
> + return 0;
> + }
> + if (peer <= alist->addrs[i].peer)
> break;
> }
>
> @@ -337,15 +350,9 @@ void afs_merge_fs_addr6(struct afs_addr_list *alist, __be32 *xdr, u16 port)
> memmove(alist->addrs + i + 1,
> alist->addrs + i,
> sizeof(alist->addrs[0]) * (alist->nr_addrs - i));
> -
> - srx = &alist->addrs[i].srx;
> - srx->srx_family = AF_RXRPC;
> - srx->transport_type = SOCK_DGRAM;
> - srx->transport_len = sizeof(srx->transport.sin6);
> - srx->transport.sin6.sin6_family = AF_INET6;
> - srx->transport.sin6.sin6_port = htons(port);
> - memcpy(&srx->transport.sin6.sin6_addr, xdr, 16);
> + alist->addrs[i].peer = peer;
> alist->nr_addrs++;
> + return 0;
> }
>
> /*
> diff --git a/fs/afs/cmservice.c b/fs/afs/cmservice.c
> index d4ddb20d6732..99a3f20bc786 100644
> --- a/fs/afs/cmservice.c
> +++ b/fs/afs/cmservice.c
> @@ -146,10 +146,11 @@ static int afs_find_cm_server_by_peer(struct afs_call *call)
> {
> struct sockaddr_rxrpc srx;
> struct afs_server *server;
> + struct rxrpc_peer *peer;
>
> - rxrpc_kernel_get_peer(call->net->socket, call->rxcall, &srx);
> + peer = rxrpc_kernel_get_call_peer(call->net->socket, call->rxcall);
>
> - server = afs_find_server(call->net, &srx);
> + server = afs_find_server(call->net, peer);
> if (!server) {
> trace_afs_cm_no_server(call, &srx);
> return 0;
> diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
> index 3dd24842f277..58d28b82571e 100644
> --- a/fs/afs/fs_probe.c
> +++ b/fs/afs/fs_probe.c
> @@ -101,6 +101,7 @@ static void afs_fs_probe_not_done(struct afs_net *net,
> void afs_fileserver_probe_result(struct afs_call *call)
> {
> struct afs_addr_list *alist = call->alist;
> + struct afs_address *addr = &alist->addrs[call->addr_ix];
> struct afs_server *server = call->server;
> unsigned int index = call->addr_ix;
> unsigned int rtt_us = 0, cap0;
> @@ -153,12 +154,12 @@ void afs_fileserver_probe_result(struct afs_call *call)
> if (call->service_id == YFS_FS_SERVICE) {
> server->probe.is_yfs = true;
> set_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
> - alist->addrs[index].srx.srx_service = call->service_id;
> + addr->service_id = call->service_id;
> } else {
> server->probe.not_yfs = true;
> if (!server->probe.is_yfs) {
> clear_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
> - alist->addrs[index].srx.srx_service = call->service_id;
> + addr->service_id = call->service_id;
> }
> cap0 = ntohl(call->tmp);
> if (cap0 & AFS3_VICED_CAPABILITY_64BITFILES)
> @@ -167,7 +168,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
> clear_bit(AFS_SERVER_FL_HAS_FS64, &server->flags);
> }
>
> - rxrpc_kernel_get_srtt(call->net->socket, call->rxcall, &rtt_us);
> + rtt_us = rxrpc_kernel_get_srtt(addr->peer);
> if (rtt_us < server->probe.rtt) {
> server->probe.rtt = rtt_us;
> server->rtt = rtt_us;
> @@ -181,8 +182,8 @@ void afs_fileserver_probe_result(struct afs_call *call)
> out:
> spin_unlock(&server->probe_lock);
>
> - _debug("probe %pU [%u] %pISpc rtt=%u ret=%d",
> - &server->uuid, index, &alist->addrs[index].srx.transport,
> + _debug("probe %pU [%u] %pISpc rtt=%d ret=%d",
> + &server->uuid, index, rxrpc_kernel_remote_addr(alist->addrs[index].peer),
> rtt_us, ret);
>
> return afs_done_one_fs_probe(call->net, server);
> diff --git a/fs/afs/internal.h b/fs/afs/internal.h
> index ae874baee249..caf89edc0644 100644
> --- a/fs/afs/internal.h
> +++ b/fs/afs/internal.h
> @@ -72,6 +72,11 @@ enum afs_call_state {
> AFS_CALL_COMPLETE, /* Completed or failed */
> };
>
> +struct afs_address {
> + struct rxrpc_peer *peer;
> + u16 service_id;
> +};
> +
> /*
> * List of server addresses.
> */
> @@ -87,9 +92,7 @@ struct afs_addr_list {
> enum dns_lookup_status status:8;
> unsigned long failed; /* Mask of addrs that failed locally/ICMP */
> unsigned long responded; /* Mask of addrs that responded */
> - struct {
> - struct sockaddr_rxrpc srx;
> - } addrs[] __counted_by(max_addrs);
> + struct afs_address addrs[] __counted_by(max_addrs);
> #define AFS_MAX_ADDRESSES ((unsigned int)(sizeof(unsigned long) * 8))
> };
>
> @@ -420,7 +423,7 @@ struct afs_vlserver {
> atomic_t probe_outstanding;
> spinlock_t probe_lock;
> struct {
> - unsigned int rtt; /* RTT in uS */
> + unsigned int rtt; /* Best RTT in uS (or UINT_MAX) */
> u32 abort_code;
> short error;
> unsigned short flags;
> @@ -537,7 +540,7 @@ struct afs_server {
> atomic_t probe_outstanding;
> spinlock_t probe_lock;
> struct {
> - unsigned int rtt; /* RTT in uS */
> + unsigned int rtt; /* Best RTT in uS (or UINT_MAX) */
> u32 abort_code;
> short error;
> bool responded:1;
> @@ -963,9 +966,7 @@ static inline struct afs_addr_list *afs_get_addrlist(struct afs_addr_list *alist
> refcount_inc(&alist->usage);
> return alist;
> }
> -extern struct afs_addr_list *afs_alloc_addrlist(unsigned int,
> - unsigned short,
> - unsigned short);
> +extern struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id);
> extern void afs_put_addrlist(struct afs_addr_list *);
> extern struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *,
> const char *, size_t, char,
> @@ -976,8 +977,10 @@ extern struct afs_vlserver_list *afs_dns_query(struct afs_cell *, time64_t *);
> extern bool afs_iterate_addresses(struct afs_addr_cursor *);
> extern int afs_end_cursor(struct afs_addr_cursor *);
>
> -extern void afs_merge_fs_addr4(struct afs_addr_list *, __be32, u16);
> -extern void afs_merge_fs_addr6(struct afs_addr_list *, __be32 *, u16);
> +extern int afs_merge_fs_addr4(struct afs_net *net, struct afs_addr_list *addr,
> + __be32 xdr, u16 port);
> +extern int afs_merge_fs_addr6(struct afs_net *net, struct afs_addr_list *addr,
> + __be32 *xdr, u16 port);
>
> /*
> * callback.c
> @@ -1404,8 +1407,7 @@ extern void __exit afs_clean_up_permit_cache(void);
> */
> extern spinlock_t afs_server_peer_lock;
>
> -extern struct afs_server *afs_find_server(struct afs_net *,
> - const struct sockaddr_rxrpc *);
> +extern struct afs_server *afs_find_server(struct afs_net *, const struct rxrpc_peer *);
> extern struct afs_server *afs_find_server_by_uuid(struct afs_net *, const uuid_t *);
> extern struct afs_server *afs_lookup_server(struct afs_cell *, struct key *, const uuid_t *, u32);
> extern struct afs_server *afs_get_server(struct afs_server *, enum afs_server_trace);
> diff --git a/fs/afs/proc.c b/fs/afs/proc.c
> index ab9cd986cfd9..8a65a06908d2 100644
> --- a/fs/afs/proc.c
> +++ b/fs/afs/proc.c
> @@ -307,7 +307,7 @@ static int afs_proc_cell_vlservers_show(struct seq_file *m, void *v)
> for (i = 0; i < alist->nr_addrs; i++)
> seq_printf(m, " %c %pISpc\n",
> alist->preferred == i ? '>' : '-',
> - &alist->addrs[i].srx.transport);
> + rxrpc_kernel_remote_addr(alist->addrs[i].peer));
> }
> seq_printf(m, " info: fl=%lx rtt=%d\n", vlserver->flags, vlserver->rtt);
> seq_printf(m, " probe: fl=%x e=%d ac=%d out=%d\n",
> @@ -398,9 +398,10 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
> seq_printf(m, " - ALIST v=%u rsp=%lx f=%lx\n",
> alist->version, alist->responded, alist->failed);
> for (i = 0; i < alist->nr_addrs; i++)
> - seq_printf(m, " [%x] %pISpc%s\n",
> - i, &alist->addrs[i].srx.transport,
> - alist->preferred == i ? "*" : "");
> + seq_printf(m, " [%x] %pISpc%s rtt=%d\n",
> + i, rxrpc_kernel_remote_addr(alist->addrs[i].peer),
> + alist->preferred == i ? "*" : "",
> + rxrpc_kernel_get_srtt(alist->addrs[i].peer));
> return 0;
> }
>
> diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
> index 993e20d752d9..1c8f26a7f128 100644
> --- a/fs/afs/rotate.c
> +++ b/fs/afs/rotate.c
> @@ -113,7 +113,7 @@ bool afs_select_fileserver(struct afs_operation *op)
> struct afs_server *server;
> struct afs_vnode *vnode = op->file[0].vnode;
> struct afs_error e;
> - u32 rtt;
> + unsigned int rtt;
> int error = op->ac.error, i;
>
> _enter("%lx[%d],%lx[%d],%d,%d",
> @@ -419,7 +419,7 @@ bool afs_select_fileserver(struct afs_operation *op)
> }
>
> op->index = -1;
> - rtt = U32_MAX;
> + rtt = UINT_MAX;
> for (i = 0; i < op->server_list->nr_servers; i++) {
> struct afs_server *s = op->server_list->servers[i].server;
>
> @@ -487,7 +487,7 @@ bool afs_select_fileserver(struct afs_operation *op)
>
> _debug("address [%u] %u/%u %pISp",
> op->index, op->ac.index, op->ac.alist->nr_addrs,
> - &op->ac.alist->addrs[op->ac.index].srx.transport);
> + rxrpc_kernel_remote_addr(op->ac.alist->addrs[op->ac.index].peer));
>
> _leave(" = t");
> return true;
> diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
> index 2b1a31b4249c..1a18fcbdec80 100644
> --- a/fs/afs/rxrpc.c
> +++ b/fs/afs/rxrpc.c
> @@ -296,7 +296,8 @@ static void afs_notify_end_request_tx(struct sock *sock,
> */
> void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
> {
> - struct sockaddr_rxrpc *srx = &ac->alist->addrs[ac->index].srx;
> + struct afs_address *addr = &ac->alist->addrs[ac->index];
> + struct rxrpc_peer *peer = addr->peer;
> struct rxrpc_call *rxcall;
> struct msghdr msg;
> struct kvec iov[1];
> @@ -304,7 +305,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
> s64 tx_total_len;
> int ret;
>
> - _enter(",{%pISp},", &srx->transport);
> + _enter(",{%pISp},", rxrpc_kernel_remote_addr(addr->peer));
>
> ASSERT(call->type != NULL);
> ASSERT(call->type->name != NULL);
> @@ -333,7 +334,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
> }
>
> /* create a call */
> - rxcall = rxrpc_kernel_begin_call(call->net->socket, srx, call->key,
> + rxcall = rxrpc_kernel_begin_call(call->net->socket, peer, call->key,
> (unsigned long)call,
> tx_total_len,
> call->max_lifespan,
> @@ -341,6 +342,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
> (call->async ?
> afs_wake_up_async_call :
> afs_wake_up_call_waiter),
> + addr->service_id,
> call->upgrade,
> (call->intr ? RXRPC_PREINTERRUPTIBLE :
> RXRPC_UNINTERRUPTIBLE),
> @@ -461,7 +463,7 @@ static void afs_log_error(struct afs_call *call, s32 remote_abort)
> max = m + 1;
> pr_notice("kAFS: Peer reported %s failure on %s [%pISp]\n",
> msg, call->type->name,
> - &call->alist->addrs[call->addr_ix].srx.transport);
> + rxrpc_kernel_remote_addr(call->alist->addrs[call->addr_ix].peer));
> }
> }
>
> diff --git a/fs/afs/server.c b/fs/afs/server.c
> index 4b98788c7b12..831254d8ef9c 100644
> --- a/fs/afs/server.c
> +++ b/fs/afs/server.c
> @@ -21,13 +21,12 @@ static void __afs_put_server(struct afs_net *, struct afs_server *);
> /*
> * Find a server by one of its addresses.
> */
> -struct afs_server *afs_find_server(struct afs_net *net,
> - const struct sockaddr_rxrpc *srx)
> +struct afs_server *afs_find_server(struct afs_net *net, const struct rxrpc_peer *peer)
> {
> const struct afs_addr_list *alist;
> struct afs_server *server = NULL;
> unsigned int i;
> - int seq = 0, diff;
> + int seq = 0;
>
> rcu_read_lock();
>
> @@ -37,37 +36,11 @@ struct afs_server *afs_find_server(struct afs_net *net,
> server = NULL;
> read_seqbegin_or_lock(&net->fs_addr_lock, &seq);
>
> - if (srx->transport.family == AF_INET6) {
> - const struct sockaddr_in6 *a = &srx->transport.sin6, *b;
> - hlist_for_each_entry_rcu(server, &net->fs_addresses6, addr6_link) {
> - alist = rcu_dereference(server->addresses);
> - for (i = alist->nr_ipv4; i < alist->nr_addrs; i++) {
> - b = &alist->addrs[i].srx.transport.sin6;
> - diff = ((u16 __force)a->sin6_port -
> - (u16 __force)b->sin6_port);
> - if (diff == 0)
> - diff = memcmp(&a->sin6_addr,
> - &b->sin6_addr,
> - sizeof(struct in6_addr));
> - if (diff == 0)
> - goto found;
> - }
> - }
> - } else {
> - const struct sockaddr_in *a = &srx->transport.sin, *b;
> - hlist_for_each_entry_rcu(server, &net->fs_addresses4, addr4_link) {
> - alist = rcu_dereference(server->addresses);
> - for (i = 0; i < alist->nr_ipv4; i++) {
> - b = &alist->addrs[i].srx.transport.sin;
> - diff = ((u16 __force)a->sin_port -
> - (u16 __force)b->sin_port);
> - if (diff == 0)
> - diff = ((u32 __force)a->sin_addr.s_addr -
> - (u32 __force)b->sin_addr.s_addr);
> - if (diff == 0)
> - goto found;
> - }
> - }
> + hlist_for_each_entry_rcu(server, &net->fs_addresses6, addr6_link) {
> + alist = rcu_dereference(server->addresses);
> + for (i = 0; i < alist->nr_addrs; i++)
> + if (alist->addrs[i].peer == peer)
> + goto found;
> }
>
> server = NULL;
> diff --git a/fs/afs/vl_alias.c b/fs/afs/vl_alias.c
> index d3c0df70a1a5..6fdf9f1bedc0 100644
> --- a/fs/afs/vl_alias.c
> +++ b/fs/afs/vl_alias.c
> @@ -32,55 +32,6 @@ static struct afs_volume *afs_sample_volume(struct afs_cell *cell, struct key *k
> return volume;
> }
>
> -/*
> - * Compare two addresses.
> - */
> -static int afs_compare_addrs(const struct sockaddr_rxrpc *srx_a,
> - const struct sockaddr_rxrpc *srx_b)
> -{
> - short port_a, port_b;
> - int addr_a, addr_b, diff;
> -
> - diff = (short)srx_a->transport_type - (short)srx_b->transport_type;
> - if (diff)
> - goto out;
> -
> - switch (srx_a->transport_type) {
> - case AF_INET: {
> - const struct sockaddr_in *a = &srx_a->transport.sin;
> - const struct sockaddr_in *b = &srx_b->transport.sin;
> - addr_a = ntohl(a->sin_addr.s_addr);
> - addr_b = ntohl(b->sin_addr.s_addr);
> - diff = addr_a - addr_b;
> - if (diff == 0) {
> - port_a = ntohs(a->sin_port);
> - port_b = ntohs(b->sin_port);
> - diff = port_a - port_b;
> - }
> - break;
> - }
> -
> - case AF_INET6: {
> - const struct sockaddr_in6 *a = &srx_a->transport.sin6;
> - const struct sockaddr_in6 *b = &srx_b->transport.sin6;
> - diff = memcmp(&a->sin6_addr, &b->sin6_addr, 16);
> - if (diff == 0) {
> - port_a = ntohs(a->sin6_port);
> - port_b = ntohs(b->sin6_port);
> - diff = port_a - port_b;
> - }
> - break;
> - }
> -
> - default:
> - WARN_ON(1);
> - diff = 1;
> - }
> -
> -out:
> - return diff;
> -}
> -
> /*
> * Compare the address lists of a pair of fileservers.
> */
> @@ -94,9 +45,9 @@ static int afs_compare_fs_alists(const struct afs_server *server_a,
> lb = rcu_dereference(server_b->addresses);
>
> while (a < la->nr_addrs && b < lb->nr_addrs) {
> - const struct sockaddr_rxrpc *srx_a = &la->addrs[a].srx;
> - const struct sockaddr_rxrpc *srx_b = &lb->addrs[b].srx;
> - int diff = afs_compare_addrs(srx_a, srx_b);
> + unsigned long pa = (unsigned long)la->addrs[a].peer;
> + unsigned long pb = (unsigned long)lb->addrs[b].peer;
> + long diff = pa - pb;
>
> if (diff < 0) {
> a++;
> diff --git a/fs/afs/vl_list.c b/fs/afs/vl_list.c
> index acc48216136a..ba89140eee9e 100644
> --- a/fs/afs/vl_list.c
> +++ b/fs/afs/vl_list.c
> @@ -83,14 +83,15 @@ static u16 afs_extract_le16(const u8 **_b)
> /*
> * Build a VL server address list from a DNS queried server list.
> */
> -static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
> +static struct afs_addr_list *afs_extract_vl_addrs(struct afs_net *net,
> + const u8 **_b, const u8 *end,
> u8 nr_addrs, u16 port)
> {
> struct afs_addr_list *alist;
> const u8 *b = *_b;
> int ret = -EINVAL;
>
> - alist = afs_alloc_addrlist(nr_addrs, VL_SERVICE, port);
> + alist = afs_alloc_addrlist(nr_addrs, VL_SERVICE);
> if (!alist)
> return ERR_PTR(-ENOMEM);
> if (nr_addrs == 0)
> @@ -109,7 +110,9 @@ static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
> goto error;
> }
> memcpy(x, b, 4);
> - afs_merge_fs_addr4(alist, x[0], port);
> + ret = afs_merge_fs_addr4(net, alist, x[0], port);
> + if (ret < 0)
> + goto error;
> b += 4;
> break;
>
> @@ -119,7 +122,9 @@ static struct afs_addr_list *afs_extract_vl_addrs(const u8 **_b, const u8 *end,
> goto error;
> }
> memcpy(x, b, 16);
> - afs_merge_fs_addr6(alist, x, port);
> + ret = afs_merge_fs_addr6(net, alist, x, port);
> + if (ret < 0)
> + goto error;
> b += 16;
> break;
>
> @@ -247,7 +252,7 @@ struct afs_vlserver_list *afs_extract_vlserver_list(struct afs_cell *cell,
> /* Extract the addresses - note that we can't skip this as we
> * have to advance the payload pointer.
> */
> - addrs = afs_extract_vl_addrs(&b, end, bs.nr_addrs, bs.port);
> + addrs = afs_extract_vl_addrs(cell->net, &b, end, bs.nr_addrs, bs.port);
> if (IS_ERR(addrs)) {
> ret = PTR_ERR(addrs);
> goto error_2;
> diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
> index bdd9372e3fb2..9551aef07cee 100644
> --- a/fs/afs/vl_probe.c
> +++ b/fs/afs/vl_probe.c
> @@ -48,6 +48,7 @@ void afs_vlserver_probe_result(struct afs_call *call)
> {
> struct afs_addr_list *alist = call->alist;
> struct afs_vlserver *server = call->vlserver;
> + struct afs_address *addr = &alist->addrs[call->addr_ix];
> unsigned int server_index = call->server_index;
> unsigned int rtt_us = 0;
> unsigned int index = call->addr_ix;
> @@ -106,16 +107,16 @@ void afs_vlserver_probe_result(struct afs_call *call)
> if (call->service_id == YFS_VL_SERVICE) {
> server->probe.flags |= AFS_VLSERVER_PROBE_IS_YFS;
> set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
> - alist->addrs[index].srx.srx_service = call->service_id;
> + addr->service_id = call->service_id;
> } else {
> server->probe.flags |= AFS_VLSERVER_PROBE_NOT_YFS;
> if (!(server->probe.flags & AFS_VLSERVER_PROBE_IS_YFS)) {
> clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
> - alist->addrs[index].srx.srx_service = call->service_id;
> + addr->service_id = call->service_id;
> }
> }
>
> - rxrpc_kernel_get_srtt(call->net->socket, call->rxcall, &rtt_us);
> + rtt_us = rxrpc_kernel_get_srtt(addr->peer);
> if (rtt_us < server->probe.rtt) {
> server->probe.rtt = rtt_us;
> server->rtt = rtt_us;
> @@ -130,8 +131,9 @@ void afs_vlserver_probe_result(struct afs_call *call)
> out:
> spin_unlock(&server->probe_lock);
>
> - _debug("probe [%u][%u] %pISpc rtt=%u ret=%d",
> - server_index, index, &alist->addrs[index].srx.transport, rtt_us, ret);
> + _debug("probe [%u][%u] %pISpc rtt=%d ret=%d",
> + server_index, index, rxrpc_kernel_remote_addr(addr->peer),
> + rtt_us, ret);
>
> afs_done_one_vl_probe(server, have_result);
> }
> diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
> index 0d0b54819128..af445e7d3a12 100644
> --- a/fs/afs/vl_rotate.c
> +++ b/fs/afs/vl_rotate.c
> @@ -86,7 +86,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
> struct afs_addr_list *alist;
> struct afs_vlserver *vlserver;
> struct afs_error e;
> - u32 rtt;
> + unsigned int rtt;
> int error = vc->ac.error, i;
>
> _enter("%lx[%d],%lx[%d],%d,%d",
> @@ -188,7 +188,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
> goto selected_server;
>
> vc->index = -1;
> - rtt = U32_MAX;
> + rtt = UINT_MAX;
> for (i = 0; i < vc->server_list->nr_servers; i++) {
> struct afs_vlserver *s = vc->server_list->servers[i].server;
>
> @@ -243,7 +243,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
>
> _debug("VL address %d/%d", vc->ac.index, vc->ac.alist->nr_addrs);
>
> - _leave(" = t %pISpc", &vc->ac.alist->addrs[vc->ac.index].srx.transport);
> + _leave(" = t %pISpc", rxrpc_kernel_remote_addr(vc->ac.alist->addrs[vc->ac.index].peer));
> return true;
>
> next_server:
> diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
> index 00fca3c66ba6..41e7932d75c6 100644
> --- a/fs/afs/vlclient.c
> +++ b/fs/afs/vlclient.c
> @@ -208,7 +208,7 @@ static int afs_deliver_vl_get_addrs_u(struct afs_call *call)
> count = ntohl(*bp);
>
> nentries = min(nentries, count);
> - alist = afs_alloc_addrlist(nentries, FS_SERVICE, AFS_FS_PORT);
> + alist = afs_alloc_addrlist(nentries, FS_SERVICE);
> if (!alist)
> return -ENOMEM;
> alist->version = uniquifier;
> @@ -230,9 +230,13 @@ static int afs_deliver_vl_get_addrs_u(struct afs_call *call)
> alist = call->ret_alist;
> bp = call->buffer;
> count = min(call->count, 4U);
> - for (i = 0; i < count; i++)
> - if (alist->nr_addrs < call->count2)
> - afs_merge_fs_addr4(alist, *bp++, AFS_FS_PORT);
> + for (i = 0; i < count; i++) {
> + if (alist->nr_addrs < call->count2) {
> + ret = afs_merge_fs_addr4(call->net, alist, *bp++, AFS_FS_PORT);
> + if (ret < 0)
> + return ret;
> + }
> + }
>
> call->count -= count;
> if (call->count > 0)
> @@ -450,7 +454,7 @@ static int afs_deliver_yfsvl_get_endpoints(struct afs_call *call)
> if (call->count > YFS_MAXENDPOINTS)
> return afs_protocol_error(call, afs_eproto_yvl_fsendpt_num);
>
> - alist = afs_alloc_addrlist(call->count, FS_SERVICE, AFS_FS_PORT);
> + alist = afs_alloc_addrlist(call->count, FS_SERVICE);
> if (!alist)
> return -ENOMEM;
> alist->version = uniquifier;
> @@ -488,14 +492,18 @@ static int afs_deliver_yfsvl_get_endpoints(struct afs_call *call)
> if (ntohl(bp[0]) != sizeof(__be32) * 2)
> return afs_protocol_error(
> call, afs_eproto_yvl_fsendpt4_len);
> - afs_merge_fs_addr4(alist, bp[1], ntohl(bp[2]));
> + ret = afs_merge_fs_addr4(call->net, alist, bp[1], ntohl(bp[2]));
> + if (ret < 0)
> + return ret;
> bp += 3;
> break;
> case YFS_ENDPOINT_IPV6:
> if (ntohl(bp[0]) != sizeof(__be32) * 5)
> return afs_protocol_error(
> call, afs_eproto_yvl_fsendpt6_len);
> - afs_merge_fs_addr6(alist, bp + 1, ntohl(bp[5]));
> + ret = afs_merge_fs_addr6(call->net, alist, bp + 1, ntohl(bp[5]));
> + if (ret < 0)
> + return ret;
> bp += 6;
> break;
> default:
> diff --git a/include/net/af_rxrpc.h b/include/net/af_rxrpc.h
> index 5531dd08061e..0754c463224a 100644
> --- a/include/net/af_rxrpc.h
> +++ b/include/net/af_rxrpc.h
> @@ -15,6 +15,7 @@ struct key;
> struct sock;
> struct socket;
> struct rxrpc_call;
> +struct rxrpc_peer;
> enum rxrpc_abort_reason;
>
> enum rxrpc_interruptibility {
> @@ -41,13 +42,14 @@ void rxrpc_kernel_new_call_notification(struct socket *,
> rxrpc_notify_new_call_t,
> rxrpc_discard_new_call_t);
> struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
> - struct sockaddr_rxrpc *srx,
> + struct rxrpc_peer *peer,
> struct key *key,
> unsigned long user_call_ID,
> s64 tx_total_len,
> u32 hard_timeout,
> gfp_t gfp,
> rxrpc_notify_rx_t notify_rx,
> + u16 service_id,
> bool upgrade,
> enum rxrpc_interruptibility interruptibility,
> unsigned int debug_id);
> @@ -60,9 +62,14 @@ bool rxrpc_kernel_abort_call(struct socket *, struct rxrpc_call *,
> u32, int, enum rxrpc_abort_reason);
> void rxrpc_kernel_shutdown_call(struct socket *sock, struct rxrpc_call *call);
> void rxrpc_kernel_put_call(struct socket *sock, struct rxrpc_call *call);
> -void rxrpc_kernel_get_peer(struct socket *, struct rxrpc_call *,
> - struct sockaddr_rxrpc *);
> -bool rxrpc_kernel_get_srtt(struct socket *, struct rxrpc_call *, u32 *);
> +struct rxrpc_peer *rxrpc_kernel_lookup_peer(struct socket *sock,
> + struct sockaddr_rxrpc *srx, gfp_t gfp);
> +void rxrpc_kernel_put_peer(struct rxrpc_peer *peer);
> +struct rxrpc_peer *rxrpc_kernel_get_peer(struct rxrpc_peer *peer);
> +struct rxrpc_peer *rxrpc_kernel_get_call_peer(struct socket *sock, struct rxrpc_call *call);
> +const struct sockaddr_rxrpc *rxrpc_kernel_remote_srx(const struct rxrpc_peer *peer);
> +const struct sockaddr *rxrpc_kernel_remote_addr(const struct rxrpc_peer *peer);
> +unsigned int rxrpc_kernel_get_srtt(const struct rxrpc_peer *);
> int rxrpc_kernel_charge_accept(struct socket *, rxrpc_notify_rx_t,
> rxrpc_user_attach_call_t, unsigned long, gfp_t,
> unsigned int);
> diff --git a/include/trace/events/rxrpc.h b/include/trace/events/rxrpc.h
> index 4c53a5ef6257..90a1e39d620e 100644
> --- a/include/trace/events/rxrpc.h
> +++ b/include/trace/events/rxrpc.h
> @@ -178,7 +178,9 @@
> #define rxrpc_peer_traces \
> EM(rxrpc_peer_free, "FREE ") \
> EM(rxrpc_peer_get_accept, "GET accept ") \
> + EM(rxrpc_peer_get_application, "GET app ") \
> EM(rxrpc_peer_get_bundle, "GET bundle ") \
> + EM(rxrpc_peer_get_call, "GET call ") \
> EM(rxrpc_peer_get_client_conn, "GET cln-conn") \
> EM(rxrpc_peer_get_input, "GET input ") \
> EM(rxrpc_peer_get_input_error, "GET inpt-err") \
> @@ -187,6 +189,7 @@
> EM(rxrpc_peer_get_service_conn, "GET srv-conn") \
> EM(rxrpc_peer_new_client, "NEW client ") \
> EM(rxrpc_peer_new_prealloc, "NEW prealloc") \
> + EM(rxrpc_peer_put_application, "PUT app ") \
> EM(rxrpc_peer_put_bundle, "PUT bundle ") \
> EM(rxrpc_peer_put_call, "PUT call ") \
> EM(rxrpc_peer_put_conn, "PUT conn ") \
> diff --git a/net/rxrpc/af_rxrpc.c b/net/rxrpc/af_rxrpc.c
> index fa8aec78f63d..465bfe5eb061 100644
> --- a/net/rxrpc/af_rxrpc.c
> +++ b/net/rxrpc/af_rxrpc.c
> @@ -258,16 +258,62 @@ static int rxrpc_listen(struct socket *sock, int backlog)
> return ret;
> }
>
> +/**
> + * rxrpc_kernel_lookup_peer - Obtain remote transport endpoint for an address
> + * @sock: The socket through which it will be accessed
> + * @srx: The network address
> + * @gfp: Allocation flags
> + *
> + * Lookup or create a remote transport endpoint record for the specified
> + * address and return it with a ref held.
> + */
> +struct rxrpc_peer *rxrpc_kernel_lookup_peer(struct socket *sock,
> + struct sockaddr_rxrpc *srx, gfp_t gfp)
> +{
> + struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
> + int ret;
> +
> + ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
> + if (ret < 0)
> + return ERR_PTR(ret);
> +
> + return rxrpc_lookup_peer(rx->local, srx, gfp);
> +}
> +EXPORT_SYMBOL(rxrpc_kernel_lookup_peer);
> +
> +/**
> + * rxrpc_kernel_get_peer - Get a reference on a peer
> + * @peer: The peer to get a reference on.
> + *
> + * Get a record for the remote peer in a call.
> + */
> +struct rxrpc_peer *rxrpc_kernel_get_peer(struct rxrpc_peer *peer)
> +{
> + return peer ? rxrpc_get_peer(peer, rxrpc_peer_get_application) : NULL;
> +}
> +EXPORT_SYMBOL(rxrpc_kernel_get_peer);
> +
> +/**
> + * rxrpc_kernel_put_peer - Allow a kernel app to drop a peer reference
> + * @peer: The peer to drop a ref on
> + */
> +void rxrpc_kernel_put_peer(struct rxrpc_peer *peer)
> +{
> + rxrpc_put_peer(peer, rxrpc_peer_put_application);
> +}
> +EXPORT_SYMBOL(rxrpc_kernel_put_peer);
> +
> /**
> * rxrpc_kernel_begin_call - Allow a kernel service to begin a call
> * @sock: The socket on which to make the call
> - * @srx: The address of the peer to contact
> + * @peer: The peer to contact
> * @key: The security context to use (defaults to socket setting)
> * @user_call_ID: The ID to use
> * @tx_total_len: Total length of data to transmit during the call (or -1)
> * @hard_timeout: The maximum lifespan of the call in sec
> * @gfp: The allocation constraints
> * @notify_rx: Where to send notifications instead of socket queue
> + * @service_id: The ID of the service to contact
> * @upgrade: Request service upgrade for call
> * @interruptibility: The call is interruptible, or can be canceled.
> * @debug_id: The debug ID for tracing to be assigned to the call
> @@ -280,13 +326,14 @@ static int rxrpc_listen(struct socket *sock, int backlog)
> * supplying @srx and @key.
> */
> struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
> - struct sockaddr_rxrpc *srx,
> + struct rxrpc_peer *peer,
> struct key *key,
> unsigned long user_call_ID,
> s64 tx_total_len,
> u32 hard_timeout,
> gfp_t gfp,
> rxrpc_notify_rx_t notify_rx,
> + u16 service_id,
> bool upgrade,
> enum rxrpc_interruptibility interruptibility,
> unsigned int debug_id)
> @@ -295,13 +342,11 @@ struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
> struct rxrpc_call_params p;
> struct rxrpc_call *call;
> struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
> - int ret;
>
> _enter(",,%x,%lx", key_serial(key), user_call_ID);
>
> - ret = rxrpc_validate_address(rx, srx, sizeof(*srx));
> - if (ret < 0)
> - return ERR_PTR(ret);
> + if (WARN_ON_ONCE(peer->local != rx->local))
> + return ERR_PTR(-EIO);
>
> lock_sock(&rx->sk);
>
> @@ -319,12 +364,13 @@ struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *sock,
>
> memset(&cp, 0, sizeof(cp));
> cp.local = rx->local;
> + cp.peer = peer;
> cp.key = key;
> cp.security_level = rx->min_sec_level;
> cp.exclusive = false;
> cp.upgrade = upgrade;
> - cp.service_id = srx->srx_service;
> - call = rxrpc_new_client_call(rx, &cp, srx, &p, gfp, debug_id);
> + cp.service_id = service_id;
> + call = rxrpc_new_client_call(rx, &cp, &p, gfp, debug_id);
> /* The socket has been unlocked. */
> if (!IS_ERR(call)) {
> call->notify_rx = notify_rx;
> diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
> index e8e14c6f904d..8eea7a487380 100644
> --- a/net/rxrpc/ar-internal.h
> +++ b/net/rxrpc/ar-internal.h
> @@ -364,6 +364,7 @@ struct rxrpc_conn_proto {
>
> struct rxrpc_conn_parameters {
> struct rxrpc_local *local; /* Representation of local endpoint */
> + struct rxrpc_peer *peer; /* Representation of remote endpoint */
> struct key *key; /* Security details */
> bool exclusive; /* T if conn is exclusive */
> bool upgrade; /* T if service ID can be upgraded */
> @@ -867,7 +868,6 @@ struct rxrpc_call *rxrpc_find_call_by_user_ID(struct rxrpc_sock *, unsigned long
> struct rxrpc_call *rxrpc_alloc_call(struct rxrpc_sock *, gfp_t, unsigned int);
> struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *,
> struct rxrpc_conn_parameters *,
> - struct sockaddr_rxrpc *,
> struct rxrpc_call_params *, gfp_t,
> unsigned int);
> void rxrpc_start_call_timer(struct rxrpc_call *call);
> diff --git a/net/rxrpc/call_object.c b/net/rxrpc/call_object.c
> index 773eecd1e979..beea25ac88f5 100644
> --- a/net/rxrpc/call_object.c
> +++ b/net/rxrpc/call_object.c
> @@ -193,7 +193,6 @@ struct rxrpc_call *rxrpc_alloc_call(struct rxrpc_sock *rx, gfp_t gfp,
> * Allocate a new client call.
> */
> static struct rxrpc_call *rxrpc_alloc_client_call(struct rxrpc_sock *rx,
> - struct sockaddr_rxrpc *srx,
> struct rxrpc_conn_parameters *cp,
> struct rxrpc_call_params *p,
> gfp_t gfp,
> @@ -211,10 +210,12 @@ static struct rxrpc_call *rxrpc_alloc_client_call(struct rxrpc_sock *rx,
> now = ktime_get_real();
> call->acks_latest_ts = now;
> call->cong_tstamp = now;
> - call->dest_srx = *srx;
> + call->dest_srx = cp->peer->srx;
> + call->dest_srx.srx_service = cp->service_id;
> call->interruptibility = p->interruptibility;
> call->tx_total_len = p->tx_total_len;
> call->key = key_get(cp->key);
> + call->peer = rxrpc_get_peer(cp->peer, rxrpc_peer_get_call);
> call->local = rxrpc_get_local(cp->local, rxrpc_local_get_call);
> call->security_level = cp->security_level;
> if (p->kernel)
> @@ -306,10 +307,6 @@ static int rxrpc_connect_call(struct rxrpc_call *call, gfp_t gfp)
>
> _enter("{%d,%lx},", call->debug_id, call->user_call_ID);
>
> - call->peer = rxrpc_lookup_peer(local, &call->dest_srx, gfp);
> - if (!call->peer)
> - goto error;
> -
> ret = rxrpc_look_up_bundle(call, gfp);
> if (ret < 0)
> goto error;
> @@ -334,7 +331,6 @@ static int rxrpc_connect_call(struct rxrpc_call *call, gfp_t gfp)
> */
> struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *rx,
> struct rxrpc_conn_parameters *cp,
> - struct sockaddr_rxrpc *srx,
> struct rxrpc_call_params *p,
> gfp_t gfp,
> unsigned int debug_id)
> @@ -349,13 +345,18 @@ struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *rx,
>
> _enter("%p,%lx", rx, p->user_call_ID);
>
> + if (WARN_ON_ONCE(!cp->peer)) {
> + release_sock(&rx->sk);
> + return ERR_PTR(-EIO);
> + }
> +
> limiter = rxrpc_get_call_slot(p, gfp);
> if (!limiter) {
> release_sock(&rx->sk);
> return ERR_PTR(-ERESTARTSYS);
> }
>
> - call = rxrpc_alloc_client_call(rx, srx, cp, p, gfp, debug_id);
> + call = rxrpc_alloc_client_call(rx, cp, p, gfp, debug_id);
> if (IS_ERR(call)) {
> release_sock(&rx->sk);
> up(limiter);
> diff --git a/net/rxrpc/peer_object.c b/net/rxrpc/peer_object.c
> index 8d7a715a0bb1..65ea57b427a1 100644
> --- a/net/rxrpc/peer_object.c
> +++ b/net/rxrpc/peer_object.c
> @@ -22,6 +22,8 @@
> #include <net/ip6_route.h>
> #include "ar-internal.h"
>
> +static const struct sockaddr_rxrpc rxrpc_null_addr;
> +
> /*
> * Hash a peer key.
> */
> @@ -457,39 +459,51 @@ void rxrpc_destroy_all_peers(struct rxrpc_net *rxnet)
> }
>
> /**
> - * rxrpc_kernel_get_peer - Get the peer address of a call
> + * rxrpc_kernel_get_call_peer - Get the peer address of a call
> * @sock: The socket on which the call is in progress.
> * @call: The call to query
> - * @_srx: Where to place the result
> *
> - * Get the address of the remote peer in a call.
> + * Get a record for the remote peer in a call.
> */
> -void rxrpc_kernel_get_peer(struct socket *sock, struct rxrpc_call *call,
> - struct sockaddr_rxrpc *_srx)
> +struct rxrpc_peer *rxrpc_kernel_get_call_peer(struct socket *sock, struct rxrpc_call *call)
> {
> - *_srx = call->peer->srx;
> + return call->peer;
> }
> -EXPORT_SYMBOL(rxrpc_kernel_get_peer);
> +EXPORT_SYMBOL(rxrpc_kernel_get_call_peer);
>
> /**
> * rxrpc_kernel_get_srtt - Get a call's peer smoothed RTT
> - * @sock: The socket on which the call is in progress.
> - * @call: The call to query
> - * @_srtt: Where to store the SRTT value.
> + * @peer: The peer to query
> *
> - * Get the call's peer smoothed RTT in uS.
> + * Get the call's peer smoothed RTT in uS or UINT_MAX if we have no samples.
> */
> -bool rxrpc_kernel_get_srtt(struct socket *sock, struct rxrpc_call *call,
> - u32 *_srtt)
> +unsigned int rxrpc_kernel_get_srtt(const struct rxrpc_peer *peer)
> {
> - struct rxrpc_peer *peer = call->peer;
> + return peer->rtt_count > 0 ? peer->srtt_us >> 3 : UINT_MAX;
> +}
> +EXPORT_SYMBOL(rxrpc_kernel_get_srtt);
>
> - if (peer->rtt_count == 0) {
> - *_srtt = 1000000; /* 1S */
> - return false;
> - }
> +/**
> + * rxrpc_kernel_remote_srx - Get the address of a peer
> + * @peer: The peer to query
> + *
> + * Get a pointer to the address from a peer record. The caller is responsible
> + * for making sure that the address is not deallocated.
> + */
> +const struct sockaddr_rxrpc *rxrpc_kernel_remote_srx(const struct rxrpc_peer *peer)
> +{
> + return peer ? &peer->srx : &rxrpc_null_addr;
> +}
>
> - *_srtt = call->peer->srtt_us >> 3;
> - return true;
> +/**
> + * rxrpc_kernel_remote_addr - Get the peer transport address of a call
> + * @peer: The peer to query
> + *
> + * Get a pointer to the transport address from a peer record. The caller is
> + * responsible for making sure that the address is not deallocated.
> + */
> +const struct sockaddr *rxrpc_kernel_remote_addr(const struct rxrpc_peer *peer)
> +{
> + return (const struct sockaddr *)
> + (peer ? &peer->srx.transport : &rxrpc_null_addr.transport);
> }
> -EXPORT_SYMBOL(rxrpc_kernel_get_srtt);
> diff --git a/net/rxrpc/sendmsg.c b/net/rxrpc/sendmsg.c
> index 8e0b94714e84..5677d5690a02 100644
> --- a/net/rxrpc/sendmsg.c
> +++ b/net/rxrpc/sendmsg.c
> @@ -572,6 +572,7 @@ rxrpc_new_client_call_for_sendmsg(struct rxrpc_sock *rx, struct msghdr *msg,
> __acquires(&call->user_mutex)
> {
> struct rxrpc_conn_parameters cp;
> + struct rxrpc_peer *peer;
> struct rxrpc_call *call;
> struct key *key;
>
> @@ -584,21 +585,29 @@ rxrpc_new_client_call_for_sendmsg(struct rxrpc_sock *rx, struct msghdr *msg,
> return ERR_PTR(-EDESTADDRREQ);
> }
>
> + peer = rxrpc_lookup_peer(rx->local, srx, GFP_KERNEL);
> + if (!peer) {
> + release_sock(&rx->sk);
> + return ERR_PTR(-ENOMEM);
> + }
> +
> key = rx->key;
> if (key && !rx->key->payload.data[0])
> key = NULL;
>
> memset(&cp, 0, sizeof(cp));
> cp.local = rx->local;
> + cp.peer = peer;
> cp.key = rx->key;
> cp.security_level = rx->min_sec_level;
> cp.exclusive = rx->exclusive | p->exclusive;
> cp.upgrade = p->upgrade;
> cp.service_id = srx->srx_service;
> - call = rxrpc_new_client_call(rx, &cp, srx, &p->call, GFP_KERNEL,
> + call = rxrpc_new_client_call(rx, &cp, &p->call, GFP_KERNEL,
> atomic_inc_return(&rxrpc_debug_id));
> /* The socket is now unlocked */
>
> + rxrpc_put_peer(peer, rxrpc_peer_put_application);
> _leave(" = %p\n", call);
> return call;
> }
rxrpc_kernel_remote_srx and rxrpc_kernel_remote_addr need an
EXPORT_SYMBOL to be available to fs/afs when built as a module.
Marc
^ permalink raw reply [flat|nested] 57+ messages in thread
* [PATCH 11/41] afs: Don't skip server addresses for which we didn't get an RTT reading
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (9 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 10/41] rxrpc, afs: Allow afs to pin rxrpc_peer objects David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 12/41] afs: Rename addr_list::failed to probe_failed David Howells
` (29 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
In the rotation algorithms for iterating over volume location servers and
file servers, don't skip servers from which we got a valid response to a
probe (either a reply DATA packet or an ABORT) even if we didn't manage to
get an RTT reading.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/rotate.c | 2 +-
fs/afs/vl_rotate.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 1c8f26a7f128..689acb0ad64b 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -426,7 +426,7 @@ bool afs_select_fileserver(struct afs_operation *op)
if (!test_bit(i, &op->untried) ||
!test_bit(AFS_SERVER_FL_RESPONDING, &s->flags))
continue;
- if (s->probe.rtt < rtt) {
+ if (s->probe.rtt <= rtt) {
op->index = i;
rtt = s->probe.rtt;
}
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index af445e7d3a12..6e29272ffa8e 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -195,7 +195,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
if (!test_bit(i, &vc->untried) ||
!test_bit(AFS_VLSERVER_FL_RESPONDING, &s->flags))
continue;
- if (s->probe.rtt < rtt) {
+ if (s->probe.rtt <= rtt) {
vc->index = i;
rtt = s->probe.rtt;
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 12/41] afs: Rename addr_list::failed to probe_failed
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (10 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 11/41] afs: Don't skip server addresses for which we didn't get an RTT reading David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 13/41] afs: Handle the VIO abort explicitly David Howells
` (28 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Rename the failed member of struct addr_list to probe_failed as it's
specifically related to probe failures.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 2 +-
fs/afs/fs_probe.c | 6 +++---
fs/afs/internal.h | 2 +-
fs/afs/proc.c | 2 +-
fs/afs/rotate.c | 2 +-
fs/afs/vl_probe.c | 4 ++--
fs/afs/vl_rotate.c | 2 +-
7 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index 519821f5aedc..b76abf500713 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -367,7 +367,7 @@ bool afs_iterate_addresses(struct afs_addr_cursor *ac)
return false;
set = ac->alist->responded;
- failed = ac->alist->failed;
+ failed = ac->alist->probe_failed;
_enter("%lx-%lx-%lx,%d", set, failed, ac->tried, ac->index);
ac->nr_iterations++;
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index 58d28b82571e..fbb91ad775b9 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -88,7 +88,7 @@ static void afs_fs_probe_not_done(struct afs_net *net,
if (server->probe.error == 0)
server->probe.error = -ENOMEM;
- set_bit(index, &alist->failed);
+ set_bit(index, &alist->probe_failed);
spin_unlock(&server->probe_lock);
return afs_done_one_fs_probe(net, server);
@@ -138,7 +138,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
case -ETIME:
default:
clear_bit(index, &alist->responded);
- set_bit(index, &alist->failed);
+ set_bit(index, &alist->probe_failed);
if (!server->probe.responded &&
(server->probe.error == 0 ||
server->probe.error == -ETIMEDOUT ||
@@ -149,7 +149,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
}
responded:
- clear_bit(index, &alist->failed);
+ clear_bit(index, &alist->probe_failed);
if (call->service_id == YFS_FS_SERVICE) {
server->probe.is_yfs = true;
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index caf89edc0644..9182bd410bbd 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -90,7 +90,7 @@ struct afs_addr_list {
unsigned char nr_ipv4; /* Number of IPv4 addresses */
enum dns_record_source source:8;
enum dns_lookup_status status:8;
- unsigned long failed; /* Mask of addrs that failed locally/ICMP */
+ unsigned long probe_failed; /* Mask of addrs that failed locally/ICMP */
unsigned long responded; /* Mask of addrs that responded */
struct afs_address addrs[] __counted_by(max_addrs);
#define AFS_MAX_ADDRESSES ((unsigned int)(sizeof(unsigned long) * 8))
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index 8a65a06908d2..16d93fa6396f 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -396,7 +396,7 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
(int)(jiffies - server->probed_at) / HZ,
atomic_read(&server->probe_outstanding));
seq_printf(m, " - ALIST v=%u rsp=%lx f=%lx\n",
- alist->version, alist->responded, alist->failed);
+ alist->version, alist->responded, alist->probe_failed);
for (i = 0; i < alist->nr_addrs; i++)
seq_printf(m, " [%x] %pISpc%s rtt=%d\n",
i, rxrpc_kernel_remote_addr(alist->addrs[i].peer),
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 689acb0ad64b..0f59f2a81f23 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -582,7 +582,7 @@ void afs_dump_edestaddrreq(const struct afs_operation *op)
a->nr_ipv4, a->nr_addrs, a->max_addrs,
a->preferred);
pr_notice("FC: - R=%lx F=%lx\n",
- a->responded, a->failed);
+ a->responded, a->probe_failed);
if (a == op->ac.alist)
pr_notice("FC: - current\n");
}
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index 9551aef07cee..44bff3a2a5ac 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -90,7 +90,7 @@ void afs_vlserver_probe_result(struct afs_call *call)
case -ETIME:
default:
clear_bit(index, &alist->responded);
- set_bit(index, &alist->failed);
+ set_bit(index, &alist->probe_failed);
if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED) &&
(server->probe.error == 0 ||
server->probe.error == -ETIMEDOUT ||
@@ -102,7 +102,7 @@ void afs_vlserver_probe_result(struct afs_call *call)
responded:
set_bit(index, &alist->responded);
- clear_bit(index, &alist->failed);
+ clear_bit(index, &alist->probe_failed);
if (call->service_id == YFS_VL_SERVICE) {
server->probe.flags |= AFS_VLSERVER_PROBE_IS_YFS;
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index 6e29272ffa8e..af5c6cd1ed44 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -312,7 +312,7 @@ static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
a->nr_ipv4, a->nr_addrs, a->max_addrs,
a->preferred);
pr_notice("VC: - R=%lx F=%lx\n",
- a->responded, a->failed);
+ a->responded, a->probe_failed);
if (a == vc->ac.alist)
pr_notice("VC: - current\n");
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 13/41] afs: Handle the VIO abort explicitly
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (11 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 12/41] afs: Rename addr_list::failed to probe_failed David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 18:12 ` Jeffrey E Altman
2023-11-09 15:39 ` [PATCH 14/41] afs: Use op->nr_iterations=-1 to indicate to begin fileserver iteration David Howells
` (27 subsequent siblings)
40 siblings, 1 reply; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
When processing the result of a call, handle the VIO abort specifically
rather than leaving it to a default case. Rather than erroring out
unconditionally, see if there's another server if the volume has more than
one server available, otherwise return -EREMOTEIO.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/rotate.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 0f59f2a81f23..342afc951fe4 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -329,6 +329,12 @@ bool afs_select_fileserver(struct afs_operation *op)
goto restart_from_beginning;
+ case VIO:
+ op->error = -EREMOTEIO;
+ if (op->volume->type != AFSVL_RWVOL)
+ goto next_server;
+ goto failed;
+
case VDISKFULL:
case UAENOSPC:
/* The partition is full. Only applies to RWVOLs.
^ permalink raw reply related [flat|nested] 57+ messages in thread* Re: [PATCH 13/41] afs: Handle the VIO abort explicitly
2023-11-09 15:39 ` [PATCH 13/41] afs: Handle the VIO abort explicitly David Howells
@ 2023-11-09 18:12 ` Jeffrey E Altman
0 siblings, 0 replies; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-09 18:12 UTC (permalink / raw)
To: David Howells, Marc Dionne; +Cc: linux-afs, linux-fsdevel, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 970 bytes --]
On 11/9/2023 10:39 AM, David Howells wrote:
> When processing the result of a call, handle the VIO abort specifically
> rather than leaving it to a default case. Rather than erroring out
> unconditionally, see if there's another server if the volume has more than
> one server available, otherwise return -EREMOTEIO.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Marc Dionne <marc.dionne@auristor.com>
> cc: linux-afs@lists.infradead.org
> ---
> fs/afs/rotate.c | 6 ++++++
> 1 file changed, 6 insertions(+)
OpenAFS fileservers can return VIO (112) either during an attempt to
load a vnode or to store a vnode. However, most IBM AFS derived cache
managers do not explicitly handle VIO errors and pass them to the vfs to
be interpreted as a local operating system error. For Linux that means
EHOSTDOWN. Therefore, AuriStorFS fileservers return UAEIO instead.
Please modify this patch to handle UAEIO the same as VIO.
Thank you.
Jeffrey Altman
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread
* [PATCH 14/41] afs: Use op->nr_iterations=-1 to indicate to begin fileserver iteration
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (12 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 13/41] afs: Handle the VIO abort explicitly David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 15/41] afs: Return ENOENT if no cell DNS record can be found David Howells
` (26 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Set op->nr_iterations to -1 to indicate that we need to begin fileserver
iteration rather than setting error to SHRT_MAX. This makes it easier to
eliminate the address cursor.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/fs_operation.c | 2 +-
fs/afs/internal.h | 2 +-
fs/afs/rotate.c | 11 ++++++-----
3 files changed, 8 insertions(+), 7 deletions(-)
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index 7a3803ce3a22..3e31fae9a149 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -41,7 +41,7 @@ struct afs_operation *afs_alloc_operation(struct key *key, struct afs_volume *vo
op->cb_v_break = volume->cb_v_break;
op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
op->error = -EDESTADDRREQ;
- op->ac.error = SHRT_MAX;
+ op->nr_iterations = -1;
_leave(" = [op=%08x]", op->debug_id);
return op;
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 9182bd410bbd..3739110fc0b5 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -858,7 +858,7 @@ struct afs_operation {
struct afs_call *call;
unsigned long untried; /* Bitmask of untried servers */
short index; /* Current server */
- unsigned short nr_iterations; /* Number of server iterations */
+ short nr_iterations; /* Number of server iterations */
unsigned int flags;
#define AFS_OPERATION_STOP 0x0001 /* Set to cease iteration */
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 342afc951fe4..beb9fd4e8f44 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -116,7 +116,10 @@ bool afs_select_fileserver(struct afs_operation *op)
unsigned int rtt;
int error = op->ac.error, i;
- _enter("%lx[%d],%lx[%d],%d,%d",
+ op->nr_iterations++;
+
+ _enter("OP=%x+%x,%llx,%lx[%d],%lx[%d],%d,%d",
+ op->debug_id, op->nr_iterations, op->volume->vid,
op->untried, op->index,
op->ac.tried, op->ac.index,
error, op->ac.abort_code);
@@ -126,13 +129,11 @@ bool afs_select_fileserver(struct afs_operation *op)
return false;
}
- op->nr_iterations++;
+ if (op->nr_iterations == 0)
+ goto start;
/* Evaluate the result of the previous operation, if there was one. */
switch (error) {
- case SHRT_MAX:
- goto start;
-
case 0:
default:
/* Success or local failure. Stop. */
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 15/41] afs: Return ENOENT if no cell DNS record can be found
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (13 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 14/41] afs: Use op->nr_iterations=-1 to indicate to begin fileserver iteration David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 16/41] afs: Wrap most op->error accesses with inline funcs David Howells
` (25 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne
Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel,
Markus Suvanto
Make AFS return error ENOENT if no cell SRV or AFSDB DNS record (or
cellservdb config file record) can be found rather than returning
EDESTADDRREQ.
Also add cell name lookup info to the cursor dump.
Reported-by: Markus Suvanto <markus.suvanto@gmail.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216637
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/vl_rotate.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index af5c6cd1ed44..01c890b3d18d 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -58,6 +58,12 @@ static bool afs_start_vl_iteration(struct afs_vl_cursor *vc)
}
/* Status load is ordered after lookup counter load */
+ if (cell->dns_status == DNS_LOOKUP_GOT_NOT_FOUND) {
+ pr_warn("No record of cell %s\n", cell->name);
+ vc->error = -ENOENT;
+ return false;
+ }
+
if (cell->dns_source == DNS_RECORD_UNAVAILABLE) {
vc->error = -EDESTADDRREQ;
return false;
@@ -285,6 +291,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
*/
static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
{
+ struct afs_cell *cell = vc->cell;
static int count;
int i;
@@ -294,6 +301,9 @@ static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
rcu_read_lock();
pr_notice("EDESTADDR occurred\n");
+ pr_notice("CELL: %s err=%d\n", cell->name, cell->error);
+ pr_notice("DNS: src=%u st=%u lc=%x\n",
+ cell->dns_source, cell->dns_status, cell->dns_lookup_count);
pr_notice("VC: ut=%lx ix=%u ni=%hu fl=%hx err=%hd\n",
vc->untried, vc->index, vc->nr_iterations, vc->flags, vc->error);
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 16/41] afs: Wrap most op->error accesses with inline funcs
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (14 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 15/41] afs: Return ENOENT if no cell DNS record can be found David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 17/41] afs: Don't put afs_call in afs_wait_for_call_to_complete() David Howells
` (24 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Wrap most op->error accesses with inline funcs which will make it easier
for a subsequent patch to replace op->error with something else. Two
functions are added to this end:
(1) afs_op_error() - Get the error code.
(2) afs_op_set_error() - Set the error code.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/dir.c | 38 +++++++++++++++---------------
fs/afs/file.c | 4 ++--
fs/afs/fs_operation.c | 21 ++++++++++-------
fs/afs/fsclient.c | 2 +-
fs/afs/inode.c | 2 +-
fs/afs/internal.h | 20 ++++++++++++----
fs/afs/rotate.c | 55 ++++++++++++++++++++++++-------------------
fs/afs/server.c | 6 ++---
fs/afs/write.c | 6 ++---
9 files changed, 87 insertions(+), 67 deletions(-)
diff --git a/fs/afs/dir.c b/fs/afs/dir.c
index 5219182e52e1..b40f7ae850a8 100644
--- a/fs/afs/dir.c
+++ b/fs/afs/dir.c
@@ -878,14 +878,14 @@ static struct inode *afs_do_lookup(struct inode *dir, struct dentry *dentry,
* lookups contained therein are stored in the reply without aborting
* the whole operation.
*/
- op->error = -ENOTSUPP;
+ afs_op_set_error(op, -ENOTSUPP);
if (!cookie->one_only) {
op->ops = &afs_inline_bulk_status_operation;
afs_begin_vnode_operation(op);
afs_wait_for_operation(op);
}
- if (op->error == -ENOTSUPP) {
+ if (afs_op_error(op) == -ENOTSUPP) {
/* We could try FS.BulkStatus next, but this aborts the entire
* op if any of the lookups fails - so, for the moment, revert
* to FS.FetchStatus for op->file[1].
@@ -895,10 +895,10 @@ static struct inode *afs_do_lookup(struct inode *dir, struct dentry *dentry,
afs_begin_vnode_operation(op);
afs_wait_for_operation(op);
}
- inode = ERR_PTR(op->error);
+ inode = ERR_PTR(afs_op_error(op));
out_op:
- if (op->error == 0) {
+ if (!afs_op_error(op)) {
inode = &op->file[1].vnode->netfs.inode;
op->file[1].vnode = NULL;
}
@@ -1273,7 +1273,7 @@ static void afs_vnode_new_inode(struct afs_operation *op)
_enter("");
- ASSERTCMP(op->error, ==, 0);
+ ASSERTCMP(afs_op_error(op), ==, 0);
inode = afs_iget(op, vp);
if (IS_ERR(inode)) {
@@ -1286,7 +1286,7 @@ static void afs_vnode_new_inode(struct afs_operation *op)
vnode = AFS_FS_I(inode);
set_bit(AFS_VNODE_NEW_CONTENT, &vnode->flags);
- if (!op->error)
+ if (!afs_op_error(op))
afs_cache_permit(vnode, op->key, vnode->cb_break, &vp->scb);
d_instantiate(op->dentry, inode);
}
@@ -1320,7 +1320,7 @@ static void afs_create_put(struct afs_operation *op)
{
_enter("op=%08x", op->debug_id);
- if (op->error)
+ if (afs_op_error(op))
d_drop(op->dentry);
}
@@ -1480,7 +1480,7 @@ static void afs_dir_remove_link(struct afs_operation *op)
struct dentry *dentry = op->dentry;
int ret;
- if (op->error != 0 ||
+ if (afs_op_error(op) ||
(op->file[1].scb.have_status && op->file[1].scb.have_error))
return;
if (d_really_is_positive(dentry))
@@ -1504,10 +1504,10 @@ static void afs_dir_remove_link(struct afs_operation *op)
ret = afs_validate(vnode, op->key);
if (ret != -ESTALE)
- op->error = ret;
+ afs_op_set_error(op, ret);
}
- _debug("nlink %d [val %d]", vnode->netfs.inode.i_nlink, op->error);
+ _debug("nlink %d [val %d]", vnode->netfs.inode.i_nlink, afs_op_error(op));
}
static void afs_unlink_success(struct afs_operation *op)
@@ -1538,7 +1538,7 @@ static void afs_unlink_edit_dir(struct afs_operation *op)
static void afs_unlink_put(struct afs_operation *op)
{
_enter("op=%08x", op->debug_id);
- if (op->unlink.need_rehash && op->error < 0 && op->error != -ENOENT)
+ if (op->unlink.need_rehash && afs_op_error(op) < 0 && afs_op_error(op) != -ENOENT)
d_rehash(op->dentry);
}
@@ -1579,7 +1579,7 @@ static int afs_unlink(struct inode *dir, struct dentry *dentry)
/* Try to make sure we have a callback promise on the victim. */
ret = afs_validate(vnode, op->key);
if (ret < 0) {
- op->error = ret;
+ afs_op_set_error(op, ret);
goto error;
}
@@ -1588,7 +1588,7 @@ static int afs_unlink(struct inode *dir, struct dentry *dentry)
spin_unlock(&dentry->d_lock);
/* Start asynchronous writeout of the inode */
write_inode_now(d_inode(dentry), 0);
- op->error = afs_sillyrename(dvnode, vnode, dentry, op->key);
+ afs_op_set_error(op, afs_sillyrename(dvnode, vnode, dentry, op->key));
goto error;
}
if (!d_unhashed(dentry)) {
@@ -1609,7 +1609,7 @@ static int afs_unlink(struct inode *dir, struct dentry *dentry)
/* If there was a conflict with a third party, check the status of the
* unlinked vnode.
*/
- if (op->error == 0 && (op->flags & AFS_OPERATION_DIR_CONFLICT)) {
+ if (afs_op_error(op) == 0 && (op->flags & AFS_OPERATION_DIR_CONFLICT)) {
op->file[1].update_ctime = false;
op->fetch_status.which = 1;
op->ops = &afs_fetch_status_operation;
@@ -1691,7 +1691,7 @@ static void afs_link_success(struct afs_operation *op)
static void afs_link_put(struct afs_operation *op)
{
_enter("op=%08x", op->debug_id);
- if (op->error)
+ if (afs_op_error(op))
d_drop(op->dentry);
}
@@ -1889,7 +1889,7 @@ static void afs_rename_put(struct afs_operation *op)
if (op->rename.rehash)
d_rehash(op->rename.rehash);
dput(op->rename.tmp);
- if (op->error)
+ if (afs_op_error(op))
d_rehash(op->dentry);
}
@@ -1934,7 +1934,7 @@ static int afs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
return PTR_ERR(op);
ret = afs_validate(vnode, op->key);
- op->error = ret;
+ afs_op_set_error(op, ret);
if (ret < 0)
goto error;
@@ -1971,7 +1971,7 @@ static int afs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
op->rename.tmp = d_alloc(new_dentry->d_parent,
&new_dentry->d_name);
if (!op->rename.tmp) {
- op->error = -ENOMEM;
+ afs_op_nomem(op);
goto error;
}
@@ -1979,7 +1979,7 @@ static int afs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
AFS_FS_I(d_inode(new_dentry)),
new_dentry, op->key);
if (ret) {
- op->error = ret;
+ afs_op_set_error(op, ret);
goto error;
}
diff --git a/fs/afs/file.c b/fs/afs/file.c
index d37dd201752b..0c81c39c32f5 100644
--- a/fs/afs/file.c
+++ b/fs/afs/file.c
@@ -243,7 +243,7 @@ static void afs_fetch_data_notify(struct afs_operation *op)
{
struct afs_read *req = op->fetch.req;
struct netfs_io_subrequest *subreq = req->subreq;
- int error = op->error;
+ int error = afs_op_error(op);
if (error == -ECONNABORTED)
error = afs_abort_to_error(op->ac.abort_code);
@@ -271,7 +271,7 @@ static void afs_fetch_data_success(struct afs_operation *op)
static void afs_fetch_data_put(struct afs_operation *op)
{
- op->fetch.req->error = op->error;
+ op->fetch.req->error = afs_op_error(op);
afs_put_read(op->fetch.req);
}
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index 3e31fae9a149..bfb9a7634bd9 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -40,8 +40,8 @@ struct afs_operation *afs_alloc_operation(struct key *key, struct afs_volume *vo
op->net = volume->cell->net;
op->cb_v_break = volume->cb_v_break;
op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
- op->error = -EDESTADDRREQ;
op->nr_iterations = -1;
+ afs_op_set_error(op, -EDESTADDRREQ);
_leave(" = [op=%08x]", op->debug_id);
return op;
@@ -71,7 +71,7 @@ static bool afs_get_io_locks(struct afs_operation *op)
swap(vnode, vnode2);
if (mutex_lock_interruptible(&vnode->io_lock) < 0) {
- op->error = -ERESTARTSYS;
+ afs_op_set_error(op, -ERESTARTSYS);
op->flags |= AFS_OPERATION_STOP;
_leave(" = f [I 0]");
return false;
@@ -80,7 +80,7 @@ static bool afs_get_io_locks(struct afs_operation *op)
if (vnode2) {
if (mutex_lock_interruptible_nested(&vnode2->io_lock, 1) < 0) {
- op->error = -ERESTARTSYS;
+ afs_op_set_error(op, -ERESTARTSYS);
op->flags |= AFS_OPERATION_STOP;
mutex_unlock(&vnode->io_lock);
op->flags &= ~AFS_OPERATION_LOCK_0;
@@ -159,11 +159,14 @@ static void afs_end_vnode_operation(struct afs_operation *op)
{
_enter("");
- if (op->error == -EDESTADDRREQ ||
- op->error == -EADDRNOTAVAIL ||
- op->error == -ENETUNREACH ||
- op->error == -EHOSTUNREACH)
+ switch (afs_op_error(op)) {
+ case -EDESTADDRREQ:
+ case -EADDRNOTAVAIL:
+ case -ENETUNREACH:
+ case -EHOSTUNREACH:
afs_dump_edestaddrreq(op);
+ break;
+ }
afs_drop_io_locks(op);
@@ -209,7 +212,7 @@ void afs_wait_for_operation(struct afs_operation *op)
afs_end_vnode_operation(op);
- if (op->error == 0 && op->ops->edit_dir) {
+ if (!afs_op_error(op) && op->ops->edit_dir) {
_debug("edit_dir");
op->ops->edit_dir(op);
}
@@ -221,7 +224,7 @@ void afs_wait_for_operation(struct afs_operation *op)
*/
int afs_put_operation(struct afs_operation *op)
{
- int i, ret = op->error;
+ int i, ret = afs_op_error(op);
_enter("op=%08x,%d", op->debug_id, ret);
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 7d37f63ef0f0..6821ce0f9d63 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -1899,7 +1899,7 @@ void afs_fs_inline_bulk_status(struct afs_operation *op)
int i;
if (test_bit(AFS_SERVER_FL_NO_IBULK, &op->server->flags)) {
- op->error = -ENOTSUPP;
+ afs_op_set_error(op, -ENOTSUPP);
return;
}
diff --git a/fs/afs/inode.c b/fs/afs/inode.c
index 1c794a1896aa..b3d76faa83c0 100644
--- a/fs/afs/inode.c
+++ b/fs/afs/inode.c
@@ -331,7 +331,7 @@ static void afs_fetch_status_success(struct afs_operation *op)
if (vnode->netfs.inode.i_state & I_NEW) {
ret = afs_inode_init_from_status(op, vp, vnode);
- op->error = ret;
+ afs_op_set_error(op, ret);
if (ret == 0)
afs_cache_permit(vnode, op->key, vp->cb_break_before, &vp->scb);
} else {
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 3739110fc0b5..08026dfd4421 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -1139,11 +1139,6 @@ extern bool afs_begin_vnode_operation(struct afs_operation *);
extern void afs_wait_for_operation(struct afs_operation *);
extern int afs_do_sync_operation(struct afs_operation *);
-static inline void afs_op_nomem(struct afs_operation *op)
-{
- op->error = -ENOMEM;
-}
-
static inline void afs_op_set_vnode(struct afs_operation *op, unsigned int n,
struct afs_vnode *vnode)
{
@@ -1237,6 +1232,21 @@ static inline void __afs_stat(atomic_t *s)
extern int afs_abort_to_error(u32);
extern void afs_prioritise_error(struct afs_error *, int, u32);
+static inline void afs_op_nomem(struct afs_operation *op)
+{
+ op->error = -ENOMEM;
+}
+
+static inline int afs_op_error(const struct afs_operation *op)
+{
+ return op->error;
+}
+
+static inline int afs_op_set_error(struct afs_operation *op, int error)
+{
+ return op->error = error;
+}
+
/*
* mntpt.c
*/
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index beb9fd4e8f44..282f9417f17d 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -51,7 +51,7 @@ static bool afs_start_fs_iteration(struct afs_operation *op,
* and have to return an error.
*/
if (op->flags & AFS_OPERATION_CUR_ONLY) {
- op->error = -ESTALE;
+ afs_op_set_error(op, -ESTALE);
return false;
}
@@ -93,7 +93,7 @@ static bool afs_sleep_and_retry(struct afs_operation *op)
if (!(op->flags & AFS_OPERATION_UNINTR)) {
msleep_interruptible(1000);
if (signal_pending(current)) {
- op->error = -ERESTARTSYS;
+ afs_op_set_error(op, -ERESTARTSYS);
return false;
}
} else {
@@ -137,7 +137,7 @@ bool afs_select_fileserver(struct afs_operation *op)
case 0:
default:
/* Success or local failure. Stop. */
- op->error = error;
+ afs_op_set_error(op, error);
op->flags |= AFS_OPERATION_STOP;
_leave(" = f [okay/local %d]", error);
return false;
@@ -174,11 +174,13 @@ bool afs_select_fileserver(struct afs_operation *op)
set_bit(AFS_VOLUME_NEEDS_UPDATE, &op->volume->flags);
error = afs_check_volume_status(op->volume, op);
- if (error < 0)
- goto failed_set_error;
+ if (error < 0) {
+ afs_op_set_error(op, error);
+ goto failed;
+ }
if (test_bit(AFS_VOLUME_DELETED, &op->volume->flags)) {
- op->error = -ENOMEDIUM;
+ afs_op_set_error(op, -ENOMEDIUM);
goto failed;
}
@@ -250,11 +252,11 @@ bool afs_select_fileserver(struct afs_operation *op)
clear_bit(AFS_VOLUME_BUSY, &op->volume->flags);
}
if (op->flags & AFS_OPERATION_NO_VSLEEP) {
- op->error = -EADV;
+ afs_op_set_error(op, -EADV);
goto failed;
}
if (op->flags & AFS_OPERATION_CUR_ONLY) {
- op->error = -ESTALE;
+ afs_op_set_error(op, -ESTALE);
goto failed;
}
goto busy;
@@ -274,7 +276,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* lock we need to maintain.
*/
if (op->flags & AFS_OPERATION_NO_VSLEEP) {
- op->error = -EBUSY;
+ afs_op_set_error(op, -EBUSY);
goto failed;
}
if (!test_and_set_bit(AFS_VOLUME_BUSY, &op->volume->flags)) {
@@ -303,7 +305,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* honour, just in case someone sets up a loop.
*/
if (op->flags & AFS_OPERATION_VMOVED) {
- op->error = -EREMOTEIO;
+ afs_op_set_error(op, -EREMOTEIO);
goto failed;
}
op->flags |= AFS_OPERATION_VMOVED;
@@ -311,8 +313,10 @@ bool afs_select_fileserver(struct afs_operation *op)
set_bit(AFS_VOLUME_WAIT, &op->volume->flags);
set_bit(AFS_VOLUME_NEEDS_UPDATE, &op->volume->flags);
error = afs_check_volume_status(op->volume, op);
- if (error < 0)
- goto failed_set_error;
+ if (error < 0) {
+ afs_op_set_error(op, error);
+ goto failed;
+ }
/* If the server list didn't change, then the VLDB is
* out of sync with the fileservers. This is hopefully
@@ -342,7 +346,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* Translate locally and return ENOSPC.
* No replicas to failover to.
*/
- op->error = -ENOSPC;
+ afs_op_set_error(op, -ENOSPC);
goto failed_but_online;
case VOVERQUOTA:
@@ -351,7 +355,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* Translate locally and return EDQUOT.
* No replicas to failover to.
*/
- op->error = -EDQUOT;
+ afs_op_set_error(op, -EDQUOT);
goto failed_but_online;
default:
@@ -364,7 +368,7 @@ bool afs_select_fileserver(struct afs_operation *op)
case -ETIMEDOUT:
case -ETIME:
- if (op->error != -EDESTADDRREQ)
+ if (afs_op_error(op) != -EDESTADDRREQ)
goto iterate_address;
fallthrough;
case -ERFKILL:
@@ -383,7 +387,7 @@ bool afs_select_fileserver(struct afs_operation *op)
fallthrough;
case -ECONNRESET:
_debug("call reset");
- op->error = error;
+ afs_op_set_error(op, error);
goto failed;
}
@@ -399,8 +403,10 @@ bool afs_select_fileserver(struct afs_operation *op)
* volume may have moved or even have been deleted.
*/
error = afs_check_volume_status(op->volume, op);
- if (error < 0)
- goto failed_set_error;
+ if (error < 0) {
+ afs_op_set_error(op, error);
+ goto failed;
+ }
if (!afs_start_fs_iteration(op, vnode))
goto failed;
@@ -411,8 +417,10 @@ bool afs_select_fileserver(struct afs_operation *op)
_debug("pick [%lx]", op->untried);
error = afs_wait_for_fs_probes(op->server_list, op->untried);
- if (error < 0)
- goto failed_set_error;
+ if (error < 0) {
+ afs_op_set_error(op, error);
+ goto failed;
+ }
/* Pick the untried server with the lowest RTT. If we have outstanding
* callbacks, we stick with the server we're already using if we can.
@@ -513,7 +521,8 @@ bool afs_select_fileserver(struct afs_operation *op)
op->flags &= ~AFS_OPERATION_RETRY_SERVER;
goto retry_server;
case -ERESTARTSYS:
- goto failed_set_error;
+ afs_op_set_error(op, error);
+ goto failed;
case -ETIME:
case -EDESTADDRREQ:
goto next_server;
@@ -542,13 +551,11 @@ bool afs_select_fileserver(struct afs_operation *op)
}
error = e.error;
-
-failed_set_error:
op->error = error;
failed:
op->flags |= AFS_OPERATION_STOP;
afs_end_cursor(&op->ac);
- _leave(" = f [failed %d]", op->error);
+ _leave(" = f [failed %d]", afs_op_error(op));
return false;
}
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 831254d8ef9c..a72192566dda 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -628,8 +628,8 @@ static noinline bool afs_update_server_record(struct afs_operation *op,
_leave(" = t [intr]");
return true;
}
- op->error = PTR_ERR(alist);
- _leave(" = f [%d]", op->error);
+ afs_op_set_error(op, PTR_ERR(alist));
+ _leave(" = f [%d]", afs_op_error(op));
return false;
}
@@ -683,7 +683,7 @@ bool afs_check_server_record(struct afs_operation *op, struct afs_server *server
(op->flags & AFS_OPERATION_UNINTR) ?
TASK_UNINTERRUPTIBLE : TASK_INTERRUPTIBLE);
if (ret == -ERESTARTSYS) {
- op->error = ret;
+ afs_op_set_error(op, ret);
_leave(" = f [intr]");
return false;
}
diff --git a/fs/afs/write.c b/fs/afs/write.c
index e1c45341719b..953f2c581afa 100644
--- a/fs/afs/write.c
+++ b/fs/afs/write.c
@@ -366,7 +366,7 @@ static void afs_store_data_success(struct afs_operation *op)
op->ctime = op->file[0].scb.status.mtime_client;
afs_vnode_commit_status(op, &op->file[0]);
- if (op->error == 0) {
+ if (!afs_op_error(op)) {
if (!op->store.laundering)
afs_pages_written_back(vnode, op->store.pos, op->store.size);
afs_stat_v(vnode, n_stores);
@@ -428,7 +428,7 @@ static int afs_store_data(struct afs_vnode *vnode, struct iov_iter *iter, loff_t
afs_wait_for_operation(op);
- switch (op->error) {
+ switch (afs_op_error(op)) {
case -EACCES:
case -EPERM:
case -ENOKEY:
@@ -447,7 +447,7 @@ static int afs_store_data(struct afs_vnode *vnode, struct iov_iter *iter, loff_t
}
afs_put_wb_key(wbk);
- _leave(" = %d", op->error);
+ _leave(" = %d", afs_op_error(op));
return afs_put_operation(op);
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 17/41] afs: Don't put afs_call in afs_wait_for_call_to_complete()
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (15 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 16/41] afs: Wrap most op->error accesses with inline funcs David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 18/41] afs: Simplify error handling David Howells
` (23 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Don't put the afs_call struct in afs_wait_for_call_to_complete() but rather
have the caller do it. This will allow the caller to fish stuff out of the
afs_call struct rather than the afs_addr_cursor struct, thereby allowing a
subsequent patch to subsume it.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/fs_operation.c | 7 +++--
fs/afs/fsclient.c | 5 ++-
fs/afs/internal.h | 2 +-
fs/afs/rxrpc.c | 73 ++++++++++++++++---------------------------
fs/afs/vlclient.c | 64 ++++++++++++++++++++++---------------
5 files changed, 75 insertions(+), 76 deletions(-)
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index bfb9a7634bd9..1c22d6e77846 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -191,8 +191,11 @@ void afs_wait_for_operation(struct afs_operation *op)
else
op->ac.error = -ENOTSUPP;
- if (op->call)
- op->error = afs_wait_for_call_to_complete(op->call, &op->ac);
+ if (op->call) {
+ afs_wait_for_call_to_complete(op->call, &op->ac);
+ op->error = op->ac.error;
+ afs_put_call(op->call);
+ }
}
switch (op->error) {
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 6821ce0f9d63..020073387111 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -1612,6 +1612,7 @@ int afs_fs_give_up_all_callbacks(struct afs_net *net,
{
struct afs_call *call;
__be32 *bp;
+ int ret;
_enter("");
@@ -1627,7 +1628,9 @@ int afs_fs_give_up_all_callbacks(struct afs_net *net,
call->server = afs_use_server(server, afs_server_trace_give_up_cb);
afs_make_call(ac, call, GFP_NOFS);
- return afs_wait_for_call_to_complete(call, ac);
+ afs_wait_for_call_to_complete(call, ac);
+ afs_put_call(call);
+ return ret;
}
/*
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 08026dfd4421..6ab353e145be 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -1290,7 +1290,7 @@ extern void __net_exit afs_close_socket(struct afs_net *);
extern void afs_charge_preallocation(struct work_struct *);
extern void afs_put_call(struct afs_call *);
extern void afs_make_call(struct afs_addr_cursor *, struct afs_call *, gfp_t);
-extern long afs_wait_for_call_to_complete(struct afs_call *, struct afs_addr_cursor *);
+void afs_wait_for_call_to_complete(struct afs_call *call, struct afs_addr_cursor *ac);
extern struct afs_call *afs_alloc_flat_call(struct afs_net *,
const struct afs_call_type *,
size_t, size_t);
diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
index 1a18fcbdec80..31044fba1bdb 100644
--- a/fs/afs/rxrpc.c
+++ b/fs/afs/rxrpc.c
@@ -575,48 +575,44 @@ static void afs_deliver_to_call(struct afs_call *call)
/*
* Wait synchronously for a call to complete and clean up the call struct.
*/
-long afs_wait_for_call_to_complete(struct afs_call *call,
- struct afs_addr_cursor *ac)
+void afs_wait_for_call_to_complete(struct afs_call *call, struct afs_addr_cursor *ac)
{
- long ret;
bool rxrpc_complete = false;
- DECLARE_WAITQUEUE(myself, current);
-
_enter("");
- ret = call->error;
- if (ret < 0)
- goto out;
+ if (!afs_check_call_state(call, AFS_CALL_COMPLETE)) {
+ DECLARE_WAITQUEUE(myself, current);
+
+ add_wait_queue(&call->waitq, &myself);
+ for (;;) {
+ set_current_state(TASK_UNINTERRUPTIBLE);
+
+ /* deliver any messages that are in the queue */
+ if (!afs_check_call_state(call, AFS_CALL_COMPLETE) &&
+ call->need_attention) {
+ call->need_attention = false;
+ __set_current_state(TASK_RUNNING);
+ afs_deliver_to_call(call);
+ continue;
+ }
- add_wait_queue(&call->waitq, &myself);
- for (;;) {
- set_current_state(TASK_UNINTERRUPTIBLE);
-
- /* deliver any messages that are in the queue */
- if (!afs_check_call_state(call, AFS_CALL_COMPLETE) &&
- call->need_attention) {
- call->need_attention = false;
- __set_current_state(TASK_RUNNING);
- afs_deliver_to_call(call);
- continue;
- }
+ if (afs_check_call_state(call, AFS_CALL_COMPLETE))
+ break;
- if (afs_check_call_state(call, AFS_CALL_COMPLETE))
- break;
+ if (!rxrpc_kernel_check_life(call->net->socket, call->rxcall)) {
+ /* rxrpc terminated the call. */
+ rxrpc_complete = true;
+ break;
+ }
- if (!rxrpc_kernel_check_life(call->net->socket, call->rxcall)) {
- /* rxrpc terminated the call. */
- rxrpc_complete = true;
- break;
+ schedule();
}
- schedule();
+ remove_wait_queue(&call->waitq, &myself);
+ __set_current_state(TASK_RUNNING);
}
- remove_wait_queue(&call->waitq, &myself);
- __set_current_state(TASK_RUNNING);
-
if (!afs_check_call_state(call, AFS_CALL_COMPLETE)) {
if (rxrpc_complete) {
afs_set_call_complete(call, call->error, call->abort_code);
@@ -635,23 +631,8 @@ long afs_wait_for_call_to_complete(struct afs_call *call,
ac->error = call->error;
spin_unlock_bh(&call->state_lock);
- ret = ac->error;
- switch (ret) {
- case 0:
- ret = call->ret0;
- call->ret0 = 0;
-
- fallthrough;
- case -ECONNABORTED:
+ if (call->error == 0 || call->error == -ECONNABORTED)
ac->responded = true;
- break;
- }
-
-out:
- _debug("call complete");
- afs_put_call(call);
- _leave(" = %p", (void *)ret);
- return ret;
}
/*
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index 41e7932d75c6..650534892a20 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -106,12 +106,6 @@ static int afs_deliver_vl_get_entry_by_name_u(struct afs_call *call)
return 0;
}
-static void afs_destroy_vl_get_entry_by_name_u(struct afs_call *call)
-{
- kfree(call->ret_vldb);
- afs_flat_call_destructor(call);
-}
-
/*
* VL.GetEntryByNameU operation type.
*/
@@ -119,7 +113,7 @@ static const struct afs_call_type afs_RXVLGetEntryByNameU = {
.name = "VL.GetEntryByNameU",
.op = afs_VL_GetEntryByNameU,
.deliver = afs_deliver_vl_get_entry_by_name_u,
- .destructor = afs_destroy_vl_get_entry_by_name_u,
+ .destructor = afs_flat_call_destructor,
};
/*
@@ -166,7 +160,13 @@ struct afs_vldb_entry *afs_vl_get_entry_by_name_u(struct afs_vl_cursor *vc,
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
- return (struct afs_vldb_entry *)afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_put_call(call);
+ if (vc->ac.error) {
+ kfree(entry);
+ return ERR_PTR(vc->ac.error);
+ }
+ return entry;
}
/*
@@ -249,12 +249,6 @@ static int afs_deliver_vl_get_addrs_u(struct afs_call *call)
return 0;
}
-static void afs_vl_get_addrs_u_destructor(struct afs_call *call)
-{
- afs_put_addrlist(call->ret_alist);
- return afs_flat_call_destructor(call);
-}
-
/*
* VL.GetAddrsU operation type.
*/
@@ -262,7 +256,7 @@ static const struct afs_call_type afs_RXVLGetAddrsU = {
.name = "VL.GetAddrsU",
.op = afs_VL_GetAddrsU,
.deliver = afs_deliver_vl_get_addrs_u,
- .destructor = afs_vl_get_addrs_u_destructor,
+ .destructor = afs_flat_call_destructor,
};
/*
@@ -273,6 +267,7 @@ struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *vc,
const uuid_t *uuid)
{
struct afs_ListAddrByAttributes__xdr *r;
+ struct afs_addr_list *alist;
const struct afs_uuid *u = (const struct afs_uuid *)uuid;
struct afs_call *call;
struct afs_net *net = vc->cell->net;
@@ -309,7 +304,14 @@ struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *vc,
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
- return (struct afs_addr_list *)afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_wait_for_call_to_complete(call, &vc->ac);
+ alist = call->ret_alist;
+ afs_put_call(call);
+ if (vc->ac.error) {
+ afs_put_addrlist(alist);
+ return ERR_PTR(vc->ac.error);
+ }
+ return alist;
}
/*
@@ -618,7 +620,7 @@ static const struct afs_call_type afs_YFSVLGetEndpoints = {
.name = "YFSVL.GetEndpoints",
.op = afs_YFSVL_GetEndpoints,
.deliver = afs_deliver_yfsvl_get_endpoints,
- .destructor = afs_vl_get_addrs_u_destructor,
+ .destructor = afs_flat_call_destructor,
};
/*
@@ -628,6 +630,7 @@ static const struct afs_call_type afs_YFSVLGetEndpoints = {
struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *vc,
const uuid_t *uuid)
{
+ struct afs_addr_list *alist;
struct afs_call *call;
struct afs_net *net = vc->cell->net;
__be32 *bp;
@@ -652,7 +655,14 @@ struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *vc,
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
- return (struct afs_addr_list *)afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_wait_for_call_to_complete(call, &vc->ac);
+ alist = call->ret_alist;
+ afs_put_call(call);
+ if (vc->ac.error) {
+ afs_put_addrlist(alist);
+ return ERR_PTR(vc->ac.error);
+ }
+ return alist;
}
/*
@@ -717,12 +727,6 @@ static int afs_deliver_yfsvl_get_cell_name(struct afs_call *call)
return 0;
}
-static void afs_destroy_yfsvl_get_cell_name(struct afs_call *call)
-{
- kfree(call->ret_str);
- afs_flat_call_destructor(call);
-}
-
/*
* VL.GetCapabilities operation type
*/
@@ -730,7 +734,7 @@ static const struct afs_call_type afs_YFSVLGetCellName = {
.name = "YFSVL.GetCellName",
.op = afs_YFSVL_GetCellName,
.deliver = afs_deliver_yfsvl_get_cell_name,
- .destructor = afs_destroy_yfsvl_get_cell_name,
+ .destructor = afs_flat_call_destructor,
};
/*
@@ -745,6 +749,7 @@ char *afs_yfsvl_get_cell_name(struct afs_vl_cursor *vc)
struct afs_call *call;
struct afs_net *net = vc->cell->net;
__be32 *bp;
+ char *cellname;
_enter("");
@@ -763,5 +768,12 @@ char *afs_yfsvl_get_cell_name(struct afs_vl_cursor *vc)
/* Can't take a ref on server */
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
- return (char *)afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_wait_for_call_to_complete(call, &vc->ac);
+ cellname = call->ret_str;
+ afs_put_call(call);
+ if (vc->ac.error) {
+ kfree(cellname);
+ return ERR_PTR(vc->ac.error);
+ }
+ return cellname;
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 18/41] afs: Simplify error handling
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (16 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 17/41] afs: Don't put afs_call in afs_wait_for_call_to_complete() David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 19/41] afs: Add a tracepoint for struct afs_addr_list David Howells
` (22 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Simplify error handling a bit by moving it from the afs_addr_cursor struct
to the afs_operation and afs_vl_cursor structs and using the error
prioritisation function for accumulating errors from multiple sources (AFS
tries to rotate between multiple fileservers, some of which may be
inaccessible or in some state of offlinedness).
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 8 ++--
fs/afs/dir.c | 14 ++++---
fs/afs/dir_silly.c | 2 +-
fs/afs/file.c | 3 --
fs/afs/fs_operation.c | 24 +++++------
fs/afs/fsclient.c | 1 +
fs/afs/internal.h | 44 +++++++++++++++------
fs/afs/misc.c | 10 ++++-
fs/afs/rotate.c | 58 ++++++++++++++-------------
fs/afs/rxrpc.c | 17 ++++----
fs/afs/server.c | 1 -
fs/afs/vl_alias.c | 2 +-
fs/afs/vl_probe.c | 7 ++--
fs/afs/vl_rotate.c | 92 +++++++++++++++++++++----------------------
fs/afs/vlclient.c | 34 ++++++++++------
15 files changed, 174 insertions(+), 143 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index b76abf500713..a1f3c995e328 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -386,26 +386,24 @@ bool afs_iterate_addresses(struct afs_addr_cursor *ac)
selected:
ac->index = index;
set_bit(index, &ac->tried);
- ac->responded = false;
+ ac->call_responded = false;
return true;
}
/*
* Release an address list cursor.
*/
-int afs_end_cursor(struct afs_addr_cursor *ac)
+void afs_end_cursor(struct afs_addr_cursor *ac)
{
struct afs_addr_list *alist;
alist = ac->alist;
if (alist) {
- if (ac->responded &&
+ if (ac->call_responded &&
ac->index != alist->preferred &&
test_bit(ac->alist->preferred, &ac->tried))
WRITE_ONCE(alist->preferred, ac->index);
afs_put_addrlist(alist);
ac->alist = NULL;
}
-
- return ac->error;
}
diff --git a/fs/afs/dir.c b/fs/afs/dir.c
index b40f7ae850a8..ecb889a269fb 100644
--- a/fs/afs/dir.c
+++ b/fs/afs/dir.c
@@ -693,8 +693,9 @@ static void afs_do_lookup_success(struct afs_operation *op)
vp = &op->file[0];
abort_code = vp->scb.status.abort_code;
if (abort_code != 0) {
- op->ac.abort_code = abort_code;
- op->error = afs_abort_to_error(abort_code);
+ op->call_abort_code = abort_code;
+ afs_op_set_error(op, afs_abort_to_error(abort_code));
+ op->cumul_error.abort_code = abort_code;
}
break;
@@ -846,13 +847,14 @@ static struct inode *afs_do_lookup(struct inode *dir, struct dentry *dentry,
_debug("nr_files %u", op->nr_files);
/* Need space for examining all the selected files */
- op->error = -ENOMEM;
if (op->nr_files > 2) {
op->more_files = kvcalloc(op->nr_files - 2,
sizeof(struct afs_vnode_param),
GFP_KERNEL);
- if (!op->more_files)
+ if (!op->more_files) {
+ afs_op_nomem(op);
goto out_op;
+ }
for (i = 2; i < op->nr_files; i++) {
vp = &op->more_files[i - 2];
@@ -1255,7 +1257,7 @@ void afs_check_for_remote_deletion(struct afs_operation *op)
{
struct afs_vnode *vnode = op->file[0].vnode;
- switch (op->ac.abort_code) {
+ switch (afs_op_abort_code(op)) {
case VNOVNODE:
set_bit(AFS_VNODE_DELETED, &vnode->flags);
afs_break_callback(vnode, afs_cb_break_for_deleted);
@@ -1280,7 +1282,7 @@ static void afs_vnode_new_inode(struct afs_operation *op)
/* ENOMEM or EINTR at a really inconvenient time - just abandon
* the new directory on the server.
*/
- op->error = PTR_ERR(inode);
+ afs_op_accumulate_error(op, PTR_ERR(inode), 0);
return;
}
diff --git a/fs/afs/dir_silly.c b/fs/afs/dir_silly.c
index bb5807e87fa4..a1e581946b93 100644
--- a/fs/afs/dir_silly.c
+++ b/fs/afs/dir_silly.c
@@ -218,7 +218,7 @@ static int afs_do_silly_unlink(struct afs_vnode *dvnode, struct afs_vnode *vnode
/* If there was a conflict with a third party, check the status of the
* unlinked vnode.
*/
- if (op->error == 0 && (op->flags & AFS_OPERATION_DIR_CONFLICT)) {
+ if (op->cumul_error.error == 0 && (op->flags & AFS_OPERATION_DIR_CONFLICT)) {
op->file[1].update_ctime = false;
op->fetch_status.which = 1;
op->ops = &afs_fetch_status_operation;
diff --git a/fs/afs/file.c b/fs/afs/file.c
index 0c81c39c32f5..8f9b42427569 100644
--- a/fs/afs/file.c
+++ b/fs/afs/file.c
@@ -245,10 +245,7 @@ static void afs_fetch_data_notify(struct afs_operation *op)
struct netfs_io_subrequest *subreq = req->subreq;
int error = afs_op_error(op);
- if (error == -ECONNABORTED)
- error = afs_abort_to_error(op->ac.abort_code);
req->error = error;
-
if (subreq) {
__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
netfs_subreq_terminated(subreq, error ?: req->actual_len, false);
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index 1c22d6e77846..cebe4fad8192 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -169,9 +169,6 @@ static void afs_end_vnode_operation(struct afs_operation *op)
}
afs_drop_io_locks(op);
-
- if (op->error == -ECONNABORTED)
- op->error = afs_abort_to_error(op->ac.abort_code);
}
/*
@@ -182,6 +179,8 @@ void afs_wait_for_operation(struct afs_operation *op)
_enter("");
while (afs_select_fileserver(op)) {
+ op->call_error = 0;
+ op->call_abort_code = 0;
op->cb_s_break = op->server->cb_s_break;
if (test_bit(AFS_SERVER_FL_IS_YFS, &op->server->flags) &&
op->ops->issue_yfs_rpc)
@@ -189,28 +188,29 @@ void afs_wait_for_operation(struct afs_operation *op)
else if (op->ops->issue_afs_rpc)
op->ops->issue_afs_rpc(op);
else
- op->ac.error = -ENOTSUPP;
+ op->call_error = -ENOTSUPP;
if (op->call) {
afs_wait_for_call_to_complete(op->call, &op->ac);
- op->error = op->ac.error;
+ op->call_abort_code = op->call->abort_code;
+ op->call_error = op->call->error;
+ op->call_responded = op->call->responded;
+ op->ac.call_responded = true;
+ WRITE_ONCE(op->ac.alist->addrs[op->ac.index].last_error,
+ op->call_error);
afs_put_call(op->call);
}
}
- switch (op->error) {
- case 0:
+ if (!afs_op_error(op)) {
_debug("success");
op->ops->success(op);
- break;
- case -ECONNABORTED:
+ } else if (op->cumul_error.aborted) {
if (op->ops->aborted)
op->ops->aborted(op);
- fallthrough;
- default:
+ } else {
if (op->ops->failed)
op->ops->failed(op);
- break;
}
afs_end_vnode_operation(op);
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 020073387111..2a56dea22519 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -1629,6 +1629,7 @@ int afs_fs_give_up_all_callbacks(struct afs_net *net,
call->server = afs_use_server(server, afs_server_trace_give_up_cb);
afs_make_call(ac, call, GFP_NOFS);
afs_wait_for_call_to_complete(call, ac);
+ ret = call->error;
afs_put_call(call);
return ret;
}
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 6ab353e145be..c5b9814e765a 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -75,6 +75,7 @@ enum afs_call_state {
struct afs_address {
struct rxrpc_peer *peer;
u16 service_id;
+ short last_error; /* Last error from this address */
};
/*
@@ -121,7 +122,6 @@ struct afs_call {
};
void *buffer; /* reply receive buffer */
union {
- long ret0; /* Value to reply with instead of 0 */
struct afs_addr_list *ret_alist;
struct afs_vldb_entry *ret_vldb;
char *ret_str;
@@ -145,6 +145,7 @@ struct afs_call {
bool upgrade; /* T to request service upgrade */
bool intr; /* T if interruptible */
bool unmarshalling_error; /* T if an unmarshalling error occurred */
+ bool responded; /* Got a response from the call (may be abort) */
u16 service_id; /* Actual service ID (after upgrade) */
unsigned int debug_id; /* Trace ID */
u32 operation_ID; /* operation ID for an incoming call */
@@ -718,8 +719,10 @@ struct afs_permits {
* Error prioritisation and accumulation.
*/
struct afs_error {
- short error; /* Accumulated error */
+ s32 abort_code; /* Cumulative abort code */
+ short error; /* Cumulative error */
bool responded; /* T if server responded */
+ bool aborted; /* T if ->error is from an abort */
};
/*
@@ -729,10 +732,8 @@ struct afs_addr_cursor {
struct afs_addr_list *alist; /* Current address list (pins ref) */
unsigned long tried; /* Tried addresses */
signed char index; /* Current address */
- bool responded; /* T if the current address responded */
unsigned short nr_iterations; /* Number of address iterations */
- short error;
- u32 abort_code;
+ bool call_responded;
};
/*
@@ -745,13 +746,16 @@ struct afs_vl_cursor {
struct afs_vlserver *server; /* Server on which this resides */
struct key *key; /* Key for the server */
unsigned long untried; /* Bitmask of untried servers */
+ struct afs_error cumul_error; /* Cumulative error */
+ s32 call_abort_code;
short index; /* Current server */
- short error;
+ short call_error; /* Error from single call */
unsigned short flags;
#define AFS_VL_CURSOR_STOP 0x0001 /* Set to cease iteration */
#define AFS_VL_CURSOR_RETRY 0x0002 /* Set to do a retry */
#define AFS_VL_CURSOR_RETRIED 0x0004 /* Set if started a retry */
- unsigned short nr_iterations; /* Number of server iterations */
+ short nr_iterations; /* Number of server iterations */
+ bool call_responded; /* T if the current address responded */
};
/*
@@ -802,8 +806,10 @@ struct afs_operation {
struct dentry *dentry_2; /* Second dentry to be altered */
struct timespec64 mtime; /* Modification time to record */
struct timespec64 ctime; /* Change time to set */
+ struct afs_error cumul_error; /* Cumulative error */
short nr_files; /* Number of entries in file[], more_files */
- short error;
+ short call_error; /* Error from single call */
+ s32 call_abort_code; /* Abort code from single call */
unsigned int debug_id;
unsigned int cb_v_break; /* Volume break counter before op */
@@ -859,6 +865,8 @@ struct afs_operation {
unsigned long untried; /* Bitmask of untried servers */
short index; /* Current server */
short nr_iterations; /* Number of server iterations */
+ bool call_responded; /* T if the current address responded */
+
unsigned int flags;
#define AFS_OPERATION_STOP 0x0001 /* Set to cease iteration */
@@ -975,7 +983,7 @@ bool afs_addr_list_same(const struct afs_addr_list *a,
const struct afs_addr_list *b);
extern struct afs_vlserver_list *afs_dns_query(struct afs_cell *, time64_t *);
extern bool afs_iterate_addresses(struct afs_addr_cursor *);
-extern int afs_end_cursor(struct afs_addr_cursor *);
+extern void afs_end_cursor(struct afs_addr_cursor *ac);
extern int afs_merge_fs_addr4(struct afs_net *net, struct afs_addr_list *addr,
__be32 xdr, u16 port);
@@ -1234,17 +1242,27 @@ extern void afs_prioritise_error(struct afs_error *, int, u32);
static inline void afs_op_nomem(struct afs_operation *op)
{
- op->error = -ENOMEM;
+ op->cumul_error.error = -ENOMEM;
}
static inline int afs_op_error(const struct afs_operation *op)
{
- return op->error;
+ return op->cumul_error.error;
+}
+
+static inline s32 afs_op_abort_code(const struct afs_operation *op)
+{
+ return op->cumul_error.abort_code;
}
static inline int afs_op_set_error(struct afs_operation *op, int error)
{
- return op->error = error;
+ return op->cumul_error.error = error;
+}
+
+static inline void afs_op_accumulate_error(struct afs_operation *op, int error, s32 abort_code)
+{
+ afs_prioritise_error(&op->cumul_error, error, abort_code);
}
/*
@@ -1617,7 +1635,7 @@ static inline void afs_update_dentry_version(struct afs_operation *op,
struct afs_vnode_param *dir_vp,
struct dentry *dentry)
{
- if (!op->error)
+ if (!op->cumul_error.error)
dentry->d_fsdata =
(void *)(unsigned long)dir_vp->scb.status.data_version;
}
diff --git a/fs/afs/misc.c b/fs/afs/misc.c
index 805328ca5428..b8180bf2281f 100644
--- a/fs/afs/misc.c
+++ b/fs/afs/misc.c
@@ -116,6 +116,8 @@ void afs_prioritise_error(struct afs_error *e, int error, u32 abort_code)
{
switch (error) {
case 0:
+ e->aborted = false;
+ e->error = 0;
return;
default:
if (e->error == -ETIMEDOUT ||
@@ -161,12 +163,16 @@ void afs_prioritise_error(struct afs_error *e, int error, u32 abort_code)
if (e->responded)
return;
e->error = error;
+ e->aborted = false;
return;
case -ECONNABORTED:
- error = afs_abort_to_error(abort_code);
- fallthrough;
+ e->error = afs_abort_to_error(abort_code);
+ e->aborted = true;
+ e->responded = true;
+ return;
case -ENETRESET: /* Responded, but we seem to have changed address */
+ e->aborted = false;
e->responded = true;
e->error = error;
return;
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 282f9417f17d..a9d9b99b441f 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -112,9 +112,9 @@ bool afs_select_fileserver(struct afs_operation *op)
struct afs_addr_list *alist;
struct afs_server *server;
struct afs_vnode *vnode = op->file[0].vnode;
- struct afs_error e;
unsigned int rtt;
- int error = op->ac.error, i;
+ s32 abort_code = op->call_abort_code;
+ int error = op->call_error, i;
op->nr_iterations++;
@@ -122,7 +122,7 @@ bool afs_select_fileserver(struct afs_operation *op)
op->debug_id, op->nr_iterations, op->volume->vid,
op->untried, op->index,
op->ac.tried, op->ac.index,
- error, op->ac.abort_code);
+ error, abort_code);
if (op->flags & AFS_OPERATION_STOP) {
_leave(" = f [stopped]");
@@ -133,8 +133,10 @@ bool afs_select_fileserver(struct afs_operation *op)
goto start;
/* Evaluate the result of the previous operation, if there was one. */
- switch (error) {
+ switch (op->call_error) {
case 0:
+ op->cumul_error.responded = true;
+ fallthrough;
default:
/* Success or local failure. Stop. */
afs_op_set_error(op, error);
@@ -151,7 +153,8 @@ bool afs_select_fileserver(struct afs_operation *op)
* errors instead. IBM AFS and OpenAFS fileservers, however, do leak
* these abort codes.
*/
- switch (op->ac.abort_code) {
+ op->cumul_error.responded = true;
+ switch (abort_code) {
case VNOVOL:
/* This fileserver doesn't know about the volume.
* - May indicate that the VL is wrong - retry once and compare
@@ -164,7 +167,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* (administrative action).
*/
if (op->flags & AFS_OPERATION_VNOVOL) {
- op->error = -EREMOTEIO;
+ afs_op_accumulate_error(op, -EREMOTEIO, abort_code);
goto next_server;
}
@@ -188,7 +191,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* it's the fileserver having trouble.
*/
if (rcu_access_pointer(op->volume->servers) == op->server_list) {
- op->error = -EREMOTEIO;
+ afs_op_accumulate_error(op, -EREMOTEIO, abort_code);
goto next_server;
}
@@ -201,8 +204,8 @@ bool afs_select_fileserver(struct afs_operation *op)
case VONLINE:
/* These should not be returned from the fileserver. */
pr_warn("Fileserver returned unexpected abort %d\n",
- op->ac.abort_code);
- op->error = -EREMOTEIO;
+ abort_code);
+ afs_op_accumulate_error(op, -EREMOTEIO, abort_code);
goto next_server;
case VNOSERVICE:
@@ -233,7 +236,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* VNOSERVICE should be treated as an alias for RX_CALL_TIMEOUT..
*/
case RX_CALL_TIMEOUT:
- op->error = -ETIMEDOUT;
+ afs_op_accumulate_error(op, -ETIMEDOUT, abort_code);
goto next_server;
case VSALVAGING: /* This error should not be leaked to cache managers
@@ -248,7 +251,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* days).
*/
if (!test_and_set_bit(AFS_VOLUME_OFFLINE, &op->volume->flags)) {
- afs_busy(op->volume, op->ac.abort_code);
+ afs_busy(op->volume, abort_code);
clear_bit(AFS_VOLUME_BUSY, &op->volume->flags);
}
if (op->flags & AFS_OPERATION_NO_VSLEEP) {
@@ -280,7 +283,7 @@ bool afs_select_fileserver(struct afs_operation *op)
goto failed;
}
if (!test_and_set_bit(AFS_VOLUME_BUSY, &op->volume->flags)) {
- afs_busy(op->volume, op->ac.abort_code);
+ afs_busy(op->volume, abort_code);
clear_bit(AFS_VOLUME_OFFLINE, &op->volume->flags);
}
busy:
@@ -328,14 +331,14 @@ bool afs_select_fileserver(struct afs_operation *op)
* TODO: Retry a few times with sleeps.
*/
if (rcu_access_pointer(op->volume->servers) == op->server_list) {
- op->error = -ENOMEDIUM;
+ afs_op_accumulate_error(op, -ENOMEDIUM, abort_code);
goto failed;
}
goto restart_from_beginning;
case VIO:
- op->error = -EREMOTEIO;
+ afs_op_accumulate_error(op, -EREMOTEIO, abort_code);
if (op->volume->type != AFSVL_RWVOL)
goto next_server;
goto failed;
@@ -359,7 +362,7 @@ bool afs_select_fileserver(struct afs_operation *op)
goto failed_but_online;
default:
- op->error = afs_abort_to_error(op->ac.abort_code);
+ afs_op_accumulate_error(op, error, abort_code);
failed_but_online:
clear_bit(AFS_VOLUME_OFFLINE, &op->volume->flags);
clear_bit(AFS_VOLUME_BUSY, &op->volume->flags);
@@ -378,7 +381,7 @@ bool afs_select_fileserver(struct afs_operation *op)
case -EHOSTDOWN:
case -ECONNREFUSED:
_debug("no conn");
- op->error = error;
+ afs_op_accumulate_error(op, error, 0);
goto iterate_address;
case -ENETRESET:
@@ -504,6 +507,7 @@ bool afs_select_fileserver(struct afs_operation *op)
op->index, op->ac.index, op->ac.alist->nr_addrs,
rxrpc_kernel_remote_addr(op->ac.alist->addrs[op->ac.index].peer));
+ op->call_responded = false;
_leave(" = t");
return true;
@@ -541,17 +545,14 @@ bool afs_select_fileserver(struct afs_operation *op)
if (op->flags & AFS_OPERATION_VBUSY)
goto restart_from_beginning;
- e.error = -EDESTADDRREQ;
- e.responded = false;
for (i = 0; i < op->server_list->nr_servers; i++) {
struct afs_server *s = op->server_list->servers[i].server;
- afs_prioritise_error(&e, READ_ONCE(s->probe.error),
- s->probe.abort_code);
+ error = READ_ONCE(s->probe.error);
+ if (error < 0)
+ afs_op_accumulate_error(op, error, s->probe.abort_code);
}
- error = e.error;
- op->error = error;
failed:
op->flags |= AFS_OPERATION_STOP;
afs_end_cursor(&op->ac);
@@ -574,11 +575,13 @@ void afs_dump_edestaddrreq(const struct afs_operation *op)
rcu_read_lock();
pr_notice("EDESTADDR occurred\n");
- pr_notice("FC: cbb=%x cbb2=%x fl=%x err=%hd\n",
+ pr_notice("OP: cbb=%x cbb2=%x fl=%x err=%hd\n",
op->file[0].cb_break_before,
- op->file[1].cb_break_before, op->flags, op->error);
- pr_notice("FC: ut=%lx ix=%d ni=%u\n",
+ op->file[1].cb_break_before, op->flags, op->cumul_error.error);
+ pr_notice("OP: ut=%lx ix=%d ni=%u\n",
op->untried, op->index, op->nr_iterations);
+ pr_notice("OP: call er=%d ac=%d r=%u\n",
+ op->call_error, op->call_abort_code, op->call_responded);
if (op->server_list) {
const struct afs_server_list *sl = op->server_list;
@@ -603,8 +606,7 @@ void afs_dump_edestaddrreq(const struct afs_operation *op)
}
}
- pr_notice("AC: t=%lx ax=%u ac=%d er=%d r=%u ni=%u\n",
- op->ac.tried, op->ac.index, op->ac.abort_code, op->ac.error,
- op->ac.responded, op->ac.nr_iterations);
+ pr_notice("AC: t=%lx ax=%u ni=%u\n",
+ op->ac.tried, op->ac.index, op->ac.nr_iterations);
rcu_read_unlock();
}
diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
index 31044fba1bdb..7408513e24d7 100644
--- a/fs/afs/rxrpc.c
+++ b/fs/afs/rxrpc.c
@@ -408,8 +408,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
rxrpc_kernel_recv_data(call->net->socket, rxcall,
&msg.msg_iter, &len, false,
&call->abort_code, &call->service_id);
- ac->abort_code = call->abort_code;
- ac->responded = true;
+ call->responded = true;
}
call->error = ret;
trace_afs_call_done(call);
@@ -429,7 +428,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
afs_put_call(call);
}
- ac->error = ret;
+ call->error = ret;
call->state = AFS_CALL_COMPLETE;
_leave(" = %d", ret);
}
@@ -510,6 +509,7 @@ static void afs_deliver_to_call(struct afs_call *call)
ret = -EBADMSG;
switch (ret) {
case 0:
+ call->responded = true;
afs_queue_call_work(call);
if (state == AFS_CALL_CL_PROC_REPLY) {
if (call->op)
@@ -524,9 +524,11 @@ static void afs_deliver_to_call(struct afs_call *call)
goto out;
case -ECONNABORTED:
ASSERTCMP(state, ==, AFS_CALL_COMPLETE);
+ call->responded = true;
afs_log_error(call, call->abort_code);
goto done;
case -ENOTSUPP:
+ call->responded = true;
abort_code = RXGEN_OPCODE;
rxrpc_kernel_abort_call(call->net->socket, call->rxcall,
abort_code, ret,
@@ -573,7 +575,7 @@ static void afs_deliver_to_call(struct afs_call *call)
}
/*
- * Wait synchronously for a call to complete and clean up the call struct.
+ * Wait synchronously for a call to complete.
*/
void afs_wait_for_call_to_complete(struct afs_call *call, struct afs_addr_cursor *ac)
{
@@ -626,13 +628,8 @@ void afs_wait_for_call_to_complete(struct afs_call *call, struct afs_addr_cursor
}
}
- spin_lock_bh(&call->state_lock);
- ac->abort_code = call->abort_code;
- ac->error = call->error;
- spin_unlock_bh(&call->state_lock);
-
if (call->error == 0 || call->error == -ECONNABORTED)
- ac->responded = true;
+ call->responded = true;
}
/*
diff --git a/fs/afs/server.c b/fs/afs/server.c
index a72192566dda..6a773dcbbf2d 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -436,7 +436,6 @@ static void afs_give_up_callbacks(struct afs_net *net, struct afs_server *server
struct afs_addr_cursor ac = {
.alist = alist,
.index = alist->preferred,
- .error = 0,
};
afs_fs_give_up_all_callbacks(net, server, &ac, NULL);
diff --git a/fs/afs/vl_alias.c b/fs/afs/vl_alias.c
index 6fdf9f1bedc0..89cadd9a69e1 100644
--- a/fs/afs/vl_alias.c
+++ b/fs/afs/vl_alias.c
@@ -236,7 +236,7 @@ static char *afs_vl_get_cell_name(struct afs_cell *cell, struct key *key)
while (afs_select_vlserver(&vc)) {
if (!test_bit(AFS_VLSERVER_FL_IS_YFS, &vc.server->flags)) {
- vc.ac.error = -EOPNOTSUPP;
+ vc.call_error = -EOPNOTSUPP;
skipped = true;
continue;
}
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index 44bff3a2a5ac..138f5715619d 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -169,10 +169,11 @@ static bool afs_do_probe_vlserver(struct afs_net *net,
call = afs_vl_get_capabilities(net, &ac, key, server,
server_index);
if (!IS_ERR(call)) {
+ afs_prioritise_error(_e, call->error, call->abort_code);
afs_put_call(call);
in_progress = true;
} else {
- afs_prioritise_error(_e, PTR_ERR(call), ac.abort_code);
+ afs_prioritise_error(_e, PTR_ERR(call), 0);
afs_done_one_vl_probe(server, false);
}
}
@@ -187,12 +188,10 @@ int afs_send_vl_probes(struct afs_net *net, struct key *key,
struct afs_vlserver_list *vllist)
{
struct afs_vlserver *server;
- struct afs_error e;
+ struct afs_error e = {};
bool in_progress = false;
int i;
- e.error = 0;
- e.responded = false;
for (i = 0; i < vllist->nr_servers; i++) {
server = vllist->servers[i].server;
if (test_bit(AFS_VLSERVER_FL_PROBED, &server->flags))
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index 01c890b3d18d..7ae73418697d 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -20,11 +20,11 @@ bool afs_begin_vlserver_operation(struct afs_vl_cursor *vc, struct afs_cell *cel
memset(vc, 0, sizeof(*vc));
vc->cell = cell;
vc->key = key;
- vc->error = -EDESTADDRREQ;
- vc->ac.error = SHRT_MAX;
+ vc->cumul_error.error = -EDESTADDRREQ;
+ vc->nr_iterations = -1;
if (signal_pending(current)) {
- vc->error = -EINTR;
+ vc->cumul_error.error = -EINTR;
vc->flags |= AFS_VL_CURSOR_STOP;
return false;
}
@@ -52,7 +52,7 @@ static bool afs_start_vl_iteration(struct afs_vl_cursor *vc)
&cell->dns_lookup_count,
smp_load_acquire(&cell->dns_lookup_count)
!= dns_lookup_count) < 0) {
- vc->error = -ERESTARTSYS;
+ vc->cumul_error.error = -ERESTARTSYS;
return false;
}
}
@@ -60,12 +60,12 @@ static bool afs_start_vl_iteration(struct afs_vl_cursor *vc)
/* Status load is ordered after lookup counter load */
if (cell->dns_status == DNS_LOOKUP_GOT_NOT_FOUND) {
pr_warn("No record of cell %s\n", cell->name);
- vc->error = -ENOENT;
+ vc->cumul_error.error = -ENOENT;
return false;
}
if (cell->dns_source == DNS_RECORD_UNAVAILABLE) {
- vc->error = -EDESTADDRREQ;
+ vc->cumul_error.error = -EDESTADDRREQ;
return false;
}
}
@@ -91,52 +91,52 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
{
struct afs_addr_list *alist;
struct afs_vlserver *vlserver;
- struct afs_error e;
unsigned int rtt;
- int error = vc->ac.error, i;
+ s32 abort_code = vc->call_abort_code;
+ int error = vc->call_error, i;
+
+ vc->nr_iterations++;
_enter("%lx[%d],%lx[%d],%d,%d",
vc->untried, vc->index,
vc->ac.tried, vc->ac.index,
- error, vc->ac.abort_code);
+ error, abort_code);
if (vc->flags & AFS_VL_CURSOR_STOP) {
_leave(" = f [stopped]");
return false;
}
- vc->nr_iterations++;
+ if (vc->nr_iterations == 0)
+ goto start;
/* Evaluate the result of the previous operation, if there was one. */
switch (error) {
- case SHRT_MAX:
- goto start;
-
default:
case 0:
/* Success or local failure. Stop. */
- vc->error = error;
+ vc->cumul_error.error = error;
vc->flags |= AFS_VL_CURSOR_STOP;
- _leave(" = f [okay/local %d]", vc->ac.error);
+ _leave(" = f [okay/local %d]", vc->cumul_error.error);
return false;
case -ECONNABORTED:
/* The far side rejected the operation on some grounds. This
* might involve the server being busy or the volume having been moved.
*/
- switch (vc->ac.abort_code) {
+ switch (abort_code) {
case AFSVL_IO:
case AFSVL_BADVOLOPER:
case AFSVL_NOMEM:
/* The server went weird. */
- vc->error = -EREMOTEIO;
+ afs_prioritise_error(&vc->cumul_error, -EREMOTEIO, abort_code);
//write_lock(&vc->cell->vl_servers_lock);
//vc->server_list->weird_mask |= 1 << vc->index;
//write_unlock(&vc->cell->vl_servers_lock);
goto next_server;
default:
- vc->error = afs_abort_to_error(vc->ac.abort_code);
+ afs_prioritise_error(&vc->cumul_error, error, abort_code);
goto failed;
}
@@ -149,12 +149,12 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
case -ETIMEDOUT:
case -ETIME:
_debug("no conn %d", error);
- vc->error = error;
+ afs_prioritise_error(&vc->cumul_error, error, 0);
goto iterate_address;
case -ECONNRESET:
_debug("call reset");
- vc->error = error;
+ afs_prioritise_error(&vc->cumul_error, error, 0);
vc->flags |= AFS_VL_CURSOR_RETRY;
goto next_server;
@@ -178,15 +178,19 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
goto failed;
error = afs_send_vl_probes(vc->cell->net, vc->key, vc->server_list);
- if (error < 0)
- goto failed_set_error;
+ if (error < 0) {
+ afs_prioritise_error(&vc->cumul_error, error, 0);
+ goto failed;
+ }
pick_server:
_debug("pick [%lx]", vc->untried);
error = afs_wait_for_vl_probes(vc->server_list, vc->untried);
- if (error < 0)
- goto failed_set_error;
+ if (error < 0) {
+ afs_prioritise_error(&vc->cumul_error, error, 0);
+ goto failed;
+ }
/* Pick the untried server with the lowest RTT. */
vc->index = vc->server_list->preferred;
@@ -249,6 +253,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
_debug("VL address %d/%d", vc->ac.index, vc->ac.alist->nr_addrs);
+ vc->call_responded = false;
_leave(" = t %pISpc", rxrpc_kernel_remote_addr(vc->ac.alist->addrs[vc->ac.index].peer));
return true;
@@ -264,25 +269,19 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
if (vc->flags & AFS_VL_CURSOR_RETRY)
goto restart_from_beginning;
- e.error = -EDESTADDRREQ;
- e.responded = false;
for (i = 0; i < vc->server_list->nr_servers; i++) {
struct afs_vlserver *s = vc->server_list->servers[i].server;
if (test_bit(AFS_VLSERVER_FL_RESPONDING, &s->flags))
- e.responded = true;
- afs_prioritise_error(&e, READ_ONCE(s->probe.error),
+ vc->cumul_error.responded = true;
+ afs_prioritise_error(&vc->cumul_error, READ_ONCE(s->probe.error),
s->probe.abort_code);
}
- error = e.error;
-
-failed_set_error:
- vc->error = error;
failed:
vc->flags |= AFS_VL_CURSOR_STOP;
afs_end_cursor(&vc->ac);
- _leave(" = f [failed %d]", vc->error);
+ _leave(" = f [failed %d]", vc->cumul_error.error);
return false;
}
@@ -305,7 +304,10 @@ static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
pr_notice("DNS: src=%u st=%u lc=%x\n",
cell->dns_source, cell->dns_status, cell->dns_lookup_count);
pr_notice("VC: ut=%lx ix=%u ni=%hu fl=%hx err=%hd\n",
- vc->untried, vc->index, vc->nr_iterations, vc->flags, vc->error);
+ vc->untried, vc->index, vc->nr_iterations, vc->flags,
+ vc->cumul_error.error);
+ pr_notice("VC: call er=%d ac=%d r=%u\n",
+ vc->call_error, vc->call_abort_code, vc->call_responded);
if (vc->server_list) {
const struct afs_vlserver_list *sl = vc->server_list;
@@ -329,9 +331,8 @@ static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
}
}
- pr_notice("AC: t=%lx ax=%u ac=%d er=%d r=%u ni=%u\n",
- vc->ac.tried, vc->ac.index, vc->ac.abort_code, vc->ac.error,
- vc->ac.responded, vc->ac.nr_iterations);
+ pr_notice("AC: t=%lx ax=%u ni=%u\n",
+ vc->ac.tried, vc->ac.index, vc->ac.nr_iterations);
rcu_read_unlock();
}
@@ -342,17 +343,16 @@ int afs_end_vlserver_operation(struct afs_vl_cursor *vc)
{
struct afs_net *net = vc->cell->net;
- if (vc->error == -EDESTADDRREQ ||
- vc->error == -EADDRNOTAVAIL ||
- vc->error == -ENETUNREACH ||
- vc->error == -EHOSTUNREACH)
+ switch (vc->cumul_error.error) {
+ case -EDESTADDRREQ:
+ case -EADDRNOTAVAIL:
+ case -ENETUNREACH:
+ case -EHOSTUNREACH:
afs_vl_dump_edestaddrreq(vc);
+ break;
+ }
afs_end_cursor(&vc->ac);
afs_put_vlserverlist(net, vc->server_list);
-
- if (vc->error == -ECONNABORTED)
- vc->error = afs_abort_to_error(vc->ac.abort_code);
-
- return vc->error;
+ return vc->cumul_error.error;
}
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index 650534892a20..db7e94584e87 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -161,10 +161,13 @@ struct afs_vldb_entry *afs_vl_get_entry_by_name_u(struct afs_vl_cursor *vc,
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
afs_wait_for_call_to_complete(call, &vc->ac);
+ vc->call_abort_code = call->abort_code;
+ vc->call_error = call->error;
+ vc->call_responded = call->responded;
afs_put_call(call);
- if (vc->ac.error) {
+ if (vc->call_error) {
kfree(entry);
- return ERR_PTR(vc->ac.error);
+ return ERR_PTR(vc->call_error);
}
return entry;
}
@@ -305,11 +308,14 @@ struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *vc,
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
afs_wait_for_call_to_complete(call, &vc->ac);
- alist = call->ret_alist;
+ vc->call_abort_code = call->abort_code;
+ vc->call_error = call->error;
+ vc->call_responded = call->responded;
+ alist = call->ret_alist;
afs_put_call(call);
- if (vc->ac.error) {
+ if (vc->call_error) {
afs_put_addrlist(alist);
- return ERR_PTR(vc->ac.error);
+ return ERR_PTR(vc->call_error);
}
return alist;
}
@@ -656,11 +662,14 @@ struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *vc,
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
afs_wait_for_call_to_complete(call, &vc->ac);
- alist = call->ret_alist;
+ vc->call_abort_code = call->abort_code;
+ vc->call_error = call->error;
+ vc->call_responded = call->responded;
+ alist = call->ret_alist;
afs_put_call(call);
- if (vc->ac.error) {
+ if (vc->call_error) {
afs_put_addrlist(alist);
- return ERR_PTR(vc->ac.error);
+ return ERR_PTR(vc->call_error);
}
return alist;
}
@@ -769,11 +778,14 @@ char *afs_yfsvl_get_cell_name(struct afs_vl_cursor *vc)
trace_afs_make_vl_call(call);
afs_make_call(&vc->ac, call, GFP_KERNEL);
afs_wait_for_call_to_complete(call, &vc->ac);
- cellname = call->ret_str;
+ vc->call_abort_code = call->abort_code;
+ vc->call_error = call->error;
+ vc->call_responded = call->responded;
+ cellname = call->ret_str;
afs_put_call(call);
- if (vc->ac.error) {
+ if (vc->call_error) {
kfree(cellname);
- return ERR_PTR(vc->ac.error);
+ return ERR_PTR(vc->call_error);
}
return cellname;
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 19/41] afs: Add a tracepoint for struct afs_addr_list
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (17 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 18/41] afs: Simplify error handling David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 20/41] afs: Rename some fields David Howells
` (21 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Add a tracepoint to track the lifetime of the afs_addr_list struct.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 33 ++++++++++++++++++++++----
fs/afs/fs_probe.c | 4 ++--
fs/afs/internal.h | 10 +++-----
fs/afs/rotate.c | 4 ++--
fs/afs/rxrpc.c | 4 ++--
fs/afs/server.c | 9 ++++----
fs/afs/vl_list.c | 11 +++++----
fs/afs/vl_rotate.c | 9 ++------
fs/afs/vlclient.c | 4 ++--
include/trace/events/afs.h | 47 ++++++++++++++++++++++++++++++++++++++
10 files changed, 100 insertions(+), 35 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index a1f3c995e328..41ef0c879239 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -20,17 +20,39 @@ static void afs_free_addrlist(struct rcu_head *rcu)
for (i = 0; i < alist->nr_addrs; i++)
rxrpc_kernel_put_peer(alist->addrs[i].peer);
+ trace_afs_alist(alist->debug_id, refcount_read(&alist->usage), afs_alist_trace_free);
+ kfree(alist);
}
/*
* Release an address list.
*/
-void afs_put_addrlist(struct afs_addr_list *alist)
+void afs_put_addrlist(struct afs_addr_list *alist, enum afs_alist_trace reason)
{
- if (alist && refcount_dec_and_test(&alist->usage))
+ unsigned int debug_id;
+ bool dead;
+ int r;
+
+ if (!alist)
+ return;
+ debug_id = alist->debug_id;
+ dead = __refcount_dec_and_test(&alist->usage, &r);
+ trace_afs_alist(debug_id, r - 1, reason);
+ if (dead)
call_rcu(&alist->rcu, afs_free_addrlist);
}
+struct afs_addr_list *afs_get_addrlist(struct afs_addr_list *alist, enum afs_alist_trace reason)
+{
+ int r;
+
+ if (alist) {
+ __refcount_inc(&alist->usage, &r);
+ trace_afs_alist(alist->debug_id, r + 1, reason);
+ }
+ return alist;
+}
+
/*
* Allocate an address list.
*/
@@ -38,6 +60,7 @@ struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id)
{
struct afs_addr_list *alist;
unsigned int i;
+ static atomic_t debug_id;
_enter("%u,%u", nr, service_id);
@@ -50,9 +73,11 @@ struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id)
refcount_set(&alist->usage, 1);
alist->max_addrs = nr;
+ alist->debug_id = atomic_inc_return(&debug_id);
for (i = 0; i < nr; i++)
alist->addrs[i].service_id = service_id;
+ trace_afs_alist(alist->debug_id, 1, afs_alist_trace_alloc);
return alist;
}
@@ -217,7 +242,7 @@ struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *net,
problem, p - text, (int)len, (int)len, text);
ret = -EINVAL;
error:
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_parse_error);
error_vl:
afs_put_vlserverlist(net, vllist);
return ERR_PTR(ret);
@@ -403,7 +428,7 @@ void afs_end_cursor(struct afs_addr_cursor *ac)
ac->index != alist->preferred &&
test_bit(ac->alist->preferred, &ac->tried))
WRITE_ONCE(alist->preferred, ac->index);
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_end_cursor);
ac->alist = NULL;
}
}
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index fbb91ad775b9..18891492c0b4 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -205,7 +205,7 @@ void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
read_lock(&server->fs_lock);
ac.alist = rcu_dereference_protected(server->addresses,
lockdep_is_held(&server->fs_lock));
- afs_get_addrlist(ac.alist);
+ afs_get_addrlist(ac.alist, afs_alist_trace_get_probe);
read_unlock(&server->fs_lock);
server->probed_at = jiffies;
@@ -226,7 +226,7 @@ void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
afs_fs_probe_not_done(net, server, &ac);
}
- afs_put_addrlist(ac.alist);
+ afs_put_addrlist(ac.alist, afs_alist_trace_put_probe);
}
/*
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index c5b9814e765a..d5a26cea2358 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -85,6 +85,7 @@ struct afs_addr_list {
struct rcu_head rcu;
refcount_t usage;
u32 version; /* Version */
+ unsigned int debug_id;
unsigned char max_addrs;
unsigned char nr_addrs;
unsigned char preferred; /* Preferred address */
@@ -968,14 +969,9 @@ static inline bool afs_is_folio_dirty_mmapped(unsigned long priv)
/*
* addr_list.c
*/
-static inline struct afs_addr_list *afs_get_addrlist(struct afs_addr_list *alist)
-{
- if (alist)
- refcount_inc(&alist->usage);
- return alist;
-}
+struct afs_addr_list *afs_get_addrlist(struct afs_addr_list *alist, enum afs_alist_trace reason);
extern struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id);
-extern void afs_put_addrlist(struct afs_addr_list *);
+extern void afs_put_addrlist(struct afs_addr_list *alist, enum afs_alist_trace reason);
extern struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *,
const char *, size_t, char,
unsigned short, unsigned short);
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index a9d9b99b441f..1767c6e67184 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -482,7 +482,7 @@ bool afs_select_fileserver(struct afs_operation *op)
read_lock(&server->fs_lock);
alist = rcu_dereference_protected(server->addresses,
lockdep_is_held(&server->fs_lock));
- afs_get_addrlist(alist);
+ afs_get_addrlist(alist, afs_alist_trace_get_fsrotate_set);
read_unlock(&server->fs_lock);
retry_server:
@@ -491,7 +491,7 @@ bool afs_select_fileserver(struct afs_operation *op)
if (!op->ac.alist)
op->ac.alist = alist;
else
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_retry_server);
op->ac.index = -1;
diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
index 7408513e24d7..ede7ab4b5dfe 100644
--- a/fs/afs/rxrpc.c
+++ b/fs/afs/rxrpc.c
@@ -187,7 +187,7 @@ void afs_put_call(struct afs_call *call)
call->type->destructor(call);
afs_unuse_server_notime(call->net, call->server, afs_server_trace_put_call);
- afs_put_addrlist(call->alist);
+ afs_put_addrlist(call->alist, afs_alist_trace_put_call);
kfree(call->request);
trace_afs_call(call->debug_id, afs_call_trace_free, 0, o,
@@ -315,7 +315,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
atomic_read(&call->net->nr_outstanding_calls));
call->addr_ix = ac->index;
- call->alist = afs_get_addrlist(ac->alist);
+ call->alist = afs_get_addrlist(ac->alist, afs_alist_trace_get_make_call);
/* Work out the length we're going to transmit. This is awkward for
* calls such as FS.StoreData where there's an extra injection of data
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 6a773dcbbf2d..e4c369132935 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -274,13 +274,13 @@ struct afs_server *afs_lookup_server(struct afs_cell *cell, struct key *key,
candidate = afs_alloc_server(cell, uuid, alist);
if (!candidate) {
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_server_oom);
return ERR_PTR(-ENOMEM);
}
server = afs_install_server(cell, candidate);
if (server != candidate) {
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_server_dup);
kfree(candidate);
} else {
/* Immediately dispatch an asynchronous probe to each interface
@@ -420,7 +420,8 @@ static void afs_server_rcu(struct rcu_head *rcu)
trace_afs_server(server->debug_id, refcount_read(&server->ref),
atomic_read(&server->active), afs_server_trace_free);
- afs_put_addrlist(rcu_access_pointer(server->addresses));
+ afs_put_addrlist(rcu_access_pointer(server->addresses),
+ afs_alist_trace_put_server);
kfree(server);
}
@@ -642,7 +643,7 @@ static noinline bool afs_update_server_record(struct afs_operation *op,
write_unlock(&server->fs_lock);
}
- afs_put_addrlist(discard);
+ afs_put_addrlist(discard, afs_alist_trace_put_server_update);
_leave(" = t");
return true;
}
diff --git a/fs/afs/vl_list.c b/fs/afs/vl_list.c
index ba89140eee9e..3a2875933261 100644
--- a/fs/afs/vl_list.c
+++ b/fs/afs/vl_list.c
@@ -33,7 +33,8 @@ static void afs_vlserver_rcu(struct rcu_head *rcu)
{
struct afs_vlserver *vlserver = container_of(rcu, struct afs_vlserver, rcu);
- afs_put_addrlist(rcu_access_pointer(vlserver->addresses));
+ afs_put_addrlist(rcu_access_pointer(vlserver->addresses),
+ afs_alist_trace_put_vlserver);
kfree_rcu(vlserver, rcu);
}
@@ -145,7 +146,7 @@ static struct afs_addr_list *afs_extract_vl_addrs(struct afs_net *net,
error:
*_b = b;
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_parse_error);
return ERR_PTR(ret);
}
@@ -260,7 +261,7 @@ struct afs_vlserver_list *afs_extract_vlserver_list(struct afs_cell *cell,
if (vllist->nr_servers >= nr_servers) {
_debug("skip %u >= %u", vllist->nr_servers, nr_servers);
- afs_put_addrlist(addrs);
+ afs_put_addrlist(addrs, afs_alist_trace_put_parse_empty);
afs_put_vlserver(cell->net, server);
continue;
}
@@ -269,7 +270,7 @@ struct afs_vlserver_list *afs_extract_vlserver_list(struct afs_cell *cell,
addrs->status = bs.status;
if (addrs->nr_addrs == 0) {
- afs_put_addrlist(addrs);
+ afs_put_addrlist(addrs, afs_alist_trace_put_parse_empty);
if (!rcu_access_pointer(server->addresses)) {
afs_put_vlserver(cell->net, server);
continue;
@@ -281,7 +282,7 @@ struct afs_vlserver_list *afs_extract_vlserver_list(struct afs_cell *cell,
old = rcu_replace_pointer(server->addresses, old,
lockdep_is_held(&server->lock));
write_unlock(&server->lock);
- afs_put_addrlist(old);
+ afs_put_addrlist(old, afs_alist_trace_put_vlserver_old);
}
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index 7ae73418697d..e8fbbeb551bb 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -231,16 +231,11 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
read_lock(&vlserver->lock);
alist = rcu_dereference_protected(vlserver->addresses,
lockdep_is_held(&vlserver->lock));
- afs_get_addrlist(alist);
+ afs_get_addrlist(alist, afs_alist_trace_get_vlrotate_set);
read_unlock(&vlserver->lock);
memset(&vc->ac, 0, sizeof(vc->ac));
-
- if (!vc->ac.alist)
- vc->ac.alist = alist;
- else
- afs_put_addrlist(alist);
-
+ vc->ac.alist = alist;
vc->ac.index = -1;
iterate_address:
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index db7e94584e87..8dea7b56b75a 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -314,7 +314,7 @@ struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *vc,
alist = call->ret_alist;
afs_put_call(call);
if (vc->call_error) {
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_getaddru);
return ERR_PTR(vc->call_error);
}
return alist;
@@ -668,7 +668,7 @@ struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *vc,
alist = call->ret_alist;
afs_put_call(call);
if (vc->call_error) {
- afs_put_addrlist(alist);
+ afs_put_addrlist(alist, afs_alist_trace_put_getaddru);
return ERR_PTR(vc->call_error);
}
return alist;
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 597677acc6b1..ed91666ca4cc 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -202,6 +202,27 @@ enum yfs_cm_operation {
EM(afs_cell_trace_use_sbi, "USE sbi ") \
E_(afs_cell_trace_wait, "WAIT ")
+#define afs_alist_traces \
+ EM(afs_alist_trace_alloc, "ALLOC ") \
+ EM(afs_alist_trace_get_fsrotate_set, "GET fs-rot") \
+ EM(afs_alist_trace_get_make_call, "GET mkcall") \
+ EM(afs_alist_trace_get_probe, "GET probe ") \
+ EM(afs_alist_trace_get_vlrotate_set, "GET vl-rot") \
+ EM(afs_alist_trace_put_call, "PUT call ") \
+ EM(afs_alist_trace_put_end_cursor, "PUT endcur") \
+ EM(afs_alist_trace_put_getaddru, "PUT GtAdrU") \
+ EM(afs_alist_trace_put_parse_empty, "PUT p-empt") \
+ EM(afs_alist_trace_put_parse_error, "PUT p-err ") \
+ EM(afs_alist_trace_put_probe, "PUT probe ") \
+ EM(afs_alist_trace_put_retry_server, "PUT retry ") \
+ EM(afs_alist_trace_put_server, "PUT server") \
+ EM(afs_alist_trace_put_server_dup, "PUT sv-dup") \
+ EM(afs_alist_trace_put_server_oom, "PUT sv-oom") \
+ EM(afs_alist_trace_put_server_update, "PUT sv-upd") \
+ EM(afs_alist_trace_put_vlserver, "PUT vlsrvr") \
+ EM(afs_alist_trace_put_vlserver_old, "PUT vs-old") \
+ E_(afs_alist_trace_free, "FREE ")
+
#define afs_fs_operations \
EM(afs_FS_FetchData, "FS.FetchData") \
EM(afs_FS_FetchStatus, "FS.FetchStatus") \
@@ -420,6 +441,7 @@ enum yfs_cm_operation {
#define EM(a, b) a,
#define E_(a, b) a
+enum afs_alist_trace { afs_alist_traces } __mode(byte);
enum afs_call_trace { afs_call_traces } __mode(byte);
enum afs_cb_break_reason { afs_cb_break_reasons } __mode(byte);
enum afs_cell_trace { afs_cell_traces } __mode(byte);
@@ -443,6 +465,7 @@ enum afs_volume_trace { afs_volume_traces } __mode(byte);
#define EM(a, b) TRACE_DEFINE_ENUM(a);
#define E_(a, b) TRACE_DEFINE_ENUM(a);
+afs_alist_traces;
afs_call_traces;
afs_server_traces;
afs_cell_traces;
@@ -1330,6 +1353,30 @@ TRACE_EVENT(afs_cell,
__entry->active)
);
+TRACE_EVENT(afs_alist,
+ TP_PROTO(unsigned int alist_debug_id, int ref, enum afs_alist_trace reason),
+
+ TP_ARGS(alist_debug_id, ref, reason),
+
+ TP_STRUCT__entry(
+ __field(unsigned int, alist)
+ __field(int, ref)
+ __field(int, active)
+ __field(int, reason)
+ ),
+
+ TP_fast_assign(
+ __entry->alist = alist_debug_id;
+ __entry->ref = ref;
+ __entry->reason = reason;
+ ),
+
+ TP_printk("AL=%08x %s r=%d",
+ __entry->alist,
+ __print_symbolic(__entry->reason, afs_alist_traces),
+ __entry->ref)
+ );
+
#endif /* _TRACE_AFS_H */
/* This part must be outside protection */
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 20/41] afs: Rename some fields
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (18 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 19/41] afs: Add a tracepoint for struct afs_addr_list David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 21/41] afs: Use peer + service_id as call address David Howells
` (20 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Rename the ->index and ->untried fields of the afs_vl_cursor and
afs_operation struct to ->server_index and ->untried_servers to avoid
confusion with address iteration fields when those get folded in.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/internal.h | 8 ++++----
fs/afs/rotate.c | 36 ++++++++++++++++++------------------
fs/afs/vl_rotate.c | 32 ++++++++++++++++----------------
3 files changed, 38 insertions(+), 38 deletions(-)
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index d5a26cea2358..b3353fe63250 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -746,11 +746,11 @@ struct afs_vl_cursor {
struct afs_vlserver_list *server_list; /* Current server list (pins ref) */
struct afs_vlserver *server; /* Server on which this resides */
struct key *key; /* Key for the server */
- unsigned long untried; /* Bitmask of untried servers */
+ unsigned long untried_servers; /* Bitmask of untried servers */
struct afs_error cumul_error; /* Cumulative error */
s32 call_abort_code;
- short index; /* Current server */
short call_error; /* Error from single call */
+ short server_index; /* Current server */
unsigned short flags;
#define AFS_VL_CURSOR_STOP 0x0001 /* Set to cease iteration */
#define AFS_VL_CURSOR_RETRY 0x0002 /* Set to do a retry */
@@ -863,8 +863,8 @@ struct afs_operation {
struct afs_server_list *server_list; /* Current server list (pins ref) */
struct afs_server *server; /* Server we're using (ref pinned by server_list) */
struct afs_call *call;
- unsigned long untried; /* Bitmask of untried servers */
- short index; /* Current server */
+ unsigned long untried_servers; /* Bitmask of untried servers */
+ short server_index; /* Current server */
short nr_iterations; /* Number of server iterations */
bool call_responded; /* T if the current address responded */
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 1767c6e67184..fdf5e45e6a0a 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -32,8 +32,8 @@ static bool afs_start_fs_iteration(struct afs_operation *op,
lockdep_is_held(&op->volume->servers_lock)));
read_unlock(&op->volume->servers_lock);
- op->untried = (1UL << op->server_list->nr_servers) - 1;
- op->index = READ_ONCE(op->server_list->preferred);
+ op->untried_servers = (1UL << op->server_list->nr_servers) - 1;
+ op->server_index = READ_ONCE(op->server_list->preferred);
cb_server = vnode->cb_server;
if (cb_server) {
@@ -41,7 +41,7 @@ static bool afs_start_fs_iteration(struct afs_operation *op,
for (i = 0; i < op->server_list->nr_servers; i++) {
server = op->server_list->servers[i].server;
if (server == cb_server) {
- op->index = i;
+ op->server_index = i;
goto found_interest;
}
}
@@ -120,7 +120,7 @@ bool afs_select_fileserver(struct afs_operation *op)
_enter("OP=%x+%x,%llx,%lx[%d],%lx[%d],%d,%d",
op->debug_id, op->nr_iterations, op->volume->vid,
- op->untried, op->index,
+ op->untried_servers, op->server_index,
op->ac.tried, op->ac.index,
error, abort_code);
@@ -172,7 +172,7 @@ bool afs_select_fileserver(struct afs_operation *op)
}
write_lock(&op->volume->servers_lock);
- op->server_list->vnovol_mask |= 1 << op->index;
+ op->server_list->vnovol_mask |= 1 << op->server_index;
write_unlock(&op->volume->servers_lock);
set_bit(AFS_VOLUME_NEEDS_UPDATE, &op->volume->flags);
@@ -417,9 +417,9 @@ bool afs_select_fileserver(struct afs_operation *op)
_debug("__ VOL %llx __", op->volume->vid);
pick_server:
- _debug("pick [%lx]", op->untried);
+ _debug("pick [%lx]", op->untried_servers);
- error = afs_wait_for_fs_probes(op->server_list, op->untried);
+ error = afs_wait_for_fs_probes(op->server_list, op->untried_servers);
if (error < 0) {
afs_op_set_error(op, error);
goto failed;
@@ -429,40 +429,40 @@ bool afs_select_fileserver(struct afs_operation *op)
* callbacks, we stick with the server we're already using if we can.
*/
if (op->server) {
- _debug("server %u", op->index);
- if (test_bit(op->index, &op->untried))
+ _debug("server %u", op->server_index);
+ if (test_bit(op->server_index, &op->untried_servers))
goto selected_server;
op->server = NULL;
_debug("no server");
}
- op->index = -1;
+ op->server_index = -1;
rtt = UINT_MAX;
for (i = 0; i < op->server_list->nr_servers; i++) {
struct afs_server *s = op->server_list->servers[i].server;
- if (!test_bit(i, &op->untried) ||
+ if (!test_bit(i, &op->untried_servers) ||
!test_bit(AFS_SERVER_FL_RESPONDING, &s->flags))
continue;
if (s->probe.rtt <= rtt) {
- op->index = i;
+ op->server_index = i;
rtt = s->probe.rtt;
}
}
- if (op->index == -1)
+ if (op->server_index == -1)
goto no_more_servers;
selected_server:
- _debug("use %d", op->index);
- __clear_bit(op->index, &op->untried);
+ _debug("use %d", op->server_index);
+ __clear_bit(op->server_index, &op->untried_servers);
/* We're starting on a different fileserver from the list. We need to
* check it, create a callback intercept, find its address list and
* probe its capabilities before we use it.
*/
ASSERTCMP(op->ac.alist, ==, NULL);
- server = op->server_list->servers[op->index].server;
+ server = op->server_list->servers[op->server_index].server;
if (!afs_check_server_record(op, server))
goto failed;
@@ -504,7 +504,7 @@ bool afs_select_fileserver(struct afs_operation *op)
goto out_of_addresses;
_debug("address [%u] %u/%u %pISp",
- op->index, op->ac.index, op->ac.alist->nr_addrs,
+ op->server_index, op->ac.index, op->ac.alist->nr_addrs,
rxrpc_kernel_remote_addr(op->ac.alist->addrs[op->ac.index].peer));
op->call_responded = false;
@@ -579,7 +579,7 @@ void afs_dump_edestaddrreq(const struct afs_operation *op)
op->file[0].cb_break_before,
op->file[1].cb_break_before, op->flags, op->cumul_error.error);
pr_notice("OP: ut=%lx ix=%d ni=%u\n",
- op->untried, op->index, op->nr_iterations);
+ op->untried_servers, op->server_index, op->nr_iterations);
pr_notice("OP: call er=%d ac=%d r=%u\n",
op->call_error, op->call_abort_code, op->call_responded);
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index e8fbbeb551bb..f895eb94129e 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -78,8 +78,8 @@ static bool afs_start_vl_iteration(struct afs_vl_cursor *vc)
if (!vc->server_list->nr_servers)
return false;
- vc->untried = (1UL << vc->server_list->nr_servers) - 1;
- vc->index = -1;
+ vc->untried_servers = (1UL << vc->server_list->nr_servers) - 1;
+ vc->server_index = -1;
return true;
}
@@ -98,7 +98,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
vc->nr_iterations++;
_enter("%lx[%d],%lx[%d],%d,%d",
- vc->untried, vc->index,
+ vc->untried_servers, vc->server_index,
vc->ac.tried, vc->ac.index,
error, abort_code);
@@ -131,7 +131,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
/* The server went weird. */
afs_prioritise_error(&vc->cumul_error, -EREMOTEIO, abort_code);
//write_lock(&vc->cell->vl_servers_lock);
- //vc->server_list->weird_mask |= 1 << vc->index;
+ //vc->server_list->weird_mask |= 1 << vc->server_index;
//write_unlock(&vc->cell->vl_servers_lock);
goto next_server;
@@ -184,46 +184,46 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
}
pick_server:
- _debug("pick [%lx]", vc->untried);
+ _debug("pick [%lx]", vc->untried_servers);
- error = afs_wait_for_vl_probes(vc->server_list, vc->untried);
+ error = afs_wait_for_vl_probes(vc->server_list, vc->untried_servers);
if (error < 0) {
afs_prioritise_error(&vc->cumul_error, error, 0);
goto failed;
}
/* Pick the untried server with the lowest RTT. */
- vc->index = vc->server_list->preferred;
- if (test_bit(vc->index, &vc->untried))
+ vc->server_index = vc->server_list->preferred;
+ if (test_bit(vc->server_index, &vc->untried_servers))
goto selected_server;
- vc->index = -1;
+ vc->server_index = -1;
rtt = UINT_MAX;
for (i = 0; i < vc->server_list->nr_servers; i++) {
struct afs_vlserver *s = vc->server_list->servers[i].server;
- if (!test_bit(i, &vc->untried) ||
+ if (!test_bit(i, &vc->untried_servers) ||
!test_bit(AFS_VLSERVER_FL_RESPONDING, &s->flags))
continue;
if (s->probe.rtt <= rtt) {
- vc->index = i;
+ vc->server_index = i;
rtt = s->probe.rtt;
}
}
- if (vc->index == -1)
+ if (vc->server_index == -1)
goto no_more_servers;
selected_server:
- _debug("use %d", vc->index);
- __clear_bit(vc->index, &vc->untried);
+ _debug("use %d", vc->server_index);
+ __clear_bit(vc->server_index, &vc->untried_servers);
/* We're starting on a different vlserver from the list. We need to
* check it, find its address list and probe its capabilities before we
* use it.
*/
ASSERTCMP(vc->ac.alist, ==, NULL);
- vlserver = vc->server_list->servers[vc->index].server;
+ vlserver = vc->server_list->servers[vc->server_index].server;
vc->server = vlserver;
_debug("USING VLSERVER: %s", vlserver->name);
@@ -299,7 +299,7 @@ static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
pr_notice("DNS: src=%u st=%u lc=%x\n",
cell->dns_source, cell->dns_status, cell->dns_lookup_count);
pr_notice("VC: ut=%lx ix=%u ni=%hu fl=%hx err=%hd\n",
- vc->untried, vc->index, vc->nr_iterations, vc->flags,
+ vc->untried_servers, vc->server_index, vc->nr_iterations, vc->flags,
vc->cumul_error.error);
pr_notice("VC: call er=%d ac=%d r=%u\n",
vc->call_error, vc->call_abort_code, vc->call_responded);
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 21/41] afs: Use peer + service_id as call address
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (19 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 20/41] afs: Rename some fields David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 22/41] afs: Fold the afs_addr_cursor struct in David Howells
` (19 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Use the rxrpc_peer plus the service ID as the call address instead of
passing in a sockaddr_srx down to rxrpc. The peer record is obtained by
using rxrpc_kernel_get_peer(). This avoids the need to repeatedly look up
the peer and allows rxrpc to hold on to resources for it.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 10 +++-------
fs/afs/fs_probe.c | 4 ++--
fs/afs/fsclient.c | 14 +++++++++-----
fs/afs/internal.h | 18 +++++++++++-------
fs/afs/rxrpc.c | 12 ++++++------
fs/afs/server.c | 1 +
fs/afs/vl_list.c | 3 ++-
fs/afs/vl_probe.c | 4 ++--
fs/afs/vlclient.c | 14 ++++++++++++--
9 files changed, 48 insertions(+), 32 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index 41ef0c879239..032e6963c5d8 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -56,13 +56,12 @@ struct afs_addr_list *afs_get_addrlist(struct afs_addr_list *alist, enum afs_ali
/*
* Allocate an address list.
*/
-struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id)
+struct afs_addr_list *afs_alloc_addrlist(unsigned int nr)
{
struct afs_addr_list *alist;
- unsigned int i;
static atomic_t debug_id;
- _enter("%u,%u", nr, service_id);
+ _enter("%u", nr);
if (nr > AFS_MAX_ADDRESSES)
nr = AFS_MAX_ADDRESSES;
@@ -74,9 +73,6 @@ struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id)
refcount_set(&alist->usage, 1);
alist->max_addrs = nr;
alist->debug_id = atomic_inc_return(&debug_id);
-
- for (i = 0; i < nr; i++)
- alist->addrs[i].service_id = service_id;
trace_afs_alist(alist->debug_id, 1, afs_alist_trace_alloc);
return alist;
}
@@ -150,7 +146,7 @@ struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *net,
if (!vllist->servers[0].server)
goto error_vl;
- alist = afs_alloc_addrlist(nr, service);
+ alist = afs_alloc_addrlist(nr);
if (!alist)
goto error;
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index 18891492c0b4..337673e65f87 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -154,12 +154,12 @@ void afs_fileserver_probe_result(struct afs_call *call)
if (call->service_id == YFS_FS_SERVICE) {
server->probe.is_yfs = true;
set_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
- addr->service_id = call->service_id;
+ server->service_id = call->service_id;
} else {
server->probe.not_yfs = true;
if (!server->probe.is_yfs) {
clear_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
- addr->service_id = call->service_id;
+ server->service_id = call->service_id;
}
cap0 = ntohl(call->tmp);
if (cap0 & AFS3_VICED_CAPABILITY_64BITFILES)
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 2a56dea22519..2b64641b20a4 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -1620,7 +1620,9 @@ int afs_fs_give_up_all_callbacks(struct afs_net *net,
if (!call)
return -ENOMEM;
- call->key = key;
+ call->key = key;
+ call->peer = rxrpc_kernel_get_peer(ac->alist->addrs[ac->index].peer);
+ call->service_id = server->service_id;
/* marshall the parameters */
bp = call->request;
@@ -1722,10 +1724,12 @@ bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
if (!call)
return false;
- call->key = key;
- call->server = afs_use_server(server, afs_server_trace_get_caps);
- call->upgrade = true;
- call->async = true;
+ call->key = key;
+ call->server = afs_use_server(server, afs_server_trace_get_caps);
+ call->peer = rxrpc_kernel_get_peer(ac->alist->addrs[ac->index].peer);
+ call->service_id = server->service_id;
+ call->upgrade = true;
+ call->async = true;
call->max_lifespan = AFS_PROBE_MAX_LIFESPAN;
/* marshall the parameters */
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index b3353fe63250..b3ff6cd078de 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -74,7 +74,6 @@ enum afs_call_state {
struct afs_address {
struct rxrpc_peer *peer;
- u16 service_id;
short last_error; /* Last error from this address */
};
@@ -108,6 +107,7 @@ struct afs_call {
struct work_struct async_work; /* async I/O processor */
struct work_struct work; /* actual work processor */
struct rxrpc_call *rxcall; /* RxRPC call handle */
+ struct rxrpc_peer *peer; /* Remote endpoint */
struct key *key; /* security for this call */
struct afs_net *net; /* The network namespace */
struct afs_server *server; /* The fileserver record if fs op (pins ref) */
@@ -435,6 +435,7 @@ struct afs_vlserver {
#define AFS_VLSERVER_PROBE_LOCAL_FAILURE 0x08 /* A local failure prevented a probe */
} probe;
+ u16 service_id; /* Service ID we're using */
u16 port;
u16 name_len; /* Length of name */
char name[]; /* Server name, case-flattened */
@@ -527,6 +528,7 @@ struct afs_server {
refcount_t ref; /* Object refcount */
atomic_t active; /* Active user count */
u32 addr_version; /* Address list version */
+ u16 service_id; /* Service ID we're using. */
unsigned int rtt; /* Server's current RTT in uS */
unsigned int debug_id; /* Debugging ID for traces */
@@ -970,7 +972,7 @@ static inline bool afs_is_folio_dirty_mmapped(unsigned long priv)
* addr_list.c
*/
struct afs_addr_list *afs_get_addrlist(struct afs_addr_list *alist, enum afs_alist_trace reason);
-extern struct afs_addr_list *afs_alloc_addrlist(unsigned int nr, u16 service_id);
+extern struct afs_addr_list *afs_alloc_addrlist(unsigned int nr);
extern void afs_put_addrlist(struct afs_addr_list *alist, enum afs_alist_trace reason);
extern struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *,
const char *, size_t, char,
@@ -1317,11 +1319,13 @@ extern int afs_protocol_error(struct afs_call *, enum afs_eproto_cause);
static inline void afs_make_op_call(struct afs_operation *op, struct afs_call *call,
gfp_t gfp)
{
- op->call = call;
- op->type = call->type;
- call->op = op;
- call->key = op->key;
- call->intr = !(op->flags & AFS_OPERATION_UNINTR);
+ op->call = call;
+ op->type = call->type;
+ call->op = op;
+ call->key = op->key;
+ call->intr = !(op->flags & AFS_OPERATION_UNINTR);
+ call->peer = rxrpc_kernel_get_peer(op->ac.alist->addrs[op->ac.index].peer);
+ call->service_id = op->server->service_id;
afs_make_call(&op->ac, call, gfp);
}
diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
index ede7ab4b5dfe..9b7ef6309184 100644
--- a/fs/afs/rxrpc.c
+++ b/fs/afs/rxrpc.c
@@ -178,6 +178,8 @@ void afs_put_call(struct afs_call *call)
ASSERT(!work_pending(&call->async_work));
ASSERT(call->type->name != NULL);
+ rxrpc_kernel_put_peer(call->peer);
+
if (call->rxcall) {
rxrpc_kernel_shutdown_call(net->socket, call->rxcall);
rxrpc_kernel_put_call(net->socket, call->rxcall);
@@ -296,8 +298,6 @@ static void afs_notify_end_request_tx(struct sock *sock,
*/
void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
{
- struct afs_address *addr = &ac->alist->addrs[ac->index];
- struct rxrpc_peer *peer = addr->peer;
struct rxrpc_call *rxcall;
struct msghdr msg;
struct kvec iov[1];
@@ -305,7 +305,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
s64 tx_total_len;
int ret;
- _enter(",{%pISp},", rxrpc_kernel_remote_addr(addr->peer));
+ _enter(",{%pISp+%u},", rxrpc_kernel_remote_addr(call->peer), call->service_id);
ASSERT(call->type != NULL);
ASSERT(call->type->name != NULL);
@@ -334,7 +334,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
}
/* create a call */
- rxcall = rxrpc_kernel_begin_call(call->net->socket, peer, call->key,
+ rxcall = rxrpc_kernel_begin_call(call->net->socket, call->peer, call->key,
(unsigned long)call,
tx_total_len,
call->max_lifespan,
@@ -342,7 +342,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
(call->async ?
afs_wake_up_async_call :
afs_wake_up_call_waiter),
- addr->service_id,
+ call->service_id,
call->upgrade,
(call->intr ? RXRPC_PREINTERRUPTIBLE :
RXRPC_UNINTERRUPTIBLE),
@@ -462,7 +462,7 @@ static void afs_log_error(struct afs_call *call, s32 remote_abort)
max = m + 1;
pr_notice("kAFS: Peer reported %s failure on %s [%pISp]\n",
msg, call->type->name,
- rxrpc_kernel_remote_addr(call->alist->addrs[call->addr_ix].peer));
+ rxrpc_kernel_remote_addr(call->peer));
}
}
diff --git a/fs/afs/server.c b/fs/afs/server.c
index e4c369132935..185d67271b3c 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -214,6 +214,7 @@ static struct afs_server *afs_alloc_server(struct afs_cell *cell,
spin_lock_init(&server->probe_lock);
server->cell = cell;
server->rtt = UINT_MAX;
+ server->service_id = FS_SERVICE;
afs_inc_servers_outstanding(net);
trace_afs_server(server->debug_id, 1, 1, afs_server_trace_alloc);
diff --git a/fs/afs/vl_list.c b/fs/afs/vl_list.c
index 3a2875933261..5c4cd71caccf 100644
--- a/fs/afs/vl_list.c
+++ b/fs/afs/vl_list.c
@@ -23,6 +23,7 @@ struct afs_vlserver *afs_alloc_vlserver(const char *name, size_t name_len,
spin_lock_init(&vlserver->probe_lock);
vlserver->rtt = UINT_MAX;
vlserver->name_len = name_len;
+ vlserver->service_id = VL_SERVICE;
vlserver->port = port;
memcpy(vlserver->name, name, name_len);
}
@@ -92,7 +93,7 @@ static struct afs_addr_list *afs_extract_vl_addrs(struct afs_net *net,
const u8 *b = *_b;
int ret = -EINVAL;
- alist = afs_alloc_addrlist(nr_addrs, VL_SERVICE);
+ alist = afs_alloc_addrlist(nr_addrs);
if (!alist)
return ERR_PTR(-ENOMEM);
if (nr_addrs == 0)
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index 138f5715619d..d9a99ba9fc78 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -107,12 +107,12 @@ void afs_vlserver_probe_result(struct afs_call *call)
if (call->service_id == YFS_VL_SERVICE) {
server->probe.flags |= AFS_VLSERVER_PROBE_IS_YFS;
set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
- addr->service_id = call->service_id;
+ server->service_id = call->service_id;
} else {
server->probe.flags |= AFS_VLSERVER_PROBE_NOT_YFS;
if (!(server->probe.flags & AFS_VLSERVER_PROBE_IS_YFS)) {
clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
- addr->service_id = call->service_id;
+ server->service_id = call->service_id;
}
}
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index 8dea7b56b75a..4bf98a38c3a1 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -149,6 +149,8 @@ struct afs_vldb_entry *afs_vl_get_entry_by_name_u(struct afs_vl_cursor *vc,
call->key = vc->key;
call->ret_vldb = entry;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
+ call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->service_id = vc->server->service_id;
/* Marshall the parameters */
bp = call->request;
@@ -211,7 +213,7 @@ static int afs_deliver_vl_get_addrs_u(struct afs_call *call)
count = ntohl(*bp);
nentries = min(nentries, count);
- alist = afs_alloc_addrlist(nentries, FS_SERVICE);
+ alist = afs_alloc_addrlist(nentries);
if (!alist)
return -ENOMEM;
alist->version = uniquifier;
@@ -288,6 +290,8 @@ struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *vc,
call->key = vc->key;
call->ret_alist = NULL;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
+ call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->service_id = vc->server->service_id;
/* Marshall the parameters */
bp = call->request;
@@ -407,6 +411,8 @@ struct afs_call *afs_vl_get_capabilities(struct afs_net *net,
call->key = key;
call->vlserver = afs_get_vlserver(server);
call->server_index = server_index;
+ call->peer = rxrpc_kernel_get_peer(ac->alist->addrs[ac->index].peer);
+ call->service_id = server->service_id;
call->upgrade = true;
call->async = true;
call->max_lifespan = AFS_PROBE_MAX_LIFESPAN;
@@ -462,7 +468,7 @@ static int afs_deliver_yfsvl_get_endpoints(struct afs_call *call)
if (call->count > YFS_MAXENDPOINTS)
return afs_protocol_error(call, afs_eproto_yvl_fsendpt_num);
- alist = afs_alloc_addrlist(call->count, FS_SERVICE);
+ alist = afs_alloc_addrlist(call->count);
if (!alist)
return -ENOMEM;
alist->version = uniquifier;
@@ -652,6 +658,8 @@ struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *vc,
call->key = vc->key;
call->ret_alist = NULL;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
+ call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->service_id = vc->server->service_id;
/* Marshall the parameters */
bp = call->request;
@@ -769,6 +777,8 @@ char *afs_yfsvl_get_cell_name(struct afs_vl_cursor *vc)
call->key = vc->key;
call->ret_str = NULL;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
+ call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->service_id = vc->server->service_id;
/* marshall the parameters */
bp = call->request;
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 22/41] afs: Fold the afs_addr_cursor struct in
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (20 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 21/41] afs: Use peer + service_id as call address David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 23/41] rxrpc: Create a procfile to display outstanding clien conn bundles David Howells
` (18 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Fold the afs_addr_cursor struct into the afs_operation struct and the
afs_vl_cursor struct and fold its operations into their callers also.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 53 -----------------------
fs/afs/fs_operation.c | 21 +++++++--
fs/afs/fs_probe.c | 41 ++++++++----------
fs/afs/fsclient.c | 31 +++++++++-----
fs/afs/internal.h | 58 ++++++++++++-------------
fs/afs/rotate.c | 71 ++++++++++++++++++------------
fs/afs/rxrpc.c | 13 ++----
fs/afs/server.c | 6 +--
fs/afs/vl_probe.c | 23 +++++-----
fs/afs/vl_rotate.c | 88 ++++++++++++++++++++++++++++----------
fs/afs/vlclient.c | 34 ++++++++-------
include/trace/events/afs.h | 18 ++++++--
12 files changed, 243 insertions(+), 214 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index 032e6963c5d8..18c286efa3a5 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -375,56 +375,3 @@ int afs_merge_fs_addr6(struct afs_net *net, struct afs_addr_list *alist,
alist->nr_addrs++;
return 0;
}
-
-/*
- * Get an address to try.
- */
-bool afs_iterate_addresses(struct afs_addr_cursor *ac)
-{
- unsigned long set, failed;
- int index;
-
- if (!ac->alist)
- return false;
-
- set = ac->alist->responded;
- failed = ac->alist->probe_failed;
- _enter("%lx-%lx-%lx,%d", set, failed, ac->tried, ac->index);
-
- ac->nr_iterations++;
-
- set &= ~(failed | ac->tried);
-
- if (!set)
- return false;
-
- index = READ_ONCE(ac->alist->preferred);
- if (test_bit(index, &set))
- goto selected;
-
- index = __ffs(set);
-
-selected:
- ac->index = index;
- set_bit(index, &ac->tried);
- ac->call_responded = false;
- return true;
-}
-
-/*
- * Release an address list cursor.
- */
-void afs_end_cursor(struct afs_addr_cursor *ac)
-{
- struct afs_addr_list *alist;
-
- alist = ac->alist;
- if (alist) {
- if (ac->call_responded &&
- ac->index != alist->preferred &&
- test_bit(ac->alist->preferred, &ac->tried))
- WRITE_ONCE(alist->preferred, ac->index);
- afs_put_addrlist(alist, afs_alist_trace_put_end_cursor);
- ac->alist = NULL;
- }
-}
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index cebe4fad8192..00e22259be36 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -179,6 +179,7 @@ void afs_wait_for_operation(struct afs_operation *op)
_enter("");
while (afs_select_fileserver(op)) {
+ op->call_responded = false;
op->call_error = 0;
op->call_abort_code = 0;
op->cb_s_break = op->server->cb_s_break;
@@ -191,17 +192,19 @@ void afs_wait_for_operation(struct afs_operation *op)
op->call_error = -ENOTSUPP;
if (op->call) {
- afs_wait_for_call_to_complete(op->call, &op->ac);
+ afs_wait_for_call_to_complete(op->call);
op->call_abort_code = op->call->abort_code;
op->call_error = op->call->error;
op->call_responded = op->call->responded;
- op->ac.call_responded = true;
- WRITE_ONCE(op->ac.alist->addrs[op->ac.index].last_error,
+ WRITE_ONCE(op->alist->addrs[op->addr_index].last_error,
op->call_error);
afs_put_call(op->call);
}
}
+ if (op->call_responded)
+ set_bit(AFS_SERVER_FL_RESPONDING, &op->server->flags);
+
if (!afs_op_error(op)) {
_debug("success");
op->ops->success(op);
@@ -227,6 +230,7 @@ void afs_wait_for_operation(struct afs_operation *op)
*/
int afs_put_operation(struct afs_operation *op)
{
+ struct afs_addr_list *alist;
int i, ret = afs_op_error(op);
_enter("op=%08x,%d", op->debug_id, ret);
@@ -249,7 +253,16 @@ int afs_put_operation(struct afs_operation *op)
kfree(op->more_files);
}
- afs_end_cursor(&op->ac);
+ alist = op->alist;
+ if (alist) {
+ if (op->call_responded &&
+ op->addr_index != alist->preferred &&
+ test_bit(alist->preferred, &op->addr_tried))
+ WRITE_ONCE(alist->preferred, op->addr_index);
+ afs_put_addrlist(alist, afs_alist_trace_put_operation);
+ op->alist = NULL;
+ }
+
afs_put_serverlist(op->net, op->server_list);
afs_put_volume(op->net, op->volume, afs_volume_trace_put_put_op);
key_put(op->key);
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index 337673e65f87..aef16ac3f577 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -74,11 +74,9 @@ static void afs_done_one_fs_probe(struct afs_net *net, struct afs_server *server
*/
static void afs_fs_probe_not_done(struct afs_net *net,
struct afs_server *server,
- struct afs_addr_cursor *ac)
+ struct afs_addr_list *alist,
+ int index)
{
- struct afs_addr_list *alist = ac->alist;
- unsigned int index = ac->index;
-
_enter("");
trace_afs_io_error(0, -ENOMEM, afs_io_error_fs_probe_fail);
@@ -100,10 +98,10 @@ static void afs_fs_probe_not_done(struct afs_net *net,
*/
void afs_fileserver_probe_result(struct afs_call *call)
{
- struct afs_addr_list *alist = call->alist;
- struct afs_address *addr = &alist->addrs[call->addr_ix];
+ struct afs_addr_list *alist = call->probe_alist;
+ struct afs_address *addr = &alist->addrs[call->probe_index];
struct afs_server *server = call->server;
- unsigned int index = call->addr_ix;
+ unsigned int index = call->probe_index;
unsigned int rtt_us = 0, cap0;
int ret = call->error;
@@ -196,37 +194,36 @@ void afs_fileserver_probe_result(struct afs_call *call)
void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
struct key *key, bool all)
{
- struct afs_addr_cursor ac = {
- .index = 0,
- };
+ struct afs_addr_list *alist;
+ unsigned int index;
_enter("%pU", &server->uuid);
read_lock(&server->fs_lock);
- ac.alist = rcu_dereference_protected(server->addresses,
- lockdep_is_held(&server->fs_lock));
- afs_get_addrlist(ac.alist, afs_alist_trace_get_probe);
+ alist = rcu_dereference_protected(server->addresses,
+ lockdep_is_held(&server->fs_lock));
+ afs_get_addrlist(alist, afs_alist_trace_get_probe);
read_unlock(&server->fs_lock);
server->probed_at = jiffies;
- atomic_set(&server->probe_outstanding, all ? ac.alist->nr_addrs : 1);
+ atomic_set(&server->probe_outstanding, all ? alist->nr_addrs : 1);
memset(&server->probe, 0, sizeof(server->probe));
server->probe.rtt = UINT_MAX;
- ac.index = ac.alist->preferred;
- if (ac.index < 0 || ac.index >= ac.alist->nr_addrs)
+ index = alist->preferred;
+ if (index < 0 || index >= alist->nr_addrs)
all = true;
if (all) {
- for (ac.index = 0; ac.index < ac.alist->nr_addrs; ac.index++)
- if (!afs_fs_get_capabilities(net, server, &ac, key))
- afs_fs_probe_not_done(net, server, &ac);
+ for (index = 0; index < alist->nr_addrs; index++)
+ if (!afs_fs_get_capabilities(net, server, alist, index, key))
+ afs_fs_probe_not_done(net, server, alist, index);
} else {
- if (!afs_fs_get_capabilities(net, server, &ac, key))
- afs_fs_probe_not_done(net, server, &ac);
+ if (!afs_fs_get_capabilities(net, server, alist, index, key))
+ afs_fs_probe_not_done(net, server, alist, index);
}
- afs_put_addrlist(ac.alist, afs_alist_trace_put_probe);
+ afs_put_addrlist(alist, afs_alist_trace_put_probe);
}
/*
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 2b64641b20a4..4f98b43b0dde 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -1605,10 +1605,8 @@ static const struct afs_call_type afs_RXFSGiveUpAllCallBacks = {
/*
* Flush all the callbacks we have on a server.
*/
-int afs_fs_give_up_all_callbacks(struct afs_net *net,
- struct afs_server *server,
- struct afs_addr_cursor *ac,
- struct key *key)
+int afs_fs_give_up_all_callbacks(struct afs_net *net, struct afs_server *server,
+ struct afs_address *addr, struct key *key)
{
struct afs_call *call;
__be32 *bp;
@@ -1621,7 +1619,7 @@ int afs_fs_give_up_all_callbacks(struct afs_net *net,
return -ENOMEM;
call->key = key;
- call->peer = rxrpc_kernel_get_peer(ac->alist->addrs[ac->index].peer);
+ call->peer = rxrpc_kernel_get_peer(addr->peer);
call->service_id = server->service_id;
/* marshall the parameters */
@@ -1629,9 +1627,11 @@ int afs_fs_give_up_all_callbacks(struct afs_net *net,
*bp++ = htonl(FSGIVEUPALLCALLBACKS);
call->server = afs_use_server(server, afs_server_trace_give_up_cb);
- afs_make_call(ac, call, GFP_NOFS);
- afs_wait_for_call_to_complete(call, ac);
+ afs_make_call(call, GFP_NOFS);
+ afs_wait_for_call_to_complete(call);
ret = call->error;
+ if (call->responded)
+ set_bit(AFS_SERVER_FL_RESPONDING, &server->flags);
afs_put_call(call);
return ret;
}
@@ -1695,6 +1695,12 @@ static int afs_deliver_fs_get_capabilities(struct afs_call *call)
return 0;
}
+static void afs_fs_get_capabilities_destructor(struct afs_call *call)
+{
+ afs_put_addrlist(call->probe_alist, afs_alist_trace_put_getcaps);
+ afs_flat_call_destructor(call);
+}
+
/*
* FS.GetCapabilities operation type
*/
@@ -1703,7 +1709,7 @@ static const struct afs_call_type afs_RXFSGetCapabilities = {
.op = afs_FS_GetCapabilities,
.deliver = afs_deliver_fs_get_capabilities,
.done = afs_fileserver_probe_result,
- .destructor = afs_flat_call_destructor,
+ .destructor = afs_fs_get_capabilities_destructor,
};
/*
@@ -1713,7 +1719,8 @@ static const struct afs_call_type afs_RXFSGetCapabilities = {
* ->done() - otherwise we return false to indicate we didn't even try.
*/
bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
- struct afs_addr_cursor *ac, struct key *key)
+ struct afs_addr_list *alist, unsigned int addr_index,
+ struct key *key)
{
struct afs_call *call;
__be32 *bp;
@@ -1726,7 +1733,9 @@ bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
call->key = key;
call->server = afs_use_server(server, afs_server_trace_get_caps);
- call->peer = rxrpc_kernel_get_peer(ac->alist->addrs[ac->index].peer);
+ call->peer = rxrpc_kernel_get_peer(alist->addrs[addr_index].peer);
+ call->probe_alist = afs_get_addrlist(alist, afs_alist_trace_get_getcaps);
+ call->probe_index = addr_index;
call->service_id = server->service_id;
call->upgrade = true;
call->async = true;
@@ -1737,7 +1746,7 @@ bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
*bp++ = htonl(FSGETCAPABILITIES);
trace_afs_make_fs_call(call, NULL);
- afs_make_call(ac, call, GFP_NOFS);
+ afs_make_call(call, GFP_NOFS);
afs_put_call(call);
return true;
}
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index b3ff6cd078de..b6659136959a 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -102,7 +102,6 @@ struct afs_addr_list {
*/
struct afs_call {
const struct afs_call_type *type; /* type of call */
- struct afs_addr_list *alist; /* Address is alist[addr_ix] */
wait_queue_head_t waitq; /* processes awaiting completion */
struct work_struct async_work; /* async I/O processor */
struct work_struct work; /* actual work processor */
@@ -123,6 +122,10 @@ struct afs_call {
};
void *buffer; /* reply receive buffer */
union {
+ struct {
+ struct afs_addr_list *probe_alist;
+ unsigned char probe_index; /* Address in ->probe_alist */
+ };
struct afs_addr_list *ret_alist;
struct afs_vldb_entry *ret_vldb;
char *ret_str;
@@ -139,7 +142,6 @@ struct afs_call {
unsigned reply_max; /* maximum size of reply */
unsigned count2; /* count used in unmarshalling */
unsigned char unmarshall; /* unmarshalling phase */
- unsigned char addr_ix; /* Address in ->alist */
bool drop_ref; /* T if need to drop ref for incoming call */
bool need_attention; /* T if RxRPC poked us */
bool async; /* T if asynchronous */
@@ -728,31 +730,23 @@ struct afs_error {
bool aborted; /* T if ->error is from an abort */
};
-/*
- * Cursor for iterating over a server's address list.
- */
-struct afs_addr_cursor {
- struct afs_addr_list *alist; /* Current address list (pins ref) */
- unsigned long tried; /* Tried addresses */
- signed char index; /* Current address */
- unsigned short nr_iterations; /* Number of address iterations */
- bool call_responded;
-};
-
/*
* Cursor for iterating over a set of volume location servers.
*/
struct afs_vl_cursor {
- struct afs_addr_cursor ac;
struct afs_cell *cell; /* The cell we're querying */
struct afs_vlserver_list *server_list; /* Current server list (pins ref) */
struct afs_vlserver *server; /* Server on which this resides */
+ struct afs_addr_list *alist; /* Current address list (pins ref) */
struct key *key; /* Key for the server */
unsigned long untried_servers; /* Bitmask of untried servers */
+ unsigned long addr_tried; /* Tried addresses */
struct afs_error cumul_error; /* Cumulative error */
+ unsigned int debug_id;
s32 call_abort_code;
short call_error; /* Error from single call */
short server_index; /* Current server */
+ signed char addr_index; /* Current address */
unsigned short flags;
#define AFS_VL_CURSOR_STOP 0x0001 /* Set to cease iteration */
#define AFS_VL_CURSOR_RETRY 0x0002 /* Set to do a retry */
@@ -811,8 +805,6 @@ struct afs_operation {
struct timespec64 ctime; /* Change time to set */
struct afs_error cumul_error; /* Cumulative error */
short nr_files; /* Number of entries in file[], more_files */
- short call_error; /* Error from single call */
- s32 call_abort_code; /* Abort code from single call */
unsigned int debug_id;
unsigned int cb_v_break; /* Volume break counter before op */
@@ -861,16 +853,19 @@ struct afs_operation {
};
/* Fileserver iteration state */
- struct afs_addr_cursor ac;
struct afs_server_list *server_list; /* Current server list (pins ref) */
struct afs_server *server; /* Server we're using (ref pinned by server_list) */
+ struct afs_addr_list *alist; /* Current address list (pins ref) */
struct afs_call *call;
unsigned long untried_servers; /* Bitmask of untried servers */
+ unsigned long addr_tried; /* Tried addresses */
+ s32 call_abort_code; /* Abort code from single call */
+ short call_error; /* Error from single call */
short server_index; /* Current server */
short nr_iterations; /* Number of server iterations */
+ signed char addr_index; /* Current address */
bool call_responded; /* T if the current address responded */
-
unsigned int flags;
#define AFS_OPERATION_STOP 0x0001 /* Set to cease iteration */
#define AFS_OPERATION_VBUSY 0x0002 /* Set if seen VBUSY */
@@ -980,8 +975,6 @@ extern struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *,
bool afs_addr_list_same(const struct afs_addr_list *a,
const struct afs_addr_list *b);
extern struct afs_vlserver_list *afs_dns_query(struct afs_cell *, time64_t *);
-extern bool afs_iterate_addresses(struct afs_addr_cursor *);
-extern void afs_end_cursor(struct afs_addr_cursor *ac);
extern int afs_merge_fs_addr4(struct afs_net *net, struct afs_addr_list *addr,
__be32 xdr, u16 port);
@@ -1122,10 +1115,11 @@ extern void afs_fs_get_volume_status(struct afs_operation *);
extern void afs_fs_set_lock(struct afs_operation *);
extern void afs_fs_extend_lock(struct afs_operation *);
extern void afs_fs_release_lock(struct afs_operation *);
-extern int afs_fs_give_up_all_callbacks(struct afs_net *, struct afs_server *,
- struct afs_addr_cursor *, struct key *);
-extern bool afs_fs_get_capabilities(struct afs_net *, struct afs_server *,
- struct afs_addr_cursor *, struct key *);
+int afs_fs_give_up_all_callbacks(struct afs_net *net, struct afs_server *server,
+ struct afs_address *addr, struct key *key);
+bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
+ struct afs_addr_list *alist, unsigned int addr_index,
+ struct key *key);
extern void afs_fs_inline_bulk_status(struct afs_operation *);
struct afs_acl {
@@ -1305,8 +1299,8 @@ extern int __net_init afs_open_socket(struct afs_net *);
extern void __net_exit afs_close_socket(struct afs_net *);
extern void afs_charge_preallocation(struct work_struct *);
extern void afs_put_call(struct afs_call *);
-extern void afs_make_call(struct afs_addr_cursor *, struct afs_call *, gfp_t);
-void afs_wait_for_call_to_complete(struct afs_call *call, struct afs_addr_cursor *ac);
+void afs_make_call(struct afs_call *call, gfp_t gfp);
+void afs_wait_for_call_to_complete(struct afs_call *call);
extern struct afs_call *afs_alloc_flat_call(struct afs_net *,
const struct afs_call_type *,
size_t, size_t);
@@ -1324,9 +1318,9 @@ static inline void afs_make_op_call(struct afs_operation *op, struct afs_call *c
call->op = op;
call->key = op->key;
call->intr = !(op->flags & AFS_OPERATION_UNINTR);
- call->peer = rxrpc_kernel_get_peer(op->ac.alist->addrs[op->ac.index].peer);
+ call->peer = rxrpc_kernel_get_peer(op->alist->addrs[op->addr_index].peer);
call->service_id = op->server->service_id;
- afs_make_call(&op->ac, call, gfp);
+ afs_make_call(call, gfp);
}
static inline void afs_extract_begin(struct afs_call *call, void *buf, size_t size)
@@ -1492,8 +1486,12 @@ extern void afs_fs_exit(void);
extern struct afs_vldb_entry *afs_vl_get_entry_by_name_u(struct afs_vl_cursor *,
const char *, int);
extern struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *, const uuid_t *);
-extern struct afs_call *afs_vl_get_capabilities(struct afs_net *, struct afs_addr_cursor *,
- struct key *, struct afs_vlserver *, unsigned int);
+struct afs_call *afs_vl_get_capabilities(struct afs_net *net,
+ struct afs_addr_list *alist,
+ unsigned int addr_index,
+ struct key *key,
+ struct afs_vlserver *server,
+ unsigned int server_index);
extern struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *, const uuid_t *);
extern char *afs_yfsvl_get_cell_name(struct afs_vl_cursor *);
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index fdf5e45e6a0a..2ad119d632fa 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -109,19 +109,20 @@ static bool afs_sleep_and_retry(struct afs_operation *op)
*/
bool afs_select_fileserver(struct afs_operation *op)
{
- struct afs_addr_list *alist;
+ struct afs_addr_list *alist = op->alist;
struct afs_server *server;
struct afs_vnode *vnode = op->file[0].vnode;
+ unsigned long set;
unsigned int rtt;
s32 abort_code = op->call_abort_code;
- int error = op->call_error, i;
+ int error = op->call_error, addr_index, i;
op->nr_iterations++;
- _enter("OP=%x+%x,%llx,%lx[%d],%lx[%d],%d,%d",
+ _enter("OP=%x+%x,%llx,%u{%lx},%u{%lx},%d,%d",
op->debug_id, op->nr_iterations, op->volume->vid,
- op->untried_servers, op->server_index,
- op->ac.tried, op->ac.index,
+ op->server_index, op->untried_servers,
+ op->addr_index, op->addr_tried,
error, abort_code);
if (op->flags & AFS_OPERATION_STOP) {
@@ -396,12 +397,14 @@ bool afs_select_fileserver(struct afs_operation *op)
restart_from_beginning:
_debug("restart");
- afs_end_cursor(&op->ac);
+ afs_put_addrlist(alist, afs_alist_trace_put_restart_rotate);
+ alist = op->alist = NULL;
op->server = NULL;
afs_put_serverlist(op->net, op->server_list);
op->server_list = NULL;
start:
_debug("start");
+ ASSERTCMP(alist, ==, NULL);
/* See if we need to do an update of the volume record. Note that the
* volume may have moved or even have been deleted.
*/
@@ -418,6 +421,7 @@ bool afs_select_fileserver(struct afs_operation *op)
pick_server:
_debug("pick [%lx]", op->untried_servers);
+ ASSERTCMP(alist, ==, NULL);
error = afs_wait_for_fs_probes(op->server_list, op->untried_servers);
if (error < 0) {
@@ -461,7 +465,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* check it, create a callback intercept, find its address list and
* probe its capabilities before we use it.
*/
- ASSERTCMP(op->ac.alist, ==, NULL);
+ ASSERTCMP(alist, ==, NULL);
server = op->server_list->servers[op->server_index].server;
if (!afs_check_server_record(op, server))
@@ -482,32 +486,34 @@ bool afs_select_fileserver(struct afs_operation *op)
read_lock(&server->fs_lock);
alist = rcu_dereference_protected(server->addresses,
lockdep_is_held(&server->fs_lock));
- afs_get_addrlist(alist, afs_alist_trace_get_fsrotate_set);
+ op->alist = afs_get_addrlist(alist, afs_alist_trace_get_fsrotate_set);
read_unlock(&server->fs_lock);
retry_server:
- memset(&op->ac, 0, sizeof(op->ac));
-
- if (!op->ac.alist)
- op->ac.alist = alist;
- else
- afs_put_addrlist(alist, afs_alist_trace_put_retry_server);
-
- op->ac.index = -1;
+ op->addr_tried = 0;
+ op->addr_index = -1;
iterate_address:
- ASSERT(op->ac.alist);
/* Iterate over the current server's address list to try and find an
* address on which it will respond to us.
*/
- if (!afs_iterate_addresses(&op->ac))
+ set = READ_ONCE(alist->responded);
+ set &= ~(READ_ONCE(alist->probe_failed) | op->addr_tried);
+ if (!set)
goto out_of_addresses;
- _debug("address [%u] %u/%u %pISp",
- op->server_index, op->ac.index, op->ac.alist->nr_addrs,
- rxrpc_kernel_remote_addr(op->ac.alist->addrs[op->ac.index].peer));
+ addr_index = READ_ONCE(alist->preferred);
+ if (!test_bit(addr_index, &set))
+ addr_index = __ffs(set);
+
+ op->addr_index = addr_index;
+ set_bit(addr_index, &op->addr_tried);
+ op->alist = alist;
op->call_responded = false;
+ _debug("address [%u] %u/%u %pISp",
+ op->server_index, addr_index, alist->nr_addrs,
+ rxrpc_kernel_remote_addr(alist->addrs[op->addr_index].peer));
_leave(" = t");
return true;
@@ -517,7 +523,6 @@ bool afs_select_fileserver(struct afs_operation *op)
*/
afs_probe_fileserver(op->net, op->server);
if (op->flags & AFS_OPERATION_RETRY_SERVER) {
- alist = op->ac.alist;
error = afs_wait_for_one_fs_probe(
op->server, !(op->flags & AFS_OPERATION_UNINTR));
switch (error) {
@@ -535,7 +540,13 @@ bool afs_select_fileserver(struct afs_operation *op)
next_server:
_debug("next");
- afs_end_cursor(&op->ac);
+ ASSERT(alist);
+ if (op->call_responded &&
+ op->addr_index != READ_ONCE(alist->preferred) &&
+ test_bit(alist->preferred, &op->addr_tried))
+ WRITE_ONCE(alist->preferred, op->addr_index);
+ afs_put_addrlist(alist, afs_alist_trace_put_next_server);
+ alist = op->alist = NULL;
goto pick_server;
no_more_servers:
@@ -555,7 +566,14 @@ bool afs_select_fileserver(struct afs_operation *op)
failed:
op->flags |= AFS_OPERATION_STOP;
- afs_end_cursor(&op->ac);
+ if (alist) {
+ if (op->call_responded &&
+ op->addr_index != READ_ONCE(alist->preferred) &&
+ test_bit(alist->preferred, &op->addr_tried))
+ WRITE_ONCE(alist->preferred, op->addr_index);
+ afs_put_addrlist(alist, afs_alist_trace_put_op_failed);
+ op->alist = NULL;
+ }
_leave(" = f [failed %d]", afs_op_error(op));
return false;
}
@@ -600,13 +618,12 @@ void afs_dump_edestaddrreq(const struct afs_operation *op)
a->preferred);
pr_notice("FC: - R=%lx F=%lx\n",
a->responded, a->probe_failed);
- if (a == op->ac.alist)
+ if (a == op->alist)
pr_notice("FC: - current\n");
}
}
}
- pr_notice("AC: t=%lx ax=%u ni=%u\n",
- op->ac.tried, op->ac.index, op->ac.nr_iterations);
+ pr_notice("AC: t=%lx ax=%u\n", op->addr_tried, op->addr_index);
rcu_read_unlock();
}
diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c
index 9b7ef6309184..568494d90ba9 100644
--- a/fs/afs/rxrpc.c
+++ b/fs/afs/rxrpc.c
@@ -189,7 +189,6 @@ void afs_put_call(struct afs_call *call)
call->type->destructor(call);
afs_unuse_server_notime(call->net, call->server, afs_server_trace_put_call);
- afs_put_addrlist(call->alist, afs_alist_trace_put_call);
kfree(call->request);
trace_afs_call(call->debug_id, afs_call_trace_free, 0, o,
@@ -296,7 +295,7 @@ static void afs_notify_end_request_tx(struct sock *sock,
* Initiate a call and synchronously queue up the parameters for dispatch. Any
* error is stored into the call struct, which the caller must check for.
*/
-void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
+void afs_make_call(struct afs_call *call, gfp_t gfp)
{
struct rxrpc_call *rxcall;
struct msghdr msg;
@@ -314,9 +313,6 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
call, call->type->name, key_serial(call->key),
atomic_read(&call->net->nr_outstanding_calls));
- call->addr_ix = ac->index;
- call->alist = afs_get_addrlist(ac->alist, afs_alist_trace_get_make_call);
-
/* Work out the length we're going to transmit. This is awkward for
* calls such as FS.StoreData where there's an extra injection of data
* after the initial fixed part.
@@ -392,7 +388,7 @@ void afs_make_call(struct afs_addr_cursor *ac, struct afs_call *call, gfp_t gfp)
/* Note that at this point, we may have received the reply or an abort
* - and an asynchronous call may already have completed.
*
- * afs_wait_for_call_to_complete(call, ac)
+ * afs_wait_for_call_to_complete(call)
* must be called to synchronously clean up.
*/
return;
@@ -577,7 +573,7 @@ static void afs_deliver_to_call(struct afs_call *call)
/*
* Wait synchronously for a call to complete.
*/
-void afs_wait_for_call_to_complete(struct afs_call *call, struct afs_addr_cursor *ac)
+void afs_wait_for_call_to_complete(struct afs_call *call)
{
bool rxrpc_complete = false;
@@ -627,9 +623,6 @@ void afs_wait_for_call_to_complete(struct afs_call *call, struct afs_addr_cursor
afs_set_call_complete(call, -EINTR, 0);
}
}
-
- if (call->error == 0 || call->error == -ECONNABORTED)
- call->responded = true;
}
/*
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 185d67271b3c..2c46f65b6bd6 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -435,12 +435,8 @@ static void __afs_put_server(struct afs_net *net, struct afs_server *server)
static void afs_give_up_callbacks(struct afs_net *net, struct afs_server *server)
{
struct afs_addr_list *alist = rcu_access_pointer(server->addresses);
- struct afs_addr_cursor ac = {
- .alist = alist,
- .index = alist->preferred,
- };
- afs_fs_give_up_all_callbacks(net, server, &ac, NULL);
+ afs_fs_give_up_all_callbacks(net, server, &alist->addrs[alist->preferred], NULL);
}
/*
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index d9a99ba9fc78..f868ae5d40e5 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -46,12 +46,12 @@ static void afs_done_one_vl_probe(struct afs_vlserver *server, bool wake_up)
*/
void afs_vlserver_probe_result(struct afs_call *call)
{
- struct afs_addr_list *alist = call->alist;
+ struct afs_addr_list *alist = call->probe_alist;
struct afs_vlserver *server = call->vlserver;
- struct afs_address *addr = &alist->addrs[call->addr_ix];
+ struct afs_address *addr = &alist->addrs[call->probe_index];
unsigned int server_index = call->server_index;
unsigned int rtt_us = 0;
- unsigned int index = call->addr_ix;
+ unsigned int index = call->probe_index;
bool have_result = false;
int ret = call->error;
@@ -148,25 +148,25 @@ static bool afs_do_probe_vlserver(struct afs_net *net,
unsigned int server_index,
struct afs_error *_e)
{
- struct afs_addr_cursor ac = {
- .index = 0,
- };
+ struct afs_addr_list *alist;
struct afs_call *call;
+ unsigned int index;
bool in_progress = false;
_enter("%s", server->name);
read_lock(&server->lock);
- ac.alist = rcu_dereference_protected(server->addresses,
- lockdep_is_held(&server->lock));
+ alist = rcu_dereference_protected(server->addresses,
+ lockdep_is_held(&server->lock));
+ afs_get_addrlist(alist, afs_alist_trace_get_vlprobe);
read_unlock(&server->lock);
- atomic_set(&server->probe_outstanding, ac.alist->nr_addrs);
+ atomic_set(&server->probe_outstanding, alist->nr_addrs);
memset(&server->probe, 0, sizeof(server->probe));
server->probe.rtt = UINT_MAX;
- for (ac.index = 0; ac.index < ac.alist->nr_addrs; ac.index++) {
- call = afs_vl_get_capabilities(net, &ac, key, server,
+ for (index = 0; index < alist->nr_addrs; index++) {
+ call = afs_vl_get_capabilities(net, alist, index, key, server,
server_index);
if (!IS_ERR(call)) {
afs_prioritise_error(_e, call->error, call->abort_code);
@@ -178,6 +178,7 @@ static bool afs_do_probe_vlserver(struct afs_net *net,
}
}
+ afs_put_addrlist(alist, afs_alist_trace_put_vlprobe);
return in_progress;
}
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index f895eb94129e..91168528179c 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -17,6 +17,8 @@
bool afs_begin_vlserver_operation(struct afs_vl_cursor *vc, struct afs_cell *cell,
struct key *key)
{
+ static atomic_t debug_ids;
+
memset(vc, 0, sizeof(*vc));
vc->cell = cell;
vc->key = key;
@@ -29,6 +31,7 @@ bool afs_begin_vlserver_operation(struct afs_vl_cursor *vc, struct afs_cell *cel
return false;
}
+ vc->debug_id = atomic_inc_return(&debug_ids);
return true;
}
@@ -89,17 +92,18 @@ static bool afs_start_vl_iteration(struct afs_vl_cursor *vc)
*/
bool afs_select_vlserver(struct afs_vl_cursor *vc)
{
- struct afs_addr_list *alist;
+ struct afs_addr_list *alist = vc->alist;
struct afs_vlserver *vlserver;
+ unsigned long set, failed;
unsigned int rtt;
s32 abort_code = vc->call_abort_code;
int error = vc->call_error, i;
vc->nr_iterations++;
- _enter("%lx[%d],%lx[%d],%d,%d",
- vc->untried_servers, vc->server_index,
- vc->ac.tried, vc->ac.index,
+ _enter("VC=%x+%x,%d{%lx},%d{%lx},%d,%d",
+ vc->debug_id, vc->nr_iterations, vc->server_index, vc->untried_servers,
+ vc->addr_index, vc->addr_tried,
error, abort_code);
if (vc->flags & AFS_VL_CURSOR_STOP) {
@@ -165,7 +169,13 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
restart_from_beginning:
_debug("restart");
- afs_end_cursor(&vc->ac);
+ if (vc->call_responded &&
+ vc->addr_index != vc->alist->preferred &&
+ test_bit(alist->preferred, &vc->addr_tried))
+ WRITE_ONCE(alist->preferred, vc->addr_index);
+ afs_put_addrlist(alist, afs_alist_trace_put_vlrotate_restart);
+ alist = vc->alist = NULL;
+
afs_put_vlserverlist(vc->cell->net, vc->server_list);
vc->server_list = NULL;
if (vc->flags & AFS_VL_CURSOR_RETRIED)
@@ -173,6 +183,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
vc->flags |= AFS_VL_CURSOR_RETRIED;
start:
_debug("start");
+ ASSERTCMP(alist, ==, NULL);
if (!afs_start_vl_iteration(vc))
goto failed;
@@ -185,6 +196,7 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
pick_server:
_debug("pick [%lx]", vc->untried_servers);
+ ASSERTCMP(alist, ==, NULL);
error = afs_wait_for_vl_probes(vc->server_list, vc->untried_servers);
if (error < 0) {
@@ -222,7 +234,6 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
* check it, find its address list and probe its capabilities before we
* use it.
*/
- ASSERTCMP(vc->ac.alist, ==, NULL);
vlserver = vc->server_list->servers[vc->server_index].server;
vc->server = vlserver;
@@ -231,30 +242,48 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
read_lock(&vlserver->lock);
alist = rcu_dereference_protected(vlserver->addresses,
lockdep_is_held(&vlserver->lock));
- afs_get_addrlist(alist, afs_alist_trace_get_vlrotate_set);
+ vc->alist = afs_get_addrlist(alist, afs_alist_trace_get_vlrotate_set);
read_unlock(&vlserver->lock);
- memset(&vc->ac, 0, sizeof(vc->ac));
- vc->ac.alist = alist;
- vc->ac.index = -1;
+ vc->addr_tried = 0;
+ vc->addr_index = -1;
iterate_address:
- ASSERT(vc->ac.alist);
/* Iterate over the current server's address list to try and find an
* address on which it will respond to us.
*/
- if (!afs_iterate_addresses(&vc->ac))
+ set = READ_ONCE(alist->responded);
+ failed = READ_ONCE(alist->probe_failed);
+ vc->addr_index = READ_ONCE(alist->preferred);
+
+ _debug("%lx-%lx-%lx,%d", set, failed, vc->addr_tried, vc->addr_index);
+
+ set &= ~(failed | vc->addr_tried);
+
+ if (!set)
goto next_server;
- _debug("VL address %d/%d", vc->ac.index, vc->ac.alist->nr_addrs);
+ if (!test_bit(vc->addr_index, &set))
+ vc->addr_index = __ffs(set);
+
+ set_bit(vc->addr_index, &vc->addr_tried);
+ vc->alist = alist;
+
+ _debug("VL address %d/%d", vc->addr_index, alist->nr_addrs);
vc->call_responded = false;
- _leave(" = t %pISpc", rxrpc_kernel_remote_addr(vc->ac.alist->addrs[vc->ac.index].peer));
+ _leave(" = t %pISpc", rxrpc_kernel_remote_addr(alist->addrs[vc->addr_index].peer));
return true;
next_server:
_debug("next");
- afs_end_cursor(&vc->ac);
+ ASSERT(alist);
+ if (vc->call_responded &&
+ vc->addr_index != alist->preferred &&
+ test_bit(alist->preferred, &vc->addr_tried))
+ WRITE_ONCE(alist->preferred, vc->addr_index);
+ afs_put_addrlist(alist, afs_alist_trace_put_vlrotate_next);
+ alist = vc->alist = NULL;
goto pick_server;
no_more_servers:
@@ -274,8 +303,15 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
}
failed:
+ if (alist) {
+ if (vc->call_responded &&
+ vc->addr_index != alist->preferred &&
+ test_bit(alist->preferred, &vc->addr_tried))
+ WRITE_ONCE(alist->preferred, vc->addr_index);
+ afs_put_addrlist(alist, afs_alist_trace_put_vlrotate_fail);
+ alist = vc->alist = NULL;
+ }
vc->flags |= AFS_VL_CURSOR_STOP;
- afs_end_cursor(&vc->ac);
_leave(" = f [failed %d]", vc->cumul_error.error);
return false;
}
@@ -299,8 +335,8 @@ static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
pr_notice("DNS: src=%u st=%u lc=%x\n",
cell->dns_source, cell->dns_status, cell->dns_lookup_count);
pr_notice("VC: ut=%lx ix=%u ni=%hu fl=%hx err=%hd\n",
- vc->untried_servers, vc->server_index, vc->nr_iterations, vc->flags,
- vc->cumul_error.error);
+ vc->untried_servers, vc->server_index, vc->nr_iterations,
+ vc->flags, vc->cumul_error.error);
pr_notice("VC: call er=%d ac=%d r=%u\n",
vc->call_error, vc->call_abort_code, vc->call_responded);
@@ -320,14 +356,13 @@ static void afs_vl_dump_edestaddrreq(const struct afs_vl_cursor *vc)
a->preferred);
pr_notice("VC: - R=%lx F=%lx\n",
a->responded, a->probe_failed);
- if (a == vc->ac.alist)
+ if (a == vc->alist)
pr_notice("VC: - current\n");
}
}
}
- pr_notice("AC: t=%lx ax=%u ni=%u\n",
- vc->ac.tried, vc->ac.index, vc->ac.nr_iterations);
+ pr_notice("AC: t=%lx ax=%u\n", vc->addr_tried, vc->addr_index);
rcu_read_unlock();
}
@@ -338,6 +373,8 @@ int afs_end_vlserver_operation(struct afs_vl_cursor *vc)
{
struct afs_net *net = vc->cell->net;
+ _enter("VC=%x+%x", vc->debug_id, vc->nr_iterations);
+
switch (vc->cumul_error.error) {
case -EDESTADDRREQ:
case -EADDRNOTAVAIL:
@@ -347,7 +384,14 @@ int afs_end_vlserver_operation(struct afs_vl_cursor *vc)
break;
}
- afs_end_cursor(&vc->ac);
+ if (vc->alist) {
+ if (vc->call_responded &&
+ vc->addr_index != vc->alist->preferred &&
+ test_bit(vc->alist->preferred, &vc->addr_tried))
+ WRITE_ONCE(vc->alist->preferred, vc->addr_index);
+ afs_put_addrlist(vc->alist, afs_alist_trace_put_vlrotate_end);
+ vc->alist = NULL;
+ }
afs_put_vlserverlist(net, vc->server_list);
return vc->cumul_error.error;
}
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index 4bf98a38c3a1..39a0b7614d05 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -149,7 +149,7 @@ struct afs_vldb_entry *afs_vl_get_entry_by_name_u(struct afs_vl_cursor *vc,
call->key = vc->key;
call->ret_vldb = entry;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
- call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->peer = rxrpc_kernel_get_peer(vc->alist->addrs[vc->addr_index].peer);
call->service_id = vc->server->service_id;
/* Marshall the parameters */
@@ -161,8 +161,8 @@ struct afs_vldb_entry *afs_vl_get_entry_by_name_u(struct afs_vl_cursor *vc,
memset((void *)bp + volnamesz, 0, padsz);
trace_afs_make_vl_call(call);
- afs_make_call(&vc->ac, call, GFP_KERNEL);
- afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_make_call(call, GFP_KERNEL);
+ afs_wait_for_call_to_complete(call);
vc->call_abort_code = call->abort_code;
vc->call_error = call->error;
vc->call_responded = call->responded;
@@ -290,7 +290,7 @@ struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *vc,
call->key = vc->key;
call->ret_alist = NULL;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
- call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->peer = rxrpc_kernel_get_peer(vc->alist->addrs[vc->addr_index].peer);
call->service_id = vc->server->service_id;
/* Marshall the parameters */
@@ -310,8 +310,8 @@ struct afs_addr_list *afs_vl_get_addrs_u(struct afs_vl_cursor *vc,
r->uuid.node[i] = htonl(u->node[i]);
trace_afs_make_vl_call(call);
- afs_make_call(&vc->ac, call, GFP_KERNEL);
- afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_make_call(call, GFP_KERNEL);
+ afs_wait_for_call_to_complete(call);
vc->call_abort_code = call->abort_code;
vc->call_error = call->error;
vc->call_responded = call->responded;
@@ -371,6 +371,7 @@ static int afs_deliver_vl_get_capabilities(struct afs_call *call)
static void afs_destroy_vl_get_capabilities(struct afs_call *call)
{
+ afs_put_addrlist(call->probe_alist, afs_alist_trace_put_vlgetcaps);
afs_put_vlserver(call->net, call->vlserver);
afs_flat_call_destructor(call);
}
@@ -394,7 +395,8 @@ static const struct afs_call_type afs_RXVLGetCapabilities = {
* other end supports.
*/
struct afs_call *afs_vl_get_capabilities(struct afs_net *net,
- struct afs_addr_cursor *ac,
+ struct afs_addr_list *alist,
+ unsigned int addr_index,
struct key *key,
struct afs_vlserver *server,
unsigned int server_index)
@@ -411,7 +413,9 @@ struct afs_call *afs_vl_get_capabilities(struct afs_net *net,
call->key = key;
call->vlserver = afs_get_vlserver(server);
call->server_index = server_index;
- call->peer = rxrpc_kernel_get_peer(ac->alist->addrs[ac->index].peer);
+ call->peer = rxrpc_kernel_get_peer(alist->addrs[addr_index].peer);
+ call->probe_alist = afs_get_addrlist(alist, afs_alist_trace_get_vlgetcaps);
+ call->probe_index = addr_index;
call->service_id = server->service_id;
call->upgrade = true;
call->async = true;
@@ -423,7 +427,7 @@ struct afs_call *afs_vl_get_capabilities(struct afs_net *net,
/* Can't take a ref on server */
trace_afs_make_vl_call(call);
- afs_make_call(ac, call, GFP_KERNEL);
+ afs_make_call(call, GFP_KERNEL);
return call;
}
@@ -658,7 +662,7 @@ struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *vc,
call->key = vc->key;
call->ret_alist = NULL;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
- call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->peer = rxrpc_kernel_get_peer(vc->alist->addrs[vc->addr_index].peer);
call->service_id = vc->server->service_id;
/* Marshall the parameters */
@@ -668,8 +672,8 @@ struct afs_addr_list *afs_yfsvl_get_endpoints(struct afs_vl_cursor *vc,
memcpy(bp, uuid, sizeof(*uuid)); /* Type opr_uuid */
trace_afs_make_vl_call(call);
- afs_make_call(&vc->ac, call, GFP_KERNEL);
- afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_make_call(call, GFP_KERNEL);
+ afs_wait_for_call_to_complete(call);
vc->call_abort_code = call->abort_code;
vc->call_error = call->error;
vc->call_responded = call->responded;
@@ -777,7 +781,7 @@ char *afs_yfsvl_get_cell_name(struct afs_vl_cursor *vc)
call->key = vc->key;
call->ret_str = NULL;
call->max_lifespan = AFS_VL_MAX_LIFESPAN;
- call->peer = rxrpc_kernel_get_peer(vc->ac.alist->addrs[vc->ac.index].peer);
+ call->peer = rxrpc_kernel_get_peer(vc->alist->addrs[vc->addr_index].peer);
call->service_id = vc->server->service_id;
/* marshall the parameters */
@@ -786,8 +790,8 @@ char *afs_yfsvl_get_cell_name(struct afs_vl_cursor *vc)
/* Can't take a ref on server */
trace_afs_make_vl_call(call);
- afs_make_call(&vc->ac, call, GFP_KERNEL);
- afs_wait_for_call_to_complete(call, &vc->ac);
+ afs_make_call(call, GFP_KERNEL);
+ afs_wait_for_call_to_complete(call);
vc->call_abort_code = call->abort_code;
vc->call_error = call->error;
vc->call_responded = call->responded;
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index ed91666ca4cc..0f68d67f52c8 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -204,21 +204,31 @@ enum yfs_cm_operation {
#define afs_alist_traces \
EM(afs_alist_trace_alloc, "ALLOC ") \
+ EM(afs_alist_trace_get_getcaps, "GET getcap") \
EM(afs_alist_trace_get_fsrotate_set, "GET fs-rot") \
- EM(afs_alist_trace_get_make_call, "GET mkcall") \
EM(afs_alist_trace_get_probe, "GET probe ") \
+ EM(afs_alist_trace_get_vlgetcaps, "GET vgtcap") \
+ EM(afs_alist_trace_get_vlprobe, "GET vprobe") \
EM(afs_alist_trace_get_vlrotate_set, "GET vl-rot") \
- EM(afs_alist_trace_put_call, "PUT call ") \
- EM(afs_alist_trace_put_end_cursor, "PUT endcur") \
EM(afs_alist_trace_put_getaddru, "PUT GtAdrU") \
+ EM(afs_alist_trace_put_getcaps, "PUT getcap") \
+ EM(afs_alist_trace_put_next_server, "PUT nx-srv") \
+ EM(afs_alist_trace_put_op_failed, "PUT op-fai") \
+ EM(afs_alist_trace_put_operation, "PUT op ") \
EM(afs_alist_trace_put_parse_empty, "PUT p-empt") \
EM(afs_alist_trace_put_parse_error, "PUT p-err ") \
EM(afs_alist_trace_put_probe, "PUT probe ") \
- EM(afs_alist_trace_put_retry_server, "PUT retry ") \
+ EM(afs_alist_trace_put_restart_rotate, "PUT rstrot") \
EM(afs_alist_trace_put_server, "PUT server") \
EM(afs_alist_trace_put_server_dup, "PUT sv-dup") \
EM(afs_alist_trace_put_server_oom, "PUT sv-oom") \
EM(afs_alist_trace_put_server_update, "PUT sv-upd") \
+ EM(afs_alist_trace_put_vlgetcaps, "PUT vgtcap") \
+ EM(afs_alist_trace_put_vlprobe, "PUT vprobe") \
+ EM(afs_alist_trace_put_vlrotate_end, "PUT vr-end") \
+ EM(afs_alist_trace_put_vlrotate_fail, "PUT vr-fai") \
+ EM(afs_alist_trace_put_vlrotate_next, "PUT vr-nxt") \
+ EM(afs_alist_trace_put_vlrotate_restart,"PUT vr-rst") \
EM(afs_alist_trace_put_vlserver, "PUT vlsrvr") \
EM(afs_alist_trace_put_vlserver_old, "PUT vs-old") \
E_(afs_alist_trace_free, "FREE ")
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 23/41] rxrpc: Create a procfile to display outstanding clien conn bundles
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (21 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 22/41] afs: Fold the afs_addr_cursor struct in David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 18:20 ` Jeffrey E Altman
2023-11-09 15:39 ` [PATCH 24/41] afs: Add some more info to /proc/net/afs/servers David Howells
` (17 subsequent siblings)
40 siblings, 1 reply; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Create /proc/net/rxrpc/bundles to display outstanding rxrpc client
connection bundles.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
net/rxrpc/ar-internal.h | 4 +++
net/rxrpc/conn_client.c | 10 ++++++
net/rxrpc/net_ns.c | 4 +++
net/rxrpc/proc.c | 76 +++++++++++++++++++++++++++++++++++++++++
4 files changed, 94 insertions(+)
diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
index 8eea7a487380..2f8b39a614c3 100644
--- a/net/rxrpc/ar-internal.h
+++ b/net/rxrpc/ar-internal.h
@@ -68,6 +68,7 @@ struct rxrpc_net {
atomic_t nr_calls; /* Count of allocated calls */
atomic_t nr_conns;
+ struct list_head bundle_proc_list; /* List of bundles for proc */
struct list_head conn_proc_list; /* List of conns in this namespace for proc */
struct list_head service_conns; /* Service conns in this namespace */
rwlock_t conn_lock; /* Lock for ->conn_proc_list, ->service_conns */
@@ -432,6 +433,7 @@ struct rxrpc_bundle {
struct rxrpc_local *local; /* Representation of local endpoint */
struct rxrpc_peer *peer; /* Remote endpoint */
struct key *key; /* Security details */
+ struct list_head proc_link; /* Link in net->bundle_proc_list */
const struct rxrpc_security *security; /* applied security module */
refcount_t ref;
atomic_t active; /* Number of active users */
@@ -445,6 +447,7 @@ struct rxrpc_bundle {
struct rb_node local_node; /* Node in local->client_conns */
struct list_head waiting_calls; /* Calls waiting for channels */
unsigned long avail_chans; /* Mask of available channels */
+ unsigned int conn_ids[4]; /* Connection IDs. */
struct rxrpc_connection *conns[4]; /* The connections in the bundle (max 4) */
};
@@ -1167,6 +1170,7 @@ void rxrpc_put_peer(struct rxrpc_peer *, enum rxrpc_peer_trace);
*/
extern const struct seq_operations rxrpc_call_seq_ops;
extern const struct seq_operations rxrpc_connection_seq_ops;
+extern const struct seq_operations rxrpc_bundle_seq_ops;
extern const struct seq_operations rxrpc_peer_seq_ops;
extern const struct seq_operations rxrpc_local_seq_ops;
diff --git a/net/rxrpc/conn_client.c b/net/rxrpc/conn_client.c
index 1d95f8bc769f..3b9b267a4431 100644
--- a/net/rxrpc/conn_client.c
+++ b/net/rxrpc/conn_client.c
@@ -91,6 +91,10 @@ static struct rxrpc_bundle *rxrpc_alloc_bundle(struct rxrpc_call *call,
atomic_set(&bundle->active, 1);
INIT_LIST_HEAD(&bundle->waiting_calls);
trace_rxrpc_bundle(bundle->debug_id, 1, rxrpc_bundle_new);
+
+ write_lock(&bundle->local->rxnet->conn_lock);
+ list_add_tail(&bundle->proc_link, &bundle->local->rxnet->bundle_proc_list);
+ write_unlock(&bundle->local->rxnet->conn_lock);
}
return bundle;
}
@@ -109,6 +113,9 @@ static void rxrpc_free_bundle(struct rxrpc_bundle *bundle)
{
trace_rxrpc_bundle(bundle->debug_id, refcount_read(&bundle->ref),
rxrpc_bundle_free);
+ write_lock(&bundle->local->rxnet->conn_lock);
+ list_del(&bundle->proc_link);
+ write_unlock(&bundle->local->rxnet->conn_lock);
rxrpc_put_peer(bundle->peer, rxrpc_peer_put_bundle);
key_put(bundle->key);
kfree(bundle);
@@ -338,6 +345,7 @@ static bool rxrpc_add_conn_to_bundle(struct rxrpc_bundle *bundle,
old = bundle->conns[slot];
if (old) {
bundle->conns[slot] = NULL;
+ bundle->conn_ids[slot] = 0;
trace_rxrpc_client(old, -1, rxrpc_client_replace);
rxrpc_put_connection(old, rxrpc_conn_put_noreuse);
}
@@ -351,6 +359,7 @@ static bool rxrpc_add_conn_to_bundle(struct rxrpc_bundle *bundle,
rxrpc_activate_bundle(bundle);
conn->bundle_shift = shift;
bundle->conns[slot] = conn;
+ bundle->conn_ids[slot] = conn->debug_id;
for (i = 0; i < RXRPC_MAXCALLS; i++)
set_bit(shift + i, &bundle->avail_chans);
return true;
@@ -671,6 +680,7 @@ static void rxrpc_unbundle_conn(struct rxrpc_connection *conn)
if (bundle->conns[bindex] == conn) {
_debug("clear slot %u", bindex);
bundle->conns[bindex] = NULL;
+ bundle->conn_ids[bindex] = 0;
for (i = 0; i < RXRPC_MAXCALLS; i++)
clear_bit(conn->bundle_shift + i, &bundle->avail_chans);
rxrpc_put_client_connection_id(bundle->local, conn);
diff --git a/net/rxrpc/net_ns.c b/net/rxrpc/net_ns.c
index a0319c040c25..a4c135d0fbcc 100644
--- a/net/rxrpc/net_ns.c
+++ b/net/rxrpc/net_ns.c
@@ -45,6 +45,7 @@ static __net_init int rxrpc_init_net(struct net *net)
atomic_set(&rxnet->nr_calls, 1);
atomic_set(&rxnet->nr_conns, 1);
+ INIT_LIST_HEAD(&rxnet->bundle_proc_list);
INIT_LIST_HEAD(&rxnet->conn_proc_list);
INIT_LIST_HEAD(&rxnet->service_conns);
rwlock_init(&rxnet->conn_lock);
@@ -78,6 +79,9 @@ static __net_init int rxrpc_init_net(struct net *net)
proc_create_net("conns", 0444, rxnet->proc_net,
&rxrpc_connection_seq_ops,
sizeof(struct seq_net_private));
+ proc_create_net("bundles", 0444, rxnet->proc_net,
+ &rxrpc_bundle_seq_ops,
+ sizeof(struct seq_net_private));
proc_create_net("peers", 0444, rxnet->proc_net,
&rxrpc_peer_seq_ops,
sizeof(struct seq_net_private));
diff --git a/net/rxrpc/proc.c b/net/rxrpc/proc.c
index 682636d3b060..6c86cbb98d1d 100644
--- a/net/rxrpc/proc.c
+++ b/net/rxrpc/proc.c
@@ -198,6 +198,82 @@ const struct seq_operations rxrpc_connection_seq_ops = {
.show = rxrpc_connection_seq_show,
};
+/*
+ * generate a list of extant virtual bundles in /proc/net/rxrpc/bundles
+ */
+static void *rxrpc_bundle_seq_start(struct seq_file *seq, loff_t *_pos)
+ __acquires(rxnet->conn_lock)
+{
+ struct rxrpc_net *rxnet = rxrpc_net(seq_file_net(seq));
+
+ read_lock(&rxnet->conn_lock);
+ return seq_list_start_head(&rxnet->bundle_proc_list, *_pos);
+}
+
+static void *rxrpc_bundle_seq_next(struct seq_file *seq, void *v,
+ loff_t *pos)
+{
+ struct rxrpc_net *rxnet = rxrpc_net(seq_file_net(seq));
+
+ return seq_list_next(v, &rxnet->bundle_proc_list, pos);
+}
+
+static void rxrpc_bundle_seq_stop(struct seq_file *seq, void *v)
+ __releases(rxnet->conn_lock)
+{
+ struct rxrpc_net *rxnet = rxrpc_net(seq_file_net(seq));
+
+ read_unlock(&rxnet->conn_lock);
+}
+
+static int rxrpc_bundle_seq_show(struct seq_file *seq, void *v)
+{
+ struct rxrpc_bundle *bundle;
+ struct rxrpc_net *rxnet = rxrpc_net(seq_file_net(seq));
+ char lbuff[50], rbuff[50];
+
+ if (v == &rxnet->bundle_proc_list) {
+ seq_puts(seq,
+ "Proto Local "
+ " Remote "
+ " SvID Ref Act Flg Key |"
+ " Bundle Conn_0 Conn_1 Conn_2 Conn_3\n"
+ );
+ return 0;
+ }
+
+ bundle = list_entry(v, struct rxrpc_bundle, proc_link);
+
+ sprintf(lbuff, "%pISpc", &bundle->local->srx.transport);
+ sprintf(rbuff, "%pISpc", &bundle->peer->srx.transport);
+ seq_printf(seq,
+ "UDP %-47.47s %-47.47s %4x %3u %3d"
+ " %c%c%c %08x | %08x %08x %08x %08x %08x\n",
+ lbuff,
+ rbuff,
+ bundle->service_id,
+ refcount_read(&bundle->ref),
+ atomic_read(&bundle->active),
+ bundle->try_upgrade ? 'U' : '-',
+ bundle->exclusive ? 'e' : '-',
+ bundle->upgrade ? 'u' : '-',
+ key_serial(bundle->key),
+ bundle->debug_id,
+ bundle->conn_ids[0],
+ bundle->conn_ids[1],
+ bundle->conn_ids[2],
+ bundle->conn_ids[3]);
+
+ return 0;
+}
+
+const struct seq_operations rxrpc_bundle_seq_ops = {
+ .start = rxrpc_bundle_seq_start,
+ .next = rxrpc_bundle_seq_next,
+ .stop = rxrpc_bundle_seq_stop,
+ .show = rxrpc_bundle_seq_show,
+};
+
/*
* generate a list of extant virtual peers in /proc/net/rxrpc/peers
*/
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 24/41] afs: Add some more info to /proc/net/afs/servers
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (22 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 23/41] rxrpc: Create a procfile to display outstanding clien conn bundles David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 25/41] afs: Remove the unimplemented afs_cmp_addr_list() David Howells
` (16 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
In /proc/net/afs/servers, show the cell name and the last error for each
address in the server's list.
cc: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/fs_operation.c | 2 --
fs/afs/fs_probe.c | 2 ++
fs/afs/proc.c | 24 ++++++++++++++++--------
fs/afs/rotate.c | 2 ++
fs/afs/vl_rotate.c | 2 ++
5 files changed, 22 insertions(+), 10 deletions(-)
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index 00e22259be36..e760e11d5bcb 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -196,8 +196,6 @@ void afs_wait_for_operation(struct afs_operation *op)
op->call_abort_code = op->call->abort_code;
op->call_error = op->call->error;
op->call_responded = op->call->responded;
- WRITE_ONCE(op->alist->addrs[op->addr_index].last_error,
- op->call_error);
afs_put_call(op->call);
}
}
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index aef16ac3f577..8008d3ecabab 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -107,6 +107,8 @@ void afs_fileserver_probe_result(struct afs_call *call)
_enter("%pU,%u", &server->uuid, index);
+ WRITE_ONCE(addr->last_error, ret);
+
spin_lock(&server->probe_lock);
switch (ret) {
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index 16d93fa6396f..0b43bb9b0260 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -377,31 +377,39 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
{
struct afs_server *server;
struct afs_addr_list *alist;
+ unsigned long failed;
int i;
if (v == SEQ_START_TOKEN) {
- seq_puts(m, "UUID REF ACT\n");
+ seq_puts(m, "UUID REF ACT CELL\n");
return 0;
}
server = list_entry(v, struct afs_server, proc_link);
alist = rcu_dereference(server->addresses);
- seq_printf(m, "%pU %3d %3d\n",
+ seq_printf(m, "%pU %3d %3d %s\n",
&server->uuid,
refcount_read(&server->ref),
- atomic_read(&server->active));
+ atomic_read(&server->active),
+ server->cell->name);
seq_printf(m, " - info: fl=%lx rtt=%u brk=%x\n",
server->flags, server->rtt, server->cb_s_break);
seq_printf(m, " - probe: last=%d out=%d\n",
(int)(jiffies - server->probed_at) / HZ,
atomic_read(&server->probe_outstanding));
+ failed = alist->probe_failed;
seq_printf(m, " - ALIST v=%u rsp=%lx f=%lx\n",
alist->version, alist->responded, alist->probe_failed);
- for (i = 0; i < alist->nr_addrs; i++)
- seq_printf(m, " [%x] %pISpc%s rtt=%d\n",
- i, rxrpc_kernel_remote_addr(alist->addrs[i].peer),
- alist->preferred == i ? "*" : "",
- rxrpc_kernel_get_srtt(alist->addrs[i].peer));
+ for (i = 0; i < alist->nr_addrs; i++) {
+ const struct afs_address *addr = &alist->addrs[i];
+
+ seq_printf(m, " [%x] %pISpc%s rtt=%d err=%d\n",
+ i, rxrpc_kernel_remote_addr(addr->peer),
+ alist->preferred == i ? "*" :
+ test_bit(i, &failed) ? "!" : "",
+ rxrpc_kernel_get_srtt(addr->peer),
+ addr->last_error);
+ }
return 0;
}
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 2ad119d632fa..8b4151eb28be 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -133,6 +133,8 @@ bool afs_select_fileserver(struct afs_operation *op)
if (op->nr_iterations == 0)
goto start;
+ WRITE_ONCE(alist->addrs[op->addr_index].last_error, error);
+
/* Evaluate the result of the previous operation, if there was one. */
switch (op->call_error) {
case 0:
diff --git a/fs/afs/vl_rotate.c b/fs/afs/vl_rotate.c
index 91168528179c..d8f79f6ada3d 100644
--- a/fs/afs/vl_rotate.c
+++ b/fs/afs/vl_rotate.c
@@ -114,6 +114,8 @@ bool afs_select_vlserver(struct afs_vl_cursor *vc)
if (vc->nr_iterations == 0)
goto start;
+ WRITE_ONCE(alist->addrs[vc->addr_index].last_error, error);
+
/* Evaluate the result of the previous operation, if there was one. */
switch (error) {
default:
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 25/41] afs: Remove the unimplemented afs_cmp_addr_list()
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (23 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 24/41] afs: Add some more info to /proc/net/afs/servers David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 26/41] afs: Provide a way to configure address priorities David Howells
` (15 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Remove afs_cmp_addr_list() as it was never implemented.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_list.c | 13 -------------
1 file changed, 13 deletions(-)
diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index 18c286efa3a5..6d42f85c6be5 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -244,19 +244,6 @@ struct afs_vlserver_list *afs_parse_text_addrs(struct afs_net *net,
return ERR_PTR(ret);
}
-/*
- * Compare old and new address lists to see if there's been any change.
- * - How to do this in better than O(Nlog(N)) time?
- * - We don't really want to sort the address list, but would rather take the
- * list as we got it so as not to undo record rotation by the DNS server.
- */
-#if 0
-static int afs_cmp_addr_list(const struct afs_addr_list *a1,
- const struct afs_addr_list *a2)
-{
-}
-#endif
-
/*
* Perform a DNS query for VL servers and build a up an address list.
*/
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 26/41] afs: Provide a way to configure address priorities
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (24 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 25/41] afs: Remove the unimplemented afs_cmp_addr_list() David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 27/41] afs: Mark address lists with configured priorities David Howells
` (14 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
AFS servers may have multiple addresses, but the client can't easily judge
between them as to which one is best. For instance, an address that has a
larger RTT might actually have a better bandwidth because it goes through a
switch rather than being directly connected - but we can't work this out
dynamically unless we push through sufficient data that we can measure it.
To allow the administrator to configure this, add a list of preference
weightings for server addresses by IPv4/IPv6 address or subnet and allow
this to be viewed through a procfile and altered by writing text commands
to that same file. Preference rules can be added/updated by:
echo "add <proto> <addr>[/<subnet>] <prior>" >/proc/fs/afs/addr_prefs
echo "add udp 1.2.3.4 1000" >/proc/fs/afs/addr_prefs
echo "add udp 192.168.0.0/16 3000" >/proc/fs/afs/addr_prefs
echo "add udp 1001:2002:0:6::/64 4000" >/proc/fs/afs/addr_prefs
and removed by:
echo "del <proto> <addr>[/<subnet>]" >/proc/fs/afs/addr_prefs
echo "del udp 1.2.3.4" >/proc/fs/afs/addr_prefs
where the priority is a number between 0 and 65535.
The list is split between IPv4 and IPv6 addresses and each sublist is kept
in numerical order, with rules that would otherwise match but have
different subnet masking being ordered with the most specific submatch
first.
A subsequent patch will apply these rules.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/Makefile | 1 +
fs/afs/addr_prefs.c | 449 ++++++++++++++++++++++++++++++++++++++++++++
fs/afs/internal.h | 29 +++
fs/afs/main.c | 1 +
fs/afs/proc.c | 55 +++++-
5 files changed, 534 insertions(+), 1 deletion(-)
create mode 100644 fs/afs/addr_prefs.c
diff --git a/fs/afs/Makefile b/fs/afs/Makefile
index e8956b65d7ff..b3849bea0553 100644
--- a/fs/afs/Makefile
+++ b/fs/afs/Makefile
@@ -5,6 +5,7 @@
kafs-y := \
addr_list.o \
+ addr_prefs.o \
callback.o \
cell.o \
cmservice.o \
diff --git a/fs/afs/addr_prefs.c b/fs/afs/addr_prefs.c
new file mode 100644
index 000000000000..c6dcff4f8aa1
--- /dev/null
+++ b/fs/afs/addr_prefs.c
@@ -0,0 +1,449 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* Address preferences management
+ *
+ * Copyright (C) 2023 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": addr_prefs: " fmt
+#include <linux/slab.h>
+#include <linux/ctype.h>
+#include <linux/inet.h>
+#include <linux/seq_file.h>
+#include <keys/rxrpc-type.h>
+#include "internal.h"
+
+static inline struct afs_net *afs_seq2net_single(struct seq_file *m)
+{
+ return afs_net(seq_file_single_net(m));
+}
+
+/*
+ * Split a NUL-terminated string up to the first newline around spaces. The
+ * source string will be modified to have NUL-terminations inserted.
+ */
+static int afs_split_string(char **pbuf, char *strv[], unsigned int maxstrv)
+{
+ unsigned int count = 0;
+ char *p = *pbuf;
+
+ maxstrv--; /* Allow for terminal NULL */
+ for (;;) {
+ /* Skip over spaces */
+ while (isspace(*p)) {
+ if (*p == '\n') {
+ p++;
+ break;
+ }
+ p++;
+ }
+ if (!*p)
+ break;
+
+ /* Mark start of word */
+ if (count >= maxstrv) {
+ pr_warn("Too many elements in string\n");
+ return -EINVAL;
+ }
+ strv[count++] = p;
+
+ /* Skip over word */
+ while (!isspace(*p))
+ p++;
+ if (!*p)
+ break;
+
+ /* Mark end of word */
+ if (*p == '\n') {
+ *p++ = 0;
+ break;
+ }
+ *p++ = 0;
+ }
+
+ *pbuf = p;
+ strv[count] = NULL;
+ return count;
+}
+
+/*
+ * Parse an address with an optional subnet mask.
+ */
+static int afs_parse_address(char *p, struct afs_addr_preference *pref)
+{
+ const char *stop;
+ unsigned long mask, tmp;
+ char *end = p + strlen(p);
+ bool bracket = false;
+
+ if (*p == '[') {
+ p++;
+ bracket = true;
+ }
+
+#if 0
+ if (*p == '[') {
+ p++;
+ q = memchr(p, ']', end - p);
+ if (!q) {
+ pr_warn("Can't find closing ']'\n");
+ return -EINVAL;
+ }
+ } else {
+ for (q = p; q < end; q++)
+ if (*q == '/')
+ break;
+ }
+#endif
+
+ if (in4_pton(p, end - p, (u8 *)&pref->ipv4_addr, -1, &stop)) {
+ pref->family = AF_INET;
+ mask = 32;
+ } else if (in6_pton(p, end - p, (u8 *)&pref->ipv6_addr, -1, &stop)) {
+ pref->family = AF_INET6;
+ mask = 128;
+ } else {
+ pr_warn("Can't determine address family\n");
+ return -EINVAL;
+ }
+
+ p = (char *)stop;
+ if (bracket) {
+ if (*p != ']') {
+ pr_warn("Can't find closing ']'\n");
+ return -EINVAL;
+ }
+ p++;
+ }
+
+ if (*p == '/') {
+ p++;
+ tmp = simple_strtoul(p, &p, 10);
+ if (tmp > mask) {
+ pr_warn("Subnet mask too large\n");
+ return -EINVAL;
+ }
+ if (tmp == 0) {
+ pr_warn("Subnet mask too small\n");
+ return -EINVAL;
+ }
+ mask = tmp;
+ }
+
+ if (*p) {
+ pr_warn("Invalid address\n");
+ return -EINVAL;
+ }
+
+ pref->subnet_mask = mask;
+ return 0;
+}
+
+enum cmp_ret {
+ CONTINUE_SEARCH,
+ INSERT_HERE,
+ EXACT_MATCH,
+ SUBNET_MATCH,
+};
+
+/*
+ * See if a candidate address matches a listed address.
+ */
+static enum cmp_ret afs_cmp_address_pref(const struct afs_addr_preference *a,
+ const struct afs_addr_preference *b)
+{
+ int subnet = min(a->subnet_mask, b->subnet_mask);
+ const __be32 *pa, *pb;
+ u32 mask, na, nb;
+ int diff;
+
+ if (a->family != b->family)
+ return INSERT_HERE;
+
+ switch (a->family) {
+ case AF_INET6:
+ pa = a->ipv6_addr.s6_addr32;
+ pb = b->ipv6_addr.s6_addr32;
+ break;
+ case AF_INET:
+ pa = &a->ipv4_addr.s_addr;
+ pb = &b->ipv4_addr.s_addr;
+ break;
+ }
+
+ while (subnet > 32) {
+ diff = ntohl(*pa++) - ntohl(*pb++);
+ if (diff < 0)
+ return INSERT_HERE; /* a<b */
+ if (diff > 0)
+ return CONTINUE_SEARCH; /* a>b */
+ subnet -= 32;
+ }
+
+ if (subnet == 0)
+ return EXACT_MATCH;
+
+ mask = 0xffffffffU << (32 - subnet);
+ na = ntohl(*pa);
+ nb = ntohl(*pb);
+ diff = (na & mask) - (nb & mask);
+ //kdebug("diff %08x %08x %08x %d", na, nb, mask, diff);
+ if (diff < 0)
+ return INSERT_HERE; /* a<b */
+ if (diff > 0)
+ return CONTINUE_SEARCH; /* a>b */
+ if (a->subnet_mask == b->subnet_mask)
+ return EXACT_MATCH;
+ if (a->subnet_mask > b->subnet_mask)
+ return SUBNET_MATCH; /* a binds tighter than b */
+ return CONTINUE_SEARCH; /* b binds tighter than a */
+}
+
+/*
+ * Insert an address preference.
+ */
+static int afs_insert_address_pref(struct afs_addr_preference_list **_preflist,
+ struct afs_addr_preference *pref,
+ int index)
+{
+ struct afs_addr_preference_list *preflist = *_preflist, *old = preflist;
+ size_t size, max_prefs;
+
+ _enter("{%u/%u/%u},%u", preflist->ipv6_off, preflist->nr, preflist->max_prefs, index);
+
+ if (preflist->nr == 255)
+ return -ENOSPC;
+ if (preflist->nr >= preflist->max_prefs) {
+ max_prefs = preflist->max_prefs + 1;
+ size = struct_size(preflist, prefs, max_prefs);
+ size = roundup_pow_of_two(size);
+ max_prefs = min_t(size_t, (size - sizeof(*preflist)) / sizeof(*pref), 255);
+ preflist = kmalloc(size, GFP_KERNEL);
+ if (!preflist)
+ return -ENOMEM;
+ *preflist = **_preflist;
+ preflist->max_prefs = max_prefs;
+ *_preflist = preflist;
+
+ if (index < preflist->nr)
+ memcpy(preflist->prefs + index + 1, old->prefs + index,
+ sizeof(*pref) * (preflist->nr - index));
+ if (index > 0)
+ memcpy(preflist->prefs, old->prefs, sizeof(*pref) * index);
+ } else {
+ if (index < preflist->nr)
+ memmove(preflist->prefs + index + 1, preflist->prefs + index,
+ sizeof(*pref) * (preflist->nr - index));
+ }
+
+ preflist->prefs[index] = *pref;
+ preflist->nr++;
+ if (pref->family == AF_INET)
+ preflist->ipv6_off++;
+ return 0;
+}
+
+/*
+ * Add an address preference.
+ * echo "add <proto> <IP>[/<mask>] <prior>" >/proc/fs/afs/addr_prefs
+ */
+static int afs_add_address_pref(struct afs_net *net, struct afs_addr_preference_list **_preflist,
+ int argc, char **argv)
+{
+ struct afs_addr_preference_list *preflist = *_preflist;
+ struct afs_addr_preference pref;
+ enum cmp_ret cmp;
+ int ret, i, stop;
+
+ if (argc != 3) {
+ pr_warn("Wrong number of params\n");
+ return -EINVAL;
+ }
+
+ if (strcmp(argv[0], "udp") != 0) {
+ pr_warn("Unsupported protocol\n");
+ return -EINVAL;
+ }
+
+ ret = afs_parse_address(argv[1], &pref);
+ if (ret < 0)
+ return ret;
+
+ ret = kstrtou16(argv[2], 10, &pref.prio);
+ if (ret < 0) {
+ pr_warn("Invalid priority\n");
+ return ret;
+ }
+
+ if (pref.family == AF_INET) {
+ i = 0;
+ stop = preflist->ipv6_off;
+ } else {
+ i = preflist->ipv6_off;
+ stop = preflist->nr;
+ }
+
+ for (; i < stop; i++) {
+ cmp = afs_cmp_address_pref(&pref, &preflist->prefs[i]);
+ switch (cmp) {
+ case CONTINUE_SEARCH:
+ continue;
+ case INSERT_HERE:
+ case SUBNET_MATCH:
+ return afs_insert_address_pref(_preflist, &pref, i);
+ case EXACT_MATCH:
+ preflist->prefs[i].prio = pref.prio;
+ return 0;
+ }
+ }
+
+ return afs_insert_address_pref(_preflist, &pref, i);
+}
+
+/*
+ * Delete an address preference.
+ */
+static int afs_delete_address_pref(struct afs_addr_preference_list **_preflist,
+ int index)
+{
+ struct afs_addr_preference_list *preflist = *_preflist;
+
+ _enter("{%u/%u/%u},%u", preflist->ipv6_off, preflist->nr, preflist->max_prefs, index);
+
+ if (preflist->nr == 0)
+ return -ENOENT;
+
+ if (index < preflist->nr - 1)
+ memmove(preflist->prefs + index, preflist->prefs + index + 1,
+ sizeof(preflist->prefs[0]) * (preflist->nr - index - 1));
+
+ if (index < preflist->ipv6_off)
+ preflist->ipv6_off--;
+ preflist->nr--;
+ return 0;
+}
+
+/*
+ * Delete an address preference.
+ * echo "del <proto> <IP>[/<mask>]" >/proc/fs/afs/addr_prefs
+ */
+static int afs_del_address_pref(struct afs_net *net, struct afs_addr_preference_list **_preflist,
+ int argc, char **argv)
+{
+ struct afs_addr_preference_list *preflist = *_preflist;
+ struct afs_addr_preference pref;
+ enum cmp_ret cmp;
+ int ret, i, stop;
+
+ if (argc != 2) {
+ pr_warn("Wrong number of params\n");
+ return -EINVAL;
+ }
+
+ if (strcmp(argv[0], "udp") != 0) {
+ pr_warn("Unsupported protocol\n");
+ return -EINVAL;
+ }
+
+ ret = afs_parse_address(argv[1], &pref);
+ if (ret < 0)
+ return ret;
+
+ if (pref.family == AF_INET) {
+ i = 0;
+ stop = preflist->ipv6_off;
+ } else {
+ i = preflist->ipv6_off;
+ stop = preflist->nr;
+ }
+
+ for (; i < stop; i++) {
+ cmp = afs_cmp_address_pref(&pref, &preflist->prefs[i]);
+ switch (cmp) {
+ case CONTINUE_SEARCH:
+ continue;
+ case INSERT_HERE:
+ case SUBNET_MATCH:
+ return 0;
+ case EXACT_MATCH:
+ return afs_delete_address_pref(_preflist, i);
+ }
+ }
+
+ return -ENOANO;
+}
+
+/*
+ * Handle writes to /proc/fs/afs/addr_prefs
+ */
+int afs_proc_addr_prefs_write(struct file *file, char *buf, size_t size)
+{
+ struct afs_addr_preference_list *preflist, *old;
+ struct seq_file *m = file->private_data;
+ struct afs_net *net = afs_seq2net_single(m);
+ size_t psize;
+ char *argv[5];
+ int ret, argc, max_prefs;
+
+ inode_lock(file_inode(file));
+
+ /* Allocate a candidate new list and initialise it from the old. */
+ old = rcu_dereference_protected(net->address_prefs,
+ lockdep_is_held(&file_inode(file)->i_rwsem));
+
+ if (old)
+ max_prefs = old->nr + 1;
+ else
+ max_prefs = 1;
+
+ psize = struct_size(old, prefs, max_prefs);
+ psize = roundup_pow_of_two(psize);
+ max_prefs = min_t(size_t, (psize - sizeof(*old)) / sizeof(old->prefs[0]), 255);
+
+ ret = -ENOMEM;
+ preflist = kmalloc(struct_size(preflist, prefs, max_prefs), GFP_KERNEL);
+ if (!preflist)
+ goto done;
+
+ if (old)
+ memcpy(preflist, old, struct_size(preflist, prefs, old->nr));
+ else
+ memset(preflist, 0, sizeof(*preflist));
+ preflist->max_prefs = max_prefs;
+
+ do {
+ argc = afs_split_string(&buf, argv, ARRAY_SIZE(argv));
+ if (argc < 0)
+ return argc;
+ if (argc < 2)
+ goto inval;
+
+ if (strcmp(argv[0], "add") == 0)
+ ret = afs_add_address_pref(net, &preflist, argc - 1, argv + 1);
+ else if (strcmp(argv[0], "del") == 0)
+ ret = afs_del_address_pref(net, &preflist, argc - 1, argv + 1);
+ else
+ goto inval;
+ if (ret < 0)
+ goto done;
+ } while (*buf);
+
+ preflist->version++;
+ rcu_assign_pointer(net->address_prefs, preflist);
+ /* Store prefs before version */
+ smp_store_release(&net->address_pref_version, preflist->version);
+ kfree_rcu(old, rcu);
+ preflist = NULL;
+ ret = 0;
+
+done:
+ kfree(preflist);
+ inode_unlock(file_inode(file));
+ _leave(" = %d", ret);
+ return ret;
+
+inval:
+ pr_warn("Invalid Command\n");
+ ret = -EINVAL;
+ goto done;
+}
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index b6659136959a..8bafe3b7529d 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -72,6 +72,28 @@ enum afs_call_state {
AFS_CALL_COMPLETE, /* Completed or failed */
};
+/*
+ * Address preferences.
+ */
+struct afs_addr_preference {
+ union {
+ struct in_addr ipv4_addr; /* AF_INET address to compare against */
+ struct in6_addr ipv6_addr; /* AF_INET6 address to compare against */
+ };
+ sa_family_t family; /* Which address to use */
+ u16 prio; /* Priority */
+ u8 subnet_mask; /* How many bits to compare */
+};
+
+struct afs_addr_preference_list {
+ struct rcu_head rcu;
+ u16 version; /* Incremented when prefs list changes */
+ u8 ipv6_off; /* Offset of IPv6 addresses */
+ u8 nr; /* Number of addresses in total */
+ u8 max_prefs; /* Number of prefs allocated */
+ struct afs_addr_preference prefs[] __counted_by(max_prefs);
+};
+
struct afs_address {
struct rxrpc_peer *peer;
short last_error; /* Last error from this address */
@@ -315,6 +337,8 @@ struct afs_net {
struct proc_dir_entry *proc_afs; /* /proc/net/afs directory */
struct afs_sysnames *sysnames;
rwlock_t sysnames_lock;
+ struct afs_addr_preference_list __rcu *address_prefs;
+ u16 address_pref_version;
/* Statistics counters */
atomic_t n_lookup; /* Number of lookups done */
@@ -981,6 +1005,11 @@ extern int afs_merge_fs_addr4(struct afs_net *net, struct afs_addr_list *addr,
extern int afs_merge_fs_addr6(struct afs_net *net, struct afs_addr_list *addr,
__be32 *xdr, u16 port);
+/*
+ * addr_prefs.c
+ */
+int afs_proc_addr_prefs_write(struct file *file, char *buf, size_t size);
+
/*
* callback.c
*/
diff --git a/fs/afs/main.c b/fs/afs/main.c
index eae288c8d40a..ab59fdf38ab5 100644
--- a/fs/afs/main.c
+++ b/fs/afs/main.c
@@ -158,6 +158,7 @@ static void __net_exit afs_net_exit(struct net *net_ns)
afs_close_socket(net);
afs_proc_cleanup(net);
afs_put_sysnames(net->sysnames);
+ kfree_rcu(net->address_prefs, rcu);
}
static struct pernet_operations afs_net_ops = {
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index 0b43bb9b0260..2e63c99a4f1e 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -146,6 +146,55 @@ static int afs_proc_cells_write(struct file *file, char *buf, size_t size)
goto done;
}
+/*
+ * Display the list of addr_prefs known to the namespace.
+ */
+static int afs_proc_addr_prefs_show(struct seq_file *m, void *v)
+{
+ struct afs_addr_preference_list *preflist;
+ struct afs_addr_preference *pref;
+ struct afs_net *net = afs_seq2net_single(m);
+ union {
+ struct sockaddr_in sin;
+ struct sockaddr_in6 sin6;
+ } addr;
+ unsigned int i;
+ char buf[44]; /* Maximum ipv6 + max subnet is 43 */
+
+ rcu_read_lock();
+ preflist = rcu_dereference(net->address_prefs);
+
+ if (!preflist) {
+ seq_puts(m, "NO PREFS\n");
+ return 0;
+ }
+
+ seq_printf(m, "PROT SUBNET PRIOR (v=%u n=%u/%u/%u)\n",
+ preflist->version, preflist->ipv6_off, preflist->nr, preflist->max_prefs);
+
+ memset(&addr, 0, sizeof(addr));
+
+ for (i = 0; i < preflist->nr; i++) {
+ pref = &preflist->prefs[i];
+
+ addr.sin.sin_family = pref->family;
+ if (pref->family == AF_INET) {
+ memcpy(&addr.sin.sin_addr, &pref->ipv4_addr,
+ sizeof(addr.sin.sin_addr));
+ snprintf(buf, sizeof(buf), "%pISc/%u", &addr.sin, pref->subnet_mask);
+ seq_printf(m, "UDP %-43.43s %5u\n", buf, pref->prio);
+ } else {
+ memcpy(&addr.sin6.sin6_addr, &pref->ipv6_addr,
+ sizeof(addr.sin6.sin6_addr));
+ snprintf(buf, sizeof(buf), "%pISc/%u", &addr.sin6, pref->subnet_mask);
+ seq_printf(m, "UDP %-43.43s %5u\n", buf, pref->prio);
+ }
+ }
+
+ rcu_read_lock();
+ return 0;
+}
+
/*
* Display the name of the current workstation cell.
*/
@@ -690,7 +739,11 @@ int afs_proc_init(struct afs_net *net)
&afs_proc_sysname_ops,
afs_proc_sysname_write,
sizeof(struct seq_net_private),
- NULL))
+ NULL) ||
+ !proc_create_net_single_write("addr_prefs", 0644, p,
+ afs_proc_addr_prefs_show,
+ afs_proc_addr_prefs_write,
+ NULL))
goto error_tree;
net->proc_afs = p;
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 27/41] afs: Mark address lists with configured priorities
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (25 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 26/41] afs: Provide a way to configure address priorities David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 28/41] afs: Dispatch fileserver probes in priority order David Howells
` (13 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Add a field to each address in an address list (afs_addr_list struct) that
records the current priority for that address according to the address
preference table. We don't want to do this every time we use an address
list, so the version number of the address preference table is recorded in
the address list too and we only re-mark the list when we see the version
change.
These numbers are then displayed through /proc/net/afs/servers.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/addr_prefs.c | 82 +++++++++++++++++++++++++++++++++++++++++++++
fs/afs/internal.h | 4 +++
fs/afs/proc.c | 9 ++---
3 files changed, 91 insertions(+), 4 deletions(-)
diff --git a/fs/afs/addr_prefs.c b/fs/afs/addr_prefs.c
index c6dcff4f8aa1..a189ff8a5034 100644
--- a/fs/afs/addr_prefs.c
+++ b/fs/afs/addr_prefs.c
@@ -447,3 +447,85 @@ int afs_proc_addr_prefs_write(struct file *file, char *buf, size_t size)
ret = -EINVAL;
goto done;
}
+
+/*
+ * Mark the priorities on an address list if the address preferences table has
+ * changed. The caller must hold the RCU read lock.
+ */
+void afs_get_address_preferences_rcu(struct afs_net *net, struct afs_addr_list *alist)
+{
+ const struct afs_addr_preference_list *preflist =
+ rcu_dereference(net->address_prefs);
+ const struct sockaddr_in6 *sin6;
+ const struct sockaddr_in *sin;
+ const struct sockaddr *sa;
+ struct afs_addr_preference test;
+ enum cmp_ret cmp;
+ int i, j;
+
+ if (!preflist || !preflist->nr || !alist->nr_addrs ||
+ smp_load_acquire(&alist->addr_pref_version) == preflist->version)
+ return;
+
+ test.family = AF_INET;
+ test.subnet_mask = 32;
+ test.prio = 0;
+ for (i = 0; i < alist->nr_ipv4; i++) {
+ sa = rxrpc_kernel_remote_addr(alist->addrs[i].peer);
+ sin = (const struct sockaddr_in *)sa;
+ test.ipv4_addr = sin->sin_addr;
+ for (j = 0; j < preflist->ipv6_off; j++) {
+ cmp = afs_cmp_address_pref(&test, &preflist->prefs[j]);
+ switch (cmp) {
+ case CONTINUE_SEARCH:
+ continue;
+ case INSERT_HERE:
+ break;
+ case EXACT_MATCH:
+ case SUBNET_MATCH:
+ WRITE_ONCE(alist->addrs[i].prio, preflist->prefs[j].prio);
+ break;
+ }
+ }
+ }
+
+ test.family = AF_INET6;
+ test.subnet_mask = 128;
+ test.prio = 0;
+ for (; i < alist->nr_addrs; i++) {
+ sa = rxrpc_kernel_remote_addr(alist->addrs[i].peer);
+ sin6 = (const struct sockaddr_in6 *)sa;
+ test.ipv6_addr = sin6->sin6_addr;
+ for (j = preflist->ipv6_off; j < preflist->nr; j++) {
+ cmp = afs_cmp_address_pref(&test, &preflist->prefs[j]);
+ switch (cmp) {
+ case CONTINUE_SEARCH:
+ continue;
+ case INSERT_HERE:
+ break;
+ case EXACT_MATCH:
+ case SUBNET_MATCH:
+ WRITE_ONCE(alist->addrs[i].prio, preflist->prefs[j].prio);
+ break;
+ }
+ }
+ }
+
+ smp_store_release(&alist->addr_pref_version, preflist->version);
+}
+
+/*
+ * Mark the priorities on an address list if the address preferences table has
+ * changed. Avoid taking the RCU read lock if we can.
+ */
+void afs_get_address_preferences(struct afs_net *net, struct afs_addr_list *alist)
+{
+ if (!net->address_prefs ||
+ /* Load version before prefs */
+ smp_load_acquire(&net->address_pref_version) == alist->addr_pref_version)
+ return;
+
+ rcu_read_lock();
+ afs_get_address_preferences_rcu(net, alist);
+ rcu_read_unlock();
+}
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 8bafe3b7529d..90b82e50da18 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -97,6 +97,7 @@ struct afs_addr_preference_list {
struct afs_address {
struct rxrpc_peer *peer;
short last_error; /* Last error from this address */
+ u16 prio; /* Address priority */
};
/*
@@ -107,6 +108,7 @@ struct afs_addr_list {
refcount_t usage;
u32 version; /* Version */
unsigned int debug_id;
+ unsigned int addr_pref_version; /* Version of address preference list */
unsigned char max_addrs;
unsigned char nr_addrs;
unsigned char preferred; /* Preferred address */
@@ -1009,6 +1011,8 @@ extern int afs_merge_fs_addr6(struct afs_net *net, struct afs_addr_list *addr,
* addr_prefs.c
*/
int afs_proc_addr_prefs_write(struct file *file, char *buf, size_t size);
+void afs_get_address_preferences_rcu(struct afs_net *net, struct afs_addr_list *alist);
+void afs_get_address_preferences(struct afs_net *net, struct afs_addr_list *alist);
/*
* callback.c
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index 2e63c99a4f1e..944eb51e75a1 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -447,17 +447,18 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
(int)(jiffies - server->probed_at) / HZ,
atomic_read(&server->probe_outstanding));
failed = alist->probe_failed;
- seq_printf(m, " - ALIST v=%u rsp=%lx f=%lx\n",
- alist->version, alist->responded, alist->probe_failed);
+ seq_printf(m, " - ALIST v=%u rsp=%lx f=%lx ap=%u\n",
+ alist->version, alist->responded, alist->probe_failed,
+ alist->addr_pref_version);
for (i = 0; i < alist->nr_addrs; i++) {
const struct afs_address *addr = &alist->addrs[i];
- seq_printf(m, " [%x] %pISpc%s rtt=%d err=%d\n",
+ seq_printf(m, " [%x] %pISpc%s rtt=%d err=%d p=%u\n",
i, rxrpc_kernel_remote_addr(addr->peer),
alist->preferred == i ? "*" :
test_bit(i, &failed) ? "!" : "",
rxrpc_kernel_get_srtt(addr->peer),
- addr->last_error);
+ addr->last_error, addr->prio);
}
return 0;
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 28/41] afs: Dispatch fileserver probes in priority order
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (26 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 27/41] afs: Mark address lists with configured priorities David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 29/41] afs: Dispatch vlserver " David Howells
` (12 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
When probing all the addresses for a fileserver, dispatch them in order of
descending priority to try and get back highest priority one first.
Also add a tracepoint to show the transmission and completion of the
probes.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/fs_probe.c | 25 +++++++++++++++++++++++--
include/trace/events/afs.h | 33 +++++++++++++++++++++++++++++++++
2 files changed, 56 insertions(+), 2 deletions(-)
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index 8008d3ecabab..c5702698b18b 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -102,7 +102,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
struct afs_address *addr = &alist->addrs[call->probe_index];
struct afs_server *server = call->server;
unsigned int index = call->probe_index;
- unsigned int rtt_us = 0, cap0;
+ unsigned int rtt_us = -1, cap0;
int ret = call->error;
_enter("%pU,%u", &server->uuid, index);
@@ -182,6 +182,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
out:
spin_unlock(&server->probe_lock);
+ trace_afs_fs_probe(server, false, alist, index, call->error, call->abort_code, rtt_us);
_debug("probe %pU [%u] %pISpc rtt=%d ret=%d",
&server->uuid, index, rxrpc_kernel_remote_addr(alist->addrs[index].peer),
rtt_us, ret);
@@ -207,6 +208,8 @@ void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
afs_get_addrlist(alist, afs_alist_trace_get_probe);
read_unlock(&server->fs_lock);
+ afs_get_address_preferences(net, alist);
+
server->probed_at = jiffies;
atomic_set(&server->probe_outstanding, all ? alist->nr_addrs : 1);
memset(&server->probe, 0, sizeof(server->probe));
@@ -217,10 +220,28 @@ void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
all = true;
if (all) {
- for (index = 0; index < alist->nr_addrs; index++)
+ unsigned long unprobed = (1UL << alist->nr_addrs) - 1;
+ unsigned int i;
+ int best_prio;
+
+ while (unprobed) {
+ best_prio = -1;
+ index = 0;
+ for (i = 0; i < alist->nr_addrs; i++) {
+ if (test_bit(i, &unprobed) &&
+ alist->addrs[i].prio > best_prio) {
+ index = i;
+ best_prio = alist->addrs[i].prio;
+ }
+ }
+ __clear_bit(index, &unprobed);
+
+ trace_afs_fs_probe(server, true, alist, index, 0, 0, 0);
if (!afs_fs_get_capabilities(net, server, alist, index, key))
afs_fs_probe_not_done(net, server, alist, index);
+ }
} else {
+ trace_afs_fs_probe(server, true, alist, index, 0, 0, 0);
if (!afs_fs_get_capabilities(net, server, alist, index, key))
afs_fs_probe_not_done(net, server, alist, index);
}
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 0f68d67f52c8..81eb87fbcfa7 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -1387,6 +1387,39 @@ TRACE_EVENT(afs_alist,
__entry->ref)
);
+TRACE_EVENT(afs_fs_probe,
+ TP_PROTO(struct afs_server *server, bool tx, struct afs_addr_list *alist,
+ unsigned int addr_index, int error, s32 abort_code, unsigned int rtt_us),
+
+ TP_ARGS(server, tx, alist, addr_index, error, abort_code, rtt_us),
+
+ TP_STRUCT__entry(
+ __field(unsigned int, server)
+ __field(bool, tx)
+ __field(u16, addr_index)
+ __field(short, error)
+ __field(s32, abort_code)
+ __field(unsigned int, rtt_us)
+ __field_struct(struct sockaddr_rxrpc, srx)
+ ),
+
+ TP_fast_assign(
+ __entry->server = server->debug_id;
+ __entry->tx = tx;
+ __entry->addr_index = addr_index;
+ __entry->error = error;
+ __entry->abort_code = abort_code;
+ __entry->rtt_us = rtt_us;
+ memcpy(&__entry->srx, rxrpc_kernel_remote_srx(alist->addrs[addr_index].peer),
+ sizeof(__entry->srx));
+ ),
+
+ TP_printk("s=%08x %s ax=%u e=%d ac=%d rtt=%d %pISpc",
+ __entry->server, __entry->tx ? "tx" : "rx", __entry->addr_index,
+ __entry->error, __entry->abort_code, __entry->rtt_us,
+ &__entry->srx.transport)
+ );
+
#endif /* _TRACE_AFS_H */
/* This part must be outside protection */
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 29/41] afs: Dispatch vlserver probes in priority order
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (27 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 28/41] afs: Dispatch fileserver probes in priority order David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 30/41] afs: Keep a record of the current fileserver endpoint state David Howells
` (11 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
When probing all the addresses for a volume location server, dispatch them
in order of descending priority to try and get back highest priority one
first.
Also add a tracepoint to show the transmission and completion of the
probes.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/internal.h | 1 +
fs/afs/vl_list.c | 2 ++
fs/afs/vl_probe.c | 20 ++++++++++++++++++--
include/trace/events/afs.h | 34 ++++++++++++++++++++++++++++++++++
4 files changed, 55 insertions(+), 2 deletions(-)
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 90b82e50da18..b88589bb4a6e 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -447,6 +447,7 @@ struct afs_vlserver {
rwlock_t lock; /* Lock on addresses */
refcount_t ref;
unsigned int rtt; /* Server's current RTT in uS */
+ unsigned int debug_id;
/* Probe state */
wait_queue_head_t probe_wq;
diff --git a/fs/afs/vl_list.c b/fs/afs/vl_list.c
index 5c4cd71caccf..9b1c20daac53 100644
--- a/fs/afs/vl_list.c
+++ b/fs/afs/vl_list.c
@@ -13,6 +13,7 @@ struct afs_vlserver *afs_alloc_vlserver(const char *name, size_t name_len,
unsigned short port)
{
struct afs_vlserver *vlserver;
+ static atomic_t debug_ids;
vlserver = kzalloc(struct_size(vlserver, name, name_len + 1),
GFP_KERNEL);
@@ -21,6 +22,7 @@ struct afs_vlserver *afs_alloc_vlserver(const char *name, size_t name_len,
rwlock_init(&vlserver->lock);
init_waitqueue_head(&vlserver->probe_wq);
spin_lock_init(&vlserver->probe_lock);
+ vlserver->debug_id = atomic_inc_return(&debug_ids);
vlserver->rtt = UINT_MAX;
vlserver->name_len = name_len;
vlserver->service_id = VL_SERVICE;
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index f868ae5d40e5..b128dc3d8af7 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -131,6 +131,7 @@ void afs_vlserver_probe_result(struct afs_call *call)
out:
spin_unlock(&server->probe_lock);
+ trace_afs_vl_probe(server, false, alist, index, call->error, call->abort_code, rtt_us);
_debug("probe [%u][%u] %pISpc rtt=%d ret=%d",
server_index, index, rxrpc_kernel_remote_addr(addr->peer),
rtt_us, ret);
@@ -150,8 +151,10 @@ static bool afs_do_probe_vlserver(struct afs_net *net,
{
struct afs_addr_list *alist;
struct afs_call *call;
- unsigned int index;
+ unsigned long unprobed;
+ unsigned int index, i;
bool in_progress = false;
+ int best_prio;
_enter("%s", server->name);
@@ -165,7 +168,20 @@ static bool afs_do_probe_vlserver(struct afs_net *net,
memset(&server->probe, 0, sizeof(server->probe));
server->probe.rtt = UINT_MAX;
- for (index = 0; index < alist->nr_addrs; index++) {
+ unprobed = (1UL << alist->nr_addrs) - 1;
+ while (unprobed) {
+ best_prio = -1;
+ index = 0;
+ for (i = 0; i < alist->nr_addrs; i++) {
+ if (test_bit(i, &unprobed) &&
+ alist->addrs[i].prio > best_prio) {
+ index = i;
+ best_prio = alist->addrs[i].prio;
+ }
+ }
+ __clear_bit(index, &unprobed);
+
+ trace_afs_vl_probe(server, true, alist, index, 0, 0, 0);
call = afs_vl_get_capabilities(net, alist, index, key, server,
server_index);
if (!IS_ERR(call)) {
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 81eb87fbcfa7..f1815b3dafb0 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -1420,6 +1420,40 @@ TRACE_EVENT(afs_fs_probe,
&__entry->srx.transport)
);
+TRACE_EVENT(afs_vl_probe,
+ TP_PROTO(struct afs_vlserver *server, bool tx, struct afs_addr_list *alist,
+ unsigned int addr_index, int error, s32 abort_code, unsigned int rtt_us),
+
+ TP_ARGS(server, tx, alist, addr_index, error, abort_code, rtt_us),
+
+ TP_STRUCT__entry(
+ __field(unsigned int, server)
+ __field(bool, tx)
+ __field(unsigned short, flags)
+ __field(u16, addr_index)
+ __field(short, error)
+ __field(s32, abort_code)
+ __field(unsigned int, rtt_us)
+ __field_struct(struct sockaddr_rxrpc, srx)
+ ),
+
+ TP_fast_assign(
+ __entry->server = server->debug_id;
+ __entry->tx = tx;
+ __entry->addr_index = addr_index;
+ __entry->error = error;
+ __entry->abort_code = abort_code;
+ __entry->rtt_us = rtt_us;
+ memcpy(&__entry->srx, rxrpc_kernel_remote_srx(alist->addrs[addr_index].peer),
+ sizeof(__entry->srx));
+ ),
+
+ TP_printk("vl=%08x %s ax=%u e=%d ac=%d rtt=%d %pISpc",
+ __entry->server, __entry->tx ? "tx" : "rx", __entry->addr_index,
+ __entry->error, __entry->abort_code, __entry->rtt_us,
+ &__entry->srx.transport)
+ );
+
#endif /* _TRACE_AFS_H */
/* This part must be outside protection */
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 30/41] afs: Keep a record of the current fileserver endpoint state
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (28 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 29/41] afs: Dispatch vlserver " David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 31/41] afs: Combine the endpoint state bools into a bitmask David Howells
` (10 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Keep a record of the current fileserver endpoint state, including the probe
state, and replace it when a new probe is started rather than just
squelching the old state and overwriting it. Clearance of the old state
can cause a race if there's another thread also currently trying to
communicate with that server.
It appears that this race might be the culprit for some occasions where
kafs complains about invalid data in the RPC reply because the rotation
algorithm fell all the way through without actually issuing an RPC call and
the error return got filled in from the probe state (which has a zero error
recorded). Whatever happens to be in the caller's reply buffer is then
taken as the response.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/fs_operation.c | 19 +--
fs/afs/fs_probe.c | 235 ++++++++++++++++++++++---------------
fs/afs/fsclient.c | 8 +-
fs/afs/internal.h | 66 +++++++----
fs/afs/proc.c | 21 ++--
fs/afs/rotate.c | 80 +++++++------
fs/afs/server.c | 69 +++++++----
fs/afs/vl_alias.c | 4 +-
fs/afs/vl_probe.c | 2 +-
fs/afs/vlclient.c | 4 +-
include/trace/events/afs.h | 69 ++++++++---
11 files changed, 366 insertions(+), 211 deletions(-)
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index e760e11d5bcb..8c6d827f999d 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -228,6 +228,7 @@ void afs_wait_for_operation(struct afs_operation *op)
*/
int afs_put_operation(struct afs_operation *op)
{
+ struct afs_endpoint_state *estate = op->estate;
struct afs_addr_list *alist;
int i, ret = afs_op_error(op);
@@ -251,14 +252,16 @@ int afs_put_operation(struct afs_operation *op)
kfree(op->more_files);
}
- alist = op->alist;
- if (alist) {
- if (op->call_responded &&
- op->addr_index != alist->preferred &&
- test_bit(alist->preferred, &op->addr_tried))
- WRITE_ONCE(alist->preferred, op->addr_index);
- afs_put_addrlist(alist, afs_alist_trace_put_operation);
- op->alist = NULL;
+ if (estate) {
+ alist = estate->addresses;
+ if (alist) {
+ if (op->call_responded &&
+ op->addr_index != alist->preferred &&
+ test_bit(alist->preferred, &op->addr_tried))
+ WRITE_ONCE(alist->preferred, op->addr_index);
+ }
+ afs_put_endpoint_state(estate, afs_estate_trace_put_operation);
+ op->estate = NULL;
}
afs_put_serverlist(op->net, op->server_list);
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index c5702698b18b..a669aee033c5 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -15,6 +15,42 @@
static unsigned int afs_fs_probe_fast_poll_interval = 30 * HZ;
static unsigned int afs_fs_probe_slow_poll_interval = 5 * 60 * HZ;
+struct afs_endpoint_state *afs_get_endpoint_state(struct afs_endpoint_state *estate,
+ enum afs_estate_trace where)
+{
+ if (estate) {
+ int r;
+
+ __refcount_inc(&estate->ref, &r);
+ trace_afs_estate(estate->server_id, estate->probe_seq, r, where);
+ }
+ return estate;
+}
+
+static void afs_endpoint_state_rcu(struct rcu_head *rcu)
+{
+ struct afs_endpoint_state *estate = container_of(rcu, struct afs_endpoint_state, rcu);
+
+ trace_afs_estate(estate->server_id, estate->probe_seq, refcount_read(&estate->ref),
+ afs_estate_trace_free);
+ afs_put_addrlist(estate->addresses, afs_alist_trace_put_estate);
+ kfree(estate);
+}
+
+void afs_put_endpoint_state(struct afs_endpoint_state *estate, enum afs_estate_trace where)
+{
+ if (estate) {
+ unsigned int server_id = estate->server_id, probe_seq = estate->probe_seq;
+ bool dead;
+ int r;
+
+ dead = __refcount_dec_and_test(&estate->ref, &r);
+ trace_afs_estate(server_id, probe_seq, r, where);
+ if (dead)
+ call_rcu(&estate->rcu, afs_endpoint_state_rcu);
+ }
+}
+
/*
* Start the probe polling timer. We have to supply it with an inc on the
* outstanding server count.
@@ -38,9 +74,10 @@ static void afs_schedule_fs_probe(struct afs_net *net,
/*
* Handle the completion of a set of probes.
*/
-static void afs_finished_fs_probe(struct afs_net *net, struct afs_server *server)
+static void afs_finished_fs_probe(struct afs_net *net, struct afs_server *server,
+ struct afs_endpoint_state *estate)
{
- bool responded = server->probe.responded;
+ bool responded = estate->responded;
write_seqlock(&net->fs_lock);
if (responded) {
@@ -50,6 +87,7 @@ static void afs_finished_fs_probe(struct afs_net *net, struct afs_server *server
clear_bit(AFS_SERVER_FL_RESPONDING, &server->flags);
list_add_tail(&server->probe_link, &net->fs_probe_fast);
}
+
write_sequnlock(&net->fs_lock);
afs_schedule_fs_probe(net, server, !responded);
@@ -58,12 +96,13 @@ static void afs_finished_fs_probe(struct afs_net *net, struct afs_server *server
/*
* Handle the completion of a probe.
*/
-static void afs_done_one_fs_probe(struct afs_net *net, struct afs_server *server)
+static void afs_done_one_fs_probe(struct afs_net *net, struct afs_server *server,
+ struct afs_endpoint_state *estate)
{
_enter("");
- if (atomic_dec_and_test(&server->probe_outstanding))
- afs_finished_fs_probe(net, server);
+ if (atomic_dec_and_test(&estate->nr_probing))
+ afs_finished_fs_probe(net, server, estate);
wake_up_all(&server->probe_wq);
}
@@ -74,7 +113,7 @@ static void afs_done_one_fs_probe(struct afs_net *net, struct afs_server *server
*/
static void afs_fs_probe_not_done(struct afs_net *net,
struct afs_server *server,
- struct afs_addr_list *alist,
+ struct afs_endpoint_state *estate,
int index)
{
_enter("");
@@ -82,14 +121,14 @@ static void afs_fs_probe_not_done(struct afs_net *net,
trace_afs_io_error(0, -ENOMEM, afs_io_error_fs_probe_fail);
spin_lock(&server->probe_lock);
- server->probe.local_failure = true;
- if (server->probe.error == 0)
- server->probe.error = -ENOMEM;
+ estate->local_failure = true;
+ if (estate->error == 0)
+ estate->error = -ENOMEM;
- set_bit(index, &alist->probe_failed);
+ set_bit(index, &estate->failed_set);
spin_unlock(&server->probe_lock);
- return afs_done_one_fs_probe(net, server);
+ return afs_done_one_fs_probe(net, server, estate);
}
/*
@@ -98,7 +137,8 @@ static void afs_fs_probe_not_done(struct afs_net *net,
*/
void afs_fileserver_probe_result(struct afs_call *call)
{
- struct afs_addr_list *alist = call->probe_alist;
+ struct afs_endpoint_state *estate = call->probe;
+ struct afs_addr_list *alist = estate->addresses;
struct afs_address *addr = &alist->addrs[call->probe_index];
struct afs_server *server = call->server;
unsigned int index = call->probe_index;
@@ -113,18 +153,18 @@ void afs_fileserver_probe_result(struct afs_call *call)
switch (ret) {
case 0:
- server->probe.error = 0;
+ estate->error = 0;
goto responded;
case -ECONNABORTED:
- if (!server->probe.responded) {
- server->probe.abort_code = call->abort_code;
- server->probe.error = ret;
+ if (!estate->responded) {
+ estate->abort_code = call->abort_code;
+ estate->error = ret;
}
goto responded;
case -ENOMEM:
case -ENONET:
- clear_bit(index, &alist->responded);
- server->probe.local_failure = true;
+ clear_bit(index, &estate->responsive_set);
+ estate->local_failure = true;
trace_afs_io_error(call->debug_id, ret, afs_io_error_fs_probe_fail);
goto out;
case -ECONNRESET: /* Responded, but call expired. */
@@ -137,28 +177,28 @@ void afs_fileserver_probe_result(struct afs_call *call)
case -ETIMEDOUT:
case -ETIME:
default:
- clear_bit(index, &alist->responded);
- set_bit(index, &alist->probe_failed);
- if (!server->probe.responded &&
- (server->probe.error == 0 ||
- server->probe.error == -ETIMEDOUT ||
- server->probe.error == -ETIME))
- server->probe.error = ret;
+ clear_bit(index, &estate->responsive_set);
+ set_bit(index, &estate->failed_set);
+ if (!estate->responded &&
+ (estate->error == 0 ||
+ estate->error == -ETIMEDOUT ||
+ estate->error == -ETIME))
+ estate->error = ret;
trace_afs_io_error(call->debug_id, ret, afs_io_error_fs_probe_fail);
goto out;
}
responded:
- clear_bit(index, &alist->probe_failed);
+ clear_bit(index, &estate->failed_set);
if (call->service_id == YFS_FS_SERVICE) {
- server->probe.is_yfs = true;
+ estate->is_yfs = true;
set_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
server->service_id = call->service_id;
} else {
- server->probe.not_yfs = true;
- if (!server->probe.is_yfs) {
- clear_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
+ estate->not_yfs = true;
+ if (!estate->is_yfs) {
+ estate->is_yfs = false;
server->service_id = call->service_id;
}
cap0 = ntohl(call->tmp);
@@ -169,84 +209,90 @@ void afs_fileserver_probe_result(struct afs_call *call)
}
rtt_us = rxrpc_kernel_get_srtt(addr->peer);
- if (rtt_us < server->probe.rtt) {
- server->probe.rtt = rtt_us;
+ if (rtt_us < estate->rtt) {
+ estate->rtt = rtt_us;
server->rtt = rtt_us;
alist->preferred = index;
}
smp_wmb(); /* Set rtt before responded. */
- server->probe.responded = true;
- set_bit(index, &alist->responded);
+ estate->responded = true;
+ set_bit(index, &estate->responsive_set);
set_bit(AFS_SERVER_FL_RESPONDING, &server->flags);
out:
spin_unlock(&server->probe_lock);
- trace_afs_fs_probe(server, false, alist, index, call->error, call->abort_code, rtt_us);
- _debug("probe %pU [%u] %pISpc rtt=%d ret=%d",
- &server->uuid, index, rxrpc_kernel_remote_addr(alist->addrs[index].peer),
+ trace_afs_fs_probe(server, false, estate, index, call->error, call->abort_code, rtt_us);
+ _debug("probe[%x] %pU [%u] %pISpc rtt=%d ret=%d",
+ estate->probe_seq, &server->uuid, index,
+ rxrpc_kernel_remote_addr(alist->addrs[index].peer),
rtt_us, ret);
- return afs_done_one_fs_probe(call->net, server);
+ return afs_done_one_fs_probe(call->net, server, estate);
}
/*
- * Probe one or all of a fileserver's addresses to find out the best route and
- * to query its capabilities.
+ * Probe all of a fileserver's addresses to find out the best route and to
+ * query its capabilities.
*/
void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
- struct key *key, bool all)
+ struct afs_addr_list *new_alist, struct key *key)
{
+ struct afs_endpoint_state *estate, *old;
struct afs_addr_list *alist;
- unsigned int index;
+ unsigned long unprobed;
_enter("%pU", &server->uuid);
- read_lock(&server->fs_lock);
- alist = rcu_dereference_protected(server->addresses,
- lockdep_is_held(&server->fs_lock));
- afs_get_addrlist(alist, afs_alist_trace_get_probe);
- read_unlock(&server->fs_lock);
+ estate = kzalloc(sizeof(*estate), GFP_KERNEL);
+ if (!estate)
+ return;
+
+ refcount_set(&estate->ref, 1);
+ estate->server_id = server->debug_id;
+ estate->rtt = UINT_MAX;
+
+ write_lock(&server->fs_lock);
+
+ old = rcu_dereference_protected(server->endpoint_state,
+ lockdep_is_held(&server->fs_lock));
+ estate->responsive_set = old->responsive_set;
+ estate->addresses = afs_get_addrlist(new_alist ?: old->addresses,
+ afs_alist_trace_get_estate);
+ alist = estate->addresses;
+ estate->probe_seq = ++server->probe_counter;
+ atomic_set(&estate->nr_probing, alist->nr_addrs);
+
+ rcu_assign_pointer(server->endpoint_state, estate);
+ old->superseded = true;
+ write_unlock(&server->fs_lock);
+
+ trace_afs_estate(estate->server_id, estate->probe_seq, refcount_read(&estate->ref),
+ afs_estate_trace_alloc_probe);
afs_get_address_preferences(net, alist);
server->probed_at = jiffies;
- atomic_set(&server->probe_outstanding, all ? alist->nr_addrs : 1);
- memset(&server->probe, 0, sizeof(server->probe));
- server->probe.rtt = UINT_MAX;
-
- index = alist->preferred;
- if (index < 0 || index >= alist->nr_addrs)
- all = true;
-
- if (all) {
- unsigned long unprobed = (1UL << alist->nr_addrs) - 1;
- unsigned int i;
- int best_prio;
-
- while (unprobed) {
- best_prio = -1;
- index = 0;
- for (i = 0; i < alist->nr_addrs; i++) {
- if (test_bit(i, &unprobed) &&
- alist->addrs[i].prio > best_prio) {
- index = i;
- best_prio = alist->addrs[i].prio;
- }
+ unprobed = (1UL << alist->nr_addrs) - 1;
+ while (unprobed) {
+ unsigned int index = 0, i;
+ int best_prio = -1;
+
+ for (i = 0; i < alist->nr_addrs; i++) {
+ if (test_bit(i, &unprobed) &&
+ alist->addrs[i].prio > best_prio) {
+ index = i;
+ best_prio = alist->addrs[i].prio;
}
- __clear_bit(index, &unprobed);
-
- trace_afs_fs_probe(server, true, alist, index, 0, 0, 0);
- if (!afs_fs_get_capabilities(net, server, alist, index, key))
- afs_fs_probe_not_done(net, server, alist, index);
}
- } else {
- trace_afs_fs_probe(server, true, alist, index, 0, 0, 0);
- if (!afs_fs_get_capabilities(net, server, alist, index, key))
- afs_fs_probe_not_done(net, server, alist, index);
+ __clear_bit(index, &unprobed);
+
+ trace_afs_fs_probe(server, true, estate, index, 0, 0, 0);
+ if (!afs_fs_get_capabilities(net, server, estate, index, key))
+ afs_fs_probe_not_done(net, server, estate, index);
}
- afs_put_addrlist(alist, afs_alist_trace_put_probe);
+ afs_put_endpoint_state(old, afs_estate_trace_put_probe);
}
/*
@@ -254,6 +300,7 @@ void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
*/
int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
{
+ struct afs_endpoint_state *estate;
struct wait_queue_entry *waits;
struct afs_server *server;
unsigned int rtt = UINT_MAX, rtt_s;
@@ -263,15 +310,18 @@ int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
_enter("%u,%lx", slist->nr_servers, untried);
/* Only wait for servers that have a probe outstanding. */
+ rcu_read_lock();
for (i = 0; i < slist->nr_servers; i++) {
if (test_bit(i, &untried)) {
server = slist->servers[i].server;
- if (!atomic_read(&server->probe_outstanding))
+ estate = rcu_dereference(server->endpoint_state);
+ if (!atomic_read(&estate->nr_probing))
__clear_bit(i, &untried);
- if (server->probe.responded)
+ if (estate->responded)
have_responders = true;
}
}
+ rcu_read_unlock();
if (have_responders || !untried)
return 0;
@@ -294,9 +344,9 @@ int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
for (i = 0; i < slist->nr_servers; i++) {
if (test_bit(i, &untried)) {
server = slist->servers[i].server;
- if (server->probe.responded)
+ if (estate->responded)
goto stop;
- if (atomic_read(&server->probe_outstanding))
+ if (atomic_read(&estate->nr_probing))
still_probing = true;
}
}
@@ -348,7 +398,7 @@ void afs_fs_probe_timer(struct timer_list *timer)
/*
* Dispatch a probe to a server.
*/
-static void afs_dispatch_fs_probe(struct afs_net *net, struct afs_server *server, bool all)
+static void afs_dispatch_fs_probe(struct afs_net *net, struct afs_server *server)
__releases(&net->fs_lock)
{
struct key *key = NULL;
@@ -361,7 +411,7 @@ static void afs_dispatch_fs_probe(struct afs_net *net, struct afs_server *server
afs_get_server(server, afs_server_trace_get_probe);
write_sequnlock(&net->fs_lock);
- afs_fs_probe_fileserver(net, server, key, all);
+ afs_fs_probe_fileserver(net, server, NULL, key);
afs_put_server(net, server, afs_server_trace_put_probe);
}
@@ -373,7 +423,7 @@ void afs_probe_fileserver(struct afs_net *net, struct afs_server *server)
{
write_seqlock(&net->fs_lock);
if (!list_empty(&server->probe_link))
- return afs_dispatch_fs_probe(net, server, true);
+ return afs_dispatch_fs_probe(net, server);
write_sequnlock(&net->fs_lock);
}
@@ -433,7 +483,7 @@ void afs_fs_probe_dispatcher(struct work_struct *work)
_debug("probe %pU", &server->uuid);
if (server && (first_pass || !need_resched())) {
- afs_dispatch_fs_probe(net, server, server == fast);
+ afs_dispatch_fs_probe(net, server);
first_pass = false;
goto again;
}
@@ -457,12 +507,13 @@ void afs_fs_probe_dispatcher(struct work_struct *work)
/*
* Wait for a probe on a particular fileserver to complete for 2s.
*/
-int afs_wait_for_one_fs_probe(struct afs_server *server, bool is_intr)
+int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_state *estate,
+ bool is_intr)
{
struct wait_queue_entry wait;
unsigned long timo = 2 * HZ;
- if (atomic_read(&server->probe_outstanding) == 0)
+ if (atomic_read(&estate->nr_probing) == 0)
goto dont_wait;
init_wait_entry(&wait, 0);
@@ -470,8 +521,8 @@ int afs_wait_for_one_fs_probe(struct afs_server *server, bool is_intr)
prepare_to_wait_event(&server->probe_wq, &wait,
is_intr ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
if (timo == 0 ||
- server->probe.responded ||
- atomic_read(&server->probe_outstanding) == 0 ||
+ estate->responded ||
+ atomic_read(&estate->nr_probing) == 0 ||
(is_intr && signal_pending(current)))
break;
timo = schedule_timeout(timo);
@@ -480,7 +531,7 @@ int afs_wait_for_one_fs_probe(struct afs_server *server, bool is_intr)
finish_wait(&server->probe_wq, &wait);
dont_wait:
- if (server->probe.responded)
+ if (estate->responded)
return 0;
if (is_intr && signal_pending(current))
return -ERESTARTSYS;
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 4f98b43b0dde..f1f879ba9cf7 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -1697,7 +1697,7 @@ static int afs_deliver_fs_get_capabilities(struct afs_call *call)
static void afs_fs_get_capabilities_destructor(struct afs_call *call)
{
- afs_put_addrlist(call->probe_alist, afs_alist_trace_put_getcaps);
+ afs_put_endpoint_state(call->probe, afs_estate_trace_put_getcaps);
afs_flat_call_destructor(call);
}
@@ -1719,7 +1719,7 @@ static const struct afs_call_type afs_RXFSGetCapabilities = {
* ->done() - otherwise we return false to indicate we didn't even try.
*/
bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
- struct afs_addr_list *alist, unsigned int addr_index,
+ struct afs_endpoint_state *estate, unsigned int addr_index,
struct key *key)
{
struct afs_call *call;
@@ -1733,8 +1733,8 @@ bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
call->key = key;
call->server = afs_use_server(server, afs_server_trace_get_caps);
- call->peer = rxrpc_kernel_get_peer(alist->addrs[addr_index].peer);
- call->probe_alist = afs_get_addrlist(alist, afs_alist_trace_get_getcaps);
+ call->peer = rxrpc_kernel_get_peer(estate->addresses->addrs[addr_index].peer);
+ call->probe = afs_get_endpoint_state(estate, afs_estate_trace_get_getcaps);
call->probe_index = addr_index;
call->service_id = server->service_id;
call->upgrade = true;
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index b88589bb4a6e..259efeeee235 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -33,6 +33,7 @@
struct pagevec;
struct afs_call;
struct afs_vnode;
+struct afs_server_probe;
/*
* Partial file-locking emulation mode. (The problem being that AFS3 only
@@ -146,14 +147,13 @@ struct afs_call {
};
void *buffer; /* reply receive buffer */
union {
- struct {
- struct afs_addr_list *probe_alist;
- unsigned char probe_index; /* Address in ->probe_alist */
- };
+ struct afs_endpoint_state *probe;
+ struct afs_addr_list *vl_probe;
struct afs_addr_list *ret_alist;
struct afs_vldb_entry *ret_vldb;
char *ret_str;
};
+ unsigned char probe_index; /* Address in ->probe_alist */
struct afs_operation *op;
unsigned int server_index;
refcount_t ref;
@@ -520,6 +520,32 @@ struct afs_vldb_entry {
u8 name[AFS_MAXVOLNAME + 1]; /* NUL-padded volume name */
};
+/*
+ * Fileserver endpoint state. The records the addresses of a fileserver's
+ * endpoints and the state and result of a round of probing on them. This
+ * allows the rotation algorithm to access those results without them being
+ * erased by a subsequent round of probing.
+ */
+struct afs_endpoint_state {
+ struct rcu_head rcu;
+ struct afs_addr_list *addresses; /* The addresses being probed */
+ unsigned long responsive_set; /* Bitset of responsive endpoints */
+ unsigned long failed_set; /* Bitset of endpoints we failed to probe */
+ refcount_t ref;
+ unsigned int server_id; /* Debug ID of server */
+ unsigned int probe_seq; /* Probe sequence (from server::probe_counter) */
+
+ atomic_t nr_probing; /* Number of outstanding probes */
+ unsigned int rtt; /* Best RTT in uS (or UINT_MAX) */
+ s32 abort_code;
+ short error;
+ bool responded:1;
+ bool is_yfs:1;
+ bool not_yfs:1;
+ bool local_failure:1;
+ bool superseded:1; /* Set if has been superseded */
+};
+
/*
* Record of fileserver with which we're actively communicating.
*/
@@ -530,7 +556,6 @@ struct afs_server {
struct afs_uuid _uuid;
};
- struct afs_addr_list __rcu *addresses;
struct afs_cell *cell; /* Cell to which belongs (pins ref) */
struct rb_node uuid_rb; /* Link in net->fs_servers */
struct afs_server __rcu *uuid_next; /* Next server with same UUID */
@@ -568,19 +593,11 @@ struct afs_server {
unsigned cb_s_break; /* Break-everything counter. */
/* Probe state */
+ struct afs_endpoint_state __rcu *endpoint_state; /* Latest endpoint/probe state */
unsigned long probed_at; /* Time last probe was dispatched (jiffies) */
wait_queue_head_t probe_wq;
- atomic_t probe_outstanding;
+ unsigned int probe_counter; /* Number of probes issued */
spinlock_t probe_lock;
- struct {
- unsigned int rtt; /* Best RTT in uS (or UINT_MAX) */
- u32 abort_code;
- short error;
- bool responded:1;
- bool is_yfs:1;
- bool not_yfs:1;
- bool local_failure:1;
- } probe;
};
/*
@@ -882,7 +899,7 @@ struct afs_operation {
/* Fileserver iteration state */
struct afs_server_list *server_list; /* Current server list (pins ref) */
struct afs_server *server; /* Server we're using (ref pinned by server_list) */
- struct afs_addr_list *alist; /* Current address list (pins ref) */
+ struct afs_endpoint_state *estate; /* Current endpoint state (pins ref) */
struct afs_call *call;
unsigned long untried_servers; /* Bitmask of untried servers */
unsigned long addr_tried; /* Tried addresses */
@@ -1152,7 +1169,7 @@ extern void afs_fs_release_lock(struct afs_operation *);
int afs_fs_give_up_all_callbacks(struct afs_net *net, struct afs_server *server,
struct afs_address *addr, struct key *key);
bool afs_fs_get_capabilities(struct afs_net *net, struct afs_server *server,
- struct afs_addr_list *alist, unsigned int addr_index,
+ struct afs_endpoint_state *estate, unsigned int addr_index,
struct key *key);
extern void afs_fs_inline_bulk_status(struct afs_operation *);
@@ -1189,12 +1206,17 @@ static inline void afs_op_set_fid(struct afs_operation *op, unsigned int n,
/*
* fs_probe.c
*/
+struct afs_endpoint_state *afs_get_endpoint_state(struct afs_endpoint_state *estate,
+ enum afs_estate_trace where);
+void afs_put_endpoint_state(struct afs_endpoint_state *estate, enum afs_estate_trace where);
extern void afs_fileserver_probe_result(struct afs_call *);
-extern void afs_fs_probe_fileserver(struct afs_net *, struct afs_server *, struct key *, bool);
+void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
+ struct afs_addr_list *new_addrs, struct key *key);
extern int afs_wait_for_fs_probes(struct afs_server_list *, unsigned long);
extern void afs_probe_fileserver(struct afs_net *, struct afs_server *);
extern void afs_fs_probe_dispatcher(struct work_struct *);
-extern int afs_wait_for_one_fs_probe(struct afs_server *, bool);
+int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_state *estate,
+ bool is_intr);
extern void afs_fs_probe_cleanup(struct afs_net *);
/*
@@ -1347,12 +1369,14 @@ extern int afs_protocol_error(struct afs_call *, enum afs_eproto_cause);
static inline void afs_make_op_call(struct afs_operation *op, struct afs_call *call,
gfp_t gfp)
{
+ struct afs_addr_list *alist = op->estate->addresses;
+
op->call = call;
op->type = call->type;
call->op = op;
call->key = op->key;
call->intr = !(op->flags & AFS_OPERATION_UNINTR);
- call->peer = rxrpc_kernel_get_peer(op->alist->addrs[op->addr_index].peer);
+ call->peer = rxrpc_kernel_get_peer(alist->addrs[op->addr_index].peer);
call->service_id = op->server->service_id;
afs_make_call(call, gfp);
}
@@ -1475,7 +1499,7 @@ extern void afs_manage_servers(struct work_struct *);
extern void afs_servers_timer(struct timer_list *);
extern void afs_fs_probe_timer(struct timer_list *);
extern void __net_exit afs_purge_servers(struct afs_net *);
-extern bool afs_check_server_record(struct afs_operation *, struct afs_server *);
+bool afs_check_server_record(struct afs_operation *op, struct afs_server *server, struct key *key);
static inline void afs_inc_servers_outstanding(struct afs_net *net)
{
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index 944eb51e75a1..a138022d8e0d 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -424,8 +424,9 @@ static const struct seq_operations afs_proc_cell_vlservers_ops = {
*/
static int afs_proc_servers_show(struct seq_file *m, void *v)
{
- struct afs_server *server;
+ struct afs_endpoint_state *estate;
struct afs_addr_list *alist;
+ struct afs_server *server;
unsigned long failed;
int i;
@@ -435,7 +436,8 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
}
server = list_entry(v, struct afs_server, proc_link);
- alist = rcu_dereference(server->addresses);
+ estate = rcu_dereference(server->endpoint_state);
+ alist = estate->addresses;
seq_printf(m, "%pU %3d %3d %s\n",
&server->uuid,
refcount_read(&server->ref),
@@ -443,13 +445,14 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
server->cell->name);
seq_printf(m, " - info: fl=%lx rtt=%u brk=%x\n",
server->flags, server->rtt, server->cb_s_break);
- seq_printf(m, " - probe: last=%d out=%d\n",
- (int)(jiffies - server->probed_at) / HZ,
- atomic_read(&server->probe_outstanding));
- failed = alist->probe_failed;
- seq_printf(m, " - ALIST v=%u rsp=%lx f=%lx ap=%u\n",
- alist->version, alist->responded, alist->probe_failed,
- alist->addr_pref_version);
+ seq_printf(m, " - probe: last=%d\n",
+ (int)(jiffies - server->probed_at) / HZ);
+ failed = estate->failed_set;
+ seq_printf(m, " - ESTATE pq=%x np=%u rsp=%lx f=%lx\n",
+ estate->probe_seq, atomic_read(&estate->nr_probing),
+ estate->responsive_set, estate->failed_set);
+ seq_printf(m, " - ALIST v=%u ap=%u\n",
+ alist->version, alist->addr_pref_version);
for (i = 0; i < alist->nr_addrs; i++) {
const struct afs_address *addr = &alist->addrs[i];
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 8b4151eb28be..56916558dd1d 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -109,10 +109,11 @@ static bool afs_sleep_and_retry(struct afs_operation *op)
*/
bool afs_select_fileserver(struct afs_operation *op)
{
- struct afs_addr_list *alist = op->alist;
+ struct afs_endpoint_state *estate = op->estate;
+ struct afs_addr_list *alist;
struct afs_server *server;
struct afs_vnode *vnode = op->file[0].vnode;
- unsigned long set;
+ unsigned long set, failed;
unsigned int rtt;
s32 abort_code = op->call_abort_code;
int error = op->call_error, addr_index, i;
@@ -133,7 +134,7 @@ bool afs_select_fileserver(struct afs_operation *op)
if (op->nr_iterations == 0)
goto start;
- WRITE_ONCE(alist->addrs[op->addr_index].last_error, error);
+ WRITE_ONCE(estate->addresses->addrs[op->addr_index].last_error, error);
/* Evaluate the result of the previous operation, if there was one. */
switch (op->call_error) {
@@ -399,14 +400,14 @@ bool afs_select_fileserver(struct afs_operation *op)
restart_from_beginning:
_debug("restart");
- afs_put_addrlist(alist, afs_alist_trace_put_restart_rotate);
- alist = op->alist = NULL;
+ afs_put_endpoint_state(estate, afs_estate_trace_put_restart_rotate);
+ estate = op->estate = NULL;
op->server = NULL;
afs_put_serverlist(op->net, op->server_list);
op->server_list = NULL;
start:
_debug("start");
- ASSERTCMP(alist, ==, NULL);
+ ASSERTCMP(estate, ==, NULL);
/* See if we need to do an update of the volume record. Note that the
* volume may have moved or even have been deleted.
*/
@@ -423,7 +424,7 @@ bool afs_select_fileserver(struct afs_operation *op)
pick_server:
_debug("pick [%lx]", op->untried_servers);
- ASSERTCMP(alist, ==, NULL);
+ ASSERTCMP(estate, ==, NULL);
error = afs_wait_for_fs_probes(op->server_list, op->untried_servers);
if (error < 0) {
@@ -450,9 +451,9 @@ bool afs_select_fileserver(struct afs_operation *op)
if (!test_bit(i, &op->untried_servers) ||
!test_bit(AFS_SERVER_FL_RESPONDING, &s->flags))
continue;
- if (s->probe.rtt <= rtt) {
+ if (s->rtt <= rtt) {
op->server_index = i;
- rtt = s->probe.rtt;
+ rtt = s->rtt;
}
}
@@ -467,10 +468,10 @@ bool afs_select_fileserver(struct afs_operation *op)
* check it, create a callback intercept, find its address list and
* probe its capabilities before we use it.
*/
- ASSERTCMP(alist, ==, NULL);
+ ASSERTCMP(estate, ==, NULL);
server = op->server_list->servers[op->server_index].server;
- if (!afs_check_server_record(op, server))
+ if (!afs_check_server_record(op, server, op->key))
goto failed;
_debug("USING SERVER: %pU", &server->uuid);
@@ -486,9 +487,9 @@ bool afs_select_fileserver(struct afs_operation *op)
}
read_lock(&server->fs_lock);
- alist = rcu_dereference_protected(server->addresses,
- lockdep_is_held(&server->fs_lock));
- op->alist = afs_get_addrlist(alist, afs_alist_trace_get_fsrotate_set);
+ estate = rcu_dereference_protected(server->endpoint_state,
+ lockdep_is_held(&server->fs_lock));
+ op->estate = afs_get_endpoint_state(estate, afs_estate_trace_get_fsrotate_set);
read_unlock(&server->fs_lock);
retry_server:
@@ -499,18 +500,20 @@ bool afs_select_fileserver(struct afs_operation *op)
/* Iterate over the current server's address list to try and find an
* address on which it will respond to us.
*/
- set = READ_ONCE(alist->responded);
- set &= ~(READ_ONCE(alist->probe_failed) | op->addr_tried);
+ set = READ_ONCE(estate->responsive_set);
+ failed = READ_ONCE(estate->failed_set);
+ _debug("iterate ES=%x rs=%lx fs=%lx", estate->probe_seq, set, failed);
+ set &= ~(failed | op->addr_tried);
if (!set)
goto out_of_addresses;
+ alist = estate->addresses;
addr_index = READ_ONCE(alist->preferred);
if (!test_bit(addr_index, &set))
addr_index = __ffs(set);
op->addr_index = addr_index;
set_bit(addr_index, &op->addr_tried);
- op->alist = alist;
op->call_responded = false;
_debug("address [%u] %u/%u %pISp",
@@ -525,8 +528,8 @@ bool afs_select_fileserver(struct afs_operation *op)
*/
afs_probe_fileserver(op->net, op->server);
if (op->flags & AFS_OPERATION_RETRY_SERVER) {
- error = afs_wait_for_one_fs_probe(
- op->server, !(op->flags & AFS_OPERATION_UNINTR));
+ error = afs_wait_for_one_fs_probe(op->server, estate,
+ !(op->flags & AFS_OPERATION_UNINTR));
switch (error) {
case 0:
op->flags &= ~AFS_OPERATION_RETRY_SERVER;
@@ -542,13 +545,14 @@ bool afs_select_fileserver(struct afs_operation *op)
next_server:
_debug("next");
- ASSERT(alist);
+ ASSERT(estate);
+ alist = estate->addresses;
if (op->call_responded &&
op->addr_index != READ_ONCE(alist->preferred) &&
test_bit(alist->preferred, &op->addr_tried))
WRITE_ONCE(alist->preferred, op->addr_index);
- afs_put_addrlist(alist, afs_alist_trace_put_next_server);
- alist = op->alist = NULL;
+ afs_put_endpoint_state(estate, afs_estate_trace_put_next_server);
+ estate = op->estate = NULL;
goto pick_server;
no_more_servers:
@@ -558,23 +562,28 @@ bool afs_select_fileserver(struct afs_operation *op)
if (op->flags & AFS_OPERATION_VBUSY)
goto restart_from_beginning;
+ rcu_read_lock();
for (i = 0; i < op->server_list->nr_servers; i++) {
+ struct afs_endpoint_state *estate;
struct afs_server *s = op->server_list->servers[i].server;
- error = READ_ONCE(s->probe.error);
+ estate = rcu_dereference(s->endpoint_state);
+ error = READ_ONCE(estate->error);
if (error < 0)
- afs_op_accumulate_error(op, error, s->probe.abort_code);
+ afs_op_accumulate_error(op, error, estate->abort_code);
}
+ rcu_read_unlock();
failed:
op->flags |= AFS_OPERATION_STOP;
- if (alist) {
+ if (estate) {
+ alist = estate->addresses;
if (op->call_responded &&
op->addr_index != READ_ONCE(alist->preferred) &&
test_bit(alist->preferred, &op->addr_tried))
WRITE_ONCE(alist->preferred, op->addr_index);
- afs_put_addrlist(alist, afs_alist_trace_put_op_failed);
- op->alist = NULL;
+ afs_put_endpoint_state(estate, afs_estate_trace_put_op_failed);
+ op->estate = NULL;
}
_leave(" = f [failed %d]", afs_op_error(op));
return false;
@@ -605,27 +614,30 @@ void afs_dump_edestaddrreq(const struct afs_operation *op)
if (op->server_list) {
const struct afs_server_list *sl = op->server_list;
+
pr_notice("FC: SL nr=%u pr=%u vnov=%hx\n",
sl->nr_servers, sl->preferred, sl->vnovol_mask);
for (i = 0; i < sl->nr_servers; i++) {
const struct afs_server *s = sl->servers[i].server;
+ const struct afs_endpoint_state *e =
+ rcu_dereference(s->endpoint_state);
+ const struct afs_addr_list *a = e->addresses;
+
pr_notice("FC: server fl=%lx av=%u %pU\n",
s->flags, s->addr_version, &s->uuid);
- if (s->addresses) {
- const struct afs_addr_list *a =
- rcu_dereference(s->addresses);
+ pr_notice("FC: - pq=%x R=%lx F=%lx\n",
+ e->probe_seq, e->responsive_set, e->failed_set);
+ if (a) {
pr_notice("FC: - av=%u nr=%u/%u/%u pr=%u\n",
a->version,
a->nr_ipv4, a->nr_addrs, a->max_addrs,
a->preferred);
- pr_notice("FC: - R=%lx F=%lx\n",
- a->responded, a->probe_failed);
- if (a == op->alist)
+ if (a == e->addresses)
pr_notice("FC: - current\n");
}
}
}
- pr_notice("AC: t=%lx ax=%u\n", op->addr_tried, op->addr_index);
+ pr_notice("AC: t=%lx ax=%d\n", op->addr_tried, op->addr_index);
rcu_read_unlock();
}
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 2c46f65b6bd6..9a524ca92636 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -23,6 +23,7 @@ static void __afs_put_server(struct afs_net *, struct afs_server *);
*/
struct afs_server *afs_find_server(struct afs_net *net, const struct rxrpc_peer *peer)
{
+ const struct afs_endpoint_state *estate;
const struct afs_addr_list *alist;
struct afs_server *server = NULL;
unsigned int i;
@@ -37,7 +38,8 @@ struct afs_server *afs_find_server(struct afs_net *net, const struct rxrpc_peer
read_seqbegin_or_lock(&net->fs_addr_lock, &seq);
hlist_for_each_entry_rcu(server, &net->fs_addresses6, addr6_link) {
- alist = rcu_dereference(server->addresses);
+ estate = rcu_dereference(server->endpoint_state);
+ alist = estate->addresses;
for (i = 0; i < alist->nr_addrs; i++)
if (alist->addrs[i].peer == peer)
goto found;
@@ -110,6 +112,7 @@ struct afs_server *afs_find_server_by_uuid(struct afs_net *net, const uuid_t *uu
static struct afs_server *afs_install_server(struct afs_cell *cell,
struct afs_server *candidate)
{
+ const struct afs_endpoint_state *estate;
const struct afs_addr_list *alist;
struct afs_server *server, *next;
struct afs_net *net = cell->net;
@@ -161,8 +164,9 @@ static struct afs_server *afs_install_server(struct afs_cell *cell,
added_dup:
write_seqlock(&net->fs_addr_lock);
- alist = rcu_dereference_protected(server->addresses,
- lockdep_is_held(&net->fs_addr_lock.lock));
+ estate = rcu_dereference_protected(server->endpoint_state,
+ lockdep_is_held(&net->fs_addr_lock.lock));
+ alist = estate->addresses;
/* Secondly, if the server has any IPv4 and/or IPv6 addresses, install
* it in the IPv4 and/or IPv6 reverse-map lists.
@@ -192,6 +196,7 @@ static struct afs_server *afs_alloc_server(struct afs_cell *cell,
const uuid_t *uuid,
struct afs_addr_list *alist)
{
+ struct afs_endpoint_state *estate;
struct afs_server *server;
struct afs_net *net = cell->net;
@@ -201,10 +206,13 @@ static struct afs_server *afs_alloc_server(struct afs_cell *cell,
if (!server)
goto enomem;
+ estate = kzalloc(sizeof(struct afs_endpoint_state), GFP_KERNEL);
+ if (!estate)
+ goto enomem_server;
+
refcount_set(&server->ref, 1);
atomic_set(&server->active, 1);
server->debug_id = atomic_inc_return(&afs_server_debug_id);
- RCU_INIT_POINTER(server->addresses, alist);
server->addr_version = alist->version;
server->uuid = *uuid;
rwlock_init(&server->fs_lock);
@@ -216,11 +224,23 @@ static struct afs_server *afs_alloc_server(struct afs_cell *cell,
server->rtt = UINT_MAX;
server->service_id = FS_SERVICE;
+ server->probe_counter = 1;
+ server->probed_at = jiffies - LONG_MAX / 2;
+ refcount_set(&estate->ref, 1);
+ estate->addresses = alist;
+ estate->server_id = server->debug_id;
+ estate->probe_seq = 1;
+ rcu_assign_pointer(server->endpoint_state, estate);
+
afs_inc_servers_outstanding(net);
trace_afs_server(server->debug_id, 1, 1, afs_server_trace_alloc);
+ trace_afs_estate(estate->server_id, estate->probe_seq, refcount_read(&estate->ref),
+ afs_estate_trace_alloc_server);
_leave(" = %p", server);
return server;
+enomem_server:
+ kfree(server);
enomem:
_leave(" = NULL [nomem]");
return NULL;
@@ -288,7 +308,7 @@ struct afs_server *afs_lookup_server(struct afs_cell *cell, struct key *key,
* on the fileserver. This will make sure the repeat-probing
* service is started.
*/
- afs_fs_probe_fileserver(cell->net, server, key, true);
+ afs_fs_probe_fileserver(cell->net, server, alist, key);
}
return server;
@@ -421,8 +441,8 @@ static void afs_server_rcu(struct rcu_head *rcu)
trace_afs_server(server->debug_id, refcount_read(&server->ref),
atomic_read(&server->active), afs_server_trace_free);
- afs_put_addrlist(rcu_access_pointer(server->addresses),
- afs_alist_trace_put_server);
+ afs_put_endpoint_state(rcu_access_pointer(server->endpoint_state),
+ afs_estate_trace_put_server);
kfree(server);
}
@@ -434,7 +454,8 @@ static void __afs_put_server(struct afs_net *net, struct afs_server *server)
static void afs_give_up_callbacks(struct afs_net *net, struct afs_server *server)
{
- struct afs_addr_list *alist = rcu_access_pointer(server->addresses);
+ struct afs_endpoint_state *estate = rcu_access_pointer(server->endpoint_state);
+ struct afs_addr_list *alist = estate->addresses;
afs_fs_give_up_all_callbacks(net, server, &alist->addrs[alist->preferred], NULL);
}
@@ -606,9 +627,12 @@ void afs_purge_servers(struct afs_net *net)
* Get an update for a server's address list.
*/
static noinline bool afs_update_server_record(struct afs_operation *op,
- struct afs_server *server)
+ struct afs_server *server,
+ struct key *key)
{
- struct afs_addr_list *alist, *discard;
+ struct afs_endpoint_state *estate;
+ struct afs_addr_list *alist;
+ bool has_addrs;
_enter("");
@@ -618,10 +642,15 @@ static noinline bool afs_update_server_record(struct afs_operation *op,
alist = afs_vl_lookup_addrs(op->volume->cell, op->key, &server->uuid);
if (IS_ERR(alist)) {
+ rcu_read_lock();
+ estate = rcu_dereference(server->endpoint_state);
+ has_addrs = estate->addresses;
+ rcu_read_unlock();
+
if ((PTR_ERR(alist) == -ERESTARTSYS ||
PTR_ERR(alist) == -EINTR) &&
(op->flags & AFS_OPERATION_UNINTR) &&
- server->addresses) {
+ has_addrs) {
_leave(" = t [intr]");
return true;
}
@@ -630,17 +659,10 @@ static noinline bool afs_update_server_record(struct afs_operation *op,
return false;
}
- discard = alist;
- if (server->addr_version != alist->version) {
- write_lock(&server->fs_lock);
- discard = rcu_dereference_protected(server->addresses,
- lockdep_is_held(&server->fs_lock));
- rcu_assign_pointer(server->addresses, alist);
- server->addr_version = alist->version;
- write_unlock(&server->fs_lock);
- }
+ if (server->addr_version != alist->version)
+ afs_fs_probe_fileserver(op->net, server, alist, key);
- afs_put_addrlist(discard, afs_alist_trace_put_server_update);
+ afs_put_addrlist(alist, afs_alist_trace_put_server_update);
_leave(" = t");
return true;
}
@@ -648,7 +670,8 @@ static noinline bool afs_update_server_record(struct afs_operation *op,
/*
* See if a server's address list needs updating.
*/
-bool afs_check_server_record(struct afs_operation *op, struct afs_server *server)
+bool afs_check_server_record(struct afs_operation *op, struct afs_server *server,
+ struct key *key)
{
bool success;
int ret, retries = 0;
@@ -668,7 +691,7 @@ bool afs_check_server_record(struct afs_operation *op, struct afs_server *server
update:
if (!test_and_set_bit_lock(AFS_SERVER_FL_UPDATING, &server->flags)) {
clear_bit(AFS_SERVER_FL_NEEDS_UPDATE, &server->flags);
- success = afs_update_server_record(op, server);
+ success = afs_update_server_record(op, server, key);
clear_bit_unlock(AFS_SERVER_FL_UPDATING, &server->flags);
wake_up_bit(&server->flags, AFS_SERVER_FL_UPDATING);
_leave(" = %d", success);
diff --git a/fs/afs/vl_alias.c b/fs/afs/vl_alias.c
index 89cadd9a69e1..43788d0c18e8 100644
--- a/fs/afs/vl_alias.c
+++ b/fs/afs/vl_alias.c
@@ -41,8 +41,8 @@ static int afs_compare_fs_alists(const struct afs_server *server_a,
const struct afs_addr_list *la, *lb;
int a = 0, b = 0, addr_matches = 0;
- la = rcu_dereference(server_a->addresses);
- lb = rcu_dereference(server_b->addresses);
+ la = rcu_dereference(server_a->endpoint_state)->addresses;
+ lb = rcu_dereference(server_b->endpoint_state)->addresses;
while (a < la->nr_addrs && b < lb->nr_addrs) {
unsigned long pa = (unsigned long)la->addrs[a].peer;
diff --git a/fs/afs/vl_probe.c b/fs/afs/vl_probe.c
index b128dc3d8af7..3d2e0c925460 100644
--- a/fs/afs/vl_probe.c
+++ b/fs/afs/vl_probe.c
@@ -46,7 +46,7 @@ static void afs_done_one_vl_probe(struct afs_vlserver *server, bool wake_up)
*/
void afs_vlserver_probe_result(struct afs_call *call)
{
- struct afs_addr_list *alist = call->probe_alist;
+ struct afs_addr_list *alist = call->vl_probe;
struct afs_vlserver *server = call->vlserver;
struct afs_address *addr = &alist->addrs[call->probe_index];
unsigned int server_index = call->server_index;
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index 39a0b7614d05..cef02a265edc 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -371,7 +371,7 @@ static int afs_deliver_vl_get_capabilities(struct afs_call *call)
static void afs_destroy_vl_get_capabilities(struct afs_call *call)
{
- afs_put_addrlist(call->probe_alist, afs_alist_trace_put_vlgetcaps);
+ afs_put_addrlist(call->vl_probe, afs_alist_trace_put_vlgetcaps);
afs_put_vlserver(call->net, call->vlserver);
afs_flat_call_destructor(call);
}
@@ -414,7 +414,7 @@ struct afs_call *afs_vl_get_capabilities(struct afs_net *net,
call->vlserver = afs_get_vlserver(server);
call->server_index = server_index;
call->peer = rxrpc_kernel_get_peer(alist->addrs[addr_index].peer);
- call->probe_alist = afs_get_addrlist(alist, afs_alist_trace_get_vlgetcaps);
+ call->vl_probe = afs_get_addrlist(alist, afs_alist_trace_get_vlgetcaps);
call->probe_index = addr_index;
call->service_id = server->service_id;
call->upgrade = true;
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index f1815b3dafb0..cf2fa4fddd5b 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -204,22 +204,14 @@ enum yfs_cm_operation {
#define afs_alist_traces \
EM(afs_alist_trace_alloc, "ALLOC ") \
- EM(afs_alist_trace_get_getcaps, "GET getcap") \
- EM(afs_alist_trace_get_fsrotate_set, "GET fs-rot") \
- EM(afs_alist_trace_get_probe, "GET probe ") \
+ EM(afs_alist_trace_get_estate, "GET estate") \
EM(afs_alist_trace_get_vlgetcaps, "GET vgtcap") \
EM(afs_alist_trace_get_vlprobe, "GET vprobe") \
EM(afs_alist_trace_get_vlrotate_set, "GET vl-rot") \
+ EM(afs_alist_trace_put_estate, "PUT estate") \
EM(afs_alist_trace_put_getaddru, "PUT GtAdrU") \
- EM(afs_alist_trace_put_getcaps, "PUT getcap") \
- EM(afs_alist_trace_put_next_server, "PUT nx-srv") \
- EM(afs_alist_trace_put_op_failed, "PUT op-fai") \
- EM(afs_alist_trace_put_operation, "PUT op ") \
EM(afs_alist_trace_put_parse_empty, "PUT p-empt") \
EM(afs_alist_trace_put_parse_error, "PUT p-err ") \
- EM(afs_alist_trace_put_probe, "PUT probe ") \
- EM(afs_alist_trace_put_restart_rotate, "PUT rstrot") \
- EM(afs_alist_trace_put_server, "PUT server") \
EM(afs_alist_trace_put_server_dup, "PUT sv-dup") \
EM(afs_alist_trace_put_server_oom, "PUT sv-oom") \
EM(afs_alist_trace_put_server_update, "PUT sv-upd") \
@@ -233,6 +225,20 @@ enum yfs_cm_operation {
EM(afs_alist_trace_put_vlserver_old, "PUT vs-old") \
E_(afs_alist_trace_free, "FREE ")
+#define afs_estate_traces \
+ EM(afs_estate_trace_alloc_probe, "ALLOC prob") \
+ EM(afs_estate_trace_alloc_server, "ALLOC srvr") \
+ EM(afs_estate_trace_get_fsrotate_set, "GET fs-rot") \
+ EM(afs_estate_trace_get_getcaps, "GET getcap") \
+ EM(afs_estate_trace_put_getcaps, "PUT getcap") \
+ EM(afs_estate_trace_put_next_server, "PUT nx-srv") \
+ EM(afs_estate_trace_put_op_failed, "PUT op-fai") \
+ EM(afs_estate_trace_put_operation, "PUT op ") \
+ EM(afs_estate_trace_put_probe, "PUT probe ") \
+ EM(afs_estate_trace_put_restart_rotate, "PUT rstrot") \
+ EM(afs_estate_trace_put_server, "PUT server") \
+ E_(afs_estate_trace_free, "FREE ")
+
#define afs_fs_operations \
EM(afs_FS_FetchData, "FS.FetchData") \
EM(afs_FS_FetchStatus, "FS.FetchStatus") \
@@ -458,6 +464,7 @@ enum afs_cell_trace { afs_cell_traces } __mode(byte);
enum afs_edit_dir_op { afs_edit_dir_ops } __mode(byte);
enum afs_edit_dir_reason { afs_edit_dir_reasons } __mode(byte);
enum afs_eproto_cause { afs_eproto_causes } __mode(byte);
+enum afs_estate_trace { afs_estate_traces } __mode(byte);
enum afs_file_error { afs_file_errors } __mode(byte);
enum afs_flock_event { afs_flock_events } __mode(byte);
enum afs_flock_operation { afs_flock_operations } __mode(byte);
@@ -486,6 +493,7 @@ yfs_cm_operations;
afs_edit_dir_ops;
afs_edit_dir_reasons;
afs_eproto_causes;
+afs_estate_traces;
afs_io_errors;
afs_file_errors;
afs_flock_types;
@@ -1387,14 +1395,43 @@ TRACE_EVENT(afs_alist,
__entry->ref)
);
+TRACE_EVENT(afs_estate,
+ TP_PROTO(unsigned int server_debug_id, unsigned int estate_debug_id,
+ int ref, enum afs_estate_trace reason),
+
+ TP_ARGS(server_debug_id, estate_debug_id, ref, reason),
+
+ TP_STRUCT__entry(
+ __field(unsigned int, server)
+ __field(unsigned int, estate)
+ __field(int, ref)
+ __field(int, active)
+ __field(int, reason)
+ ),
+
+ TP_fast_assign(
+ __entry->server = server_debug_id;
+ __entry->estate = estate_debug_id;
+ __entry->ref = ref;
+ __entry->reason = reason;
+ ),
+
+ TP_printk("ES=%08x[%x] %s r=%d",
+ __entry->server,
+ __entry->estate,
+ __print_symbolic(__entry->reason, afs_estate_traces),
+ __entry->ref)
+ );
+
TRACE_EVENT(afs_fs_probe,
- TP_PROTO(struct afs_server *server, bool tx, struct afs_addr_list *alist,
+ TP_PROTO(struct afs_server *server, bool tx, struct afs_endpoint_state *estate,
unsigned int addr_index, int error, s32 abort_code, unsigned int rtt_us),
- TP_ARGS(server, tx, alist, addr_index, error, abort_code, rtt_us),
+ TP_ARGS(server, tx, estate, addr_index, error, abort_code, rtt_us),
TP_STRUCT__entry(
__field(unsigned int, server)
+ __field(unsigned int, estate)
__field(bool, tx)
__field(u16, addr_index)
__field(short, error)
@@ -1404,7 +1441,9 @@ TRACE_EVENT(afs_fs_probe,
),
TP_fast_assign(
+ struct afs_addr_list *alist = estate->addresses;
__entry->server = server->debug_id;
+ __entry->estate = estate->probe_seq;
__entry->tx = tx;
__entry->addr_index = addr_index;
__entry->error = error;
@@ -1414,9 +1453,9 @@ TRACE_EVENT(afs_fs_probe,
sizeof(__entry->srx));
),
- TP_printk("s=%08x %s ax=%u e=%d ac=%d rtt=%d %pISpc",
- __entry->server, __entry->tx ? "tx" : "rx", __entry->addr_index,
- __entry->error, __entry->abort_code, __entry->rtt_us,
+ TP_printk("s=%08x %s pq=%x ax=%u e=%d ac=%d rtt=%d %pISpc",
+ __entry->server, __entry->tx ? "tx" : "rx", __entry->estate,
+ __entry->addr_index, __entry->error, __entry->abort_code, __entry->rtt_us,
&__entry->srx.transport)
);
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 31/41] afs: Combine the endpoint state bools into a bitmask
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (29 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 30/41] afs: Keep a record of the current fileserver endpoint state David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 32/41] afs: Fix file locking on R/O volumes to operate in local mode David Howells
` (9 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Combine the endpoint state bool-type members into a bitmask so that some of
them can be waited upon more easily.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/fs_probe.c | 30 +++++++++++++++---------------
fs/afs/internal.h | 12 ++++++------
2 files changed, 21 insertions(+), 21 deletions(-)
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index a669aee033c5..c00d38b98a67 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -77,7 +77,7 @@ static void afs_schedule_fs_probe(struct afs_net *net,
static void afs_finished_fs_probe(struct afs_net *net, struct afs_server *server,
struct afs_endpoint_state *estate)
{
- bool responded = estate->responded;
+ bool responded = test_bit(AFS_ESTATE_RESPONDED, &estate->flags);
write_seqlock(&net->fs_lock);
if (responded) {
@@ -121,7 +121,7 @@ static void afs_fs_probe_not_done(struct afs_net *net,
trace_afs_io_error(0, -ENOMEM, afs_io_error_fs_probe_fail);
spin_lock(&server->probe_lock);
- estate->local_failure = true;
+ set_bit(AFS_ESTATE_LOCAL_FAILURE, &estate->flags);
if (estate->error == 0)
estate->error = -ENOMEM;
@@ -156,7 +156,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
estate->error = 0;
goto responded;
case -ECONNABORTED:
- if (!estate->responded) {
+ if (!test_bit(AFS_ESTATE_RESPONDED, &estate->flags)) {
estate->abort_code = call->abort_code;
estate->error = ret;
}
@@ -164,7 +164,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
case -ENOMEM:
case -ENONET:
clear_bit(index, &estate->responsive_set);
- estate->local_failure = true;
+ set_bit(AFS_ESTATE_LOCAL_FAILURE, &estate->flags);
trace_afs_io_error(call->debug_id, ret, afs_io_error_fs_probe_fail);
goto out;
case -ECONNRESET: /* Responded, but call expired. */
@@ -179,7 +179,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
default:
clear_bit(index, &estate->responsive_set);
set_bit(index, &estate->failed_set);
- if (!estate->responded &&
+ if (!test_bit(AFS_ESTATE_RESPONDED, &estate->flags) &&
(estate->error == 0 ||
estate->error == -ETIMEDOUT ||
estate->error == -ETIME))
@@ -192,13 +192,13 @@ void afs_fileserver_probe_result(struct afs_call *call)
clear_bit(index, &estate->failed_set);
if (call->service_id == YFS_FS_SERVICE) {
- estate->is_yfs = true;
+ set_bit(AFS_ESTATE_IS_YFS, &estate->flags);
set_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
server->service_id = call->service_id;
} else {
- estate->not_yfs = true;
- if (!estate->is_yfs) {
- estate->is_yfs = false;
+ set_bit(AFS_ESTATE_NOT_YFS, &estate->flags);
+ if (!test_bit(AFS_ESTATE_IS_YFS, &estate->flags)) {
+ clear_bit(AFS_SERVER_FL_IS_YFS, &server->flags);
server->service_id = call->service_id;
}
cap0 = ntohl(call->tmp);
@@ -216,7 +216,7 @@ void afs_fileserver_probe_result(struct afs_call *call)
}
smp_wmb(); /* Set rtt before responded. */
- estate->responded = true;
+ set_bit(AFS_ESTATE_RESPONDED, &estate->flags);
set_bit(index, &estate->responsive_set);
set_bit(AFS_SERVER_FL_RESPONDING, &server->flags);
out:
@@ -264,7 +264,7 @@ void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
atomic_set(&estate->nr_probing, alist->nr_addrs);
rcu_assign_pointer(server->endpoint_state, estate);
- old->superseded = true;
+ set_bit(AFS_ESTATE_SUPERSEDED, &old->flags);
write_unlock(&server->fs_lock);
trace_afs_estate(estate->server_id, estate->probe_seq, refcount_read(&estate->ref),
@@ -317,7 +317,7 @@ int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
estate = rcu_dereference(server->endpoint_state);
if (!atomic_read(&estate->nr_probing))
__clear_bit(i, &untried);
- if (estate->responded)
+ if (test_bit(AFS_ESTATE_RESPONDED, &estate->flags))
have_responders = true;
}
}
@@ -344,7 +344,7 @@ int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
for (i = 0; i < slist->nr_servers; i++) {
if (test_bit(i, &untried)) {
server = slist->servers[i].server;
- if (estate->responded)
+ if (test_bit(AFS_ESTATE_RESPONDED, &estate->flags))
goto stop;
if (atomic_read(&estate->nr_probing))
still_probing = true;
@@ -521,7 +521,7 @@ int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_sta
prepare_to_wait_event(&server->probe_wq, &wait,
is_intr ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
if (timo == 0 ||
- estate->responded ||
+ test_bit(AFS_ESTATE_RESPONDED, &estate->flags) ||
atomic_read(&estate->nr_probing) == 0 ||
(is_intr && signal_pending(current)))
break;
@@ -531,7 +531,7 @@ int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_sta
finish_wait(&server->probe_wq, &wait);
dont_wait:
- if (estate->responded)
+ if (test_bit(AFS_ESTATE_RESPONDED, &estate->flags))
return 0;
if (is_intr && signal_pending(current))
return -ERESTARTSYS;
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 259efeeee235..8410ec4d6035 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -534,16 +534,16 @@ struct afs_endpoint_state {
refcount_t ref;
unsigned int server_id; /* Debug ID of server */
unsigned int probe_seq; /* Probe sequence (from server::probe_counter) */
-
atomic_t nr_probing; /* Number of outstanding probes */
unsigned int rtt; /* Best RTT in uS (or UINT_MAX) */
s32 abort_code;
short error;
- bool responded:1;
- bool is_yfs:1;
- bool not_yfs:1;
- bool local_failure:1;
- bool superseded:1; /* Set if has been superseded */
+ unsigned long flags;
+#define AFS_ESTATE_RESPONDED 0 /* Set if the server responded */
+#define AFS_ESTATE_SUPERSEDED 1 /* Set if this record has been superseded */
+#define AFS_ESTATE_IS_YFS 2 /* Set if probe upgraded to YFS */
+#define AFS_ESTATE_NOT_YFS 3 /* Set if probe didn't upgrade to YFS */
+#define AFS_ESTATE_LOCAL_FAILURE 4 /* Set if there was a local failure (eg. ENOMEM) */
};
/*
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 32/41] afs: Fix file locking on R/O volumes to operate in local mode
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (30 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 31/41] afs: Combine the endpoint state bools into a bitmask David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 33/41] afs: Mark a superblock for an R/O or Backup volume as SB_RDONLY David Howells
` (8 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
AFS doesn't really do locking on R/O volumes as fileservers don't maintain
state with each other and thus a lock on a R/O volume file on one
fileserver will not be be visible to someone looking at the same file on
another fileserver.
Further, the server may return an error if you try it.
Fix this by doing what other AFS clients do and handle filelocking on R/O
volume files entirely within the client and don't touch the server.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/super.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/fs/afs/super.c b/fs/afs/super.c
index 95d713074dc8..e95fb4cb4fcd 100644
--- a/fs/afs/super.c
+++ b/fs/afs/super.c
@@ -407,6 +407,8 @@ static int afs_validate_fc(struct fs_context *fc)
return PTR_ERR(volume);
ctx->volume = volume;
+ if (volume->type != AFSVL_RWVOL)
+ ctx->flock_mode = afs_flock_mode_local;
}
return 0;
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 33/41] afs: Mark a superblock for an R/O or Backup volume as SB_RDONLY
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (31 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 32/41] afs: Fix file locking on R/O volumes to operate in local mode David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 34/41] afs: Make it possible to find the volumes that are using a server David Howells
` (7 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Mark a superblock that is for for an R/O or Backup volume as SB_RDONLY when
mounting it.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/super.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/fs/afs/super.c b/fs/afs/super.c
index e95fb4cb4fcd..a01a0fb2cdbb 100644
--- a/fs/afs/super.c
+++ b/fs/afs/super.c
@@ -407,8 +407,10 @@ static int afs_validate_fc(struct fs_context *fc)
return PTR_ERR(volume);
ctx->volume = volume;
- if (volume->type != AFSVL_RWVOL)
+ if (volume->type != AFSVL_RWVOL) {
ctx->flock_mode = afs_flock_mode_local;
+ fc->sb_flags |= SB_RDONLY;
+ }
}
return 0;
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 34/41] afs: Make it possible to find the volumes that are using a server
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (32 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 33/41] afs: Mark a superblock for an R/O or Backup volume as SB_RDONLY David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 35/41] afs: Defer volume record destruction to a workqueue David Howells
` (6 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Make it possible to find the afs_volume structs that are using an
afs_server struct to aid in breaking volume callbacks.
The way this is done is that each afs_volume already has an array of
afs_server_entry records that point to the servers where that volume might
be found. An afs_volume backpointer and a list node is added to each entry
and each entry is then added to an RCU-traversable list on the afs_server
to which it points.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/cell.c | 1 +
fs/afs/internal.h | 23 +++++----
fs/afs/server.c | 1 +
fs/afs/server_list.c | 112 +++++++++++++++++++++++++++++++++++++++----
fs/afs/vl_alias.c | 2 +-
fs/afs/volume.c | 36 ++++++++------
6 files changed, 143 insertions(+), 32 deletions(-)
diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index 988c2ac7cece..69716fc0ee36 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -161,6 +161,7 @@ static struct afs_cell *afs_alloc_cell(struct afs_net *net,
refcount_set(&cell->ref, 1);
atomic_set(&cell->active, 0);
INIT_WORK(&cell->manager, afs_manage_cell_work);
+ spin_lock_init(&cell->vs_lock);
cell->volumes = RB_ROOT;
INIT_HLIST_HEAD(&cell->proc_volumes);
seqlock_init(&cell->volume_lock);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 8410ec4d6035..a085492a7167 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -414,6 +414,7 @@ struct afs_cell {
unsigned int debug_id;
/* The volumes belonging to this cell */
+ spinlock_t vs_lock; /* Lock for server->volumes */
struct rb_root volumes; /* Tree of volumes on this server */
struct hlist_head proc_volumes; /* procfs volume list */
seqlock_t volume_lock; /* For volumes */
@@ -564,6 +565,7 @@ struct afs_server {
struct hlist_node addr4_link; /* Link in net->fs_addresses4 */
struct hlist_node addr6_link; /* Link in net->fs_addresses6 */
struct hlist_node proc_link; /* Link in net->fs_proc */
+ struct list_head volumes; /* RCU list of afs_server_entry objects */
struct work_struct initcb_work; /* Work for CB.InitCallBackState* */
struct afs_server *gc_next; /* Next server in manager's list */
time64_t unuse_time; /* Time at which last unused */
@@ -605,12 +607,14 @@ struct afs_server {
*/
struct afs_server_entry {
struct afs_server *server;
+ struct afs_volume *volume;
+ struct list_head slink; /* Link in server->volumes */
};
struct afs_server_list {
struct rcu_head rcu;
- afs_volid_t vids[AFS_MAXTYPES]; /* Volume IDs */
refcount_t usage;
+ bool attached; /* T if attached to servers */
unsigned char nr_servers;
unsigned char preferred; /* Preferred server */
unsigned short vnovol_mask; /* Servers to be skipped due to VNOVOL */
@@ -623,10 +627,9 @@ struct afs_server_list {
* Live AFS volume management.
*/
struct afs_volume {
- union {
- struct rcu_head rcu;
- afs_volid_t vid; /* volume ID */
- };
+ struct rcu_head rcu;
+ afs_volid_t vid; /* The volume ID of this volume */
+ afs_volid_t vids[AFS_MAXTYPES]; /* All associated volume IDs */
refcount_t ref;
time64_t update_at; /* Time at which to next update */
struct afs_cell *cell; /* Cell to which belongs (pins ref) */
@@ -1527,10 +1530,14 @@ static inline struct afs_server_list *afs_get_serverlist(struct afs_server_list
}
extern void afs_put_serverlist(struct afs_net *, struct afs_server_list *);
-extern struct afs_server_list *afs_alloc_server_list(struct afs_cell *, struct key *,
- struct afs_vldb_entry *,
- u8);
+struct afs_server_list *afs_alloc_server_list(struct afs_volume *volume,
+ struct key *key,
+ struct afs_vldb_entry *vldb);
extern bool afs_annotate_server_list(struct afs_server_list *, struct afs_server_list *);
+void afs_attach_volume_to_servers(struct afs_volume *volume, struct afs_server_list *slist);
+void afs_reattach_volume_to_servers(struct afs_volume *volume, struct afs_server_list *slist,
+ struct afs_server_list *old);
+void afs_detach_volume_from_servers(struct afs_volume *volume, struct afs_server_list *slist);
/*
* super.c
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 9a524ca92636..811dbda88f9c 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -216,6 +216,7 @@ static struct afs_server *afs_alloc_server(struct afs_cell *cell,
server->addr_version = alist->version;
server->uuid = *uuid;
rwlock_init(&server->fs_lock);
+ INIT_LIST_HEAD(&server->volumes);
INIT_WORK(&server->initcb_work, afs_server_init_callback_work);
init_waitqueue_head(&server->probe_wq);
INIT_LIST_HEAD(&server->probe_link);
diff --git a/fs/afs/server_list.c b/fs/afs/server_list.c
index b59896b1de0a..4d6369477f54 100644
--- a/fs/afs/server_list.c
+++ b/fs/afs/server_list.c
@@ -24,13 +24,13 @@ void afs_put_serverlist(struct afs_net *net, struct afs_server_list *slist)
/*
* Build a server list from a VLDB record.
*/
-struct afs_server_list *afs_alloc_server_list(struct afs_cell *cell,
+struct afs_server_list *afs_alloc_server_list(struct afs_volume *volume,
struct key *key,
- struct afs_vldb_entry *vldb,
- u8 type_mask)
+ struct afs_vldb_entry *vldb)
{
struct afs_server_list *slist;
struct afs_server *server;
+ unsigned int type_mask = 1 << volume->type;
int ret = -ENOMEM, nr_servers = 0, i, j;
for (i = 0; i < vldb->nr_servers; i++)
@@ -44,15 +44,12 @@ struct afs_server_list *afs_alloc_server_list(struct afs_cell *cell,
refcount_set(&slist->usage, 1);
rwlock_init(&slist->lock);
- for (i = 0; i < AFS_MAXTYPES; i++)
- slist->vids[i] = vldb->vid[i];
-
/* Make sure a records exists for each server in the list. */
for (i = 0; i < vldb->nr_servers; i++) {
if (!(vldb->fs_mask[i] & type_mask))
continue;
- server = afs_lookup_server(cell, key, &vldb->fs_server[i],
+ server = afs_lookup_server(volume->cell, key, &vldb->fs_server[i],
vldb->addr_version[i]);
if (IS_ERR(server)) {
ret = PTR_ERR(server);
@@ -70,7 +67,7 @@ struct afs_server_list *afs_alloc_server_list(struct afs_cell *cell,
break;
if (j < slist->nr_servers) {
if (slist->servers[j].server == server) {
- afs_put_server(cell->net, server,
+ afs_put_server(volume->cell->net, server,
afs_server_trace_put_slist_isort);
continue;
}
@@ -81,6 +78,7 @@ struct afs_server_list *afs_alloc_server_list(struct afs_cell *cell,
}
slist->servers[j].server = server;
+ slist->servers[j].volume = volume;
slist->nr_servers++;
}
@@ -92,7 +90,7 @@ struct afs_server_list *afs_alloc_server_list(struct afs_cell *cell,
return slist;
error_2:
- afs_put_serverlist(cell->net, slist);
+ afs_put_serverlist(volume->cell->net, slist);
error:
return ERR_PTR(ret);
}
@@ -127,3 +125,99 @@ bool afs_annotate_server_list(struct afs_server_list *new,
return true;
}
+
+/*
+ * Attach a volume to the servers it is going to use.
+ */
+void afs_attach_volume_to_servers(struct afs_volume *volume, struct afs_server_list *slist)
+{
+ struct afs_server_entry *se, *pe;
+ struct afs_server *server;
+ struct list_head *p;
+ unsigned int i;
+
+ spin_lock(&volume->cell->vs_lock);
+
+ for (i = 0; i < slist->nr_servers; i++) {
+ se = &slist->servers[i];
+ server = se->server;
+
+ list_for_each(p, &server->volumes) {
+ pe = list_entry(p, struct afs_server_entry, slink);
+ if (volume->vid <= pe->volume->vid)
+ break;
+ }
+ list_add_tail_rcu(&se->slink, p);
+ }
+
+ slist->attached = true;
+ spin_unlock(&volume->cell->vs_lock);
+}
+
+/*
+ * Reattach a volume to the servers it is going to use when server list is
+ * replaced. We try to switch the attachment points to avoid rewalking the
+ * lists.
+ */
+void afs_reattach_volume_to_servers(struct afs_volume *volume, struct afs_server_list *new,
+ struct afs_server_list *old)
+{
+ unsigned int n = 0, o = 0;
+
+ spin_lock(&volume->cell->vs_lock);
+
+ while (n < new->nr_servers || o < old->nr_servers) {
+ struct afs_server_entry *pn = n < new->nr_servers ? &new->servers[n] : NULL;
+ struct afs_server_entry *po = o < old->nr_servers ? &old->servers[o] : NULL;
+ struct afs_server_entry *s;
+ struct list_head *p;
+ int diff;
+
+ if (pn && po && pn->server == po->server) {
+ list_replace_rcu(&po->slink, &pn->slink);
+ n++;
+ o++;
+ continue;
+ }
+
+ if (pn && po)
+ diff = memcmp(&pn->server->uuid, &po->server->uuid,
+ sizeof(pn->server->uuid));
+ else
+ diff = pn ? -1 : 1;
+
+ if (diff < 0) {
+ list_for_each(p, &pn->server->volumes) {
+ s = list_entry(p, struct afs_server_entry, slink);
+ if (volume->vid <= s->volume->vid)
+ break;
+ }
+ list_add_tail_rcu(&pn->slink, p);
+ n++;
+ } else {
+ list_del_rcu(&po->slink);
+ o++;
+ }
+ }
+
+ spin_unlock(&volume->cell->vs_lock);
+}
+
+/*
+ * Detach a volume from the servers it has been using.
+ */
+void afs_detach_volume_from_servers(struct afs_volume *volume, struct afs_server_list *slist)
+{
+ unsigned int i;
+
+ if (!slist->attached)
+ return;
+
+ spin_lock(&volume->cell->vs_lock);
+
+ for (i = 0; i < slist->nr_servers; i++)
+ list_del_rcu(&slist->servers[i].slink);
+
+ slist->attached = false;
+ spin_unlock(&volume->cell->vs_lock);
+}
diff --git a/fs/afs/vl_alias.c b/fs/afs/vl_alias.c
index 43788d0c18e8..63e7ed324af9 100644
--- a/fs/afs/vl_alias.c
+++ b/fs/afs/vl_alias.c
@@ -77,7 +77,7 @@ static int afs_compare_volume_slists(const struct afs_volume *vol_a,
lb = rcu_dereference(vol_b->servers);
for (i = 0; i < AFS_MAXTYPES; i++)
- if (la->vids[i] != lb->vids[i])
+ if (vol_a->vids[i] != vol_b->vids[i])
return 0;
while (a < la->nr_servers && b < lb->nr_servers) {
diff --git a/fs/afs/volume.c b/fs/afs/volume.c
index 29d483c80281..25d2b44c7aa6 100644
--- a/fs/afs/volume.c
+++ b/fs/afs/volume.c
@@ -66,11 +66,11 @@ static void afs_remove_volume_from_cell(struct afs_volume *volume)
*/
static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
struct afs_vldb_entry *vldb,
- unsigned long type_mask)
+ struct afs_server_list **_slist)
{
struct afs_server_list *slist;
struct afs_volume *volume;
- int ret = -ENOMEM;
+ int ret = -ENOMEM, i;
volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
if (!volume)
@@ -89,13 +89,16 @@ static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
rwlock_init(&volume->cb_v_break_lock);
memcpy(volume->name, vldb->name, vldb->name_len + 1);
- slist = afs_alloc_server_list(params->cell, params->key, vldb, type_mask);
+ for (i = 0; i < AFS_MAXTYPES; i++)
+ volume->vids[i] = vldb->vid[i];
+
+ slist = afs_alloc_server_list(volume, params->key, vldb);
if (IS_ERR(slist)) {
ret = PTR_ERR(slist);
goto error_1;
}
- refcount_set(&slist->usage, 1);
+ *_slist = slist;
rcu_assign_pointer(volume->servers, slist);
trace_afs_volume(volume->vid, 1, afs_volume_trace_alloc);
return volume;
@@ -111,17 +114,19 @@ static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
* Look up or allocate a volume record.
*/
static struct afs_volume *afs_lookup_volume(struct afs_fs_context *params,
- struct afs_vldb_entry *vldb,
- unsigned long type_mask)
+ struct afs_vldb_entry *vldb)
{
+ struct afs_server_list *slist;
struct afs_volume *candidate, *volume;
- candidate = afs_alloc_volume(params, vldb, type_mask);
+ candidate = afs_alloc_volume(params, vldb, &slist);
if (IS_ERR(candidate))
return candidate;
volume = afs_insert_volume_into_cell(params->cell, candidate);
- if (volume != candidate)
+ if (volume == candidate)
+ afs_attach_volume_to_servers(volume, slist);
+ else
afs_put_volume(params->net, candidate, afs_volume_trace_put_cell_dup);
return volume;
}
@@ -202,8 +207,7 @@ struct afs_volume *afs_create_volume(struct afs_fs_context *params)
goto error;
}
- type_mask = 1UL << params->type;
- volume = afs_lookup_volume(params, vldb, type_mask);
+ volume = afs_lookup_volume(params, vldb);
error:
kfree(vldb);
@@ -215,14 +219,17 @@ struct afs_volume *afs_create_volume(struct afs_fs_context *params)
*/
static void afs_destroy_volume(struct afs_net *net, struct afs_volume *volume)
{
+ struct afs_server_list *slist = rcu_access_pointer(volume->servers);
+
_enter("%p", volume);
#ifdef CONFIG_AFS_FSCACHE
ASSERTCMP(volume->cache, ==, NULL);
#endif
+ afs_detach_volume_from_servers(volume, slist);
afs_remove_volume_from_cell(volume);
- afs_put_serverlist(net, rcu_access_pointer(volume->servers));
+ afs_put_serverlist(net, slist);
afs_put_cell(volume->cell, afs_cell_trace_put_vol);
trace_afs_volume(volume->vid, refcount_read(&volume->ref),
afs_volume_trace_free);
@@ -342,8 +349,7 @@ static int afs_update_volume_status(struct afs_volume *volume, struct key *key)
}
/* See if the volume's server list got updated. */
- new = afs_alloc_server_list(volume->cell, key,
- vldb, (1 << volume->type));
+ new = afs_alloc_server_list(volume, key, vldb);
if (IS_ERR(new)) {
ret = PTR_ERR(new);
goto error_vldb;
@@ -364,9 +370,11 @@ static int afs_update_volume_status(struct afs_volume *volume, struct key *key)
volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
write_unlock(&volume->servers_lock);
- ret = 0;
+ if (discard == old)
+ afs_reattach_volume_to_servers(volume, new, old);
afs_put_serverlist(volume->cell->net, discard);
+ ret = 0;
error_vldb:
kfree(vldb);
error:
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 35/41] afs: Defer volume record destruction to a workqueue
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (33 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 34/41] afs: Make it possible to find the volumes that are using a server David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:39 ` [PATCH 36/41] afs: Move the vnode/volume validity checking code into its own file David Howells
` (5 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Defer volume record destruction to a workqueue so that afs_put_volume()
isn't going to run the destruction process in the callback workqueue whilst
the server is holding up other clients whilst waiting for us to reply to a
CB.CallBack notification RPC.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/cell.c | 2 +-
fs/afs/fs_operation.c | 2 +-
fs/afs/internal.h | 3 ++-
fs/afs/super.c | 7 +++----
fs/afs/vl_alias.c | 6 +++---
fs/afs/volume.c | 15 +++++++++------
6 files changed, 19 insertions(+), 16 deletions(-)
diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index 69716fc0ee36..6b389f2bcd0c 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -816,7 +816,7 @@ static void afs_manage_cell(struct afs_cell *cell)
final_destruction:
/* The root volume is pinning the cell */
- afs_put_volume(cell->net, cell->root_volume, afs_volume_trace_put_cell_root);
+ afs_put_volume(cell->root_volume, afs_volume_trace_put_cell_root);
cell->root_volume = NULL;
afs_put_cell(cell, afs_cell_trace_put_destroy);
}
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index 8c6d827f999d..10137681aa7d 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -265,7 +265,7 @@ int afs_put_operation(struct afs_operation *op)
}
afs_put_serverlist(op->net, op->server_list);
- afs_put_volume(op->net, op->volume, afs_volume_trace_put_put_op);
+ afs_put_volume(op->volume, afs_volume_trace_put_put_op);
key_put(op->key);
kfree(op);
return ret;
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index a085492a7167..22cf782fe924 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -636,6 +636,7 @@ struct afs_volume {
struct rb_node cell_node; /* Link in cell->volumes */
struct hlist_node proc_link; /* Link in cell->proc_volumes */
struct super_block __rcu *sb; /* Superblock on which inodes reside */
+ struct work_struct destructor; /* Deferred destructor */
unsigned long flags;
#define AFS_VOLUME_NEEDS_UPDATE 0 /* - T if an update needs performing */
#define AFS_VOLUME_UPDATING 1 /* - T if an update is in progress */
@@ -1611,7 +1612,7 @@ extern struct afs_volume *afs_create_volume(struct afs_fs_context *);
extern int afs_activate_volume(struct afs_volume *);
extern void afs_deactivate_volume(struct afs_volume *);
extern struct afs_volume *afs_get_volume(struct afs_volume *, enum afs_volume_trace);
-extern void afs_put_volume(struct afs_net *, struct afs_volume *, enum afs_volume_trace);
+void afs_put_volume(struct afs_volume *volume, enum afs_volume_trace reason);
extern int afs_check_volume_status(struct afs_volume *, struct afs_operation *);
/*
diff --git a/fs/afs/super.c b/fs/afs/super.c
index a01a0fb2cdbb..ae2d66a52add 100644
--- a/fs/afs/super.c
+++ b/fs/afs/super.c
@@ -381,8 +381,7 @@ static int afs_validate_fc(struct fs_context *fc)
ctx->key = key;
if (ctx->volume) {
- afs_put_volume(ctx->net, ctx->volume,
- afs_volume_trace_put_validate_fc);
+ afs_put_volume(ctx->volume, afs_volume_trace_put_validate_fc);
ctx->volume = NULL;
}
@@ -529,7 +528,7 @@ static void afs_destroy_sbi(struct afs_super_info *as)
{
if (as) {
struct afs_net *net = afs_net(as->net_ns);
- afs_put_volume(net, as->volume, afs_volume_trace_put_destroy_sbi);
+ afs_put_volume(as->volume, afs_volume_trace_put_destroy_sbi);
afs_unuse_cell(net, as->cell, afs_cell_trace_unuse_sbi);
put_net(as->net_ns);
kfree(as);
@@ -615,7 +614,7 @@ static void afs_free_fc(struct fs_context *fc)
struct afs_fs_context *ctx = fc->fs_private;
afs_destroy_sbi(fc->s_fs_info);
- afs_put_volume(ctx->net, ctx->volume, afs_volume_trace_put_free_fc);
+ afs_put_volume(ctx->volume, afs_volume_trace_put_free_fc);
afs_unuse_cell(ctx->net, ctx->cell, afs_cell_trace_unuse_fc);
key_put(ctx->key);
kfree(ctx);
diff --git a/fs/afs/vl_alias.c b/fs/afs/vl_alias.c
index 63e7ed324af9..9f36e14f1c2d 100644
--- a/fs/afs/vl_alias.c
+++ b/fs/afs/vl_alias.c
@@ -156,7 +156,7 @@ static int afs_query_for_alias_one(struct afs_cell *cell, struct key *key,
/* And see if it's in the new cell. */
volume = afs_sample_volume(cell, key, pvol->name, pvol->name_len);
if (IS_ERR(volume)) {
- afs_put_volume(cell->net, pvol, afs_volume_trace_put_query_alias);
+ afs_put_volume(pvol, afs_volume_trace_put_query_alias);
if (PTR_ERR(volume) != -ENOMEDIUM)
return PTR_ERR(volume);
/* That volume is not in the new cell, so not an alias */
@@ -174,8 +174,8 @@ static int afs_query_for_alias_one(struct afs_cell *cell, struct key *key,
rcu_read_unlock();
}
- afs_put_volume(cell->net, volume, afs_volume_trace_put_query_alias);
- afs_put_volume(cell->net, pvol, afs_volume_trace_put_query_alias);
+ afs_put_volume(volume, afs_volume_trace_put_query_alias);
+ afs_put_volume(pvol, afs_volume_trace_put_query_alias);
return ret;
}
diff --git a/fs/afs/volume.c b/fs/afs/volume.c
index 25d2b44c7aa6..1d3bd14dfa2f 100644
--- a/fs/afs/volume.c
+++ b/fs/afs/volume.c
@@ -11,6 +11,8 @@
static unsigned __read_mostly afs_volume_record_life = 60 * 60;
+static void afs_destroy_volume(struct work_struct *work);
+
/*
* Insert a volume into a cell. If there's an existing volume record, that is
* returned instead with a ref held.
@@ -85,6 +87,7 @@ static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
refcount_set(&volume->ref, 1);
INIT_HLIST_NODE(&volume->proc_link);
+ INIT_WORK(&volume->destructor, afs_destroy_volume);
rwlock_init(&volume->servers_lock);
rwlock_init(&volume->cb_v_break_lock);
memcpy(volume->name, vldb->name, vldb->name_len + 1);
@@ -127,7 +130,7 @@ static struct afs_volume *afs_lookup_volume(struct afs_fs_context *params,
if (volume == candidate)
afs_attach_volume_to_servers(volume, slist);
else
- afs_put_volume(params->net, candidate, afs_volume_trace_put_cell_dup);
+ afs_put_volume(candidate, afs_volume_trace_put_cell_dup);
return volume;
}
@@ -217,8 +220,9 @@ struct afs_volume *afs_create_volume(struct afs_fs_context *params)
/*
* Destroy a volume record
*/
-static void afs_destroy_volume(struct afs_net *net, struct afs_volume *volume)
+static void afs_destroy_volume(struct work_struct *work)
{
+ struct afs_volume *volume = container_of(work, struct afs_volume, destructor);
struct afs_server_list *slist = rcu_access_pointer(volume->servers);
_enter("%p", volume);
@@ -229,7 +233,7 @@ static void afs_destroy_volume(struct afs_net *net, struct afs_volume *volume)
afs_detach_volume_from_servers(volume, slist);
afs_remove_volume_from_cell(volume);
- afs_put_serverlist(net, slist);
+ afs_put_serverlist(volume->cell->net, slist);
afs_put_cell(volume->cell, afs_cell_trace_put_vol);
trace_afs_volume(volume->vid, refcount_read(&volume->ref),
afs_volume_trace_free);
@@ -257,8 +261,7 @@ struct afs_volume *afs_get_volume(struct afs_volume *volume,
/*
* Drop a reference on a volume record.
*/
-void afs_put_volume(struct afs_net *net, struct afs_volume *volume,
- enum afs_volume_trace reason)
+void afs_put_volume(struct afs_volume *volume, enum afs_volume_trace reason)
{
if (volume) {
afs_volid_t vid = volume->vid;
@@ -268,7 +271,7 @@ void afs_put_volume(struct afs_net *net, struct afs_volume *volume,
zero = __refcount_dec_and_test(&volume->ref, &r);
trace_afs_volume(vid, r - 1, reason);
if (zero)
- afs_destroy_volume(net, volume);
+ schedule_work(&volume->destructor);
}
}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 36/41] afs: Move the vnode/volume validity checking code into its own file
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (34 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 35/41] afs: Defer volume record destruction to a workqueue David Howells
@ 2023-11-09 15:39 ` David Howells
2023-11-09 15:40 ` [PATCH 37/41] afs: Apply server breaks to mmap'd files in the call processor David Howells
` (4 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:39 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Move the code that does validity checking of vnodes and volumes with
respect to third-party changes into its own file.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/Makefile | 1 +
fs/afs/inode.c | 174 -----------------------------------------
fs/afs/internal.h | 10 ++-
fs/afs/validation.c | 185 ++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 193 insertions(+), 177 deletions(-)
create mode 100644 fs/afs/validation.c
diff --git a/fs/afs/Makefile b/fs/afs/Makefile
index b3849bea0553..dcdc0f1bb76f 100644
--- a/fs/afs/Makefile
+++ b/fs/afs/Makefile
@@ -28,6 +28,7 @@ kafs-y := \
server.o \
server_list.o \
super.o \
+ validation.o \
vlclient.o \
vl_alias.o \
vl_list.o \
diff --git a/fs/afs/inode.c b/fs/afs/inode.c
index b3d76faa83c0..8dbb09a79c4d 100644
--- a/fs/afs/inode.c
+++ b/fs/afs/inode.c
@@ -572,180 +572,6 @@ struct inode *afs_root_iget(struct super_block *sb, struct key *key)
return ERR_PTR(ret);
}
-/*
- * mark the data attached to an inode as obsolete due to a write on the server
- * - might also want to ditch all the outstanding writes and dirty pages
- */
-static void afs_zap_data(struct afs_vnode *vnode)
-{
- _enter("{%llx:%llu}", vnode->fid.vid, vnode->fid.vnode);
-
- afs_invalidate_cache(vnode, 0);
-
- /* nuke all the non-dirty pages that aren't locked, mapped or being
- * written back in a regular file and completely discard the pages in a
- * directory or symlink */
- if (S_ISREG(vnode->netfs.inode.i_mode))
- invalidate_remote_inode(&vnode->netfs.inode);
- else
- invalidate_inode_pages2(vnode->netfs.inode.i_mapping);
-}
-
-/*
- * Check to see if we have a server currently serving this volume and that it
- * hasn't been reinitialised or dropped from the list.
- */
-static bool afs_check_server_good(struct afs_vnode *vnode)
-{
- struct afs_server_list *slist;
- struct afs_server *server;
- bool good;
- int i;
-
- if (vnode->cb_fs_s_break == atomic_read(&vnode->volume->cell->fs_s_break))
- return true;
-
- rcu_read_lock();
-
- slist = rcu_dereference(vnode->volume->servers);
- for (i = 0; i < slist->nr_servers; i++) {
- server = slist->servers[i].server;
- if (server == vnode->cb_server) {
- good = (vnode->cb_s_break == server->cb_s_break);
- rcu_read_unlock();
- return good;
- }
- }
-
- rcu_read_unlock();
- return false;
-}
-
-/*
- * Check the validity of a vnode/inode.
- */
-bool afs_check_validity(struct afs_vnode *vnode)
-{
- enum afs_cb_break_reason need_clear = afs_cb_break_no_break;
- time64_t now = ktime_get_real_seconds();
- unsigned int cb_break;
- int seq = 0;
-
- do {
- read_seqbegin_or_lock(&vnode->cb_lock, &seq);
- cb_break = vnode->cb_break;
-
- if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
- if (vnode->cb_v_break != vnode->volume->cb_v_break)
- need_clear = afs_cb_break_for_v_break;
- else if (!afs_check_server_good(vnode))
- need_clear = afs_cb_break_for_s_reinit;
- else if (test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
- need_clear = afs_cb_break_for_zap;
- else if (vnode->cb_expires_at - 10 <= now)
- need_clear = afs_cb_break_for_lapsed;
- } else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
- ;
- } else {
- need_clear = afs_cb_break_no_promise;
- }
-
- } while (need_seqretry(&vnode->cb_lock, seq));
-
- done_seqretry(&vnode->cb_lock, seq);
-
- if (need_clear == afs_cb_break_no_break)
- return true;
-
- write_seqlock(&vnode->cb_lock);
- if (need_clear == afs_cb_break_no_promise)
- vnode->cb_v_break = vnode->volume->cb_v_break;
- else if (cb_break == vnode->cb_break)
- __afs_break_callback(vnode, need_clear);
- else
- trace_afs_cb_miss(&vnode->fid, need_clear);
- write_sequnlock(&vnode->cb_lock);
- return false;
-}
-
-/*
- * Returns true if the pagecache is still valid. Does not sleep.
- */
-bool afs_pagecache_valid(struct afs_vnode *vnode)
-{
- if (unlikely(test_bit(AFS_VNODE_DELETED, &vnode->flags))) {
- if (vnode->netfs.inode.i_nlink)
- clear_nlink(&vnode->netfs.inode);
- return true;
- }
-
- if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags) &&
- afs_check_validity(vnode))
- return true;
-
- return false;
-}
-
-/*
- * validate a vnode/inode
- * - there are several things we need to check
- * - parent dir data changes (rm, rmdir, rename, mkdir, create, link,
- * symlink)
- * - parent dir metadata changed (security changes)
- * - dentry data changed (write, truncate)
- * - dentry metadata changed (security changes)
- */
-int afs_validate(struct afs_vnode *vnode, struct key *key)
-{
- int ret;
-
- _enter("{v={%llx:%llu} fl=%lx},%x",
- vnode->fid.vid, vnode->fid.vnode, vnode->flags,
- key_serial(key));
-
- if (afs_pagecache_valid(vnode))
- goto valid;
-
- down_write(&vnode->validate_lock);
-
- /* if the promise has expired, we need to check the server again to get
- * a new promise - note that if the (parent) directory's metadata was
- * changed then the security may be different and we may no longer have
- * access */
- if (!test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
- _debug("not promised");
- ret = afs_fetch_status(vnode, key, false, NULL);
- if (ret < 0) {
- if (ret == -ENOENT) {
- set_bit(AFS_VNODE_DELETED, &vnode->flags);
- ret = -ESTALE;
- }
- goto error_unlock;
- }
- _debug("new promise [fl=%lx]", vnode->flags);
- }
-
- if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
- _debug("file already deleted");
- ret = -ESTALE;
- goto error_unlock;
- }
-
- /* if the vnode's data version number changed then its contents are
- * different */
- if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
- afs_zap_data(vnode);
- up_write(&vnode->validate_lock);
-valid:
- _leave(" = 0");
- return 0;
-
-error_unlock:
- up_write(&vnode->validate_lock);
- _leave(" = %d", ret);
- return ret;
-}
-
/*
* read the attributes of an inode
*/
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 22cf782fe924..0a4ecfc250bf 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -1234,9 +1234,6 @@ extern int afs_ilookup5_test_by_fid(struct inode *, void *);
extern struct inode *afs_iget_pseudo_dir(struct super_block *, bool);
extern struct inode *afs_iget(struct afs_operation *, struct afs_vnode_param *);
extern struct inode *afs_root_iget(struct super_block *, struct key *);
-extern bool afs_check_validity(struct afs_vnode *);
-extern int afs_validate(struct afs_vnode *, struct key *);
-bool afs_pagecache_valid(struct afs_vnode *);
extern int afs_getattr(struct mnt_idmap *idmap, const struct path *,
struct kstat *, u32, unsigned int);
extern int afs_setattr(struct mnt_idmap *idmap, struct dentry *, struct iattr *);
@@ -1546,6 +1543,13 @@ void afs_detach_volume_from_servers(struct afs_volume *volume, struct afs_server
extern int __init afs_fs_init(void);
extern void afs_fs_exit(void);
+/*
+ * validation.c
+ */
+bool afs_check_validity(struct afs_vnode *vnode);
+bool afs_pagecache_valid(struct afs_vnode *vnode);
+int afs_validate(struct afs_vnode *vnode, struct key *key);
+
/*
* vlclient.c
*/
diff --git a/fs/afs/validation.c b/fs/afs/validation.c
new file mode 100644
index 000000000000..188ccc8e0fcf
--- /dev/null
+++ b/fs/afs/validation.c
@@ -0,0 +1,185 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* vnode and volume validity verification.
+ *
+ * Copyright (C) 2023 Red Hat, Inc. All Rights Reserved.
+ * Written by David Howells (dhowells@redhat.com)
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include "internal.h"
+
+/*
+ * mark the data attached to an inode as obsolete due to a write on the server
+ * - might also want to ditch all the outstanding writes and dirty pages
+ */
+static void afs_zap_data(struct afs_vnode *vnode)
+{
+ _enter("{%llx:%llu}", vnode->fid.vid, vnode->fid.vnode);
+
+ afs_invalidate_cache(vnode, 0);
+
+ /* nuke all the non-dirty pages that aren't locked, mapped or being
+ * written back in a regular file and completely discard the pages in a
+ * directory or symlink */
+ if (S_ISREG(vnode->netfs.inode.i_mode))
+ invalidate_remote_inode(&vnode->netfs.inode);
+ else
+ invalidate_inode_pages2(vnode->netfs.inode.i_mapping);
+}
+
+/*
+ * Check to see if we have a server currently serving this volume and that it
+ * hasn't been reinitialised or dropped from the list.
+ */
+static bool afs_check_server_good(struct afs_vnode *vnode)
+{
+ struct afs_server_list *slist;
+ struct afs_server *server;
+ bool good;
+ int i;
+
+ if (vnode->cb_fs_s_break == atomic_read(&vnode->volume->cell->fs_s_break))
+ return true;
+
+ rcu_read_lock();
+
+ slist = rcu_dereference(vnode->volume->servers);
+ for (i = 0; i < slist->nr_servers; i++) {
+ server = slist->servers[i].server;
+ if (server == vnode->cb_server) {
+ good = (vnode->cb_s_break == server->cb_s_break);
+ rcu_read_unlock();
+ return good;
+ }
+ }
+
+ rcu_read_unlock();
+ return false;
+}
+
+/*
+ * Check the validity of a vnode/inode.
+ */
+bool afs_check_validity(struct afs_vnode *vnode)
+{
+ enum afs_cb_break_reason need_clear = afs_cb_break_no_break;
+ time64_t now = ktime_get_real_seconds();
+ unsigned int cb_break;
+ int seq = 0;
+
+ do {
+ read_seqbegin_or_lock(&vnode->cb_lock, &seq);
+ cb_break = vnode->cb_break;
+
+ if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
+ if (vnode->cb_v_break != vnode->volume->cb_v_break)
+ need_clear = afs_cb_break_for_v_break;
+ else if (!afs_check_server_good(vnode))
+ need_clear = afs_cb_break_for_s_reinit;
+ else if (test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
+ need_clear = afs_cb_break_for_zap;
+ else if (vnode->cb_expires_at - 10 <= now)
+ need_clear = afs_cb_break_for_lapsed;
+ } else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
+ ;
+ } else {
+ need_clear = afs_cb_break_no_promise;
+ }
+
+ } while (need_seqretry(&vnode->cb_lock, seq));
+
+ done_seqretry(&vnode->cb_lock, seq);
+
+ if (need_clear == afs_cb_break_no_break)
+ return true;
+
+ write_seqlock(&vnode->cb_lock);
+ if (need_clear == afs_cb_break_no_promise)
+ vnode->cb_v_break = vnode->volume->cb_v_break;
+ else if (cb_break == vnode->cb_break)
+ __afs_break_callback(vnode, need_clear);
+ else
+ trace_afs_cb_miss(&vnode->fid, need_clear);
+ write_sequnlock(&vnode->cb_lock);
+ return false;
+}
+
+/*
+ * Returns true if the pagecache is still valid. Does not sleep.
+ */
+bool afs_pagecache_valid(struct afs_vnode *vnode)
+{
+ if (unlikely(test_bit(AFS_VNODE_DELETED, &vnode->flags))) {
+ if (vnode->netfs.inode.i_nlink)
+ clear_nlink(&vnode->netfs.inode);
+ return true;
+ }
+
+ if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags) &&
+ afs_check_validity(vnode))
+ return true;
+
+ return false;
+}
+
+/*
+ * validate a vnode/inode
+ * - there are several things we need to check
+ * - parent dir data changes (rm, rmdir, rename, mkdir, create, link,
+ * symlink)
+ * - parent dir metadata changed (security changes)
+ * - dentry data changed (write, truncate)
+ * - dentry metadata changed (security changes)
+ */
+int afs_validate(struct afs_vnode *vnode, struct key *key)
+{
+ int ret;
+
+ _enter("{v={%llx:%llu} fl=%lx},%x",
+ vnode->fid.vid, vnode->fid.vnode, vnode->flags,
+ key_serial(key));
+
+ if (afs_pagecache_valid(vnode))
+ goto valid;
+
+ down_write(&vnode->validate_lock);
+
+ /* if the promise has expired, we need to check the server again to get
+ * a new promise - note that if the (parent) directory's metadata was
+ * changed then the security may be different and we may no longer have
+ * access */
+ if (!test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
+ _debug("not promised");
+ ret = afs_fetch_status(vnode, key, false, NULL);
+ if (ret < 0) {
+ if (ret == -ENOENT) {
+ set_bit(AFS_VNODE_DELETED, &vnode->flags);
+ ret = -ESTALE;
+ }
+ goto error_unlock;
+ }
+ _debug("new promise [fl=%lx]", vnode->flags);
+ }
+
+ if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
+ _debug("file already deleted");
+ ret = -ESTALE;
+ goto error_unlock;
+ }
+
+ /* if the vnode's data version number changed then its contents are
+ * different */
+ if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
+ afs_zap_data(vnode);
+ up_write(&vnode->validate_lock);
+valid:
+ _leave(" = 0");
+ return 0;
+
+error_unlock:
+ up_write(&vnode->validate_lock);
+ _leave(" = %d", ret);
+ return ret;
+}
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 37/41] afs: Apply server breaks to mmap'd files in the call processor
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (35 preceding siblings ...)
2023-11-09 15:39 ` [PATCH 36/41] afs: Move the vnode/volume validity checking code into its own file David Howells
@ 2023-11-09 15:40 ` David Howells
2023-11-09 15:40 ` [PATCH 38/41] afs: Parse the VolSync record in the reply of a number of RPC ops David Howells
` (3 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:40 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Apply server breaks to mmap'd files that are being used from that server
from the call processor work function rather than punting it off to a
workqueue. The work item, afs_server_init_callback(), then bumps each
individual inode off to its own work item introducing a potentially lengthy
delay. This reduces that delay at the cost of extending the amount of time
we delay replying to the CB.InitCallBack3 notification RPC from the server.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/callback.c | 33 +++++++++++++++++++--------------
fs/afs/cell.c | 2 +-
fs/afs/internal.h | 6 +++---
fs/afs/server.c | 2 --
fs/afs/server_list.c | 22 +++++++++++-----------
fs/afs/volume.c | 15 +++++++++++++++
include/trace/events/afs.h | 2 ++
7 files changed, 51 insertions(+), 31 deletions(-)
diff --git a/fs/afs/callback.c b/fs/afs/callback.c
index a484fa642808..535477f88a4f 100644
--- a/fs/afs/callback.c
+++ b/fs/afs/callback.c
@@ -33,9 +33,8 @@ void afs_invalidate_mmap_work(struct work_struct *work)
unmap_mapping_pages(vnode->netfs.inode.i_mapping, 0, 0, false);
}
-void afs_server_init_callback_work(struct work_struct *work)
+static void afs_server_init_callback(struct afs_server *server)
{
- struct afs_server *server = container_of(work, struct afs_server, initcb_work);
struct afs_vnode *vnode;
struct afs_cell *cell = server->cell;
@@ -57,15 +56,19 @@ void afs_server_init_callback_work(struct work_struct *work)
*/
void afs_init_callback_state(struct afs_server *server)
{
- rcu_read_lock();
+ struct afs_cell *cell = server->cell;
+
+ down_read(&cell->vs_lock);
+
do {
server->cb_s_break++;
atomic_inc(&server->cell->fs_s_break);
if (!list_empty(&server->cell->fs_open_mmaps))
- queue_work(system_unbound_wq, &server->initcb_work);
+ afs_server_init_callback(server);
} while ((server = rcu_dereference(server->uuid_next)));
- rcu_read_unlock();
+
+ up_read(&cell->vs_lock);
}
/*
@@ -112,7 +115,7 @@ static struct afs_volume *afs_lookup_volume_rcu(struct afs_cell *cell,
struct rb_node *p;
int seq = 0;
- do {
+ for (;;) {
/* Unfortunately, rbtree walking doesn't give reliable results
* under just the RCU read lock, so we have to check for
* changes.
@@ -132,7 +135,12 @@ static struct afs_volume *afs_lookup_volume_rcu(struct afs_cell *cell,
volume = NULL;
}
- } while (need_seqretry(&cell->volume_lock, seq));
+ if (volume && afs_try_get_volume(volume, afs_volume_trace_get_callback))
+ break;
+ if (!need_seqretry(&cell->volume_lock, seq))
+ break;
+ seq |= 1; /* Want a lock next time */
+ }
done_seqretry(&cell->volume_lock, seq);
return volume;
@@ -187,12 +195,11 @@ static void afs_break_some_callbacks(struct afs_server *server,
afs_volid_t vid = cbb->fid.vid;
size_t i;
+ rcu_read_lock();
volume = afs_lookup_volume_rcu(server->cell, vid);
-
/* TODO: Find all matching volumes if we couldn't match the server and
* break them anyway.
*/
-
for (i = *_count; i > 0; cbb++, i--) {
if (cbb->fid.vid == vid) {
_debug("- Fid { vl=%08llx n=%llu u=%u }",
@@ -206,6 +213,9 @@ static void afs_break_some_callbacks(struct afs_server *server,
*residue++ = *cbb;
}
}
+
+ rcu_read_unlock();
+ afs_put_volume(volume, afs_volume_trace_put_callback);
}
/*
@@ -218,11 +228,6 @@ void afs_break_callbacks(struct afs_server *server, size_t count,
ASSERT(server != NULL);
- rcu_read_lock();
-
while (count > 0)
afs_break_some_callbacks(server, callbacks, &count);
-
- rcu_read_unlock();
- return;
}
diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index 6b389f2bcd0c..1e5cb0f6ee07 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -161,7 +161,7 @@ static struct afs_cell *afs_alloc_cell(struct afs_net *net,
refcount_set(&cell->ref, 1);
atomic_set(&cell->active, 0);
INIT_WORK(&cell->manager, afs_manage_cell_work);
- spin_lock_init(&cell->vs_lock);
+ init_rwsem(&cell->vs_lock);
cell->volumes = RB_ROOT;
INIT_HLIST_HEAD(&cell->proc_volumes);
seqlock_init(&cell->volume_lock);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 0a4ecfc250bf..73f0a456ab98 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -414,7 +414,7 @@ struct afs_cell {
unsigned int debug_id;
/* The volumes belonging to this cell */
- spinlock_t vs_lock; /* Lock for server->volumes */
+ struct rw_semaphore vs_lock; /* Lock for server->volumes */
struct rb_root volumes; /* Tree of volumes on this server */
struct hlist_head proc_volumes; /* procfs volume list */
seqlock_t volume_lock; /* For volumes */
@@ -566,7 +566,6 @@ struct afs_server {
struct hlist_node addr6_link; /* Link in net->fs_addresses6 */
struct hlist_node proc_link; /* Link in net->fs_proc */
struct list_head volumes; /* RCU list of afs_server_entry objects */
- struct work_struct initcb_work; /* Work for CB.InitCallBackState* */
struct afs_server *gc_next; /* Next server in manager's list */
time64_t unuse_time; /* Time at which last unused */
unsigned long flags;
@@ -1040,7 +1039,6 @@ void afs_get_address_preferences(struct afs_net *net, struct afs_addr_list *alis
* callback.c
*/
extern void afs_invalidate_mmap_work(struct work_struct *);
-extern void afs_server_init_callback_work(struct work_struct *work);
extern void afs_init_callback_state(struct afs_server *);
extern void __afs_break_callback(struct afs_vnode *, enum afs_cb_break_reason);
extern void afs_break_callback(struct afs_vnode *, enum afs_cb_break_reason);
@@ -1619,6 +1617,8 @@ extern struct afs_volume *afs_get_volume(struct afs_volume *, enum afs_volume_tr
void afs_put_volume(struct afs_volume *volume, enum afs_volume_trace reason);
extern int afs_check_volume_status(struct afs_volume *, struct afs_operation *);
+bool afs_try_get_volume(struct afs_volume *volume, enum afs_volume_trace reason);
+
/*
* write.c
*/
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 811dbda88f9c..c97617366557 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -217,7 +217,6 @@ static struct afs_server *afs_alloc_server(struct afs_cell *cell,
server->uuid = *uuid;
rwlock_init(&server->fs_lock);
INIT_LIST_HEAD(&server->volumes);
- INIT_WORK(&server->initcb_work, afs_server_init_callback_work);
init_waitqueue_head(&server->probe_wq);
INIT_LIST_HEAD(&server->probe_link);
spin_lock_init(&server->probe_lock);
@@ -469,7 +468,6 @@ static void afs_destroy_server(struct afs_net *net, struct afs_server *server)
if (test_bit(AFS_SERVER_FL_MAY_HAVE_CB, &server->flags))
afs_give_up_callbacks(net, server);
- flush_work(&server->initcb_work);
afs_put_server(net, server, afs_server_trace_destroy);
}
diff --git a/fs/afs/server_list.c b/fs/afs/server_list.c
index 4d6369477f54..cfd900eb09ed 100644
--- a/fs/afs/server_list.c
+++ b/fs/afs/server_list.c
@@ -136,7 +136,7 @@ void afs_attach_volume_to_servers(struct afs_volume *volume, struct afs_server_l
struct list_head *p;
unsigned int i;
- spin_lock(&volume->cell->vs_lock);
+ down_write(&volume->cell->vs_lock);
for (i = 0; i < slist->nr_servers; i++) {
se = &slist->servers[i];
@@ -147,11 +147,11 @@ void afs_attach_volume_to_servers(struct afs_volume *volume, struct afs_server_l
if (volume->vid <= pe->volume->vid)
break;
}
- list_add_tail_rcu(&se->slink, p);
+ list_add_tail(&se->slink, p);
}
slist->attached = true;
- spin_unlock(&volume->cell->vs_lock);
+ up_write(&volume->cell->vs_lock);
}
/*
@@ -164,7 +164,7 @@ void afs_reattach_volume_to_servers(struct afs_volume *volume, struct afs_server
{
unsigned int n = 0, o = 0;
- spin_lock(&volume->cell->vs_lock);
+ down_write(&volume->cell->vs_lock);
while (n < new->nr_servers || o < old->nr_servers) {
struct afs_server_entry *pn = n < new->nr_servers ? &new->servers[n] : NULL;
@@ -174,7 +174,7 @@ void afs_reattach_volume_to_servers(struct afs_volume *volume, struct afs_server
int diff;
if (pn && po && pn->server == po->server) {
- list_replace_rcu(&po->slink, &pn->slink);
+ list_replace(&po->slink, &pn->slink);
n++;
o++;
continue;
@@ -192,15 +192,15 @@ void afs_reattach_volume_to_servers(struct afs_volume *volume, struct afs_server
if (volume->vid <= s->volume->vid)
break;
}
- list_add_tail_rcu(&pn->slink, p);
+ list_add_tail(&pn->slink, p);
n++;
} else {
- list_del_rcu(&po->slink);
+ list_del(&po->slink);
o++;
}
}
- spin_unlock(&volume->cell->vs_lock);
+ up_write(&volume->cell->vs_lock);
}
/*
@@ -213,11 +213,11 @@ void afs_detach_volume_from_servers(struct afs_volume *volume, struct afs_server
if (!slist->attached)
return;
- spin_lock(&volume->cell->vs_lock);
+ down_write(&volume->cell->vs_lock);
for (i = 0; i < slist->nr_servers; i++)
- list_del_rcu(&slist->servers[i].slink);
+ list_del(&slist->servers[i].slink);
slist->attached = false;
- spin_unlock(&volume->cell->vs_lock);
+ up_write(&volume->cell->vs_lock);
}
diff --git a/fs/afs/volume.c b/fs/afs/volume.c
index 1d3bd14dfa2f..00253032ccca 100644
--- a/fs/afs/volume.c
+++ b/fs/afs/volume.c
@@ -242,6 +242,21 @@ static void afs_destroy_volume(struct work_struct *work)
_leave(" [destroyed]");
}
+
+/*
+ * Try to get a reference on a volume record.
+ */
+bool afs_try_get_volume(struct afs_volume *volume, enum afs_volume_trace reason)
+{
+ int r;
+
+ if (__refcount_inc_not_zero(&volume->ref, &r)) {
+ trace_afs_volume(volume->vid, r + 1, reason);
+ return true;
+ }
+ return false;
+}
+
/*
* Get a reference on a volume record.
*/
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index cf2fa4fddd5b..63ab23876be8 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -151,9 +151,11 @@ enum yfs_cm_operation {
EM(afs_volume_trace_alloc, "ALLOC ") \
EM(afs_volume_trace_free, "FREE ") \
EM(afs_volume_trace_get_alloc_sbi, "GET sbi-alloc ") \
+ EM(afs_volume_trace_get_callback, "GET callback ") \
EM(afs_volume_trace_get_cell_insert, "GET cell-insrt") \
EM(afs_volume_trace_get_new_op, "GET op-new ") \
EM(afs_volume_trace_get_query_alias, "GET cell-alias") \
+ EM(afs_volume_trace_put_callback, "PUT callback ") \
EM(afs_volume_trace_put_cell_dup, "PUT cell-dup ") \
EM(afs_volume_trace_put_cell_root, "PUT cell-root ") \
EM(afs_volume_trace_put_destroy_sbi, "PUT sbi-destry") \
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 38/41] afs: Parse the VolSync record in the reply of a number of RPC ops
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (36 preceding siblings ...)
2023-11-09 15:40 ` [PATCH 37/41] afs: Apply server breaks to mmap'd files in the call processor David Howells
@ 2023-11-09 15:40 ` David Howells
2023-11-09 15:40 ` [PATCH 39/41] afs: Overhaul invalidation handling to better support RO volumes David Howells
` (2 subsequent siblings)
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:40 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
A number of fileserver RPC operations return a VolSync record as part of
their reply that gives some information about the state of the volume being
accessed, including:
(1) A volume Creation timestamp. For an RW volume, this is the time at
which the volume was created; if it changes, the RW volume was
presumably restored from a backup and all cached data should be
scrubbed as Data Version numbers could regress on the files in the
volume.
For an RO volume, this is the time it was last snapshotted from the RW
volume. It is expected to advance each time this happens; if it
regresses, cached data should be scrubbed.
(2) A volume Update timestamp (Auristor only). For an RW volume, this is
updated any time any change is made to a volume or its contents. If
it regresses, all cached data must be scrubbed.
For an RO volume, this is a copy of the RW volume's Update timestamp
at the point of snapshotting. It can be used as a version number when
checking to see if a callback on a RO volume was due to a snapshot.
If it regresses, all cached data must be scrubbed.
but this is currently not made use of by the in-kernel afs filesystem.
Make the afs filesystem use this by:
(1) Add a mask to the afs_volsync struct to record what fields are set
there and add an update time.
(2) Add an afs_volsync struct to the afs_volume struct and use this to
track the two timestamps in the afs_volume struct.
(3) Add a 'pre-op volsync' struct to the afs_operation struct to record
the state of the volume tracking before the op.
(4) Add a new counter, cb_scrub, to the afs_volume struct to count events
that require all data to be scrubbed. A copy is placed in the
afs_vnode struct (inode) and if they no longer match, a scrub takes
place.
(5) When the result of an operation is being parsed, parse the VolSync
data too, if it is provided. Note that the two timestamps are handled
separately, since they don't work in quite the same way.
- If the afs_volume tracking is unset, just set it and do nothing
else.
- If the result timestamps are the same as the ones in afs_volume, do
nothing.
- If the timestamps regress, propose incrementing cb_scrub.
- If the RW creation timestamp changes, propose incrementing cb_scrub.
- If any timestamp changes, propose incrementing cb_v_break,
indicating updates may be required.
- If an update cb_scrub or cb_v_break is proposed, try to cmpxchg the
copy of the timestamp in afs_volume from the pre-op version to the
result version and only update if successful. This should help
mitigate the case where a bunch of parallel ops all see the change
and all try crank the counters.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/afs.h | 4 ++
fs/afs/callback.c | 7 +--
fs/afs/fs_operation.c | 15 +++---
fs/afs/fsclient.c | 9 +++-
fs/afs/inode.c | 4 +-
fs/afs/internal.h | 12 +++--
fs/afs/rotate.c | 2 +-
fs/afs/validation.c | 107 ++++++++++++++++++++++++++++++++++++-
fs/afs/yfsclient.c | 6 ++-
include/trace/events/afs.h | 27 ++++++++++
10 files changed, 174 insertions(+), 19 deletions(-)
diff --git a/fs/afs/afs.h b/fs/afs/afs.h
index 81815724db6c..a6328a9ec9b0 100644
--- a/fs/afs/afs.h
+++ b/fs/afs/afs.h
@@ -165,7 +165,11 @@ struct afs_status_cb {
* AFS volume synchronisation information
*/
struct afs_volsync {
+ unsigned int mask; /* Bitmask of supplied fields */
+#define AFS_VOLSYNC_CREATION 0x01
+#define AFS_VOLSYNC_UPDATE 0x02
time64_t creation; /* volume creation time */
+ time64_t update; /* Volume update time */
};
/*
diff --git a/fs/afs/callback.c b/fs/afs/callback.c
index 535477f88a4f..62332cfa8661 100644
--- a/fs/afs/callback.c
+++ b/fs/afs/callback.c
@@ -81,7 +81,7 @@ void __afs_break_callback(struct afs_vnode *vnode, enum afs_cb_break_reason reas
clear_bit(AFS_VNODE_NEW_CONTENT, &vnode->flags);
if (test_and_clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
vnode->cb_break++;
- vnode->cb_v_break = vnode->volume->cb_v_break;
+ vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
afs_clear_permits(vnode);
if (vnode->lock_state == AFS_VNODE_LOCK_WAITING_FOR_CB)
@@ -158,12 +158,13 @@ static void afs_break_one_callback(struct afs_volume *volume,
struct super_block *sb;
struct afs_vnode *vnode;
struct inode *inode;
+ unsigned int cb_v_break;
if (fid->vnode == 0 && fid->unique == 0) {
/* The callback break applies to an entire volume. */
write_lock(&volume->cb_v_break_lock);
- volume->cb_v_break++;
- trace_afs_cb_break(fid, volume->cb_v_break,
+ cb_v_break = atomic_inc_return(&volume->cb_v_break);
+ trace_afs_cb_break(fid, cb_v_break,
afs_cb_break_for_volume_callback, false);
write_unlock(&volume->cb_v_break_lock);
return;
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index 10137681aa7d..7f1ee13ebce0 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -35,11 +35,14 @@ struct afs_operation *afs_alloc_operation(struct key *key, struct afs_volume *vo
key_get(key);
}
- op->key = key;
- op->volume = afs_get_volume(volume, afs_volume_trace_get_new_op);
- op->net = volume->cell->net;
- op->cb_v_break = volume->cb_v_break;
- op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
+ op->key = key;
+ op->volume = afs_get_volume(volume, afs_volume_trace_get_new_op);
+ op->net = volume->cell->net;
+ op->cb_v_break = atomic_read(&volume->cb_v_break);
+ op->pre_volsync.mask = READ_ONCE(volume->volsync.mask);
+ op->pre_volsync.creation = READ_ONCE(volume->volsync.creation);
+ op->pre_volsync.update = READ_ONCE(volume->volsync.update);
+ op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
op->nr_iterations = -1;
afs_op_set_error(op, -EDESTADDRREQ);
@@ -147,7 +150,7 @@ bool afs_begin_vnode_operation(struct afs_operation *op)
afs_prepare_vnode(op, &op->file[0], 0);
afs_prepare_vnode(op, &op->file[1], 1);
- op->cb_v_break = op->volume->cb_v_break;
+ op->cb_v_break = atomic_read(&op->volume->cb_v_break);
_leave(" = true");
return true;
}
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index f1f879ba9cf7..1bd1614688f7 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -165,8 +165,10 @@ static void xdr_decode_AFSVolSync(const __be32 **_bp,
bp++; /* spare6 */
*_bp = bp;
- if (volsync)
+ if (volsync) {
volsync->creation = creation;
+ volsync->mask |= AFS_VOLSYNC_CREATION;
+ }
}
/*
@@ -1870,7 +1872,10 @@ static int afs_deliver_fs_inline_bulk_status(struct afs_call *call)
return ret;
bp = call->buffer;
- xdr_decode_AFSVolSync(&bp, &op->volsync);
+ /* Unfortunately, prior to OpenAFS-1.6, volsync here is filled
+ * with rubbish.
+ */
+ xdr_decode_AFSVolSync(&bp, NULL);
call->unmarshall++;
fallthrough;
diff --git a/fs/afs/inode.c b/fs/afs/inode.c
index 8dbb09a79c4d..43c71fd50202 100644
--- a/fs/afs/inode.c
+++ b/fs/afs/inode.c
@@ -329,6 +329,8 @@ static void afs_fetch_status_success(struct afs_operation *op)
struct afs_vnode *vnode = vp->vnode;
int ret;
+ afs_update_volume_state(op, vp);
+
if (vnode->netfs.inode.i_state & I_NEW) {
ret = afs_inode_init_from_status(op, vp, vnode);
afs_op_set_error(op, ret);
@@ -542,7 +544,7 @@ struct inode *afs_root_iget(struct super_block *sb, struct key *key)
BUG_ON(!(inode->i_state & I_NEW));
vnode = AFS_FS_I(inode);
- vnode->cb_v_break = as->volume->cb_v_break,
+ vnode->cb_v_break = atomic_read(&as->volume->cb_v_break),
afs_set_netfs_context(vnode);
op = afs_alloc_operation(key, as->volume);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 73f0a456ab98..6561a1946be2 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -631,6 +631,7 @@ struct afs_volume {
afs_volid_t vids[AFS_MAXTYPES]; /* All associated volume IDs */
refcount_t ref;
time64_t update_at; /* Time at which to next update */
+ struct afs_volsync volsync; /* Volume info from server */
struct afs_cell *cell; /* Cell to which belongs (pins ref) */
struct rb_node cell_node; /* Link in cell->volumes */
struct hlist_node proc_link; /* Link in cell->proc_volumes */
@@ -651,7 +652,10 @@ struct afs_volume {
rwlock_t servers_lock; /* Lock for ->servers */
unsigned int servers_seq; /* Incremented each time ->servers changes */
- unsigned cb_v_break; /* Break-everything counter. */
+ /* Callback management */
+ atomic_t cb_ro_snapshot; /* RO volume update-from-snapshot counter */
+ atomic_t cb_v_break; /* Volume-break event counter. */
+ atomic_t cb_scrub; /* Scrub-all-data event counter. */
rwlock_t cb_v_break_lock;
afs_voltype_t type; /* type of volume */
@@ -845,7 +849,8 @@ struct afs_operation {
struct afs_volume *volume; /* Volume being accessed */
struct afs_vnode_param file[2];
struct afs_vnode_param *more_files;
- struct afs_volsync volsync;
+ struct afs_volsync pre_volsync; /* Volsync before op */
+ struct afs_volsync volsync; /* Volsync returned by op */
struct dentry *dentry; /* Dentry to be altered */
struct dentry *dentry_2; /* Second dentry to be altered */
struct timespec64 mtime; /* Modification time to record */
@@ -1052,7 +1057,7 @@ static inline unsigned int afs_calc_vnode_cb_break(struct afs_vnode *vnode)
static inline bool afs_cb_is_broken(unsigned int cb_break,
const struct afs_vnode *vnode)
{
- return cb_break != (vnode->cb_break + vnode->volume->cb_v_break);
+ return cb_break != (vnode->cb_break + atomic_read(&vnode->volume->cb_v_break));
}
/*
@@ -1544,6 +1549,7 @@ extern void afs_fs_exit(void);
/*
* validation.c
*/
+void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp);
bool afs_check_validity(struct afs_vnode *vnode);
bool afs_pagecache_valid(struct afs_vnode *vnode);
int afs_validate(struct afs_vnode *vnode, struct key *key);
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 56916558dd1d..f2f7e57a4bdd 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -482,7 +482,7 @@ bool afs_select_fileserver(struct afs_operation *op)
vnode->cb_server = server;
vnode->cb_s_break = server->cb_s_break;
vnode->cb_fs_s_break = atomic_read(&server->cell->fs_s_break);
- vnode->cb_v_break = vnode->volume->cb_v_break;
+ vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
}
diff --git a/fs/afs/validation.c b/fs/afs/validation.c
index 188ccc8e0fcf..f6bdaa121cf6 100644
--- a/fs/afs/validation.c
+++ b/fs/afs/validation.c
@@ -10,6 +10,109 @@
#include <linux/sched.h>
#include "internal.h"
+/*
+ * Handle a change to the volume creation time in the VolSync record.
+ */
+static void afs_update_volume_creation_time(struct afs_operation *op, struct afs_volume *volume)
+{
+ enum afs_cb_break_reason reason = afs_cb_break_for_vos_release;
+ unsigned int snap;
+ time64_t cur = READ_ONCE(volume->volsync.creation);
+ time64_t old = op->pre_volsync.creation;
+ time64_t new = op->volsync.creation;
+
+ _enter("%llx,%llx,%llx->%llx", volume->vid, cur, old, new);
+
+ if (!(op->volsync.mask & AFS_VOLSYNC_CREATION) ||
+ !(volume->volsync.mask & AFS_VOLSYNC_CREATION)) {
+ WRITE_ONCE(volume->volsync.creation, new);
+ volume->volsync.mask |= AFS_VOLSYNC_CREATION;
+ return;
+ }
+
+ if (likely(new == cur))
+ return;
+
+ /* If the creation time changes in an unexpected way, we need to scrub
+ * our caches. For a RW vol, this will only change if the volume is
+ * restored from a backup; for a RO/Backup vol, this will advance when
+ * the volume is updated to a new snapshot (eg. "vos release").
+ */
+ if (volume->type == AFSVL_RWVOL || new < old)
+ reason = afs_cb_break_for_creation_regress;
+
+ /* Try to advance the creation timestamp from what we had before the
+ * operation to what we got back from the server. This should
+ * hopefully ensure that in a race between multiple operations only one
+ * of them will do this.
+ */
+ if (try_cmpxchg(&volume->volsync.creation, &old, new)) {
+ if (reason == afs_cb_break_for_creation_regress)
+ atomic_inc(&volume->cb_scrub);
+ else if (volume->type != AFSVL_RWVOL)
+ snap = atomic_inc_return(&volume->cb_ro_snapshot);
+ trace_afs_cb_v_break(volume->vid, snap, reason);
+ }
+}
+
+/*
+ * Handle a change to the volume update time in the VolSync record.
+ */
+static void afs_update_volume_update_time(struct afs_operation *op, struct afs_volume *volume)
+{
+ enum afs_cb_break_reason reason = afs_cb_break_no_break;
+ time64_t cur = READ_ONCE(volume->volsync.update);
+ time64_t old = op->pre_volsync.update;
+ time64_t new = op->volsync.update;
+
+ _enter("%llx,%llx,%llx->%llx", volume->vid, cur, old, new);
+
+ if (!(op->volsync.mask & AFS_VOLSYNC_UPDATE) ||
+ !(volume->volsync.mask & AFS_VOLSYNC_UPDATE)) {
+ WRITE_ONCE(volume->volsync.update, new);
+ volume->volsync.mask |= AFS_VOLSYNC_UPDATE;
+ return;
+ }
+
+ if (likely(new == cur))
+ return;
+
+ /* If the volume update time changes in an unexpected way, we need to
+ * scrub our caches. For a RW vol, this will advance on every
+ * modification op; for a RO/Backup vol, this will advance when the
+ * volume is updated to a new snapshot (eg. "vos release").
+ */
+ if (new < old)
+ reason = afs_cb_break_for_update_regress;
+
+ /* Try to advance the update timestamp from what we had before the
+ * operation to what we got back from the server. This should
+ * hopefully ensure that in a race between multiple operations only one
+ * of them will do this.
+ */
+ if (try_cmpxchg(&volume->volsync.update, &old, new)) {
+ if (reason == afs_cb_break_for_update_regress) {
+ atomic_inc(&volume->cb_scrub);
+ trace_afs_cb_v_break(volume->vid, 0, reason);
+ }
+ }
+}
+
+/*
+ * Update the state of a volume.
+ */
+void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp)
+{
+ struct afs_volume *volume = op->volume;
+
+ _enter("%llx", op->volume->vid);
+
+ if (op->volsync.mask & AFS_VOLSYNC_CREATION)
+ afs_update_volume_creation_time(op, volume);
+ if (op->volsync.mask & AFS_VOLSYNC_UPDATE)
+ afs_update_volume_update_time(op, volume);
+}
+
/*
* mark the data attached to an inode as obsolete due to a write on the server
* - might also want to ditch all the outstanding writes and dirty pages
@@ -74,7 +177,7 @@ bool afs_check_validity(struct afs_vnode *vnode)
cb_break = vnode->cb_break;
if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
- if (vnode->cb_v_break != vnode->volume->cb_v_break)
+ if (vnode->cb_v_break != atomic_read(&vnode->volume->cb_v_break))
need_clear = afs_cb_break_for_v_break;
else if (!afs_check_server_good(vnode))
need_clear = afs_cb_break_for_s_reinit;
@@ -97,7 +200,7 @@ bool afs_check_validity(struct afs_vnode *vnode)
write_seqlock(&vnode->cb_lock);
if (need_clear == afs_cb_break_no_promise)
- vnode->cb_v_break = vnode->volume->cb_v_break;
+ vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
else if (cb_break == vnode->cb_break)
__afs_break_callback(vnode, need_clear);
else
diff --git a/fs/afs/yfsclient.c b/fs/afs/yfsclient.c
index 11571cca86c1..7c013070ef51 100644
--- a/fs/afs/yfsclient.c
+++ b/fs/afs/yfsclient.c
@@ -245,12 +245,16 @@ static void xdr_decode_YFSVolSync(const __be32 **_bp,
struct afs_volsync *volsync)
{
struct yfs_xdr_YFSVolSync *x = (void *)*_bp;
- u64 creation;
+ u64 creation, update;
if (volsync) {
creation = xdr_to_u64(x->vol_creation_date);
do_div(creation, 10 * 1000 * 1000);
volsync->creation = creation;
+ update = xdr_to_u64(x->vol_update_date);
+ do_div(update, 10 * 1000 * 1000);
+ volsync->update = update;
+ volsync->mask |= AFS_VOLSYNC_CREATION | AFS_VOLSYNC_UPDATE;
}
*_bp += xdr_size(x);
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 63ab23876be8..201dac3be54f 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -440,12 +440,15 @@ enum yfs_cm_operation {
EM(afs_cb_break_no_break, "no-break") \
EM(afs_cb_break_no_promise, "no-promise") \
EM(afs_cb_break_for_callback, "break-cb") \
+ EM(afs_cb_break_for_creation_regress, "creation-regress") \
EM(afs_cb_break_for_deleted, "break-del") \
EM(afs_cb_break_for_lapsed, "break-lapsed") \
EM(afs_cb_break_for_s_reinit, "s-reinit") \
EM(afs_cb_break_for_unlink, "break-unlink") \
+ EM(afs_cb_break_for_update_regress, "update-regress") \
EM(afs_cb_break_for_v_break, "break-v") \
EM(afs_cb_break_for_volume_callback, "break-v-cb") \
+ EM(afs_cb_break_for_vos_release, "break-vos-release") \
E_(afs_cb_break_for_zap, "break-zap")
/*
@@ -1249,6 +1252,30 @@ TRACE_EVENT(afs_get_tree,
__entry->cell, __entry->volume, __entry->vid)
);
+TRACE_EVENT(afs_cb_v_break,
+ TP_PROTO(afs_volid_t vid, unsigned int cb_v_break,
+ enum afs_cb_break_reason reason),
+
+ TP_ARGS(vid, cb_v_break, reason),
+
+ TP_STRUCT__entry(
+ __field(afs_volid_t, vid)
+ __field(unsigned int, cb_v_break)
+ __field(enum afs_cb_break_reason, reason)
+ ),
+
+ TP_fast_assign(
+ __entry->vid = vid;
+ __entry->cb_v_break = cb_v_break;
+ __entry->reason = reason;
+ ),
+
+ TP_printk("%llx vb=%x %s",
+ __entry->vid,
+ __entry->cb_v_break,
+ __print_symbolic(__entry->reason, afs_cb_break_reasons))
+ );
+
TRACE_EVENT(afs_cb_break,
TP_PROTO(struct afs_fid *fid, unsigned int cb_break,
enum afs_cb_break_reason reason, bool skipped),
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 39/41] afs: Overhaul invalidation handling to better support RO volumes
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (37 preceding siblings ...)
2023-11-09 15:40 ` [PATCH 38/41] afs: Parse the VolSync record in the reply of a number of RPC ops David Howells
@ 2023-11-09 15:40 ` David Howells
2023-11-09 19:00 ` Jeffrey E Altman
2023-11-13 15:58 ` [PATCH 42/41] afs: Fix the handling of " David Howells
2023-11-09 15:40 ` [PATCH 40/41] afs: Fix fileserver rotation David Howells
2023-11-09 15:40 ` [PATCH 41/41] afs: Fix offline and busy handling David Howells
40 siblings, 2 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:40 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Overhaul the third party-induced invalidation handling, making use of the
previously added volume-level event counters (cb_scrub and cb_ro_snapshot)
that are now being parsed out of the VolSync record returned by the
fileserver in many of its replies.
This allows better handling of RO (and Backup) volumes. Since these are
snapshot of a RW volume that are updated atomically simultantanously across
all servers that host them, they only require a single callback promise for
the entire volume. The currently upstream code assumes that RO volumes
operate in the same manner as RW volumes, and that each file has its own
individual callback - which means that it does a status fetch for *every*
file in a RO volume, whether or not the volume got "released" (volume
callback breaks can occur for other reasons too, such as the volumeserver
taking ownership of a volume from a fileserver).
To this end, make the following changes:
(1) Change the meaning of the volume's cb_v_break counter so that it is
now a hint that we need to issue a status fetch to work out the state
of a volume. cb_v_break is incremented by volume break callbacks and
by server initialisation callbacks.
(2) Add a second counter, cb_v_check, to the afs_volume struct such that
if this differs from cb_v_break, we need to do a check. When the
check is complete, cb_v_check is advanced to what cb_v_break was at
the start of the status fetch.
(3) Move the list of mmap'd vnodes to the volume and trigger removal of
PTEs that map to files on a volume break rather than on a server
break.
(4) When a server reinitialisation callback comes in, use the
server-to-volume reverse mapping added in a preceding patch to iterate
over all the volumes using that server and clear the volume callback
promises for that server and the general volume promise as a whole to
trigger reanalysis.
(5) Replace the AFS_VNODE_CB_PROMISED flag with an AFS_NO_CB_PROMISE
(TIME64_MIN) value in the cb_expires_at field, reducing the number of
checks we need to make.
(6) Change afs_check_validity() to quickly see if various event counters
have been incremented or if the vnode or volume callback promise is
due to expire/has expired without making any changes to the state.
That is now left to afs_validate() as this may get more complicated in
future as we may have to examine server records too.
(7) Overhaul afs_validate() so that it does a single status fetch if we
need to check the state of either the vnode or the volume - and do so
under appropriate locking. The function does the following steps:
(A) If the vnode/volume is no longer seen as valid, then we take the
vnode validation lock and, if the volume promise has expired, the
volume check lock also. The latter prevents redundant checks being
made to find out if a new version of the volume got released.
(B) If a previous RPC call found that the volsync changed unexpectedly
or that a RO volume was updated, then we unmap all PTEs pointing to
the file to stop mmap being used for access.
(C) If the vnode is still seen to be of uncertain validity, then we
perform an FS.FetchStatus RPC op to jointly update the volume status
and the vnode status. This assessment is done as part of parsing the
reply:
If the RO volume creation timestamp advances, cb_ro_snapshot is
incremented; if either the creation or update timestamps changes in
an unexpected way, the cb_scrub counter is incremented
If the Data Version returned doesn't match the copy we have
locally, then we ask for the pagecache to be zapped. This takes
care of handling RO update.
(D) If cb_scrub differs between volume and vnode, the vnode's
pagecache is zapped and the vnode's cb_scrub is updated unless the
file is marked as having been deleted.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/callback.c | 122 +++++++++------
fs/afs/cell.c | 2 -
fs/afs/dir.c | 3 +-
fs/afs/file.c | 13 +-
fs/afs/fs_operation.c | 3 +-
fs/afs/inode.c | 20 ++-
fs/afs/internal.h | 32 ++--
fs/afs/proc.c | 4 +-
fs/afs/rotate.c | 8 +-
fs/afs/server_list.c | 2 +
fs/afs/validation.c | 302 ++++++++++++++++++++++++-------------
fs/afs/volume.c | 2 +
include/trace/events/afs.h | 6 +-
13 files changed, 315 insertions(+), 204 deletions(-)
diff --git a/fs/afs/callback.c b/fs/afs/callback.c
index 62332cfa8661..1b4f8bb86f6b 100644
--- a/fs/afs/callback.c
+++ b/fs/afs/callback.c
@@ -33,21 +33,20 @@ void afs_invalidate_mmap_work(struct work_struct *work)
unmap_mapping_pages(vnode->netfs.inode.i_mapping, 0, 0, false);
}
-static void afs_server_init_callback(struct afs_server *server)
+static void afs_volume_init_callback(struct afs_volume *volume)
{
struct afs_vnode *vnode;
- struct afs_cell *cell = server->cell;
- down_read(&cell->fs_open_mmaps_lock);
+ down_read(&volume->open_mmaps_lock);
- list_for_each_entry(vnode, &cell->fs_open_mmaps, cb_mmap_link) {
- if (vnode->cb_server == server) {
- clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
+ list_for_each_entry(vnode, &volume->open_mmaps, cb_mmap_link) {
+ if (vnode->cb_v_check != atomic_read(&volume->cb_v_break)) {
+ vnode->cb_expires_at = AFS_NO_CB_PROMISE;
queue_work(system_unbound_wq, &vnode->cb_work);
}
}
- up_read(&cell->fs_open_mmaps_lock);
+ up_read(&volume->open_mmaps_lock);
}
/*
@@ -56,19 +55,20 @@ static void afs_server_init_callback(struct afs_server *server)
*/
void afs_init_callback_state(struct afs_server *server)
{
- struct afs_cell *cell = server->cell;
+ struct afs_server_entry *se;
- down_read(&cell->vs_lock);
+ down_read(&server->cell->vs_lock);
- do {
- server->cb_s_break++;
- atomic_inc(&server->cell->fs_s_break);
- if (!list_empty(&server->cell->fs_open_mmaps))
- afs_server_init_callback(server);
-
- } while ((server = rcu_dereference(server->uuid_next)));
+ list_for_each_entry(se, &server->volumes, slink) {
+ se->cb_expires_at = AFS_NO_CB_PROMISE;
+ se->volume->cb_expires_at = AFS_NO_CB_PROMISE;
+ trace_afs_cb_v_break(se->volume->vid, atomic_read(&se->volume->cb_v_break),
+ afs_cb_break_for_s_reinit);
+ if (!list_empty(&se->volume->open_mmaps))
+ afs_volume_init_callback(se->volume);
+ }
- up_read(&cell->vs_lock);
+ up_read(&server->cell->vs_lock);
}
/*
@@ -79,9 +79,9 @@ void __afs_break_callback(struct afs_vnode *vnode, enum afs_cb_break_reason reas
_enter("");
clear_bit(AFS_VNODE_NEW_CONTENT, &vnode->flags);
- if (test_and_clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
+ if (xchg(&vnode->cb_expires_at, AFS_NO_CB_PROMISE) != AFS_NO_CB_PROMISE) {
vnode->cb_break++;
- vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
+ vnode->cb_v_check = atomic_read(&vnode->volume->cb_v_break);
afs_clear_permits(vnode);
if (vnode->lock_state == AFS_VNODE_LOCK_WAITING_FOR_CB)
@@ -146,29 +146,51 @@ static struct afs_volume *afs_lookup_volume_rcu(struct afs_cell *cell,
return volume;
}
+/*
+ * Allow the fileserver to break callbacks at the volume-level. This is
+ * typically done when, for example, a R/W volume is snapshotted to a R/O
+ * volume (the only way to change an R/O volume). It may also, however, happen
+ * when a volserver takes control of a volume (offlining it, moving it, etc.).
+ *
+ * Every file in that volume will need to be reevaluated.
+ */
+static void afs_break_volume_callback(struct afs_server *server,
+ struct afs_volume *volume)
+ __releases(RCU)
+{
+ struct afs_server_list *slist = rcu_dereference(volume->servers);
+ unsigned int i, cb_v_break;
+
+ write_lock(&volume->cb_v_break_lock);
+
+ for (i = 0; i < slist->nr_servers; i++)
+ if (slist->servers[i].server == server)
+ slist->servers[i].cb_expires_at = AFS_NO_CB_PROMISE;
+ volume->cb_expires_at = AFS_NO_CB_PROMISE;
+
+ cb_v_break = atomic_inc_return_release(&volume->cb_v_break);
+ trace_afs_cb_v_break(volume->vid, cb_v_break, afs_cb_break_for_volume_callback);
+
+ write_unlock(&volume->cb_v_break_lock);
+ rcu_read_unlock();
+
+ if (!list_empty(&volume->open_mmaps))
+ afs_volume_init_callback(volume);
+}
+
/*
* allow the fileserver to explicitly break one callback
* - happens when
* - the backing file is changed
* - a lock is released
*/
-static void afs_break_one_callback(struct afs_volume *volume,
+static void afs_break_one_callback(struct afs_server *server,
+ struct afs_volume *volume,
struct afs_fid *fid)
{
struct super_block *sb;
struct afs_vnode *vnode;
struct inode *inode;
- unsigned int cb_v_break;
-
- if (fid->vnode == 0 && fid->unique == 0) {
- /* The callback break applies to an entire volume. */
- write_lock(&volume->cb_v_break_lock);
- cb_v_break = atomic_inc_return(&volume->cb_v_break);
- trace_afs_cb_break(fid, cb_v_break,
- afs_cb_break_for_volume_callback, false);
- write_unlock(&volume->cb_v_break_lock);
- return;
- }
/* See if we can find a matching inode - even an I_NEW inode needs to
* be marked as it can have its callback broken before we finish
@@ -198,24 +220,32 @@ static void afs_break_some_callbacks(struct afs_server *server,
rcu_read_lock();
volume = afs_lookup_volume_rcu(server->cell, vid);
- /* TODO: Find all matching volumes if we couldn't match the server and
- * break them anyway.
- */
- for (i = *_count; i > 0; cbb++, i--) {
- if (cbb->fid.vid == vid) {
- _debug("- Fid { vl=%08llx n=%llu u=%u }",
- cbb->fid.vid,
- cbb->fid.vnode,
- cbb->fid.unique);
- --*_count;
- if (volume)
- afs_break_one_callback(volume, &cbb->fid);
- } else {
- *residue++ = *cbb;
+ if (cbb->fid.vnode == 0 && cbb->fid.unique == 0) {
+ afs_break_volume_callback(server, volume);
+ *_count -= 1;
+ if (*_count)
+ memmove(cbb, cbb + 1, sizeof(*cbb) * *_count);
+ } else {
+ /* TODO: Find all matching volumes if we couldn't match the server and
+ * break them anyway.
+ */
+
+ for (i = *_count; i > 0; cbb++, i--) {
+ if (cbb->fid.vid == vid) {
+ _debug("- Fid { vl=%08llx n=%llu u=%u }",
+ cbb->fid.vid,
+ cbb->fid.vnode,
+ cbb->fid.unique);
+ --*_count;
+ if (volume)
+ afs_break_one_callback(server, volume, &cbb->fid);
+ } else {
+ *residue++ = *cbb;
+ }
}
+ rcu_read_unlock();
}
- rcu_read_unlock();
afs_put_volume(volume, afs_volume_trace_put_callback);
}
diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index 1e5cb0f6ee07..ed990e2934a0 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -167,8 +167,6 @@ static struct afs_cell *afs_alloc_cell(struct afs_net *net,
seqlock_init(&cell->volume_lock);
cell->fs_servers = RB_ROOT;
seqlock_init(&cell->fs_lock);
- INIT_LIST_HEAD(&cell->fs_open_mmaps);
- init_rwsem(&cell->fs_open_mmaps_lock);
rwlock_init(&cell->vl_servers_lock);
cell->flags = (1 << AFS_CELL_FL_CHECK_ALIAS);
diff --git a/fs/afs/dir.c b/fs/afs/dir.c
index ecb889a269fb..190d1ff42e60 100644
--- a/fs/afs/dir.c
+++ b/fs/afs/dir.c
@@ -1260,6 +1260,7 @@ void afs_check_for_remote_deletion(struct afs_operation *op)
switch (afs_op_abort_code(op)) {
case VNOVNODE:
set_bit(AFS_VNODE_DELETED, &vnode->flags);
+ clear_nlink(&vnode->netfs.inode);
afs_break_callback(vnode, afs_cb_break_for_deleted);
}
}
@@ -1375,7 +1376,7 @@ static void afs_dir_remove_subdir(struct dentry *dentry)
clear_nlink(&vnode->netfs.inode);
set_bit(AFS_VNODE_DELETED, &vnode->flags);
- clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
+ vnode->cb_expires_at = AFS_NO_CB_PROMISE;
clear_bit(AFS_VNODE_DIR_VALID, &vnode->flags);
}
}
diff --git a/fs/afs/file.c b/fs/afs/file.c
index 8f9b42427569..30914e0d9cb2 100644
--- a/fs/afs/file.c
+++ b/fs/afs/file.c
@@ -514,13 +514,12 @@ static bool afs_release_folio(struct folio *folio, gfp_t gfp)
static void afs_add_open_mmap(struct afs_vnode *vnode)
{
if (atomic_inc_return(&vnode->cb_nr_mmap) == 1) {
- down_write(&vnode->volume->cell->fs_open_mmaps_lock);
+ down_write(&vnode->volume->open_mmaps_lock);
if (list_empty(&vnode->cb_mmap_link))
- list_add_tail(&vnode->cb_mmap_link,
- &vnode->volume->cell->fs_open_mmaps);
+ list_add_tail(&vnode->cb_mmap_link, &vnode->volume->open_mmaps);
- up_write(&vnode->volume->cell->fs_open_mmaps_lock);
+ up_write(&vnode->volume->open_mmaps_lock);
}
}
@@ -529,12 +528,12 @@ static void afs_drop_open_mmap(struct afs_vnode *vnode)
if (!atomic_dec_and_test(&vnode->cb_nr_mmap))
return;
- down_write(&vnode->volume->cell->fs_open_mmaps_lock);
+ down_write(&vnode->volume->open_mmaps_lock);
if (atomic_read(&vnode->cb_nr_mmap) == 0)
list_del_init(&vnode->cb_mmap_link);
- up_write(&vnode->volume->cell->fs_open_mmaps_lock);
+ up_write(&vnode->volume->open_mmaps_lock);
flush_work(&vnode->cb_work);
}
@@ -570,7 +569,7 @@ static vm_fault_t afs_vm_map_pages(struct vm_fault *vmf, pgoff_t start_pgoff, pg
{
struct afs_vnode *vnode = AFS_FS_I(file_inode(vmf->vma->vm_file));
- if (afs_pagecache_valid(vnode))
+ if (afs_check_validity(vnode))
return filemap_map_pages(vmf, start_pgoff, end_pgoff);
return 0;
}
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index 7f1ee13ebce0..c9e85fa94d58 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -43,7 +43,7 @@ struct afs_operation *afs_alloc_operation(struct key *key, struct afs_volume *vo
op->pre_volsync.creation = READ_ONCE(volume->volsync.creation);
op->pre_volsync.update = READ_ONCE(volume->volsync.update);
op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
- op->nr_iterations = -1;
+ op->nr_iterations = -1;
afs_op_set_error(op, -EDESTADDRREQ);
_leave(" = [op=%08x]", op->debug_id);
@@ -185,7 +185,6 @@ void afs_wait_for_operation(struct afs_operation *op)
op->call_responded = false;
op->call_error = 0;
op->call_abort_code = 0;
- op->cb_s_break = op->server->cb_s_break;
if (test_bit(AFS_SERVER_FL_IS_YFS, &op->server->flags) &&
op->ops->issue_yfs_rpc)
op->ops->issue_yfs_rpc(op);
diff --git a/fs/afs/inode.c b/fs/afs/inode.c
index 43c71fd50202..5e6426062364 100644
--- a/fs/afs/inode.c
+++ b/fs/afs/inode.c
@@ -85,8 +85,7 @@ static int afs_inode_init_from_status(struct afs_operation *op,
write_seqlock(&vnode->cb_lock);
- vnode->cb_v_break = op->cb_v_break;
- vnode->cb_s_break = op->cb_s_break;
+ vnode->cb_v_check = op->cb_v_break;
vnode->status = *status;
t = status->mtime_client;
@@ -146,11 +145,10 @@ static int afs_inode_init_from_status(struct afs_operation *op,
if (!vp->scb.have_cb) {
/* it's a symlink we just created (the fileserver
* didn't give us a callback) */
- vnode->cb_expires_at = ktime_get_real_seconds();
+ vnode->cb_expires_at = AFS_NO_CB_PROMISE;
} else {
- vnode->cb_expires_at = vp->scb.callback.expires_at;
vnode->cb_server = op->server;
- set_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
+ vnode->cb_expires_at = vp->scb.callback.expires_at;
}
write_sequnlock(&vnode->cb_lock);
@@ -214,7 +212,7 @@ static void afs_apply_status(struct afs_operation *op,
vnode->status = *status;
if (vp->dv_before + vp->dv_delta != status->data_version) {
- if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags))
+ if (vnode->cb_expires_at != AFS_NO_CB_PROMISE)
pr_warn("kAFS: vnode modified {%llx:%llu} %llx->%llx %s (op=%x)\n",
vnode->fid.vid, vnode->fid.vnode,
(unsigned long long)vp->dv_before + vp->dv_delta,
@@ -268,9 +266,9 @@ static void afs_apply_callback(struct afs_operation *op,
struct afs_vnode *vnode = vp->vnode;
if (!afs_cb_is_broken(vp->cb_break_before, vnode)) {
- vnode->cb_expires_at = cb->expires_at;
- vnode->cb_server = op->server;
- set_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
+ if (op->volume->type == AFSVL_RWVOL)
+ vnode->cb_server = op->server;
+ vnode->cb_expires_at = cb->expires_at;
}
}
@@ -544,7 +542,7 @@ struct inode *afs_root_iget(struct super_block *sb, struct key *key)
BUG_ON(!(inode->i_state & I_NEW));
vnode = AFS_FS_I(inode);
- vnode->cb_v_break = atomic_read(&as->volume->cb_v_break),
+ vnode->cb_v_check = atomic_read(&as->volume->cb_v_break),
afs_set_netfs_context(vnode);
op = afs_alloc_operation(key, as->volume);
@@ -589,7 +587,7 @@ int afs_getattr(struct mnt_idmap *idmap, const struct path *path,
if (vnode->volume &&
!(query_flags & AT_STATX_DONT_SYNC) &&
- !test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
+ vnode->cb_expires_at == AFS_NO_CB_PROMISE) {
key = afs_request_key(vnode->volume->cell);
if (IS_ERR(key))
return PTR_ERR(key);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 6561a1946be2..d9ec8cdca169 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -422,9 +422,6 @@ struct afs_cell {
/* Active fileserver interaction state. */
struct rb_root fs_servers; /* afs_server (by server UUID) */
seqlock_t fs_lock; /* For fs_servers */
- struct rw_semaphore fs_open_mmaps_lock;
- struct list_head fs_open_mmaps; /* List of vnodes that are mmapped */
- atomic_t fs_s_break; /* Counter of CB.InitCallBackState messages */
/* VL server list. */
rwlock_t vl_servers_lock; /* Lock on vl_servers */
@@ -590,9 +587,6 @@ struct afs_server {
/* file service access */
rwlock_t fs_lock; /* access lock */
- /* callback promise management */
- unsigned cb_s_break; /* Break-everything counter. */
-
/* Probe state */
struct afs_endpoint_state __rcu *endpoint_state; /* Latest endpoint/probe state */
unsigned long probed_at; /* Time last probe was dispatched (jiffies) */
@@ -608,6 +602,7 @@ struct afs_server_entry {
struct afs_server *server;
struct afs_volume *volume;
struct list_head slink; /* Link in server->volumes */
+ time64_t cb_expires_at; /* Time at which volume-level callback expires */
};
struct afs_server_list {
@@ -653,10 +648,15 @@ struct afs_volume {
unsigned int servers_seq; /* Incremented each time ->servers changes */
/* Callback management */
+ struct mutex cb_check_lock; /* Lock to control race to check after v_break */
+ time64_t cb_expires_at; /* Earliest volume callback expiry time */
atomic_t cb_ro_snapshot; /* RO volume update-from-snapshot counter */
atomic_t cb_v_break; /* Volume-break event counter. */
+ atomic_t cb_v_check; /* Volume-break has-been-checked counter. */
atomic_t cb_scrub; /* Scrub-all-data event counter. */
rwlock_t cb_v_break_lock;
+ struct rw_semaphore open_mmaps_lock;
+ struct list_head open_mmaps; /* List of vnodes that are mmapped */
afs_voltype_t type; /* type of volume */
char type_force; /* force volume type (suppress R/O -> R/W) */
@@ -695,7 +695,6 @@ struct afs_vnode {
spinlock_t wb_lock; /* lock for wb_keys */
spinlock_t lock; /* waitqueue/flags lock */
unsigned long flags;
-#define AFS_VNODE_CB_PROMISED 0 /* Set if vnode has a callback promise */
#define AFS_VNODE_UNSET 1 /* set if vnode attributes not yet set */
#define AFS_VNODE_DIR_VALID 2 /* Set if dir contents are valid */
#define AFS_VNODE_ZAP_DATA 3 /* set if vnode's data should be invalidated */
@@ -721,13 +720,14 @@ struct afs_vnode {
struct list_head cb_mmap_link; /* Link in cell->fs_open_mmaps */
void *cb_server; /* Server with callback/filelock */
atomic_t cb_nr_mmap; /* Number of mmaps */
- unsigned int cb_fs_s_break; /* Mass server break counter (cell->fs_s_break) */
- unsigned int cb_s_break; /* Mass break counter on ->server */
- unsigned int cb_v_break; /* Mass break counter on ->volume */
+ unsigned int cb_ro_snapshot; /* RO volume release counter on ->volume */
+ unsigned int cb_scrub; /* Scrub counter on ->volume */
unsigned int cb_break; /* Break counter on vnode */
+ unsigned int cb_v_check; /* Break check counter on ->volume */
seqlock_t cb_lock; /* Lock for ->cb_server, ->status, ->cb_*break */
time64_t cb_expires_at; /* time at which callback expires */
+#define AFS_NO_CB_PROMISE TIME64_MIN
};
static inline struct fscache_cookie *afs_vnode_cache(struct afs_vnode *vnode)
@@ -824,7 +824,7 @@ struct afs_vnode_param {
struct afs_fid fid; /* Fid to access */
struct afs_status_cb scb; /* Returned status and callback promise */
afs_dataversion_t dv_before; /* Data version before the call */
- unsigned int cb_break_before; /* cb_break + cb_s_break before the call */
+ unsigned int cb_break_before; /* cb_break before the call */
u8 dv_delta; /* Expected change in data version */
bool put_vnode:1; /* T if we have a ref on the vnode */
bool need_io_lock:1; /* T if we need the I/O lock on this */
@@ -860,7 +860,6 @@ struct afs_operation {
unsigned int debug_id;
unsigned int cb_v_break; /* Volume break counter before op */
- unsigned int cb_s_break; /* Server break counter before op */
union {
struct {
@@ -1051,13 +1050,15 @@ extern void afs_break_callbacks(struct afs_server *, size_t, struct afs_callback
static inline unsigned int afs_calc_vnode_cb_break(struct afs_vnode *vnode)
{
- return vnode->cb_break + vnode->cb_v_break;
+ return vnode->cb_break + vnode->cb_ro_snapshot + vnode->cb_scrub;
}
static inline bool afs_cb_is_broken(unsigned int cb_break,
const struct afs_vnode *vnode)
{
- return cb_break != (vnode->cb_break + atomic_read(&vnode->volume->cb_v_break));
+ return cb_break != (vnode->cb_break +
+ atomic_read(&vnode->volume->cb_ro_snapshot) +
+ atomic_read(&vnode->volume->cb_scrub));
}
/*
@@ -1549,9 +1550,8 @@ extern void afs_fs_exit(void);
/*
* validation.c
*/
+bool afs_check_validity(const struct afs_vnode *vnode);
void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp);
-bool afs_check_validity(struct afs_vnode *vnode);
-bool afs_pagecache_valid(struct afs_vnode *vnode);
int afs_validate(struct afs_vnode *vnode, struct key *key);
/*
diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index a138022d8e0d..3bd02571f30d 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -443,8 +443,8 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
refcount_read(&server->ref),
atomic_read(&server->active),
server->cell->name);
- seq_printf(m, " - info: fl=%lx rtt=%u brk=%x\n",
- server->flags, server->rtt, server->cb_s_break);
+ seq_printf(m, " - info: fl=%lx rtt=%u\n",
+ server->flags, server->rtt);
seq_printf(m, " - probe: last=%d\n",
(int)(jiffies - server->probed_at) / HZ);
failed = estate->failed_set;
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index f2f7e57a4bdd..e0267547f305 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -59,7 +59,7 @@ static bool afs_start_fs_iteration(struct afs_operation *op,
write_seqlock(&vnode->cb_lock);
ASSERTCMP(cb_server, ==, vnode->cb_server);
vnode->cb_server = NULL;
- if (test_and_clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags))
+ if (xchg(&vnode->cb_expires_at, AFS_NO_CB_PROMISE) != AFS_NO_CB_PROMISE)
vnode->cb_break++;
write_sequnlock(&vnode->cb_lock);
}
@@ -480,10 +480,8 @@ bool afs_select_fileserver(struct afs_operation *op)
op->server = server;
if (vnode->cb_server != server) {
vnode->cb_server = server;
- vnode->cb_s_break = server->cb_s_break;
- vnode->cb_fs_s_break = atomic_read(&server->cell->fs_s_break);
- vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
- clear_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
+ vnode->cb_v_check = atomic_read(&vnode->volume->cb_v_break);
+ vnode->cb_expires_at = AFS_NO_CB_PROMISE;
}
read_lock(&server->fs_lock);
diff --git a/fs/afs/server_list.c b/fs/afs/server_list.c
index cfd900eb09ed..c2c2b429ade2 100644
--- a/fs/afs/server_list.c
+++ b/fs/afs/server_list.c
@@ -79,6 +79,7 @@ struct afs_server_list *afs_alloc_server_list(struct afs_volume *volume,
slist->servers[j].server = server;
slist->servers[j].volume = volume;
+ slist->servers[j].cb_expires_at = AFS_NO_CB_PROMISE;
slist->nr_servers++;
}
@@ -174,6 +175,7 @@ void afs_reattach_volume_to_servers(struct afs_volume *volume, struct afs_server
int diff;
if (pn && po && pn->server == po->server) {
+ pn->cb_expires_at = po->cb_expires_at;
list_replace(&po->slink, &pn->slink);
n++;
o++;
diff --git a/fs/afs/validation.c b/fs/afs/validation.c
index f6bdaa121cf6..39c83e7dafb1 100644
--- a/fs/afs/validation.c
+++ b/fs/afs/validation.c
@@ -10,6 +10,131 @@
#include <linux/sched.h>
#include "internal.h"
+/*
+ * Data validation is managed through a number of mechanisms from the server:
+ *
+ * (1) On first contact with a server (such as if it has just been rebooted),
+ * the server sends us a CB.InitCallBackState* request.
+ *
+ * (2) On a RW volume, in response to certain vnode (inode)-accessing RPC
+ * calls, the server maintains a time-limited per-vnode promise that it
+ * will send us a CB.CallBack request if a third party alters the vnodes
+ * accessed.
+ *
+ * Note that a vnode-level callbacks may also be sent for other reasons,
+ * such as filelock release.
+ *
+ * (3) On a RO (or Backup) volume, in response to certain vnode-accessing RPC
+ * calls, each server maintains a time-limited per-volume promise that it
+ * will send us a CB.CallBack request if the RO volume is updated to a
+ * snapshot of the RW volume ("vos release"). This is an atomic event
+ * that cuts over all instances of the RO volume across multiple servers
+ * simultaneously.
+ *
+ * Note that a volume-level callbacks may also be sent for other reasons,
+ * such as the volumeserver taking over control of the volume from the
+ * fileserver.
+ *
+ * Note also that each server maintains an independent time limit on an
+ * independent callback.
+ *
+ * (4) Certain RPC calls include a volume information record "VolSync" in
+ * their reply. This contains a creation date for the volume that should
+ * remain unchanged for a RW volume (but will be changed if the volume is
+ * restored from backup) or will be bumped to the time of snapshotting
+ * when a RO volume is released.
+ *
+ * In order to track this events, the following are provided:
+ *
+ * ->cb_v_break. A counter of events that might mean that the contents of
+ * a volume have been altered since we last checked a vnode.
+ *
+ * ->cb_v_check. A counter of the number of events that we've sent a
+ * query to the server for. Everything's up to date if this equals
+ * cb_v_break.
+ *
+ * ->cb_scrub. A counter of the number of regression events for which we
+ * have to completely wipe the cache.
+ *
+ * ->cb_ro_snapshot. A counter of the number of times that we've
+ * recognised that a RO volume has been updated.
+ *
+ * ->cb_break. A counter of events that might mean that the contents of a
+ * vnode have been altered.
+ *
+ * ->cb_expires_at. The time at which the callback promise expires or
+ * AFS_NO_CB_PROMISE if we have no promise.
+ *
+ * The way we manage things is:
+ *
+ * (1) When a volume-level CB.CallBack occurs, we increment ->cb_v_break on
+ * the volume and reset ->cb_expires_at (ie. set AFS_NO_CB_PROMISE) on the
+ * volume and volume's server record.
+ *
+ * (2) When a CB.InitCallBackState occurs, we treat this as a volume-level
+ * callback break on all the volumes that have been using that volume
+ * (ie. increment ->cb_v_break and reset ->cb_expires_at).
+ *
+ * (3) When a vnode-level CB.CallBack occurs, we increment ->cb_break on the
+ * vnode and reset its ->cb_expires_at. If the vnode is mmapped, we also
+ * dispatch a work item to unmap all PTEs to the vnode's pagecache to
+ * force reentry to the filesystem for revalidation.
+ *
+ * (4) When entering the filesystem, we call afs_validate() to check the
+ * validity of a vnode. This first checks to see if ->cb_v_check and
+ * ->cb_v_break match, and if they don't, we lock volume->cb_check_lock
+ * exclusively and perform an FS.FetchStatus on the vnode.
+ *
+ * After checking the volume, we check the vnode. If there's a mismatch
+ * between the volume counters and the vnode's mirrors of those counters,
+ * we lock vnode->validate_lock and issue an FS.FetchStatus on the vnode.
+ *
+ * (5) When the reply from FS.FetchStatus arrives, the VolSync record is
+ * parsed:
+ *
+ * (A) If the Creation timestamp has changed on a RW volume or regressed
+ * on a RO volume, we try to increment ->cb_scrub; if it advances on a
+ * RO volume, we assume "vos release" happened and try to increment
+ * ->cb_ro_snapshot.
+ *
+ * (B) If the Update timestamp has regressed, we try to increment
+ * ->cb_scrub.
+ *
+ * Note that in both of these cases, we only do the increment if we can
+ * cmpxchg the value of the timestamp from the value we noted before the
+ * op. This tries to prevent parallel ops from fighting one another.
+ *
+ * volume->cb_v_check is then set to ->cb_v_break.
+ *
+ * (6) The AFSCallBack record included in the FS.FetchStatus reply is also
+ * parsed and used to set the promise in ->cb_expires_at for the vnode,
+ * the volume and the volume's server record.
+ *
+ * (7) If ->cb_scrub is seen to have advanced, we invalidate the pagecache for
+ * the vnode.
+ */
+
+/*
+ * Check the validity of a vnode/inode and its parent volume.
+ */
+bool afs_check_validity(const struct afs_vnode *vnode)
+{
+ const struct afs_volume *volume = vnode->volume;
+ time64_t deadline = ktime_get_real_seconds() + 10;
+
+ if (atomic_read(&volume->cb_v_check) != atomic_read(&volume->cb_v_break) ||
+ vnode->cb_expires_at <= deadline ||
+ volume->cb_expires_at <= deadline ||
+ vnode->cb_ro_snapshot != atomic_read(&volume->cb_ro_snapshot) ||
+ vnode->cb_scrub != atomic_read(&volume->cb_scrub) ||
+ test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags)) {
+ _debug("inval");
+ return false;
+ }
+
+ return true;
+}
+
/*
* Handle a change to the volume creation time in the VolSync record.
*/
@@ -99,11 +224,17 @@ static void afs_update_volume_update_time(struct afs_operation *op, struct afs_v
}
/*
- * Update the state of a volume.
+ * Update the state of a volume, including recording the expiration time of the
+ * callback promise.
*/
void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp)
{
+ struct afs_server_list *slist = op->server_list;
+ struct afs_server_entry *se = &slist->servers[op->server_index];
+ struct afs_callback *cb = &vp->scb.callback;
struct afs_volume *volume = op->volume;
+ unsigned int cb_v_break = atomic_read(&volume->cb_v_break);
+ unsigned int cb_v_check = atomic_read(&volume->cb_v_check);
_enter("%llx", op->volume->vid);
@@ -111,6 +242,13 @@ void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *v
afs_update_volume_creation_time(op, volume);
if (op->volsync.mask & AFS_VOLSYNC_UPDATE)
afs_update_volume_update_time(op, volume);
+
+ if (op->cb_v_break == cb_v_break) {
+ se->cb_expires_at = cb->expires_at;
+ volume->cb_expires_at = cb->expires_at;
+ }
+ if (cb_v_check < op->cb_v_break)
+ atomic_cmpxchg(&volume->cb_v_check, cb_v_check, op->cb_v_break);
}
/*
@@ -132,101 +270,6 @@ static void afs_zap_data(struct afs_vnode *vnode)
invalidate_inode_pages2(vnode->netfs.inode.i_mapping);
}
-/*
- * Check to see if we have a server currently serving this volume and that it
- * hasn't been reinitialised or dropped from the list.
- */
-static bool afs_check_server_good(struct afs_vnode *vnode)
-{
- struct afs_server_list *slist;
- struct afs_server *server;
- bool good;
- int i;
-
- if (vnode->cb_fs_s_break == atomic_read(&vnode->volume->cell->fs_s_break))
- return true;
-
- rcu_read_lock();
-
- slist = rcu_dereference(vnode->volume->servers);
- for (i = 0; i < slist->nr_servers; i++) {
- server = slist->servers[i].server;
- if (server == vnode->cb_server) {
- good = (vnode->cb_s_break == server->cb_s_break);
- rcu_read_unlock();
- return good;
- }
- }
-
- rcu_read_unlock();
- return false;
-}
-
-/*
- * Check the validity of a vnode/inode.
- */
-bool afs_check_validity(struct afs_vnode *vnode)
-{
- enum afs_cb_break_reason need_clear = afs_cb_break_no_break;
- time64_t now = ktime_get_real_seconds();
- unsigned int cb_break;
- int seq = 0;
-
- do {
- read_seqbegin_or_lock(&vnode->cb_lock, &seq);
- cb_break = vnode->cb_break;
-
- if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
- if (vnode->cb_v_break != atomic_read(&vnode->volume->cb_v_break))
- need_clear = afs_cb_break_for_v_break;
- else if (!afs_check_server_good(vnode))
- need_clear = afs_cb_break_for_s_reinit;
- else if (test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
- need_clear = afs_cb_break_for_zap;
- else if (vnode->cb_expires_at - 10 <= now)
- need_clear = afs_cb_break_for_lapsed;
- } else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
- ;
- } else {
- need_clear = afs_cb_break_no_promise;
- }
-
- } while (need_seqretry(&vnode->cb_lock, seq));
-
- done_seqretry(&vnode->cb_lock, seq);
-
- if (need_clear == afs_cb_break_no_break)
- return true;
-
- write_seqlock(&vnode->cb_lock);
- if (need_clear == afs_cb_break_no_promise)
- vnode->cb_v_break = atomic_read(&vnode->volume->cb_v_break);
- else if (cb_break == vnode->cb_break)
- __afs_break_callback(vnode, need_clear);
- else
- trace_afs_cb_miss(&vnode->fid, need_clear);
- write_sequnlock(&vnode->cb_lock);
- return false;
-}
-
-/*
- * Returns true if the pagecache is still valid. Does not sleep.
- */
-bool afs_pagecache_valid(struct afs_vnode *vnode)
-{
- if (unlikely(test_bit(AFS_VNODE_DELETED, &vnode->flags))) {
- if (vnode->netfs.inode.i_nlink)
- clear_nlink(&vnode->netfs.inode);
- return true;
- }
-
- if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags) &&
- afs_check_validity(vnode))
- return true;
-
- return false;
-}
-
/*
* validate a vnode/inode
* - there are several things we need to check
@@ -238,23 +281,48 @@ bool afs_pagecache_valid(struct afs_vnode *vnode)
*/
int afs_validate(struct afs_vnode *vnode, struct key *key)
{
+ struct afs_volume *volume = vnode->volume;
+ unsigned int cb_ro_snapshot, cb_scrub;
+ time64_t deadline = ktime_get_real_seconds() + 10;
+ bool zap = false, locked_vol = false;
int ret;
_enter("{v={%llx:%llu} fl=%lx},%x",
vnode->fid.vid, vnode->fid.vnode, vnode->flags,
key_serial(key));
- if (afs_pagecache_valid(vnode))
- goto valid;
+ if (afs_check_validity(vnode))
+ return true;
+
+ ret = down_write_killable(&vnode->validate_lock);
+ if (ret < 0)
+ goto error;
- down_write(&vnode->validate_lock);
+ /* Validate a volume after the v_break has changed or the volume
+ * callback expired. We only want to do this once per volume per
+ * v_break change. The actual work will be done when parsing the
+ * status fetch reply.
+ */
+ if (volume->cb_expires_at <= deadline ||
+ atomic_read(&volume->cb_v_check) != atomic_read(&volume->cb_v_break)) {
+ ret = mutex_lock_interruptible(&volume->cb_check_lock);
+ if (ret < 0)
+ goto error_unlock;
+ locked_vol = true;
+ }
- /* if the promise has expired, we need to check the server again to get
- * a new promise - note that if the (parent) directory's metadata was
- * changed then the security may be different and we may no longer have
- * access */
- if (!test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
- _debug("not promised");
+ cb_ro_snapshot = atomic_read(&volume->cb_ro_snapshot);
+ cb_scrub = atomic_read(&volume->cb_scrub);
+ if (vnode->cb_ro_snapshot != cb_ro_snapshot ||
+ vnode->cb_scrub != cb_scrub)
+ unmap_mapping_pages(vnode->netfs.inode.i_mapping, 0, 0, false);
+
+ if (vnode->cb_ro_snapshot != cb_ro_snapshot ||
+ vnode->cb_scrub != cb_scrub ||
+ volume->cb_expires_at <= deadline ||
+ atomic_read(&volume->cb_v_check) != atomic_read(&volume->cb_v_break) ||
+ vnode->cb_expires_at <= deadline
+ ) {
ret = afs_fetch_status(vnode, key, false, NULL);
if (ret < 0) {
if (ret == -ENOENT) {
@@ -263,9 +331,26 @@ int afs_validate(struct afs_vnode *vnode, struct key *key)
}
goto error_unlock;
}
+
_debug("new promise [fl=%lx]", vnode->flags);
}
+ /* We can drop the volume lock now as. */
+ if (locked_vol) {
+ mutex_unlock(&volume->cb_check_lock);
+ locked_vol = false;
+ }
+
+ cb_ro_snapshot = atomic_read(&volume->cb_ro_snapshot);
+ cb_scrub = atomic_read(&volume->cb_scrub);
+ _debug("vnode inval %x==%x %x==%x",
+ vnode->cb_ro_snapshot, cb_ro_snapshot,
+ vnode->cb_scrub, cb_scrub);
+ if (vnode->cb_scrub != cb_scrub)
+ zap = true;
+ vnode->cb_ro_snapshot = cb_ro_snapshot;
+ vnode->cb_scrub = cb_scrub;
+
if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
_debug("file already deleted");
ret = -ESTALE;
@@ -274,15 +359,18 @@ int afs_validate(struct afs_vnode *vnode, struct key *key)
/* if the vnode's data version number changed then its contents are
* different */
- if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
+ zap |= test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags);
+ if (zap)
afs_zap_data(vnode);
up_write(&vnode->validate_lock);
-valid:
_leave(" = 0");
return 0;
error_unlock:
+ if (locked_vol)
+ mutex_unlock(&volume->cb_check_lock);
up_write(&vnode->validate_lock);
+error:
_leave(" = %d", ret);
return ret;
}
diff --git a/fs/afs/volume.c b/fs/afs/volume.c
index 00253032ccca..aebdeeed450a 100644
--- a/fs/afs/volume.c
+++ b/fs/afs/volume.c
@@ -90,6 +90,8 @@ static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
INIT_WORK(&volume->destructor, afs_destroy_volume);
rwlock_init(&volume->servers_lock);
rwlock_init(&volume->cb_v_break_lock);
+ INIT_LIST_HEAD(&volume->open_mmaps);
+ init_rwsem(&volume->open_mmaps_lock);
memcpy(volume->name, vldb->name, vldb->name_len + 1);
for (i = 0; i < AFS_MAXTYPES; i++)
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 201dac3be54f..3905fd9c957f 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -438,18 +438,14 @@ enum yfs_cm_operation {
#define afs_cb_break_reasons \
EM(afs_cb_break_no_break, "no-break") \
- EM(afs_cb_break_no_promise, "no-promise") \
EM(afs_cb_break_for_callback, "break-cb") \
EM(afs_cb_break_for_creation_regress, "creation-regress") \
EM(afs_cb_break_for_deleted, "break-del") \
- EM(afs_cb_break_for_lapsed, "break-lapsed") \
EM(afs_cb_break_for_s_reinit, "s-reinit") \
EM(afs_cb_break_for_unlink, "break-unlink") \
EM(afs_cb_break_for_update_regress, "update-regress") \
- EM(afs_cb_break_for_v_break, "break-v") \
EM(afs_cb_break_for_volume_callback, "break-v-cb") \
- EM(afs_cb_break_for_vos_release, "break-vos-release") \
- E_(afs_cb_break_for_zap, "break-zap")
+ E_(afs_cb_break_for_vos_release, "break-vos-release")
/*
* Generate enums for tracing information.
^ permalink raw reply related [flat|nested] 57+ messages in thread* Re: [PATCH 39/41] afs: Overhaul invalidation handling to better support RO volumes
2023-11-09 15:40 ` [PATCH 39/41] afs: Overhaul invalidation handling to better support RO volumes David Howells
@ 2023-11-09 19:00 ` Jeffrey E Altman
2023-11-13 15:58 ` [PATCH 42/41] afs: Fix the handling of " David Howells
1 sibling, 0 replies; 57+ messages in thread
From: Jeffrey E Altman @ 2023-11-09 19:00 UTC (permalink / raw)
To: David Howells, Marc Dionne; +Cc: linux-afs, linux-fsdevel, linux-kernel
[-- Attachment #1: Type: text/plain, Size: 2237 bytes --]
On 11/9/2023 10:40 AM, David Howells wrote:
> This allows better handling of RO (and Backup) volumes. Since these are
> snapshot of a RW volume that are updated atomically simultantanously across
> all servers that host them, they only require a single callback promise for
> the entire volume.
The atomic visibility of a volume cloning operation is limited to
individual volume locations.
It is untrue that a "vos release" updates all RO volume locations
simultaneously. If that were
to occur then there would be the potential for all volume locations to
be offline simultaneously
causing an outage. Instead, the traditional "volume release" process
updates the RO volume
site co-located with the RW volume site and makes that version visible
to cache managers.
Then 50% of the remaining sites are updated in parallel followed by the
remaining sites.
Each time a site is updated to the new version the volume location entry
is updated to flag
the site as VLSF_NEWREPSITE. However, changes to the volume location
site flags might
not be known to the cache manager. Prior to completion of the "volume
release" process
it is possible for a cache manager to failover from a newer snapshot to
an older one.
> The currently upstream code assumes that RO volumes
> operate in the same manner as RW volumes, and that each file has its own
> individual callback - which means that it does a status fetch for *every*
> file in a RO volume, whether or not the volume got "released" (volume
> callback breaks can occur for other reasons too, such as the volumeserver
> taking ownership of a volume from a fileserver).
Knowledge that the volume snapshot has not changed can be used to avoid
individual
FetchStatus queries when accessing vnodes anonymously. In that case
the vnode's
FetchStatus.AnonymousAccess can be trusted to be unchanged. However, for
authenticated
access the individual FetchStatus or InlineBulkStat queries must still
be issued because
the FetchStatus.CallerAccess rights are determined not only by the
access control list
contents but the caller identity's group memberships which might have
changed. This
distinction only matters when evaluating authorization decisions.
Jeffrey Altman
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4039 bytes --]
^ permalink raw reply [flat|nested] 57+ messages in thread
* [PATCH 42/41] afs: Fix the handling of RO volumes
2023-11-09 15:40 ` [PATCH 39/41] afs: Overhaul invalidation handling to better support RO volumes David Howells
2023-11-09 19:00 ` Jeffrey E Altman
@ 2023-11-13 15:58 ` David Howells
1 sibling, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-13 15:58 UTC (permalink / raw)
To: Jeffrey E Altman
Cc: dhowells, Marc Dionne, linux-afs, linux-fsdevel, linux-kernel
Jeffrey E Altman <jaltman@auristor.com> wrote:
> The atomic visibility of a volume cloning operation is limited to individual
> volume locations. It is untrue that a "vos release" updates all RO volume
> locations simultaneously.
[This need folding down into the preceding patches]
Fix the handling of RO volumes which aren't actually updated all at once,
but are rather updated in stages.
Switch from using the old servers to the new servers when at least half the
servers are marked as new replication sites (VLSF_NEWREPSITE).
If an operation gets a reply from a server that indicates that an RO volume
is of a more recent snapshot, we regenerate the server list and that
excludes whichever bunch of servers we don't want to look at. If the
server we got the reply from is now excluded, we reissue the operation,
which will cause the rotation algorithm to start agian.
Whilst it thinks that RO replication is in progress, it rechecks the VLDB
record every 2s. The problem is that as we don't then maintain a callback
on an excluded server, we won't get a notification of completion of the
replication process and will maintain the exclusion until the normal VLDB
recheck happens after an hour.
To simplify things, the volume creation and update times are set to
TIME64_MIN when not valid rather than using a separate flag, and a mutex is
used to serialise attempts to modify them.
David
---
fs/afs/afs.h | 7 -
fs/afs/dir.c | 4 -
fs/afs/fs_operation.c | 5 -
fs/afs/fsclient.c | 4 -
fs/afs/inode.c | 5 -
fs/afs/internal.h | 17 +++-
fs/afs/rotate.c | 22 +++++-
fs/afs/server_list.c | 52 ++++++++++++--
fs/afs/validation.c | 164 +++++++++++++++++++++++++++++++++++----------
fs/afs/vlclient.c | 19 -----
fs/afs/volume.c | 10 ++
fs/afs/yfsclient.c | 1
include/trace/events/afs.h | 3
13 files changed, 230 insertions(+), 83 deletions(-)
diff --git a/fs/afs/afs.h b/fs/afs/afs.h
index a6328a9ec9b0..b488072aee87 100644
--- a/fs/afs/afs.h
+++ b/fs/afs/afs.h
@@ -165,11 +165,8 @@ struct afs_status_cb {
* AFS volume synchronisation information
*/
struct afs_volsync {
- unsigned int mask; /* Bitmask of supplied fields */
-#define AFS_VOLSYNC_CREATION 0x01
-#define AFS_VOLSYNC_UPDATE 0x02
- time64_t creation; /* volume creation time */
- time64_t update; /* Volume update time */
+ time64_t creation; /* Volume creation time (or TIME64_MIN) */
+ time64_t update; /* Volume update time (or TIME64_MIN) */
};
/*
diff --git a/fs/afs/dir.c b/fs/afs/dir.c
index 190d1ff42e60..7ea5803c05da 100644
--- a/fs/afs/dir.c
+++ b/fs/afs/dir.c
@@ -807,8 +807,8 @@ static struct inode *afs_do_lookup(struct inode *dir, struct dentry *dentry,
cookie->fids[i].vid = dvnode->fid.vid;
cookie->ctx.actor = afs_lookup_filldir;
cookie->name = dentry->d_name;
- cookie->nr_fids = 2; /* slot 0 is saved for the fid we actually want
- * and slot 1 for the directory */
+ cookie->nr_fids = 2; /* slot 1 is saved for the fid we actually want
+ * and slot 0 for the directory */
if (!afs_server_supports_ibulk(dvnode))
cookie->one_only = true;
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index e0f5920a5aef..3546b087e791 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -39,9 +39,8 @@ struct afs_operation *afs_alloc_operation(struct key *key, struct afs_volume *vo
op->volume = afs_get_volume(volume, afs_volume_trace_get_new_op);
op->net = volume->cell->net;
op->cb_v_break = atomic_read(&volume->cb_v_break);
- op->pre_volsync.mask = READ_ONCE(volume->volsync_mask);
- op->pre_volsync.creation = atomic64_read(&volume->creation_time);
- op->pre_volsync.update = atomic64_read(&volume->update_time);
+ op->pre_volsync.creation = volume->creation_time;
+ op->pre_volsync.update = volume->update_time;
op->debug_id = atomic_inc_return(&afs_operation_debug_counter);
op->nr_iterations = -1;
afs_op_set_error(op, -EDESTADDRREQ);
diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c
index 1bd1614688f7..80f7d9e796e3 100644
--- a/fs/afs/fsclient.c
+++ b/fs/afs/fsclient.c
@@ -165,10 +165,8 @@ static void xdr_decode_AFSVolSync(const __be32 **_bp,
bp++; /* spare6 */
*_bp = bp;
- if (volsync) {
+ if (volsync)
volsync->creation = creation;
- volsync->mask |= AFS_VOLSYNC_CREATION;
- }
}
/*
diff --git a/fs/afs/inode.c b/fs/afs/inode.c
index 04c335fab1a5..0fe1c63c9af5 100644
--- a/fs/afs/inode.c
+++ b/fs/afs/inode.c
@@ -212,7 +212,8 @@ static void afs_apply_status(struct afs_operation *op,
vnode->status = *status;
if (vp->dv_before + vp->dv_delta != status->data_version) {
- if (vnode->cb_expires_at != AFS_NO_CB_PROMISE)
+ if (vnode->cb_ro_snapshot == atomic_read(&vnode->volume->cb_ro_snapshot) &&
+ vnode->cb_expires_at != AFS_NO_CB_PROMISE)
pr_warn("kAFS: vnode modified {%llx:%llu} %llx->%llx %s (op=%x)\n",
vnode->fid.vid, vnode->fid.vnode,
(unsigned long long)vp->dv_before + vp->dv_delta,
@@ -327,8 +328,6 @@ static void afs_fetch_status_success(struct afs_operation *op)
struct afs_vnode *vnode = vp->vnode;
int ret;
- afs_update_volume_state(op, vp);
-
if (vnode->netfs.inode.i_state & I_NEW) {
ret = afs_inode_init_from_status(op, vp, vnode);
afs_op_set_error(op, ret);
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index 43e438697baa..0c41601324fc 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -512,6 +512,7 @@ struct afs_vldb_entry {
#define AFS_VOL_VTM_RW 0x01 /* R/W version of the volume is available (on this server) */
#define AFS_VOL_VTM_RO 0x02 /* R/O version of the volume is available (on this server) */
#define AFS_VOL_VTM_BAK 0x04 /* backup version of the volume is available (on this server) */
+ u8 vlsf_flags[AFS_NMAXNSERVERS];
short error;
u8 nr_servers; /* Number of server records */
u8 name_len;
@@ -595,6 +596,12 @@ struct afs_server {
spinlock_t probe_lock;
};
+enum afs_ro_replicating {
+ AFS_RO_NOT_REPLICATING, /* Not doing replication */
+ AFS_RO_REPLICATING_USE_OLD, /* Replicating; use old version */
+ AFS_RO_REPLICATING_USE_NEW, /* Replicating; switch to new version */
+} __mode(byte);
+
/*
* Replaceable volume server list.
*/
@@ -606,12 +613,14 @@ struct afs_server_entry {
unsigned long flags;
#define AFS_SE_VOLUME_OFFLINE 0 /* Set if volume offline notice given */
#define AFS_SE_VOLUME_BUSY 1 /* Set if volume busy notice given */
+#define AFS_SE_EXCLUDED 2 /* Set if server is to be excluded in rotation */
};
struct afs_server_list {
struct rcu_head rcu;
refcount_t usage;
bool attached; /* T if attached to servers */
+ enum afs_ro_replicating ro_replicating; /* RW->RO update (probably) in progress */
unsigned char nr_servers;
unsigned short vnovol_mask; /* Servers to be skipped due to VNOVOL */
unsigned int seq; /* Set to ->servers_seq when installed */
@@ -647,9 +656,9 @@ struct afs_volume {
unsigned int servers_seq; /* Incremented each time ->servers changes */
/* RO release tracking */
- unsigned int volsync_mask; /* Mask of what values we have obtained */
- atomic64_t creation_time; /* Volume creation time (time64_t) */
- atomic64_t update_time; /* Volume update time (time64_t) */
+ struct mutex volsync_lock; /* Time/state evaluation lock */
+ time64_t creation_time; /* Volume creation time (or TIME64_MIN) */
+ time64_t update_time; /* Volume update time (or TIME64_MIN) */
/* Callback management */
struct mutex cb_check_lock; /* Lock to control race to check after v_break */
@@ -1571,7 +1580,7 @@ extern void afs_fs_exit(void);
* validation.c
*/
bool afs_check_validity(const struct afs_vnode *vnode);
-void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp);
+int afs_update_volume_state(struct afs_operation *op);
int afs_validate(struct afs_vnode *vnode, struct key *key);
/*
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 099e0d1caa67..79c736eeca29 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -187,6 +187,22 @@ bool afs_select_fileserver(struct afs_operation *op)
clear_bit(AFS_SE_VOLUME_BUSY,
&op->server_list->servers[op->server_index].flags);
op->cumul_error.responded = true;
+
+ /* We succeeded, but we may need to check the VLDB and redo the
+ * op from another server if we're looking at a set of RO
+ * volumes where some of the servers have not yet been brought
+ * up to date to avoid regressing the data. We only switch to
+ * the new version once >=50% of the servers are updated.
+ */
+ error = afs_update_volume_state(op);
+ if (error != 0) {
+ if (error == 1) {
+ afs_sleep_and_retry(op);
+ goto restart_from_beginning;
+ }
+ afs_op_set_error(op, error);
+ goto failed;
+ }
fallthrough;
default:
/* Success or local failure. Stop. */
@@ -517,10 +533,12 @@ bool afs_select_fileserver(struct afs_operation *op)
best_prio = -1;
for (i = 0; i < op->server_list->nr_servers; i++) {
struct afs_endpoint_state *es;
+ struct afs_server_entry *se = &op->server_list->servers[i];
struct afs_addr_list *sal;
- struct afs_server *s = op->server_list->servers[i].server;
+ struct afs_server *s = se->server;
if (!test_bit(i, &op->untried_servers) ||
+ test_bit(AFS_SE_EXCLUDED, &se->flags) ||
!test_bit(AFS_SERVER_FL_RESPONDING, &s->flags))
continue;
es = op->server_states->endpoint_state;
@@ -598,6 +616,8 @@ bool afs_select_fileserver(struct afs_operation *op)
op->addr_index = addr_index;
set_bit(addr_index, &op->addr_tried);
+ op->volsync.creation = TIME64_MIN;
+ op->volsync.update = TIME64_MIN;
op->call_responded = false;
_debug("address [%u] %u/%u %pISp",
op->server_index, addr_index, alist->nr_addrs,
diff --git a/fs/afs/server_list.c b/fs/afs/server_list.c
index c9f146f69adf..5d45ff95946c 100644
--- a/fs/afs/server_list.c
+++ b/fs/afs/server_list.c
@@ -31,23 +31,55 @@ struct afs_server_list *afs_alloc_server_list(struct afs_volume *volume,
struct afs_server_list *slist;
struct afs_server *server;
unsigned int type_mask = 1 << volume->type;
- int ret = -ENOMEM, nr_servers = 0, i, j;
-
- for (i = 0; i < vldb->nr_servers; i++)
- if (vldb->fs_mask[i] & type_mask)
- nr_servers++;
+ bool use_newrepsites = false;
+ int ret = -ENOMEM, nr_servers = 0, newrep = 0, i, j;
+ u32 tmp;
+
+ /* Work out if we're going to restrict to NEWREPSITE-marked servers or
+ * not. If at least one site is marked as NEWREPSITE, then it's likely
+ * that "vos release" is busy updating RO sites. We cut over from one
+ * to the other when >=50% of the sites have been updated. Sites that
+ * are in the process of being updated are marked DONTUSE.
+ */
+ for (i = 0; i < vldb->nr_servers; i++) {
+ if (!(vldb->fs_mask[i] & type_mask))
+ continue;
+ nr_servers++;
+ if (vldb->vlsf_flags[i] & AFS_VLSF_DONTUSE)
+ continue;
+ if (vldb->vlsf_flags[i] & AFS_VLSF_NEWREPSITE)
+ newrep++;
+ }
slist = kzalloc(struct_size(slist, servers, nr_servers), GFP_KERNEL);
if (!slist)
goto error;
+ if (newrep) {
+ if (newrep < nr_servers / 2) {
+ kdebug("USE-OLD");
+ slist->ro_replicating = AFS_RO_REPLICATING_USE_OLD;
+ } else {
+ kdebug("USE-NEW");
+ slist->ro_replicating = AFS_RO_REPLICATING_USE_NEW;
+ use_newrepsites = true;
+ }
+ }
+
refcount_set(&slist->usage, 1);
rwlock_init(&slist->lock);
/* Make sure a records exists for each server in the list. */
for (i = 0; i < vldb->nr_servers; i++) {
+ unsigned long se_flags = 0;
+ bool newrepsite = tmp & AFS_VLSF_NEWREPSITE;
+
if (!(vldb->fs_mask[i] & type_mask))
continue;
+ if (tmp & AFS_VLSF_DONTUSE)
+ __set_bit(AFS_SE_EXCLUDED, &se_flags);
+ if (newrep && (newrepsite ^ use_newrepsites))
+ __set_bit(AFS_SE_EXCLUDED, &se_flags);
server = afs_lookup_server(volume->cell, key, &vldb->fs_server[i],
vldb->addr_version[i]);
@@ -79,6 +111,7 @@ struct afs_server_list *afs_alloc_server_list(struct afs_volume *volume,
slist->servers[j].server = server;
slist->servers[j].volume = volume;
+ slist->servers[j].flags = se_flags;
slist->servers[j].cb_expires_at = AFS_NO_CB_PROMISE;
slist->nr_servers++;
}
@@ -102,14 +135,19 @@ struct afs_server_list *afs_alloc_server_list(struct afs_volume *volume,
bool afs_annotate_server_list(struct afs_server_list *new,
struct afs_server_list *old)
{
+ unsigned long mask = 1UL << AFS_SE_EXCLUDED;
int i;
- if (old->nr_servers != new->nr_servers)
+ if (old->nr_servers != new->nr_servers ||
+ old->ro_replicating != new->ro_replicating)
goto changed;
- for (i = 0; i < old->nr_servers; i++)
+ for (i = 0; i < old->nr_servers; i++) {
if (old->servers[i].server != new->servers[i].server)
goto changed;
+ if ((old->servers[i].flags & mask) != (new->servers[i].flags & mask))
+ goto changed;
+ }
return false;
changed:
return true;
diff --git a/fs/afs/validation.c b/fs/afs/validation.c
index 3802790fe4de..f8e24439727a 100644
--- a/fs/afs/validation.c
+++ b/fs/afs/validation.c
@@ -135,49 +135,115 @@ bool afs_check_validity(const struct afs_vnode *vnode)
return true;
}
+/*
+ * See if the server we've just talked to is currently excluded.
+ */
+static bool __afs_is_server_excluded(struct afs_operation *op, struct afs_volume *volume)
+{
+ const struct afs_server_entry *se;
+ const struct afs_server_list *slist;
+ bool is_excluded = true;
+ int i;
+
+ rcu_read_lock();
+
+ slist = rcu_dereference(volume->servers);
+ for (i = 0; i < slist->nr_servers; i++) {
+ se = &slist->servers[i];
+ if (op->server == se->server) {
+ is_excluded = test_bit(AFS_SE_EXCLUDED, &se->flags);
+ break;
+ }
+ }
+
+ rcu_read_unlock();
+ return is_excluded;
+}
+
+/*
+ * Update the volume's server list when the creation time changes and see if
+ * the server we've just talked to is currently excluded.
+ */
+static int afs_is_server_excluded(struct afs_operation *op, struct afs_volume *volume)
+{
+ int ret;
+
+ if (__afs_is_server_excluded(op, volume))
+ return 1;
+
+ set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
+ ret = afs_check_volume_status(op->volume, op);
+ if (ret < 0)
+ return ret;
+
+ return __afs_is_server_excluded(op, volume);
+}
+
/*
* Handle a change to the volume creation time in the VolSync record.
*/
-static void afs_update_volume_creation_time(struct afs_operation *op, struct afs_volume *volume)
+static int afs_update_volume_creation_time(struct afs_operation *op, struct afs_volume *volume)
{
- enum afs_cb_break_reason reason = afs_cb_break_for_vos_release;
unsigned int snap;
- time64_t cur = atomic64_read(&volume->creation_time);
+ time64_t cur = volume->creation_time;
time64_t old = op->pre_volsync.creation;
time64_t new = op->volsync.creation;
+ int ret;
_enter("%llx,%llx,%llx->%llx", volume->vid, cur, old, new);
- if (!(op->volsync.mask & AFS_VOLSYNC_CREATION) ||
- !(volume->volsync_mask & AFS_VOLSYNC_CREATION)) {
- atomic64_set(&volume->creation_time, new);
- volume->volsync_mask |= AFS_VOLSYNC_CREATION;
- return;
+ if (cur == TIME64_MIN) {
+ volume->creation_time = new;
+ return 0;
}
- if (likely(new == cur))
- return;
+ if (new == cur)
+ return 0;
+
+ /* Try to advance the creation timestamp from what we had before the
+ * operation to what we got back from the server. This should
+ * hopefully ensure that in a race between multiple operations only one
+ * of them will do this.
+ */
+ if (cur != old)
+ return 0;
/* If the creation time changes in an unexpected way, we need to scrub
* our caches. For a RW vol, this will only change if the volume is
* restored from a backup; for a RO/Backup vol, this will advance when
* the volume is updated to a new snapshot (eg. "vos release").
*/
- if (volume->type == AFSVL_RWVOL || new < old)
- reason = afs_cb_break_for_creation_regress;
+ if (volume->type == AFSVL_RWVOL)
+ goto regressed;
+ if (volume->type == AFSVL_BACKVOL) {
+ if (new < old)
+ goto regressed;
+ goto advance;
+ }
- /* Try to advance the creation timestamp from what we had before the
- * operation to what we got back from the server. This should
- * hopefully ensure that in a race between multiple operations only one
- * of them will do this.
+ /* We have an RO volume, we need to query the VL server and look at the
+ * server flags to see if RW->RO replication is in progress.
*/
- if (atomic64_try_cmpxchg(&volume->creation_time, &old, new)) {
- if (reason == afs_cb_break_for_creation_regress)
- atomic_inc(&volume->cb_scrub);
- else if (volume->type != AFSVL_RWVOL)
- snap = atomic_inc_return(&volume->cb_ro_snapshot);
- trace_afs_cb_v_break(volume->vid, snap, reason);
+ ret = afs_is_server_excluded(op, volume);
+ if (ret < 0)
+ return ret;
+ if (ret > 0) {
+ snap = atomic_read(&volume->cb_ro_snapshot);
+ trace_afs_cb_v_break(volume->vid, snap, afs_cb_break_volume_excluded);
+ return ret;
}
+
+advance:
+ snap = atomic_inc_return(&volume->cb_ro_snapshot);
+ trace_afs_cb_v_break(volume->vid, snap, afs_cb_break_for_vos_release);
+ volume->creation_time = new;
+ return 0;
+
+regressed:
+ atomic_inc(&volume->cb_scrub);
+ trace_afs_cb_v_break(volume->vid, 0, afs_cb_break_for_creation_regress);
+ volume->creation_time = new;
+ return 0;
}
/*
@@ -186,20 +252,18 @@ static void afs_update_volume_creation_time(struct afs_operation *op, struct afs
static void afs_update_volume_update_time(struct afs_operation *op, struct afs_volume *volume)
{
enum afs_cb_break_reason reason = afs_cb_break_no_break;
- time64_t cur = atomic64_read(&volume->update_time);
+ time64_t cur = volume->update_time;
time64_t old = op->pre_volsync.update;
time64_t new = op->volsync.update;
_enter("%llx,%llx,%llx->%llx", volume->vid, cur, old, new);
- if (!(op->volsync.mask & AFS_VOLSYNC_UPDATE) ||
- !(volume->volsync_mask & AFS_VOLSYNC_UPDATE)) {
- atomic64_set(&volume->update_time, new);
- volume->volsync_mask |= AFS_VOLSYNC_UPDATE;
+ if (cur == TIME64_MIN) {
+ volume->update_time = new;
return;
}
- if (likely(new == cur))
+ if (new == cur)
return;
/* If the volume update time changes in an unexpected way, we need to
@@ -215,33 +279,59 @@ static void afs_update_volume_update_time(struct afs_operation *op, struct afs_v
* hopefully ensure that in a race between multiple operations only one
* of them will do this.
*/
- if (atomic64_try_cmpxchg(&volume->update_time, &old, new)) {
+ if (cur == old) {
if (reason == afs_cb_break_for_update_regress) {
atomic_inc(&volume->cb_scrub);
trace_afs_cb_v_break(volume->vid, 0, reason);
}
+ volume->update_time = new;
+ }
+}
+
+static int afs_update_volume_times(struct afs_operation *op, struct afs_volume *volume)
+{
+ int ret = 0;
+
+ if (likely(op->volsync.creation == volume->creation_time &&
+ op->volsync.update == volume->update_time))
+ return 0;
+
+ mutex_lock(&volume->volsync_lock);
+ if (op->volsync.creation != volume->creation_time) {
+ ret = afs_update_volume_creation_time(op, volume);
+ if (ret < 0)
+ goto out;
}
+ if (op->volsync.update != volume->update_time)
+ afs_update_volume_update_time(op, volume);
+out:
+ mutex_unlock(&volume->volsync_lock);
+ return ret;
}
/*
* Update the state of a volume, including recording the expiration time of the
- * callback promise.
+ * callback promise. Returns 1 to redo the operation from the start.
*/
-void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *vp)
+int afs_update_volume_state(struct afs_operation *op)
{
struct afs_server_list *slist = op->server_list;
struct afs_server_entry *se = &slist->servers[op->server_index];
- struct afs_callback *cb = &vp->scb.callback;
+ struct afs_callback *cb = &op->file[0].scb.callback;
struct afs_volume *volume = op->volume;
unsigned int cb_v_break = atomic_read(&volume->cb_v_break);
unsigned int cb_v_check = atomic_read(&volume->cb_v_check);
+ int ret;
_enter("%llx", op->volume->vid);
- if (op->volsync.mask & AFS_VOLSYNC_CREATION)
- afs_update_volume_creation_time(op, volume);
- if (op->volsync.mask & AFS_VOLSYNC_UPDATE)
- afs_update_volume_update_time(op, volume);
+ if (op->volsync.creation != TIME64_MIN || op->volsync.update != TIME64_MIN) {
+ ret = afs_update_volume_times(op, volume);
+ if (ret != 0) {
+ _leave(" = %d", ret);
+ return ret;
+ }
+ }
if (op->cb_v_break == cb_v_break) {
se->cb_expires_at = cb->expires_at;
@@ -249,6 +339,8 @@ void afs_update_volume_state(struct afs_operation *op, struct afs_vnode_param *v
}
if (cb_v_check < op->cb_v_break)
atomic_cmpxchg(&volume->cb_v_check, cb_v_check, op->cb_v_break);
+ _leave(" = %d", ret);
+ return ret;
}
/*
diff --git a/fs/afs/vlclient.c b/fs/afs/vlclient.c
index cef02a265edc..cac75f89b64a 100644
--- a/fs/afs/vlclient.c
+++ b/fs/afs/vlclient.c
@@ -18,8 +18,7 @@ static int afs_deliver_vl_get_entry_by_name_u(struct afs_call *call)
{
struct afs_uvldbentry__xdr *uvldb;
struct afs_vldb_entry *entry;
- bool new_only = false;
- u32 tmp, nr_servers, vlflags;
+ u32 nr_servers, vlflags;
int i, ret;
_enter("");
@@ -41,27 +40,14 @@ static int afs_deliver_vl_get_entry_by_name_u(struct afs_call *call)
entry->name[i] = 0;
entry->name_len = strlen(entry->name);
- /* If there is a new replication site that we can use, ignore all the
- * sites that aren't marked as new.
- */
- for (i = 0; i < nr_servers; i++) {
- tmp = ntohl(uvldb->serverFlags[i]);
- if (!(tmp & AFS_VLSF_DONTUSE) &&
- (tmp & AFS_VLSF_NEWREPSITE))
- new_only = true;
- }
-
vlflags = ntohl(uvldb->flags);
for (i = 0; i < nr_servers; i++) {
struct afs_uuid__xdr *xdr;
struct afs_uuid *uuid;
+ u32 tmp = ntohl(uvldb->serverFlags[i]);
int j;
int n = entry->nr_servers;
- tmp = ntohl(uvldb->serverFlags[i]);
- if (tmp & AFS_VLSF_DONTUSE ||
- (new_only && !(tmp & AFS_VLSF_NEWREPSITE)))
- continue;
if (tmp & AFS_VLSF_RWVOL) {
entry->fs_mask[n] |= AFS_VOL_VTM_RW;
if (vlflags & AFS_VLF_BACKEXISTS)
@@ -82,6 +68,7 @@ static int afs_deliver_vl_get_entry_by_name_u(struct afs_call *call)
for (j = 0; j < 6; j++)
uuid->node[j] = (u8)ntohl(xdr->node[j]);
+ entry->vlsf_flags[n] = tmp;
entry->addr_version[n] = ntohl(uvldb->serverUnique[i]);
entry->nr_servers++;
}
diff --git a/fs/afs/volume.c b/fs/afs/volume.c
index aebdeeed450a..fd5478e792da 100644
--- a/fs/afs/volume.c
+++ b/fs/afs/volume.c
@@ -84,11 +84,15 @@ static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
volume->type = params->type;
volume->type_force = params->force;
volume->name_len = vldb->name_len;
+ volume->creation_time = TIME64_MIN;
+ volume->update_time = TIME64_MIN;
refcount_set(&volume->ref, 1);
INIT_HLIST_NODE(&volume->proc_link);
INIT_WORK(&volume->destructor, afs_destroy_volume);
rwlock_init(&volume->servers_lock);
+ mutex_init(&volume->volsync_lock);
+ mutex_init(&volume->cb_check_lock);
rwlock_init(&volume->cb_v_break_lock);
INIT_LIST_HEAD(&volume->open_mmaps);
init_rwsem(&volume->open_mmaps_lock);
@@ -388,7 +392,11 @@ static int afs_update_volume_status(struct afs_volume *volume, struct key *key)
discard = old;
}
- volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
+ /* Check more often if replication is ongoing. */
+ if (new->ro_replicating)
+ volume->update_at = ktime_get_real_seconds() + 2;
+ else
+ volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
write_unlock(&volume->servers_lock);
if (discard == old)
diff --git a/fs/afs/yfsclient.c b/fs/afs/yfsclient.c
index 7c013070ef51..2d6943f05ea5 100644
--- a/fs/afs/yfsclient.c
+++ b/fs/afs/yfsclient.c
@@ -254,7 +254,6 @@ static void xdr_decode_YFSVolSync(const __be32 **_bp,
update = xdr_to_u64(x->vol_update_date);
do_div(update, 10 * 1000 * 1000);
volsync->update = update;
- volsync->mask |= AFS_VOLSYNC_CREATION | AFS_VOLSYNC_UPDATE;
}
*_bp += xdr_size(x);
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index ac50fa687429..b2e0847eca47 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -442,7 +442,8 @@ enum yfs_cm_operation {
EM(afs_cb_break_for_unlink, "break-unlink") \
EM(afs_cb_break_for_update_regress, "update-regress") \
EM(afs_cb_break_for_volume_callback, "break-v-cb") \
- E_(afs_cb_break_for_vos_release, "break-vos-release")
+ EM(afs_cb_break_for_vos_release, "break-vos-release") \
+ E_(afs_cb_break_volume_excluded, "vol-excluded")
#define afs_rotate_traces \
EM(afs_rotate_trace_aborted, "Abortd") \
^ permalink raw reply related [flat|nested] 57+ messages in thread
* [PATCH 40/41] afs: Fix fileserver rotation
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (38 preceding siblings ...)
2023-11-09 15:40 ` [PATCH 39/41] afs: Overhaul invalidation handling to better support RO volumes David Howells
@ 2023-11-09 15:40 ` David Howells
2023-11-09 15:40 ` [PATCH 41/41] afs: Fix offline and busy handling David Howells
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:40 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
Fix the fileserver rotation so that it doesn't use RTT as the basis for
deciding which server and address to use as this doesn't necessarily give a
good indication of the best path. Instead, use the configurable preference
list in conjunction with whatever probes have succeeded at the time of
looking.
To this end, make the following changes:
(1) Keep an array of "server states" to track what addresses we've tried
on each server and move the waitqueue entries there that we'll need
for probing.
(2) Each afs_server_state struct is made to pin the corresponding server's
endpoint state rather than the afs_operation struct carrying a pin on
the server we're currently looking at.
(3) Drop the server list preference; we now always rescan the server list.
(4) afs_wait_for_probes() now uses the server state list to guide it in
what it waits for (and to provide the waitqueue entries) and returns
an indication of whether we'd got a response, run out of responsive
addresses or the endpoint state had been superseded and we need to
restart the iteration.
(5) We call afs_get_address_preferences*() occasionally to refresh the
preference values.
(6) When picking a server, scan the addresses of the servers for which we
have as-yet untested communications, looking for the highest priority
one and use that instead of trying all the addresses for a particular
server in ascending-RTT order.
Also add a tracepoint to track progress through the rotation algorithm.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/fs_operation.c | 8 +-
fs/afs/fs_probe.c | 103 +++++++++--------------
fs/afs/internal.h | 23 ++++-
fs/afs/rotate.c | 168 +++++++++++++++++++++++++++----------
fs/afs/server_list.c | 14 +---
include/trace/events/afs.h | 80 +++++++++++++++---
6 files changed, 256 insertions(+), 140 deletions(-)
diff --git a/fs/afs/fs_operation.c b/fs/afs/fs_operation.c
index c9e85fa94d58..4c2ac403751d 100644
--- a/fs/afs/fs_operation.c
+++ b/fs/afs/fs_operation.c
@@ -230,7 +230,6 @@ void afs_wait_for_operation(struct afs_operation *op)
*/
int afs_put_operation(struct afs_operation *op)
{
- struct afs_endpoint_state *estate = op->estate;
struct afs_addr_list *alist;
int i, ret = afs_op_error(op);
@@ -254,18 +253,17 @@ int afs_put_operation(struct afs_operation *op)
kfree(op->more_files);
}
- if (estate) {
- alist = estate->addresses;
+ if (op->estate) {
+ alist = op->estate->addresses;
if (alist) {
if (op->call_responded &&
op->addr_index != alist->preferred &&
test_bit(alist->preferred, &op->addr_tried))
WRITE_ONCE(alist->preferred, op->addr_index);
}
- afs_put_endpoint_state(estate, afs_estate_trace_put_operation);
- op->estate = NULL;
}
+ afs_clear_server_states(op);
afs_put_serverlist(op->net, op->server_list);
afs_put_volume(op->volume, afs_volume_trace_put_put_op);
key_put(op->key);
diff --git a/fs/afs/fs_probe.c b/fs/afs/fs_probe.c
index c00d38b98a67..580de4adaaf6 100644
--- a/fs/afs/fs_probe.c
+++ b/fs/afs/fs_probe.c
@@ -296,58 +296,48 @@ void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
}
/*
- * Wait for the first as-yet untried fileserver to respond.
+ * Wait for the first as-yet untried fileserver to respond, for the probe state
+ * to be superseded or for all probes to finish.
*/
-int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
+int afs_wait_for_fs_probes(struct afs_operation *op, struct afs_server_state *states, bool intr)
{
struct afs_endpoint_state *estate;
- struct wait_queue_entry *waits;
- struct afs_server *server;
- unsigned int rtt = UINT_MAX, rtt_s;
- bool have_responders = false;
- int pref = -1, i;
+ struct afs_server_list *slist = op->server_list;
+ bool still_probing = true;
+ int ret = 0, i;
- _enter("%u,%lx", slist->nr_servers, untried);
+ _enter("%u", slist->nr_servers);
- /* Only wait for servers that have a probe outstanding. */
- rcu_read_lock();
for (i = 0; i < slist->nr_servers; i++) {
- if (test_bit(i, &untried)) {
- server = slist->servers[i].server;
- estate = rcu_dereference(server->endpoint_state);
- if (!atomic_read(&estate->nr_probing))
- __clear_bit(i, &untried);
- if (test_bit(AFS_ESTATE_RESPONDED, &estate->flags))
- have_responders = true;
- }
+ estate = states[i].endpoint_state;
+ if (test_bit(AFS_ESTATE_SUPERSEDED, &estate->flags))
+ return 2;
+ if (atomic_read(&estate->nr_probing))
+ still_probing = true;
+ if (estate->responsive_set & states[i].untried_addrs)
+ return 1;
}
- rcu_read_unlock();
- if (have_responders || !untried)
+ if (!still_probing)
return 0;
- waits = kmalloc(array_size(slist->nr_servers, sizeof(*waits)), GFP_KERNEL);
- if (!waits)
- return -ENOMEM;
-
- for (i = 0; i < slist->nr_servers; i++) {
- if (test_bit(i, &untried)) {
- server = slist->servers[i].server;
- init_waitqueue_entry(&waits[i], current);
- add_wait_queue(&server->probe_wq, &waits[i]);
- }
- }
+ for (i = 0; i < slist->nr_servers; i++)
+ add_wait_queue(&slist->servers[i].server->probe_wq, &states[i].probe_waiter);
for (;;) {
- bool still_probing = false;
+ still_probing = false;
- set_current_state(TASK_INTERRUPTIBLE);
+ set_current_state(intr ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
for (i = 0; i < slist->nr_servers; i++) {
- if (test_bit(i, &untried)) {
- server = slist->servers[i].server;
- if (test_bit(AFS_ESTATE_RESPONDED, &estate->flags))
- goto stop;
- if (atomic_read(&estate->nr_probing))
- still_probing = true;
+ estate = states[i].endpoint_state;
+ if (test_bit(AFS_ESTATE_SUPERSEDED, &estate->flags)) {
+ ret = 2;
+ goto stop;
+ }
+ if (atomic_read(&estate->nr_probing))
+ still_probing = true;
+ if (estate->responsive_set & states[i].untried_addrs) {
+ ret = 1;
+ goto stop;
}
}
@@ -359,28 +349,12 @@ int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
stop:
set_current_state(TASK_RUNNING);
- for (i = 0; i < slist->nr_servers; i++) {
- if (test_bit(i, &untried)) {
- server = slist->servers[i].server;
- rtt_s = READ_ONCE(server->rtt);
- if (test_bit(AFS_SERVER_FL_RESPONDING, &server->flags) &&
- rtt_s < rtt) {
- pref = i;
- rtt = rtt_s;
- }
-
- remove_wait_queue(&server->probe_wq, &waits[i]);
- }
- }
-
- kfree(waits);
-
- if (pref == -1 && signal_pending(current))
- return -ERESTARTSYS;
+ for (i = 0; i < slist->nr_servers; i++)
+ remove_wait_queue(&slist->servers[i].server->probe_wq, &states[i].probe_waiter);
- if (pref >= 0)
- slist->preferred = pref;
- return 0;
+ if (!ret && signal_pending(current))
+ ret = -ERESTARTSYS;
+ return ret;
}
/*
@@ -508,7 +482,7 @@ void afs_fs_probe_dispatcher(struct work_struct *work)
* Wait for a probe on a particular fileserver to complete for 2s.
*/
int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_state *estate,
- bool is_intr)
+ unsigned long exclude, bool is_intr)
{
struct wait_queue_entry wait;
unsigned long timo = 2 * HZ;
@@ -521,7 +495,8 @@ int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_sta
prepare_to_wait_event(&server->probe_wq, &wait,
is_intr ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE);
if (timo == 0 ||
- test_bit(AFS_ESTATE_RESPONDED, &estate->flags) ||
+ test_bit(AFS_ESTATE_SUPERSEDED, &estate->flags) ||
+ (estate->responsive_set & ~exclude) ||
atomic_read(&estate->nr_probing) == 0 ||
(is_intr && signal_pending(current)))
break;
@@ -531,7 +506,9 @@ int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_sta
finish_wait(&server->probe_wq, &wait);
dont_wait:
- if (test_bit(AFS_ESTATE_RESPONDED, &estate->flags))
+ if (estate->responsive_set & ~exclude)
+ return 1;
+ if (test_bit(AFS_ESTATE_SUPERSEDED, &estate->flags))
return 0;
if (is_intr && signal_pending(current))
return -ERESTARTSYS;
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index d9ec8cdca169..d1031241d11b 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -610,7 +610,6 @@ struct afs_server_list {
refcount_t usage;
bool attached; /* T if attached to servers */
unsigned char nr_servers;
- unsigned char preferred; /* Preferred server */
unsigned short vnovol_mask; /* Servers to be skipped due to VNOVOL */
unsigned int seq; /* Set to ->servers_seq when installed */
rwlock_t lock;
@@ -806,6 +805,20 @@ struct afs_vl_cursor {
bool call_responded; /* T if the current address responded */
};
+/*
+ * Fileserver state tracking for an operation. An array of these is kept,
+ * indexed by server index.
+ */
+struct afs_server_state {
+ /* Tracking of fileserver probe state. Other operations may interfere
+ * by probing a fileserver when accessing other volumes.
+ */
+ unsigned int probe_seq;
+ unsigned long untried_addrs; /* Addresses we haven't tried yet */
+ struct wait_queue_entry probe_waiter;
+ struct afs_endpoint_state *endpoint_state; /* Endpoint state being monitored */
+};
+
/*
* Fileserver operation methods.
*/
@@ -906,7 +919,8 @@ struct afs_operation {
/* Fileserver iteration state */
struct afs_server_list *server_list; /* Current server list (pins ref) */
struct afs_server *server; /* Server we're using (ref pinned by server_list) */
- struct afs_endpoint_state *estate; /* Current endpoint state (pins ref) */
+ struct afs_endpoint_state *estate; /* Current endpoint state (doesn't pin ref) */
+ struct afs_server_state *server_states; /* States of the servers involved */
struct afs_call *call;
unsigned long untried_servers; /* Bitmask of untried servers */
unsigned long addr_tried; /* Tried addresses */
@@ -1220,11 +1234,11 @@ void afs_put_endpoint_state(struct afs_endpoint_state *estate, enum afs_estate_t
extern void afs_fileserver_probe_result(struct afs_call *);
void afs_fs_probe_fileserver(struct afs_net *net, struct afs_server *server,
struct afs_addr_list *new_addrs, struct key *key);
-extern int afs_wait_for_fs_probes(struct afs_server_list *, unsigned long);
+int afs_wait_for_fs_probes(struct afs_operation *op, struct afs_server_state *states, bool intr);
extern void afs_probe_fileserver(struct afs_net *, struct afs_server *);
extern void afs_fs_probe_dispatcher(struct work_struct *);
int afs_wait_for_one_fs_probe(struct afs_server *server, struct afs_endpoint_state *estate,
- bool is_intr);
+ unsigned long exclude, bool is_intr);
extern void afs_fs_probe_cleanup(struct afs_net *);
/*
@@ -1348,6 +1362,7 @@ static inline void afs_put_sysnames(struct afs_sysnames *sysnames) {}
/*
* rotate.c
*/
+void afs_clear_server_states(struct afs_operation *op);
extern bool afs_select_fileserver(struct afs_operation *);
extern void afs_dump_edestaddrreq(const struct afs_operation *);
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index e0267547f305..3f656dcb0adf 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -15,6 +15,18 @@
#include "afs_fs.h"
#include "protocol_uae.h"
+void afs_clear_server_states(struct afs_operation *op)
+{
+ unsigned int i;
+
+ if (op->server_states) {
+ for (i = 0; i < op->server_list->nr_servers; i++)
+ afs_put_endpoint_state(op->server_states[i].endpoint_state,
+ afs_estate_trace_put_server_state);
+ kfree(op->server_states);
+ }
+}
+
/*
* Begin iteration through a server list, starting with the vnode's last used
* server if possible, or the last recorded good server if not.
@@ -26,14 +38,41 @@ static bool afs_start_fs_iteration(struct afs_operation *op,
void *cb_server;
int i;
+ trace_afs_rotate(op, afs_rotate_trace_start, 0);
+
read_lock(&op->volume->servers_lock);
op->server_list = afs_get_serverlist(
rcu_dereference_protected(op->volume->servers,
lockdep_is_held(&op->volume->servers_lock)));
read_unlock(&op->volume->servers_lock);
+ op->server_states = kcalloc(op->server_list->nr_servers, sizeof(op->server_states[0]),
+ GFP_KERNEL);
+ if (!op->server_states) {
+ afs_op_nomem(op);
+ trace_afs_rotate(op, afs_rotate_trace_nomem, 0);
+ return false;
+ }
+
+ rcu_read_lock();
+ for (i = 0; i < op->server_list->nr_servers; i++) {
+ struct afs_endpoint_state *estate;
+ struct afs_server_state *s = &op->server_states[i];
+
+ server = op->server_list->servers[i].server;
+ estate = rcu_dereference(server->endpoint_state);
+ s->endpoint_state = afs_get_endpoint_state(estate,
+ afs_estate_trace_get_server_state);
+ s->probe_seq = estate->probe_seq;
+ s->untried_addrs = (1UL << estate->addresses->nr_addrs) - 1;
+ init_waitqueue_entry(&s->probe_waiter, current);
+ afs_get_address_preferences(op->net, estate->addresses);
+ }
+ rcu_read_unlock();
+
+
op->untried_servers = (1UL << op->server_list->nr_servers) - 1;
- op->server_index = READ_ONCE(op->server_list->preferred);
+ op->server_index = -1;
cb_server = vnode->cb_server;
if (cb_server) {
@@ -52,6 +91,7 @@ static bool afs_start_fs_iteration(struct afs_operation *op,
*/
if (op->flags & AFS_OPERATION_CUR_ONLY) {
afs_op_set_error(op, -ESTALE);
+ trace_afs_rotate(op, afs_rotate_trace_stale_lock, 0);
return false;
}
@@ -109,14 +149,13 @@ static bool afs_sleep_and_retry(struct afs_operation *op)
*/
bool afs_select_fileserver(struct afs_operation *op)
{
- struct afs_endpoint_state *estate = op->estate;
struct afs_addr_list *alist;
struct afs_server *server;
struct afs_vnode *vnode = op->file[0].vnode;
unsigned long set, failed;
- unsigned int rtt;
s32 abort_code = op->call_abort_code;
- int error = op->call_error, addr_index, i;
+ int best_prio = 0;
+ int error = op->call_error, addr_index, i, j;
op->nr_iterations++;
@@ -127,6 +166,7 @@ bool afs_select_fileserver(struct afs_operation *op)
error, abort_code);
if (op->flags & AFS_OPERATION_STOP) {
+ trace_afs_rotate(op, afs_rotate_trace_stopped, 0);
_leave(" = f [stopped]");
return false;
}
@@ -134,7 +174,8 @@ bool afs_select_fileserver(struct afs_operation *op)
if (op->nr_iterations == 0)
goto start;
- WRITE_ONCE(estate->addresses->addrs[op->addr_index].last_error, error);
+ WRITE_ONCE(op->estate->addresses->addrs[op->addr_index].last_error, error);
+ trace_afs_rotate(op, afs_rotate_trace_iter, op->call_error);
/* Evaluate the result of the previous operation, if there was one. */
switch (op->call_error) {
@@ -145,6 +186,7 @@ bool afs_select_fileserver(struct afs_operation *op)
/* Success or local failure. Stop. */
afs_op_set_error(op, error);
op->flags |= AFS_OPERATION_STOP;
+ trace_afs_rotate(op, afs_rotate_trace_stop, error);
_leave(" = f [okay/local %d]", error);
return false;
@@ -157,6 +199,7 @@ bool afs_select_fileserver(struct afs_operation *op)
* errors instead. IBM AFS and OpenAFS fileservers, however, do leak
* these abort codes.
*/
+ trace_afs_rotate(op, afs_rotate_trace_aborted, abort_code);
op->cumul_error.responded = true;
switch (abort_code) {
case VNOVOL:
@@ -399,19 +442,22 @@ bool afs_select_fileserver(struct afs_operation *op)
}
restart_from_beginning:
+ trace_afs_rotate(op, afs_rotate_trace_restart, 0);
_debug("restart");
- afs_put_endpoint_state(estate, afs_estate_trace_put_restart_rotate);
- estate = op->estate = NULL;
+ op->estate = NULL;
op->server = NULL;
+ afs_clear_server_states(op);
+ op->server_states = NULL;
afs_put_serverlist(op->net, op->server_list);
op->server_list = NULL;
start:
_debug("start");
- ASSERTCMP(estate, ==, NULL);
+ ASSERTCMP(op->estate, ==, NULL);
/* See if we need to do an update of the volume record. Note that the
* volume may have moved or even have been deleted.
*/
error = afs_check_volume_status(op->volume, op);
+ trace_afs_rotate(op, afs_rotate_trace_check_vol_status, error);
if (error < 0) {
afs_op_set_error(op, error);
goto failed;
@@ -424,16 +470,29 @@ bool afs_select_fileserver(struct afs_operation *op)
pick_server:
_debug("pick [%lx]", op->untried_servers);
- ASSERTCMP(estate, ==, NULL);
+ ASSERTCMP(op->estate, ==, NULL);
- error = afs_wait_for_fs_probes(op->server_list, op->untried_servers);
- if (error < 0) {
+ error = afs_wait_for_fs_probes(op, op->server_states,
+ !(op->flags & AFS_OPERATION_UNINTR));
+ switch (error) {
+ case 0: /* No untried responsive servers and no outstanding probes */
+ trace_afs_rotate(op, afs_rotate_trace_probe_none, 0);
+ goto no_more_servers;
+ case 1: /* Got a response */
+ trace_afs_rotate(op, afs_rotate_trace_probe_response, 0);
+ break;
+ case 2: /* Probe data superseded */
+ trace_afs_rotate(op, afs_rotate_trace_probe_superseded, 0);
+ goto restart_from_beginning;
+ default:
+ trace_afs_rotate(op, afs_rotate_trace_probe_error, error);
afs_op_set_error(op, error);
goto failed;
}
- /* Pick the untried server with the lowest RTT. If we have outstanding
- * callbacks, we stick with the server we're already using if we can.
+ /* Pick the untried server with the highest priority untried endpoint.
+ * If we have outstanding callbacks, we stick with the server we're
+ * already using if we can.
*/
if (op->server) {
_debug("server %u", op->server_index);
@@ -443,32 +502,45 @@ bool afs_select_fileserver(struct afs_operation *op)
_debug("no server");
}
+ rcu_read_lock();
op->server_index = -1;
- rtt = UINT_MAX;
+ best_prio = -1;
for (i = 0; i < op->server_list->nr_servers; i++) {
+ struct afs_endpoint_state *es;
+ struct afs_addr_list *sal;
struct afs_server *s = op->server_list->servers[i].server;
if (!test_bit(i, &op->untried_servers) ||
!test_bit(AFS_SERVER_FL_RESPONDING, &s->flags))
continue;
- if (s->rtt <= rtt) {
- op->server_index = i;
- rtt = s->rtt;
+ es = op->server_states->endpoint_state;
+ sal = es->addresses;
+
+ afs_get_address_preferences_rcu(op->net, sal);
+ for (j = 0; j < sal->nr_addrs; j++) {
+ if (!sal->addrs[j].peer)
+ continue;
+ if (sal->addrs[j].prio > best_prio) {
+ op->server_index = i;
+ best_prio = sal->addrs[j].prio;
+ }
}
}
+ rcu_read_unlock();
if (op->server_index == -1)
goto no_more_servers;
selected_server:
- _debug("use %d", op->server_index);
+ trace_afs_rotate(op, afs_rotate_trace_selected_server, best_prio);
+ _debug("use %d prio %u", op->server_index, best_prio);
__clear_bit(op->server_index, &op->untried_servers);
/* We're starting on a different fileserver from the list. We need to
* check it, create a callback intercept, find its address list and
* probe its capabilities before we use it.
*/
- ASSERTCMP(estate, ==, NULL);
+ ASSERTCMP(op->estate, ==, NULL);
server = op->server_list->servers[op->server_index].server;
if (!afs_check_server_record(op, server, op->key))
@@ -484,12 +556,6 @@ bool afs_select_fileserver(struct afs_operation *op)
vnode->cb_expires_at = AFS_NO_CB_PROMISE;
}
- read_lock(&server->fs_lock);
- estate = rcu_dereference_protected(server->endpoint_state,
- lockdep_is_held(&server->fs_lock));
- op->estate = afs_get_endpoint_state(estate, afs_estate_trace_get_fsrotate_set);
- read_unlock(&server->fs_lock);
-
retry_server:
op->addr_tried = 0;
op->addr_index = -1;
@@ -498,14 +564,23 @@ bool afs_select_fileserver(struct afs_operation *op)
/* Iterate over the current server's address list to try and find an
* address on which it will respond to us.
*/
- set = READ_ONCE(estate->responsive_set);
- failed = READ_ONCE(estate->failed_set);
- _debug("iterate ES=%x rs=%lx fs=%lx", estate->probe_seq, set, failed);
+ op->estate = op->server_states[op->server_index].endpoint_state;
+ set = READ_ONCE(op->estate->responsive_set);
+ failed = READ_ONCE(op->estate->failed_set);
+ _debug("iterate ES=%x rs=%lx fs=%lx", op->estate->probe_seq, set, failed);
set &= ~(failed | op->addr_tried);
+ trace_afs_rotate(op, afs_rotate_trace_iterate_addr, set);
if (!set)
- goto out_of_addresses;
+ goto wait_for_more_probe_results;
+
+ alist = op->estate->addresses;
+ for (i = 0; i < alist->nr_addrs; i++) {
+ if (alist->addrs[i].prio > best_prio) {
+ addr_index = i;
+ best_prio = alist->addrs[i].prio;
+ }
+ }
- alist = estate->addresses;
addr_index = READ_ONCE(alist->preferred);
if (!test_bit(addr_index, &set))
addr_index = __ffs(set);
@@ -520,17 +595,24 @@ bool afs_select_fileserver(struct afs_operation *op)
_leave(" = t");
return true;
-out_of_addresses:
+wait_for_more_probe_results:
+ error = afs_wait_for_one_fs_probe(op->server, op->estate, op->addr_tried,
+ !(op->flags & AFS_OPERATION_UNINTR));
+ if (!error)
+ goto iterate_address;
+
/* We've now had a failure to respond on all of a server's addresses -
* immediately probe them again and consider retrying the server.
*/
+ trace_afs_rotate(op, afs_rotate_trace_probe_fileserver, 0);
afs_probe_fileserver(op->net, op->server);
if (op->flags & AFS_OPERATION_RETRY_SERVER) {
- error = afs_wait_for_one_fs_probe(op->server, estate,
+ error = afs_wait_for_one_fs_probe(op->server, op->estate, op->addr_tried,
!(op->flags & AFS_OPERATION_UNINTR));
switch (error) {
case 0:
op->flags &= ~AFS_OPERATION_RETRY_SERVER;
+ trace_afs_rotate(op, afs_rotate_trace_retry_server, 0);
goto retry_server;
case -ERESTARTSYS:
afs_op_set_error(op, error);
@@ -542,30 +624,30 @@ bool afs_select_fileserver(struct afs_operation *op)
}
next_server:
+ trace_afs_rotate(op, afs_rotate_trace_next_server, 0);
_debug("next");
- ASSERT(estate);
- alist = estate->addresses;
+ ASSERT(op->estate);
+ alist = op->estate->addresses;
if (op->call_responded &&
op->addr_index != READ_ONCE(alist->preferred) &&
test_bit(alist->preferred, &op->addr_tried))
WRITE_ONCE(alist->preferred, op->addr_index);
- afs_put_endpoint_state(estate, afs_estate_trace_put_next_server);
- estate = op->estate = NULL;
+ op->estate = NULL;
goto pick_server;
no_more_servers:
/* That's all the servers poked to no good effect. Try again if some
* of them were busy.
*/
+ trace_afs_rotate(op, afs_rotate_trace_no_more_servers, 0);
if (op->flags & AFS_OPERATION_VBUSY)
goto restart_from_beginning;
rcu_read_lock();
for (i = 0; i < op->server_list->nr_servers; i++) {
struct afs_endpoint_state *estate;
- struct afs_server *s = op->server_list->servers[i].server;
- estate = rcu_dereference(s->endpoint_state);
+ estate = op->server_states->endpoint_state;
error = READ_ONCE(estate->error);
if (error < 0)
afs_op_accumulate_error(op, error, estate->abort_code);
@@ -573,14 +655,14 @@ bool afs_select_fileserver(struct afs_operation *op)
rcu_read_unlock();
failed:
+ trace_afs_rotate(op, afs_rotate_trace_failed, 0);
op->flags |= AFS_OPERATION_STOP;
- if (estate) {
- alist = estate->addresses;
+ if (op->estate) {
+ alist = op->estate->addresses;
if (op->call_responded &&
op->addr_index != READ_ONCE(alist->preferred) &&
test_bit(alist->preferred, &op->addr_tried))
WRITE_ONCE(alist->preferred, op->addr_index);
- afs_put_endpoint_state(estate, afs_estate_trace_put_op_failed);
op->estate = NULL;
}
_leave(" = f [failed %d]", afs_op_error(op));
@@ -613,8 +695,8 @@ void afs_dump_edestaddrreq(const struct afs_operation *op)
if (op->server_list) {
const struct afs_server_list *sl = op->server_list;
- pr_notice("FC: SL nr=%u pr=%u vnov=%hx\n",
- sl->nr_servers, sl->preferred, sl->vnovol_mask);
+ pr_notice("FC: SL nr=%u vnov=%hx\n",
+ sl->nr_servers, sl->vnovol_mask);
for (i = 0; i < sl->nr_servers; i++) {
const struct afs_server *s = sl->servers[i].server;
const struct afs_endpoint_state *e =
diff --git a/fs/afs/server_list.c b/fs/afs/server_list.c
index c2c2b429ade2..c9f146f69adf 100644
--- a/fs/afs/server_list.c
+++ b/fs/afs/server_list.c
@@ -102,8 +102,7 @@ struct afs_server_list *afs_alloc_server_list(struct afs_volume *volume,
bool afs_annotate_server_list(struct afs_server_list *new,
struct afs_server_list *old)
{
- struct afs_server *cur;
- int i, j;
+ int i;
if (old->nr_servers != new->nr_servers)
goto changed;
@@ -111,19 +110,8 @@ bool afs_annotate_server_list(struct afs_server_list *new,
for (i = 0; i < old->nr_servers; i++)
if (old->servers[i].server != new->servers[i].server)
goto changed;
-
return false;
-
changed:
- /* Maintain the same preferred server as before if possible. */
- cur = old->servers[old->preferred].server;
- for (j = 0; j < new->nr_servers; j++) {
- if (new->servers[j].server == cur) {
- new->preferred = j;
- break;
- }
- }
-
return true;
}
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 3905fd9c957f..804d9e147314 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -230,15 +230,12 @@ enum yfs_cm_operation {
#define afs_estate_traces \
EM(afs_estate_trace_alloc_probe, "ALLOC prob") \
EM(afs_estate_trace_alloc_server, "ALLOC srvr") \
- EM(afs_estate_trace_get_fsrotate_set, "GET fs-rot") \
+ EM(afs_estate_trace_get_server_state, "GET srv-st") \
EM(afs_estate_trace_get_getcaps, "GET getcap") \
EM(afs_estate_trace_put_getcaps, "PUT getcap") \
- EM(afs_estate_trace_put_next_server, "PUT nx-srv") \
- EM(afs_estate_trace_put_op_failed, "PUT op-fai") \
- EM(afs_estate_trace_put_operation, "PUT op ") \
EM(afs_estate_trace_put_probe, "PUT probe ") \
- EM(afs_estate_trace_put_restart_rotate, "PUT rstrot") \
EM(afs_estate_trace_put_server, "PUT server") \
+ EM(afs_estate_trace_put_server_state, "PUT srv-st") \
E_(afs_estate_trace_free, "FREE ")
#define afs_fs_operations \
@@ -447,6 +444,28 @@ enum yfs_cm_operation {
EM(afs_cb_break_for_volume_callback, "break-v-cb") \
E_(afs_cb_break_for_vos_release, "break-vos-release")
+#define afs_rotate_traces \
+ EM(afs_rotate_trace_aborted, "Abortd") \
+ EM(afs_rotate_trace_check_vol_status, "VolStt") \
+ EM(afs_rotate_trace_failed, "Failed") \
+ EM(afs_rotate_trace_iter, "Iter ") \
+ EM(afs_rotate_trace_iterate_addr, "ItAddr") \
+ EM(afs_rotate_trace_next_server, "NextSv") \
+ EM(afs_rotate_trace_no_more_servers, "NoMore") \
+ EM(afs_rotate_trace_nomem, "Nomem ") \
+ EM(afs_rotate_trace_probe_error, "PrbErr") \
+ EM(afs_rotate_trace_probe_fileserver, "PrbFsv") \
+ EM(afs_rotate_trace_probe_none, "PrbNon") \
+ EM(afs_rotate_trace_probe_response, "PrbRsp") \
+ EM(afs_rotate_trace_probe_superseded, "PrbSup") \
+ EM(afs_rotate_trace_restart, "Rstart") \
+ EM(afs_rotate_trace_retry_server, "RtrySv") \
+ EM(afs_rotate_trace_selected_server, "SlctSv") \
+ EM(afs_rotate_trace_stale_lock, "StlLck") \
+ EM(afs_rotate_trace_start, "Start ") \
+ EM(afs_rotate_trace_stop, "Stop ") \
+ E_(afs_rotate_trace_stopped, "Stoppd")
+
/*
* Generate enums for tracing information.
*/
@@ -470,6 +489,7 @@ enum afs_file_error { afs_file_errors } __mode(byte);
enum afs_flock_event { afs_flock_events } __mode(byte);
enum afs_flock_operation { afs_flock_operations } __mode(byte);
enum afs_io_error { afs_io_errors } __mode(byte);
+enum afs_rotate_trace { afs_rotate_traces } __mode(byte);
enum afs_server_trace { afs_server_traces } __mode(byte);
enum afs_volume_trace { afs_volume_traces } __mode(byte);
@@ -485,21 +505,22 @@ enum afs_volume_trace { afs_volume_traces } __mode(byte);
afs_alist_traces;
afs_call_traces;
-afs_server_traces;
+afs_cb_break_reasons;
afs_cell_traces;
-afs_fs_operations;
-afs_vl_operations;
afs_cm_operations;
-yfs_cm_operations;
afs_edit_dir_ops;
afs_edit_dir_reasons;
afs_eproto_causes;
afs_estate_traces;
-afs_io_errors;
afs_file_errors;
-afs_flock_types;
afs_flock_operations;
-afs_cb_break_reasons;
+afs_flock_types;
+afs_fs_operations;
+afs_io_errors;
+afs_rotate_traces;
+afs_server_traces;
+afs_vl_operations;
+yfs_cm_operations;
/*
* Now redefine the EM() and E_() macros to map the enums to the strings that
@@ -1518,6 +1539,41 @@ TRACE_EVENT(afs_vl_probe,
&__entry->srx.transport)
);
+TRACE_EVENT(afs_rotate,
+ TP_PROTO(struct afs_operation *op, enum afs_rotate_trace reason, unsigned int extra),
+
+ TP_ARGS(op, reason, extra),
+
+ TP_STRUCT__entry(
+ __field(unsigned int, op)
+ __field(unsigned int, flags)
+ __field(unsigned int, extra)
+ __field(unsigned short, iteration)
+ __field(short, server_index)
+ __field(short, addr_index)
+ __field(enum afs_rotate_trace, reason)
+ ),
+
+ TP_fast_assign(
+ __entry->op = op->debug_id;
+ __entry->flags = op->flags;
+ __entry->iteration = op->nr_iterations;
+ __entry->server_index = op->server_index;
+ __entry->addr_index = op->addr_index;
+ __entry->reason = reason;
+ __entry->extra = extra;
+ ),
+
+ TP_printk("OP=%08x it=%02x %s fl=%x sx=%d ax=%d ext=%d",
+ __entry->op,
+ __entry->iteration,
+ __print_symbolic(__entry->reason, afs_rotate_traces),
+ __entry->flags,
+ __entry->server_index,
+ __entry->addr_index,
+ __entry->extra)
+ );
+
#endif /* _TRACE_AFS_H */
/* This part must be outside protection */
^ permalink raw reply related [flat|nested] 57+ messages in thread* [PATCH 41/41] afs: Fix offline and busy handling
2023-11-09 15:39 [PATCH 00/41] afs: Fix probe handling, server rotation and RO volume callback handling David Howells
` (39 preceding siblings ...)
2023-11-09 15:40 ` [PATCH 40/41] afs: Fix fileserver rotation David Howells
@ 2023-11-09 15:40 ` David Howells
40 siblings, 0 replies; 57+ messages in thread
From: David Howells @ 2023-11-09 15:40 UTC (permalink / raw)
To: Marc Dionne; +Cc: David Howells, linux-afs, linux-fsdevel, linux-kernel
The current code assumes offline and busy volume states apply to all
instances of a volume, not just the one on the server that returned
VOFFLINE or VBUSY.
Fix that by moving the flags recording this to the afs_server_entry struct
that is used to represent a particular instance of a volume on a specific
server.
Further, add a sleep for when we have iterated through all the servers so
that we don't keep poking the server every few milliseconds.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
---
fs/afs/internal.h | 7 ++++---
fs/afs/rotate.c | 40 ++++++++++++++++++++++++--------------
include/trace/events/afs.h | 1 +
3 files changed, 30 insertions(+), 18 deletions(-)
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index d1031241d11b..eb59b0487f8b 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -603,6 +603,9 @@ struct afs_server_entry {
struct afs_volume *volume;
struct list_head slink; /* Link in server->volumes */
time64_t cb_expires_at; /* Time at which volume-level callback expires */
+ unsigned long flags;
+#define AFS_SE_VOLUME_OFFLINE 0 /* Set if volume offline notice given */
+#define AFS_SE_VOLUME_BUSY 1 /* Set if volume busy notice given */
};
struct afs_server_list {
@@ -636,9 +639,7 @@ struct afs_volume {
#define AFS_VOLUME_UPDATING 1 /* - T if an update is in progress */
#define AFS_VOLUME_WAIT 2 /* - T if users must wait for update */
#define AFS_VOLUME_DELETED 3 /* - T if volume appears deleted */
-#define AFS_VOLUME_OFFLINE 4 /* - T if volume offline notice given */
-#define AFS_VOLUME_BUSY 5 /* - T if volume busy notice given */
-#define AFS_VOLUME_MAYBE_NO_IBULK 6 /* - T if some servers don't have InlineBulkStatus */
+#define AFS_VOLUME_MAYBE_NO_IBULK 4 /* - T if some servers don't have InlineBulkStatus */
#ifdef CONFIG_AFS_FSCACHE
struct fscache_volume *cache; /* Caching cookie */
#endif
diff --git a/fs/afs/rotate.c b/fs/afs/rotate.c
index 3f656dcb0adf..cb0ab1c2c401 100644
--- a/fs/afs/rotate.c
+++ b/fs/afs/rotate.c
@@ -111,7 +111,7 @@ static bool afs_start_fs_iteration(struct afs_operation *op,
/*
* Post volume busy note.
*/
-static void afs_busy(struct afs_volume *volume, u32 abort_code)
+static void afs_busy(struct afs_operation *op, u32 abort_code)
{
const char *m;
@@ -122,7 +122,8 @@ static void afs_busy(struct afs_volume *volume, u32 abort_code)
default: m = "busy"; break;
}
- pr_notice("kAFS: Volume %llu '%s' is %s\n", volume->vid, volume->name, m);
+ pr_notice("kAFS: Volume %llu '%s' on server %pU is %s\n",
+ op->volume->vid, op->volume->name, &op->server->uuid, m);
}
/*
@@ -130,6 +131,7 @@ static void afs_busy(struct afs_volume *volume, u32 abort_code)
*/
static bool afs_sleep_and_retry(struct afs_operation *op)
{
+ trace_afs_rotate(op, afs_rotate_trace_busy_sleep, 0);
if (!(op->flags & AFS_OPERATION_UNINTR)) {
msleep_interruptible(1000);
if (signal_pending(current)) {
@@ -180,6 +182,10 @@ bool afs_select_fileserver(struct afs_operation *op)
/* Evaluate the result of the previous operation, if there was one. */
switch (op->call_error) {
case 0:
+ clear_bit(AFS_SE_VOLUME_OFFLINE,
+ &op->server_list->servers[op->server_index].flags);
+ clear_bit(AFS_SE_VOLUME_BUSY,
+ &op->server_list->servers[op->server_index].flags);
op->cumul_error.responded = true;
fallthrough;
default:
@@ -297,18 +303,16 @@ bool afs_select_fileserver(struct afs_operation *op)
* expected to come back but it might take a long time (could be
* days).
*/
- if (!test_and_set_bit(AFS_VOLUME_OFFLINE, &op->volume->flags)) {
- afs_busy(op->volume, abort_code);
- clear_bit(AFS_VOLUME_BUSY, &op->volume->flags);
+ if (!test_and_set_bit(AFS_SE_VOLUME_OFFLINE,
+ &op->server_list->servers[op->server_index].flags)) {
+ afs_busy(op, abort_code);
+ clear_bit(AFS_SE_VOLUME_BUSY,
+ &op->server_list->servers[op->server_index].flags);
}
if (op->flags & AFS_OPERATION_NO_VSLEEP) {
afs_op_set_error(op, -EADV);
goto failed;
}
- if (op->flags & AFS_OPERATION_CUR_ONLY) {
- afs_op_set_error(op, -ESTALE);
- goto failed;
- }
goto busy;
case VRESTARTING: /* The fileserver is either shutting down or starting up. */
@@ -329,9 +333,11 @@ bool afs_select_fileserver(struct afs_operation *op)
afs_op_set_error(op, -EBUSY);
goto failed;
}
- if (!test_and_set_bit(AFS_VOLUME_BUSY, &op->volume->flags)) {
- afs_busy(op->volume, abort_code);
- clear_bit(AFS_VOLUME_OFFLINE, &op->volume->flags);
+ if (!test_and_set_bit(AFS_SE_VOLUME_BUSY,
+ &op->server_list->servers[op->server_index].flags)) {
+ afs_busy(op, abort_code);
+ clear_bit(AFS_SE_VOLUME_OFFLINE,
+ &op->server_list->servers[op->server_index].flags);
}
busy:
if (op->flags & AFS_OPERATION_CUR_ONLY) {
@@ -411,8 +417,10 @@ bool afs_select_fileserver(struct afs_operation *op)
default:
afs_op_accumulate_error(op, error, abort_code);
failed_but_online:
- clear_bit(AFS_VOLUME_OFFLINE, &op->volume->flags);
- clear_bit(AFS_VOLUME_BUSY, &op->volume->flags);
+ clear_bit(AFS_SE_VOLUME_OFFLINE,
+ &op->server_list->servers[op->server_index].flags);
+ clear_bit(AFS_SE_VOLUME_BUSY,
+ &op->server_list->servers[op->server_index].flags);
goto failed;
}
@@ -640,8 +648,10 @@ bool afs_select_fileserver(struct afs_operation *op)
* of them were busy.
*/
trace_afs_rotate(op, afs_rotate_trace_no_more_servers, 0);
- if (op->flags & AFS_OPERATION_VBUSY)
+ if (op->flags & AFS_OPERATION_VBUSY) {
+ afs_sleep_and_retry(op);
goto restart_from_beginning;
+ }
rcu_read_lock();
for (i = 0; i < op->server_list->nr_servers; i++) {
diff --git a/include/trace/events/afs.h b/include/trace/events/afs.h
index 804d9e147314..ac50fa687429 100644
--- a/include/trace/events/afs.h
+++ b/include/trace/events/afs.h
@@ -446,6 +446,7 @@ enum yfs_cm_operation {
#define afs_rotate_traces \
EM(afs_rotate_trace_aborted, "Abortd") \
+ EM(afs_rotate_trace_busy_sleep, "BsySlp") \
EM(afs_rotate_trace_check_vol_status, "VolStt") \
EM(afs_rotate_trace_failed, "Failed") \
EM(afs_rotate_trace_iter, "Iter ") \
^ permalink raw reply related [flat|nested] 57+ messages in thread