* Re: [PATCH v2 06/16] thunderbolt: Add support for XDomain discovery protocol
From: David Miller @ 2017-09-27 4:35 UTC (permalink / raw)
To: mika.westerberg
Cc: gregkh, andreas.noever, michael.jamet, yehezkel.bernat,
amir.jer.levy, Mario.Limonciello, lukas, andriy.shevchenko,
andrew, linux-kernel, netdev
In-Reply-To: <20170925110738.68382-7-mika.westerberg@linux.intel.com>
From: Mika Westerberg <mika.westerberg@linux.intel.com>
Date: Mon, 25 Sep 2017 14:07:28 +0300
> +struct icm_fr_event_xdomain_connected {
> + struct icm_pkg_header hdr;
> + u16 reserved;
> + u16 link_info;
> + uuid_t remote_uuid;
> + uuid_t local_uuid;
> + u32 local_route_hi;
> + u32 local_route_lo;
> + u32 remote_route_hi;
> + u32 remote_route_lo;
> +} __packed;
> +
> +struct icm_fr_event_xdomain_disconnected {
> + struct icm_pkg_header hdr;
> + u16 reserved;
> + u16 link_info;
> + uuid_t remote_uuid;
> +} __packed;
> +
> struct icm_fr_pkg_add_device_key {
> struct icm_pkg_header hdr;
> uuid_t ep_uuid;
Again, __packed should be avoided unless absolutely necessary.
Thank you.
^ permalink raw reply
* Re: [PATCH v2 02/16] thunderbolt: Add support for XDomain properties
From: David Miller @ 2017-09-27 4:33 UTC (permalink / raw)
To: mika.westerberg
Cc: gregkh, andreas.noever, michael.jamet, yehezkel.bernat,
amir.jer.levy, Mario.Limonciello, lukas, andriy.shevchenko,
andrew, linux-kernel, netdev
In-Reply-To: <20170925110738.68382-3-mika.westerberg@linux.intel.com>
From: Mika Westerberg <mika.westerberg@linux.intel.com>
Date: Mon, 25 Sep 2017 14:07:24 +0300
> +struct tb_property_entry {
> + u32 key_hi;
> + u32 key_lo;
> + u16 length;
> + u8 reserved;
> + u8 type;
> + u32 value;
> +} __packed;
> +
> +struct tb_property_rootdir_entry {
> + u32 magic;
> + u32 length;
> + struct tb_property_entry entries[];
> +} __packed;
> +
> +struct tb_property_dir_entry {
> + u32 uuid[4];
> + struct tb_property_entry entries[];
> +} __packed;
There is no apparent need for __packed here, and __packed should be
avoided unless absolutely necessary as it pessimizes the code
significantly on some architectures.
Please remove __packed from these datastructures unless you can
prove it is absolutely needed and, in such case, please document
in a comment why that requirement exists. Because from the layout
of these types, everything will be packed in just fine without
__packed.
Thank you.
^ permalink raw reply
* [PATCH] net-ipv6: add support for sockopt(SOL_IPV6, IPV6_FREEBIND)
From: Maciej Żenczykowski @ 2017-09-27 4:32 UTC (permalink / raw)
To: Maciej Żenczykowski, David S . Miller; +Cc: netdev
From: Maciej Żenczykowski <maze@google.com>
So far we've been relying on sockopt(SOL_IP, IP_FREEBIND) being usable
even on IPv6 sockets.
However, it turns out it is perfectly reasonable to want to set freebind
on an AF_INET6 SOCK_RAW socket - but there is no way to set any SOL_IP
socket option on such a socket (they're all blindly errored out).
One use case for this is to allow spoofing src ip on a raw socket
via sendmsg cmsg.
Tested:
built, and booted
# python
>>> import socket
>>> SOL_IP = socket.SOL_IP
>>> SOL_IPV6 = socket.IPPROTO_IPV6
>>> IP_FREEBIND = 15
>>> IPV6_FREEBIND = 78
>>> s = socket.socket(socket.AF_INET6, socket.SOCK_DGRAM, 0)
>>> s.getsockopt(SOL_IP, IP_FREEBIND)
0
>>> s.getsockopt(SOL_IPV6, IPV6_FREEBIND)
0
>>> s.setsockopt(SOL_IPV6, IPV6_FREEBIND, 1)
>>> s.getsockopt(SOL_IP, IP_FREEBIND)
1
>>> s.getsockopt(SOL_IPV6, IPV6_FREEBIND)
1
Signed-off-by: Maciej Żenczykowski <maze@google.com>
---
include/uapi/linux/in6.h | 1 +
net/ipv6/ipv6_sockglue.c | 12 ++++++++++++
2 files changed, 13 insertions(+)
diff --git a/include/uapi/linux/in6.h b/include/uapi/linux/in6.h
index 46444f8fbee4..4f8f3eb0699f 100644
--- a/include/uapi/linux/in6.h
+++ b/include/uapi/linux/in6.h
@@ -284,6 +284,7 @@ struct in6_flowlabel_req {
#define IPV6_TRANSPARENT 75
#define IPV6_UNICAST_IF 76
#define IPV6_RECVFRAGSIZE 77
+#define IPV6_FREEBIND 78
/*
* Multicast Routing:
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index a5e466d4e093..b9404feabd78 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c
@@ -377,6 +377,14 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
retv = 0;
break;
+ case IPV6_FREEBIND:
+ if (optlen < sizeof(int))
+ goto e_inval;
+ /* we also don't have a separate freebind bit for IPV6 */
+ inet_sk(sk)->freebind = valbool;
+ retv = 0;
+ break;
+
case IPV6_RECVORIGDSTADDR:
if (optlen < sizeof(int))
goto e_inval;
@@ -1214,6 +1222,10 @@ static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
val = inet_sk(sk)->transparent;
break;
+ case IPV6_FREEBIND:
+ val = inet_sk(sk)->freebind;
+ break;
+
case IPV6_RECVORIGDSTADDR:
val = np->rxopt.bits.rxorigdstaddr;
break;
--
2.14.1.992.g2c7b836f3a-goog
^ permalink raw reply related
* Re: [PATCH] net: stmmac: Meet alignment requirements for DMA
From: Paul Burton @ 2017-09-27 4:30 UTC (permalink / raw)
To: David Miller
Cc: matt.redfearn, netdev, alexandre.torgue, peppe.cavallaro,
linux-kernel, linux-mips, james.hogan
In-Reply-To: <20170926.195244.506518182147628099.davem@davemloft.net>
[-- Attachment #1: Type: text/plain, Size: 1433 bytes --]
Hi David,
On Tuesday, 26 September 2017 19:52:44 PDT David Miller wrote:
> From: Paul Burton <paul.burton@imgtec.com>
> Date: Tue, 26 Sep 2017 14:30:33 -0700
>
> > I'd suggest that at a minimum if you're unwilling to obey the API as
> > described in Documentation/DMA-API.txt then it would be beneficial
> > if you could propose a change to it such that it works for you, and
> > perhaps we can extend the API & its documentation to allow your
> > usage whilst also allowing us to catch broken uses.
>
> The networking driver code works fine as is.
>
> I also didn't write that ill-advised documentation in the DMA docs,
> nor the non-merged new MIPS assertion.
>
> So I'm trying to figure out on what basis I am required to do
> anything.
>
> Thank you.
Nobody said you wrote the documentation, but you do maintain code which
disobeys the documented DMA API & now you're being an ass about it
unnecessarily.
Nobody said that you are required to do anything, I suggested that it would be
beneficial if you were to suggest a change to the documented DMA API such that
it allows your usage where it currently does not. If you don't want to have
any input into that, and you actually think that your current approach of
ignoring the documented API is the best path forwards, then we're probably
done here & I'll be making a note to avoid yourself & anything under net/ to
whatever extent is possible...
Thanks,
Paul
[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
* Re: [PATCH net-next 0/7] nfp: flower vxlan tunnel offload
From: David Miller @ 2017-09-27 4:29 UTC (permalink / raw)
To: simon.horman; +Cc: jakub.kicinski, netdev, oss-drivers
In-Reply-To: <1506335021-32024-1-git-send-email-simon.horman@netronome.com>
From: Simon Horman <simon.horman@netronome.com>
Date: Mon, 25 Sep 2017 12:23:34 +0200
> From: Simon Horman <simon.horman@netronome.com>
>
> John says:
>
> This patch set allows offloading of TC flower match and set tunnel fields
> to the NFP. The initial focus is on VXLAN traffic. Due to the current
> state of the NFP firmware, only VXLAN traffic on well known port 4789 is
> handled. The match and action fields must explicity set this value to be
> supported. Tunnel end point information is also offloaded to the NFP for
> both encapsulation and decapsulation. The NFP expects 3 separate data sets
> to be supplied.
...
Series applied, thanks.
I see there is some discussion about ipv6 flow dissector key handling
and ND keepalives, but those should be addressable in follow-on changes.
Thanks.
^ permalink raw reply
* Re: [PATCH net v2] sctp: Fix a big endian bug in sctp_diag_dump()
From: David Miller @ 2017-09-27 4:17 UTC (permalink / raw)
To: dan.carpenter
Cc: vyasevich, lucien.xin, nhorman, linux-sctp, netdev,
kernel-janitors
In-Reply-To: <20170925101926.db4f6x4hblh7tcvo@mwanda>
From: Dan Carpenter <dan.carpenter@oracle.com>
Date: Mon, 25 Sep 2017 13:19:26 +0300
> The sctp_for_each_transport() function takes an pointer to int. The
> cb->args[] array holds longs so it's only using the high 32 bits. It
> works on little endian system but will break on big endian 64 bit
> machines.
>
> Fixes: d25adbeb0cdb ("sctp: fix an use-after-free issue in sctp_sock_dump")
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> ---
> v2: The v1 patch changed the function to take a long pointer, but v2
> just changes the caller.
Applied, thanks.
^ permalink raw reply
* Re: [PATCH net] net/ncsi: Don't assume last available channel exists
From: Samuel Mendoza-Jonas @ 2017-09-27 4:12 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-kernel, Benjamin Herrenschmidt
In-Reply-To: <20170921.181111.2086501653519863373.davem@davemloft.net>
On Thu, 2017-09-21 at 18:11 -0700, David Miller wrote:
> From: Samuel Mendoza-Jonas <sam@mendozajonas.com>
> Date: Fri, 22 Sep 2017 11:00:00 +1000
>
> > If we haven't configured a channel yet (or are in the process of doing
> > so) we won't have a hot_channel - does it make more sense to
> > - check against the hot_channel as currently done,
> > - only check the filter size at configure time for /each/ channel,
> > - only conditionally enable the .ndo_vlan_rx_add_vid net_device callback
> > once we've configured a channel (eg. for ftgmac100 in the
> > ftgmac100_ncsi_handler() callback?)
>
> The last isn't so feasible.
>
> The device shouldn't be marked attached until a channel is available,
> because it seems like communication cannot occur until one is. Right?
Yes that's right.
>
> You could experiment with netif_device_detach()/netif_device_attach().
>
> When the device is in the detached state, callbacks such as
> ->ndo_vlan_rx_add_vid() will not be invoked.
This looked like the way at first, but _detach() ceases any tx/rx on the
interface right?
NCSI still needs the interface to be active since the 'channels' are on a
separate network controller that the interface is connected to, eg on the
machines I'm using:
BMC 'Host' network controller
---------------------- ----------------------------
|ftgmac100 interface | <---- NCSI Link ----> | BCM5719 interface | --> external interface
---------------------- ----------------------------
Looking at the NCSI init path I believe we're guaranteed to have an ndp
struct by the time ndo_vlan_rx_add_vid() is called, making some of those
checks overly cautious. It might be easiest to just track new vids as we
see them (up to the NCSI spec limit), and then deal with configured
channels on a case by case basis since their limits can be different.
I'll work on a V2 but hopefully I haven't misinterpreted
_detach()/_attach() :)
Sam
^ permalink raw reply
* [PATCH] net-ipv6: remove unused IP6_ECN_clear() function
From: Maciej Żenczykowski @ 2017-09-27 3:37 UTC (permalink / raw)
To: Maciej Żenczykowski, David S . Miller; +Cc: netdev
From: Maciej Żenczykowski <maze@google.com>
This function is unused, and furthermore it is buggy since it suffers
from the same issue that requires IP6_ECN_set_ce() to take a pointer
to the skb so that it may (in case of CHECKSUM_COMPLETE) update skb->csum
Instead of fixing it, let's just outright remove it.
Tested: builds, and 'git grep IP6_ECN_clear' comes up empty
Signed-off-by: Maciej Żenczykowski <maze@google.com>
---
include/net/inet_ecn.h | 5 -----
1 file changed, 5 deletions(-)
diff --git a/include/net/inet_ecn.h b/include/net/inet_ecn.h
index dce2d586d9ce..f5ff16d72fe6 100644
--- a/include/net/inet_ecn.h
+++ b/include/net/inet_ecn.h
@@ -133,11 +133,6 @@ static inline int IP6_ECN_set_ce(struct sk_buff *skb, struct ipv6hdr *iph)
return 1;
}
-static inline void IP6_ECN_clear(struct ipv6hdr *iph)
-{
- *(__be32*)iph &= ~htonl(INET_ECN_MASK << 20);
-}
-
static inline void ipv6_copy_dscp(unsigned int dscp, struct ipv6hdr *inner)
{
dscp &= ~INET_ECN_MASK;
--
2.14.1.992.g2c7b836f3a-goog
^ permalink raw reply related
* [PATCH v5 3/4] ipv4: Namespaceify tcp_fastopen_key knob
From: Haishuang Yan @ 2017-09-27 3:35 UTC (permalink / raw)
To: David S. Miller, Alexey Kuznetsov, Eric Dumazet, Wei Wang,
Luca BRUNO
Cc: netdev, linux-kernel, Haishuang Yan
In-Reply-To: <1506483343-11544-1-git-send-email-yanhaishuang@cmss.chinamobile.com>
Different namespace application might require different tcp_fastopen_key
independently of the host.
David Miller pointed out there is a leak without releasing the context
of tcp_fastopen_key during netns teardown. So add the release action in
exit_batch path.
Tested:
1. Container namespace:
# cat /proc/sys/net/ipv4/tcp_fastopen_key:
2817fff2-f803cf97-eadfd1f3-78c0992b
cookie key in tcp syn packets:
Fast Open Cookie
Kind: TCP Fast Open Cookie (34)
Length: 10
Fast Open Cookie: 1e5dd82a8c492ca9
2. Host:
# cat /proc/sys/net/ipv4/tcp_fastopen_key:
107d7c5f-68eb2ac7-02fb06e6-ed341702
cookie key in tcp syn packets:
Fast Open Cookie
Kind: TCP Fast Open Cookie (34)
Length: 10
Fast Open Cookie: e213c02bf0afbc8a
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
---
include/net/netns/ipv4.h | 4 +++
include/net/tcp.h | 6 ++---
net/ipv4/af_inet.c | 2 +-
net/ipv4/sysctl_net_ipv4.c | 21 ++++++++-------
net/ipv4/tcp.c | 2 +-
net/ipv4/tcp_fastopen.c | 64 +++++++++++++++++++++++++++++++---------------
net/ipv4/tcp_ipv4.c | 6 +++++
7 files changed, 70 insertions(+), 35 deletions(-)
diff --git a/include/net/netns/ipv4.h b/include/net/netns/ipv4.h
index ce6dde0..66b8335 100644
--- a/include/net/netns/ipv4.h
+++ b/include/net/netns/ipv4.h
@@ -36,6 +36,8 @@ struct inet_timewait_death_row {
int sysctl_max_tw_buckets;
};
+struct tcp_fastopen_context;
+
struct netns_ipv4 {
#ifdef CONFIG_SYSCTL
struct ctl_table_header *forw_hdr;
@@ -128,6 +130,8 @@ struct netns_ipv4 {
struct inet_timewait_death_row tcp_death_row;
int sysctl_max_syn_backlog;
int sysctl_tcp_fastopen;
+ struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
+ spinlock_t tcp_fastopen_ctx_lock;
#ifdef CONFIG_NET_L3_MASTER_DEV
int sysctl_udp_l3mdev_accept;
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 061c128..e27bd18 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -1556,13 +1556,13 @@ struct tcp_fastopen_request {
};
void tcp_free_fastopen_req(struct tcp_sock *tp);
-extern struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
-int tcp_fastopen_reset_cipher(void *key, unsigned int len);
+void tcp_fastopen_ctx_destroy(struct net *net);
+int tcp_fastopen_reset_cipher(struct net *net, void *key, unsigned int len);
void tcp_fastopen_add_skb(struct sock *sk, struct sk_buff *skb);
struct sock *tcp_try_fastopen(struct sock *sk, struct sk_buff *skb,
struct request_sock *req,
struct tcp_fastopen_cookie *foc);
-void tcp_fastopen_init_key_once(void);
+void tcp_fastopen_init_key_once(struct net *net);
bool tcp_fastopen_cookie_check(struct sock *sk, u16 *mss,
struct tcp_fastopen_cookie *cookie);
bool tcp_fastopen_defer_connect(struct sock *sk, int *err);
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index e73ce79..43a1bbe 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -222,7 +222,7 @@ int inet_listen(struct socket *sock, int backlog)
(tcp_fastopen & TFO_SERVER_ENABLE) &&
!inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
fastopen_queue_tune(sk, backlog);
- tcp_fastopen_init_key_once();
+ tcp_fastopen_init_key_once(sock_net(sk));
}
err = inet_csk_listen_start(sk, backlog);
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index f6324ea..20e19fe 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -251,10 +251,12 @@ static int proc_allowed_congestion_control(struct ctl_table *ctl,
return ret;
}
-static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
+static int proc_tcp_fastopen_key(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp,
loff_t *ppos)
{
+ struct net *net = container_of(table->data, struct net,
+ ipv4.sysctl_tcp_fastopen);
struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
struct tcp_fastopen_context *ctxt;
int ret;
@@ -265,7 +267,7 @@ static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
return -ENOMEM;
rcu_read_lock();
- ctxt = rcu_dereference(tcp_fastopen_ctx);
+ ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
if (ctxt)
memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
else
@@ -282,7 +284,7 @@ static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
ret = -EINVAL;
goto bad_key;
}
- tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
+ tcp_fastopen_reset_cipher(net, user_key, TCP_FASTOPEN_KEY_LENGTH);
}
bad_key:
@@ -396,12 +398,6 @@ static int proc_tcp_available_ulp(struct ctl_table *ctl,
.proc_handler = proc_dointvec
},
{
- .procname = "tcp_fastopen_key",
- .mode = 0600,
- .maxlen = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
- .proc_handler = proc_tcp_fastopen_key,
- },
- {
.procname = "tcp_fastopen_blackhole_timeout_sec",
.data = &sysctl_tcp_fastopen_blackhole_timeout,
.maxlen = sizeof(int),
@@ -1080,6 +1076,13 @@ static int proc_tcp_available_ulp(struct ctl_table *ctl,
.mode = 0644,
.proc_handler = proc_dointvec,
},
+ {
+ .procname = "tcp_fastopen_key",
+ .mode = 0600,
+ .data = &init_net.ipv4.sysctl_tcp_fastopen,
+ .maxlen = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
+ .proc_handler = proc_tcp_fastopen_key,
+ },
#ifdef CONFIG_IP_ROUTE_MULTIPATH
{
.procname = "fib_multipath_use_neigh",
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 4e39545..23225c9 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2749,7 +2749,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
case TCP_FASTOPEN:
if (val >= 0 && ((1 << sk->sk_state) & (TCPF_CLOSE |
TCPF_LISTEN))) {
- tcp_fastopen_init_key_once();
+ tcp_fastopen_init_key_once(net);
fastopen_queue_tune(sk, val);
} else {
diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
index 8c8f0f0..4eae44a 100644
--- a/net/ipv4/tcp_fastopen.c
+++ b/net/ipv4/tcp_fastopen.c
@@ -9,13 +9,18 @@
#include <net/inetpeer.h>
#include <net/tcp.h>
-struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
-
-static DEFINE_SPINLOCK(tcp_fastopen_ctx_lock);
-
-void tcp_fastopen_init_key_once(void)
+void tcp_fastopen_init_key_once(struct net *net)
{
- static u8 key[TCP_FASTOPEN_KEY_LENGTH];
+ u8 key[TCP_FASTOPEN_KEY_LENGTH];
+ struct tcp_fastopen_context *ctxt;
+
+ rcu_read_lock();
+ ctxt = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
+ if (ctxt) {
+ rcu_read_unlock();
+ return;
+ }
+ rcu_read_unlock();
/* tcp_fastopen_reset_cipher publishes the new context
* atomically, so we allow this race happening here.
@@ -23,8 +28,8 @@ void tcp_fastopen_init_key_once(void)
* All call sites of tcp_fastopen_cookie_gen also check
* for a valid cookie, so this is an acceptable risk.
*/
- if (net_get_random_once(key, sizeof(key)))
- tcp_fastopen_reset_cipher(key, sizeof(key));
+ get_random_bytes(key, sizeof(key));
+ tcp_fastopen_reset_cipher(net, key, sizeof(key));
}
static void tcp_fastopen_ctx_free(struct rcu_head *head)
@@ -35,7 +40,22 @@ static void tcp_fastopen_ctx_free(struct rcu_head *head)
kfree(ctx);
}
-int tcp_fastopen_reset_cipher(void *key, unsigned int len)
+void tcp_fastopen_ctx_destroy(struct net *net)
+{
+ struct tcp_fastopen_context *ctxt;
+
+ spin_lock(&net->ipv4.tcp_fastopen_ctx_lock);
+
+ ctxt = rcu_dereference_protected(net->ipv4.tcp_fastopen_ctx,
+ lockdep_is_held(&net->ipv4.tcp_fastopen_ctx_lock));
+ rcu_assign_pointer(net->ipv4.tcp_fastopen_ctx, NULL);
+ spin_unlock(&net->ipv4.tcp_fastopen_ctx_lock);
+
+ if (ctxt)
+ call_rcu(&ctxt->rcu, tcp_fastopen_ctx_free);
+}
+
+int tcp_fastopen_reset_cipher(struct net *net, void *key, unsigned int len)
{
int err;
struct tcp_fastopen_context *ctx, *octx;
@@ -59,26 +79,27 @@ int tcp_fastopen_reset_cipher(void *key, unsigned int len)
}
memcpy(ctx->key, key, len);
- spin_lock(&tcp_fastopen_ctx_lock);
+ spin_lock(&net->ipv4.tcp_fastopen_ctx_lock);
- octx = rcu_dereference_protected(tcp_fastopen_ctx,
- lockdep_is_held(&tcp_fastopen_ctx_lock));
- rcu_assign_pointer(tcp_fastopen_ctx, ctx);
- spin_unlock(&tcp_fastopen_ctx_lock);
+ octx = rcu_dereference_protected(net->ipv4.tcp_fastopen_ctx,
+ lockdep_is_held(&net->ipv4.tcp_fastopen_ctx_lock));
+ rcu_assign_pointer(net->ipv4.tcp_fastopen_ctx, ctx);
+ spin_unlock(&net->ipv4.tcp_fastopen_ctx_lock);
if (octx)
call_rcu(&octx->rcu, tcp_fastopen_ctx_free);
return err;
}
-static bool __tcp_fastopen_cookie_gen(const void *path,
+static bool __tcp_fastopen_cookie_gen(struct net *net,
+ const void *path,
struct tcp_fastopen_cookie *foc)
{
struct tcp_fastopen_context *ctx;
bool ok = false;
rcu_read_lock();
- ctx = rcu_dereference(tcp_fastopen_ctx);
+ ctx = rcu_dereference(net->ipv4.tcp_fastopen_ctx);
if (ctx) {
crypto_cipher_encrypt_one(ctx->tfm, foc->val, path);
foc->len = TCP_FASTOPEN_COOKIE_SIZE;
@@ -94,7 +115,8 @@ static bool __tcp_fastopen_cookie_gen(const void *path,
*
* XXX (TFO) - refactor when TCP_FASTOPEN_COOKIE_SIZE != AES_BLOCK_SIZE.
*/
-static bool tcp_fastopen_cookie_gen(struct request_sock *req,
+static bool tcp_fastopen_cookie_gen(struct net *net,
+ struct request_sock *req,
struct sk_buff *syn,
struct tcp_fastopen_cookie *foc)
{
@@ -102,7 +124,7 @@ static bool tcp_fastopen_cookie_gen(struct request_sock *req,
const struct iphdr *iph = ip_hdr(syn);
__be32 path[4] = { iph->saddr, iph->daddr, 0, 0 };
- return __tcp_fastopen_cookie_gen(path, foc);
+ return __tcp_fastopen_cookie_gen(net, path, foc);
}
#if IS_ENABLED(CONFIG_IPV6)
@@ -110,13 +132,13 @@ static bool tcp_fastopen_cookie_gen(struct request_sock *req,
const struct ipv6hdr *ip6h = ipv6_hdr(syn);
struct tcp_fastopen_cookie tmp;
- if (__tcp_fastopen_cookie_gen(&ip6h->saddr, &tmp)) {
+ if (__tcp_fastopen_cookie_gen(net, &ip6h->saddr, &tmp)) {
struct in6_addr *buf = &tmp.addr;
int i;
for (i = 0; i < 4; i++)
buf->s6_addr32[i] ^= ip6h->daddr.s6_addr32[i];
- return __tcp_fastopen_cookie_gen(buf, foc);
+ return __tcp_fastopen_cookie_gen(net, buf, foc);
}
}
#endif
@@ -296,7 +318,7 @@ struct sock *tcp_try_fastopen(struct sock *sk, struct sk_buff *skb,
goto fastopen;
if (foc->len >= 0 && /* Client presents or requests a cookie */
- tcp_fastopen_cookie_gen(req, skb, &valid_foc) &&
+ tcp_fastopen_cookie_gen(sock_net(sk), req, skb, &valid_foc) &&
foc->len == TCP_FASTOPEN_COOKIE_SIZE &&
foc->len == valid_foc.len &&
!memcmp(foc->val, valid_foc.val, foc->len)) {
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 88409b1..49c74c0 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -2473,6 +2473,7 @@ static int __net_init tcp_sk_init(struct net *net)
net->ipv4.sysctl_tcp_timestamps = 1;
net->ipv4.sysctl_tcp_fastopen = TFO_CLIENT_ENABLE;
+ spin_lock_init(&net->ipv4.tcp_fastopen_ctx_lock);
return 0;
fail:
@@ -2483,7 +2484,12 @@ static int __net_init tcp_sk_init(struct net *net)
static void __net_exit tcp_sk_exit_batch(struct list_head *net_exit_list)
{
+ struct net *net;
+
inet_twsk_purge(&tcp_hashinfo, AF_INET);
+
+ list_for_each_entry(net, net_exit_list, exit_list)
+ tcp_fastopen_ctx_destroy(net);
}
static struct pernet_operations __net_initdata tcp_sk_ops = {
--
1.8.3.1
^ permalink raw reply related
* [PATCH v5 4/4] ipv4: Namespaceify tcp_fastopen_blackhole_timeout knob
From: Haishuang Yan @ 2017-09-27 3:35 UTC (permalink / raw)
To: David S. Miller, Alexey Kuznetsov, Eric Dumazet, Wei Wang,
Luca BRUNO
Cc: netdev, linux-kernel, Haishuang Yan
In-Reply-To: <1506483343-11544-1-git-send-email-yanhaishuang@cmss.chinamobile.com>
Different namespace application might require different time period in
second to disable Fastopen on active TCP sockets.
Tested:
Simulate following similar situation that the server's data gets dropped
after 3WHS.
C ---- syn-data ---> S
C <--- syn/ack ----- S
C ---- ack --------> S
S (accept & write)
C? X <- data ------ S
[retry and timeout]
And then print netstat of TCPFastOpenBlackhole, the counter increased as
expected when the firewall blackhole issue is detected and active TFO is
disabled.
# cat /proc/net/netstat | awk '{print $91}'
TCPFastOpenBlackhole
1
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
---
include/net/netns/ipv4.h | 3 +++
net/ipv4/sysctl_net_ipv4.c | 20 +++++++++++---------
net/ipv4/tcp_fastopen.c | 30 +++++++++++-------------------
net/ipv4/tcp_ipv4.c | 2 ++
4 files changed, 27 insertions(+), 28 deletions(-)
diff --git a/include/net/netns/ipv4.h b/include/net/netns/ipv4.h
index 66b8335..d76edde 100644
--- a/include/net/netns/ipv4.h
+++ b/include/net/netns/ipv4.h
@@ -132,6 +132,9 @@ struct netns_ipv4 {
int sysctl_tcp_fastopen;
struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
spinlock_t tcp_fastopen_ctx_lock;
+ unsigned int sysctl_tcp_fastopen_blackhole_timeout;
+ atomic_t tfo_active_disable_times;
+ unsigned long tfo_active_disable_stamp;
#ifdef CONFIG_NET_L3_MASTER_DEV
int sysctl_udp_l3mdev_accept;
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index 20e19fe..cac8dd3 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -355,11 +355,13 @@ static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
void __user *buffer,
size_t *lenp, loff_t *ppos)
{
+ struct net *net = container_of(table->data, struct net,
+ ipv4.sysctl_tcp_fastopen_blackhole_timeout);
int ret;
ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
if (write && ret == 0)
- tcp_fastopen_active_timeout_reset();
+ atomic_set(&net->ipv4.tfo_active_disable_times, 0);
return ret;
}
@@ -398,14 +400,6 @@ static int proc_tcp_available_ulp(struct ctl_table *ctl,
.proc_handler = proc_dointvec
},
{
- .procname = "tcp_fastopen_blackhole_timeout_sec",
- .data = &sysctl_tcp_fastopen_blackhole_timeout,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = proc_tfo_blackhole_detect_timeout,
- .extra1 = &zero,
- },
- {
.procname = "tcp_abort_on_overflow",
.data = &sysctl_tcp_abort_on_overflow,
.maxlen = sizeof(int),
@@ -1083,6 +1077,14 @@ static int proc_tcp_available_ulp(struct ctl_table *ctl,
.maxlen = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
.proc_handler = proc_tcp_fastopen_key,
},
+ {
+ .procname = "tcp_fastopen_blackhole_timeout_sec",
+ .data = &init_net.ipv4.sysctl_tcp_fastopen_blackhole_timeout,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = proc_tfo_blackhole_detect_timeout,
+ .extra1 = &zero,
+ },
#ifdef CONFIG_IP_ROUTE_MULTIPATH
{
.procname = "fib_multipath_use_neigh",
diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
index 4eae44a..de470e7 100644
--- a/net/ipv4/tcp_fastopen.c
+++ b/net/ipv4/tcp_fastopen.c
@@ -422,25 +422,16 @@ bool tcp_fastopen_defer_connect(struct sock *sk, int *err)
* TFO connection with data exchanges.
*/
-/* Default to 1hr */
-unsigned int sysctl_tcp_fastopen_blackhole_timeout __read_mostly = 60 * 60;
-static atomic_t tfo_active_disable_times __read_mostly = ATOMIC_INIT(0);
-static unsigned long tfo_active_disable_stamp __read_mostly;
-
/* Disable active TFO and record current jiffies and
* tfo_active_disable_times
*/
void tcp_fastopen_active_disable(struct sock *sk)
{
- atomic_inc(&tfo_active_disable_times);
- tfo_active_disable_stamp = jiffies;
- NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPFASTOPENBLACKHOLE);
-}
+ struct net *net = sock_net(sk);
-/* Reset tfo_active_disable_times to 0 */
-void tcp_fastopen_active_timeout_reset(void)
-{
- atomic_set(&tfo_active_disable_times, 0);
+ atomic_inc(&net->ipv4.tfo_active_disable_times);
+ net->ipv4.tfo_active_disable_stamp = jiffies;
+ NET_INC_STATS(net, LINUX_MIB_TCPFASTOPENBLACKHOLE);
}
/* Calculate timeout for tfo active disable
@@ -449,17 +440,18 @@ void tcp_fastopen_active_timeout_reset(void)
*/
bool tcp_fastopen_active_should_disable(struct sock *sk)
{
- int tfo_da_times = atomic_read(&tfo_active_disable_times);
- int multiplier;
+ unsigned int tfo_bh_timeout = sock_net(sk)->ipv4.sysctl_tcp_fastopen_blackhole_timeout;
+ int tfo_da_times = atomic_read(&sock_net(sk)->ipv4.tfo_active_disable_times);
unsigned long timeout;
+ int multiplier;
if (!tfo_da_times)
return false;
/* Limit timout to max: 2^6 * initial timeout */
multiplier = 1 << min(tfo_da_times - 1, 6);
- timeout = multiplier * sysctl_tcp_fastopen_blackhole_timeout * HZ;
- if (time_before(jiffies, tfo_active_disable_stamp + timeout))
+ timeout = multiplier * tfo_bh_timeout * HZ;
+ if (time_before(jiffies, sock_net(sk)->ipv4.tfo_active_disable_stamp + timeout))
return true;
/* Mark check bit so we can check for successful active TFO
@@ -495,10 +487,10 @@ void tcp_fastopen_active_disable_ofo_check(struct sock *sk)
}
}
} else if (tp->syn_fastopen_ch &&
- atomic_read(&tfo_active_disable_times)) {
+ atomic_read(&sock_net(sk)->ipv4.tfo_active_disable_times)) {
dst = sk_dst_get(sk);
if (!(dst && dst->dev && (dst->dev->flags & IFF_LOOPBACK)))
- tcp_fastopen_active_timeout_reset();
+ atomic_set(&sock_net(sk)->ipv4.tfo_active_disable_times, 0);
dst_release(dst);
}
}
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 49c74c0..ad3b5bb 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -2474,6 +2474,8 @@ static int __net_init tcp_sk_init(struct net *net)
net->ipv4.sysctl_tcp_fastopen = TFO_CLIENT_ENABLE;
spin_lock_init(&net->ipv4.tcp_fastopen_ctx_lock);
+ net->ipv4.sysctl_tcp_fastopen_blackhole_timeout = 60 * 60;
+ atomic_set(&net->ipv4.tfo_active_disable_times, 0);
return 0;
fail:
--
1.8.3.1
^ permalink raw reply related
* [PATCH v5 2/4] ipv4: Remove the 'publish' logic in tcp_fastopen_init_key_once
From: Haishuang Yan @ 2017-09-27 3:35 UTC (permalink / raw)
To: David S. Miller, Alexey Kuznetsov, Eric Dumazet, Wei Wang,
Luca BRUNO
Cc: netdev, linux-kernel, Haishuang Yan
In-Reply-To: <1506483343-11544-1-git-send-email-yanhaishuang@cmss.chinamobile.com>
The 'publish' logic is not necessary after commit dfea2aa65424 ("tcp:
Do not call tcp_fastopen_reset_cipher from interrupt context"), because
in tcp_fastopen_cookie_gen，it wouldn't call tcp_fastopen_init_key_once.
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
---
include/net/tcp.h | 2 +-
net/ipv4/af_inet.c | 2 +-
net/ipv4/sysctl_net_ipv4.c | 5 -----
net/ipv4/tcp.c | 2 +-
net/ipv4/tcp_fastopen.c | 4 ++--
5 files changed, 5 insertions(+), 10 deletions(-)
diff --git a/include/net/tcp.h b/include/net/tcp.h
index f628967..061c128 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -1562,7 +1562,7 @@ struct tcp_fastopen_request {
struct sock *tcp_try_fastopen(struct sock *sk, struct sk_buff *skb,
struct request_sock *req,
struct tcp_fastopen_cookie *foc);
-void tcp_fastopen_init_key_once(bool publish);
+void tcp_fastopen_init_key_once(void);
bool tcp_fastopen_cookie_check(struct sock *sk, u16 *mss,
struct tcp_fastopen_cookie *cookie);
bool tcp_fastopen_defer_connect(struct sock *sk, int *err);
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index ddd126d..e73ce79 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -222,7 +222,7 @@ int inet_listen(struct socket *sock, int backlog)
(tcp_fastopen & TFO_SERVER_ENABLE) &&
!inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
fastopen_queue_tune(sk, backlog);
- tcp_fastopen_init_key_once(true);
+ tcp_fastopen_init_key_once();
}
err = inet_csk_listen_start(sk, backlog);
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index e31e853c..f6324ea 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -282,11 +282,6 @@ static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
ret = -EINVAL;
goto bad_key;
}
- /* Generate a dummy secret but don't publish it. This
- * is needed so we don't regenerate a new key on the
- * first invocation of tcp_fastopen_cookie_gen
- */
- tcp_fastopen_init_key_once(false);
tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
}
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index dac56c4..4e39545 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2749,7 +2749,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
case TCP_FASTOPEN:
if (val >= 0 && ((1 << sk->sk_state) & (TCPF_CLOSE |
TCPF_LISTEN))) {
- tcp_fastopen_init_key_once(true);
+ tcp_fastopen_init_key_once();
fastopen_queue_tune(sk, val);
} else {
diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
index 31b08ec..8c8f0f0 100644
--- a/net/ipv4/tcp_fastopen.c
+++ b/net/ipv4/tcp_fastopen.c
@@ -13,7 +13,7 @@
static DEFINE_SPINLOCK(tcp_fastopen_ctx_lock);
-void tcp_fastopen_init_key_once(bool publish)
+void tcp_fastopen_init_key_once(void)
{
static u8 key[TCP_FASTOPEN_KEY_LENGTH];
@@ -23,7 +23,7 @@ void tcp_fastopen_init_key_once(bool publish)
* All call sites of tcp_fastopen_cookie_gen also check
* for a valid cookie, so this is an acceptable risk.
*/
- if (net_get_random_once(key, sizeof(key)) && publish)
+ if (net_get_random_once(key, sizeof(key)))
tcp_fastopen_reset_cipher(key, sizeof(key));
}
--
1.8.3.1
^ permalink raw reply related
* [PATCH v5 1/4] ipv4: Namespaceify tcp_fastopen knob
From: Haishuang Yan @ 2017-09-27 3:35 UTC (permalink / raw)
To: David S. Miller, Alexey Kuznetsov, Eric Dumazet, Wei Wang,
Luca BRUNO
Cc: netdev, linux-kernel, Haishuang Yan
Different namespace application might require enable TCP Fast Open
feature independently of the host.
This patch series continues making more of the TCP Fast Open related
sysctl knobs be per net-namespace.
Reported-by: Luca BRUNO <lucab@debian.org>
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
---
Changes since v5:
* Splite patch #2 into two changes
* Fix potential leak issue
* Fix typo error
* Split the patches to be per-sysctl
* Remove unrelated change by mistake
---
include/net/netns/ipv4.h | 1 +
include/net/tcp.h | 1 -
net/ipv4/af_inet.c | 7 ++++---
net/ipv4/sysctl_net_ipv4.c | 14 +++++++-------
net/ipv4/tcp.c | 4 ++--
net/ipv4/tcp_fastopen.c | 11 +++++------
net/ipv4/tcp_ipv4.c | 2 ++
7 files changed, 21 insertions(+), 19 deletions(-)
diff --git a/include/net/netns/ipv4.h b/include/net/netns/ipv4.h
index 20d061c..ce6dde0 100644
--- a/include/net/netns/ipv4.h
+++ b/include/net/netns/ipv4.h
@@ -127,6 +127,7 @@ struct netns_ipv4 {
int sysctl_tcp_timestamps;
struct inet_timewait_death_row tcp_death_row;
int sysctl_max_syn_backlog;
+ int sysctl_tcp_fastopen;
#ifdef CONFIG_NET_L3_MASTER_DEV
int sysctl_udp_l3mdev_accept;
diff --git a/include/net/tcp.h b/include/net/tcp.h
index b510f28..f628967 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -240,7 +240,6 @@
/* sysctl variables for tcp */
-extern int sysctl_tcp_fastopen;
extern int sysctl_tcp_retrans_collapse;
extern int sysctl_tcp_stdurg;
extern int sysctl_tcp_rfc1337;
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index e31108e..ddd126d 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -195,7 +195,7 @@ int inet_listen(struct socket *sock, int backlog)
{
struct sock *sk = sock->sk;
unsigned char old_state;
- int err;
+ int err, tcp_fastopen;
lock_sock(sk);
@@ -217,8 +217,9 @@ int inet_listen(struct socket *sock, int backlog)
* because the socket was in TCP_LISTEN state previously but
* was shutdown() rather than close().
*/
- if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) &&
- (sysctl_tcp_fastopen & TFO_SERVER_ENABLE) &&
+ tcp_fastopen = sock_net(sk)->ipv4.sysctl_tcp_fastopen;
+ if ((tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) &&
+ (tcp_fastopen & TFO_SERVER_ENABLE) &&
!inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
fastopen_queue_tune(sk, backlog);
tcp_fastopen_init_key_once(true);
diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index 0d3c038..e31e853c 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -401,13 +401,6 @@ static int proc_tcp_available_ulp(struct ctl_table *ctl,
.proc_handler = proc_dointvec
},
{
- .procname = "tcp_fastopen",
- .data = &sysctl_tcp_fastopen,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = proc_dointvec,
- },
- {
.procname = "tcp_fastopen_key",
.mode = 0600,
.maxlen = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
@@ -1085,6 +1078,13 @@ static int proc_tcp_available_ulp(struct ctl_table *ctl,
.mode = 0644,
.proc_handler = proc_dointvec
},
+ {
+ .procname = "tcp_fastopen",
+ .data = &init_net.ipv4.sysctl_tcp_fastopen,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = proc_dointvec,
+ },
#ifdef CONFIG_IP_ROUTE_MULTIPATH
{
.procname = "fib_multipath_use_neigh",
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 5091402..dac56c4 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1126,7 +1126,7 @@ static int tcp_sendmsg_fastopen(struct sock *sk, struct msghdr *msg,
struct sockaddr *uaddr = msg->msg_name;
int err, flags;
- if (!(sysctl_tcp_fastopen & TFO_CLIENT_ENABLE) ||
+ if (!(sock_net(sk)->ipv4.sysctl_tcp_fastopen & TFO_CLIENT_ENABLE) ||
(uaddr && msg->msg_namelen >= sizeof(uaddr->sa_family) &&
uaddr->sa_family == AF_UNSPEC))
return -EOPNOTSUPP;
@@ -2759,7 +2759,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
case TCP_FASTOPEN_CONNECT:
if (val > 1 || val < 0) {
err = -EINVAL;
- } else if (sysctl_tcp_fastopen & TFO_CLIENT_ENABLE) {
+ } else if (net->ipv4.sysctl_tcp_fastopen & TFO_CLIENT_ENABLE) {
if (sk->sk_state == TCP_CLOSE)
tp->fastopen_connect = val;
else
diff --git a/net/ipv4/tcp_fastopen.c b/net/ipv4/tcp_fastopen.c
index e3c3322..31b08ec 100644
--- a/net/ipv4/tcp_fastopen.c
+++ b/net/ipv4/tcp_fastopen.c
@@ -9,8 +9,6 @@
#include <net/inetpeer.h>
#include <net/tcp.h>
-int sysctl_tcp_fastopen __read_mostly = TFO_CLIENT_ENABLE;
-
struct tcp_fastopen_context __rcu *tcp_fastopen_ctx;
static DEFINE_SPINLOCK(tcp_fastopen_ctx_lock);
@@ -279,21 +277,22 @@ struct sock *tcp_try_fastopen(struct sock *sk, struct sk_buff *skb,
struct request_sock *req,
struct tcp_fastopen_cookie *foc)
{
- struct tcp_fastopen_cookie valid_foc = { .len = -1 };
bool syn_data = TCP_SKB_CB(skb)->end_seq != TCP_SKB_CB(skb)->seq + 1;
+ int tcp_fastopen = sock_net(sk)->ipv4.sysctl_tcp_fastopen;
+ struct tcp_fastopen_cookie valid_foc = { .len = -1 };
struct sock *child;
if (foc->len == 0) /* Client requests a cookie */
NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPFASTOPENCOOKIEREQD);
- if (!((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) &&
+ if (!((tcp_fastopen & TFO_SERVER_ENABLE) &&
(syn_data || foc->len >= 0) &&
tcp_fastopen_queue_check(sk))) {
foc->len = -1;
return NULL;
}
- if (syn_data && (sysctl_tcp_fastopen & TFO_SERVER_COOKIE_NOT_REQD))
+ if (syn_data && (tcp_fastopen & TFO_SERVER_COOKIE_NOT_REQD))
goto fastopen;
if (foc->len >= 0 && /* Client presents or requests a cookie */
@@ -347,7 +346,7 @@ bool tcp_fastopen_cookie_check(struct sock *sk, u16 *mss,
return false;
}
- if (sysctl_tcp_fastopen & TFO_CLIENT_NO_COOKIE) {
+ if (sock_net(sk)->ipv4.sysctl_tcp_fastopen & TFO_CLIENT_NO_COOKIE) {
cookie->len = -1;
return true;
}
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index d9416b5..88409b1 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -2472,6 +2472,8 @@ static int __net_init tcp_sk_init(struct net *net)
net->ipv4.sysctl_tcp_window_scaling = 1;
net->ipv4.sysctl_tcp_timestamps = 1;
+ net->ipv4.sysctl_tcp_fastopen = TFO_CLIENT_ENABLE;
+
return 0;
fail:
tcp_sk_exit(net);
--
1.8.3.1
^ permalink raw reply related
* Re: [patch net-next 0/3] mlxsw: Introduce support for "pass" gact action offloading
From: David Miller @ 2017-09-27 3:35 UTC (permalink / raw)
To: jiri; +Cc: netdev, idosch, mlxsw, jhs, xiyou.wangcong
In-Reply-To: <20170925085822.1358-1-jiri@resnulli.us>
From: Jiri Pirko <jiri@resnulli.us>
Date: Mon, 25 Sep 2017 10:58:19 +0200
> From: Jiri Pirko <jiri@mellanox.com>
>
> Very simple patchset adds ability for user to insert filters with "pass"
> gact action and offload it. That allows scenarios like this:
>
> $ tc filter add dev enp3s0np19 ingress protocol ip pref 10 flower skip_sw dst_ip 192.168.101.0/24 action drop
> $ tc filter add dev enp3s0np19 ingress protocol ip pref 9 flower skip_sw dst_ip 192.168.101.1 action pass
Series applied, thanks.
^ permalink raw reply
* Re: pull-request: wireless-drivers 2017-09-25
From: David Miller @ 2017-09-27 3:22 UTC (permalink / raw)
To: kvalo; +Cc: linux-wireless, netdev, linux-kernel
In-Reply-To: <878th36ttx.fsf@kamboji.qca.qualcomm.com>
From: Kalle Valo <kvalo@codeaurora.org>
Date: Mon, 25 Sep 2017 11:55:22 +0300
> here a pull request to net for 4.14, more info in the signed tag below.
> Please let me know if there are any problems.
Pulled, thanks Kalle.
^ permalink raw reply
* Re: [PATCH 6/6] net: dsa: mv88e6xxx: Forward broadcast frames to cpu and dsa ports
From: David Miller @ 2017-09-27 3:20 UTC (permalink / raw)
To: andrew; +Cc: vivien.didelot, netdev
In-Reply-To: <1506464764-12699-8-git-send-email-andrew@lunn.ch>
From: Andrew Lunn <andrew@lunn.ch>
Date: Wed, 27 Sep 2017 00:26:04 +0200
> +
> + for (port = 0; port < mv88e6xxx_num_ports(chip); ++port) {
Just a nit pick since you are going to respin this anyways:
for (x; cond(x); x++)
seems to be the canonical way to express this in the kernel, so please
make "++port" into "port++"
Thank you.
^ permalink raw reply
* Re: [PATCH net-next 0/5] net: dsa: use generic slave phydev
From: David Miller @ 2017-09-27 3:06 UTC (permalink / raw)
To: vivien.didelot; +Cc: netdev, linux-kernel, kernel, f.fainelli, andrew
In-Reply-To: <20170926211535.21273-1-vivien.didelot@savoirfairelinux.com>
From: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Date: Tue, 26 Sep 2017 17:15:30 -0400
> DSA currently stores a phy_device pointer in each slave private
> structure. This requires to implement our own ethtool ksettings
> accessors and such.
>
> This patchset removes the private phy_device in favor of the one
> provided in the net_device structure, and thus allows us to use the
> generic phy_ethtool_* functions.
Series applied, thanks Vivien.
^ permalink raw reply
* Re: [patch net-next 00/10] mlxsw: Add router adjacency dpipe table
From: David Miller @ 2017-09-27 3:05 UTC (permalink / raw)
To: jiri; +Cc: netdev, arkadis, idosch, mlxsw
In-Reply-To: <20170925083230.1193-1-jiri@resnulli.us>
From: Jiri Pirko <jiri@resnulli.us>
Date: Mon, 25 Sep 2017 10:32:20 +0200
> From: Jiri Pirko <jiri@mellanox.com>
>
> Arkadi says:
>
> This patchset adds router adjacency dpipe table support. This will provide
> the ability to observe the hardware offloaded IPv4/6 nexthops.
Series applied, thanks everyone.
^ permalink raw reply
* Re: [PATCH] net/ipv4: Update sk_for_each_entry_offset_rcu macro to utilize rcu methods hlist_next_rcu. This fixes the warnings thrown by sparse regarding net/ipv4/udp.c on line 1974.
From: David Miller @ 2017-09-27 3:03 UTC (permalink / raw)
To: devtimhansen; +Cc: netdev, linux-kernel
In-Reply-To: <20170927005406.5935-1-devtimhansen@gmail.com>
From: Tim Hansen <devtimhansen@gmail.com>
Date: Tue, 26 Sep 2017 20:54:05 -0400
> Signed-off-by: Tim Hansen <devtimhansen@gmail.com>
This is a poor patch submission on many levels.
But the main problem, is that there is no use of
sk_for_each_entry_offset_rcu() in any of my networking kernel trees.
Referencing code by line number never works, you have to mention
what version of the kernel, what tree, and where in what fucntion
the problem is occurring.
Secondly, sk_for_each_entry_offset_rcu() is not meant to be used
in _raw() contexts. This is why it's not called
sk_for_each_entry_offset_rcu_raw().
The sparse warning is probably legitimate, and points to a bug.
But nobody can tell where becuase you haven't told us what tree
and where this happens.
^ permalink raw reply
* Re: [PATCH] net: stmmac: Meet alignment requirements for DMA
From: David Miller @ 2017-09-27 2:52 UTC (permalink / raw)
To: paul.burton
Cc: matt.redfearn, netdev, alexandre.torgue, peppe.cavallaro,
linux-kernel, linux-mips, james.hogan
In-Reply-To: <1752163.uYYNevmZpH@np-p-burton>
From: Paul Burton <paul.burton@imgtec.com>
Date: Tue, 26 Sep 2017 14:30:33 -0700
> I'd suggest that at a minimum if you're unwilling to obey the API as
> described in Documentation/DMA-API.txt then it would be beneficial
> if you could propose a change to it such that it works for you, and
> perhaps we can extend the API & its documentation to allow your
> usage whilst also allowing us to catch broken uses.
The networking driver code works fine as is.
I also didn't write that ill-advised documentation in the DMA docs,
nor the non-merged new MIPS assertion.
So I'm trying to figure out on what basis I am required to do
anything.
Thank you.
^ permalink raw reply
* Re: [PATCH net-next RFC 0/5] batched tx processing in vhost_net
From: Jason Wang @ 2017-09-27 2:06 UTC (permalink / raw)
To: Michael S. Tsirkin; +Cc: virtualization, netdev, linux-kernel, kvm
In-Reply-To: <20170926222528-mutt-send-email-mst@kernel.org>
On 2017年09月27日 03:26, Michael S. Tsirkin wrote:
> On Fri, Sep 22, 2017 at 04:02:30PM +0800, Jason Wang wrote:
>> Hi:
>>
>> This series tries to implement basic tx batched processing. This is
>> done by prefetching descriptor indices and update used ring in a
>> batch. This intends to speed up used ring updating and improve the
>> cache utilization. Test shows about ~22% improvement in tx pss.
>
>
>
>> Please review.
>>
>> Jason Wang (5):
>> vhost: split out ring head fetching logic
>> vhost: introduce helper to prefetch desc index
>> vhost: introduce vhost_add_used_idx()
> Please squash these new APIs into where they are used.
> This split-up just makes review harder for me as
> I can't figure out how the new APIs are used.
Ok.
Thanks
>
>
>> vhost_net: rename VHOST_RX_BATCH to VHOST_NET_BATCH
> This is ok as a separate patch.
>
>> vhost_net: basic tx virtqueue batched processing
>>
>> drivers/vhost/net.c | 221 ++++++++++++++++++++++++++++----------------------
>> drivers/vhost/vhost.c | 165 +++++++++++++++++++++++++++++++------
>> drivers/vhost/vhost.h | 9 ++
>> 3 files changed, 270 insertions(+), 125 deletions(-)
>>
>> --
>> 2.7.4
^ permalink raw reply
* Re: [PATCH net-next RFC 5/5] vhost_net: basic tx virtqueue batched processing
From: Jason Wang @ 2017-09-27 2:04 UTC (permalink / raw)
To: Michael S. Tsirkin; +Cc: virtualization, netdev, linux-kernel, kvm
In-Reply-To: <20170926222223-mutt-send-email-mst@kernel.org>
On 2017年09月27日 03:25, Michael S. Tsirkin wrote:
> On Fri, Sep 22, 2017 at 04:02:35PM +0800, Jason Wang wrote:
>> This patch implements basic batched processing of tx virtqueue by
>> prefetching desc indices and updating used ring in a batch. For
>> non-zerocopy case, vq->heads were used for storing the prefetched
>> indices and updating used ring. It is also a requirement for doing
>> more batching on top. For zerocopy case and for simplicity, batched
>> processing were simply disabled by only fetching and processing one
>> descriptor at a time, this could be optimized in the future.
>>
>> XDP_DROP (without touching skb) on tun (with Moongen in guest) with
>> zercopy disabled:
>>
>> Intel(R) Xeon(R) CPU E5-2650 0 @ 2.00GHz:
>> Before: 3.20Mpps
>> After: 3.90Mpps (+22%)
>>
>> No differences were seen with zerocopy enabled.
>>
>> Signed-off-by: Jason Wang <jasowang@redhat.com>
> So where is the speedup coming from? I'd guess the ring is
> hot in cache, it's faster to access it in one go, then
> pass many packets to net stack. Is that right?
>
> Another possibility is better code cache locality.
Yes, I think the speed up comes from:
- less cache misses
- less cache line bounce when virtqueue is about to be full (guest is
faster than host which is the case of MoonGen)
- less memory barriers
- possible faster copy speed by using copy_to_user() on modern CPUs
>
> So how about this patchset is refactored:
>
> 1. use existing APIs just first get packets then
> transmit them all then use them all
Looks like current API can not get packets first, it only support get
packet one by one (if you mean vhost_get_vq_desc()). And used ring
updating may get more misses in this case.
> 2. add new APIs and move the loop into vhost core
> for more speedups
I don't see any advantages, looks like just need some e.g callbacks in
this case.
Thanks
^ permalink raw reply
* [PATCH v3 net-next 09/10] net: hns3: Setting for fc_mode and dcb enable flag in TM module
From: Yunsheng Lin @ 2017-09-27 1:45 UTC (permalink / raw)
To: davem
Cc: huangdaode, xuwei5, liguozhu, Yisen.Zhuang, gabriele.paoloni,
john.garry, linuxarm, yisen.zhuang, salil.mehta, lipeng321,
netdev, linux-kernel
In-Reply-To: <1506476732-128130-1-git-send-email-linyunsheng@huawei.com>
After the DCB feature is supported, fc_mode and dcb enable flag
must be set according to the DCB parameter.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
---
.../net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c | 34 +++++++++++++++++++---
1 file changed, 30 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c
index 8295684..359ee67 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c
@@ -486,7 +486,11 @@ static void hclge_tm_tc_info_init(struct hclge_dev *hdev)
hdev->tm_info.prio_tc[i] =
(i >= hdev->tm_info.num_tc) ? 0 : i;
- hdev->flag &= ~HCLGE_FLAG_DCB_ENABLE;
+ /* DCB is enabled if we have more than 1 TC */
+ if (hdev->tm_info.num_tc > 1)
+ hdev->flag |= HCLGE_FLAG_DCB_ENABLE;
+ else
+ hdev->flag &= ~HCLGE_FLAG_DCB_ENABLE;
}
static void hclge_tm_pg_info_init(struct hclge_dev *hdev)
@@ -512,6 +516,24 @@ static void hclge_tm_pg_info_init(struct hclge_dev *hdev)
}
}
+static void hclge_pfc_info_init(struct hclge_dev *hdev)
+{
+ if (!(hdev->flag & HCLGE_FLAG_DCB_ENABLE)) {
+ if (hdev->fc_mode_last_time == HCLGE_FC_PFC)
+ dev_warn(&hdev->pdev->dev,
+ "DCB is disable, but last mode is FC_PFC\n");
+
+ hdev->tm_info.fc_mode = hdev->fc_mode_last_time;
+ } else if (hdev->tm_info.fc_mode != HCLGE_FC_PFC) {
+ /* fc_mode_last_time record the last fc_mode when
+ * DCB is enabled, so that fc_mode can be set to
+ * the correct value when DCB is disabled.
+ */
+ hdev->fc_mode_last_time = hdev->tm_info.fc_mode;
+ hdev->tm_info.fc_mode = HCLGE_FC_PFC;
+ }
+}
+
static int hclge_tm_schd_info_init(struct hclge_dev *hdev)
{
if ((hdev->tx_sch_mode != HCLGE_FLAG_TC_BASE_SCH_MODE) &&
@@ -524,8 +546,7 @@ static int hclge_tm_schd_info_init(struct hclge_dev *hdev)
hclge_tm_vport_info_update(hdev);
- hdev->tm_info.fc_mode = HCLGE_FC_NONE;
- hdev->fc_mode_last_time = hdev->tm_info.fc_mode;
+ hclge_pfc_info_init(hdev);
return 0;
}
@@ -1158,8 +1179,13 @@ int hclge_tm_init_hw(struct hclge_dev *hdev)
int hclge_tm_schd_init(struct hclge_dev *hdev)
{
- int ret = hclge_tm_schd_info_init(hdev);
+ int ret;
+
+ /* fc_mode is HCLGE_FC_FULL on reset */
+ hdev->tm_info.fc_mode = HCLGE_FC_FULL;
+ hdev->fc_mode_last_time = hdev->tm_info.fc_mode;
+ ret = hclge_tm_schd_info_init(hdev);
if (ret)
return ret;
--
1.9.1
^ permalink raw reply related
* [PATCH v3 net-next 06/10] net: hns3: Add some interface for the support of DCB feature
From: Yunsheng Lin @ 2017-09-27 1:45 UTC (permalink / raw)
To: davem
Cc: huangdaode, xuwei5, liguozhu, Yisen.Zhuang, gabriele.paoloni,
john.garry, linuxarm, yisen.zhuang, salil.mehta, lipeng321,
netdev, linux-kernel
In-Reply-To: <1506476732-128130-1-git-send-email-linyunsheng@huawei.com>
This patch add some interface and export some interface from
hclge_tm and hclgc_main to support the upcoming DCB feature.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
---
.../ethernet/hisilicon/hns3/hns3pf/hclge_main.c | 3 +-
.../ethernet/hisilicon/hns3/hns3pf/hclge_main.h | 3 ++
.../net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c | 48 ++++++++++++++++++++--
.../net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.h | 6 +++
4 files changed, 55 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
index b345070..ca7e3bc 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
@@ -30,7 +30,6 @@
#define HCLGE_64BIT_STATS_FIELD_OFF(f) (offsetof(struct hclge_64_bit_stats, f))
#define HCLGE_32BIT_STATS_FIELD_OFF(f) (offsetof(struct hclge_32_bit_stats, f))
-static int hclge_rss_init_hw(struct hclge_dev *hdev);
static int hclge_set_mta_filter_mode(struct hclge_dev *hdev,
enum hclge_mta_dmac_sel_type mta_mac_sel,
bool enable);
@@ -2655,7 +2654,7 @@ static int hclge_get_tc_size(struct hnae3_handle *handle)
return hdev->rss_size_max;
}
-static int hclge_rss_init_hw(struct hclge_dev *hdev)
+int hclge_rss_init_hw(struct hclge_dev *hdev)
{
const u8 hfunc = HCLGE_RSS_HASH_ALGO_TOEPLITZ;
struct hclge_vport *vport = hdev->vport;
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.h b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.h
index 4fc36f0..394b587 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.h
@@ -515,4 +515,7 @@ static inline int hclge_get_queue_id(struct hnae3_queue *queue)
int hclge_cfg_mac_speed_dup(struct hclge_dev *hdev, int speed, u8 duplex);
int hclge_set_vf_vlan_common(struct hclge_dev *vport, int vfid,
bool is_kill, u16 vlan, u8 qos, __be16 proto);
+
+int hclge_buffer_alloc(struct hclge_dev *hdev);
+int hclge_rss_init_hw(struct hclge_dev *hdev);
#endif
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c
index ea94d23..8295684 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.c
@@ -883,10 +883,14 @@ static int hclge_tm_pri_dwrr_cfg(struct hclge_dev *hdev)
return 0;
}
-static int hclge_tm_map_cfg(struct hclge_dev *hdev)
+int hclge_tm_map_cfg(struct hclge_dev *hdev)
{
int ret;
+ ret = hclge_up_to_tc_map(hdev);
+ if (ret)
+ return ret;
+
ret = hclge_tm_pg_to_pri_map(hdev);
if (ret)
return ret;
@@ -994,7 +998,7 @@ static int hclge_tm_lvl34_schd_mode_cfg(struct hclge_dev *hdev)
return 0;
}
-static int hclge_tm_schd_mode_hw(struct hclge_dev *hdev)
+int hclge_tm_schd_mode_hw(struct hclge_dev *hdev)
{
int ret;
@@ -1092,7 +1096,45 @@ int hclge_pause_setup_hw(struct hclge_dev *hdev)
return ret;
}
- return hclge_up_to_tc_map(hdev);
+ return 0;
+}
+
+int hclge_tm_prio_tc_info_update(struct hclge_dev *hdev, u8 *prio_tc)
+{
+ struct hclge_vport *vport = hdev->vport;
+ struct hnae3_knic_private_info *kinfo;
+ u32 i, k;
+
+ for (i = 0; i < HNAE3_MAX_USER_PRIO; i++) {
+ if (prio_tc[i] >= hdev->tm_info.num_tc)
+ return -EINVAL;
+ hdev->tm_info.prio_tc[i] = prio_tc[i];
+
+ for (k = 0; k < hdev->num_alloc_vport; k++) {
+ kinfo = &vport[k].nic.kinfo;
+ kinfo->prio_tc[i] = prio_tc[i];
+ }
+ }
+ return 0;
+}
+
+void hclge_tm_schd_info_update(struct hclge_dev *hdev, u8 num_tc)
+{
+ u8 i, bit_map = 0;
+
+ hdev->tm_info.num_tc = num_tc;
+
+ for (i = 0; i < hdev->tm_info.num_tc; i++)
+ bit_map |= BIT(i);
+
+ if (!bit_map) {
+ bit_map = 1;
+ hdev->tm_info.num_tc = 1;
+ }
+
+ hdev->hw_tc_map = bit_map;
+
+ hclge_tm_schd_info_init(hdev);
}
int hclge_tm_init_hw(struct hclge_dev *hdev)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.h b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.h
index 19a01e4..bf59961 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_tm.h
@@ -112,4 +112,10 @@ struct hclge_port_shapping_cmd {
int hclge_tm_schd_init(struct hclge_dev *hdev);
int hclge_pause_setup_hw(struct hclge_dev *hdev);
+int hclge_tm_schd_mode_hw(struct hclge_dev *hdev);
+int hclge_tm_prio_tc_info_update(struct hclge_dev *hdev, u8 *prio_tc);
+void hclge_tm_schd_info_update(struct hclge_dev *hdev, u8 num_tc);
+int hclge_tm_dwrr_cfg(struct hclge_dev *hdev);
+int hclge_tm_map_cfg(struct hclge_dev *hdev);
+int hclge_tm_init_hw(struct hclge_dev *hdev);
#endif
--
1.9.1
^ permalink raw reply related
* [PATCH v3 net-next 08/10] net: hns3: Add dcb netlink interface for the support of DCB feature
From: Yunsheng Lin @ 2017-09-27 1:45 UTC (permalink / raw)
To: davem
Cc: huangdaode, xuwei5, liguozhu, Yisen.Zhuang, gabriele.paoloni,
john.garry, linuxarm, yisen.zhuang, salil.mehta, lipeng321,
netdev, linux-kernel
In-Reply-To: <1506476732-128130-1-git-send-email-linyunsheng@huawei.com>
This patch add dcb netlink interface by calling the interface from
hclge_dcb module.
This patch also update Makefile in order to build hns3_dcbnl module.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
---
.../net/ethernet/hisilicon/hns3/hns3pf/Makefile | 2 +
.../ethernet/hisilicon/hns3/hns3pf/hns3_dcbnl.c | 106 +++++++++++++++++++++
.../net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.c | 2 +
.../net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.h | 7 ++
4 files changed, 117 insertions(+)
create mode 100644 drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_dcbnl.c
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/Makefile b/drivers/net/ethernet/hisilicon/hns3/hns3pf/Makefile
index 7023dc87..d2b20d0 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/Makefile
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/Makefile
@@ -11,3 +11,5 @@ hclge-$(CONFIG_HNS3_DCB) += hclge_dcb.o
obj-$(CONFIG_HNS3_ENET) += hns3.o
hns3-objs = hns3_enet.o hns3_ethtool.o
+
+hns3-$(CONFIG_HNS3_DCB) += hns3_dcbnl.o
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_dcbnl.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_dcbnl.c
new file mode 100644
index 0000000..9832172
--- /dev/null
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_dcbnl.c
@@ -0,0 +1,106 @@
+/*
+ * Copyright (c) 2016-2017 Hisilicon Limited.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include "hnae3.h"
+#include "hns3_enet.h"
+
+static
+int hns3_dcbnl_ieee_getets(struct net_device *ndev, struct ieee_ets *ets)
+{
+ struct hns3_nic_priv *priv = netdev_priv(ndev);
+ struct hnae3_handle *h = priv->ae_handle;
+
+ if (h->kinfo.dcb_ops->ieee_getets)
+ return h->kinfo.dcb_ops->ieee_getets(h, ets);
+
+ return -EOPNOTSUPP;
+}
+
+static
+int hns3_dcbnl_ieee_setets(struct net_device *ndev, struct ieee_ets *ets)
+{
+ struct hns3_nic_priv *priv = netdev_priv(ndev);
+ struct hnae3_handle *h = priv->ae_handle;
+
+ if (h->kinfo.dcb_ops->ieee_setets)
+ return h->kinfo.dcb_ops->ieee_setets(h, ets);
+
+ return -EOPNOTSUPP;
+}
+
+static
+int hns3_dcbnl_ieee_getpfc(struct net_device *ndev, struct ieee_pfc *pfc)
+{
+ struct hns3_nic_priv *priv = netdev_priv(ndev);
+ struct hnae3_handle *h = priv->ae_handle;
+
+ if (h->kinfo.dcb_ops->ieee_getpfc)
+ return h->kinfo.dcb_ops->ieee_getpfc(h, pfc);
+
+ return -EOPNOTSUPP;
+}
+
+static
+int hns3_dcbnl_ieee_setpfc(struct net_device *ndev, struct ieee_pfc *pfc)
+{
+ struct hns3_nic_priv *priv = netdev_priv(ndev);
+ struct hnae3_handle *h = priv->ae_handle;
+
+ if (h->kinfo.dcb_ops->ieee_setpfc)
+ return h->kinfo.dcb_ops->ieee_setpfc(h, pfc);
+
+ return -EOPNOTSUPP;
+}
+
+/* DCBX configuration */
+static u8 hns3_dcbnl_getdcbx(struct net_device *ndev)
+{
+ struct hns3_nic_priv *priv = netdev_priv(ndev);
+ struct hnae3_handle *h = priv->ae_handle;
+
+ if (h->kinfo.dcb_ops->getdcbx)
+ return h->kinfo.dcb_ops->getdcbx(h);
+
+ return 0;
+}
+
+/* return 0 if successful, otherwise fail */
+static u8 hns3_dcbnl_setdcbx(struct net_device *ndev, u8 mode)
+{
+ struct hns3_nic_priv *priv = netdev_priv(ndev);
+ struct hnae3_handle *h = priv->ae_handle;
+
+ if (h->kinfo.dcb_ops->setdcbx)
+ return h->kinfo.dcb_ops->setdcbx(h, mode);
+
+ return 1;
+}
+
+static const struct dcbnl_rtnl_ops hns3_dcbnl_ops = {
+ .ieee_getets = hns3_dcbnl_ieee_getets,
+ .ieee_setets = hns3_dcbnl_ieee_setets,
+ .ieee_getpfc = hns3_dcbnl_ieee_getpfc,
+ .ieee_setpfc = hns3_dcbnl_ieee_setpfc,
+ .getdcbx = hns3_dcbnl_getdcbx,
+ .setdcbx = hns3_dcbnl_setdcbx,
+};
+
+/* hclge_dcbnl_setup - DCBNL setup
+ * @handle: the corresponding vport handle
+ * Set up DCBNL
+ */
+void hns3_dcbnl_setup(struct hnae3_handle *handle)
+{
+ struct net_device *dev = handle->kinfo.netdev;
+
+ if (!handle->kinfo.dcb_ops)
+ return;
+
+ dev->dcbnl_ops = &hns3_dcbnl_ops;
+}
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.c
index 35369e1..11dab26 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.c
@@ -2790,6 +2790,8 @@ static int hns3_client_init(struct hnae3_handle *handle)
goto out_reg_netdev_fail;
}
+ hns3_dcbnl_setup(handle);
+
/* MTU range: (ETH_MIN_MTU(kernel default) - 9706) */
netdev->max_mtu = HNS3_MAX_MTU - (ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN);
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.h b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.h
index 7e87461..481eada 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hns3_enet.h
@@ -590,4 +590,11 @@ static inline void hns3_write_reg(void __iomem *base, u32 reg, u32 value)
void hns3_ethtool_set_ops(struct net_device *netdev);
int hns3_clean_tx_ring(struct hns3_enet_ring *ring, int budget);
+
+#ifdef CONFIG_HNS3_DCB
+void hns3_dcbnl_setup(struct hnae3_handle *handle);
+#else
+static inline void hns3_dcbnl_setup(struct hnae3_handle *handle) {}
+#endif
+
#endif
--
1.9.1
^ permalink raw reply related
* [PATCH v3 net-next 01/10] net: hns3: Support for dynamically assigning tx buffer to TC
From: Yunsheng Lin @ 2017-09-27 1:45 UTC (permalink / raw)
To: davem
Cc: huangdaode, xuwei5, liguozhu, Yisen.Zhuang, gabriele.paoloni,
john.garry, linuxarm, yisen.zhuang, salil.mehta, lipeng321,
netdev, linux-kernel
In-Reply-To: <1506476732-128130-1-git-send-email-linyunsheng@huawei.com>
This patch add support of dynamically assigning tx buffer to
TC when the TC is enabled.
It will save buffer for rx direction to avoid packet loss.
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
---
V2:
Fix for not defining variables in local loop.
V1:
Initial submit.
---
.../net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h | 1 +
.../ethernet/hisilicon/hns3/hns3pf/hclge_main.c | 64 ++++++++++++++++++----
2 files changed, 55 insertions(+), 10 deletions(-)
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
index 758cf39..a81c6cb 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
@@ -311,6 +311,7 @@ struct hclge_tc_thrd {
struct hclge_priv_buf {
struct hclge_waterline wl; /* Waterline for low and high*/
u32 buf_size; /* TC private buffer size */
+ u32 tx_buf_size;
u32 enable; /* Enable TC private buffer or not */
};
diff --git a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
index d27618b..02da3d5 100644
--- a/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
+++ b/drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
@@ -1324,7 +1324,7 @@ static int hclge_alloc_vport(struct hclge_dev *hdev)
return 0;
}
-static int hclge_cmd_alloc_tx_buff(struct hclge_dev *hdev, u16 buf_size)
+static int hclge_cmd_alloc_tx_buff(struct hclge_dev *hdev)
{
/* TX buffer size is unit by 128 byte */
#define HCLGE_BUF_SIZE_UNIT_SHIFT 7
@@ -1337,10 +1337,13 @@ static int hclge_cmd_alloc_tx_buff(struct hclge_dev *hdev, u16 buf_size)
req = (struct hclge_tx_buff_alloc *)desc.data;
hclge_cmd_setup_basic_desc(&desc, HCLGE_OPC_TX_BUFF_ALLOC, 0);
- for (i = 0; i < HCLGE_TC_NUM; i++)
+ for (i = 0; i < HCLGE_TC_NUM; i++) {
+ u32 buf_size = hdev->priv_buf[i].tx_buf_size;
+
req->tx_pkt_buff[i] =
cpu_to_le16((buf_size >> HCLGE_BUF_SIZE_UNIT_SHIFT) |
HCLGE_BUF_SIZE_UPDATE_EN_MSK);
+ }
ret = hclge_cmd_send(&hdev->hw, &desc, 1);
if (ret) {
@@ -1352,9 +1355,9 @@ static int hclge_cmd_alloc_tx_buff(struct hclge_dev *hdev, u16 buf_size)
return 0;
}
-static int hclge_tx_buffer_alloc(struct hclge_dev *hdev, u32 buf_size)
+static int hclge_tx_buffer_alloc(struct hclge_dev *hdev)
{
- int ret = hclge_cmd_alloc_tx_buff(hdev, buf_size);
+ int ret = hclge_cmd_alloc_tx_buff(hdev);
if (ret) {
dev_err(&hdev->pdev->dev,
@@ -1433,6 +1436,16 @@ static u32 hclge_get_rx_priv_buff_alloced(struct hclge_dev *hdev)
return rx_priv;
}
+static u32 hclge_get_tx_buff_alloced(struct hclge_dev *hdev)
+{
+ u32 i, total_tx_size = 0;
+
+ for (i = 0; i < HCLGE_MAX_TC_NUM; i++)
+ total_tx_size += hdev->priv_buf[i].tx_buf_size;
+
+ return total_tx_size;
+}
+
static bool hclge_is_rx_buf_ok(struct hclge_dev *hdev, u32 rx_all)
{
u32 shared_buf_min, shared_buf_tc, shared_std;
@@ -1477,18 +1490,43 @@ static bool hclge_is_rx_buf_ok(struct hclge_dev *hdev, u32 rx_all)
return true;
}
+static int hclge_tx_buffer_calc(struct hclge_dev *hdev)
+{
+ u32 i, total_size;
+
+ total_size = hdev->pkt_buf_size;
+
+ /* alloc tx buffer for all enabled tc */
+ for (i = 0; i < HCLGE_MAX_TC_NUM; i++) {
+ struct hclge_priv_buf *priv = &hdev->priv_buf[i];
+
+ if (total_size < HCLGE_DEFAULT_TX_BUF)
+ return -ENOMEM;
+
+ if (hdev->hw_tc_map & BIT(i))
+ priv->tx_buf_size = HCLGE_DEFAULT_TX_BUF;
+ else
+ priv->tx_buf_size = 0;
+
+ total_size -= priv->tx_buf_size;
+ }
+
+ return 0;
+}
+
/* hclge_rx_buffer_calc: calculate the rx private buffer size for all TCs
* @hdev: pointer to struct hclge_dev
- * @tx_size: the allocated tx buffer for all TCs
* @return: 0: calculate sucessful, negative: fail
*/
-int hclge_rx_buffer_calc(struct hclge_dev *hdev, u32 tx_size)
+int hclge_rx_buffer_calc(struct hclge_dev *hdev)
{
- u32 rx_all = hdev->pkt_buf_size - tx_size;
+ u32 rx_all = hdev->pkt_buf_size;
int no_pfc_priv_num, pfc_priv_num;
struct hclge_priv_buf *priv;
int i;
+ rx_all -= hclge_get_tx_buff_alloced(hdev);
+
/* When DCB is not supported, rx private
* buffer is not allocated.
*/
@@ -1771,7 +1809,6 @@ static int hclge_common_wl_config(struct hclge_dev *hdev)
int hclge_buffer_alloc(struct hclge_dev *hdev)
{
- u32 tx_buf_size = HCLGE_DEFAULT_TX_BUF;
int ret;
hdev->priv_buf = devm_kmalloc_array(&hdev->pdev->dev, HCLGE_MAX_TC_NUM,
@@ -1780,14 +1817,21 @@ int hclge_buffer_alloc(struct hclge_dev *hdev)
if (!hdev->priv_buf)
return -ENOMEM;
- ret = hclge_tx_buffer_alloc(hdev, tx_buf_size);
+ ret = hclge_tx_buffer_calc(hdev);
+ if (ret) {
+ dev_err(&hdev->pdev->dev,
+ "could not calc tx buffer size for all TCs %d\n", ret);
+ return ret;
+ }
+
+ ret = hclge_tx_buffer_alloc(hdev);
if (ret) {
dev_err(&hdev->pdev->dev,
"could not alloc tx buffers %d\n", ret);
return ret;
}
- ret = hclge_rx_buffer_calc(hdev, tx_buf_size);
+ ret = hclge_rx_buffer_calc(hdev);
if (ret) {
dev_err(&hdev->pdev->dev,
"could not calc rx priv buffer size for all TCs %d\n",
--
1.9.1
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox