* [PATCH net-next 2/3] net/vlan: introduce __vlan_hwaccel_copy_tag() helper
From: Michał Mirosław @ 2017-01-03 23:53 UTC (permalink / raw)
To: netdev
In-Reply-To: <cover.1483487429.git.mirq-linux@rere.qmqm.pl>
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
include/linux/if_vlan.h | 13 +++++++++++++
1 file changed, 13 insertions(+)
diff --git a/include/linux/if_vlan.h b/include/linux/if_vlan.h
index 38be9041cde4..75e839b84a63 100644
--- a/include/linux/if_vlan.h
+++ b/include/linux/if_vlan.h
@@ -393,6 +393,19 @@ static inline void __vlan_hwaccel_clear_tag(struct sk_buff *skb)
skb->vlan_tci = 0;
}
+/**
+ * __vlan_hwaccel_copy_tag - copy hardware accelerated VLAN info from another skb
+ * @dst: skbuff to copy to
+ * @src: skbuff to copy from
+ *
+ * Copies VLAN information from @src to @dst (for branchless code)
+ */
+static inline void __vlan_hwaccel_copy_tag(struct sk_buff *dst, const struct sk_buff *src)
+{
+ dst->vlan_proto = src->vlan_proto;
+ dst->vlan_tci = src->vlan_tci;
+}
+
/*
* __vlan_hwaccel_push_inside - pushes vlan tag to the payload
* @skb: skbuff to tag
--
2.11.0
^ permalink raw reply related
* [PATCH net-next 0/3] Preparation for VLAN_TAG_PRESENT cleanup
From: Michał Mirosław @ 2017-01-03 23:53 UTC (permalink / raw)
To: netdev
This introduces additional and fixes existing helpers for accessing
skb->vlan_tag.
Michał Mirosław (3):
net/vlan: introduce __vlan_hwaccel_clear_tag() helper
net/vlan: introduce __vlan_hwaccel_copy_tag() helper
net/vlan: include the shift in skb_vlan_tag_get_prio()
include/linux/if_vlan.h | 28 ++++++++++++++++++++++++++--
net/core/flow_dissector.c | 3 +--
2 files changed, 27 insertions(+), 4 deletions(-)
--
2.11.0
^ permalink raw reply
* [PATCH net-next 3/3] net/vlan: include the shift in skb_vlan_tag_get_prio()
From: Michał Mirosław @ 2017-01-03 23:53 UTC (permalink / raw)
To: netdev
In-Reply-To: <cover.1483487429.git.mirq-linux@rere.qmqm.pl>
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
include/linux/if_vlan.h | 2 +-
net/core/flow_dissector.c | 3 +--
2 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/include/linux/if_vlan.h b/include/linux/if_vlan.h
index 75e839b84a63..bc4e8c2533da 100644
--- a/include/linux/if_vlan.h
+++ b/include/linux/if_vlan.h
@@ -81,7 +81,7 @@ static inline bool is_vlan_dev(const struct net_device *dev)
#define skb_vlan_tag_present(__skb) ((__skb)->vlan_tci & VLAN_TAG_PRESENT)
#define skb_vlan_tag_get(__skb) ((__skb)->vlan_tci & ~VLAN_TAG_PRESENT)
#define skb_vlan_tag_get_id(__skb) ((__skb)->vlan_tci & VLAN_VID_MASK)
-#define skb_vlan_tag_get_prio(__skb) ((__skb)->vlan_tci & VLAN_PRIO_MASK)
+#define skb_vlan_tag_get_prio(__skb) (((__skb)->vlan_tci & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT)
/**
* struct vlan_pcpu_stats - VLAN percpu rx/tx stats
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index d6447dc10371..d93768ac39a1 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -295,8 +295,7 @@ bool __skb_flow_dissect(const struct sk_buff *skb,
if (vlan_tag_present) {
key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
- key_vlan->vlan_priority =
- (skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
+ key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
} else {
key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
VLAN_VID_MASK;
--
2.11.0
^ permalink raw reply related
* [PATCH net-next 1/3] net/vlan: introduce __vlan_hwaccel_clear_tag() helper
From: Michał Mirosław @ 2017-01-03 23:53 UTC (permalink / raw)
To: netdev
In-Reply-To: <cover.1483487429.git.mirq-linux@rere.qmqm.pl>
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
include/linux/if_vlan.h | 13 ++++++++++++-
1 file changed, 12 insertions(+), 1 deletion(-)
diff --git a/include/linux/if_vlan.h b/include/linux/if_vlan.h
index 8d5fcd6284ce..38be9041cde4 100644
--- a/include/linux/if_vlan.h
+++ b/include/linux/if_vlan.h
@@ -382,6 +382,17 @@ static inline struct sk_buff *vlan_insert_tag_set_proto(struct sk_buff *skb,
return skb;
}
+/**
+ * __vlan_hwaccel_clear_tag - clear hardware accelerated VLAN info
+ * @skb: skbuff to clear
+ *
+ * Clears the VLAN information from @skb
+ */
+static inline void __vlan_hwaccel_clear_tag(struct sk_buff *skb)
+{
+ skb->vlan_tci = 0;
+}
+
/*
* __vlan_hwaccel_push_inside - pushes vlan tag to the payload
* @skb: skbuff to tag
@@ -396,7 +407,7 @@ static inline struct sk_buff *__vlan_hwaccel_push_inside(struct sk_buff *skb)
skb = vlan_insert_tag_set_proto(skb, skb->vlan_proto,
skb_vlan_tag_get(skb));
if (likely(skb))
- skb->vlan_tci = 0;
+ __vlan_hwaccel_clear_tag(skb);
return skb;
}
--
2.11.0
^ permalink raw reply related
* Re: [PATCH net-next v2 00/27] Allow passing of VLAN CFI bit through network stack
From: Michał Mirosław @ 2017-01-03 23:36 UTC (permalink / raw)
To: David Miller; +Cc: netdev
In-Reply-To: <20170103232151.hq65poblr4bfvc4z@rere.qmqm.pl>
On Wed, Jan 04, 2017 at 12:21:51AM +0100, Michał Mirosław wrote:
> On Tue, Jan 03, 2017 at 04:32:17PM -0500, David Miller wrote:
> > From: Michał Mirosław <mirq-linux@rere.qmqm.pl>
> > Date: Tue, 3 Jan 2017 21:52:33 +0100 (CET)
> >
> > > Dear NetDevs
> > >
> > > This series removes an abuse of VLAN CFI bit in Linux networking stack.
> > > Currently Linux always clears the bit on outgoing traffic and presents
> > > it cleared to userspace (even via AF_PACKET/tcpdump when hw-accelerated).
> > >
> > > This uses a new vlan_present bit in struct skbuff, and removes an assumption
> > > that vlan_proto != 0 when VLAN tag is present.
> > >
> > > As I can't test most of the driver changes, please look at them carefully.
> > >
> > > The series is supposed to be bisect-friendly and that requires temporary
> > > insertion of #define VLAN_TAG_PRESENT in BPF code to be able to split
> > > JIT changes per architecture.
> > >
> > > Best Regards,
> > > Michał Mirosław
> > >
> > > v2: rebase onto net-next
> >
> > This patch series is really way too large.
> >
> > You're going to have to find a way to combine related changes, or submit
> > this as a series of logical sets, one at a time.
>
> The dependency graph is really sparse: main patch 25 depends on all previous,
> 19-23 all depend only on 18 and 26 depends on 25.
>
> That's it. So the question is: how would it be easier for you to manage?
Ah, I forgot about patches 1-2 that most depend on... So if you get those two
early, then most others can be split.
Best Regards,
Michał Mirosław
^ permalink raw reply
* [PATCH] tc: make tc linking depend on libtc.a
From: David Michael @ 2017-01-03 23:32 UTC (permalink / raw)
To: netdev
There was a race condition where the command to link the tc binary
could (rarely) run before the libtc.a archive existed.
---
Hi,
I've hit a random failure during parallel builds:
ld: cannot find -ltc
The issue seems to be that both tc and libtc.a are permitted to build in
parallel, when tc actually depends on libtc.a. This patch adjusts the
Make prerequisites accordingly. Can something like this be applied?
Thanks.
David
tc/Makefile | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/tc/Makefile b/tc/Makefile
index bb90114..7fd0c4a 100644
--- a/tc/Makefile
+++ b/tc/Makefile
@@ -96,7 +96,7 @@ ifneq ($(TC_CONFIG_NO_XT),y)
endif
TCOBJ += $(TCMODULES)
-LDLIBS += -L. -ltc -lm
+LDLIBS += -L. -lm
ifeq ($(SHARED_LIBS),y)
LDLIBS += -ldl
@@ -124,13 +124,13 @@ MODDESTDIR := $(DESTDIR)$(LIBDIR)/tc
$(QUIET_CC)$(CC) $(CFLAGS) $(LDFLAGS) -shared -fpic $< -o $@
-all: libtc.a tc $(TCSO)
+all: tc $(TCSO)
-tc: $(TCOBJ) $(TCLIB)
+tc: $(TCOBJ) libtc.a
$(QUIET_LINK)$(CC) $^ $(LDFLAGS) $(LDLIBS) -o $@
libtc.a: $(TCLIB)
- $(QUIET_AR)$(AR) rcs $@ $(TCLIB)
+ $(QUIET_AR)$(AR) rcs $@ $^
install: all
mkdir -p $(MODDESTDIR)
--
2.7.4
^ permalink raw reply related
* [PATCH net-next 2/2] tools: psock_tpacket: verify that packet was received on lo before counting it
From: Sowmini Varadhan @ 2017-01-03 23:27 UTC (permalink / raw)
To: linux-kselftest, netdev, sowmini.varadhan; +Cc: daniel, willemb, davem, shuah
In-Reply-To: <cover.1483482971.git.sowmini.varadhan@oracle.com>
Packets from any/all interfaces may be queued up on the PF_PACKET socket
before it is bound to the loopback interface by psock_tpacket, and
when these are passed up by the kernel, they should not be counted
toward the conditions needed to pass/fail the Rx tests.
psock_tpacket discards these packets by examining the sll_ifindex sent
up in each frame and ensuring that this is the same as the ifindex
that was used in bind_ring()
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
---
tools/testing/selftests/net/psock_tpacket.c | 25 ++++++++++++++++++++-----
1 files changed, 20 insertions(+), 5 deletions(-)
diff --git a/tools/testing/selftests/net/psock_tpacket.c b/tools/testing/selftests/net/psock_tpacket.c
index 4a1bc64..3c3f1aa 100644
--- a/tools/testing/selftests/net/psock_tpacket.c
+++ b/tools/testing/selftests/net/psock_tpacket.c
@@ -235,6 +235,7 @@ static void walk_v1_v2_rx(int sock, struct ring *ring)
int udp_sock[2];
union frame_map ppd;
unsigned int frame_num = 0;
+ int ifindex = ring->ll.sll_ifindex;
bug_on(ring->type != PACKET_RX_RING);
@@ -255,12 +256,18 @@ static void walk_v1_v2_rx(int sock, struct ring *ring)
switch (ring->version) {
case TPACKET_V1:
+ if (ppd.v1->s_ll.sll_ifindex != ifindex)
+ goto skip;
+
test_payload((uint8_t *) ppd.raw + ppd.v1->tp_h.tp_mac,
ppd.v1->tp_h.tp_snaplen);
total_bytes += ppd.v1->tp_h.tp_snaplen;
break;
case TPACKET_V2:
+ if (ppd.v2->s_ll.sll_ifindex != ifindex)
+ goto skip;
+
test_payload((uint8_t *) ppd.raw + ppd.v2->tp_h.tp_mac,
ppd.v2->tp_h.tp_snaplen);
total_bytes += ppd.v2->tp_h.tp_snaplen;
@@ -270,6 +277,7 @@ static void walk_v1_v2_rx(int sock, struct ring *ring)
status_bar_update();
total_packets++;
+skip:
__v1_v2_rx_user_ready(ppd.raw, ring->version);
frame_num = (frame_num + 1) % ring->rd_num;
@@ -553,11 +561,13 @@ static void __v3_test_block_header(struct block_desc *pbd, const int block_num)
__v3_test_block_seq_num(pbd);
}
-static void __v3_walk_block(struct block_desc *pbd, const int block_num)
+static void __v3_walk_block(struct block_desc *pbd, const int block_num,
+ int ifindex)
{
int num_pkts = pbd->h1.num_pkts, i;
unsigned long bytes = 0, bytes_with_padding = ALIGN_8(sizeof(*pbd));
struct tpacket3_hdr *ppd;
+ struct sockaddr_ll *s_ll;
__v3_test_block_header(pbd, block_num);
@@ -572,10 +582,15 @@ static void __v3_walk_block(struct block_desc *pbd, const int block_num)
else
bytes_with_padding += ALIGN_8(ppd->tp_snaplen + ppd->tp_mac);
- test_payload((uint8_t *) ppd + ppd->tp_mac, ppd->tp_snaplen);
+ s_ll = (struct sockaddr_ll *)&ppd[1];
- status_bar_update();
- total_packets++;
+ if (ifindex == s_ll->sll_ifindex) {
+ test_payload((uint8_t *) ppd + ppd->tp_mac,
+ ppd->tp_snaplen);
+
+ status_bar_update();
+ total_packets++;
+ }
ppd = (struct tpacket3_hdr *) ((uint8_t *) ppd + ppd->tp_next_offset);
__sync_synchronize();
@@ -616,7 +631,7 @@ static void walk_v3_rx(int sock, struct ring *ring)
while ((pbd->h1.block_status & TP_STATUS_USER) == 0)
poll(&pfd, 1, 1);
- __v3_walk_block(pbd, block_num);
+ __v3_walk_block(pbd, block_num, ring->ll.sll_ifindex);
__v3_flush_block(pbd);
block_num = (block_num + 1) % ring->rd_num;
--
1.7.1
^ permalink raw reply related
* [PATCH net-next 1/2] tools: psock_lib: tighten conditions checked in sock_setfilter
From: Sowmini Varadhan @ 2017-01-03 23:27 UTC (permalink / raw)
To: linux-kselftest, netdev, sowmini.varadhan; +Cc: daniel, willemb, davem, shuah
In-Reply-To: <cover.1483482971.git.sowmini.varadhan@oracle.com>
The bpf_prog used in sock_setfilter() only attempts to check for
ip pktlen, and verifies that the contents of the 80'th packet in
the ethernet frame is 'a' or 'b'. Offsets used for checking these
conditions are incorrectly computed. Thus many non-udp packets
could incorrectly pass through this filter and cause the test to
fail.
This commit tightens the conditions checked by the filter so
that only UDP/IPv4 packets with the matching length and test-character
will be permitted by the filter. The filter has been cleaned up
to explicitly use the BPF macros to make it more readable.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
---
tools/testing/selftests/net/psock_lib.h | 28 +++++++++++++++++++++-------
1 files changed, 21 insertions(+), 7 deletions(-)
diff --git a/tools/testing/selftests/net/psock_lib.h b/tools/testing/selftests/net/psock_lib.h
index 24bc7ec..e62540e 100644
--- a/tools/testing/selftests/net/psock_lib.h
+++ b/tools/testing/selftests/net/psock_lib.h
@@ -27,6 +27,7 @@
#include <string.h>
#include <arpa/inet.h>
#include <unistd.h>
+#include <netinet/udp.h>
#define DATA_LEN 100
#define DATA_CHAR 'a'
@@ -40,14 +41,27 @@
static __maybe_unused void sock_setfilter(int fd, int lvl, int optnum)
{
+ uint16_t ip_len = DATA_LEN +
+ sizeof(struct iphdr) + sizeof(struct udphdr);
+ /* the filter below checks for all of the following conditions that
+ * are based on the contents of create_payload()
+ * ether type 0x800 and
+ * ip proto udp and
+ * ip len == ip_len and
+ * udp[38] == 'a' or udp[38] == 'b'
+ */
struct sock_filter bpf_filter[] = {
- { 0x80, 0, 0, 0x00000000 }, /* LD pktlen */
- { 0x35, 0, 4, DATA_LEN }, /* JGE DATA_LEN [f goto nomatch]*/
- { 0x30, 0, 0, 0x00000050 }, /* LD ip[80] */
- { 0x15, 1, 0, DATA_CHAR }, /* JEQ DATA_CHAR [t goto match]*/
- { 0x15, 0, 1, DATA_CHAR_1}, /* JEQ DATA_CHAR_1 [t goto match]*/
- { 0x06, 0, 0, 0x00000060 }, /* RET match */
- { 0x06, 0, 0, 0x00000000 }, /* RET no match */
+ BPF_STMT(BPF_LD | BPF_H | BPF_ABS, 12), /* LD ethertype */
+ BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, ETH_P_IP, 0, 8),
+ BPF_STMT(BPF_LD|BPF_B|BPF_ABS, 23), /* LD ip_proto */
+ BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, IPPROTO_UDP, 0, 6),
+ BPF_STMT(BPF_LD|BPF_H|BPF_ABS, 16), /* LD ip_len */
+ BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, ip_len, 0, 4),
+ BPF_STMT(BPF_LD|BPF_B|BPF_ABS, 80), /* LD udp[38] */
+ BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, DATA_CHAR, 1, 0),
+ BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, DATA_CHAR_1, 0, 1),
+ BPF_STMT(BPF_RET | BPF_K, ~0), /* match */
+ BPF_STMT(BPF_RET | BPF_K, 0) /* no match */
};
struct sock_fprog bpf_prog;
--
1.7.1
^ permalink raw reply related
* [PATCH net-next 0/2] tools: psock_tpacket bug fixes
From: Sowmini Varadhan @ 2017-01-03 23:27 UTC (permalink / raw)
To: linux-kselftest, netdev, sowmini.varadhan; +Cc: daniel, willemb, davem, shuah
This patchset includes fixes to psock_tpacket for false-negatives
sporadically reported by the test when it was run concurrently with
other heavy network traffic (e.g., over an ssh session, as opposed
to running the test from the console of the test machine). The
test sometimes failed with errors reporting more recvd packets than
expected (e.g., "walk_v0_rx: received 201 out of 100 pkts") or
the reception of non-IP packets (e.g., ARP packets).
There are 2 problems with the test:
1. set_sockfilter() has many bugs (e.g. it will allow TCP packets
that may happen to match on the length field and happen to have
the 'a' or 'b' at the location tested by the filter). In addition,
it is very hard to read, making these bugs harder to find.
2. There is a race-window between packet_create() and packet_do_bind()
in which packets from any interface (e.g., eth0) will get queued
for Rx on the test socket. These packets should not be counted
toward the test.
Patch 1 fixes the first issue by cleaing up set_sockfilter() and
hardening it to make sure that it only permits UDP/IPv4 packets.
Patch 2 fixes the second issue by checking the Rx ifindex passed up
with the tpacket rx frame and making sure that packets counted
toward the test are those received on lo.
Sowmini Varadhan (2):
tools: tighten conditions checked in sock_setfilter
tools: psock_tpacket: verify that packet was received on lo before
counting it
tools/testing/selftests/net/psock_lib.h | 28 ++++++++++++++++++++------
tools/testing/selftests/net/psock_tpacket.c | 25 +++++++++++++++++++----
2 files changed, 41 insertions(+), 12 deletions(-)
^ permalink raw reply
* Re: [PATCH net-next v2 00/27] Allow passing of VLAN CFI bit through network stack
From: Michał Mirosław @ 2017-01-03 23:21 UTC (permalink / raw)
To: David Miller; +Cc: netdev
In-Reply-To: <20170103.163217.697324652959765171.davem@davemloft.net>
On Tue, Jan 03, 2017 at 04:32:17PM -0500, David Miller wrote:
> From: Michał Mirosław <mirq-linux@rere.qmqm.pl>
> Date: Tue, 3 Jan 2017 21:52:33 +0100 (CET)
>
> > Dear NetDevs
> >
> > This series removes an abuse of VLAN CFI bit in Linux networking stack.
> > Currently Linux always clears the bit on outgoing traffic and presents
> > it cleared to userspace (even via AF_PACKET/tcpdump when hw-accelerated).
> >
> > This uses a new vlan_present bit in struct skbuff, and removes an assumption
> > that vlan_proto != 0 when VLAN tag is present.
> >
> > As I can't test most of the driver changes, please look at them carefully.
> >
> > The series is supposed to be bisect-friendly and that requires temporary
> > insertion of #define VLAN_TAG_PRESENT in BPF code to be able to split
> > JIT changes per architecture.
> >
> > Best Regards,
> > Michał Mirosław
> >
> > v2: rebase onto net-next
>
> This patch series is really way too large.
>
> You're going to have to find a way to combine related changes, or submit
> this as a series of logical sets, one at a time.
The dependency graph is really sparse: main patch 25 depends on all previous,
19-23 all depend only on 18 and 26 depends on 25.
That's it. So the question is: how would it be easier for you to manage?
Best Regards,
Michał Mirosław
^ permalink raw reply
* [PATCH net-next v3 10/27] benet: use __vlan_hwaccel helpers
From: Michał Mirosław @ 2017-01-03 23:11 UTC (permalink / raw)
To: netdev; +Cc: Ajit Khaparde
In-Reply-To: <CACZ4nhsUxWYvM5HoASHb7-m2uZtnk3DN6cQigp+cObyLqPJXdA@mail.gmail.com>
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
v3: fix adapter->pvid handling
drivers/net/ethernet/emulex/benet/be_main.c | 13 +++++++++----
1 file changed, 9 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/emulex/benet/be_main.c b/drivers/net/ethernet/emulex/benet/be_main.c
index 7e1633bf5a22..cd12c9a7664b 100644
--- a/drivers/net/ethernet/emulex/benet/be_main.c
+++ b/drivers/net/ethernet/emulex/benet/be_main.c
@@ -1035,30 +1035,35 @@ static struct sk_buff *be_insert_vlan_in_pkt(struct be_adapter *adapter,
struct be_wrb_params
*wrb_params)
{
+ bool insert_vlan = false;
u16 vlan_tag = 0;
skb = skb_share_check(skb, GFP_ATOMIC);
if (unlikely(!skb))
return skb;
- if (skb_vlan_tag_present(skb))
+ if (skb_vlan_tag_present(skb)) {
vlan_tag = be_get_tx_vlan_tag(adapter, skb);
+ insert_vlan = true;
+ }
if (qnq_async_evt_rcvd(adapter) && adapter->pvid) {
- if (!vlan_tag)
+ if (!insert_vlan) {
vlan_tag = adapter->pvid;
+ insert_vlan = true;
+ }
/* f/w workaround to set skip_hw_vlan = 1, informs the F/W to
* skip VLAN insertion
*/
BE_WRB_F_SET(wrb_params->features, VLAN_SKIP_HW, 1);
}
- if (vlan_tag) {
+ if (insert_vlan) {
skb = vlan_insert_tag_set_proto(skb, htons(ETH_P_8021Q),
vlan_tag);
if (unlikely(!skb))
return skb;
- skb->vlan_tci = 0;
+ __vlan_hwaccel_clear_tag(skb);
}
/* Insert the outer VLAN, if any */
--
2.11.0
^ permalink raw reply related
* Re: [PATCH] drop_monitor: consider inserted data in genlmsg_end
From: Wolfgang Reiter @ 2017-01-03 23:09 UTC (permalink / raw)
To: Neil Horman, David Miller; +Cc: netdev, linux-kernel
In-Reply-To: <20170103160443.GC11735@hmsreliant.think-freely.org>
Yes, genlmsg_end changes nlmsg_len field dependent on skb->tail.
After allocation in reset_per_cpu_data skb->tail is modified in
trace_drop_common via __nla_reserve_nohdr.
Best place for setting nlmsg_len to its final value is after being
swapped out in reset_per_cpu_data.
Neil Horman <nhorman@tuxdriver.com> writes:
> On Tue, Jan 03, 2017 at 09:54:19AM -0500, David Miller wrote:
>> From: Reiter Wolfgang <wr0112358@gmail.com>
>> Date: Tue, 3 Jan 2017 01:39:10 +0100
>>
>> > Final nlmsg_len field update must reflect inserted net_dm_drop_point
>> > data.
>> >
>> > This patch depends on previous patch:
>> > "drop_monitor: add missing call to genlmsg_end"
>> >
>> > Signed-off-by: Reiter Wolfgang <wr0112358@gmail.com>
>>
>> I don't understand why the current code doesn't work properly.
>>
>> All over the tree, the pattern is:
>>
>> x = genlmsg_put(skb, ...);
>> ...
>> genlmsg_end(skb, x);
>>
>> And that is exactly what the code is doing right now.
>>
>
> Because reset_per_cpu_data should close the use of of the established skb
> that was being written to. Without this patch we add the END tlv to the skb
> that is just getting started for use in the drop monitor, rather than for the
> skb that is getting returned for use in sending up to user space listeners.
>
> Or am I missing something?
^ permalink raw reply
* Re: [net-next PATCH 5/6] i40e: Add TX and RX support in switchdev mode.
From: Or Gerlitz @ 2017-01-03 23:03 UTC (permalink / raw)
To: Samudrala, Sridhar
Cc: Alexander Duyck, John Fastabend, Anjali Singhai Jain,
jakub.kicinski, intel-wired-lan, Linux Netdev List
In-Reply-To: <586693A6.5090105@intel.com>
On Fri, Dec 30, 2016 at 7:04 PM, Samudrala, Sridhar
<sridhar.samudrala@intel.com> wrote:
> On 12/30/2016 7:31 AM, Or Gerlitz wrote:
>> Are you exposing switchdev ops for the representators? didn't see that
>> or maybe it's in the 4th patch which didn't make it to the list?
> Not at this time. In the future patches when we offload fdb/vlan
> functionality, we could use switchdev ops.
but wait, this is the switchdev mode... even before doing any
offloading, you want (need) your representor netdevices to have the
same HW ID marking they are all ports of the same ASIC, this you can
do with the switchdev parent ID attribute.
^ permalink raw reply
* Re: [PATCH 2/2] isdn: i4l: move active-isdn drivers to staging
From: Arnd Bergmann @ 2017-01-03 23:03 UTC (permalink / raw)
To: Paul Bolle
Cc: devel, Karsten Keil, Jonathan Corbet, Greg Kroah-Hartman,
linux-doc, linux-kernel, netdev, David S. Miller,
linux-arm-kernel
In-Reply-To: <1483484256.13649.13.camel@tiscali.nl>
On Tuesday, January 3, 2017 11:57:36 PM CET Paul Bolle wrote:
> On Tue, 2017-01-03 at 23:25 +0100, Arnd Bergmann wrote:
> > As far as I'm concerned, we are totally fine as long as there exists a
> > longterm supported kernel that has i4l in drivers/staging.
>
> Or in drivers/isdn, right?
Right, I was assuming that we would first move it to staging and then
delete it, both at future points in time that we can debate. With the
existing longterm kernels that have i4l in drivers/isdn, the few remaining
users still have access to a supported kernel release until at least
2020.
Arnd
^ permalink raw reply
* Re: [PATCH 2/2] isdn: i4l: move active-isdn drivers to staging
From: Paul Bolle @ 2017-01-03 22:57 UTC (permalink / raw)
To: Arnd Bergmann
Cc: Greg Kroah-Hartman, devel, Karsten Keil, linux-doc, netdev,
Jonathan Corbet, linux-kernel, David S. Miller, linux-arm-kernel
In-Reply-To: <3780968.pTT7cvIH4p@wuerfel>
On Tue, 2017-01-03 at 23:25 +0100, Arnd Bergmann wrote:
> As far as I'm concerned, we are totally fine as long as there exists a
> longterm supported kernel that has i4l in drivers/staging.
Or in drivers/isdn, right?
Paul Bolle
^ permalink raw reply
* Re: [PATCH net-next 2/3] net: dsa: move HWMON support to its own file
From: kbuild test robot @ 2017-01-03 22:33 UTC (permalink / raw)
To: Vivien Didelot
Cc: kbuild-all, netdev, linux-kernel, kernel, David S. Miller,
Florian Fainelli, Andrew Lunn, cphealy, Vivien Didelot
In-Reply-To: <20170103181538.2593-3-vivien.didelot@savoirfairelinux.com>
[-- Attachment #1: Type: text/plain, Size: 1793 bytes --]
Hi Vivien,
[auto build test WARNING on net-next/master]
url: https://github.com/0day-ci/linux/commits/Vivien-Didelot/net-dsa-restore-HWMON-support-in-dsa2/20170104-055351
config: i386-randconfig-x005-201701 (attached as .config)
compiler: gcc-6 (Debian 6.2.0-3) 6.2.0 20160901
reproduce:
# save the attached .config to linux build tree
make ARCH=i386
All warnings (new ones prefixed by >>):
In file included from net/dsa/dsa.c:26:0:
>> net/dsa/dsa_priv.h:61:5: warning: "CONFIG_NET_DSA_HWMON" is not defined [-Wundef]
#if CONFIG_NET_DSA_HWMON
^~~~~~~~~~~~~~~~~~~~
vim +/CONFIG_NET_DSA_HWMON +61 net/dsa/dsa_priv.h
45 struct net_device *bridge_dev;
46 #ifdef CONFIG_NET_POLL_CONTROLLER
47 struct netpoll *netpoll;
48 #endif
49 };
50
51 /* dsa.c */
52 extern char dsa_driver_version[];
53 int dsa_cpu_dsa_setup(struct dsa_switch *ds, struct device *dev,
54 struct device_node *port_dn, int port);
55 void dsa_cpu_dsa_destroy(struct device_node *port_dn);
56 const struct dsa_device_ops *dsa_resolve_tag_protocol(int tag_protocol);
57 int dsa_cpu_port_ethtool_setup(struct dsa_switch *ds);
58 void dsa_cpu_port_ethtool_restore(struct dsa_switch *ds);
59
60 /* hwmon.c */
> 61 #if CONFIG_NET_DSA_HWMON
62 void dsa_hwmon_register(struct dsa_switch *ds);
63 void dsa_hwmon_unregister(struct dsa_switch *ds);
64 #else
65 static inline void dsa_hwmon_register(struct dsa_switch *ds) { }
66 static inline void dsa_hwmon_unregister(struct dsa_switch *ds) { }
67 #endif
68
69 /* slave.c */
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 37536 bytes --]
^ permalink raw reply
* Re: [PATCH 2/2] isdn: i4l: move active-isdn drivers to staging
From: Arnd Bergmann @ 2017-01-03 22:25 UTC (permalink / raw)
To: Paul Bolle
Cc: Greg Kroah-Hartman, devel, Karsten Keil, linux-doc, netdev,
Jonathan Corbet, linux-kernel, David S. Miller, linux-arm-kernel
In-Reply-To: <1483480459.10534.8.camel@tiscali.nl>
On Tuesday, January 3, 2017 10:54:19 PM CET Paul Bolle wrote:
> On Tue, 2017-01-03 at 22:19 +0100, Arnd Bergmann wrote:
> > isdn: move isdnhdlc out of i4l
> > isdn: i4l: move hisax driver to staging
> > isdn: move i4l to staging
> >
> > I can post those as well, at least I think the first two are helpful
> > for untangling i4l from the rest of ISDN. I also still think that
> > moving hisax and i4l to staging is reasonable given the state of
> > that code, even if there are a couple of users today.
>
> There are? And even if there are: is there any reason to expect that moving
> the rest of i4l to staging will result in anything other than a stream of
> checkpatch cleanups?
To clarify: Karsten's concern was about the loss of features that are
present in i4l but not in mISDN. There were active users of those features
last year, so I assumed that there are still a few this year. However,
whether any of those users would ever need to move to a 4.11 kernel or
newer is an entirely different question.
As far as I'm concerned, we are totally fine as long as there exists a
longterm supported kernel that has i4l in drivers/staging. If we move
i4l to staging for v4.11 with the intention of removing it after the
2018 longterm release (i.e. after Deutsche Telekom turns off their
ISDN network), that gives us at least until 2020. I assume there will
be at least one older kernel with a longer end-of-support date.
> How often did a bunch of drivers re-enter the tree after being sent to
> staging?
Greg can probably answer that. I'm sure it's either never or very rare.
The only case of removed code coming back later is arch/h8300, which
was removed in 2013 and replaced with a much nicer implementation
in 2015.
Arnd
^ permalink raw reply
* Re: [RFC PATCH] virtio_net: XDP support for adjust_head
From: Michael S. Tsirkin @ 2017-01-03 22:16 UTC (permalink / raw)
To: Jason Wang
Cc: John Fastabend, john.r.fastabend, netdev, alexei.starovoitov,
daniel
In-Reply-To: <73715f7a-eeeb-679f-a7b8-7b1fefe1757e@redhat.com>
On Tue, Jan 03, 2017 at 02:01:27PM +0800, Jason Wang wrote:
>
>
> On 2017年01月03日 03:44, John Fastabend wrote:
> > Add support for XDP adjust head by allocating a 256B header region
> > that XDP programs can grow into. This is only enabled when a XDP
> > program is loaded.
> >
> > In order to ensure that we do not have to unwind queue headroom push
> > queue setup below bpf_prog_add. It reads better to do a prog ref
> > unwind vs another queue setup call.
> >
> > : There is a problem with this patch as is. When xdp prog is loaded
> > the old buffers without the 256B headers need to be flushed so that
> > the bpf prog has the necessary headroom. This patch does this by
> > calling the virtqueue_detach_unused_buf() and followed by the
> > virtnet_set_queues() call to reinitialize the buffers. However I
> > don't believe this is safe per comment in virtio_ring this API
> > is not valid on an active queue and the only thing we have done
> > here is napi_disable/napi_enable wrappers which doesn't do anything
> > to the emulation layer.
> >
> > So the RFC is really to find the best solution to this problem.
> > A couple things come to mind, (a) always allocate the necessary
> > headroom but this is a bit of a waste (b) add some bit somewhere
> > to check if the buffer has headroom but this would mean XDP programs
> > would be broke for a cycle through the ring, (c) figure out how
> > to deactivate a queue, free the buffers and finally reallocate.
> > I think (c) is the best choice for now but I'm not seeing the
> > API to do this so virtio/qemu experts anyone know off-hand
> > how to make this work? I started looking into the PCI callbacks
> > reset() and virtio_device_ready() or possibly hitting the right
> > set of bits with vp_set_status() but my first attempt just hung
> > the device.
>
> Hi John:
>
> AFAIK, disabling a specific queue was supported only by virtio 1.0 through
> queue_enable field in pci common cfg.
In fact 1.0 only allows enabling queues selectively.
We can add disabling by a spec enhancement but
for now reset is the only way.
> But unfortunately, qemu does not
> emulate this at all and legacy device does not even support this. So the
> safe way is probably reset the device and redo the initialization here.
You will also have to re-apply rx filtering if you do this.
Probably sending notification uplink.
> >
> > Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
> > ---
> > drivers/net/virtio_net.c | 106 +++++++++++++++++++++++++++++++++++-----------
> > 1 file changed, 80 insertions(+), 26 deletions(-)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 5deeda6..fcc5bd7 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -159,6 +159,9 @@ struct virtnet_info {
> > /* Ethtool settings */
> > u8 duplex;
> > u32 speed;
> > +
> > + /* Headroom allocated in RX Queue */
> > + unsigned int headroom;
> > };
> > struct padded_vnet_hdr {
> > @@ -355,6 +358,7 @@ static void virtnet_xdp_xmit(struct virtnet_info *vi,
> > }
> > if (vi->mergeable_rx_bufs) {
> > + xdp->data -= sizeof(struct virtio_net_hdr_mrg_rxbuf);
> > /* Zero header and leave csum up to XDP layers */
> > hdr = xdp->data;
> > memset(hdr, 0, vi->hdr_len);
> > @@ -371,7 +375,7 @@ static void virtnet_xdp_xmit(struct virtnet_info *vi,
> > num_sg = 2;
> > sg_init_table(sq->sg, 2);
> > sg_set_buf(sq->sg, hdr, vi->hdr_len);
> > - skb_to_sgvec(skb, sq->sg + 1, 0, skb->len);
> > + skb_to_sgvec(skb, sq->sg + 1, vi->headroom, xdp->data_end - xdp->data);
>
> vi->headroom look suspicious, should it be xdp->data - xdp->data_hard_start?
>
> > }
> > err = virtqueue_add_outbuf(sq->vq, sq->sg, num_sg,
> > data, GFP_ATOMIC);
> > @@ -393,34 +397,39 @@ static u32 do_xdp_prog(struct virtnet_info *vi,
> > struct bpf_prog *xdp_prog,
> > void *data, int len)
> > {
> > - int hdr_padded_len;
> > struct xdp_buff xdp;
> > - void *buf;
> > unsigned int qp;
> > u32 act;
> > +
> > if (vi->mergeable_rx_bufs) {
> > - hdr_padded_len = sizeof(struct virtio_net_hdr_mrg_rxbuf);
> > - xdp.data = data + hdr_padded_len;
> > + int desc_room = sizeof(struct virtio_net_hdr_mrg_rxbuf);
> > +
> > + /* Allow consuming headroom but reserve enough space to push
> > + * the descriptor on if we get an XDP_TX return code.
> > + */
> > + xdp.data_hard_start = data - vi->headroom + desc_room;
> > + xdp.data = data + desc_room;
> > xdp.data_end = xdp.data + (len - vi->hdr_len);
> > - buf = data;
> > } else { /* small buffers */
> > struct sk_buff *skb = data;
> > - xdp.data = skb->data;
> > + xdp.data_hard_start = skb->data;
> > + xdp.data = skb->data + vi->headroom;
> > xdp.data_end = xdp.data + len;
> > - buf = skb->data;
> > }
> > act = bpf_prog_run_xdp(xdp_prog, &xdp);
> > switch (act) {
> > case XDP_PASS:
> > + if (!vi->mergeable_rx_bufs)
> > + __skb_pull((struct sk_buff *) data,
> > + xdp.data - xdp.data_hard_start);
>
> Instead of doing things here and virtnet_xdp_xmit(). How about always making
> skb->data point to the buffer head like:
>
> 1) reserve headroom in add_recvbuf_small()
> 2) skb_push(xdp->data - xdp_data_hard_start, skb) if we detect xdp->data was
> modified afer bpf_prog_run_xdp()
>
> Then there's no special code in either XDP_PASS or XDP_TX?
>
> > return XDP_PASS;
> > case XDP_TX:
> > qp = vi->curr_queue_pairs -
> > vi->xdp_queue_pairs +
> > smp_processor_id();
>
> [...]
>
> > +#define VIRTIO_XDP_HEADROOM 256
> > +
> > static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog)
> > {
> > unsigned long int max_sz = PAGE_SIZE - sizeof(struct padded_vnet_hdr);
> > struct virtnet_info *vi = netdev_priv(dev);
> > struct bpf_prog *old_prog;
> > u16 xdp_qp = 0, curr_qp;
> > + unsigned int old_hr;
> > int i, err;
> > if (virtio_has_feature(vi->vdev, VIRTIO_NET_F_GUEST_TSO4) ||
> > @@ -1736,19 +1751,58 @@ static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog)
> > return -ENOMEM;
> > }
> > + old_hr = vi->headroom;
> > + if (prog) {
> > + prog = bpf_prog_add(prog, vi->max_queue_pairs - 1);
> > + if (IS_ERR(prog))
> > + return PTR_ERR(prog);
> > + vi->headroom = VIRTIO_XDP_HEADROOM;
> > + } else {
> > + vi->headroom = 0;
> > + }
> > +
> > + /* Changing the headroom in buffers is a disruptive operation because
> > + * existing buffers must be flushed and reallocated. This will happen
> > + * when a xdp program is initially added or xdp is disabled by removing
> > + * the xdp program.
> > + */
>
> We probably need reset the device here, but maybe Michale has more ideas.
> And if we do this, another interesting thing to do is to disable EWMA and
> always use a single page for each packet, this could almost eliminate
> linearizing.
>
> Thanks
^ permalink raw reply
* [for-next 06/10] net/mlx5: Add interface to get reference to a UAR
From: Saeed Mahameed @ 2017-01-03 21:55 UTC (permalink / raw)
To: David S. Miller, Doug Ledford
Cc: netdev, linux-rdma, Leon Romanovsky, Eli Cohen, Matan Barak,
Leon Romanovsky, Saeed Mahameed
In-Reply-To: <1483480528-22622-1-git-send-email-saeedm@mellanox.com>
From: Eli Cohen <eli@mellanox.com>
A reference to a UAR is required to generate CQ or EQ doorbells. Since
CQ or EQ doorbells can all be generated using the same UAR area without
any effect on performance, we are just getting a reference to any
available UAR, If one is not available we allocate it but we don't waste
the blue flame registers it can provide and we will use them for
subsequent allocations.
We get a reference to such UAR and put in mlx5_priv so any kernel
consumer can make use of it.
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
drivers/net/ethernet/mellanox/mlx5/core/eq.c | 14 ++++-------
drivers/net/ethernet/mellanox/mlx5/core/main.c | 22 ++++++++++++++----
drivers/net/ethernet/mellanox/mlx5/core/uar.c | 32 ++++++++++++++++++++++++++
include/linux/mlx5/driver.h | 5 +++-
4 files changed, 59 insertions(+), 14 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eq.c b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
index 11a8d63..9849ee9 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eq.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eq.c
@@ -512,7 +512,7 @@ static void init_eq_buf(struct mlx5_eq *eq)
int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
int nent, u64 mask, const char *name,
- struct mlx5_uar *uar, enum mlx5_eq_type type)
+ enum mlx5_eq_type type)
{
u32 out[MLX5_ST_SZ_DW(create_eq_out)] = {0};
struct mlx5_priv *priv = &dev->priv;
@@ -556,7 +556,7 @@ int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
eqc = MLX5_ADDR_OF(create_eq_in, in, eq_context_entry);
MLX5_SET(eqc, eqc, log_eq_size, ilog2(eq->nent));
- MLX5_SET(eqc, eqc, uar_page, uar->index);
+ MLX5_SET(eqc, eqc, uar_page, priv->uar->index);
MLX5_SET(eqc, eqc, intr, vecidx);
MLX5_SET(eqc, eqc, log_page_size,
eq->buf.page_shift - MLX5_ADAPTER_PAGE_SHIFT);
@@ -571,7 +571,7 @@ int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
eq->eqn = MLX5_GET(create_eq_out, out, eq_number);
eq->irqn = priv->msix_arr[vecidx].vector;
eq->dev = dev;
- eq->doorbell = uar->map + MLX5_EQ_DOORBEL_OFFSET;
+ eq->doorbell = priv->uar->map + MLX5_EQ_DOORBEL_OFFSET;
err = request_irq(eq->irqn, handler, 0,
priv->irq_info[vecidx].name, eq);
if (err)
@@ -686,8 +686,7 @@ int mlx5_start_eqs(struct mlx5_core_dev *dev)
err = mlx5_create_map_eq(dev, &table->cmd_eq, MLX5_EQ_VEC_CMD,
MLX5_NUM_CMD_EQE, 1ull << MLX5_EVENT_TYPE_CMD,
- "mlx5_cmd_eq", &dev->priv.bfregi.uars[0],
- MLX5_EQ_TYPE_ASYNC);
+ "mlx5_cmd_eq", MLX5_EQ_TYPE_ASYNC);
if (err) {
mlx5_core_warn(dev, "failed to create cmd EQ %d\n", err);
return err;
@@ -697,8 +696,7 @@ int mlx5_start_eqs(struct mlx5_core_dev *dev)
err = mlx5_create_map_eq(dev, &table->async_eq, MLX5_EQ_VEC_ASYNC,
MLX5_NUM_ASYNC_EQE, async_event_mask,
- "mlx5_async_eq", &dev->priv.bfregi.uars[0],
- MLX5_EQ_TYPE_ASYNC);
+ "mlx5_async_eq", MLX5_EQ_TYPE_ASYNC);
if (err) {
mlx5_core_warn(dev, "failed to create async EQ %d\n", err);
goto err1;
@@ -708,7 +706,6 @@ int mlx5_start_eqs(struct mlx5_core_dev *dev)
MLX5_EQ_VEC_PAGES,
/* TODO: sriov max_vf + */ 1,
1 << MLX5_EVENT_TYPE_PAGE_REQUEST, "mlx5_pages_eq",
- &dev->priv.bfregi.uars[0],
MLX5_EQ_TYPE_ASYNC);
if (err) {
mlx5_core_warn(dev, "failed to create pages EQ %d\n", err);
@@ -722,7 +719,6 @@ int mlx5_start_eqs(struct mlx5_core_dev *dev)
MLX5_NUM_ASYNC_EQE,
1 << MLX5_EVENT_TYPE_PAGE_FAULT,
"mlx5_page_fault_eq",
- &dev->priv.bfregi.uars[0],
MLX5_EQ_TYPE_PF);
if (err) {
mlx5_core_warn(dev, "failed to create page fault EQ %d\n",
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index 634e96a..2882d04 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -753,8 +753,7 @@ static int alloc_comp_eqs(struct mlx5_core_dev *dev)
snprintf(name, MLX5_MAX_IRQ_NAME, "mlx5_comp%d", i);
err = mlx5_create_map_eq(dev, eq,
i + MLX5_EQ_VEC_COMP_BASE, nent, 0,
- name, &dev->priv.bfregi.uars[0],
- MLX5_EQ_TYPE_COMP);
+ name, MLX5_EQ_TYPE_COMP);
if (err) {
kfree(eq);
goto clean;
@@ -1094,12 +1093,18 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv,
goto err_cleanup_once;
}
- err = mlx5_alloc_bfregs(dev, &priv->bfregi);
- if (err) {
+ dev->priv.uar = mlx5_get_uars_page(dev);
+ if (!dev->priv.uar) {
dev_err(&pdev->dev, "Failed allocating uar, aborting\n");
goto err_disable_msix;
}
+ err = mlx5_alloc_bfregs(dev, &priv->bfregi);
+ if (err) {
+ dev_err(&pdev->dev, "Failed allocating uuars, aborting\n");
+ goto err_uar_cleanup;
+ }
+
err = mlx5_start_eqs(dev);
if (err) {
dev_err(&pdev->dev, "Failed to start pages and async EQs\n");
@@ -1172,6 +1177,9 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv,
err_free_uar:
mlx5_free_bfregs(dev, &priv->bfregi);
+err_uar_cleanup:
+ mlx5_put_uars_page(dev, priv->uar);
+
err_disable_msix:
mlx5_disable_msix(dev);
@@ -1231,6 +1239,7 @@ static int mlx5_unload_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv,
free_comp_eqs(dev);
mlx5_stop_eqs(dev);
mlx5_free_bfregs(dev, &priv->bfregi);
+ mlx5_put_uars_page(dev, priv->uar);
mlx5_disable_msix(dev);
if (cleanup)
mlx5_cleanup_once(dev);
@@ -1305,6 +1314,11 @@ static int init_one(struct pci_dev *pdev,
goto clean_dev;
}
#endif
+ mutex_init(&priv->bfregs.reg_head.lock);
+ mutex_init(&priv->bfregs.wc_head.lock);
+ INIT_LIST_HEAD(&priv->bfregs.reg_head.list);
+ INIT_LIST_HEAD(&priv->bfregs.wc_head.list);
+
err = mlx5_pci_init(dev, priv);
if (err) {
dev_err(&pdev->dev, "mlx5_pci_init failed with error code %d\n", err);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/uar.c b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
index 6a081a8..fcc0270 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/uar.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
@@ -332,6 +332,38 @@ static struct mlx5_uars_page *alloc_uars_page(struct mlx5_core_dev *mdev,
return ERR_PTR(err);
}
+struct mlx5_uars_page *mlx5_get_uars_page(struct mlx5_core_dev *mdev)
+{
+ struct mlx5_uars_page *ret;
+
+ mutex_lock(&mdev->priv.bfregs.reg_head.lock);
+ if (list_empty(&mdev->priv.bfregs.reg_head.list)) {
+ ret = alloc_uars_page(mdev, false);
+ if (IS_ERR(ret)) {
+ ret = NULL;
+ goto out;
+ }
+ list_add(&ret->list, &mdev->priv.bfregs.reg_head.list);
+ } else {
+ ret = list_first_entry(&mdev->priv.bfregs.reg_head.list,
+ struct mlx5_uars_page, list);
+ kref_get(&ret->ref_count);
+ }
+out:
+ mutex_unlock(&mdev->priv.bfregs.reg_head.lock);
+
+ return ret;
+}
+EXPORT_SYMBOL(mlx5_get_uars_page);
+
+void mlx5_put_uars_page(struct mlx5_core_dev *mdev, struct mlx5_uars_page *up)
+{
+ mutex_lock(&mdev->priv.bfregs.reg_head.lock);
+ kref_put(&up->ref_count, up_rel_func);
+ mutex_unlock(&mdev->priv.bfregs.reg_head.lock);
+}
+EXPORT_SYMBOL(mlx5_put_uars_page);
+
static unsigned long map_offset(struct mlx5_core_dev *mdev, int dbi)
{
/* return the offset in bytes from the start of the page to the
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 969aa1f..9a3a095 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -679,6 +679,7 @@ struct mlx5_priv {
struct srcu_struct pfault_srcu;
#endif
struct mlx5_bfreg_data bfregs;
+ struct mlx5_uars_page *uar;
};
enum mlx5_device_state {
@@ -1007,7 +1008,7 @@ void mlx5_cmd_comp_handler(struct mlx5_core_dev *dev, u64 vec);
void mlx5_cq_event(struct mlx5_core_dev *dev, u32 cqn, int event_type);
int mlx5_create_map_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq, u8 vecidx,
int nent, u64 mask, const char *name,
- struct mlx5_uar *uar, enum mlx5_eq_type type);
+ enum mlx5_eq_type type);
int mlx5_destroy_unmap_eq(struct mlx5_core_dev *dev, struct mlx5_eq *eq);
int mlx5_start_eqs(struct mlx5_core_dev *dev);
int mlx5_stop_eqs(struct mlx5_core_dev *dev);
@@ -1118,6 +1119,8 @@ int mlx5_cmd_create_vport_lag(struct mlx5_core_dev *dev);
int mlx5_cmd_destroy_vport_lag(struct mlx5_core_dev *dev);
bool mlx5_lag_is_active(struct mlx5_core_dev *dev);
struct net_device *mlx5_lag_get_roce_netdev(struct mlx5_core_dev *dev);
+struct mlx5_uars_page *mlx5_get_uars_page(struct mlx5_core_dev *mdev);
+void mlx5_put_uars_page(struct mlx5_core_dev *mdev, struct mlx5_uars_page *up);
struct mlx5_profile {
u64 mask;
--
2.7.4
^ permalink raw reply related
* [for-next 09/10] IB/mlx5: Support 4k UAR for libmlx5
From: Saeed Mahameed @ 2017-01-03 21:55 UTC (permalink / raw)
To: David S. Miller, Doug Ledford
Cc: netdev, linux-rdma, Leon Romanovsky, Eli Cohen, Matan Barak,
Leon Romanovsky, Saeed Mahameed
In-Reply-To: <1483480528-22622-1-git-send-email-saeedm@mellanox.com>
From: Eli Cohen <eli@mellanox.com>
Add fields to structs to convey to kernel an indication whether the
library supports multi UARs per page and return to the library the size
of a UAR based on the queried value.
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
drivers/infiniband/hw/mlx5/main.c | 21 +++++++-
drivers/net/ethernet/mellanox/mlx5/core/cq.c | 2 +
drivers/net/ethernet/mellanox/mlx5/core/en.h | 9 ++--
.../net/ethernet/mellanox/mlx5/core/en_common.c | 12 +----
drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 21 ++++----
drivers/net/ethernet/mellanox/mlx5/core/uar.c | 56 ----------------------
include/linux/mlx5/cq.h | 2 +-
include/linux/mlx5/driver.h | 12 -----
include/uapi/rdma/mlx5-abi.h | 7 +++
9 files changed, 42 insertions(+), 100 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 6640672..a191b93 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -992,6 +992,12 @@ static int mlx5_ib_modify_port(struct ib_device *ibdev, u8 port, int mask,
return err;
}
+static void print_lib_caps(struct mlx5_ib_dev *dev, u64 caps)
+{
+ mlx5_ib_dbg(dev, "MLX5_LIB_CAP_4K_UAR = %s\n",
+ caps & MLX5_LIB_CAP_4K_UAR ? "y" : "n");
+}
+
static int calc_total_bfregs(struct mlx5_ib_dev *dev, bool lib_uar_4k,
struct mlx5_ib_alloc_ucontext_req_v2 *req,
u32 *num_sys_pages)
@@ -1122,6 +1128,10 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
resp.cqe_version = min_t(__u8,
(__u8)MLX5_CAP_GEN(dev->mdev, cqe_version),
req.max_cqe_version);
+ resp.log_uar_size = MLX5_CAP_GEN(dev->mdev, uar_4k) ?
+ MLX5_ADAPTER_PAGE_SHIFT : PAGE_SHIFT;
+ resp.num_uars_per_page = MLX5_CAP_GEN(dev->mdev, uar_4k) ?
+ MLX5_CAP_GEN(dev->mdev, num_of_uars_per_page) : 1;
resp.response_length = min(offsetof(typeof(resp), response_length) +
sizeof(resp.response_length), udata->outlen);
@@ -1129,7 +1139,7 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
if (!context)
return ERR_PTR(-ENOMEM);
- lib_uar_4k = false;
+ lib_uar_4k = req.lib_caps & MLX5_LIB_CAP_4K_UAR;
bfregi = &context->bfregi;
/* updates req->total_num_bfregs */
@@ -1209,6 +1219,12 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
sizeof(resp.reserved2);
}
+ if (field_avail(typeof(resp), log_uar_size, udata->outlen))
+ resp.response_length += sizeof(resp.log_uar_size);
+
+ if (field_avail(typeof(resp), num_uars_per_page, udata->outlen))
+ resp.response_length += sizeof(resp.num_uars_per_page);
+
err = ib_copy_to_udata(udata, &resp, resp.response_length);
if (err)
goto out_td;
@@ -1216,7 +1232,8 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
bfregi->ver = ver;
bfregi->num_low_latency_bfregs = req.num_low_latency_bfregs;
context->cqe_version = resp.cqe_version;
- context->lib_caps = false;
+ context->lib_caps = req.lib_caps;
+ print_lib_caps(dev, context->lib_caps);
return &context->ibucontext;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/cq.c b/drivers/net/ethernet/mellanox/mlx5/core/cq.c
index 32d4af9..336d473 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/cq.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/cq.c
@@ -179,6 +179,8 @@ int mlx5_core_create_cq(struct mlx5_core_dev *dev, struct mlx5_core_cq *cq,
mlx5_core_dbg(dev, "failed adding CP 0x%x to debug file system\n",
cq->cqn);
+ cq->uar = dev->priv.uar;
+
return 0;
err_cmd:
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index e3ef0b5..396c63d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -465,7 +465,6 @@ struct mlx5e_sq {
/* read only */
struct mlx5_wq_cyc wq;
u32 dma_fifo_mask;
- void __iomem *uar_map;
struct netdev_queue *txq;
u32 sqn;
u16 bf_buf_size;
@@ -479,7 +478,7 @@ struct mlx5e_sq {
/* control path */
struct mlx5_wq_ctrl wq_ctrl;
- struct mlx5_uar uar;
+ struct mlx5_sq_bfreg bfreg;
struct mlx5e_channel *channel;
int tc;
u32 rate_limit;
@@ -806,7 +805,7 @@ void mlx5e_set_rx_cq_mode_params(struct mlx5e_params *params,
static inline void mlx5e_tx_notify_hw(struct mlx5e_sq *sq,
struct mlx5_wqe_ctrl_seg *ctrl, int bf_sz)
{
- u16 ofst = MLX5_BF_OFFSET + sq->bf_offset;
+ u16 ofst = sq->bf_offset;
/* ensure wqe is visible to device before updating doorbell record */
dma_wmb();
@@ -818,9 +817,9 @@ static inline void mlx5e_tx_notify_hw(struct mlx5e_sq *sq,
*/
wmb();
if (bf_sz)
- __iowrite64_copy(sq->uar_map + ofst, ctrl, bf_sz);
+ __iowrite64_copy(sq->bfreg.map + ofst, ctrl, bf_sz);
else
- mlx5_write64((__be32 *)ctrl, sq->uar_map + ofst);
+ mlx5_write64((__be32 *)ctrl, sq->bfreg.map + ofst);
/* flush the write-combining mapped buffer */
wmb();
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_common.c b/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
index f175518..bd898d8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
@@ -89,16 +89,10 @@ int mlx5e_create_mdev_resources(struct mlx5_core_dev *mdev)
struct mlx5e_resources *res = &mdev->mlx5e_res;
int err;
- err = mlx5_alloc_map_uar(mdev, &res->cq_uar, false);
- if (err) {
- mlx5_core_err(mdev, "alloc_map uar failed, %d\n", err);
- return err;
- }
-
err = mlx5_core_alloc_pd(mdev, &res->pdn);
if (err) {
mlx5_core_err(mdev, "alloc pd failed, %d\n", err);
- goto err_unmap_free_uar;
+ return err;
}
err = mlx5_core_alloc_transport_domain(mdev, &res->td.tdn);
@@ -121,9 +115,6 @@ int mlx5e_create_mdev_resources(struct mlx5_core_dev *mdev)
mlx5_core_dealloc_transport_domain(mdev, res->td.tdn);
err_dealloc_pd:
mlx5_core_dealloc_pd(mdev, res->pdn);
-err_unmap_free_uar:
- mlx5_unmap_free_uar(mdev, &res->cq_uar);
-
return err;
}
@@ -134,7 +125,6 @@ void mlx5e_destroy_mdev_resources(struct mlx5_core_dev *mdev)
mlx5_core_destroy_mkey(mdev, &res->mkey);
mlx5_core_dealloc_transport_domain(mdev, res->td.tdn);
mlx5_core_dealloc_pd(mdev, res->pdn);
- mlx5_unmap_free_uar(mdev, &res->cq_uar);
}
int mlx5e_refresh_tirs_self_loopback(struct mlx5_core_dev *mdev,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 5ff86f0..c32754b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -991,7 +991,7 @@ static int mlx5e_create_sq(struct mlx5e_channel *c,
sq->channel = c;
sq->tc = tc;
- err = mlx5_alloc_map_uar(mdev, &sq->uar, !!MLX5_CAP_GEN(mdev, bf));
+ err = mlx5_alloc_bfreg(mdev, &sq->bfreg, MLX5_CAP_GEN(mdev, bf), false);
if (err)
return err;
@@ -1003,12 +1003,9 @@ static int mlx5e_create_sq(struct mlx5e_channel *c,
goto err_unmap_free_uar;
sq->wq.db = &sq->wq.db[MLX5_SND_DBR];
- if (sq->uar.bf_map) {
+ if (sq->bfreg.wc)
set_bit(MLX5E_SQ_STATE_BF_ENABLE, &sq->state);
- sq->uar_map = sq->uar.bf_map;
- } else {
- sq->uar_map = sq->uar.map;
- }
+
sq->bf_buf_size = (1 << MLX5_CAP_GEN(mdev, log_bf_reg_size)) / 2;
sq->max_inline = param->max_inline;
sq->min_inline_mode =
@@ -1036,7 +1033,7 @@ static int mlx5e_create_sq(struct mlx5e_channel *c,
mlx5_wq_destroy(&sq->wq_ctrl);
err_unmap_free_uar:
- mlx5_unmap_free_uar(mdev, &sq->uar);
+ mlx5_free_bfreg(mdev, &sq->bfreg);
return err;
}
@@ -1048,7 +1045,7 @@ static void mlx5e_destroy_sq(struct mlx5e_sq *sq)
mlx5e_free_sq_db(sq);
mlx5_wq_destroy(&sq->wq_ctrl);
- mlx5_unmap_free_uar(priv->mdev, &sq->uar);
+ mlx5_free_bfreg(priv->mdev, &sq->bfreg);
}
static int mlx5e_enable_sq(struct mlx5e_sq *sq, struct mlx5e_sq_param *param)
@@ -1082,7 +1079,7 @@ static int mlx5e_enable_sq(struct mlx5e_sq *sq, struct mlx5e_sq_param *param)
MLX5_SET(sqc, sqc, tis_lst_sz, param->type == MLX5E_SQ_ICO ? 0 : 1);
MLX5_SET(wq, wq, wq_type, MLX5_WQ_TYPE_CYCLIC);
- MLX5_SET(wq, wq, uar_page, sq->uar.index);
+ MLX5_SET(wq, wq, uar_page, sq->bfreg.index);
MLX5_SET(wq, wq, log_wq_pg_sz, sq->wq_ctrl.buf.page_shift -
MLX5_ADAPTER_PAGE_SHIFT);
MLX5_SET64(wq, wq, dbr_addr, sq->wq_ctrl.db.dma);
@@ -1240,7 +1237,6 @@ static int mlx5e_create_cq(struct mlx5e_channel *c,
mcq->comp = mlx5e_completion_event;
mcq->event = mlx5e_cq_error_event;
mcq->irqn = irqn;
- mcq->uar = &mdev->mlx5e_res.cq_uar;
for (i = 0; i < mlx5_cqwq_get_size(&cq->wq); i++) {
struct mlx5_cqe64 *cqe = mlx5_cqwq_get_wqe(&cq->wq, i);
@@ -1289,7 +1285,7 @@ static int mlx5e_enable_cq(struct mlx5e_cq *cq, struct mlx5e_cq_param *param)
MLX5_SET(cqc, cqc, cq_period_mode, param->cq_period_mode);
MLX5_SET(cqc, cqc, c_eqn, eqn);
- MLX5_SET(cqc, cqc, uar_page, mcq->uar->index);
+ MLX5_SET(cqc, cqc, uar_page, mdev->priv.uar->index);
MLX5_SET(cqc, cqc, log_page_size, cq->wq_ctrl.frag_buf.page_shift -
MLX5_ADAPTER_PAGE_SHIFT);
MLX5_SET64(cqc, cqc, dbr_addr, cq->wq_ctrl.db.dma);
@@ -1701,7 +1697,7 @@ static void mlx5e_build_common_cq_param(struct mlx5e_priv *priv,
{
void *cqc = param->cqc;
- MLX5_SET(cqc, cqc, uar_page, priv->mdev->mlx5e_res.cq_uar.index);
+ MLX5_SET(cqc, cqc, uar_page, priv->mdev->priv.uar->index);
}
static void mlx5e_build_rx_cq_param(struct mlx5e_priv *priv,
@@ -2320,7 +2316,6 @@ static int mlx5e_create_drop_cq(struct mlx5e_priv *priv,
mcq->comp = mlx5e_completion_event;
mcq->event = mlx5e_cq_error_event;
mcq->irqn = irqn;
- mcq->uar = &mdev->mlx5e_res.cq_uar;
cq->priv = priv;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/uar.c b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
index 07b273c..2e6b0f2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/uar.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
@@ -37,11 +37,6 @@
#include <linux/mlx5/cmd.h>
#include "mlx5_core.h"
-enum {
- NUM_DRIVER_UARS = 4,
- NUM_LOW_LAT_BFREGS = 4,
-};
-
int mlx5_cmd_alloc_uar(struct mlx5_core_dev *dev, u32 *uarn)
{
u32 out[MLX5_ST_SZ_DW(alloc_uar_out)] = {0};
@@ -67,57 +62,6 @@ int mlx5_cmd_free_uar(struct mlx5_core_dev *dev, u32 uarn)
}
EXPORT_SYMBOL(mlx5_cmd_free_uar);
-int mlx5_alloc_map_uar(struct mlx5_core_dev *mdev, struct mlx5_uar *uar,
- bool map_wc)
-{
- phys_addr_t pfn;
- phys_addr_t uar_bar_start;
- int err;
-
- err = mlx5_cmd_alloc_uar(mdev, &uar->index);
- if (err) {
- mlx5_core_warn(mdev, "mlx5_cmd_alloc_uar() failed, %d\n", err);
- return err;
- }
-
- uar_bar_start = pci_resource_start(mdev->pdev, 0);
- pfn = (uar_bar_start >> PAGE_SHIFT) + uar->index;
-
- if (map_wc) {
- uar->bf_map = ioremap_wc(pfn << PAGE_SHIFT, PAGE_SIZE);
- if (!uar->bf_map) {
- mlx5_core_warn(mdev, "ioremap_wc() failed\n");
- uar->map = ioremap(pfn << PAGE_SHIFT, PAGE_SIZE);
- if (!uar->map)
- goto err_free_uar;
- }
- } else {
- uar->map = ioremap(pfn << PAGE_SHIFT, PAGE_SIZE);
- if (!uar->map)
- goto err_free_uar;
- }
-
- return 0;
-
-err_free_uar:
- mlx5_core_warn(mdev, "ioremap() failed\n");
- err = -ENOMEM;
- mlx5_cmd_free_uar(mdev, uar->index);
-
- return err;
-}
-EXPORT_SYMBOL(mlx5_alloc_map_uar);
-
-void mlx5_unmap_free_uar(struct mlx5_core_dev *mdev, struct mlx5_uar *uar)
-{
- if (uar->map)
- iounmap(uar->map);
- else
- iounmap(uar->bf_map);
- mlx5_cmd_free_uar(mdev, uar->index);
-}
-EXPORT_SYMBOL(mlx5_unmap_free_uar);
-
static int uars_per_sys_page(struct mlx5_core_dev *mdev)
{
if (MLX5_CAP_GEN(mdev, uar_4k))
diff --git a/include/linux/mlx5/cq.h b/include/linux/mlx5/cq.h
index 1ed50f7..fbce184 100644
--- a/include/linux/mlx5/cq.h
+++ b/include/linux/mlx5/cq.h
@@ -42,13 +42,13 @@ struct mlx5_core_cq {
int cqe_sz;
__be32 *set_ci_db;
__be32 *arm_db;
+ struct mlx5_uars_page *uar;
atomic_t refcount;
struct completion free;
unsigned vector;
unsigned int irqn;
void (*comp) (struct mlx5_core_cq *);
void (*event) (struct mlx5_core_cq *, enum mlx5_event);
- struct mlx5_uar *uar;
u32 cons_index;
unsigned arm_sn;
struct mlx5_rsc_debug *dbg;
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 7e7394f..10e6325 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -467,12 +467,6 @@ struct mlx5_sq_bfreg {
unsigned int offset;
};
-struct mlx5_uar {
- u32 index;
- void __iomem *map;
- void __iomem *bf_map;
-};
-
struct mlx5_core_health {
struct health_buffer __iomem *health;
__be32 __iomem *health_counter;
@@ -725,7 +719,6 @@ struct mlx5_td {
};
struct mlx5e_resources {
- struct mlx5_uar cq_uar;
u32 pdn;
struct mlx5_td td;
struct mlx5_core_mkey mkey;
@@ -915,11 +908,6 @@ void mlx5_cmd_mbox_status(void *out, u8 *status, u32 *syndrome);
int mlx5_core_get_caps(struct mlx5_core_dev *dev, enum mlx5_cap_type cap_type);
int mlx5_cmd_alloc_uar(struct mlx5_core_dev *dev, u32 *uarn);
int mlx5_cmd_free_uar(struct mlx5_core_dev *dev, u32 uarn);
-int mlx5_alloc_bfregs(struct mlx5_core_dev *dev, struct mlx5_bfreg_info *bfregi);
-int mlx5_free_bfregs(struct mlx5_core_dev *dev, struct mlx5_bfreg_info *bfregi);
-int mlx5_alloc_map_uar(struct mlx5_core_dev *mdev, struct mlx5_uar *uar,
- bool map_wc);
-void mlx5_unmap_free_uar(struct mlx5_core_dev *mdev, struct mlx5_uar *uar);
void mlx5_health_cleanup(struct mlx5_core_dev *dev);
int mlx5_health_init(struct mlx5_core_dev *dev);
void mlx5_start_health_poll(struct mlx5_core_dev *dev);
diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h
index 86a8f30..85dc966 100644
--- a/include/uapi/rdma/mlx5-abi.h
+++ b/include/uapi/rdma/mlx5-abi.h
@@ -65,6 +65,10 @@ struct mlx5_ib_alloc_ucontext_req {
__u32 num_low_latency_bfregs;
};
+enum mlx5_lib_caps {
+ MLX5_LIB_CAP_4K_UAR = (u64)1 << 0,
+};
+
struct mlx5_ib_alloc_ucontext_req_v2 {
__u32 total_num_bfregs;
__u32 num_low_latency_bfregs;
@@ -74,6 +78,7 @@ struct mlx5_ib_alloc_ucontext_req_v2 {
__u8 reserved0;
__u16 reserved1;
__u32 reserved2;
+ __u64 lib_caps;
};
enum mlx5_ib_alloc_ucontext_resp_mask {
@@ -103,6 +108,8 @@ struct mlx5_ib_alloc_ucontext_resp {
__u8 cmds_supp_uhw;
__u16 reserved2;
__u64 hca_core_clock_offset;
+ __u32 log_uar_size;
+ __u32 num_uars_per_page;
};
struct mlx5_ib_alloc_pd_resp {
--
2.7.4
^ permalink raw reply related
* [for-next 04/10] IB/mlx5: Fix retrieval of index to first hi class bfreg
From: Saeed Mahameed @ 2017-01-03 21:55 UTC (permalink / raw)
To: David S. Miller, Doug Ledford
Cc: netdev, linux-rdma, Leon Romanovsky, Eli Cohen, Matan Barak,
Leon Romanovsky, Saeed Mahameed
In-Reply-To: <1483480528-22622-1-git-send-email-saeedm@mellanox.com>
From: Eli Cohen <eli@mellanox.com>
First the function retrieving the index of the first hi latency class
blue flame register. High latency class bfregs are located right above
medium latency class bfregs.
Fixes: c1be5232d21d ('IB/mlx5: Fix micro UAR allocator')
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
drivers/infiniband/hw/mlx5/qp.c | 24 ++++++++++++++----------
1 file changed, 14 insertions(+), 10 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index fbea9bd..240fbb0 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -490,12 +490,21 @@ static int next_bfreg(int n)
return n;
}
+enum {
+ /* this is the first blue flame register in the array of bfregs assigned
+ * to a processes. Since we do not use it for blue flame but rather
+ * regular 64 bit doorbells, we do not need a lock for maintaiing
+ * "odd/even" order
+ */
+ NUM_NON_BLUE_FLAME_BFREGS = 1,
+};
+
static int num_med_bfreg(struct mlx5_bfreg_info *bfregi)
{
int n;
n = bfregi->num_uars * MLX5_NON_FP_BFREGS_PER_UAR -
- bfregi->num_low_latency_bfregs - 1;
+ bfregi->num_low_latency_bfregs - NUM_NON_BLUE_FLAME_BFREGS;
return n >= 0 ? n : 0;
}
@@ -508,17 +517,9 @@ static int max_bfregi(struct mlx5_bfreg_info *bfregi)
static int first_hi_bfreg(struct mlx5_bfreg_info *bfregi)
{
int med;
- int i;
- int t;
med = num_med_bfreg(bfregi);
- for (t = 0, i = first_med_bfreg();; i = next_bfreg(i)) {
- t++;
- if (t == med)
- return next_bfreg(i);
- }
-
- return 0;
+ return next_bfreg(med);
}
static int alloc_high_class_bfreg(struct mlx5_bfreg_info *bfregi)
@@ -544,6 +545,8 @@ static int alloc_med_class_bfreg(struct mlx5_bfreg_info *bfregi)
for (i = first_med_bfreg(); i < first_hi_bfreg(bfregi); i = next_bfreg(i)) {
if (bfregi->count[i] < bfregi->count[minidx])
minidx = i;
+ if (!bfregi->count[minidx])
+ break;
}
bfregi->count[minidx]++;
@@ -558,6 +561,7 @@ static int alloc_bfreg(struct mlx5_bfreg_info *bfregi,
mutex_lock(&bfregi->lock);
switch (lat) {
case MLX5_IB_LATENCY_CLASS_LOW:
+ BUILD_BUG_ON(NUM_NON_BLUE_FLAME_BFREGS != 1);
bfregn = 0;
bfregi->count[bfregn]++;
break;
--
2.7.4
^ permalink raw reply related
* [for-next 02/10] IB/mlx5: Fix error handling order in create_kernel_qp
From: Saeed Mahameed @ 2017-01-03 21:55 UTC (permalink / raw)
To: David S. Miller, Doug Ledford
Cc: netdev, linux-rdma, Leon Romanovsky, Eli Cohen, Matan Barak,
Leon Romanovsky, Saeed Mahameed
In-Reply-To: <1483480528-22622-1-git-send-email-saeedm@mellanox.com>
From: Eli Cohen <eli@mellanox.com>
Make sure order of cleanup is exactly the opposite of initialization.
Fixes: 9603b61de1ee ('mlx5: Move pci device handling from mlx5_ib to mlx5_core')
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
drivers/infiniband/hw/mlx5/qp.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 53f4dd3..42d021cd 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -994,12 +994,12 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
return 0;
err_wrid:
- mlx5_db_free(dev->mdev, &qp->db);
kfree(qp->sq.wqe_head);
kfree(qp->sq.w_list);
kfree(qp->sq.wrid);
kfree(qp->sq.wr_data);
kfree(qp->rq.wrid);
+ mlx5_db_free(dev->mdev, &qp->db);
err_free:
kvfree(*in);
@@ -1014,12 +1014,12 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
static void destroy_qp_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
{
- mlx5_db_free(dev->mdev, &qp->db);
kfree(qp->sq.wqe_head);
kfree(qp->sq.w_list);
kfree(qp->sq.wrid);
kfree(qp->sq.wr_data);
kfree(qp->rq.wrid);
+ mlx5_db_free(dev->mdev, &qp->db);
mlx5_buf_free(dev->mdev, &qp->buf);
free_uuar(&dev->mdev->priv.uuari, qp->bf->uuarn);
}
--
2.7.4
^ permalink raw reply related
* [for-next 10/10] net/mlx5: Activate support for 4K UARs
From: Saeed Mahameed @ 2017-01-03 21:55 UTC (permalink / raw)
To: David S. Miller, Doug Ledford
Cc: netdev-u79uwXL29TY76Z2rM5mHXA, linux-rdma-u79uwXL29TY76Z2rM5mHXA,
Leon Romanovsky, Eli Cohen, Matan Barak, Leon Romanovsky,
Saeed Mahameed
In-Reply-To: <1483480528-22622-1-git-send-email-saeedm-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
From: Eli Cohen <eli-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Activate 4K UAR support for firmware versions that support it.
Signed-off-by: Eli Cohen <eli-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Leon Romanovsky <leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
Signed-off-by: Saeed Mahameed <saeedm-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
drivers/net/ethernet/mellanox/mlx5/core/main.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index ff1f144..a16ee16 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -530,6 +530,10 @@ static int handle_hca_cap(struct mlx5_core_dev *dev)
/* disable cmdif checksum */
MLX5_SET(cmd_hca_cap, set_hca_cap, cmdif_checksum, 0);
+ /* If the HCA supports 4K UARs use it */
+ if (MLX5_CAP_GEN_MAX(dev, uar_4k))
+ MLX5_SET(cmd_hca_cap, set_hca_cap, uar_4k, 1);
+
MLX5_SET(cmd_hca_cap, set_hca_cap, log_uar_page_sz, PAGE_SHIFT - 12);
err = set_caps(dev, set_ctx, set_sz,
--
2.7.4
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [for-next 08/10] IB/mlx5: Allow future extension of libmlx5 input data
From: Saeed Mahameed @ 2017-01-03 21:55 UTC (permalink / raw)
To: David S. Miller, Doug Ledford
Cc: netdev-u79uwXL29TY76Z2rM5mHXA, linux-rdma-u79uwXL29TY76Z2rM5mHXA,
Leon Romanovsky, Eli Cohen, Matan Barak, Leon Romanovsky,
Saeed Mahameed
In-Reply-To: <1483480528-22622-1-git-send-email-saeedm-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
From: Eli Cohen <eli-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Current check requests that new fields in struct
mlx5_ib_alloc_ucontext_req_v2 that are not known to the driver be zero.
This was introduced so new libraries passing additional information to
the kernel through struct mlx5_ib_alloc_ucontext_req_v2 will be notified
by old kernels that do not support their request by failing the
operation. This schecme is problematic since it requires libmlx5 to issue
the requests with descending input size for struct
mlx5_ib_alloc_ucontext_req_v2.
To avoid this, we require that new features that will obey the following
rules:
If the feature requires one or more fields in the response and the at
least one of the fields can be encoded such that a zero value means the
kernel ignored the request then this field will provide the indication
to the library. If no response is required or if zero is a valid
response, a new field should be added that indicates to the library
whether its request was processed.
Fixes: b368d7cb8ceb ('IB/mlx5: Add hca_core_clock_offset to udata in init_ucontext')
Signed-off-by: Eli Cohen <eli-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Leon Romanovsky <leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
Signed-off-by: Saeed Mahameed <saeedm-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
drivers/infiniband/hw/mlx5/cq.c | 2 +-
drivers/infiniband/hw/mlx5/main.c | 201 ++++++++++++++++++++++-------------
drivers/infiniband/hw/mlx5/mlx5_ib.h | 15 ++-
drivers/infiniband/hw/mlx5/qp.c | 133 ++++++++++-------------
include/linux/mlx5/device.h | 12 ++-
include/linux/mlx5/driver.h | 12 +--
6 files changed, 209 insertions(+), 166 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index a28ec33..31803b3 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -788,7 +788,7 @@ static int create_cq_user(struct mlx5_ib_dev *dev, struct ib_udata *udata,
MLX5_SET(cqc, cqc, log_page_size,
page_shift - MLX5_ADAPTER_PAGE_SHIFT);
- *index = to_mucontext(context)->bfregi.uars[0].index;
+ *index = to_mucontext(context)->bfregi.sys_pages[0];
if (ucmd.cqe_comp_en == 1) {
if (unlikely((*cqe_size != 64) ||
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index e9f0830..6640672 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -992,6 +992,80 @@ static int mlx5_ib_modify_port(struct ib_device *ibdev, u8 port, int mask,
return err;
}
+static int calc_total_bfregs(struct mlx5_ib_dev *dev, bool lib_uar_4k,
+ struct mlx5_ib_alloc_ucontext_req_v2 *req,
+ u32 *num_sys_pages)
+{
+ int uars_per_sys_page;
+ int bfregs_per_sys_page;
+ int ref_bfregs = req->total_num_bfregs;
+
+ if (req->total_num_bfregs == 0)
+ return -EINVAL;
+
+ BUILD_BUG_ON(MLX5_MAX_BFREGS % MLX5_NON_FP_BFREGS_IN_PAGE);
+ BUILD_BUG_ON(MLX5_MAX_BFREGS < MLX5_NON_FP_BFREGS_IN_PAGE);
+
+ if (req->total_num_bfregs > MLX5_MAX_BFREGS)
+ return -ENOMEM;
+
+ uars_per_sys_page = get_uars_per_sys_page(dev, lib_uar_4k);
+ bfregs_per_sys_page = uars_per_sys_page * MLX5_NON_FP_BFREGS_PER_UAR;
+ req->total_num_bfregs = ALIGN(req->total_num_bfregs, bfregs_per_sys_page);
+ *num_sys_pages = req->total_num_bfregs / bfregs_per_sys_page;
+
+ if (req->num_low_latency_bfregs > req->total_num_bfregs - 1)
+ return -EINVAL;
+
+ mlx5_ib_dbg(dev, "uar_4k: fw support %s, lib support %s, user requested %d bfregs, alloated %d, using %d sys pages\n",
+ MLX5_CAP_GEN(dev->mdev, uar_4k) ? "yes" : "no",
+ lib_uar_4k ? "yes" : "no", ref_bfregs,
+ req->total_num_bfregs, *num_sys_pages);
+
+ return 0;
+}
+
+static int allocate_uars(struct mlx5_ib_dev *dev, struct mlx5_ib_ucontext *context)
+{
+ struct mlx5_bfreg_info *bfregi;
+ int err;
+ int i;
+
+ bfregi = &context->bfregi;
+ for (i = 0; i < bfregi->num_sys_pages; i++) {
+ err = mlx5_cmd_alloc_uar(dev->mdev, &bfregi->sys_pages[i]);
+ if (err)
+ goto error;
+
+ mlx5_ib_dbg(dev, "allocated uar %d\n", bfregi->sys_pages[i]);
+ }
+ return 0;
+
+error:
+ for (--i; i >= 0; i--)
+ if (mlx5_cmd_free_uar(dev->mdev, bfregi->sys_pages[i]))
+ mlx5_ib_warn(dev, "failed to free uar %d\n", i);
+
+ return err;
+}
+
+static int deallocate_uars(struct mlx5_ib_dev *dev, struct mlx5_ib_ucontext *context)
+{
+ struct mlx5_bfreg_info *bfregi;
+ int err;
+ int i;
+
+ bfregi = &context->bfregi;
+ for (i = 0; i < bfregi->num_sys_pages; i++) {
+ err = mlx5_cmd_free_uar(dev->mdev, bfregi->sys_pages[i]);
+ if (err) {
+ mlx5_ib_warn(dev, "failed to free uar %d\n", i);
+ return err;
+ }
+ }
+ return 0;
+}
+
static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
struct ib_udata *udata)
{
@@ -1000,16 +1074,12 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
struct mlx5_ib_alloc_ucontext_resp resp = {};
struct mlx5_ib_ucontext *context;
struct mlx5_bfreg_info *bfregi;
- struct mlx5_uar *uars;
- int gross_bfregs;
- int num_uars;
int ver;
- int bfregn;
int err;
- int i;
size_t reqlen;
size_t min_req_v2 = offsetof(struct mlx5_ib_alloc_ucontext_req_v2,
max_cqe_version);
+ bool lib_uar_4k;
if (!dev->ib_active)
return ERR_PTR(-EAGAIN);
@@ -1032,27 +1102,14 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
if (req.flags)
return ERR_PTR(-EINVAL);
- if (req.total_num_bfregs > MLX5_MAX_BFREGS)
- return ERR_PTR(-ENOMEM);
-
- if (req.total_num_bfregs == 0)
- return ERR_PTR(-EINVAL);
-
if (req.comp_mask || req.reserved0 || req.reserved1 || req.reserved2)
return ERR_PTR(-EOPNOTSUPP);
- if (reqlen > sizeof(req) &&
- !ib_is_udata_cleared(udata, sizeof(req),
- reqlen - sizeof(req)))
- return ERR_PTR(-EOPNOTSUPP);
-
req.total_num_bfregs = ALIGN(req.total_num_bfregs,
MLX5_NON_FP_BFREGS_PER_UAR);
if (req.num_low_latency_bfregs > req.total_num_bfregs - 1)
return ERR_PTR(-EINVAL);
- num_uars = req.total_num_bfregs / MLX5_NON_FP_BFREGS_PER_UAR;
- gross_bfregs = num_uars * MLX5_BFREGS_PER_UAR;
resp.qp_tab_size = 1 << MLX5_CAP_GEN(dev->mdev, log_max_qp);
if (mlx5_core_is_pf(dev->mdev) && MLX5_CAP_GEN(dev->mdev, bf))
resp.bf_reg_size = 1 << MLX5_CAP_GEN(dev->mdev, log_bf_reg_size);
@@ -1072,42 +1129,34 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
if (!context)
return ERR_PTR(-ENOMEM);
+ lib_uar_4k = false;
bfregi = &context->bfregi;
- mutex_init(&bfregi->lock);
- uars = kcalloc(num_uars, sizeof(*uars), GFP_KERNEL);
- if (!uars) {
- err = -ENOMEM;
+
+ /* updates req->total_num_bfregs */
+ err = calc_total_bfregs(dev, lib_uar_4k, &req, &bfregi->num_sys_pages);
+ if (err)
goto out_ctx;
- }
- bfregi->bitmap = kcalloc(BITS_TO_LONGS(gross_bfregs),
- sizeof(*bfregi->bitmap),
+ mutex_init(&bfregi->lock);
+ bfregi->lib_uar_4k = lib_uar_4k;
+ bfregi->count = kcalloc(req.total_num_bfregs, sizeof(*bfregi->count),
GFP_KERNEL);
- if (!bfregi->bitmap) {
+ if (!bfregi->count) {
err = -ENOMEM;
- goto out_uar_ctx;
- }
- /*
- * clear all fast path bfregs
- */
- for (i = 0; i < gross_bfregs; i++) {
- bfregn = i & 3;
- if (bfregn == 2 || bfregn == 3)
- set_bit(i, bfregi->bitmap);
+ goto out_ctx;
}
- bfregi->count = kcalloc(gross_bfregs,
- sizeof(*bfregi->count), GFP_KERNEL);
- if (!bfregi->count) {
+ bfregi->sys_pages = kcalloc(bfregi->num_sys_pages,
+ sizeof(*bfregi->sys_pages),
+ GFP_KERNEL);
+ if (!bfregi->sys_pages) {
err = -ENOMEM;
- goto out_bitmap;
+ goto out_count;
}
- for (i = 0; i < num_uars; i++) {
- err = mlx5_cmd_alloc_uar(dev->mdev, &uars[i].index);
- if (err)
- goto out_count;
- }
+ err = allocate_uars(dev, context);
+ if (err)
+ goto out_sys_pages;
#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
context->ibucontext.invalidate_range = &mlx5_ib_invalidate_range;
@@ -1166,9 +1215,8 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
bfregi->ver = ver;
bfregi->num_low_latency_bfregs = req.num_low_latency_bfregs;
- bfregi->uars = uars;
- bfregi->num_uars = num_uars;
context->cqe_version = resp.cqe_version;
+ context->lib_caps = false;
return &context->ibucontext;
@@ -1180,19 +1228,17 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
free_page(context->upd_xlt_page);
out_uars:
- for (i--; i >= 0; i--)
- mlx5_cmd_free_uar(dev->mdev, uars[i].index);
-out_count:
- kfree(bfregi->count);
+ deallocate_uars(dev, context);
-out_bitmap:
- kfree(bfregi->bitmap);
+out_sys_pages:
+ kfree(bfregi->sys_pages);
-out_uar_ctx:
- kfree(uars);
+out_count:
+ kfree(bfregi->count);
out_ctx:
kfree(context);
+
return ERR_PTR(err);
}
@@ -1200,31 +1246,31 @@ static int mlx5_ib_dealloc_ucontext(struct ib_ucontext *ibcontext)
{
struct mlx5_ib_ucontext *context = to_mucontext(ibcontext);
struct mlx5_ib_dev *dev = to_mdev(ibcontext->device);
- struct mlx5_bfreg_info *bfregi = &context->bfregi;
- int i;
+ struct mlx5_bfreg_info *bfregi;
+ bfregi = &context->bfregi;
if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain))
mlx5_core_dealloc_transport_domain(dev->mdev, context->tdn);
free_page(context->upd_xlt_page);
-
- for (i = 0; i < bfregi->num_uars; i++) {
- if (mlx5_cmd_free_uar(dev->mdev, bfregi->uars[i].index))
- mlx5_ib_warn(dev, "Failed to free UAR 0x%x\n",
- bfregi->uars[i].index);
- }
-
+ deallocate_uars(dev, context);
+ kfree(bfregi->sys_pages);
kfree(bfregi->count);
- kfree(bfregi->bitmap);
- kfree(bfregi->uars);
kfree(context);
return 0;
}
-static phys_addr_t uar_index2pfn(struct mlx5_ib_dev *dev, int index)
+static phys_addr_t uar_index2pfn(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi,
+ int idx)
{
- return (pci_resource_start(dev->mdev->pdev, 0) >> PAGE_SHIFT) + index;
+ int fw_uars_per_page;
+
+ fw_uars_per_page = MLX5_CAP_GEN(dev->mdev, uar_4k) ? MLX5_UARS_IN_PAGE : 1;
+
+ return (pci_resource_start(dev->mdev->pdev, 0) >> PAGE_SHIFT) +
+ bfregi->sys_pages[idx] / fw_uars_per_page;
}
static int get_command(unsigned long offset)
@@ -1384,6 +1430,18 @@ static int uar_mmap(struct mlx5_ib_dev *dev, enum mlx5_ib_mmap_cmd cmd,
unsigned long idx;
phys_addr_t pfn, pa;
pgprot_t prot;
+ int uars_per_page;
+
+ if (vma->vm_end - vma->vm_start != PAGE_SIZE)
+ return -EINVAL;
+
+ uars_per_page = get_uars_per_sys_page(dev, bfregi->lib_uar_4k);
+ idx = get_index(vma->vm_pgoff);
+ if (idx % uars_per_page ||
+ idx * uars_per_page >= bfregi->num_sys_pages) {
+ mlx5_ib_warn(dev, "invalid uar index %lu\n", idx);
+ return -EINVAL;
+ }
switch (cmd) {
case MLX5_IB_MMAP_WC_PAGE:
@@ -1406,14 +1464,7 @@ static int uar_mmap(struct mlx5_ib_dev *dev, enum mlx5_ib_mmap_cmd cmd,
return -EINVAL;
}
- if (vma->vm_end - vma->vm_start != PAGE_SIZE)
- return -EINVAL;
-
- idx = get_index(vma->vm_pgoff);
- if (idx >= bfregi->num_uars)
- return -EINVAL;
-
- pfn = uar_index2pfn(dev, bfregi->uars[idx].index);
+ pfn = uar_index2pfn(dev, bfregi, idx);
mlx5_ib_dbg(dev, "uar idx 0x%lx, pfn %pa\n", idx, &pfn);
vma->vm_page_prot = prot;
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index ae3bc4a..e1a4b93d 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -90,7 +90,6 @@ enum mlx5_ib_latency_class {
MLX5_IB_LATENCY_CLASS_LOW,
MLX5_IB_LATENCY_CLASS_MEDIUM,
MLX5_IB_LATENCY_CLASS_HIGH,
- MLX5_IB_LATENCY_CLASS_FAST_PATH
};
enum mlx5_ib_mad_ifc_flags {
@@ -129,6 +128,7 @@ struct mlx5_ib_ucontext {
unsigned long upd_xlt_page;
/* protect ODP/KSM */
struct mutex upd_xlt_page_mutex;
+ u64 lib_caps;
};
static inline struct mlx5_ib_ucontext *to_mucontext(struct ib_ucontext *ibucontext)
@@ -975,4 +975,17 @@ static inline int get_srq_user_index(struct mlx5_ib_ucontext *ucontext,
return verify_assign_uidx(cqe_version, ucmd->uidx, user_index);
}
+
+static inline int get_uars_per_sys_page(struct mlx5_ib_dev *dev, bool lib_support)
+{
+ return lib_support && MLX5_CAP_GEN(dev->mdev, uar_4k) ?
+ MLX5_UARS_IN_PAGE : 1;
+}
+
+static inline int get_num_uars(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi)
+{
+ return get_uars_per_sys_page(dev, bfregi->lib_uar_4k) * bfregi->num_sys_pages;
+}
+
#endif /* MLX5_IB_H */
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index b38240e..ecef3c0 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -480,16 +480,6 @@ static int first_med_bfreg(void)
return 1;
}
-static int next_bfreg(int n)
-{
- n++;
-
- while (((n % 4) & 2))
- n++;
-
- return n;
-}
-
enum {
/* this is the first blue flame register in the array of bfregs assigned
* to a processes. Since we do not use it for blue flame but rather
@@ -499,36 +489,38 @@ enum {
NUM_NON_BLUE_FLAME_BFREGS = 1,
};
-static int num_med_bfreg(struct mlx5_bfreg_info *bfregi)
+static int max_bfregs(struct mlx5_ib_dev *dev, struct mlx5_bfreg_info *bfregi)
+{
+ return get_num_uars(dev, bfregi) * MLX5_NON_FP_BFREGS_PER_UAR;
+}
+
+static int num_med_bfreg(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi)
{
int n;
- n = bfregi->num_uars * MLX5_NON_FP_BFREGS_PER_UAR -
- bfregi->num_low_latency_bfregs - NUM_NON_BLUE_FLAME_BFREGS;
+ n = max_bfregs(dev, bfregi) - bfregi->num_low_latency_bfregs -
+ NUM_NON_BLUE_FLAME_BFREGS;
return n >= 0 ? n : 0;
}
-static int max_bfregi(struct mlx5_bfreg_info *bfregi)
-{
- return bfregi->num_uars * 4;
-}
-
-static int first_hi_bfreg(struct mlx5_bfreg_info *bfregi)
+static int first_hi_bfreg(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi)
{
int med;
- med = num_med_bfreg(bfregi);
- return next_bfreg(med);
+ med = num_med_bfreg(dev, bfregi);
+ return ++med;
}
-static int alloc_high_class_bfreg(struct mlx5_bfreg_info *bfregi)
+static int alloc_high_class_bfreg(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi)
{
int i;
- for (i = first_hi_bfreg(bfregi); i < max_bfregi(bfregi); i = next_bfreg(i)) {
- if (!test_bit(i, bfregi->bitmap)) {
- set_bit(i, bfregi->bitmap);
+ for (i = first_hi_bfreg(dev, bfregi); i < max_bfregs(dev, bfregi); i++) {
+ if (!bfregi->count[i]) {
bfregi->count[i]++;
return i;
}
@@ -537,12 +529,13 @@ static int alloc_high_class_bfreg(struct mlx5_bfreg_info *bfregi)
return -ENOMEM;
}
-static int alloc_med_class_bfreg(struct mlx5_bfreg_info *bfregi)
+static int alloc_med_class_bfreg(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi)
{
int minidx = first_med_bfreg();
int i;
- for (i = first_med_bfreg(); i < first_hi_bfreg(bfregi); i = next_bfreg(i)) {
+ for (i = first_med_bfreg(); i < first_hi_bfreg(dev, bfregi); i++) {
if (bfregi->count[i] < bfregi->count[minidx])
minidx = i;
if (!bfregi->count[minidx])
@@ -553,7 +546,8 @@ static int alloc_med_class_bfreg(struct mlx5_bfreg_info *bfregi)
return minidx;
}
-static int alloc_bfreg(struct mlx5_bfreg_info *bfregi,
+static int alloc_bfreg(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi,
enum mlx5_ib_latency_class lat)
{
int bfregn = -EINVAL;
@@ -570,18 +564,14 @@ static int alloc_bfreg(struct mlx5_bfreg_info *bfregi,
if (bfregi->ver < 2)
bfregn = -ENOMEM;
else
- bfregn = alloc_med_class_bfreg(bfregi);
+ bfregn = alloc_med_class_bfreg(dev, bfregi);
break;
case MLX5_IB_LATENCY_CLASS_HIGH:
if (bfregi->ver < 2)
bfregn = -ENOMEM;
else
- bfregn = alloc_high_class_bfreg(bfregi);
- break;
-
- case MLX5_IB_LATENCY_CLASS_FAST_PATH:
- bfregn = 2;
+ bfregn = alloc_high_class_bfreg(dev, bfregi);
break;
}
mutex_unlock(&bfregi->lock);
@@ -589,37 +579,10 @@ static int alloc_bfreg(struct mlx5_bfreg_info *bfregi,
return bfregn;
}
-static void free_med_class_bfreg(struct mlx5_bfreg_info *bfregi, int bfregn)
+static void free_bfreg(struct mlx5_ib_dev *dev, struct mlx5_bfreg_info *bfregi, int bfregn)
{
- clear_bit(bfregn, bfregi->bitmap);
- --bfregi->count[bfregn];
-}
-
-static void free_high_class_bfreg(struct mlx5_bfreg_info *bfregi, int bfregn)
-{
- clear_bit(bfregn, bfregi->bitmap);
- --bfregi->count[bfregn];
-}
-
-static void free_bfreg(struct mlx5_bfreg_info *bfregi, int bfregn)
-{
- int nbfregs = bfregi->num_uars * MLX5_BFREGS_PER_UAR;
- int high_bfreg = nbfregs - bfregi->num_low_latency_bfregs;
-
mutex_lock(&bfregi->lock);
- if (bfregn == 0) {
- --bfregi->count[bfregn];
- goto out;
- }
-
- if (bfregn < high_bfreg) {
- free_med_class_bfreg(bfregi, bfregn);
- goto out;
- }
-
- free_high_class_bfreg(bfregi, bfregn);
-
-out:
+ bfregi->count[bfregn]--;
mutex_unlock(&bfregi->lock);
}
@@ -661,9 +624,20 @@ static void mlx5_ib_lock_cqs(struct mlx5_ib_cq *send_cq,
static void mlx5_ib_unlock_cqs(struct mlx5_ib_cq *send_cq,
struct mlx5_ib_cq *recv_cq);
-static int bfregn_to_uar_index(struct mlx5_bfreg_info *bfregi, int bfregn)
+static int bfregn_to_uar_index(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi, int bfregn)
{
- return bfregi->uars[bfregn / MLX5_BFREGS_PER_UAR].index;
+ int bfregs_per_sys_page;
+ int index_of_sys_page;
+ int offset;
+
+ bfregs_per_sys_page = get_uars_per_sys_page(dev, bfregi->lib_uar_4k) *
+ MLX5_NON_FP_BFREGS_PER_UAR;
+ index_of_sys_page = bfregn / bfregs_per_sys_page;
+
+ offset = bfregn % bfregs_per_sys_page / MLX5_NON_FP_BFREGS_PER_UAR;
+
+ return bfregi->sys_pages[index_of_sys_page] + offset;
}
static int mlx5_ib_umem_get(struct mlx5_ib_dev *dev,
@@ -766,6 +740,13 @@ static int create_user_rq(struct mlx5_ib_dev *dev, struct ib_pd *pd,
return err;
}
+static int adjust_bfregn(struct mlx5_ib_dev *dev,
+ struct mlx5_bfreg_info *bfregi, int bfregn)
+{
+ return bfregn / MLX5_NON_FP_BFREGS_PER_UAR * MLX5_BFREGS_PER_UAR +
+ bfregn % MLX5_NON_FP_BFREGS_PER_UAR;
+}
+
static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
struct mlx5_ib_qp *qp, struct ib_udata *udata,
struct ib_qp_init_attr *attr,
@@ -800,15 +781,15 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
/* In CROSS_CHANNEL CQ and QP must use the same UAR */
bfregn = MLX5_CROSS_CHANNEL_BFREG;
else {
- bfregn = alloc_bfreg(&context->bfregi, MLX5_IB_LATENCY_CLASS_HIGH);
+ bfregn = alloc_bfreg(dev, &context->bfregi, MLX5_IB_LATENCY_CLASS_HIGH);
if (bfregn < 0) {
mlx5_ib_dbg(dev, "failed to allocate low latency BFREG\n");
mlx5_ib_dbg(dev, "reverting to medium latency\n");
- bfregn = alloc_bfreg(&context->bfregi, MLX5_IB_LATENCY_CLASS_MEDIUM);
+ bfregn = alloc_bfreg(dev, &context->bfregi, MLX5_IB_LATENCY_CLASS_MEDIUM);
if (bfregn < 0) {
mlx5_ib_dbg(dev, "failed to allocate medium latency BFREG\n");
mlx5_ib_dbg(dev, "reverting to high latency\n");
- bfregn = alloc_bfreg(&context->bfregi, MLX5_IB_LATENCY_CLASS_LOW);
+ bfregn = alloc_bfreg(dev, &context->bfregi, MLX5_IB_LATENCY_CLASS_LOW);
if (bfregn < 0) {
mlx5_ib_warn(dev, "bfreg allocation failed\n");
return bfregn;
@@ -817,7 +798,7 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
}
}
- uar_index = bfregn_to_uar_index(&context->bfregi, bfregn);
+ uar_index = bfregn_to_uar_index(dev, &context->bfregi, bfregn);
mlx5_ib_dbg(dev, "bfregn 0x%x, uar_index 0x%x\n", bfregn, uar_index);
qp->rq.offset = 0;
@@ -858,7 +839,7 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
MLX5_SET(qpc, qpc, page_offset, offset);
MLX5_SET(qpc, qpc, uar_page, uar_index);
- resp->bfreg_index = bfregn;
+ resp->bfreg_index = adjust_bfregn(dev, &context->bfregi, bfregn);
qp->bfregn = bfregn;
err = mlx5_ib_db_map_user(context, ucmd.db_addr, &qp->db);
@@ -887,12 +868,12 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
ib_umem_release(ubuffer->umem);
err_bfreg:
- free_bfreg(&context->bfregi, bfregn);
+ free_bfreg(dev, &context->bfregi, bfregn);
return err;
}
-static void destroy_qp_user(struct ib_pd *pd, struct mlx5_ib_qp *qp,
- struct mlx5_ib_qp_base *base)
+static void destroy_qp_user(struct mlx5_ib_dev *dev, struct ib_pd *pd,
+ struct mlx5_ib_qp *qp, struct mlx5_ib_qp_base *base)
{
struct mlx5_ib_ucontext *context;
@@ -900,7 +881,7 @@ static void destroy_qp_user(struct ib_pd *pd, struct mlx5_ib_qp *qp,
mlx5_ib_db_unmap_user(context, &qp->db);
if (base->ubuffer.umem)
ib_umem_release(base->ubuffer.umem);
- free_bfreg(&context->bfregi, qp->bfregn);
+ free_bfreg(dev, &context->bfregi, qp->bfregn);
}
static int create_kernel_qp(struct mlx5_ib_dev *dev,
@@ -1784,7 +1765,7 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
err_create:
if (qp->create_type == MLX5_QP_USER)
- destroy_qp_user(pd, qp, base);
+ destroy_qp_user(dev, pd, qp, base);
else if (qp->create_type == MLX5_QP_KERNEL)
destroy_qp_kernel(dev, qp);
@@ -1962,7 +1943,7 @@ static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
if (qp->create_type == MLX5_QP_KERNEL)
destroy_qp_kernel(dev, qp);
else if (qp->create_type == MLX5_QP_USER)
- destroy_qp_user(&get_pd(qp)->ibpd, qp, base);
+ destroy_qp_user(dev, &get_pd(qp)->ibpd, qp, base);
}
static const char *ib_qp_type_str(enum ib_qp_type type)
diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h
index db1b928..dd345e8 100644
--- a/include/linux/mlx5/device.h
+++ b/include/linux/mlx5/device.h
@@ -212,6 +212,11 @@ enum {
};
enum {
+ MLX5_ADAPTER_PAGE_SHIFT = 12,
+ MLX5_ADAPTER_PAGE_SIZE = 1 << MLX5_ADAPTER_PAGE_SHIFT,
+};
+
+enum {
MLX5_BFREGS_PER_UAR = 4,
MLX5_MAX_UARS = 1 << 8,
MLX5_NON_FP_BFREGS_PER_UAR = 2,
@@ -219,6 +224,8 @@ enum {
MLX5_NON_FP_BFREGS_PER_UAR,
MLX5_MAX_BFREGS = MLX5_MAX_UARS *
MLX5_NON_FP_BFREGS_PER_UAR,
+ MLX5_UARS_IN_PAGE = PAGE_SIZE / MLX5_ADAPTER_PAGE_SIZE,
+ MLX5_NON_FP_BFREGS_IN_PAGE = MLX5_NON_FP_BFREGS_PER_UAR * MLX5_UARS_IN_PAGE,
};
enum {
@@ -392,11 +399,6 @@ enum {
};
enum {
- MLX5_ADAPTER_PAGE_SHIFT = 12,
- MLX5_ADAPTER_PAGE_SIZE = 1 << MLX5_ADAPTER_PAGE_SHIFT,
-};
-
-enum {
MLX5_CAP_OFF_CMDIF_CSUM = 46,
};
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index bb362f5..7e7394f 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -189,18 +189,17 @@ enum mlx5_eq_type {
};
struct mlx5_bfreg_info {
- struct mlx5_uar *uars;
- int num_uars;
+ u32 *sys_pages;
int num_low_latency_bfregs;
- unsigned long *bitmap;
unsigned int *count;
- struct mlx5_bf *bfs;
/*
* protect bfreg allocation data structs
*/
struct mutex lock;
u32 ver;
+ bool lib_uar_4k;
+ u32 num_sys_pages;
};
struct mlx5_cmd_first {
@@ -470,13 +469,10 @@ struct mlx5_sq_bfreg {
struct mlx5_uar {
u32 index;
- struct list_head bf_list;
- unsigned free_bf_bmap;
- void __iomem *bf_map;
void __iomem *map;
+ void __iomem *bf_map;
};
-
struct mlx5_core_health {
struct health_buffer __iomem *health;
__be32 __iomem *health_counter;
--
2.7.4
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply related
* [for-next 07/10] IB/mlx5: Use blue flame register allocator in mlx5_ib
From: Saeed Mahameed @ 2017-01-03 21:55 UTC (permalink / raw)
To: David S. Miller, Doug Ledford
Cc: netdev-u79uwXL29TY76Z2rM5mHXA, linux-rdma-u79uwXL29TY76Z2rM5mHXA,
Leon Romanovsky, Eli Cohen, Matan Barak, Leon Romanovsky,
Saeed Mahameed
In-Reply-To: <1483480528-22622-1-git-send-email-saeedm-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
From: Eli Cohen <eli-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Make use of the blue flame registers allocator at mlx5_ib. Since blue
flame was not really supported we remove all the code that is related to
blue flame and we let all consumers to use the same blue flame register.
Once blue flame is supported we will add the code. As part of this patch
we also move the definition of struct mlx5_bf to mlx5_ib.h as it is only
used by mlx5_ib.
Signed-off-by: Eli Cohen <eli-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Leon Romanovsky <leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
Signed-off-by: Saeed Mahameed <saeedm-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
drivers/infiniband/hw/mlx5/cq.c | 8 +-
drivers/infiniband/hw/mlx5/main.c | 28 +++++-
drivers/infiniband/hw/mlx5/mlx5_ib.h | 11 ++-
drivers/infiniband/hw/mlx5/qp.c | 73 +++------------
drivers/net/ethernet/mellanox/mlx5/core/Kconfig | 2 +-
drivers/net/ethernet/mellanox/mlx5/core/en.h | 4 +-
drivers/net/ethernet/mellanox/mlx5/core/main.c | 16 +---
drivers/net/ethernet/mellanox/mlx5/core/uar.c | 114 ------------------------
include/linux/mlx5/cq.h | 3 +-
include/linux/mlx5/doorbell.h | 32 +------
include/linux/mlx5/driver.h | 19 ----
11 files changed, 58 insertions(+), 252 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index bb7e91c..a28ec33 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -689,7 +689,7 @@ int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
{
struct mlx5_core_dev *mdev = to_mdev(ibcq->device)->mdev;
struct mlx5_ib_cq *cq = to_mcq(ibcq);
- void __iomem *uar_page = mdev->priv.bfregi.uars[0].map;
+ void __iomem *uar_page = mdev->priv.uar->map;
unsigned long irq_flags;
int ret = 0;
@@ -704,9 +704,7 @@ int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
mlx5_cq_arm(&cq->mcq,
(flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ?
MLX5_CQ_DB_REQ_NOT_SOL : MLX5_CQ_DB_REQ_NOT,
- uar_page,
- MLX5_GET_DOORBELL_LOCK(&mdev->priv.cq_uar_lock),
- to_mcq(ibcq)->mcq.cons_index);
+ uar_page, to_mcq(ibcq)->mcq.cons_index);
return ret;
}
@@ -886,7 +884,7 @@ static int create_cq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
MLX5_SET(cqc, cqc, log_page_size,
cq->buf.buf.page_shift - MLX5_ADAPTER_PAGE_SHIFT);
- *index = dev->mdev->priv.bfregi.uars[0].index;
+ *index = dev->mdev->priv.uar->index;
return 0;
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index d5cf82b..e9f0830 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -3074,8 +3074,6 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
if (mlx5_use_mad_ifc(dev))
get_ext_port_caps(dev);
- MLX5_INIT_DOORBELL_LOCK(&dev->uar_lock);
-
if (!mlx5_lag_is_active(mdev))
name = "mlx5_%d";
else
@@ -3251,9 +3249,21 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
if (err)
goto err_odp;
+ dev->mdev->priv.uar = mlx5_get_uars_page(dev->mdev);
+ if (!dev->mdev->priv.uar)
+ goto err_q_cnt;
+
+ err = mlx5_alloc_bfreg(dev->mdev, &dev->bfreg, false, false);
+ if (err)
+ goto err_uar_page;
+
+ err = mlx5_alloc_bfreg(dev->mdev, &dev->fp_bfreg, false, true);
+ if (err)
+ goto err_bfreg;
+
err = ib_register_device(&dev->ib_dev, NULL);
if (err)
- goto err_q_cnt;
+ goto err_fp_bfreg;
err = create_umr_res(dev);
if (err)
@@ -3276,6 +3286,15 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
err_dev:
ib_unregister_device(&dev->ib_dev);
+err_fp_bfreg:
+ mlx5_free_bfreg(dev->mdev, &dev->fp_bfreg);
+
+err_bfreg:
+ mlx5_free_bfreg(dev->mdev, &dev->bfreg);
+
+err_uar_page:
+ mlx5_put_uars_page(dev->mdev, dev->mdev->priv.uar);
+
err_q_cnt:
mlx5_ib_dealloc_q_counters(dev);
@@ -3307,6 +3326,9 @@ static void mlx5_ib_remove(struct mlx5_core_dev *mdev, void *context)
mlx5_remove_netdev_notifier(dev);
ib_unregister_device(&dev->ib_dev);
+ mlx5_free_bfreg(dev->mdev, &dev->fp_bfreg);
+ mlx5_free_bfreg(dev->mdev, &dev->bfreg);
+ mlx5_put_uars_page(dev->mdev, mdev->priv.uar);
mlx5_ib_dealloc_q_counters(dev);
destroy_umrc_res(dev);
mlx5_ib_odp_remove_one(dev);
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index d4d1329..ae3bc4a 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -324,6 +324,12 @@ struct mlx5_ib_raw_packet_qp {
struct mlx5_ib_rq rq;
};
+struct mlx5_bf {
+ int buf_size;
+ unsigned long offset;
+ struct mlx5_sq_bfreg *bfreg;
+};
+
struct mlx5_ib_qp {
struct ib_qp ibqp;
union {
@@ -349,7 +355,7 @@ struct mlx5_ib_qp {
int wq_sig;
int scat_cqe;
int max_inline_data;
- struct mlx5_bf *bf;
+ struct mlx5_bf bf;
int has_rq;
/* only for user space QPs. For kernel
@@ -591,7 +597,6 @@ struct mlx5_ib_dev {
struct ib_device ib_dev;
struct mlx5_core_dev *mdev;
struct mlx5_roce roce;
- MLX5_DECLARE_DOORBELL_LOCK(uar_lock);
int num_ports;
/* serialize update of capability mask
*/
@@ -621,6 +626,8 @@ struct mlx5_ib_dev {
struct list_head qp_list;
/* Array with num_ports elements */
struct mlx5_ib_port *port;
+ struct mlx5_sq_bfreg bfreg;
+ struct mlx5_sq_bfreg fp_bfreg;
};
static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 240fbb0..b38240e 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -909,14 +909,10 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
u32 **in, int *inlen,
struct mlx5_ib_qp_base *base)
{
- enum mlx5_ib_latency_class lc = MLX5_IB_LATENCY_CLASS_LOW;
- struct mlx5_bfreg_info *bfregi;
int uar_index;
void *qpc;
- int bfregn;
int err;
- bfregi = &dev->mdev->priv.bfregi;
if (init_attr->create_flags & ~(IB_QP_CREATE_SIGNATURE_EN |
IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK |
IB_QP_CREATE_IPOIB_UD_LSO |
@@ -924,21 +920,17 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
return -EINVAL;
if (init_attr->qp_type == MLX5_IB_QPT_REG_UMR)
- lc = MLX5_IB_LATENCY_CLASS_FAST_PATH;
-
- bfregn = alloc_bfreg(bfregi, lc);
- if (bfregn < 0) {
- mlx5_ib_dbg(dev, "\n");
- return -ENOMEM;
- }
+ qp->bf.bfreg = &dev->fp_bfreg;
+ else
+ qp->bf.bfreg = &dev->bfreg;
- qp->bf = &bfregi->bfs[bfregn];
- uar_index = qp->bf->uar->index;
+ qp->bf.buf_size = 1 << MLX5_CAP_GEN(dev->mdev, log_bf_reg_size);
+ uar_index = qp->bf.bfreg->index;
err = calc_sq_size(dev, init_attr, qp);
if (err < 0) {
mlx5_ib_dbg(dev, "err %d\n", err);
- goto err_bfreg;
+ return err;
}
qp->rq.offset = 0;
@@ -948,7 +940,7 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
err = mlx5_buf_alloc(dev->mdev, base->ubuffer.buf_size, &qp->buf);
if (err) {
mlx5_ib_dbg(dev, "err %d\n", err);
- goto err_bfreg;
+ return err;
}
qp->sq.qend = mlx5_get_send_wqe(qp, qp->sq.wqe_cnt);
@@ -1010,9 +1002,6 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev,
err_buf:
mlx5_buf_free(dev->mdev, &qp->buf);
-
-err_bfreg:
- free_bfreg(&dev->mdev->priv.bfregi, bfregn);
return err;
}
@@ -1025,7 +1014,6 @@ static void destroy_qp_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
kfree(qp->rq.wrid);
mlx5_db_free(dev->mdev, &qp->db);
mlx5_buf_free(dev->mdev, &qp->buf);
- free_bfreg(&dev->mdev->priv.bfregi, qp->bf->bfregn);
}
static u32 get_rx_type(struct mlx5_ib_qp *qp, struct ib_qp_init_attr *attr)
@@ -3744,24 +3732,6 @@ static void dump_wqe(struct mlx5_ib_qp *qp, int idx, int size_16)
}
}
-static void mlx5_bf_copy(u64 __iomem *dst, u64 *src,
- unsigned bytecnt, struct mlx5_ib_qp *qp)
-{
- while (bytecnt > 0) {
- __iowrite64_copy(dst++, src++, 8);
- __iowrite64_copy(dst++, src++, 8);
- __iowrite64_copy(dst++, src++, 8);
- __iowrite64_copy(dst++, src++, 8);
- __iowrite64_copy(dst++, src++, 8);
- __iowrite64_copy(dst++, src++, 8);
- __iowrite64_copy(dst++, src++, 8);
- __iowrite64_copy(dst++, src++, 8);
- bytecnt -= 64;
- if (unlikely(src == qp->sq.qend))
- src = mlx5_get_send_wqe(qp, 0);
- }
-}
-
static u8 get_fence(u8 fence, struct ib_send_wr *wr)
{
if (unlikely(wr->opcode == IB_WR_LOCAL_INV &&
@@ -3857,7 +3827,7 @@ int mlx5_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
return mlx5_ib_gsi_post_send(ibqp, wr, bad_wr);
qp = to_mqp(ibqp);
- bf = qp->bf;
+ bf = &qp->bf;
qend = qp->sq.qend;
spin_lock_irqsave(&qp->sq.lock, flags);
@@ -4130,28 +4100,13 @@ int mlx5_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
* we hit doorbell */
wmb();
- if (bf->need_lock)
- spin_lock(&bf->lock);
- else
- __acquire(&bf->lock);
-
- /* TBD enable WC */
- if (0 && nreq == 1 && bf->bfregn && inl && size > 1 && size <= bf->buf_size / 16) {
- mlx5_bf_copy(bf->reg + bf->offset, (u64 *)ctrl, ALIGN(size * 16, 64), qp);
- /* wc_wmb(); */
- } else {
- mlx5_write64((__be32 *)ctrl, bf->regreg + bf->offset,
- MLX5_GET_DOORBELL_LOCK(&bf->lock32));
- /* Make sure doorbells don't leak out of SQ spinlock
- * and reach the HCA out of order.
- */
- mmiowb();
- }
+ /* currently we support only regular doorbells */
+ mlx5_write64((__be32 *)ctrl, bf->bfreg->map + bf->offset);
+ /* Make sure doorbells don't leak out of SQ spinlock
+ * and reach the HCA out of order.
+ */
+ mmiowb();
bf->offset ^= bf->buf_size;
- if (bf->need_lock)
- spin_unlock(&bf->lock);
- else
- __release(&bf->lock);
}
spin_unlock_irqrestore(&qp->sq.lock, flags);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/Kconfig b/drivers/net/ethernet/mellanox/mlx5/core/Kconfig
index ddb4ca4..39505ac 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/Kconfig
+++ b/drivers/net/ethernet/mellanox/mlx5/core/Kconfig
@@ -5,7 +5,7 @@
config MLX5_CORE
tristate "Mellanox Technologies ConnectX-4 and Connect-IB core driver"
depends on MAY_USE_DEVLINK
- depends on PCI
+ depends on PCI && 64BIT
default n
---help---
Core driver for low level functionality of the ConnectX-4 and
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 951dbd5..e3ef0b5 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -820,7 +820,7 @@ static inline void mlx5e_tx_notify_hw(struct mlx5e_sq *sq,
if (bf_sz)
__iowrite64_copy(sq->uar_map + ofst, ctrl, bf_sz);
else
- mlx5_write64((__be32 *)ctrl, sq->uar_map + ofst, NULL);
+ mlx5_write64((__be32 *)ctrl, sq->uar_map + ofst);
/* flush the write-combining mapped buffer */
wmb();
@@ -832,7 +832,7 @@ static inline void mlx5e_cq_arm(struct mlx5e_cq *cq)
struct mlx5_core_cq *mcq;
mcq = &cq->mcq;
- mlx5_cq_arm(mcq, MLX5_CQ_DB_REQ_NOT, mcq->uar->map, NULL, cq->wq.cc);
+ mlx5_cq_arm(mcq, MLX5_CQ_DB_REQ_NOT, mcq->uar->map, cq->wq.cc);
}
static inline u32 mlx5e_get_wqe_mtt_offset(struct mlx5e_rq *rq, u16 wqe_ix)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index 2882d04..ff1f144 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -913,8 +913,6 @@ static int mlx5_init_once(struct mlx5_core_dev *dev, struct mlx5_priv *priv)
goto out;
}
- MLX5_INIT_DOORBELL_LOCK(&priv->cq_uar_lock);
-
err = mlx5_init_cq_table(dev);
if (err) {
dev_err(&pdev->dev, "failed to initialize cq table\n");
@@ -1099,16 +1097,10 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv,
goto err_disable_msix;
}
- err = mlx5_alloc_bfregs(dev, &priv->bfregi);
- if (err) {
- dev_err(&pdev->dev, "Failed allocating uuars, aborting\n");
- goto err_uar_cleanup;
- }
-
err = mlx5_start_eqs(dev);
if (err) {
dev_err(&pdev->dev, "Failed to start pages and async EQs\n");
- goto err_free_uar;
+ goto err_put_uars;
}
err = alloc_comp_eqs(dev);
@@ -1174,10 +1166,7 @@ static int mlx5_load_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv,
err_stop_eqs:
mlx5_stop_eqs(dev);
-err_free_uar:
- mlx5_free_bfregs(dev, &priv->bfregi);
-
-err_uar_cleanup:
+err_put_uars:
mlx5_put_uars_page(dev, priv->uar);
err_disable_msix:
@@ -1238,7 +1227,6 @@ static int mlx5_unload_one(struct mlx5_core_dev *dev, struct mlx5_priv *priv,
mlx5_irq_clear_affinity_hints(dev);
free_comp_eqs(dev);
mlx5_stop_eqs(dev);
- mlx5_free_bfregs(dev, &priv->bfregi);
mlx5_put_uars_page(dev, priv->uar);
mlx5_disable_msix(dev);
if (cleanup)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/uar.c b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
index fcc0270..07b273c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/uar.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/uar.c
@@ -67,120 +67,6 @@ int mlx5_cmd_free_uar(struct mlx5_core_dev *dev, u32 uarn)
}
EXPORT_SYMBOL(mlx5_cmd_free_uar);
-static int need_bfreg_lock(int bfregn)
-{
- int tot_bfregs = NUM_DRIVER_UARS * MLX5_BFREGS_PER_UAR;
-
- if (bfregn == 0 || tot_bfregs - NUM_LOW_LAT_BFREGS)
- return 0;
-
- return 1;
-}
-
-int mlx5_alloc_bfregs(struct mlx5_core_dev *dev, struct mlx5_bfreg_info *bfregi)
-{
- int tot_bfregs = NUM_DRIVER_UARS * MLX5_BFREGS_PER_UAR;
- struct mlx5_bf *bf;
- phys_addr_t addr;
- int err;
- int i;
-
- bfregi->num_uars = NUM_DRIVER_UARS;
- bfregi->num_low_latency_bfregs = NUM_LOW_LAT_BFREGS;
-
- mutex_init(&bfregi->lock);
- bfregi->uars = kcalloc(bfregi->num_uars, sizeof(*bfregi->uars), GFP_KERNEL);
- if (!bfregi->uars)
- return -ENOMEM;
-
- bfregi->bfs = kcalloc(tot_bfregs, sizeof(*bfregi->bfs), GFP_KERNEL);
- if (!bfregi->bfs) {
- err = -ENOMEM;
- goto out_uars;
- }
-
- bfregi->bitmap = kcalloc(BITS_TO_LONGS(tot_bfregs), sizeof(*bfregi->bitmap),
- GFP_KERNEL);
- if (!bfregi->bitmap) {
- err = -ENOMEM;
- goto out_bfs;
- }
-
- bfregi->count = kcalloc(tot_bfregs, sizeof(*bfregi->count), GFP_KERNEL);
- if (!bfregi->count) {
- err = -ENOMEM;
- goto out_bitmap;
- }
-
- for (i = 0; i < bfregi->num_uars; i++) {
- err = mlx5_cmd_alloc_uar(dev, &bfregi->uars[i].index);
- if (err)
- goto out_count;
-
- addr = dev->iseg_base + ((phys_addr_t)(bfregi->uars[i].index) << PAGE_SHIFT);
- bfregi->uars[i].map = ioremap(addr, PAGE_SIZE);
- if (!bfregi->uars[i].map) {
- mlx5_cmd_free_uar(dev, bfregi->uars[i].index);
- err = -ENOMEM;
- goto out_count;
- }
- mlx5_core_dbg(dev, "allocated uar index 0x%x, mmaped at %p\n",
- bfregi->uars[i].index, bfregi->uars[i].map);
- }
-
- for (i = 0; i < tot_bfregs; i++) {
- bf = &bfregi->bfs[i];
-
- bf->buf_size = (1 << MLX5_CAP_GEN(dev, log_bf_reg_size)) / 2;
- bf->uar = &bfregi->uars[i / MLX5_BFREGS_PER_UAR];
- bf->regreg = bfregi->uars[i / MLX5_BFREGS_PER_UAR].map;
- bf->reg = NULL; /* Add WC support */
- bf->offset = (i % MLX5_BFREGS_PER_UAR) *
- (1 << MLX5_CAP_GEN(dev, log_bf_reg_size)) +
- MLX5_BF_OFFSET;
- bf->need_lock = need_bfreg_lock(i);
- spin_lock_init(&bf->lock);
- spin_lock_init(&bf->lock32);
- bf->bfregn = i;
- }
-
- return 0;
-
-out_count:
- for (i--; i >= 0; i--) {
- iounmap(bfregi->uars[i].map);
- mlx5_cmd_free_uar(dev, bfregi->uars[i].index);
- }
- kfree(bfregi->count);
-
-out_bitmap:
- kfree(bfregi->bitmap);
-
-out_bfs:
- kfree(bfregi->bfs);
-
-out_uars:
- kfree(bfregi->uars);
- return err;
-}
-
-int mlx5_free_bfregs(struct mlx5_core_dev *dev, struct mlx5_bfreg_info *bfregi)
-{
- int i = bfregi->num_uars;
-
- for (i--; i >= 0; i--) {
- iounmap(bfregi->uars[i].map);
- mlx5_cmd_free_uar(dev, bfregi->uars[i].index);
- }
-
- kfree(bfregi->count);
- kfree(bfregi->bitmap);
- kfree(bfregi->bfs);
- kfree(bfregi->uars);
-
- return 0;
-}
-
int mlx5_alloc_map_uar(struct mlx5_core_dev *mdev, struct mlx5_uar *uar,
bool map_wc)
{
diff --git a/include/linux/mlx5/cq.h b/include/linux/mlx5/cq.h
index 7c3c0d3..1ed50f7 100644
--- a/include/linux/mlx5/cq.h
+++ b/include/linux/mlx5/cq.h
@@ -144,7 +144,6 @@ enum {
static inline void mlx5_cq_arm(struct mlx5_core_cq *cq, u32 cmd,
void __iomem *uar_page,
- spinlock_t *doorbell_lock,
u32 cons_index)
{
__be32 doorbell[2];
@@ -164,7 +163,7 @@ static inline void mlx5_cq_arm(struct mlx5_core_cq *cq, u32 cmd,
doorbell[0] = cpu_to_be32(sn << 28 | cmd | ci);
doorbell[1] = cpu_to_be32(cq->cqn);
- mlx5_write64(doorbell, uar_page + MLX5_CQ_DOORBELL, doorbell_lock);
+ mlx5_write64(doorbell, uar_page + MLX5_CQ_DOORBELL);
}
int mlx5_init_cq_table(struct mlx5_core_dev *dev);
diff --git a/include/linux/mlx5/doorbell.h b/include/linux/mlx5/doorbell.h
index afc78a3..9aa84a2 100644
--- a/include/linux/mlx5/doorbell.h
+++ b/include/linux/mlx5/doorbell.h
@@ -36,44 +36,14 @@
#define MLX5_BF_OFFSET 0x800
#define MLX5_CQ_DOORBELL 0x20
-#if BITS_PER_LONG == 64
/* Assume that we can just write a 64-bit doorbell atomically. s390
* actually doesn't have writeq() but S/390 systems don't even have
* PCI so we won't worry about it.
*/
-#define MLX5_DECLARE_DOORBELL_LOCK(name)
-#define MLX5_INIT_DOORBELL_LOCK(ptr) do { } while (0)
-#define MLX5_GET_DOORBELL_LOCK(ptr) (NULL)
-
-static inline void mlx5_write64(__be32 val[2], void __iomem *dest,
- spinlock_t *doorbell_lock)
+static inline void mlx5_write64(__be32 val[2], void __iomem *dest)
{
__raw_writeq(*(u64 *)val, dest);
}
-#else
-
-/* Just fall back to a spinlock to protect the doorbell if
- * BITS_PER_LONG is 32 -- there's no portable way to do atomic 64-bit
- * MMIO writes.
- */
-
-#define MLX5_DECLARE_DOORBELL_LOCK(name) spinlock_t name;
-#define MLX5_INIT_DOORBELL_LOCK(ptr) spin_lock_init(ptr)
-#define MLX5_GET_DOORBELL_LOCK(ptr) (ptr)
-
-static inline void mlx5_write64(__be32 val[2], void __iomem *dest,
- spinlock_t *doorbell_lock)
-{
- unsigned long flags;
-
- spin_lock_irqsave(doorbell_lock, flags);
- __raw_writel((__force u32) val[0], dest);
- __raw_writel((__force u32) val[1], dest + 4);
- spin_unlock_irqrestore(doorbell_lock, flags);
-}
-
-#endif
-
#endif /* MLX5_DOORBELL_H */
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 9a3a095..bb362f5 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -203,23 +203,6 @@ struct mlx5_bfreg_info {
u32 ver;
};
-struct mlx5_bf {
- void __iomem *reg;
- void __iomem *regreg;
- int buf_size;
- struct mlx5_uar *uar;
- unsigned long offset;
- int need_lock;
- /* protect blue flame buffer selection when needed
- */
- spinlock_t lock;
-
- /* serialize 64 bit writes when done as two 32 bit accesses
- */
- spinlock_t lock32;
- int bfregn;
-};
-
struct mlx5_cmd_first {
__be32 data[4];
};
@@ -612,8 +595,6 @@ struct mlx5_priv {
struct mlx5_eq_table eq_table;
struct msix_entry *msix_arr;
struct mlx5_irq_info *irq_info;
- struct mlx5_bfreg_info bfregi;
- MLX5_DECLARE_DOORBELL_LOCK(cq_uar_lock);
/* pages stuff */
struct workqueue_struct *pg_wq;
--
2.7.4
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ 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