Age | Commit message (Collapse) | Author | Files | Lines |
|
Drop the smc_sendpage() code as smc_sendmsg() just passes the call down to
the underlying TCP socket and smc_tx_sendpage() is just a wrapper around
its sendmsg implementation.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Karsten Graul <kgraul@linux.ibm.com>
cc: Wenjia Zhang <wenjia@linux.ibm.com>
cc: Jan Karcher <jaka@linux.ibm.com>
cc: "D. Wythe" <alibuda@linux.alibaba.com>
cc: Tony Lu <tonylu@linux.alibaba.com>
cc: Wen Gu <guwen@linux.alibaba.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
Link: https://lore.kernel.org/r/20230623225513.2732256-10-dhowells@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
As MSG_SENDPAGE_NOTLAST is being phased out along with sendpage(), don't
use it further in than the sendpage methods, but rather translate it to
MSG_MORE and use that instead.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
cc: Bernard Metzler <bmt@zurich.ibm.com>
cc: Jason Gunthorpe <jgg@ziepe.ca>
cc: Leon Romanovsky <leon@kernel.org>
cc: John Fastabend <john.fastabend@gmail.com>
cc: Jakub Sitnicki <jakub@cloudflare.com>
cc: David Ahern <dsahern@kernel.org>
cc: Karsten Graul <kgraul@linux.ibm.com>
cc: Wenjia Zhang <wenjia@linux.ibm.com>
cc: Jan Karcher <jaka@linux.ibm.com>
cc: "D. Wythe" <alibuda@linux.alibaba.com>
cc: Tony Lu <tonylu@linux.alibaba.com>
cc: Wen Gu <guwen@linux.alibaba.com>
cc: Boris Pismenny <borisp@nvidia.com>
cc: Steffen Klassert <steffen.klassert@secunet.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
Link: https://lore.kernel.org/r/20230623225513.2732256-2-dhowells@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
__condition is evaluated twice in sk_wait_event() macro.
First invocation is lockless, and reads can race with writes,
as spotted by syzbot.
BUG: KCSAN: data-race in sk_stream_wait_connect / tcp_disconnect
write to 0xffff88812d83d6a0 of 4 bytes by task 9065 on cpu 1:
tcp_disconnect+0x2cd/0xdb0
inet_shutdown+0x19e/0x1f0 net/ipv4/af_inet.c:911
__sys_shutdown_sock net/socket.c:2343 [inline]
__sys_shutdown net/socket.c:2355 [inline]
__do_sys_shutdown net/socket.c:2363 [inline]
__se_sys_shutdown+0xf8/0x140 net/socket.c:2361
__x64_sys_shutdown+0x31/0x40 net/socket.c:2361
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x41/0xc0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
read to 0xffff88812d83d6a0 of 4 bytes by task 9040 on cpu 0:
sk_stream_wait_connect+0x1de/0x3a0 net/core/stream.c:75
tcp_sendmsg_locked+0x2e4/0x2120 net/ipv4/tcp.c:1266
tcp_sendmsg+0x30/0x50 net/ipv4/tcp.c:1484
inet6_sendmsg+0x63/0x80 net/ipv6/af_inet6.c:651
sock_sendmsg_nosec net/socket.c:724 [inline]
sock_sendmsg net/socket.c:747 [inline]
__sys_sendto+0x246/0x300 net/socket.c:2142
__do_sys_sendto net/socket.c:2154 [inline]
__se_sys_sendto net/socket.c:2150 [inline]
__x64_sys_sendto+0x78/0x90 net/socket.c:2150
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x41/0xc0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
value changed: 0x00000000 -> 0x00000068
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
READ/WRITE proved to be actively confusing - the meanings are
"data destination, as used with read(2)" and "data source, as
used with write(2)", but people keep interpreting those as
"we read data from it" and "we write data to it", i.e. exactly
the wrong way.
Call them ITER_DEST and ITER_SOURCE - at least that is harder
to misinterpret...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
|
|
This struct is used in a single place only, and its usage generates
inefficient code. Time to clean up!
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-and-tested-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Wenjia Zhang < wenjia@linux.ibm.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
On long-running enterprise production servers, high-order contiguous
memory pages are usually very rare and in most cases we can only get
fragmented pages.
When replacing TCP with SMC-R in such production scenarios, attempting
to allocate high-order physically contiguous sndbufs and RMBs may result
in frequent memory compaction, which will cause unexpected hung issue
and further stability risks.
So this patch is aimed to allow SMC-R link group to use virtually
contiguous sndbufs and RMBs to avoid potential issues mentioned above.
Whether to use physically or virtually contiguous buffers can be set
by sysctl smcr_buf_type.
Note that using virtually contiguous buffers will bring an acceptable
performance regression, which can be mainly divided into two parts:
1) regression in data path, which is brought by additional address
translation of sndbuf by RNIC in Tx. But in general, translating
address through MTT is fast.
Taking 256KB sndbuf and RMB as an example, the comparisons in qperf
latency and bandwidth test with physically and virtually contiguous
buffers are as follows:
- client:
smc_run taskset -c <cpu> qperf <server> -oo msg_size:1:64K:*2\
-t 5 -vu tcp_{bw|lat}
- server:
smc_run taskset -c <cpu> qperf
[latency]
msgsize tcp smcr smcr-use-virt-buf
1 11.17 us 7.56 us 7.51 us (-0.67%)
2 10.65 us 7.74 us 7.56 us (-2.31%)
4 11.11 us 7.52 us 7.59 us ( 0.84%)
8 10.83 us 7.55 us 7.51 us (-0.48%)
16 11.21 us 7.46 us 7.51 us ( 0.71%)
32 10.65 us 7.53 us 7.58 us ( 0.61%)
64 10.95 us 7.74 us 7.80 us ( 0.76%)
128 11.14 us 7.83 us 7.87 us ( 0.47%)
256 10.97 us 7.94 us 7.92 us (-0.28%)
512 11.23 us 7.94 us 8.20 us ( 3.25%)
1024 11.60 us 8.12 us 8.20 us ( 0.96%)
2048 14.04 us 8.30 us 8.51 us ( 2.49%)
4096 16.88 us 9.13 us 9.07 us (-0.64%)
8192 22.50 us 10.56 us 11.22 us ( 6.26%)
16384 28.99 us 12.88 us 13.83 us ( 7.37%)
32768 40.13 us 16.76 us 16.95 us ( 1.16%)
65536 68.70 us 24.68 us 24.85 us ( 0.68%)
[bandwidth]
msgsize tcp smcr smcr-use-virt-buf
1 1.65 MB/s 1.59 MB/s 1.53 MB/s (-3.88%)
2 3.32 MB/s 3.17 MB/s 3.08 MB/s (-2.67%)
4 6.66 MB/s 6.33 MB/s 6.09 MB/s (-3.85%)
8 13.67 MB/s 13.45 MB/s 11.97 MB/s (-10.99%)
16 25.36 MB/s 27.15 MB/s 24.16 MB/s (-11.01%)
32 48.22 MB/s 54.24 MB/s 49.41 MB/s (-8.89%)
64 106.79 MB/s 107.32 MB/s 99.05 MB/s (-7.71%)
128 210.21 MB/s 202.46 MB/s 201.02 MB/s (-0.71%)
256 400.81 MB/s 416.81 MB/s 393.52 MB/s (-5.59%)
512 746.49 MB/s 834.12 MB/s 809.99 MB/s (-2.89%)
1024 1292.33 MB/s 1641.96 MB/s 1571.82 MB/s (-4.27%)
2048 2007.64 MB/s 2760.44 MB/s 2717.68 MB/s (-1.55%)
4096 2665.17 MB/s 4157.44 MB/s 4070.76 MB/s (-2.09%)
8192 3159.72 MB/s 4361.57 MB/s 4270.65 MB/s (-2.08%)
16384 4186.70 MB/s 4574.13 MB/s 4501.17 MB/s (-1.60%)
32768 4093.21 MB/s 4487.42 MB/s 4322.43 MB/s (-3.68%)
65536 4057.14 MB/s 4735.61 MB/s 4555.17 MB/s (-3.81%)
2) regression in buffer initialization and destruction path, which is
brought by additional MR operations of sndbufs. But thanks to link
group buffer reuse mechanism, the impact of this kind of regression
decreases as times of buffer reuse increases.
Taking 256KB sndbuf and RMB as an example, latency of some key SMC-R
buffer-related function obtained by bpftrace are as follows:
Function Phys-bufs Virt-bufs
smcr_new_buf_create() 67154 ns 79164 ns
smc_ib_buf_map_sg() 525 ns 928 ns
smc_ib_get_memory_region() 162294 ns 161191 ns
smc_wr_reg_send() 9957 ns 9635 ns
smc_ib_put_memory_region() 203548 ns 198374 ns
smc_ib_buf_unmap_sg() 508 ns 1158 ns
------------
Test environment notes:
1. Above tests run on 2 VMs within the same Host.
2. The NIC is ConnectX-4Lx, using SRIOV and passing through 2 VFs to
the each VM respectively.
3. VMs' vCPUs are binded to different physical CPUs, and the binded
physical CPUs are isolated by `isolcpus=xxx` cmdline.
4. NICs' queue number are set to 1.
Signed-off-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
smc_ib_sync_sg_for_cpu/device are the ops used for dma memory cache
consistency. Smc sndbufs are dma buffers, where CPU writes data to
it and PCIE device reads data from it. So for sndbufs,
smc_ib_sync_sg_for_device is needed and smc_ib_sync_sg_for_cpu is
redundant as PCIE device will not write the buffers. Smc rmbs
are dma buffers, where PCIE device write data to it and CPU read
data from it. So for rmbs, smc_ib_sync_sg_for_cpu is needed and
smc_ib_sync_sg_for_device is redundant as CPU will not write the buffers.
Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Rdma write with inline flag when sending small packages,
whose length is shorter than the qp's max_inline_data, can
help reducing latency.
In my test environment, which are 2 VMs running on the same
physical host and whose NICs(ConnectX-4Lx) are working on
SR-IOV mode, qperf shows 0.5us-0.7us improvement in latency.
Test command:
server: smc_run taskset -c 1 qperf
client: smc_run taskset -c 1 qperf <server ip> -oo \
msg_size:1:2K:*2 -t 30 -vu tcp_lat
The results shown below:
msgsize before after
1B 11.2 us 10.6 us (-0.6 us)
2B 11.2 us 10.7 us (-0.5 us)
4B 11.3 us 10.7 us (-0.6 us)
8B 11.2 us 10.6 us (-0.6 us)
16B 11.3 us 10.7 us (-0.6 us)
32B 11.3 us 10.6 us (-0.7 us)
64B 11.2 us 11.2 us (0 us)
128B 11.2 us 11.2 us (0 us)
256B 11.2 us 11.2 us (0 us)
512B 11.4 us 11.3 us (-0.1 us)
1KB 11.4 us 11.5 us (0.1 us)
2KB 11.5 us 11.5 us (0 us)
Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Tested-by: kernel test robot <lkp@intel.com>
Acked-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Fix build:
include/linux/minmax.h:45:25: note: in expansion of macro ‘__careful_cmp’
45 | #define min(x, y) __careful_cmp(x, y, <)
| ^~~~~~~~~~~~~
net/smc/smc_tx.c:150:24: note: in expansion of macro ‘min’
150 | corking_size = min(sock_net(&smc->sk)->smc.sysctl_autocorking_size,
| ^~~
Fixes: 12bbb0d163a9 ("net/smc: add sysctl for autocorking")
Link: https://lore.kernel.org/r/20220301222446.1271127-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This add a new sysctl: net.smc.autocorking_size
We can dynamically change the behaviour of autocorking
by change the value of autocorking_size.
Setting to 0 disables autocorking in SMC
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds autocorking support for SMC which could improve
throughput for small message by x3+.
The main idea is borrowed from TCP autocorking with some RDMA
specific modification:
1. The first message should never cork to make sure we won't
bring extra latency
2. If we have posted any Tx WRs to the NIC that have not
completed, cork the new messages until:
a) Receive CQE for the last Tx WR
b) We have corked enough message on the connection
3. Try to push the corked data out when we receive CQE of
the last Tx WR to prevent the corked messages hang in
the send queue.
Both SMC autocorking and TCP autocorking check the TX completion
to decide whether we should cork or not. The difference is
when we got a SMC Tx WR completion, the data have been confirmed
by the RNIC while TCP TX completion just tells us the data
have been sent out by the local NIC.
Add an atomic variable tx_pushing in smc_connection to make
sure only one can send to let it cork more and save CDC slot.
SMC autocorking should not bring extra latency since the first
message will always been sent out immediately.
The qperf tcp_bw test shows more than x4 increase under small
message size with Mellanox connectX4-Lx, same result with other
throughput benchmarks like sockperf/netperf.
The qperf tcp_lat test shows SMC autocorking has not increase any
ping-pong latency.
Test command:
client: smc_run taskset -c 1 qperf smc-server -oo msg_size:1:64K:*2 \
-t 30 -vu tcp_{bw|lat}
server: smc_run taskset -c 1 qperf
=== Bandwidth ====
MsgSize(Bytes) SMC-NoCork TCP SMC-AutoCorking
1 0.578 MB/s 2.392 MB/s(313.57%) 2.647 MB/s(357.72%)
2 1.159 MB/s 4.780 MB/s(312.53%) 5.153 MB/s(344.71%)
4 2.283 MB/s 10.266 MB/s(349.77%) 10.363 MB/s(354.02%)
8 4.668 MB/s 19.040 MB/s(307.86%) 21.215 MB/s(354.45%)
16 9.147 MB/s 38.904 MB/s(325.31%) 41.740 MB/s(356.32%)
32 18.369 MB/s 79.587 MB/s(333.25%) 82.392 MB/s(348.52%)
64 36.562 MB/s 148.668 MB/s(306.61%) 161.564 MB/s(341.89%)
128 72.961 MB/s 274.913 MB/s(276.80%) 325.363 MB/s(345.94%)
256 144.705 MB/s 512.059 MB/s(253.86%) 633.743 MB/s(337.96%)
512 288.873 MB/s 884.977 MB/s(206.35%) 1250.681 MB/s(332.95%)
1024 574.180 MB/s 1337.736 MB/s(132.98%) 2246.121 MB/s(291.19%)
2048 1095.192 MB/s 1865.952 MB/s( 70.38%) 2057.767 MB/s( 87.89%)
4096 2066.157 MB/s 2380.337 MB/s( 15.21%) 2173.983 MB/s( 5.22%)
8192 3717.198 MB/s 2733.073 MB/s(-26.47%) 3491.223 MB/s( -6.08%)
16384 4742.221 MB/s 2958.693 MB/s(-37.61%) 4637.692 MB/s( -2.20%)
32768 5349.550 MB/s 3061.285 MB/s(-42.77%) 5385.796 MB/s( 0.68%)
65536 5162.919 MB/s 3731.408 MB/s(-27.73%) 5223.890 MB/s( 1.18%)
==== Latency ====
MsgSize(Bytes) SMC-NoCork TCP SMC-AutoCorking
1 10.540 us 11.938 us( 13.26%) 10.573 us( 0.31%)
2 10.996 us 11.992 us( 9.06%) 10.269 us( -6.61%)
4 10.229 us 11.687 us( 14.25%) 10.240 us( 0.11%)
8 10.203 us 11.653 us( 14.21%) 10.402 us( 1.95%)
16 10.530 us 11.313 us( 7.44%) 10.599 us( 0.66%)
32 10.241 us 11.586 us( 13.13%) 10.223 us( -0.18%)
64 10.693 us 11.652 us( 8.97%) 10.251 us( -4.13%)
128 10.597 us 11.579 us( 9.27%) 10.494 us( -0.97%)
256 10.409 us 11.957 us( 14.87%) 10.710 us( 2.89%)
512 11.088 us 12.505 us( 12.78%) 10.547 us( -4.88%)
1024 11.240 us 12.255 us( 9.03%) 10.787 us( -4.03%)
2048 11.485 us 16.970 us( 47.76%) 11.256 us( -1.99%)
4096 12.077 us 13.948 us( 15.49%) 12.230 us( 1.27%)
8192 13.683 us 16.693 us( 22.00%) 13.786 us( 0.75%)
16384 16.470 us 23.615 us( 43.38%) 16.459 us( -0.07%)
32768 22.540 us 40.966 us( 81.75%) 23.284 us( 3.30%)
65536 34.192 us 73.003 us(113.51%) 34.233 us( 0.12%)
With SMC autocorking support, we can archive better throughput
than TCP in most message sizes without any latency trade-off.
Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This also calls trace_smc_tx_sendmsg() even if data is corked. For ease
of understanding, if statements are not expanded here.
Link: https://lore.kernel.org/all/f4166712-9a1e-51a0-409d-b7df25a66c52@linux.ibm.com/
Fixes: 139653bc6635 ("net/smc: Remove corked dealyed work")
Suggested-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The previous patch introduces a lock-free version of smc_tx_work() to
solve unnecessary lock contention, which is expected to be held lock.
So this adds comment to remind people to keep an eye out for locks.
Suggested-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This introduces a new corked flag, MSG_SENDPAGE_NOTLAST, which is
involved in syscall sendfile() [1], it indicates this is not the last
page. So we can cork the data until the page is not specify this flag.
It has the same effect as MSG_MORE, but existed in sendfile() only.
This patch adds a option MSG_SENDPAGE_NOTLAST for corking data, try to
cork more data before sending when using sendfile(), which acts like
TCP's behaviour. Also, this reimplements the default sendpage to inform
that it is supported to some extent.
[1] https://man7.org/linux/man-pages/man2/sendfile.2.html
Signed-off-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Based on the manual of TCP_CORK [1] and MSG_MORE [2], these two options
have the same effect. Applications can set these options and informs the
kernel to pend the data, and send them out only when the socket or
syscall does not specify this flag. In other words, there's no need to
send data out by a delayed work, which will queue a lot of work.
This removes corked delayed work with SMC_TX_CORK_DELAY (250ms), and the
applications control how/when to send them out. It improves the
performance for sendfile and throughput, and remove unnecessary race of
lock_sock(). This also unlocks the limitation of sndbuf, and try to fill
it up before sending.
[1] https://linux.die.net/man/7/tcp
[2] https://man7.org/linux/man-pages/man2/send.2.html
Signed-off-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
According to the man page of TCP_CORK [1], if set, don't send out
partial frames. All queued partial frames are sent when option is
cleared again.
When applications call setsockopt to disable TCP_CORK, this call is
protected by lock_sock(), and tries to mod_delayed_work() to 0, in order
to send pending data right now. However, the delayed work smc_tx_work is
also protected by lock_sock(). There introduces lock contention for
sending data.
To fix it, send pending data directly which acts like TCP, without
lock_sock() protected in the context of setsockopt (already lock_sock()ed),
and cancel unnecessary dealyed work, which is protected by lock.
[1] https://linux.die.net/man/7/tcp
Signed-off-by: Tony Lu <tonylu@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This introduce two tracepoints for smc tx and rx msg to help us
diagnosis issues of data path. These two tracepoitns don't cover the
path of CORK or MSG_MORE in tx, just the top half of data path.
Signed-off-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Wen Gu <guwen@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Commit 8f3d65c16679 ("net/smc: fix wait on already cleared link")
introduced link refcounting to avoid waits on already cleared links.
This patch extents and improves the refcounting to cover all
remaining possible cases for this kind of error situation.
Fixes: 15e1b99aadfb ("net/smc: no WR buffer wait for terminating link group")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There can be a race between the waiters for a tx work request buffer
and the link down processing that finally clears the link. Although
all waiters are woken up before the link is cleared there might be
waiters which did not yet get back control and are still waiting.
This results in an access to a cleared wait queue head.
Fix this by introducing atomic reference counting around the wait calls,
and wait with the link clear processing until all waiters have finished.
Move the work request layer related calls into smc_wr.c and set the
link state to INACTIVE before calling smcr_link_clear() in
smc_llc_srv_add_link().
Fixes: 15e1b99aadfb ("net/smc: no WR buffer wait for terminating link group")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When smc_sendmsg() is called before the SMC socket initialization has
completed, smc_tx_sendmsg() will access un-initialized fields of the
SMC socket which results in a null-pointer dereference.
Fix this by checking the socket state first in smc_tx_sendmsg().
Fixes: e0e4b8fa5338 ("net/smc: Add SMC statistics support")
Reported-by: syzbot+5dda108b672b54141857@syzkaller.appspotmail.com
Reviewed-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the gathered SMC statistics network namespace aware, for each
namespace collect an own set of statistic information.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add the ability to collect SMC statistics information. Per-cpu
variables are used to collect the statistic information for better
performance and for reducing concurrency pitfalls. The code that is
collecting statistic data is implemented in macros to increase code
reuse and readability.
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There are 6 types of workers which exist per smc connection. 3 of them
are used for listen and handshake processing, another 2 are used for
close and abort processing and 1 is the tx worker that moves calls to
sleeping functions into a worker.
To prevent flooding of the system work queue when many connections are
opened or closed at the same time (some pattern uperf implements), move
those workers to one of 3 smc-specific work queues. Two work queues are
module-global and used for handshake and close workers. The third work
queue is defined per link group and used by the tx workers that may
sleep waiting for resources of this link group.
And in smc_llc_enqueue() queue the llc_event_work work to the system
prio work queue because its critical that this work is started fast.
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add smc_switch_conns() to switch all connections from a link that is
going down. Find an other link to switch the connections to, and
switch each connection to the new link. smc_switch_cursor() updates the
cursors of a connection to the state of the last successfully sent CDC
message. When there is no link to switch to, terminate the link group.
Call smc_switch_conns() when a link is going down.
And with the possibility that links of connections can switch adapt CDC
and TX functions to detect and handle link switches.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use the introduced link down processing in all places where the link
group is terminated and take down the affected link only.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
As a preparation for the support of multiple links remove the usage of
a static link id (SMC_SINGLE_LINK) and allow dynamic link ids.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
smc_tx_rdma_write() is called under the send_lock and should not call
smc_lgr_terminate() directly. Call smc_lgr_terminate_sched() instead
which schedules a worker.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
A final cleanup due to SMCD device removal means immediate freeing
of all link groups belonging to this device in interrupt context.
This patch introduces a separate SMCD link group termination routine,
which terminates all link groups of an SMCD device.
This new routine smcd_terminate_all ()is reused if the smc module is
unloaded.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The resources for a terminated socket are being cleaned up.
This patch makes sure
* no more data is received for an actively terminated socket
* no more data is sent for an actively or passively terminated socket
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
|
|
Currently, we are only explicitly setting SOCK_NOSPACE on a write timeout
for non-blocking sockets. Epoll() edge-trigger mode relies on SOCK_NOSPACE
being set when -EAGAIN is returned to ensure that EPOLLOUT is raised.
Expand the setting of SOCK_NOSPACE to non-blocking sockets as well that can
use SO_SNDTIMEO to adjust their write timeout. This mirrors the behavior
that Eric Dumazet introduced for tcp sockets.
Signed-off-by: Jason Baron <jbaron@akamai.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Ursula Braun <ubraun@linux.ibm.com>
Cc: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Use local variable pflags from the beginning of function
smcr_tx_sndbuf_nonempty
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When no free transfer buffers are available then a work to call
smc_tx_work() is scheduled. Set the schedule delay to zero, because for
the out-of-buffers condition the work can start immediately and will
block in the called function smc_wr_tx_get_free_slot(), waiting for free
buffers.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Move the call to smc_close_wake_tx_prepared() (which wakes up a possibly
waiting close processing that might wait for 'all data sent') to
smc_tx_sndbuf_nonempty() (which is the main function to send data).
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When an updated rx_cursor_confirmed field was sent to the peer then
reset the cons_curs_upd_req flag. And remove the duplicate reset and
cursor update in smc_tx_consumer_update().
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In some scenarios a separate consumer cursor update is necessary.
The decision is made in smc_tx_consumer_cursor_update(). The
sender_free computation could be wrong:
The rx confirmed cursor is always smaller than or equal to the
rx producer cursor. The parameters in the smc_curs_diff() call
have to be exchanged, otherwise sender_free might even be negative.
And if more data arrives local_rx_ctrl.prod might be updated, enabling
a cursor difference between local_rx_ctrl.prod and rx confirmed cursor
larger than the RMB size. This case is not covered by smc_curs_diff().
Thus function smc_curs_diff_large() is introduced here.
If a recvmsg() is processed in parallel, local_tx_ctrl.cons might
change during smc_cdc_msg_send. Make sure rx_curs_confirmed is updated
with the actually sent local_tx_ctrl.cons value.
Fixes: e82f2e31f559 ("net/smc: optimize consumer cursor updates")
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The work requests for rdma writes are built in local variables within
function smc_tx_rdma_write(). This violates the rule that the work
request storage has to stay till the work request is confirmed by
a completion queue response.
This patch introduces preallocated memory for these work requests.
The storage is allocated, once a link (and thus a queue pair) is
established.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
smc_cdc_get_free_slot() might wait for free transfer buffers when using
SMC-R. This wait should not be done under the send_lock, which is a
spin_lock. This fixes a cpu loop in parallel threads waiting for the
send_lock.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When there is no more send buffer space and at least 1 byte was already
sent then return to user space. The wait is only done when no data was
sent by the sendmsg() call.
This fixes smc_tx_sendmsg() which tried to always send all user data and
started to wait for free send buffer space when needed. During this wait
the user space program was blocked in the sendmsg() call and hence not
able to receive incoming data. When both sides were in such a situation
then the connection stalled forever.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
rdma.git merge resolution for the 4.19 merge window
Conflicts:
drivers/infiniband/core/rdma_core.c
- Use the rdma code and revise with the new spelling for
atomic_fetch_add_unless
drivers/nvme/host/rdma.c
- Replace max_sge with max_send_sge in new blk code
drivers/nvme/target/rdma.c
- Use the blk code and revise to use NULL for ib_post_recv when
appropriate
- Replace max_sge with max_recv_sge in new blk code
net/rds/ib_send.c
- Use the net code and revise to use NULL for ib_post_recv when
appropriate
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
|
|
Resolve merge conflicts from the -rc cycle against the rdma.git tree:
Conflicts:
drivers/infiniband/core/uverbs_cmd.c
- New ifs added to ib_uverbs_ex_create_flow in -rc and for-next
- Merge removal of file->ucontext in for-next with new code in -rc
drivers/infiniband/core/uverbs_main.c
- for-next removed code from ib_uverbs_write() that was modified
in for-rc
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
|
|
Instead of declaring and passing a dummy 'bad_wr' pointer, pass NULL
as third argument to ib_post_(send|recv|srq_recv)().
Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Acked-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
|
|
The functions to read and write cursors are exclusively used to copy
cursors. Therefore switch to a respective function instead.
Signed-off-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The SMC protocol requires to send a separate consumer cursor update,
if it cannot be piggybacked to updates of the producer cursor.
Currently the decision to send a separate consumer cursor update
just considers the amount of data already received by the socket
program. It does not consider the amount of data already arrived, but
not yet consumed by the receiver. Basing the decision on the
difference between already confirmed and already arrived data
(instead of difference between already confirmed and already consumed
data), may lead to a somewhat earlier consumer cursor update send in
fast unidirectional traffic scenarios, and thus to better throughput.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Suggested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The data transfer and CDC message headers differ in SMC-R and SMC-D.
This patch adds support for the SMC-D data transfer to the existing SMC
code. It consists of the following:
* SMC-D CDC support
* SMC-D tx support
* SMC-D rx support
The CDC header is stored at the beginning of the receive buffer. Thus, a
rx_offset variable is added for the CDC header offset within the buffer
(0 for SMC-R).
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Suggested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The SMC protocol requires to send a separate consumer cursor update,
if it cannot be piggybacked to updates of the producer cursor.
Currently the decision to send a separate consumer cursor update
just considers the amount of data already received by the socket
program. It does not consider the amount of data already arrived, but
not yet consumed by the receiver. Basing the decision on the
difference between already confirmed and already arrived data
(instead of difference between already confirmed and already consumed
data), may lead to a somewhat earlier consumer cursor update send in
fast unidirectional traffic scenarios, and thus to better throughput.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Suggested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add support for out of band data send and receive.
Signed-off-by: Stefan Raspl <raspl@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently, the write offset within the RMB is calculated on each write
operation although it is fixed for each connection. With this patch, the
offset is calculated once and stored in a connection specific variable.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The connection index is actually a RMBE index. So, this patch changes
the name accordingly.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
In addition to the buffer references, SMC currently stores the sizes of
the receive and send buffers in each connection as separate variables.
This patch introduces a buffer length variable in the common buffer
descriptor and uses this length instead.
Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
syzkaller found that following program crashes the host :
{
int fd = socket(AF_SMC, SOCK_STREAM, 0);
int val = 1;
listen(fd, 0);
shutdown(fd, SHUT_RDWR);
setsockopt(fd, 6, TCP_NODELAY, &val, 4);
}
Simply initialize conn.tx_work & conn.send_lock at socket creation,
rather than deeper in the stack.
ODEBUG: assert_init not available (active state 0) object type: timer_list hint: (null)
WARNING: CPU: 1 PID: 13988 at lib/debugobjects.c:329 debug_print_object+0x16a/0x210 lib/debugobjects.c:326
Kernel panic - not syncing: panic_on_warn set ...
CPU: 1 PID: 13988 Comm: syz-executor0 Not tainted 4.17.0-rc4+ #46
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0x1b9/0x294 lib/dump_stack.c:113
panic+0x22f/0x4de kernel/panic.c:184
__warn.cold.8+0x163/0x1b3 kernel/panic.c:536
report_bug+0x252/0x2d0 lib/bug.c:186
fixup_bug arch/x86/kernel/traps.c:178 [inline]
do_error_trap+0x1de/0x490 arch/x86/kernel/traps.c:296
do_invalid_op+0x1b/0x20 arch/x86/kernel/traps.c:315
invalid_op+0x14/0x20 arch/x86/entry/entry_64.S:992
RIP: 0010:debug_print_object+0x16a/0x210 lib/debugobjects.c:326
RSP: 0018:ffff880197a37880 EFLAGS: 00010086
RAX: 0000000000000061 RBX: 0000000000000005 RCX: ffffc90001ed0000
RDX: 0000000000004aaf RSI: ffffffff8160f6f1 RDI: 0000000000000001
RBP: ffff880197a378c0 R08: ffff8801aa7a0080 R09: ffffed003b5e3eb2
R10: ffffed003b5e3eb2 R11: ffff8801daf1f597 R12: 0000000000000001
R13: ffffffff88d96980 R14: ffffffff87fa19a0 R15: ffffffff81666ec0
debug_object_assert_init+0x309/0x500 lib/debugobjects.c:692
debug_timer_assert_init kernel/time/timer.c:724 [inline]
debug_assert_init kernel/time/timer.c:776 [inline]
del_timer+0x74/0x140 kernel/time/timer.c:1198
try_to_grab_pending+0x439/0x9a0 kernel/workqueue.c:1223
mod_delayed_work_on+0x91/0x250 kernel/workqueue.c:1592
mod_delayed_work include/linux/workqueue.h:541 [inline]
smc_setsockopt+0x387/0x6d0 net/smc/af_smc.c:1367
__sys_setsockopt+0x1bd/0x390 net/socket.c:1903
__do_sys_setsockopt net/socket.c:1914 [inline]
__se_sys_setsockopt net/socket.c:1911 [inline]
__x64_sys_setsockopt+0xbe/0x150 net/socket.c:1911
do_syscall_64+0x1b1/0x800 arch/x86/entry/common.c:287
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Fixes: 01d2f7e2cdd3 ("net/smc: sockopts TCP_NODELAY and TCP_CORK")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ursula Braun <ubraun@linux.ibm.com>
Cc: linux-s390@vger.kernel.org
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|