| Message ID | 1648447836-111521-1-git-send-email-guwen@linux.alibaba.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 906b3d64913c19a50d5c553f21b54d2f4ce3ded7 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] net/smc: Send out the remaining data in sndbuf before close | expand |
On Mon, Mar 28, 2022 at 02:10:36PM +0800, Wen Gu wrote: >The current autocork algorithms will delay the data transmission >in BH context to smc_release_cb() when sock_lock is hold by user. > >So there is a possibility that when connection is being actively >closed (sock_lock is hold by user now), some corked data still >remains in sndbuf, waiting to be sent by smc_release_cb(). This >will cause: > >- smc_close_stream_wait(), which is called under the sock_lock, > has a high probability of timeout because data transmission is > delayed until sock_lock is released. > >- Unexpected data sends may happen after connction closed and use > the rtoken which has been deleted by remote peer through > LLC_DELETE_RKEY messages. > >So this patch will try to send out the remaining corked data in >sndbuf before active close process, to ensure data integrity and >avoid unexpected data transmission after close. I think this issue should also happen if TCP_CORK is set and autocorking is not enabled ? Autocorking and delaying the TX from BH to smc_release_cb() greatly increased the probability of this problem. > >Reported-by: Guangguan Wang <guangguan.wang@linux.alibaba.com> >Fixes: 6b88af839d20 ("net/smc: don't send in the BH context if sock_owned_by_user") >Signed-off-by: Wen Gu <guwen@linux.alibaba.com> >--- > net/smc/smc_close.c | 3 +++ > 1 file changed, 3 insertions(+) > >diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c >index 292e4d9..676cb23 100644 >--- a/net/smc/smc_close.c >+++ b/net/smc/smc_close.c >@@ -57,6 +57,9 @@ static void smc_close_stream_wait(struct smc_sock *smc, long timeout) > if (!smc_tx_prepared_sends(&smc->conn)) > return; > >+ /* Send out corked data remaining in sndbuf */ >+ smc_tx_pending(&smc->conn); >+ > smc->wait_close_tx_prepared = 1; > add_wait_queue(sk_sleep(sk), &wait); > while (!signal_pending(current) && timeout) { >-- >1.8.3.1
Thanks for your comments. On 2022/3/28 5:04 pm, dust.li wrote: > > I think this issue should also happen if TCP_CORK is set and > autocorking is not enabled ? Yes, setting TCP_CORK also works. > > Autocorking and delaying the TX from BH to smc_release_cb() greatly > increased the probability of this problem. > Thanks, Wen Gu
On 28/03/2022 08:10, Wen Gu wrote: > The current autocork algorithms will delay the data transmission > in BH context to smc_release_cb() when sock_lock is hold by user. > > So there is a possibility that when connection is being actively > closed (sock_lock is hold by user now), some corked data still > remains in sndbuf, waiting to be sent by smc_release_cb(). This > will cause: > > - smc_close_stream_wait(), which is called under the sock_lock, > has a high probability of timeout because data transmission is > delayed until sock_lock is released. > > - Unexpected data sends may happen after connction closed and use > the rtoken which has been deleted by remote peer through > LLC_DELETE_RKEY messages. > > So this patch will try to send out the remaining corked data in > sndbuf before active close process, to ensure data integrity and > avoid unexpected data transmission after close. > > Reported-by: Guangguan Wang <guangguan.wang@linux.alibaba.com> > Fixes: 6b88af839d20 ("net/smc: don't send in the BH context if sock_owned_by_user") > Signed-off-by: Wen Gu <guwen@linux.alibaba.com> > --- Thank you, Acked-by: Karsten Graul <kgraul@linux.ibm.com>
Hello: This patch was applied to netdev/net.git (master) by Jakub Kicinski <kuba@kernel.org>: On Mon, 28 Mar 2022 14:10:36 +0800 you wrote: > The current autocork algorithms will delay the data transmission > in BH context to smc_release_cb() when sock_lock is hold by user. > > So there is a possibility that when connection is being actively > closed (sock_lock is hold by user now), some corked data still > remains in sndbuf, waiting to be sent by smc_release_cb(). This > will cause: > > [...] Here is the summary with links: - [net] net/smc: Send out the remaining data in sndbuf before close https://git.kernel.org/netdev/net/c/906b3d64913c You are awesome, thank you!
diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c index 292e4d9..676cb23 100644 --- a/net/smc/smc_close.c +++ b/net/smc/smc_close.c @@ -57,6 +57,9 @@ static void smc_close_stream_wait(struct smc_sock *smc, long timeout) if (!smc_tx_prepared_sends(&smc->conn)) return; + /* Send out corked data remaining in sndbuf */ + smc_tx_pending(&smc->conn); + smc->wait_close_tx_prepared = 1; add_wait_queue(sk_sleep(sk), &wait); while (!signal_pending(current) && timeout) {
The current autocork algorithms will delay the data transmission in BH context to smc_release_cb() when sock_lock is hold by user. So there is a possibility that when connection is being actively closed (sock_lock is hold by user now), some corked data still remains in sndbuf, waiting to be sent by smc_release_cb(). This will cause: - smc_close_stream_wait(), which is called under the sock_lock, has a high probability of timeout because data transmission is delayed until sock_lock is released. - Unexpected data sends may happen after connction closed and use the rtoken which has been deleted by remote peer through LLC_DELETE_RKEY messages. So this patch will try to send out the remaining corked data in sndbuf before active close process, to ensure data integrity and avoid unexpected data transmission after close. Reported-by: Guangguan Wang <guangguan.wang@linux.alibaba.com> Fixes: 6b88af839d20 ("net/smc: don't send in the BH context if sock_owned_by_user") Signed-off-by: Wen Gu <guwen@linux.alibaba.com> --- net/smc/smc_close.c | 3 +++ 1 file changed, 3 insertions(+)