| Message ID | 3de97b2d-1c15-5dda-4fe2-78311a91d861@I-love.SAKURA.ne.jp (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | a91b750fd6629354460282bbf5146c01b05c4859 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | net: rds: don't hold sock lock when cancelling work from rds_tcp_reset_callbacks() | expand |
Hello: This patch was applied to netdev/net.git (master) by David S. Miller <davem@davemloft.net>: On Thu, 29 Sep 2022 00:25:37 +0900 you wrote: > syzbot is reporting lockdep warning at rds_tcp_reset_callbacks() [1], for > commit ac3615e7f3cffe2a ("RDS: TCP: Reduce code duplication in > rds_tcp_reset_callbacks()") added cancel_delayed_work_sync() into a section > protected by lock_sock() without realizing that rds_send_xmit() might call > lock_sock(). > > We don't need to protect cancel_delayed_work_sync() using lock_sock(), for > even if rds_{send,recv}_worker() re-queued this work while __flush_work() > from cancel_delayed_work_sync() was waiting for this work to complete, > retried rds_{send,recv}_worker() is no-op due to the absence of RDS_CONN_UP > bit. > > [...] Here is the summary with links: - net: rds: don't hold sock lock when cancelling work from rds_tcp_reset_callbacks() https://git.kernel.org/netdev/net/c/a91b750fd662 You are awesome, thank you!
diff --git a/net/rds/tcp.c b/net/rds/tcp.c index 73ee2771093d..d0ff413f697c 100644 --- a/net/rds/tcp.c +++ b/net/rds/tcp.c @@ -166,10 +166,10 @@ void rds_tcp_reset_callbacks(struct socket *sock, */ atomic_set(&cp->cp_state, RDS_CONN_RESETTING); wait_event(cp->cp_waitq, !test_bit(RDS_IN_XMIT, &cp->cp_flags)); - lock_sock(osock->sk); /* reset receive side state for rds_tcp_data_recv() for osock */ cancel_delayed_work_sync(&cp->cp_send_w); cancel_delayed_work_sync(&cp->cp_recv_w); + lock_sock(osock->sk); if (tc->t_tinc) { rds_inc_put(&tc->t_tinc->ti_inc); tc->t_tinc = NULL;
syzbot is reporting lockdep warning at rds_tcp_reset_callbacks() [1], for commit ac3615e7f3cffe2a ("RDS: TCP: Reduce code duplication in rds_tcp_reset_callbacks()") added cancel_delayed_work_sync() into a section protected by lock_sock() without realizing that rds_send_xmit() might call lock_sock(). We don't need to protect cancel_delayed_work_sync() using lock_sock(), for even if rds_{send,recv}_worker() re-queued this work while __flush_work() from cancel_delayed_work_sync() was waiting for this work to complete, retried rds_{send,recv}_worker() is no-op due to the absence of RDS_CONN_UP bit. Link: https://syzkaller.appspot.com/bug?extid=78c55c7bc6f66e53dce2 [1] Reported-by: syzbot <syzbot+78c55c7bc6f66e53dce2@syzkaller.appspotmail.com> Co-developed-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Tested-by: syzbot <syzbot+78c55c7bc6f66e53dce2@syzkaller.appspotmail.com> Fixes: ac3615e7f3cffe2a ("RDS: TCP: Reduce code duplication in rds_tcp_reset_callbacks()") --- Hillf, why don't you propose as a formal patch after syzbot tested your patch? Explaining as a formal patch helps us with understanding/reviewing what you thought and how you came to your patch. I feel sorry for stealing result of your trial and error... net/rds/tcp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)