On 01/23, Jason Baron wrote: > > >It was added by 3c7151275c0c9a "tcp: add memory barriers to write space > >paths" > >and the patch looks correct in that we need the barriers in tcp_check_space() > >and tcp_poll() in theory, so it seems tcp_check_space() needs smp_mb() ? > > > > Yes, I think it should be upgraded to an smp_mb() there. If you agree with > this analysis, I will send a patch to upgrade it.
I think this makes sense, if nothing else to remove the obviously wrong smp_mb__after_atomic() ;) > Note, I did not actually > run into this race in practice. Yes, I am not sure it is actually necessary in practice, and even if tcp_check_space() races with tcp_poll() and misses SOCK_NOSPACE it should be probably called again later, but I know nothing about networking code. We noticed this by accident, we have a bug report which really looks as a missed wakeup in epoll, but this kernel is old and in particular it lacks the commit 128dd1759 "epoll: prevent missed events on EPOLL_CTL_MOD" which looks more promising. But we are not sure it can fully explain the problem we can't reproduce, so we are looking for anything else which may contribute to the problem. Oleg.
