On Tue, Mar 7, 2023 at 4:17 PM Keller, Jacob E <jacob.e.keller@xxxxxxxxx> wrote: > > > > > -----Original Message----- > > From: Paolo Abeni <pabeni@xxxxxxxxxx> > > Sent: Tuesday, March 7, 2023 10:47 AM > > To: netdev@xxxxxxxxxxxxxxx > > Cc: Soheil Hassas Yeganeh <soheil@xxxxxxxxxx>; Al Viro > > <viro@xxxxxxxxxxxxxxxxxx>; Carlos Maiolino <cmaiolino@xxxxxxxxxx>; Eric > > Biggers <ebiggers@xxxxxxxxxx>; Keller, Jacob E <jacob.e.keller@xxxxxxxxx>; > > Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>; Jens Axboe <axboe@xxxxxxxxx>; > > Christian Brauner <brauner@xxxxxxxxxx>; linux-fsdevel@xxxxxxxxxxxxxxx > > Subject: [PATCH v4 RESEND] epoll: use refcount to reduce ep_mutex contention > > > > We are observing huge contention on the epmutex during an http > > connection/rate test: > > > > 83.17% 0.25% nginx [kernel.kallsyms] [k] > > entry_SYSCALL_64_after_hwframe > > [...] > > |--66.96%--__fput > > |--60.04%--eventpoll_release_file > > |--58.41%--__mutex_lock.isra.6 > > |--56.56%--osq_lock > > > > The application is multi-threaded, creates a new epoll entry for > > each incoming connection, and does not delete it before the > > connection shutdown - that is, before the connection's fd close(). > > > > Many different threads compete frequently for the epmutex lock, > > affecting the overall performance. > > > > To reduce the contention this patch introduces explicit reference counting > > for the eventpoll struct. Each registered event acquires a reference, > > and references are released at ep_remove() time. > > > > Additionally, this introduces a new 'dying' flag to prevent races between > > the EP file close() and the monitored file close(). > > ep_eventpoll_release() marks, under f_lock spinlock, each epitem as before > > removing it, while EP file close() does not touch dying epitems. > > > > The eventpoll struct is released by whoever - among EP file close() and > > and the monitored file close() drops its last reference. > > > > With all the above in place, we can drop the epmutex usage at disposal time. > > > > Overall this produces a significant performance improvement in the > > mentioned connection/rate scenario: the mutex operations disappear from > > the topmost offenders in the perf report, and the measured connections/rate > > grows by ~60%. > > > > To make the change more readable this additionally renames ep_free() to > > ep_clear_and_put(), and moves the actual memory cleanup in a separate > > ep_free() helper. > > > > Tested-by: Xiumei Mu <xmu@xxxxxxxxxxx> > > Reviewed-by: Jacob Keller <jacob.e.keller@xxxxxxxxx> > > Acked-by: Soheil Hassas Yeganeh <soheil@xxxxxxxxxx> > > Signed-off-by: Paolo Abeni <pabeni@xxxxxxxxxx> > > --- > > This is a repost of v4, with no changes. Kindly asking if FS maintainers > > could have a look. > > This (still) looks good to me. > > Thanks, > Jake Thank you! Still looks great to me as well.
