On 10.09.2020 14:54, Al Viro wrote:
On Thu, Sep 10, 2020 at 12:48:34PM +0300, Sergey Nikitin wrote:Hi! epoll does not report an event to all the threads running epoll_wait() on the same epoll descriptor. The behavior appeared in recent kernel versions starting with 5.6 probably. How to reproduce: - create a pair of sockets - create epoll instance - register the socket on the epoll instance, listen for EPOLLIN events - start 2 threads running epoll_wait() - send some data to the socket - see that epoll_wait() within one of the threads reported an event, unlike another.Could you reproduce it on mainline kernel and try to bisect it?
I rechecked the f4d51dffc6c0 Linux 5.9-rc4. The issue is still reproducible. Bisect result: 339ddb53d373baee6e7946aec17c739c4924d6d9 is the first bad commit commit 339ddb53d373baee6e7946aec17c739c4924d6d9 Author: Heiher <r@xxxxxx> Date: Wed Dec 4 16:52:15 2019 -0800 fs/epoll: remove unnecessary wakeups of nested epoll Take the case where we have: t0 | (ew) e0 | (et) e1 | (lt) s0 t0: thread 0 e0: epoll fd 0 e1: epoll fd 1 s0: socket fd 0 ew: epoll_wait et: edge-trigger lt: level-triggerWe remove unnecessary wakeups to prevent the nested epoll that working in edge-
triggered mode to waking up continuously.
Test code:
#include <unistd.h>
#include <sys/epoll.h>
#include <sys/socket.h>
int main(int argc, char *argv[])
{
int sfd[2];
int efd[2];
struct epoll_event e;
if (socketpair(AF_UNIX, SOCK_STREAM, 0, sfd) < 0)
goto out;
efd[0] = epoll_create(1);
if (efd[0] < 0)
goto out;
efd[1] = epoll_create(1);
if (efd[1] < 0)
goto out;
e.events = EPOLLIN;
if (epoll_ctl(efd[1], EPOLL_CTL_ADD, sfd[0], &e) < 0)
goto out;
e.events = EPOLLIN | EPOLLET;
if (epoll_ctl(efd[0], EPOLL_CTL_ADD, efd[1], &e) < 0)
goto out;
if (write(sfd[1], "w", 1) != 1)
goto out;
if (epoll_wait(efd[0], &e, 1, 0) != 1)
goto out;
if (epoll_wait(efd[0], &e, 1, 0) != 0)
goto out;
close(efd[0]);
close(efd[1]);
close(sfd[0]);
close(sfd[1]);
return 0;
out:
return -1;
}
More tests:
https://github.com/heiher/epoll-wakeup
Link: http://lkml.kernel.org/r/20191009060516.3577-1-r@xxxxxx
Signed-off-by: hev <r@xxxxxx>
Reviewed-by: Roman Penyaev <rpenyaev@xxxxxxx>
Cc: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
Cc: Davide Libenzi <davidel@xxxxxxxxxxxxxxx>
Cc: Davidlohr Bueso <dave@xxxxxxxxxxxx>
Cc: Dominik Brodowski <linux@xxxxxxxxxxxxxxxxxxxx>
Cc: Eric Wong <e@xxxxxxxxx>
Cc: Jason Baron <jbaron@xxxxxxxxxx>
Cc: Sridhar Samudrala <sridhar.samudrala@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
fs/eventpoll.c | 16 ----------------
1 file changed, 16 deletions(-)
Attaching a C reproducer which I was using to bisect.
--
Best regards,
Sergey Nikitin
Attachment:
reproducer.tar
Description: Unix tar archive
