BUG/MINOR: select: fix excess number of dead/skip reported

In 1.8, commit ab62f5195 ("MINOR: polling: Use fd_update_events to update
events seen for a fd") updated the pollers to rely on fd_update_events(),
but the modification delayed the test of presence of the FD in the report,
resulting in owner/thread_mask and possibly event updates being performed
for each FD appearing in a block of 32 FDs around an active one. This
caused the request rate to be ~3 times lower with select() than poll()
under 6 threads.

This can be backported as far as 1.8.
This commit is contained in:
Willy Tarreau 2021-07-30 13:55:36 +02:00
parent c37ccd70b4
commit fcc5281513

View File

@ -197,6 +197,15 @@ static void _do_poll(struct poller *p, int exp, int wake)
for (count = BITS_PER_INT, fd = fds * BITS_PER_INT; count && fd < maxfd; count--, fd++) {
unsigned int n = 0;
if (FD_ISSET(fd, tmp_evts[DIR_RD]))
n |= FD_EV_READY_R;
if (FD_ISSET(fd, tmp_evts[DIR_WR]))
n |= FD_EV_READY_W;
if (!n)
continue;
#ifdef DEBUG_FD
_HA_ATOMIC_INC(&fdtab[fd].event_count);
#endif
@ -210,12 +219,6 @@ static void _do_poll(struct poller *p, int exp, int wake)
continue;
}
if (FD_ISSET(fd, tmp_evts[DIR_RD]))
n |= FD_EV_READY_R;
if (FD_ISSET(fd, tmp_evts[DIR_WR]))
n |= FD_EV_READY_W;
fd_update_events(fd, n);
}
}