* [PATCH 0/2] httpd/async race fixes @ 2016-07-09 4:51 Eric Wong 2016-07-09 4:51 ` [PATCH 1/2] httpd/async: do not attempt future writes on closed sockets Eric Wong 2016-07-09 4:51 ` [PATCH 2/2] httpd/async: reinstate D::S timer usage for cleanup Eric Wong 0 siblings, 2 replies; 3+ messages in thread From: Eric Wong @ 2016-07-09 4:51 UTC (permalink / raw) To: meta Oops, these are some tricky fixes which took a while to notice and find during stress testing. Eric Wong (2): httpd/async: do not attempt future writes on closed sockets httpd/async: reinstate D::S timer usage for cleanup lib/PublicInbox/EvCleanup.pm | 9 +++++++++ lib/PublicInbox/HTTPD/Async.pm | 3 ++- 2 files changed, 11 insertions(+), 1 deletion(-) ^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH 1/2] httpd/async: do not attempt future writes on closed sockets 2016-07-09 4:51 [PATCH 0/2] httpd/async race fixes Eric Wong @ 2016-07-09 4:51 ` Eric Wong 2016-07-09 4:51 ` [PATCH 2/2] httpd/async: reinstate D::S timer usage for cleanup Eric Wong 1 sibling, 0 replies; 3+ messages in thread From: Eric Wong @ 2016-07-09 4:51 UTC (permalink / raw) To: meta Danga::Socket::close does not clear the write_buf_size field, so it's conceivable we could attempt to queue up data and callbacks we can never flush out. --- lib/PublicInbox/HTTPD/Async.pm | 1 + 1 file changed, 1 insertion(+) diff --git a/lib/PublicInbox/HTTPD/Async.pm b/lib/PublicInbox/HTTPD/Async.pm index a936d9b..880cf02 100644 --- a/lib/PublicInbox/HTTPD/Async.pm +++ b/lib/PublicInbox/HTTPD/Async.pm @@ -41,6 +41,7 @@ sub async_pass { my $r = sysread($self->{sock}, $$bref, 8192); if ($r) { $fh->write($$bref); + return if $io->{closed}; if ($io->{write_buf_size}) { $self->watch_read(0); $io->write($restart_read); # D::S::write -- EW ^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH 2/2] httpd/async: reinstate D::S timer usage for cleanup 2016-07-09 4:51 [PATCH 0/2] httpd/async race fixes Eric Wong 2016-07-09 4:51 ` [PATCH 1/2] httpd/async: do not attempt future writes on closed sockets Eric Wong @ 2016-07-09 4:51 ` Eric Wong 1 sibling, 0 replies; 3+ messages in thread From: Eric Wong @ 2016-07-09 4:51 UTC (permalink / raw) To: meta EvCleanup::asap events are not guaranteed to run after Danga::Socket closes sockets at the event loop. Thus we must use slower Danga::Socket timers which are guaranteed to run at the end of the event loop. --- lib/PublicInbox/EvCleanup.pm | 9 +++++++++ lib/PublicInbox/HTTPD/Async.pm | 2 +- 2 files changed, 10 insertions(+), 1 deletion(-) diff --git a/lib/PublicInbox/EvCleanup.pm b/lib/PublicInbox/EvCleanup.pm index 61837b8..2b77c61 100644 --- a/lib/PublicInbox/EvCleanup.pm +++ b/lib/PublicInbox/EvCleanup.pm @@ -9,6 +9,7 @@ use base qw(Danga::Socket); use fields qw(rd); my $singleton; my $asapq = [ [], undef ]; +my $nextq = [ [], undef ]; my $laterq = [ [], undef ]; sub once_init () { @@ -30,6 +31,7 @@ sub _run_all ($) { } sub _run_asap () { _run_all($asapq) } +sub _run_next () { _run_all($nextq) } sub _run_later () { _run_all($laterq) } # Called by Danga::Socket @@ -51,6 +53,12 @@ sub asap ($) { $asapq->[1] ||= _asap_timer(); } +sub next_tick ($) { + my ($cb) = @_; + push @{$nextq->[0]}, $cb; + $nextq->[1] ||= Danga::Socket->AddTimer(0, *_run_next); +} + sub later ($) { my ($cb) = @_; push @{$laterq->[0]}, $cb; @@ -59,6 +67,7 @@ sub later ($) { END { _run_asap(); + _run_next(); _run_later(); } diff --git a/lib/PublicInbox/HTTPD/Async.pm b/lib/PublicInbox/HTTPD/Async.pm index 880cf02..68514f5 100644 --- a/lib/PublicInbox/HTTPD/Async.pm +++ b/lib/PublicInbox/HTTPD/Async.pm @@ -72,7 +72,7 @@ sub close { $self->SUPER::close(@_); # we defer this to the next timer loop since close is deferred - PublicInbox::EvCleanup::asap($cleanup) if $cleanup; + PublicInbox::EvCleanup::next_tick($cleanup) if $cleanup; } 1; -- EW ^ permalink raw reply related [flat|nested] 3+ messages in thread
end of thread, other threads:[~2016-07-09 4:51 UTC | newest] Thread overview: 3+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2016-07-09 4:51 [PATCH 0/2] httpd/async race fixes Eric Wong 2016-07-09 4:51 ` [PATCH 1/2] httpd/async: do not attempt future writes on closed sockets Eric Wong 2016-07-09 4:51 ` [PATCH 2/2] httpd/async: reinstate D::S timer usage for cleanup Eric Wong
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).