unofficial mirror of meta@public-inbox.org
 help / color / mirror / Atom feed
* [PATCH 0/3] sig handling fixes for non-signalfd/kqueue
@ 2023-09-08 10:51 Eric Wong
  2023-09-08 10:51 ` [PATCH 1/3] ds: fix signals unblock " Eric Wong
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Eric Wong @ 2023-09-08 10:51 UTC (permalink / raw)
  To: meta

The first two fix t/imapd.t getting stuck when waiting for
-watch to reload after setting imap.pollInterval.  This only
affects non-Linux users without IO::KQueue installed.

3/3 doesn't fix anything exposed in tests, but is defense
against PID typos by someone trying to trigger a reload/rescan
on the toplevel -watch process.

Eric Wong (3):
  ds: unblock signals for non-signalfd/kqueue correctly
  watch: set %SIG for non-signalfd/kqueue
  watch: reset HUP + USR1 signal handlers in children

 lib/PublicInbox/DS.pm     | 16 ++++++++++------
 lib/PublicInbox/Watch.pm  |  7 ++++---
 script/public-inbox-watch |  3 ++-
 3 files changed, 16 insertions(+), 10 deletions(-)

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH 1/3] ds: fix signals unblock for non-signalfd/kqueue
  2023-09-08 10:51 [PATCH 0/3] sig handling fixes for non-signalfd/kqueue Eric Wong
@ 2023-09-08 10:51 ` Eric Wong
  2023-09-08 10:51 ` [PATCH 2/3] watch: set %SIG " Eric Wong
  2023-09-08 10:51 ` [PATCH 3/3] watch: reset HUP + USR1 signal handlers in children Eric Wong
  2 siblings, 0 replies; 4+ messages in thread
From: Eric Wong @ 2023-09-08 10:51 UTC (permalink / raw)
  To: meta

Using the sigset result of allowset() isn't appropriate for
SIG_UNBLOCK.  We must generate a new signal set off of the $sig
dispatch map for use with SIG_UNBLOCK to actually unblock the
signals.

This is the first part in getting t/imapd.t to pass the
reload-after-setting--imap.pollInterval-test when neither
signalfd nor kqueue are usable.
---
 lib/PublicInbox/DS.pm | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/lib/PublicInbox/DS.pm b/lib/PublicInbox/DS.pm
index 5168a6ee..ff10c9c0 100644
--- a/lib/PublicInbox/DS.pm
+++ b/lib/PublicInbox/DS.pm
@@ -259,18 +259,21 @@ sub PostEventLoop () {
 			: 1
 }
 
-sub allowset ($) {
-	my ($sig) = @_; # { signame => whatever }
+sub sigset_prep ($$$) {
+	my ($sig, $init, $each) = @_; # $sig: { signame => whatever }
 	my $ret = POSIX::SigSet->new;
-	$ret->fillset or die "fillset: $!";
+	$ret->$init or die "$init: $!";
 	for my $s (keys %$sig) {
 		my $num = $SIGNUM{$s} // POSIX->can("SIG$s")->();
-		$ret->delset($num) or die "delset ($s => $num): $!";
+		$ret->$each($num) or die "$each ($s => $num): $!";
 	}
-	for (@UNBLOCKABLE) { $ret->delset($_) or die "delset($_): $!" }
+	for (@UNBLOCKABLE) { $ret->$each($_) or die "$each ($_): $!" }
 	$ret;
 }
 
+sub allowset ($) { sigset_prep $_[0], 'fillset', 'delset' }
+sub unblockset ($) { sigset_prep $_[0], 'emptyset', 'addset' }
+
 # Start processing IO events. In most daemon programs this never exits. See
 # C<post_loop_do> for how to exit the loop.
 sub event_loop (;$$) {
@@ -293,7 +296,8 @@ sub event_loop (;$$) {
 		# wake up every second to accept signals if we don't
 		# have signalfd or IO::KQueue:
 		sig_setmask($oldset) if $oldset;
-		sigprocmask(SIG_UNBLOCK, allowset($sig)) or die "SIG_UNBLOCK: $!";
+		sigprocmask(SIG_UNBLOCK, unblockset($sig)) or
+			die "SIG_UNBLOCK: $!";
 		PublicInbox::DS->SetLoopTimeout(1000);
 	}
 	$_[0] = $sigfd = $sig = undef; # $_[0] == sig

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [PATCH 2/3] watch: set %SIG for non-signalfd/kqueue
  2023-09-08 10:51 [PATCH 0/3] sig handling fixes for non-signalfd/kqueue Eric Wong
  2023-09-08 10:51 ` [PATCH 1/3] ds: fix signals unblock " Eric Wong
@ 2023-09-08 10:51 ` Eric Wong
  2023-09-08 10:51 ` [PATCH 3/3] watch: reset HUP + USR1 signal handlers in children Eric Wong
  2 siblings, 0 replies; 4+ messages in thread
From: Eric Wong @ 2023-09-08 10:51 UTC (permalink / raw)
  To: meta

We need to ensure there isn't a window where we lose $SIG{CHLD}
handling.  This is the second part in getting t/imapd.t to pass
the reload-after-setting-imap.pollInterval test

That said, I'm not entirely happy with the way -watch jumps
in and out of the event loop.  It's historical baggage from
the pre-event_loop days.
---
 script/public-inbox-watch | 1 +
 1 file changed, 1 insertion(+)

diff --git a/script/public-inbox-watch b/script/public-inbox-watch
index 75a9a36b..870cd31b 100755
--- a/script/public-inbox-watch
+++ b/script/public-inbox-watch
@@ -52,6 +52,7 @@ if ($watch) {
 		CHLD => \&PublicInbox::DS::enqueue_reap,
 	};
 	$sig->{QUIT} = $sig->{TERM} = $sig->{INT} = $quit;
+	local @SIG{keys %$sig} = values(%$sig); # for non-signalfd/kqueue
 
 	# --no-scan is only intended for testing atm, undocumented.
 	PublicInbox::DS::requeue($scan) if $do_scan;

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [PATCH 3/3] watch: reset HUP + USR1 signal handlers in children
  2023-09-08 10:51 [PATCH 0/3] sig handling fixes for non-signalfd/kqueue Eric Wong
  2023-09-08 10:51 ` [PATCH 1/3] ds: fix signals unblock " Eric Wong
  2023-09-08 10:51 ` [PATCH 2/3] watch: set %SIG " Eric Wong
@ 2023-09-08 10:51 ` Eric Wong
  2 siblings, 0 replies; 4+ messages in thread
From: Eric Wong @ 2023-09-08 10:51 UTC (permalink / raw)
  To: meta

Child processes handling IMAP/NNTP aren't going to want
to handle config reloads nor forced rescans, those are
exclusively for the parent.  We'll leave a note that
QUIT/TERM/INT can safely use the same callback for both
parent and children, as I nearly made the mistake of
resetting those to their default values in the child.
---
 lib/PublicInbox/Watch.pm  | 7 ++++---
 script/public-inbox-watch | 2 +-
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/lib/PublicInbox/Watch.pm b/lib/PublicInbox/Watch.pm
index a2dc125f..cf0720e3 100644
--- a/lib/PublicInbox/Watch.pm
+++ b/lib/PublicInbox/Watch.pm
@@ -250,7 +250,7 @@ sub quit_done ($) {
 	$live == 0;
 }
 
-sub quit {
+sub quit { # may be called in IMAP/NNTP children
 	my ($self) = @_;
 	$self->{quit} = 1;
 	%{$self->{opendirs}} = ();
@@ -260,7 +260,7 @@ sub quit {
 		my $di = $PublicInbox::DS::DescriptorMap{$fd};
 		$di->close if $di && $di->can('add_watches');
 	}
-	if (my $idle_mic = delete $self->{idle_mic}) {
+	if (my $idle_mic = delete $self->{idle_mic}) { # IMAP child
 		return unless $idle_mic->IsConnected && $idle_mic->Socket;
 		eval { $idle_mic->done };
 		if ($@ && $idle_mic->IsConnected && $idle_mic->Socket) {
@@ -387,7 +387,8 @@ sub watch_atfork_child ($) {
 	delete $self->{opendirs};
 	PublicInbox::DS->Reset;
 	my $sig = delete $self->{sig};
-	$sig->{CHLD} = 'DEFAULT';
+	$sig->{CHLD} = $sig->{HUP} = $sig->{USR1} = 'DEFAULT';
+	# TERM/QUIT/INT call ->quit, which works in both parent+child
 	@SIG{keys %$sig} = values %$sig;
 	PublicInbox::DS::sig_setmask(PublicInbox::DS::allowset($sig));
 }
diff --git a/script/public-inbox-watch b/script/public-inbox-watch
index 870cd31b..d9215de9 100755
--- a/script/public-inbox-watch
+++ b/script/public-inbox-watch
@@ -41,7 +41,7 @@ if ($watch) {
 		warn "# scanning\n";
 		$watch->trigger_scan('full');
 	};
-	my $quit = sub {
+	my $quit = sub { # may be called in IMAP/NNTP children
 		$watch->quit if $watch;
 		$watch = undef;
 		$0 .= ' quitting';

^ permalink raw reply related	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2023-09-08 10:51 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-09-08 10:51 [PATCH 0/3] sig handling fixes for non-signalfd/kqueue Eric Wong
2023-09-08 10:51 ` [PATCH 1/3] ds: fix signals unblock " Eric Wong
2023-09-08 10:51 ` [PATCH 2/3] watch: set %SIG " Eric Wong
2023-09-08 10:51 ` [PATCH 3/3] watch: reset HUP + USR1 signal handlers in children 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).