lei: fix inadvertant FD sharing

$wq->{-ipc_atfork_child_close} neededed to be initialized properly.
And start setting $0 in workers to improve visibility.

5 files changed, 47 insertions, 34 deletions

diff --git a/lib/PublicInbox/IPC.pm b/lib/PublicInbox/IPC.pm
index 8fec2e62..24f45e03 100644
--- a/lib/PublicInbox/IPC.pm
+++ b/lib/PublicInbox/IPC.pm
@@ -134,6 +134,12 @@ sub ipc_worker_reap { # dwaitpid callback
         warn "PID:$pid died with \$?=$?\n" if $? && ($? & 127) != 15;
 }
 
+sub wq_wait_old {
+        my ($self) = @_;
+        my $pids = delete $self->{"-wq_old_pids.$$"} or return;
+        dwaitpid($_, \&ipc_worker_reap, $self) for @$pids;
+}
+
 # for base class, override in sub classes
 sub ipc_atfork_prepare {}
 
@@ -370,17 +376,25 @@ sub wq_workers {
 }
 
 sub wq_close {
-        my ($self) = @_;
+        my ($self, $nohang) = @_;
         delete @$self{qw(-wq_s1 -wq_s2)} or return;
         my $ppid = delete $self->{-wq_ppid} or return;
         my $workers = delete $self->{-wq_workers} // die 'BUG: no wq_workers';
         return if $ppid != $$; # can't reap siblings or parents
-        return (keys %$workers) if wantarray; # caller will reap
-        for my $pid (keys %$workers) {
-                dwaitpid($pid, \&ipc_worker_reap, $self);
+        my @pids = map { $_ + 0 } keys %$workers;
+        if ($nohang) {
+                push @{$self->{"-wq_old_pids.$$"}}, @pids;
+        } else {
+                dwaitpid($_, \&ipc_worker_reap, $self) for @pids;
         }
 }
 
+sub wq_kill_old {
+        my ($self) = @_;
+        my $pids = $self->{"-wq_old_pids.$$"} or return;
+        kill 'TERM', @$pids;
+}
+
 sub wq_kill {
         my ($self, $sig) = @_;
         my $workers = $self->{-wq_workers} or return;
diff --git a/lib/PublicInbox/LEI.pm b/lib/PublicInbox/LEI.pm
index 6be6d10b..2cb2bf40 100644
--- a/lib/PublicInbox/LEI.pm
+++ b/lib/PublicInbox/LEI.pm
@@ -281,11 +281,14 @@ sub fail ($$;$) {
 
 sub atfork_prepare_wq {
         my ($self, $wq) = @_;
-        my $tcafc = $wq->{-ipc_atfork_child_close};
+        my $tcafc = $wq->{-ipc_atfork_child_close} //= [];
         push @$tcafc, @TO_CLOSE_ATFORK_CHILD;
         if (my $sock = $self->{sock}) {
                 push @$tcafc, @$self{qw(0 1 2)}, $sock;
         }
+        if (my $pgr = $self->{pgr}) {
+                push @$tcafc, @$pgr[1,2];
+        }
         for my $f (qw(lxs l2m)) {
                 my $ipc = $self->{$f} or next;
                 push @$tcafc, grep { defined }
@@ -335,9 +338,7 @@ sub atfork_parent_wq {
         my $l2m = $ret->{l2m};
         if ($l2m && $l2m != $wq) { # $wq == lxs
                 $io[4] = $l2m->{-wq_s1} if $l2m->{-wq_s1};
-                if (my @pids = $l2m->wq_close) {
-                        $wq->{l2m_pids} = \@pids;
-                }
+                $l2m->wq_close(1);
         }
         ($ret, @io);
 }
diff --git a/lib/PublicInbox/LeiQuery.pm b/lib/PublicInbox/LeiQuery.pm
index 941bc299..7d634b5e 100644
--- a/lib/PublicInbox/LeiQuery.pm
+++ b/lib/PublicInbox/LeiQuery.pm
@@ -32,24 +32,25 @@ sub lei_q {
                 my $sto = $self->_lei_store(1);
                 push @srcs, $sto->search;
         }
-        my $lxs = PublicInbox::LeiXSearch->new;
 
+        my $lxs = $self->{lxs} = PublicInbox::LeiXSearch->new;
         # --external is enabled by default, but allow --no-external
-        if ($opt->{external} // 1) {
+        if ($opt->{external} //= 1) {
                 $self->_externals_each(\&_vivify_external, \@srcs);
         }
-        my $j = $opt->{jobs} // (scalar(@srcs) > 3 ? 3 : scalar(@srcs));
-        $j = 1 if !$opt->{thread};
-        $self->atfork_prepare_wq($lxs);
-        $lxs->wq_workers_start('lei_xsearch', $j, $self->oldset);
-        $self->{lxs} = $lxs;
-
+        my $xj = $opt->{jobs} // (scalar(@srcs) > 3 ? 3 : scalar(@srcs));
+        $xj = 1 if !$opt->{thread};
         my $ovv = PublicInbox::LeiOverview->new($self) or return;
+        $self->atfork_prepare_wq($lxs);
+        $lxs->wq_workers_start('lei_xsearch', $xj, $self->oldset);
+        delete $lxs->{-ipc_atfork_child_close};
         if (my $l2m = $self->{l2m}) {
-                $j = 4 if $j <= 4; # TODO configurable
+                my $mj = 4; # TODO: configurable
                 $self->atfork_prepare_wq($l2m);
-                $l2m->wq_workers_start('lei2mail', $j, $self->oldset);
+                $l2m->wq_workers_start('lei2mail', $mj, $self->oldset);
+                delete $l2m->{-ipc_atfork_child_close};
         }
+
         # no forking workers after this
 
         my %mset_opt = map { $_ => $opt->{$_} } qw(thread limit offset);
diff --git a/lib/PublicInbox/LeiToMail.pm b/lib/PublicInbox/LeiToMail.pm
index 3dcce9e7..87cc9c47 100644
--- a/lib/PublicInbox/LeiToMail.pm
+++ b/lib/PublicInbox/LeiToMail.pm
@@ -467,7 +467,7 @@ sub write_mail { # via ->wq_do
 
 sub ipc_atfork_prepare {
         my ($self) = @_;
-        # (qry_status_wr, stdout|mbox, stderr, 3: sock, 4: each_smsg_done_wr)
+        # (done_wr, stdout|mbox, stderr, 3: sock, 4: each_smsg_done_wr)
         $self->wq_set_recv_modes(qw[+<&= >&= >&= +<&= >&=]);
         $self->SUPER::ipc_atfork_prepare; # PublicInbox::IPC
 }
diff --git a/lib/PublicInbox/LeiXSearch.pm b/lib/PublicInbox/LeiXSearch.pm
index 13611882..7b33677e 100644
--- a/lib/PublicInbox/LeiXSearch.pm
+++ b/lib/PublicInbox/LeiXSearch.pm
@@ -110,6 +110,7 @@ sub wait_startq ($) {
 
 sub query_thread_mset { # for --thread
         my ($self, $lei, $ibxish) = @_;
+        local $0 = "$0 query_thread_mset";
         my $startq = delete $self->{5};
         my %sig = $lei->atfork_child_wq($self);
         local @SIG{keys %sig} = values %sig;
@@ -148,6 +149,7 @@ sub query_thread_mset { # for --thread
 
 sub query_mset { # non-parallel for non-"--thread" users
         my ($self, $lei, $srcs) = @_;
+        local $0 = "$0 query_mset";
         my $startq = delete $self->{5};
         my %sig = $lei->atfork_child_wq($self);
         local @SIG{keys %sig} = values %sig;
@@ -192,12 +194,10 @@ sub git {
 sub query_done { # EOF callback
         my ($self, $lei) = @_;
         my $l2m = delete $lei->{l2m};
-        if (my $pids = delete $self->{l2m_pids}) {
-                my $ipc_worker_reap = $self->can('ipc_worker_reap');
-                dwaitpid($_, $ipc_worker_reap, $l2m) for @$pids;
-        }
+        $l2m->wq_wait_old if $l2m;
+        $self->wq_wait_old;
         $lei->{ovv}->ovv_end($lei);
-        if ($l2m) { # calls LeiToMail reap_compress
+        if ($l2m) { # close() calls LeiToMail reap_compress
                 close(delete($lei->{1})) if $lei->{1};
                 $lei->start_mua;
         }
@@ -232,12 +232,12 @@ sub start_query { # always runs in main (lei-daemon) process
         for my $rmt (@$remotes) {
                 $self->wq_do('query_thread_mbox', $io, $lei, $rmt);
         }
-        close $io->[0]; # qry_status_wr
         @$io = ();
 }
 
 sub query_prepare { # called by wq_do
         my ($self, $lei) = @_;
+        local $0 = "$0 query_prepare";
         my %sig = $lei->atfork_child_wq($self);
         -p $lei->{0} or die "BUG: \$done pipe expected";
         local @SIG{keys %sig} = values %sig;
@@ -246,11 +246,11 @@ sub query_prepare { # called by wq_do
         syswrite($lei->{0}, '.') == 1 or die "do_post_augment trigger: $!";
 }
 
-sub sigpipe_handler {
-        my ($self, $lei_orig, $pids) = @_;
-        if ($pids) { # one-shot (no event loop)
-                kill 'TERM', @$pids;
+sub sigpipe_handler { # handles SIGPIPE from wq workers
+        my ($self, $lei_orig) = @_;
+        if ($self->wq_kill_old) {
                 kill 'PIPE', $$;
+                $self->wq_wait_old;
         } else {
                 $self->wq_kill;
                 $self->wq_close;
@@ -287,19 +287,16 @@ sub do_query {
                 $io[1] = $zpipe->[1] if $zpipe;
         }
         start_query($self, \@io, $lei, $srcs);
+        $self->wq_close(1);
         unless ($in_loop) {
-                my @pids = $self->wq_close;
                 # for the $lei->atfork_child_wq PIPE handler:
-                $done_op->{'!'}->[3] = \@pids;
                 while ($done->{sock}) { $done->event_step }
-                my $ipc_worker_reap = $self->can('ipc_worker_reap');
-                dwaitpid($_, $ipc_worker_reap, $self) for @pids;
         }
 }
 
 sub ipc_atfork_prepare {
         my ($self) = @_;
-        # (0: qry_status_wr, 1: stdout|mbox, 2: stderr,
+        # (0: done_wr, 1: stdout|mbox, 2: stderr,
         #  3: sock, 4: $l2m->{-wq_s1}, 5: $startq)
         $self->wq_set_recv_modes(qw[+<&= >&= >&= +<&= +<&= <&=]);
         $self->SUPER::ipc_atfork_prepare; # PublicInbox::IPC