diff options
Diffstat (limited to 'lib/PublicInbox/DS.pm')
| -rw-r--r-- | lib/PublicInbox/DS.pm | 699 |
1 files changed, 362 insertions, 337 deletions
diff --git a/lib/PublicInbox/DS.pm b/lib/PublicInbox/DS.pm index 9c278307..a6fec954 100644 --- a/lib/PublicInbox/DS.pm +++ b/lib/PublicInbox/DS.pm @@ -21,42 +21,43 @@ # (tmpio = [ GLOB, offset, [ length ] ]) package PublicInbox::DS; use strict; -use bytes; -use POSIX qw(WNOHANG); -use IO::Handle qw(); +use v5.10.1; +use parent qw(Exporter); +use bytes qw(length substr); # FIXME(?): needed for PublicInbox::NNTP +use POSIX qw(WNOHANG sigprocmask SIG_SETMASK SIG_UNBLOCK); use Fcntl qw(SEEK_SET :DEFAULT O_APPEND); use Time::HiRes qw(clock_gettime CLOCK_MONOTONIC); -use parent qw(Exporter); -our @EXPORT_OK = qw(now msg_more); -use 5.010_001; use Scalar::Util qw(blessed); -use PublicInbox::Syscall qw(:epoll); +use PublicInbox::Syscall qw(%SIGNUM + EPOLLIN EPOLLOUT EPOLLONESHOT EPOLLEXCLUSIVE); use PublicInbox::Tmpfile; -use Errno qw(EAGAIN EINVAL); -use Carp qw(confess carp); +use PublicInbox::Select; +use PublicInbox::OnDestroy; +use Errno qw(EAGAIN EINVAL ECHILD); +use Carp qw(carp croak); +our @EXPORT_OK = qw(now msg_more awaitpid add_timer add_uniq_timer); my $nextq; # queue for next_tick -my $wait_pids; # list of [ pid, callback, callback_arg ] -my $later_queue; # list of callbacks to run at some later interval -my $EXPMAP; # fd -> idle_time -our $EXPTIME = 180; # 3 minutes -my ($later_timer, $reap_armed, $exp_timer); -my $ToClose; # sockets to close when event loop is done -our ( - %DescriptorMap, # fd (num) -> PublicInbox::DS object - $Epoll, # Global epoll fd (or DSKQXS ref) - $_io, # IO::Handle for Epoll - - $PostLoopCallback, # subref to call at the end of each loop, if defined (global) - - $LoopTimeout, # timeout of event loop in milliseconds - $DoneInit, # if we've done the one-time module init yet +my $reap_armed; +my @active; # FDs (or objs) returned by epoll/kqueue +our (%AWAIT_PIDS, # pid => [ $callback, @args ] + $cur_runq, # only set inside next_tick + @FD_MAP, # fd (num) -> PublicInbox::DS object + $Poller, # global Select, Epoll, DSPoll, or DSKQXS ref + + @post_loop_do, # subref + args to call at the end of each loop + + $loop_timeout, # timeout of event loop in milliseconds @Timers, # timers + %UniqTimer, $in_loop, ); Reset(); +# clobber everything explicitly to avoid DESTROY ordering problems w/ DBI +END { Reset() } + ##################################################################### ### C L A S S M E T H O D S ##################################################################### @@ -67,180 +68,156 @@ Reset all state =cut sub Reset { - %DescriptorMap = (); - $in_loop = $wait_pids = $later_queue = $reap_armed = undef; - $EXPMAP = {}; - $nextq = $ToClose = $later_timer = $exp_timer = undef; - $LoopTimeout = -1; # no timeout by default - @Timers = (); - - $PostLoopCallback = undef; - $DoneInit = 0; - - $_io = undef; # closes real $Epoll FD - $Epoll = undef; # may call DSKQXS::DESTROY + $Poller = bless [], 'PublicInbox::DummyPoller'; + do { + $in_loop = undef; # first in case DESTROY callbacks use this + # clobbering $Poller may call DSKQXS::DESTROY, + # we must always have this set to something to avoid + # needing branches before ep_del/ep_mod calls (via ->close). + @FD_MAP = (); + @Timers = (); + %UniqTimer = (); + @post_loop_do = (); + + # we may be called from an *atfork_child inside next_tick: + @$cur_runq = () if $cur_runq; + @active = (); + $nextq = undef; # may call ep_del + %AWAIT_PIDS = (); + } while (@Timers || $nextq || keys(%AWAIT_PIDS) || + @active || @FD_MAP || + @post_loop_do || keys(%UniqTimer) || + scalar(@{$cur_runq // []})); # do not vivify cur_runq - *EventLoop = *FirstTimeEventLoop; -} - -=head2 C<< CLASS->SetLoopTimeout( $timeout ) >> - -Set the loop timeout for the event loop to some value in milliseconds. - -A timeout of 0 (zero) means poll forever. A timeout of -1 means poll and return -immediately. - -=cut -sub SetLoopTimeout { - return $LoopTimeout = $_[1] + 0; + $reap_armed = undef; + $loop_timeout = -1; # no timeout by default + $Poller = PublicInbox::Select->new; } -=head2 C<< PublicInbox::DS::add_timer( $seconds, $coderef, $arg) >> - -Add a timer to occur $seconds from now. $seconds may be fractional, but timers -are not guaranteed to fire at the exact time you ask for. - -=cut -sub add_timer ($$;$) { - my ($secs, $coderef, $arg) = @_; - - my $fire_time = now() + $secs; - - my $timer = [$fire_time, $coderef, $arg]; - - if (!@Timers || $fire_time >= $Timers[-1][0]) { - push @Timers, $timer; - return $timer; - } - - # Now, where do we insert? (NOTE: this appears slow, algorithm-wise, - # but it was compared against calendar queues, heaps, naive push/sort, - # and a bunch of other versions, and found to be fastest with a large - # variety of datasets.) - for (my $i = 0; $i < @Timers; $i++) { - if ($Timers[$i][0] > $fire_time) { - splice(@Timers, $i, 0, $timer); - return $timer; - } - } - - die "Shouldn't get here."; -} +sub _add_named_timer { + my ($name, $secs, $coderef, @args) = @_; + my $fire_time = now() + $secs; + my $timer = [$fire_time, $name, $coderef, @args]; -# keeping this around in case we support other FD types for now, -# epoll_create1(EPOLL_CLOEXEC) requires Linux 2.6.27+... -sub set_cloexec ($) { - my ($fd) = @_; + if (!@Timers || $fire_time >= $Timers[-1][0]) { + push @Timers, $timer; + return $timer; + } - $_io = IO::Handle->new_from_fd($fd, 'r+') or return; - defined(my $fl = fcntl($_io, F_GETFD, 0)) or return; - fcntl($_io, F_SETFD, $fl | FD_CLOEXEC); + # Now, where do we insert? (NOTE: this appears slow, algorithm-wise, + # but it was compared against calendar queues, heaps, naive push/sort, + # and a bunch of other versions, and found to be fastest with a large + # variety of datasets.) + for (my $i = 0; $i < @Timers; $i++) { + if ($Timers[$i][0] > $fire_time) { + splice(@Timers, $i, 0, $timer); + return $timer; + } + } + die "Shouldn't get here."; } -sub _InitPoller -{ - return if $DoneInit; - $DoneInit = 1; +sub add_timer { _add_named_timer(undef, @_) } - if (PublicInbox::Syscall::epoll_defined()) { - $Epoll = epoll_create(); - set_cloexec($Epoll) if (defined($Epoll) && $Epoll >= 0); - } else { - my $cls; - for (qw(DSKQXS DSPoll)) { - $cls = "PublicInbox::$_"; - last if eval "require $cls"; - } - $cls->import(qw(epoll_ctl epoll_wait)); - $Epoll = $cls->new; - } - *EventLoop = *EpollEventLoop; +sub add_uniq_timer { # ($name, $secs, $coderef, @args) = @_; + $UniqTimer{$_[0]} //= _add_named_timer(@_); } -=head2 C<< CLASS->EventLoop() >> - -Start processing IO events. In most daemon programs this never exits. See -C<PostLoopCallback> below for how to exit the loop. - -=cut -sub FirstTimeEventLoop { - my $class = shift; - - _InitPoller(); - - EventLoop($class); +# caller sets return value to $Poller +sub _InitPoller () { + my @try = ($^O eq 'linux' ? 'Epoll' : 'DSKQXS'); + my $cls; + for (@try, 'DSPoll') { + $cls = "PublicInbox::$_"; + last if eval "require $cls"; + } + $cls->new; } sub now () { clock_gettime(CLOCK_MONOTONIC) } sub next_tick () { - my $q = $nextq or return; - $nextq = undef; - for (@$q) { - # we avoid "ref" on blessed refs to workaround a Perl 5.16.3 leak: - # https://rt.perl.org/Public/Bug/Display.html?id=114340 - if (blessed($_)) { - $_->event_step; - } else { - $_->(); - } - } + $cur_runq = $nextq or return; + $nextq = undef; + while (my $obj = shift @$cur_runq) { + # avoid "ref" on blessed refs to workaround a Perl 5.16.3 leak: + # https://rt.perl.org/Public/Bug/Display.html?id=114340 + blessed($obj) ? $obj->event_step : $obj->(); + } + 1; } # runs timers and returns milliseconds for next one, or next event loop sub RunTimers { - next_tick(); + my $ran = next_tick(); - return (($nextq || $ToClose) ? 0 : $LoopTimeout) unless @Timers; + return ($nextq || $ran ? 0 : $loop_timeout) unless @Timers; - my $now = now(); + my $now = now(); - # Run expired timers - while (@Timers && $Timers[0][0] <= $now) { - my $to_run = shift(@Timers); - $to_run->[1]->($to_run->[2]); - } + # Run expired timers + while (@Timers && $Timers[0][0] <= $now) { + my $to_run = shift(@Timers); + delete $UniqTimer{$to_run->[1] // ''}; + $to_run->[2]->(@$to_run[3..$#$to_run]); + $ran = 1; + } - # timers may enqueue into nextq: - return 0 if ($nextq || $ToClose); + # timers may enqueue into nextq: + return 0 if $nextq || $ran; - return $LoopTimeout unless @Timers; + return $loop_timeout unless @Timers; - # convert time to an even number of milliseconds, adding 1 - # extra, otherwise floating point fun can occur and we'll - # call RunTimers like 20-30 times, each returning a timeout - # of 0.0000212 seconds - my $timeout = int(($Timers[0][0] - $now) * 1000) + 1; + # convert time to an even number of milliseconds, adding 1 + # extra, otherwise floating point fun can occur and we'll + # call RunTimers like 20-30 times, each returning a timeout + # of 0.0000212 seconds + my $t = int(($Timers[0][0] - $now) * 1000) + 1; - # -1 is an infinite timeout, so prefer a real timeout - return $timeout if $LoopTimeout == -1; + # -1 is an infinite timeout, so prefer a real timeout + ($loop_timeout < 0 || $loop_timeout >= $t) ? $t : $loop_timeout +} - # otherwise pick the lower of our regular timeout and time until - # the next timer - return $LoopTimeout if $LoopTimeout < $timeout; - return $timeout; +sub sig_setmask { sigprocmask(SIG_SETMASK, @_) or die "sigprocmask: $!" } + +# ensure we detect bugs, HW problems and user rlimits +our @UNBLOCKABLE = (POSIX::SIGABRT, POSIX::SIGBUS, POSIX::SIGFPE, + POSIX::SIGILL, POSIX::SIGSEGV, POSIX::SIGXCPU, POSIX::SIGXFSZ); + +sub block_signals { # anything in @_ stays unblocked + my $newset = POSIX::SigSet->new; + $newset->fillset or die "fillset: $!"; + for (@_, @UNBLOCKABLE) { $newset->delset($_) or die "delset($_): $!" } + my $oldset = POSIX::SigSet->new; + sig_setmask($newset, $oldset); + $oldset; } -# We can't use waitpid(-1) safely here since it can hit ``, system(), -# and other things. So we scan the $wait_pids list, which is hopefully -# not too big. We keep $wait_pids small by not calling dwaitpid() -# until we've hit EOF when reading the stdout of the child. +sub await_cb ($;@) { + my ($pid, @cb_args) = @_; + my $cb = shift @cb_args or return; + eval { $cb->($pid, @cb_args) }; + warn "E: awaitpid($pid): $@" if $@; +} +# This relies on our Perl process being single-threaded, or at least +# no threads spawning and waiting on processes (``, system(), etc...) +# Threads are officially discouraged by the Perl5 team, and I expect +# that to remain the case. sub reap_pids { $reap_armed = undef; - my $tmp = $wait_pids or return; - $wait_pids = undef; - foreach my $ary (@$tmp) { - my ($pid, $cb, $arg) = @$ary; - my $ret = waitpid($pid, WNOHANG); - if ($ret == 0) { - push @$wait_pids, $ary; # autovivifies @$wait_pids - } elsif ($cb) { - eval { $cb->($arg, $pid) }; + while (1) { + my $pid = waitpid(-1, WNOHANG) or return; + if (defined(my $cb_args = delete $AWAIT_PIDS{$pid})) { + await_cb($pid, @$cb_args) if $cb_args; + } elsif ($pid == -1 && $! == ECHILD) { + return requeue(\&dflush); # force @post_loop_do to run + } elsif ($pid > 0) { + warn "W: reaped unknown PID=$pid: \$?=$?\n"; + } else { # does this happen? + return warn("W: waitpid(-1, WNOHANG) => $pid ($!)"); } } - # we may not be done, yet, and could've missed/masked a SIGCHLD: - $reap_armed //= requeue(\&reap_pids) if $wait_pids; } # reentrant SIGCHLD handler (since reap_pids is not reentrant) @@ -248,65 +225,80 @@ sub enqueue_reap () { $reap_armed //= requeue(\&reap_pids) } sub in_loop () { $in_loop } +# use inside @post_loop_do, returns number of busy clients +sub close_non_busy () { + my $n = 0; + for my $s (grep defined, @FD_MAP) { + # close as much as possible, early as possible + ($s->busy ? ++$n : $s->close) if $s->can('busy'); + } + $n; +} + # Internal function: run the post-event callback, send read events # for pushed-back data, and close pending connections. returns 1 # if event loop should continue, or 0 to shut it all down. sub PostEventLoop () { - # now we can close sockets that wanted to close during our event - # processing. (we didn't want to close them during the loop, as we - # didn't want fd numbers being reused and confused during the event - # loop) - if (my $close_now = $ToClose) { - $ToClose = undef; # will be autovivified on push - @$close_now = map { fileno($_) } @$close_now; - - # order matters, destroy expiry times, first: - delete @$EXPMAP{@$close_now}; - - # ->DESTROY methods may populate ToClose - delete @DescriptorMap{@$close_now}; - } - # by default we keep running, unless a postloop callback cancels it - $PostLoopCallback ? $PostLoopCallback->(\%DescriptorMap) : 1; -} - -sub EpollEventLoop { - local $in_loop = 1; - do { - my @events; - my $i; - my $timeout = RunTimers(); - - # get up to 1000 events - my $evcount = epoll_wait($Epoll, 1000, $timeout, \@events); - for ($i=0; $i<$evcount; $i++) { - # it's possible epoll_wait returned many events, including some at the end - # that ones in the front triggered unregister-interest actions. if we - # can't find the %sock entry, it's because we're no longer interested - # in that event. - $DescriptorMap{$events[$i]->[0]}->event_step; - } - } while (PostEventLoop()); - _run_later(); + @post_loop_do ? $post_loop_do[0]->(@post_loop_do[1..$#post_loop_do]) : 1 } -=head2 C<< CLASS->SetPostLoopCallback( CODEREF ) >> - -Sets post loop callback function. Pass a subref and it will be -called every time the event loop finishes. - -Return 1 (or any true value) from the sub to make the loop continue, 0 or false -and it will exit. +sub sigset_prep ($$$) { + my ($sig, $init, $each) = @_; # $sig: { signame => whatever } + my $ret = POSIX::SigSet->new; + $ret->$init or die "$init: $!"; + for my $s (keys %$sig) { + my $num = $SIGNUM{$s} // POSIX->can("SIG$s")->(); + $ret->$each($num) or die "$each ($s => $num): $!"; + } + 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 (;$$) { + my ($sig, $oldset) = @_; + $Poller //= _InitPoller(); + require PublicInbox::Sigfd if $sig; + my $sigfd = $sig ? PublicInbox::Sigfd->new($sig) : undef; + if ($sigfd && $sigfd->{is_kq}) { + my $tmp = allowset($sig); + local @SIG{keys %$sig} = values(%$sig); + sig_setmask($tmp, my $old = POSIX::SigSet->new); + # Unlike Linux signalfd, EVFILT_SIGNAL can't handle + # signals received before the filter is created, + # so we peek at signals here. + sig_setmask($old); + } + local @SIG{keys %$sig} = values(%$sig) if $sig && !$sigfd; + local $SIG{PIPE} = 'IGNORE'; + if (!$sigfd && $sig) { + # wake up every second to accept signals if we don't + # have signalfd or IO::KQueue: + sig_setmask($oldset) if $oldset; + sigprocmask(SIG_UNBLOCK, unblockset($sig)) or + die "SIG_UNBLOCK: $!"; + $loop_timeout = 1000; + } + $_[0] = $sigfd = $sig = undef; # $_[0] == sig + local $in_loop = 1; + do { + my $timeout = RunTimers(); -The callback function will be passed two parameters: \%DescriptorMap + # grab whatever FDs are ready + $Poller->ep_wait($timeout, \@active); -=cut -sub SetPostLoopCallback { - my ($class, $ref) = @_; + # map all FDs to their associated Perl object + @active = @FD_MAP[@active]; - # global callback - $PostLoopCallback = (defined $ref && ref $ref eq 'CODE') ? $ref : undef; + while (my $obj = shift @active) { + $obj->event_step; + } + } while (PostEventLoop()); } ##################################################################### @@ -318,7 +310,7 @@ sub SetPostLoopCallback { =head2 C<< CLASS->new( $socket ) >> Create a new PublicInbox::DS subclass object for the given I<socket> which will -react to events on it during the C<EventLoop>. +react to events on it during the C<event_loop>. This is normally (always?) called from your subclass via: @@ -330,62 +322,54 @@ sub new { $self->{sock} = $sock; my $fd = fileno($sock); - _InitPoller(); - - if (epoll_ctl($Epoll, EPOLL_CTL_ADD, $fd, $ev)) { + $Poller //= _InitPoller(); +retry: + if ($Poller->ep_add($sock, $ev)) { if ($! == EINVAL && ($ev & EPOLLEXCLUSIVE)) { $ev &= ~EPOLLEXCLUSIVE; goto retry; } - die "couldn't add epoll watch for $fd: $!\n"; + die "EPOLL_CTL_ADD $self/$sock/$fd: $!"; } - confess("DescriptorMap{$fd} defined ($DescriptorMap{$fd})") - if defined($DescriptorMap{$fd}); + defined($FD_MAP[$fd]) and + croak("BUG: FD:$fd in use by $FD_MAP[$fd] (for $self/$sock)"); - $DescriptorMap{$fd} = $self; + $FD_MAP[$fd] = $self; } - -##################################################################### -### I N S T A N C E M E T H O D S -##################################################################### +# for IMAP, NNTP, and POP3 which greet clients upon connect +sub greet { + my ($self, $sock) = @_; + my $ev = EPOLLIN; + my $wbuf; + if ($sock->can('accept_SSL') && !$sock->accept_SSL) { + return if $! != EAGAIN || !($ev = PublicInbox::TLS::epollbit()); + $wbuf = [ \&accept_tls_step, $self->can('do_greet')]; + } + new($self, $sock, $ev | EPOLLONESHOT); + if ($wbuf) { + $self->{wbuf} = $wbuf; + } else { + $self->do_greet; + } + $self; +} sub requeue ($) { push @$nextq, $_[0] } # autovivifies -=head2 C<< $obj->close >> - -Close the socket. - -=cut +# drop the IO::Handle ref, true if successful, false if not (or already dropped) +# (this is closer to CORE::close than Danga::Socket::close) sub close { - my ($self) = @_; - my $sock = delete $self->{sock} or return; - - # we need to flush our write buffer, as there may - # be self-referential closures (sub { $client->close }) - # preventing the object from being destroyed - delete $self->{wbuf}; - - # if we're using epoll, we have to remove this from our epoll fd so we stop getting - # notifications about it - my $fd = fileno($sock); - epoll_ctl($Epoll, EPOLL_CTL_DEL, $fd, 0) and - confess("EPOLL_CTL_DEL: $!"); - - # we explicitly don't delete from DescriptorMap here until we - # actually close the socket, as we might be in the middle of - # processing an epoll_wait/etc that returned hundreds of fds, one - # of which is not yet processed and is what we're closing. if we - # keep it in DescriptorMap, then the event harnesses can just - # looked at $pob->{sock} == undef and ignore it. but if it's an - # un-accounted for fd, then it (understandably) freak out a bit - # and emit warnings, thinking their state got off. + my ($self) = @_; + my $sock = delete $self->{sock} or return; - # defer closing the actual socket until the event loop is done - # processing this round of events. (otherwise we might reuse fds) - push @$ToClose, $sock; # autovivifies $ToClose + # we need to clear our write buffer, as there may + # be self-referential closures (sub { $client->close }) + # preventing the object from being destroyed + delete $self->{wbuf}; + $FD_MAP[fileno($sock)] = undef; - return 0; + !$Poller->ep_del($sock); # stop getting notifications } # portable, non-thread-safe sendfile emulation (no pread, yet) @@ -431,8 +415,8 @@ next_buf: shift @$wbuf; goto next_buf; } - } elsif ($! == EAGAIN) { - epwait($sock, epbit($sock, EPOLLOUT) | EPOLLONESHOT); + } elsif ($! == EAGAIN && (my $ev = epbit($sock, EPOLLOUT))) { + epwait($sock, $ev | EPOLLONESHOT); return 0; } else { return $self->close; @@ -461,39 +445,40 @@ sub rbuf_idle ($$) { } } +# returns true if bytes are read, false otherwise sub do_read ($$$;$) { - my ($self, $rbuf, $len, $off) = @_; - my $r = sysread(my $sock = $self->{sock}, $$rbuf, $len, $off // 0); - return ($r == 0 ? $self->close : $r) if defined $r; - # common for clients to break connections without warning, - # would be too noisy to log here: - if ($! == EAGAIN) { - epwait($sock, epbit($sock, EPOLLIN) | EPOLLONESHOT); - rbuf_idle($self, $rbuf); - 0; - } else { - $self->close; - } + my ($self, $rbuf, $len, $off) = @_; + my ($ev, $r, $s); + $r = sysread($s = $self->{sock}, $$rbuf, $len, $off // 0) and return $r; + + if (!defined($r) && $! == EAGAIN && ($ev = epbit($s, EPOLLIN))) { + epwait($s, $ev | EPOLLONESHOT); + rbuf_idle($self, $rbuf); + } else { + $self->close; + } + undef; } # drop the socket if we hit unrecoverable errors on our system which # require BOFH attention: ENOSPC, EFBIG, EIO, EMFILE, ENFILE... sub drop { - my $self = shift; - carp(@_); - $self->close; + my $self = shift; + carp(@_); + $self->close; + undef; } -# n.b.: use ->write/->read for this buffer to allow compatibility with -# PerlIO::mmap or PerlIO::scalar if needed sub tmpio ($$$) { - my ($self, $bref, $off) = @_; - my $fh = tmpfile('wbuf', $self->{sock}, O_APPEND) or - return drop($self, "tmpfile $!"); - $fh->autoflush(1); - my $len = bytes::length($$bref) - $off; - $fh->write($$bref, $len, $off) or return drop($self, "write ($len): $!"); - [ $fh, 0 ] # [1] = offset, [2] = length, not set by us + my ($self, $bref, $off) = @_; + my $fh = tmpfile('wbuf', $self->{sock}, O_APPEND) or + return drop($self, "tmpfile $!"); + $fh->autoflush(1); + my $len = length($$bref) - $off; + my $n = syswrite($fh, $$bref, $len, $off) // + return drop($self, "write ($len): $!"); + $n == $len or return drop($self, "wrote $n < $len bytes"); + [ $fh, 0 ] # [1] = offset, [2] = length, not set by us } =head2 C<< $obj->write( $data ) >> @@ -523,7 +508,8 @@ sub write { push @$wbuf, $bref; } else { my $tmpio = $wbuf->[-1]; - if ($tmpio && !defined($tmpio->[2])) { # append to tmp file buffer + if (ref($tmpio) eq 'ARRAY' && !defined($tmpio->[2])) { + # append to tmp file buffer $tmpio->[0]->print($$bref) or return drop($self, "print: $!"); } else { my $tmpio = tmpio($self, $bref, 0) or return 0; @@ -535,14 +521,15 @@ sub write { $bref->($self); return 1; } else { - my $to_write = bytes::length($$bref); + my $to_write = length($$bref); my $written = syswrite($sock, $$bref, $to_write); if (defined $written) { return 1 if $written == $to_write; requeue($self); # runs: event_step -> flush_write } elsif ($! == EAGAIN) { - epwait($sock, epbit($sock, EPOLLOUT) | EPOLLONESHOT); + my $ev = epbit($sock, EPOLLOUT) or return $self->close; + epwait($sock, $ev | EPOLLONESHOT); $written = 0; } else { return $self->close; @@ -569,7 +556,7 @@ sub msg_more ($$) { !$sock->can('stop_SSL')) { my $n = send($sock, $_[1], MSG_MORE); if (defined $n) { - my $nlen = bytes::length($_[1]) - $n; + my $nlen = length($_[1]) - $n; return 1 if $nlen == 0; # all done! # queue up the unwritten substring: my $tmpio = tmpio($self, \($_[1]), $n) or return 0; @@ -584,9 +571,8 @@ sub msg_more ($$) { } sub epwait ($$) { - my ($sock, $ev) = @_; - epoll_ctl($Epoll, EPOLL_CTL_MOD, fileno($sock), $ev) and - confess("EPOLL_CTL_MOD $!"); + my ($io, $ev) = @_; + $Poller->ep_mod($io, $ev) and croak("EPOLL_CTL_MOD($io): $!"); } # return true if complete, false if incomplete (or failure) @@ -595,84 +581,123 @@ sub accept_tls_step ($) { my $sock = $self->{sock} or return; return 1 if $sock->accept_SSL; return $self->close if $! != EAGAIN; - epwait($sock, PublicInbox::TLS::epollbit() | EPOLLONESHOT); + my $ev = PublicInbox::TLS::epollbit() or return $self->close; + epwait($sock, $ev | EPOLLONESHOT); unshift(@{$self->{wbuf}}, \&accept_tls_step); # autovivifies 0; } -# return true if complete, false if incomplete (or failure) +# return value irrelevant sub shutdn_tls_step ($) { my ($self) = @_; my $sock = $self->{sock} or return; return $self->close if $sock->stop_SSL(SSL_fast_shutdown => 1); return $self->close if $! != EAGAIN; - epwait($sock, PublicInbox::TLS::epollbit() | EPOLLONESHOT); + my $ev = PublicInbox::TLS::epollbit() or return $self->close; + epwait($sock, $ev | EPOLLONESHOT); unshift(@{$self->{wbuf}}, \&shutdn_tls_step); # autovivifies - 0; } # don't bother with shutdown($sock, 2), we don't fork+exec w/o CLOEXEC # or fork w/o exec, so no inadvertent socket sharing sub shutdn ($) { - my ($self) = @_; - my $sock = $self->{sock} or return; - if ($sock->can('stop_SSL')) { - shutdn_tls_step($self); - } else { - $self->close; - } + my ($self) = @_; + my $sock = $self->{sock} or return; + $sock->can('stop_SSL') ? shutdn_tls_step($self) : $self->close; } -# must be called with eval, PublicInbox::DS may not be loaded (see t/qspawn.t) -sub dwaitpid ($$$) { - die "Not in EventLoop\n" unless $in_loop; - push @$wait_pids, [ @_ ]; # [ $pid, $cb, $arg ] +sub dflush {} # overridden by DSdeflate +sub compressed {} # overridden by DSdeflate +sub long_response_done {} # overridden by Net::NNTP - # We could've just missed our SIGCHLD, cover it, here: - enqueue_reap(); +sub long_step { + my ($self) = @_; + # wbuf is unset or empty, here; {long} may add to it + my ($fd, $cb, $t0, @args) = @{$self->{long_cb}}; + my $more = eval { $cb->($self, @args) }; + if ($@ || !$self->{sock}) { # something bad happened... + delete $self->{long_cb}; + my $elapsed = now() - $t0; + $@ and warn("$@ during long response[$fd] - ", + sprintf('%0.6f', $elapsed),"\n"); + $self->out(" deferred[$fd] aborted - %0.6f", $elapsed); + $self->close; + } elsif ($more) { # $self->{wbuf}: + # control passed to ibx_async_cat if $more == \undef + requeue_once($self) if !ref($more); + } else { # all done! + delete $self->{long_cb}; + $self->long_response_done; + my $elapsed = now() - $t0; + $self->out(" deferred[$fd] done - %0.6f", $elapsed); + my $wbuf = $self->{wbuf}; # do NOT autovivify + requeue($self) unless $wbuf && @$wbuf; + } } -sub _run_later () { - my $run = $later_queue or return; - $later_timer = $later_queue = undef; - $_->() for @$run; -} +sub requeue_once { + my ($self) = @_; + # COMPRESS users all share the same DEFLATE context. + # Flush it here to ensure clients don't see each other's data + $self->dflush; -sub later ($) { - push @$later_queue, $_[0]; # autovivifies @$later_queue - $later_timer //= add_timer(60, \&_run_later); + # no recursion, schedule another call ASAP, + # but only after all pending writes are done. + # autovivify wbuf. wbuf may be populated by $cb, + # no need to rearm if so: (push returns new size of array) + $self->requeue if push(@{$self->{wbuf}}, \&long_step) == 1; } -sub expire_old () { - my $now = now(); - my $exp = $EXPTIME; - my $old = $now - $exp; - my %new; - while (my ($fd, $idle_at) = each %$EXPMAP) { - if ($idle_at < $old) { - my $ds_obj = $DescriptorMap{$fd}; - $new{$fd} = $idle_at if !$ds_obj->shutdn; +sub long_response ($$;@) { + my ($self, $cb, @args) = @_; # cb returns true if more, false if done + my $sock = $self->{sock} or return; + # make sure we disable reading during a long response, + # clients should not be sending us stuff and making us do more + # work while we are stream a response to them + $self->{long_cb} = [ fileno($sock), $cb, now(), @args ]; + long_step($self); # kick off! + undef; +} + +sub awaitpid { + my ($pid, @cb_args) = @_; # @cb_args = ($cb, @args), $cb may be undef + $AWAIT_PIDS{$pid} = \@cb_args if @cb_args; + # provide synchronous API + if (defined(wantarray) || (!$in_loop && !@cb_args)) { + my $ret = waitpid($pid, 0); + if ($ret == $pid) { + my $cb_args = delete $AWAIT_PIDS{$pid}; + @cb_args = @$cb_args if !@cb_args && $cb_args; + await_cb($pid, @cb_args); } else { - $new{$fd} = $idle_at; + carp "waitpid($pid) => $ret ($!)"; + delete $AWAIT_PIDS{$pid}; } + return $ret; + } elsif ($in_loop) { # We could've just missed our SIGCHLD, cover it, here: + enqueue_reap(); } - $EXPMAP = \%new; - $exp_timer = scalar(keys %new) ? later(\&expire_old) : undef; } -sub update_idle_time { - my ($self) = @_; - my $sock = $self->{sock} or return; - $EXPMAP->{fileno($sock)} = now(); - $exp_timer //= later(\&expire_old); +# for persistent child process +sub fork_persist () { + my $seed = rand(0xffffffff); + my $pid = PublicInbox::OnDestroy::fork_tmp; + if ($pid == 0) { + srand($seed); + eval { Net::SSLeay::randomize() }; # may not be loaded + Reset(); + } + $pid; } -sub not_idle_long { - my ($self, $now) = @_; - my $sock = $self->{sock} or return; - my $idle_at = $EXPMAP->{fileno($sock)} or return; - ($idle_at + $EXPTIME) > $now; -} +package PublicInbox::DummyPoller; # only used during Reset +use v5.12; + +sub ep_del {} +no warnings 'once'; +*ep_add = \&ep_del; +*ep_mod = \&ep_del; 1; |