1 # Copyright (C) 2016-2021 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # ref: https://cr.yp.to/proto/maildir.html
5 # httsp://wiki2.dovecot.org/MailboxFormat/Maildir
6 package PublicInbox::Watch;
10 use PublicInbox::InboxWritable qw(eml_from_path);
11 use PublicInbox::MdirReader;
12 use PublicInbox::NetReader;
13 use PublicInbox::Filter::Base qw(REJECT);
14 use PublicInbox::Spamcheck;
15 use PublicInbox::Sigfd;
16 use PublicInbox::DS qw(now add_timer);
17 use PublicInbox::MID qw(mids);
18 use PublicInbox::ContentHash qw(content_hash);
19 use PublicInbox::EOFpipe;
20 use POSIX qw(_exit WNOHANG);
22 sub compile_watchheaders ($) {
25 if (my $whs = $ibx->{watchheader}) {
27 my ($k, $v) = split(/:/, $_, 2);
28 # XXX should this be case-insensitive?
29 # Or, mutt-style, case-sensitive iff
30 # a capital letter exists?
31 push @$watch_hdrs, [ $k, qr/\Q$v\E/ ];
34 if (my $list_ids = $ibx->{listid}) {
36 # RFC2919 section 6 stipulates
37 # "case insensitive equality"
38 my $re = qr/<[ \t]*\Q$_\E[ \t]*>/i;
39 push @$watch_hdrs, ['List-Id', $re ];
42 $ibx->{-watchheaders} = $watch_hdrs if scalar @$watch_hdrs;
46 my ($class, $cfg) = @_;
48 my (%imap, %nntp); # url => [inbox objects] or 'watchspam'
51 # "publicinboxwatch" is the documented namespace
52 # "publicinboxlearn" is legacy but may be supported
54 foreach my $pfx (qw(publicinboxwatch publicinboxlearn)) {
55 my $k = "$pfx.watchspam";
56 defined(my $dirs = $cfg->{$k}) or next;
57 $dirs = PublicInbox::Config::_array($dirs);
58 for my $dir (@$dirs) {
60 if (is_maildir($dir)) {
61 # skip "new", no MUA has seen it, yet.
62 $mdmap{"$dir/cur"} = 'watchspam';
63 } elsif ($uri = imap_uri($dir)) {
64 $imap{$$uri} = 'watchspam';
66 } elsif ($uri = nntp_uri($dir)) {
67 $nntp{$$uri} = 'watchspam';
70 warn "unsupported $k=$dir\n";
75 my $k = 'publicinboxwatch.spamcheck';
77 my $spamcheck = PublicInbox::Spamcheck::get($cfg, $k, $default);
78 $spamcheck = _spamcheck_cb($spamcheck) if $spamcheck;
80 $cfg->each_inbox(sub {
81 # need to make all inboxes writable for spam removal:
82 my $ibx = $_[0] = PublicInbox::InboxWritable->new($_[0]);
84 my $watches = $ibx->{watch} or return;
85 $watches = PublicInbox::Config::_array($watches);
86 for my $watch (@$watches) {
88 if (is_maildir($watch)) {
89 compile_watchheaders($ibx);
90 my ($new, $cur) = ("$watch/new", "$watch/cur");
91 my $cur_dst = $mdmap{$cur} //= [];
92 return if is_watchspam($cur, $cur_dst, $ibx);
93 push @{$mdmap{$new} //= []}, $ibx;
95 } elsif ($uri = imap_uri($watch)) {
96 my $cur_dst = $imap{$$uri} //= [];
97 return if is_watchspam($uri, $cur_dst, $ibx);
98 compile_watchheaders($ibx);
99 push(@imap, $uri) if 1 == push(@$cur_dst, $ibx);
100 } elsif ($uri = nntp_uri($watch)) {
101 my $cur_dst = $nntp{$$uri} //= [];
102 return if is_watchspam($uri, $cur_dst, $ibx);
103 compile_watchheaders($ibx);
104 push(@nntp, $uri) if 1 == push(@$cur_dst, $ibx);
106 warn "watch unsupported: $k=$watch\n";
112 if (scalar keys %mdmap) {
113 $mdre = join('|', map { quotemeta($_) } keys %mdmap);
114 $mdre = qr!\A($mdre)/!;
116 return unless $mdre || scalar(keys %imap) || scalar(keys %nntp);
119 max_batch => 10, # avoid hogging locks for too long
120 spamcheck => $spamcheck,
124 imap => scalar keys %imap ? \%imap : undef,
125 nntp => scalar keys %nntp? \%nntp : undef,
126 imap_order => scalar(@imap) ? \@imap : undef,
127 nntp_order => scalar(@nntp) ? \@nntp: undef,
129 opendirs => {}, # dirname => dirhandle (in progress scans)
130 ops => [], # 'quit', 'full'
136 local $PublicInbox::DS::in_loop = 0; # waitpid() synchronously
137 for my $im (values %{$self->{importers}}) {
138 next if !$im; # $im may be undef during cleanup
140 warn "$im->{ibx}->{name} ->done: $@\n" if $@;
144 sub remove_eml_i { # each_inbox callback
145 my ($ibx, $self, $eml, $loc) = @_;
148 # try to avoid taking a lock or unnecessary spawning
149 my $im = $self->{importers}->{"$ibx"};
151 if ((!$im || !$im->active) && $ibx->over) {
152 if (content_exists($ibx, $eml)) {
154 } elsif (my $scrub = $ibx->filter($im)) {
155 $scrubbed = $scrub->scrub($eml, 1);
156 if ($scrubbed && $scrubbed != REJECT &&
157 !content_exists($ibx, $scrubbed)) {
165 $im //= _importer_for($self, $ibx); # may spawn fast-import
166 $im->remove($eml, 'spam');
168 my $scrub = $ibx->filter($im);
169 $scrub ? $scrub->scrub($eml, 1) : undef;
171 if ($scrubbed && $scrubbed != REJECT) {
172 $im->remove($scrubbed, 'spam');
176 warn "error removing spam at: $loc from $ibx->{name}: $@\n";
177 _done_for_now($self);
182 my ($self, $path) = @_;
183 # path must be marked as (S)een
184 $path =~ /:2,[A-R]*S[T-Za-z]*\z/ or return;
185 my $eml = eml_from_path($path) or return;
186 local $SIG{__WARN__} = PublicInbox::Eml::warn_ignore_cb();
187 $self->{pi_cfg}->each_inbox(\&remove_eml_i, $self, $eml, $path);
190 sub import_eml ($$$) {
191 my ($self, $ibx, $eml) = @_;
193 # any header match means it's eligible for the inbox:
194 if (my $watch_hdrs = $ibx->{-watchheaders}) {
196 for my $wh (@$watch_hdrs) {
197 my @v = $eml->header_raw($wh->[0]);
198 $ok = grep(/$wh->[1]/, @v) and last;
203 my $im = _importer_for($self, $ibx);
204 if (my $scrub = $ibx->filter($im)) {
205 my $scrubbed = $scrub->scrub($eml) or return;
206 $scrubbed == REJECT and return;
209 $im->add($eml, $self->{spamcheck});
212 warn "$ibx->{name} add failed: $@\n";
213 _done_for_now($self);
218 my ($self, $path) = @_;
219 my $fl = PublicInbox::MdirReader::maildir_path_flags($path) // return;
220 return if $fl =~ /[DT]/; # no Drafts or Trash
221 if ($path !~ $self->{mdre}) {
222 warn "unrecognized path: $path\n";
225 my $inboxes = $self->{mdmap}->{$1};
227 warn "unmappable dir: $1\n";
230 my $warn_cb = $SIG{__WARN__} || \&CORE::warn;
231 local $SIG{__WARN__} = sub {
232 my $pfx = ($_[0] // '') =~ /^([A-Z]: )/g ? $1 : '';
233 $warn_cb->($pfx, "path: $path\n", @_);
235 if (!ref($inboxes) && $inboxes eq 'watchspam') {
236 return _remove_spam($self, $path);
238 foreach my $ibx (@$inboxes) {
239 my $eml = eml_from_path($path) or next;
240 import_eml($self, $ibx, $eml);
246 return unless $self->{quit};
248 # don't have reliable wakeups, keep signalling
250 for (qw(idle_pids poll_pids)) {
251 my $pids = $self->{$_} or next;
253 $done = undef if kill('QUIT', $_);
262 %{$self->{opendirs}} = ();
263 _done_for_now($self);
265 if (my $idle_mic = $self->{idle_mic}) {
266 eval { $idle_mic->done };
268 warn "IDLE DONE error: $@\n";
269 eval { $idle_mic->disconnect };
270 warn "IDLE LOGOUT error: $@\n" if $@;
275 sub watch_fs_init ($) {
278 delete $self->{done_timer};
279 _done_for_now($self);
281 my $cb = sub { # called by PublicInbox::DirIdle::event_step
282 _try_path($self, $_[0]->fullname);
283 $self->{done_timer} //= PublicInbox::DS::requeue($done);
285 require PublicInbox::DirIdle;
286 # inotify_create + EPOLL_CTL_ADD
287 PublicInbox::DirIdle->new([keys %{$self->{mdmap}}], $cb);
290 sub imap_import_msg ($$$$$) {
291 my ($self, $uri, $uid, $raw, $flags) = @_;
292 # our target audience expects LF-only, save storage
293 $$raw =~ s/\r\n/\n/sg;
295 my $inboxes = $self->{imap}->{$$uri};
297 for my $ibx (@$inboxes) {
298 my $eml = PublicInbox::Eml->new($$raw);
299 import_eml($self, $ibx, $eml);
301 } elsif ($inboxes eq 'watchspam') {
302 return if $flags !~ /\\Seen\b/; # don't remove unseen messages
303 local $SIG{__WARN__} = PublicInbox::Eml::warn_ignore_cb();
304 my $eml = PublicInbox::Eml->new($raw);
305 $self->{pi_cfg}->each_inbox(\&remove_eml_i,
306 $self, $eml, "$uri UID:$uid");
308 die "BUG: destination unknown $inboxes";
312 sub net_cb { # NetReader::(nntp|imap)_each callback
313 my ($uri, $art, $kw, $eml, $self, $inboxes) = @_;
314 local $self->{cur_uid} = $art; # IMAP UID or NNTP article
319 my $tmp = PublicInbox::Eml->new(\($eml->as_string));
320 import_eml($self, $ibx, $tmp);
322 import_eml($self, $last, $eml);
323 } elsif ($inboxes eq 'watchspam') {
324 $self->{pi_cfg}->each_inbox(\&remove_eml_i,
325 $self, $eml, "$uri #$art");
327 die "BUG: destination unknown $inboxes";
331 sub imap_fetch_all ($$) {
332 my ($self, $uri) = @_;
333 my $warn_cb = $SIG{__WARN__} || \&CORE::warn;
334 $self->{incremental} = 1;
335 $self->{on_commit} = [ \&_done_for_now, $self ];
336 local $self->{cur_uid};
337 local $SIG{__WARN__} = sub {
338 my $pfx = ($_[0] // '') =~ /^([A-Z]: |# )/g ? $1 : '';
339 my $uid = $self->{cur_uid};
340 $warn_cb->("$pfx$uri", $uid ? ("UID:$uid") : (), "\n", @_);
342 PublicInbox::NetReader::imap_each($self, $uri, \&net_cb, $self,
343 $self->{imap}->{$$uri});
346 sub imap_idle_once ($$$$) {
347 my ($self, $mic, $intvl, $uri) = @_;
348 my $i = $intvl //= (29 * 60);
349 my $end = now() + $intvl;
350 warn "I: $uri idling for ${intvl}s\n";
351 local $0 = "IDLE $0";
352 unless ($mic->idle) {
353 return if $self->{quit};
354 return "E: IDLE failed on $uri: $!";
356 $self->{idle_mic} = $mic; # for ->quit
358 until ($self->{quit} || !$mic->IsConnected ||
359 grep(/^\* [0-9]+ EXISTS/, @res) || $i <= 0) {
360 @res = $mic->idle_data($i);
363 delete $self->{idle_mic};
364 unless ($self->{quit}) {
365 $mic->IsConnected or return "E: IDLE disconnected on $uri";
366 $mic->done or return "E: IDLE DONE failed on $uri: $!";
371 # idles on a single URI
372 sub watch_imap_idle_1 ($$$) {
373 my ($self, $uri, $intvl) = @_;
374 my $sec = uri_section($uri);
375 my $mic_arg = $self->{mic_arg}->{$sec} or
376 die "BUG: no Mail::IMAPClient->new arg for $sec";
378 local $0 = $uri->mailbox." $sec";
379 until ($self->{quit}) {
380 $mic //= PublicInbox::IMAPClient->new(%$mic_arg);
382 if ($mic && $mic->IsConnected) {
383 local $self->{mics_cached}->{$sec} = $mic;
384 my $m = imap_fetch_all($self, $uri);
385 $m == $mic or die "BUG: wrong mic";
386 $mic->IsConnected and
387 $err = imap_idle_once($self, $mic, $intvl, $uri)
389 $err = "E: not connected: $!";
391 if ($err && !$self->{quit}) {
394 sleep 60 unless $self->{quit};
399 sub watch_atfork_child ($) {
401 delete $self->{idle_pids};
402 delete $self->{poll_pids};
403 delete $self->{opendirs};
404 PublicInbox::DS->Reset;
405 %SIG = (%SIG, %{$self->{sig}}, CHLD => 'DEFAULT');
406 PublicInbox::DS::sig_setmask($self->{oldset});
409 sub watch_atfork_parent ($) {
411 _done_for_now($self);
412 PublicInbox::DS::block_signals();
415 sub imap_idle_requeue { # DS::add_timer callback
416 my ($self, $uri_intvl) = @_;
417 return if $self->{quit};
418 push @{$self->{idle_todo}}, $uri_intvl;
422 sub imap_idle_reap { # PublicInbox::DS::dwaitpid callback
423 my ($self, $pid) = @_;
424 my $uri_intvl = delete $self->{idle_pids}->{$pid} or
425 die "BUG: PID=$pid (unknown) reaped: \$?=$?\n";
427 my ($uri, $intvl) = @$uri_intvl;
428 return if $self->{quit};
429 warn "W: PID=$pid on $uri died: \$?=$?\n" if $?;
430 add_timer(60, \&imap_idle_requeue, $self, $uri_intvl);
433 sub reap { # callback for EOFpipe
434 my ($pid, $cb, $self) = @{$_[0]};
435 my $ret = waitpid($pid, 0);
437 $cb->($self, $pid); # poll_fetch_reap || imap_idle_reap
439 warn "W: waitpid($pid) => ", $ret // "($!)", "\n";
443 sub imap_idle_fork ($$) {
444 my ($self, $uri_intvl) = @_;
445 my ($uri, $intvl) = @$uri_intvl;
446 pipe(my ($r, $w)) or die "pipe: $!";
447 my $seed = rand(0xffffffff);
448 my $pid = fork // die "fork: $!";
451 eval { Net::SSLeay::randomize() };
453 watch_atfork_child($self);
454 watch_imap_idle_1($self, $uri, $intvl);
458 $self->{idle_pids}->{$pid} = $uri_intvl;
459 PublicInbox::EOFpipe->new($r, \&reap, [$pid, \&imap_idle_reap, $self]);
464 return if $self->{quit};
465 my $idle_todo = $self->{idle_todo};
466 if ($idle_todo && @$idle_todo) {
467 my $oldset = watch_atfork_parent($self);
469 while (my $uri_intvl = shift(@$idle_todo)) {
470 imap_idle_fork($self, $uri_intvl);
473 PublicInbox::DS::sig_setmask($oldset);
476 fs_scan_step($self) if $self->{mdre};
479 sub watch_imap_fetch_all ($$) {
480 my ($self, $uris) = @_;
481 for my $uri (@$uris) {
482 imap_fetch_all($self, $uri);
483 last if $self->{quit};
487 sub watch_nntp_fetch_all ($$) {
488 my ($self, $uris) = @_;
489 $self->{incremental} = 1;
490 $self->{on_commit} = [ \&_done_for_now, $self ];
491 my $warn_cb = $SIG{__WARN__} || \&CORE::warn;
492 local $self->{cur_uid};
494 local $SIG{__WARN__} = sub {
495 my $pfx = ($_[0] // '') =~ /^([A-Z]: |# )/g ? $1 : '';
496 my $art = $self->{cur_uid};
497 $warn_cb->("$pfx$uri", $art ? ("ARTICLE $art") : (), "\n", @_);
500 PublicInbox::NetReader::nntp_each($self, $uri, \&net_cb, $self,
501 $self->{nntp}->{$$uri});
502 last if $self->{quit};
506 sub poll_fetch_fork { # DS::add_timer callback
507 my ($self, $intvl, $uris) = @_;
508 return if $self->{quit};
509 pipe(my ($r, $w)) or die "pipe: $!";
510 my $oldset = watch_atfork_parent($self);
511 my $seed = rand(0xffffffff);
513 if (defined($pid) && $pid == 0) {
515 eval { Net::SSLeay::randomize() };
517 watch_atfork_child($self);
518 if ($uris->[0]->scheme =~ m!\Aimaps?!i) {
519 watch_imap_fetch_all($self, $uris);
521 watch_nntp_fetch_all($self, $uris);
526 PublicInbox::DS::sig_setmask($oldset);
527 die "fork: $!" unless defined $pid;
528 $self->{poll_pids}->{$pid} = [ $intvl, $uris ];
529 PublicInbox::EOFpipe->new($r, \&reap, [$pid, \&poll_fetch_reap, $self]);
532 sub poll_fetch_reap {
533 my ($self, $pid) = @_;
534 my $intvl_uris = delete $self->{poll_pids}->{$pid} or
535 die "BUG: PID=$pid (unknown) reaped: \$?=$?\n";
536 return if $self->{quit};
537 my ($intvl, $uris) = @$intvl_uris;
539 warn "W: PID=$pid died: \$?=$?\n", map { "$_\n" } @$uris;
541 warn("I: will check $_ in ${intvl}s\n") for @$uris;
542 add_timer($intvl, \&poll_fetch_fork, $self, $intvl, $uris);
545 sub watch_imap_init ($$) {
546 my ($self, $poll) = @_;
547 my $mics = PublicInbox::NetReader::imap_common_init($self);
548 my $idle = []; # [ [ uri1, intvl1 ], [uri2, intvl2] ]
549 for my $uri (@{$self->{imap_order}}) {
550 my $sec = uri_section($uri);
551 my $mic = $mics->{$sec};
552 my $intvl = $self->{imap_opt}->{$sec}->{pollInterval};
553 if ($mic->has_capability('IDLE') && !$intvl) {
554 $intvl = $self->{imap_opt}->{$sec}->{idleInterval};
555 push @$idle, [ $uri, $intvl // () ];
557 push @{$poll->{$intvl || 120}}, $uri;
561 $self->{idle_todo} = $idle;
562 PublicInbox::DS::requeue($self); # ->event_step to fork
566 sub watch_nntp_init ($$) {
567 my ($self, $poll) = @_;
568 PublicInbox::NetReader::nntp_common_init($self);
569 for my $uri (@{$self->{nntp_order}}) {
570 my $sec = uri_section($uri);
571 my $intvl = $self->{nntp_opt}->{$sec}->{pollInterval};
572 push @{$poll->{$intvl || 120}}, $uri;
576 sub watch { # main entry point
577 my ($self, $sig, $oldset) = @_;
578 $self->{oldset} = $oldset;
580 my $poll = {}; # intvl_seconds => [ uri1, uri2 ]
581 watch_imap_init($self, $poll) if $self->{imap};
582 watch_nntp_init($self, $poll) if $self->{nntp};
583 while (my ($intvl, $uris) = each %$poll) {
584 # poll all URIs for a given interval sequentially
585 add_timer(0, \&poll_fetch_fork, $self, $intvl, $uris);
587 watch_fs_init($self) if $self->{mdre};
588 PublicInbox::DS->SetPostLoopCallback(sub { !$self->quit_done });
589 PublicInbox::DS->EventLoop; # calls ->event_step
590 _done_for_now($self);
594 my ($self, $op) = @_;
595 push @{$self->{ops}}, $op;
596 PublicInbox::DS::requeue($self);
601 return if $self->{quit};
602 my $op = shift @{$self->{ops}};
603 local $PublicInbox::DS::in_loop = 0; # waitpid() synchronously
605 # continue existing scan
606 my $opendirs = $self->{opendirs};
607 my @dirnames = keys %$opendirs;
608 foreach my $dir (@dirnames) {
609 my $dh = delete $opendirs->{$dir};
610 my $n = $self->{max_batch};
611 while (my $fn = readdir($dh)) {
612 _try_path($self, "$dir/$fn");
615 $opendirs->{$dir} = $dh if $n < 0;
617 if ($op && $op eq 'full') {
618 foreach my $dir (keys %{$self->{mdmap}}) {
619 next if $opendirs->{$dir}; # already in progress
620 my $ok = opendir(my $dh, $dir);
622 warn "failed to open $dir: $!\n";
625 my $n = $self->{max_batch};
626 while (my $fn = readdir($dh)) {
627 _try_path($self, "$dir/$fn");
630 $opendirs->{$dir} = $dh if $n < 0;
633 _done_for_now($self);
634 # do we have more work to do?
635 PublicInbox::DS::requeue($self) if keys %$opendirs;
639 my ($self, $op) = @_;
640 push @{$self->{ops}}, $op;
645 my ($self, $ibx) = @_;
646 my $importers = $self->{importers};
647 my $im = $importers->{"$ibx"} ||= $ibx->importer(0);
648 if (scalar(keys(%$importers)) > 2) {
649 delete $importers->{"$ibx"};
650 _done_for_now($self);
653 $importers->{"$ibx"} = $im;
656 # XXX consider sharing with V2Writable, this only requires read-only access
657 sub content_exists ($$) {
658 my ($ibx, $eml) = @_;
659 my $over = $ibx->over or return;
660 my $mids = mids($eml);
661 my $chash = content_hash($eml);
663 for my $mid (@$mids) {
664 while (my $smsg = $over->next_by_mid($mid, \$id, \$prev)) {
665 my $cmp = $ibx->smsg_eml($smsg) or return;
666 return 1 if $chash eq content_hash($cmp);
674 sub { # this gets called by (V2Writable||Import)->add
675 my ($mime, $ibx) = @_;
676 return if content_exists($ibx, $mime);
678 if ($sc->spamcheck($mime, \$tmp)) {
679 return PublicInbox::Eml->new(\$tmp);
681 warn $mime->header('Message-ID')." failed spam check\n";
687 $_[0] =~ s!\Amaildir:!! or return;
694 my ($cur, $ws, $ibx) = @_;
695 if ($ws && !ref($ws) && $ws eq 'watchspam') {
697 E: $cur is a spam folder and cannot be used for `$ibx->{name}' input