1 # Copyright (C) 2016-2020 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 # http://wiki2.dovecot.org/MailboxFormat/Maildir
6 package PublicInbox::WatchMaildir;
10 use PublicInbox::InboxWritable;
11 use PublicInbox::Filter::Base qw(REJECT);
12 use PublicInbox::Spamcheck;
13 use PublicInbox::Sigfd;
14 use PublicInbox::DS qw(now);
15 use POSIX qw(_exit WNOHANG);
16 *mime_from_path = \&PublicInbox::InboxWritable::mime_from_path;
18 sub compile_watchheaders ($) {
21 if (my $whs = $ibx->{watchheader}) {
23 my ($k, $v) = split(/:/, $_, 2);
24 # XXX should this be case-insensitive?
25 # Or, mutt-style, case-sensitive iff
26 # a capital letter exists?
27 push @$watch_hdrs, [ $k, qr/\Q$v\E/ ];
30 if (my $list_ids = $ibx->{listid}) {
32 # RFC2919 section 6 stipulates
33 # "case insensitive equality"
34 my $re = qr/<[ \t]*\Q$_\E[ \t]*>/i;
35 push @$watch_hdrs, ['List-Id', $re ];
38 $ibx->{-watchheaders} = $watch_hdrs if scalar @$watch_hdrs;
42 my ($class, $config) = @_;
43 my (%mdmap, @mdir, $spamc);
44 my %uniq; # directory => count
45 my %imap; # url => [inbox objects] or 'watchspam'
47 # "publicinboxwatch" is the documented namespace
48 # "publicinboxlearn" is legacy but may be supported
50 foreach my $pfx (qw(publicinboxwatch publicinboxlearn)) {
51 my $k = "$pfx.watchspam";
52 defined(my $dirs = $config->{$k}) or next;
53 $dirs = [ $dirs ] if !ref($dirs);
54 for my $dir (@$dirs) {
55 if (is_maildir($dir)) {
56 # skip "new", no MUA has seen it, yet.
60 $mdmap{$cur} = 'watchspam';
61 } elsif (my $url = imap_url($dir)) {
62 $imap{$url} = 'watchspam';
64 warn "unsupported $k=$dir\n";
69 my $k = 'publicinboxwatch.spamcheck';
71 my $spamcheck = PublicInbox::Spamcheck::get($config, $k, $default);
72 $spamcheck = _spamcheck_cb($spamcheck) if $spamcheck;
74 $config->each_inbox(sub {
75 # need to make all inboxes writable for spam removal:
76 my $ibx = $_[0] = PublicInbox::InboxWritable->new($_[0]);
78 my $watch = $ibx->{watch} or return;
79 if (is_maildir($watch)) {
80 compile_watchheaders($ibx);
81 my ($new, $cur) = ("$watch/new", "$watch/cur");
82 return if is_watchspam($cur, $mdmap{$cur}, $ibx);
83 push @mdir, $new unless $uniq{$new}++;
84 push @mdir, $cur unless $uniq{$cur}++;
85 push @{$mdmap{$new} ||= []}, $ibx;
86 push @{$mdmap{$cur} ||= []}, $ibx;
87 } elsif (my $url = imap_url($watch)) {
88 return if is_watchspam($url, $imap{$url}, $ibx);
89 compile_watchheaders($ibx);
90 push @{$imap{$url} ||= []}, $ibx;
92 warn "watch unsupported: $k=$watch\n";
95 return unless scalar(@mdir) || scalar(keys %imap);
99 $mdre = join('|', map { quotemeta($_) } @mdir);
100 $mdre = qr!\A($mdre)/!;
103 spamcheck => $spamcheck,
108 imap => scalar keys %imap ? \%imap : undef,
110 opendirs => {}, # dirname => dirhandle (in progress scans)
111 ops => [], # 'quit', 'full'
117 my $importers = $self->{importers};
118 foreach my $im (values %$importers) {
123 sub remove_eml_i { # each_inbox callback
124 my ($ibx, $arg) = @_;
125 my ($self, $eml, $loc) = @$arg;
127 my $im = _importer_for($self, $ibx);
128 $im->remove($eml, 'spam');
129 if (my $scrub = $ibx->filter($im)) {
130 my $scrubbed = $scrub->scrub($eml, 1);
132 $scrubbed == REJECT() and return;
133 $im->remove($scrubbed, 'spam');
136 warn "error removing spam at: $loc from $ibx->{name}: $@\n" if $@;
140 my ($self, $path) = @_;
141 # path must be marked as (S)een
142 $path =~ /:2,[A-R]*S[T-Za-z]*\z/ or return;
143 my $eml = mime_from_path($path) or return;
144 $self->{config}->each_inbox(\&remove_eml_i, [ $self, $eml, $path ]);
147 sub import_eml ($$$) {
148 my ($self, $ibx, $eml) = @_;
149 my $im = _importer_for($self, $ibx);
151 # any header match means it's eligible for the inbox:
152 if (my $watch_hdrs = $ibx->{-watchheaders}) {
154 my $hdr = $eml->header_obj;
155 for my $wh (@$watch_hdrs) {
156 my @v = $hdr->header_raw($wh->[0]);
157 $ok = grep(/$wh->[1]/, @v) and last;
162 if (my $scrub = $ibx->filter($im)) {
163 my $ret = $scrub->scrub($eml) or return;
164 $ret == REJECT() and return;
167 $im->add($eml, $self->{spamcheck});
171 my ($self, $path) = @_;
172 return unless PublicInbox::InboxWritable::is_maildir_path($path);
173 if ($path !~ $self->{mdre}) {
174 warn "unrecognized path: $path\n";
177 my $inboxes = $self->{mdmap}->{$1};
179 warn "unmappable dir: $1\n";
182 if (!ref($inboxes) && $inboxes eq 'watchspam') {
183 return _remove_spam($self, $path);
186 my $warn_cb = $SIG{__WARN__} || sub { print STDERR @_ };
187 local $SIG{__WARN__} = sub {
188 $warn_cb->("path: $path\n");
191 foreach my $ibx (@$inboxes) {
192 my $eml = mime_from_path($path) or next;
193 import_eml($self, $ibx, $eml);
200 %{$self->{opendirs}} = ();
201 _done_for_now($self);
202 if (my $imap_pid = $self->{-imap_pid}) {
203 kill('QUIT', $imap_pid);
205 if (my $idle_pids = $self->{idle_pids}) {
206 kill('QUIT', $_) for (keys %$idle_pids);
208 if (my $idle_mic = $self->{idle_mic}) {
209 eval { $idle_mic->done };
210 warn "IDLE DONE error: $@\n" if $@;
211 eval { $idle_mic->disconnect };
212 warn "IDLE LOGOUT error: $@\n" if $@;
218 require PublicInbox::DirIdle;
220 delete $self->{done_timer};
221 _done_for_now($self);
224 _try_path($self, $_[0]->fullname);
225 $self->{done_timer} //= PublicInbox::DS::requeue($done);
227 my $di = PublicInbox::DirIdle->new($self->{mdir}, $cb);
228 PublicInbox::DS->SetPostLoopCallback(sub { !$self->{quit} });
229 PublicInbox::DS->EventLoop;
230 _done_for_now($self);
233 # returns the git config section name, e.g [imap "imaps://user@example.com"]
234 # without the mailbox, so we can share connections between different inboxes
235 sub imap_section ($) {
237 $uri->scheme . '://' . $uri->authority;
241 my ($cfg, $key) = @_;
242 defined(my $v = $cfg->{lc($key)}) or return;
243 $v =~ /\A[0-9]+\z/s and return $v + 0;
244 if (ref($v) eq 'ARRAY') {
245 $v = join(', ', @$v);
246 warn "W: $key has multiple values: $v\nW: $key ignored\n";
248 warn "W: $key=$v is not an integer value in seconds\n";
252 # flesh out common IMAP-specific data structures
253 sub imap_common_init ($) {
255 my $cfg = $self->{config};
256 my $mic_args = {}; # scheme://authority => Mail:IMAPClient arg
257 for my $url (sort keys %{$self->{imap}}) {
258 my $uri = PublicInbox::URIimap->new($url);
259 my $sec = imap_section($uri);
260 for my $k (qw(Starttls Debug Compress)) {
261 my $key = lc("imap.$sec.$k");
262 defined(my $orig = $cfg->{$key}) or next;
263 my $v = PublicInbox::Config::_git_config_bool($orig);
265 $mic_args->{$sec}->{$k} = $v;
267 warn "W: $key=$orig is not boolean\n";
270 my $to = cfg_intvl($cfg, "imap.$sec.Timeout");
271 $mic_args->{$sec}->{Timeout} = $to if $to;
272 $to = cfg_intvl($cfg, "imap.$sec.PollInterval");
273 $self->{imap_opt}->{$sec}->{poll_intvl} = $to if $to;
274 $to = cfg_intvl($cfg, "imap.$sec.IdleInterval");
275 $self->{imap_opt}->{$sec}->{idle_intvl} = $to if $to;
280 sub auth_anon_cb { '' }; # for Mail::IMAPClient::Authcallback
282 sub mic_for ($$$) { # mic = Mail::IMAPClient
283 my ($self, $uri, $mic_args) = @_;
284 my $url = $uri->as_string;
287 protocol => $uri->scheme,
289 username => $uri->user,
290 password => $uri->password,
292 my $common = $mic_args->{imap_section($uri)} // {};
293 my $host = $cred->{host};
296 # IMAPClient mishandles `0', so we pass `127.0.0.1'
297 Server => $host eq '0' ? '127.0.0.1' : $host,
298 Ssl => $uri->scheme eq 'imaps',
299 Keepalive => 1, # SO_KEEPALIVE
300 %$common, # may set Starttls, Compress, Debug ....
302 my $mic = PublicInbox::IMAPClient->new(%$mic_arg) or
303 die "E: <$url> new: $@\n";
305 # default to using STARTTLS if it's available, but allow
306 # it to be disabled since I usually connect to localhost
307 if (!$mic_arg->{Ssl} && !defined($mic_arg->{Starttls}) &&
308 $mic->has_capability('STARTTLS') &&
309 $mic->can('starttls')) {
310 $mic->starttls or die "E: <$url> STARTTLS: $@\n";
313 # do we even need credentials?
314 if (!defined($cred->{username}) &&
315 $mic->has_capability('AUTH=ANONYMOUS')) {
319 Git::credential($cred, 'fill'); # may prompt user here
320 $mic->User($mic_arg->{User} = $cred->{username});
321 $mic->Password($mic_arg->{Password} = $cred->{password});
322 } else { # AUTH=ANONYMOUS
323 $mic->Authmechanism($mic_arg->{Authmechanism} = 'ANONYMOUS');
324 $mic->Authcallback($mic_arg->{Authcallback} = \&auth_anon_cb);
326 if ($mic->login && $mic->IsAuthenticated) {
327 # success! keep IMAPClient->new arg in case we get disconnected
328 $self->{mic_arg}->{imap_section($uri)} = $mic_arg;
330 warn "E: <$url> LOGIN: $@\n";
333 Git::credential($cred, $mic ? 'approve' : 'reject') if $cred;
339 eval { require PublicInbox::IMAPClient } or
340 die "Mail::IMAPClient is required for IMAP:\n$@\n";
341 eval { require Git } or
342 die "Git (Perl module) is required for IMAP:\n$@\n";
343 eval { require PublicInbox::IMAPTracker } or
344 die "DBD::SQLite is required for IMAP\n:$@\n";
346 my $mic_args = imap_common_init($self);
347 # make sure we can connect and cache the credentials in memory
348 $self->{mic_arg} = {}; # schema://authority => IMAPClient->new args
349 my $mics = $self->{mics} = {}; # schema://authority => IMAPClient obj
350 for my $url (sort keys %{$self->{imap}}) {
351 my $uri = PublicInbox::URIimap->new($url);
352 $mics->{imap_section($uri)} //= mic_for($self, $uri, $mic_args);
356 sub imap_fetch_all ($$$) {
357 my ($self, $mic, $uri) = @_;
358 my $sec = imap_section($uri);
359 my $mbx = $uri->mailbox;
360 my $url = $uri->as_string;
361 $mic->Clear(1); # trim results history
362 $mic->examine($mbx) or return "E: EXAMINE $mbx ($sec) failed: $!";
363 my ($r_uidval, $r_uidnext);
364 for ($mic->Results) {
365 /^\* OK \[UIDVALIDITY ([0-9]+)\].*/ and $r_uidval = $1;
366 /^\* OK \[UIDNEXT ([0-9]+)\].*/ and $r_uidnext = $1;
367 last if $r_uidval && $r_uidnext;
369 $r_uidval //= $mic->uidvalidity($mbx) //
370 return "E: $url cannot get UIDVALIDITY";
371 $r_uidnext //= $mic->uidnext($mbx) //
372 return "E: $url cannot get UIDNEXT";
373 my $itrk = PublicInbox::IMAPTracker->new;
374 my ($l_uidval, $l_uid) = $itrk->get_last($url);
375 $l_uidval //= $r_uidval; # first time
377 if ($l_uidval != $r_uidval) {
378 return "E: $url UIDVALIDITY mismatch\n".
379 "E: local=$l_uidval != remote=$r_uidval";
381 my $r_uid = $r_uidnext - 1;
382 if ($l_uid != 1 && $l_uid > $r_uid) {
383 return "E: $url local UID exceeds remote ($l_uid > $r_uid)\n".
384 "E: $url strangely, UIDVALIDLITY matches ($l_uidval)\n";
386 return if $l_uid >= $r_uid; # nothing to do
388 $mic->Uid(1); # the default, we hope
389 my $req = $mic->imap4rev1 ? 'BODY.PEEK[]' : 'RFC822.PEEK';
392 my $inboxes = $self->{imap}->{$url};
393 warn "I: $url fetching $l_uid..$r_uid\n";
395 my $warn_cb = $SIG{__WARN__} || sub { print STDERR @_ };
396 local $SIG{__WARN__} = sub {
397 $warn_cb->("$url UID:$uid\n");
401 $itrk->{dbh}->begin_work;
402 for my $u ($l_uid..$r_uid) {
404 local $0 = "UID:$uid $mbx $sec";
405 my $r = $mic->fetch_hash($uid, $req);
406 unless ($r) { # network error?
407 $err = "E: $url UID FETCH $uid error: $!\n";
411 # messages get deleted, so holes appear
412 defined(my $raw = delete $r->{$uid}->{$key}) or next;
414 # our target audience expects LF-only, save storage
415 $raw =~ s/\r\n/\n/sg;
418 for my $ibx (@$inboxes) {
419 my $eml = PublicInbox::Eml->new($raw);
420 my $x = import_eml($self, $ibx, $eml);
422 } elsif ($inboxes eq 'watchspam') {
423 my $eml = PublicInbox::Eml->new($raw);
424 my $arg = [ $self, $eml, "$uri UID:$uid" ];
425 $self->{config}->each_inbox(\&remove_eml_i, $arg);
427 die "BUG: destination unknown $inboxes";
429 $itrk->update_last($url, $r_uidval, $uid);
430 last if $self->{quit};
432 _done_for_now($self);
433 $itrk->{dbh}->commit;
437 sub imap_idle_once ($$$$) {
438 my ($self, $mic, $intvl, $url) = @_;
439 my $i = $intvl //= (29 * 60);
440 my $end = now() + $intvl;
441 warn "I: $url idling for ${intvl}s\n";
442 local $0 = "IDLE $0";
443 unless ($mic->idle) {
444 return if $self->{quit};
445 return "E: IDLE failed on $url: $!";
447 $self->{idle_mic} = $mic; # for ->quit
449 until ($self->{quit} || grep(/^\* [0-9]+ EXISTS/, @res) || $i <= 0) {
450 @res = $mic->idle_data($i);
453 delete $self->{idle_mic};
454 unless ($self->{quit}) {
455 $mic->IsConnected or return "E: IDLE disconnected on $url";
456 $mic->done or return "E: IDLE DONE failed on $url: $!";
461 # idles on a single URI
462 sub watch_imap_idle_1 ($$$) {
463 my ($self, $uri, $intvl) = @_;
464 my $sec = imap_section($uri);
465 my $mic_arg = $self->{mic_arg}->{$sec} or
466 die "BUG: no Mail::IMAPClient->new arg for $sec";
468 local $0 = $uri->mailbox." $sec";
469 until ($self->{quit}) {
470 $mic //= delete($self->{mics}->{$sec}) //
471 PublicInbox::IMAPClient->new(%$mic_arg);
472 my $err = imap_fetch_all($self, $mic, $uri);
473 $err //= imap_idle_once($self, $mic, $intvl, $uri->as_string);
474 if ($err && !$self->{quit}) {
477 sleep 60 unless $self->{quit};
482 sub watch_atfork_child ($) {
484 PublicInbox::Sigfd::sig_setmask($self->{oldset});
485 %SIG = (%SIG, %{$self->{sig}});
488 sub watch_imap_idle_all ($$) {
489 my ($self, $idle) = @_; # $idle = [[ uri1, intvl1 ], [ uri2, intvl2 ]]
490 $self->{mics} = {}; # going to be forking, so disconnect
491 my $idle_pids = $self->{idle_pids} = {};
492 until ($self->{quit}) {
493 while (my $uri_intvl = shift @$idle) {
494 my ($uri, $intvl) = @$uri_intvl;
495 defined(my $pid = fork) or die "fork: $!";
497 watch_atfork_child($self);
498 delete $self->{idle_pids};
499 watch_imap_idle_1($self, $uri, $intvl);
502 $idle_pids->{$pid} = $uri_intvl;
504 my $pid = waitpid(-1, 0) or next;
506 warn "W: no idling children: $!";
510 warn "W: nothing to respawn, quitting IDLE\n";
514 if (my $uri_intvl = delete $idle_pids->{$pid}) {
515 my ($uri, $intvl) = @$uri_intvl;
516 my $url = $uri->as_string;
517 if ($? || !$self->{quit}) {
518 warn "W: PID=$pid on $url died: \$?=$?\n";
520 push @$idle, $uri_intvl;
522 warn "W: PID=$pid (unknown) reaped: \$?=$?\n";
527 kill('QUIT', $_) for (keys %$idle_pids);
528 while (scalar keys %$idle_pids) {
529 if (my $pid = waitpid(-1, WNOHANG)) {
531 warn "E: no children? $! (PIDs: ",
532 join(', ', keys %$idle_pids),")\n";
535 delete $idle_pids->{$pid};
537 } else { # signals aren't that reliable w/o signalfd/kevent
539 kill('QUIT', $_) for (keys %$idle_pids);
546 my $idle = []; # [ [ uri1, intvl1 ], [uri2, intvl2] ];
547 my $poll = {}; # intvl_seconds => [ uri1, uri2 ]
548 for my $url (keys %{$self->{imap}}) {
549 my $uri = PublicInbox::URIimap->new($url);
550 my $sec = imap_section($uri);
551 my $mic = $self->{mics}->{$sec};
552 my $intvl = $self->{imap_opt}->{$sec}->{poll_intvl};
553 if ($mic->has_capability('IDLE') && !$intvl) {
554 $intvl = $self->{imap_opt}->{$sec}->{idle_intvl};
555 push @$idle, [ $uri, $intvl // () ];
557 push @{$poll->{$intvl || 120}}, $uri;
560 my $nr_poll = scalar keys %$poll;
561 if (scalar @$idle && !$nr_poll) { # multiple idlers, need fork
562 watch_imap_idle_all($self, $idle);
568 my ($self, $sig, $oldset) = @_;
569 $self->{oldset} = $oldset;
571 if ($self->{mdre} && $self->{imap}) {
572 defined(my $pid = fork) or die "fork: $!";
574 watch_atfork_child($self);
578 $self->{-imap_pid} = $pid;
579 } elsif ($self->{imap}) {
580 # not a child process, but no signalfd, yet:
581 watch_atfork_child($self);
589 my ($self, $op) = @_;
590 push @{$self->{ops}}, $op;
591 PublicInbox::DS::requeue($self);
594 # called directly, and by PublicInbox::DS
597 return if $self->{quit};
598 my $op = shift @{$self->{ops}};
600 # continue existing scan
602 my $opendirs = $self->{opendirs};
603 my @dirnames = keys %$opendirs;
604 foreach my $dir (@dirnames) {
605 my $dh = delete $opendirs->{$dir};
607 while (my $fn = readdir($dh)) {
608 _try_path($self, "$dir/$fn");
611 $opendirs->{$dir} = $dh if $n < 0;
613 if ($op && $op eq 'full') {
614 foreach my $dir (@{$self->{mdir}}) {
615 next if $opendirs->{$dir}; # already in progress
616 my $ok = opendir(my $dh, $dir);
618 warn "failed to open $dir: $!\n";
622 while (my $fn = readdir($dh)) {
623 _try_path($self, "$dir/$fn");
626 $opendirs->{$dir} = $dh if $n < 0;
629 _done_for_now($self);
630 # do we have more work to do?
631 PublicInbox::DS::requeue($self) if keys %$opendirs;
635 my ($self, $op) = @_;
636 push @{$self->{ops}}, $op;
641 my ($self, $ibx) = @_;
642 my $importers = $self->{importers};
643 my $im = $importers->{"$ibx"} ||= $ibx->importer(0);
644 if (scalar(keys(%$importers)) > 2) {
645 delete $importers->{"$ibx"};
646 _done_for_now($self);
649 $importers->{"$ibx"} = $im;
657 if ($sc->spamcheck($mime, \$tmp)) {
658 return PublicInbox::Eml->new(\$tmp);
660 warn $mime->header('Message-ID')." failed spam check\n";
666 $_[0] =~ s!\Amaildir:!! or return;
673 my ($cur, $ws, $ibx) = @_;
674 if ($ws && !ref($ws) && $ws eq 'watchspam') {
676 E: $cur is a spam folder and cannot be used for `$ibx->{name}' input
685 require PublicInbox::URIimap;
686 my $uri = PublicInbox::URIimap->new($url);
687 $uri ? $uri->canonical->as_string : undef;