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);
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 for (qw(idle_pids poll_pids)) {
206 my $pids = $self->{$_} or next;
207 kill('QUIT', $_) for (keys %$pids);
209 if (my $idle_mic = $self->{idle_mic}) {
210 eval { $idle_mic->done };
211 warn "IDLE DONE error: $@\n" if $@;
212 eval { $idle_mic->disconnect };
213 warn "IDLE LOGOUT error: $@\n" if $@;
217 sub watch_fs_init ($) {
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 require PublicInbox::DirIdle;
228 PublicInbox::DirIdle->new($self->{mdir}, $cb); # EPOLL_CTL_ADD
231 # returns the git config section name, e.g [imap "imaps://user@example.com"]
232 # without the mailbox, so we can share connections between different inboxes
233 sub imap_section ($) {
235 $uri->scheme . '://' . $uri->authority;
238 sub cfg_intvl ($$$$$) {
239 my ($cfg, $cfg_section, $cfg_key, $imap_section, $url) = @_;
240 my $key = "$cfg_section.$imap_section.$cfg_key";
241 my $v = $cfg->{lc($key)} //
242 $cfg->urlmatch("$cfg_section.$cfg_key", $url) // return;
243 $v =~ /\A[0-9]+(?:\.[0-9]+)?\z/s and return $v + 0;
244 if (ref($v) eq 'ARRAY') {
245 $v = join(', ', @$v);
246 warn "W: $key has multiple values: $v\nW: $key ignored\n";
248 warn "W: $key=$v is not a numeric value in seconds\n";
252 # flesh out common IMAP-specific data structures
253 sub imap_common_init ($) {
255 my $cfg = $self->{config};
256 my $mic_args = {}; # scheme://authority => Mail:IMAPClient arg
257 for my $url (sort keys %{$self->{imap}}) {
258 my $uri = PublicInbox::URIimap->new($url);
259 my $sec = imap_section($uri);
260 for my $k (qw(Starttls Debug Compress)) {
261 my $key = lc("imap.$sec.$k");
262 my $orig = $cfg->{$key} //
263 $cfg->urlmatch("imap.$k", $url) // next;
264 my $v = PublicInbox::Config::_git_config_bool($orig);
266 $mic_args->{$sec}->{$k} = $v;
268 warn "W: $key=$orig is not boolean\n";
271 my $to = cfg_intvl($cfg, 'imap', 'Timeout', $sec, $url);
272 $mic_args->{$sec}->{Timeout} = $to if $to;
273 $to = cfg_intvl($cfg, 'imap', 'PollInterval', $sec, $url);
274 $self->{imap_opt}->{$sec}->{poll_intvl} = $to if $to;
275 $to = cfg_intvl($cfg, 'imap', 'IdleInterval', $sec, $url);
276 $self->{imap_opt}->{$sec}->{idle_intvl} = $to if $to;
281 sub auth_anon_cb { '' }; # for Mail::IMAPClient::Authcallback
283 sub mic_for ($$$) { # mic = Mail::IMAPClient
284 my ($self, $uri, $mic_args) = @_;
285 my $url = $uri->as_string;
288 protocol => $uri->scheme,
290 username => $uri->user,
291 password => $uri->password,
293 my $common = $mic_args->{imap_section($uri)} // {};
294 my $host = $cred->{host};
297 # IMAPClient mishandles `0', so we pass `127.0.0.1'
298 Server => $host eq '0' ? '127.0.0.1' : $host,
299 Ssl => $uri->scheme eq 'imaps',
300 Keepalive => 1, # SO_KEEPALIVE
301 %$common, # may set Starttls, Compress, Debug ....
303 my $mic = PublicInbox::IMAPClient->new(%$mic_arg) or
304 die "E: <$url> new: $@\n";
306 # default to using STARTTLS if it's available, but allow
307 # it to be disabled since I usually connect to localhost
308 if (!$mic_arg->{Ssl} && !defined($mic_arg->{Starttls}) &&
309 $mic->has_capability('STARTTLS') &&
310 $mic->can('starttls')) {
311 $mic->starttls or die "E: <$url> STARTTLS: $@\n";
314 # do we even need credentials?
315 if (!defined($cred->{username}) &&
316 $mic->has_capability('AUTH=ANONYMOUS')) {
320 Git::credential($cred, 'fill'); # may prompt user here
321 $mic->User($mic_arg->{User} = $cred->{username});
322 $mic->Password($mic_arg->{Password} = $cred->{password});
323 } else { # AUTH=ANONYMOUS
324 $mic->Authmechanism($mic_arg->{Authmechanism} = 'ANONYMOUS');
325 $mic->Authcallback($mic_arg->{Authcallback} = \&auth_anon_cb);
327 if ($mic->login && $mic->IsAuthenticated) {
328 # success! keep IMAPClient->new arg in case we get disconnected
329 $self->{mic_arg}->{imap_section($uri)} = $mic_arg;
331 warn "E: <$url> LOGIN: $@\n";
334 Git::credential($cred, $mic ? 'approve' : 'reject') if $cred;
338 sub imap_fetch_all ($$$) {
339 my ($self, $mic, $uri) = @_;
340 my $sec = imap_section($uri);
341 my $mbx = $uri->mailbox;
342 my $url = $uri->as_string;
343 $mic->Clear(1); # trim results history
344 $mic->examine($mbx) or return "E: EXAMINE $mbx ($sec) failed: $!";
345 my ($r_uidval, $r_uidnext);
346 for ($mic->Results) {
347 /^\* OK \[UIDVALIDITY ([0-9]+)\].*/ and $r_uidval = $1;
348 /^\* OK \[UIDNEXT ([0-9]+)\].*/ and $r_uidnext = $1;
349 last if $r_uidval && $r_uidnext;
351 $r_uidval //= $mic->uidvalidity($mbx) //
352 return "E: $url cannot get UIDVALIDITY";
353 $r_uidnext //= $mic->uidnext($mbx) //
354 return "E: $url cannot get UIDNEXT";
355 my $itrk = PublicInbox::IMAPTracker->new;
356 my ($l_uidval, $l_uid) = $itrk->get_last($url);
357 $l_uidval //= $r_uidval; # first time
359 if ($l_uidval != $r_uidval) {
360 return "E: $url UIDVALIDITY mismatch\n".
361 "E: local=$l_uidval != remote=$r_uidval";
363 my $r_uid = $r_uidnext - 1;
364 if ($l_uid != 1 && $l_uid > $r_uid) {
365 return "E: $url local UID exceeds remote ($l_uid > $r_uid)\n".
366 "E: $url strangely, UIDVALIDLITY matches ($l_uidval)\n";
368 return if $l_uid >= $r_uid; # nothing to do
370 $mic->Uid(1); # the default, we hope
371 my $req = $mic->imap4rev1 ? 'BODY.PEEK[]' : 'RFC822.PEEK';
374 my $inboxes = $self->{imap}->{$url};
375 warn "I: $url fetching $l_uid..$r_uid\n";
377 my $warn_cb = $SIG{__WARN__} || sub { print STDERR @_ };
378 local $SIG{__WARN__} = sub {
379 $warn_cb->("$url UID:$uid\n");
383 $itrk->{dbh}->begin_work;
384 for my $u ($l_uid..$r_uid) {
386 local $0 = "UID:$uid $mbx $sec";
387 my $r = $mic->fetch_hash($uid, $req);
388 unless ($r) { # network error?
389 $err = "E: $url UID FETCH $uid error: $!\n";
393 # messages get deleted, so holes appear
394 defined(my $raw = delete $r->{$uid}->{$key}) or next;
396 # our target audience expects LF-only, save storage
397 $raw =~ s/\r\n/\n/sg;
400 for my $ibx (@$inboxes) {
401 my $eml = PublicInbox::Eml->new($raw);
402 my $x = import_eml($self, $ibx, $eml);
404 } elsif ($inboxes eq 'watchspam') {
405 my $eml = PublicInbox::Eml->new($raw);
406 my $arg = [ $self, $eml, "$uri UID:$uid" ];
407 $self->{config}->each_inbox(\&remove_eml_i, $arg);
409 die "BUG: destination unknown $inboxes";
411 $itrk->update_last($url, $r_uidval, $uid);
412 last if $self->{quit};
414 _done_for_now($self);
415 $itrk->{dbh}->commit;
419 sub imap_idle_once ($$$$) {
420 my ($self, $mic, $intvl, $url) = @_;
421 my $i = $intvl //= (29 * 60);
422 my $end = now() + $intvl;
423 warn "I: $url idling for ${intvl}s\n";
424 local $0 = "IDLE $0";
425 unless ($mic->idle) {
426 return if $self->{quit};
427 return "E: IDLE failed on $url: $!";
429 $self->{idle_mic} = $mic; # for ->quit
431 until ($self->{quit} || grep(/^\* [0-9]+ EXISTS/, @res) || $i <= 0) {
432 @res = $mic->idle_data($i);
435 delete $self->{idle_mic};
436 unless ($self->{quit}) {
437 $mic->IsConnected or return "E: IDLE disconnected on $url";
438 $mic->done or return "E: IDLE DONE failed on $url: $!";
443 # idles on a single URI
444 sub watch_imap_idle_1 ($$$) {
445 my ($self, $uri, $intvl) = @_;
446 my $sec = imap_section($uri);
447 my $mic_arg = $self->{mic_arg}->{$sec} or
448 die "BUG: no Mail::IMAPClient->new arg for $sec";
450 local $0 = $uri->mailbox." $sec";
451 until ($self->{quit}) {
452 $mic //= delete($self->{mics}->{$sec}) //
453 PublicInbox::IMAPClient->new(%$mic_arg);
454 my $err = imap_fetch_all($self, $mic, $uri);
455 $err //= imap_idle_once($self, $mic, $intvl, $uri->as_string);
456 if ($err && !$self->{quit}) {
459 sleep 60 unless $self->{quit};
464 sub watch_atfork_child ($) {
466 delete $self->{idle_pids};
467 delete $self->{poll_pids};
468 delete $self->{opendirs};
469 PublicInbox::DS->Reset;
470 PublicInbox::Sigfd::sig_setmask($self->{oldset});
471 %SIG = (%SIG, %{$self->{sig}});
474 sub watch_atfork_parent ($) {
476 _done_for_now($self);
477 $self->{mics} = {}; # going to be forking, so disconnect
480 sub imap_idle_reap { # PublicInbox::DS::dwaitpid callback
481 my ($self, $pid) = @_;
482 my $uri_intvl = delete $self->{idle_pids}->{$pid} or
483 die "BUG: PID=$pid (unknown) reaped: \$?=$?\n";
485 my ($uri, $intvl) = @$uri_intvl;
486 my $url = $uri->as_string;
487 return if $self->{quit};
488 warn "W: PID=$pid on $url died: \$?=$?\n" if $?;
489 push @{$self->{idle_todo}}, $uri_intvl;
490 PubicInbox::DS::requeue($self); # call ->event_step to respawn
493 sub imap_idle_fork ($$) {
494 my ($self, $uri_intvl) = @_;
495 my ($uri, $intvl) = @$uri_intvl;
496 defined(my $pid = fork) or die "fork: $!";
498 watch_atfork_child($self);
499 watch_imap_idle_1($self, $uri, $intvl);
502 $self->{idle_pids}->{$pid} = $uri_intvl;
503 PublicInbox::DS::dwaitpid($pid, \&imap_idle_reap, $self);
508 return if $self->{quit};
509 my $idle_todo = $self->{idle_todo};
510 if ($idle_todo && @$idle_todo) {
511 watch_atfork_parent($self);
512 while (my $uri_intvl = shift(@$idle_todo)) {
513 imap_idle_fork($self, $uri_intvl);
516 goto(&fs_scan_step) if $self->{mdre};
519 sub watch_imap_fetch_all ($$) {
520 my ($self, $uris) = @_;
521 for my $uri (@$uris) {
522 my $sec = imap_section($uri);
523 my $mic_arg = $self->{mic_arg}->{$sec} or
524 die "BUG: no Mail::IMAPClient->new arg for $sec";
525 my $mic = PublicInbox::IMAPClient->new(%$mic_arg) or next;
526 my $err = imap_fetch_all($self, $mic, $uri);
527 last if $self->{quit};
528 warn $err, "\n" if $err;
532 sub imap_fetch_fork ($$$) {
533 my ($self, $intvl, $uris) = @_;
534 return if $self->{quit};
535 watch_atfork_parent($self);
536 defined(my $pid = fork) or die "fork: $!";
538 watch_atfork_child($self);
539 watch_imap_fetch_all($self, $uris);
542 $self->{poll_pids}->{$pid} = [ $intvl, $uris ];
543 PublicInbox::DS::dwaitpid($pid, \&imap_fetch_reap, $self);
546 sub imap_fetch_cb ($$$) {
547 my ($self, $intvl, $uris) = @_;
548 sub { imap_fetch_fork($self, $intvl, $uris) };
551 sub imap_fetch_reap { # PublicInbox::DS::dwaitpid callback
552 my ($self, $pid) = @_;
553 my $intvl_uris = delete $self->{poll_pids}->{$pid} or
554 die "BUG: PID=$pid (unknown) reaped: \$?=$?\n";
555 return if $self->{quit};
556 my ($intvl, $uris) = @$intvl_uris;
558 warn "W: PID=$pid died: \$?=$?\n",
559 map { $_->as_string."\n" } @$uris;
561 warn('I: will check ', $_->as_string, " in ${intvl}s\n") for @$uris;
562 PublicInbox::DS::add_timer($intvl, imap_fetch_cb($self, $intvl, $uris));
565 sub watch_imap_init ($) {
567 eval { require PublicInbox::IMAPClient } or
568 die "Mail::IMAPClient is required for IMAP:\n$@\n";
569 eval { require Git } or
570 die "Git (Perl module) is required for IMAP:\n$@\n";
571 eval { require PublicInbox::IMAPTracker } or
572 die "DBD::SQLite is required for IMAP\n:$@\n";
574 my $mic_args = imap_common_init($self); # read args from config
576 # make sure we can connect and cache the credentials in memory
577 $self->{mic_arg} = {}; # schema://authority => IMAPClient->new args
578 my $mics = $self->{mics} = {}; # schema://authority => IMAPClient obj
579 for my $url (sort keys %{$self->{imap}}) {
580 my $uri = PublicInbox::URIimap->new($url);
581 $mics->{imap_section($uri)} //= mic_for($self, $uri, $mic_args);
584 my $idle = []; # [ [ uri1, intvl1 ], [uri2, intvl2] ]
585 my $poll = {}; # intvl_seconds => [ uri1, uri2 ]
586 for my $url (keys %{$self->{imap}}) {
587 my $uri = PublicInbox::URIimap->new($url);
588 my $sec = imap_section($uri);
589 my $mic = $mics->{$sec};
590 my $intvl = $self->{imap_opt}->{$sec}->{poll_intvl};
591 if ($mic->has_capability('IDLE') && !$intvl) {
592 $intvl = $self->{imap_opt}->{$sec}->{idle_intvl};
593 push @$idle, [ $uri, $intvl // () ];
595 push @{$poll->{$intvl || 120}}, $uri;
599 $self->{idle_pids} = {};
600 $self->{idle_todo} = $idle;
601 PublicInbox::DS::requeue($self); # ->event_step to fork
603 return unless scalar keys %$poll;
604 $self->{poll_pids} = {};
606 # poll all URIs for a given interval sequentially
607 while (my ($intvl, $uris) = each %$poll) {
608 PublicInbox::DS::requeue(imap_fetch_cb($self, $intvl, $uris));
613 my ($self, $sig, $oldset) = @_;
614 $self->{oldset} = $oldset;
616 watch_imap_init($self) if $self->{imap};
617 watch_fs_init($self) if $self->{mdre};
618 PublicInbox::DS->SetPostLoopCallback(sub {});
619 PublicInbox::DS->EventLoop until $self->{quit};
620 _done_for_now($self);
624 my ($self, $op) = @_;
625 push @{$self->{ops}}, $op;
626 PublicInbox::DS::requeue($self);
631 return if $self->{quit};
632 my $op = shift @{$self->{ops}};
634 # continue existing scan
636 my $opendirs = $self->{opendirs};
637 my @dirnames = keys %$opendirs;
638 foreach my $dir (@dirnames) {
639 my $dh = delete $opendirs->{$dir};
641 while (my $fn = readdir($dh)) {
642 _try_path($self, "$dir/$fn");
645 $opendirs->{$dir} = $dh if $n < 0;
647 if ($op && $op eq 'full') {
648 foreach my $dir (@{$self->{mdir}}) {
649 next if $opendirs->{$dir}; # already in progress
650 my $ok = opendir(my $dh, $dir);
652 warn "failed to open $dir: $!\n";
656 while (my $fn = readdir($dh)) {
657 _try_path($self, "$dir/$fn");
660 $opendirs->{$dir} = $dh if $n < 0;
663 _done_for_now($self);
664 # do we have more work to do?
665 PublicInbox::DS::requeue($self) if keys %$opendirs;
669 my ($self, $op) = @_;
670 push @{$self->{ops}}, $op;
675 my ($self, $ibx) = @_;
676 my $importers = $self->{importers};
677 my $im = $importers->{"$ibx"} ||= $ibx->importer(0);
678 if (scalar(keys(%$importers)) > 2) {
679 delete $importers->{"$ibx"};
680 _done_for_now($self);
683 $importers->{"$ibx"} = $im;
691 if ($sc->spamcheck($mime, \$tmp)) {
692 return PublicInbox::Eml->new(\$tmp);
694 warn $mime->header('Message-ID')." failed spam check\n";
700 $_[0] =~ s!\Amaildir:!! or return;
707 my ($cur, $ws, $ibx) = @_;
708 if ($ws && !ref($ws) && $ws eq 'watchspam') {
710 E: $cur is a spam folder and cannot be used for `$ibx->{name}' input
719 require PublicInbox::URIimap;
720 my $uri = PublicInbox::URIimap->new($url);
721 $uri ? $uri->canonical->as_string : undef;