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 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 $@;
216 sub watch_fs_init ($) {
219 delete $self->{done_timer};
220 _done_for_now($self);
223 _try_path($self, $_[0]->fullname);
224 $self->{done_timer} //= PublicInbox::DS::requeue($done);
226 require PublicInbox::DirIdle;
227 PublicInbox::DirIdle->new($self->{mdir}, $cb); # EPOLL_CTL_ADD
230 # returns the git config section name, e.g [imap "imaps://user@example.com"]
231 # without the mailbox, so we can share connections between different inboxes
232 sub imap_section ($) {
234 $uri->scheme . '://' . $uri->authority;
238 my ($cfg, $key) = @_;
239 defined(my $v = $cfg->{lc($key)}) or return;
240 $v =~ /\A[0-9]+\z/s and return $v + 0;
241 if (ref($v) eq 'ARRAY') {
242 $v = join(', ', @$v);
243 warn "W: $key has multiple values: $v\nW: $key ignored\n";
245 warn "W: $key=$v is not an integer value in seconds\n";
249 # flesh out common IMAP-specific data structures
250 sub imap_common_init ($) {
252 my $cfg = $self->{config};
253 my $mic_args = {}; # scheme://authority => Mail:IMAPClient arg
254 for my $url (sort keys %{$self->{imap}}) {
255 my $uri = PublicInbox::URIimap->new($url);
256 my $sec = imap_section($uri);
257 for my $k (qw(Starttls Debug Compress)) {
258 my $key = lc("imap.$sec.$k");
259 defined(my $orig = $cfg->{$key}) or next;
260 my $v = PublicInbox::Config::_git_config_bool($orig);
262 $mic_args->{$sec}->{$k} = $v;
264 warn "W: $key=$orig is not boolean\n";
267 my $to = cfg_intvl($cfg, "imap.$sec.Timeout");
268 $mic_args->{$sec}->{Timeout} = $to if $to;
269 $to = cfg_intvl($cfg, "imap.$sec.PollInterval");
270 $self->{imap_opt}->{$sec}->{poll_intvl} = $to if $to;
271 $to = cfg_intvl($cfg, "imap.$sec.IdleInterval");
272 $self->{imap_opt}->{$sec}->{idle_intvl} = $to if $to;
277 sub auth_anon_cb { '' }; # for Mail::IMAPClient::Authcallback
279 sub mic_for ($$$) { # mic = Mail::IMAPClient
280 my ($self, $uri, $mic_args) = @_;
281 my $url = $uri->as_string;
284 protocol => $uri->scheme,
286 username => $uri->user,
287 password => $uri->password,
289 my $common = $mic_args->{imap_section($uri)} // {};
290 my $host = $cred->{host};
293 # IMAPClient mishandles `0', so we pass `127.0.0.1'
294 Server => $host eq '0' ? '127.0.0.1' : $host,
295 Ssl => $uri->scheme eq 'imaps',
296 Keepalive => 1, # SO_KEEPALIVE
297 %$common, # may set Starttls, Compress, Debug ....
299 my $mic = PublicInbox::IMAPClient->new(%$mic_arg) or
300 die "E: <$url> new: $@\n";
302 # default to using STARTTLS if it's available, but allow
303 # it to be disabled since I usually connect to localhost
304 if (!$mic_arg->{Ssl} && !defined($mic_arg->{Starttls}) &&
305 $mic->has_capability('STARTTLS') &&
306 $mic->can('starttls')) {
307 $mic->starttls or die "E: <$url> STARTTLS: $@\n";
310 # do we even need credentials?
311 if (!defined($cred->{username}) &&
312 $mic->has_capability('AUTH=ANONYMOUS')) {
316 Git::credential($cred, 'fill'); # may prompt user here
317 $mic->User($mic_arg->{User} = $cred->{username});
318 $mic->Password($mic_arg->{Password} = $cred->{password});
319 } else { # AUTH=ANONYMOUS
320 $mic->Authmechanism($mic_arg->{Authmechanism} = 'ANONYMOUS');
321 $mic->Authcallback($mic_arg->{Authcallback} = \&auth_anon_cb);
323 if ($mic->login && $mic->IsAuthenticated) {
324 # success! keep IMAPClient->new arg in case we get disconnected
325 $self->{mic_arg}->{imap_section($uri)} = $mic_arg;
327 warn "E: <$url> LOGIN: $@\n";
330 Git::credential($cred, $mic ? 'approve' : 'reject') if $cred;
334 sub imap_fetch_all ($$$) {
335 my ($self, $mic, $uri) = @_;
336 my $sec = imap_section($uri);
337 my $mbx = $uri->mailbox;
338 my $url = $uri->as_string;
339 $mic->Clear(1); # trim results history
340 $mic->examine($mbx) or return "E: EXAMINE $mbx ($sec) failed: $!";
341 my ($r_uidval, $r_uidnext);
342 for ($mic->Results) {
343 /^\* OK \[UIDVALIDITY ([0-9]+)\].*/ and $r_uidval = $1;
344 /^\* OK \[UIDNEXT ([0-9]+)\].*/ and $r_uidnext = $1;
345 last if $r_uidval && $r_uidnext;
347 $r_uidval //= $mic->uidvalidity($mbx) //
348 return "E: $url cannot get UIDVALIDITY";
349 $r_uidnext //= $mic->uidnext($mbx) //
350 return "E: $url cannot get UIDNEXT";
351 my $itrk = PublicInbox::IMAPTracker->new;
352 my ($l_uidval, $l_uid) = $itrk->get_last($url);
353 $l_uidval //= $r_uidval; # first time
355 if ($l_uidval != $r_uidval) {
356 return "E: $url UIDVALIDITY mismatch\n".
357 "E: local=$l_uidval != remote=$r_uidval";
359 my $r_uid = $r_uidnext - 1;
360 if ($l_uid != 1 && $l_uid > $r_uid) {
361 return "E: $url local UID exceeds remote ($l_uid > $r_uid)\n".
362 "E: $url strangely, UIDVALIDLITY matches ($l_uidval)\n";
364 return if $l_uid >= $r_uid; # nothing to do
366 $mic->Uid(1); # the default, we hope
367 my $req = $mic->imap4rev1 ? 'BODY.PEEK[]' : 'RFC822.PEEK';
370 my $inboxes = $self->{imap}->{$url};
371 warn "I: $url fetching $l_uid..$r_uid\n";
373 my $warn_cb = $SIG{__WARN__} || sub { print STDERR @_ };
374 local $SIG{__WARN__} = sub {
375 $warn_cb->("$url UID:$uid\n");
379 $itrk->{dbh}->begin_work;
380 for my $u ($l_uid..$r_uid) {
382 local $0 = "UID:$uid $mbx $sec";
383 my $r = $mic->fetch_hash($uid, $req);
384 unless ($r) { # network error?
385 $err = "E: $url UID FETCH $uid error: $!\n";
389 # messages get deleted, so holes appear
390 defined(my $raw = delete $r->{$uid}->{$key}) or next;
392 # our target audience expects LF-only, save storage
393 $raw =~ s/\r\n/\n/sg;
396 for my $ibx (@$inboxes) {
397 my $eml = PublicInbox::Eml->new($raw);
398 my $x = import_eml($self, $ibx, $eml);
400 } elsif ($inboxes eq 'watchspam') {
401 my $eml = PublicInbox::Eml->new($raw);
402 my $arg = [ $self, $eml, "$uri UID:$uid" ];
403 $self->{config}->each_inbox(\&remove_eml_i, $arg);
405 die "BUG: destination unknown $inboxes";
407 $itrk->update_last($url, $r_uidval, $uid);
408 last if $self->{quit};
410 _done_for_now($self);
411 $itrk->{dbh}->commit;
415 sub imap_idle_once ($$$$) {
416 my ($self, $mic, $intvl, $url) = @_;
417 my $i = $intvl //= (29 * 60);
418 my $end = now() + $intvl;
419 warn "I: $url idling for ${intvl}s\n";
420 local $0 = "IDLE $0";
421 unless ($mic->idle) {
422 return if $self->{quit};
423 return "E: IDLE failed on $url: $!";
425 $self->{idle_mic} = $mic; # for ->quit
427 until ($self->{quit} || grep(/^\* [0-9]+ EXISTS/, @res) || $i <= 0) {
428 @res = $mic->idle_data($i);
431 delete $self->{idle_mic};
432 unless ($self->{quit}) {
433 $mic->IsConnected or return "E: IDLE disconnected on $url";
434 $mic->done or return "E: IDLE DONE failed on $url: $!";
439 # idles on a single URI
440 sub watch_imap_idle_1 ($$$) {
441 my ($self, $uri, $intvl) = @_;
442 my $sec = imap_section($uri);
443 my $mic_arg = $self->{mic_arg}->{$sec} or
444 die "BUG: no Mail::IMAPClient->new arg for $sec";
446 local $0 = $uri->mailbox." $sec";
447 until ($self->{quit}) {
448 $mic //= delete($self->{mics}->{$sec}) //
449 PublicInbox::IMAPClient->new(%$mic_arg);
450 my $err = imap_fetch_all($self, $mic, $uri);
451 $err //= imap_idle_once($self, $mic, $intvl, $uri->as_string);
452 if ($err && !$self->{quit}) {
455 sleep 60 unless $self->{quit};
460 sub watch_atfork_child ($) {
462 delete $self->{idle_pids};
463 PublicInbox::DS->Reset;
464 PublicInbox::Sigfd::sig_setmask($self->{oldset});
465 %SIG = (%SIG, %{$self->{sig}});
468 sub imap_idle_reap { # PublicInbox::DS::dwaitpid callback
469 my ($self, $pid) = @_;
470 my $uri_intvl = delete $self->{idle_pids}->{$pid} or
471 die "BUG: PID=$pid (unknown) reaped: \$?=$?\n";
473 my ($uri, $intvl) = @$uri_intvl;
474 my $url = $uri->as_string;
475 return if $self->{quit};
476 warn "W: PID=$pid on $url died: \$?=$?\n" if $?;
477 push @{$self->{idle_todo}}, $uri_intvl;
478 PubicInbox::DS::requeue($self); # call ->event_step to respawn
481 sub imap_idle_fork ($$) {
482 my ($self, $uri_intvl) = @_;
483 my ($uri, $intvl) = @$uri_intvl;
484 defined(my $pid = fork) or die "fork: $!";
486 watch_atfork_child($self);
487 watch_imap_idle_1($self, $uri, $intvl);
490 $self->{idle_pids}->{$pid} = $uri_intvl;
491 PublicInbox::DS::dwaitpid($pid, \&imap_idle_reap, $self);
496 return if $self->{quit};
497 my $idle_todo = $self->{idle_todo};
498 if ($idle_todo && @$idle_todo) {
499 $self->{mics} = {}; # going to be forking, so disconnect
500 while (my $uri_intvl = shift(@$idle_todo)) {
501 imap_idle_fork($self, $uri_intvl);
504 goto(&fs_scan_step) if $self->{mdre};
507 sub watch_imap_init ($) {
509 eval { require PublicInbox::IMAPClient } or
510 die "Mail::IMAPClient is required for IMAP:\n$@\n";
511 eval { require Git } or
512 die "Git (Perl module) is required for IMAP:\n$@\n";
513 eval { require PublicInbox::IMAPTracker } or
514 die "DBD::SQLite is required for IMAP\n:$@\n";
516 my $mic_args = imap_common_init($self); # read args from config
518 # make sure we can connect and cache the credentials in memory
519 $self->{mic_arg} = {}; # schema://authority => IMAPClient->new args
520 my $mics = $self->{mics} = {}; # schema://authority => IMAPClient obj
521 for my $url (sort keys %{$self->{imap}}) {
522 my $uri = PublicInbox::URIimap->new($url);
523 $mics->{imap_section($uri)} //= mic_for($self, $uri, $mic_args);
526 my $idle = []; # [ [ uri1, intvl1 ], [uri2, intvl2] ]
527 my $poll = {}; # intvl_seconds => [ uri1, uri2 ]
528 for my $url (keys %{$self->{imap}}) {
529 my $uri = PublicInbox::URIimap->new($url);
530 my $sec = imap_section($uri);
531 my $mic = $mics->{$sec};
532 my $intvl = $self->{imap_opt}->{$sec}->{poll_intvl};
533 if ($mic->has_capability('IDLE') && !$intvl) {
534 $intvl = $self->{imap_opt}->{$sec}->{idle_intvl};
535 push @$idle, [ $uri, $intvl // () ];
537 push @{$poll->{$intvl || 120}}, $uri;
541 $self->{idle_pids} = {};
542 $self->{idle_todo} = $idle;
543 PublicInbox::DS::requeue($self); # ->event_step to fork
549 my ($self, $sig, $oldset) = @_;
550 $self->{oldset} = $oldset;
552 watch_imap_init($self) if $self->{imap};
553 watch_fs_init($self) if $self->{mdre};
554 PublicInbox::DS->SetPostLoopCallback(sub {});
555 PublicInbox::DS->EventLoop until $self->{quit};
556 _done_for_now($self);
560 my ($self, $op) = @_;
561 push @{$self->{ops}}, $op;
562 PublicInbox::DS::requeue($self);
567 return if $self->{quit};
568 my $op = shift @{$self->{ops}};
570 # continue existing scan
572 my $opendirs = $self->{opendirs};
573 my @dirnames = keys %$opendirs;
574 foreach my $dir (@dirnames) {
575 my $dh = delete $opendirs->{$dir};
577 while (my $fn = readdir($dh)) {
578 _try_path($self, "$dir/$fn");
581 $opendirs->{$dir} = $dh if $n < 0;
583 if ($op && $op eq 'full') {
584 foreach my $dir (@{$self->{mdir}}) {
585 next if $opendirs->{$dir}; # already in progress
586 my $ok = opendir(my $dh, $dir);
588 warn "failed to open $dir: $!\n";
592 while (my $fn = readdir($dh)) {
593 _try_path($self, "$dir/$fn");
596 $opendirs->{$dir} = $dh if $n < 0;
599 _done_for_now($self);
600 # do we have more work to do?
601 PublicInbox::DS::requeue($self) if keys %$opendirs;
605 my ($self, $op) = @_;
606 push @{$self->{ops}}, $op;
611 my ($self, $ibx) = @_;
612 my $importers = $self->{importers};
613 my $im = $importers->{"$ibx"} ||= $ibx->importer(0);
614 if (scalar(keys(%$importers)) > 2) {
615 delete $importers->{"$ibx"};
616 _done_for_now($self);
619 $importers->{"$ibx"} = $im;
627 if ($sc->spamcheck($mime, \$tmp)) {
628 return PublicInbox::Eml->new(\$tmp);
630 warn $mime->header('Message-ID')." failed spam check\n";
636 $_[0] =~ s!\Amaildir:!! or return;
643 my ($cur, $ws, $ibx) = @_;
644 if ($ws && !ref($ws) && $ws eq 'watchspam') {
646 E: $cur is a spam folder and cannot be used for `$ibx->{name}' input
655 require PublicInbox::URIimap;
656 my $uri = PublicInbox::URIimap->new($url);
657 $uri ? $uri->canonical->as_string : undef;