1 # Copyright (C) all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # Each instance of this represents an IMAP client connected to
5 # public-inbox-imapd. Much of this was taken from NNTP, but
6 # further refined while experimenting on future ideas to handle
11 # * NNTP article numbers are UIDs, mm->created_at is UIDVALIDITY
13 # * public-inboxes are sliced into mailboxes of 50K messages
14 # to not overload MUAs: $NEWSGROUP_NAME.$SLICE_INDEX
15 # Slices are similar in concept to v2 "epochs". Epochs
16 # are for the limitations of git clients, while slices are
17 # for the limitations of IMAP clients.
19 # * We also take advantage of slices being only 50K to store
20 # "UID offset" to message sequence number (MSN) mapping
21 # as a 50K uint16_t array (via pack("S*", ...)). "UID offset"
22 # is the offset from {uid_base} which determines the start of
26 # imapd: PublicInbox::IMAPD ref
27 # ibx: PublicInbox::Inbox ref
28 # long_cb: long_response private data
29 # uid_base: base UID for mailbox slice (0-based)
30 # -login_tag: IMAP TAG for LOGIN
31 # -idle_tag: IMAP response tag for IDLE
32 # uo2m: UID-to-MSN mapping
33 package PublicInbox::IMAP;
35 use parent qw(PublicInbox::DS);
37 use PublicInbox::EmlContentFoo qw(parse_content_disposition);
38 use PublicInbox::DS qw(now);
39 use PublicInbox::GitAsyncCat;
40 use Text::ParseWords qw(parse_line);
42 use PublicInbox::IMAPsearchqp;
45 for my $mod (qw(Email::Address::XS Mail::Address)) {
46 eval "require $mod" or next;
47 $Address = $mod and last;
49 die "neither Email::Address::XS nor Mail::Address loaded: $@" if !$Address;
51 sub LINE_MAX () { 8000 } # RFC 2683 3.2.1.5
53 # Changing UID_SLICE will cause grief for clients which cache.
54 # This also needs to be <64K: we pack it into a uint16_t
55 # for long_response UID (offset) => MSN mappings
56 sub UID_SLICE () { 50_000 }
58 # these values area also used for sorting
59 sub NEED_SMSG () { 1 }
60 sub NEED_BLOB () { NEED_SMSG|2 }
61 sub CRLF_BREF () { 4 }
63 sub CRLF_HDR () { 16 }
65 sub CRLF_BDY () { 64 }
66 my $OP_EML_NEW = [ EML_HDR - 1, \&op_eml_new ];
67 my $OP_CRLF_BREF = [ CRLF_BREF, \&op_crlf_bref ];
68 my $OP_CRLF_HDR = [ CRLF_HDR, \&op_crlf_hdr ];
69 my $OP_CRLF_BDY = [ CRLF_BDY, \&op_crlf_bdy ];
72 'BODY[HEADER]' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ],
73 'BODY[TEXT]' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ],
74 'BODY[]' => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ],
75 'RFC822.HEADER' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ],
76 'RFC822.TEXT' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ],
77 'RFC822.SIZE' => [ NEED_SMSG, \&emit_rfc822_size ],
78 RFC822 => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ],
79 BODY => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_body ],
80 BODYSTRUCTURE => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_bodystructure ],
81 ENVELOPE => [ NEED_BLOB|EML_HDR, \&emit_envelope ],
82 FLAGS => [ 0, \&emit_flags ],
83 INTERNALDATE => [ NEED_SMSG, \&emit_internaldate ],
85 my %FETCH_ATT = map { $_ => [ $_ ] } keys %FETCH_NEED;
87 # aliases (RFC 3501 section 6.4.5)
88 $FETCH_ATT{FAST} = [ qw(FLAGS INTERNALDATE RFC822.SIZE) ];
89 $FETCH_ATT{ALL} = [ @{$FETCH_ATT{FAST}}, 'ENVELOPE' ];
90 $FETCH_ATT{FULL} = [ @{$FETCH_ATT{ALL}}, 'BODY' ];
92 for my $att (keys %FETCH_ATT) {
93 my %h = map { $_ => $FETCH_NEED{$_} } @{$FETCH_ATT{$att}};
94 $FETCH_ATT{$att} = \%h;
98 my $valid_range = '[0-9]+|[0-9]+:[0-9]+|[0-9]+:\*';
99 $valid_range = qr/\A(?:$valid_range)(?:,(?:$valid_range))*\z/;
103 my $capa = capa($self);
104 $self->write(\"* OK [$capa] public-inbox-imapd ready\r\n");
108 my (undef, $sock, $imapd) = @_;
109 (bless { imapd => $imapd }, 'PublicInbox::IMAP_preauth')->greet($sock)
117 # dovecot advertises IDLE pre-login; perhaps because some clients
118 # depend on it, so we'll do the same
119 my $capa = 'CAPABILITY IMAP4rev1 IDLE';
120 if ($self->logged_in) {
121 $capa .= ' COMPRESS=DEFLATE';
123 if (!($self->{sock} // $self)->can('accept_SSL') &&
124 $self->{imapd}->{accept_tls}) {
125 $capa .= ' STARTTLS';
127 $capa .= ' AUTH=ANONYMOUS';
131 sub login_success ($$) {
132 my ($self, $tag) = @_;
133 bless $self, 'PublicInbox::IMAP';
134 my $capa = capa($self);
135 "$tag OK [$capa] Logged in\r\n";
138 sub auth_challenge_ok ($) {
140 my $tag = delete($self->{-login_tag}) or return;
141 login_success($self, $tag);
144 sub cmd_login ($$$$) {
145 my ($self, $tag) = @_; # ignore ($user, $password) = ($_[2], $_[3])
146 login_success($self, $tag);
150 my ($self, $tag) = @_;
151 delete @$self{qw(uid_base uo2m)};
152 delete $self->{ibx} ? "$tag OK Close done\r\n"
153 : "$tag BAD No mailbox\r\n";
156 sub cmd_logout ($$) {
157 my ($self, $tag) = @_;
158 delete $self->{-idle_tag};
159 $self->write(\"* BYE logging out\r\n$tag OK Logout done\r\n");
160 $self->shutdn; # PublicInbox::DS::shutdn
164 sub cmd_authenticate ($$$) {
165 my ($self, $tag) = @_; # $method = $_[2], should be "ANONYMOUS"
166 $self->{-login_tag} = $tag;
170 sub cmd_capability ($$) {
171 my ($self, $tag) = @_;
172 '* '.capa($self)."\r\n$tag OK Capability done\r\n";
175 # uo2m: UID Offset to MSN, this is an arrayref by default,
176 # but uo2m_hibernate can compact and deduplicate it
177 sub uo2m_ary_new ($;$) {
178 my ($self, $exists) = @_;
179 my $ub = $self->{uid_base};
180 my $uids = $self->{ibx}->over(1)->uid_range($ub + 1, $ub + UID_SLICE);
182 # convert UIDs to offsets from {base}
183 my @tmp; # [$UID_OFFSET] => $MSN
186 $tmp[$_ - $ub] = ++$msn for @$uids;
187 $$exists = $msn if $exists;
191 # changes UID-offset-to-MSN mapping into a deduplicated scalar:
192 # uint16_t uo2m[UID_SLICE].
193 # May be swapped out for idle clients if THP is disabled.
194 sub uo2m_hibernate ($) {
196 ref(my $uo2m = $self->{uo2m}) or return;
197 my %dedupe = ( uo2m_pack($uo2m) => undef );
198 $self->{uo2m} = (keys(%dedupe))[0];
202 sub uo2m_last_uid ($) {
204 defined(my $uo2m = $self->{uo2m}) or die 'BUG: uo2m_last_uid w/o {uo2m}';
205 (ref($uo2m) ? @$uo2m : (length($uo2m) >> 1)) + $self->{uid_base};
209 # $_[0] is an arrayref of MSNs, it may have undef gaps if there
210 # are gaps in the corresponding UIDs: [ msn1, msn2, undef, msn3 ]
211 no warnings 'uninitialized';
212 pack('S*', @{$_[0]});
215 # extend {uo2m} to account for new messages which arrived since
216 # {uo2m} was created.
217 sub uo2m_extend ($$;$) {
218 my ($self, $new_uid_max) = @_;
219 defined(my $uo2m = $self->{uo2m}) or
220 return($self->{uo2m} = uo2m_ary_new($self));
221 my $beg = uo2m_last_uid($self); # last UID we've learned
222 return $uo2m if $beg >= $new_uid_max; # fast path
224 # need to extend the current range:
225 my $base = $self->{uid_base};
227 my $uids = $self->{ibx}->over(1)->uid_range($beg, $base + UID_SLICE);
228 return $uo2m if !scalar(@$uids);
229 my @tmp; # [$UID_OFFSET] => $MSN
230 my $write_method = $_[2] // 'msg_more';
232 my $msn = $uo2m->[-1];
233 $tmp[$_ - $beg] = ++$msn for @$uids;
234 $self->$write_method("* $msn EXISTS\r\n");
238 my $msn = unpack('S', substr($uo2m, -2, 2));
239 $tmp[$_ - $beg] = ++$msn for @$uids;
240 $self->$write_method("* $msn EXISTS\r\n");
241 $uo2m .= uo2m_pack(\@tmp);
242 my %dedupe = ($uo2m => undef);
243 $self->{uo2m} = (keys %dedupe)[0];
248 my ($self, $tag) = @_;
249 defined($self->{uid_base}) and
250 uo2m_extend($self, $self->{uid_base} + UID_SLICE);
251 \"$tag OK Noop done\r\n";
254 # the flexible version which works on scalars and array refs.
255 # Must call uo2m_extend before this
257 my ($self, $uid) = @_;
258 my $uo2m = $self->{uo2m};
259 my $off = $uid - $self->{uid_base} - 1;
260 ref($uo2m) ? $uo2m->[$off] : unpack('S', substr($uo2m, $off << 1, 2));
263 # returns an arrayref of UIDs, so MSNs can be translated to UIDs via:
264 # $msn2uid->[$MSN-1] => $UID. The result of this is always ephemeral
265 # and does not live beyond the event loop.
268 my $base = $self->{uid_base};
269 my $uo2m = uo2m_extend($self, $base + UID_SLICE);
270 $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
274 for my $msn (@$uo2m) {
276 $msn2uid[$msn - 1] = $uo + $base if $msn;
281 # converts a set of message sequence numbers in requests to UIDs:
282 sub msn_to_uid_range ($$) {
284 $_[1] =~ s!([0-9]+)!$msn2uid->[$1 - 1] // ($msn2uid->[-1] // 0 + 1)!sge;
287 # called by PublicInbox::InboxIdle
288 sub on_inbox_unlock {
289 my ($self, $ibx) = @_;
290 my $uid_end = $self->{uid_base} + UID_SLICE;
291 uo2m_extend($self, $uid_end, 'write');
292 my $new = uo2m_last_uid($self);
293 if ($new == $uid_end) { # max exceeded $uid_end
294 # continue idling w/o inotify
295 my $sock = $self->{sock} or return;
296 $ibx->unsubscribe_unlock(fileno($sock));
300 # called every minute or so by PublicInbox::DS::later
301 my $IDLERS; # fileno($obj->{sock}) => PublicInbox::IMAP
305 for my $i (values %$old) {
306 next if ($i->{wbuf} || !exists($i->{-idle_tag}));
307 $IDLERS->{fileno($i->{sock})} = $i;
308 $i->write(\"* OK Still here\r\n");
311 PublicInbox::DS::add_uniq_timer('idle', 60, \&idle_tick_all);
315 my ($self, $tag) = @_;
316 # IDLE seems allowed by dovecot w/o a mailbox selected *shrug*
317 my $ibx = $self->{ibx} or return "$tag BAD no mailbox selected\r\n";
318 my $uid_end = $self->{uid_base} + UID_SLICE;
319 uo2m_extend($self, $uid_end);
320 my $sock = $self->{sock} or return;
321 my $fd = fileno($sock);
322 $self->{-idle_tag} = $tag;
323 # only do inotify on most recent slice
324 if ($ibx->over(1)->max < $uid_end) {
325 $ibx->subscribe_unlock($fd, $self);
326 $self->{imapd}->idler_start;
328 PublicInbox::DS::add_uniq_timer('idle', 60, \&idle_tick_all);
329 $IDLERS->{$fd} = $self;
334 my ($self, $ibx) = @_;
335 my $sock = $self->{sock} or return;
336 my $fd = fileno($sock);
337 delete $IDLERS->{$fd};
338 $ibx->unsubscribe_unlock($fd);
342 my ($self, $tag) = @_; # $tag is "DONE" (case-insensitive)
343 defined(my $idle_tag = delete $self->{-idle_tag}) or
344 return "$tag BAD not idle\r\n";
345 my $ibx = $self->{ibx} or do {
346 warn "BUG: idle_tag set w/o inbox";
347 return "$tag BAD internal bug\r\n";
349 stop_idle($self, $ibx);
350 "$idle_tag OK Idle done\r\n";
353 sub ensure_slices_exist ($$$) {
354 my ($imapd, $ibx, $max) = @_;
355 defined(my $mb_top = $ibx->{newsgroup}) or return;
356 my $mailboxes = $imapd->{mailboxes};
358 for (my $i = int($max/UID_SLICE); $i >= 0; --$i) {
359 my $sub_mailbox = "$mb_top.$i";
360 last if exists $mailboxes->{$sub_mailbox};
361 $mailboxes->{$sub_mailbox} = $ibx;
362 $sub_mailbox =~ s/\Ainbox\./INBOX./i; # more familiar to users
363 push @created, $sub_mailbox;
365 return unless @created;
366 my $l = $imapd->{mailboxlist} or return;
367 push @$l, map { qq[* LIST (\\HasNoChildren) "." $_\r\n] } @created;
370 sub inbox_lookup ($$;$) {
371 my ($self, $mailbox, $examine) = @_;
372 my ($ibx, $exists, $uidmax, $uid_base) = (undef, 0, 0, 0);
373 $mailbox = lc $mailbox;
374 $ibx = $self->{imapd}->{mailboxes}->{$mailbox} or return;
375 my $over = $ibx->over(1);
376 if ($over != $ibx) { # not a dummy
377 $mailbox =~ /\.([0-9]+)\z/ or
378 die "BUG: unexpected dummy mailbox: $mailbox\n";
379 $uid_base = $1 * UID_SLICE;
381 $uidmax = $ibx->mm->num_highwater // 0;
383 $self->{uid_base} = $uid_base;
385 $self->{uo2m} = uo2m_ary_new($self, \$exists);
387 my $uid_end = $uid_base + UID_SLICE;
388 $exists = $over->imap_exists($uid_base, $uid_end);
390 ensure_slices_exist($self->{imapd}, $ibx, $over->max);
393 $self->{uid_base} = $uid_base;
395 delete $self->{uo2m};
397 # if "INBOX.foo.bar" is selected and "INBOX.foo.bar.0",
398 # check for new UID ranges (e.g. "INBOX.foo.bar.1")
399 if (my $z = $self->{imapd}->{mailboxes}->{"$mailbox.0"}) {
400 ensure_slices_exist($self->{imapd}, $z,
404 ($ibx, $exists, $uidmax + 1, $uid_base);
407 sub cmd_examine ($$$) {
408 my ($self, $tag, $mailbox) = @_;
409 # XXX: do we need this? RFC 5162/7162
410 my $ret = $self->{ibx} ? "* OK [CLOSED] previous closed\r\n" : '';
411 my ($ibx, $exists, $uidnext, $base) = inbox_lookup($self, $mailbox, 1);
412 return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
417 * OK [PERMANENTFLAGS ()] Read-only mailbox\r
418 * OK [UNSEEN $exists]\r
419 * OK [UIDNEXT $uidnext]\r
420 * OK [UIDVALIDITY $ibx->{uidvalidity}]\r
421 $tag OK [READ-ONLY] EXAMINE/SELECT done\r
429 } elsif ($v =~ /[{"\r\n%*\\\[]/) { # literal string
430 '{' . length($v) . "}\r\n" . $v;
431 } else { # quoted string
436 sub addr_envelope ($$;$) {
437 my ($eml, $x, $y) = @_;
438 my $v = $eml->header_raw($x) //
439 ($y ? $eml->header_raw($y) : undef) // return 'NIL';
441 my @x = $Address->parse($v) or return 'NIL';
443 map { '(' . join(' ',
444 _esc($_->name), 'NIL',
445 _esc($_->user), _esc($_->host)
451 sub eml_envelope ($) {
454 _esc($eml->header_raw('Date')),
455 _esc($eml->header_raw('Subject')),
456 addr_envelope($eml, 'From'),
457 addr_envelope($eml, 'Sender', 'From'),
458 addr_envelope($eml, 'Reply-To', 'From'),
459 addr_envelope($eml, 'To'),
460 addr_envelope($eml, 'Cc'),
461 addr_envelope($eml, 'Bcc'),
462 _esc($eml->header_raw('In-Reply-To')),
463 _esc($eml->header_raw('Message-ID')),
469 if ($hash && scalar keys %$hash) {
470 $hash = [ %$hash ]; # flatten hash into 1-dimensional array
471 '(' . join(' ', map { _esc($_) } @$hash) . ')';
477 sub body_disposition ($) {
479 my $cd = $eml->header_raw('Content-Disposition') or return 'NIL';
480 $cd = parse_content_disposition($cd);
481 my $buf = '('._esc($cd->{type});
482 $buf .= ' ' . _esc_hash($cd->{attributes});
486 sub body_leaf ($$;$) {
487 my ($eml, $structure, $hold) = @_;
489 $eml->{is_submsg} and # parent was a message/(rfc822|news|global)
490 $buf .= eml_envelope($eml). ' ';
492 $buf .= '('._esc($ct->{type}).' ';
493 $buf .= _esc($ct->{subtype});
494 $buf .= ' ' . _esc_hash($ct->{attributes});
495 $buf .= ' ' . _esc($eml->header_raw('Content-ID'));
496 $buf .= ' ' . _esc($eml->header_raw('Content-Description'));
497 my $cte = $eml->header_raw('Content-Transfer-Encoding') // '7bit';
498 $buf .= ' ' . _esc($cte);
499 $buf .= ' ' . $eml->{imap_body_len};
500 $buf .= ' '.($eml->body_raw =~ tr/\n/\n/) if lc($ct->{type}) eq 'text';
502 # for message/(rfc822|global|news), $hold[0] should have envelope
503 $buf .= ' ' . (@$hold ? join('', @$hold) : 'NIL') if $hold;
506 $buf .= ' '._esc($eml->header_raw('Content-MD5'));
507 $buf .= ' '. body_disposition($eml);
508 $buf .= ' '._esc($eml->header_raw('Content-Language'));
509 $buf .= ' '._esc($eml->header_raw('Content-Location'));
514 sub body_parent ($$$) {
515 my ($eml, $structure, $hold) = @_;
517 my $type = lc($ct->{type});
518 if ($type eq 'multipart') {
520 $buf .= @$hold ? join('', @$hold) : 'NIL';
521 $buf .= ' '._esc($ct->{subtype});
523 $buf .= ' '._esc_hash($ct->{attributes});
524 $buf .= ' '.body_disposition($eml);
525 $buf .= ' '._esc($eml->header_raw('Content-Language'));
526 $buf .= ' '._esc($eml->header_raw('Content-Location'));
530 } else { # message/(rfc822|global|news)
531 @$hold = (body_leaf($eml, $structure, $hold));
535 # this is gross, but we need to process the parent part AFTER
536 # the child parts are done
537 sub bodystructure_prep {
539 my ($eml, $depth) = @$p; # ignore idx
540 # set length here, as $eml->{bdy} gets deleted for message/rfc822
541 $eml->{imap_body_len} = length($eml->body_raw);
542 push @$q, $eml, $depth;
545 # for FETCH BODY and FETCH BODYSTRUCTURE
546 sub fetch_body ($;$) {
547 my ($eml, $structure) = @_;
549 $eml->each_part(\&bodystructure_prep, \@q, 0, 1);
553 my ($part, $depth) = splice(@q, -2);
554 my $is_mp_parent = $depth == ($cur_depth - 1);
558 body_parent($part, $structure, \@hold);
560 unshift @hold, body_leaf($part, $structure);
567 my ($self, $smsg, $bref, $ops, $partial) = @_;
568 my $uid = $smsg->{num};
569 $self->msg_more('* '.uid2msn($self, $uid)." FETCH (UID $uid");
571 for (my $i = 0; $i < @$ops;) {
573 $ops->[$i++]->($self, $k, $smsg, $bref, $eml);
575 partial_emit($self, $partial, $eml) if $partial;
576 $self->msg_more(")\r\n");
579 sub fetch_blob_cb { # called by git->cat_async via ibx_async_cat
580 my ($bref, $oid, $type, $size, $fetch_arg) = @_;
581 my ($self, undef, $msgs, $range_info, $ops, $partial) = @$fetch_arg;
582 my $ibx = $self->{ibx} or return $self->close; # client disconnected
583 my $smsg = shift @$msgs or die 'BUG: no smsg';
584 if (!defined($oid)) {
585 # it's possible to have TOCTOU if an admin runs
586 # public-inbox-(edit|purge), just move onto the next message
587 warn "E: $smsg->{blob} missing in $ibx->{inboxdir}\n";
588 return $self->requeue_once;
590 $smsg->{blob} eq $oid or die "BUG: $smsg->{blob} != $oid";
593 if (!$self->{wbuf} && (my $nxt = $msgs->[0])) {
594 $pre = ibx_async_prefetch($ibx, $nxt->{blob},
595 \&fetch_blob_cb, $fetch_arg);
597 fetch_run_ops($self, $smsg, $bref, $ops, $partial);
598 $pre ? $self->zflush : $self->requeue_once;
602 my ($self, $k, undef, $bref) = @_;
603 $self->msg_more(" $k {" . length($$bref)."}\r\n");
604 $self->msg_more($$bref);
607 # Mail::IMAPClient::message_string cares about this by default,
608 # (->Ignoresizeerrors attribute). Admins are encouraged to
609 # --reindex for IMAP support, anyways.
610 sub emit_rfc822_size {
611 my ($self, $k, $smsg) = @_;
612 $self->msg_more(' RFC822.SIZE ' . $smsg->{bytes});
615 sub emit_internaldate {
616 my ($self, undef, $smsg) = @_;
617 $self->msg_more(' INTERNALDATE "'.$smsg->internaldate.'"');
620 sub emit_flags { $_[0]->msg_more(' FLAGS ()') }
623 my ($self, undef, undef, undef, $eml) = @_;
624 $self->msg_more(' ENVELOPE '.eml_envelope($eml));
627 sub emit_rfc822_header {
628 my ($self, $k, undef, undef, $eml) = @_;
629 $self->msg_more(" $k {".length(${$eml->{hdr}})."}\r\n");
630 $self->msg_more(${$eml->{hdr}});
633 # n.b. this is sorted to be after any emit_eml_new ops
634 sub emit_rfc822_text {
635 my ($self, $k, undef, $bref) = @_;
636 $self->msg_more(" $k {".length($$bref)."}\r\n");
637 $self->msg_more($$bref);
640 sub emit_bodystructure {
641 my ($self, undef, undef, undef, $eml) = @_;
642 $self->msg_more(' BODYSTRUCTURE '.fetch_body($eml, 1));
646 my ($self, undef, undef, undef, $eml) = @_;
647 $self->msg_more(' BODY '.fetch_body($eml));
650 # set $eml once ($_[4] == $eml, $_[3] == $bref)
651 sub op_eml_new { $_[4] = PublicInbox::Eml->new($_[3]) }
653 # s/From / fixes old bug from import (pre-a0c07cba0e5d8b6a)
655 ${$_[0]} =~ s/(?<!\r)\n/\r\n/sg;
656 ${$_[0]} =~ s/\A[\r\n]*From [^\r\n]*\r\n//s;
659 sub op_crlf_bref { to_crlf_full($_[3]) }
661 sub op_crlf_hdr { to_crlf_full($_[4]->{hdr}) }
663 sub op_crlf_bdy { ${$_[4]->{bdy}} =~ s/(?<!\r)\n/\r\n/sg if $_[4]->{bdy} }
665 sub uid_clamp ($$$) {
666 my ($self, $beg, $end) = @_;
667 my $uid_min = $self->{uid_base} + 1;
668 my $uid_end = $uid_min + UID_SLICE - 1;
669 $$beg = $uid_min if $$beg < $uid_min;
670 $$end = $uid_end if $$end > $uid_end;
673 sub range_step ($$) {
674 my ($self, $range_csv) = @_;
675 my ($beg, $end, $range);
676 if ($$range_csv =~ s/\A([^,]+),//) {
679 $range = $$range_csv;
682 my $uid_base = $self->{uid_base};
683 my $uid_end = $uid_base + UID_SLICE;
684 if ($range =~ /\A([0-9]+):([0-9]+)\z/) {
685 ($beg, $end) = ($1 + 0, $2 + 0);
686 uid_clamp($self, \$beg, \$end);
687 } elsif ($range =~ /\A([0-9]+):\*\z/) {
689 $end = $self->{ibx}->over(1)->max;
690 $end = $uid_end if $end > $uid_end;
691 $beg = $end if $beg > $end;
692 uid_clamp($self, \$beg, \$end);
693 } elsif ($range =~ /\A[0-9]+\z/) {
694 $beg = $end = $range + 0;
695 # just let the caller do an out-of-range query if a single
696 # UID is out-of-range
697 ++$beg if ($beg <= $uid_base || $end > $uid_end);
699 return 'BAD fetch range';
701 [ $beg, $end, $$range_csv ];
704 sub refill_range ($$$) {
705 my ($self, $msgs, $range_info) = @_;
706 my ($beg, $end, $range_csv) = @$range_info;
707 if (scalar(@$msgs = @{$self->{ibx}->over(1)->query_xover($beg, $end)})){
708 $range_info->[0] = $msgs->[-1]->{num} + 1;
711 return 'OK Fetch done' if !$range_csv;
712 my $next_range = range_step($self, \$range_csv);
713 return $next_range if !ref($next_range); # error
714 @$range_info = @$next_range;
715 undef; # keep looping
718 sub fetch_blob { # long_response
719 my ($self, $tag, $msgs, $range_info, $ops, $partial) = @_;
720 while (!@$msgs) { # rare
721 if (my $end = refill_range($self, $msgs, $range_info)) {
722 $self->write(\"$tag $end\r\n");
726 uo2m_extend($self, $msgs->[-1]->{num});
727 ibx_async_cat($self->{ibx}, $msgs->[0]->{blob},
728 \&fetch_blob_cb, \@_);
731 sub fetch_smsg { # long_response
732 my ($self, $tag, $msgs, $range_info, $ops) = @_;
733 while (!@$msgs) { # rare
734 if (my $end = refill_range($self, $msgs, $range_info)) {
735 $self->write(\"$tag $end\r\n");
739 uo2m_extend($self, $msgs->[-1]->{num});
740 fetch_run_ops($self, $_, undef, $ops) for @$msgs;
745 sub refill_uids ($$$;$) {
746 my ($self, $uids, $range_info, $sql) = @_;
747 my ($beg, $end, $range_csv) = @$range_info;
748 my $over = $self->{ibx}->over(1);
750 if (scalar(@$uids = @{$over->uid_range($beg, $end, $sql)})) {
751 $range_info->[0] = $uids->[-1] + 1; # update $beg
753 } elsif (!$range_csv) {
756 my $next_range = range_step($self, \$range_csv);
757 return $next_range if !ref($next_range); # error
758 ($beg, $end, $range_csv) = @$range_info = @$next_range;
764 sub fetch_uid { # long_response
765 my ($self, $tag, $uids, $range_info, $ops) = @_;
766 if (defined(my $err = refill_uids($self, $uids, $range_info))) {
767 $err ||= 'OK Fetch done';
768 $self->write("$tag $err\r\n");
771 my $adj = $self->{uid_base} + 1;
772 my $uo2m = uo2m_extend($self, $uids->[-1]);
773 $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
776 $self->msg_more("* $uo2m->[$_ - $adj] FETCH (UID $_");
777 for ($i = 0; $i < @$ops;) {
779 $ops->[$i++]->($self, $k);
781 $self->msg_more(")\r\n");
787 sub cmd_status ($$$;@) {
788 my ($self, $tag, $mailbox, @items) = @_;
789 return "$tag BAD no items\r\n" if !scalar(@items);
790 ($items[0] !~ s/\A\(//s || $items[-1] !~ s/\)\z//s) and
791 return "$tag BAD invalid args\r\n";
792 my ($ibx, $exists, $uidnext) = inbox_lookup($self, $mailbox);
793 return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
795 for my $it (@items) {
798 if ($it =~ /\A(?:MESSAGES|UNSEEN|RECENT)\z/) {
800 } elsif ($it eq 'UIDNEXT') {
802 } elsif ($it eq 'UIDVALIDITY') {
803 push @it, $ibx->{uidvalidity};
805 return "$tag BAD invalid item\r\n";
808 return "$tag BAD no items\r\n" if !@it;
809 "* STATUS $mailbox (".join(' ', @it).")\r\n" .
810 "$tag OK Status done\r\n";
813 my %patmap = ('*' => '.*', '%' => '[^\.]*');
814 sub cmd_list ($$$$) {
815 my ($self, $tag, $refname, $wildcard) = @_;
816 my $l = $self->{imapd}->{mailboxlist};
817 if ($refname eq '' && $wildcard eq '') {
818 # request for hierarchy delimiter
819 $l = [ qq[* LIST (\\Noselect) "." ""\r\n] ];
820 } elsif ($refname ne '' || $wildcard ne '*') {
821 $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!egi;
822 $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/is, @$l) ];
824 \(join('', @$l, "$tag OK List done\r\n"));
827 sub cmd_lsub ($$$$) {
828 my (undef, $tag) = @_; # same args as cmd_list
829 "$tag OK Lsub done\r\n";
832 sub eml_index_offs_i { # PublicInbox::Eml::each_part callback
834 my ($eml, undef, $idx) = @$p;
835 if ($idx && lc($eml->ct->{type}) eq 'multipart') {
836 $eml->{imap_bdy} = $eml->{bdy} // \'';
838 $all->{$idx} = $eml; # $idx => Eml
841 # prepares an index for BODY[$SECTION_IDX] fetches
842 sub eml_body_idx ($$) {
843 my ($eml, $section_idx) = @_;
844 my $idx = $eml->{imap_all_parts} // do {
846 $eml->each_part(\&eml_index_offs_i, $all, 0, 1);
847 # top-level of multipart, BODY[0] not allowed (nz-number)
849 $eml->{imap_all_parts} = $all;
851 $idx->{$section_idx};
854 # BODY[($SECTION_IDX)?(.$SECTION_NAME)?]<$offset.$bytes>
856 my ($eml, $section_idx, $section_name) = @_;
857 if (defined $section_idx) {
858 $eml = eml_body_idx($eml, $section_idx) or return;
860 if (defined $section_name) {
861 if ($section_name eq 'MIME') {
862 # RFC 3501 6.4.5 states:
863 # The MIME part specifier MUST be prefixed
864 # by one or more numeric part specifiers
865 return unless defined $section_idx;
866 return $eml->header_obj->as_string . "\r\n";
868 my $bdy = $eml->{bdy} // $eml->{imap_bdy} // \'';
869 $eml = PublicInbox::Eml->new($$bdy);
870 if ($section_name eq 'TEXT') {
871 return $eml->body_raw;
872 } elsif ($section_name eq 'HEADER') {
873 return $eml->header_obj->as_string . "\r\n";
875 die "BUG: bad section_name=$section_name";
878 ${$eml->{bdy} // $eml->{imap_bdy} // \''};
881 # similar to what's in PublicInbox::Eml::re_memo, but doesn't memoize
882 # to avoid OOM with malicious users
883 sub hdrs_regexp ($) {
885 my $names = join('|', map { "\Q$_" } split(/[ \t]+/, $hdrs));
886 qr/^(?:$names):[ \t]*[^\n]*\r?\n # 1st line
887 # continuation lines:
888 (?:[^:\n]*?[ \t]+[^\n]*\r?\n)*
892 # BODY[($SECTION_IDX.)?HEADER.FIELDS.NOT ($HDRS)]<$offset.$bytes>
893 sub partial_hdr_not {
894 my ($eml, $section_idx, $hdrs_re) = @_;
895 if (defined $section_idx) {
896 $eml = eml_body_idx($eml, $section_idx) or return;
898 my $str = $eml->header_obj->as_string;
899 $str =~ s/$hdrs_re//g;
900 $str =~ s/(?<!\r)\n/\r\n/sg;
904 # BODY[($SECTION_IDX.)?HEADER.FIELDS ($HDRS)]<$offset.$bytes>
905 sub partial_hdr_get {
906 my ($eml, $section_idx, $hdrs_re) = @_;
907 if (defined $section_idx) {
908 $eml = eml_body_idx($eml, $section_idx) or return;
910 my $str = $eml->header_obj->as_string;
911 $str = join('', ($str =~ m/($hdrs_re)/g));
912 $str =~ s/(?<!\r)\n/\r\n/sg;
916 sub partial_prepare ($$$$) {
917 my ($need, $partial, $want, $att) = @_;
919 # recombine [ "BODY[1.HEADER.FIELDS", "(foo", "bar)]" ]
920 # back to: "BODY[1.HEADER.FIELDS (foo bar)]"
921 return unless $att =~ /\ABODY\[/s;
922 until (rindex($att, ']') >= 0) {
923 my $next = shift @$want or return;
924 $att .= ' ' . uc($next);
926 if ($att =~ /\ABODY\[([0-9]+(?:\.[0-9]+)*)? # 1 - section_idx
927 (?:\.(HEADER|MIME|TEXT))? # 2 - section_name
928 \](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 3, 4
929 $partial->{$att} = [ \&partial_body, $1, $2, $3, $4 ];
930 $$need |= CRLF_BREF|EML_HDR|EML_BDY;
931 } elsif ($att =~ /\ABODY\[(?:([0-9]+(?:\.[0-9]+)*)\.)? # 1 - section_idx
932 (?:HEADER\.FIELDS(\.NOT)?)\x20 # 2
933 \(([A-Z0-9\-\x20]+)\) # 3 - hdrs
934 \](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 4 5
935 my $tmp = $partial->{$att} = [ $2 ? \&partial_hdr_not
938 $tmp->[2] = hdrs_regexp($3);
940 # don't emit CRLF_HDR instruction, here, partial_hdr_*
941 # will do CRLF conversion with only the extracted result
942 # and not waste time converting lines we don't care about.
949 sub partial_emit ($$$) {
950 my ($self, $partial, $eml) = @_;
952 my ($k, $cb, @args) = @$_;
953 my ($offset, $len) = splice(@args, -2);
954 # $cb is partial_body|partial_hdr_get|partial_hdr_not
955 my $str = $cb->($eml, @args) // '';
956 if (defined $offset) {
958 $str = substr($str, $offset, $len);
959 $k =~ s/\.$len>\z/>/ or warn
960 "BUG: unable to remove `.$len>' from `$k'";
962 $str = substr($str, $offset);
968 $self->msg_more(" $k {$len}\r\n");
969 $self->msg_more($str);
973 sub fetch_compile ($) {
975 if ($want->[0] =~ s/\A\(//s) {
976 $want->[-1] =~ s/\)\z//s or return 'BAD no rparen';
978 my (%partial, %seen, @op);
980 while (defined(my $att = shift @$want)) {
982 next if $att eq 'UID'; # always returned
983 $att =~ s/\ABODY\.PEEK\[/BODY\[/; # we're read-only
984 my $x = $FETCH_ATT{$att};
986 while (my ($k, $fl_cb) = each %$x) {
988 $need |= $fl_cb->[0];
989 push @op, [ @$fl_cb, $k ];
991 } elsif (!partial_prepare(\$need, \%partial, $want, $att)) {
992 return "BAD param: $att";
997 # stabilize partial order for consistency and ease-of-debugging:
998 if (scalar keys %partial) {
1000 $r[2] = [ map { [ $_, @{$partial{$_}} ] } sort keys %partial ];
1003 push @op, $OP_EML_NEW if ($need & (EML_HDR|EML_BDY));
1005 # do we need CRLF conversion?
1006 if ($need & CRLF_BREF) {
1007 push @op, $OP_CRLF_BREF;
1008 } elsif (my $crlf = ($need & (CRLF_HDR|CRLF_BDY))) {
1009 if ($crlf == (CRLF_HDR|CRLF_BDY)) {
1010 push @op, $OP_CRLF_BREF;
1011 } elsif ($need & CRLF_HDR) {
1012 push @op, $OP_CRLF_HDR;
1014 push @op, $OP_CRLF_BDY;
1018 $r[0] = $need & NEED_BLOB ? \&fetch_blob :
1019 ($need & NEED_SMSG ? \&fetch_smsg : \&fetch_uid);
1021 # r[1] = [ $key1, $cb1, $key2, $cb2, ... ]
1022 use sort 'stable'; # makes output more consistent
1023 $r[1] = [ map { ($_->[2], $_->[1]) } sort { $a->[0] <=> $b->[0] } @op ];
1027 sub cmd_uid_fetch ($$$$;@) {
1028 my ($self, $tag, $range_csv, @want) = @_;
1029 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1030 my ($cb, $ops, $partial) = fetch_compile(\@want);
1031 return "$tag $cb\r\n" unless $ops;
1033 # cb is one of fetch_blob, fetch_smsg, fetch_uid
1034 $range_csv = 'bad' if $range_csv !~ $valid_range;
1035 my $range_info = range_step($self, \$range_csv);
1036 return "$tag $range_info\r\n" if !ref($range_info);
1037 uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
1038 $self->long_response($cb, $tag, [], $range_info, $ops, $partial);
1041 sub cmd_fetch ($$$$;@) {
1042 my ($self, $tag, $range_csv, @want) = @_;
1043 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1044 my ($cb, $ops, $partial) = fetch_compile(\@want);
1045 return "$tag $cb\r\n" unless $ops;
1047 # cb is one of fetch_blob, fetch_smsg, fetch_uid
1048 $range_csv = 'bad' if $range_csv !~ $valid_range;
1049 msn_to_uid_range(msn2uid($self), $range_csv);
1050 my $range_info = range_step($self, \$range_csv);
1051 return "$tag $range_info\r\n" if !ref($range_info);
1052 uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
1053 $self->long_response($cb, $tag, [], $range_info, $ops, $partial);
1056 sub msn_convert ($$) {
1057 my ($self, $uids) = @_;
1058 my $adj = $self->{uid_base} + 1;
1059 my $uo2m = uo2m_extend($self, $uids->[-1]);
1060 $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
1061 $_ = $uo2m->[$_ - $adj] for @$uids;
1064 sub search_uid_range { # long_response
1065 my ($self, $tag, $sql, $range_info, $want_msn) = @_;
1067 if (defined(my $err = refill_uids($self, $uids, $range_info, $sql))) {
1068 $err ||= 'OK Search done';
1069 $self->write("\r\n$tag $err\r\n");
1072 msn_convert($self, $uids) if $want_msn;
1073 $self->msg_more(join(' ', '', @$uids));
1077 sub parse_imap_query ($$) {
1078 my ($self, $query) = @_;
1079 my $q = PublicInbox::IMAPsearchqp::parse($self, $query);
1081 my $max = $self->{ibx}->over(1)->max;
1083 uid_clamp($self, \$beg, \$max);
1084 $q->{range_info} = [ $beg, $max ];
1090 my ($self, $tag, $query, $want_msn) = @_;
1091 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1092 my $q = parse_imap_query($self, $query);
1093 return "$tag $q\r\n" if !ref($q);
1094 my ($sql, $range_info) = delete @$q{qw(sql range_info)};
1095 if (!scalar(keys %$q)) { # overview.sqlite3
1096 $self->msg_more('* SEARCH');
1097 $self->long_response(\&search_uid_range,
1098 $tag, $sql, $range_info, $want_msn);
1099 } elsif ($q = $q->{xap}) {
1100 my $srch = $self->{ibx}->isrch or
1101 return "$tag BAD search not available for mailbox\r\n";
1105 uid_range => $range_info
1107 my $mset = $srch->mset($q, $opt);
1108 my $uids = $srch->mset_to_artnums($mset, $opt);
1109 msn_convert($self, $uids) if scalar(@$uids) && $want_msn;
1110 "* SEARCH @$uids\r\n$tag OK Search done\r\n";
1112 "$tag BAD Error\r\n";
1116 sub cmd_uid_search ($$$) {
1117 my ($self, $tag, $query) = @_;
1118 search_common($self, $tag, $query);
1121 sub cmd_search ($$$;) {
1122 my ($self, $tag, $query) = @_;
1123 search_common($self, $tag, $query, 1);
1126 # returns 1 if we can continue, 0 if not due to buffered writes or disconnect
1127 sub process_line ($$) {
1128 my ($self, $l) = @_;
1130 # TODO: IMAP allows literals for big requests to upload messages
1131 # (which we don't support) but maybe some big search queries use it.
1132 # RFC 3501 9 (2) doesn't permit TAB or multiple SP
1133 my ($tag, $req, @args) = parse_line('[ \t]+', 0, $l);
1134 pop(@args) if (@args && !defined($args[-1]));
1135 if (@args && uc($req) eq 'UID') {
1136 $req .= "_".(shift @args);
1139 if (defined(my $idle_tag = $self->{-idle_tag})) {
1140 (uc($tag // '') eq 'DONE' && !defined($req)) ?
1141 idle_done($self, $tag) :
1142 "$idle_tag BAD expected DONE\r\n";
1143 } elsif (my $cmd = $self->can('cmd_'.lc($req // ''))) {
1144 if ($cmd == \&cmd_uid_search || $cmd == \&cmd_search) {
1145 # preserve user-supplied quotes for search
1146 (undef, @args) = split(/ search /i, $l, 2);
1148 $cmd->($self, $tag, @args);
1149 } else { # this is weird
1150 auth_challenge_ok($self) //
1152 ' BAD Error in IMAP command '.
1154 ": Unknown command\r\n";
1158 if ($err && $self->{sock}) {
1160 err($self, 'error from: %s (%s)', $l, $err);
1162 $res = "$tag BAD program fault - command not performed\r\n";
1164 return 0 unless defined $res;
1169 my ($self, $fmt, @args) = @_;
1170 printf { $self->{imapd}->{err} } $fmt."\n", @args;
1174 my ($self, $fmt, @args) = @_;
1175 printf { $self->{imapd}->{out} } $fmt."\n", @args;
1178 # callback used by PublicInbox::DS for any (e)poll (in/out/hup/err)
1182 return unless $self->flush_write && $self->{sock} && !$self->{long_cb};
1184 # only read more requests if we've drained the write buffer,
1185 # otherwise we can be buffering infinitely w/o backpressure
1187 my $rbuf = $self->{rbuf} // \(my $x = '');
1188 my $line = index($$rbuf, "\n");
1190 if (length($$rbuf) >= LINE_MAX) {
1191 $self->write(\"\* BAD request too long\r\n");
1192 return $self->close;
1194 $self->do_read($rbuf, LINE_MAX, length($$rbuf)) or
1195 return uo2m_hibernate($self);
1196 $line = index($$rbuf, "\n");
1198 $line = substr($$rbuf, 0, $line + 1, '');
1199 $line =~ s/\r?\n\z//s;
1200 return $self->close if $line =~ /[[:cntrl:]]/s;
1202 my $fd = fileno($self->{sock});
1203 my $r = eval { process_line($self, $line) };
1204 my $pending = $self->{wbuf} ? ' pending' : '';
1205 out($self, "[$fd] %s - %0.6f$pending - $r", $line, now() - $t0);
1207 return $self->close if $r < 0;
1208 $self->rbuf_idle($rbuf);
1210 # maybe there's more pipelined data, or we'll have
1211 # to register it for socket-readiness notifications
1212 $self->requeue unless $pending;
1216 sub cmd_compress ($$$) {
1217 my ($self, $tag, $alg) = @_;
1218 return "$tag BAD DEFLATE only\r\n" if uc($alg) ne "DEFLATE";
1219 return "$tag BAD COMPRESS active\r\n" if $self->compressed;
1221 # CRIME made TLS compression obsolete
1222 # return "$tag NO [COMPRESSIONACTIVE]\r\n" if $self->tls_compressed;
1224 PublicInbox::IMAPdeflate->enable($self) or return
1225 \"$tag BAD failed to activate compression\r\n";
1226 PublicInbox::DS::write($self, \"$tag OK DEFLATE active\r\n");
1231 sub cmd_starttls ($$) {
1232 my ($self, $tag) = @_;
1233 my $sock = $self->{sock} or return;
1234 if ($sock->can('stop_SSL') || $self->compressed) {
1235 return "$tag BAD TLS or compression already enabled\r\n";
1237 my $opt = $self->{imapd}->{accept_tls} or
1238 return "$tag BAD can not initiate TLS negotiation\r\n";
1239 $self->write(\"$tag OK begin TLS negotiation now\r\n");
1240 $self->{sock} = IO::Socket::SSL->start_SSL($sock, %$opt);
1241 $self->requeue if PublicInbox::DS::accept_tls_step($self);
1245 sub busy { # for graceful shutdown in PublicInbox::Daemon:
1247 if (defined($self->{-idle_tag})) {
1248 $self->write(\"* BYE server shutting down\r\n");
1249 return; # not busy anymore
1251 defined($self->{rbuf}) || defined($self->{wbuf}) ||
1252 !$self->write(\"* BYE server shutting down\r\n");
1257 if (my $ibx = delete $self->{ibx}) {
1258 stop_idle($self, $ibx);
1260 $self->SUPER::close; # PublicInbox::DS::close
1263 # we're read-only, so SELECT and EXAMINE do the same thing
1265 *cmd_select = \&cmd_examine;
1267 package PublicInbox::IMAP_preauth;
1268 our @ISA = qw(PublicInbox::IMAP);
1272 package PublicInbox::IMAPdeflate;
1273 use PublicInbox::DSdeflate;
1274 our @ISA = qw(PublicInbox::DSdeflate PublicInbox::IMAP);