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);
566 sub requeue_once ($) {
568 # COMPRESS users all share the same DEFLATE context.
569 # Flush it here to ensure clients don't see
573 # no recursion, schedule another call ASAP,
574 # but only after all pending writes are done.
576 my $new_size = push(@{$self->{wbuf}}, \&long_step);
578 # wbuf may be populated by $cb, no need to rearm if so:
579 $self->requeue if $new_size == 1;
583 my ($self, $smsg, $bref, $ops, $partial) = @_;
584 my $uid = $smsg->{num};
585 $self->msg_more('* '.uid2msn($self, $uid)." FETCH (UID $uid");
587 for (my $i = 0; $i < @$ops;) {
589 $ops->[$i++]->($self, $k, $smsg, $bref, $eml);
591 partial_emit($self, $partial, $eml) if $partial;
592 $self->msg_more(")\r\n");
595 sub fetch_blob_cb { # called by git->cat_async via ibx_async_cat
596 my ($bref, $oid, $type, $size, $fetch_arg) = @_;
597 my ($self, undef, $msgs, $range_info, $ops, $partial) = @$fetch_arg;
598 my $ibx = $self->{ibx} or return $self->close; # client disconnected
599 my $smsg = shift @$msgs or die 'BUG: no smsg';
600 if (!defined($oid)) {
601 # it's possible to have TOCTOU if an admin runs
602 # public-inbox-(edit|purge), just move onto the next message
603 warn "E: $smsg->{blob} missing in $ibx->{inboxdir}\n";
604 return requeue_once($self);
606 $smsg->{blob} eq $oid or die "BUG: $smsg->{blob} != $oid";
609 if (!$self->{wbuf} && (my $nxt = $msgs->[0])) {
610 $pre = ibx_async_prefetch($ibx, $nxt->{blob},
611 \&fetch_blob_cb, $fetch_arg);
613 fetch_run_ops($self, $smsg, $bref, $ops, $partial);
614 $pre ? $self->zflush : requeue_once($self);
618 my ($self, $k, undef, $bref) = @_;
619 $self->msg_more(" $k {" . length($$bref)."}\r\n");
620 $self->msg_more($$bref);
623 # Mail::IMAPClient::message_string cares about this by default,
624 # (->Ignoresizeerrors attribute). Admins are encouraged to
625 # --reindex for IMAP support, anyways.
626 sub emit_rfc822_size {
627 my ($self, $k, $smsg) = @_;
628 $self->msg_more(' RFC822.SIZE ' . $smsg->{bytes});
631 sub emit_internaldate {
632 my ($self, undef, $smsg) = @_;
633 $self->msg_more(' INTERNALDATE "'.$smsg->internaldate.'"');
636 sub emit_flags { $_[0]->msg_more(' FLAGS ()') }
639 my ($self, undef, undef, undef, $eml) = @_;
640 $self->msg_more(' ENVELOPE '.eml_envelope($eml));
643 sub emit_rfc822_header {
644 my ($self, $k, undef, undef, $eml) = @_;
645 $self->msg_more(" $k {".length(${$eml->{hdr}})."}\r\n");
646 $self->msg_more(${$eml->{hdr}});
649 # n.b. this is sorted to be after any emit_eml_new ops
650 sub emit_rfc822_text {
651 my ($self, $k, undef, $bref) = @_;
652 $self->msg_more(" $k {".length($$bref)."}\r\n");
653 $self->msg_more($$bref);
656 sub emit_bodystructure {
657 my ($self, undef, undef, undef, $eml) = @_;
658 $self->msg_more(' BODYSTRUCTURE '.fetch_body($eml, 1));
662 my ($self, undef, undef, undef, $eml) = @_;
663 $self->msg_more(' BODY '.fetch_body($eml));
666 # set $eml once ($_[4] == $eml, $_[3] == $bref)
667 sub op_eml_new { $_[4] = PublicInbox::Eml->new($_[3]) }
669 # s/From / fixes old bug from import (pre-a0c07cba0e5d8b6a)
671 ${$_[0]} =~ s/(?<!\r)\n/\r\n/sg;
672 ${$_[0]} =~ s/\A[\r\n]*From [^\r\n]*\r\n//s;
675 sub op_crlf_bref { to_crlf_full($_[3]) }
677 sub op_crlf_hdr { to_crlf_full($_[4]->{hdr}) }
679 sub op_crlf_bdy { ${$_[4]->{bdy}} =~ s/(?<!\r)\n/\r\n/sg if $_[4]->{bdy} }
681 sub uid_clamp ($$$) {
682 my ($self, $beg, $end) = @_;
683 my $uid_min = $self->{uid_base} + 1;
684 my $uid_end = $uid_min + UID_SLICE - 1;
685 $$beg = $uid_min if $$beg < $uid_min;
686 $$end = $uid_end if $$end > $uid_end;
689 sub range_step ($$) {
690 my ($self, $range_csv) = @_;
691 my ($beg, $end, $range);
692 if ($$range_csv =~ s/\A([^,]+),//) {
695 $range = $$range_csv;
698 my $uid_base = $self->{uid_base};
699 my $uid_end = $uid_base + UID_SLICE;
700 if ($range =~ /\A([0-9]+):([0-9]+)\z/) {
701 ($beg, $end) = ($1 + 0, $2 + 0);
702 uid_clamp($self, \$beg, \$end);
703 } elsif ($range =~ /\A([0-9]+):\*\z/) {
705 $end = $self->{ibx}->over(1)->max;
706 $end = $uid_end if $end > $uid_end;
707 $beg = $end if $beg > $end;
708 uid_clamp($self, \$beg, \$end);
709 } elsif ($range =~ /\A[0-9]+\z/) {
710 $beg = $end = $range + 0;
711 # just let the caller do an out-of-range query if a single
712 # UID is out-of-range
713 ++$beg if ($beg <= $uid_base || $end > $uid_end);
715 return 'BAD fetch range';
717 [ $beg, $end, $$range_csv ];
720 sub refill_range ($$$) {
721 my ($self, $msgs, $range_info) = @_;
722 my ($beg, $end, $range_csv) = @$range_info;
723 if (scalar(@$msgs = @{$self->{ibx}->over(1)->query_xover($beg, $end)})){
724 $range_info->[0] = $msgs->[-1]->{num} + 1;
727 return 'OK Fetch done' if !$range_csv;
728 my $next_range = range_step($self, \$range_csv);
729 return $next_range if !ref($next_range); # error
730 @$range_info = @$next_range;
731 undef; # keep looping
734 sub fetch_blob { # long_response
735 my ($self, $tag, $msgs, $range_info, $ops, $partial) = @_;
736 while (!@$msgs) { # rare
737 if (my $end = refill_range($self, $msgs, $range_info)) {
738 $self->write(\"$tag $end\r\n");
742 uo2m_extend($self, $msgs->[-1]->{num});
743 ibx_async_cat($self->{ibx}, $msgs->[0]->{blob},
744 \&fetch_blob_cb, \@_);
747 sub fetch_smsg { # long_response
748 my ($self, $tag, $msgs, $range_info, $ops) = @_;
749 while (!@$msgs) { # rare
750 if (my $end = refill_range($self, $msgs, $range_info)) {
751 $self->write(\"$tag $end\r\n");
755 uo2m_extend($self, $msgs->[-1]->{num});
756 fetch_run_ops($self, $_, undef, $ops) for @$msgs;
761 sub refill_uids ($$$;$) {
762 my ($self, $uids, $range_info, $sql) = @_;
763 my ($beg, $end, $range_csv) = @$range_info;
764 my $over = $self->{ibx}->over(1);
766 if (scalar(@$uids = @{$over->uid_range($beg, $end, $sql)})) {
767 $range_info->[0] = $uids->[-1] + 1; # update $beg
769 } elsif (!$range_csv) {
772 my $next_range = range_step($self, \$range_csv);
773 return $next_range if !ref($next_range); # error
774 ($beg, $end, $range_csv) = @$range_info = @$next_range;
780 sub fetch_uid { # long_response
781 my ($self, $tag, $uids, $range_info, $ops) = @_;
782 if (defined(my $err = refill_uids($self, $uids, $range_info))) {
783 $err ||= 'OK Fetch done';
784 $self->write("$tag $err\r\n");
787 my $adj = $self->{uid_base} + 1;
788 my $uo2m = uo2m_extend($self, $uids->[-1]);
789 $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
792 $self->msg_more("* $uo2m->[$_ - $adj] FETCH (UID $_");
793 for ($i = 0; $i < @$ops;) {
795 $ops->[$i++]->($self, $k);
797 $self->msg_more(")\r\n");
803 sub cmd_status ($$$;@) {
804 my ($self, $tag, $mailbox, @items) = @_;
805 return "$tag BAD no items\r\n" if !scalar(@items);
806 ($items[0] !~ s/\A\(//s || $items[-1] !~ s/\)\z//s) and
807 return "$tag BAD invalid args\r\n";
808 my ($ibx, $exists, $uidnext) = inbox_lookup($self, $mailbox);
809 return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
811 for my $it (@items) {
814 if ($it =~ /\A(?:MESSAGES|UNSEEN|RECENT)\z/) {
816 } elsif ($it eq 'UIDNEXT') {
818 } elsif ($it eq 'UIDVALIDITY') {
819 push @it, $ibx->{uidvalidity};
821 return "$tag BAD invalid item\r\n";
824 return "$tag BAD no items\r\n" if !@it;
825 "* STATUS $mailbox (".join(' ', @it).")\r\n" .
826 "$tag OK Status done\r\n";
829 my %patmap = ('*' => '.*', '%' => '[^\.]*');
830 sub cmd_list ($$$$) {
831 my ($self, $tag, $refname, $wildcard) = @_;
832 my $l = $self->{imapd}->{mailboxlist};
833 if ($refname eq '' && $wildcard eq '') {
834 # request for hierarchy delimiter
835 $l = [ qq[* LIST (\\Noselect) "." ""\r\n] ];
836 } elsif ($refname ne '' || $wildcard ne '*') {
837 $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!egi;
838 $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/is, @$l) ];
840 \(join('', @$l, "$tag OK List done\r\n"));
843 sub cmd_lsub ($$$$) {
844 my (undef, $tag) = @_; # same args as cmd_list
845 "$tag OK Lsub done\r\n";
848 sub eml_index_offs_i { # PublicInbox::Eml::each_part callback
850 my ($eml, undef, $idx) = @$p;
851 if ($idx && lc($eml->ct->{type}) eq 'multipart') {
852 $eml->{imap_bdy} = $eml->{bdy} // \'';
854 $all->{$idx} = $eml; # $idx => Eml
857 # prepares an index for BODY[$SECTION_IDX] fetches
858 sub eml_body_idx ($$) {
859 my ($eml, $section_idx) = @_;
860 my $idx = $eml->{imap_all_parts} // do {
862 $eml->each_part(\&eml_index_offs_i, $all, 0, 1);
863 # top-level of multipart, BODY[0] not allowed (nz-number)
865 $eml->{imap_all_parts} = $all;
867 $idx->{$section_idx};
870 # BODY[($SECTION_IDX)?(.$SECTION_NAME)?]<$offset.$bytes>
872 my ($eml, $section_idx, $section_name) = @_;
873 if (defined $section_idx) {
874 $eml = eml_body_idx($eml, $section_idx) or return;
876 if (defined $section_name) {
877 if ($section_name eq 'MIME') {
878 # RFC 3501 6.4.5 states:
879 # The MIME part specifier MUST be prefixed
880 # by one or more numeric part specifiers
881 return unless defined $section_idx;
882 return $eml->header_obj->as_string . "\r\n";
884 my $bdy = $eml->{bdy} // $eml->{imap_bdy} // \'';
885 $eml = PublicInbox::Eml->new($$bdy);
886 if ($section_name eq 'TEXT') {
887 return $eml->body_raw;
888 } elsif ($section_name eq 'HEADER') {
889 return $eml->header_obj->as_string . "\r\n";
891 die "BUG: bad section_name=$section_name";
894 ${$eml->{bdy} // $eml->{imap_bdy} // \''};
897 # similar to what's in PublicInbox::Eml::re_memo, but doesn't memoize
898 # to avoid OOM with malicious users
899 sub hdrs_regexp ($) {
901 my $names = join('|', map { "\Q$_" } split(/[ \t]+/, $hdrs));
902 qr/^(?:$names):[ \t]*[^\n]*\r?\n # 1st line
903 # continuation lines:
904 (?:[^:\n]*?[ \t]+[^\n]*\r?\n)*
908 # BODY[($SECTION_IDX.)?HEADER.FIELDS.NOT ($HDRS)]<$offset.$bytes>
909 sub partial_hdr_not {
910 my ($eml, $section_idx, $hdrs_re) = @_;
911 if (defined $section_idx) {
912 $eml = eml_body_idx($eml, $section_idx) or return;
914 my $str = $eml->header_obj->as_string;
915 $str =~ s/$hdrs_re//g;
916 $str =~ s/(?<!\r)\n/\r\n/sg;
920 # BODY[($SECTION_IDX.)?HEADER.FIELDS ($HDRS)]<$offset.$bytes>
921 sub partial_hdr_get {
922 my ($eml, $section_idx, $hdrs_re) = @_;
923 if (defined $section_idx) {
924 $eml = eml_body_idx($eml, $section_idx) or return;
926 my $str = $eml->header_obj->as_string;
927 $str = join('', ($str =~ m/($hdrs_re)/g));
928 $str =~ s/(?<!\r)\n/\r\n/sg;
932 sub partial_prepare ($$$$) {
933 my ($need, $partial, $want, $att) = @_;
935 # recombine [ "BODY[1.HEADER.FIELDS", "(foo", "bar)]" ]
936 # back to: "BODY[1.HEADER.FIELDS (foo bar)]"
937 return unless $att =~ /\ABODY\[/s;
938 until (rindex($att, ']') >= 0) {
939 my $next = shift @$want or return;
940 $att .= ' ' . uc($next);
942 if ($att =~ /\ABODY\[([0-9]+(?:\.[0-9]+)*)? # 1 - section_idx
943 (?:\.(HEADER|MIME|TEXT))? # 2 - section_name
944 \](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 3, 4
945 $partial->{$att} = [ \&partial_body, $1, $2, $3, $4 ];
946 $$need |= CRLF_BREF|EML_HDR|EML_BDY;
947 } elsif ($att =~ /\ABODY\[(?:([0-9]+(?:\.[0-9]+)*)\.)? # 1 - section_idx
948 (?:HEADER\.FIELDS(\.NOT)?)\x20 # 2
949 \(([A-Z0-9\-\x20]+)\) # 3 - hdrs
950 \](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 4 5
951 my $tmp = $partial->{$att} = [ $2 ? \&partial_hdr_not
954 $tmp->[2] = hdrs_regexp($3);
956 # don't emit CRLF_HDR instruction, here, partial_hdr_*
957 # will do CRLF conversion with only the extracted result
958 # and not waste time converting lines we don't care about.
965 sub partial_emit ($$$) {
966 my ($self, $partial, $eml) = @_;
968 my ($k, $cb, @args) = @$_;
969 my ($offset, $len) = splice(@args, -2);
970 # $cb is partial_body|partial_hdr_get|partial_hdr_not
971 my $str = $cb->($eml, @args) // '';
972 if (defined $offset) {
974 $str = substr($str, $offset, $len);
975 $k =~ s/\.$len>\z/>/ or warn
976 "BUG: unable to remove `.$len>' from `$k'";
978 $str = substr($str, $offset);
984 $self->msg_more(" $k {$len}\r\n");
985 $self->msg_more($str);
989 sub fetch_compile ($) {
991 if ($want->[0] =~ s/\A\(//s) {
992 $want->[-1] =~ s/\)\z//s or return 'BAD no rparen';
994 my (%partial, %seen, @op);
996 while (defined(my $att = shift @$want)) {
998 next if $att eq 'UID'; # always returned
999 $att =~ s/\ABODY\.PEEK\[/BODY\[/; # we're read-only
1000 my $x = $FETCH_ATT{$att};
1002 while (my ($k, $fl_cb) = each %$x) {
1003 next if $seen{$k}++;
1004 $need |= $fl_cb->[0];
1005 push @op, [ @$fl_cb, $k ];
1007 } elsif (!partial_prepare(\$need, \%partial, $want, $att)) {
1008 return "BAD param: $att";
1013 # stabilize partial order for consistency and ease-of-debugging:
1014 if (scalar keys %partial) {
1016 $r[2] = [ map { [ $_, @{$partial{$_}} ] } sort keys %partial ];
1019 push @op, $OP_EML_NEW if ($need & (EML_HDR|EML_BDY));
1021 # do we need CRLF conversion?
1022 if ($need & CRLF_BREF) {
1023 push @op, $OP_CRLF_BREF;
1024 } elsif (my $crlf = ($need & (CRLF_HDR|CRLF_BDY))) {
1025 if ($crlf == (CRLF_HDR|CRLF_BDY)) {
1026 push @op, $OP_CRLF_BREF;
1027 } elsif ($need & CRLF_HDR) {
1028 push @op, $OP_CRLF_HDR;
1030 push @op, $OP_CRLF_BDY;
1034 $r[0] = $need & NEED_BLOB ? \&fetch_blob :
1035 ($need & NEED_SMSG ? \&fetch_smsg : \&fetch_uid);
1037 # r[1] = [ $key1, $cb1, $key2, $cb2, ... ]
1038 use sort 'stable'; # makes output more consistent
1039 $r[1] = [ map { ($_->[2], $_->[1]) } sort { $a->[0] <=> $b->[0] } @op ];
1043 sub cmd_uid_fetch ($$$$;@) {
1044 my ($self, $tag, $range_csv, @want) = @_;
1045 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1046 my ($cb, $ops, $partial) = fetch_compile(\@want);
1047 return "$tag $cb\r\n" unless $ops;
1049 # cb is one of fetch_blob, fetch_smsg, fetch_uid
1050 $range_csv = 'bad' if $range_csv !~ $valid_range;
1051 my $range_info = range_step($self, \$range_csv);
1052 return "$tag $range_info\r\n" if !ref($range_info);
1053 uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
1054 long_response($self, $cb, $tag, [], $range_info, $ops, $partial);
1057 sub cmd_fetch ($$$$;@) {
1058 my ($self, $tag, $range_csv, @want) = @_;
1059 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1060 my ($cb, $ops, $partial) = fetch_compile(\@want);
1061 return "$tag $cb\r\n" unless $ops;
1063 # cb is one of fetch_blob, fetch_smsg, fetch_uid
1064 $range_csv = 'bad' if $range_csv !~ $valid_range;
1065 msn_to_uid_range(msn2uid($self), $range_csv);
1066 my $range_info = range_step($self, \$range_csv);
1067 return "$tag $range_info\r\n" if !ref($range_info);
1068 uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
1069 long_response($self, $cb, $tag, [], $range_info, $ops, $partial);
1072 sub msn_convert ($$) {
1073 my ($self, $uids) = @_;
1074 my $adj = $self->{uid_base} + 1;
1075 my $uo2m = uo2m_extend($self, $uids->[-1]);
1076 $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
1077 $_ = $uo2m->[$_ - $adj] for @$uids;
1080 sub search_uid_range { # long_response
1081 my ($self, $tag, $sql, $range_info, $want_msn) = @_;
1083 if (defined(my $err = refill_uids($self, $uids, $range_info, $sql))) {
1084 $err ||= 'OK Search done';
1085 $self->write("\r\n$tag $err\r\n");
1088 msn_convert($self, $uids) if $want_msn;
1089 $self->msg_more(join(' ', '', @$uids));
1093 sub parse_imap_query ($$) {
1094 my ($self, $query) = @_;
1095 my $q = PublicInbox::IMAPsearchqp::parse($self, $query);
1097 my $max = $self->{ibx}->over(1)->max;
1099 uid_clamp($self, \$beg, \$max);
1100 $q->{range_info} = [ $beg, $max ];
1106 my ($self, $tag, $query, $want_msn) = @_;
1107 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1108 my $q = parse_imap_query($self, $query);
1109 return "$tag $q\r\n" if !ref($q);
1110 my ($sql, $range_info) = delete @$q{qw(sql range_info)};
1111 if (!scalar(keys %$q)) { # overview.sqlite3
1112 $self->msg_more('* SEARCH');
1113 long_response($self, \&search_uid_range,
1114 $tag, $sql, $range_info, $want_msn);
1115 } elsif ($q = $q->{xap}) {
1116 my $srch = $self->{ibx}->isrch or
1117 return "$tag BAD search not available for mailbox\r\n";
1121 uid_range => $range_info
1123 my $mset = $srch->mset($q, $opt);
1124 my $uids = $srch->mset_to_artnums($mset, $opt);
1125 msn_convert($self, $uids) if scalar(@$uids) && $want_msn;
1126 "* SEARCH @$uids\r\n$tag OK Search done\r\n";
1128 "$tag BAD Error\r\n";
1132 sub cmd_uid_search ($$$) {
1133 my ($self, $tag, $query) = @_;
1134 search_common($self, $tag, $query);
1137 sub cmd_search ($$$;) {
1138 my ($self, $tag, $query) = @_;
1139 search_common($self, $tag, $query, 1);
1142 # returns 1 if we can continue, 0 if not due to buffered writes or disconnect
1143 sub process_line ($$) {
1144 my ($self, $l) = @_;
1146 # TODO: IMAP allows literals for big requests to upload messages
1147 # (which we don't support) but maybe some big search queries use it.
1148 # RFC 3501 9 (2) doesn't permit TAB or multiple SP
1149 my ($tag, $req, @args) = parse_line('[ \t]+', 0, $l);
1150 pop(@args) if (@args && !defined($args[-1]));
1151 if (@args && uc($req) eq 'UID') {
1152 $req .= "_".(shift @args);
1155 if (defined(my $idle_tag = $self->{-idle_tag})) {
1156 (uc($tag // '') eq 'DONE' && !defined($req)) ?
1157 idle_done($self, $tag) :
1158 "$idle_tag BAD expected DONE\r\n";
1159 } elsif (my $cmd = $self->can('cmd_'.lc($req // ''))) {
1160 if ($cmd == \&cmd_uid_search || $cmd == \&cmd_search) {
1161 # preserve user-supplied quotes for search
1162 (undef, @args) = split(/ search /i, $l, 2);
1164 $cmd->($self, $tag, @args);
1165 } else { # this is weird
1166 auth_challenge_ok($self) //
1168 ' BAD Error in IMAP command '.
1170 ": Unknown command\r\n";
1174 if ($err && $self->{sock}) {
1176 err($self, 'error from: %s (%s)', $l, $err);
1178 $res = "$tag BAD program fault - command not performed\r\n";
1180 return 0 unless defined $res;
1186 # wbuf is unset or empty, here; {long} may add to it
1187 my ($fd, $cb, $t0, @args) = @{$self->{long_cb}};
1188 my $more = eval { $cb->($self, @args) };
1189 if ($@ || !$self->{sock}) { # something bad happened...
1190 delete $self->{long_cb};
1191 my $elapsed = now() - $t0;
1194 "%s during long response[$fd] - %0.6f",
1197 out($self, " deferred[$fd] aborted - %0.6f", $elapsed);
1199 } elsif ($more) { # $self->{wbuf}:
1200 # control passed to ibx_async_cat if $more == \undef
1201 requeue_once($self) if !ref($more);
1202 } else { # all done!
1203 delete $self->{long_cb};
1204 my $elapsed = now() - $t0;
1205 my $fd = fileno($self->{sock});
1206 out($self, " deferred[$fd] done - %0.6f", $elapsed);
1207 my $wbuf = $self->{wbuf}; # do NOT autovivify
1209 $self->requeue unless $wbuf && @$wbuf;
1214 my ($self, $fmt, @args) = @_;
1215 printf { $self->{imapd}->{err} } $fmt."\n", @args;
1219 my ($self, $fmt, @args) = @_;
1220 printf { $self->{imapd}->{out} } $fmt."\n", @args;
1223 sub long_response ($$;@) {
1224 my ($self, $cb, @args) = @_; # cb returns true if more, false if done
1226 my $sock = $self->{sock} or return;
1227 # make sure we disable reading during a long response,
1228 # clients should not be sending us stuff and making us do more
1229 # work while we are stream a response to them
1230 $self->{long_cb} = [ fileno($sock), $cb, now(), @args ];
1231 long_step($self); # kick off!
1235 # callback used by PublicInbox::DS for any (e)poll (in/out/hup/err)
1239 return unless $self->flush_write && $self->{sock} && !$self->{long_cb};
1241 # only read more requests if we've drained the write buffer,
1242 # otherwise we can be buffering infinitely w/o backpressure
1244 my $rbuf = $self->{rbuf} // \(my $x = '');
1245 my $line = index($$rbuf, "\n");
1247 if (length($$rbuf) >= LINE_MAX) {
1248 $self->write(\"\* BAD request too long\r\n");
1249 return $self->close;
1251 $self->do_read($rbuf, LINE_MAX, length($$rbuf)) or
1252 return uo2m_hibernate($self);
1253 $line = index($$rbuf, "\n");
1255 $line = substr($$rbuf, 0, $line + 1, '');
1256 $line =~ s/\r?\n\z//s;
1257 return $self->close if $line =~ /[[:cntrl:]]/s;
1259 my $fd = fileno($self->{sock});
1260 my $r = eval { process_line($self, $line) };
1261 my $pending = $self->{wbuf} ? ' pending' : '';
1262 out($self, "[$fd] %s - %0.6f$pending - $r", $line, now() - $t0);
1264 return $self->close if $r < 0;
1265 $self->rbuf_idle($rbuf);
1267 # maybe there's more pipelined data, or we'll have
1268 # to register it for socket-readiness notifications
1269 $self->requeue unless $pending;
1272 sub compressed { undef }
1274 sub zflush {} # overridden by IMAPdeflate
1277 sub cmd_compress ($$$) {
1278 my ($self, $tag, $alg) = @_;
1279 return "$tag BAD DEFLATE only\r\n" if uc($alg) ne "DEFLATE";
1280 return "$tag BAD COMPRESS active\r\n" if $self->compressed;
1282 # CRIME made TLS compression obsolete
1283 # return "$tag NO [COMPRESSIONACTIVE]\r\n" if $self->tls_compressed;
1285 PublicInbox::IMAPdeflate->enable($self, $tag);
1290 sub cmd_starttls ($$) {
1291 my ($self, $tag) = @_;
1292 my $sock = $self->{sock} or return;
1293 if ($sock->can('stop_SSL') || $self->compressed) {
1294 return "$tag BAD TLS or compression already enabled\r\n";
1296 my $opt = $self->{imapd}->{accept_tls} or
1297 return "$tag BAD can not initiate TLS negotiation\r\n";
1298 $self->write(\"$tag OK begin TLS negotiation now\r\n");
1299 $self->{sock} = IO::Socket::SSL->start_SSL($sock, %$opt);
1300 $self->requeue if PublicInbox::DS::accept_tls_step($self);
1304 sub busy { # for graceful shutdown in PublicInbox::Daemon:
1306 if (defined($self->{-idle_tag})) {
1307 $self->write(\"* BYE server shutting down\r\n");
1308 return; # not busy anymore
1310 defined($self->{rbuf}) || defined($self->{wbuf}) ||
1311 !$self->write(\"* BYE server shutting down\r\n");
1316 if (my $ibx = delete $self->{ibx}) {
1317 stop_idle($self, $ibx);
1319 $self->SUPER::close; # PublicInbox::DS::close
1322 # we're read-only, so SELECT and EXAMINE do the same thing
1324 *cmd_select = \&cmd_examine;
1326 package PublicInbox::IMAP_preauth;
1327 our @ISA = qw(PublicInbox::IMAP);