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}->{ssl_ctx_opt}) {
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) = @_;
355 my $mb_top = $ibx->{newsgroup} // return;
356 my $mailboxes = $imapd->{mailboxes};
358 for (my $i = int($ibx->art_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 ($$;$) {
371 my ($self, $mailbox, $examine) = @_;
372 my ($ibx, $exists, $uidmax, $uid_base) = (undef, 0, 0, 0);
373 $mailbox = lc $mailbox;
374 $ibx = $self->{imapd}->{mailboxes}->{$mailbox} or return;
375 my $over = $ibx->over(1);
376 if ($over != $ibx) { # not a dummy
377 $mailbox =~ /\.([0-9]+)\z/ or
378 die "BUG: unexpected dummy mailbox: $mailbox\n";
379 $uid_base = $1 * UID_SLICE;
381 $uidmax = $ibx->mm->num_highwater // 0;
383 $self->{uid_base} = $uid_base;
385 $self->{uo2m} = uo2m_ary_new($self, \$exists);
387 my $uid_end = $uid_base + UID_SLICE;
388 $exists = $over->imap_exists($uid_base, $uid_end);
390 delete $ibx->{-art_max};
391 ensure_slices_exist($self->{imapd}, $ibx);
394 $self->{uid_base} = $uid_base;
396 delete $self->{uo2m};
398 # if "INBOX.foo.bar" is selected and "INBOX.foo.bar.0",
399 # check for new UID ranges (e.g. "INBOX.foo.bar.1")
400 if (my $ibx = $self->{imapd}->{mailboxes}->{"$mailbox.0"}) {
401 delete $ibx->{-art_max};
402 ensure_slices_exist($self->{imapd}, $ibx);
405 ($ibx, $exists, $uidmax + 1, $uid_base);
408 sub cmd_examine ($$$) {
409 my ($self, $tag, $mailbox) = @_;
410 # XXX: do we need this? RFC 5162/7162
411 my $ret = $self->{ibx} ? "* OK [CLOSED] previous closed\r\n" : '';
412 my ($ibx, $exists, $uidnext, $base) = inbox_lookup($self, $mailbox, 1);
413 return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
418 * OK [PERMANENTFLAGS ()] Read-only mailbox\r
419 * OK [UNSEEN $exists]\r
420 * OK [UIDNEXT $uidnext]\r
421 * OK [UIDVALIDITY $ibx->{uidvalidity}]\r
422 $tag OK [READ-ONLY] EXAMINE/SELECT done\r
430 } elsif ($v =~ /[{"\r\n%*\\\[]/) { # literal string
431 '{' . length($v) . "}\r\n" . $v;
432 } else { # quoted string
437 sub addr_envelope ($$;$) {
438 my ($eml, $x, $y) = @_;
439 my $v = $eml->header_raw($x) //
440 ($y ? $eml->header_raw($y) : undef) // return 'NIL';
442 my @x = $Address->parse($v) or return 'NIL';
444 map { '(' . join(' ',
445 _esc($_->name), 'NIL',
446 _esc($_->user), _esc($_->host)
452 sub eml_envelope ($) {
455 _esc($eml->header_raw('Date')),
456 _esc($eml->header_raw('Subject')),
457 addr_envelope($eml, 'From'),
458 addr_envelope($eml, 'Sender', 'From'),
459 addr_envelope($eml, 'Reply-To', 'From'),
460 addr_envelope($eml, 'To'),
461 addr_envelope($eml, 'Cc'),
462 addr_envelope($eml, 'Bcc'),
463 _esc($eml->header_raw('In-Reply-To')),
464 _esc($eml->header_raw('Message-ID')),
470 if ($hash && scalar keys %$hash) {
471 $hash = [ %$hash ]; # flatten hash into 1-dimensional array
472 '(' . join(' ', map { _esc($_) } @$hash) . ')';
478 sub body_disposition ($) {
480 my $cd = $eml->header_raw('Content-Disposition') or return 'NIL';
481 $cd = parse_content_disposition($cd);
482 my $buf = '('._esc($cd->{type});
483 $buf .= ' ' . _esc_hash($cd->{attributes});
487 sub body_leaf ($$;$) {
488 my ($eml, $structure, $hold) = @_;
490 $eml->{is_submsg} and # parent was a message/(rfc822|news|global)
491 $buf .= eml_envelope($eml). ' ';
493 $buf .= '('._esc($ct->{type}).' ';
494 $buf .= _esc($ct->{subtype});
495 $buf .= ' ' . _esc_hash($ct->{attributes});
496 $buf .= ' ' . _esc($eml->header_raw('Content-ID'));
497 $buf .= ' ' . _esc($eml->header_raw('Content-Description'));
498 my $cte = $eml->header_raw('Content-Transfer-Encoding') // '7bit';
499 $buf .= ' ' . _esc($cte);
500 $buf .= ' ' . $eml->{imap_body_len};
501 $buf .= ' '.($eml->body_raw =~ tr/\n/\n/) if lc($ct->{type}) eq 'text';
503 # for message/(rfc822|global|news), $hold[0] should have envelope
504 $buf .= ' ' . (@$hold ? join('', @$hold) : 'NIL') if $hold;
507 $buf .= ' '._esc($eml->header_raw('Content-MD5'));
508 $buf .= ' '. body_disposition($eml);
509 $buf .= ' '._esc($eml->header_raw('Content-Language'));
510 $buf .= ' '._esc($eml->header_raw('Content-Location'));
515 sub body_parent ($$$) {
516 my ($eml, $structure, $hold) = @_;
518 my $type = lc($ct->{type});
519 if ($type eq 'multipart') {
521 $buf .= @$hold ? join('', @$hold) : 'NIL';
522 $buf .= ' '._esc($ct->{subtype});
524 $buf .= ' '._esc_hash($ct->{attributes});
525 $buf .= ' '.body_disposition($eml);
526 $buf .= ' '._esc($eml->header_raw('Content-Language'));
527 $buf .= ' '._esc($eml->header_raw('Content-Location'));
531 } else { # message/(rfc822|global|news)
532 @$hold = (body_leaf($eml, $structure, $hold));
536 # this is gross, but we need to process the parent part AFTER
537 # the child parts are done
538 sub bodystructure_prep {
540 my ($eml, $depth) = @$p; # ignore idx
541 # set length here, as $eml->{bdy} gets deleted for message/rfc822
542 $eml->{imap_body_len} = length($eml->body_raw);
543 push @$q, $eml, $depth;
546 # for FETCH BODY and FETCH BODYSTRUCTURE
547 sub fetch_body ($;$) {
548 my ($eml, $structure) = @_;
550 $eml->each_part(\&bodystructure_prep, \@q, 0, 1);
554 my ($part, $depth) = splice(@q, -2);
555 my $is_mp_parent = $depth == ($cur_depth - 1);
559 body_parent($part, $structure, \@hold);
561 unshift @hold, body_leaf($part, $structure);
568 my ($self, $smsg, $bref, $ops, $partial) = @_;
569 my $uid = $smsg->{num};
570 $self->msg_more('* '.uid2msn($self, $uid)." FETCH (UID $uid");
572 for (my $i = 0; $i < @$ops;) {
574 $ops->[$i++]->($self, $k, $smsg, $bref, $eml);
576 partial_emit($self, $partial, $eml) if $partial;
577 $self->msg_more(")\r\n");
580 sub fetch_blob_cb { # called by git->cat_async via ibx_async_cat
581 my ($bref, $oid, $type, $size, $fetch_arg) = @_;
582 my ($self, undef, $msgs, $range_info, $ops, $partial) = @$fetch_arg;
583 my $ibx = $self->{ibx} or return $self->close; # client disconnected
584 my $smsg = shift @$msgs or die 'BUG: no smsg';
585 if (!defined($oid)) {
586 # it's possible to have TOCTOU if an admin runs
587 # public-inbox-(edit|purge), just move onto the next message
588 warn "E: $smsg->{blob} missing in $ibx->{inboxdir}\n";
589 return $self->requeue_once;
591 $smsg->{blob} eq $oid or die "BUG: $smsg->{blob} != $oid";
594 if (!$self->{wbuf} && (my $nxt = $msgs->[0])) {
595 $pre = ibx_async_prefetch($ibx, $nxt->{blob},
596 \&fetch_blob_cb, $fetch_arg);
598 fetch_run_ops($self, $smsg, $bref, $ops, $partial);
599 $pre ? $self->dflush : $self->requeue_once;
603 my ($self, $k, undef, $bref) = @_;
604 $self->msg_more(" $k {" . length($$bref)."}\r\n");
605 $self->msg_more($$bref);
608 # Mail::IMAPClient::message_string cares about this by default,
609 # (->Ignoresizeerrors attribute). Admins are encouraged to
610 # --reindex for IMAP support, anyways.
611 sub emit_rfc822_size {
612 my ($self, $k, $smsg) = @_;
613 $self->msg_more(' RFC822.SIZE ' . $smsg->{bytes});
616 sub emit_internaldate {
617 my ($self, undef, $smsg) = @_;
618 $self->msg_more(' INTERNALDATE "'.$smsg->internaldate.'"');
621 sub emit_flags { $_[0]->msg_more(' FLAGS ()') }
624 my ($self, undef, undef, undef, $eml) = @_;
625 $self->msg_more(' ENVELOPE '.eml_envelope($eml));
628 sub emit_rfc822_header {
629 my ($self, $k, undef, undef, $eml) = @_;
630 $self->msg_more(" $k {".length(${$eml->{hdr}})."}\r\n");
631 $self->msg_more(${$eml->{hdr}});
634 # n.b. this is sorted to be after any emit_eml_new ops
635 sub emit_rfc822_text {
636 my ($self, $k, undef, $bref) = @_;
637 $self->msg_more(" $k {".length($$bref)."}\r\n");
638 $self->msg_more($$bref);
641 sub emit_bodystructure {
642 my ($self, undef, undef, undef, $eml) = @_;
643 $self->msg_more(' BODYSTRUCTURE '.fetch_body($eml, 1));
647 my ($self, undef, undef, undef, $eml) = @_;
648 $self->msg_more(' BODY '.fetch_body($eml));
651 # set $eml once ($_[4] == $eml, $_[3] == $bref)
652 sub op_eml_new { $_[4] = PublicInbox::Eml->new($_[3]) }
654 # s/From / fixes old bug from import (pre-a0c07cba0e5d8b6a)
656 ${$_[0]} =~ s/(?<!\r)\n/\r\n/sg;
657 ${$_[0]} =~ s/\A[\r\n]*From [^\r\n]*\r\n//s;
660 sub op_crlf_bref { to_crlf_full($_[3]) }
662 sub op_crlf_hdr { to_crlf_full($_[4]->{hdr}) }
664 sub op_crlf_bdy { ${$_[4]->{bdy}} =~ s/(?<!\r)\n/\r\n/sg if $_[4]->{bdy} }
666 sub uid_clamp ($$$) {
667 my ($self, $beg, $end) = @_;
668 my $uid_min = $self->{uid_base} + 1;
669 my $uid_end = $uid_min + UID_SLICE - 1;
670 $$beg = $uid_min if $$beg < $uid_min;
671 $$end = $uid_end if $$end > $uid_end;
674 sub range_step ($$) {
675 my ($self, $range_csv) = @_;
676 my ($beg, $end, $range);
677 if ($$range_csv =~ s/\A([^,]+),//) {
680 $range = $$range_csv;
683 my $uid_base = $self->{uid_base};
684 my $uid_end = $uid_base + UID_SLICE;
685 if ($range =~ /\A([0-9]+):([0-9]+)\z/) {
686 ($beg, $end) = ($1 + 0, $2 + 0);
687 uid_clamp($self, \$beg, \$end);
688 } elsif ($range =~ /\A([0-9]+):\*\z/) {
690 $end = $self->{ibx}->over(1)->max;
691 $end = $uid_end if $end > $uid_end;
692 $beg = $end if $beg > $end;
693 uid_clamp($self, \$beg, \$end);
694 } elsif ($range =~ /\A[0-9]+\z/) {
695 $beg = $end = $range + 0;
696 # just let the caller do an out-of-range query if a single
697 # UID is out-of-range
698 ++$beg if ($beg <= $uid_base || $end > $uid_end);
700 return 'BAD fetch range';
702 [ $beg, $end, $$range_csv ];
705 sub refill_range ($$$) {
706 my ($self, $msgs, $range_info) = @_;
707 my ($beg, $end, $range_csv) = @$range_info;
708 if (scalar(@$msgs = @{$self->{ibx}->over(1)->query_xover($beg, $end)})){
709 $range_info->[0] = $msgs->[-1]->{num} + 1;
712 return 'OK Fetch done' if !$range_csv;
713 my $next_range = range_step($self, \$range_csv);
714 return $next_range if !ref($next_range); # error
715 @$range_info = @$next_range;
716 undef; # keep looping
719 sub fetch_blob { # long_response
720 my ($self, $tag, $msgs, $range_info, $ops, $partial) = @_;
721 while (!@$msgs) { # rare
722 if (my $end = refill_range($self, $msgs, $range_info)) {
723 $self->write(\"$tag $end\r\n");
727 uo2m_extend($self, $msgs->[-1]->{num});
728 ibx_async_cat($self->{ibx}, $msgs->[0]->{blob},
729 \&fetch_blob_cb, \@_);
732 sub fetch_smsg { # long_response
733 my ($self, $tag, $msgs, $range_info, $ops) = @_;
734 while (!@$msgs) { # rare
735 if (my $end = refill_range($self, $msgs, $range_info)) {
736 $self->write(\"$tag $end\r\n");
740 uo2m_extend($self, $msgs->[-1]->{num});
741 fetch_run_ops($self, $_, undef, $ops) for @$msgs;
746 sub refill_uids ($$$;$) {
747 my ($self, $uids, $range_info, $sql) = @_;
748 my ($beg, $end, $range_csv) = @$range_info;
749 my $over = $self->{ibx}->over(1);
751 if (scalar(@$uids = @{$over->uid_range($beg, $end, $sql)})) {
752 $range_info->[0] = $uids->[-1] + 1; # update $beg
754 } elsif (!$range_csv) {
757 my $next_range = range_step($self, \$range_csv);
758 return $next_range if !ref($next_range); # error
759 ($beg, $end, $range_csv) = @$range_info = @$next_range;
765 sub fetch_uid { # long_response
766 my ($self, $tag, $uids, $range_info, $ops) = @_;
767 if (defined(my $err = refill_uids($self, $uids, $range_info))) {
768 $err ||= 'OK Fetch done';
769 $self->write("$tag $err\r\n");
772 my $adj = $self->{uid_base} + 1;
773 my $uo2m = uo2m_extend($self, $uids->[-1]);
774 $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
777 $self->msg_more("* $uo2m->[$_ - $adj] FETCH (UID $_");
778 for ($i = 0; $i < @$ops;) {
780 $ops->[$i++]->($self, $k);
782 $self->msg_more(")\r\n");
788 sub cmd_status ($$$;@) {
789 my ($self, $tag, $mailbox, @items) = @_;
790 return "$tag BAD no items\r\n" if !scalar(@items);
791 ($items[0] !~ s/\A\(//s || $items[-1] !~ s/\)\z//s) and
792 return "$tag BAD invalid args\r\n";
793 my ($ibx, $exists, $uidnext) = inbox_lookup($self, $mailbox);
794 return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
796 for my $it (@items) {
799 if ($it =~ /\A(?:MESSAGES|UNSEEN|RECENT)\z/) {
801 } elsif ($it eq 'UIDNEXT') {
803 } elsif ($it eq 'UIDVALIDITY') {
804 push @it, $ibx->{uidvalidity};
806 return "$tag BAD invalid item\r\n";
809 return "$tag BAD no items\r\n" if !@it;
810 "* STATUS $mailbox (".join(' ', @it).")\r\n" .
811 "$tag OK Status done\r\n";
814 my %patmap = ('*' => '.*', '%' => '[^\.]*');
815 sub cmd_list ($$$$) {
816 my ($self, $tag, $refname, $wildcard) = @_;
817 my $l = $self->{imapd}->{mailboxlist};
818 if ($refname eq '' && $wildcard eq '') {
819 # request for hierarchy delimiter
820 $l = [ qq[* LIST (\\Noselect) "." ""\r\n] ];
821 } elsif ($refname ne '' || $wildcard ne '*') {
822 $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!egi;
823 $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/is, @$l) ];
825 \(join('', @$l, "$tag OK List done\r\n"));
828 sub cmd_lsub ($$$$) {
829 my (undef, $tag) = @_; # same args as cmd_list
830 "$tag OK Lsub done\r\n";
833 sub eml_index_offs_i { # PublicInbox::Eml::each_part callback
835 my ($eml, undef, $idx) = @$p;
836 if ($idx && lc($eml->ct->{type}) eq 'multipart') {
837 $eml->{imap_bdy} = $eml->{bdy} // \'';
839 $all->{$idx} = $eml; # $idx => Eml
842 # prepares an index for BODY[$SECTION_IDX] fetches
843 sub eml_body_idx ($$) {
844 my ($eml, $section_idx) = @_;
845 my $idx = $eml->{imap_all_parts} // do {
847 $eml->each_part(\&eml_index_offs_i, $all, 0, 1);
848 # top-level of multipart, BODY[0] not allowed (nz-number)
850 $eml->{imap_all_parts} = $all;
852 $idx->{$section_idx};
855 # BODY[($SECTION_IDX)?(.$SECTION_NAME)?]<$offset.$bytes>
857 my ($eml, $section_idx, $section_name) = @_;
858 if (defined $section_idx) {
859 $eml = eml_body_idx($eml, $section_idx) or return;
861 if (defined $section_name) {
862 if ($section_name eq 'MIME') {
863 # RFC 3501 6.4.5 states:
864 # The MIME part specifier MUST be prefixed
865 # by one or more numeric part specifiers
866 return unless defined $section_idx;
867 return $eml->header_obj->as_string . "\r\n";
869 my $bdy = $eml->{bdy} // $eml->{imap_bdy} // \'';
870 $eml = PublicInbox::Eml->new($$bdy);
871 if ($section_name eq 'TEXT') {
872 return $eml->body_raw;
873 } elsif ($section_name eq 'HEADER') {
874 return $eml->header_obj->as_string . "\r\n";
876 die "BUG: bad section_name=$section_name";
879 ${$eml->{bdy} // $eml->{imap_bdy} // \''};
882 # similar to what's in PublicInbox::Eml::re_memo, but doesn't memoize
883 # to avoid OOM with malicious users
884 sub hdrs_regexp ($) {
886 my $names = join('|', map { "\Q$_" } split(/[ \t]+/, $hdrs));
887 qr/^(?:$names):[ \t]*[^\n]*\r?\n # 1st line
888 # continuation lines:
889 (?:[^:\n]*?[ \t]+[^\n]*\r?\n)*
893 # BODY[($SECTION_IDX.)?HEADER.FIELDS.NOT ($HDRS)]<$offset.$bytes>
894 sub partial_hdr_not {
895 my ($eml, $section_idx, $hdrs_re) = @_;
896 if (defined $section_idx) {
897 $eml = eml_body_idx($eml, $section_idx) or return;
899 my $str = $eml->header_obj->as_string;
900 $str =~ s/$hdrs_re//g;
901 $str =~ s/(?<!\r)\n/\r\n/sg;
905 # BODY[($SECTION_IDX.)?HEADER.FIELDS ($HDRS)]<$offset.$bytes>
906 sub partial_hdr_get {
907 my ($eml, $section_idx, $hdrs_re) = @_;
908 if (defined $section_idx) {
909 $eml = eml_body_idx($eml, $section_idx) or return;
911 my $str = $eml->header_obj->as_string;
912 $str = join('', ($str =~ m/($hdrs_re)/g));
913 $str =~ s/(?<!\r)\n/\r\n/sg;
917 sub partial_prepare ($$$$) {
918 my ($need, $partial, $want, $att) = @_;
920 # recombine [ "BODY[1.HEADER.FIELDS", "(foo", "bar)]" ]
921 # back to: "BODY[1.HEADER.FIELDS (foo bar)]"
922 return unless $att =~ /\ABODY\[/s;
923 until (rindex($att, ']') >= 0) {
924 my $next = shift @$want or return;
925 $att .= ' ' . uc($next);
927 if ($att =~ /\ABODY\[([0-9]+(?:\.[0-9]+)*)? # 1 - section_idx
928 (?:\.(HEADER|MIME|TEXT))? # 2 - section_name
929 \](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 3, 4
930 $partial->{$att} = [ \&partial_body, $1, $2, $3, $4 ];
931 $$need |= CRLF_BREF|EML_HDR|EML_BDY;
932 } elsif ($att =~ /\ABODY\[(?:([0-9]+(?:\.[0-9]+)*)\.)? # 1 - section_idx
933 (?:HEADER\.FIELDS(\.NOT)?)\x20 # 2
934 \(([A-Z0-9\-\x20]+)\) # 3 - hdrs
935 \](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 4 5
936 my $tmp = $partial->{$att} = [ $2 ? \&partial_hdr_not
939 $tmp->[2] = hdrs_regexp($3);
941 # don't emit CRLF_HDR instruction, here, partial_hdr_*
942 # will do CRLF conversion with only the extracted result
943 # and not waste time converting lines we don't care about.
950 sub partial_emit ($$$) {
951 my ($self, $partial, $eml) = @_;
953 my ($k, $cb, @args) = @$_;
954 my ($offset, $len) = splice(@args, -2);
955 # $cb is partial_body|partial_hdr_get|partial_hdr_not
956 my $str = $cb->($eml, @args) // '';
957 if (defined $offset) {
959 $str = substr($str, $offset, $len);
960 $k =~ s/\.$len>\z/>/ or warn
961 "BUG: unable to remove `.$len>' from `$k'";
963 $str = substr($str, $offset);
969 $self->msg_more(" $k {$len}\r\n");
970 $self->msg_more($str);
974 sub fetch_compile ($) {
976 if ($want->[0] =~ s/\A\(//s) {
977 $want->[-1] =~ s/\)\z//s or return 'BAD no rparen';
979 my (%partial, %seen, @op);
981 while (defined(my $att = shift @$want)) {
983 next if $att eq 'UID'; # always returned
984 $att =~ s/\ABODY\.PEEK\[/BODY\[/; # we're read-only
985 my $x = $FETCH_ATT{$att};
987 while (my ($k, $fl_cb) = each %$x) {
989 $need |= $fl_cb->[0];
990 push @op, [ @$fl_cb, $k ];
992 } elsif (!partial_prepare(\$need, \%partial, $want, $att)) {
993 return "BAD param: $att";
998 # stabilize partial order for consistency and ease-of-debugging:
999 if (scalar keys %partial) {
1001 $r[2] = [ map { [ $_, @{$partial{$_}} ] } sort keys %partial ];
1004 push @op, $OP_EML_NEW if ($need & (EML_HDR|EML_BDY));
1006 # do we need CRLF conversion?
1007 if ($need & CRLF_BREF) {
1008 push @op, $OP_CRLF_BREF;
1009 } elsif (my $crlf = ($need & (CRLF_HDR|CRLF_BDY))) {
1010 if ($crlf == (CRLF_HDR|CRLF_BDY)) {
1011 push @op, $OP_CRLF_BREF;
1012 } elsif ($need & CRLF_HDR) {
1013 push @op, $OP_CRLF_HDR;
1015 push @op, $OP_CRLF_BDY;
1019 $r[0] = $need & NEED_BLOB ? \&fetch_blob :
1020 ($need & NEED_SMSG ? \&fetch_smsg : \&fetch_uid);
1022 # r[1] = [ $key1, $cb1, $key2, $cb2, ... ]
1023 use sort 'stable'; # makes output more consistent
1024 $r[1] = [ map { ($_->[2], $_->[1]) } sort { $a->[0] <=> $b->[0] } @op ];
1028 sub cmd_uid_fetch ($$$$;@) {
1029 my ($self, $tag, $range_csv, @want) = @_;
1030 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1031 my ($cb, $ops, $partial) = fetch_compile(\@want);
1032 return "$tag $cb\r\n" unless $ops;
1034 # cb is one of fetch_blob, fetch_smsg, fetch_uid
1035 $range_csv = 'bad' if $range_csv !~ $valid_range;
1036 my $range_info = range_step($self, \$range_csv);
1037 return "$tag $range_info\r\n" if !ref($range_info);
1038 uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
1039 $self->long_response($cb, $tag, [], $range_info, $ops, $partial);
1042 sub cmd_fetch ($$$$;@) {
1043 my ($self, $tag, $range_csv, @want) = @_;
1044 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1045 my ($cb, $ops, $partial) = fetch_compile(\@want);
1046 return "$tag $cb\r\n" unless $ops;
1048 # cb is one of fetch_blob, fetch_smsg, fetch_uid
1049 $range_csv = 'bad' if $range_csv !~ $valid_range;
1050 msn_to_uid_range(msn2uid($self), $range_csv);
1051 my $range_info = range_step($self, \$range_csv);
1052 return "$tag $range_info\r\n" if !ref($range_info);
1053 uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM
1054 $self->long_response($cb, $tag, [], $range_info, $ops, $partial);
1057 sub msn_convert ($$) {
1058 my ($self, $uids) = @_;
1059 my $adj = $self->{uid_base} + 1;
1060 my $uo2m = uo2m_extend($self, $uids->[-1]);
1061 $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m);
1062 $_ = $uo2m->[$_ - $adj] for @$uids;
1065 sub search_uid_range { # long_response
1066 my ($self, $tag, $sql, $range_info, $want_msn) = @_;
1068 if (defined(my $err = refill_uids($self, $uids, $range_info, $sql))) {
1069 $err ||= 'OK Search done';
1070 $self->write("\r\n$tag $err\r\n");
1073 msn_convert($self, $uids) if $want_msn;
1074 $self->msg_more(join(' ', '', @$uids));
1078 sub parse_imap_query ($$) {
1079 my ($self, $query) = @_;
1080 my $q = PublicInbox::IMAPsearchqp::parse($self, $query);
1082 my $max = $self->{ibx}->over(1)->max;
1084 uid_clamp($self, \$beg, \$max);
1085 $q->{range_info} = [ $beg, $max ];
1091 my ($self, $tag, $query, $want_msn) = @_;
1092 my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
1093 my $q = parse_imap_query($self, $query);
1094 return "$tag $q\r\n" if !ref($q);
1095 my ($sql, $range_info) = delete @$q{qw(sql range_info)};
1096 if (!scalar(keys %$q)) { # overview.sqlite3
1097 $self->msg_more('* SEARCH');
1098 $self->long_response(\&search_uid_range,
1099 $tag, $sql, $range_info, $want_msn);
1100 } elsif ($q = $q->{xap}) {
1101 my $srch = $self->{ibx}->isrch or
1102 return "$tag BAD search not available for mailbox\r\n";
1106 uid_range => $range_info
1108 my $mset = $srch->mset($q, $opt);
1109 my $uids = $srch->mset_to_artnums($mset, $opt);
1110 msn_convert($self, $uids) if scalar(@$uids) && $want_msn;
1111 "* SEARCH @$uids\r\n$tag OK Search done\r\n";
1113 "$tag BAD Error\r\n";
1117 sub cmd_uid_search ($$$) {
1118 my ($self, $tag, $query) = @_;
1119 search_common($self, $tag, $query);
1122 sub cmd_search ($$$;) {
1123 my ($self, $tag, $query) = @_;
1124 search_common($self, $tag, $query, 1);
1127 # returns 1 if we can continue, 0 if not due to buffered writes or disconnect
1128 sub process_line ($$) {
1129 my ($self, $l) = @_;
1131 # TODO: IMAP allows literals for big requests to upload messages
1132 # (which we don't support) but maybe some big search queries use it.
1133 # RFC 3501 9 (2) doesn't permit TAB or multiple SP
1134 my ($tag, $req, @args) = parse_line('[ \t]+', 0, $l);
1135 pop(@args) if (@args && !defined($args[-1]));
1136 if (@args && uc($req) eq 'UID') {
1137 $req .= "_".(shift @args);
1140 if (defined(my $idle_tag = $self->{-idle_tag})) {
1141 (uc($tag // '') eq 'DONE' && !defined($req)) ?
1142 idle_done($self, $tag) :
1143 "$idle_tag BAD expected DONE\r\n";
1144 } elsif (my $cmd = $self->can('cmd_'.lc($req // ''))) {
1145 if ($cmd == \&cmd_uid_search || $cmd == \&cmd_search) {
1146 # preserve user-supplied quotes for search
1147 (undef, @args) = split(/ search /i, $l, 2);
1149 $cmd->($self, $tag, @args);
1150 } else { # this is weird
1151 auth_challenge_ok($self) //
1153 ' BAD Error in IMAP command '.
1155 ": Unknown command\r\n";
1159 if ($err && $self->{sock}) {
1161 err($self, 'error from: %s (%s)', $l, $err);
1163 $res = "$tag BAD program fault - command not performed\r\n";
1165 return 0 unless defined $res;
1170 my ($self, $fmt, @args) = @_;
1171 printf { $self->{imapd}->{err} } $fmt."\n", @args;
1175 my ($self, $fmt, @args) = @_;
1176 printf { $self->{imapd}->{out} } $fmt."\n", @args;
1179 # callback used by PublicInbox::DS for any (e)poll (in/out/hup/err)
1183 return unless $self->flush_write && $self->{sock} && !$self->{long_cb};
1185 # only read more requests if we've drained the write buffer,
1186 # otherwise we can be buffering infinitely w/o backpressure
1188 my $rbuf = $self->{rbuf} // \(my $x = '');
1189 my $line = index($$rbuf, "\n");
1191 if (length($$rbuf) >= LINE_MAX) {
1192 $self->write(\"\* BAD request too long\r\n");
1193 return $self->close;
1195 $self->do_read($rbuf, LINE_MAX, length($$rbuf)) or
1196 return uo2m_hibernate($self);
1197 $line = index($$rbuf, "\n");
1199 $line = substr($$rbuf, 0, $line + 1, '');
1200 $line =~ s/\r?\n\z//s;
1201 return $self->close if $line =~ /[[:cntrl:]]/s;
1203 my $fd = fileno($self->{sock});
1204 my $r = eval { process_line($self, $line) };
1205 my $pending = $self->{wbuf} ? ' pending' : '';
1206 out($self, "[$fd] %s - %0.6f$pending - $r", $line, now() - $t0);
1208 return $self->close if $r < 0;
1209 $self->rbuf_idle($rbuf);
1211 # maybe there's more pipelined data, or we'll have
1212 # to register it for socket-readiness notifications
1213 $self->requeue unless $pending;
1217 sub cmd_compress ($$$) {
1218 my ($self, $tag, $alg) = @_;
1219 return "$tag BAD DEFLATE only\r\n" if uc($alg) ne "DEFLATE";
1220 return "$tag BAD COMPRESS active\r\n" if $self->compressed;
1222 # CRIME made TLS compression obsolete
1223 # return "$tag NO [COMPRESSIONACTIVE]\r\n" if $self->tls_compressed;
1225 PublicInbox::IMAPdeflate->enable($self) or return
1226 \"$tag BAD failed to activate compression\r\n";
1227 PublicInbox::DS::write($self, \"$tag OK DEFLATE active\r\n");
1232 sub cmd_starttls ($$) {
1233 my ($self, $tag) = @_;
1234 (($self->{sock} // return)->can('stop_SSL') || $self->compressed) and
1235 return "$tag BAD TLS or compression already enabled\r\n";
1236 $self->{imapd}->{ssl_ctx_opt} or
1237 return "$tag BAD can not initiate TLS negotiation\r\n";
1238 $self->write(\"$tag OK begin TLS negotiation now\r\n");
1239 PublicInbox::TLS::start($self->{sock}, $self->{imapd});
1240 $self->requeue if PublicInbox::DS::accept_tls_step($self);
1244 sub busy { # for graceful shutdown in PublicInbox::Daemon:
1246 if (defined($self->{-idle_tag})) {
1247 $self->write(\"* BYE server shutting down\r\n");
1248 return; # not busy anymore
1250 defined($self->{rbuf}) || defined($self->{wbuf}) ||
1251 !$self->write(\"* BYE server shutting down\r\n");
1256 if (my $ibx = delete $self->{ibx}) {
1257 stop_idle($self, $ibx);
1259 $self->SUPER::close; # PublicInbox::DS::close
1262 # we're read-only, so SELECT and EXAMINE do the same thing
1264 *cmd_select = \&cmd_examine;
1266 package PublicInbox::IMAP_preauth;
1267 our @ISA = qw(PublicInbox::IMAP);
1271 package PublicInbox::IMAPdeflate;
1272 use PublicInbox::DSdeflate;
1273 our @ISA = qw(PublicInbox::DSdeflate PublicInbox::IMAP);