X-Git-Url: http://www.git.stargrave.org/?p=public-inbox.git;a=blobdiff_plain;f=lib%2FPublicInbox%2FIMAP.pm;h=ce0dce0f317e8fb40a2a0b9da160594cafeb4fcf;hp=3815141a15ec399716b3e1fc58dbccd4d4636bb0;hb=d07ba9c30800225052d17ccca458afbfa05a8ff0;hpb=f1c9ad532f8dd46df172cfde85329a6e00ed1eab diff --git a/lib/PublicInbox/IMAP.pm b/lib/PublicInbox/IMAP.pm index 3815141a..ce0dce0f 100644 --- a/lib/PublicInbox/IMAP.pm +++ b/lib/PublicInbox/IMAP.pm @@ -1,4 +1,4 @@ -# Copyright (C) 2020 all contributors +# Copyright (C) all contributors # License: AGPL-3.0+ # # Each instance of this represents an IMAP client connected to @@ -7,23 +7,39 @@ # slow storage. # # data notes: -# * NNTP article numbers are UIDs and message sequence numbers (MSNs) -# * Message sequence numbers (MSNs) can be stable since we're read-only. -# Most IMAP clients use UIDs (I hope), and we can return a dummy -# message if a client requests a non-existent MSN. - +# +# * NNTP article numbers are UIDs, mm->created_at is UIDVALIDITY +# +# * public-inboxes are sliced into mailboxes of 50K messages +# to not overload MUAs: $NEWSGROUP_NAME.$SLICE_INDEX +# Slices are similar in concept to v2 "epochs". Epochs +# are for the limitations of git clients, while slices are +# for the limitations of IMAP clients. +# +# * We also take advantage of slices being only 50K to store +# "UID offset" to message sequence number (MSN) mapping +# as a 50K uint16_t array (via pack("S*", ...)). "UID offset" +# is the offset from {uid_base} which determines the start of +# the mailbox slice. +# +# fields: +# imapd: PublicInbox::IMAPD ref +# ibx: PublicInbox::Inbox ref +# long_cb: long_response private data +# uid_base: base UID for mailbox slice (0-based) +# -login_tag: IMAP TAG for LOGIN +# -idle_tag: IMAP response tag for IDLE +# uo2m: UID-to-MSN mapping package PublicInbox::IMAP; use strict; -use base qw(PublicInbox::DS); -use fields qw(imapd logged_in ibx long_cb -login_tag - -idle_tag -idle_max); +use parent qw(PublicInbox::DS); use PublicInbox::Eml; use PublicInbox::EmlContentFoo qw(parse_content_disposition); use PublicInbox::DS qw(now); -use PublicInbox::Syscall qw(EPOLLIN EPOLLONESHOT); use PublicInbox::GitAsyncCat; use Text::ParseWords qw(parse_line); use Errno qw(EAGAIN); +use PublicInbox::IMAPsearchqp; my $Address; for my $mod (qw(Email::Address::XS Mail::Address)) { @@ -32,24 +48,41 @@ for my $mod (qw(Email::Address::XS Mail::Address)) { } die "neither Email::Address::XS nor Mail::Address loaded: $@" if !$Address; -sub LINE_MAX () { 512 } # does RFC 3501 have a limit like RFC 977? - -my %FETCH_NEED_BLOB = ( # for future optimization - 'BODY[HEADER]' => 1, - 'BODY[TEXT]' => 1, - 'BODY[]' => 1, - 'RFC822.HEADER' => 1, - 'RFC822.SIZE' => 1, # needs CRLF conversion :< - 'RFC822.TEXT' => 1, - BODY => 1, - BODYSTRUCTURE => 1, - ENVELOPE => 1, - FLAGS => 0, - INTERNALDATE => 0, - RFC822 => 1, - UID => 0, +sub LINE_MAX () { 8000 } # RFC 2683 3.2.1.5 + +# Changing UID_SLICE will cause grief for clients which cache. +# This also needs to be <64K: we pack it into a uint16_t +# for long_response UID (offset) => MSN mappings +sub UID_SLICE () { 50_000 } + +# these values area also used for sorting +sub NEED_SMSG () { 1 } +sub NEED_BLOB () { NEED_SMSG|2 } +sub CRLF_BREF () { 4 } +sub EML_HDR () { 8 } +sub CRLF_HDR () { 16 } +sub EML_BDY () { 32 } +sub CRLF_BDY () { 64 } +my $OP_EML_NEW = [ EML_HDR - 1, \&op_eml_new ]; +my $OP_CRLF_BREF = [ CRLF_BREF, \&op_crlf_bref ]; +my $OP_CRLF_HDR = [ CRLF_HDR, \&op_crlf_hdr ]; +my $OP_CRLF_BDY = [ CRLF_BDY, \&op_crlf_bdy ]; + +my %FETCH_NEED = ( + 'BODY[HEADER]' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ], + 'BODY[TEXT]' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ], + 'BODY[]' => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ], + 'RFC822.HEADER' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ], + 'RFC822.TEXT' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ], + 'RFC822.SIZE' => [ NEED_SMSG, \&emit_rfc822_size ], + RFC822 => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ], + BODY => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_body ], + BODYSTRUCTURE => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_bodystructure ], + ENVELOPE => [ NEED_BLOB|EML_HDR, \&emit_envelope ], + FLAGS => [ 0, \&emit_flags ], + INTERNALDATE => [ NEED_SMSG, \&emit_internaldate ], ); -my %FETCH_ATT = map { $_ => [ $_ ] } keys %FETCH_NEED_BLOB; +my %FETCH_ATT = map { $_ => [ $_ ] } keys %FETCH_NEED; # aliases (RFC 3501 section 6.4.5) $FETCH_ATT{FAST} = [ qw(FLAGS INTERNALDATE RFC822.SIZE) ]; @@ -57,47 +90,34 @@ $FETCH_ATT{ALL} = [ @{$FETCH_ATT{FAST}}, 'ENVELOPE' ]; $FETCH_ATT{FULL} = [ @{$FETCH_ATT{ALL}}, 'BODY' ]; for my $att (keys %FETCH_ATT) { - my %h = map { $_ => 1 } @{$FETCH_ATT{$att}}; + my %h = map { $_ => $FETCH_NEED{$_} } @{$FETCH_ATT{$att}}; $FETCH_ATT{$att} = \%h; } +undef %FETCH_NEED; my $valid_range = '[0-9]+|[0-9]+:[0-9]+|[0-9]+:\*'; $valid_range = qr/\A(?:$valid_range)(?:,(?:$valid_range))*\z/; -sub greet ($) { +sub do_greet { my ($self) = @_; my $capa = capa($self); $self->write(\"* OK [$capa] public-inbox-imapd ready\r\n"); } -sub new ($$$) { - my ($class, $sock, $imapd) = @_; - my $self = fields::new($class); - my $ev = EPOLLIN; - my $wbuf; - if ($sock->can('accept_SSL') && !$sock->accept_SSL) { - return CORE::close($sock) if $! != EAGAIN; - $ev = PublicInbox::TLS::epollbit(); - $wbuf = [ \&PublicInbox::DS::accept_tls_step, \&greet ]; - } - $self->SUPER::new($sock, $ev | EPOLLONESHOT); - $self->{imapd} = $imapd; - if ($wbuf) { - $self->{wbuf} = $wbuf; - } else { - greet($self); - } - $self->update_idle_time; - $self; +sub new { + my (undef, $sock, $imapd) = @_; + (bless { imapd => $imapd }, 'PublicInbox::IMAP_preauth')->greet($sock) } +sub logged_in { 1 } + sub capa ($) { my ($self) = @_; # dovecot advertises IDLE pre-login; perhaps because some clients # depend on it, so we'll do the same my $capa = 'CAPABILITY IMAP4rev1 IDLE'; - if ($self->{logged_in}) { + if ($self->logged_in) { $capa .= ' COMPRESS=DEFLATE'; } else { if (!($self->{sock} // $self)->can('accept_SSL') && @@ -110,7 +130,7 @@ sub capa ($) { sub login_success ($$) { my ($self, $tag) = @_; - $self->{logged_in} = 1; + bless $self, 'PublicInbox::IMAP'; my $capa = capa($self); "$tag OK [$capa] Logged in\r\n"; } @@ -128,13 +148,14 @@ sub cmd_login ($$$$) { sub cmd_close ($$) { my ($self, $tag) = @_; + delete @$self{qw(uid_base uo2m)}; delete $self->{ibx} ? "$tag OK Close done\r\n" : "$tag BAD No mailbox\r\n"; } sub cmd_logout ($$) { my ($self, $tag) = @_; - delete $self->{logged_in}; + delete $self->{-idle_tag}; $self->write(\"* BYE logging out\r\n$tag OK Logout done\r\n"); $self->shutdn; # PublicInbox::DS::shutdn undef; @@ -151,32 +172,173 @@ sub cmd_capability ($$) { '* '.capa($self)."\r\n$tag OK Capability done\r\n"; } -sub cmd_noop ($$) { "$_[1] OK Noop done\r\n" } +# uo2m: UID Offset to MSN, this is an arrayref by default, +# but uo2m_hibernate can compact and deduplicate it +sub uo2m_ary_new ($;$) { + my ($self, $exists) = @_; + my $ub = $self->{uid_base}; + my $uids = $self->{ibx}->over(1)->uid_range($ub + 1, $ub + UID_SLICE); + + # convert UIDs to offsets from {base} + my @tmp; # [$UID_OFFSET] => $MSN + my $msn = 0; + ++$ub; + $tmp[$_ - $ub] = ++$msn for @$uids; + $$exists = $msn if $exists; + \@tmp; +} + +# changes UID-offset-to-MSN mapping into a deduplicated scalar: +# uint16_t uo2m[UID_SLICE]. +# May be swapped out for idle clients if THP is disabled. +sub uo2m_hibernate ($) { + my ($self) = @_; + ref(my $uo2m = $self->{uo2m}) or return; + my %dedupe = ( uo2m_pack($uo2m) => undef ); + $self->{uo2m} = (keys(%dedupe))[0]; + undef; +} + +sub uo2m_last_uid ($) { + my ($self) = @_; + defined(my $uo2m = $self->{uo2m}) or die 'BUG: uo2m_last_uid w/o {uo2m}'; + (ref($uo2m) ? @$uo2m : (length($uo2m) >> 1)) + $self->{uid_base}; +} + +sub uo2m_pack ($) { + # $_[0] is an arrayref of MSNs, it may have undef gaps if there + # are gaps in the corresponding UIDs: [ msn1, msn2, undef, msn3 ] + no warnings 'uninitialized'; + pack('S*', @{$_[0]}); +} + +# extend {uo2m} to account for new messages which arrived since +# {uo2m} was created. +sub uo2m_extend ($$;$) { + my ($self, $new_uid_max) = @_; + defined(my $uo2m = $self->{uo2m}) or + return($self->{uo2m} = uo2m_ary_new($self)); + my $beg = uo2m_last_uid($self); # last UID we've learned + return $uo2m if $beg >= $new_uid_max; # fast path + + # need to extend the current range: + my $base = $self->{uid_base}; + ++$beg; + my $uids = $self->{ibx}->over(1)->uid_range($beg, $base + UID_SLICE); + return $uo2m if !scalar(@$uids); + my @tmp; # [$UID_OFFSET] => $MSN + my $write_method = $_[2] // 'msg_more'; + if (ref($uo2m)) { + my $msn = $uo2m->[-1]; + $tmp[$_ - $beg] = ++$msn for @$uids; + $self->$write_method("* $msn EXISTS\r\n"); + push @$uo2m, @tmp; + $uo2m; + } else { + my $msn = unpack('S', substr($uo2m, -2, 2)); + $tmp[$_ - $beg] = ++$msn for @$uids; + $self->$write_method("* $msn EXISTS\r\n"); + $uo2m .= uo2m_pack(\@tmp); + my %dedupe = ($uo2m => undef); + $self->{uo2m} = (keys %dedupe)[0]; + } +} + +sub cmd_noop ($$) { + my ($self, $tag) = @_; + defined($self->{uid_base}) and + uo2m_extend($self, $self->{uid_base} + UID_SLICE); + \"$tag OK Noop done\r\n"; +} + +# the flexible version which works on scalars and array refs. +# Must call uo2m_extend before this +sub uid2msn ($$) { + my ($self, $uid) = @_; + my $uo2m = $self->{uo2m}; + my $off = $uid - $self->{uid_base} - 1; + ref($uo2m) ? $uo2m->[$off] : unpack('S', substr($uo2m, $off << 1, 2)); +} + +# returns an arrayref of UIDs, so MSNs can be translated to UIDs via: +# $msn2uid->[$MSN-1] => $UID. The result of this is always ephemeral +# and does not live beyond the event loop. +sub msn2uid ($) { + my ($self) = @_; + my $base = $self->{uid_base}; + my $uo2m = uo2m_extend($self, $base + UID_SLICE); + $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m); + + my $uo = 0; + my @msn2uid; + for my $msn (@$uo2m) { + ++$uo; + $msn2uid[$msn - 1] = $uo + $base if $msn; + } + \@msn2uid; +} + +# converts a set of message sequence numbers in requests to UIDs: +sub msn_to_uid_range ($$) { + my $msn2uid = $_[0]; + $_[1] =~ s!([0-9]+)!$msn2uid->[$1 - 1] // ($msn2uid->[-1] // 0 + 1)!sge; +} # called by PublicInbox::InboxIdle sub on_inbox_unlock { my ($self, $ibx) = @_; - my $new = $ibx->mm->max; - defined(my $old = $self->{-idle_max}) or die 'BUG: -idle_max unset'; - if ($new > $old) { - $self->{-idle_max} = $new; - $self->msg_more("* $_ EXISTS\r\n") for (($old + 1)..($new - 1)); - $self->write(\"* $new EXISTS\r\n"); + my $uid_end = $self->{uid_base} + UID_SLICE; + uo2m_extend($self, $uid_end, 'write'); + my $new = uo2m_last_uid($self); + if ($new == $uid_end) { # max exceeded $uid_end + # continue idling w/o inotify + my $sock = $self->{sock} or return; + $ibx->unsubscribe_unlock(fileno($sock)); + } +} + +# called every minute or so by PublicInbox::DS::later +my $IDLERS; # fileno($obj->{sock}) => PublicInbox::IMAP +sub idle_tick_all { + my $old = $IDLERS; + $IDLERS = undef; + for my $i (values %$old) { + next if ($i->{wbuf} || !exists($i->{-idle_tag})); + $IDLERS->{fileno($i->{sock})} = $i; + $i->write(\"* OK Still here\r\n"); } + $IDLERS and + PublicInbox::DS::add_uniq_timer('idle', 60, \&idle_tick_all); } sub cmd_idle ($$) { my ($self, $tag) = @_; # IDLE seems allowed by dovecot w/o a mailbox selected *shrug* my $ibx = $self->{ibx} or return "$tag BAD no mailbox selected\r\n"; - $ibx->subscribe_unlock(fileno($self->{sock}), $self); - $self->{imapd}->idler_start; + my $uid_end = $self->{uid_base} + UID_SLICE; + uo2m_extend($self, $uid_end); + my $sock = $self->{sock} or return; + my $fd = fileno($sock); $self->{-idle_tag} = $tag; - $self->{-idle_max} = $ibx->mm->max // 0; - "+ idling\r\n" + # only do inotify on most recent slice + if ($ibx->over(1)->max < $uid_end) { + $ibx->subscribe_unlock($fd, $self); + $self->{imapd}->idler_start; + } + PublicInbox::DS::add_uniq_timer('idle', 60, \&idle_tick_all); + $IDLERS->{$fd} = $self; + \"+ idling\r\n" +} + +sub stop_idle ($$) { + my ($self, $ibx) = @_; + my $sock = $self->{sock} or return; + my $fd = fileno($sock); + delete $IDLERS->{$fd}; + $ibx->unsubscribe_unlock($fd); } -sub cmd_done ($$) { +sub idle_done ($$) { my ($self, $tag) = @_; # $tag is "DONE" (case-insensitive) defined(my $idle_tag = delete $self->{-idle_tag}) or return "$tag BAD not idle\r\n"; @@ -184,35 +346,80 @@ sub cmd_done ($$) { warn "BUG: idle_tag set w/o inbox"; return "$tag BAD internal bug\r\n"; }; - $ibx->unsubscribe_unlock(fileno($self->{sock})); + stop_idle($self, $ibx); "$idle_tag OK Idle done\r\n"; } +sub ensure_slices_exist ($$$) { + my ($imapd, $ibx, $max) = @_; + defined(my $mb_top = $ibx->{newsgroup}) or return; + my $mailboxes = $imapd->{mailboxes}; + my @created; + for (my $i = int($max/UID_SLICE); $i >= 0; --$i) { + my $sub_mailbox = "$mb_top.$i"; + last if exists $mailboxes->{$sub_mailbox}; + $mailboxes->{$sub_mailbox} = $ibx; + $sub_mailbox =~ s/\Ainbox\./INBOX./i; # more familiar to users + push @created, $sub_mailbox; + } + return unless @created; + my $l = $imapd->{mailboxlist} or return; + push @$l, map { qq[* LIST (\\HasNoChildren) "." $_\r\n] } @created; +} + +sub inbox_lookup ($$;$) { + my ($self, $mailbox, $examine) = @_; + my ($ibx, $exists, $uidmax, $uid_base) = (undef, 0, 0, 0); + $mailbox = lc $mailbox; + $ibx = $self->{imapd}->{mailboxes}->{$mailbox} or return; + my $over = $ibx->over(1); + if ($over != $ibx) { # not a dummy + $mailbox =~ /\.([0-9]+)\z/ or + die "BUG: unexpected dummy mailbox: $mailbox\n"; + $uid_base = $1 * UID_SLICE; + + $uidmax = $ibx->mm->num_highwater // 0; + if ($examine) { + $self->{uid_base} = $uid_base; + $self->{ibx} = $ibx; + $self->{uo2m} = uo2m_ary_new($self, \$exists); + } else { + my $uid_end = $uid_base + UID_SLICE; + $exists = $over->imap_exists($uid_base, $uid_end); + } + ensure_slices_exist($self->{imapd}, $ibx, $over->max); + } else { + if ($examine) { + $self->{uid_base} = $uid_base; + $self->{ibx} = $ibx; + delete $self->{uo2m}; + } + # if "INBOX.foo.bar" is selected and "INBOX.foo.bar.0", + # check for new UID ranges (e.g. "INBOX.foo.bar.1") + if (my $z = $self->{imapd}->{mailboxes}->{"$mailbox.0"}) { + ensure_slices_exist($self->{imapd}, $z, + $z->over(1)->max); + } + } + ($ibx, $exists, $uidmax + 1, $uid_base); +} + sub cmd_examine ($$$) { my ($self, $tag, $mailbox) = @_; - my $ibx = $self->{imapd}->{groups}->{$mailbox} or - return "$tag NO Mailbox doesn't exist: $mailbox\r\n"; - my $mm = $ibx->mm; - my $max = $mm->max // 0; - # RFC 3501 2.3.1.1 - "A good UIDVALIDITY value to use in - # this case is a 32-bit representation of the creation - # date/time of the mailbox" - my $uidvalidity = $mm->created_at or return "$tag BAD UIDVALIDITY\r\n"; - my $uidnext = $max + 1; - # XXX: do we need this? RFC 5162/7162 my $ret = $self->{ibx} ? "* OK [CLOSED] previous closed\r\n" : ''; - $self->{ibx} = $ibx; + my ($ibx, $exists, $uidnext, $base) = inbox_lookup($self, $mailbox, 1); + return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx; $ret .= <{uidvalidity}]\r +$tag OK [READ-ONLY] EXAMINE/SELECT done\r EOF - $ret .= "* OK [UNSEEN $max]\r\n" if $max; - $ret .= "* OK [UIDNEXT $uidnext]\r\n" if defined $uidnext; - $ret .= "* OK [UIDVALIDITY $uidvalidity]\r\n" if defined $uidvalidity; - $ret .= "$tag OK [READ-ONLY] EXAMINE/SELECT done\r\n"; } sub _esc ($) { @@ -272,7 +479,7 @@ sub body_disposition ($) { my $cd = $eml->header_raw('Content-Disposition') or return 'NIL'; $cd = parse_content_disposition($cd); my $buf = '('._esc($cd->{type}); - $buf .= ' ' . _esc_hash(delete $cd->{attributes}); + $buf .= ' ' . _esc_hash($cd->{attributes}); $buf .= ')'; } @@ -284,7 +491,7 @@ sub body_leaf ($$;$) { my $ct = $eml->ct; $buf .= '('._esc($ct->{type}).' '; $buf .= _esc($ct->{subtype}); - $buf .= ' ' . _esc_hash(delete $ct->{attributes}); + $buf .= ' ' . _esc_hash($ct->{attributes}); $buf .= ' ' . _esc($eml->header_raw('Content-ID')); $buf .= ' ' . _esc($eml->header_raw('Content-Description')); my $cte = $eml->header_raw('Content-Transfer-Encoding') // '7bit'; @@ -313,7 +520,7 @@ sub body_parent ($$$) { $buf .= @$hold ? join('', @$hold) : 'NIL'; $buf .= ' '._esc($ct->{subtype}); if ($structure) { - $buf .= ' '._esc_hash(delete $ct->{attributes}); + $buf .= ' '._esc_hash($ct->{attributes}); $buf .= ' '.body_disposition($eml); $buf .= ' '._esc($eml->header_raw('Content-Language')); $buf .= ' '._esc($eml->header_raw('Content-Location')); @@ -356,97 +563,115 @@ sub fetch_body ($;$) { join('', @hold); } -sub dummy_message ($$) { - my ($seqno, $ibx) = @_; - my $ret = <{newsgroup}>\r -Subject: dummy message #$seqno\r -\r -You're seeing this message because your IMAP client didn't use UIDs.\r -The message which used to use this sequence number was likely spam\r -and removed by the administrator.\r -EOF - \$ret; -} - -sub requeue_once ($) { - my ($self) = @_; - # COMPRESS users all share the same DEFLATE context. - # Flush it here to ensure clients don't see - # each other's data - $self->zflush; - - # no recursion, schedule another call ASAP, - # but only after all pending writes are done. - # autovivify wbuf: - my $new_size = push(@{$self->{wbuf}}, \&long_step); - - # wbuf may be populated by $cb, no need to rearm if so: - $self->requeue if $new_size == 1; +sub fetch_run_ops { + my ($self, $smsg, $bref, $ops, $partial) = @_; + my $uid = $smsg->{num}; + $self->msg_more('* '.uid2msn($self, $uid)." FETCH (UID $uid"); + my ($eml, $k); + for (my $i = 0; $i < @$ops;) { + $k = $ops->[$i++]; + $ops->[$i++]->($self, $k, $smsg, $bref, $eml); + } + partial_emit($self, $partial, $eml) if $partial; + $self->msg_more(")\r\n"); } -sub uid_fetch_cb { # called by git->cat_async via git_async_cat - my ($bref, $oid, $type, $size, $fetch_m_arg) = @_; - my ($self, undef, $ibx, $msgs, undef, $want) = @$fetch_m_arg; +sub fetch_blob_cb { # called by git->cat_async via ibx_async_cat + my ($bref, $oid, $type, $size, $fetch_arg) = @_; + my ($self, undef, $msgs, $range_info, $ops, $partial) = @$fetch_arg; + my $ibx = $self->{ibx} or return $self->close; # client disconnected my $smsg = shift @$msgs or die 'BUG: no smsg'; if (!defined($oid)) { # it's possible to have TOCTOU if an admin runs # public-inbox-(edit|purge), just move onto the next message - return requeue_once($self) unless defined $want->{-seqno}; - $bref = dummy_message($smsg->{num}, $ibx); + warn "E: $smsg->{blob} missing in $ibx->{inboxdir}\n"; + return $self->requeue_once; } else { $smsg->{blob} eq $oid or die "BUG: $smsg->{blob} != $oid"; } + my $pre; + if (!$self->{wbuf} && (my $nxt = $msgs->[0])) { + $pre = ibx_async_prefetch($ibx, $nxt->{blob}, + \&fetch_blob_cb, $fetch_arg); + } + fetch_run_ops($self, $smsg, $bref, $ops, $partial); + $pre ? $self->zflush : $self->requeue_once; +} - $$bref =~ s/(?msg_more(" $k {" . length($$bref)."}\r\n"); + $self->msg_more($$bref); +} - # fixup old bug from import (pre-a0c07cba0e5d8b6a) - $$bref =~ s/\A[\r\n]*From [^\r\n]*\r\n//s; +# Mail::IMAPClient::message_string cares about this by default, +# (->Ignoresizeerrors attribute). Admins are encouraged to +# --reindex for IMAP support, anyways. +sub emit_rfc822_size { + my ($self, $k, $smsg) = @_; + $self->msg_more(' RFC822.SIZE ' . $smsg->{bytes}); +} - $self->msg_more("* $smsg->{num} FETCH (UID $smsg->{num}"); +sub emit_internaldate { + my ($self, undef, $smsg) = @_; + $self->msg_more(' INTERNALDATE "'.$smsg->internaldate.'"'); +} - $want->{'RFC822.SIZE'} and - $self->msg_more(' RFC822.SIZE '.length($$bref)); - $want->{INTERNALDATE} and - $self->msg_more(' INTERNALDATE "'.$smsg->internaldate.'"'); - $want->{FLAGS} and $self->msg_more(' FLAGS ()'); - for ('RFC822', 'BODY[]') { - $want->{$_} or next; - $self->msg_more(" $_ {".length($$bref)."}\r\n"); - $self->msg_more($$bref); - } +sub emit_flags { $_[0]->msg_more(' FLAGS ()') } - my $eml = PublicInbox::Eml->new($bref); +sub emit_envelope { + my ($self, undef, undef, undef, $eml) = @_; + $self->msg_more(' ENVELOPE '.eml_envelope($eml)); +} - $want->{ENVELOPE} and - $self->msg_more(' ENVELOPE '.eml_envelope($eml)); +sub emit_rfc822_header { + my ($self, $k, undef, undef, $eml) = @_; + $self->msg_more(" $k {".length(${$eml->{hdr}})."}\r\n"); + $self->msg_more(${$eml->{hdr}}); +} - for ('RFC822.HEADER', 'BODY[HEADER]') { - $want->{$_} or next; - $self->msg_more(" $_ {".length(${$eml->{hdr}})."}\r\n"); - $self->msg_more(${$eml->{hdr}}); - } - for ('RFC822.TEXT', 'BODY[TEXT]') { - $want->{$_} or next; - $self->msg_more(" $_ {".length($$bref)."}\r\n"); - $self->msg_more($$bref); - } - $want->{BODYSTRUCTURE} and - $self->msg_more(' BODYSTRUCTURE '.fetch_body($eml, 1)); - $want->{BODY} and - $self->msg_more(' BODY '.fetch_body($eml)); - if (my $partial = $want->{-partial}) { - partial_emit($self, $partial, $eml); - } - $self->msg_more(")\r\n"); - requeue_once($self); +# n.b. this is sorted to be after any emit_eml_new ops +sub emit_rfc822_text { + my ($self, $k, undef, $bref) = @_; + $self->msg_more(" $k {".length($$bref)."}\r\n"); + $self->msg_more($$bref); +} + +sub emit_bodystructure { + my ($self, undef, undef, undef, $eml) = @_; + $self->msg_more(' BODYSTRUCTURE '.fetch_body($eml, 1)); +} + +sub emit_body { + my ($self, undef, undef, undef, $eml) = @_; + $self->msg_more(' BODY '.fetch_body($eml)); +} + +# set $eml once ($_[4] == $eml, $_[3] == $bref) +sub op_eml_new { $_[4] = PublicInbox::Eml->new($_[3]) } + +# s/From / fixes old bug from import (pre-a0c07cba0e5d8b6a) +sub to_crlf_full { + ${$_[0]} =~ s/(?{hdr}) } + +sub op_crlf_bdy { ${$_[4]->{bdy}} =~ s/(?{bdy} } + +sub uid_clamp ($$$) { + my ($self, $beg, $end) = @_; + my $uid_min = $self->{uid_base} + 1; + my $uid_end = $uid_min + UID_SLICE - 1; + $$beg = $uid_min if $$beg < $uid_min; + $$end = $uid_end if $$end > $uid_end; } sub range_step ($$) { - my ($ibx, $range_csv) = @_; + my ($self, $range_csv) = @_; my ($beg, $end, $range); if ($$range_csv =~ s/\A([^,]+),//) { $range = $1; @@ -454,12 +679,22 @@ sub range_step ($$) { $range = $$range_csv; $$range_csv = undef; } + my $uid_base = $self->{uid_base}; + my $uid_end = $uid_base + UID_SLICE; if ($range =~ /\A([0-9]+):([0-9]+)\z/) { - ($beg, $end) = ($1, $2); + ($beg, $end) = ($1 + 0, $2 + 0); + uid_clamp($self, \$beg, \$end); } elsif ($range =~ /\A([0-9]+):\*\z/) { - ($beg, $end) = ($1, $ibx->mm->max // 0); + $beg = $1 + 0; + $end = $self->{ibx}->over(1)->max; + $end = $uid_end if $end > $uid_end; + $beg = $end if $beg > $end; + uid_clamp($self, \$beg, \$end); } elsif ($range =~ /\A[0-9]+\z/) { - $beg = $end = $range; + $beg = $end = $range + 0; + # just let the caller do an out-of-range query if a single + # UID is out-of-range + ++$beg if ($beg <= $uid_base || $end > $uid_end); } else { return 'BAD fetch range'; } @@ -467,50 +702,105 @@ sub range_step ($$) { } sub refill_range ($$$) { - my ($ibx, $msgs, $range_info) = @_; + my ($self, $msgs, $range_info) = @_; my ($beg, $end, $range_csv) = @$range_info; - if (scalar(@$msgs = @{$ibx->over->query_xover($beg, $end)})) { + if (scalar(@$msgs = @{$self->{ibx}->over(1)->query_xover($beg, $end)})){ $range_info->[0] = $msgs->[-1]->{num} + 1; return; } return 'OK Fetch done' if !$range_csv; - my $next_range = range_step($ibx, \$range_csv); + my $next_range = range_step($self, \$range_csv); return $next_range if !ref($next_range); # error @$range_info = @$next_range; undef; # keep looping } -sub uid_fetch_m { # long_response - my ($self, $tag, $ibx, $msgs, $range_info, $want) = @_; +sub fetch_blob { # long_response + my ($self, $tag, $msgs, $range_info, $ops, $partial) = @_; + while (!@$msgs) { # rare + if (my $end = refill_range($self, $msgs, $range_info)) { + $self->write(\"$tag $end\r\n"); + return; + } + } + uo2m_extend($self, $msgs->[-1]->{num}); + ibx_async_cat($self->{ibx}, $msgs->[0]->{blob}, + \&fetch_blob_cb, \@_); +} + +sub fetch_smsg { # long_response + my ($self, $tag, $msgs, $range_info, $ops) = @_; while (!@$msgs) { # rare - if (my $end = refill_range($ibx, $msgs, $range_info)) { + if (my $end = refill_range($self, $msgs, $range_info)) { $self->write(\"$tag $end\r\n"); return; } } - git_async_cat($ibx->git, $msgs->[0]->{blob}, \&uid_fetch_cb, \@_); + uo2m_extend($self, $msgs->[-1]->{num}); + fetch_run_ops($self, $_, undef, $ops) for @$msgs; + @$msgs = (); + 1; # more +} + +sub refill_uids ($$$;$) { + my ($self, $uids, $range_info, $sql) = @_; + my ($beg, $end, $range_csv) = @$range_info; + my $over = $self->{ibx}->over(1); + while (1) { + if (scalar(@$uids = @{$over->uid_range($beg, $end, $sql)})) { + $range_info->[0] = $uids->[-1] + 1; # update $beg + return; + } elsif (!$range_csv) { + return 0; + } else { + my $next_range = range_step($self, \$range_csv); + return $next_range if !ref($next_range); # error + ($beg, $end, $range_csv) = @$range_info = @$next_range; + # continue looping + } + } +} + +sub fetch_uid { # long_response + my ($self, $tag, $uids, $range_info, $ops) = @_; + if (defined(my $err = refill_uids($self, $uids, $range_info))) { + $err ||= 'OK Fetch done'; + $self->write("$tag $err\r\n"); + return; + } + my $adj = $self->{uid_base} + 1; + my $uo2m = uo2m_extend($self, $uids->[-1]); + $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m); + my ($i, $k); + for (@$uids) { + $self->msg_more("* $uo2m->[$_ - $adj] FETCH (UID $_"); + for ($i = 0; $i < @$ops;) { + $k = $ops->[$i++]; + $ops->[$i++]->($self, $k); + } + $self->msg_more(")\r\n"); + } + @$uids = (); + 1; # more } sub cmd_status ($$$;@) { my ($self, $tag, $mailbox, @items) = @_; - my $ibx = $self->{imapd}->{groups}->{$mailbox} or - return "$tag NO Mailbox doesn't exist: $mailbox\r\n"; return "$tag BAD no items\r\n" if !scalar(@items); ($items[0] !~ s/\A\(//s || $items[-1] !~ s/\)\z//s) and return "$tag BAD invalid args\r\n"; - - my $mm = $ibx->mm; - my ($max, @it); + my ($ibx, $exists, $uidnext) = inbox_lookup($self, $mailbox); + return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx; + my @it; for my $it (@items) { $it = uc($it); push @it, $it; if ($it =~ /\A(?:MESSAGES|UNSEEN|RECENT)\z/) { - push(@it, ($max //= $mm->max // 0)); + push @it, $exists; } elsif ($it eq 'UIDNEXT') { - push(@it, ($max //= $mm->max // 0) + 1); + push @it, $uidnext; } elsif ($it eq 'UIDVALIDITY') { - push(@it, $mm->created_at // - return("$tag BAD UIDVALIDITY\r\n")); + push @it, $ibx->{uidvalidity}; } else { return "$tag BAD invalid item\r\n"; } @@ -523,13 +813,13 @@ sub cmd_status ($$$;@) { my %patmap = ('*' => '.*', '%' => '[^\.]*'); sub cmd_list ($$$$) { my ($self, $tag, $refname, $wildcard) = @_; - my $l = $self->{imapd}->{inboxlist}; + my $l = $self->{imapd}->{mailboxlist}; if ($refname eq '' && $wildcard eq '') { # request for hierarchy delimiter $l = [ qq[* LIST (\\Noselect) "." ""\r\n] ]; } elsif ($refname ne '' || $wildcard ne '*') { - $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!eig; - $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/s, @$l) ]; + $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!egi; + $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/is, @$l) ]; } \(join('', @$l, "$tag OK List done\r\n")); } @@ -551,12 +841,12 @@ sub eml_index_offs_i { # PublicInbox::Eml::each_part callback # prepares an index for BODY[$SECTION_IDX] fetches sub eml_body_idx ($$) { my ($eml, $section_idx) = @_; - my $idx = $eml->{imap_all_parts} //= do { + my $idx = $eml->{imap_all_parts} // do { my $all = {}; $eml->each_part(\&eml_index_offs_i, $all, 0, 1); # top-level of multipart, BODY[0] not allowed (nz-number) delete $all->{0}; - $all; + $eml->{imap_all_parts} = $all; }; $idx->{$section_idx}; } @@ -607,6 +897,7 @@ sub partial_hdr_not { } my $str = $eml->header_obj->as_string; $str =~ s/$hdrs_re//g; + $str =~ s/(?header_obj->as_string; - join('', ($str =~ m/($hdrs_re)/g), "\r\n"); + $str = join('', ($str =~ m/($hdrs_re)/g)); + $str =~ s/(?)?\z/sx) { # 3, 4 $partial->{$att} = [ \&partial_body, $1, $2, $3, $4 ]; + $$need |= CRLF_BREF|EML_HDR|EML_BDY; } elsif ($att =~ /\ABODY\[(?:([0-9]+(?:\.[0-9]+)*)\.)? # 1 - section_idx (?:HEADER\.FIELDS(\.NOT)?)\x20 # 2 \(([A-Z0-9\-\x20]+)\) # 3 - hdrs @@ -642,6 +936,11 @@ sub partial_prepare ($$$) { : \&partial_hdr_get, $1, undef, $4, $5 ]; $tmp->[2] = hdrs_regexp($3); + + # don't emit CRLF_HDR instruction, here, partial_hdr_* + # will do CRLF conversion with only the extracted result + # and not waste time converting lines we don't care about. + $$need |= EML_HDR; } else { undef; } @@ -671,199 +970,201 @@ sub partial_emit ($$$) { } } -sub fetch_common ($$$$) { - my ($self, $tag, $range_csv, $want) = @_; - my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n"; +sub fetch_compile ($) { + my ($want) = @_; if ($want->[0] =~ s/\A\(//s) { - $want->[-1] =~ s/\)\z//s or return "$tag BAD no rparen\r\n"; + $want->[-1] =~ s/\)\z//s or return 'BAD no rparen'; } - my (%partial, %want); + my (%partial, %seen, @op); + my $need = 0; while (defined(my $att = shift @$want)) { $att = uc($att); + next if $att eq 'UID'; # always returned $att =~ s/\ABODY\.PEEK\[/BODY\[/; # we're read-only my $x = $FETCH_ATT{$att}; if ($x) { - %want = (%want, %$x); - } elsif (!partial_prepare(\%partial, $want, $att)) { - return "$tag BAD param: $att\r\n"; + while (my ($k, $fl_cb) = each %$x) { + next if $seen{$k}++; + $need |= $fl_cb->[0]; + push @op, [ @$fl_cb, $k ]; + } + } elsif (!partial_prepare(\$need, \%partial, $want, $att)) { + return "BAD param: $att"; } } + my @r; # stabilize partial order for consistency and ease-of-debugging: if (scalar keys %partial) { - $want{-partial} = [ map {; - [ $_, @{$partial{$_}} ] - } sort keys %partial ]; + $need |= NEED_BLOB; + $r[2] = [ map { [ $_, @{$partial{$_}} ] } sort keys %partial ]; + } + + push @op, $OP_EML_NEW if ($need & (EML_HDR|EML_BDY)); + + # do we need CRLF conversion? + if ($need & CRLF_BREF) { + push @op, $OP_CRLF_BREF; + } elsif (my $crlf = ($need & (CRLF_HDR|CRLF_BDY))) { + if ($crlf == (CRLF_HDR|CRLF_BDY)) { + push @op, $OP_CRLF_BREF; + } elsif ($need & CRLF_HDR) { + push @op, $OP_CRLF_HDR; + } else { + push @op, $OP_CRLF_BDY; + } } + + $r[0] = $need & NEED_BLOB ? \&fetch_blob : + ($need & NEED_SMSG ? \&fetch_smsg : \&fetch_uid); + + # r[1] = [ $key1, $cb1, $key2, $cb2, ... ] + use sort 'stable'; # makes output more consistent + $r[1] = [ map { ($_->[2], $_->[1]) } sort { $a->[0] <=> $b->[0] } @op ]; + @r; +} + +sub cmd_uid_fetch ($$$$;@) { + my ($self, $tag, $range_csv, @want) = @_; + my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n"; + my ($cb, $ops, $partial) = fetch_compile(\@want); + return "$tag $cb\r\n" unless $ops; + + # cb is one of fetch_blob, fetch_smsg, fetch_uid $range_csv = 'bad' if $range_csv !~ $valid_range; - my $range_info = range_step($ibx, \$range_csv); + my $range_info = range_step($self, \$range_csv); return "$tag $range_info\r\n" if !ref($range_info); - [ $tag, $ibx, [], $range_info, \%want ]; + uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM + $self->long_response($cb, $tag, [], $range_info, $ops, $partial); } -sub cmd_uid_fetch ($$$;@) { +sub cmd_fetch ($$$$;@) { my ($self, $tag, $range_csv, @want) = @_; - my $args = fetch_common($self, $tag, $range_csv, \@want); - ref($args) eq 'ARRAY' ? - long_response($self, \&uid_fetch_m, @$args) : - $args; # error + my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n"; + my ($cb, $ops, $partial) = fetch_compile(\@want); + return "$tag $cb\r\n" unless $ops; + + # cb is one of fetch_blob, fetch_smsg, fetch_uid + $range_csv = 'bad' if $range_csv !~ $valid_range; + msn_to_uid_range(msn2uid($self), $range_csv); + my $range_info = range_step($self, \$range_csv); + return "$tag $range_info\r\n" if !ref($range_info); + uo2m_hibernate($self) if $cb == \&fetch_blob; # slow, save RAM + $self->long_response($cb, $tag, [], $range_info, $ops, $partial); } -sub seq_fetch_m { # long_response - my ($self, $tag, $ibx, $msgs, $range_info, $want) = @_; - while (!@$msgs) { # rare - if (my $end = refill_range($ibx, $msgs, $range_info)) { - $self->write(\"$tag $end\r\n"); - return; - } - } - my $seq = $want->{-seqno}++; - my $cur_num = $msgs->[0]->{num}; - if ($cur_num == $seq) { # as expected - git_async_cat($ibx->git, $msgs->[0]->{blob}, - \&uid_fetch_cb, \@_); - } elsif ($cur_num > $seq) { - # send dummy messages until $seq catches up to $cur_num - my $smsg = bless { num => $seq, ts => 0 }, 'PublicInbox::Smsg'; - unshift @$msgs, $smsg; - my $bref = dummy_message($seq, $ibx); - uid_fetch_cb($bref, undef, undef, undef, \@_); - $smsg; # blessed response since uid_fetch_cb requeues - } else { # should not happen - die "BUG: cur_num=$cur_num < seq=$seq"; - } +sub msn_convert ($$) { + my ($self, $uids) = @_; + my $adj = $self->{uid_base} + 1; + my $uo2m = uo2m_extend($self, $uids->[-1]); + $uo2m = [ unpack('S*', $uo2m) ] if !ref($uo2m); + $_ = $uo2m->[$_ - $adj] for @$uids; } -sub cmd_fetch ($$$;@) { - my ($self, $tag, $range_csv, @want) = @_; - my $args = fetch_common($self, $tag, $range_csv, \@want); - ref($args) eq 'ARRAY' ? do { - my $want = $args->[-1]; - $want->{-seqno} = $args->[3]->[0]; # $beg == $range_info->[0]; - long_response($self, \&seq_fetch_m, @$args) - } : $args; # error -} - -sub uid_search_all { # long_response - my ($self, $tag, $ibx, $num) = @_; - my $uids = $ibx->mm->ids_after($num); - if (scalar(@$uids)) { - $self->msg_more(join(' ', '', @$uids)); - } else { - $self->write(\"\r\n$tag OK Search done\r\n"); - undef; +sub search_uid_range { # long_response + my ($self, $tag, $sql, $range_info, $want_msn) = @_; + my $uids = []; + if (defined(my $err = refill_uids($self, $uids, $range_info, $sql))) { + $err ||= 'OK Search done'; + $self->write("\r\n$tag $err\r\n"); + return; } + msn_convert($self, $uids) if $want_msn; + $self->msg_more(join(' ', '', @$uids)); + 1; # more } -sub uid_search_uid_range { # long_response - my ($self, $tag, $ibx, $beg, $end) = @_; - my $uids = $ibx->mm->msg_range($beg, $end, 'num'); - if (@$uids) { - $self->msg_more(join('', map { " $_->[0]" } @$uids)); - } else { - $self->write(\"\r\n$tag OK Search done\r\n"); - undef; +sub parse_imap_query ($$) { + my ($self, $query) = @_; + my $q = PublicInbox::IMAPsearchqp::parse($self, $query); + if (ref($q)) { + my $max = $self->{ibx}->over(1)->max; + my $beg = 1; + uid_clamp($self, \$beg, \$max); + $q->{range_info} = [ $beg, $max ]; } + $q; } -sub cmd_uid_search ($$$;) { - my ($self, $tag, $arg, @rest) = @_; +sub search_common { + my ($self, $tag, $query, $want_msn) = @_; my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n"; - $arg = uc($arg); - if ($arg eq 'ALL' && !@rest) { + my $q = parse_imap_query($self, $query); + return "$tag $q\r\n" if !ref($q); + my ($sql, $range_info) = delete @$q{qw(sql range_info)}; + if (!scalar(keys %$q)) { # overview.sqlite3 $self->msg_more('* SEARCH'); - my $num = 0; - long_response($self, \&uid_search_all, $tag, $ibx, \$num); - } elsif ($arg eq 'UID' && scalar(@rest) == 1) { - if ($rest[0] =~ /\A([0-9]+):([0-9]+|\*)\z/s) { - my ($beg, $end) = ($1, $2); - $end = $ibx->mm->max if $end eq '*'; - $self->msg_more('* SEARCH'); - long_response($self, \&uid_search_uid_range, - $tag, $ibx, \$beg, $end); - } elsif ($rest[0] =~ /\A[0-9]+\z/s) { - my $uid = $rest[0]; - $uid = $ibx->over->get_art($uid) ? " $uid" : ''; - "* SEARCH$uid\r\n$tag OK Search done\r\n"; - } else { - "$tag BAD Error\r\n"; - } + $self->long_response(\&search_uid_range, + $tag, $sql, $range_info, $want_msn); + } elsif ($q = $q->{xap}) { + my $srch = $self->{ibx}->isrch or + return "$tag BAD search not available for mailbox\r\n"; + my $opt = { + relevance => -1, + limit => UID_SLICE, + uid_range => $range_info + }; + my $mset = $srch->mset($q, $opt); + my $uids = $srch->mset_to_artnums($mset, $opt); + msn_convert($self, $uids) if scalar(@$uids) && $want_msn; + "* SEARCH @$uids\r\n$tag OK Search done\r\n"; } else { "$tag BAD Error\r\n"; } } -sub args_ok ($$) { # duplicated from PublicInbox::NNTP - my ($cb, $argc) = @_; - my $tot = prototype $cb; - my ($nreq, undef) = split(';', $tot); - $nreq = ($nreq =~ tr/$//) - 1; - $tot = ($tot =~ tr/$//) - 1; - ($argc <= $tot && $argc >= $nreq); +sub cmd_uid_search ($$$) { + my ($self, $tag, $query) = @_; + search_common($self, $tag, $query); +} + +sub cmd_search ($$$;) { + my ($self, $tag, $query) = @_; + search_common($self, $tag, $query, 1); } # returns 1 if we can continue, 0 if not due to buffered writes or disconnect sub process_line ($$) { my ($self, $l) = @_; + + # TODO: IMAP allows literals for big requests to upload messages + # (which we don't support) but maybe some big search queries use it. + # RFC 3501 9 (2) doesn't permit TAB or multiple SP my ($tag, $req, @args) = parse_line('[ \t]+', 0, $l); pop(@args) if (@args && !defined($args[-1])); if (@args && uc($req) eq 'UID') { $req .= "_".(shift @args); } my $res = eval { - if (my $cmd = $self->can('cmd_'.lc($req // ''))) { - defined($self->{-idle_tag}) ? - "$self->{-idle_tag} BAD expected DONE\r\n" : - $cmd->($self, $tag, @args); - } elsif (uc($tag // '') eq 'DONE' && !defined($req)) { - cmd_done($self, $tag); + if (defined(my $idle_tag = $self->{-idle_tag})) { + (uc($tag // '') eq 'DONE' && !defined($req)) ? + idle_done($self, $tag) : + "$idle_tag BAD expected DONE\r\n"; + } elsif (my $cmd = $self->can('cmd_'.lc($req // ''))) { + if ($cmd == \&cmd_uid_search || $cmd == \&cmd_search) { + # preserve user-supplied quotes for search + (undef, @args) = split(/ search /i, $l, 2); + } + $cmd->($self, $tag, @args); } else { # this is weird auth_challenge_ok($self) // - "$tag BAD Error in IMAP command $req: ". - "Unknown command\r\n"; + ($tag // '*') . + ' BAD Error in IMAP command '. + ($req // '(???)'). + ": Unknown command\r\n"; } }; my $err = $@; if ($err && $self->{sock}) { $l =~ s/\r?\n//s; err($self, 'error from: %s (%s)', $l, $err); + $tag //= '*'; $res = "$tag BAD program fault - command not performed\r\n"; } return 0 unless defined $res; $self->write($res); } -sub long_step { - my ($self) = @_; - # wbuf is unset or empty, here; {long} may add to it - my ($fd, $cb, $t0, @args) = @{$self->{long_cb}}; - my $more = eval { $cb->($self, @args) }; - if ($@ || !$self->{sock}) { # something bad happened... - delete $self->{long_cb}; - my $elapsed = now() - $t0; - if ($@) { - err($self, - "%s during long response[$fd] - %0.6f", - $@, $elapsed); - } - out($self, " deferred[$fd] aborted - %0.6f", $elapsed); - $self->close; - } elsif ($more) { # $self->{wbuf}: - $self->update_idle_time; - - # control passed to $more may be a GitAsyncCat object - requeue_once($self) if !ref($more); - } else { # all done! - delete $self->{long_cb}; - my $elapsed = now() - $t0; - my $fd = fileno($self->{sock}); - out($self, " deferred[$fd] done - %0.6f", $elapsed); - my $wbuf = $self->{wbuf}; # do NOT autovivify - - $self->requeue unless $wbuf && @$wbuf; - } -} - sub err ($$;@) { my ($self, $fmt, @args) = @_; printf { $self->{imapd}->{err} } $fmt."\n", @args; @@ -874,33 +1175,24 @@ sub out ($$;@) { printf { $self->{imapd}->{out} } $fmt."\n", @args; } -sub long_response ($$;@) { - my ($self, $cb, @args) = @_; # cb returns true if more, false if done - - my $sock = $self->{sock} or return; - # make sure we disable reading during a long response, - # clients should not be sending us stuff and making us do more - # work while we are stream a response to them - $self->{long_cb} = [ fileno($sock), $cb, now(), @args ]; - long_step($self); # kick off! - undef; -} - # callback used by PublicInbox::DS for any (e)poll (in/out/hup/err) sub event_step { my ($self) = @_; return unless $self->flush_write && $self->{sock} && !$self->{long_cb}; - $self->update_idle_time; # only read more requests if we've drained the write buffer, # otherwise we can be buffering infinitely w/o backpressure my $rbuf = $self->{rbuf} // \(my $x = ''); my $line = index($$rbuf, "\n"); while ($line < 0) { - return $self->close if length($$rbuf) >= LINE_MAX; - $self->do_read($rbuf, LINE_MAX, length($$rbuf)) or return; + if (length($$rbuf) >= LINE_MAX) { + $self->write(\"\* BAD request too long\r\n"); + return $self->close; + } + $self->do_read($rbuf, LINE_MAX, length($$rbuf)) or + return uo2m_hibernate($self); $line = index($$rbuf, "\n"); } $line = substr($$rbuf, 0, $line + 1, ''); @@ -914,7 +1206,6 @@ sub event_step { return $self->close if $r < 0; $self->rbuf_idle($rbuf); - $self->update_idle_time; # maybe there's more pipelined data, or we'll have # to register it for socket-readiness notifications @@ -923,8 +1214,6 @@ sub event_step { sub compressed { undef } -sub zflush {} # overridden by IMAPdeflate - # RFC 4978 sub cmd_compress ($$$) { my ($self, $tag, $alg) = @_; @@ -953,18 +1242,20 @@ sub cmd_starttls ($$) { undef; } -# for graceful shutdown in PublicInbox::Daemon: -sub busy { - my ($self, $now) = @_; - ($self->{rbuf} || $self->{wbuf} || $self->not_idle_long($now)); +sub busy { # for graceful shutdown in PublicInbox::Daemon: + my ($self) = @_; + if (defined($self->{-idle_tag})) { + $self->write(\"* BYE server shutting down\r\n"); + return; # not busy anymore + } + defined($self->{rbuf}) || defined($self->{wbuf}) || + !$self->write(\"* BYE server shutting down\r\n"); } sub close { my ($self) = @_; if (my $ibx = delete $self->{ibx}) { - if (my $sock = $self->{sock}) {; - $ibx->unsubscribe_unlock(fileno($sock)); - } + stop_idle($self, $ibx); } $self->SUPER::close; # PublicInbox::DS::close } @@ -973,4 +1264,9 @@ sub close { no warnings 'once'; *cmd_select = \&cmd_examine; +package PublicInbox::IMAP_preauth; +our @ISA = qw(PublicInbox::IMAP); + +sub logged_in { 0 } + 1;