X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FGitHTTPBackend.pm;h=c8878145b8c35488aa2ad32748bc5049dbda7c3b;hb=9bd675d33ad1e49bd2ebe12a1d216216e61380de;hp=a069fd9473649c09c970e358aff3b620307be2b0;hpb=15e14f11af0b9919f11e0c186b365ae0154e7e77;p=public-inbox.git diff --git a/lib/PublicInbox/GitHTTPBackend.pm b/lib/PublicInbox/GitHTTPBackend.pm index a069fd94..c8878145 100644 --- a/lib/PublicInbox/GitHTTPBackend.pm +++ b/lib/PublicInbox/GitHTTPBackend.pm @@ -1,8 +1,8 @@ -# Copyright (C) 2016 all contributors +# Copyright (C) 2016-2019 all contributors # License: AGPL-3.0+ # when no endpoints match, fallback to this and serve a static file -# or smart HTTP +# or smart HTTP. This is our wrapper for git-http-backend(1) package PublicInbox::GitHTTPBackend; use strict; use warnings; @@ -18,7 +18,7 @@ my $default_limiter = PublicInbox::Qspawn::Limiter->new(32); # n.b. serving "description" and "cloneurl" should be innocuous enough to # not cause problems. serving "config" might... -my @text = qw[HEAD info/refs +my @text = qw[HEAD info/refs info/attributes objects/info/(?:http-alternates|alternates|packs) cloneurl description]; @@ -46,10 +46,13 @@ sub r ($;$) { sub serve { my ($env, $git, $path) = @_; + # XXX compatibility... ugh, can we stop supporting this? + $git = PublicInbox::Git->new($git) unless ref($git); + # Documentation/technical/http-protocol.txt in git.git # requires one and exactly one query parameter: - if ($env->{QUERY_STRING} =~ /\Aservice=git-\w+-pack\z/ || - $path =~ /\Agit-\w+-pack\z/) { + if ($env->{QUERY_STRING} =~ /\Aservice=git-[A-Za-z0-9_]+-pack\z/ || + $path =~ /\Agit-[A-Za-z0-9_]+-pack\z/) { my $ok = serve_smart($env, $git, $path); return $ok if $ok; } @@ -64,7 +67,7 @@ sub err ($@) { sub drop_client ($) { if (my $io = $_[0]->{'psgix.io'}) { - $io->close; # this is Danga::Socket::close + $io->close; # this is PublicInbox::DS::close } } @@ -77,46 +80,26 @@ sub cache_one_year { 'Cache-Control', 'public, max-age=31536000'; } -sub serve_dumb { - my ($env, $git, $path) = @_; - - my @h; - my $type; - if ($path =~ m!\Aobjects/[a-f0-9]{2}/[a-f0-9]{38}\z!) { - $type = 'application/x-git-loose-object'; - cache_one_year(\@h); - } elsif ($path =~ m!\Aobjects/pack/pack-[a-f0-9]{40}\.pack\z!) { - $type = 'application/x-git-packed-objects'; - cache_one_year(\@h); - } elsif ($path =~ m!\Aobjects/pack/pack-[a-f0-9]{40}\.idx\z!) { - $type = 'application/x-git-packed-objects-toc'; - cache_one_year(\@h); - } elsif ($path =~ /\A(?:$TEXT)\z/o) { - $type = 'text/plain'; - push @h, @no_cache; - } else { - return r(404); - } - - my $f = (ref $git ? $git->{git_dir} : $git) . '/' . $path; +sub static_result ($$$$) { + my ($env, $h, $f, $type) = @_; return r(404) unless -f $f && -r _; # just in case it's a FIFO :P - my $size = -s _; # TODO: If-Modified-Since and Last-Modified? open my $in, '<', $f or return r(404); + my $size = -s $in; my $len = $size; my $code = 200; - push @h, 'Content-Type', $type; - if (($env->{HTTP_RANGE} || '') =~ /\bbytes=(\d*)-(\d*)\z/) { - ($code, $len) = prepare_range($env, $in, \@h, $1, $2, $size); + push @$h, 'Content-Type', $type; + if (($env->{HTTP_RANGE} || '') =~ /\bbytes=([0-9]*)-([0-9]*)\z/) { + ($code, $len) = prepare_range($env, $in, $h, $1, $2, $size); if ($code == 416) { - push @h, 'Content-Range', "bytes */$size"; - return [ 416, \@h, [] ]; + push @$h, 'Content-Range', "bytes */$size"; + return [ 416, $h, [] ]; } } - push @h, 'Content-Length', $len; + push @$h, 'Content-Length', $len; my $n = 65536; - [ $code, \@h, Plack::Util::inline_object(close => sub { close $in }, + [ $code, $h, Plack::Util::inline_object(close => sub { close $in }, getline => sub { return if $len == 0; $n = $len if $len < $n; @@ -135,6 +118,30 @@ sub serve_dumb { })] } +sub serve_dumb { + my ($env, $git, $path) = @_; + + my $h = []; + my $type; + if ($path =~ m!\Aobjects/[a-f0-9]{2}/[a-f0-9]{38}\z!) { + $type = 'application/x-git-loose-object'; + cache_one_year($h); + } elsif ($path =~ m!\Aobjects/pack/pack-[a-f0-9]{40}\.pack\z!) { + $type = 'application/x-git-packed-objects'; + cache_one_year($h); + } elsif ($path =~ m!\Aobjects/pack/pack-[a-f0-9]{40}\.idx\z!) { + $type = 'application/x-git-packed-objects-toc'; + cache_one_year($h); + } elsif ($path =~ /\A(?:$TEXT)\z/o) { + $type = 'text/plain'; + push @$h, @no_cache; + } else { + return r(404); + } + + static_result($env, $h, "$git->{git_dir}/$path", $type); +} + sub prepare_range { my ($env, $in, $h, $beg, $end, $size) = @_; my $code = 200; @@ -179,11 +186,6 @@ sub prepare_range { # returns undef if 403 so it falls back to dumb HTTP sub serve_smart { my ($env, $git, $path) = @_; - my $in = $env->{'psgi.input'}; - my $fd = eval { fileno($in) }; - unless (defined $fd && $fd >= 0) { - $in = input_to_file($env) or return r(500); - } my %env = %ENV; # GIT_COMMITTER_NAME, GIT_COMMITTER_EMAIL # may be set in the server-process and are passed as-is @@ -196,88 +198,31 @@ sub serve_smart { my $val = $env->{$name}; $env{$name} = $val if defined $val; } - my ($git_dir, $limiter); - if (ref $git) { - $limiter = $git->{-httpbackend_limiter} || $default_limiter; - $git_dir = $git->{git_dir}; - } else { - $limiter = $default_limiter; - $git_dir = $git; - } + my $limiter = $git->{-httpbackend_limiter} || $default_limiter; $env{GIT_HTTP_EXPORT_ALL} = '1'; - $env{PATH_TRANSLATED} = "$git_dir/$path"; - my %rdr = ( 0 => fileno($in) ); - my $x = PublicInbox::Qspawn->new([qw(git http-backend)], \%env, \%rdr); - my ($fh, $rpipe); - my $end = sub { - if (my $err = $x->finish) { - err($env, "git http-backend ($git_dir): $err"); - } - $fh->close if $fh; # async-only - }; - - # Danga::Socket users, we queue up the read_enable callback to - # fire after pending writes are complete: - my $buf = ''; - my $rd_hdr = sub { - my $r = sysread($rpipe, $buf, 1024, length($buf)); - return if !defined($r) && ($!{EINTR} || $!{EAGAIN}); - return r(500, 'http-backend error') unless $r; - $r = parse_cgi_headers(\$buf) or return; # incomplete headers - $r->[0] == 403 ? serve_dumb($env, $git, $path) : $r; - }; - my $res; - my $async = $env->{'pi-httpd.async'}; # XXX unstable API - my $io = $env->{'psgix.io'}; - my $cb = sub { - my $r = $rd_hdr->() or return; - $rd_hdr = undef; - if (scalar(@$r) == 3) { # error: - if ($async) { - $async->close; # calls rpipe->close - } else { - $rpipe->close; - $end->(); - } - return $res->($r); - } - if ($async) { - $fh = $res->($r); - return $async->async_pass($io, $fh, \$buf); - } - - # for synchronous PSGI servers - require PublicInbox::GetlineBody; - $r->[2] = PublicInbox::GetlineBody->new($rpipe, $end, $buf); - $res->($r); - }; - sub { - ($res) = @_; - - # hopefully this doesn't break any middlewares, - # holding the input here is a waste of FDs and memory - $env->{'psgi.input'} = undef; - - $x->start($limiter, sub { # may run later, much later... - ($rpipe) = @_; - $in = undef; - if ($async) { - $async = $async->($rpipe, $cb, $end); - } else { # generic PSGI - $cb->() while $rd_hdr; - } - }); - }; + $env{PATH_TRANSLATED} = "$git->{git_dir}/$path"; + my $rdr = input_prepare($env) or return r(500); + my $qsp = PublicInbox::Qspawn->new([qw(git http-backend)], \%env, $rdr); + $qsp->psgi_return($env, $limiter, sub { + my ($r, $bref) = @_; + my $res = parse_cgi_headers($r, $bref) or return; # incomplete + $res->[0] == 403 ? serve_dumb($env, $git, $path) : $res; + }); } -sub input_to_file { +sub input_prepare { my ($env) = @_; + + my $input = $env->{'psgi.input'}; + my $fd = eval { fileno($input) }; + if (defined $fd && $fd >= 0) { + return { 0 => $fd }; + } open(my $in, '+>', undef); unless (defined $in) { err($env, "could not open temporary file: $!"); return; } - my $input = $env->{'psgi.input'}; my $buf; while (1) { my $r = $input->read($buf, 8192); @@ -285,35 +230,35 @@ sub input_to_file { err($env, "error reading input: $!"); return; } - my $off = 0; - while ($r > 0) { - my $w = syswrite($in, $buf, $r, $off); - if (defined $w) { - $r -= $w; - $off += $w; - } else { - err($env, "error writing temporary file: $!"); - return; - } + last if $r == 0; + unless (print $in $buf) { + err($env, "error writing temporary file: $!"); + return; } } + # ensure it's visible to git-http-backend(1): + unless ($in->flush) { + err($env, "error writing temporary file: $!"); + return; + } unless (defined(sysseek($in, 0, SEEK_SET))) { err($env, "error seeking temporary file: $!"); return; } - return $in; + { 0 => fileno($in), -hold => $in }; } sub parse_cgi_headers { - my ($bref) = @_; - $$bref =~ s/\A(.*?)\r\n\r\n//s or return; + my ($r, $bref) = @_; + return r(500) unless defined $r && $r >= 0; + $$bref =~ s/\A(.*?)\r?\n\r?\n//s or return $r == 0 ? r(500) : undef; my $h = $1; my $code = 200; my @h; - foreach my $l (split(/\r\n/, $h)) { + foreach my $l (split(/\r?\n/, $h)) { my ($k, $v) = split(/:\s*/, $l, 2); if ($k =~ /\AStatus\z/i) { - ($code) = ($v =~ /\b(\d+)\b/); + ($code) = ($v =~ /\b([0-9]+)\b/); } else { push @h, $k, $v; }