X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FGitHTTPBackend.pm;h=09411048db25a8d4fcb41237d4285fa46abd9960;hb=cd50d183273c105a7f08b1875ba6f7a51d9f8e9a;hp=30efa839ac3abb2cd0b8692478c29c8839df13dd;hpb=0b0d9b7f6237593c69dc38ec64566343fd20191c;p=public-inbox.git diff --git a/lib/PublicInbox/GitHTTPBackend.pm b/lib/PublicInbox/GitHTTPBackend.pm index 30efa839..09411048 100644 --- a/lib/PublicInbox/GitHTTPBackend.pm +++ b/lib/PublicInbox/GitHTTPBackend.pm @@ -1,26 +1,24 @@ -# Copyright (C) 2016 all contributors +# Copyright (C) 2016-2018 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; use Fcntl qw(:seek); -use IO::File; -use PublicInbox::Spawn qw(spawn); +use IO::Handle; +use HTTP::Date qw(time2str); +use HTTP::Status qw(status_message); +use Plack::Util; +use PublicInbox::Qspawn; -# TODO: make configurable, but keep in mind it's better to have -# multiple -httpd worker processes which are already scaled to -# the proper number of CPUs and memory. git-pack-objects(1) may -# also use threads and bust memory limits, too, so I recommend -# limiting threads to 1 (via `pack.threads` knob in git) for serving. -my $LIMIT = 1; -my $nr_running = 0; +# 32 is same as the git-daemon connection limit +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]; @@ -29,78 +27,123 @@ my @binary = qw! objects/pack/pack-[a-f0-9]{40}\.(?:pack|idx) !; -our $ANY = join('|', @binary, @text); +our $ANY = join('|', @binary, @text, 'git-upload-pack'); my $BIN = join('|', @binary); my $TEXT = join('|', @text); -sub r { - [ $_[0] , [qw(Content-Type text/plain Content-Length 0) ], [] ] +my @no_cache = ('Expires', 'Fri, 01 Jan 1980 00:00:00 GMT', + 'Pragma', 'no-cache', + 'Cache-Control', 'no-cache, max-age=0, must-revalidate'); + +sub r ($;$) { + my ($code, $msg) = @_; + $msg ||= status_message($code); + my $len = length($msg); + [ $code, [qw(Content-Type text/plain Content-Length), $len, @no_cache], + [$msg] ] } sub serve { - my ($cgi, $git, $path) = @_; - return serve_dumb($cgi, $git, $path) if $nr_running >= $LIMIT; + my ($env, $git, $path) = @_; + + # XXX compatibility... ugh, can we stop supporting this? + $git = PublicInbox::Git->new($git) unless ref($git); - my $service = $cgi->param('service') || ''; - if ($service =~ /\Agit-\w+-pack\z/ || $path =~ /\Agit-\w+-pack\z/) { - my $ok = serve_smart($cgi, $git, $path); + # Documentation/technical/http-protocol.txt in git.git + # requires one and exactly one query parameter: + if ($env->{QUERY_STRING} =~ /\Aservice=git-\w+-pack\z/ || + $path =~ /\Agit-\w+-pack\z/) { + my $ok = serve_smart($env, $git, $path); return $ok if $ok; } - serve_dumb($cgi, $git, $path); + serve_dumb($env, $git, $path); } -sub serve_dumb { - my ($cgi, $git, $path) = @_; +sub err ($@) { + my ($env, @msg) = @_; + $env->{'psgi.errors'}->print(@msg, "\n"); +} - my $type; - if ($path =~ /\A(?:$BIN)\z/o) { - $type = 'application/octet-stream'; - } elsif ($path =~ /\A(?:$TEXT)\z/o) { - $type = 'text/plain'; - } else { - return r(404); +sub drop_client ($) { + if (my $io = $_[0]->{'psgix.io'}) { + $io->close; # this is PublicInbox::DS::close } - my $f = "$git->{git_dir}/$path"; - return r(404) unless -f $f && -r _; - my @st = stat(_); - my $size = $st[7]; +} + +my $prev = 0; +my $exp; +sub cache_one_year { + my ($h) = @_; + my $t = time + 31536000; + push @$h, 'Expires', $t == $prev ? $exp : ($exp = time2str($prev = $t)), + 'Cache-Control', 'public, max-age=31536000'; +} - # TODO: If-Modified-Since and Last-Modified +sub static_result ($$$$) { + my ($env, $h, $f, $type) = @_; + return r(404) unless -f $f && -r _; # just in case it's a FIFO :P + + # TODO: If-Modified-Since and Last-Modified? open my $in, '<', $f or return r(404); - my $code = 200; + my $size = -s $in; my $len = $size; - my @h; - - my $env = $cgi->{env}; - my $range = $env->{HTTP_RANGE}; - if (defined $range && $range =~ /\bbytes=(\d*)-(\d*)\z/) { - ($code, $len) = prepare_range($cgi, $in, \@h, $1, $2, $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); if ($code == 416) { - push @h, 'Content-Range', "bytes */$size"; - return [ 416, \@h, [] ]; + push @$h, 'Content-Range', "bytes */$size"; + return [ 416, $h, [] ]; } } - - push @h, 'Content-Type', $type, 'Content-Length', $len; - sub { - my ($res) = @_; # Plack callback - my $fh = $res->([ $code, \@h ]); - my $buf; - my $n = 8192; - while ($len > 0) { + push @$h, 'Content-Length', $len; + my $n = 65536; + [ $code, $h, Plack::Util::inline_object(close => sub { close $in }, + getline => sub { + return if $len == 0; $n = $len if $len < $n; - my $r = sysread($in, $buf, $n); - last if (!defined($r) || $r <= 0); - $len -= $r; - $fh->write($buf); - } - $fh->close; + my $r = sysread($in, my $buf, $n); + if (!defined $r) { + err($env, "$f read error: $!"); + } elsif ($r <= 0) { + err($env, "$f EOF with $len bytes left"); + } else { + $len -= $r; + $n = 8192; + return $buf; + } + drop_client($env); + return; + })] +} + +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 ($cgi, $in, $h, $beg, $end, $size) = @_; + my ($env, $in, $h, $beg, $end, $size) = @_; my $code = 200; my $len = $size; if ($beg eq '') { @@ -129,12 +172,12 @@ sub prepare_range { if ($len <= 0) { $code = 416; } else { - seek($in, $beg, SEEK_SET) or return [ 500, [], [] ]; + sysseek($in, $beg, SEEK_SET) or return [ 500, [], [] ]; push @$h, qw(Accept-Ranges bytes Content-Range); push @$h, "bytes $beg-$end/$size"; # FIXME: Plack::Middleware::Deflater bug? - $cgi->{env}->{'psgix.no-compress'} = 1; + $env->{'psgix.no-compress'} = 1; } } ($code, $len); @@ -142,24 +185,7 @@ sub prepare_range { # returns undef if 403 so it falls back to dumb HTTP sub serve_smart { - my ($cgi, $git, $path) = @_; - my $env = $cgi->{env}; - - my $input = $env->{'psgi.input'}; - my $buf; - my $in; - my $err = $env->{'psgi.errors'}; - my $fd = eval { fileno($input) }; - if (defined $fd && $fd >= 0) { - $in = $input; - } else { - $in = input_to_file($env) or return r(500); - } - my ($rpipe, $wpipe); - unless (pipe($rpipe, $wpipe)) { - $err->print("error creating pipe: $! - going static\n"); - return; - } + my ($env, $git, $path) = @_; my %env = %ENV; # GIT_COMMITTER_NAME, GIT_COMMITTER_EMAIL # may be set in the server-process and are passed as-is @@ -172,109 +198,74 @@ sub serve_smart { my $val = $env->{$name}; $env{$name} = $val if defined $val; } - my $git_dir = $git->{git_dir}; + my $limiter = $git->{-httpbackend_limiter} || $default_limiter; $env{GIT_HTTP_EXPORT_ALL} = '1'; - $env{PATH_TRANSLATED} = "$git_dir/$path"; - my %rdr = ( 0 => fileno($in), 1 => fileno($wpipe) ); - my $pid = spawn([qw(git http-backend)], \%env, \%rdr); - unless (defined $pid) { - $err->print("error spawning: $! - going static\n"); - return; - } - $wpipe = $in = undef; - $buf = ''; - my ($vin, $fh, $res); - $nr_running++; - my $end = sub { - if ($fh) { - $fh->close; - $fh = undef; - } - if ($rpipe) { - $rpipe->close; # _may_ be Danga::Socket::close - $rpipe = undef; - $nr_running--; - } - if (defined $pid && $pid != waitpid($pid, 0)) { - $err->print("git http-backend ($git_dir): $?\n"); - } else { - $pid = undef; - } - return unless $res; - my $dumb = serve_dumb($cgi, $git, $path); - ref($dumb) eq 'ARRAY' ? $res->($dumb) : $dumb->($res); - }; - my $fail = sub { - if ($!{EAGAIN} || $!{EINTR}) { - select($vin, undef, undef, undef) if defined $vin; - # $vin is undef on async, so this is a noop on EAGAIN - return; - } - my $e = $!; - $end->(); - $err->print("git http-backend ($git_dir): $e\n"); - }; - my $cb = sub { # read git-http-backend output and stream to client - my $r = $rpipe ? $rpipe->sysread($buf, 8192, length($buf)) : 0; - return $fail->() unless defined $r; - return $end->() if $r == 0; # EOF - if ($fh) { # stream body from git-http-backend to HTTP client - $fh->write($buf); - $buf = ''; - } elsif ($buf =~ s/\A(.*?)\r\n\r\n//s) { # parse headers - my $h = $1; - my $code = 200; - my @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/); - } else { - push @h, $k, $v; - } - } - if ($code == 403) { - # smart cloning disabled, serve dumbly - # in $end since we never undef $res in here - } else { # write response header: - $fh = $res->([ $code, \@h ]); - $res = undef; - $fh->write($buf); - } - $buf = ''; - } # else { keep reading ... } - }; - if (my $async = $env->{'pi-httpd.async'}) { - $rpipe = $async->($rpipe, $cb); - sub { ($res) = @_ } # let Danga::Socket handle the rest. - } else { # synchronous loop for other PSGI servers - $vin = ''; - vec($vin, fileno($rpipe), 1) = 1; - sub { - ($res) = @_; - while ($rpipe) { $cb->() } - } - } + $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 $in = IO::File->new_tmpfile; + 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 $buf; while (1) { my $r = $input->read($buf, 8192); unless (defined $r) { - my $err = $env->{'psgi.errors'}; - $err->print("error reading input: $!\n"); + err($env, "error reading input: $!"); return; } - last if ($r == 0); - $in->write($buf); + last if $r == 0; + 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; + } + } + } + unless (defined(sysseek($in, 0, SEEK_SET))) { + err($env, "error seeking temporary file: $!"); + return; + } + { 0 => fileno($in), -hold => $in }; +} + +sub parse_cgi_headers { + 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/, $h)) { + my ($k, $v) = split(/:\s*/, $l, 2); + if ($k =~ /\AStatus\z/i) { + ($code) = ($v =~ /\b(\d+)\b/); + } else { + push @h, $k, $v; + } } - $in->flush; - $in->sysseek(0, SEEK_SET); - return $in; + [ $code, \@h ] } 1;