user/dev discussion of public-inbox itself
 help / color / mirror / code / Atom feed
508d43fd762b2d7c59c2480097465efca1424fef blob 8261 bytes (raw)

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
 
# Copyright (C) 2016-2020 all contributors <meta@public-inbox.org>
# License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
#
# This allows vfork to be used for spawning subprocesses if
# ~/.cache/public-inbox/inline-c is writable or if PERL_INLINE_DIRECTORY
# is explicitly defined in the environment (and writable).
# Under Linux, vfork can make a big difference in spawning performance
# as process size increases (fork still needs to mark pages for CoW use).
# Currently, we only use this for code intended for long running
# daemons (inside the PSGI code (-httpd) and -nntpd).  The short-lived
# scripts (-mda, -index, -learn, -init) either use IPC::run or standard
# Perl routines.
#
# There'll probably be more OS-level C stuff here, down the line.
# We don't want too many DSOs: https://udrepper.livejournal.com/8790.html

package PublicInbox::Spawn;
use strict;
use parent qw(Exporter);
use Symbol qw(gensym);
use PublicInbox::ProcessPipe;
our @EXPORT_OK = qw/which spawn popen_rd/;
our @RLIMITS = qw(RLIMIT_CPU RLIMIT_CORE RLIMIT_DATA);

my $vfork_spawn = <<'VFORK_SPAWN';
#include <sys/types.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>

/* some platforms need alloca.h, but some don't */
#if defined(__GNUC__) && !defined(alloca)
#  define alloca(sz) __builtin_alloca(sz)
#endif

#include <signal.h>
#include <assert.h>

/*
 * From the av_len apidoc:
 *   Note that, unlike what the name implies, it returns
 *   the highest index in the array, so to get the size of
 *   the array you need to use "av_len(av) + 1".
 *   This is unlike "sv_len", which returns what you would expect.
 */
#define AV2C_COPY(dst, src) do { \
	I32 i; \
	I32 top_index = av_len(src); \
	I32 real_len = top_index + 1; \
	I32 capa = real_len + 1; \
	dst = alloca(capa * sizeof(char *)); \
	for (i = 0; i < real_len; i++) { \
		SV **sv = av_fetch(src, i, 0); \
		dst[i] = SvPV_nolen(*sv); \
	} \
	dst[real_len] = 0; \
} while (0)

/* needs to be safe inside a vfork'ed process */
static void exit_err(int *cerrnum)
{
	*cerrnum = errno;
	_exit(1);
}

/*
 * unstable internal API.  It'll be updated depending on
 * whatever we'll need in the future.
 * Be sure to update PublicInbox::SpawnPP if this changes
 */
int pi_fork_exec(SV *redirref, SV *file, SV *cmdref, SV *envref, SV *rlimref,
		 const char *cd)
{
	AV *redir = (AV *)SvRV(redirref);
	AV *cmd = (AV *)SvRV(cmdref);
	AV *env = (AV *)SvRV(envref);
	AV *rlim = (AV *)SvRV(rlimref);
	const char *filename = SvPV_nolen(file);
	pid_t pid;
	char **argv, **envp;
	sigset_t set, old, cset;
	int ret, perrnum, cerrnum = 0;

	AV2C_COPY(argv, cmd);
	AV2C_COPY(envp, env);

	ret = sigfillset(&set);
	assert(ret == 0 && "BUG calling sigfillset");
	ret = sigprocmask(SIG_SETMASK, &set, &old);
	assert(ret == 0 && "BUG calling sigprocmask to block");
	ret = sigemptyset(&cset);
	assert(ret == 0 && "BUG calling sigemptyset");
	ret = sigaddset(&cset, SIGCHLD);
	assert(ret == 0 && "BUG calling sigaddset for SIGCHLD");
	pid = vfork();
	if (pid == 0) {
		int sig;
		I32 i, child_fd, max = av_len(redir);

		for (child_fd = 0; child_fd <= max; child_fd++) {
			SV **parent = av_fetch(redir, child_fd, 0);
			int parent_fd = SvIV(*parent);
			if (parent_fd == child_fd)
				continue;
			if (dup2(parent_fd, child_fd) < 0)
				exit_err(&cerrnum);
		}
		for (sig = 1; sig < NSIG; sig++)
			signal(sig, SIG_DFL); /* ignore errors on signals */
		if (*cd && chdir(cd) < 0)
			exit_err(&cerrnum);

		max = av_len(rlim);
		for (i = 0; i < max; i += 3) {
			struct rlimit rl;
			SV **res = av_fetch(rlim, i, 0);
			SV **soft = av_fetch(rlim, i + 1, 0);
			SV **hard = av_fetch(rlim, i + 2, 0);

			rl.rlim_cur = SvIV(*soft);
			rl.rlim_max = SvIV(*hard);
			if (setrlimit(SvIV(*res), &rl) < 0)
				exit_err(&cerrnum);
		}

		/*
		 * don't bother unblocking other signals for now, just SIGCHLD.
		 * we don't want signals to the group taking out a subprocess
		 */
		(void)sigprocmask(SIG_UNBLOCK, &cset, NULL);
		execve(filename, argv, envp);
		exit_err(&cerrnum);
	}
	perrnum = errno;
	ret = sigprocmask(SIG_SETMASK, &old, NULL);
	assert(ret == 0 && "BUG calling sigprocmask to restore");
	if (cerrnum) {
		if (pid > 0)
			waitpid(pid, NULL, 0);
		pid = -1;
		errno = cerrnum;
	} else if (perrnum) {
		errno = perrnum;
	}
	return (int)pid;
}
VFORK_SPAWN

# btrfs on Linux is copy-on-write (COW) by default.  As of Linux 5.7,
# this still leads to fragmentation for SQLite and Xapian files where
# random I/O happens, so we disable COW just for SQLite files and Xapian
# directories.  Disabling COW disables checksumming, so we only do this
# for regeneratable files, and not canonical git storage (git doesn't
# checksum refs, only data under $GIT_DIR/objects).
my $set_nodatacow = $^O eq 'linux' ? <<'SET_NODATACOW' : '';
#include <sys/ioctl.h>
#include <sys/vfs.h>
#include <linux/magic.h>
#include <linux/fs.h>
#include <errno.h>
#include <stdio.h>
#include <string.h>

void set_nodatacow(int fd)
{
	struct statfs buf;
	int val = 0;

	if (fstatfs(fd, &buf) < 0) {
		fprintf(stderr, "fstatfs: %s\\n", strerror(errno));
		return;
	}

	/* only btrfs is known to have this problem, so skip for non-btrfs */
	if (buf.f_type != BTRFS_SUPER_MAGIC)
		return;

	if (ioctl(fd, FS_IOC_GETFLAGS, &val) < 0) {
		fprintf(stderr, "FS_IOC_GET_FLAGS: %s\\n", strerror(errno));
		return;
	}
	val |= FS_NOCOW_FL;
	if (ioctl(fd, FS_IOC_SETFLAGS, &val) < 0)
		fprintf(stderr, "FS_IOC_SET_FLAGS: %s\\n", strerror(errno));
}
SET_NODATACOW

my $inline_dir = $ENV{PERL_INLINE_DIRECTORY} //= (
		$ENV{XDG_CACHE_HOME} //
		( ($ENV{HOME} // '/nonexistent').'/.cache' )
	).'/public-inbox/inline-c';

$set_nodatacow = $vfork_spawn = undef unless -d $inline_dir && -w _;
if (defined $vfork_spawn) {
	# Inline 0.64 or later has locking in multi-process env,
	# but we support 0.5 on Debian wheezy
	use Fcntl qw(:flock);
	eval {
		my $f = "$inline_dir/.public-inbox.lock";
		open my $fh, '>', $f or die "failed to open $f: $!\n";
		flock($fh, LOCK_EX) or die "LOCK_EX failed on $f: $!\n";
		eval 'use Inline C => $vfork_spawn . $set_nodatacow';
		my $err = $@;
		my $ndc_err;
		if ($err && $set_nodatacow) { # missing Linux kernel headers
			$ndc_err = $err;
			undef $set_nodatacow;
			eval 'use Inline C => $vfork_spawn';
		}
		flock($fh, LOCK_UN) or die "LOCK_UN failed on $f: $!\n";
		die $err if $err;
		warn $ndc_err if $ndc_err;
	};
	if ($@) {
		warn "Inline::C failed for vfork: $@\n";
		$set_nodatacow = $vfork_spawn = undef;
	}
}

unless (defined $vfork_spawn) {
	require PublicInbox::SpawnPP;
	*pi_fork_exec = \&PublicInbox::SpawnPP::pi_fork_exec
}
unless ($set_nodatacow) {
	require PublicInbox::NDC_PP;
	no warnings 'once';
	*set_nodatacow = \&PublicInbox::NDC_PP::set_nodatacow;
}
undef $set_nodatacow;
undef $vfork_spawn;

sub which ($) {
	my ($file) = @_;
	return $file if index($file, '/') >= 0;
	foreach my $p (split(':', $ENV{PATH})) {
		$p .= "/$file";
		return $p if -x $p;
	}
	undef;
}

sub spawn ($;$$) {
	my ($cmd, $env, $opts) = @_;
	my $f = which($cmd->[0]);
	defined $f or die "$cmd->[0]: command not found\n";
	my @env;
	$opts ||= {};

	my %env = $env ? (%ENV, %$env) : %ENV;
	while (my ($k, $v) = each %env) {
		push @env, "$k=$v";
	}
	my $redir = [];
	for my $child_fd (0..2) {
		my $parent_fd = $opts->{$child_fd};
		if (defined($parent_fd) && $parent_fd !~ /\A[0-9]+\z/) {
			defined(my $fd = fileno($parent_fd)) or
					die "$parent_fd not an IO GLOB? $!";
			$parent_fd = $fd;
		}
		$redir->[$child_fd] = $parent_fd // $child_fd;
	}
	my $rlim = [];

	foreach my $l (@RLIMITS) {
		defined(my $v = $opts->{$l}) or next;
		my $r = eval "require BSD::Resource; BSD::Resource::$l();";
		unless (defined $r) {
			warn "$l undefined by BSD::Resource: $@\n";
			next;
		}
		push @$rlim, $r, @$v;
	}
	my $cd = $opts->{'-C'} // ''; # undef => NULL mapping doesn't work?
	my $pid = pi_fork_exec($redir, $f, $cmd, \@env, $rlim, $cd);
	die "fork_exec @$cmd failed: $!\n" unless $pid > 0;
	$pid;
}

sub popen_rd {
	my ($cmd, $env, $opts) = @_;
	pipe(my ($r, $w)) or die "pipe: $!\n";
	$opts ||= {};
	$opts->{1} = fileno($w);
	my $pid = spawn($cmd, $env, $opts);
	return ($r, $pid) if wantarray;
	my $ret = gensym;
	tie *$ret, 'PublicInbox::ProcessPipe', $pid, $r;
	$ret;
}

1;
debug log:

solving 508d43fd7 ...
found 508d43fd7 in public-inbox.git

user/dev discussion of public-inbox itself

This inbox may be cloned and mirrored by anyone:

	git clone --mirror https://public-inbox.org/meta
	git clone --mirror http://czquwvybam4bgbro.onion/meta
	git clone --mirror http://hjrcffqmbrq6wope.onion/meta
	git clone --mirror http://ou63pmih66umazou.onion/meta

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V1 meta meta/ https://public-inbox.org/meta \
		meta@public-inbox.org
	public-inbox-index meta

Example config snippet for mirrors.
Newsgroups are available over NNTP:
	nntp://news.public-inbox.org/inbox.comp.mail.public-inbox.meta
	nntp://7fh6tueqddpjyxjmgtdiueylzoqt6pt7hec3pukyptlmohoowvhde4yd.onion/inbox.comp.mail.public-inbox.meta
	nntp://ie5yzdi7fg72h7s4sdcztq5evakq23rdt33mfyfcddc5u3ndnw24ogqd.onion/inbox.comp.mail.public-inbox.meta
	nntp://4uok3hntl7oi7b4uf4rtfwefqeexfzil2w6kgk2jn5z2f764irre7byd.onion/inbox.comp.mail.public-inbox.meta
	nntp://news.gmane.io/gmane.mail.public-inbox.general
 note: .onion URLs require Tor: https://www.torproject.org/

code repositories for project(s) associated with this inbox:

	https://80x24.org/public-inbox.git

AGPL code for this site: git clone https://public-inbox.org/public-inbox.git