-# Copyright (C) 2016-2019 all contributors <meta@public-inbox.org>
+# Copyright (C) 2016-2021 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
-# PERL_INLINE_DIRECTORY is explicitly defined in the environment.
+# ~/.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 warnings;
-use base qw(Exporter);
+use parent qw(Exporter);
use Symbol qw(gensym);
-use IO::Handle;
use PublicInbox::ProcessPipe;
-our @EXPORT_OK = qw/which spawn popen_rd/;
-sub RLIMITS () { qw(RLIMIT_CPU RLIMIT_CORE RLIMIT_DATA) }
+our @EXPORT_OK = qw(which spawn popen_rd run_die nodatacow_dir);
+our @RLIMITS = qw(RLIMIT_CPU RLIMIT_CORE RLIMIT_DATA);
-my $vfork_spawn = <<'VFORK_SPAWN';
+my $all_libc = <<'ALL_LIBC'; # all *nix systems we support
+#include <sys/resource.h>
+#include <sys/socket.h>
#include <sys/types.h>
-#include <sys/uio.h>
#include <sys/time.h>
-#include <sys/resource.h>
+#include <sys/uio.h>
#include <unistd.h>
#include <stdlib.h>
+#include <errno.h>
/* some platforms need alloca.h, but some don't */
#if defined(__GNUC__) && !defined(alloca)
dst[real_len] = 0; \
} while (0)
-static void *deconst(const char *s)
-{
- union { const char *in; void *out; } u;
- u.in = s;
- return u.out;
-}
-
/* needs to be safe inside a vfork'ed process */
-static void xerr(const char *msg)
+static void exit_err(const char *fn, volatile int *cerrnum)
{
- struct iovec iov[3];
- const char *err = strerror(errno); /* should be safe in practice */
-
- iov[0].iov_base = deconst(msg);
- iov[0].iov_len = strlen(msg);
- iov[1].iov_base = deconst(err);
- iov[1].iov_len = strlen(err);
- iov[2].iov_base = deconst("\n");
- iov[2].iov_len = 1;
- writev(2, iov, 3);
+ *cerrnum = errno;
+ write(2, fn, strlen(fn));
_exit(1);
}
* 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)
+ const char *cd, int pgid)
{
AV *redir = (AV *)SvRV(redirref);
AV *cmd = (AV *)SvRV(cmdref);
pid_t pid;
char **argv, **envp;
sigset_t set, old;
- int ret, errnum;
+ int ret, perrnum;
+ volatile int cerrnum = 0; /* shared due to vfork */
+ int chld_is_member;
+ I32 max_fd = av_len(redir);
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");
+ if (sigfillset(&set)) return -1;
+ if (sigprocmask(SIG_SETMASK, &set, &old)) return -1;
+ chld_is_member = sigismember(&old, SIGCHLD);
+ if (chld_is_member < 0) return -1;
+ if (chld_is_member > 0)
+ sigdelset(&old, SIGCHLD);
+
pid = vfork();
if (pid == 0) {
int sig;
- I32 i, child_fd, max = av_len(redir);
+ I32 i, child_fd, max_rlim;
- for (child_fd = 0; child_fd <= max; child_fd++) {
+ for (child_fd = 0; child_fd <= max_fd; 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)
- xerr("dup2");
+ exit_err("dup2", &cerrnum);
}
+ if (pgid >= 0 && setpgid(0, pgid) < 0)
+ exit_err("setpgid", &cerrnum);
for (sig = 1; sig < NSIG; sig++)
signal(sig, SIG_DFL); /* ignore errors on signals */
if (*cd && chdir(cd) < 0)
- xerr("chdir");
+ exit_err("chdir", &cerrnum);
- max = av_len(rlim);
- for (i = 0; i < max; i += 3) {
+ max_rlim = av_len(rlim);
+ for (i = 0; i < max_rlim; i += 3) {
struct rlimit rl;
SV **res = av_fetch(rlim, i, 0);
SV **soft = av_fetch(rlim, i + 1, 0);
rl.rlim_cur = SvIV(*soft);
rl.rlim_max = SvIV(*hard);
if (setrlimit(SvIV(*res), &rl) < 0)
- xerr("sertlimit");
+ exit_err("setrlimit", &cerrnum);
}
- /*
- * don't bother unblocking, we don't want signals
- * to the group taking out a subprocess
- */
+ (void)sigprocmask(SIG_SETMASK, &old, NULL);
execve(filename, argv, envp);
- xerr("execve failed");
+ exit_err("execve", &cerrnum);
}
- errnum = errno;
+ perrnum = errno;
+ if (chld_is_member > 0)
+ sigaddset(&old, SIGCHLD);
ret = sigprocmask(SIG_SETMASK, &old, NULL);
assert(ret == 0 && "BUG calling sigprocmask to restore");
- errno = errnum;
-
+ if (cerrnum) {
+ int err_fd = STDERR_FILENO;
+ if (err_fd <= max_fd) {
+ SV **parent = av_fetch(redir, err_fd, 0);
+ err_fd = SvIV(*parent);
+ }
+ if (pid > 0)
+ waitpid(pid, NULL, 0);
+ pid = -1;
+ /* continue message started by exit_err in child */
+ dprintf(err_fd, ": %s\n", strerror(cerrnum));
+ errno = cerrnum;
+ } else if (perrnum) {
+ errno = perrnum;
+ }
return (int)pid;
}
-VFORK_SPAWN
-my $inline_dir = $ENV{PERL_INLINE_DIRECTORY};
-$vfork_spawn = undef unless defined $inline_dir && -d $inline_dir && -w _;
-if (defined $vfork_spawn) {
+#if defined(CMSG_SPACE) && defined(CMSG_LEN)
+#define SEND_FD_CAPA 10
+#define SEND_FD_SPACE (SEND_FD_CAPA * sizeof(int))
+union my_cmsg {
+ struct cmsghdr hdr;
+ char pad[sizeof(struct cmsghdr) + 16 + SEND_FD_SPACE];
+};
+
+SV *send_cmd4(PerlIO *s, SV *svfds, SV *data, int flags)
+{
+ struct msghdr msg = { 0 };
+ union my_cmsg cmsg = { 0 };
+ STRLEN dlen = 0;
+ struct iovec iov;
+ ssize_t sent;
+ AV *fds = (AV *)SvRV(svfds);
+ I32 i, nfds = av_len(fds) + 1;
+ int *fdp;
+
+ if (SvOK(data)) {
+ iov.iov_base = SvPV(data, dlen);
+ iov.iov_len = dlen;
+ }
+ if (!dlen) { /* must be non-zero */
+ iov.iov_base = &msg.msg_namelen; /* whatever */
+ iov.iov_len = 1;
+ }
+ msg.msg_iov = &iov;
+ msg.msg_iovlen = 1;
+ if (nfds) {
+ if (nfds > SEND_FD_CAPA) {
+ fprintf(stderr, "FIXME: bump SEND_FD_CAPA=%d\n", nfds);
+ nfds = SEND_FD_CAPA;
+ }
+ msg.msg_control = &cmsg.hdr;
+ msg.msg_controllen = CMSG_SPACE(nfds * sizeof(int));
+ cmsg.hdr.cmsg_level = SOL_SOCKET;
+ cmsg.hdr.cmsg_type = SCM_RIGHTS;
+ cmsg.hdr.cmsg_len = CMSG_LEN(nfds * sizeof(int));
+ fdp = (int *)CMSG_DATA(&cmsg.hdr);
+ for (i = 0; i < nfds; i++) {
+ SV **fd = av_fetch(fds, i, 0);
+ *fdp++ = SvIV(*fd);
+ }
+ }
+ sent = sendmsg(PerlIO_fileno(s), &msg, flags);
+ return sent >= 0 ? newSViv(sent) : &PL_sv_undef;
+}
+
+void recv_cmd4(PerlIO *s, SV *buf, STRLEN n)
+{
+ union my_cmsg cmsg = { 0 };
+ struct msghdr msg = { 0 };
+ struct iovec iov;
+ ssize_t i;
+ Inline_Stack_Vars;
+ Inline_Stack_Reset;
+
+ if (!SvOK(buf))
+ sv_setpvn(buf, "", 0);
+ iov.iov_base = SvGROW(buf, n + 1);
+ iov.iov_len = n;
+ msg.msg_iov = &iov;
+ msg.msg_iovlen = 1;
+ msg.msg_control = &cmsg.hdr;
+ msg.msg_controllen = CMSG_SPACE(SEND_FD_SPACE);
+
+ i = recvmsg(PerlIO_fileno(s), &msg, 0);
+ if (i < 0)
+ Inline_Stack_Push(&PL_sv_undef);
+ else
+ SvCUR_set(buf, i);
+ if (i > 0 && cmsg.hdr.cmsg_level == SOL_SOCKET &&
+ cmsg.hdr.cmsg_type == SCM_RIGHTS) {
+ size_t len = cmsg.hdr.cmsg_len;
+ int *fdp = (int *)CMSG_DATA(&cmsg.hdr);
+ for (i = 0; CMSG_LEN((i + 1) * sizeof(int)) <= len; i++)
+ Inline_Stack_Push(sv_2mortal(newSViv(*fdp++)));
+ }
+ Inline_Stack_Done;
+}
+#endif /* defined(CMSG_SPACE) && defined(CMSG_LEN) */
+ALL_LIBC
+
+# 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 <dirent.h>
+#include <errno.h>
+#include <stdio.h>
+#include <string.h>
+
+void nodatacow_fd(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));
+}
+
+void nodatacow_dir(const char *dir)
+{
+ DIR *dh = opendir(dir);
+ int fd;
+
+ if (!dh) croak("opendir(%s): %s", dir, strerror(errno));
+ fd = dirfd(dh);
+ if (fd >= 0)
+ nodatacow_fd(fd);
+ /* ENOTSUP probably won't happen under Linux... */
+ closedir(dh);
+}
+SET_NODATACOW
+
+my $inline_dir = $ENV{PERL_INLINE_DIRECTORY} //= (
+ $ENV{XDG_CACHE_HOME} //
+ ( ($ENV{HOME} // '/nonexistent').'/.cache' )
+ ).'/public-inbox/inline-c';
+
+$set_nodatacow = $all_libc = undef unless -d $inline_dir && -w _;
+if (defined $all_libc) {
# Inline 0.64 or later has locking in multi-process env,
# but we support 0.5 on Debian wheezy
use Fcntl qw(:flock);
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'; #, BUILD_NOISY => 1';
+ eval 'use Inline C => $all_libc.$set_nodatacow';
+ # . ', BUILD_NOISY => 1';
my $err = $@;
+ my $ndc_err;
+ if ($err && $set_nodatacow) { # missing Linux kernel headers
+ $ndc_err = $err;
+ undef $set_nodatacow;
+ eval 'use Inline C => $all_libc';
+ }
flock($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";
- $vfork_spawn = undef;
+ $set_nodatacow = $all_libc = undef;
}
}
-unless (defined $vfork_spawn) {
+unless ($all_libc) {
require PublicInbox::SpawnPP;
*pi_fork_exec = \&PublicInbox::SpawnPP::pi_fork_exec
}
+unless ($set_nodatacow) {
+ require PublicInbox::NDC_PP;
+ no warnings 'once';
+ *nodatacow_fd = \&PublicInbox::NDC_PP::nodatacow_fd;
+ *nodatacow_dir = \&PublicInbox::NDC_PP::nodatacow_dir;
+}
+
+undef $set_nodatacow;
+undef $all_libc;
sub which ($) {
my ($file) = @_;
return $file if index($file, '/') >= 0;
- foreach my $p (split(':', $ENV{PATH})) {
+ for my $p (split(/:/, $ENV{PATH})) {
$p .= "/$file";
return $p if -x $p;
}
sub spawn ($;$$) {
my ($cmd, $env, $opts) = @_;
- my $f = which($cmd->[0]);
- defined $f or die "$cmd->[0]: command not found\n";
+ my $f = which($cmd->[0]) // die "$cmd->[0]: command not found\n";
my @env;
$opts ||= {};
-
- my %env = $env ? (%ENV, %$env) : %ENV;
+ my %env = (%ENV, $env ? %$env : ());
while (my ($k, $v) = each %env) {
- push @env, "$k=$v";
+ push @env, "$k=$v" if defined($v);
}
my $redir = [];
for my $child_fd (0..2) {
my $parent_fd = $opts->{$child_fd};
if (defined($parent_fd) && $parent_fd !~ /\A[0-9]+\z/) {
- defined(my $fd = fileno($parent_fd)) or
+ my $fd = fileno($parent_fd) //
die "$parent_fd not an IO GLOB? $!";
$parent_fd = $fd;
}
}
my $rlim = [];
- foreach my $l (RLIMITS()) {
- defined(my $v = $opts->{$l}) or next;
+ foreach my $l (@RLIMITS) {
+ my $v = $opts->{$l} // next;
my $r = eval "require BSD::Resource; BSD::Resource::$l();";
unless (defined $r) {
warn "$l undefined by BSD::Resource: $@\n";
push @$rlim, $r, @$v;
}
my $cd = $opts->{'-C'} // ''; # undef => NULL mapping doesn't work?
- my $pid = pi_fork_exec($redir, $f, $cmd, \@env, $rlim, $cd);
- $pid < 0 ? undef : $pid;
+ my $pgid = $opts->{pgid} // -1;
+ my $pid = pi_fork_exec($redir, $f, $cmd, \@env, $rlim, $cd, $pgid);
+ die "fork_exec @$cmd failed: $!\n" unless $pid > 0;
+ $pid;
}
sub popen_rd {
- my ($cmd, $env, $opts) = @_;
+ my ($cmd, $env, $opt) = @_;
pipe(my ($r, $w)) or die "pipe: $!\n";
- $opts ||= {};
- $opts->{1} = fileno($w);
- my $pid = spawn($cmd, $env, $opts);
- return unless defined $pid;
+ $opt ||= {};
+ $opt->{1} = fileno($w);
+ my $pid = spawn($cmd, $env, $opt);
return ($r, $pid) if wantarray;
my $ret = gensym;
- tie *$ret, 'PublicInbox::ProcessPipe', $pid, $r;
+ tie *$ret, 'PublicInbox::ProcessPipe', $pid, $r, @$opt{qw(cb arg)};
$ret;
}
+sub run_die ($;$$) {
+ my ($cmd, $env, $rdr) = @_;
+ my $pid = spawn($cmd, $env, $rdr);
+ waitpid($pid, 0) == $pid or die "@$cmd did not finish";
+ $? == 0 or die "@$cmd failed: \$?=$?\n";
+}
+
1;