Date: Tue, 13 Aug 2013 23:00:45 +0200 From: Jilles Tjoelker <jilles@stack.nl> To: Peter Wemm <peter@FreeBSD.org> Cc: svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r254297 - head/lib/libc/stdlib Message-ID: <20130813210045.GB68244@stack.nl> In-Reply-To: <201308132038.r7DKcuH7082570@svn.freebsd.org> References: <201308132038.r7DKcuH7082570@svn.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Aug 13, 2013 at 08:38:56PM +0000, Peter Wemm wrote: > Author: peter > Date: Tue Aug 13 20:38:55 2013 > New Revision: 254297 > URL: http://svnweb.freebsd.org/changeset/base/254297 > Log: > vfork(2) was listed as deprecated in 1994 (r1573) and was the false > reports of its impending demise were removed in 2009 (r199257). > However, in 1996 (r16117) system(3) was switched from vfork(2) to > fork(2) based partly on this. Switch back to vfork(2). This has a > dramatic effect in cases of extreme mmap use - such as excessive > abuse (500+) of shared libraries. > popen(3) has used vfork(2) for a while. vfork(2) isn't going anywhere. > Modified: > head/lib/libc/stdlib/system.c > Modified: head/lib/libc/stdlib/system.c > ============================================================================== > --- head/lib/libc/stdlib/system.c Tue Aug 13 20:33:50 2013 (r254296) > +++ head/lib/libc/stdlib/system.c Tue Aug 13 20:38:55 2013 (r254297) > @@ -39,6 +39,7 @@ __FBSDID("$FreeBSD$"); > #include <signal.h> > #include <stdlib.h> > #include <stddef.h> > +#include <string.h> > #include <unistd.h> > #include <paths.h> > #include <errno.h> > @@ -56,37 +57,36 @@ __system(const char *command) > if (!command) /* just checking... */ > return(1); > > - /* > - * Ignore SIGINT and SIGQUIT, block SIGCHLD. Remember to save > - * existing signal dispositions. > - */ > - ign.sa_handler = SIG_IGN; > - (void)sigemptyset(&ign.sa_mask); > - ign.sa_flags = 0; > - (void)_sigaction(SIGINT, &ign, &intact); > - (void)_sigaction(SIGQUIT, &ign, &quitact); > (void)sigemptyset(&newsigblock); > (void)sigaddset(&newsigblock, SIGCHLD); > (void)_sigprocmask(SIG_BLOCK, &newsigblock, &oldsigblock); > - switch(pid = fork()) { > + switch(pid = vfork()) { > case -1: /* error */ > - break; > + (void)_sigprocmask(SIG_SETMASK, &oldsigblock, NULL); > + return (-1); > case 0: /* child */ > /* > * Restore original signal dispositions and exec the command. > */ > - (void)_sigaction(SIGINT, &intact, NULL); > - (void)_sigaction(SIGQUIT, &quitact, NULL); > (void)_sigprocmask(SIG_SETMASK, &oldsigblock, NULL); > execl(_PATH_BSHELL, "sh", "-c", command, (char *)NULL); > _exit(127); > - default: /* parent */ > - savedpid = pid; > - do { > - pid = _wait4(savedpid, &pstat, 0, (struct rusage *)0); > - } while (pid == -1 && errno == EINTR); > - break; > } > + /* > + * If we are running means that the child has either completed > + * its execve, or has failed. > + * Block SIGINT/QUIT because sh -c handles it and wait for > + * it to clean up. > + */ > + memset(&ign, 0, sizeof(ign)); > + ign.sa_handler = SIG_IGN; > + (void)sigemptyset(&ign.sa_mask); > + (void)_sigaction(SIGINT, &ign, &intact); > + (void)_sigaction(SIGQUIT, &ign, &quitact); > + savedpid = pid; > + do { > + pid = _wait4(savedpid, &pstat, 0, (struct rusage *)0); > + } while (pid == -1 && errno == EINTR); > (void)_sigaction(SIGINT, &intact, NULL); > (void)_sigaction(SIGQUIT, &quitact, NULL); > (void)_sigprocmask(SIG_SETMASK, &oldsigblock, NULL); This is definitely good, but I think ignoring SIGINT/SIGQUIT only after the vfork() puts back an old bug. The ignore was deliberately moved to before fork(). You can avoid it more simply by adding SIGINT and SIGQUIT to the signal mask above. The kernel then keeps any SIGINT/SIGQUIT pending until the signal is ignored after vfork(). -- Jilles Tjoelker
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130813210045.GB68244>