During my recent university work, I have noticed the man (3) page for system (https://man7.org/linux/man-pages/man3/system.3.html) states: The system() library function uses fork(2) to create a child process that executes the shell command specified in command using execl(3) as follows: execl("/bin/sh", "sh", "-c", command, (char *) NULL); On reading the source code for system, this caused me some confusion. Eventually I found a commit from 2018 (https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=5fb7fc96350575c9adb1316833e48ca11553be49), in which fork and exec were replaced by posix_spawn. The explains why I could not aline the man page with the source code.
[CC += Glibc, Adhemerval] Hi Adhemerval, On 10/29/21 22:37, bugzilla-daemon@bugzilla.kernel.org wrote: > https://bugzilla.kernel.org/show_bug.cgi?id=214871 > > Bug ID: 214871 > Summary: Man(3) system outdated > Product: Documentation > Version: unspecified > Hardware: All > OS: Linux > Status: NEW > Severity: normal > Priority: P1 > Component: man-pages > Assignee: documentation_man-pages@kernel-bugs.osdl.org > Reporter: mikekearney85@hotmail.com > Regression: No > > During my recent university work, I have noticed the man (3) page for system > (https://man7.org/linux/man-pages/man3/system.3.html) states: > > The system() library function uses fork(2) to create a child process that > executes the shell command specified in command using execl(3) as follows: > execl("/bin/sh", "sh", "-c", command, (char *) NULL); > > On reading the source code for system, this caused me some confusion. > Eventually I found a commit from 2018 > > (https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=5fb7fc96350575c9adb1316833e48ca11553be49), > in which fork and exec were replaced by posix_spawn. The explains why I could > not aline the man page with the source code. > I was about to change that line to say that it works "as if" using fork(2) and execl(3), as POSIX says, but then, reading the page further, it talks about many implementation details that seem to relate to when glibc did actually use fork(2). They may be outdated now, so I think it's better if someone from glibc that knows its current details reviews the full page and proposes changes. Thanks, Alex
Created attachment 299385 [details] attachment-27459-0.html On Sat, Oct 30, 2021 at 8:55 AM Alejandro Colomar (man-pages) < alx.manpages@gmail.com> wrote: > [CC += Glibc, Adhemerval] > > Hi Adhemerval, > > On 10/29/21 22:37, bugzilla-daemon@bugzilla.kernel.org wrote: > > https://bugzilla.kernel.org/show_bug.cgi?id=214871 > > > > Bug ID: 214871 > > Summary: Man(3) system outdated > > Product: Documentation > > Version: unspecified > > Hardware: All > > OS: Linux > > Status: NEW > > Severity: normal > > Priority: P1 > > Component: man-pages > > Assignee: documentation_man-pages@kernel-bugs.osdl.org > > Reporter: mikekearney85@hotmail.com > > Regression: No > > > > During my recent university work, I have noticed the man (3) page for > system > > (https://man7.org/linux/man-pages/man3/system.3.html) states: > > > > The system() library function uses fork(2) to create a child process that > > executes the shell command specified in command using execl(3) as > follows: > > execl("/bin/sh", "sh", "-c", command, (char *) NULL); > > > > On reading the source code for system, this caused me some confusion. > > Eventually I found a commit from 2018 > > ( > > https://sourceware.org/git/?p=glibc.git;a=commitdiff;h=5fb7fc96350575c9adb1316833e48ca11553be49 > ), > > in which fork and exec were replaced by posix_spawn. The explains why I > could > > not aline the man page with the source code. > > > > I was about to change that line to say that it works "as if" using > fork(2) and execl(3), as POSIX says, but then, reading the page further, > it talks about many implementation details that seem to relate to when > glibc did actually use fork(2). They may be outdated now, so I think > it's better if someone from glibc that knows its current details reviews > the full page and proposes changes. > Hi Alejandro, Indeed we changed glibc to use posix_spawn on system and popen on 2.29 since to use posix_spawn (which uses clone (CLONE_VFORK | CLONE_VM)). It plays way better with Linux memory subsystem than fork() + exec() specially when such calls are used with process with large RSS.
Hi Adhemerval, Michael, On 11/1/21 13:06, zatrazz wrote: > Hi Alejandro, > > Indeed we changed glibc to use posix_spawn on system and popen on 2.29 > since > to use posix_spawn (which uses clone (CLONE_VFORK | CLONE_VM)). It plays > way better with Linux memory subsystem than fork() + exec() specially > when such > calls are used with process with large RSS. I applied the following patch. I'll close the bug. Thanks, Alex --- system.3: It doesn't call fork anymore Reported-by: Michael Kearney <mikekearney85@hotmail.com> Cc: Adhemerval Zanella <zatrazz@gmail.com> Signed-off-by: Alejandro Colomar <alx.manpages@gmail.com> diff --git a/man3/system.3 b/man3/system.3 index 880fe93c6..bde45ed82 100644 --- a/man3/system.3 +++ b/man3/system.3 @@ -40,9 +40,9 @@ system \- execute a shell command .SH DESCRIPTION The .BR system () -library function uses +library function behaves as if it used .BR fork (2) -to create a child process that executes the shell command specified in +to create a child process that executed the shell command specified in .I command using .BR execl (3)
On 02/11/2021 06:58, Alejandro Colomar (man-pages) via Libc-alpha wrote: > Hi Adhemerval, Michael, > > On 11/1/21 13:06, zatrazz wrote: >> Hi Alejandro, >> >> Indeed we changed glibc to use posix_spawn on system and popen on 2.29 >> since >> to use posix_spawn (which uses clone (CLONE_VFORK | CLONE_VM)). It plays >> way better with Linux memory subsystem than fork() + exec() specially when >> such >> calls are used with process with large RSS. > > I applied the following patch. I'll close the bug. > > Thanks, > > Alex I forgot to add that neither system(), posix_spawn(), nor popen() call atfork handlers [1]. POSIX recently changed their description to make it implementation-defined instead of unspecified [2], so I think it is worth to change that is not really behave as called by fork() (since it will run atfork handlers). [1] https://sourceware.org/bugzilla/show_bug.cgi?id=17490 [2] https://www.austingroupbugs.net/view.php?id=1317 > > --- > system.3: It doesn't call fork anymore > > Reported-by: Michael Kearney <mikekearney85@hotmail.com> > Cc: Adhemerval Zanella <zatrazz@gmail.com> > Signed-off-by: Alejandro Colomar <alx.manpages@gmail.com> > > diff --git a/man3/system.3 b/man3/system.3 > index 880fe93c6..bde45ed82 100644 > --- a/man3/system.3 > +++ b/man3/system.3 > @@ -40,9 +40,9 @@ system \- execute a shell command > .SH DESCRIPTION > The > .BR system () > -library function uses > +library function behaves as if it used > .BR fork (2) > -to create a child process that executes the shell command specified in > +to create a child process that executed the shell command specified in > .I command > using > .BR execl (3) > > >