Re: Possible kernel bug with SYS_clone / CLONE_PARENT

[Louis Rilling <Louis.Rilling@kerlabs.com>]

[-- Attachment #1: Type: text/plain, Size: 4455 bytes --]

On 03/08/09 14:32 +1000, Nicholas Vinen wrote:
> Hello,
> 
> I have a program where process A forks process B and process B forks
> process C. I want process A to be notified if/when process C terminates.
> (Process B terminates almost immediately after forking process C and a
> number of other siblings.)
> 
> According to the manual page for clone() I should be able to do this
> with CLONE_PARENT:
> 
> -----------------
>     CLONE_PARENT (since Linux 2.3.12)
>     If CLONE_PARENT is set, then the parent of the new child (as
> returned by getppid(2)) will be the same as that of the calling process.
>     If CLONE_PARENT is not set, then (as with fork(2)) the child's
> parent is the calling process.
>     Note that it is the parent process, as returned by getppid(2), which
> is signaled when the child terminates, so that if CLONE_PARENT is set,
> then the parent of the calling process, rather than the calling process
> itself, will be signaled.
> -----------------
> 
> 
> I am using kernel 2.6.29-gentoo-r5 (not the most recent, I know, but
> relatively new), glibc 2.9_p20081201-r2 and gcc 4.3.2-r3. Here is my
> test program:
> 
> -----------------
> #include <asm/unistd.h>
> #include <sys/syscall.h>
> #include <sched.h>
> #include <signal.h>
> #include <stdio.h>
> #include <string.h>
> #include <stdlib.h>
> 
> int fork_but_keep_ppid() {
>   return syscall(SYS_clone, CLONE_PARENT, (void*)0);
> }

You should use CLONE_PARENT|SIGCHLD. CLONE_PARENT alone deactivates
signal notification to original parent.

Louis

> 
> void sigchld_handler(int signum, siginfo_t* info, void* ucontext) {
>   fprintf(stderr, "SIGCHLD (PID=%d)\n", info->si_pid);
> }
> 
> int main(void) {
>   struct sigaction act;
>   memset(&act, 0, sizeof(act));
>   act.sa_sigaction = sigchld_handler;
>   act.sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_SIGINFO;
>   sigaction(SIGCHLD, &act, 0);
> 
>   fprintf(stderr, "Main PID = %d\n", syscall(SYS_getpid));
> 
>   pid_t a = fork();
>   if( a == 0 ) {
>     fprintf(stderr, "first fork PID = %d, PPID = %d\n",
> syscall(SYS_getpid), syscall(SYS_getppid));
>     pid_t b = fork_but_keep_ppid();
>     if( b == 0 ) {
>       fprintf(stderr, "second fork PID = %d, PPID = %d\n",
> syscall(SYS_getpid), syscall(SYS_getppid));
>       exit(-2);
>     } else {
>       fprintf(stderr, "second fork returned %d\n", b);
>     }
>     exit(-1);
>   } else {
>     fprintf(stderr, "first fork returned %d\n", a);
>   }
>   sleep(1);
>   fprintf(stderr, "Creating & terminating another child to check
> SIGCHILD still works...\n");
>   pid_t c = fork();
>   if( c == 0 ) {
>     fprintf(stderr, "Child's PID is %d\n", syscall(SYS_getpid));
>     exit(-3);
>   }
>   sleep(1);
> 
>   return 0;
> }
> -----------------
> 
> Note that I am making the syscall directly because I want fork-like
> semantics and these are not provided by the clone() call. I also don't
> want to have to provide a separate stack for the child and it seems that
> (at least according to the man page) for clone() you have to, even if
> you don't use CLONE_VM.
> 
> The output from this program is:
> 
> -----------------
> Main PID = 13125
> first fork returned 13126
> first fork PID = 13126, PPID = 13125
> second fork PID = 13127, PPID = 13125
> second fork returned 13127
> SIGCHLD (PID=13126)
> Creating & terminating another child to check SIGCHILD still works...
> Child's PID is 13128
> SIGCHLD (PID=13128)
> -----------------
> 
> 
> There should be a SIGCHLD for the second fork's child PID but there
> isn't. ps shows the process is a zombie with PPID=13125 (in this case)
> while the test program is still running, yet the "parent" does not seem
> to receive the SIGCHLD.
> 
> I could be doing something wrong but I can't see what it might be. Any
> suggestions? Please CC me on any reply as I am not currently subscribed
> to LKML.
> 
> 
> 
> Thanks,
> 
> Nicholas.
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
Dr Louis Rilling			Kerlabs
Skype: louis.rilling			Batiment Germanium
Phone: (+33|0) 6 80 89 08 23		80 avenue des Buttes de Coesmes
http://www.kerlabs.com/			35700 Rennes

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 197 bytes --]