[Devel] Re: [PATCH user-cr] use syscall(NR_getpid) to avoid cached pid

[Oren Laadan]

Got it, thanks.

Serge E. Hallyn wrote:
> otherwise glibc caches getpid() result and we put the parent
> process in the new cgroup, in the case of 'nsexec -cg /bin/sleep 10'
> 
> Signed-off-by: Serge Hallyn <serue at us.ibm.com>
> ---
>  nsexec.c |    4 ++--
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/nsexec.c b/nsexec.c
> index 3a825f1..79fcc70 100644
> --- a/nsexec.c
> +++ b/nsexec.c
> @@ -154,7 +154,7 @@ int move_to_new_cgroup(int newcgroup)
>  	fout = fopen(tasksfname, "w");
>  	if (!fout)
>  		return 0;
> -	fprintf(fout, "%d\n", getpid());
> +	fprintf(fout, "%d\n", syscall(__NR_getpid));
>  	fclose(fout);
>  	return 1;
>  }
> @@ -193,7 +193,7 @@ int do_child(void *vargv)
>  
>  	/* if pid == 1 then remount /proc */
>  	/* But if the container has no /proc don't fret */
> -	if (getpid() == 1) {
> +	if (syscall(__NR_getpid) == 1) {
>  		umount2("/proc", MNT_DETACH);
>  		mount("proc", "/proc", "proc", 0, NULL);
>  	}
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers