Re: [Qemu-devel] [PATCH v3] network scripts: don't block SIGCHLD before forking

[Paolo Bonzini <pbonzini@redhat.com>]

On 12/08/2011 04:48 AM, Michael Roth wrote:
> This patch fixes a bug where child processes of launch_script() can
> misbehave due to SIGCHLD being blocked. In the case of `sudo`, this
> causes a permanent hang.
>
> Previously a SIGCHLD handler was added to reap fork_exec()'d zombie
> processes by calling waitpid(-1, ...). This required other
> fork()/waitpid() callers to temporarilly block SIGCHILD to avoid
> having the final wait status being intercepted by the SIGCHLD
> handler:
>
> 7c3370d4fe3fa6cda8655f109e4659afc8ca4269
>
> Since then, the qemu_add_child_watch() interface was added to allow
> registration of such processes and reap only from that specific set
> of PIDs:
>
> 4d54ec7898bd951007cb6122d5315584bd41d0c4
>
> As a result, we can now avoid blocking SIGCHLD in launch_script(), so
> drop that behavior.
>
> Signed-off-by: Michael Roth<mdroth@linux.vnet.ibm.com>
> ---
>   net/tap.c |    6 ------
>   1 files changed, 0 insertions(+), 6 deletions(-)
>
> diff --git a/net/tap.c b/net/tap.c
> index 1f26dc9..6c27a94 100644
> --- a/net/tap.c
> +++ b/net/tap.c
> @@ -346,15 +346,10 @@ static TAPState *net_tap_fd_init(VLANState *vlan,
>
>   static int launch_script(const char *setup_script, const char *ifname, int fd)
>   {
> -    sigset_t oldmask, mask;
>       int pid, status;
>       char *args[3];
>       char **parg;
>
> -    sigemptyset(&mask);
> -    sigaddset(&mask, SIGCHLD);
> -    sigprocmask(SIG_BLOCK,&mask,&oldmask);
> -
>       /* try to launch network script */
>       pid = fork();
>       if (pid == 0) {
> @@ -378,7 +373,6 @@ static int launch_script(const char *setup_script, const char *ifname, int fd)
>           while (waitpid(pid,&status, 0) != pid) {
>               /* loop */
>           }
> -        sigprocmask(SIG_SETMASK,&oldmask, NULL);
>
>           if (WIFEXITED(status)&&  WEXITSTATUS(status) == 0) {
>               return 0;

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>

Paolo