Re: [PATCH] Increase fork signal safety for single-threaded processes [BZ #19703]

[Florian Weimer <fweimer at redhat dot com>]

On 05/11/2016 09:59 PM, Tulio Magno Quites Machado Filho wrote:
> Florian Weimer <fweimer@redhat.com> writes:
>
> > This provides a band-aid and addresses the scenario where fork is
> > called from a signal handler while the process is in the malloc
> > subsystem (or has acquired the libio list lock).  It does not
> > address the general issue of async-signal-safety of fork;
> > multi-threaded processes are not covered, and some glibc
> > subsystems have fork handlers which are not async-signal-safe.
> >
> > 2016-05-11  Florian Weimer  <fweimer@redhat.com>
> >
> > 	[BZ #19703]
> > 	Partially async-signal-safe fork for single-threaded processes.
> > 	* sysdeps/nptl/fork.c (__libc_fork): Introduce multiple_threads
> > 	variable.  Do not acquire and reset/release malloc and libio locks
> > 	in single-threaded processes.
> > 	* malloc/tst-mallocfork2.c: New file.
>
> Do you have plans to remove malloc/tst-mallocfork?

No, I would just change exit to _exit there.

> > diff --git a/malloc/Makefile b/malloc/Makefile
> > index 3283f4f..fa1730e 100644
> >
> > +  sigusr1_sender_pid = fork ();
> > +  if (sigusr1_sender_pid == 0)
> > +    signal_sender (SIGUSR1, false);
> > +  pid_t sigusr2_sender_pid = fork ();
> > +  if (sigusr2_sender_pid == 0)
> > +    signal_sender (SIGUSR2, true);
>
> I suggest to fork the SIGUSR2 sender first.
> Sometimes the SIGUSR1 sender floods the parent to the point it takes seconds
> to fork the second sender.

Interesting.  I have made the change locally, as you suggested.

Any comments on the change to the fork implementation?

Thanks,
Florian