CLONE from pthread_create fails with ERESTARTNOINTR and application crashes

CLONE from pthread_create fails with ERESTARTNOINTR and application crashes

[Raghu Pv]

Overview:



CLONE from pthread_create fails with ERESTARTNOINTR and application crashes



Steps to reproduce:



Executing a multithreaded application that creates and terminates
threads and also there a lot of signals



If a signal arrives during execution of clone(), clone is returning
ERESTARTNOINTR to userspace.



We see the following implementation that sets ERESTARTNOINTR:



        /*

         * Process group and session signals need to be delivered to just the

         * parent before the fork or both the parent and the child after the

         * fork. Restart if a signal comes in before we add the new process to

         * it's process group.

         * A fatal signal pending means that current will exit, so the new

         * thread can't slip out of an OOM kill (or normal SIGKILL).

         */

        recalc_sigpending();

        if (signal_pending(current)) {

                spin_unlock(&current->sighand->siglock);

                write_unlock_irq(&tasklist_lock);

                retval = -ERESTARTNOINTR;

                goto bad_fork_cleanup_namespace;

        }





Leaking ERESTARTNOINTR to userspace is a big concern because this a
kernel internal error code.



Immediately after that the application crashes.



Actual result: Application crashes



Expected Result: Application should not crash and clone system call
should restart if clone is previously interrupted by a signal.

CLONE from pthread_create fails with ERESTARTNOINTR and application crashes

[Will Deacon]

On Wed, May 21, 2014 at 02:40:16PM +0100, Raghu Pv wrote:
> CLONE from pthread_create fails with ERESTARTNOINTR and application crashes

[...]

> Executing a multithreaded application that creates and terminates
> threads and also there a lot of signals

Do you have sample source for that?

> If a signal arrives during execution of clone(), clone is returning
> ERESTARTNOINTR to userspace.

With which kernel version? Your quote from fork.c doesn't match mainline.
Can you reproduce with -rc6?

Will

CLONE from pthread_create fails with ERESTARTNOINTR and application crashes

[Raghu Pv]

Hi WiIl,

Many Thanks for responding to my issue.

I am currently using 2.6.18.8 kernel for  armv4tl platform.

We are seeing this issue on our larger system codebase which happens
when pthread_create/clone call is interrupted by a signal.

I am currently working out to come up a sample code segment that can
reproduce the issue.

If you are aware of this issue in 2.6.18.8 kernel and if it is fixed
in subsequent commits please help me to find the relevant patches that
can resolve the problem.

I look forward for your response.

Thanks in advance.

Thanks,
Raghu

On 22 May 2014 10:04, Will Deacon <will.deacon@arm.com> wrote:
> On Wed, May 21, 2014 at 02:40:16PM +0100, Raghu Pv wrote:
>> CLONE from pthread_create fails with ERESTARTNOINTR and application crashes
>
> [...]
>
>> Executing a multithreaded application that creates and terminates
>> threads and also there a lot of signals
>
> Do you have sample source for that?
>
>> If a signal arrives during execution of clone(), clone is returning
>> ERESTARTNOINTR to userspace.
>
> With which kernel version? Your quote from fork.c doesn't match mainline.
> Can you reproduce with -rc6?
>
> Will