Re: linux-kernel@vger.kernel.org

[Sudarshan Soma]

It looks like i can't call exit(-1)in signal handlers as its
Asynchronous—signal-safe function. As my application wont register any
exit handler functions, is this safe to use??

Please suggest if there are better alternatives.

On Mon, May 18, 2009 at 11:44 AM, Sudarshan Soma <sudarshan12s@xxxxxxxxx> wrote:
> --- On Sat, 16/5/09, Robert Hancock <hancockrwd@xxxxxxxxx> wrote:
>
>
>    From: Robert Hancock <hancockrwd@xxxxxxxxx>
>    Subject: Re: patch related to SIGSEGV handling in multiple threads
> and hung threads
>    To: "Sudarshan Soma" <sudarshan12s@xxxxxxxxx>
>    Cc: linux-kernel@xxxxxxxxxxxxxxx
>    Date: Saturday, 16 May, 2009, 6:45 AM
>
>    Sudarshan Soma wrote:
>    > Hi ,
>    > I have a program which will create multiple threads. One of the thread
>    > receives SIGSEGV and it prints stacktrace in its handler and calls
>    > _exit(-1). The call to exit wont terminate all threads in  a process
>    > and hence thread hangs here. This causes resouce leaks and other
>    > problems. It is seen only on linux 2.4.20.
>    >
>    > For printing stacktrace it uses
>    > __builtin_return_address and __builtin_frame_address
>    >
>    >
>    > The same program works fine on linux 2.6.26 kernel. It  will exit the
>    > whole process when one of its thread dies as above.
>    >
>    >
>    > Please let me know the patch details or any workaround, if any. This
>    > is a bit urgent, so please let me know for any pointers , if any
>    >
>    > Thanks and Best Regards,
>    > Pavan.
>
> [Robert]
>    That doesn't seem surprising, the old LinuxThreads thread
> implementation normally used with the 2.4 kernels was pretty poor.
> However it would likely be better to re-raise the signal rather than
> trying to exit out of the signal handler.
>
>
> [Pavan]
> Thanks Robert.
> using exit(-1) instead of exit(0) does terminate all the threads in
> the process in my sample program.  Please suggest if this workaround
> is safe for now.
>
> I would like to kill whole process(and its threads) when one of the
> thread gets SIGSEGV signal. Could you please eloborate , how could i
> achieve the same by raising SIGSEGV signal again..
>
> Best Regards,
> Pavan.
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/