Re: [PATCH] [RESEND 2] Take over futex of dead task only if FUTEX_WAITERSis not set

[Darren Hart]

On 10/24/2012 11:08 AM, Thomas Gleixner wrote:
> On Wed, 24 Oct 2012, Siddhesh Poyarekar wrote:
> 
>>> Now there is a different solution to that problem. Do not look at the
>>> user space value at all and enforce a lookup of possibly available
>>> pi_state. If pi_state can be found, then the new incoming locker T3
>>> blocks on that pi_state and legitimately races with T2 to acquire the
>>> rt_mutex and the pi_state and therefor the proper ownership of the
>>> user space futex.
>>
>> That works. Thanks for the detailed explanation too.
> 
> Thanks for the reproducer and finding the trouble spot in the first
> place!


Absolutely, that was great. Siddhesh, any objection to this test being
incorporated into futextest?

http://git.kernel.org/?p=linux/kernel/git/dvhart/futextest.git;a=summary

> I'll queue that if Darren has no objections and mark it for stable as
> well.

I would mostly like to understand the stale waiters case you mentioned.
Otherwise, it seems sound - but changing what appears to be a workaround
for an undocumented cornercase in code this complex does make me a bit
nervous. I'd feel better if we could get Siddhesh's and this stale
waiters covered in futextest.

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Technical Lead - Linux Kernel
--
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/