3.4.106: futex: Wrong back port of 9e9aab5dbf?

3.4.106: futex: Wrong back port of 9e9aab5dbf?

[Dirk Behme]

Hello,

from a private -stable review we think that the commit 9e9aab5dbf 
("futex: Ensure get_futex_key_refs() always implies a barrier" [1]) is 
not correctly back ported to 3.4.x:

The issue is that the memory barrier is introduced into 
drop_futex_key_refs() function instead of get_futex_key_refs() function.

Additionally, we are unsure if the futex commits this commit is fixing 
are really in < 3.4.106. I.e. we are unsure if this fix is needed at 
all. Or at least if some additional futex commits are needed.

Do you like to check?

Best regards

Dirk

[1] 
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/?id=9e9aab5dbf6f13478cd996692ed9679af2404fc7

Re: 3.4.106: futex: Wrong back port of 9e9aab5dbf?

[Zefan Li]

Hi Dirk,

On 2015/8/8 14:12, Dirk Behme wrote:
> Hello,
>
> from a private -stable review we think that the commit 9e9aab5dbf ("futex: Ensure get_futex_key_refs() always implies a barrier" [1]) is not correctly back ported to 3.4.x:
>
> The issue is that the memory barrier is introduced into drop_futex_key_refs() function instead of get_futex_key_refs() function.
>
> Additionally, we are unsure if the futex commits this commit is fixing are really in < 3.4.106. I.e. we are unsure if this fix is needed at all. Or at least if some additional futex commits are needed.
>
> Do you like to check?
>

I've confired this commit isn't needed for 3.4.y. It was backported quite long ago,
and at that time I hadn't improved my scripts to check the Fixes tag. Fortunately
seems this mistake won't cause serious issue.

I'll revert it in the coming 3.4.109. Thanks for the review!