Re: [PATCH v4 0/3] [v4.9.y] coredump: fix race condition between mmget_not_zero()/get_task_mm() and core dumping

[Ajay Kaher]

ïOn 25/06/19, 1:51 AM, "Sasha Levin" <sashal@xxxxxxxxxx> wrote:
    
> On Tue, Jun 25, 2019 at 02:33:06AM +0530, Ajay Kaher wrote:
> > coredump: fix race condition between mmget_not_zero()/get_task_mm()
> > and core dumping
> >
> > [PATCH v4 1/3]:
> > Backporting of commit 04f5866e41fb70690e28397487d8bd8eea7d712a upstream.
> >
> > [PATCH v4 2/3]:
> > Extension of commit 04f5866e41fb to fix the race condition between
> > get_task_mm() and core dumping for IB->mlx4 and IB->mlx5 drivers.
> >
> > [PATCH v4 3/3]
> > Backporting of commit 59ea6d06cfa9247b586a695c21f94afa7183af74 upstream.
> >
> > [diff from v3]:
> > - added [PATCH v4 3/3]
        
> Why do all the patches have the same subject line?
Thanks for catching this. I will correct in next version of these patches,
along with review comments if any.
    
        
> I guess it's correct for the first one, but can you explain what's up
> with #2 and #3?
>
> If the second one isn't upstream, please explain in detail why not and
> how 4.9 differs from upstream so that it requires a custom backport.

#2 applied to 4.14.y:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-4.14/infiniband-fix-race-condition-between-infiniband-mlx4-mlx5-driver-and-core-dumping.patch?id=e4041a3f6b569140549fe7b41ed527c5c1e38ec9

And then to 4.9.y (some part as requires). 
4.18 and onwards doesn't have  mmap_sem locking in mlx4 and mlx5, 
so no need of #2 in 4.18 and onwards.

> The third one just looks like a different patch altogether with a wrong
> subject line?
#3 was in discussion here (during v1), so added here. 
  
> --
> Thanks,
> Sasha