Re: [3.16.y-ckt stable] Linux 3.16.7-ckt4 stable review

[Luis Henriques]

On Tue, Jan 13, 2015 at 12:35:58AM +0100, Thomas Voegtle wrote:
> On Mon, 12 Jan 2015, Luis Henriques wrote:
> 
> >On Mon, Jan 12, 2015 at 09:30:45PM +0100, Thomas Voegtle wrote:
> >>On Mon, 12 Jan 2015, Luis Henriques wrote:
> >>
> >>>This is the start of the review cycle for the Linux 3.16.7-ckt4 stable kernel.
> >>>
> >>>This version contains 216 new patches, summarized below.  The new patches are
> >>>posted as replies to this message and also available in this git branch:
> >>>
> >>>http://kernel.ubuntu.com/git?p=ubuntu/linux.git;h=linux-3.16.y-review;a=shortlog
> >>>
> >>>git://kernel.ubuntu.com/ubuntu/linux.git  linux-3.16.y-review
> >>>
> >>>The review period for version 3.16.7-ckt4 will be open for the next three days.
> >>>To report a problem, please reply to the relevant follow-up patch message.
> >>>
> >>>For more information about the Linux 3.16.y-ckt extended stable kernel version,
> >>>see https://wiki.ubuntu.com/Kernel/Dev/ExtendedStable .
> >>
> >>
> >>Something is wrong with drm/i915 (I guess). I had a terrible memleak on Xorg
> >>when using mplayer using more and more RAM, and then the system is swapping
> >>itself to death.
> >>
> >>I'm using a openSUSE 13.1 (x86_64) on a Baytrail J1900 (this is Intel HD
> >>Gen7).
> >>
> >>
> >
> >Thank you for reporting, Thomas.  Can you please confirm that this is
> >actually a regression in 3.16.7-ckt4 (i.e., that you can't reproduce
> >it in 3.16.7-ckt3)?  If so, is it possible to bisecting it?
> 
> 
> 3.16.7-ckt3 was fine for me.
> 
> I tried to revert the drm/i915 patches in the review branch, and I got lucky
> with (only) reverting cb58c663d940a "drm/i915: Disallow pin ioctl completely
> for kms drivers".
> 
> No clue, why reverting these few lines helps, but then I have no memleak.
> 
> 
>      Thomas
> 

Thank a lot for narrowing this down, Thomas!  The text in that commit
(upstream commit d472fcc8379c) may actually provide an hint for this
issue.  It asks backporters to make sure they include both

commit b45305fce5bb1abec263fcff9d81ebecd6306ede
Author: Daniel Vetter <daniel.vetter@xxxxxxxx>
Date:   Mon Dec 17 16:21:27 2012 +0100

    drm/i915: Implement workaround for broken CS tlb on i830/845

and

commit c4d69da167fa967749aeb70bc0e94a457e5d00c1
Author: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Date:   Mon Sep 8 14:25:41 2014 +0100

    drm/i915: Evict CS TLBs between batches

For some reason, this second commit (which was tagged for stable!)
isn't in 3.16, and I completely missed that.

Would you be able to verify that adding this commit (backport attached
bellow) fixes the issue?

Cheers,
--
Luís