linux-kernel - Re: [Intel-gfx] [PATCH v6 08/17] drm/ttm: use gem vma

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20190814055827.6hrxj6daovxxnnvw@sirius.home.kraxel.org>
Date:   Wed, 14 Aug 2019 07:58:27 +0200
From:   Gerd Hoffmann <kraxel@...hat.com>
To:     Thierry Reding <thierry.reding@...il.com>
Cc:     dri-devel@...ts.freedesktop.org, David Airlie <airlied@...ux.ie>,
        "open list:DRM DRIVER FOR NVIDIA GEFORCE/QUADRO GPUS" 
        <nouveau@...ts.freedesktop.org>,
        "open list:DRM DRIVER FOR QXL VIRTUAL GPU" 
        <virtualization@...ts.linux-foundation.org>,
        Huang Rui <ray.huang@....com>,
        "David (ChunMing) Zhou" <David1.Zhou@....com>,
        Thomas Hellstrom <thellstrom@...are.com>,
        "open list:RADEON and AMDGPU DRM DRIVERS" 
        <amd-gfx@...ts.freedesktop.org>,
        Maxime Ripard <maxime.ripard@...tlin.com>,
        VMware Graphics <linux-graphics-maintainer@...are.com>,
        bskeggs@...hat.com, Dave Airlie <airlied@...hat.com>,
        thomas@...pmail.org, tzimmermann@...e.de,
        ckoenig.leichtzumerken@...il.com, intel-gfx@...ts.freedesktop.org,
        "open list:DRM DRIVER FOR QXL VIRTUAL GPU" 
        <spice-devel@...ts.freedesktop.org>,
        open list <linux-kernel@...r.kernel.org>,
        Alex Deucher <alexander.deucher@....com>,
        Christian König <christian.koenig@....com>
Subject: Re: [Intel-gfx] [PATCH v6 08/17] drm/ttm: use gem vma_node

> Hi Gerd,
> 
> I've been seeing a regression on Nouveau with recent linux-next releases
> and git bisect points at this commit as the first bad one. If I revert
> it (there's a tiny conflict with a patch that was merged subsequently),
> things are back to normal.
> 
> I think the reason for this issue is that Nouveau doesn't use GEM
> objects for all buffer objects,

That shouldn't be a problem ...

> and even when it uses GEM objects, the
> code will not initialize the GEM object until after the buffer objects
> and the backing TTM objects have been created.

... but the initialization order is.

ttm_bo_uses_embedded_gem_object() assumes gem gets initialized first.

drm_gem_object_init() init calling drm_vma_node_reset() again is
probably the root cause for the breakage.

> I tried to fix that by making sure drm_gem_object_init() gets called by
> Nouveau before ttm_bo_init(), but the changes are fairly involved and I
> was unable to get the GEM reference counting right. I can look into the
> proper fix some more, but it might be worth reverting this patch for
> now to get Nouveau working again.

Changing the order doesn't look hard.  Patch attached (untested, have no
test hardware).  But maybe I missed some detail ...

The other patch attached works around the issue with a flag, to avoid
drm_vma_node_reset() being called twice.

cheers,
  Gerd


View attachment "0001-try-unbreak-nouveau-1.patch" of type "text/plain" (2332 bytes)

View attachment "0002-try-unbreak-nouveau-2.patch" of type "text/plain" (8981 bytes)