DRM framebuffer removal broken since 13803132

DRM framebuffer removal broken since 13803132

[Thomas Hellstrom]

Hi!

The commit

From 13803132818cf8084d169617be060fd8e3411a98 Mon Sep 17 00:00:00 2001
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Date: Wed, 9 Sep 2015 16:40:56 +0200
Subject: [PATCH] drm/core: Preserve the framebuffer after removing it.

badly breaks vmwgfx multimon since the crtcs are no longer turned off at
framebuffer removal,
which is part of the user-space contract.

Also isn't this a security issue, since at, for example, Xorg crash, the
crtcs can be left scanning out from memory
DRM no longer owns?

/Thomas


_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: DRM framebuffer removal broken since 13803132

[Maarten Lankhorst]

Hey,

Op 16-03-16 om 09:34 schreef Thomas Hellstrom:
> Hi!
>
> The commit
>
> From 13803132818cf8084d169617be060fd8e3411a98 Mon Sep 17 00:00:00 2001
> From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Date: Wed, 9 Sep 2015 16:40:56 +0200
> Subject: [PATCH] drm/core: Preserve the framebuffer after removing it.
>
> badly breaks vmwgfx multimon since the crtcs are no longer turned off at
> framebuffer removal,
> which is part of the user-space contract.
Is this about rmfb or lastclose?

Lastclose is handled by kms fbdev emulation, rmfb behavioral change was intentional, but can be reverted.
> Also isn't this a security issue, since at, for example, Xorg crash, the
> crtcs can be left scanning out from memory
> DRM no longer owns?
>
Not in general, the drivers I've worked with have the framebuffers holding a reference to the underlying
object storage. Else this would be a security issue without the rmfb behavior too.

~Maarten
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: DRM framebuffer removal broken since 13803132

[Thomas Hellstrom]

On 03/16/2016 11:17 AM, Maarten Lankhorst wrote:
> Hey,
>
> Op 16-03-16 om 09:34 schreef Thomas Hellstrom:
>> Hi!
>>
>> The commit
>>
>> From 13803132818cf8084d169617be060fd8e3411a98 Mon Sep 17 00:00:00 2001
>> From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
>> Date: Wed, 9 Sep 2015 16:40:56 +0200
>> Subject: [PATCH] drm/core: Preserve the framebuffer after removing it.
>>
>> badly breaks vmwgfx multimon since the crtcs are no longer turned off at
>> framebuffer removal,
>> which is part of the user-space contract.
> Is this about rmfb or lastclose?
>
> Lastclose is handled by kms fbdev emulation, rmfb behavioral change was intentional, but can be reverted.

OK, this is about rmfb.

/Thomas

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel