[Mesa-dev] [PATCH] anv/entrypoints: VkGetDeviceProcAddr returns NULL for core instance commands

Iago Toral itoral at igalia.com
Tue Mar 6 12:09:18 UTC 2018


On Tue, 2018-03-06 at 11:21 +0000, Emil Velikov wrote:
> On 6 March 2018 at 09:57, Bas Nieuwenhuizen <basni at chromium.org>
> wrote:
> > 
> > 
> > On Tue, Mar 6, 2018 at 8:02 AM, Iago Toral <itoral at igalia.com>
> > wrote:
> > > 
> > > On Mon, 2018-03-05 at 12:11 +0000, Emil Velikov wrote:
> > > > Hi Iago,
> > > > 
> > > > Top level questions:
> > > > 
> > > > I think this and the original commit should go to stable right?
> > > 
> > > I am not sure if this qualifies for stable: these patches don't
> > > fix any
> > > user-visible bugs. If an application was calling
> > > vkGetDeviceProcAddr to
> > > get pointers to non-device functions (which is incorrect by the
> > > spec)
> > > the previous behavior would allow it to get away with it without
> > > issues, bit with these patches it will start to crash since it
> > > will
> > > receive NULL pointers.
> > > 
> 
> According to Lenny's comment in the github issue there's nothing to
> be
> concerned. Namely:
>  - "The pointers being returned are invalid. ... trying to use them
> will result in a crash."
>  - "Wolfenstein was acquiring, but not using the pointers."

Because it is not using the pointers :), if some other app is using
them it will start to crash.

But that was not my point, my point was that this doesn't fix anything
for users, so I was questioning whether it was material for stable
based on that.

> 
> > > > The dispatch in 17.3 is very different making, yet 18.0 should
> > > > be
> > > > perfectly fine.
> > > > 
> > > > Mildly related - anv is missing a special case for following
> > > > three
> > > > extensions. Should we port those over from radv?
> > > > 
> > > > vkCreateInstance vkEnumerateInstanceExtensionProperties
> > > > vkEnumerateInstanceLayerProperties
> > > 
> > > What is the exception exactly? Right now we report NULL for these
> > > from
> > > vkGetDeviceProcAddr which is the right thing to do.
> > 
> > 
> > He's talking about the 3 instance commands that we should return in
> > the
> > GetInstanceProcAddr even if instance is NULL, but anv handles that
> > inside
> > anv_device.c anyway, so I don't think there is anything to be done
> > there.
> 
> Indeed. Thanks Bas.
> 
> HTH
> Emil
> 


More information about the mesa-dev mailing list