[PATCH] [RFC] xf86: add glamor preferred probing

[Dave Airlie]

On 2 September 2015 at 18:37, Hans de Goede <hdegoede at redhat.com> wrote:
> Hi Dave,
>
> On 02-09-15 06:57, Dave Airlie wrote:
>>
>> From: Dave Airlie <airlied at redhat.com>
>>
>> This patch adds an interface to query and ask DDX if
>> they would prefer to use modesetting/glamor for a particular
>> hw configuration, and if so to fail probing on that hw,
>> so that modesetting gets picked up.
>>
>> It adds a new driverFunc query to see if the driver supports
>> glamor/modesetting preference, then passes a new flag to the
>> platform probe asking for it, if the X server is configured
>> to ask. It adds a new serverflag that can be disabled in xorg.conf.
>>
>> The only ugly bit is if we do platform probing and fail and the
>> driver supports glamor preferences we don't do pci or old school
>> probing as they have no nice flags parameter, and really we should
>> be using platform probing anywhere this matters.
>>
>> I've written -intel support for this, but -ati and -nouveau support
>> should be fairly easy. I'll send the intel patch just for reference.
>>
>> Signed-off-by: Dave Airlie <airlied at redhat.com>
>
>
> Overall the code looks good to me, but I wonder do we really need
> to ask the driver if it supports the new PLATFORM_PROBE_PREFER_GLAMOR
> flag? I would expect drivers which do not support it to simple ignore
> the (to them unknown) flag, and then we can avoid this whole
> ask the driver go-around routine.

Intel currently fails to load if it gets a flag that it doesn't know
about, and after years of vague ioctl interfaces it's probably for the
best to ask first :-)

Dave.