[Intel-xe] [PATCH v6 1/2] drm/xe/display: Don't try to use vram if not available
Juha-Pekka Heikkila
juhapekka.heikkila at gmail.com
Tue Oct 17 11:13:14 UTC 2023
On 16.10.2023 18.03, Ruhl, Michael J wrote:
>> -----Original Message-----
>> From: Juha-Pekka Heikkila <juhapekka.heikkila at gmail.com>
>> Sent: Monday, October 16, 2023 10:54 AM
>> To: Ruhl, Michael J <michael.j.ruhl at intel.com>
>> Subject: Fwd: [PATCH v6 1/2] drm/xe/display: Don't try to use vram if not
>> available
>>
>> Trying to get bo from vram when vram not available will cause
>> WARN_ON() hence avoid touching vram if not available.
>>
>> Signed-off-by: Juha-Pekka Heikkila <juhapekka.heikkila at gmail.com>
>> ---
>> drivers/gpu/drm/xe/display/xe_fb_pin.c | 11 ++++++-----
>> 1 file changed, 6 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/display/xe_fb_pin.c
>> b/drivers/gpu/drm/xe/display/xe_fb_pin.c
>> index 2c36e5032433..89cf65c58191 100644
>> --- a/drivers/gpu/drm/xe/display/xe_fb_pin.c
>> +++ b/drivers/gpu/drm/xe/display/xe_fb_pin.c
>> @@ -62,11 +62,12 @@ static int __xe_pin_fb_vma_dpt(struct
>> intel_framebuffer *fb,
>> dpt_size = ALIGN(intel_rotation_info_size(&view->rotated) * 8,
>> XE_PAGE_SIZE);
>> - dpt = xe_bo_create_pin_map(xe, tile0, NULL, dpt_size,
>> - ttm_bo_type_kernel,
>> - XE_BO_CREATE_VRAM0_BIT |
>> - XE_BO_CREATE_GGTT_BIT);
>> - if (IS_ERR(dpt))
>> + if (IS_DGFX(xe))
>> + dpt = xe_bo_create_pin_map(xe, tile0, NULL, dpt_size,
>> + ttm_bo_type_kernel,
>> + XE_BO_CREATE_VRAM0_BIT |
>> + XE_BO_CREATE_GGTT_BIT);
>> + else
>> dpt = xe_bo_create_pin_map(xe, tile0, NULL, dpt_size,
>> ttm_bo_type_kernel,
>> XE_BO_CREATE_STOLEN_BIT |
>
> Hi Juha,
>
> You have changed this code from:
>
> "Best effort allocation" (try vram, and then fallback to stolen), to "Discrete gets vram, !discrete gets
> stolen".
Hi Michael,
this is correct. vram is for discrete and stolen is found on integrated.
Earlier pattern was copied from i915 way of setting up display page
table but it had to be changed because asking for vram on Xe while on
discrete will cause WARN_ON() hence exact same pattern from i915
couldn't be used.
/Juha-Pekka
>
> If this is correct:
>
> Reviewed-by: Michael J. Ruhl <michael.j.ruhl at intel.com>
>
> Is this new sematic the "correct" allocation method?
>
> Thanks,
>
> Mike
>
>
>> --
>> 2.25.1
>
More information about the Intel-xe
mailing list