Re: [PATCH v2 1/3] drm/xe: Enable Wa_14025515070

[Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>]

On 2/2/2026 5:37 PM, Daniele Ceraolo Spurio wrote:
>
>
> On 2/2/2026 5:02 PM, Daniele Ceraolo Spurio wrote:
>>
>>
>> On 1/30/2026 12:50 PM, Julia Filipchuk wrote:
>>> Corrects a failure on context switch. On registration of context with
>>> indirect ring state will correct state of targeted and idle command
>>> streamers.
>>>
>>> If Wa is unavailable, notify and disable indirect_ring_state.
>>>
>>> Signed-off-by: Julia Filipchuk <julia.filipchuk@intel.com>
>>> ---
>>>   drivers/gpu/drm/xe/abi/guc_klvs_abi.h |  1 +
>>>   drivers/gpu/drm/xe/xe_guc_ads.c       | 10 ++++++++++
>>>   drivers/gpu/drm/xe/xe_wa_oob.rules    |  5 +++++
>>>   3 files changed, 16 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/xe/abi/guc_klvs_abi.h 
>>> b/drivers/gpu/drm/xe/abi/guc_klvs_abi.h
>>> index e33bd622ab44..69e6d9f9650c 100644
>>> --- a/drivers/gpu/drm/xe/abi/guc_klvs_abi.h
>>> +++ b/drivers/gpu/drm/xe/abi/guc_klvs_abi.h
>>> @@ -496,6 +496,7 @@ enum xe_guc_klv_ids {
>>> GUC_WA_KLV_WAKE_POWER_DOMAINS_FOR_OUTBOUND_MMIO = 0x900a,
>>> GUC_WA_KLV_RESET_BB_STACK_PTR_ON_VF_SWITCH = 0x900b,
>>> GUC_WA_KLV_RESTORE_UNSAVED_MEDIA_CONTROL_REG = 0x900c,
>>> + 
>>> GUC_WA_KLV_CLEAR_CS_INDIRECT_RING_STATE_IF_IDLE_AT_CONTEXT_REGISTRATION 
>>> = 0x900e,
>>
>> Maybe we could try to make this define a bit shorter? something like:
>> GUC_WA_KLV_CLR_CS_INDIRECT_RING_STATE_IF_IDLE_AT_CTX_REG
>>
>>>   };
>>>     #endif
>>> diff --git a/drivers/gpu/drm/xe/xe_guc_ads.c 
>>> b/drivers/gpu/drm/xe/xe_guc_ads.c
>>> index f4cbc030f4c8..1c563a01c12c 100644
>>> --- a/drivers/gpu/drm/xe/xe_guc_ads.c
>>> +++ b/drivers/gpu/drm/xe/xe_guc_ads.c
>>> @@ -360,6 +360,16 @@ static void guc_waklv_init(struct xe_guc_ads *ads)
>>>       if (XE_GT_WA(gt, 14020001231))
>>>           guc_waklv_enable(ads, NULL, 0, &offset, &remain,
>>> GUC_WORKAROUND_KLV_DISABLE_PSMI_INTERRUPTS_AT_C6_ENTRY_RESTORE_AT_EXIT); 
>>>
>>> +    if (XE_GT_WA(gt, 14025515070)) {
>>> +        if (GUC_FIRMWARE_VER_AT_LEAST(&gt->uc.guc, 70, 53))
>>> +            guc_waklv_enable(ads, NULL, 0, &offset, &remain,
>>> + 
>>> GUC_WA_KLV_CLEAR_CS_INDIRECT_RING_STATE_IF_IDLE_AT_CONTEXT_REGISTRATION);
>>> +        else {
>>> +            /* Disable indirect_ring_state when workaround not 
>>> available. */
>>> +            gt->info.has_indirect_ring_state = 0;
>>> +            xe_gt_notice(gt, "indirect ring state requires 
>>> workaround present in GuC firmware v70.53.+\n");
>>
>> With v70.53.+ I can't understand if you mean "70.53 or newer" or 
>> "70.53.*" (although both mean the same thing). Maybe just v70.53+ 
>> would be clearer?
>> Also, shortening some of the words should make this fit within the 
>> 100 char limit, e.g. "indirect ring state requires a WA present in 
>> GuC v70.53+"
>>
>>> +        }
>>> +    }
>>>         size = guc_ads_waklv_size(ads) - remain;
>>>       if (!size)
>>> diff --git a/drivers/gpu/drm/xe/xe_wa_oob.rules 
>>> b/drivers/gpu/drm/xe/xe_wa_oob.rules
>>> index 5cd7fa6d2a5c..1da3a8f96f98 100644
>>> --- a/drivers/gpu/drm/xe/xe_wa_oob.rules
>>> +++ b/drivers/gpu/drm/xe/xe_wa_oob.rules
>>> @@ -69,6 +69,11 @@
>>>           MEDIA_VERSION(3002), FUNC(xe_rtp_match_psmi_enabled)
>>>   16023683509    MEDIA_VERSION(2000), FUNC(xe_rtp_match_psmi_enabled)
>>>           MEDIA_VERSION(3000), MEDIA_STEP(A0, B0), 
>>> FUNC(xe_rtp_match_psmi_enabled)
>>> +14025515070    GRAPHICS_VERSION(2004)
>>> +        MEDIA_VERSION_RANGE(1301, 3000)
>>
>> We usually avoid using big ranges that cover multiple gens, because a 
>> new derivative might be added for an older gen that doesn't require 
>> the WA.
>
> To clarify, this isn't a blocker, since we do have exceptions. As long 
> as you've confirmed we don't expect this to change for derivatives 
> we're good.
>
> Daniele
>
>>
>>> +        MEDIA_VERSION(3002)
>>> +        GRAPHICS_VERSION_RANGE(3000, 3001)
>>> +        GRAPHICS_VERSION(3003)
>>
>> According to the WA database this also applies to Graphics 3004, 3005 
>> and 3511 and Media 3500 and 3503

Looked at this again with Julia earlier today and confirmed this doesn't 
actually apply to G3511 and M3503, I had just misread the database. It 
does still apply to the other missing versions.

Daniele

>>
>> Daniele
>>
>>>     15015404425_disable PLATFORM(PANTHERLAKE), MEDIA_STEP(B0, FOREVER)
>>>   16026007364    MEDIA_VERSION(3000)
>>
>