Re: [PATCH v2] drm/xe/pf: Always expose VRAM provisioning data on discrete GPUs

["Laguna, Lukasz" <lukasz.laguna@intel.com>]

On 10/17/2025 12:27, Michal Wajdeczko wrote:
>
> On 10/16/2025 2:22 PM, Lukasz Laguna wrote:
>> Currently, VRAM provisioning data is only exposed if the device supports
>> LMTT. While it should not be possible to modify VRAM provisioning on
>> platforms without LMTT, it is still useful to be able to read the VRAM
>> provisioning data on all discrete GPU platforms.
>>
>> Expose the VRAM debugfs attributes whenever running on dGFX, adjusting
>> file permissions to read only when LMTT is not available.
>>
>> Fixes: 921ddb37d87c ("drm/xe/pf: Don't allow LMEM provisioning if LMTT isn't available on the device")
> all our current DGFX SR-IOV platforms are using LMTT, so IMO there is nothing to fix

OK, we can remove the Fixes tag.

>
>> Signed-off-by: Lukasz Laguna <lukasz.laguna@intel.com>
>> ---
>> v2:
>>   - add extra protection against changing the LMEM provisioning (Piotr)
>> ---
>>   drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c    | 3 ++-
>>   drivers/gpu/drm/xe/xe_tile_sriov_pf_debugfs.c | 9 ++++++---
>>   2 files changed, 8 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c b/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c
>> index b2e5c52978e6..c0c0215c0703 100644
>> --- a/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c
>> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c
>> @@ -1548,7 +1548,8 @@ int xe_gt_sriov_pf_config_set_lmem(struct xe_gt *gt, unsigned int vfid, u64 size
>>   {
>>   	int err;
>>   
>> -	xe_gt_assert(gt, xe_device_has_lmtt(gt_to_xe(gt)));
>> +	if (!xe_device_has_lmtt(gt_to_xe(gt)))
>> +		return -EPERM;
>>   
>>   	mutex_lock(xe_gt_sriov_pf_master_mutex(gt));
>>   	if (vfid)
>> diff --git a/drivers/gpu/drm/xe/xe_tile_sriov_pf_debugfs.c b/drivers/gpu/drm/xe/xe_tile_sriov_pf_debugfs.c
>> index c8df18af4d00..6ba2332c77d4 100644
>> --- a/drivers/gpu/drm/xe/xe_tile_sriov_pf_debugfs.c
>> +++ b/drivers/gpu/drm/xe/xe_tile_sriov_pf_debugfs.c
>> @@ -165,14 +165,17 @@ DEFINE_SRIOV_TILE_CONFIG_DEBUGFS_ATTRIBUTE(vram, lmem, u64, "%llu\n");
>>   
>>   static void pf_add_config_attrs(struct xe_tile *tile, struct dentry *dent, unsigned int vfid)
>>   {
>> +	struct xe_device *xe = tile->xe;
>> +
>>   	xe_tile_assert(tile, tile == extract_tile(dent));
>>   	xe_tile_assert(tile, vfid == extract_vfid(dent));
>>   
>>   	debugfs_create_file_unsafe(vfid ? "ggtt_quota" : "ggtt_spare",
>>   				   0644, dent, dent, &ggtt_fops);
>> -	if (xe_device_has_lmtt(tile->xe))
>> +	if (IS_DGFX(xe))
>>   		debugfs_create_file_unsafe(vfid ? "vram_quota" : "vram_spare",
>> -					   0644, dent, dent, &vram_fops);
>> +					   xe_device_has_lmtt(xe) ? 0644 : 0444,
>> +					   dent, dent, &vram_fops);
> for platforms without LMTT, does it make sense to expose PF's "spare" attribute?

In my opinion yes - if it's exposed we can check the size of the region 
assigned to PF.

>>   }
>>   
>>   static void pf_populate_tile(struct xe_tile *tile, struct dentry *dent, unsigned int vfid)
>> @@ -188,7 +191,7 @@ static void pf_populate_tile(struct xe_tile *tile, struct dentry *dent, unsigned
>>   		drm_debugfs_create_files(pf_ggtt_info,
>>   					 ARRAY_SIZE(pf_ggtt_info),
>>   					 dent, minor);
>> -		if (xe_device_has_lmtt(xe))
>> +		if (IS_DGFX(xe))
>>   			drm_debugfs_create_files(pf_vram_info,
>>   						 ARRAY_SIZE(pf_vram_info),
>>   						 dent, minor);