Re: [PATCH v2 3/4] drm/xe: Wire up dmem cgroup reclaim for VRAM manager

[Maarten Lankhorst]

Hey,

Den 2026-04-28 kl. 12:02, skrev Thomas Hellström:
> On Tue, 2026-04-28 at 11:50 +0200, Maarten Lankhorst wrote:
>>
>>
>> Den 2026-04-28 kl. 09:31, skrev Thomas Hellström:
>>> Register the VRAM manager with the dmem cgroup reclaim
>>> infrastructure
>>> so that lowering dmem.max below current VRAM usage triggers TTM
>>> eviction rather than failing with -EBUSY.
>>>
>>> Assisted-by: GitHub Copilot:claude-sonnet-4.6
>>> Signed-off-by: Thomas Hellström <thomas.hellstrom@xxxxxxxxxxxxxxx>
>>> ---
>>>  drivers/gpu/drm/xe/xe_ttm_vram_mgr.c | 19 ++++++++++++-------
>>>  1 file changed, 12 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
>>> b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
>>> index 5fd0d5506a7e..1bdcb3fee901 100644
>>> --- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
>>> +++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
>>> @@ -303,13 +303,6 @@ int __xe_ttm_vram_mgr_init(struct xe_device
>>> *xe, struct xe_ttm_vram_mgr *mgr,
>>>  	struct ttm_resource_manager *man = &mgr->manager;
>>>  	int err;
>>>  
>>> -	if (mem_type != XE_PL_STOLEN) {
>>> -		const char *name = mem_type == XE_PL_VRAM0 ?
>>> "vram0" : "vram1";
>>> -		man->cg = drmm_cgroup_register_region(&xe->drm,
>>> name, size);
>>> -		if (IS_ERR(man->cg))
>>> -			return PTR_ERR(man->cg);
>>> -	}
>>> -
>>>  	man->func = &xe_ttm_vram_mgr_func;
>>>  	mgr->mem_type = mem_type;
>>>  	mutex_init(&mgr->lock);
>>> @@ -318,6 +311,18 @@ int __xe_ttm_vram_mgr_init(struct xe_device
>>> *xe, struct xe_ttm_vram_mgr *mgr,
>>>  	mgr->visible_avail = io_size;
>>>  
>>>  	ttm_resource_manager_init(man, &xe->ttm, size);
>>> +
>>> +	if (mem_type != XE_PL_STOLEN) {
>>> +		const char *name = mem_type == XE_PL_VRAM0 ?
>>> "vram0" : "vram1";
>>> +		struct dmem_cgroup_region *cg =
>>> +			drmm_cgroup_register_region(&xe->drm,
>>> name, size);
>>> +
>>> +		if (IS_ERR(cg))
>>> +			return PTR_ERR(cg);
>>> +
>>> +		ttm_resource_manager_set_dmem_region(man, cg);
>>> +	}
>>> +
>>>  	err = gpu_buddy_init(&mgr->mm, man->size,
>>> default_page_size);
>>>  	if (err)
>>>  		return err;
>>
>> This patch will conflict with 
>> https://patchwork.freedesktop.org/series/164694/ which removes
>> stolen support, can we merge that patch first while we wait for AMD
>> acks?
> 
> Sure, np.
> 
>>
>> Do I need an ack to get the series through drm-misc?
> 
> Which series? The stolen support or this cgroup series?
This one, I think it might be easiest to just merge stolen through xe,
and then backmerge into drm-misc.