Re: [PATCH i-g-t v6 2/9] lib/xe: Add purgeable memory ioctl support

["Yadav, Arvind" <arvind.yadav@intel.com>]

On 06-04-2026 12:29, Sharma, Nishit wrote:
>
> On 3/25/2026 6:14 PM, Arvind Yadav wrote:
>> Add xe_vm_madvise_purgeable() helper function to support purgeable
>> memory management through the XE madvise ioctl. This allows applications
>> to hint to the kernel about buffer object usage patterns for better
>> memory management under pressure.
>>
>> The function provides a clean interface to:
>> - Mark buffer objects as DONTNEED (purgeable)
>> - Mark buffer objects as WILLNEED (not purgeable)
>>
>> Returns the retained value directly (1 if backing store exists, 0 if
>> purged).
>>
>> Also update __xe_vm_madvise() to reject purgeable state operations
>> and direct users to the dedicated helper.
>
> Here you can also mention why this new ioctl when __xe_vm_madvise() 
> already available.
>
> You can mention the changes required.


Noted, I will update.
The existing __xe_vm_madvise() uses a generic type/op_val pattern that 
maps to a single scalar field (e.g., pat_index.val, atomic.val). 
Purgeable state requires a different ioctl layout — it populates 
purge_state_val.val (the state) and purge_state_val.retained_ptr (an 
output pointer the kernel writes back to), which doesn't fit the generic 
op_val model. so I have added this new one.

Thanks,
Arvind

>
>>
>> v2:
>>    - retained must be initialized to 0(Thomas)
>>
>> v5:
>>    - Rename retained to retained_ptr. (Jose)
>>
>> Cc: Nishit Sharma <nishit.sharma@intel.com>
>> Cc: Matthew Brost <matthew.brost@intel.com>
>> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
>> Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
>> Cc: Pravalika Gurram <pravalika.gurram@intel.com>
>> Signed-off-by: Arvind Yadav <arvind.yadav@intel.com>
>> ---
>>   lib/xe/xe_ioctl.c | 33 +++++++++++++++++++++++++++++++++
>>   lib/xe/xe_ioctl.h |  2 ++
>>   2 files changed, 35 insertions(+)
>>
>> diff --git a/lib/xe/xe_ioctl.c b/lib/xe/xe_ioctl.c
>> index ea3f2fcaa..955607f93 100644
>> --- a/lib/xe/xe_ioctl.c
>> +++ b/lib/xe/xe_ioctl.c
>> @@ -785,6 +785,9 @@ int __xe_vm_madvise(int fd, uint32_t vm, uint64_t 
>> addr, uint64_t range,
>>       case DRM_XE_MEM_RANGE_ATTR_PAT:
>>           madvise.pat_index.val = op_val;
>>           break;
>> +    case DRM_XE_VMA_ATTR_PURGEABLE_STATE:
>> +        /* Purgeable state handled by xe_vm_madvise_purgeable */
>> +        return -EINVAL;
>>       default:
>>           igt_warn("Unknown attribute\n");
>>           return -EINVAL;
>> @@ -821,6 +824,36 @@ void xe_vm_madvise(int fd, uint32_t vm, uint64_t 
>> addr, uint64_t range,
>>                         instance), 0);
>>   }
>>   +/**
>> + * xe_vm_madvise_purgeable:
>> + * @fd: xe device fd
>> + * @vm_id: vm_id of the virtual range
>> + * @start: start of the virtual address range
>> + * @range: size of the virtual address range
>> + * @state: purgeable state (DRM_XE_VMA_PURGEABLE_STATE_WILLNEED or 
>> DONTNEED)
>> + *
>> + * Sets the purgeable state for a virtual memory range. This allows 
>> applications
>> + * to hint to the kernel about buffer object usage patterns for 
>> better memory management.
>> + *
>> + * Returns: retained value (1 if backing store exists, 0 if purged)
>> + */
>> +uint32_t xe_vm_madvise_purgeable(int fd, uint32_t vm_id, uint64_t 
>> start,
>> +                 uint64_t range, uint32_t state)
>> +{
>> +    uint32_t retained_val = 0;
>> +    struct drm_xe_madvise madvise = {
>> +        .vm_id = vm_id,
>> +        .start = start,
>> +        .range = range,
>> +        .type = DRM_XE_VMA_ATTR_PURGEABLE_STATE,
>> +        .purge_state_val.val = state,
>> +        .purge_state_val.retained_ptr = 
>> (uint64_t)(uintptr_t)&retained_val,
>> +    };
>> +
>> +    igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_MADVISE, &madvise), 0);
>> +    return retained_val;
>> +}
>> +
>>   #define    BIND_SYNC_VAL    0x686868
>>   void xe_vm_bind_lr_sync(int fd, uint32_t vm, uint32_t bo, uint64_t 
>> offset,
>>               uint64_t addr, uint64_t size, uint32_t flags)
>> diff --git a/lib/xe/xe_ioctl.h b/lib/xe/xe_ioctl.h
>> index b62d259fd..a02d68cfe 100644
>> --- a/lib/xe/xe_ioctl.h
>> +++ b/lib/xe/xe_ioctl.h
>> @@ -108,6 +108,8 @@ int __xe_vm_madvise(int fd, uint32_t vm, uint64_t 
>> addr, uint64_t range, uint64_t
>>               uint32_t type, uint32_t op_val, uint16_t policy, 
>> uint16_t instance);
>>   void xe_vm_madvise(int fd, uint32_t vm, uint64_t addr, uint64_t 
>> range, uint64_t ext,
>>              uint32_t type, uint32_t op_val, uint16_t policy, 
>> uint16_t instance);
>> +uint32_t xe_vm_madvise_purgeable(int fd, uint32_t vm_id, uint64_t 
>> start,
>> +                 uint64_t range, uint32_t state);
>>   int xe_vm_number_vmas_in_range(int fd, struct 
>> drm_xe_vm_query_mem_range_attr *vmas_attr);
>>   int xe_vm_vma_attrs(int fd, struct drm_xe_vm_query_mem_range_attr 
>> *vmas_attr,
>>               struct drm_xe_mem_range_attr *mem_attr);
>
> with above change in commit message LGTM:
>
> Reviewed-by: Nishit Sharma <nishit.sharma@intel.com>
>