[PATCH v4 19/20] drm/xe/uapi: Add UAPI for querying VMA count and memory attributes
Matthew Brost
matthew.brost at intel.com
Mon Jun 23 22:43:08 UTC 2025
On Fri, Jun 13, 2025 at 06:25:57PM +0530, Himal Prasad Ghimiray wrote:
> Introduce the DRM_IOCTL_XE_VM_QUERY_VMAS_ATTRS ioctl to allow userspace
> to query memory attributes of VMAs within a specified virtual address
> range.
> If num_vmas == 0 and vector_of_vma_mem_attr == NULL, the ioctl returns
> the number of VMAs in the specified range.
> If num_vmas > 0 and a valid user pointer is provided in
> vector_of_vma_mem_attr, the ioctl fills the buffer with memory
> attributes for each VMA.
> This two-step interface allows userspace to first query the required
> buffer size, then retrieve detailed attributes efficiently.
>
> v2 (Matthew Brost)
> - Use same ioctl to overload functionality
>
> v3
> - Add kernel-doc
>
> Cc: Matthew Brost <matthew.brost at intel.com>
> Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray at intel.com>
> ---
> drivers/gpu/drm/xe/xe_device.c | 1 +
> drivers/gpu/drm/xe/xe_vm.c | 89 ++++++++++++++++++++++
> drivers/gpu/drm/xe/xe_vm.h | 2 +-
> include/uapi/drm/xe_drm.h | 130 +++++++++++++++++++++++++++++++++
> 4 files changed, 221 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
> index 632d3ab12392..1f9969acbcd9 100644
> --- a/drivers/gpu/drm/xe/xe_device.c
> +++ b/drivers/gpu/drm/xe/xe_device.c
> @@ -199,6 +199,7 @@ static const struct drm_ioctl_desc xe_ioctls[] = {
> DRM_RENDER_ALLOW),
> DRM_IOCTL_DEF_DRV(XE_OBSERVATION, xe_observation_ioctl, DRM_RENDER_ALLOW),
> DRM_IOCTL_DEF_DRV(XE_MADVISE, xe_vm_madvise_ioctl, DRM_RENDER_ALLOW),
> + DRM_IOCTL_DEF_DRV(XE_VM_QUERY_VMAS_ATTRS, xe_vm_query_vmas_attrs_ioctl, DRM_RENDER_ALLOW),
> };
>
> static long xe_drm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index 14e554f3f4d5..d9ce25f3abf4 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -2171,6 +2171,95 @@ int xe_vm_destroy_ioctl(struct drm_device *dev, void *data,
> return err;
> }
>
> +static void xe_vm_query_vmas(struct xe_vm *vm, u32 *num_vmas, u64 start, u64 end)
> +{
> + struct drm_gpuva *gpuva;
> +
> + lockdep_assert_held(&vm->lock);
> + drm_gpuvm_for_each_va_range(gpuva, &vm->gpuvm, start, end)
> + (*num_vmas)++;
Return num_vmas rather than pass by reference?
> +}
> +
> +static int get_mem_attrs(struct xe_vm *vm, u32 *num_vmas, u64 start,
> + u64 end, struct drm_xe_vma_mem_attr *attrs)
> +{
> + struct drm_gpuva *gpuva;
> + int i = 0;
> +
> + lockdep_assert_held(&vm->lock);
> +
> + drm_gpuvm_for_each_va_range(gpuva, &vm->gpuvm, start, end) {
> + struct xe_vma *vma = gpuva_to_vma(gpuva);
> +
> + if (i == *num_vmas)
> + return -EINVAL;
Hmm, if the number of num_vmas has changed behind users back (e.g.,
binds happen), I think it is fair to return an error. But I think
-ENOSPC would better and also document what this error code means (e.g.,
call the query IOCTL again with num_vmas == 0 and try again.
> +
> + attrs[i].start = xe_vma_start(vma);
> + attrs[i].end = xe_vma_end(vma);
> + attrs[i].atomic.val = vma->attr.atomic_access;
> + attrs[i].pat_index.val = vma->attr.pat_index;
> + attrs[i].preferred_mem_loc.devmem_fd = vma->attr.preferred_loc.devmem_fd;
> + attrs[i].preferred_mem_loc.migration_policy = vma->attr.preferred_loc.migration_policy;
You should either explictly set the reserved bits to zero or kvzalloc
the attrs.
> +
> + i++;
> + }
> +
> + if (i < (*num_vmas - 1))
> + *num_vmas = i;
> + return 0;
> +}
> +
> +int xe_vm_query_vmas_attrs_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
> +{
> + struct xe_device *xe = to_xe_device(dev);
> + struct xe_file *xef = to_xe_file(file);
> + struct drm_xe_vma_mem_attr *mem_attrs;
> + struct drm_xe_vm_query_vmas_attr *args = data;
> + u64 __user *attrs_user = u64_to_user_ptr(args->vector_of_vma_mem_attr);
> + struct xe_vm *vm;
> + int err = 0;
> +
> + if (XE_IOCTL_DBG(xe,
> + (args->num_vmas == 0 && attrs_user) ||
> + (args->num_vmas > 0 && !attrs_user)))
> + return -EINVAL;
> +
> + vm = xe_vm_lookup(xef, args->vm_id);
> + if (XE_IOCTL_DBG(xe, !vm))
> + return -EINVAL;
> +
> + down_read(&vm->lock);
down_read_interruptable
> +
> + attrs_user = u64_to_user_ptr(args->vector_of_vma_mem_attr);
> +
> + if (args->num_vmas == 0 && !attrs_user) {
> + xe_vm_query_vmas(vm, &args->num_vmas, args->start, args->start + args->range);
> + goto unlock_vm;
> + }
> +
> + mem_attrs = kvmalloc_array(args->num_vmas, sizeof(struct drm_xe_vma_mem_attr),
> + GFP_KERNEL | __GFP_ACCOUNT |
> + __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
> + if (!mem_attrs) {
> + err = args->num_vmas > 1 ? -ENOBUFS : -ENOMEM;
> + goto unlock_vm;
> + }
> +
> + err = get_mem_attrs(vm, &args->num_vmas, args->start,
> + args->start + args->range, mem_attrs);
> + if (err)
> + goto free_mem_attrs;
> +
> + err = __copy_to_user(attrs_user, mem_attrs,
I think copy_to_user is preferred. This is fixed in a bunch of places in
Xe recently.
> + sizeof(struct drm_xe_vma_mem_attr) * args->num_vmas);
> +
> +free_mem_attrs:
> + kvfree(mem_attrs);
> +unlock_vm:
> + up_read(&vm->lock);
> + return err;
> +}
> +
> static bool vma_matches(struct xe_vma *vma, u64 page_addr)
> {
> if (page_addr > xe_vma_end(vma) - 1 ||
> diff --git a/drivers/gpu/drm/xe/xe_vm.h b/drivers/gpu/drm/xe/xe_vm.h
> index 1fb639a33ffb..e4a601a3583f 100644
> --- a/drivers/gpu/drm/xe/xe_vm.h
> +++ b/drivers/gpu/drm/xe/xe_vm.h
> @@ -195,7 +195,7 @@ int xe_vm_destroy_ioctl(struct drm_device *dev, void *data,
> struct drm_file *file);
> int xe_vm_bind_ioctl(struct drm_device *dev, void *data,
> struct drm_file *file);
> -
> +int xe_vm_query_vmas_attrs_ioctl(struct drm_device *dev, void *data, struct drm_file *file);
> void xe_vm_close_and_put(struct xe_vm *vm);
>
> static inline bool xe_vm_in_fault_mode(struct xe_vm *vm)
> diff --git a/include/uapi/drm/xe_drm.h b/include/uapi/drm/xe_drm.h
> index 81e90270313d..c0aa7c19870e 100644
> --- a/include/uapi/drm/xe_drm.h
> +++ b/include/uapi/drm/xe_drm.h
> @@ -82,6 +82,7 @@ extern "C" {
> * - &DRM_IOCTL_XE_WAIT_USER_FENCE
> * - &DRM_IOCTL_XE_OBSERVATION
> * - &DRM_IOCTL_XE_MADVISE
> + * - &DRM_IOCTL_XE_VM_QUERY_VMAS_ATTRS
> */
>
> /*
> @@ -104,6 +105,7 @@ extern "C" {
> #define DRM_XE_WAIT_USER_FENCE 0x0a
> #define DRM_XE_OBSERVATION 0x0b
> #define DRM_XE_MADVISE 0x0c
> +#define DRM_XE_VM_QUERY_VMAS_ATTRS 0x0d
>
> /* Must be kept compact -- no holes */
>
> @@ -120,6 +122,7 @@ extern "C" {
> #define DRM_IOCTL_XE_WAIT_USER_FENCE DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_WAIT_USER_FENCE, struct drm_xe_wait_user_fence)
> #define DRM_IOCTL_XE_OBSERVATION DRM_IOW(DRM_COMMAND_BASE + DRM_XE_OBSERVATION, struct drm_xe_observation_param)
> #define DRM_IOCTL_XE_MADVISE DRM_IOW(DRM_COMMAND_BASE + DRM_XE_MADVISE, struct drm_xe_madvise)
> +#define DRM_IOCTL_XE_VM_QUERY_VMAS_ATTRS DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_VM_QUERY_VMAS_ATTRS, struct drm_xe_vm_query_vmas_attr)
>
> /**
> * DOC: Xe IOCTL Extensions
> @@ -2093,6 +2096,133 @@ struct drm_xe_madvise {
> __u64 reserved[2];
> };
>
> +/**
> + * struct drm_xe_vma_mem_attr - Output of &DRM_IOCTL_XE_VM_QUERY_VMAS_ATTRS
> + *
> + * This structure is provided by userspace and filled by KMD in response to the
> + * DRM_IOCTL_XE_VM_QUERY_VMAS_ATTRS ioctl. It describes memory attributes of
> + * a VMA within a specified address range in a VM.
> + *
> + * The structure includes information such as atomic access policy, purgeable
> + * state, page attribute table (PAT) index, and preferred memory location.
> + * Userspace allocates an array of these structures and passes a pointer to the
> + * ioctl to retrieve attributes for each VMA
> + *
> + * @extensions: Pointer to the first extension struct, if any
> + * @start: Start address of the virtual memory area
> + * @end: End address of the virtual memory area
> + *
> + * @atomic.val: Value of atomic operation policy
> + * @atomic.reserved: Reserved, must be zero
> + *
> + * @purge_state_val.val: Value for DRM_XE_VMA_ATTR_PURGEABLE_STATE
> + * @purge_state_val.reserved: Reserved, must be zero
> + *
> + * @pat_index.val: Page Attribute Table (PAT) index
> + * @pat_index.reserved: Reserved, must be zero
> + *
> + * @preferred_mem_loc.devmem_fd: File descriptor for preferred device memory
> + * @preferred_mem_loc.migration_policy: Migration policy for memory placement
Double kernel doc for the fields?
> + *
> + */
> +struct drm_xe_vma_mem_attr {
> + /** @extensions: Pointer to the first extension struct, if any */
> + __u64 extensions;
> +
> + /** @start: start of the vma */
> + __u64 start;
> +
> + /** @end: end of the vma */
> + __u64 end;
> +
> + struct {
Same thing here, match the order of the defines for these.
Also no need for the outer struct here.
> + struct {
> + /** @atomic.val: atomic attribute for vma*/
> + __u32 val;
> +
> + /** @atomic.reserved: Reserved */
Alignment is off in my editor.
> + __u32 reserved;
> + } atomic;
> +
> + struct {
> + /** @pat_index.val: PAT index of vma */
> + __u32 val;
> +
> + /** @pat_index.reserved: Reserved */
> + __u32 reserved;
> + } pat_index;
> +
> + /** @preferred_mem_loc: preferred memory location */
> + struct {
> + /** @preferred_mem_loc.devmem_fd: fd for preferred loc */
> + __u32 devmem_fd;
> +
> + /** @preferred_mem_loc.migration_policy: Page migration policy */
> + __u32 migration_policy;
> + } preferred_mem_loc;
> + };
> +
> + /** @reserved: Reserved */
> + __u64 reserved[2];
> +};
> +
> +/**
> + * struct drm_xe_vm_query_vmas_attr - Input of &DRM_IOCTL_XE_VM_QUERY_MEM_ATTRIBUTES
> + *
> + * This structure is used to query memory attributes of virtual memory areas
> + * (VMAs) within a specified address range in a VM. It provides detailed
> + * information about each VMA, including atomic access policy, purgeable state,
> + * page attribute table (PAT) index, and preferred memory location.
> + *
> + * Userspace first calls the ioctl with @num_vmas = 0 and
> + * @vector_of_vma_mem_attr = NULL to retrieve the number of VMAs in the range.
> + * Then, it allocates a buffer of that size and calls the ioctl again to fill
> + * the buffer with VMA attributes.
> + *
> + * Example:
> + *
> + * .. code-block:: C
> + *
I'd show how query is initialized. See drm_xe_device_query kernel doc
for an example.
Matt
> + * struct drm_xe_vma_mem_attr *attrs;
> + * __u32 num_vmas;
> + *
> + * // First call to get number of VMAs
> + * ioctl(fd, DRM_IOCTL_XE_VM_QUERY_VMAS_ATTRS, &query);
> + * num_vmas = query.num_vmas;
> + *
> + * // Allocate and query attributes
> + * attrs = malloc(num_vmas, sizeof(*attrs));
> + * query.vector_of_vma_mem_attr = (__u64)(uintptr_t)attrs;
> + * query.num_vmas = num_vmas;
> + * ioctl(fd, DRM_IOCTL_XE_VM_QUERY_VMAS_ATTRS, &query);
> + */
> +struct drm_xe_vm_query_vmas_attr {
> + /** @extensions: Pointer to the first extension struct, if any */
> + __u64 extensions;
> +
> + /** @vm_id: vm_id of the virtual range */
> + __u32 vm_id;
> +
> + /** @num_vmas: number of vmas in range */
> + __u32 num_vmas;
> +
> + /** @start: start of the virtual address range */
> + __u64 start;
> +
> + /** @size: size of the virtual address range */
> + __u64 range;
> +
> + /**
> + * @vector_of_ops: userptr to array of struct
> + * drm_xe_vma_mem_attr
> + */
> + __u64 vector_of_vma_mem_attr;
> +
> + /** @reserved: Reserved */
> + __u64 reserved[2];
> +
> +};
> +
> #if defined(__cplusplus)
> }
> #endif
> --
> 2.34.1
>
More information about the Intel-xe
mailing list