[RFC PATCH 0/5] Add memory page offlining support

[RFC PATCH 0/5] Add memory page offlining support

[Tejas Upadhyay]

This functionality represents a significant step in making
the xe driver gracefully handle hardware memory degradation.
By integrating with the DRM Buddy allocator, the driver
can permanently "carve out" faulty memory so it isn't reused
by subsequent allocations.

This series adds memory page offlining support with following:
1. Link and track ttm BO's with physical addresses
2. Handle the generated physical address error by reserving addresses 4K page
3. Adds supporting debugfs to inject manual physcal address error
4. Add buddy block allocation dump for debuggin buddy related issues
5. Sysfs entry to provide statistics of bad gpu vram pages for user info

Opens:
1. mm->avail under drm_buddy throwing WARN_ON(mm->avail != mm->size) with no leaks 
in memory, mutliple bind/ubind works fine. Debug in progress.
2. dump_allocated_blocks() and xe_ttm_vram_addr_to_tbo() API will move under drm_buddy, 
right now just to showcase concept its part of xe code

Tejas Upadhyay (5):
  drm/xe: Implement VRAM object tracking ability using physical address
  drm/xe: Handle physical memory address error
  [DO_NOT_REVIEW]drm/xe/cri: Add debugfs to inject faulty vram address
  drm/xe: Add routine to dump allocated VRAM blocks
  [DO NOT REVIEW]drm/xe/cri: Add sysfs interface for bad gpu vram pages

 drivers/gpu/drm/xe/xe_debugfs.c            |  49 +++
 drivers/gpu/drm/xe/xe_device_sysfs.c       |   2 +
 drivers/gpu/drm/xe/xe_tile_sysfs.c         |   1 +
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.c       | 366 +++++++++++++++++++++
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.h       |   6 +-
 drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h |  23 ++
 6 files changed, 446 insertions(+), 1 deletion(-)

-- 
2.52.0

[RFC PATCH 1/5] drm/xe: Implement VRAM object tracking ability using physical address

[Tejas Upadhyay]

Implement the capability to track and identify TTM buffer objects
using a specific faulty memory address in VRAM. This functionality
is critical for supporting the memory page offline feature on CRI,
where identified faulty pages must be traced back to their
originating buffer for safe removal.

Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
---
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.c | 75 ++++++++++++++++++++++++++++
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.h |  2 +-
 2 files changed, 76 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
index d6aa61e55f4d..4e852eed5170 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
@@ -56,6 +56,7 @@ static int xe_ttm_vram_mgr_new(struct ttm_resource_manager *man,
 	u64 size, min_page_size;
 	unsigned long lpfn;
 	int err;
+	struct drm_buddy_block *block;
 
 	lpfn = place->lpfn;
 	if (!lpfn || lpfn > man->size >> PAGE_SHIFT)
@@ -137,6 +138,8 @@ static int xe_ttm_vram_mgr_new(struct ttm_resource_manager *man,
 	}
 
 	mgr->visible_avail -= vres->used_visible_size;
+	list_for_each_entry(block, &vres->blocks, link)
+		block->private = tbo;
 	mutex_unlock(&mgr->lock);
 
 	if (!(vres->base.placement & TTM_PL_FLAG_CONTIGUOUS) &&
@@ -467,3 +470,75 @@ u64 xe_ttm_vram_get_avail(struct ttm_resource_manager *man)
 
 	return avail;
 }
+
+static inline bool overlaps(u64 s1, u64 e1, u64 s2, u64 e2)
+{
+	return s1 <= e2 && e1 >= s2;
+}
+
+static inline bool contains(u64 s1, u64 e1, u64 s2, u64 e2)
+{
+	return s1 <= s2 && e1 <= e2;
+}
+
+static struct ttm_buffer_object *xe_ttm_vram_addr_to_tbo(struct drm_buddy *mm, u64 start)
+{
+	struct drm_buddy_block *block;
+	u64 end;
+	LIST_HEAD(dfs);
+	int i;
+
+	end = start + SZ_4K - 1;
+	for (i = 0; i < mm->n_roots; ++i)
+		list_add_tail(&mm->roots[i]->tmp_link, &dfs);
+
+	do {
+		u64 block_start;
+		u64 block_end;
+
+		block = list_first_entry_or_null(&dfs,
+						 struct drm_buddy_block,
+						 tmp_link);
+		if (!block)
+			break;
+
+		list_del(&block->tmp_link);
+
+		block_start = drm_buddy_block_offset(block);
+		block_end = block_start + drm_buddy_block_size(mm, block) - 1;
+
+		if (!overlaps(start, end, block_start, block_end))
+			continue;
+
+		if (contains(start, end, block_start, block_end) &&
+		    !drm_buddy_block_is_split(block)) {
+			if (drm_buddy_block_is_free(block)) {
+				return NULL;
+			} else if (drm_buddy_block_is_allocated(block) && !mm->clear_avail) {
+				struct ttm_buffer_object *tbo = block->private;
+
+				WARN_ON(!tbo);
+				return tbo;
+			}
+		}
+
+		if (drm_buddy_block_is_split(block)) {
+			list_add(&block->right->tmp_link, &dfs);
+			list_add(&block->left->tmp_link, &dfs);
+		}
+	} while (1);
+
+	return NULL;
+}
+
+int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr)
+{
+	struct xe_ttm_vram_mgr *vram_mgr = &tile->mem.vram->ttm;
+	struct drm_buddy mm = vram_mgr->mm;
+	struct ttm_buffer_object *tbo;
+
+	tbo = xe_ttm_vram_addr_to_tbo(&mm, addr);
+
+	return 0;
+}
+EXPORT_SYMBOL(xe_ttm_tbo_handle_addr_fault);
diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
index 87b7fae5edba..1d6075411ebf 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
@@ -30,7 +30,7 @@ u64 xe_ttm_vram_get_avail(struct ttm_resource_manager *man);
 u64 xe_ttm_vram_get_cpu_visible_size(struct ttm_resource_manager *man);
 void xe_ttm_vram_get_used(struct ttm_resource_manager *man,
 			  u64 *used, u64 *used_visible);
-
+int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr);
 static inline struct xe_ttm_vram_mgr_resource *
 to_xe_ttm_vram_mgr_resource(struct ttm_resource *res)
 {
-- 
2.52.0

Re: [RFC PATCH 1/5] drm/xe: Implement VRAM object tracking ability using physical address

[Matthew Brost]

On Wed, Feb 11, 2026 at 10:31:34AM +0530, Tejas Upadhyay wrote:

It will take a bit of time to get through this but spotted an issue.

> Implement the capability to track and identify TTM buffer objects
> using a specific faulty memory address in VRAM. This functionality
> is critical for supporting the memory page offline feature on CRI,
> where identified faulty pages must be traced back to their
> originating buffer for safe removal.
> 
> Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_ttm_vram_mgr.c | 75 ++++++++++++++++++++++++++++
>  drivers/gpu/drm/xe/xe_ttm_vram_mgr.h |  2 +-
>  2 files changed, 76 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> index d6aa61e55f4d..4e852eed5170 100644
> --- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> +++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> @@ -56,6 +56,7 @@ static int xe_ttm_vram_mgr_new(struct ttm_resource_manager *man,
>  	u64 size, min_page_size;
>  	unsigned long lpfn;
>  	int err;
> +	struct drm_buddy_block *block;
>  
>  	lpfn = place->lpfn;
>  	if (!lpfn || lpfn > man->size >> PAGE_SHIFT)
> @@ -137,6 +138,8 @@ static int xe_ttm_vram_mgr_new(struct ttm_resource_manager *man,
>  	}
>  
>  	mgr->visible_avail -= vres->used_visible_size;
> +	list_for_each_entry(block, &vres->blocks, link)
> +		block->private = tbo;
>  	mutex_unlock(&mgr->lock);
>  
>  	if (!(vres->base.placement & TTM_PL_FLAG_CONTIGUOUS) &&
> @@ -467,3 +470,75 @@ u64 xe_ttm_vram_get_avail(struct ttm_resource_manager *man)
>  
>  	return avail;
>  }
> +
> +static inline bool overlaps(u64 s1, u64 e1, u64 s2, u64 e2)
> +{
> +	return s1 <= e2 && e1 >= s2;
> +}
> +
> +static inline bool contains(u64 s1, u64 e1, u64 s2, u64 e2)
> +{
> +	return s1 <= s2 && e1 <= e2;
> +}
> +
> +static struct ttm_buffer_object *xe_ttm_vram_addr_to_tbo(struct drm_buddy *mm, u64 start)
> +{
> +	struct drm_buddy_block *block;
> +	u64 end;
> +	LIST_HEAD(dfs);
> +	int i;
> +
> +	end = start + SZ_4K - 1;
> +	for (i = 0; i < mm->n_roots; ++i)
> +		list_add_tail(&mm->roots[i]->tmp_link, &dfs);
> +
> +	do {
> +		u64 block_start;
> +		u64 block_end;
> +
> +		block = list_first_entry_or_null(&dfs,
> +						 struct drm_buddy_block,
> +						 tmp_link);
> +		if (!block)
> +			break;
> +
> +		list_del(&block->tmp_link);
> +
> +		block_start = drm_buddy_block_offset(block);
> +		block_end = block_start + drm_buddy_block_size(mm, block) - 1;
> +
> +		if (!overlaps(start, end, block_start, block_end))
> +			continue;
> +
> +		if (contains(start, end, block_start, block_end) &&
> +		    !drm_buddy_block_is_split(block)) {
> +			if (drm_buddy_block_is_free(block)) {
> +				return NULL;
> +			} else if (drm_buddy_block_is_allocated(block) && !mm->clear_avail) {
> +				struct ttm_buffer_object *tbo = block->private;

The SVM code sets block->private to 'struct xe_vram_region' but we
definitely don't need too as there are other ways to look this up. Want
to fixup xe_svm.c in an independent patch to setting or looking at
block->private?

Matt

> +
> +				WARN_ON(!tbo);
> +				return tbo;
> +			}
> +		}
> +
> +		if (drm_buddy_block_is_split(block)) {
> +			list_add(&block->right->tmp_link, &dfs);
> +			list_add(&block->left->tmp_link, &dfs);
> +		}
> +	} while (1);
> +
> +	return NULL;
> +}
> +
> +int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr)
> +{
> +	struct xe_ttm_vram_mgr *vram_mgr = &tile->mem.vram->ttm;
> +	struct drm_buddy mm = vram_mgr->mm;
> +	struct ttm_buffer_object *tbo;
> +
> +	tbo = xe_ttm_vram_addr_to_tbo(&mm, addr);
> +
> +	return 0;
> +}
> +EXPORT_SYMBOL(xe_ttm_tbo_handle_addr_fault);
> diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
> index 87b7fae5edba..1d6075411ebf 100644
> --- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
> +++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
> @@ -30,7 +30,7 @@ u64 xe_ttm_vram_get_avail(struct ttm_resource_manager *man);
>  u64 xe_ttm_vram_get_cpu_visible_size(struct ttm_resource_manager *man);
>  void xe_ttm_vram_get_used(struct ttm_resource_manager *man,
>  			  u64 *used, u64 *used_visible);
> -
> +int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr);
>  static inline struct xe_ttm_vram_mgr_resource *
>  to_xe_ttm_vram_mgr_resource(struct ttm_resource *res)
>  {
> -- 
> 2.52.0
> 

RE: [RFC PATCH 1/5] drm/xe: Implement VRAM object tracking ability using physical address

[Upadhyay, Tejas]

> -----Original Message-----
> From: Brost, Matthew <matthew.brost@intel.com>
> Sent: 11 February 2026 11:57
> To: Upadhyay, Tejas <tejas.upadhyay@intel.com>
> Cc: intel-xe@lists.freedesktop.org; Auld, Matthew
> <matthew.auld@intel.com>; Ghimiray, Himal Prasad
> <himal.prasad.ghimiray@intel.com>
> Subject: Re: [RFC PATCH 1/5] drm/xe: Implement VRAM object tracking ability
> using physical address
> 
> On Wed, Feb 11, 2026 at 10:31:34AM +0530, Tejas Upadhyay wrote:
> 
> It will take a bit of time to get through this but spotted an issue.
> 
> > Implement the capability to track and identify TTM buffer objects
> > using a specific faulty memory address in VRAM. This functionality is
> > critical for supporting the memory page offline feature on CRI, where
> > identified faulty pages must be traced back to their originating
> > buffer for safe removal.
> >
> > Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
> > ---
> >  drivers/gpu/drm/xe/xe_ttm_vram_mgr.c | 75
> > ++++++++++++++++++++++++++++
> drivers/gpu/drm/xe/xe_ttm_vram_mgr.h |
> > 2 +-
> >  2 files changed, 76 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> > b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> > index d6aa61e55f4d..4e852eed5170 100644
> > --- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> > +++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> > @@ -56,6 +56,7 @@ static int xe_ttm_vram_mgr_new(struct
> ttm_resource_manager *man,
> >  	u64 size, min_page_size;
> >  	unsigned long lpfn;
> >  	int err;
> > +	struct drm_buddy_block *block;
> >
> >  	lpfn = place->lpfn;
> >  	if (!lpfn || lpfn > man->size >> PAGE_SHIFT) @@ -137,6 +138,8 @@
> > static int xe_ttm_vram_mgr_new(struct ttm_resource_manager *man,
> >  	}
> >
> >  	mgr->visible_avail -= vres->used_visible_size;
> > +	list_for_each_entry(block, &vres->blocks, link)
> > +		block->private = tbo;
> >  	mutex_unlock(&mgr->lock);
> >
> >  	if (!(vres->base.placement & TTM_PL_FLAG_CONTIGUOUS) && @@ -
> 467,3
> > +470,75 @@ u64 xe_ttm_vram_get_avail(struct ttm_resource_manager
> *man)
> >
> >  	return avail;
> >  }
> > +
> > +static inline bool overlaps(u64 s1, u64 e1, u64 s2, u64 e2) {
> > +	return s1 <= e2 && e1 >= s2;
> > +}
> > +
> > +static inline bool contains(u64 s1, u64 e1, u64 s2, u64 e2) {
> > +	return s1 <= s2 && e1 <= e2;
> > +}
> > +
> > +static struct ttm_buffer_object *xe_ttm_vram_addr_to_tbo(struct
> > +drm_buddy *mm, u64 start) {
> > +	struct drm_buddy_block *block;
> > +	u64 end;
> > +	LIST_HEAD(dfs);
> > +	int i;
> > +
> > +	end = start + SZ_4K - 1;
> > +	for (i = 0; i < mm->n_roots; ++i)
> > +		list_add_tail(&mm->roots[i]->tmp_link, &dfs);
> > +
> > +	do {
> > +		u64 block_start;
> > +		u64 block_end;
> > +
> > +		block = list_first_entry_or_null(&dfs,
> > +						 struct drm_buddy_block,
> > +						 tmp_link);
> > +		if (!block)
> > +			break;
> > +
> > +		list_del(&block->tmp_link);
> > +
> > +		block_start = drm_buddy_block_offset(block);
> > +		block_end = block_start + drm_buddy_block_size(mm, block)
> - 1;
> > +
> > +		if (!overlaps(start, end, block_start, block_end))
> > +			continue;
> > +
> > +		if (contains(start, end, block_start, block_end) &&
> > +		    !drm_buddy_block_is_split(block)) {
> > +			if (drm_buddy_block_is_free(block)) {
> > +				return NULL;
> > +			} else if (drm_buddy_block_is_allocated(block) &&
> !mm->clear_avail) {
> > +				struct ttm_buffer_object *tbo = block-
> >private;
> 
> The SVM code sets block->private to 'struct xe_vram_region' but we definitely
> don't need too as there are other ways to look this up. Want to fixup xe_svm.c
> in an independent patch to setting or looking at
> block->private?

Sure, will have a look.

Tejas
> 
> Matt
> 
> > +
> > +				WARN_ON(!tbo);
> > +				return tbo;
> > +			}
> > +		}
> > +
> > +		if (drm_buddy_block_is_split(block)) {
> > +			list_add(&block->right->tmp_link, &dfs);
> > +			list_add(&block->left->tmp_link, &dfs);
> > +		}
> > +	} while (1);
> > +
> > +	return NULL;
> > +}
> > +
> > +int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long
> > +addr) {
> > +	struct xe_ttm_vram_mgr *vram_mgr = &tile->mem.vram->ttm;
> > +	struct drm_buddy mm = vram_mgr->mm;
> > +	struct ttm_buffer_object *tbo;
> > +
> > +	tbo = xe_ttm_vram_addr_to_tbo(&mm, addr);
> > +
> > +	return 0;
> > +}
> > +EXPORT_SYMBOL(xe_ttm_tbo_handle_addr_fault);
> > diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
> > b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
> > index 87b7fae5edba..1d6075411ebf 100644
> > --- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
> > +++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
> > @@ -30,7 +30,7 @@ u64 xe_ttm_vram_get_avail(struct
> > ttm_resource_manager *man);
> >  u64 xe_ttm_vram_get_cpu_visible_size(struct ttm_resource_manager
> > *man);  void xe_ttm_vram_get_used(struct ttm_resource_manager *man,
> >  			  u64 *used, u64 *used_visible);
> > -
> > +int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long
> > +addr);
> >  static inline struct xe_ttm_vram_mgr_resource *
> > to_xe_ttm_vram_mgr_resource(struct ttm_resource *res)  {
> > --
> > 2.52.0
> >

[RFC PATCH 2/5] drm/xe: Handle physical memory address error

[Tejas Upadhyay]

This functionality represents a significant step in making
the xe driver gracefully handle hardware memory degradation.
By integrating with the DRM Buddy allocator, the driver
can permanently "carve out" faulty memory so it isn't reused
by subsequent allocations.

Buddy Block Reservation:
----------------------
When a memory address is reported as faulty, the driver instructs
the DRM Buddy allocator to reserve a block of the specific page
size (typically 4KB). This marks the memory as "dirty/used"
indefinitely.

Two-Stage Tracking:
-----------------
Offlined Pages:
Pages that have been successfully isolated and removed from the
available memory pool.

Queued Pages:
Addresses that have been flagged as faulty but are currently in
use by a process. These are tracked until the associated buffer
object (BO) is released or migrated, at which point they move
to the "offlined" state.

Sysfs Reporting:
--------------
The patch exposes these metrics through a standard interface,
allowing administrators to monitor VRAM health:
/sys/bus/pci/devices/<device_id>/vram_bad_bad_pages

Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
---
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.c       | 187 ++++++++++++++++++++-
 drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h |  21 +++
 2 files changed, 203 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
index 4e852eed5170..ab8243b17cf3 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
@@ -276,6 +276,26 @@ static const struct ttm_resource_manager_func xe_ttm_vram_mgr_func = {
 	.debug	= xe_ttm_vram_mgr_debug
 };
 
+static void xe_ttm_vram_free_bad_pages(struct drm_device *dev, struct xe_ttm_vram_mgr *mgr)
+{
+	struct xe_ttm_offline_resource *pos, *n;
+
+	mutex_lock(&mgr->lock);
+	list_for_each_entry_safe(pos, n, &mgr->offlined_pages, offlined_link) {
+		--mgr->n_offlined_pages;
+		drm_buddy_free_list(&mgr->mm, &pos->blocks, 0);
+		mgr->visible_avail += pos->used_visible_size;
+		list_del(&pos->offlined_link);
+		kfree(pos);
+	}
+	list_for_each_entry_safe(pos, n, &mgr->queued_pages, queued_link) {
+		list_del(&pos->queued_link);
+		mgr->n_queued_pages--;
+		kfree(pos);
+	}
+	mutex_unlock(&mgr->lock);
+}
+
 static void xe_ttm_vram_mgr_fini(struct drm_device *dev, void *arg)
 {
 	struct xe_device *xe = to_xe_device(dev);
@@ -287,6 +307,8 @@ static void xe_ttm_vram_mgr_fini(struct drm_device *dev, void *arg)
 	if (ttm_resource_manager_evict_all(&xe->ttm, man))
 		return;
 
+	xe_ttm_vram_free_bad_pages(dev, mgr);
+
 	WARN_ON_ONCE(mgr->visible_avail != mgr->visible_size);
 
 	drm_buddy_fini(&mgr->mm);
@@ -315,6 +337,8 @@ int __xe_ttm_vram_mgr_init(struct xe_device *xe, struct xe_ttm_vram_mgr *mgr,
 	man->func = &xe_ttm_vram_mgr_func;
 	mgr->mem_type = mem_type;
 	mutex_init(&mgr->lock);
+	INIT_LIST_HEAD(&mgr->offlined_pages);
+	INIT_LIST_HEAD(&mgr->queued_pages);
 	mgr->default_page_size = default_page_size;
 	mgr->visible_size = io_size;
 	mgr->visible_avail = io_size;
@@ -531,14 +555,167 @@ static struct ttm_buffer_object *xe_ttm_vram_addr_to_tbo(struct drm_buddy *mm, u
 	return NULL;
 }
 
-int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr)
+static int xe_ttm_vram_reserve_page_at_addr(struct xe_device *xe, unsigned long addr,
+					    struct xe_ttm_vram_mgr *vram_mgr, struct drm_buddy *mm)
 {
-	struct xe_ttm_vram_mgr *vram_mgr = &tile->mem.vram->ttm;
-	struct drm_buddy mm = vram_mgr->mm;
+	int ret = 0;
+	u64 size = SZ_4K;
 	struct ttm_buffer_object *tbo;
+	struct xe_ttm_offline_resource *nentry;
+
+	mutex_lock(&vram_mgr->lock);
+	tbo = xe_ttm_vram_addr_to_tbo(mm, addr);
+	if (IS_ERR(tbo)) {
+		drm_err(&xe->drm, "Could not traverse blocks ret:%ld... Exiting\n",
+			PTR_ERR(tbo));
+		mutex_unlock(&vram_mgr->lock);
+		return -EFAULT;
+	}
 
-	tbo = xe_ttm_vram_addr_to_tbo(&mm, addr);
+	nentry = kzalloc(sizeof(*nentry), GFP_KERNEL);
+	if (!nentry)
+		return -ENOMEM;
 
-	return 0;
+	INIT_LIST_HEAD(&nentry->blocks);
+	if (tbo) {
+		struct xe_ttm_vram_mgr_resource *pvres;
+		struct ttm_placement place = {};
+		struct ttm_operation_ctx ctx = {
+			.interruptible = false,
+			.gfp_retry_mayfail = false,
+		};
+		bool locked;
+		struct xe_ttm_offline_resource *pos, *n;
+		struct xe_bo *pbo = ttm_to_xe_bo(tbo);
+
+		xe_bo_get(pbo);
+		/* Critical kernel BO? */
+		if (pbo->ttm.type == ttm_bo_type_kernel &&
+		    !(pbo->flags & XE_BO_FLAG_FORCE_USER_VRAM)) {
+			mutex_unlock(&vram_mgr->lock);
+			xe_bo_put(pbo);
+			drm_warn(&xe->drm,
+				 "%s: corrupt addr: %lx in critical kernel bo, wedge now\n",
+				__func__, addr);
+			/* Wedge the device */
+			xe_device_declare_wedged(xe);
+			return -EIO;
+		}
+		pvres = to_xe_ttm_vram_mgr_resource(pbo->ttm.resource);
+		nentry->id = ++vram_mgr->n_queued_pages;
+		nentry->blocks = pvres->blocks;
+		list_add(&nentry->queued_link, &vram_mgr->queued_pages);
+		mutex_unlock(&vram_mgr->lock);
+
+		/* Purge BO containing address */
+		spin_lock(&pbo->ttm.bdev->lru_lock);
+		locked = dma_resv_trylock(pbo->ttm.base.resv);
+		spin_unlock(&pbo->ttm.bdev->lru_lock);
+		WARN_ON(!locked);
+		ret = ttm_bo_validate(&pbo->ttm, &place, &ctx);
+		drm_WARN_ON(&xe->drm, ret);
+		xe_bo_put(pbo);
+		if (locked)
+			dma_resv_unlock(pbo->ttm.base.resv);
+
+		/* Reserve page at address addr*/
+		mutex_lock(&vram_mgr->lock);
+		ret = drm_buddy_alloc_blocks(mm, addr, addr + size,
+					     size, size, &nentry->blocks,
+					     DRM_BUDDY_RANGE_ALLOCATION);
+
+		if (ret) {
+			drm_warn(&xe->drm, "Could not reserve page at addr:%lx, ret:%d\n",
+				 addr, ret);
+			mutex_unlock(&vram_mgr->lock);
+			return ret;
+		}
+		if ((addr + size) <= vram_mgr->visible_size) {
+			nentry->used_visible_size = size;
+		} else {
+			struct drm_buddy_block *block;
+
+			list_for_each_entry(block, &nentry->blocks, link) {
+				u64 start = drm_buddy_block_offset(block);
+
+				if (start < vram_mgr->visible_size) {
+					u64 end = start + drm_buddy_block_size(mm, block);
+
+					nentry->used_visible_size +=
+						min(end, vram_mgr->visible_size) - start;
+				}
+			}
+		}
+		vram_mgr->visible_avail -= nentry->used_visible_size;
+		list_for_each_entry_safe(pos, n, &vram_mgr->queued_pages, queued_link) {
+			if (pos->id == nentry->id) {
+				--vram_mgr->n_queued_pages;
+				list_del(&pos->queued_link);
+				break;
+			}
+		}
+		list_add(&nentry->offlined_link, &vram_mgr->offlined_pages);
+		++vram_mgr->n_offlined_pages;
+		mutex_unlock(&vram_mgr->lock);
+		return ret;
+
+	} else {
+		ret = drm_buddy_alloc_blocks(mm, addr, addr + size,
+					     size, size, &nentry->blocks,
+					     DRM_BUDDY_RANGE_ALLOCATION);
+		if (ret) {
+			drm_warn(&xe->drm, "Could not reserve page at addr:%lx, ret:%d\n",
+				 addr, ret);
+			list_del(&nentry->offlined_link);
+			kfree(nentry);
+			mutex_unlock(&vram_mgr->lock);
+			return ret;
+		}
+		if ((addr + size) <= vram_mgr->visible_size) {
+			nentry->used_visible_size = size;
+		} else {
+			struct drm_buddy_block *block;
+
+			list_for_each_entry(block, &nentry->blocks, link) {
+				u64 start = drm_buddy_block_offset(block);
+
+				if (start < vram_mgr->visible_size) {
+					u64 end = start + drm_buddy_block_size(mm, block);
+
+					nentry->used_visible_size +=
+						min(end, vram_mgr->visible_size) - start;
+				}
+			}
+		}
+		vram_mgr->visible_avail -= nentry->used_visible_size;
+		nentry->id = ++vram_mgr->n_offlined_pages;
+		list_add(&nentry->offlined_link, &vram_mgr->offlined_pages);
+		mutex_unlock(&vram_mgr->lock);
+	}
+	/* Success */
+	return ret;
+}
+
+/**
+ * xe_ttm_tbo_handle_addr_fault - Handle vram physical address error flaged
+ * @tile: pointer to tile where address belongs
+ * @addr: physical faulty address
+ *
+ * Handle the physcial faulty address error on specific tile.
+ *
+ * Returns 0 for success, negative error code otherwise.
+ */
+int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr)
+{
+	struct xe_ttm_vram_mgr *vram_mgr = &tile->mem.vram->ttm;
+	struct drm_buddy mm = vram_mgr->mm;
+	struct xe_device *xe = tile_to_xe(tile);
+	int ret;
+
+	 /* Reserve page at address */
+	ret = xe_ttm_vram_reserve_page_at_addr(xe, addr, vram_mgr, &mm);
+	if (ret == -EIO)
+		return 0; /* success, wedged by kernel. */
+	return ret;
 }
 EXPORT_SYMBOL(xe_ttm_tbo_handle_addr_fault);
diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h b/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h
index a71e14818ec2..85511b51af75 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h
@@ -19,6 +19,14 @@ struct xe_ttm_vram_mgr {
 	struct ttm_resource_manager manager;
 	/** @mm: DRM buddy allocator which manages the VRAM */
 	struct drm_buddy mm;
+	/** @offlined_pages: List of offlined pages */
+	struct list_head offlined_pages;
+	/** @n_offlined_pages: Number of offlined pages */
+	u16 n_offlined_pages;
+	/** @queued_pages: List of queued pages */
+	struct list_head queued_pages;
+       /** @n_queued_pages: Number of queued pages */
+	u16 n_queued_pages;
 	/** @visible_size: Proped size of the CPU visible portion */
 	u64 visible_size;
 	/** @visible_avail: CPU visible portion still unallocated */
@@ -45,4 +53,17 @@ struct xe_ttm_vram_mgr_resource {
 	unsigned long flags;
 };
 
+struct xe_ttm_offline_resource {
+	/** @offlined_link: Link to offlined pages */
+	struct list_head offlined_link;
+	/** @queued_link: Link to queued pages */
+	struct list_head queued_link;
+	/** @blocks: list of DRM buddy blocks */
+	struct list_head blocks;
+	/** @used_visible_size: How many CPU visible bytes this resource is using */
+	u64 used_visible_size;
+	/** @id: The id of an offline resource */
+	u16 id;
+};
+
 #endif
-- 
2.52.0

[RFC PATCH 3/5] [DO_NOT_REVIEW]drm/xe/cri: Add debugfs to inject faulty vram address

[Tejas Upadhyay]

Add debugfs which can help testing feature with manual error injection.
Adding a debugfs interface to the drm/xe driver allows manual injection
of faulty VRAM addresses, facilitating the testing of the CRI memory
page offline feature before it is fully functional. The implementation
involves creating a debugfs entry, likely under
/sys/kernel/debug/dri/bdf/invalid_addr_vram0,
to accept specific faulty addresses for validation.

For example,
echo 0x1000 > /sys/kernel/debug/dri/bdf/invalid_addr_vram0
where 0x1000 is faulty adress being injected.

Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
---
 drivers/gpu/drm/xe/xe_debugfs.c            | 49 ++++++++++++++++++++++
 drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h |  2 +
 2 files changed, 51 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_debugfs.c b/drivers/gpu/drm/xe/xe_debugfs.c
index 844cfafe1ec7..d9dc1acebbce 100644
--- a/drivers/gpu/drm/xe/xe_debugfs.c
+++ b/drivers/gpu/drm/xe/xe_debugfs.c
@@ -27,6 +27,7 @@
 #include "xe_sriov_vf.h"
 #include "xe_step.h"
 #include "xe_tile_debugfs.h"
+#include "xe_ttm_vram_mgr.h"
 #include "xe_vsec.h"
 #include "xe_wa.h"
 
@@ -509,12 +510,48 @@ static const struct file_operations disable_late_binding_fops = {
 	.write = disable_late_binding_set,
 };
 
+static ssize_t addr_fault_reporting_show(struct file *f, char __user *ubuf,
+					 size_t size, loff_t *pos)
+{
+	struct xe_device *xe = file_inode(f)->i_private;
+	char buf[32];
+	int len;
+
+	len = scnprintf(buf, sizeof(buf), "%lld\n", xe->mem.vram->ttm.fault_addr);
+
+	return simple_read_from_buffer(ubuf, size, pos, buf, len);
+}
+
+static ssize_t addr_fault_reporting_set(struct file *f, const char __user *ubuf,
+					size_t size, loff_t *pos)
+{
+	struct xe_device *xe = file_inode(f)->i_private;
+	u64 addr;
+	int ret;
+
+	ret = kstrtou64_from_user(ubuf, size, 0, &addr);
+	if (ret)
+		return ret;
+
+	xe->mem.vram->ttm.fault_addr = addr;
+	xe_ttm_tbo_handle_addr_fault(xe_device_get_root_tile(xe), xe->mem.vram->ttm.fault_addr);
+
+	return size;
+}
+
+static const struct file_operations addr_fault_reporting_fops = {
+	.owner = THIS_MODULE,
+	.read = addr_fault_reporting_show,
+	.write = addr_fault_reporting_set,
+};
+
 void xe_debugfs_register(struct xe_device *xe)
 {
 	struct ttm_device *bdev = &xe->ttm;
 	struct drm_minor *minor = xe->drm.primary;
 	struct dentry *root = minor->debugfs_root;
 	struct ttm_resource_manager *man;
+	u8 mem_type = XE_PL_VRAM1;
 	struct xe_tile *tile;
 	struct xe_gt *gt;
 	u8 tile_id;
@@ -565,6 +602,18 @@ void xe_debugfs_register(struct xe_device *xe)
 	if (man)
 		ttm_resource_manager_create_debugfs(man, root, "stolen_mm");
 
+	do {
+		man = ttm_manager_type(bdev, mem_type);
+		if (man) {
+			char name[20];
+
+			snprintf(name, sizeof(name), "invalid_addr_vram%d", mem_type - XE_PL_VRAM0);
+			debugfs_create_file(name, 0600, root, xe,
+					    &addr_fault_reporting_fops);
+		}
+		--mem_type;
+	} while (mem_type >= XE_PL_VRAM0);
+
 	for_each_tile(tile, xe, tile_id)
 		xe_tile_debugfs_register(tile);
 
diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h b/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h
index 85511b51af75..c93573b9aab2 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr_types.h
@@ -37,6 +37,8 @@ struct xe_ttm_vram_mgr {
 	struct mutex lock;
 	/** @mem_type: The TTM memory type */
 	u32 mem_type;
+	/** @fault_addr: debugfs hook for setting faulty address */
+	u64 fault_addr;
 };
 
 /**
-- 
2.52.0

[RFC PATCH 4/5] drm/xe: Add routine to dump allocated VRAM blocks

[Tejas Upadhyay]

To implement the ability to see allocated blocks under a specific VRAM
instance in the drm/xe driver, new api is introduced. While existing structs
often show the free block list, this addition provides a comprehensive view
of all currently resident VRAM allocations.

Dump will look like,

[  +0.000003] xe 0000:03:00.0: [drm] 0x00000002f8000000-0x00000002f8800000: 8388608
[  +0.000005] xe 0000:03:00.0: [drm] 0x00000002f8800000-0x00000002f8840000: 262144
[  +0.000004] xe 0000:03:00.0: [drm] 0x00000002f8840000-0x00000002f8860000: 131072
[  +0.000004] xe 0000:03:00.0: [drm] 0x00000002f8860000-0x00000002f8870000: 65536
[  +0.000005] xe 0000:03:00.0: [drm] 0x00000002f9000000-0x00000002f9800000: 8388608
[  +0.000004] xe 0000:03:00.0: [drm] 0x00000002f9800000-0x00000002f9880000: 524288
[  +0.000005] xe 0000:03:00.0: [drm] 0x00000002f9880000-0x00000002f9884000: 16384
[  +0.000004] xe 0000:03:00.0: [drm] 0x00000002f9900000-0x00000002f9980000: 524288
[  +0.000005] xe 0000:03:00.0: [drm] 0x00000002f9980000-0x00000002f9988000: 32768
[  +0.000004] xe 0000:03:00.0: [drm] 0x00000002f9988000-0x00000002f998c000: 16384

Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
---
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.c | 36 ++++++++++++++++++++++++++++
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.h |  3 +++
 2 files changed, 39 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
index ab8243b17cf3..1ad98ef8b54e 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
@@ -505,6 +505,42 @@ static inline bool contains(u64 s1, u64 e1, u64 s2, u64 e2)
 	return s1 <= s2 && e1 <= e2;
 }
 
+void xe_ttm_vram_dump_allocated_blocks(struct drm_device *dev, struct drm_buddy *mm,
+				       struct drm_printer *p)
+{
+	struct drm_buddy_block *block;
+	LIST_HEAD(dfs);
+	int i;
+
+	for (i = 0; i < mm->n_roots; ++i)
+		list_add_tail(&mm->roots[i]->tmp_link, &dfs);
+
+	do {
+		u64 block_start;
+		u64 block_end;
+
+		block = list_first_entry_or_null(&dfs,
+						 struct drm_buddy_block,
+						 tmp_link);
+		if (!block)
+			break;
+
+		list_del(&block->tmp_link);
+
+		block_start = drm_buddy_block_offset(block);
+		block_end = block_start + drm_buddy_block_size(mm, block) - 1;
+
+		if (drm_buddy_block_is_allocated(block))
+			drm_buddy_block_print(mm, block, p);
+
+		if (drm_buddy_block_is_split(block)) {
+			list_add(&block->right->tmp_link, &dfs);
+			list_add(&block->left->tmp_link, &dfs);
+		}
+	} while (1);
+}
+EXPORT_SYMBOL(xe_ttm_vram_dump_allocated_blocks);
+
 static struct ttm_buffer_object *xe_ttm_vram_addr_to_tbo(struct drm_buddy *mm, u64 start)
 {
 	struct drm_buddy_block *block;
diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
index 1d6075411ebf..5872e8b48779 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
@@ -12,6 +12,7 @@ enum dma_data_direction;
 struct xe_device;
 struct xe_tile;
 struct xe_vram_region;
+struct drm_device;
 
 int __xe_ttm_vram_mgr_init(struct xe_device *xe, struct xe_ttm_vram_mgr *mgr,
 			   u32 mem_type, u64 size, u64 io_size,
@@ -31,6 +32,8 @@ u64 xe_ttm_vram_get_cpu_visible_size(struct ttm_resource_manager *man);
 void xe_ttm_vram_get_used(struct ttm_resource_manager *man,
 			  u64 *used, u64 *used_visible);
 int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr);
+void xe_ttm_vram_dump_allocated_blocks(struct drm_device *dev, struct drm_buddy *mm,
+				       struct drm_printer *p);
 static inline struct xe_ttm_vram_mgr_resource *
 to_xe_ttm_vram_mgr_resource(struct ttm_resource *res)
 {
-- 
2.52.0

[RFC PATCH 5/5] [DO NOT REVIEW]drm/xe/cri: Add sysfs interface for bad gpu vram pages

[Tejas Upadhyay]

Starting CRI, Include a sysfs interface designed to expose information
about bad VRAM pages—those identified as having hardware faults
(e.g., ECC errors). This interface allows userspace tools and
administrators to monitor the health of the GPU's local memory and
track the status of page retirement.To get details on bad gpu vram
pages can be found under /sys/bus/pci/devices/bdf/vram_bad_pages.

Where The format is, pfn : gpu page size : flags

flags:
R: reserved, this gpu page is reserved.
P: pending for reserve, this gpu page is marked as bad, will be reserved in next window of page_reserve.
F: unable to reserve. this gpu page can’t be reserved due to some reasons.

For example if you read using cat /sys/bus/pci/devices/bdf/vram_bad_pages,
0x00000000 : 0x00001000 : R
0x00001234 : 0x00001000 : P

Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
---
 drivers/gpu/drm/xe/xe_device_sysfs.c |  2 +
 drivers/gpu/drm/xe/xe_tile_sysfs.c   |  1 +
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.c | 78 ++++++++++++++++++++++++++++
 drivers/gpu/drm/xe/xe_ttm_vram_mgr.h |  1 +
 4 files changed, 82 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_device_sysfs.c b/drivers/gpu/drm/xe/xe_device_sysfs.c
index a73e0e957cb0..e6a017601428 100644
--- a/drivers/gpu/drm/xe/xe_device_sysfs.c
+++ b/drivers/gpu/drm/xe/xe_device_sysfs.c
@@ -14,6 +14,7 @@
 #include "xe_pcode_api.h"
 #include "xe_pcode.h"
 #include "xe_pm.h"
+#include "xe_ttm_vram_mgr.h"
 
 /**
  * DOC: Xe device sysfs
@@ -284,6 +285,7 @@ int xe_device_sysfs_init(struct xe_device *xe)
 		if (ret)
 			return ret;
 	}
+	xe_ttm_vram_sysfs_init(xe);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/xe/xe_tile_sysfs.c b/drivers/gpu/drm/xe/xe_tile_sysfs.c
index 9e1236a9ec67..beaf20ecddb4 100644
--- a/drivers/gpu/drm/xe/xe_tile_sysfs.c
+++ b/drivers/gpu/drm/xe/xe_tile_sysfs.c
@@ -10,6 +10,7 @@
 #include "xe_pm.h"
 #include "xe_tile.h"
 #include "xe_tile_sysfs.h"
+#include "xe_ttm_vram_mgr.h"
 #include "xe_vram_freq.h"
 
 static void xe_tile_sysfs_kobj_release(struct kobject *kobj)
diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
index 1ad98ef8b54e..f06416b92962 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
@@ -296,6 +296,37 @@ static void xe_ttm_vram_free_bad_pages(struct drm_device *dev, struct xe_ttm_vra
 	mutex_unlock(&mgr->lock);
 }
 
+static void xe_ttm_vram_dump_bad_pages_info(char *buf, struct xe_ttm_vram_mgr *mgr)
+{
+	const unsigned int element_size = sizeof("0xabcdabcd : 0x12345678 : R\n") - 1;
+	struct xe_ttm_offline_resource *pos, *n;
+	struct drm_buddy_block *block;
+	ssize_t s = 0;
+
+	mutex_lock(&mgr->lock);
+	list_for_each_entry_safe(pos, n, &mgr->offlined_pages, offlined_link) {
+		block = list_first_entry(&pos->blocks,
+					 struct drm_buddy_block,
+					 link);
+		s += scnprintf(&buf[s], element_size + 1,
+			       "0x%08llx : 0x%08llx : %1s\n",
+			       drm_buddy_block_offset(block) >> PAGE_SHIFT,
+			       drm_buddy_block_size(&mgr->mm, block),
+			       "R");
+	}
+	list_for_each_entry_safe(pos, n, &mgr->queued_pages, queued_link) {
+		block = list_first_entry(&pos->blocks,
+					 struct drm_buddy_block,
+					 link);
+		s += scnprintf(&buf[s], element_size + 1,
+			       "0x%08llx : 0x%08llx : %1s\n",
+			       drm_buddy_block_offset(block) >> PAGE_SHIFT,
+			       drm_buddy_block_size(&mgr->mm, block),
+			       "P");
+	}
+	mutex_unlock(&mgr->lock);
+}
+
 static void xe_ttm_vram_mgr_fini(struct drm_device *dev, void *arg)
 {
 	struct xe_device *xe = to_xe_device(dev);
@@ -755,3 +786,50 @@ int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr)
 	return ret;
 }
 EXPORT_SYMBOL(xe_ttm_tbo_handle_addr_fault);
+
+static ssize_t vram_bad_pages_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct xe_device *xe = pdev_to_xe_device(pdev);
+	struct ttm_resource_manager *man;
+	u8 mem_type = XE_PL_VRAM1;
+
+	do {
+		man = ttm_manager_type(&xe->ttm, mem_type);
+		struct xe_ttm_vram_mgr *mgr = to_xe_ttm_vram_mgr(man);
+
+		if (man)
+			xe_ttm_vram_dump_bad_pages_info(buf, mgr);
+		--mem_type;
+	} while (mem_type >= XE_PL_VRAM0);
+
+	return sysfs_emit(buf, "%s\n", buf);
+}
+static DEVICE_ATTR_RO(vram_bad_pages);
+
+static void xe_ttm_vram_sysfs_fini(void *arg)
+{
+	struct xe_device *xe = arg;
+
+	device_remove_file(xe->drm.dev, &dev_attr_vram_bad_pages);
+}
+
+/**
+ * xe_ttm_vram_sysfs_init - Initialize vram sysfs component
+ * @tile: Xe Tile object
+ *
+ * It needs to be initialized after the main tile component is ready
+ *
+ * Returns: 0 on success, negative error code on error.
+ */
+int xe_ttm_vram_sysfs_init(struct xe_device *xe)
+{
+	int err;
+
+	err = device_create_file(xe->drm.dev, &dev_attr_vram_bad_pages);
+	if (err)
+		return 0;
+
+	return devm_add_action_or_reset(xe->drm.dev, xe_ttm_vram_sysfs_fini, xe);
+}
+EXPORT_SYMBOL(xe_ttm_vram_sysfs_init);
diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
index 5872e8b48779..6e69140c0be8 100644
--- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
+++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.h
@@ -34,6 +34,7 @@ void xe_ttm_vram_get_used(struct ttm_resource_manager *man,
 int xe_ttm_tbo_handle_addr_fault(struct xe_tile *tile, unsigned long addr);
 void xe_ttm_vram_dump_allocated_blocks(struct drm_device *dev, struct drm_buddy *mm,
 				       struct drm_printer *p);
+int xe_ttm_vram_sysfs_init(struct xe_device *xe);
 static inline struct xe_ttm_vram_mgr_resource *
 to_xe_ttm_vram_mgr_resource(struct ttm_resource *res)
 {
-- 
2.52.0