Re: [PATCH v2] drm/xe: Convert stolen memory over to ttm_range_manager

["Govindapillai, Vinod" <vinod.govindapillai@intel.com>]

On Thu, 2026-04-16 at 20:49 +0530, Sanjay Yadav wrote:
> Stolen memory requires physically contiguous allocations for display
> scanout and compressed framebuffers. The stolen memory manager was
> sharing the gpu_buddy allocator backend with the VRAM manager, but
> buddy manages non-contiguous power-of-two blocks making it a poor
> fit.
> Stolen memory also has fundamentally different allocation patterns:
> 
> - Allocation sizes are not power-of-two. Since buddy rounds up to the
>   next power-of-two block size, a ~17MB request can fail even with
>   ~22MB free, because the free space is fragmented across non-fitting
>   power-of-two blocks.
> - Hardware restrictions prevent using the first 4K page of stolen for
>   certain allocations (e.g., FBC). The display code sets fpfn=1 to
>   enforce this, but when fpfn != 0, gpu_buddy enables
>   GPU_BUDDY_RANGE_ALLOCATION mode which disables the try_harder
>   coalescing path, further reducing allocation success.
> 
> This combination caused FBC compressed framebuffer (CFB) allocation
> failures on platforms like NVL/PTL. In case of NVL where stolen
> memory
> is ~56MB and the initial plane framebuffer consumes ~34MB at probe
> time,
> leaving ~22MB for subsequent allocations.
> 
> Use ttm_range_man_init_nocheck() to set up a drm_mm-backed TTM
> resource
> manager for stolen memory. This reuses the TTM core's
> ttm_range_manager
> callbacks, avoiding duplicate implementations.
> 
> Tested on NVL with a 4K DP display: stolen_mm shows a single ~22MB
> contiguous free hole after initial plane framebuffer allocation, and
> FBC successfully allocates its CFB from that region. The
> corresponding
> IGT was previously skipped and now passes.
> 
> v2:
> - Clarify that stolen memory requires contiguous allocations (Matt B)
> - Properly handle xe_ttm_resource_visible() for stolen instead of
>   unconditionally returning true (Matt A)
> 
> Closes:
> https://gitlab.freedesktop.org/drm/xe/kernel/-/work_items/7631
> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Cc: Matthew Brost <matthew.brost@intel.com>
> Suggested-by: Matthew Auld <matthew.auld@intel.com>
> Assisted-by: GitHub Copilot:claude-sonnet-4.6
> Signed-off-by: Sanjay Yadav <sanjay.kumar.yadav@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_bo.c             | 16 +++++--
>  drivers/gpu/drm/xe/xe_device_types.h   |  3 ++
>  drivers/gpu/drm/xe/xe_res_cursor.h     | 14 +++++-
>  drivers/gpu/drm/xe/xe_ttm_stolen_mgr.c | 64 +++++++++---------------
> --
>  drivers/gpu/drm/xe/xe_ttm_stolen_mgr.h |  9 ++++
>  drivers/gpu/drm/xe/xe_ttm_vram_mgr.c   | 11 ++---
>  6 files changed, 63 insertions(+), 54 deletions(-)
> 

FP16 stolen memory allocations are working ~35mb

Acked-by: Vinod Govindapillai <vinod.govindapillai@intel.com>

> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index a7c2dc7f224c..b737edf04d60 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -599,11 +599,17 @@ static void xe_ttm_tt_destroy(struct ttm_device
> *ttm_dev, struct ttm_tt *tt)
>  	kfree(tt);
>  }
>  
> -static bool xe_ttm_resource_visible(struct ttm_resource *mem)
> +static bool xe_ttm_resource_visible(struct xe_device *xe, struct
> ttm_resource *mem)
>  {
> -	struct xe_ttm_vram_mgr_resource *vres =
> -		to_xe_ttm_vram_mgr_resource(mem);
> +	struct xe_ttm_vram_mgr_resource *vres;
>  
> +	if (mem->mem_type == XE_PL_STOLEN) {
> +		struct xe_ttm_stolen_mgr *mgr = xe->mem.stolen_mgr;
> +
> +		return mgr->io_base &&
> !xe_ttm_stolen_cpu_access_needs_ggtt(xe);
> +	}
> +
> +	vres = to_xe_ttm_vram_mgr_resource(mem);
>  	return vres->used_visible_size == mem->size;
>  }
>  
> @@ -621,7 +627,7 @@ bool xe_bo_is_visible_vram(struct xe_bo *bo)
>  	if (drm_WARN_ON(bo->ttm.base.dev, !xe_bo_is_vram(bo)))
>  		return false;
>  
> -	return xe_ttm_resource_visible(bo->ttm.resource);
> +	return xe_ttm_resource_visible(xe_bo_device(bo), bo-
> >ttm.resource);
>  }
>  
>  static int xe_ttm_io_mem_reserve(struct ttm_device *bdev,
> @@ -637,7 +643,7 @@ static int xe_ttm_io_mem_reserve(struct
> ttm_device *bdev,
>  	case XE_PL_VRAM1: {
>  		struct xe_vram_region *vram =
> res_to_mem_region(mem);
>  
> -		if (!xe_ttm_resource_visible(mem))
> +		if (!xe_ttm_resource_visible(xe, mem))
>  			return -EINVAL;
>  
>  		mem->bus.offset = mem->start << PAGE_SHIFT;
> diff --git a/drivers/gpu/drm/xe/xe_device_types.h
> b/drivers/gpu/drm/xe/xe_device_types.h
> index 150c76b2acaf..ffb84659c413 100644
> --- a/drivers/gpu/drm/xe/xe_device_types.h
> +++ b/drivers/gpu/drm/xe/xe_device_types.h
> @@ -42,6 +42,7 @@ struct xe_ggtt;
>  struct xe_i2c;
>  struct xe_pat_ops;
>  struct xe_pxp;
> +struct xe_ttm_stolen_mgr;
>  struct xe_vram_region;
>  
>  /**
> @@ -278,6 +279,8 @@ struct xe_device {
>  		struct ttm_resource_manager sys_mgr;
>  		/** @mem.sys_mgr: system memory shrinker. */
>  		struct xe_shrinker *shrinker;
> +		/** @mem.stolen_mgr: stolen memory manager. */
> +		struct xe_ttm_stolen_mgr *stolen_mgr;
>  	} mem;
>  
>  	/** @sriov: device level virtualization data */
> diff --git a/drivers/gpu/drm/xe/xe_res_cursor.h
> b/drivers/gpu/drm/xe/xe_res_cursor.h
> index 5f4ab08c0686..0522caafd89d 100644
> --- a/drivers/gpu/drm/xe/xe_res_cursor.h
> +++ b/drivers/gpu/drm/xe/xe_res_cursor.h
> @@ -101,7 +101,15 @@ static inline void xe_res_first(struct
> ttm_resource *res,
>  	cur->mem_type = res->mem_type;
>  
>  	switch (cur->mem_type) {
> -	case XE_PL_STOLEN:
> +	case XE_PL_STOLEN: {
> +		/* res->start is in pages (ttm_range_manager). */
> +		cur->start = (res->start << PAGE_SHIFT) + start;
> +		cur->size = size;
> +		cur->remaining = size;
> +		cur->node = NULL;
> +		cur->mm = NULL;
> +		break;
> +	}
>  	case XE_PL_VRAM0:
>  	case XE_PL_VRAM1: {
>  		struct gpu_buddy_block *block;
> @@ -289,6 +297,10 @@ static inline void xe_res_next(struct
> xe_res_cursor *cur, u64 size)
>  
>  	switch (cur->mem_type) {
>  	case XE_PL_STOLEN:
> +		/* Just advance within the contiguous region. */
> +		cur->start += size;
> +		cur->size = cur->remaining;
> +		break;
>  	case XE_PL_VRAM0:
>  	case XE_PL_VRAM1:
>  		start = size - cur->size;
> diff --git a/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.c
> b/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.c
> index 27c9d72222cf..5e9070739e65 100644
> --- a/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.c
> +++ b/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.c
> @@ -19,30 +19,11 @@
>  #include "xe_device.h"
>  #include "xe_gt_printk.h"
>  #include "xe_mmio.h"
> -#include "xe_res_cursor.h"
>  #include "xe_sriov.h"
>  #include "xe_ttm_stolen_mgr.h"
> -#include "xe_ttm_vram_mgr.h"
>  #include "xe_vram.h"
>  #include "xe_wa.h"
>  
> -struct xe_ttm_stolen_mgr {
> -	struct xe_ttm_vram_mgr base;
> -
> -	/* PCI base offset */
> -	resource_size_t io_base;
> -	/* GPU base offset */
> -	resource_size_t stolen_base;
> -
> -	void __iomem *mapping;
> -};
> -
> -static inline struct xe_ttm_stolen_mgr *
> -to_stolen_mgr(struct ttm_resource_manager *man)
> -{
> -	return container_of(man, struct xe_ttm_stolen_mgr,
> base.manager);
> -}
> -
>  /**
>   * xe_ttm_stolen_cpu_access_needs_ggtt() - If we can't directly CPU
> access
>   * stolen, can we then fallback to mapping through the GGTT.
> @@ -210,12 +191,19 @@ static u64 detect_stolen(struct xe_device *xe,
> struct xe_ttm_stolen_mgr *mgr)
>  #endif
>  }
>  
> +static void xe_ttm_stolen_mgr_fini(struct drm_device *dev, void
> *arg)
> +{
> +	struct xe_device *xe = to_xe_device(dev);
> +
> +	ttm_range_man_fini_nocheck(&xe->ttm, XE_PL_STOLEN);
> +}
> +
>  int xe_ttm_stolen_mgr_init(struct xe_device *xe)
>  {
>  	struct pci_dev *pdev = to_pci_dev(xe->drm.dev);
>  	struct xe_ttm_stolen_mgr *mgr;
>  	u64 stolen_size, io_size;
> -	int err;
> +	int ret;
>  
>  	mgr = drmm_kzalloc(&xe->drm, sizeof(*mgr), GFP_KERNEL);
>  	if (!mgr)
> @@ -244,12 +232,12 @@ int xe_ttm_stolen_mgr_init(struct xe_device
> *xe)
>  	if (mgr->io_base &&
> !xe_ttm_stolen_cpu_access_needs_ggtt(xe))
>  		io_size = stolen_size;
>  
> -	err = __xe_ttm_vram_mgr_init(xe, &mgr->base, XE_PL_STOLEN,
> stolen_size,
> -				     io_size, PAGE_SIZE);
> -	if (err) {
> -		drm_dbg_kms(&xe->drm, "Stolen mgr init failed:
> %i\n", err);
> -		return err;
> -	}
> +	ret = ttm_range_man_init_nocheck(&xe->ttm, XE_PL_STOLEN,
> false,
> +					 stolen_size >> PAGE_SHIFT);
> +	if (ret)
> +		return ret;
> +
> +	xe->mem.stolen_mgr = mgr;
>  
>  	drm_dbg_kms(&xe->drm, "Initialized stolen memory support
> with %llu bytes\n",
>  		    stolen_size);
> @@ -257,36 +245,32 @@ int xe_ttm_stolen_mgr_init(struct xe_device
> *xe)
>  	if (io_size)
>  		mgr->mapping = devm_ioremap_wc(&pdev->dev, mgr-
> >io_base, io_size);
>  
> -	return 0;
> +	return drmm_add_action_or_reset(&xe->drm,
> xe_ttm_stolen_mgr_fini, mgr);
>  }
>  
>  u64 xe_ttm_stolen_io_offset(struct xe_bo *bo, u32 offset)
>  {
>  	struct xe_device *xe = xe_bo_device(bo);
> -	struct ttm_resource_manager *ttm_mgr = ttm_manager_type(&xe-
> >ttm, XE_PL_STOLEN);
> -	struct xe_ttm_stolen_mgr *mgr = to_stolen_mgr(ttm_mgr);
> -	struct xe_res_cursor cur;
> +	struct xe_ttm_stolen_mgr *mgr = xe->mem.stolen_mgr;
>  
>  	XE_WARN_ON(!mgr->io_base);
>  
>  	if (xe_ttm_stolen_cpu_access_needs_ggtt(xe))
>  		return mgr->io_base + xe_bo_ggtt_addr(bo) + offset;
>  
> -	xe_res_first(bo->ttm.resource, offset, 4096, &cur);
> -	return mgr->io_base + cur.start;
> +	/* Range allocator: res->start is in pages. */
> +	return mgr->io_base + (bo->ttm.resource->start <<
> PAGE_SHIFT) + offset;
>  }
>  
>  static int __xe_ttm_stolen_io_mem_reserve_bar2(struct xe_device *xe,
>  					       struct
> xe_ttm_stolen_mgr *mgr,
>  					       struct ttm_resource
> *mem)
>  {
> -	struct xe_res_cursor cur;
> -
>  	if (!mgr->io_base)
>  		return -EIO;
>  
> -	xe_res_first(mem, 0, 4096, &cur);
> -	mem->bus.offset = cur.start;
> +	/* Range allocator always produces contiguous allocations.
> */
> +	mem->bus.offset = mem->start << PAGE_SHIFT;
>  
>  	drm_WARN_ON(&xe->drm, !(mem->placement &
> TTM_PL_FLAG_CONTIGUOUS));
>  
> @@ -329,8 +313,7 @@ static int
> __xe_ttm_stolen_io_mem_reserve_stolen(struct xe_device *xe,
>  
>  int xe_ttm_stolen_io_mem_reserve(struct xe_device *xe, struct
> ttm_resource *mem)
>  {
> -	struct ttm_resource_manager *ttm_mgr = ttm_manager_type(&xe-
> >ttm, XE_PL_STOLEN);
> -	struct xe_ttm_stolen_mgr *mgr = ttm_mgr ?
> to_stolen_mgr(ttm_mgr) : NULL;
> +	struct xe_ttm_stolen_mgr *mgr = xe->mem.stolen_mgr;
>  
>  	if (!mgr || !mgr->io_base)
>  		return -EIO;
> @@ -343,8 +326,5 @@ int xe_ttm_stolen_io_mem_reserve(struct xe_device
> *xe, struct ttm_resource *mem)
>  
>  u64 xe_ttm_stolen_gpu_offset(struct xe_device *xe)
>  {
> -	struct xe_ttm_stolen_mgr *mgr =
> -		to_stolen_mgr(ttm_manager_type(&xe->ttm,
> XE_PL_STOLEN));
> -
> -	return mgr->stolen_base;
> +	return xe->mem.stolen_mgr->stolen_base;
>  }
> diff --git a/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.h
> b/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.h
> index 8e877d1e839b..049e91e77326 100644
> --- a/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.h
> +++ b/drivers/gpu/drm/xe/xe_ttm_stolen_mgr.h
> @@ -12,6 +12,15 @@ struct ttm_resource;
>  struct xe_bo;
>  struct xe_device;
>  
> +struct xe_ttm_stolen_mgr {
> +	/* PCI base offset */
> +	resource_size_t io_base;
> +	/* GPU base offset */
> +	resource_size_t stolen_base;
> +
> +	void __iomem *mapping;
> +};
> +
>  int xe_ttm_stolen_mgr_init(struct xe_device *xe);
>  int xe_ttm_stolen_io_mem_reserve(struct xe_device *xe, struct
> ttm_resource *mem);
>  bool xe_ttm_stolen_cpu_access_needs_ggtt(struct xe_device *xe);
> diff --git a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> index 5fd0d5506a7e..79ef8e1b5e5c 100644
> --- a/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> +++ b/drivers/gpu/drm/xe/xe_ttm_vram_mgr.c
> @@ -301,14 +301,13 @@ int __xe_ttm_vram_mgr_init(struct xe_device
> *xe, struct xe_ttm_vram_mgr *mgr,
>  			   u64 default_page_size)
>  {
>  	struct ttm_resource_manager *man = &mgr->manager;
> +	const char *name;
>  	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);
> -	}
> +	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;