1 files changed, 425 insertions, 196 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
index 2fd77c36a1ff..9d934c07fa6b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
@@ -24,16 +24,20 @@
 
 #include <linux/dma-mapping.h>
 #include <drm/ttm/ttm_range_manager.h>
+#include <drm/drm_drv.h>
 
 #include "amdgpu.h"
 #include "amdgpu_vm.h"
 #include "amdgpu_res_cursor.h"
-#include "amdgpu_atomfirmware.h"
 #include "atom.h"
 
+#define AMDGPU_MAX_SG_SEGMENT_SIZE	(2UL << 30)
+
 struct amdgpu_vram_reservation {
-	struct list_head node;
-	struct drm_mm_node mm_node;
+	u64 start;
+	u64 size;
+	struct list_head allocated;
+	struct list_head blocks;
 };
 
 static inline struct amdgpu_vram_mgr *
@@ -48,6 +52,44 @@ to_amdgpu_device(struct amdgpu_vram_mgr *mgr)
 	return container_of(mgr, struct amdgpu_device, mman.vram_mgr);
 }
 
+static inline struct drm_buddy_block *
+amdgpu_vram_mgr_first_block(struct list_head *list)
+{
+	return list_first_entry_or_null(list, struct drm_buddy_block, link);
+}
+
+static inline bool amdgpu_is_vram_mgr_blocks_contiguous(struct list_head *head)
+{
+	struct drm_buddy_block *block;
+	u64 start, size;
+
+	block = amdgpu_vram_mgr_first_block(head);
+	if (!block)
+		return false;
+
+	while (head != block->link.next) {
+		start = amdgpu_vram_mgr_block_start(block);
+		size = amdgpu_vram_mgr_block_size(block);
+
+		block = list_entry(block->link.next, struct drm_buddy_block, link);
+		if (start + size != amdgpu_vram_mgr_block_start(block))
+			return false;
+	}
+
+	return true;
+}
+
+static inline u64 amdgpu_vram_mgr_blocks_size(struct list_head *head)
+{
+	struct drm_buddy_block *block;
+	u64 size = 0;
+
+	list_for_each_entry(block, head, link)
+		size += amdgpu_vram_mgr_block_size(block);
+
+	return size;
+}
+
 /**
  * DOC: mem_info_vram_total
  *
@@ -96,10 +138,9 @@ static ssize_t amdgpu_mem_info_vram_used_show(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
-	struct ttm_resource_manager *man;
+	struct ttm_resource_manager *man = &adev->mman.vram_mgr.manager;
 
-	man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
-	return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_usage(man));
+	return sysfs_emit(buf, "%llu\n", ttm_resource_manager_usage(man));
 }
 
 /**
@@ -116,10 +157,9 @@ static ssize_t amdgpu_mem_info_vis_vram_used_show(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
-	struct ttm_resource_manager *man;
 
-	man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
-	return sysfs_emit(buf, "%llu\n", amdgpu_vram_mgr_vis_usage(man));
+	return sysfs_emit(buf, "%llu\n",
+			  amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr));
 }
 
 /**
@@ -183,23 +223,41 @@ static struct attribute *amdgpu_vram_mgr_attributes[] = {
 	NULL
 };
 
+static umode_t amdgpu_vram_attrs_is_visible(struct kobject *kobj,
+					    struct attribute *attr, int i)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (attr == &dev_attr_mem_info_vram_vendor.attr &&
+	    !adev->gmc.vram_vendor)
+		return 0;
+
+	if (!ttm_resource_manager_used(&adev->mman.vram_mgr.manager))
+		return 0;
+
+	return attr->mode;
+}
+
 const struct attribute_group amdgpu_vram_mgr_attr_group = {
-	.attrs = amdgpu_vram_mgr_attributes
+	.attrs = amdgpu_vram_mgr_attributes,
+	.is_visible = amdgpu_vram_attrs_is_visible
 };
 
 /**
- * amdgpu_vram_mgr_vis_size - Calculate visible node size
+ * amdgpu_vram_mgr_vis_size - Calculate visible block size
  *
  * @adev: amdgpu_device pointer
- * @node: MM node structure
+ * @block: DRM BUDDY block structure
  *
- * Calculate how many bytes of the MM node are inside visible VRAM
+ * Calculate how many bytes of the DRM BUDDY block are inside visible VRAM
  */
 static u64 amdgpu_vram_mgr_vis_size(struct amdgpu_device *adev,
-				    struct drm_mm_node *node)
+				    struct drm_buddy_block *block)
 {
-	uint64_t start = node->start << PAGE_SHIFT;
-	uint64_t end = (node->size + node->start) << PAGE_SHIFT;
+	u64 start = amdgpu_vram_mgr_block_start(block);
+	u64 end = start + amdgpu_vram_mgr_block_size(block);
 
 	if (start >= adev->gmc.visible_vram_size)
 		return 0;
@@ -220,9 +278,9 @@ u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct ttm_resource *res = bo->tbo.resource;
-	unsigned pages = res->num_pages;
-	struct drm_mm_node *mm;
-	u64 usage;
+	struct amdgpu_vram_mgr_resource *vres = to_amdgpu_vram_mgr_resource(res);
+	struct drm_buddy_block *block;
+	u64 usage = 0;
 
 	if (amdgpu_gmc_vram_full_visible(&adev->gmc))
 		return amdgpu_bo_size(bo);
@@ -230,9 +288,8 @@ u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo)
 	if (res->start >= adev->gmc.visible_vram_size >> PAGE_SHIFT)
 		return 0;
 
-	mm = &container_of(res, struct ttm_range_mgr_node, base)->mm_nodes[0];
-	for (usage = 0; pages; pages -= mm->size, mm++)
-		usage += amdgpu_vram_mgr_vis_size(adev, mm);
+	list_for_each_entry(block, &vres->blocks, link)
+		usage += amdgpu_vram_mgr_vis_size(adev, block);
 
 	return usage;
 }
@@ -242,51 +299,61 @@ static void amdgpu_vram_mgr_do_reserve(struct ttm_resource_manager *man)
 {
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_device *adev = to_amdgpu_device(mgr);
-	struct drm_mm *mm = &mgr->mm;
+	struct drm_buddy *mm = &mgr->mm;
 	struct amdgpu_vram_reservation *rsv, *temp;
+	struct drm_buddy_block *block;
 	uint64_t vis_usage;
 
-	list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node) {
-		if (drm_mm_reserve_node(mm, &rsv->mm_node))
+	list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, blocks) {
+		if (drm_buddy_alloc_blocks(mm, rsv->start, rsv->start + rsv->size,
+					   rsv->size, mm->chunk_size, &rsv->allocated,
+					   DRM_BUDDY_RANGE_ALLOCATION))
+			continue;
+
+		block = amdgpu_vram_mgr_first_block(&rsv->allocated);
+		if (!block)
 			continue;
 
 		dev_dbg(adev->dev, "Reservation 0x%llx - %lld, Succeeded\n",
-			rsv->mm_node.start, rsv->mm_node.size);
+			rsv->start, rsv->size);
 
-		vis_usage = amdgpu_vram_mgr_vis_size(adev, &rsv->mm_node);
+		vis_usage = amdgpu_vram_mgr_vis_size(adev, block);
 		atomic64_add(vis_usage, &mgr->vis_usage);
-		atomic64_add(rsv->mm_node.size << PAGE_SHIFT, &mgr->usage);
-		list_move(&rsv->node, &mgr->reserved_pages);
+		spin_lock(&man->bdev->lru_lock);
+		man->usage += rsv->size;
+		spin_unlock(&man->bdev->lru_lock);
+		list_move(&rsv->blocks, &mgr->reserved_pages);
 	}
 }
 
 /**
  * amdgpu_vram_mgr_reserve_range - Reserve a range from VRAM
  *
- * @man: TTM memory type manager
+ * @mgr: amdgpu_vram_mgr pointer
  * @start: start address of the range in VRAM
  * @size: size of the range
  *
- * Reserve memory from start addess with the specified size in VRAM
+ * Reserve memory from start address with the specified size in VRAM
  */
-int amdgpu_vram_mgr_reserve_range(struct ttm_resource_manager *man,
+int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr,
 				  uint64_t start, uint64_t size)
 {
-	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_vram_reservation *rsv;
 
 	rsv = kzalloc(sizeof(*rsv), GFP_KERNEL);
 	if (!rsv)
 		return -ENOMEM;
 
-	INIT_LIST_HEAD(&rsv->node);
-	rsv->mm_node.start = start >> PAGE_SHIFT;
-	rsv->mm_node.size = size >> PAGE_SHIFT;
+	INIT_LIST_HEAD(&rsv->allocated);
+	INIT_LIST_HEAD(&rsv->blocks);
 
-	spin_lock(&mgr->lock);
-	list_add_tail(&mgr->reservations_pending, &rsv->node);
-	amdgpu_vram_mgr_do_reserve(man);
-	spin_unlock(&mgr->lock);
+	rsv->start = start;
+	rsv->size = size;
+
+	mutex_lock(&mgr->lock);
+	list_add_tail(&rsv->blocks, &mgr->reservations_pending);
+	amdgpu_vram_mgr_do_reserve(&mgr->manager);
+	mutex_unlock(&mgr->lock);
 
 	return 0;
 }
@@ -294,7 +361,7 @@ int amdgpu_vram_mgr_reserve_range(struct ttm_resource_manager *man,
 /**
  * amdgpu_vram_mgr_query_page_status - query the reservation status
  *
- * @man: TTM memory type manager
+ * @mgr: amdgpu_vram_mgr pointer
  * @start: start address of a page in VRAM
  *
  * Returns:
@@ -302,26 +369,25 @@ int amdgpu_vram_mgr_reserve_range(struct ttm_resource_manager *man,
  *	0: the page has been reserved
  *	-ENOENT: the input page is not a reservation
  */
-int amdgpu_vram_mgr_query_page_status(struct ttm_resource_manager *man,
+int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr,
 				      uint64_t start)
 {
-	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_vram_reservation *rsv;
 	int ret;
 
-	spin_lock(&mgr->lock);
+	mutex_lock(&mgr->lock);
 
-	list_for_each_entry(rsv, &mgr->reservations_pending, node) {
-		if ((rsv->mm_node.start <= start) &&
-		    (start < (rsv->mm_node.start + rsv->mm_node.size))) {
+	list_for_each_entry(rsv, &mgr->reservations_pending, blocks) {
+		if (rsv->start <= start &&
+		    (start < (rsv->start + rsv->size))) {
 			ret = -EBUSY;
 			goto out;
 		}
 	}
 
-	list_for_each_entry(rsv, &mgr->reserved_pages, node) {
-		if ((rsv->mm_node.start <= start) &&
-		    (start < (rsv->mm_node.start + rsv->mm_node.size))) {
+	list_for_each_entry(rsv, &mgr->reserved_pages, blocks) {
+		if (rsv->start <= start &&
+		    (start < (rsv->start + rsv->size))) {
 			ret = 0;
 			goto out;
 		}
@@ -329,30 +395,37 @@ int amdgpu_vram_mgr_query_page_status(struct ttm_resource_manager *man,
 
 	ret = -ENOENT;
 out:
-	spin_unlock(&mgr->lock);
+	mutex_unlock(&mgr->lock);
 	return ret;
 }
 
-/**
- * amdgpu_vram_mgr_virt_start - update virtual start address
- *
- * @mem: ttm_resource to update
- * @node: just allocated node
- *
- * Calculate a virtual BO start address to easily check if everything is CPU
- * accessible.
- */
-static void amdgpu_vram_mgr_virt_start(struct ttm_resource *mem,
-				       struct drm_mm_node *node)
+int amdgpu_vram_mgr_query_address_block_info(struct amdgpu_vram_mgr *mgr,
+			uint64_t address, struct amdgpu_vram_block_info *info)
 {
-	unsigned long start;
+	struct amdgpu_vram_mgr_resource *vres;
+	struct drm_buddy_block *block;
+	u64 start, size;
+	int ret = -ENOENT;
+
+	mutex_lock(&mgr->lock);
+	list_for_each_entry(vres, &mgr->allocated_vres_list, vres_node) {
+		list_for_each_entry(block, &vres->blocks, link) {
+			start = amdgpu_vram_mgr_block_start(block);
+			size = amdgpu_vram_mgr_block_size(block);
+			if ((start <= address) && (address < (start + size))) {
+				info->start = start;
+				info->size = size;
+				memcpy(&info->task, &vres->task, sizeof(vres->task));
+				ret = 0;
+				goto out;
+			}
+		}
+	}
 
-	start = node->start + node->size;
-	if (start > mem->num_pages)
-		start -= mem->num_pages;
-	else
-		start = 0;
-	mem->start = max(mem->start, start);
+out:
+	mutex_unlock(&mgr->lock);
+
+	return ret;
 }
 
 /**
@@ -361,7 +434,7 @@ static void amdgpu_vram_mgr_virt_start(struct ttm_resource *mem,
  * @man: TTM memory type manager
  * @tbo: TTM BO we need this range for
  * @place: placement flags and restrictions
- * @mem: the resulting mem object
+ * @res: the resulting mem object
  *
  * Allocate VRAM for the given BO.
  */
@@ -370,116 +443,183 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 			       const struct ttm_place *place,
 			       struct ttm_resource **res)
 {
-	unsigned long lpfn, num_nodes, pages_per_node, pages_left, pages;
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_device *adev = to_amdgpu_device(mgr);
-	uint64_t vis_usage = 0, mem_bytes, max_bytes;
-	struct ttm_range_mgr_node *node;
-	struct drm_mm *mm = &mgr->mm;
-	enum drm_mm_insert_mode mode;
-	unsigned i;
+	struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo);
+	u64 vis_usage = 0, max_bytes, min_block_size;
+	struct amdgpu_vram_mgr_resource *vres;
+	u64 size, remaining_size, lpfn, fpfn;
+	unsigned int adjust_dcc_size = 0;
+	struct drm_buddy *mm = &mgr->mm;
+	struct drm_buddy_block *block;
+	unsigned long pages_per_block;
 	int r;
 
-	lpfn = place->lpfn;
-	if (!lpfn)
+	lpfn = (u64)place->lpfn << PAGE_SHIFT;
+	if (!lpfn || lpfn > man->size)
 		lpfn = man->size;
 
+	fpfn = (u64)place->fpfn << PAGE_SHIFT;
+
 	max_bytes = adev->gmc.mc_vram_size;
 	if (tbo->type != ttm_bo_type_kernel)
 		max_bytes -= AMDGPU_VM_RESERVED_VRAM;
 
-	/* bail out quickly if there's likely not enough VRAM for this BO */
-	mem_bytes = tbo->base.size;
-	if (atomic64_add_return(mem_bytes, &mgr->usage) > max_bytes) {
-		r = -ENOSPC;
-		goto error_sub;
-	}
-
-	if (place->flags & TTM_PL_FLAG_CONTIGUOUS) {
-		pages_per_node = ~0ul;
-		num_nodes = 1;
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) {
+		pages_per_block = ~0ul;
 	} else {
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
-		pages_per_node = HPAGE_PMD_NR;
+		pages_per_block = HPAGE_PMD_NR;
 #else
 		/* default to 2MB */
-		pages_per_node = 2UL << (20UL - PAGE_SHIFT);
+		pages_per_block = 2UL << (20UL - PAGE_SHIFT);
 #endif
-		pages_per_node = max_t(uint32_t, pages_per_node,
-				       tbo->page_alignment);
-		num_nodes = DIV_ROUND_UP_ULL(PFN_UP(mem_bytes), pages_per_node);
+		pages_per_block = max_t(u32, pages_per_block,
+					tbo->page_alignment);
 	}
 
-	node = kvmalloc(struct_size(node, mm_nodes, num_nodes),
-			GFP_KERNEL | __GFP_ZERO);
-	if (!node) {
-		r = -ENOMEM;
-		goto error_sub;
+	vres = kzalloc(sizeof(*vres), GFP_KERNEL);
+	if (!vres)
+		return -ENOMEM;
+
+	ttm_resource_init(tbo, place, &vres->base);
+
+	/* bail out quickly if there's likely not enough VRAM for this BO */
+	if (ttm_resource_manager_usage(man) > max_bytes) {
+		r = -ENOSPC;
+		goto error_fini;
 	}
 
-	ttm_resource_init(tbo, place, &node->base);
+	INIT_LIST_HEAD(&vres->blocks);
 
-	mode = DRM_MM_INSERT_BEST;
 	if (place->flags & TTM_PL_FLAG_TOPDOWN)
-		mode = DRM_MM_INSERT_HIGH;
-
-	pages_left = node->base.num_pages;
-
-	/* Limit maximum size to 2GB due to SG table limitations */
-	pages = min(pages_left, 2UL << (30 - PAGE_SHIFT));
-
-	i = 0;
-	spin_lock(&mgr->lock);
-	while (pages_left) {
-		uint32_t alignment = tbo->page_alignment;
-
-		if (pages >= pages_per_node)
-			alignment = pages_per_node;
-
-		r = drm_mm_insert_node_in_range(mm, &node->mm_nodes[i], pages,
-						alignment, 0, place->fpfn,
-						lpfn, mode);
-		if (unlikely(r)) {
-			if (pages > pages_per_node) {
-				if (is_power_of_2(pages))
-					pages = pages / 2;
-				else
-					pages = rounddown_pow_of_two(pages);
-				continue;
-			}
-			goto error_free;
+		vres->flags |= DRM_BUDDY_TOPDOWN_ALLOCATION;
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
+		vres->flags |= DRM_BUDDY_CONTIGUOUS_ALLOCATION;
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CLEARED)
+		vres->flags |= DRM_BUDDY_CLEAR_ALLOCATION;
+
+	if (fpfn || lpfn != mgr->mm.size)
+		/* Allocate blocks in desired range */
+		vres->flags |= DRM_BUDDY_RANGE_ALLOCATION;
+
+	if (bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC &&
+	    adev->gmc.gmc_funcs->get_dcc_alignment)
+		adjust_dcc_size = amdgpu_gmc_get_dcc_alignment(adev);
+
+	remaining_size = (u64)vres->base.size;
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		unsigned int dcc_size;
+
+		dcc_size = roundup_pow_of_two(vres->base.size + adjust_dcc_size);
+		remaining_size = (u64)dcc_size;
+
+		vres->flags |= DRM_BUDDY_TRIM_DISABLE;
+	}
+
+	mutex_lock(&mgr->lock);
+	while (remaining_size) {
+		if (tbo->page_alignment)
+			min_block_size = (u64)tbo->page_alignment << PAGE_SHIFT;
+		else
+			min_block_size = mgr->default_page_size;
+
+		size = remaining_size;
+
+		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size)
+			min_block_size = size;
+		else if ((size >= (u64)pages_per_block << PAGE_SHIFT) &&
+			 !(size & (((u64)pages_per_block << PAGE_SHIFT) - 1)))
+			min_block_size = (u64)pages_per_block << PAGE_SHIFT;
+
+		BUG_ON(min_block_size < mm->chunk_size);
+
+		r = drm_buddy_alloc_blocks(mm, fpfn,
+					   lpfn,
+					   size,
+					   min_block_size,
+					   &vres->blocks,
+					   vres->flags);
+
+		if (unlikely(r == -ENOSPC) && pages_per_block == ~0ul &&
+		    !(place->flags & TTM_PL_FLAG_CONTIGUOUS)) {
+			vres->flags &= ~DRM_BUDDY_CONTIGUOUS_ALLOCATION;
+			pages_per_block = max_t(u32, 2UL << (20UL - PAGE_SHIFT),
+						tbo->page_alignment);
+
+			continue;
 		}
 
-		vis_usage += amdgpu_vram_mgr_vis_size(adev, &node->mm_nodes[i]);
-		amdgpu_vram_mgr_virt_start(&node->base, &node->mm_nodes[i]);
-		pages_left -= pages;
-		++i;
+		if (unlikely(r))
+			goto error_free_blocks;
+
+		if (size > remaining_size)
+			remaining_size = 0;
+		else
+			remaining_size -= size;
+	}
+
+	vres->task.pid = task_pid_nr(current);
+	get_task_comm(vres->task.comm, current);
+	list_add_tail(&vres->vres_node, &mgr->allocated_vres_list);
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		struct drm_buddy_block *dcc_block;
+		unsigned long dcc_start;
+		u64 trim_start;
+
+		dcc_block = amdgpu_vram_mgr_first_block(&vres->blocks);
+		/* Adjust the start address for DCC buffers only */
+		dcc_start =
+			roundup((unsigned long)amdgpu_vram_mgr_block_start(dcc_block),
+				adjust_dcc_size);
+		trim_start = (u64)dcc_start;
+		drm_buddy_block_trim(mm, &trim_start,
+				     (u64)vres->base.size,
+				     &vres->blocks);
+	}
+	mutex_unlock(&mgr->lock);
+
+	vres->base.start = 0;
+	size = max_t(u64, amdgpu_vram_mgr_blocks_size(&vres->blocks),
+		     vres->base.size);
+	list_for_each_entry(block, &vres->blocks, link) {
+		unsigned long start;
 
-		if (pages > pages_left)
-			pages = pages_left;
+		start = amdgpu_vram_mgr_block_start(block) +
+			amdgpu_vram_mgr_block_size(block);
+		start >>= PAGE_SHIFT;
+
+		if (start > PFN_UP(size))
+			start -= PFN_UP(size);
+		else
+			start = 0;
+		vres->base.start = max(vres->base.start, start);
+
+		vis_usage += amdgpu_vram_mgr_vis_size(adev, block);
 	}
-	spin_unlock(&mgr->lock);
 
-	if (i == 1)
-		node->base.placement |= TTM_PL_FLAG_CONTIGUOUS;
+	if (amdgpu_is_vram_mgr_blocks_contiguous(&vres->blocks))
+		vres->base.placement |= TTM_PL_FLAG_CONTIGUOUS;
 
 	if (adev->gmc.xgmi.connected_to_cpu)
-		node->base.bus.caching = ttm_cached;
+		vres->base.bus.caching = ttm_cached;
 	else
-		node->base.bus.caching = ttm_write_combined;
+		vres->base.bus.caching = ttm_write_combined;
 
 	atomic64_add(vis_usage, &mgr->vis_usage);
-	*res = &node->base;
+	*res = &vres->base;
 	return 0;
 
-error_free:
-	while (i--)
-		drm_mm_remove_node(&node->mm_nodes[i]);
-	spin_unlock(&mgr->lock);
-	kvfree(node);
+error_free_blocks:
+	drm_buddy_free_list(mm, &vres->blocks, 0);
+	mutex_unlock(&mgr->lock);
+error_fini:
+	ttm_resource_fini(man, &vres->base);
+	kfree(vres);
 
-error_sub:
-	atomic64_sub(mem_bytes, &mgr->usage);
 	return r;
 }
 
@@ -487,42 +627,43 @@ error_sub:
  * amdgpu_vram_mgr_del - free ranges
  *
  * @man: TTM memory type manager
- * @mem: TTM memory object
+ * @res: TTM memory object
  *
  * Free the allocated VRAM again.
  */
 static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man,
 				struct ttm_resource *res)
 {
-	struct ttm_range_mgr_node *node = to_ttm_range_mgr_node(res);
+	struct amdgpu_vram_mgr_resource *vres = to_amdgpu_vram_mgr_resource(res);
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_device *adev = to_amdgpu_device(mgr);
-	uint64_t usage = 0, vis_usage = 0;
-	unsigned i, pages;
+	struct drm_buddy *mm = &mgr->mm;
+	struct drm_buddy_block *block;
+	uint64_t vis_usage = 0;
 
-	spin_lock(&mgr->lock);
-	for (i = 0, pages = res->num_pages; pages;
-	     pages -= node->mm_nodes[i].size, ++i) {
-		struct drm_mm_node *mm = &node->mm_nodes[i];
+	mutex_lock(&mgr->lock);
 
-		drm_mm_remove_node(mm);
-		usage += mm->size << PAGE_SHIFT;
-		vis_usage += amdgpu_vram_mgr_vis_size(adev, mm);
-	}
+	list_del(&vres->vres_node);
+	memset(&vres->task, 0, sizeof(vres->task));
+
+	list_for_each_entry(block, &vres->blocks, link)
+		vis_usage += amdgpu_vram_mgr_vis_size(adev, block);
+
+	drm_buddy_free_list(mm, &vres->blocks, vres->flags);
 	amdgpu_vram_mgr_do_reserve(man);
-	spin_unlock(&mgr->lock);
+	mutex_unlock(&mgr->lock);
 
-	atomic64_sub(usage, &mgr->usage);
 	atomic64_sub(vis_usage, &mgr->vis_usage);
 
-	kvfree(node);
+	ttm_resource_fini(man, res);
+	kfree(vres);
 }
 
 /**
  * amdgpu_vram_mgr_alloc_sgt - allocate and fill a sg table
  *
  * @adev: amdgpu device pointer
- * @mem: TTM memory object
+ * @res: TTM memory object
  * @offset: byte offset from the base of VRAM BO
  * @length: number of bytes to export in sg_table
  * @dev: the other device
@@ -547,11 +688,11 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 	if (!*sgt)
 		return -ENOMEM;
 
-	/* Determine the number of DRM_MM nodes to export */
+	/* Determine the number of DRM_BUDDY blocks to export */
 	amdgpu_res_first(res, offset, length, &cursor);
 	while (cursor.remaining) {
 		num_entries++;
-		amdgpu_res_next(&cursor, cursor.size);
+		amdgpu_res_next(&cursor, min(cursor.size, AMDGPU_MAX_SG_SEGMENT_SIZE));
 	}
 
 	r = sg_alloc_table(*sgt, num_entries, GFP_KERNEL);
@@ -563,15 +704,15 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 		sg->length = 0;
 
 	/*
-	 * Walk down DRM_MM nodes to populate scatterlist nodes
-	 * @note: Use iterator api to get first the DRM_MM node
+	 * Walk down DRM_BUDDY blocks to populate scatterlist nodes
+	 * @note: Use iterator api to get first the DRM_BUDDY block
 	 * and the number of bytes from it. Access the following
-	 * DRM_MM node(s) if more buffer needs to exported
+	 * DRM_BUDDY block(s) if more buffer needs to exported
 	 */
 	amdgpu_res_first(res, offset, length, &cursor);
 	for_each_sgtable_sg((*sgt), sg, i) {
 		phys_addr_t phys = cursor.start + adev->gmc.aper_base;
-		size_t size = cursor.size;
+		unsigned long size = min(cursor.size, AMDGPU_MAX_SG_SEGMENT_SIZE);
 		dma_addr_t addr;
 
 		addr = dma_map_resource(dev, phys, size, dir,
@@ -584,7 +725,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 		sg_dma_address(sg) = addr;
 		sg_dma_len(sg) = size;
 
-		amdgpu_res_next(&cursor, cursor.size);
+		amdgpu_res_next(&cursor, size);
 	}
 
 	return 0;
@@ -630,31 +771,98 @@ void amdgpu_vram_mgr_free_sgt(struct device *dev,
 }
 
 /**
- * amdgpu_vram_mgr_usage - how many bytes are used in this domain
+ * amdgpu_vram_mgr_vis_usage - how many bytes are used in the visible part
+ *
+ * @mgr: amdgpu_vram_mgr pointer
+ *
+ * Returns how many bytes are used in the visible part of VRAM
+ */
+uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr)
+{
+	return atomic64_read(&mgr->vis_usage);
+}
+
+/**
+ * amdgpu_vram_mgr_clear_reset_blocks - reset clear blocks
+ *
+ * @adev: amdgpu device pointer
+ *
+ * Reset the cleared drm buddy blocks.
+ */
+void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev)
+{
+	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
+	struct drm_buddy *mm = &mgr->mm;
+
+	mutex_lock(&mgr->lock);
+	drm_buddy_reset_clear(mm, false);
+	mutex_unlock(&mgr->lock);
+}
+
+/**
+ * amdgpu_vram_mgr_intersects - test each drm buddy block for intersection
  *
  * @man: TTM memory type manager
+ * @res: The resource to test
+ * @place: The place to test against
+ * @size: Size of the new allocation
  *
- * Returns how many bytes are used in this domain.
+ * Test each drm buddy block for intersection for eviction decision.
  */
-uint64_t amdgpu_vram_mgr_usage(struct ttm_resource_manager *man)
+static bool amdgpu_vram_mgr_intersects(struct ttm_resource_manager *man,
+				       struct ttm_resource *res,
+				       const struct ttm_place *place,
+				       size_t size)
 {
-	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
+	struct amdgpu_vram_mgr_resource *mgr = to_amdgpu_vram_mgr_resource(res);
+	struct drm_buddy_block *block;
+
+	/* Check each drm buddy block individually */
+	list_for_each_entry(block, &mgr->blocks, link) {
+		unsigned long fpfn =
+			amdgpu_vram_mgr_block_start(block) >> PAGE_SHIFT;
+		unsigned long lpfn = fpfn +
+			(amdgpu_vram_mgr_block_size(block) >> PAGE_SHIFT);
+
+		if (place->fpfn < lpfn &&
+		    (!place->lpfn || place->lpfn > fpfn))
+			return true;
+	}
 
-	return atomic64_read(&mgr->usage);
+	return false;
 }
 
 /**
- * amdgpu_vram_mgr_vis_usage - how many bytes are used in the visible part
+ * amdgpu_vram_mgr_compatible - test each drm buddy block for compatibility
  *
  * @man: TTM memory type manager
+ * @res: The resource to test
+ * @place: The place to test against
+ * @size: Size of the new allocation
  *
- * Returns how many bytes are used in the visible part of VRAM
+ * Test each drm buddy block for placement compatibility.
  */
-uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_resource_manager *man)
+static bool amdgpu_vram_mgr_compatible(struct ttm_resource_manager *man,
+				       struct ttm_resource *res,
+				       const struct ttm_place *place,
+				       size_t size)
 {
-	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
+	struct amdgpu_vram_mgr_resource *mgr = to_amdgpu_vram_mgr_resource(res);
+	struct drm_buddy_block *block;
+
+	/* Check each drm buddy block individually */
+	list_for_each_entry(block, &mgr->blocks, link) {
+		unsigned long fpfn =
+			amdgpu_vram_mgr_block_start(block) >> PAGE_SHIFT;
+		unsigned long lpfn = fpfn +
+			(amdgpu_vram_mgr_block_size(block) >> PAGE_SHIFT);
+
+		if (fpfn < place->fpfn ||
+		    (place->lpfn && lpfn > place->lpfn))
+			return false;
+	}
 
-	return atomic64_read(&mgr->vis_usage);
+	return true;
 }
 
 /**
@@ -669,19 +877,30 @@ static void amdgpu_vram_mgr_debug(struct ttm_resource_manager *man,
 				  struct drm_printer *printer)
 {
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
+	struct drm_buddy *mm = &mgr->mm;
+	struct amdgpu_vram_reservation *rsv;
+
+	drm_printf(printer, "  vis usage:%llu\n",
+		   amdgpu_vram_mgr_vis_usage(mgr));
 
-	spin_lock(&mgr->lock);
-	drm_mm_print(&mgr->mm, printer);
-	spin_unlock(&mgr->lock);
+	mutex_lock(&mgr->lock);
+	drm_printf(printer, "default_page_size: %lluKiB\n",
+		   mgr->default_page_size >> 10);
 
-	drm_printf(printer, "man size:%llu pages, ram usage:%lluMB, vis usage:%lluMB\n",
-		   man->size, amdgpu_vram_mgr_usage(man) >> 20,
-		   amdgpu_vram_mgr_vis_usage(man) >> 20);
+	drm_buddy_print(mm, printer);
+
+	drm_printf(printer, "reserved:\n");
+	list_for_each_entry(rsv, &mgr->reserved_pages, blocks)
+		drm_printf(printer, "%#018llx-%#018llx: %llu\n",
+			rsv->start, rsv->start + rsv->size, rsv->size);
+	mutex_unlock(&mgr->lock);
 }
 
 static const struct ttm_resource_manager_func amdgpu_vram_mgr_func = {
 	.alloc	= amdgpu_vram_mgr_new,
 	.free	= amdgpu_vram_mgr_del,
+	.intersects = amdgpu_vram_mgr_intersects,
+	.compatible = amdgpu_vram_mgr_compatible,
 	.debug	= amdgpu_vram_mgr_debug
 };
 
@@ -696,15 +915,24 @@ int amdgpu_vram_mgr_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
 	struct ttm_resource_manager *man = &mgr->manager;
+	int err;
 
-	ttm_resource_manager_init(man, adev->gmc.real_vram_size >> PAGE_SHIFT);
+	man->cg = drmm_cgroup_register_region(adev_to_drm(adev), "vram", adev->gmc.real_vram_size);
+	if (IS_ERR(man->cg))
+		return PTR_ERR(man->cg);
+	ttm_resource_manager_init(man, &adev->mman.bdev,
+				  adev->gmc.real_vram_size);
 
-	man->func = &amdgpu_vram_mgr_func;
-
-	drm_mm_init(&mgr->mm, 0, man->size);
-	spin_lock_init(&mgr->lock);
+	mutex_init(&mgr->lock);
 	INIT_LIST_HEAD(&mgr->reservations_pending);
 	INIT_LIST_HEAD(&mgr->reserved_pages);
+	INIT_LIST_HEAD(&mgr->allocated_vres_list);
+	mgr->default_page_size = PAGE_SIZE;
+
+	man->func = &amdgpu_vram_mgr_func;
+	err = drm_buddy_init(&mgr->mm, man->size, PAGE_SIZE);
+	if (err)
+		return err;
 
 	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager);
 	ttm_resource_manager_set_used(man, true);
@@ -732,16 +960,17 @@ void amdgpu_vram_mgr_fini(struct amdgpu_device *adev)
 	if (ret)
 		return;
 
-	spin_lock(&mgr->lock);
-	list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, node)
+	mutex_lock(&mgr->lock);
+	list_for_each_entry_safe(rsv, temp, &mgr->reservations_pending, blocks)
 		kfree(rsv);
 
-	list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, node) {
-		drm_mm_remove_node(&rsv->mm_node);
+	list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, blocks) {
+		drm_buddy_free_list(&mgr->mm, &rsv->allocated, 0);
 		kfree(rsv);
 	}
-	drm_mm_takedown(&mgr->mm);
-	spin_unlock(&mgr->lock);
+	if (!adev->gmc.is_app_apu)
+		drm_buddy_fini(&mgr->mm);
+	mutex_unlock(&mgr->lock);
 
 	ttm_resource_manager_cleanup(man);
 	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, NULL);