1 files changed, 156 insertions, 86 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
index 83e344fbb50a..e55508b39496 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c
@@ -28,6 +28,21 @@
  *    Christian König <christian.koenig@amd.com>
  */
 
+/**
+ * DOC: MMU Notifier
+ *
+ * For coherent userptr handling registers an MMU notifier to inform the driver
+ * about updates on the page tables of a process.
+ *
+ * When somebody tries to invalidate the page tables we block the update until
+ * all operations on the pages in question are completed, then those pages are
+ * marked as accessed and also dirty if it wasn't a read only access.
+ *
+ * New command submissions using the userptrs in question are delayed until all
+ * page table invalidation are completed and we once more see a coherent process
+ * address space.
+ */
+
 #include <linux/firmware.h>
 #include <linux/module.h>
 #include <linux/mmu_notifier.h>
@@ -38,6 +53,22 @@
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
 
+/**
+ * struct amdgpu_mn
+ *
+ * @adev: amdgpu device pointer
+ * @mm: process address space
+ * @mn: MMU notifier structure
+ * @type: type of MMU notifier
+ * @work: destruction work item
+ * @node: hash table node to find structure by adev and mn
+ * @lock: rw semaphore protecting the notifier nodes
+ * @objects: interval tree containing amdgpu_mn_nodes
+ * @read_lock: mutex for recursive locking of @lock
+ * @recursion: depth of recursion
+ *
+ * Data for each amdgpu device and process address space.
+ */
 struct amdgpu_mn {
 	/* constant after initialisation */
 	struct amdgpu_device	*adev;
@@ -58,13 +89,21 @@ struct amdgpu_mn {
 	atomic_t		recursion;
 };
 
+/**
+ * struct amdgpu_mn_node
+ *
+ * @it: interval node defining start-last of the affected address range
+ * @bos: list of all BOs in the affected address range
+ *
+ * Manages all BOs which are affected of a certain range of address space.
+ */
 struct amdgpu_mn_node {
 	struct interval_tree_node	it;
 	struct list_head		bos;
 };
 
 /**
- * amdgpu_mn_destroy - destroy the rmn
+ * amdgpu_mn_destroy - destroy the MMU notifier
  *
  * @work: previously sheduled work item
  *
@@ -72,47 +111,50 @@ struct amdgpu_mn_node {
  */
 static void amdgpu_mn_destroy(struct work_struct *work)
 {
-	struct amdgpu_mn *rmn = container_of(work, struct amdgpu_mn, work);
-	struct amdgpu_device *adev = rmn->adev;
+	struct amdgpu_mn *amn = container_of(work, struct amdgpu_mn, work);
+	struct amdgpu_device *adev = amn->adev;
 	struct amdgpu_mn_node *node, *next_node;
 	struct amdgpu_bo *bo, *next_bo;
 
 	mutex_lock(&adev->mn_lock);
-	down_write(&rmn->lock);
-	hash_del(&rmn->node);
+	down_write(&amn->lock);
+	hash_del(&amn->node);
 	rbtree_postorder_for_each_entry_safe(node, next_node,
-					     &rmn->objects.rb_root, it.rb) {
+					     &amn->objects.rb_root, it.rb) {
 		list_for_each_entry_safe(bo, next_bo, &node->bos, mn_list) {
 			bo->mn = NULL;
 			list_del_init(&bo->mn_list);
 		}
 		kfree(node);
 	}
-	up_write(&rmn->lock);
+	up_write(&amn->lock);
 	mutex_unlock(&adev->mn_lock);
-	mmu_notifier_unregister_no_release(&rmn->mn, rmn->mm);
-	kfree(rmn);
+	mmu_notifier_unregister_no_release(&amn->mn, amn->mm);
+	kfree(amn);
 }
 
 /**
  * amdgpu_mn_release - callback to notify about mm destruction
  *
  * @mn: our notifier
- * @mn: the mm this callback is about
+ * @mm: the mm this callback is about
  *
  * Shedule a work item to lazy destroy our notifier.
  */
 static void amdgpu_mn_release(struct mmu_notifier *mn,
 			      struct mm_struct *mm)
 {
-	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
-	INIT_WORK(&rmn->work, amdgpu_mn_destroy);
-	schedule_work(&rmn->work);
+	struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
+
+	INIT_WORK(&amn->work, amdgpu_mn_destroy);
+	schedule_work(&amn->work);
 }
 
 
 /**
- * amdgpu_mn_lock - take the write side lock for this mn
+ * amdgpu_mn_lock - take the write side lock for this notifier
+ *
+ * @mn: our notifier
  */
 void amdgpu_mn_lock(struct amdgpu_mn *mn)
 {
@@ -121,7 +163,9 @@ void amdgpu_mn_lock(struct amdgpu_mn *mn)
 }
 
 /**
- * amdgpu_mn_unlock - drop the write side lock for this mn
+ * amdgpu_mn_unlock - drop the write side lock for this notifier
+ *
+ * @mn: our notifier
  */
 void amdgpu_mn_unlock(struct amdgpu_mn *mn)
 {
@@ -130,40 +174,44 @@ void amdgpu_mn_unlock(struct amdgpu_mn *mn)
 }
 
 /**
- * amdgpu_mn_read_lock - take the rmn read lock
+ * amdgpu_mn_read_lock - take the read side lock for this notifier
  *
- * @rmn: our notifier
- *
- * Take the rmn read side lock.
+ * @amn: our notifier
  */
-static void amdgpu_mn_read_lock(struct amdgpu_mn *rmn)
+static int amdgpu_mn_read_lock(struct amdgpu_mn *amn, bool blockable)
 {
-	mutex_lock(&rmn->read_lock);
-	if (atomic_inc_return(&rmn->recursion) == 1)
-		down_read_non_owner(&rmn->lock);
-	mutex_unlock(&rmn->read_lock);
+	if (blockable)
+		mutex_lock(&amn->read_lock);
+	else if (!mutex_trylock(&amn->read_lock))
+		return -EAGAIN;
+
+	if (atomic_inc_return(&amn->recursion) == 1)
+		down_read_non_owner(&amn->lock);
+	mutex_unlock(&amn->read_lock);
+
+	return 0;
 }
 
 /**
- * amdgpu_mn_read_unlock - drop the rmn read lock
+ * amdgpu_mn_read_unlock - drop the read side lock for this notifier
  *
- * @rmn: our notifier
- *
- * Drop the rmn read side lock.
+ * @amn: our notifier
  */
-static void amdgpu_mn_read_unlock(struct amdgpu_mn *rmn)
+static void amdgpu_mn_read_unlock(struct amdgpu_mn *amn)
 {
-	if (atomic_dec_return(&rmn->recursion) == 0)
-		up_read_non_owner(&rmn->lock);
+	if (atomic_dec_return(&amn->recursion) == 0)
+		up_read_non_owner(&amn->lock);
 }
 
 /**
  * amdgpu_mn_invalidate_node - unmap all BOs of a node
  *
  * @node: the node with the BOs to unmap
+ * @start: start of address range affected
+ * @end: end of address range affected
  *
- * We block for all BOs and unmap them by move them
- * into system domain again.
+ * Block for operations on BOs to finish and mark pages as accessed and
+ * potentially dirty.
  */
 static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node,
 				      unsigned long start,
@@ -190,42 +238,54 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node,
  * amdgpu_mn_invalidate_range_start_gfx - callback to notify about mm change
  *
  * @mn: our notifier
- * @mn: the mm this callback is about
+ * @mm: the mm this callback is about
  * @start: start of updated range
  * @end: end of updated range
  *
- * We block for all BOs between start and end to be idle and
- * unmap them by move them into system domain again.
+ * Block for operations on BOs to finish and mark pages as accessed and
+ * potentially dirty.
  */
-static void amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn,
+static int amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn,
 						 struct mm_struct *mm,
 						 unsigned long start,
-						 unsigned long end)
+						 unsigned long end,
+						 bool blockable)
 {
-	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
+	struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
 	struct interval_tree_node *it;
 
 	/* notification is exclusive, but interval is inclusive */
 	end -= 1;
 
-	amdgpu_mn_read_lock(rmn);
+	/* TODO we should be able to split locking for interval tree and
+	 * amdgpu_mn_invalidate_node
+	 */
+	if (amdgpu_mn_read_lock(amn, blockable))
+		return -EAGAIN;
 
-	it = interval_tree_iter_first(&rmn->objects, start, end);
+	it = interval_tree_iter_first(&amn->objects, start, end);
 	while (it) {
 		struct amdgpu_mn_node *node;
 
+		if (!blockable) {
+			amdgpu_mn_read_unlock(amn);
+			return -EAGAIN;
+		}
+
 		node = container_of(it, struct amdgpu_mn_node, it);
 		it = interval_tree_iter_next(it, start, end);
 
 		amdgpu_mn_invalidate_node(node, start, end);
 	}
+
+	return 0;
 }
 
 /**
  * amdgpu_mn_invalidate_range_start_hsa - callback to notify about mm change
  *
  * @mn: our notifier
- * @mn: the mm this callback is about
+ * @mm: the mm this callback is about
  * @start: start of updated range
  * @end: end of updated range
  *
@@ -233,24 +293,31 @@ static void amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn,
  * necessitates evicting all user-mode queues of the process. The BOs
  * are restorted in amdgpu_mn_invalidate_range_end_hsa.
  */
-static void amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn,
+static int amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn,
 						 struct mm_struct *mm,
 						 unsigned long start,
-						 unsigned long end)
+						 unsigned long end,
+						 bool blockable)
 {
-	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
+	struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
 	struct interval_tree_node *it;
 
 	/* notification is exclusive, but interval is inclusive */
 	end -= 1;
 
-	amdgpu_mn_read_lock(rmn);
+	if (amdgpu_mn_read_lock(amn, blockable))
+		return -EAGAIN;
 
-	it = interval_tree_iter_first(&rmn->objects, start, end);
+	it = interval_tree_iter_first(&amn->objects, start, end);
 	while (it) {
 		struct amdgpu_mn_node *node;
 		struct amdgpu_bo *bo;
 
+		if (!blockable) {
+			amdgpu_mn_read_unlock(amn);
+			return -EAGAIN;
+		}
+
 		node = container_of(it, struct amdgpu_mn_node, it);
 		it = interval_tree_iter_next(it, start, end);
 
@@ -262,13 +329,15 @@ static void amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn,
 				amdgpu_amdkfd_evict_userptr(mem, mm);
 		}
 	}
+
+	return 0;
 }
 
 /**
  * amdgpu_mn_invalidate_range_end - callback to notify about mm change
  *
  * @mn: our notifier
- * @mn: the mm this callback is about
+ * @mm: the mm this callback is about
  * @start: start of updated range
  * @end: end of updated range
  *
@@ -279,9 +348,9 @@ static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn,
 					   unsigned long start,
 					   unsigned long end)
 {
-	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
+	struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn);
 
-	amdgpu_mn_read_unlock(rmn);
+	amdgpu_mn_read_unlock(amn);
 }
 
 static const struct mmu_notifier_ops amdgpu_mn_ops[] = {
@@ -315,7 +384,7 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
 				enum amdgpu_mn_type type)
 {
 	struct mm_struct *mm = current->mm;
-	struct amdgpu_mn *rmn;
+	struct amdgpu_mn *amn;
 	unsigned long key = AMDGPU_MN_KEY(mm, type);
 	int r;
 
@@ -325,41 +394,41 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
 		return ERR_PTR(-EINTR);
 	}
 
-	hash_for_each_possible(adev->mn_hash, rmn, node, key)
-		if (AMDGPU_MN_KEY(rmn->mm, rmn->type) == key)
+	hash_for_each_possible(adev->mn_hash, amn, node, key)
+		if (AMDGPU_MN_KEY(amn->mm, amn->type) == key)
 			goto release_locks;
 
-	rmn = kzalloc(sizeof(*rmn), GFP_KERNEL);
-	if (!rmn) {
-		rmn = ERR_PTR(-ENOMEM);
+	amn = kzalloc(sizeof(*amn), GFP_KERNEL);
+	if (!amn) {
+		amn = ERR_PTR(-ENOMEM);
 		goto release_locks;
 	}
 
-	rmn->adev = adev;
-	rmn->mm = mm;
-	init_rwsem(&rmn->lock);
-	rmn->type = type;
-	rmn->mn.ops = &amdgpu_mn_ops[type];
-	rmn->objects = RB_ROOT_CACHED;
-	mutex_init(&rmn->read_lock);
-	atomic_set(&rmn->recursion, 0);
+	amn->adev = adev;
+	amn->mm = mm;
+	init_rwsem(&amn->lock);
+	amn->type = type;
+	amn->mn.ops = &amdgpu_mn_ops[type];
+	amn->objects = RB_ROOT_CACHED;
+	mutex_init(&amn->read_lock);
+	atomic_set(&amn->recursion, 0);
 
-	r = __mmu_notifier_register(&rmn->mn, mm);
+	r = __mmu_notifier_register(&amn->mn, mm);
 	if (r)
-		goto free_rmn;
+		goto free_amn;
 
-	hash_add(adev->mn_hash, &rmn->node, AMDGPU_MN_KEY(mm, type));
+	hash_add(adev->mn_hash, &amn->node, AMDGPU_MN_KEY(mm, type));
 
 release_locks:
 	up_write(&mm->mmap_sem);
 	mutex_unlock(&adev->mn_lock);
 
-	return rmn;
+	return amn;
 
-free_rmn:
+free_amn:
 	up_write(&mm->mmap_sem);
 	mutex_unlock(&adev->mn_lock);
-	kfree(rmn);
+	kfree(amn);
 
 	return ERR_PTR(r);
 }
@@ -379,14 +448,14 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	enum amdgpu_mn_type type =
 		bo->kfd_bo ? AMDGPU_MN_TYPE_HSA : AMDGPU_MN_TYPE_GFX;
-	struct amdgpu_mn *rmn;
+	struct amdgpu_mn *amn;
 	struct amdgpu_mn_node *node = NULL, *new_node;
 	struct list_head bos;
 	struct interval_tree_node *it;
 
-	rmn = amdgpu_mn_get(adev, type);
-	if (IS_ERR(rmn))
-		return PTR_ERR(rmn);
+	amn = amdgpu_mn_get(adev, type);
+	if (IS_ERR(amn))
+		return PTR_ERR(amn);
 
 	new_node = kmalloc(sizeof(*new_node), GFP_KERNEL);
 	if (!new_node)
@@ -394,12 +463,12 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 
 	INIT_LIST_HEAD(&bos);
 
-	down_write(&rmn->lock);
+	down_write(&amn->lock);
 
-	while ((it = interval_tree_iter_first(&rmn->objects, addr, end))) {
+	while ((it = interval_tree_iter_first(&amn->objects, addr, end))) {
 		kfree(node);
 		node = container_of(it, struct amdgpu_mn_node, it);
-		interval_tree_remove(&node->it, &rmn->objects);
+		interval_tree_remove(&node->it, &amn->objects);
 		addr = min(it->start, addr);
 		end = max(it->last, end);
 		list_splice(&node->bos, &bos);
@@ -410,7 +479,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 	else
 		kfree(new_node);
 
-	bo->mn = rmn;
+	bo->mn = amn;
 
 	node->it.start = addr;
 	node->it.last = end;
@@ -418,9 +487,9 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 	list_splice(&bos, &node->bos);
 	list_add(&bo->mn_list, &node->bos);
 
-	interval_tree_insert(&node->it, &rmn->objects);
+	interval_tree_insert(&node->it, &amn->objects);
 
-	up_write(&rmn->lock);
+	up_write(&amn->lock);
 
 	return 0;
 }
@@ -435,18 +504,18 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 void amdgpu_mn_unregister(struct amdgpu_bo *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	struct amdgpu_mn *rmn;
+	struct amdgpu_mn *amn;
 	struct list_head *head;
 
 	mutex_lock(&adev->mn_lock);
 
-	rmn = bo->mn;
-	if (rmn == NULL) {
+	amn = bo->mn;
+	if (amn == NULL) {
 		mutex_unlock(&adev->mn_lock);
 		return;
 	}
 
-	down_write(&rmn->lock);
+	down_write(&amn->lock);
 
 	/* save the next list entry for later */
 	head = bo->mn_list.next;
@@ -456,12 +525,13 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo)
 
 	if (list_empty(head)) {
 		struct amdgpu_mn_node *node;
+
 		node = container_of(head, struct amdgpu_mn_node, bos);
-		interval_tree_remove(&node->it, &rmn->objects);
+		interval_tree_remove(&node->it, &amn->objects);
 		kfree(node);
 	}
 
-	up_write(&rmn->lock);
+	up_write(&amn->lock);
 	mutex_unlock(&adev->mn_lock);
 }