drm/ttm: Fix up io_mem_reserve / io_mem_free calling

This patch attempts to fix up shortcomings with the current calling
sequences.

1) There's a fastpath where no locking occurs and only io_mem_reserved is
   called to obtain needed info for mapping. The fastpath is set per
   memory type manager.
2) If the fastpath is disabled, io_mem_reserve and io_mem_free will be exactly
   balanced and not called recursively for the same struct ttm_mem_reg.
3) Optionally the driver can choose to enable a per memory type manager LRU
   eviction mechanism that, when io_mem_reserve returns -EAGAIN will attempt
   to kill user-space mappings of memory in that manager to free up needed
   resources

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

3 files changed, 166 insertions, 30 deletions

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 25e4c2a..cf2ec56 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -378,8 +378,13 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 	int ret = 0;
 
 	if (old_is_pci || new_is_pci ||
-	    ((mem->placement & bo->mem.placement & TTM_PL_MASK_CACHING) == 0))
-		ttm_bo_unmap_virtual(bo);
+	    ((mem->placement & bo->mem.placement & TTM_PL_MASK_CACHING) == 0)) {
+		ret = ttm_mem_io_lock(old_man, true);
+		if (unlikely(ret != 0))
+			goto out_err;
+		ttm_bo_unmap_virtual_locked(bo);
+		ttm_mem_io_unlock(old_man);
+	}
 
 	/*
 	 * Create and bind a ttm if required.
@@ -466,7 +471,6 @@ static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo)
 		ttm_tt_destroy(bo->ttm);
 		bo->ttm = NULL;
 	}
-
 	ttm_bo_mem_put(bo, &bo->mem);
 
 	atomic_set(&bo->reserved, 0);
@@ -665,6 +669,7 @@ static void ttm_bo_release(struct kref *kref)
 	struct ttm_buffer_object *bo =
 	    container_of(kref, struct ttm_buffer_object, kref);
 	struct ttm_bo_device *bdev = bo->bdev;
+	struct ttm_mem_type_manager *man = &bdev->man[bo->mem.mem_type];
 
 	if (likely(bo->vm_node != NULL)) {
 		rb_erase(&bo->vm_rb, &bdev->addr_space_rb);
@@ -672,6 +677,9 @@ static void ttm_bo_release(struct kref *kref)
 		bo->vm_node = NULL;
 	}
 	write_unlock(&bdev->vm_lock);
+	ttm_mem_io_lock(man, false);
+	ttm_mem_io_free_vm(bo);
+	ttm_mem_io_unlock(man);
 	ttm_bo_cleanup_refs_or_queue(bo);
 	kref_put(&bo->list_kref, ttm_bo_release_list);
 	write_lock(&bdev->vm_lock);
@@ -728,7 +736,8 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo, bool interruptible,
 
 	evict_mem = bo->mem;
 	evict_mem.mm_node = NULL;
-	evict_mem.bus.io_reserved = false;
+	evict_mem.bus.io_reserved_vm = false;
+	evict_mem.bus.io_reserved_count = 0;
 
 	placement.fpfn = 0;
 	placement.lpfn = 0;
@@ -1065,7 +1074,8 @@ int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
 	mem.num_pages = bo->num_pages;
 	mem.size = mem.num_pages << PAGE_SHIFT;
 	mem.page_alignment = bo->mem.page_alignment;
-	mem.bus.io_reserved = false;
+	mem.bus.io_reserved_vm = false;
+	mem.bus.io_reserved_count = 0;
 	/*
 	 * Determine where to move the buffer.
 	 */
@@ -1184,6 +1194,7 @@ int ttm_bo_init(struct ttm_bo_device *bdev,
 	INIT_LIST_HEAD(&bo->lru);
 	INIT_LIST_HEAD(&bo->ddestroy);
 	INIT_LIST_HEAD(&bo->swap);
+	INIT_LIST_HEAD(&bo->io_reserve_lru);
 	bo->bdev = bdev;
 	bo->glob = bdev->glob;
 	bo->type = type;
@@ -1193,7 +1204,8 @@ int ttm_bo_init(struct ttm_bo_device *bdev,
 	bo->mem.num_pages = bo->num_pages;
 	bo->mem.mm_node = NULL;
 	bo->mem.page_alignment = page_alignment;
-	bo->mem.bus.io_reserved = false;
+	bo->mem.bus.io_reserved_vm = false;
+	bo->mem.bus.io_reserved_count = 0;
 	bo->buffer_start = buffer_start & PAGE_MASK;
 	bo->priv_flags = 0;
 	bo->mem.placement = (TTM_PL_FLAG_SYSTEM | TTM_PL_FLAG_CACHED);
@@ -1367,6 +1379,10 @@ int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type,
 	BUG_ON(type >= TTM_NUM_MEM_TYPES);
 	man = &bdev->man[type];
 	BUG_ON(man->has_type);
+	man->io_reserve_fastpath = true;
+	man->use_io_reserve_lru = false;
+	mutex_init(&man->io_reserve_mutex);
+	INIT_LIST_HEAD(&man->io_reserve_lru);
 
 	ret = bdev->driver->init_mem_type(bdev, type, man);
 	if (ret)
@@ -1574,7 +1590,7 @@ bool ttm_mem_reg_is_pci(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem)
 	return true;
 }
 
-void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo)
+void ttm_bo_unmap_virtual_locked(struct ttm_buffer_object *bo)
 {
 	struct ttm_bo_device *bdev = bo->bdev;
 	loff_t offset = (loff_t) bo->addr_space_offset;
@@ -1583,8 +1599,20 @@ void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo)
 	if (!bdev->dev_mapping)
 		return;
 	unmap_mapping_range(bdev->dev_mapping, offset, holelen, 1);
-	ttm_mem_io_free(bdev, &bo->mem);
+	ttm_mem_io_free_vm(bo);
 }
+
+void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo)
+{
+	struct ttm_bo_device *bdev = bo->bdev;
+	struct ttm_mem_type_manager *man = &bdev->man[bo->mem.mem_type];
+
+	ttm_mem_io_lock(man, false);
+	ttm_bo_unmap_virtual_locked(bo);
+	ttm_mem_io_unlock(man);
+}
+
+
 EXPORT_SYMBOL(ttm_bo_unmap_virtual);
 
 static void ttm_bo_vm_insert_rb(struct ttm_buffer_object *bo)
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 4b75133..a89839f 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -75,37 +75,123 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_bo_move_ttm);
 
-int ttm_mem_io_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem)
+int ttm_mem_io_lock(struct ttm_mem_type_manager *man, bool interruptible)
 {
-	int ret;
+	if (likely(man->io_reserve_fastpath))
+		return 0;
+
+	if (interruptible)
+		return mutex_lock_interruptible(&man->io_reserve_mutex);
+
+	mutex_lock(&man->io_reserve_mutex);
+	return 0;
+}
 
-	if (!mem->bus.io_reserved) {
-		mem->bus.io_reserved = true;
+void ttm_mem_io_unlock(struct ttm_mem_type_manager *man)
+{
+	if (likely(man->io_reserve_fastpath))
+		return;
+
+	mutex_unlock(&man->io_reserve_mutex);
+}
+
+static int ttm_mem_io_evict(struct ttm_mem_type_manager *man)
+{
+	struct ttm_buffer_object *bo;
+
+	if (!man->use_io_reserve_lru || list_empty(&man->io_reserve_lru))
+		return -EAGAIN;
+
+	bo = list_first_entry(&man->io_reserve_lru,
+			      struct ttm_buffer_object,
+			      io_reserve_lru);
+	list_del_init(&bo->io_reserve_lru);
+	ttm_bo_unmap_virtual_locked(bo);
+
+	return 0;
+}
+
+static int ttm_mem_io_reserve(struct ttm_bo_device *bdev,
+			      struct ttm_mem_reg *mem)
+{
+	struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type];
+	int ret = 0;
+
+	if (!bdev->driver->io_mem_reserve)
+		return 0;
+	if (likely(man->io_reserve_fastpath))
+		return bdev->driver->io_mem_reserve(bdev, mem);
+
+	if (bdev->driver->io_mem_reserve &&
+	    mem->bus.io_reserved_count++ == 0) {
+retry:
 		ret = bdev->driver->io_mem_reserve(bdev, mem);
+		if (ret == -EAGAIN) {
+			ret = ttm_mem_io_evict(man);
+			if (ret == 0)
+				goto retry;
+		}
+	}
+	return ret;
+}
+
+static void ttm_mem_io_free(struct ttm_bo_device *bdev,
+			    struct ttm_mem_reg *mem)
+{
+	struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type];
+
+	if (likely(man->io_reserve_fastpath))
+		return;
+
+	if (bdev->driver->io_mem_reserve &&
+	    --mem->bus.io_reserved_count == 0 &&
+	    bdev->driver->io_mem_free)
+		bdev->driver->io_mem_free(bdev, mem);
+
+}
+
+int ttm_mem_io_reserve_vm(struct ttm_buffer_object *bo)
+{
+	struct ttm_mem_reg *mem = &bo->mem;
+	int ret;
+
+	if (!mem->bus.io_reserved_vm) {
+		struct ttm_mem_type_manager *man =
+			&bo->bdev->man[mem->mem_type];
+
+		ret = ttm_mem_io_reserve(bo->bdev, mem);
 		if (unlikely(ret != 0))
 			return ret;
+		mem->bus.io_reserved_vm = true;
+		if (man->use_io_reserve_lru)
+			list_add_tail(&bo->io_reserve_lru,
+				      &man->io_reserve_lru);
 	}
 	return 0;
 }
 
-void ttm_mem_io_free(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem)
+void ttm_mem_io_free_vm(struct ttm_buffer_object *bo)
 {
-	if (bdev->driver->io_mem_reserve) {
-		if (mem->bus.io_reserved) {
-			mem->bus.io_reserved = false;
-			bdev->driver->io_mem_free(bdev, mem);
-		}
+	struct ttm_mem_reg *mem = &bo->mem;
+
+	if (mem->bus.io_reserved_vm) {
+		mem->bus.io_reserved_vm = false;
+		list_del_init(&bo->io_reserve_lru);
+		ttm_mem_io_free(bo->bdev, mem);
 	}
 }
 
 int ttm_mem_reg_ioremap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 			void **virtual)
 {
+	struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type];
 	int ret;
 	void *addr;
 
 	*virtual = NULL;
+	(void) ttm_mem_io_lock(man, false);
 	ret = ttm_mem_io_reserve(bdev, mem);
+	ttm_mem_io_unlock(man);
 	if (ret || !mem->bus.is_iomem)
 		return ret;
 
@@ -117,7 +203,9 @@ int ttm_mem_reg_ioremap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 		else
 			addr = ioremap_nocache(mem->bus.base + mem->bus.offset, mem->bus.size);
 		if (!addr) {
+			(void) ttm_mem_io_lock(man, false);
 			ttm_mem_io_free(bdev, mem);
+			ttm_mem_io_unlock(man);
 			return -ENOMEM;
 		}
 	}
@@ -134,7 +222,9 @@ void ttm_mem_reg_iounmap(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem,
 
 	if (virtual && mem->bus.addr == NULL)
 		iounmap(virtual);
+	(void) ttm_mem_io_lock(man, false);
 	ttm_mem_io_free(bdev, mem);
+	ttm_mem_io_unlock(man);
 }
 
 static int ttm_copy_io_page(void *dst, void *src, unsigned long page)
@@ -231,7 +321,7 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	struct ttm_mem_type_manager *man = &bdev->man[new_mem->mem_type];
 	struct ttm_tt *ttm = bo->ttm;
 	struct ttm_mem_reg *old_mem = &bo->mem;
-	struct ttm_mem_reg old_copy = *old_mem;
+	struct ttm_mem_reg old_copy;
 	void *old_iomap;
 	void *new_iomap;
 	int ret;
@@ -281,7 +371,7 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	mb();
 out2:
 	ttm_bo_free_old_node(bo);
-
+	old_copy = *old_mem;
 	*old_mem = *new_mem;
 	new_mem->mm_node = NULL;
 
@@ -292,7 +382,7 @@ out2:
 	}
 
 out1:
-	ttm_mem_reg_iounmap(bdev, new_mem, new_iomap);
+	ttm_mem_reg_iounmap(bdev, old_mem, new_iomap);
 out:
 	ttm_mem_reg_iounmap(bdev, &old_copy, old_iomap);
 	return ret;
@@ -341,6 +431,7 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
 	INIT_LIST_HEAD(&fbo->ddestroy);
 	INIT_LIST_HEAD(&fbo->lru);
 	INIT_LIST_HEAD(&fbo->swap);
+	INIT_LIST_HEAD(&fbo->io_reserve_lru);
 	fbo->vm_node = NULL;
 	atomic_set(&fbo->cpu_writers, 0);
 
@@ -452,6 +543,8 @@ int ttm_bo_kmap(struct ttm_buffer_object *bo,
 		unsigned long start_page, unsigned long num_pages,
 		struct ttm_bo_kmap_obj *map)
 {
+	struct ttm_mem_type_manager *man =
+		&bo->bdev->man[bo->mem.mem_type];
 	unsigned long offset, size;
 	int ret;
 
@@ -466,7 +559,9 @@ int ttm_bo_kmap(struct ttm_buffer_object *bo,
 	if (num_pages > 1 && !DRM_SUSER(DRM_CURPROC))
 		return -EPERM;
 #endif
+	(void) ttm_mem_io_lock(man, false);
 	ret = ttm_mem_io_reserve(bo->bdev, &bo->mem);
+	ttm_mem_io_unlock(man);
 	if (ret)
 		return ret;
 	if (!bo->mem.bus.is_iomem) {
@@ -481,12 +576,15 @@ EXPORT_SYMBOL(ttm_bo_kmap);
 
 void ttm_bo_kunmap(struct ttm_bo_kmap_obj *map)
 {
+	struct ttm_buffer_object *bo = map->bo;
+	struct ttm_mem_type_manager *man =
+		&bo->bdev->man[bo->mem.mem_type];
+
 	if (!map->virtual)
 		return;
 	switch (map->bo_kmap_type) {
 	case ttm_bo_map_iomap:
 		iounmap(map->virtual);
-		ttm_mem_io_free(map->bo->bdev, &map->bo->mem);
 		break;
 	case ttm_bo_map_vmap:
 		vunmap(map->virtual);
@@ -499,6 +597,9 @@ void ttm_bo_kunmap(struct ttm_bo_kmap_obj *map)
 	default:
 		BUG();
 	}
+	(void) ttm_mem_io_lock(man, false);
+	ttm_mem_io_free(map->bo->bdev, &map->bo->mem);
+	ttm_mem_io_unlock(man);
 	map->virtual = NULL;
 	map->page = NULL;
 }
diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c
index 8dd446c..221b924 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_vm.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c
@@ -83,6 +83,8 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 	int i;
 	unsigned long address = (unsigned long)vmf->virtual_address;
 	int retval = VM_FAULT_NOPAGE;
+	struct ttm_mem_type_manager *man =
+		&bdev->man[bo->mem.mem_type];
 
 	/*
 	 * Work around locking order reversal in fault / nopfn
@@ -130,12 +132,16 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 	} else
 		spin_unlock(&bdev->fence_lock);
 
-
-	ret = ttm_mem_io_reserve(bdev, &bo->mem);
-	if (ret) {
-		retval = VM_FAULT_SIGBUS;
+	ret = ttm_mem_io_lock(man, true);
+	if (unlikely(ret != 0)) {
+		retval = VM_FAULT_NOPAGE;
 		goto out_unlock;
 	}
+	ret = ttm_mem_io_reserve_vm(bo);
+	if (unlikely(ret != 0)) {
+		retval = VM_FAULT_SIGBUS;
+		goto out_io_unlock;
+	}
 
 	page_offset = ((address - vma->vm_start) >> PAGE_SHIFT) +
 	    bo->vm_node->start - vma->vm_pgoff;
@@ -144,7 +150,7 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 
 	if (unlikely(page_offset >= bo->num_pages)) {
 		retval = VM_FAULT_SIGBUS;
-		goto out_unlock;
+		goto out_io_unlock;
 	}
 
 	/*
@@ -182,7 +188,7 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 			page = ttm_tt_get_page(ttm, page_offset);
 			if (unlikely(!page && i == 0)) {
 				retval = VM_FAULT_OOM;
-				goto out_unlock;
+				goto out_io_unlock;
 			} else if (unlikely(!page)) {
 				break;
 			}
@@ -200,14 +206,15 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 		else if (unlikely(ret != 0)) {
 			retval =
 			    (ret == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS;
-			goto out_unlock;
+			goto out_io_unlock;
 		}
 
 		address += PAGE_SIZE;
 		if (unlikely(++page_offset >= page_last))
 			break;
 	}
-
+out_io_unlock:
+	ttm_mem_io_unlock(man);
 out_unlock:
 	ttm_bo_unreserve(bo);
 	return retval;