Skip to contentSkip to conversation

btrfs: Make btrfs_finish_chunk_alloc private to block-group.c

PATCH Superseded Nikolay Borisov 3 participants 5 messages Jul 2, 2021 Btrfs file system

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 btrfs: Make btrfs_finish_chunk_alloc private to block-group.c

One of the final things that must be done to add a new chunk is allocating its device extents and inserting them in the extent tree. This is currently done in btrfs_finish_chunk_alloc whose name isn't very informative. What's more this functino is only used in block-group.c but is defined as public. There isn't anything special about it that would warrant it being defined in volumes.c.

Just move btrfs_finish_chunk_alloc and alloc_chunk_dev_extent to block-group.c, make the former static and it to alloc_chunk_dev_extents. The new name is a lot more explicit about the purpose of the function.

---
 fs/btrfs/block-group.c | 98 +++++++++++++++++++++++++++++++++++++++++-
 fs/btrfs/volumes.c     | 92 ---------------------------------------
 fs/btrfs/volumes.h     |  2 -
 3 files changed, 96 insertions(+), 96 deletions(-)
fs/btrfs/block-group.c +96 -2
index c557327b4545..2e133ee61d83 100644
@@ -2236,6 +2236,100 @@ static int insert_block_group_item(struct btrfs_trans_handle *trans,
 	return btrfs_insert_item(trans, root, &key, &bgi, sizeof(bgi));
 }
 
+
+static int alloc_dev_extent(struct btrfs_trans_handle *trans,
+			    struct btrfs_device *device, u64 chunk_offset,
+			    u64 start, u64 num_bytes)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_fs_info *fs_info = device->fs_info;
+	struct btrfs_root *root = fs_info->dev_root;
+	struct btrfs_dev_extent *extent;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+
+	WARN_ON(!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &device->dev_state));
+	WARN_ON(test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state));
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = device->devid;
+	key.offset = start;
+	key.type = BTRFS_DEV_EXTENT_KEY;
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(*extent));
+	if (ret)
+		goto out;
+
+	leaf = path->nodes[0];
+	extent = btrfs_item_ptr(leaf, path->slots[0],
+				struct btrfs_dev_extent);
+	btrfs_set_dev_extent_chunk_tree(leaf, extent,
+					BTRFS_CHUNK_TREE_OBJECTID);
+	btrfs_set_dev_extent_chunk_objectid(leaf, extent,
+					    BTRFS_FIRST_CHUNK_TREE_OBJECTID);
+	btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset);
+
+	btrfs_set_dev_extent_length(leaf, extent, num_bytes);
+	btrfs_mark_buffer_dirty(leaf);
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+
+/*
+ * This function belongs to phase 2.
+ *
+ * See the comment at btrfs_chunk_alloc() for details about the chunk allocation
+ * phases.
+ */
+static int alloc_chunk_dev_extents(struct btrfs_trans_handle *trans,
+				   u64 chunk_offset, u64 chunk_size)
+{
+	struct btrfs_fs_info *fs_info = trans->fs_info;
+	struct btrfs_device *device;
+	struct extent_map *em;
+	struct map_lookup *map;
+	u64 dev_offset;
+	u64 stripe_size;
+	int i;
+	int ret = 0;
+
+	em = btrfs_get_chunk_map(fs_info, chunk_offset, chunk_size);
+	if (IS_ERR(em))
+		return PTR_ERR(em);
+
+	map = em->map_lookup;
+	stripe_size = em->orig_block_len;
+
+	/*
+	 * Take the device list mutex to prevent races with the final phase of
+	 * a device replace operation that replaces the device object associated
+	 * with the map's stripes, because the device object's id can change
+	 * at any time during that final phase of the device replace operation
+	 * (dev-replace.c:btrfs_dev_replace_finishing()), so we could grab the
+	 * replaced device and then see it with an ID of BTRFS_DEV_REPLACE_DEVID,
+	 * resulting in persisting a device extent item with such ID.
+	 */
+	mutex_lock(&fs_info->fs_devices->device_list_mutex);
+	for (i = 0; i < map->num_stripes; i++) {
+		device = map->stripes[i].dev;
+		dev_offset = map->stripes[i].physical;
+
+		ret = alloc_dev_extent(trans, device, chunk_offset, dev_offset,
+				       stripe_size);
+		if (ret)
+			break;
+	}
+	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+
+	free_extent_map(em);
+	return ret;
+}
+
 /*
  * This function, btrfs_create_pending_block_groups(), belongs to the phase 2 of
  * chunk allocation.
@@ -2270,8 +2364,8 @@ void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans)
 			if (ret)
 				btrfs_abort_transaction(trans, ret);
 		}
-		ret = btrfs_finish_chunk_alloc(trans, block_group->start,
-					block_group->length);
+		ret = alloc_chunk_dev_extents(trans, block_group->start,
+					      block_group->length);
 		if (ret)
 			btrfs_abort_transaction(trans, ret);
 		add_block_group_free_space(trans, block_group);
fs/btrfs/volumes.c +0 -92

104 lines not on this page · raw message

fs/btrfs/volumes.h +0 -2
index 55a8ba244716..70c749eee3ad 100644
@@ -508,8 +508,6 @@ int btrfs_is_parity_mirror(struct btrfs_fs_info *fs_info,
 			   u64 logical, u64 len);
 unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,
 				    u64 logical);
-int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
-			     u64 chunk_offset, u64 chunk_size);
 int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
 				     struct btrfs_block_group *bg);
 int btrfs_remove_chunk(struct btrfs_trans_handle *trans, u64 chunk_offset);
Signature · 1 line
2.17.1
#1 lore raw
Quoted 10 lines · Nikolay Borisov

One of the final things that must be done to add a new chunk is allocating its device extents and inserting them in the extent tree. This is currently done in btrfs_finish_chunk_alloc whose name isn't very
informative. What's more this functino is only used in block-group.c bu function ~^ but ~^

is defined as public. There isn't anything special about it that would warrant it being defined in volumes.c.

Just move btrfs_finish_chunk_alloc and alloc_chunk_dev_extent to block-group.c, make the former static and it to alloc_chunk_dev_extents.

rename ~^

Quoted 15 lines

The new name is a lot more explicit about the purpose of the function.

---
 fs/btrfs/block-group.c | 98 +++++++++++++++++++++++++++++++++++++++++-
 fs/btrfs/volumes.c     | 92 ---------------------------------------
 fs/btrfs/volumes.h     |  2 -
 3 files changed, 96 insertions(+), 96 deletions(-)
fs/btrfs/block-group.c +0 -0
index c557327b4545..2e133ee61d83 100644

fs/btrfs/block-group.c static int insert_block_group_item(struct btrfs_trans_handle *trans, · line 2239

Quoted 4 lines
fs/btrfs/block-group.c +1 -0
@@ -2236,6 +2236,100 @@ static int insert_block_group_item(struct btrfs_trans_handle *trans,
 	return btrfs_insert_item(trans, root, &key, &bgi, sizeof(bgi));
 }
 
+

double newline

#2 lore raw
Quoted 7 lines · Johannes Thumshirn
Quoted 4 lines · Nikolay Borisov

One of the final things that must be done to add a new chunk is allocating its device extents and inserting them in the extent tree. This is currently done in btrfs_finish_chunk_alloc whose name isn't very
informative. What's more this functino is only used in block-group.c bu function ~^ but ~^

Quoted 9 lines
Quoted 5 lines

is defined as public. There isn't anything special about it that would warrant it being defined in volumes.c.

Just move btrfs_finish_chunk_alloc and alloc_chunk_dev_extent to block-group.c, make the former static and it to alloc_chunk_dev_extents.

rename ~^

Quoted 1 line

The new name is a lot more explicit about the purpose of the function.

I find the new name confusing as well.

alloc_chunk_dev_extents() suggests it's allocating device extents, but it's not, all it does is insert the device extent items into the devices btree.
Allocation of extents happened in phase 1, through btrfs_alloc_chunk().

Thanks.

Quoted 21 lines
Quoted 14 lines
---
 fs/btrfs/block-group.c | 98 +++++++++++++++++++++++++++++++++++++++++-
 fs/btrfs/volumes.c     | 92 ---------------------------------------
 fs/btrfs/volumes.h     |  2 -
 3 files changed, 96 insertions(+), 96 deletions(-)
fs/btrfs/block-group.c +0 -0
index c557327b4545..2e133ee61d83 100644

fs/btrfs/block-group.c static int insert_block_group_item(struct btrfs_trans_handle *trans, · line 2239

Quoted 4 lines
fs/btrfs/block-group.c +1 -0
@@ -2236,6 +2236,100 @@ static int insert_block_group_item(struct btrfs_trans_handle *trans,
      return btrfs_insert_item(trans, root, &key, &bgi, sizeof(bgi));
 }
 
+

double newline

Signature · 3 lines
Filipe David Manana,

“Whether you think you can, or you think you can't — you're right.”
#3 lore raw
Quoted 17 lines · Filipe Manana
Quoted 15 lines · Johannes Thumshirn
Quoted 10 lines · Nikolay Borisov

One of the final things that must be done to add a new chunk is allocating its device extents and inserting them in the extent tree. This is currently done in btrfs_finish_chunk_alloc whose name isn't very
informative. What's more this functino is only used in block-group.c bu function ~^ but ~^

is defined as public. There isn't anything special about it that would warrant it being defined in volumes.c.

Just move btrfs_finish_chunk_alloc and alloc_chunk_dev_extent to block-group.c, make the former static and it to alloc_chunk_dev_extents.

rename ~^

Quoted 8 lines
Quoted 1 line
Quoted 1 line

The new name is a lot more explicit about the purpose of the function.

I find the new name confusing as well.

alloc_chunk_dev_extents() suggests it's allocating device extents, but it's not, all it does is insert the device extent items into the devices btree.
Allocation of extents happened in phase 1, through btrfs_alloc_chunk().

The reason I give it plural name is because the inner function is called alloc_dev_extent hence I gave the top level functino a plural form. I'm fine simply renaming both functions to insert_dev_extent[s] respectively?

<snip>

#4 lore raw
Quoted 25 lines · Nikolay Borisov
Quoted 21 lines · Filipe Manana
Quoted 7 lines · Johannes Thumshirn
Quoted 4 lines · Nikolay Borisov

One of the final things that must be done to add a new chunk is allocating its device extents and inserting them in the extent tree. This is currently done in btrfs_finish_chunk_alloc whose name isn't very
informative. What's more this functino is only used in block-group.c bu function ~^ but ~^

Quoted 9 lines
Quoted 5 lines

is defined as public. There isn't anything special about it that would warrant it being defined in volumes.c.

Just move btrfs_finish_chunk_alloc and alloc_chunk_dev_extent to block-group.c, make the former static and it to alloc_chunk_dev_extents.

rename ~^

Quoted 1 line

The new name is a lot more explicit about the purpose of the function.

I find the new name confusing as well.

Quoted 8 lines
Quoted 4 lines

alloc_chunk_dev_extents() suggests it's allocating device extents, but it's not, all it does is insert the device extent items into the devices btree.
Allocation of extents happened in phase 1, through btrfs_alloc_chunk().

The reason I give it plural name is because the inner function is called alloc_dev_extent hence I gave the top level functino a plural form. I'm fine simply renaming both functions to insert_dev_extent[s] respectively?

Yes, that's much better. Either "insert" or "add".

Also please fix the changelog:

"One of the final things that must be done to add a new chunk is allocating its device extents and inserting them in the extent tree."

The allocation part is not done in this phase, as I mentioned earlier, and the insertion is in the devices btree and not in the extent btree.

Thanks.

Quoted 2 lines

<snip>

Signature · 3 lines
Filipe David Manana,

“Whether you think you can, or you think you can't — you're right.”