btrfs: update btrfs_space_info's bytes_may_use timely

This patch can fix some false ENOSPC errors, below test script can
reproduce one false ENOSPC error:
	#!/bin/bash
	dd if=/dev/zero of=fs.img bs=$((1024*1024)) count=128
	dev=$(losetup --show -f fs.img)
	mkfs.btrfs -f -M $dev
	mkdir /tmp/mntpoint
	mount $dev /tmp/mntpoint
	cd /tmp/mntpoint
	xfs_io -f -c "falloc 0 $((64*1024*1024))" testfile

Above script will fail for ENOSPC reason, but indeed fs still has free
space to satisfy this request. Please see call graph:
btrfs_fallocate()
|-> btrfs_alloc_data_chunk_ondemand()
|   bytes_may_use += 64M
|-> btrfs_prealloc_file_range()
    |-> btrfs_reserve_extent()
        |-> btrfs_add_reserved_bytes()
        |   alloc_type is RESERVE_ALLOC_NO_ACCOUNT, so it does not
        |   change bytes_may_use, and bytes_reserved += 64M. Now
        |   bytes_may_use + bytes_reserved == 128M, which is greater
        |   than btrfs_space_info's total_bytes, false enospc occurs.
        |   Note, the bytes_may_use decrease operation will be done in
        |   end of btrfs_fallocate(), which is too late.

Here is another simple case for buffered write:
                    CPU 1              |              CPU 2
                                       |
|-> cow_file_range()                   |-> __btrfs_buffered_write()
    |-> btrfs_reserve_extent()         |   |
    |                                  |   |
    |                                  |   |
    |    .....                         |   |-> btrfs_check_data_free_space()
    |                                  |
    |                                  |
    |-> extent_clear_unlock_delalloc() |

In CPU 1, btrfs_reserve_extent()->find_free_extent()->
btrfs_add_reserved_bytes() do not decrease bytes_may_use, the decrease
operation will be delayed to be done in extent_clear_unlock_delalloc().
Assume in this case, btrfs_reserve_extent() reserved 128MB data, CPU2's
btrfs_check_data_free_space() tries to reserve 100MB data space.
If
	100MB > data_sinfo->total_bytes - data_sinfo->bytes_used -
		data_sinfo->bytes_reserved - data_sinfo->bytes_pinned -
		data_sinfo->bytes_readonly - data_sinfo->bytes_may_use
btrfs_check_data_free_space() will try to allcate new data chunk or call
btrfs_start_delalloc_roots(), or commit current transaction in order to
reserve some free space, obviously a lot of work. But indeed it's not
necessary as long as decreasing bytes_may_use timely, we still have
free space, decreasing 128M from bytes_may_use.

To fix this issue, this patch chooses to update bytes_may_use for both
data and metadata in btrfs_add_reserved_bytes(). For compress path, real
extent length may not be equal to file content length, so introduce a
ram_bytes argument for btrfs_reserve_extent(), find_free_extent() and
btrfs_add_reserved_bytes(), it's becasue bytes_may_use is increased by
file content length. Then compress path can update bytes_may_use
correctly. Also now we can discard RESERVE_ALLOC_NO_ACCOUNT, RESERVE_ALLOC
and RESERVE_FREE.

As we know, usually EXTENT_DO_ACCOUNTING is used for error path. In
run_delalloc_nocow(), for inode marked as NODATACOW or extent marked as
PREALLOC, we also need to update bytes_may_use, but can not pass
EXTENT_DO_ACCOUNTING, because it also clears metadata reservation, so
here we introduce EXTENT_CLEAR_DATA_RESV flag to indicate btrfs_clear_bit_hook()
to update btrfs_space_info's bytes_may_use.

Meanwhile __btrfs_prealloc_file_range() will call
btrfs_free_reserved_data_space() internally for both sucessful and failed
path, btrfs_prealloc_file_range()'s callers does not need to call
btrfs_free_reserved_data_space() any more.

Signed-off-by: Wang Xiaoguang <wangxg.fnst@cn.fujitsu.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Chris Mason <clm@fb.com>

fs/btrfs/ctree.h

@@ -2579,7 +2579,7 @@ int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root,
 				   u64 root_objectid, u64 owner, u64 offset,
 				   struct btrfs_key *ins);
-int btrfs_reserve_extent(struct btrfs_root *root, u64 num_bytes,
+int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes, u64 num_bytes,
 			 u64 min_alloc_size, u64 empty_size, u64 hint_byte,
 			 struct btrfs_key *ins, int is_data, int delalloc);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,

fs/btrfs/extent-tree.c

@@ -60,21 +60,6 @@ enum {
 	CHUNK_ALLOC_FORCE = 2,
 };
 
-/*
- * Control how reservations are dealt with.
- *
- * RESERVE_FREE - freeing a reservation.
- * RESERVE_ALLOC - allocating space and we need to update bytes_may_use for
- *   ENOSPC accounting
- * RESERVE_ALLOC_NO_ACCOUNT - allocating space and we should not update
- *   bytes_may_use as the ENOSPC accounting is done elsewhere
- */
-enum {
-	RESERVE_FREE = 0,
-	RESERVE_ALLOC = 1,
-	RESERVE_ALLOC_NO_ACCOUNT = 2,
-};
-
 static int update_block_group(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root, u64 bytenr,
 			      u64 num_bytes, int alloc);
@@ -105,7 +90,7 @@ static int find_next_key(struct btrfs_path *path, int level,
 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
 			    int dump_block_groups);
 static int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
-				    u64 num_bytes, int reserve, int delalloc);
+				    u64 ram_bytes, u64 num_bytes, int delalloc);
 static int btrfs_free_reserved_bytes(struct btrfs_block_group_cache *cache,
 				     u64 num_bytes, int delalloc);
 static int block_rsv_use_bytes(struct btrfs_block_rsv *block_rsv,
@@ -3502,7 +3487,6 @@ again:
 		dcs = BTRFS_DC_SETUP;
 	else if (ret == -ENOSPC)
 		set_bit(BTRFS_TRANS_CACHE_ENOSPC, &trans->transaction->flags);
-	btrfs_free_reserved_data_space(inode, 0, num_pages);
 
 out_put:
 	iput(inode);
@@ -6500,8 +6484,9 @@ void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg)
 /**
  * btrfs_add_reserved_bytes - update the block_group and space info counters
  * @cache:	The cache we are manipulating
+ * @ram_bytes:  The number of bytes of file content, and will be same to
+ *              @num_bytes except for the compress path.
  * @num_bytes:	The number of bytes in question
- * @reserve:	One of the reservation enums
  * @delalloc:   The blocks are allocated for the delalloc write
  *
  * This is called by the allocator when it reserves space. Metadata
@@ -6516,7 +6501,7 @@ void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg)
  * succeeds.
  */
 static int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
-				    u64 num_bytes, int reserve, int delalloc)
+				    u64 ram_bytes, u64 num_bytes, int delalloc)
 {
 	struct btrfs_space_info *space_info = cache->space_info;
 	int ret = 0;
@@ -6528,13 +6513,11 @@ static int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
 	} else {
 		cache->reserved += num_bytes;
 		space_info->bytes_reserved += num_bytes;
-		if (reserve == RESERVE_ALLOC) {
-			trace_btrfs_space_reservation(cache->fs_info,
-					"space_info", space_info->flags,
-					num_bytes, 0);
-			space_info->bytes_may_use -= num_bytes;
-		}
 
+		trace_btrfs_space_reservation(cache->fs_info,
+				"space_info", space_info->flags,
+				ram_bytes, 0);
+		space_info->bytes_may_use -= ram_bytes;
 		if (delalloc)
 			cache->delalloc_bytes += num_bytes;
 	}
@@ -7433,9 +7416,9 @@ btrfs_release_block_group(struct btrfs_block_group_cache *cache,
  * the free space extent currently.
  */
 static noinline int find_free_extent(struct btrfs_root *orig_root,
-				     u64 num_bytes, u64 empty_size,
-				     u64 hint_byte, struct btrfs_key *ins,
-				     u64 flags, int delalloc)
+				u64 ram_bytes, u64 num_bytes, u64 empty_size,
+				u64 hint_byte, struct btrfs_key *ins,
+				u64 flags, int delalloc)
 {
 	int ret = 0;
 	struct btrfs_root *root = orig_root->fs_info->extent_root;
@@ -7447,8 +7430,6 @@ static noinline int find_free_extent(struct btrfs_root *orig_root,
 	struct btrfs_space_info *space_info;
 	int loop = 0;
 	int index = __get_raid_index(flags);
-	int alloc_type = (flags & BTRFS_BLOCK_GROUP_DATA) ?
-		RESERVE_ALLOC_NO_ACCOUNT : RESERVE_ALLOC;
 	bool failed_cluster_refill = false;
 	bool failed_alloc = false;
 	bool use_cluster = true;
@@ -7780,8 +7761,8 @@ checks:
 					     search_start - offset);
 		BUG_ON(offset > search_start);
 
-		ret = btrfs_add_reserved_bytes(block_group, num_bytes,
-				alloc_type, delalloc);
+		ret = btrfs_add_reserved_bytes(block_group, ram_bytes,
+				num_bytes, delalloc);
 		if (ret == -EAGAIN) {
 			btrfs_add_free_space(block_group, offset, num_bytes);
 			goto loop;
@@ -7953,7 +7934,7 @@ again:
 	up_read(&info->groups_sem);
 }
 
-int btrfs_reserve_extent(struct btrfs_root *root,
+int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
 			 u64 num_bytes, u64 min_alloc_size,
 			 u64 empty_size, u64 hint_byte,
 			 struct btrfs_key *ins, int is_data, int delalloc)
@@ -7965,8 +7946,8 @@ int btrfs_reserve_extent(struct btrfs_root *root,
 	flags = btrfs_get_alloc_profile(root, is_data);
 again:
 	WARN_ON(num_bytes < root->sectorsize);
-	ret = find_free_extent(root, num_bytes, empty_size, hint_byte, ins,
-			       flags, delalloc);
+	ret = find_free_extent(root, ram_bytes, num_bytes, empty_size,
+			       hint_byte, ins, flags, delalloc);
 	if (!ret && !is_data) {
 		btrfs_dec_block_group_reservations(root->fs_info,
 						   ins->objectid);
@@ -7975,6 +7956,7 @@ again:
 			num_bytes = min(num_bytes >> 1, ins->offset);
 			num_bytes = round_down(num_bytes, root->sectorsize);
 			num_bytes = max(num_bytes, min_alloc_size);
+			ram_bytes = num_bytes;
 			if (num_bytes == min_alloc_size)
 				final_tried = true;
 			goto again;
@@ -8241,7 +8223,7 @@ int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
 		return -EINVAL;
 
 	ret = btrfs_add_reserved_bytes(block_group, ins->offset,
-			RESERVE_ALLOC_NO_ACCOUNT, 0);
+				       ins->offset, 0);
 	BUG_ON(ret); /* logic error */
 	ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
 					 0, owner, offset, ins, 1);
@@ -8385,7 +8367,7 @@ struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
 	if (IS_ERR(block_rsv))
 		return ERR_CAST(block_rsv);
 
-	ret = btrfs_reserve_extent(root, blocksize, blocksize,
+	ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
 				   empty_size, hint, &ins, 0, 0);
 	if (ret)
 		goto out_unuse;

fs/btrfs/extent_io.h

@@ -20,6 +20,7 @@
 #define EXTENT_DAMAGED		(1U << 14)
 #define EXTENT_NORESERVE	(1U << 15)
 #define EXTENT_QGROUP_RESERVED	(1U << 16)
+#define EXTENT_CLEAR_DATA_RESV	(1U << 17)
 #define EXTENT_IOBITS		(EXTENT_LOCKED | EXTENT_WRITEBACK)
 #define EXTENT_CTLBITS		(EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC)
 

fs/btrfs/file.c

@@ -2675,6 +2675,7 @@ static long btrfs_fallocate(struct file *file, int mode,
 
 	alloc_start = round_down(offset, blocksize);
 	alloc_end = round_up(offset + len, blocksize);
+	cur_offset = alloc_start;
 
 	/* Make sure we aren't being give some crap mode */
 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
@@ -2767,7 +2768,6 @@ static long btrfs_fallocate(struct file *file, int mode,
 
 	/* First, check if we exceed the qgroup limit */
 	INIT_LIST_HEAD(&reserve_list);
-	cur_offset = alloc_start;
 	while (1) {
 		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
 				      alloc_end - cur_offset, 0);
@@ -2794,6 +2794,14 @@ static long btrfs_fallocate(struct file *file, int mode,
 					last_byte - cur_offset);
 			if (ret < 0)
 				break;
+		} else {
+			/*
+			 * Do not need to reserve unwritten extent for this
+			 * range, free reserved data space first, otherwise
+			 * it'll result in false ENOSPC error.
+			 */
+			btrfs_free_reserved_data_space(inode, cur_offset,
+				last_byte - cur_offset);
 		}
 		free_extent_map(em);
 		cur_offset = last_byte;
@@ -2811,6 +2819,9 @@ static long btrfs_fallocate(struct file *file, int mode,
 					range->start,
 					range->len, 1 << inode->i_blkbits,
 					offset + len, &alloc_hint);
+		else
+			btrfs_free_reserved_data_space(inode, range->start,
+						       range->len);
 		list_del(&range->list);
 		kfree(range);
 	}
@@ -2845,18 +2856,11 @@ out_unlock:
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
 			     &cached_state, GFP_KERNEL);
 out:
-	/*
-	 * As we waited the extent range, the data_rsv_map must be empty
-	 * in the range, as written data range will be released from it.
-	 * And for prealloacted extent, it will also be released when
-	 * its metadata is written.
-	 * So this is completely used as cleanup.
-	 */
-	btrfs_qgroup_free_data(inode, alloc_start, alloc_end - alloc_start);
 	inode_unlock(inode);
 	/* Let go of our reservation. */
-	btrfs_free_reserved_data_space(inode, alloc_start,
-				       alloc_end - alloc_start);
+	if (ret != 0)
+		btrfs_free_reserved_data_space(inode, alloc_start,
+				       alloc_end - cur_offset);
 	return ret;
 }
 

fs/btrfs/inode-map.c

@@ -495,10 +495,9 @@ again:
 	ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc,
 					      prealloc, prealloc, &alloc_hint);
 	if (ret) {
-		btrfs_delalloc_release_space(inode, 0, prealloc);
+		btrfs_delalloc_release_metadata(inode, prealloc);
 		goto out_put;
 	}
-	btrfs_free_reserved_data_space(inode, 0, prealloc);
 
 	ret = btrfs_write_out_ino_cache(root, trans, path, inode);
 out_put:

fs/btrfs/inode.c

@@ -566,6 +566,8 @@ cont:
 						     PAGE_SET_WRITEBACK |
 						     page_error_op |
 						     PAGE_END_WRITEBACK);
+			btrfs_free_reserved_data_space_noquota(inode, start,
+						end - start + 1);
 			goto free_pages_out;
 		}
 	}
@@ -742,7 +744,7 @@ retry:
 		lock_extent(io_tree, async_extent->start,
 			    async_extent->start + async_extent->ram_size - 1);
 
-		ret = btrfs_reserve_extent(root,
+		ret = btrfs_reserve_extent(root, async_extent->ram_size,
 					   async_extent->compressed_size,
 					   async_extent->compressed_size,
 					   0, alloc_hint, &ins, 1, 1);
@@ -969,7 +971,8 @@ static noinline int cow_file_range(struct inode *inode,
 				     EXTENT_DEFRAG, PAGE_UNLOCK |
 				     PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK |
 				     PAGE_END_WRITEBACK);
-
+			btrfs_free_reserved_data_space_noquota(inode, start,
+						end - start + 1);
 			*nr_written = *nr_written +
 			     (end - start + PAGE_SIZE) / PAGE_SIZE;
 			*page_started = 1;
@@ -989,7 +992,7 @@ static noinline int cow_file_range(struct inode *inode,
 		unsigned long op;
 
 		cur_alloc_size = disk_num_bytes;
-		ret = btrfs_reserve_extent(root, cur_alloc_size,
+		ret = btrfs_reserve_extent(root, cur_alloc_size, cur_alloc_size,
 					   root->sectorsize, 0, alloc_hint,
 					   &ins, 1, 1);
 		if (ret < 0)
@@ -1489,8 +1492,10 @@ out_check:
 		extent_clear_unlock_delalloc(inode, cur_offset,
 					     cur_offset + num_bytes - 1,
 					     locked_page, EXTENT_LOCKED |
-					     EXTENT_DELALLOC, PAGE_UNLOCK |
-					     PAGE_SET_PRIVATE2);
+					     EXTENT_DELALLOC |
+					     EXTENT_CLEAR_DATA_RESV,
+					     PAGE_UNLOCK | PAGE_SET_PRIVATE2);
+
 		if (!nolock && nocow)
 			btrfs_end_write_no_snapshoting(root);
 		cur_offset = extent_end;
@@ -1807,7 +1812,9 @@ static void btrfs_clear_bit_hook(struct inode *inode,
 			return;
 
 		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
-		    && do_list && !(state->state & EXTENT_NORESERVE))
+		    && do_list && !(state->state & EXTENT_NORESERVE)
+		    && (*bits & (EXTENT_DO_ACCOUNTING |
+		    EXTENT_CLEAR_DATA_RESV)))
 			btrfs_free_reserved_data_space_noquota(inode,
 					state->start, len);
 
@@ -7252,7 +7259,7 @@ static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
 	int ret;
 
 	alloc_hint = get_extent_allocation_hint(inode, start, len);
-	ret = btrfs_reserve_extent(root, len, root->sectorsize, 0,
+	ret = btrfs_reserve_extent(root, len, len, root->sectorsize, 0,
 				   alloc_hint, &ins, 1, 1);
 	if (ret)
 		return ERR_PTR(ret);
@@ -7752,6 +7759,13 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
 				ret = PTR_ERR(em2);
 				goto unlock_err;
 			}
+			/*
+			 * For inode marked NODATACOW or extent marked PREALLOC,
+			 * use the existing or preallocated extent, so does not
+			 * need to adjust btrfs_space_info's bytes_may_use.
+			 */
+			btrfs_free_reserved_data_space_noquota(inode,
+					start, len);
 			goto unlock;
 		}
 	}
@@ -7786,7 +7800,6 @@ unlock:
 			i_size_write(inode, start + len);
 
 		adjust_dio_outstanding_extents(inode, dio_data, len);
-		btrfs_free_reserved_data_space(inode, start, len);
 		WARN_ON(dio_data->reserve < len);
 		dio_data->reserve -= len;
 		dio_data->unsubmitted_oe_range_end = start + len;
@@ -10306,6 +10319,7 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
 	u64 last_alloc = (u64)-1;
 	int ret = 0;
 	bool own_trans = true;
+	u64 end = start + num_bytes - 1;
 
 	if (trans)
 		own_trans = false;
@@ -10327,8 +10341,8 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
 		 * sized chunks.
 		 */
 		cur_bytes = min(cur_bytes, last_alloc);
-		ret = btrfs_reserve_extent(root, cur_bytes, min_size, 0,
-					   *alloc_hint, &ins, 1, 0);
+		ret = btrfs_reserve_extent(root, cur_bytes, cur_bytes,
+				min_size, 0, *alloc_hint, &ins, 1, 0);
 		if (ret) {
 			if (own_trans)
 				btrfs_end_transaction(trans, root);
@@ -10414,6 +10428,9 @@ next:
 		if (own_trans)
 			btrfs_end_transaction(trans, root);
 	}
+	if (cur_offset < end)
+		btrfs_free_reserved_data_space(inode, cur_offset,
+			end - cur_offset + 1);
 	return ret;
 }
 

fs/btrfs/relocation.c

@@ -3040,6 +3040,7 @@ int prealloc_file_extent_cluster(struct inode *inode,
 	int ret = 0;
 	u64 prealloc_start = cluster->start - offset;
 	u64 prealloc_end = cluster->end - offset;
+	u64 cur_offset;
 
 	BUG_ON(cluster->start != cluster->boundary[0]);
 	inode_lock(inode);
@@ -3049,6 +3050,7 @@ int prealloc_file_extent_cluster(struct inode *inode,
 	if (ret)
 		goto out;
 
+	cur_offset = prealloc_start;
 	while (nr < cluster->nr) {
 		start = cluster->boundary[nr] - offset;
 		if (nr + 1 < cluster->nr)
@@ -3058,16 +3060,21 @@ int prealloc_file_extent_cluster(struct inode *inode,
 
 		lock_extent(&BTRFS_I(inode)->io_tree, start, end);
 		num_bytes = end + 1 - start;
+		if (cur_offset < start)
+			btrfs_free_reserved_data_space(inode, cur_offset,
+					start - cur_offset);
 		ret = btrfs_prealloc_file_range(inode, 0, start,
 						num_bytes, num_bytes,
 						end + 1, &alloc_hint);
+		cur_offset = end + 1;
 		unlock_extent(&BTRFS_I(inode)->io_tree, start, end);
 		if (ret)
 			break;
 		nr++;
 	}
-	btrfs_free_reserved_data_space(inode, prealloc_start,
-				       prealloc_end + 1 - prealloc_start);
+	if (cur_offset < prealloc_end)
+		btrfs_free_reserved_data_space(inode, cur_offset,
+				       prealloc_end + 1 - cur_offset);
 out:
 	inode_unlock(inode);
 	return ret;