diff mbox series

[RFC,v4,20/34] ext4: use reserved metadata blocks when splitting extent on endio

Message ID 20240410132818.2812377-21-yi.zhang@huaweicloud.com (mailing list archive)
State New
Headers show
Series ext4: use iomap for regular file's buffered IO path and enable large folio | expand

Commit Message

Zhang Yi April 10, 2024, 1:28 p.m. UTC
From: Zhang Yi <yi.zhang@huawei.com>

ext4 only reserved space for delalloc for data blocks, doesn't reserve
space for metadata blocks in ext4_da_reserve_space(). Besides, if we
enable dioread_nolock mount option, it also doesn't reserve metadata
blocks for the extent status conversion.

In order to prevent data loss caused by fail to allocate metadata blocks
on writeback, we reserve 2% space or 4096 blocks for meta data, and use
EXT4_GET_BLOCKS_PRE_IO to do the potential split in advance. But all
these two methods were just best efforts, if it's really running out of
sapce, there is no difference between splitting extent on writeback and
on IO completed, both will lead to data loss.

The best way is to reserve enough space for metadata. Before that, we
can at least make sure that things won't get worse if we postpone
splitting extent to endio. So let's use reserved sapce in endio too.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
---
 fs/ext4/extents.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
diff mbox series

Patch

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 8bc8a519f745..fcb1916a7c29 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -3722,7 +3722,8 @@  static int ext4_convert_unwritten_extents_endio(handle_t *handle,
 			     (unsigned long long)map->m_lblk, map->m_len);
 #endif
 		err = ext4_split_convert_extents(handle, inode, map, ppath,
-						 EXT4_GET_BLOCKS_CONVERT);
+					EXT4_GET_BLOCKS_CONVERT |
+					EXT4_GET_BLOCKS_METADATA_NOFAIL);
 		if (err < 0)
 			return err;
 		path = ext4_find_extent(inode, map->m_lblk, ppath, 0);