OSDN Git Service

btrfs: make free space cache size consistent across different PAGE_SIZE
authorQu Wenruo <wqu@suse.com>
Tue, 13 Apr 2021 06:23:14 +0000 (14:23 +0800)
committerDavid Sterba <dsterba@suse.com>
Mon, 21 Jun 2021 13:19:08 +0000 (15:19 +0200)
Currently free space cache inode size is determined by two factors:

- block group size
- PAGE_SIZE

This means, for the same sized block groups, with different PAGE_SIZE,
it will result in different inode sizes.

This will not be a good thing for subpage support, so change the
requirement for PAGE_SIZE to sectorsize.

Now for the same 4K sectorsize btrfs, it should result the same inode
size no matter what the PAGE_SIZE is.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/block-group.c

index 6d5c4e4..c42b652 100644 (file)
@@ -2505,7 +2505,7 @@ static int cache_save_setup(struct btrfs_block_group *block_group,
        struct extent_changeset *data_reserved = NULL;
        u64 alloc_hint = 0;
        int dcs = BTRFS_DC_ERROR;
-       u64 num_pages = 0;
+       u64 cache_size = 0;
        int retries = 0;
        int ret = 0;
 
@@ -2617,20 +2617,20 @@ again:
         * taking up quite a bit since it's not folded into the other space
         * cache.
         */
-       num_pages = div_u64(block_group->length, SZ_256M);
-       if (!num_pages)
-               num_pages = 1;
+       cache_size = div_u64(block_group->length, SZ_256M);
+       if (!cache_size)
+               cache_size = 1;
 
-       num_pages *= 16;
-       num_pages *= PAGE_SIZE;
+       cache_size *= 16;
+       cache_size *= fs_info->sectorsize;
 
        ret = btrfs_check_data_free_space(BTRFS_I(inode), &data_reserved, 0,
-                                         num_pages);
+                                         cache_size);
        if (ret)
                goto out_put;
 
-       ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages,
-                                             num_pages, num_pages,
+       ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, cache_size,
+                                             cache_size, cache_size,
                                              &alloc_hint);
        /*
         * Our cache requires contiguous chunks so that we don't modify a bunch