ext4: Calculate metadata requirements more accurately

In the past, ext4_calc_metadata_amount(), and its sub-functions
ext4_ext_calc_metadata_amount() and ext4_indirect_calc_metadata_amount()
badly over-estimated the number of metadata blocks that might be
required for delayed allocation blocks.  This didn't matter as much
when functions which managed the reserved metadata blocks were more
aggressive about dropping reserved metadata blocks as delayed
allocation blocks were written, but unfortunately they were too
aggressive.  This was fixed in commit 0637c6f, but as a result the
over-estimation by ext4_calc_metadata_amount() would lead to reserving
2-3 times the number of pending delayed allocation blocks as
potentially required metadata blocks.  So if there are 1 megabytes of
blocks which have been not yet been allocation, up to 3 megabytes of
space would get reserved out of the user's quota and from the file
system free space pool until all of the inode's data blocks have been
allocated.

This commit addresses this problem by much more accurately estimating
the number of metadata blocks that will be required.  It will still
somewhat over-estimate the number of blocks needed, since it must make
a worst case estimate not knowing which physical blocks will be
needed, but it is much more accurate than before.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

1 files changed, 32 insertions, 17 deletions

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 91ae46098ea4..7d7b74e94687 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -296,29 +296,44 @@ static inline int ext4_ext_space_root_idx(struct inode *inode, int check)
  * to allocate @blocks
  * Worse case is one block per extent
  */
-int ext4_ext_calc_metadata_amount(struct inode *inode, int blocks)
+int ext4_ext_calc_metadata_amount(struct inode *inode, sector_t lblock)
 {
-        int lcap, icap, rcap, leafs, idxs, num;
-        int newextents = blocks;
-
-        rcap = ext4_ext_space_root_idx(inode, 0);
-        lcap = ext4_ext_space_block(inode, 0);
-        icap = ext4_ext_space_block_idx(inode, 0);
+        struct ext4_inode_info *ei = EXT4_I(inode);
+        int idxs, num = 0;
 
-        /* number of new leaf blocks needed */
-        num = leafs = (newextents + lcap - 1) / lcap;
+        idxs = ((inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
+                / sizeof(struct ext4_extent_idx));
 
         /*
-         * Worse case, we need separate index block(s)
-         * to link all new leaf blocks
+         * If the new delayed allocation block is contiguous with the
+         * previous da block, it can share index blocks with the
+         * previous block, so we only need to allocate a new index
+         * block every idxs leaf blocks.  At ldxs**2 blocks, we need
+         * an additional index block, and at ldxs**3 blocks, yet
+         * another index blocks.
          */
-        idxs = (leafs + icap - 1) / icap;
-        do {
-                num += idxs;
-                idxs = (idxs + icap - 1) / icap;
-        } while (idxs > rcap);
+        if (ei->i_da_metadata_calc_len &&
+            ei->i_da_metadata_calc_last_lblock+1 == lblock) {
+                if ((ei->i_da_metadata_calc_len % idxs) == 0)
+                        num++;
+                if ((ei->i_da_metadata_calc_len % (idxs*idxs)) == 0)
+                        num++;
+                if ((ei->i_da_metadata_calc_len % (idxs*idxs*idxs)) == 0) {
+                        num++;
+                        ei->i_da_metadata_calc_len = 0;
+                } else
+                        ei->i_da_metadata_calc_len++;
+                ei->i_da_metadata_calc_last_lblock++;
+                return num;
+        }
 
-        return num;
+        /*
+         * In the worst case we need a new set of index blocks at
+         * every level of the inode's extent tree.
+         */
+        ei->i_da_metadata_calc_len = 1;
+        ei->i_da_metadata_calc_last_lblock = lblock;
+        return ext_depth(inode) + 1;
 }
 
 static int