ext4: correct best extent lstart adjustment logic
authorBaokun Li <libaokun1@huawei.com>
Thu, 1 Feb 2024 14:18:45 +0000 (22:18 +0800)
committerTheodore Ts'o <tytso@mit.edu>
Thu, 22 Feb 2024 03:46:39 +0000 (22:46 -0500)
When yangerkun review commit 93cdf49f6eca ("ext4: Fix best extent lstart
adjustment logic in ext4_mb_new_inode_pa()"), it was found that the best
extent did not completely cover the original request after adjusting the
best extent lstart in ext4_mb_new_inode_pa() as follows:

  original request: 2/10(8)
  normalized request: 0/64(64)
  best extent: 0/9(9)

When we check if best ex can be kept at start of goal, ac_o_ex.fe_logical
is 2 less than the adjusted best extent logical end 9, so we think the
adjustment is done. But obviously 0/9(9) doesn't cover 2/10(8), so we
should determine here if the original request logical end is less than or
equal to the adjusted best extent logical end.

In addition, add a comment stating when adjusted best_ex will not cover
the original request, and remove the duplicate assertion because adjusting
lstart makes no change to b_ex.fe_len.

Link: https://lore.kernel.org/r/3630fa7f-b432-7afd-5f79-781bc3b2c5ea@huawei.com
Fixes: 93cdf49f6eca ("ext4: Fix best extent lstart adjustment logic in ext4_mb_new_inode_pa()")
Cc: <stable@kernel.org>
Signed-off-by: yangerkun <yangerkun@huawei.com>
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Link: https://lore.kernel.org/r/20240201141845.1879253-1-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
fs/ext4/mballoc.c

index bce82e1..bd26f2a 100644 (file)
@@ -5172,10 +5172,16 @@ ext4_mb_new_inode_pa(struct ext4_allocation_context *ac)
                        .fe_len = ac->ac_orig_goal_len,
                };
                loff_t orig_goal_end = extent_logical_end(sbi, &ex);
+               loff_t o_ex_end = extent_logical_end(sbi, &ac->ac_o_ex);
 
-               /* we can't allocate as much as normalizer wants.
-                * so, found space must get proper lstart
-                * to cover original request */
+               /*
+                * We can't allocate as much as normalizer wants, so we try
+                * to get proper lstart to cover the original request, except
+                * when the goal doesn't cover the original request as below:
+                *
+                * orig_ex:2045/2055(10), isize:8417280 -> normalized:0/2048
+                * best_ex:0/200(200) -> adjusted: 1848/2048(200)
+                */
                BUG_ON(ac->ac_g_ex.fe_logical > ac->ac_o_ex.fe_logical);
                BUG_ON(ac->ac_g_ex.fe_len < ac->ac_o_ex.fe_len);
 
@@ -5187,7 +5193,7 @@ ext4_mb_new_inode_pa(struct ext4_allocation_context *ac)
                 * 1. Check if best ex can be kept at end of goal (before
                 *    cr_best_avail trimmed it) and still cover original start
                 * 2. Else, check if best ex can be kept at start of goal and
-                *    still cover original start
+                *    still cover original end
                 * 3. Else, keep the best ex at start of original request.
                 */
                ex.fe_len = ac->ac_b_ex.fe_len;
@@ -5197,7 +5203,7 @@ ext4_mb_new_inode_pa(struct ext4_allocation_context *ac)
                        goto adjust_bex;
 
                ex.fe_logical = ac->ac_g_ex.fe_logical;
-               if (ac->ac_o_ex.fe_logical < extent_logical_end(sbi, &ex))
+               if (o_ex_end <= extent_logical_end(sbi, &ex))
                        goto adjust_bex;
 
                ex.fe_logical = ac->ac_o_ex.fe_logical;
@@ -5205,7 +5211,6 @@ adjust_bex:
                ac->ac_b_ex.fe_logical = ex.fe_logical;
 
                BUG_ON(ac->ac_o_ex.fe_logical < ac->ac_b_ex.fe_logical);
-               BUG_ON(ac->ac_o_ex.fe_len > ac->ac_b_ex.fe_len);
                BUG_ON(extent_logical_end(sbi, &ex) > orig_goal_end);
        }