]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
ext4: fix dirtyclusters double decrement on fs shutdown
authorBrian Foster <bfoster@redhat.com>
Tue, 13 Jan 2026 17:19:05 +0000 (12:19 -0500)
committerTheodore Ts'o <tytso@mit.edu>
Tue, 20 Jan 2026 03:51:23 +0000 (22:51 -0500)
fstests test generic/388 occasionally reproduces a warning in
ext4_put_super() associated with the dirty clusters count:

  WARNING: CPU: 7 PID: 76064 at fs/ext4/super.c:1324 ext4_put_super+0x48c/0x590 [ext4]

Tracing the failure shows that the warning fires due to an
s_dirtyclusters_counter value of -1. IOW, this appears to be a
spurious decrement as opposed to some sort of leak. Further tracing
of the dirty cluster count deltas and an LLM scan of the resulting
output identified the cause as a double decrement in the error path
between ext4_mb_mark_diskspace_used() and the caller
ext4_mb_new_blocks().

First, note that generic/388 is a shutdown vs. fsstress test and so
produces a random set of operations and shutdown injections. In the
problematic case, the shutdown triggers an error return from the
ext4_handle_dirty_metadata() call(s) made from
ext4_mb_mark_context(). The changed value is non-zero at this point,
so ext4_mb_mark_diskspace_used() does not exit after the error
bubbles up from ext4_mb_mark_context(). Instead, the former
decrements both cluster counters and returns the error up to
ext4_mb_new_blocks(). The latter falls into the !ar->len out path
which decrements the dirty clusters counter a second time, creating
the inconsistency.

To avoid this problem and simplify ownership of the cluster
reservation in this codepath, lift the counter reduction to a single
place in the caller. This makes it more clear that
ext4_mb_new_blocks() is responsible for acquiring cluster
reservation (via ext4_claim_free_clusters()) in the !delalloc case
as well as releasing it, regardless of whether it ends up consumed
or returned due to failure.

Fixes: 0087d9fb3f29 ("ext4: Fix s_dirty_blocks_counter if block allocation failed with nodelalloc")
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Baokun Li <libaokun1@huawei.com>
Link: https://patch.msgid.link/20260113171905.118284-1-bfoster@redhat.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
fs/ext4/mballoc-test.c
fs/ext4/mballoc.c

index a9416b20ff64c930d90fc177d01cd6a4639d8333..4abb40d4561ce3aeffb20307ee3733ae3a23eab0 100644 (file)
@@ -567,7 +567,7 @@ test_mark_diskspace_used_range(struct kunit *test,
 
        bitmap = mbt_ctx_bitmap(sb, TEST_GOAL_GROUP);
        memset(bitmap, 0, sb->s_blocksize);
-       ret = ext4_mb_mark_diskspace_used(ac, NULL, 0);
+       ret = ext4_mb_mark_diskspace_used(ac, NULL);
        KUNIT_ASSERT_EQ(test, ret, 0);
 
        max = EXT4_CLUSTERS_PER_GROUP(sb);
index de4cacb740b33e1e554068e6de0a4c762ded79b3..dd29558ad753b94d826f73b725f826a99e6be75a 100644 (file)
@@ -4186,8 +4186,7 @@ out_err:
  * Returns 0 if success or error code
  */
 static noinline_for_stack int
-ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
-                               handle_t *handle, unsigned int reserv_clstrs)
+ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac, handle_t *handle)
 {
        struct ext4_group_desc *gdp;
        struct ext4_sb_info *sbi;
@@ -4242,13 +4241,6 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
        BUG_ON(changed != ac->ac_b_ex.fe_len);
 #endif
        percpu_counter_sub(&sbi->s_freeclusters_counter, ac->ac_b_ex.fe_len);
-       /*
-        * Now reduce the dirty block count also. Should not go negative
-        */
-       if (!(ac->ac_flags & EXT4_MB_DELALLOC_RESERVED))
-               /* release all the reserved blocks if non delalloc */
-               percpu_counter_sub(&sbi->s_dirtyclusters_counter,
-                                  reserv_clstrs);
 
        return err;
 }
@@ -6333,7 +6325,7 @@ repeat:
                        ext4_mb_pa_put_free(ac);
        }
        if (likely(ac->ac_status == AC_STATUS_FOUND)) {
-               *errp = ext4_mb_mark_diskspace_used(ac, handle, reserv_clstrs);
+               *errp = ext4_mb_mark_diskspace_used(ac, handle);
                if (*errp) {
                        ext4_discard_allocated_blocks(ac);
                        goto errout;
@@ -6364,12 +6356,9 @@ errout:
 out:
        if (inquota && ar->len < inquota)
                dquot_free_block(ar->inode, EXT4_C2B(sbi, inquota - ar->len));
-       if (!ar->len) {
-               if ((ar->flags & EXT4_MB_DELALLOC_RESERVED) == 0)
-                       /* release all the reserved blocks if non delalloc */
-                       percpu_counter_sub(&sbi->s_dirtyclusters_counter,
-                                               reserv_clstrs);
-       }
+       /* release any reserved blocks */
+       if (reserv_clstrs)
+               percpu_counter_sub(&sbi->s_dirtyclusters_counter, reserv_clstrs);
 
        trace_ext4_allocate_blocks(ar, (unsigned long long)block);