On Sat, Mar 04, 2023 at 01:21:01AM +0800, Kemeng Shi wrote:
> We need to set ac_g_ex to notify the goal start used in
> ext4_mb_find_by_goal. Set ac_g_ex instead of ac_f_ex in
> ext4_mb_normalize_request.
> Besides we should assure goal start is in range [first_data_block,
> blocks_count) as ext4_mb_initialize_context does.
>
> Signed-off-by: Kemeng Shi <shikemeng@xxxxxxxxxxxxxxx>
> Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@xxxxxxxxx>
> ---
> fs/ext4/mballoc.c | 15 +++++++++------
> 1 file changed, 9 insertions(+), 6 deletions(-)
>
> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
> index 5b2ae37a8b80..36cd545f5ab4 100644
> --- a/fs/ext4/mballoc.c
> +++ b/fs/ext4/mballoc.c
> @@ -3993,6 +3993,7 @@ ext4_mb_normalize_request(struct ext4_allocation_context *ac,
> struct ext4_allocation_request *ar)
> {
> struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb);
> + struct ext4_super_block *es = sbi->s_es;
> int bsbits, max;
> ext4_lblk_t end;
> loff_t size, start_off;
> @@ -4188,18 +4189,20 @@ ext4_mb_normalize_request(struct ext4_allocation_context *ac,
> ac->ac_g_ex.fe_len = EXT4_NUM_B2C(sbi, size);
>
> /* define goal start in order to merge */
> - if (ar->pright && (ar->lright == (start + size))) {
> + if (ar->pright && (ar->lright == (start + size)) &&
> + ar->pright - size >= le32_to_cpu(es->s_first_data_block)) {
> /* merge to the right */
I had to ammend this commit to add this check:
/* define goal start in order to merge */
if (ar->pright && (ar->lright == (start + size)) &&
+ ar->pright >= size &&
ar->pright - size >= le32_to_cpu(es->s_first_data_block)) {
Without this check, it's possible for ar->pright - size to go negative
(well, underflow since it's an unsigned value). This will later
trigger a BUG_ON, which was easily reproduced via:
kvm-xfstests -c ext4/ext3conv generic/231
Cheers,
- Ted
