On Wed, Jan 5, 2022 at 6:45 PM Xin Yin <yinxin.x@xxxxxxxxxxxxx> wrote:
>
> during fast commit replay procedure, we clear inode blocks bitmap in
> ext4_ext_clear_bb(), this may cause ext4_mb_new_blocks_simple() allocate
> blocks still in use.
>
> make ext4_fc_record_regions() also record physical disk regions used by
> inodes during replay procedure. Then ext4_mb_new_blocks_simple() can
> excludes these blocks in use.
>
> Signed-off-by: Xin Yin <yinxin.x@xxxxxxxxxxxxx>
> ---
> fs/ext4/ext4.h | 2 ++
> fs/ext4/extents.c | 4 ++++
> fs/ext4/fast_commit.c | 11 ++++++++---
> 3 files changed, 14 insertions(+), 3 deletions(-)
>
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 82fa51d6f145..7b0686758691 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -2932,6 +2932,8 @@ bool ext4_fc_replay_check_excluded(struct super_block *sb, ext4_fsblk_t block);
> void ext4_fc_replay_cleanup(struct super_block *sb);
> int ext4_fc_commit(journal_t *journal, tid_t commit_tid);
> int __init ext4_fc_init_dentry_cache(void);
> +int ext4_fc_record_regions(struct super_block *sb, int ino,
> + ext4_lblk_t lblk, ext4_fsblk_t pblk, int len, int replay);
>
> /* mballoc.c */
> extern const struct seq_operations ext4_mb_seq_groups_ops;
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index c3e76a5de661..9b6c76629c93 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -6096,11 +6096,15 @@ int ext4_ext_clear_bb(struct inode *inode)
>
> ext4_mb_mark_bb(inode->i_sb,
> path[j].p_block, 1, 0);
> + ext4_fc_record_regions(inode->i_sb, inode->i_ino,
> + 0, path[j].p_block, 1, 1);
> }
> ext4_ext_drop_refs(path);
> kfree(path);
> }
> ext4_mb_mark_bb(inode->i_sb, map.m_pblk, map.m_len, 0);
> + ext4_fc_record_regions(inode->i_sb, inode->i_ino,
> + map.m_lblk, map.m_pblk, map.m_len, 1);
> }
> cur = cur + map.m_len;
> }
> diff --git a/fs/ext4/fast_commit.c b/fs/ext4/fast_commit.c
> index 23d13983a281..f0cd20f5fe5e 100644
> --- a/fs/ext4/fast_commit.c
> +++ b/fs/ext4/fast_commit.c
> @@ -1567,13 +1567,15 @@ static int ext4_fc_replay_create(struct super_block *sb, struct ext4_fc_tl *tl,
> * Record physical disk regions which are in use as per fast commit area. Our
> * simple replay phase allocator excludes these regions from allocation.
> */
> -static int ext4_fc_record_regions(struct super_block *sb, int ino,
> - ext4_lblk_t lblk, ext4_fsblk_t pblk, int len)
> +int ext4_fc_record_regions(struct super_block *sb, int ino,
> + ext4_lblk_t lblk, ext4_fsblk_t pblk, int len, int replay)
Can you explain a bit why this replay parameter is needed here? This
function simply reallocs the regions array if it doesn't have enough
space. I am not sure why we need to change that behavior.
> {
> struct ext4_fc_replay_state *state;
> struct ext4_fc_alloc_region *region;
>
> state = &EXT4_SB(sb)->s_fc_replay_state;
> + if (replay && state->fc_regions_used != state->fc_regions_valid)
> + state->fc_regions_used = state->fc_regions_valid;
> if (state->fc_regions_used == state->fc_regions_size) {
> state->fc_regions_size +=
> EXT4_FC_REPLAY_REALLOC_INCREMENT;
> @@ -1591,6 +1593,9 @@ static int ext4_fc_record_regions(struct super_block *sb, int ino,
> region->pblk = pblk;
> region->len = len;
>
> + if (replay)
> + state->fc_regions_valid++;
> +
> return 0;
> }
>
> @@ -1938,7 +1943,7 @@ static int ext4_fc_replay_scan(journal_t *journal,
> ret = ext4_fc_record_regions(sb,
> le32_to_cpu(ext.fc_ino),
> le32_to_cpu(ex->ee_block), ext4_ext_pblock(ex),
> - ext4_ext_get_actual_len(ex));
> + ext4_ext_get_actual_len(ex), 0);
> if (ret < 0)
> break;
> ret = JBD2_FC_REPLAY_CONTINUE;
> --
> 2.20.1
>
