fs/ext4/ext4.h | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-)
From: Jonas Zhou <jonaszhou@zhaoxin.com>
In the tested x86-64 layout, s_journal and s_ext4_flags share a cache
line with s_orphan_lock and s_orphan. Legacy orphan-list updates
invalidate the same line used by unrelated journal and inode paths.
Move s_orphan_lock, s_orphan and s_orphan_info together below
s_journal_triggers. This separates orphan-list writes from journal
and flag readers while keeping the related fields in logical groups.
Add a comment identifying the orphan tracking group and preserve the
existing member comments.
On a system with two Xeon Silver 4208 processors and a filesystem
without orphan_file enabled, repeated tests against v7.3-rc2 recorded
mean throughput increases of 11.06% for stress-ng dnotify and 35.93%
for unlink.
Tests with orphan_file enabled showed no clear throughput change
within the observed boot-to-boot variation.
perf c2c shows remote HITM samples on the original shared line, but
none on the separate journal line after the change. Local HITM samples
and orphan-lock contention remain.
Suggested-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/all/575wyoks3ldokhrg6v43rxwsegaecod4bgrpbii6jeczy7czbi@a3yq26u52orp/
Signed-off-by: Jonas Zhou <jonaszhou@zhaoxin.com>
---
Changes since v2:
- Follow Jan's suggestion to move the orphan tracking fields together
below s_journal_triggers, instead of exchanging journal state with
the error-reporting and lazy-init fields.
- Add an Orphan inode tracking comment.
- Rebase and retest against v7.3-rc2.
v2:
https://lore.kernel.org/all/a33183c1c24ea47a98afdd4756b49365593a2762.1788779199.git.jonaszhou@zhaoxin.com/
v1:
https://lore.kernel.org/all/20260903021836.962624-1-jonaszhou-oc@zhaoxin.com/
fs/ext4/ext4.h | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 724a27e8be61..166910914973 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1644,10 +1644,6 @@ struct ext4_sb_info {
/* Journaling */
struct journal_s *s_journal;
unsigned long s_ext4_flags; /* Ext4 superblock flags */
- struct mutex s_orphan_lock; /* Protects on disk list changes */
- struct list_head s_orphan; /* List of orphaned inodes in on disk
- list */
- struct ext4_orphan_info s_orphan_info;
unsigned long s_commit_interval;
u32 s_max_batch_time;
u32 s_min_batch_time;
@@ -1786,6 +1782,12 @@ struct ext4_sb_info {
/* Journal triggers for checksum computation */
struct ext4_journal_trigger s_journal_triggers[EXT4_JOURNAL_TRIGGER_COUNT];
+ /* Orphan inode tracking */
+ struct mutex s_orphan_lock; /* Protects on disk list changes */
+ struct list_head s_orphan; /* List of orphaned inodes in on disk
+ list */
+ struct ext4_orphan_info s_orphan_info;
+
/* Ratelimit ext4 messages. */
struct ratelimit_state s_err_ratelimit_state;
struct ratelimit_state s_warning_ratelimit_state;
base-commit: df2908090cda368b01ff43709f51890076c56157
--
2.43.0
On Fri 11-09-26 11:31:13, JonasZhou-oc wrote:
> From: Jonas Zhou <jonaszhou@zhaoxin.com>
>
> In the tested x86-64 layout, s_journal and s_ext4_flags share a cache
> line with s_orphan_lock and s_orphan. Legacy orphan-list updates
> invalidate the same line used by unrelated journal and inode paths.
>
> Move s_orphan_lock, s_orphan and s_orphan_info together below
> s_journal_triggers. This separates orphan-list writes from journal
> and flag readers while keeping the related fields in logical groups.
> Add a comment identifying the orphan tracking group and preserve the
> existing member comments.
>
> On a system with two Xeon Silver 4208 processors and a filesystem
> without orphan_file enabled, repeated tests against v7.3-rc2 recorded
> mean throughput increases of 11.06% for stress-ng dnotify and 35.93%
> for unlink.
>
> Tests with orphan_file enabled showed no clear throughput change
> within the observed boot-to-boot variation.
>
> perf c2c shows remote HITM samples on the original shared line, but
> none on the separate journal line after the change. Local HITM samples
> and orphan-lock contention remain.
>
> Suggested-by: Jan Kara <jack@suse.cz>
> Link: https://lore.kernel.org/all/575wyoks3ldokhrg6v43rxwsegaecod4bgrpbii6jeczy7czbi@a3yq26u52orp/
> Signed-off-by: Jonas Zhou <jonaszhou@zhaoxin.com>
Looks good. Feel free to add:
Reviewed-by: Jan Kara <jack@suse.cz>
Honza
> ---
> Changes since v2:
> - Follow Jan's suggestion to move the orphan tracking fields together
> below s_journal_triggers, instead of exchanging journal state with
> the error-reporting and lazy-init fields.
> - Add an Orphan inode tracking comment.
> - Rebase and retest against v7.3-rc2.
>
> v2:
> https://lore.kernel.org/all/a33183c1c24ea47a98afdd4756b49365593a2762.1788779199.git.jonaszhou@zhaoxin.com/
> v1:
> https://lore.kernel.org/all/20260903021836.962624-1-jonaszhou-oc@zhaoxin.com/
>
> fs/ext4/ext4.h | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 724a27e8be61..166910914973 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -1644,10 +1644,6 @@ struct ext4_sb_info {
> /* Journaling */
> struct journal_s *s_journal;
> unsigned long s_ext4_flags; /* Ext4 superblock flags */
> - struct mutex s_orphan_lock; /* Protects on disk list changes */
> - struct list_head s_orphan; /* List of orphaned inodes in on disk
> - list */
> - struct ext4_orphan_info s_orphan_info;
> unsigned long s_commit_interval;
> u32 s_max_batch_time;
> u32 s_min_batch_time;
> @@ -1786,6 +1782,12 @@ struct ext4_sb_info {
> /* Journal triggers for checksum computation */
> struct ext4_journal_trigger s_journal_triggers[EXT4_JOURNAL_TRIGGER_COUNT];
>
> + /* Orphan inode tracking */
> + struct mutex s_orphan_lock; /* Protects on disk list changes */
> + struct list_head s_orphan; /* List of orphaned inodes in on disk
> + list */
> + struct ext4_orphan_info s_orphan_info;
> +
> /* Ratelimit ext4 messages. */
> struct ratelimit_state s_err_ratelimit_state;
> struct ratelimit_state s_warning_ratelimit_state;
>
> base-commit: df2908090cda368b01ff43709f51890076c56157
> --
> 2.43.0
>
>
--
Jan Kara <jack@suse.com>
SUSE Labs, CR
© 2016 - 2026 Red Hat, Inc.