* [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats
@ 2026-04-22 1:50 Baolin Liu
2026-04-22 1:50 ` [PATCH v3 v3 1/2] ext4: add blocks_allocated to mb_stats output Baolin Liu
` (2 more replies)
0 siblings, 3 replies; 10+ messages in thread
From: Baolin Liu @ 2026-04-22 1:50 UTC (permalink / raw)
To: tytso, adilger.kernel, ojaswin, ritesh.list, yi.zhang
Cc: linux-ext4, linux-kernel, wangguanyu, liubaolin12138
The series contains two patches:
- add blocks_allocated to /proc/fs/ext4/<dev>/mb_stats
- allow writing 0 to /proc/fs/ext4/<dev>/mb_stats to clear the current
mballoc statistics
Changes since v2:
- Add mb_stats documentation to patch 2
- Add Reviewed-by tags
Baolin Liu (2):
ext4: add blocks_allocated to mb_stats output
ext4: allow clearing mballoc stats through mb_stats
Documentation/admin-guide/ext4.rst | 5 ++++
Documentation/filesystems/proc.rst | 3 +++
fs/ext4/ext4.h | 1 +
fs/ext4/mballoc.c | 31 +++++++++++++++++++++++
fs/ext4/sysfs.c | 40 ++++++++++++++++++++++++++++--
5 files changed, 78 insertions(+), 2 deletions(-)
--
2.51.0
^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v3 v3 1/2] ext4: add blocks_allocated to mb_stats output
2026-04-22 1:50 [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats Baolin Liu
@ 2026-04-22 1:50 ` Baolin Liu
2026-04-24 2:35 ` Baokun Li
2026-04-22 1:50 ` [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats Baolin Liu
2026-04-22 9:55 ` [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats liubaolin
2 siblings, 1 reply; 10+ messages in thread
From: Baolin Liu @ 2026-04-22 1:50 UTC (permalink / raw)
To: tytso, adilger.kernel, ojaswin, ritesh.list, yi.zhang
Cc: linux-ext4, linux-kernel, wangguanyu, liubaolin12138, Baolin Liu,
Andreas Dilger
From: Baolin Liu <liubaolin@kylinos.cn>
Add blocks_allocated to /proc/fs/ext4/<dev>/mb_stats so that the
reported statistics match the mballoc summary printed at unmount time.
Reviewed-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Reviewed-by: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Baolin Liu <liubaolin@kylinos.cn>
---
fs/ext4/mballoc.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 20e9fdaf4301..1e13ef62cb9d 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -3211,6 +3211,8 @@ int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset)
"\tTo enable, please write \"1\" to sysfs file mb_stats.\n");
return 0;
}
+ seq_printf(seq, "\tblocks_allocated: %u\n",
+ atomic_read(&sbi->s_bal_allocated));
seq_printf(seq, "\treqs: %u\n", atomic_read(&sbi->s_bal_reqs));
seq_printf(seq, "\tsuccess: %u\n", atomic_read(&sbi->s_bal_success));
--
2.51.0
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats
2026-04-22 1:50 [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats Baolin Liu
2026-04-22 1:50 ` [PATCH v3 v3 1/2] ext4: add blocks_allocated to mb_stats output Baolin Liu
@ 2026-04-22 1:50 ` Baolin Liu
2026-04-23 16:19 ` Theodore Tso
2026-04-22 9:55 ` [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats liubaolin
2 siblings, 1 reply; 10+ messages in thread
From: Baolin Liu @ 2026-04-22 1:50 UTC (permalink / raw)
To: tytso, adilger.kernel, ojaswin, ritesh.list, yi.zhang
Cc: linux-ext4, linux-kernel, wangguanyu, liubaolin12138, Baolin Liu,
Andreas Dilger
From: Baolin Liu <liubaolin@kylinos.cn>
Make /proc/fs/ext4/<dev>/mb_stats writable and clear the runtime
mballoc statistics when 0 is written.
Update the related documentation accordingly.
Reviewed-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Reviewed-by: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Baolin Liu <liubaolin@kylinos.cn>
---
Documentation/admin-guide/ext4.rst | 5 ++++
Documentation/filesystems/proc.rst | 3 +++
fs/ext4/ext4.h | 1 +
fs/ext4/mballoc.c | 29 ++++++++++++++++++++++
fs/ext4/sysfs.c | 40 ++++++++++++++++++++++++++++--
5 files changed, 76 insertions(+), 2 deletions(-)
diff --git a/Documentation/admin-guide/ext4.rst b/Documentation/admin-guide/ext4.rst
index ac0c709ea9e7..fb3887cd5e44 100644
--- a/Documentation/admin-guide/ext4.rst
+++ b/Documentation/admin-guide/ext4.rst
@@ -436,6 +436,11 @@ Files in /proc/fs/ext4/<devname>
mb_groups
details of multiblock allocator buddy cache of free blocks
+ mb_stats
+ reports runtime statistics from the multiblock allocator
+ (mballoc). Writing 0 to this file clears the current
+ statistics.
+
/sys entries
============
diff --git a/Documentation/filesystems/proc.rst b/Documentation/filesystems/proc.rst
index b0c0d1b45b99..7ce02573a3d9 100644
--- a/Documentation/filesystems/proc.rst
+++ b/Documentation/filesystems/proc.rst
@@ -1634,6 +1634,9 @@ directory are shown in Table 1-12, below.
============== ==========================================================
File Content
mb_groups details of multiblock allocator buddy cache of free blocks
+ mb_stats reports runtime statistics from the multiblock allocator
+ (mballoc). Writing 0 to this file clears the current
+ statistics.
============== ==========================================================
1.9 /proc/consoles
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 293f698b7042..3223e73612ae 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -2994,6 +2994,7 @@ int ext4_fc_record_regions(struct super_block *sb, int ino,
extern const struct seq_operations ext4_mb_seq_groups_ops;
extern const struct seq_operations ext4_mb_seq_structs_summary_ops;
extern int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset);
+extern void ext4_mb_stats_clear(struct ext4_sb_info *sbi);
extern int ext4_mb_init(struct super_block *);
extern void ext4_mb_release(struct super_block *);
extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 1e13ef62cb9d..79ddfa935813 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -4723,6 +4723,35 @@ static void ext4_mb_collect_stats(struct ext4_allocation_context *ac)
trace_ext4_mballoc_prealloc(ac);
}
+void ext4_mb_stats_clear(struct ext4_sb_info *sbi)
+{
+ int i;
+
+ atomic_set(&sbi->s_bal_reqs, 0);
+ atomic_set(&sbi->s_bal_success, 0);
+ atomic_set(&sbi->s_bal_allocated, 0);
+ atomic_set(&sbi->s_bal_groups_scanned, 0);
+
+ for (i = 0; i < EXT4_MB_NUM_CRS; i++) {
+ atomic64_set(&sbi->s_bal_cX_hits[i], 0);
+ atomic64_set(&sbi->s_bal_cX_groups_considered[i], 0);
+ atomic_set(&sbi->s_bal_cX_ex_scanned[i], 0);
+ atomic64_set(&sbi->s_bal_cX_failed[i], 0);
+ }
+
+ atomic_set(&sbi->s_bal_ex_scanned, 0);
+ atomic_set(&sbi->s_bal_goals, 0);
+ atomic_set(&sbi->s_bal_stream_goals, 0);
+ atomic_set(&sbi->s_bal_len_goals, 0);
+ atomic_set(&sbi->s_bal_2orders, 0);
+ atomic_set(&sbi->s_bal_breaks, 0);
+ atomic_set(&sbi->s_mb_lost_chunks, 0);
+ atomic_set(&sbi->s_mb_buddies_generated, 0);
+ atomic64_set(&sbi->s_mb_generation_time, 0);
+ atomic_set(&sbi->s_mb_preallocated, 0);
+ atomic_set(&sbi->s_mb_discarded, 0);
+}
+
/*
* Called on failure; free up any blocks from the inode PA for this
* context. We don't need this for MB_GROUP_PA because we only change
diff --git a/fs/ext4/sysfs.c b/fs/ext4/sysfs.c
index b87d7bdab06a..e90885d470ab 100644
--- a/fs/ext4/sysfs.c
+++ b/fs/ext4/sysfs.c
@@ -52,6 +52,42 @@ typedef enum {
static const char proc_dirname[] = "fs/ext4";
static struct proc_dir_entry *ext4_proc_root;
+static int ext4_mb_stats_open(struct inode *inode, struct file *file)
+{
+ return single_open(file, ext4_seq_mb_stats_show, pde_data(inode));
+}
+
+static ssize_t ext4_mb_stats_write(struct file *file, const char __user *buf,
+ size_t count, loff_t *ppos)
+{
+ struct super_block *sb = pde_data(file_inode(file));
+ char kbuf[2];
+
+ if (count == 0 || count > sizeof(kbuf))
+ return -EINVAL;
+
+ if (copy_from_user(kbuf, buf, count))
+ return -EFAULT;
+
+ if (count == 2) {
+ if (kbuf[0] != '0' || kbuf[1] != '\n')
+ return -EINVAL;
+ } else if (kbuf[0] != '0') {
+ return -EINVAL;
+ }
+
+ ext4_mb_stats_clear(EXT4_SB(sb));
+ return count;
+}
+
+static const struct proc_ops ext4_mb_stats_proc_ops = {
+ .proc_open = ext4_mb_stats_open,
+ .proc_read = seq_read,
+ .proc_lseek = seq_lseek,
+ .proc_release = single_release,
+ .proc_write = ext4_mb_stats_write,
+};
+
struct ext4_attr {
struct attribute attr;
short attr_id;
@@ -630,8 +666,8 @@ int ext4_register_sysfs(struct super_block *sb)
ext4_fc_info_show, sb);
proc_create_seq_data("mb_groups", S_IRUGO, sbi->s_proc,
&ext4_mb_seq_groups_ops, sb);
- proc_create_single_data("mb_stats", 0444, sbi->s_proc,
- ext4_seq_mb_stats_show, sb);
+ proc_create_data("mb_stats", 0644, sbi->s_proc,
+ &ext4_mb_stats_proc_ops, sb);
proc_create_seq_data("mb_structs_summary", 0444, sbi->s_proc,
&ext4_mb_seq_structs_summary_ops, sb);
}
--
2.51.0
^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats
2026-04-22 1:50 [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats Baolin Liu
2026-04-22 1:50 ` [PATCH v3 v3 1/2] ext4: add blocks_allocated to mb_stats output Baolin Liu
2026-04-22 1:50 ` [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats Baolin Liu
@ 2026-04-22 9:55 ` liubaolin
2 siblings, 0 replies; 10+ messages in thread
From: liubaolin @ 2026-04-22 9:55 UTC (permalink / raw)
To: tytso, adilger.kernel, ojaswin, ritesh.list, yi.zhang
Cc: linux-ext4, linux-kernel, wangguanyu
在 2026/4/22 9:50, Baolin Liu 写道:
> The series contains two patches:
> - add blocks_allocated to /proc/fs/ext4/<dev>/mb_stats
> - allow writing 0 to /proc/fs/ext4/<dev>/mb_stats to clear the current
> mballoc statistics
>
> Changes since v2:
> - Add mb_stats documentation to patch 2
> - Add Reviewed-by tags
>
> Baolin Liu (2):
> ext4: add blocks_allocated to mb_stats output
> ext4: allow clearing mballoc stats through mb_stats
>
> Documentation/admin-guide/ext4.rst | 5 ++++
> Documentation/filesystems/proc.rst | 3 +++
> fs/ext4/ext4.h | 1 +
> fs/ext4/mballoc.c | 31 +++++++++++++++++++++++
> fs/ext4/sysfs.c | 40 ++++++++++++++++++++++++++++--
> 5 files changed, 78 insertions(+), 2 deletions(-)
>
Dear All,
This commit adds the description of the ext4 proc parameter mb_stats
to the corresponding documentation.
I noticed that the documentation also lacks descriptions for the
es_shrinker_info, fc_info, mb_structs_summary, and options parameters.
However, these parameters are irrelevant to the work done in this
series of patches.
I will commit a separate patch later to explain the other parameters.
This time, I only explain mb_stats.
Thanks,
Baolin
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats
2026-04-22 1:50 ` [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats Baolin Liu
@ 2026-04-23 16:19 ` Theodore Tso
2026-04-24 3:12 ` Baokun Li
2026-04-24 8:09 ` liubaolin
0 siblings, 2 replies; 10+ messages in thread
From: Theodore Tso @ 2026-04-23 16:19 UTC (permalink / raw)
To: Baolin Liu
Cc: adilger.kernel, ojaswin, ritesh.list, yi.zhang, linux-ext4,
linux-kernel, wangguanyu, Baolin Liu, Andreas Dilger
On Wed, Apr 22, 2026 at 09:50:25AM +0800, Baolin Liu wrote:
> From: Baolin Liu <liubaolin@kylinos.cn>
>
> Make /proc/fs/ext4/<dev>/mb_stats writable and clear the runtime
> mballoc statistics when 0 is written.
At the moment to enable mb_stats the system administrator needs to
write "1" to /sys/fs/ext4/<dev>/mb_stats, and writing "0" to the sysfs
file will pauce the statistics colleciton (but not clear the
statistics). Adding a way to clear the statistics by writing to the
procfs file might be confusing to users.
So.... as a suggestion, if you're adding to the ability to write to
/proc/fs/.../mb_stats, what if we make things work by
* Write 1 to /proc/fs/.../mb_stats to enable statistics collection
* Write 0 to /proc/fs/.../mb_stats to disable statistics collection
* Write -1 to /proc/fs/.../mb_stats to clear statistics counters
And then deprecate the /sys/fs/.../mb_stats variable (but we probably
won't be able to remove it for at least a year or two).
- Ted
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v3 v3 1/2] ext4: add blocks_allocated to mb_stats output
2026-04-22 1:50 ` [PATCH v3 v3 1/2] ext4: add blocks_allocated to mb_stats output Baolin Liu
@ 2026-04-24 2:35 ` Baokun Li
0 siblings, 0 replies; 10+ messages in thread
From: Baokun Li @ 2026-04-24 2:35 UTC (permalink / raw)
To: Baolin Liu
Cc: tytso, adilger.kernel, ojaswin, ritesh.list, yi.zhang, linux-ext4,
linux-kernel, wangguanyu, Baolin Liu, Andreas Dilger
On 2026/4/22 09:50, Baolin Liu wrote:
> From: Baolin Liu <liubaolin@kylinos.cn>
>
> Add blocks_allocated to /proc/fs/ext4/<dev>/mb_stats so that the
> reported statistics match the mballoc summary printed at unmount time.
>
> Reviewed-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
> Reviewed-by: Andreas Dilger <adilger@dilger.ca>
> Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
> Reviewed-by: Zhang Yi <yi.zhang@huawei.com>
> Signed-off-by: Baolin Liu <liubaolin@kylinos.cn>
Looks good, feel free to add:
Reviewed-by: Baokun Li <libaokun@linux.alibaba.com>
> ---
> fs/ext4/mballoc.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
> index 20e9fdaf4301..1e13ef62cb9d 100644
> --- a/fs/ext4/mballoc.c
> +++ b/fs/ext4/mballoc.c
> @@ -3211,6 +3211,8 @@ int ext4_seq_mb_stats_show(struct seq_file *seq, void *offset)
> "\tTo enable, please write \"1\" to sysfs file mb_stats.\n");
> return 0;
> }
> + seq_printf(seq, "\tblocks_allocated: %u\n",
> + atomic_read(&sbi->s_bal_allocated));
> seq_printf(seq, "\treqs: %u\n", atomic_read(&sbi->s_bal_reqs));
> seq_printf(seq, "\tsuccess: %u\n", atomic_read(&sbi->s_bal_success));
>
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats
2026-04-23 16:19 ` Theodore Tso
@ 2026-04-24 3:12 ` Baokun Li
2026-04-24 8:09 ` liubaolin
1 sibling, 0 replies; 10+ messages in thread
From: Baokun Li @ 2026-04-24 3:12 UTC (permalink / raw)
To: Theodore Tso, Baolin Liu
Cc: adilger.kernel, ojaswin, ritesh.list, yi.zhang, linux-ext4,
linux-kernel, wangguanyu, Baolin Liu, Andreas Dilger
On 2026/4/24 00:19, Theodore Tso wrote:
> On Wed, Apr 22, 2026 at 09:50:25AM +0800, Baolin Liu wrote:
>> From: Baolin Liu <liubaolin@kylinos.cn>
>>
>> Make /proc/fs/ext4/<dev>/mb_stats writable and clear the runtime
>> mballoc statistics when 0 is written.
> At the moment to enable mb_stats the system administrator needs to
> write "1" to /sys/fs/ext4/<dev>/mb_stats, and writing "0" to the sysfs
> file will pauce the statistics colleciton (but not clear the
> statistics). Adding a way to clear the statistics by writing to the
> procfs file might be confusing to users.
>
> So.... as a suggestion, if you're adding to the ability to write to
> /proc/fs/.../mb_stats, what if we make things work by
>
> * Write 1 to /proc/fs/.../mb_stats to enable statistics collection
> * Write 0 to /proc/fs/.../mb_stats to disable statistics collection
> * Write -1 to /proc/fs/.../mb_stats to clear statistics counters
>
> And then deprecate the /sys/fs/.../mb_stats variable (but we probably
> won't be able to remove it for at least a year or two).
I like this idea. Consolidating everything into /proc/fs/.../mb_stats
and using -1 to reset the counters is much cleaner than the current
split approach.
Cheers,
Baokun
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats
2026-04-23 16:19 ` Theodore Tso
2026-04-24 3:12 ` Baokun Li
@ 2026-04-24 8:09 ` liubaolin
2026-04-24 9:34 ` Baokun Li
2026-04-24 12:07 ` Theodore Tso
1 sibling, 2 replies; 10+ messages in thread
From: liubaolin @ 2026-04-24 8:09 UTC (permalink / raw)
To: Theodore Tso, libaokun
Cc: adilger.kernel, ojaswin, ritesh.list, yi.zhang, linux-ext4,
linux-kernel, wangguanyu, Baolin Liu, Andreas Dilger
在 2026/4/24 0:19, Theodore Tso 写道:
> On Wed, Apr 22, 2026 at 09:50:25AM +0800, Baolin Liu wrote:
>> From: Baolin Liu <liubaolin@kylinos.cn>
>>
>> Make /proc/fs/ext4/<dev>/mb_stats writable and clear the runtime
>> mballoc statistics when 0 is written.
>
> At the moment to enable mb_stats the system administrator needs to
> write "1" to /sys/fs/ext4/<dev>/mb_stats, and writing "0" to the sysfs
> file will pauce the statistics colleciton (but not clear the
> statistics). Adding a way to clear the statistics by writing to the
> procfs file might be confusing to users.
>
> So.... as a suggestion, if you're adding to the ability to write to
> /proc/fs/.../mb_stats, what if we make things work by
>
> * Write 1 to /proc/fs/.../mb_stats to enable statistics collection
> * Write 0 to /proc/fs/.../mb_stats to disable statistics collection
> * Write -1 to /proc/fs/.../mb_stats to clear statistics counters
>
> And then deprecate the /sys/fs/.../mb_stats variable (but we probably
> won't be able to remove it for at least a year or two).
>
> - Ted
Dear Ted, Baokun,
Thank you for your review and suggestions.
Since you mentioned that /sys/fs/.../mb_stats cannot be deleted in
the short term,
I plan to modify and submit a v4 patch according to the following
strategy.
1. Change `/proc/fs/.../mb_stats` to read-write mode.
* Read `/proc/fs/.../mb_stats` to show statistics counters.
* Write 0 to `/proc/fs/.../mb_stats` to disable statistics collection.
* Write 1 to `/proc/fs/.../mb_stats` to enable statistics collection.
* Write 2 to `/proc/fs/.../mb_stats` to clear statistics counters.
2. Do not delete the `/sys/fs/.../mb_stats` node for now; implement
the same write control logic.
* Write 0 to `/sys/fs/.../mb_stats` to disable statistics collection.
* Write 1 to `/sys/fs/.../mb_stats` to enable statistics collection.
* Write 2 to `/sys/fs/.../mb_stats` to clear statistics counters.
Delete `/sys/fs/.../mb_stats` later when it is possible to delete it.
3. Modify the relevant documentation for `mb_stats`.
Documentation/ABI/testing/sysfs-fs-ext4
Documentation/admin-guide/ext4.rst
Documentation/filesystems/proc.rst
Compared to your suggestion, I recommend using the value 2 for the
clear operation because s_mb_stats is an unsigned int variable, and
using -1 requires changing the variable type.
I suggest avoiding changing the s_mb_stats variable type unless
absolutely necessary.
Do you think this modification is appropriate?
If there are no problems, I will start modifying the code and submit
the v4 patch as soon as possible.
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats
2026-04-24 8:09 ` liubaolin
@ 2026-04-24 9:34 ` Baokun Li
2026-04-24 12:07 ` Theodore Tso
1 sibling, 0 replies; 10+ messages in thread
From: Baokun Li @ 2026-04-24 9:34 UTC (permalink / raw)
To: liubaolin, Theodore Tso
Cc: adilger.kernel, ojaswin, ritesh.list, yi.zhang, linux-ext4,
linux-kernel, wangguanyu, Baolin Liu, Andreas Dilger
On 2026/4/24 16:09, liubaolin wrote:
>
>
> 在 2026/4/24 0:19, Theodore Tso 写道:
>> On Wed, Apr 22, 2026 at 09:50:25AM +0800, Baolin Liu wrote:
>>> From: Baolin Liu <liubaolin@kylinos.cn>
>>>
>>> Make /proc/fs/ext4/<dev>/mb_stats writable and clear the runtime
>>> mballoc statistics when 0 is written.
>>
>> At the moment to enable mb_stats the system administrator needs to
>> write "1" to /sys/fs/ext4/<dev>/mb_stats, and writing "0" to the sysfs
>> file will pauce the statistics colleciton (but not clear the
>> statistics). Adding a way to clear the statistics by writing to the
>> procfs file might be confusing to users.
>>
>> So.... as a suggestion, if you're adding to the ability to write to
>> /proc/fs/.../mb_stats, what if we make things work by
>>
>> * Write 1 to /proc/fs/.../mb_stats to enable statistics collection
>> * Write 0 to /proc/fs/.../mb_stats to disable statistics collection
>> * Write -1 to /proc/fs/.../mb_stats to clear statistics counters
>>
>> And then deprecate the /sys/fs/.../mb_stats variable (but we probably
>> won't be able to remove it for at least a year or two).
>>
>> - Ted
> Dear Ted, Baokun,
> Thank you for your review and suggestions.
> Since you mentioned that /sys/fs/.../mb_stats cannot be deleted in
> the short term,
> I plan to modify and submit a v4 patch according to the following
> strategy.
>
> 1. Change `/proc/fs/.../mb_stats` to read-write mode.
> * Read `/proc/fs/.../mb_stats` to show statistics counters.
> * Write 0 to `/proc/fs/.../mb_stats` to disable statistics
> collection.
> * Write 1 to `/proc/fs/.../mb_stats` to enable statistics collection.
> * Write 2 to `/proc/fs/.../mb_stats` to clear statistics counters.
>
> 2. Do not delete the `/sys/fs/.../mb_stats` node for now; implement
> the same write control logic.
> * Write 0 to `/sys/fs/.../mb_stats` to disable statistics collection.
> * Write 1 to `/sys/fs/.../mb_stats` to enable statistics collection.
> * Write 2 to `/sys/fs/.../mb_stats` to clear statistics counters.
>
> Delete `/sys/fs/.../mb_stats` later when it is possible to delete it.
>
> 3. Modify the relevant documentation for `mb_stats`.
> Documentation/ABI/testing/sysfs-fs-ext4
> Documentation/admin-guide/ext4.rst
> Documentation/filesystems/proc.rst
>
> Compared to your suggestion, I recommend using the value 2 for the
> clear operation because s_mb_stats is an unsigned int variable, and
> using -1 requires changing the variable type.
> I suggest avoiding changing the s_mb_stats variable type unless
> absolutely necessary.
>
> Do you think this modification is appropriate?
> If there are no problems, I will start modifying the code and
> submit the v4 patch as soon as possible.
For the clear command, we only handle it without storing it, so s_mb_stats
remains unchanged and still stores only 0 and non-zero values to represent
disabled and enabled, respectively. Otherwise, you will have to deal with
a large number of s_mb_stats checks
That means the /sys/fs/.../mb_stats interface does not need to support
clearing, but it might make sense to add a deprecation warning there.
Then in `/proc/fs/.../mb_stats`, writing 0 or a positive number passes
it to s_mb_stats, writing -1 performs a reset, and other negative values
return -EINVAL.
Cheers,
Baokun
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats
2026-04-24 8:09 ` liubaolin
2026-04-24 9:34 ` Baokun Li
@ 2026-04-24 12:07 ` Theodore Tso
1 sibling, 0 replies; 10+ messages in thread
From: Theodore Tso @ 2026-04-24 12:07 UTC (permalink / raw)
To: liubaolin
Cc: libaokun, adilger.kernel, ojaswin, ritesh.list, yi.zhang,
linux-ext4, linux-kernel, wangguanyu, Baolin Liu, Andreas Dilger
On Fri, Apr 24, 2026 at 04:09:31PM +0800, liubaolin wrote:
>
> 2. Do not delete the `/sys/fs/.../mb_stats` node for now; implement the
> same write control logic.
> * Write 0 to `/sys/fs/.../mb_stats` to disable statistics collection.
> * Write 1 to `/sys/fs/.../mb_stats` to enable statistics collection.
> * Write 2 to `/sys/fs/.../mb_stats` to clear statistics counters.
We could do that, but note that currently writing to
/sys/fs/.../mb_stats just sets an unsigned integer in
EXT4(sb)->s_mb_stats. There is no ext4-specific function that runs
when /sys/fs/.../mb_stats is updated.
So either you have to add some check in fs/ext4/mballoc.c which gets
called every single time a block allocation happens --- and consider
the race condition where two CPU's are checking s_mb_stats at the same
time, and the desireability of adding a spinlock that would need to be
taken every single time a block allocation happens ---- or you have
add an ext4-specific function in fs/ext4/sysfs.c.
> Compared to your suggestion, I recommend using the value 2 for the clear
> operation because s_mb_stats is an unsigned int variable, and using -1
> requires changing the variable type.
Well, since you have introduced an ext4-specific function which gets
called when writing to the procfs file, that function can clear the
statistics counter when -1 is written to the file --- and then set
s_mbi_stats to 1.
Cheers,
- Ted
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2026-04-24 12:08 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-04-22 1:50 [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats Baolin Liu
2026-04-22 1:50 ` [PATCH v3 v3 1/2] ext4: add blocks_allocated to mb_stats output Baolin Liu
2026-04-24 2:35 ` Baokun Li
2026-04-22 1:50 ` [PATCH v3 v3 2/2] ext4: allow clearing mballoc stats through mb_stats Baolin Liu
2026-04-23 16:19 ` Theodore Tso
2026-04-24 3:12 ` Baokun Li
2026-04-24 8:09 ` liubaolin
2026-04-24 9:34 ` Baokun Li
2026-04-24 12:07 ` Theodore Tso
2026-04-22 9:55 ` [PATCH v3 v3 0/2] add blocks_allocated to mb_stats and clear mb_stats liubaolin
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox