Re: [PATCH] mm/ksm: prepare to remove the redundant ksm_merging_pages in procfs

Re: [PATCH] mm/ksm: prepare to remove the redundant ksm_merging_pages in procfs

[David Hildenbrand]

On 06.07.23 11:49, Nanyong Sun wrote:
> Since the ksm_merging_pages information already included in
> /proc/<pid>/ksm_stat, we could remove /proc/<pid>/ksm_merging_pages
> to make the directory more clean, and can save a little bit resources.
> 
> To delete this interface more smoothly and avoid userspace break,
> retain this interface temporarily and modify its function to hint
> users to use ksm_stat instead.
> 
> Signed-off-by: Nanyong Sun <sunnanyong@huawei.com>
> ---
>   fs/proc/base.c | 9 +--------
>   1 file changed, 1 insertion(+), 8 deletions(-)
> 
> diff --git a/fs/proc/base.c b/fs/proc/base.c
> index eb2e498e3b8d..d080c58cbe6c 100644
> --- a/fs/proc/base.c
> +++ b/fs/proc/base.c
> @@ -3189,14 +3189,7 @@ static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
>   static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns,
>   				struct pid *pid, struct task_struct *task)
>   {
> -	struct mm_struct *mm;
> -
> -	mm = get_task_mm(task);
> -	if (mm) {
> -		seq_printf(m, "%lu\n", mm->ksm_merging_pages);
> -		mmput(mm);
> -	}
> -
> +	seq_puts(m, "please use /proc/<pid>/ksm_stat instead\n");
>   	return 0;
>   }
>   static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,


Why do we care so much about removing 15 simple LOC? That change here 
will already mess with user space.

Sorry, but IMHO it's all not worth the churn.

-- 
Cheers,

David / dhildenb

[PATCH] mm/ksm: prepare to remove the redundant ksm_merging_pages in procfs

[Nanyong Sun]

Since the ksm_merging_pages information already included in
/proc/<pid>/ksm_stat, we could remove /proc/<pid>/ksm_merging_pages
to make the directory more clean, and can save a little bit resources.

To delete this interface more smoothly and avoid userspace break,
retain this interface temporarily and modify its function to hint
users to use ksm_stat instead.

Signed-off-by: Nanyong Sun <sunnanyong@huawei.com>
---
 fs/proc/base.c | 9 +--------
 1 file changed, 1 insertion(+), 8 deletions(-)

diff --git a/fs/proc/base.c b/fs/proc/base.c
index eb2e498e3b8d..d080c58cbe6c 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -3189,14 +3189,7 @@ static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
 static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns,
 				struct pid *pid, struct task_struct *task)
 {
-	struct mm_struct *mm;
-
-	mm = get_task_mm(task);
-	if (mm) {
-		seq_printf(m, "%lu\n", mm->ksm_merging_pages);
-		mmput(mm);
-	}
-
+	seq_puts(m, "please use /proc/<pid>/ksm_stat instead\n");
 	return 0;
 }
 static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
-- 
2.25.1

Re: [PATCH] mm/ksm: prepare to remove the redundant ksm_merging_pages in procfs

[Nanyong Sun]

On 2023/7/6 16:55, David Hildenbrand wrote:

> On 06.07.23 11:49, Nanyong Sun wrote:
>> Since the ksm_merging_pages information already included in
>> /proc/<pid>/ksm_stat, we could remove /proc/<pid>/ksm_merging_pages
>> to make the directory more clean, and can save a little bit resources.
>>
>> To delete this interface more smoothly and avoid userspace break,
>> retain this interface temporarily and modify its function to hint
>> users to use ksm_stat instead.
>>
>> Signed-off-by: Nanyong Sun <sunnanyong@huawei.com>
>> ---
>>   fs/proc/base.c | 9 +--------
>>   1 file changed, 1 insertion(+), 8 deletions(-)
>>
>> diff --git a/fs/proc/base.c b/fs/proc/base.c
>> index eb2e498e3b8d..d080c58cbe6c 100644
>> --- a/fs/proc/base.c
>> +++ b/fs/proc/base.c
>> @@ -3189,14 +3189,7 @@ static int proc_pid_patch_state(struct 
>> seq_file *m, struct pid_namespace *ns,
>>   static int proc_pid_ksm_merging_pages(struct seq_file *m, struct 
>> pid_namespace *ns,
>>                   struct pid *pid, struct task_struct *task)
>>   {
>> -    struct mm_struct *mm;
>> -
>> -    mm = get_task_mm(task);
>> -    if (mm) {
>> -        seq_printf(m, "%lu\n", mm->ksm_merging_pages);
>> -        mmput(mm);
>> -    }
>> -
>> +    seq_puts(m, "please use /proc/<pid>/ksm_stat instead\n");
>>       return 0;
>>   }
>>   static int proc_pid_ksm_stat(struct seq_file *m, struct 
>> pid_namespace *ns,
>
>
> Why do we care so much about removing 15 simple LOC? That change here 
> will already mess with user space.
>
> Sorry, but IMHO it's all not worth the churn.
>
We do not pay attention to these 15 LOC. We pay more attention to the 
redundant interface under procfs.

This interface is used by developers and has a short history. Therefore, 
changing it now has no big impact.

Re: [PATCH] mm/ksm: prepare to remove the redundant ksm_merging_pages in procfs

[David Hildenbrand]

On 06.07.23 13:59, Nanyong Sun wrote:
> On 2023/7/6 16:55, David Hildenbrand wrote:
> 
>> On 06.07.23 11:49, Nanyong Sun wrote:
>>> Since the ksm_merging_pages information already included in
>>> /proc/<pid>/ksm_stat, we could remove /proc/<pid>/ksm_merging_pages
>>> to make the directory more clean, and can save a little bit resources.
>>>
>>> To delete this interface more smoothly and avoid userspace break,
>>> retain this interface temporarily and modify its function to hint
>>> users to use ksm_stat instead.
>>>
>>> Signed-off-by: Nanyong Sun <sunnanyong@huawei.com>
>>> ---
>>>    fs/proc/base.c | 9 +--------
>>>    1 file changed, 1 insertion(+), 8 deletions(-)
>>>
>>> diff --git a/fs/proc/base.c b/fs/proc/base.c
>>> index eb2e498e3b8d..d080c58cbe6c 100644
>>> --- a/fs/proc/base.c
>>> +++ b/fs/proc/base.c
>>> @@ -3189,14 +3189,7 @@ static int proc_pid_patch_state(struct
>>> seq_file *m, struct pid_namespace *ns,
>>>    static int proc_pid_ksm_merging_pages(struct seq_file *m, struct
>>> pid_namespace *ns,
>>>                    struct pid *pid, struct task_struct *task)
>>>    {
>>> -    struct mm_struct *mm;
>>> -
>>> -    mm = get_task_mm(task);
>>> -    if (mm) {
>>> -        seq_printf(m, "%lu\n", mm->ksm_merging_pages);
>>> -        mmput(mm);
>>> -    }
>>> -
>>> +    seq_puts(m, "please use /proc/<pid>/ksm_stat instead\n");
>>>        return 0;
>>>    }
>>>    static int proc_pid_ksm_stat(struct seq_file *m, struct
>>> pid_namespace *ns,
>>
>>
>> Why do we care so much about removing 15 simple LOC? That change here
>> will already mess with user space.
>>
>> Sorry, but IMHO it's all not worth the churn.
>>
> We do not pay attention to these 15 LOC. We pay more attention to the
> redundant interface under procfs.
> 
> This interface is used by developers and has a short history. Therefore,
> changing it now has no big impact.
> 

Again, already 1 year old. And returning garbage instead of ripping it 
out might be even nastier.

-- 
Cheers,

David / dhildenb