[v1] fs/proc: do_task_stat: use __for_each_thread()

[RESEND PATCH 5.10] fs/proc: do_task_stat: use __for_each_thread()

Posted by Heyne, Maximilian 2 months, 4 weeks ago

From: Oleg Nesterov <oleg@redhat.com>

[ Upstream commit 7904e53ed5a20fc678c01d5d1b07ec486425bb6a ]

do/while_each_thread should be avoided when possible.

Link: https://lkml.kernel.org/r/20230909164501.GA11581@redhat.com
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Stable-dep-of: 7601df8031fd ("fs/proc: do_task_stat: use sig->stats_lock to gather the threads/children stats")
Cc: stable@vger.kernel.org
[mheyne: adjusted context]
Signed-off-by: Maximilian Heyne <mheyne@amazon.de>
---

Compile-tested only.
We're seeing soft lock-ups with 5.10.237 because of the backport of
commit 4fe85bdaabd6 ("fs/proc: do_task_stat: use sig->stats_lock to
gather the threads/children stats").

---
 fs/proc/array.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/proc/array.c b/fs/proc/array.c
index 8fba6d39e776..77b94c04e4af 100644
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -512,18 +512,18 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 		cgtime = sig->cgtime;
 
 		if (whole) {
-			struct task_struct *t = task;
+			struct task_struct *t;
 
 			min_flt = sig->min_flt;
 			maj_flt = sig->maj_flt;
 			gtime = sig->gtime;
 
 			rcu_read_lock();
-			do {
+			__for_each_thread(sig, t) {
 				min_flt += t->min_flt;
 				maj_flt += t->maj_flt;
 				gtime += task_gtime(t);
-			} while_each_thread(task, t);
+			}
 			rcu_read_unlock();
 
 			thread_group_cputime_adjusted(task, &utime, &stime);
-- 
2.47.1




Amazon Web Services Development Center Germany GmbH
Tamara-Danz-Str. 13
10243 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Jonathan Weiss
Eingetragen am Amtsgericht Charlottenburg unter HRB 257764 B
Sitz: Berlin
Ust-ID: DE 365 538 597

Re: [RESEND PATCH 5.10] fs/proc: do_task_stat: use __for_each_thread()

Posted by Greg KH 2 months, 4 weeks ago

On Thu, Jul 10, 2025 at 12:35:43PM +0000, Heyne, Maximilian wrote:
> From: Oleg Nesterov <oleg@redhat.com>
> 
> [ Upstream commit 7904e53ed5a20fc678c01d5d1b07ec486425bb6a ]
> 
> do/while_each_thread should be avoided when possible.
> 
> Link: https://lkml.kernel.org/r/20230909164501.GA11581@redhat.com
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
> Cc: Eric W. Biederman <ebiederm@xmission.com>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> Stable-dep-of: 7601df8031fd ("fs/proc: do_task_stat: use sig->stats_lock to gather the threads/children stats")
> Cc: stable@vger.kernel.org
> [mheyne: adjusted context]
> Signed-off-by: Maximilian Heyne <mheyne@amazon.de>
> ---
> 
> Compile-tested only.
> We're seeing soft lock-ups with 5.10.237 because of the backport of
> commit 4fe85bdaabd6 ("fs/proc: do_task_stat: use sig->stats_lock to
> gather the threads/children stats").

And this fixes it?

How?

> 
> ---
>  fs/proc/array.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/proc/array.c b/fs/proc/array.c
> index 8fba6d39e776..77b94c04e4af 100644
> --- a/fs/proc/array.c
> +++ b/fs/proc/array.c
> @@ -512,18 +512,18 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
>  		cgtime = sig->cgtime;
>  
>  		if (whole) {
> -			struct task_struct *t = task;
> +			struct task_struct *t;
>  
>  			min_flt = sig->min_flt;
>  			maj_flt = sig->maj_flt;
>  			gtime = sig->gtime;
>  
>  			rcu_read_lock();
> -			do {
> +			__for_each_thread(sig, t) {
>  				min_flt += t->min_flt;
>  				maj_flt += t->maj_flt;
>  				gtime += task_gtime(t);
> -			} while_each_thread(task, t);
> +			}

Ideally, the code generated here should be identical as before, so why
is this change needed?

confused,

greg k-h

Re: [RESEND PATCH 5.10] fs/proc: do_task_stat: use __for_each_thread()

Posted by Harshit Mogalapalli 2 months, 4 weeks ago

Hi Greg,

On 10/07/25 18:39, Greg KH wrote:
> On Thu, Jul 10, 2025 at 12:35:43PM +0000, Heyne, Maximilian wrote:
>> From: Oleg Nesterov <oleg@redhat.com>
>>
>> [ Upstream commit 7904e53ed5a20fc678c01d5d1b07ec486425bb6a ]
>>
>> do/while_each_thread should be avoided when possible.
>>
>> Link: https://lkml.kernel.org/r/20230909164501.GA11581@redhat.com
>> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
>> Cc: Eric W. Biederman <ebiederm@xmission.com>
>> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
>> Stable-dep-of: 7601df8031fd ("fs/proc: do_task_stat: use sig->stats_lock to gather the threads/children stats")
>> Cc: stable@vger.kernel.org
>> [mheyne: adjusted context]
>> Signed-off-by: Maximilian Heyne <mheyne@amazon.de>
>> ---
>>
>> Compile-tested only.
>> We're seeing soft lock-ups with 5.10.237 because of the backport of
>> commit 4fe85bdaabd6 ("fs/proc: do_task_stat: use sig->stats_lock to
>> gather the threads/children stats").
> 
> And this fixes it?
> 

Our testing also showed that after the backport of this commit(on 5.15.y 
based release), we don't see the soft lockup anymore.

> How?

I think __for_each_thread() is safe whereas while_each_thread() is not safe.

This thread 
https://lore.kernel.org/all/20131202152437.GA10896@redhat.com/ explains 
why while_each_thread() is unsafe.

Thanks,
Harshit

> 
>>
>> ---
>>   fs/proc/array.c | 6 +++---
>>   1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/fs/proc/array.c b/fs/proc/array.c
>> index 8fba6d39e776..77b94c04e4af 100644
>> --- a/fs/proc/array.c
>> +++ b/fs/proc/array.c
>> @@ -512,18 +512,18 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
>>   		cgtime = sig->cgtime;
>>   
>>   		if (whole) {
>> -			struct task_struct *t = task;
>> +			struct task_struct *t;
>>   
>>   			min_flt = sig->min_flt;
>>   			maj_flt = sig->maj_flt;
>>   			gtime = sig->gtime;
>>   
>>   			rcu_read_lock();
>> -			do {
>> +			__for_each_thread(sig, t) {
>>   				min_flt += t->min_flt;
>>   				maj_flt += t->maj_flt;
>>   				gtime += task_gtime(t);
>> -			} while_each_thread(task, t);
>> +			}
> 
> Ideally, the code generated here should be identical as before, so why
> is this change needed?
> > confused,>
> greg k-h