linux-kernel - Re: [PATCH] hung_task: add warning counter to blocked task report

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1bbddb30-05bb-43c4-b104-c127430b3915@linux.dev>
Date: Wed, 23 Jul 2025 15:31:18 +0800
From: Ye Liu <ye.liu@...ux.dev>
To: Lance Yang <lance.yang@...ux.dev>
Cc: Ye Liu <liuye@...inos.cn>, linux-kernel@...r.kernel.org,
 Andrew Morton <akpm@...ux-foundation.org>, Zi Li <zi.li@...ux.dev>
Subject: Re: [PATCH] hung_task: add warning counter to blocked task report



在 2025/7/21 14:19, Lance Yang 写道:
> 
> 
> On 2025/7/21 13:45, Ye Liu wrote:
>>
>>
>> On 2025/7/21 12:56, Lance Yang wrote:
>>> Hi Ye,
>>>
>>> Thanks for your patch!
>>>
>>> On 2025/7/21 11:17, Ye Liu wrote:
>>>> From: Ye Liu <liuye@...inos.cn>
>>>>
>>>> Add a warning counter to each hung task message to make it easier
>>>> to analyze and locate issues in the logs.
>>>>
>>>> Signed-off-by: Ye Liu <liuye@...inos.cn>
>>>> ---
>>>>    kernel/hung_task.c | 6 ++++--
>>>>    1 file changed, 4 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/kernel/hung_task.c b/kernel/hung_task.c
>>>> index 8708a1205f82..9e5f86148d47 100644
>>>> --- a/kernel/hung_task.c
>>>> +++ b/kernel/hung_task.c
>>>> @@ -58,6 +58,7 @@ EXPORT_SYMBOL_GPL(sysctl_hung_task_timeout_secs);
>>>>    static unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
>>>>      static int __read_mostly sysctl_hung_task_warnings = 10;
>>>> +static int hung_task_warning_count;
>>>>      static int __read_mostly did_panic;
>>>>    static bool hung_task_show_lock;
>>>> @@ -232,8 +233,9 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
>>>>        if (sysctl_hung_task_warnings || hung_task_call_panic) {
>>>>            if (sysctl_hung_task_warnings > 0)
>>>>                sysctl_hung_task_warnings--;
>>>> -        pr_err("INFO: task %s:%d blocked for more than %ld seconds.\n",
>>>> -               t->comm, t->pid, (jiffies - t->last_switch_time) / HZ);
>>>> +        pr_err("INFO: task %s:%d blocked for more than %ld seconds. [Warning #%d]\n",
>>>> +               t->comm, t->pid, (jiffies - t->last_switch_time) / HZ,
>>>> +               ++hung_task_warning_count);
>>>>            pr_err("      %s %s %.*s\n",
>>>>                print_tainted(), init_utsname()->release,
>>>>                (int)strcspn(init_utsname()->version, " "),
>>>
>>> A quick thought on this: we already have the hung_task_detect_count
>>> counter, which tracks the total number of hung tasks detected since
>>> boot ;)
>>>
>>> While this patch adds a counter inline with the warning message, the
>>> existing counter already provides a way to know how many hung task
>>> events have occurred.
>>>
>>> Could you elaborate on the specific benefit of printing this count
>>> directly in the log, compared to checking the global hung_task_detect_count?
>>>
>>> Also, if the goal is to give each warning a unique sequence number,
>>> I think the dmesg timestamp prefix serves the same purpose ;)
>>>
>>> Thanks,
>>> Lance
>>
>> Sorry for not noticing sysctl_hung_task_detect_count.
>> I just thought adding it directly to the warning message would make the
>> log easier to read and more intuitive than relying on timestamps.
>>
>> If accepted, I will send V2, like this:
> 
> Let's step back and considet the practical use case. when we are
> troubleshooting hung task issues in a production log, what information
> do we actually use?
> 
> Typically, we look for:
> 1) The timestamp, to correlate with other system events
> 2) The task name and PID (%s:%d)
> 3) The kernel stack trace that follows, to see where it's stuck
> 
> So, my question is: in what specific troubleshooting scenario would
> knowing the sequence number, like [#N], provide actionable information
> that the above data points do not?
> 
> Unless there's a compelling use case I'm missing, I'd prefer to keep
> the code as it is ;)

There is no specific use case—the system currently defaults to 10 printouts.
When multiple logs are present in the same window, I simply want to quickly
identify which one is the Nth instance (e.g., the 3rd out of 10). That’s all.

> Thanks,
> Lance
> 
>>
>> diff --git a/kernel/hung_task.c b/kernel/hung_task.c
>> index 8708a1205f82..231afdb68bb2 100644
>> --- a/kernel/hung_task.c
>> +++ b/kernel/hung_task.c
>> @@ -232,8 +232,9 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
>>          if (sysctl_hung_task_warnings || hung_task_call_panic) {
>>                  if (sysctl_hung_task_warnings > 0)
>>                          sysctl_hung_task_warnings--;
>> -               pr_err("INFO: task %s:%d blocked for more than %ld seconds.\n",
>> -                      t->comm, t->pid, (jiffies - t->last_switch_time) / HZ);
>> +               pr_err("INFO: task %s:%d blocked for more than %ld seconds. [#%ld]\n",
>> +                      t->comm, t->pid, (jiffies - t->last_switch_time) / HZ,
>> +                      sysctl_hung_task_detect_count);
>>                  pr_err("      %s %s %.*s\n",
>>                          print_tainted(), init_utsname()->release,
>>                          (int)strcspn(init_utsname()->version, " "),
>>
>>
>>
>>
> 

-- 
Thanks,
Ye Liu