Re: [PATCH 1/2] hung_task: Show the blocker task if the task is hung on mutex

From: Google
Date: Wed Feb 19 2025 - 17:57:08 EST


On Wed, 19 Feb 2025 17:44:11 -0500
Waiman Long <llong@xxxxxxxxxx> wrote:

>
> On 2/19/25 3:24 PM, Steven Rostedt wrote:
> > On Wed, 19 Feb 2025 15:18:57 -0500
> > Waiman Long <llong@xxxxxxxxxx> wrote:
> >
> >> It is tricky to access the mutex_waiter structure which is allocated
> >> from stack. So another way to work around this issue is to add a new
> >> blocked_on_mutex field in task_struct to directly point to relevant
> >> mutex. Yes, that increase the size of task_struct by 8 bytes, but it is
> >> a pretty large structure anyway. Using READ_ONCE/WRITE_ONCE() to access
> > And it's been on my TODO list for some time to try to make that structure
> > smaller again :-/

I agree to add the field, actually it was my first prototype :)

> >
> >> this field, we don't need to take lock, though taking the wait_lock may
> >> still be needed to examine other information inside the mutex.

Do we need to take it just for accessing owner, which is in an atomic?

> > But perhaps if we add a new config option for this feature, we could just
> > add the lock that a task is blocked on before it goes to sleep and
> > reference that instead. That would be easier than trying to play games
> > getting the lock owner from the blocked_on field.
>
> Yes, it could be a new config option. This will be a useful feature that
> I believe most distros will turn it on. Or we may just include that in
> the core code without any option.

Do we need another option? or just extend DETECT_HUNG_TASK?

Thanks,

>
> BTW, this field can also be shared by other sleeping locks like rwsem
> and rt_mutex as a task can only be blocked on one of them. We do need
> another type field to identify the type of the blocked lock.
>
> Cheers,
> Longman
>
>


--
Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>