Re: [PATCH] livepatch: Avoid hard lockup caused by klp_try_switch_task()

[zhang warden <zhangwarden@gmail.com>]

> On Jan 22, 2025, at 20:50, Petr Mladek <pmladek@suse.com> wrote:
> 
> With this patch, any operation which takes the tasklist_lock might
> break klp_try_complete_transition(). I am afraid that this might
> block the transition for a long time on huge systems with some
> specific loads.
> 
> And the problem is caused by a printk() added just for debugging.
> I wonder if you even use a slow serial port.
> 
> You might try to use printk_deferred() instead. Also you might need
> to disable interrupts around the read_lock()/read_unlock() to
> make sure that the console handling will be deferred after
> the tasklist_lock gets released.
> 
> Anyway, I am against this patch.
> 
> Best Regards,
> Petr

Hi, Petr.

I am unfamiliar with the function `rwlock_is_contended`, but it seems this function will not block and just only check the status of the rw_lock.

If I understand it right, the problem would raise from the `break` which will stop the process of `for_each_process_thread`, right?

Thanks.
Wardenjohn.