Re: [RT PATCH] locking/rwsem-rt: Remove might_sleep() in __up_read()

[Thomas Gleixner]

Andrew,

On Thu, Jul 08 2021 at 14:51, Andrew Halaney wrote:

Cc+ stable-rt folks and leave context untrimmed.

> On Mon, Apr 26, 2021 at 08:16:34PM +0200, Thomas Gleixner wrote:
>> On Mon, Apr 26 2021 at 11:42, Andrew Halaney wrote:
>> > On Tue, Apr 06, 2021 at 05:19:52PM -0500, Andrew Halaney wrote:
>> >> There's no chance of sleeping here, the reader is giving up the
>> >> lock and possibly waking up the writer who is waiting on it.
>> >> 
>> >> Reported-by: Chunyu Hu <chuhu@xxxxxxxxxx>
>> >> Signed-off-by: Andrew Halaney <ahalaney@xxxxxxxxxx>
>> >> ---
>> >> Hello,
>> >> 
>> >> I ran into a warning caused by this, and I think the warning is
>> >> incorrect. Please let me know if I'm wrong!
>> >> I'm working off of linux-5.12.y-rt, but this applies cleanly to older
>> >> stable branches as well.
>> >>
>> >> Thanks,
>> >> Andrew
>> >> 
>> >>  kernel/locking/rwsem-rt.c | 1 -
>> >>  1 file changed, 1 deletion(-)
>> >> 
>> >> diff --git a/kernel/locking/rwsem-rt.c b/kernel/locking/rwsem-rt.c
>> >> index 274172d5bb3a..b61edc4dcb73 100644
>> >> --- a/kernel/locking/rwsem-rt.c
>> >> +++ b/kernel/locking/rwsem-rt.c
>> >> @@ -198,7 +198,6 @@ void __up_read(struct rw_semaphore *sem)
>> >>  	if (!atomic_dec_and_test(&sem->readers))
>> >>  		return;
>> >>  
>> >> -	might_sleep();
>> >>  	raw_spin_lock_irq(&m->wait_lock);
>> >>  	/*
>> >>  	 * Wake the writer, i.e. the rtmutex owner. It might release the
>> >> -- 
>> >> 2.30.2
>> >
>> > Just a gentle follow up, any feedback?
>> 
>> Looks correct. Will go into the next rt-release.
>> 
>> Thanks,
>> 
>>         tglx
>> 
>
> Hi Thomas,
>
> I see a new release (v5.13-rt1) was created, and with it rwsem
> was overhauled entirely making this patch pointless for linux-rt-devel.
>
> That being said, it's a little unclear to me how RT only patches
> make their way to the maintained branches over in linux-stable-rt.
> I think it should be applied to:
>  v4.9-rt
>  v4.14-rt
>  v4.19-rt
>  v5.4-rt
>  v5.10-rt
> to remove the incorrect warning, unless those branches plan to backport
> the latest RT patchset with the new rwsem implementation. Is there a
> proper way for me signal that?

See Cc.

Thanks,

        tglx