Re: [PATCH] locking/lockdep: Dump state of percpu_rwsem upon hung up.

[Tetsuo Handa]

On 2018/07/10 18:30, Peter Zijlstra wrote:
>> diff --git a/include/linux/rwsem-spinlock.h b/include/linux/rwsem-spinlock.h
>> index e475683..1164965 100644
>> --- a/include/linux/rwsem-spinlock.h
>> +++ b/include/linux/rwsem-spinlock.h
>> @@ -22,7 +22,7 @@
>>   * - if wait_list is not empty, then there are processes waiting for the semaphore
>>   */
>>  struct rw_semaphore {
>> -	__s32			count;
>> +	atomic_long_t		count;
> 
> Uhhh... how does this even build? rwsem-spinlock.c doesn't use
> atomic_long primitives to change count.

Sorry, I forgot to build test rwsem-spinlock.c side.

> 
> And the atomic_long count usage is completely private to rwsem-xadd.c,
> nothing outside of that should use that field, ever.
> 

The reason I made above patch is to make
"[PATCH] locking/lockdep: Dump state of percpu_rwsem upon hung up." patch
possible to build on x86_64. Although "nothing outside of that should use
that field", I want to check "struct rw_semaphore"->count for isolating
the problem.

Should I update
"[PATCH] locking/lockdep: Dump state of percpu_rwsem upon hung up." side
to use "#ifndef CONFIG_RWSEM_GENERIC_SPINLOCK" ?