Re: [patch 4/4] ipc: sem optimise simple operations

[Cyrill Gorcunov <gorcunov@gmail.com>]

On 8/12/09, Nick Piggin <npiggin@suse.de> wrote:
> On Wed, Aug 12, 2009 at 12:07:11AM +0400, Cyrill Gorcunov wrote:
>> [npiggin@suse.de - Tue, Aug 11, 2009 at 09:09:06PM +1000]
>> ...
>> | +static void update_queue_simple(struct sem_array *sma, ushort semnum)
>> | +{
>> | +	if (unlikely(sma->complex_count)) {
>> | +		update_queue(sma);
>> | +	} else {
>> | +		struct sem *sem;
>> | +
>> | +		sem = &sma->sem_base[semnum];
>> | +		if (sem->semval > 0)
>> | +			update_negv_queue(sma, sem);
>> | +		if (sem->semval == 0)
>> | +			update_zero_queue(sma, sem);
>> | +	}
>> | +}
>> | +
>> ...
>>
>> Hi Nick,
>>
>> mostly probably miss something but can't we trgigger BUG_ON at updating
>> zero queue if semaphore was created with undo list and via new operation
>> reached -ERANGE on undo value?
>>
>> Again, I could be missing something or plain wrong. Just a thought.
>
> Hi Cyrill,
>
> Thanks for looking... Hmm, you mean BUG_ON(error) due to try_atomic_semop
> returning -ERANGE? I think it should not be possible because it should
> prevent any operation from bringing the undo list to -ERANGE so then any
> operation which does not modify the sem value should not go out of range
> I think.
>
> (I think it would be a bug if we ever return -ERANGE for a wait-for-zero
> operation).
>
> Thanks,
> Nick
>
Thanks for explanation, Nick! I meant exactly that.