Re: [PATCH RESEND v2 5/5] sbitmap: correct wake_batch recalculation to avoid potential IO hung

[Kemeng Shi]

on 1/3/2023 10:12 AM, Kemeng Shi wrote:
> Friendly ping...
> 
> on 12/26/2022 4:57 PM, Yu Kuai wrote:
>> 在 2022/12/26 15:50, Yu Kuai 写道:
>>
>>>> why using
>>>> wake batch of 4 is safe for cards with say 32 tags in case active_users is
>>>> currently 32. Because I don't see why that is correct either.
>>>>
>>
>> I see, you guys are worried that during the period that some hctx
>> complete all it's not idle yet. It's right waiter might wait for
>> other hctx to become idle to be awaken in this case. However, I'm
>> not sure which way is better.
>>
>> Ming, do you have any idea?
>>
>> Thanks,
>> Kuai
>>
> 
Hi Jan. The magic batch 4 seems just for performance initially while
lacks of full consideration. And there is no better solution provided
in futher. Do you have any suggestion that I can do to make more
progress.

-- 
Best wishes
Kemeng Shi