Re: [PATCH v2] blk-iocost: fix shift-out-of-bounds in iocg_hick_delay()
From: Li Jinlin
Date: Mon Nov 28 2022 - 21:49:20 EST
On 2022/11/29 9:14, Yu Kuai wrote:
> Hi,
>
> 在 2022/11/29 3:58, Tejun Heo 写道:
>> On Mon, Nov 28, 2022 at 11:04:13AM +0800, Li Jinlin wrote:
>>> /* calculate the current delay in effect - 1/2 every second */
>>> tdelta = now->now - iocg->delay_at;
>>> if (iocg->delay)
>>> - delay = iocg->delay >> div64_u64(tdelta, USEC_PER_SEC);
>>> + delay = iocg->delay >>
>>> + min_t(u64, div64_u64(tdelta, USEC_PER_SEC), 63);
>>
>> I replied earlier but the right thing to do here is setting delay to 0 if
>> the shift is >= 64.
>
> Perhaps following change will make more sense?
>
> @@ -1322,18 +1323,19 @@ static bool iocg_kick_delay(struct ioc_gq *iocg, struct ioc_now *now)
> {
> struct ioc *ioc = iocg->ioc;
> struct blkcg_gq *blkg = iocg_to_blkg(iocg);
> - u64 tdelta, delay, new_delay;
> + u64 delay = 0;
> + u64 new_delay;
> s64 vover, vover_pct;
> u32 hwa;
>
> lockdep_assert_held(&iocg->waitq.lock);
>
> /* calculate the current delay in effect - 1/2 every second */
> - tdelta = now->now - iocg->delay_at;
> - if (iocg->delay)
> + if (iocg->delay && now->now > iocg->delay_at) {
> + u64 tdelta = now->now - iocg->delay_at;
> +
> delay = iocg->delay >> div64_u64(tdelta, USEC_PER_SEC);
> - else
> - delay = 0;
> + }
>
I think "now->now > iocg->delay_at" is unnecessary, it is almost inevitable.
What about the following change for setting delay to 0 if the shift is >= 64.
@@ -1329,11 +1329,9 @@ static bool iocg_kick_delay(struct ioc_gq *iocg, struct ioc_now *now)
lockdep_assert_held(&iocg->waitq.lock);
/* calculate the current delay in effect - 1/2 every second */
- tdelta = now->now - iocg->delay_at;
if (iocg->delay)
- delay = iocg->delay >> div64_u64(tdelta, USEC_PER_SEC);
- else
- delay = 0;
+ tdelta = div64_u64(now->now - iocg->delay_at, USEC_PER_SEC);
+ delay = (iocg->delay && tdelta < 64) ? iocg->delay >> tdelta : 0;
/* calculate the new delay from the debt amount */
current_hweight(iocg, &hwa, NULL);
Jinlin
Thanks.
>>
>> Thanks.
>>
>