Re: [PATCH 2/3] blk-iocost: clear delay state when freeing policy data

[Tejun Heo]

> Does this unconditional call to blkcg_clear_delay() overwrite state managed
> by other policies?
> 
> Since use_delay is shared among block cgroup policies, if iocost never
> actually applied a delay, this might forcefully zero out the delay state
> managed by other active policies like blk-iolatency. 
> 
> Would it be safer to check if iocost had actually set a delay before clearing
> it here?

Yeah, this is fine. The two mechanisms aren't usually used together and even
when they step on each other, use_delay being cleared spuriously is
transient and nothing really breaks.

Thanks.

-- 
tejun