Re: [PATCH v2 4/5] blk-iocost: fix sleeping in atomic context warnning

[Tejun Heo <tj@kernel.org>]

On Mon, Dec 05, 2022 at 05:39:33PM +0800, Yu Kuai wrote:
> Hi, Tejun
> 
> 在 2022/11/23 18:22, Yu Kuai 写道:
> > Hi, Tejun
> > 
> > 在 2022/11/23 8:42, Tejun Heo 写道:
> > > On Tue, Nov 22, 2022 at 05:14:29PM -0700, Jens Axboe wrote:
> > > > > > Then match_strdup() and kfree() in match_NUMBER() can be replaced with
> > > > > > get_buffer() and put_buffer().
> > > > > 
> > > > > Sorry about the late reply. Yeah, something like this.
> > 
> 
> I wonder can we just use arary directly in stack? The max size is just
> 24 bytes, which should be fine:
> 
> HEX: "0xFFFFFFFFFFFFFFFF" --> 18
> DEC: "18446744073709551615" --> 20
> OCT: "01777777777777777777777" --> 23
> 
> Something like:
> #define U64_MAX_SIZE 23
> static int match_strdup_local(const substring_t *s, char *buf)
> {
> 	size_t len = s->to - s->from;
> 
> 	if (len > U64_MAX_SIZE)
> 		return -ERANGE;
> 
> 	if (!s->from)
> 		return -EINVAL;
> 
> 	memcpy(buf, s->from, len);
> 	buf[len] = '\0';
> 	return 0;
> }
> 
>  static int match_u64int(substring_t *s, u64 *result, int base)
>  {
> 	char buf[U64_MAX_SIZE + 1];
>  	int ret;
>  	u64 val;
> 
> 	ret = match_strdup_local(s, buf);
> 	if (ret)
> 		return ret;
>  	ret = kstrtoull(buf, base, &val);
>  	if (!ret)
>  		*result = val;;
>  	return ret;
>  }

Oh yeah, absolutely. That's much better.

Thanks.

-- 
tejun