> On 2021/06/04 16:53, Changheun Lee wrote:
> >> On 2021/06/04 14:22, Changheun Lee wrote:
> >>> + * @q: the request queue for the device
> >>> + * @bytes : bio max bytes to be set
> >>> + *
> >>> + * Description:
> >>> + * Set proper bio max size to optimize queue operating.
> >>> + **/
> >>> +void blk_queue_max_bio_bytes(struct request_queue *q, unsigned int bytes)
> >>> +{
> >>> + struct queue_limits *limits = &q->limits;
> >>> + unsigned int max_bio_bytes = round_up(bytes, PAGE_SIZE);
> >>> +
> >>> + limits->max_bio_bytes = max_t(unsigned int, max_bio_bytes,
> >>> + BIO_MAX_VECS * PAGE_SIZE);
> >>> +}
> >>> +EXPORT_SYMBOL(blk_queue_max_bio_bytes);
> >>
> >> Setting of the stacked limits is still missing.
> >
> > max_bio_bytes for stacked device is just default(UINT_MAX) in this patch.
> > Because blk_set_stacking_limits() call blk_set_default_limits().
> > I'll work continue for stacked device after this patchowork.
>
> Why ? Without that added now, anybody using this performance fix will see no
> benefits if a device mapper is used. The stacking limit should be super simple.
> In blk_stack_limits(), just add:
>
> t->max_bio_bytes = min(t->max_bio_bytes, b->max_bio_bytes);
>
>
> --
> Damien Le Moal
> Western Digital Research
I had tried like as your comment at first. But I got many feedbacks that
applying for all device is not good idea. So I'll try to control bio size
in each stacked driver like as setting max_bio_bytes in LLD as you
recommended. I'm trying to find some target stacked devices to contorl
bio size like as dm-crypt, dm-liner, ... etc. And I'll try to find some
method to control bio max size include using of blk_queue_max_bio_bytes().
Thank you,
Changheun Lee
