On Thu, Aug 29 2019, Yufen Yu wrote:
> In RAID5, if issued bio has sectors more than STRIPE_SIZE,
> it will be split in unit of STRIPE_SIZE. For bio sectors
> less then STRIPE_SIZE, raid5 issue bio to disk in the size
> of STRIPE_SIZE.
>
> For now, STRIPE_SIZE is equal to the value of PAGE_SIZE.
> That means, RAID5 will issus echo bio to disk at least 64KB
> when use 64KB page size in arm64.
>
> However, filesystem usually issue bio in the unit of 4KB.
> Then, RAID5 will waste resource of disk bandwidth and compute xor.
>
> To avoding the waste, we can add a CONFIG option to set
> the default STRIPE_SIZE as 4096. User can also set the value
> bigger than 4KB for some special requirements, such as we know
> the issued io size is more than 4KB.
>
> Signed-off-by: Yufen Yu <yuyufen@xxxxxxxxxx>
> ---
> drivers/md/Kconfig | 9 +++++++++
> drivers/md/raid5.h | 2 +-
> 2 files changed, 10 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig
> index 3834332f4963..4263a655dbbb 100644
> --- a/drivers/md/Kconfig
> +++ b/drivers/md/Kconfig
> @@ -157,6 +157,15 @@ config MD_RAID456
>
> If unsure, say Y.
>
> +config MD_RAID456_STRIPE_SIZE
> + int "RAID4/RAID5/RAID6 stripe size"
> + default "4096"
> + depends on MD_RAID456
> + help
> + The default value is 4096. Only change this if you know
> + what you are doing.
> +
> +
> config MD_MULTIPATH
> tristate "Multipath I/O support"
> depends on BLK_DEV_MD
> diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
> index cf991f13403e..2fcf5853b719 100644
> --- a/drivers/md/raid5.h
> +++ b/drivers/md/raid5.h
> @@ -473,7 +473,7 @@ struct disk_info {
> */
>
> #define NR_STRIPES 256
> -#define STRIPE_SIZE PAGE_SIZE
> +#define STRIPE_SIZE CONFIG_MD_RAID456_STRIPE_SIZE
> #define STRIPE_SHIFT (PAGE_SHIFT - 9)
> #define STRIPE_SECTORS (STRIPE_SIZE>>9)
> #define IO_THRESHOLD 1
> --
> 2.17.2
While I agree that this is a problem that should be fixed, I think your
fix is rather simplistic.
grow_buffers() uses alloc_page() allocate the buffers for each
stripe_head()
With this change, it will allocate 64K buffers and use 4K of them.
This is a waste. Unfortunately fixing it would be a lot more work.
I think that is best way forward though.
Also, I don't think this should be a config option - just make it
always 4096.
Thanks,
NeilBrown
Attachment:
signature.asc
Description: PGP signature
