On 11/30/2016 07:38 PM, Damien Le Moal wrote:
>
> Jens,
>
> On 12/1/16 10:40, Jens Axboe wrote:
>> This looks better, thanks. Are the zone sizes mandated by spec to be a
>> power-of-2?
>
> No, the standards allow any zone size, and different sizes for the zones
> too. However, the sd_zbc code down in the SCSI stack limits support to
> HM & HA drives that have a power of 2 zone size, with all zones of the
> same size, except for an eventual smaller last zone (Seagate drives have
> that). This restriction was necessary so that limits.chunk_sectors can
> be used to avoid BIO spawning zones.
>
> See 89d9475610771b5e5fe1879075f0fc9ba6e3755f:
>
> diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
> new file mode 100644
> index 0000000..16d3fa6
> --- /dev/null
> +++ b/drivers/scsi/sd_zbc.c
> @@ -0,0 +1,642 @@
> +/*
> + * SCSI Zoned Block commands
> ...
> + if (!is_power_of_2(zone_blocks)) {
> + if (sdkp->first_scan)
> + sd_printk(KERN_NOTICE, sdkp,
> + "Devices with non power of 2 zone "
> + "size are not supported\n");
> + return -ENODEV;
> + }
> +
>
> Martin, Shaun and myself agreed on the restriction. All ZBC/ZAC disks on
> the market today match these so there are no problems. And it is
> unlikely that we will ever see weirdly sized SMR drives (customers
> generally do not want that).
I'm fine with that, my only concern is that part_zone_aligned() assumes
this, and it's a bit fragile. If we remove the power-of-2 restriction.
Not sure what the best way to fix it is. Ideally it'd have a
WARN_ON_ONCE() and a fallback to a modulo calculation instead of the
AND.
--
Jens Axboe
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
