Re: [PATCH] bcache: make stripe_size configurable and persistent for hardware raid5/6

[Eric Wheeler]

On Mon, 24 Jun 2019, Martin K. Petersen wrote:
> > Perhaps they do not set stripe_width using io_opt? I did a grep to see
> > if any of them did, but I didn't see them. How is stripe_width
> > indicated by RAID controllers?
> 
> The values are reported in the Block Limits VPD page for each SCSI block
> device and are thus set by the SCSI disk driver. IOW, the RAID
> controller device drivers have nothing to do with this.
> 
> For RAID controllers specifically, the controller firmware will fill out
> the VPD fields for each virtual SCSI disk when you configure a RAID
> set. For pretty much everything else, the Block Limits come straight
> from the device itself.
> 
> Also note that these values aren't specific to RAID controllers at
> all. Most new SCSI devices, including disk drives and SSDs, will fill
> out the Block Limits VPD page one way or the other. Even some USB
> storage devices are providing this page.

Thanks, that makes sense.  Interesting about USB.

> > If they do set io_opt, then at least my Areca 1883 does not set io_opt
> > as of 4.19.x. I also have a LSI MegaRAID 3108 which does not report
> > io_opt as of 4.1.x, but that is an older kernel so maybe support has
> > been added since then.
> 
> I have several MegaRAIDs that all report it. But it depends on the
> controller firmware.
> 
> > Is it visible through sysfs or debugfs so I can check my hardware
> > support without hacking debugging the kernel?
> 
> To print the block device topology:
> 
>   # lsblk -t
> 
> or look up io_opt in sysfs:
> 
>   # grep . /sys/block/sdX/queue/optimal_io_size
> 
> You can also query a SCSI device's Block Limits directly:
> 
>   # sg_vpd -p bl /dev/sdX

Perfect, thank you for that.  I've tried the following controllers that I 
have access to.  One worked (hspa/HP Gen8 Smart Array Controller), but the 
others I tried are not providing VPDs:

* LSI 2108 (Supermicro)
* LSI 3108 (Dell)
* Areca 1882
* Areca 1883
* Fibrechannel 8gbe connected to a Storwize 3700

~]# sg_vpd -p bl /dev/sdb
VPD page=0xb0
fetching VPD page failed

> If you want to tinker, you can simulate a SCSI disk with your choice of
> io_opt:
> 
>   # modprobe scsi_debug opt_blks=N
> 
> where N is the number of logical blocks to report as being the optimal
> I/O size.

Neat, thanks for the hint!

-Eric

> 
> -- 
> Martin K. Petersen	Oracle Linux Engineering
>