[Parted-maintainers] Bug#923561: parted: Incorrect optimal alignment for USB device

[Kevin Locke]

After a bit more research, I found that this issue has been mentioned
on linux-usb[1] and that Martin K. Petersen weighed in on the issue in
a util-linux-ng thread[2]:

On 17 June 2015 at 01:08, Martin K. Petersen <martin.petersen at oracle.com> wrote:
> There's only so much we can do about devices that report garbage.
> 
> Also, the kernel only reports things. It is up to Karel to decide
> whether to sanity check the values before he uses them.

If we are going to bring it up again, I'd like to have a specific
recommendation or request.  More discussion below:

On Mon, 2019-03-04 at 13:42 -0500, Phillip Susi wrote:
> On 3/4/2019 11:50 AM, Kevin Locke wrote:
>> optimal_io_size is defined in the kernel ABI[1] as "the preferred
>> request size for workloads where sustained throughput is desired".  In
>> this case, I think it comes from scsi_disk.opt_xfer_blocks[2] which
>> comes from the VPD block limits page[3] which my disk reports as 65535
>> blocks (according to sg_vpd).
> 
> Yes, I was looking over the definition earlier in the kernel
> documentation, and it seems to me that part of the definition indicates
> it is simply an optimal size to transfer in, and has nothing to do with
> alignment.  But parted treats it as having to do with alignment, because
> the same kernel documentation mentions that is what the md driver uses
> the field for.

Could you point me to the kernel documentation that mentions how the
md driver uses optimal_io_size?

>> However, whether or not this is changed in the kernel, I would argue
>> it should be handled in parted.  Both to support previous+current
>> kernels and because I think the optimal_io_size ABI definition allows
>> optimal_io_size which is not a multiple of minimum_io_size for
>> transports where that truly is the case.
> 
> Unfortunately, that means parted needs to simply ignore the value of
> optimal io size as totally meaningless garbage, and I don't much like
> the thought of that.  If parted has always been wrong and the intended
> meaning of the kernel parameter has never had anything to do with
> alignment, then yes, parted needs to ignore it because it read meaning
> into it that was never intended.  But if it is supposed to have
> something to do with optimal alignment, then md is right and usb is
> wrong, and usb needs to be fixed.

My current reading is that optimal_io_size has the same definition as
SCSI VPD Optimal Transfer Length.  It has a loosely defined meaning,
but its value for any particular use is contingent on sanity checking.

Do you have a particular proposal for how to improve the kernel
documentation or how optimal_io_size behaves?  I worked up a patch
which only uses logical_to_bytes(sdp, sdkp->opt_xfer_blocks) for
io_opt if it is a multiple of sdkp->physical_block_size, but I am not
convinced enough that it is universally applicable to advocate for it.
Any alternative suggestions?

Thanks,
Kevin

[1]: https://www.spinics.net/lists/linux-usb/msg125988.html
[2]: https://www.spinics.net/lists/util-linux-ng/msg11702.html