Re: [PATCH 4/5] brd: Request from fdisk 4k alignment

[Boaz Harrosh <boaz@plexistor.com>]

On 11/06/2014 07:25 PM, Martin K. Petersen wrote:
>>>>>> "Boaz" == Boaz Harrosh <boaz@plexistor.com> writes:
> 
> Boaz,
> 
> Boaz> Note that blk_queue_physical_block_size() also trashes io_min, but
> Boaz> we can leave this one to be 512. io_min as opposed to
> Boaz> physical_block_size will actually change Kernel behavior
> 
>>> Care to elaborate?
> 
> I wanted you to elaborate on how io_min changes kernel behavior.
> 

OK You are correct. io_min changes behavior of Kernel in exactly the same
way as physical_block_size.

Through the call to queue_limit_alignment_offset() inside add_partition()
by setting hd_struct->alignment_offset

I do not know why I thought that only io_min does that, I can see now that
both effect the Kernel the same way. Which scares me a bit.

Will I have problems?

<>
> 
> It really sounds like either your fdisk is way too old or you are
> running it in DOS compat mode.
> 
> I don't have a fundamental issue reporting pbs of 4K. But if you are
> only doing it to force a certain partition alignment then it sounds like
> a kernel fix for a userland problem.
> 
> I've CC:ed Karel who can comment on fdisk partition alignment issues.
> 

Thanks Martin. I agree, we should not fix user-mode problems. Please see
the other email with the exact numbers I get from fdisk. It looks like
when everything is 512 like default it would not align my partitions, but
with the 4k-phisical thing it would. What is the expected behavior we want?

Thanks
Boaz