Re: [PATCH] loop: Increase bsize variable from unsigned short to unsigned int

[John Garry <john.g.garry@oracle.com>]

On 27/08/2024 08:35, Li Wang wrote:
> On Tue, Aug 27, 2024 at 3:20 PM John Garry <john.g.garry@oracle.com> wrote:
>>
>> On 27/08/2024 04:22, Li Wang wrote:
>>
>> +linux-block, Jens
>>
>>> This change allows the loopback driver to handle larger block sizes
>>
>> larger than what? PAGE_SIZE?
> 
> Yes, 

Please then explicitly mention that

> system should return EINVAL when the tested bsize is larger than PAGE_SIZE.
> But due to the loop_reconfigure_limits() cast it to 'unsined short' that
 > never gets an expected error when testing invalid logical block size.>
> That's why LTP/ioctl_loop06 failed on a system with 64k (ppc64le) pagesize
> (since kernel-v6.11-rc1 with patches):
> 
>    9423c653fe6110 ("loop: Don't bother validating blocksize")
>    fe3d508ba95bc6 ("block: Validate logical block size in blk_validate_limits()")
> 
> 
> 

I feel that you should be adding a fixes tag, but it seems that those 
commits only expose the issue, and whatever introduced unsigned short 
usage was not right. Maybe you can just get this included for 6.11, 
where 9423c653fe6110 was introduced.

Thanks,
John