Re: [PATCH 0/6] xfs: quota fixes and enhancements

[Eric Sandeen <sandeen@sandeen.net>]

On 4/7/18 8:37 PM, Darrick J. Wong wrote:
> On Sat, Apr 07, 2018 at 05:00:13PM -0500, Eric Sandeen wrote:
>> On 4/4/18 1:47 PM, Eric Sandeen wrote:
>>> A semi-random smattering of quota stuff.  First three seem quite
>>> good to go, the rest are more along the lines of a suggestion
>>> or conversation-starter.  ;)
>>>
>>> (the first patch is just removing an unused arg).
>>>
>>> xfs_repair doesn't look at quota blocks.  At all.  It relies
>>> on quotacheck in the kernel to fix them up as needed.
>>
>> I'm starting to rethink a lot of this hackery.  Why doesn't xfs_repair
>> just fix things up?  (leave quotacheck to the next mount, but the
>> "repair" stuff in the kernel seems like a really strange wart.)
>>
>> I think I'll look at teaching repair to sanity check the quota
>> inodes, but if anyone knows why that's a bad idea please let me
>> know.  ;)
> 
> /me shrugs, we still need to fix the kernel's quota verifiers to check
> the uuid and all that, right?  Which means that both are going to need
> patches, afaict.

The verifiers do check UUID:

xfs_dquot_buf_read_verify
	xfs_dquot_buf_verify_crc
                if (!uuid_equal(&d->dd_uuid, &mp->m_sb.sb_meta_uuid))
                        return false;

the issue is that neither the kernel nor userspace repairs this error
if it's detected.

-Eric