nvme completion handling refactor and fix v2

[Christoph Hellwig]

On Mon, Aug 17, 2020 at 11:10:33AM -0700, Sagi Grimberg wrote:
>
>> Hi all,
>>
>> the first two patches refactor the common nvme completion code
>> to be a little less obsfucated, and the third one is the fix
>> from Chao to redirect to a different path ported on top of these
>> changes, which now make the fix very simple.
>>
>> Changes since v1:
>>   - add a new patch to handle the path related errors using the
>>     SCT class
>
> This looks good to me,
>
> Do you think this is 5.9-rc material? I'm thinking this should go
> to 5.10...

I think 5.9 should be fine.  We're early in the cycle, we've go a fix,
and more important the building blocks for the persistent reservation
fix, and also a similar one in ->report_zones that I found now.