[PATCH 2/2] nvme: don't set iosqes,iocqes for discovery controllers

[Sagi Grimberg]

>> The annoying thing here is that this issue will manifest in the host
>> waiting on nvme_wait_ready for a long 7.5 seconds to understand that
>> its maybe incompatible and re-attempt (nvme_enable_ctrl writes to cap
>> and then polls CSTS)...
>>
>> What I have now looks like:
> 
> In terms of what to do about the retry this is what we'll have to
> do.  That being said I think for existing controllers we should just
> stick to the existing wrong value, and only set the correct one
> for versions > 1.3.  This isn't exactly correct, but probably causes
> the least harm.  It will need an extensive comment, and maybe we need
> to add an option to override it if controllers show up that do not
> like the wrong value.
So you suggest that we keep the retry w.a. but continue to send the
wrong value in 1.3 or lower?

I personally think that this would be extremely confusing to users, if
they do trip on this, but I agree that they are less likely to hit
it because existing controllers obviously accept wrong values...

As for option to override. I really hate it.. What is the least exposure
we can get for this but still allow the user to control this?

Wish we had a good and reliable quirk mechanism for fabrics...