[PATCH] NVMe: Add reset controller sysfs entry

[keith.busch@intel.com (Keith Busch)]

On Thu, 4 Jun 2015, Jens Axboe wrote:

> On 06/04/2015 02:00 PM, Jens Axboe wrote:
>> On 06/04/2015 09:39 AM, Christoph Hellwig wrote:
>>> On Thu, Jun 04, 2015@03:26:03PM +0000, Keith Busch wrote:
>>>> On Thu, 4 Jun 2015, Brandon Schulz wrote:
>>>>> Do you plan to merge this into your legacy tree somewhere that we can
>>>>> reference in Bugzillas/etc. with the Linux distro vendors?
>>>> 
>>>> There is a strong preference with vendors to see fixes upstream prior
>>>> to backporting but this is a long-standing percieved weakness in Linux
>>>> vs. other operating systems; no one wants to reboot their machine for
>>>> a f/w upgrade ... so, okay, I'll apply it to legacy and see if I can
>>>> lobby OSVs to take out-of-stream fixes.
>>> 
>>> Eww.  Jens, can you please just apply this patch so that we're not
>>> having such a mess?
>> 
>> Yeah, I'll apply it now for 4.2.
>
> Well, that was a bit of a mess. Attached is my quick attempt at combining the 
> two, and reusing the reset code from the ioctl path as well. One functional 
> change was NOT flushing the work if we don't queue it in nvme_reset(), the 
> previous code from Keith did that. Otherwise it should be straight forward.
>
> Please review, I'll hold off on committing this.

Hi Jens,

Thanks for fixing the merge. This tests successfully on my machine. I
didn't have a strong opinion either way on what a user sees when
requesting resets while a reset is in progress, and what you've merged
looks good to me.