[PATCH 2/2] nvme: add 'queue_if_no_path' semantics

[Hannes Reinecke]

On 10/6/20 10:27 AM, Christoph Hellwig wrote:
> On Tue, Oct 06, 2020 at 10:07:49AM +0200, Hannes Reinecke wrote:
>> On 10/6/20 9:51 AM, Christoph Hellwig wrote:
>>> On Tue, Oct 06, 2020 at 07:48:40AM +0200, Hannes Reinecke wrote:
>>>> Well, because it's established behaviour of the current code.
>>>> Changing it now has the potential to break existing scenarios.
>>>>
>>>> For PCI (ie non-fabrics) the current behaviour is arguably a corner case
>>>> (as one needs to have a PCI-NVMe with CMIC bit set), but once you have it
>>>> PCI hotplug is done for. So there we really want to change the behaviour to
>>>> get the same user experience for all NVMe drives.
>>>
>>> Dual port PCI NVMe devices are pretty common in the enterprise space.
>>> I really don't see any good reason to differ here.
>>>
>> ?
>> Differ from what?
>>
>> Normal (non-dual ported) PCI-NVMe?
>> NVMe-oF?
> 
> All multipath devices should behave the same.  No special casing for
> PCIe, please.
> 
Even if the default behaviour breaks PCI hotplug?

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		           Kernel Storage Architect
hare at suse.de			                  +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer