[PATCH v7] nvme-fabrics: reject I/O to offline device

[Hannes Reinecke]

On 7/10/20 6:50 AM, Sagi Grimberg wrote:
> 
>> Looks good to me.  Only nit I see is - why are we checking ctrl->opts 
>> in the xxx_failfast_work() routines ?
> 
> Which brings me to the question, is there any desire to have this
> behavior in pci as well?

I was under the impression that an offline NVMe-PCI device is a pretty 
terminal condition. This patch only would make sense if a reconnect 
would be able to recover this situation.
For NVMe-PCI we would notify the application earlier; device's still 
dead, though, so I'm not sure if that makes a huge difference.
Nor should it be a common occurence.

Unless you talk about ARM servers, for which frying NVMe devices seems 
to be a hobby :-)

Cheers,

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