On 9/24/2018 2:32 AM, Hannes Reinecke wrote:

quoted

+??? nvmerq->status = NVME_SC_ABORT_REQ;

Shouldn't we use '0x371' (as per TP 4028) for this?
Description said 'Command Aborted By host: The command was aborted as 
a result of host action (e.g., the host disconnected the Fabric 
connection).' which seem to be a pretty good match ...

Possibly - but...? If it was the transport that did a real io abort, it 
would make sense. But in this case, the io never even got started, so 
why mark it aborted ?? If it really was aborted, then doing a ctlr reset 
and incrementing retry counters may make sense. But if just bounced due 
to not being able to talk to the controller right now - those things 
should happen. So I still think we need a different value specific for 
the linux stack.

-- james