Re: [PATCH for-next 4/4] nvme-multipath: add multipathing for uring-passthrough commands

[Sagi Grimberg <sagi@grimberg.me>]

>> I view uring passthru somewhat as a different thing than sending SG_IO
>> ioctls to dm-mpath. But it can be argued otherwise.
>>
>> BTW, the only consumer of it that I'm aware of commented that he
>> expects dm-mpath to retry SG_IO when dm-mpath retry for SG_IO submission
>> was attempted (https://www.spinics.net/lists/dm-devel/msg46924.html).
> 
> Yeah.  But the point is that if we have a path failure, the kernel
> will pick a new path next time anyway, both in dm-mpath and nvme-mpath.

If such a path is available at all.

>> I still think that there is a problem with the existing semantics for
>> passthru requests over mpath device nodes.
>>
>> Again, I think it will actually be cleaner not to expose passthru
>> devices for mpath at all if we are not going to support retry/failover.
> 
> I think they are very useful here.  Users of passthrough interface
> need to be able to retry anyway, even on non-multipath setups.  And
> a dumb retry will do the right thing.

I think you are painting a simple picture while this is not the case
necessarily. It is not a dumb retry, because the user needs to determine
if an available path for this particular namespace exists or wait for
one if it doesn't want to do a submit/fail constant loop.

A passthru interface does not mean that the user by definition needs to
understand multipathing, ana/ctrl/ns states/mappings, etc. The user may
just want to be able issue vendor-specific commands to the device.

If the user needs to understand multipathing by definition, he/she has
zero use of a mpath passthru device if it doesn't retry IMO.