[RFC v1 0/3] Unifying fabrics drivers
Chaitanya Kulkarni
chaitanyak at nvidia.com
Tue Mar 7 15:55:52 PST 2023
>> start_queue():
>> can't common code do the nvmf_connect_admin_queue() call ?
>> If its too tied to the LIVE bit - then perhaps the queue LIVE
>> bits should actually be getting set by the common code after
>> the successful start_queue - or has routines transport can call
>> from start/stop to set/clear it.
>> If LIVE bit can be dealt with - this whole block in rdma/tcp moves to
>> common code.
> Good point, the LIVE bit is currently transport specific which it dosen't need
> to be.
>
> What would be the best strategy with this kind of change for review? Should I
> add stuff ontop v2 or should I try to add the common code first and then migrate
> the transports over? The current approach with moving and updating in the same
> time is obviously a bit hard to review (and has a good chance to introduce
> regressions).
please avoid it moving and updating at the same time..
I'd first do a prep patches and move the bits/code around
and establish the stability of code then only I'd make the
actual changes for unification...
-ck
More information about the Linux-nvme
mailing list