On 17/09/2020 09:26, Kashyap Desai wrote:
Hi John -
We have driver code of <mpt3sas> ready for this feature. Due to some other
priority, we are not able to make any progress on this. Primarily we want to
cover more unit testing and performance on this feature.
We will be sending <mpt3sas> driver version soon.
ok, great. Not sure why we missed this one originally.
Thanks!
Kashyap
-----Original Message-----
From: John Garry [mailto:john.garry@xxxxxxxxxx]
Sent: Thursday, September 17, 2020 1:48 PM
To: Sathya Prakash <sathya.prakash@xxxxxxxxxxxx>;
sreekanth.reddy@xxxxxxxxxxxx; Suganath Prabu Subramani <suganath-
prabu.subramani@xxxxxxxxxxxx>
Cc: MPT-FusionLinux.pdl@xxxxxxxxxxxx; linux-scsi@xxxxxxxxxxxxxxx;
Kashyap Desai <kashyap.desai@xxxxxxxxxxxx>; Hannes Reinecke
<hare@xxxxxxxx>; Martin K . Petersen <martin.petersen@xxxxxxxxxx>;
axboe@xxxxxxxxx; Ming Lei <ming.lei@xxxxxxxxxx>
Subject: mpt3sas hostwide tagset?
Hi guys,
You may have noticed patchset "blk-mq/scsi: Provide hostwide shared tags
for SCSI HBAs", where we allow SCSI HBAs which have restriction of
hostwide
tagset to expose hw queues to blk-mq. Main motivation is to take advantage
of blk-mq CPU hotplug handling support [0], and also possibly [1].
From looking at this driver again, I now notice that mpt3sas seems to
have all
the characteristics of a driver which could make this change:
uses managed interrupts for some completion queues, manages internally
CPU<->queue mapping, does not set Scsi_Host.nr_hw_queues, and also uses
request->tag.
Have you considered making this transition? Not sure if there are
technical
reasons not to.
Thanks,
John
[0] https://lore.kernel.org/linux-block/20200529135315.199230-1-
hch@xxxxxx/
[1]
https://lore.kernel.org/linux-
scsi/CAHsXFKFy+ZVvaCr=H224VGA755k45fAJhz5TaMz+tOP6hNpj1g@xxxxxxx
ail.com/
