Re: [PATCH 1/3] blk-mq: add an API to estimate hardware queue node

[Jens Axboe]

On 03/29/2016 10:47 AM, Shaohua Li wrote:
> On Tue, Mar 29, 2016 at 12:24:43AM -0700, Christoph Hellwig wrote:
> > On Fri, Mar 25, 2016 at 02:36:30PM -0700, Shaohua Li wrote:
> > > we allocate most data structure in device's node, but some data
> > > structures are not for DMA and mostly used by specific cpus/node which
> > > could diff from device's node. Allocating such hot data in device's
> > > node doesn't make sense. Add an API to estimate hardware queue node.
> > > This can be used before blk-mq actually establishes the mapping. This
> > > API runs slow, but it only used in initialization time.
> >
> > I think this is the wrong way around.  I've got some proprotype code
> > that just leaves the cpu assignments to the drivers and picks it up
> > in blk-mq.  Give me a few days to post it..
>
> This looks weird, shouldn't the cpu assignment be determined by block
> core (blk-mq) because block core decides how to use the queue?

I agree, that belongs in the blk-mq proper, the driver should just 
follow the rules outlined, not impose their own in this regard. It'll 
also help with irq affinity mappings, once we get that in.

--
Jens Axboe