Re: nvme: utilize two queue maps, one for reads and one for writes

[Mike Snitzer]

On Tue, Nov 13 2018 at  8:28pm -0500,
Mike Snitzer <snitzer@xxxxxxxxxx> wrote:

> On Tue, Nov 13 2018 at  7:51pm -0500,
> Jens Axboe <axboe@xxxxxxxxx> wrote:
> 
> > On 11/13/18 5:41 PM, Guenter Roeck wrote:
> > > Hi,
> > > 
> > > On Wed, Oct 31, 2018 at 08:36:31AM -0600, Jens Axboe wrote:
> > >> NVMe does round-robin between queues by default, which means that
> > >> sharing a queue map for both reads and writes can be problematic
> > >> in terms of read servicing. It's much easier to flood the queue
> > >> with writes and reduce the read servicing.
> > >>
> > >> Implement two queue maps, one for reads and one for writes. The
> > >> write queue count is configurable through the 'write_queues'
> > >> parameter.
> > >>
> > >> By default, we retain the previous behavior of having a single
> > >> queue set, shared between reads and writes. Setting 'write_queues'
> > >> to a non-zero value will create two queue sets, one for reads and
> > >> one for writes, the latter using the configurable number of
> > >> queues (hardware queue counts permitting).
> > >>
> > >> Reviewed-by: Hannes Reinecke <hare@xxxxxxxx>
> > >> Reviewed-by: Keith Busch <keith.busch@xxxxxxxxx>
> > >> Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
> > > 
> > > This patch causes hangs when running recent versions of
> > > -next with several architectures; see the -next column at
> > > kerneltests.org/builders for details.  Bisect log below; this
> > > was run with qemu on alpha. Reverting this patch as well as
> > > "nvme: add separate poll queue map" fixes the problem.
> > 
> > I don't see anything related to what hung, the trace, and so on.
> > Can you clue me in? Where are the test results with dmesg?
> > 
> > How to reproduce?
> 
> Think Guenter should've provided a full kerneltests.org url, but I had a
> look and found this for powerpc with -next:
> https://kerneltests.org/builders/next-powerpc-next/builds/998/steps/buildcommand/logs/stdio
> 
> Has useful logs of the build failure due to block.

Take that back, of course I only had a quick look and first scrolled to
this fragment and thought "yeap shows block build failure" (not
_really_):

opt/buildbot/slave/next-next/build/kernel/sched/psi.c: In function 'cgroup_move_task':
/opt/buildbot/slave/next-next/build/include/linux/spinlock.h:273:32: warning: 'rq' may be used uninitialized in this function [-Wmaybe-uninitialized]
 #define raw_spin_unlock(lock)  _raw_spin_unlock(lock)
                                ^~~~~~~~~~~~~~~~
/opt/buildbot/slave/next-next/build/kernel/sched/psi.c:639:13: note: 'rq' was declared here
  struct rq *rq;
             ^~