Re: [PATCH 06/14] block: lift setting the readahead size into the block layer

From: Mike Snitzer <snitzer@redhat.com>
To: Christoph Hellwig <hch@lst.de>
Cc: Jens Axboe <axboe@kernel.dk>,
	linux-block@vger.kernel.org, martin.petersen@oracle.com,
	Hans de Goede <hdegoede@redhat.com>, Song Liu <song@kernel.org>,
	Richard Weinberger <richard@nod.at>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-raid@vger.kernel.org, Minchan Kim <minchan@kernel.org>,
	dm-devel@redhat.com, linux-mtd@lists.infradead.org,
	linux-mm@kvack.org, drbd-dev@tron.linbit.com,
	cgroups@vger.kernel.org
Subject: Re: [PATCH 06/14] block: lift setting the readahead size into the block layer
Date: Wed, 2 Sep 2020 12:20:07 -0400	[thread overview]
Message-ID: <20200902162007.GB5513@redhat.com> (raw)
In-Reply-To: <20200902151144.GA1738@lst.de>

On Wed, Sep 02 2020 at 11:11am -0400,
Christoph Hellwig <hch@lst.de> wrote:

> On Wed, Aug 26, 2020 at 06:07:38PM -0400, Mike Snitzer wrote:
> > On Sun, Jul 26 2020 at 11:03am -0400,
> > Christoph Hellwig <hch@lst.de> wrote:
> > 
> > > Drivers shouldn't really mess with the readahead size, as that is a VM
> > > concept.  Instead set it based on the optimal I/O size by lifting the
> > > algorithm from the md driver when registering the disk.  Also set
> > > bdi->io_pages there as well by applying the same scheme based on
> > > max_sectors.
> > > 
> > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > > ---
> > >  block/blk-settings.c         |  5 ++---
> > >  block/blk-sysfs.c            |  1 -
> > >  block/genhd.c                | 13 +++++++++++--
> > >  drivers/block/aoe/aoeblk.c   |  2 --
> > >  drivers/block/drbd/drbd_nl.c | 12 +-----------
> > >  drivers/md/bcache/super.c    |  4 ----
> > >  drivers/md/dm-table.c        |  3 ---
> > >  drivers/md/raid0.c           | 16 ----------------
> > >  drivers/md/raid10.c          | 24 +-----------------------
> > >  drivers/md/raid5.c           | 13 +------------
> > >  10 files changed, 16 insertions(+), 77 deletions(-)
> > 
> > 
> > In general these changes need a solid audit relative to stacking
> > drivers.  That is, the limits stacking methods (blk_stack_limits)
> > vs lower level allocation methods (__device_add_disk).
> > 
> > You optimized for lowlevel __device_add_disk establishing the bdi's
> > ra_pages and io_pages.  That is at the beginning of disk allocation,
> > well before any build up of stacking driver's queue_io_opt() -- which
> > was previously done in disk_stack_limits or driver specific methods
> > (e.g. dm_table_set_restrictions) that are called _after_ all the limits
> > stacking occurs.
> > 
> > By inverting the setting of the bdi's ra_pages and io_pages to be done
> > so early in __device_add_disk it'll break properly setting these values
> > for at least DM afaict.
> 
> ra_pages never got inherited by stacking drivers, check it by modifying
> it on an underlying device and then creating a trivial dm or md one.

Sure, not saying that it did.  But if the goal is to set ra_pages based
on io_opt then to do that correctly on stacking drivers it must be done
in terms of limits stacking right?  Or at least done at a location that
is after the limits stacking has occurred?  So should DM just open-code
setting ra_pages like it did for io_pages?

Because setting ra_pages in __device_add_disk() is way too early for DM
-- given it uses device_add_disk_no_queue_reg via add_disk_no_queue_reg
at DM device creation (before stacking all underlying devices' limits).

> And I think that is a good thing - in general we shouldn't really mess
> with this thing from drivers if we can avoid it.  I've kept the legacy
> aoe and md parity raid cases, out of which the first looks pretty weird
> and the md one at least remotely sensible.

I don't want drivers, like DM, to have to worry about these.  So I agree
with that goal ;)

> ->io_pages is still inherited in disk_stack_limits, just like before
> so no change either.

I'm missing where, but I only looked closer at this 06/14 patch.
In it I see io_pages is no longer adjusted in disk_stack_limits().

Mike