Re: [PATCH v8 0/7] Add non-strict mode support for iommu-dma

[John Garry]

On 21/09/2018 12:03, Robin Murphy wrote:
> On 21/09/18 10:29, Robin Murphy wrote:
> > Hi John,
> >
> > On 2018-09-21 10:20 AM, John Garry wrote:
> > > On 20/09/2018 17:10, Robin Murphy wrote:
> > > > Hi all,
> > > >
> > > > Hopefully this is the last spin of the series - I've now dropped my
> > > > light
> > > > touch and fixed up all the various prose text, plus implemented the
> > > > proper
> > > > quirk support for short-descriptor because it's actually just a trivial
> > > > cut-and-paste job.
> > > >
> > > > Robin.
> > >
> > > Hi Robin,
> > >
> > > JFYI, I'm trying to test this patchset to get some figures and
> > > provide a tested-by tag, but 4/8 seems to rely on
> > > "iommu/io-pgtable-arm: Fix race handling in split_blk_unmap()" - more
> > > specifically, it seems to rely on the version which Will rewrote in
> > > your patch review, and I am not sure on what branch it exists on, if
> > > any.
> >
> > Sorry, I should have said this is based on Will's iommu/devel branch:
> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/will/linux.git/log/?h=iommu/devel
>
>
> FWIW I've now pushed out a complete branch here:
>
>   git://linux-arm.org/linux-rm iommu/non-strict

Cheers

So for my network test scenario I was getting a boost @ 250K vs 160K 
packet(s)/second with strict off/on

For NVMe single disk performance, I was getting a boost @ 582K vs 370K 
IOPS with strict off/on.

I wasn't seeing such a boost for other storage controller scenario 
(that's with 6 SSDs), with 776K vs 740K IOPS for strict off/on, but SMMU 
off was ~800K IOPS.

FWIW:
Tested-by: John Garry <john.garry@xxxxxxxxxx>

Thanks,
John

> Robin.
>
> .