Re: [PATCH 4/6] iommu: Combine device strictness requests with the global default

From: Doug Anderson <dianders@chromium.org>
To: Lu Baolu <baolu.lu@linux.intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	"Rafael J. Wysocki" <rafael@kernel.org>,
	"Rafael J. Wysocki" <rafael.j.wysocki@intel.com>,
	Will Deacon <will@kernel.org>,
	Robin Murphy <robin.murphy@arm.com>,
	Joerg Roedel <joro@8bytes.org>,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	Ulf Hansson <ulf.hansson@linaro.org>,
	Adrian Hunter <adrian.hunter@intel.com>,
	Bjorn Helgaas <bhelgaas@google.com>,
	Rob Clark <robdclark@chromium.org>,
	LKML <linux-kernel@vger.kernel.org>,
	Saravana Kannan <saravanak@google.com>,
	linux-arm-msm <linux-arm-msm@vger.kernel.org>,
	Linux MMC List <linux-mmc@vger.kernel.org>,
	quic_c_gdjako@quicinc.com,
	"list@263.net:IOMMU DRIVERS <iommu@lists.linux-foundation.org>,
	Joerg Roedel <joro@8bytes.org>,"
	<iommu@lists.linux-foundation.org>,
	linux-pci@vger.kernel.org,
	Joel Fernandes <joel@joelfernandes.org>,
	Rajat Jain <rajatja@google.com>,
	Sonny Rao <sonnyrao@chromium.org>,
	Veerabhadrarao Badiganti <vbadigan@codeaurora.org>
Subject: Re: [PATCH 4/6] iommu: Combine device strictness requests with the global default
Date: Tue, 22 Jun 2021 10:01:51 -0700	[thread overview]
Message-ID: <CAD=FV=Xi8Qai5sKwuT4-4B=K5i7f4BWQfp9+TxR1x3VSt7dkGA@mail.gmail.com> (raw)
In-Reply-To: <CAD=FV=XpYkUqGNcumJ0NvoXqbSkBaV5ZadUnpsKTMPx7tSjGig@mail.gmail.com>

Hi,

On Tue, Jun 22, 2021 at 9:53 AM Doug Anderson <dianders@chromium.org> wrote:
>
> Hi,
>
> On Mon, Jun 21, 2021 at 7:05 PM Lu Baolu <baolu.lu@linux.intel.com> wrote:
> >
> > On 6/22/21 7:52 AM, Douglas Anderson wrote:
> > > @@ -1519,7 +1542,8 @@ static int iommu_get_def_domain_type(struct device *dev)
> > >
> > >   static int iommu_group_alloc_default_domain(struct bus_type *bus,
> > >                                           struct iommu_group *group,
> > > -                                         unsigned int type)
> > > +                                         unsigned int type,
> > > +                                         struct device *dev)
> > >   {
> > >       struct iommu_domain *dom;
> > >
> > > @@ -1534,6 +1558,12 @@ static int iommu_group_alloc_default_domain(struct bus_type *bus,
> > >       if (!dom)
> > >               return -ENOMEM;
> > >
> > > +     /* Save the strictness requests from the device */
> > > +     if (dev && type == IOMMU_DOMAIN_DMA) {
> > > +             dom->request_non_strict = dev->request_non_strict_iommu;
> > > +             dom->force_strict = dev->force_strict_iommu;
> > > +     }
> > > +
> >
> > An iommu default domain might be used by multiple devices which might
> > have different "strict" attributions. Then who could override who?
>
> My gut instinct would be that if multiple devices were part of a given
> domain that it would be combined like this:
>
> 1. Any device that requests strict makes the domain strict force strict.
>
> 2. To request non-strict all of the devices in the domain would have
> to request non-strict.
>
> To do that I'd have to change my patchset obviously, but I don't think
> it should be hard. We can just keep a count of devices and a count of
> the strict vs. non-strict requests? If there are no other blockers
> I'll try to do that in my v2.

One issue, I guess, is that we might need to transition a non-strict
domain to become strict. Is that possible? We'd end up with an extra
"flush queue" that we didn't need to allocate, but are there other
problems?

Actually, in general would it be possible to transition between strict
and non-strict at runtime as long as we called
init_iova_flush_queue()? Maybe that's a better solution than all
this--we just boot in strict mode and can just transition to
non-strict mode after-the-fact?

-Doug