[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZMOzs1/xoEPX2+vA@nvidia.com>
Date: Fri, 28 Jul 2023 09:25:23 -0300
From: Jason Gunthorpe <jgg@...dia.com>
To: Michael Shavit <mshavit@...gle.com>
Cc: Will Deacon <will@...nel.org>, Robin Murphy <robin.murphy@....com>,
Joerg Roedel <joro@...tes.org>, jean-philippe@...aro.org,
nicolinc@...dia.com, baolu.lu@...ux.intel.com,
linux-arm-kernel@...ts.infradead.org, iommu@...ts.linux.dev,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v1 7/7] iommu/arm-smmu-v3: Move CD table to
arm_smmu_master
On Fri, Jul 28, 2023 at 07:11:46PM +0800, Michael Shavit wrote:
> On Fri, Jul 28, 2023 at 2:43 AM Jason Gunthorpe <jgg@...dia.com> wrote:
>
> > Then you can put the free in a detach op once the CD table becomes
> > empty and it behaves much like it already does.
>
> This turns out to be a bit tricky; the SMMU driver detaches the
> currently attached domain whenever a new domain is attached with
> attach_dev().
Oh, yeah, it is a bit quirky that way
> More generally, do we really want to be de-allocating
> the table whenever we switch between an S1 domain and other domain
> types that don't make use of the table (such as IDENTITY or NESTED)?
Probably.
> One solution is to defer the allocation to the first attach_op, but
> only free when the master is freed (this patch's v1 behavior).
That seems reasonable, just don't allocate it at probe time since
PASID is very rarely used.
Jason
Powered by blists - more mailing lists