linux-kernel - Re: [PATCH v1 0/5] Add PASID support to SMMUv3 unmanaged domains

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <ZF0PsybSn54A4f/+@nvidia.com>
Date:   Thu, 11 May 2023 12:54:27 -0300
From:   Jason Gunthorpe <jgg@...dia.com>
To:     Robin Murphy <robin.murphy@....com>
Cc:     Michael Shavit <mshavit@...gle.com>, Will Deacon <will@...nel.org>,
        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 0/5] Add PASID support to SMMUv3 unmanaged domains

On Thu, May 11, 2023 at 01:33:58PM +0100, Robin Murphy wrote:
> > > 2. arm_smmu_master should own an s1cfg (which holds a cdtable) that is
> > > used by unmanaged/dma and sva domains attached to this master.
> > 
> > The arm_smmu_master's cd table can be inserted into a steering table
> 
> Not sure what you mean there... STE.S1ContextPtr is essentially just a
> pointer to an array of CD structures (which only contains 1 element when
> PASIDs aren't enabled), so every master must own its own CD table directly.
> There is no viable indirection if you want the abstraction to bear any
> relation to reality.

Yes, this is what I mean. Whenever we need a kernel owned CD table it
comes from the smmu master and is inserted into the steering table
owned by the arm_smmu_device.

"Insert" is just the usual verb we tend to use when talking about
these kinds of structures. Ie a PTE is inserted into a page table and
points at a page - a page table doesn't hold a PTE owned by the page.

So we have, basically, three kinds of tables, Steering/CD/IOPTE, they
are owned by their respective objects
arm_smmu_device/arm_smmu_master/arm_smmu_domain

And we insert pointers from Steering -> CD -> IOPTE as appropriate.

The only case a CD table is not in the arm_smmu_master is for nesting,
but we can still say that the nesting domain is inserted into the
steering table.

Jason