linux-kernel - Re: [PATCH 7/7] iommu: Add iommu_domain::domain

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]

Date:   Tue, 25 Jan 2022 11:00:46 -0400
From:   Jason Gunthorpe <jgg@...dia.com>
To:     Robin Murphy <robin.murphy@....com>
Cc:     Lu Baolu <baolu.lu@...ux.intel.com>,
        Joerg Roedel <joro@...tes.org>,
        Christoph Hellwig <hch@...radead.org>,
        Ben Skeggs <bskeggs@...hat.com>,
        Kevin Tian <kevin.tian@...el.com>,
        Ashok Raj <ashok.raj@...el.com>, Will Deacon <will@...nel.org>,
        Alex Williamson <alex.williamson@...hat.com>,
        Eric Auger <eric.auger@...hat.com>,
        Liu Yi L <yi.l.liu@...el.com>,
        Jacob jun Pan <jacob.jun.pan@...el.com>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>,
        Thierry Reding <thierry.reding@...il.com>,
        Jonathan Hunter <jonathanh@...dia.com>,
        iommu@...ts.linux-foundation.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 7/7] iommu: Add iommu_domain::domain_ops

On Tue, Jan 25, 2022 at 02:23:52PM +0000, Robin Murphy wrote:
> On 2022-01-25 06:27, Lu Baolu wrote:

> Where it's just about which operations are valid for which domains, it's
> even simpler for the core interface wrappers to validate the domain type,
> rather than forcing drivers to implement multiple ops structures purely for
> the sake of having different callbacks populated. We already have this in
> places, e.g. where iommu_map() checks for __IOMMU_DOMAIN_PAGING.

In my experience it is usually much clearer to directly test the op
for NULL to know if a feature is supported than to invent flags to do
the same test.  eg ops->map/etc == NULL means no paging.

I think we should not be afraid to have multiple ops in drivers for
things that are actually different in the driver. This is usually a
net win vs tying to handle all the cases with different 'if' flows.

eg identity domains and others really would ideally eventually have a
NULL ops for map/unmap too.

The 'type' should conceptually be part of the ops, not the mutable
struct - but we don't have to get there all at once.

> Paging domains are also effectively the baseline level of IOMMU API
> functionality. All drivers support them, and for the majority of drivers
> it's all they will ever support. Those drivers really don't benefit from any
> of the churn and boilerplate in this patch as-is, and it's so easy to
> compromise with a couple of lines of core code to handle the common case by
> default when the driver *isn't* one of the handful which ever actually cares
> to install their own per-domain ops. Consider how much cleaner this patch
> would look if the typical driver diff could be something completely minimal
> like this:

It is clever, but I'm not sure if hoisting a single assignment out of
the driver is worth the small long term complexity of having different
driver flows?

Jason