linux-kernel - Re: [PATCH v5 5/9] iommu/arm-smmu-v3: Refactor write_ctx

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, 15 Aug 2023 19:40:14 +0800
From:   Michael Shavit <mshavit@...gle.com>
To:     Will Deacon <will@...nel.org>
Cc:     iommu@...ts.linux.dev, linux-arm-kernel@...ts.infradead.org,
        linux-kernel@...r.kernel.org, robin.murphy@....com,
        nicolinc@...dia.com, jgg@...dia.com, jean-philippe@...aro.org
Subject: Re: [PATCH v5 5/9] iommu/arm-smmu-v3: Refactor write_ctx_desc

On Tue, Aug 15, 2023 at 6:19 PM Will Deacon <will@...nel.org> wrote:
>
> On Tue, Aug 15, 2023 at 01:04:43PM +0800, Michael Shavit wrote:
> > On Thu, Aug 10, 2023 at 5:15 PM Michael Shavit <mshavit@...gle.com> wrote:
> > > On Wed, Aug 9, 2023 at 9:50 PM Will Deacon <will@...nel.org> wrote:
> > > >
> > > > Since you're dropping this and relying on the lock being taken higher up
> > > > callstack, can we add a lockdep assertion that we do actually hold the
> > > > devices_lock, please?
> > >
> > > Will do!
> >
> > I spoke too soon; the point of this change was to remove the
> > dependency on the arm_smmu_domain, piping the devices_lock would
> > defeat this. In fact, this section is really depending on the iommu
> > group lock not the devices_lock.
>
> Sorry, but I'm not following you here. What is the devices_lock protecting
> if we're depending on the group lock?
>
> Will

Ugh, yes sorry, we do need the devices_lock held *in the SVA* call
flow. The group lock is ineffective there because SVA performs
cross-master operations on iommu_domain callbacks. There, the
devices_lock is needed to ensure that the list of masters that a
domain is attached to doesn't mutate while we operate on that domain.

Nonetheless, it feels awkward to require the devices_lock to be held
outside SVA, since there we operate on a single master/CD table
rather. We don't care what other masters that domain is attached to.