linux-kernel - Re: [PATCH] of/device: add blacklist for iommu dma

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:   Mon, 3 Jun 2019 08:48:51 -0600
From:   Jordan Crouse <jcrouse@...eaurora.org>
To:     Thierry Reding <thierry.reding@...il.com>
Cc:     Rob Clark <robdclark@...il.com>, devicetree@...r.kernel.org,
        David Airlie <airlied@...ux.ie>,
        Frank Rowand <frowand.list@...il.com>,
        Will Deacon <will.deacon@....com>,
        Doug Anderson <dianders@...omium.org>,
        dri-devel <dri-devel@...ts.freedesktop.org>,
        Linux IOMMU <iommu@...ts.linux-foundation.org>,
        Rob Herring <robh+dt@...nel.org>,
        Sean Paul <seanpaul@...omium.org>,
        linux-arm-msm <linux-arm-msm@...r.kernel.org>,
        freedreno <freedreno@...ts.freedesktop.org>,
        Robin Murphy <robin.murphy@....com>,
        Christoph Hellwig <hch@....de>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] of/device: add blacklist for iommu dma_ops

> It shouldn't be a problem to hook something else up to the IOMMU
> subsystem. Hopefully it's something that people are going to standardize
> on.
> 
> > 3) The automatic attach of DMA domain is also causing a different
> >    problem for us on the GPU side, preventing us from supporting per-
> >    context pagetables (since we end up with a disagreement about
> >    which context bank is used between arm-smmu and the firmware).
> 
> I'm not sure I understand this issue. Is the context bank hard-coded in
> the firmware somehow? Or is it possible to rewrite which one is going to
> be used at runtime? Do you switch out the actual page tables rather than
> the IOMMU domains for context switching?
 
We have a rather long history on this but the tl;dr is that the GPU microcode
switches the pagetables by rewriting TTBR0 on the fly (since this is
arm-smmu-v2 we have no better option) and yes, unfortunately it is hard coded
to use context bank 0. [1] is the current patchset to support all this,
including my own take on avoiding the dma-domain (all the cool kids have one).

Jordan

[1] https://patchwork.freedesktop.org/series/57441/

-- 
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project