linux-kernel - Re: [PATCH] iommu/arm-smmu-v3: Handle duplicated Stream IDs from other masters

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:   Thu, 7 Jan 2021 13:03:19 +0000
From:   Will Deacon <will@...nel.org>
To:     Ajay Kumar <ajaykumar.rs@...sung.com>
Cc:     iommu@...ts.linux-foundation.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        joro@...tes.org, robh+dt@...nel.org, mark.rutland@....com,
        robin.murphy@....com, john.garry@...wei.com
Subject: Re: [PATCH] iommu/arm-smmu-v3: Handle duplicated Stream IDs from
 other masters

On Thu, Jan 07, 2021 at 03:03:40PM +0530, Ajay Kumar wrote:
> When PCI function drivers(ex:pci-endpoint-test) are probed for already
> initialized PCIe-RC(Root Complex), and PCIe-RC is already bound to SMMU,
> then we encounter a situation where the function driver tries to attach
> itself to the smmu with the same stream-id as PCIe-RC and re-initialize
> an already initialized STE. This causes ste_live BUG_ON() in the driver.

I don't understand why the endpoint is using the same stream ID as the root
complex in this case. Why is that? Is the grouping logic not working
properly?

> There is an already existing check in the driver to manage duplicated ids
> if duplicated ids are added in same master device, but there can be
> scenarios like above where we need to extend the check for other masters
> using the same stream-id.
> 
> Signed-off-by: Ajay Kumar <ajaykumar.rs@...sung.com>
> ---
>  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 33 +++++++++++++++++++++
>  1 file changed, 33 insertions(+)

It doesn't feel like the driver is the right place to fix this, as the same
issue could surely occur for other IOMMUs too, right? In which case, I think
we should avoid getting into the situation where different groups have
overlapping stream IDs.

Will