linux-kernel - Re: [RFC PATCH 2/5] iommu/vt-d: Add generic IO page table support

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] [day] [month] [year] [list]

Message-ID: <20231109003304.GG4634@ziepe.ca>
Date:   Wed, 8 Nov 2023 20:33:04 -0400
From:   Jason Gunthorpe <jgg@...pe.ca>
To:     "Zhang, Tina" <tina.zhang@...el.com>
Cc:     Jean-Philippe Brucker <jean-philippe@...aro.org>,
        "Tian, Kevin" <kevin.tian@...el.com>,
        Lu Baolu <baolu.lu@...ux.intel.com>,
        "joro@...tes.org" <joro@...tes.org>,
        "will@...nel.org" <will@...nel.org>,
        "Liu, Yi L" <yi.l.liu@...el.com>,
        "virtualization@...ts.linux-foundation.org" 
        <virtualization@...ts.linux-foundation.org>,
        "iommu@...ts.linux.dev" <iommu@...ts.linux.dev>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [RFC PATCH 2/5] iommu/vt-d: Add generic IO page table support

On Thu, Nov 09, 2023 at 12:10:59AM +0000, Zhang, Tina wrote:

> > If this is going to happen can we also convert vt-d to actually use the io page
> > table stuff directly and shuffle the code around so it is structured like the rest of
> > the io page table implementations?

> Converting VT-d driver to use io page table involves much code
> change. I made a local version of it, and it didn't prove much
> benefit.

Well, it structures the code in a similar way to all the other
drivers, though I admit to having not looked closely at the io page
table stuff.
 
> VT-d only supports one 1st-stage IO pgtable format and one 2nd-stage
> IO pgtable format. So, the current IO pgtable handling operations
> seems more efficient comparing to adding another layer callbacks in
> them.

I would like to de-virtualize those callbacks, is is completely
pointless when we have per-format iommu_domain ops now.

Jason