linux-kernel - Re: [PATCH] iommu/vt-d: Check for non-NULL domain on device release

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: Wed, 21 Feb 2024 11:40:12 -0400
From: Jason Gunthorpe <jgg@...pe.ca>
To: Baolu Lu <baolu.lu@...ux.intel.com>
Cc: Eric Badger <ebadger@...estorage.com>,
	David Woodhouse <dwmw2@...radead.org>,
	Joerg Roedel <joro@...tes.org>, Will Deacon <will@...nel.org>,
	Robin Murphy <robin.murphy@....com>,
	"open list:INTEL IOMMU (VT-d)" <iommu@...ts.linux.dev>,
	open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] iommu/vt-d: Check for non-NULL domain on device release

On Wed, Jan 31, 2024 at 03:10:53PM +0800, Baolu Lu wrote:
> I like this suggestion.
> 
> Currently, the device_release callback for an iommu driver does the
> following:
> 
> a) Silent IOMMU DMA translation. This is done by detaching the existing
>    domain from the IOMMU and bringing the IOMMU into a blocking state
>    (some drivers might be in identity state);
> b) Releases the resources allocated in the probe callback and restores
>    the device to its previous state before the probe.
> 
> From my understanding of your suggestion, we should move a) out of the
> release callback and make it a special domain, which could be a blocking
> domain or identity domain, depending on the iommu hardware.
> 
> In the end, the release_device callback of an iommu driver should focus
> on the opposite operation of device_probe. This makes the concept
> clearer.

Right

Can someone make some patches to fix Eric's bug? We don't really need
to do the release_domain stuff if the driver just self-attaches one of
its known static domain types (blocking/identity)

Jason