Re: amd-iommu / agpgart-amd64 problem: Resources present before probing

[Lukas Wunner <lukas@wunner.de>]

On Tue, Jun 17, 2025 at 10:47:48PM +0300, Fedor Pchelkin wrote:
> Hello,
> 
> there is a 
> 
>   [    0.579232] pci 0000:00:00.2: Resources present before probing
> 
> error message observed after
> 
>   commit 3be5fa236649da6404f1bca1491bf02d4b0d5cce
>   Author: Lukas Wunner <lukas@wunner.de>
>   Date:   Fri Apr 25 11:24:21 2025 +0200
>   
>       Revert "iommu/amd: Prevent binding other PCI drivers to IOMMU PCI devices"

For the record, the reporter of the above-quoted issue appears to be
working for an OFAC sanctioned entity:

https://sanctionssearch.ofac.treas.gov/Details.aspx?id=50890

This prohibits me from two-way engagement with the reporter:

https://www.linuxfoundation.org/blog/navigating-global-regulations-and-open-source-us-ofac-sanctions

   "Reviewing an unsolicited patch from a contributor in a sanctioned
    region should generally be fine, but actively engaging them to
    better understand their issue, diagnose the problem, or help
    improve a patch or modify code would likely cross the line.
    If the contributor is linked to a sanctioned entity or region,
    in general, it is best to keep communications strictly one-way.
    If a patch is received and you improve it and submit it upstream,
    that should be fine, but going back and forth in communications
    with the SDN developer likely would not."

Hence I am removing the reporter and the lvc-project@linuxtesting.org
address (hosted by ispras.ru) from the To: and Cc: headers.

I note that prior to 6fd024893911, the amd64-agp.c driver was only bound
to devices with a PCI_CAP_ID_AGP capability.

agp_amd64_probe() does check for presence of the capability, but that's
too late to avoid the error message emitted by really_probe().

What we could do however is to first check for presence of a device with
PCI_CAP_ID_AGP, and only if one is found would we try to bind to any
device.  That should avoid the message on any halfway modern system.

Thoughts?

Thanks,

Lukas