linux-kernel - Re: [PATCH v3 4/8] x86/psp: Add IRQ 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] [thread-next>] [day] [month] [year] [list]

Message-ID: <87h6ue9z0p.ffs@tglx>
Date:   Tue, 21 Mar 2023 11:31:18 +0100
From:   Thomas Gleixner <tglx@...utronix.de>
To:     Jeremi Piotrowski <jpiotrowski@...ux.microsoft.com>,
        linux-kernel@...r.kernel.org
Cc:     Jeremi Piotrowski <jpiotrowski@...ux.microsoft.com>,
        Brijesh Singh <brijesh.singh@....com>,
        Tom Lendacky <thomas.lendacky@....com>,
        "Kalra, Ashish" <ashish.kalra@....com>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        Dave Hansen <dave.hansen@...ux.intel.com>, x86@...nel.org
Subject: Re: [PATCH v3 4/8] x86/psp: Add IRQ support

On Mon, Mar 20 2023 at 19:19, Jeremi Piotrowski wrote:
> The ACPI PSP device provides a mailbox irq that needs to be configured
> through the ACPI mailbox register first. This requires passing a CPU
> vector and physical CPU id and then enabling interrupt delivery.
> Allocate the irq directly from the default irq domain
> (x86_vector_domain) to get access to the required information. By
> passing a cpumask through irq_alloc_info the vector is immediately
> allocated (and not later during activation) and can be retrieved.

Sorry, but this is a horrible hack which violates _all_ design rules
for interrupts in one go.

 1) What's so special about this PSP device that it requires a vector
    directly from the vector domain and evades interrupt remapping?

 2) Why is this interrupt enabled _before_ it is actually requested?

 3) Why is this interrupt required to be bound to CPU0 and still exposes
    a disfunctional and broken affinity setter interface in /proc?

There is absolutely zero reason and justification to fiddle in the guts
of the x86 vector configuration data just because it's possible.

This is clearly a custom MSI implementation and the obvious solution is
a per device MSI interrupt domain.

Thanks,

        tglx