linux-kernel - Re: [PATCH RESEND] x86/cpu/topology: remove limit of CPUs due to noapic on x86

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: <8e7f8783-5745-46fd-9a95-a5ed36065f77@riseup.net>
Date: Mon, 25 Nov 2024 16:13:39 +0100
From: "Fernando F. Mancera" <ffmancera@...eup.net>
To: Ingo Molnar <mingo@...nel.org>
Cc: linux-kernel@...r.kernel.org, tglx@...utronix.de, x86@...nel.org
Subject: Re: [PATCH RESEND] x86/cpu/topology: remove limit of CPUs due to
 noapic on x86_64

On 25/11/2024 09:37, Ingo Molnar wrote:
> 
> * Fernando Fernandez Mancera <ffmancera@...eup.net> wrote:
> 
>> On x86_64 the command line parameter "noapic" should not limit the
>> number of possible CPUs, as it only limits the use of IRQ sharing or
>> device IRQ remapping. Only on x86_32 the command line parameter
>> "nolapic" limits the number of possible CPUs to one. This restores the
>> behavior previous to the rework of possible CPU management.
> 
> So what's the motivation? Arguably the x86-64 boot option behavior was
> weird: a working local APIC is very much needed to have an SMP system.
> 

Sorry if I am wrong here but I am not an expert on the matter. I 
believed that "noapic" disables I/O APIC which handles external 
interrupts while the local APICs are still enabled as they are managed 
by "nolapic". If that is the case, SMP should still be possible.

If both I/O APIC and Local APICs are disabled then the kernel should 
fallback to a single CPU mode. That is the behavior that kernel has with 
my patch.

The motivation is to fix multiple users with systems that requires 
"noapic" to work and after the rework their systems are using a single CPU.

References:
[1] https://bugzilla.redhat.com/show_bug.cgi?id=2295026
[2] 
https://discussion.fedoraproject.org/t/fedora-sees-only-1-cpu-core-after-updating-the-kernel-from-6-8-x-to-6-9-x/121385/17

> If we want to disable IRQ sharing or device IRQ remapping, then that
> should have an appropriately named boot command line option. Does some
> system require that perhaps?
> 
> Thanks,
> 
> 	Ingo