linux-kernel - Re: [PATCH] leds: trigger: ledtrig-cpu:: enforce NR

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: <20250625103631.GU795775@google.com>
Date: Wed, 25 Jun 2025 11:36:31 +0100
From: Lee Jones <lee@...nel.org>
To: Arnd Bergmann <arnd@...db.de>
Cc: Arnd Bergmann <arnd@...nel.org>, Pavel Machek <pavel@...nel.org>,
	Hans de Goede <hansg@...nel.org>, linux-leds@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] leds: trigger: ledtrig-cpu:: enforce NR_CPUS limit in
 Kconfig

On Wed, 25 Jun 2025, Arnd Bergmann wrote:

> On Wed, Jun 25, 2025, at 11:24, Lee Jones wrote:
> > On Fri, 20 Jun 2025, Arnd Bergmann wrote:
> >> 
> >> Enforce this limit in Kconfig instead to avoid the build failure.
> >
> > I have so many questions!
> >
> > - This number seems arbitrary - what is the limiting factor?
> >   - Character size for printing?
> 
> That is my best guess, yes. The original commit 8f88731d052d
> ("led-triggers: create a trigger for CPU activity") already
> had this, plus
> 
> #define MAX_NAME_LEN 8
> snprintf(trig->name, MAX_NAME_LEN, "cpu%d", cpu);
> 
> which comes out to 10000 cpus (0 through 9999) in a NUL-terminated
> string. Obviously we could have a higher limit here but would
> technically still need to check that.
> 
> > - What platform did you test this on to blow through that number?
> 
> I was only compile-testing with a crazy number of CPUs, to see
> what would break, in particular when there are per-cpu structures
> or cpumask_t variables on the stack. I sent fixes for the ones
> that broke the build, but don't expect anything to actually
> need this at runtime. Some of the fixes I sent also address
> possible stack overflows that happen with a more realistic
> number of CPUs.
> 
> > - What if we really do want 10000+ CPUs?
> >   - And what will that LED array look like to support them all?
> 
> abcc131292aa ("ledtrig-cpu: Limit to 8 CPUs") actually added a
> runtime limit to eight CPUS, which is probably more reasonable.
> 
> > - If we're enforcing here, is the BUILD_BUG_ON() now superfluous?
> 
> Yes, but it documents the requirement better than just the
> Kconfig dependency. Since there is a runtime limit, we could
> probably just drop the compile-time check and not add the
> Kconfig check.

Your call.  I trust your judgement.

All of this was just a thought experiment.

-- 
Lee Jones [李琼斯]