linux-kernel - Re: [PATCH v3 2/2] x86/smpboot: avoid SMT domain attach/destroy if SMT is not enabled

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: <20250625082932.GA1613200@noisy.programming.kicks-ass.net>
Date: Wed, 25 Jun 2025 10:29:32 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Li Chen <me@...ux.beauty>
Cc: Thomas Gleixner <tglx@...utronix.de>, Ingo Molnar <mingo@...hat.com>,
	Borislav Petkov <bp@...en8.de>,
	Dave Hansen <dave.hansen@...ux.intel.com>, x86@...nel.org,
	"H . Peter Anvin" <hpa@...or.com>,
	"Rafael J . Wysocki" <rafael.j.wysocki@...el.com>,
	K Prateek Nayak <kprateek.nayak@....com>,
	Sohil Mehta <sohil.mehta@...el.com>,
	Brian Gerst <brgerst@...il.com>,
	Patryk Wlazlyn <patryk.wlazlyn@...ux.intel.com>,
	linux-kernel@...r.kernel.org,
	"Gautham R. Shenoy" <gautham.shenoy@....com>,
	Li Chen <chenl311@...natelecom.cn>
Subject: Re: [PATCH v3 2/2] x86/smpboot: avoid SMT domain attach/destroy if
 SMT is not enabled

On Wed, Jun 25, 2025 at 11:45:50AM +0800, Li Chen wrote:
> From: Li Chen <chenl311@...natelecom.cn>
> 
> Currently, the SMT domain is added into sched_domain_topology by default.
> 
> If cpu_attach_domain() finds that the CPU SMT domain’s cpumask_weight
> is just 1, it will destroy it.
> 
> On a large machine, such as one with 512 cores, this results in
> 512 redundant domain attach/destroy operations.
> 
> Avoid these unnecessary operations by simply checking
> cpu_smt_num_threads and remove SMT domain if the SMT domain is not
> enabled, and adjust the PKG index accordingly if NUMA-in-package
> invalidates that level as well.
> 
> Signed-off-by: Li Chen <chenl311@...natelecom.cn>
> ---
> changelog:
> v2: fix wording issue as suggested by Thomas [1]
> v3: remove pointless memset and adjust PKG index accordingly,
>     as suggested by Thomas [2] 
> 
> [1]: https://lore.kernel.org/all/87msa2r018.ffs@tglx/
> [2]: https://lore.kernel.org/all/875xglntx1.ffs@tglx/
> 
>  arch/x86/kernel/smpboot.c | 18 +++++++++++++++++-
>  1 file changed, 17 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
> index 7d202f9785362..4b6daa1545445 100644
> --- a/arch/x86/kernel/smpboot.c
> +++ b/arch/x86/kernel/smpboot.c
> @@ -494,13 +494,29 @@ static struct sched_domain_topology_level x86_topology[] = {
>  
>  static void __init build_sched_topology(void)
>  {
> +	bool smt_dropped = false;
> +
> +	if (cpu_smt_num_threads <= 1) {
> +		/*
> +		 * SMT level is x86_topology[0].  Shift the array left by one,
> +		 */
> +		memmove(&x86_topology[0], &x86_topology[1],
> +			sizeof(x86_topology) - sizeof(x86_topology[0]));
> +		smt_dropped = true;
> +	}
> +
>  	/*
>  	 * When there is NUMA topology inside the package invalidate the
>  	 * PKG domain since the NUMA domains will auto-magically create the
>  	 * right spanning domains based on the SLIT.
>  	 */
>  	if (x86_has_numa_in_package) {
> -		unsigned int pkgdom = ARRAY_SIZE(x86_topology) - 2;
> +		unsigned int pkgdom;
> +
> +		if (smt_dropped)
> +			pkgdom = ARRAY_SIZE(x86_topology) - 3;
> +		else
> +			pkgdom = ARRAY_SIZE(x86_topology) - 2;
>  
>  		memset(&x86_topology[pkgdom], 0, sizeof(x86_topology[pkgdom]));
>  	}

Oh gawd, and you all really think this is better than dynamically
creating that array?

This is quite terrible.