linux-kernel - Re: [PATCH v1 1/7] sched/isolation: Add infrastructure to adjust affinity for dynamic CPU isolation

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]

Date: Fri, 17 May 2024 23:37:57 +0200
From: Thomas Gleixner <tglx@...utronix.de>
To: Costa Shulyupin <costa.shul@...hat.com>, longman@...hat.com,
 pauld@...hat.com, juri.lelli@...hat.com, prarit@...hat.com,
 vschneid@...hat.com, Anna-Maria Behnsen <anna-maria@...utronix.de>,
 Frederic Weisbecker <frederic@...nel.org>, Zefan Li
 <lizefan.x@...edance.com>, Tejun Heo <tj@...nel.org>, Johannes Weiner
 <hannes@...xchg.org>, Ingo Molnar <mingo@...hat.com>, Peter Zijlstra
 <peterz@...radead.org>, Vincent Guittot <vincent.guittot@...aro.org>,
 Dietmar Eggemann <dietmar.eggemann@....com>, Steven Rostedt
 <rostedt@...dmis.org>, Ben Segall <bsegall@...gle.com>, Mel Gorman
 <mgorman@...e.de>, Daniel Bristot de Oliveira <bristot@...hat.com>, Petr
 Mladek <pmladek@...e.com>, Andrew Morton <akpm@...ux-foundation.org>,
 Masahiro Yamada <masahiroy@...nel.org>, Randy Dunlap
 <rdunlap@...radead.org>, Yoann Congal <yoann.congal@...le.fr>, "Gustavo A.
 R. Silva" <gustavoars@...nel.org>, Nhat Pham <nphamcs@...il.com>, Costa
 Shulyupin <costa.shul@...hat.com>, linux-kernel@...r.kernel.org,
 cgroups@...r.kernel.org
Subject: Re: [PATCH v1 1/7] sched/isolation: Add infrastructure to adjust
 affinity for dynamic CPU isolation

On Thu, May 16 2024 at 22:04, Costa Shulyupin wrote:
> Introduce infrastructure function housekeeping_update() to change
> housekeeping_cpumask during runtime and adjust affinities of depended
> subsystems.
>
> Affinity adjustments of subsystems follow in subsequent patches.
>
> Parent patch:
> "sched/isolation: Exclude dynamically isolated CPUs from housekeeping masks"
> https://lore.kernel.org/lkml/20240229021414.508972-2-longman@redhat.com/
>
> Test example for cgroup2:
>
> cd /sys/fs/cgroup/
> echo +cpuset > cgroup.subtree_control
> mkdir test
> echo isolated > test/cpuset.cpus.partition
> echo $isolate > test/cpuset.cpus

This changelog is not telling me anything. Please see
Documentation/process/ what changelogs should contain.

> +/*
> + * housekeeping_update - change housekeeping.cpumasks[type] and propagate the
> + * change.
> + *
> + * Assuming cpuset_mutex is held in sched_partition_write or
> + * cpuset_write_resmask.

Locking cannot be assumed. lockdep_assert_held() is there to document
and enforce such requirements.

> + */
> +static int housekeeping_update(enum hk_type type, cpumask_var_t update)

Please us 'struct cpumask *update' as it makes it clear what this is
about. cpumask_var_t is a hack to make onstack and embedded cpumask and
their allocated counterparts possible without #ifdeffery in the code.

But any function which is not related to alloc/free of cpumask_var_t
should simply use 'struct cpumask *' as argument type.

> +	housekeeping.flags |= BIT(type);

The existing code uses WRITE_ONCE() probably for a reason. Why is that
not longer required here?

>  static int __init housekeeping_setup(char *str, unsigned long flags)
>  {
>  	cpumask_var_t non_housekeeping_mask, housekeeping_staging;
> @@ -314,9 +347,12 @@ int housekeeping_exlude_isolcpus(const struct cpumask *isolcpus, unsigned long f
>  		/*
>  		 * Reset housekeeping to bootup default
>  		 */
> -		for_each_set_bit(type, &housekeeping_boot.flags, HK_TYPE_MAX)
> -			cpumask_copy(housekeeping.cpumasks[type],
> -				     housekeeping_boot.cpumasks[type]);
> +		for_each_set_bit(type, &housekeeping_boot.flags, HK_TYPE_MAX) {
> +			int err = housekeeping_update(type, housekeeping_boot.cpumasks[type]);
> +
> +			if (err)
> +				return err;
> +		}
>  
>  		WRITE_ONCE(housekeeping.flags, housekeeping_boot.flags);
>  		if (!housekeeping_boot.flags &&
> @@ -344,9 +380,11 @@ int housekeeping_exlude_isolcpus(const struct cpumask *isolcpus, unsigned long f
>  		cpumask_andnot(tmp_mask, src_mask, isolcpus);
>  		if (!cpumask_intersects(tmp_mask, cpu_online_mask))
>  			return -EINVAL;	/* Invalid isolated CPUs */
> -		cpumask_copy(housekeeping.cpumasks[type], tmp_mask);
> +		int err = housekeeping_update(type, tmp_mask);
> +
> +		if (err)
> +			return err;

Do we really need two places to define 'int err' or might it be possible
to have one instance defined at function scope?

Thanks,

        tglx