Received-SPF: pass (google.com: domain of linux-kernel+bounces-108668-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Message-ID: <5a6f58c5-40f3-4c89-8988-a9fee932e3ca@rasmusvillemoes.dk>
Date: Wed, 20 Mar 2024 10:03:02 +0100
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] cpumask: Create dedicated kmem cache for cpumask var
Content-Language: en-US, da
To: Dawei Li <dawei.li@shingroup.cn>, yury.norov@gmail.com,
 andriy.shevchenko@linux.intel.com, akpm@linux-foundation.org
Cc: linux-kernel@vger.kernel.org
References: <20240319122400.399516-1-dawei.li@shingroup.cn>
From: Rasmus Villemoes <linux@rasmusvillemoes.dk>
In-Reply-To: <20240319122400.399516-1-dawei.li@shingroup.cn>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit

On 19/03/2024 13.24, Dawei Li wrote:
> alloc_cpumask_var_node() and friends allocate cpumask var dynamically
> for CONFIG_CPUMASK_OFFSTACK=y kernel. The allocated size of cpumask var
> is cpumask_size(), which is runtime constant after nr_cpu_ids is
> freezed.
> 
> Create a dedicated kmem cache for dynamic allocation of cpumask var.

Why?

> The window for creation of cache is somewhat narrow:
> - After last update of nr_cpu_ids(via set_nr_cpu_ids())
> - After kmem cache is available.
> - Before any alloc_cpumask_var_node() invocations(sched_init() e.g).

OK, so this sounds somewhat fragile. It's maybe correct, but I fail to
see what is gained by this, and the commit message does not provide any
hints.

Rasmus