Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 19 Sep 2022 16:34:40 +0200
From:   Peter Zijlstra <peterz@infradead.org>
To:     Kees Cook <keescook@chromium.org>
Cc:     Josh Triplett <josh@joshtriplett.org>,
        Eric Biederman <ebiederm@xmission.com>,
        Alexander Viro <viro@zeniv.linux.org.uk>, linux-mm@kvack.org,
        linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
        Mel Gorman <mgorman@suse.de>
Subject: Re: [PATCH] fs/exec.c: Add fast path for ENOENT on PATH search
 before allocating mm
Message-ID: <Yyh+AFYJJwvx3iun@hirez.programming.kicks-ass.net>
References: <5c7333ea4bec2fad1b47a8fa2db7c31e4ffc4f14.1663334978.git.josh@joshtriplett.org>
 <202209160727.5FC78B735@keescook>
 <YyTY+OaClK+JHCOw@localhost>
 <202209161637.9EDAF6B18@keescook>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <202209161637.9EDAF6B18@keescook>
Precedence: bulk

On Fri, Sep 16, 2022 at 05:11:18PM -0700, Kees Cook wrote:

> The interaction with sched_exec() should be no worse (the file is opened
> before it in either case), but in reading that function, it talks about
> taking the opportunity to move the process to another CPU (IIUC) since,
> paraphrasing, "it is at its lowest memory/cache size." But I wonder if
> there is an existing accidental pessimistic result in that the process
> stack has already been allocated. I am only passingly familiar with how
> tasks get moved around under NUMA -- is the scheduler going to move
> this process onto a different NUMA node and now it will be forced to
> have the userspace process stack on one node and the program text and
> heap on another? Or is that totally lost in the noise?

Probably noise; text is going to be a crap-shoot anyway due to DSOs only
having a single copy in the page-cache. And the stack will be relatively
small at this point and also, numa-balance can migrate those pages
around if they matter.

> More specifically, I was wondering if processes would benefit from having
> sched_exec() moved before the mm creation?

Can't hurt I think.