Re: [PATCH] x86/mm: Do not verify W^X at boot up

From: Peter Zijlstra <peterz@infradead.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Steven Rostedt <rostedt@goodmis.org>,
	LKML <linux-kernel@vger.kernel.org>,
	Kees Cook <keescook@chromium.org>,
	Dave Hansen <dave.hansen@intel.com>,
	Sean Christopherson <seanjc@google.com>
Subject: Re: [PATCH] x86/mm: Do not verify W^X at boot up
Date: Tue, 25 Oct 2022 20:46:27 +0200	[thread overview]
Message-ID: <Y1gvA9gHUIiQFogw@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <CAHk-=wgvuanYPtndJR2iDpn60ry4FMueEg2DCRdQresvkdRgHQ@mail.gmail.com>

On Tue, Oct 25, 2022 at 11:14:07AM -0700, Linus Torvalds wrote:
> On Tue, Oct 25, 2022 at 10:48 AM Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > mm_alloc() uses allocate_mm() which requires a kmem_cache to be set-up.
> 
> Well, that seems to be just a strange effect of mm_cachep being set up
> by the oddly named "proc_caches_init" (I say oddly named because these
> days I associate "proc" with proc-fs, but I think it actually comes
> from "process").
> 
> That would actually probably make more sense if it was part of
> mm_init(), much earlier (where we do "kmem_cache_init()")
> 
> So this is another oddity in how we do "mm_init()", but we haven't
> actually initialized _that_ part of the mm setup.
> 
> Extra bonus points for another strange thing: we have "fork_init()",
> but that too doesn't actually initialize the mm_cachep that fork()
> actually uses. It does initialize the process one
> (task_struct_cachep). So that kind of makes sense, but yeah, the
> mm_alloc() cachep should have been set up by mm_init.
> 
> I think this is all "we just ended up randomly initializing things due
> to hysterical raisins"

OK, I'll go make all that happen.