linux-kernel - Re: [PATCH 4/5] x86/ftrace: Remove SYSTEM

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: <Y1jbl8SQ9EQ/Cw/a@hirez.programming.kicks-ass.net>
Date:   Wed, 26 Oct 2022 09:02:47 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Steven Rostedt <rostedt@...dmis.org>
Cc:     torvalds@...ux-foundation.org, dave.hansen@...el.com,
        linux-kernel@...r.kernel.org, x86@...nel.org,
        keescook@...omium.org, seanjc@...gle.com
Subject: Re: [PATCH 4/5] x86/ftrace: Remove SYSTEM_BOOTING exceptions

On Tue, Oct 25, 2022 at 04:59:56PM -0400, Steven Rostedt wrote:
> On Tue, 25 Oct 2022 22:07:00 +0200
> Peter Zijlstra <peterz@...radead.org> wrote:
> 
> > Now that text_poke is available before ftrace, remove the
> > SYSTEM_BOOTING exceptions.
> > 
> > Specifically, this cures a W+X case during boot.
> 
> We have W+X all over the place (the entire kernel text). And I don't think
> we really want this.
> 
> This will slow down boots in general, as it will cause all static_branches
> to use this memory page logic. And I don't think we really want to do
> that at boot up when we don't need to.

Both static_call and jump_label explicitly call text_poke_early() when
appropriate.

> I would change this to:
> 
> 	if (unlikely(system_state == SYSTEM_BOOTING) &&
> 	    core_kernel_text((unsigned long)addr)) {
> 
> This way we still do memcpy() on all core kernel text which is still
> writable. It was the ftrace allocated trampoline that caused issues, not
> the locations that were being updated.

I would suggest changing ftrace to call text_poke_early() when
appropriate if it matters (it already does a little of that); doing a
boot test with and without my patch 4 on shows no noticable overhead
over being horribly slow either way.