From: Kees Cook <keescook@chromium.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Alexander Popov <alex.popov@linux.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
Thomas Gleixner <tglx@linutronix.de>,
Josh Poimboeuf <jpoimboe@redhat.com>,
Borislav Petkov <bp@alien8.de>,
linux-kernel@vger.kernel.org, linux-hardening@vger.kernel.org
Subject: Re: [PATCH] gcc-plugins/stackleak: Use noinstr in favor of notrace
Date: Sun, 6 Feb 2022 18:57:59 -0800 [thread overview]
Message-ID: <202202061856.EBC48533B@keescook> (raw)
In-Reply-To: <YgAyL2D25nweODX3@hirez.programming.kicks-ass.net>
On Sun, Feb 06, 2022 at 09:40:15PM +0100, Peter Zijlstra wrote:
> On Sun, Feb 06, 2022 at 08:46:47AM -0800, Kees Cook wrote:
> > On Sun, Feb 06, 2022 at 12:58:16PM +0100, Peter Zijlstra wrote:
> > > On Tue, Feb 01, 2022 at 04:19:18PM -0800, Kees Cook wrote:
> > > > Is it correct to exclude .noinstr.text here? That means any functions called in
> > > > there will have their stack utilization untracked. This doesn't seem right to me,
> > > > though. Shouldn't stackleak_track_stack() just be marked noinstr instead?
> > >
> > > This patch is right. stackleak_track_stack() cannot be marked noinstr
> > > becaues it accesses things that might not be there.
> >
> > Hmm, as in "current()" may not be available/sane?
>
> Exactly the case; if we lift the PTI address space swizzle, we start
> with C without having the kernel mapped or even the per-cpu segment
> offset set. So things like current will explode.
>
> The whole noinstr thing was invented to get back to C as portable
> Assembler, with the express purpose to lift a bunch of entry code to C.
>
> > > Consider what happens if we pull the PTI page-table swap into the
> > > noinstr C part.
> >
> > Yeah, I see your point. I suspect the reason this all currently works
> > is because stackleak is supposed to only instrument leaf functions that
> > have sufficiently large (default 100 bytes) stack usage.
> >
> > What sorts of things may end up in .noinstr.text that are 100+ byte stack
> > leaf functions that would be end up deeper in the call stack? (i.e. what
> > could get missed from stack depth tracking?) Interrupt handling comes
> > to mind, but I'd expect that to make further calls (i.e. not a leaf).
>
> All the syscall/exception/interrupt entry stuff is noinstr; I don't
> think we have huge stackframes, but with all that in C that's much
> easier to do than with then in asm.
>
> If you worry about this, it should be possible to have objtool warn
> about excessive stack frames for noinstr code I suppose, it already
> tracks the stack anyway.
Yeah, I think we should be okay at least for now.
Let me know what you think of
https://lore.kernel.org/linux-hardening/20220206174508.2425076-1-keescook@chromium.org/
and if you like it I can send a v2 Linus's way...
-Kees
--
Kees Cook
prev parent reply other threads:[~2022-02-07 5:56 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-02-02 0:19 [PATCH] gcc-plugins/stackleak: Use noinstr in favor of notrace Kees Cook
2022-02-02 10:45 ` Mark Rutland
2022-02-03 19:33 ` Linus Torvalds
2022-02-06 11:58 ` Peter Zijlstra
2022-02-06 16:46 ` Kees Cook
2022-02-06 20:40 ` Peter Zijlstra
2022-02-07 2:57 ` Kees Cook [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=202202061856.EBC48533B@keescook \
--to=keescook@chromium.org \
--cc=alex.popov@linux.com \
--cc=bp@alien8.de \
--cc=jpoimboe@redhat.com \
--cc=linux-hardening@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=peterz@infradead.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).