linux-kernel - Re: [PATCH] ftrace: Allow WITH_ARGS flavour of graph tracer with shadow call stack

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]

Date:   Mon, 12 Dec 2022 10:46:12 +0000
From:   Mark Rutland <mark.rutland@....com>
To:     Masami Hiramatsu <mhiramat@...nel.org>
Cc:     Ard Biesheuvel <ardb@...nel.org>, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, will@...nel.org,
        rostedt@...dmis.org, samitolvanen@...gle.com, keescook@...omium.org
Subject: Re: [PATCH] ftrace: Allow WITH_ARGS flavour of graph tracer with
 shadow call stack

On Sun, Dec 11, 2022 at 12:27:31PM +0900, Masami Hiramatsu wrote:
> On Fri,  9 Dec 2022 15:34:02 +0100
> Ard Biesheuvel <ardb@...nel.org> wrote:
> 
> > The recent switch on arm64 from DYNAMIC_FTRACE_WITH_REGS to
> > DYNAMIC_FTRACE_WITH_ARGS failed to take into account that we currently
> > require the former in order to allow the function graph tracer to be
> > enabled in combination with shadow call stacks. This means that this is
> > no longer permitted at all, in spite of the fact that either flavour of
> > ftrace works perfectly fine in this combination.
> > 
> > Given that arm64 is the only arch that implements shadow call stacks in
> > the first place, let's update the condition to just reflect the arm64
> > change. When other architectures adopt shadow call stack support, this
> > can be revisited if needed.
> 
> This brings a question. Is the SCS safe if kretprobe(rethook) is enabled?
> it also changes the stack entry after a calling function.

That should be safe.

The SCS push is just an instruction somewhere in the function, and since
kretprobe instruments the first instruction of a function, that intrumentation
will run *before* the SCS push occurs, and so it can safely override the return
address.

The difficulty with ftrace is that the old mcount implementation calls into
ftrace *after* the function prologue, after saving some GPRs to the stack,
signing the return address with pointer authentication, and/or pushing the
return address to the SCS.

The DYNAMIC_FTRACE_WITH_{ARGS,REGS} forms use patchable-function-entry to hook
functions *before* any of that happens, and are safe for the same reason as
kretprobes.

Thanks,
Mark.

> 
> Thank you,
> 
> > 
> > Signed-off-by: Ard Biesheuvel <ardb@...nel.org>
> > ---
> >  arch/Kconfig | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/arch/Kconfig b/arch/Kconfig
> > index 072a1b39e3afd0d1..683f365b5e31c856 100644
> > --- a/arch/Kconfig
> > +++ b/arch/Kconfig
> > @@ -635,7 +635,7 @@ config ARCH_SUPPORTS_SHADOW_CALL_STACK
> >  config SHADOW_CALL_STACK
> >  	bool "Shadow Call Stack"
> >  	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
> > -	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
> > +	depends on DYNAMIC_FTRACE_WITH_ARGS || !FUNCTION_GRAPH_TRACER
> >  	help
> >  	  This option enables the compiler's Shadow Call Stack, which
> >  	  uses a shadow stack to protect function return addresses from
> > -- 
> > 2.35.1
> > 
> 
> 
> -- 
> Masami Hiramatsu (Google) <mhiramat@...nel.org>