Re: [PATCH] tracepoint: Do not fail unregistering a probe due to memory allocation

From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
To: rostedt <rostedt@goodmis.org>
Cc: Kees Cook <keescook@chromium.org>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	Matt Mullins <mmullins@mmlx.us>, Ingo Molnar <mingo@redhat.com>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Dmitry Vyukov <dvyukov@google.com>,
	Martin KaFai Lau <kafai@fb.com>, Song Liu <songliubraving@fb.com>,
	Yonghong Song <yhs@fb.com>, Andrii Nakryiko <andriin@fb.com>,
	John Fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@chromium.org>, netdev <netdev@vger.kernel.org>,
	bpf <bpf@vger.kernel.org>
Subject: Re: [PATCH] tracepoint: Do not fail unregistering a probe due to memory allocation
Date: Tue, 17 Nov 2020 18:12:54 -0500 (EST)	[thread overview]
Message-ID: <1473764147.48847.1605654774757.JavaMail.zimbra@efficios.com> (raw)
In-Reply-To: <20201117171904.2d455699@gandalf.local.home>

----- On Nov 17, 2020, at 5:19 PM, rostedt rostedt@goodmis.org wrote:

> On Tue, 17 Nov 2020 13:33:42 -0800
> Kees Cook <keescook@chromium.org> wrote:
> 
>> As I think got discussed in the thread, what you had here wouldn't work
>> in a CFI build if the function prototype of the call site and the
>> function don't match. (Though I can't tell if .func() is ever called?)
>> 
>> i.e. .func's prototype must match tp_stub_func()'s.
>> 
> 
> 
> Hmm, I wonder how you handle tracepoints? This is called here:
> 
> include/linux/tracepoint.h:
> 
> 
> #define DEFINE_TRACE_FN(_name, _reg, _unreg, proto, args)		\
>	static const char __tpstrtab_##_name[]				\
>	__section("__tracepoints_strings") = #_name;			\
>	extern struct static_call_key STATIC_CALL_KEY(tp_func_##_name);	\
>	int __traceiter_##_name(void *__data, proto);			\
>	struct tracepoint __tracepoint_##_name	__used			\
>	__section("__tracepoints") = {					\
>		.name = __tpstrtab_##_name,				\
>		.key = STATIC_KEY_INIT_FALSE,				\
>		.static_call_key = &STATIC_CALL_KEY(tp_func_##_name),	\
>		.static_call_tramp = STATIC_CALL_TRAMP_ADDR(tp_func_##_name), \
>		.iterator = &__traceiter_##_name,			\
>		.regfunc = _reg,					\
>		.unregfunc = _unreg,					\
>		.funcs = NULL };					\
>	__TRACEPOINT_ENTRY(_name);					\
>	int __traceiter_##_name(void *__data, proto)			\
>	{								\
>		struct tracepoint_func *it_func_ptr;			\
>		void *it_func;						\
>									\
>		it_func_ptr =						\
>			rcu_dereference_raw((&__tracepoint_##_name)->funcs); \
>		do {							\
>			it_func = (it_func_ptr)->func;			\
>			__data = (it_func_ptr)->data;			\
> 
>			((void(*)(void *, proto))(it_func))(__data, args); \
> 
>			^^^^ called above ^^^^
> 
> Where args is unique for every tracepoint, but func is simply a void
> pointer.

That being said, the called functions have a prototype which match the
caller prototype exactly. So within the tracepoint internal data structures,
this function pointer is indeed a void pointer, but it is cast to a prototype
matching the callees to perform the calls. I suspect that as long as CFI checks
that caller/callees prototypes are compatible at runtime when the actual
calls happen, this all works fine.

Thanks,

Mathieu

> 
> -- Steve
> 
> 
>		} while ((++it_func_ptr)->func);			\
>		return 0;						\
> 	}								\

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com