[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <54B5D5B9.4060409@hitachi.com>
Date: Wed, 14 Jan 2015 11:34:33 +0900
From: Masami Hiramatsu <masami.hiramatsu.pt@...achi.com>
To: Jiri Kosina <jkosina@...e.cz>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
Stephen Rothwell <sfr@...b.auug.org.au>,
Josh Poimboeuf <jpoimboe@...hat.com>,
Christoph Hellwig <hch@...radead.org>,
linux-kernel@...r.kernel.org, live-patching@...r.kernel.org,
linux-next@...r.kernel.org, Steven Rostedt <rostedt@...dmis.org>
Subject: Re: [PATCH] ftrace: don't allow IPMODIFY without proper compiler
support (was Re: Re: livepatching tree for linux-next)
(2015/01/14 7:47), Jiri Kosina wrote:
> On Mon, 12 Jan 2015, Masami Hiramatsu wrote:
>
>>> In any case, Masami, I really think you would like to do something
>>> like that for IPMODIFY as well ... or are you deliberately defering
>>> the responsibility to handle the possible mcount fallout to the
>>> ftrace_ops owner?
>>
>> Ah, good point. I just tried to use ftrace and WARN if not possible
>> to use it. I'll see it tomorrow. Anyway, I'd prefer to have this
>> kind of checking functionality in ftrace.
>
> Okay, so how about something like this, for example ... ?
Thanks! Could you read my comments?
> From: Jiri Kosina <jkosina@...e.cz>
> Subject: [PATCH] ftrace: don't allow IPMODIFY without proper compiler support
>
> Using IPMODIFY needs to be allowed only with compilers which are
> guaranteed to generate function prologues compatible with function
> redirection through changing instruction pointer in saved regs.
>
> For example changing regs->ip on x86_64 in cases when gcc is using mcount
> (and not fentry) is not allowed, because at the time mcount call is
> issued, the original function's prologue has already been executed, which
> leads to all kinds of inconsistent havoc.
>
> There is currently no way to express dependency on gcc features in
> Kconfig, (CC_USING_FENTRY is defined only during build, so it's not
> possible for Kconfig symbol to depend on it) so this needs to be checked
> in runtime.
>
> Mark x86_64 with fentry supported for now. Other archs can be added
> gradually.
>
> Signed-off-by: Jiri Kosina <jkosina@...e.cz>
> ---
> arch/x86/include/asm/ftrace.h | 2 ++
> include/linux/ftrace.h | 4 ++++
> kernel/trace/ftrace.c | 5 +++++
> 3 files changed, 11 insertions(+)
>
> diff --git a/arch/x86/include/asm/ftrace.h b/arch/x86/include/asm/ftrace.h
> index f45acad..29fa417 100644
> --- a/arch/x86/include/asm/ftrace.h
> +++ b/arch/x86/include/asm/ftrace.h
> @@ -4,8 +4,10 @@
> #ifdef CONFIG_FUNCTION_TRACER
> #ifdef CC_USING_FENTRY
> # define MCOUNT_ADDR ((long)(__fentry__))
> +# define arch_ftrace_ipmodify_compiler_support(void) ({ 1; })
> #else
> # define MCOUNT_ADDR ((long)(mcount))
> +#define arch_ftrace_ipmodify_compiler_support(void) ({ 0; })
Hmm, can we just define ARCH_FTRACE_SUPPORT_IPMODIFY here?
> #endif
> #define MCOUNT_INSN_SIZE 5 /* sizeof mcount call */
>
> diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
> index 1da6029..655ba99 100644
> --- a/include/linux/ftrace.h
> +++ b/include/linux/ftrace.h
> @@ -244,6 +244,10 @@ static inline int ftrace_function_local_disabled(struct ftrace_ops *ops)
> extern void ftrace_stub(unsigned long a0, unsigned long a1,
> struct ftrace_ops *op, struct pt_regs *regs);
>
> +#ifndef arch_ftrace_ipmodify_compiler_support
> +/* let's not make any implicit assumptions about profiling call placement */
> +# define arch_ftrace_ipmodify_compiler_support() { 0; }
> +#endif
> #else /* !CONFIG_FUNCTION_TRACER */
> /*
> * (un)register_ftrace_function must be a macro since the ops parameter
> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
> index 929a733..11370fd 100644
> --- a/kernel/trace/ftrace.c
> +++ b/kernel/trace/ftrace.c
> @@ -1809,6 +1809,11 @@ static int __ftrace_hash_update_ipmodify(struct ftrace_ops *ops,
> if (!(ops->flags & FTRACE_OPS_FL_IPMODIFY))
> return 0;
>
> + if (!arch_ftrace_ipmodify_compiler_support()) {
> + WARN(1, "Your compiler doesn't support features necessary for IPMODIFY");
> + return 0;
> + }
Actually, if ftrace doesn't support IPMODIFY, I would like to just drop
the entire code for CONFIG_KPROBES_ON_FTRACE(this is a hidden config),
instead of checking at runtime.
So, there are 2 ifdefs of code in kernel/kprobes.c for
CONFIG_KPROBES_ON_FTRACE, those should also check
ARCH_FTRACE_SUPPORT_IPMODIFY too.
Thank you,
> +
> /*
> * Since the IPMODIFY is a very address sensitive action, we do not
> * allow ftrace_ops to set all functions to new hash.
>
--
Masami HIRAMATSU
Software Platform Research Dept. Linux Technology Research Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@...achi.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists