Re: [PATCH] x86: Pin task-stack in __get_wchan()

From: Qi Zheng <zhengqi.arch@bytedance.com>
To: Peter Zijlstra <peterz@infradead.org>,
	Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
	Holger Hoffst??tte <holger@applied-asynchrony.com>,
	Kees Cook <keescook@chromium.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Justin Forbes <jmforbes@linuxtx.org>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Guenter Roeck <linux@roeck-us.net>, Shuah Khan <shuah@kernel.org>,
	patches@kernelci.org, lkft-triage@lists.linaro.org,
	Pavel Machek <pavel@denx.de>, Jon Hunter <jonathanh@nvidia.com>,
	Florian Fainelli <f.fainelli@gmail.com>,
	stable <stable@vger.kernel.org>
Subject: Re: [PATCH] x86: Pin task-stack in __get_wchan()
Date: Fri, 19 Nov 2021 18:02:50 +0800	[thread overview]
Message-ID: <759cd319-990f-af23-2f1c-aba55d0768b8@bytedance.com> (raw)
In-Reply-To: <YZduix64h64cDa7R@hirez.programming.kicks-ass.net>

On 11/19/21 5:29 PM, Peter Zijlstra wrote:
> On Thu, Nov 18, 2021 at 06:04:27PM -0800, Josh Poimboeuf wrote:
>> On Thu, Nov 18, 2021 at 01:11:09PM +0100, Peter Zijlstra wrote:
> 
>>> I now have the below, the only thing missing is that there's a
>>> user_mode() call on a stack based regs. Now on x86_64 we can
>>> __get_kernel_nofault() regs->cs and call it a day, but on i386 we have
>>> to also fetch regs->flags.
>>>
>>> Is this really the way to go?
>>
>> Please no.  Can we just add a check in unwind_start() to ensure the
>> caller did try_get_task_stack()?
> 
> I tried; but at best it's fundamentally racy and in practise its worse
> because init_task doesn't seem to believe in refcounts and kthreads are
> odd for some raisin. Now those are fixable, but given the fundamental
> races, I don't see how it's ever going to be reliable.
> 
> I don't mind the __get_kernel_nofault() usage and think I can do a
> better implementation that will allow us to get rid of the
> pagefault_{dis,en}able() sprinkling, but that's for another day. It's
> just the user_mode(regs) usage that's going to be somewhat ugleh.
> 
> Anyway, below is the minimal fix for the situation at hand. I'm not
> going to be around much today, so if Linus wants to pick that up instead
> of mass revert things that's obviously fine too.
> 
> ---
> Subject: x86: Pin task-stack in __get_wchan()
> 
> When commit 5d1ceb3969b6 ("x86: Fix __get_wchan() for !STACKTRACE")
> moved from stacktrace to native unwind_*() usage, the
> try_get_task_stack() got lost, leading to use-after-free issues for
> dying tasks.
> 
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> ---
>   arch/x86/kernel/process.c | 5 +++++
>   1 file changed, 5 insertions(+)
> 
> diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
> index e9ee8b526319..04143a653a8a 100644
> --- a/arch/x86/kernel/process.c
> +++ b/arch/x86/kernel/process.c
> @@ -964,6 +964,9 @@ unsigned long __get_wchan(struct task_struct *p)
>   	struct unwind_state state;
>   	unsigned long addr = 0;
>   
> +	if (!try_get_task_stack(p))
> +		return 0;
> +
>   	for (unwind_start(&state, p, NULL, NULL); !unwind_done(&state);
>   	     unwind_next_frame(&state)) {
>   		addr = unwind_get_return_address(&state);
> @@ -974,6 +977,8 @@ unsigned long __get_wchan(struct task_struct *p)
>   		break;
>   	}
>   
> +	put_task_stack(p);
> +
>   	return addr;
>   }
>   
> 

This implementation is very similar to stack_trace_save_tsk(), maybe we
can just move stack_trace_save_tsk() out of CONFIG_STACKTRACE and reuse
it.

-- 
Thanks,
Qi