Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <2cac23c2-ffe2-f83d-5df0-2f851e09c4f8@linux.com>
Date:   Sun, 15 May 2022 19:32:50 +0300
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.7.0
Reply-To: alex.popov@linux.com
Subject: Re: [PATCH v2 06/13] stackleak: rework stack high bound handling
Content-Language: en-US
To:     Mark Rutland <mark.rutland@arm.com>
Cc:     linux-arm-kernel@lists.infradead.org, akpm@linux-foundation.org,
        catalin.marinas@arm.com, keescook@chromium.org,
        linux-kernel@vger.kernel.org, luto@kernel.org, will@kernel.org
References: <20220427173128.2603085-1-mark.rutland@arm.com>
 <20220427173128.2603085-7-mark.rutland@arm.com>
 <ff36f701-6950-b923-c4f5-d6182bdc7534@linux.com>
 <YnpK6ju2FkzagbPY@FVFF77S0Q05N>
From:   Alexander Popov <alex.popov@linux.com>
In-Reply-To: <YnpK6ju2FkzagbPY@FVFF77S0Q05N>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 10.05.2022 14:22, Mark Rutland wrote:
> On Mon, May 09, 2022 at 12:27:22AM +0300, Alexander Popov wrote:
>> On 27.04.2022 20:31, Mark Rutland wrote:
>>> Prior to returning to userpace, we reset current->lowest_stack to a
>>> reasonable high bound. Currently we do this by subtracting the arbitrary
>>> value `THREAD_SIZE/64` from the top of the stack, for reasons lost to
>>> history.
>>>
>>> Looking at configurations today:
>>>
>>> * On i386 where THREAD_SIZE is 8K, the bound will be 128 bytes. The
>>>     pt_regs at the top of the stack is 68 bytes (with 0 to 16 bytes of
>>>     padding above), and so this covers an additional portion of 44 to 60
>>>     bytes.
>>>
>>> * On x86_64 where THREAD_SIZE is at least 16K (up to 32K with KASAN) the
>>>     bound will be at least 256 bytes (up to 512 with KASAN). The pt_regs
>>>     at the top of the stack is 168 bytes, and so this cover an additional
>>>     88 bytes of stack (up to 344 with KASAN).
>>>
>>> * On arm64 where THREAD_SIZE is at least 16K (up to 64K with 64K pages
>>>     and VMAP_STACK), the bound will be at least 256 bytes (up to 1024 with
>>>     KASAN). The pt_regs at the top of the stack is 336 bytes, so this can
>>>     fall within the pt_regs, or can cover an additional 688 bytes of
>>>     stack.
>>>
>>> Clearly the `THREAD_SIZE/64` value doesn't make much sense -- in the
>>> worst case, this will cause more than 600 bytes of stack to be erased
>>> for every syscall, even if actual stack usage were substantially
>>> smaller.
>>>
>>> This patches makes this slightly less nonsensical by consistently
>>> resetting current->lowest_stack to the base of the task pt_regs. For
>>> clarity and for consistency with the handling of the low bound, the
>>> generation of the high bound is split into a helper with commentary
>>> explaining why.
>>>
>>> Since the pt_regs at the top of the stack will be clobbered upon the
>>> next exception entry, we don't need to poison these at exception exit.
>>> By using task_pt_regs() as the high stack boundary instead of
>>> current_top_of_stack() we avoid some redundant poisoning, and the
>>> compiler can share the address generation between the poisoning and
>>> restting of `current->lowest_stack`, making the generated code more
>>> optimal.
>>>
>>> It's not clear to me whether the existing `THREAD_SIZE/64` offset was a
>>> dodgy heuristic to skip the pt_regs, or whether it was attempting to
>>> minimize the number of times stackleak_check_stack() would have to
>>> update `current->lowest_stack` when stack usage was shallow at the cost
>>> of unconditionally poisoning a small portion of the stack for every exit
>>> to userspace.
>>
>> I inherited this 'THREAD_SIZE/64' logic is from the original grsecurity patch.
>> As I mentioned, originally this was written in asm.
>>
>> For x86_64:
>> 	mov	TASK_thread_sp0(%r11), %rdi
>> 	sub	$256, %rdi
>> 	mov	%rdi, TASK_lowest_stack(%r11)
>>
>> For x86_32:
>> 	mov TASK_thread_sp0(%ebp), %edi
>> 	sub $128, %edi
>> 	mov %edi, TASK_lowest_stack(%ebp)
>>
>> 256 bytes for x86_64 and 128 bytes for x86_32 are exactly THREAD_SIZE/64.
> 
> To check my understanding, did you come up with the `THREAD_SIZE/64`
> calculation from reverse-engineering the assembly?

Yes, exactly.

> I strongly suspect that has nothing to do with THEAD_SIZE, and was trying to
> fall just below the task's pt_regs (and maybe some unconditional stack usage
> just below that).

Aha, that's probable.

>> I think this value was chosen as optimal for minimizing poison scanning.
>> It's possible that stackleak_track_stack() is not called during the syscall
>> because all the called functions have small stack frames.
> 
> I can believe that, but given the clearing cost appears to dominate the
> scanning cost, I suspect we can live with making this precisely the bottom of
> the pt_regs.

Sounds good!

>>> For now I've simply removed the offset, and if we need/want to minimize
>>> updates for shallow stack usage it should be easy to add a better
>>> heuristic atop, with appropriate commentary so we know what's going on.
>>
>> I like your idea to erase the thread stack up to pt_regs if we call the
>> stackleak erasing from the trampoline stack.
>>
>> But here I don't understand where task_pt_regs() points to...
> 
> As mentioned in the commit message, there's a struct pt_regs at the top of each
> task stack (created at exception entry, and consumed by exception return),
> which contains the user register state. On arm64 and x86_64, that's *right* at
> the top of the stack, but on i386 there can be some padding above that.
> task_pt_regs(tsk) points at the base of that pt_regs.
> 
> That looks something like the following (with increasing addresses going
> upwards):
> 
>    ----------------   <--- task_stack_page(tsk) + THREAD_SIZE
>    (optional padding)
>    ----------------   <--- task_top_of_stack(tsk) // x86 only
>        /\
>        ||
>        || pt_regs
>        ||
>        \/
>    ----------------   <--- task_pt_regs(tsk)
>        /\
>        ||
>        ||
>        ||
>        || Usable task stack
>        ||
>        ||
>        ||
>        \/
>    ----------------
>    STACK_END_MAGIC
>    ----------------   <--- task_stack_page(tsk)
> 
> Thanks,
> Mark.
> 
>>> Signed-off-by: Mark Rutland <mark.rutland@arm.com>
>>> Cc: Alexander Popov <alex.popov@linux.com>
>>> Cc: Andrew Morton <akpm@linux-foundation.org>
>>> Cc: Andy Lutomirski <luto@kernel.org>
>>> Cc: Kees Cook <keescook@chromium.org>
>>> ---
>>>    include/linux/stackleak.h | 14 ++++++++++++++
>>>    kernel/stackleak.c        | 19 ++++++++++++++-----
>>>    2 files changed, 28 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/include/linux/stackleak.h b/include/linux/stackleak.h
>>> index 67430faa5c518..467661aeb4136 100644
>>> --- a/include/linux/stackleak.h
>>> +++ b/include/linux/stackleak.h
>>> @@ -28,6 +28,20 @@ stackleak_task_low_bound(const struct task_struct *tsk)
>>>    	return (unsigned long)end_of_stack(tsk) + sizeof(unsigned long);
>>>    }
>>> +/*
>>> + * The address immediately after the highest address on tsk's stack which we
>>> + * can plausibly erase.
>>> + */
>>> +static __always_inline unsigned long
>>> +stackleak_task_high_bound(const struct task_struct *tsk)
>>> +{
>>> +	/*
>>> +	 * The task's pt_regs lives at the top of the task stack and will be
>>> +	 * overwritten by exception entry, so there's no need to erase them.
>>> +	 */
>>> +	return (unsigned long)task_pt_regs(tsk);
>>> +}
>>> +
>>>    static inline void stackleak_task_init(struct task_struct *t)
>>>    {
>>>    	t->lowest_stack = stackleak_task_low_bound(t);
>>> diff --git a/kernel/stackleak.c b/kernel/stackleak.c
>>> index d5f684dc0a2d9..ba346d46218f5 100644
>>> --- a/kernel/stackleak.c
>>> +++ b/kernel/stackleak.c
>>> @@ -73,6 +73,7 @@ late_initcall(stackleak_sysctls_init);
>>>    static __always_inline void __stackleak_erase(void)
>>>    {
>>>    	const unsigned long task_stack_low = stackleak_task_low_bound(current);
>>> +	const unsigned long task_stack_high = stackleak_task_high_bound(current);
>>>    	unsigned long erase_low = current->lowest_stack;
>>>    	unsigned long erase_high;
>>>    	unsigned int poison_count = 0;
>>> @@ -93,14 +94,22 @@ static __always_inline void __stackleak_erase(void)
>>>    #endif
>>>    	/*
>>> -	 * Now write the poison value to the kernel stack between 'erase_low'
>>> -	 * and 'erase_high'. We assume that the stack pointer doesn't change
>>> -	 * when we write poison.
>>> +	 * Write poison to the task's stack between 'erase_low' and
>>> +	 * 'erase_high'.
>>> +	 *
>>> +	 * If we're running on a different stack (e.g. an entry trampoline
>>> +	 * stack) we can erase everything below the pt_regs at the top of the
>>> +	 * task stack.
>>> +	 *
>>> +	 * If we're running on the task stack itself, we must not clobber any
>>> +	 * stack used by this function and its caller. We assume that this
>>> +	 * function has a fixed-size stack frame, and the current stack pointer
>>> +	 * doesn't change while we write poison.
>>>    	 */
>>>    	if (on_thread_stack())
>>>    		erase_high = current_stack_pointer;
>>>    	else
>>> -		erase_high = current_top_of_stack();
>>> +		erase_high = task_stack_high;
>>>    	while (erase_low < erase_high) {
>>>    		*(unsigned long *)erase_low = STACKLEAK_POISON;
>>> @@ -108,7 +117,7 @@ static __always_inline void __stackleak_erase(void)
>>>    	}
>>>    	/* Reset the 'lowest_stack' value for the next syscall */
>>> -	current->lowest_stack = current_top_of_stack() - THREAD_SIZE/64;
>>> +	current->lowest_stack = task_stack_high;
>>>    }
>>>    asmlinkage void noinstr stackleak_erase(void)
>>