Re: [PATCH 06/17] x86/alternative: use temporary mm for text poking

From: Nadav Amit <nadav.amit@gmail.com>
To: Andy Lutomirski <luto@kernel.org>
Cc: Rick Edgecombe <rick.p.edgecombe@intel.com>,
	Ingo Molnar <mingo@redhat.com>,
	LKML <linux-kernel@vger.kernel.org>, X86 ML <x86@kernel.org>,
	"H. Peter Anvin" <hpa@zytor.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Borislav Petkov <bp@alien8.de>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Damian Tometzki <linux_dti@icloud.com>,
	linux-integrity <linux-integrity@vger.kernel.org>,
	LSM List <linux-security-module@vger.kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Kernel Hardening <kernel-hardening@lists.openwall.com>,
	Linux-MM <linux-mm@kvack.org>, Will Deacon <will.deacon@arm.com>,
	Ard Biesheuvel <ard.biesheuvel@linaro.org>,
	Kristen Carlson Accardi <kristen@linux.intel.com>,
	"Dock, Deneen T" <deneen.t.dock@intel.com>,
	Kees Cook <keescook@chromium.org>,
	Dave Hansen <dave.hansen@intel.com>,
	Masami Hiramatsu <mhiramat@kernel.org>
Subject: Re: [PATCH 06/17] x86/alternative: use temporary mm for text poking
Date: Thu, 17 Jan 2019 14:29:36 -0800	[thread overview]
Message-ID: <32219CAE-7D49-4848-9497-A17E0D809B3E@gmail.com> (raw)
In-Reply-To: <B8C39C5A-A669-4F80-9BAE-7C11A4379ECF@gmail.com>

> On Jan 17, 2019, at 1:43 PM, Nadav Amit <nadav.amit@gmail.com> wrote:
> 
>> On Jan 17, 2019, at 12:47 PM, Andy Lutomirski <luto@kernel.org> wrote:
>> 
>> On Thu, Jan 17, 2019 at 12:27 PM Andy Lutomirski <luto@kernel.org> wrote:
>>> On Wed, Jan 16, 2019 at 4:33 PM Rick Edgecombe
>>> <rick.p.edgecombe@intel.com> wrote:
>>>> From: Nadav Amit <namit@vmware.com>
>>>> 
>>>> text_poke() can potentially compromise the security as it sets temporary
>>>> PTEs in the fixmap. These PTEs might be used to rewrite the kernel code
>>>> from other cores accidentally or maliciously, if an attacker gains the
>>>> ability to write onto kernel memory.
>>> 
>>> i think this may be sufficient, but barely.
>>> 
>>>> +       pte_clear(poking_mm, poking_addr, ptep);
>>>> +
>>>> +       /*
>>>> +        * __flush_tlb_one_user() performs a redundant TLB flush when PTI is on,
>>>> +        * as it also flushes the corresponding "user" address spaces, which
>>>> +        * does not exist.
>>>> +        *
>>>> +        * Poking, however, is already very inefficient since it does not try to
>>>> +        * batch updates, so we ignore this problem for the time being.
>>>> +        *
>>>> +        * Since the PTEs do not exist in other kernel address-spaces, we do
>>>> +        * not use __flush_tlb_one_kernel(), which when PTI is on would cause
>>>> +        * more unwarranted TLB flushes.
>>>> +        *
>>>> +        * There is a slight anomaly here: the PTE is a supervisor-only and
>>>> +        * (potentially) global and we use __flush_tlb_one_user() but this
>>>> +        * should be fine.
>>>> +        */
>>>> +       __flush_tlb_one_user(poking_addr);
>>>> +       if (cross_page_boundary) {
>>>> +               pte_clear(poking_mm, poking_addr + PAGE_SIZE, ptep + 1);
>>>> +               __flush_tlb_one_user(poking_addr + PAGE_SIZE);
>>>> +       }
>>> 
>>> In principle, another CPU could still have the old translation.  Your
>>> mutex probably makes this impossible, but it makes me nervous.
>>> Ideally you'd use flush_tlb_mm_range(), but I guess you can't do that
>>> with IRQs off.  Hmm.  I think you should add an inc_mm_tlb_gen() here.
>>> Arguably, if you did that, you could omit the flushes, but maybe
>>> that's silly.
>>> 
>>> If we start getting new users of use_temporary_mm(), we should give
>>> some serious thought to the SMP semantics.
>>> 
>>> Also, you're using PAGE_KERNEL.  Please tell me that the global bit
>>> isn't set in there.
>> 
>> Much better solution: do unuse_temporary_mm() and *then*
>> flush_tlb_mm_range().  This is entirely non-sketchy and should be just
>> about optimal, too.
> 
> This solution sounds nice and clean. The fact the global-bit was set didn’t
> matter before (since __flush_tlb_one_user would get rid of it no matter
> what), but would matter now, so I’ll change it too.

Err.. so actually text_poke() might be called with disabled IRQs (by kgdb).
flush_tlb_mm_range() should still work fine even with disabled IRQs since no
core would use poking_mm at this point. I can add a comment to
flush_tlb_mm_range(), but all in all it is actually not very pretty.