linux-kernel - Re: [PATCH 1/5] mmu: mark spte present if the x bit is set

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <jpgoa6lf2jy.fsf@linux.bootlegged.copy>
Date:	Tue, 28 Jun 2016 17:04:17 -0400
From:	Bandan Das <bsd@...hat.com>
To:	Paolo Bonzini <pbonzini@...hat.com>
Cc:	kvm@...r.kernel.org, guangrong.xiao@...ux.intel.com,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/5] mmu: mark spte present if the x bit is set

Paolo Bonzini <pbonzini@...hat.com> writes:

> On 28/06/2016 22:37, Bandan Das wrote:
>> Paolo Bonzini <pbonzini@...hat.com> writes:
>> 
>>> On 28/06/2016 19:33, Bandan Das wrote:
>>>>>>>>  static int is_shadow_present_pte(u64 pte)
>>>>>>>>  {
>>>>>>>> -	return pte & PT_PRESENT_MASK && !is_mmio_spte(pte);
>>>>>>>> +	return pte & (PT_PRESENT_MASK | shadow_x_mask) &&
>>>>>>>> +		!is_mmio_spte(pte);
>>>>>>
>>>>>> This should really be pte & 7 when using EPT.  But this is okay as an
>>>>>> alternative to a new shadow_present_mask.
>>>> I could revive shadow_xonly_valid probably... Anyway, for now I will
>>>> add a TODO comment here.
>>>
>>> It's okay to it like this, because the only invalid PTEs reaching this
>>> point are those that is_mmio_spte filters away.  Hence you'll never get
>>> -W- PTEs here, and pte & 7 is really the same as how you wrote it.  It's
>>> pretty clever, and doesn't need a TODO at all. :)
>> 
>> Thanks, understood. So, the way it is written now covers all cases for
>> pte & 7. Let's still add a comment - clever things are usually
>> confusing to many!
>
> I think another way to write it is "(pte & 0xFFFFFFFFull) &&
> !is_mmio_spte(pte)", since non-present/non-MMIO SPTEs never use bits
> 1..31 (they can have non-zero bits 32..63 on 32-bit CPUs where we don't
> update the PTEs atomically).  Guangrong, what do you prefer?

Actually, I like this one better although until now, I was not sure if it's
a safe assumption for non-ept cases. From your description, it looks like
it is.

> Paolo