Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: Memory zeroed when made available to user process
To:     Richard Weinberger <richard@sigma-star.at>,
        Michal Hocko <mhocko@kernel.org>
Cc:     Richard Weinberger <richard.weinberger@gmail.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
References: <71d13d76-4591-9206-ebbb-5e9599f10c7c@gmail.com>
 <CAFLxGvyJsjWum34Q67UP+EGkg9+W0nPsMpu8mS9PaH8wL3Rd9g@mail.gmail.com>
 <20180627131248.GA3032@dhcp22.suse.cz> <2634793.Cb5OrAPIO5@blindfold>
From:   Jefferson Carpenter <jeffersoncarpenter2@gmail.com>
Message-ID: <c1cda925-72bd-5937-a73d-617fd6f50c8e@gmail.com>
Date:   Fri, 29 Jun 2018 00:52:16 +0000
User-Agent: Mozilla/5.0 (Windows NT 6.2; WOW64; rv:52.0) Gecko/20100101
 Thunderbird/52.8.0
MIME-Version: 1.0
In-Reply-To: <2634793.Cb5OrAPIO5@blindfold>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 6/27/2018 1:18 PM, Richard Weinberger wrote:
> Am Mittwoch, 27. Juni 2018, 15:12:48 CEST schrieb Michal Hocko:
>> On Wed 27-06-18 13:29:05, Richard Weinberger wrote:
>>> On Wed, Jun 27, 2018 at 11:34 AM, Jefferson Carpenter
>>> <jeffersoncarpenter2@gmail.com> wrote:
>>>> Is there a way for a user process to mark memory as 'sensitive' or
>>>> 'non-sensitive' when it is allocated?  That could allow it not to have to be
>>>> zeroed before being allocated to another process.
>>>
>>> Isn't this what we have Meltdown and Spectre for? ;-)
>>>
>>> No, memory from the kernel is always zeroed.
>>> libc offers malloc() and calloc() for this purpose.

Interesting.  Let's say

Process 1:
free(use_memory(malloc(1024)));

Then Process 2:
malloc(1024);

The physical RAM used to service Process 2's malloc call has to be 
zeroed to prevent it from leaking data from Process 1.  However, if 
Process 1 could mark that memory as non-sensitive, then it would not 
have to be zeroed, saving the time it takes to do that.  However, this 
would require at least a bit per memory page, so maybe it's not worth it.

>>
>> Well, except for the weird MAP_UNINITIALIZED. Anyway agreed that this is
>> a bad idea and the flag should have never been merged. I've just
>> mentioned it for completness.
> 
> Oh, I forgot about the crazy nommu world. :-)
> 
> Thanks,
> //richard
>