linux-kernel - Re: [PATCH] Fix region lost in /proc/self/smaps

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 for Android: free password hash cracker in your pocket

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

Message-ID: <8b800d72-9b28-237c-47a6-604d98a40315@linux.intel.com>
Date:   Thu, 8 Sep 2016 11:36:11 +0800
From:   Xiao Guangrong <guangrong.xiao@...ux.intel.com>
To:     Dave Hansen <dave.hansen@...el.com>, pbonzini@...hat.com,
        akpm@...ux-foundation.org, mhocko@...e.com,
        dan.j.williams@...el.com
Cc:     gleb@...nel.org, mtosatti@...hat.com, kvm@...r.kernel.org,
        linux-kernel@...r.kernel.org, stefanha@...hat.com,
        yuhuang@...hat.com, linux-mm@...ck.org,
        ross.zwisler@...ux.intel.com
Subject: Re: [PATCH] Fix region lost in /proc/self/smaps



On 09/08/2016 12:34 AM, Dave Hansen wrote:
> On 09/06/2016 11:51 PM, Xiao Guangrong wrote:
>> In order to fix this bug, we make 'file->version' indicate the next VMA
>> we want to handle
>
> This new approach makes it more likely that we'll skip a new VMA that
> gets inserted in between the read()s.  But, I guess that's OK.  We don't
> exactly claim to be giving super up-to-date data at the time of read().

Yes, I completely agree with you. :)

>
> With the old code, was there also a case that we could print out the
> same virtual address range more than once?  It seems like that could
> happen if we had a VMA split between two reads.

Yes.

>
> I think this introduces one oddity: if you have a VMA merge between two
> reads(), you might get the same virtual address range twice in your
> output.  This didn't happen before because we would have just skipped
> over the area that got merged.
>
> Take two example VMAs:
>
> 	vma-A: (0x1000 -> 0x2000)
> 	vma-B: (0x2000 -> 0x3000)
>
> read() #1: prints vma-A, sets m->version=0x2000
>
> Now, merge A/B to make C:
>
> 	vma-C: (0x1000 -> 0x3000)
>
> read() #2: find_vma(m->version=0x2000), returns vma-C, prints vma-C
>
> The user will see two VMAs in their output:
>
> 	A: 0x1000->0x2000
> 	C: 0x1000->0x3000
>
> Will it confuse them to see the same virtual address range twice?  Or is
> there something preventing that happening that I'm missing?
>

You are right. Nothing can prevent it.

However, it is not easy to handle the case that the new VMA overlays with the old VMA
already got by userspace. I think we have some choices:
1: One way is completely skipping the new VMA region as current kernel code does but i
    do not think this is good as the later VMAs will be dropped.

2: show the un-overlayed portion of new VMA. In your case, we just show the region
    (0x2000 -> 0x3000), however, it can not work well if the VMA is a new created
    region with different attributions.

3: completely show the new VMA as this patch does.

Which one do you prefer?

Thanks!