linux-kernel - Re: [PATCH] lib: vsprintf: Avoid 32-bit truncation in vsscanf number parsing

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] [day] [month] [year] [list]

Message-ID: <efd71e08-2df2-db4e-4448-a096bf05b667@opensource.cirrus.com>
Date:   Mon, 16 Nov 2020 10:47:37 +0000
From:   Richard Fitzgerald <rf@...nsource.cirrus.com>
To:     Petr Mladek <pmladek@...e.com>,
        Steven Rostedt <rostedt@...dmis.org>
CC:     <sergey.senozhatsky@...il.com>, <linux-kernel@...r.kernel.org>,
        <patches@...nsource.cirrus.com>
Subject: Re: [PATCH] lib: vsprintf: Avoid 32-bit truncation in vsscanf number
 parsing

On 13/11/2020 14:00, Petr Mladek wrote:
> On Thu 2020-11-12 12:04:27, Steven Rostedt wrote:
>> On Thu, 12 Nov 2020 15:46:46 +0000
>> Richard Fitzgerald <rf@...nsource.cirrus.com> wrote:
>>
>>> See this thread from 2014 where the field width problem was raised and
>>> explained:
>>> http://lkml.iu.edu/hypermail/linux/kernel/1401.1/03443.html
>>>
>>> and the reply from Linus Torvalds that was against fixing field width
>>> handling:
>>> http://lkml.iu.edu/hypermail/linux/kernel/1401.1/03488.html
>>
>> Thanks for the pointers, but note, that references to older emails should
>> use https://lore.kernel.org/ as these links format the output really
>> horribly.
>>
>>>
>>> which I assume is why the field handling wasn't unoptimized to be
>>> strictly correct.
> 
> Honestly, the handling of the number width by div does not look like
> a real optimization to me. It avoids the need of the temporary buffer
> by expensive and error-prone operation.
> 
> IMHO, it is safe to assume that the width will be limited so that
> the value would never overflow.
> 
> The longest supported number would be (2^64 - 1) in octal. If I am
> counting correctly, it is
> 
>       01777777777777777777777
> 
> and it fits into buf[24] including the trailing '\0'.
> 
> We could call WARN_ON_ONCE() when the width >= 24 is higher.
> And we could add a compiler check when long long is bigger
> than 64-bit.
> 
>> Yes, but perhaps its time to fix the real problem and not just add
>> band-aids. That thread is over 6 years old (the email was from Jan 14, 2014)
>>
>> $ git diff `git rev-list --before 'Jan 14 2014' HEAD --max-count=1` |
>>    grep '^+' | grep sscanf | wc -l
>> 622
>>
>> There's been over 600 new additions of sscanf(). Now is the time to just
>> fix it correctly.
> 
> And the following one might suffer from this problem:
> 
> drivers/soundwire/slave.c:              ret = sscanf(compat, "sdw%01x%04hx%04hx%02hhx", &sdw_version,
> 

That's exactly the bug I have.
I'll look at reworking the code to handle number field widths properly.

> I agree with Steven that it is time to fix it properly.
> 
> Best Regards,
> Petr
>