linux-kernel - Re: [PATCH v3] vsprintf: Check real user/group id for %pK

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]

Date:	Thu, 10 Oct 2013 09:25:24 +1100
From:	Ryan Mallon <rmallon@...il.com>
To:	Joe Perches <joe@...ches.com>
CC:	Andrew Morton <akpm@...ux-foundation.org>, eldad@...refinery.com,
	Jiri Kosina <jkosina@...e.cz>, jgunthorpe@...idianresearch.com,
	Dan Rosenberg <dan.j.rosenberg@...il.com>,
	Kees Cook <keescook@...omium.org>,
	Alexander Viro <viro@...iv.linux.org.uk>,
	"Eric W. Biederman" <ebiederm@...ssion.com>,
	George Spelvin <linux@...izon.com>,
	"kernel-hardening@...ts.openwall.com" 
	<kernel-hardening@...ts.openwall.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v3] vsprintf: Check real user/group id for %pK

On 10/10/13 09:14, Joe Perches wrote:
> On Thu, 2013-10-10 at 09:04 +1100, Ryan Mallon wrote:
>> On 10/10/13 09:00, Joe Perches wrote:
> []
>>> Move the interrupt tests and pK-error printk
>>> into case 1:
>>>
>>> It's the only case where CAP_SYSLOG needs to be
>>> tested so it doesn't need to be above the switch.
>>
>> Like I said, I think it is useful to do the pK-error check anyway. It is
>> checking for internal kernel bugs, since if 'pK-error' ever gets
>> printed, then some kernel code is doing the wrong thing.
> 
> I think you don't quite understand how kptr_restrict works.
> 
> If it's 0, then the ptr value is always emitted naturally.
> if it's 2, then the ptr value is always emitted as 0.

I understand this.

> 
>> Therefore, I
>> think it is useful to print it always (I would argue it even makes sense
>> when kptr_restrict=0).
> 
> How?  Maybe it's me that doesn't quite understand.

This check:

	if (kptr_restrict && (in_irq() || in_serving_softirq() ||
			      in_nmi())) {

Is making sure that you don't have kernel code doing something like this:

	irqreturn_t some_irq_handler(int irq, void *data)
	{
		struct seq_file *seq = to_seq(data);

		seq_printf(seq, "value = %pK\n");
		return IRQ_HANDLED;
	}

Because that obviously won't work when kptr_restrict=1 (because the
CAP_SYSLOG check is meaningless). However, the code is broken regardless
of the kptr_restrict value. Since the default value of kptr_restrict is
0, this kind of bug can go over-looked because the seq file will print
the pointer value correctly when kptr_restrict=0, and it will correctly
print 0's when kptr_restrict=2, but it will print 'pK-error' when
kptr_restrict=1. Doing the check in all cases makes it more likely that
bugs like this get found. In fact, doing something like:

	if (WARN_ON(in_irq() || in_serving_softirq() || in_nmi())) {

Might be better, since that will print a stack-trace showing where the
offending vsprintf is.

~Ryan

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/