[kernel-hardening] Re: [PATCH 3/3] x86: kernel base offset ASLR

[Borislav Petkov <bp@alien8.de>]

On Fri, Apr 05, 2013 at 04:18:49PM -0700, Kees Cook wrote:
> There is already a hook in the patch that prints the offset:
> 
> +dump_kernel_offset(struct notifier_block *self, unsigned long v, void *p)
> +{
> +       pr_emerg("Kernel Offset: 0x%lx\n",
> +                (unsigned long)&_text - __START_KERNEL);
> ...
> +       atomic_notifier_chain_register(&panic_notifier_list,
> +                                       &kernel_offset_notifier);
> 
> But of course, this can get improved.

Yeah, this should probably be added to dump_trace(), i.e. something
which walks stack frames and dumps addresses. Because, in the panic
notifier, you're missing all those WARN* callsites, for example.

Also, I wonder whether it wouldn't be too hard to go even a step further
and compute the original, linker vmlinux addresses from the offsets
and dump a stack trace which looks exactly the same as if KASLR is
off. It'll probably need something to say KASLR was on when the trace
happened, though:

[  790.253365] Call Trace (KASLR):
[  790.254121]  [<ffffffff8110bc90>] ? __smpboot_create_thread+0x180/0x180
[  790.255428]  [<ffffffff810ff1df>] kthread+0xef/0x100
...

so that people who stare at this, know.

Because, in that case, you don't need both the panic notifier or the
userspace script massaging stack trace output anymore.

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--