linux-kernel - Re: [kernel-hardening] Re: [PATCH] usercopy: Add tests for all get

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]

Message-ID: <878tp3j12z.fsf@concordia.ellerman.id.au>
Date:   Sat, 18 Feb 2017 20:32:52 +1100
From:   Michael Ellerman <mpe@...erman.id.au>
To:     Kees Cook <keescook@...omium.org>,
        Geert Uytterhoeven <geert@...ux-m68k.org>
Cc:     "linux-kernel\@vger.kernel.org" <linux-kernel@...r.kernel.org>,
        Hoeun Ryu <hoeun.ryu@...il.com>,
        "kernel-hardening\@lists.openwall.com" 
        <kernel-hardening@...ts.openwall.com>
Subject: Re: [kernel-hardening] Re: [PATCH] usercopy: Add tests for all get_user() sizes

Kees Cook <keescook@...omium.org> writes:

> On Wed, Feb 15, 2017 at 12:50 AM, Geert Uytterhoeven
> <geert@...ux-m68k.org> wrote:
>> On Tue, Feb 14, 2017 at 9:40 PM, Kees Cook <keescook@...omium.org> wrote:
>>> The existing test was only exercising native unsigned long size
>>> get_user(). For completeness, we should check all sizes.
>>>
>>> Signed-off-by: Kees Cook <keescook@...omium.org>
>>> ---
>>>  lib/test_user_copy.c | 45 ++++++++++++++++++++++++++++++++++-----------
>>>  1 file changed, 34 insertions(+), 11 deletions(-)
>>>
>>> diff --git a/lib/test_user_copy.c b/lib/test_user_copy.c
>>> index ac3a60ba9331..49569125b7c5 100644
>>> --- a/lib/test_user_copy.c
>>> +++ b/lib/test_user_copy.c
>>> @@ -40,8 +40,11 @@ static int __init test_user_copy_init(void)
>>>         char __user *usermem;
>>>         char *bad_usermem;
>>>         unsigned long user_addr;
>>> -       unsigned long value = 0x5A;
>>>         char *zerokmem;
>>> +       u8 val_u8;
>>> +       u16 val_u16;
>>> +       u32 val_u32;
>>> +       u64 val_u64;
>>>
>>>         kmem = kmalloc(PAGE_SIZE * 2, GFP_KERNEL);
>>>         if (!kmem)
>>> @@ -72,10 +75,20 @@ static int __init test_user_copy_init(void)
>>>                     "legitimate copy_from_user failed");
>>>         ret |= test(copy_to_user(usermem, kmem, PAGE_SIZE),
>>>                     "legitimate copy_to_user failed");
>>> -       ret |= test(get_user(value, (unsigned long __user *)usermem),
>>> -                   "legitimate get_user failed");
>>> -       ret |= test(put_user(value, (unsigned long __user *)usermem),
>>> -                   "legitimate put_user failed");
>>> +
>>> +#define test_legit(size)                                                 \
>>> +       do {                                                              \
>>> +               ret |= test(get_user(val_##size, (size __user *)usermem), \
>>> +                   "legitimate get_user (" #size ") failed");            \
>>> +               ret |= test(put_user(val_##size, (size __user *)usermem), \
>>> +                   "legitimate put_user (" #size ") failed");            \
>>> +       } while (0)
>>> +
>>> +       test_legit(u8);
>>> +       test_legit(u16);
>>> +       test_legit(u32);
>>> +       test_legit(u64);
>>> +#undef test_legit
>>
>> ERROR: "__get_user_bad" [lib/test_user_copy.ko] undefined!
>>
>> http://kisskb.ellerman.id.au/kisskb/buildresult/12936728/
>>
>> So 64-bit get_user() support is mandatory now?
>
> That's not my intention. :) In my sampling of architectures, I missed
> a couple 32-bit archs that don't support 64-bit getuser(). I'm not
> sure how to correctly write these tests, though, since it seems rather
> ad-hoc. e.g. m68k has 64-bit getuser() commented out due to an old gcc
> bug...
>
> Should I just universally skip 64-bit getuser on 32-bit archs?

I think you should just make it opt-in for 32-bit arches.

cheers