Re: BUG: bad usercopy in __check_object_size (2)

[Dmitry Vyukov]

On Fri, Sep 7, 2018 at 9:57 PM, Kees Cook <keescook@xxxxxxxxxx> wrote:
> On Fri, Sep 7, 2018 at 9:17 AM, Tetsuo Handa
> <penguin-kernel@xxxxxxxxxxxxxxxxxxx> wrote:
>> On 2018/09/08 0:29, syzbot wrote:
>>> syzbot has found a reproducer for the following crash on:
>>>
>>> HEAD commit:    28619527b8a7 Merge git://git.kernel.org/pub/scm/linux/kern..
>>> git tree:       bpf
>>> console output: https://syzkaller.appspot.com/x/log.txt?x=124e64d1400000
>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=62e9b447c16085cf
>>> dashboard link: https://syzkaller.appspot.com/bug?extid=a3c9d2673837ccc0f22b
>>> compiler:       gcc (GCC) 8.0.1 20180413 (experimental)
>>> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=179f9cd1400000
>>> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=11b3e8be400000
>>>
>>> IMPORTANT: if you fix the bug, please add the following tag to the commit:
>>> Reported-by: syzbot+a3c9d2673837ccc0f22b@xxxxxxxxxxxxxxxxxxxxxxxxx
>>>
>>>  entry_SYSCALL_64_after_hwframe+0x49/0xbe
>>> RIP: 0033:0x440479
>>> usercopy: Kernel memory overwrite attempt detected to spans multiple pages (offset 0, size 64)!
>>
>> Kees, is this because check_page_span() is failing to allow on-stack variable
>>
>>    u8 opcodes[OPCODE_BUFSIZE];
>>
>> which by chance crossed PAGE_SIZE boundary?
>
> There are a lot of failure conditions for the PAGESPAN check. This
> might be one (and one that I'm hoping to solve separately).

Disabled CONFIG_HARDENED_USERCOPY_PAGESPAN on syzbot:
https://github.com/google/syzkaller/commit/be20da425029ecd45b18e99fa5f09691ba0658ea

#syz invalid