Re: [mm] 6128b3af2a: UBSAN:shift-out-of-bounds_in(null)

[David Hildenbrand]

On 19.10.21 17:49, Eric W. Biederman wrote:
> kernel test robot <oliver.sang@xxxxxxxxx> writes:
> 
>> Greeting,
>>
>> FYI, we noticed the following commit (built with clang-14):
>>
>> commit: 6128b3af2a5e42386aa7faf37609b57f39fb7d00 ("mm: ignore MAP_DENYWRITE in ksys_mmap_pgoff()")
>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
> 
> I believe this failure is misattributed.  Perhaps your reproducer
> only intermittently reproduces the problem?
> 
> The change in question only contains
> 
> 	flags &= ~MAP_DENYWRITE
> 
> After all of the other users of MAP_DENYWRITE had been removed from the
> kernel.  So I don't see how it could possibly be responsible for the
> reported shift out of bounds problem.
> 
> Eric

Thanks for looking into this Eric while I spent the last couple of days
in bed feeling miserable. :)


So we get 9 new instances of "UBSAN:shift-out-of-bounds_in(null)" (NULL
pointer dereference) on 6128b3af2a compared to 6128b3af2a^ (8d0920bde5),
apparently inside ksys_mmap_pgoff() on 32bit.

As we're dealing with a fuzzer, is there any reproducer as sometimes
provided by syzkaller? The report itself is not very helpful when
judging if that patch is actually responsible for what we're seeing.

I agree with Eric that it's rather unlikely that when we stop masking
off a bit that's ignored throughout the kernel, that we suddenly trigger
a NULL pointer de-reference. But I learned that everything is possible ;)

-- 
Thanks,

David / dhildenb