Re: [kernel-hardening] [PATCH v4 02/10] asm/nospec, array_ptr: sanitize speculative array de-references

[Dan Williams]

On Fri, Jan 19, 2018 at 10:18 AM, Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Fri, Jan 19, 2018 at 2:20 AM, Jann Horn <jannh@xxxxxxxxxx> wrote:
>>> +                                                                       \
>>> +       __u._ptr = _arr + (_i & _mask);                                 \
>>> +       __u._bit &= _mask;                                              \
>>
>> AFAICS, if `idx` is out of bounds, you first zero out the index
>> (`_i & _mask`) and then immediately afterwards zero out
>> the whole pointer (`_u._bit &= _mask`).
>> Is there a reason for the `_i & _mask`, and if so, can you
>> add a comment explaining that?
>
> I think that's just leftovers from my original (untested) thing that
> also did the access itself. So that __u._bit masking wasn't masking
> the pointer, it was masking the value that was *read* from the
> pointer, so that you could know that an invalid access returned
> 0/NULL, not just the first value in the array.

Yes, the index masking can be dropped since we're returning a
sanitized array element pointer now.