Re: [PATCH] seccomp: fix populating a0-a5 syscall args in 32-bit x86 BPF

From: Alexei Starovoitov
Date: Tue Apr 15 2014 - 02:32:02 EST


On Mon, Apr 14, 2014 at 1:28 PM, Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
> On Mon, Apr 14, 2014 at 1:24 PM, David Miller <davem@xxxxxxxxxxxxx> wrote:
>> From: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
>> Date: Mon, 14 Apr 2014 13:13:45 -0700
>>
>>> I think this description is wrong. (unsigned long *) &sd->args[1] is
>>> the right location, at least on 32-bit little-endian architectures.
>>
>> It absolutely is not.
>
> Huh? It's a pointer to the right address, but the type is wrong.
>
> The changelog says "on 32-bit x86 (or any other 32bit arch), it would
> result in storing a0-a5 at wrong offsets in args[] member". Unless
> I'm mistaken, this is incorrect: a0-a5 are are the correct offsets,
> but they are stored with the wrong type, so the other bits in there
> are garbage.

agree. your above description is more correct than the log.
We were focusing on the bug itself and the log came a bit misleading
as a result of multiple iterations back and forth between me and Daniel.

also the log says:
"gcc is clever and optimizes the copy away in other cases, e.g. x86_64"
since we actually checked assembler, so the fix doesn't pessimize
64-bit architectures :)
This function is in critical path for seccomp, so performance definitely
matters.

>>
>> The thing is a u64, and we must respect that type in a completely
>> portable way.
>>
>> Daniel's change is %100 correct, portable, and doesn't have any
>> ugly ifdef crap.
>>
>
> I have no problem with the patch itself. I'm suggesting that a better
> changelog message would confuse other people reading the same patch
> less.
>
> --Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/