netdev - Re: [PATCH bpf-next] selftests/bpf: don't use bpf helpers in non-bpf environment

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]

Date:   Wed, 12 Dec 2018 10:04:59 +0100
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Stanislav Fomichev <sdf@...ichev.me>
Cc:     Stanislav Fomichev <sdf@...gle.com>, netdev@...r.kernel.org,
        davem@...emloft.net, ast@...nel.org
Subject: Re: [PATCH bpf-next] selftests/bpf: don't use bpf helpers in non-bpf
 environment

On 12/12/2018 04:13 AM, Stanislav Fomichev wrote:
> On 12/12, Daniel Borkmann wrote:
>> On 12/11/2018 10:49 PM, Stanislav Fomichev wrote:
>>> On 12/11, Stanislav Fomichev wrote:
>>>> We're using bpf_htons in test_progs.c to initialize some static
>>>> global data and I think I hit some weird case on an older compiler
>>>> which doesn't have __builtin_bswap16 (and __builtin_constant_p
>>>> expands to false).
>>>>
>>>> In this case I see:
>>>> error: implicit declaration of function '__builtin_bswap16'
>>
>> Is that gcc < 4.8?
> Yes.
> 
>>>> Let's explicitly use __constant_htons which should be exposed by the
>>>> linux/byteorder.h uapi header.
>>>
>>> Forgot to mention, that using simple htons produces the following:
>>>     test_progs.c:54:17: error: braced-group within expression allowed only
>>>     inside a function
>>>       .eth.h_proto = htons(ETH_P_IP),
>>>
>>>> Signed-off-by: Stanislav Fomichev <sdf@...gle.com>
>>>> ---
>>>>  tools/testing/selftests/bpf/test_progs.c | 8 ++++----
>>>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/tools/testing/selftests/bpf/test_progs.c b/tools/testing/selftests/bpf/test_progs.c
>>>> index 26f1fdf3e2bf..61593d319c0e 100644
>>>> --- a/tools/testing/selftests/bpf/test_progs.c
>>>> +++ b/tools/testing/selftests/bpf/test_progs.c
>>>> @@ -51,10 +51,10 @@ static struct {
>>>>  	struct iphdr iph;
>>>>  	struct tcphdr tcp;
>>>>  } __packed pkt_v4 = {
>>>> -	.eth.h_proto = bpf_htons(ETH_P_IP),
>>>> +	.eth.h_proto = __constant_htons(ETH_P_IP),
>>
>> If the __builtin_constant_p() evaluated to false on the constants (?),
>> wouldn't using the __bpf_constant_htons() directly work as well given
>> it's not using a builtin either? Should be fine either way though using
>> the same api/header might be slightly nicer.
> I got it wrong, __builtin_constant_p() evaluates correctly, I played
> with it a bit. But for some reason it still complains about that branch
> that it doesn't take :-/
> 
> Using __bpf_constant_htons() is a good idea, I'll follow up with a v2.

Ok, sounds good, thanks!