netdev - Re: [iproute2 net-next 3/8] bpf: Add BPF

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:   Mon, 12 Dec 2016 12:48:52 +0100
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Hannes Frederic Sowa <hannes@...essinduktion.org>,
        David Ahern <dsa@...ulusnetworks.com>, netdev@...r.kernel.org,
        stephen@...workplumber.org
Subject: Re: [iproute2 net-next 3/8] bpf: Add BPF_ macros

On 12/12/2016 12:28 PM, Hannes Frederic Sowa wrote:
> On 12.12.2016 01:53, David Ahern wrote:
>> Based on version in kernel repo, samples/bpf/libbpf.h
>>
>> Signed-off-by: David Ahern <dsa@...ulusnetworks.com>
>> ---
>>   include/bpf_util.h | 179 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 179 insertions(+)
>>
>> diff --git a/include/bpf_util.h b/include/bpf_util.h
>> index 726e34777755..5361dab1933d 100644
>
> Maybe this was already discussed: why are those not part of uapi? They
> get used in the bpf manpage.

The reason why these macros haven't been exposed to uapi header was because
they are not strict part of kernel abi. So really only the very minimum that
had to go there (such as insn opcodes) were placed into the header. That way,
kernel is still free to, for example, change or rename the macros once new
insns would get added, and user space can just define what it really needs.
That way, both don't get into each others way long term. Often it's also not
really needed if you have an elf loader, f.e. iproute2 could live without
them (up to this point).

Just took a look at the man page example, we don't even define BPF_CALL_FUNC()
in the kernel, so it's also not exactly the same. Probably makes sense to
fix/complete that example and in general add a description there wrt opcodes.
Last time I spoke with Michael, he planned to make a bigger revision to the
man page to improve documentation.

Thanks,
Daniel