[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <2D15A822-5BE1-4C4A-84B2-46FFA27AC32B@fb.com>
Date: Thu, 20 May 2021 17:56:30 +0000
From: Song Liu <songliubraving@...com>
To: Alexei Starovoitov <alexei.starovoitov@...il.com>
CC: Dmitrii Banshchikov <me@...que.spb.ru>,
"open list:BPF (Safe dynamic programs and tools)"
<bpf@...r.kernel.org>, Alexei Starovoitov <ast@...nel.org>,
"davem@...emloft.net" <davem@...emloft.net>,
"daniel@...earbox.net" <daniel@...earbox.net>,
"andrii@...nel.org" <andrii@...nel.org>, Martin Lau <kafai@...com>,
Yonghong Song <yhs@...com>,
"john.fastabend@...il.com" <john.fastabend@...il.com>,
"kpsingh@...nel.org" <kpsingh@...nel.org>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
Andrey Ignatov <rdna@...com>
Subject: Re: [PATCH bpf-next 00/11] bpfilter
> On May 20, 2021, at 9:55 AM, Alexei Starovoitov <alexei.starovoitov@...il.com> wrote:
>
> On Thu, May 20, 2021 at 12:53 AM Dmitrii Banshchikov <me@...que.spb.ru> wrote:
>>
>> On Thu, May 20, 2021 at 04:54:45AM +0000, Song Liu wrote:
>>>
>>>
>>>> On May 17, 2021, at 3:52 PM, Dmitrii Banshchikov <me@...que.spb.ru> wrote:
>>>>
>>>> The patchset is based on the patches from David S. Miller [1] and Daniel
>>>> Borkmann [2].
>>>>
>>>> The main goal of the patchset is to prepare bpfilter for iptables'
>>>> configuration blob parsing and code generation.
>>>>
>>>> The patchset introduces data structures and code for matches, targets, rules
>>>> and tables.
>>>>
>>>> It seems inconvenient to continue to use the same blob internally in bpfilter
>>>> in parts other than the blob parsing. That is why a superstructure with native
>>>> types is introduced. It provides a more convenient way to iterate over the blob
>>>> and limit the crazy structs widespread in the bpfilter code.
>>>>
>>>
>>> [...]
>>>
>>>>
>>>>
>>>> 1. https://lore.kernel.org/patchwork/patch/902785/
>>>
>>> [1] used bpfilter_ prefix on struct definitions, like "struct bpfilter_target"
>>> I think we should do the same in this version. (Or were there discussions on
>>> removing the prefix?).
>>
>> There were no discussions about it.
>> As those structs are private to bpfilter I assumed that it is
>> safe to save some characters.
>> I will add the prefix to all internal structs in the next
>> iteration.
>
> For internal types it's ok to skip the prefix otherwise it's too verbose.
> In libbpf we skip 'bpf_' prefix in such cases.
Do we have plan to put some of this logic in a library? If that is the case, the
effort now may save some pain in the future.
Thanks,
Song
Powered by blists - more mailing lists