lkml.org 
[lkml]   [2014]   [Mar]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH v9 net-next 1/3] filter: add Extended BPF interpreter and converter
From
On Wed, Mar 12, 2014 at 3:16 PM, Cong Wang <cwang@twopensource.com> wrote:
> (Sorry for jumping into this thread late.)
>
> On Mon, Mar 10, 2014 at 9:41 PM, Alexei Starovoitov <ast@plumgrid.com> wrote:
>>
>> 3. tracing filters systemtap-like with extended BPF
>>
>> 4. OVS with extended BPF
>>
>> 5. nftables with extended BPF
>
> If you plan to use extended BPF out of networking, does
> it make sense to make it independent of networking?
> That is, rename those sk_*() APIs?

well, v1 and v2 series already demonstrate how ebpf can be used for tracing.
That's why in v1 series they were under kernel/ dir, but since it's
a natural bpf extension that reuses a lot of bpf pieces, it makes sense
for the whole thing be in net/core/filter.c and in linux/filter.h
I believe that was a consensus in v2-v3 series.
I don't want to rename the whole thing once again.
Also seccomp is not networking, but it already uses bpf and
after these patches ebpf. So net/core/filter.c location is fine.

>>
>> +int bpf_ext_enable __read_mostly;
>> +
>
> This could be converted to a jump label, no?

yes, but it's not in critical path, so no need.

Thank you for taking a look!
Even late feedback is always appreciated.


\
 
 \ /
  Last update: 2014-03-13 00:21    [W:0.107 / U:1.016 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site