Messages in this thread | | | Date | Wed, 23 Jul 2014 09:49:16 -0700 | From | Alexei Starovoitov <> | Subject | Re: [PATCH net-next 1/2] net: filter: split filter.c into two files |
| |
On Wed, Jul 23, 2014 at 10:54:04AM +0200, Daniel Borkmann wrote: > On 07/23/2014 08:11 AM, Varka Bhadram wrote: > >On 07/23/2014 11:31 AM, Alexei Starovoitov wrote: > >>BPF is used in several kernel components. This split creates logical boundary > >>between generic eBPF core and the rest > >> > >>kernel/bpf/core.c: eBPF interpreter > >> > >>net/core/filter.c: classic->eBPF converter, classic verifiers, socket filters > >> > >>This patch only moves functions. > > > >If we are moving the code also its good to do cleanup. > > > >Run checkpatch.pl on this... > > Not sure what you understand as a cleanup, but there's nothing > wrong with it and most things reported are either intentional or > false positives as you might have seen already. The only thing
+1. interpreter code is quite polished already. Few lines are longer than 80 char, but wrapping will look ugly. The rest are false positives, since checkpatch is struggling to understand jump-threaded code.
> one could do later on perhaps is to rename related functions to > have consistently a 'bpf' prefix instead of 'sk' since they > won't be under net/ anymore.
yes. it's on my todo list. The sk_* functions and 'struct sk_filter' look out of place in kernel/bpf/* The first renaming patch of sock_filter_int->bpf_insn I've sent as rfc: http://patchwork.ozlabs.org/patch/371332/
'struct sk_filter' I'm thinking to rename as 'struct bpf_prog'
Speaking of kernel/bpf/* comment style... I would prefer to keep networking comments to save lines.
| |