On Thu, Aug 17, 2023 at 02:05 PM +02, Ferenc Fejes wrote:
> Hi Liu!
>
> On Fri, 2023-08-11 at 17:32 +0800, Liu Jian wrote:
>> If the sockmap msg redirection function is used only to forward
>> packets
>> and no other operation, the execution result of the
>> BPF_SK_MSG_VERDICT
>> program is the same each time. In this case, the BPF program only
>> needs to
>> be run once. Add BPF_F_PERMANENTLY flag to bpf_msg_redirect_map() and
>> bpf_msg_redirect_hash() to implement this ability.
>
> Did you considered other names for this flag e.g. BPF_F_SPLICED or
> BPF_F_PIPED?
Ferenc,
A reference to splice/pipe syscall certainly paints a picture.
But I'm not sure if it makes it more intutive or more confusing in the
context of bpf_{msg,sk}_redirect_{hash,map}. Consider:
bpf_msg_redirect_map(..., BPF_F_SPLICE)
vs
bpf_msg_redirect_map(..., BPF_F_PERMANENTLY)
Liu,
No need to go for the adverb form ("PERMANENTLY"). An adjective
("PERMANENT") will as expressive here. So BPF_F_PERMANENT is what I'm
suggesting.
Also, I'm thinking maybe it's time for a dedicated prefix to avoid name
clashes, like BPF_F_ADJ_ROOM_*.
BPF_F_INGRESS, which is also accepted by other helpers. But that won't
be the case with the new flag. BPF_F_SK_REDIR_*? That would make it
BPF_F_SK_REDIR_PERMANENT.
Alternatively, BPF_F_SK_REDIR_FIXED comes to mind. Naming is hard.
[...]
