Hello!
> When you e.g. have a TCP sniffer it makes sense to only bind it to ETH_P_IP.
For what purpose? To add a small underdeveloped copy of BPF?
If it was an optimization I would understand this of course.
But you propose deoptimization. :-)
> Do you worry about the handling of hundreds of packet sockets?
I worry about _one_ packet socket, which implements a protocol
in user space. And only about this. It is what packet sockets
are used for.
And I do want to see any refs to it in irrelevant place, which output path is.
To summarize: I wanted to see a patch allowing to detect that
nobody listens on outpu (or even splitting input and output ptype_all.)
So that it becomes possible to use ETH_P_ALL to listen
for all frames, but not to abuse output path.
Opposite is just non-sense with no applications.
Alexey
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Nov 07 2001 - 21:00:16 EST