Re: Is PACKET_RX_RING broken in 2.6.0-test?

[P]

Oded Comay wrote:
> The PACKET_RX_RING mechanism (enabled by CONFIG_PACKET_MMAP) appears to be 
> somewhat broken in 2.6.0-test* (2.6 for short). There are 3 issues:
>
> 1. Performance in 2.6 is much lower than in 2.4. As an example, using the 
> same hardware and under the same traffic load, the attached sample program 
> could process 177K packets/sec in 2.4, but only 70K packets/sec in 2.6.

I'm guessing the scheduling changes are the cause?
Try messing with sched_setscheduler (like:
http://awgn.antifork.org/codes/brute.c)

Anyone else notice that the packet buffer messes up if
packets are being received while it's being created.
I worked around it here by bringing the interfaces
up AFTER the buffer is created.

Pádraig.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/