On Wednesday, September 18, 2002, at 08:03 AM, Petri Helenius wrote:
>
> I played around adjusting udp socket buffers for a while and noticed
> that if the input buffer is set to a value, packets start getting
> dropped
> when npkt*MTU > SO_RCVBUF so if a socket receives 100 byte packets over
> an ethernet interface of 1500 byte MTU and receive buffer of 100k the
> packets
> start dropping at less than 10k received data in a buffer.
This is, I think, normal behavior. Check Wright/Stevens (TCP/IP
Illustrated, V2), Ch. 2, where this is discussed (as I recall). A
socket buffer counts not only the valid data bytes enqueued, but also
the size of the mbufs used. The reasoning is clear: in order to avoid
having all the mbufs in the system end up on a single queue, because
very small packets are being received, counting mbuf space limits the
number of mbufs that can be sucked up by one direction for one socket.
Regards,
Justin
--
Justin C. Walker, Curmudgeon-At-Large *
Institute for General Semantics | It's not whether you win or
lose...
| It's whether *I* win or lose.
*--------------------------------------*-------------------------------*
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-net" in the body of the message
