Re: kern/60453

["Erik LaBine via gnats" <gnats-admin%NetBSD.org@localhost>]

The following reply was made to PR kern/60453; it has been noted by GNATS.

From: Erik LaBine <erik%eriklabine.com@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: kern/60453
Date: Mon, 19 Oct 2026 14:54:32 -0400

 I'm terribly sorry for the delay in responding to this issue; I had hoped to 
 do a thorough test on the hardware post-merge, and I've only just today had 
 the chance to get back at it.
 
 I've run a series of tests on a SparcStation 10 running a Ross hyperSparc 125 
 MHz processor. The iperf3 tests were against a local network server and were 
 run for 120 seconds to ensure they could handle a sustained load. Parallel 
 performance was about the same as single, as you'd expect.
 
 iperf3 sender: 334MB, 23.3 Mbit/s
 iperf3 receiver: 368 MB, 25.7 Mbit/s
 ftp of current pkgsrc: 101534035 bytes retrieved in 03:21 (490.96 KiB/s)
 
 That's all consistent with the numbers I was getting earlier.
 
 For the iperf3 receive mode test, it's giving a decent number of 
 retransmissions from the server, consistently bout 7 per second for my setup. 
 Still not quite sure what that's about, but the one time I thought to record 
 the data from netstat it was 268,680 input packets to 705 input errors.
 
 My theory is that the interface is simply receiving data faster than the 20KB 
 RX fifo can handle it, since downloading pkgsrc at ~1/6 the data rate didn't 
 cause additional errors.
 
 Haven't had a chance to check any of this on another machine as yet, although 
 I think the interface is working quite well at this point!
 
 -Erik