Dear NetBSD/alpha:

I thought I'd report some tlp errors which appeared on my AS200 
4/166's console yesterday and today, viz.

tlp0: incoming packet spilled, resetting
tlp0: receive ring overrun

appeared repeatedly on the console until I rebooted it, after which

tlp0: receive error: MII error
tlp0: receive error: dribbling bit
tlp0: receive error: CRC error

showed up a couple of times.

I'd been using de for some weeks before switching to tlp and hadn't 
seen any errors.  More details are appended to this message.


Ray



Further particulars:

Shortly before the "packet spilled errors" appeared I'd been ftping a 
~ 470 MB file to the machine a number of times.  I didn't notice, but 
it's possible they began to occur during those sessions--the transfer 
rate certainly slowed down a lot, e.g. from 902.79 KB/s to 125.18 
KB/s (averaged over the whole transfer).

The AS200 has 144 MB RAM, is using its built in network card, and is 
running NetBSD/alpha 1.5.2.  This is part of its dmesg's output:

tlp0 at pci0 dev 11 function 0: DECchip 21040 Ethernet, pass 2.3
tlp0: interrupting at isa irq 5
tlp0: Ethernet address 08:00:2b:e7:b5:17
tlp0: 10baseT, 10baseT-FDX, 10base5, manual

The source of the file was a SparcStation 10 with 448 MB RAM using 
the le0 driver for its internal 10baseT interface, and running 
NetBSD/sparc 1.5.2.  I also tried fetching it from a PC with 128 MB 
RAM, a 10/100baseT card which was using the rtk driver, and running 
NetBSD/i386 1.5.2.

The point of the ftp'ing exercise was to see if there was a 
noticeable difference in performance between the de and tlp drivers: 
I moved the file between the machines a few times with both the de 
and tlp drivers.

To make the two kernels for the tests, I applied the SA2002-001 
ptrace patch to the 1.5.2 sys source tree I made a copy of the 
GENERIC configuration file and enabled KERNFS in it; the de kernel 
was built from that.  The tlp kernel config file was the same except 
that

de*     at      pci? dev ? function ?           # 21x4x-based Ethernet

was commented out and

tlp*    at      pci? dev ? function ?           # DECchip 21x4x and clones
options        TLP_MATCH_21040

were uncommented.