Subject: Re: ie0: DMA underrun
To: David Jones <dej@achilles.net>
From: Jason Thorpe <thorpej@SJ.Xenotropic.COM>
List: port-sparc
Date: 11/19/1995 09:01:55
On Sun, 19 Nov 1995 08:36:49 -0500 (EST)
dej@achilles.net (David Jones) wrote:
I haven't experienced any data lossage yet. The world has been building
on my MVME-147 (for which my 4/260 is the fileserver) all night, and it
seems to be OK; still going, all the binaries work, etc.
> Funny thing - I don't get DMA underruns running Gordon's latest incarnation
> of the si driver. It might be worse if X is running, I dunno.
Yah, I've pulled in Gordon's latest patches. While they work terrific, I
still get _lots_ of DMA underrun errors on the ethernet.
Speaking of which, I have some changes to the "si_options" handling I'll
forward on to Gordon; I rather liked the idea of changing them with the
"flags" directive in the kernel config file...
> What I find really disturbing is that you get underruns even without
> SCSI DMA. Is the Ethernet on-board, or is it a separate VME card?
The ie is an onboard. Here's dmesg output (I had to beboot to get it
:-/). Ignore the framebuffer stuff; that's something else I've been
hacking on.
NetBSD 1.0A (BIGSBY) #93: Sat Nov 18 16:41:00 PST 1995
thorpej@bigsby:/tmp_mnt/basalt/work/netbsd/src/sys/arch/sparc/compile/BIGSBY
real mem = 25141248
avail mem = 22413312
using 153 buffers containing 1253376 bytes of memory
bootpath: /vmes0/si0/sd@0,0
mainbus0 (root)
cpu0 at mainbus0: SUN-4/200 series (MB86910 or WTL1164/5 FPU)
cpu0: 131072 byte write-back, 16 bytes/line, sw flush cache enabled
obio0 at mainbus0
oclock0 at obio0 addr f3000000
eeprom0 at obio0 addr f2000000
memreg0 at obio0 addr f4000000
zs0 at obio0 addr f1000000 pri 12, softpri 6
zs1 at obio0 addr f0000000 pri 12, softpri 6
bwtwo0 at obio0 addr fd000000: bwtwo, 1600 x 1280 (console)
bwtwo0: attached to /dev/fb
ie0 at obio0 addr f6000000 pri 6 address 08:00:20:40:31:17, type onboard
vmel0 at mainbus0
vmes0 at mainbus0
si0 at vmes0 addr ff200000 vec 0x40 pri 3
scsibus0 at si0
si0 targ 0 lun 0: <HITACHI, DK516C, NO34> SCSI2 0/direct fixed
sd0 at scsibus0: 1272MB, 2179 cyl, 15 head, 79 sec, 512 bytes/sec
root on sd0a
Another thing I've noticed: probing for drives _really_ takes a long
time; much longer than on my 3/60. This happened with the old "si"
driver, too. Might be a DELAY() thing. Dunno.
------------------------------------------------------------------------------
Jason R. Thorpe thorpej@Xenotropic.COM
Just me and my collection of obsolete computer gear(s).