Re: v100 ethernet problems

[Marc Horowitz]

Justin A <[EMAIL PROTECTED]> writes:

>> On Sun, 2004-03-14 at 17:41, Marc Horowitz wrote:
>> 
>> > Problem 3: Full-duplex doesn't work right.  At boot, the interface
>> > autonegotiates to 100baseTx-HD.  If I use mii-tool to force it to full
>> > duplex, I see reduced throughput, and a transmit error on every packet
>> > is reported (but the packets are sent).
>> > 
>> > Marc
>> 
>> My ultra2 did this when it was plugged into my old 10mbit switch(it went
>> to 10 hd) If I forced it to 10 fd, it barely worked.  Since I've gotten
>> a new 100mbit switch, I haven't had any problems

I've got an HP procurve 2524 managed 10/100 switch.  I've tried many
combinations of settings on both ends, but I couldn't win.  Specific
suggestions of settings are encouraged.

Marc

v100 ethernet problems

[Marc Horowitz]

I'm seeing several problems with the network on my v100 machine.  I'm
running 2.4.21.  I compared the sources between this and 2.4.25, and
it looks like nothing relevant has changed, but I'm willing to try a
newer kernel if someone tells me it fixes any of these problems.

Problem 1: Somewhat correlated with periods of moderate to heavy
traffic, the interface stops working and I get an infinite series of
these:

NETDEV WATCHDOG: eth1: transmit timed out
NETDEV WATCHDOG: eth1: transmit timed out
NETDEV WATCHDOG: eth1: transmit timed out
NETDEV WATCHDOG: eth1: transmit timed out
NETDEV WATCHDOG: eth1: transmit timed out
NETDEV WATCHDOG: eth1: transmit timed out
NETDEV WATCHDOG: eth1: transmit timed out
NETDEV WATCHDOG: eth1: transmit timed out

if I log into the serial console and ifdown/ifup the interface,
everything works ok until it happens again.

Problem 2: Sometimes, after many timeouts as in problem 1, the host
panics:

  \|/  \|/
  "@'/ .. \`@"
  /_| \__/ |_\
 \__U_/
swapper(0): Kernel bad sw trap 5
TSTATE: 004480f09600 TPC: 00428cb0 TNPC: 00428cb4 Y: 
Not tainted
Using defaults from ksymoops -t elf32-sparc -a sparc
g0: 002d g1:  g2:  g3: 
f800659b
g4: f800 g5: 1de2 g6: 00414000 g7: 

o0:  o1: 0001 o2: 003f o3: 
007ab000
o4: 007ab230 o5:  sp: 00417091 ret_pc: 
00428be8
l0: f80066ff8220 l1: f8006783 l2: f8006058e010 l3: 
000f
l4: f80067eb8820 l5: 0003 l6: 0003 l7: 
f800650a7650
i0: 1fff i1: f80067831de2 i2: 0001 i3: 
0001
i4: 001b i5: 06898702 i6: 00417151 i7: 
005463f8
Caller[005463f8]
Caller[005f9df0]
Caller[005f0c8c]
Caller[0044de84]
Caller[0040ef40]
Caller[0041a504]
Caller[007206f4]
Caller[00404638]
Caller[]
Instruction DUMP: 10680004  0100  9194c000 <91d02005> 9194c000  
81cfe008  91316000  9de3bf40  83366000 

ksymoops says:

>>PC;  00428cb0<=

>>g6; 00414000 
>>o3; 007ab000 
>>o4; 007ab230 
>>sp; 00417091 
>>ret_pc; 00428be8 
>>i6; 00417151 
>>i7; 005463f8 

Trace; 005463f8 
Trace; 005f9df0 
Trace; 005f0c8c 
Trace; 0044de84 
Trace; 0040ef40 <__handle_softirq+0/10>
Trace; 0041a504 
Trace; 007206f4 
Trace; 00404638 
Trace;  Before first symbol

Code;  00428ca4 
 <_PC>:
Code;  00428ca4 
   0:   10 68 00 04   unknown
Code;  00428ca8 
   4:   01 00 00 00   nop 
Code;  00428cac 
   8:   91 94 c0 00   unknown
Code;  00428cb0<=
   c:   91 d0 20 05   ta  5   <=
Code;  00428cb4 
  10:   91 94 c0 00   unknown
Code;  00428cb8 
  14:   81 cf e0 08   rett  %i7 + 8
Code;  00428cbc 
  18:   91 31 60 00   srl  %g5, 0, %o0
Code;  00428cc0 
  1c:   9d e3 bf 40   save  %sp, -192, %sp
Code;  00428cc4 
  20:   83 36 60 00   srl  %i1, 0, %g1

Problem 3: Full-duplex doesn't work right.  At boot, the interface
autonegotiates to 100baseTx-HD.  If I use mii-tool to force it to full
duplex, I see reduced throughput, and a transmit error on every packet
is reported (but the packets are sent).

Marc

Re: Losing my mind, RAID1 on Sparc completely broken?

[Marc Horowitz]

I'm having the exact same problem.  Check out the thread starting with
.

booting with nodma has made my ide subsystem completely stable,
including an md raid1 mirror pair.  I'm still trying to get the bug
workaround described in that thread working, but my source has been
out of town, so I'm stalled for now.

Marc

SunBlade 100 install hangs

[Marc Horowitz]

I just tried following the instructions posted here

a few weeks ago, and I cannot make it work.

When I netboot either the 4 Dec or 5 Dec tftpboot.img from
phunnypharm, the OBP (version 4.10.6) says

Remapping the kernel... done.
Booting Linux...

Then hangs forever.  If I try to use

or the 5 Dec variant, either the linux or rescue target, I get

Uncompressing image...

Decompression error: uncompressed image too long - wouldn't fit into 
destination


Error loading /dists/stable/main/disks-sparc/current/sparc64/vmlinuz


mentions a 2.4.23 iso at
, but
it isn't there.

What else should I try to get some sort of install going?

Marc

Re: cd boot fails: kernel can't find root fs

[Marc Horowitz]

I got my ss20 to boot the install cdrom!

I had 4x32mb memory modules in the machine.  On a lark, I pulled two
of them out, hoping to affect the memory allocation algorithm, and it
worked.

If anybody else out there has an ss20 they've been unable to boot off
cdrom, try pulling memory until it's got only 64mb, or perhaps
anywhere than 128mb.  I'm curious if my solution is reproducible.

There's still a bug here, but this is a sufficient workaround for me.
When I'm done installing, I'll put the memory back in.

Marc

Re: cd boot fails: kernel can't find root fs

[Marc Horowitz]

On Tue, Dec 03, 2002 at 12:36:16AM -0500, Marc Horowitz wrote:
> I have an ss10 with a cdrom drive and no floppy.  I used jigdo to
> download and burn a 3.0 cdrom image, which worked without a hitch.
> When I put the media in and boot it, SILO loads, and I hit enter to
> load the default kernel image.  the kernel boots, then things stop
> working so perfectly:

I had time to look into this, and I noticed a telltale error at the
beginning of the boot sequence:

Uncompressing image...
Loading initial ramdisk
PROMLIB: obio_ranges 5
Booting Linux...
PROMLIB: Sun Boot Prom Version 3 Revision 2
Linux version 2.2.20 ([EMAIL PROTECTED]) (gcc version 2.95.2 2220 
(Debian GNU/Linux)) #1 Fri Nov 16 15:48:02 EST 2001
ARCH: SUN4M
TYPE: Sun4m SparcStation10/20
Ethernet address: 8:0:20:73:46:22
Boot time fixup v1.6. 4/Mar/98 Jakub Jelinek ([EMAIL PROTECTED]). Patching 
kernel for srmmu[TI Viking]/iommu
initrd extends beyond end of memory (0xf811adbd > 0xf800)
disabling initrd
[0]: v[f000,f100](100) p[]
[1]: v[f100,f200](100) p[0200]
[2]: v[f200,f300](100) p[0800]
[3]: v[f300,f400](100) p[0a00]
[4]: v[f400,f500](100) p[0c00]
[5]: v[f500,f600](100) p[0e00]
[6]: v[f600,f700](100) p[1400]
[7]: v[f700,f7f4f000](f4f000) p[1600]

It appears the end of memory (in virtual address terms) is at
0xf800.  The "initrd extends beyond end of memory" error indicates
that my initrd image ends at 0xf811adbd.  This looks like a random
number, but 0x11adbd == 1158589, which is the size of the root image:

-r--r--r--3 root root  1158589 May 18  2002 
dists/stable/main/disks-sparc/current/images-1.44/root.bin

So, the kernel is apparently looking for the root image right at the
end of memory.  The boot loader seems to load the ramdisk (at least,
it spends some time doing something between the second and third lines
above).  So, either silo is writing the root image into nowhere, or
silo isn't passing the address of the root image into the kernel
peoperly.

I took some look at the silo code, but I don't know nearly enough
about the sparc to know what to do to fix this.

Marc

Re: cd boot fails: kernel can't find root fs

[Marc Horowitz]

Nate Campi <[EMAIL PROTECTED]> writes:

>> I just type 'rescue' at the CD's SILO prompt and install successfully
>> from there. I can't seem to use it as an actual rescue CD, though :(

I tried this, but it fails the same way as before.

Marc

cd boot fails: kernel can't find root fs

[Marc Horowitz]

I have an ss10 with a cdrom drive and no floppy.  I used jigdo to
download and burn a 3.0 cdrom image, which worked without a hitch.
When I put the media in and boot it, SILO loads, and I hit enter to
load the default kernel image.  the kernel boots, then things stop
working so perfectly:

scsi : 1 host.
  Vendor: QUANTUM   Model: FB1080J  SUN1.05  Rev: 630E
  Type:   Direct-Access  ANSI SCSI revision: 02
Detected scsi disk sda at scsi0, channel 0, id 3, lun 0
  Vendor: TOSHIBA   Model: XM-4101TASUNSLCD  Rev: 1084
  Type:   CD-ROM ANSI SCSI revision: 02
Detected scsi CD-ROM sr0 at scsi0, channel 0, id 6, lun 0
scsi : detected 1 SCSI cdrom 1 SCSI disk total.
esp0: target 6 asynchronous
Uniform CD-ROM driver Revision: 3.11
esp0: target 3 [period 100ns offset 8 10.00MHz FAST SCSI-II]
SCSI device sda: hdwr sector= 512 bytes. Sectors= 2134305 [1042 MB] [1.0 GB]
sunlance.c:v1.12 11/Mar/99 Miguel de Icaza ([EMAIL PROTECTED])
eth0: LANCE 08:00:20:73:46:22 
eth0: using auto-carrier-detection.
Partition check:
 sda: unknown partition table
Root-NFS: No NFS server available, giving up.
VFS: Unable to mount root fs via NFS, trying floppy.
VFS: Insert root floppy and press ENTER

ENTER, not surprisingly, fails.

request_module[block-major-2]: Root fs not mounted
VFS: Cannot open root device 02:00
Kernel panic: VFS: Unable to mount root fs on 02:00
Press L1-A to return to the boot prom

The host is obviously seeing the cdrom drive.  The kernel appears to
be seeing and loading proper drivers for the cdrom drive.  But it's
loading the root fs from the wrong place.  /boot/silo.conf has an
initrd line which points to a file which contains a compressed ext2 fs
root image, as I would expect, but the systme doesn't appear to be
loading it.

Does anybody have any suggestions?

Thanks!

Marc