ANECDOTALLY, network driver geom times out under

moderate steady load on a Sun Blade 150 within an hour of use.

(as in network stops working, everything else is ok, and

dmesg says "geom0: device timeout" about every 30 seconds; and

reboot fixes it)
Haven't seen the problem yet under Linux.

Have seen it multiple times under OpenBSD 4.2.
The machine was able to be up for days no problem busy with cd

/usr/ports/x11/kde

&& sudo make, so generally all else is well.
I will try 4.3 soon.
I know this is poor problem report. It's about all I have.

I can do my work equally well for now under any BSD or Linux so punting is ok

for me.

(I'm porting some stuff to all of OpenBSD, NetBSD, FreeBSD, Linux.)
lspci from Linux:
s00:00.0 Host bridge: Sun Microsystems Computer Corp. Psycho UPA-PCI Bus

Module [pcipsy]

00:03.0 Non-VGA unclassified device: ALi Corporation M7101 Power Management

Controller [PMU]

00:05.0 PCI bridge: Intel Corporation 21152 PCI-to-PCI Bridge

00:07.0 ISA bridge: ALi Corporation M1533/M1535 PCI to ISA Bridge [Aladdin

IV/V/V+]

00:08.0 Multimedia audio controller: ALi Corporation M5451 PCI AC-Link

Controller Audio Device (rev 01)

00:0c.0 Bridge: Sun Microsystems Computer Corp. RIO EBUS (rev 01)

00:0c.1 Ethernet controller: Sun Microsystems Computer Corp. RIO 10/100

Ethernet [eri] (rev 01)

00:0c.2 FireWire (IEEE 1394): Sun Microsystems Computer Corp. RIO 1394 (rev

01)

00:0c.3 USB Controller: Sun Microsystems Computer Corp. RIO USB (rev 01)

00:0d.0 IDE interface: ALi Corporation M5229 IDE (rev c3)

00:13.0 VGA compatible controller: ATI Technologies Inc Rage XL (rev 27)

01:00.0 PCI bridge: Intel Corporation 21152 PCI-to-PCI Bridge

02:01.0 Bridge: Xerox Corporation Unknown device 0017

02:04.0 FireWire (IEEE 1394): Texas Instruments PCILynx/PCILynx2 IEEE 1394

Link Layer Controller (rev 04)
I assume Linux dmesg isn't interesting, and I can send OpenBSD dmesg if it

matters.
Fuller disclosure: The first install I saw this on I had partly done cd

/usr/src &a...
[ message continues ]

Are you running the latest version of OBP on the system?

_________________________________________________________________
All new Live Search at Live.com
http://clk.atdmt.com/UKM/go/msnnkmgl0010000006ukm/direct/01/

My OBP is definitely out of date, but I haven't yet been able to fix that.

I'm not sure if it requires installing Solaris (which file systems can OB

access?), which I've been unable to do, and I'm not sure if it requires

cracking open the machine, which I am reluctant to do, but might do.
Someone did tell me they see this in "current", however I finally noticed this

promising note:
http://www.openbsd.org/plus43.html

Changes made between OpenBSD 4.2 and OpenBSD 4.3:  >> Fixed watchdog timeouts

on gem(4).
Wireless is working surprisingly well, and my 4.3 CDs are en route. :)
 - Jay

Subject: RE: geom network driver times out on sparc 4.2?> Date: Sun, 11 May

2008 17:54:25 +0100> > > Are you running the latest version of OBP on the

system?> _________________________________________________________________

I gets a little more interesting..Linux/sparc was working fine for a few

days.Now it seems, merely plugging this machine into the network makes

thenetwork barely work -- very slow, connections dropped. Even ifI'm just

sitting in various installers. Plugging in the wired on board ethernet.

apt-get upgrade left Linux failing to boot.Solaris 9 install fails very early,

something about newfs failing.I really tried a bunch of manual parameters,

besides that itis run automatically.I think 10 is in the mail. (I have 10

onDVD but so far only a CD drive on the machine).

I think it's something about the disklabel/partition left by  either OpenBSD

or Linux.

I'm not sure how to get the OBP to the machine, since I can't install Solaris.

Not sure the others will do. Let alone the possible need to open it up to

enable writing.

So, anyway, for now, I have wireless working and am back in OpenBSD.It's a

little unfortunate because the Linux stuff was about a third done, now I'll

try to complete OpenBSD before going back to Linux to finish.

I was surprised how easy wireless was to get working.Granted I had to use the

command line and read the man page, butgiven that, not bad. I'm going to try

no X on the machine, just ssh. :)Of course, all my file editing will be on

Windows and I'll scp files over.Or maybe try NFS or Samba..
Oh, and the OBP is definitely out of date.

It's like version 4.6.5 from 2002, when they are up to like 4.10 or so fro

2005.
Oh, also, a little complaint about OpenBSD installer.

Solaris 9 installer pushes you to have swap first.

So then with that in mind, i did that for OpenBSD.

But OpenBSD defaults to installing to the first partition, and doesn't

let you change that, so with 512meg, 10gig / after it, setup runs

out of room installing to swap. Fixed by taking the default disklabel

and using a for add and going more with its defaults, instead of what

Solaris wanted. For once, I want more manual override..or for

setup to notice..and maybe guess there isn't enough space...

 - Ja...
[ message continues ]

if you mean the installer complains about disk failure then try this:

http://www.geeklan.co.uk/?p=61
otherwise I'd highly recommend you upgrade the OBP!
Sevan / Venture37

_________________________________________________________________

Great deals on almost anything at eBay.co.uk. Search, bid, find and win on

eBay today!

http://clk.atdmt.com/UKM/go/msnnkmgl0010000004ukm/direct/01/

I don't think that is it, but I will try, thank you.

I'm pretty sure the label that OpenBSD and/or Ubuntu left me with

is a Solaris type.

And dumb me, I didn't consider OBP as helping the install too.

So the questions remain if I can install OBP without Solaris, and if I'll have

to crack open the case.

I'll see...
The error from setup is like "newfs failed".

If I run newfs manually, ALMOST no matter I try, it always seems to tell me my

numbers are wrong.

I've tried various -f and -i values. But I did find some that worked, but then

I guess setup

still wants to re-newfs it and gets it wrong. Setup doesn't show the whole

message but

I assume it is using the wrong numbers or something.
Thanks,

Subject: RE: geom network driver times out on sparc 4.2?> Date: Thu, 15 May

2008 01:40:00 +0100> > > > > apt-get upgrade left Linux failing to boot.> >

Solaris 9 install fails very early, something about newfs failing.> > I really

tried a bunch of manual parameters, besides that it> > is run automatically.I

think 10 is in the mail. (I have 10 on> > DVD but so far only a CD drive on

the machine).> > if you mean the installer complains about disk failure then

try this:> http://www.geeklan.co.uk/?p=61> > otherwise I'd highly recommend

you upgrade the OBP!> > > Sevan / Venture37>

_________________________________________________________________> Great deals

on almost anything at eBay.co.uk. Search, bid, find and win on eBay today!>

http://clk.atdmt.com/UKM/go/msnnkmgl0010000004ukm/direct/01/

From the OBP update page:

"Note 1: This utility is *not* OS-dependent.  The list of releases shown under

the

           "Solaris Release" and "SunOS Release" sections may not be complete:

The

           absence of a valid Solaris Release or SunOS  Release from the lists

above

           does not preclude the installation of this patch against the

hardware."
_________________________________________________________________

Be a Hero and Win with Iron Man

http://clk.atdmt.com/UKM/go/msnnkmgl0010000009ukm/direct/01/

Hi,
Sorry, that went only to Sevan.  Sorry Sevan.
You can netboot the OBP upgrade.  I've done it.  What I can't seem to do

is put my hands on how to do this right now.  I seem to remember that the key

is to get the right boot loader from ?solaris? maybe?
Note that this is not without some risk.  You probably want to play around

to make sure that the boot works well before trying the upgrade.
cheers
bruce

So, just as I say this, the page is at:
http://www.SMTPS.net/netboot_flash_obp.html
I did an Ultra 10 this way with no problems.  I may have done

an Ultra2 as well.
cheers
bruce
On Thu, May 15, 2008 at 07:59:48AM +0100, Sevan /

Venture37

well, of course OB can read the file system.It loads the kernel after all.The

instructions are encouraging:

OB> boot disk /flash-update

Cool, like, the flash-update is a kernel?Well, not that, but a program

runnable as if it is a kernel?

But it looks like "the OS", er, the OS installer, is between OB and the

kernel;Specifically there are some "boot blocks" installed by the OS, and the

OpenBSD ones don't recognize the file format of the flash-update.

Darn.

Maybe there is a way?

 - Jay
From: jayk123@hotmail.comTo: openbsd@pckswarms.ch; venture37@hotmail.comCC:

misc@openbsd.orgSubject: RE: geom network driver times out on sparc 4.2?Date:

Fri, 16 May 2008 20:17:26 +0000
Awesome, thanks!Normally I would have said "I have never netbooted; it seems

too hard to setup" but those look like great instructions.And I was almost

right in my paranoia about needing Solaris.I still wonder though -- if

OpenBSD's UFS is the same format as Solaris's, or if OpenBSD can create a

Solaris format of file system, then I think OpenBoot can read the files. And

if not, not. I think.Mitigating factors:  I got 4.3 in the mail that

supposedly fixes this.  The wireless networking is working fine.  The OPB

isn't even known to fix this, but hopefully. But yeah, running old OBP/BIOS

venture37@hotmail.com> CC: jayk123@hotmail.com; misc@openbsd.org> Subject: Re:

geom network driver times out on sparc 4.2?> > So, just as I say this, the

page is at:> > http://www.SMTPS.net/netboot_flash_obp.html> > I did an Ultra

10 this way with no problems. I may have done> an Ultra2 as well.> > cheers> >

bruce> > On Thu, May 15, 2008 at 07:59:48AM +0100, Sevan / > Venture37 >

have to crack open the case.> > > I'll see...> > > > From the OBP update

page:> > "Note 1: This utility is *not* OS-dependent. The list of releases

shown under> > the> > "Solaris Release" and "SunOS Release" sections may not

be complete:> > The> > absenc...
[ message continues ]

Hi,
I was unable to get the obp upgrade to boot when put on a openbsd disk.  I guess,

but don't know for sure, that the sequence is:
- obp starts

- obp reads "something"

- something starts

- something reads the rest of the command line and reads the kernel.
This "something" doesn't read the obp upgrade.
If you compare the upgrade OBP with netboot to the diskless man page, the thing

loaded across the network via tftp when openbsd boots is ofwboot.net.  The thing

loaded with tftp is the actuall upgrade program of obp in that case.
cheers
bruce

Ok! It is done.
I think there might be a reasonable "bug" or "feature request" here to enable

the OpenBSD/sparc64 bootblk to be able to boot the flash updates, like if it

is "just" a matter of supporting ELF32 or something. But I don't know.
Solaris still won't install. It actually got worse, before the OBP update.

Solaris setup had brought up X, now it fails to.

The machine came with Solaris, but when it booted, and went graphical, the LCD

couldn't keep up.. Oh well.
 - Jay
From: jayk123@hotmail.comTo: openbsd@pckswarms.chCC: venture37@hotmail.com;

misc@openbsd.orgSubject: RE: geom network driver times out on sparc 4.2?Date:

Sat, 17 May 2008 18:56:38 +0000
Ok, much progress.I got to the point where it boots the flash update and I

believe I have to fix the jumper now.

Boot the machine whilst holding the STOP & N key on your keyboard,that will

reset your obp to defaults, then hook up a null modem cable to the sun &

another box, run a terminal emulator on the other box, power cycle the sun &

hold STOP & D

this will cause the sun to do a full hardware diag.

_________________________________________________________________
http://clk.atdmt.com/UKM/go/msnnkmgl0010000002ukm/direct/01/

Ok, much progress.

I got to the point where it boots the flash update and I believe I have to fix

the jumper now.
Here are some tricks.If you read the footnote of the instructions, you realize

that RARPD and DHCP are applesand apples. You must pick just one. And it isn't

up to you. It is how the Sun boots.

So, extreme measure:  Take both machines off the main network. No more dhcp,

temporarily.  Run one cable between them. No more wireless, temporarily.  edit

/etc/hostname.if (hostname.gem0 for me) on the rarpd/tftpd server to give it a

static address   I used 10.0.0.1 -- right from the start of "man hostname.if"

edit /etc/hosts as instructed, I usd 10.0.0.2.

I'm not sure how you "really" set up network booting. This can't be it.I know

more modern systems to have dhcp in the boot environment. That should help

completely.

This got me to the point of rarpd sending a reply and then the Sun waiting and

telling meto double check the tftpd server.

Now, I varied a few things flailing around, but I think the main one wasthat

the files in /tftpboot should be named in all caps.

I also killed and restarted inetd, not just -1 (sighup), but that's probably

not needed.

I also ran inetd -d and it reported starting tftpd and then shortly after

reaping it.If I ran tftpd under gdb, it exited with 1 after a short run.I was

considering building it from source and debugging, but I haven't built OpenBSD

yet.I THINK it was the CAPS in the file names, but not sure.

AHA the instructions to use a capital X. I mistyped that.

Now to open the machine and deal with the jumper...

 - Jay

<whining>
  Ugh, this is not so easy.

  First of all, I am able to write the Solaris and OpenBSD bootblocks.    I

could not find any documentation on saving/restoring them, but    I could find

how to set them to a specified set. It's not difficult.   You boot the Solaris

CD and like    /blah/installboot /blah/`uname -i`/blah/bootblk blah

  And when you are done, to get OpenBSD back, boot the openBSD CD and like:

     mount /dev/wd0 /mnt          /mnt/mdec/blah/installboot /mnt/mdec/bootblk

/dev/rwd0

     Actually I got an error so out of paranoia I did more like:

   mount /dev/wd0 /mnt      cp /mnt/mdec/blah/* /tmp     umount /mnt

/tmp/installboot /tmp/bootblk /dev/rwd0

   It took me a little while to find the OpenBSD installboot, buried in "mdec"

instead of any of /usr/bin, /usr/sbin, /bin, /sbin..even thought to check

/stand.     (damn there are too many of these directories! I know people like

to fragment up

their hard drives into multiple partitions in order to make it harder to

decide how large

to make the partitions, and so then there is /bin and /usr/bin,

but must we have sbin too?, and on a single partition system, can't they all

just

be in /bin and /usr/bin a symlink to /bin, and on a multi partition system,

put them

where they are "needed" and then fill the others with symlinks? I realize

that's wasteful

of storage and $path search...I know these are not great suggestions, but

I do often wish it was all just in /bin.)

   find is not present in the shell when you boot the OpenBSD CD, and the one

in /mnt/blah crashes.

   All that, and the Solaris boot blocks won't boot the flash updater either.

They say something like "file just loaded does not appear to be an executable"

or somesuch. This is surprising to me. I really thought this would work.

 Ok, so let's try the net boot approach.

 Well, there's a step "edit /etc/hosts in the usual way". The usual way? I

always use dhcp. The usual way is not at all. So I tried my usual way..

 At first I forgot to switch the Su...
[ message continues ]

Agreed. "something" is "boot blocks" and they are "installed" by "the OS".

The flash-update is a 32bit ELF file and I imagine the OpenBSD/sparc64 boot

blocks only like 64bit ELF.
(Per my other unrelated question -- I was wrong, OpenBSD/sparc64 is pure 64

bit, gcc -m32 doesn't "work" (from a certain point of view, yes I realize it

does exactly what it is meant to do, and it is arguably superior this way,

rather than open a can of worms as to just what is "the architecture" of "the

OS", some hard to pin down hybrid, or simply only SPARC64.)
It is probably possible and not difficult to temporarily install the Solaris

boot blocks (such as from the environment booting the Solaris install CD gives

you), boot the flash-update, and then put back the OpenBSD boot blocks. I

haven't really tried yet.
It might even be possible, like, to say boot cdrom /blahblah/ or boot floppy

/blahblahblah where /blahblahblah is, you know, normally just like bsd or

/update-flash, the kernel or the program to run, relative to the device, but

maybe you can use a "device path" there at the start and have the boot blocks

on one device read the "kernel" (or rather update-flash) from another device.
The flash-update is also 1.4something meg in size, which I thought therefore

might fit on a floppy and be bootable completely from there, but I didn't have

luck with that. The floppy drive wasn't working from OpenBSD and the floppy I

produced on NT doesn't work. The size is maybe just a coincidence, and heck

maybe I misread the number of digits, it was 14<something>. I was too lazy to

determine the actual value of 1.44meg -- LAZY of me, so easy to have done...
I'll experiment later. I blew away my Linux/macppc and started OpenBSD/macppc

install so I can try the netboot (which is something I want to try anyway). I

know those directions aren't specific to macppc, or even OpenBSD, but I have

no other OpenBSD machines currently, the Mac was a good candidate, and I might

as well not risk Linux or MacOSX varying in...
[ message continues ]

I have a Netra T1 that unexpectedly fell off the net. I think it was

alive but without working network. It was power cycled before I had a

chance to look at it. After the event I found this:
Apr 16 16:41:11 sun /bsd: gem0: device timeout

Apr 16 16:41:44 sun last message repeated 3 times

Apr 16 16:43:45 sun last message repeated 13 times

Apr 16 16:53:48 sun last message repeated 111 times

Apr 16 16:59:09 sun syslogd: start

Apr 16 16:59:09 sun /bsd: console is /pci@1f,0/pci@1,1/isa@7/serial@0,3f8
I am still waiting for it to do it again. If I had access to it I'd

update OBP not pontificate about boot loaders and file systems.
OpenBSD 4.3-current (GENERIC) #0: Fri Mar 28 19:16:20 GMT 2008

    pedro@sun.dns:/usr/src/sys/arch/sparc64/compile/GENERIC

real mem = 1073741824 (1024MB)

avail mem = 1027112960 (979MB)

mainbus0 at root: Netra T1 200 (UltraSPARC-IIe 500MHz)

cpu0 at mainbus0: SUNW,UltraSPARC-IIe (rev 1.4) @ 500 MHz

cpu0: physical 16K instruction (32 b/l), 16K data (32 b/l), 256K external (64 b/l)

psycho0 at mainbus0: SUNW,sabre, impl 0, version 0, ign 7c0

psycho0: bus range 0-2, PCI bus 0

psycho0: dvma map c0000000-dfffffff, iotdb 1278000-12f8000

pci0 at psycho0

ppb0 at pci0 dev 1 function 1 "Sun Simba PCI-PCI" rev 0x13

pci1 at ppb0 bus 1

ebus0 at pci1 dev 12 function 0 "Sun RIO EBus" rev 0x01

"flashprom" at ebus0 addr 0-fffff not configured

clock1 at ebus0 addr 0-1fff: mk48t59

"SUNW,lomh" at ebus0 addr 200000-200003 ivec 0x2a not configured

alipm0 at pci1 dev 3 function 0 "Acer Labs M7101 Power" rev 0x00: 74KHz clock

iic0 at alipm0

"max1617" at alipm0 addr 0x18 skipped due to alipm0 bugs

spdmem0 at iic0 addr 0x54: 256MB SDRAM registered ECC PC133CL2

spdmem1 at iic0 addr 0x55: 256MB SDRAM registered ECC PC133CL2

spdmem2 at iic0 addr 0x56: 512MB SDRAM registered ECC PC133CL2

ebus1 at pci1 dev 7 function 0 "Acer Labs M1533 ISA" rev 0x00

power0 at ebus1 addr 2000-2007 ivec 0x25

com0 at ebus1 addr 3f8-3ff ivec 0x2b: ns16550a, 16 byte fifo

com0: console

com1 at ebus1 addr 2...
[ message continues ]

Awesome, thanks!

Normally I would have said "I have never netbooted; it seems too hard to

setup" but those look like great instructions.

And I was almost right in my paranoia about needing Solaris.

I still wonder though -- if OpenBSD's UFS is the same format as Solaris's, or

if OpenBSD can create a Solaris format of file system, then I think OpenBoot

can read the files. And if not, not. I think.

Mitigating factors:

  I got 4.3 in the mail that supposedly fixes this.

  The wireless networking is working fine.

  The OPB isn't even known to fix this, but hopefully.
But yeah, running old OBP/BIOS not great.
venture37@hotmail.com> CC: jayk123@hotmail.com; misc@openbsd.org> Subject: Re:

geom network driver times out on sparc 4.2?> > So, just as I say this, the

page is at:> > http://www.SMTPS.net/netboot_flash_obp.html> > I did an Ultra

10 this way with no problems. I may have done> an Ultra2 as well.> > cheers> >

bruce> > On Thu, May 15, 2008 at 07:59:48AM +0100, Sevan / > Venture37 >

have to crack open the case.> > > I'll see...> > > > From the OBP update

page:> > "Note 1: This utility is *not* OS-dependent. The list of releases

shown under> > the> > "Solaris Release" and "SunOS Release" sections may not

be complete:> > The> > absence of a valid Solaris Release or SunOS Release

from the lists> > above> > does not preclude the installation of this patch

against the> > hardware."> > > > > > > >

_________________________________________________________________> > Be a Hero

and Win with Iron Man> >

http://clk.atdmt.com/UKM/go/msnnkmgl0010000009ukm/direct/01/> >

I have no idea. OBP = "on board PROM"?
I'll research this, but not right away.
Anecdotally, Linux is running well, been up for a few days, with the same sort

of scenarios I was attempting in OpenBSD, just  a bunch of

ssh/scp/cvs/gcc/ld.
Is it possible OpenBSD is sensitive to having the latest OBP and Linux is

not?

   Yes.
Is it possible for OpenBSD to lose this sensitivity and work as well as Linux,

in this regard?

  (Far be it for me to suggest that OpenBSD does not work better than Linux in

general. I do appreciate the OpenBSD attitude, like the web page with reasons

to build your own kernel, every single reason starts with the same

discouragement. :) )

In this case, Linux == Ubuntu 7.10.
Thanks,

Subject: RE: geom network driver times out on sparc 4.2?> Date: Sun, 11 May

2008 17:54:25 +0100> > > Are you running the latest version of OBP on the

system?> _________________________________________________________________> >

All new Live Search at Live.com> >

http://clk.atdmt.com/UKM/go/msnnkmgl0010000006ukm/direct/01/

> I have no idea. OBP = "on board PROM"?
nearly

OpenBoot Prom

http://sunsolve.sun.com/search/advsearch.do?collection=PATCH&type=collec...

&queryKey5=119235&toDocument=yes

_________________________________________________________________
Discover and Win with Live Search
http://clk.atdmt.com/UKM/go/msnnkmgl0010000007ukm/direct/01/