XpPro-32 guest on 12.2-64 installation gets a blue Xp stop screen!

English Virtualization
21

Thank you lwfinger!

I agree with most everything you wrote! In fact I think I have to start from scratch with a new install, if/when I get this memory problem fixed. The vb stuff I mentioned in my previous post was done prior to the memory problem discovery.

Yes the test that fails is the best test to confirm or condemn the fix.

Believe me I’m happy to have the openSuse memory test on the DVD!

I wasn’t asking to find a test that wouldn’t fail, I was asking about a memory diagnostic test that can isolate a memory problem to a module, etc. Heboland.

22

On 03/06/2013 12:16 PM, heboland wrote:This wasn’t the way when I was in the business.

I too was in the business for a lot of years and never heard anything like that. I would be changing vendors if I’d heard it.

My opinion too.

I wasn’t asking to find a test that wouldn’t fail, I was asking about a memory diagnostic test that can isolate a memory problem to a module, etc. Heboland.

I’d test one module at a time until you find one that fails, and I’d test them all individually no matter what, 'cause they may all fail in which case they may or may not be defective modules. It could also be that the particular chips are not compatible with the motherboard in which case you’ll want to exchange all the modules for the right ones. Some motherboards very fussy about wanting exactly the right speed and timing from the SDRAM. (that’s not a bad thing). You can check with the motherboard maker, Asus in this instance, for the exact specs for your motherboard. Interestingly enough I used to see that issue most often with Asus and Intel boards.

When you buy RAM it’s not the maker of the module you have to worry about. It’s the chip maker from whom they bought the chips that are used to build the module that can be the problem. Some chip makers are very good about adhering to specs and others can be very careless.

The “brand name” modules generally use pretty good chips, but the generic modules are often assembled with what ever chips are cheap that week. The generic modules that are made one week may not feature the same chips that were used on the “same” module the week before.

23

On Wed, 06 Mar 2013 21:46:03 +0000, heboland wrote:

> I wasn’t asking to find a test that wouldn’t fail, I was asking about a
> memory diagnostic test that can isolate a memory problem to a module,
> etc.

Easiest way to do this is remove the modules from the system, then
install the minimum number required (1 or 2, depending on whether the
system requires paired modules or not) and run the test. Repeat the test
as necessary, putting new modules in.

Your motherboard manufacturer may have documentation as well to tell you
which slots are used for which addresses, but I find it usually is easier
to just narrow the problem down by physically inserting the modules you
want to test.

Jim

–
Jim Henderson
openSUSE Forums Administrator
Forum Use Terms & Conditions at http://tinyurl.com/openSUSE-T-C

24

Thanks for the responses, guys! This hasn’t been much of a day!

Jim, I would love to try new memory modules in my mobo, but I’m 60 miles from the nearest store and they wouldn’t have anything that would be approved for my mobo.

In my mobo manual, I don’t see any reference as to what slots relate to what addresses. Actually I found the 12.2-64 DVD memory test is pretty easy to debug with.

Robin, I did try your approach switching modules, but the results as I unterpret them suggest two conclusions. Let me first share my conclusions. If you are still with me afterward, i’ll give you my preliminary observations.

Either the memory test 7 is invalid for my hw, or the memory modules are not compatible with my mobo.

While I type this with my 12.1 Pentium3 backup system, my new pride and joy is crunching out the 12.2-64 DVD memory test 8. There’s no errors yet, like tests 1-6. I think there are only 9 or 10 tests total in this suite, but maybe there’s more.

Test 7 fails solidly from the first pass every four addresses starting with 0x00008100000 then 4 then 8, etc. The “good data” is random like 0xf5ab9d7fa. The “bad data” for this pattern is f-f. Other addresses and patterns give 0-0.

The 0s Vs fs may have to do with swapping modules, one module Vs two. Either of the modules fail the same way in the same memory slot and both also fail the same addresses when both are plugged in together.

It’s like test 7 fell off an address cliff! I would dearly love to know that this test passes on similar hw!

Of course, I don’t know what test is on my DVD, so even if someone was willing to verify this for me, what could I tell them to run?

If you’re sill with me, the second point is unqualified mobo modules. My barebones was a kind of kit that the seller assembled and did a smoke test on. They have been easy to work with so far, but that may be coming to an end!

The memory modules that came assembled into the mobo are 8GB Super-Talent DDR3 -1600 Micron W1600UB8GM 8GI512x8.

The mobo qualified list shows 6GB Super-Talent DDR3 -1600 Micron WA160UX6G9 6GB 3x2GB. The date on the mobo manual is 9-12.

So when I call my seller tomorrow he’s going to tell me his part is qualified now and test 7 is broke.

What are the chances that my mobo is addressing the 2GB of memory that the qualified memory wouldn’t have when test 7 runs? And farther that test 7 is the only test in the suite that addresses this area?

To me the failure is catastrophic whether the test or the hw.

It seems I have a hw problem because of the virtualbox behavior. The computer thinks it has 16GB of memory. Any more suggestions? Heboland.

25

Try the memory test from 12.1:
https://forums.opensuse.org/english/get-technical-help-here/hardware/483911-memtest86-bad-memory-bad-controller.html#post2532581

26

Your module: 8GB Super-Talent DDR3 -1600 Micron W1600UB8GM 8GI512x8
The “approved module”: 6GB Super-Talent DDR3 -1600 Micron WA160UX6G9 6GB 3x2GB

Remember what I wrote about the chips on the module? The chips on these
modules are not the same at all. Notice the last part of the descriptions.
That’s where the chip layout is described… 8GI512x8 vs. 3x2GB, different
layout types, using different chips with unknown characteristics.

Unfortunately neither the maker of the chips, Micron nor the maker of the
modules, Super-Talent give any relavent info on those chips on their website,
so I go to the Asus site where I find next to no useful information other than
the simple statement that the ram should be unbuffered, non-ECC.

The bottom line, If you’re unable to get to a store, buy some other rame on
line, from a vendor that has a good return policy just in case. If it works
better and your vendor won’t take it back then sell it on eBay.

If it were me I’d follow this link.

27

Thanks again guys for the responses!

Sauerland I like your idea, but there’s a caveat. my 12.1 is 32bits. What are the chances that different DVD version have different versions of the test?

The 12.1-32bit installation DVD memory test does pass on the 64-box. The test does recognize that cpu the same way as the 12-2-64 does for whatever that means (Intel Core Gen2 3410 MHz).

I have one more memory test experiment to run. That will be to see if the 12.2-64 memory test will run on my P3.

caprus thanks for all your searching on my behalf! I think you are correct that I should get some other memory for my 64 bit box. First I’ll try to get the barebones seller to exchange what I have. Failing that, I should order some online. Kingston definitely has the most qualified parts for my mobo.

Yes the layout types of the two Super-Talent modules are different, but one is a 6GB part Vs and 8GB part. Of course the 6GB part is qualified while the 8GB part is what I have.

I think it’s time to call my BB seller! Heboland.

28

On Thu, 07 Mar 2013 06:56:04 +0000, heboland wrote:

> Jim, I would love to try new memory modules in my mobo, but I’m 60 miles
> from the nearest store and they wouldn’t have anything that would be
> approved for my mobo.

That’s not what I was suggesting.

Let’s say you’ve got 4 memory slots, and you can install one memory
module at a time.

Remove all the memory from the system. Install one module in slot 1.
Test it. Remove it, add the second module to slot 1, and test that.
Repeat until you have run out of memory modules.

That’ll tell you which module you have that is bad.

Jim

–
Jim Henderson
openSUSE Forums Administrator
Forum Use Terms & Conditions at http://tinyurl.com/openSUSE-T-C

29

I don’t want to belabor the point, but it’s not just the layout that’s different. The individual chips that make up the layout are different too… not just a little bit different. Apples & Oranges

By way of illustration: Years ago I built a few hundred servers for a customer who used to buy 14 at a time. He insisted that, since the design had been qualified for use with his custom software, all the component part #'s must stay the same. One day I’d built another batch of 14, but when I tested them I found that none of the servers were stable. It turned out the RAM modules I’d used were all the same part #as before, but the maker had switched to a different brand of chip on the boards. Similar, but not the same. I returned them, found the old style module still in inventory at another supplier, bought up 100 of them… problem solved. I retested the servers and all passed.

This sort of problem is quite rare, and I’m not surprised others with a lot of experience have never seen it, but…

30

caprus, thanks again to you and all the responders on this thread.

Yes it’s true on occasion that even “equivalents” aren’t equal!

My BB seller had no problem with taking back my two Super-Talent ram parts on an RMA. I asked that they be replaced with the Kingston pair that’s listed in the latest Intel QVL. Like you experienced on my behalf, Asus links their QVL to Intel’s.

So with the memory in the snailmail, I’m back to my P3! Since it’ll be a couple weeks before my new memory arrives, that may give me time to order a 12.3-64 install DVD.

On another thread I started in the 64bit section of this site prior to this thread, a modulator there posted that 12.2-64 install DVD memory test is defective. He suggested to use 12.1 or 12.3.

I posted back the question as to whether the 12.1-32 install DVD memory was equivalent to the 12.1-64. That moderator may have answered that question there, but if so, I don’t have his answer yet.

So there is a good chance my memory in the mail never failed a valid memory test! One other point on the install DVD memory test. The 12.2-64 won’t run a memory test on 32 bit hw. In advance it detects a 32bit system and offers only a reboot button.

To me this is the end of this thread. I won’t be surprised if virtualbox crashes again in 12.3-64. Maybe Oracle will have something newer than the rpm on the 11.4 repository. Also it would be wonderful to see k3b morphed into something up to date. Heboland.

31

Replying to myself and the list,

Something I would like some feedback on is should I try to reactivate this thread? Instead of XP BSODs, virtualbox is crashing the host OS 12.2-64. Thats not only with an XP CD, but also an 11.4-32 DVD. So the BSOD has changed to host crash, but the problem may well be the same.

I have my box running again. This time the Kingston memory modules straight from the Intel QVL haven’t changed much of anything for me. I ran the 12.1-32 memory overnight with no errors. The hdd got all reformated and the 12.2-64 I reloaded from scratch.

In the spirit of KISS, I only have the Oracle VirtualBox-4.2-4.2.10_84104_openSUSE114-1.x86_64 installed, tho the rpm grep also finds openbox-3.5.0-18.1.2.x86_64 and dosbox-0.74-13.1.2.x86_64. I downloaded the Oracle VirtualBox-4.2-4.2.10 extension, but haven’t installed it yet.

This is a second point on which I would like some feedback! After a careful rereading of the sticky intro, I concluded the extension pack is a feature set to the vm which is optional. Thus to keep things simple I didn’t install it.

The host crash occurs now during the vm format step, as tho the formatting undermined the host. I’ve looked at vbox logfiles and dmesg, but the latter gets rewriten on the crash recovery reboot. vbox log seems unaware of impending doom!

Is there something assumed about the host hdd partitioning and/or format. Except for my /boot partition which is ext2, the other partitions are ext4 and of course swap.

The oracle debug is described as an information collecting option, the output of which only oracle can decipher.

What am I doing wrong? Heboland.

32

Hello,
Here are some suggestions…

First, although I wouldn’t discount the possibility of bad hardware entirely, I’d say you’d have to be very unlucky to actually experience a problem. Rather than bad RAM, I’ve far more often seen mis-matched RAM. Best if possible for all modules to be as same as possible. Depending on hardware, you may also need to modify BIOS timing/speed, etc.

When you have new hardware, particularly CPU, RAM, or motherboard you may want to “burn in.” This means to push the limits of the system for a day or so continuously. The reasons for this are… If parts are going to fail you want that to happen within warranty, if there is a failure you want it to happen when not doing something critical.

So, if you have some patience, don’t install and run VB right away. Do some other stress jobs like transcoding, benchmarking, running zillions of windows and apps doing many things at the same time. While you’re doing this you may also get a better feel whether everything is running smoothly or you’re seeing anomalies which could be warnings and precursors.

Then, after a day or two of heavy, continuous use re-visit VB.

BTW- no Guest VM should ever crash the Host unless the app(VB) is faulty.

HTH,
TSU

33

Thank you tsu2 for the suggestions!

I would doubt the memory module mismatch in my case. There are just two matched 8GB Kingston modules that came in a kit. These modules are on the Intel QVL for the Z77 chipset.

What I was told by the seller is these modules each contain an information chip that tells the BIOS what to set.

Otherwise this HW is very Plain Jane. Wired internet connection from a router, usb wires for mouse, kb, and printer. Display is LCD monitor on vesa cable.

Regarding the “crash”, maybe this behavior is instead a reboot requested by the guest installation. The screen goes black followed by the total system reboot. There’s no fisking of the drive on reboot. Something apparently sent a shutdown signal. There should be a log of this somewhere.

Understand since I’ve never got past this point, I really don’t know what’s supposed to happen. Maybe I can find a tutorial to see the rest of the story.

Like you confirmed, I wouldn’t expect the guest to crash the host.

In the sticky intro posts, registeration is necessary to download the tarball. The rpm installation doesn’t display any legal form requiring terms of acceptance like it does for adobe. My guess is that I’m missing something everyone else thought was too obvious to mention.

Log files should be the way to track this behavior down. Maybe I’ll join an Oracle list and ask there.

Meanwhile I have a 12.3-64 DVD in the mail. I have plans to install it on this box as a dual boot with the 12.2. I wouldn’t expect the 12-3 to behave any different with the Oracle VB. Maybe the OSS vb will work better than the 12.2. If I can get past the kernel module problem with the 12.2, there should be logs to check.

Probably this will be the last post on this thread, but let me ask this question anyway. How do OpenSuse developers log errors to analyze behavior after a crash? This would take some kind of system file that gets appended to rather than overwritten. Thanks again to all the responders! Heboland.

34

Errors seen on the Host will be logged in the syslog.

On 12.3, that’s the journal but may still be piped to the original syslog.

Eg

journalctl -b

tail -n 300 /var/log/messages

The only cause I can think of is that you may not be configuring a virtual disk file for your Guest, if you point to your Host file system you could cause a conflict.

You need to describe in extreme detail the steps you’re taking in creating your Guest. Are you deviating from the default options in any way?

TSU

35

I’m no expert on VirtualBox by any means, but I’m just curious:

Why “Oracle VirtualBox-4.2-4.2.10_84104_openSUSE114-1.x86_64”? Why not 4.1.22-1.10.1 from the openSUSE-12.2-Update repo?

Two reasons for asking:

  1. Seems like a violation of the KISS principle.
  2. I suspect that the version in the openSUSE repo has been selected for, and may have had more testing for compatibility with your specific OS.
36

Thanks caprus, tsu2! I’m encouraged that I’m still getting help on this thread.

caprus, regarding the vbox version, I have the Oracle VirtualBox openSUSE added at Index of /virtualbox/rpm/opensuse/11.4/. This is still with 12.2-64. When I tell the yast SWmanager to search for virtualbox, it finds the version I have and two earler versions. The sticky instructions are to use the latest version. Finally, to use the extension version that matches the vb version.

Maybe going with the latest release complicates the picture. OTOH upgrading to the latest available version is often a prerequisite to getting help.

tsu2 thanks for the system error information. That will be a very useful place to get a hint at the problem. What’s in the log now is hard to sort out, but I may post back a sample in a code block after some more analysis.

Regarding the vb setup, do I have control of where I point the guest virtual storage? Do I need to leave an unformatted partition? I just expect vb to use space in ~ or create a file system in /usr.

You both should know that I HAVE NOT installed the Oracle extension pack. I’ll be happy to run after I can walk. Taking all the defaults causes a crash, and it is a real crash. The reboot following does fsk the ext2 boot partition.

Yes, I can keep detailed notes regarding the vm creation steps.

What would be the vote between you two regarding the version? It’s an easy get to drop back from 4.2 to 4.1. tsu2, what version are you running?

On the concurrent topic of a hardware problem, what do you guys think of building a new 12.2 kernel as a stress test for the hardware? I’ve never built a kernel before, but I’ve seen instructions.

My concern with this is getting all the pieces over my slow internet <1Mb/sec. Would there be less downloading if I built a 12.3 kernel instead? There might not be as many changes required on something newrer. Heboland.

37

I’m very, very far from an expert on VirtualBox. My comment was intended only to reflect my experience that, when troubleshooting what is likely to be an incompatibility issue of some sort, it’s helpful to use software from only the standard openSUSE community supported repositories that correspond to your OS version (not a generic 11.4 repo with 12.2, and not the factory ones either). You could try getting the latest available version, currently 4.1.22, from the openSUSE-12.2-Update repo - Index of /update/12.2

On the concurrent topic of a hardware problem, what do you guys think of building a new 12.2 kernel as a stress test for the hardware? I’ve never built a kernel before, but I’ve seen instructions.

My concern with this is getting all the pieces over my slow internet <1Mb/sec. Would there be less downloading if I built a 12.3 kernel instead? There might not be as many changes required on something newer. Heboland.

I appreciate your frustration with that <1Mbs speed, however I remember downloading Linux OSs on a 44K dialup connection. It can be done, and I think the advantage of having a stock install of 12.3 to explore the possibility of incompatibility between your 12.2 and the bleeding edge features of your mobo would make it worthwhile to set up the download and just let it run. I’d use a torrent, particularly if your connection is unstable. It won’t take as long as you think. (Of course that will mean getting VirtualBox from the repo for 12.3 - Index of /distribution/12.3/repo/oss/suse/x86_64 (Note: you’ll find 4.2.6 there)

Note: On our little island we just got the ability to have a WiMax connection running at 10Mbs +/-. Until six weeks ago we thought our 750Kbs DSL with <75% uptime was the best we could hope for.

38

Thanks caprus!

You may still have a point about the vb version. The sticky post instructions are very specific about not mixing SW.

The sticky recommendations seem to be for Oracle, hence the Oracle repro. I think you are referring to the OSS version of virtualbox.

If you are referring to OSS vb, then I have some exposure with it. After gutting all the Oracle and disabling that repository, I did install the OSS vb. It uses a kernel module with the same name as Oracle (vbox.drv).

The OSS vb install did add its kernel module, but the vb kept complaining about the kernel module missing. Almost nowhere in the sticky posts is there any information about the OSS version. Poking around since, there is an OSE website that does have a vb forum. In this virtulation section, the OSS vb is not a darling!

Monday I should have my 12.3 install DVD. My thought was that to build a new 12.3 kernel for a release so new, I should be able to use many files off the DVD.

Another thought for “make work” to stress HW, building fonts can soak up some cpu cycles. As I recall the fonts on the DVD are already “built” but maybe I could rebuild some.

Back to bad HW, I’m not sure if my ACPI errors in dmesg were mentioned. In another thread a moderator with HW similar to mine gets those errors also and doesn’t worry about them. The look like this:
0.997633] ACPI Error: [DSSP] Namespace lookup failure, AE_NOT_FOUND (20120320/psargs-359)

The info I got about these errors is that they are DSDT table bugs. They are caused by a mismatch between the BIOS and the HW. My BIOS is up to date.

I did find in the vm details, they use a dsdt also. Heboland.

39

I may be mistaken, as the documentation I’ve seen is confusing at best, but I understood what I read to mean that as of version 4 there’s no longer any difference in the “core package”, however the extensions are still "released under a proprietary Personal Use and Evaluation License (PUEL) * "

If I’m wrong then I suspect that someone better informed than I will correct me.

40

IMO
Possible causes of your problems in ascending likelihood (last one most likely)

Hardware issues. But perhaps most important. Although least likely, HW problems will cause other things to not work.

OS problems(including installed libraries). Make sure a new install is rock solid before installing apps. Run apps you’re familiar with to know and recognise anomalies.

App problems, in this case Virtual Box. If you’re not familiar with the app, install and configure using the simplest and most automated way possible. On openSUSE, this means installing from the recognised repos and RPM packages, <not> from the original vendor, not from source and not following some cookbook of instuctions found on the Internet. These other options should be tried only after installing openSUSE repo RPMs using YAST or zypper.

Lastly, and where I recommend you start troubleshooting is whatever you’re doing when creating your Guest. In fact, I’d ask you to find the configuration file for your latest attempt to build a Guest and post that file in its entirety (within CODE tags).

TSU