UBCD 5.1 kernel panic running Prime Test

[bob3248]

When I try to run the Mersenne Prime Test v24.14 in UBCD 5.1 I get an error message that says:
[1.168034] kernel panic - not syncing:VFS:Unable to mount root fs on unknown block (1,0)

I receive this error on 2 different computers and if I run UBCD in a VirtualBox virtual machine. I tried changing the SATA configuration in the BIOS to both AHCI and IDE and it made no difference. UBCD 5.03 works fine.

[Icecube]

As a workaround, boot UBCD and press TAB when you are on the Mersenne Prime Test v24.14 entry and change the kernel file from:

Code: Select all

./linux /pmagic/pmagic/bzImage

to:

Code: Select all

.linux /ubcd/boot/cpustress/bzImage

[ady]

I think that UBCD should include both possible menu entries, the "original", and the one using PMagic's bzImage.

For some systems, the PMagic entry would work, and for other (probably older) systems, the "original" entry would work.

In any case, the tests are still the same, so for newer systems, the resulting tests will not actually test the complete set (a newer CPU would have additional "functions" that are not being tested).

Ideally, Mersenne could be updated ( nowadays, to version 26 ), but currently that means some "Icecube's time" .

[Icecube]

The current Parted Magic kernel included in UBCD5.1 won't work with the CPUstress initrd on any PC. The current Parted Magic kernel doesn't support mounting the ext2 formatted initrd, but probably only initramfs. The CPUstress initrd should move to the initramfs format anyway.

[bob3248]

That workaround solved the problem. Thanks.

[LethalDuck]

Doesn't work for me.
I've tried all the CPU stress test tools, and get the same or very similar error each time.

I tried what you suggested. Now I get:
Lots of log, then the last 3 lines look like this...

VFS: Cannot open root devicd "<NULL>" or unknown-block(8,5)
Please append a correct "root=" boot option; here are the available partitions:
Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(8,5)

With the default I'm getting the following:
where n is a digit.

[ n.nnnnnn] drivers/rtc/hctosys.c: unable to open rtc device (rtc0)
[ n.nnnnnn] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(1,0)
[ n.nnnnnn] Pid: 1, comm: swapper Not tainted 3.0.0-pmagic #1
[ n.nnnnnn] Call Trace:
[ n.nnnnnn] [<nnnnnnnn>] ? printk+0x14/0x1d
[ n.nnnnnn] [<nnnnnnnn>] panic+0x55/0x144
[ n.nnnnnn] [<nnnnnnnn>] mount_block_root+0x1ee/0x20e
[ n.nnnnnn] [<nnnnnnnn>] ? sys_mknod+0x18/0x1a
[ n.nnnnnn] [<nnnnnnnn>] mount_root+0x4c/0x54
[ n.nnnnnn] [<nnnnnnnn>] prepare_namespace+0x140/0x171
[ n.nnnnnn] [<nnnnnnnn>] kernel_init+0x114/0x123
[ n.nnnnnn] [<nnnnnnnn>] ? start_kernel+0x340/0x340
[ n.nnnnnn] [<nnnnnnnn>] kernel_thread_healper+0x6/0x10

Can anyone decifer this?
Thanks.

The reason I'm running tests is because my system was randomly stopping, starting.
It would sometimes stop, then start after a while.
PSU tested non loaded and loaded and well within tolerances.

My system hardware:

ASUS p5kpl/epu Mobo
Celeron 1.8Ghz single core
Corsair 2GB KIT (2x1GB) DDR2 800Mhz DIMM PC6400 - Desktop RAM - TWIN2X2048-6400C4
P/S ZM750-HP
2 x HDD swap trays. 3 SATA 3.5" in 2 5.25" bays
4 x 750 GB WD HDD's
4GB AData thumbdrive.

[LethalDuck]

Bump?

[ady]

@LethalDuck,

The workaround might not work for your CPU, as Icecube says in his post above.

You could use an older version of UBCD, but that also means older CPUs compatibility. Your CPU might be supported in those older versions.

You could also try to use Mersenne's CPU test by itself (not from UBCD). You would need to check their website on how to use their latest tools (check the home page of UBCD for the links).

Other alternative would be to use some hardware test specialized Linux distro.

[Explorer09]

To everyone who has the same problem,

I have tried repacked the cpustress initrd image to a cpio format. And I got a good news that it works!

However I still need some people to test my image to see if it works for everyone.
(http://dl.dropbox.com/u/70170658/ubcd/newinitrd.gz)

I got some messages saying that can't log to /dev/tty5 and I'm not sure whether it was a big deal or not.

When using kernel "/pmagic/pmagic/bzImage" :

Code: Select all

[    2.743183] drivers/rtc/hctosys.c: unable to open rtc devices (rtc0)
init: can't log to /dev/tty5
init started: BusyBox v1.13.2 (2009-02-25 19:25:29 CET)
init: can't log to /dev/tty5
starting pid 682, tty '': '/etc/init.d/rcS'

When using kernel "/ubcd/boot/cpustress/bzImage": (sic)

Code: Select all

IO APIC resources could be not be allocated.
init: can't log to /dev/tty5
init started: BusyBox v1.13.2 (2009-02-25 19:25:29 CET)
init: can't log to /dev/tty5
starting pid 869, tty '': '/etc/init.d/rcS'

Can anyone explain?

[Explorer09]

Well, Icecube did not update the image for a long time. So I made a new version for myself. I think it will be useful if I share it here.

CPUstress test v2.1expl3
(I call my version 2.1expl3 in order to not conflict with Icecube's version. If Icecube and others want this image to be v2.2, tell me and I'll change it.)
(EDIT: These download links has been changed to v2.1expl4. See below post.)
The image:
http://dl.dropbox.com/u/70170658/ubcd/expl4/initrd.gz
The build.tar:
http://dl.dropbox.com/u/70170658/ubcd/e ... ild.tar.xz

----

Updated programs:
- Updated Mersenne Prime test to v27.7 build 2 (from v24.14). Note that v23.9.2 is still available for people who run Pentium III or older CPUs.
- Updated Stress to v1.0.4 (from v1.0.0)
- Updated Busybox to v1.20.2 (from v1.13.2)
- Updated ld-linux, libc, libm (part of eglibc) to v2.16 (from v2.8.90)
- Added libdl and libpthread (also part of eglibc) v2.16. These two files are required for mprime27.

Fixes:
- Repacked ubcd/boot/cpustress/initrd.gz to cpio+gzip format. This prevents a kernel panic when booting. (viewtopic.php?f=12&t=6488)
- The version number of Mersenne Prime test 23 is actually 23.9.2. Fixed the documentation bugs in various scripts and help files.
- Add signal traps to terminate cpuburn-in or mprime when user terminates the wrapper scripts via Ctrl+C (SIGINT) or Ctrl+\ (SIGQUIT). (viewtopic.php?f=12&t=14429)
- Minor bug fixes in scripts.

Changes:
- Added "help" as an alias to "helpinfo" command.
- In "menu" the user can now type the command name as the choice rather than number.
- Added a hint "Type 'menu' to return to main menu." in "menu".
- Use dynamically linked executable for mprime23, since the CPUstress image have all the libraries it depends on already.
- Change the path in /lib/* symlinks from absolute to relative paths.

----

[Victor Chew]

Just to let you know, I am in the process of giving UBCD a much-needed update, and I will be using your updated image. Thanks for chipping in where IceCube left off!

BTW, in V5.1.1, all CPUStress' entries were updated to use PMagic's bzImage for better system compatibility. Are you using that as well for yourself?

[Explorer09]

Just to let you know, I am in the process of giving UBCD a much-needed update, and I will be using your updated image. Thanks for chipping in where IceCube left off!

BTW, in V5.1.1, all CPUStress' entries were updated to use PMagic's bzImage for better system compatibility. Are you using that as well for yourself?

First, I'm going to update this image again to correct some minor errors.

Second, yes, my updated image is meant to run with PMagic's bzImage. Because I cannot have the kernel compiled when using IceCube's config file. I have no choice.

EDIT: It's done.

CPUstress test v2.1expl4
http://dl.dropbox.com/u/70170658/ubcd/expl4/initrd.gz
http://dl.dropbox.com/u/70170658/ubcd/e ... ild.tar.xz

Changes:
- I re-downloaded and re-compiled eglibc and Busybox. The new binaries should have some upstream bugs fixed.
- Fixed this bug more completely in cpuburn and mprime: viewtopic.php?f=12&t=14429

[Don Manuel]

If I had found this thread earlier, I would maybe not have seeked a solution on my own.
However I downloaded your version and was very surprised that your binary for Mersenne27 is about 10% in size than what I've downloaded here. Could you explain why?

[Explorer09]

If I had found this thread earlier, I would maybe not have seeked a solution on my own.
However I downloaded your version and was very surprised that your binary for Mersenne27 is about 10% in size than what I've downloaded here. Could you explain why?

It seem that you didn't download my build.tar.xz, and read my "compile notes".

Anyway, here's what I've done:
1. Strip the binary.
2. Compress the binary with UPX.

Yes, the binary of Mersenne27 is significantly larger than the old versions, and that's why I decide to compress it.

[Don Manuel]

Thanks a lot for the info, I have yet downloaded the build-file. Didn't imagine the binary could be compressed, I just compressed the whole filesystem, but with gzip only my compressed fs still is ~20% larger than yours .

[Explorer09]

CPUstress test v2.1expl5
http://dl.dropbox.com/u/70170658/ubcd/c ... .1expl5.7z
[del]sha1sum: cd3f80d5122b32977a45c8d2f27f8a334754c6b2 cpustress-2.1expl5.7z[/del]
[del]md5sum: e7c87f80e5917cb1ab9c170d93cdd14c cpustress-2.1expl5.7z[/del]
md5sum: aa044424f0b6883797dceafc58fa5b1b
sha1sum: 93a309f85c4a1f21dd0ebdad1236cad22b63c573

Changes:
- Updated Mersenne Prime Test (prime95) to v27.9 build 1 (from 27.7 build 2).
- Updated eglibc to 2.17 (from 2.16.1).
- Changed the isolinux entries in various documentation and scripts. CPUstress should now be booted using Parted Magic's kernel image.
- Updated readme.txt and added a changelog (changes.txt) for my own changes. (These can be put into UBCD if Victor wishes to.)

[Don Manuel]

Great! And thanks for distributing the info about upx in your build.gz. Didn't know about this very usefull executable-compression-utility.

[ady]

- Changed the isolinux entries in various documentation and scripts. CPUstress should now be booted using Parted Magic's kernel image.

@Explorer09,

Thank you for your work on this.

Just a little detail...

For the inclusion of this set of programs in UBCD using pmagic's kernel, it would probably be better (or even necessary) to use

Code: Select all

COM32 /boot/syslinux/linux.c32 /pmagic/bzImage

instead of using "

kernel /pmagic/bzImage

" in the Syslinux boot entry.

Also, adding a help text as DISPLAY file would be recommended (let me know if I can help).

[Explorer09]

@Explorer09,
For the inclusion of this set of programs in UBCD using pmagic's kernel, it would probably be better (or even necessary) to use

Code: Select all

COM32 /boot/syslinux/linux.c32 /pmagic/bzImage

instead of using "

kernel /pmagic/bzImage

" in the Syslinux boot entry.

I don't understand, why is this change necessary?
What's the difference between the two?

Also, adding a help text as DISPLAY file would be recommended (let me know if I can help).

What do you mean as a display file?

[ady]

@Explorer09,
For the inclusion of this set of programs in UBCD using pmagic's kernel, it would probably be better (or even necessary) to use

Code: Select all

COM32 /boot/syslinux/linux.c32 /pmagic/bzImage

instead of using "

kernel /pmagic/bzImage

" in the Syslinux boot entry.

I don't understand, why is this change necessary?
What's the difference between the two?

1_ The key word "KERNEL" in Syslinux cfg files is more generic. Its behavior depends on the file extension of the specific kernel. For example, to boot linux kernels that use the "*.bin" file extension, the "KERNEL" key word should NOT be used. Use "LINUX" instead.

2_ The second difference is more relevant to this particular case. In Syslinux 4.06 (and older), linux.c32 is capable of booting newer (bigger) Linux kernels (like the one used in PMagic). The "traditional" directives ("KERNEL" or "LINUX") might fail in these cases. Note that the linux.c32 module included in Syslinux 4.06 (used by UBCD 5.2 pre-releases), although capable of booting newer Linux kernels, is still under development, so there are cases where it might fail. Having said that, I strongly suggest using it for pmagic's kernels, since the original PMagic distro already uses it (and has been using it for a long time).

Also, adding a help text as DISPLAY file would be recommended (let me know if I can help).

What do you mean as a display file?

In Syslinux, DISPLAY files are those "help text" files that can be displayed before the boot prompt, or by pressing F1-F12 keys (among other possible options).

Instead of (or in addition to) using different boot entries in the main boot menu, there could be a help screen (help text in a DISPLAY file) where the user gets all the available options for each stress-like test, including explanations for each "boot cheat code".

UBCD has some examples of DISPLAY files. "F1" and "F2" for pmagic, when booted from UBCD's menu. PMagic itself has them too, with its own content, of course. DBAN and memtest86 4.1.0 (included in UBCD 5.2b3) have their respective DISPLAY files. There are several alternative ways to show their contents, and the aforementioned examples use different variations of them.

I guess I should first wait for Victor to include the new set of stress-like tools and then I may try to come up with some possible DISPLAY files for them, if there is any real interest / utility for UBCD.