Question #155035 “Bogus MCE errors with 11.04 install” : Questions : linux package : Ubuntu

Revision history for this message

Tomasz Kusmierz (wally-tm) said on 2011-04-30:

#1

BTW,
pretty much same stuff happens on server kernel ...

Revision history for this message

Seanyoung247 (seanyoung247) said on 2011-04-30:

#2

I get the exact same problem with similar hardware, ASUS Z7S WS, 2.33Ghz Quad Xeon, 8Gb FBDimms (4x2gb) and ATI 5850. This also occurs running the Fedora 15 Beta Live CD, that uses the same Kernel as Ubuntu 11.04. I've tried booting with nomce and nomodeset parameters set with no change.

Revision history for this message

Tomasz Kusmierz (wally-tm) said on 2011-05-01:

#3

damn, so shall we take this to kernel mail list ? and also what kernel fetora (sic) 15 is using on beta 15 live cd ? (mainly concerned for 2.6.38-xx bit)

BTW 2,
I've tested bios 0401 and 0501 - both with exactly same problem !

So, is there a chance to get new ubuntu flawored kernel from 2.6.38 range in any near future ?

Revision history for this message

Eliah Kagan (degeneracypressure) said on 2011-05-02:

#4

@Tomasz Kusmierz
I recommend you report this as a bug in linux in Ubuntu. To do that, first read https://help.ubuntu.com/community/ReportingBugs carefully. Then search to see if the bug has already been reported (among many other things, that page explains how.) Then report the bug by running "ubuntu-bug linux". In your bug report, include all the information about the bug that you've posted in this question, especially your explanation of why there is very good reason to believe that these machine check exceptions are spurious. For your bug report, please find the [Hardware Error] lines in the file /var/log/dmesg, and post them in your bug report. You should also attach the file /var/log/dmesg, if Apport does not attach it automatically. If you are unable to find those lines in /var/log/dmesg, then either they were't written, or the last crash occurred long enough ago that they would be the rotated dmesg logs (/var/log/dmesg.0, /var/log/dmesg.1.gz, /var/log/dmesg.2.gz, and so forth). You only have to check the contents of /var/log/dmesg.0 (or others) if you cannot find this in /var/log/dmesg and the file modification times on the rotated log files are such that they could have been current enough to have covered the most recent instance of the crash. (It's a reasonably good bet that if the crash occurred once and wasn't logged, then on your system it is not ever logged when it occurs.) If this was not recorded in any log files, then you should go ahead and include the copy of the error message lines from your photo, and in addition, you should attach the photo as an image file (unless you took it with an analog camera and have no way of scanning or otherwise digitizing it) to the bug report.

After you file your bug report, you can link this question and the bug to each other using the "Link existing bug" link on this question page. While it is a good idea to do this, your bug report should be complete and self-contained--it should not be necessary for the Ubuntu developers to refer to this question in order to completely understand and appreciate your bug report.

Even though this problem is unlikely to specific to Ubuntu (in part because Seanyoung247 has also experienced it with Fedora), it is still very appropriate to file a Launchpad bug about this against linux in Ubuntu. At some point soon, it would also be good to file a bug at https://bugzilla.kernel.org (whoever does that may want to read through https://bugzilla.kernel.org/docs/en/html/using.html first and should certainly at least take a look at it). If you're reporting this against linux in Ubuntu, then by way of division of labor, it would perhaps be reasonable for Seanyoung247, who has experienced it on multiple distros, to file the upstream bug. But that's just a suggestion. Similarly, either of you could file the downstream bug here on Launchpad, too. (Or you both could, and then one could get marked a duplicate of the other.) Once the upstream bug is filed (or found already to have been filed, by a search of the bugs at https://bugzilla.kernel.org), the downstream bug should be linked to it using the "Also affects project" link on the downstream (Launchpad) bug page.

@Seanyoung247
Do you also have good reason to believe that the machine check exceptions aren't real, on your system? For example, do you run another operating system for a significant number of hours each day and *not* get machine check exceptions, like Tomasz Kusmierz does? If you run a Linux-based OS on the machine, what kernel does it use?

@Tomasz Kusmierz
I recommend you report this as a bug in linux in Ubuntu. To do that, first read https://help.ubuntu.com/community/ReportingBugs carefully. Then search to see if the bug has already been reported (among many other things, that page explains how.) Then report the bug by running "ubuntu-bug linux". In your bug report, include all the information about the bug that you've posted in this question, especially your explanation of why there is very good reason to believe that these machine check exceptions are spurious. For your bug report, please find the [Hardware Error] lines in the file /var/log/dmesg, and post them in your bug report. You should also attach the file /var/log/dmesg, if Apport does not attach it automatically. If you are unable to find those lines in /var/log/dmesg, then either they were't written, or the last crash occurred long enough ago that they would be the rotated dmesg logs (/var/log/dmesg.0, /var/log/dmesg.1.gz, /var/log/dmesg.2.gz, and so forth). You only have to check the contents of /var/log/dmesg.0 (or others) if you cannot find this in /var/log/dmesg and the file modification times on the rotated log files are such that they could have been current enough to have covered the most recent instance of the crash. (It's a reasonably good bet that if the crash occurred once and wasn't logged, then on your system it is not ever logged when it occurs.) If this was not recorded in any log files, then you should go ahead and include the copy of the error message lines from your photo, and in addition, you should attach the photo as an image file (unless you took it with an analog camera and have no way of scanning or otherwise digitizing it) to the bug report.

After you file your bug report, you can link this question and the bug to each other using the "Link existing bug" link on this question page. While it is a good idea to do this, your bug report should be complete and self-contained--it should not be necessary for the Ubuntu developers to refer to this question in order to completely understand and appreciate your bug report.

Even though this problem is unlikely to specific to Ubuntu (in part because Seanyoung247 has also experienced it with Fedora), it is still very appropriate to file a Launchpad bug about this against linux in Ubuntu. At some point soon, it would also be good to file a bug at https://bugzilla.kernel.org (whoever does that may want to read through https://bugzilla.kernel.org/docs/en/html/using.html first and should certainly at least take a look at it). If you're reporting this against linux in Ubuntu, then by way of division of labor, it would perhaps be reasonable for Seanyoung247, who has experienced it on multiple distros, to file the upstream bug. But that's just a suggestion. Similarly, either of you could file the downstream bug here on Launchpad, too. (Or you both could, and then one could get marked a duplicate of the other.) Once the upstream bug is filed (or found already to have been filed, by a search of the bugs at https://bugzilla.kernel.org), the downstream bug should be linked to it using the "Also affects project" link on the downstream (Launchpad) bug page.

@Seanyoung247
Do you also have good reason to believe that the machine check exceptions aren't real, on your system? For example, do you run another operating system for a significant number of hours each day and *not* get machine check exceptions, like Tomasz Kusmierz does? If you run a Linux-based OS on the machine, what kernel does it use?

Revision history for this message

Eliah Kagan (degeneracypressure) said on 2011-05-02:

#5

"So, is there a chance to get new ubuntu flawored kernel from 2.6.38 range in any near future ?"

I don't know which 2.3.38 kernel you're running, but there is a new version in natty-proposed. The version in main is 2.6.38-8.42. The version in proposed is 2.6.38-9.43. The changelog (you can read it at https://launchpad.net/ubuntu/+source/linux) doesn't say anything suggesting that a bug like this would have been (intentionally) fixed, and I doubt that will fix this problem, but you could try it and find out (see https://wiki.ubuntu.com/Testing/EnableProposed). If you try the version in proposed, that could be either before or after you report your bug, but if you're going to try it and you haven't reported the bug yet, I suppose it may as well be before. If you're running the version from proposed when you report the bug, you should make sure to mention in your bug report that it also occurs in the version in main (and give the full version number of the -main kernel for clarity, which will probably be 2.6.38-8.42), so as to avoid creating the impression that this bug might have originated in kernel 2.6.38-9.43

Revision history for this message

Seanyoung247 (seanyoung247) said on 2011-05-02:

#6

@Tomasz Kusmierz

The Kernel in use by Fedora 15 Beta is 2.6.38.2-9. It may be worth mentioning that I've tried the 32 bit version of the Natty liveCD as well, and it also crashes, though before completing boot and without giving any Kernel messages before rebooting.

@Eliah Kagan
I believe these messages are spurious also, yes. I have been running Windows Vista as a dual boot for over a year on this system with no problems. I have been, and I am currently back using, Ubuntu 10.10 with the 2.6.35-28-generic Kernel and it's never had any problems.

Revision history for this message

Tomasz Kusmierz (wally-tm) said on 2011-05-03:

#7

Dear
@Eliah Kagan
- cheers for advice on https://help.ubuntu.com/community/ReportingBugs
- ubuntu-bug linux tells me that I can NOT report this since this is "not original ubuntu package" (when running on 2.6.35) and when running on 2.6.38 it does not manage to get to sending process because kernel crashes.
- /var/log/dmesg does not shows any messages [Hardware Error] same as any other log file && / || rotated log file - anyway it would be VERY suspicious if kernel would log anything after biting the bullet of 'processor context corrupt' (writing corrupted data to fs ?). Also, have you read through the line stating "There seems to be nothing in any log files about this crash..." in first post?
- it's fairly hard to provide dmesg from crashed installation process.
- rather than throwing some lame photos from my mobile, I've preferred to put some effort and at least copy vital text in so anyone searching for this type of bug will find what he/she look for.

Kind regards for guidance & support.

Anyway, @Seanyoung247
cheers for info, I've tried 32 bit install as well because I was keen to believe that 64bit build might got contaminated. Also I've tried 2.6.38-9 from proposed repo (as proposed by Eliah Kagan) with no luck. For time being I'll run 11.04 with 2.6.35 kernel ... when I'll get some time I'll post this bug on kernel mail list or bugzila (whatever). If so are you ok to provide your setup as additional reference ?

Btw, I'm not trying to blame any kernel dev's for having this bug for that long in main stream - after all this mobo in not really a common hardware.

Cheers, Tom.

Revision history for this message

Seanyoung247 (seanyoung247) said on 2011-05-03:

#8

@Tomasz Kusmierz
Yeah that's perfectly fine, let me know if you need any more information from me.

Revision history for this message

Tomasz Kusmierz (wally-tm) said on 2011-05-03:

#9

Btw, in first post, last line:

kernel is 2.6.28-8-generic #42-Ubuntu

should be:

kernel is 2.6.38-8-generic #42-Ubuntu

my bad, sorry.

Revision history for this message

Eliah Kagan (degeneracypressure) said on 2011-05-03:

#10

@Tomasz Kusmierz
"- ubuntu-bug linux tells me that I can NOT report this since this is "not original ubuntu package" (when running on 2.6.35) and when running on 2.6.38 it does not manage to get to sending process because kernel crashes."

If you run a custom kernel--i.e., one you build yourself, with any changes of any kind in the configuration--then you cannot report a bug in it against linux in Ubuntu. (Actually, you can, but you shouldn't, and Apport won't facilitate it easily, and if you do, your bug will be summarily marked Invalid.) To be reported downstream, bugs in a custom kernel must be reproduced in a vanilla (i.e., non-customized) kernel. You can still report bugs affecting a custom kernel upstream at https://bugzilla.kernel.org; when doing so, you should make sure to explain the configuration changes and/or unofficial patches you used when you compiled it.

However, from your description, it seem that the bug you're experiencing is not present in 2.6.35. Therefore, even if "ubuntu-bug linux" worked fine when running that kernel, that would not be a good way to report this bug, as it would attach information about the wrong kernel. Therefore, to report the bug against a 2.6.38 kernel when running the 2.6.35 kernel, you should run:

ubuntu-bug linux-image-2.6.38-9-generic

Or, if you've already removed that version (which is the one from -proposed):

ubuntu-bug linux-image-2.6.38-8-generic

As an alternative technique--and this is arguably better, though only marginally--you could boot into the kernel where the bug occurs and run:

apport-cli linux

That will collect information, upload it, and provide a URL. Tell it not to open a web browser, but copy the URL. This should only take a couple of minutes at most, so with one or two tries, you should be able to accomplish this on the system running the bad kernel. Then you can boot with the good kernel (or use another machine) and go to the URL provided. You can do this by running "ubuntu-bug linux" too, but the added time and resources associated with actually spawning a browser window and going to the URL might make it harder to complete the operation and get the URL in time.

"- /var/log/dmesg does not shows any messages [Hardware Error] ...."

That's not surprising, though it was worth checking. Since you had not specified which log files you'd checked, I wasn't sure you'd checked the right ones--I see now that you have. Since prior to Natty, Ubuntu put the dmesg in /var/log/messages instead of /var/log/dmesg, there have been some questions posted expressing confusion about where to find the kernel's logs on disk.

I recommend attaching the dmesg anyway, even though it doesn't contain *direct* documentation of the panic. It may still contain information that might shed light on the problem. I do recommend pointing out in the report that it doesn't contain a log of the problem, and attaching the text you wrote based on your photo, as well as the image itself, to your bug report, as described above. (When you run ubuntu-bug or apport-cli or the like, it might automatically attach the dmesg, so then you needn't do it.)

"it's fairly hard to provide dmesg from crashed installation process."

And also fairly hard to run a different kernel in the installation process, yes. But since you said you were booting into the 2.6.25 kernel, presumably you've either overcome that (via live CD customization?) or you've got an installed system on which the problem occurs. Is that the case?

"rather than throwing some lame photos from my mobile, I've preferred to put some effort and at least copy vital text in so anyone searching for this type of bug will find what he/she look for."

Yes, if you can copy all the text precisely and without error, you're quite right that this is preferable. If you get all the text from the screen exactly, then I agree that you should not bother attaching the image.

@Tomasz Kusmierz
"- ubuntu-bug linux tells me that I can NOT report this since this is "not original ubuntu package" (when running on 2.6.35) and when running on 2.6.38 it does not manage to get to sending process because kernel crashes."
 
If you run a custom kernel--i.e., one you build yourself, with any changes of any kind in the configuration--then you cannot report a bug in it against linux in Ubuntu. (Actually, you can, but you shouldn't, and Apport won't facilitate it easily, and if you do, your bug will be summarily marked Invalid.) To be reported downstream, bugs in a custom kernel must be reproduced in a vanilla (i.e., non-customized) kernel. You can still report bugs affecting a custom kernel upstream at https://bugzilla.kernel.org; when doing so, you should make sure to explain the configuration changes and/or unofficial patches you used when you compiled it.