|
Bugzilla – Full Text Bug Listing |
| Summary: | inconsistent freeze patterns at 10.0 boot | ||
|---|---|---|---|
| Product: | [openSUSE] SUSE LINUX 10.0 | Reporter: | Dzmitry Prakapenka <4maillists> |
| Component: | Kernel | Assignee: | Thomas Renninger <trenn> |
| Status: | RESOLVED WONTFIX | QA Contact: | E-mail List <qa-bugs> |
| Severity: | Critical | ||
| Priority: | P5 - None | CC: | mark.langsdorf |
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Hardware: | x86-64 | ||
| OS: | SuSE Linux 10.0 | ||
| Whiteboard: | |||
| Found By: | Customer | Services Priority: | |
| Business Priority: | Blocker: | --- | |
| Marketing QA Status: | --- | IT Deployment: | --- |
|
Description
Dzmitry Prakapenka
2005-11-06 22:16:11 UTC
I tried installing the 32-bit version of Suse 10 and it's been working fine for a few days already without any of the above issues. John Gelnaw contacted me with a possible solution for the problems described above: "The solution on my system was to disable powersaved via: chkconfig powersaved off I finally had to install with the default run level set to 3, then was able to log in via text mode to make the change. Also, disabling "USB Keyboard+Storeage" (their typo, not mine) in the BIOS helped. Didn't solve the problem, but it allowed my system to boot far enough that I could get into text mode." Thomas, this looks like a problem with x86_64 cpufreq - can you take a look please? AMD athlon 64 3000+(overclocked to 3200+ and working stable). How have you overclocked, by hardware (jumpers?), by BIOS (FSB freq), by OS (cpufreq drivers)? Sorry, this one is invalid. If you alter the default frequencies, it is no wonder that your machine freezes. I expect that BIOS exports still the same default frequencies? It should be possible to patch the powernow-k8 driver, ignore ACPI frequency/volt pairs and statically compile in your preferred overclocked/underclocked values. There was a nice introduction/report on cpufreq@lists.linux.org.uk, hmm I just cannot find the original mail (something with "overclocking and undervoltaging", maybe it was another mailinglist), here is a comment from Dominik the cpufreq maintainer: The values reported by cpufreq are the values the CPU is supposed to run at, that is the hardcoded FSB speed mutliplied by the "multiplier factor". If you modify the FSB (and not the multiplier), this isn't detectable by cpufreq. However, if you modify the multiplier, you'd need to override the PSB or ACPI tables, and the frequency table helpers are the wrong place to look for that. You also might find some info about the overclocking/undervoltaging stuff somewhere at this site: http://fab51.com/cpu/barton/athlon-e24.html I overcocked that machine by changing the FSB freq. in BIOS... I tried installing SuSe 10.0 x86_64 on a different machine (a clone of the 1st one except the CPU) with the following specs: AMD athlon 64 3200+ (not overclocked) NVidia GeForce FX5900XT 1.0 GB DDR pc 3200 DFI lanpartyUT 250gb motherboard ... and the same problems persist. After that I tried installing the 32-bit version of Suse 10 on that machine and it is working fine without any of the above issues... can you connect a null modem with serial console and boot with console=ttyS0,baud and boot with powersaved enabled and add the log upto the freeze to the bug? Unfortunately I cannot do it as I do not have a null modem with serial console... So this is a -default not an -smp kernel? What is the value for CPUFREQ_CONTROL="" in /etc/sysconfig/powersave/cpufreq. Does it help if you switch (userspace/kernel) and restart powersaved or reboot? Any error messages still written to disk in /var/log/messages before the system froze totally? Hmmm, rereading: We still don't know for sure whether it is cpufreq or ACPI access of powersaved that causes the freeze? You can check by setting CPUFREQD_MODULE="off" (same config file as above). If it does not freeze anymore it's probably powersaved accessing /proc/acpi/... That's even more likely, cpufreq on AMD UP is rather stable. If it is ACPI, please attach acpidmp output. Maybe you should also attach dmesg output, maybe there is some kind of hint to the problem ... I'm really sorry but the hardware on which the issue was occuring is longer accessible to me, so I cannot troubleshoot it anymore. :( Some answers to your questions: Yes, that was the default, not the -smp kernel. It did not help to switch (userspace/kernel) and restart powersaved or reboot - the system froze every time. As this is the only such bugreport with -default kernel I am closing this one. Would be nice if you reopen it again if you encounter something similar. Especially with newer kernels like recent OpenSuse 10.1. As sooner we get those reports as higher is the chance we could still fix those... Thanks for reporting! |