Latest working kernel version: could not find any, which uses Tickless Earliest failing kernel version: afaik since Tickless was introduced Distribution: Ubuntu Hardware Environment: HP/Compaq 6715s Software Environment: Problem Description: When Tickless and/or HighResTimer is enabled my machine hangs, while saying: * Reading files reeded to boot. Screen turns black after some seconds. I have to turn off the power. On one ocasion boot progress went further. Lockup happened a few seconds later, however I could not reproduce this Passing nohz=off to the kernel (or don't enable the options in the .config before compile) solves the issue Happens with 2.6.25.2 from kernel.org and 2.6.24 from Ubuntu hardy I suppose this problem was already present in 2.6.22 (32-bit), since only the 64-bit version worked on this machine, which did not include the named featueres. Steps to reproduce: Compile the 2.25.2 kernel with Tickless or/and HighResTimer and start the machine. Thank you in advance! Best Regards, Simon
Created attachment 16250 [details] lspci output
Hello again everybody, I forgot: passing noapic to the kernel will also allow the system to start. Although in this case hdd-performance degrades such, that the system gets unusable. Best Regards, Simon
Simon: Does 2.6.26 improve things? Thomas: any guesses on this one?
Hi everyone! No, Kernel 2.6.26 does not improve anything. The first time I started my PC with Kernel 2.6.26, the problem seemed to be gone. The boot-process proceeded normally, for about 20 seconds, then the system hung again. Now, it hangs at the usual moment. Best Regards, Simon
On Sun, 17 Aug 2008, bugme-daemon@bugzilla.kernel.org wrote: > No, Kernel 2.6.26 does not improve anything. > The first time I started my PC with Kernel 2.6.26, the problem seemed to be > gone. The boot-process proceeded normally, for about 20 seconds, then the > system hung again. Now, it hangs at the usual moment. Can you please provide a full bootlog from that machine with 2.6.26 ? Please add apic=verbose to the kernel command line. If it's not possible to get a bootlog with nohz/highres enabled then send one with both disabled on the kernel command line. Any chance to give 2.6.27-rc4 a try ? It contains lots of updates in that area. Thanks, tglx
Created attachment 17653 [details] System-log kernel 2.6.26
Created attachment 17654 [details] System-log kernel 2.6.27-rc5
Hello Thomas! Thank you for your effort. Today I tried the linux-kernel 2.6.27-rc5. Now the System starts without passing any parameters to the kernel. However the system-performance is very low (GNOME is too slow to be usable). The Processor speed is locked to 800MHz (the lowest speed) and the CPU-fan always runs at its maximum. It seems incredible to me, but today also 2.6.26 started the same way 2.6.27-rc5 did. Attached you can find the bootlogs. I believe the following bug is related: https://bugs.launchpad.net/bugs/134856 Thank you a lot! best regards, Simon
Simon, could you please apply the patch: http://bugzilla.kernel.org/attachment.cgi?id=17644&action=view on top of 2.6.27-rc5 and check whether there is any difference ? Thanks, tglx
Another thing. Can you please provide the output of cat /proc/acpi/processor/CPU0/power Thanks, tglx
Created attachment 17656 [details] output of cat /proc/acpi/processor/C000/power
Sorry forgot to ask for the output of cat /sys/devices/system/clocksource/clocksource0/available_clocksource and cat /proc/timer_list
stupid me, wanted to have cat /sys/devices/system/clocksource/clocksource0/current_clocksource as well
Here we go ;-): #sudo cat /sys/devices/system/clocksource/clocksource0/current_clocksource acpi_pm #sudo cat /sys/devices/system/clocksource/clocksource0/available_clocksource acpi_pm jiffies tsc cat /proc/timer_list is attached. Everything under 2.6.25.2 (no highres timers and no tickless) I'll tell you the results of the patch tomorrow. The kernel is still compilig and It's time to go to bed. Best regards! Simon
Created attachment 17657 [details] sudo cat /proc/timer_list
Thanks, sleep well !
Ok, my pc was faster than I thougt! Nothing has chnaged unfortunately. The result is attached! Good night ;-)!
Created attachment 17658 [details] Syslog 2.6.27-rc5 patched
Created attachment 17715 [details] Test patch to find out whether C1E and ACPI C-States do not mix Can you please test the attached patch on top of 2.6.27-rc6 ?
Created attachment 17754 [details] yslog 2.6.27-rc6 patched Hi Thomas! Here is the result! I just noted, that this version of the Kernel won't even work properly with nohz=off. Is this due to the Patch? Best regards, Simon
This should be fixed on recent kernels (2.6.27.7 or 2.6.28-rcX). Note that the 6715s with 2.6.27.7 is very slow, but this is related to another bug: http://bugzilla.kernel.org/show_bug.cgi?id=10778 Can someone other confirm this?
Someone should close this bug.
Hi, I just tested 2.6.29-rc2. No go! The systems hangs even without Tickless/HiRes Timers, but probably I'm doing someting wrong! Anyway, I opened the bug, so now I'll close it. Thank you for your help! Best regards, Simon