Description Jacob Thompson 2007-04-06 17:06:09 UTC
Most recent kernel where this bug did *NOT* occur: N/A Distribution: Fedora Core 6.92 (Rawhide) Hardware Environment: Acer Veriton 6900 Pro, Intel Pentium D dual-core Software Environment: Linux 22.214.171.124 Problem Description: The system will hang while the kernel is initializing if the option acpi=off is not specified on the kernel command line. After the hang occurs, the Num Lock key on the keyboard no longer turns Num Lock on or off. This is an Acer Veriton 6900 Pro system, Intel Pentium D dual-core CPU, Phoenix-Award BIOS. Steps to reproduce: Boot kernel without acpi=off option
Comment 1 Jacob Thompson 2007-04-06 17:07:08 UTC
Created attachment 11086 [details] Kernel messages with options: ro root=LABEL=/ console=ttyS0 acpi_dbg_layer=1 acpi_dbg_level=1 apic=debug debug
Comment 2 Jacob Thompson 2007-04-06 17:07:48 UTC
Created attachment 11087 [details] Kernel messages with options: ro root=LABEL=/ console=ttyS0 acpi=ht acpi_dbg_layer=1 acpi_dbg_level=1 apic=debug debug
Comment 3 Jacob Thompson 2007-04-06 17:08:19 UTC
Created attachment 11088 [details] Kernel messages with options: ro root=LABEL=/ console=ttyS0 acpi=off acpi_dbg_layer=1 acpi_dbg_level=1 apic=debug debug
Comment 4 Len Brown 2007-04-11 19:29:21 UTC
What is the most recent kernel that worked properly without any cmdline workarounds? Does the latest upstream kernel (2.6.21-rc6) work any better? any luck with "pci=nommconf" or with "noapic"?
Comment 5 Jacob Thompson 2007-04-12 08:43:19 UTC
I will have access to this machine on Apr. 13. It is a new machine and I have not tried a kernel earlier than 2.6.18 (Fedora version). I plan on trying kernel.org versions of 2.6.21-rc6, 126.96.36.199, and 188.8.131.52 with and without the suggested options. Are there any other versions that would be good baselines to try (e.g. before major ACPI changes)?
Comment 6 Jacob Thompson 2007-04-13 20:20:29 UTC
I tried kernel 2.6.21-rc6 tonight. With just pci=nommconf, the kernel prints an irq 19: nobody cared message, suggests trying irqpoll, and gets stuck trying to detect the CD-ROM drive. It is not a complete hang as SysRq combinations still work. With both the pci=nommconf and irqpoll options, the kernel boots and runs fine.
Comment 7 Shaohua 2007-04-18 19:07:17 UTC
please attach the dmesg and acpidump output with pci=nommconf. mmconf is known broken in a lot systems. irqpoll suggests there is a irq mis-routing.
Comment 8 ykzhao 2007-08-30 23:13:28 UTC
Will you please upload the acpidump after the system is booted with option acpi=off ?
Comment 9 Jacob Thompson 2007-10-06 15:18:52 UTC
Here is an update: pci=nommconf with no other options is working (most of the time, see paragraph below) as of 184.108.40.206-91.fc7 and I am attaching dmesg output and the requested acpidump. I am using the output from a Fedora kernel, if that is not OK, I can compile one from kernel.org. I do still have problems with intermittent hangs on bootup even when using pci=nommconf. The kernel prints a detailed "oops" style error message and prints another error message every 5 seconds. This is not consistently reproducible can sometimes be fixed by disconnecting all USB devices and rebooting. I can try to capture the kernel output if I can get it to happen again.
Comment 10 Jacob Thompson 2007-10-06 15:19:46 UTC
Created attachment 13061 [details] Kernel messages with options: ro root=LABEL=/ pci=nommconf
Comment 11 Jacob Thompson 2007-10-06 15:20:15 UTC
Created attachment 13062 [details] acpidump with options: ro root=LABEL=/ pci=nommconf
Comment 12 Jacob Thompson 2007-10-06 16:49:48 UTC
Created attachment 13064 [details] Kernel messages for intermittent hang with pci=nommconf
Comment 13 Jacob Thompson 2007-10-06 16:50:12 UTC
Created attachment 13065 [details] Kernel messages for intermittent hang with pci=nommconf pci=routeirq
Comment 14 Jacob Thompson 2007-10-06 16:57:18 UTC
Created attachment 13066 [details] Kernel messages for successful boot with pci=nommconf pci=routeirq This was literally just a few minutes after attempting to boot with pci=nommconf pci=routeirq and getting the errors in attachment 13065 [details]. After attachment 13065 [details] I rebooted the system and one of the BIOS screens that flashed by showed "Update Successful", so I suspect that the BIOS is changing something in ACPI/ESCD/whatever in between the boot attempts.
Comment 15 Fu Michael 2007-10-22 06:11:10 UTC
could you please check to see if http://bugzilla.kernel.org/show_bug.cgi?id=8833#c12 can fix your problem?
Comment 16 Jacob Thompson 2007-10-26 17:19:02 UTC
Using the patch at http://marc.info/?l=linux-kernel&m=118809338631160&w=2 against 220.127.116.11 the hang appears to be resolved.
Comment 17 Fu Michael 2007-10-26 17:22:49 UTC
great! mark it as dup then. *** This bug has been marked as a duplicate of bug 8838 ***
Comment 19 Len Brown 2008-01-12 16:04:58 UTC