Most recent kernel where this bug did not occur: None found Distribution: Gentoo Hardware Environment: MSI rs480m2 motherboard + AMD64, Antec Aria case Software Environment: Gentoo linux Problem Description: MSI rs480m2 system (AMD64 processor) will not boot when either acpi or apic are enabled in the boot arguments. This has been tested in many kernels (gentoo and vanilla) up and including 2.6.16-rc1. The info here is from the 2.6.16-rc1 kernel 1) When booting with command line root=/dev/ram0 init=/linuxrc ramdisk=8192 real_root=/dev/sda5 udev report_lost_ticks=100 noapictimer the following is a screenshot of what happens. First it pauses at "sata_sil" and then http://img66.imageshack.us/img66/2489/screenacpismall4my.jpg 2) When booting with commmand line root=/dev/ram0 init=/linuxrc ramdisk=8192 real_root=/dev/sda5 udev report_lost_ticks=100 acpi=off the following is a screenshot of what happens. http://img204.imageshack.us/img204/9753/screenapicsmall5tx.jpg A litte more investigation shows that the system will in fact boot with Command line root=/dev/ram0 init=/linuxrc ramdisk=8192 real_root=/dev/sda5 udev report_lost_ticks=100 noapictimer acpi=noirq Steps to reproduce: Attempt to boot machine with APIC or ACPI enabled. Raphael
Created attachment 7088 [details] dmesg output with noapictimer acpi=noirq
Created attachment 7089 [details] lspci -vv output
Created attachment 7090 [details] acpidump.out
Created attachment 7091 [details] ./acpidump/acpidump -t DSDT -b -o DSDT;output of iasl -d DSDT
Created attachment 7092 [details] dmidecode output
cat /proc/interrupts CPU0 0: 22159 XT-PIC timer 1: 122 XT-PIC i8042 2: 0 XT-PIC cascade 3: 1 XT-PIC ATI IXP 5: 554 XT-PIC eth0 8: 0 XT-PIC rtc 9: 0 XT-PIC acpi 10: 1242 XT-PIC ehci_hcd:usb1, ohci_hcd:usb2, ohci_hcd:usb3, libata 11: 7585 XT-PIC libata 14: 25 XT-PIC ide0 NMI: 38 LOC: 0 ERR: 0 MIS: 0
Downstream report: http://bugs.gentoo.org/119833
It seems I made a brown paper bad testing error! At some point in the 2.6.x kernel series after 2.6.14 you can in fact boot the system if BOTH acpi and apic are on (not if one is on and one is off). So 2.6.16-rc1 in fact does boot with both acpi and apic on. Unfortunately this doesn't fix the main problem which is the interrupt/timing issue when writing/reading cds/dvds for example :( However, that is probably for another bug report. I attach the dmesg in any case just for completeness.
Created attachment 7115 [details] dmesg with apic and acpi enabled I can't understand which warnings are bad in dmesg (ignore fglrx errors which seems to be an incompatibility with 2.6.16-rc1) but note that time.c: Lost 3 timer tick(s)! rip msleep+0x3/0x3a) time.c: Lost 2 timer tick(s)! rip msleep+0x3/0x3a) is still there and mounting and ripping a dvd ( for example) results in time.c: Lost 6 timer tick(s)! rip default_idle+0x2a/0x52) hda: irq timeout: status=0x90 { Busy } ide: failed opcode was: unknown hda: ATAPI reset complete UDF-fs INFO UDF 0.9.8.1 (2004/29/09) Mounting volume 'DUNE_CVR_MNT', timestamp 2036/02/07 01:58 (1000) time.c: Lost 6 timer tick(s)! rip default_idle+0x2a/0x52) time.c: Lost 4 timer tick(s)! rip default_idle+0x2a/0x52) time.c: Lost 5 timer tick(s)! rip default_idle+0x2a/0x52) [...] Losing some ticks... checking if CPU frequency changed. [...] The clock is also running far too fast now (more than double speed).
Created attachment 7116 [details] dmesg with no_timer_check As suggested this is a boot with no_timer_check added. Note time.c: Lost 1 timer tick(s)! rip acpi_os_write_port+0x1a/0x34) ..MP-BIOS bug: 8254 timer not connected to IO-APIC failed. timer doesn't work through the IO-APIC - disabling NMI Watchdog! Uhhuh. NMI received for unknown reason 2d. Dazed and confused, but trying to continue Do you have a strange power saving mode enabled? works. Using local APIC timer interrupts. and time.c errors still there. Ripping a DVD gives roughly the same problems as before but the clock is no longer too fast.
The system will boot with the latest kernel. I will report the remaining problems in a separate bug.