This is the same problem as bug #152630: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=152630 The bug is still present in the latest RHEL4 update kernel (2.6.9-5.0.3.EL). Note that my initial guess about the problem in RHEL3 must have been wrong. The ACPI code in RHEL4 is more up to date, and includes all the patches to mpparse.c which I was trying in the RHEL3 kernel. Something else must be causing the timer interrupt to be received twice. The workaround must have fixed things by using the XT-PIC to handle the timer instead of the APIC.
Created attachment 112601 [details] boot dmesg from latest RHEL4 kernel (2.6.9-5.0.3.EL) this is the boot dmesg from the latest update kernel.
I also get the following error messages from the kernel, seemingly when the system is under load. (network or disk controller) APIC error on CPU0: 00(40) APIC error on CPU0: 40(40)
Confirmed that this problem exists in the latest official 2.6 kernel (2.6.11.6). I downloaded linux-2.6.11.6.tar.bz2 from kernel.org and used the existing RHEL4 kernel source RPM to build the kernel. (I removed all the Red Hat %patches, and used 'make config' to translate the existing Red Hat kernel configs into something that would work with vanilla 2.6.11.6) My new 2.6.11.6 kernel RPM built, installed, and booted fine. The clock still runs at twice normal speed though. $ uname -r 2.6.11.6-5.EL.wing $ cat /proc/interrupts; sleep 10; cat /proc/interrupts (5 seconds go by) 0: 1215570 IO-APIC-edge timer LOC: 607691 0: 1225584 IO-APIC-edge timer LOC: 612698 (2000 timer ints/second, 1000 local APIC ints/second) I still get: APIC error on CPU0: 00(40) APIC error on CPU0: 40(40) I may report this to the linux-kernel mailing list, and see if this is a Linux bug or a bad BIOS.
The problem goes away when I disable APIC mode entirely (via a BIOS option). This corresponds to my earlier experience with RHEL3 and the 2.4 kernel; the patches I tried ended up somehow making only the timer interrupt handled via the XT-PIC, but left the other interrupts going through the APIC. A BIOS upgrade did not resolve the problem. (when APIC mode is enabled) I entered a support request with the system manufacturer.
Created attachment 112625 [details] boot dmesg from latest RHEL4 kernel (2.6.9-5.0.3.EL), with APIC mode disabled in BIOS The clock works properly when APIC mode is disabled in the BIOS; all interrupts in /proc/interrupts are labeled 'XT-PIC'.
I encountered this bug in FC4_x86-64 (kernel version 2.6.12-1.1398) on an Acer Aspire 5024WLMi (AMD Turion64 ML-34 prcoessor, ATI Mobility Radeon XPRESS 200 series chipset). The output of cat /proc/interrupts is similar to that of Comment #2 above. However, the Phoenix BIOS (version 1.11) of this laptop does not offer the opportunity to disable APIC mode as suggested above. Is there any opportunity to correct my system clock? Am I to post a new bug as this concerns FC4 (another "product")?
Also encountered this problem with a new Compaq Presario SR1030Z. Also can't disable APIC mode in BIOS. System clock situation improves when booting with 'noapic' kernel parameter--but LAN connectivity stops working. Tried a few kernels with same result: 2.6.9-22.0.1.EL 2.6.14-1.1644_FC4 2.6.14-1.1739_FC5 I need to get this machine in a useable state. Will buying a different video adapter and disabling the mainboard ATI work around the problem?
FYI this is the same as FC4 bug 152170. Much information there.
This bug is the same issue as Bug 173236, which was fixed in RHEL4.4 *** This bug has been marked as a duplicate of 173236 ***