From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.1) Gecko/20031014 Description of problem: See above. Version-Release number of selected component (if applicable): kernel-smp-2.4.22-1.2093.nptl How reproducible: Always Steps to Reproduce: 1. Install "kernel-smp-2.4.22-1.2093.nptl" on a PR440FX dual Pentium Pro system. 2. Reboot system. 3. check system activity by means of "top" or "xosview" Actual Results: Only one processor is displayed. Expected Results: Both processors should be up. Additional info:
Created attachment 95280 [details] "dmesg" boot log for "kernel-smp-2.4.22-1.2093.nptl"
Same behaviour for "kernel-smp-2.4.22-1.2097.nptl" kernel package :(
What was the last kernel that worked for you ?
Well, the last version working normally was the direct precursor of the "kernel-smp-2.4.22-1.2093.nptl" kernel package, which was "kernel-smp-2.4.22-1.2087.nptl", I think.
Oops, finally it was rather "kernel-smp-2.4.22-1.2088.nptl". "dmesg" output follows.
Created attachment 95439 [details] "dmesg" boot log for "kernel-smp-2.4.22-1.2088.nptl"
Does 2108 fare any better ?
Created attachment 95503 [details] kernel-smp-2.4.22-1.2108.nptl-dmesg.output After installing kernel-smp-2.4.22-1.2108 ht sopped working on my 2,6 ghz pentium 4 box. I think this is related to this bug, eventhough it worked up to kernel-smp-2.4.22-1.2105.
No, unfortunately, "kernel-smp-2.4.22-1.2108.nptl" behaves exactly as any SMP kernel later than "kernel-smp-2.4.22-1.2088.nptl" which means it does not boot the 2nd CPU.
I am also running the 2108 kernel on a 3.06 GHz P4 box with hyper threading and though the 2nd cpu is recognized, no processes get assigned to it and it sits idle at 0% usage all the time. Dmesg assigns both cpu's, starts the migration threads, but just never runs any processes on the 2nd.
#8 and #10 are different bugs to this report, please open new ones for those. In the original report, the 2nd CPU wasn't detected at all. From the look of your dmesg, your CPUs are being detected.
"kernel-smp-2.4.22-1.2111.nptl" still only boots one CPU. "dmesg" output identical to previous versions.
"kernel-smp-2.4.22-1.2115.nptl" still only boots one CPU.
Is it possible for you to answer these three things?: 1. if you boot with "acpi=force" does that fair any better? If it works, we know this is an MPS issue, rather than an SMP issue. If it doesn't work it may or may not help us debug the problem, depending on how it fails. 2. If you re-build your kernel with the following line changed - #define APIC_DEBUG 0 + #define APIC_DEBUG 1 in include/asm-i386/apic.h, the dmesg output may explain why we don't enable the 2nd processor. 3. if you boot the latest 2.4.23 kernel with acpi=off, does it work? (ie. did the latest baseline regress also, or did just Fedora regress?) thanks, -Len
I have performed step 1 and 2 of the suggestions made by Len: 1. Adding the kernel option "acpi=force" leads to a kernel panic while the machine is booting. 2. The "dmesg" ouput for "APIC_DEBUG=1" is attached hereafter. I will try step 3 once the final version 2.4.23 of the Linux kernel is released. BTW, 2 years ago, I had reported a still pending issue concerning APIC IRQ-routing in bug #55223 related to the INTEL PR440FX platform.
Created attachment 95965 [details] "dmesg" boot log for "kernel-smp-2.4.22-1.2115.nptlcustom"
"kernel-smp-2.4.22-1.2129.nptl" boots the 2nd CPU correctly. "dmesg" output now essentially identical to that of version "kernel-smp-2.4.22-1.2088.nptl" again. The original IRQ mapping which had been corrupted by the later kernel versions has successfully been reestablished!