Description of problem: The latest kernel (kernel-2.6.15-1.1831_FC4) won't boot on our system, while the default kernel that comes with FC4 DVD iso boots without problems (kernel-2.6.11-1.1369_FC4). The error reported is: BUG: Spinlock bad magic on CPU#0, swapper/1 (Not tainted) BUG: Spinlock lockup on CPU#0, (several garbage characters here)/0, ffffffff80450690 (Not tainted) Version-Release number of selected component (if applicable): kernel-2.6.15-1.1831_FC4 How reproducible: Always Steps to Reproduce: 1. boot system with latest kernel 2. watch it die with the above error Actual results: It doens't boot Expected results: It should boot Additional info: system hardware is as follows: motherboard: Sapphire PURE Innovation PI-A9RX480 cpu: AMD Athlon 64 3500+ memory: 512MB ram drives: 2x 80gig drives in software raid One additional detail: No linux kernel boots if "APIC" is enabled. Thus the APIC option is disabled from the bios.
I've just tested a live CD (PCLinuxOS) with kernel 2.6.12 and that worked fine. In addition, the 2.6.12 kernel also adds support for the internal ethernet chipset (Marvel) which isn't supported by the default FC4 kernel.
Based on some discussions in IRC about this problem, people have been asking me if this is an SMP system or if i'm running an SMP kernel. No, this is NOT an SMP system, its a UP, running a single Athlon 64 3500+. I don't believe i'm running an SMP kernel, 'yum' should have installed the correct 2.6.15 UP kernel i believe. Here is what i've got: # rpm -q kernel kernel-2.6.11-1.1369_FC4 kernel-2.6.15-1.1831_FC4
I just tested another kernel, got it from the fedora/redhat ftp and its the kernel 2.6.14-1.1656_FC4, which also boots without problems. This leads me to believe that 2.6.15 has something specific that makes it unbootable on this machine.
Hello? anyone reading these bug reports?
WOOOOW!!! Apparently the .15 kernels i've been trying to use are SMP while all the others are NOT SMP. I've been told that redhat decided to use SMP kernels on UP systems because the difference in performance is insignificant. I dont know, but everyone tells me that my problem is related to the SMP being enabled on the kernel....
that decision has been made for FC5. The change accidentally got backported to the last rebase for the FC4 kernel. The next errata will make the UP x86-64 kernel UP again. The work-in-progress kernel at http://people.redhat.com/davej/kernels/Fedora/FC-4/ has this fixed already.
Its alive! its alive!.... erm sorry :P so, where was i... ah yes, well those work-in-progress kernels aren't any help to me, since my Marvel ethernet chipset only works with either a .16 kernel or the -netdev kernels someone has made from redhat. So i'll have to wait a while for either a .16 kernel or for the -netdev kernels to become non-smp. i've been told that sk98lin has been deprecated and that a new driver (ske2 or skge or something) will have support for the Marvel gigabit ethernet. The new driver only exists in .16 kernels. Thank you anyway.