Created attachment 69432 [details] dmesg output I have an Intel Core 2 Duo T5750 2.0Ghz on which 32-bit and 64-bit kernels fail to boot CPU #1. (Annoyingly, Vista is fine.) This problem is present in kernel 3.1.0-rc2 and all kernels tried (at least back to 2.6.24.4). Attached is data from dmesg, acpidump & mptable. I wondered if this was an ACPI or MP table problem but believe not. They looked OK to me so I adapted a standalone program (to be grub-bootable) which uses the Intel example way of booting (an APIC INIT IPI followed by STARTUP IPI twice, sent using broadcast to all except self). This works, so next I hacked arch/x86/kernel/smpboot.c's wakeup_secondary_cpu_via_init to use the same way - which also works. The relevant Intel manuals appear to include the "Intel 64 and IA-32 Architectures Software Developer's Manual Vol 3A", Chapter 8, and "x86 System Programming Guide", Chapter 7. I would be grateful for contact from anyone with ideas or with knowledge of why smpboot.c is structured as it is instead of using Intel's algorithm.
Created attachment 69442 [details] acpidump output
Created attachment 69452 [details] mptable output - looks like the kernel ignores it but maybe useful
dmesg shows "multiple APIC/MADT found"; it does not seem to matter whether acpi_apic_instance is set to 0 or 2.
This may be the same problem affecting various MacBooks such as with Ubuntu. google the dmesg message if interested!
Please test this bug and see if it's fixed with newer kernels. Cheers Nick