> I'd bet resonable sums of money that we are (once again) accessing > per-cpu variables in code before ia64 has a chance to initialize > things so that they can actually work. :-( I tried this simple hack to check ... per-cpu variables are not usable until cpu_init() sets up the "ar.k3" register on ia64. So this hack checks to see if it is set before calling cpu_clock(): --- a/kernel/printk.c 2008-08-01 10:20:47.000000000 -0700 +++ b/kernel/printk.c 2008-08-04 14:49:18.000000000 -0700 @@ -34,6 +34,7 @@ #include <linux/syscalls.h> #include <asm/uaccess.h> +#include <asm/kregs.h> /* * Architectures can override it: @@ -737,7 +738,10 @@ unsigned long long t; unsigned long nanosec_rem; - t = cpu_clock(printk_cpu); + if (ia64_get_kr(IA64_KR_PER_CPU_DATA)) + t = cpu_clock(printk_cpu); + else + t = 0; nanosec_rem = do_div(t, 1000000000); tlen = sprintf(tbuf, "[%5lu.%06lu] ", (unsigned long) t, With this hack the kernel boots (with CONFIG_PRINTK_TIME=y). This is effectively how the code used to operate before the architecture hooks were cleaned away by: commit b842271fbb9c8b5fd0e1c3e1895a3b67ba5bcc54 Author: Ingo Molnar <mingo@xxxxxxx> Date: Fri Jan 25 21:07:59 2008 +0100 sched: remove printk_clock() printk_clock() is obsolete - it has been replaced with cpu_clock(). -Tony -- To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html