This fixed oprofile being broken on x86-64 SMP. The current reservation runs on all CPUs, but the ones following the first, will fail since the reservation bitmap has an 1 in the MSR entry. The reservation code should really run once, and their addresses used on other CPUs. There are other solutions to this, but this is the simplest and shorter one that came in my mind. Tested in my dual Opteron 252, and working fine here. Andi, if you don't like both Ben and my patch, would you mind fixing oprofile yourself in some way?
Signed-off-by: Davide Libenzi <[EMAIL PROTECTED]> - Davide Index: linux-2.6.22-rc1-git6.oprof/arch/i386/oprofile/nmi_int.c =================================================================== --- linux-2.6.22-rc1-git6.oprof.orig/arch/i386/oprofile/nmi_int.c 2007-05-17 18:32:48.000000000 -0700 +++ linux-2.6.22-rc1-git6.oprof/arch/i386/oprofile/nmi_int.c 2007-05-17 18:51:28.000000000 -0700 @@ -130,8 +130,19 @@ static void nmi_save_registers(void * dummy) { int cpu = smp_processor_id(); - struct op_msrs * msrs = &cpu_msrs[cpu]; - model->fill_in_addresses(msrs); + struct op_msrs *msrs = &cpu_msrs[cpu]; + + /* + * Use the CPU#0 already allocated addresses as reference. + */ + if (cpu != 0) { + const struct op_msrs *rmsrs = &cpu_msrs[0]; + + memcpy(msrs->counters, rmsrs->counters, + model->num_counters * sizeof(msrs->counters[0])); + memcpy(msrs->controls, rmsrs->controls, + model->num_controls * sizeof(msrs->controls[0])); + } nmi_cpu_save_registers(msrs); } @@ -195,6 +206,7 @@ static int nmi_setup(void) { int err=0; + struct op_msrs *msrs; if (!allocate_msrs()) return -ENOMEM; @@ -203,6 +215,14 @@ free_msrs(); return err; } + /* + * MSR reservation should happen only once with the current API, + * otherwise CPU other then the first would fail to allocate. + * We do it here, and then we propagate the CPU#0 addresses to + * other CPUs. + */ + msrs = &cpu_msrs[0]; + model->fill_in_addresses(msrs); /* We need to serialize save and setup for HT because the subset * of msrs are distinct for save and setup operations - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/