Fix no_timer_check on x86_64

fixed upstream in 2.6.28 in merge of ioapic*.c for x86

In io_apic_32.c the logic of no_timer_check is "always make timer_irq_works
return 1".

Io_apic_64.c on the other hand checks for
  if (!no_timer_check && timer_irq_works())
basically meaning "make timer_irq_works fail" in the crucial first check.

Now, in order to not move too much code, we can just reverse the logic here
and should be fine off, basically rendering no_timer_check useful again.

This issue seems to be resolved as of 2.6.28 by the merge of io_apic*.c,
but still exists for at least 2.6.27.

Signed-off-by: Alexander Graf <agraf@suse.de>
Acked-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

diff --git a/arch/x86/kernel/io_apic_64.c b/arch/x86/kernel/io_apic_64.c
index 61a83b7..6ad530f 100644 (file)
--- a/arch/x86/kernel/io_apic_64.c
+++ b/arch/x86/kernel/io_apic_64.c
@@ -1729,7 +1729,7 @@ static inline void __init check_timer(void)
                        setup_timer_IRQ0_pin(apic1, pin1, cfg->vector);
                }
                unmask_IO_APIC_irq(0);
-               if (!no_timer_check && timer_irq_works()) {
+               if (no_timer_check || timer_irq_works()) {
                        if (nmi_watchdog == NMI_IO_APIC) {
                                setup_nmi();
                                enable_8259A_irq(0);