Hi,
We recently encountered a failure in the code of test_apic_change_mode()
(x86/apic.c of kvm-unit-test). This failure, so far, only happened on
Intel Xeon "E3-1220 V2" CPU.
Here is the code snip where the failure happens:
static void test_apic_change_mode()
{
...
apic_write(APIC_TMICT, tmict);
wait_until_tmcct_is_zero(tmict, true);
apic_change_mode(APIC_LVT_TIMER_PERIODIC);
report("TMCCT should have a non-zero value", apic_read(APIC_TMCCT));
/*
* After the change of mode, the counter should not be reset and
* continue counting down from where it was
*/
report("TMCCT should not be reset to TMICT value",
apic_read(APIC_TMCCT) < (tmict / 2));
wait_until_tmcct_is_zero(tmict, false);
report("TMCCT should be reset to the initial-count",
apic_read(APIC_TMCCT) > (tmict / 2)); <=== FAILURE
...
}
The last statement above is where the failure happened. Obviously the
code is expecting the APIC_TMCCT counter will be reset to the initial
value stored in APIC_TMICT, right after the counter reach zero (it is a
periodic timer anyway). However this assumption shouldn't be held true
all the time. It is possible the test code will reads APIC_TMCCT before
the reset happens. If this happens, the test will fail. In fact, in my
experiment, it is quite common (>20% of chance) to see APIC_TMCCT 0 on
Intel "E3-1220 V2" CPU.
My suggestion is to remove the following two statements.
- wait_until_tmcct_is_zero(tmict, false);
- report("TMCCT should be reset to the initial-count",
- apic_read(APIC_TMCCT) > (tmict / 2));
Thanks,
-Wei
