]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
x86: fix hpet timer reinit for x86_64
authorPavel Emelyanov <xemul@openvz.org>
Wed, 4 Feb 2009 10:40:31 +0000 (13:40 +0300)
committerIngo Molnar <mingo@elte.hu>
Thu, 5 Feb 2009 00:04:16 +0000 (01:04 +0100)
There's a small problem with hpet_rtc_reinit function - it checks
for the:

hpet_readl(HPET_COUNTER) - hpet_t1_cmp > 0

to continue increasing both the HPET_T1_CMP (register) and the
hpet_t1_cmp (variable).

But since the HPET_COUNTER is always 32-bit, if the hpet_t1_cmp
is 64-bit this condition will always be FALSE once the latter hits
the 32-bit boundary, and we can have a situation, when we don't
increase the HPET_T1_CMP register high enough.

The result - timer stops ticking, since HPET_T1_CMP becomes less,
than the COUNTER and never increased again.

The solution is (based on Linus's suggestion) to not compare 64-bits
(on 64-bit x86), but to do the comparison on 32-bit signed
integers.

Reported-by: Kirill Korotaev <dev@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
arch/x86/kernel/hpet.c

index 64d5ad0b8add85190ef600bb4f4887889729ef29..c761f914430ad3a463430cdd41ce5ba32d247932 100644 (file)
@@ -1075,7 +1075,7 @@ static void hpet_rtc_timer_reinit(void)
                hpet_t1_cmp += delta;
                hpet_writel(hpet_t1_cmp, HPET_T1_CMP);
                lost_ints++;
-       } while ((long)(hpet_readl(HPET_COUNTER) - hpet_t1_cmp) > 0);
+       } while ((s32)(hpet_readl(HPET_COUNTER) - hpet_t1_cmp) > 0);
 
        if (lost_ints) {
                if (hpet_rtc_flags & RTC_PIE)