Skip to contentSkip to conversation

arm: fix mach-versatile booting in qemu

PATCH New Arnd Bergmann 3 participants 4 messages Jul 14, 2010

Subsystems (2)
Arm (32-bit) Arm sub-architectures
lore
Share & tools
lore

Arnd Bergmann

author
#0 lore raw

PATCH 1/1 arm: fix mach-versatile booting in qemu

The versatile platform has not been able to boot in qemu since 2.6.34-rc2, because e388771 "ARM: Realview/Versatile: separate out common SP804 timer code" changed the behavior of the versatile_set_mode function.

This restores the original way it worked until 2.6.34-rc1.

Signed-off-by Arnd Bergmann <[email protected]>

---
I found this with git-bisect. I do not know wether the old behavior that this reverts to is sensible, but it's how both versatile and realview have been doing it since they first started using clockevents.

I also don't know if the same problem exists on actual versatile hardware or if it's really a qemu bug.

Please apply in combination with common sense. If it's correct, do we also want the same change in 2.6.34.x-stable?

arch/arm/plat-versatile/timer-sp.c +1 -2
@@ -90,8 +90,6 @@ static void sp804_set_mode(enum clock_event_mode mode,
 {
 	unsigned long ctrl = TIMER_CTRL_32BIT | TIMER_CTRL_IE;
 
-	writel(ctrl, clkevt_base + TIMER_CTRL);
-
 	switch (mode) {
 	case CLOCK_EVT_MODE_PERIODIC:
 		writel(TIMER_RELOAD, clkevt_base + TIMER_LOAD);

@@ -106,6 +104,7 @@ static void sp804_set_mode(enum clock_event_mode mode,
 	case CLOCK_EVT_MODE_UNUSED:
 	case CLOCK_EVT_MODE_SHUTDOWN:
 	default:
+		ctrl = 0;
 		break;
 	}
#1 lore raw

PATCH 1/1 arm: fix mach-versatile booting in qemu

Quoted 6 lines · Arnd Bergmann

The versatile platform has not been able to boot in qemu since 2.6.34-rc2, because e388771 "ARM: Realview/Versatile: separate out common SP804 timer code" changed the behavior of the versatile_set_mode function.

This restores the original way it worked until 2.6.34-rc1.

Quoted 8 lines
Signed-off-by Arnd Bergmann <[email protected]>

---
I found this with git-bisect. I do not know wether the old behavior that this reverts to is sensible, but it's how both versatile and realview have been doing it since they first started using clockevents.

I also don't know if the same problem exists on actual versatile hardware or if it's really a qemu bug.

It's a QEMU bug. It was fixed with this patch to QEMU:

http://www.mail-archive.com/qemu-devel at nongnu.org/msg30785.html

Rabin

author
#2 lore raw

PATCH 1/1 arm: fix mach-versatile booting in qemu

Quoted 7 lines · On Wednesday 14 July 2010 19:23:19 Rabin Vincent wrote:
Quoted 2 lines · Arnd Bergmann

I also don't know if the same problem exists on actual versatile hardware or if it's really a qemu bug.

It's a QEMU bug. It was fixed with this patch to QEMU:

http://www.mail-archive.com/qemu-devel at nongnu.org/msg30785.html

Ok, thanks for the info. I was sure I was using the latest qemu checkout, but apparently I made a mistake there.

Arnd

#3 lore raw

PATCH 1/1 arm: fix mach-versatile booting in qemu

In reply to Arnd Bergmann
Quoted 3 lines · Arnd Bergmann

The versatile platform has not been able to boot in qemu since 2.6.34-rc2, because e388771 "ARM: Realview/Versatile: separate out common SP804 timer code" changed the behavior of the versatile_set_mode function.

NAK. qemu is broken.