[PATCH v4 21/24] arm/v2m: initialize arch_timers even if v2m_timer is not present

On 07/01/13 17:21, Stefano Stabellini wrote:

On Fri, 14 Sep 2012, Stefano Stabellini wrote:

quoted

On Fri, 14 Sep 2012, Pawel Moll wrote:

quoted

On Fri, 2012-09-14 at 12:13 +0100, Stefano Stabellini wrote:

quoted

Signed-off-by: Stefano Stabellini <redacted>
CC: Russell King <redacted>
CC: Pawel Moll <redacted>
CC: Marc Zyngier <redacted>
---
 arch/arm/mach-vexpress/v2m.c |   11 ++++++-----
 1 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/arch/arm/mach-vexpress/v2m.c b/arch/arm/mach-vexpress/v2m.c
index 37608f2..4e567f7 100644
--- a/arch/arm/mach-vexpress/v2m.c
+++ b/arch/arm/mach-vexpress/v2m.c

@@ -621,16 +621,17 @@ static void __init v2m_dt_timer_init(void)
 
 	v2m_clk_init();
 
-	err = of_property_read_string(of_aliases, "arm,v2m_timer", &path);
-	if (WARN_ON(err))
-		return;
-	node = of_find_node_by_path(path);
-	v2m_sp804_init(of_iomap(node, 0), irq_of_parse_and_map(node, 0));
 	if (arch_timer_of_register() != 0)
 		twd_local_timer_of_register();
 
 	if (arch_timer_sched_clock_init() != 0)
 		versatile_sched_clock_init(v2m_sysreg_base + V2M_SYS_24MHZ, 24000000);
+
+	err = of_property_read_string(of_aliases, "arm,v2m_timer", &path);
+	if (WARN_ON(err))
+		return;
+	node = of_find_node_by_path(path);
+	v2m_sp804_init(of_iomap(node, 0), irq_of_parse_and_map(node, 0));
 }
 
 static struct sys_timer v2m_dt_timer = {

Fair point. The alias is going to disappear anyway (I'm working on a VE
platform rework right now), but in case I won't get it on time for 3.7,
I'll make sure this one is merged instead.

Great, thanks!
Should I leave this patch out of the Xen on ARM series for 3.7 then?

I have just noticed that this patch hasn't been merged yet, but still we
need it to boot Linux on Xen.
Do you want me to rebase and resent?

I'm looking at v3.8-rc2, and the code have changed significantly (see
433683a66401adb0150792e725cc4f631c94de46).

If you don't have an "arm,sp804" device in your DT, it shouldn't get
registered and carry on registering the per-cpu timers. Can you check
that it fixes your issue?

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...