diff mbox

ARM: tegra30: fix power up sequence for boot_secondary

Message ID 1355970603-28531-1-git-send-email-josephl@nvidia.com (mailing list archive)
State New, archived
Headers show

Commit Message

Joseph Lo Dec. 20, 2012, 2:30 a.m. UTC
The power up sequence is different on the cold boot CPU and the CPU
that resumed from the hotplug. For the cold boot CPU, it was been power
gated as default. To power up the cold boot CPU, the power should be
un-gated by un toggling the power gate register manually.

For the CPU that resumed from the hotplug, after un-halted the CPU. The
flow controller will un-gate the power of the CPU. No need to manually
control, just wait the power be resumed and continue the power up
sequence after the CPU power is ready.

Based on the work by:
Varun Wadekar <vwadekar@nvidia.com>

Signed-off-by: Joseph Lo <josephl@nvidia.com>
---
 arch/arm/mach-tegra/platsmp.c |   31 ++++++++++++++++++++++++++++++-
 1 files changed, 30 insertions(+), 1 deletions(-)

Comments

Stephen Warren Dec. 20, 2012, 4:55 p.m. UTC | #1
On 12/19/2012 07:30 PM, Joseph Lo wrote:
> The power up sequence is different on the cold boot CPU and the CPU
> that resumed from the hotplug. For the cold boot CPU, it was been power
> gated as default. To power up the cold boot CPU, the power should be
> un-gated by un toggling the power gate register manually.
> 
> For the CPU that resumed from the hotplug, after un-halted the CPU. The
> flow controller will un-gate the power of the CPU. No need to manually
> control, just wait the power be resumed and continue the power up
> sequence after the CPU power is ready.

I'd like Peter to review this too (CC'd). I guess it looks OK...

> diff --git a/arch/arm/mach-tegra/platsmp.c b/arch/arm/mach-tegra/platsmp.c

> @@ -72,7 +75,27 @@ static int tegra30_power_up_cpu(unsigned int cpu)
>  	if (pwrgateid < 0)
>  		return pwrgateid;
>  
> -	/* If this is the first boot, toggle powergates directly. */
> +	/*
> +	 * If the CPU had booted and went through here after CPU
> +	 * had been already un-gated by flow controller. Wait
> +	 * for confirmation that the CPU is powered then remove
> +	 * the IO clamps. On the cold boot entry, do not wait.
> +	 */

That comment is quite unclear. In particular the first sentence says "if
something" rather than "if something then something". Can you please
re-write it?

Also, it'd be good to explain (or at least briefly reference) the boot
path for a cold-boot CPU and a warm boot CPU; I assume that somehow a
cold boot CPU doesn't execute tegra_secondary_init(), but a warm boot
CPU does?

> +	if (cpumask_test_cpu(cpu, &tegra_cpu_init_mask)) {
> +		timeout = jiffies + 5;

Are the units/frequency of the jiffies variable always constant? Should
"5" be not a constant, but instead a calculation involving HZ or
msecs_to_jiffies()? Certainly, the few other uses of time_before() I
looked at use HZ or msecs_to_jiffies().
Joseph Lo Dec. 21, 2012, 8:07 a.m. UTC | #2
On Fri, 2012-12-21 at 00:55 +0800, Stephen Warren wrote:
> On 12/19/2012 07:30 PM, Joseph Lo wrote:
> > The power up sequence is different on the cold boot CPU and the CPU
> > that resumed from the hotplug. For the cold boot CPU, it was been power
> > gated as default. To power up the cold boot CPU, the power should be
> > un-gated by un toggling the power gate register manually.
> > 
> > For the CPU that resumed from the hotplug, after un-halted the CPU. The
> > flow controller will un-gate the power of the CPU. No need to manually
> > control, just wait the power be resumed and continue the power up
> > sequence after the CPU power is ready.
> 
> I'd like Peter to review this too (CC'd). I guess it looks OK...
> 
> > diff --git a/arch/arm/mach-tegra/platsmp.c b/arch/arm/mach-tegra/platsmp.c
> 
> > @@ -72,7 +75,27 @@ static int tegra30_power_up_cpu(unsigned int cpu)
> >  	if (pwrgateid < 0)
> >  		return pwrgateid;
> >  
> > -	/* If this is the first boot, toggle powergates directly. */
> > +	/*
> > +	 * If the CPU had booted and went through here after CPU
> > +	 * had been already un-gated by flow controller. Wait
> > +	 * for confirmation that the CPU is powered then remove
> > +	 * the IO clamps. On the cold boot entry, do not wait.
> > +	 */
> 
> That comment is quite unclear. In particular the first sentence says "if
> something" rather than "if something then something". Can you please
> re-write it?
> 
OK.
> Also, it'd be good to explain (or at least briefly reference) the boot
> path for a cold-boot CPU and a warm boot CPU; I assume that somehow a
> cold boot CPU doesn't execute tegra_secondary_init(), but a warm boot
> CPU does?
> 
The boot path was the same. Both of the conditions were booted from
tegra_boot_secondary. Just the power sequence has been different. We add
the code to recover the power up sequence of warm boot CPU.

Thanks,
Joseph
Stephen Warren Dec. 21, 2012, 9:12 p.m. UTC | #3
On 12/21/2012 01:07 AM, Joseph Lo wrote:
> On Fri, 2012-12-21 at 00:55 +0800, Stephen Warren wrote:
>> On 12/19/2012 07:30 PM, Joseph Lo wrote:
>>> The power up sequence is different on the cold boot CPU and the CPU
>>> that resumed from the hotplug. For the cold boot CPU, it was been power
>>> gated as default. To power up the cold boot CPU, the power should be
>>> un-gated by un toggling the power gate register manually.
>>>
>>> For the CPU that resumed from the hotplug, after un-halted the CPU. The
>>> flow controller will un-gate the power of the CPU. No need to manually
>>> control, just wait the power be resumed and continue the power up
>>> sequence after the CPU power is ready.
...
>> Also, it'd be good to explain (or at least briefly reference) the boot
>> path for a cold-boot CPU and a warm boot CPU; I assume that somehow a
>> cold boot CPU doesn't execute tegra_secondary_init(), but a warm boot
>> CPU does?
>>
> The boot path was the same. Both of the conditions were booted from
> tegra_boot_secondary. Just the power sequence has been different. We add
> the code to recover the power up sequence of warm boot CPU.

Ah, I see - tegra20_power_up_cpu() runs first to boot the secondary CPU,
and tegra_secondary_init() runs on the secondary CPU during the boot
process, which then sets tegra_cpu_init_mask which influences what
tegra20_power_up_cpu() does next time around. It'd be useful to say that
in the patch description.
Peter De Schrijver Jan. 2, 2013, 12:15 p.m. UTC | #4
On Thu, Dec 20, 2012 at 03:30:03AM +0100, Joseph Lo wrote:
> The power up sequence is different on the cold boot CPU and the CPU
> that resumed from the hotplug. For the cold boot CPU, it was been power
> gated as default. To power up the cold boot CPU, the power should be
> un-gated by un toggling the power gate register manually.
> 
> For the CPU that resumed from the hotplug, after un-halted the CPU. The
> flow controller will un-gate the power of the CPU. No need to manually
> control, just wait the power be resumed and continue the power up
> sequence after the CPU power is ready.
> 
> Based on the work by:
> Varun Wadekar <vwadekar@nvidia.com>
> 
> Signed-off-by: Joseph Lo <josephl@nvidia.com>

Acked-by: Peter De Schrijver <pdeschrijver@nvidia.com>
diff mbox

Patch

diff --git a/arch/arm/mach-tegra/platsmp.c b/arch/arm/mach-tegra/platsmp.c
index 6cedb3e..9bd4898 100644
--- a/arch/arm/mach-tegra/platsmp.c
+++ b/arch/arm/mach-tegra/platsmp.c
@@ -23,6 +23,7 @@ 
 #include <asm/hardware/gic.h>
 #include <asm/mach-types.h>
 #include <asm/smp_scu.h>
+#include <asm/smp_plat.h>
 
 #include <mach/powergate.h>
 
@@ -36,6 +37,7 @@ 
 
 extern void tegra_secondary_startup(void);
 
+static cpumask_t tegra_cpu_init_mask;
 static void __iomem *scu_base = IO_ADDRESS(TEGRA_ARM_PERIF_BASE);
 
 #define EVP_CPU_RESET_VECTOR \
@@ -50,6 +52,7 @@  static void __cpuinit tegra_secondary_init(unsigned int cpu)
 	 */
 	gic_secondary_init(0);
 
+	cpumask_set_cpu(cpu, &tegra_cpu_init_mask);
 }
 
 static int tegra20_power_up_cpu(unsigned int cpu)
@@ -72,7 +75,27 @@  static int tegra30_power_up_cpu(unsigned int cpu)
 	if (pwrgateid < 0)
 		return pwrgateid;
 
-	/* If this is the first boot, toggle powergates directly. */
+	/*
+	 * If the CPU had booted and went through here after CPU
+	 * had been already un-gated by flow controller. Wait
+	 * for confirmation that the CPU is powered then remove
+	 * the IO clamps. On the cold boot entry, do not wait.
+	 */
+	if (cpumask_test_cpu(cpu, &tegra_cpu_init_mask)) {
+		timeout = jiffies + 5;
+		do {
+			if (!tegra_powergate_is_powered(pwrgateid))
+				goto remove_clamps;
+			udelay(10);
+		} while (time_before(jiffies, timeout));
+	}
+
+	/*
+	 * The power status of the cold boot CPU is power gating as
+	 * default. To power up the cold boot CPU, the power should
+	 * be un-gated by un-toggling the power gate register
+	 * manually.
+	 */
 	if (!tegra_powergate_is_powered(pwrgateid)) {
 		ret = tegra_powergate_power_on(pwrgateid);
 		if (ret)
@@ -87,6 +110,7 @@  static int tegra30_power_up_cpu(unsigned int cpu)
 		}
 	}
 
+remove_clamps:
 	/* CPU partition is powered. Enable the CPU clock. */
 	tegra_enable_cpu_clock(cpu);
 	udelay(10);
@@ -105,6 +129,8 @@  static int __cpuinit tegra_boot_secondary(unsigned int cpu, struct task_struct *
 {
 	int status;
 
+	cpu = cpu_logical_map(cpu);
+
 	/*
 	 * Force the CPU into reset. The CPU must remain in reset when the
 	 * flow controller state is cleared (which will cause the flow
@@ -165,6 +191,9 @@  static void __init tegra_smp_init_cpus(void)
 
 static void __init tegra_smp_prepare_cpus(unsigned int max_cpus)
 {
+	/* Always mark the boot CPU (CPU0) as initialized. */
+	cpumask_set_cpu(0, &tegra_cpu_init_mask);
+
 	tegra_cpu_reset_handler_init();
 	scu_enable(scu_base);
 }