ARM: tegra30: cpuidle: add powered-down state for CPU0

This is a power gating idle mode. It support power gating vdd_cpu rail
after all cpu cores in "powered-down" status. For Tegra30, the CPU0 can
enter this state only when all secondary CPU is offline. We need to take
care and make sure whole secondary CPUs were offline and checking the
CPU power gate status. After that, the CPU0 can go into "powered-down"
state safely. Then shut off the CPU rail.

Be aware of that, you may see the legacy power state "LP2" in the code
which is exactly the same meaning of "CPU power down".

Base on the work by:
Scott Williams <scwilliams@nvidia.com>
Signed-off-by: Joseph Lo <josephl@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>

arch/arm/mach-tegra/cpuidle-tegra30.c

@@ -32,6 +32,7 @@
 
 #include "pm.h"
 #include "sleep.h"
+#include "tegra_cpu_car.h"
 
 #ifdef CONFIG_PM_SLEEP
 static int tegra30_idle_lp2(struct cpuidle_device *dev,
@@ -67,6 +68,31 @@ static struct cpuidle_driver tegra_idle_driver = {
 static DEFINE_PER_CPU(struct cpuidle_device, tegra_idle_device);
 
 #ifdef CONFIG_PM_SLEEP
+static bool tegra30_cpu_cluster_power_down(struct cpuidle_device *dev,
+					   struct cpuidle_driver *drv,
+					   int index)
+{
+	struct cpuidle_state *state = &drv->states[index];
+	u32 cpu_on_time = state->exit_latency;
+	u32 cpu_off_time = state->target_residency - state->exit_latency;
+
+	/* All CPUs entering LP2 is not working.
+	 * Don't let CPU0 enter LP2 when any secondary CPU is online.
+	 */
+	if (num_online_cpus() > 1 || !tegra_cpu_rail_off_ready()) {
+		cpu_do_idle();
+		return false;
+	}
+
+	clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &dev->cpu);
+
+	tegra_idle_lp2_last(cpu_on_time, cpu_off_time);
+
+	clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &dev->cpu);
+
+	return true;
+}
+
 #ifdef CONFIG_SMP
 static bool tegra30_cpu_core_power_down(struct cpuidle_device *dev,
 					struct cpuidle_driver *drv,
@@ -101,16 +127,22 @@ static int __cpuinit tegra30_idle_lp2(struct cpuidle_device *dev,
 {
 	u32 cpu = is_smp() ? cpu_logical_map(dev->cpu) : dev->cpu;
 	bool entered_lp2 = false;
+	bool last_cpu;
 
 	local_fiq_disable();
 
-	tegra_set_cpu_in_lp2(cpu);
+	last_cpu = tegra_set_cpu_in_lp2(cpu);
 	cpu_pm_enter();
 
-	if (cpu == 0)
-		cpu_do_idle();
-	else
+	if (cpu == 0) {
+		if (last_cpu)
+			entered_lp2 = tegra30_cpu_cluster_power_down(dev, drv,
+								     index);
+		else
+			cpu_do_idle();
+	} else {
 		entered_lp2 = tegra30_cpu_core_power_down(dev, drv, index);
+	}
 
 	cpu_pm_exit();
 	tegra_clear_cpu_in_lp2(cpu);
@@ -130,6 +162,10 @@ int __init tegra30_cpuidle_init(void)
 	struct cpuidle_device *dev;
 	struct cpuidle_driver *drv = &tegra_idle_driver;
 
+#ifdef CONFIG_PM_SLEEP
+	tegra_tear_down_cpu = tegra30_tear_down_cpu;
+#endif
+
 	ret = cpuidle_register_driver(&tegra_idle_driver);
 	if (ret) {
 		pr_err("CPUidle driver registration failed\n");

arch/arm/mach-tegra/pm.c

@@ -20,13 +20,36 @@
 #include <linux/spinlock.h>
 #include <linux/io.h>
 #include <linux/cpumask.h>
+#include <linux/delay.h>
+#include <linux/cpu_pm.h>
+#include <linux/clk.h>
+#include <linux/err.h>
+
+#include <asm/smp_plat.h>
+#include <asm/cacheflush.h>
+#include <asm/suspend.h>
+#include <asm/idmap.h>
+#include <asm/proc-fns.h>
+#include <asm/tlbflush.h>
 
 #include "iomap.h"
 #include "reset.h"
+#include "flowctrl.h"
+#include "sleep.h"
+#include "tegra_cpu_car.h"
+
+#define TEGRA_POWER_CPU_PWRREQ_OE	(1 << 16)  /* CPU pwr req enable */
+
+#define PMC_CTRL		0x0
+#define PMC_CPUPWRGOOD_TIMER	0xc8
+#define PMC_CPUPWROFF_TIMER	0xcc
 
 #ifdef CONFIG_PM_SLEEP
 static unsigned int g_diag_reg;
 static DEFINE_SPINLOCK(tegra_lp2_lock);
+static void __iomem *pmc = IO_ADDRESS(TEGRA_PMC_BASE);
+static struct clk *tegra_pclk;
+void (*tegra_tear_down_cpu)(void);
 
 void save_cpu_arch_register(void)
 {
@@ -42,6 +65,89 @@ void restore_cpu_arch_register(void)
 	return;
 }
 
+static void set_power_timers(unsigned long us_on, unsigned long us_off)
+{
+	unsigned long long ticks;
+	unsigned long long pclk;
+	unsigned long rate;
+	static unsigned long tegra_last_pclk;
+
+	if (tegra_pclk == NULL) {
+		tegra_pclk = clk_get_sys(NULL, "pclk");
+		WARN_ON(IS_ERR(tegra_pclk));
+	}
+
+	rate = clk_get_rate(tegra_pclk);
+
+	if (WARN_ON_ONCE(rate <= 0))
+		pclk = 100000000;
+	else
+		pclk = rate;
+
+	if ((rate != tegra_last_pclk)) {
+		ticks = (us_on * pclk) + 999999ull;
+		do_div(ticks, 1000000);
+		writel((unsigned long)ticks, pmc + PMC_CPUPWRGOOD_TIMER);
+
+		ticks = (us_off * pclk) + 999999ull;
+		do_div(ticks, 1000000);
+		writel((unsigned long)ticks, pmc + PMC_CPUPWROFF_TIMER);
+		wmb();
+	}
+	tegra_last_pclk = pclk;
+}
+
+/*
+ * restore_cpu_complex
+ *
+ * restores cpu clock setting, clears flow controller
+ *
+ * Always called on CPU 0.
+ */
+static void restore_cpu_complex(void)
+{
+	int cpu = smp_processor_id();
+
+	BUG_ON(cpu != 0);
+
+#ifdef CONFIG_SMP
+	cpu = cpu_logical_map(cpu);
+#endif
+
+	/* Restore the CPU clock settings */
+	tegra_cpu_clock_resume();
+
+	flowctrl_cpu_suspend_exit(cpu);
+
+	restore_cpu_arch_register();
+}
+
+/*
+ * suspend_cpu_complex
+ *
+ * saves pll state for use by restart_plls, prepares flow controller for
+ * transition to suspend state
+ *
+ * Must always be called on cpu 0.
+ */
+static void suspend_cpu_complex(void)
+{
+	int cpu = smp_processor_id();
+
+	BUG_ON(cpu != 0);
+
+#ifdef CONFIG_SMP
+	cpu = cpu_logical_map(cpu);
+#endif
+
+	/* Save the CPU clock settings */
+	tegra_cpu_clock_suspend();
+
+	flowctrl_cpu_suspend_enter(cpu);
+
+	save_cpu_arch_register();
+}
+
 void __cpuinit tegra_clear_cpu_in_lp2(int phy_cpu_id)
 {
 	u32 *cpu_in_lp2 = tegra_cpu_lp2_mask;
@@ -71,4 +177,42 @@ bool __cpuinit tegra_set_cpu_in_lp2(int phy_cpu_id)
 	spin_unlock(&tegra_lp2_lock);
 	return last_cpu;
 }
+
+static int tegra_sleep_cpu(unsigned long v2p)
+{
+	/* Switch to the identity mapping. */
+	cpu_switch_mm(idmap_pgd, &init_mm);
+
+	/* Flush the TLB. */
+	local_flush_tlb_all();
+
+	tegra_sleep_cpu_finish(v2p);
+
+	/* should never here */
+	BUG();
+
+	return 0;
+}
+
+void tegra_idle_lp2_last(u32 cpu_on_time, u32 cpu_off_time)
+{
+	u32 mode;
+
+	/* Only the last cpu down does the final suspend steps */
+	mode = readl(pmc + PMC_CTRL);
+	mode |= TEGRA_POWER_CPU_PWRREQ_OE;
+	writel(mode, pmc + PMC_CTRL);
+
+	set_power_timers(cpu_on_time, cpu_off_time);
+
+	cpu_cluster_pm_enter();
+	suspend_cpu_complex();
+	outer_disable();
+
+	cpu_suspend(PHYS_OFFSET - PAGE_OFFSET, &tegra_sleep_cpu);
+
+	outer_resume();
+	restore_cpu_complex();
+	cpu_cluster_pm_exit();
+}
 #endif

arch/arm/mach-tegra/pm.h

@@ -27,4 +27,7 @@ void restore_cpu_arch_register(void);
 void tegra_clear_cpu_in_lp2(int phy_cpu_id);
 bool tegra_set_cpu_in_lp2(int phy_cpu_id);
 
+void tegra_idle_lp2_last(u32 cpu_on_time, u32 cpu_off_time);
+extern void (*tegra_tear_down_cpu)(void);
+
 #endif /* _MACH_TEGRA_PM_H_ */

arch/arm/mach-tegra/sleep-tegra30.S

@@ -124,4 +124,48 @@ ENTRY(tegra30_sleep_cpu_secondary_finish)
 	mov	r0, #1                          @ never return here
 	mov	pc, r7
 ENDPROC(tegra30_sleep_cpu_secondary_finish)
+
+/*
+ * tegra30_tear_down_cpu
+ *
+ * Switches the CPU to enter sleep.
+ */
+ENTRY(tegra30_tear_down_cpu)
+	mov32	r6, TEGRA_FLOW_CTRL_BASE
+
+	b	tegra30_enter_sleep
+ENDPROC(tegra30_tear_down_cpu)
+
+/*
+ * tegra30_enter_sleep
+ *
+ * uses flow controller to enter sleep state
+ * executes from IRAM with SDRAM in selfrefresh when target state is LP0 or LP1
+ * executes from SDRAM with target state is LP2
+ * r6 = TEGRA_FLOW_CTRL_BASE
+ */
+tegra30_enter_sleep:
+	cpu_id	r1
+
+	cpu_to_csr_reg	r2, r1
+	ldr	r0, [r6, r2]
+	orr	r0, r0, #FLOW_CTRL_CSR_INTR_FLAG | FLOW_CTRL_CSR_EVENT_FLAG
+	orr	r0, r0, #FLOW_CTRL_CSR_ENABLE
+	str	r0, [r6, r2]
+
+	mov	r0, #FLOW_CTRL_WAIT_FOR_INTERRUPT
+	orr	r0, r0, #FLOW_CTRL_HALT_CPU_IRQ | FLOW_CTRL_HALT_CPU_FIQ
+	cpu_to_halt_reg r2, r1
+	str	r0, [r6, r2]
+	dsb
+	ldr	r0, [r6, r2] /* memory barrier */
+
+halted:
+	isb
+	dsb
+	wfi	/* CPU should be power gated here */
+
+	/* !!!FIXME!!! Implement halt failure handler */
+	b	halted
+
 #endif

arch/arm/mach-tegra/sleep.S

@@ -25,6 +25,7 @@
 #include <linux/linkage.h>
 
 #include <asm/assembler.h>
+#include <asm/cache.h>
 #include <asm/cp15.h>
 
 #include "iomap.h"
@@ -59,4 +60,45 @@ ENTRY(tegra_disable_clean_inv_dcache)
 	ldmfd	sp!, {r0, r4-r5, r7, r9-r11, pc}
 ENDPROC(tegra_disable_clean_inv_dcache)
 
+/*
+ * tegra_sleep_cpu_finish(unsigned long v2p)
+ *
+ * enters suspend in LP2 by turning off the mmu and jumping to
+ * tegra?_tear_down_cpu
+ */
+ENTRY(tegra_sleep_cpu_finish)
+	/* Flush and disable the L1 data cache */
+	bl	tegra_disable_clean_inv_dcache
+
+	mov32	r6, tegra_tear_down_cpu
+	ldr	r1, [r6]
+	add	r1, r1, r0
+
+	mov32	r3, tegra_shut_off_mmu
+	add	r3, r3, r0
+	mov	r0, r1
+
+	mov	pc, r3
+ENDPROC(tegra_sleep_cpu_finish)
+
+/*
+ * tegra_shut_off_mmu
+ *
+ * r0 = physical address to jump to with mmu off
+ *
+ * called with VA=PA mapping
+ * turns off MMU, icache, dcache and branch prediction
+ */
+	.align	L1_CACHE_SHIFT
+	.pushsection	.idmap.text, "ax"
+ENTRY(tegra_shut_off_mmu)
+	mrc	p15, 0, r3, c1, c0, 0
+	movw	r2, #CR_I | CR_Z | CR_C | CR_M
+	bic	r3, r3, r2
+	dsb
+	mcr	p15, 0, r3, c1, c0, 0
+	isb
+	mov	pc, r0
+ENDPROC(tegra_shut_off_mmu)
+	.popsection
 #endif

arch/arm/mach-tegra/sleep.h

@@ -73,6 +73,7 @@
 .endm
 #else
 void tegra_resume(void);
+int tegra_sleep_cpu_finish(unsigned long);
 
 #ifdef CONFIG_HOTPLUG_CPU
 void tegra20_hotplug_init(void);
@@ -83,6 +84,7 @@ static inline void tegra30_hotplug_init(void) {}
 #endif
 
 int tegra30_sleep_cpu_secondary_finish(unsigned long);
+void tegra30_tear_down_cpu(void);
 
 #endif
 #endif