ARM: rockchip: rename osc_switch_to_32k variable

The variable name is misleading, as the deep suspend mode always switches the main supplying clock to the 32kHz source. Additionally the main oscillator remain running in some cases, which this var indicates.

So rename it to osc_disable to clarity.

Signed-off-by Heiko Stuebner <[email protected]>

Currently the stabilization thresholds for the oscillator and external pmu are statically set to 30ms based on a 32kHz clock rate. This leaves out the case when we don't switch to the 32kHz clock when only entering the shallow suspend mode where the logic keeps running.

So, set the correct threshold after we have determined if we switch to the 32kHz clock or stay with the 24MHz one. Also set the oscillator- stabilization to 0 if it is kept running during suspend, as it of course does not need to stabilize then.

Reported-by Chris Zhong <[email protected]>

Signed-off-by Heiko Stuebner <[email protected]>

---
changes since v2:
- describe 32kHz vs 24MHz
- don't wait for oscillator stabilization if it's still running, as explained by Chris Zhong in v2
changes since v1:
- 24MHz oriented threshold is only needed in shallow suspend, the deep suspend always switches to 32kHz and only leaves the 24MHz oscillator running if needed for stuff like usb wakeup

  arch/arm/mach-rockchip/pm.c | 22 +++++++++++++++++++---
  arch/arm/mach-rockchip/pm.h |  4 ----
  2 files changed, 19 insertions(+), 7 deletions(-)

arch/arm/mach-rockchip/pm.c +19 -3

index 2ca1170..c11a30b 100644
@@ -145,6 +145,19 @@ static void rk3288_slp_mode_set(int level)
 
  		mode_set1 |= BIT(PMU_CLR_ALIVE) | BIT(PMU_CLR_BUS) |
  			     BIT(PMU_CLR_PERI) | BIT(PMU_CLR_DMA);
+
+		/*
+		 * In deep suspend we use PMU_PMU_USE_LF to let the rk3288
+		 * switch its main clock supply to the alternative 32kHz
+		 * source. Therefore set 30ms on a 32kHz clock for pmic
+		 * stabilization. Similar 30ms on 24MHz for the other
+		 * mode below.
+		 */
+		regmap_write(pmu_regmap, RK3288_PMU_STABL_CNT, 32 * 30);
+
+		/* only wait for stabilization, if we turned the osc off */
+		regmap_write(pmu_regmap, RK3288_PMU_OSC_CNT,
+					 osc_disable ? 32 * 30 : 0);
  	} else {
  		/*
  		 * arm off, logic normal

@@ -152,6 +165,12 @@ static void rk3288_slp_mode_set(int level)
  		 * wakeup will be error
  		 */
  		mode_set |= BIT(PMU_CLK_CORE_SRC_GATE_EN);
+
+		/* 30ms on a 24MHz clock for pmic stabilization */
+		regmap_write(pmu_regmap, RK3288_PMU_STABL_CNT, 24000 * 30);
+
+		/* oscillator is still running, so no need to wait */
+		regmap_write(pmu_regmap, RK3288_PMU_OSC_CNT, 0);
  	}
 
  	regmap_write(pmu_regmap, RK3288_PMU_PWRMODE_CON, mode_set);

@@ -262,9 +281,6 @@ static int rk3288_suspend_init(struct device_node *np)
  	memcpy(rk3288_bootram_base, rockchip_slp_cpu_resume,
  	       rk3288_bootram_sz);
 
-	regmap_write(pmu_regmap, RK3288_PMU_OSC_CNT, OSC_STABL_CNT_THRESH);
-	regmap_write(pmu_regmap, RK3288_PMU_STABL_CNT, PMU_STABL_CNT_THRESH);
-
  	return 0;
  }
 

arch/arm/mach-rockchip/pm.h +0 -0

index b6494c2..8a55ee2 100644

Currently the stabilization thresholds for the oscillator and external pmu are statically set to 30ms based on a 32kHz clock rate. This leaves out the case when we don't switch to the 32kHz clock when only entering the shallow suspend mode where the logic keeps running.

So, set the correct threshold after we have determined if we switch to the 32kHz clock or stay with the 24MHz one. Also set the oscillator- stabilization to 0 if it is kept running during suspend, as it of course does not need to stabilize then.

Reported-by Chris Zhong <[email protected]>

Signed-off-by Heiko Stuebner <[email protected]>

---
changes since v2:
- describe 32kHz vs 24MHz
- don't wait for oscillator stabilization if it's still running, as explained by Chris Zhong in v2
changes since v1:
- 24MHz oriented threshold is only needed in shallow suspend, the deep suspend always switches to 32kHz and only leaves the 24MHz oscillator running if needed for stuff like usb wakeup

 arch/arm/mach-rockchip/pm.c | 22 +++++++++++++++++++---
 arch/arm/mach-rockchip/pm.h |  4 ----
 2 files changed, 19 insertions(+), 7 deletions(-)

arch/arm/mach-rockchip/pm.c +0 -0

index 2ca1170..c11a30b 100644
@@ -145,6 +145,19 @@ static void rk3288_slp_mode_set(int level)
 
                mode_set1 |= BIT(PMU_CLR_ALIVE) | BIT(PMU_CLR_BUS) |

The variable name is misleading, as the deep suspend mode always switches the main supplying clock to the 32kHz source. Additionally the main oscillator remain running in some cases, which this var indicates.

So rename it to osc_disable to clarity.

Signed-off-by Heiko Stuebner <[email protected]>

---
  arch/arm/mach-rockchip/pm.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)

arch/arm/mach-rockchip/pm.c +1 -1

index 892bace..2ca1170 100644
@@ -96,7 +96,7 @@ static bool rk3288_slp_disable_osc(void)
  static void rk3288_slp_mode_set(int level)
  {
  	u32 mode_set, mode_set1;
-	bool osc_switch_to_32k = rk3288_slp_disable_osc();
+	bool osc_disable = rk3288_slp_disable_osc();
 
  	regmap_read(sgrf_regmap, RK3288_SGRF_CPU_CON0, &rk3288_sgrf_cpu_con0);
  	regmap_read(sgrf_regmap, RK3288_SGRF_SOC_CON0, &rk3288_sgrf_soc_con0);

The variable name is misleading, as the deep suspend mode always switches the main supplying clock to the 32kHz source. Additionally the main oscillator remain running in some cases, which this var indicates.

So rename it to osc_disable to clarity.

Signed-off-by Heiko Stuebner <[email protected]>

---
 arch/arm/mach-rockchip/pm.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

PMU_GPIOINT_WAKEUP_EN seems needed when entering the shallow suspend (with logic staying on) but does not seem to be needed for the deep suspend for unknown reasons.
Testing revealed that this setting really is necessary to reliably resume the veyron devices from suspend.

Reported-by Chris Zhong <[email protected]>

Signed-off-by Heiko Stuebner <[email protected]>

---
  arch/arm/mach-rockchip/pm.c | 9 ++++++---
  arch/arm/mach-rockchip/pm.h | 1 +
  2 files changed, 7 insertions(+), 3 deletions(-)

arch/arm/mach-rockchip/pm.c +6 -3

index c11a30b..156cd23 100644
@@ -123,9 +123,6 @@ static void rk3288_slp_mode_set(int level)
  	regmap_write(sgrf_regmap, RK3288_SGRF_FAST_BOOT_ADDR,
  		     rk3288_bootram_phy);
 
-	regmap_write(pmu_regmap, RK3288_PMU_WAKEUP_CFG1,
-		     PMU_ARMINT_WAKEUP_EN);
-
  	mode_set = BIT(PMU_GLOBAL_INT_DISABLE) | BIT(PMU_L2FLUSH_EN) |
  		   BIT(PMU_SREF0_ENTER_EN) | BIT(PMU_SREF1_ENTER_EN) |
  		   BIT(PMU_DDR0_GATING_EN) | BIT(PMU_DDR1_GATING_EN) |

@@ -146,6 +143,9 @@ static void rk3288_slp_mode_set(int level)
  		mode_set1 |= BIT(PMU_CLR_ALIVE) | BIT(PMU_CLR_BUS) |
  			     BIT(PMU_CLR_PERI) | BIT(PMU_CLR_DMA);
 
+		regmap_write(pmu_regmap, RK3288_PMU_WAKEUP_CFG1,
+			     PMU_ARMINT_WAKEUP_EN);
+
  		/*
  		 * In deep suspend we use PMU_PMU_USE_LF to let the rk3288
  		 * switch its main clock supply to the alternative 32kHz

@@ -166,6 +166,9 @@ static void rk3288_slp_mode_set(int level)
  		 */
  		mode_set |= BIT(PMU_CLK_CORE_SRC_GATE_EN);
 
+		regmap_write(pmu_regmap, RK3288_PMU_WAKEUP_CFG1,
+			     PMU_ARMINT_WAKEUP_EN | PMU_GPIOINT_WAKEUP_EN);
+
  		/* 30ms on a 24MHz clock for pmic stabilization */
  		regmap_write(pmu_regmap, RK3288_PMU_STABL_CNT, 24000 * 30);
 

arch/arm/mach-rockchip/pm.h +0 -0

index 8a55ee2..b5af26f 100644

PMU_GPIOINT_WAKEUP_EN seems needed when entering the shallow suspend (with logic staying on) but does not seem to be needed for the deep suspend for unknown reasons.
Testing revealed that this setting really is necessary to reliably resume the veyron devices from suspend.

Reported-by Chris Zhong <[email protected]>

Signed-off-by Heiko Stuebner <[email protected]>

---
 arch/arm/mach-rockchip/pm.c | 9 ++++++---
 arch/arm/mach-rockchip/pm.h | 1 +
 2 files changed, 7 insertions(+), 3 deletions(-)

Heiko,

Quoted 10 lines · Heiko St?bner

PMU_GPIOINT_WAKEUP_EN seems needed when entering the shallow suspend (with logic staying on) but does not seem to be needed for the deep suspend for unknown reasons.
Testing revealed that this setting really is necessary to reliably resume the veyron devices from suspend.

Reported-by Chris Zhong <[email protected]>

Signed-off-by Heiko Stuebner <[email protected]>

---

Quoted 3 lines

 arch/arm/mach-rockchip/pm.c | 9 ++++++---
 arch/arm/mach-rockchip/pm.h | 1 +
 2 files changed, 7 insertions(+), 3 deletions(-)

Very odd, but OK by me...

Hopefully we'll eventually end up being
able to use deep suspend on mainline. That'd be nice, wouldn't it?