[U-Boot] [PATCH v3 0/9] arm: exynos: Fix reboot on Odroid HC1
Hi,
Changes since v2 ================ 1. Add Lukasz review tags. 2. Patch 7: Return on error, as suggested by Simon. 2. Patch 3: Use IS_ENABLED() to run revision detection only once - either during late display board or misc_init_r.
Changes since v1 ================ 1. Move fixes to beginning of patchset. 2. Patch 3: Rework the idea - split revision detection to be executed later. 3. Patch 4: New patch. 4. Patch 6: Apply Simon's comments. 5. Patch 6: Do not delay when changing voltage if regulator is disabled. 6. Patch 6: Do not delay when disabling the regulator.
Description =========== Odroid HC1 does not reboot properly (at least from SD card but I do not expect difference on eMMC), if LDO4/VDD_ADC was turned off by Linux kernel. This condition happens so far always, because Linux kernel did not enable ADC on Odroid HC1, therefore the VDD_ADC regulator was turned off as unused.
The issue is in detection of revision which later is used to load proper DTB.
The revision is obtained by ADC read of a voltage depending on VDD_ADC. Therefore: 1. VDD_ADC has to be turned on (but board detection happens before power is initialized), 2. Turning VDD_ADC should wait with ramp delay, 3. Reading the value from ADC should wait for it to stabilize.
Tested on Odroid XU3-Lite and Odroid HC1.
Commends and testing are welcomed.
Best regards, Krzysztof
Krzysztof Kozlowski (9): adc: exynos-adc: Fix wrong bit operation used to stop the ADC power: regulator: s2mps11: Fix step for LDO27 and LDO35 arm: exynos: Detect revision later, when all resources are ready arm: exynos: odroid-xu3: Display info late to have proper type arm: exynos: Wait till ADC stabilizes before checking Odroid HC1 revision regulator: Add support for ramp delay power: regulator: s2mps11: Add enable delay arm: dts: exynos: Add supply for ADC block to Odroid XU3 family arm: dts: exynos: Add ramp delay property to LDO regulators to Odroid XU3 family
arch/arm/dts/exynos5422-odroidxu3.dts | 20 +++++++ board/samsung/common/board.c | 24 ++++++++- board/samsung/common/exynos5-dt-types.c | 54 +++++++++++++++++-- board/samsung/odroid/odroid.c | 8 +++ configs/odroid-xu3_defconfig | 2 + .../regulator/regulator.txt | 2 + drivers/adc/exynos-adc.c | 2 +- drivers/power/regulator/regulator-uclass.c | 47 +++++++++++++++- drivers/power/regulator/s2mps11_regulator.c | 15 +++++- include/power/regulator.h | 2 + include/samsung/misc.h | 1 + 11 files changed, 167 insertions(+), 10 deletions(-)
When stopping the ADC_V2_CON1_STC_EN should be cleared.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org Reviewed-by: Lukasz Majewski lukma@denx.de --- drivers/adc/exynos-adc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/adc/exynos-adc.c b/drivers/adc/exynos-adc.c index d33e3d632afc..12c49fc8cefb 100644 --- a/drivers/adc/exynos-adc.c +++ b/drivers/adc/exynos-adc.c @@ -62,7 +62,7 @@ int exynos_adc_stop(struct udevice *dev)
/* Stop conversion */ cfg = readl(®s->con1); - cfg |= ~ADC_V2_CON1_STC_EN; + cfg &= ~ADC_V2_CON1_STC_EN;
writel(cfg, ®s->con1);
LDO27 and LDO35 have 25 mV step, not 50 mV.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org Reviewed-by: Lukasz Majewski lukma@denx.de --- drivers/power/regulator/s2mps11_regulator.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/drivers/power/regulator/s2mps11_regulator.c b/drivers/power/regulator/s2mps11_regulator.c index ced504eb1476..723d27f67c9a 100644 --- a/drivers/power/regulator/s2mps11_regulator.c +++ b/drivers/power/regulator/s2mps11_regulator.c @@ -346,6 +346,8 @@ static int s2mps11_ldo_hex2volt(int ldo, int hex) case 11: case 22: case 23: + case 27: + case 35: uV = hex * S2MPS11_LDO_STEP + S2MPS11_LDO_UV_MIN; break; default: @@ -366,6 +368,8 @@ static int s2mps11_ldo_volt2hex(int ldo, int uV) case 11: case 22: case 23: + case 27: + case 35: hex = (uV - S2MPS11_LDO_UV_MIN) / S2MPS11_LDO_STEP; break; default:
Detection of board revision is done early - before power setup. In case of Odroid XU3/XU4/HC1 family, the detection is done using ADC which is supplied by LDO4/VDD_ADC regulator. This regulator could be turned off (e.g. by kernel before reboot). If ADC is used early, the regulators are not yet available and the detection won't work.
Split the revision detection out of set_board_type() into separate function called later - either when displaying board info (in late mode) or during misc_init_r. The idea is that set_board_type() will be called early so its method of detection are limited to flattened device tree (exynos5-dt-types.c for Exynos5) or GPIO (odroid.c for Exynos4412). The newly added set_board_revision() can be called only later, when resources like regulator are available.
This is necessary to fix the detection of Odroid HC1 after reboot, if kernel turned off the LDO4 regulator.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org --- board/samsung/common/board.c | 22 +++++++++++++++++++++- board/samsung/common/exynos5-dt-types.c | 16 +++++++++++++--- board/samsung/odroid/odroid.c | 8 ++++++++ include/samsung/misc.h | 1 + 4 files changed, 43 insertions(+), 4 deletions(-)
diff --git a/board/samsung/common/board.c b/board/samsung/common/board.c index 96228a86a117..ef2204742e1d 100644 --- a/board/samsung/common/board.c +++ b/board/samsung/common/board.c @@ -253,7 +253,18 @@ int board_eth_init(bd_t *bis) int checkboard(void) { if (IS_ENABLED(CONFIG_BOARD_TYPES)) { - const char *board_info = get_board_type(); + const char *board_info; + + if (IS_ENABLED(CONFIG_DISPLAY_BOARDINFO_LATE)) { + /* + * Printing type requires having revision, although + * this will succeed only if done late. + * Otherwise revision will be set in misc_init_r(). + */ + set_board_revision(); + } + + board_info = get_board_type();
if (board_info) printf("Type: %s\n", board_info); @@ -287,6 +298,15 @@ int board_late_init(void) #ifdef CONFIG_MISC_INIT_R int misc_init_r(void) { + if (!IS_ENABLED(CONFIG_DISPLAY_BOARDINFO_LATE)) { + /* + * If revision was not set by late display boardinfo, + * set it here. At this point regulators should be already + * available. + */ + set_board_revision(); + } + #ifdef CONFIG_ENV_VARS_UBOOT_RUNTIME_CONFIG set_board_info(); #endif diff --git a/board/samsung/common/exynos5-dt-types.c b/board/samsung/common/exynos5-dt-types.c index 7a86e9187768..7c1271d6547a 100644 --- a/board/samsung/common/exynos5-dt-types.c +++ b/board/samsung/common/exynos5-dt-types.c @@ -192,8 +192,11 @@ const char *get_board_type(void)
/** * set_board_type() - set board type in gd->board_type. - * As default type set EXYNOS5_BOARD_GENERIC, if detect Odroid, - * then set its proper type. + * As default type set EXYNOS5_BOARD_GENERIC. If Odroid is detected, + * set its proper type based on device tree. + * + * This might be called early when some more specific ways to detect revision + * are not yet available. */ void set_board_type(void) { @@ -211,8 +214,15 @@ void set_board_type(void) gd->board_type = of_match->data; break; } +}
- /* If Odroid, then check its revision */ +/** + * set_board_revision() - set detailed board type in gd->board_type. + * Should be called when resources (e.g. regulators) are available + * so ADC can be used to detect the specific revision of a board. + */ +void set_board_revision(void) +{ if (board_is_odroidxu3()) gd->board_type = odroid_get_board_type(); } diff --git a/board/samsung/odroid/odroid.c b/board/samsung/odroid/odroid.c index 552333fe869d..4be8cc9826c3 100644 --- a/board/samsung/odroid/odroid.c +++ b/board/samsung/odroid/odroid.c @@ -54,6 +54,14 @@ void set_board_type(void) gd->board_type = ODROID_TYPE_U3; }
+void set_board_revision(void) +{ + /* + * Revision already set by set_board_type() because it can be + * executed early. + */ +} + const char *get_board_type(void) { const char *board_type[] = {"u3", "x2"}; diff --git a/include/samsung/misc.h b/include/samsung/misc.h index 017560c25662..4ff28a1df0e8 100644 --- a/include/samsung/misc.h +++ b/include/samsung/misc.h @@ -33,6 +33,7 @@ char *get_dfu_alt_system(char *interface, char *devstr); char *get_dfu_alt_boot(char *interface, char *devstr); #endif void set_board_type(void); +void set_board_revision(void); const char *get_board_type(void);
#endif /* __SAMSUNG_MISC_COMMON_H__ */
On Sat, 16 Feb 2019 10:45:42 +0100 Krzysztof Kozlowski krzk@kernel.org wrote:
Detection of board revision is done early - before power setup. In case of Odroid XU3/XU4/HC1 family, the detection is done using ADC which is supplied by LDO4/VDD_ADC regulator. This regulator could be turned off (e.g. by kernel before reboot). If ADC is used early, the regulators are not yet available and the detection won't work.
Split the revision detection out of set_board_type() into separate function called later - either when displaying board info (in late mode) or during misc_init_r. The idea is that set_board_type() will be called early so its method of detection are limited to flattened device tree (exynos5-dt-types.c for Exynos5) or GPIO (odroid.c for Exynos4412). The newly added set_board_revision() can be called only later, when resources like regulator are available.
This is necessary to fix the detection of Odroid HC1 after reboot, if kernel turned off the LDO4 regulator.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org
board/samsung/common/board.c | 22 +++++++++++++++++++++- board/samsung/common/exynos5-dt-types.c | 16 +++++++++++++--- board/samsung/odroid/odroid.c | 8 ++++++++ include/samsung/misc.h | 1 + 4 files changed, 43 insertions(+), 4 deletions(-)
diff --git a/board/samsung/common/board.c b/board/samsung/common/board.c index 96228a86a117..ef2204742e1d 100644 --- a/board/samsung/common/board.c +++ b/board/samsung/common/board.c @@ -253,7 +253,18 @@ int board_eth_init(bd_t *bis) int checkboard(void) { if (IS_ENABLED(CONFIG_BOARD_TYPES)) {
const char *board_info = get_board_type();
const char *board_info;if (IS_ENABLED(CONFIG_DISPLAY_BOARDINFO_LATE)) {/** Printing type requires having revision,although
* this will succeed only if done late.* Otherwise revision will be set inmisc_init_r().
*/set_board_revision();}board_info = get_board_type();if (board_info) printf("Type: %s\n", board_info);
@@ -287,6 +298,15 @@ int board_late_init(void) #ifdef CONFIG_MISC_INIT_R int misc_init_r(void) {
- if (!IS_ENABLED(CONFIG_DISPLAY_BOARDINFO_LATE)) {
/** If revision was not set by late display boardinfo,* set it here. At this point regulators should bealready
* available.*/set_board_revision();- }
#ifdef CONFIG_ENV_VARS_UBOOT_RUNTIME_CONFIG set_board_info(); #endif diff --git a/board/samsung/common/exynos5-dt-types.c b/board/samsung/common/exynos5-dt-types.c index 7a86e9187768..7c1271d6547a 100644 --- a/board/samsung/common/exynos5-dt-types.c +++ b/board/samsung/common/exynos5-dt-types.c @@ -192,8 +192,11 @@ const char *get_board_type(void) /**
- set_board_type() - set board type in gd->board_type.
- As default type set EXYNOS5_BOARD_GENERIC, if detect Odroid,
- then set its proper type.
- As default type set EXYNOS5_BOARD_GENERIC. If Odroid is detected,
- set its proper type based on device tree.
- This might be called early when some more specific ways to detect
revision
*/
- are not yet available.
void set_board_type(void) { @@ -211,8 +214,15 @@ void set_board_type(void) gd->board_type = of_match->data; break; } +}
- /* If Odroid, then check its revision */
+/**
- set_board_revision() - set detailed board type in gd->board_type.
- Should be called when resources (e.g. regulators) are available
- so ADC can be used to detect the specific revision of a board.
- */
+void set_board_revision(void) +{ if (board_is_odroidxu3()) gd->board_type = odroid_get_board_type(); } diff --git a/board/samsung/odroid/odroid.c b/board/samsung/odroid/odroid.c index 552333fe869d..4be8cc9826c3 100644 --- a/board/samsung/odroid/odroid.c +++ b/board/samsung/odroid/odroid.c @@ -54,6 +54,14 @@ void set_board_type(void) gd->board_type = ODROID_TYPE_U3; }
+void set_board_revision(void) +{
- /*
* Revision already set by set_board_type() because it can be* executed early.*/+}
const char *get_board_type(void) { const char *board_type[] = {"u3", "x2"}; diff --git a/include/samsung/misc.h b/include/samsung/misc.h index 017560c25662..4ff28a1df0e8 100644 --- a/include/samsung/misc.h +++ b/include/samsung/misc.h @@ -33,6 +33,7 @@ char *get_dfu_alt_system(char *interface, char *devstr); char *get_dfu_alt_boot(char *interface, char *devstr); #endif void set_board_type(void); +void set_board_revision(void); const char *get_board_type(void);
#endif /* __SAMSUNG_MISC_COMMON_H__ */
Reviewed-by: Lukasz Majewski lukma@denx.de
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma@denx.de
Printing the "Type" of board requires proper detection of revision which can happen only late because regulators are needed.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org --- board/samsung/common/board.c | 2 +- configs/odroid-xu3_defconfig | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-)
diff --git a/board/samsung/common/board.c b/board/samsung/common/board.c index ef2204742e1d..4ffbd7254205 100644 --- a/board/samsung/common/board.c +++ b/board/samsung/common/board.c @@ -249,7 +249,7 @@ int board_eth_init(bd_t *bis) return 0; }
-#ifdef CONFIG_DISPLAY_BOARDINFO +#if defined(CONFIG_DISPLAY_BOARDINFO) || defined(CONFIG_DISPLAY_BOARDINFO_LATE) int checkboard(void) { if (IS_ENABLED(CONFIG_BOARD_TYPES)) { diff --git a/configs/odroid-xu3_defconfig b/configs/odroid-xu3_defconfig index f6f05b294833..57aca5e015fc 100644 --- a/configs/odroid-xu3_defconfig +++ b/configs/odroid-xu3_defconfig @@ -11,6 +11,8 @@ CONFIG_FIT=y CONFIG_FIT_BEST_MATCH=y CONFIG_SILENT_CONSOLE=y CONFIG_CONSOLE_MUX=y +# CONFIG_DISPLAY_BOARDINFO is not set +CONFIG_DISPLAY_BOARDINFO_LATE=y CONFIG_MISC_INIT_R=y CONFIG_SYS_PROMPT="ODROID-XU3 # " CONFIG_CMD_THOR_DOWNLOAD=y
Fix detection of Odroid HC1 (Exynos5422) after reboot if kernel disabled the LDO4/VDD_ADC regulator.
The LDO4 supplies both ADC block and the ADC input AIN9. Voltage on AIN9 will rise slowly, so be patient and wait for it to stabilize.
First reads on Odroid HC1 return 305, 1207, 1297 and finally 1308 (reference value is 1309).
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org --- board/samsung/common/exynos5-dt-types.c | 38 ++++++++++++++++++++++++- 1 file changed, 37 insertions(+), 1 deletion(-)
diff --git a/board/samsung/common/exynos5-dt-types.c b/board/samsung/common/exynos5-dt-types.c index 7c1271d6547a..516c32923e44 100644 --- a/board/samsung/common/exynos5-dt-types.c +++ b/board/samsung/common/exynos5-dt-types.c @@ -57,12 +57,48 @@ static unsigned int odroid_get_rev(void) return 0; }
+/* + * Read ADC at least twice and check the resuls. If regulator providing voltage + * on to measured point was just turned on, first reads might require time + * to stabilize. + */ +static int odroid_get_adc_val(unsigned int *adcval) +{ + unsigned int adcval_prev = 0; + int ret, retries = 20; + + ret = adc_channel_single_shot("adc", CONFIG_ODROID_REV_AIN, + &adcval_prev); + if (ret) + return ret; + + while (retries--) { + mdelay(5); + + ret = adc_channel_single_shot("adc", CONFIG_ODROID_REV_AIN, + adcval); + if (ret) + return ret; + + /* + * If difference between ADC reads is less than 3%, + * accept the result + */ + if ((100 * abs(*adcval - adcval_prev) / adcval_prev) < 3) + return ret; + + adcval_prev = *adcval; + } + + return ret; +} + static int odroid_get_board_type(void) { unsigned int adcval; int ret, i;
- ret = adc_channel_single_shot("adc", CONFIG_ODROID_REV_AIN, &adcval); + ret = odroid_get_adc_val(&adcval); if (ret) goto rev_default;
Changing voltage and enabling regulator might require delays so the regulator stabilizes at expected level.
Add support for "regulator-ramp-delay" binding which can introduce required time to both enabling the regulator and to changing the voltage.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org --- .../regulator/regulator.txt | 2 + drivers/power/regulator/regulator-uclass.c | 47 ++++++++++++++++++- include/power/regulator.h | 2 + 3 files changed, 49 insertions(+), 2 deletions(-)
diff --git a/doc/device-tree-bindings/regulator/regulator.txt b/doc/device-tree-bindings/regulator/regulator.txt index 65b69c427899..4ba642b7c77f 100644 --- a/doc/device-tree-bindings/regulator/regulator.txt +++ b/doc/device-tree-bindings/regulator/regulator.txt @@ -35,6 +35,7 @@ Optional properties: - regulator-max-microamp: a maximum allowed Current value - regulator-always-on: regulator should never be disabled - regulator-boot-on: enabled by bootloader/firmware +- regulator-ramp-delay: ramp delay for regulator (in uV/us)
Note The "regulator-name" constraint is used for setting the device's uclass @@ -60,4 +61,5 @@ ldo0 { regulator-max-microamp = <100000>; regulator-always-on; regulator-boot-on; + regulator-ramp-delay = <12000>; }; diff --git a/drivers/power/regulator/regulator-uclass.c b/drivers/power/regulator/regulator-uclass.c index 6f355b969a6d..363c6e6441fa 100644 --- a/drivers/power/regulator/regulator-uclass.c +++ b/drivers/power/regulator/regulator-uclass.c @@ -35,10 +35,22 @@ int regulator_get_value(struct udevice *dev) return ops->get_value(dev); }
+static void regulator_set_value_delay(struct udevice *dev, int old_uV, + int new_uV, unsigned int ramp_delay) +{ + int delay = DIV_ROUND_UP(abs(new_uV - old_uV), ramp_delay); + + debug("regulator %s: delay %u us (%d uV -> %d uV)\n", dev->name, delay, + old_uV, new_uV); + + udelay(delay); +} + int regulator_set_value(struct udevice *dev, int uV) { const struct dm_regulator_ops *ops = dev_get_driver_ops(dev); struct dm_regulator_uclass_platdata *uc_pdata; + int ret, old_uV = uV, is_enabled = 0;
uc_pdata = dev_get_uclass_platdata(dev); if (uc_pdata->min_uV != -ENODATA && uV < uc_pdata->min_uV) @@ -49,7 +61,20 @@ int regulator_set_value(struct udevice *dev, int uV) if (!ops || !ops->set_value) return -ENOSYS;
- return ops->set_value(dev, uV); + if (uc_pdata->ramp_delay) { + is_enabled = regulator_get_enable(dev); + old_uV = regulator_get_value(dev); + } + + ret = ops->set_value(dev, uV); + + if (!ret) { + if (uc_pdata->ramp_delay && old_uV > 0 && is_enabled) + regulator_set_value_delay(dev, old_uV, uV, + uc_pdata->ramp_delay); + } + + return ret; }
/* @@ -107,6 +132,7 @@ int regulator_set_enable(struct udevice *dev, bool enable) { const struct dm_regulator_ops *ops = dev_get_driver_ops(dev); struct dm_regulator_uclass_platdata *uc_pdata; + int ret, old_enable = 0;
if (!ops || !ops->set_enable) return -ENOSYS; @@ -115,7 +141,22 @@ int regulator_set_enable(struct udevice *dev, bool enable) if (!enable && uc_pdata->always_on) return -EACCES;
- return ops->set_enable(dev, enable); + if (uc_pdata->ramp_delay) + old_enable = regulator_get_enable(dev); + + ret = ops->set_enable(dev, enable); + if (!ret) { + if (uc_pdata->ramp_delay && !old_enable && enable) { + int uV = regulator_get_value(dev); + + if (uV > 0) { + regulator_set_value_delay(dev, 0, uV, + uc_pdata->ramp_delay); + } + } + } + + return ret; }
int regulator_set_enable_if_allowed(struct udevice *dev, bool enable) @@ -335,6 +376,8 @@ static int regulator_pre_probe(struct udevice *dev) -ENODATA); uc_pdata->always_on = dev_read_bool(dev, "regulator-always-on"); uc_pdata->boot_on = dev_read_bool(dev, "regulator-boot-on"); + uc_pdata->ramp_delay = dev_read_u32_default(dev, "regulator-ramp-delay", + 0);
/* Those values are optional (-ENODATA if unset) */ if ((uc_pdata->min_uV != -ENODATA) && diff --git a/include/power/regulator.h b/include/power/regulator.h index 314160a894b7..6c6e2cd4f996 100644 --- a/include/power/regulator.h +++ b/include/power/regulator.h @@ -150,6 +150,7 @@ enum regulator_flag { * @always_on* - bool type, true or false * @boot_on* - bool type, true or false * TODO(sjg@chromium.org): Consider putting the above two into @flags + * @ramp_delay - Time to settle down after voltage change (unit: uV/us) * @flags: - flags value (see REGULATOR_FLAG_...) * @name** - fdt regulator name - should be taken from the device tree * ctrl_reg: - Control register offset used to enable/disable regulator @@ -169,6 +170,7 @@ struct dm_regulator_uclass_platdata { int max_uV; int min_uA; int max_uA; + unsigned int ramp_delay; bool always_on; bool boot_on; const char *name;
On Sat, Feb 16, 2019 at 10:45:45AM +0100, Krzysztof Kozlowski wrote:
Changing voltage and enabling regulator might require delays so the regulator stabilizes at expected level.
Add support for "regulator-ramp-delay" binding which can introduce required time to both enabling the regulator and to changing the voltage.
I'm surprised that such a thing doesn't exist already.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org
--- a/doc/device-tree-bindings/regulator/regulator.txt +++ b/doc/device-tree-bindings/regulator/regulator.txt @@ -35,6 +35,7 @@ Optional properties:
- regulator-max-microamp: a maximum allowed Current value
- regulator-always-on: regulator should never be disabled
- regulator-boot-on: enabled by bootloader/firmware
+- regulator-ramp-delay: ramp delay for regulator (in uV/us)
I guess you mean s/V, not V/s; at least the code suggests so. But my main point is: is the required delay always a linear function of the voltage jump? Depending on the dampening and load on the rail this could be an overshoot and settle, no?
So I suggest to make that an array with 2 elements: a fixed part and a time per voltage change. Does that make sense?
Torsten
On Mon, 18 Feb 2019 at 15:03, Torsten Duwe duwe@lst.de wrote:
On Sat, Feb 16, 2019 at 10:45:45AM +0100, Krzysztof Kozlowski wrote:
Changing voltage and enabling regulator might require delays so the regulator stabilizes at expected level.
Add support for "regulator-ramp-delay" binding which can introduce required time to both enabling the regulator and to changing the voltage.
I'm surprised that such a thing doesn't exist already.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org
--- a/doc/device-tree-bindings/regulator/regulator.txt +++ b/doc/device-tree-bindings/regulator/regulator.txt @@ -35,6 +35,7 @@ Optional properties:
- regulator-max-microamp: a maximum allowed Current value
- regulator-always-on: regulator should never be disabled
- regulator-boot-on: enabled by bootloader/firmware
+- regulator-ramp-delay: ramp delay for regulator (in uV/us)
I guess you mean s/V, not V/s; at least the code suggests so.
uV/uS. It is correct in the code: int delay = DIV_ROUND_UP(abs(new_uV - old_uV), ramp_delay); delay = (uV - uV) / (uV / uS) = uS
But my main point is: is the required delay always a linear function of the voltage jump? Depending on the dampening and load on the rail this could be an overshoot and settle, no?
So I suggest to make that an array with 2 elements: a fixed part and a time per voltage change. Does that make sense?
Just to make it clear - then we do not follow Linux kernel DT bindings. The voltage change might have exponential characteristic and/or have additional fixed delay time (see patch 7 here). We could re-use some properties from Linux bindings for that purpose: https://elixir.bootlin.com/linux/latest/source/Documentation/devicetree/bind... https://elixir.bootlin.com/linux/latest/source/Documentation/devicetree/bind...
Best regards, Krzysztof
On Mon, Feb 18, 2019 at 03:28:46PM +0100, Krzysztof Kozlowski wrote:
On Mon, 18 Feb 2019 at 15:03, Torsten Duwe duwe@lst.de wrote:
--- a/doc/device-tree-bindings/regulator/regulator.txt +++ b/doc/device-tree-bindings/regulator/regulator.txt @@ -35,6 +35,7 @@ Optional properties:
- regulator-max-microamp: a maximum allowed Current value
- regulator-always-on: regulator should never be disabled
- regulator-boot-on: enabled by bootloader/firmware
+- regulator-ramp-delay: ramp delay for regulator (in uV/us)
I guess you mean s/V, not V/s; at least the code suggests so.
uV/uS. It is correct in the code: int delay = DIV_ROUND_UP(abs(new_uV - old_uV), ramp_delay); delay = (uV - uV) / (uV / uS) = uS
You're right. _divide_ by that value; somhow this has escaped me. Sorry for the noise.
nit: "s" is for seconds, "S" is for conductance.
But my main point is: is the required delay always a linear function of the voltage jump? Depending on the dampening and load on the rail this could be an overshoot and settle, no?
So I suggest to make that an array with 2 elements: a fixed part and a time per voltage change. Does that make sense?
Just to make it clear - then we do not follow Linux kernel DT bindings. The voltage change might have exponential characteristic and/or have additional fixed delay time (see patch 7 here). We could re-use some properties from Linux bindings for that purpose: https://elixir.bootlin.com/linux/latest/source/Documentation/devicetree/bind... https://elixir.bootlin.com/linux/latest/source/Documentation/devicetree/bind...
I see. But then "static void regulator_set_value_delay(...)" should either at least have a "ramp" somewhere in its name or it should discover the device properties on its own, in order to be able to handle regulator-settling-time* and regulator-enable-ramp-delay as well in the future. (i.e. pass it uc_pdata instead of uc_pdata->ramp_delay and also let it handle the condition).
Torsten
On Mon, 18 Feb 2019 at 16:27, Torsten Duwe duwe@lst.de wrote:
But my main point is: is the required delay always a linear function of the voltage jump? Depending on the dampening and load on the rail this could be an overshoot and settle, no?
So I suggest to make that an array with 2 elements: a fixed part and a time per voltage change. Does that make sense?
Just to make it clear - then we do not follow Linux kernel DT bindings. The voltage change might have exponential characteristic and/or have additional fixed delay time (see patch 7 here). We could re-use some properties from Linux bindings for that purpose: https://elixir.bootlin.com/linux/latest/source/Documentation/devicetree/bind... https://elixir.bootlin.com/linux/latest/source/Documentation/devicetree/bind...
I see. But then "static void regulator_set_value_delay(...)" should either at least have a "ramp" somewhere in its name or it should discover the device properties on its own, in order to be able to handle regulator-settling-time* and regulator-enable-ramp-delay as well in the future. (i.e. pass it uc_pdata instead of uc_pdata->ramp_delay and also let it handle the condition).
Makes sense, so let me add the ramp keyword. I will also mention in comment that other delays are not yet handled.
Thanks for feedback!
Best regards, Krzysztof
According to datasheet, the output on LDO regulators will start appearing after 10-15 us.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org --- drivers/power/regulator/s2mps11_regulator.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/drivers/power/regulator/s2mps11_regulator.c b/drivers/power/regulator/s2mps11_regulator.c index 723d27f67c9a..67d1f9689de3 100644 --- a/drivers/power/regulator/s2mps11_regulator.c +++ b/drivers/power/regulator/s2mps11_regulator.c @@ -551,7 +551,16 @@ static int ldo_get_enable(struct udevice *dev)
static int ldo_set_enable(struct udevice *dev, bool enable) { - return s2mps11_ldo_enable(dev, PMIC_OP_SET, &enable); + int ret; + + ret = s2mps11_ldo_enable(dev, PMIC_OP_SET, &enable); + if (ret) + return ret; + + /* Wait the "enable delay" for voltage to start to rise */ + udelay(15); + + return 0; }
static int ldo_get_mode(struct udevice *dev)
The ADC block requires VDD supply to be on so provide one.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org Reviewed-by: Lukasz Majewski lukma@denx.de --- arch/arm/dts/exynos5422-odroidxu3.dts | 1 + 1 file changed, 1 insertion(+)
diff --git a/arch/arm/dts/exynos5422-odroidxu3.dts b/arch/arm/dts/exynos5422-odroidxu3.dts index e859dd1b981a..9dfae90667cf 100644 --- a/arch/arm/dts/exynos5422-odroidxu3.dts +++ b/arch/arm/dts/exynos5422-odroidxu3.dts @@ -32,6 +32,7 @@
adc@12D10000 { u-boot,dm-pre-reloc; + vdd-supply = <&ldo4_reg>; status = "okay"; };
Add startup time to LDO regulators of S2MPS11 PMIC on Odroid XU3/XU4/HC1 family of boards to be sure the voltage is proper before relying on the regulator.
The datasheet for all the S2MPS1x family is inconsistent here and does not specify unambiguously the value of ramp delay for LDO. It mentions 30 mV/us in one timing diagram but then omits it completely in LDO regulator characteristics table (it is specified for bucks).
However the vendor kernels for Galaxy S5 and Odroid XU3 use values of 12 mV/us or 24 mV/us.
Without the ramp delay value the consumers do not wait for voltage settle after changing it. Although the proper value of ramp delay for LDOs is unknown, it seems safer to use at least some value from reference kernel than to leave it unset.
Signed-off-by: Krzysztof Kozlowski krzk@kernel.org Reviewed-by: Lukasz Majewski lukma@denx.de --- arch/arm/dts/exynos5422-odroidxu3.dts | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
diff --git a/arch/arm/dts/exynos5422-odroidxu3.dts b/arch/arm/dts/exynos5422-odroidxu3.dts index 9dfae90667cf..04ecc404f907 100644 --- a/arch/arm/dts/exynos5422-odroidxu3.dts +++ b/arch/arm/dts/exynos5422-odroidxu3.dts @@ -45,6 +45,7 @@ regulator-name = "vdd_ldo1"; regulator-min-microvolt = <1000000>; regulator-max-microvolt = <1000000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -52,18 +53,21 @@ regulator-name = "vddq_mmc0"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; };
ldo4_reg: LDO4 { regulator-name = "vdd_adc"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; };
ldo5_reg: LDO5 { regulator-name = "vdd_ldo5"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -71,6 +75,7 @@ regulator-name = "vdd_ldo6"; regulator-min-microvolt = <1000000>; regulator-max-microvolt = <1000000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -78,6 +83,7 @@ regulator-name = "vdd_ldo7"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -85,6 +91,7 @@ regulator-name = "vdd_ldo8"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -92,6 +99,7 @@ regulator-name = "vdd_ldo9"; regulator-min-microvolt = <3000000>; regulator-max-microvolt = <3000000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -99,6 +107,7 @@ regulator-name = "vdd_ldo10"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -106,6 +115,7 @@ regulator-name = "vdd_ldo11"; regulator-min-microvolt = <1000000>; regulator-max-microvolt = <1000000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -113,6 +123,7 @@ regulator-name = "vdd_ldo12"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -120,12 +131,14 @@ regulator-name = "vddq_mmc2"; regulator-min-microvolt = <2800000>; regulator-max-microvolt = <2800000>; + regulator-ramp-delay = <12000>; };
ldo15_reg: LDO15 { regulator-name = "vdd_ldo15"; regulator-min-microvolt = <3300000>; regulator-max-microvolt = <3300000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -133,6 +146,7 @@ regulator-name = "vdd_ldo16"; regulator-min-microvolt = <2200000>; regulator-max-microvolt = <2200000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -140,6 +154,7 @@ regulator-name = "vdd_ldo17"; regulator-min-microvolt = <3300000>; regulator-max-microvolt = <3300000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -147,18 +162,21 @@ regulator-name = "vdd_emmc_1V8"; regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; + regulator-ramp-delay = <12000>; };
ldo19_reg: LDO19 { regulator-name = "vdd_sd"; regulator-min-microvolt = <2800000>; regulator-max-microvolt = <2800000>; + regulator-ramp-delay = <12000>; };
ldo24_reg: LDO24 { regulator-name = "tsp_io"; regulator-min-microvolt = <2800000>; regulator-max-microvolt = <2800000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
@@ -166,6 +184,7 @@ regulator-name = "vdd_ldo26"; regulator-min-microvolt = <3000000>; regulator-max-microvolt = <3000000>; + regulator-ramp-delay = <12000>; regulator-always-on; };
Hi Krzysztof,
Please add my. on Odroid HC1
Tested-by: Anand Moon linux.amoon@gmail.com
Best Regards -Anand
On Sat, 16 Feb 2019 at 15:16, Krzysztof Kozlowski krzk@kernel.org wrote:
Hi,
Changes since v2
- Add Lukasz review tags.
- Patch 7: Return on error, as suggested by Simon.
- Patch 3: Use IS_ENABLED() to run revision detection only once - either during late display board or misc_init_r.
Changes since v1
- Move fixes to beginning of patchset.
- Patch 3: Rework the idea - split revision detection to be executed later.
- Patch 4: New patch.
- Patch 6: Apply Simon's comments.
- Patch 6: Do not delay when changing voltage if regulator is disabled.
- Patch 6: Do not delay when disabling the regulator.
Description
Odroid HC1 does not reboot properly (at least from SD card but I do not expect difference on eMMC), if LDO4/VDD_ADC was turned off by Linux kernel. This condition happens so far always, because Linux kernel did not enable ADC on Odroid HC1, therefore the VDD_ADC regulator was turned off as unused.
The issue is in detection of revision which later is used to load proper DTB.
The revision is obtained by ADC read of a voltage depending on VDD_ADC. Therefore:
- VDD_ADC has to be turned on (but board detection happens before power is initialized),
- Turning VDD_ADC should wait with ramp delay,
- Reading the value from ADC should wait for it to stabilize.
Tested on Odroid XU3-Lite and Odroid HC1.
Commends and testing are welcomed.
Best regards, Krzysztof
Krzysztof Kozlowski (9): adc: exynos-adc: Fix wrong bit operation used to stop the ADC power: regulator: s2mps11: Fix step for LDO27 and LDO35 arm: exynos: Detect revision later, when all resources are ready arm: exynos: odroid-xu3: Display info late to have proper type arm: exynos: Wait till ADC stabilizes before checking Odroid HC1 revision regulator: Add support for ramp delay power: regulator: s2mps11: Add enable delay arm: dts: exynos: Add supply for ADC block to Odroid XU3 family arm: dts: exynos: Add ramp delay property to LDO regulators to Odroid XU3 family
arch/arm/dts/exynos5422-odroidxu3.dts | 20 +++++++ board/samsung/common/board.c | 24 ++++++++- board/samsung/common/exynos5-dt-types.c | 54 +++++++++++++++++-- board/samsung/odroid/odroid.c | 8 +++ configs/odroid-xu3_defconfig | 2 + .../regulator/regulator.txt | 2 + drivers/adc/exynos-adc.c | 2 +- drivers/power/regulator/regulator-uclass.c | 47 +++++++++++++++- drivers/power/regulator/s2mps11_regulator.c | 15 +++++- include/power/regulator.h | 2 + include/samsung/misc.h | 1 + 11 files changed, 167 insertions(+), 10 deletions(-)
-- 2.17.1
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
Dear Krzysztof Kozlowski,
On 24/02/2019 21:55, Anand Moon wrote:
Hi Krzysztof,
Please add my. on Odroid HC1
Tested-by: Anand Moon linux.amoon@gmail.com
Best Regards -Anand
On Sat, 16 Feb 2019 at 15:16, Krzysztof Kozlowski krzk@kernel.org wrote:
Hi,
Changes since v2
- Add Lukasz review tags.
- Patch 7: Return on error, as suggested by Simon.
- Patch 3: Use IS_ENABLED() to run revision detection only once - either during late display board or misc_init_r.
Changes since v1
- Move fixes to beginning of patchset.
- Patch 3: Rework the idea - split revision detection to be executed later.
- Patch 4: New patch.
- Patch 6: Apply Simon's comments.
- Patch 6: Do not delay when changing voltage if regulator is disabled.
- Patch 6: Do not delay when disabling the regulator.
Description
Odroid HC1 does not reboot properly (at least from SD card but I do not expect difference on eMMC), if LDO4/VDD_ADC was turned off by Linux kernel. This condition happens so far always, because Linux kernel did not enable ADC on Odroid HC1, therefore the VDD_ADC regulator was turned off as unused.
The issue is in detection of revision which later is used to load proper DTB.
The revision is obtained by ADC read of a voltage depending on VDD_ADC. Therefore:
- VDD_ADC has to be turned on (but board detection happens before power is initialized),
- Turning VDD_ADC should wait with ramp delay,
- Reading the value from ADC should wait for it to stabilize.
Tested on Odroid XU3-Lite and Odroid HC1.
Commends and testing are welcomed.
Best regards, Krzysztof
Krzysztof Kozlowski (9): adc: exynos-adc: Fix wrong bit operation used to stop the ADC power: regulator: s2mps11: Fix step for LDO27 and LDO35 arm: exynos: Detect revision later, when all resources are ready arm: exynos: odroid-xu3: Display info late to have proper type arm: exynos: Wait till ADC stabilizes before checking Odroid HC1 revision regulator: Add support for ramp delay power: regulator: s2mps11: Add enable delay arm: dts: exynos: Add supply for ADC block to Odroid XU3 family arm: dts: exynos: Add ramp delay property to LDO regulators to Odroid XU3 family
arch/arm/dts/exynos5422-odroidxu3.dts | 20 +++++++ board/samsung/common/board.c | 24 ++++++++- board/samsung/common/exynos5-dt-types.c | 54 +++++++++++++++++-- board/samsung/odroid/odroid.c | 8 +++ configs/odroid-xu3_defconfig | 2 + .../regulator/regulator.txt | 2 + drivers/adc/exynos-adc.c | 2 +- drivers/power/regulator/regulator-uclass.c | 47 +++++++++++++++- drivers/power/regulator/s2mps11_regulator.c | 15 +++++- include/power/regulator.h | 2 + include/samsung/misc.h | 1 + 11 files changed, 167 insertions(+), 10 deletions(-)
-- 2.17.1
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
Patch looks good but could you please rebase on latest u-boot-samsung tree?
Applying: adc: exynos-adc: Fix wrong bit operation used to stop the ADC Applying: power: regulator: s2mps11: Fix step for LDO27 and LDO35 Applying: arm: exynos: Detect revision later, when all resources are ready Applying: arm: exynos: odroid-xu3: Display info late to have proper type error: patch failed: configs/odroid-xu3_defconfig:11 error: configs/odroid-xu3_defconfig: patch does not apply Patch failed at 0004 arm: exynos: odroid-xu3: Display info late to have proper type
Thanks, Minkyu Kang.
On Tue, 5 Mar 2019 at 11:16, Minkyu Kang mk7.kang@samsung.com wrote:
Dear Krzysztof Kozlowski,
On 24/02/2019 21:55, Anand Moon wrote:
Hi Krzysztof,
Please add my. on Odroid HC1
Tested-by: Anand Moon linux.amoon@gmail.com
Best Regards -Anand
On Sat, 16 Feb 2019 at 15:16, Krzysztof Kozlowski krzk@kernel.org wrote:
Hi,
Changes since v2
- Add Lukasz review tags.
- Patch 7: Return on error, as suggested by Simon.
- Patch 3: Use IS_ENABLED() to run revision detection only once - either during late display board or misc_init_r.
Changes since v1
- Move fixes to beginning of patchset.
- Patch 3: Rework the idea - split revision detection to be executed later.
- Patch 4: New patch.
- Patch 6: Apply Simon's comments.
- Patch 6: Do not delay when changing voltage if regulator is disabled.
- Patch 6: Do not delay when disabling the regulator.
Description
Odroid HC1 does not reboot properly (at least from SD card but I do not expect difference on eMMC), if LDO4/VDD_ADC was turned off by Linux kernel. This condition happens so far always, because Linux kernel did not enable ADC on Odroid HC1, therefore the VDD_ADC regulator was turned off as unused.
The issue is in detection of revision which later is used to load proper DTB.
The revision is obtained by ADC read of a voltage depending on VDD_ADC. Therefore:
- VDD_ADC has to be turned on (but board detection happens before power is initialized),
- Turning VDD_ADC should wait with ramp delay,
- Reading the value from ADC should wait for it to stabilize.
Tested on Odroid XU3-Lite and Odroid HC1.
Commends and testing are welcomed.
Best regards, Krzysztof
Krzysztof Kozlowski (9): adc: exynos-adc: Fix wrong bit operation used to stop the ADC power: regulator: s2mps11: Fix step for LDO27 and LDO35 arm: exynos: Detect revision later, when all resources are ready arm: exynos: odroid-xu3: Display info late to have proper type arm: exynos: Wait till ADC stabilizes before checking Odroid HC1 revision regulator: Add support for ramp delay power: regulator: s2mps11: Add enable delay arm: dts: exynos: Add supply for ADC block to Odroid XU3 family arm: dts: exynos: Add ramp delay property to LDO regulators to Odroid XU3 family
arch/arm/dts/exynos5422-odroidxu3.dts | 20 +++++++ board/samsung/common/board.c | 24 ++++++++- board/samsung/common/exynos5-dt-types.c | 54 +++++++++++++++++-- board/samsung/odroid/odroid.c | 8 +++ configs/odroid-xu3_defconfig | 2 + .../regulator/regulator.txt | 2 + drivers/adc/exynos-adc.c | 2 +- drivers/power/regulator/regulator-uclass.c | 47 +++++++++++++++- drivers/power/regulator/s2mps11_regulator.c | 15 +++++- include/power/regulator.h | 2 + include/samsung/misc.h | 1 + 11 files changed, 167 insertions(+), 10 deletions(-)
-- 2.17.1
U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
Patch looks good but could you please rebase on latest u-boot-samsung tree?
Applying: adc: exynos-adc: Fix wrong bit operation used to stop the ADC Applying: power: regulator: s2mps11: Fix step for LDO27 and LDO35 Applying: arm: exynos: Detect revision later, when all resources are ready Applying: arm: exynos: odroid-xu3: Display info late to have proper type error: patch failed: configs/odroid-xu3_defconfig:11 error: configs/odroid-xu3_defconfig: patch does not apply Patch failed at 0004 arm: exynos: odroid-xu3: Display info late to have proper type
Sure, let me rebase.
Best regards, Krzysztof
participants (5)
-
Anand Moon -
Krzysztof Kozlowski -
Lukasz Majewski -
Minkyu Kang -
Torsten Duwe