Add SW support for the vibrator module inside PMI632, PM7250B, PM7325B, PM7550BA.
It is very similar to the vibrator module inside PM8916 which is supported in
pm8xxx-vib driver but just the drive amplitude is controlled with 2 registers,
and the register base offset in each PMIC is different.
Changes in v8:
1. Remove hw_type, and still keep the register info in match data
2. Update to use register offset in pm8xxx_regs, and the base address
defined in DT for SPMI vibrator will be added in register access
3. Update voltage output range for SPMI vibrator which has 2 bytes drive
registers
Changes in v7:
1. Fix a typo: SSBL_VIB_DRV_REG --> SSBI_VIB_DRV_REG
2. Move the hw_type switch case in pm8xxx_vib_set() to the refactoring
change.
Changes in v6:
1. Add "qcom,pmi632-vib" as a standalone compatible string.
Changes in v5:
1. Drop "qcom,spmi-vib-gen2" generic compatible string as requested
and use device specific compatible strings only.
Changes in v4:
1. Update to use the combination of the HW type and register offset
as the constant match data, the register base address defined in
'reg' property will be added when accessing SPMI registers using
regmap APIs.
2. Remove 'qcom,spmi-vib-gen1' generic compatible string.
Changes in v3:
1. Refactor the driver to support different type of the vibrators with
better flexibility by introducing the HW type with corresponding
register fields definitions.
2. Add 'qcom,spmi-vib-gen1' and 'qcom,spmi-vib-gen2' compatible
strings, and add PMI632, PM7250B, PM7325B, PM7550BA as compatbile as
spmi-vib-gen2.
Changes in v2:
Remove the "pm7550ba-vib" compatible string as it's compatible with pm7325b.
Signed-off-by: Fenglin Wu <[email protected]>
---
Fenglin Wu (3):
input: pm8xxx-vibrator: refactor to support new SPMI vibrator
dt-bindings: input: qcom,pm8xxx-vib: add new SPMI vibrator module
input: pm8xxx-vibrator: add new SPMI vibrator support
.../devicetree/bindings/input/qcom,pm8xxx-vib.yaml | 16 +++-
drivers/input/misc/pm8xxx-vibrator.c | 87 ++++++++++++++++------
2 files changed, 78 insertions(+), 25 deletions(-)
---
base-commit: 650cda2ce25f08e8fae391b3ba6be27e7296c6a5
change-id: 20240328-pm8xxx-vibrator-new-design-e5811ad59e8a
Best regards,
--
Fenglin Wu <[email protected]>
From: Fenglin Wu <[email protected]>
Currently, vibrator control register addresses are hard coded,
including the base address and offsets, it's not flexible to
support new SPMI vibrator module which is usually included in
different PMICs with different base address. Refactor it by using
the base address defined in devicetree.
Signed-off-by: Fenglin Wu <[email protected]>
---
drivers/input/misc/pm8xxx-vibrator.c | 40 ++++++++++++++++++++++++------------
1 file changed, 27 insertions(+), 13 deletions(-)
diff --git a/drivers/input/misc/pm8xxx-vibrator.c b/drivers/input/misc/pm8xxx-vibrator.c
index 04cb87efd799..3b6a2e949f30 100644
--- a/drivers/input/misc/pm8xxx-vibrator.c
+++ b/drivers/input/misc/pm8xxx-vibrator.c
@@ -19,26 +19,26 @@
#define MAX_FF_SPEED 0xff
struct pm8xxx_regs {
- unsigned int enable_addr;
+ unsigned int enable_offset;
unsigned int enable_mask;
- unsigned int drv_addr;
+ unsigned int drv_offset;
unsigned int drv_mask;
unsigned int drv_shift;
unsigned int drv_en_manual_mask;
};
static const struct pm8xxx_regs pm8058_regs = {
- .drv_addr = 0x4A,
+ .drv_offset = 0x4A,
.drv_mask = 0xf8,
.drv_shift = 3,
.drv_en_manual_mask = 0xfc,
};
static struct pm8xxx_regs pm8916_regs = {
- .enable_addr = 0xc046,
+ .enable_offset = 0x46,
.enable_mask = BIT(7),
- .drv_addr = 0xc041,
+ .drv_offset = 0x41,
.drv_mask = 0x1F,
.drv_shift = 0,
.drv_en_manual_mask = 0,
@@ -50,6 +50,8 @@ static struct pm8xxx_regs pm8916_regs = {
* @work: work structure to set the vibration parameters
* @regmap: regmap for register read/write
* @regs: registers' info
+ * @enable_addr: vibrator enable register
+ * @drv_addr: vibrator drive strength register
* @speed: speed of vibration set from userland
* @active: state of vibrator
* @level: level of vibration to set in the chip
@@ -60,6 +62,8 @@ struct pm8xxx_vib {
struct work_struct work;
struct regmap *regmap;
const struct pm8xxx_regs *regs;
+ unsigned int enable_addr;
+ unsigned int drv_addr;
int speed;
int level;
bool active;
@@ -82,14 +86,14 @@ static int pm8xxx_vib_set(struct pm8xxx_vib *vib, bool on)
else
val &= ~regs->drv_mask;
- rc = regmap_write(vib->regmap, regs->drv_addr, val);
+ rc = regmap_write(vib->regmap, vib->drv_addr, val);
if (rc < 0)
return rc;
vib->reg_vib_drv = val;
if (regs->enable_mask)
- rc = regmap_update_bits(vib->regmap, regs->enable_addr,
+ rc = regmap_update_bits(vib->regmap, vib->enable_addr,
regs->enable_mask, on ? ~0 : 0);
return rc;
@@ -102,11 +106,10 @@ static int pm8xxx_vib_set(struct pm8xxx_vib *vib, bool on)
static void pm8xxx_work_handler(struct work_struct *work)
{
struct pm8xxx_vib *vib = container_of(work, struct pm8xxx_vib, work);
- const struct pm8xxx_regs *regs = vib->regs;
- int rc;
unsigned int val;
+ int rc;
- rc = regmap_read(vib->regmap, regs->drv_addr, &val);
+ rc = regmap_read(vib->regmap, vib->drv_addr, &val);
if (rc < 0)
return;
@@ -169,7 +172,7 @@ static int pm8xxx_vib_probe(struct platform_device *pdev)
struct pm8xxx_vib *vib;
struct input_dev *input_dev;
int error;
- unsigned int val;
+ unsigned int val, reg_base = 0;
const struct pm8xxx_regs *regs;
vib = devm_kzalloc(&pdev->dev, sizeof(*vib), GFP_KERNEL);
@@ -189,13 +192,24 @@ static int pm8xxx_vib_probe(struct platform_device *pdev)
regs = of_device_get_match_data(&pdev->dev);
+ if (regs->enable_offset != 0) {
+ error = fwnode_property_read_u32(pdev->dev.fwnode, "reg", ®_base);
+ if (error < 0) {
+ dev_err(&pdev->dev, "Failed to read reg address, rc=%d\n", error);
+ return error;
+ }
+ }
+
+ vib->enable_addr = reg_base + regs->enable_offset;
+ vib->drv_addr = reg_base + regs->drv_offset;
+
/* operate in manual mode */
- error = regmap_read(vib->regmap, regs->drv_addr, &val);
+ error = regmap_read(vib->regmap, vib->drv_addr, &val);
if (error < 0)
return error;
val &= regs->drv_en_manual_mask;
- error = regmap_write(vib->regmap, regs->drv_addr, val);
+ error = regmap_write(vib->regmap, vib->drv_addr, val);
if (error < 0)
return error;
--
2.25.1
From: Fenglin Wu <[email protected]>
Add support for a new SPMI vibrator module which is very similar
to the vibrator module inside PM8916 but has a finer drive voltage
step and different output voltage range, its drive level control
is expanded across 2 registers. The vibrator module can be found
in following Qualcomm PMICs: PMI632, PM7250B, PM7325B, PM7550BA.
Signed-off-by: Fenglin Wu <[email protected]>
---
drivers/input/misc/pm8xxx-vibrator.c | 49 +++++++++++++++++++++++++++++-------
1 file changed, 40 insertions(+), 9 deletions(-)
diff --git a/drivers/input/misc/pm8xxx-vibrator.c b/drivers/input/misc/pm8xxx-vibrator.c
index 3b6a2e949f30..59548cd9331c 100644
--- a/drivers/input/misc/pm8xxx-vibrator.c
+++ b/drivers/input/misc/pm8xxx-vibrator.c
@@ -12,9 +12,9 @@
#include <linux/regmap.h>
#include <linux/slab.h>
-#define VIB_MAX_LEVEL_mV (3100)
-#define VIB_MIN_LEVEL_mV (1200)
-#define VIB_MAX_LEVELS (VIB_MAX_LEVEL_mV - VIB_MIN_LEVEL_mV)
+#define VIB_MAX_LEVEL_mV(vib) (vib->drv2_addr ? (3544) : (3100))
+#define VIB_MIN_LEVEL_mV(vib) (vib->drv2_addr ? (1504) : (1200))
+#define VIB_MAX_LEVELS(vib) (VIB_MAX_LEVEL_mV(vib) - VIB_MIN_LEVEL_mV(vib))
#define MAX_FF_SPEED 0xff
@@ -25,6 +25,9 @@ struct pm8xxx_regs {
unsigned int drv_offset;
unsigned int drv_mask;
unsigned int drv_shift;
+ unsigned int drv2_offset;
+ unsigned int drv2_mask;
+ unsigned int drv2_shift;
unsigned int drv_en_manual_mask;
};
@@ -44,6 +47,18 @@ static struct pm8xxx_regs pm8916_regs = {
.drv_en_manual_mask = 0,
};
+static struct pm8xxx_regs pmi632_regs = {
+ .enable_offset = 0x46,
+ .enable_mask = BIT(7),
+ .drv_offset = 0x40,
+ .drv_mask = 0xFF,
+ .drv_shift = 0,
+ .drv2_offset = 0x41,
+ .drv2_mask = 0x0F,
+ .drv2_shift = 8,
+ .drv_en_manual_mask = 0,
+};
+
/**
* struct pm8xxx_vib - structure to hold vibrator data
* @vib_input_dev: input device supporting force feedback
@@ -52,6 +67,7 @@ static struct pm8xxx_regs pm8916_regs = {
* @regs: registers' info
* @enable_addr: vibrator enable register
* @drv_addr: vibrator drive strength register
+ * @drv2_addr: vibrator drive strength upper byte register
* @speed: speed of vibration set from userland
* @active: state of vibrator
* @level: level of vibration to set in the chip
@@ -64,6 +80,7 @@ struct pm8xxx_vib {
const struct pm8xxx_regs *regs;
unsigned int enable_addr;
unsigned int drv_addr;
+ unsigned int drv2_addr;
int speed;
int level;
bool active;
@@ -92,6 +109,16 @@ static int pm8xxx_vib_set(struct pm8xxx_vib *vib, bool on)
vib->reg_vib_drv = val;
+ if (regs->drv2_mask) {
+ if (on)
+ val = (vib->level << regs->drv2_shift) & regs->drv2_mask;
+ else
+ val = 0;
+ rc = regmap_write(vib->regmap, vib->drv2_addr, val);
+ if (rc < 0)
+ return rc;
+ }
+
if (regs->enable_mask)
rc = regmap_update_bits(vib->regmap, vib->enable_addr,
regs->enable_mask, on ? ~0 : 0);
@@ -114,19 +141,22 @@ static void pm8xxx_work_handler(struct work_struct *work)
return;
/*
- * pmic vibrator supports voltage ranges from 1.2 to 3.1V, so
+ * pmic vibrator supports voltage ranges from MIN_LEVEL to MAX_LEVEL, so
* scale the level to fit into these ranges.
*/
if (vib->speed) {
vib->active = true;
- vib->level = ((VIB_MAX_LEVELS * vib->speed) / MAX_FF_SPEED) +
- VIB_MIN_LEVEL_mV;
- vib->level /= 100;
+ vib->level = ((VIB_MAX_LEVELS(vib) * vib->speed) / MAX_FF_SPEED) +
+ VIB_MIN_LEVEL_mV(vib);
} else {
vib->active = false;
- vib->level = VIB_MIN_LEVEL_mV / 100;
+ vib->level = VIB_MIN_LEVEL_mV(vib);
+
}
+ if (!vib->drv2_addr)
+ vib->level /= 100;
+
pm8xxx_vib_set(vib, vib->active);
}
@@ -202,7 +232,7 @@ static int pm8xxx_vib_probe(struct platform_device *pdev)
vib->enable_addr = reg_base + regs->enable_offset;
vib->drv_addr = reg_base + regs->drv_offset;
-
+ vib->drv2_addr = reg_base + regs->drv2_offset;
/* operate in manual mode */
error = regmap_read(vib->regmap, vib->drv_addr, &val);
if (error < 0)
@@ -256,6 +286,7 @@ static const struct of_device_id pm8xxx_vib_id_table[] = {
{ .compatible = "qcom,pm8058-vib", .data = &pm8058_regs },
{ .compatible = "qcom,pm8921-vib", .data = &pm8058_regs },
{ .compatible = "qcom,pm8916-vib", .data = &pm8916_regs },
+ { .compatible = "qcom,pmi632-vib", .data = &pmi632_regs },
{ }
};
MODULE_DEVICE_TABLE(of, pm8xxx_vib_id_table);
--
2.25.1
On 1.04.2024 10:38 AM, Fenglin Wu via B4 Relay wrote:
> From: Fenglin Wu <[email protected]>
>
> Currently, vibrator control register addresses are hard coded,
> including the base address and offsets, it's not flexible to
> support new SPMI vibrator module which is usually included in
> different PMICs with different base address. Refactor it by using
> the base address defined in devicetree.
>
> Signed-off-by: Fenglin Wu <[email protected]>
> ---
[...]
> if (regs->enable_mask)
> - rc = regmap_update_bits(vib->regmap, regs->enable_addr,
> + rc = regmap_update_bits(vib->regmap, vib->enable_addr,
> regs->enable_mask, on ? ~0 : 0);
The idiomatic way across the kernel seems to be writing the mask value
instead of ~0 (which also saves like 2 cpu instructions)
Not sure about how ssbi addressing works, but except for that lgtm
Konrad
On 4/2/2024 11:21 PM, Konrad Dybcio wrote:
> On 1.04.2024 10:38 AM, Fenglin Wu via B4 Relay wrote:
>> From: Fenglin Wu <[email protected]>
>>
>> Currently, vibrator control register addresses are hard coded,
>> including the base address and offsets, it's not flexible to
>> support new SPMI vibrator module which is usually included in
>> different PMICs with different base address. Refactor it by using
>> the base address defined in devicetree.
>>
>> Signed-off-by: Fenglin Wu <[email protected]>
>> ---
>
> [...]
>
>> if (regs->enable_mask)
>> - rc = regmap_update_bits(vib->regmap, regs->enable_addr,
>> + rc = regmap_update_bits(vib->regmap, vib->enable_addr,
>> regs->enable_mask, on ? ~0 : 0);
>
> The idiomatic way across the kernel seems to be writing the mask value
> instead of ~0 (which also saves like 2 cpu instructions)
>
>
> Not sure about how ssbi addressing works, but except for that lgtm
>
Agree.
SSBI driver doesn't provide reg_update_bits function call so similar
mathematics is done on the value before writing to the register, I can
update it to use enable_mask directly in next version.
> Konrad
On 4/1/24 10:38, Fenglin Wu via B4 Relay wrote:
> From: Fenglin Wu <[email protected]>
>
> Add support for a new SPMI vibrator module which is very similar
> to the vibrator module inside PM8916 but has a finer drive voltage
> step and different output voltage range, its drive level control
> is expanded across 2 registers. The vibrator module can be found
> in following Qualcomm PMICs: PMI632, PM7250B, PM7325B, PM7550BA.
>
> Signed-off-by: Fenglin Wu <[email protected]>
> ---
[...]
>
> -#define VIB_MAX_LEVEL_mV (3100)
> -#define VIB_MIN_LEVEL_mV (1200)
> -#define VIB_MAX_LEVELS (VIB_MAX_LEVEL_mV - VIB_MIN_LEVEL_mV)
> +#define VIB_MAX_LEVEL_mV(vib) (vib->drv2_addr ? (3544) : (3100))
You shouldn't need the additional inside parentheses
Also, is this really a good discriminator for the voltage ranges? Do *all*
PMIC vibrators with a drv2_addr operate within this range? If not, consider
a struct field here
> +#define VIB_MIN_LEVEL_mV(vib) (vib->drv2_addr ? (1504) : (1200))
> +#define VIB_MAX_LEVELS(vib) (VIB_MAX_LEVEL_mV(vib) - VIB_MIN_LEVEL_mV(vib))
If the ranges are supposed to be inclusive, this is off-by-one. But looking
at the driver, it seems like MIN_LEVEL may be "off"? I'm not sure though.
Either way, this would be a separate fix.
[...]
> +static struct pm8xxx_regs pmi632_regs = {
> + .enable_offset = 0x46,
> + .enable_mask = BIT(7),
> + .drv_offset = 0x40,
> + .drv_mask = 0xFF,
GENMASK(7, 0)
> + .drv_shift = 0,
> + .drv2_offset = 0x41,
> + .drv2_mask = 0x0F,
GENMASK(3, 0)
[...]
>
> + if (regs->drv2_mask) {
> + if (on)
> + val = (vib->level << regs->drv2_shift) & regs->drv2_mask;
> + else
> + val = 0;
> + rc = regmap_write(vib->regmap, vib->drv2_addr, val);
Are you purposefuly zeroing out the other bits?
If yes, consider regmap_write_bits here
If not, consider regmap_update_bits here
> + if (rc < 0)
> + return rc;
Ignore regmap_r/w errors, these mean a complete failure of the API and
we don't generally assume MMIO accesses can fail
Unless SPMI is known to have issues here
> + }
> +
> if (regs->enable_mask)
> rc = regmap_update_bits(vib->regmap, vib->enable_addr,
> regs->enable_mask, on ? ~0 : 0);
> @@ -114,19 +141,22 @@ static void pm8xxx_work_handler(struct work_struct *work)
> return;
>
> /*
> - * pmic vibrator supports voltage ranges from 1.2 to 3.1V, so
> + * pmic vibrator supports voltage ranges from MIN_LEVEL to MAX_LEVEL, so
> * scale the level to fit into these ranges.
> */
> if (vib->speed) {
> vib->active = true;
> - vib->level = ((VIB_MAX_LEVELS * vib->speed) / MAX_FF_SPEED) +
> - VIB_MIN_LEVEL_mV;
> - vib->level /= 100;
> + vib->level = ((VIB_MAX_LEVELS(vib) * vib->speed) / MAX_FF_SPEED) +
mult_frac()
> + VIB_MIN_LEVEL_mV(vib);
vib->level = VIB_MIN_LEVEL_mV;
vib->level += the other thing
for readability?
> } else {
> vib->active = false;
> - vib->level = VIB_MIN_LEVEL_mV / 100;
> + vib->level = VIB_MIN_LEVEL_mV(vib);
> +
> }
>
> + if (!vib->drv2_addr)
> + vib->level /= 100;
Maybe this could be moved to pm8xxx_vib_set() instead
> +
> pm8xxx_vib_set(vib, vib->active);
> }
>
> @@ -202,7 +232,7 @@ static int pm8xxx_vib_probe(struct platform_device *pdev)
>
> vib->enable_addr = reg_base + regs->enable_offset;
> vib->drv_addr = reg_base + regs->drv_offset;
> -
> + vib->drv2_addr = reg_base + regs->drv2_offset;
It would be nice to preserve a newline between assignments and rw
functions here
Thanks for working on this!
Konrad
Hi Konrad,
On 4/11/2024 2:10 AM, Konrad Dybcio wrote:
>
>
>> -#define VIB_MAX_LEVEL_mV (3100)
>> -#define VIB_MIN_LEVEL_mV (1200)
>> -#define VIB_MAX_LEVELS (VIB_MAX_LEVEL_mV - VIB_MIN_LEVEL_mV)
>> +#define VIB_MAX_LEVEL_mV(vib) (vib->drv2_addr ? (3544) : (3100))
>
> You shouldn't need the additional inside parentheses
>
> Also, is this really a good discriminator for the voltage ranges? Do *all*
> PMIC vibrators with a drv2_addr operate within this range? If not, consider
> a struct field here
>
Currently, yes, all PMIC vibrators with a drv2_addr (PMI632, PM7250B,
PM7325B, PM7550BA) operate within the same range because they are the
same type.
>
>> +#define VIB_MIN_LEVEL_mV(vib) (vib->drv2_addr ? (1504) : (1200))
>> +#define VIB_MAX_LEVELS(vib) (VIB_MAX_LEVEL_mV(vib) -
>> VIB_MIN_LEVEL_mV(vib))
>
> If the ranges are supposed to be inclusive, this is off-by-one. But looking
> at the driver, it seems like MIN_LEVEL may be "off"? I'm not sure though.
>
> Either way, this would be a separate fix.
> [...]
The voltage range [1504, 3544] for the new SPMI vibrator is inclusive. I
checked the voltage range [1200 3100] for PM8916 SPMI vibrator is also
inclusive. I couldn't find any document to confirm if the SSBI vibrator
but I assume it is the same as PM8916. I will make change before the
series to address it.
Thanks for reviewing the changes!
Fenglin
Hi Konrad,
On 4/11/2024 2:10 AM, Konrad Dybcio wrote:
>
>
>> + if (regs->drv2_mask) {
>> + if (on)
>> + val = (vib->level << regs->drv2_shift) & regs->drv2_mask;
>> + else
>> + val = 0;
>> + rc = regmap_write(vib->regmap, vib->drv2_addr, val);
>
> Are you purposefuly zeroing out the other bits?
>
> If yes, consider regmap_write_bits here
> If not, consider regmap_update_bits here
>
>> + if (rc < 0)
>> + return rc;
>
> Ignore regmap_r/w errors, these mean a complete failure of the API and
> we don't generally assume MMIO accesses can fail
>
> Unless SPMI is known to have issues here
>
Sorry, forgot to reply on this comment. Yes, SPMI transaction would fail
(even with very low odds) on some boards if the layout of SPMI lines is
not good enough. I'd like to keep the consistence since the whole driver
also checks the regmap_r/w errors.