extcon: sm5502: Clear pending interrupts during initialization
PATCH 1/1 extcon: sm5502: Clear pending interrupts during initialization
On some devices (e.g. Samsung Galaxy A5 (2015)), the bootloader seems to keep interrupts enabled for SM5502 when booting Linux. Changing the cable state (i.e. plugging in a cable) - until the driver is loaded - will therefore produce an interrupt that is never read.
In this situation, the cable state will be stuck forever on the initial state because SM5502 stops sending interrupts. This can be avoided by clearing those pending interrupts after the driver has been loaded.
Reading the interrupt status registers twice seems to be sufficient to make interrupts work in this situation.
---
This makes interrupts work on the Samsung Galaxy A5 (2015), which has recently gained mainline support [1].
I was not able to find a datasheet for SM5502, so this patch is merely based on testing and comparison with the downstream driver [2].
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=1329c1ab0730b521e6cd3051c56a2ff3d55f21e6
[2]: https://github.com/msm8916-mainline/android_kernel_qcom_msm8916/blob/SM-A500FU/drivers/misc/sm5502.c#L1566-L1578
--- drivers/extcon/extcon-sm5502.c | 6 ++++++ 1 file changed, 6 insertions(+)
drivers/extcon/extcon-sm5502.c +6 -0
@@ -539,6 +539,12 @@ static void sm5502_init_dev_type(struct sm5502_muic_info *info) val = info->reg_data[i].val;regmap_write (info->regmap, info->reg_data[i].reg, val); } + +/* Clear pending interrupts */ +regmap_read (info->regmap,SM5502_REG_INT1 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT2 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT1 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT2 , ®_data); } static int sm5022_muic_i2c_probe(struct i2c_client *i2c,
Signature · 1 line
2.23.0
Chanwoo Choi
maintainerQuoted 32 lines · Stephan Gerhold
On some devices (e.g. Samsung Galaxy A5 (2015)), the bootloader seems to keep interrupts enabled for SM5502 when booting Linux. Changing the cable state (i.e. plugging in a cable) - until the driver is loaded - will therefore produce an interrupt that is never read.
In this situation, the cable state will be stuck forever on the initial state because SM5502 stops sending interrupts. This can be avoided by clearing those pending interrupts after the driver has been loaded.
Reading the interrupt status registers twice seems to be sufficient to make interrupts work in this situation.
Signed-off-by Stephan Gerhold <[email protected]>---
This makes interrupts work on the Samsung Galaxy A5 (2015), which has recently gained mainline support [1].I was not able to find a datasheet for SM5502, so this patch is merely based on testing and comparison with the downstream driver [2].
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=1329c1ab0730b521e6cd3051c56a2ff3d55f21e6
[2]: https://protect2.fireeye.com/url?k=029d0042-5ffa4464-029c8b0d-0cc47a31384a-14ac0bce09798d1f&u=https://github.com/msm8916-mainline/android_kernel_qcom_msm8916/blob/SM-A500FU/drivers/misc/sm5502.c#L1566-L1578--- drivers/extcon/extcon-sm5502.c | 6 ++++++ 1 file changed, 6 insertions(+)drivers/extcon/extcon-sm5502.c +0
-0
drivers/extcon/extcon-sm5502.c static void sm5502_init_dev_type(struct sm5502_muic_info *info) · line 547
Quoted 9 lines
drivers/extcon/extcon-sm5502.c +6
-0@@ -539,6 +539,12 @@ static void sm5502_init_dev_type(struct sm5502_muic_info *info) val = info->reg_data[i].val;regmap_write (info->regmap, info->reg_data[i].reg, val); } + +/* Clear pending interrupts */ +regmap_read (info->regmap,SM5502_REG_INT1 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT2 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT1 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT2 , ®_data);
It is not proper. Instead, initialize the SM5502_RET_INT1/2 with zero.
The reset value of SM5502_RET_INT1/2 are zero (0x00) as following:
If you can test it with h/w, please try to test it and then send the modified patch.
drivers/extcon/extcon-sm5502.c +8 -0
@@ -68,6 +68,14 @@ static struct reg_data sm5502_reg_data[] = { .reg = SM5502_REG_CONTROL, .val = SM5502_REG_CONTROL_MASK_INT_MASK, .invert = false, + }, { + .reg = SM5502_REG_INT1, + .val = SM5502_RET_INT1_MASK, + .invert = true, + }, { + .reg = SM5502_REG_INT2, + .val = SM5502_RET_INT1_MASK, + .invert = true, }, { .reg = SM5502_REG_INTMASK1, .val = SM5502_REG_INTM1_KP_MASK
drivers/extcon/extcon-sm5502.h +2 -0
@@ -93,6 +93,8 @@ enum sm5502_reg {#define SM5502_REG_CONTROL_RAW_DATA_MASK (0x1 << SM5502_REG_CONTROL_RAW_DATA_SHIFT)#define SM5502_REG_CONTROL_SW_OPEN_MASK (0x1 << SM5502_REG_CONTROL_SW_OPEN_SHIFT) +#define SM5502_RET_INT1_MASK (0xff) +#define SM5502_REG_INTM1_ATTACH_SHIFT 0#define SM5502_REG_INTM1_DETACH_SHIFT 1#define SM5502_REG_INTM1_KP_SHIFT 2
Quoted 4 lines
drivers/extcon/extcon-sm5502.c +0
-0} static int sm5022_muic_i2c_probe(struct i2c_client *i2c,
Signature · 3 lines
Best Regards, Chanwoo Choi Samsung Electronics
Chanwoo Choi
maintainerQuoted 82 lines · Chanwoo Choi
Quoted 32 lines · Stephan Gerhold
On some devices (e.g. Samsung Galaxy A5 (2015)), the bootloader seems to keep interrupts enabled for SM5502 when booting Linux. Changing the cable state (i.e. plugging in a cable) - until the driver is loaded - will therefore produce an interrupt that is never read.
In this situation, the cable state will be stuck forever on the initial state because SM5502 stops sending interrupts. This can be avoided by clearing those pending interrupts after the driver has been loaded.
Reading the interrupt status registers twice seems to be sufficient to make interrupts work in this situation.
Signed-off-by Stephan Gerhold <[email protected]>---
This makes interrupts work on the Samsung Galaxy A5 (2015), which has recently gained mainline support [1].I was not able to find a datasheet for SM5502, so this patch is merely based on testing and comparison with the downstream driver [2].
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=1329c1ab0730b521e6cd3051c56a2ff3d55f21e6
[2]: https://protect2.fireeye.com/url?k=029d0042-5ffa4464-029c8b0d-0cc47a31384a-14ac0bce09798d1f&u=https://github.com/msm8916-mainline/android_kernel_qcom_msm8916/blob/SM-A500FU/drivers/misc/sm5502.c#L1566-L1578--- drivers/extcon/extcon-sm5502.c | 6 ++++++ 1 file changed, 6 insertions(+)drivers/extcon/extcon-sm5502.c +0
-0drivers/extcon/extcon-sm5502.c static void sm5502_init_dev_type(struct sm5502_muic_info *info) · line 547
Quoted 9 lines
drivers/extcon/extcon-sm5502.c +6
-0@@ -539,6 +539,12 @@ static void sm5502_init_dev_type(struct sm5502_muic_info *info) val = info->reg_data[i].val;regmap_write (info->regmap, info->reg_data[i].reg, val); } + +/* Clear pending interrupts */ +regmap_read (info->regmap,SM5502_REG_INT1 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT2 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT1 , ®_data); +regmap_read (info->regmap,SM5502_REG_INT2 , ®_data);It is not proper. Instead, initialize the SM5502_RET_INT1/2 with zero.
The reset value of SM5502_RET_INT1/2 are zero (
0x00) as following:
If you can test it with h/w, please try to test it and then send the modified patch.drivers/extcon/extcon-sm5502.c +8
-0@@ -68,6 +68,14 @@ static struct reg_data sm5502_reg_data[] = { .reg = SM5502_REG_CONTROL, .val = SM5502_REG_CONTROL_MASK_INT_MASK, .invert = false, + }, { + .reg = SM5502_REG_INT1, + .val = SM5502_RET_INT1_MASK, + .invert = true, + }, { + .reg = SM5502_REG_INT2, + .val = SM5502_RET_INT1_MASK, + .invert = true, }, { .reg = SM5502_REG_INTMASK1, .val = SM5502_REG_INTM1_KP_MASKdrivers/extcon/extcon-sm5502.h +2
-0@@ -93,6 +93,8 @@ enum sm5502_reg {#define SM5502_REG_CONTROL_RAW_DATA_MASK (0x1 << SM5502_REG_CONTROL_RAW_DATA_SHIFT)#define SM5502_REG_CONTROL_SW_OPEN_MASK (0x1 << SM5502_REG_CONTROL_SW_OPEN_SHIFT) +#define SM5502_RET_INT1_MASK (0xff) +#define SM5502_REG_INTM1_ATTACH_SHIFT 0#define SM5502_REG_INTM1_DETACH_SHIFT 1#define SM5502_REG_INTM1_KP_SHIFT 2
Quoted 6 lines
drivers/extcon/extcon-sm5502.c static void sm5502_init_dev_type(struct sm5502_muic_info *info) · line 550
Quoted 4 lines
drivers/extcon/extcon-sm5502.c +0
-0} static int sm5022_muic_i2c_probe(struct i2c_client *i2c,
When write 0x1 to SM5502_REG_RESET, reset the device.
So, you can reset the all registers by writing 1 to SM5502_REG_RESET as following:
drivers/extcon/extcon-sm5502.c +13 -0
@@ -65,9 +65,22 @@ struct sm5502_muic_info {/* Default value of SM5502 register to bring up MUIC device. */ static struct reg_data sm5502_reg_data[] = { { + { + .reg =SM5502_REG_RESET , + .val =SM5502_REG_RESET_MASK , + .invert = true, + }, { .reg =SM5502_REG_CONTROL , .val =SM5502_REG_CONTROL_MASK_INT_MASK , .invert = false, + }, { + .reg =SM5502_REG_INT1 , + .val =SM5502_REG_INT1_MASK , + .invert = true, + }, { + .reg =SM5502_REG_INT2 , + .val =SM5502_REG_INT1_MASK , + .invert = true, }, { .reg =SM5502_REG_INTMASK1 , .val =SM5502_REG_INTM1_KP_MASK
drivers/extcon/extcon-sm5502.h +4 -0
@@ -93,6 +93,8 @@ enum sm5502_reg {#define SM5502_REG_CONTROL_RAW_DATA_MASK (0x1 << SM5502_REG_CONTROL_RAW_DATA_SHIFT)#define SM5502_REG_CONTROL_SW_OPEN_MASK (0x1 << SM5502_REG_CONTROL_SW_OPEN_SHIFT) +#define SM5502_REG_INT1_MASK (0xff) +#define SM5502_REG_INTM1_ATTACH_SHIFT 0#define SM5502_REG_INTM1_DETACH_SHIFT 1#define SM5502_REG_INTM1_KP_SHIFT 2
@@ -237,6 +239,8 @@ enum sm5502_reg {#define DM_DP_SWITCH_UART ((DM_DP_CON_SWITCH_UART <<SM5502_REG_MANUAL_SW1_DP_SHIFT) \ | (DM_DP_CON_SWITCH_UART <<SM5502_REG_MANUAL_SW1_DM_SHIFT)) +#define SM5502_RER_RESET_MASK (0x1) +/* SM5502 Interrupts */ enum sm5502_irq {/* INT1 */
Signature · 3 lines
Best Regards, Chanwoo Choi Samsung Electronics
Hi Chanwoo,
thank you for your suggestions!
Quoted 2 lines · Chanwoo Choi
Quoted 1 line · Chanwoo Choi
It is not proper. Instead, initialize the SM5502_RET_INT1/2 with zero.
Sorry about that. I don't have a datasheet, so I wasn't sure what's the best way to fix the problem.
Quoted 38 lines
Quoted 38 lines
The reset value of SM5502_RET_INT1/2 are zero (
0x00) as following:
If you can test it with h/w, please try to test it and then send the modified patch.drivers/extcon/extcon-sm5502.c +8
-0@@ -68,6 +68,14 @@ static struct reg_data sm5502_reg_data[] = { .reg = SM5502_REG_CONTROL, .val = SM5502_REG_CONTROL_MASK_INT_MASK, .invert = false, + }, { + .reg = SM5502_REG_INT1, + .val = SM5502_RET_INT1_MASK, + .invert = true, + }, { + .reg = SM5502_REG_INT2, + .val = SM5502_RET_INT1_MASK, + .invert = true, }, { .reg = SM5502_REG_INTMASK1, .val = SM5502_REG_INTM1_KP_MASKdrivers/extcon/extcon-sm5502.h +2
-0@@ -93,6 +93,8 @@ enum sm5502_reg {#define SM5502_REG_CONTROL_RAW_DATA_MASK (0x1 << SM5502_REG_CONTROL_RAW_DATA_SHIFT)#define SM5502_REG_CONTROL_SW_OPEN_MASK (0x1 << SM5502_REG_CONTROL_SW_OPEN_SHIFT) +#define SM5502_RET_INT1_MASK (0xff) +#define SM5502_REG_INTM1_ATTACH_SHIFT 0#define SM5502_REG_INTM1_DETACH_SHIFT 1#define SM5502_REG_INTM1_KP_SHIFT 2
Quoted 6 lines
Quoted 6 lines
Quoted 4 lines
}
static int sm5022_muic_i2c_probe(struct
i2c_client*i2c,
This patch (i.e. writing to SM5502_REG_INT1 and SM5502_REG_INT2) does not result in any difference.
There are still no interrupts being sent.
On the other hand, your other suggestion fixes the problem:
Quoted 4 lines
When write
0x1toSM5502_REG_RESET, reset the device.
So, you can reset the all registers by writing 1 toSM5502_REG_RESETas following:
Writing 0x1 to SM5502_REG_RESET seems to make interrupts work, so writing to SM5502_REG_INT1 and SM5502_REG_INT2 is not even necessary in this case.
Would you still prefer initializing SM5502_REG_INT1/2 or is the patch below enough?
drivers/extcon/extcon-sm5502.c +4 -0
@@ -65,6 +65,10 @@ struct sm5502_muic_info {/* Default value of SM5502 register to bring up MUIC device. */ static struct reg_data sm5502_reg_data[] = { { + .reg =SM5502_REG_RESET , + .val =SM5502_REG_RESET_MASK , + .invert = true, + }, { .reg =SM5502_REG_CONTROL , .val =SM5502_REG_CONTROL_MASK_INT_MASK , .invert = false,
drivers/extcon/extcon-sm5502.h +2 -0
@@ -237,6 +237,8 @@ enum sm5502_reg {#define DM_DP_SWITCH_UART ((DM_DP_CON_SWITCH_UART <<SM5502_REG_MANUAL_SW1_DP_SHIFT) \ | (DM_DP_CON_SWITCH_UART <<SM5502_REG_MANUAL_SW1_DM_SHIFT)) +#define SM5502_REG_RESET_MASK (0x1) +/* SM5502 Interrupts */ enum sm5502_irq {/* INT1 */
Chanwoo Choi
maintainerHi Stephan,
Quoted 63 lines · Stephan Gerhold
Hi Chanwoo,
thank you for your suggestions!
Quoted 2 lines · Chanwoo Choi
Quoted 1 line · Chanwoo Choi
It is not proper. Instead, initialize the SM5502_RET_INT1/2 with zero.
Sorry about that. I don't have a datasheet, so I wasn't sure what's the best way to fix the problem.
Quoted 38 lines
Quoted 38 lines
The reset value of SM5502_RET_INT1/2 are zero (
0x00) as following:
If you can test it with h/w, please try to test it and then send the modified patch.drivers/extcon/extcon-sm5502.c +8
-0@@ -68,6 +68,14 @@ static struct reg_data sm5502_reg_data[] = { .reg = SM5502_REG_CONTROL, .val = SM5502_REG_CONTROL_MASK_INT_MASK, .invert = false, + }, { + .reg = SM5502_REG_INT1, + .val = SM5502_RET_INT1_MASK, + .invert = true, + }, { + .reg = SM5502_REG_INT2, + .val = SM5502_RET_INT1_MASK, + .invert = true, }, { .reg = SM5502_REG_INTMASK1, .val = SM5502_REG_INTM1_KP_MASKdrivers/extcon/extcon-sm5502.h +2
-0@@ -93,6 +93,8 @@ enum sm5502_reg {#define SM5502_REG_CONTROL_RAW_DATA_MASK (0x1 << SM5502_REG_CONTROL_RAW_DATA_SHIFT)#define SM5502_REG_CONTROL_SW_OPEN_MASK (0x1 << SM5502_REG_CONTROL_SW_OPEN_SHIFT) +#define SM5502_RET_INT1_MASK (0xff) +#define SM5502_REG_INTM1_ATTACH_SHIFT 0#define SM5502_REG_INTM1_DETACH_SHIFT 1#define SM5502_REG_INTM1_KP_SHIFT 2Quoted 6 lines
Quoted 6 lines
Quoted 4 lines
}
static int sm5022_muic_i2c_probe(struct
i2c_client*i2c,This patch (i.e. writing to
SM5502_REG_INT1andSM5502_REG_INT2) does not result in any difference.
There are still no interrupts being sent.On the other hand, your other suggestion fixes the problem:
Quoted 1 line
Quoted 6 lines
Quoted 3 lines
When write
0x1toSM5502_REG_RESET, reset the device.
So, you can reset the all registers by writing 1 toSM5502_REG_RESETas following:Writing
0x1toSM5502_REG_RESETseems to make interrupts work, so writing toSM5502_REG_INT1andSM5502_REG_INT2is not even necessary in this case.
OK.
Quoted 3 lines
Would you still prefer initializing SM5502_REG_INT1/2 or is the patch below enough?
No. If SM5502_REG_RESET is enough to fix this issue, I'm OK. Thanks.
Quoted 31 lines
drivers/extcon/extcon-sm5502.c +4
-0@@ -65,6 +65,10 @@ struct sm5502_muic_info {/* Default value of SM5502 register to bring up MUIC device. */ static struct reg_data sm5502_reg_data[] = { { + .reg =SM5502_REG_RESET , + .val =SM5502_REG_RESET_MASK , + .invert = true, + }, { .reg =SM5502_REG_CONTROL , .val =SM5502_REG_CONTROL_MASK_INT_MASK , .invert = false,drivers/extcon/extcon-sm5502.h +2
-0@@ -237,6 +237,8 @@ enum sm5502_reg {#define DM_DP_SWITCH_UART ((DM_DP_CON_SWITCH_UART <<SM5502_REG_MANUAL_SW1_DP_SHIFT) \ | (DM_DP_CON_SWITCH_UART <<SM5502_REG_MANUAL_SW1_DM_SHIFT)) +#define SM5502_REG_RESET_MASK (0x1) +/* SM5502 Interrupts */ enum sm5502_irq {/* INT1 */
Signature · 3 lines
Best Regards, Chanwoo Choi Samsung Electronics