hwmon: pmbus: Reduce clear fault page invocations
PATCH 1/3 hwmon: pmbus: Reduce clear fault page invocations
From: Patrick Rudolph <[email protected]>
Observing I2C traffic revealed consecutive transmission of CLEAR_FAULT commands. While this doesn't cause issues, it extends driver probe time.
Avoid invoking pmbus_clear_fault_page for virtual registers, as they're managed by the driver, not the chip.
TEST: Verified using an I2C bus analyzer that only one CLEAR_FAULT
command is send instead 5 in a row.
--- drivers/hwmon/pmbus/pmbus_core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/hwmon/pmbus/pmbus_core.c +2 -1
@@ -561,7 +561,8 @@ static bool pmbus_check_register(struct i2c_client *client, rv =pmbus_check_status_cml (client);if (rv <0 && (data->flags &PMBUS_READ_STATUS_AFTER_FAILED_CHECK )) data->read_status (client,-1 ); -pmbus_clear_fault_page (client,-1 ); +if (reg <PMBUS_VIRT_BASE ) +pmbus_clear_fault_page (client,-1 );return rv >=0 ; }
base-commit: fc8df28be7155ddcf8731b4c85fd3372811ba5ac
Signature · 1 line
2.41.0
PATCH 2/3 hwmon: pmbus: Drop unnecessary clear fault page
From: Patrick Rudolph <[email protected]>
The pmbus_check_byte_register function already calls clear fault page, so there's no need to do it again in pmbus_identify_common.
TEST: Verified using an I2C bus analyser to confirm that only
one clear fault page is being send instead of two in a row.
--- drivers/hwmon/pmbus/pmbus_core.c | 1 - 1 file changed, 1 deletion(-)
drivers/hwmon/pmbus/pmbus_core.c +0 -1
@@ -2541,7 +2541,6 @@ static int pmbus_identify_common(struct i2c_client *client, } } -pmbus_clear_fault_page (client, page);return 0 ; }
Signature · 1 line
2.41.0
Re: [PATCH 2/3] hwmon: pmbus: Drop unnecessary clear fault page
Quoted 11 lines · Naresh Solanki
From: Patrick Rudolph <[email protected]>
The
pmbus_check_byte_registerfunction already calls clear fault page, so there's no need to do it again inpmbus_identify_common.TEST: Verified using an I2C bus analyser to confirm that only
one clear fault page is being send instead of two in a row.Signed-off-by Patrick Rudolph <[email protected]>Signed-off-by Naresh Solanki <[email protected]>
Applied.
Thanks,
Guenter
Quoted 16 lines
--- drivers/hwmon/pmbus/pmbus_core.c | 1 - 1 file changed, 1 deletion(-)drivers/hwmon/pmbus/pmbus_core.c +0
-1@@ -2541,7 +2541,6 @@ static int pmbus_identify_common(struct i2c_client *client, } } -pmbus_clear_fault_page (client, page);return 0 ; }
PATCH 3/3 hwmon: pmbus: Fix -EIO seen on pli1209
In reply to Naresh SolankiFrom: Patrick Rudolph <[email protected]>
After doing performance optimizations the pli1209 driver failed to probe with a probabilty of 2%. It wasn't able to read the PMBUS_OPERATION register due to an -EIO error.
An investigation showed that the PLI1209 takes 230 usec to execute the CLEAR_FAULTS command. During that time it's busy and NACKs all requests on the SMBUS interface. It also NACKs reads on PMBUS_STATUS_BYTE making it impossible to poll the BUSY flag.
Add a custom write_data function to just wait for not BUSY unconditionally after sending a CLEAR_FAULTS command.
TEST: Verified using an I2C bus analyser that no more NACKs are seen after
sending a CLEAR_FAULTS command.
--- drivers/hwmon/pmbus/pli1209bc.c | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+)
drivers/hwmon/pmbus/pli1209bc.c +26 -0
@@ -5,6 +5,7 @@* Copyright (c) 2022 9elements GmbH */ +#include <linux/delay.h> #include <linux/i2c.h> #include <linux/module.h> #include <linux/pmbus.h>
@@ -53,6 +54,30 @@ static int pli1209bc_read_word_data(struct i2c_client *client, int page, } } +static int pli1209bc_write_byte (struct i2c_client *client,int page,u8 reg) +{ +int ret; + +switch (reg) { +case PMBUS_CLEAR_FAULTS : + ret =pmbus_write_byte (client, page, reg); +/* +* PLI1209 takes 230 usec to execute the CLEAR_FAULTS command. +* During that time it's busy and NACKs all requests on the +* SMBUS interface. It also NACKs reads on PMBUS_STATUS_BYTE +* making it impossible to poll the BUSY flag. +* +* Just wait for not BUSY unconditionally. +*/ +usleep_range (250 ,300 ); +break ; +default : + ret = -ENODATA ; +break ; + } +return ret; +} +#if IS_ENABLED (CONFIG_SENSORS_PLI1209BC_REGULATOR )static const struct regulator_desc pli1209bc_reg_desc = { .name = "vout2",
@@ -102,6 +127,7 @@ static struct pmbus_driver_info pli1209bc_info = { | PMBUS_HAVE_TEMP | PMBUS_HAVE_STATUS_TEMP | PMBUS_HAVE_STATUS_IOUT | PMBUS_HAVE_STATUS_INPUT, .read_word_data = pli1209bc_read_word_data, + .write_byte = pli1209bc_write_byte, #if IS_ENABLED(CONFIG_SENSORS_PLI1209BC_REGULATOR) .num_regulators =1 , .reg_desc = &pli1209bc_reg_desc,
Signature · 1 line
2.41.0
Re: [PATCH 3/3] hwmon: pmbus: Fix -EIO seen on pli1209
Quoted 11 lines · Naresh Solanki
From: Patrick Rudolph <[email protected]>
After doing performance optimizations the pli1209 driver failed to probe with a probabilty of 2%. It wasn't able to read the
PMBUS_OPERATIONregister due to an-EIOerror.An investigation showed that the PLI1209 takes 230 usec to execute the
CLEAR_FAULTScommand. During that time it's busy and NACKs all requests on the SMBUS interface. It also NACKs reads onPMBUS_STATUS_BYTEmaking it impossible to poll the BUSY flag.
Quoted 8 lines
Add a custom
write_datafunction to just wait for not BUSY unconditionally after sending aCLEAR_FAULTScommand.TEST: Verified using an I2C bus analyser that no more NACKs are seen after
sending aCLEAR_FAULTScommand.Signed-off-by Patrick Rudolph <[email protected]>Signed-off-by Naresh Solanki <[email protected]>
Applied.
Thanks,
Guenter
Quoted 55 lines
--- drivers/hwmon/pmbus/pli1209bc.c | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+)drivers/hwmon/pmbus/pli1209bc.c +26
-0@@ -5,6 +5,7 @@* Copyright (c) 2022 9elements GmbH */ +#include <linux/delay.h> #include <linux/i2c.h> #include <linux/module.h> #include <linux/pmbus.h> @@ -53,6 +54,30 @@ static int pli1209bc_read_word_data(struct i2c_client *client, int page, } } +static int pli1209bc_write_byte (struct i2c_client *client,int page,u8 reg) +{ +int ret; + +switch (reg) { +case PMBUS_CLEAR_FAULTS : + ret =pmbus_write_byte (client, page, reg); +/* +* PLI1209 takes 230 usec to execute the CLEAR_FAULTS command. +* During that time it's busy and NACKs all requests on the +* SMBUS interface. It also NACKs reads on PMBUS_STATUS_BYTE +* making it impossible to poll the BUSY flag. +* +* Just wait for not BUSY unconditionally. +*/ +usleep_range (250 ,300 ); +break ; +default : + ret = -ENODATA ; +break ; + } +return ret; +} +#if IS_ENABLED (CONFIG_SENSORS_PLI1209BC_REGULATOR )static const struct regulator_desc pli1209bc_reg_desc = { .name = "vout2",@@ -102,6 +127,7 @@ static struct pmbus_driver_info pli1209bc_info = { | PMBUS_HAVE_TEMP | PMBUS_HAVE_STATUS_TEMP | PMBUS_HAVE_STATUS_IOUT | PMBUS_HAVE_STATUS_INPUT, .read_word_data = pli1209bc_read_word_data, + .write_byte = pli1209bc_write_byte, #if IS_ENABLED(CONFIG_SENSORS_PLI1209BC_REGULATOR) .num_regulators =1 , .reg_desc = &pli1209bc_reg_desc,
Quoted 5 lines · Naresh Solanki
From: Patrick Rudolph <[email protected]>
Observing I2C traffic revealed consecutive transmission of
CLEAR_FAULTcommands. While this doesn't cause issues, it extends driver probe time.
Quoted 8 lines
Avoid invoking
pmbus_clear_fault_pagefor virtual registers, as they're managed by the driver, not the chip.TEST: Verified using an I2C bus analyzer that only one
CLEAR_FAULT
command is send instead 5 in a row.Signed-off-by Patrick Rudolph <[email protected]>Signed-off-by Naresh Solanki <[email protected]>
Applied.
Thanks,
Guenter
Quoted 21 lines
--- drivers/hwmon/pmbus/pmbus_core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)base-commit:
fc8df28be7155ddcf8731b4c85fd3372811ba5acdrivers/hwmon/pmbus/pmbus_core.c +2
-1@@ -561,7 +561,8 @@ static bool pmbus_check_register(struct i2c_client *client, rv =pmbus_check_status_cml (client);if (rv <0 && (data->flags &PMBUS_READ_STATUS_AFTER_FAILED_CHECK )) data->read_status (client,-1 ); -pmbus_clear_fault_page (client,-1 ); +if (reg <PMBUS_VIRT_BASE ) +pmbus_clear_fault_page (client,-1 );return rv >=0 ; }