Skip to contentSkip to conversation

acpi_power_meter: clean up duplicate code

PATCH In review Kyle McMartin 3 participants 4 messages Apr 2, 2012 – Apr 4, 2012

lore
Share & tools
lore
#0 lore raw

cover acpi_power_meter: clean up duplicate code

From: Kyle McMartin <[email protected]>

acpi_power_meter currently duplicates a lot of code between the rw and ro sensor templates unnecessarily. The attached five part series cleans things up to use the same code path for both, deciding which is which by the use of the .set field.

Additionally, (imho at least) it makes the struct initializers a bit more pretty and uses a macro to reduce the repetition of each member.

It's been compile tested across the series for bisectability, but I do not currently have access to any device which uses this driver for testing.

regards, Kyle McMartin

Kyle McMartin (5):
acpi_power_meter: use the same struct {rw,ro}_sensor_template for
both
acpi_power_meter: use a {RW,RO}_SENSOR_TEMPLATE macro to clean things
up
acpi_power_meter: remove duplicate code between
register_{ro,rw}_attrs
acpi_power_meter: drop meter_rw_attrs, use common meter_attrs
acpi_power_meter: clean up code around setup_attrs

 drivers/hwmon/acpi_power_meter.c |  165 +++++++++++++++++---------------------
 1 files changed, 72 insertions(+), 93 deletions(-)
Signature · 7 lines
1.7.7.6


_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
#1 lore raw
Quoted 26 lines · Kyle McMartin

From: Kyle McMartin <[email protected]>

acpi_power_meter currently duplicates a lot of code between the rw and ro sensor templates unnecessarily. The attached five part series cleans things up to use the same code path for both, deciding which is which by the use of the .set field.

Additionally, (imho at least) it makes the struct initializers a bit more pretty and uses a macro to reduce the repetition of each member.

It's been compile tested across the series for bisectability, but I do not currently have access to any device which uses this driver for testing.

regards, Kyle McMartin

Kyle McMartin (5):
acpi_power_meter: use the same struct {rw,ro}_sensor_template for
both
acpi_power_meter: use a {RW,RO}_SENSOR_TEMPLATE macro to clean things
up
acpi_power_meter: remove duplicate code between
register_{ro,rw}_attrs
acpi_power_meter: drop meter_rw_attrs, use common meter_attrs
acpi_power_meter: clean up code around setup_attrs

Quoted 3 lines
 drivers/hwmon/acpi_power_meter.c |  165 +++++++++++++++++---------------------
 1 files changed, 72 insertions(+), 93 deletions(-)

Hi Kyle,

nice cleanup series.

Couple of comments / questions. That won't prevent me from applying the series, just some things to think about.

- You could have used a single macro instead of two, with _set set to NULL for read-only attributes.
- Why not just use SENSOR_DEVICE_ATTR, struct attribute, and sysfs_create_group/sysfs_remove_group ?
- I think there may be a race condition in acpi_power_meter_notify(), in the METER_NOTIFY_CONFIG path. It frees some data, re-allocates it, then removes the attributes and re-creates them. If someone were to read the power1_model_number attribute after the call to free_capabilities, show_str() might potentially access freed memory.

Thanks,
Guenter

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#2 lore raw
In reply to Kyle McMartin
Quoted 5 lines · Guenter Roeck

nice cleanup series.

Couple of comments / questions. That won't prevent me from applying the series, just some things to think about.

Quoted 7 lines

- You could have used a single macro instead of two, with _set set to NULL for read-only attributes.
- Why not just use SENSOR_DEVICE_ATTR, struct attribute, and sysfs_create_group/sysfs_remove_group ?
- I think there may be a race condition in acpi_power_meter_notify(), in the METER_NOTIFY_CONFIG path. It frees some data, re-allocates it, then removes the attributes and re-creates them. If someone were to read the power1_model_number attribute after the call to free_capabilities, show_str() might potentially access freed memory.

Thanks for the review, Guenter. I'll take a look at some of these other fixes this week.

regards, Kyle

Quoted 3 lines

Thanks,
Guenter

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#3 lore raw
In reply to Kyle McMartin
Quoted 26 lines · Kyle McMartin

From: Kyle McMartin <[email protected]>

acpi_power_meter currently duplicates a lot of code between the rw and ro sensor templates unnecessarily. The attached five part series cleans things up to use the same code path for both, deciding which is which by the use of the .set field.

Additionally, (imho at least) it makes the struct initializers a bit more pretty and uses a macro to reduce the repetition of each member.

It's been compile tested across the series for bisectability, but I do not currently have access to any device which uses this driver for testing.

regards, Kyle McMartin

Kyle McMartin (5):
acpi_power_meter: use the same struct {rw,ro}_sensor_template for
both
acpi_power_meter: use a {RW,RO}_SENSOR_TEMPLATE macro to clean things
up
acpi_power_meter: remove duplicate code between
register_{ro,rw}_attrs
acpi_power_meter: drop meter_rw_attrs, use common meter_attrs
acpi_power_meter: clean up code around setup_attrs

Quoted 3 lines
 drivers/hwmon/acpi_power_meter.c |  165 +++++++++++++++++---------------------
 1 files changed, 72 insertions(+), 93 deletions(-)

Series applied to -next.

Thanks,
Guenter

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors