Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1405579 > unrolled thread

[RFC 0/7] Deal with iio trigger names

Back to article view | Back to linux.kernel

Contents

  [RFC 0/7] Deal with iio trigger names Crestez Dan Leonard <leonard.crestez@intel.com> - 2016-05-23 20:50 +0200
    Re: [RFC 0/7] Deal with iio trigger names Jonathan Cameron <jic23@kernel.org> - 2016-05-29 21:50 +0200

#1405579 — [RFC 0/7] Deal with iio trigger names

IIO documents that trigger names are unique but does not actually guarantee
this. You can easily create a software trigger with a duplicate name if you
enable CONFIG_IIO_HRTIMER_TRIGGER:

	mkdir /sys/kernel/config/iio/triggers/hrtimer/\
		`cat /sys/bus/iio/devices/trigger0/name`

You can also get in this situation if you connect two identical st_sensors.

My attempt to fix this is by adding a new 'current_trigger_id' ABI which is a
numeric ID (X from triggerX) rather than a non-unique string. I tested that
this works as expected when connecting two LIS3DH to the same system. This is
in patches 4,5.

In theory the current_trigger ABI could be overloaded to looks for something
matching "trigger[0-9]+" but this would be nastier. There would be a need to
prevent registering triggers that match that expression in order to prevent
stuff like:

	mkdir /sys/kernel/config/iio/triggers/hrtimer/trigger1

Such an overload wouldn't work on read and make it impossible to unambigiously
determine the currently selected trigger.


Patch 6 disallows registering duplicate trigger names. This makes drivers
which currently do that break at probe time.

Patch 7 attempts to ensure that all drivers include indio_dev->id when calling
iio_trigger_alloc. This obviously changes trigger names.


Either 4,5 or 6,7 fix this issue, but 4,5 causes fewer compat issues.


Patches 1,2,3 are just minor features in tools/generic_buffer. They supersede
the v2 I posted earlier:

	https://www.spinics.net/lists/linux-iio/msg24867.html

Crestez Dan Leonard (7):
  iio: generic_buffer: Cleanup when receiving signals
  iio: generic_buffer: Add --device-num option
  iio: generic_buffer: Add --trigger-num option
  iio: Add current_trigger_id alternative
  iio: generic_buffer: Use current_trigger_id
  iio: Refuse to register triggers with duplicate names
  iio: Make trigger names unique

 Documentation/ABI/testing/sysfs-bus-iio            |   9 +
 Documentation/DocBook/iio.tmpl                     |   4 +-
 drivers/iio/adc/max1027.c                          |   4 +-
 drivers/iio/common/st_sensors/st_sensors_trigger.c |   2 +-
 drivers/iio/industrialio-trigger.c                 | 136 ++++++++---
 drivers/iio/light/gp2ap020a00f.c                   |   4 +-
 tools/iio/generic_buffer.c                         | 269 ++++++++++++++-------
 7 files changed, 309 insertions(+), 119 deletions(-)

-- 
2.5.5

[toc] | [next] | [standalone]

#1408659

On 23/05/16 19:39, Crestez Dan Leonard wrote:
> IIO documents that trigger names are unique but does not actually guarantee
> this. You can easily create a software trigger with a duplicate name if you
> enable CONFIG_IIO_HRTIMER_TRIGGER:
> 
> 	mkdir /sys/kernel/config/iio/triggers/hrtimer/\
> 		`cat /sys/bus/iio/devices/trigger0/name`
> 
> You can also get in this situation if you connect two identical st_sensors.
> 
> My attempt to fix this is by adding a new 'current_trigger_id' ABI which is a
> numeric ID (X from triggerX) rather than a non-unique string. I tested that
> this works as expected when connecting two LIS3DH to the same system. This is
> in patches 4,5.
> 
> In theory the current_trigger ABI could be overloaded to looks for something
> matching "trigger[0-9]+" but this would be nastier. There would be a need to
> prevent registering triggers that match that expression in order to prevent
> stuff like:
> 
> 	mkdir /sys/kernel/config/iio/triggers/hrtimer/trigger1
> 
> Such an overload wouldn't work on read and make it impossible to unambigiously
> determine the currently selected trigger.
> 
> 
> Patch 6 disallows registering duplicate trigger names. This makes drivers
> which currently do that break at probe time.
> 
> Patch 7 attempts to ensure that all drivers include indio_dev->id when calling
> iio_trigger_alloc. This obviously changes trigger names.
> 
> 
> Either 4,5 or 6,7 fix this issue, but 4,5 causes fewer compat issues.
> 
> 
> Patches 1,2,3 are just minor features in tools/generic_buffer. They supersede
> the v2 I posted earlier:
Please don't do this in future.  When reviewing lots of series, one tends
to relying on sorting by name to make sure you find the latest version of
a series.  Do this and people don't know there is a new version embedded in
a different series (in this case I picked up V2 when you had found an issue
in it).

Jonathan
> 
> 	https://www.spinics.net/lists/linux-iio/msg24867.html
> 
> Crestez Dan Leonard (7):
>   iio: generic_buffer: Cleanup when receiving signals
>   iio: generic_buffer: Add --device-num option
>   iio: generic_buffer: Add --trigger-num option
>   iio: Add current_trigger_id alternative
>   iio: generic_buffer: Use current_trigger_id
>   iio: Refuse to register triggers with duplicate names
>   iio: Make trigger names unique
> 
>  Documentation/ABI/testing/sysfs-bus-iio            |   9 +
>  Documentation/DocBook/iio.tmpl                     |   4 +-
>  drivers/iio/adc/max1027.c                          |   4 +-
>  drivers/iio/common/st_sensors/st_sensors_trigger.c |   2 +-
>  drivers/iio/industrialio-trigger.c                 | 136 ++++++++---
>  drivers/iio/light/gp2ap020a00f.c                   |   4 +-
>  tools/iio/generic_buffer.c                         | 269 ++++++++++++++-------
>  7 files changed, 309 insertions(+), 119 deletions(-)
> 

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web