On 24.02.2016 21:43, Alison Schofield wrote:

Daniel,

I'm working on IIO coding task #2.

In 'auditing IIO device state updates to locking' I believe that
in addition to "Updates to state changes must always take mlock",
any checks of device state must also take mlock.  And that is,
what you are asking us to do in this task.
(iio_buffer_enabled is checking device state, not setting)

Basically, anywhere iio_dev->currentmode is set|checked use mlock.

Right track?

Correct.

For this task, I'm not clear on why we are doing RFC.  In practice, 
mlock is used very consistently for protecting device state.
Can you clarify the controversial aspect?

As agreed on this thread: http://www.spinics.net/lists/linux-iio/msg18540.html
we will try to propose some API like:

iio_device_claim_direct_mode
iio_device_release_direct_mode

which will try to make code easier to read and hide the inner details of taking/releasing mlock.

We will make this in two steps:

* introduce API
* find a place where locking is not done
* use new API

After IIO maintainer says this is OK:
* update all places with the new API.

thanks,
Daniel.