Am 28.09.2016 um 17:12 schrieb Tony Lindgren [off-list ref]:

* H. Nikolaus Schaller [off-list ref] [160927 23:50]:

quoted

Hi,

quoted

Am 28.09.2016 um 05:37 schrieb Tony Lindgren [off-list ref]:

* H. Nikolaus Schaller [off-list ref] [160927 13:11]:

quoted

quoted

Am 27.09.2016 um 21:49 schrieb Tony Lindgren [off-list ref]:
How about this for defaults:

- heartbeat for led3
- cpu0 for led4
- cpu1 for led5

Good idea. Will try.

What I don't exactly know is if these gpios based on an I2C-expander
can handle cpu activity triggers or if they are locked up if this i2c
processing triggers another cpu activity...

Oh right, if the GPIOs are on the i2c bus

yes, they are all gpio9 which is the tca6424 on i2c5.

quoted

it's probably not a good
idea :) Or at least will be inaccurate if the bus can sleep.

I have tested a little by writing to /sys/class/leds/.../trigger.

If I make one LED trigger by "cpu0" it is always on and if I make
a second one triggered by "cpu1" the heartbeat becomes irregular.

So indeed this does not work.

So should we just leave the default trigger unpopulated then?

It is a matter of taste and something not easy to decide...

And everything can be changed from user-space by some script
running during the boot process to a more meaningful setting.

So the initial setting doesn't matter really much.

Having some of the LEDs as heartbeat gives an (additional)
feedback that the kernel has booted.

We could even take led1, led3 and led5 as heartbeat and the other
two as default-on. Would give a nice pattern showing that the LEDs
are not broken...

So let me know how you would like to have it and I add it to a PATCH v2.

BR and thanks,
Nikolaus