therm_bulk_read not working when other (non w1_therm) w1 devices are present

[bublath]

While testing multiple one-wire devices I detected the issue that
echo trigger >therm_bulk_read
that typically will take ~1 second (still don't understand why it can't do this asynchronously and return right away) will return immediately and not trigger conversions (cat w1_slave on any device takes ~1 second which with working bulk_read will returns instantly).
I could reproduce this with both adding a DS2423 and a DS2438 device (single or both). Once the devices get removed it starts working again.

I'm using kernel 5.10.63-v7l+ on a fully updated Raspbian buster on a Raspberry Pi 4

[akira215]

I assume that the bulk read command is colliding with other device activity. Sysfs could not be asynchronous, if required, you need to access direct register.

[bublath]

If it would collide with activity, it would probably sometimes work and sometimes not, wouldn't it?
For me it is always failing, which makes me rather assume that bulk_read tries to access other devices than w1_therm and fails (since they don't support this).

As for asyncronous:
My understanding of bulk_read is, that it puts a trigger and you then need to read therm_bulk_read to find out if all your conversions are finished. However even putting the trigger takes almost 1s (pretty much the conversion time with precision 12) before the "echo trigger > therm_bulk_read" returns. Shouldn't this just trigger the conversions and return instantly?

• or is triggering the conversions already taking so much time?
• or does it wait for the conversions to be finished (but what's the point of the query option then?)

Just trying to understand this (I need to work around this anyway).