udev settle

[edgardogho]

Hi, I have this weird problem using TC5.2
On my .xsession file I added:
sudo udevadm control --reload-rules
sudo udevadm trigger --action=add
sudo udevadm settle

The settle part hangs at least one out of 3 boots and shows unprocessed stuff in the queue (there is no pattern on the stuff, its not always the same thing, and not even the same size, could be just a few events, just one, or over 30).

I realized that if I kill udev (with sudo kill -9 $(pidof udevd) ) and relaunch it with sudo udevd --daemon, after reload/trigger/settle it works most of the time (1 out of 10 hangs at settle again, but usually with just a couple of events in the queue). If I do it a third time (kill udev, reload/retrigger/settle) it wont' fail.

I'm using kernel 3.12.15 (compiled myself).
Udevd version is 173. I'm adding some rules of my own but I don't think they are the problem in here since they seem to work some of the cases and after killing udev and reloading/retriggering/settle it hardly ever fails.

I think there is some race condition or something like that. I also noticed that if I do ps ax when the settle hangs, there is usually a lot of udevd processes running (I Believe these are worker threads).

Did someone had a similar problem?

[edgardogho]

I can confirm that this will also happen on the 3.8 kernel used on TC5.2
It took about 60 reboots but on the sixtieths reboot it hanged at udevadm settle and printed some stuff in the queue.

[edgardogho]

I ended up compiling udev 174 and that seems work. No udevadm settle hangs in over 40 reboots.