That happens because there is a short period of time between the loading of modules in /etc/modules.autoload and the apparition of device nodes in /dev/input. On my system /dev/input/mouse0 appears after "/etc/init.d/gpm start" is run. So that gpm fails with oops /dev/input/mouse0 not found. After this, logging restarting gpm works fine. my uggly workaround is a /bin/sleep 1 just before gpm is run in /etc/init.d/gpm -- nohar
a note about this, it isn't really a devfs issue, it is just a USB issue, the USB controller takes about that long to probe devices once it is activated.
Created attachment 1335 [details] /etc/init.d/gpm new version This script detects (hopefully) if we are using a dynamically loaded mouse device, then attempts to wait for that device to come online for 3 seconds, and then attempts to start gpm regardless. Also updated to properly detect the exit status of gpm (gpm doesn't return 1 on a failed device open like a civilized program would)
Could work. Why the second sleep loop (after gpm start)? Also, it should prob test for USB as well. To be honest though, I am not sure this is the correct fix. If this is a universal problem with USB, we should rather fix it USB side .. maybe a USB rc-script that handles the module probing and sleep if needed.
This works fine for me with current kernels.