The stock devfsd.conf file installed during a stage1-3 build on a Ultrasparc30 (sparc64) sets up devfsd.conf for a ps2 mouse (/dev/psaux). AFAIK most sparcs don't have PS2 ports. Due to this setup X will NOT work without manually editing hte devfsd.conf file and restarting devfsd, and deleting the symlink in /dev (/dev/mouse) upon deletion the devfsd daemon will recreate the proper link. The proper lines for devfsd.conf are: # Create /dev/mouse LOOKUP ^mouse$ CFUNCTION GLOBAL mksymlink misc/sunmouse mouse REGISTER ^misc/sunmouse$ CFUNCTION GLOBAL mksymlink $devname mouse UNREGISTER ^misc/sunmouse$ CFUNCTION GLOBAL unlink mouse Reproducible: Always Steps to Reproduce: 1.install gentoo on a sparc64 (ultra 30/60/80 tested) 2.Attempt to run X 3. Actual Results: X dies complaing that the mouse is not accessible. Cimpley deleting /dev/mouse and relinking it to the proper place WILL NOT WORK due to devfsd recreating the link inappropriately as described above. Expected Results: installation of baselayout should have setup /etc/devfsd.conf properly for the hardware. Perhaps a simple shell script to query entries in /proc or dmesg to determine if hte machien has pcmcia ports would be enough to allow the installation of baselayout on sparc to determine which devfsd.conf entries should be used. (AFAIK newer sparc hardware may have PS2 ports whereas older ones did not) System is an UltraSparc30 running gentoo-1.4RC4 (the perpetual RC canidate) KB/Mouse is a standard sun keyboard with the mouse chained off of it. (non optical mouse)
An additional note about this. On some sparc64 models (namely the Blade series), they use USB mice rather than the traditional Sun busmouse. This ends up showing up as /dev/usbmouse. So I'm guessing some sed-fu is in order for sparc64 to check for the existance of either /dev/sunmouse or /dev/usbmouse and set /dev/mouse accordingly. Also users can add an USB card to PCI based sparc64s, which can add to the confusion and cause both to be present, but for now I'd just go with the above until somebody brings up the issue. Also this can be applied to sparc as well, though the USB related issues don't exist there (as there are no sbus USB controllers that I'm aware of, nor did Sun ship a sparc with USB).
Compiling the Kernel with 'sunmouse' support (CONFIG_SUN_MOUSE) provides /dev/sunmouse. I never had to change devfsd.conf on my sparc64 boxes (U5, U1) so I cant reproduce this error.
Dave, does this problem still exist for you?
This guy's issue is that devfsd.conf isn't setting /dev/mouse to point to whatever device his sun mouse happens to be. It could be /dev/sunmouse, /dev/misc/sunmouse, /dev/usbmouse, etc. I'm not sure that this is really a bug, so much as something a user should be configuring on thier system (all my X configs point to /dev/misc/sunmouse)... So the question is, is this a bug, if so then lets make a patch and get it resolved, if not lets resolve it otherwise...
resolved. (user error)