hdtodd@eagle.wesleyan.edu (01/24/91)
This is a followup answer to a question I posted a couple of weeks ago. Running 10.2, BSD4.3 on a 2500 disked node supporting two diskless nodes, I set up the systems to run with xownsroot and to start up uwm from startup_login. The login process hung for non-root users but succeeded for root on the diskless nodes but worked just fine for both on the disked nodes. After spending a couple of days working on the problem, I asked for insights. I didn't get any responses, so this must be a rare environment. In case anyone else runs into the problem, here was the cause and solution for mine. During bootup, the rc file clears /tmp. Apparently, uwm needs a file /tmp/X11-unix/X0 --- and it needs for that file to be writable by the display user. Unfortunately, it seems to create it srwxr-xr-x (at least on the diskless nodes). So, root works just fine, others hang on login (uwm hangs). My fix, as a novice UNIX user, was to create the directory and file at the end of the rc script and to set the protection to 4777. There must be a cleaner way than this, but it works ... and I'd spent more time on this than I could afford as it was. Suggestions for cleaner fixes (or obvious mistakes in my setup of the system) would be welcome. I hope this helps someone else considering this environment.
krowitz@RICHTER.MIT.EDU (David Krowitz) (01/25/91)
An oberservation that may help you ... The /tmp directory is actually a link to `node_data/tmp on Domain/OS systems. For machines with their own boot disk, this translates into the directory //my_disk/sys/node_data. For diskless machines, this translates into the //my_partners_disk/sys/node_data.my_node_id/tmp directory (sigh, I missed the "/tmp" on the end of the first example ...). The default file protection for files created in /tmp is taken from the ACL of the /sys/node_data/tmp directory for machines with disks and from the /sys/node_data.my_node_id/tmp directory for machines without disks. The ACLs for the /sys/node_data directory (and it's subdirectories) should have been setup when the OS was installed; but the ACL's for the /sys/node_data.my_node_id directory used by the diskless machines are set up when the diskless machine is first booted off of the partner according to the /sys/net/netman.rc shell script. This is usually a link to either netman.bin_sh or netman.com_sh, which results in the diskless node's /tmp directory receiving either Unix-style or Aegis style ACL's when the diskless node is first booted. My guess is that the ACL's on you /sys/node_data/tmp directory are different from those on your /sys/node_data.my_node_id/tmp directory, and that the culprit is the /sys/net/netman.rc file either being linked to the wrong environment, or simply having the wrong file protections being set by the file. -- David Krowitz krowitz@richter.mit.edu (18.83.0.109) krowitz%richter.mit.edu@eddie.mit.edu krowitz%richter.mit.edu@mitvma.bitnet (in order of decreasing preference)