matt@iquery.UUCP (Matt Reedy) (10/26/89)
We recently upgraded our 3B2/300 from 2MB of RAM to 4MB. Since we did that, the machine seems to periodically (3 or 4 times a day) go into an infinite loop for anywhere from 10 seconds to 2 minutes, even when there is only one user, on the console. Pressing "break" or "delete" does nothing until the machine decides to come back from la-la land. However, pressing "return" does cause new-lines to be echoed to the terminal (indicating that at least something is awake in there). I did notice that our swap partition is only 6000 blocks (3MB). Could this be too small now that we have upgraded the RAM? We have never seen any messages on the console indicating "out of swap space" or some such. Any and all suggestions are greatly appreciated. -- Matthew Reedy UUCP: cs.utexas.edu!swrinde!dpmizar!iquery!matt Programmed Intelligence Corp. 400 N Loop 1604 E, Suite 100 San Antonio, TX 78232 (512) 490 6684 Fax: (512) 490-3590
rich@jolnet.ORPK.IL.US (Rich Andrews) (10/27/89)
In article <170@iquery.UUCP> matt@iquery.UUCP (Matt Reedy) writes: >We recently upgraded our 3B2/300 from 2MB of RAM to 4MB. Since we did that, the >machine seems to periodically (3 or 4 times a day) go into an infinite loop >for anywhere from 10 seconds to 2 minutes, even when there is only one user, >on the console. Pressing "break" or "delete" does nothing until the machine >decides to come back from la-la land. However, pressing "return" does cause >new-lines to be echoed to the terminal (indicating that at least something >is awake in there). It appears that the real problem is a bug in the hard disk driver. here is the problem and solution..... Some 3B2's with two integral disk drives occasionally "hang" for anywhere between 10 seconds and a few hours. Then, for no obvious reason, the hang will clear itself up. Closer inspection reveals that processes are hung on one of the integral disk drives, usually the second one. Not all processes are hung, though it usually seems like it to the users. A patch to the integral disk driver accompanies this README. This is the second published patch for this problem. The original patch would not eliminate the hangs entirely, but would reduce their duration to at most 20 seconds. THIS patch prevents the hangs entirely. This patch is for UNIX System V Release 3.1 on a 3B2. It includes one file besides the README. $ ls -l -rwxrwxr-x 1 rbc ucs 36248 Oct 27 17:21 idisk.o $ sum -r idisk.o 05778 71 idisk.o To install: Login as root. Save a copy of the original driver and /unix (Hopefully you have room): # mv /boot/IDISK / # cp /unix /unix.old Note: Do NOT save a copy of IDISK under a different name in the /boot directory! Use mkboot to format the idisk.o file into the /boot directory. # mkboot idisk.o Reboot the system in a way to make a new /unix. # touch /etc/system # shutdown -g0 -i6 -y That should do it. To monitor use: # /etc/crash > od idscancnt 2 40253824: 00000002 00000001 ^--------^- This shows 2 resets on drive 0 and 1 reset on drive 1. If no patch is in, idscancnt will not exist. If the old patch is in, these numbers show the number of hangs that were noticed and reset within 20 seconds. If this (the new) patch is in, these numbers should always be 0 unless there is a hardware problem. write if you need more info.... rich -- "Hey Rocky, watch me pull a rabbit out of my hat." Bullwinkle Moose "oh...a wise guy....Come here....I'll murder ya..." Moe Howard "I'm hunting wabbits...." Elmer Fudd "If I could typ I'd be dangerous" Rich Andrews