pinard@IRO.UMontreal.CA (Francois Pinard) (08/25/89)
Hello. I would like to share some facts and eventually receive some feedback about a few freezes and crashes I experienced with 386/ix Unix 2.0.1. Other people may have had similar adventures. Even then, abstract these glitches, I would be quite satisfied with this system, which is on several other aspects well behaved beyond my expectations. 1) The system freezes at various places while surface testing the disk. In fact, character echo is being done, so the system is not really dead. The very bad side effect of these is that alternate sector tables never get written, which further induce the system in unstable operations and prohibit the development of confidence. 2) The probability of a system crash increase drastically if I start several processes requiring a lot of memory each, to the extent it is worrying to launch big tasks without killing others first. Those are the two main problems. The following are, by comparison, of subsidiary importance: 3) Getting out of cu may cause, sometimes, loss of a virtual terminal and of UUCP operations. When cu starts, it establishes two tasks, ending cu will have one of these tasks to terminate, the other will hang unkillable, with a small, but non negligible probability. An fsck will be mandatory at the next reboot. This might be related to ~%nostop, but I've no evidence. 4) If there is any bad block in the free list of a file system, the system is not mountable. If there is a bad block in the free list of the root partition, you've lost a system. It would be nice if fsck was a least able, after having read a bad block, not to keep it part of its reconstructed free list. 5) A ttymap reset will hang the virtual terminal if done from a non-root user. I'm not using nor even testing the reset option anymore. -- Franc,ois Pinard pinard@iro.umontreal.ca (514) 588-4656 ``Vivement GNU!'' ...!uunet!iros1!pinard