loc@tmsoft.UUCP (Leigh Clayton) (08/04/90)
I have a Magneto-Optical disk, which forces me to deal an a daily basis with the grisly details of 386/ix VTOC processing (the things 'labelit' writes). My guess as to the rules actually used by 386/ix is as follows: 1/ If I edit /etc/partitions to a correct value, and try to mount a file system that has an incorrect VTOC entry, mount does a careful check on the VTOC data, sees that they're wrong, and ignores the data in /etc/partitions. 2/ If the on-disk VTOC data is correct, mount writes whatever is in /etc/partitions over it (utterly no validation here). I won't go into the opinion I have of these rules (I expect you can guess), but I wonder if anyone can definitely verify that this is the way it works, and even better, suggest a way to fix up the VTOC when it's been clobbered (which seems destined to happen at least weekly for the rest of the time I run 386/ix). In case you care, the fundamental problem is that the MO media can be changed, they're like floppies with 290 meg on each (:-) except that you *never* get real disk errors (data good for 15-20 years :-) :-). The only way I have been able to get them 'supported' by 386/ix, however, is to basically put a little file system on each one in a fixed location, and keep /etc/partitions data in a file in this little file system. I then use a script to mount a MO disk that makes a new /etc/partitions from a fixed stub plus this data, before I do any mounts. Sounds good, but in practise I keep losing VTOC entries on the MO disks. If I could figure out a reasonable way to rebuild them, this would be OK (well, I could deal with it). By the way, I'm relatively new to Unix (only a year or so), so don't decline to suggest something because it's so simple I must have tried it - I very well may not have. ----------------------------------------------------------- loc@tmsoft.UUCP (Leigh Clayton) uunet!mnetor!tmsoft!loc