In article <[EMAIL PROTECTED]>, Lars Wirzenius <[EMAIL PROTECTED]> wrote: >However, I should point out at once that I don't know if my >lockfile.c will work properly in all situations. I've done the >best I can, based on manual pages, but I haven't been able to >test it over NFS when various errors happen. I'm not an expert >on this. I think I have some notes from Miquel somewhere, >and will try to find them. The problem is that one needs to >understand NFS and other failure modes to design the locking >code properly, and I don't understand NFS. > >And it isn't enough to just create a lock file. If the creator >of the lock file crashes without removing the lock, everyone >still alive will starve. There should be a method for automatically >recovering from this, and that method must not rely on a daemon >purging lock files at suitable intervals. There's no good way >of doing this, as far as I can tell.
SystemV has a function "touchlock" or somesuch. You just have to touch the lockfile regulary. If the delta time between creation time and modification time gets over 5 minutes, the lock is considered stale. Read "man maillock" on Solaris, together with /usr/include/maillock.h. Interesting stuff. Mike. -- | Miquel van Smoorenburg | "I need more space" "Well, why not move to Texas" | | [EMAIL PROTECTED] | "No, on my account, stupid." "Stupid? Uh-oh.." | | PGP fingerprint: FE 66 52 4F CD 59 A5 36 7F 39 8B 20 F1 D6 74 02 | -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] .