Messages in this thread | | | From | Eric.Schenk@dna ... | Subject | Re: 2.0.23 finally goes down :( | Date | Tue, 08 Jul 1997 15:51:46 +0200 |
| |
Jon Lewis <jlewis@inorganic5.fdt.net> writes: >I had a system with long uptime (248 days) running 2.0.23 (ping protection >and a few other patches added) get weird on me the other day. > >root 1 99.9 0.1 868 84 ? S 22:00 1349:41 init >root 14 99.9 0.0 840 32 ? S 22:00 625:47 update (bdflush) > >init and update were going wild on CPU usage, but other than that, the >system seemed quite normal. Unfortunately, init (2.64) seemed to hang >during shutdown -rf now and I had to have someone hit the reset button. > >Anyone know what might have caused this?
Taking a stab in the dark, you just crossed the boundary where the jiffies clock rolls past 2^31. If init and update are using the times() system call to keep track of when they should be doing things (to be immune to changes in the wall clock), then things can go nuts at this point, since time_t is declared to be type "long". There could also be things internal to the kernel that might be breaking at this point, as there could also be at the 497 day mark. Perhaps it is time to again consider starting kernels with the jiffies count just about to roll over, and also just about to go past the 2^31 mark so that we can start hunting down these types of problems.
The times() man page should probably also mention that the ticks counts returned do wrap after a sufficiently long time, and should not be used as an absolute time measure.
-- Eric Schenk www: http://www.dna.lth.se/~erics Dept. of Comp. Sci., Lund University email: Eric.Schenk@dna.lth.se Box 118, S-221 00 LUND, Sweden fax: +46-46 13 10 21 ph: +46-46 222 96 38
| |