Recalling a recent comment by Linus relative to handling of free-inode
lists in the kernel, has something changed in 1.3.87? Last night, while
deleting a large number of files (make clean in multiple directories),
the system began complaining "virtual memory exhausted" - somewhat
unlikely, as I have 32-Meg of ram and a 20-meg swap partition (whose
disk never indicated any access). At this point, I could not run
anything, e.g. 'ps' and 'top' segfaulted immediately and I could not
even shutdown! Tellingly perhaps, one succesful invocation of 'df'
appeared to indicate almost no increase in free disk space at a point
where almost 20-megs had been deleted.
One red-button reset (with accompanying crossed fingers) later, the
system came back up unscathed amd with the expected amount of free disk
space. Nonetheless, this was a bit scary and very untypical of a
rock-solid machine. Is it possible that a race condition or misdirected
pointer reference exists within kernel code?
Any input appreciated!
- Steve
-- ___________________________________________________________ |Steven N. Hirsch "Anything worth doing is worth | |University of Vermont overdoing.." - Hunter S. Thompson | |Computer Science / EE | ------------------------------------------------------------