On Mon, Nov 20, 2000 at 12:11:01PM -0800, Konrad Schroder wrote:
> The patches to lfs_cleanerd that are not on the branch, though, have only
> to do with memory usage/leakage, and shouldn't exhibit this particular
> problem, I think.  It is possible (even likely) that lfs_avail accounting
> is messed up, or that the recorded number of clean segments is too high,
> causing the cleaner not to run and running the filesystem out of actual
> clean segments.

Yes, with the last set of patches I appended to PR 11468 which
fixes the last of the known lfs_avail accounting leaks, I can no
longer reproduce this panic.  At least, not yet anyway.  I haven't
tried all the different ways to fill the filesystem yet, but its
at least _harder_ to provoke.

Konrad, I just sent you these patches last night.  As of right now,
-current still has some serious problems with lfs_avail accounting.

//Jesse Off

//Jesse Off