DragonFly kernel List (threaded) for 2004-11
Re: HEADS up on VM fix now in stable
What corruption would you see?
Matthew Dillon wrote:
David Rhodus and I finally tracked down the dogged filesystem corruption
and 'page not found in hash' panics that he has been getting for over
a year. The fix has been committed and the DragonFly_Stable tag slipped
for it. The bug was in contigmalloc() and particularly hit by people
using the 'twe' driver, not through any fault of the driver which is
why we were never able to find the problem before.
Anyone with weird random occassional filesystem corruption or 'page not
found in hash' panics should cvsup and get the latest stable, and the
problem should go away. The bug was inherited from FreeBSD-4 and even
FreeBSD-5 still had a portion of the buggy code in place (though it may
have been masked by other work done in FreeBSD-5).
Many thanks to David Rhodus for providing the many kernel cores (dozens
over the last year) related to this bug.
This bug is believed to have been the cause of *all* known occassdional
filesystem and/or VM corruption issues in DragonFly so it is with great
relief that we've finally been able to track it down and fix it.