Daily panic in getnewvnode

From: Greg 'groggy' Lehey <grog_at_FreeBSD.org>
Date: Wed, 29 Sep 2004 17:44:09 +0930
For the last month I've been getting a daily panic out of getvnode.  A
brief summary:

- Panic message is "Cleaned vnode isn't".  This means that the vnode
  pulled off the free list and cleaned with vtryrecycle still has the
  vp->v_data field set.
- It happens during the nightly cron jobs.
- The program in question is find, going through a large disk (200 GB,
  several million files).
- The disk does not seem to be getting fsck'd.
- The test in question is part of the INVARIANTS checking code, so if
  I turned INVARIANTS off, the panic would no longer occur.

I suspect that if I ran fsck on the disk, it would also solve the
problem.  The reason I haven't done so is because I chose this
particular panic to document for my next kernel debugging tutorial.
Unfortunately, I'm running out of time, and I'd really like to find
the *real* cause of this bug.  If anybody's seen anything like this,
please let me know.  I have a draft of my tutorial notes at
http://www.lemis.com/grog/handout.pdf, and the description starts on
page 86.  If anybody can help me clarify the problem, I'd be very
grateful.

Greg
--
See complete headers for address and phone numbers.

Received on Wed Sep 29 2004 - 06:14:13 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:14 UTC