On Mon, 22 Apr 2002, Bruce Evans wrote:

> Yes; I forgot to write the clause about it only being equivalent to a
> panic if certain options (mainly INVARIANTS) are configured.
> > do see that falloc does lock the file descriptor table too, though ---
> > I wonder how it ever worked.
> INVARIANTS causes a check of an assertion that non-recursive locks
> like fd_mtx aren't actually misused recursively.  The check causes more
> panics than the misuse since the misuse is rarely fatal.  I may be
> missing something since I rarely use INVARIANTS and haven't checked
> that it causes the panic deterministically.

 I think you're talking about the WITNESS option rather than
INVARIANTS, that's the one who's at least documented in GENERIC as
being responsible for mutex and deadlock checks. And this is somewhat
confirmed by the face that an alternative kernel that I built doesn't
panic and the difference between them two is that WITNESS is disabled in
the non-panicking one.

 Happy debuggin'
 Ady (@freebsd.ady.ro)
| Programming in BASIC causes brain damage.                           |
|   (Edsger Wybe Dijkstra)                                            |

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message

Reply via email to