On 03-Mar-2003 Andrew Gallatin wrote:
> I'm developing a character driver which tracks a lot of state on a
> per-open basis.  I've got several mutexes in there which are
> initialzed at open, and destroyed at close.  After a few
> dozen opens, witness seems to croak with:
>     witness_get: witness exhausted
> Am I leaking something?  Or is the witness code?  I looked at
> subr_witness.c, and I don't see witness_free() being called from
> witness_destroy().  There's probably some design constraint that
> I don't understand.

Unfortunately dead witnesses may still be stuck in the lock order
hierarchy and I haven't figured out yet how to properly handle the
case of free'ing a witness structure from the tree while preserving
the correct lock orders.  You can try
http://www.freebsd.org/~jhb/patches/witness.patch  but I don't think
it will help with your specific case.

> FWIW, Witness (and the FreeBSD debugging environment in general) is
> why I've gotten approval to co-develop this driver on FreeBSD (in
> addition to linux).  Its already caught several locking bugs.

Glad to hear it is at least working in some cases. :)


John Baldwin <[EMAIL PROTECTED]>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve!"  -  http://www.FreeBSD.org/

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

Reply via email to