Hello Andrew,
sorry for the late answer and thank you for your remarks.
I like your idea of the static hash iterator and have got this implemented in
svn revision 1499. This should remove the problems about the memory leaks
with dangling hash iterators. What do you think?
As the patch obviously
Hi Simon,
I just gave your patch a try on my laptop and could successfully,
reproduceably crash my kernel in the following way:
Setting up wifi to ad-hoc mode and connecting it to other
batman-wifi-nodes, insmodding batman-adv on my laptop and adding
this wifi interface to batman - kernel hangs
On Thu, Dec 03, 2009 at 02:31:22AM +0100, Linus L??ssing wrote:
Hi Simon,
I just gave your patch a try on my laptop and could successfully,
reproduceably crash my kernel in the following way:
This looks like a deadlock.
Simon: Did you try lockdep on this new code?
On Sun, Nov 29, 2009 at 09:09:50PM +0100, Simon Wunderlich wrote:
I did some testing, including loading, unloading, killing individual
nodes etc, which seems to be clean so far. However there might be
more race conditions introduced by this large patch, and i'm therefore
requesting a careful
The orig_hash_lock locks big sections of the batman-adv code base, which
might lead to unneccesary performance degradation at some points.
Therefore this patch moves the locking from the whole originator hash
table to individual orig nodes, introducing a reference count based
system to identify