Re: [B.A.T.M.A.N.] [PATCHv2] [batman-adv] Removing the big batman lock
Simon Wunderlich wrote: @@ -131,9 +197,12 @@ struct orig_node *get_orig_node(uint8_t *addr) INIT_LIST_HEAD(orig_node-neigh_list); memcpy(orig_node-orig, addr, ETH_ALEN); + + orig_node-lock = __SPIN_LOCK_UNLOCKED(device_client-lock); orig_node-router = NULL; orig_node-batman_if = NULL; orig_node-hna_b The initialisation is wrong here. You must use spin_lock_init and not use the static initializer - because it isn't a static variable. Best regards, Sven signature.asc Description: This is a digitally signed message part.
Re: [B.A.T.M.A.N.] [PATCHv2] [batman-adv] Removing the big batman lock
It seems to deadlock according to https://lists.open-mesh.net/pipermail/b.a.t.m.a.n/2009-December/001938.html Patch was modified by Sven Eckelmann sven.eckelm...@gmx.de to apply cleanly against r1490. Is this just modification so that it cleanly applies? Or does it also fix the deadlock? Thanks Andrew
Re: [B.A.T.M.A.N.] [PATCHv2] [batman-adv] Removing the big batman lock
Andrew Lunn wrote: It seems to deadlock according to https://lists.open-mesh.net/pipermail/b.a.t.m.a.n/2009-December/001938.ht ml Patch was modified by Sven Eckelmann sven.eckelm...@gmx.de to apply cleanly against r1490. Is this just modification so that it cleanly applies? Or does it also fix the deadlock? No, deadlock isn't fixed yet. Just was nerved that it didn't apply when I wanted to test it. :) And to the lockdep stuff. It is a real nice feature (I used it partly when debugging another kernel module), but as I noticed the mathematical correctness check is disabled on uml and cannot be enabled due to some dependency problems. :( It is real a problem because uml is a preemption-less uniprocessor kernel and thus the spinlocks are just NOPs - so it is quite useless for any debugging of them. So back to the lets crash my working machine and/or qemu-based tests. Best regards, Sven signature.asc Description: This is a digitally signed message part.
