Hi, I got the famous "unregister_netdevice: waiting for br0 to become free. Usage count = 1" deadlock with a linux-2.6.28 kernel when trying to destroy a bridge. I read in previous posts that it should have been fixed already. Some posts mentionned that IPv6 and/or ip_queue could cause this to happen. IPv6 is not compiled in my kernel, but I am using IP_QUEUE to mangle some packets Here is my environment: - Debian etch, with a non-patched original kernel 2.6.28 downloaded from www.kernel.org - netfilter/iptables, used to queue packets - a home-made program using ipqueue, which always sets a verdict to ip_queue, and reinjecting packets as fast as it can. The bad point is that I tried for a week to reproduce the bug without success; it just happens "sometimes" in heavy load. And here are my tests: - Tried to lose packets (no verdict set), until the queue is full (test failure: doesn't crash the kernel) - Tried to overload the system (network overload), sending SIGKILL from times to times to the home-made program (test "sometimes" crashed the kernel) - Any other nicer tests failed to crash the kernel too. - The home-made program may use up to 99% CPU time, as told by the `top` command; the hardware is a bi-processor, so I don't know the meaning of CPU time in this case - Since I try to find a way to reproduce the bug, I reboot after each test I also had kernel messages like: "nf_queue: full at 1024 entries, dropping packets(s). Dropped: 5" and "__ratelimit: 16 callbacks suppressed" at crash and non-crash times, which should not be alarming since it just tells that ip_queue can't handle the load I don't have any clues on how to track down this bug and the fact it seems random to me doesn't help. Thanks for the help -- To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
