On Sun, 2013-03-17 at 09:33 -0700, Eric Dumazet wrote: > On Sun, 2013-03-17 at 02:21 -0700, dormando wrote: > > > Hope you don't mind a screenshot: > > http://www.dormando.me/p/3.8.2-trace-crash.jpg > > > > (I put the patches on 3.8.2). box is on another continent so screenshot > > via IPMI is what I get. If this isn't enough or isn't right I'll try > > harder to get the trace logged, I guess? > > Thanks a lot, this gives another useful input, no need for more traces > for the moment. > > Could you send me the disassembly of tcp_release_cb() ? > > (objdump -d vmlinux | filter to get tcp_release_cb() body) > >
It seems we can have a refcnt imbalance because of atomic_set(&newsk->sk_refcnt, 2); we do in sk_clone_lock() Somehow, something is wrong, because at this point the socket should not be found. We perform a sock_hold() somewhere while the socket is already dead. I'll send a patch asap -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/