Hi S, On Fri, Feb 22, 2013 at 11:53 PM, శ్రీధర్ <[email protected]> wrote: > I have the following setup three nodes configured to form a mesh (secure), i > enabled kememleak in the kernel, and after starting all the nodes and > verifying that all of them established the mesh correctly. I left the > system, no external data traffic introduced. After few hour i noticed a > bunch of messages on the console "kmemleak: 6 new suspected memory leaks > (see /sys/kernel/debug/kmemleak)", the number varied, attached is the out > put of kmemleak file. > > Looks like the memory allocated in mesh_queue_preq() > preq_node = kmalloc(sizeof(struct mesh_preq_queue), GFP_ATOMIC); > if (!preq_node) { > mhwmp_dbg("could not allocate PREQ node"); > return; > }
Thanks for the analysis, guess we should run kmemleak once in a while as well :). There does indeed seem to be a race in the code where two preqs may get queued up, but only one of them will later get dequeued. A patch for fixing this behavior would be welcome. On another note, wireless-testing HEAD doesn't seem to be triggering a path refresh at all! Exactly which compat-wireless are you running? -- Thomas _______________________________________________ Devel mailing list [email protected] http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel
