Return-path: Received: from mail-vc0-f178.google.com ([209.85.220.178]:42013 "EHLO mail-vc0-f178.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753266AbaIZGPu convert rfc822-to-8bit (ORCPT ); Fri, 26 Sep 2014 02:15:50 -0400 Received: by mail-vc0-f178.google.com with SMTP id lf12so6506084vcb.37 for ; Thu, 25 Sep 2014 23:15:49 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <542448C0.9040505@Informatik.TU-Cottbus.DE> References: <542448C0.9040505@Informatik.TU-Cottbus.DE> Date: Fri, 26 Sep 2014 08:15:49 +0200 Message-ID: (sfid-20140926_081554_425788_67A72224) Subject: Re: Question about debugging mac80211 without printk From: Michal Kazior To: Pink Cc: linux-wireless Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 25 September 2014 18:54, Pink wrote: > Hi, > > actually if have the following problem, if have a lot of debug messages > using printk inside the mac layer but if there is a > critical bug in the source code of the mac80211 module > the system goes into kernel panic state during execution. Unfortunately so > far i now, it isn't possible > after a system restart to identify the rough position of the bug looking for > the debug msg's written before the bug has occured in > dmesg or syslog because the messages were not saved in follow of the crash. > > Now i thought user mode linux (uml) is maybe a potential solution to > identify the bug, if the kernel running the mac layer > is executed as virtual machine in uml. But i didn't found any kind of > description how to forward the physical wireless card except using TAP/TUN > device to the vm so > that the kernel inside the vm can run its mac layer directly on the card. > > Now my question what tools are you use for the development and testing of > the ath wireless driver > as well as the complete wireless stack including mac80211. Is it user mode > linux or something else? > If user mode linux is used, is there any howto available to forward the > atheros card to the vm. You can try using KVM and pass through devices to a virtual machine. Passing USB through has no extra requirements albeit from my experience some USB controllers and USB devices have some issues sometimes. PCI pass-through requires VT-d or AMD-v to be supported on both CPU and motherboard. I actually use this kind of setup (I'm using QEMU) myself. I can attach gdb, emulate serial console, pass-through rootfs via 9p and just boot a kernel in the VM omitting bootloader which is really convenient. If it crashes I can get pretty much all info I need. Michał