Return-path: Received: from e36.co.us.ibm.com ([32.97.110.154]:51618 "EHLO e36.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751101AbYH2RXC (ORCPT ); Fri, 29 Aug 2008 13:23:02 -0400 Received: from d03relay02.boulder.ibm.com (d03relay02.boulder.ibm.com [9.17.195.227]) by e36.co.us.ibm.com (8.13.8/8.13.8) with ESMTP id m7THN1P0031239 for ; Fri, 29 Aug 2008 13:23:01 -0400 Received: from d03av01.boulder.ibm.com (d03av01.boulder.ibm.com [9.17.195.167]) by d03relay02.boulder.ibm.com (8.13.8/8.13.8/NCO v9.0) with ESMTP id m7THN1JV193978 for ; Fri, 29 Aug 2008 11:23:01 -0600 Received: from d03av01.boulder.ibm.com (loopback [127.0.0.1]) by d03av01.boulder.ibm.com (8.12.11.20060308/8.13.3) with ESMTP id m7THN0tx004260 for ; Fri, 29 Aug 2008 11:23:01 -0600 Subject: Re: [TESTING] Wireless mesh networking (802.11s) draft support From: Subrata Modak Reply-To: subrata@linux.vnet.ibm.com To: Jouni Malinen Cc: "Luis R. Rodriguez" , Luis Carlos Cobo , Johannes Berg , "John W. Linville" , ltp-list , "linux-wireless@vger.kernel.org" , Jouni.Malinen@atheros.com In-Reply-To: <20080828165618.GK4863@jm.kir.nu> References: <1219698032.5434.6.camel@subratamodak.linux.ibm.com> <20080825214625.GL6007@tesla> <20080828165618.GK4863@jm.kir.nu> Content-Type: text/plain Date: Fri, 29 Aug 2008 22:52:55 +0530 Message-Id: <1220030575.880.54.camel@subratamodak.linux.ibm.com> (sfid-20080829_192308_174885_C0420747) Mime-Version: 1.0 Sender: linux-wireless-owner@vger.kernel.org List-ID: Thanks Jouni and Luis, On Thu, 2008-08-28 at 19:56 +0300, Jouni Malinen wrote: > On Mon, Aug 25, 2008 at 02:46:25PM -0700, Luis R. Rodriguez wrote: > > > Mesh code is GPLv2 right now. What I'd recommend is to get mesh support > > added into mac80211_hwsim driver and then you can test mesh code without > > even having any wireless hardware. > > That would be quite useful addition for automated tests that anyone can > run. Yes, this is what we also aim to achieve - fully automated testing, even if it requires some initial human intervention. Normally this happens even in the normal LTP network tests: http://ltp.cvs.sourceforge.net/ltp/ltp/testcases/network, Somebody needs to do the following: 1) Setup some network daemons to run on the server. This is done manually, 2) The LTP source code should be installed at the same directory location in both the client and the server, 3) From the Client side, the user-id/password/server-ip is exported, Once this is done, then the network tests begin and then everything in automated. You just need to analyze the logs after the test is over. This typical scenario makes this test cases to run separately from other default LTP tests, yet they reside inside LTP. > > > At the third Linux wireless summit in Ottawa we briefly spoke about > > automating testing for wireless. We should try to automate this first > > with mac80211_hwsim and some scripts to do basic tests. Mesh testing can > > be added once mac8021_hwsim gets support for Mesh. > > I've started collecting test cases for mac80211_hwsim. It would be > interesting to see whether these could be integrated into LTP somehow > and allow people to run a set of tests in fully scripted way. These Yes, this is the way we should target things. Say we have your wireless test cases inside a directory: http://ltp.cvs.sourceforge.net/ltp/ltp/testcases/network/wireless/mesh_networking/ This should have some bare minimum entries, like: 1) Makefile -> which should be able to build any binary required and also install the binaries (actual test cases and wrapper scripts, etc) here: http://ltp.cvs.sourceforge.net/ltp/ltp/testcases/bin/ 2) A master wrapper script which does: 2.1) Checks out the existence of the driver(s) in the system, 2.1.1) If does not exist, then the script should also be able to download it from the required location, 2.2) If driver exists, then load it, 2.2.1) handle driver loading/unloading errors and report accordingly, 2.3) Then actually run those tests(other binaries/scripts) and record results, 2.4) Unload the driver once the tests are over ; report unloading errors if any, Starting initially, we will prevent your Makefile to get called from the root LTP Makefile (when make is invoked from LTP root directory). This allows only selected testers (who will be actually interested in testing this component) to drill down to this sub-directory and issue make. And then invoke the master wrapper script to run tests and see results. You can look in to the following to see how SElinux tests are intergrated inside LTP. This will help you to understand the scenario: http://ltp.cvs.sourceforge.net/ltp/ltp/testcases/kernel/security/, Once this process gets stabilized, we can integrate the make with the LTP make so that your test cases gets built and installed during the normal LTP build/install. Now in one of your READMEs, you should mention the following things very properly: 1) Ways to get/download/install the source(test code), 2) Ways to build/install and run, 3) Any other special requirement/setup the tests need to execute, like Glibc version ??, Kernel version ?? (which options in kernel should have been enabled to get this feature running), Architecture Supported, etc. I hope we can start in bits and pieces. You can create a small patch and send it across. I am not in a hurry. Please take your time and do things in bits and pieces. I wanted to stay in touch and make sure that we have started to do somethings, however small they may be. Regards-- Subrata > tests are described in hostap.git tree mac80211_hwsim/tests > subdirectories: > http://w1.fi/gitweb/gitweb.cgi?p=hostap.git;a=tree;f=mac80211_hwsim/tests > > The current instructions for running the commands (*/test.txt) are more > for a person than a script, but the changes to get test running would be > trivial (i.e., just run hostapd and wpa_supplicant on background). > Larger task is in adding some functionality to verify that the test > passed and then finish with cleaning up (killing daemons from background > and unload mac80211_hwsim). >