From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Rusty Lynch" Date: Wed, 20 Nov 2002 17:30:39 +0000 Subject: Re: Hotplug Test Cases or Test Procedures Message-Id: List-Id: References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable To: linux-hotplug@vger.kernel.org > Hi Rusty, > > What would you want the testcases or procedures to cover? Are there > tools, like 'cardctl eject' for cardbus/pcmcia, that interact with > hotplug events in testable ways? Particular drivers/hardware to test, > maybe with setup scripts, or just certain core functionality? We are just starting to scope this work out, so we don't know what it will entail. Initially we need to validate the operation of a couple of cPCI systems ranging from basic functionality of the zt5550 driver to system wide testing to verify once a given peripheral board is inserted, all of the hotplug actions (like loading a driver) happen. Eventually there are other bladed architectures with hotplug capibilities people are dreaming up that my group will need to help implement and test on Linux. I'm sure a lot of these test will be very specific a given setup (i.e. a given platform hotpluging a specific device), but at the very least it would be nice to provide a common way for people to upload other test procedures. This ping is really my attempt at getting my group to work effectivly in open source. It is just too easy to go off and implement something behind close doors only to find out you did it in a way that is not very useful to the rest of the world. > > I don't know of any particular tests, but it'd be good to have some! > > It's easy enough to handle the "do modules X, Y, and Z load" parts, > maybe even use a 'cardctl' analogue to simulate card insert/eject, > and that part is driver-independent. Stuff like setup scripts is > harder to characterize, as well as driver-specific (except for the > essentials like "did the script get run"). > > > It's not the same, but there is some info about debugging at the sf.net > site (http://linux-hotplug.sourceforge.net/?selected=DEbug) which talks > about applying one test technique when debugging ... it could easily get > used in regression testing: > > - save events > - play back later > - do they produce the intended result? > > I suspect a "save the events" routine should be part of the standard > hotplug toolset, it'd have more uses than just debug, test, or logging. > Cool. Thanks for the pointer. > > >>>Would this be something people would like to see in the hotplug CVS > >>>repository? > > If it's really addressing the ways that user mode tools are > alerted through /sbin/hotplug and /etc/hotplug/* support, > I'd hope so! Certainly everything that's shared with the > other hotpluggable pci busses should be there, and I agree > with you that > > > >>.. Or we can just put it up on the web site. > > > > > > Writing the procedures as html and making them accessable via the web site > > sounds like a good way to start. > > Yes. Some would need to stay that way (where people need to do > things like swap cards around), no matter how much can be automated. > > Are you volunteering to maintain those pages? > Yeap. I can do that. Actually I will be pulling in Guomin to work on this. Guomin is just starting to get familiar with hotplug and pci hotplug, so there will be some ramp up time. > > >>>Maybe I could approach the LTP? > >> > >>Traditionally they have only wanted tests that they can automate, and > >>unfortunately, pci hotplug tests (and most driver tests) can't be > >>automated very easily. But it can't hurt to try asking them :) > > > > > > I'll ping the LTP people to see if they are interested, but that shouldn't > > slow us down any. > > See above. Some kinds of testing _can_ be automated without > needing industrial robotics! BTW, I talked with Stephanie Glass of the LTP project, and she believes that there is a place we can put manual test procedures. There test will just not be part of the "run all" test cases. I'm thinking adding test to LTP could be a long term goal, but we should know more after something materializes. > > - Dave > > ------------------------------------------------------- This sf.net email is sponsored by: To learn the basics of securing=20 your web site with SSL, click here to get a FREE TRIAL of a Thawte=20 Server Certificate: http://www.gothawte.com/rd524.html _______________________________________________ Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net Linux-hotplug-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel