From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Mon, 26 Jan 2015 20:54:09 +0100 From: Gilles Chanteperdrix Message-ID: <20150126195409.GA1491@hermes.click-hack.org> References: <20150126190241.GA10160@bugfree> <20150126190911.GN12812@hermes.click-hack.org> <20150126195219.GA10959@bugfree> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150126195219.GA10959@bugfree> Subject: Re: [Xenomai] Porting Xenomai on Odroid-U3 List-Id: Discussions about the Xenomai project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Arnaud Degroote Cc: xenomai@xenomai.org On Mon, Jan 26, 2015 at 08:52:19PM +0100, Arnaud Degroote wrote: > On 26/Jan - 20:09, Gilles Chanteperdrix wrote: > > On Mon, Jan 26, 2015 at 08:02:41PM +0100, Arnaud Degroote wrote: > > > Hi list, > > > > > > I'm trying to port Xenomai on Odroid-U3. I read on the list that there > > > is a first try by GP Orcullo, so I start from its work. > > > > > > I'm working on a 3.17.8 kernel on which I ported the Ipipe / Xenomai > > > patch. I validated it on two omaps platform where I don't see regression > > > against previous version of the patch, so I'm fairly confident that the > > > core part is correct. In a second time, I ported the previous effort of > > > GP Orcullo. The current state of the effort is available on > > > > > > https://github.com/adegroote/linux/tree/linux-stable-xenomai-odroid-3.17.y > > > > > > I tested without IPIPE, and with IPIPE. Troubles start with the XENOMAI > > > option. Basically, the kernel seems to hangs when I load any "skin" (I > > > compiled them in modules to be sure). For example, > > > > > > modprobe xeno-native > > > > > > prints > > > > > > Xenomai: starting native API services. > > > > > > but I cannot get back the shell prompt. > > > > > > From the Troubleshooting part, I suppose something is wrong with the > > > ipipe_timer, but I don't have much idea where to look for. > > > > > > Can someone give me some inputs about how to trace / debug this issue > > > or/and quickly check if I miss something obvious in my patch. > > > > > > https://github.com/adegroote/linux/commit/81c6ebeb3127efdceb3e31f9d3f1f400ecbb579f.patch > > > > Well, you should put some printks in the timer code to see what is > > the last action done. Is the ->ack function called, does it really > > acknowledge the timer interrupt at timer level (because if it does > > not, the interrupt will trigger again at the end of the current > > interrupt). > > Thanks Gilles, > I will try tomorrow morning, when I will be back to work. > > > > Could you post a link to the patched file handling the timer ? > > The timer used is exynos-mct, available here > > https://github.com/adegroote/linux/blob/linux-stable-xenomai-odroid-3.17.y/drivers/clocksource/exynos_mct.c The ISR is wrong. The acknowledge should be skipped when the timer is stolen, not the event handler. -- Gilles.