On Tue, 2008-09-16 at 21:47 +0100, John Allman wrote: > On Thu, Sep 11, 2008 at 1:23 AM, Andy Walls <[EMAIL PROTECTED]> wrote: > > > > You try to disable loading DRM via the Xorg.conf file and maybe > > blacklist and unload the drm modules to see if you can get things > > working. > > > > Hi Andy, > > Thanks for your very detailed response - I really appreciate the > effort. After reading your post, I carefully looked at the lsmod > output and it showed that drm was loaded and used by "via". I couldn't > unload it as it was used. I commented out the Load dri and Load glx > lines out of my xorg.conf and the section about dri permissions. I > also noticed that both drm and via were explicitly listed in > /etc/modules. I removed them there and rebooted. > > It would appear now that I have things working (though I'll have to > confirm). Does this feel like the "right" fix? Should I be doing > something else? Feels a bit weird to just disable drm...
Yup. That's why it's called a work-around. What you do as a work-around depends on your needs and the trade offs you are willing to make: 1. Using irqpoll gives you both drm/via and ivtv working at the expense of some processor cycles 2. Disabling DRM gives you two working encoders for capture, but a performance hit on playback 3. Leaving things as it was leaves you without one encoder and a possible performance hit in DRM when interrupt driven events are needed. 4. If you can fine tune the PCI interrupt routing so that the on-board graphics chip doesn't share an interrupt, you can get everything working, at the small expense of not sharing that one interrupt line. 5. You might be able to delay the load of the ivtv driver until late in the boot process (blacklist and then load in /etc/rc.d/rc.local) and if the agpart code only stimulates one unhandled interrupt from the on-board graphics controller at boot, you could get away with only needing to tweak the system startup scripts. Utlimately the problem is at system level. All the pieces (drm, via, ivtv, agpgart, kernel interrupt error recovery, the on-board graphics controller, and PCI interrupt routing) working in concert to give you a very annoying failure mode of the system. :) The "correct" fix is to have the modules that manipulate/stimulate the on-board VIA graphics controller, causing it generate an interrupt, really acknowledge the interrupt as handled. That is best done in the agpgart, via-agp, drm, or via(drm) code by their respective maintainers. You really need to take the problem to them for the correct fix IMO. I'd only screw it up. :) Regards, Andy _______________________________________________ ivtv-users mailing list [email protected] http://ivtvdriver.org/mailman/listinfo/ivtv-users
