Am Saturday 18 December 2010 schrieben Sie:
> On Sat, 2010-12-18 at 14:40 +0100, Boris Cuber wrote:
> > Am Friday 17 December 2010 schrieben Sie:
> > > On Fri, 2010-12-17 at 12:19 +0100, Boris Cuber wrote:
> > > > Hello linux-media people!
> > > >
> > > > I have to problems with my dvb card ("TeVii S470"). I already
> > > > filed 2 bug reports some time ago, but no one seems to have
> > > > noticed/read them, so i'm trying it here now.
> > > > If you need a "full" dmesg, then please take a look at
> > > > https://bugzilla.kernel.org/attachment.cgi?id=40552
> > > >
> > > > 1) "TeVii S470 dvbs-2 card (cx23885) is not usable after
> > > > pm-suspend/resume" https://bugzilla.kernel.org/show_bug.cgi?id=16467
> > >
> > > The cx23885 driver does not implement power management. It would
> > > likely take many, many hours of coding and testing to implement it
> > > properly.
> > >
> > > If you need resume/suspend, use the power management scripts on your
> > > machine to kill all the applications using the TeVii S470, and then
> > > unload the cx23885 module just before suspend.
> > >
> > > On resume, have the power management scripts reload the cx23885 module.
> >
> > Well, this doesn't work. If i did tune a channel before or used the dvb
> > card somehow for watching tv, unloading and reloading the cx23885
> > module also makes the card unuseable.
> > In dmesg there's lots of "do_IRQ: 1.161 No irq handler for vector (irq
> > -1)" messages then. This can only be fixed by rebooting the computer.
>
> That is s a known issue with the CX2388[578] chip and PCIe MSI.
>
> The CX2388[578] will not accept a different value for its "MSI Data"
> field in its PCI config space, when MSI has been enabled on the hardware
> once.
>
> The kernel will always try to give a different value for the "MSI Data"
> field to the CX2388[578] chip, on cx23885 module unload and reload.
>
> So suspend and then resume didn't reset the chip hardware?
>
> You can set "pci=nomsi" on your kernel command line to prevent the
> cx23885 driver, and your whole system unfortunately, from using MSI.
>
Ah, now i got it. Simply reloading the module kills the card with that nasty
do_IRQ thing (until reboot), but
-> 1) unloading module 2) suspending 3) resuming 4) loading module
actually works. It's kinda dirty solution (with some hackish script in
/etc/pm/sleep.d/), but it works somehow.
Perhaps someday there will be a better solution (power management?).
> Regards,
> Andy
Thanks for your time,
Boris
--
http://boris64.net 20xx ;)
Attachment:
signature.asc
Description: This is a digitally signed message part.
