Re: Q77 IGD instantly crashes on xen-pciback bind.

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Fri, Nov 28, 2014 at 08:00:40PM -0600, Dr. Greg Wettstein wrote:
> On Nov 27, 12:11pm, Sander Eikelenboom wrote:
> } Subject: Re: [Xen-devel] Q77 IGD instantly crashes on xen-pciback bind.
> 
> Hi, hope the week has gone well for everyone.
> 
> > > So we are obviously working with qemu-dm-traditional and with the
> > > IGD/LVDS BIOS configuration issue fixed the adapater passthrough is
> > > working and Windows7 is coming up and detecting the IGD as a standard
> > > VGA display adapter.  Additional invocations of the VM after the first
> > > one result in failed passthrough with a garbled display.
> 
> > This is probably due to the current lack of slot/bus reset in
> > xen-pciback, Konrad has a preliminary kernel patch for xen-pciback
> > that does this.  I have attached the patch, though it has some rough
> > edges in the design :-)
> >
> > I'm currently running with his 3.19 xen-pciback patches series + the
> > preliminary patch for slot/bus reset and rebooting a guest with
> > vga/pci passthrough now works. (i'm running with a radeon card,
> > passed through as a secondary card to the emulated qemu one, in a
> > linux guest using qemu-xen, so i can't help you with your other
> > questions and problems).
> 
> Thanks for taking the time for respond and forward along the patch.
> 
> I back ported the do_flr patch into the 3.14.x kernel and spent some
> time working with it.  I thought it might be useful to others to
> document what we ran into.
> 
> First of all the issue with the unsuccessful boot of Windows after the
> first invocation doesn't appear to have anything to do with resetting
> the card.  This was fixed by installing the most recent version of the
> Intel HD drivers in the Windows guest.
> 
> If IGD passthrough is done without the HD drivers Windows 7 appears to
> use its standard VGA driver which seems to be able to initialize and
> run the IGD device but does not appear to shutdown the device in a
> manner in which it can be re-started.  After the first invocation of
> the guest is shutdown the screen goes to a solid color.  Subsequent
> invocations result in the flashing multi-color screens which others
> have documented.
> 
> With the HD drivers installed IGD passthrough works fine through
> multiple invocations of a guest with the stock xen-pciback in 3.14.x.
> We ran 40-50 repetitive Windows guest invocations and every one was
> completely deterministic.
> 
> That being said we ran into an issue which we wanted to bounce off the
> list in the context of this thread.
> 
> One of the things we were not able to do was to successfuly re-start
> the IGD display for dom0.  After spending a lot of time going through
> Konrad's reset code it appears the IGD devices in the Q77 and Q87
> chipset implementations we are looking at are not advertising reset
> functions which can be used.
> 
> It appears that the IGD devices are advertising that they need a
> Device Specific Initialization (DSI) sequence in order to be enabled
> or powered up, if we interpret the output of lscpi properly, ie:
> 
> ---------------------------------------------------------------------------
> 00:02.0 VGA compatible controller: Intel Corporation Device 0152 (rev 09) (prog-if 00 [VGA controller])
>         Subsystem: Intel Corporation Device 2036
>         Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
>         Status: Cap+ 66MHz- UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort-
> <MAbort- >SERR- <PERR- INTx-
>         Latency: 0
>         Interrupt: pin A routed to IRQ 11
>         Region 0: Memory at f7800000 (64-bit, non-prefetchable) [size=4M]
>         Region 2: Memory at e0000000 (64-bit, prefetchable) [size=256M]
>         Region 4: I/O ports at f000 [size=64]
>         Expansion ROM at <unassigned> [disabled]
>         Capabilities: [90] MSI: Mask- 64bit- Count=1/1 Enable-
>                 Address: 00000000  Data: 0000
>         Capabilities: [d0] Power Management version 2
>                 Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
>                 Status: D0 PME-Enable- DSel=0 DScale=0 PME-
>         Capabilities: [a4] PCIe advanced features <?>
> ---------------------------------------------------------------------------
> 
> The ATI adapters which we have a lot of passthrough experience with
> are FLR capable and can be re-enabled after passthrough by writing a
> '1' to the enable pseudo-file in the /sys/bus/pci/devices/BDF
> directory.  The IGD adapters seem to fail to respond to a request to
> be re-enabled.
> 
> We would certainly be interested in any suggestions the collective may
> have with respect to how to get the adapter turned back on and/or
> implement a reset.

CC-ing Chien who has been looking at adding IGD passthrough support
in the qemu-upstream. Perhaps he can provide some ideas.

Thanks.
> 
> > Sander
> 
> Have a good weekend.
> 
> Greg
> 
> }-- End of excerpt from Sander Eikelenboom
> 
> As always,
> Dr. G.W. Wettstein, Ph.D.   Enjellic Systems Development, LLC.
> 4206 N. 19th Ave.           Specializing in information infra-structure
> Fargo, ND  58102            development.
> PH: 701-281-1686
> FAX: 701-281-3949           EMAIL: greg@enjellic.com
> ------------------------------------------------------------------------------
> "Laugh now but you won't be laughing when we find you laying on the
>  side of the road dead."
>                                 -- Betty Wettstein
>                                    At the Lake