[vfio-users] vfio_pci_read_config failed: Bad address

[Alex Williamson]

On Tue, 2015-11-17 at 22:35 +0100, Ronnie Swanink wrote:
> Hello,
> 
> For the last week I've been trying to get GPU passthrough to work. My 
> goal is to have a simple MSI Geforce 210 as the host graphics adapter, 
> and pass a MSI GTX 970 GAMING to a Windows 8 guest. However, I can't get 
> it to work. The 970 does give output when the drivers are not installed, 
> but after installing the drivers qemu gives me the following error when 
> Windows boots (the vm restarts after that and Windows goes into 
> automatic recovery):
> 
> 2015-11-17T21:08:59.910448Z qemu-system-x86_64: 
> vfio_pci_read_config(0000:02:00.0, 0x60c, 0x4) failed: Bad address
> 2015-11-17T21:08:59.910566Z qemu-system-x86_64: 
> vfio_pci_read_config(0000:02:00.0, 0x610, 0x4) failed: Bad address
> 2015-11-17T21:09:00.404715Z qemu-system-x86_64: 
> vfio_pci_read_config(0000:02:00.0, 0x150, 0x4) failed: Bad address


Note that these (and the bazillion others included) are all to offsets
>= 0x100, PCI extended config space.

> 
> My hardware (as you can see, the 970 is claimed by vfio-pci):
> 
> [root at localhost Documents]# lspci -v

And none of the listing here show any extended config space for devices.
For example, your GPU should look more like this:

01:00.0 VGA compatible controller: NVIDIA Corporation GM107 [GeForce GTX 750] (rev a2) (prog-if 00 [VGA controller])
	Subsystem: eVga.com. Corp. Device 2753
	Flags: fast devsel, IRQ 16
	Memory at f6000000 (32-bit, non-prefetchable) [size=16M]
	Memory at e0000000 (64-bit, prefetchable) [size=256M]
	Memory at f0000000 (64-bit, prefetchable) [size=32M]
	I/O ports at e000 [size=128]
	Expansion ROM at f7000000 [disabled] [size=512K]
	Capabilities: [60] Power Management version 3
	Capabilities: [68] MSI: Enable- Count=1/1 Maskable- 64bit+
	Capabilities: [78] Express Legacy Endpoint, MSI 00
	Capabilities: [100] Virtual Channel
	Capabilities: [250] Latency Tolerance Reporting
	Capabilities: [258] L1 PM Substates
	Capabilities: [128] Power Budgeting <?>
	Capabilities: [600] Vendor Specific Information: ID=0001 Rev=1 Len=024 <?>
	Capabilities: [900] #19
	Kernel driver in use: vfio-pci
	Kernel modules: nouveau

So, there's likely a QEMU bug that we're passing through a config
request beyond the end of what the kernel reported the size of config
space to be, and that results in a -EFAULT from the kernel, but there
must be a reason that extended config space is not present.  Assuming
this happens on Fedora kernels, what kernel options are you providing?
A pastebin of dmesg might provide some clues.  Things like pci=nommconf
would certainly explain this.  Thanks,

Alex