IO_PAGE_FAULT while booting.

IO_PAGE_FAULT while booting.

[Prasad Joshi]

Hello,

I saw this error logged while booting the kernel
prasad@prasad-kvm:~$ dmesg | grep PAGE
[   13.132547] AMD-Vi: Event logged [IO_PAGE_FAULT device=06:00.1
domain=0x0000 address=0x00000000bb402000 flags=0x0050]

The kernel I have is cloned copy of the KVM repository.

menuentry 'Ubuntu, with Linux 2.6.37-rc6+' --class ubuntu --class
gnu-linux --class gnu --class os {
    recordfail
    insmod part_msdos
    insmod ext2
    set root='(hd0,msdos5)'
    search --no-floppy --fs-uuid --set ab9dfe8f-8e58-43ca-b16f-a42e0f02da72
    linux   /vmlinuz-2.6.37-rc6+ root=/dev/sda1 ro iommu=pt iommu=1
amd_iommu_dump
    initrd  /initrd.img-2.6.37-rc6+
}

prasad@prasad-kvm:~$ uname -a
Linux prasad-kvm 2.6.37-rc6+ #19 SMP Mon Jan 10 19:03:36 GMT 2011
x86_64 GNU/Linux

Output of lspci -v

06:00.0 SATA controller: JMicron Technology Corp. JMB361 AHCI/IDE (rev
02) (prog-if 01 [AHCI 1.0])
	Subsystem: ASUSTeK Computer Inc. Device 843f
	Flags: bus master, fast devsel, latency 0, IRQ 44
	Memory at f9ffe000 (32-bit, non-prefetchable) [size=8K]
	Capabilities: [68] Power Management version 2
	Capabilities: [50] Express Legacy Endpoint, MSI 01
	Kernel driver in use: ahci

06:00.1 IDE interface: JMicron Technology Corp. JMB361 AHCI/IDE (rev
02) (prog-if 85 [Master SecO PriO])
	Subsystem: ASUSTeK Computer Inc. Device 843f
	Flags: bus master, fast devsel, latency 0, IRQ 45
	I/O ports at dc00 [size=8]
	I/O ports at d880 [size=4]
	I/O ports at d800 [size=8]
	I/O ports at d480 [size=4]
	I/O ports at d400 [size=16]
	Capabilities: [68] Power Management version 2
	Kernel driver in use: JMicron IDE

Please let me know if you need more information from my side.

Thanks and Regards,
Prasad

Re: IO_PAGE_FAULT while booting.

[Joerg Roedel]

Hi Prasad,

On Mon, Jan 17, 2011 at 12:30:48PM +0000, Prasad Joshi wrote:
> Hello,
> 
> I saw this error logged while booting the kernel
> prasad@prasad-kvm:~$ dmesg | grep PAGE
> [   13.132547] AMD-Vi: Event logged [IO_PAGE_FAULT device=06:00.1
> domain=0x0000 address=0x00000000bb402000 flags=0x0050]

Can you please boot with amd_iommu_dump added on the kernel command line
and send full dmesg, your kernels .config and dmidecode please? That
would help analyzing the problem.

	Joerg

Re: IO_PAGE_FAULT while booting.

[Joerg Roedel]

On Mon, Jan 17, 2011 at 05:08:42PM +0000, Prasad Joshi wrote:
> Please find the information you requested attached with this mail.
> Also attaching the output of lspci -vvv -xxx incase it is useful.

Thanks, I'll have a look at it.

> I would be interested in understanding the problem. If possible can
> you please let me know what does this error indicate?

The error indicates that the device in question tried a dma transfer to
a location which it is not allowed to access. To most likely reason is
that the IOMMU was setup before the BIOS was giving away the device to
the OS. But other reasons are also possible. I'll let you know when I
know more.

	Joerg

Re: IO_PAGE_FAULT while booting.

[Joerg Roedel]

On Mon, Jan 17, 2011 at 05:08:42PM +0000, Prasad Joshi wrote:
> >> [   13.132547] AMD-Vi: Event logged [IO_PAGE_FAULT device=06:00.1 domain=0x0000 address=0x00000000bb402000 flags=0x0050]

Ok, the problem is, that the device 06:00.1 is not described in the
ACPI table for the IOMMU driver. So the driver does not feel responsible
for it and does not create mappings. It just blocks all DMA from that
device which results in the IO_PAGE_FAULT you have seen.
You will not be able to use the IDE controler in DMA mode, so if you
attach any drives to it you are sticked to slow PIO access mode. The
problem was already seen by another person and he reported the issue to
the board manufacturer to get a new BIOS where this issue is fixed.

Regards,
	Joerg

Re: IO_PAGE_FAULT while booting.

[Joerg Roedel]

On Fri, Jan 28, 2011 at 03:31:17PM +0000, Prasad Joshi wrote:
>On Fri, Jan 28, 2011 at 3:15 PM, Joerg Roedel <joro@8bytes.org> wrote:
>
>  On Thu, Jan 27, 2011 at 06:43:19PM +0000, Prasad Joshi wrote:
>  >    AMD-Vi: Event logged [IO_PAGE_FAULT device=02:00.0 domain=0x0002
>  >    address=0x00000000bb447000 flags=0x0000]
>  >    AMD-Vi: Event logged [IO_PAGE_FAULT device=02:00.0 domain=0x0002
>  >    address=0x00000000bb447040 flags=0x0000]
>  >    AMD-Vi: Event logged [IO_PAGE_FAULT device=02:00.0 domain=0x0002
>  >    address=0x00000000bb447080 flags=0x0000]
>
>  Hard to tell whats going wrong here. How many memory have you given to
>  the guest?
>
>I am allocating 1 GB memory to the VM

The address of the page-faults is around 3GB, so the guest sends DMA
commands to the GFX card which contains addresses outside its memory. I
Cc the list again, maybe they can bring some light into this issue.
In general GFX passthrough is not expected to work. But if it works for
you with -net none and doesn't without there must be another issue.

	Joerg

> 
>
>  What type of GFX card do you try to pass through?
>
># lspci -vvv
>02:00.0 VGA compatible controller: ATI Technologies Inc RV370 5B64 [FireGL
>V3100 (PCIE)] (rev 80) (prog-if 00 [VGA controller])
>    Subsystem: ATI Technologies Inc Device 0102
>    Interrupt: pin A routed to IRQ 32
>    Region 0: Memory at c0000000 (32-bit, prefetchable) [size=128M]
>    Region 1: I/O ports at a000 [size=256]
>    Region 2: Memory at f9cf0000 (32-bit, non-prefetchable) [size=64K]
>    Expansion ROM at f9cc0000 [disabled] [size=128K]
>
># dmidecode
>Handle 0x002E, DMI type 9, 17 bytes
>System Slot Information
>        Designation: PCIEX16_2
>        Type: 32-bit PCI Express
>        Current Usage: In Use
>        Length: Short
>        ID: 27
>        Characteristics:
>                3.3 V is provided
>                Opening is shared
>                PME signal is supported
>        Bus Address: ffff:02:00.0
>
> 
>
>         Joerg