Guillaume Thouvenin
2013-07-09 09:18:10 UTC
Hello,
I'd like to test a nvidia K1 with pci assignement by using the pci-stub
module or the vfio module. I'm running an ubuntu raring as the host and
a version 1.4.0 of qemu-system-x86 1.4.0. The nvidia K1 has a grid
architecture and I see four K1 on one card.
I have unbind the device from the old driver and bind it to the
vfio-pci. Then started the VM with the following command:
$ sudo sh -c "qemu-system-x86_64 -M q35 -m 4096 --enable-kvm \
-net nic,model=virtio,macaddr=52:54:00:82:69:75 -net tap,ifname=tap0 \
-drive file=/home/thouveng/ubu13.04_amd64_base.qcow2,if=virtio \
-device vfio-pci,host=87:00.0"
Then I logged into the VM and I can see the nvidia device:
$ lspci | grep -i nvi
00:03.0 VGA compatible controller: NVIDIA Corporation GK107GL [GRID
K1] (rev a1)
I installed the nvidia driver that I downloaded from their website.
Every things seems ok. Now to validate that everyhting is working I
tried to run the command "nvidia-smi -q" and I got the following error:
$ nvidia-smi -q
NVIDIA: could not open the device file /dev/nvidia0 (Input/output error).
Unable to determine the device handle for GPU 0000:00:03.0: Unknown Error
and in the syslog of the guest I can see:
[ 1502.459123] NVRM: RmInitAdapter failed! (0x26:0x38:1170)
[ 1502.459150] NVRM: rm_init_adapter(0) failed
I checked the BAR registers in the guest but as far as I understand
them, memory regions seem correct:
$ lspci -v
00:03.0 VGA compatible controller: NVIDIA Corporation GK107GL [GRID
K1] (rev a1) (prog-if 00 [VGA controller])
Subsystem: NVIDIA Corporation Device 099d
Flags: bus master, fast devsel, latency 0, IRQ 23
Memory at fd000000 (32-bit, non-prefetchable) [size=16M]
Memory at f0000000 (64-bit, prefetchable) [size=128M]
Memory at fa000000 (64-bit, prefetchable) [size=32M]
I/O ports at c000 [size=128]
Expansion ROM at fe000000 [disabled] [size=512K]
Capabilities: <access denied>
Kernel driver in use: nvidia
I also tried to do the assignement with pci-stub but I have the same
problem.
The onlu clue that I could find on the web was on the page
http://us.download.nvidia.com/XFree86/Linux-x86/304.60/README/commonproblems.html where they said that the same problem occured when the "VBIOS fail to load on my Optimus system". They said about this error "Such problems are typically beyond the control of the NVIDIA driver, which relies on proper cooperation of ACPI and the System BIOS to retrieve important information about the GPU, including the Video
BIOS."
I don't really understand the problem. I will really appreciate any help to
troubleshoot the problem and any links that can help are more than welcome :).
I will also appreciate any comment about a succesfull story with PCI
assignement with a K1. I don't have all of them but I maybe will have
the opportunity to test them so any comments are welcome :).
Best regards,
Guillaume
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to ***@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
I'd like to test a nvidia K1 with pci assignement by using the pci-stub
module or the vfio module. I'm running an ubuntu raring as the host and
a version 1.4.0 of qemu-system-x86 1.4.0. The nvidia K1 has a grid
architecture and I see four K1 on one card.
I have unbind the device from the old driver and bind it to the
vfio-pci. Then started the VM with the following command:
$ sudo sh -c "qemu-system-x86_64 -M q35 -m 4096 --enable-kvm \
-net nic,model=virtio,macaddr=52:54:00:82:69:75 -net tap,ifname=tap0 \
-drive file=/home/thouveng/ubu13.04_amd64_base.qcow2,if=virtio \
-device vfio-pci,host=87:00.0"
Then I logged into the VM and I can see the nvidia device:
$ lspci | grep -i nvi
00:03.0 VGA compatible controller: NVIDIA Corporation GK107GL [GRID
K1] (rev a1)
I installed the nvidia driver that I downloaded from their website.
Every things seems ok. Now to validate that everyhting is working I
tried to run the command "nvidia-smi -q" and I got the following error:
$ nvidia-smi -q
NVIDIA: could not open the device file /dev/nvidia0 (Input/output error).
Unable to determine the device handle for GPU 0000:00:03.0: Unknown Error
and in the syslog of the guest I can see:
[ 1502.459123] NVRM: RmInitAdapter failed! (0x26:0x38:1170)
[ 1502.459150] NVRM: rm_init_adapter(0) failed
I checked the BAR registers in the guest but as far as I understand
them, memory regions seem correct:
$ lspci -v
00:03.0 VGA compatible controller: NVIDIA Corporation GK107GL [GRID
K1] (rev a1) (prog-if 00 [VGA controller])
Subsystem: NVIDIA Corporation Device 099d
Flags: bus master, fast devsel, latency 0, IRQ 23
Memory at fd000000 (32-bit, non-prefetchable) [size=16M]
Memory at f0000000 (64-bit, prefetchable) [size=128M]
Memory at fa000000 (64-bit, prefetchable) [size=32M]
I/O ports at c000 [size=128]
Expansion ROM at fe000000 [disabled] [size=512K]
Capabilities: <access denied>
Kernel driver in use: nvidia
I also tried to do the assignement with pci-stub but I have the same
problem.
The onlu clue that I could find on the web was on the page
http://us.download.nvidia.com/XFree86/Linux-x86/304.60/README/commonproblems.html where they said that the same problem occured when the "VBIOS fail to load on my Optimus system". They said about this error "Such problems are typically beyond the control of the NVIDIA driver, which relies on proper cooperation of ACPI and the System BIOS to retrieve important information about the GPU, including the Video
BIOS."
I don't really understand the problem. I will really appreciate any help to
troubleshoot the problem and any links that can help are more than welcome :).
I will also appreciate any comment about a succesfull story with PCI
assignement with a K1. I don't have all of them but I maybe will have
the opportunity to test them so any comments are welcome :).
Best regards,
Guillaume
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to ***@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html