VGA device assignment - Revision history

Ckotichas: /* Open issues */

2017-01-20T02:44:19Z

Open issues

← Older revision		Revision as of 22:44, 19 January 2017
Line 5:		Line 5:
	== Open issues ==		== Open issues ==

	* '''Insufficient PCI window size'''<br>This is a fixable SeaBIOS limitation. Patches are currently under development by [http://www.kraxel.org/cgit/seabios/log/?h=kraxel.q35&id=99ff4d141d5a460eaf20fe2d4d13117c888eae15 Gerd Hoffmann]		* '''Insufficient PCI window size'''<br>This is a fixable SeaBIOS limitation. Patches are currently under development by [http://www.kraxel.org/cgit/seabios/log/?h=kraxel.q35&id=99ff4d141d5a460eaf20fe2d4d13117c888eae15 Gerd Hoffmann].
	* '''Non-standard I/O region remapping'''<br>Some NVIDIA Quadro adapters are known to be affected by this, but probably only during early boot. A workaround could be identity mapping, i.e. locating the memory BARs in the guest at the same locations as on the host. Of course, this breaks if the guest OS or guest drivers decide to remap them.		* '''Non-standard I/O region remapping'''<br>Some NVIDIA Quadro adapters are known to be affected by this, but probably only during early boot. A workaround could be identity mapping, i.e. locating the memory BARs in the guest at the same locations as on the host. Of course, this breaks if the guest OS or guest drivers decide to remap them.
	* '''Legacy VGA support'''<br>In order to get boot messages of the guest BIOS, boot loader, or early OS stages, access to the legacy VGA PIO and MMIO regions need to be forwarded to the passed-through adapter. Moreover, the VBIOS of the adapter must be bootable inside the guest. As a result, the following aspects will need to be addressed:		* '''Legacy VGA support'''<br>In order to get boot messages of the guest BIOS, boot loader, or early OS stages, access to the legacy VGA PIO and MMIO regions need to be forwarded to the passed-through adapter. Moreover, the VBIOS of the adapter must be bootable inside the guest. As a result, the following aspects will need to be addressed:

Ckotichas: /* Assigning physical VGA adapters */

2017-01-20T02:43:28Z

Assigning physical VGA adapters

Espes: other successes

2013-03-03T17:09:17Z

other successes

← Older revision		Revision as of 13:09, 3 March 2013
Line 22:		Line 22:

	* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/94524 Radeon HD 7870]		* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/94524 Radeon HD 7870]
			* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/104760 Radeon HD 7750]
	* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/72987 Radeon HD 6950 with patched SeaBIOS]		* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/72987 Radeon HD 6950 with patched SeaBIOS]
			* [http://tavi-tech.blogspot.com.au/2012/05/vga-passthrough-kvm-fedora-17-and.html Radeon HD 6770]
	* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/71981 Radeon HD 5850]		* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/71981 Radeon HD 5850]

El Presidente at 14:01, 18 July 2012

2012-07-18T14:01:21Z

← Older revision		Revision as of 10:01, 18 July 2012
Line 21:		Line 21:
	Despite all existing problems, some users already succeeded passing through VGA adapters:		Despite all existing problems, some users already succeeded passing through VGA adapters:

			* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/94524 Radeon HD 7870]
	* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/72987 Radeon HD 6950 with patched SeaBIOS]		* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/72987 Radeon HD 6950 with patched SeaBIOS]
	* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/71981 Radeon HD 5850]		* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/71981 Radeon HD 5850]

Jan at 10:58, 27 May 2011

2011-05-27T10:58:48Z

New page

= Assigning physical VGA adapters =

On first glance, assigning modern PCI/PCIe VGA adapters should not be different from [[How_to_assign_devices_with_VT-d_in_KVM|passing through normal PCI devices]]. But there are additional challenges to solve, specifically related to legacy VGA requirement and proprietary mechanisms to relocate I/O regions of the adapters. Moreover, some QEMU and SeaBIOS deficits contribute to the current situation that things rarely work out of the box.

== Open issues ==

* '''Insufficient PCI window size''' This is a fixable SeaBIOS limitations. Patches are under development, e.g. by [http://www.kraxel.org/cgit/seabios/log/?h=kraxel.q35&id=99ff4d141d5a460eaf20fe2d4d13117c888eae15 Gerd Hoffmann]
* '''Non-standard I/O region remapping''' At least NVIDIA adapters of the Quadro series a known to be affected by this, but probably only during early boot. A workaround could be identity mapping, i.e. locating the memory BARs in the guest at the same locations as on the host. Of course, this breaks if guest OS or guest driver decided to remap them for some reason.
* '''Legacy VGA support''' In order to get boot messages of the guest BIOS, boot loader, or early OS stages, access to the legacy VGA PIO and MMIO regions need to be forwarded to the passed-through adapter. Moreover, the VBIOS of the adapter must be bootable inside the guest. In details, the following aspects need to be addressed:
** '''Legacy VGA pass-through''' QEMU requires patches to forward guest access to legacy VGA to the assigned adapter.
** '''PAM/SRAM QEMU breakage''' Current QEMU contains broken code that prevents access to legacy VGA even if the adapter is emulated but differently initialized (test case: -vga none -device VGA)
** '''Legacy VGA arbitration''' As the legacy IO resources cannot be remapped, access need to be switched between guests and host on demand. The Linux kernel provides a VGA arbiter for this purpose, but it is not used consistently by all involved parties. Moreover, QEMU requires additional patches to support this.
** '''VBIOS ROM access''' To re-run the POST procedures of the assigned adapter inside the guest, the proper VBIOS ROM image has to be used. However, when passing through the primary adapter of the host, Linux provides only access to the shadowed version of the VBIOS which may differ from the pre-POST version (due to modification applied during POST). This has be been observed with NVDIA Quadro adapters. A workaround is to retrieve the VBIOS from the adapter while it is in secondary mode and use this saved image (-device pci-assign,...,romfile=...). But even that may fail, either due to problems of the host chipset or BIOS (host kernel complains about unmappable ROM BAR).

== Potential tweaks & tricks ==

* Some adapters may advertise MSI support even if that feature does not work properly. In that case, KVM will use MSI on the host side unconditionally unless instructed to follow the guest driver usage via -device pci-assign,...,prefer_msi=off.

== Success stories ==

Despite all existing problems, some users already succeeded passing through VGA adapters:

* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/72987 Radeon HD 6950 with patched SeaBIOS]
* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/71981 Radeon HD 5850]

← Older revision		Revision as of 22:43, 19 January 2017
Line 1:		Line 1:
	= Assigning physical VGA adapters =		= Assigning physical VGA adapters =

	~~On first glance,~~ assigning modern PCI/PCIe VGA adapters should not be different from [[How_to_assign_devices_with_VT-d_in_KVM\|passing through normal PCI devices]]. But there are additional challenges to solve, specifically related to legacy VGA ~~requirement~~ and proprietary mechanisms to relocate I/O regions of the adapters. Moreover, some QEMU and SeaBIOS deficits contribute to the current situation that things rarely work out of the box.		It would appear that assigning modern PCI/PCIe VGA adapters should not be any different from [[How_to_assign_devices_with_VT-d_in_KVM\|passing through normal PCI devices]]. But there are additional challenges to solve, specifically related to legacy VGA requirements and proprietary mechanisms to relocate I/O regions of the adapters. Moreover, some QEMU and SeaBIOS deficits contribute to the current situation that things rarely work out of the box.

	== Open issues ==		== Open issues ==

	* '''Insufficient PCI window size'''<br>This is a fixable SeaBIOS ~~limitations~~. Patches are under development~~, e.g.~~ by [http://www.kraxel.org/cgit/seabios/log/?h=kraxel.q35&id=99ff4d141d5a460eaf20fe2d4d13117c888eae15 Gerd Hoffmann]		* '''Insufficient PCI window size'''<br>This is a fixable SeaBIOS limitation. Patches are currently under development by [http://www.kraxel.org/cgit/seabios/log/?h=kraxel.q35&id=99ff4d141d5a460eaf20fe2d4d13117c888eae15 Gerd Hoffmann]
	* '''Non-standard I/O region remapping'''<br>~~At least~~ NVIDIA adapters ~~of the Quadro series a~~ known to be affected by this, but probably only during early boot. A workaround could be identity mapping, i.e. locating the memory BARs in the guest at the same locations as on the host. Of course, this breaks if guest OS or guest ~~driver decided~~ to remap them ~~for some reason~~.		* '''Non-standard I/O region remapping'''<br>Some NVIDIA Quadro adapters are known to be affected by this, but probably only during early boot. A workaround could be identity mapping, i.e. locating the memory BARs in the guest at the same locations as on the host. Of course, this breaks if the guest OS or guest drivers decide to remap them.
	* '''Legacy VGA support'''<br>In order to get boot messages of the guest BIOS, boot loader, or early OS stages, access to the legacy VGA PIO and MMIO regions need to be forwarded to the passed-through adapter. Moreover, the VBIOS of the adapter must be bootable inside the guest. ~~In details~~, the following aspects need to be addressed:		* '''Legacy VGA support'''<br>In order to get boot messages of the guest BIOS, boot loader, or early OS stages, access to the legacy VGA PIO and MMIO regions need to be forwarded to the passed-through adapter. Moreover, the VBIOS of the adapter must be bootable inside the guest. As a result, the following aspects will need to be addressed:
	** '''Legacy VGA pass-through'''<br>QEMU requires patches to forward guest access to legacy VGA to the assigned adapter.		** '''Legacy VGA pass-through'''<br>QEMU requires patches to forward guest access to legacy VGA via the assigned adapter.
	** '''PAM/SRAM QEMU breakage'''<br>Current QEMU contains broken code that prevents access to legacy VGA even if the adapter is emulated but differently initialized (test case: -vga none -device VGA)		** '''PAM/SRAM QEMU breakage'''<br>Current QEMU contains broken code that prevents access to legacy VGA even if the adapter is emulated but differently initialized (test case: -vga none -device VGA)
	** '''Legacy VGA arbitration'''<br>As the legacy IO resources cannot be remapped, access ~~need~~ to be switched between guests and host on demand. The Linux kernel provides a VGA arbiter for this purpose, but it is not used consistently by all involved parties. Moreover, QEMU requires additional patches to support this.		** '''Legacy VGA arbitration'''<br>As the legacy IO resources cannot be remapped, access needs to be switched between guests and the host on demand. The Linux kernel provides a VGA arbiter for this purpose, but it is not used consistently by all involved parties. Moreover, QEMU also requires additional patches to support this.
	** '''VBIOS ROM access'''<br>To re-run the POST procedures of the assigned adapter inside the guest, the proper VBIOS ROM image has to be used. However, when passing through the primary adapter of the host, Linux provides only access to the shadowed version of the VBIOS which may differ from the pre-POST version (due to ~~modification~~ applied during POST). This has be been observed with ~~NVDIA~~ Quadro adapters. A workaround is to retrieve the VBIOS from the adapter while it is in secondary mode and use this saved image (-device pci-assign,...,romfile=...). But even that may fail, either due to problems of the host chipset or BIOS (host kernel complains about unmappable ROM BAR).		** '''VBIOS ROM access'''<br>To re-run the POST procedures of the assigned adapter inside the guest, the proper VBIOS ROM image has to be used. However, when passing through the primary adapter of the host, Linux provides only access to the shadowed version of the VBIOS which may differ from the pre-POST version (due to modifications applied during POST). This has been observed with NVIDIA Quadro adapters. A workaround is to retrieve the VBIOS from the adapter while it is in secondary mode and use this saved image (-device pci-assign,...,romfile=...). But even that may fail, either due to problems of the host chipset or BIOS (e.g. host kernel complains about an unmappable ROM BAR).

	== Potential tweaks & tricks ==		== Potential tweaks & tricks ==
Line 19:		Line 19:
	== Success stories ==		== Success stories ==

	Despite all existing problems, some users already succeeded ~~passing~~ through VGA adapters:		Despite all existing problems, some users have already succeeded in utilizing pass-through functionality for various VGA adapters:

	* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/94524 Radeon HD 7870]		* [http://thread.gmane.org/gmane.comp.emulators.kvm.devel/94524 Radeon HD 7870]