A few simple questions. - How many VM can be setup for HVM? - Would it make sense to have more than one? - I heard that PCI passthrough or what ever it is called is not supported in HVM. Is that a permanent limitation or will it soon become available? Unfortunately, I need 1 Windows VM that can use dedicated PCI card. Since Windows is only supported in HVM, then this is why I ask this question. - I am seeing conflicting posts and wiki entries about AM2 Sempron support for SVM. Does the AM2 Sempron support HVM? Thanks, Greg _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> -----Original Message----- > From: xen-users-bounces@lists.xensource.com > [mailto:xen-users-bounces@lists.xensource.com] On Behalf Of > Gregory Gee > Sent: 10 July 2006 22:59 > To: xen-users@lists.xensource.com > Subject: [Xen-users] HVM questions > > > A few simple questions. > > - How many VM can be setup for HVM?As many as you can afford memory for.> > - Would it make sense to have more than one? > > - I heard that PCI passthrough or what ever it is called is not > supported in HVM. Is that a permanent limitation or will it > soon become > available? Unfortunately, I need 1 Windows VM that can use dedicated > PCI card. Since Windows is only supported in HVM, then this is why I > ask this question.Yes, PCI passthrough is currently only supported for Para-virtual guests. It is expected to be solved (there''s no technical limitation for this, just that someone has to write some code to support it, basically - and the principle is not difficult - just fill in the PCI-table in the guest so that the device looks like its there). There are some other limitations tho''. If the PCI-card uses direct memory access, you''d need to use a para-virtual driver (i.e a driver that is aware of the fact that physical memory isn''t where the guest-OS thinks it is). PIO-type card will work fine without special drivers - unfortunately, very few PCI-cards are PIO-based...> > - I am seeing conflicting posts and wiki entries about AM2 Sempron > support for SVM. Does the AM2 Sempron support HVM?As far as I can tell from the rather poorly documented feature-differences, Sempron will not support SVM, but all other RevF products (Turion, Athlon64 and the to-be-released Opterons) do. Sorry about that... -- Mats> > Thanks, > Greg > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
>> - I heard that PCI passthrough or what ever it is called is not >> supported in HVM. Is that a permanent limitation or will it >> soon become >> available? Unfortunately, I need 1 Windows VM that can use dedicated >> PCI card. Since Windows is only supported in HVM, then this is why I >> ask this question. > > Yes, PCI passthrough is currently only supported for Para-virtual > guests. > It is expected to be solved (there''s no technical limitation for this, > just that someone has to write some code to support it, basically - and > the principle is not difficult - just fill in the PCI-table in the guest > so that the device looks like its there). There are some other > limitations tho''. If the PCI-card uses direct memory access, you''d need > to use a para-virtual driver (i.e a driver that is aware of the fact > that physical memory isn''t where the guest-OS thinks it is). > > PIO-type card will work fine without special drivers - unfortunately, > very few PCI-cards are PIO-based... > -- > MatsInteresting. Mixed answer. So it sounds like it will be supported, but for specific PCI cards only or a para-vitual driver(SOL if guest in Windows I guess?). Do vendors usually list if a card is PIO? Thanks, Greg _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> -----Original Message----- > From: xen-users-bounces@lists.xensource.com > [mailto:xen-users-bounces@lists.xensource.com] On Behalf Of > Gregory Gee > Sent: 11 July 2006 02:22 > To: xen-users@lists.xensource.com > Subject: RE: [Xen-users] HVM questions > > > >> - I heard that PCI passthrough or what ever it is called is not > >> supported in HVM. Is that a permanent limitation or will it > >> soon become > >> available? Unfortunately, I need 1 Windows VM that can > use dedicated > >> PCI card. Since Windows is only supported in HVM, then > this is why I > >> ask this question. > > > > Yes, PCI passthrough is currently only supported for Para-virtual > > guests. > > It is expected to be solved (there''s no technical > limitation for this, > > just that someone has to write some code to support it, > basically - and > > the principle is not difficult - just fill in the PCI-table > in the guest > > so that the device looks like its there). There are some other > > limitations tho''. If the PCI-card uses direct memory > access, you''d need > > to use a para-virtual driver (i.e a driver that is aware of the fact > > that physical memory isn''t where the guest-OS thinks it is). > > > > PIO-type card will work fine without special drivers - > unfortunately, > > very few PCI-cards are PIO-based... > > -- > > Mats > > Interesting. Mixed answer. So it sounds like it will be > supported, but > for specific PCI cards only or a para-vitual driver(SOL if guest in > Windows I guess?). Do vendors usually list if a card is PIO?Probably not - but reading the specs or looking at the driver would tell (to someone who understands how these things work). PIO would be used for cards that handle very little data at a time, serial cards or simple cards for intefacing with external sensors for example. Anything that does high-speed/mass-transfer work, like grapics, network- and disk-controllers would definitely use DMA. In the future, machines will have a IOMMU-device so the hypervisor should be able to set up a mapping of what the guest thinks is physical addresses and what the REAL physical address is, so that PCI-cards can be used without modifications to the driver. -- Mats> > Thanks, > Greg > > > > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xensource.com > http://lists.xensource.com/xen-users > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
To Mats, Gregory, and Xen-Users: Does the below limitation also explain the current issues with accessing pass-through video cards in domU via X as described in Julian Davison''s post "X11 in domU" (4 July 2006)? Petersson''s E-mail:>Yes, PCI passthrough is currently only supported for Para-virtual guests. >It is expected to be solved (there''s no technical limitation for this, just>that someone has to write some code to support it, basically - and the >principle is not difficult - just fill in the PCI-table in the guest sothat>the device looks like its there). There are some other limitations tho''. If>the PCI-card uses direct memory access, you''d need to use a para-virtual >driver (i.e a driver that is aware of the fact that physical memory isn''t >where the guest-OS thinks it is).>PIO-type card will work fine without special drivers - unfortunately, >very few PCI-cards are PIO-based...Davison''s E-mail:>Using "X -novtswitch -sharevts vt1" I ended up being told: > >-------------- copy of end of X output starts -------------- >(EE) I810(0): Cannot map SYS BIOS > >Fatal server error: >xf86MapVidMem: Could not mmap framebuffer (0x000a0000,0x10000) (Badaddress)>-------------- copy of end of X output ends --------------Thank-you. John Helder gtg032i@mail.gatech.edu _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> -----Original Message----- > From: John Helder [mailto:gtg032i@mail.gatech.edu] > Sent: 11 July 2006 14:18 > To: Petersson, Mats; ''Gregory Gee''; xen-users@lists.xensource.com > Subject: RE: [Xen-users] HVM questions > > To Mats, Gregory, and Xen-Users: > > > Does the below limitation also explain the current issues > with accessing > pass-through video cards in domU via X as described in Julian > Davison''s post > "X11 in domU" (4 July 2006)?Only if you try to do it in HVM domain - which you can''t do anyways, as you can''t pass the PCI (AGP/PCIe are PCI in this case) to the DomU in HVM-mode. It''s probably broken in some other way... -- Mats> > > Petersson''s E-mail: > >Yes, PCI passthrough is currently only supported for > Para-virtual guests. > >It is expected to be solved (there''s no technical limitation > for this, just > > >that someone has to write some code to support it, basically > - and the > >principle is not difficult - just fill in the PCI-table in > the guest so > that > >the device looks like its there). There are some other > limitations tho''. If > > >the PCI-card uses direct memory access, you''d need to use a > para-virtual > >driver (i.e a driver that is aware of the fact that physical > memory isn''t > >where the guest-OS thinks it is). > > >PIO-type card will work fine without special drivers - > unfortunately, > >very few PCI-cards are PIO-based... > > > > Davison''s E-mail: > >Using "X -novtswitch -sharevts vt1" I ended up being told: > > > >-------------- copy of end of X output starts -------------- > >(EE) I810(0): Cannot map SYS BIOS > > > >Fatal server error: > >xf86MapVidMem: Could not mmap framebuffer (0x000a0000,0x10000) (Bad > address) > >-------------- copy of end of X output ends -------------- > > > Thank-you. > > John Helder > gtg032i@mail.gatech.edu > > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users