Hi list, 1. Is it possible to share contiguous memory regions between a hvm domain and dom0 using the grant tables? 2. Why do we need the piix bridge in the ioemu? is it being used at all? Thanks in advance for your help, Eric. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 26/10/06 6:27 pm, "Eric Benton" <benton71@gmail.com> wrote:> 1. Is it possible to share contiguous memory regions between a hvm domain and > dom0 using the grant tables? > 2. Why do we need the piix bridge in the ioemu? is it being used at all?1. Yes. The PV-on-HVM drivers do this (from HVM to dom0). Other way (dom0 pages mapped in HVM) will not work as yet. 2. It¹s used for PCI-ISA IRQ routing. I wondered about removing it but I think I¹ve changed my mind for the time being. If you boot an HVM guest in APIC mode, it will not have the 4 IRQ routers¹ limitation. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Thanks Keir for the fast response! To implement the PV-on-HVM, have you used a PCI probing in the hypervisor? I don''t get how you left the HVM operating system unmodified? Is it a FE/BE model? Does the HVM uses the "old" dom0 FEs? Thanks again, Eric. On 10/26/06, Keir Fraser <Keir.Fraser@cl.cam.ac.uk> wrote:> > > > > On 26/10/06 6:27 pm, "Eric Benton" <benton71@gmail.com> wrote: > > 1. Is it possible to share contiguous memory regions between a hvm domain > and dom0 using the grant tables? > 2. Why do we need the piix bridge in the ioemu? is it being used at all? > > > > 1. Yes. The PV-on-HVM drivers do this (from HVM to dom0). Other way > (dom0 pages mapped in HVM) will not work as yet. > 2. It''s used for PCI-ISA IRQ routing. I wondered about removing it > but I think I''ve changed my mind for the time being. If you boot an HVM > guest in APIC mode, it will not have the ''4 IRQ routers'' limitation. > > > -- Keir >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 26/10/06 7:31 pm, "Eric Benton" <benton71@gmail.com> wrote:> Thanks Keir for the fast response! > > To implement the PV-on-HVM, have you used a PCI probing in the hypervisor? > I don''t get how you left the HVM operating system unmodified? Is it a FE/BE > model? Does the HVM uses the "old" dom0 FEs?There¹s a fake PCI xen platform¹ device that triggers loading of our PCI driver in the HVM guest, and we hook everything else off that event. All the magic to do PV-on-HVM is hidden in our xen-platform driver, so there is no modification to the core HVM OS kernel. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ross Maxfield
2006-Oct-26 23:25 UTC
[Xen-devel] Spurious Interrupts with PV drivers for HVM guests
I''ve been testing PV drivers on SLES9 and SLES10 HVM guests and have been very impressed with the performance results. During this testing, however, I happened to cat out /proc/interrupts and noticed that ERR: was counting up nearly as fast as INT10 - millions/minute. I expect INT10 to clip along at this rate as my testing is LAN and disk intensive, but a similar rate for spurious interrupts was a great surprise. I''m curious to know if others are seeing the same thing. BTW, I''m using NetBench in this particular test case so I can''t imagine this having anything to do with the test directly. I also see spurious interrupt occur by simply pinging the the guest from a remote client. -- Ross _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Steven Smith
2006-Oct-31 20:48 UTC
Re: [Xen-devel] Spurious Interrupts with PV drivers for HVM guests
> I''ve been testing PV drivers on SLES9 and SLES10 HVM guests and have > been very impressed with the performance results. During this > testing, however, I happened to cat out /proc/interrupts and noticed > that ERR: was counting up nearly as fast as INT10 - millions/minute.Okay, this is hopefully fixed by cset 12040 of xen-unstable. Thanks! Steven. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel