After yesterday''s checkins to unstable, I can no longer boot an fully virtualized guest (Intel VT). Here is an attempt at booting a RHEL4U2-32bit uniprocessor kernel guest. The flood of ''(XEN) HVM_PIT:missed ticks < 0'' messages actually occur before getting to the guest''s grub menu so the guest should not matter. (XEN) (GUEST: 5) HVM Loader (XEN) (GUEST: 5) Loading ROMBIOS ... (XEN) (GUEST: 5) Loading Cirrus VGABIOS ... (XEN) (GUEST: 5) Loading VMXAssist ... (XEN) (GUEST: 5) VMX go ... (XEN) (GUEST: 5) VMXAssist (Apr 23 2006) (XEN) (GUEST: 5) Memory size 256 MB (XEN) (GUEST: 5) E820 map: (XEN) (GUEST: 5) 0000000000000000 - 000000000009F800 (RAM) (XEN) (GUEST: 5) 000000000009F800 - 00000000000A0000 (Reserved) (XEN) (GUEST: 5) 00000000000A0000 - 00000000000C0000 (Type 16) (XEN) (GUEST: 5) 00000000000F0000 - 0000000000100000 (Reserved) (XEN) (GUEST: 5) 0000000000100000 - 000000000FFFE000 (RAM) (XEN) (GUEST: 5) 000000000FFFE000 - 000000000FFFF000 (Type 18) (XEN) (GUEST: 5) 000000000FFFF000 - 0000000010000000 (Type 17) (XEN) (GUEST: 5) 0000000010000000 - 0000000010003000 (ACPI NVS) (XEN) (GUEST: 5) 0000000010003000 - 000000001000D000 (ACPI Data) (XEN) (GUEST: 5) 00000000FEC00000 - 0000000100000000 (Type 16) (XEN) (GUEST: 5) (XEN) (GUEST: 5) Start BIOS ... (XEN) (GUEST: 5) Starting emulated 16-bit real-mode: ip=F000:FFF0 (XEN) (GUEST: 5) rombios.c,v 1.138 2005/05/07 15:55:26 vruppert Exp $ (XEN) (GUEST: 5) Remapping master: ICW2 0x8 -> 0x20 (XEN) (GUEST: 5) Remapping slave: ICW2 0x70 -> 0x28 (XEN) (GUEST: 5) VGABios $Id: vgabios.c,v 1.61 2005/05/24 16:50:50 vruppert Exp $ (XEN) (GUEST: 5) HVMAssist BIOS, 8 cpus, $Revision: 1.138 $ $Date: 2005/05/07 15:55:26 $ (XEN) (GUEST: 5) (XEN) HVM_PIT:missed ticks < 0 (XEN) HVM_PIT:missed ticks < 0 (XEN) HVM_PIT:missed ticks < 0 _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 23 Apr 2006, at 14:03, Ed Smith wrote:> After yesterday''s checkins to unstable, I can no longer boot an fully > virtualized guest (Intel VT). Here is an attempt at booting a > RHEL4U2-32bit uniprocessor kernel guest. The flood of ''(XEN) > HVM_PIT:missed ticks < 0'' messages actually occur before getting to > the guest''s grub menu so the guest should not matter.I think this has started since Edwin Zhai''s PIT emulation changes. I applied a further patch from him to quieten down debug output for SMP guests, but if those messages are coming out that early for you then it looks like maybe the patch will only hide the underlying problem. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Sun, Apr 23, 2006 at 02:39:27PM +0100, Keir Fraser wrote:> > On 23 Apr 2006, at 14:03, Ed Smith wrote: > > >After yesterday''s checkins to unstable, I can no longer boot an fully > >virtualized guest (Intel VT).? Here is an attempt at booting a > >RHEL4U2-32bit uniprocessor kernel guest.? The flood of ''(XEN) > >HVM_PIT:missed ticks < 0'' messages actually occur before getting to > >the guest''s grub menu so the guest should not matter. > > I think this has started since Edwin Zhai''s PIT emulation changes. I > applied a further patch from him to quieten down debug output for SMP > guests, but if those messages are coming out that early for you then it > looks like maybe the patch will only hide the underlying problem. >we have an optimization in original code: pit actimer will be stopped when the vcpu is sched out, so we have to check the missed ticks and rearm actimer when each vcpu is ched in with an inactive pit actimer. in smp environment, if cpu 0 has checked missed ticks in pit actimer function, then cpu 1 will probably find no missed time(now < schedule time) and complain when sched in. in summary, this redundant debug info can be safely removed. but we''ll come with perfect pit solution for smp.> -- Keir-- thanks, edwin _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> guests, but if those messages are coming out that early for you then > it looks like maybe the patch will only hide the underlying problem. >Yes, we have collected bunch of SMP time virtualization issue and the design is ongoing now. Eddie _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel