Hi All, It seems my upgrade to rc4 of xen-unstable went pretty smoothly ... only two things i noticed so far: - Had to disable startup of xend by hand, perhaps a bit strange if it''s deprecated ? - Messages i haven''t seen in dmesg with xen-4.1.3 in combination with 3.6-rc? kernel: [ 3.333195] tseg: 0000000000 [ 3.333231] CPU: Physical Processor ID: 0 [ 3.333236] CPU: Processor Core ID: 0 [ 3.333240] mce: CPU supports 2 MCE banks [ 3.333258] LVT offset 0 assigned for vector 0xf9 [ 3.333263] [Firmware Bug]: cpu 0, try to use APIC500 (LVT offset 0) for vector 0xf9, but the register is already in use for vector 0x0 on this cpu [ 3.333272] [Firmware Bug]: cpu 0, failed to setup threshold interrupt for bank 4, block 0 (MSR00000413=0xc008000001000000) [ 3.333281] [Firmware Bug]: cpu 0, try to use APIC500 (LVT offset 0) for vector 0xf9, but the register is already in use for vector 0x0 on this cpu [ 3.333290] [Firmware Bug]: cpu 0, failed to setup threshold interrupt for bank 4, block 1 (MSRC0000408=0xc000000001000000) [ 3.333306] [Firmware Bug]: cpu 0, try to use APIC500 (LVT offset 0) for vector 0xf9, but the register is already in use for vector 0x0 on this cpu [ 3.333315] [Firmware Bug]: cpu 0, failed to setup threshold interrupt for bank 4, block 2 (MSRC0000409=0xc000000001000000) [ 3.333331] Last level iTLB entries: 4KB 0, 2MB 0, 4MB 0 [ 3.333331] Last level dTLB entries: 4KB 0, 2MB 0, 4MB 0 [ 3.333331] tlb_flushall_shift is 0xffffffff [ 3.334057] ACPI: Core revision 20120711 The kernel gives these firmware bugs for the other CPUS later on, this is on a AMD Phenom x6, but everything seems to be working so perhaps a minor thing. For the rest, all my domains and pci-passthrough seem to be working :-) Thx, -- Sander
On Fri, 2012-08-31 at 10:44 +0100, Sander Eikelenboom wrote:> Hi All, > > It seems my upgrade to rc4 of xen-unstable went pretty smoothly ... only two things i noticed so far: > > - Had to disable startup of xend by hand, perhaps a bit strange if it''s deprecated ?I can''t for the life of me find the bit of Xen''s build+install system which actually enables the initscripts. My installation notes include an explicit step to do this, but I build and deploy on different machines so perhaps I miss out on part of the usual infrastructure. Are you sure this isn''t part of your packaging or surrounding scripts? Or perhaps just left over rc?.d symlinks from before you upgraded?> - Messages i haven''t seen in dmesg with xen-4.1.3 in combination with 3.6-rc? kernel:[...ducks...]> For the rest, all my domains and pci-passthrough seem to be working :-)Excellent new! Thanks for the report. Ian.
>>> On 31.08.12 at 11:44, Sander Eikelenboom <linux@eikelenboom.it> wrote: > - Messages i haven''t seen in dmesg with xen-4.1.3 in combination with 3.6-rc? > kernel: > > [ 3.333195] tseg: 0000000000 > [ 3.333231] CPU: Physical Processor ID: 0 > [ 3.333236] CPU: Processor Core ID: 0 > [ 3.333240] mce: CPU supports 2 MCE banks > [ 3.333258] LVT offset 0 assigned for vector 0xf9 > [ 3.333263] [Firmware Bug]: cpu 0, try to use APIC500 (LVT offset 0) for > vector 0xf9, but the register is already in use for vector 0x0 on this cpu > [ 3.333272] [Firmware Bug]: cpu 0, failed to setup threshold interrupt > for bank 4, block 0 (MSR00000413=0xc008000001000000) > [ 3.333281] [Firmware Bug]: cpu 0, try to use APIC500 (LVT offset 0) for > vector 0xf9, but the register is already in use for vector 0x0 on this cpu > [ 3.333290] [Firmware Bug]: cpu 0, failed to setup threshold interrupt > for bank 4, block 1 (MSRC0000408=0xc000000001000000) > [ 3.333306] [Firmware Bug]: cpu 0, try to use APIC500 (LVT offset 0) for > vector 0xf9, but the register is already in use for vector 0x0 on this cpu > [ 3.333315] [Firmware Bug]: cpu 0, failed to setup threshold interrupt > for bank 4, block 2 (MSRC0000409=0xc000000001000000) > [ 3.333331] Last level iTLB entries: 4KB 0, 2MB 0, 4MB 0 > [ 3.333331] Last level dTLB entries: 4KB 0, 2MB 0, 4MB 0 > [ 3.333331] tlb_flushall_shift is 0xffffffff > [ 3.334057] ACPI: Core revision 20120711 > > The kernel gives these firmware bugs for the other CPUS later on, this is on > a AMD Phenom x6, but everything seems to be working so perhaps a minor thing.This is the usual effect of keeping all sorts of native code enabled in the pv-ops kernel that tries to control what really is being controlled by Xen (MCE in this case). Jan
> > - Messages i haven''t seen in dmesg with xen-4.1.3 in combination with 3.6-rc? kernel: > [...ducks...]I think that can easily be fixed. I seem to be seeing those too.> > > For the rest, all my domains and pci-passthrough seem to be working :-)Oh nice. My AMD machine reboots when I tried HVM and PCI passthrough :-( Hadn''t yet actually done any git bisection or tried to get a serial output.
Friday, August 31, 2012, 7:29:11 PM, you wrote:>> > - Messages i haven''t seen in dmesg with xen-4.1.3 in combination with 3.6-rc? kernel: >> [...ducks...]> I think that can easily be fixed. I seem to be seeing those too. >> >> > For the rest, all my domains and pci-passthrough seem to be working :-)> Oh nice. My AMD machine reboots when I tried HVM and PCI passthrough :-( > Hadn''t yet actually done any git bisection or tried to get a serial > output.I''m mostly using PV and PCI passthrough, haven''t tried HVM and PCI passthrough yet, but will somewhere this weekend ..
On Fri, Aug 31, 2012 at 07:32:35PM +0200, Sander Eikelenboom wrote:> > Friday, August 31, 2012, 7:29:11 PM, you wrote: > > >> > - Messages i haven''t seen in dmesg with xen-4.1.3 in combination with 3.6-rc? kernel: > >> [...ducks...] > > > I think that can easily be fixed. I seem to be seeing those too. > >> > >> > For the rest, all my domains and pci-passthrough seem to be working :-) > > > Oh nice. My AMD machine reboots when I tried HVM and PCI passthrough :-( > > Hadn''t yet actually done any git bisection or tried to get a serial > > output. > > I''m mostly using PV and PCI passthrough, haven''t tried HVM and PCI passthrough yet, but will somewhere this weekend ..For v3.5 and later there is a bug in the Linux code (introduced by me) which makes MSI in HVM PCI passthrough guests not work. You would need to revert cd9db80e5257682a7f7ab245a2459648b3c8d268.> > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel >
Hello Konrad, Friday, August 31, 2012, 7:39:42 PM, you wrote:> On Fri, Aug 31, 2012 at 07:32:35PM +0200, Sander Eikelenboom wrote: >> >> Friday, August 31, 2012, 7:29:11 PM, you wrote: >> >> >> > - Messages i haven''t seen in dmesg with xen-4.1.3 in combination with 3.6-rc? kernel: >> >> [...ducks...] >> >> > I think that can easily be fixed. I seem to be seeing those too. >> >> >> >> > For the rest, all my domains and pci-passthrough seem to be working :-) >> >> > Oh nice. My AMD machine reboots when I tried HVM and PCI passthrough :-( >> > Hadn''t yet actually done any git bisection or tried to get a serial >> > output. >> >> I''m mostly using PV and PCI passthrough, haven''t tried HVM and PCI passthrough yet, but will somewhere this weekend ..> For v3.5 and later there is a bug in the Linux code (introduced by me) > which makes MSI in HVM PCI passthrough guests not work. You would need > to revert cd9db80e5257682a7f7ab245a2459648b3c8d268.Ok Found a problem with PCI passthrough to the PV guest as well, seems to be IOMMU related, never seen IO_PAGE_FAULT''s on 4.1.x: xl dmesg: (XEN) [2012-08-31 09:46:16] AMD-Vi: IO_PAGE_FAULT: domain = 14, device id = 0x0700, fault address = 0xa99bfee0 (XEN) [2012-08-31 09:46:16] AMD-Vi: IO_PAGE_FAULT: domain = 14, device id = 0x0700, fault address = 0xa99bff00 (XEN) [2012-08-31 09:46:16] AMD-Vi: IO_PAGE_FAULT: domain = 14, device id = 0x0700, fault address = 0xa99bff40 (XEN) [2012-08-31 09:46:16] AMD-Vi: IO_PAGE_FAULT: domain = 14, device id = 0x0700, fault address = 0xa99bff60 (XEN) [2012-08-31 09:46:16] AMD-Vi: IO_PAGE_FAULT: domain = 14, device id = 0x0700, fault address = 0xa99bffc0 (XEN) [2012-08-31 09:46:16] AMD-Vi: IO_PAGE_FAULT: domain = 14, device id = 0x0700, fault address = 0xa99bff80 (XEN) [2012-08-31 09:46:16] AMD-Vi: IO_PAGE_FAULT: domain = 14, device id = 0x0700, fault address = 0xa99bffe0 (XEN) [2012-08-31 09:56:20] grant_table.c:254:d0 Increased maptrack size to 2 frames>> >> >> _______________________________________________ >> Xen-devel mailing list >> Xen-devel@lists.xen.org >> http://lists.xen.org/xen-devel >>-- Best regards, Sander mailto:linux@eikelenboom.it