[Bug 473217] New: PCI Pass-Through IRQ trouble XEN Dom0/DomU
https://bugzilla.novell.com/show_bug.cgi?id=473217 User wolf@womaro.ch added comment https://bugzilla.novell.com/show_bug.cgi?id=473217#c463596 Summary: PCI Pass-Through IRQ trouble XEN Dom0/DomU Classification: openSUSE Product: openSUSE 11.1 Version: Final Platform: 64bit OS/Version: openSUSE 11.1 Status: NEW Severity: Normal Priority: P5 - None Component: Xen AssignedTo: cgriffin@novell.com ReportedBy: wolf@womaro.ch QAContact: qa@suse.de CC: jbeulich@novell.com, lbendixs@novell.com, frank.mueller@rhoen.de, wolf@womaro.ch Depends on: 463596 Found By: Customer Created an attachment (id=270722) --> (https://bugzilla.novell.com/attachment.cgi?id=270722) Dom0 (xenos) and DomU (infra) log +++ This bug was initially created as a clone of Bug #463596 +++ The fix supplied by Jan Beulich for bug #463596 solved the original problem in the combination of the stock 11.1 2.6.27.9 Dom0 kernel with a patched 2.6.27.10 DomU. I've now upgraded the Dom0 kernel from 2.6.27.9 to 2.6.27.10 (factory) - and again run into problems with the delegation of PCI HW (motherboard USB controller): With a 2.6.27.10 Dom0 kernel from factory the patched 2.6.27.10 DomU starts up ok and has access to the PCI hardware, albeit with slightly different behaviour: Where only IRQ23 was visibly used in DomU before, IRQ18 now is visibly used, too (although it is shared with the Dom0 eth2 device). When the DomU is shutdown, cleanup of the shared interrupt does not seem to work - the DomU loops in the final phase of the shutdown and the Dom0 produces a stack trace. This behaviour has only been observed on the second "power-cycle" for the DomU (at least up to now). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=473217 Jan Beulich <jbeulich@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P3 - Medium Status|NEW |ASSIGNED AssignedTo|cgriffin@novell.com |jbeulich@novell.com -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=473217 User jbeulich@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=473217#c1 Jan Beulich <jbeulich@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |wolf@womaro.ch --- Comment #1 from Jan Beulich <jbeulich@novell.com> 2009-02-18 04:40:50 MST --- Could you try whether disabling PCI-MSI in the DomU helps (pci=nomsi)? Albeit seemingly unrelated (as you don't use MSI at all), it seems there is some MSI code getting executed during shutdown even for devices not actually using MSI. If it does make a difference, does it also address the Dom0 issue (where apparently IRQ 23 no longer gets delivered)? In either case (i.e. also without the workaround above applied) it'd be helpful to know whether no interrupts on IRQ 23 appear anymore, or whether they just come through very sporadically. Also, once the guest got killed, does Dom0 behavior turn back to normal? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=473217 User wolf@womaro.ch added comment https://bugzilla.novell.com/show_bug.cgi?id=473217#c2 --- Comment #2 from Wolf Geldmacher <wolf@womaro.ch> 2009-03-02 11:09:21 MST --- Sorry for not reacting - I didn't get any notice from bugzilla.. (?) I've just migrated to the brand new 2.6.27.19 kernels in both Dom0 and DomU and I'm happy to report that everthing seems to work now without a hitch! I can restart the domain, attach/detach USB hardware while running, I see the interrupts, IRQ sharing seems to work, too. For reference here is the /proc/interrupts from Dom0: CPU0 CPU1 1: 12709 0 Phys-level i8042 8: 0 0 Phys-level rtc0 9: 0 0 Phys-level acpi 12: 2059 0 Phys-level i8042 16: 269237 0 Phys-level pata_jmicron 18: 108412152 267107 Phys-level eth2 19: 5048870 16324 Phys-level ahci, ata_piix, ata_piix 20: 2368670 24503 Phys-level eth1 22: 192 0 Phys-level HDA Intel 23: 196 0 Phys-level 1276: 8248099 1847675 Phys-level eth0 1280: 37462957 2698386 Dynamic-level timer 1281: 177959 109593 Dynamic-level resched 1282: 0 0 Dynamic-level callfunc 1283: 2062 3101 Dynamic-level call1func 1284: 9144 7538 Dynamic-level spinlock 1285: 3061 1326 Dynamic-level xenbus 1286: 0 0 Dynamic-level console 1287: 0 0 Dynamic-level suspend 1288: 417 0 Dynamic-level pciback 1289: 204917 44422 Dynamic-level blkif-backend 1290: 112249 53157 Dynamic-level blkif-backend 1291: 4163173 4063432 Dynamic-level vif3.0 1292: 445951 330498 Dynamic-level blkif-backend 1293: 768974 227531 Dynamic-level vif2.0 1294: 116993 19219 Dynamic-level vif2.1 NMI: 0 0 Non-maskable interrupts RES: 177959 109594 Rescheduling interrupts CAL: 2062 3101 function call interrupts and from the DomU: CPU0 16: 0 Phys-irq-level uhci_hcd:usb1 18: 108518274 Phys-irq-level uhci_hcd:usb5, ehci_hcd:usb6 19: 0 Phys-irq-level uhci_hcd:usb4 21: 0 Phys-irq-level uhci_hcd:usb2 23: 24371800 Phys-irq-level uhci_hcd:usb3, ehci_hcd:usb7 256: 66495050 Dynamic-irq-level timer0 257: 0 Dynamic-irq-level resched0 258: 0 Dynamic-irq-level callfunc0 259: 459 Dynamic-irq-level xenbus 260: 4100 Dynamic-irq-level xencons 261: 513191 Dynamic-irq-level blkif 262: 1002777 Dynamic-irq-level blkif 263: 29691447 Dynamic-irq-level eth0 NMI: 0 IRQ 18 is shared and works ok, IRQ 23 is unique to the DomU and works as well. Thanks again for your support! Cheers, Wolf -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=473217 User jbeulich@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=473217#c3 --- Comment #3 from Jan Beulich <jbeulich@novell.com> 2009-03-04 01:30:03 MST --- I'd be happy to close this, but - did this survive a shutdown of the guest? Also, you state that IRQ 23 is unique to DomU, which the output you provided contradicts to. Did, after a shutdown of the guest, devices on IRQs 18 and 23 indeed continue to work? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠novell.com