Description of problem: While booting the 2.6.18-76.el5 kernel the e1000e reports an error: 0000:00:1c.0: eth0: Unable to allocate MSI interrupt Error: -1 Version-Release number of selected component (if applicable): 2.6.18-76.el5 How reproducible: Always Several system in RHTS use this driver: ixen-01.rhts.boston.redhat.com ixen-02.rhts.boston.redhat.com Actual results: 01:00.0 Ethernet controller: Intel Corporation 82573E Gigabit Ethernet Controller (Copper) (rev 03) Checking module information e1000e: Intel(R) PRO/1000 Network Driver 0.2.0 Expected results: System should not report errors during normal boot operations Additional info: I have not seen any adverse affect as a result of the Error message being printed.
This bugzilla has Keywords: Regression. Since no regressions are allowed between releases, it is also being proposed as a blocker for this release. Please resolve ASAP.
This hardware is one of the ones that switched from e1000 to e1000e on this update. I tried to login to ixen-01.rhts.boston.redhat.com, but cannot and ixen-02.rhts.boston.redhat.com seems to be running an 2.6.24-based rt kernel (so still using e1000 for some reason). Do you know if these systems were not able to use MSI for handling interrupts when using e1000e? Can I reboot ixen-02 and try it out?
The message that shows up is simply to inform the user that MSI doesn't work on this system. The older e1000 driver didn't print a message when MSI didn't work. I've reserved ixen-01 from rhts and will test 5.1 and 5.2beta kernels to make sure they operate the same (both with INTx mode rather than one with MSI and one without).
Looks like everything is fine on the 2.6.18-53 and a 2.6.18-75-based kernel. They both have MSI enabled on the NIC (at least accoring to /proc/interrupts). There really isn't any difference in the code that initializes MSI between the e1000 and e1000e driver except for the extra error message if MSI fails in e1000e. I'm going to check and see if there is a difference between 53xen and 75xen just to be sure these kernels are behaving the same.
[root@ixen-01 ~]# uname -a Linux ixen-01.rhts.boston.redhat.com 2.6.18-53.el5xen #1 SMP Wed Oct 10 16:48:44 EDT 2007 x86_64 x86_64 x86_64 GNU/Linux [root@ixen-01 ~]# cat /proc/interrupts CPU0 CPU1 1: 8 0 Phys-irq i8042 6: 5 0 Phys-irq floppy 7: 0 0 Phys-irq parport0 8: 0 0 Phys-irq rtc 9: 0 0 Phys-irq acpi 12: 103 0 Phys-irq i8042 17: 0 0 Phys-irq uhci_hcd:usb3 18: 20048 0 Phys-irq uhci_hcd:usb2, libata 19: 1175 140 Phys-irq uhci_hcd:usb4, eth0 20: 0 0 Phys-irq uhci_hcd:usb1 21: 148 0 Phys-irq HDA Intel 256: 14217 0 Dynamic-irq timer0 257: 2807 0 Dynamic-irq resched0 258: 42 0 Dynamic-irq callfunc0 259: 0 3377 Dynamic-irq resched1 260: 0 94 Dynamic-irq callfunc1 261: 0 6783 Dynamic-irq timer1 262: 0 0 Dynamic-irq xenbus 263: 61 0 Dynamic-irq console NMI: 0 0 LOC: 0 0 ERR: 0 MIS: 0 # uname -a Linux ixen-01.rhts.boston.redhat.com 2.6.18-75.el5.gtest.37xen #1 SMP Tue Jan 29 23:13:12 EST 2008 x86_64 x86_64 x86_64 GNU/Linux [root@ixen-01 ~]# cat /proc/interrupts CPU0 CPU1 1: 8 0 Phys-irq i8042 6: 5 0 Phys-irq floppy 7: 0 0 Phys-irq parport0 8: 0 0 Phys-irq rtc 9: 0 0 Phys-irq acpi 12: 103 0 Phys-irq i8042 17: 0 0 Phys-irq uhci_hcd:usb3 18: 7315 0 Phys-irq uhci_hcd:usb2, libata 19: 134 141 Phys-irq uhci_hcd:usb4, eth0 20: 0 0 Phys-irq uhci_hcd:usb1 21: 148 0 Phys-irq HDA Intel 256: 6737 0 Dynamic-irq timer0 257: 1747 0 Dynamic-irq resched0 258: 44 0 Dynamic-irq callfunc0 259: 0 2375 Dynamic-irq resched1 260: 0 93 Dynamic-irq callfunc1 261: 0 4993 Dynamic-irq timer1 262: 0 0 Dynamic-irq xenbus 263: 23 0 Dynamic-irq console NMI: 0 0 LOC: 0 0 ERR: 0 MIS: 0 No real difference between the 2 here. As long as it can be confirmed that Xen kernels don't allow the traditional use of MSI interrupts (at least as far as the running kernel is concerned) then this should be closed as NOTABUG.
Created attachment 293662 [details] e1000e-irq-minimize.patch I posted the following upstream and when we get some confirmation from one of the 'xen folks' that not having MSI interrupts is expected feel free to transfer this back to me and I'll take it and post the patch.
Created attachment 293663 [details] e1000e-irq-minimize.patch I've got patch issues.... :-)
Anyone familiar with Xen and interrupts want to make a statement regarding comment #5?
My test kernels have been updated to include a patch for this bugzilla. http://people.redhat.com/agospoda/#rhel5 Please test them and report back your results.
in 2.6.18-79.el5 You can download this test kernel from http://people.redhat.com/dzickus/el5
confirmed fix is in the -90 .el5 kernel
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2008-0314.html