Package: linux-image-2.6.18-2-xen-vserver-686 Version: 2.6.18-5 I use Debians 2.6.18-5 kernel sources to build my own xen-vserver kernel image for my Sarge XEN/vserver system. The problems looks like the following: DomUs with pass-through PCI devices seem to have problems with interrupt handling like the following dmesg excerpt shows (from domU named 'trustedvserver'): irq 16: nobody cared (try booting with the "irqpoll" option) [<c014f4aa>] __report_bad_irq+0x2a/0x90 [<c014f5e7>] note_interrupt+0xb7/0xe0 [<c014ebc8>] __do_IRQ+0x138/0x150 [<c0106ce8>] do_IRQ+0x48/0xa0 [<c0257ee3>] evtchn_do_upcall+0x93/0x110 [<c0105285>] hypervisor_callback+0x3d/0x48 [<c01099f7>] raw_safe_halt+0x27/0x60 [<c0102ad3>] xen_idle+0x23/0x40 [<c0102b73>] cpu_idle+0x83/0xd0 [<c0406a0f>] start_kernel+0x1cf/0x230 [<c0406300>] unknown_bootoption+0x0/0x1e0 handlers: [<c02aead0>] (usb_hcd_irq+0x0/0x70) Disabling IRQ #16 (This is the onboard USB controller with an usb sound apdater attached to it) The bug also seems to kill the kernel NFS server (it doesn't respond after such a kernel oops and the domU has to be restartet). Or this dmesg excerpt is from another domU called 'router' with 2 NICs (a 3Com 10/100 (ethdsl) and an Intel e1000 (eth0)): NETDEV WATCHDOG: eth0: transmit timed out e1000: eth0: e1000_watchdog: NIC Link is Up 1000 Mbps Full Duplex NETDEV WATCHDOG: ethdsl: transmit timed out ethdsl: transmit timed out, tx_status 00 status e601. diagnostics: net 0ccc media 8880 dma 0000003a fifo 0000 ethdsl: Interrupt posted but not delivered -- IRQ blocked by another device? Flags; bus-master 1, dirty 16894(14) current 16894(14) Transmit list 00000000 vs. c43d7ac0. 0: @c43d7200 length 80000024 status 00010024 1: @c43d72a0 length 80000024 status 00010024 2: @c43d7340 length 80000024 status 00010024 3: @c43d73e0 length 80000024 status 00010024 4: @c43d7480 length 80000024 status 00010024 5: @c43d7520 length 80000024 status 00010024 6: @c43d75c0 length 80000024 status 00010024 7: @c43d7660 length 80000024 status 00010024 8: @c43d7700 length 80000024 status 00010024 9: @c43d77a0 length 80000020 status 00010020 10: @c43d7840 length 80000020 status 00010020 11: @c43d78e0 length 80000020 status 00010020 12: @c43d7980 length 80000020 status 80010020 13: @c43d7a20 length 80000020 status 80010020 14: @c43d7ac0 length 80000034 status 00010034 15: @c43d7b60 length 80000024 status 00010024 ethdsl: Resetting the Tx ring pointer. A third domU (called 'dmzvserver)' without any physical devices does not suffer from this problems. I'm using XEN 3.0.3 (no Debian packages, the stock XEN release tarball). When booting the domUs with my previous 2.6.16.27 kernel (for XEN 3.0.2) there is no problem although I have to use kernel 2.6.18.2 for dom0 (XEN 3.0.3 doesn't want to boot dom0 with my 2.6.16.27). I also suffered from this problem with the 2.6.18-3/2.6.18-4 sources with patch mentioned in #397281 applied. More details, more complete dmesg output and various other config files can be found here: http://markus.schuster.name/2.6.18.2-xen-vserver/ Regards, Markus Schuster
Attachment:
pgpGlaotUvVcH.pgp
Description: PGP signature