[Bug 1003999] New: booting xen guest hangs
http://bugzilla.suse.com/show_bug.cgi?id=1003999 Bug ID: 1003999 Summary: booting xen guest hangs Classification: openSUSE Product: openSUSE Distribution Version: Leap 42.2 Hardware: Other OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: Xen Assignee: xen-bugs@suse.de Reporter: ro@suse.com QA Contact: qa-bugs@suse.de CC: adrian@suse.com Found By: --- Blocker: --- this is happening on the xen OBS build worker machines CPU model name : Dual-Core AMD Opteron(tm) Processor 8214 (no vme, that's actually why these boxes are running xen while everything else does kvm) booting a xen vm goes up to this point: [ 2.421079] emc: device handler registered [ 2.421462] rdac: device handler registered [ 2.435194] device-mapper: uevent: version 1.0.3 [ 2.435392] device-mapper: ioctl: 4.34.0-ioctl (2015-10-28) initialised: dm-devel@redhat.com [ 2.455299] device-mapper: multipath: version 1.11.0 loaded [ 3.070552] loop: unknown parameter 'lbs' ignored [ 3.220213] loop: module loaded [ 3.316259] fuse init (API version 7.23) [ 3.373081] squashfs: version 4.0 (2009/01/31) Phillip Lougher [ 3.524229] nf_conntrack version 0.5.0 (16384 buckets, 65536 max) [ 117.136041] random: nonblocking pool is initialized --------------------- and there it hangs. The same setup works on leap42.1 (kernel 4.1.21-14-xen, xen-4.5.2_06) on leap42.2 (kernel 4.4.22-1-default, xen-4.7.0_12) I get the stuck VM This happens with every single guest, regardless if 32 or 64bit. config file: kernel = "/boot/vmlinuz" ramdisk = "/boot/initrd" memory = 1500 vcpus = 4 root = "/dev/hda1 ro" extra = "init=/bin/bash console=ttyS0 panic=1 udev_timeout=10" on_poweroff = "destroy" on_reboot = "destroy" on_crash = "destroy" xl call: xl create -c /var/tmp/build.xen.conf-ShCtQtZiH name="\"build_Z\"" "disk=[ \"phy:/dev/dm-2,hda1,w\", \"phy:/dev/dm-3,hda2,w\" ]" "extra=\"panic=1 init=/.build/build rd.driver.pre=binfmt_misc elevator=noop console=ttyS0 udev_timeout=10 rd.debug rdudevdebug rd.shell\"" -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 Ludwig Nussel <lnussel@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P2 - High -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 Martin Kravec <mkravec@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |mkravec@suse.com -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 http://bugzilla.suse.com/show_bug.cgi?id=1003999#c1 Olaf Hering <ohering@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |ro@suse.com Flags| |needinfo?(ro@suse.com) --- Comment #1 from Olaf Hering <ohering@suse.com> --- Please add 'initcall_debug debug' to the VM kernel cmdline to see how far it gets. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 http://bugzilla.suse.com/show_bug.cgi?id=1003999#c2 --- Comment #2 from Ruediger Oertel <ro@suse.com> --- okay, I'll try this next. some data: - guest tumbleweed hangs - guest 13.2 runs (kernel-obs-build from openSUSE:13.2:Update/i586) -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 http://bugzilla.suse.com/show_bug.cgi?id=1003999#c3 Ruediger Oertel <ro@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Flags|needinfo?(ro@suse.com) | --- Comment #3 from Ruediger Oertel <ro@suse.com> --- Created attachment 697172 --> http://bugzilla.suse.com/attachment.cgi?id=697172&action=edit logfile booting xen VM, guest kernel 4.4.22-1-default -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 http://bugzilla.suse.com/show_bug.cgi?id=1003999#c5 --- Comment #5 from Olaf Hering <ohering@suse.com> --- Is this a Xen VM image? Looks like virtion drivers get loaded, and not pv frontend drivers. [ 4.023210] xenbus_probe_frontend: Device with no driver: device/vbd/769 Perhaps the image creation misses kernel-xen and does a fallback to KVM. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 http://bugzilla.suse.com/show_bug.cgi?id=1003999#c7 --- Comment #7 from Olaf Hering <ohering@suse.com> --- I think forcing the module 'xen:vbd' into every initrd will solve this for xenlinux and pvops. -- You are receiving this mail because: You are on the CC list for the bug.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 http://bugzilla.suse.com/show_bug.cgi?id=1003999#c8 --- Comment #8 from Olaf Hering <ohering@suse.com> --- Beside the missing frontend drivers, how did that /dev/dm-N work for KVM guests? Was that just a special raw filesystem/swap for Xen guests? After changes made for bug #979002 a guest can now convert a domU.cfg 'hda' entry into a symlink to xvda. This is done in xen-tools-domU. However, this case is not covered because the rule considers only disk names: /usr/lib/udev/rules.d/60-persistent-xvd.rules KERNEL=="xvd*[!0-9]", ENV{VBD_HD_SYMLINK}=="hd[a-d]", SYMLINK+="$env{VBD_HD_SYMLINK}" As we discovered, hda2 used to work because the xenlinux based blkfront driver tried to claim the name from the config file (hda2) in the guest and provide a device node 'hda2' with major:minor 3:2. This does not work anymore with pvops, it always uses xvd* als namespace. However, a udevrule can be used to create a compat symlink. Looks like the rule has to be extended to cover also hd?[0-9]* -- You are receiving this mail because: You are on the CC list for the bug.
how did that /dev/dm-N work for KVM guests? any block device works, this is just a device mapper device. for qemu/kvm we are using virtio for basically everything, even console.
http://bugzilla.suse.com/show_bug.cgi?id=1003999 http://bugzilla.suse.com/show_bug.cgi?id=1003999#c9 Ruediger Oertel <ro@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution|--- |WORKSFORME --- Comment #9 from Ruediger Oertel <ro@suse.com> --- the classic worker setup had logical volumes for both kvm and xen. (current workers just use files in a tmpfs instead, but you need enough ram to do this ...) anyway, fixed github.com:openSUSE/obs-build.git commit d65f0f936ffc4632f49ca2906cbf554f0474938b has the changes. (plus the missing xenblk (xen-blkfront) in the initrd used, which was the first problem encountered in the bug.) -- You are receiving this mail because: You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠novell.com