Description of problem: kernel reports a BUG memory.c on the target system upon "xen migrate --live" of a hvm guest between two identical machines. The guest OS is running from an iscsi-target (exported from third source) Version-Release number of selected component (if applicable): 2931.fc7, 2934.fc7, 2936.fc7 How reproducible: frequently Steps to Reproduce: 1. use to iscsi initiator to logon to the lun on both hosts 2. start hvm guest (wait till finished booting) 3. initiate a live migration (xm migrate --live guest target) Actual results: migration starts and prior/upon last part of completion the bug occurs sending the guest os to nirvana Expected results: no BUG causing crash of guest ;-) Additional info: ------------[ cut here ]------------ kernel BUG at mm/memory.c:2292! invalid opcode: 0000 [#1] SMP last sysfs file: /class/misc/evtchn/dev Modules linked in: tun crc32c libcrc32c nfs lockd nfs_acl netbk xenblktap blkbk ipt_MASQUERADE iptable_nat nf_nat nf_conntrack_ipv4 xt_state nf_conntrack nfnetlink ipt_REJECT xt_tcpudp iptable_filter ip_tables x_tables bridge autofs4 nls_utf8 cifs sunrpc ipv6 ib_iser rdma_cm ib_cm iw_cm ib_sa ib_mad ib_core ib_addr iscsi_tcp libiscsi scsi_transport_iscsi dm_multipath video sbs i2c_ec dock button battery asus_acpi backlight ac snd_hda_intel snd_hda_codec snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss snd_mixer_oss snd_pcm e1000 snd_timer i2c_i801 snd soundcore snd_page_alloc i2c_core serial_core serio_raw sr_mod cdrom pcspkr sg dm_snapshot dm_zero dm_mirror dm_mod pata_marvell ata_generic ata_piix libata sd_mod scsi_mod ext3 jbd mbcache ehci_hcd ohci_hcd uhci_hcd CPU: 0 EIP: 0061:[<c105c740>] Not tainted VLI EFLAGS: 00010202 (2.6.20-2936.fc7xen #1) EIP is at __handle_mm_fault+0x29c/0x1169 eax: c12e48b8 ebx: 00000000 ecx: 00000000 edx: 00000178 esi: eb6b4cd4 edi: 00000000 ebp: eb06af10 esp: eb06ae0c ds: 007b es: 007b ss: 0069 Process qemu-dm (pid: 3718, ti=eb06a000 task=ebb174f0 task.ti=eb06a000) Stack: 00000000 00697d60 c12cbe0c c12cbd80 eb191000 00000001 fffffff2 eb06ae30 c10529e5 b522f000 eb6b4cd4 c0795a80 00000000 00000000 00000000 00000000 00000000 00000000 c2063000 c1697d48 eb190178 c0795a80 eb191000 00000001 Call Trace: [<c1005e65>] show_trace_log_lvl+0x1a/0x2f [<c1005f15>] show_stack_log_lvl+0x9b/0xa3 [<c10060b1>] show_registers+0x194/0x26a [<c10062c1>] die+0x13a/0x24f [<c11fb943>] do_trap+0x79/0x91 [<c100686a>] do_invalid_op+0x97/0xa1 [<c11fb82d>] error_code+0x35/0x3c [<c11fd168>] do_page_fault+0x70a/0xbe0 [<c11fb82d>] error_code+0x35/0x3c ======================= Code: 07 00 00 8b b5 24 ff ff ff 8b 46 44 85 c0 0f 84 9a 04 00 00 83 78 08 00 0f 84 54 04 00 00 c7 45 f0 02 00 00 00 f6 46 19 04 74 04 <0f> 0b eb fe 8b bd 24 ff ff ff 8b 47 4c c7 85 6c ff ff ff 00 00 EIP: [<c105c740>] __handle_mm_fault+0x29c/0x1169 SS:ESP 0069:eb06ae0c
It seems to be the same BUG_ON() of bug #254208.
Could you test the package at: http://koji.fedoraproject.org/koji/taskinfo?taskID=186943 It includes a fix for the Oops you are getting. It is possible that the problem move to user-space, as the expected behaviour instead of the Oops is the process getting a SIGBUS, and I don't know if qemu-dm is really expecting a SIGBUS.
The BUG doesn't occur anymore. But while the migration is now successful most of the time, it occasionally leaves a dead (no status, all dashes) guest behind without any visible error in any log. However, I guess this is a problem with xen, not the kernel.
I am marking as MODIFIED as the fix is on CVS and will go to the next F-7 update. This bug can be cloned to the 'xen' component for the dead domain problem. When the dead guest appears, is the migration successful, or it is aborted?
The dead guest is on the target, so the migration itself is successful. Thanks for the fix!
*** Bug 245314 has been marked as a duplicate of this bug. ***
kernel-xen-2.6-2.6.20-2943.fc7 has been pushed to the Fedora 7 testing repository. If problems still persist, please make note of it in this bug report. If you want to test the update, you can install it with su -c 'yum --enablerepo=updates-testing update kernel-xen-2.6'
kernel-xen-2.6-2.6.20-2943.fc7 has been pushed to the Fedora 7 stable repository. If problems still persist, please make note of it in this bug report.