Bug 68051
| Summary: | Kernel stack trace when attempt to boot into 3.13.0-0.rc6.git0.1 results in "kernel BUG at mm/page_alloc.c:2788" | ||
|---|---|---|---|
| Product: | Virtualization | Reporter: | Kashyap Chamarthy (kashyap.cv) |
| Component: | kvm | Assignee: | virtualization_kvm |
| Status: | RESOLVED CODE_FIX | ||
| Severity: | normal | CC: | mtosatti, xerofoify |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 3.13.0-0.rc6.git0.1.fc21.x86_64+debug | Subsystem: | |
| Regression: | No | Bisected commit-id: | |
| Attachments: |
Complete stdout of dmesg
Complete stdout of dmidecode |
||
|
Description
Kashyap Chamarthy
2014-01-02 14:35:23 UTC
Created attachment 120651 [details]
Complete stdout of dmesg
Created attachment 120661 [details]
Complete stdout of dmidecode
I should add some more context: I noticed this issue when I found an un-killable defunct qemu process (I ended up with it when I tried to force power-off the guest via `virsh`), and I had to reboot the host.
$ virsh destroy ostack-compute
error: Failed to destroy domain 3
error: Failed to terminate process 3152 with SIGKILL: Device or resource busy
NOTE - The disk image is *not* on NFS mount. It's on local disk:
$ virsh domblklist ostack-compute
Target Source
------------------------------------------------
vda /home/kashyap/vmimages/ostack-compute.qcow2
Try to destroy the guest again, with LIBVIRT_DEBUG enabled, that's what I see in a loop:
=============
.
.
.
2014-01-02 12:42:54.012+0000: 27462: debug : virKeepAliveCheckMessage:395 : Got keepalive request from client 0x7f4dfaca75d0
2014-01-02 12:42:54.012+0000: 27462: debug : virNetMessageNew:44 : msg=0x7f4dfaca7770 tracked=0
2014-01-02 12:42:54.012+0000: 27462: debug : virNetMessageEncodePayloadEmpty:479 : Encode length as 28
2014-01-02 12:42:54.012+0000: 27462: debug : virKeepAliveMessage:101 : Sending keepalive response to client 0x7f4dfaca75d0
2014-01-02 12:42:54.012+0000: 27462: debug : virKeepAliveMessage:104 : RPC_KEEPALIVE_SEND: ka=0x7f4dfaca78c0 client=0x7f4dfaca75d0 prog=1801807216 vers=1 proc=2
2014-01-02 12:42:54.012+0000: 27462: debug : virNetClientQueueNonBlocking:1926 : RPC_CLIENT_MSG_TX_QUEUE: client=0x7f4dfaca75d0 len=28 prog=1801807216 vers=1 proc=2 type=2 status=0 serial=0
2014-01-02 12:42:54.012+0000: 27462: debug : virNetClientCallNew:1905 : New call 0x7f4dfaca7ae0: msg=0x7f4dfaca7770, expectReply=0, nonBlock=1
2014-01-02 12:42:54.012+0000: 27462: debug : virNetMessageClear:55 : msg=0x7f4dfaca7638 nfds=0
2014-01-02 12:42:54.012+0000: 27462: debug : virNetClientIOEventLoopRemoveDone:1379 : Removing completed call 0x7f4dfaca7ae0
2014-01-02 12:42:59.172+0000: 27462: debug : virNetMessageDecodeLength:149 : Got length, now need 28 total (24 more)
2014-01-02 12:42:59.172+0000: 27462: debug : virNetClientCallDispatch:1123 : RPC_CLIENT_MSG_RX: client=0x7f4dfaca75d0 len=28 prog=1801807216 vers=1 proc=1 type=2 status=0 serial=0
2014-01-02 12:42:59.172+0000: 27462: debug : virKeepAliveCheckMessage:374 : ka=0x7f4dfaca78c0, client=0x7f4dfaca75d0, msg=0x7f4dfaca7638
2014-01-02 12:42:59.172+0000: 27462: debug : virKeepAliveCheckMessage:391 : RPC_KEEPALIVE_RECEIVED: ka=0x7f4dfaca78c0 client=0x7f4dfaca75d0 prog=1801807216 vers=1 proc=1
.
.
.
=============
And:
$ ps -ef | grep qemu
qemu 3152 1 30 2013 ? 13:30:42 [qemu-system-x86] <defunct>
$ pstree 3152
qemu-system-x86───{qemu-system-x86}
Libvirt & QEMU version on the host:
$ rpm -q libvirt-daemon-kvm qemu-system-x86
libvirt-daemon-kvm-1.2.0-1.fc21.x86_64
qemu-system-x86-1.7.0-3.fc21.x86_64
Seems fixed after reading the comments here. Can you close this as the bug seems to be close still in June of 2014. Thanks Nick Thanks Marcelo, Xerofoify. I missed to update this. Yes, this can be closed. With current Fedora Rawhide kernels (3.15.0-0.rc7.git4.2.fc21.x86_64) works in my current nested environment. For reference, here's the associated Fedora Kernel bug (that's closed) -- https://bugzilla.redhat.com/show_bug.cgi?id=1047892 |