Bug 504711
| Summary: | "vcpu not ready for apic_round_robin" loop on guest panic | ||||||
|---|---|---|---|---|---|---|---|
| Product: | Red Hat Enterprise Linux 5 | Reporter: | david ahern <daahern> | ||||
| Component: | kvm | Assignee: | john cooper <john.cooper> | ||||
| Status: | CLOSED NOTABUG | QA Contact: | Lawrence Lim <llim> | ||||
| Severity: | high | Docs Contact: | |||||
| Priority: | low | ||||||
| Version: | 5.4 | CC: | bburns, ehabkost, nobody, sghosh, tburke, tools-bugs, virt-maint | ||||
| Target Milestone: | rc | ||||||
| Target Release: | 5.4 | ||||||
| Hardware: | x86_64 | ||||||
| OS: | Linux | ||||||
| Whiteboard: | |||||||
| Fixed In Version: | Doc Type: | Bug Fix | |||||
| Doc Text: | Story Points: | --- | |||||
| Clone Of: | Environment: | ||||||
| Last Closed: | 2009-07-14 00:35:16 UTC | Type: | --- | ||||
| Regression: | --- | Mount Type: | --- | ||||
| Documentation: | --- | CRM: | |||||
| Verified Versions: | Category: | --- | |||||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||
| Cloudforms Team: | --- | Target Upstream Version: | |||||
| Embargoed: | |||||||
| Attachments: |
|
||||||
Able to reproduce under a recent 5.4 kernel tree with the following: qemu/x86_64-softmmu/qemu-system-x86_64 -cdrom /workspace/bz504711/kvm-bug.iso -boot d Unable to handle kernel NULL pointer dereference at 0000000000000028 RIP: [<ffffffff887bf886>] :kvm:kvm_get_intr_delivery_bitmask+0x4e/0x86 PGD 4dd99067 PUD 4dc9d067 PMD 0 Oops: 0000 [1] SMP last sysfs file: /class/misc/kvm/dev CPU 3 Modules linked in: kvm_intel(U) kvm(U) i915(U) drm(U) ipt_MASQUERADE(U) iptable) Pid: 7250, comm: qemu-system-x86 Tainted: G 2.6.18-prepgupf #7 RIP: 0010:[<ffffffff887bf886>] [<ffffffff887bf886>] :kvm:kvm_get_intr_delivery6 RSP: 0018:ffff810052f4dbb8 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff810052f4dbe8 RCX: ffffffff80307c28 RDX: ffffffff80307c28 RSI: 0000000000001000 RDI: ffffffff80307c20 RBP: ffff810052f4dbd8 R08: ffffffff80307c28 R09: 0000000000000046 R10: 0000000000000000 R11: 0000000000000080 R12: ffff81006cceda00 R13: ffff81006cceda20 R14: 0000000000000001 R15: 0000000000001000 FS: 0000000041af6940(0063) GS:ffff81007d63e640(0000) knlGS:0000000000000000 CS: 0010 DS: 002b ES: 002b CR0: 000000008005003b CR2: 0000000000000028 CR3: 000000004fd12000 CR4: 00000000000026e0 Process qemu-system-x86 (pid: 7250, threadinfo ffff810052f4c000, task ffff81004) Stack: ffff810064935e80 ffff8100508a8000 ffff81006cceda00 ffffffff887be06b 0000000000000001 ffff810000000001 0100000000000939 0000000000000001 ffff810064935e80 ffff8100508a8000 00000000ffffffff 0000000000000001 Call Trace: [<ffffffff887be06b>] :kvm:ioapic_service+0x55/0x12a [<ffffffff887bfa6e>] :kvm:kvm_set_irq+0x65/0xa3 [<ffffffff887b0c48>] :kvm:kvm_arch_vm_ioctl+0x37e/0x62e [<ffffffff8008cbf1>] sched_clock_cpu+0x116/0x124 [<ffffffff8008cd1b>] update_rq_clock+0x17/0x20 [<ffffffff80063489>] __sched_text_start+0xf9/0xbef [<ffffffff887aa643>] :kvm:kvm_vm_ioctl+0xa87/0xade [<ffffffff887a84ed>] :kvm:kvm_io_bus_find_dev+0x35/0x50 [<ffffffff8008c4eb>] preempt_notifier_unregister+0x92/0xac [<ffffffff887a9ac4>] :kvm:vcpu_put+0x16/0x1f [<ffffffff887af789>] :kvm:kvm_arch_vcpu_ioctl_run+0x5fd/0x60b [<ffffffff887ab3d0>] :kvm:kvm_vcpu_ioctl+0x435/0x448 [<ffffffff8008cf62>] default_wake_function+0x0/0xe [<ffffffff800424cf>] do_ioctl+0x21/0x6b [<ffffffff800306c7>] vfs_ioctl+0x457/0x4b9 [<ffffffff800b7498>] audit_syscall_entry+0x16e/0x1a1 [<ffffffff8004cae5>] sys_ioctl+0x59/0x78 [<ffffffff8005e28d>] tracesys+0xd5/0xe0 Code: 8b 40 28 0f ab 45 00 eb 2a e8 5e 2e 8d f7 85 c0 74 19 40 b6 RIP [<ffffffff887bf886>] :kvm:kvm_get_intr_delivery_bitmask+0x4e/0x86 RSP <ffff810052f4dbb8> CR2: 0000000000000028 <0>Kernel panic - not syncing: Fatal exception Looks like kernel/x86/irq_comm.c:kvm_get_intr_delivery_bitmask() isn't laundering the vcpu * return from kvm_get_lowest_prio_vcpu() and dereferences a NULL pointer. It seems to be a duplicate of bug #504018. Could you test with a newer KVM package? (kvm-83-65.el5 or later). The host side panic did not happen with kvm-83-82, but the host is spewing the following (from dmesg): vcpu not ready for apic_round_robin (In reply to comment #3) > The host side panic did not happen with kvm-83-82, but the host is spewing the > following (from dmesg): > > vcpu not ready for apic_round_robin Thanks, I'm updating the summary accordingly. It looks like a less serious issue, but allowing a guest to flood dmesg sounds undesirable too. Correction: the host does not panic in this scenario built from a tree after commit fb6eef649d07cc27680ee073fe5d9b5fbf4093c1. I had reproduced the above failure inadvertently with an older tree. I didn't find the "vcpu not ready for apic_round_robin" diagnostic generated either but that by itself isn't conclusive w/r/t comment #3 above. Except for this nit I think we can close the case. Fixed in current tree by fb6eef649d07cc27680ee073fe5d9b5fbf4093c1. |
Created attachment 346942 [details] iso which causes panic Description of problem: The guest is started and running a rhel5.2 kernel with a homegrown initrd. The init script for the initrd is a bash script with a typo in it and caused the guest to halt with a guest side panic. That in turn caused the host to panic. kdump is setup in the host, but it is running off a USB key which does not have enough space for a vmcore. Version-Release number of selected component (if applicable): kvm-83-59 How reproducible: Very. See uploaded iso. Steps to Reproduce: qemu-kvm -m 2048 -smp 2 -vnc :2 -cdrom /tmp/kvm-bug.iso -boot d Actual results: guest panics, host panics. Expected results: guest to panic (my typo); host should not.