Guests running OpenIndiana (and relatives) fail to boot on AMD hardware
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
QEMU |
Expired
|
Undecided
|
Unassigned |
Bug Description
First observed with OpenSolaris 2009.06, and also applies to the latest OpenIndiana release.
Version: qemu-kvm 1.1.1
Hardware:
2 x AMD Opteron 6128 8-core processors, 64GB RAM.
These guests boot on equivalent Intel hardware.
To reproduce:
qemu-kvm -nodefaults -m 512 -cpu host -vga cirrus -usbdevice tablet -vnc :99 -monitor stdio -hda drive.img -cdrom oi-dev-
I've tested with "-vga std" and various different emulated CPU types, to no effect.
What happens:
GRUB loads, and offers multiple boot options, but none work. Some kind of kernel panic flies by very fast before restarting the VM, and careful use of the screenshot button reveals that it reads as follows:
panic[cpu0]
#df Double fault
pid=0, pc=0xault
pid=0, pc=0xfe800377, sp=0xfec40090, eflags=0x202
cr0: 80050011<
cr2: 0cr3: ae2f000
gs: 1b0 fs: 0 es: 160 ds: 160
edi: 0 esi: 0 ebp: 0 esp: fec2b4c4
ebx: c0010015 edx: 0 ecx: 0 eax: fec40400
trp: 8 err: 0 eip: fe800377 cs: 158
efl: 202 usp: fec40090 ss: 160
tss.tss_link: 0x0
tss.tss_esp0: 0x0
tss.tss_ss0: 0x160
tss.tss_esp1: 0x0
tss.tss_ss1: 0x0
tss.tss esp2: 0x0
tss.tss_ss2: 0x0
tss.tss_cr3: 0xae2f000
tss.tss_eip: 0xfec40400
tss.tss_eflags: 0x202
tss.tss_eax: 0xfec40400
tss.tss_ebx: 0xc0010015
tss.tss_ecx: 0xc0010000
tss.tss_edx: 0x0
tss.tss_esp: 0xfec40090
Warning - stack not written to the dumpbuf
fec2b3c8 unix:due+e4 (8, fec2b48c, 0, 0)
fec2b478 unix:trap+12fa (fec2b48c, 0, 0)
fec2b48c unix:_cmntrap+7c (1b0, 0, 160, 160, 0)
If there's any more, I haven't managed to catch it.
Solaris 11 does not seem to suffer from the same issue, although the first message that appears at boot (after the version info) is "trap: Unkown trap type 8 in user mode". Could be related?
As always, thanks in advance and please let me know if I can help to test, or provide any more information.
Triaging old bug tickets ... can you still reproduce this issue with the latest version of QEMU (currently v2.9)?