Bug#488284: Same failure with the Sid hypervisor
maximilian attems <max@stro.at> writes:
> On Tue, Jul 01, 2008 at 04:06:03PM +0200, Ferenc Wagner wrote:
>
>> I upgraded the hypervisor to the Sid version (3.2.1-2) but that fails
>> all the same. :(
>
> sorry i'm lost. what are you trying to do?
> you know that the sid linux-2.6 linux images only work as guest and
> not as dom0 ?
Yes, I know that. The dom0 is an Etch system, with a stock Etch kernel:
ii linux-image-2.6.18-6-xen-686 2.6.18.dfsg.1-18etch6 Linux 2.6.18 image on i686
I tried the stock Etch hypervisor
ii xen-hypervisor-3.0.3-1-i386-pae 3.0.3-0-4 The Xen Hypervisor on i386 with pae
then the backports.org hypervisor (xen-hypervisor-3.2-1-i386, version
3.2.0-3~bpo4+2) and finally the Sid hypervisor (xen-hypervisor-3.2-1-i386,
version 3.2.1-2). Each hypervisor booted the same dom0 kernel, the
stock Etch one. And each hypervisor could run a given guest with the
stock Etch kernel, but failed to run the same guest with the current
Sid kernel (linux-image-2.6.25-2-xen-686, version 2.6.25-5). I also
tried some earlier Sid kernels (*-1-xen-686) with basically the same
result: the guest system (Lenny) starts up, but crashes before
finishing the boot process. To be more precise: with the 3.0.3
hypervisor the guest does not crash, but reports:
Starting web server: apache2[ 18.873026] 1 multicall(s) failed: cpu 0
[ 18.873072] call 1/1: op=14 arg=[b7f57000] result=-22
[ 18.873108] ------------[ cut here ]------------
[ 18.873119] kernel BUG at arch/x86/xen/multicalls.c:103!
[ 18.873128] invalid opcode: 0000 [#1] SMP
[ 18.873138] Modules linked in: xt_multiport xt_limit ip6t_LOG ipt_LOG ip6t_REJECT ipt_REJECT xt_tcpudp nf_conntrack_ipv6 nf_conntrack_ipv4 xt_state nf_conntrack ip6table_filter ip6_tables iptable_filter ip_tables x_tables ipv6 evdev pcspkr ext3 jbd mbcache dm_mirror dm_snapshot dm_mod
[ 18.873191]
[ 18.873191] Pid: 1589, comm: jsvc Not tainted (2.6.25-2-xen-686 #1)
[ 18.873191] EIP: 0061:[<c010203d>] EFLAGS: 00010202 CPU: 0
[ 18.873191] EIP is at xen_mc_flush+0x131/0x13b
[ 18.873191] EAX: 00000000 EBX: c1402a60 ECX: 00000000 EDX: 00000000
[ 18.873191] ESI: 00000000 EDI: 00000001 EBP: c1402060 ESP: ded17f00
[ 18.873191] DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0069
[ 18.873191] Process jsvc (pid: 1589, ti=ded16000 task=df4d7290 task.ti=ded16000)
[ 18.873191] Stack: 00000200 00000000 00000000 00000000 b7f58000 c0164e91 c1000000 dece6188
[ 18.873191] decf13d8 c0103661 c1324360 b7f58000 df1a9b1c 196630e0 00000000 df548080
[ 18.873191] 00000070 00000001 00000000 df4aa010 df548080 decf9ab8 c13d9f2c df488df8
[ 18.873191] Call Trace:
[ 18.873191] [<c0164e91>] mprotect_fixup+0x461/0x5b9
[ 18.873191] [<c0103661>] enable_dynirq+0x2d/0xd2
[ 18.873191] [<c0165163>] sys_mprotect+0x17a/0x1df
[ 18.873191] [<c0106862>] syscall_call+0x7/0xb
[ 18.873191] =======================
[ 18.873191] Code: 5a 59 31 f6 8d 9d 00 0a 00 00 eb 09 8b 43 04 46 ff 13 83 c3 08 3b b5 08 0b 00 00 72 ef 85 ff c7 85 08 0b 00 00 00 00 00 00 74 04 <0f> 0b eb fe 58 5b 5e 5f 5d c3 57 89 d7 56 89 c6 53 b8 60 c0 34
[ 18.873191] EIP: [<c010203d>] xen_mc_flush+0x131/0x13b SS:ESP 0069:ded17f00
[ 18.873218] ---[ end trace da09e02a2f03a22d ]---
Then I can ssh into it, but freezes if I try to halt it.
Interestingly, I've got another Lenny domU running a 2.6.25-1-xen-686
kernel, which doesn't exhibit this problem. I even tried installing
Apache on it, still no failure... It's still much less loaded, though
(no tomcat, for example). But I can experiment with it, like giving
it the same amount of memory, installing similar packages, and so on,
if you see fit.
A final note: when I tested the Sid hypervisor, I did not upgrade the
user-space Xen tools, rather stayed with those from backports.org.
They seemingly weren't upset at all by the patchlevel difference.
--
Regards,
Feri.
Reply to: