Talking about regressions...

Thanks for running those tests.
Does kvm complains ( dmesg ) ?

-----Original Message-----
From: kvm-devel-bounces-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org on behalf of Michael Riepe
Sent: Sun 14/01/2007 19:09
To: [email protected]
Subject: [kvm-devel] Talking about regressions...

When I did a couple of tests the other day, I noticed that one or more bugs must have been introduced at least into the vmx part in kvm-10:

- netbsd 3.1 used to work with kvm-9. In kvm-10, the system hangs during boot. One of the last things I can see is that a shell process dies with a SIGSEGV. Then the load rises to 100% (sys) on the host and stays there until I kill qemu (yes I've waited several minutes).

- opensolaris displays a "Time of Day clock error" during boot, claiming the clock has jumped by 0x5 (no unit given, I suppose that means seconds). Everything else seems to work.

- qemu-puppy-2.13-1 sometimes works and sometimes hangs somewhere inside the initial ramdisk. The exact place seems to differ, but usually it happens while the kernel modules are loaded. Again, host load rises to 100%. An older version (2.01-3) still works fine. kvm-9 works, too.

(Core Duo T2400, 32-bit SMP host, 32-bit guest, Linux 2.6.19 + kvm trunk revision 4290)

Any idea what to look for? The SIGSEGV on netbsd makes me suspect that there's a problem with the page tables, but I couldn't narrow down the cause yet.

--
Michael "Tired" Riepe <[email protected]>
X-Tired: Each morning I get up I die a little

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV>
_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

Thanks for running those tests.
Does kvm complains ( dmesg ) ?

-----Original Message-----
From: kvm-devel-bounces-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org on behalf of Michael Riepe
Sent: Sun 14/01/2007 19:09
To: [email protected]
Subject: [kvm-devel] Talking about regressions...

When I did a couple of tests the other day, I noticed that one or more bugs must have been introduced at least into the vmx part in kvm-10:

- netbsd 3.1 used to work with kvm-9. In kvm-10, the system hangs during boot. One of the last things I can see is that a shell process dies with a SIGSEGV. Then the load rises to 100% (sys) on the host and stays there until I kill qemu (yes I've waited several minutes).

- opensolaris displays a "Time of Day clock error" during boot, claiming the clock has jumped by 0x5 (no unit given, I suppose that means seconds). Everything else seems to work.

- qemu-puppy-2.13-1 sometimes works and sometimes hangs somewhere inside the initial ramdisk. The exact place seems to differ, but usually it happens while the kernel modules are loaded. Again, host load rises to 100%. An older version (2.01-3) still works fine. kvm-9 works, too.

(Core Duo T2400, 32-bit SMP host, 32-bit guest, Linux 2.6.19 + kvm trunk revision 4290)

Any idea what to look for? The SIGSEGV on netbsd makes me suspect that there's a problem with the page tables, but I couldn't narrow down the cause yet.

--
Michael "Tired" Riepe <[email protected]>
X-Tired: Each morning I get up I die a little

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV>
_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

O.K.
I'll try to reproduce it with netbsd, and investigate a little. Uri.

------------------------------------------------------------------------
*From:* Michael Riepe [mailto:[email protected]] *Sent:* Mon 15/01/2007 20:38
*To:* Uri Lublin
*Cc:* [email protected] *Subject:* Re: [kvm-devel] Talking about regressions...

Hi!

Quoted 3 lines · Uri Lublin

Thanks for running those tests.
Does kvm complains ( dmesg ) ?

Not in the cases I listed. The host just becomes very busy, for a long time. After ~30 minutes I lost patience and killed qemu.

I see a single "kvm: emulating exchange as write" when I try to boot FreeBSD 6.1, and with Knoppix 5.1.1 the kernel reports "kvm: unhandled
wrmsr: 0xc1" once. Neither of these worked with earlier kvm versions, though.

Interestingly, Knoppix 5.1.1 *does* work on an AMD Quad Dual-Core Opteron we have at work (running Suse Linux Enterprise Server 10, as far as I remember). Knoppix 5.0, on the other hand, which works fine on the Core Duo, needs "acpi=off" as a boot parameter on the AMD.

More AMD results: FreeBSD, NetBSD and OpenSolaris (32-bit versions) all fail with exceptions, usually page faults. OpenSolaris catches lots of SIGILL's. OpenSuse 10.2 (32 bit as well) works, even with gfxmenu enabled. Qemu-puppy behaves the same as on the Intel CPU - the old version works, the new one sometimes hangs (and at the same point).

The next thing I'm going to try is the Dual Quad-Core Xeon... as soon as I can get my hands on it. :-)

Quoted 30 lines

-----Original Message-----
From: kvm-devel-bounces-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org on behalf of Michael Riepe
Sent: Sun 14/01/2007 19:09
To: [email protected]
Subject: [kvm-devel] Talking about regressions...

When I did a couple of tests the other day, I noticed that one or more bugs must have been introduced at least into the vmx part in kvm-10:

- netbsd 3.1 used to work with kvm-9. In kvm-10, the system hangs during boot. One of the last things I can see is that a shell process dies with a SIGSEGV. Then the load rises to 100% (sys) on the host and stays there until I kill qemu (yes I've waited several minutes).

- opensolaris displays a "Time of Day clock error" during boot, claiming the clock has jumped by 0x5 (no unit given, I suppose that means seconds). Everything else seems to work.

- qemu-puppy-2.13-1 sometimes works and sometimes hangs somewhere inside the initial ramdisk. The exact place seems to differ, but usually it happens while the kernel modules are loaded. Again, host load rises to 100%. An older version (2.01-3) still works fine. kvm-9 works, too.

(Core Duo T2400, 32-bit SMP host, 32-bit guest, Linux 2.6.19 + kvm trunk revision 4290)

Any idea what to look for? The SIGSEGV on netbsd makes me suspect that there's a problem with the page tables, but I couldn't narrow down the cause yet.

Quoted 7 lines

--
Michael "Tired" Riepe <[email protected]>
X-Tired: Each morning I get up I die a little

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to

share your

Quoted 2 lines

opinions on IT & business topics through brief surveys - and earn cash

http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV>

Quoted 1 line

<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV>>

Quoted 6 lines

_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

--
Michael "Tired" Riepe <[email protected]>
X-Tired: Each morning I get up I die a little

O.K.
I'll try to reproduce it with netbsd, and investigate a little. Uri.

------------------------------------------------------------------------
*From:* Michael Riepe [mailto:[email protected]] *Sent:* Mon 15/01/2007 20:38
*To:* Uri Lublin
*Cc:* [email protected] *Subject:* Re: [kvm-devel] Talking about regressions...

Hi!

Quoted 3 lines · Uri Lublin

Thanks for running those tests.
Does kvm complains ( dmesg ) ?

Not in the cases I listed. The host just becomes very busy, for a long time. After ~30 minutes I lost patience and killed qemu.

I see a single "kvm: emulating exchange as write" when I try to boot FreeBSD 6.1, and with Knoppix 5.1.1 the kernel reports "kvm: unhandled
wrmsr: 0xc1" once. Neither of these worked with earlier kvm versions, though.

Interestingly, Knoppix 5.1.1 *does* work on an AMD Quad Dual-Core Opteron we have at work (running Suse Linux Enterprise Server 10, as far as I remember). Knoppix 5.0, on the other hand, which works fine on the Core Duo, needs "acpi=off" as a boot parameter on the AMD.

More AMD results: FreeBSD, NetBSD and OpenSolaris (32-bit versions) all fail with exceptions, usually page faults. OpenSolaris catches lots of SIGILL's. OpenSuse 10.2 (32 bit as well) works, even with gfxmenu enabled. Qemu-puppy behaves the same as on the Intel CPU - the old version works, the new one sometimes hangs (and at the same point).

The next thing I'm going to try is the Dual Quad-Core Xeon... as soon as I can get my hands on it. :-)

Quoted 30 lines

-----Original Message-----
From: kvm-devel-bounces-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org on behalf of Michael Riepe
Sent: Sun 14/01/2007 19:09
To: [email protected]
Subject: [kvm-devel] Talking about regressions...

When I did a couple of tests the other day, I noticed that one or more bugs must have been introduced at least into the vmx part in kvm-10:

- netbsd 3.1 used to work with kvm-9. In kvm-10, the system hangs during boot. One of the last things I can see is that a shell process dies with a SIGSEGV. Then the load rises to 100% (sys) on the host and stays there until I kill qemu (yes I've waited several minutes).

- opensolaris displays a "Time of Day clock error" during boot, claiming the clock has jumped by 0x5 (no unit given, I suppose that means seconds). Everything else seems to work.

- qemu-puppy-2.13-1 sometimes works and sometimes hangs somewhere inside the initial ramdisk. The exact place seems to differ, but usually it happens while the kernel modules are loaded. Again, host load rises to 100%. An older version (2.01-3) still works fine. kvm-9 works, too.

(Core Duo T2400, 32-bit SMP host, 32-bit guest, Linux 2.6.19 + kvm trunk revision 4290)

Any idea what to look for? The SIGSEGV on netbsd makes me suspect that there's a problem with the page tables, but I couldn't narrow down the cause yet.

Quoted 7 lines

--
Michael "Tired" Riepe <[email protected]>
X-Tired: Each morning I get up I die a little

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to

share your

Quoted 2 lines

opinions on IT & business topics through brief surveys - and earn cash

http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV>

Quoted 1 line

<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
<http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV>>

Quoted 6 lines

_______________________________________________
kvm-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/kvm-devel

--
Michael "Tired" Riepe <[email protected]>
X-Tired: Each morning I get up I die a little

Hi!

Just for the records: On the Xeon 53xx, netbsd produces a few kernel messages before it dies with a protection fault:

kvm: msrs: 6
kvm: unhandled rdmsr: 0x1a0
inject_general_protection: rip 0xc0433a1c
kvm: unhandled rdmsr: 0x1a0
inject_general_protection: rip 0xc0433a1c

When I did a couple of tests the other day, I noticed that one or more bugs must have been introduced at least into the vmx part in kvm-10:

- netbsd 3.1 used to work with kvm-9. In kvm-10, the system hangs during boot. One of the last things I can see is that a shell process dies with a SIGSEGV. Then the load rises to 100% (sys) on the host and stays there until I kill qemu (yes I've waited several minutes).

- opensolaris displays a "Time of Day clock error" during boot, claiming the clock has jumped by 0x5 (no unit given, I suppose that means seconds). Everything else seems to work.

- qemu-puppy-2.13-1 sometimes works and sometimes hangs somewhere inside the initial ramdisk. The exact place seems to differ, but usually it happens while the kernel modules are loaded. Again, host load rises to 100%. An older version (2.01-3) still works fine. kvm-9 works, too.

(Core Duo T2400, 32-bit SMP host, 32-bit guest, Linux 2.6.19 + kvm trunk revision 4290)

Any idea what to look for? The SIGSEGV on netbsd makes me suspect that there's a problem with the page tables, but I couldn't narrow down the cause yet.