DRIVER_IRQL_NOT_LESS_OR_EQUAL booting WIndows XP with Synaptics driver installed

[Michal Suchanek]

Positng the issue here since I did not get any reply on the ML.

I was trying to update some windows XP (SP3) images in kvm.

It worked fine several times but last time I added mass storage
drivers to sysprep and now on the second boot after reseal (the first
is mini-setup) I get a BSOD with message
DRIVER_IRQL_NOT_LESS_OR_EQUAL.

It turns out that the error is unrelated to storage drivers. It is triggered by Synaptics driver installing for the PS2 mouse in kvm (which does not happen in VirtualBox or on real hardware).

The image is originally created on hardware with MP acpi (not virtualization).

qemu-kvm 0.12.5+dfsg-2

[Michal Suchanek (hramrach)]

Note that most of the section is generated by sysprep (the drivers referencing C:\windows), only small part at the end is generated by my script for additional drivers (references C:\drivers).

The previous images that did not have this section worked only on controllers compatible with the MS generic PCI IDE driver (and also on KVM).

[Jes Sorensen (jes-sorensen)]

Just to be sure, you are not using the virtio-blk driver for Windows here?

I have seen similar crashes with the older version of virtio-blk when used on
recent versions of KVM.

[Michal Suchanek (hramrach)]

On 7 October 2010 11:05, Jes Sorensen <email address hidden> wrote:
> Just to be sure, you are not using the virtio-blk driver for Windows
> here?
>
> I have seen similar crashes with the older version of virtio-blk when used on
> recent versions of KVM.
>
> --
> DRIVER_IRQL_NOT_LESS_OR_EQUAL booting WIndows XP with Synaptics driver installed
> https://bugs.launchpad.net/bugs/639651
> You received this bug notification because you are a member of qemu-
> devel-ml, which is subscribed to QEMU.
>
> Status in QEMU: New
> Status in Debian GNU/Linux: New
>
> Bug description:
> Positng the issue here since I did not get any reply on the ML.
>
> I was trying to update some windows XP (SP3) images in kvm.
>
> It worked fine several times but last time I added mass storage
> drivers to sysprep and now on the second boot after reseal (the first
> is mini-setup) I get a BSOD with message
> DRIVER_IRQL_NOT_LESS_OR_EQUAL.
>
> It turns out that the error is unrelated to storage drivers. It is triggered by Synaptics driver installing for the PS2 mouse in kvm (which does not happen in VirtualBox or on real hardware).
>
> The image is originally created on hardware with MP acpi (not virtualization).
>
> qemu-kvm                  0.12.5+dfsg-2
>
Actually the issue is caused by the Synaptics touchpad driver binding
to the PS/2 mouse device in qemu.

I have no idea how PS/2 devices are detected but the one present in
qemu is misdetected as a synaptics touchapd by the Synaptics driver
for Windows.

As a workaround I have patched my qemu to not enable the PS/2 mouse device.

Thanks

Michal

[Jes Sorensen (jes-sorensen)]

On 10/07/10 11:51, Michal Suchanek wrote:
> Actually the issue is caused by the Synaptics touchpad driver binding
> to the PS/2 mouse device in qemu.
>
> I have no idea how PS/2 devices are detected but the one present in
> qemu is misdetected as a synaptics touchapd by the Synaptics driver
> for Windows.
>
> As a workaround I have patched my qemu to not enable the PS/2 mouse device.

Hi Michal,

If you have the time to look, it would be interesting to see what
actually goes over the wire to/from the PS/2 driver in QEMU just prior
to the crash. It would be good to get this fixed.

Cheers,
Jes

[Michal Suchanek (hramrach)]

I have no idea how to log the data.

I looked at the qemu man page but it does not even mention the PS/2
mouse as a chardev nor offers an option to log traffic of chardevs
without attaching them to a file and thus detaching them from the
emulated device.

Thanks

Michal

[Jes Sorensen (jes-sorensen)]

On 10/07/10 12:17, Michal Suchanek wrote:
> I have no idea how to log the data.
>
> I looked at the qemu man page but it does not even mention the PS/2
> mouse as a chardev nor offers an option to log traffic of chardevs
> without attaching them to a file and thus detaching them from the
> emulated device.

I would attack it by hacking hw/ps.c a bit to monitor the transactions
in ps2_queue() and ps2_read_data().

Cheers,
Jes

[Michal Suchanek (hramrach)]

Attaching logged PS/2 port communication.

[Michal Suchanek (hramrach)]

For reference attaching the patch used for obtaining the log.

[Michal Suchanek (hramrach)]

I guess the problem here is that qemu emulates some very basic mouse (reported as PS/2 compatible mouse in Windows) whereas real hardware usually has a fancy mouse (reported as MS Explorer compatible mouse in Windows).

I don't know how PS/2 mice are detected but due to different mice being detected differently there is obviously some detection mechanism.

The Windos device IDs in a Touchpad driver that does not exhibit the problem (synaptics v8)

[SynMfg]
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN010D
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN010E
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN010F
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0110
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN0111
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0112
%PS2.SynDeviceDesc% = HP__0100__PS2_Inst,*SYN0113
%PS2.SynDeviceDesc% = HP__0100__PS2_Inst,*SYN0114
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0115
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN0116
%PS2.SynDeviceDesc% = HP_GROUP1_PS2_Inst,*SYN0117
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0118
%PS2.SynDeviceDesc% = HP_GROUP5_PS2_Inst,*SYN0119
%PS2.SynDeviceDesc% = HP_GROUP6_PS2_Inst,*SYN011A
%PS2.SynDeviceDesc% = HP_GROUP4_PS2_Inst,*SYN011B
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN011C
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN011D
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN011E
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN011F
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0120
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN0121
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0122
%PS2.SynDeviceDesc% = HP_GROUP7_PS2_Inst,*SYN0123
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0124
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN0125
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0126
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0127
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN0128
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0129
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN012A
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN012B
%PS2.SynDeviceDesc% = HP_GROUP3_PS2_Inst,*SYN012C

and in drivers that do exhibit the problem (note the first device):

Synaptics v9
[SynMfg]
%PS2.SynDeviceDesc% = HP__0100__PS2_Inst,*PNP0F13
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN010D
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN010E
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN010F
%PS2.SynDeviceDesc% = HP_GROUP8_PS2_Inst,*SYN0110
%PS2.SynDeviceDesc% = HP_GROUP9_PS2_Inst,*SYN0111
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0112
%PS2.SynDeviceDesc% = HP__0100__PS2_Inst,*SYN0113
%PS2.SynDeviceDesc% = HP__0100__PS2_Inst,*SYN0114
%PS2.SynDeviceDesc% = HP_GROUP8_PS2_Inst,*SYN0115
%PS2.SynDeviceDesc% = HP_GROUP9_PS2_Inst,*SYN0116
%PS2.SynDeviceDesc% = HP_GROUP1_PS2_Inst,*SYN0117
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN0118
%PS2.SynDeviceDesc% = HP_GROUP5_PS2_Inst,*SYN0119
%PS2.SynDeviceDesc% = HP_GROUP6_PS2_Inst,*SYN011A
%PS2.SynDeviceDesc% = HP_GROUP4_PS2_Inst,*SYN011B
%PS2.SynDeviceDesc% = HP_GROUP9_PS2_Inst,*SYN011C
%PS2.SynDeviceDesc% = HP_GROUP2_PS2_Inst,*SYN011D
%PS2.SynDeviceDesc% = HP_GROUP8_PS2_Inst,*SYN011E
%PS2.SynDeviceDesc% = HP_GROUP9_PS...

[Thomas Huth (th-huth)]

QEMU 0.12 is pretty much outdated nowadays ... can you still reproduce this problem with the latest version of QEMU, or can we close this ticket nowadays?

[Michal Suchanek (hramrach)]

Does qemu allow to disable the PS/2 port now?

If so then there is easy workaround in case something similar ever happens.

[Thomas Huth (th-huth)]

Hmm, I'm not aware of a way to disable the PS2 mouse in QEMU yet. Looking at your other related bug (https://bugs.launchpad.net/qemu/+bug/813546), I think you might need to discuss that patch on the qemu-devel mailing list.

[Launchpad Janitor (janitor)]

[Expired for QEMU because there has been no activity for 60 days.]