[Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Erik Rull]

 Hi all,

first of all I'm a bit confused:

What is the difference between qemu with command line option --enable-kvm
and qemu-kvm?
It seems to be a difference in code so far, from the performance point of
view it seems to be the same...

Now my issue that lead me to a git bisect on qemu-kvm:
The following commit / merge breaks my windows guest boot sequence and
causes resets infinitely:
ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge' into
next
Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity <avi@redhat.com>


Interesting: qemu with --enable-kvm master and the same command line
options as qemu-kvm runs perfect.
My command line options are:
qemu-system-x86_64 -serial /dev/ttyS2 -readconfig /etc/ich9-ehci-uhci.cfg
-device usb-host,bus=ehci.0 -device usb-tablet -drive
file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L /usr/X11R6/share/qemu
-boot c -localtime


Best regards,

Erik

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Avi Kivity]

On 02/01/2012 02:52 PM, Erik Rull wrote:
>  Hi all,
>
> first of all I'm a bit confused:
>
> What is the difference between qemu with command line option --enable-kvm
> and qemu-kvm?
> It seems to be a difference in code so far, from the performance point of
> view it seems to be the same...

The differences are being reduced rapidly, thanks to Jan's efforts. 
Right now what remains is PIT performance and accuracy, device
assignment, and Windows XP performance.  Most guests should see the same
performance.

> Now my issue that lead me to a git bisect on qemu-kvm:
> The following commit / merge breaks my windows guest boot sequence and
> causes resets infinitely:
> ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge' into
> next
> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity <avi@redhat.com>
>
>
> Interesting: qemu with --enable-kvm master and the same command line
> options as qemu-kvm runs perfect.
> My command line options are:
> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig /etc/ich9-ehci-uhci.cfg
> -device usb-host,bus=ehci.0 -device usb-tablet -drive
> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L /usr/X11R6/share/qemu
> -boot c -localtime
>
>

What version of Windows are you using?  What's the contents of
/etc/ich9-ehci-uhci.cfg?

-- 
error compiling committee.c: too many arguments to function

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Erik Rull]

On February 1, 2012 at 2:40 PM Avi Kivity <avi@redhat.com> wrote:

> On 02/01/2012 02:52 PM, Erik Rull wrote:
> >  Hi all,
> >
> > first of all I'm a bit confused:
> >
> > What is the difference between qemu with command line option
--enable-kvm
> > and qemu-kvm?
> > It seems to be a difference in code so far, from the performance point
of
> > view it seems to be the same...
>
> The differences are being reduced rapidly, thanks to Jan's efforts.
> Right now what remains is PIT performance and accuracy, device
> assignment, and Windows XP performance.  Most guests should see the same
> performance.
>
> > Now my issue that lead me to a git bisect on qemu-kvm:
> > The following commit / merge breaks my windows guest boot sequence and
> > causes resets infinitely:
> > ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge'
into
> > next
> > Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity
<avi@redhat.com>
> >
> >
> > Interesting: qemu with --enable-kvm master and the same command line
> > options as qemu-kvm runs perfect.
> > My command line options are:
> > qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
/etc/ich9-ehci-uhci.cfg
> > -device usb-host,bus=ehci.0 -device usb-tablet -drive
> > file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
> > tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
/usr/X11R6/share/qemu
> > -boot c -localtime
> >
> >
>
> What version of Windows are you using?  What's the contents of
> /etc/ich9-ehci-uhci.cfg?
>

Hi Avi,

the contents from the .cfg are located in docs/ich9-ehci-uhci.cfg

I tried two versions of Windows XP: One is the default Windows XP SP3 that
you get from MSDN, one is Windows Embedded Standard (embedded customized
XP).

Both show the same behavior: boots with qemu -enable-kvm and continuously
reboots with qemu-kvm.

Best regards,

Erik

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Jan Kiszka]

On 2012-02-01 15:02, Erik Rull wrote:
> 
> On February 1, 2012 at 2:40 PM Avi Kivity <avi@redhat.com> wrote:
> 
>> On 02/01/2012 02:52 PM, Erik Rull wrote:
>>>  Hi all,
>>>
>>> first of all I'm a bit confused:
>>>
>>> What is the difference between qemu with command line option
> --enable-kvm
>>> and qemu-kvm?
>>> It seems to be a difference in code so far, from the performance point
> of
>>> view it seems to be the same...
>>
>> The differences are being reduced rapidly, thanks to Jan's efforts.
>> Right now what remains is PIT performance and accuracy, device
>> assignment, and Windows XP performance.  Most guests should see the same
>> performance.

MSI performance is expected to be worse with upstream as well, thus virtio.

>>
>>> Now my issue that lead me to a git bisect on qemu-kvm:
>>> The following commit / merge breaks my windows guest boot sequence and
>>> causes resets infinitely:
>>> ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge'
> into
>>> next
>>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity
> <avi@redhat.com>
>>>
>>>
>>> Interesting: qemu with --enable-kvm master and the same command line
>>> options as qemu-kvm runs perfect.
>>> My command line options are:
>>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
> /etc/ich9-ehci-uhci.cfg
>>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
>>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
>>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
> /usr/X11R6/share/qemu
>>> -boot c -localtime
>>>
>>>
>>
>> What version of Windows are you using?  What's the contents of
>> /etc/ich9-ehci-uhci.cfg?
>>
> 
> Hi Avi,
> 
> the contents from the .cfg are located in docs/ich9-ehci-uhci.cfg
> 
> I tried two versions of Windows XP: One is the default Windows XP SP3 that
> you get from MSDN, one is Windows Embedded Standard (embedded customized
> XP).
> 
> Both show the same behavior: boots with qemu -enable-kvm and continuously
> reboots with qemu-kvm.

What does qemu-kvm with -no-kvm-irqchip do?

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Erik Rull]

 On February 1, 2012 at 3:42 PM Jan Kiszka <jan.kiszka@siemens.com> wrote:

> On 2012-02-01 15:02, Erik Rull wrote:
> >
> > On February 1, 2012 at 2:40 PM Avi Kivity <avi@redhat.com> wrote:
> >
> >> On 02/01/2012 02:52 PM, Erik Rull wrote:
> >>>  Hi all,
> >>>
> >>> first of all I'm a bit confused:
> >>>
> >>> What is the difference between qemu with command line option
> > --enable-kvm
> >>> and qemu-kvm?
> >>> It seems to be a difference in code so far, from the performance
point
> > of
> >>> view it seems to be the same...
> >>
> >> The differences are being reduced rapidly, thanks to Jan's efforts.
> >> Right now what remains is PIT performance and accuracy, device
> >> assignment, and Windows XP performance.  Most guests should see the
same
> >> performance.
>
> MSI performance is expected to be worse with upstream as well, thus
virtio.
>
> >>
> >>> Now my issue that lead me to a git bisect on qemu-kvm:
> >>> The following commit / merge breaks my windows guest boot sequence
and
> >>> causes resets infinitely:
> >>> ec757c67c40a56492001487e69272f62144fd124 Merge branch
'upstream-merge'
> > into
> >>> next
> >>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity
> > <avi@redhat.com>
> >>>
> >>>
> >>> Interesting: qemu with --enable-kvm master and the same command line
> >>> options as qemu-kvm runs perfect.
> >>> My command line options are:
> >>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
> > /etc/ich9-ehci-uhci.cfg
> >>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
> >>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
> >>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
> > /usr/X11R6/share/qemu
> >>> -boot c -localtime
> >>>
> >>>
> >>
> >> What version of Windows are you using?  What's the contents of
> >> /etc/ich9-ehci-uhci.cfg?
> >>
> >
> > Hi Avi,
> >
> > the contents from the .cfg are located in docs/ich9-ehci-uhci.cfg
> >
> > I tried two versions of Windows XP: One is the default Windows XP SP3
that
> > you get from MSDN, one is Windows Embedded Standard (embedded
customized
> > XP).
> >
> > Both show the same behavior: boots with qemu -enable-kvm and
continuously
> > reboots with qemu-kvm.
>
> What does qemu-kvm with -no-kvm-irqchip do?
>
> Jan
>

Wow - that works!
Does this influence the guest performance?

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Jan Kiszka]

On 2012-02-01 16:43, Erik Rull wrote:
>  On February 1, 2012 at 3:42 PM Jan Kiszka <jan.kiszka@siemens.com> wrote:
> 
>> On 2012-02-01 15:02, Erik Rull wrote:
>>>
>>> On February 1, 2012 at 2:40 PM Avi Kivity <avi@redhat.com> wrote:
>>>
>>>> On 02/01/2012 02:52 PM, Erik Rull wrote:
>>>>>  Hi all,
>>>>>
>>>>> first of all I'm a bit confused:
>>>>>
>>>>> What is the difference between qemu with command line option
>>> --enable-kvm
>>>>> and qemu-kvm?
>>>>> It seems to be a difference in code so far, from the performance
> point
>>> of
>>>>> view it seems to be the same...
>>>>
>>>> The differences are being reduced rapidly, thanks to Jan's efforts.
>>>> Right now what remains is PIT performance and accuracy, device
>>>> assignment, and Windows XP performance.  Most guests should see the
> same
>>>> performance.
>>
>> MSI performance is expected to be worse with upstream as well, thus
> virtio.
>>
>>>>
>>>>> Now my issue that lead me to a git bisect on qemu-kvm:
>>>>> The following commit / merge breaks my windows guest boot sequence
> and
>>>>> causes resets infinitely:
>>>>> ec757c67c40a56492001487e69272f62144fd124 Merge branch
> 'upstream-merge'
>>> into
>>>>> next

So commit aad3b517a1b83561f2755dc4451596a421399c19, i.e. the last merge
before that one is still fine?

>>>>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity
>>> <avi@redhat.com>
>>>>>
>>>>>
>>>>> Interesting: qemu with --enable-kvm master and the same command line
>>>>> options as qemu-kvm runs perfect.
>>>>> My command line options are:
>>>>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
>>> /etc/ich9-ehci-uhci.cfg
>>>>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
>>>>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
>>>>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
>>> /usr/X11R6/share/qemu
>>>>> -boot c -localtime
>>>>>
>>>>>
>>>>
>>>> What version of Windows are you using?  What's the contents of
>>>> /etc/ich9-ehci-uhci.cfg?
>>>>
>>>
>>> Hi Avi,
>>>
>>> the contents from the .cfg are located in docs/ich9-ehci-uhci.cfg
>>>
>>> I tried two versions of Windows XP: One is the default Windows XP SP3
> that
>>> you get from MSDN, one is Windows Embedded Standard (embedded
> customized
>>> XP).
>>>
>>> Both show the same behavior: boots with qemu -enable-kvm and
> continuously
>>> reboots with qemu-kvm.
>>
>> What does qemu-kvm with -no-kvm-irqchip do?
>>
>> Jan
>>
> 
> Wow - that works!
> Does this influence the guest performance?
> 

Yes, how much depends on your workload.

Still strange, though. We should try to understand this issue. It stays
like this up to and including current qemu-kvm.git master?

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Erik Rull]

 On February 1, 2012 at 5:01 PM Jan Kiszka <jan.kiszka@siemens.com> wrote:

> On 2012-02-01 16:43, Erik Rull wrote:
> >  On February 1, 2012 at 3:42 PM Jan Kiszka <jan.kiszka@siemens.com>
wrote:
> >
> >> On 2012-02-01 15:02, Erik Rull wrote:
> >>>
> >>> On February 1, 2012 at 2:40 PM Avi Kivity <avi@redhat.com> wrote:
> >>>
> >>>> On 02/01/2012 02:52 PM, Erik Rull wrote:
> >>>>>  Hi all,
> >>>>>
> >>>>> first of all I'm a bit confused:
> >>>>>
> >>>>> What is the difference between qemu with command line option
> >>> --enable-kvm
> >>>>> and qemu-kvm?
> >>>>> It seems to be a difference in code so far, from the performance
> > point
> >>> of
> >>>>> view it seems to be the same...
> >>>>
> >>>> The differences are being reduced rapidly, thanks to Jan's efforts.
> >>>> Right now what remains is PIT performance and accuracy, device
> >>>> assignment, and Windows XP performance.  Most guests should see the
> > same
> >>>> performance.
> >>
> >> MSI performance is expected to be worse with upstream as well, thus
> > virtio.
> >>
> >>>>
> >>>>> Now my issue that lead me to a git bisect on qemu-kvm:
> >>>>> The following commit / merge breaks my windows guest boot sequence
> > and
> >>>>> causes resets infinitely:
> >>>>> ec757c67c40a56492001487e69272f62144fd124 Merge branch
> > 'upstream-merge'
> >>> into
> >>>>> next
>
> So commit aad3b517a1b83561f2755dc4451596a421399c19, i.e. the last merge
> before that one is still fine?
>
> >>>>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity
> >>> <avi@redhat.com>
> >>>>>
> >>>>>
> >>>>> Interesting: qemu with --enable-kvm master and the same command
line
> >>>>> options as qemu-kvm runs perfect.
> >>>>> My command line options are:
> >>>>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
> >>> /etc/ich9-ehci-uhci.cfg
> >>>>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
> >>>>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
> >>>>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
> >>> /usr/X11R6/share/qemu
> >>>>> -boot c -localtime
> >>>>>
> >>>>>
> >>>>
> >>>> What version of Windows are you using?  What's the contents of
> >>>> /etc/ich9-ehci-uhci.cfg?
> >>>>
> >>>
> >>> Hi Avi,
> >>>
> >>> the contents from the .cfg are located in docs/ich9-ehci-uhci.cfg
> >>>
> >>> I tried two versions of Windows XP: One is the default Windows XP SP3
> > that
> >>> you get from MSDN, one is Windows Embedded Standard (embedded
> > customized
> >>> XP).
> >>>
> >>> Both show the same behavior: boots with qemu -enable-kvm and
> > continuously
> >>> reboots with qemu-kvm.
> >>
> >> What does qemu-kvm with -no-kvm-irqchip do?
> >>
> >> Jan
> >>
> >
> > Wow - that works!
> > Does this influence the guest performance?
> >
>
> Yes, how much depends on your workload.
>
> Still strange, though. We should try to understand this issue. It stays
> like this up to and including current qemu-kvm.git master?
>
> Jan
>

Hi Jan,

I didn't follow the branch, only the master between 1.0 and head.

Here my bisect log:

git bisect start
# good: [30c044521889195f54a9f2c21310894f545994e8] Merge commit 'v1.0' into
next
git bisect good 30c044521889195f54a9f2c21310894f545994e8
# bad: [2793248c5427c0bc585fdf9c101680bab29f4839] Merge remote-tracking
branch 'upstream' into next
git bisect bad 2793248c5427c0bc585fdf9c101680bab29f4839
# good: [262db38871b9a2613761cc5f05c4cf697e246a68] qemu-nbd: asynchronous
operation
git bisect good 262db38871b9a2613761cc5f05c4cf697e246a68
# good: [9737383beb515a583fdb6f2aafa631fcd6797068] qerror: add
check-qerror.sh to verify alphabetical order
git bisect good 9737383beb515a583fdb6f2aafa631fcd6797068
# skip: [fb5458cd10a199e55e622a906b24f8085d922c0f] qmp: add
query-block-jobs
git bisect skip fb5458cd10a199e55e622a906b24f8085d922c0f
# skip: [aa398a5c3a4c0fc29baf02aee5283a7fa0f202a3] blockdev: make image
streaming safe across hotplug
git bisect skip aa398a5c3a4c0fc29baf02aee5283a7fa0f202a3
# good: [506b7ddf889312659b36c667f7ae17bc9e909418] prep: Use i82378
PCI->ISA bridge for 'prep' machine
git bisect good 506b7ddf889312659b36c667f7ae17bc9e909418
# good: [31841e463fc069f3370f85060fbe450b3ff664b4] Merge remote-tracking
branch 'pmaydell/arm-devs.for-upstream' into staging
git bisect good 31841e463fc069f3370f85060fbe450b3ff664b4
# bad: [5fc4ecdf107133914f97590c8a27144ac0852d63] Merge commit
'5b4448d27d7c6ff6e18a1edc8245cb1db783e37c' into queues/qemu-merge
git bisect bad 5fc4ecdf107133914f97590c8a27144ac0852d63
# bad: [289a1056a2b902e25f8d6ec5e17984aa48d201c1] pci-assign: Fix
multifunction support
git bisect bad 289a1056a2b902e25f8d6ec5e17984aa48d201c1
# good: [7c82a98d17a2e86ce6d933dbdd01a570c7bb4ea8] Merge commit
'03ecd2c80a64d030a22fe67cc7a60f24e17ff211' into upstream-merge
git bisect good 7c82a98d17a2e86ce6d933dbdd01a570c7bb4ea8
# bad: [372951014b5008ca047e4dfbfaf4003bc27a2f6b] qemu-kvm: Fix
save/restore of in-kernel i8259
git bisect bad 372951014b5008ca047e4dfbfaf4003bc27a2f6b
# good: [682a3c07f0e28d2532c911a44a9b6142d6299cc2] Merge commit
'c5705a7728b4a6bc9e4f2d35911adbaf28042b25' into upstream-merge
git bisect good 682a3c07f0e28d2532c911a44a9b6142d6299cc2
# good: [d33666a58d6da4554e69ed6036a052f7261c9298] Merge commit
'c47f3223658119219bbe0b8d09da733d1c06e76f' into upstream-merge
git bisect good d33666a58d6da4554e69ed6036a052f7261c9298
# bad: [ec757c67c40a56492001487e69272f62144fd124] Merge branch
'upstream-merge' into next
git bisect bad ec757c67c40a56492001487e69272f62144fd124


Best regards,

Erik

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Jan Kiszka]

On 2012-02-01 13:52, Erik Rull wrote:
>  Hi all,
> 
> first of all I'm a bit confused:
> 
> What is the difference between qemu with command line option --enable-kvm
> and qemu-kvm?
> It seems to be a difference in code so far, from the performance point of
> view it seems to be the same...
> 
> Now my issue that lead me to a git bisect on qemu-kvm:
> The following commit / merge breaks my windows guest boot sequence and
> causes resets infinitely:

Cannot confirm yet, but I have no ACPI-free Windows installation at
hand. Where does it reset, after the BIOS?

> ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge' into
> next
> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity <avi@redhat.com>
> 
> 
> Interesting: qemu with --enable-kvm master and the same command line
> options as qemu-kvm runs perfect.
> My command line options are:
> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig /etc/ich9-ehci-uhci.cfg
> -device usb-host,bus=ehci.0 -device usb-tablet -drive
> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L /usr/X11R6/share/qemu
> -boot c -localtime

Is the BIOS at /usr/X11R6/share/qemu in sync with the qemu version you
try? Does leaving out options change the picture?

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Erik Rull]

Jan Kiszka wrote:
> On 2012-02-01 13:52, Erik Rull wrote:
>>   Hi all,
>>
>> first of all I'm a bit confused:
>>
>> What is the difference between qemu with command line option --enable-kvm
>> and qemu-kvm?
>> It seems to be a difference in code so far, from the performance point of
>> view it seems to be the same...
>>
>> Now my issue that lead me to a git bisect on qemu-kvm:
>> The following commit / merge breaks my windows guest boot sequence and
>> causes resets infinitely:
>
> Cannot confirm yet, but I have no ACPI-free Windows installation at
> hand. Where does it reset, after the BIOS?
>
>> ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge' into
>> next
>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi Kivity<avi@redhat.com>
>>
>>
>> Interesting: qemu with --enable-kvm master and the same command line
>> options as qemu-kvm runs perfect.
>> My command line options are:
>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig /etc/ich9-ehci-uhci.cfg
>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L /usr/X11R6/share/qemu
>> -boot c -localtime
>
> Is the BIOS at /usr/X11R6/share/qemu in sync with the qemu version you
> try? Does leaving out options change the picture?
>
> Jan
>

It happens directly after the windows boot progress bar is completed (I 
boot without logo)

With the -no-kvm-irqchip it seems to be fine...

Best regards,

Erik

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Erik Rull]

On February 1, 2012 at 11:05 PM Erik Rull <erik.rull@rdsoftware.de> wrote:

> Jan Kiszka wrote:
> > On 2012-02-01 13:52, Erik Rull wrote:
> >>   Hi all,
> >>
> >> first of all I'm a bit confused:
> >>
> >> What is the difference between qemu with command line option
--enable-kvm
> >> and qemu-kvm?
> >> It seems to be a difference in code so far, from the performance point
of
> >> view it seems to be the same...
> >>
> >> Now my issue that lead me to a git bisect on qemu-kvm:
> >> The following commit / merge breaks my windows guest boot sequence and
> >> causes resets infinitely:
> >
> > Cannot confirm yet, but I have no ACPI-free Windows installation at
> > hand. Where does it reset, after the BIOS?
> >
> >> ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge'
into
> >> next
> >> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi
Kivity<avi@redhat.com>
> >>
> >>
> >> Interesting: qemu with --enable-kvm master and the same command line
> >> options as qemu-kvm runs perfect.
> >> My command line options are:
> >> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
/etc/ich9-ehci-uhci.cfg
> >> -device usb-host,bus=ehci.0 -device usb-tablet -drive
> >> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
> >> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
/usr/X11R6/share/qemu
> >> -boot c -localtime
> >
> > Is the BIOS at /usr/X11R6/share/qemu in sync with the qemu version you
> > try? Does leaving out options change the picture?
> >
> > Jan
> >
>
> It happens directly after the windows boot progress bar is completed (I
> boot without logo)
>
> With the -no-kvm-irqchip it seems to be fine...
>
> Best regards,
>
> Erik


Hi Jan,

I tested with an ACPI-enabled windows.
Results:
-no-acpi: Continuous reboots like the no-acpi-windows-version
without -no-acpi: boots!

So I tested the no-acpi-windows-version without -no-acpi option - still
rebooting
And without -no-acpi and -no-kvm-irqchip => works again

Best regards,

Erik

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Jan Kiszka]

On 2012-02-02 14:18, Erik Rull wrote:
> 
> On February 1, 2012 at 11:05 PM Erik Rull <erik.rull@rdsoftware.de> wrote:
> 
>> Jan Kiszka wrote:
>>> On 2012-02-01 13:52, Erik Rull wrote:
>>>>   Hi all,
>>>>
>>>> first of all I'm a bit confused:
>>>>
>>>> What is the difference between qemu with command line option
> --enable-kvm
>>>> and qemu-kvm?
>>>> It seems to be a difference in code so far, from the performance point
> of
>>>> view it seems to be the same...
>>>>
>>>> Now my issue that lead me to a git bisect on qemu-kvm:
>>>> The following commit / merge breaks my windows guest boot sequence and
>>>> causes resets infinitely:
>>>
>>> Cannot confirm yet, but I have no ACPI-free Windows installation at
>>> hand. Where does it reset, after the BIOS?
>>>
>>>> ec757c67c40a56492001487e69272f62144fd124 Merge branch 'upstream-merge'
> into
>>>> next
>>>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi
> Kivity<avi@redhat.com>
>>>>
>>>>
>>>> Interesting: qemu with --enable-kvm master and the same command line
>>>> options as qemu-kvm runs perfect.
>>>> My command line options are:
>>>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
> /etc/ich9-ehci-uhci.cfg
>>>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
>>>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
>>>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
> /usr/X11R6/share/qemu
>>>> -boot c -localtime
>>>
>>> Is the BIOS at /usr/X11R6/share/qemu in sync with the qemu version you
>>> try? Does leaving out options change the picture?
>>>
>>> Jan
>>>
>>
>> It happens directly after the windows boot progress bar is completed (I
>> boot without logo)
>>
>> With the -no-kvm-irqchip it seems to be fine...
>>
>> Best regards,
>>
>> Erik
> 
> 
> Hi Jan,
> 
> I tested with an ACPI-enabled windows.
> Results:
> -no-acpi: Continuous reboots like the no-acpi-windows-version
> without -no-acpi: boots!
> 
> So I tested the no-acpi-windows-version without -no-acpi option - still
> rebooting
> And without -no-acpi and -no-kvm-irqchip => works again

Interesting. Need to install such a version, I guess.

What about no-acpi-windows and upstream qemu with kvm and -machine
kernel_irqchip=on?

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Erik Rull]

On February 2, 2012 at 2:21 PM Jan Kiszka <jan.kiszka@siemens.com> wrote:

> On 2012-02-02 14:18, Erik Rull wrote:
> >
> > On February 1, 2012 at 11:05 PM Erik Rull <erik.rull@rdsoftware.de>
wrote:
> >
> >> Jan Kiszka wrote:
> >>> On 2012-02-01 13:52, Erik Rull wrote:
> >>>>   Hi all,
> >>>>
> >>>> first of all I'm a bit confused:
> >>>>
> >>>> What is the difference between qemu with command line option
> > --enable-kvm
> >>>> and qemu-kvm?
> >>>> It seems to be a difference in code so far, from the performance
point
> > of
> >>>> view it seems to be the same...
> >>>>
> >>>> Now my issue that lead me to a git bisect on qemu-kvm:
> >>>> The following commit / merge breaks my windows guest boot sequence
and
> >>>> causes resets infinitely:
> >>>
> >>> Cannot confirm yet, but I have no ACPI-free Windows installation at
> >>> hand. Where does it reset, after the BIOS?
> >>>
> >>>> ec757c67c40a56492001487e69272f62144fd124 Merge branch
'upstream-merge'
> > into
> >>>> next
> >>>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi
> > Kivity<avi@redhat.com>
> >>>>
> >>>>
> >>>> Interesting: qemu with --enable-kvm master and the same command line
> >>>> options as qemu-kvm runs perfect.
> >>>> My command line options are:
> >>>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
> > /etc/ich9-ehci-uhci.cfg
> >>>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
> >>>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
> >>>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
> > /usr/X11R6/share/qemu
> >>>> -boot c -localtime
> >>>
> >>> Is the BIOS at /usr/X11R6/share/qemu in sync with the qemu version
you
> >>> try? Does leaving out options change the picture?
> >>>
> >>> Jan
> >>>
> >>
> >> It happens directly after the windows boot progress bar is completed
(I
> >> boot without logo)
> >>
> >> With the -no-kvm-irqchip it seems to be fine...
> >>
> >> Best regards,
> >>
> >> Erik
> >
> >
> > Hi Jan,
> >
> > I tested with an ACPI-enabled windows.
> > Results:
> > -no-acpi: Continuous reboots like the no-acpi-windows-version
> > without -no-acpi: boots!
> >
> > So I tested the no-acpi-windows-version without -no-acpi option - still
> > rebooting
> > And without -no-acpi and -no-kvm-irqchip => works again
>
> Interesting. Need to install such a version, I guess.
>
> What about no-acpi-windows and upstream qemu with kvm and -machine
> kernel_irqchip=on?
>
> Jan

Boots with and without -no-acpi

Best regards,

Erik

P.S. Too many options for me :-)

Re: [Qemu-devel] git bisect results: ec757c67c40a56492001487e69272f62144fd124 breaks windows boot in qemu-kvm

[Jan Kiszka]

On 2012-02-02 15:07, Erik Rull wrote:
> 
> On February 2, 2012 at 2:21 PM Jan Kiszka <jan.kiszka@siemens.com> wrote:
> 
>> On 2012-02-02 14:18, Erik Rull wrote:
>>>
>>> On February 1, 2012 at 11:05 PM Erik Rull <erik.rull@rdsoftware.de>
> wrote:
>>>
>>>> Jan Kiszka wrote:
>>>>> On 2012-02-01 13:52, Erik Rull wrote:
>>>>>>   Hi all,
>>>>>>
>>>>>> first of all I'm a bit confused:
>>>>>>
>>>>>> What is the difference between qemu with command line option
>>> --enable-kvm
>>>>>> and qemu-kvm?
>>>>>> It seems to be a difference in code so far, from the performance
> point
>>> of
>>>>>> view it seems to be the same...
>>>>>>
>>>>>> Now my issue that lead me to a git bisect on qemu-kvm:
>>>>>> The following commit / merge breaks my windows guest boot sequence
> and
>>>>>> causes resets infinitely:
>>>>>
>>>>> Cannot confirm yet, but I have no ACPI-free Windows installation at
>>>>> hand. Where does it reset, after the BIOS?
>>>>>
>>>>>> ec757c67c40a56492001487e69272f62144fd124 Merge branch
> 'upstream-merge'
>>> into
>>>>>> next
>>>>>> Thu, 5 Jan 2012 11:00:07 +0000 (13:00 +0200)    Avi
>>> Kivity<avi@redhat.com>
>>>>>>
>>>>>>
>>>>>> Interesting: qemu with --enable-kvm master and the same command line
>>>>>> options as qemu-kvm runs perfect.
>>>>>> My command line options are:
>>>>>> qemu-system-x86_64 -serial /dev/ttyS2 -readconfig
>>> /etc/ich9-ehci-uhci.cfg
>>>>>> -device usb-host,bus=ehci.0 -device usb-tablet -drive
>>>>>> file=/dev/sda2,cache=off -m 1024 -net nic,macaddr=$MACADDR -net
>>>>>> tap,script=/etc/qemu-ifup -no-acpi -monitor stdio -L
>>> /usr/X11R6/share/qemu
>>>>>> -boot c -localtime
>>>>>
>>>>> Is the BIOS at /usr/X11R6/share/qemu in sync with the qemu version
> you
>>>>> try? Does leaving out options change the picture?
>>>>>
>>>>> Jan
>>>>>
>>>>
>>>> It happens directly after the windows boot progress bar is completed
> (I
>>>> boot without logo)
>>>>
>>>> With the -no-kvm-irqchip it seems to be fine...
>>>>
>>>> Best regards,
>>>>
>>>> Erik
>>>
>>>
>>> Hi Jan,
>>>
>>> I tested with an ACPI-enabled windows.
>>> Results:
>>> -no-acpi: Continuous reboots like the no-acpi-windows-version
>>> without -no-acpi: boots!
>>>
>>> So I tested the no-acpi-windows-version without -no-acpi option - still
>>> rebooting
>>> And without -no-acpi and -no-kvm-irqchip => works again
>>
>> Interesting. Need to install such a version, I guess.
>>
>> What about no-acpi-windows and upstream qemu with kvm and -machine
>> kernel_irqchip=on?
>>
>> Jan
> 
> Boots with and without -no-acpi

Ok... Either I fixed something magically, or the feature gap makes the
difference. In any case, need to reproduce.

> 
> Best regards,
> 
> Erik
> 
> P.S. Too many options for me :-)

Not just for you. :)

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux