(pv)hvm, upstream qemu 1.7.1

classic Classic list List threaded Threaded
13 messages Options
Reply | Threaded
Open this post in threaded view
|

(pv)hvm, upstream qemu 1.7.1

Jacek Konieczny
Hi,

I wanted to try running a PVHVM linux VM under Xen 4.4.0, using upstream
qemu 1.7.1.

I have prepared a system image, which I can successfully run in the
following modes:
- PV
– PVH (using fixes from Xen 4.4 git branch to prevent Xen lock-up on the
  domU shutdown)
- PVHVM, using device_model_version="qemu-xen-traditional"

What does not work is (PV)HVM with device_model_version="qemu-xen". My
Xen is compiled with '--with-system-qemu=...' and uses the system
qemu-system-x86_64 binary, which is qemu 1.7.1.

This is supposed to work, according to
http://wiki.xen.org/wiki/QEMU_Upstream.

The domain won't start. Or, rather, it crashes/reboots immediately (I
have stopped this with the 'on_reboot/on_crash' settings).
There is little interesting in the logs, except the one error in 'xl
dmesg':

(XEN) io.c:204:d58 MMIO emulation failed @ 0008:ffff34d1: 10 89 f9 1e 04
83 ff ff 06 02


The config file ('pvhtest.cfg'):

memory = 256
vcpus = 1
name = "pvhtest"
vif = [ 'mac=02:00:0f:ff:00:1E, bridge=xenbr0']
disk = [ 'phy:/dev/vg/pvhtest,hda,w' ]
#bootloader = 'pygrub'
#pvh = 1
builder = 'hvm'
xen_platform_pci=1
boot="c"
paused = 1
pae=1
acpi=1
apic=1
stdvga=0
vnc=1
vncdisplay=1
vncpasswd="dupa"
serial='pty'
on_poweroff = 'destroy'
on_reboot   = 'preserve'
on_crash    = 'preserve'
device_model_version="qemu-xen"

The start command:

# xl -v create pvhtest.cfg
Parsing config from pvhtest.cfg
libxl: detail: libxl_dom.c:195:numa_place_domain: NUMA placement
candidate with 1 nodes, 4 cpus and 14117 KB free selected
xc: detail: elf_parse_binary: phdr: paddr=0x100000 memsz=0x9ef68
xc: detail: elf_parse_binary: memory: 0x100000 -> 0x19ef68
xc: detail: VIRTUAL MEMORY ARRANGEMENT:
  Loader:        0000000000100000->000000000019ef68
  Modules:       0000000000000000->0000000000000000
  TOTAL:         0000000000000000->000000000f800000
  ENTRY ADDRESS: 0000000000100620
xc: detail: PHYSICAL MEMORY ALLOCATION:
  4KB PAGES: 0x0000000000000200
  2MB PAGES: 0x000000000000007b
  1GB PAGES: 0x0000000000000000
xc: detail: elf_load_binary: phdr 0 at 0x7fd78ab46000 -> 0x7fd78abdbded

logs:

qemu-dm-pvhtest.log:

char device redirected to /dev/pts/4 (label serial0)

xl-pvhtest.log:

Waiting for domain pvhtest (domid 59) to die [pid 4914]
Domain 59 has shut down, reason code 1 0x1
Action for shutdown reason code 1 is preserve
Done. Exiting now

xl dmesg:

(XEN) io.c:204:d58 MMIO emulation failed @ 0008:ffff34d1: 10 89 f9 1e 04
83 ff ff 06 02


What is going wrong here? How do I debug that?

Greets,
        Jacek

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Ian Campbell-10
Anthony, Stefano: Any idea what this issue is?

On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:

> Hi,
>
> I wanted to try running a PVHVM linux VM under Xen 4.4.0, using upstream
> qemu 1.7.1.
>
> I have prepared a system image, which I can successfully run in the
> following modes:
> - PV
> – PVH (using fixes from Xen 4.4 git branch to prevent Xen lock-up on the
>   domU shutdown)
> - PVHVM, using device_model_version="qemu-xen-traditional"
>
> What does not work is (PV)HVM with device_model_version="qemu-xen". My
> Xen is compiled with '--with-system-qemu=...' and uses the system
> qemu-system-x86_64 binary, which is qemu 1.7.1.
>
> This is supposed to work, according to
> http://wiki.xen.org/wiki/QEMU_Upstream.
>
> The domain won't start. Or, rather, it crashes/reboots immediately (I
> have stopped this with the 'on_reboot/on_crash' settings).
> There is little interesting in the logs, except the one error in 'xl
> dmesg':
>
> (XEN) io.c:204:d58 MMIO emulation failed @ 0008:ffff34d1: 10 89 f9 1e 04
> 83 ff ff 06 02
>
>
> The config file ('pvhtest.cfg'):
>
> memory = 256
> vcpus = 1
> name = "pvhtest"
> vif = [ 'mac=02:00:0f:ff:00:1E, bridge=xenbr0']
> disk = [ 'phy:/dev/vg/pvhtest,hda,w' ]
> #bootloader = 'pygrub'
> #pvh = 1
> builder = 'hvm'
> xen_platform_pci=1
> boot="c"
> paused = 1
> pae=1
> acpi=1
> apic=1
> stdvga=0
> vnc=1
> vncdisplay=1
> vncpasswd="dupa"
> serial='pty'
> on_poweroff = 'destroy'
> on_reboot   = 'preserve'
> on_crash    = 'preserve'
> device_model_version="qemu-xen"
>
> The start command:
>
> # xl -v create pvhtest.cfg
> Parsing config from pvhtest.cfg
> libxl: detail: libxl_dom.c:195:numa_place_domain: NUMA placement
> candidate with 1 nodes, 4 cpus and 14117 KB free selected
> xc: detail: elf_parse_binary: phdr: paddr=0x100000 memsz=0x9ef68
> xc: detail: elf_parse_binary: memory: 0x100000 -> 0x19ef68
> xc: detail: VIRTUAL MEMORY ARRANGEMENT:
>   Loader:        0000000000100000->000000000019ef68
>   Modules:       0000000000000000->0000000000000000
>   TOTAL:         0000000000000000->000000000f800000
>   ENTRY ADDRESS: 0000000000100620
> xc: detail: PHYSICAL MEMORY ALLOCATION:
>   4KB PAGES: 0x0000000000000200
>   2MB PAGES: 0x000000000000007b
>   1GB PAGES: 0x0000000000000000
> xc: detail: elf_load_binary: phdr 0 at 0x7fd78ab46000 -> 0x7fd78abdbded
>
> logs:
>
> qemu-dm-pvhtest.log:
>
> char device redirected to /dev/pts/4 (label serial0)
>
> xl-pvhtest.log:
>
> Waiting for domain pvhtest (domid 59) to die [pid 4914]
> Domain 59 has shut down, reason code 1 0x1
> Action for shutdown reason code 1 is preserve
> Done. Exiting now
>
> xl dmesg:
>
> (XEN) io.c:204:d58 MMIO emulation failed @ 0008:ffff34d1: 10 89 f9 1e 04
> 83 ff ff 06 02
>
>
> What is going wrong here? How do I debug that?
>
> Greets,
> Jacek
>
> _______________________________________________
> Xen-users mailing list
> [hidden email]
> http://lists.xen.org/xen-users



_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Stefano Stabellini-3
On Mon, 14 Apr 2014, Ian Campbell wrote:

> Anthony, Stefano: Any idea what this issue is?
>
> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
> > Hi,
> >
> > I wanted to try running a PVHVM linux VM under Xen 4.4.0, using upstream
> > qemu 1.7.1.
> >
> > I have prepared a system image, which I can successfully run in the
> > following modes:
> > - PV
> > – PVH (using fixes from Xen 4.4 git branch to prevent Xen lock-up on the
> >   domU shutdown)
> > - PVHVM, using device_model_version="qemu-xen-traditional"
> >
> > What does not work is (PV)HVM with device_model_version="qemu-xen". My
> > Xen is compiled with '--with-system-qemu=...' and uses the system
> > qemu-system-x86_64 binary, which is qemu 1.7.1.
> >
> > This is supposed to work, according to
> > http://wiki.xen.org/wiki/QEMU_Upstream.
Sorry for the late reply.
If you compile qemu-system-i386 instead of qemu-system-x86_64, does that
work for you?



> > The domain won't start. Or, rather, it crashes/reboots immediately (I
> > have stopped this with the 'on_reboot/on_crash' settings).
> > There is little interesting in the logs, except the one error in 'xl
> > dmesg':
> >
> > (XEN) io.c:204:d58 MMIO emulation failed @ 0008:ffff34d1: 10 89 f9 1e 04
> > 83 ff ff 06 02
> >
> >
> > The config file ('pvhtest.cfg'):
> >
> > memory = 256
> > vcpus = 1
> > name = "pvhtest"
> > vif = [ 'mac=02:00:0f:ff:00:1E, bridge=xenbr0']
> > disk = [ 'phy:/dev/vg/pvhtest,hda,w' ]
> > #bootloader = 'pygrub'
> > #pvh = 1
> > builder = 'hvm'
> > xen_platform_pci=1
> > boot="c"
> > paused = 1
> > pae=1
> > acpi=1
> > apic=1
> > stdvga=0
> > vnc=1
> > vncdisplay=1
> > vncpasswd="dupa"
> > serial='pty'
> > on_poweroff = 'destroy'
> > on_reboot   = 'preserve'
> > on_crash    = 'preserve'
> > device_model_version="qemu-xen"
> >
> > The start command:
> >
> > # xl -v create pvhtest.cfg
> > Parsing config from pvhtest.cfg
> > libxl: detail: libxl_dom.c:195:numa_place_domain: NUMA placement
> > candidate with 1 nodes, 4 cpus and 14117 KB free selected
> > xc: detail: elf_parse_binary: phdr: paddr=0x100000 memsz=0x9ef68
> > xc: detail: elf_parse_binary: memory: 0x100000 -> 0x19ef68
> > xc: detail: VIRTUAL MEMORY ARRANGEMENT:
> >   Loader:        0000000000100000->000000000019ef68
> >   Modules:       0000000000000000->0000000000000000
> >   TOTAL:         0000000000000000->000000000f800000
> >   ENTRY ADDRESS: 0000000000100620
> > xc: detail: PHYSICAL MEMORY ALLOCATION:
> >   4KB PAGES: 0x0000000000000200
> >   2MB PAGES: 0x000000000000007b
> >   1GB PAGES: 0x0000000000000000
> > xc: detail: elf_load_binary: phdr 0 at 0x7fd78ab46000 -> 0x7fd78abdbded
> >
> > logs:
> >
> > qemu-dm-pvhtest.log:
> >
> > char device redirected to /dev/pts/4 (label serial0)
> >
> > xl-pvhtest.log:
> >
> > Waiting for domain pvhtest (domid 59) to die [pid 4914]
> > Domain 59 has shut down, reason code 1 0x1
> > Action for shutdown reason code 1 is preserve
> > Done. Exiting now
> >
> > xl dmesg:
> >
> > (XEN) io.c:204:d58 MMIO emulation failed @ 0008:ffff34d1: 10 89 f9 1e 04
> > 83 ff ff 06 02
> >
> >
> > What is going wrong here? How do I debug that?
> >
> > Greets,
> > Jacek
> >
> > _______________________________________________
> > Xen-users mailing list
> > [hidden email]
> > http://lists.xen.org/xen-users
>
>
_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Jacek Konieczny
On 05/07/14 18:55, Stefano Stabellini wrote:

> On Mon, 14 Apr 2014, Ian Campbell wrote:
>> Anthony, Stefano: Any idea what this issue is?
>>
>> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
>>>
>>> What does not work is (PV)HVM with device_model_version="qemu-xen". My
>>> Xen is compiled with '--with-system-qemu=...' and uses the system
>>> qemu-system-x86_64 binary, which is qemu 1.7.1.
>>>
>>> This is supposed to work, according to
>>> http://wiki.xen.org/wiki/QEMU_Upstream.
>
> Sorry for the late reply.
> If you compile qemu-system-i386 instead of qemu-system-x86_64, does that
> work for you?

That only changes the default DM path, right?

So, I have tried:

device_model_override = "/usr/bin/qemu-system-i386"

And this changed nothing.

I have noticed something else though…

The config file I have been playing with contains:

memory = 256

And trying to boot the domain causes 'MMIO emulation failed' errors with
various code:

(XEN) io.c:204:d1 MMIO emulation failed @ 0008:ffff34d1: 10 90 39 00 04
83 ff ff 06 02
(XEN) io.c:204:d2 MMIO emulation failed @ 0008:ffff34d1: e0 29 00 08 e0
82 ff ff cd 18
(XEN) io.c:204:d3 MMIO emulation failed @ 0008:ffff34d1: 50 16 57 00 04
83 ff ff 06 02
(XEN) io.c:204:d4 MMIO emulation failed @ 0008:ffff34d1: 20 5a 57 00 04
83 ff ff 06 02

(XEN) io.c:204:d16 MMIO emulation failed @ 0008:ffff34d1: e0 29 00 08 e0
82 ff ff cd 18
(XEN) io.c:204:d17 MMIO emulation failed @ 0008:ffff34d1: 10 29 57 00 04
83 ff ff 06 02
(XEN) io.c:204:d18 MMIO emulation failed @ 0008:ffff34d1: e0 29 00 08 e0
82 ff ff cd 18

but when I changed it to 'memory = 4096' (with an assumption that 256MB
may be
a bit to little for a 64-bit HVM guest) the VM still failed, but with
more consistent error messages:

(XEN) io.c:204:d10 MMIO emulation failed @ 0008:ffff34d1: 6a 05 40 00 00
00 00 00 01 00
(XEN) io.c:204:d11 MMIO emulation failed @ 0008:ffff34d1: 6e 05 40 00 00
00 00 00 01 00
(XEN) io.c:204:d12 MMIO emulation failed @ 0008:ffff34d1: 6e 05 40 00 00
00 00 00 01 00
(XEN) io.c:204:d13 MMIO emulation failed @ 0008:ffff34d1: 6d 05 40 00 00
00 00 00 01 00
(XEN) io.c:204:d14 MMIO emulation failed @ 0008:ffff34d1: 6d 05 40 00 00
00 00 00 01 00
(XEN) io.c:204:d15 MMIO emulation failed @ 0008:ffff34d1: 6d 05 40 00 00
00 00 00 01 00

I guess this may be of little help, but maybe…

Greets,
        Jacek

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Stefano Stabellini-3
On Thu, 8 May 2014, Jacek Konieczny wrote:

> On 05/07/14 18:55, Stefano Stabellini wrote:
> > On Mon, 14 Apr 2014, Ian Campbell wrote:
> >> Anthony, Stefano: Any idea what this issue is?
> >>
> >> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
> >>>
> >>> What does not work is (PV)HVM with device_model_version="qemu-xen". My
> >>> Xen is compiled with '--with-system-qemu=...' and uses the system
> >>> qemu-system-x86_64 binary, which is qemu 1.7.1.
> >>>
> >>> This is supposed to work, according to
> >>> http://wiki.xen.org/wiki/QEMU_Upstream.
> >
> > Sorry for the late reply.
> > If you compile qemu-system-i386 instead of qemu-system-x86_64, does that
> > work for you?
>
> That only changes the default DM path, right?
>
> So, I have tried:
>
> device_model_override = "/usr/bin/qemu-system-i386"

That basically defaults to whatever qemu was installed by your distro:
Xen installs qemu to /usr/lib/xen/bin, so you would need:

device_model_override = "/usr/lib/xen/bin/qemu-system-i386"

But actually if you just add:

device_model_version = "qemu-xen"

xl should automatically look for qemu-system-i386 under
/usr/lib/xen/bin.

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Ian Campbell-10
On Fri, 2014-05-09 at 15:08 +0100, Stefano Stabellini wrote:

> On Thu, 8 May 2014, Jacek Konieczny wrote:
> > On 05/07/14 18:55, Stefano Stabellini wrote:
> > > On Mon, 14 Apr 2014, Ian Campbell wrote:
> > >> Anthony, Stefano: Any idea what this issue is?
> > >>
> > >> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
> > >>>
> > >>> What does not work is (PV)HVM with device_model_version="qemu-xen". My
> > >>> Xen is compiled with '--with-system-qemu=...' and uses the system
> > >>> qemu-system-x86_64 binary, which is qemu 1.7.1.
> > >>>
> > >>> This is supposed to work, according to
> > >>> http://wiki.xen.org/wiki/QEMU_Upstream.
> > >
> > > Sorry for the late reply.
> > > If you compile qemu-system-i386 instead of qemu-system-x86_64, does that
> > > work for you?
> >
> > That only changes the default DM path, right?
> >
> > So, I have tried:
> >
> > device_model_override = "/usr/bin/qemu-system-i386"
>
> That basically defaults to whatever qemu was installed by your distro:

FWIW Jacek did exactly what I thought you had asked him to do, i.e. try
the qemu-system-i386 compiled on his system. Were you really trying to
ask if the Xen version of qemu worked?

> Xen installs qemu to /usr/lib/xen/bin, so you would need:
>
> device_model_override = "/usr/lib/xen/bin/qemu-system-i386"
>
> But actually if you just add:
>
> device_model_version = "qemu-xen"
>
> xl should automatically look for qemu-system-i386 under
> /usr/lib/xen/bin.

Note that using the distro provided qemu is *supposed* to work...

Ian.


_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Stefano Stabellini-3
On Fri, 9 May 2014, Ian Campbell wrote:

> On Fri, 2014-05-09 at 15:08 +0100, Stefano Stabellini wrote:
> > On Thu, 8 May 2014, Jacek Konieczny wrote:
> > > On 05/07/14 18:55, Stefano Stabellini wrote:
> > > > On Mon, 14 Apr 2014, Ian Campbell wrote:
> > > >> Anthony, Stefano: Any idea what this issue is?
> > > >>
> > > >> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
> > > >>>
> > > >>> What does not work is (PV)HVM with device_model_version="qemu-xen". My
> > > >>> Xen is compiled with '--with-system-qemu=...' and uses the system
> > > >>> qemu-system-x86_64 binary, which is qemu 1.7.1.
> > > >>>
> > > >>> This is supposed to work, according to
> > > >>> http://wiki.xen.org/wiki/QEMU_Upstream.
> > > >
> > > > Sorry for the late reply.
> > > > If you compile qemu-system-i386 instead of qemu-system-x86_64, does that
> > > > work for you?
> > >
> > > That only changes the default DM path, right?
> > >
> > > So, I have tried:
> > >
> > > device_model_override = "/usr/bin/qemu-system-i386"
> >
> > That basically defaults to whatever qemu was installed by your distro:
>
> FWIW Jacek did exactly what I thought you had asked him to do, i.e. try
> the qemu-system-i386 compiled on his system. Were you really trying to
> ask if the Xen version of qemu worked?

I realize now that I should have been much clearer in my reply: what I
meant is please try to make sure that you are actually testing with the
qemu you have compiled yourself, rather than another version. Keep in
mind that if you use the Xen default paths, it would be installed under
/usr/lib/xen/bin rather than /usr/bin.

In fact could you please post the QEMU configure line and output so that
I can try to repro the issue myself?


> > Xen installs qemu to /usr/lib/xen/bin, so you would need:
> >
> > device_model_override = "/usr/lib/xen/bin/qemu-system-i386"
> >
> > But actually if you just add:
> >
> > device_model_version = "qemu-xen"
> >
> > xl should automatically look for qemu-system-i386 under
> > /usr/lib/xen/bin.
>
> Note that using the distro provided qemu is *supposed* to work...
 
That is true, but if we are trying to debug a specific version of QEMU
is best to avoid comparing apples with oranges.

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Alvin Starr-2
In reply to this post by Ian Campbell-10
As a side comment.

I just ran the test of device_model = "/usr/bin/qemu-system-i386"
it is from qemu-system-x86-1.6.2-4.fc20.x86_64.
So that it does work in at least one situation.

On 05/09/2014 10:11 AM, Ian Campbell wrote:

> On Fri, 2014-05-09 at 15:08 +0100, Stefano Stabellini wrote:
>> On Thu, 8 May 2014, Jacek Konieczny wrote:
>>> On 05/07/14 18:55, Stefano Stabellini wrote:
>>>> On Mon, 14 Apr 2014, Ian Campbell wrote:
>>>>> Anthony, Stefano: Any idea what this issue is?
>>>>>
>>>>> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
>>>>>> What does not work is (PV)HVM with device_model_version="qemu-xen". My
>>>>>> Xen is compiled with '--with-system-qemu=...' and uses the system
>>>>>> qemu-system-x86_64 binary, which is qemu 1.7.1.
>>>>>>
>>>>>> This is supposed to work, according to
>>>>>> http://wiki.xen.org/wiki/QEMU_Upstream.
>>>> Sorry for the late reply.
>>>> If you compile qemu-system-i386 instead of qemu-system-x86_64, does that
>>>> work for you?
>>> That only changes the default DM path, right?
>>>
>>> So, I have tried:
>>>
>>> device_model_override = "/usr/bin/qemu-system-i386"
>> That basically defaults to whatever qemu was installed by your distro:
> FWIW Jacek did exactly what I thought you had asked him to do, i.e. try
> the qemu-system-i386 compiled on his system. Were you really trying to
> ask if the Xen version of qemu worked?
>
>> Xen installs qemu to /usr/lib/xen/bin, so you would need:
>>
>> device_model_override = "/usr/lib/xen/bin/qemu-system-i386"
>>
>> But actually if you just add:
>>
>> device_model_version = "qemu-xen"
>>
>> xl should automatically look for qemu-system-i386 under
>> /usr/lib/xen/bin.
> Note that using the distro provided qemu is *supposed* to work...
>
> Ian.
>
>
> _______________________________________________
> Xen-users mailing list
> [hidden email]
> http://lists.xen.org/xen-users


--
Alvin Starr                   ||   voice: (905)513-7688
Netvel Inc.                   ||   Cell:  (416)806-0133
[hidden email]              ||


_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: (pv)hvm, upstream qemu 1.7.1

Jacek Konieczny
In reply to this post by Stefano Stabellini-3
On 2014-05-09 16:21, Stefano Stabellini wrote:

> On Fri, 9 May 2014, Ian Campbell wrote:
>> On Fri, 2014-05-09 at 15:08 +0100, Stefano Stabellini wrote:
>>> On Thu, 8 May 2014, Jacek Konieczny wrote:
>>>>
>>>> That only changes the default DM path, right?
>>>>
>>>> So, I have tried:
>>>>
>>>> device_model_override = "/usr/bin/qemu-system-i386"
>>>
>>> That basically defaults to whatever qemu was installed by your distro:
>>
>> FWIW Jacek did exactly what I thought you had asked him to do, i.e. try
>> the qemu-system-i386 compiled on his system. Were you really trying to
>> ask if the Xen version of qemu worked?
>
> I realize now that I should have been much clearer in my reply: what I
> meant is please try to make sure that you are actually testing with the
> qemu you have compiled yourself, rather than another version.

I can understand the confusion. In this case I am a Linux distribution
developer and this is the exact binary I want to debug and use.
Currently I don't even know if the QEMU or the Xen package is to blame.

> In fact could you please post the QEMU configure line and output so that
> I can try to repro the issue myself?

The binary was built as a part of an RPM package, so the 'configure'
line used is not simple:

./configure '--extra-cflags=-O2 -fwrapv -pipe -Wformat \
  -Werror=format-security -gdwarf-4 -fno-debug-types-section \
  -fvar-tracking-assignments -g2 -Wp,-D_FORTIFY_SOURCE=2 -fstack-protector \
  --param=ssp-buffer-size=4 -fPIC -march=x86-64 -gdwarf-4 \
  -fno-debug-types-section -fvar-tracking-assignments -g2 \
  -I/usr/include/ncurses' '--extra-ldflags=-Wl,--as-needed \
  -Wl,--no-copy-dt-needed-entries -Wl,-z,relro -Wl,-z,combreloc ' \
  '--prefix=/usr' '--sysconfdir=/etc' '--cc=x86_64-pld-linux-gcc' \
  '--host-cc=x86_64-pld-linux-gcc' --disable-strip --enable-attr \
  --disable-bluez --disable-brlapi --enable-cap-ng --enable-curl \
  --enable-curses --enable-docs --enable-fdt --disable-glusterfs \
  --disable-libiscsi --disable-glx --disable-rbd --disable-rdma \
  --disable-sdl --disable-seccomp --disable-spice --disable-smartcard-nss \
  --enable-tpm --disable-usb-redir --enable-uuid --enable-vde \
  --enable-virtfs --enable-vnc-jpeg --enable-vnc-png --enable-vnc-sasl \
  --enable-vnc-tls --enable-kvm --enable-xen '--audio-drv-list=alsa' \
  '--interp-prefix=/usr/lib64/qemu/lib-%M' --disable-gtk

The full build log is available here:
https://www.dropbox.com/s/3yjaq3lcq6sui6m/qemu%2Cc6a3873e-59bb-49e5-9839-c8e0bf679085.bz2

Greets,
  Jacek

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: [Solved] (pv)hvm, upstream qemu 1.7.1

Jacek Konieczny
In reply to this post by Stefano Stabellini-3
On 2014-05-09 16:21, Stefano Stabellini wrote:

> On Fri, 9 May 2014, Ian Campbell wrote:
>> On Fri, 2014-05-09 at 15:08 +0100, Stefano Stabellini wrote:
>>> On Thu, 8 May 2014, Jacek Konieczny wrote:
>>>> On 05/07/14 18:55, Stefano Stabellini wrote:
>>>>> On Mon, 14 Apr 2014, Ian Campbell wrote:
>>>>>> Anthony, Stefano: Any idea what this issue is?
>>>>>>
>>>>>> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
>>>>>>>
>>>>>>> What does not work is (PV)HVM with device_model_version="qemu-xen". My
>>>>>>> Xen is compiled with '--with-system-qemu=...' and uses the system
>>>>>>> qemu-system-x86_64 binary, which is qemu 1.7.1.
>>>>>>>
>>>>>>> This is supposed to work, according to
>>>>>>> http://wiki.xen.org/wiki/QEMU_Upstream.

After reading some other, related posts on the xen-devel list I managed
to fix my problem.

I rebuilt my Xen with:
    --with-system-seabios=/usr/share/qemu/bios.bin
to force it to use the SeaBIOS provided by QEMU. Now the 'MMIO emulation'
errors are gone and the domU boots properly.

It seems the SeaBIOS used by default in the Xen build process is not
compatible with QEMU 1.7.1.

Now, I am going to prepare a separate package with SeaBIOS for my
distribution (PLD Linux) and make both QEMU (on runtime) and Xen (during
boot) use that.

Greets,
        Jacek

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: [Solved] (pv)hvm, upstream qemu 1.7.1

Stefano Stabellini-3
On Sun, 11 May 2014, Jacek Konieczny wrote:

> On 2014-05-09 16:21, Stefano Stabellini wrote:
> > On Fri, 9 May 2014, Ian Campbell wrote:
> >> On Fri, 2014-05-09 at 15:08 +0100, Stefano Stabellini wrote:
> >>> On Thu, 8 May 2014, Jacek Konieczny wrote:
> >>>> On 05/07/14 18:55, Stefano Stabellini wrote:
> >>>>> On Mon, 14 Apr 2014, Ian Campbell wrote:
> >>>>>> Anthony, Stefano: Any idea what this issue is?
> >>>>>>
> >>>>>> On Fri, 2014-04-11 at 21:44 +0200, Jacek Konieczny wrote:
> >>>>>>>
> >>>>>>> What does not work is (PV)HVM with device_model_version="qemu-xen". My
> >>>>>>> Xen is compiled with '--with-system-qemu=...' and uses the system
> >>>>>>> qemu-system-x86_64 binary, which is qemu 1.7.1.
> >>>>>>>
> >>>>>>> This is supposed to work, according to
> >>>>>>> http://wiki.xen.org/wiki/QEMU_Upstream.
>
> After reading some other, related posts on the xen-devel list I managed
> to fix my problem.
>
> I rebuilt my Xen with:
>     --with-system-seabios=/usr/share/qemu/bios.bin
> to force it to use the SeaBIOS provided by QEMU. Now the 'MMIO emulation'
> errors are gone and the domU boots properly.
>
> It seems the SeaBIOS used by default in the Xen build process is not
> compatible with QEMU 1.7.1.
>
> Now, I am going to prepare a separate package with SeaBIOS for my
> distribution (PLD Linux) and make both QEMU (on runtime) and Xen (during
> boot) use that.

Thanks for narrowing it down by yourself, and sorry that I couldn't help you.
Cheers,

Stefano

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: [Solved] (pv)hvm, upstream qemu 1.7.1

Sylvain Munaut
Hi,


Interestingly I had the same "MMIO emulation" issue when trying to
build xen-4.4 package for debian wheezy (based on the xen 4.3 package
for jessie).

But in that case, it was kind of the opposite, I was trying to use the
system seabios and etherboot (like in xen 4.3 debian package), but
this caused the compiled hvmloader to be broken. By letting the xen
build use its own seabios and etherboot to build hvmloader, it then
worked fine. And then I'm using the distribution provided qemu 2.0
(which should be using the system seabios if it needs one).


Cheers,

    Sylvain

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users
Reply | Threaded
Open this post in threaded view
|

Re: [Solved] (pv)hvm, upstream qemu 1.7.1

Jacek Konieczny
On 05/14/14 10:05, Sylvain Munaut wrote:

> Interestingly I had the same "MMIO emulation" issue when trying to
> build xen-4.4 package for debian wheezy (based on the xen 4.3 package
> for jessie).
>
> But in that case, it was kind of the opposite, I was trying to use the
> system seabios and etherboot (like in xen 4.3 debian package), but
> this caused the compiled hvmloader to be broken. By letting the xen
> build use its own seabios and etherboot to build hvmloader, it then
> worked fine. And then I'm using the distribution provided qemu 2.0
> (which should be using the system seabios if it needs one).

There is a difference in the BIOS binaries used by qemu.

qemu 1.7.1 provides a single, 256kB seabios image in the 'bios.bin'
file.  qemu 2.0.0 provides two seabios images: a stripped one 128kB
'bios.bin' and another one: 256kB 'bios-256k.bin'.

I guess that older qemu version could use only a 128kB 'bios.bin' image.

I am not sure what is the seabios image included in the Xen build by
default, but if it is 128k, then this could explain the differences.

Of course, 'the system seabios' may be very distribution-specific and
different from what upstream qemu or Xen provide.

Greets,
        Jacek

_______________________________________________
Xen-users mailing list
[hidden email]
http://lists.xen.org/xen-users