Xen devel - Jun 2013 - Latest Arndale Xen, dom0 kernel stuck on ASIX mac.

Hi,

I was trying to update my Xen version and noticed I needed to use a
newer dom0 kernel too ( took the one from Julien''s arndale branch ). I
got both and compiled them but I can''t get the new kernel to boot. It
keeps getting stuck at ( attached the whole boot log too ):

asix 1-3.2.4:1.0 (unregistered net_device): invalid hw address, using random
asix 1-3.2.4:1.0 eth0: register ''asix'' at
usb-12110000.usb-3.2.4, ASIX
AX88772 USB 2.0 Ethernet, d2:a5:dc:ad:09:b5

I tried setting another mac address in u-boot as stated on the wiki,
one starting with ''00:16:3e:...'' but without result.

So I was wondering if anyone ran into the same issue yet? The kernel
getting stuck seemingly because of a mac address?

Also I noticed that even using saveenv u-boot isn''t keeping my changes
to ethaddr and usbethaddr after a reboot. I''m using U-Boot
2013.01.-rc1-00001-g30dd996 (Apr 29 2013 - 11:09:40) for ARNDALE5250
which still seems to be the one the Linaro guide says to use. Is this
a bug in u-boot?

Thanks,
Sander


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On 06/01/2013 03:31 PM, Sander Bogaert wrote:
> Hi,
> 
> I was trying to update my Xen version and noticed I needed to use a
> newer dom0 kernel too ( took the one from Julien''s arndale branch
). I
> got both and compiled them but I can''t get the new kernel to boot.
It
> keeps getting stuck at ( attached the whole boot log too ):
> 
> asix 1-3.2.4:1.0 (unregistered net_device): invalid hw address, using
random
> asix 1-3.2.4:1.0 eth0: register ''asix'' at
usb-12110000.usb-3.2.4, ASIX
> AX88772 USB 2.0 Ethernet, d2:a5:dc:ad:09:b5
> 
> I tried setting another mac address in u-boot as stated on the wiki,
> one starting with ''00:16:3e:...'' but without result.
> 
> So I was wondering if anyone ran into the same issue yet? The kernel
> getting stuck seemingly because of a mac address?

If you look at few lines before you can see:
Waiting for root device /dev/sda1...

It seems that linux is unable to start the hard drive. Unfortunately, I
never try this configuration with Xen.

Does Xen hang? You can try triple CTRL-a and see if Xen answers.
Is linux/xen receive the SATA interrupt?
> Also I noticed that even using saveenv u-boot isn''t keeping my
changes
> to ethaddr and usbethaddr after a reboot. I''m using U-Boot
> 2013.01.-rc1-00001-g30dd996 (Apr 29 2013 - 11:09:40) for ARNDALE5250
> which still seems to be the one the Linaro guide says to use. Is this
> a bug in u-boot?

I didn''t try recent U-boot. FYI, I''m using U-boot 2012.10-rc1.

Cheers,

Julien

On 3 June 2013 14:28, Julien Grall <julien.grall@linaro.org>
wrote:
> On 06/01/2013 03:31 PM, Sander Bogaert wrote:
>
>> Hi,
>>
>> I was trying to update my Xen version and noticed I needed to use a
>> newer dom0 kernel too ( took the one from Julien''s arndale
branch ). I
>> got both and compiled them but I can''t get the new kernel to
boot. It
>> keeps getting stuck at ( attached the whole boot log too ):
>>
>> asix 1-3.2.4:1.0 (unregistered net_device): invalid hw address, using
random
>> asix 1-3.2.4:1.0 eth0: register ''asix'' at
usb-12110000.usb-3.2.4, ASIX
>> AX88772 USB 2.0 Ethernet, d2:a5:dc:ad:09:b5
>>
>> I tried setting another mac address in u-boot as stated on the wiki,
>> one starting with ''00:16:3e:...'' but without result.
>>
>> So I was wondering if anyone ran into the same issue yet? The kernel
>> getting stuck seemingly because of a mac address?
>
>
> If you look at few lines before you can see:
> Waiting for root device /dev/sda1...
>
> It seems that linux is unable to start the hard drive. Unfortunately, I
> never try this configuration with Xen.
>
> Does Xen hang? You can try triple CTRL-a and see if Xen answers.
> Is linux/xen receive the SATA interrupt?
Xen does not hang. How can I check who got the sata interrupt?
>
>> Also I noticed that even using saveenv u-boot isn''t keeping my
changes
>> to ethaddr and usbethaddr after a reboot. I''m using U-Boot
>> 2013.01.-rc1-00001-g30dd996 (Apr 29 2013 - 11:09:40) for ARNDALE5250
>> which still seems to be the one the Linaro guide says to use. Is this
>> a bug in u-boot?
>
>
> I didn''t try recent U-boot. FYI, I''m using U-boot
2012.10-rc1.
I had to update u-boot because the older one doesn'' work with this SSD
drive. There is a post to the mailinglist about that and updating
u-boot seemed to solve the issue.
>
> Cheers,
>
> Julien

On 4 June 2013 20:48, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
> On 3 June 2013 14:28, Julien Grall <julien.grall@linaro.org> wrote:
>> On 06/01/2013 03:31 PM, Sander Bogaert wrote:
>>
>>> Hi,
>>>
>>> I was trying to update my Xen version and noticed I needed to use a
>>> newer dom0 kernel too ( took the one from Julien''s arndale
branch ). I
>>> got both and compiled them but I can''t get the new kernel
to boot. It
>>> keeps getting stuck at ( attached the whole boot log too ):
>>>
>>> asix 1-3.2.4:1.0 (unregistered net_device): invalid hw address,
using random
>>> asix 1-3.2.4:1.0 eth0: register ''asix'' at
usb-12110000.usb-3.2.4, ASIX
>>> AX88772 USB 2.0 Ethernet, d2:a5:dc:ad:09:b5
>>>
>>> I tried setting another mac address in u-boot as stated on the
wiki,
>>> one starting with ''00:16:3e:...'' but without
result.
>>>
>>> So I was wondering if anyone ran into the same issue yet? The
kernel
>>> getting stuck seemingly because of a mac address?
>>
>>
>> If you look at few lines before you can see:
>> Waiting for root device /dev/sda1...
>>
>> It seems that linux is unable to start the hard drive. Unfortunately, I
>> never try this configuration with Xen.
>>
>> Does Xen hang? You can try triple CTRL-a and see if Xen answers.
>> Is linux/xen receive the SATA interrupt?
>
> Xen does not hang. How can I check who got the sata interrupt?
If you cherry-pick fb61df1 from my tree (julieng/xen-unstable.git),
you will be able to know
which IRQ has been sent to the guest and when. You will need to do
triple CTRL-a then ''q''.

--
Julien Grall

>>> Does Xen hang? You can try triple CTRL-a and see if Xen answers.
>>> Is linux/xen receive the SATA interrupt?
>>
>> Xen does not hang. How can I check who got the sata interrupt?
>
> If you cherry-pick fb61df1 from my tree (julieng/xen-unstable.git),
> you will be able to know
> which IRQ has been sent to the guest and when. You will need to do
> triple CTRL-a then ''q''.
I attached the interrupt information and I got this from the manual:

SPI Port No     ID
109                141       SATAPMEREQ
108                140       SATAPHY

These where the only ones I could find with "SATA" in the description
but I''m not sure these are all the relevant ones. I''m not sure
which
is the ID your patch prints but I''m guessing the SPI Port No. If that
is the case 109 - SATAPMEREQ shows up in there and 108 doesn''t. There
is a 107 but the exynos manual states that one as reserved...

If I''m right about the ID printed does this mean the SATA interrupt
doesn''t make it through Xen to dom0?

Thanks,
Sander


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On 5 June 2013 20:23, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
>>>> Does Xen hang? You can try triple CTRL-a and see if Xen
answers.
>>>> Is linux/xen receive the SATA interrupt?
>>>
>>> Xen does not hang. How can I check who got the sata interrupt?
>>
>> If you cherry-pick fb61df1 from my tree (julieng/xen-unstable.git),
>> you will be able to know
>> which IRQ has been sent to the guest and when. You will need to do
>> triple CTRL-a then ''q''.
>
> I attached the interrupt information and I got this from the manual:
>
> SPI Port No     ID
> 109                141       SATAPMEREQ
> 108                140       SATAPHY
> These where the only ones I could find with "SATA" in the
description
> but I''m not sure these are all the relevant ones. I''m not
sure which
> is the ID your patch prints but I''m guessing the SPI Port No. If
that
> is the case 109 - SATAPMEREQ shows up in there and 108 doesn''t.
There
> is a 107 but the exynos manual states that one as reserved...
> If I''m right about the ID printed does this mean the SATA
interrupt
> doesn''t make it through Xen to dom0?
vgic dump printed ID not SPI. I have just noticed that my current branch (based
on linux upstream 3.9-rc4) doesn''t support exynos ATA.
It seems the driver is not yet upstreamed in the kernel.

I have updated my tree with a branch based on linaro''s tree (which is
a linux 3.10-rc2)
and there is a driver for the exynos ATA. The command to clone and compile the
kernel:

git clone -b linaro-3.10
git://xenbits.xen.org/people/julieng/linux-arm.git linux
cd linux
make arndale_xen_dom0_defconfig
make zImage
make dtbs

The device tree is: arch/arm/boot/dts/exynos5250-arndale.dtb

If it works for you I will update the wiki page and stick to the linaro tree for
the next update.

Cheers,

--
Julien Grall

> vgic dump printed ID not SPI. I have just noticed that my current branch
(based
> on linux upstream 3.9-rc4) doesn''t support exynos ATA.
> It seems the driver is not yet upstreamed in the kernel.
>
> I have updated my tree with a branch based on linaro''s tree (which
is
> a linux 3.10-rc2)
> and there is a driver for the exynos ATA. The command to clone and compile
the
> kernel:
>
> git clone -b linaro-3.10
> git://xenbits.xen.org/people/julieng/linux-arm.git linux
> cd linux
> make arndale_xen_dom0_defconfig
> make zImage
> make dtbs
>
> The device tree is: arch/arm/boot/dts/exynos5250-arndale.dtb
>
> If it works for you I will update the wiki page and stick to the linaro
tree for
> the next update.
>
> Cheers,
>
> --
> Julien Grall
No luck... I attached the crashing dom0 log.
Please note I used the  arndale_xen_dom0_defconfig make target
guessing that''s the one you intended to write :-)

This is a different crash however so it did change things...

It seems like the sda drive isn''t detected at all? Maybe the driver is
broke? In the dts I just change root=mmc... into root=/dev/sda1 which
works for my older kernel/xen versions.

Thanks,
Sander

On 6 June 2013 10:51, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
>> vgic dump printed ID not SPI. I have just noticed that my current
branch (based
>> on linux upstream 3.9-rc4) doesn''t support exynos ATA.
>> It seems the driver is not yet upstreamed in the kernel.
>>
>> I have updated my tree with a branch based on linaro''s tree
(which is
>> a linux 3.10-rc2)
>> and there is a driver for the exynos ATA. The command to clone and
compile the
>> kernel:
>>
>> git clone -b linaro-3.10
>> git://xenbits.xen.org/people/julieng/linux-arm.git linux
>> cd linux
>> make arndale_xen_dom0_defconfig
>> make zImage
>> make dtbs
>>
>> The device tree is: arch/arm/boot/dts/exynos5250-arndale.dtb
>>
>> If it works for you I will update the wiki page and stick to the linaro
tree for
>> the next update.
>>
>> Cheers,
>>
>> --
>> Julien Grall
>
> No luck... I attached the crashing dom0 log.
> Please note I used the  arndale_xen_dom0_defconfig make target
> guessing that''s the one you intended to write :-)
>
> This is a different crash however so it did change things...
>
> It seems like the sda drive isn''t detected at all? Maybe the
driver is
> broke? In the dts I just change root=mmc... into root=/dev/sda1 which
> works for my older kernel/xen versions.
>
> Thanks,
> Sander
Sorry for spamming, forgot to attach....


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On 06/06/2013 09:52 AM, Sander Bogaert wrote:

>> No luck... I attached the crashing dom0 log.
>> Please note I used the  arndale_xen_dom0_defconfig make target
>> guessing that''s the one you intended to write :-)
>>
>> This is a different crash however so it did change things...
>>
>> It seems like the sda drive isn''t detected at all? Maybe the
driver is
>> broke? In the dts I just change root=mmc... into root=/dev/sda1 which
>> works for my older kernel/xen versions.

[    5.366313] exynos-sata 122f0000.sata: failed to get sata phy for port 0

...
[    6.084555] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
[    6.084606] platform 122f0000.sata: Driver exynos-sata requests probe
deferral
[    6.085869] input: gpio_keys.5 as /devices/gpio_keys.5/input/input0
[    6.086535] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
[    6.086584] platform 122f0000.sata: Driver exynos-sata requests probe
deferral

Could you try the same kernel on bare metal and see what happens?

Cheers,

-- 
Julien

On 6 June 2013 12:37, Julien Grall <julien.grall@linaro.org>
wrote:
> On 06/06/2013 09:52 AM, Sander Bogaert wrote:
>
>
>>> No luck... I attached the crashing dom0 log.
>>> Please note I used the  arndale_xen_dom0_defconfig make target
>>> guessing that''s the one you intended to write :-)
>>>
>>> This is a different crash however so it did change things...
>>>
>>> It seems like the sda drive isn''t detected at all? Maybe
the driver is
>>> broke? In the dts I just change root=mmc... into root=/dev/sda1
which
>>> works for my older kernel/xen versions.
>
>
> [    5.366313] exynos-sata 122f0000.sata: failed to get sata phy for port 0
>
> ...
> [    6.084555] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
> [    6.084606] platform 122f0000.sata: Driver exynos-sata requests probe
deferral
> [    6.085869] input: gpio_keys.5 as /devices/gpio_keys.5/input/input0
> [    6.086535] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
> [    6.086584] platform 122f0000.sata: Driver exynos-sata requests probe
deferral
>
> Could you try the same kernel on bare metal and see what happens?
>
all seems to be fine for native Linux, the partitions on the ssd get
detected so I''m guessing it must be a Xen issue. It''s an issue
I
didn''t have with older versions... ( end of April ).

I attached the a log of booting the kernel native. It does have a
fault at the end but that''s because sda5 doesn''t exist. I
didn''t want
to possibly mess up my working sda1 partition.

Regards,
Sander


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On 06/07/2013 03:40 PM, Sander Bogaert wrote:
> On 6 June 2013 12:37, Julien Grall <julien.grall@linaro.org> wrote:
>> On 06/06/2013 09:52 AM, Sander Bogaert wrote:
>>
>>
>>>> No luck... I attached the crashing dom0 log.
>>>> Please note I used the  arndale_xen_dom0_defconfig make target
>>>> guessing that''s the one you intended to write :-)
>>>>
>>>> This is a different crash however so it did change things...
>>>>
>>>> It seems like the sda drive isn''t detected at all?
Maybe the driver is
>>>> broke? In the dts I just change root=mmc... into root=/dev/sda1
which
>>>> works for my older kernel/xen versions.
>>
>>
>> [    5.366313] exynos-sata 122f0000.sata: failed to get sata phy for
port 0
>>
>> ...
>> [    6.084555] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
>> [    6.084606] platform 122f0000.sata: Driver exynos-sata requests
probe deferral
>> [    6.085869] input: gpio_keys.5 as /devices/gpio_keys.5/input/input0
>> [    6.086535] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
>> [    6.086584] platform 122f0000.sata: Driver exynos-sata requests
probe deferral
>>
>> Could you try the same kernel on bare metal and see what happens?
>>
> 
> all seems to be fine for native Linux, the partitions on the ssd get
> detected so I''m guessing it must be a Xen issue. It''s an
issue I
> didn''t have with older versions... ( end of April ).
> 
> I attached the a log of booting the kernel native. It does have a
> fault at the end but that''s because sda5 doesn''t exist. I
didn''t want
> to possibly mess up my working sda1 partition.

Thanks, I think I have found this issue. The i2c tries to register
cpufreq callback. I disabled both cpufreq and cpuidle to avoid specific
board callback.

Could you try this small patch?

==================================================================diff --git
a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
index bcf4e76..3a8cee3 100644
--- a/arch/arm/xen/enlighten.c
+++ b/arch/arm/xen/enlighten.c
@@ -299,7 +299,6 @@ static int __init xen_pm_init(void)
         * cpu idle and cpu freq
         */
        disable_cpuidle();
-       disable_cpufreq();

        return 0;
 }
==============================================================
Cheers,

-- 
Julien

On 7 June 2013 17:03, Julien Grall <julien.grall@linaro.org>
wrote:
> On 06/07/2013 03:40 PM, Sander Bogaert wrote:
>
>> On 6 June 2013 12:37, Julien Grall <julien.grall@linaro.org>
wrote:
>>> On 06/06/2013 09:52 AM, Sander Bogaert wrote:
>>>
>>>
>>>>> No luck... I attached the crashing dom0 log.
>>>>> Please note I used the  arndale_xen_dom0_defconfig make
target
>>>>> guessing that''s the one you intended to write :-)
>>>>>
>>>>> This is a different crash however so it did change
things...
>>>>>
>>>>> It seems like the sda drive isn''t detected at all?
Maybe the driver is
>>>>> broke? In the dts I just change root=mmc... into
root=/dev/sda1 which
>>>>> works for my older kernel/xen versions.
>>>
>>>
>>> [    5.366313] exynos-sata 122f0000.sata: failed to get sata phy
for port 0
>>>
>>> ...
>>> [    6.084555] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
>>> [    6.084606] platform 122f0000.sata: Driver exynos-sata requests
probe deferral
>>> [    6.085869] input: gpio_keys.5 as
/devices/gpio_keys.5/input/input0
>>> [    6.086535] exynos-sata 122f0000.sata: forcing PORTS_IMPL to 0x1
>>> [    6.086584] platform 122f0000.sata: Driver exynos-sata requests
probe deferral
>>>
>>> Could you try the same kernel on bare metal and see what happens?
>>>
>>
>> all seems to be fine for native Linux, the partitions on the ssd get
>> detected so I''m guessing it must be a Xen issue. It''s
an issue I
>> didn''t have with older versions... ( end of April ).
>>
>> I attached the a log of booting the kernel native. It does have a
>> fault at the end but that''s because sda5 doesn''t
exist. I didn''t want
>> to possibly mess up my working sda1 partition.
>
>
> Thanks, I think I have found this issue. The i2c tries to register
> cpufreq callback. I disabled both cpufreq and cpuidle to avoid specific
> board callback.
>
> Could you try this small patch?
>
> ==================================================================> diff
--git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
> index bcf4e76..3a8cee3 100644
> --- a/arch/arm/xen/enlighten.c
> +++ b/arch/arm/xen/enlighten.c
> @@ -299,7 +299,6 @@ static int __init xen_pm_init(void)
>          * cpu idle and cpu freq
>          */
>         disable_cpuidle();
> -       disable_cpufreq();
>
>         return 0;
>  }
> ==============================================================>
Applying these changes allow me to boot but I end up with a read-only
filesystem so something must still be wrong.

I first removed both disable_* calls the way you described and not the
way you showed in the patch. After that I only removed cpufreq as the
patch shows. Both tries have the same result, I attached a log.

Normally I''d think this is a Linux / configuration issue but since the
same filesystem works fine with the older Xen I suspect Xen is
involved.

Thanks for the patch!


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On 06/07/2013 07:09 PM, Sander Bogaert wrote:
> On 7 June 2013 17:03, Julien Grall <julien.grall@linaro.org> wrote:
>> On 06/07/2013 03:40 PM, Sander Bogaert wrote:
>>
>>> On 6 June 2013 12:37, Julien Grall <julien.grall@linaro.org>
wrote:
>>>> On 06/06/2013 09:52 AM, Sander Bogaert wrote:
>>>>
>>>>
>>>>>> No luck... I attached the crashing dom0 log.
>>>>>> Please note I used the  arndale_xen_dom0_defconfig make
target
>>>>>> guessing that''s the one you intended to write
:-)
>>>>>>
>>>>>> This is a different crash however so it did change
things...
>>>>>>
>>>>>> It seems like the sda drive isn''t detected at
all? Maybe the driver is
>>>>>> broke? In the dts I just change root=mmc... into
root=/dev/sda1 which
>>>>>> works for my older kernel/xen versions.
>>>>
>>>>
>>>> [    5.366313] exynos-sata 122f0000.sata: failed to get sata
phy for port 0
>>>>
>>>> ...
>>>> [    6.084555] exynos-sata 122f0000.sata: forcing PORTS_IMPL to
0x1
>>>> [    6.084606] platform 122f0000.sata: Driver exynos-sata
requests probe deferral
>>>> [    6.085869] input: gpio_keys.5 as
/devices/gpio_keys.5/input/input0
>>>> [    6.086535] exynos-sata 122f0000.sata: forcing PORTS_IMPL to
0x1
>>>> [    6.086584] platform 122f0000.sata: Driver exynos-sata
requests probe deferral
>>>>
>>>> Could you try the same kernel on bare metal and see what
happens?
>>>>
>>>
>>> all seems to be fine for native Linux, the partitions on the ssd
get
>>> detected so I''m guessing it must be a Xen issue.
It''s an issue I
>>> didn''t have with older versions... ( end of April ).
>>>
>>> I attached the a log of booting the kernel native. It does have a
>>> fault at the end but that''s because sda5 doesn''t
exist. I didn''t want
>>> to possibly mess up my working sda1 partition.
>>
>>
>> Thanks, I think I have found this issue. The i2c tries to register
>> cpufreq callback. I disabled both cpufreq and cpuidle to avoid specific
>> board callback.
>>
>> Could you try this small patch?
>>
>>
==================================================================>> diff
--git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c
>> index bcf4e76..3a8cee3 100644
>> --- a/arch/arm/xen/enlighten.c
>> +++ b/arch/arm/xen/enlighten.c
>> @@ -299,7 +299,6 @@ static int __init xen_pm_init(void)
>>          * cpu idle and cpu freq
>>          */
>>         disable_cpuidle();
>> -       disable_cpufreq();
>>
>>         return 0;
>>  }
>> ==============================================================>>
> 
> Applying these changes allow me to boot but I end up with a read-only
> filesystem so something must still be wrong.
> 
> I first removed both disable_* calls the way you described and not the
> way you showed in the patch. After that I only removed cpufreq as the
> patch shows. Both tries have the same result, I attached a log.
> 

Is rw on the linux command line change something? Otherwise, you can try
to boot up to login shell on bare metal and see if it''s still
read-only.

Cheers,

-- 
Julien

> Is rw on the linux command line change something? Otherwise, you can try
> to boot up to login shell on bare metal and see if it''s still
read-only.
>
> Cheers,
>
> --
> Julien
Thanks, after doing this and some other stuff ( hostname was gone
resulting in trouble and some problems with xencommons daemon ) I''m
only left with:

# xl list
xc: error: Could not obtain handle on privileged command interface (2
= No such file or directory): Internal error
libxl: error: libxl.c:94:libxl_ctx_alloc: cannot open libxc handle: No
such file or directory
cannot init xl context

The solutions suggested by googling didn''t do the trick. Am I supposed
to get newer tools? There is no /proc/xen so no privcmd interface I
guess..

All these changes seem rather strange, possibly pointing to a bigger
underlying issue. Did you test the kernelconfig with the latest linaro
kernel?

Sander

On 7 June 2013 20:58, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
> # xl list
> xc: error: Could not obtain handle on privileged command interface (2
> = No such file or directory): Internal error
> libxl: error: libxl.c:94:libxl_ctx_alloc: cannot open libxc handle: No
> such file or directory
> cannot init xl context
>
> The solutions suggested by googling didn''t do the trick. Am I
supposed
> to get newer tools? There is no /proc/xen so no privcmd interface I
> guess..
The xencommons initscript should mount /proc/xen. What happens if you
try to mount manually?

mount -t xenfs xen /proc/xen
> All these changes seem rather strange, possibly pointing to a bigger
> underlying issue. Did you test the kernelconfig with the latest linaro
> kernel?
If you are talking about my branch linaro-3.10, I''m using it every day
on the arndale board and the versatile express.

--
Julien Grall

On 7 June 2013 22:25, Julien Grall <julien.grall@linaro.org>
wrote:
> On 7 June 2013 20:58, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
>> # xl list
>> xc: error: Could not obtain handle on privileged command interface (2
>> = No such file or directory): Internal error
>> libxl: error: libxl.c:94:libxl_ctx_alloc: cannot open libxc handle: No
>> such file or directory
>> cannot init xl context
>>
>> The solutions suggested by googling didn''t do the trick. Am I
supposed
>> to get newer tools? There is no /proc/xen so no privcmd interface I
>> guess..
>
> The xencommons initscript should mount /proc/xen. What happens if you
> try to mount manually?
>
> mount -t xenfs xen /proc/xen
I tried that, it doesn''t exist. To be exact:

mount: mount point /proc/xen does not exist
>
>> All these changes seem rather strange, possibly pointing to a bigger
>> underlying issue. Did you test the kernelconfig with the latest linaro
>> kernel?
>
> If you are talking about my branch linaro-3.10, I''m using it every
day
> on the arndale board and the versatile express.
>
> --
> Julien Grall

On 7 June 2013 21:35, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
> On 7 June 2013 22:25, Julien Grall <julien.grall@linaro.org> wrote:
>> On 7 June 2013 20:58, Sander Bogaert
<sander.bogaert@elis.ugent.be> wrote:
>>> # xl list
>>> xc: error: Could not obtain handle on privileged command interface
(2
>>> = No such file or directory): Internal error
>>> libxl: error: libxl.c:94:libxl_ctx_alloc: cannot open libxc handle:
No
>>> such file or directory
>>> cannot init xl context
>>>
>>> The solutions suggested by googling didn''t do the trick.
Am I supposed
>>> to get newer tools? There is no /proc/xen so no privcmd interface I
>>> guess..
>>
>> The xencommons initscript should mount /proc/xen. What happens if you
>> try to mount manually?
>>
>> mount -t xenfs xen /proc/xen
>
> I tried that, it doesn''t exist. To be exact:
>
> mount: mount point /proc/xen does not exist
You are right,  I have a mismatch between my .config and
arndale_xen_dom0_defconfig.

I updated my branch linaro-3.10 with the cpufreq patch and add XENFS
in the configuration
file.

Thanks,

--
Julien Grall

On 7 June 2013 22:56, Julien Grall <julien.grall@linaro.org>
wrote:
> On 7 June 2013 21:35, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
>> On 7 June 2013 22:25, Julien Grall <julien.grall@linaro.org>
wrote:
>>> On 7 June 2013 20:58, Sander Bogaert
<sander.bogaert@elis.ugent.be> wrote:
>>>> # xl list
>>>> xc: error: Could not obtain handle on privileged command
interface (2
>>>> = No such file or directory): Internal error
>>>> libxl: error: libxl.c:94:libxl_ctx_alloc: cannot open libxc
handle: No
>>>> such file or directory
>>>> cannot init xl context
>>>>
>>>> The solutions suggested by googling didn''t do the
trick. Am I supposed
>>>> to get newer tools? There is no /proc/xen so no privcmd
interface I
>>>> guess..
>>>
>>> The xencommons initscript should mount /proc/xen. What happens if
you
>>> try to mount manually?
>>>
>>> mount -t xenfs xen /proc/xen
>>
>> I tried that, it doesn''t exist. To be exact:
>>
>> mount: mount point /proc/xen does not exist
>
> You are right,  I have a mismatch between my .config and
> arndale_xen_dom0_defconfig.
>
> I updated my branch linaro-3.10 with the cpufreq patch and add XENFS
> in the configuration
> file.
>
Again no luck. Tried some different kernel config options to in order to fix it.

First time around still proc related I guess (hangs after):

warning: failed to read mtab
warning: failed to read mtab
mount: can''t find /run in /etc/fstab or /etc/mtab
warning: failed to read mtab
Segmentation fault
warning: failed to read mtab
mount: can''t find /run/shm in /etc/fstab or /etc/mtab
[warn] Filesystem type ''devpts'' is not supported. Skipping
mount. ... (warning).
/etc/init.d/mtab.sh: 48: /etc/init.d/mtab.sh: cannot open
/proc/mounts: No such file
[ ok ] Mounting local filesystems...done.
[ ok ] Activating swapfile swap...done.
df: Warning: cannot read table of mounted file systems: No such file
or directory
[....] Setting kernel variables ...[....] Cleaning up temporary files...

Added to fstab: proc /proc proc rw,nosuid,nodev,noexec,relatime 0 0

Segmentation fault
[ ok ] Starting enhanced syslogd: rsyslogd[....] Starting periodic
command scheduler: cron.
[ ok ] Starting OpenBSD Secure Shell server: sshd.
[FAIL] startpar: service(s) returned failure: rsyslog ... failed!

Now /proc is empty and hostname gone ( like before ). Rebooted to get
a boot without the segfault ( guessing this is the know problem with
Xen atm ), same issue.

I tried this with non modular kernel config ( don''t think I had them
before ) and with modules, doesn''t make a difference.

Sander

On 7 June 2013 22:51, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
> On 7 June 2013 22:56, Julien Grall <julien.grall@linaro.org> wrote:
>> On 7 June 2013 21:35, Sander Bogaert
<sander.bogaert@elis.ugent.be> wrote:
>>> On 7 June 2013 22:25, Julien Grall <julien.grall@linaro.org>
wrote:
>>>> On 7 June 2013 20:58, Sander Bogaert
<sander.bogaert@elis.ugent.be> wrote:
>>>>> # xl list
>>>>> xc: error: Could not obtain handle on privileged command
interface (2
>>>>> = No such file or directory): Internal error
>>>>> libxl: error: libxl.c:94:libxl_ctx_alloc: cannot open libxc
handle: No
>>>>> such file or directory
>>>>> cannot init xl context
>>>>>
>>>>> The solutions suggested by googling didn''t do the
trick. Am I supposed
>>>>> to get newer tools? There is no /proc/xen so no privcmd
interface I
>>>>> guess..
>>>>
>>>> The xencommons initscript should mount /proc/xen. What happens
if you
>>>> try to mount manually?
>>>>
>>>> mount -t xenfs xen /proc/xen
>>>
>>> I tried that, it doesn''t exist. To be exact:
>>>
>>> mount: mount point /proc/xen does not exist
>>
>> You are right,  I have a mismatch between my .config and
>> arndale_xen_dom0_defconfig.
>>
>> I updated my branch linaro-3.10 with the cpufreq patch and add XENFS
>> in the configuration
>> file.
>>
>
> Again no luck. Tried some different kernel config options to in order to
fix it.
Do you have all theses configuration variable enabled on your .config?
CONFIG_XEN_DOM0=y
CONFIG_XEN=y
CONFIG_XEN_BLKDEV_FRONTEND=y
CONFIG_XEN_BLKDEV_BACKEND=y
CONFIG_XEN_NETDEV_FRONTEND=y
CONFIG_XEN_NETDEV_BACKEND=y
CONFIG_HVC_XEN=y
CONFIG_HVC_XEN_FRONTEND=y
CONFIG_XEN_DEV_EVTCHN=y
CONFIG_XEN_BACKEND=y
CONFIG_XENFS=y
CONFIG_XEN_COMPAT_XENFS=y
CONFIG_XEN_SYS_HYPERVISOR=y
CONFIG_XEN_XENBUS_FRONTEND=y
CONFIG_XEN_GNTDEV=m
CONFIG_XEN_GRANT_DEV_ALLOC=m
CONFIG_XEN_PRIVCMD=y

/proc/xen is created by CONFIG_XEN_COMPAT_XENFS. I won''t have access to
an
arndale board until monday, so right now I can''t give a try.
> Now /proc is empty and hostname gone ( like before ). Rebooted to get
> a boot without the segfault ( guessing this is the know problem with
> Xen atm ), same issue.
To avoid segfault you can add on both command line:
  - Xen: dom0_max_vcpus = 1
  - Linux: nosmp

It will disable SMP but should avoid segfault. I hope linux will be able to
boot on your arndale. On mine, linux 3.10 does something wrong with the
mmc driver.

BTW, on linux 3.7 did you have SMP support on dom0?

-- 
Julien Grall

On 8 June 2013 00:44, Julien Grall <julien.grall@linaro.org>
wrote:
> On 7 June 2013 22:51, Sander Bogaert <sander.bogaert@elis.ugent.be>
wrote:
>> On 7 June 2013 22:56, Julien Grall <julien.grall@linaro.org>
wrote:
>>> On 7 June 2013 21:35, Sander Bogaert
<sander.bogaert@elis.ugent.be> wrote:
>>>> On 7 June 2013 22:25, Julien Grall
<julien.grall@linaro.org> wrote:
>>>>> On 7 June 2013 20:58, Sander Bogaert
<sander.bogaert@elis.ugent.be> wrote:
>>>>>> # xl list
>>>>>> xc: error: Could not obtain handle on privileged
command interface (2
>>>>>> = No such file or directory): Internal error
>>>>>> libxl: error: libxl.c:94:libxl_ctx_alloc: cannot open
libxc handle: No
>>>>>> such file or directory
>>>>>> cannot init xl context
>>>>>>
>>>>>> The solutions suggested by googling didn''t do
the trick. Am I supposed
>>>>>> to get newer tools? There is no /proc/xen so no privcmd
interface I
>>>>>> guess..
>>>>>
>>>>> The xencommons initscript should mount /proc/xen. What
happens if you
>>>>> try to mount manually?
>>>>>
>>>>> mount -t xenfs xen /proc/xen
>>>>
>>>> I tried that, it doesn''t exist. To be exact:
>>>>
>>>> mount: mount point /proc/xen does not exist
>>>
>>> You are right,  I have a mismatch between my .config and
>>> arndale_xen_dom0_defconfig.
>>>
>>> I updated my branch linaro-3.10 with the cpufreq patch and add
XENFS
>>> in the configuration
>>> file.
>>>
>>
>> Again no luck. Tried some different kernel config options to in order
to fix it.
>
> Do you have all theses configuration variable enabled on your .config?
> CONFIG_XEN_DOM0=y
> CONFIG_XEN=y
> CONFIG_XEN_BLKDEV_FRONTEND=y
> CONFIG_XEN_BLKDEV_BACKEND=y
> CONFIG_XEN_NETDEV_FRONTEND=y
> CONFIG_XEN_NETDEV_BACKEND=y
> CONFIG_HVC_XEN=y
> CONFIG_HVC_XEN_FRONTEND=y
> CONFIG_XEN_DEV_EVTCHN=y
> CONFIG_XEN_BACKEND=y
> CONFIG_XENFS=y
> CONFIG_XEN_COMPAT_XENFS=y
> CONFIG_XEN_SYS_HYPERVISOR=y
> CONFIG_XEN_XENBUS_FRONTEND=y
> CONFIG_XEN_GNTDEV=m
> CONFIG_XEN_GRANT_DEV_ALLOC=m
> CONFIG_XEN_PRIVCMD=y
>
> /proc/xen is created by CONFIG_XEN_COMPAT_XENFS. I won''t have
access to an
> arndale board until monday, so right now I can''t give a try.
I''m testing with the config from your latest push, all seem to be
there:

% grep XEN .config
CONFIG_XEN_DOM0=y
CONFIG_XEN=y
CONFIG_XEN_BLKDEV_FRONTEND=y
CONFIG_XEN_BLKDEV_BACKEND=y
CONFIG_XEN_NETDEV_FRONTEND=y
CONFIG_XEN_NETDEV_BACKEND=y
CONFIG_INPUT_XEN_KBDDEV_FRONTEND=y
CONFIG_HVC_XEN=y
CONFIG_HVC_XEN_FRONTEND=y
CONFIG_XEN_FBDEV_FRONTEND=y
CONFIG_XEN_DEV_EVTCHN=y
CONFIG_XEN_BACKEND=y
CONFIG_XENFS=y
CONFIG_XEN_COMPAT_XENFS=y
CONFIG_XEN_SYS_HYPERVISOR=y
CONFIG_XEN_XENBUS_FRONTEND=y
CONFIG_XEN_GNTDEV=m
CONFIG_XEN_GRANT_DEV_ALLOC=m
CONFIG_XEN_PRIVCMD=y
# CONFIG_DEBUG_XEN is not set

I''ll try some other configurations later this weekend. Are you also
using Debian for a dom0 fs?
>
>> Now /proc is empty and hostname gone ( like before ). Rebooted to get
>> a boot without the segfault ( guessing this is the know problem with
>> Xen atm ), same issue.
>
> To avoid segfault you can add on both command line:
>   - Xen: dom0_max_vcpus = 1
>   - Linux: nosmp
>
> It will disable SMP but should avoid segfault. I hope linux will be able to
> boot on your arndale. On mine, linux 3.10 does something wrong with the
> mmc driver.
Thanks, I read about this in the other mailthread.
>
> BTW, on linux 3.7 did you have SMP support on dom0?
I did and I also had the random segfaults in case you were wondering :-)

Sander

On 06/08/2013 08:57 AM, Sander Bogaert wrote:
> 
> I''ll try some other configurations later this weekend. Are you
also
> using Debian for a dom0 fs?

No. I''m using an OpenSuse dom0.

-- 
Julien