[Xen-devel] Problem booting Debian buster on arndale

[Xen-devel] Problem booting Debian buster on arndale

[Ian Jackson]

Hi.  As you know I am working on getting osstest to work with Debian
buster.  This is important because Debian's support for oldstable
(which is what stretch is now) is not always very good.  Relying on
the LTS caused us serious problems before.

One of the issues I encountered is that the Xen and Linux that osstest
built, did not boot on arndale:
  http://logs.test-lab.xenproject.org/osstest/logs/144312/test-armhf-armhf-xl-arndale/info.html

It works on cubietruck:
  http://logs.test-lab.xenproject.org/osstest/logs/144312/test-armhf-armhf-xl-cubietruck/info.html
(Please disregard the guest-start failure; I think that's something
else, and architecture-independent.)

It seems to have hung during boot.  NB that I don't know whether this
is a one-off.

I would appreciate it if someone from the ARM community could take a
look.  If you would like me to set up a repro I on one of our arndales
can do that for you.  I haven't done that yet because it books the
machine out so I want to do it when you say you're read for it...

Thanks,
Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Problem booting Debian buster on arndale

[Julien Grall]

Hi Ian,

On 06/12/2019 16:47, Ian Jackson wrote:
> Hi.  As you know I am working on getting osstest to work with Debian
> buster.  This is important because Debian's support for oldstable
> (which is what stretch is now) is not always very good.  Relying on
> the LTS caused us serious problems before.
> 
> One of the issues I encountered is that the Xen and Linux that osstest
> built, did not boot on arndale:
>    http://logs.test-lab.xenproject.org/osstest/logs/144312/test-armhf-armhf-xl-arndale/info.html
> 
> It works on cubietruck:
>    http://logs.test-lab.xenproject.org/osstest/logs/144312/test-armhf-armhf-xl-cubietruck/info.html
> (Please disregard the guest-start failure; I think that's something
> else, and architecture-independent.)
> 
> It seems to have hung during boot.  NB that I don't know whether this
> is a one-off.

Looking at [1], most of the recent flight have managed to boot Xen on 
the arndale. However, I somehow can't find the flight 144312 in that 
list. Do you have any idea why?

> 
> I would appreciate it if someone from the ARM community could take a
> look.  If you would like me to set up a repro I on one of our arndales
> can do that for you.  I haven't done that yet because it books the
> machine out so I want to do it when you say you're read for it...

This seems to be a one-off, so I would wait to see if it happens again 
before setting up a repro. I will keep an eye on osstest result.

Meanwhile, could you try a more recent flight?

Cheers,

[1] 
http://logs.test-lab.xenproject.org/osstest/results/history/test-armhf-armhf-xl-arndale/xen-unstable.html

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Problem booting Debian buster on arndale

[Ian Jackson]

Julien Grall writes ("Re: Problem booting Debian buster on arndale"):
> On 06/12/2019 16:47, Ian Jackson wrote:
> > It seems to have hung during boot.  NB that I don't know whether this
> > is a one-off.
> 
> Looking at [1], most of the recent flight have managed to boot Xen on 
> the arndale. However, I somehow can't find the flight 144312 in that 
> list. Do you have any idea why?

The flight in question was not on the xen-unstable "branch", and not
officially blessed, because it's part of my buster work.

I found a second instance in the same flight:
  http://logs.test-lab.xenproject.org/osstest/logs/144312/test-armhf-armhf-examine/info.html

I think it must be specific to buster, but that is rather mysterious.
After all it is supposed to be the same kernel and Xen as the current
osstest mainline is using.  Maybe the new compiler is doing something
unexpected.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Problem booting Debian buster on arndale

[Stefano Stabellini]

On Fri, 6 Dec 2019, Ian Jackson wrote:
> Julien Grall writes ("Re: Problem booting Debian buster on arndale"):
> > On 06/12/2019 16:47, Ian Jackson wrote:
> > > It seems to have hung during boot.  NB that I don't know whether this
> > > is a one-off.
> > 
> > Looking at [1], most of the recent flight have managed to boot Xen on 
> > the arndale. However, I somehow can't find the flight 144312 in that 
> > list. Do you have any idea why?
> 
> The flight in question was not on the xen-unstable "branch", and not
> officially blessed, because it's part of my buster work.
> 
> I found a second instance in the same flight:
>   http://logs.test-lab.xenproject.org/osstest/logs/144312/test-armhf-armhf-examine/info.html
> 
> I think it must be specific to buster, but that is rather mysterious.
> After all it is supposed to be the same kernel and Xen as the current
> osstest mainline is using.  Maybe the new compiler is doing something
> unexpected.

That's very interesting and mysterious indeed!

Let me summarize the situation for clarity. Debian Buster can install on
Arndale correctly. It builds on target Xen and Linux, based on the
provided branches. The binaries produced by the build of Xen and Linux
fail to boot correctly on Arndale, even though the same version of Linux
and Xen have been tested successfully previously on Arndale as part of
regular OSSTest runs. Yeah, I cannot make any sense of that :-/

I assume that the device tree used is still also the same?

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Problem booting Debian buster on arndale

[Julien Grall]

Hi,

On 06/12/2019 18:08, Ian Jackson wrote:
> Julien Grall writes ("Re: Problem booting Debian buster on arndale"):
>> On 06/12/2019 16:47, Ian Jackson wrote:
>>> It seems to have hung during boot.  NB that I don't know whether this
>>> is a one-off.
>>
>> Looking at [1], most of the recent flight have managed to boot Xen on
>> the arndale. However, I somehow can't find the flight 144312 in that
>> list. Do you have any idea why?
> 
> The flight in question was not on the xen-unstable "branch", and not
> officially blessed, because it's part of my buster work.
> 
> I found a second instance in the same flight:
>    http://logs.test-lab.xenproject.org/osstest/logs/144312/test-armhf-armhf-examine/info.html
> 
> I think it must be specific to buster, but that is rather mysterious.
> After all it is supposed to be the same kernel and Xen as the current
> osstest mainline is using.  Maybe the new compiler is doing something
> unexpected.

Looking at the command line for Linux, we don't pass clk_ignore_unused. 
Without it, the Linux may disable the clock of the UART if it wasn't 
shared with another device. This would explain the sudden loss of the 
console.

The following line in Osstest/Debian.pm would need to be updated:

      push @xenkopt, "clk_ignore_unused"
             if $ho->{Suite} =~ m/wheezy|jessie|stretch/;

Cheers,

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Problem booting Debian buster on arndale

[Ian Jackson]

Julien Grall writes ("Re: Problem booting Debian buster on arndale"):
> Looking at the command line for Linux, we don't pass clk_ignore_unused. 
> Without it, the Linux may disable the clock of the UART if it wasn't 
> shared with another device. This would explain the sudden loss of the 
> console.
> 
> The following line in Osstest/Debian.pm would need to be updated:
> 
>       push @xenkopt, "clk_ignore_unused"
>              if $ho->{Suite} =~ m/wheezy|jessie|stretch/;

Well spotted.  I will try updating this.  Thanks.

(Is there any possibility that this bug will be fixed at some point?)

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel

Re: [Xen-devel] Problem booting Debian buster on arndale

[Julien Grall]

Hi,

On 09/12/2019 11:39, Ian Jackson wrote:
> Julien Grall writes ("Re: Problem booting Debian buster on arndale"):
>> Looking at the command line for Linux, we don't pass clk_ignore_unused.
>> Without it, the Linux may disable the clock of the UART if it wasn't
>> shared with another device. This would explain the sudden loss of the
>> console.
>>
>> The following line in Osstest/Debian.pm would need to be updated:
>>
>>        push @xenkopt, "clk_ignore_unused"
>>               if $ho->{Suite} =~ m/wheezy|jessie|stretch/;
> 
> Well spotted.  I will try updating this.  Thanks.
> 
> (Is there any possibility that this bug will be fixed at some point?)

We had the same discussion a few months ago for Stretch :).

Someone attempted it to fix it a few years ago but this never reached 
Linux upstream.

 From my perspective, this is a low hanging-fruit as there are more 
concerning bugs that can't be workaround. But the problem is documented, 
so someone could easily pick it up if he/she wanted.

Cheers,

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel