Re: [Clonezilla-live] Problems with live-getty and live-netdev

[Santiago B. <san...@gm...>]

On Thu, Dec 24, 2009 at 4:46 AM, Steven Shiau <st...@nc...> wrote:
>
>
> Santiago Bruno wrote:
>>
>> On Wed, Dec 23, 2009 at 11:59 AM, Steven Shiau <st...@nc...> wrote:
>>>
>>> Santiago Bruno wrote:
>>>>
>>>> On Mon, Dec 14, 2009 at 4:47 AM, Steven Shiau <st...@nc...>
>>>> wrote:
>>>>
>>>>> Hi,
>>>>> Thanks for the bug reports. I reply your question sin the following.
>>>>>
>>>>> Santiago Bruno wrote:
>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> I'm trying to use Clonezilla live in a machine that will boot using
>>>>>> PXE through its interface eth1. This machine can redirect the console
>>>>>> using Serial over Lan with its own hardware and is configured to do
>>>>>> so.
>>>>>>
>>>>>> I added live-getty and console=ttyS0,115200n8 to the kernel command
>>>>>> line but I'm having a strange problem.
>>>>>>
>>>>>> When I'm looking at the serial console, after booting clonezilla,
>>>>>> there is output at the console as expected, but some output also goes
>>>>>> to the VGA. And it's not the same output. Is like there are two
>>>>>> different clonezilla instances in tty1 and ttyS0. They even start
>>>>>> cloning the same partitions at diferent times messing all up. At the
>>>>>> end one can go to the command line on both consoles and evidently they
>>>>>> are two different processes.
>>>>>>
>>>>> How did you configure your clonezilla live ? Please show us the config
>>>>> files, e.g. syslinux.cfg... So it's easier for me to reproduce the
>>>>> problem
>>>>> then we can fix this problem.
>>>>>
>>>> Well, I'm using some custom scripts, but now I reduced the
>>>> configuration to try to use a script that does not exists to see if
>>>> the problem still happened so it could be easier to reproduce.
>>>>
>>>> The pxe file for the machine would be like this:
>>>>
>>>> default Clonezilla
>>>> label Clonezilla
>>>> kernel /clonezilla/vmlinuz
>>>> append initrd=/clonezilla/initrd.img boot=live live-getty
>>>> console=ttyS0,115200n8 union=aufs noswap noprompt nosplash
>>>> live-netdev="eth1"
>>>> fetch=tftp://<SERVER_IP>/clonezilla/filesystem.squashfs
>>>> ocs_live_keymap="NONE" ocs_live_batch="yes" ocs_lang="en_US.UTF-8"
>>>> ocs_live_run="/blablablabla"
>>>>
>>>> and what happens when booting is the following.
>>>>
>>>> VGA starts displaying the pxe boot, "Decompressing Linux..."
>>>> decompressing the initrd, and the last lines are
>>>> "Ready
>>>> Probing EDD (...) ok"
>>>>
>>>> It stops there, and then, text starts coming out through ttyS0
>>>> instead. squashfs is decompressed, init starts, I can see that it
>>>> tries to set up eth0 through dhcp because I see some "DHCPDISCOVER..."
>>>> lines until it times out.
>>>> Then I think it sets up eth1 very fast and then the problem starts.
>>>> I think it is when the ocs scripts begin executing.
>>>> The VGA changes from the typical console text to another font,
>>>> probably a framebuffer, and some scripts begin executing both in tty1
>>>> and ttyS0
>>>> obviously they fail inmediately because there is no blablablabla
>>>> script, but I get the prompt to select what to do next in both
>>>> terminals and they are different processes.
>>>> If I type "tty" one gives /dev/tty1 and the other /dev/ttyS0.
>>>> I would expect that everything happens in ttyS0 if I specified
>>>> console=ttyS0 on the command line, or to have an option to do that. Is
>>>> it possible? I would expect in that case to keep tty1 blocked or in a
>>>> login prompt.
>>>>
>>>>
>>> Yes. Here I think the best solution is in your customized script, add one
>>> line in the beginning:
>>> ================================
>>> [ "$(tty)" != "/dev/ttyS0" ] && exit 1
>>> ================================
>>>
>>> Then the rest of your script will only be run in the /dev/ttyS0.
>>>
>>> Hope this helps.
>>
>> Yes, I considered implementing exactly that workaround. The only
>> problem is that I will be using the script as part of a bigger system
>> with an user interface where the user may choose to redirect the
>> output or not. So I will have two different customized scripts, one
>> for redirecting and another for not redirecting, and selecting which
>> script to use through the generated pxe config file.
>> I thought of reporting the problem so I had to work less :P
>> But this should work.
>>
>> Thank you!
>>
>> Santiago.
>
> Another possibility is we can implement another ocs option to force
> $ocs_live_run to be run in specific tty, e.g. ttyS0. Now it will be run in
> tty1 and ttyS0.
> E.g. with boot parameter
> ocs_live_tty="/dev/ttyS0"
> then the $ocs_live_run will be run in ttyS0 only. Not in tty1.
>
> This should work. Or maybe you have better idea?

That would be something nice to have. I have no other idea. But I
don´t know how complicated it will be to implement it. I can live with
the workaround you suggested ;)
But let me know if you implement it in some future release or if I can
help in some way, like testing.

Cheers,

Santiago.