Le 18/05/2016 20:38, Steve Litt a écrit :
> On Wed, 18 May 2016 12:34:22 +0100
> KatolaZ <katolaz@???> wrote:
>
>> On Wed, May 18, 2016 at 12:21:11PM +0100, Rainer Weikusat wrote:
>>> I have to build one of these things as part of packaging 'our'
>>> kernel (for more streamlined installations, ie, without compiling a
>>> dedicated kernel for each). I remember reading various criticisms
>>> about the impenetrable obscurity of the mechanism/ arrangement here
>>> in the past but so far, it seems reasonably straight-forward to me
>>> (the prereq arrangement is a bit weird but that's about it).
>>>
>>> Is there some hidden booby-trap I'm missing?
>> You are right Rainer. There is no trap at all, and no witchcraft
>> involved. Building an initramfs has always been quite
>> straigthforward. You just put on a directory (DIR) the filesystem you
>> want to have available before mounting root,
> I was the guy voicing the criticisms about impenetrable obscurity, but
> I wasn't referring to the process of *making* an initramfs. I was
> referring to the process of *troubleshooting* a system whose initramfs
> does not complete.
>
> I know there are some periscope programs (I think the main one is
> sponsored by Redhat and is part of systemd) with which you can pop up
> in an incomplete initramfs and look around. But it's still a real pain
> in the ass.
>
> SteveT
>

     Hack the initramfs to give the user the chance to escape to an 
interactive shell. I did that to my hand-crafted versions of initramfs 
because it speeds up debugging by a HUGE factor. I imagine it could be 
done on the stock initramfs.

     Didier