Re: [SSI-devel] Re: init/main.c changes

[Aneesh K. K.V <ane...@hp...>]

Aneesh Kumar K.V wrote:

> John Byrne wrote:
>
>>
>> I've forgotten the exact reason we need to launch the initrd in the 
>> place we do instead of where the base does it, but there is one. 
>> We've modified the mkinitrd on redhat appropriately to do with this.
>>
>
> I guess we doesn't want init to be started in the normal way at all. 
> Since for us init is a special case. Even if we reach that particualr 
> code section we start /linuxrc instead of init.
>

I looked at the code again. For SSI kernel initrd is not handled by 
handle_initrd(). We handle it in a special case. As david said we 
consider ramdisk as real root. That also means that /linuxrc will not 
get executed at all.  Instead the kernel will go ahead and start 
executing /sbin/init from the real root which in our case will be 
ramdisk. So we put a hack to execute /linuxrc instead of /sbin/init. 
That explains the CONFIG_SSI in  init/main.c:607

-aneesh