Having problems with init

[Denys Vlasenko]

On Tuesday 02 November 2010 19:17, Dallas Clement wrote:
> Hello again,
> 
> I found the answer to my own problem.  It turns out that 'init' was
> not reading inittab or running rcS because my PID != 1.  So in other
> words, init needs to be run as the special init process ID 1.  I
> actually already knew that, but never suspected that my PID would not
> be 1 since I was doing my pivot_root from my initrd / linuxrc script.
> 
> It turns out that two options *must* be explicitly added to the linux
> command line for your PID to be 1.
> 
> You should add root=/dev/ram0 and init=/linuxrc if you want your PID
> to be 1 and busybox init to work.
> 
> It would sure be nice if busybox init produced some sort of error
> message if PID != 1.

But it does produce a message:

> > write(2, "BusyBox v1.15.3 (2010-04-23 20:1"..., 41BusyBox v1.15.3
> > (2010-04-23 20:10:06 JST)) = 41
> > write(2, " multi-call binary\n", 19 multi-call binary
> > )    = 19
> > write(2, "\nUsage: ", 8
> > Usage: )                = 8
> > write(2, "init", 4init)                     = 4
> > write(2, " ", 1 )                        = 1
> > write(2, "\n\nInit is the parent of all proc"..., 37
> >
> > Init is the parent of all processes) = 37
> > write(2, "\n\n", 2


Just isn't the clearest one, right? The message doesn't say
"...but you didn't run me as PID 1 process. Bad boy"

How about this fixed behavior? Is it enough? -

# ./busybox init
init: must be run as PID 1


-- 
vda