This helps significantly.... sys_open tells me that /bin/sh is
opening /dev/ttyS0 - aside from my console being on /dev/ttys03 -
I would have expected hello world to open /dev/ttyS00
Am I missing something glaringly obvious?
FYI - it turns out you can define your console device number by
setting the device field in struct console to the appropriate device
number. This kind of strange knowledge is what happens when
you read the code. You also need to define console.flags
as CON_CONSDEV | CON_PRINTBUFFER.
> Hi all,
> On 13-Sep-98 Thomas Riemer wrote:
> > So've gotten the dz code to tell me that it thinks that it has opened
> > the /dev/console device.
> > In fact at this point, it goes all the way to trying to execute init -
> > what do I have to do to get the kernel to execute the "hello" world
> > program? What's the patch that I need to give it - or what do command
> > line do I have to give the kernel?
> Excuse me, but playing around with three or four different source trees
> simultaneously I somewhat lost track of what I have published so far. :-)
> IIRC you don't need any patches or command line options if you are using the
> latest snapshot from September 5th. The "hello world" program is /bin/sh on
> ramdisk image and will be executed automagically.
> A simple test to check wether it is really executed is an additional printk in
> sys_open(...) in fs/open.c.
> Obviously you need working serial interrupts to see the hello message itself.