[PATCH] ARM: rpi: add SD card environment support

[Andre Heider]

On Tue, Oct 15, 2013 at 02:53:12PM +0200, Sascha Hauer wrote:
> On Tue, Oct 15, 2013 at 01:54:50PM +0200, Andre Heider wrote:
> > Hi,
> > 
> > > +static int rpi_env_init(void)
> > > +{
> > > +	struct stat s;
> > > +	const char *diskdev = "/dev/disk0.0";
> > > +	int ret;
> > > +
> > > +	device_detect_by_name("mci0");
> > > +
> > > +	ret = stat(diskdev, &s);
> > > +	if (ret) {
> > > +		printf("no %s. using default env\n", diskdev);
> > > +		return 0;
> > > +	}
> > > +
> > > +	mkdir("/boot", 0666);
> > > +	ret = mount(diskdev, "fat", "/boot");
> > > +	if (ret) {
> > > +		printf("failed to mount %s\n", diskdev);
> > > +		return 0;
> > > +	}
> > > +
> > > +	default_environment_path = "/boot/barebox.env";
> > > +
> > > +	return 0;
> > > +}
> > 
> > I was solving the same thing just yesterday ;)
> > 
> > What I did was to create a env/init/automount script, what's the advantage of this approach?
> 
> This approach also sets default_environment_path to /boot/barebox.env so
> that loadenv/saveenv works and barebox comes up with the saved
> environment.
> 
> You could probably do similar with scripts doing something like:
> 
> automount -d /boot "detect mci0 && mount /dev/disk0.0 /boot"
> loadenv -s /boot/barebox.env /env
> /env/bin/init
> 
> However, I made the experience that in scripts proper error checking is
> hard to archieve and therefore often skipped. Also with C Code a single
> git commitish makes the system behaviour clear. When instead scripts are
> used which might have been modified by a user, or which might come from
> an older barebox version the situation gets much more complex.
> 
> I generally try to reduce the usage (or better: necessity) of scripts.
> They pretend a high level of flexibility but this comes at the cost of
> reduced consistency across different boards (or even different instances
> of the same board)
> 

Alright, thanks!

Regards,
Andre