[PATCH] ARM: Add lsxhl machine type to mach-types

Russell King - ARM Linux linux at arm.linux.org.uk
Sun Nov 13 16:06:20 EST 2011


On Sun, Nov 13, 2011 at 05:20:22PM +0100, Marek Vasut wrote:
> > On Sun, Nov 13, 2011 at 01:29:56AM +0100, Marek Vasut wrote:
> > > > On Sun, Nov 13, 2011 at 12:40:42AM +0100, Michael Walle wrote:
> > > > > then, i guess, the problem is on the uboot side, at least if they
> > > > > want to support both device tree based and old-fashioned setup code
> > > > > (eg. backported device support.)
> > > > 
> > > > If uboot wants to carry the full list, rather than copying it from
> > > > the kernel source, they should be going straight to the main source
> > > > of the file (which is given inside the file itself.)
> > > > 
> > > > Otherwise, if they want to be purely dependent on the version shipped
> > > > with the kernel, which will have entries for platforms not merged into
> > > > the kernel source deleted from it once they're older than 12 months -
> > > > but which may be in u-boot, they can continue taking it out of the
> > > > kernel tree.  I think that's sub-optimal as they'll see regressions
> > > > from time to time when platforms have been merged into u-boot but not
> > > > the kernel.
> > > 
> > > Hey,
> > > 
> > > U-Boot follows the in-kernel mach-types file. We had a few issues with
> > > boards not building due to the removal of old mach types.
> > 
> > As the kernel version of the file is now specifically customised for
> > the kernel, u-boot should _not_ follow the kernel version anymore, as
> > I pointed out.  u-boot need to change this policy of theirs.
> 
> Why would that be so? If the users want to support the boards and the boards are 
> broken/removed from linux, they just define the mach id by hand. That also makes 
> it easier to detect such boards. It is also a warning that noone gives a crap 
> about such boards and they might as well soon be removed from uboot too.

That is a policy for uboot folk to decide upon.

However, given that the policy for the kernel changed from "lets include
everything" to "lets include only a limited amount" effectively that
change of kernel policy has also changed the uboot policy without any
uboot people apparantly being aware of the change.

If uboot people are happy with the idea that entries get removed, then
fine - but be absolutely clear that I won't be adding entries to the
kernel's file to satisfy the cravings of someone in the uboot camp
complaining that their entry has been deleted for whatever reason.

And note that _sometimes_ entries get deleted not because they aren't
in mainline, but because they were found to be non-conformant and were
missed during the period when I manually checked them.  So, the lack of
an entry in the kernel's file _doesn't_ actually tell uboot why it was
removed - it might even still be in the mainline kernel.

Case in point there: Sheeva eSATA platform.



More information about the linux-arm-kernel mailing list