Re: [PATCH 17/19] bootwrapper: compatibility layer for old U-Boots (a.k.a. cuImage, cuboot)

[David Gibson <david@gibson.dropbear.id.au>]

On Thu, Mar 15, 2007 at 01:02:34PM +1100, David Gibson wrote:
> On Wed, Mar 14, 2007 at 06:59:04PM -0700, Mark A. Greer wrote:
> > On Thu, Mar 15, 2007 at 11:04:45AM +1100, David Gibson wrote:
> > > On Wed, Mar 14, 2007 at 04:23:39PM -0700, Mark A. Greer wrote:
> > > > On Wed, Mar 14, 2007 at 04:48:49PM -0500, Scott Wood wrote:
> > > > > Mark A. Greer wrote:
> > > > > >Are you sure that '_end' (which is the end of the zImage/cuImage)
> > > > > >is safe to use?  If the kernel is large enough (e.g., INITRAMFS)
> > > > > >it will overwrite your dtb when its decompressed and relocated to 0.
> > > > > >You need to grok the elfheader to figure out where the kernel will end
> > > > > >and take the max of that and _end.
> > > > > 
> > > > > Wouldn't it overwrite the bootwrapper itself before overwriting the heap?
> > > > 
> > > > Sure but that doesn't matter--the kernel is running so the bootwrapper's
> > > > life is over but the dtb's life isn't.
> > > 
> > > The bootloader now expands the kernel directly at 0, rather than
> > > allocating space for it, except on platforms that can't (OF).
> > 
> > Well, its not at 0 if you have a platform_ops.vmlinux_alloc()
> > defined which could be anyone.  But I get your point.
> 
> Well, yes, but cuboot doesn't.  The platforms always going to have to
> define a malloc() and vmlinux_alloc() that don't conflict with each
> other.
> 
> > > So we
> > > *would* clobber the bootloader during the bootloader's life if the
> > > kernel was too large.  There's a check for a too-large kernel as we
> > > decompress it.
> > 
> > This seems like a problem to me.  Premably, INITRAMFS will be used
> > more & more often so this check is going to fail more & more often.
> > We'll end up havng to keep moving the addr it loads at further & further
> > back or define platform_ops.vmlinux_alloc() and have the issue I
> > mentioned.
> 
> Err... won't the initramfs image go in the same place as the initrd
> image, not as part of the vmlinux.

Although, speaking of that.  Long term, I think it would be a good
idea for the wrapper script to base the zImage's link address on the
size of the vmlinux.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson