[U-Boot] [PATCH 1/2] image: Support relative-addresses property in FIT images

[Stephen Warren]

Wolfgang Denk wrote at Thursday, October 06, 2011 2:52 PM:
> In message <74CDBE0F657A3D45AFBB94109FB122FF173A2C75C4 at HQMAIL01.nvidia.com> you wrote:
> >
> > > I suggest we define a new image type (say, IH_TYPE_KERNEL_REL) for
> > > this purpose.  This would probably cause the least conflicts.
> >
> > OK. Wouldn't we also need e.g. IH_TYPE_STANDALONE_REL, IH_TYPE_RAMDISK_REL
> > and perhaps other too?
> 
> I don't think so - either these don't have (and use) load and entry
> point addesses (like in the ramdisk case), or they are not easily
> relocatable in the same way as the Linux kernel.

Sorry, I don't quite understand then; in both boot_get_ramdisk() and
boot_get_fdt(), I see calls to image_get_load() and fit_image_get_load().
I guess in the ramdisk case, this just sets rd_load, which isn't used,
but in the FDT case, it sets load_start, which is used to derive load_end
and fdt_blob.

> No, at the moment only the Linux kernel images need to support tis
> property.
> 
> OK, eventually this might be useful for multi-part images, too, but
> these have been deprecated for a long time, and here the user should
> just use a FIT image instead,

-- 
nvpublic