Did you see my qemu r2d memory patch?

It was in 
https://lists.gnu.org/archive/html/qemu-devel/2026-01/msg06318.html and 
probably needs to be reposted as a separate thing, but with it I booted 
an sh4 kernel with over 200 megs of memory, enough to act as a 
reasonable native build node (especially using the distcc trick). No 
kernel patch required, just a .config change setting two symbols it's 
already got to different values.

I need to port the r2d target over to device tree (well, extend the 
turtle device tree stuff to support r2d and its devices) and teach qemu 
to supply a device tree overlay with the current memory map so you don't 
have to match kernel with qemu -m layout (the only way to get the extra 
memory in one contiguous block is to move the START of the memory range, 
which means moving the initrd and command line string locations which 
are relative to start of memory, so the kernel needs to know where to 
expect start of memory and thus needs to be recompiled for the different 
memory amount rather than just autodetecting. Device tree blob location 
is loaded into a register before kernel init...)

Anyway, thought I'd ask what you thought,

Rob