Fix loader link address for tqma8xqp1gb

Seems like this has been regressed for a while unfortunately.
The issue is that there are reserved regions that conflict with
the loader itself which causes the tool to error out for this
platform.

Testing locally, the loader being loaded at the previous address
of 0x80280000 seems to relocate itself to 0x90000000 fine, so
doing `go 0x80280000` seems to still work.

In general this whole loader link address thing is too fragile
and doens't work for really large Microkit images (since what happens
is that the reserved region for the initial task gets larger and
larger until it ends up overlapping with where the loader should go).

Will do this hot-fix for now, and open an issue for making this part
of the loader less fragile in the future.

Signed-off-by: Ivan Velickovic <i.velickovic@unsw.edu.au>

Author: Ivan-Velickovic

build_sdk.py

@@ -167,7 +167,7 @@ class KernelPath:
         name="tqma8xqp1gb",
         arch=KernelArch.AARCH64,
         gcc_cpu="cortex-a35",
-        loader_link_address=0x80280000,
+        loader_link_address=0x90000000,
         smp_cores=4,
         kernel_options={
             "KernelPlatform": "tqma8xqp1gb",

docs/manual.md

@@ -1520,7 +1520,7 @@ For example:
 To use tftp you also need to set the file to load and the memory address to load it to:
 
     => env set bootfile loader.img
-    => env set loadaddr 0x80280000
+    => env set loadaddr 0x90000000
     => env save
 
 The system image generated by the Microkit tool by default is a raw binary file.