elfloader: fixup load from stack on aarch64

It seems that the 'ldp x0, x1, [sp, #-16]!' syntax used to load the
registers stored to the stack before calling the fixup_image_base
function does not in fact load the registers as expected (at least
on my Raspberry Pi 3B).
This causes the dtb address supplied by the bootloader to be lost when
elfloader loads the dtb image which is an issue if a dtb is not built
into the kernel binary image.

In the example usage from arm
(https://developer.arm.com/documentation/
102374/0101/Loads-and-stores---load-pair-and-store-pair)
the 'ldp x0, x1, [sp], #16' is used for poping the stack.
This also seems to also work as expected on my Raspberry Pi 3B
resulting in the dtb address to be loaded as expected in sys_boot.c.

Signed-off-by: Viktor Sannum <[email protected]>

elfloader-tool/src/arch-arm/64/crt0.S

@@ -29,7 +29,7 @@ BEGIN_FUNC(_start)
     bl      fixup_image_base
     mov     x2, x0
     /* restore original arguments for next step */
-    ldp     x0, x1, [sp, #-16]!
+    ldp     x0, x1, [sp], #16
     /* fixup_image_base returns 1 if no need to move */
     cmp     x2, #1
     beq     1f