The encoding of the sub instruction used to handle CONFIG_SYS_MALLOC_F_LEN
can only accept certain values, and the set of acceptable values differs
between the AArch32 and AArch64 instructions sets. The default value of
CONFIG_SYS_MALLOC_F_LEN works with either ISA. Tegra uses a non-default
value that can only be encoded in the AArch32 ISA. Fix the AArch64 crt0
assembly so it can handle completely arbitrary values.
Signed-off-by: Thierry Reding <treding@nvidia.com>
[twarren: trimmed Thierry's patch to remove changes already present]
Signed-off-by: Tom Warren <twarren@nvidia.com>
[swarren, cleaned up patch, wrote description, re-wrote subject]
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
cmp x0, x18
b.gt zero_gd
#if defined(CONFIG_SYS_MALLOC_F_LEN)
- sub x0, x18, #CONFIG_SYS_MALLOC_F_LEN
+ ldr x0, =CONFIG_SYS_MALLOC_F_LEN
+ sub x0, x18, x0
str x0, [x18, #GD_MALLOC_BASE]
#endif
bic sp, x0, #0xf /* 16-byte alignment for ABI compliance */