[PATCH] libgloss, aarch64: Fix off-by-one in exception handler

There is a catch-all trap handler in the EL3 boot code found in
libgloss/aarch64/cpu-init/rdimon-aem-el3.S.  It makes a call to write
which is equivalent to the following C code:

write(STDERR_FILENO, "Terminated by exception.\n", 26);

the problem is that 26 is the length of the string + 1, the correct
length is 25:

$ python3 -c 'print(len("Terminated by exception.\n"))'
25

Therefore, as things stand, the trailing NUL byte is also written to
stderr; this can be seen by inspecting the output of binaries built with
e.g. -specs=aem-ve.specs before and after this patch is applied, as in
the below:

$ cat run_qemu.sh
#!/bin/bash
qemu-system-aarch64 -machine virt,secure=on -cpu neoverse-v1 -m 2g \
  -nographic -semihosting -device loader,file=$1,cpu-num=0
$ ./run_qemu.sh before.exe 2>&1 | xxd
00000000: 5465 726d 696e 6174 6564 2062 7920 6578  Terminated by ex
00000010: 6365 7074 696f 6e2e 0a00                 ception...
$ ./run_qemu.sh after.exe 2>&1 | xxd
00000000: 5465 726d 696e 6174 6564 2062 7920 6578  Terminated by ex
00000010: 6365 7074 696f 6e2e 0a                   ception..

This simple patch fixes the off-by-one error, passing the correct length
to write in the exception handler.
---

OK to commit?  I'll need someone to commit on my behalf as I don't have
write access.

Thanks,
Alex

---
 libgloss/aarch64/cpu-init/rdimon-aem-el3.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/libgloss/aarch64/cpu-init/rdimon-aem-el3.S b/libgloss/aarch64/cpu-init/rdimon-aem-el3.S
index 0296a8054..19dfee4f3 100644
--- a/libgloss/aarch64/cpu-init/rdimon-aem-el3.S
+++ b/libgloss/aarch64/cpu-init/rdimon-aem-el3.S
@@ -68,7 +68,7 @@ lower_a32_fiq:
 lower_a32_serror:
 	mov	x0, 2
 	adr	x1, .LC3
-	mov	x2, 26
+	mov	x2, 25
 	bl	FUNCTION (write)
 	mov	x0,  126
 	b	FUNCTION (exit)		/* Cannot return.  */