arch/x86: Save resume vector to stack in x86_64 mode

In x86_64, the first function parameter is passed in rdi register, and
the 32-bit code after exiting long mode reads the resume vector in
4(%esp), so it's needed to save the resume vector from rdi to 4(%rsp).

Also note that the function attribute "regparm" only works on x86-32
targets according to the GCC manual, so "asmlinkage" doesn't change
the ABI of an x86_64 function.

Tested on HP EliteBook 2560p. The laptop can resume from S3 in x86_64
mode after this change.

Change-Id: I45f2678071b2511c0af5dce9d9b73ac70dfd7252
Signed-off-by: Iru Cai <mytbk920423@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/55947
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>

src/arch/x86/wakeup.S

@@ -15,6 +15,13 @@
 	.globl __wakeup
 __wakeup:
 #if ENV_X86_64
+	/* When called in x86_64 mode, the resume vector is in %rdi
+	 * instead of the stack, save it in 4(%rsp) for the 32-bit code.
+	 * It's OK to overwrite the return address at (%rsp) because this
+	 * function doesn't return.
+	 */
+	mov	%edi, 4(%rsp)
+
 	xor	%rax,%rax
 	mov	%ss, %ax
 	push	%rax