perf(bl31): convert cpu_data fetching to C

The assembly routines are opaque to the compiler and it can't inline
them. There is also no requirement for them to be called without a
stack - each of their calls has a stack available. So convert them to C
so that the compiler can do its inlining magic.

On AArch32 we need to be able to call _cpu_data from the entrypoint so
it has to stay as a slight exception.

We can also straighten out the type of the cpu_ops_ptr member so we
don't have to cast it everywhere.

Change-Id: I9c2939a955b396edf26b99ef36318eebeaab13e6
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
diff --git a/lib/el3_runtime/aarch32/cpu_data.S b/lib/el3_runtime/aarch32/cpu_data.S
deleted file mode 100644
index e59b7fd..0000000
--- a/lib/el3_runtime/aarch32/cpu_data.S
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2016, Arm Limited and Contributors. All rights reserved.
- *
- * SPDX-License-Identifier: BSD-3-Clause
- */
-
-#include <asm_macros.S>
-#include <lib/el3_runtime/cpu_data.h>
-
-	.globl	_cpu_data
-	.globl	_cpu_data_by_index
-
-/* -----------------------------------------------------------------
- * cpu_data_t *_cpu_data(void)
- *
- * Return the cpu_data structure for the current CPU.
- * -----------------------------------------------------------------
- */
-func _cpu_data
-	/* r12 is pushed to meet the 8 byte stack alignment requirement */
-	push	{r12, lr}
-	bl	plat_my_core_pos
-	pop	{r12, lr}
-	b	_cpu_data_by_index
-endfunc _cpu_data
-
-/* -----------------------------------------------------------------
- * cpu_data_t *_cpu_data_by_index(uint32_t cpu_index)
- *
- * Return the cpu_data structure for the CPU with given linear index
- *
- * This can be called without a valid stack.
- * clobbers: r0, r1
- * -----------------------------------------------------------------
- */
-func _cpu_data_by_index
-	mov_imm	r1, CPU_DATA_SIZE
-	mul	r0, r0, r1
-	ldr	r1, =percpu_data
-	add	r0, r0, r1
-	bx	lr
-endfunc _cpu_data_by_index
diff --git a/lib/el3_runtime/aarch64/cpu_data.S b/lib/el3_runtime/aarch64/cpu_data.S
deleted file mode 100644
index 02d9415..0000000
--- a/lib/el3_runtime/aarch64/cpu_data.S
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2014-2025, Arm Limited and Contributors. All rights reserved.
- *
- * SPDX-License-Identifier: BSD-3-Clause
- */
-
-#include <asm_macros.S>
-#include <lib/el3_runtime/cpu_data.h>
-
-.globl	_cpu_data_by_index
-
-/* -----------------------------------------------------------------
- * cpu_data_t *_cpu_data_by_index(uint32_t cpu_index)
- *
- * Return the cpu_data structure for the CPU with given linear index
- *
- * This can be called without a valid stack.
- * clobbers: x0, x1
- * -----------------------------------------------------------------
- */
-func _cpu_data_by_index
-	mov_imm	x1, CPU_DATA_SIZE
-	mul	x0, x0, x1
-	adrp	x1, percpu_data
-	add	x1, x1, :lo12:percpu_data
-	add	x0, x0, x1
-	ret
-endfunc _cpu_data_by_index
diff --git a/lib/el3_runtime/cpu_data_array.c b/lib/el3_runtime/cpu_data_array.c
index 2056182..f2e97f0 100644
--- a/lib/el3_runtime/cpu_data_array.c
+++ b/lib/el3_runtime/cpu_data_array.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014-2016, Arm Limited and Contributors. All rights reserved.
+ * Copyright (c) 2014-2025, Arm Limited and Contributors. All rights reserved.
  *
  * SPDX-License-Identifier: BSD-3-Clause
  */
@@ -8,6 +8,14 @@
 
 #include <lib/cassert.h>
 #include <lib/el3_runtime/cpu_data.h>
+#include <plat/common/platform.h>
 
 /* The per_cpu_ptr_cache_t space allocation */
 cpu_data_t percpu_data[PLATFORM_CORE_COUNT];
+
+#ifndef __aarch64__
+cpu_data_t *_cpu_data(void)
+{
+	return _cpu_data_by_index(plat_my_core_pos());
+}
+#endif