refactor(mpam): enable FEAT_MPAM for FEAT_STATE_CHECKED

At the moment we only support FEAT_MPAM to be either unconditionally
compiled in, or to be not supported at all.

Add support for runtime detection (ENABLE_MPAM_FOR_LOWER_ELS=2), by
splitting get_mpam_version() into an ID register reading
function and a second function to report the support status. That
function considers both build time settings and runtime information (if
needed), and is used before we access MPAM related registers.
Also move the context saving code from assembly to C, and use the new
is_feat_mpam_supported() function to guard its execution.

ENABLE_MPAM_FOR_LOWER_ELS defaults to 0, so add a stub enable function
to cover builds with compiler optimisations turned off. The unused
mpam_enable() function call will normally be optimised away (because it
would never be called), but with -O0 the compiler will leave the symbol
in the object file.

Change-Id: I531d87cb855a7c43471f861f625b5a6d4bc61313
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
diff --git a/lib/el3_runtime/aarch64/context.S b/lib/el3_runtime/aarch64/context.S
index b058334..7bb0f99 100644
--- a/lib/el3_runtime/aarch64/context.S
+++ b/lib/el3_runtime/aarch64/context.S
@@ -17,10 +17,6 @@
 	.global	el2_sysregs_context_save_mte
 	.global	el2_sysregs_context_restore_mte
 #endif /* CTX_INCLUDE_MTE_REGS */
-#if ENABLE_MPAM_FOR_LOWER_ELS
-	.global	el2_sysregs_context_save_mpam
-	.global	el2_sysregs_context_restore_mpam
-#endif /* ENABLE_MPAM_FOR_LOWER_ELS */
 #if ENABLE_FEAT_ECV
 	.global	el2_sysregs_context_save_ecv
 	.global	el2_sysregs_context_restore_ecv
@@ -230,208 +226,6 @@
 endfunc el2_sysregs_context_restore_mte
 #endif /* CTX_INCLUDE_MTE_REGS */
 
-#if ENABLE_MPAM_FOR_LOWER_ELS
-func el2_sysregs_context_save_mpam
-	mrs	x10, MPAM2_EL2
-	str	x10, [x0, #CTX_MPAM2_EL2]
-
-	mrs	x10, MPAMIDR_EL1
-
-	/*
-	 * The context registers that we intend to save would be part of the
-	 * PE's system register frame only if MPAMIDR_EL1.HAS_HCR == 1.
-	 */
-	tbz	w10, #MPAMIDR_EL1_HAS_HCR_SHIFT, 3f
-
-	/*
-	 * MPAMHCR_EL2, MPAMVPMV_EL2 and MPAMVPM0_EL2 would be present in the
-	 * system register frame if MPAMIDR_EL1.HAS_HCR == 1. Proceed to save
-	 * the context of these registers.
-	 */
-	mrs	x11, MPAMHCR_EL2
-	mrs	x12, MPAMVPM0_EL2
-	stp	x11, x12, [x0, #CTX_MPAMHCR_EL2]
-
-	mrs	x13, MPAMVPMV_EL2
-	str	x13, [x0, #CTX_MPAMVPMV_EL2]
-
-	/*
-	 * MPAMIDR_EL1.VPMR_MAX has to be probed to obtain the maximum supported
-	 * VPMR value. Proceed to save the context of registers from
-	 * MPAMVPM1_EL2 to MPAMVPM<x>_EL2 where x is VPMR_MAX. From MPAM spec,
-	 * VPMR_MAX should not be zero if HAS_HCR == 1.
-	 */
-	ubfx	x10, x10, #MPAMIDR_EL1_VPMR_MAX_SHIFT, \
-		#MPAMIDR_EL1_VPMR_MAX_WIDTH
-
-	/*
-	 * Once VPMR_MAX has been identified, calculate the offset relative to
-	 * PC to jump to so that relevant context can be saved. The offset is
-	 * calculated as (VPMR_POSSIBLE_MAX - VPMR_MAX) * (instruction size for
-	 * saving one VPM register) + (absolute address of label "1").
-	 */
-	mov	w11, #MPAMIDR_EL1_VPMR_MAX_POSSIBLE
-	sub	w10, w11, w10
-
-	/* Calculate the size of one block of MPAMVPM*_EL2 save */
-	adr	x11, 1f
-	adr	x12, 2f
-	sub	x12, x12, x11
-
-	madd	x10, x10, x12, x11
-	br	x10
-
-	/*
-	 * The branch above would land properly on one of the blocks following
-	 * label "1". Make sure that the order of save is retained.
-	 */
-1:
-#if ENABLE_BTI
-	bti	j
-#endif
-	mrs	x10, MPAMVPM7_EL2
-	str	x10, [x0, #CTX_MPAMVPM7_EL2]
-2:
-#if ENABLE_BTI
-	bti	j
-#endif
-	mrs	x11, MPAMVPM6_EL2
-	str	x11, [x0, #CTX_MPAMVPM6_EL2]
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	mrs	x12, MPAMVPM5_EL2
-	str	x12, [x0, #CTX_MPAMVPM5_EL2]
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	mrs	x13, MPAMVPM4_EL2
-	str	x13, [x0, #CTX_MPAMVPM4_EL2]
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	mrs	x14, MPAMVPM3_EL2
-	str	x14, [x0, #CTX_MPAMVPM3_EL2]
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	mrs	x15, MPAMVPM2_EL2
-	str	x15, [x0, #CTX_MPAMVPM2_EL2]
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	mrs	x16, MPAMVPM1_EL2
-	str	x16, [x0, #CTX_MPAMVPM1_EL2]
-
-3:	ret
-endfunc el2_sysregs_context_save_mpam
-
-func el2_sysregs_context_restore_mpam
-	ldr	x10, [x0, #CTX_MPAM2_EL2]
-	msr	MPAM2_EL2, x10
-
-	mrs	x10, MPAMIDR_EL1
-	/*
-	 * The context registers that we intend to restore would be part of the
-	 * PE's system register frame only if MPAMIDR_EL1.HAS_HCR == 1.
-	 */
-	tbz	w10, #MPAMIDR_EL1_HAS_HCR_SHIFT, 3f
-
-	/*
-	 * MPAMHCR_EL2, MPAMVPMV_EL2 and MPAMVPM0_EL2 would be present in the
-	 * system register frame if MPAMIDR_EL1.HAS_HCR == 1. Proceed to restore
-	 * the context of these registers
-	 */
-	ldp	x11, x12, [x0, #CTX_MPAMHCR_EL2]
-	msr	MPAMHCR_EL2, x11
-	msr	MPAMVPM0_EL2, x12
-
-	ldr	x13, [x0, #CTX_MPAMVPMV_EL2]
-	msr	MPAMVPMV_EL2, x13
-
-	/*
-	 * MPAMIDR_EL1.VPMR_MAX has to be probed to obtain the maximum supported
-	 * VPMR value. Proceed to restore the context of registers from
-	 * MPAMVPM1_EL2 to MPAMVPM<x>_EL2 where x is VPMR_MAX. from MPAM spec,
-	 * VPMR_MAX should not be zero if HAS_HCR == 1.
-	 */
-	ubfx	x10, x10, #MPAMIDR_EL1_VPMR_MAX_SHIFT,	\
-		#MPAMIDR_EL1_VPMR_MAX_WIDTH
-
-	/*
-	 * Once VPMR_MAX has been identified, calculate the offset relative to
-	 * PC to jump to so that relevant context can be restored. The offset is
-	 * calculated as (VPMR_POSSIBLE_MAX - VPMR_MAX) * (instruction size for
-	 * restoring one VPM register) + (absolute address of label "1").
-	 */
-	mov	w11, #MPAMIDR_EL1_VPMR_MAX_POSSIBLE
-	sub	w10, w11, w10
-
-	/* Calculate the size of one block of MPAMVPM*_EL2 restore */
-	adr	x11, 1f
-	adr	x12, 2f
-	sub	x12, x12, x11
-
-	madd	x10, x10, x12, x11
-	br	x10
-
-	/*
-	 * The branch above would land properly on one of the blocks following
-	 * label "1". Make sure that the order of restore is retained.
-	 */
-1:
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	ldr	x10, [x0, #CTX_MPAMVPM7_EL2]
-	msr	MPAMVPM7_EL2, x10
-2:
-#if ENABLE_BTI
-	bti	j
-#endif
-	ldr	x11, [x0, #CTX_MPAMVPM6_EL2]
-	msr	MPAMVPM6_EL2, x11
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	ldr	x12, [x0, #CTX_MPAMVPM5_EL2]
-	msr	MPAMVPM5_EL2, x12
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	ldr	x13, [x0, #CTX_MPAMVPM4_EL2]
-	msr	MPAMVPM4_EL2, x13
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	ldr	x14, [x0, #CTX_MPAMVPM3_EL2]
-	msr	MPAMVPM3_EL2, x14
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	ldr	x15, [x0, #CTX_MPAMVPM2_EL2]
-	msr	MPAMVPM2_EL2, x15
-
-#if ENABLE_BTI
-	bti	j
-#endif
-	ldr	x16, [x0, #CTX_MPAMVPM1_EL2]
-	msr	MPAMVPM1_EL2, x16
-
-3:	ret
-endfunc el2_sysregs_context_restore_mpam
-#endif /* ENABLE_MPAM_FOR_LOWER_ELS */
-
 #if ENABLE_FEAT_ECV
 func el2_sysregs_context_save_ecv
 	mrs	x11, CNTPOFF_EL2
diff --git a/lib/el3_runtime/aarch64/context_mgmt.c b/lib/el3_runtime/aarch64/context_mgmt.c
index c5f20ad..eb5d1db 100644
--- a/lib/el3_runtime/aarch64/context_mgmt.c
+++ b/lib/el3_runtime/aarch64/context_mgmt.c
@@ -498,9 +498,9 @@
 	sve_enable(ctx);
 #endif
 
-#if ENABLE_MPAM_FOR_LOWER_ELS
-	mpam_enable(el2_unused);
-#endif
+	if (is_feat_mpam_supported()) {
+		mpam_enable(el2_unused);
+	}
 
 	if (is_feat_trbe_supported()) {
 		trbe_enable();
@@ -834,6 +834,96 @@
 	write_hfgwtr_el2(read_ctx_reg(ctx, CTX_HFGWTR_EL2));
 }
 
+static void el2_sysregs_context_save_mpam(el2_sysregs_t *ctx)
+{
+	u_register_t mpam_idr = read_mpamidr_el1();
+
+	write_ctx_reg(ctx, CTX_MPAM2_EL2, read_mpam2_el2());
+
+	/*
+	 * The context registers that we intend to save would be part of the
+	 * PE's system register frame only if MPAMIDR_EL1.HAS_HCR == 1.
+	 */
+	if ((mpam_idr & MPAMIDR_HAS_HCR_BIT) == 0U) {
+		return;
+	}
+
+	/*
+	 * MPAMHCR_EL2, MPAMVPMV_EL2 and MPAMVPM0_EL2 are always present if
+	 * MPAMIDR_HAS_HCR_BIT == 1.
+	 */
+	write_ctx_reg(ctx, CTX_MPAMHCR_EL2, read_mpamhcr_el2());
+	write_ctx_reg(ctx, CTX_MPAMVPM0_EL2, read_mpamvpm0_el2());
+	write_ctx_reg(ctx, CTX_MPAMVPMV_EL2, read_mpamvpmv_el2());
+
+	/*
+	 * The number of MPAMVPM registers is implementation defined, their
+	 * number is stored in the MPAMIDR_EL1 register.
+	 */
+	switch ((mpam_idr >> MPAMIDR_EL1_VPMR_MAX_SHIFT) & MPAMIDR_EL1_VPMR_MAX_MASK) {
+	case 7:
+		write_ctx_reg(ctx, CTX_MPAMVPM7_EL2, read_mpamvpm7_el2());
+		__fallthrough;
+	case 6:
+		write_ctx_reg(ctx, CTX_MPAMVPM6_EL2, read_mpamvpm6_el2());
+		__fallthrough;
+	case 5:
+		write_ctx_reg(ctx, CTX_MPAMVPM5_EL2, read_mpamvpm5_el2());
+		__fallthrough;
+	case 4:
+		write_ctx_reg(ctx, CTX_MPAMVPM4_EL2, read_mpamvpm4_el2());
+		__fallthrough;
+	case 3:
+		write_ctx_reg(ctx, CTX_MPAMVPM3_EL2, read_mpamvpm3_el2());
+		__fallthrough;
+	case 2:
+		write_ctx_reg(ctx, CTX_MPAMVPM2_EL2, read_mpamvpm2_el2());
+		__fallthrough;
+	case 1:
+		write_ctx_reg(ctx, CTX_MPAMVPM1_EL2, read_mpamvpm1_el2());
+		break;
+	}
+}
+
+static void el2_sysregs_context_restore_mpam(el2_sysregs_t *ctx)
+{
+	u_register_t mpam_idr = read_mpamidr_el1();
+
+	write_mpam2_el2(read_ctx_reg(ctx, CTX_MPAM2_EL2));
+
+	if ((mpam_idr & MPAMIDR_HAS_HCR_BIT) == 0U) {
+		return;
+	}
+
+	write_mpamhcr_el2(read_ctx_reg(ctx, CTX_MPAMHCR_EL2));
+	write_mpamvpm0_el2(read_ctx_reg(ctx, CTX_MPAMVPM0_EL2));
+	write_mpamvpmv_el2(read_ctx_reg(ctx, CTX_MPAMVPMV_EL2));
+
+	switch ((mpam_idr >> MPAMIDR_EL1_VPMR_MAX_SHIFT) & MPAMIDR_EL1_VPMR_MAX_MASK) {
+	case 7:
+		write_mpamvpm7_el2(read_ctx_reg(ctx, CTX_MPAMVPM7_EL2));
+		__fallthrough;
+	case 6:
+		write_mpamvpm6_el2(read_ctx_reg(ctx, CTX_MPAMVPM6_EL2));
+		__fallthrough;
+	case 5:
+		write_mpamvpm5_el2(read_ctx_reg(ctx, CTX_MPAMVPM5_EL2));
+		__fallthrough;
+	case 4:
+		write_mpamvpm4_el2(read_ctx_reg(ctx, CTX_MPAMVPM4_EL2));
+		__fallthrough;
+	case 3:
+		write_mpamvpm3_el2(read_ctx_reg(ctx, CTX_MPAMVPM3_EL2));
+		__fallthrough;
+	case 2:
+		write_mpamvpm2_el2(read_ctx_reg(ctx, CTX_MPAMVPM2_EL2));
+		__fallthrough;
+	case 1:
+		write_mpamvpm1_el2(read_ctx_reg(ctx, CTX_MPAMVPM1_EL2));
+		break;
+	}
+}
+
 /*******************************************************************************
  * Save EL2 sysreg context
  ******************************************************************************/
@@ -859,9 +949,9 @@
 #if CTX_INCLUDE_MTE_REGS
 		el2_sysregs_context_save_mte(el2_sysregs_ctx);
 #endif
-#if ENABLE_MPAM_FOR_LOWER_ELS
-		el2_sysregs_context_save_mpam(el2_sysregs_ctx);
-#endif
+		if (is_feat_mpam_supported()) {
+			el2_sysregs_context_save_mpam(el2_sysregs_ctx);
+		}
 
 		if (is_feat_fgt_supported()) {
 			el2_sysregs_context_save_fgt(el2_sysregs_ctx);
@@ -919,9 +1009,9 @@
 #if CTX_INCLUDE_MTE_REGS
 		el2_sysregs_context_restore_mte(el2_sysregs_ctx);
 #endif
-#if ENABLE_MPAM_FOR_LOWER_ELS
-		el2_sysregs_context_restore_mpam(el2_sysregs_ctx);
-#endif
+		if (is_feat_mpam_supported()) {
+			el2_sysregs_context_restore_mpam(el2_sysregs_ctx);
+		}
 
 		if (is_feat_fgt_supported()) {
 			el2_sysregs_context_restore_fgt(el2_sysregs_ctx);