kselftest/arm64: Verify SME only ABI in syscall-abi

Currently syscall-abi only covers SME in the case where the system supports
SVE however it is architecturally valid to support SME without SVE. Update
the program to cover this case, this requires adjustments in the code to
check for SVCR.SM being set when deciding if we're handling the FPSIMD or
SVE registers and the addition of new test cases for the SME only case.

Note that in the SME only case we should not save the SVE registers after a
syscall since even if we were in streaming mode and therefore set them the
syscall should have exited streaming mode, we check that we have done so by
looking at SVCR.

Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20221223-arm64-syscall-abi-sme-only-v1-3-4fabfbd62087@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>

2 files changed, 42 insertions, 6 deletions

diff --git a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S
index acd5e9f3bc0b..cdfafc939a9e 100644
--- a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S
+++ b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S
@@ -92,8 +92,11 @@ do_syscall:
 	str	x29, [x2], #8		// FP
 	str	x30, [x2], #8		// LR
 
-	// Load FPRs if we're not doing SVE
+	// Load FPRs if we're not doing neither SVE nor streaming SVE
 	cbnz	x0, 1f
+	ldr	x2, =svcr_in
+	tbnz	x2, #SVCR_SM_SHIFT, 1f
+
 	ldr	x2, =fpr_in
 	ldp	q0, q1, [x2]
 	ldp	q2, q3, [x2, #16 * 2]
@@ -111,10 +114,11 @@ do_syscall:
 	ldp	q26, q27, [x2, #16 * 26]
 	ldp	q28, q29, [x2, #16 * 28]
 	ldp	q30, q31, [x2, #16 * 30]
+
+	b	2f
 1:
 
 	// Load the SVE registers if we're doing SVE/SME
-	cbz	x0, 1f
 
 	ldr	x2, =z_in
 	ldr	z0, [x2, #0, MUL VL]
@@ -155,9 +159,9 @@ do_syscall:
 	ldr	x2, =ffr_in
 	ldr	p0, [x2]
 	ldr	x2, [x2, #0]
-	cbz	x2, 2f
+	cbz	x2, 1f
 	wrffr	p0.b
-2:
+1:
 
 	ldr	x2, =p_in
 	ldr	p0, [x2, #0, MUL VL]
@@ -176,7 +180,7 @@ do_syscall:
 	ldr	p13, [x2, #13, MUL VL]
 	ldr	p14, [x2, #14, MUL VL]
 	ldr	p15, [x2, #15, MUL VL]
-1:
+2:
 
 	// Do the syscall
 	svc	#0
diff --git a/tools/testing/selftests/arm64/abi/syscall-abi.c b/tools/testing/selftests/arm64/abi/syscall-abi.c
index 45fdcbe3e909..7c9b6e947040 100644
--- a/tools/testing/selftests/arm64/abi/syscall-abi.c
+++ b/tools/testing/selftests/arm64/abi/syscall-abi.c
@@ -88,6 +88,7 @@ static int check_gpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, uint64_t s
 #define NUM_FPR 32
 uint64_t fpr_in[NUM_FPR * 2];
 uint64_t fpr_out[NUM_FPR * 2];
+uint64_t fpr_zero[NUM_FPR * 2];
 
 static void setup_fpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl,
 		      uint64_t svcr)
@@ -102,7 +103,7 @@ static int check_fpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl,
 	int errors = 0;
 	int i;
 
-	if (!sve_vl) {
+	if (!sve_vl && !(svcr & SVCR_SM_MASK)) {
 		for (i = 0; i < ARRAY_SIZE(fpr_in); i++) {
 			if (fpr_in[i] != fpr_out[i]) {
 				ksft_print_msg("%s Q%d/%d mismatch %llx != %llx\n",
@@ -114,6 +115,18 @@ static int check_fpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl,
 		}
 	}
 
+	/*
+	 * In streaming mode the whole register set should be cleared
+	 * by the transition out of streaming mode.
+	 */
+	if (svcr & SVCR_SM_MASK) {
+		if (memcmp(fpr_zero, fpr_out, sizeof(fpr_out)) != 0) {
+			ksft_print_msg("%s FPSIMD registers non-zero exiting SM\n",
+				       cfg->name);
+			errors++;
+		}
+	}
+
 	return errors;
 }
 
@@ -400,6 +413,24 @@ static void test_one_syscall(struct syscall_cfg *cfg)
 					 sme_vls[sme]);
 		}
 	}
+
+	for (sme = 0; sme < sme_vl_count; sme++) {
+		ret = prctl(PR_SME_SET_VL, sme_vls[sme]);
+		if (ret == -1)
+			ksft_exit_fail_msg("PR_SME_SET_VL failed: %s (%d)\n",
+						   strerror(errno), errno);
+
+		ksft_test_result(do_test(cfg, 0, sme_vls[sme],
+					 SVCR_ZA_MASK | SVCR_SM_MASK),
+				 "%s SME VL %d SM+ZA\n",
+				 cfg->name, sme_vls[sme]);
+		ksft_test_result(do_test(cfg, 0, sme_vls[sme], SVCR_SM_MASK),
+				 "%s SME VL %d SM\n",
+				 cfg->name, sme_vls[sme]);
+		ksft_test_result(do_test(cfg, 0, sme_vls[sme], SVCR_ZA_MASK),
+				 "%s SME VL %d ZA\n",
+				 cfg->name, sme_vls[sme]);
+	}
 }
 
 void sve_count_vls(void)
@@ -474,6 +505,7 @@ int main(void)
 	sme_count_vls();
 
 	tests += sve_vl_count;
+	tests += sme_vl_count * 3;
 	tests += (sve_vl_count * sme_vl_count) * 3;
 	ksft_set_plan(ARRAY_SIZE(syscalls) * tests);