[kvm-unit-tests GIT PULL 0/4] s390x: storage key and CMM concurrent tests

[kvm-unit-tests GIT PULL 0/4] s390x: storage key and CMM concurrent tests

[Claudio Imbrenda]

Hi Paolo and/or Thomas,


please merge the following changes:

* storage key and cmm concurrent migration test
* new utility macros for PSWs


MERGE: https://gitlab.com/kvm-unit-tests/kvm-unit-tests/-/merge_requests/38

PIPELINE: https://gitlab.com/imbrenda/kvm-unit-tests/-/pipelines/739008304

PULL: https://gitlab.com/imbrenda/kvm-unit-tests.git s390x-next-2023-01

Claudio Imbrenda (2):
  lib: s390x: add PSW and PSW_WITH_CUR_MASK macros
  s390x: use the new PSW and PSW_WITH_CUR_MASK macros

Nico Boehr (2):
  s390x: add parallel skey migration test
  s390x: add CMM test during migration

 lib/s390x/asm/arch_def.h |   4 +
 s390x/adtl-status.c      |  24 +---
 s390x/firq.c             |   5 +-
 s390x/migration-cmm.c    | 258 ++++++++++++++++++++++++++++++++++-----
 s390x/migration-skey.c   | 218 ++++++++++++++++++++++++++++++---
 s390x/migration.c        |   6 +-
 s390x/skrf.c             |   7 +-
 s390x/smp.c              |  53 ++------
 s390x/uv-host.c          |   5 +-
 s390x/unittests.cfg      |  30 +++--
 10 files changed, 473 insertions(+), 137 deletions(-)

-- 
2.39.0

[kvm-unit-tests GIT PULL 1/4] lib: s390x: add PSW and PSW_WITH_CUR_MASK macros

[Claudio Imbrenda]

Since a lot of code starts new CPUs using the current PSW mask, add two
macros to streamline the creation of generic PSWs and PSWs with the
current program mask.

Reviewed-by: Nico Boehr <nrb@linux.ibm.com>
Reviewed-by: Janis Schoetterl-Glausch <scgl@linux.ibm.com>
Reviewed-by: Janosch Frank <frankja@linux.ibm.com>
Link: https://lore.kernel.org/r/20221130154038.70492-2-imbrenda@linux.ibm.com
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Message-Id: <20221130154038.70492-2-imbrenda@linux.ibm.com>
---
 lib/s390x/asm/arch_def.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/lib/s390x/asm/arch_def.h b/lib/s390x/asm/arch_def.h
index 783a7eaa..bb26e008 100644
--- a/lib/s390x/asm/arch_def.h
+++ b/lib/s390x/asm/arch_def.h
@@ -41,6 +41,8 @@ struct psw {
 	uint64_t	addr;
 };
 
+#define PSW(m, a) ((struct psw){ .mask = (m), .addr = (uint64_t)(a) })
+
 struct short_psw {
 	uint32_t	mask;
 	uint32_t	addr;
@@ -321,6 +323,8 @@ static inline uint64_t extract_psw_mask(void)
 	return (uint64_t) mask_upper << 32 | mask_lower;
 }
 
+#define PSW_WITH_CUR_MASK(addr) PSW(extract_psw_mask(), (addr))
+
 static inline void load_psw_mask(uint64_t mask)
 {
 	struct psw psw = {
-- 
2.39.0

[kvm-unit-tests GIT PULL 2/4] s390x: use the new PSW and PSW_WITH_CUR_MASK macros

[Claudio Imbrenda]

Use the new macros in the existing code. No functional changes intended.

Reviewed-by: Janis Schoetterl-Glausch <scgl@linux.ibm.com>
Reviewed-by: Janosch Frank <frankja@linux.ibm.com>
Link: https://lore.kernel.org/r/20221130154038.70492-3-imbrenda@linux.ibm.com
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Message-Id: <20221130154038.70492-3-imbrenda@linux.ibm.com>
---
 s390x/adtl-status.c | 24 ++++----------------
 s390x/firq.c        |  5 +----
 s390x/migration.c   |  6 +----
 s390x/skrf.c        |  7 +-----
 s390x/smp.c         | 53 ++++++++++-----------------------------------
 s390x/uv-host.c     |  5 +----
 6 files changed, 19 insertions(+), 81 deletions(-)

diff --git a/s390x/adtl-status.c b/s390x/adtl-status.c
index 9fb76319..2db650a1 100644
--- a/s390x/adtl-status.c
+++ b/s390x/adtl-status.c
@@ -205,16 +205,11 @@ static void restart_write_vector(void)
 
 static void cpu_write_magic_to_vector_regs(uint16_t cpu_idx)
 {
-	struct psw new_psw;
-
 	smp_cpu_stop(cpu_idx);
 
-	new_psw.mask = extract_psw_mask();
-	new_psw.addr = (unsigned long)restart_write_vector;
-
 	set_flag(0);
 
-	smp_cpu_start(cpu_idx, new_psw);
+	smp_cpu_start(cpu_idx, PSW_WITH_CUR_MASK(restart_write_vector));
 
 	wait_for_flag();
 }
@@ -241,7 +236,6 @@ static int adtl_status_check_unmodified_fields_for_lc(unsigned long lc)
 static void __store_adtl_status_vector_lc(unsigned long lc)
 {
 	uint32_t status = -1;
-	struct psw psw;
 	int cc;
 
 	report_prefix_pushf("LC %lu", lc);
@@ -274,9 +268,7 @@ static void __store_adtl_status_vector_lc(unsigned long lc)
 	 * Destroy and re-initalize the CPU to fix that.
 	 */
 	smp_cpu_destroy(1);
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)test_func;
-	smp_cpu_setup(1, psw);
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(test_func));
 
 out:
 	report_prefix_pop();
@@ -325,16 +317,11 @@ static void restart_write_gs_regs(void)
 
 static void cpu_write_to_gs_regs(uint16_t cpu_idx)
 {
-	struct psw new_psw;
-
 	smp_cpu_stop(cpu_idx);
 
-	new_psw.mask = extract_psw_mask();
-	new_psw.addr = (unsigned long)restart_write_gs_regs;
-
 	set_flag(0);
 
-	smp_cpu_start(cpu_idx, new_psw);
+	smp_cpu_start(cpu_idx, PSW_WITH_CUR_MASK(restart_write_gs_regs));
 
 	wait_for_flag();
 }
@@ -382,7 +369,6 @@ out:
 
 int main(void)
 {
-	struct psw psw;
 	report_prefix_push("adtl_status");
 
 	if (smp_query_num_cpus() == 1) {
@@ -391,9 +377,7 @@ int main(void)
 	}
 
 	/* Setting up the cpu to give it a stack and lowcore */
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)test_func;
-	smp_cpu_setup(1, psw);
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(test_func));
 	smp_cpu_stop(1);
 
 	test_store_adtl_status_unavail();
diff --git a/s390x/firq.c b/s390x/firq.c
index b4b3542e..0d65d87f 100644
--- a/s390x/firq.c
+++ b/s390x/firq.c
@@ -33,7 +33,6 @@ static void wait_for_sclp_int(void)
  */
 static void test_wait_state_delivery(void)
 {
-	struct psw psw;
 	SCCBHeader *h;
 	int ret;
 
@@ -55,9 +54,7 @@ static void test_wait_state_delivery(void)
 	sclp_mark_busy();
 
 	/* Start CPU #1 and let it wait for the interrupt. */
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)wait_for_sclp_int;
-	ret = smp_cpu_setup(1, psw);
+	ret = smp_cpu_setup(1, PSW_WITH_CUR_MASK(wait_for_sclp_int));
 	/* This must not fail because we have at least 3 CPUs */
 	assert(!ret);
 
diff --git a/s390x/migration.c b/s390x/migration.c
index fe6ea836..269e272d 100644
--- a/s390x/migration.c
+++ b/s390x/migration.c
@@ -159,8 +159,6 @@ static void test_func(void)
 
 int main(void)
 {
-	struct psw psw;
-
 	/* don't say migrate here otherwise we will migrate right away */
 	report_prefix_push("migration");
 
@@ -170,9 +168,7 @@ int main(void)
 	}
 
 	/* Second CPU does the actual tests */
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)test_func;
-	smp_cpu_setup(1, psw);
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(test_func));
 
 	/* wait for thread setup */
 	while(!flag_thread_complete)
diff --git a/s390x/skrf.c b/s390x/skrf.c
index 4cb563c3..e72a2b19 100644
--- a/s390x/skrf.c
+++ b/s390x/skrf.c
@@ -151,11 +151,6 @@ static void ecall_setup(void)
 
 static void test_exception_ext_new(void)
 {
-	struct psw psw = {
-		.mask = extract_psw_mask(),
-		.addr = (unsigned long)ecall_setup
-	};
-
 	report_prefix_push("exception external new");
 	if (smp_query_num_cpus() < 2) {
 		report_skip("Need second cpu for exception external new test.");
@@ -163,7 +158,7 @@ static void test_exception_ext_new(void)
 		return;
 	}
 
-	smp_cpu_setup(1, psw);
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(ecall_setup));
 	wait_for_flag();
 	set_flag(0);
 
diff --git a/s390x/smp.c b/s390x/smp.c
index 91f3e3bc..2b03bf0c 100644
--- a/s390x/smp.c
+++ b/s390x/smp.c
@@ -101,12 +101,8 @@ static void test_func(void)
 
 static void test_start(void)
 {
-	struct psw psw;
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)test_func;
-
 	set_flag(0);
-	smp_cpu_start(1, psw);
+	smp_cpu_start(1, PSW_WITH_CUR_MASK(test_func));
 	wait_for_flag();
 	report_pass("start");
 }
@@ -120,8 +116,7 @@ static void test_restart(void)
 	report_prefix_push("restart");
 	report_prefix_push("stopped");
 
-	lc->restart_new_psw.mask = extract_psw_mask();
-	lc->restart_new_psw.addr = (unsigned long)test_func;
+	lc->restart_new_psw = PSW_WITH_CUR_MASK(test_func);
 
 	/* Make sure cpu is stopped */
 	smp_cpu_stop(1);
@@ -256,7 +251,6 @@ static void test_set_prefix(void)
 	struct lowcore *new_lc = alloc_pages_flags(1, AREA_DMA31);
 	struct cpu *cpu1 = smp_cpu_from_idx(1);
 	uint32_t status = 0;
-	struct psw new_psw;
 	int cc;
 
 	report_prefix_push("set prefix");
@@ -268,9 +262,7 @@ static void test_set_prefix(void)
 
 	report_prefix_push("running");
 	set_flag(0);
-	new_psw.addr = (unsigned long)stpx_and_set_flag;
-	new_psw.mask = extract_psw_mask();
-	smp_cpu_start(1, new_psw);
+	smp_cpu_start(1, PSW_WITH_CUR_MASK(stpx_and_set_flag));
 	wait_for_flag();
 	cpu1_prefix = 0xFFFFFFFF;
 
@@ -326,7 +318,6 @@ static void call_received(void)
 static void test_calls(void)
 {
 	int i;
-	struct psw psw;
 
 	for (i = 0; i < ARRAY_SIZE(cases_sigp_call); i++) {
 		current_sigp_call_case = &cases_sigp_call[i];
@@ -339,9 +330,7 @@ static void test_calls(void)
 		}
 
 		set_flag(0);
-		psw.mask = extract_psw_mask();
-		psw.addr = (unsigned long)call_received;
-		smp_cpu_start(1, psw);
+		smp_cpu_start(1, PSW_WITH_CUR_MASK(call_received));
 
 		/* Wait until the receiver has finished setup */
 		wait_for_flag();
@@ -389,7 +378,6 @@ static void call_in_wait_cleanup(void)
 static void test_calls_in_wait(void)
 {
 	int i;
-	struct psw psw;
 
 	report_prefix_push("psw wait");
 	for (i = 0; i < ARRAY_SIZE(cases_sigp_call); i++) {
@@ -404,9 +392,7 @@ static void test_calls_in_wait(void)
 
 		/* Let the secondary CPU setup the external mask and the external interrupt cleanup function */
 		set_flag(0);
-		psw.mask = extract_psw_mask();
-		psw.addr = (unsigned long)call_in_wait_setup;
-		smp_cpu_start(1, psw);
+		smp_cpu_start(1, PSW_WITH_CUR_MASK(call_in_wait_setup));
 
 		/* Wait until the receiver has finished setup */
 		wait_for_flag();
@@ -422,9 +408,7 @@ static void test_calls_in_wait(void)
 		 * wait until the CPU becomes operating (done by smp_cpu_start).
 		 */
 		smp_cpu_stop(1);
-		psw.mask = extract_psw_mask() | PSW_MASK_EXT | PSW_MASK_WAIT;
-		psw.addr = (unsigned long)call_in_wait_received;
-		smp_cpu_start(1, psw);
+		smp_cpu_start(1, PSW(extract_psw_mask() | PSW_MASK_EXT | PSW_MASK_WAIT, call_in_wait_received));
 
 		smp_sigp(1, current_sigp_call_case->call, 0, NULL);
 
@@ -439,9 +423,7 @@ static void test_calls_in_wait(void)
 		 * to catch an interrupt that is presented twice since we would
 		 * disable the external call on the first interrupt.
 		 */
-		psw.mask = extract_psw_mask();
-		psw.addr = (unsigned long)call_in_wait_cleanup;
-		smp_cpu_start(1, psw);
+		smp_cpu_start(1, PSW_WITH_CUR_MASK(call_in_wait_cleanup));
 
 		/* Wait until the cleanup has been completed */
 		wait_for_flag();
@@ -478,15 +460,11 @@ static void test_func_initial(void)
 static void test_reset_initial(void)
 {
 	struct cpu_status *status = alloc_pages_flags(0, AREA_DMA31);
-	struct psw psw;
 	int i;
 
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)test_func_initial;
-
 	report_prefix_push("reset initial");
 	set_flag(0);
-	smp_cpu_start(1, psw);
+	smp_cpu_start(1, PSW_WITH_CUR_MASK(test_func_initial));
 	wait_for_flag();
 
 	smp_sigp(1, SIGP_INITIAL_CPU_RESET, 0, NULL);
@@ -525,22 +503,16 @@ static void test_local_ints(void)
 
 static void test_reset(void)
 {
-	struct psw psw;
-
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)test_func;
-
 	report_prefix_push("cpu reset");
 	smp_sigp(1, SIGP_EMERGENCY_SIGNAL, 0, NULL);
 	smp_sigp(1, SIGP_EXTERNAL_CALL, 0, NULL);
-	smp_cpu_start(1, psw);
+	smp_cpu_start(1, PSW_WITH_CUR_MASK(test_func));
 
 	smp_sigp(1, SIGP_CPU_RESET, 0, NULL);
 	report(smp_cpu_stopped(1), "cpu stopped");
 
 	set_flag(0);
-	psw.addr = (unsigned long)test_local_ints;
-	smp_cpu_start(1, psw);
+	smp_cpu_start(1, PSW_WITH_CUR_MASK(test_local_ints));
 	wait_for_flag();
 	report_pass("local interrupts cleared");
 	report_prefix_pop();
@@ -548,7 +520,6 @@ static void test_reset(void)
 
 int main(void)
 {
-	struct psw psw;
 	report_prefix_push("smp");
 
 	if (smp_query_num_cpus() == 1) {
@@ -557,9 +528,7 @@ int main(void)
 	}
 
 	/* Setting up the cpu to give it a stack and lowcore */
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)test_func;
-	smp_cpu_setup(1, psw);
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(test_func));
 	smp_cpu_stop(1);
 
 	test_start();
diff --git a/s390x/uv-host.c b/s390x/uv-host.c
index 9f09f80e..33e6eec6 100644
--- a/s390x/uv-host.c
+++ b/s390x/uv-host.c
@@ -501,7 +501,6 @@ static void test_init(void)
 {
 	int rc;
 	uint64_t mem;
-	struct psw psw;
 
 	/* Donated storage needs to be over 2GB */
 	mem = (uint64_t)memalign_pages_flags(SZ_1M, uvcb_qui.uv_base_stor_len, AREA_NORMAL);
@@ -547,9 +546,7 @@ static void test_init(void)
 	       "storage below 2GB");
 	uvcb_init.stor_origin = mem;
 
-	psw.mask = extract_psw_mask();
-	psw.addr = (unsigned long)cpu_loop;
-	smp_cpu_setup(1, psw);
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(cpu_loop));
 	rc = uv_call(0, (uint64_t)&uvcb_init);
 	report(rc == 1 && uvcb_init.header.rc == 0x102,
 	       "too many running cpus");
-- 
2.39.0

[kvm-unit-tests GIT PULL 3/4] s390x: add parallel skey migration test

[Claudio Imbrenda]

From: Nico Boehr <nrb@linux.ibm.com>

Right now, we have a test which sets storage keys, then migrates the VM
and - after migration finished - verifies the skeys are still there.

Add a new version of the test which changes storage keys while the
migration is in progress. This is achieved by adding a command line
argument to the existing migration-skey test.

Signed-off-by: Nico Boehr <nrb@linux.ibm.com>
Reviewed-by: Nina Schoetterl-Glausch <nsg@linux.ibm.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Link: https://lore.kernel.org/r/20221220083030.30153-2-nrb@linux.ibm.com
Message-Id: <20221220083030.30153-2-nrb@linux.ibm.com>
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
---
 s390x/migration-skey.c | 218 +++++++++++++++++++++++++++++++++++++----
 s390x/unittests.cfg    |  15 ++-
 2 files changed, 210 insertions(+), 23 deletions(-)

diff --git a/s390x/migration-skey.c b/s390x/migration-skey.c
index a91eb6b5..8d6d8ecf 100644
--- a/s390x/migration-skey.c
+++ b/s390x/migration-skey.c
@@ -2,6 +2,12 @@
 /*
  * Storage Key migration tests
  *
+ * There are two variants of this test:
+ * - sequential: set storage keys on some pages, migrates the VM and then
+ *   verifies that the storage keys are still as we expect.
+ * - parallel: start migration and set and check storage keys on some
+ *   pages while migration is in process.
+ *
  * Copyright IBM Corp. 2022
  *
  * Authors:
@@ -13,16 +19,45 @@
 #include <asm/facility.h>
 #include <asm/page.h>
 #include <asm/mem.h>
-#include <asm/interrupt.h>
+#include <asm/barrier.h>
 #include <hardware.h>
+#include <smp.h>
+
+struct verify_result {
+	bool verify_failed;
+	union skey expected_key;
+	union skey actual_key;
+	unsigned long page_mismatch_idx;
+	unsigned long page_mismatch_addr;
+};
 
 #define NUM_PAGES 128
-static uint8_t pagebuf[NUM_PAGES][PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
+static uint8_t pagebuf[NUM_PAGES * PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
+
+static struct verify_result result;
+
+static unsigned int thread_iters;
+static bool thread_should_exit;
+static bool thread_exited;
+
+static enum {
+	TEST_INVALID,
+	TEST_SEQUENTIAL,
+	TEST_PARALLEL
+} arg_test_to_run;
 
-static void test_migration(void)
+/*
+ * Set storage key test pattern on pagebuf with a seed for the storage keys.
+ *
+ * Each page's storage key is generated by taking the page's index in pagebuf,
+ * XOR-ing that with the given seed and then multipling the result with two.
+ *
+ * Only the lower seven bits of the seed are considered.
+ */
+static void set_test_pattern(unsigned char seed)
 {
-	union skey expected_key, actual_key;
-	int i, key_to_set, key_mismatches = 0;
+	unsigned char key_to_set;
+	unsigned long i;
 
 	for (i = 0; i < NUM_PAGES; i++) {
 		/*
@@ -32,15 +67,34 @@ static void test_migration(void)
 		 * protection as well as reference and change indication for
 		 * some keys.
 		 */
-		key_to_set = i * 2;
-		set_storage_key(pagebuf[i], key_to_set, 1);
+		key_to_set = (i ^ seed) * 2;
+		set_storage_key(pagebuf + i * PAGE_SIZE, key_to_set, 1);
 	}
+}
 
-	migrate_once();
+/*
+ * Verify storage keys on pagebuf.
+ * Storage keys must have been set by set_test_pattern on pagebuf before.
+ * set_test_pattern must have been called with the same seed value.
+ *
+ * If storage keys match the expected result, will return a verify_result
+ * with verify_failed false. All other fields are then invalid.
+ * If there is a mismatch, returned struct will have verify_failed true and will
+ * be filled with the details on the first mismatch encountered.
+ */
+static struct verify_result verify_test_pattern(unsigned char seed)
+{
+	union skey expected_key, actual_key;
+	struct verify_result result = {
+		.verify_failed = true
+	};
+	uint8_t *cur_page;
+	unsigned long i;
 
 	for (i = 0; i < NUM_PAGES; i++) {
-		actual_key.val = get_storage_key(pagebuf[i]);
-		expected_key.val = i * 2;
+		cur_page = pagebuf + i * PAGE_SIZE;
+		actual_key.val = get_storage_key(cur_page);
+		expected_key.val = (i ^ seed) * 2;
 
 		/*
 		 * The PoP neither gives a guarantee that the reference bit is
@@ -51,27 +105,153 @@ static void test_migration(void)
 		actual_key.str.rf = 0;
 		expected_key.str.rf = 0;
 
-		/* don't log anything when key matches to avoid spamming the log */
 		if (actual_key.val != expected_key.val) {
-			key_mismatches++;
-			report_fail("page %d expected_key=0x%x actual_key=0x%x", i, expected_key.val, actual_key.val);
+			result.expected_key.val = expected_key.val;
+			result.actual_key.val = actual_key.val;
+			result.page_mismatch_idx = i;
+			result.page_mismatch_addr = (unsigned long)cur_page;
+			return result;
 		}
 	}
 
-	report(!key_mismatches, "skeys after migration match");
+	result.verify_failed = false;
+	return result;
+}
+
+static void report_verify_result(const struct verify_result * result)
+{
+	if (result->verify_failed)
+		report_fail("page skey mismatch: first page idx = %lu, addr = 0x%lx, "
+			    "expected_key = 0x%02x, actual_key = 0x%02x",
+			    result->page_mismatch_idx, result->page_mismatch_addr,
+			    result->expected_key.val, result->actual_key.val);
+	else
+		report_pass("skeys match");
+}
+
+static void test_skey_migration_sequential(void)
+{
+	report_prefix_push("sequential");
+
+	set_test_pattern(0);
+
+	migrate_once();
+
+	result = verify_test_pattern(0);
+	report_verify_result(&result);
+
+	report_prefix_pop();
+}
+
+static void set_skeys_thread(void)
+{
+	while (!READ_ONCE(thread_should_exit)) {
+		set_test_pattern(thread_iters);
+
+		result = verify_test_pattern(thread_iters);
+
+		/*
+		 * Always increment even if the verify fails. This ensures primary CPU knows where
+		 * we left off and can do an additional verify round after migration finished.
+		 */
+		thread_iters++;
+
+		if (result.verify_failed)
+			break;
+	}
+
+	WRITE_ONCE(thread_exited, 1);
+}
+
+static void test_skey_migration_parallel(void)
+{
+	report_prefix_push("parallel");
+
+	if (smp_query_num_cpus() == 1) {
+		report_skip("need at least 2 cpus for this test");
+		goto error;
+	}
+
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(set_skeys_thread));
+
+	migrate_once();
+
+	WRITE_ONCE(thread_should_exit, 1);
+
+	while (!READ_ONCE(thread_exited))
+		;
+
+	/* Ensure we read result and thread_iters below from memory after thread exited */
+	mb();
+	report_info("thread completed %u iterations", thread_iters);
+
+	report_prefix_push("during migration");
+	report_verify_result(&result);
+	report_prefix_pop();
+
+	/*
+	 * Verification of skeys occurs on the thread. We don't know if we
+	 * were still migrating during the verification.
+	 * To be sure, make another verification round after the migration
+	 * finished to catch skeys which might not have been migrated
+	 * correctly.
+	 */
+	report_prefix_push("after migration");
+	assert(thread_iters > 0);
+	result = verify_test_pattern(thread_iters - 1);
+	report_verify_result(&result);
+	report_prefix_pop();
+
+error:
+	report_prefix_pop();
+}
+
+static void print_usage(void)
+{
+	report_info("Usage: migration-skey [--parallel|--sequential]");
+}
+
+static void parse_args(int argc, char **argv)
+{
+	if (argc < 2) {
+		/* default to sequential since it only needs one cpu */
+		arg_test_to_run = TEST_SEQUENTIAL;
+		return;
+	}
+
+	if (!strcmp("--parallel", argv[1]))
+		arg_test_to_run = TEST_PARALLEL;
+	else if (!strcmp("--sequential", argv[1]))
+		arg_test_to_run = TEST_SEQUENTIAL;
+	else
+		arg_test_to_run = TEST_INVALID;
 }
 
-int main(void)
+int main(int argc, char **argv)
 {
 	report_prefix_push("migration-skey");
 
-	if (test_facility(169))
+	if (test_facility(169)) {
 		report_skip("storage key removal facility is active");
-	else
-		test_migration();
+		goto error;
+	}
 
-	migrate_once();
+	parse_args(argc, argv);
+
+	switch (arg_test_to_run) {
+	case TEST_SEQUENTIAL:
+		test_skey_migration_sequential();
+		break;
+	case TEST_PARALLEL:
+		test_skey_migration_parallel();
+		break;
+	default:
+		print_usage();
+		break;
+	}
 
+error:
+	migrate_once();
 	report_prefix_pop();
 	return report_summary();
 }
diff --git a/s390x/unittests.cfg b/s390x/unittests.cfg
index 3caf81ed..d97eb5e9 100644
--- a/s390x/unittests.cfg
+++ b/s390x/unittests.cfg
@@ -185,10 +185,6 @@ smp = 2
 file = migration-cmm.elf
 groups = migration
 
-[migration-skey]
-file = migration-skey.elf
-groups = migration
-
 [panic-loop-extint]
 file = panic-loop-extint.elf
 groups = panic
@@ -208,3 +204,14 @@ groups = migration
 [exittime]
 file = exittime.elf
 smp = 2
+
+[migration-skey-sequential]
+file = migration-skey.elf
+groups = migration
+extra_params = -append '--sequential'
+
+[migration-skey-parallel]
+file = migration-skey.elf
+smp = 2
+groups = migration
+extra_params = -append '--parallel'
-- 
2.39.0

[kvm-unit-tests GIT PULL 4/4] s390x: add CMM test during migration

[Claudio Imbrenda]

From: Nico Boehr <nrb@linux.ibm.com>

Add a test which modifies CMM page states while migration is in
progress.

This is added to the existing migration-cmm test, which gets a new
command line argument for the sequential and parallel variants.

Signed-off-by: Nico Boehr <nrb@linux.ibm.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Link: https://lore.kernel.org/r/20221221090953.341247-2-nrb@linux.ibm.com
Message-Id: <20221221090953.341247-2-nrb@linux.ibm.com>
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
---
 s390x/migration-cmm.c | 258 +++++++++++++++++++++++++++++++++++++-----
 s390x/unittests.cfg   |  15 ++-
 2 files changed, 240 insertions(+), 33 deletions(-)

diff --git a/s390x/migration-cmm.c b/s390x/migration-cmm.c
index 43673f18..2d46c6be 100644
--- a/s390x/migration-cmm.c
+++ b/s390x/migration-cmm.c
@@ -2,6 +2,12 @@
 /*
  * CMM migration tests (ESSA)
  *
+ * There are two variants of this test:
+ * - sequential: sets CMM page states, then migrates the VM and - after
+ *   migration finished - verifies that page states have been preserved.
+ * - parallel: migrate VM and - while migration is in progress - change
+ *   page states and verify that they are preserved.
+ *
  * Copyright IBM Corp. 2022
  *
  * Authors:
@@ -13,55 +19,249 @@
 #include <asm/interrupt.h>
 #include <asm/page.h>
 #include <asm/cmm.h>
+#include <asm/barrier.h>
 #include <bitops.h>
+#include <smp.h>
+
+struct verify_result {
+	bool verify_failed;
+	char expected_mask;
+	char actual_mask;
+	unsigned long page_mismatch_idx;
+	unsigned long page_mismatch_addr;
+};
+
+static enum {
+	TEST_INVLALID,
+	TEST_SEQUENTIAL,
+	TEST_PARALLEL
+} arg_test_to_run;
 
 #define NUM_PAGES 128
-static uint8_t pagebuf[NUM_PAGES][PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
 
-static void test_migration(void)
+/*
+ * Allocate 3 pages more than we need so we can start at different offsets.
+ * For the parallel test, this ensures page states change on every loop iteration.
+ */
+static uint8_t pagebuf[(NUM_PAGES + 3) * PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
+
+static struct verify_result result;
+
+static unsigned int thread_iters;
+static int thread_should_exit;
+static int thread_exited;
+
+/*
+ * Maps ESSA actions to states the page is allowed to be in after the
+ * respective action was executed.
+ */
+static const unsigned long allowed_essa_state_masks[4] = {
+	BIT(ESSA_USAGE_STABLE),					/* ESSA_SET_STABLE */
+	BIT(ESSA_USAGE_UNUSED),					/* ESSA_SET_UNUSED */
+	BIT(ESSA_USAGE_VOLATILE),				/* ESSA_SET_VOLATILE */
+	BIT(ESSA_USAGE_VOLATILE) | BIT(ESSA_USAGE_POT_VOLATILE) /* ESSA_SET_POT_VOLATILE */
+};
+
+/*
+ * Set CMM page state test pattern on pagebuf.
+ * pagebuf must point to page_count consecutive pages.
+ * page_count must be a multiple of 4.
+ */
+static void set_test_pattern(uint8_t *pagebuf, size_t page_count)
+{
+	unsigned long addr = (unsigned long)pagebuf;
+	size_t i;
+
+	assert(page_count % 4 == 0);
+	for (i = 0; i < page_count; i += 4) {
+		essa(ESSA_SET_STABLE, addr + i * PAGE_SIZE);
+		essa(ESSA_SET_UNUSED, addr + (i + 1) * PAGE_SIZE);
+		essa(ESSA_SET_VOLATILE, addr + (i + 2) * PAGE_SIZE);
+		essa(ESSA_SET_POT_VOLATILE, addr + (i + 3) * PAGE_SIZE);
+	}
+}
+
+/*
+ * Verify CMM page states on pagebuf.
+ * Page states must have been set by set_test_pattern on pagebuf before.
+ * page_count must be a multiple of 4.
+ *
+ * If page states match the expected result, will return a verify_result
+ * with verify_failed false. All other fields are then invalid.
+ * If there is a mismatch, the returned struct will have verify_failed true
+ * and will be filled with details on the first mismatch encountered.
+ */
+static struct verify_result verify_page_states(uint8_t *pagebuf, size_t page_count)
+{
+	unsigned long expected_mask, actual_mask;
+	struct verify_result result = {
+		.verify_failed = true
+	};
+	unsigned long addr;
+	size_t i;
+
+	assert(page_count % 4 == 0);
+	for (i = 0; i < page_count; i++) {
+		addr = (unsigned long)(pagebuf + i * PAGE_SIZE);
+		actual_mask = essa(ESSA_GET_STATE, addr);
+		/* usage state in bits 60 and 61 */
+		actual_mask = BIT((actual_mask >> 2) & 0x3);
+		expected_mask = allowed_essa_state_masks[i % ARRAY_SIZE(allowed_essa_state_masks)];
+		if (!(actual_mask & expected_mask)) {
+			result.page_mismatch_idx = i;
+			result.page_mismatch_addr = addr;
+			result.expected_mask = expected_mask;
+			result.actual_mask = actual_mask;
+			return result;
+		}
+	}
+
+	result.verify_failed = false;
+	return result;
+}
+
+static void report_verify_result(const struct verify_result *result)
+{
+	if (result->verify_failed)
+		report_fail("page state mismatch: first page idx = %lu, addr = %lx, "
+			    "expected_mask = 0x%x, actual_mask = 0x%x",
+			    result->page_mismatch_idx, result->page_mismatch_addr,
+			    result->expected_mask, result->actual_mask);
+	else
+		report_pass("page states match");
+}
+
+static void test_cmm_migration_sequential(void)
+{
+	report_prefix_push("sequential");
+
+	set_test_pattern(pagebuf, NUM_PAGES);
+
+	migrate_once();
+
+	result = verify_page_states(pagebuf, NUM_PAGES);
+	report_verify_result(&result);
+
+	report_prefix_pop();
+}
+
+static void set_cmm_thread(void)
 {
-	int i, state_mask, actual_state;
+	uint8_t *pagebuf_start;
 	/*
-	 * Maps ESSA actions to states the page is allowed to be in after the
-	 * respective action was executed.
+	 * The second CPU must not print to the console, otherwise it will race with
+	 * the primary CPU on the SCLP buffer.
 	 */
-	int allowed_essa_state_masks[4] = {
-		BIT(ESSA_USAGE_STABLE),					/* ESSA_SET_STABLE */
-		BIT(ESSA_USAGE_UNUSED),					/* ESSA_SET_UNUSED */
-		BIT(ESSA_USAGE_VOLATILE),				/* ESSA_SET_VOLATILE */
-		BIT(ESSA_USAGE_VOLATILE) | BIT(ESSA_USAGE_POT_VOLATILE) /* ESSA_SET_POT_VOLATILE */
-	};
+	while (!READ_ONCE(thread_should_exit)) {
+		/*
+		 * Start on a offset different from the last iteration so page states change with
+		 * every iteration. This is why pagebuf has 3 extra pages.
+		 */
+		pagebuf_start = pagebuf + (thread_iters % 4) * PAGE_SIZE;
+		set_test_pattern(pagebuf_start, NUM_PAGES);
+
+		/*
+		 * Always increment even if the verify fails. This ensures primary CPU knows where
+		 * we left off and can do an additional verify round after migration finished.
+		 */
+		thread_iters++;
 
-	assert(NUM_PAGES % 4 == 0);
-	for (i = 0; i < NUM_PAGES; i += 4) {
-		essa(ESSA_SET_STABLE, (unsigned long)pagebuf[i]);
-		essa(ESSA_SET_UNUSED, (unsigned long)pagebuf[i + 1]);
-		essa(ESSA_SET_VOLATILE, (unsigned long)pagebuf[i + 2]);
-		essa(ESSA_SET_POT_VOLATILE, (unsigned long)pagebuf[i + 3]);
+		result = verify_page_states(pagebuf_start, NUM_PAGES);
+		if (result.verify_failed)
+			break;
 	}
 
+	WRITE_ONCE(thread_exited, 1);
+}
+
+static void test_cmm_migration_parallel(void)
+{
+	report_prefix_push("parallel");
+
+	if (smp_query_num_cpus() == 1) {
+		report_skip("need at least 2 cpus for this test");
+		goto error;
+	}
+
+	smp_cpu_setup(1, PSW_WITH_CUR_MASK(set_cmm_thread));
+
 	migrate_once();
 
-	for (i = 0; i < NUM_PAGES; i++) {
-		actual_state = essa(ESSA_GET_STATE, (unsigned long)pagebuf[i]);
-		/* extract the usage state in bits 60 and 61 */
-		actual_state = (actual_state >> 2) & 0x3;
-		state_mask = allowed_essa_state_masks[i % ARRAY_SIZE(allowed_essa_state_masks)];
-		report(BIT(actual_state) & state_mask, "page %d state: expected_mask=0x%x actual_mask=0x%lx", i, state_mask, BIT(actual_state));
+	WRITE_ONCE(thread_should_exit, 1);
+
+	while (!READ_ONCE(thread_exited))
+		;
+
+	/* Ensure thread_iters and result below are read from memory after thread completed */
+	mb();
+
+	report_info("thread completed %u iterations", thread_iters);
+
+	report_prefix_push("during migration");
+	report_verify_result(&result);
+	report_prefix_pop();
+
+	/*
+	 * Verification of page states occurs on the thread. We don't know if we
+	 * were still migrating during the verification.
+	 * To be sure, make another verification round after the migration
+	 * finished to catch page states which might not have been migrated
+	 * correctly.
+	 */
+	report_prefix_push("after migration");
+	assert(thread_iters > 0);
+	result = verify_page_states(pagebuf + ((thread_iters - 1) % 4) * PAGE_SIZE, NUM_PAGES);
+	report_verify_result(&result);
+	report_prefix_pop();
+
+error:
+	report_prefix_pop();
+}
+
+static void print_usage(void)
+{
+	report_info("Usage: migration-cmm [--parallel|--sequential]");
+}
+
+static void parse_args(int argc, char **argv)
+{
+	if (argc < 2) {
+		/* default to sequential since it only needs one CPU */
+		arg_test_to_run = TEST_SEQUENTIAL;
+		return;
 	}
+
+	if (!strcmp("--parallel", argv[1]))
+		arg_test_to_run = TEST_PARALLEL;
+	else if (!strcmp("--sequential", argv[1]))
+		arg_test_to_run = TEST_SEQUENTIAL;
+	else
+		arg_test_to_run = TEST_INVLALID;
 }
 
-int main(void)
+int main(int argc, char **argv)
 {
 	report_prefix_push("migration-cmm");
-
-	if (!check_essa_available())
+	if (!check_essa_available()) {
 		report_skip("ESSA is not available");
-	else
-		test_migration();
+		goto error;
+	}
 
-	migrate_once();
+	parse_args(argc, argv);
+
+	switch (arg_test_to_run) {
+	case TEST_SEQUENTIAL:
+		test_cmm_migration_sequential();
+		break;
+	case TEST_PARALLEL:
+		test_cmm_migration_parallel();
+		break;
+	default:
+		print_usage();
+	}
 
+error:
+	migrate_once();
 	report_prefix_pop();
 	return report_summary();
 }
diff --git a/s390x/unittests.cfg b/s390x/unittests.cfg
index d97eb5e9..96977f2b 100644
--- a/s390x/unittests.cfg
+++ b/s390x/unittests.cfg
@@ -181,10 +181,6 @@ file = migration.elf
 groups = migration
 smp = 2
 
-[migration-cmm]
-file = migration-cmm.elf
-groups = migration
-
 [panic-loop-extint]
 file = panic-loop-extint.elf
 groups = panic
@@ -215,3 +211,14 @@ file = migration-skey.elf
 smp = 2
 groups = migration
 extra_params = -append '--parallel'
+
+[migration-cmm-sequential]
+file = migration-cmm.elf
+groups = migration
+extra_params = -append '--sequential'
+
+[migration-cmm-parallel]
+file = migration-cmm.elf
+groups = migration
+smp = 2
+extra_params = -append '--parallel'
-- 
2.39.0

Re: [kvm-unit-tests GIT PULL 4/4] s390x: add CMM test during migration

[Thomas Huth]

On 05/01/2023 13.15, Claudio Imbrenda wrote:
> From: Nico Boehr <nrb@linux.ibm.com>
> 
> Add a test which modifies CMM page states while migration is in
> progress.
> 
> This is added to the existing migration-cmm test, which gets a new
> command line argument for the sequential and parallel variants.
> 
> Signed-off-by: Nico Boehr <nrb@linux.ibm.com>
> Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> Link: https://lore.kernel.org/r/20221221090953.341247-2-nrb@linux.ibm.com
> Message-Id: <20221221090953.341247-2-nrb@linux.ibm.com>
> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> ---
>   s390x/migration-cmm.c | 258 +++++++++++++++++++++++++++++++++++++-----
>   s390x/unittests.cfg   |  15 ++-
>   2 files changed, 240 insertions(+), 33 deletions(-)

  Hi!

While this works fine on my z15 LPAR, I'm getting a failure when running 
this test on my z13 LPAR:

$ cat logs/migration-cmm-parallel.log
run_migration timeout -k 1s --foreground 90s /usr/local/bin/qemu-kvm 
-nodefaults -nographic -machine s390-ccw-virtio,accel=kvm -chardev 
stdio,id=con0 -device sclpconsole,chardev=con0 -kernel 
s390x/migration-cmm.elf -smp 2 -append --parallel # -initrd /tmp/tmp.YKFTGTHnwt
SMP: Initializing, found 2 cpus
Now migrate the VM, then press a key to continue...
INFO: migration-cmm: parallel: Migration complete
INFO: migration-cmm: parallel: thread completed 65308 iterations
FAIL: migration-cmm: parallel: during migration: page state mismatch: first 
page idx = 0, addr = 28000, expected_mask = 0x1, actual_mask = 0x2
FAIL: migration-cmm: parallel: after migration: page state mismatch: first 
page idx = 0, addr = 28000, expected_mask = 0x1, actual_mask = 0x2
SUMMARY: 2 tests, 2 unexpected failures

EXIT: STATUS=3

Could you please fix that first?

  Thanks,
   Thomas

Re: [kvm-unit-tests GIT PULL 4/4] s390x: add CMM test during migration

[Nico Boehr]

Quoting Thomas Huth (2023-01-05 13:24:03)
> On 05/01/2023 13.15, Claudio Imbrenda wrote:
> > From: Nico Boehr <nrb@linux.ibm.com>
> > 
> > Add a test which modifies CMM page states while migration is in
> > progress.
> > 
> > This is added to the existing migration-cmm test, which gets a new
> > command line argument for the sequential and parallel variants.
> > 
> > Signed-off-by: Nico Boehr <nrb@linux.ibm.com>
> > Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> > Link: https://lore.kernel.org/r/20221221090953.341247-2-nrb@linux.ibm.com
> > Message-Id: <20221221090953.341247-2-nrb@linux.ibm.com>
> > Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> > ---
> >   s390x/migration-cmm.c | 258 +++++++++++++++++++++++++++++++++++++-----
> >   s390x/unittests.cfg   |  15 ++-
> >   2 files changed, 240 insertions(+), 33 deletions(-)
> 
>   Hi!
> 
> While this works fine on my z15 LPAR, I'm getting a failure when running 
> this test on my z13 LPAR:

I can _sometimes_ reproduce this on z16, z13 and misc older machines. The older the machine, the more often it seems to happen.

I think we may have a bug somewhere. While I investigate, feel free to leave out this patch if you prefer.