Skip to contentSkip to conversation

s390x: Extend instruction interception tests

PATCH Superseded Nico Boehr 2 participants 11 messages Feb 21, 2022 – Feb 23, 2022

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Nico Boehr

author
#0 lore raw

v2 cover s390x: Extend instruction interception tests

This series extends the instruction interception tests for s390x.

For most instructions, there is already coverage in existing tests, but they are not covering some failure cases, e.g. bad alignment. In this case, the existing tests were extended.

SCK was not under test anywhere yet, hence a new test file was added.

The EPSW test gets it's own file, too, because it requires a I/O device, more details in the respective commit.

Changelog from v1:
----
- Reset pmcw flags at test end
- Rebase

Nico Boehr (8):
s390x: Add more tests for MSCH
s390x: Add test for PFMF low-address protection
s390x: Add sck tests
s390x: Add tests for STCRW
s390x: Add more tests for SSCH
s390x: Add more tests for STSCH
s390x: Add tests for TSCH
s390x: Add EPSW test

 lib/s390x/css.h     |  17 +++
 lib/s390x/css_lib.c |  60 ++++++++++
 s390x/Makefile      |   2 +
 s390x/css.c         | 278 ++++++++++++++++++++++++++++++++++++++++++++
 s390x/epsw.c        | 113 ++++++++++++++++++
 s390x/pfmf.c        |  29 +++++
 s390x/sck.c         | 127 ++++++++++++++++++++
 s390x/unittests.cfg |   7 ++
 8 files changed, 633 insertions(+)
 create mode 100644 s390x/epsw.c
 create mode 100644 s390x/sck.c
Signature · 1 line
2.31.1
author
#1 lore raw

PATCH v2 1/8 s390x: Add more tests for MSCH

We already have some coverage for MSCH, but there are more cases to test for:

- invalid SCHIB structure. We cover that by setting reserved bits 0, 1, 6 and 7 in the flags of the PMCW.
This test currently fails because of a QEMU bug, a fix
is available (see "[PATCH qemu] s390x/css: fix PMCW invalid mask")
- a pointer to an unaligned SCHIB. We cover misalignment by 1 and 2 bytes. Using pointer to valid memory avoids messing up random memory in case of test failures.

Here's the QEMU PMCW invalid mask fix: https://lists.nongnu.org/archive/html/qemu-s390x/2021-12/msg00100.html

---
 s390x/css.c | 53 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 53 insertions(+)
s390x/css.c +53 -0
index 881206ba1cef..932daf69bb36 100644
@@ -27,6 +27,8 @@ static int test_device_sid;
 static struct senseid *senseid;
 struct ccw1 *ccw;
 
+char alignment_test_page[PAGE_SIZE] __attribute__((aligned(PAGE_SIZE)));
+
 static void test_enumerate(void)
 {
 	test_device_sid = css_enumerate();
@@ -331,6 +333,56 @@ static void test_schm_fmt1(void)
 	free_io_mem(mb1, sizeof(struct measurement_block_format1));
 }
 
+static void test_msch(void)
+{
+	const int align_to = 4;
+	int cc;
+	int invalid_pmcw_flags[] = {0, 1, 6, 7};
+	int invalid_flag;
+	uint16_t old_pmcw_flags;
+
+	if (!test_device_sid) {
+		report_skip("No device");
+		return;
+	}
+
+	cc = stsch(test_device_sid, &schib);
+	if (cc) {
+		report_fail("stsch: sch %08x failed with cc=%d", test_device_sid, cc);
+		return;
+	}
+
+	report_prefix_push("Unaligned");
+	for (int i = 1; i < align_to; i *= 2) {
+		report_prefix_pushf("%d", i);
+
+		expect_pgm_int();
+		msch(test_device_sid, (struct schib *)(alignment_test_page + i));
+		check_pgm_int_code(PGM_INT_CODE_SPECIFICATION);
+
+		report_prefix_pop();
+	}
+	report_prefix_pop();
+
+	report_prefix_push("Invalid SCHIB");
+	old_pmcw_flags = schib.pmcw.flags;
+	for (int i = 0; i < ARRAY_SIZE(invalid_pmcw_flags); i++) {
+		invalid_flag = invalid_pmcw_flags[i];
+
+		report_prefix_pushf("PMCW flag bit %d set", invalid_flag);
+
+		schib.pmcw.flags = old_pmcw_flags | BIT(15 - invalid_flag);
+		expect_pgm_int();
+		msch(test_device_sid, &schib);
+		check_pgm_int_code(PGM_INT_CODE_OPERAND);
+
+		report_prefix_pop();
+	}
+	report_prefix_pop();
+
+	schib.pmcw.flags = old_pmcw_flags;
+}
+
 static struct {
 	const char *name;
 	void (*func)(void);
@@ -343,6 +395,7 @@ static struct {
 	{ "measurement block (schm)", test_schm },
 	{ "measurement block format0", test_schm_fmt0 },
 	{ "measurement block format1", test_schm_fmt1 },
+	{ "msch", test_msch },
 	{ NULL, NULL }
 };
 
Signature · 1 line
2.31.1
author
#2 lore raw

PATCH v2 2/8 s390x: Add test for PFMF low-address protection

In reply to Nico Boehr

PFMF should respect the low-address protection when clearing pages, hence add some tests for it.

When low-address protection fails, clearing frame 0 is a destructive operation. It messes up interrupts and thus printing test results won't work properly. Hence, we first attempt to clear frame 1 which is not as destructive.

Doing it this way around increases the chances for the user to see a proper failure message instead of QEMU randomly quitting in the middle of the test run.

---
 s390x/pfmf.c | 29 +++++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)
s390x/pfmf.c +29 -0
index 2f3cb110dc4c..aa1305292ee8 100644
@@ -113,6 +113,34 @@ static void test_1m_clear(void)
 	report_prefix_pop();
 }
 
+static void test_low_addr_prot(void)
+{
+	union pfmf_r1 r1 = {
+		.reg.cf = 1,
+		.reg.fsc = PFMF_FSC_4K
+	};
+
+	report_prefix_push("low-address protection");
+
+	report_prefix_push("0x1000");
+	expect_pgm_int();
+	low_prot_enable();
+	pfmf(r1.val, (void *)0x1000);
+	low_prot_disable();
+	check_pgm_int_code(PGM_INT_CODE_PROTECTION);
+	report_prefix_pop();
+
+	report_prefix_push("0x0");
+	expect_pgm_int();
+	low_prot_enable();
+	pfmf(r1.val, 0);
+	low_prot_disable();
+	check_pgm_int_code(PGM_INT_CODE_PROTECTION);
+	report_prefix_pop();
+
+	report_prefix_pop();
+}
+
 int main(void)
 {
 	bool has_edat = test_facility(8);
@@ -124,6 +152,7 @@ int main(void)
 	}
 
 	test_priv();
+	test_low_addr_prot();
 	/* Force the buffer pages in */
 	memset(pagebuf, 0, PAGE_SIZE * 256);
 
Signature · 1 line
2.31.1
author
#3 lore raw

PATCH v2 3/8 s390x: Add sck tests

In reply to Nico Boehr

SCK updates the guest's TOD clock. It needs to be intercepted, hence we should have some tests for it:

- its operand must be 8-byte aligned. We test for 1, 2 and 4 byte misalignment.
- it should actually update the clock value. As guests are unable to stop their clock, we can only check the clock value is at least higher than the value we set.
- it is a privileged call, so expect it to fail when in problem state. We also test the clock value is not changed in this case. We do so by first adjusting it (privileged) to a high value. We then attempt, in nonprivileged mode, to set the clock to a comparatively small value. We can then store the clock value and test its value is strictly higher than the high value we set previously.

---
 s390x/Makefile      |   1 +
 s390x/sck.c         | 127 ++++++++++++++++++++++++++++++++++++++++++++
 s390x/unittests.cfg |   3 ++
 3 files changed, 131 insertions(+)
 create mode 100644 s390x/sck.c
s390x/Makefile +1 -0
index 53b0fe044fe7..a76b78e5a011 100644
@@ -17,6 +17,7 @@ tests += $(TEST_DIR)/stsi.elf
 tests += $(TEST_DIR)/skrf.elf
 tests += $(TEST_DIR)/smp.elf
 tests += $(TEST_DIR)/sclp.elf
+tests += $(TEST_DIR)/sck.elf
 tests += $(TEST_DIR)/css.elf
 tests += $(TEST_DIR)/uv-guest.elf
 tests += $(TEST_DIR)/sie.elf
s390x/sck.c new file +127 -0

127 lines not on this page · raw message

s390x/unittests.cfg +3 -0
index 1600e714c8b9..8b148fe31ac0 100644
@@ -136,3 +136,6 @@ file = firq.elf
 timeout = 20
 extra_params = -smp 1,maxcpus=3 -cpu qemu -device qemu-s390x-cpu,core-id=2 -device qemu-s390x-cpu,core-id=1
 accel = tcg
+
+[sck]
+file = sck.elf
Signature · 1 line
2.31.1
author
#4 lore raw

PATCH v2 4/8 s390x: Add tests for STCRW

In reply to Nico Boehr

The test consists of two parts: First a simple check to ensure we enforce an aligned address. We test misalignment by 1 and 2 bytes.

The second part tests the handling of pending Channel Reports (CR). We first assume no CR is initally pending and check STCRW returns accordingly. Then, we generate a CR by resetting a Channel Path using RCHP and make sure this results in exactly one CRW being generated which has a Reporting-Source Code (RSC) corresponding to the Channel Path facility.

---
 lib/s390x/css.h     | 17 ++++++++++
 lib/s390x/css_lib.c | 60 +++++++++++++++++++++++++++++++++
 s390x/css.c         | 82 +++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 159 insertions(+)
lib/s390x/css.h +17 -0
index 0db8a28166f6..a6a68577248b 100644
@@ -266,6 +266,20 @@ static inline int rchp(unsigned long chpid)
 	return cc;
 }
 
+static inline int stcrw(uint32_t *crw)
+{
+	int cc;
+
+	asm volatile(
+		"	stcrw	%[crw]\n"
+		"	ipm	%[cc]\n"
+		"	srl	%[cc],28"
+		: [cc] "=d" (cc)
+		: [crw] "Q" (*crw)
+		: "cc", "memory");
+	return cc;
+}
+
 /* Debug functions */
 char *dump_pmcw_flags(uint16_t f);
 char *dump_scsw_flags(uint32_t f);
@@ -294,6 +308,9 @@ int css_residual_count(unsigned int schid);
 void enable_io_isc(uint8_t isc);
 int wait_and_check_io_completion(int schid);
 
+int css_find_installed_chpid(int sid, uint8_t *chpid_out);
+int css_generate_crw(int sid);
+
 /*
  * CHSC definitions
  */
lib/s390x/css_lib.c +60 -0

63 lines not on this page · raw message

s390x/css.c +82 -0

94 lines not on this page · raw message

Signature · 1 line
2.31.1
author
#5 lore raw

PATCH v2 5/8 s390x: Add more tests for SSCH

In reply to Nico Boehr

We already have some coverage of SSCH due to its use in start_ccw1_chain() in css_lib, but two more cases deserve our attention:

- unaligned operand address. We check for misalignment by 1 and 2 bytes.
- an invalid ORB structure.

---
 s390x/css.c | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)
s390x/css.c +33 -0
index 6a2db79c8097..a90a0cd64e2b 100644
@@ -464,6 +464,38 @@ static void test_stcrw(void)
 	report_prefix_pop();
 }
 
+static void test_ssch(void)
+{
+	const int align_to = 4;
+	struct orb orb;
+
+	if (!test_device_sid) {
+		report_skip("No device");
+		return;
+	}
+
+	report_prefix_push("Unaligned");
+	for (int i = 1; i < align_to; i *= 2) {
+		report_prefix_pushf("%d", i);
+
+		expect_pgm_int();
+		ssch(test_device_sid, (struct orb *)(alignment_test_page + i));
+		check_pgm_int_code(PGM_INT_CODE_SPECIFICATION);
+
+		report_prefix_pop();
+	}
+	report_prefix_pop();
+
+	report_prefix_push("Invalid ORB");
+
+	memset(&orb, 0xff, sizeof(orb));
+	expect_pgm_int();
+	ssch(test_device_sid, &orb);
+	check_pgm_int_code(PGM_INT_CODE_OPERAND);
+
+	report_prefix_pop();
+}
+
 static struct {
 	const char *name;
 	void (*func)(void);
@@ -478,6 +510,7 @@ static struct {
 	{ "measurement block format1", test_schm_fmt1 },
 	{ "msch", test_msch },
 	{ "stcrw", test_stcrw },
+	{ "ssch", test_ssch },
 	{ NULL, NULL }
 };
 
Signature · 1 line
2.31.1
author
#6 lore raw

PATCH v2 6/8 s390x: Add more tests for STSCH

In reply to Nico Boehr

css_lib extensively uses STSCH, but two more cases deserve their own tests:

- unaligned address for SCHIB. We check for misalignment by 1 and 2 bytes.
- channel not operational
- bit 47 in SID not set
- bit 5 of PMCW flags.
As per the principles of operation, bit 5 of the PMCW flags shall be ignored by msch and always stored as zero by stsch.

Currently, QEMU requires this bit to always be zero on msch, which is why this test currently fails. A fix was posted upstream ("[PATCH qemu] s390x/css: fix PMCW invalid mask").

Here's the QEMU PMCW invalid mask fix: https://lists.nongnu.org/archive/html/qemu-s390x/2021-12/msg00100.html

---
 s390x/css.c | 74 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 74 insertions(+)
s390x/css.c +74 -0
index a90a0cd64e2b..021eb12573c0 100644
@@ -496,6 +496,78 @@ static void test_ssch(void)
 	report_prefix_pop();
 }
 
+static void test_stsch(void)
+{
+	const int align_to = 4;
+	struct schib schib;
+	int cc;
+
+	if (!test_device_sid) {
+		report_skip("No device");
+		return;
+	}
+
+	report_prefix_push("Unaligned");
+	for (int i = 1; i < align_to; i *= 2) {
+		report_prefix_pushf("%d", i);
+
+		expect_pgm_int();
+		stsch(test_device_sid, (struct schib *)(alignment_test_page + i));
+		check_pgm_int_code(PGM_INT_CODE_SPECIFICATION);
+
+		report_prefix_pop();
+	}
+	report_prefix_pop();
+
+	report_prefix_push("Invalid subchannel number");
+	cc = stsch(0x0001ffff, &schib);
+	report(cc == 3, "Channel not operational");
+	report_prefix_pop();
+
+	report_prefix_push("Bit 47 in SID is zero");
+	expect_pgm_int();
+	stsch(0x0000ffff, &schib);
+	check_pgm_int_code(PGM_INT_CODE_OPERAND);
+	report_prefix_pop();
+}
+
+static void test_pmcw_bit5(void)
+{
+	int cc;
+	uint16_t old_pmcw_flags;
+
+	cc = stsch(test_device_sid, &schib);
+	if (cc) {
+		report_fail("stsch: sch %08x failed with cc=%d", test_device_sid, cc);
+		return;
+	}
+	old_pmcw_flags = schib.pmcw.flags;
+
+	report_prefix_push("Bit 5 set");
+
+	schib.pmcw.flags = old_pmcw_flags | BIT(15 - 5);
+	cc = msch(test_device_sid, &schib);
+	report(!cc, "MSCH cc == 0");
+
+	cc = stsch(test_device_sid, &schib);
+	report(!cc, "STSCH cc == 0");
+	report(!(schib.pmcw.flags & BIT(15 - 5)), "STSCH PMCW Bit 5 is clear");
+
+	report_prefix_pop();
+
+	report_prefix_push("Bit 5 clear");
+
+	schib.pmcw.flags = old_pmcw_flags & ~BIT(15 - 5);
+	cc = msch(test_device_sid, &schib);
+	report(!cc, "MSCH cc == 0");
+
+	cc = stsch(test_device_sid, &schib);
+	report(!cc, "STSCH cc == 0");
+	report(!(schib.pmcw.flags & BIT(15 - 5)), "STSCH PMCW Bit 5 is clear");
+
+	report_prefix_pop();
+}
+
 static struct {
 	const char *name;
 	void (*func)(void);
@@ -511,6 +583,8 @@ static struct {
 	{ "msch", test_msch },
 	{ "stcrw", test_stcrw },
 	{ "ssch", test_ssch },
+	{ "stsch", test_stsch },
+	{ "pmcw bit 5 ignored", test_pmcw_bit5 },
 	{ NULL, NULL }
 };
 
Signature · 1 line
2.31.1
author
#7 lore raw

PATCH v2 7/8 s390x: Add tests for TSCH

In reply to Nico Boehr

TSCH has two special cases which need to be tested:

- unaligned address. We test for misalignment by 1 and 2 bytes.
- channel not operational
- bit 47 in SID not set

---
 s390x/css.c | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
s390x/css.c +36 -0
index 021eb12573c0..b2e027c99848 100644
@@ -568,6 +568,41 @@ static void test_pmcw_bit5(void)
 	report_prefix_pop();
 }
 
+static void test_tsch(void)
+{
+	const int align_to = 4;
+	struct irb irb;
+	int cc;
+
+	if (!test_device_sid) {
+		report_skip("No device");
+		return;
+	}
+
+	report_prefix_push("Unaligned");
+	for (int i = 1; i < align_to; i *= 2) {
+		report_prefix_pushf("%d", i);
+
+		expect_pgm_int();
+		tsch(test_device_sid, (struct irb *)(alignment_test_page + i));
+		check_pgm_int_code(PGM_INT_CODE_SPECIFICATION);
+
+		report_prefix_pop();
+	}
+	report_prefix_pop();
+
+	report_prefix_push("Invalid subchannel number");
+	cc = tsch(0x0001ffff, &irb);
+	report(cc == 3, "Channel not operational");
+	report_prefix_pop();
+
+	report_prefix_push("Bit 47 in SID is zero");
+	expect_pgm_int();
+	tsch(0x0000ffff, &irb);
+	check_pgm_int_code(PGM_INT_CODE_OPERAND);
+	report_prefix_pop();
+}
+
 static struct {
 	const char *name;
 	void (*func)(void);
@@ -585,6 +620,7 @@ static struct {
 	{ "ssch", test_ssch },
 	{ "stsch", test_stsch },
 	{ "pmcw bit 5 ignored", test_pmcw_bit5 },
+	{ "tsch", test_tsch },
 	{ NULL, NULL }
 };
 
Signature · 1 line
2.31.1
author
#8 lore raw

PATCH v2 8/8 s390x: Add EPSW test

In reply to Nico Boehr

EPSW is only intercepted in certain cases. One of these cases is if we have a CRW pending and machine check interrupts are masked. This can be achieved by issuing a RCHP on a valid channel path. This is why we need the CSS lib and an IO device in this test and hence need to skip it when not running under QEMU.

Three special cases deserve our attention:

- upper 32 bits of both operands are never modified,
- second operand is not modified if it is zero.
- when both operands are zero, bits 0-11 and 13-31 of the PSW are stored in r0.

We also verify we get the correct contents when the second operand is zero. To do so, we save the data stored at the first operand in the first case as a reference. As we don't mess with the PSW, the only thing that might change is the Condition Code (CC) due to some instruction in between, so we zero it out using zero_out_cc_from_epsw_op1().

This test must be fenced when running in non-QEMU.

---
 s390x/Makefile      |   1 +
 s390x/epsw.c        | 113 ++++++++++++++++++++++++++++++++++++++++++++
 s390x/unittests.cfg |   4 ++
 3 files changed, 118 insertions(+)
 create mode 100644 s390x/epsw.c
s390x/Makefile +1 -0
index a76b78e5a011..25449708da0d 100644
@@ -27,6 +27,7 @@ tests += $(TEST_DIR)/edat.elf
 tests += $(TEST_DIR)/mvpg-sie.elf
 tests += $(TEST_DIR)/spec_ex-sie.elf
 tests += $(TEST_DIR)/firq.elf
+tests += $(TEST_DIR)/epsw.elf
 
 pv-tests += $(TEST_DIR)/pv-diags.elf
 
s390x/epsw.c new file +113 -0

113 lines not on this page · raw message

s390x/unittests.cfg +4 -0
index 8b148fe31ac0..aeb82246dddf 100644
@@ -139,3 +139,7 @@ accel = tcg
 
 [sck]
 file = sck.elf
+
+[epsw]
+file = epsw.elf
+extra_params = -device virtio-net-ccw
Signature · 1 line
2.31.1
#9 lore raw
In reply to Nico Boehr
Quoted 10 lines · Nico Boehr

This series extends the instruction interception tests for s390x.

For most instructions, there is already coverage in existing tests, but they are not covering some failure cases, e.g. bad alignment. In this case, the existing tests were extended.

SCK was not under test anywhere yet, hence a new test file was added.

The EPSW test gets it's own file, too, because it requires a I/O device, more details in the respective commit.

Could you please push this to devel so we can get CI data?

To me it seems like only STSCH needs review so we should be done with this soonish.

Quoted 28 lines

Changelog from v1:
----
- Reset pmcw flags at test end
- Rebase

Nico Boehr (8):
s390x: Add more tests for MSCH
s390x: Add test for PFMF low-address protection
s390x: Add sck tests
s390x: Add tests for STCRW
s390x: Add more tests for SSCH
s390x: Add more tests for STSCH
s390x: Add tests for TSCH
s390x: Add EPSW test

  lib/s390x/css.h     |  17 +++
  lib/s390x/css_lib.c |  60 ++++++++++
  s390x/Makefile      |   2 +
  s390x/css.c         | 278 ++++++++++++++++++++++++++++++++++++++++++++
  s390x/epsw.c        | 113 ++++++++++++++++++
  s390x/pfmf.c        |  29 +++++
  s390x/sck.c         | 127 ++++++++++++++++++++
  s390x/unittests.cfg |   7 ++
  8 files changed, 633 insertions(+)

create mode 100644 s390x/epsw.c
create mode 100644 s390x/sck.c

author
#10 lore raw
Quoted 1 line · Janosch Frank

Could you please push this to devel so we can get CI data?

Thanks Janosch, we found an issue in the CI. A fix is in the making.