Skip to contentSkip to conversation

KVM: selftests: aarch64: page_fault_test S1PTW related fixes

PATCH Applied Ricardo Koller 3 participants 9 messages Jan 27, 2023 – Jan 29, 2023

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v2 cover KVM: selftests: aarch64: page_fault_test S1PTW related fixes

Commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots") changed the way S1PTW faults were handled by KVM. Before this fix, KVM treated any fault due to any S1PTW as a write, even S1PTW that didn't update the PTE. Some tests in page_fault_test mistakenly check for this KVM behavior and are currently failing. For example, there's a dirty log test that asserts that a S1PTW without AF or DA results in the PTE page getting dirty.

The first commit fixes the userfaultfd check by relaxing all read vs. write checks. The second commit fixes the dirtylog tests by only asserting dirty bits when the AF bit is set. The third commit fixes an issue found after fixing the previous two: the dirty log test was checking for the first page in the PT region. Finally, commit "KVM:
arm64: Fix S1PTW handling on RO memslots" allows for having readonly memslots holding page tables, so commit 4 add tests for it.

Changes from v1:
- added sha1's for commit "KVM: arm64: Fix S1PTW handling on RO memslots" in all messages. (Oliver)
- removed _with_af from RO macro. (Oliver)

Ricardo Koller (4):
KVM: selftests: aarch64: Relax userfaultfd read vs. write checks
KVM: selftests: aarch64: Do not default to dirty PTE pages on all
S1PTWs
KVM: selftests: aarch64: Fix check of dirty log PT write
KVM: selftests: aarch64: Test read-only PT memory regions

 .../selftests/kvm/aarch64/page_fault_test.c   | 187 ++++++++++--------
 1 file changed, 103 insertions(+), 84 deletions(-)
Signature · 1 line
2.39.1.456.gfc5497dd1b-goog
#1 lore raw

PATCH v2 1/4 KVM: selftests: aarch64: Relax userfaultfd read vs. write checks

Only Stage1 Page table walks (S1PTW) writing a PTE on an unmapped page should result in a userfaultfd write. However, the userfaultfd tests in page_fault_test wrongly assert that any S1PTW is a PTE write.

Fix this by relaxing the read vs. write checks in all userfaultfd handlers. Note that this is also an attempt to focus less on KVM (and userfaultfd) behavior, and more on architectural behavior. Also note that after commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots"), the userfaultfd fault (S1PTW with AF on an unmaped PTE page) is actually a read: the translation fault that comes before the permission fault.

---
 .../selftests/kvm/aarch64/page_fault_test.c   | 83 ++++++++-----------
 1 file changed, 34 insertions(+), 49 deletions(-)
tools/testing/selftests/kvm/aarch64/page_fault_test.c +34 -49
index beb944fa6fd4..0dda58766185 100644
@@ -304,7 +304,7 @@ static struct uffd_args {
 
 /* Returns true to continue the test, and false if it should be skipped. */
 static int uffd_generic_handler(int uffd_mode, int uffd, struct uffd_msg *msg,
-				struct uffd_args *args, bool expect_write)
+				struct uffd_args *args)
 {
 	uint64_t addr = msg->arg.pagefault.address;
 	uint64_t flags = msg->arg.pagefault.flags;
@@ -313,7 +313,6 @@ static int uffd_generic_handler(int uffd_mode, int uffd, struct uffd_msg *msg,
 
 	TEST_ASSERT(uffd_mode == UFFDIO_REGISTER_MODE_MISSING,
 		    "The only expected UFFD mode is MISSING");
-	ASSERT_EQ(!!(flags & UFFD_PAGEFAULT_FLAG_WRITE), expect_write);
 	ASSERT_EQ(addr, (uint64_t)args->hva);
 
 	pr_debug("uffd fault: addr=%p write=%d\n",
@@ -337,19 +336,14 @@ static int uffd_generic_handler(int uffd_mode, int uffd, struct uffd_msg *msg,
 	return 0;
 }
 
-static int uffd_pt_write_handler(int mode, int uffd, struct uffd_msg *msg)
+static int uffd_pt_handler(int mode, int uffd, struct uffd_msg *msg)
 {
-	return uffd_generic_handler(mode, uffd, msg, &pt_args, true);
+	return uffd_generic_handler(mode, uffd, msg, &pt_args);
 }
 
-static int uffd_data_write_handler(int mode, int uffd, struct uffd_msg *msg)
+static int uffd_data_handler(int mode, int uffd, struct uffd_msg *msg)
 {
-	return uffd_generic_handler(mode, uffd, msg, &data_args, true);
-}
-
-static int uffd_data_read_handler(int mode, int uffd, struct uffd_msg *msg)
-{
-	return uffd_generic_handler(mode, uffd, msg, &data_args, false);
+	return uffd_generic_handler(mode, uffd, msg, &data_args);
 }
 
 static void setup_uffd_args(struct userspace_mem_region *region,
@@ -822,7 +816,7 @@ static void help(char *name)
 	.mem_mark_cmd		= CMD_HOLE_DATA | CMD_HOLE_PT,			\
 	.guest_test_check	= { _CHECK(_with_af), _test_check },		\
 	.uffd_data_handler	= _uffd_data_handler,				\
-	.uffd_pt_handler	= uffd_pt_write_handler,			\
+	.uffd_pt_handler	= uffd_pt_handler,				\
 	.expected_events	= { .uffd_faults = _uffd_faults, },		\
 }
 
@@ -878,7 +872,7 @@ static void help(char *name)
 	.guest_prepare		= { _PREPARE(_access) },			\
 	.guest_test		= _access,					\
 	.uffd_data_handler	= _uffd_data_handler,				\
-	.uffd_pt_handler	= uffd_pt_write_handler,			\
+	.uffd_pt_handler	= uffd_pt_handler,				\
 	.mmio_handler		= _mmio_handler,				\
 	.expected_events	= { .mmio_exits = _mmio_exits,			\
 				    .uffd_faults = _uffd_faults },		\
@@ -892,7 +886,7 @@ static void help(char *name)
 	.mem_mark_cmd		= CMD_HOLE_DATA | CMD_HOLE_PT,			\
 	.guest_test		= _access,					\
 	.uffd_data_handler	= _uffd_data_handler,				\
-	.uffd_pt_handler	= uffd_pt_write_handler,			\
+	.uffd_pt_handler	= uffd_pt_handler,			\
 	.fail_vcpu_run_handler	= fail_vcpu_run_mmio_no_syndrome_handler,	\
 	.expected_events	= { .fail_vcpu_runs = 1,			\
 				    .uffd_faults = _uffd_faults },		\
@@ -933,29 +927,27 @@ static struct test_desc tests[] = {
 	 * (S1PTW).
 	 */
 	TEST_UFFD(guest_read64, with_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_read_handler, uffd_pt_write_handler, 2),
-	/* no_af should also lead to a PT write. */
+		  uffd_data_handler, uffd_pt_handler, 2),
 	TEST_UFFD(guest_read64, no_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_read_handler, uffd_pt_write_handler, 2),
-	/* Note how that cas invokes the read handler. */
+		  uffd_data_handler, uffd_pt_handler, 2),
 	TEST_UFFD(guest_cas, with_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_read_handler, uffd_pt_write_handler, 2),
+		  uffd_data_handler, uffd_pt_handler, 2),
 	/*
 	 * Can't test guest_at with_af as it's IMPDEF whether the AF is set.
 	 * The S1PTW fault should still be marked as a write.
 	 */
 	TEST_UFFD(guest_at, no_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_read_handler, uffd_pt_write_handler, 1),
+		  uffd_no_handler, uffd_pt_handler, 1),
 	TEST_UFFD(guest_ld_preidx, with_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_read_handler, uffd_pt_write_handler, 2),
+		  uffd_data_handler, uffd_pt_handler, 2),
 	TEST_UFFD(guest_write64, with_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_write_handler, uffd_pt_write_handler, 2),
+		  uffd_data_handler, uffd_pt_handler, 2),
 	TEST_UFFD(guest_dc_zva, with_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_write_handler, uffd_pt_write_handler, 2),
+		  uffd_data_handler, uffd_pt_handler, 2),
 	TEST_UFFD(guest_st_preidx, with_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_write_handler, uffd_pt_write_handler, 2),
+		  uffd_data_handler, uffd_pt_handler, 2),
 	TEST_UFFD(guest_exec, with_af, CMD_HOLE_DATA | CMD_HOLE_PT,
-		  uffd_data_read_handler, uffd_pt_write_handler, 2),
+		  uffd_data_handler, uffd_pt_handler, 2),
 
 	/*
 	 * Try accesses when the data and PT memory regions are both
@@ -980,25 +972,25 @@ static struct test_desc tests[] = {
 	 * fault, and nothing in the dirty log.  Any S1PTW should result in
 	 * a write in the dirty log and a userfaultfd write.
 	 */
-	TEST_UFFD_AND_DIRTY_LOG(guest_read64, with_af, uffd_data_read_handler, 2,
+	TEST_UFFD_AND_DIRTY_LOG(guest_read64, with_af, uffd_data_handler, 2,
 				guest_check_no_write_in_dirty_log),
 	/* no_af should also lead to a PT write. */
-	TEST_UFFD_AND_DIRTY_LOG(guest_read64, no_af, uffd_data_read_handler, 2,
+	TEST_UFFD_AND_DIRTY_LOG(guest_read64, no_af, uffd_data_handler, 2,
 				guest_check_no_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_ld_preidx, with_af, uffd_data_read_handler,
+	TEST_UFFD_AND_DIRTY_LOG(guest_ld_preidx, with_af, uffd_data_handler,
 				2, guest_check_no_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_at, with_af, 0, 1,
+	TEST_UFFD_AND_DIRTY_LOG(guest_at, with_af, uffd_no_handler, 1,
 				guest_check_no_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_exec, with_af, uffd_data_read_handler, 2,
+	TEST_UFFD_AND_DIRTY_LOG(guest_exec, with_af, uffd_data_handler, 2,
 				guest_check_no_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_write64, with_af, uffd_data_write_handler,
+	TEST_UFFD_AND_DIRTY_LOG(guest_write64, with_af, uffd_data_handler,
 				2, guest_check_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_cas, with_af, uffd_data_read_handler, 2,
+	TEST_UFFD_AND_DIRTY_LOG(guest_cas, with_af, uffd_data_handler, 2,
 				guest_check_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_dc_zva, with_af, uffd_data_write_handler,
+	TEST_UFFD_AND_DIRTY_LOG(guest_dc_zva, with_af, uffd_data_handler,
 				2, guest_check_write_in_dirty_log),
 	TEST_UFFD_AND_DIRTY_LOG(guest_st_preidx, with_af,
-				uffd_data_write_handler, 2,
+				uffd_data_handler, 2,
 				guest_check_write_in_dirty_log),
 
 	/*
@@ -1051,22 +1043,15 @@ static struct test_desc tests[] = {
 	 * no userfaultfd write fault. Reads result in userfaultfd getting
 	 * triggered.
 	 */
-	TEST_RO_MEMSLOT_AND_UFFD(guest_read64, 0, 0,
-				 uffd_data_read_handler, 2),
-	TEST_RO_MEMSLOT_AND_UFFD(guest_ld_preidx, 0, 0,
-				 uffd_data_read_handler, 2),
-	TEST_RO_MEMSLOT_AND_UFFD(guest_at, 0, 0,
-				 uffd_no_handler, 1),
-	TEST_RO_MEMSLOT_AND_UFFD(guest_exec, 0, 0,
-				 uffd_data_read_handler, 2),
+	TEST_RO_MEMSLOT_AND_UFFD(guest_read64, 0, 0, uffd_data_handler, 2),
+	TEST_RO_MEMSLOT_AND_UFFD(guest_ld_preidx, 0, 0, uffd_data_handler, 2),
+	TEST_RO_MEMSLOT_AND_UFFD(guest_at, 0, 0, uffd_no_handler, 1),
+	TEST_RO_MEMSLOT_AND_UFFD(guest_exec, 0, 0, uffd_data_handler, 2),
 	TEST_RO_MEMSLOT_AND_UFFD(guest_write64, mmio_on_test_gpa_handler, 1,
-				 uffd_data_write_handler, 2),
-	TEST_RO_MEMSLOT_NO_SYNDROME_AND_UFFD(guest_cas,
-					     uffd_data_read_handler, 2),
-	TEST_RO_MEMSLOT_NO_SYNDROME_AND_UFFD(guest_dc_zva,
-					     uffd_no_handler, 1),
-	TEST_RO_MEMSLOT_NO_SYNDROME_AND_UFFD(guest_st_preidx,
-					     uffd_no_handler, 1),
+				 uffd_data_handler, 2),
+	TEST_RO_MEMSLOT_NO_SYNDROME_AND_UFFD(guest_cas, uffd_data_handler, 2),
+	TEST_RO_MEMSLOT_NO_SYNDROME_AND_UFFD(guest_dc_zva, uffd_no_handler, 1),
+	TEST_RO_MEMSLOT_NO_SYNDROME_AND_UFFD(guest_st_preidx, uffd_no_handler, 1),
 
 	{ 0 }
 };
Signature · 1 line
2.39.1.456.gfc5497dd1b-goog
#2 lore raw

PATCH v2 2/4 KVM: selftests: aarch64: Do not default to dirty PTE pages on all S1PTWs

In reply to Ricardo Koller

Only Stage1 Page table walks (S1PTW) trying to write into a PTE should result in the PTE page being dirty in the log. However, the dirty log tests in page_fault_test default to treat all S1PTW accesses as writes. Fix the relevant tests by asserting dirty pages only for S1PTW writes, which in these tests only applies to when Hardware management of the Access Flag is enabled.

---
 .../selftests/kvm/aarch64/page_fault_test.c   | 93 ++++++++++++-------
 1 file changed, 60 insertions(+), 33 deletions(-)
tools/testing/selftests/kvm/aarch64/page_fault_test.c +60 -33
index 0dda58766185..1a3bb2bd8657 100644
@@ -237,6 +237,11 @@ static void guest_check_s1ptw_wr_in_dirty_log(void)
 	GUEST_SYNC(CMD_CHECK_S1PTW_WR_IN_DIRTY_LOG);
 }
 
+static void guest_check_no_s1ptw_wr_in_dirty_log(void)
+{
+	GUEST_SYNC(CMD_CHECK_NO_S1PTW_WR_IN_DIRTY_LOG);
+}
+
 static void guest_exec(void)
 {
 	int (*code)(void) = (int (*)(void))TEST_EXEC_GVA;
@@ -791,7 +796,7 @@ static void help(char *name)
 	.expected_events	= { .uffd_faults = _uffd_faults, },		\
 }
 
-#define TEST_DIRTY_LOG(_access, _with_af, _test_check)				\
+#define TEST_DIRTY_LOG(_access, _with_af, _test_check, _pt_check)		\
 {										\
 	.name			= SCAT3(dirty_log, _access, _with_af),		\
 	.data_memslot_flags	= KVM_MEM_LOG_DIRTY_PAGES,			\
@@ -799,13 +804,12 @@ static void help(char *name)
 	.guest_prepare		= { _PREPARE(_with_af),				\
 				    _PREPARE(_access) },			\
 	.guest_test		= _access,					\
-	.guest_test_check	= { _CHECK(_with_af), _test_check,		\
-				    guest_check_s1ptw_wr_in_dirty_log},		\
+	.guest_test_check	= { _CHECK(_with_af), _test_check, _pt_check },	\
 	.expected_events	= { 0 },					\
 }
 
 #define TEST_UFFD_AND_DIRTY_LOG(_access, _with_af, _uffd_data_handler,		\
-				_uffd_faults, _test_check)			\
+				_uffd_faults, _test_check, _pt_check)		\
 {										\
 	.name			= SCAT3(uffd_and_dirty_log, _access, _with_af),	\
 	.data_memslot_flags	= KVM_MEM_LOG_DIRTY_PAGES,			\
@@ -814,7 +818,7 @@ static void help(char *name)
 				    _PREPARE(_access) },			\
 	.guest_test		= _access,					\
 	.mem_mark_cmd		= CMD_HOLE_DATA | CMD_HOLE_PT,			\
-	.guest_test_check	= { _CHECK(_with_af), _test_check },		\
+	.guest_test_check	= { _CHECK(_with_af), _test_check, _pt_check },	\
 	.uffd_data_handler	= _uffd_data_handler,				\
 	.uffd_pt_handler	= uffd_pt_handler,				\
 	.expected_events	= { .uffd_faults = _uffd_faults, },		\
@@ -953,16 +957,25 @@ static struct test_desc tests[] = {
 	 * Try accesses when the data and PT memory regions are both
 	 * tracked for dirty logging.
 	 */
-	TEST_DIRTY_LOG(guest_read64, with_af, guest_check_no_write_in_dirty_log),
-	/* no_af should also lead to a PT write. */
-	TEST_DIRTY_LOG(guest_read64, no_af, guest_check_no_write_in_dirty_log),
-	TEST_DIRTY_LOG(guest_ld_preidx, with_af, guest_check_no_write_in_dirty_log),
-	TEST_DIRTY_LOG(guest_at, no_af, guest_check_no_write_in_dirty_log),
-	TEST_DIRTY_LOG(guest_exec, with_af, guest_check_no_write_in_dirty_log),
-	TEST_DIRTY_LOG(guest_write64, with_af, guest_check_write_in_dirty_log),
-	TEST_DIRTY_LOG(guest_cas, with_af, guest_check_write_in_dirty_log),
-	TEST_DIRTY_LOG(guest_dc_zva, with_af, guest_check_write_in_dirty_log),
-	TEST_DIRTY_LOG(guest_st_preidx, with_af, guest_check_write_in_dirty_log),
+	TEST_DIRTY_LOG(guest_read64, with_af, guest_check_no_write_in_dirty_log,
+		       guest_check_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_read64, no_af, guest_check_no_write_in_dirty_log,
+		       guest_check_no_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_ld_preidx, with_af,
+		       guest_check_no_write_in_dirty_log,
+		       guest_check_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_at, no_af, guest_check_no_write_in_dirty_log,
+		       guest_check_no_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_exec, with_af, guest_check_no_write_in_dirty_log,
+		       guest_check_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_write64, with_af, guest_check_write_in_dirty_log,
+		       guest_check_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_cas, with_af, guest_check_write_in_dirty_log,
+		       guest_check_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_dc_zva, with_af, guest_check_write_in_dirty_log,
+		       guest_check_s1ptw_wr_in_dirty_log),
+	TEST_DIRTY_LOG(guest_st_preidx, with_af, guest_check_write_in_dirty_log,
+		       guest_check_s1ptw_wr_in_dirty_log),
 
 	/*
 	 * Access when the data and PT memory regions are both marked for
@@ -972,27 +985,41 @@ static struct test_desc tests[] = {
 	 * fault, and nothing in the dirty log.  Any S1PTW should result in
 	 * a write in the dirty log and a userfaultfd write.
 	 */
-	TEST_UFFD_AND_DIRTY_LOG(guest_read64, with_af, uffd_data_handler, 2,
-				guest_check_no_write_in_dirty_log),
-	/* no_af should also lead to a PT write. */
-	TEST_UFFD_AND_DIRTY_LOG(guest_read64, no_af, uffd_data_handler, 2,
-				guest_check_no_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_ld_preidx, with_af, uffd_data_handler,
-				2, guest_check_no_write_in_dirty_log),
+	TEST_UFFD_AND_DIRTY_LOG(guest_read64, with_af,
+				uffd_data_handler, 2,
+				guest_check_no_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
+	TEST_UFFD_AND_DIRTY_LOG(guest_read64, no_af,
+				uffd_data_handler, 2,
+				guest_check_no_write_in_dirty_log,
+				guest_check_no_s1ptw_wr_in_dirty_log),
+	TEST_UFFD_AND_DIRTY_LOG(guest_ld_preidx, with_af,
+				uffd_data_handler,
+				2, guest_check_no_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
 	TEST_UFFD_AND_DIRTY_LOG(guest_at, with_af, uffd_no_handler, 1,
-				guest_check_no_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_exec, with_af, uffd_data_handler, 2,
-				guest_check_no_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_write64, with_af, uffd_data_handler,
-				2, guest_check_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_cas, with_af, uffd_data_handler, 2,
-				guest_check_write_in_dirty_log),
-	TEST_UFFD_AND_DIRTY_LOG(guest_dc_zva, with_af, uffd_data_handler,
-				2, guest_check_write_in_dirty_log),
+				guest_check_no_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
+	TEST_UFFD_AND_DIRTY_LOG(guest_exec, with_af,
+				uffd_data_handler, 2,
+				guest_check_no_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
+	TEST_UFFD_AND_DIRTY_LOG(guest_write64, with_af,
+				uffd_data_handler,
+				2, guest_check_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
+	TEST_UFFD_AND_DIRTY_LOG(guest_cas, with_af,
+				uffd_data_handler, 2,
+				guest_check_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
+	TEST_UFFD_AND_DIRTY_LOG(guest_dc_zva, with_af,
+				uffd_data_handler,
+				2, guest_check_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
 	TEST_UFFD_AND_DIRTY_LOG(guest_st_preidx, with_af,
 				uffd_data_handler, 2,
-				guest_check_write_in_dirty_log),
-
+				guest_check_write_in_dirty_log,
+				guest_check_s1ptw_wr_in_dirty_log),
 	/*
 	 * Try accesses when the data memory region is marked read-only
 	 * (with KVM_MEM_READONLY). Writes with a syndrome result in an
Signature · 1 line
2.39.1.456.gfc5497dd1b-goog
#3 lore raw

PATCH v2 3/4 KVM: selftests: aarch64: Fix check of dirty log PT write

In reply to Ricardo Koller

The dirty log checks are mistakenly testing the first page in the page table (PT) memory region instead of the page holding the test data page PTE. This wasn't an issue before commit 406504c7b040 ("KVM:
arm64: Fix S1PTW handling on RO memslots") as all PT pages (including the first page) were treated as writes.

Fix the page_fault_test dirty logging tests by checking for the right
page: the one for the PTE of the data test page.

Fixes a4edf25b3e25 ("KVM: selftests: aarch64: Add dirty logging tests into page_fault_test")
---
 tools/testing/selftests/kvm/aarch64/page_fault_test.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
tools/testing/selftests/kvm/aarch64/page_fault_test.c +5 -2
index 1a3bb2bd8657..2e2178a7d0d8 100644
@@ -470,9 +470,12 @@ static bool handle_cmd(struct kvm_vm *vm, int cmd)
 {
 	struct userspace_mem_region *data_region, *pt_region;
 	bool continue_test = true;
+	uint64_t pte_gpa, pte_pg;
 
 	data_region = vm_get_mem_region(vm, MEM_REGION_TEST_DATA);
 	pt_region = vm_get_mem_region(vm, MEM_REGION_PT);
+	pte_gpa = addr_hva2gpa(vm, virt_get_pte_hva(vm, TEST_GVA));
+	pte_pg = (pte_gpa - pt_region->region.guest_phys_addr) / getpagesize();
 
 	if (cmd == CMD_SKIP_TEST)
 		continue_test = false;
@@ -485,13 +488,13 @@ static bool handle_cmd(struct kvm_vm *vm, int cmd)
 		TEST_ASSERT(check_write_in_dirty_log(vm, data_region, 0),
 			    "Missing write in dirty log");
 	if (cmd & CMD_CHECK_S1PTW_WR_IN_DIRTY_LOG)
-		TEST_ASSERT(check_write_in_dirty_log(vm, pt_region, 0),
+		TEST_ASSERT(check_write_in_dirty_log(vm, pt_region, pte_pg),
 			    "Missing s1ptw write in dirty log");
 	if (cmd & CMD_CHECK_NO_WRITE_IN_DIRTY_LOG)
 		TEST_ASSERT(!check_write_in_dirty_log(vm, data_region, 0),
 			    "Unexpected write in dirty log");
 	if (cmd & CMD_CHECK_NO_S1PTW_WR_IN_DIRTY_LOG)
-		TEST_ASSERT(!check_write_in_dirty_log(vm, pt_region, 0),
+		TEST_ASSERT(!check_write_in_dirty_log(vm, pt_region, pte_pg),
 			    "Unexpected s1ptw write in dirty log");
 
 	return continue_test;
Signature · 1 line
2.39.1.456.gfc5497dd1b-goog
#4 lore raw

PATCH v2 4/4 KVM: selftests: aarch64: Test read-only PT memory regions

In reply to Ricardo Koller

Extend the read-only memslot tests in page_fault_test to test read-only PT (Page table) memslots. Note that this was not allowed before commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots") as all S1PTW faults were treated as writes which resulted in an (unrecoverable) exception inside the guest.

---
 .../selftests/kvm/aarch64/page_fault_test.c    | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)
tools/testing/selftests/kvm/aarch64/page_fault_test.c +11 -7
index 2e2178a7d0d8..54680dc5887f 100644
@@ -829,8 +829,9 @@ static void help(char *name)
 
 #define TEST_RO_MEMSLOT(_access, _mmio_handler, _mmio_exits)			\
 {										\
-	.name			= SCAT3(ro_memslot, _access, _with_af),		\
+	.name			= SCAT2(ro_memslot, _access),			\
 	.data_memslot_flags	= KVM_MEM_READONLY,				\
+	.pt_memslot_flags	= KVM_MEM_READONLY,				\
 	.guest_prepare		= { _PREPARE(_access) },			\
 	.guest_test		= _access,					\
 	.mmio_handler		= _mmio_handler,				\
@@ -841,6 +842,7 @@ static void help(char *name)
 {										\
 	.name			= SCAT2(ro_memslot_no_syndrome, _access),	\
 	.data_memslot_flags	= KVM_MEM_READONLY,				\
+	.pt_memslot_flags	= KVM_MEM_READONLY,				\
 	.guest_test		= _access,					\
 	.fail_vcpu_run_handler	= fail_vcpu_run_mmio_no_syndrome_handler,	\
 	.expected_events	= { .fail_vcpu_runs = 1 },			\
@@ -849,9 +851,9 @@ static void help(char *name)
 #define TEST_RO_MEMSLOT_AND_DIRTY_LOG(_access, _mmio_handler, _mmio_exits,	\
 				      _test_check)				\
 {										\
-	.name			= SCAT3(ro_memslot, _access, _with_af),		\
+	.name			= SCAT2(ro_memslot, _access),			\
 	.data_memslot_flags	= KVM_MEM_READONLY | KVM_MEM_LOG_DIRTY_PAGES,	\
-	.pt_memslot_flags	= KVM_MEM_LOG_DIRTY_PAGES,			\
+	.pt_memslot_flags	= KVM_MEM_READONLY | KVM_MEM_LOG_DIRTY_PAGES,	\
 	.guest_prepare		= { _PREPARE(_access) },			\
 	.guest_test		= _access,					\
 	.guest_test_check	= { _test_check },				\
@@ -863,7 +865,7 @@ static void help(char *name)
 {										\
 	.name			= SCAT2(ro_memslot_no_syn_and_dlog, _access),	\
 	.data_memslot_flags	= KVM_MEM_READONLY | KVM_MEM_LOG_DIRTY_PAGES,	\
-	.pt_memslot_flags	= KVM_MEM_LOG_DIRTY_PAGES,			\
+	.pt_memslot_flags	= KVM_MEM_READONLY | KVM_MEM_LOG_DIRTY_PAGES,	\
 	.guest_test		= _access,					\
 	.guest_test_check	= { _test_check },				\
 	.fail_vcpu_run_handler	= fail_vcpu_run_mmio_no_syndrome_handler,	\
@@ -875,6 +877,7 @@ static void help(char *name)
 {										\
 	.name			= SCAT2(ro_memslot_uffd, _access),		\
 	.data_memslot_flags	= KVM_MEM_READONLY,				\
+	.pt_memslot_flags	= KVM_MEM_READONLY,				\
 	.mem_mark_cmd		= CMD_HOLE_DATA | CMD_HOLE_PT,			\
 	.guest_prepare		= { _PREPARE(_access) },			\
 	.guest_test		= _access,					\
@@ -890,6 +893,7 @@ static void help(char *name)
 {										\
 	.name			= SCAT2(ro_memslot_no_syndrome, _access),	\
 	.data_memslot_flags	= KVM_MEM_READONLY,				\
+	.pt_memslot_flags	= KVM_MEM_READONLY,				\
 	.mem_mark_cmd		= CMD_HOLE_DATA | CMD_HOLE_PT,			\
 	.guest_test		= _access,					\
 	.uffd_data_handler	= _uffd_data_handler,				\
@@ -1024,7 +1028,7 @@ static struct test_desc tests[] = {
 				guest_check_write_in_dirty_log,
 				guest_check_s1ptw_wr_in_dirty_log),
 	/*
-	 * Try accesses when the data memory region is marked read-only
+	 * Access when both the PT and data regions are marked read-only
 	 * (with KVM_MEM_READONLY). Writes with a syndrome result in an
 	 * MMIO exit, writes with no syndrome (e.g., CAS) result in a
 	 * failed vcpu run, and reads/execs with and without syndroms do
@@ -1040,7 +1044,7 @@ static struct test_desc tests[] = {
 	TEST_RO_MEMSLOT_NO_SYNDROME(guest_st_preidx),
 
 	/*
-	 * Access when both the data region is both read-only and marked
+	 * The PT and data regions are both read-only and marked
 	 * for dirty logging at the same time. The expected result is that
 	 * for writes there should be no write in the dirty log. The
 	 * readonly handling is the same as if the memslot was not marked
@@ -1065,7 +1069,7 @@ static struct test_desc tests[] = {
 						  guest_check_no_write_in_dirty_log),
 
 	/*
-	 * Access when the data region is both read-only and punched with
+	 * The PT and data regions are both read-only and punched with
 	 * holes tracked with userfaultfd.  The expected result is the
 	 * union of both userfaultfd and read-only behaviors. For example,
 	 * write accesses result in a userfaultfd write fault and an MMIO
Signature · 1 line
2.39.1.456.gfc5497dd1b-goog
#5 lore raw

Re: [PATCH v2 4/4] KVM: selftests: aarch64: Test read-only PT memory regions

Hi Ricardo,

Quoted 5 lines · Ricardo Koller

Extend the read-only memslot tests in page_fault_test to test read-only PT (Page table) memslots. Note that this was not allowed before commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots") as all S1PTW faults were treated as writes which resulted in an (unrecoverable) exception inside the guest.

More of a style nit going forward (don't bother respinning):

Quoted 10 lines
---
 .../selftests/kvm/aarch64/page_fault_test.c    | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)
tools/testing/selftests/kvm/aarch64/page_fault_test.c +0 -0
index 2e2178a7d0d8..54680dc5887f 100644

tools/testing/selftests/kvm/aarch64/page_fault_test.c static void help(char *name) · line 832

Quoted 5 lines
tools/testing/selftests/kvm/aarch64/page_fault_test.c +1 -1
@@ -829,8 +829,9 @@ static void help(char *name)
 
 #define TEST_RO_MEMSLOT(_access, _mmio_handler, _mmio_exits)			\
 {										\
-	.name			= SCAT3(ro_memslot, _access, _with_af),		\
+	.name			= SCAT2(ro_memslot, _access),			\

You should explicitly call out these sort of drive-by/opportunistic changes in the commit message as being just that reviewers don't get lost figuring out how it relates to the functional change of this patch.

Signature · 2 lines
Thanks,
Oliver
#6 lore raw

Re: [PATCH v2 4/4] KVM: selftests: aarch64: Test read-only PT memory regions

Quoted 23 lines · Oliver Upton

Hi Ricardo,

Quoted 5 lines · Ricardo Koller

Extend the read-only memslot tests in page_fault_test to test read-only PT (Page table) memslots. Note that this was not allowed before commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots") as all S1PTW faults were treated as writes which resulted in an (unrecoverable) exception inside the guest.

More of a style nit going forward (don't bother respinning):

Quoted 11 lines
---
 .../selftests/kvm/aarch64/page_fault_test.c    | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)
tools/testing/selftests/kvm/aarch64/page_fault_test.c +0 -0
index 2e2178a7d0d8..54680dc5887f 100644
@@ -829,8 +829,9 @@ static void help(char *name)
 
Quoted 9 lines

tools/testing/selftests/kvm/aarch64/page_fault_test.c static void help(char *name) · line 832

Quoted 4 lines
tools/testing/selftests/kvm/aarch64/page_fault_test.c +1 -1
 #define TEST_RO_MEMSLOT(_access, _mmio_handler, _mmio_exits)                 \
 {                                                                            \
-     .name                   = SCAT3(ro_memslot, _access, _with_af),         \
+     .name                   = SCAT2(ro_memslot, _access),                   \

You should explicitly call out these sort of drive-by/opportunistic changes in the commit message as being just that reviewers don't get lost figuring out how it relates to the functional change of this patch.

Ah, sorry for that. Forgot about mentioning this in the commit message.

Ricardo

Quoted 4 lines

--
Thanks,
Oliver

Reviewed-by on the v2 series
#7 lore raw
In reply to Ricardo Koller

On Fri, Jan 27, 2023 at 09:43:49PM +0000, Ricardo Koller wrote:

[...]

Quoted 11 lines

Changes from v1:
- added sha1's for commit "KVM: arm64: Fix S1PTW handling on RO memslots" in all messages. (Oliver)
- removed _with_af from RO macro. (Oliver)

Ricardo Koller (4):
KVM: selftests: aarch64: Relax userfaultfd read vs. write checks
KVM: selftests: aarch64: Do not default to dirty PTE pages on all
S1PTWs
KVM: selftests: aarch64: Fix check of dirty log PT write
KVM: selftests: aarch64: Test read-only PT memory regions

These all LGTM, thanks for respinning. Tested these patches on an Altra machine.

Marc, I leave it up to you if you want to grab these patches for /fixes, otherwise I'll take them for 6.3. If you wind up applying these, for the series:

Signature · 2 lines
Thanks,
Oliver
maintainer Applied
#8 lore raw
In reply to Ricardo Koller
Quoted 9 lines · Ricardo Koller

Commit 406504c7b040 ("KVM: arm64: Fix S1PTW handling on RO memslots") changed the way S1PTW faults were handled by KVM. Before this fix, KVM treated any fault due to any S1PTW as a write, even S1PTW that didn't update the PTE. Some tests in page_fault_test mistakenly check for this KVM behavior and are currently failing. For example, there's a dirty log test that asserts that a S1PTW without AF or DA results in the PTE page getting dirty.

[...]

Applied to fixes, thanks!

[1/4] KVM: selftests: aarch64: Relax userfaultfd read vs. write checks

      commit: 0dd8d22a887a473f6f11abf556c4f7944ab5ef1d
[2/4] KVM: selftests: aarch64: Do not default to dirty PTE pages on all S1PTWs
      commit: 42561751ea918d8f3f54412622735e1f887cb360
[3/4] KVM: selftests: aarch64: Fix check of dirty log PT write
      commit: 8b03c97fa6fd442b949b71aeb7545b970b968fe3
[4/4] KVM: selftests: aarch64: Test read-only PT memory regions
      commit: 08ddbbdf0b55839ca93a12677a30a1ef24634969

Cheers,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.