Skip to contentSkip to conversation

selftests: powerpc: Fixes and execute-disable test for pkeys

PATCH Applied Sandipan Das 2 participants 5 messages Jun 4, 2020 – Jul 16, 2020

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

v3 cover selftests: powerpc: Fixes and execute-disable test for pkeys

This fixes the way the Authority Mask Register (AMR) is updated by the existing pkey tests and adds a new test to verify the functionality of execute-disabled pkeys.

Previous versions can be found at:
v2: https://lore.kernel.org/linuxppc-dev/[email protected]/
v1: https://lore.kernel.org/linuxppc-dev/[email protected]/

Changes in v3:
- Fixed AMR writes for existing pkey tests (new patch).
- Moved Hash MMU check under utilities (new patch) and removed duplicate code.
- Fixed comments on why the pkey permission bits were redefined.
- Switched to existing mfspr() macro for reading AMR.
- Switched to sig_atomic_t as data type for variables updated in the signal handlers.
- Switched to exit()-ing if the signal handlers come across an unexpected condition instead of trying to reset page and pkey permissions.
- Switched to write() from printf() for printing error messages from the signal handlers.
- Switched to getpagesize().
- Renamed fault counter to denote remaining faults.
- Dropped unnecessary randomization for choosing an address to fault at.
- Added additional information on change in permissions due to AMR and IAMR bits in comments.
- Switched the first instruction word of the executable region to a trap to test if it is actually overwritten by a no-op later.
- Added an new test scenario where the pkey imposes no restrictions and an attempt is made to jump to the executable region again.

Changes in v2:
- Added .gitignore entry for test binary.
- Fixed builds for older distros where siginfo_t might not have si_pkey as a formal member based on discussion with Michael.

Sandipan Das (3):
selftests: powerpc: Fix pkey access right updates
selftests: powerpc: Move Hash MMU check to utilities
selftests: powerpc: Add test for execute-disabled pkeys

 tools/testing/selftests/powerpc/include/reg.h |   6 +
 .../testing/selftests/powerpc/include/utils.h |   1 +
 tools/testing/selftests/powerpc/mm/.gitignore |   1 +
 tools/testing/selftests/powerpc/mm/Makefile   |   5 +-
 .../selftests/powerpc/mm/bad_accesses.c       |  28 --
 .../selftests/powerpc/mm/pkey_exec_prot.c     | 388 ++++++++++++++++++
 .../selftests/powerpc/ptrace/core-pkey.c      |   2 +-
 .../selftests/powerpc/ptrace/ptrace-pkey.c    |   2 +-
 tools/testing/selftests/powerpc/utils.c       |  28 ++
 9 files changed, 429 insertions(+), 32 deletions(-)
 create mode 100644 tools/testing/selftests/powerpc/mm/pkey_exec_prot.c
Signature · 1 line
2.25.1
#1 lore raw

PATCH v3 1/3 selftests: powerpc: Fix pkey access right updates

The Power ISA mandates that all writes to the Authority Mask Register (AMR) must always be preceded as well as succeeded by a context synchronizing instruction.

This makes sure that the tests follow this requirement when attempting to update a pkey's access rights.

---
 tools/testing/selftests/powerpc/include/reg.h        | 6 ++++++
 tools/testing/selftests/powerpc/ptrace/core-pkey.c   | 2 +-
 tools/testing/selftests/powerpc/ptrace/ptrace-pkey.c | 2 +-
 3 files changed, 8 insertions(+), 2 deletions(-)
tools/testing/selftests/powerpc/include/reg.h +6 -0
index 022c5076b2c5..c0f2742a3a59 100644
@@ -57,6 +57,12 @@
 #define SPRN_PPR       896	/* Program Priority Register */
 #define SPRN_AMR       13	/* Authority Mask Register - problem state */
 
+#define set_amr(v)	asm volatile("isync;" \
+				     "mtspr " __stringify(SPRN_AMR) ",%0;" \
+				     "isync" : \
+				    : "r" ((unsigned long)(v)) \
+				    : "memory")
+
 /* TEXASR register bits */
 #define TEXASR_FC	0xFE00000000000000
 #define TEXASR_FP	0x0100000000000000
tools/testing/selftests/powerpc/ptrace/core-pkey.c +1 -1
index d5c64fee032d..bbc05ffc5860 100644
@@ -150,7 +150,7 @@ static int child(struct shared_info *info)
 	printf("%-30s AMR: %016lx pkey1: %d pkey2: %d pkey3: %d\n",
 	       user_write, info->amr, pkey1, pkey2, pkey3);
 
-	mtspr(SPRN_AMR, info->amr);
+	set_amr(info->amr);
 
 	/*
 	 * We won't use pkey3. This tests whether the kernel restores the UAMOR
tools/testing/selftests/powerpc/ptrace/ptrace-pkey.c +1 -1
index bdbbbe8431e0..904c04f8c919 100644
@@ -126,7 +126,7 @@ static int child(struct shared_info *info)
 	printf("%-30s AMR: %016lx pkey1: %d pkey2: %d pkey3: %d\n",
 	       user_write, info->amr1, pkey1, pkey2, pkey3);
 
-	mtspr(SPRN_AMR, info->amr1);
+	set_amr(info->amr1);
 
 	/* Wait for parent to read our AMR value and write a new one. */
 	ret = prod_parent(&info->child_sync);
Signature · 1 line
2.25.1
#2 lore raw

PATCH v3 2/3 selftests: powerpc: Move Hash MMU check to utilities

In reply to Sandipan Das

This moves a function to test if the MMU is in Hash mode under the generic test utilities.

---
 .../testing/selftests/powerpc/include/utils.h |  1 +
 tools/testing/selftests/powerpc/mm/Makefile   |  2 +-
 .../selftests/powerpc/mm/bad_accesses.c       | 28 -------------------
 tools/testing/selftests/powerpc/utils.c       | 28 +++++++++++++++++++
 4 files changed, 30 insertions(+), 29 deletions(-)
tools/testing/selftests/powerpc/include/utils.h +1 -0
index e089a0c30d9a..ad2728736ae5 100644
@@ -60,6 +60,7 @@ static inline bool have_hwcap2(unsigned long ftr2)
 #endif
 
 bool is_ppc64le(void);
+int using_hash_mmu(bool *using_hash);
 
 /* Yes, this is evil */
 #define FAIL_IF(x)						\
tools/testing/selftests/powerpc/mm/Makefile +1 -1
index b9103c4bb414..2389bf791fd6 100644
@@ -10,7 +10,7 @@ TEST_GEN_FILES := tempfile
 top_srcdir = ../../../../..
 include ../../lib.mk
 
-$(TEST_GEN_PROGS): ../harness.c
+$(TEST_GEN_PROGS): ../harness.c ../utils.c
 
 $(OUTPUT)/prot_sao: ../utils.c
 
tools/testing/selftests/powerpc/mm/bad_accesses.c +0 -28
index adc465f499ef..a864ed7e2008 100644
@@ -64,34 +64,6 @@ int bad_access(char *p, bool write)
 	return 0;
 }
 
-static int using_hash_mmu(bool *using_hash)
-{
-	char line[128];
-	FILE *f;
-	int rc;
-
-	f = fopen("/proc/cpuinfo", "r");
-	FAIL_IF(!f);
-
-	rc = 0;
-	while (fgets(line, sizeof(line), f) != NULL) {
-		if (strcmp(line, "MMU		: Hash\n") == 0) {
-			*using_hash = true;
-			goto out;
-		}
-
-		if (strcmp(line, "MMU		: Radix\n") == 0) {
-			*using_hash = false;
-			goto out;
-		}
-	}
-
-	rc = -1;
-out:
-	fclose(f);
-	return rc;
-}
-
 static int test(void)
 {
 	unsigned long i, j, addr, region_shift, page_shift, page_size;
tools/testing/selftests/powerpc/utils.c +28 -0
index 5ee0e98c4896..933678f1ed0a 100644
@@ -293,3 +293,31 @@ void set_dscr(unsigned long val)
 
 	asm volatile("mtspr %1,%0" : : "r" (val), "i" (SPRN_DSCR));
 }
+
+int using_hash_mmu(bool *using_hash)
+{
+	char line[128];
+	FILE *f;
+	int rc;
+
+	f = fopen("/proc/cpuinfo", "r");
+	FAIL_IF(!f);
+
+	rc = 0;
+	while (fgets(line, sizeof(line), f) != NULL) {
+		if (strcmp(line, "MMU		: Hash\n") == 0) {
+			*using_hash = true;
+			goto out;
+		}
+
+		if (strcmp(line, "MMU		: Radix\n") == 0) {
+			*using_hash = false;
+			goto out;
+		}
+	}
+
+	rc = -1;
+out:
+	fclose(f);
+	return rc;
+}
Signature · 1 line
2.25.1
#3 lore raw

PATCH v3 3/3 selftests: powerpc: Add test for execute-disabled pkeys

In reply to Sandipan Das

Apart from read and write access, memory protection keys can also be used for restricting execute permission of pages on powerpc. This adds a test to verify if the feature works as expected.

---
 tools/testing/selftests/powerpc/mm/.gitignore |   1 +
 tools/testing/selftests/powerpc/mm/Makefile   |   3 +-
 .../selftests/powerpc/mm/pkey_exec_prot.c     | 388 ++++++++++++++++++
 3 files changed, 391 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/selftests/powerpc/mm/pkey_exec_prot.c
tools/testing/selftests/powerpc/mm/.gitignore +1 -0
index 2ca523255b1b..8f841f925baa 100644
@@ -8,3 +8,4 @@ wild_bctr
 large_vm_fork_separation
 bad_accesses
 tlbie_test
+pkey_exec_prot
tools/testing/selftests/powerpc/mm/Makefile +2 -1
index 2389bf791fd6..f9fa0ba7435c 100644
@@ -3,7 +3,7 @@ noarg:
 	$(MAKE) -C ../
 
 TEST_GEN_PROGS := hugetlb_vs_thp_test subpage_prot prot_sao segv_errors wild_bctr \
-		  large_vm_fork_separation bad_accesses
+		  large_vm_fork_separation bad_accesses pkey_exec_prot
 TEST_GEN_PROGS_EXTENDED := tlbie_test
 TEST_GEN_FILES := tempfile
 
@@ -17,6 +17,7 @@ $(OUTPUT)/prot_sao: ../utils.c
 $(OUTPUT)/wild_bctr: CFLAGS += -m64
 $(OUTPUT)/large_vm_fork_separation: CFLAGS += -m64
 $(OUTPUT)/bad_accesses: CFLAGS += -m64
+$(OUTPUT)/pkey_exec_prot: CFLAGS += -m64
 
 $(OUTPUT)/tempfile:
 	dd if=/dev/zero of=$@ bs=64k count=1
tools/testing/selftests/powerpc/mm/pkey_exec_prot.c new file +388 -0

388 lines not on this page · raw message

Signature · 1 line
2.25.1
#4 lore raw
In reply to Sandipan Das
Quoted 9 lines · Sandipan Das

This fixes the way the Authority Mask Register (AMR) is updated by the existing pkey tests and adds a new test to verify the functionality of execute-disabled pkeys.

Previous versions can be found at:
v2: https://lore.kernel.org/linuxppc-dev/[email protected]/
v1: https://lore.kernel.org/linuxppc-dev/[email protected]/

[...]

Applied to powerpc/next.

[1/3] selftests/powerpc: Fix pkey access right updates

      https://git.kernel.org/powerpc/c/828ca4320d130bbe1d12866152600c49ff6a9f79
[2/3] selftests/powerpc: Move Hash MMU check to utilities
      https://git.kernel.org/powerpc/c/c405b738daf9d8e8a5aedfeb6be851681e65e54b
[3/3] selftests/powerpc: Add test for execute-disabled pkeys
      https://git.kernel.org/powerpc/c/1addb6444791f9e87fce0eb9882ec96a4a76e615

cheers