summaryrefslogtreecommitdiffstats
path: root/tools/testing/selftests/powerpc/tm
diff options
context:
space:
mode:
Diffstat (limited to 'tools/testing/selftests/powerpc/tm')
-rw-r--r--tools/testing/selftests/powerpc/tm/.gitignore22
-rw-r--r--tools/testing/selftests/powerpc/tm/Makefile32
-rw-r--r--tools/testing/selftests/powerpc/tm/settings1
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-exec.c66
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-fork.c42
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-poison.c182
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-resched-dscr.c100
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-context-chk-fpu.c110
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-context-chk-gpr.c112
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vmx.c135
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vsx.c184
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-context-force-tm.c180
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-msr-resv.c74
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-pagefault.c284
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-sigreturn-nt.c50
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal-stack.c76
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-signal.S110
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-sigreturn.c93
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-syscall-asm.S28
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-syscall.c106
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-tar.c91
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-tmspr.c143
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-trap.c333
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-unavailable.c410
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-vmx-unavail.c118
-rw-r--r--tools/testing/selftests/powerpc/tm/tm-vmxcopy.c104
-rw-r--r--tools/testing/selftests/powerpc/tm/tm.h96
27 files changed, 3282 insertions, 0 deletions
diff --git a/tools/testing/selftests/powerpc/tm/.gitignore b/tools/testing/selftests/powerpc/tm/.gitignore
new file mode 100644
index 000000000..d8900a0c4
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/.gitignore
@@ -0,0 +1,22 @@
+# SPDX-License-Identifier: GPL-2.0-only
+tm-resched-dscr
+tm-syscall
+tm-signal-msr-resv
+tm-signal-stack
+tm-vmxcopy
+tm-fork
+tm-tar
+tm-tmspr
+tm-exec
+tm-signal-context-chk-fpu
+tm-signal-context-chk-gpr
+tm-signal-context-chk-vmx
+tm-signal-context-chk-vsx
+tm-signal-context-force-tm
+tm-signal-sigreturn-nt
+tm-signal-pagefault
+tm-vmx-unavail
+tm-unavailable
+tm-trap
+tm-sigreturn
+tm-poison
diff --git a/tools/testing/selftests/powerpc/tm/Makefile b/tools/testing/selftests/powerpc/tm/Makefile
new file mode 100644
index 000000000..5881e97c7
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/Makefile
@@ -0,0 +1,32 @@
+# SPDX-License-Identifier: GPL-2.0
+SIGNAL_CONTEXT_CHK_TESTS := tm-signal-context-chk-gpr tm-signal-context-chk-fpu \
+ tm-signal-context-chk-vmx tm-signal-context-chk-vsx
+
+TEST_GEN_PROGS := tm-resched-dscr tm-syscall tm-signal-msr-resv tm-signal-stack \
+ tm-vmxcopy tm-fork tm-tar tm-tmspr tm-vmx-unavail tm-unavailable tm-trap \
+ $(SIGNAL_CONTEXT_CHK_TESTS) tm-sigreturn tm-signal-sigreturn-nt \
+ tm-signal-context-force-tm tm-poison tm-signal-pagefault
+
+TEST_FILES := settings
+
+top_srcdir = ../../../../..
+include ../../lib.mk
+
+$(TEST_GEN_PROGS): ../harness.c ../utils.c
+
+CFLAGS += -mhtm
+
+$(OUTPUT)/tm-syscall: tm-syscall-asm.S
+$(OUTPUT)/tm-syscall: CFLAGS += -I../../../../../usr/include
+$(OUTPUT)/tm-tmspr: CFLAGS += -pthread
+$(OUTPUT)/tm-vmx-unavail: CFLAGS += -pthread -m64
+$(OUTPUT)/tm-resched-dscr: ../pmu/lib.c
+$(OUTPUT)/tm-unavailable: CFLAGS += -O0 -pthread -m64 -Wno-error=uninitialized -mvsx
+$(OUTPUT)/tm-trap: CFLAGS += -O0 -pthread -m64
+$(OUTPUT)/tm-signal-context-force-tm: CFLAGS += -pthread -m64
+$(OUTPUT)/tm-signal-pagefault: CFLAGS += -pthread -m64
+$(OUTPUT)/tm-poison: CFLAGS += -m64
+
+SIGNAL_CONTEXT_CHK_TESTS := $(patsubst %,$(OUTPUT)/%,$(SIGNAL_CONTEXT_CHK_TESTS))
+$(SIGNAL_CONTEXT_CHK_TESTS): tm-signal.S
+$(SIGNAL_CONTEXT_CHK_TESTS): CFLAGS += -mhtm -m64 -mvsx
diff --git a/tools/testing/selftests/powerpc/tm/settings b/tools/testing/selftests/powerpc/tm/settings
new file mode 100644
index 000000000..e7b941753
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/settings
@@ -0,0 +1 @@
+timeout=0
diff --git a/tools/testing/selftests/powerpc/tm/tm-exec.c b/tools/testing/selftests/powerpc/tm/tm-exec.c
new file mode 100644
index 000000000..260cfdb97
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-exec.c
@@ -0,0 +1,66 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016, Cyril Bur, IBM Corp.
+ *
+ * Syscalls can be performed provided the transactions are suspended.
+ * The exec() class of syscall is unique as a new process is loaded.
+ *
+ * It makes little sense for after an exec() call for the previously
+ * suspended transaction to still exist.
+ */
+
+#define _GNU_SOURCE
+#include <errno.h>
+#include <inttypes.h>
+#include <libgen.h>
+#include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#include "utils.h"
+#include "tm.h"
+
+static char *path;
+
+static int test_exec(void)
+{
+ SKIP_IF(!have_htm());
+
+ asm __volatile__(
+ "tbegin.;"
+ "blt 1f; "
+ "tsuspend.;"
+ "1: ;"
+ : : : "memory");
+
+ execl(path, "tm-exec", "--child", NULL);
+
+ /* Shouldn't get here */
+ perror("execl() failed");
+ return 1;
+}
+
+static int after_exec(void)
+{
+ asm __volatile__(
+ "tbegin.;"
+ "blt 1f;"
+ "tsuspend.;"
+ "1: ;"
+ : : : "memory");
+
+ FAIL_IF(failure_is_nesting());
+ return 0;
+}
+
+int main(int argc, char *argv[])
+{
+ path = argv[0];
+
+ if (argc > 1 && strcmp(argv[1], "--child") == 0)
+ return after_exec();
+
+ return test_harness(test_exec, "tm_exec");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-fork.c b/tools/testing/selftests/powerpc/tm/tm-fork.c
new file mode 100644
index 000000000..6efa5a685
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-fork.c
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2015, Michael Neuling, IBM Corp.
+ *
+ * Edited: Rashmica Gupta, Nov 2015
+ *
+ * This test does a fork syscall inside a transaction. Basic sniff test
+ * to see if we can enter the kernel during a transaction.
+ */
+
+#include <errno.h>
+#include <inttypes.h>
+#include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include "utils.h"
+#include "tm.h"
+
+int test_fork(void)
+{
+ SKIP_IF(!have_htm());
+
+ asm __volatile__(
+ "tbegin.;"
+ "blt 1f; "
+ "li 0, 2;" /* fork syscall */
+ "sc ;"
+ "tend.;"
+ "1: ;"
+ : : : "memory", "r0");
+ /* If we reach here, we've passed. Otherwise we've probably crashed
+ * the kernel */
+
+ return 0;
+}
+
+int main(int argc, char *argv[])
+{
+ return test_harness(test_fork, "tm_fork");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-poison.c b/tools/testing/selftests/powerpc/tm/tm-poison.c
new file mode 100644
index 000000000..29e5f26af
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-poison.c
@@ -0,0 +1,182 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2019, Gustavo Romero, Michael Neuling, IBM Corp.
+ *
+ * This test will spawn two processes. Both will be attached to the same
+ * CPU (CPU 0). The child will be in a loop writing to FP register f31 and
+ * VMX/VEC/Altivec register vr31 a known value, called poison, calling
+ * sched_yield syscall after to allow the parent to switch on the CPU.
+ * Parent will set f31 and vr31 to 1 and in a loop will check if f31 and
+ * vr31 remain 1 as expected until a given timeout (2m). If the issue is
+ * present child's poison will leak into parent's f31 or vr31 registers,
+ * otherwise, poison will never leak into parent's f31 and vr31 registers.
+ */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <inttypes.h>
+#include <sched.h>
+#include <sys/types.h>
+#include <signal.h>
+#include <inttypes.h>
+
+#include "tm.h"
+
+int tm_poison_test(void)
+{
+ int cpu, pid;
+ cpu_set_t cpuset;
+ uint64_t poison = 0xdeadbeefc0dec0fe;
+ uint64_t unknown = 0;
+ bool fail_fp = false;
+ bool fail_vr = false;
+
+ SKIP_IF(!have_htm());
+
+ cpu = pick_online_cpu();
+ FAIL_IF(cpu < 0);
+
+ // Attach both Child and Parent to the same CPU
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu, &cpuset);
+ FAIL_IF(sched_setaffinity(0, sizeof(cpuset), &cpuset) != 0);
+
+ pid = fork();
+ if (!pid) {
+ /**
+ * child
+ */
+ while (1) {
+ sched_yield();
+ asm (
+ "mtvsrd 31, %[poison];" // f31 = poison
+ "mtvsrd 63, %[poison];" // vr31 = poison
+
+ : : [poison] "r" (poison) : );
+ }
+ }
+
+ /**
+ * parent
+ */
+ asm (
+ /*
+ * Set r3, r4, and f31 to known value 1 before entering
+ * in transaction. They won't be written after that.
+ */
+ " li 3, 0x1 ;"
+ " li 4, 0x1 ;"
+ " mtvsrd 31, 4 ;"
+
+ /*
+ * The Time Base (TB) is a 64-bit counter register that is
+ * independent of the CPU clock and which is incremented
+ * at a frequency of 512000000 Hz, so every 1.953125ns.
+ * So it's necessary 120s/0.000000001953125s = 61440000000
+ * increments to get a 2 minutes timeout. Below we set that
+ * value in r5 and then use r6 to track initial TB value,
+ * updating TB values in r7 at every iteration and comparing it
+ * to r6. When r7 (current) - r6 (initial) > 61440000000 we bail
+ * out since for sure we spent already 2 minutes in the loop.
+ * SPR 268 is the TB register.
+ */
+ " lis 5, 14 ;"
+ " ori 5, 5, 19996 ;"
+ " sldi 5, 5, 16 ;" // r5 = 61440000000
+
+ " mfspr 6, 268 ;" // r6 (TB initial)
+ "1: mfspr 7, 268 ;" // r7 (TB current)
+ " subf 7, 6, 7 ;" // r7 - r6 > 61440000000 ?
+ " cmpd 7, 5 ;"
+ " bgt 3f ;" // yes, exit
+
+ /*
+ * Main loop to check f31
+ */
+ " tbegin. ;" // no, try again
+ " beq 1b ;" // restart if no timeout
+ " mfvsrd 3, 31 ;" // read f31
+ " cmpd 3, 4 ;" // f31 == 1 ?
+ " bne 2f ;" // broken :-(
+ " tabort. 3 ;" // try another transaction
+ "2: tend. ;" // commit transaction
+ "3: mr %[unknown], 3 ;" // record r3
+
+ : [unknown] "=r" (unknown)
+ :
+ : "cr0", "r3", "r4", "r5", "r6", "r7", "vs31"
+
+ );
+
+ /*
+ * On leak 'unknown' will contain 'poison' value from child,
+ * otherwise (no leak) 'unknown' will contain the same value
+ * as r3 before entering in transactional mode, i.e. 0x1.
+ */
+ fail_fp = unknown != 0x1;
+ if (fail_fp)
+ printf("Unknown value %#"PRIx64" leaked into f31!\n", unknown);
+ else
+ printf("Good, no poison or leaked value into FP registers\n");
+
+ asm (
+ /*
+ * Set r3, r4, and vr31 to known value 1 before entering
+ * in transaction. They won't be written after that.
+ */
+ " li 3, 0x1 ;"
+ " li 4, 0x1 ;"
+ " mtvsrd 63, 4 ;"
+
+ " lis 5, 14 ;"
+ " ori 5, 5, 19996 ;"
+ " sldi 5, 5, 16 ;" // r5 = 61440000000
+
+ " mfspr 6, 268 ;" // r6 (TB initial)
+ "1: mfspr 7, 268 ;" // r7 (TB current)
+ " subf 7, 6, 7 ;" // r7 - r6 > 61440000000 ?
+ " cmpd 7, 5 ;"
+ " bgt 3f ;" // yes, exit
+
+ /*
+ * Main loop to check vr31
+ */
+ " tbegin. ;" // no, try again
+ " beq 1b ;" // restart if no timeout
+ " mfvsrd 3, 63 ;" // read vr31
+ " cmpd 3, 4 ;" // vr31 == 1 ?
+ " bne 2f ;" // broken :-(
+ " tabort. 3 ;" // try another transaction
+ "2: tend. ;" // commit transaction
+ "3: mr %[unknown], 3 ;" // record r3
+
+ : [unknown] "=r" (unknown)
+ :
+ : "cr0", "r3", "r4", "r5", "r6", "r7", "vs63"
+
+ );
+
+ /*
+ * On leak 'unknown' will contain 'poison' value from child,
+ * otherwise (no leak) 'unknown' will contain the same value
+ * as r3 before entering in transactional mode, i.e. 0x1.
+ */
+ fail_vr = unknown != 0x1;
+ if (fail_vr)
+ printf("Unknown value %#"PRIx64" leaked into vr31!\n", unknown);
+ else
+ printf("Good, no poison or leaked value into VEC registers\n");
+
+ kill(pid, SIGKILL);
+
+ return (fail_fp | fail_vr);
+}
+
+int main(int argc, char *argv[])
+{
+ /* Test completes in about 4m */
+ test_harness_set_timeout(250);
+ return test_harness(tm_poison_test, "tm_poison_test");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-resched-dscr.c b/tools/testing/selftests/powerpc/tm/tm-resched-dscr.c
new file mode 100644
index 000000000..4cdb83964
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-resched-dscr.c
@@ -0,0 +1,100 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Test context switching to see if the DSCR SPR is correctly preserved
+ * when within a transaction.
+ *
+ * Note: We assume that the DSCR has been left at the default value (0)
+ * for all CPUs.
+ *
+ * Method:
+ *
+ * Set a value into the DSCR.
+ *
+ * Start a transaction, and suspend it (*).
+ *
+ * Hard loop checking to see if the transaction has become doomed.
+ *
+ * Now that we *may* have been preempted, record the DSCR and TEXASR SPRS.
+ *
+ * If the abort was because of a context switch, check the DSCR value.
+ * Otherwise, try again.
+ *
+ * (*) If the transaction is not suspended we can't see the problem because
+ * the transaction abort handler will restore the DSCR to it's checkpointed
+ * value before we regain control.
+ */
+
+#include <inttypes.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <assert.h>
+#include <asm/tm.h>
+
+#include "utils.h"
+#include "tm.h"
+#include "../pmu/lib.h"
+
+#define SPRN_DSCR 0x03
+
+int test_body(void)
+{
+ uint64_t rv, dscr1 = 1, dscr2, texasr;
+
+ SKIP_IF(!have_htm());
+
+ printf("Check DSCR TM context switch: ");
+ fflush(stdout);
+ for (;;) {
+ asm __volatile__ (
+ /* set a known value into the DSCR */
+ "ld 3, %[dscr1];"
+ "mtspr %[sprn_dscr], 3;"
+
+ "li %[rv], 1;"
+ /* start and suspend a transaction */
+ "tbegin.;"
+ "beq 1f;"
+ "tsuspend.;"
+
+ /* hard loop until the transaction becomes doomed */
+ "2: ;"
+ "tcheck 0;"
+ "bc 4, 0, 2b;"
+
+ /* record DSCR and TEXASR */
+ "mfspr 3, %[sprn_dscr];"
+ "std 3, %[dscr2];"
+ "mfspr 3, %[sprn_texasr];"
+ "std 3, %[texasr];"
+
+ "tresume.;"
+ "tend.;"
+ "li %[rv], 0;"
+ "1: ;"
+ : [rv]"=r"(rv), [dscr2]"=m"(dscr2), [texasr]"=m"(texasr)
+ : [dscr1]"m"(dscr1)
+ , [sprn_dscr]"i"(SPRN_DSCR), [sprn_texasr]"i"(SPRN_TEXASR)
+ : "memory", "r3"
+ );
+ assert(rv); /* make sure the transaction aborted */
+ if ((texasr >> 56) != TM_CAUSE_RESCHED) {
+ continue;
+ }
+ if (dscr2 != dscr1) {
+ printf(" FAIL\n");
+ return 1;
+ } else {
+ printf(" OK\n");
+ return 0;
+ }
+ }
+}
+
+static int tm_resched_dscr(void)
+{
+ return eat_cpu(test_body);
+}
+
+int main(int argc, const char *argv[])
+{
+ return test_harness(tm_resched_dscr, "tm_resched_dscr");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-fpu.c b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-fpu.c
new file mode 100644
index 000000000..254f912ad
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-fpu.c
@@ -0,0 +1,110 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016, Cyril Bur, IBM Corp.
+ *
+ * Test the kernel's signal frame code.
+ *
+ * The kernel sets up two sets of ucontexts if the signal was to be
+ * delivered while the thread was in a transaction (referred too as
+ * first and second contexts).
+ * Expected behaviour is that the checkpointed state is in the user
+ * context passed to the signal handler (first context). The speculated
+ * state can be accessed with the uc_link pointer (second context).
+ *
+ * The rationale for this is that if TM unaware code (which linked
+ * against TM libs) installs a signal handler it will not know of the
+ * speculative nature of the 'live' registers and may infer the wrong
+ * thing.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <signal.h>
+#include <unistd.h>
+
+#include <altivec.h>
+
+#include "utils.h"
+#include "tm.h"
+
+#define MAX_ATTEMPT 500000
+
+#define NV_FPU_REGS 18 /* Number of non-volatile FP registers */
+#define FPR14 14 /* First non-volatile FP register to check in f14-31 subset */
+
+long tm_signal_self_context_load(pid_t pid, long *gprs, double *fps, vector int *vms, vector int *vss);
+
+/* Test only non-volatile registers, i.e. 18 fpr registers from f14 to f31 */
+static double fps[] = {
+ /* First context will be set with these values, i.e. non-speculative */
+ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,
+ /* Second context will be set with these values, i.e. speculative */
+ -1,-2,-3,-4,-5,-6,-7,-8,-9,-10,-11,-12,-13,-14,-15,-16,-17,-18
+};
+
+static sig_atomic_t fail, broken;
+
+static void signal_usr1(int signum, siginfo_t *info, void *uc)
+{
+ int i;
+ ucontext_t *ucp = uc;
+ ucontext_t *tm_ucp = ucp->uc_link;
+
+ for (i = 0; i < NV_FPU_REGS; i++) {
+ /* Check first context. Print all mismatches. */
+ fail = (ucp->uc_mcontext.fp_regs[FPR14 + i] != fps[i]);
+ if (fail) {
+ broken = 1;
+ printf("FPR%d (1st context) == %g instead of %g (expected)\n",
+ FPR14 + i, ucp->uc_mcontext.fp_regs[FPR14 + i], fps[i]);
+ }
+ }
+
+ for (i = 0; i < NV_FPU_REGS; i++) {
+ /* Check second context. Print all mismatches. */
+ fail = (tm_ucp->uc_mcontext.fp_regs[FPR14 + i] != fps[NV_FPU_REGS + i]);
+ if (fail) {
+ broken = 1;
+ printf("FPR%d (2nd context) == %g instead of %g (expected)\n",
+ FPR14 + i, tm_ucp->uc_mcontext.fp_regs[FPR14 + i], fps[NV_FPU_REGS + i]);
+ }
+ }
+}
+
+static int tm_signal_context_chk_fpu()
+{
+ struct sigaction act;
+ int i;
+ long rc;
+ pid_t pid = getpid();
+
+ SKIP_IF(!have_htm());
+
+ act.sa_sigaction = signal_usr1;
+ sigemptyset(&act.sa_mask);
+ act.sa_flags = SA_SIGINFO;
+ if (sigaction(SIGUSR1, &act, NULL) < 0) {
+ perror("sigaction sigusr1");
+ exit(1);
+ }
+
+ i = 0;
+ while (i < MAX_ATTEMPT && !broken) {
+ /*
+ * tm_signal_self_context_load will set both first and second
+ * contexts accordingly to the values passed through non-NULL
+ * array pointers to it, in that case 'fps', and invoke the
+ * signal handler installed for SIGUSR1.
+ */
+ rc = tm_signal_self_context_load(pid, NULL, fps, NULL, NULL);
+ FAIL_IF(rc != pid);
+ i++;
+ }
+
+ return (broken);
+}
+
+int main(void)
+{
+ return test_harness(tm_signal_context_chk_fpu, "tm_signal_context_chk_fpu");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-gpr.c b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-gpr.c
new file mode 100644
index 000000000..0cc680f61
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-gpr.c
@@ -0,0 +1,112 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016, Cyril Bur, IBM Corp.
+ *
+ * Test the kernel's signal frame code.
+ *
+ * The kernel sets up two sets of ucontexts if the signal was to be
+ * delivered while the thread was in a transaction (referred too as
+ * first and second contexts).
+ * Expected behaviour is that the checkpointed state is in the user
+ * context passed to the signal handler (first context). The speculated
+ * state can be accessed with the uc_link pointer (second context).
+ *
+ * The rationale for this is that if TM unaware code (which linked
+ * against TM libs) installs a signal handler it will not know of the
+ * speculative nature of the 'live' registers and may infer the wrong
+ * thing.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <signal.h>
+#include <unistd.h>
+
+#include <altivec.h>
+
+#include "utils.h"
+#include "tm.h"
+
+#define MAX_ATTEMPT 500000
+
+#define NV_GPR_REGS 18 /* Number of non-volatile GPR registers */
+#define R14 14 /* First non-volatile register to check in r14-r31 subset */
+
+long tm_signal_self_context_load(pid_t pid, long *gprs, double *fps, vector int *vms, vector int *vss);
+
+static sig_atomic_t fail, broken;
+
+/* Test only non-volatile general purpose registers, i.e. r14-r31 */
+static long gprs[] = {
+ /* First context will be set with these values, i.e. non-speculative */
+ /* R14, R15, ... */
+ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,
+ /* Second context will be set with these values, i.e. speculative */
+ /* R14, R15, ... */
+ -1,-2,-3,-4,-5,-6,-7,-8,-9,-10,-11,-12,-13,-14,-15,-16,-17,-18
+};
+
+static void signal_usr1(int signum, siginfo_t *info, void *uc)
+{
+ int i;
+ ucontext_t *ucp = uc;
+ ucontext_t *tm_ucp = ucp->uc_link;
+
+ /* Check first context. Print all mismatches. */
+ for (i = 0; i < NV_GPR_REGS; i++) {
+ fail = (ucp->uc_mcontext.gp_regs[R14 + i] != gprs[i]);
+ if (fail) {
+ broken = 1;
+ printf("GPR%d (1st context) == %lu instead of %lu (expected)\n",
+ R14 + i, ucp->uc_mcontext.gp_regs[R14 + i], gprs[i]);
+ }
+ }
+
+ /* Check second context. Print all mismatches. */
+ for (i = 0; i < NV_GPR_REGS; i++) {
+ fail = (tm_ucp->uc_mcontext.gp_regs[R14 + i] != gprs[NV_GPR_REGS + i]);
+ if (fail) {
+ broken = 1;
+ printf("GPR%d (2nd context) == %lu instead of %lu (expected)\n",
+ R14 + i, tm_ucp->uc_mcontext.gp_regs[R14 + i], gprs[NV_GPR_REGS + i]);
+ }
+ }
+}
+
+static int tm_signal_context_chk_gpr()
+{
+ struct sigaction act;
+ int i;
+ long rc;
+ pid_t pid = getpid();
+
+ SKIP_IF(!have_htm());
+
+ act.sa_sigaction = signal_usr1;
+ sigemptyset(&act.sa_mask);
+ act.sa_flags = SA_SIGINFO;
+ if (sigaction(SIGUSR1, &act, NULL) < 0) {
+ perror("sigaction sigusr1");
+ exit(1);
+ }
+
+ i = 0;
+ while (i < MAX_ATTEMPT && !broken) {
+ /*
+ * tm_signal_self_context_load will set both first and second
+ * contexts accordingly to the values passed through non-NULL
+ * array pointers to it, in that case 'gprs', and invoke the
+ * signal handler installed for SIGUSR1.
+ */
+ rc = tm_signal_self_context_load(pid, gprs, NULL, NULL, NULL);
+ FAIL_IF(rc != pid);
+ i++;
+ }
+
+ return broken;
+}
+
+int main(void)
+{
+ return test_harness(tm_signal_context_chk_gpr, "tm_signal_context_chk_gpr");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vmx.c b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vmx.c
new file mode 100644
index 000000000..b6d52730a
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vmx.c
@@ -0,0 +1,135 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016, Cyril Bur, IBM Corp.
+ *
+ * Test the kernel's signal frame code.
+ *
+ * The kernel sets up two sets of ucontexts if the signal was to be
+ * delivered while the thread was in a transaction (referred too as
+ * first and second contexts).
+ * Expected behaviour is that the checkpointed state is in the user
+ * context passed to the signal handler (first context). The speculated
+ * state can be accessed with the uc_link pointer (second context).
+ *
+ * The rationale for this is that if TM unaware code (which linked
+ * against TM libs) installs a signal handler it will not know of the
+ * speculative nature of the 'live' registers and may infer the wrong
+ * thing.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <signal.h>
+#include <unistd.h>
+
+#include <altivec.h>
+
+#include "utils.h"
+#include "tm.h"
+
+#define MAX_ATTEMPT 500000
+
+#define NV_VMX_REGS 12 /* Number of non-volatile VMX registers */
+#define VMX20 20 /* First non-volatile register to check in vr20-31 subset */
+
+long tm_signal_self_context_load(pid_t pid, long *gprs, double *fps, vector int *vms, vector int *vss);
+
+static sig_atomic_t fail, broken;
+
+/* Test only non-volatile registers, i.e. 12 vmx registers from vr20 to vr31 */
+vector int vms[] = {
+ /* First context will be set with these values, i.e. non-speculative */
+ /* VMX20 , VMX21 , ... */
+ { 1, 2, 3, 4},{ 5, 6, 7, 8},{ 9,10,11,12},
+ {13,14,15,16},{17,18,19,20},{21,22,23,24},
+ {25,26,27,28},{29,30,31,32},{33,34,35,36},
+ {37,38,39,40},{41,42,43,44},{45,46,47,48},
+ /* Second context will be set with these values, i.e. speculative */
+ /* VMX20 , VMX21 , ... */
+ { -1, -2, -3, -4},{ -5, -6, -7, -8},{ -9,-10,-11,-12},
+ {-13,-14,-15,-16},{-17,-18,-19,-20},{-21,-22,-23,-24},
+ {-25,-26,-27,-28},{-29,-30,-31,-32},{-33,-34,-35,-36},
+ {-37,-38,-39,-40},{-41,-42,-43,-44},{-45,-46,-47,-48}
+};
+
+static void signal_usr1(int signum, siginfo_t *info, void *uc)
+{
+ int i, j;
+ ucontext_t *ucp = uc;
+ ucontext_t *tm_ucp = ucp->uc_link;
+
+ for (i = 0; i < NV_VMX_REGS; i++) {
+ /* Check first context. Print all mismatches. */
+ fail = memcmp(ucp->uc_mcontext.v_regs->vrregs[VMX20 + i],
+ &vms[i], sizeof(vector int));
+ if (fail) {
+ broken = 1;
+ printf("VMX%d (1st context) == 0x", VMX20 + i);
+ /* Print actual value in first context. */
+ for (j = 0; j < 4; j++)
+ printf("%08x", ucp->uc_mcontext.v_regs->vrregs[VMX20 + i][j]);
+ printf(" instead of 0x");
+ /* Print expected value. */
+ for (j = 0; j < 4; j++)
+ printf("%08x", vms[i][j]);
+ printf(" (expected)\n");
+ }
+ }
+
+ for (i = 0; i < NV_VMX_REGS; i++) {
+ /* Check second context. Print all mismatches. */
+ fail = memcmp(tm_ucp->uc_mcontext.v_regs->vrregs[VMX20 + i],
+ &vms[NV_VMX_REGS + i], sizeof (vector int));
+ if (fail) {
+ broken = 1;
+ printf("VMX%d (2nd context) == 0x", NV_VMX_REGS + i);
+ /* Print actual value in second context. */
+ for (j = 0; j < 4; j++)
+ printf("%08x", tm_ucp->uc_mcontext.v_regs->vrregs[VMX20 + i][j]);
+ printf(" instead of 0x");
+ /* Print expected value. */
+ for (j = 0; j < 4; j++)
+ printf("%08x", vms[NV_VMX_REGS + i][j]);
+ printf(" (expected)\n");
+ }
+ }
+}
+
+static int tm_signal_context_chk()
+{
+ struct sigaction act;
+ int i;
+ long rc;
+ pid_t pid = getpid();
+
+ SKIP_IF(!have_htm());
+
+ act.sa_sigaction = signal_usr1;
+ sigemptyset(&act.sa_mask);
+ act.sa_flags = SA_SIGINFO;
+ if (sigaction(SIGUSR1, &act, NULL) < 0) {
+ perror("sigaction sigusr1");
+ exit(1);
+ }
+
+ i = 0;
+ while (i < MAX_ATTEMPT && !broken) {
+ /*
+ * tm_signal_self_context_load will set both first and second
+ * contexts accordingly to the values passed through non-NULL
+ * array pointers to it, in that case 'vms', and invoke the
+ * signal handler installed for SIGUSR1.
+ */
+ rc = tm_signal_self_context_load(pid, NULL, NULL, vms, NULL);
+ FAIL_IF(rc != pid);
+ i++;
+ }
+
+ return (broken);
+}
+
+int main(void)
+{
+ return test_harness(tm_signal_context_chk, "tm_signal_context_chk_vmx");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vsx.c b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vsx.c
new file mode 100644
index 000000000..8e25e2072
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-context-chk-vsx.c
@@ -0,0 +1,184 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016, Cyril Bur, IBM Corp.
+ *
+ * Test the kernel's signal frame code.
+ *
+ * The kernel sets up two sets of ucontexts if the signal was to be
+ * delivered while the thread was in a transaction (referred too as
+ * first and second contexts).
+ * Expected behaviour is that the checkpointed state is in the user
+ * context passed to the signal handler (first context). The speculated
+ * state can be accessed with the uc_link pointer (second context).
+ *
+ * The rationale for this is that if TM unaware code (which linked
+ * against TM libs) installs a signal handler it will not know of the
+ * speculative nature of the 'live' registers and may infer the wrong
+ * thing.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <signal.h>
+#include <unistd.h>
+
+#include <altivec.h>
+
+#include "utils.h"
+#include "tm.h"
+
+#define MAX_ATTEMPT 500000
+
+#define NV_VSX_REGS 12 /* Number of VSX registers to check. */
+#define VSX20 20 /* First VSX register to check in vsr20-vsr31 subset */
+#define FPR20 20 /* FPR20 overlaps VSX20 most significant doubleword */
+
+long tm_signal_self_context_load(pid_t pid, long *gprs, double *fps, vector int *vms, vector int *vss);
+
+static sig_atomic_t fail, broken;
+
+/* Test only 12 vsx registers from vsr20 to vsr31 */
+vector int vsxs[] = {
+ /* First context will be set with these values, i.e. non-speculative */
+ /* VSX20 , VSX21 , ... */
+ { 1, 2, 3, 4},{ 5, 6, 7, 8},{ 9,10,11,12},
+ {13,14,15,16},{17,18,19,20},{21,22,23,24},
+ {25,26,27,28},{29,30,31,32},{33,34,35,36},
+ {37,38,39,40},{41,42,43,44},{45,46,47,48},
+ /* Second context will be set with these values, i.e. speculative */
+ /* VSX20 , VSX21 , ... */
+ {-1, -2, -3, -4 },{-5, -6, -7, -8 },{-9, -10,-11,-12},
+ {-13,-14,-15,-16},{-17,-18,-19,-20},{-21,-22,-23,-24},
+ {-25,-26,-27,-28},{-29,-30,-31,-32},{-33,-34,-35,-36},
+ {-37,-38,-39,-40},{-41,-42,-43,-44},{-45,-46,-47,-48}
+};
+
+static void signal_usr1(int signum, siginfo_t *info, void *uc)
+{
+ int i, j;
+ uint8_t vsx[sizeof(vector int)];
+ uint8_t vsx_tm[sizeof(vector int)];
+ ucontext_t *ucp = uc;
+ ucontext_t *tm_ucp = ucp->uc_link;
+
+ /*
+ * FP registers and VMX registers overlap the VSX registers.
+ *
+ * FP registers (f0-31) overlap the most significant 64 bits of VSX
+ * registers vsr0-31, whilst VMX registers vr0-31, being 128-bit like
+ * the VSX registers, overlap fully the other half of VSX registers,
+ * i.e. vr0-31 overlaps fully vsr32-63.
+ *
+ * Due to compatibility and historical reasons (VMX/Altivec support
+ * appeared first on the architecture), VMX registers vr0-31 (so VSX
+ * half vsr32-63 too) are stored right after the v_regs pointer, in an
+ * area allocated for 'vmx_reverse' array (please see
+ * arch/powerpc/include/uapi/asm/sigcontext.h for details about the
+ * mcontext_t structure on Power).
+ *
+ * The other VSX half (vsr0-31) is hence stored below vr0-31/vsr32-63
+ * registers, but only the least significant 64 bits of vsr0-31. The
+ * most significant 64 bits of vsr0-31 (f0-31), as it overlaps the FP
+ * registers, is kept in fp_regs.
+ *
+ * v_regs is a 16 byte aligned pointer at the start of vmx_reserve
+ * (vmx_reserve may or may not be 16 aligned) where the v_regs structure
+ * exists, so v_regs points to where vr0-31 / vsr32-63 registers are
+ * fully stored. Since v_regs type is elf_vrregset_t, v_regs + 1
+ * skips all the slots used to store vr0-31 / vsr32-64 and points to
+ * part of one VSX half, i.e. v_regs + 1 points to the least significant
+ * 64 bits of vsr0-31. The other part of this half (the most significant
+ * part of vsr0-31) is stored in fp_regs.
+ *
+ */
+ /* Get pointer to least significant doubleword of vsr0-31 */
+ long *vsx_ptr = (long *)(ucp->uc_mcontext.v_regs + 1);
+ long *tm_vsx_ptr = (long *)(tm_ucp->uc_mcontext.v_regs + 1);
+
+ /* Check first context. Print all mismatches. */
+ for (i = 0; i < NV_VSX_REGS; i++) {
+ /*
+ * Copy VSX most significant doubleword from fp_regs and
+ * copy VSX least significant one from 64-bit slots below
+ * saved VMX registers.
+ */
+ memcpy(vsx, &ucp->uc_mcontext.fp_regs[FPR20 + i], 8);
+ memcpy(vsx + 8, &vsx_ptr[VSX20 + i], 8);
+
+ fail = memcmp(vsx, &vsxs[i], sizeof(vector int));
+
+ if (fail) {
+ broken = 1;
+ printf("VSX%d (1st context) == 0x", VSX20 + i);
+ for (j = 0; j < 16; j++)
+ printf("%02x", vsx[j]);
+ printf(" instead of 0x");
+ for (j = 0; j < 4; j++)
+ printf("%08x", vsxs[i][j]);
+ printf(" (expected)\n");
+ }
+ }
+
+ /* Check second context. Print all mismatches. */
+ for (i = 0; i < NV_VSX_REGS; i++) {
+ /*
+ * Copy VSX most significant doubleword from fp_regs and
+ * copy VSX least significant one from 64-bit slots below
+ * saved VMX registers.
+ */
+ memcpy(vsx_tm, &tm_ucp->uc_mcontext.fp_regs[FPR20 + i], 8);
+ memcpy(vsx_tm + 8, &tm_vsx_ptr[VSX20 + i], 8);
+
+ fail = memcmp(vsx_tm, &vsxs[NV_VSX_REGS + i], sizeof(vector int));
+
+ if (fail) {
+ broken = 1;
+ printf("VSX%d (2nd context) == 0x", VSX20 + i);
+ for (j = 0; j < 16; j++)
+ printf("%02x", vsx_tm[j]);
+ printf(" instead of 0x");
+ for (j = 0; j < 4; j++)
+ printf("%08x", vsxs[NV_VSX_REGS + i][j]);
+ printf("(expected)\n");
+ }
+ }
+}
+
+static int tm_signal_context_chk()
+{
+ struct sigaction act;
+ int i;
+ long rc;
+ pid_t pid = getpid();
+
+ SKIP_IF(!have_htm());
+
+ act.sa_sigaction = signal_usr1;
+ sigemptyset(&act.sa_mask);
+ act.sa_flags = SA_SIGINFO;
+ if (sigaction(SIGUSR1, &act, NULL) < 0) {
+ perror("sigaction sigusr1");
+ exit(1);
+ }
+
+ i = 0;
+ while (i < MAX_ATTEMPT && !broken) {
+ /*
+ * tm_signal_self_context_load will set both first and second
+ * contexts accordingly to the values passed through non-NULL
+ * array pointers to it, in that case 'vsxs', and invoke the
+ * signal handler installed for SIGUSR1.
+ */
+ rc = tm_signal_self_context_load(pid, NULL, NULL, NULL, vsxs);
+ FAIL_IF(rc != pid);
+ i++;
+ }
+
+ return (broken);
+}
+
+int main(void)
+{
+ return test_harness(tm_signal_context_chk, "tm_signal_context_chk_vsx");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-context-force-tm.c b/tools/testing/selftests/powerpc/tm/tm-signal-context-force-tm.c
new file mode 100644
index 000000000..421cb082f
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-context-force-tm.c
@@ -0,0 +1,180 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright 2018, Breno Leitao, Gustavo Romero, IBM Corp.
+ *
+ * This test raises a SIGUSR1 signal, and toggle the MSR[TS]
+ * fields at the signal handler. With MSR[TS] being set, the kernel will
+ * force a recheckpoint, which may cause a segfault when returning to
+ * user space. Since the test needs to re-run, the segfault needs to be
+ * caught and handled.
+ *
+ * In order to continue the test even after a segfault, the context is
+ * saved prior to the signal being raised, and it is restored when there is
+ * a segmentation fault. This happens for COUNT_MAX times.
+ *
+ * This test never fails (as returning EXIT_FAILURE). It either succeeds,
+ * or crash the kernel (on a buggy kernel).
+ */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <signal.h>
+#include <string.h>
+#include <ucontext.h>
+#include <unistd.h>
+#include <sys/mman.h>
+
+#include "tm.h"
+#include "utils.h"
+#include "reg.h"
+
+#define COUNT_MAX 5000 /* Number of interactions */
+
+/*
+ * This test only runs on 64 bits system. Unsetting MSR_TS_S to avoid
+ * compilation issue on 32 bits system. There is no side effect, since the
+ * whole test will be skipped if it is not running on 64 bits system.
+ */
+#ifndef __powerpc64__
+#undef MSR_TS_S
+#define MSR_TS_S 0
+#endif
+
+/* Setting contexts because the test will crash and we want to recover */
+ucontext_t init_context;
+
+/* count is changed in the signal handler, so it must be volatile */
+static volatile int count;
+
+void usr_signal_handler(int signo, siginfo_t *si, void *uc)
+{
+ ucontext_t *ucp = uc;
+ int ret;
+
+ /*
+ * Allocating memory in a signal handler, and never freeing it on
+ * purpose, forcing the heap increase, so, the memory leak is what
+ * we want here.
+ */
+ ucp->uc_link = mmap(NULL, sizeof(ucontext_t),
+ PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
+ if (ucp->uc_link == (void *)-1) {
+ perror("Mmap failed");
+ exit(-1);
+ }
+
+ /* Forcing the page to be allocated in a page fault */
+ ret = madvise(ucp->uc_link, sizeof(ucontext_t), MADV_DONTNEED);
+ if (ret) {
+ perror("madvise failed");
+ exit(-1);
+ }
+
+ memcpy(&ucp->uc_link->uc_mcontext, &ucp->uc_mcontext,
+ sizeof(ucp->uc_mcontext));
+
+ /* Forcing to enable MSR[TM] */
+ UCONTEXT_MSR(ucp) |= MSR_TS_S;
+
+ /*
+ * A fork inside a signal handler seems to be more efficient than a
+ * fork() prior to the signal being raised.
+ */
+ if (fork() == 0) {
+ /*
+ * Both child and parent will return, but, child returns
+ * with count set so it will exit in the next segfault.
+ * Parent will continue to loop.
+ */
+ count = COUNT_MAX;
+ }
+
+ /*
+ * If the change above does not hit the bug, it will cause a
+ * segmentation fault, since the ck structures are NULL.
+ */
+}
+
+void seg_signal_handler(int signo, siginfo_t *si, void *uc)
+{
+ count++;
+
+ /* Reexecute the test */
+ setcontext(&init_context);
+}
+
+void tm_trap_test(void)
+{
+ struct sigaction usr_sa, seg_sa;
+ stack_t ss;
+
+ usr_sa.sa_flags = SA_SIGINFO | SA_ONSTACK;
+ usr_sa.sa_sigaction = usr_signal_handler;
+
+ seg_sa.sa_flags = SA_SIGINFO;
+ seg_sa.sa_sigaction = seg_signal_handler;
+
+ /*
+ * Set initial context. Will get back here from
+ * seg_signal_handler()
+ */
+ getcontext(&init_context);
+
+ while (count < COUNT_MAX) {
+ /* Allocated an alternative signal stack area */
+ ss.ss_sp = mmap(NULL, SIGSTKSZ, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
+ ss.ss_size = SIGSTKSZ;
+ ss.ss_flags = 0;
+
+ if (ss.ss_sp == (void *)-1) {
+ perror("mmap error\n");
+ exit(-1);
+ }
+
+ /* Force the allocation through a page fault */
+ if (madvise(ss.ss_sp, SIGSTKSZ, MADV_DONTNEED)) {
+ perror("madvise\n");
+ exit(-1);
+ }
+
+ /*
+ * Setting an alternative stack to generate a page fault when
+ * the signal is raised.
+ */
+ if (sigaltstack(&ss, NULL)) {
+ perror("sigaltstack\n");
+ exit(-1);
+ }
+
+ /* The signal handler will enable MSR_TS */
+ sigaction(SIGUSR1, &usr_sa, NULL);
+ /* If it does not crash, it might segfault, avoid it to retest */
+ sigaction(SIGSEGV, &seg_sa, NULL);
+
+ raise(SIGUSR1);
+ count++;
+ }
+}
+
+int tm_signal_context_force_tm(void)
+{
+ SKIP_IF(!have_htm());
+ /*
+ * Skipping if not running on 64 bits system, since I think it is
+ * not possible to set mcontext's [MSR] with TS, due to it being 32
+ * bits.
+ */
+ SKIP_IF(!is_ppc64le());
+
+ tm_trap_test();
+
+ return EXIT_SUCCESS;
+}
+
+int main(int argc, char **argv)
+{
+ test_harness(tm_signal_context_force_tm, "tm_signal_context_force_tm");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-msr-resv.c b/tools/testing/selftests/powerpc/tm/tm-signal-msr-resv.c
new file mode 100644
index 000000000..4a61e9bd1
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-msr-resv.c
@@ -0,0 +1,74 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2015, Michael Neuling, IBM Corp.
+ *
+ * Test the kernel's signal return code to ensure that it doesn't
+ * crash when both the transactional and suspend MSR bits are set in
+ * the signal context.
+ *
+ * For this test, we send ourselves a SIGUSR1. In the SIGUSR1 handler
+ * we modify the signal context to set both MSR TM S and T bits (which
+ * is "reserved" by the PowerISA). When we return from the signal
+ * handler (implicit sigreturn), the kernel should detect reserved MSR
+ * value and send us with a SIGSEGV.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <signal.h>
+#include <unistd.h>
+
+#include "utils.h"
+#include "tm.h"
+
+int segv_expected = 0;
+
+void signal_segv(int signum)
+{
+ if (segv_expected && (signum == SIGSEGV))
+ _exit(0);
+ _exit(1);
+}
+
+void signal_usr1(int signum, siginfo_t *info, void *uc)
+{
+ ucontext_t *ucp = uc;
+
+ /* Link tm checkpointed context to normal context */
+ ucp->uc_link = ucp;
+ /* Set all TM bits so that the context is now invalid */
+#ifdef __powerpc64__
+ ucp->uc_mcontext.gp_regs[PT_MSR] |= (7ULL << 32);
+#else
+ ucp->uc_mcontext.uc_regs->gregs[PT_MSR] |= (7ULL);
+#endif
+ /* Should segv on return becuase of invalid context */
+ segv_expected = 1;
+}
+
+int tm_signal_msr_resv()
+{
+ struct sigaction act;
+
+ SKIP_IF(!have_htm());
+
+ act.sa_sigaction = signal_usr1;
+ sigemptyset(&act.sa_mask);
+ act.sa_flags = SA_SIGINFO;
+ if (sigaction(SIGUSR1, &act, NULL) < 0) {
+ perror("sigaction sigusr1");
+ exit(1);
+ }
+ if (signal(SIGSEGV, signal_segv) == SIG_ERR)
+ exit(1);
+
+ raise(SIGUSR1);
+
+ /* We shouldn't get here as we exit in the segv handler */
+ return 1;
+}
+
+int main(void)
+{
+ return test_harness(tm_signal_msr_resv, "tm_signal_msr_resv");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-pagefault.c b/tools/testing/selftests/powerpc/tm/tm-signal-pagefault.c
new file mode 100644
index 000000000..5908bc6ab
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-pagefault.c
@@ -0,0 +1,284 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright 2020, Gustavo Luiz Duarte, IBM Corp.
+ *
+ * This test starts a transaction and triggers a signal, forcing a pagefault to
+ * happen when the kernel signal handling code touches the user signal stack.
+ *
+ * In order to avoid pre-faulting the signal stack memory and to force the
+ * pagefault to happen precisely in the kernel signal handling code, the
+ * pagefault handling is done in userspace using the userfaultfd facility.
+ *
+ * Further pagefaults are triggered by crafting the signal handler's ucontext
+ * to point to additional memory regions managed by the userfaultfd, so using
+ * the same mechanism used to avoid pre-faulting the signal stack memory.
+ *
+ * On failure (bug is present) kernel crashes or never returns control back to
+ * userspace. If bug is not present, tests completes almost immediately.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <linux/userfaultfd.h>
+#include <poll.h>
+#include <unistd.h>
+#include <sys/ioctl.h>
+#include <sys/syscall.h>
+#include <fcntl.h>
+#include <sys/mman.h>
+#include <pthread.h>
+#include <signal.h>
+#include <errno.h>
+
+#include "tm.h"
+
+
+#define UF_MEM_SIZE 655360 /* 10 x 64k pages */
+
+/* Memory handled by userfaultfd */
+static char *uf_mem;
+static size_t uf_mem_offset = 0;
+
+/*
+ * Data that will be copied into the faulting pages (instead of zero-filled
+ * pages). This is used to make the test more reliable and avoid segfaulting
+ * when we return from the signal handler. Since we are making the signal
+ * handler's ucontext point to newly allocated memory, when that memory is
+ * paged-in it will contain the expected content.
+ */
+static char backing_mem[UF_MEM_SIZE];
+
+static size_t pagesize;
+
+/*
+ * Return a chunk of at least 'size' bytes of memory that will be handled by
+ * userfaultfd. If 'backing_data' is not NULL, its content will be save to
+ * 'backing_mem' and then copied into the faulting pages when the page fault
+ * is handled.
+ */
+void *get_uf_mem(size_t size, void *backing_data)
+{
+ void *ret;
+
+ if (uf_mem_offset + size > UF_MEM_SIZE) {
+ fprintf(stderr, "Requesting more uf_mem than expected!\n");
+ exit(EXIT_FAILURE);
+ }
+
+ ret = &uf_mem[uf_mem_offset];
+
+ /* Save the data that will be copied into the faulting page */
+ if (backing_data != NULL)
+ memcpy(&backing_mem[uf_mem_offset], backing_data, size);
+
+ /* Reserve the requested amount of uf_mem */
+ uf_mem_offset += size;
+ /* Keep uf_mem_offset aligned to the page size (round up) */
+ uf_mem_offset = (uf_mem_offset + pagesize - 1) & ~(pagesize - 1);
+
+ return ret;
+}
+
+void *fault_handler_thread(void *arg)
+{
+ struct uffd_msg msg; /* Data read from userfaultfd */
+ long uffd; /* userfaultfd file descriptor */
+ struct uffdio_copy uffdio_copy;
+ struct pollfd pollfd;
+ ssize_t nread, offset;
+
+ uffd = (long) arg;
+
+ for (;;) {
+ pollfd.fd = uffd;
+ pollfd.events = POLLIN;
+ if (poll(&pollfd, 1, -1) == -1) {
+ perror("poll() failed");
+ exit(EXIT_FAILURE);
+ }
+
+ nread = read(uffd, &msg, sizeof(msg));
+ if (nread == 0) {
+ fprintf(stderr, "read(): EOF on userfaultfd\n");
+ exit(EXIT_FAILURE);
+ }
+
+ if (nread == -1) {
+ perror("read() failed");
+ exit(EXIT_FAILURE);
+ }
+
+ /* We expect only one kind of event */
+ if (msg.event != UFFD_EVENT_PAGEFAULT) {
+ fprintf(stderr, "Unexpected event on userfaultfd\n");
+ exit(EXIT_FAILURE);
+ }
+
+ /*
+ * We need to handle page faults in units of pages(!).
+ * So, round faulting address down to page boundary.
+ */
+ uffdio_copy.dst = msg.arg.pagefault.address & ~(pagesize-1);
+
+ offset = (char *) uffdio_copy.dst - uf_mem;
+ uffdio_copy.src = (unsigned long) &backing_mem[offset];
+
+ uffdio_copy.len = pagesize;
+ uffdio_copy.mode = 0;
+ uffdio_copy.copy = 0;
+ if (ioctl(uffd, UFFDIO_COPY, &uffdio_copy) == -1) {
+ perror("ioctl-UFFDIO_COPY failed");
+ exit(EXIT_FAILURE);
+ }
+ }
+}
+
+void setup_uf_mem(void)
+{
+ long uffd; /* userfaultfd file descriptor */
+ pthread_t thr;
+ struct uffdio_api uffdio_api;
+ struct uffdio_register uffdio_register;
+ int ret;
+
+ pagesize = sysconf(_SC_PAGE_SIZE);
+
+ /* Create and enable userfaultfd object */
+ uffd = syscall(__NR_userfaultfd, O_CLOEXEC | O_NONBLOCK);
+ if (uffd == -1) {
+ perror("userfaultfd() failed");
+ exit(EXIT_FAILURE);
+ }
+ uffdio_api.api = UFFD_API;
+ uffdio_api.features = 0;
+ if (ioctl(uffd, UFFDIO_API, &uffdio_api) == -1) {
+ perror("ioctl-UFFDIO_API failed");
+ exit(EXIT_FAILURE);
+ }
+
+ /*
+ * Create a private anonymous mapping. The memory will be demand-zero
+ * paged, that is, not yet allocated. When we actually touch the memory
+ * the related page will be allocated via the userfaultfd mechanism.
+ */
+ uf_mem = mmap(NULL, UF_MEM_SIZE, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ if (uf_mem == MAP_FAILED) {
+ perror("mmap() failed");
+ exit(EXIT_FAILURE);
+ }
+
+ /*
+ * Register the memory range of the mapping we've just mapped to be
+ * handled by the userfaultfd object. In 'mode' we request to track
+ * missing pages (i.e. pages that have not yet been faulted-in).
+ */
+ uffdio_register.range.start = (unsigned long) uf_mem;
+ uffdio_register.range.len = UF_MEM_SIZE;
+ uffdio_register.mode = UFFDIO_REGISTER_MODE_MISSING;
+ if (ioctl(uffd, UFFDIO_REGISTER, &uffdio_register) == -1) {
+ perror("ioctl-UFFDIO_REGISTER");
+ exit(EXIT_FAILURE);
+ }
+
+ /* Create a thread that will process the userfaultfd events */
+ ret = pthread_create(&thr, NULL, fault_handler_thread, (void *) uffd);
+ if (ret != 0) {
+ fprintf(stderr, "pthread_create(): Error. Returned %d\n", ret);
+ exit(EXIT_FAILURE);
+ }
+}
+
+/*
+ * Assumption: the signal was delivered while userspace was in transactional or
+ * suspended state, i.e. uc->uc_link != NULL.
+ */
+void signal_handler(int signo, siginfo_t *si, void *uc)
+{
+ ucontext_t *ucp = uc;
+
+ /* Skip 'trap' after returning, otherwise we get a SIGTRAP again */
+ ucp->uc_link->uc_mcontext.regs->nip += 4;
+
+ ucp->uc_mcontext.v_regs =
+ get_uf_mem(sizeof(elf_vrreg_t), ucp->uc_mcontext.v_regs);
+
+ ucp->uc_link->uc_mcontext.v_regs =
+ get_uf_mem(sizeof(elf_vrreg_t), ucp->uc_link->uc_mcontext.v_regs);
+
+ ucp->uc_link = get_uf_mem(sizeof(ucontext_t), ucp->uc_link);
+}
+
+bool have_userfaultfd(void)
+{
+ long rc;
+
+ errno = 0;
+ rc = syscall(__NR_userfaultfd, -1);
+
+ return rc == 0 || errno != ENOSYS;
+}
+
+int tm_signal_pagefault(void)
+{
+ struct sigaction sa;
+ stack_t ss;
+
+ SKIP_IF(!have_htm());
+ SKIP_IF(!have_userfaultfd());
+
+ setup_uf_mem();
+
+ /*
+ * Set an alternative stack that will generate a page fault when the
+ * signal is raised. The page fault will be treated via userfaultfd,
+ * i.e. via fault_handler_thread.
+ */
+ ss.ss_sp = get_uf_mem(SIGSTKSZ, NULL);
+ ss.ss_size = SIGSTKSZ;
+ ss.ss_flags = 0;
+ if (sigaltstack(&ss, NULL) == -1) {
+ perror("sigaltstack() failed");
+ exit(EXIT_FAILURE);
+ }
+
+ sa.sa_flags = SA_SIGINFO | SA_ONSTACK;
+ sa.sa_sigaction = signal_handler;
+ if (sigaction(SIGTRAP, &sa, NULL) == -1) {
+ perror("sigaction() failed");
+ exit(EXIT_FAILURE);
+ }
+
+ /* Trigger a SIGTRAP in transactional state */
+ asm __volatile__(
+ "tbegin.;"
+ "beq 1f;"
+ "trap;"
+ "1: ;"
+ : : : "memory");
+
+ /* Trigger a SIGTRAP in suspended state */
+ asm __volatile__(
+ "tbegin.;"
+ "beq 1f;"
+ "tsuspend.;"
+ "trap;"
+ "tresume.;"
+ "1: ;"
+ : : : "memory");
+
+ return EXIT_SUCCESS;
+}
+
+int main(int argc, char **argv)
+{
+ /*
+ * Depending on kernel config, the TM Bad Thing might not result in a
+ * crash, instead the kernel never returns control back to userspace, so
+ * set a tight timeout. If the test passes it completes almost
+ * immediately.
+ */
+ test_harness_set_timeout(2);
+ return test_harness(tm_signal_pagefault, "tm_signal_pagefault");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-sigreturn-nt.c b/tools/testing/selftests/powerpc/tm/tm-signal-sigreturn-nt.c
new file mode 100644
index 000000000..07c388147
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-sigreturn-nt.c
@@ -0,0 +1,50 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright 2018, Breno Leitao, Gustavo Romero, IBM Corp.
+ *
+ * A test case that creates a signal and starts a suspended transaction
+ * inside the signal handler.
+ *
+ * It returns from the signal handler with the CPU at suspended state, but
+ * without setting usercontext MSR Transaction State (TS) fields.
+ */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <signal.h>
+
+#include "utils.h"
+#include "tm.h"
+
+void trap_signal_handler(int signo, siginfo_t *si, void *uc)
+{
+ ucontext_t *ucp = (ucontext_t *) uc;
+
+ asm("tbegin.; tsuspend.;");
+
+ /* Skip 'trap' instruction if it succeed */
+ ucp->uc_mcontext.regs->nip += 4;
+}
+
+int tm_signal_sigreturn_nt(void)
+{
+ struct sigaction trap_sa;
+
+ SKIP_IF(!have_htm());
+
+ trap_sa.sa_flags = SA_SIGINFO;
+ trap_sa.sa_sigaction = trap_signal_handler;
+
+ sigaction(SIGTRAP, &trap_sa, NULL);
+
+ raise(SIGTRAP);
+
+ return EXIT_SUCCESS;
+}
+
+int main(int argc, char **argv)
+{
+ test_harness(tm_signal_sigreturn_nt, "tm_signal_sigreturn_nt");
+}
+
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal-stack.c b/tools/testing/selftests/powerpc/tm/tm-signal-stack.c
new file mode 100644
index 000000000..cdcf8c5bb
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal-stack.c
@@ -0,0 +1,76 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2015, Michael Neuling, IBM Corp.
+ *
+ * Test the kernel's signal delievery code to ensure that we don't
+ * trelaim twice in the kernel signal delivery code. This can happen
+ * if we trigger a signal when in a transaction and the stack pointer
+ * is bogus.
+ *
+ * This test case registers a SEGV handler, sets the stack pointer
+ * (r1) to NULL, starts a transaction and then generates a SEGV. The
+ * SEGV should be handled but we exit here as the stack pointer is
+ * invalid and hance we can't sigreturn. We only need to check that
+ * this flow doesn't crash the kernel.
+ */
+
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <signal.h>
+
+#include "utils.h"
+#include "tm.h"
+
+void signal_segv(int signum)
+{
+ /* This should never actually run since stack is foobar */
+ exit(1);
+}
+
+int tm_signal_stack()
+{
+ int pid;
+
+ SKIP_IF(!have_htm());
+
+ pid = fork();
+ if (pid < 0)
+ exit(1);
+
+ if (pid) { /* Parent */
+ /*
+ * It's likely the whole machine will crash here so if
+ * the child ever exits, we are good.
+ */
+ wait(NULL);
+ return 0;
+ }
+
+ /*
+ * The flow here is:
+ * 1) register a signal handler (so signal delievery occurs)
+ * 2) make stack pointer (r1) = NULL
+ * 3) start transaction
+ * 4) cause segv
+ */
+ if (signal(SIGSEGV, signal_segv) == SIG_ERR)
+ exit(1);
+ asm volatile("li 1, 0 ;" /* stack ptr == NULL */
+ "1:"
+ "tbegin.;"
+ "beq 1b ;" /* retry forever */
+ "tsuspend.;"
+ "ld 2, 0(1) ;" /* trigger segv" */
+ : : : "memory");
+
+ /* This should never get here due to above segv */
+ return 1;
+}
+
+int main(void)
+{
+ return test_harness(tm_signal_stack, "tm_signal_stack");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-signal.S b/tools/testing/selftests/powerpc/tm/tm-signal.S
new file mode 100644
index 000000000..c80c91366
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-signal.S
@@ -0,0 +1,110 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright 2015, Cyril Bur, IBM Corp.
+ */
+
+#include "basic_asm.h"
+#include "gpr_asm.h"
+#include "fpu_asm.h"
+#include "vmx_asm.h"
+#include "vsx_asm.h"
+
+/*
+ * Large caveat here being that the caller cannot expect the
+ * signal to always be sent! The hardware can (AND WILL!) abort
+ * the transaction between the tbegin and the tsuspend (however
+ * unlikely it seems or infrequently it actually happens).
+ * You have been warned.
+ */
+/* long tm_signal_self(pid_t pid, long *gprs, double *fps, vector *vms, vector *vss); */
+FUNC_START(tm_signal_self_context_load)
+ PUSH_BASIC_STACK(512)
+ /*
+ * Don't strictly need to save and restore as it depends on if
+ * we're going to use them, however this reduces messy logic
+ */
+ PUSH_VMX(STACK_FRAME_LOCAL(5,0),r8)
+ PUSH_FPU(512)
+ PUSH_NVREGS_BELOW_FPU(512)
+ std r3, STACK_FRAME_PARAM(0)(sp) /* pid */
+ std r4, STACK_FRAME_PARAM(1)(sp) /* gps */
+ std r5, STACK_FRAME_PARAM(2)(sp) /* fps */
+ std r6, STACK_FRAME_PARAM(3)(sp) /* vms */
+ std r7, STACK_FRAME_PARAM(4)(sp) /* vss */
+
+ ld r3, STACK_FRAME_PARAM(1)(sp)
+ cmpdi r3, 0
+ beq skip_gpr_lc
+ bl load_gpr
+skip_gpr_lc:
+ ld r3, STACK_FRAME_PARAM(2)(sp)
+ cmpdi r3, 0
+ beq skip_fpu_lc
+ bl load_fpu
+skip_fpu_lc:
+ ld r3, STACK_FRAME_PARAM(3)(sp)
+ cmpdi r3, 0
+ beq skip_vmx_lc
+ bl load_vmx
+skip_vmx_lc:
+ ld r3, STACK_FRAME_PARAM(4)(sp)
+ cmpdi r3, 0
+ beq skip_vsx_lc
+ bl load_vsx
+skip_vsx_lc:
+ /*
+ * Set r3 (return value) before tbegin. Use the pid as a known
+ * 'all good' return value, zero is used to indicate a non-doomed
+ * transaction.
+ */
+ ld r3, STACK_FRAME_PARAM(0)(sp)
+ tbegin.
+ beq 1f
+ tsuspend. /* Can't enter a syscall transactionally */
+ ld r3, STACK_FRAME_PARAM(1)(sp)
+ cmpdi r3, 0
+ beq skip_gpr_lt
+ /* Get the second half of the array */
+ addi r3, r3, 8 * 18
+ bl load_gpr
+skip_gpr_lt:
+ ld r3, STACK_FRAME_PARAM(2)(sp)
+ cmpdi r3, 0
+ beq skip_fpu_lt
+ /* Get the second half of the array */
+ addi r3, r3, 8 * 18
+ bl load_fpu
+skip_fpu_lt:
+ ld r3, STACK_FRAME_PARAM(3)(sp)
+ cmpdi r3, 0
+ beq skip_vmx_lt
+ /* Get the second half of the array */
+ addi r3, r3, 16 * 12
+ bl load_vmx
+skip_vmx_lt:
+ ld r3, STACK_FRAME_PARAM(4)(sp)
+ cmpdi r3, 0
+ beq skip_vsx_lt
+ /* Get the second half of the array */
+ addi r3, r3, 16 * 12
+ bl load_vsx
+skip_vsx_lt:
+ li r0, 37 /* sys_kill */
+ ld r3, STACK_FRAME_PARAM(0)(sp) /* pid */
+ li r4, 10 /* SIGUSR1 */
+ sc /* Taking the signal will doom the transaction */
+ tabort. 0
+ tresume. /* Be super sure we abort */
+ /*
+ * This will cause us to resume doomed transaction and cause
+ * hardware to cleanup, we'll end up at 1: anything between
+ * tresume. and 1: shouldn't ever run.
+ */
+ li r3, 0
+ 1:
+ POP_VMX(STACK_FRAME_LOCAL(5,0),r4)
+ POP_FPU(512)
+ POP_NVREGS_BELOW_FPU(512)
+ POP_BASIC_STACK(512)
+ blr
+FUNC_END(tm_signal_self_context_load)
diff --git a/tools/testing/selftests/powerpc/tm/tm-sigreturn.c b/tools/testing/selftests/powerpc/tm/tm-sigreturn.c
new file mode 100644
index 000000000..9a6017a1d
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-sigreturn.c
@@ -0,0 +1,93 @@
+// SPDX-License-Identifier: GPL-2.0
+
+/*
+ * Copyright 2015, Laurent Dufour, IBM Corp.
+ *
+ * Test the kernel's signal returning code to check reclaim is done if the
+ * sigreturn() is called while in a transaction (suspended since active is
+ * already dropped trough the system call path).
+ *
+ * The kernel must discard the transaction when entering sigreturn, since
+ * restoring the potential TM SPRS from the signal frame is requiring to not be
+ * in a transaction.
+ */
+
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+
+#include "tm.h"
+#include "utils.h"
+
+
+void handler(int sig)
+{
+ uint64_t ret;
+
+ asm __volatile__(
+ "li 3,1 ;"
+ "tbegin. ;"
+ "beq 1f ;"
+ "li 3,0 ;"
+ "tsuspend. ;"
+ "1: ;"
+ "std%X[ret] 3, %[ret] ;"
+ : [ret] "=m"(ret)
+ :
+ : "memory", "3", "cr0");
+
+ if (ret)
+ exit(1);
+
+ /*
+ * We return from the signal handle while in a suspended transaction
+ */
+}
+
+
+int tm_sigreturn(void)
+{
+ struct sigaction sa;
+ uint64_t ret = 0;
+
+ SKIP_IF(!have_htm());
+ SKIP_IF(!is_ppc64le());
+
+ memset(&sa, 0, sizeof(sa));
+ sa.sa_handler = handler;
+ sigemptyset(&sa.sa_mask);
+
+ if (sigaction(SIGSEGV, &sa, NULL))
+ exit(1);
+
+ asm __volatile__(
+ "tbegin. ;"
+ "beq 1f ;"
+ "li 3,0 ;"
+ "std 3,0(3) ;" /* trigger SEGV */
+ "li 3,1 ;"
+ "std%X[ret] 3,%[ret] ;"
+ "tend. ;"
+ "b 2f ;"
+ "1: ;"
+ "li 3,2 ;"
+ "std%X[ret] 3,%[ret] ;"
+ "2: ;"
+ : [ret] "=m"(ret)
+ :
+ : "memory", "3", "cr0");
+
+ if (ret != 2)
+ exit(1);
+
+ exit(0);
+}
+
+int main(void)
+{
+ return test_harness(tm_sigreturn, "tm_sigreturn");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-syscall-asm.S b/tools/testing/selftests/powerpc/tm/tm-syscall-asm.S
new file mode 100644
index 000000000..bd1ca25fe
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-syscall-asm.S
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#include <ppc-asm.h>
+#include <asm/unistd.h>
+
+ .text
+FUNC_START(getppid_tm_active)
+ tbegin.
+ beq 1f
+ li r0, __NR_getppid
+ sc
+ tend.
+ blr
+1:
+ li r3, -1
+ blr
+
+FUNC_START(getppid_tm_suspended)
+ tbegin.
+ beq 1f
+ li r0, __NR_getppid
+ tsuspend.
+ sc
+ tresume.
+ tend.
+ blr
+1:
+ li r3, -1
+ blr
diff --git a/tools/testing/selftests/powerpc/tm/tm-syscall.c b/tools/testing/selftests/powerpc/tm/tm-syscall.c
new file mode 100644
index 000000000..becb8207b
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-syscall.c
@@ -0,0 +1,106 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2015, Sam Bobroff, IBM Corp.
+ *
+ * Test the kernel's system call code to ensure that a system call
+ * made from within an active HTM transaction is aborted with the
+ * correct failure code.
+ * Conversely, ensure that a system call made from within a
+ * suspended transaction can succeed.
+ */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <asm/tm.h>
+#include <sys/time.h>
+#include <stdlib.h>
+
+#include "utils.h"
+#include "tm.h"
+
+extern int getppid_tm_active(void);
+extern int getppid_tm_suspended(void);
+
+unsigned retries = 0;
+
+#define TEST_DURATION 10 /* seconds */
+#define TM_RETRIES 100
+
+pid_t getppid_tm(bool suspend)
+{
+ int i;
+ pid_t pid;
+
+ for (i = 0; i < TM_RETRIES; i++) {
+ if (suspend)
+ pid = getppid_tm_suspended();
+ else
+ pid = getppid_tm_active();
+
+ if (pid >= 0)
+ return pid;
+
+ if (failure_is_persistent()) {
+ if (failure_is_syscall())
+ return -1;
+
+ printf("Unexpected persistent transaction failure.\n");
+ printf("TEXASR 0x%016lx, TFIAR 0x%016lx.\n",
+ __builtin_get_texasr(), __builtin_get_tfiar());
+ exit(-1);
+ }
+
+ retries++;
+ }
+
+ printf("Exceeded limit of %d temporary transaction failures.\n", TM_RETRIES);
+ printf("TEXASR 0x%016lx, TFIAR 0x%016lx.\n",
+ __builtin_get_texasr(), __builtin_get_tfiar());
+
+ exit(-1);
+}
+
+int tm_syscall(void)
+{
+ unsigned count = 0;
+ struct timeval end, now;
+
+ SKIP_IF(!have_htm_nosc());
+
+ setbuf(stdout, NULL);
+
+ printf("Testing transactional syscalls for %d seconds...\n", TEST_DURATION);
+
+ gettimeofday(&end, NULL);
+ now.tv_sec = TEST_DURATION;
+ now.tv_usec = 0;
+ timeradd(&end, &now, &end);
+
+ for (count = 0; timercmp(&now, &end, <); count++) {
+ /*
+ * Test a syscall within a suspended transaction and verify
+ * that it succeeds.
+ */
+ FAIL_IF(getppid_tm(true) == -1); /* Should succeed. */
+
+ /*
+ * Test a syscall within an active transaction and verify that
+ * it fails with the correct failure code.
+ */
+ FAIL_IF(getppid_tm(false) != -1); /* Should fail... */
+ FAIL_IF(!failure_is_persistent()); /* ...persistently... */
+ FAIL_IF(!failure_is_syscall()); /* ...with code syscall. */
+ gettimeofday(&now, 0);
+ }
+
+ printf("%d active and suspended transactions behaved correctly.\n", count);
+ printf("(There were %d transaction retries.)\n", retries);
+
+ return 0;
+}
+
+int main(void)
+{
+ return test_harness(tm_syscall, "tm_syscall");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-tar.c b/tools/testing/selftests/powerpc/tm/tm-tar.c
new file mode 100644
index 000000000..03be8c472
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-tar.c
@@ -0,0 +1,91 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2015, Michael Neuling, IBM Corp.
+ * Original: Michael Neuling 19/7/2013
+ * Edited: Rashmica Gupta 01/12/2015
+ *
+ * Do some transactions, see if the tar is corrupted.
+ * If the transaction is aborted, the TAR should be rolled back to the
+ * checkpointed value before the transaction began. The value written to
+ * TAR in suspended mode should only remain in TAR if the transaction
+ * completes.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+
+#include "tm.h"
+#include "utils.h"
+
+int num_loops = 10000;
+
+int test_tar(void)
+{
+ int i;
+
+ SKIP_IF(!have_htm());
+ SKIP_IF(!is_ppc64le());
+
+ for (i = 0; i < num_loops; i++)
+ {
+ uint64_t result = 0;
+ asm __volatile__(
+ "li 7, 1;"
+ "mtspr %[tar], 7;" /* tar = 1 */
+ "tbegin.;"
+ "beq 3f;"
+ "li 4, 0x7000;" /* Loop lots, to use time */
+ "2:;" /* Start loop */
+ "li 7, 2;"
+ "mtspr %[tar], 7;" /* tar = 2 */
+ "tsuspend.;"
+ "li 7, 3;"
+ "mtspr %[tar], 7;" /* tar = 3 */
+ "tresume.;"
+ "subi 4, 4, 1;"
+ "cmpdi 4, 0;"
+ "bne 2b;"
+ "tend.;"
+
+ /* Transaction sucess! TAR should be 3 */
+ "mfspr 7, %[tar];"
+ "ori %[res], 7, 4;" // res = 3|4 = 7
+ "b 4f;"
+
+ /* Abort handler. TAR should be rolled back to 1 */
+ "3:;"
+ "mfspr 7, %[tar];"
+ "ori %[res], 7, 8;" // res = 1|8 = 9
+ "4:;"
+
+ : [res]"=r"(result)
+ : [tar]"i"(SPRN_TAR)
+ : "memory", "r0", "r4", "r7");
+
+ /* If result is anything else other than 7 or 9, the tar
+ * value must have been corrupted. */
+ if ((result != 7) && (result != 9))
+ return 1;
+ }
+ return 0;
+}
+
+int main(int argc, char *argv[])
+{
+ /* A low number of iterations (eg 100) can cause a false pass */
+ if (argc > 1) {
+ if (strcmp(argv[1], "-h") == 0) {
+ printf("Syntax:\n\t%s [<num loops>]\n",
+ argv[0]);
+ return 1;
+ } else {
+ num_loops = atoi(argv[1]);
+ }
+ }
+
+ printf("Starting, %d loops\n", num_loops);
+
+ return test_harness(test_tar, "tm_tar");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-tmspr.c b/tools/testing/selftests/powerpc/tm/tm-tmspr.c
new file mode 100644
index 000000000..794d574db
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-tmspr.c
@@ -0,0 +1,143 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2015, Michael Neuling, IBM Corp.
+ *
+ * Original: Michael Neuling 3/4/2014
+ * Modified: Rashmica Gupta 8/12/2015
+ *
+ * Check if any of the Transaction Memory SPRs get corrupted.
+ * - TFIAR - stores address of location of transaction failure
+ * - TFHAR - stores address of software failure handler (if transaction
+ * fails)
+ * - TEXASR - lots of info about the transacion(s)
+ *
+ * (1) create more threads than cpus
+ * (2) in each thread:
+ * (a) set TFIAR and TFHAR a unique value
+ * (b) loop for awhile, continually checking to see if
+ * either register has been corrupted.
+ *
+ * (3) Loop:
+ * (a) begin transaction
+ * (b) abort transaction
+ * (c) check TEXASR to see if FS has been corrupted
+ */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <pthread.h>
+#include <string.h>
+
+#include "utils.h"
+#include "tm.h"
+
+int num_loops = 1000000;
+int passed = 1;
+
+void tfiar_tfhar(void *in)
+{
+ unsigned long tfhar, tfhar_rd, tfiar, tfiar_rd;
+ int i;
+
+ /* TFIAR: Last bit has to be high so userspace can read register */
+ tfiar = ((unsigned long)in) + 1;
+ tfiar += 2;
+ mtspr(SPRN_TFIAR, tfiar);
+
+ /* TFHAR: Last two bits are reserved */
+ tfhar = ((unsigned long)in);
+ tfhar &= ~0x3UL;
+ tfhar += 4;
+ mtspr(SPRN_TFHAR, tfhar);
+
+ for (i = 0; i < num_loops; i++) {
+ tfhar_rd = mfspr(SPRN_TFHAR);
+ tfiar_rd = mfspr(SPRN_TFIAR);
+ if ( (tfhar != tfhar_rd) || (tfiar != tfiar_rd) ) {
+ passed = 0;
+ return;
+ }
+ }
+ return;
+}
+
+void texasr(void *in)
+{
+ unsigned long i;
+ uint64_t result = 0;
+
+ for (i = 0; i < num_loops; i++) {
+ asm __volatile__(
+ "tbegin.;"
+ "beq 3f ;"
+ "tabort. 0 ;"
+ "tend.;"
+
+ /* Abort handler */
+ "3: ;"
+ ::: "memory");
+
+ /* Check the TEXASR */
+ result = mfspr(SPRN_TEXASR);
+ if ((result & TEXASR_FS) == 0) {
+ passed = 0;
+ return;
+ }
+ }
+ return;
+}
+
+int test_tmspr()
+{
+ pthread_t *thread;
+ int thread_num;
+ unsigned long i;
+
+ SKIP_IF(!have_htm());
+
+ /* To cause some context switching */
+ thread_num = 10 * sysconf(_SC_NPROCESSORS_ONLN);
+
+ thread = malloc(thread_num * sizeof(pthread_t));
+ if (thread == NULL)
+ return EXIT_FAILURE;
+
+ /* Test TFIAR and TFHAR */
+ for (i = 0; i < thread_num; i += 2) {
+ if (pthread_create(&thread[i], NULL, (void *)tfiar_tfhar,
+ (void *)i))
+ return EXIT_FAILURE;
+ }
+ /* Test TEXASR */
+ for (i = 1; i < thread_num; i += 2) {
+ if (pthread_create(&thread[i], NULL, (void *)texasr, (void *)i))
+ return EXIT_FAILURE;
+ }
+
+ for (i = 0; i < thread_num; i++) {
+ if (pthread_join(thread[i], NULL) != 0)
+ return EXIT_FAILURE;
+ }
+
+ free(thread);
+
+ if (passed)
+ return 0;
+ else
+ return 1;
+}
+
+int main(int argc, char *argv[])
+{
+ if (argc > 1) {
+ if (strcmp(argv[1], "-h") == 0) {
+ printf("Syntax:\t [<num loops>]\n");
+ return 0;
+ } else {
+ num_loops = atoi(argv[1]);
+ }
+ }
+ return test_harness(test_tmspr, "tm_tmspr");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-trap.c b/tools/testing/selftests/powerpc/tm/tm-trap.c
new file mode 100644
index 000000000..c75960af8
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-trap.c
@@ -0,0 +1,333 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2017, Gustavo Romero, IBM Corp.
+ *
+ * Check if thread endianness is flipped inadvertently to BE on trap
+ * caught in TM whilst MSR.FP and MSR.VEC are zero (i.e. just after
+ * load_fp and load_vec overflowed).
+ *
+ * The issue can be checked on LE machines simply by zeroing load_fp
+ * and load_vec and then causing a trap in TM. Since the endianness
+ * changes to BE on return from the signal handler, 'nop' is
+ * thread as an illegal instruction in following sequence:
+ * tbegin.
+ * beq 1f
+ * trap
+ * tend.
+ * 1: nop
+ *
+ * However, although the issue is also present on BE machines, it's a
+ * bit trickier to check it on BE machines because MSR.LE bit is set
+ * to zero which determines a BE endianness that is the native
+ * endianness on BE machines, so nothing notably critical happens,
+ * i.e. no illegal instruction is observed immediately after returning
+ * from the signal handler (as it happens on LE machines). Thus to test
+ * it on BE machines LE endianness is forced after a first trap and then
+ * the endianness is verified on subsequent traps to determine if the
+ * endianness "flipped back" to the native endianness (BE).
+ */
+
+#define _GNU_SOURCE
+#include <error.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <htmintrin.h>
+#include <inttypes.h>
+#include <pthread.h>
+#include <sched.h>
+#include <signal.h>
+#include <stdbool.h>
+
+#include "tm.h"
+#include "utils.h"
+
+#define pr_error(error_code, format, ...) \
+ error_at_line(1, error_code, __FILE__, __LINE__, format, ##__VA_ARGS__)
+
+#define MSR_LE 1UL
+#define LE 1UL
+
+pthread_t t0_ping;
+pthread_t t1_pong;
+
+int exit_from_pong;
+
+int trap_event;
+int le;
+
+bool success;
+
+void trap_signal_handler(int signo, siginfo_t *si, void *uc)
+{
+ ucontext_t *ucp = uc;
+ uint64_t thread_endianness;
+
+ /* Get thread endianness: extract bit LE from MSR */
+ thread_endianness = MSR_LE & ucp->uc_mcontext.gp_regs[PT_MSR];
+
+ /***
+ * Little-Endian Machine
+ */
+
+ if (le) {
+ /* First trap event */
+ if (trap_event == 0) {
+ /* Do nothing. Since it is returning from this trap
+ * event that endianness is flipped by the bug, so just
+ * let the process return from the signal handler and
+ * check on the second trap event if endianness is
+ * flipped or not.
+ */
+ }
+ /* Second trap event */
+ else if (trap_event == 1) {
+ /*
+ * Since trap was caught in TM on first trap event, if
+ * endianness was still LE (not flipped inadvertently)
+ * after returning from the signal handler instruction
+ * (1) is executed (basically a 'nop'), as it's located
+ * at address of tbegin. +4 (rollback addr). As (1) on
+ * LE endianness does in effect nothing, instruction (2)
+ * is then executed again as 'trap', generating a second
+ * trap event (note that in that case 'trap' is caught
+ * not in transacional mode). On te other hand, if after
+ * the return from the signal handler the endianness in-
+ * advertently flipped, instruction (1) is tread as a
+ * branch instruction, i.e. b .+8, hence instruction (3)
+ * and (4) are executed (tbegin.; trap;) and we get sim-
+ * ilaly on the trap signal handler, but now in TM mode.
+ * Either way, it's now possible to check the MSR LE bit
+ * once in the trap handler to verify if endianness was
+ * flipped or not after the return from the second trap
+ * event. If endianness is flipped, the bug is present.
+ * Finally, getting a trap in TM mode or not is just
+ * worth noting because it affects the math to determine
+ * the offset added to the NIP on return: the NIP for a
+ * trap caught in TM is the rollback address, i.e. the
+ * next instruction after 'tbegin.', whilst the NIP for
+ * a trap caught in non-transactional mode is the very
+ * same address of the 'trap' instruction that generated
+ * the trap event.
+ */
+
+ if (thread_endianness == LE) {
+ /* Go to 'success', i.e. instruction (6) */
+ ucp->uc_mcontext.gp_regs[PT_NIP] += 16;
+ } else {
+ /*
+ * Thread endianness is BE, so it flipped
+ * inadvertently. Thus we flip back to LE and
+ * set NIP to go to 'failure', instruction (5).
+ */
+ ucp->uc_mcontext.gp_regs[PT_MSR] |= 1UL;
+ ucp->uc_mcontext.gp_regs[PT_NIP] += 4;
+ }
+ }
+ }
+
+ /***
+ * Big-Endian Machine
+ */
+
+ else {
+ /* First trap event */
+ if (trap_event == 0) {
+ /*
+ * Force thread endianness to be LE. Instructions (1),
+ * (3), and (4) will be executed, generating a second
+ * trap in TM mode.
+ */
+ ucp->uc_mcontext.gp_regs[PT_MSR] |= 1UL;
+ }
+ /* Second trap event */
+ else if (trap_event == 1) {
+ /*
+ * Do nothing. If bug is present on return from this
+ * second trap event endianness will flip back "automat-
+ * ically" to BE, otherwise thread endianness will
+ * continue to be LE, just as it was set above.
+ */
+ }
+ /* A third trap event */
+ else {
+ /*
+ * Once here it means that after returning from the sec-
+ * ond trap event instruction (4) (trap) was executed
+ * as LE, generating a third trap event. In that case
+ * endianness is still LE as set on return from the
+ * first trap event, hence no bug. Otherwise, bug
+ * flipped back to BE on return from the second trap
+ * event and instruction (4) was executed as 'tdi' (so
+ * basically a 'nop') and branch to 'failure' in
+ * instruction (5) was taken to indicate failure and we
+ * never get here.
+ */
+
+ /*
+ * Flip back to BE and go to instruction (6), i.e. go to
+ * 'success'.
+ */
+ ucp->uc_mcontext.gp_regs[PT_MSR] &= ~1UL;
+ ucp->uc_mcontext.gp_regs[PT_NIP] += 8;
+ }
+ }
+
+ trap_event++;
+}
+
+void usr1_signal_handler(int signo, siginfo_t *si, void *not_used)
+{
+ /* Got a USR1 signal from ping(), so just tell pong() to exit */
+ exit_from_pong = 1;
+}
+
+void *ping(void *not_used)
+{
+ uint64_t i;
+
+ trap_event = 0;
+
+ /*
+ * Wait an amount of context switches so load_fp and load_vec overflows
+ * and MSR_[FP|VEC|V] is 0.
+ */
+ for (i = 0; i < 1024*1024*512; i++)
+ ;
+
+ asm goto(
+ /*
+ * [NA] means "Native Endianness", i.e. it tells how a
+ * instruction is executed on machine's native endianness (in
+ * other words, native endianness matches kernel endianness).
+ * [OP] means "Opposite Endianness", i.e. on a BE machine, it
+ * tells how a instruction is executed as a LE instruction; con-
+ * versely, on a LE machine, it tells how a instruction is
+ * executed as a BE instruction. When [NA] is omitted, it means
+ * that the native interpretation of a given instruction is not
+ * relevant for the test. Likewise when [OP] is omitted.
+ */
+
+ " tbegin. ;" /* (0) tbegin. [NA] */
+ " tdi 0, 0, 0x48;" /* (1) nop [NA]; b (3) [OP] */
+ " trap ;" /* (2) trap [NA] */
+ ".long 0x1D05007C;" /* (3) tbegin. [OP] */
+ ".long 0x0800E07F;" /* (4) trap [OP]; nop [NA] */
+ " b %l[failure] ;" /* (5) b [NA]; MSR.LE flipped (bug) */
+ " b %l[success] ;" /* (6) b [NA]; MSR.LE did not flip (ok)*/
+
+ : : : : failure, success);
+
+failure:
+ success = false;
+ goto exit_from_ping;
+
+success:
+ success = true;
+
+exit_from_ping:
+ /* Tell pong() to exit before leaving */
+ pthread_kill(t1_pong, SIGUSR1);
+ return NULL;
+}
+
+void *pong(void *not_used)
+{
+ while (!exit_from_pong)
+ /*
+ * Induce context switches on ping() thread
+ * until ping() finishes its job and signs
+ * to exit from this loop.
+ */
+ sched_yield();
+
+ return NULL;
+}
+
+int tm_trap_test(void)
+{
+ uint16_t k = 1;
+ int cpu, rc;
+
+ pthread_attr_t attr;
+ cpu_set_t cpuset;
+
+ struct sigaction trap_sa;
+
+ SKIP_IF(!have_htm());
+
+ trap_sa.sa_flags = SA_SIGINFO;
+ trap_sa.sa_sigaction = trap_signal_handler;
+ sigaction(SIGTRAP, &trap_sa, NULL);
+
+ struct sigaction usr1_sa;
+
+ usr1_sa.sa_flags = SA_SIGINFO;
+ usr1_sa.sa_sigaction = usr1_signal_handler;
+ sigaction(SIGUSR1, &usr1_sa, NULL);
+
+ cpu = pick_online_cpu();
+ FAIL_IF(cpu < 0);
+
+ // Set only one CPU in the mask. Both threads will be bound to that CPU.
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu, &cpuset);
+
+ /* Init pthread attribute */
+ rc = pthread_attr_init(&attr);
+ if (rc)
+ pr_error(rc, "pthread_attr_init()");
+
+ /*
+ * Bind thread ping() and pong() both to CPU 0 so they ping-pong and
+ * speed up context switches on ping() thread, speeding up the load_fp
+ * and load_vec overflow.
+ */
+ rc = pthread_attr_setaffinity_np(&attr, sizeof(cpu_set_t), &cpuset);
+ if (rc)
+ pr_error(rc, "pthread_attr_setaffinity()");
+
+ /* Figure out the machine endianness */
+ le = (int) *(uint8_t *)&k;
+
+ printf("%s machine detected. Checking if endianness flips %s",
+ le ? "Little-Endian" : "Big-Endian",
+ "inadvertently on trap in TM... ");
+
+ rc = fflush(0);
+ if (rc)
+ pr_error(rc, "fflush()");
+
+ /* Launch ping() */
+ rc = pthread_create(&t0_ping, &attr, ping, NULL);
+ if (rc)
+ pr_error(rc, "pthread_create()");
+
+ exit_from_pong = 0;
+
+ /* Launch pong() */
+ rc = pthread_create(&t1_pong, &attr, pong, NULL);
+ if (rc)
+ pr_error(rc, "pthread_create()");
+
+ rc = pthread_join(t0_ping, NULL);
+ if (rc)
+ pr_error(rc, "pthread_join()");
+
+ rc = pthread_join(t1_pong, NULL);
+ if (rc)
+ pr_error(rc, "pthread_join()");
+
+ if (success) {
+ printf("no.\n"); /* no, endianness did not flip inadvertently */
+ return EXIT_SUCCESS;
+ }
+
+ printf("yes!\n"); /* yes, endianness did flip inadvertently */
+ return EXIT_FAILURE;
+}
+
+int main(int argc, char **argv)
+{
+ return test_harness(tm_trap_test, "tm_trap_test");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-unavailable.c b/tools/testing/selftests/powerpc/tm/tm-unavailable.c
new file mode 100644
index 000000000..a1348a5f7
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-unavailable.c
@@ -0,0 +1,410 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2017, Gustavo Romero, Breno Leitao, Cyril Bur, IBM Corp.
+ *
+ * Force FP, VEC and VSX unavailable exception during transaction in all
+ * possible scenarios regarding the MSR.FP and MSR.VEC state, e.g. when FP
+ * is enable and VEC is disable, when FP is disable and VEC is enable, and
+ * so on. Then we check if the restored state is correctly set for the
+ * FP and VEC registers to the previous state we set just before we entered
+ * in TM, i.e. we check if it corrupts somehow the recheckpointed FP and
+ * VEC/Altivec registers on abortion due to an unavailable exception in TM.
+ * N.B. In this test we do not test all the FP/Altivec/VSX registers for
+ * corruption, but only for registers vs0 and vs32, which are respectively
+ * representatives of FP and VEC/Altivec reg sets.
+ */
+
+#define _GNU_SOURCE
+#include <error.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <inttypes.h>
+#include <stdbool.h>
+#include <pthread.h>
+#include <sched.h>
+
+#include "tm.h"
+
+#define DEBUG 0
+
+/* Unavailable exceptions to test in HTM */
+#define FP_UNA_EXCEPTION 0
+#define VEC_UNA_EXCEPTION 1
+#define VSX_UNA_EXCEPTION 2
+
+#define NUM_EXCEPTIONS 3
+#define err_at_line(status, errnum, format, ...) \
+ error_at_line(status, errnum, __FILE__, __LINE__, format ##__VA_ARGS__)
+
+#define pr_warn(code, format, ...) err_at_line(0, code, format, ##__VA_ARGS__)
+#define pr_err(code, format, ...) err_at_line(1, code, format, ##__VA_ARGS__)
+
+struct Flags {
+ int touch_fp;
+ int touch_vec;
+ int result;
+ int exception;
+} flags;
+
+bool expecting_failure(void)
+{
+ if (flags.touch_fp && flags.exception == FP_UNA_EXCEPTION)
+ return false;
+
+ if (flags.touch_vec && flags.exception == VEC_UNA_EXCEPTION)
+ return false;
+
+ /*
+ * If both FP and VEC are touched it does not mean that touching VSX
+ * won't raise an exception. However since FP and VEC state are already
+ * correctly loaded, the transaction is not aborted (i.e.
+ * treclaimed/trecheckpointed) and MSR.VSX is just set as 1, so a TM
+ * failure is not expected also in this case.
+ */
+ if ((flags.touch_fp && flags.touch_vec) &&
+ flags.exception == VSX_UNA_EXCEPTION)
+ return false;
+
+ return true;
+}
+
+/* Check if failure occurred whilst in transaction. */
+bool is_failure(uint64_t condition_reg)
+{
+ /*
+ * When failure handling occurs, CR0 is set to 0b1010 (0xa). Otherwise
+ * transaction completes without failure and hence reaches out 'tend.'
+ * that sets CR0 to 0b0100 (0x4).
+ */
+ return ((condition_reg >> 28) & 0xa) == 0xa;
+}
+
+void *tm_una_ping(void *input)
+{
+
+ /*
+ * Expected values for vs0 and vs32 after a TM failure. They must never
+ * change, otherwise they got corrupted.
+ */
+ uint64_t high_vs0 = 0x5555555555555555;
+ uint64_t low_vs0 = 0xffffffffffffffff;
+ uint64_t high_vs32 = 0x5555555555555555;
+ uint64_t low_vs32 = 0xffffffffffffffff;
+
+ /* Counter for busy wait */
+ uint64_t counter = 0x1ff000000;
+
+ /*
+ * Variable to keep a copy of CR register content taken just after we
+ * leave the transactional state.
+ */
+ uint64_t cr_ = 0;
+
+ /*
+ * Wait a bit so thread can get its name "ping". This is not important
+ * to reproduce the issue but it's nice to have for systemtap debugging.
+ */
+ if (DEBUG)
+ sleep(1);
+
+ printf("If MSR.FP=%d MSR.VEC=%d: ", flags.touch_fp, flags.touch_vec);
+
+ if (flags.exception != FP_UNA_EXCEPTION &&
+ flags.exception != VEC_UNA_EXCEPTION &&
+ flags.exception != VSX_UNA_EXCEPTION) {
+ printf("No valid exception specified to test.\n");
+ return NULL;
+ }
+
+ asm (
+ /* Prepare to merge low and high. */
+ " mtvsrd 33, %[high_vs0] ;"
+ " mtvsrd 34, %[low_vs0] ;"
+
+ /*
+ * Adjust VS0 expected value after an TM failure,
+ * i.e. vs0 = 0x5555555555555555555FFFFFFFFFFFFFFFF
+ */
+ " xxmrghd 0, 33, 34 ;"
+
+ /*
+ * Adjust VS32 expected value after an TM failure,
+ * i.e. vs32 = 0x5555555555555555555FFFFFFFFFFFFFFFF
+ */
+ " xxmrghd 32, 33, 34 ;"
+
+ /*
+ * Wait an amount of context switches so load_fp and load_vec
+ * overflow and MSR.FP, MSR.VEC, and MSR.VSX become zero (off).
+ */
+ " mtctr %[counter] ;"
+
+ /* Decrement CTR branch if CTR non zero. */
+ "1: bdnz 1b ;"
+
+ /*
+ * Check if we want to touch FP prior to the test in order
+ * to set MSR.FP = 1 before provoking an unavailable
+ * exception in TM.
+ */
+ " cmpldi %[touch_fp], 0 ;"
+ " beq no_fp ;"
+ " fadd 10, 10, 10 ;"
+ "no_fp: ;"
+
+ /*
+ * Check if we want to touch VEC prior to the test in order
+ * to set MSR.VEC = 1 before provoking an unavailable
+ * exception in TM.
+ */
+ " cmpldi %[touch_vec], 0 ;"
+ " beq no_vec ;"
+ " vaddcuw 10, 10, 10 ;"
+ "no_vec: ;"
+
+ /*
+ * Perhaps it would be a better idea to do the
+ * compares outside transactional context and simply
+ * duplicate code.
+ */
+ " tbegin. ;"
+ " beq trans_fail ;"
+
+ /* Do we do FP Unavailable? */
+ " cmpldi %[exception], %[ex_fp] ;"
+ " bne 1f ;"
+ " fadd 10, 10, 10 ;"
+ " b done ;"
+
+ /* Do we do VEC Unavailable? */
+ "1: cmpldi %[exception], %[ex_vec] ;"
+ " bne 2f ;"
+ " vaddcuw 10, 10, 10 ;"
+ " b done ;"
+
+ /*
+ * Not FP or VEC, therefore VSX. Ensure this
+ * instruction always generates a VSX Unavailable.
+ * ISA 3.0 is tricky here.
+ * (xxmrghd will on ISA 2.07 and ISA 3.0)
+ */
+ "2: xxmrghd 10, 10, 10 ;"
+
+ "done: tend. ;"
+
+ "trans_fail: ;"
+
+ /* Give values back to C. */
+ " mfvsrd %[high_vs0], 0 ;"
+ " xxsldwi 3, 0, 0, 2 ;"
+ " mfvsrd %[low_vs0], 3 ;"
+ " mfvsrd %[high_vs32], 32 ;"
+ " xxsldwi 3, 32, 32, 2 ;"
+ " mfvsrd %[low_vs32], 3 ;"
+
+ /* Give CR back to C so that it can check what happened. */
+ " mfcr %[cr_] ;"
+
+ : [high_vs0] "+r" (high_vs0),
+ [low_vs0] "+r" (low_vs0),
+ [high_vs32] "=r" (high_vs32),
+ [low_vs32] "=r" (low_vs32),
+ [cr_] "+r" (cr_)
+ : [touch_fp] "r" (flags.touch_fp),
+ [touch_vec] "r" (flags.touch_vec),
+ [exception] "r" (flags.exception),
+ [ex_fp] "i" (FP_UNA_EXCEPTION),
+ [ex_vec] "i" (VEC_UNA_EXCEPTION),
+ [ex_vsx] "i" (VSX_UNA_EXCEPTION),
+ [counter] "r" (counter)
+
+ : "cr0", "ctr", "v10", "vs0", "vs10", "vs3", "vs32", "vs33",
+ "vs34", "fr10"
+
+ );
+
+ /*
+ * Check if we were expecting a failure and it did not occur by checking
+ * CR0 state just after we leave the transaction. Either way we check if
+ * vs0 or vs32 got corrupted.
+ */
+ if (expecting_failure() && !is_failure(cr_)) {
+ printf("\n\tExpecting the transaction to fail, %s",
+ "but it didn't\n\t");
+ flags.result++;
+ }
+
+ /* Check if we were not expecting a failure and a it occurred. */
+ if (!expecting_failure() && is_failure(cr_) &&
+ !failure_is_reschedule()) {
+ printf("\n\tUnexpected transaction failure 0x%02lx\n\t",
+ failure_code());
+ return (void *) -1;
+ }
+
+ /*
+ * Check if TM failed due to the cause we were expecting. 0xda is a
+ * TM_CAUSE_FAC_UNAV cause, otherwise it's an unexpected cause, unless
+ * it was caused by a reschedule.
+ */
+ if (is_failure(cr_) && !failure_is_unavailable() &&
+ !failure_is_reschedule()) {
+ printf("\n\tUnexpected failure cause 0x%02lx\n\t",
+ failure_code());
+ return (void *) -1;
+ }
+
+ /* 0x4 is a success and 0xa is a fail. See comment in is_failure(). */
+ if (DEBUG)
+ printf("CR0: 0x%1lx ", cr_ >> 28);
+
+ /* Check FP (vs0) for the expected value. */
+ if (high_vs0 != 0x5555555555555555 || low_vs0 != 0xFFFFFFFFFFFFFFFF) {
+ printf("FP corrupted!");
+ printf(" high = %#16" PRIx64 " low = %#16" PRIx64 " ",
+ high_vs0, low_vs0);
+ flags.result++;
+ } else
+ printf("FP ok ");
+
+ /* Check VEC (vs32) for the expected value. */
+ if (high_vs32 != 0x5555555555555555 || low_vs32 != 0xFFFFFFFFFFFFFFFF) {
+ printf("VEC corrupted!");
+ printf(" high = %#16" PRIx64 " low = %#16" PRIx64,
+ high_vs32, low_vs32);
+ flags.result++;
+ } else
+ printf("VEC ok");
+
+ putchar('\n');
+
+ return NULL;
+}
+
+/* Thread to force context switch */
+void *tm_una_pong(void *not_used)
+{
+ /* Wait thread get its name "pong". */
+ if (DEBUG)
+ sleep(1);
+
+ /* Classed as an interactive-like thread. */
+ while (1)
+ sched_yield();
+}
+
+/* Function that creates a thread and launches the "ping" task. */
+void test_fp_vec(int fp, int vec, pthread_attr_t *attr)
+{
+ int retries = 2;
+ void *ret_value;
+ pthread_t t0;
+
+ flags.touch_fp = fp;
+ flags.touch_vec = vec;
+
+ /*
+ * Without luck it's possible that the transaction is aborted not due to
+ * the unavailable exception caught in the middle as we expect but also,
+ * for instance, due to a context switch or due to a KVM reschedule (if
+ * it's running on a VM). Thus we try a few times before giving up,
+ * checking if the failure cause is the one we expect.
+ */
+ do {
+ int rc;
+
+ /* Bind to CPU 0, as specified in 'attr'. */
+ rc = pthread_create(&t0, attr, tm_una_ping, (void *) &flags);
+ if (rc)
+ pr_err(rc, "pthread_create()");
+ rc = pthread_setname_np(t0, "tm_una_ping");
+ if (rc)
+ pr_warn(rc, "pthread_setname_np");
+ rc = pthread_join(t0, &ret_value);
+ if (rc)
+ pr_err(rc, "pthread_join");
+
+ retries--;
+ } while (ret_value != NULL && retries);
+
+ if (!retries) {
+ flags.result = 1;
+ if (DEBUG)
+ printf("All transactions failed unexpectedly\n");
+
+ }
+}
+
+int tm_unavailable_test(void)
+{
+ int cpu, rc, exception; /* FP = 0, VEC = 1, VSX = 2 */
+ pthread_t t1;
+ pthread_attr_t attr;
+ cpu_set_t cpuset;
+
+ SKIP_IF(!have_htm());
+
+ cpu = pick_online_cpu();
+ FAIL_IF(cpu < 0);
+
+ // Set only one CPU in the mask. Both threads will be bound to that CPU.
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu, &cpuset);
+
+ /* Init pthread attribute. */
+ rc = pthread_attr_init(&attr);
+ if (rc)
+ pr_err(rc, "pthread_attr_init()");
+
+ /* Set CPU 0 mask into the pthread attribute. */
+ rc = pthread_attr_setaffinity_np(&attr, sizeof(cpu_set_t), &cpuset);
+ if (rc)
+ pr_err(rc, "pthread_attr_setaffinity_np()");
+
+ rc = pthread_create(&t1, &attr /* Bind to CPU 0 */, tm_una_pong, NULL);
+ if (rc)
+ pr_err(rc, "pthread_create()");
+
+ /* Name it for systemtap convenience */
+ rc = pthread_setname_np(t1, "tm_una_pong");
+ if (rc)
+ pr_warn(rc, "pthread_create()");
+
+ flags.result = 0;
+
+ for (exception = 0; exception < NUM_EXCEPTIONS; exception++) {
+ printf("Checking if FP/VEC registers are sane after");
+
+ if (exception == FP_UNA_EXCEPTION)
+ printf(" a FP unavailable exception...\n");
+
+ else if (exception == VEC_UNA_EXCEPTION)
+ printf(" a VEC unavailable exception...\n");
+
+ else
+ printf(" a VSX unavailable exception...\n");
+
+ flags.exception = exception;
+
+ test_fp_vec(0, 0, &attr);
+ test_fp_vec(1, 0, &attr);
+ test_fp_vec(0, 1, &attr);
+ test_fp_vec(1, 1, &attr);
+
+ }
+
+ if (flags.result > 0) {
+ printf("result: failed!\n");
+ exit(1);
+ } else {
+ printf("result: success\n");
+ exit(0);
+ }
+}
+
+int main(int argc, char **argv)
+{
+ test_harness_set_timeout(220);
+ return test_harness(tm_unavailable_test, "tm_unavailable_test");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-vmx-unavail.c b/tools/testing/selftests/powerpc/tm/tm-vmx-unavail.c
new file mode 100644
index 000000000..e2a0c07e8
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-vmx-unavail.c
@@ -0,0 +1,118 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2017, Michael Neuling, IBM Corp.
+ * Original: Breno Leitao <brenohl@br.ibm.com> &
+ * Gustavo Bueno Romero <gromero@br.ibm.com>
+ * Edited: Michael Neuling
+ *
+ * Force VMX unavailable during a transaction and see if it corrupts
+ * the checkpointed VMX register state after the abort.
+ */
+
+#include <inttypes.h>
+#include <htmintrin.h>
+#include <string.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <pthread.h>
+#include <sys/mman.h>
+#include <unistd.h>
+#include <pthread.h>
+
+#include "tm.h"
+#include "utils.h"
+
+int passed;
+
+void *worker(void *unused)
+{
+ __int128 vmx0;
+ uint64_t texasr;
+
+ asm goto (
+ "li 3, 1;" /* Stick non-zero value in VMX0 */
+ "std 3, 0(%[vmx0_ptr]);"
+ "lvx 0, 0, %[vmx0_ptr];"
+
+ /* Wait here a bit so we get scheduled out 255 times */
+ "lis 3, 0x3fff;"
+ "1: ;"
+ "addi 3, 3, -1;"
+ "cmpdi 3, 0;"
+ "bne 1b;"
+
+ /* Kernel will hopefully turn VMX off now */
+
+ "tbegin. ;"
+ "beq failure;"
+
+ /* Cause VMX unavail. Any VMX instruction */
+ "vaddcuw 0,0,0;"
+
+ "tend. ;"
+ "b %l[success];"
+
+ /* Check VMX0 sanity after abort */
+ "failure: ;"
+ "lvx 1, 0, %[vmx0_ptr];"
+ "vcmpequb. 2, 0, 1;"
+ "bc 4, 24, %l[value_mismatch];"
+ "b %l[value_match];"
+ :
+ : [vmx0_ptr] "r"(&vmx0)
+ : "r3"
+ : success, value_match, value_mismatch
+ );
+
+ /* HTM aborted and VMX0 is corrupted */
+value_mismatch:
+ texasr = __builtin_get_texasr();
+
+ printf("\n\n==============\n\n");
+ printf("Failure with error: %lx\n", _TEXASR_FAILURE_CODE(texasr));
+ printf("Summary error : %lx\n", _TEXASR_FAILURE_SUMMARY(texasr));
+ printf("TFIAR exact : %lx\n\n", _TEXASR_TFIAR_EXACT(texasr));
+
+ passed = 0;
+ return NULL;
+
+ /* HTM aborted but VMX0 is correct */
+value_match:
+// printf("!");
+ return NULL;
+
+success:
+// printf(".");
+ return NULL;
+}
+
+int tm_vmx_unavail_test()
+{
+ int threads;
+ pthread_t *thread;
+
+ SKIP_IF(!have_htm());
+
+ passed = 1;
+
+ threads = sysconf(_SC_NPROCESSORS_ONLN) * 4;
+ thread = malloc(sizeof(pthread_t)*threads);
+ if (!thread)
+ return EXIT_FAILURE;
+
+ for (uint64_t i = 0; i < threads; i++)
+ pthread_create(&thread[i], NULL, &worker, NULL);
+
+ for (uint64_t i = 0; i < threads; i++)
+ pthread_join(thread[i], NULL);
+
+ free(thread);
+
+ return passed ? EXIT_SUCCESS : EXIT_FAILURE;
+}
+
+
+int main(int argc, char **argv)
+{
+ return test_harness(tm_vmx_unavail_test, "tm_vmx_unavail_test");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm-vmxcopy.c b/tools/testing/selftests/powerpc/tm/tm-vmxcopy.c
new file mode 100644
index 000000000..c1e788a6d
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm-vmxcopy.c
@@ -0,0 +1,104 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2015, Michael Neuling, IBM Corp.
+ *
+ * Original: Michael Neuling 4/12/2013
+ * Edited: Rashmica Gupta 4/12/2015
+ *
+ * See if the altivec state is leaked out of an aborted transaction due to
+ * kernel vmx copy loops.
+ *
+ * When the transaction aborts, VSR values should rollback to the values
+ * they held before the transaction commenced. Using VSRs while transaction
+ * is suspended should not affect the checkpointed values.
+ *
+ * (1) write A to a VSR
+ * (2) start transaction
+ * (3) suspend transaction
+ * (4) change the VSR to B
+ * (5) trigger kernel vmx copy loop
+ * (6) abort transaction
+ * (7) check that the VSR value is A
+ */
+
+#include <inttypes.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <sys/mman.h>
+#include <string.h>
+#include <assert.h>
+
+#include "tm.h"
+#include "utils.h"
+
+int test_vmxcopy()
+{
+ long double vecin = 1.3;
+ long double vecout;
+ unsigned long pgsize = getpagesize();
+ int i;
+ int fd;
+ int size = pgsize*16;
+ char tmpfile[] = "/tmp/page_faultXXXXXX";
+ char buf[pgsize];
+ char *a;
+ uint64_t aborted = 0;
+
+ SKIP_IF(!have_htm());
+ SKIP_IF(!is_ppc64le());
+
+ fd = mkstemp(tmpfile);
+ assert(fd >= 0);
+
+ memset(buf, 0, pgsize);
+ for (i = 0; i < size; i += pgsize)
+ assert(write(fd, buf, pgsize) == pgsize);
+
+ unlink(tmpfile);
+
+ a = mmap(NULL, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
+ assert(a != MAP_FAILED);
+
+ asm __volatile__(
+ "lxvd2x 40,0,%[vecinptr];" /* set 40 to initial value*/
+ "tbegin.;"
+ "beq 3f;"
+ "tsuspend.;"
+ "xxlxor 40,40,40;" /* set 40 to 0 */
+ "std 5, 0(%[map]);" /* cause kernel vmx copy page */
+ "tabort. 0;"
+ "tresume.;"
+ "tend.;"
+ "li %[res], 0;"
+ "b 5f;"
+
+ /* Abort handler */
+ "3:;"
+ "li %[res], 1;"
+
+ "5:;"
+ "stxvd2x 40,0,%[vecoutptr];"
+ : [res]"=&r"(aborted)
+ : [vecinptr]"r"(&vecin),
+ [vecoutptr]"r"(&vecout),
+ [map]"r"(a)
+ : "memory", "r0", "r3", "r4", "r5", "r6", "r7");
+
+ if (aborted && (vecin != vecout)){
+ printf("FAILED: vector state leaked on abort %f != %f\n",
+ (double)vecin, (double)vecout);
+ return 1;
+ }
+
+ munmap(a, size);
+
+ close(fd);
+
+ return 0;
+}
+
+int main(void)
+{
+ return test_harness(test_vmxcopy, "tm_vmxcopy");
+}
diff --git a/tools/testing/selftests/powerpc/tm/tm.h b/tools/testing/selftests/powerpc/tm/tm.h
new file mode 100644
index 000000000..c5a1e5c16
--- /dev/null
+++ b/tools/testing/selftests/powerpc/tm/tm.h
@@ -0,0 +1,96 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright 2015, Michael Ellerman, IBM Corp.
+ */
+
+#ifndef _SELFTESTS_POWERPC_TM_TM_H
+#define _SELFTESTS_POWERPC_TM_TM_H
+
+#include <stdbool.h>
+#include <asm/tm.h>
+
+#include "utils.h"
+
+static inline bool have_htm(void)
+{
+#ifdef PPC_FEATURE2_HTM
+ return have_hwcap2(PPC_FEATURE2_HTM);
+#else
+ printf("PPC_FEATURE2_HTM not defined, can't check AT_HWCAP2\n");
+ return false;
+#endif
+}
+
+static inline bool have_htm_nosc(void)
+{
+#ifdef PPC_FEATURE2_HTM_NOSC
+ return have_hwcap2(PPC_FEATURE2_HTM_NOSC);
+#else
+ printf("PPC_FEATURE2_HTM_NOSC not defined, can't check AT_HWCAP2\n");
+ return false;
+#endif
+}
+
+static inline long failure_code(void)
+{
+ return __builtin_get_texasru() >> 24;
+}
+
+static inline bool failure_is_persistent(void)
+{
+ return (failure_code() & TM_CAUSE_PERSISTENT) == TM_CAUSE_PERSISTENT;
+}
+
+static inline bool failure_is_syscall(void)
+{
+ return (failure_code() & TM_CAUSE_SYSCALL) == TM_CAUSE_SYSCALL;
+}
+
+static inline bool failure_is_unavailable(void)
+{
+ return (failure_code() & TM_CAUSE_FAC_UNAV) == TM_CAUSE_FAC_UNAV;
+}
+
+static inline bool failure_is_reschedule(void)
+{
+ if ((failure_code() & TM_CAUSE_RESCHED) == TM_CAUSE_RESCHED ||
+ (failure_code() & TM_CAUSE_KVM_RESCHED) == TM_CAUSE_KVM_RESCHED ||
+ (failure_code() & TM_CAUSE_KVM_FAC_UNAV) == TM_CAUSE_KVM_FAC_UNAV)
+ return true;
+
+ return false;
+}
+
+static inline bool failure_is_nesting(void)
+{
+ return (__builtin_get_texasru() & 0x400000);
+}
+
+static inline int tcheck(void)
+{
+ long cr;
+ asm volatile ("tcheck 0" : "=r"(cr) : : "cr0");
+ return (cr >> 28) & 4;
+}
+
+static inline bool tcheck_doomed(void)
+{
+ return tcheck() & 8;
+}
+
+static inline bool tcheck_active(void)
+{
+ return tcheck() & 4;
+}
+
+static inline bool tcheck_suspended(void)
+{
+ return tcheck() & 2;
+}
+
+static inline bool tcheck_transactional(void)
+{
+ return tcheck() & 6;
+}
+
+#endif /* _SELFTESTS_POWERPC_TM_TM_H */