Snap for 8192738 from d4063865cc3ab201ab3cc24965b1d2896902cb8b to tm-frc-extservices-release

Change-Id: Ibcf844e6f1566d5c56ea83277e9f136f0b8b73ce

10 files changed, 519 insertions, 31 deletions

diff --git a/Android.bp b/Android.bp
index e56904c3722e..0b9166c56315 100644
--- a/Android.bp
+++ b/Android.bp
@@ -96,6 +96,14 @@ cc_defaults {
     test_per_src: true,
 }
 
+// binderfs test
+cc_test {
+    name: "kselftest_binderfs_tests",
+    relative_install_path: "linux-kselftest/filesystems/binderfs",
+    srcs: ["tools/testing/selftests/filesystems/binderfs/binderfs_test.c"],
+    defaults: ["kselftest_defaults"],
+}
+
 // breakpoints test
 cc_test {
     name: "kselftest_breakpoints_tests",
@@ -575,6 +583,7 @@ cc_test {
         "tools/testing/selftests/vm/mlock-random-test.c",
         "tools/testing/selftests/vm/mlock2-tests.c",
         "tools/testing/selftests/vm/on-fault-limit.c",
+        "tools/testing/selftests/vm/userfaultfd.c",
         "tools/testing/selftests/vm/thuge-gen.c",
         "tools/testing/selftests/vm/transhuge-stress.c",
     ],
diff --git a/OWNERS b/OWNERS
index 64bf926733d4..3ae489ae8e6c 100644
--- a/OWNERS
+++ b/OWNERS
@@ -1,2 +1,3 @@
 drosen@google.com
+smuckle@google.com
 willmcvicker@google.com
diff --git a/android/README.md b/android/README.md
index 2c94e05b4cdf..7a12a1d3aa21 100644
--- a/android/README.md
+++ b/android/README.md
@@ -334,9 +334,6 @@ kernel module is not required by Android.
 The hugepage-mmap, hugepage-shm, compaction, and map_hugetlb tests rely on
 hugetlbfs/hugetlb page support which is not required by Android.
 
-The userfaultfd test relies on the userfaultfd syscall which is not required by
-Android.
-
 ### watchdog/
 
 The watchdog test depends on CONFIG_WATCHDOG which is not required by Android.
diff --git a/android/kselftest_test_list.mk b/android/kselftest_test_list.mk
index 5445bd12d26f..07df6fc5d758 100644
--- a/android/kselftest_test_list.mk
+++ b/android/kselftest_test_list.mk
@@ -12,6 +12,7 @@ kselftest_modules := \
 
 # From Android.bp
 kselftest_modules += \
+  kselftest_binderfs_tests_binderfs_test \
   kselftest_breakpoints_tests_breakpoint_test \
   kselftest_breakpoints_tests_breakpoint_test_arm64 \
   kselftest_breakpoints_tests_step_after_suspend_test \
@@ -52,6 +53,7 @@ kselftest_modules += \
   kselftest_vdso_test \
   kselftest_vm_tests_mlock2-tests \
   kselftest_vm_tests_on-fault-limit \
+  kselftest_vm_tests_userfaultfd \
   kselftest_vm64_tests_virtual_address_range \
   kselftest_x86_ptrace_syscall \
   kselftest_x86_check_initial_reg_state \
diff --git a/android/patches/0019-userfaultfd.patch b/android/patches/0019-userfaultfd.patch
new file mode 100644
index 000000000000..67fe8edc9848
--- /dev/null
+++ b/android/patches/0019-userfaultfd.patch
@@ -0,0 +1,261 @@
+--- b/tools/testing/selftests/vm/userfaultfd.c
++++ b/tools/testing/selftests/vm/userfaultfd.c
+@@ -82,9 +82,11 @@
+ static char *huge_fd_off0;
+ static unsigned long long *count_verify;
+ static int uffd, uffd_flags, finished, *pipefd;
++static volatile bool ready_for_fork;
+ static char *area_src, *area_src_alias, *area_dst, *area_dst_alias;
+ static char *zeropage;
+ pthread_attr_t attr;
++pthread_key_t long_jmp_key;
+ 
+ /* pthread_mutex_t starts at page offset 0 */
+ #define area_mutex(___area, ___nr)					\
+@@ -139,8 +141,11 @@
+ 
+ static void anon_allocate_area(void **alloc_area)
+ {
+-	if (posix_memalign(alloc_area, page_size, nr_pages * page_size)) {
+-		fprintf(stderr, "out of memory\n");
++	// We can't use posix_memalign due to pointer-tagging used in bionic.
++	*alloc_area = mmap(NULL, nr_pages * page_size, PROT_READ | PROT_WRITE,
++			   MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
++	if (*alloc_area == MAP_FAILED) {
++		fprintf(stderr, "anon memory mmap failed\n");
+ 		*alloc_area = NULL;
+ 	}
+ }
+@@ -284,23 +289,11 @@
+ static void *locking_thread(void *arg)
+ {
+ 	unsigned long cpu = (unsigned long) arg;
+-	struct random_data rand;
+ 	unsigned long page_nr = *(&(page_nr)); /* uninitialized warning */
+-	int32_t rand_nr;
+ 	unsigned long long count;
+-	char randstate[64];
+-	unsigned int seed;
+ 	time_t start;
+ 
+-	if (bounces & BOUNCE_RANDOM) {
+-		seed = (unsigned int) time(NULL) - bounces;
+-		if (!(bounces & BOUNCE_RACINGFAULTS))
+-			seed += cpu;
+-		bzero(&rand, sizeof(rand));
+-		bzero(&randstate, sizeof(randstate));
+-		if (initstate_r(seed, randstate, sizeof(randstate), &rand))
+-			fprintf(stderr, "srandom_r error\n"), exit(1);
+-	} else {
++	if (!(bounces & BOUNCE_RANDOM)) {
+ 		page_nr = -bounces;
+ 		if (!(bounces & BOUNCE_RACINGFAULTS))
+ 			page_nr += cpu * nr_pages_per_cpu;
+@@ -308,13 +301,9 @@
+ 
+ 	while (!finished) {
+ 		if (bounces & BOUNCE_RANDOM) {
+-			if (random_r(&rand, &rand_nr))
+-				fprintf(stderr, "random_r 1 error\n"), exit(1);
+-			page_nr = rand_nr;
+-			if (sizeof(page_nr) > sizeof(rand_nr)) {
+-				if (random_r(&rand, &rand_nr))
+-					fprintf(stderr, "random_r 2 error\n"), exit(1);
+-				page_nr |= (((unsigned long) rand_nr) << 16) <<
++			page_nr = random();
++			if (sizeof(page_nr) > sizeof(uint32_t)) {
++				page_nr |= (((unsigned long) random()) << 16) <<
+ 					   16;
+ 			}
+ 		} else
+@@ -501,6 +490,9 @@
+ 	pollfd[1].fd = pipefd[cpu*2];
+ 	pollfd[1].events = POLLIN;
+ 
++	// Notify the main thread that it can now fork.
++	ready_for_fork = true;
++
+ 	for (;;) {
+ 		ret = poll(pollfd, 2, -1);
+ 		if (!ret)
+@@ -548,18 +540,31 @@
+ 
+ pthread_mutex_t uffd_read_mutex = PTHREAD_MUTEX_INITIALIZER;
+ 
++static void sigusr1_handler(int signum, siginfo_t *siginfo, void *ptr)
++{
++	jmp_buf *env;
++	env = pthread_getspecific(long_jmp_key);
++	longjmp(*env, 1);
++}
++
+ static void *uffd_read_thread(void *arg)
+ {
+ 	unsigned long *this_cpu_userfaults;
+ 	struct uffd_msg msg;
++	jmp_buf env;
++	int setjmp_ret;
++
++	pthread_setspecific(long_jmp_key, &env);
+ 
+ 	this_cpu_userfaults = (unsigned long *) arg;
+ 	*this_cpu_userfaults = 0;
+ 
+ 	pthread_mutex_unlock(&uffd_read_mutex);
+-	/* from here cancellation is ok */
+ 
+-	for (;;) {
++	// One first return setjmp return 0. On second (fake) return from
++	// longjmp() it returns the provided value, which will be 1 in our case.
++	setjmp_ret = setjmp(env);
++	while (!setjmp_ret) {
+ 		if (uffd_read_msg(uffd, &msg))
+ 			continue;
+ 		(*this_cpu_userfaults) += uffd_handle_page_fault(&msg);
+@@ -608,6 +613,7 @@
+ 				   background_thread, (void *)cpu))
+ 			return 1;
+ 	}
++
+ 	for (cpu = 0; cpu < nr_cpus; cpu++)
+ 		if (pthread_join(background_threads[cpu], NULL))
+ 			return 1;
+@@ -640,7 +646,7 @@
+ 			if (pthread_join(uffd_threads[cpu], &_userfaults[cpu]))
+ 				return 1;
+ 		} else {
+-			if (pthread_cancel(uffd_threads[cpu]))
++			if (pthread_kill(uffd_threads[cpu], SIGUSR1))
+ 				return 1;
+ 			if (pthread_join(uffd_threads[cpu], NULL))
+ 				return 1;
+@@ -654,7 +660,7 @@
+ {
+ 	struct uffdio_api uffdio_api;
+ 
+-	uffd = syscall(__NR_userfaultfd, O_CLOEXEC | O_NONBLOCK);
++	uffd = syscall(__NR_userfaultfd, O_CLOEXEC | O_NONBLOCK | UFFD_USER_MODE_ONLY);
+ 	if (uffd < 0) {
+ 		fprintf(stderr,
+ 			"userfaultfd syscall not available in this kernel\n");
+@@ -914,6 +920,10 @@
+ 	pid_t pid;
+ 	char c;
+ 
++	// All the syscalls below up to pthread_create will ensure that this
++	// write is completed before, the uffd_thread sets it to true.
++	ready_for_fork = false;
++
+ 	printf("testing events (fork, remap, remove): ");
+ 	fflush(stdout);
+ 
+@@ -942,6 +952,11 @@
+ 	if (pthread_create(&uffd_mon, &attr, uffd_poll_thread, NULL))
+ 		perror("uffd_poll_thread create"), exit(1);
+ 
++	// Wait for the poll_thread to start executing before forking. This is
++	// required to avoid a deadlock, which can happen if poll_thread doesn't
++	// start getting executed by the time fork is invoked.
++	while (!ready_for_fork);
++
+ 	pid = fork();
+ 	if (pid < 0)
+ 		perror("fork"), exit(1);
+@@ -974,6 +989,10 @@
+ 	pid_t pid;
+ 	char c;
+ 
++	// All the syscalls below up to pthread_create will ensure that this
++	// write is completed before, the uffd_thread sets it to true.
++	ready_for_fork = false;
++
+ 	printf("testing signal delivery: ");
+ 	fflush(stdout);
+ 
+@@ -1007,6 +1026,11 @@
+ 	if (pthread_create(&uffd_mon, &attr, uffd_poll_thread, NULL))
+ 		perror("uffd_poll_thread create"), exit(1);
+ 
++	// Wait for the poll_thread to start executing before forking. This is
++	// required to avoid a deadlock, which can happen if poll_thread doesn't
++	// start getting executed by the time fork is invoked.
++	while (!ready_for_fork);
++
+ 	pid = fork();
+ 	if (pid < 0)
+ 		perror("fork"), exit(1);
+@@ -1036,6 +1060,7 @@
+ 	char *tmp_area;
+ 	unsigned long nr;
+ 	struct uffdio_register uffdio_register;
++	struct sigaction act;
+ 	unsigned long cpu;
+ 	int err;
+ 	unsigned long userfaults[nr_cpus];
+@@ -1094,6 +1119,17 @@
+ 	pthread_attr_init(&attr);
+ 	pthread_attr_setstacksize(&attr, 16*1024*1024);
+ 
++	// For handling thread termination of read thread in the absense of
++	// pthread_cancel().
++	pthread_key_create(&long_jmp_key, NULL);
++	memset(&act, 0, sizeof(act));
++	act.sa_sigaction = sigusr1_handler;
++	act.sa_flags = SA_SIGINFO;
++	if (sigaction(SIGUSR1, &act, 0)) {
++		perror("sigaction");
++		return 1;
++	}
++
+ 	err = 0;
+ 	while (bounces--) {
+ 		unsigned long expected_ioctls;
+@@ -1215,6 +1251,8 @@
+ 		printf("\n");
+ 	}
+ 
++	pthread_key_delete(long_jmp_key);
++
+ 	if (err)
+ 		return err;
+ 
+@@ -1291,6 +1329,9 @@
+ 
+ int main(int argc, char **argv)
+ {
++	char randstate[64];
++	unsigned int seed;
++
+ 	if (argc < 4)
+ 		usage();
+ 
+@@ -1332,6 +1373,12 @@
+ 	}
+ 	printf("nr_pages: %lu, nr_pages_per_cpu: %lu\n",
+ 	       nr_pages, nr_pages_per_cpu);
++
++	seed = (unsigned int) time(NULL);
++	bzero(&randstate, sizeof(randstate));
++	if (!initstate(seed, randstate, sizeof(randstate)))
++		fprintf(stderr, "srandom_r error\n"), exit(1);
++
+ 	return userfaultfd_stress();
+ }
+ 
+--- a/tools/testing/selftests/vm/userfaultfd.c
++++ b/tools/testing/selftests/vm/userfaultfd.c
+@@ -662,8 +662,13 @@ static int userfaultfd_open(int features)
+ 
+ 	uffd = syscall(__NR_userfaultfd, O_CLOEXEC | O_NONBLOCK | UFFD_USER_MODE_ONLY);
+ 	if (uffd < 0) {
++		if (errno == ENOSYS) {
++			fprintf(stderr,
++				"userfaultfd syscall not available in this kernel\n");
++			exit(KSFT_PASS);
++		}
+ 		fprintf(stderr,
+-			"userfaultfd syscall not available in this kernel\n");
++			"userfaultfd syscall failed with errno: %d\n", errno);
+ 		return 1;
+ 	}
+ 	uffd_flags = fcntl(uffd, F_GETFD, NULL);
diff --git a/android/patches/0020-selftests-binderfs-fix-Wunused-label-error.patch b/android/patches/0020-selftests-binderfs-fix-Wunused-label-error.patch
new file mode 100644
index 000000000000..8b872f18e51f
--- /dev/null
+++ b/android/patches/0020-selftests-binderfs-fix-Wunused-label-error.patch
@@ -0,0 +1,31 @@
+From c4970a0b5461b65ce3b7d1c40a76672bd18d3ba5 Mon Sep 17 00:00:00 2001
+From: Carlos Llamas <cmllamas@google.com>
+Date: Mon, 26 Jul 2021 17:06:02 +0000
+Subject: [PATCH 1/2] selftests/binderfs: fix Wunused-label error
+
+Fix the following build issue:
+  tools/testing/selftests/filesystems/binderfs/binderfs_test.c:244:1:
+  error: unused label 'on_error' [-Werror,-Wunused-label]
+  on_error:
+  ^~~~~~~~~
+
+Signed-off-by: Carlos Llamas <cmllamas@google.com>
+---
+ tools/testing/selftests/filesystems/binderfs/binderfs_test.c | 1 -
+ 1 file changed, 1 deletion(-)
+
+diff --git a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
+index 8c2ed962e1c7..540ad4840495 100644
+--- a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
++++ b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
+@@ -241,7 +241,6 @@ static void __do_binderfs_test(void)
+ 	/* binder-control device removal failed as expected */
+ 	ksft_inc_xfail_cnt();
+ 
+-on_error:
+ 	ret = umount2("/dev/binderfs", MNT_DETACH);
+ 	keep ?: rmdir_protect_errno("/dev/binderfs");
+ 	if (ret < 0)
+-- 
+2.32.0.432.gabb21c7263-goog
+
diff --git a/android/patches/0021-selftests-binderfs-skip-unprivileged-test-when-USER_.patch b/android/patches/0021-selftests-binderfs-skip-unprivileged-test-when-USER_.patch
new file mode 100644
index 000000000000..8a5af5794c01
--- /dev/null
+++ b/android/patches/0021-selftests-binderfs-skip-unprivileged-test-when-USER_.patch
@@ -0,0 +1,36 @@
+From e51806a0b6918cefbd068cbb8100f410e7b01970 Mon Sep 17 00:00:00 2001
+From: Carlos Llamas <cmllamas@google.com>
+Date: Mon, 26 Jul 2021 17:12:42 +0000
+Subject: [PATCH 2/2] selftests/binderfs: skip unprivileged test when !USER_NS
+
+Binderfs unprivileged test requires CONFIG_USER_NS to be set. However,
+this option is not required by Android platforms. Lets skip the test
+as the underlying routine __do_binderfs_test() has already been checked
+by the binderfs_test_privileged() anyway.
+
+Signed-off-by: Carlos Llamas <cmllamas@google.com>
+---
+ .../testing/selftests/filesystems/binderfs/binderfs_test.c | 7 +++++++
+ 1 file changed, 7 insertions(+)
+
+diff --git a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
+index 540ad4840495..397bf265ba97 100644
+--- a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
++++ b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
+@@ -262,6 +262,13 @@ static void binderfs_test_privileged()
+ 
+ static void binderfs_test_unprivileged()
+ {
++	/* This test depends on USER_NS which is not required by Android */
++	if (access("/proc/self/ns/user", F_OK) != 0) {
++		ksft_test_result_skip("%s: user namespace not supported\n",
++				__func__);
++		return;
++	}
++
+ 	change_to_userns();
+ 	__do_binderfs_test();
+ }
+-- 
+2.32.0.432.gabb21c7263-goog
+
diff --git a/android/patches/0022-selftests-binderfs-skip-tests-for-unsupported-kernel.patch b/android/patches/0022-selftests-binderfs-skip-tests-for-unsupported-kernel.patch
new file mode 100644
index 000000000000..7b1e1a1c32bc
--- /dev/null
+++ b/android/patches/0022-selftests-binderfs-skip-tests-for-unsupported-kernel.patch
@@ -0,0 +1,65 @@
+From 57b74c1b318625057398ccbe479a157507488d8c Mon Sep 17 00:00:00 2001
+From: Carlos Llamas <cmllamas@google.com>
+Date: Tue, 14 Dec 2021 05:58:48 +0000
+Subject: [PATCH] selftests/binderfs: skip tests for unsupported kernels
+
+Binderfs was first introduced in Linux 5.0 release, so we need to skip
+these tests for older kernels. Also, we force a successful exit as the
+current framework doesn't handle KSFT_SKIP return codes.
+
+Signed-off-by: Carlos Llamas <cmllamas@google.com>
+---
+ .../filesystems/binderfs/binderfs_test.c      | 28 +++++++++++++++++++
+ 1 file changed, 28 insertions(+)
+
+diff --git a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
+index 8c2ed962e1c7..be5a8f446066 100644
+--- a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
++++ b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
+@@ -12,6 +12,7 @@
+ #include <sys/mount.h>
+ #include <sys/stat.h>
+ #include <sys/types.h>
++#include <sys/utsname.h>
+ #include <unistd.h>
+ #include <linux/android/binder.h>
+ #include <linux/android/binderfs.h>
+@@ -267,8 +268,35 @@ static void binderfs_test_unprivileged()
+ 	__do_binderfs_test();
+ }
+ 
++/* binderfs first introduced in Linux 5.0 */
++bool binderfs_supported(void)
++{
++	const int min_kernel_version = 5;
++	struct utsname utsname;
++	int ret, version;
++
++	ret = uname(&utsname);
++	if (ret)
++		ksft_exit_fail_msg("%s - Failed to get kernel version\n",
++				   strerror(errno));
++
++	ret = sscanf(utsname.release, "%d.", &version);
++	if (ret != 1)
++		ksft_exit_fail_msg("%s - Failed to parse uname: %s\n",
++				   strerror(errno), utsname.release);
++
++	return version >= min_kernel_version? true: false;
++}
++
++
+ int main(int argc, char *argv[])
+ {
++	/* Force success exit for older kernels */
++	if (!binderfs_supported()) {
++		ksft_print_msg("Skipping tests - binderfs not supported\n");
++		ksft_exit_pass();
++	}
++
+ 	binderfs_test_privileged();
+ 	binderfs_test_unprivileged();
+ 	ksft_exit_pass();
+-- 
+2.34.1.173.g76aa8bc2d0-goog
+
diff --git a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
index 8c2ed962e1c7..4465b4f6a81a 100644
--- a/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
+++ b/tools/testing/selftests/filesystems/binderfs/binderfs_test.c
@@ -12,6 +12,7 @@
 #include <sys/mount.h>
 #include <sys/stat.h>
 #include <sys/types.h>
+#include <sys/utsname.h>
 #include <unistd.h>
 #include <linux/android/binder.h>
 #include <linux/android/binderfs.h>
@@ -241,7 +242,6 @@ static void __do_binderfs_test(void)
 	/* binder-control device removal failed as expected */
 	ksft_inc_xfail_cnt();
 
-on_error:
 	ret = umount2("/dev/binderfs", MNT_DETACH);
 	keep ?: rmdir_protect_errno("/dev/binderfs");
 	if (ret < 0)
@@ -263,12 +263,46 @@ static void binderfs_test_privileged()
 
 static void binderfs_test_unprivileged()
 {
+	/* This test depends on USER_NS which is not required by Android */
+	if (access("/proc/self/ns/user", F_OK) != 0) {
+		ksft_test_result_skip("%s: user namespace not supported\n",
+				__func__);
+		return;
+	}
+
 	change_to_userns();
 	__do_binderfs_test();
 }
 
+/* binderfs first introduced in Linux 5.0 */
+bool binderfs_supported(void)
+{
+	const int min_kernel_version = 5;
+	struct utsname utsname;
+	int ret, version;
+
+	ret = uname(&utsname);
+	if (ret)
+		ksft_exit_fail_msg("%s - Failed to get kernel version\n",
+				   strerror(errno));
+
+	ret = sscanf(utsname.release, "%d.", &version);
+	if (ret != 1)
+		ksft_exit_fail_msg("%s - Failed to parse uname: %s\n",
+				   strerror(errno), utsname.release);
+
+	return version >= min_kernel_version? true: false;
+}
+
+
 int main(int argc, char *argv[])
 {
+	/* Force success exit for older kernels */
+	if (!binderfs_supported()) {
+		ksft_print_msg("Skipping tests - binderfs not supported\n");
+		ksft_exit_pass();
+	}
+
 	binderfs_test_privileged();
 	binderfs_test_unprivileged();
 	ksft_exit_pass();
diff --git a/tools/testing/selftests/vm/userfaultfd.c b/tools/testing/selftests/vm/userfaultfd.c
index d3362777a425..ffc78f3e4444 100644
--- a/tools/testing/selftests/vm/userfaultfd.c
+++ b/tools/testing/selftests/vm/userfaultfd.c
@@ -82,9 +82,11 @@ static int huge_fd;
 static char *huge_fd_off0;
 static unsigned long long *count_verify;
 static int uffd, uffd_flags, finished, *pipefd;
+static volatile bool ready_for_fork;
 static char *area_src, *area_src_alias, *area_dst, *area_dst_alias;
 static char *zeropage;
 pthread_attr_t attr;
+pthread_key_t long_jmp_key;
 
 /* pthread_mutex_t starts at page offset 0 */
 #define area_mutex(___area, ___nr)					\
@@ -139,8 +141,11 @@ static int anon_release_pages(char *rel_area)
 
 static void anon_allocate_area(void **alloc_area)
 {
-	if (posix_memalign(alloc_area, page_size, nr_pages * page_size)) {
-		fprintf(stderr, "out of memory\n");
+	// We can't use posix_memalign due to pointer-tagging used in bionic.
+	*alloc_area = mmap(NULL, nr_pages * page_size, PROT_READ | PROT_WRITE,
+			   MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
+	if (*alloc_area == MAP_FAILED) {
+		fprintf(stderr, "anon memory mmap failed\n");
 		*alloc_area = NULL;
 	}
 }
@@ -284,23 +289,11 @@ static int my_bcmp(char *str1, char *str2, size_t n)
 static void *locking_thread(void *arg)
 {
 	unsigned long cpu = (unsigned long) arg;
-	struct random_data rand;
 	unsigned long page_nr = *(&(page_nr)); /* uninitialized warning */
-	int32_t rand_nr;
 	unsigned long long count;
-	char randstate[64];
-	unsigned int seed;
 	time_t start;
 
-	if (bounces & BOUNCE_RANDOM) {
-		seed = (unsigned int) time(NULL) - bounces;
-		if (!(bounces & BOUNCE_RACINGFAULTS))
-			seed += cpu;
-		bzero(&rand, sizeof(rand));
-		bzero(&randstate, sizeof(randstate));
-		if (initstate_r(seed, randstate, sizeof(randstate), &rand))
-			fprintf(stderr, "srandom_r error\n"), exit(1);
-	} else {
+	if (!(bounces & BOUNCE_RANDOM)) {
 		page_nr = -bounces;
 		if (!(bounces & BOUNCE_RACINGFAULTS))
 			page_nr += cpu * nr_pages_per_cpu;
@@ -308,13 +301,9 @@ static void *locking_thread(void *arg)
 
 	while (!finished) {
 		if (bounces & BOUNCE_RANDOM) {
-			if (random_r(&rand, &rand_nr))
-				fprintf(stderr, "random_r 1 error\n"), exit(1);
-			page_nr = rand_nr;
-			if (sizeof(page_nr) > sizeof(rand_nr)) {
-				if (random_r(&rand, &rand_nr))
-					fprintf(stderr, "random_r 2 error\n"), exit(1);
-				page_nr |= (((unsigned long) rand_nr) << 16) <<
+			page_nr = random();
+			if (sizeof(page_nr) > sizeof(uint32_t)) {
+				page_nr |= (((unsigned long) random()) << 16) <<
 					   16;
 			}
 		} else
@@ -501,6 +490,9 @@ static void *uffd_poll_thread(void *arg)
 	pollfd[1].fd = pipefd[cpu*2];
 	pollfd[1].events = POLLIN;
 
+	// Notify the main thread that it can now fork.
+	ready_for_fork = true;
+
 	for (;;) {
 		ret = poll(pollfd, 2, -1);
 		if (!ret)
@@ -548,18 +540,31 @@ static void *uffd_poll_thread(void *arg)
 
 pthread_mutex_t uffd_read_mutex = PTHREAD_MUTEX_INITIALIZER;
 
+static void sigusr1_handler(int signum, siginfo_t *siginfo, void *ptr)
+{
+	jmp_buf *env;
+	env = pthread_getspecific(long_jmp_key);
+	longjmp(*env, 1);
+}
+
 static void *uffd_read_thread(void *arg)
 {
 	unsigned long *this_cpu_userfaults;
 	struct uffd_msg msg;
+	jmp_buf env;
+	int setjmp_ret;
+
+	pthread_setspecific(long_jmp_key, &env);
 
 	this_cpu_userfaults = (unsigned long *) arg;
 	*this_cpu_userfaults = 0;
 
 	pthread_mutex_unlock(&uffd_read_mutex);
-	/* from here cancellation is ok */
 
-	for (;;) {
+	// One first return setjmp return 0. On second (fake) return from
+	// longjmp() it returns the provided value, which will be 1 in our case.
+	setjmp_ret = setjmp(env);
+	while (!setjmp_ret) {
 		if (uffd_read_msg(uffd, &msg))
 			continue;
 		(*this_cpu_userfaults) += uffd_handle_page_fault(&msg);
@@ -608,6 +613,7 @@ static int stress(unsigned long *userfaults)
 				   background_thread, (void *)cpu))
 			return 1;
 	}
+
 	for (cpu = 0; cpu < nr_cpus; cpu++)
 		if (pthread_join(background_threads[cpu], NULL))
 			return 1;
@@ -640,7 +646,7 @@ static int stress(unsigned long *userfaults)
 			if (pthread_join(uffd_threads[cpu], &_userfaults[cpu]))
 				return 1;
 		} else {
-			if (pthread_cancel(uffd_threads[cpu]))
+			if (pthread_kill(uffd_threads[cpu], SIGUSR1))
 				return 1;
 			if (pthread_join(uffd_threads[cpu], NULL))
 				return 1;
@@ -654,10 +660,15 @@ static int userfaultfd_open(int features)
 {
 	struct uffdio_api uffdio_api;
 
-	uffd = syscall(__NR_userfaultfd, O_CLOEXEC | O_NONBLOCK);
+	uffd = syscall(__NR_userfaultfd, O_CLOEXEC | O_NONBLOCK | UFFD_USER_MODE_ONLY);
 	if (uffd < 0) {
+		if (errno == ENOSYS) {
+			fprintf(stderr,
+				"userfaultfd syscall not available in this kernel\n");
+			exit(KSFT_PASS);
+		}
 		fprintf(stderr,
-			"userfaultfd syscall not available in this kernel\n");
+			"userfaultfd syscall failed with errno: %d\n", errno);
 		return 1;
 	}
 	uffd_flags = fcntl(uffd, F_GETFD, NULL);
@@ -914,6 +925,10 @@ static int userfaultfd_events_test(void)
 	pid_t pid;
 	char c;
 
+	// All the syscalls below up to pthread_create will ensure that this
+	// write is completed before, the uffd_thread sets it to true.
+	ready_for_fork = false;
+
 	printf("testing events (fork, remap, remove): ");
 	fflush(stdout);
 
@@ -942,6 +957,11 @@ static int userfaultfd_events_test(void)
 	if (pthread_create(&uffd_mon, &attr, uffd_poll_thread, NULL))
 		perror("uffd_poll_thread create"), exit(1);
 
+	// Wait for the poll_thread to start executing before forking. This is
+	// required to avoid a deadlock, which can happen if poll_thread doesn't
+	// start getting executed by the time fork is invoked.
+	while (!ready_for_fork);
+
 	pid = fork();
 	if (pid < 0)
 		perror("fork"), exit(1);
@@ -974,6 +994,10 @@ static int userfaultfd_sig_test(void)
 	pid_t pid;
 	char c;
 
+	// All the syscalls below up to pthread_create will ensure that this
+	// write is completed before, the uffd_thread sets it to true.
+	ready_for_fork = false;
+
 	printf("testing signal delivery: ");
 	fflush(stdout);
 
@@ -1007,6 +1031,11 @@ static int userfaultfd_sig_test(void)
 	if (pthread_create(&uffd_mon, &attr, uffd_poll_thread, NULL))
 		perror("uffd_poll_thread create"), exit(1);
 
+	// Wait for the poll_thread to start executing before forking. This is
+	// required to avoid a deadlock, which can happen if poll_thread doesn't
+	// start getting executed by the time fork is invoked.
+	while (!ready_for_fork);
+
 	pid = fork();
 	if (pid < 0)
 		perror("fork"), exit(1);
@@ -1036,6 +1065,7 @@ static int userfaultfd_stress(void)
 	char *tmp_area;
 	unsigned long nr;
 	struct uffdio_register uffdio_register;
+	struct sigaction act;
 	unsigned long cpu;
 	int err;
 	unsigned long userfaults[nr_cpus];
@@ -1094,6 +1124,17 @@ static int userfaultfd_stress(void)
 	pthread_attr_init(&attr);
 	pthread_attr_setstacksize(&attr, 16*1024*1024);
 
+	// For handling thread termination of read thread in the absense of
+	// pthread_cancel().
+	pthread_key_create(&long_jmp_key, NULL);
+	memset(&act, 0, sizeof(act));
+	act.sa_sigaction = sigusr1_handler;
+	act.sa_flags = SA_SIGINFO;
+	if (sigaction(SIGUSR1, &act, 0)) {
+		perror("sigaction");
+		return 1;
+	}
+
 	err = 0;
 	while (bounces--) {
 		unsigned long expected_ioctls;
@@ -1215,6 +1256,8 @@ static int userfaultfd_stress(void)
 		printf("\n");
 	}
 
+	pthread_key_delete(long_jmp_key);
+
 	if (err)
 		return err;
 
@@ -1291,6 +1334,9 @@ static void sigalrm(int sig)
 
 int main(int argc, char **argv)
 {
+	char randstate[64];
+	unsigned int seed;
+
 	if (argc < 4)
 		usage();
 
@@ -1332,6 +1378,12 @@ int main(int argc, char **argv)
 	}
 	printf("nr_pages: %lu, nr_pages_per_cpu: %lu\n",
 	       nr_pages, nr_pages_per_cpu);
+
+	seed = (unsigned int) time(NULL);
+	bzero(&randstate, sizeof(randstate));
+	if (!initstate(seed, randstate, sizeof(randstate)))
+		fprintf(stderr, "srandom_r error\n"), exit(1);
+
 	return userfaultfd_stress();
 }