@@ -41,6 +41,7 @@ TEST_GEN_FILES += map_fixed_noreplace
TEST_GEN_FILES += map_hugetlb
TEST_GEN_FILES += map_populate
TEST_GEN_FILES += memfd_secret
+TEST_GEN_FILES += migration
TEST_GEN_FILES += mlock-random-test
TEST_GEN_FILES += mlock2-tests
TEST_GEN_FILES += mremap_dontunmap
@@ -149,6 +150,8 @@ $(OUTPUT)/hmm-tests: LDLIBS += $(HMM_EXTRA_LIBS)
$(OUTPUT)/ksm_tests: LDLIBS += -lnuma
+$(OUTPUT)/migration: LDLIBS += -lnuma
+
local_config.mk local_config.h: check_config.sh
/bin/sh ./check_config.sh $(CC)
new file mode 100644
@@ -0,0 +1,193 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * The main purpose of the tests here is to exercise the migration entry code
+ * paths in the kernel.
+ */
+
+#include "../kselftest_harness.h"
+#include <strings.h>
+#include <pthread.h>
+#include <numa.h>
+#include <numaif.h>
+#include <sys/mman.h>
+#include <sys/types.h>
+#include <signal.h>
+#include <time.h>
+
+#define TWOMEG (2<<20)
+#define RUNTIME (60)
+
+#define ALIGN(x, a) (((x) + (a - 1)) & (~((a) - 1)))
+
+FIXTURE(migration)
+{
+ pthread_t *threads;
+ pid_t *pids;
+ int nthreads;
+ int n1;
+ int n2;
+};
+
+FIXTURE_SETUP(migration)
+{
+ int n;
+
+ ASSERT_EQ(numa_available(), 0);
+ self->nthreads = numa_num_task_cpus() - 1;
+ self->n1 = -1;
+ self->n2 = -1;
+
+ for (n = 0; n < numa_max_possible_node(); n++)
+ if (numa_bitmask_isbitset(numa_all_nodes_ptr, n)) {
+ if (self->n1 == -1) {
+ self->n1 = n;
+ } else {
+ self->n2 = n;
+ break;
+ }
+ }
+
+ self->threads = malloc(self->nthreads * sizeof(*self->threads));
+ ASSERT_NE(self->threads, NULL);
+ self->pids = malloc(self->nthreads * sizeof(*self->pids));
+ ASSERT_NE(self->pids, NULL);
+};
+
+FIXTURE_TEARDOWN(migration)
+{
+ free(self->threads);
+ free(self->pids);
+}
+
+int migrate(uint64_t *ptr, int n1, int n2)
+{
+ int ret, tmp;
+ int status = 0;
+ struct timespec ts1, ts2;
+
+ if (clock_gettime(CLOCK_MONOTONIC, &ts1))
+ return -1;
+
+ while (1) {
+ if (clock_gettime(CLOCK_MONOTONIC, &ts2))
+ return -1;
+
+ if (ts2.tv_sec - ts1.tv_sec >= RUNTIME)
+ return 0;
+
+ ret = move_pages(0, 1, (void **) &ptr, &n2, &status,
+ MPOL_MF_MOVE_ALL);
+ if (ret) {
+ if (ret > 0)
+ printf("Didn't migrate %d pages\n", ret);
+ else
+ perror("Couldn't migrate pages");
+ return -2;
+ }
+
+ tmp = n2;
+ n2 = n1;
+ n1 = tmp;
+ }
+
+ return 0;
+}
+
+void *access_mem(void *ptr)
+{
+ uint64_t y = 0;
+ volatile uint64_t *x = ptr;
+
+ while (1) {
+ pthread_testcancel();
+ y += *x;
+ }
+
+ return NULL;
+}
+
+/*
+ * Basic migration entry testing. One thread will move pages back and forth
+ * between nodes whilst other threads try and access them triggering the
+ * migration entry wait paths in the kernel.
+ */
+TEST_F_TIMEOUT(migration, private_anon, 2*RUNTIME)
+{
+ uint64_t *ptr;
+ int i;
+
+ if (self->nthreads < 2 || self->n1 < 0 || self->n2 < 0)
+ SKIP(return, "Not enough threads or NUMA nodes available");
+
+ ptr = mmap(NULL, TWOMEG, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ ASSERT_NE(ptr, MAP_FAILED);
+
+ memset(ptr, 0xde, TWOMEG);
+ for (i = 0; i < self->nthreads - 1; i++)
+ if (pthread_create(&self->threads[i], NULL, access_mem, ptr))
+ perror("Couldn't create thread");
+
+ ASSERT_EQ(migrate(ptr, self->n1, self->n2), 0);
+ for (i = 0; i < self->nthreads - 1; i++)
+ ASSERT_EQ(pthread_cancel(self->threads[i]), 0);
+}
+
+/*
+ * Same as the previous test but with shared memory.
+ */
+TEST_F_TIMEOUT(migration, shared_anon, 2*RUNTIME)
+{
+ pid_t pid;
+ uint64_t *ptr;
+ int i;
+
+ if (self->nthreads < 2 || self->n1 < 0 || self->n2 < 0)
+ SKIP(return, "Not enough threads or NUMA nodes available");
+
+ ptr = mmap(NULL, TWOMEG, PROT_READ | PROT_WRITE,
+ MAP_SHARED | MAP_ANONYMOUS, -1, 0);
+ ASSERT_NE(ptr, MAP_FAILED);
+
+ memset(ptr, 0xde, TWOMEG);
+ for (i = 0; i < self->nthreads - 1; i++) {
+ pid = fork();
+ if (!pid)
+ access_mem(ptr);
+ else
+ self->pids[i] = pid;
+ }
+
+ ASSERT_EQ(migrate(ptr, self->n1, self->n2), 0);
+ for (i = 0; i < self->nthreads - 1; i++)
+ ASSERT_EQ(kill(self->pids[i], SIGTERM), 0);
+}
+
+/*
+ * Tests the pmd migration entry paths.
+ */
+TEST_F_TIMEOUT(migration, private_anon_thp, 2*RUNTIME)
+{
+ uint64_t *ptr;
+ int i;
+
+ if (self->nthreads < 2 || self->n1 < 0 || self->n2 < 0)
+ SKIP(return, "Not enough threads or NUMA nodes available");
+
+ ptr = mmap(NULL, 2*TWOMEG, PROT_READ | PROT_WRITE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+ ASSERT_NE(ptr, MAP_FAILED);
+
+ ptr = (uint64_t *) ALIGN((uintptr_t) ptr, TWOMEG);
+ ASSERT_EQ(madvise(ptr, TWOMEG, MADV_HUGEPAGE), 0);
+ memset(ptr, 0xde, TWOMEG);
+ for (i = 0; i < self->nthreads - 1; i++)
+ if (pthread_create(&self->threads[i], NULL, access_mem, ptr))
+ perror("Couldn't create thread");
+
+ ASSERT_EQ(migrate(ptr, self->n1, self->n2), 0);
+ for (i = 0; i < self->nthreads - 1; i++)
+ ASSERT_EQ(pthread_cancel(self->threads[i]), 0);
+}
+
+TEST_HARNESS_MAIN
Add some basic migration tests and in particular tests that will stress both the pte and pmd migration entry wait paths. Signed-off-by: Alistair Popple <apopple@nvidia.com> --- This is a clean-up of an ad-hoc test I had written to test and understand the migration entry code paths. I think it would also be useful to have upstream so these code paths could be more regularly and easily exerciesed. For example these tests directly found an issue with linux-next f886cdb76920 ("mm: pvmw: add support for walking devmap pages"). tools/testing/selftests/vm/Makefile | 3 + tools/testing/selftests/vm/migration.c | 193 +++++++++++++++++++++++++ 2 files changed, 196 insertions(+) create mode 100644 tools/testing/selftests/vm/migration.c