Re: [PATCH] selftests/mm: refactore split_huge_page_test with kselftest_harness

On 9/5/25 8:54 AM, Wei Yang wrote:

Refactor split_huge_page_test with kselftest_harness, since there is a magic counting about valid tests.

The idea is simple:

For standalone test, put it into TEST().

For tests iterating order/offset, define fixture and variant with order and offset. And skip it if order/offset is not valid.

No functional change is expected.

Signed-off-by: Wei Yang richard.weiyang@gmail.com Cc: David Hildenbrand david@redhat.com Cc: Zi Yan ziy@nvidia.com Cc: Donet Tom donettom@linux.ibm.com Cc: Ryan Roberts ryan.roberts@arm.com Cc: Baolin Wang baolin.wang@linux.alibaba.com

---

.../selftests/mm/split_huge_page_test.c | 308 ++++++++++++++---- 1 file changed, 238 insertions(+), 70 deletions(-)

diff --git a/tools/testing/selftests/mm/split_huge_page_test.c b/tools/testing/selftests/mm/split_huge_page_test.c index 7731191cc8e9..2376c3530b26 100644 --- a/tools/testing/selftests/mm/split_huge_page_test.c +++ b/tools/testing/selftests/mm/split_huge_page_test.c @@ -20,14 +20,16 @@ #include <stdbool.h> #include <time.h> #include "vm_util.h" -#include "../kselftest.h" +#include "../kselftest_harness.h" uint64_t pagesize; -unsigned int pageshift; uint64_t pmd_pagesize; unsigned int pmd_order; int *expected_orders; +char *optional_xfs_path; +char fs_loc_template[] = "/tmp/thp_fs_XXXXXX";

• #define SPLIT_DEBUGFS "/sys/kernel/debug/split_huge_pages" #define SMAP_PATH "/proc/self/smaps" #define INPUT_MAX 80

@@ -41,6 +43,20 @@ const char *kpageflags_proc = "/proc/kpageflags"; int pagemap_fd; int kpageflags_fd; +#define ADD_VARIANT_ORDER(fixture_name, ord) \

• FIXTURE_VARIANT_ADD(fixture_name, order_##ord) \
• { \

• .order = ord,					\
  

• }

+#define ADD_VARIANT_ORDER_OFFSET(fixture_name, ord, stp) \

• FIXTURE_VARIANT_ADD(fixture_name, order_##ord##_stp_##stp) \
• { \

• .order = ord,						\
  

• .step = stp,						\
  

• }
• static bool is_backed_by_folio(char *vaddr, int order, int pagemap_fd, int kpageflags_fd) {

@@ -255,6 +271,26 @@ static int check_after_split_folio_orders(char *vaddr_start, size_t len, return status; } +void prepare_proc_fd(void) +{

• pagemap_fd = open(pagemap_proc, O_RDONLY);
• if (pagemap_fd == -1)

• ksft_exit_fail_msg("read pagemap: %s\n", strerror(errno));
  

• kpageflags_fd = open(kpageflags_proc, O_RDONLY);
• if (kpageflags_fd == -1)

• ksft_exit_fail_msg("read kpageflags: %s\n", strerror(errno));
  

+}

+void cleanup_proc_fd(void) +{

• if (pagemap_fd != -1)

• close(pagemap_fd);
  

• if (kpageflags_fd != -1)

• close(kpageflags_fd);
  

+}

• static void write_file(const char *path, const char *buf, size_t buflen) { int fd;

@@ -292,10 +328,8 @@ static char *allocate_zero_filled_hugepage(size_t len) size_t i; result = memalign(pmd_pagesize, len);

• if (!result) {

• printf("Fail to allocate memory\n");
  

• exit(EXIT_FAILURE);
  

• }

• if (!result)

• ksft_exit_fail_msg("Fail to allocate memory: %s\n", strerror(errno));
  

madvise(result, len, MADV_HUGEPAGE); @@ -334,16 +368,19 @@ static void verify_rss_anon_split_huge_page_all_zeroes(char *one_page, int nr_hp rss_anon_before, rss_anon_after); } -static void split_pmd_zero_pages(void) +TEST(split_pmd_zero_pages) { char *one_page; int nr_hpages = 4; size_t len = nr_hpages * pmd_pagesize;

• prepare_proc_fd();
• one_page = allocate_zero_filled_hugepage(len); verify_rss_anon_split_huge_page_all_zeroes(one_page, nr_hpages, len);

• ksft_test_result_pass("Split zero filled huge pages successful\n"); free(one_page);

• cleanup_proc_fd(); }

static void split_pmd_thp_to_order(int order) @@ -383,11 +420,10 @@ static void split_pmd_thp_to_order(int order) if (!check_huge_anon(one_page, 0, pmd_pagesize)) ksft_exit_fail_msg("Still AnonHugePages not split\n");

• ksft_test_result_pass("Split huge pages to order %d successful\n", order); free(one_page); }

-static void split_pte_mapped_thp(void) +TEST(split_pte_mapped_thp) { const size_t nr_thps = 4; const size_t thp_area_size = nr_thps * pmd_pagesize; @@ -395,6 +431,8 @@ static void split_pte_mapped_thp(void) char *thp_area, *tmp, *page_area = MAP_FAILED; size_t i;

• prepare_proc_fd();
• thp_area = mmap((void *)(1UL << 30), thp_area_size, PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0); if (thp_area == MAP_FAILED) {

@@ -470,14 +508,43 @@ static void split_pte_mapped_thp(void) ksft_test_result_fail("THP %zu not split\n", i); }

• ksft_test_result_pass("Split PTE-mapped huge pages successful\n"); out: munmap(thp_area, thp_area_size); if (page_area != MAP_FAILED) munmap(page_area, page_area_size);

• cleanup_proc_fd();

+}

+FIXTURE(split_file_backed_thp) { +};

+FIXTURE_VARIANT(split_file_backed_thp) {

• int order;

+};

+FIXTURE_SETUP(split_file_backed_thp) +{

• if (variant->order >= pmd_order)

• SKIP(return, "order %d is not supported", variant->order);
  

• prepare_proc_fd();

+}

+FIXTURE_TEARDOWN(split_file_backed_thp) +{

• cleanup_proc_fd(); }

-static void split_file_backed_thp(int order) +ADD_VARIANT_ORDER(split_file_backed_thp, 0); +ADD_VARIANT_ORDER(split_file_backed_thp, 1); +ADD_VARIANT_ORDER(split_file_backed_thp, 2); +ADD_VARIANT_ORDER(split_file_backed_thp, 3); +ADD_VARIANT_ORDER(split_file_backed_thp, 4); +ADD_VARIANT_ORDER(split_file_backed_thp, 5); +ADD_VARIANT_ORDER(split_file_backed_thp, 6); +ADD_VARIANT_ORDER(split_file_backed_thp, 7); +ADD_VARIANT_ORDER(split_file_backed_thp, 8);

+TEST_F(split_file_backed_thp, orders) { int status; int fd; @@ -487,7 +554,7 @@ static void split_file_backed_thp(int order) ssize_t num_written, num_read; char *file_buf1, *file_buf2; uint64_t pgoff_start = 0, pgoff_end = 1024;

• int i;

• int i, order = variant->order;

ksft_print_msg("Please enable pr_debug in split_huge_pages_in_file() for more info.\n"); @@ -567,7 +634,6 @@ static void split_file_backed_thp(int order) ksft_exit_fail_msg("cannot remove tmp dir: %s\n", strerror(errno)); ksft_print_msg("Please check dmesg for more information\n");

• ksft_test_result_pass("File-backed THP split to order %d test done\n", order); return;

close_file: @@ -579,6 +645,45 @@ static void split_file_backed_thp(int order) ksft_exit_fail_msg("Error occurred\n"); } +FIXTURE(split_pmd_thp_to_order) { +};

+FIXTURE_VARIANT(split_pmd_thp_to_order) {

• int order;

+};

+FIXTURE_SETUP(split_pmd_thp_to_order) +{

• if (variant->order >= pmd_order)

• SKIP(return, "order %d is not supported", variant->order);
  

On a 64 KB page size system, the page order is 5, so the remaining tests are getting skipped. On such systems, these tests are effectively invalid. Instead of skipping them, would it be better to simply not account those tests?

# PASSED: 64 / 64 tests passed.  # 30 skipped test(s) detected. Consider enabling relevant config options to improve coverage.  # Totals: pass:34 fail:0 xfail:0 xpass:0 skip:30 error:0

Here, the 30 skipped tests are invalid, and the total valid tests are 34. Would it make sense to exclude the 30 invalid tests entirely for 64KB page size systems?

• expected_orders = (int *)malloc(sizeof(int) * (pmd_order + 1));
• if (!expected_orders)

• ksft_exit_fail_msg("Fail to allocate memory: %s\n", strerror(errno));
  

• prepare_proc_fd();

+}

+FIXTURE_TEARDOWN(split_pmd_thp_to_order) +{

• free(expected_orders);
• cleanup_proc_fd();

+}

+ADD_VARIANT_ORDER(split_pmd_thp_to_order, 0); +ADD_VARIANT_ORDER(split_pmd_thp_to_order, 2); +ADD_VARIANT_ORDER(split_pmd_thp_to_order, 3); +ADD_VARIANT_ORDER(split_pmd_thp_to_order, 4); +ADD_VARIANT_ORDER(split_pmd_thp_to_order, 5); +ADD_VARIANT_ORDER(split_pmd_thp_to_order, 6); +ADD_VARIANT_ORDER(split_pmd_thp_to_order, 7); +ADD_VARIANT_ORDER(split_pmd_thp_to_order, 8);

+TEST_F(split_pmd_thp_to_order, order) +{

• split_pmd_thp_to_order(variant->order);

+}

• static bool prepare_thp_fs(const char *xfs_path, char *thp_fs_template, const char **thp_fs_loc) {

@@ -757,81 +862,144 @@ static void split_thp_in_pagecache_to_order_at(size_t fd_size, } } -int main(int argc, char **argv) -{

• int i;
• size_t fd_size;
• char *optional_xfs_path = NULL;
• char fs_loc_template[] = "/tmp/thp_fs_XXXXXX";

+FIXTURE(split_thp_in_pagecache_to_order) {

• bool created_tmp; const char *fs_loc;

+};

+FIXTURE_VARIANT(split_thp_in_pagecache_to_order) {

• int order;

+};

+FIXTURE_SETUP(split_thp_in_pagecache_to_order) +{

• /* limit order to less than pmd_order */
• if (variant->order >= pmd_order)

• SKIP(return, "order %d is not supported", variant->order);
  

• expected_orders = (int *)malloc(sizeof(int) * (pmd_order + 1));
• if (!expected_orders)

• ksft_exit_fail_msg("Fail to allocate memory: %s\n", strerror(errno));
  

• self->created_tmp = prepare_thp_fs(optional_xfs_path, fs_loc_template,

• 			&self->fs_loc);
  

• prepare_proc_fd();

+}

+FIXTURE_TEARDOWN(split_thp_in_pagecache_to_order) +{

• cleanup_proc_fd();
• cleanup_thp_fs(self->fs_loc, self->created_tmp);
• free(expected_orders);

+}

+ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 8); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 7); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 6); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 5); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 4); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 3); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 2); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 1); +ADD_VARIANT_ORDER(split_thp_in_pagecache_to_order, 0);

+TEST_F(split_thp_in_pagecache_to_order, order) +{

• size_t fd_size = 2 * pmd_pagesize;
• split_thp_in_pagecache_to_order_at(fd_size, self->fs_loc, variant->order, -1);

+}

+FIXTURE(split_thp_in_pagecache_to_order_offset) { bool created_tmp;

• const char *fs_loc; int offset;

• unsigned int nr_pages;
• unsigned int tests;

+};

• ksft_print_header();

+FIXTURE_VARIANT(split_thp_in_pagecache_to_order_offset) {

• int order;
• int step;

+};

• if (geteuid() != 0) {

• ksft_print_msg("Please run the benchmark as root\n");
  

• ksft_finished();
  

• }
• if (argc > 1)

• optional_xfs_path = argv[1];
  

+FIXTURE_SETUP(split_thp_in_pagecache_to_order_offset) +{

• int nr_pages = pmd_pagesize / pagesize;
• int offset = variant->step * MAX(nr_pages / 4, 1 << variant->order);

• pagesize = getpagesize();
• pageshift = ffs(pagesize) - 1;
• pmd_pagesize = read_pmd_pagesize();
• if (!pmd_pagesize)

• ksft_exit_fail_msg("Reading PMD pagesize failed\n");
  

• /* limit order to less than pmd_order */
• if (variant->order >= pmd_order)

• SKIP(return, "order %d is not supported", variant->order);
  

• nr_pages = pmd_pagesize / pagesize;
• pmd_order = sz2ord(pmd_pagesize, pagesize);

• if (offset < nr_pages)

• self->offset = offset;
  

• else

• SKIP(return, "offset out of thp range");
  

expected_orders = (int *)malloc(sizeof(int) * (pmd_order + 1)); if (!expected_orders) ksft_exit_fail_msg("Fail to allocate memory: %s\n", strerror(errno));

• tests = 2 + (pmd_order - 1) + (2 * pmd_order) + (pmd_order - 1) * 4 + 2;
• ksft_set_plan(tests);
• pagemap_fd = open(pagemap_proc, O_RDONLY);
• if (pagemap_fd == -1)

• ksft_exit_fail_msg("read pagemap: %s\n", strerror(errno));
  

• kpageflags_fd = open(kpageflags_proc, O_RDONLY);
• if (kpageflags_fd == -1)

• ksft_exit_fail_msg("read kpageflags: %s\n", strerror(errno));
  

• self->created_tmp = prepare_thp_fs(optional_xfs_path, fs_loc_template,

• 			&self->fs_loc);
  

• prepare_proc_fd();

+}

• fd_size = 2 * pmd_pagesize;

+FIXTURE_TEARDOWN(split_thp_in_pagecache_to_order_offset) +{

• cleanup_proc_fd();
• cleanup_thp_fs(self->fs_loc, self->created_tmp);
• free(expected_orders);

+}

• split_pmd_zero_pages();

+#define SPLIT_IN_PAGECACHE_ORDER_OFFSETS(ord) \

• ADD_VARIANT_ORDER_OFFSET(split_thp_in_pagecache_to_order_offset, ord, 0);\
• ADD_VARIANT_ORDER_OFFSET(split_thp_in_pagecache_to_order_offset, ord, 1);\
• ADD_VARIANT_ORDER_OFFSET(split_thp_in_pagecache_to_order_offset, ord, 2);\
• ADD_VARIANT_ORDER_OFFSET(split_thp_in_pagecache_to_order_offset, ord, 3)\

+SPLIT_IN_PAGECACHE_ORDER_OFFSETS(0); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(1); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(2); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(3); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(4); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(5); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(6); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(7); +SPLIT_IN_PAGECACHE_ORDER_OFFSETS(8);

+TEST_F(split_thp_in_pagecache_to_order_offset, order_offset) +{

• size_t fd_size = 2 * pmd_pagesize;

• for (i = 0; i < pmd_order; i++)

• if (i != 1)
  

• 	split_pmd_thp_to_order(i);
  

• split_thp_in_pagecache_to_order_at(fd_size, self->fs_loc,

• 	variant->order, self->offset);
  

+}

• split_pte_mapped_thp();
• for (i = 0; i < pmd_order; i++)

• split_file_backed_thp(i);
  

+int main(int argc, char **argv) +{

• int i;

• created_tmp = prepare_thp_fs(optional_xfs_path, fs_loc_template,

• 	&fs_loc);
  

• for (i = pmd_order - 1; i >= 0; i--)

• split_thp_in_pagecache_to_order_at(fd_size, fs_loc, i, -1);
  

• if (geteuid() != 0) {

• ksft_print_msg("Please run the benchmark as root\n");
  

• ksft_finished();
  

• }

• for (i = 0; i < pmd_order; i++)

• for (offset = 0;
  

•      offset < nr_pages;
  

•      offset += MAX(nr_pages / 4, 1 << i))
  

• 	split_thp_in_pagecache_to_order_at(fd_size, fs_loc, i, offset);
  

• cleanup_thp_fs(fs_loc, created_tmp);

• for (i = 1; i < argc; i++) {

• /* only one parameter supported */
  

• if (*argv[i] != '-') {
  

• 	optional_xfs_path = argv[i];
  

• 	break;
  

• }
  

• close(pagemap_fd);
• close(kpageflags_fd);
• free(expected_orders);

• /* option -l/-h has no parameter */
  

• if (*(argv[i] + 1) != 'l' && *(argv[i] + 1) != 'h')
  

• 	i++;
  

• }

• ksft_finished();

• pagesize = getpagesize();
• pmd_pagesize = read_pmd_pagesize();
• if (!pmd_pagesize)

• ksft_exit_fail_msg("Reading PMD pagesize failed\n");
  

• pmd_order = sz2ord(pmd_pagesize, pagesize);

• return 0;

• return test_harness_run(argc, argv); }