[PATCH v2 0/3] selftests/mm: avoid false failures in hugetlb and KSM tests

[PATCH v2 0/3] selftests/mm: avoid false failures in hugetlb and KSM tests

[Sayali Patil]

Hi all,

This series fixes issues in the hugetlb and KSM MM selftest categories
that can report failures when the prerequisites for the tests are not
satisfied.

Patch 1 updates the hugetlb selftest helpers to handle -EINVAL when
attempting to configure gigantic HugeTLB pages via nr_hugepages. 
PowerPC hash MMU pSeries systems expose gigantic hugepage sizes 
but do not allow runtime allocation of such pages,
causing the sysfs write to fail. Handle this case gracefully and
continue running the test instead of aborting.

Patch 2 fixes the KSM NUMA merge test on systems with memoryless NUMA
nodes. The test currently relies on the number of configured NUMA nodes
and may attempt allocations on nodes that have no memory, resulting in
spurious failures. Use the existing helpers to identify
NUMA nodes that contain memory and skip the test when fewer than two
such nodes are available.

Patch 3 fixes a pre-existing operator precedence issue in ksm_tests,
where a ternary expression combined with logical OR operators could be
evaluated differently than intended. Added parentheses to ensure the
correct evaluation order.

These changes improve handling of unsupported test configurations and
unmet test prerequisites, avoiding spurious failures.

Thanks,
Sayali
---

V1 -> V2:
- For "selftests/mm: handle EINVAL when configuring gigantic hugepages":
  Added a dedicated hugetlb_write_num() helper to handle the expected EINVAL
  returned by gigantic hugepage configuration, in the 
  hugepage setup code rather than modifying the generic write_file() helper.
- For "selftests/mm: fix ksm NUMA merge test for systems with memoryless NUMA nodes":
  Updated implementation as per David's review comment to build upon 
  existing helpers get_next_mem_node().
- Added new patch "selftests/mm: fix ternary operator precedence in ksm_tests"
  to address a pre-existing issue identified during Sahiko's review.

V1: https://lore.kernel.org/all/cover.1782365671.git.sayalip@linux.ibm.com/

---

Sayali Patil (3):
  selftests/mm: handle EINVAL when configuring gigantic hugepages
  selftests/mm: fix ksm NUMA merge test for systems with memoryless NUMA
    nodes
  selftests/mm: fix ternary operator precedence in ksm_tests

 .../testing/selftests/mm/hugepage_settings.c  | 32 ++++++++++++++++++-
 .../testing/selftests/mm/hugepage_settings.h  |  1 +
 tools/testing/selftests/mm/ksm_tests.c        | 28 ++++++++--------
 3 files changed, 47 insertions(+), 14 deletions(-)

-- 
2.52.0

[PATCH v2 1/3] selftests/mm: handle EINVAL when configuring gigantic hugepages

[Sayali Patil]

Some MM selftests attempt to configure the amount of
HugeTLB pages of different sizes by writing to nr_hugepages.

PowerPC hash MMU pSeries systems advertise gigantic hugepage sizes
but do not support runtime allocation of such pages, writes
to the corresponding nr_hugepages file fail with -EINVAL.
This causes the test to bail out even though the failure is due
to a platform limitation rather than the
functionality being tested.

Treat -EINVAL from the sysfs write as a skipped configuration request
and continue running the test instead of failing.

Before patch:
   -------------------------
   running ./hugetlb-madvise
   -------------------------
   TAP version 13
   1..1
     [INFO] detected hugetlb page size: 16777216 KiB
     [INFO] detected hugetlb page size: 16384 KiB
    ok 1 MADV_DONTNEED and MADV_REMOVE on hugetlb
    Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
    Bail out! /sys/kernel/mm/hugepages/hugepages-16777216kB/nr_hugepages
    write(0) failed: Invalid argument
    Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
    [FAIL]

After patch:
   -------------------------
   running ./hugetlb-madvise
   -------------------------
   TAP version 13
   1..1
    [INFO] detected hugetlb page size: 16777216 KiB
    [INFO] detected hugetlb page size: 16384 KiB
   ok 1 MADV_DONTNEED and MADV_REMOVE on hugetlb
   Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
   /sys/kernel/mm/hugepages/hugepages-16777216kB/nr_hugepages
   write(0) failed: Invalid argument
   [PASS]

Fixes: 27477b28b74f ("selftests/mm: hugepage_settings: add APIs to get and set nr_hugepages")
Signed-off-by: Sayali Patil <sayalip@linux.ibm.com>
---
 .../testing/selftests/mm/hugepage_settings.c  | 32 ++++++++++++++++++-
 .../testing/selftests/mm/hugepage_settings.h  |  1 +
 2 files changed, 32 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/mm/hugepage_settings.c b/tools/testing/selftests/mm/hugepage_settings.c
index 2eab2110ac6a..ce38ae3da01a 100644
--- a/tools/testing/selftests/mm/hugepage_settings.c
+++ b/tools/testing/selftests/mm/hugepage_settings.c
@@ -422,6 +422,36 @@ static void hugetlb_sysfs_path(char *buf, size_t buflen,
 		 size / 1024, attr);
 }
 
+void hugetlb_write_num(const char *path, unsigned long num)
+{
+	int fd, saved_errno;
+	ssize_t numwritten;
+	char buf[21];
+
+	sprintf(buf, "%lu", num);
+
+	fd = open(path, O_WRONLY);
+	if (fd == -1)
+		ksft_exit_fail_msg("%s open failed: %s\n", path, strerror(errno));
+
+	numwritten = write(fd, buf, strlen(buf));
+	saved_errno = errno;
+	close(fd);
+	errno = saved_errno;
+
+	/* Treat EINVAL as a skipped configuration (e.g., unsupported gigantic pages) */
+	if (numwritten < 0 && errno == EINVAL) {
+		ksft_print_msg("%s write(%s) failed: %s\n", path, buf, strerror(errno));
+		return;
+	}
+
+	if (numwritten < 0)
+		ksft_exit_fail_msg("%s write(%s) failed: %s\n", path, buf, strerror(errno));
+	if (numwritten != strlen(buf))
+		ksft_exit_fail_msg("%s write(%s) is truncated, expected %zu bytes, got %zd bytes\n",
+				   path, buf, strlen(buf), numwritten);
+}
+
 unsigned long hugetlb_nr_pages(unsigned long size)
 {
 	char path[PATH_MAX];
@@ -437,7 +467,7 @@ void hugetlb_set_nr_pages(unsigned long size, unsigned long nr)
 
 	hugetlb_sysfs_path(path, sizeof(path), size, "nr_hugepages");
 
-	write_num(path, nr);
+	hugetlb_write_num(path, nr);
 }
 
 unsigned long hugetlb_free_pages(unsigned long size)
diff --git a/tools/testing/selftests/mm/hugepage_settings.h b/tools/testing/selftests/mm/hugepage_settings.h
index 726c73c43c05..390fcc483b9e 100644
--- a/tools/testing/selftests/mm/hugepage_settings.h
+++ b/tools/testing/selftests/mm/hugepage_settings.h
@@ -95,6 +95,7 @@ bool thp_is_enabled(void);
 int detect_hugetlb_page_sizes(unsigned long sizes[], int max);
 unsigned long default_huge_page_size(void);
 
+void hugetlb_write_num(const char *path, unsigned long num);
 unsigned long hugetlb_nr_pages(unsigned long size);
 void hugetlb_set_nr_pages(unsigned long size, unsigned long nr);
 unsigned long hugetlb_free_pages(unsigned long size);
-- 
2.52.0

[PATCH v2 2/3] selftests/mm: fix ksm NUMA merge test for systems with memoryless NUMA nodes

[Sayali Patil]

The KSM NUMA merge test allocates identical pages on different NUMA
nodes and verifies KSM behavior with merge_across_nodes enabled and
disabled.

On systems with memoryless NUMA nodes, for example:
 #numactl  -H
      available: 2 nodes (0,4)
      .....
      node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
      node 0 size: 14825 MB
      node 0 free: 1382 MB
      node 4 cpus:
      node 4 size: 0 MB
      node 4 free: 0 MB

the test may attempt to allocate memory on a node without memory,
causing numa_alloc_onnode() to fail and resulting in a spurious test
failure.

The test currently checks numa_num_configured_nodes() to determine
whether sufficient NUMA nodes are available. However, configured nodes
do not necessarily have memory.

Reuse the existing get_first_mem_node() and get_next_mem_node()
helpers to locate NUMA nodes that actually contain memory, and skip
the test when fewer than two such nodes are available.

Before patch:
       ---------------------------
	running ./ksm_tests -N -m 1
       ---------------------------
        mbind: Invalid argument
        ok 1 KSM NUMA merging
	Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
        [PASS]
       ok 1 ksm_tests -N -m 1
       ---------------------------
        running ./ksm_tests -N -m 0
       ---------------------------
        mbind: Invalid argument
        not ok 1 KSM NUMA merging
	Totals: pass:0 fail:1 xfail:0 xpass:0 skip:0 error:0
        [FAIL]
       not ok 2 ksm_tests -N -m 0 # exit=1

After patch:
       ---------------------------
        running ./ksm_tests -N -m 1
       ---------------------------
        At least 2 NUMA nodes with memory must be available
	ok 1
	SKIP KSM NUMA merging
	Totals: pass:0 fail:0 xfail:0 xpass:0 skip:1 error:0
        [PASS]
        ok 1 ksm_tests -N -m 1
       ---------------------------
        running ./ksm_tests -N -m 0
       ---------------------------
        At least 2 NUMA nodes with memory must be available
	ok 1
	SKIP KSM NUMA merging
	Totals: pass:0 fail:0 xfail:0 xpass:0 skip:1 error:0
        [PASS]
        ok 2 ksm_tests -N -m 0

Fixes: e3820ab252dd ("selftest/vm: fix ksm selftest to run with different NUMA topologies")
Co-developed-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Sayali Patil <sayalip@linux.ibm.com>
---
 tools/testing/selftests/mm/ksm_tests.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/tools/testing/selftests/mm/ksm_tests.c b/tools/testing/selftests/mm/ksm_tests.c
index a050f4840cfa..2ebbb544c671 100644
--- a/tools/testing/selftests/mm/ksm_tests.c
+++ b/tools/testing/selftests/mm/ksm_tests.c
@@ -440,9 +440,9 @@ static int get_next_mem_node(int node)
 		mem_node = i % (max_node + 1);
 		node_size = numa_node_size(mem_node, NULL);
 		if (node_size > 0)
-			break;
+			return mem_node;
 	}
-	return mem_node;
+	return -ENODEV;
 }
 
 static int get_first_mem_node(void)
@@ -455,8 +455,8 @@ static int check_ksm_numa_merge(int merge_type, int mapping, int prot, int timeo
 {
 	void *numa1_map_ptr, *numa2_map_ptr;
 	struct timespec start_time;
+	int first_node, second_node;
 	int page_count = 2;
-	int first_node;
 
 	if (clock_gettime(CLOCK_MONOTONIC_RAW, &start_time)) {
 		ksft_perror("clock_gettime");
@@ -467,17 +467,19 @@ static int check_ksm_numa_merge(int merge_type, int mapping, int prot, int timeo
 		ksft_print_msg("NUMA support not enabled\n");
 		return KSFT_SKIP;
 	}
-	if (numa_num_configured_nodes() <= 1) {
-		ksft_print_msg("At least 2 NUMA nodes must be available\n");
+	first_node = get_first_mem_node();
+	second_node = get_next_mem_node(first_node);
+
+	if (second_node < 0) {
+		ksft_print_msg("At least 2 NUMA nodes with memory must be available\n");
 		return KSFT_SKIP;
 	}
 	if (ksm_write_sysfs(KSM_FP("merge_across_nodes"), merge_across_nodes))
 		return KSFT_FAIL;
 
 	/* allocate 2 pages in 2 different NUMA nodes and fill them with the same data */
-	first_node = get_first_mem_node();
 	numa1_map_ptr = numa_alloc_onnode(page_size, first_node);
-	numa2_map_ptr = numa_alloc_onnode(page_size, get_next_mem_node(first_node));
+	numa2_map_ptr = numa_alloc_onnode(page_size, second_node);
 	if (!numa1_map_ptr || !numa2_map_ptr) {
 		ksft_perror("numa_alloc_onnode");
 		return KSFT_FAIL;
-- 
2.52.0

[PATCH v2 3/3] selftests/mm: fix ternary operator precedence in ksm_tests

[Sayali Patil]

The KSM selftest uses conditional expressions to skip accesses to
merge_across_nodes on systems without NUMA support. However, the
ternary operator is combined with logical OR without parentheses:

a || numa_available() ? 0 : b || c

Due to operator precedence rules, this is parsed as:

(a || numa_available()) ? 0 : (b || c)

instead of the intended:

a || (numa_available() ? 0 : b) || c

Add parentheses around the conditional expressions to ensure the
correct evaluation order.

Fixes: 9aa1af954db0 ("selftests: vm: check numa_available() before operating "merge_across_nodes" in ksm_tests")
Signed-off-by: Sayali Patil <sayalip@linux.ibm.com>
---
 tools/testing/selftests/mm/ksm_tests.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/tools/testing/selftests/mm/ksm_tests.c b/tools/testing/selftests/mm/ksm_tests.c
index 2ebbb544c671..5fd7792a0d47 100644
--- a/tools/testing/selftests/mm/ksm_tests.c
+++ b/tools/testing/selftests/mm/ksm_tests.c
@@ -288,8 +288,8 @@ static bool assert_ksm_pages_count(long dupl_page_count)
 static int ksm_save_def(struct ksm_sysfs *ksm_sysfs)
 {
 	if (ksm_read_sysfs(KSM_FP("max_page_sharing"), &ksm_sysfs->max_page_sharing) ||
-	    numa_available() ? 0 :
-		ksm_read_sysfs(KSM_FP("merge_across_nodes"), &ksm_sysfs->merge_across_nodes) ||
+	    (numa_available() ? 0 :
+		ksm_read_sysfs(KSM_FP("merge_across_nodes"), &ksm_sysfs->merge_across_nodes)) ||
 	    ksm_read_sysfs(KSM_FP("sleep_millisecs"), &ksm_sysfs->sleep_millisecs) ||
 	    ksm_read_sysfs(KSM_FP("pages_to_scan"), &ksm_sysfs->pages_to_scan) ||
 	    ksm_read_sysfs(KSM_FP("run"), &ksm_sysfs->run) ||
@@ -304,8 +304,8 @@ static int ksm_save_def(struct ksm_sysfs *ksm_sysfs)
 static int ksm_restore(struct ksm_sysfs *ksm_sysfs)
 {
 	if (ksm_write_sysfs(KSM_FP("max_page_sharing"), ksm_sysfs->max_page_sharing) ||
-	    numa_available() ? 0 :
-		ksm_write_sysfs(KSM_FP("merge_across_nodes"), ksm_sysfs->merge_across_nodes) ||
+	    (numa_available() ? 0 :
+		ksm_write_sysfs(KSM_FP("merge_across_nodes"), ksm_sysfs->merge_across_nodes)) ||
 	    ksm_write_sysfs(KSM_FP("pages_to_scan"), ksm_sysfs->pages_to_scan) ||
 	    ksm_write_sysfs(KSM_FP("run"), ksm_sysfs->run) ||
 	    ksm_write_sysfs(KSM_FP("sleep_millisecs"), ksm_sysfs->sleep_millisecs) ||
@@ -846,8 +846,8 @@ int main(int argc, char *argv[])
 
 	if (ksm_write_sysfs(KSM_FP("run"), 2) ||
 	    ksm_write_sysfs(KSM_FP("sleep_millisecs"), 0) ||
-	    numa_available() ? 0 :
-		ksm_write_sysfs(KSM_FP("merge_across_nodes"), 1) ||
+	    (numa_available() ? 0 :
+		ksm_write_sysfs(KSM_FP("merge_across_nodes"), 1)) ||
 	    ksm_write_sysfs(KSM_FP("pages_to_scan"), page_count))
 		ksft_exit_fail_msg("Cannot set up KSM tunables\n");
 
-- 
2.52.0

Re: [PATCH v2 3/3] selftests/mm: fix ternary operator precedence in ksm_tests

[David Hildenbrand (Arm)]

On 6/30/26 11:32, Sayali Patil wrote:
> The KSM selftest uses conditional expressions to skip accesses to
> merge_across_nodes on systems without NUMA support. However, the
> ternary operator is combined with logical OR without parentheses:
> 
> a || numa_available() ? 0 : b || c
> 
> Due to operator precedence rules, this is parsed as:
> 
> (a || numa_available()) ? 0 : (b || c)
> 
> instead of the intended:
> 
> a || (numa_available() ? 0 : b) || c
> 
> Add parentheses around the conditional expressions to ensure the
> correct evaluation order.
> 
> Fixes: 9aa1af954db0 ("selftests: vm: check numa_available() before operating "merge_across_nodes" in ksm_tests")
> Signed-off-by: Sayali Patil <sayalip@linux.ibm.com>
> ---

LGTM, although the code is a bit ugly (already before your changes).

Acked-by: David Hildenbrand (Arm) <david@kernel.org>

-- 
Cheers,

David

Re: [PATCH v2 2/3] selftests/mm: fix ksm NUMA merge test for systems with memoryless NUMA nodes

[Sayali Patil]

On 30/06/26 15:02, Sayali Patil wrote:
> The KSM NUMA merge test allocates identical pages on different NUMA
> nodes and verifies KSM behavior with merge_across_nodes enabled and
> disabled.
> 
> On systems with memoryless NUMA nodes, for example:
>   #numactl  -H
>        available: 2 nodes (0,4)
>        .....
>        node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
>        node 0 size: 14825 MB
>        node 0 free: 1382 MB
>        node 4 cpus:
>        node 4 size: 0 MB
>        node 4 free: 0 MB
> 
> the test may attempt to allocate memory on a node without memory,
> causing numa_alloc_onnode() to fail and resulting in a spurious test
> failure.
> 
> The test currently checks numa_num_configured_nodes() to determine
> whether sufficient NUMA nodes are available. However, configured nodes
> do not necessarily have memory.
> 
> Reuse the existing get_first_mem_node() and get_next_mem_node()
> helpers to locate NUMA nodes that actually contain memory, and skip
> the test when fewer than two such nodes are available.
> 
> Before patch:
>         ---------------------------
> 	running ./ksm_tests -N -m 1
>         ---------------------------
>          mbind: Invalid argument
>          ok 1 KSM NUMA merging
> 	Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
>          [PASS]
>         ok 1 ksm_tests -N -m 1
>         ---------------------------
>          running ./ksm_tests -N -m 0
>         ---------------------------
>          mbind: Invalid argument
>          not ok 1 KSM NUMA merging
> 	Totals: pass:0 fail:1 xfail:0 xpass:0 skip:0 error:0
>          [FAIL]
>         not ok 2 ksm_tests -N -m 0 # exit=1
> 
> After patch:
>         ---------------------------
>          running ./ksm_tests -N -m 1
>         ---------------------------
>          At least 2 NUMA nodes with memory must be available
> 	ok 1
> 	SKIP KSM NUMA merging
> 	Totals: pass:0 fail:0 xfail:0 xpass:0 skip:1 error:0
>          [PASS]
>          ok 1 ksm_tests -N -m 1
>         ---------------------------
>          running ./ksm_tests -N -m 0
>         ---------------------------
>          At least 2 NUMA nodes with memory must be available
> 	ok 1
> 	SKIP KSM NUMA merging
> 	Totals: pass:0 fail:0 xfail:0 xpass:0 skip:1 error:0
>          [PASS]
>          ok 2 ksm_tests -N -m 0
> 
> Fixes: e3820ab252dd ("selftest/vm: fix ksm selftest to run with different NUMA topologies")
> Co-developed-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
> Signed-off-by: Sayali Patil <sayalip@linux.ibm.com>
> ---
>   tools/testing/selftests/mm/ksm_tests.c | 16 +++++++++-------
>   1 file changed, 9 insertions(+), 7 deletions(-)
> 
> diff --git a/tools/testing/selftests/mm/ksm_tests.c b/tools/testing/selftests/mm/ksm_tests.c
> index a050f4840cfa..2ebbb544c671 100644
> --- a/tools/testing/selftests/mm/ksm_tests.c
> +++ b/tools/testing/selftests/mm/ksm_tests.c
> @@ -440,9 +440,9 @@ static int get_next_mem_node(int node)
>   		mem_node = i % (max_node + 1);
>   		node_size = numa_node_size(mem_node, NULL);
>   		if (node_size > 0)
> -			break;
> +			return mem_node;
>   	}
> -	return mem_node;
> +	return -ENODEV;
>   }
>   
>   static int get_first_mem_node(void)
> @@ -455,8 +455,8 @@ static int check_ksm_numa_merge(int merge_type, int mapping, int prot, int timeo
>   {
>   	void *numa1_map_ptr, *numa2_map_ptr;
>   	struct timespec start_time;
> +	int first_node, second_node;
>   	int page_count = 2;
> -	int first_node;
>   
>   	if (clock_gettime(CLOCK_MONOTONIC_RAW, &start_time)) {
>   		ksft_perror("clock_gettime");
> @@ -467,17 +467,19 @@ static int check_ksm_numa_merge(int merge_type, int mapping, int prot, int timeo
>   		ksft_print_msg("NUMA support not enabled\n");
>   		return KSFT_SKIP;
>   	}
> -	if (numa_num_configured_nodes() <= 1) {
> -		ksft_print_msg("At least 2 NUMA nodes must be available\n");
> +	first_node = get_first_mem_node();
> +	second_node = get_next_mem_node(first_node);
> +
> +	if (second_node < 0) {
> +		ksft_print_msg("At least 2 NUMA nodes with memory must be available\n");
>   		return KSFT_SKIP;
>   	}
>   	if (ksm_write_sysfs(KSM_FP("merge_across_nodes"), merge_across_nodes))
>   		return KSFT_FAIL;
>   
>   	/* allocate 2 pages in 2 different NUMA nodes and fill them with the same data */
> -	first_node = get_first_mem_node();
>   	numa1_map_ptr = numa_alloc_onnode(page_size, first_node);
> -	numa2_map_ptr = numa_alloc_onnode(page_size, get_next_mem_node(first_node));
> +	numa2_map_ptr = numa_alloc_onnode(page_size, second_node);
>   	if (!numa1_map_ptr || !numa2_map_ptr) {
>   		ksft_perror("numa_alloc_onnode");
>   		return KSFT_FAIL;

AI review comment:
> If get_first_mem_node() doesn't find a node with memory, will it
> return -ENODEV and pass it directly into get_next_mem_node()?
> Looking at get_next_mem_node() with a negative node parameter:
> tools/testing/selftests/mm/ksm_tests.c:get_next_mem_node() {
>    ...
>    for (i = node + 1; i <= max_node + node; i++) {
>        mem_node = i % (max_node + 1);
>        node_size = numa_node_size(mem_node, NULL);
>    ...
> }
> Because the modulo operator preserves the sign of a negative dividend,
> starting the loop with a negative node value causes mem_node to become
> negative. This means a negative node ID is then passed into libnuma's
> numa_node_size() function.
> Could we check if first_node is valid before attempting to find the
> second node?

The additional check for get_first_mem_node() is not needed. After 
numa_available() succeeds, a running system should always have at least 
one NUMA node with memory, so get_first_mem_node() is expected to return 
a valid node ID. The actual prerequisite for this test is the presence 
of two NUMA nodes with memory, which is already validated by checking 
the return value of get_next_mem_node().

Thanks,
Sayali

Re: [PATCH v2 1/3] selftests/mm: handle EINVAL when configuring gigantic hugepages

[David Hildenbrand (Arm)]

On 6/30/26 11:32, Sayali Patil wrote:
> Some MM selftests attempt to configure the amount of
> HugeTLB pages of different sizes by writing to nr_hugepages.
> 
> PowerPC hash MMU pSeries systems advertise gigantic hugepage sizes
> but do not support runtime allocation of such pages, writes
> to the corresponding nr_hugepages file fail with -EINVAL.
> This causes the test to bail out even though the failure is due
> to a platform limitation rather than the
> functionality being tested.
> 
> Treat -EINVAL from the sysfs write as a skipped configuration request
> and continue running the test instead of failing.
> 
> Before patch:
>    -------------------------
>    running ./hugetlb-madvise
>    -------------------------
>    TAP version 13
>    1..1
>      [INFO] detected hugetlb page size: 16777216 KiB
>      [INFO] detected hugetlb page size: 16384 KiB
>     ok 1 MADV_DONTNEED and MADV_REMOVE on hugetlb
>     Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
>     Bail out! /sys/kernel/mm/hugepages/hugepages-16777216kB/nr_hugepages
>     write(0) failed: Invalid argument
>     Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
>     [FAIL]
> 
> After patch:
>    -------------------------
>    running ./hugetlb-madvise
>    -------------------------
>    TAP version 13
>    1..1
>     [INFO] detected hugetlb page size: 16777216 KiB
>     [INFO] detected hugetlb page size: 16384 KiB
>    ok 1 MADV_DONTNEED and MADV_REMOVE on hugetlb
>    Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
>    /sys/kernel/mm/hugepages/hugepages-16777216kB/nr_hugepages
>    write(0) failed: Invalid argument
>    [PASS]
> 
> Fixes: 27477b28b74f ("selftests/mm: hugepage_settings: add APIs to get and set nr_hugepages")
> Signed-off-by: Sayali Patil <sayalip@linux.ibm.com>
> ---
>  .../testing/selftests/mm/hugepage_settings.c  | 32 ++++++++++++++++++-
>  .../testing/selftests/mm/hugepage_settings.h  |  1 +
>  2 files changed, 32 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/mm/hugepage_settings.c b/tools/testing/selftests/mm/hugepage_settings.c
> index 2eab2110ac6a..ce38ae3da01a 100644
> --- a/tools/testing/selftests/mm/hugepage_settings.c
> +++ b/tools/testing/selftests/mm/hugepage_settings.c
> @@ -422,6 +422,36 @@ static void hugetlb_sysfs_path(char *buf, size_t buflen,
>  		 size / 1024, attr);
>  }
>  
> +void hugetlb_write_num(const char *path, unsigned long num)
> +{
> +	int fd, saved_errno;
> +	ssize_t numwritten;
> +	char buf[21];
> +
> +	sprintf(buf, "%lu", num);
> +
> +	fd = open(path, O_WRONLY);
> +	if (fd == -1)
> +		ksft_exit_fail_msg("%s open failed: %s\n", path, strerror(errno));
> +
> +	numwritten = write(fd, buf, strlen(buf));
> +	saved_errno = errno;
> +	close(fd);
> +	errno = saved_errno;
> +
> +	/* Treat EINVAL as a skipped configuration (e.g., unsupported gigantic pages) */
> +	if (numwritten < 0 && errno == EINVAL) {
> +		ksft_print_msg("%s write(%s) failed: %s\n", path, buf, strerror(errno));

Should we even print anything here? Rather confusing. It's just like we cannot
allocate anything (no memory).

In general, you are copy-pasting a lot of write_num()+write_file() content,
which is really suboptimal.

All you want is an option for write_num -> write_file to skip on -EINVAL, correct?

There are not that many write_num / write_file users ...

-- 
Cheers,

David

Re: [PATCH v2 1/3] selftests/mm: handle EINVAL when configuring gigantic hugepages

[Sayali Patil]

On 30/06/26 16:15, David Hildenbrand (Arm) wrote:
> On 6/30/26 11:32, Sayali Patil wrote:
>> Some MM selftests attempt to configure the amount of
>> HugeTLB pages of different sizes by writing to nr_hugepages.
>>
>> PowerPC hash MMU pSeries systems advertise gigantic hugepage sizes
>> but do not support runtime allocation of such pages, writes
>> to the corresponding nr_hugepages file fail with -EINVAL.
>> This causes the test to bail out even though the failure is due
>> to a platform limitation rather than the
>> functionality being tested.
>>
>> Treat -EINVAL from the sysfs write as a skipped configuration request
>> and continue running the test instead of failing.
>>
>> Before patch:
>>     -------------------------
>>     running ./hugetlb-madvise
>>     -------------------------
>>     TAP version 13
>>     1..1
>>       [INFO] detected hugetlb page size: 16777216 KiB
>>       [INFO] detected hugetlb page size: 16384 KiB
>>      ok 1 MADV_DONTNEED and MADV_REMOVE on hugetlb
>>      Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
>>      Bail out! /sys/kernel/mm/hugepages/hugepages-16777216kB/nr_hugepages
>>      write(0) failed: Invalid argument
>>      Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
>>      [FAIL]
>>
>> After patch:
>>     -------------------------
>>     running ./hugetlb-madvise
>>     -------------------------
>>     TAP version 13
>>     1..1
>>      [INFO] detected hugetlb page size: 16777216 KiB
>>      [INFO] detected hugetlb page size: 16384 KiB
>>     ok 1 MADV_DONTNEED and MADV_REMOVE on hugetlb
>>     Totals: pass:1 fail:0 xfail:0 xpass:0 skip:0 error:0
>>     /sys/kernel/mm/hugepages/hugepages-16777216kB/nr_hugepages
>>     write(0) failed: Invalid argument
>>     [PASS]
>>
>> Fixes: 27477b28b74f ("selftests/mm: hugepage_settings: add APIs to get and set nr_hugepages")
>> Signed-off-by: Sayali Patil <sayalip@linux.ibm.com>
>> ---
>>   .../testing/selftests/mm/hugepage_settings.c  | 32 ++++++++++++++++++-
>>   .../testing/selftests/mm/hugepage_settings.h  |  1 +
>>   2 files changed, 32 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/mm/hugepage_settings.c b/tools/testing/selftests/mm/hugepage_settings.c
>> index 2eab2110ac6a..ce38ae3da01a 100644
>> --- a/tools/testing/selftests/mm/hugepage_settings.c
>> +++ b/tools/testing/selftests/mm/hugepage_settings.c
>> @@ -422,6 +422,36 @@ static void hugetlb_sysfs_path(char *buf, size_t buflen,
>>   		 size / 1024, attr);
>>   }
>>   
>> +void hugetlb_write_num(const char *path, unsigned long num)
>> +{
>> +	int fd, saved_errno;
>> +	ssize_t numwritten;
>> +	char buf[21];
>> +
>> +	sprintf(buf, "%lu", num);
>> +
>> +	fd = open(path, O_WRONLY);
>> +	if (fd == -1)
>> +		ksft_exit_fail_msg("%s open failed: %s\n", path, strerror(errno));
>> +
>> +	numwritten = write(fd, buf, strlen(buf));
>> +	saved_errno = errno;
>> +	close(fd);
>> +	errno = saved_errno;
>> +
>> +	/* Treat EINVAL as a skipped configuration (e.g., unsupported gigantic pages) */
>> +	if (numwritten < 0 && errno == EINVAL) {
>> +		ksft_print_msg("%s write(%s) failed: %s\n", path, buf, strerror(errno));
> 
> Should we even print anything here? Rather confusing. It's just like we cannot
> allocate anything (no memory).
> 
> In general, you are copy-pasting a lot of write_num()+write_file() content,
> which is really suboptimal.
> 
> All you want is an option for write_num -> write_file to skip on -EINVAL, correct?
> 
> There are not that many write_num / write_file users ...
> 

Hi David,

Yes, all I need is to ignore the expected -EINVAL when attempting to
configure gigantic hugepages via nr_hugepages.

I looked at extending write_num()/write_file() for this as in v1
(https://lore.kernel.org/all/8bfa921e30eb94072685103f6496784aa23bb166.1782365671.git.sayalip@linux.ibm.com/),
but these helpers are shared by several other selftests.
For example, write_file() is used by split_huge_page_test setup and by
khugepaged tests for drop_caches, and is also used for various THP and
khugepaged settings where -EINVAL would indicate a genuine setup
failure. This concern was also raised during the v1 review.

Because the expected -EINVAL is specific to gigantic hugepage runtime
allocation, I kept the handling local to the hugetlb setup path rather
than changing the semantics of the common helpers.

I also agree that printing a message is not particularly useful in this
case, and we can simply return without emitting any output.

Please let me know if you would prefer a different approach.

Thanks,
Sayali