diff mbox series

[v3,2/2] selftests: vm: add COW time test for KSM pages

Message ID 1d03ee0d1b341959d4b61672c6401d498bff5652.1629386192.git.zhansayabagdaulet@gmail.com (mailing list archive)
State New
Headers show
Series add KSM performance tests | expand

Commit Message

Zhansaya Bagdauletkyzy Aug. 19, 2021, 3:33 p.m. UTC
Since merged pages are copied every time they need to be modified,
the write access time is different between shared and non-shared pages.
Add ksm_cow_time() function which evaluates latency of these COW
breaks. First, 4000 pages are allocated and the time, required to modify
1 byte in every other page, is measured. After this, the pages are
merged into 2000 pairs and in each pair, 1 page is modified (i.e. they
are decoupled) to detect COW breaks. The time needed to break COW of
merged pages is then compared with performance of non-shared pages.

The test is run as follows: ./ksm_tests -C
The output:
	Total size:    15 MiB

	Not merged pages:
	Total time:     0.002185489 s
	Average speed:  3202.945 MiB/s

	Merged pages:
	Total time:     0.004386872 s
	Average speed:  1595.670 MiB/s

Signed-off-by: Zhansaya Bagdauletkyzy <zhansayabagdaulet@gmail.com>
---
v2 -> v3:
 - address Tyler's feedback

Tyler's comments: https://lkml.org/lkml/2021/8/16/931

v1 -> v2:
 As suggested by Pavel,
 - add baseline figures with non-shared pages
 - instead of having all pages merged together, create pairs of
   duplicated pages

Pavel's comments: https://lkml.org/lkml/2021/8/3/1363

 tools/testing/selftests/vm/ksm_tests.c | 86 +++++++++++++++++++++++++-
 1 file changed, 83 insertions(+), 3 deletions(-)

Comments

Tyler Hicks Aug. 19, 2021, 4:31 p.m. UTC | #1
On 2021-08-19 21:33:43, Zhansaya Bagdauletkyzy wrote:
> Since merged pages are copied every time they need to be modified,
> the write access time is different between shared and non-shared pages.
> Add ksm_cow_time() function which evaluates latency of these COW
> breaks. First, 4000 pages are allocated and the time, required to modify
> 1 byte in every other page, is measured. After this, the pages are
> merged into 2000 pairs and in each pair, 1 page is modified (i.e. they
> are decoupled) to detect COW breaks. The time needed to break COW of
> merged pages is then compared with performance of non-shared pages.
> 
> The test is run as follows: ./ksm_tests -C
> The output:
> 	Total size:    15 MiB
> 
> 	Not merged pages:
> 	Total time:     0.002185489 s
> 	Average speed:  3202.945 MiB/s
> 
> 	Merged pages:
> 	Total time:     0.004386872 s
> 	Average speed:  1595.670 MiB/s
> 
> Signed-off-by: Zhansaya Bagdauletkyzy <zhansayabagdaulet@gmail.com>
> ---
> v2 -> v3:
>  - address Tyler's feedback

Thanks for incorporating my suggestions! I can confirm that my feedback
was completely addressed.

 Reviewed-by: Tyler Hicks <tyhicks@linux.microsoft.com>

Tyler

> 
> Tyler's comments: https://lkml.org/lkml/2021/8/16/931
> 
> v1 -> v2:
>  As suggested by Pavel,
>  - add baseline figures with non-shared pages
>  - instead of having all pages merged together, create pairs of
>    duplicated pages
> 
> Pavel's comments: https://lkml.org/lkml/2021/8/3/1363
> 
>  tools/testing/selftests/vm/ksm_tests.c | 86 +++++++++++++++++++++++++-
>  1 file changed, 83 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/testing/selftests/vm/ksm_tests.c b/tools/testing/selftests/vm/ksm_tests.c
> index 432dfe615e50..c14927417759 100644
> --- a/tools/testing/selftests/vm/ksm_tests.c
> +++ b/tools/testing/selftests/vm/ksm_tests.c
> @@ -33,7 +33,8 @@ enum ksm_test_name {
>  	CHECK_KSM_UNMERGE,
>  	CHECK_KSM_ZERO_PAGE_MERGE,
>  	CHECK_KSM_NUMA_MERGE,
> -	KSM_MERGE_TIME
> +	KSM_MERGE_TIME,
> +	KSM_COW_TIME
>  };
>  
>  static int ksm_write_sysfs(const char *file_path, unsigned long val)
> @@ -98,7 +99,8 @@ static void print_help(void)
>  	       " -U (page unmerging)\n"
>  	       " -P evaluate merging time and speed.\n"
>  	       "    For this test, the size of duplicated memory area (in MiB)\n"
> -	       "    must be provided using -s option\n\n");
> +	       "    must be provided using -s option\n"
> +	       " -C evaluate the time required to break COW of merged pages.\n\n");
>  
>  	printf(" -a: specify the access protections of pages.\n"
>  	       "     <prot> must be of the form [rwx].\n"
> @@ -455,6 +457,77 @@ static int ksm_merge_time(int mapping, int prot, int timeout, size_t map_size)
>  	return KSFT_FAIL;
>  }
>  
> +static int ksm_cow_time(int mapping, int prot, int timeout, size_t page_size)
> +{
> +	void *map_ptr;
> +	struct timespec start_time, end_time;
> +	unsigned long cow_time_ns;
> +
> +	/* page_count must be less than 2*page_size */
> +	size_t page_count = 4000;
> +
> +	map_ptr = allocate_memory(NULL, prot, mapping, '*', page_size * page_count);
> +	if (!map_ptr)
> +		return KSFT_FAIL;
> +
> +	if (clock_gettime(CLOCK_MONOTONIC_RAW, &start_time)) {
> +		perror("clock_gettime");
> +		return KSFT_FAIL;
> +	}
> +	for (size_t i = 0; i < page_count - 1; i = i + 2)
> +		memset(map_ptr + page_size * i, '-', 1);
> +	if (clock_gettime(CLOCK_MONOTONIC_RAW, &end_time)) {
> +		perror("clock_gettime");
> +		return KSFT_FAIL;
> +	}
> +
> +	cow_time_ns = (end_time.tv_sec - start_time.tv_sec) * NSEC_PER_SEC +
> +		       (end_time.tv_nsec - start_time.tv_nsec);
> +
> +	printf("Total size:    %lu MiB\n\n", (page_size * page_count) / MB);
> +	printf("Not merged pages:\n");
> +	printf("Total time:     %ld.%09ld s\n", cow_time_ns / NSEC_PER_SEC,
> +	       cow_time_ns % NSEC_PER_SEC);
> +	printf("Average speed:  %.3f MiB/s\n\n", ((page_size * (page_count / 2)) / MB) /
> +					       ((double)cow_time_ns / NSEC_PER_SEC));
> +
> +	/* Create 2000 pairs of duplicate pages */
> +	for (size_t i = 0; i < page_count - 1; i = i + 2) {
> +		memset(map_ptr + page_size * i, '+', i / 2 + 1);
> +		memset(map_ptr + page_size * (i + 1), '+', i / 2 + 1);
> +	}
> +	if (ksm_merge_pages(map_ptr, page_size * page_count, start_time, timeout))
> +		goto err_out;
> +
> +	if (clock_gettime(CLOCK_MONOTONIC_RAW, &start_time)) {
> +		perror("clock_gettime");
> +		goto err_out;
> +	}
> +	for (size_t i = 0; i < page_count - 1; i = i + 2)
> +		memset(map_ptr + page_size * i, '-', 1);
> +	if (clock_gettime(CLOCK_MONOTONIC_RAW, &end_time)) {
> +		perror("clock_gettime");
> +		goto err_out;
> +	}
> +
> +	cow_time_ns = (end_time.tv_sec - start_time.tv_sec) * NSEC_PER_SEC +
> +		       (end_time.tv_nsec - start_time.tv_nsec);
> +
> +	printf("Merged pages:\n");
> +	printf("Total time:     %ld.%09ld s\n", cow_time_ns / NSEC_PER_SEC,
> +	       cow_time_ns % NSEC_PER_SEC);
> +	printf("Average speed:  %.3f MiB/s\n", ((page_size * (page_count / 2)) / MB) /
> +					       ((double)cow_time_ns / NSEC_PER_SEC));
> +
> +	munmap(map_ptr, page_size * page_count);
> +	return KSFT_PASS;
> +
> +err_out:
> +	printf("Not OK\n");
> +	munmap(map_ptr, page_size * page_count);
> +	return KSFT_FAIL;
> +}
> +
>  int main(int argc, char *argv[])
>  {
>  	int ret, opt;
> @@ -468,7 +541,7 @@ int main(int argc, char *argv[])
>  	bool merge_across_nodes = KSM_MERGE_ACROSS_NODES_DEFAULT;
>  	long size_MB = 0;
>  
> -	while ((opt = getopt(argc, argv, "ha:p:l:z:m:s:MUZNP")) != -1) {
> +	while ((opt = getopt(argc, argv, "ha:p:l:z:m:s:MUZNPC")) != -1) {
>  		switch (opt) {
>  		case 'a':
>  			prot = str_to_prot(optarg);
> @@ -522,6 +595,9 @@ int main(int argc, char *argv[])
>  		case 'P':
>  			test_name = KSM_MERGE_TIME;
>  			break;
> +		case 'C':
> +			test_name = KSM_COW_TIME;
> +			break;
>  		default:
>  			return KSFT_FAIL;
>  		}
> @@ -571,6 +647,10 @@ int main(int argc, char *argv[])
>  		ret = ksm_merge_time(MAP_PRIVATE | MAP_ANONYMOUS, prot, ksm_scan_limit_sec,
>  				     size_MB);
>  		break;
> +	case KSM_COW_TIME:
> +		ret = ksm_cow_time(MAP_PRIVATE | MAP_ANONYMOUS, prot, ksm_scan_limit_sec,
> +				   page_size);
> +		break;
>  	}
>  
>  	if (ksm_restore(&ksm_sysfs_old)) {
> -- 
> 2.25.1
>
Pasha Tatashin Aug. 20, 2021, 5:20 p.m. UTC | #2
On Thu, Aug 19, 2021 at 12:31 PM Tyler Hicks
<tyhicks@linux.microsoft.com> wrote:
>
> On 2021-08-19 21:33:43, Zhansaya Bagdauletkyzy wrote:
> > Since merged pages are copied every time they need to be modified,
> > the write access time is different between shared and non-shared pages.
> > Add ksm_cow_time() function which evaluates latency of these COW
> > breaks. First, 4000 pages are allocated and the time, required to modify
> > 1 byte in every other page, is measured. After this, the pages are
> > merged into 2000 pairs and in each pair, 1 page is modified (i.e. they
> > are decoupled) to detect COW breaks. The time needed to break COW of
> > merged pages is then compared with performance of non-shared pages.
> >
> > The test is run as follows: ./ksm_tests -C
> > The output:
> >       Total size:    15 MiB
> >
> >       Not merged pages:
> >       Total time:     0.002185489 s
> >       Average speed:  3202.945 MiB/s
> >
> >       Merged pages:
> >       Total time:     0.004386872 s
> >       Average speed:  1595.670 MiB/s
> >
> > Signed-off-by: Zhansaya Bagdauletkyzy <zhansayabagdaulet@gmail.com>
> > ---
> > v2 -> v3:
> >  - address Tyler's feedback
>
> Thanks for incorporating my suggestions! I can confirm that my feedback
> was completely addressed.
>
>  Reviewed-by: Tyler Hicks <tyhicks@linux.microsoft.com>
>
> Tyler

Reviewed-by: Pavel Tatashin <pasha.tatashin@soleen.com>
diff mbox series

Patch

diff --git a/tools/testing/selftests/vm/ksm_tests.c b/tools/testing/selftests/vm/ksm_tests.c
index 432dfe615e50..c14927417759 100644
--- a/tools/testing/selftests/vm/ksm_tests.c
+++ b/tools/testing/selftests/vm/ksm_tests.c
@@ -33,7 +33,8 @@  enum ksm_test_name {
 	CHECK_KSM_UNMERGE,
 	CHECK_KSM_ZERO_PAGE_MERGE,
 	CHECK_KSM_NUMA_MERGE,
-	KSM_MERGE_TIME
+	KSM_MERGE_TIME,
+	KSM_COW_TIME
 };
 
 static int ksm_write_sysfs(const char *file_path, unsigned long val)
@@ -98,7 +99,8 @@  static void print_help(void)
 	       " -U (page unmerging)\n"
 	       " -P evaluate merging time and speed.\n"
 	       "    For this test, the size of duplicated memory area (in MiB)\n"
-	       "    must be provided using -s option\n\n");
+	       "    must be provided using -s option\n"
+	       " -C evaluate the time required to break COW of merged pages.\n\n");
 
 	printf(" -a: specify the access protections of pages.\n"
 	       "     <prot> must be of the form [rwx].\n"
@@ -455,6 +457,77 @@  static int ksm_merge_time(int mapping, int prot, int timeout, size_t map_size)
 	return KSFT_FAIL;
 }
 
+static int ksm_cow_time(int mapping, int prot, int timeout, size_t page_size)
+{
+	void *map_ptr;
+	struct timespec start_time, end_time;
+	unsigned long cow_time_ns;
+
+	/* page_count must be less than 2*page_size */
+	size_t page_count = 4000;
+
+	map_ptr = allocate_memory(NULL, prot, mapping, '*', page_size * page_count);
+	if (!map_ptr)
+		return KSFT_FAIL;
+
+	if (clock_gettime(CLOCK_MONOTONIC_RAW, &start_time)) {
+		perror("clock_gettime");
+		return KSFT_FAIL;
+	}
+	for (size_t i = 0; i < page_count - 1; i = i + 2)
+		memset(map_ptr + page_size * i, '-', 1);
+	if (clock_gettime(CLOCK_MONOTONIC_RAW, &end_time)) {
+		perror("clock_gettime");
+		return KSFT_FAIL;
+	}
+
+	cow_time_ns = (end_time.tv_sec - start_time.tv_sec) * NSEC_PER_SEC +
+		       (end_time.tv_nsec - start_time.tv_nsec);
+
+	printf("Total size:    %lu MiB\n\n", (page_size * page_count) / MB);
+	printf("Not merged pages:\n");
+	printf("Total time:     %ld.%09ld s\n", cow_time_ns / NSEC_PER_SEC,
+	       cow_time_ns % NSEC_PER_SEC);
+	printf("Average speed:  %.3f MiB/s\n\n", ((page_size * (page_count / 2)) / MB) /
+					       ((double)cow_time_ns / NSEC_PER_SEC));
+
+	/* Create 2000 pairs of duplicate pages */
+	for (size_t i = 0; i < page_count - 1; i = i + 2) {
+		memset(map_ptr + page_size * i, '+', i / 2 + 1);
+		memset(map_ptr + page_size * (i + 1), '+', i / 2 + 1);
+	}
+	if (ksm_merge_pages(map_ptr, page_size * page_count, start_time, timeout))
+		goto err_out;
+
+	if (clock_gettime(CLOCK_MONOTONIC_RAW, &start_time)) {
+		perror("clock_gettime");
+		goto err_out;
+	}
+	for (size_t i = 0; i < page_count - 1; i = i + 2)
+		memset(map_ptr + page_size * i, '-', 1);
+	if (clock_gettime(CLOCK_MONOTONIC_RAW, &end_time)) {
+		perror("clock_gettime");
+		goto err_out;
+	}
+
+	cow_time_ns = (end_time.tv_sec - start_time.tv_sec) * NSEC_PER_SEC +
+		       (end_time.tv_nsec - start_time.tv_nsec);
+
+	printf("Merged pages:\n");
+	printf("Total time:     %ld.%09ld s\n", cow_time_ns / NSEC_PER_SEC,
+	       cow_time_ns % NSEC_PER_SEC);
+	printf("Average speed:  %.3f MiB/s\n", ((page_size * (page_count / 2)) / MB) /
+					       ((double)cow_time_ns / NSEC_PER_SEC));
+
+	munmap(map_ptr, page_size * page_count);
+	return KSFT_PASS;
+
+err_out:
+	printf("Not OK\n");
+	munmap(map_ptr, page_size * page_count);
+	return KSFT_FAIL;
+}
+
 int main(int argc, char *argv[])
 {
 	int ret, opt;
@@ -468,7 +541,7 @@  int main(int argc, char *argv[])
 	bool merge_across_nodes = KSM_MERGE_ACROSS_NODES_DEFAULT;
 	long size_MB = 0;
 
-	while ((opt = getopt(argc, argv, "ha:p:l:z:m:s:MUZNP")) != -1) {
+	while ((opt = getopt(argc, argv, "ha:p:l:z:m:s:MUZNPC")) != -1) {
 		switch (opt) {
 		case 'a':
 			prot = str_to_prot(optarg);
@@ -522,6 +595,9 @@  int main(int argc, char *argv[])
 		case 'P':
 			test_name = KSM_MERGE_TIME;
 			break;
+		case 'C':
+			test_name = KSM_COW_TIME;
+			break;
 		default:
 			return KSFT_FAIL;
 		}
@@ -571,6 +647,10 @@  int main(int argc, char *argv[])
 		ret = ksm_merge_time(MAP_PRIVATE | MAP_ANONYMOUS, prot, ksm_scan_limit_sec,
 				     size_MB);
 		break;
+	case KSM_COW_TIME:
+		ret = ksm_cow_time(MAP_PRIVATE | MAP_ANONYMOUS, prot, ksm_scan_limit_sec,
+				   page_size);
+		break;
 	}
 
 	if (ksm_restore(&ksm_sysfs_old)) {