[PATCH v6 0/2] selftests/resctrl: SNC kernel support discovery

[PATCH v6 0/2] selftests/resctrl: SNC kernel support discovery

[Maciej Wieczor-Retman]

Changes v6:
- Rebase onto latest kselftests-next.
- Looking at the two patches with a fresh eye decided to make a split
  along the lines of:
	- Patch 1/2 contains all of the code that relates to SNC mode
	  detection and checking that detection's reliability.
	- Patch 2/2 contains checking kernel support for SNC and
	  modifying the messages at the end of affected tests.

Changes v5:
- Tests are skipped if snc_unreliable was set.
- Moved resctrlfs.c changes from patch 2/2 to 1/2.
- Removed CAT changes since it's not impacted by SNC in the selftest.
- Updated various comments.
- Fixed a bunch of minor issues pointed out in the review.

Changes v4:
- Printing SNC warnings at the start of every test.
- Printing SNC warnings at the end of every relevant test.
- Remove global snc_mode variable, consolidate snc detection functions
  into one.
- Correct minor mistakes.

Changes v3:
- Reworked patch 2.
- Changed minor things in patch 1 like function name and made
  corrections to the patch message.

Changes v2:
- Removed patches 2 and 3 since now this part will be supported by the
  kernel.

Sub-Numa Clustering (SNC) allows splitting CPU cores, caches and memory
into multiple NUMA nodes. When enabled, NUMA-aware applications can
achieve better performance on bigger server platforms.

SNC support in the kernel was merged into x86/cache [1]. With SNC enabled
and kernel support in place all the tests will function normally (aside
from effective cache size). There might be a problem when SNC is enabled
but the system is still using an older kernel version without SNC
support. Currently the only message displayed in that situation is a
guess that SNC might be enabled and is causing issues. That message also
is displayed whenever the test fails on an Intel platform.

Add a mechanism to discover kernel support for SNC which will add more
meaning and certainty to the error message.

Add runtime SNC mode detection and verify how reliable that information
is.

Series was tested on Ice Lake server platforms with SNC disabled, SNC-2
and SNC-4. The tests were also ran with and without kernel support for
SNC.

Series applies cleanly on kselftest/next.

[1] https://lore.kernel.org/all/20240628215619.76401-1-tony.luck@intel.com/

Previous versions of this series:
[v1] https://lore.kernel.org/all/cover.1709721159.git.maciej.wieczor-retman@intel.com/
[v2] https://lore.kernel.org/all/cover.1715769576.git.maciej.wieczor-retman@intel.com/
[v3] https://lore.kernel.org/all/cover.1719842207.git.maciej.wieczor-retman@intel.com/
[v4] https://lore.kernel.org/all/cover.1720774981.git.maciej.wieczor-retman@intel.com/
[v5] https://lore.kernel.org/all/cover.1730206468.git.maciej.wieczor-retman@intel.com/

Maciej Wieczor-Retman (2):
  selftests/resctrl: Adjust effective L3 cache size with SNC enabled
  selftests/resctrl: Discover SNC kernel support and adjust messages

 tools/testing/selftests/resctrl/cmt_test.c    |   4 +-
 tools/testing/selftests/resctrl/mba_test.c    |   2 +
 tools/testing/selftests/resctrl/mbm_test.c    |   4 +-
 tools/testing/selftests/resctrl/resctrl.h     |   5 +
 .../testing/selftests/resctrl/resctrl_tests.c |   9 +-
 tools/testing/selftests/resctrl/resctrlfs.c   | 129 ++++++++++++++++++
 6 files changed, 148 insertions(+), 5 deletions(-)

-- 
2.47.1

[PATCH v6 1/2] selftests/resctrl: Adjust effective L3 cache size with SNC enabled

[Maciej Wieczor-Retman]

Sub-NUMA Cluster divides CPUs sharing an L3 cache into separate NUMA
nodes. Systems may support splitting into either two, three, four or six
nodes. When SNC mode is enabled the effective amount of L3 cache
available for allocation is divided by the number of nodes per L3.

It's possible to detect which SNC mode is active by comparing the number
of CPUs that share a cache with CPU0, with the number of CPUs on node0.

Detect SNC mode once and let other tests inherit that information.

To check if SNC detection is reliable one can check the
/sys/devices/system/cpu/offline file. If it's empty, it means all cores
are operational and the ratio should be calculated correctly. If it has
any contents, it means the detected SNC mode can't be trusted and should
be disabled.

Check if detection was not reliable due to offline cpus. If it was skip
running tests since the results couldn't be trusted.

Co-developed-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Maciej Wieczor-Retman <maciej.wieczor-retman@intel.com>
---
Changelog v6:
- Redo snc_nodes_per_l3_cache() by removing old logic and taking into
  account existance of cpus_offline_empty().
- Remove MAX_SNC since now we just divide cache cpus by node cpus
  instead of trying to estimate the ratio.
- Skip tests if snc_unreliable was set and get_vendor() returns
  ARCH_INTEL.
- Move extern int snc_unreliable; from the next patch to this one since
  it's needed here already.
- Set snc_unreliable if node or cache cpus cannot be counted for some
  reason.
- Redo patch message after code changes.

Changelog v5:
- Set snc_mode to 1 at the start of if(!snc_mode) block.
- Move all resctrlfs.c code from patch 2/2 to this one. (Reinette)
- Fix Co-developed-by tag.
- Update SNC discovery comments for the case where it gets disabled by
  being unreliable.
- Remove exclamation mark from ksft_perror() and add full file path
  instead of "offline CPUs file".
- bit map -> bitmap.
- Remove unnecessary empty line.

Changelog v4:
- Make returned value a static local variable so the function only runs
  the logic once. (Reinette)

Changelog v3:
- Add comparison between present and online cpus to test if the
  calculated SNC mode is credible. (Reinette)
- Added comment to cache size modification to better explain why it is
  needed there. (Reinette)
- Fix facts in patch message. (Reinette)
- Change snc_ways() to snc_nodes_per_l3_cache(). (Reinette)

 tools/testing/selftests/resctrl/resctrl.h     |  4 +
 .../testing/selftests/resctrl/resctrl_tests.c |  9 +-
 tools/testing/selftests/resctrl/resctrlfs.c   | 97 +++++++++++++++++++
 3 files changed, 109 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/resctrl/resctrl.h b/tools/testing/selftests/resctrl/resctrl.h
index dab1953fc7a0..38dfe5a03fcd 100644
--- a/tools/testing/selftests/resctrl/resctrl.h
+++ b/tools/testing/selftests/resctrl/resctrl.h
@@ -11,6 +11,7 @@
 #include <signal.h>
 #include <dirent.h>
 #include <stdbool.h>
+#include <ctype.h>
 #include <sys/stat.h>
 #include <sys/ioctl.h>
 #include <sys/mount.h>
@@ -156,8 +157,11 @@ struct perf_event_read {
  */
 extern volatile int *value_sink;
 
+extern int snc_unreliable;
+
 extern char llc_occup_path[1024];
 
+int snc_nodes_per_l3_cache(void);
 int get_vendor(void);
 bool check_resctrlfs_support(void);
 int filter_dmesg(void);
diff --git a/tools/testing/selftests/resctrl/resctrl_tests.c b/tools/testing/selftests/resctrl/resctrl_tests.c
index 3335af815b21..5154ffd821c4 100644
--- a/tools/testing/selftests/resctrl/resctrl_tests.c
+++ b/tools/testing/selftests/resctrl/resctrl_tests.c
@@ -118,7 +118,7 @@ static bool test_vendor_specific_check(const struct resctrl_test *test)
 
 static void run_single_test(const struct resctrl_test *test, const struct user_params *uparams)
 {
-	int ret;
+	int ret, snc_mode;
 
 	if (test->disabled)
 		return;
@@ -128,8 +128,15 @@ static void run_single_test(const struct resctrl_test *test, const struct user_p
 		return;
 	}
 
+	snc_mode = snc_nodes_per_l3_cache();
+
 	ksft_print_msg("Starting %s test ...\n", test->name);
 
+	if (snc_mode == 1 && snc_unreliable && get_vendor() == ARCH_INTEL) {
+		ksft_test_result_skip("SNC detection unreliable due to offline CPUs. Test results may not be accurate if SNC enabled.\n");
+		return;
+	}
+
 	if (test_prepare(test)) {
 		ksft_exit_fail_msg("Abnormal failure when preparing for the test\n");
 		return;
diff --git a/tools/testing/selftests/resctrl/resctrlfs.c b/tools/testing/selftests/resctrl/resctrlfs.c
index d38d6dd90be4..50561993d37c 100644
--- a/tools/testing/selftests/resctrl/resctrlfs.c
+++ b/tools/testing/selftests/resctrl/resctrlfs.c
@@ -13,6 +13,8 @@
 
 #include "resctrl.h"
 
+int snc_unreliable;
+
 static int find_resctrl_mount(char *buffer)
 {
 	FILE *mounts;
@@ -156,6 +158,90 @@ int get_domain_id(const char *resource, int cpu_no, int *domain_id)
 	return 0;
 }
 
+/*
+ * Count number of CPUs in a /sys bitmap
+ */
+static unsigned int count_sys_bitmap_bits(char *name)
+{
+	FILE *fp = fopen(name, "r");
+	int count = 0, c;
+
+	if (!fp)
+		return 0;
+
+	while ((c = fgetc(fp)) != EOF) {
+		if (!isxdigit(c))
+			continue;
+		switch (c) {
+		case 'f':
+			count++;
+		case '7': case 'b': case 'd': case 'e':
+			count++;
+		case '3': case '5': case '6': case '9': case 'a': case 'c':
+			count++;
+		case '1': case '2': case '4': case '8':
+			count++;
+		}
+	}
+	fclose(fp);
+
+	return count;
+}
+
+static bool cpus_offline_empty(void)
+{
+	char offline_cpus_str[64];
+	FILE *fp;
+
+	fp = fopen("/sys/devices/system/cpu/offline", "r");
+	if (fscanf(fp, "%s", offline_cpus_str) < 0) {
+		if (!errno) {
+			fclose(fp);
+			return 1;
+		}
+		ksft_perror("Could not read /sys/devices/system/cpu/offline");
+	}
+
+	fclose(fp);
+
+	return 0;
+}
+
+/*
+ * Detect SNC by comparing #CPUs in node0 with #CPUs sharing LLC with CPU0.
+ * If any CPUs are offline declare the detection as unreliable and skip the
+ * tests.
+ */
+int snc_nodes_per_l3_cache(void)
+{
+	int node_cpus, cache_cpus;
+	static int snc_mode;
+
+	if (!snc_mode) {
+		snc_mode = 1;
+		if (!cpus_offline_empty()) {
+			ksft_print_msg("Runtime SNC detection unreliable due to offline CPUs.\n");
+			ksft_print_msg("Setting SNC mode to disabled.\n");
+			snc_unreliable = 1;
+			return snc_mode;
+		}
+		node_cpus = count_sys_bitmap_bits("/sys/devices/system/node/node0/cpumap");
+		cache_cpus = count_sys_bitmap_bits("/sys/devices/system/cpu/cpu0/cache/index3/shared_cpu_map");
+
+		if (!node_cpus || !cache_cpus) {
+			ksft_print_msg("Could not determine Sub-NUMA Cluster mode.\n");
+			snc_unreliable = 1;
+			return snc_mode;
+		}
+		snc_mode = cache_cpus / node_cpus;
+
+		if (snc_mode > 1)
+			ksft_print_msg("SNC-%d mode discovered.\n", snc_mode);
+	}
+
+	return snc_mode;
+}
+
 /*
  * get_cache_size - Get cache size for a specified CPU
  * @cpu_no:	CPU number
@@ -211,6 +297,17 @@ int get_cache_size(int cpu_no, const char *cache_type, unsigned long *cache_size
 			break;
 	}
 
+	/*
+	 * The amount of cache represented by each bit in the masks
+	 * in the schemata file is reduced by a factor equal to SNC
+	 * nodes per L3 cache.
+	 * E.g. on a SNC-2 system with a 100MB L3 cache a test that
+	 * allocates memory from its local SNC node (default behavior
+	 * without using libnuma) will only see 50 MB llc_occupancy
+	 * with a fully populated L3 mask in the schemata file.
+	 */
+	if (cache_num == 3)
+		*cache_size /= snc_nodes_per_l3_cache();
 	return 0;
 }
 
-- 
2.47.1

[PATCH v6 2/2] selftests/resctrl: Discover SNC kernel support and adjust messages

[Maciej Wieczor-Retman]

Resctrl selftest prints a message on test failure that Sub-Numa
Clustering (SNC) could be enabled and points the user to check their BIOS
settings. No actual check is performed before printing that message so
it is not very accurate in pinpointing a problem.

When there is SNC support for kernel's resctrl subsystem and SNC is
enabled then sub node files are created for each node in the resctrlfs.
The sub node files exist in each regular node's L3 monitoring directory.
The reliable path to check for existence of sub node files is
/sys/fs/resctrl/mon_data/mon_L3_00/mon_sub_L3_00.

Add helper that checks for mon_sub_L3_00 existence.

Correct old messages to account for kernel support of SNC in
resctrl.

Signed-off-by: Maciej Wieczor-Retman <maciej.wieczor-retman@intel.com>
---
Changelog v6:
- Remove snc_unreliable checks at the end of the tests since now the
  tests are skipped if snc_unreliable was set.
- Move snc_kernel_support() back to this patch since it's not used in
  the previous patch anymore.
- Redo the patch message after code changes.

Changelog v5:
- Move all resctrlfs.c code from this patch to 1/2. (Reinette)
- Remove kernel support check and error message from CAT since it can't
  be happen.
- Remove snc checks in CAT since snc doesn't affect it here.
- Skip MBM, MBA and CMT tests if snc is unreliable.

Changelog v4:
- Change messages at the end of tests and at the start of
  run_single_test. (Reinette)
- Add messages at the end of CAT since it can also fail due to enabled
  SNC + lack of kernel support.
- Remove snc_mode global variable. (Reinette)
- Fix wrong description of snc_kernel_support(). (Reinette)
- Move call to cpus_offline_empty() into snc_nodes_per_l3_cache() so the
  whole detection flow is in one place as discussed. (Reinette)

Changelog v3:
- Change snc_ways() to snc_nodes_per_l3_cache(). (Reinette)
- Add printing the discovered SNC mode. (Reinette)
- Change method of kernel support discovery from cache sizes to
  existance of sub node files.
- Check if SNC detection is unreliable.
- Move SNC detection to only the first run_single_test() instead on
  error at the end of test runs.
- Add global value to remind user at the end of relevant tests if SNC
  detection was found to be unreliable.
- Redo the patch message after the changes.

Changelog v2:
- Move snc_ways() checks from individual tests into
  snc_kernel_support().
- Write better comment for snc_kernel_support().

 tools/testing/selftests/resctrl/cmt_test.c  |  4 +--
 tools/testing/selftests/resctrl/mba_test.c  |  2 ++
 tools/testing/selftests/resctrl/mbm_test.c  |  4 +--
 tools/testing/selftests/resctrl/resctrl.h   |  1 +
 tools/testing/selftests/resctrl/resctrlfs.c | 32 +++++++++++++++++++++
 5 files changed, 39 insertions(+), 4 deletions(-)

diff --git a/tools/testing/selftests/resctrl/cmt_test.c b/tools/testing/selftests/resctrl/cmt_test.c
index 3bbf3042fb06..d09e693dc739 100644
--- a/tools/testing/selftests/resctrl/cmt_test.c
+++ b/tools/testing/selftests/resctrl/cmt_test.c
@@ -169,8 +169,8 @@ static int cmt_run_test(const struct resctrl_test *test, const struct user_param
 		return ret;
 
 	ret = check_results(&param, span, n);
-	if (ret && (get_vendor() == ARCH_INTEL))
-		ksft_print_msg("Intel CMT may be inaccurate when Sub-NUMA Clustering is enabled. Check BIOS configuration.\n");
+	if (ret && (get_vendor() == ARCH_INTEL) && !snc_kernel_support())
+		ksft_print_msg("Kernel doesn't support Sub-NUMA Clustering but it is enabled on the system.\n");
 
 	return ret;
 }
diff --git a/tools/testing/selftests/resctrl/mba_test.c b/tools/testing/selftests/resctrl/mba_test.c
index 536d9089d2f6..c7e9adc0368f 100644
--- a/tools/testing/selftests/resctrl/mba_test.c
+++ b/tools/testing/selftests/resctrl/mba_test.c
@@ -201,6 +201,8 @@ static int mba_run_test(const struct resctrl_test *test, const struct user_param
 		return ret;
 
 	ret = check_results();
+	if (ret && (get_vendor() == ARCH_INTEL) && !snc_kernel_support())
+		ksft_print_msg("Kernel doesn't support Sub-NUMA Clustering but it is enabled on the system.\n");
 
 	return ret;
 }
diff --git a/tools/testing/selftests/resctrl/mbm_test.c b/tools/testing/selftests/resctrl/mbm_test.c
index 315b2ef3b3bc..84d8bc250539 100644
--- a/tools/testing/selftests/resctrl/mbm_test.c
+++ b/tools/testing/selftests/resctrl/mbm_test.c
@@ -160,8 +160,8 @@ static int mbm_run_test(const struct resctrl_test *test, const struct user_param
 		return ret;
 
 	ret = check_results(param.fill_buf ? param.fill_buf->buf_size : 0);
-	if (ret && (get_vendor() == ARCH_INTEL))
-		ksft_print_msg("Intel MBM may be inaccurate when Sub-NUMA Clustering is enabled. Check BIOS configuration.\n");
+	if (ret && (get_vendor() == ARCH_INTEL) && !snc_kernel_support())
+		ksft_print_msg("Kernel doesn't support Sub-NUMA Clustering but it is enabled on the system.\n");
 
 	return ret;
 }
diff --git a/tools/testing/selftests/resctrl/resctrl.h b/tools/testing/selftests/resctrl/resctrl.h
index 38dfe5a03fcd..560171f66332 100644
--- a/tools/testing/selftests/resctrl/resctrl.h
+++ b/tools/testing/selftests/resctrl/resctrl.h
@@ -202,6 +202,7 @@ void ctrlc_handler(int signum, siginfo_t *info, void *ptr);
 int signal_handler_register(const struct resctrl_test *test);
 void signal_handler_unregister(void);
 unsigned int count_bits(unsigned long n);
+int snc_kernel_support(void);
 
 void perf_event_attr_initialize(struct perf_event_attr *pea, __u64 config);
 void perf_event_initialize_read_format(struct perf_event_read *pe_read);
diff --git a/tools/testing/selftests/resctrl/resctrlfs.c b/tools/testing/selftests/resctrl/resctrlfs.c
index 50561993d37c..d0a5c0f78397 100644
--- a/tools/testing/selftests/resctrl/resctrlfs.c
+++ b/tools/testing/selftests/resctrl/resctrlfs.c
@@ -949,3 +949,35 @@ unsigned int count_bits(unsigned long n)
 
 	return count;
 }
+
+/**
+ * snc_kernel_support - Check for existence of mon_sub_L3_00 file that indicates
+ * SNC resctrl support on the kernel side.
+ *
+ * Return: 0 if not supported, 1 if SNC is disabled or SNC discovery is
+ * unreliable or SNC is both enabled and supported.
+ */
+int snc_kernel_support(void)
+{
+	char node_path[PATH_MAX];
+	struct stat statbuf;
+	int ret;
+
+	ret = snc_nodes_per_l3_cache();
+	/*
+	 * If SNC is disabled then its kernel support isn't important. If SNC
+	 * got disabled because the discovery process was unreliable the
+	 * snc_unreliable variable was set. It can be used to verify the SNC
+	 * discovery reliability elsewhere in the selftest.
+	 */
+	if (ret == 1)
+		return ret;
+
+	snprintf(node_path, sizeof(node_path), "%s/%s/%s", RESCTRL_PATH, "mon_data",
+		 "mon_L3_00/mon_sub_L3_00");
+
+	if (!stat(node_path, &statbuf))
+		return 1;
+
+	return 0;
+}
-- 
2.47.1

Re: [PATCH v6 1/2] selftests/resctrl: Adjust effective L3 cache size with SNC enabled

[Reinette Chatre]

Hi Maciej,

On 12/2/24 3:08 AM, Maciej Wieczor-Retman wrote:

> diff --git a/tools/testing/selftests/resctrl/resctrlfs.c b/tools/testing/selftests/resctrl/resctrlfs.c
> index d38d6dd90be4..50561993d37c 100644
> --- a/tools/testing/selftests/resctrl/resctrlfs.c
> +++ b/tools/testing/selftests/resctrl/resctrlfs.c
> @@ -13,6 +13,8 @@
>  
>  #include "resctrl.h"
>  
> +int snc_unreliable;
> +
>  static int find_resctrl_mount(char *buffer)
>  {
>  	FILE *mounts;
> @@ -156,6 +158,90 @@ int get_domain_id(const char *resource, int cpu_no, int *domain_id)
>  	return 0;
>  }
>  
> +/*
> + * Count number of CPUs in a /sys bitmap
> + */
> +static unsigned int count_sys_bitmap_bits(char *name)
> +{
> +	FILE *fp = fopen(name, "r");
> +	int count = 0, c;
> +
> +	if (!fp)
> +		return 0;
> +
> +	while ((c = fgetc(fp)) != EOF) {
> +		if (!isxdigit(c))
> +			continue;
> +		switch (c) {
> +		case 'f':
> +			count++;
> +		case '7': case 'b': case 'd': case 'e':
> +			count++;
> +		case '3': case '5': case '6': case '9': case 'a': case 'c':
> +			count++;
> +		case '1': case '2': case '4': case '8':
> +			count++;
> +		}
> +	}
> +	fclose(fp);
> +

running this through a syntax checker triggers a couple of complaints due to the
missing break statements. I think this can be made more robust by making use of
"fallthrough" and "break". It looks like this can be obtained by including
linux/compiler.h ... but from what I can tell care should be taken to set
the include directory _after_ includink lib.mk so that top_srcdir is set
correctly.

> +	return count;
> +}
> +
> +static bool cpus_offline_empty(void)
> +{
> +	char offline_cpus_str[64];
> +	FILE *fp;
> +
> +	fp = fopen("/sys/devices/system/cpu/offline", "r");

Please check fp before use.

> +	if (fscanf(fp, "%s", offline_cpus_str) < 0) {

This needs something equivalent to 
46058430fc5d ("selftests/resctrl: Protect against array overflow when reading strings")

> +		if (!errno) {
> +			fclose(fp);
> +			return 1;
> +		}
> +		ksft_perror("Could not read /sys/devices/system/cpu/offline");
> +	}
> +
> +	fclose(fp);
> +
> +	return 0;
> +}
> +
> +/*
> + * Detect SNC by comparing #CPUs in node0 with #CPUs sharing LLC with CPU0.
> + * If any CPUs are offline declare the detection as unreliable and skip the
> + * tests.

nit: "and skip the tests" can be dropped since the function need not make
assumption about how callers will use it.

> + */
> +int snc_nodes_per_l3_cache(void)
> +{
> +	int node_cpus, cache_cpus;
> +	static int snc_mode;
> +
> +	if (!snc_mode) {
> +		snc_mode = 1;
> +		if (!cpus_offline_empty()) {
> +			ksft_print_msg("Runtime SNC detection unreliable due to offline CPUs.\n");
> +			ksft_print_msg("Setting SNC mode to disabled.\n");
> +			snc_unreliable = 1;
> +			return snc_mode;
> +		}
> +		node_cpus = count_sys_bitmap_bits("/sys/devices/system/node/node0/cpumap");
> +		cache_cpus = count_sys_bitmap_bits("/sys/devices/system/cpu/cpu0/cache/index3/shared_cpu_map");
> +
> +		if (!node_cpus || !cache_cpus) {
> +			ksft_print_msg("Could not determine Sub-NUMA Cluster mode.\n");
> +			snc_unreliable = 1;
> +			return snc_mode;
> +		}
> +		snc_mode = cache_cpus / node_cpus;
> +
> +		if (snc_mode > 1)
> +			ksft_print_msg("SNC-%d mode discovered.\n", snc_mode);
> +	}
> +
> +	return snc_mode;
> +}
> +
>  /*
>   * get_cache_size - Get cache size for a specified CPU
>   * @cpu_no:	CPU number
> @@ -211,6 +297,17 @@ int get_cache_size(int cpu_no, const char *cache_type, unsigned long *cache_size
>  			break;
>  	}
>  
> +	/*
> +	 * The amount of cache represented by each bit in the masks
> +	 * in the schemata file is reduced by a factor equal to SNC
> +	 * nodes per L3 cache.
> +	 * E.g. on a SNC-2 system with a 100MB L3 cache a test that
> +	 * allocates memory from its local SNC node (default behavior
> +	 * without using libnuma) will only see 50 MB llc_occupancy
> +	 * with a fully populated L3 mask in the schemata file.
> +	 */
> +	if (cache_num == 3)
> +		*cache_size /= snc_nodes_per_l3_cache();
>  	return 0;
>  }
>  

Reinette

Re: [PATCH v6 2/2] selftests/resctrl: Discover SNC kernel support and adjust messages

[Reinette Chatre]

Hi Maciej,

On 12/2/24 3:08 AM, Maciej Wieczor-Retman wrote:

> diff --git a/tools/testing/selftests/resctrl/resctrlfs.c b/tools/testing/selftests/resctrl/resctrlfs.c
> index 50561993d37c..d0a5c0f78397 100644
> --- a/tools/testing/selftests/resctrl/resctrlfs.c
> +++ b/tools/testing/selftests/resctrl/resctrlfs.c
> @@ -949,3 +949,35 @@ unsigned int count_bits(unsigned long n)
>  
>  	return count;
>  }
> +
> +/**
> + * snc_kernel_support - Check for existence of mon_sub_L3_00 file that indicates
> + * SNC resctrl support on the kernel side.
> + *
> + * Return: 0 if not supported, 1 if SNC is disabled or SNC discovery is
> + * unreliable or SNC is both enabled and supported.
> + */
> +int snc_kernel_support(void)
> +{
> +	char node_path[PATH_MAX];
> +	struct stat statbuf;
> +	int ret;
> +
> +	ret = snc_nodes_per_l3_cache();
> +	/*
> +	 * If SNC is disabled then its kernel support isn't important. If SNC
> +	 * got disabled because the discovery process was unreliable the
> +	 * snc_unreliable variable was set. It can be used to verify the SNC
> +	 * discovery reliability elsewhere in the selftest.
> +	 */
> +	if (ret == 1)
> +		return ret;
> +
> +	snprintf(node_path, sizeof(node_path), "%s/%s/%s", RESCTRL_PATH, "mon_data",
> +		 "mon_L3_00/mon_sub_L3_00");

The patch looks good. I am curious why the string constants are split in two and not,
for example, a single constant of "mon_data/mon_L3_00/mon_sub_L3_00"?

> +
> +	if (!stat(node_path, &statbuf))
> +		return 1;
> +
> +	return 0;
> +}

Reinette

Re: [PATCH v6 1/2] selftests/resctrl: Adjust effective L3 cache size with SNC enabled

[Maciej Wieczor-Retman]

On 2024-12-03 at 15:26:41 -0800, Reinette Chatre wrote:
>Hi Maciej,
>
>On 12/2/24 3:08 AM, Maciej Wieczor-Retman wrote:
>
>> diff --git a/tools/testing/selftests/resctrl/resctrlfs.c b/tools/testing/selftests/resctrl/resctrlfs.c
>> index d38d6dd90be4..50561993d37c 100644
>> --- a/tools/testing/selftests/resctrl/resctrlfs.c
>> +++ b/tools/testing/selftests/resctrl/resctrlfs.c
>> @@ -13,6 +13,8 @@
>>  
>>  #include "resctrl.h"
>>  
>> +int snc_unreliable;
>> +
>>  static int find_resctrl_mount(char *buffer)
>>  {
>>  	FILE *mounts;
>> @@ -156,6 +158,90 @@ int get_domain_id(const char *resource, int cpu_no, int *domain_id)
>>  	return 0;
>>  }
>>  
>> +/*
>> + * Count number of CPUs in a /sys bitmap
>> + */
>> +static unsigned int count_sys_bitmap_bits(char *name)
>> +{
>> +	FILE *fp = fopen(name, "r");
>> +	int count = 0, c;
>> +
>> +	if (!fp)
>> +		return 0;
>> +
>> +	while ((c = fgetc(fp)) != EOF) {
>> +		if (!isxdigit(c))
>> +			continue;
>> +		switch (c) {
>> +		case 'f':
>> +			count++;
>> +		case '7': case 'b': case 'd': case 'e':
>> +			count++;
>> +		case '3': case '5': case '6': case '9': case 'a': case 'c':
>> +			count++;
>> +		case '1': case '2': case '4': case '8':
>> +			count++;
>> +		}
>> +	}
>> +	fclose(fp);
>> +
>
>running this through a syntax checker triggers a couple of complaints due to the
>missing break statements. I think this can be made more robust by making use of
>"fallthrough" and "break". It looks like this can be obtained by including
>linux/compiler.h ... but from what I can tell care should be taken to set
>the include directory _after_ includink lib.mk so that top_srcdir is set
>correctly.

Sure, I'll look into it

>
>> +	return count;
>> +}
>> +
>> +static bool cpus_offline_empty(void)
>> +{
>> +	char offline_cpus_str[64];
>> +	FILE *fp;
>> +
>> +	fp = fopen("/sys/devices/system/cpu/offline", "r");
>
>Please check fp before use.

Will do

>
>> +	if (fscanf(fp, "%s", offline_cpus_str) < 0) {
>
>This needs something equivalent to 
>46058430fc5d ("selftests/resctrl: Protect against array overflow when reading strings")

Thanks, I'll add that protection here.

>
>> +		if (!errno) {
>> +			fclose(fp);
>> +			return 1;
>> +		}
>> +		ksft_perror("Could not read /sys/devices/system/cpu/offline");
>> +	}
>> +
>> +	fclose(fp);
>> +
>> +	return 0;
>> +}
>> +
>> +/*
>> + * Detect SNC by comparing #CPUs in node0 with #CPUs sharing LLC with CPU0.
>> + * If any CPUs are offline declare the detection as unreliable and skip the
>> + * tests.
>
>nit: "and skip the tests" can be dropped since the function need not make
>assumption about how callers will use it.

Right, sorry, will remove it.

>
>> + */
>> +int snc_nodes_per_l3_cache(void)
>> +{
>> +	int node_cpus, cache_cpus;
>> +	static int snc_mode;
>> +
>> +	if (!snc_mode) {
>> +		snc_mode = 1;
>> +		if (!cpus_offline_empty()) {
>> +			ksft_print_msg("Runtime SNC detection unreliable due to offline CPUs.\n");
>> +			ksft_print_msg("Setting SNC mode to disabled.\n");
>> +			snc_unreliable = 1;
>> +			return snc_mode;
>> +		}
>> +		node_cpus = count_sys_bitmap_bits("/sys/devices/system/node/node0/cpumap");
>> +		cache_cpus = count_sys_bitmap_bits("/sys/devices/system/cpu/cpu0/cache/index3/shared_cpu_map");
>> +
>> +		if (!node_cpus || !cache_cpus) {
>> +			ksft_print_msg("Could not determine Sub-NUMA Cluster mode.\n");
>> +			snc_unreliable = 1;
>> +			return snc_mode;
>> +		}
>> +		snc_mode = cache_cpus / node_cpus;
>> +
>> +		if (snc_mode > 1)
>> +			ksft_print_msg("SNC-%d mode discovered.\n", snc_mode);
>> +	}
>> +
>> +	return snc_mode;
>> +}
>> +
>>  /*
>>   * get_cache_size - Get cache size for a specified CPU
>>   * @cpu_no:	CPU number
>> @@ -211,6 +297,17 @@ int get_cache_size(int cpu_no, const char *cache_type, unsigned long *cache_size
>>  			break;
>>  	}
>>  
>> +	/*
>> +	 * The amount of cache represented by each bit in the masks
>> +	 * in the schemata file is reduced by a factor equal to SNC
>> +	 * nodes per L3 cache.
>> +	 * E.g. on a SNC-2 system with a 100MB L3 cache a test that
>> +	 * allocates memory from its local SNC node (default behavior
>> +	 * without using libnuma) will only see 50 MB llc_occupancy
>> +	 * with a fully populated L3 mask in the schemata file.
>> +	 */
>> +	if (cache_num == 3)
>> +		*cache_size /= snc_nodes_per_l3_cache();
>>  	return 0;
>>  }
>>  
>
>Reinette
>

-- 
Kind regards
Maciej Wieczór-Retman

Re: [PATCH v6 2/2] selftests/resctrl: Discover SNC kernel support and adjust messages

[Maciej Wieczor-Retman]

On 2024-12-03 at 15:30:56 -0800, Reinette Chatre wrote:
>Hi Maciej,
>
>On 12/2/24 3:08 AM, Maciej Wieczor-Retman wrote:
>
>> diff --git a/tools/testing/selftests/resctrl/resctrlfs.c b/tools/testing/selftests/resctrl/resctrlfs.c
>> index 50561993d37c..d0a5c0f78397 100644
>> --- a/tools/testing/selftests/resctrl/resctrlfs.c
>> +++ b/tools/testing/selftests/resctrl/resctrlfs.c
>> @@ -949,3 +949,35 @@ unsigned int count_bits(unsigned long n)
>>  
>>  	return count;
>>  }
>> +
>> +/**
>> + * snc_kernel_support - Check for existence of mon_sub_L3_00 file that indicates
>> + * SNC resctrl support on the kernel side.
>> + *
>> + * Return: 0 if not supported, 1 if SNC is disabled or SNC discovery is
>> + * unreliable or SNC is both enabled and supported.
>> + */
>> +int snc_kernel_support(void)
>> +{
>> +	char node_path[PATH_MAX];
>> +	struct stat statbuf;
>> +	int ret;
>> +
>> +	ret = snc_nodes_per_l3_cache();
>> +	/*
>> +	 * If SNC is disabled then its kernel support isn't important. If SNC
>> +	 * got disabled because the discovery process was unreliable the
>> +	 * snc_unreliable variable was set. It can be used to verify the SNC
>> +	 * discovery reliability elsewhere in the selftest.
>> +	 */
>> +	if (ret == 1)
>> +		return ret;
>> +
>> +	snprintf(node_path, sizeof(node_path), "%s/%s/%s", RESCTRL_PATH, "mon_data",
>> +		 "mon_L3_00/mon_sub_L3_00");
>
>The patch looks good. I am curious why the string constants are split in two and not,
>for example, a single constant of "mon_data/mon_L3_00/mon_sub_L3_00"?

I think I introduced this inconsistency by accident when trying to follow the
example of other snprintf calls. You're right, there's not reason it can't be
just "mon_data/mon_L3_00/mon_sub_L3_00". I'll fix it in the next version. Thanks :)

>
>> +
>> +	if (!stat(node_path, &statbuf))
>> +		return 1;
>> +
>> +	return 0;
>> +}
>
>Reinette

-- 
Kind regards
Maciej Wieczór-Retman