Skip to contentSkip to conversation

tst_fuzzy_sync: refine code and comment

PATCH In review Li Wang 2 participants 8 messages Dec 22, 2024 – Jan 8, 2025

lore
Share & tools
lore

Li Wang

author
#0 lore raw

PATCH 1/3 tst_fuzzy_sync: refine code and comment

I just polished some things preventing me from understanding the test logic. Hopefully, it makes things easier to read.

---
 lib/newlib_tests/tst_fuzzy_sync01.c | 22 ++++++------
 lib/newlib_tests/tst_fuzzy_sync02.c | 53 ++++++++++++++++++++++++-----
 2 files changed, 55 insertions(+), 20 deletions(-)
lib/newlib_tests/tst_fuzzy_sync01.c +11 -11
index d0748958c..d510bd2dd 100644
@@ -10,7 +10,7 @@
  *
  * We make the simplifying assumptions that:
  * - Each thread contains a single contiguous critical section.
- * - The threads only interact through a single variable.
+ * - The threads only interact through a single variable(H: Hit).
  * - The various timings are constant except for variations introduced
  *   by the environment.
  *
@@ -50,9 +50,9 @@
  * range is required.
  *
  * When entering their critical sections, both threads increment the
- * 'c' counter variable atomically. They both also increment it when
- * leaving their critical sections. We record the value of 'c' when A
- * increments it. From the recorded values of 'c' we can deduce if the
+ * 'H' counter variable atomically. They both also increment it when
+ * leaving their critical sections. We record the value of 'H' when A
+ * increments it. From the recorded values of 'H' we can deduce if the
  * critical sections overlap and their ordering.
  *
  * 	Start (cs)	| End (ct)	| Ordering
@@ -90,7 +90,7 @@ struct race {
 	const struct window b;
 };
 
-static int c;
+static int H;
 static struct tst_fzsync_pair pair;
 
 static const struct race races[] = {
@@ -162,15 +162,15 @@ static void *worker(void *v)
 	const struct window b = races[i].b;
 
 	while (tst_fzsync_run_b(&pair)) {
-		if (tst_atomic_load(&c))
+		if (tst_atomic_load(&H))
 			tst_brk(TBROK, "Counter should now be zero");
 
 		tst_fzsync_start_race_b(&pair);
 		delay(b.critical_s);
 
-		tst_atomic_add_return(1, &c);
+		tst_atomic_add_return(1, &H);
 		delay(b.critical_t);
-		tst_atomic_add_return(1, &c);
+		tst_atomic_add_return(1, &H);
 
 		delay(b.return_t);
 		tst_fzsync_end_race_b(&pair);
@@ -192,9 +192,9 @@ static void run(unsigned int i)
 		tst_fzsync_start_race_a(&pair);
 		delay(a.critical_s);
 
-		cs = tst_atomic_add_return(1, &c);
+		cs = tst_atomic_add_return(1, &H);
 		delay(a.critical_t);
-		ct = tst_atomic_add_return(1, &c);
+		ct = tst_atomic_add_return(1, &H);
 
 		delay(a.return_t);
 		tst_fzsync_end_race_a(&pair);
@@ -206,7 +206,7 @@ static void run(unsigned int i)
 		else
 			critical++;
 
-		r = tst_atomic_add_return(-4, &c);
+		r = tst_atomic_add_return(-4, &H);
 		if (r)
 			tst_brk(TBROK, "cs = %d, ct = %d, r = %d", cs, ct, r);
 
lib/newlib_tests/tst_fuzzy_sync02.c +44 -9

99 lines not on this page · raw message

Signature · 5 lines
2.47.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#1 lore raw

PATCH 2/3 tst_fuzzy_sync: ingnore test result if loop termination occurs

This patch updates the behavior of tst_fuzzy_sync tests to address scenarios where test results may be unreliable due to resource constraints.

Specifically, if pair->exit is set (indicating exceeded loop termination), the test result is ignored, and a TCONF status is returned. This ensures invalid results are not falsely reported as failures or successes.

tst_fuzzy_sync.h:654: TINFO: Exceeded execution loops, requesting exit /lib/newlib_tests/tst_fuzzy_sync02.c:164: TFAIL: 2| =:15

---
 lib/newlib_tests/tst_fuzzy_sync01.c | 21 +++++++++++++++++++++
 lib/newlib_tests/tst_fuzzy_sync02.c | 21 +++++++++++++++++++++
 2 files changed, 42 insertions(+)
lib/newlib_tests/tst_fuzzy_sync01.c +21 -0
index d510bd2dd..c8c195859 100644
@@ -212,10 +212,31 @@ static void run(unsigned int i)
 
 		if (critical > 100) {
 			tst_fzsync_pair_cleanup(&pair);
+			tst_atomic_store(0, &pair.exit);
 			break;
 		}
 	}
 
+	/*
+	 * If `pair->exit` is true, the test may fail to meet expected
+	 * results due to resource constraints in shared CI environments
+	 * (e.g., GitHub Actions). Limited control over CPU allocation
+	 * can cause delays or interruptions in CPU time slices due to
+	 * contention with other jobs.
+	 *
+	 * Binding the test to a single CPU core (e.g., via `taskset -c 0`)
+	 * can worsen this by increasing contention, leading to performance
+	 * degradation and premature loop termination.
+	 *
+	 * To ensure valid and reliable results in scenarios (e.g., HW, VM, CI),
+	 * it is best to ingore test result when loop termination occurs,
+	 * avoiding unnecessary false postive.
+	 */
+	if (pair.exit) {
+		tst_res(TCONF, "Test may not be able to generate a valid result");
+		return;
+	}
+
 	tst_res(critical > 50 ? TPASS : TFAIL,
 		"acs:%-2d act:%-2d art:%-2d | =:%-4d -:%-4d +:%-4d",
 		a.critical_s, a.critical_t, a.return_t,
lib/newlib_tests/tst_fuzzy_sync02.c +21 -0
index f95394371..ad19a675e 100644
@@ -192,10 +192,31 @@ static void run(unsigned int i)
 
 		if (critical > 100) {
 			tst_fzsync_pair_cleanup(&pair);
+			tst_atomic_store(0, &pair.exit);
 			break;
 		}
 	}
 
+	/*
+	 * If `pair->exit` is true, the test may fail to meet expected
+	 * results due to resource constraints in shared CI environments
+	 * (e.g., GitHub Actions). Limited control over CPU allocation
+	 * can cause delays or interruptions in CPU time slices due to
+	 * contention with other jobs.
+	 *
+	 * Binding the test to a single CPU core (e.g., via `taskset -c 0`)
+	 * can worsen this by increasing contention, leading to performance
+	 * degradation and premature loop termination.
+	 *
+	 * To ensure valid and reliable results in scenarios (e.g., HW, VM, CI),
+	 * it is best to ingore test result when loop termination occurs,
+	 * avoiding unnecessary false postive.
+	 */
+	if (pair.exit) {
+		tst_res(TCONF, "Test may not be able to generate a valid result");
+		return;
+	}
+
 	tst_res(critical > 50 ? TPASS : TFAIL, "%d| =:%-4d", i, critical);
 }
 
Signature · 5 lines
2.47.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 2/3
#2 lore raw

Re: [LTP] [PATCH 2/3] tst_fuzzy_sync: ingnore test result if loop termination occurs

Hi Li,

Quoted 9 lines

This patch updates the behavior of tst_fuzzy_sync tests to address scenarios where test results may be unreliable due to resource constraints.

Specifically, if pair->exit is set (indicating exceeded loop termination), the test result is ignored, and a TCONF status is returned. This ensures invalid results are not falsely reported as failures or successes.

tst_fuzzy_sync.h:654: TINFO: Exceeded execution loops, requesting exit /lib/newlib_tests/tst_fuzzy_sync02.c:164: TFAIL: 2| =:15

LGTM, good catch.

Below are noted typos, could you please fix them before merge?

Quoted 22 lines
---
 lib/newlib_tests/tst_fuzzy_sync01.c | 21 +++++++++++++++++++++
 lib/newlib_tests/tst_fuzzy_sync02.c | 21 +++++++++++++++++++++
 2 files changed, 42 insertions(+)
lib/newlib_tests/tst_fuzzy_sync01.c +4 -0
index d510bd2dd..c8c195859 100644
@@ -212,10 +212,31 @@ static void run(unsigned int i)
 
 		if (critical > 100) {
 			tst_fzsync_pair_cleanup(&pair);
+			tst_atomic_store(0, &pair.exit);
 			break;
 		}
 	}
 
+	/*
+	 * If `pair->exit` is true, the test may fail to meet expected
+	 * results due to resource constraints in shared CI environments

lib/newlib_tests/tst_fuzzy_sync01.c static void run(unsigned int i) · line 232

Quoted 10 lines
lib/newlib_tests/tst_fuzzy_sync01.c +10 -0
+	 * (e.g., GitHub Actions). Limited control over CPU allocation
+	 * can cause delays or interruptions in CPU time slices due to
+	 * contention with other jobs.
+	 *
+	 * Binding the test to a single CPU core (e.g., via `taskset -c 0`)
+	 * can worsen this by increasing contention, leading to performance
+	 * degradation and premature loop termination.
+	 *
+	 * To ensure valid and reliable results in scenarios (e.g., HW, VM, CI),
+	 * it is best to ingore test result when loop termination occurs,

s/ingore/ignore/

lib/newlib_tests/tst_fuzzy_sync01.c static void run(unsigned int i) · line 233

Quoted 1 line
lib/newlib_tests/tst_fuzzy_sync01.c +1 -0
+	 * avoiding unnecessary false postive.

s/postive/positive/

Quoted 25 lines
lib/newlib_tests/tst_fuzzy_sync01.c +6 -0
+	 */
+	if (pair.exit) {
+		tst_res(TCONF, "Test may not be able to generate a valid result");
+		return;
+	}
+
 	tst_res(critical > 50 ? TPASS : TFAIL,
 		"acs:%-2d act:%-2d art:%-2d | =:%-4d -:%-4d +:%-4d",
 		a.critical_s, a.critical_t, a.return_t,
lib/newlib_tests/tst_fuzzy_sync02.c +4 -0
index f95394371..ad19a675e 100644
@@ -192,10 +192,31 @@ static void run(unsigned int i)
 
 		if (critical > 100) {
 			tst_fzsync_pair_cleanup(&pair);
+			tst_atomic_store(0, &pair.exit);
 			break;
 		}
 	}
 
+	/*
+	 * If `pair->exit` is true, the test may fail to meet expected
+	 * results due to resource constraints in shared CI environments

lib/newlib_tests/tst_fuzzy_sync02.c static void run(unsigned int i) · line 212

Quoted 10 lines
lib/newlib_tests/tst_fuzzy_sync02.c +10 -0
+	 * (e.g., GitHub Actions). Limited control over CPU allocation
+	 * can cause delays or interruptions in CPU time slices due to
+	 * contention with other jobs.
+	 *
+	 * Binding the test to a single CPU core (e.g., via `taskset -c 0`)
+	 * can worsen this by increasing contention, leading to performance
+	 * degradation and premature loop termination.
+	 *
+	 * To ensure valid and reliable results in scenarios (e.g., HW, VM, CI),
+	 * it is best to ingore test result when loop termination occurs,

s/ingore/ignore/

lib/newlib_tests/tst_fuzzy_sync02.c static void run(unsigned int i) · line 213

Quoted 1 line
lib/newlib_tests/tst_fuzzy_sync02.c +1 -0
+	 * avoiding unnecessary false postive.

s/postive/positive/

Also there is original typo from Richie:
s/Emperically/Empirically/

Kind regards,
Petr

Quoted 5 lines
lib/newlib_tests/tst_fuzzy_sync02.c +5 -0
+	 */
+	if (pair.exit) {
+		tst_res(TCONF, "Test may not be able to generate a valid result");
+		return;
+	}
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#3 lore raw

PATCH 3/3 newlib_test: enable all tst_fuzzy_sync tests

In reply to Li Wang

---

Notes:

    This patch based on Peter's work:
    Message-ID: <[email protected]>
    
    CI: https://github.com/wangli5665/ltp/actions/runs/12452324779/
 lib/newlib_tests/runtest.sh | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
lib/newlib_tests/runtest.sh +3 -1
index 66aaa9361..b00965b16 100755
@@ -2,7 +2,7 @@
 # Copyright (c) 2021-2024 Petr Vorel <[email protected]>
 
 # TODO "unknown failure, exit code": test_assert test08 tst_cgroup01 tst_cgroup02 tst_res_flags variant
-# TODO TFAIL: tst_fuzzy_sync01 tst_fuzzy_sync02 test_macros0[1-6] test23 test26
+# TODO TFAIL: test_macros0[1-6] test23 test26
 # TODO TBROK: test_exec_child test_kconfig01 test_kconfig02 tst_needs_cmds04 tst_needs_cmds05 test_runtime02 test01 test02 test03 test04 test06 test11 test13 test22 test25 tst_safe_fileops
 # TODO TWARN: test_guarded_buf test14 tst_capability01 tst_print_result
 LTP_C_API_TESTS="${LTP_C_API_TESTS:-
@@ -20,6 +20,8 @@ tst_bool_expr
 tst_capability02
 tst_device
 tst_expiration_timer
+tst_fuzzy_sync01
+tst_fuzzy_sync02
 tst_fuzzy_sync03
 tst_needs_cmds0[1-36-8]
 tst_res_hexd
Signature · 5 lines
2.47.1


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 3/3
#4 lore raw

Re: [LTP] [PATCH 3/3] newlib_test: enable all tst_fuzzy_sync tests

Hi Li,

Quoted 14 lines

Notes:

    This patch based on Peter's work:
    Message-ID: <[email protected]>

    CI: https://github.com/wangli5665/ltp/actions/runs/12452324779/
 lib/newlib_tests/runtest.sh | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
lib/newlib_tests/runtest.sh +0 -0
index 66aaa9361..b00965b16 100755

lib/newlib_tests/runtest.sh line 5

Quoted 4 lines
lib/newlib_tests/runtest.sh +0 -1
@@ -2,7 +2,7 @@
 # Copyright (c) 2021-2024 Petr Vorel <[email protected]>
 
 # TODO "unknown failure, exit code": test_assert test08 tst_cgroup01 tst_cgroup02 tst_res_flags variant
-# TODO TFAIL: tst_fuzzy_sync01 tst_fuzzy_sync02 test_macros0[1-6] test23 test26

Ah, I should have mark it as tst_fuzzy_sync0[12]

Anyway, feel free to merge before I merge my patchset, I have no problem to rebase mine.

Kind regards,
Petr

Quoted 13 lines
lib/newlib_tests/runtest.sh +3 -0
+# TODO TFAIL: test_macros0[1-6] test23 test26
 # TODO TBROK: test_exec_child test_kconfig01 test_kconfig02 tst_needs_cmds04 tst_needs_cmds05 test_runtime02 test01 test02 test03 test04 test06 test11 test13 test22 test25 tst_safe_fileops
 # TODO TWARN: test_guarded_buf test14 tst_capability01 tst_print_result
 LTP_C_API_TESTS="${LTP_C_API_TESTS:-
@@ -20,6 +20,8 @@ tst_bool_expr
 tst_capability02
 tst_device
 tst_expiration_timer
+tst_fuzzy_sync01
+tst_fuzzy_sync02
 tst_fuzzy_sync03
 tst_needs_cmds0[1-36-8]
 tst_res_hexd
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Reviewed-by on v1 1/3
#5 lore raw
In reply to Li Wang

Hi Li,

Quoted 2 lines

I just polished some things preventing me from understanding the test logic. Hopefully, it makes things easier to read.

LGTM, thanks!

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
#6 lore raw
In reply to Li Wang

Hi Li,

would you mind to change before merge trailing "\*/" to "*/" (removing "/") in both files and extra blank line with "*" in tst_fuzzy_sync02.c?

Kind regards,
Petr

Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
author
#7 lore raw
Quoted 6 lines · Petr Vorel

Hi Li,

would you mind to change before merge trailing "\*/" to "*/" (removing "/") in both files and extra blank line with "*" in tst_fuzzy_sync02.c?

Thanks for the review, I merge the patch set (with your suggestions).

Signature · 5 lines
Regards,
Li Wang

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp