Skip to contentSkip to conversation

enable lcores test on Windows

PATCH Superseded Tyler Retzlaff 3 participants 12 messages Dec 6, 2022 – Dec 21, 2022

lore
Share & tools
lore
#0 lore raw

cover enable lcores test on Windows

Enable the lcores test on Windows instead of skipping it.

Two bugs are fixed to allow this test to build, run & pass.
* Mark memory configuration complete during rte_eal_init()
* Use rte thread api to get a proper implementation of thread join.

Tyler Retzlaff (2):
eal: add missing call marking memory config complete
test: enable lcores test on Windows

 app/test/test_lcores.c | 28 +++++++++++++---------------
 lib/eal/windows/eal.c  |  3 +++
 2 files changed, 16 insertions(+), 15 deletions(-)
Signature · 1 line
1.8.3.1
#1 lore raw

PATCH 1/2 eal: add missing call marking memory config complete

Memory configuration was not being marked as completed add the missing call to rte_eal_init() for Windows.

---
 lib/eal/windows/eal.c | 3 +++
 1 file changed, 3 insertions(+)
lib/eal/windows/eal.c +3 -0
index adb929a..56fadc7 100644
@@ -462,6 +462,9 @@ enum rte_proc_type_t
 	 */
 	rte_eal_mp_remote_launch(sync_func, NULL, SKIP_MAIN);
 	rte_eal_mp_wait_lcore();
+
+	eal_mcfg_complete();
+
 	return fctret;
 }
 
Signature · 1 line
1.8.3.1
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] eal: add missing call marking memory config complete

Quoted 13 lines · On Mon, 5 Dec 2022 16:39:28 -0800 Tyler Retzlaff…

Memory configuration was not being marked as completed add the missing call to rte_eal_init() for Windows.

---
 lib/eal/windows/eal.c | 3 +++
 1 file changed, 3 insertions(+)
lib/eal/windows/eal.c +0 -0
index adb929a..56fadc7 100644

lib/eal/windows/eal.c enum rte_proc_type_t · line 470

Quoted 9 lines
lib/eal/windows/eal.c +3 -0
@@ -462,6 +462,9 @@ enum rte_proc_type_t
 	 */
 	rte_eal_mp_remote_launch(sync_func, NULL, SKIP_MAIN);
 	rte_eal_mp_wait_lcore();
+
+	eal_mcfg_complete();
+
 	return fctret;
 }
 

Should this have a Fixes tag?

#3 lore raw

Re: [PATCH 1/2] eal: add missing call marking memory config complete

In reply to Tyler Retzlaff
Quoted 5 lines · Tyler Retzlaff

Memory configuration was not being marked as completed add the missing call to rte_eal_init() for Windows.

The change is ok.

The commitlog misses some context: I guess the point is to let rte_thread_register work on Windows.
So we can mark this change as:

Fixes 5c307ba2a5b1 ("eal: register non-EAL threads as lcores")

WDYT?

In any case, this change won't fix anything related to mp. If mp is supported one day on Windows, other calls to eal_mcfg* would be required.

Signature · 1 line
David Marchand
#4 lore raw

Re: [PATCH 1/2] eal: add missing call marking memory config complete

Quoted 6 lines · David Marchand
Quoted 4 lines · Tyler Retzlaff

Memory configuration was not being marked as completed add the missing call to rte_eal_init() for Windows.

Quoted 6 lines
Quoted 1 line

The change is ok.

The commitlog misses some context: I guess the point is to let rte_thread_register work on Windows.

oh i guess i better provide a better message. the reason is i want to run the test_ctrl_thread() (in test_lcores.c) which is an api available on windows. maybe when i add the replacement control thread creation it should be moved to / live in test_threads.c?

it is nice as a side-effect it did pick up that the mcfg hadn't been marked as done. (though i understand that's not the purpose of the test).

i'll have another whack at changing the log.

Quoted 4 lines

So we can mark this change as:

Fixes 5c307ba2a5b1 ("eal: register non-EAL threads as lcores")

WDYT?

yep, v2 will fix the message & the Fixes tag.

Quoted 5 lines

In any case, this change won't fix anything related to mp. If mp is supported one day on Windows, other calls to eal_mcfg* would be required.

agreed.

Quoted 3 lines
Signature · 2 lines
-- 
David Marchand
#5 lore raw

PATCH 2/2 test: enable lcores test on Windows

In reply to Tyler Retzlaff

Stop using pthread and convert the test to use EAL thread APIs. Because the EAL thread APIs provide more than just a stub for thread join on Windows the tests now pass and need not be skipped.

---
 app/test/test_lcores.c | 28 +++++++++++++---------------
 1 file changed, 13 insertions(+), 15 deletions(-)
app/test/test_lcores.c +13 -15
index a6bb412..5b43aa5 100644
@@ -8,17 +8,18 @@
 #include <rte_common.h>
 #include <rte_errno.h>
 #include <rte_lcore.h>
+#include <rte_thread.h>
 
 #include "test.h"
 
 struct thread_context {
 	enum { Thread_INIT, Thread_ERROR, Thread_DONE } state;
 	bool lcore_id_any;
-	pthread_t id;
+	rte_thread_t id;
 	unsigned int *registered_count;
 };
 
-static void *thread_loop(void *arg)
+static uint32_t thread_loop(void *arg)
 {
 	struct thread_context *t = arg;
 	unsigned int lcore_id;
@@ -55,7 +56,7 @@ static void *thread_loop(void *arg)
 	if (t->state != Thread_ERROR)
 		t->state = Thread_DONE;
 
-	return NULL;
+	return 0;
 }
 
 static int
@@ -77,7 +78,7 @@ static void *thread_loop(void *arg)
 		t->state = Thread_INIT;
 		t->registered_count = &registered_count;
 		t->lcore_id_any = false;
-		if (pthread_create(&t->id, NULL, thread_loop, t) != 0)
+		if (rte_thread_create(&t->id, NULL, thread_loop, t) != 0)
 			break;
 		non_eal_threads_count++;
 	}
@@ -96,7 +97,7 @@ static void *thread_loop(void *arg)
 	t->state = Thread_INIT;
 	t->registered_count = &registered_count;
 	t->lcore_id_any = true;
-	if (pthread_create(&t->id, NULL, thread_loop, t) == 0) {
+	if (rte_thread_create(&t->id, NULL, thread_loop, t) == 0) {
 		non_eal_threads_count++;
 		printf("non-EAL threads count: %u\n", non_eal_threads_count);
 		while (__atomic_load_n(&registered_count, __ATOMIC_ACQUIRE) !=
@@ -110,7 +111,7 @@ static void *thread_loop(void *arg)
 	ret = 0;
 	for (i = 0; i < non_eal_threads_count; i++) {
 		t = &thread_contexts[i];
-		pthread_join(t->id, NULL);
+		rte_thread_join(t->id, NULL);
 		if (t->state != Thread_DONE)
 			ret = -1;
 	}
@@ -262,7 +263,7 @@ struct limit_lcore_context {
 	t->state = Thread_INIT;
 	t->registered_count = &registered_count;
 	t->lcore_id_any = false;
-	if (pthread_create(&t->id, NULL, thread_loop, t) != 0)
+	if (rte_thread_create(&t->id, NULL, thread_loop, t) != 0)
 		goto cleanup_threads;
 	non_eal_threads_count++;
 	while (__atomic_load_n(&registered_count, __ATOMIC_ACQUIRE) !=
@@ -285,7 +286,7 @@ struct limit_lcore_context {
 	t->state = Thread_INIT;
 	t->registered_count = &registered_count;
 	t->lcore_id_any = true;
-	if (pthread_create(&t->id, NULL, thread_loop, t) != 0)
+	if (rte_thread_create(&t->id, NULL, thread_loop, t) != 0)
 		goto cleanup_threads;
 	non_eal_threads_count++;
 	while (__atomic_load_n(&registered_count, __ATOMIC_ACQUIRE) !=
@@ -309,7 +310,7 @@ struct limit_lcore_context {
 	ret = 0;
 	for (i = 0; i < non_eal_threads_count; i++) {
 		t = &thread_contexts[i];
-		pthread_join(t->id, NULL);
+		rte_thread_join(t->id, NULL);
 		if (t->state != Thread_DONE)
 			ret = -1;
 	}
@@ -330,7 +331,7 @@ struct limit_lcore_context {
 	__atomic_store_n(&registered_count, 0, __ATOMIC_RELEASE);
 	for (i = 0; i < non_eal_threads_count; i++) {
 		t = &thread_contexts[i];
-		pthread_join(t->id, NULL);
+		rte_thread_join(t->id, NULL);
 	}
 error:
 	if (handle[1] != NULL)
@@ -361,7 +362,7 @@ static void *ctrl_thread_loop(void *arg)
 	/* Create one control thread */
 	t = &ctrl_thread_context;
 	t->state = Thread_INIT;
-	if (rte_ctrl_thread_create(&t->id, "test_ctrl_threads",
+	if (rte_ctrl_thread_create((pthread_t *)&t->id, "test_ctrl_threads",
 					NULL, ctrl_thread_loop, t) != 0)
 		return -1;
 
@@ -369,7 +370,7 @@ static void *ctrl_thread_loop(void *arg)
 	 * This also acts as the barrier such that the memory operations
 	 * in control thread are visible to this thread.
 	 */
-	pthread_join(t->id, NULL);
+	rte_thread_join(t->id, NULL);
 
 	/* Check if the control thread set the correct state */
 	if (t->state != Thread_DONE)
@@ -384,9 +385,6 @@ static void *ctrl_thread_loop(void *arg)
 	unsigned int eal_threads_count = 0;
 	unsigned int i;
 
-	if (RTE_EXEC_ENV_IS_WINDOWS)
-		return TEST_SKIPPED;
-
 	for (i = 0; i < RTE_MAX_LCORE; i++) {
 		if (!rte_lcore_has_role(i, ROLE_OFF))
 			eal_threads_count++;
Signature · 1 line
1.8.3.1
Acked-by on v1 2/2
#6 lore raw

Re: [PATCH 2/2] test: enable lcores test on Windows

Quoted 5 lines · On Mon, 5 Dec 2022 16:39:29 -0800 Tyler Retzlaff…

Stop using pthread and convert the test to use EAL thread APIs. Because the EAL thread APIs provide more than just a stub for thread join on Windows the tests now pass and need not be skipped.

This makes sense to cover the API's even if not on Windows.

#7 lore raw
In reply to Tyler Retzlaff

hey folks,

a quick ping looking for review here. simple fix and increases test coverage. fairly straight forward.

thanks!

Quoted 16 lines · Tyler Retzlaff

Enable the lcores test on Windows instead of skipping it.

Two bugs are fixed to allow this test to build, run & pass.
* Mark memory configuration complete during rte_eal_init()
* Use rte thread api to get a proper implementation of thread join.

Tyler Retzlaff (2):
eal: add missing call marking memory config complete
test: enable lcores test on Windows

 app/test/test_lcores.c | 28 +++++++++++++---------------
 lib/eal/windows/eal.c  |  3 +++
 2 files changed, 16 insertions(+), 15 deletions(-)
Signature · 1 line
1.8.3.1
#8 lore raw

v2 cover Enable the lcores test on Windows instead of skipping it.

In reply to Tyler Retzlaff

Two bugs are fixed to allow this test to build, run & pass.
* Mark memory configuration complete during rte_eal_init()
* Use rte thread api to get a proper implementation of thread join.

v2:
* update commit message to clarify why this test is beneficial on Windows.
* add missing Fixes tag

Tyler Retzlaff (2):
eal: add missing call marking memory config complete
test: enable lcores test on Windows

 app/test/test_lcores.c | 28 +++++++++++++---------------
 lib/eal/windows/eal.c  |  3 +++
 2 files changed, 16 insertions(+), 15 deletions(-)
Signature · 1 line
1.8.3.1
#9 lore raw

PATCH v2 1/2 eal: add missing call marking memory config complete

Memory configuration was not being marked as completed add the missing call to rte_eal_init() for Windows.

Allows rte_thread_register to work on Windows and lcores_autotest to be built and run Windows which also exercises the rte_ctrl_thread_create API on Windows.

Fixes 5c307ba2a5b1 ("eal: register non-EAL threads as lcores")
---
 lib/eal/windows/eal.c | 3 +++
 1 file changed, 3 insertions(+)
lib/eal/windows/eal.c +3 -0
index adb929a..56fadc7 100644
@@ -462,6 +462,9 @@ enum rte_proc_type_t
 	 */
 	rte_eal_mp_remote_launch(sync_func, NULL, SKIP_MAIN);
 	rte_eal_mp_wait_lcore();
+
+	eal_mcfg_complete();
+
 	return fctret;
 }
 
Signature · 1 line
1.8.3.1
#10 lore raw

PATCH v2 2/2 test: enable lcores test on Windows

In reply to Tyler Retzlaff

Stop using pthread and convert the test to use EAL thread APIs. Because the EAL thread APIs provide more than just a stub for thread join on Windows the tests now pass and need not be skipped.

---
 app/test/test_lcores.c | 28 +++++++++++++---------------
 1 file changed, 13 insertions(+), 15 deletions(-)
app/test/test_lcores.c +13 -15
index a6bb412..5b43aa5 100644
@@ -8,17 +8,18 @@
 #include <rte_common.h>
 #include <rte_errno.h>
 #include <rte_lcore.h>
+#include <rte_thread.h>
 
 #include "test.h"
 
 struct thread_context {
 	enum { Thread_INIT, Thread_ERROR, Thread_DONE } state;
 	bool lcore_id_any;
-	pthread_t id;
+	rte_thread_t id;
 	unsigned int *registered_count;
 };
 
-static void *thread_loop(void *arg)
+static uint32_t thread_loop(void *arg)
 {
 	struct thread_context *t = arg;
 	unsigned int lcore_id;
@@ -55,7 +56,7 @@ static void *thread_loop(void *arg)
 	if (t->state != Thread_ERROR)
 		t->state = Thread_DONE;
 
-	return NULL;
+	return 0;
 }
 
 static int
@@ -77,7 +78,7 @@ static void *thread_loop(void *arg)
 		t->state = Thread_INIT;
 		t->registered_count = &registered_count;
 		t->lcore_id_any = false;
-		if (pthread_create(&t->id, NULL, thread_loop, t) != 0)
+		if (rte_thread_create(&t->id, NULL, thread_loop, t) != 0)
 			break;
 		non_eal_threads_count++;
 	}
@@ -96,7 +97,7 @@ static void *thread_loop(void *arg)
 	t->state = Thread_INIT;
 	t->registered_count = &registered_count;
 	t->lcore_id_any = true;
-	if (pthread_create(&t->id, NULL, thread_loop, t) == 0) {
+	if (rte_thread_create(&t->id, NULL, thread_loop, t) == 0) {
 		non_eal_threads_count++;
 		printf("non-EAL threads count: %u\n", non_eal_threads_count);
 		while (__atomic_load_n(&registered_count, __ATOMIC_ACQUIRE) !=
@@ -110,7 +111,7 @@ static void *thread_loop(void *arg)
 	ret = 0;
 	for (i = 0; i < non_eal_threads_count; i++) {
 		t = &thread_contexts[i];
-		pthread_join(t->id, NULL);
+		rte_thread_join(t->id, NULL);
 		if (t->state != Thread_DONE)
 			ret = -1;
 	}
@@ -262,7 +263,7 @@ struct limit_lcore_context {
 	t->state = Thread_INIT;
 	t->registered_count = &registered_count;
 	t->lcore_id_any = false;
-	if (pthread_create(&t->id, NULL, thread_loop, t) != 0)
+	if (rte_thread_create(&t->id, NULL, thread_loop, t) != 0)
 		goto cleanup_threads;
 	non_eal_threads_count++;
 	while (__atomic_load_n(&registered_count, __ATOMIC_ACQUIRE) !=
@@ -285,7 +286,7 @@ struct limit_lcore_context {
 	t->state = Thread_INIT;
 	t->registered_count = &registered_count;
 	t->lcore_id_any = true;
-	if (pthread_create(&t->id, NULL, thread_loop, t) != 0)
+	if (rte_thread_create(&t->id, NULL, thread_loop, t) != 0)
 		goto cleanup_threads;
 	non_eal_threads_count++;
 	while (__atomic_load_n(&registered_count, __ATOMIC_ACQUIRE) !=
@@ -309,7 +310,7 @@ struct limit_lcore_context {
 	ret = 0;
 	for (i = 0; i < non_eal_threads_count; i++) {
 		t = &thread_contexts[i];
-		pthread_join(t->id, NULL);
+		rte_thread_join(t->id, NULL);
 		if (t->state != Thread_DONE)
 			ret = -1;
 	}
@@ -330,7 +331,7 @@ struct limit_lcore_context {
 	__atomic_store_n(&registered_count, 0, __ATOMIC_RELEASE);
 	for (i = 0; i < non_eal_threads_count; i++) {
 		t = &thread_contexts[i];
-		pthread_join(t->id, NULL);
+		rte_thread_join(t->id, NULL);
 	}
 error:
 	if (handle[1] != NULL)
@@ -361,7 +362,7 @@ static void *ctrl_thread_loop(void *arg)
 	/* Create one control thread */
 	t = &ctrl_thread_context;
 	t->state = Thread_INIT;
-	if (rte_ctrl_thread_create(&t->id, "test_ctrl_threads",
+	if (rte_ctrl_thread_create((pthread_t *)&t->id, "test_ctrl_threads",
 					NULL, ctrl_thread_loop, t) != 0)
 		return -1;
 
@@ -369,7 +370,7 @@ static void *ctrl_thread_loop(void *arg)
 	 * This also acts as the barrier such that the memory operations
 	 * in control thread are visible to this thread.
 	 */
-	pthread_join(t->id, NULL);
+	rte_thread_join(t->id, NULL);
 
 	/* Check if the control thread set the correct state */
 	if (t->state != Thread_DONE)
@@ -384,9 +385,6 @@ static void *ctrl_thread_loop(void *arg)
 	unsigned int eal_threads_count = 0;
 	unsigned int i;
 
-	if (RTE_EXEC_ENV_IS_WINDOWS)
-		return TEST_SKIPPED;
-
 	for (i = 0; i < RTE_MAX_LCORE; i++) {
 		if (!rte_lcore_has_role(i, ROLE_OFF))
 			eal_threads_count++;
Signature · 1 line
1.8.3.1
Acked-by on the v2 series
#11 lore raw

Re: [PATCH v2 0/2] Enable the lcores test on Windows instead of skipping it.

In reply to Tyler Retzlaff
Quoted 9 lines · Tyler Retzlaff

Two bugs are fixed to allow this test to build, run & pass.
* Mark memory configuration complete during rte_eal_init()
* Use rte thread api to get a proper implementation of thread join.

v2:
* update commit message to clarify why this test is beneficial on Windows.
* add missing Fixes tag

Quoted 7 lines

Tyler Retzlaff (2):
eal: add missing call marking memory config complete
test: enable lcores test on Windows

 app/test/test_lcores.c | 28 +++++++++++++---------------
 lib/eal/windows/eal.c  |  3 +++
 2 files changed, 16 insertions(+), 15 deletions(-)

For the series,

I added review tags from Stephen from v1.
Series applied, thanks.

Signature · 1 line
David Marchand