[RFC PATCH v2 04/10] perf workqueue: add threadpool execute and wait functions

[Riccardo Mancini <rickyman7@gmail.com>]

This patch adds:
 - threadpool__execute: assigns a task to the threads to execute
   asynchronously.
 - threadpool__wait: waits for the task to complete on all threads.
Furthermore, testing for these new functions is added.

This patch completes the threadpool.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
---
 tools/perf/tests/workqueue.c           |  85 +++++++++++++-
 tools/perf/util/workqueue/threadpool.c | 146 ++++++++++++++++++++++++-
 tools/perf/util/workqueue/threadpool.h |   1 +
 3 files changed, 230 insertions(+), 2 deletions(-)

diff --git a/tools/perf/tests/workqueue.c b/tools/perf/tests/workqueue.c
index 767e4fb60be4b190..87bf8fc71c346653 100644
--- a/tools/perf/tests/workqueue.c
+++ b/tools/perf/tests/workqueue.c
@@ -1,14 +1,60 @@
 // SPDX-License-Identifier: GPL-2.0
+#include <stdlib.h>
 #include <linux/kernel.h>
 #include <linux/err.h>
+#include <linux/zalloc.h>
 #include "tests.h"
 #include "util/debug.h"
 #include "util/workqueue/threadpool.h"
 
+#define DUMMY_FACTOR 100000
+#define N_DUMMY_WORK_SIZES 7
+
 struct threadpool_test_args_t {
 	int pool_size;
 };
 
+struct test_task {
+	struct task_struct task;
+	int n_threads;
+	int *array;
+};
+
+/**
+ * dummy_work - calculates DUMMY_FACTOR * (idx % N_DUMMY_WORK_SIZES) inefficiently
+ *
+ * This function uses modulus to create work items of different sizes.
+ */
+static void dummy_work(int idx)
+{
+	volatile int prod = 0;	/* prevent possible compiler optimizations */
+	int k = idx % N_DUMMY_WORK_SIZES;
+	int i, j;
+
+	for (i = 0; i < DUMMY_FACTOR; i++)
+		for (j = 0; j < k; j++)
+			prod++;
+
+	pr_debug3("dummy: %d * %d = %d\n", DUMMY_FACTOR, k, prod);
+}
+
+static void test_task_fn1(int tidx, struct task_struct *task)
+{
+	struct test_task *mtask = container_of(task, struct test_task, task);
+
+	dummy_work(tidx);
+	mtask->array[tidx] = tidx+1;
+}
+
+static void test_task_fn2(int tidx, struct task_struct *task)
+{
+	struct test_task *mtask = container_of(task, struct test_task, task);
+
+	dummy_work(tidx);
+	mtask->array[tidx] = tidx*2;
+}
+
+
 static int __threadpool__prepare(struct threadpool **pool, int pool_size)
 {
 	int ret;
@@ -39,22 +85,59 @@ static int __threadpool__teardown(struct threadpool *pool)
 	return TEST_OK;
 }
 
+static int __threadpool__exec_wait(struct threadpool *pool,
+				struct task_struct *task)
+{
+	int ret = threadpool__execute(pool, task);
+
+	TEST_ASSERT_VAL("threadpool execute failure", ret == 0);
+	TEST_ASSERT_VAL("threadpool is not executing", threadpool__is_busy(pool));
+
+	ret = threadpool__wait(pool);
+	TEST_ASSERT_VAL("threadpool wait failure", ret == 0);
+	TEST_ASSERT_VAL("waited threadpool is not ready", threadpool__is_ready(pool));
+
+	return TEST_OK;
+}
 
 static int __test__threadpool(void *_args)
 {
 	struct threadpool_test_args_t *args = _args;
 	struct threadpool *pool;
-	int ret;
+	int ret, i;
+	struct test_task task;
+
+	task.task.fn = test_task_fn1;
+	task.n_threads = args->pool_size;
+	task.array = calloc(args->pool_size, sizeof(*task.array));
+	TEST_ASSERT_VAL("calloc failure", task.array);
 
 	ret = __threadpool__prepare(&pool, args->pool_size);
 	if (ret)
 		goto out;
 
+	ret = __threadpool__exec_wait(pool, &task.task);
+	if (ret)
+		goto out;
+
+	for (i = 0; i < args->pool_size; i++)
+		TEST_ASSERT_VAL("failed array check (1)", task.array[i] == i+1);
+
+	task.task.fn = test_task_fn2;
+
+	ret = __threadpool__exec_wait(pool, &task.task);
+	if (ret)
+		goto out;
+
+	for (i = 0; i < args->pool_size; i++)
+		TEST_ASSERT_VAL("failed array check (2)", task.array[i] == 2*i);
+
 	ret = __threadpool__teardown(pool);
 	if (ret)
 		goto out;
 
 out:
+	free(task.array);
 	return ret;
 }
 
diff --git a/tools/perf/util/workqueue/threadpool.c b/tools/perf/util/workqueue/threadpool.c
index 850ef7e110566536..33320dd0fb44fe38 100644
--- a/tools/perf/util/workqueue/threadpool.c
+++ b/tools/perf/util/workqueue/threadpool.c
@@ -26,6 +26,7 @@ static inline pid_t gettid(void)
 enum threadpool_status {
 	THREADPOOL_STATUS__STOPPED,		/* no threads */
 	THREADPOOL_STATUS__READY,		/* threads are ready but idle */
+	THREADPOOL_STATUS__BUSY,		/* threads are busy */
 	THREADPOOL_STATUS__ERROR,		/* errors */
 	THREADPOOL_STATUS__MAX
 };
@@ -33,6 +34,7 @@ enum threadpool_status {
 static const char * const threadpool_status_tags[] = {
 	"stopped",
 	"ready",
+	"busy",
 	"error"
 };
 
@@ -181,6 +183,28 @@ static int threadpool__terminate_thread(struct threadpool_entry *thread)
 	return threadpool__wait_thread(thread);
 }
 
+/**
+ * threadpool__wake_thread - send wake msg to @thread
+ *
+ * This function does not wait for the thread to actually wake
+ * NB: call only from main thread!
+ */
+static int threadpool__wake_thread(struct threadpool_entry *thread)
+{
+	int res;
+	enum threadpool_msg msg = THREADPOOL_MSG__WAKE;
+
+	res = writen(thread->pipes.cmd[1], &msg, sizeof(msg));
+	if (res < 0) {
+		pr_debug2("threadpool: error sending wake msg: %s\n", strerror(errno));
+		return -THREADPOOL_ERROR__WRITEPIPE;
+	}
+
+	pr_debug2("threadpool: sent wake msg %s to tid=%d\n",
+		threadpool_msg_tags[msg], thread->tid);
+	return 0;
+}
+
 /**
  * threadpool_entry__function - send ack to main thread
  */
@@ -249,6 +273,15 @@ static void *threadpool_entry__function(void *args)
 
 		if (cmd == THREADPOOL_MSG__STOP)
 			break;
+
+		if (!thread->pool->current_task) {
+			pr_debug("threadpool[%d]: received wake without task\n",
+				thread->tid);
+			break;
+		}
+
+		pr_debug("threadpool[%d]: executing task\n", thread->tid);
+		thread->pool->current_task->fn(thread->idx, thread->pool->current_task);
 	}
 
 	pr_debug2("threadpool[%d]: exit\n", thread->tid);
@@ -502,10 +535,14 @@ static int __threadpool__stop(struct threadpool *pool)
  * threadpool__stop - stop all threads in the pool.
  *
  * This function blocks waiting for ack from all threads.
+ * If the pool was busy, it will first wait for the task to finish.
  */
 int threadpool__stop(struct threadpool *pool)
 {
-	int ret;
+	int ret = threadpool__wait(pool);
+
+	if (ret)
+		return ret;
 
 	if (pool->status != THREADPOOL_STATUS__READY) {
 		pr_debug2("threadpool: stopping not ready pool\n");
@@ -528,3 +565,110 @@ bool threadpool__is_ready(struct threadpool *pool)
 {
 	return pool->status == THREADPOOL_STATUS__READY;
 }
+
+/**
+ * __threadpool__execute - execute @task on all threads of the @pool
+ *
+ * NB: use threadpool__execute. This function does not change @pool->status.
+ */
+static int __threadpool__execute(struct threadpool *pool, struct task_struct *task)
+{
+	int t, ret;
+
+	pool->current_task = task;
+
+	for (t = 0; t < pool->nr_threads; t++) {
+		ret = threadpool__wake_thread(&pool->threads[t]);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+/**
+ * threadpool__execute - execute @task on all threads of the @pool
+ *
+ * The task will run asynchronously wrt the main thread.
+ * The task can be waited with threadpool__wait.
+ *
+ * NB: make sure the pool is ready before calling this, since no queueing is
+ *     performed. If you need queueing, have a look at the workqueue.
+ */
+int threadpool__execute(struct threadpool *pool, struct task_struct *task)
+{
+	int ret;
+
+	if (pool->status != THREADPOOL_STATUS__READY) {
+		pr_debug2("threadpool: executing on not ready pool\n");
+		return -THREADPOOL_ERROR__NOTALLOWED;
+	}
+
+	ret = __threadpool__execute(pool, task);
+	if (ret) {
+		pool->status = THREADPOOL_STATUS__ERROR;
+		return ret;
+	}
+	pool->status = THREADPOOL_STATUS__BUSY;
+	return 0;
+}
+
+/**
+ * __threadpool__wait - wait until all threads in @pool are done
+ *
+ * NB: use threadpool__wait. This function does not change @pool->status.
+ */
+static int __threadpool__wait(struct threadpool *pool)
+{
+	int t, err = 0, ret;
+
+	for (t = 0; t < pool->nr_threads; t++) {
+		ret = threadpool__wait_thread(&pool->threads[t]);
+		if (ret)
+			err = ret;
+	}
+
+	pool->current_task = NULL;
+	return err;
+}
+
+/**
+ * threadpool__wait - wait until all threads in @pool are done
+ *
+ * This function will wait for all threads to finish execution and send their
+ * ack message.
+ *
+ * NB: call only from main thread!
+ */
+int threadpool__wait(struct threadpool *pool)
+{
+	int ret;
+
+	switch (pool->status) {
+	case THREADPOOL_STATUS__BUSY:
+		break;
+	case THREADPOOL_STATUS__READY:
+		return 0;
+	default:
+	case THREADPOOL_STATUS__STOPPED:
+	case THREADPOOL_STATUS__ERROR:
+	case THREADPOOL_STATUS__MAX:
+		return -THREADPOOL_ERROR__NOTALLOWED;
+	}
+
+	ret = __threadpool__wait(pool);
+	if (ret) {
+		pool->status = THREADPOOL_STATUS__ERROR;
+		return ret;
+	}
+	pool->status = THREADPOOL_STATUS__READY;
+	return 0;
+}
+
+/**
+ * threadpool__is_busy - check if the pool is busy
+ */
+int threadpool__is_busy(struct threadpool *pool)
+{
+	return pool->status == THREADPOOL_STATUS__BUSY;
+}
diff --git a/tools/perf/util/workqueue/threadpool.h b/tools/perf/util/workqueue/threadpool.h
index 7d56e5450fac605b..f28b3856d30e91e5 100644
--- a/tools/perf/util/workqueue/threadpool.h
+++ b/tools/perf/util/workqueue/threadpool.h
@@ -22,6 +22,7 @@ extern int threadpool__wait(struct threadpool *pool);
 
 extern int threadpool__size(struct threadpool *pool);
 extern bool threadpool__is_ready(struct threadpool *pool);
+extern int threadpool__is_busy(struct threadpool *pool);
 
 /* Error management */
 #define THREADPOOL_STRERR_BUFSIZE (128+STRERR_BUFSIZE)
-- 
2.31.1