Merge commit 'refs/top-bases/linux-kernel-disk' into linux-kernel-disk

ChangeLog

@@ -54,9 +54,6 @@
 	* module/zcommon/zfs_prop.c: Export new quota related symbols.
 	* lib/libzfs/libzfs_util.c: Increase buffer size for nvlist which
 	is needed for large configurations.
-	* lib/libzpool/kernel.c: Reimplement user kthreads based on pthread
-	thread specific data.  This resolves previous scalabily concerns about
-	the cost of calling curthread which previously required a list walk.
 
 	* Bug fixes:
 	* lib/libspl/include/sys/zfs_debug.h: Removed duplicate file.

cmd/ztest/ztest.c

@@ -141,6 +141,7 @@ typedef struct ztest_args {
 	objset_t	*za_os;
 	zilog_t		*za_zilog;
 	kthread_t	*za_thread;
+	kt_did_t	za_threadid;
 	uint64_t	za_instance;
 	uint64_t	za_random;
 	uint64_t	za_diroff;
@@ -156,7 +157,6 @@ typedef struct ztest_args {
 	ztest_block_tag_t za_wbt;
 	dmu_object_info_t za_doi;
 	dmu_buf_t	*za_dbuf;
-	boolean_t	za_exited;
 } ztest_args_t;
 
 typedef void ztest_func_t(ztest_args_t *);
@@ -253,8 +253,6 @@ typedef struct ztest_shared {
 	kmutex_t		zs_sync_lock[ZTEST_SYNC_LOCKS];
 	uint64_t		zs_seq[ZTEST_SYNC_LOCKS];
 	ztest_cb_list_t		zs_cb_list;
-	kmutex_t		zs_thr_lock;
-	kcondvar_t		zs_thr_cv;
 } ztest_shared_t;
 
 static char ztest_dev_template[] = "%s/%s.%llua";
@@ -266,7 +264,6 @@ static int ztest_dump_core = 1;
 
 static uint64_t metaslab_sz;
 static boolean_t ztest_exiting;
-static boolean_t resume_thr_exited;
 
 extern uint64_t metaslab_gang_bang;
 extern uint64_t metaslab_df_alloc_threshold;
@@ -3822,8 +3819,6 @@ ztest_resume_thread(void *arg)
 		ztest_resume(spa);
 	}
 
-	resume_thr_exited = B_TRUE;
-
 	thread_exit();
 	return (NULL);
 }
@@ -3889,13 +3884,6 @@ ztest_thread(void *arg)
 			break;
 	}
 
-	mutex_enter(&zs->zs_thr_lock);
-	za->za_exited = B_TRUE;
-	mutex_exit(&zs->zs_thr_lock);
-
-	/* Announce that the thread has finished */
-	cv_broadcast(&zs->zs_thr_cv);
-
 	thread_exit();
 	return (NULL);
 }
@@ -3912,14 +3900,13 @@ ztest_run(char *pool)
 	spa_t *spa;
 	char name[100];
 	kthread_t *resume_thread;
+	kt_did_t resume_id;
 
 	ztest_exiting = B_FALSE;
 
 	mutex_init(&zs->zs_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
 	rw_init(&zs->zs_name_lock, NULL, RW_DEFAULT, NULL);
 	mutex_init(&zs->zs_cb_list.zcl_callbacks_lock,NULL,MUTEX_DEFAULT,NULL);
-	mutex_init(&zs->zs_thr_lock, NULL, MUTEX_DEFAULT, NULL);
-	cv_init(&zs->zs_thr_cv, NULL, CV_DEFAULT, NULL);
 
 	list_create(&zs->zs_cb_list.zcl_callbacks, sizeof (ztest_cb_data_t),
 	    offsetof(ztest_cb_data_t, zcd_node));
@@ -3992,9 +3979,9 @@ ztest_run(char *pool)
 	/*
 	 * Create a thread to periodically resume suspended I/O.
 	 */
-	resume_thr_exited = B_FALSE;
 	VERIFY3P((resume_thread = thread_create(NULL, 0, ztest_resume_thread,
-	    spa, TS_RUN, NULL, 0, 0)), !=, NULL);
+	         spa, THR_BOUND, NULL, 0, 0)), !=, NULL);
+	resume_id = resume_thread->t_tid;
 
 	/*
 	 * Verify that we can safely inquire about about any object,
@@ -4070,18 +4057,13 @@ ztest_run(char *pool)
 			za[d].za_zilog = zil_open(za[d].za_os, NULL);
 		}
 
-		za[t].za_exited = B_FALSE;
-
 		VERIFY3P((za[t].za_thread = thread_create(NULL, 0, ztest_thread,
-		    &za[t], TS_RUN, NULL, 0, 0)), !=, NULL);
+		         &za[t], THR_BOUND, NULL, 0, 0)), !=, NULL);
+		za[t].za_threadid = za[t].za_thread->t_tid;
 	}
 
 	while (--t >= 0) {
-		mutex_enter(&zs->zs_thr_lock);
-		while (!za[t].za_exited)
-			cv_wait(&zs->zs_thr_cv, &zs->zs_thr_lock);
-		mutex_exit(&zs->zs_thr_lock);
-
+		VERIFY(thread_join(za[t].za_threadid, NULL, NULL) == 0);
 		if (t < zopt_datasets) {
 			zil_close(za[t].za_zilog);
 			dmu_objset_close(za[t].za_os);
@@ -4120,11 +4102,7 @@ ztest_run(char *pool)
 
 	/* Kill the resume thread */
 	ztest_exiting = B_TRUE;
-
-	/* Wait for the resume thread to exit */
-	while (!resume_thr_exited)
-		(void) poll(NULL, 0, 200);
-
+	VERIFY(thread_join(resume_id, NULL, NULL) == 0);
 	ztest_resume(spa);
 
 	/*
@@ -4140,8 +4118,6 @@ ztest_run(char *pool)
 
 	list_destroy(&zs->zs_cb_list.zcl_callbacks);
 
-	cv_destroy(&zs->zs_thr_cv);
-	mutex_destroy(&zs->zs_thr_lock);
 	mutex_destroy(&zs->zs_cb_list.zcl_callbacks_lock);
 	rw_destroy(&zs->zs_name_lock);
 	mutex_destroy(&zs->zs_vdev_lock);

lib/libzpool/include/sys/zfs_context.h

@@ -198,34 +198,34 @@ _NOTE(CONSTCOND) } while (0)
 /*
  * Threads
  */
+#define THR_BOUND		0x00000001
 #define TS_RUN			0x00000002
 
-#define STACK_SIZE 8192 /* x86/x64 */
-
-typedef void (*thread_func_t)(void);
-typedef void (*thread_func_arg_t)(void *);
+typedef void (*thread_func_t)(void *);
 typedef pthread_t kt_did_t;
 
 typedef struct kthread {
+        list_node_t	t_node;
 	kt_did_t	t_tid;
-	thread_func_t	t_func;
-	void *		t_arg;
+	pthread_attr_t	t_attr;
 } kthread_t;
 
-/* XXX tsd_create()/tsd_destroy() missing */
 #define	tsd_get(key)		pthread_getspecific(key)
 #define	tsd_set(key, val)	pthread_setspecific(key, val)
 #define	curthread		zk_thread_current()
 #define thread_exit		zk_thread_exit
 #define thread_create(stk, stksize, func, arg, len, pp, state, pri)	\
-	zk_thread_create(stk, stksize, (thread_func_t) func, arg, len,  \
-	    NULL, state, pri)
+	zk_thread_create(stk, stksize, (thread_func_t)func, arg,	\
+			 len, NULL, state, pri)
+#define thread_join(tid, dtid, status)					\
+	zk_thread_join(tid, dtid, status)
 
 extern kthread_t *zk_thread_current(void);
 extern void zk_thread_exit(void);
 extern kthread_t *zk_thread_create(caddr_t stk, size_t  stksize,
 	thread_func_t func, void *arg, size_t len,
 	void *pp, int state, pri_t pri);
+extern int zk_thread_join(kt_did_t tid, kthread_t *dtid, void **status);
 
 #define	issig(why)	(FALSE)
 #define	ISSIG(thr, why)	(FALSE)
@@ -357,7 +357,7 @@ extern taskq_t	*taskq_create(const char *, int, pri_t, int, int, uint_t);
 extern taskqid_t taskq_dispatch(taskq_t *, task_func_t, void *, uint_t);
 extern void	taskq_destroy(taskq_t *);
 extern void	taskq_wait(taskq_t *);
-extern int	taskq_member(taskq_t *, kthread_t *);
+extern int	taskq_member(taskq_t *, void *);
 extern void	system_taskq_init(void);
 extern void	system_taskq_fini(void);