Update core ZFS code from build 121 to build 141.

lib/libzpool/include/sys/zfs_context.h

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2010 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -75,6 +75,7 @@ extern "C" {
 #include <sys/u8_textprep.h>
 #include <sys/sysevent/eventdefs.h>
 #include <sys/sysevent/dev.h>
+#include <sys/sunddi.h>
 
 /*
  * Debugging
@@ -105,21 +106,27 @@ extern void vpanic(const char *, __va_list);
 
 #define	fm_panic	panic
 
+extern int aok;
+
 /* This definition is copied from assert.h. */
 #if defined(__STDC__)
 #if __STDC_VERSION__ - 0 >= 199901L
-#define	verify(EX) (void)((EX) || \
+#define	zverify(EX) (void)((EX) || (aok) || \
 	(__assert_c99(#EX, __FILE__, __LINE__, __func__), 0))
 #else
-#define	verify(EX) (void)((EX) || (__assert(#EX, __FILE__, __LINE__), 0))
+#define	zverify(EX) (void)((EX) || (aok) || \
+	(__assert(#EX, __FILE__, __LINE__), 0))
 #endif /* __STDC_VERSION__ - 0 >= 199901L */
 #else
-#define	verify(EX) (void)((EX) || (_assert("EX", __FILE__, __LINE__), 0))
+#define	zverify(EX) (void)((EX) || (aok) || \
+	(_assert("EX", __FILE__, __LINE__), 0))
 #endif	/* __STDC__ */
 
 
-#define	VERIFY	verify
-#define	ASSERT	assert
+#define	VERIFY	zverify
+#define	ASSERT	zverify
+#undef	assert
+#define	assert	zverify
 
 extern void __assert(const char *, const char *, int);
 
@@ -130,7 +137,7 @@ extern void __assert(const char *, const char *, int);
 #define	VERIFY3_IMPL(LEFT, OP, RIGHT, TYPE) do { \
 	const TYPE __left = (TYPE)(LEFT); \
 	const TYPE __right = (TYPE)(RIGHT); \
-	if (!(__left OP __right)) { \
+	if (!(__left OP __right) && (!aok)) { \
 		char *__buf = alloca(256); \
 		(void) snprintf(__buf, 256, "%s %s %s (0x%llx %s 0x%llx)", \
 			#LEFT, #OP, #RIGHT, \
@@ -196,6 +203,18 @@ typedef struct kthread kthread_t;
 #define	thread_create(stk, stksize, func, arg, len, pp, state, pri)	\
 	zk_thread_create(func, arg)
 #define	thread_exit() thr_exit(NULL)
+#define	thread_join(t)	panic("libzpool cannot join threads")
+
+#define	newproc(f, a, cid, pri, ctp, pid)	(ENOSYS)
+
+/* in libzpool, p0 exists only to have its address taken */
+struct proc {
+	uintptr_t	this_is_never_used_dont_dereference_it;
+};
+
+extern struct proc p0;
+
+#define	PS_NONE		-1
 
 extern kthread_t *zk_thread_create(void (*func)(), void *arg);
 
@@ -318,20 +337,27 @@ typedef void (task_func_t)(void *);
 #define	TASKQ_PREPOPULATE	0x0001
 #define	TASKQ_CPR_SAFE		0x0002	/* Use CPR safe protocol */
 #define	TASKQ_DYNAMIC		0x0004	/* Use dynamic thread scheduling */
-#define	TASKQ_THREADS_CPU_PCT	0x0008	/* Use dynamic thread scheduling */
+#define	TASKQ_THREADS_CPU_PCT	0x0008	/* Scale # threads by # cpus */
+#define	TASKQ_DC_BATCH		0x0010	/* Mark threads as batch */
 
 #define	TQ_SLEEP	KM_SLEEP	/* Can block for memory */
 #define	TQ_NOSLEEP	KM_NOSLEEP	/* cannot block for memory; may fail */
-#define	TQ_NOQUEUE	0x02	/* Do not enqueue if can't dispatch */
+#define	TQ_NOQUEUE	0x02		/* Do not enqueue if can't dispatch */
+#define	TQ_FRONT	0x08		/* Queue in front */
 
 extern taskq_t *system_taskq;
 
 extern taskq_t	*taskq_create(const char *, int, pri_t, int, int, uint_t);
+#define	taskq_create_proc(a, b, c, d, e, p, f) \
+	    (taskq_create(a, b, c, d, e, f))
+#define	taskq_create_sysdc(a, b, d, e, p, dc, f) \
+	    (taskq_create(a, b, maxclsyspri, d, e, f))
 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 *, void *);
 extern void	system_taskq_init(void);
+extern void	system_taskq_fini(void);
 
 #define	XVA_MAPSIZE	3
 #define	XVA_MAGIC	0x78766174
@@ -345,6 +371,7 @@ typedef struct vnode {
 	char		*v_path;
 } vnode_t;
 
+#define	AV_SCANSTAMP_SZ	32		/* length of anti-virus scanstamp */
 
 typedef struct xoptattr {
 	timestruc_t	xoa_createtime;	/* Create time of file */
@@ -360,6 +387,8 @@ typedef struct xoptattr {
 	uint8_t		xoa_opaque;
 	uint8_t		xoa_av_quarantined;
 	uint8_t		xoa_av_modified;
+	uint8_t		xoa_av_scanstamp[AV_SCANSTAMP_SZ];
+	uint8_t		xoa_reparse;
 } xoptattr_t;
 
 typedef struct vattr {
@@ -406,9 +435,11 @@ typedef struct vsecattr {
 
 #define	CRCREAT		0
 
+extern int fop_getattr(vnode_t *vp, vattr_t *vap);
+
 #define	VOP_CLOSE(vp, f, c, o, cr, ct)	0
 #define	VOP_PUTPAGE(vp, of, sz, fl, cr, ct)	0
-#define	VOP_GETATTR(vp, vap, fl, cr, ct)  ((vap)->va_size = (vp)->v_size, 0)
+#define	VOP_GETATTR(vp, vap, fl, cr, ct)  fop_getattr((vp), (vap));
 
 #define	VOP_FSYNC(vp, f, cr, ct)	fsync((vp)->v_fd)
 
@@ -433,13 +464,18 @@ extern vnode_t *rootdir;
 /*
  * Random stuff
  */
-#define	lbolt	(gethrtime() >> 23)
-#define	lbolt64	(gethrtime() >> 23)
+#define	ddi_get_lbolt()		(gethrtime() >> 23)
+#define	ddi_get_lbolt64()	(gethrtime() >> 23)
 #define	hz	119	/* frequency when using gethrtime() >> 23 for lbolt */
 
 extern void delay(clock_t ticks);
 
 #define	gethrestime_sec() time(NULL)
+#define	gethrestime(t) \
+	do {\
+		(t)->tv_sec = gethrestime_sec();\
+		(t)->tv_nsec = 0;\
+	} while (0);
 
 #define	max_ncpus	64
 
@@ -490,6 +526,9 @@ typedef struct callb_cpr {
 #define	zone_dataset_visible(x, y)	(1)
 #define	INGLOBALZONE(z)			(1)
 
+extern char *kmem_asprintf(const char *fmt, ...);
+#define	strfree(str) kmem_free((str), strlen(str)+1)
+
 /*
  * Hostname information
  */
@@ -497,6 +536,9 @@ extern char hw_serial[];	/* for userland-emulated hostid access */
 extern int ddi_strtoul(const char *str, char **nptr, int base,
     unsigned long *result);
 
+extern int ddi_strtoull(const char *str, char **nptr, int base,
+    u_longlong_t *result);
+
 /* ZFS Boot Related stuff. */
 
 struct _buf {

lib/libzpool/kernel.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2010 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -42,6 +42,7 @@
  * Emulation of kernel services in userland.
  */
 
+int aok;
 uint64_t physmem;
 vnode_t *rootdir = (vnode_t *)0xabcd1234;
 char hw_serial[HW_HOSTID_LEN];
@@ -50,6 +51,9 @@ struct utsname utsname = {
 	"userland", "libzpool", "1", "1", "na"
 };
 
+/* this only exists to have its address taken */
+struct proc p0;
+
 /*
  * =========================================================================
  * threads
@@ -269,7 +273,7 @@ cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
 	clock_t delta;
 
 top:
-	delta = abstime - lbolt;
+	delta = abstime - ddi_get_lbolt();
 	if (delta <= 0)
 		return (-1);
 
@@ -444,6 +448,24 @@ vn_close(vnode_t *vp)
 	umem_free(vp, sizeof (vnode_t));
 }
 
+/*
+ * At a minimum we need to update the size since vdev_reopen()
+ * will no longer call vn_openat().
+ */
+int
+fop_getattr(vnode_t *vp, vattr_t *vap)
+{
+	struct stat64 st;
+
+	if (fstat64(vp->v_fd, &st) == -1) {
+		close(vp->v_fd);
+		return (errno);
+	}
+
+	vap->va_size = st.st_size;
+	return (0);
+}
+
 #ifdef ZFS_DEBUG
 
 /*
@@ -754,6 +776,17 @@ ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
 	return (0);
 }
 
+int
+ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
+{
+	char *end;
+
+	*result = strtoull(str, &end, base);
+	if (*result == 0)
+		return (errno);
+	return (0);
+}
+
 /*
  * =========================================================================
  * kernel emulation setup & teardown
@@ -779,7 +812,8 @@ kernel_init(int mode)
 	dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
 	    (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
 
-	(void) snprintf(hw_serial, sizeof (hw_serial), "%ld", gethostid());
+	(void) snprintf(hw_serial, sizeof (hw_serial), "%ld",
+	    (mode & FWRITE) ? gethostid() : 0);
 
 	VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1);
 	VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1);
@@ -794,6 +828,8 @@ kernel_fini(void)
 {
 	spa_fini();
 
+	system_taskq_fini();
+
 	close(random_fd);
 	close(urandom_fd);
 
@@ -884,3 +920,27 @@ ksiddomain_rele(ksiddomain_t *ksid)
 	spa_strfree(ksid->kd_name);
 	umem_free(ksid, sizeof (ksiddomain_t));
 }
+
+/*
+ * Do not change the length of the returned string; it must be freed
+ * with strfree().
+ */
+char *
+kmem_asprintf(const char *fmt, ...)
+{
+	int size;
+	va_list adx;
+	char *buf;
+
+	va_start(adx, fmt);
+	size = vsnprintf(NULL, 0, fmt, adx) + 1;
+	va_end(adx);
+
+	buf = kmem_alloc(size, KM_SLEEP);
+
+	va_start(adx, fmt);
+	size = vsnprintf(buf, size, fmt, adx);
+	va_end(adx);
+
+	return (buf);
+}

lib/libzpool/taskq.c

@@ -19,7 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2010 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -49,6 +49,8 @@ struct taskq {
 	int		tq_nalloc;
 	int		tq_minalloc;
 	int		tq_maxalloc;
+	kcondvar_t	tq_maxalloc_cv;
+	int		tq_maxalloc_wait;
 	task_t		*tq_freelist;
 	task_t		tq_task;
 };
@@ -57,26 +59,36 @@ static task_t *
 task_alloc(taskq_t *tq, int tqflags)
 {
 	task_t *t;
+	int rv;
 
-	if ((t = tq->tq_freelist) != NULL && tq->tq_nalloc >= tq->tq_minalloc) {
+again:	if ((t = tq->tq_freelist) != NULL && tq->tq_nalloc >= tq->tq_minalloc) {
 		tq->tq_freelist = t->task_next;
 	} else {
-		mutex_exit(&tq->tq_lock);
 		if (tq->tq_nalloc >= tq->tq_maxalloc) {
-			if (!(tqflags & KM_SLEEP)) {
-				mutex_enter(&tq->tq_lock);
+			if (!(tqflags & KM_SLEEP))
 				return (NULL);
-			}
+
 			/*
 			 * We don't want to exceed tq_maxalloc, but we can't
 			 * wait for other tasks to complete (and thus free up
 			 * task structures) without risking deadlock with
 			 * the caller.  So, we just delay for one second
-			 * to throttle the allocation rate.
+			 * to throttle the allocation rate. If we have tasks
+			 * complete before one second timeout expires then
+			 * taskq_ent_free will signal us and we will
+			 * immediately retry the allocation.
 			 */
-			delay(hz);
+			tq->tq_maxalloc_wait++;
+			rv = cv_timedwait(&tq->tq_maxalloc_cv,
+			    &tq->tq_lock, ddi_get_lbolt() + hz);
+			tq->tq_maxalloc_wait--;
+			if (rv > 0)
+				goto again;		/* signaled */
 		}
+		mutex_exit(&tq->tq_lock);
+
 		t = kmem_alloc(sizeof (task_t), tqflags);
+
 		mutex_enter(&tq->tq_lock);
 		if (t != NULL)
 			tq->tq_nalloc++;
@@ -96,6 +108,9 @@ task_free(taskq_t *tq, task_t *t)
 		kmem_free(t, sizeof (task_t));
 		mutex_enter(&tq->tq_lock);
 	}
+
+	if (tq->tq_maxalloc_wait)
+		cv_signal(&tq->tq_maxalloc_cv);
 }
 
 taskqid_t
@@ -114,8 +129,13 @@ taskq_dispatch(taskq_t *tq, task_func_t func, void *arg, uint_t tqflags)
 		mutex_exit(&tq->tq_lock);
 		return (0);
 	}
-	t->task_next = &tq->tq_task;
-	t->task_prev = tq->tq_task.task_prev;
+	if (tqflags & TQ_FRONT) {
+		t->task_next = tq->tq_task.task_next;
+		t->task_prev = &tq->tq_task;
+	} else {
+		t->task_next = &tq->tq_task;
+		t->task_prev = tq->tq_task.task_prev;
+	}
 	t->task_next->task_prev = t;
 	t->task_prev->task_next = t;
 	t->task_func = func;
@@ -191,6 +211,7 @@ taskq_create(const char *name, int nthreads, pri_t pri,
 	mutex_init(&tq->tq_lock, NULL, MUTEX_DEFAULT, NULL);
 	cv_init(&tq->tq_dispatch_cv, NULL, CV_DEFAULT, NULL);
 	cv_init(&tq->tq_wait_cv, NULL, CV_DEFAULT, NULL);
+	cv_init(&tq->tq_maxalloc_cv, NULL, CV_DEFAULT, NULL);
 	tq->tq_flags = flags | TASKQ_ACTIVE;
 	tq->tq_active = nthreads;
 	tq->tq_nthreads = nthreads;
@@ -247,6 +268,7 @@ taskq_destroy(taskq_t *tq)
 	mutex_destroy(&tq->tq_lock);
 	cv_destroy(&tq->tq_dispatch_cv);
 	cv_destroy(&tq->tq_wait_cv);
+	cv_destroy(&tq->tq_maxalloc_cv);
 
 	kmem_free(tq, sizeof (taskq_t));
 }
@@ -272,3 +294,10 @@ system_taskq_init(void)
 	system_taskq = taskq_create("system_taskq", 64, minclsyspri, 4, 512,
 	    TASKQ_DYNAMIC | TASKQ_PREPOPULATE);
 }
+
+void
+system_taskq_fini(void)
+{
+	taskq_destroy(system_taskq);
+	system_taskq = NULL; /* defensive */
+}

lib/libzpool/util.c

@@ -19,8 +19,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
- * Use is subject to license terms.
+ * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  */
 
 #include <assert.h>
@@ -90,7 +89,7 @@ show_vdev_stats(const char *desc, const char *ctype, nvlist_t *nv, int indent)
 		if (is_log)
 			prefix = "log ";
 
-		if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS,
+		if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
 		    (uint64_t **)&vs, &c) != 0)
 			vs = &v0;