Merge branch 'zfs_master_03.04.2015' into zfs_master_20.03.2015_2129+…

…3115_WIP_clean

sync against latest upstream master from April 3rd to benefit from the
changes from openzfs#3225

cmd/mount_zfs/mount_zfs.c

@@ -239,7 +239,7 @@ parse_dataset(char *dataset)
 		if (fd < 0)
 			goto out;
 
-		error = zpool_read_label(fd, &config);
+		error = zpool_read_label(fd, &config, NULL);
 		(void) close(fd);
 		if (error)
 			goto out;

cmd/zdb/zdb.c

@@ -1115,7 +1115,9 @@ snprintf_blkptr_compact(char *blkbuf, size_t buflen, const blkptr_t *bp)
 
 	if (BP_IS_HOLE(bp)) {
 		(void) snprintf(blkbuf + strlen(blkbuf),
-		    buflen - strlen(blkbuf), "B=%llu",
+		    buflen - strlen(blkbuf),
+		    "%llxL B=%llu",
+		    (u_longlong_t)BP_GET_LSIZE(bp),
 		    (u_longlong_t)bp->blk_birth);
 	} else {
 		(void) snprintf(blkbuf + strlen(blkbuf),

cmd/zfs/zfs_main.c

@@ -5641,7 +5641,6 @@ share_mount_one(zfs_handle_t *zhp, int op, int flags, char *protocol,
 	char mountpoint[ZFS_MAXPROPLEN];
 	char shareopts[ZFS_MAXPROPLEN];
 	char smbshareopts[ZFS_MAXPROPLEN];
-	char overlay[ZFS_MAXPROPLEN];
 	const char *cmdname = op == OP_SHARE ? "share" : "mount";
 	struct mnttab mnt;
 	uint64_t zoned, canmount;
@@ -5748,19 +5747,6 @@ share_mount_one(zfs_handle_t *zhp, int op, int flags, char *protocol,
 		return (0);
 	}
 
-	/*
-	 * Overlay mounts are disabled by default but may be enabled
-	 * via the 'overlay' property or the 'zfs mount -O' option.
-	 */
-	if (!(flags & MS_OVERLAY)) {
-		if (zfs_prop_get(zhp, ZFS_PROP_OVERLAY, overlay,
-			    sizeof (overlay), NULL, NULL, 0, B_FALSE) == 0) {
-			if (strcmp(overlay, "on") == 0) {
-				flags |= MS_OVERLAY;
-			}
-		}
-	}
-
 	/*
 	 * At this point, we have verified that the mountpoint and/or
 	 * shareopts are appropriate for auto management. If the

cmd/zpool/zpool_vdev.c

@@ -597,7 +597,7 @@ is_spare(nvlist_t *config, const char *path)
 	if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) != 0 ||
 	    !inuse ||
 	    state != POOL_STATE_SPARE ||
-	    zpool_read_label(fd, &label) != 0) {
+	    zpool_read_label(fd, &label, NULL) != 0) {
 		free(name);
 		(void) close(fd);
 		return (B_FALSE);

config/zfs-build.m4

@@ -151,7 +151,7 @@ AC_DEFUN([ZFS_AC_RPM], [
 	RPM_SPEC_DIR="rpm/generic"
 	AC_ARG_WITH([spec],
 		AS_HELP_STRING([--with-spec=SPEC],
-		[Spec files 'generic|fedora']),
+		[Spec files 'generic|redhat']),
 		[RPM_SPEC_DIR="rpm/$withval"])
 
 	AC_MSG_CHECKING([whether spec files are available])

configure.ac

@@ -134,10 +134,10 @@ AC_CONFIG_FILES([
 	scripts/zpool-config/Makefile
 	scripts/common.sh
 	rpm/Makefile
-	rpm/fedora/Makefile
-	rpm/fedora/zfs.spec
-	rpm/fedora/zfs-kmod.spec
-	rpm/fedora/zfs-dkms.spec
+	rpm/redhat/Makefile
+	rpm/redhat/zfs.spec
+	rpm/redhat/zfs-kmod.spec
+	rpm/redhat/zfs-dkms.spec
 	rpm/generic/Makefile
 	rpm/generic/zfs.spec
 	rpm/generic/zfs-kmod.spec

dracut/90zfs/mount-zfs.sh.in

@@ -12,6 +12,9 @@ if getargbool 0 zfs_force -y zfs.force -y zfsforce ; then
 	ZPOOL_FORCE="-f"
 fi
 
+# Delay until all required block devices are present.
+udevadm settle
+
 case "$root" in
 	zfs:*)
 		# We have ZFS modules loaded, so we're able to import pools now.

include/libzfs.h

@@ -757,7 +757,7 @@ extern int zpool_in_use(libzfs_handle_t *, int, pool_state_t *, char **,
 /*
  * Label manipulation.
  */
-extern int zpool_read_label(int, nvlist_t **);
+extern int zpool_read_label(int, nvlist_t **, int *);
 extern int zpool_clear_label(int);
 
 /*

include/sys/spa.h

@@ -501,12 +501,13 @@ _NOTE(CONSTCOND) } while (0)
 	if (bp == NULL) {						\
 		len += func(buf + len, size - len, "<NULL>");		\
 	} else if (BP_IS_HOLE(bp)) {					\
-		len += func(buf + len, size - len, "<hole>");		\
-		if (bp->blk_birth > 0) {				\
-			len += func(buf + len, size - len,		\
-			    " birth=%lluL",				\
-			    (u_longlong_t)bp->blk_birth);		\
-		}							\
+		len += func(buf + len, size - len,			\
+		    "HOLE [L%llu %s] "					\
+		    "size=%llxL birth=%lluL",				\
+		    (u_longlong_t)BP_GET_LEVEL(bp),			\
+		    type,						\
+		    (u_longlong_t)BP_GET_LSIZE(bp),			\
+		    (u_longlong_t)bp->blk_birth);			\
 	} else if (BP_IS_EMBEDDED(bp)) {				\
 		len = func(buf + len, size - len,			\
 		    "EMBEDDED [L%llu %s] et=%u %s "			\

include/sys/trace_zrlock.h

@@ -42,28 +42,28 @@ DECLARE_EVENT_CLASS(zfs_zrlock_class,
 	TP_PROTO(zrlock_t *zrl, uint32_t n),
 	TP_ARGS(zrl, n),
 	TP_STRUCT__entry(
-	    __field(int32_t,		zr_refcount)
+	    __field(int32_t,		refcount)
 #ifdef	ZFS_DEBUG
-	    __field(pid_t,		zr_owner_pid)
-	    __field(const char *,	zr_caller)
+	    __field(pid_t,		owner_pid)
+	    __field(const char *,	caller)
 #endif
 	    __field(uint32_t,		n)
 	),
 	TP_fast_assign(
-	    __entry->zr_refcount	= zrl->zr_refcount;
+	    __entry->refcount	= zrl->zr_refcount;
 #ifdef	ZFS_DEBUG
-	    __entry->zr_owner_pid	= zrl->zr_owner->pid;
-	    __entry->zr_caller		= zrl->zr_caller;
+	    __entry->owner_pid	= zrl->zr_owner ? zrl->zr_owner->pid : 0;
+	    __entry->caller	= zrl->zr_caller;
 #endif
-	    __entry->n			= n;
+	    __entry->n		= n;
 	),
 #ifdef	ZFS_DEBUG
 	TP_printk("zrl { refcount %d owner_pid %d caller %s } n %u",
-	    __entry->zr_refcount, __entry->zr_owner_pid, __entry->zr_caller,
+	    __entry->refcount, __entry->owner_pid, __entry->caller,
 	    __entry->n)
 #else
 	TP_printk("zrl { refcount %d } n %u",
-	    __entry->zr_refcount, __entry->n)
+	    __entry->refcount, __entry->n)
 #endif
 );
 

include/sys/zfs_context.h

@@ -273,7 +273,6 @@ typedef struct kmutex {
 } kmutex_t;
 
 #define	MUTEX_DEFAULT	0
-#define	MUTEX_FSTRANS	MUTEX_DEFAULT
 #define	MUTEX_HELD(m)	((m)->m_owner == curthread)
 #define	MUTEX_NOT_HELD(m) (!MUTEX_HELD(m))
 

include/sys/zfs_vfsops.h

@@ -92,7 +92,7 @@ typedef struct zfs_sb {
 	uint64_t	z_replay_eof;	/* New end of file - replay only */
 	sa_attr_type_t	*z_attr_table;	/* SA attr mapping->id */
 #define	ZFS_OBJ_MTX_SZ	256
-	kmutex_t	z_hold_mtx[ZFS_OBJ_MTX_SZ];	/* znode hold locks */
+	kmutex_t	*z_hold_mtx;	/* znode hold locks */
 } zfs_sb_t;
 
 #define	ZFS_SUPER_MAGIC	0x2fc12fc1

lib/libzfs/libzfs_import.c

@@ -88,6 +88,7 @@ typedef struct name_entry {
 	char			*ne_name;
 	uint64_t		ne_guid;
 	uint64_t		ne_order;
+	uint64_t		ne_num_labels;
 	struct name_entry	*ne_next;
 } name_entry_t;
 
@@ -173,8 +174,23 @@ fix_paths(nvlist_t *nv, name_entry_t *names)
 				break;
 			}
 
-			if (best == NULL || ne->ne_order < best->ne_order)
+			if (best == NULL) {
 				best = ne;
+				continue;
+			}
+
+			/* Prefer paths with move vdev labels. */
+			if (ne->ne_num_labels > best->ne_num_labels) {
+				best = ne;
+				continue;
+			}
+
+			/* Prefer paths earlier in the search order. */
+			if (best->ne_num_labels == best->ne_num_labels &&
+			    ne->ne_order < best->ne_order) {
+				best = ne;
+				continue;
+			}
 		}
 	}
 
@@ -200,7 +216,7 @@ fix_paths(nvlist_t *nv, name_entry_t *names)
  */
 static int
 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
-    int order, nvlist_t *config)
+    int order, int num_labels, nvlist_t *config)
 {
 	uint64_t pool_guid, vdev_guid, top_guid, txg, state;
 	pool_entry_t *pe;
@@ -226,6 +242,7 @@ add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
 		}
 		ne->ne_guid = vdev_guid;
 		ne->ne_order = order;
+		ne->ne_num_labels = num_labels;
 		ne->ne_next = pl->names;
 		pl->names = ne;
 		return (0);
@@ -328,6 +345,7 @@ add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
 
 	ne->ne_guid = vdev_guid;
 	ne->ne_order = order;
+	ne->ne_num_labels = num_labels;
 	ne->ne_next = pl->names;
 	pl->names = ne;
 
@@ -843,15 +861,17 @@ label_offset(uint64_t size, int l)
 
 /*
  * Given a file descriptor, read the label information and return an nvlist
- * describing the configuration, if there is one.
+ * describing the configuration, if there is one.  The number of valid
+ * labels found will be returned in num_labels when non-NULL.
  */
 int
-zpool_read_label(int fd, nvlist_t **config)
+zpool_read_label(int fd, nvlist_t **config, int *num_labels)
 {
 	struct stat64 statbuf;
-	int l;
+	int l, count = 0;
 	vdev_label_t *label;
-	uint64_t state, txg, size;
+	nvlist_t *expected_config = NULL;
+	uint64_t expected_guid = 0, size;
 
 	*config = NULL;
 
@@ -863,6 +883,8 @@ zpool_read_label(int fd, nvlist_t **config)
 		return (-1);
 
 	for (l = 0; l < VDEV_LABELS; l++) {
+		uint64_t state, guid, txg;
+
 		if (pread64(fd, label, sizeof (vdev_label_t),
 		    label_offset(size, l)) != sizeof (vdev_label_t))
 			continue;
@@ -871,6 +893,12 @@ zpool_read_label(int fd, nvlist_t **config)
 		    sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
 			continue;
 
+		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID,
+		    &guid) != 0 || guid == 0) {
+			nvlist_free(*config);
+			continue;
+		}
+
 		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
 		    &state) != 0 || state > POOL_STATE_L2CACHE) {
 			nvlist_free(*config);
@@ -884,12 +912,24 @@ zpool_read_label(int fd, nvlist_t **config)
 			continue;
 		}
 
-		free(label);
-		return (0);
+		if (expected_guid) {
+			if (expected_guid == guid)
+				count++;
+
+			nvlist_free(*config);
+		} else {
+			expected_config = *config;
+			expected_guid = guid;
+			count++;
+		}
 	}
 
+	if (num_labels != NULL)
+		*num_labels = count;
+
 	free(label);
-	*config = NULL;
+	*config = expected_config;
+
 	return (0);
 }
 
@@ -937,7 +977,7 @@ zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
 	blkid_dev dev;
 	const char *devname;
 	nvlist_t *config;
-	int fd, err;
+	int fd, err, num_labels;
 
 	err = blkid_get_cache(&cache, NULL);
 	if (err != 0) {
@@ -972,7 +1012,7 @@ zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
 		if ((fd = open64(devname, O_RDONLY)) < 0)
 			continue;
 
-		err = zpool_read_label(fd, &config);
+		err = zpool_read_label(fd, &config, &num_labels);
 		(void) close(fd);
 
 		if (err != 0) {
@@ -981,7 +1021,8 @@ zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
 		}
 
 		if (config != NULL) {
-			err = add_config(hdl, pools, devname, 0, config);
+			err = add_config(hdl, pools, devname, 0,
+			    num_labels, config);
 			if (err != 0)
 				goto err_blkid3;
 		}
@@ -1017,7 +1058,7 @@ zpool_default_import_path[DEFAULT_IMPORT_PATH_SIZE] = {
 static nvlist_t *
 zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
 {
-	int i, dirs = iarg->paths;
+	int i, num_labels, dirs = iarg->paths;
 	DIR *dirp = NULL;
 	struct dirent64 *dp;
 	char path[MAXPATHLEN];
@@ -1143,7 +1184,7 @@ zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
 			if ((fd = openat64(dfd, name, O_RDONLY)) < 0)
 				continue;
 
-			if ((zpool_read_label(fd, &config)) != 0) {
+			if ((zpool_read_label(fd, &config, &num_labels))) {
 				(void) close(fd);
 				(void) no_memory(hdl);
 				goto error;
@@ -1177,7 +1218,8 @@ zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
 				}
 				/* use the non-raw path for the config */
 				(void) strlcpy(end, name, pathleft);
-				if (add_config(hdl, &pools, path, i+1, config))
+				if (add_config(hdl, &pools, path, i+1,
+				    num_labels, config))
 					goto error;
 			}
 		}
@@ -1461,7 +1503,7 @@ zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
 
 	*inuse = B_FALSE;
 
-	if (zpool_read_label(fd, &config) != 0) {
+	if (zpool_read_label(fd, &config, NULL) != 0) {
 		(void) no_memory(hdl);
 		return (-1);
 	}

lib/libzfs/libzfs_mount.c

@@ -388,6 +388,7 @@ zfs_mount(zfs_handle_t *zhp, const char *options, int flags)
 	struct stat buf;
 	char mountpoint[ZFS_MAXPROPLEN];
 	char mntopts[MNT_LINE_MAX];
+	char overlay[ZFS_MAXPROPLEN];
 	libzfs_handle_t *hdl = zhp->zfs_hdl;
 	int remount = 0, rc;
 
@@ -441,6 +442,19 @@ zfs_mount(zfs_handle_t *zhp, const char *options, int flags)
 		}
 	}
 
+	/*
+	 * Overlay mounts are disabled by default but may be enabled
+	 * via the 'overlay' property or the 'zfs mount -O' option.
+	 */
+	if (!(flags & MS_OVERLAY)) {
+		if (zfs_prop_get(zhp, ZFS_PROP_OVERLAY, overlay,
+			    sizeof (overlay), NULL, NULL, 0, B_FALSE) == 0) {
+			if (strcmp(overlay, "on") == 0) {
+				flags |= MS_OVERLAY;
+			}
+		}
+	}
+
 	/*
 	 * Determine if the mountpoint is empty.  If so, refuse to perform the
 	 * mount.  We don't perform this check if 'remount' is