Prepare to release 0.6.5.11

META file and RPM release log updated.

Signed-off-by: Tony Hutter <hutter2@llnl.gov>

META

@@ -1,7 +1,7 @@
 Meta:         1
 Name:         zfs
 Branch:       1.0
-Version:      0.6.5.6
+Version:      0.6.5.11
 Release:      1
 Release-Tags: relext
 License:      CDDL

Makefile.am

@@ -55,6 +55,13 @@ shellcheck:
 		 done; \
 	 fi
 
+lint: cppcheck
+
+cppcheck:
+	@if type cppcheck > /dev/null 2>&1; then \
+		cppcheck --quiet --force ${top_srcdir}; \
+	fi
+
 ctags:
 	$(RM) tags
 	find $(top_srcdir) -name .git -prune -o -name '*.[hc]' | xargs ctags

cmd/mount_zfs/mount_zfs.c

@@ -77,7 +77,10 @@ static const option_map_t option_map[] = {
 	{ MNTOPT_RELATIME,	MS_RELATIME,	ZS_COMMENT	},
 #endif
 #ifdef MS_STRICTATIME
-	{ MNTOPT_DFRATIME,	MS_STRICTATIME,	ZS_COMMENT	},
+	{ MNTOPT_STRICTATIME,	MS_STRICTATIME,	ZS_COMMENT	},
+#endif
+#ifdef MS_LAZYTIME
+	{ MNTOPT_LAZYTIME,	MS_LAZYTIME,	ZS_COMMENT	},
 #endif
 	{ MNTOPT_CONTEXT,	MS_COMMENT,	ZS_COMMENT	},
 	{ MNTOPT_FSCONTEXT,	MS_COMMENT,	ZS_COMMENT	},
@@ -605,10 +608,23 @@ main(int argc, char **argv)
 				    "failed for unknown reason.\n"), dataset);
 			}
 			return (MOUNT_SYSERR);
+#ifdef MS_MANDLOCK
+		case EPERM:
+			if (mntflags & MS_MANDLOCK) {
+				(void) fprintf(stderr, gettext("filesystem "
+				    "'%s' has the 'nbmand=on' property set, "
+				    "this mount\noption may be disabled in "
+				    "your kernel.  Use 'zfs set nbmand=off'\n"
+				    "to disable this option and try to "
+				    "mount the filesystem again.\n"), dataset);
+				return (MOUNT_SYSERR);
+			}
+			/* fallthru */
+#endif
 		default:
 			(void) fprintf(stderr, gettext("filesystem "
-			    "'%s' can not be mounted due to error "
-			    "%d\n"), dataset, errno);
+			    "'%s' can not be mounted: %s\n"), dataset,
+			    strerror(errno));
 			return (MOUNT_USAGE);
 		}
 	}

cmd/vdev_id/vdev_id

@@ -184,9 +184,9 @@ sas_handler() {
 			return
 		fi
 
-		# Get the raw scsi device name from multipath -l.  Strip off
+		# Get the raw scsi device name from multipath -ll. Strip off
 		# leading pipe symbols to make field numbering consistent.
-		DEV=`multipath -l $DM_NAME |
+		DEV=`multipath -ll $DM_NAME |
 			awk '/running/{gsub("^[|]"," "); print $3 ; exit}'`
 		if [ -z "$DEV" ] ; then
 			return

cmd/zdb/zdb.c

@@ -67,13 +67,22 @@
 	zio_compress_table[(idx)].ci_name : "UNKNOWN")
 #define	ZDB_CHECKSUM_NAME(idx) ((idx) < ZIO_CHECKSUM_FUNCTIONS ?	\
 	zio_checksum_table[(idx)].ci_name : "UNKNOWN")
-#define	ZDB_OT_NAME(idx) ((idx) < DMU_OT_NUMTYPES ?	\
-	dmu_ot[(idx)].ot_name : DMU_OT_IS_VALID(idx) ?	\
-	dmu_ot_byteswap[DMU_OT_BYTESWAP(idx)].ob_name : "UNKNOWN")
 #define	ZDB_OT_TYPE(idx) ((idx) < DMU_OT_NUMTYPES ? (idx) :		\
 	(((idx) == DMU_OTN_ZAP_DATA || (idx) == DMU_OTN_ZAP_METADATA) ?	\
 	DMU_OT_ZAP_OTHER : DMU_OT_NUMTYPES))
 
+static char *
+zdb_ot_name(dmu_object_type_t type)
+{
+	if (type < DMU_OT_NUMTYPES)
+		return (dmu_ot[type].ot_name);
+	else if ((type & DMU_OT_NEWTYPE) &&
+		((type & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS))
+		return (dmu_ot_byteswap[type & DMU_OT_BYTESWAP_MASK].ob_name);
+	else
+		return ("UNKNOWN");
+}
+
 #ifndef lint
 extern int zfs_recover;
 extern uint64_t zfs_arc_max, zfs_arc_meta_limit;
@@ -469,7 +478,7 @@ static void
 dump_bpobj_subobjs(objset_t *os, uint64_t object, void *data, size_t size)
 {
 	dmu_object_info_t doi;
-	uint64_t i;
+	int64_t i;
 
 	VERIFY0(dmu_object_info(os, object, &doi));
 	uint64_t *subobjs = kmem_alloc(doi.doi_max_offset, KM_SLEEP);
@@ -488,7 +497,7 @@ dump_bpobj_subobjs(objset_t *os, uint64_t object, void *data, size_t size)
 	}
 
 	for (i = 0; i <= last_nonzero; i++) {
-		(void) printf("\t%llu\n", (longlong_t)subobjs[i]);
+		(void) printf("\t%llu\n", (u_longlong_t)subobjs[i]);
 	}
 	kmem_free(subobjs, doi.doi_max_offset);
 }
@@ -1925,12 +1934,12 @@ dump_object(objset_t *os, uint64_t object, int verbosity, int *print_header)
 
 	(void) printf("%10lld  %3u  %5s  %5s  %5s  %5s  %6s  %s%s\n",
 	    (u_longlong_t)object, doi.doi_indirection, iblk, dblk,
-	    asize, lsize, fill, ZDB_OT_NAME(doi.doi_type), aux);
+	    asize, lsize, fill, zdb_ot_name(doi.doi_type), aux);
 
 	if (doi.doi_bonus_type != DMU_OT_NONE && verbosity > 3) {
 		(void) printf("%10s  %3s  %5s  %5s  %5s  %5s  %6s  %s\n",
 		    "", "", "", "", "", bonus_size, "bonus",
-		    ZDB_OT_NAME(doi.doi_bonus_type));
+		    zdb_ot_name(doi.doi_bonus_type));
 	}
 
 	if (verbosity >= 4) {

cmd/zfs/zfs_iter.c

@@ -444,13 +444,13 @@ zfs_for_each(int argc, char **argv, int flags, zfs_type_t types,
 
 		/*
 		 * If we're recursive, then we always allow filesystems as
-		 * arguments.  If we also are interested in snapshots, then we
-		 * can take volumes as well.
+		 * arguments.  If we also are interested in snapshots or
+		 * bookmarks, then we can take volumes as well.
 		 */
 		argtype = types;
 		if (flags & ZFS_ITER_RECURSE) {
 			argtype |= ZFS_TYPE_FILESYSTEM;
-			if (types & ZFS_TYPE_SNAPSHOT)
+			if (types & (ZFS_TYPE_SNAPSHOT | ZFS_TYPE_BOOKMARK))
 				argtype |= ZFS_TYPE_VOLUME;
 		}
 

cmd/zpool/zpool_main.c

@@ -6038,6 +6038,7 @@ main(int argc, char **argv)
 
 	(void) setlocale(LC_ALL, "");
 	(void) textdomain(TEXT_DOMAIN);
+	srand(time(NULL));
 
 	dprintf_setup(&argc, argv);
 

cmd/zpool/zpool_vdev.c

@@ -1206,12 +1206,10 @@ make_disks(zpool_handle_t *zhp, nvlist_t *nv)
 
 		/*
 		 * Remove any previously existing symlink from a udev path to
-		 * the device before labeling the disk.  This makes
-		 * zpool_label_disk_wait() truly wait for the new link to show
-		 * up instead of returning if it finds an old link still in
-		 * place.  Otherwise there is a window between when udev
-		 * deletes and recreates the link during which access attempts
-		 * will fail with ENOENT.
+		 * the device before labeling the disk.  This ensures that
+		 * only newly created links are used.  Otherwise there is a
+		 * window between when udev deletes and recreates the link
+		 * during which access attempts will fail with ENOENT.
 		 */
 		strncpy(udevpath, path, MAXPATHLEN);
 		(void) zfs_append_partition(udevpath, MAXPATHLEN);
@@ -1235,6 +1233,8 @@ make_disks(zpool_handle_t *zhp, nvlist_t *nv)
 		 * and then block until udev creates the new link.
 		 */
 		if (!is_exclusive || !is_spare(NULL, udevpath)) {
+			char *devnode = strrchr(devpath, '/') + 1;
+
 			ret = strncmp(udevpath, UDISK_ROOT, strlen(UDISK_ROOT));
 			if (ret == 0) {
 				ret = lstat64(udevpath, &statbuf);
@@ -1242,18 +1242,29 @@ make_disks(zpool_handle_t *zhp, nvlist_t *nv)
 					(void) unlink(udevpath);
 			}
 
-			if (zpool_label_disk(g_zfs, zhp,
-			    strrchr(devpath, '/') + 1) == -1)
+			/*
+			 * When labeling a pool the raw device node name
+			 * is provided as it appears under /dev/.
+			 */
+			if (zpool_label_disk(g_zfs, zhp, devnode) == -1)
 				return (-1);
 
+			/*
+			 * Wait for udev to signal the device is available
+			 * by the provided path.
+			 */
 			ret = zpool_label_disk_wait(udevpath, DISK_LABEL_WAIT);
 			if (ret) {
-				(void) fprintf(stderr, gettext("cannot "
-				    "resolve path '%s': %d\n"), udevpath, ret);
-				return (-1);
+				(void) fprintf(stderr,
+				    gettext("missing link: %s was "
+				    "partitioned but %s is missing\n"),
+				    devnode, udevpath);
+				return (ret);
 			}
 
-			(void) zero_label(udevpath);
+			ret = zero_label(udevpath);
+			if (ret)
+				return (ret);
 		}
 
 		/*

cmd/ztest/Makefile.am

@@ -1,6 +1,8 @@
 include $(top_srcdir)/config/Rules.am
 
-AM_CFLAGS += $(DEBUG_STACKFLAGS) $(FRAME_LARGER_THAN)
+# -Wnoformat-truncation to get rid of compiler warning for unchecked
+#  truncating snprintfs on gcc 7.1.1.
+AM_CFLAGS += $(DEBUG_STACKFLAGS) $(FRAME_LARGER_THAN) $(NO_FORMAT_TRUNCATION)
 
 DEFAULT_INCLUDES += \
 	-I$(top_srcdir)/include \

config/Rules.am

@@ -7,7 +7,8 @@ AM_CFLAGS += ${NO_BOOL_COMPARE}
 AM_CFLAGS += -fno-strict-aliasing
 AM_CPPFLAGS  = -D_GNU_SOURCE -D__EXTENSIONS__ -D_REENTRANT
 AM_CPPFLAGS += -D_POSIX_PTHREAD_SEMANTICS -D_FILE_OFFSET_BITS=64
-AM_CPPFLAGS += -D_LARGEFILE64_SOURCE -DTEXT_DOMAIN=\"zfs-linux-user\"
+AM_CPPFLAGS += -D_LARGEFILE64_SOURCE -DHAVE_LARGE_STACKS=1
+AM_CPPFLAGS += -DTEXT_DOMAIN=\"zfs-linux-user\"
 AM_CPPFLAGS += -DLIBEXECDIR=\"$(libexecdir)\"
 AM_CPPFLAGS += -DRUNSTATEDIR=\"$(runstatedir)\"
 AM_CPPFLAGS += -DSBINDIR=\"$(sbindir)\"

config/kernel-acl.m4

@@ -39,6 +39,35 @@ AC_DEFUN([ZFS_AC_KERNEL_POSIX_ACL_RELEASE], [
 	])
 ])
 
+dnl #
+dnl # 3.14 API change,
+dnl # set_cached_acl() and forget_cached_acl() changed from inline to
+dnl # EXPORT_SYMBOL. In the former case, they may not be usable because of
+dnl # posix_acl_release. In the latter case, we can always use them.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SET_CACHED_ACL_USABLE], [
+	AC_MSG_CHECKING([whether set_cached_acl() is usable])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/module.h>
+		#include <linux/cred.h>
+		#include <linux/fs.h>
+		#include <linux/posix_acl.h>
+
+		MODULE_LICENSE("$ZFS_META_LICENSE");
+	],[
+		struct inode *ip = NULL;
+		struct posix_acl *acl = posix_acl_alloc(1, 0);
+		set_cached_acl(ip, ACL_TYPE_ACCESS, acl);
+		forget_cached_acl(ip, ACL_TYPE_ACCESS);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SET_CACHED_ACL_USABLE, 1,
+		    [posix_acl_release() is usable])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
 dnl #
 dnl # 3.1 API change,
 dnl # posix_acl_chmod_masq() is not exported anymore and posix_acl_chmod()
@@ -75,27 +104,6 @@ AC_DEFUN([ZFS_AC_KERNEL_POSIX_ACL_CHMOD], [
 	])
 ])
 
-dnl #
-dnl # 2.6.30 API change,
-dnl # caching of ACL into the inode was added in this version.
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_POSIX_ACL_CACHING], [
-	AC_MSG_CHECKING([whether inode has i_acl and i_default_acl])
-	ZFS_LINUX_TRY_COMPILE([
-		#include <linux/fs.h>
-	],[
-		struct inode ino;
-		ino.i_acl = NULL;
-		ino.i_default_acl = NULL;
-	],[
-		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_POSIX_ACL_CACHING, 1,
-		    [inode contains i_acl and i_default_acl])
-	],[
-		AC_MSG_RESULT(no)
-	])
-])
-
 dnl #
 dnl # 3.1 API change,
 dnl # posix_acl_equiv_mode now wants an umode_t* instead of a mode_t*
@@ -117,6 +125,30 @@ AC_DEFUN([ZFS_AC_KERNEL_POSIX_ACL_EQUIV_MODE_WANTS_UMODE_T], [
 	])
 ])
 
+dnl #
+dnl # 4.8 API change,
+dnl # The function posix_acl_valid now must be passed a namespace.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_POSIX_ACL_VALID_WITH_NS], [
+	AC_MSG_CHECKING([whether posix_acl_valid() wants user namespace])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+		#include <linux/posix_acl.h>
+	],[
+		struct user_namespace *user_ns = NULL;
+		const struct posix_acl *acl = NULL;
+		int error;
+
+		error = posix_acl_valid(user_ns, acl);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_POSIX_ACL_VALID_WITH_NS, 1,
+		    [posix_acl_valid() wants user namespace])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
 dnl #
 dnl # 2.6.27 API change,
 dnl # Check if inode_operations contains the function permission
@@ -247,18 +279,45 @@ AC_DEFUN([ZFS_AC_KERNEL_INODE_OPERATIONS_GET_ACL], [
 ])
 
 dnl #
-dnl # 2.6.30 API change,
-dnl # current_umask exists only since this version.
+dnl # 3.14 API change,
+dnl # Check if inode_operations contains the function set_acl
 dnl #
-AC_DEFUN([ZFS_AC_KERNEL_CURRENT_UMASK], [
-	AC_MSG_CHECKING([whether current_umask exists])
+AC_DEFUN([ZFS_AC_KERNEL_INODE_OPERATIONS_SET_ACL], [
+	AC_MSG_CHECKING([whether iops->set_acl() exists])
 	ZFS_LINUX_TRY_COMPILE([
 		#include <linux/fs.h>
+
+		int set_acl_fn(struct inode *inode, struct posix_acl *acl, int type)
+		    { return 0; }
+
+		static const struct inode_operations
+		    iops __attribute__ ((unused)) = {
+			.set_acl = set_acl_fn,
+		};
 	],[
-		current_umask();
 	],[
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_CURRENT_UMASK, 1, [current_umask() exists])
+		AC_DEFINE(HAVE_SET_ACL, 1, [iops->set_acl() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # 4.7 API change,
+dnl # The kernel get_acl will now check cache before calling i_op->get_acl and
+dnl # do set_cached_acl after that, so i_op->get_acl don't need to do that
+dnl # anymore.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_GET_ACL_HANDLE_CACHE], [
+	AC_MSG_CHECKING([whether uncached_acl_sentinel() exists])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+	],[
+		void *sentinel __attribute__ ((unused)) = uncached_acl_sentinel(NULL);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_KERNEL_GET_ACL_HANDLE_CACHE, 1, [uncached_acl_sentinel() exists])
 	],[
 		AC_MSG_RESULT(no)
 	])

config/kernel-aio-fsync.m4

@@ -0,0 +1,21 @@
+dnl #
+dnl # Linux 4.9-rc5+ ABI, removal of the .aio_fsync field
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_AIO_FSYNC], [
+	AC_MSG_CHECKING([whether fops->aio_fsync() exists])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+
+		static const struct file_operations
+		    fops __attribute__ ((unused)) = {
+			.aio_fsync = NULL,
+		};
+	],[
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_FILE_AIO_FSYNC, 1, [fops->aio_fsync() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+

config/kernel-bdi-setup-and-register.m4

@@ -1,38 +0,0 @@
-dnl #
-dnl # 2.6.32 - 2.6.33, bdi_setup_and_register() is not exported.
-dnl # 2.6.34 - 3.19, bdi_setup_and_register() takes 3 arguments.
-dnl # 4.0 - x.y, bdi_setup_and_register() takes 2 arguments.
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_BDI_SETUP_AND_REGISTER], [
-	AC_MSG_CHECKING([whether bdi_setup_and_register() wants 2 args])
-	ZFS_LINUX_TRY_COMPILE_SYMBOL([
-		#include <linux/backing-dev.h>
-		struct backing_dev_info bdi;
-	], [
-		char *name = "bdi";
-		int error __attribute__((unused)) =
-		    bdi_setup_and_register(&bdi, name);
-	], [bdi_setup_and_register], [mm/backing-dev.c], [
-		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_2ARGS_BDI_SETUP_AND_REGISTER, 1,
-		    [bdi_setup_and_register() wants 2 args])
-	], [
-		AC_MSG_RESULT(no)
-		AC_MSG_CHECKING([whether bdi_setup_and_register() wants 3 args])
-		ZFS_LINUX_TRY_COMPILE_SYMBOL([
-			#include <linux/backing-dev.h>
-			struct backing_dev_info bdi;
-		], [
-			char *name = "bdi";
-			unsigned int cap = BDI_CAP_MAP_COPY;
-			int error __attribute__((unused)) =
-			    bdi_setup_and_register(&bdi, name, cap);
-		], [bdi_setup_and_register], [mm/backing-dev.c], [
-			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_3ARGS_BDI_SETUP_AND_REGISTER, 1,
-			    [bdi_setup_and_register() wants 3 args])
-		], [
-			AC_MSG_RESULT(no)
-		])
-	])
-])

config/kernel-bdi.m4

@@ -0,0 +1,56 @@
+dnl #
+dnl # 2.6.32 - 2.6.33, bdi_setup_and_register() is not exported.
+dnl # 2.6.34 - 3.19, bdi_setup_and_register() takes 3 arguments.
+dnl # 4.0 - 4.11, bdi_setup_and_register() takes 2 arguments.
+dnl # 4.12 - x.y, super_setup_bdi_name() new interface.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_BDI], [
+	AC_MSG_CHECKING([whether super_setup_bdi_name() exists])
+	ZFS_LINUX_TRY_COMPILE_SYMBOL([
+		#include <linux/fs.h>
+		struct super_block sb;
+	], [
+		char *name = "bdi";
+		int error __attribute__((unused)) =
+		    super_setup_bdi_name(&sb, name);
+	], [super_setup_bdi_name], [fs/super.c], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SUPER_SETUP_BDI_NAME, 1,
+                    [super_setup_bdi_name() exits])
+	], [
+		AC_MSG_RESULT(no)
+		AC_MSG_CHECKING(
+		    [whether bdi_setup_and_register() wants 2 args])
+		ZFS_LINUX_TRY_COMPILE_SYMBOL([
+			#include <linux/backing-dev.h>
+			struct backing_dev_info bdi;
+		], [
+			char *name = "bdi";
+			int error __attribute__((unused)) =
+			    bdi_setup_and_register(&bdi, name);
+		], [bdi_setup_and_register], [mm/backing-dev.c], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_2ARGS_BDI_SETUP_AND_REGISTER, 1,
+			    [bdi_setup_and_register() wants 2 args])
+		], [
+			AC_MSG_RESULT(no)
+			AC_MSG_CHECKING(
+			    [whether bdi_setup_and_register() wants 3 args])
+			ZFS_LINUX_TRY_COMPILE_SYMBOL([
+				#include <linux/backing-dev.h>
+				struct backing_dev_info bdi;
+			], [
+				char *name = "bdi";
+				unsigned int cap = BDI_CAP_MAP_COPY;
+				int error __attribute__((unused)) =
+				    bdi_setup_and_register(&bdi, name, cap);
+			], [bdi_setup_and_register], [mm/backing-dev.c], [
+				AC_MSG_RESULT(yes)
+				AC_DEFINE(HAVE_3ARGS_BDI_SETUP_AND_REGISTER, 1,
+				    [bdi_setup_and_register() wants 3 args])
+			], [
+				AC_MSG_RESULT(no)
+			])
+		])
+	])
+])

config/kernel-bio-op.m4

@@ -0,0 +1,84 @@
+dnl #
+dnl # Linux 4.8 API,
+dnl #
+dnl # The bio_op() helper was introduced as a replacement for explicitly
+dnl # checking the bio->bi_rw flags.  The following checks are used to
+dnl # detect if a specific operation is supported.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_REQ_OP_DISCARD], [
+	AC_MSG_CHECKING([whether REQ_OP_DISCARD is defined])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/blk_types.h>
+	],[
+		int op __attribute__ ((unused)) = REQ_OP_DISCARD;
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_REQ_OP_DISCARD, 1,
+		    [REQ_OP_DISCARD is defined])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_REQ_OP_SECURE_ERASE], [
+	AC_MSG_CHECKING([whether REQ_OP_SECURE_ERASE is defined])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/blk_types.h>
+	],[
+		int op __attribute__ ((unused)) = REQ_OP_SECURE_ERASE;
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_REQ_OP_SECURE_ERASE, 1,
+		    [REQ_OP_SECURE_ERASE is defined])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+
+AC_DEFUN([ZFS_AC_KERNEL_REQ_OP_FLUSH], [
+	AC_MSG_CHECKING([whether REQ_OP_FLUSH is defined])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/blk_types.h>
+	],[
+		int op __attribute__ ((unused)) = REQ_OP_FLUSH;
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_REQ_OP_FLUSH, 1,
+		    [REQ_OP_FLUSH is defined])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BIO_BI_OPF], [
+	AC_MSG_CHECKING([whether bio->bi_opf is defined])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/bio.h>
+	],[
+		struct bio bio __attribute__ ((unused));
+		bio.bi_opf = 0;
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BIO_BI_OPF, 1, [bio->bi_opf is defined])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_HAVE_BIO_SET_OP_ATTRS], [
+	AC_MSG_CHECKING([whether bio_set_op_attrs is available])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/bio.h>
+	],[
+		struct bio *bio __attribute__ ((unused)) = NULL;
+
+		bio_set_op_attrs(bio, 0, 0);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BIO_SET_OP_ATTRS, 1,
+		    [bio_set_op_attrs is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-blk-queue-flush.m4

@@ -22,25 +22,64 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_FLUSH], [
 		AC_MSG_RESULT(yes)
 		AC_DEFINE(HAVE_BLK_QUEUE_FLUSH, 1,
 		          [blk_queue_flush() is available])
+
+		AC_MSG_CHECKING([whether blk_queue_flush() is GPL-only])
+		ZFS_LINUX_TRY_COMPILE([
+			#include <linux/module.h>
+			#include <linux/blkdev.h>
+
+			MODULE_LICENSE("$ZFS_META_LICENSE");
+		],[
+			struct request_queue *q = NULL;
+			(void) blk_queue_flush(q, REQ_FLUSH);
+		],[
+			AC_MSG_RESULT(no)
+		],[
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_BLK_QUEUE_FLUSH_GPL_ONLY, 1,
+				  [blk_queue_flush() is GPL-only])
+		])
 	],[
 		AC_MSG_RESULT(no)
 	])
 
-	AC_MSG_CHECKING([whether blk_queue_flush() is GPL-only])
+	dnl #
+	dnl # 4.7 API change
+	dnl # Replace blk_queue_flush with blk_queue_write_cache
+	dnl #
+	AC_MSG_CHECKING([whether blk_queue_write_cache() exists])
 	ZFS_LINUX_TRY_COMPILE([
-		#include <linux/module.h>
+		#include <linux/kernel.h>
 		#include <linux/blkdev.h>
 
-		MODULE_LICENSE("$ZFS_META_LICENSE");
 	],[
 		struct request_queue *q = NULL;
-		(void) blk_queue_flush(q, REQ_FLUSH);
-	],[
-		AC_MSG_RESULT(no)
+		blk_queue_write_cache(q, true, true);
 	],[
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_BLK_QUEUE_FLUSH_GPL_ONLY, 1,
-		          [blk_queue_flush() is GPL-only])
+		AC_DEFINE(HAVE_BLK_QUEUE_WRITE_CACHE, 1,
+			[blk_queue_write_cache() exists])
+
+		AC_MSG_CHECKING([whether blk_queue_write_cache() is GPL-only])
+		ZFS_LINUX_TRY_COMPILE([
+			#include <linux/kernel.h>
+			#include <linux/module.h>
+			#include <linux/blkdev.h>
+
+			MODULE_LICENSE("$ZFS_META_LICENSE");
+		],[
+			struct request_queue *q = NULL;
+			blk_queue_write_cache(q, true, true);
+		],[
+			AC_MSG_RESULT(no)
+		],[
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY, 1,
+				  [blk_queue_write_cache() is GPL-only])
+		])
+	],[
+		AC_MSG_RESULT(no)
 	])
+
 	EXTRA_KCFLAGS="$tmp_flags"
 ])

config/kernel-blk-queue-unplug.m4

@@ -0,0 +1,44 @@
+dnl #
+dnl # 2.6.32-2.6.35 API - The BIO_RW_UNPLUG enum can be used as a hint
+dnl # to unplug the queue.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_HAVE_BIO_RW_UNPLUG], [
+	AC_MSG_CHECKING([whether the BIO_RW_UNPLUG enum is available])
+	tmp_flags="$EXTRA_KCFLAGS"
+	EXTRA_KCFLAGS="${NO_UNUSED_BUT_SET_VARIABLE}"
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/blkdev.h>
+	],[
+		extern enum bio_rw_flags rw;
+
+		rw = BIO_RW_UNPLUG;
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BLK_QUEUE_HAVE_BIO_RW_UNPLUG, 1,
+		          [BIO_RW_UNPLUG is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+	EXTRA_KCFLAGS="$tmp_flags"
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_HAVE_BLK_PLUG], [
+	AC_MSG_CHECKING([whether struct blk_plug is available])
+	tmp_flags="$EXTRA_KCFLAGS"
+	EXTRA_KCFLAGS="${NO_UNUSED_BUT_SET_VARIABLE}"
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/blkdev.h>
+	],[
+		struct blk_plug plug;
+
+		blk_start_plug(&plug);
+		blk_finish_plug(&plug);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BLK_QUEUE_HAVE_BLK_PLUG, 1,
+		          [struct blk_plug is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+	EXTRA_KCFLAGS="$tmp_flags"
+])

config/kernel-current-time.m4

@@ -0,0 +1,19 @@
+dnl #
+dnl # 4.9, current_time() added
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_CURRENT_TIME],
+	[AC_MSG_CHECKING([whether current_time() exists])
+	ZFS_LINUX_TRY_COMPILE_SYMBOL([
+		#include <linux/fs.h>
+	], [
+		struct inode ip;
+		struct timespec now __attribute__ ((unused));
+
+		now = current_time(&ip);
+	], [current_time], [fs/inode.c], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_CURRENT_TIME, 1, [current_time() exists])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-generic_readlink.m4

@@ -0,0 +1,22 @@
+dnl #
+dnl # 4.10 API
+dnl #
+dnl # NULL inode_operations.readlink implies generic_readlink(), which
+dnl # has been made static.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_GENERIC_READLINK_GLOBAL], [
+	AC_MSG_CHECKING([whether generic_readlink is global])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+	],[
+		int i __attribute__ ((unused));
+
+		i = generic_readlink(NULL, NULL, 0);
+	],[
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_GENERIC_READLINK, 1,
+		          [generic_readlink is global])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel-inode-getattr.m4

@@ -0,0 +1,67 @@
+dnl #
+dnl # Linux 4.11 API
+dnl # See torvalds/linux@a528d35
+dnl #
+AC_DEFUN([ZFS_AC_PATH_KERNEL_IOPS_GETATTR], [
+	AC_MSG_CHECKING([whether iops->getattr() takes a path])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+
+		int test_getattr(
+		    const struct path *p, struct kstat *k,
+		    u32 request_mask, unsigned int query_flags)
+		    { return 0; }
+
+		static const struct inode_operations
+		    iops __attribute__ ((unused)) = {
+			.getattr = test_getattr,
+		};
+	],[
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_PATH_IOPS_GETATTR, 1,
+		    [iops->getattr() takes a path])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+
+
+dnl #
+dnl # Linux 3.9 - 4.10 API
+dnl #
+AC_DEFUN([ZFS_AC_VFSMOUNT_KERNEL_IOPS_GETATTR], [
+	AC_MSG_CHECKING([whether iops->getattr() takes a vfsmount])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+
+		int test_getattr(
+		    struct vfsmount *mnt, struct dentry *d,
+		    struct kstat *k)
+		    { return 0; }
+
+		static const struct inode_operations
+		    iops __attribute__ ((unused)) = {
+			.getattr = test_getattr,
+		};
+	],[
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_VFSMOUNT_IOPS_GETATTR, 1,
+		    [iops->getattr() takes a vfsmount])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+
+dnl #
+dnl # The interface of the getattr callback from the inode_operations
+dnl # structure changed.  Also, the interface of the simple_getattr()
+dnl # function provided by the kernel changed.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_INODE_OPERATIONS_GETATTR], [
+	ZFS_AC_PATH_KERNEL_IOPS_GETATTR
+	ZFS_AC_VFSMOUNT_KERNEL_IOPS_GETATTR
+])

config/kernel-lookup-bdev.m4

@@ -1,17 +1,29 @@
 dnl #
-dnl # 2.6.27 API change
-dnl # lookup_bdev() was exported.
+dnl # 2.6.27, lookup_bdev() was exported.
+dnl # 4.4.0-6.21 - x.y on Ubuntu, lookup_bdev() takes 2 arguments.
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_LOOKUP_BDEV],
-	[AC_MSG_CHECKING([whether lookup_bdev() is available])
+	[AC_MSG_CHECKING([whether lookup_bdev() wants 1 arg])
 	ZFS_LINUX_TRY_COMPILE_SYMBOL([
 		#include <linux/fs.h>
 	], [
 		lookup_bdev(NULL);
 	], [lookup_bdev], [fs/block_dev.c], [
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_LOOKUP_BDEV, 1, [lookup_bdev() is available])
+		AC_DEFINE(HAVE_1ARG_LOOKUP_BDEV, 1, [lookup_bdev() wants 1 arg])
 	], [
 		AC_MSG_RESULT(no)
+		AC_MSG_CHECKING([whether lookup_bdev() wants 2 args])
+		ZFS_LINUX_TRY_COMPILE_SYMBOL([
+			#include <linux/fs.h>
+		], [
+			lookup_bdev(NULL, FMODE_READ);
+		], [lookup_bdev], [fs/block_dev.c], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_2ARGS_LOOKUP_BDEV, 1,
+			    [lookup_bdev() wants 2 args])
+		], [
+			AC_MSG_RESULT(no)
+		])
 	])
-])
+])

config/kernel-rename.m4

@@ -0,0 +1,25 @@
+dnl #
+dnl # 4.9 API change,
+dnl # iops->rename2() merged into iops->rename(), and iops->rename() now wants
+dnl # flags.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_RENAME_WANTS_FLAGS], [
+	AC_MSG_CHECKING([whether iops->rename() wants flags])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+		int rename_fn(struct inode *sip, struct dentry *sdp,
+			struct inode *tip, struct dentry *tdp,
+			unsigned int flags) { return 0; }
+
+		static const struct inode_operations
+		    iops __attribute__ ((unused)) = {
+			.rename = rename_fn,
+		};
+	],[
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_RENAME_WANTS_FLAGS, 1, [iops->rename() wants flags])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-setattr-prepare.m4

@@ -0,0 +1,23 @@
+dnl #
+dnl # 4.9 API change
+dnl # The inode_change_ok() function has been renamed setattr_prepare()
+dnl # and updated to take a dentry rather than an inode.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SETATTR_PREPARE],
+	[AC_MSG_CHECKING([whether setattr_prepare() is available])
+	ZFS_LINUX_TRY_COMPILE_SYMBOL([
+		#include <linux/fs.h>
+	], [
+		struct dentry *dentry = NULL;
+		struct iattr *attr = NULL;
+		int error;
+
+		error = setattr_prepare(dentry, attr);
+	], [setattr_prepare], [fs/attr.c], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_SETATTR_PREPARE, 1,
+		    [setattr_prepare() is available])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-submit_bio.m4

@@ -0,0 +1,20 @@
+dnl #
+dnl # 4.8 API change
+dnl # The rw argument has been removed from submit_bio/submit_bio_wait.
+dnl # Callers are now expected to set bio->bi_rw instead of passing it in.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SUBMIT_BIO], [
+	AC_MSG_CHECKING([whether submit_bio() wants 1 arg])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/bio.h>
+	],[
+		blk_qc_t blk_qc;
+		struct bio *bio = NULL;
+		blk_qc = submit_bio(bio);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_1ARG_SUBMIT_BIO, 1, [submit_bio() wants 1 arg])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel-vfs-iterate.m4

@@ -1,8 +1,8 @@
-dnl #
-dnl # 3.11 API change
-dnl #
 AC_DEFUN([ZFS_AC_KERNEL_VFS_ITERATE], [
-	AC_MSG_CHECKING([whether fops->iterate() is available])
+	dnl #
+	dnl # 4.7 API change
+	dnl #
+	AC_MSG_CHECKING([whether fops->iterate_shared() is available])
 	ZFS_LINUX_TRY_COMPILE([
 		#include <linux/fs.h>
 		int iterate(struct file *filp, struct dir_context * context)
@@ -10,34 +10,55 @@ AC_DEFUN([ZFS_AC_KERNEL_VFS_ITERATE], [
 
 		static const struct file_operations fops
 		    __attribute__ ((unused)) = {
-			.iterate	 = iterate,
+			.iterate_shared	 = iterate,
 		};
 	],[
 	],[
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_VFS_ITERATE, 1,
-		          [fops->iterate() is available])
+		AC_DEFINE(HAVE_VFS_ITERATE_SHARED, 1,
+		          [fops->iterate_shared() is available])
 	],[
 		AC_MSG_RESULT(no)
 
-		AC_MSG_CHECKING([whether fops->readdir() is available])
+		dnl #
+		dnl # 3.11 API change
+		dnl #
+		AC_MSG_CHECKING([whether fops->iterate() is available])
 		ZFS_LINUX_TRY_COMPILE([
 			#include <linux/fs.h>
-			int readdir(struct file *filp, void *entry, filldir_t func)
+			int iterate(struct file *filp, struct dir_context * context)
 			    { return 0; }
 
 			static const struct file_operations fops
 			    __attribute__ ((unused)) = {
-				.readdir = readdir,
+				.iterate	 = iterate,
 			};
 		],[
 		],[
 			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_VFS_READDIR, 1,
-				  [fops->readdir() is available])
+			AC_DEFINE(HAVE_VFS_ITERATE, 1,
+				  [fops->iterate() is available])
 		],[
-			AC_MSG_ERROR(no; file a bug report with ZFSOnLinux)
-		])
+			AC_MSG_RESULT(no)
 
+			AC_MSG_CHECKING([whether fops->readdir() is available])
+			ZFS_LINUX_TRY_COMPILE([
+				#include <linux/fs.h>
+				int readdir(struct file *filp, void *entry, filldir_t func)
+				    { return 0; }
+
+				static const struct file_operations fops
+				    __attribute__ ((unused)) = {
+					.readdir = readdir,
+				};
+			],[
+			],[
+				AC_MSG_RESULT(yes)
+				AC_DEFINE(HAVE_VFS_READDIR, 1,
+					  [fops->readdir() is available])
+			],[
+				AC_MSG_ERROR(no; file a bug report with ZFSOnLinux)
+			])
+		])
 	])
 ])

config/kernel-vfs-rw-iterate.m4

@@ -1,5 +1,5 @@
 dnl #
-dnl # Linux 4.1.x API
+dnl # Linux 3.16 API
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_VFS_RW_ITERATE],
 	[AC_MSG_CHECKING([whether fops->read/write_iter() are available])
@@ -21,6 +21,47 @@ AC_DEFUN([ZFS_AC_KERNEL_VFS_RW_ITERATE],
 		AC_MSG_RESULT(yes)
 		AC_DEFINE(HAVE_VFS_RW_ITERATE, 1,
 			[fops->read/write_iter() are available])
+
+		ZFS_AC_KERNEL_NEW_SYNC_READ
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # Linux 4.1 API
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_NEW_SYNC_READ],
+	[AC_MSG_CHECKING([whether new_sync_read() is available])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+	],[
+		new_sync_read(NULL, NULL, 0, NULL);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_NEW_SYNC_READ, 1,
+			[new_sync_read() is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # Linux 4.1.x API
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_GENERIC_WRITE_CHECKS],
+	[AC_MSG_CHECKING([whether generic_write_checks() takes kiocb])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+
+	],[
+		struct kiocb *iocb = NULL;
+		struct iov_iter *iov = NULL;
+		generic_write_checks(iocb, iov);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_GENERIC_WRITE_CHECKS_KIOCB, 1,
+			[generic_write_checks() takes kiocb])
 	],[
 		AC_MSG_RESULT(no)
 	])

config/kernel-xattr-handler.m4

@@ -32,23 +32,72 @@ AC_DEFUN([ZFS_AC_KERNEL_CONST_XATTR_HANDLER], [
 	])
 ])
 
+dnl #
+dnl # 4.5 API change,
+dnl # struct xattr_handler added new member "name".
+dnl # xattr_handler which matches to whole name rather than prefix should use
+dnl # "name" instead of "prefix", e.g. "system.posix_acl_access"
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_NAME], [
+	AC_MSG_CHECKING([whether xattr_handler has name])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/xattr.h>
+
+		static const struct xattr_handler
+		    xops __attribute__ ((unused)) = {
+			.name = XATTR_NAME_POSIX_ACL_ACCESS,
+		};
+	],[
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_XATTR_HANDLER_NAME, 1,
+		    [xattr_handler has name])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # 4.9 API change,
+dnl # iops->{set,get,remove}xattr and generic_{set,get,remove}xattr are
+dnl # removed. xattr operations will directly go through sb->s_xattr.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_HAVE_GENERIC_SETXATTR], [
+	AC_MSG_CHECKING([whether generic_setxattr() exists])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/fs.h>
+		#include <linux/xattr.h>
+
+		static const struct inode_operations
+		    iops __attribute__ ((unused)) = {
+			.setxattr = generic_setxattr
+		};
+	],[
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_GENERIC_SETXATTR, 1,
+		    [generic_setxattr() exists])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
 dnl #
 dnl # Supported xattr handler get() interfaces checked newest to oldest.
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_GET], [
 	dnl #
-	dnl # 4.4 API change,
-	dnl # The xattr_handler->get() callback was changed to take a
-	dnl # attr_handler, and handler_flags argument was removed and
-	dnl # should be accessed by handler->flags.
+	dnl # 4.7 API change,
+	dnl # The xattr_handler->get() callback was changed to take both
+	dnl # dentry and inode.
 	dnl #
-	AC_MSG_CHECKING([whether xattr_handler->get() wants xattr_handler])
+	AC_MSG_CHECKING([whether xattr_handler->get() wants both dentry and inode])
 	ZFS_LINUX_TRY_COMPILE([
 		#include <linux/xattr.h>
 
 		int get(const struct xattr_handler *handler,
-		    struct dentry *dentry, const char *name,
-		    void *buffer, size_t size) { return 0; }
+		    struct dentry *dentry, struct inode *inode,
+		    const char *name, void *buffer, size_t size) { return 0; }
 		static const struct xattr_handler
 		    xops __attribute__ ((unused)) = {
 			.get = get,
@@ -56,23 +105,22 @@ AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_GET], [
 	],[
 	],[
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_XATTR_GET_HANDLER, 1,
-		    [xattr_handler->get() wants xattr_handler])
+		AC_DEFINE(HAVE_XATTR_GET_DENTRY_INODE, 1,
+		    [xattr_handler->get() wants both dentry and inode])
 	],[
 		dnl #
-		dnl # 2.6.33 API change,
-		dnl # The xattr_handler->get() callback was changed to take
-		dnl # a dentry instead of an inode, and a handler_flags
-		dnl # argument was added.
+		dnl # 4.4 API change,
+		dnl # The xattr_handler->get() callback was changed to take a
+		dnl # attr_handler, and handler_flags argument was removed and
+		dnl # should be accessed by handler->flags.
 		dnl #
-		AC_MSG_RESULT(no)
-		AC_MSG_CHECKING([whether xattr_handler->get() wants dentry])
+		AC_MSG_CHECKING([whether xattr_handler->get() wants xattr_handler])
 		ZFS_LINUX_TRY_COMPILE([
 			#include <linux/xattr.h>
 
-			int get(struct dentry *dentry, const char *name,
-			    void *buffer, size_t size, int handler_flags)
-			    { return 0; }
+			int get(const struct xattr_handler *handler,
+			    struct dentry *dentry, const char *name,
+			    void *buffer, size_t size) { return 0; }
 			static const struct xattr_handler
 			    xops __attribute__ ((unused)) = {
 				.get = get,
@@ -80,20 +128,23 @@ AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_GET], [
 		],[
 		],[
 			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_XATTR_GET_DENTRY, 1,
-			    [xattr_handler->get() wants dentry])
+			AC_DEFINE(HAVE_XATTR_GET_HANDLER, 1,
+			    [xattr_handler->get() wants xattr_handler])
 		],[
 			dnl #
-			dnl # 2.6.32 API
+			dnl # 2.6.33 API change,
+			dnl # The xattr_handler->get() callback was changed to take
+			dnl # a dentry instead of an inode, and a handler_flags
+			dnl # argument was added.
 			dnl #
 			AC_MSG_RESULT(no)
-			AC_MSG_CHECKING(
-			    [whether xattr_handler->get() wants inode])
+			AC_MSG_CHECKING([whether xattr_handler->get() wants dentry])
 			ZFS_LINUX_TRY_COMPILE([
 				#include <linux/xattr.h>
 
-				int get(struct inode *ip, const char *name,
-				    void *buffer, size_t size) { return 0; }
+				int get(struct dentry *dentry, const char *name,
+				    void *buffer, size_t size, int handler_flags)
+				    { return 0; }
 				static const struct xattr_handler
 				    xops __attribute__ ((unused)) = {
 					.get = get,
@@ -101,10 +152,32 @@ AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_GET], [
 			],[
 			],[
 				AC_MSG_RESULT(yes)
-				AC_DEFINE(HAVE_XATTR_GET_INODE, 1,
-				    [xattr_handler->get() wants inode])
+				AC_DEFINE(HAVE_XATTR_GET_DENTRY, 1,
+				    [xattr_handler->get() wants dentry])
 			],[
-	                        AC_MSG_ERROR([no; please file a bug report])
+				dnl #
+				dnl # 2.6.32 API
+				dnl #
+				AC_MSG_RESULT(no)
+				AC_MSG_CHECKING(
+				    [whether xattr_handler->get() wants inode])
+				ZFS_LINUX_TRY_COMPILE([
+					#include <linux/xattr.h>
+
+					int get(struct inode *ip, const char *name,
+					    void *buffer, size_t size) { return 0; }
+					static const struct xattr_handler
+					    xops __attribute__ ((unused)) = {
+						.get = get,
+					};
+				],[
+				],[
+					AC_MSG_RESULT(yes)
+					AC_DEFINE(HAVE_XATTR_GET_INODE, 1,
+					    [xattr_handler->get() wants inode])
+				],[
+					AC_MSG_ERROR([no; please file a bug report])
+				])
 			])
 		])
 	])
@@ -115,18 +188,18 @@ dnl # Supported xattr handler set() interfaces checked newest to oldest.
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_SET], [
 	dnl #
-	dnl # 4.4 API change,
-	dnl # The xattr_handler->set() callback was changed to take a
-	dnl # xattr_handler, and handler_flags argument was removed and
-	dnl # should be accessed by handler->flags.
+	dnl # 4.7 API change,
+	dnl # The xattr_handler->set() callback was changed to take both
+	dnl # dentry and inode.
 	dnl #
-	AC_MSG_CHECKING([whether xattr_handler->set() wants xattr_handler])
+	AC_MSG_CHECKING([whether xattr_handler->set() wants both dentry and inode])
 	ZFS_LINUX_TRY_COMPILE([
 		#include <linux/xattr.h>
 
 		int set(const struct xattr_handler *handler,
-		    struct dentry *dentry, const char *name,
-		    const void *buffer, size_t size, int flags)
+		    struct dentry *dentry, struct inode *inode,
+		    const char *name, const void *buffer,
+		    size_t size, int flags)
 		    { return 0; }
 		static const struct xattr_handler
 		    xops __attribute__ ((unused)) = {
@@ -135,23 +208,23 @@ AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_SET], [
 	],[
 	],[
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_XATTR_SET_HANDLER, 1,
-		    [xattr_handler->set() wants xattr_handler])
+		AC_DEFINE(HAVE_XATTR_SET_DENTRY_INODE, 1,
+		    [xattr_handler->set() wants both dentry and inode])
 	],[
 		dnl #
-		dnl # 2.6.33 API change,
+		dnl # 4.4 API change,
 		dnl # The xattr_handler->set() callback was changed to take a
-		dnl # dentry instead of an inode, and a handler_flags
-		dnl # argument was added.
+		dnl # xattr_handler, and handler_flags argument was removed and
+		dnl # should be accessed by handler->flags.
 		dnl #
-		AC_MSG_RESULT(no)
-		AC_MSG_CHECKING([whether xattr_handler->set() wants dentry])
+		AC_MSG_CHECKING([whether xattr_handler->set() wants xattr_handler])
 		ZFS_LINUX_TRY_COMPILE([
 			#include <linux/xattr.h>
 
-			int set(struct dentry *dentry, const char *name,
-			    const void *buffer, size_t size, int flags,
-			    int handler_flags) { return 0; }
+			int set(const struct xattr_handler *handler,
+			    struct dentry *dentry, const char *name,
+			    const void *buffer, size_t size, int flags)
+			    { return 0; }
 			static const struct xattr_handler
 			    xops __attribute__ ((unused)) = {
 				.set = set,
@@ -159,21 +232,23 @@ AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_SET], [
 		],[
 		],[
 			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_XATTR_SET_DENTRY, 1,
-			    [xattr_handler->set() wants dentry])
+			AC_DEFINE(HAVE_XATTR_SET_HANDLER, 1,
+			    [xattr_handler->set() wants xattr_handler])
 		],[
 			dnl #
-			dnl # 2.6.32 API
+			dnl # 2.6.33 API change,
+			dnl # The xattr_handler->set() callback was changed to take a
+			dnl # dentry instead of an inode, and a handler_flags
+			dnl # argument was added.
 			dnl #
 			AC_MSG_RESULT(no)
-			AC_MSG_CHECKING(
-			    [whether xattr_handler->set() wants inode])
+			AC_MSG_CHECKING([whether xattr_handler->set() wants dentry])
 			ZFS_LINUX_TRY_COMPILE([
 				#include <linux/xattr.h>
 
-				int set(struct inode *ip, const char *name,
-				    const void *buffer, size_t size, int flags)
-				    { return 0; }
+				int set(struct dentry *dentry, const char *name,
+				    const void *buffer, size_t size, int flags,
+				    int handler_flags) { return 0; }
 				static const struct xattr_handler
 				    xops __attribute__ ((unused)) = {
 					.set = set,
@@ -181,10 +256,33 @@ AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_SET], [
 			],[
 			],[
 				AC_MSG_RESULT(yes)
-				AC_DEFINE(HAVE_XATTR_SET_INODE, 1,
-				    [xattr_handler->set() wants inode])
+				AC_DEFINE(HAVE_XATTR_SET_DENTRY, 1,
+				    [xattr_handler->set() wants dentry])
 			],[
-	                        AC_MSG_ERROR([no; please file a bug report])
+				dnl #
+				dnl # 2.6.32 API
+				dnl #
+				AC_MSG_RESULT(no)
+				AC_MSG_CHECKING(
+				    [whether xattr_handler->set() wants inode])
+				ZFS_LINUX_TRY_COMPILE([
+					#include <linux/xattr.h>
+
+					int set(struct inode *ip, const char *name,
+					    const void *buffer, size_t size, int flags)
+					    { return 0; }
+					static const struct xattr_handler
+					    xops __attribute__ ((unused)) = {
+						.set = set,
+					};
+				],[
+				],[
+					AC_MSG_RESULT(yes)
+					AC_DEFINE(HAVE_XATTR_SET_INODE, 1,
+					    [xattr_handler->set() wants inode])
+				],[
+		                        AC_MSG_ERROR([no; please file a bug report])
+				])
 			])
 		])
 	])

config/kernel.m4

@@ -8,6 +8,7 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_CONFIG
 	ZFS_AC_KERNEL_DECLARE_EVENT_CLASS
 	ZFS_AC_KERNEL_CURRENT_BIO_TAIL
+	ZFS_AC_KERNEL_SUBMIT_BIO
 	ZFS_AC_KERNEL_BDEV_BLOCK_DEVICE_OPERATIONS
 	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID
 	ZFS_AC_KERNEL_TYPE_FMODE_T
@@ -22,31 +23,43 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_BIO_BVEC_ITER
 	ZFS_AC_KERNEL_BIO_FAILFAST_DTD
 	ZFS_AC_KERNEL_REQ_FAILFAST_MASK
+	ZFS_AC_KERNEL_REQ_OP_DISCARD
+	ZFS_AC_KERNEL_REQ_OP_SECURE_ERASE
+	ZFS_AC_KERNEL_REQ_OP_FLUSH
+	ZFS_AC_KERNEL_BIO_BI_OPF
 	ZFS_AC_KERNEL_BIO_END_IO_T_ARGS
 	ZFS_AC_KERNEL_BIO_RW_BARRIER
 	ZFS_AC_KERNEL_BIO_RW_DISCARD
 	ZFS_AC_KERNEL_BLK_QUEUE_FLUSH
 	ZFS_AC_KERNEL_BLK_QUEUE_MAX_HW_SECTORS
 	ZFS_AC_KERNEL_BLK_QUEUE_MAX_SEGMENTS
+	ZFS_AC_KERNEL_BLK_QUEUE_HAVE_BIO_RW_UNPLUG
+	ZFS_AC_KERNEL_BLK_QUEUE_HAVE_BLK_PLUG
 	ZFS_AC_KERNEL_GET_DISK_RO
 	ZFS_AC_KERNEL_GET_GENDISK
+	ZFS_AC_KERNEL_HAVE_BIO_SET_OP_ATTRS
+	ZFS_AC_KERNEL_GENERIC_READLINK_GLOBAL
 	ZFS_AC_KERNEL_DISCARD_GRANULARITY
 	ZFS_AC_KERNEL_CONST_XATTR_HANDLER
+	ZFS_AC_KERNEL_XATTR_HANDLER_NAME
 	ZFS_AC_KERNEL_XATTR_HANDLER_GET
 	ZFS_AC_KERNEL_XATTR_HANDLER_SET
 	ZFS_AC_KERNEL_XATTR_HANDLER_LIST
 	ZFS_AC_KERNEL_INODE_OWNER_OR_CAPABLE
 	ZFS_AC_KERNEL_POSIX_ACL_FROM_XATTR_USERNS
 	ZFS_AC_KERNEL_POSIX_ACL_RELEASE
+	ZFS_AC_KERNEL_SET_CACHED_ACL_USABLE
 	ZFS_AC_KERNEL_POSIX_ACL_CHMOD
-	ZFS_AC_KERNEL_POSIX_ACL_CACHING
 	ZFS_AC_KERNEL_POSIX_ACL_EQUIV_MODE_WANTS_UMODE_T
+	ZFS_AC_KERNEL_POSIX_ACL_VALID_WITH_NS
 	ZFS_AC_KERNEL_INODE_OPERATIONS_PERMISSION
 	ZFS_AC_KERNEL_INODE_OPERATIONS_PERMISSION_WITH_NAMEIDATA
 	ZFS_AC_KERNEL_INODE_OPERATIONS_CHECK_ACL
 	ZFS_AC_KERNEL_INODE_OPERATIONS_CHECK_ACL_WITH_FLAGS
 	ZFS_AC_KERNEL_INODE_OPERATIONS_GET_ACL
-	ZFS_AC_KERNEL_CURRENT_UMASK
+	ZFS_AC_KERNEL_INODE_OPERATIONS_SET_ACL
+	ZFS_AC_KERNEL_INODE_OPERATIONS_GETATTR
+	ZFS_AC_KERNEL_GET_ACL_HANDLE_CACHE
 	ZFS_AC_KERNEL_SHOW_OPTIONS
 	ZFS_AC_KERNEL_FILE_INODE
 	ZFS_AC_KERNEL_FSYNC
@@ -55,6 +68,7 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_NR_CACHED_OBJECTS
 	ZFS_AC_KERNEL_FREE_CACHED_OBJECTS
 	ZFS_AC_KERNEL_FALLOCATE
+	ZFS_AC_KERNEL_AIO_FSYNC
 	ZFS_AC_KERNEL_MKDIR_UMODE_T
 	ZFS_AC_KERNEL_LOOKUP_NAMEIDATA
 	ZFS_AC_KERNEL_CREATE_NAMEIDATA
@@ -65,6 +79,7 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_ENCODE_FH_WITH_INODE
 	ZFS_AC_KERNEL_COMMIT_METADATA
 	ZFS_AC_KERNEL_CLEAR_INODE
+	ZFS_AC_KERNEL_SETATTR_PREPARE
 	ZFS_AC_KERNEL_INSERT_INODE_LOCKED
 	ZFS_AC_KERNEL_D_MAKE_ROOT
 	ZFS_AC_KERNEL_D_OBTAIN_ALIAS
@@ -81,17 +96,21 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_SHRINK_CONTROL_HAS_NID
 	ZFS_AC_KERNEL_S_INSTANCES_LIST_HEAD
 	ZFS_AC_KERNEL_S_D_OP
-	ZFS_AC_KERNEL_BDI_SETUP_AND_REGISTER
+	ZFS_AC_KERNEL_BDI
 	ZFS_AC_KERNEL_SET_NLINK
 	ZFS_AC_KERNEL_ELEVATOR_CHANGE
 	ZFS_AC_KERNEL_5ARG_SGET
 	ZFS_AC_KERNEL_LSEEK_EXECUTE
 	ZFS_AC_KERNEL_VFS_ITERATE
 	ZFS_AC_KERNEL_VFS_RW_ITERATE
+	ZFS_AC_KERNEL_GENERIC_WRITE_CHECKS
 	ZFS_AC_KERNEL_KMAP_ATOMIC_ARGS
 	ZFS_AC_KERNEL_FOLLOW_DOWN_ONE
 	ZFS_AC_KERNEL_MAKE_REQUEST_FN
 	ZFS_AC_KERNEL_GENERIC_IO_ACCT
+	ZFS_AC_KERNEL_RENAME_WANTS_FLAGS
+	ZFS_AC_KERNEL_HAVE_GENERIC_SETXATTR
+	ZFS_AC_KERNEL_CURRENT_TIME
 
 	AS_IF([test "$LINUX_OBJ" != "$LINUX"], [
 		KERNELMAKE_PARAMS="$KERNELMAKE_PARAMS O=$LINUX_OBJ"
@@ -462,9 +481,35 @@ AC_DEFUN([ZFS_AC_KERNEL_CONFIG], [
 		])
 	])
 
+	ZFS_AC_KERNEL_CONFIG_THREAD_SIZE
 	ZFS_AC_KERNEL_CONFIG_DEBUG_LOCK_ALLOC
 ])
 
+dnl #
+dnl # Check configured THREAD_SIZE
+dnl #
+dnl # The stack size will vary by architecture, but as of Linux 3.15 on x86_64
+dnl # the default thread stack size was increased to 16K from 8K.  Therefore,
+dnl # on newer kernels and some architectures stack usage optimizations can be
+dnl # conditionally applied to improve performance without negatively impacting
+dnl # stability.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_CONFIG_THREAD_SIZE], [
+	AC_MSG_CHECKING([whether kernel was built with 16K or larger stacks])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/module.h>
+	],[
+		#if (THREAD_SIZE < 16384)
+		#error "THREAD_SIZE is less than 16K"
+		#endif
+	],[
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_LARGE_STACKS, 1, [kernel has large stacks])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])
+
 dnl #
 dnl # Check CONFIG_DEBUG_LOCK_ALLOC
 dnl #
@@ -574,7 +619,7 @@ dnl #
 dnl # ZFS_LINUX_CONFIG
 dnl #
 AC_DEFUN([ZFS_LINUX_CONFIG],
-	[AC_MSG_CHECKING([whether Linux was built with CONFIG_$1])
+	[AC_MSG_CHECKING([whether kernel was built with CONFIG_$1])
 	ZFS_LINUX_TRY_COMPILE([
 		#include <linux/module.h>
 	],[

config/user-makedev.m4

@@ -0,0 +1,39 @@
+dnl #
+dnl # glibc 2.25
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_USER_MAKEDEV_IN_SYSMACROS], [
+	AC_MSG_CHECKING([makedev() is declared in sys/sysmacros.h])
+	AC_TRY_COMPILE(
+	[
+		#include <sys/sysmacros.h>
+	],[
+		int k;
+		k = makedev(0,0);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_MAKEDEV_IN_SYSMACROS, 1,
+		    [makedev() is declared in sys/sysmacros.h])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # glibc X < Y < 2.25
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_USER_MAKEDEV_IN_MKDEV], [
+	AC_MSG_CHECKING([makedev() is declared in sys/mkdev.h])
+	AC_TRY_COMPILE(
+	[
+		#include <sys/mkdev.h>
+	],[
+		int k;
+		k = makedev(0,0);
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_MAKEDEV_IN_MKDEV, 1,
+		    [makedev() is declared in sys/mkdev.h])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])

config/user-no-format-truncation.m4

@@ -0,0 +1,22 @@
+dnl #
+dnl # Check if gcc supports -Wno-format-truncation option.
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_USER_NO_FORMAT_TRUNCATION], [
+	AC_MSG_CHECKING([for -Wno-format-truncation support])
+
+	saved_flags="$CFLAGS"
+	CFLAGS="$CFLAGS -Wno-format-truncation"
+
+	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])],
+	[
+		NO_FORMAT_TRUNCATION=-Wno-format-truncation
+		AC_MSG_RESULT([yes])
+	],
+	[
+		NO_FORMAT_TRUNCATION=
+		AC_MSG_RESULT([no])
+	])
+
+	CFLAGS="$saved_flags"
+	AC_SUBST([NO_FORMAT_TRUNCATION])
+])

config/user.m4

@@ -13,6 +13,9 @@ AC_DEFUN([ZFS_AC_CONFIG_USER], [
 	ZFS_AC_CONFIG_USER_LIBBLKID
 	ZFS_AC_CONFIG_USER_FRAME_LARGER_THAN
 	ZFS_AC_CONFIG_USER_RUNSTATEDIR
+	ZFS_AC_CONFIG_USER_MAKEDEV_IN_SYSMACROS
+	ZFS_AC_CONFIG_USER_MAKEDEV_IN_MKDEV
+	ZFS_AC_CONFIG_USER_NO_FORMAT_TRUNCATION
 dnl #
 dnl #	Checks for library functions
 	AC_CHECK_FUNCS([mlockall])

etc/modules-load.d/zfs.conf.in

@@ -1 +1,3 @@
-zfs
+# Always load kernel modules at boot.  The default behavior is to load the
+# kernel modules in the zfs-import-*.service or when blkid(8) detects a pool.
+#zfs

etc/systemd/system/50-zfs.preset.in

@@ -1,2 +1,7 @@
 # ZFS is enabled by default
-enable zfs.*
+enable zfs-import-cache.service
+disable zfs-import-scan.service
+enable zfs-mount.service
+enable zfs-share.service
+enable zfs-zed.service
+enable zfs.target

etc/systemd/system/Makefile.am

@@ -2,7 +2,7 @@ systemdpreset_DATA = \
 	50-zfs.preset
 
 systemdunit_DATA = \
-	zed.service \
+	zfs-zed.service \
 	zfs-import-cache.service \
 	zfs-import-scan.service \
 	zfs-mount.service \
@@ -10,7 +10,7 @@ systemdunit_DATA = \
 	zfs.target
 
 EXTRA_DIST = \
-	$(top_srcdir)/etc/systemd/system/zed.service.in \
+	$(top_srcdir)/etc/systemd/system/zfs-zed.service.in \
 	$(top_srcdir)/etc/systemd/system/zfs-import-cache.service.in \
 	$(top_srcdir)/etc/systemd/system/zfs-import-scan.service.in \
 	$(top_srcdir)/etc/systemd/system/zfs-mount.service.in \

etc/systemd/system/zfs-import-cache.service.in

@@ -4,6 +4,7 @@ DefaultDependencies=no
 Requires=systemd-udev-settle.service
 After=systemd-udev-settle.service
 After=cryptsetup.target
+After=systemd-remount-fs.service
 ConditionPathExists=@sysconfdir@/zfs/zpool.cache
 
 [Service]
@@ -11,3 +12,7 @@ Type=oneshot
 RemainAfterExit=yes
 ExecStartPre=/sbin/modprobe zfs
 ExecStart=@sbindir@/zpool import -c @sysconfdir@/zfs/zpool.cache -aN
+
+[Install]
+WantedBy=zfs-mount.service
+WantedBy=zfs.target

etc/systemd/system/zfs-import-scan.service.in

@@ -10,4 +10,8 @@ ConditionPathExists=!@sysconfdir@/zfs/zpool.cache
 Type=oneshot
 RemainAfterExit=yes
 ExecStartPre=/sbin/modprobe zfs
-ExecStart=@sbindir@/zpool import -d /dev/disk/by-id -aN
+ExecStart=@sbindir@/zpool import -aN -o cachefile=none
+
+[Install]
+WantedBy=zfs-mount.service
+WantedBy=zfs.target

etc/systemd/system/zfs-mount.service.in

@@ -1,15 +1,18 @@
 [Unit]
 Description=Mount ZFS filesystems
 DefaultDependencies=no
-Wants=zfs-import-cache.service
-Wants=zfs-import-scan.service
-Requires=systemd-udev-settle.service
 After=systemd-udev-settle.service
 After=zfs-import-cache.service
 After=zfs-import-scan.service
+After=systemd-remount-fs.service
 Before=local-fs.target
 
 [Service]
 Type=oneshot
 RemainAfterExit=yes
 ExecStart=@sbindir@/zfs mount -a
+WorkingDirectory=-/sbin/
+
+[Install]
+WantedBy=zfs-share.service
+WantedBy=zfs.target

etc/systemd/system/zfs-share.service.in

@@ -1,14 +1,16 @@
 [Unit]
 Description=ZFS file system shares
-After=nfs-server.service
+After=nfs-server.service nfs-kernel-server.service
 After=smb.service
 After=zfs-mount.service
-Requires=zfs-mount.service
-PartOf=nfs-server.service
+PartOf=nfs-server.service nfs-kernel-server.service
 PartOf=smb.service
 
 [Service]
 Type=oneshot
 RemainAfterExit=yes
-ExecStartPre=-@bindir@/rm /etc/dfs/sharetab
+ExecStartPre=-@bindir@/rm -f /etc/dfs/sharetab
 ExecStart=@sbindir@/zfs share -a
+
+[Install]
+WantedBy=zfs.target

etc/systemd/system/zfs-zed.service.in

@@ -7,3 +7,7 @@ After=zfs-import-scan.service
 [Service]
 ExecStart=@sbindir@/zed -F
 Restart=on-abort
+
+[Install]
+Alias=zed.service
+WantedBy=zfs.target

etc/systemd/system/zfs.target.in

@@ -1,8 +1,5 @@
 [Unit]
 Description=ZFS startup target
-Requires=zfs-mount.service
-Requires=zfs-share.service
-Wants=zed.service
 
 [Install]
 WantedBy=multi-user.target

include/libzfs_impl.h

@@ -46,11 +46,6 @@
 extern "C" {
 #endif
 
-#ifdef	VERIFY
-#undef	VERIFY
-#endif
-#define	VERIFY	verify
-
 typedef struct libzfs_fru {
 	char *zf_device;
 	char *zf_fru;

include/linux/blkdev_compat.h

@@ -37,21 +37,48 @@ typedef unsigned __bitwise__ fmode_t;
 #endif /* HAVE_FMODE_T */
 
 /*
- * 2.6.36 API change,
+ * 4.7 - 4.x API,
+ * The blk_queue_write_cache() interface has replaced blk_queue_flush()
+ * interface.  However, the new interface is GPL-only thus we implement
+ * our own trivial wrapper when the GPL-only version is detected.
+ *
+ * 2.6.36 - 4.6 API,
  * The blk_queue_flush() interface has replaced blk_queue_ordered()
  * interface.  However, while the old interface was available to all the
  * new one is GPL-only.   Thus if the GPL-only version is detected we
- * implement our own trivial helper compatibility funcion.   The hope is
- * that long term this function will be opened up.
+ * implement our own trivial helper.
+ *
+ * 2.6.x - 2.6.35
+ * Legacy blk_queue_ordered() interface.
  */
-#if defined(HAVE_BLK_QUEUE_FLUSH) && defined(HAVE_BLK_QUEUE_FLUSH_GPL_ONLY)
-#define	blk_queue_flush __blk_queue_flush
 static inline void
-__blk_queue_flush(struct request_queue *q, unsigned int flags)
+blk_queue_set_write_cache(struct request_queue *q, bool wc, bool fua)
 {
-	q->flush_flags = flags & (REQ_FLUSH | REQ_FUA);
+#if defined(HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY)
+	spin_lock_irq(q->queue_lock);
+	if (wc)
+		queue_flag_set(QUEUE_FLAG_WC, q);
+	else
+		queue_flag_clear(QUEUE_FLAG_WC, q);
+	if (fua)
+		queue_flag_set(QUEUE_FLAG_FUA, q);
+	else
+		queue_flag_clear(QUEUE_FLAG_FUA, q);
+	spin_unlock_irq(q->queue_lock);
+#elif defined(HAVE_BLK_QUEUE_WRITE_CACHE)
+	blk_queue_write_cache(q, wc, fua);
+#elif defined(HAVE_BLK_QUEUE_FLUSH_GPL_ONLY)
+	if (wc)
+		q->flush_flags |= REQ_FLUSH;
+	if (fua)
+		q->flush_flags |= REQ_FUA;
+#elif defined(HAVE_BLK_QUEUE_FLUSH)
+	blk_queue_flush(q, (wc ? REQ_FLUSH : 0) | (fua ? REQ_FUA : 0));
+#else
+	blk_queue_ordered(q, QUEUE_ORDERED_DRAIN, NULL);
+#endif
 }
-#endif /* HAVE_BLK_QUEUE_FLUSH && HAVE_BLK_QUEUE_FLUSH_GPL_ONLY */
+
 /*
  * Most of the blk_* macros were removed in 2.6.36.  Ostensibly this was
  * done to improve readability and allow easier grepping.  However, from
@@ -234,12 +261,21 @@ bio_set_flags_failfast(struct block_device *bdev, int *flags)
 
 /*
  * 2.6.27 API change
- * The function was exported for use, prior to this it existed by the
+ * The function was exported for use, prior to this it existed but the
  * symbol was not exported.
+ *
+ * 4.4.0-6.21 API change for Ubuntu
+ * lookup_bdev() gained a second argument, FMODE_*, to check inode permissions.
  */
-#ifndef HAVE_LOOKUP_BDEV
-#define	lookup_bdev(path)		ERR_PTR(-ENOTSUP)
-#endif
+#ifdef HAVE_1ARG_LOOKUP_BDEV
+#define	vdev_lookup_bdev(path)	lookup_bdev(path)
+#else
+#ifdef HAVE_2ARGS_LOOKUP_BDEV
+#define	vdev_lookup_bdev(path)	lookup_bdev(path, 0)
+#else
+#define	vdev_lookup_bdev(path)	ERR_PTR(-ENOTSUP)
+#endif /* HAVE_2ARGS_LOOKUP_BDEV */
+#endif /* HAVE_1ARG_LOOKUP_BDEV */
 
 /*
  * 2.6.30 API change
@@ -265,48 +301,172 @@ bio_set_flags_failfast(struct block_device *bdev, int *flags)
 #endif /* HAVE_BDEV_LOGICAL_BLOCK_SIZE */
 #endif /* HAVE_BDEV_PHYSICAL_BLOCK_SIZE */
 
+#ifndef HAVE_BIO_SET_OP_ATTRS
 /*
- * 2.6.37 API change
- * The WRITE_FLUSH, WRITE_FUA, and WRITE_FLUSH_FUA flags have been
- * introduced as a replacement for WRITE_BARRIER.  This was done to
- * allow richer semantics to be expressed to the block layer.  It is
- * the block layers responsibility to choose the correct way to
- * implement these semantics.
- *
- * The existence of these flags implies that REQ_FLUSH an REQ_FUA are
- * defined.  Thus we can safely define VDEV_REQ_FLUSH and VDEV_REQ_FUA
- * compatibility macros.
+ * Kernels without bio_set_op_attrs use bi_rw for the bio flags.
  */
-#ifdef WRITE_FLUSH_FUA
-#define	VDEV_WRITE_FLUSH_FUA		WRITE_FLUSH_FUA
-#define	VDEV_REQ_FLUSH			REQ_FLUSH
-#define	VDEV_REQ_FUA			REQ_FUA
-#else
-#define	VDEV_WRITE_FLUSH_FUA		WRITE_BARRIER
-#ifdef HAVE_BIO_RW_BARRIER
-#define	VDEV_REQ_FLUSH			(1 << BIO_RW_BARRIER)
-#define	VDEV_REQ_FUA			(1 << BIO_RW_BARRIER)
-#else
-#define	VDEV_REQ_FLUSH			REQ_HARDBARRIER
-#define	VDEV_REQ_FUA			REQ_FUA
-#endif
+static inline void
+bio_set_op_attrs(struct bio *bio, unsigned rw, unsigned flags)
+{
+	bio->bi_rw |= rw | flags;
+}
 #endif
 
 /*
- * 2.6.32 API change
- * Use the normal I/O patch for discards.
+ * bio_set_flush - Set the appropriate flags in a bio to guarantee
+ * data are on non-volatile media on completion.
+ *
+ * 2.6.X - 2.6.36 API,
+ *   WRITE_BARRIER - Tells the block layer to commit all previously submitted
+ *   writes to stable storage before this one is started and that the current
+ *   write is on stable storage upon completion.  Also prevents reordering
+ *   on both sides of the current operation.
+ *
+ * 2.6.37 - 4.8 API,
+ *   Introduce  WRITE_FLUSH, WRITE_FUA, and WRITE_FLUSH_FUA flags as a
+ *   replacement for WRITE_BARRIER to allow expressing richer semantics
+ *   to the block layer.  It's up to the block layer to implement the
+ *   semantics correctly. Use the WRITE_FLUSH_FUA flag combination.
+ *
+ * 4.8 - 4.9 API,
+ *   REQ_FLUSH was renamed to REQ_PREFLUSH.  For consistency with previous
+ *   ZoL releases, prefer the WRITE_FLUSH_FUA flag set if it's available.
+ *
+ * 4.10 API,
+ *   The read/write flags and their modifiers, including WRITE_FLUSH,
+ *   WRITE_FUA and WRITE_FLUSH_FUA were removed from fs.h in
+ *   torvalds/linux@70fd7614 and replaced by direct flag modification
+ *   of the REQ_ flags in bio->bi_opf.  Use REQ_PREFLUSH.
  */
-#ifdef QUEUE_FLAG_DISCARD
-#ifdef HAVE_BIO_RW_DISCARD
-#define	VDEV_REQ_DISCARD		(1 << BIO_RW_DISCARD)
+static inline void
+bio_set_flush(struct bio *bio)
+{
+#if defined(REQ_PREFLUSH)	/* >= 4.10 */
+	bio_set_op_attrs(bio, 0, REQ_PREFLUSH);
+#elif defined(WRITE_FLUSH_FUA)	/* >= 2.6.37 and <= 4.9 */
+	bio_set_op_attrs(bio, 0, WRITE_FLUSH_FUA);
+#elif defined(WRITE_BARRIER)	/* < 2.6.37 */
+	bio_set_op_attrs(bio, 0, WRITE_BARRIER);
 #else
-#define	VDEV_REQ_DISCARD		REQ_DISCARD
+#error	"Allowing the build will cause bio_set_flush requests to be ignored."
+	"Please file an issue report at: "
+	"https://github.com/zfsonlinux/zfs/issues/new"
 #endif
+}
+
+/*
+ * 4.8 - 4.x API,
+ *   REQ_OP_FLUSH
+ *
+ * 4.8-rc0 - 4.8-rc1,
+ *   REQ_PREFLUSH
+ *
+ * 2.6.36 - 4.7 API,
+ *   REQ_FLUSH
+ *
+ * 2.6.x - 2.6.35 API,
+ *   HAVE_BIO_RW_BARRIER
+ *
+ * Used to determine if a cache flush has been requested.  This check has
+ * been left intentionally broad in order to cover both a legacy flush
+ * and the new preflush behavior introduced in Linux 4.8.  This is correct
+ * in all cases but may have a performance impact for some kernels.  It
+ * has the advantage of minimizing kernel specific changes in the zvol code.
+ *
+ */
+static inline boolean_t
+bio_is_flush(struct bio *bio)
+{
+#if defined(HAVE_REQ_OP_FLUSH) && defined(HAVE_BIO_BI_OPF)
+	return ((bio_op(bio) == REQ_OP_FLUSH) || (bio->bi_opf & REQ_PREFLUSH));
+#elif defined(REQ_PREFLUSH) && defined(HAVE_BIO_BI_OPF)
+	return (bio->bi_opf & REQ_PREFLUSH);
+#elif defined(REQ_PREFLUSH) && !defined(HAVE_BIO_BI_OPF)
+	return (bio->bi_rw & REQ_PREFLUSH);
+#elif defined(REQ_FLUSH)
+	return (bio->bi_rw & REQ_FLUSH);
+#elif defined(HAVE_BIO_RW_BARRIER)
+	return (bio->bi_rw & (1 << BIO_RW_BARRIER));
+#else
+#error	"Allowing the build will cause flush requests to be ignored. Please "
+	"file an issue report at: https://github.com/zfsonlinux/zfs/issues/new"
+#endif
+}
+
+/*
+ * 4.8 - 4.x API,
+ *   REQ_FUA flag moved to bio->bi_opf
+ *
+ * 2.6.x - 4.7 API,
+ *   REQ_FUA
+ */
+static inline boolean_t
+bio_is_fua(struct bio *bio)
+{
+#if defined(HAVE_BIO_BI_OPF)
+	return (bio->bi_opf & REQ_FUA);
+#elif defined(REQ_FUA)
+	return (bio->bi_rw & REQ_FUA);
+#else
+#error	"Allowing the build will cause fua requests to be ignored. Please "
+	"file an issue report at: https://github.com/zfsonlinux/zfs/issues/new"
+#endif
+}
+
+/*
+ * 4.8 - 4.x API,
+ *   REQ_OP_DISCARD
+ *
+ * 2.6.36 - 4.7 API,
+ *   REQ_DISCARD
+ *
+ * 2.6.28 - 2.6.35 API,
+ *   BIO_RW_DISCARD
+ *
+ * In all cases the normal I/O path is used for discards.  The only
+ * difference is how the kernel tags individual I/Os as discards.
+ *
+ * Note that 2.6.32 era kernels provide both BIO_RW_DISCARD and REQ_DISCARD,
+ * where BIO_RW_DISCARD is the correct interface.  Therefore, it is important
+ * that the HAVE_BIO_RW_DISCARD check occur before the REQ_DISCARD check.
+ */
+static inline boolean_t
+bio_is_discard(struct bio *bio)
+{
+#if defined(HAVE_REQ_OP_DISCARD)
+	return (bio_op(bio) == REQ_OP_DISCARD);
+#elif defined(HAVE_BIO_RW_DISCARD)
+	return (bio->bi_rw & (1 << BIO_RW_DISCARD));
+#elif defined(REQ_DISCARD)
+	return (bio->bi_rw & REQ_DISCARD);
 #else
 #error	"Allowing the build will cause discard requests to become writes "
-	"potentially triggering the DMU_MAX_ACCESS assertion. Please file a "
+	"potentially triggering the DMU_MAX_ACCESS assertion. Please file "
 	"an issue report at: https://github.com/zfsonlinux/zfs/issues/new"
 #endif
+}
+
+/*
+ * 4.8 - 4.x API,
+ *   REQ_OP_SECURE_ERASE
+ *
+ * 2.6.36 - 4.7 API,
+ *   REQ_SECURE
+ *
+ * 2.6.x - 2.6.35 API,
+ *   Unsupported by kernel
+ */
+static inline boolean_t
+bio_is_secure_erase(struct bio *bio)
+{
+#if defined(HAVE_REQ_OP_SECURE_ERASE)
+	return (bio_op(bio) == REQ_OP_SECURE_ERASE);
+#elif defined(REQ_SECURE)
+	return (bio->bi_rw & REQ_SECURE);
+#else
+	return (0);
+#endif
+}
 
 /*
  * 2.6.33 API change

include/linux/vfs_compat.h

@@ -69,45 +69,115 @@ truncate_setsize(struct inode *ip, loff_t new)
 /*
  * 2.6.32 - 2.6.33, bdi_setup_and_register() is not available.
  * 2.6.34 - 3.19, bdi_setup_and_register() takes 3 arguments.
- * 4.0 - x.y, bdi_setup_and_register() takes 2 arguments.
+ * 4.0 - 4.11, bdi_setup_and_register() takes 2 arguments.
+ * 4.12 - x.y, super_setup_bdi_name() new interface.
  */
-#if defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER)
+#if defined(HAVE_SUPER_SETUP_BDI_NAME)
+extern atomic_long_t zfs_bdi_seq;
+
 static inline int
-zpl_bdi_setup_and_register(struct backing_dev_info *bdi, char *name)
+zpl_bdi_setup(struct super_block *sb, char *name)
 {
-	return (bdi_setup_and_register(bdi, name));
+	return super_setup_bdi_name(sb, "%.28s-%ld", name,
+	    atomic_long_inc_return(&zfs_bdi_seq));
+}
+static inline void
+zpl_bdi_destroy(struct super_block *sb)
+{
+}
+#elif defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER)
+static inline int
+zpl_bdi_setup(struct super_block *sb, char *name)
+{
+	struct backing_dev_info *bdi;
+	int error;
+
+	bdi = kmem_zalloc(sizeof (struct backing_dev_info), KM_SLEEP);
+	error = bdi_setup_and_register(bdi, name);
+	if (error) {
+		kmem_free(bdi, sizeof (struct backing_dev_info));
+		return (error);
+	}
+
+	sb->s_bdi = bdi;
+
+	return (0);
+}
+static inline void
+zpl_bdi_destroy(struct super_block *sb)
+{
+	struct backing_dev_info *bdi = sb->s_bdi;
+
+	bdi_destroy(bdi);
+	kmem_free(bdi, sizeof (struct backing_dev_info));
+	sb->s_bdi = NULL;
 }
 #elif defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER)
 static inline int
-zpl_bdi_setup_and_register(struct backing_dev_info *bdi, char *name)
+zpl_bdi_setup(struct super_block *sb, char *name)
 {
-	return (bdi_setup_and_register(bdi, name, BDI_CAP_MAP_COPY));
+	struct backing_dev_info *bdi;
+	int error;
+
+	bdi = kmem_zalloc(sizeof (struct backing_dev_info), KM_SLEEP);
+	error = bdi_setup_and_register(bdi, name, BDI_CAP_MAP_COPY);
+	if (error) {
+		kmem_free(sb->s_bdi, sizeof (struct backing_dev_info));
+		return (error);
+	}
+
+	sb->s_bdi = bdi;
+
+	return (0);
+}
+static inline void
+zpl_bdi_destroy(struct super_block *sb)
+{
+	struct backing_dev_info *bdi = sb->s_bdi;
+
+	bdi_destroy(bdi);
+	kmem_free(bdi, sizeof (struct backing_dev_info));
+	sb->s_bdi = NULL;
 }
 #else
 extern atomic_long_t zfs_bdi_seq;
 
 static inline int
-zpl_bdi_setup_and_register(struct backing_dev_info *bdi, char *name)
+zpl_bdi_setup(struct super_block *sb, char *name)
 {
-	char tmp[32];
+	struct backing_dev_info *bdi;
 	int error;
 
+	bdi = kmem_zalloc(sizeof (struct backing_dev_info), KM_SLEEP);
 	bdi->name = name;
 	bdi->capabilities = BDI_CAP_MAP_COPY;
 
 	error = bdi_init(bdi);
-	if (error)
-		return (error);
-
-	sprintf(tmp, "%.28s%s", name, "-%d");
-	error = bdi_register(bdi, NULL, tmp,
-	    atomic_long_inc_return(&zfs_bdi_seq));
 	if (error) {
-		bdi_destroy(bdi);
+		kmem_free(bdi, sizeof (struct backing_dev_info));
 		return (error);
 	}
 
-	return (error);
+	error = bdi_register(bdi, NULL, "%.28s-%ld", name,
+	    atomic_long_inc_return(&zfs_bdi_seq));
+	if (error) {
+		bdi_destroy(bdi);
+		kmem_free(bdi, sizeof (struct backing_dev_info));
+		return (error);
+	}
+
+	sb->s_bdi = bdi;
+
+	return (0);
+}
+static inline void
+zpl_bdi_destroy(struct super_block *sb)
+{
+	struct backing_dev_info *bdi = sb->s_bdi;
+
+	bdi_destroy(bdi);
+	kmem_free(bdi, sizeof (struct backing_dev_info));
+	sb->s_bdi = NULL;
 }
 #endif
 
@@ -202,22 +272,11 @@ lseek_execute(
  * At 60 seconds the kernel will also begin issuing RCU stall warnings.
  */
 #include <linux/posix_acl.h>
-#ifndef HAVE_POSIX_ACL_CACHING
-#define	ACL_NOT_CACHED ((void *)(-1))
-#endif /* HAVE_POSIX_ACL_CACHING */
 
 #if defined(HAVE_POSIX_ACL_RELEASE) && !defined(HAVE_POSIX_ACL_RELEASE_GPL_ONLY)
-
 #define	zpl_posix_acl_release(arg)		posix_acl_release(arg)
-#define	zpl_set_cached_acl(ip, ty, n)		set_cached_acl(ip, ty, n)
-#define	zpl_forget_cached_acl(ip, ty)		forget_cached_acl(ip, ty)
-
 #else
-
-static inline void
-zpl_posix_acl_free(void *arg) {
-	kfree(arg);
-}
+void zpl_posix_acl_release_impl(struct posix_acl *);
 
 static inline void
 zpl_posix_acl_release(struct posix_acl *acl)
@@ -225,15 +284,17 @@ zpl_posix_acl_release(struct posix_acl *acl)
 	if ((acl == NULL) || (acl == ACL_NOT_CACHED))
 		return;
 
-	if (atomic_dec_and_test(&acl->a_refcount)) {
-		taskq_dispatch_delay(system_taskq, zpl_posix_acl_free, acl,
-		    TQ_SLEEP, ddi_get_lbolt() + 60*HZ);
-	}
+	if (atomic_dec_and_test(&acl->a_refcount))
+		zpl_posix_acl_release_impl(acl);
 }
+#endif /* HAVE_POSIX_ACL_RELEASE */
 
+#ifdef HAVE_SET_CACHED_ACL_USABLE
+#define	zpl_set_cached_acl(ip, ty, n)		set_cached_acl(ip, ty, n)
+#define	zpl_forget_cached_acl(ip, ty)		forget_cached_acl(ip, ty)
+#else
 static inline void
 zpl_set_cached_acl(struct inode *ip, int type, struct posix_acl *newer) {
-#ifdef HAVE_POSIX_ACL_CACHING
 	struct posix_acl *older = NULL;
 
 	spin_lock(&ip->i_lock);
@@ -255,14 +316,13 @@ zpl_set_cached_acl(struct inode *ip, int type, struct posix_acl *newer) {
 	spin_unlock(&ip->i_lock);
 
 	zpl_posix_acl_release(older);
-#endif /* HAVE_POSIX_ACL_CACHING */
 }
 
 static inline void
 zpl_forget_cached_acl(struct inode *ip, int type) {
 	zpl_set_cached_acl(ip, type, (struct posix_acl *)ACL_NOT_CACHED);
 }
-#endif /* HAVE_POSIX_ACL_RELEASE */
+#endif /* HAVE_SET_CACHED_ACL_USABLE */
 
 #ifndef HAVE___POSIX_ACL_CHMOD
 #ifdef HAVE_POSIX_ACL_CHMOD
@@ -320,15 +380,19 @@ typedef umode_t zpl_equivmode_t;
 #else
 typedef mode_t zpl_equivmode_t;
 #endif /* HAVE_POSIX_ACL_EQUIV_MODE_UMODE_T */
-#endif /* CONFIG_FS_POSIX_ACL */
 
-#ifndef HAVE_CURRENT_UMASK
-static inline int
-current_umask(void)
-{
-	return (current->fs->umask);
-}
-#endif /* HAVE_CURRENT_UMASK */
+/*
+ * 4.8 API change,
+ * posix_acl_valid() now must be passed a namespace, the namespace from
+ * from super block associated with the given inode is used for this purpose.
+ */
+#ifdef HAVE_POSIX_ACL_VALID_WITH_NS
+#define	zpl_posix_acl_valid(ip, acl)  posix_acl_valid(ip->i_sb->s_user_ns, acl)
+#else
+#define	zpl_posix_acl_valid(ip, acl)  posix_acl_valid(acl)
+#endif
+
+#endif /* CONFIG_FS_POSIX_ACL */
 
 /*
  * 2.6.38 API change,
@@ -363,4 +427,69 @@ static inline struct inode *file_inode(const struct file *f)
 #define	zpl_follow_up(path)			follow_up(path)
 #endif
 
+/*
+ * 4.9 API change
+ */
+#ifndef HAVE_SETATTR_PREPARE
+static inline int
+setattr_prepare(struct dentry *dentry, struct iattr *ia)
+{
+	return (inode_change_ok(dentry->d_inode, ia));
+}
+#endif
+
+/*
+ * 4.11 API change
+ * These macros are defined by kernel 4.11.  We define them so that the same
+ * code builds under kernels < 4.11 and >= 4.11.  The macros are set to 0 so
+ * that it will create obvious failures if they are accidentally used when built
+ * against a kernel >= 4.11.
+ */
+
+#ifndef STATX_BASIC_STATS
+#define	STATX_BASIC_STATS	0
+#endif
+
+#ifndef AT_STATX_SYNC_AS_STAT
+#define	AT_STATX_SYNC_AS_STAT	0
+#endif
+
+/*
+ * 4.11 API change
+ * 4.11 takes struct path *, < 4.11 takes vfsmount *
+ */
+
+#ifdef HAVE_VFSMOUNT_IOPS_GETATTR
+#define	ZPL_GETATTR_WRAPPER(func)					\
+static int								\
+func(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)	\
+{									\
+	struct path path = { .mnt = mnt, .dentry = dentry };		\
+	return func##_impl(&path, stat, STATX_BASIC_STATS,		\
+	    AT_STATX_SYNC_AS_STAT);					\
+}
+#elif defined(HAVE_PATH_IOPS_GETATTR)
+#define	ZPL_GETATTR_WRAPPER(func)					\
+static int								\
+func(const struct path *path, struct kstat *stat, u32 request_mask,	\
+    unsigned int query_flags)						\
+{									\
+	return (func##_impl(path, stat, request_mask, query_flags));	\
+}
+#else
+#error
+#endif
+
+/*
+ * 4.9 API change
+ * Preferred interface to get the current FS time.
+ */
+#if !defined(HAVE_CURRENT_TIME)
+static inline struct timespec
+current_time(struct inode *ip)
+{
+	return (timespec_trunc(current_kernel_time(), ip->i_sb->s_time_gran));
+}
+#endif
+
 #endif /* _ZFS_VFS_H */

include/linux/xattr_compat.h

@@ -101,13 +101,26 @@ fn(struct inode *ip, char *list, size_t list_size,			\
 }
 #endif
 
+/*
+ * 4.7 API change,
+ * The xattr_handler->get() callback was changed to take a both dentry and
+ * inode, because the dentry might not be attached to an inode yet.
+ */
+#if defined(HAVE_XATTR_GET_DENTRY_INODE)
+#define	ZPL_XATTR_GET_WRAPPER(fn)					\
+static int								\
+fn(const struct xattr_handler *handler, struct dentry *dentry,		\
+    struct inode *inode, const char *name, void *buffer, size_t size)	\
+{									\
+	return (__ ## fn(inode, name, buffer, size));			\
+}
 /*
  * 4.4 API change,
  * The xattr_handler->get() callback was changed to take a xattr_handler,
  * and handler_flags argument was removed and should be accessed by
  * handler->flags.
  */
-#if defined(HAVE_XATTR_GET_HANDLER)
+#elif defined(HAVE_XATTR_GET_HANDLER)
 #define	ZPL_XATTR_GET_WRAPPER(fn)					\
 static int								\
 fn(const struct xattr_handler *handler, struct dentry *dentry,		\
@@ -140,13 +153,27 @@ fn(struct inode *ip, const char *name, void *buffer, size_t size)	\
 }
 #endif
 
+/*
+ * 4.7 API change,
+ * The xattr_handler->set() callback was changed to take a both dentry and
+ * inode, because the dentry might not be attached to an inode yet.
+ */
+#if defined(HAVE_XATTR_SET_DENTRY_INODE)
+#define	ZPL_XATTR_SET_WRAPPER(fn)					\
+static int								\
+fn(const struct xattr_handler *handler, struct dentry *dentry,		\
+    struct inode *inode, const char *name, const void *buffer,		\
+    size_t size, int flags)						\
+{									\
+	return (__ ## fn(inode, name, buffer, size, flags));		\
+}
 /*
  * 4.4 API change,
  * The xattr_handler->set() callback was changed to take a xattr_handler,
  * and handler_flags argument was removed and should be accessed by
  * handler->flags.
  */
-#if defined(HAVE_XATTR_SET_HANDLER)
+#elif defined(HAVE_XATTR_SET_HANDLER)
 #define	ZPL_XATTR_SET_WRAPPER(fn)					\
 static int								\
 fn(const struct xattr_handler *handler, struct dentry *dentry,		\
@@ -190,20 +217,20 @@ fn(struct inode *ip, const char *name, const void *buffer,		\
 
 /*
  * Linux 3.7 API change. posix_acl_{from,to}_xattr gained the user_ns
- * parameter.  For the HAVE_POSIX_ACL_FROM_XATTR_USERNS version the
- * userns _may_ not be correct because it's used outside the RCU.
+ * parameter.  All callers are expected to pass the &init_user_ns which
+ * is available through the init credential (kcred).
  */
 #ifdef HAVE_POSIX_ACL_FROM_XATTR_USERNS
 static inline struct posix_acl *
 zpl_acl_from_xattr(const void *value, int size)
 {
-	return (posix_acl_from_xattr(CRED()->user_ns, value, size));
+	return (posix_acl_from_xattr(kcred->user_ns, value, size));
 }
 
 static inline int
 zpl_acl_to_xattr(struct posix_acl *acl, void *value, int size)
 {
-	return (posix_acl_to_xattr(CRED()->user_ns, acl, value, size));
+	return (posix_acl_to_xattr(kcred->user_ns, acl, value, size));
 }
 
 #else

include/sys/dsl_dataset.h

@@ -221,6 +221,7 @@ int dsl_dataset_own_obj(struct dsl_pool *dp, uint64_t dsobj,
 void dsl_dataset_disown(dsl_dataset_t *ds, void *tag);
 void dsl_dataset_name(dsl_dataset_t *ds, char *name);
 boolean_t dsl_dataset_tryown(dsl_dataset_t *ds, void *tag);
+int dsl_dataset_namelen(dsl_dataset_t *ds);
 uint64_t dsl_dataset_create_sync(dsl_dir_t *pds, const char *lastname,
     dsl_dataset_t *origin, uint64_t flags, cred_t *, dmu_tx_t *);
 uint64_t dsl_dataset_create_sync_dd(dsl_dir_t *dd, dsl_dataset_t *origin,

include/sys/fs/zfs.h

@@ -69,6 +69,7 @@ typedef enum dmu_objset_type {
 #define	ZAP_MAXNAMELEN 256
 #define	ZAP_MAXVALUELEN (1024 * 8)
 #define	ZAP_OLDMAXVALUELEN 1024
+#define	ZFS_MAX_DATASET_NAME_LEN 256
 
 /*
  * Dataset properties are identified by these constants and must be added to

include/sys/mntent.h

@@ -68,8 +68,9 @@
 #define	MNTOPT_NOFAIL	"nofail"	/* no failure */
 #define	MNTOPT_RELATIME	"relatime"	/* allow relative time updates */
 #define	MNTOPT_NORELATIME "norelatime"	/* do not allow relative time updates */
-#define	MNTOPT_DFRATIME	"strictatime"	/* Deferred access time updates */
-#define	MNTOPT_NODFRATIME "nostrictatime" /* No Deferred access time updates */
+#define	MNTOPT_STRICTATIME "strictatime" /* strict access time updates */
+#define	MNTOPT_NOSTRICTATIME "nostrictatime" /* No strict access time updates */
+#define	MNTOPT_LAZYTIME "lazytime"	/* Defer access time writing */
 #define	MNTOPT_SETUID	"suid"		/* Both setuid and devices allowed */
 #define	MNTOPT_NOSETUID	"nosuid"	/* Neither setuid nor devices allowed */
 #define	MNTOPT_OWNER	"owner"		/* allow owner mount */

include/sys/refcount.h

@@ -40,6 +40,17 @@ extern "C" {
  */
 #define	FTAG ((char *)__func__)
 
+/*
+ * Starting with 4.11, torvalds/linux@f405df5, the linux kernel defines a
+ * refcount_t type of its own.  The macro below effectively changes references
+ * in the ZFS code from refcount_t to zfs_refcount_t at compile time, so that
+ * existing code need not be altered, reducing conflicts when landing openZFS
+ * patches.
+ */
+
+#define	refcount_t	zfs_refcount_t
+#define	refcount_add	zfs_refcount_add
+
 #ifdef	ZFS_DEBUG
 typedef struct reference {
 	list_node_t ref_link;
@@ -55,7 +66,7 @@ typedef struct refcount {
 	list_t rc_removed;
 	int64_t rc_count;
 	int64_t rc_removed_count;
-} refcount_t;
+} zfs_refcount_t;
 
 /* Note: refcount_t must be initialized with refcount_create[_untracked]() */
 
@@ -65,7 +76,7 @@ void refcount_destroy(refcount_t *rc);
 void refcount_destroy_many(refcount_t *rc, uint64_t number);
 int refcount_is_zero(refcount_t *rc);
 int64_t refcount_count(refcount_t *rc);
-int64_t refcount_add(refcount_t *rc, void *holder_tag);
+int64_t zfs_refcount_add(refcount_t *rc, void *holder_tag);
 int64_t refcount_remove(refcount_t *rc, void *holder_tag);
 int64_t refcount_add_many(refcount_t *rc, uint64_t number, void *holder_tag);
 int64_t refcount_remove_many(refcount_t *rc, uint64_t number, void *holder_tag);
@@ -86,7 +97,7 @@ typedef struct refcount {
 #define	refcount_destroy_many(rc, number) ((rc)->rc_count = 0)
 #define	refcount_is_zero(rc) ((rc)->rc_count == 0)
 #define	refcount_count(rc) ((rc)->rc_count)
-#define	refcount_add(rc, holder) atomic_add_64_nv(&(rc)->rc_count, 1)
+#define	zfs_refcount_add(rc, holder) atomic_add_64_nv(&(rc)->rc_count, 1)
 #define	refcount_remove(rc, holder) atomic_add_64_nv(&(rc)->rc_count, -1)
 #define	refcount_add_many(rc, number, holder) \
 	atomic_add_64_nv(&(rc)->rc_count, number)

include/sys/spa.h

@@ -446,6 +446,10 @@ _NOTE(CONSTCOND) } while (0)
 	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
 	((zc1).zc_word[3] - (zc2).zc_word[3])))
 
+#define	ZIO_CHECKSUM_IS_ZERO(zc) \
+	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
+	(zc)->zc_word[2] | (zc)->zc_word[3]))
+
 #define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
 
 #define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\

include/sys/trace_acl.h

@@ -56,7 +56,6 @@ DECLARE_EVENT_CLASS(zfs_ace_class,
 	    __field(uint64_t,		z_mapcnt)
 	    __field(uint64_t,		z_gen)
 	    __field(uint64_t,		z_size)
-	    __array(uint64_t,		z_atime, 2)
 	    __field(uint64_t,		z_links)
 	    __field(uint64_t,		z_pflags)
 	    __field(uint64_t,		z_uid)
@@ -64,7 +63,6 @@ DECLARE_EVENT_CLASS(zfs_ace_class,
 	    __field(uint32_t,		z_sync_cnt)
 	    __field(mode_t,		z_mode)
 	    __field(boolean_t,		z_is_sa)
-	    __field(boolean_t,		z_is_zvol)
 	    __field(boolean_t,		z_is_mapped)
 	    __field(boolean_t,		z_is_ctldir)
 	    __field(boolean_t,		z_is_stale)
@@ -95,8 +93,6 @@ DECLARE_EVENT_CLASS(zfs_ace_class,
 	    __entry->z_mapcnt		= zn->z_mapcnt;
 	    __entry->z_gen		= zn->z_gen;
 	    __entry->z_size		= zn->z_size;
-	    __entry->z_atime[0]		= zn->z_atime[0];
-	    __entry->z_atime[1]		= zn->z_atime[1];
 	    __entry->z_links		= zn->z_links;
 	    __entry->z_pflags		= zn->z_pflags;
 	    __entry->z_uid		= zn->z_uid;
@@ -104,7 +100,6 @@ DECLARE_EVENT_CLASS(zfs_ace_class,
 	    __entry->z_sync_cnt		= zn->z_sync_cnt;
 	    __entry->z_mode		= zn->z_mode;
 	    __entry->z_is_sa		= zn->z_is_sa;
-	    __entry->z_is_zvol		= zn->z_is_zvol;
 	    __entry->z_is_mapped	= zn->z_is_mapped;
 	    __entry->z_is_ctldir	= zn->z_is_ctldir;
 	    __entry->z_is_stale		= zn->z_is_stale;
@@ -126,9 +121,9 @@ DECLARE_EVENT_CLASS(zfs_ace_class,
 	),
 	TP_printk("zn { id %llu unlinked %u atime_dirty %u "
 	    "zn_prefetch %u moved %u blksz %u seq %u "
-	    "mapcnt %llu gen %llu size %llu atime 0x%llx:0x%llx "
+	    "mapcnt %llu gen %llu size %llu "
 	    "links %llu pflags %llu uid %llu gid %llu "
-	    "sync_cnt %u mode 0x%x is_sa %d is_zvol %d "
+	    "sync_cnt %u mode 0x%x is_sa %d "
 	    "is_mapped %d is_ctldir %d is_stale %d inode { "
 	    "ino %lu nlink %u version %llu size %lli blkbits %u "
 	    "bytes %u mode 0x%x generation %x } } ace { type %u "
@@ -136,10 +131,10 @@ DECLARE_EVENT_CLASS(zfs_ace_class,
 	    __entry->z_id, __entry->z_unlinked, __entry->z_atime_dirty,
 	    __entry->z_zn_prefetch, __entry->z_moved, __entry->z_blksz,
 	    __entry->z_seq, __entry->z_mapcnt, __entry->z_gen,
-	    __entry->z_size, __entry->z_atime[0], __entry->z_atime[1],
+	    __entry->z_size,
 	    __entry->z_links, __entry->z_pflags, __entry->z_uid,
 	    __entry->z_gid, __entry->z_sync_cnt, __entry->z_mode,
-	    __entry->z_is_sa, __entry->z_is_zvol, __entry->z_is_mapped,
+	    __entry->z_is_sa, __entry->z_is_mapped,
 	    __entry->z_is_ctldir, __entry->z_is_stale, __entry->i_ino,
 	    __entry->i_nlink, __entry->i_version, __entry->i_size,
 	    __entry->i_blkbits, __entry->i_bytes, __entry->i_mode,

include/sys/vdev_disk.h

@@ -37,9 +37,5 @@ typedef struct vdev_disk {
 	struct block_device	*vd_bdev;
 } vdev_disk_t;
 
-extern int vdev_disk_physio(struct block_device *, caddr_t,
-			    size_t, uint64_t, int);
-extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **);
-
 #endif /* _KERNEL */
 #endif /* _SYS_VDEV_DISK_H */

include/sys/xvattr.h

@@ -225,7 +225,7 @@ typedef struct xvattr {
  * of requested attributes (xva_reqattrmap[]).
  */
 #define	XVA_SET_REQ(xvap, attr)					\
-	ASSERT((xvap)->xva_vattr.va_mask | AT_XVATTR);		\
+	ASSERT((xvap)->xva_vattr.va_mask & AT_XVATTR);		\
 	ASSERT((xvap)->xva_magic == XVA_MAGIC);			\
 	(xvap)->xva_reqattrmap[XVA_INDEX(attr)] |= XVA_ATTRBIT(attr)
 /*
@@ -233,7 +233,7 @@ typedef struct xvattr {
  * of requested attributes (xva_reqattrmap[]).
  */
 #define	XVA_CLR_REQ(xvap, attr)					\
-	ASSERT((xvap)->xva_vattr.va_mask | AT_XVATTR);		\
+	ASSERT((xvap)->xva_vattr.va_mask & AT_XVATTR);		\
 	ASSERT((xvap)->xva_magic == XVA_MAGIC);			\
 	(xvap)->xva_reqattrmap[XVA_INDEX(attr)] &= ~XVA_ATTRBIT(attr)
 
@@ -242,7 +242,7 @@ typedef struct xvattr {
  * of returned attributes (xva_rtnattrmap[]).
  */
 #define	XVA_SET_RTN(xvap, attr)					\
-	ASSERT((xvap)->xva_vattr.va_mask | AT_XVATTR);		\
+	ASSERT((xvap)->xva_vattr.va_mask & AT_XVATTR);		\
 	ASSERT((xvap)->xva_magic == XVA_MAGIC);			\
 	(XVA_RTNATTRMAP(xvap))[XVA_INDEX(attr)] |= XVA_ATTRBIT(attr)
 
@@ -251,7 +251,7 @@ typedef struct xvattr {
  * to see of the corresponding attribute bit is set.  If so, returns non-zero.
  */
 #define	XVA_ISSET_REQ(xvap, attr)					\
-	((((xvap)->xva_vattr.va_mask | AT_XVATTR) &&			\
+	((((xvap)->xva_vattr.va_mask & AT_XVATTR) &&			\
 		((xvap)->xva_magic == XVA_MAGIC) &&			\
 		((xvap)->xva_mapsize > XVA_INDEX(attr))) ?		\
 	((xvap)->xva_reqattrmap[XVA_INDEX(attr)] & XVA_ATTRBIT(attr)) :	0)
@@ -261,7 +261,7 @@ typedef struct xvattr {
  * to see of the corresponding attribute bit is set.  If so, returns non-zero.
  */
 #define	XVA_ISSET_RTN(xvap, attr)					\
-	((((xvap)->xva_vattr.va_mask | AT_XVATTR) &&			\
+	((((xvap)->xva_vattr.va_mask & AT_XVATTR) &&			\
 		((xvap)->xva_magic == XVA_MAGIC) &&			\
 		((xvap)->xva_mapsize > XVA_INDEX(attr))) ?		\
 	((XVA_RTNATTRMAP(xvap))[XVA_INDEX(attr)] & XVA_ATTRBIT(attr)) : 0)

include/sys/zfs_rlock.h

@@ -32,7 +32,9 @@ extern "C" {
 
 #ifdef _KERNEL
 
-#include <sys/zfs_znode.h>
+#include <sys/list.h>
+#include <sys/avl.h>
+#include <sys/condvar.h>
 
 typedef enum {
 	RL_READER,
@@ -40,8 +42,16 @@ typedef enum {
 	RL_APPEND
 } rl_type_t;
 
+typedef struct zfs_rlock {
+	kmutex_t zr_mutex;	/* protects changes to zr_avl */
+	avl_tree_t zr_avl;	/* avl tree of range locks */
+	uint64_t *zr_size;	/* points to znode->z_size */
+	uint_t *zr_blksz;	/* points to znode->z_blksz */
+	uint64_t *zr_max_blksz; /* points to zsb->z_max_blksz */
+} zfs_rlock_t;
+
 typedef struct rl {
-	znode_t *r_zp;		/* znode this lock applies to */
+	zfs_rlock_t *r_zrl;
 	avl_node_t r_node;	/* avl node link */
 	uint64_t r_off;		/* file range offset */
 	uint64_t r_len;		/* file range length */
@@ -61,7 +71,8 @@ typedef struct rl {
  * is converted to RL_WRITER that specified to lock from the start of the
  * end of file.  Returns the range lock structure.
  */
-rl_t *zfs_range_lock(znode_t *zp, uint64_t off, uint64_t len, rl_type_t type);
+rl_t *zfs_range_lock(zfs_rlock_t *zrl, uint64_t off, uint64_t len,
+    rl_type_t type);
 
 /* Unlock range and destroy range lock structure. */
 void zfs_range_unlock(rl_t *rl);
@@ -78,6 +89,23 @@ void zfs_range_reduce(rl_t *rl, uint64_t off, uint64_t len);
  */
 int zfs_range_compare(const void *arg1, const void *arg2);
 
+static inline void
+zfs_rlock_init(zfs_rlock_t *zrl)
+{
+	mutex_init(&zrl->zr_mutex, NULL, MUTEX_DEFAULT, NULL);
+	avl_create(&zrl->zr_avl, zfs_range_compare,
+	    sizeof (rl_t), offsetof(rl_t, r_node));
+	zrl->zr_size = NULL;
+	zrl->zr_blksz = NULL;
+	zrl->zr_max_blksz = NULL;
+}
+
+static inline void
+zfs_rlock_destroy(zfs_rlock_t *zrl)
+{
+	avl_destroy(&zrl->zr_avl);
+	mutex_destroy(&zrl->zr_mutex);
+}
 #endif /* _KERNEL */
 
 #ifdef	__cplusplus

include/sys/zfs_vfsops.h

@@ -64,7 +64,6 @@ typedef struct zfs_mntopts {
 
 typedef struct zfs_sb {
 	struct super_block *z_sb;	/* generic super_block */
-	struct backing_dev_info z_bdi;	/* generic backing dev info */
 	struct zfs_sb	*z_parent;	/* parent fs */
 	objset_t	*z_os;		/* objset reference */
 	zfs_mntopts_t	*z_mntopts;	/* passed mount options */

include/sys/zfs_znode.h

@@ -37,6 +37,7 @@
 #include <sys/rrwlock.h>
 #include <sys/zfs_sa.h>
 #include <sys/zfs_stat.h>
+#include <sys/zfs_rlock.h>
 #endif
 #include <sys/zfs_acl.h>
 #include <sys/zil.h>
@@ -187,8 +188,7 @@ typedef struct znode {
 	krwlock_t	z_parent_lock;	/* parent lock for directories */
 	krwlock_t	z_name_lock;	/* "master" lock for dirent locks */
 	zfs_dirlock_t	*z_dirlocks;	/* directory entry lock list */
-	kmutex_t	z_range_lock;	/* protects changes to z_range_avl */
-	avl_tree_t	z_range_avl;	/* avl tree of file range locks */
+	zfs_rlock_t	z_range_lock;	/* file range lock */
 	uint8_t		z_unlinked;	/* file has been unlinked */
 	uint8_t		z_atime_dirty;	/* atime needs to be synced */
 	uint8_t		z_zn_prefetch;	/* Prefetch znodes? */
@@ -198,7 +198,6 @@ typedef struct znode {
 	uint64_t	z_mapcnt;	/* number of pages mapped to file */
 	uint64_t	z_gen;		/* generation (cached) */
 	uint64_t	z_size;		/* file size (cached) */
-	uint64_t	z_atime[2];	/* atime (cached) */
 	uint64_t	z_links;	/* file links (cached) */
 	uint64_t	z_pflags;	/* pflags (cached) */
 	uint64_t	z_uid;		/* uid fuid (cached) */
@@ -209,11 +208,9 @@ typedef struct znode {
 	zfs_acl_t	*z_acl_cached;	/* cached acl */
 	krwlock_t	z_xattr_lock;	/* xattr data lock */
 	nvlist_t	*z_xattr_cached; /* cached xattrs */
-	struct znode	*z_xattr_parent; /* xattr parent znode */
 	list_node_t	z_link_node;	/* all znodes in fs link */
 	sa_handle_t	*z_sa_hdl;	/* handle to sa data */
 	boolean_t	z_is_sa;	/* are we native sa? */
-	boolean_t	z_is_zvol;	/* are we used by the zvol */
 	boolean_t	z_is_mapped;	/* are we mmap'ed */
 	boolean_t	z_is_ctldir;	/* are we .zfs entry */
 	boolean_t	z_is_stale;	/* are we stale due to rollback? */
@@ -306,16 +303,12 @@ extern unsigned int zfs_object_mutex_size;
 #define	STATE_CHANGED		(ATTR_CTIME)
 #define	CONTENT_MODIFIED	(ATTR_MTIME | ATTR_CTIME)
 
-#define	ZFS_ACCESSTIME_STAMP(zsb, zp) \
-	if ((zsb)->z_atime && !(zfs_is_readonly(zsb))) \
-		zfs_tstamp_update_setup(zp, ACCESSED, NULL, NULL, B_FALSE);
-
 extern int	zfs_init_fs(zfs_sb_t *, znode_t **);
 extern void	zfs_set_dataprop(objset_t *);
 extern void	zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *,
     dmu_tx_t *tx);
 extern void	zfs_tstamp_update_setup(znode_t *, uint_t, uint64_t [2],
-    uint64_t [2], boolean_t);
+    uint64_t [2]);
 extern void	zfs_grow_blocksize(znode_t *, uint64_t, dmu_tx_t *);
 extern int	zfs_freesp(znode_t *, uint64_t, uint64_t, int, boolean_t);
 extern void	zfs_znode_init(void);

include/sys/zpl.h

@@ -76,7 +76,7 @@ extern ssize_t zpl_xattr_list(struct dentry *dentry, char *buf, size_t size);
 extern int zpl_xattr_security_init(struct inode *ip, struct inode *dip,
     const struct qstr *qstr);
 #if defined(CONFIG_FS_POSIX_ACL)
-extern int zpl_set_acl(struct inode *ip, int type, struct posix_acl *acl);
+extern int zpl_set_acl(struct inode *ip, struct posix_acl *acl, int type);
 extern struct posix_acl *zpl_get_acl(struct inode *ip, int type);
 #if !defined(HAVE_GET_ACL)
 #if defined(HAVE_CHECK_ACL_WITH_FLAGS)
@@ -123,7 +123,7 @@ extern const struct inode_operations zpl_ops_snapdirs;
 extern const struct file_operations zpl_fops_shares;
 extern const struct inode_operations zpl_ops_shares;
 
-#ifdef HAVE_VFS_ITERATE
+#if defined(HAVE_VFS_ITERATE) || defined(HAVE_VFS_ITERATE_SHARED)
 
 #define	DIR_CONTEXT_INIT(_dirent, _actor, _pos) {	\
 	.actor = _actor,				\

lib/libspl/asm-i386/atomic.S

@@ -507,7 +507,7 @@
 	movl	16(%esp), %ebx
 	movl	20(%esp), %ecx
 	subl	%eax, %ebx
-	adcl	%edx, %ecx
+	sbbl	%edx, %ecx
 	lock
 	cmpxchg8b (%edi)
 	jne	1b

lib/libspl/getmntany.c

@@ -34,6 +34,7 @@
 #include <sys/mnttab.h>
 
 #include <sys/types.h>
+#include <sys/sysmacros.h>
 #include <sys/stat.h>
 #include <unistd.h>
 

lib/libspl/include/assert.h

@@ -31,69 +31,66 @@
 
 #include <stdio.h>
 #include <stdlib.h>
-
-#ifndef __assert_c99
-static inline void
-__assert_c99(const char *expr, const char *file, int line, const char *func)
-{
-	fprintf(stderr, "%s:%i: %s: Assertion `%s` failed.\n",
-		file, line, func, expr);
-	abort();
-}
-#endif  /* __assert_c99 */
-
-#ifndef verify
-#if defined(__STDC__)
-#if __STDC_VERSION__ - 0 >= 199901L
-#define	verify(EX) (void)((EX) || \
-	(__assert_c99(#EX, __FILE__, __LINE__, __func__), 0))
-#else
-#define	verify(EX) (void)((EX) || (__assert(#EX, __FILE__, __LINE__), 0))
-#endif  /* __STDC_VERSION__ - 0 >= 199901L */
-#else
-#define	verify(EX) (void)((EX) || (_assert("EX", __FILE__, __LINE__), 0))
-#endif	/* __STDC__ */
-#endif  /* verify */
-
-#undef VERIFY
-#undef ASSERT
-
-#define	VERIFY	verify
-#define	ASSERT	assert
-
-extern void __assert(const char *, const char *, int);
+#include <stdarg.h>
 
 static inline int
-assfail(const char *buf, const char *file, int line)
+libspl_assert(const char *buf, const char *file, const char *func, int line)
 {
-	__assert(buf, file, line);
-	return (0);
+	fprintf(stderr, "%s\n", buf);
+	fprintf(stderr, "ASSERT at %s:%d:%s()", file, line, func);
+	abort();
 }
 
-/* BEGIN CSTYLED */
-#define	VERIFY3_IMPL(LEFT, OP, RIGHT, TYPE) do { \
-	const TYPE __left = (TYPE)(LEFT); \
-	const TYPE __right = (TYPE)(RIGHT); \
-	if (!(__left OP __right)) { \
-		char *__buf = alloca(256); \
-		(void) snprintf(__buf, 256, "%s %s %s (0x%llx %s 0x%llx)", \
-			#LEFT, #OP, #RIGHT, \
-			(u_longlong_t)__left, #OP, (u_longlong_t)__right); \
-		assfail(__buf, __FILE__, __LINE__); \
-	} \
+/* printf version of libspl_assert */
+static inline void
+libspl_assertf(const char *file, const char *func, int line, char *format, ...)
+{
+	va_list args;
+
+	va_start(args, format);
+	vfprintf(stderr, format, args);
+	fprintf(stderr, "\n");
+	fprintf(stderr, "ASSERT at %s:%d:%s()", file, line, func);
+	va_end(args);
+	abort();
+}
+
+#ifdef verify
+#undef verify
+#endif
+
+#define	VERIFY(cond)							\
+	(void) ((!(cond)) &&						\
+	    libspl_assert(#cond, __FILE__, __FUNCTION__, __LINE__))
+#define	verify(cond)							\
+	(void) ((!(cond)) &&						\
+	    libspl_assert(#cond, __FILE__, __FUNCTION__, __LINE__))
+
+#define	VERIFY3_IMPL(LEFT, OP, RIGHT, TYPE)				\
+do {									\
+	const TYPE __left = (TYPE)(LEFT);				\
+	const TYPE __right = (TYPE)(RIGHT);				\
+	if (!(__left OP __right))					\
+		libspl_assertf(__FILE__, __FUNCTION__, __LINE__,	\
+		    "%s %s %s (0x%llx %s 0x%llx)", #LEFT, #OP, #RIGHT);	\
 } while (0)
-/* END CSTYLED */
 
 #define	VERIFY3S(x, y, z)	VERIFY3_IMPL(x, y, z, int64_t)
 #define	VERIFY3U(x, y, z)	VERIFY3_IMPL(x, y, z, uint64_t)
 #define	VERIFY3P(x, y, z)	VERIFY3_IMPL(x, y, z, uintptr_t)
 #define	VERIFY0(x)		VERIFY3_IMPL(x, ==, 0, uint64_t)
 
+#ifdef assert
+#undef assert
+#endif
+
 #ifdef NDEBUG
 #define	ASSERT3S(x, y, z)	((void)0)
 #define	ASSERT3U(x, y, z)	((void)0)
 #define	ASSERT3P(x, y, z)	((void)0)
 #define	ASSERT0(x)		((void)0)
+#define	ASSERT(x)		((void)0)
+#define	assert(x)		((void)0)
 #define	ASSERTV(x)
 #define	IMPLY(A, B)		((void)0)
 #define	EQUIV(A, B)		((void)0)
@@ -102,13 +99,17 @@ assfail(const char *buf, const char *file, int line)
 #define	ASSERT3U(x, y, z)	VERIFY3U(x, y, z)
 #define	ASSERT3P(x, y, z)	VERIFY3P(x, y, z)
 #define	ASSERT0(x)		VERIFY0(x)
+#define	ASSERT(x)		VERIFY(x)
+#define	assert(x)		VERIFY(x)
 #define	ASSERTV(x)		x
 #define	IMPLY(A, B) \
 	((void)(((!(A)) || (B)) || \
-	    assfail("(" #A ") implies (" #B ")", __FILE__, __LINE__)))
+	    libspl_assert("(" #A ") implies (" #B ")", \
+	    __FILE__, __FUNCTION__, __LINE__)))
 #define	EQUIV(A, B) \
 	((void)((!!(A) == !!(B)) || \
-	    assfail("(" #A ") is equivalent to (" #B ")", __FILE__, __LINE__)))
+	    libspl_assert("(" #A ") is equivalent to (" #B ")", \
+	    __FILE__, __FUNCTION__, __LINE__)))
 
 #endif  /* NDEBUG */
 

lib/libspl/include/sys/sysmacros.h

@@ -42,7 +42,6 @@
 
 #define	makedevice(maj, min)	makedev(maj, min)
 #define	_sysconf(a)		sysconf(a)
-#define	__NORETURN		__attribute__((noreturn))
 
 /*
  * Compatibility macros/typedefs needed for Solaris -> Linux port

lib/libspl/include/sys/types.h

@@ -27,10 +27,15 @@
 #ifndef _LIBSPL_SYS_TYPES_H
 #define	_LIBSPL_SYS_TYPES_H
 
+#if defined(HAVE_MAKEDEV_IN_SYSMACROS)
+#include <sys/sysmacros.h>
+#elif defined(HAVE_MAKEDEV_IN_MKDEV)
+#include <sys/mkdev.h>
+#endif
+
 #include <sys/isa_defs.h>
 #include <sys/feature_tests.h>
 #include_next <sys/types.h>
-#include <sys/param.h> /* for NBBY */
 #include <sys/types32.h>
 #include <sys/va_list.h>
 
@@ -95,4 +100,6 @@ typedef union {
 } lloff_t;
 #endif
 
+#include <sys/param.h> /* for NBBY */
+
 #endif

lib/libzfs/libzfs_dataset.c

@@ -3315,8 +3315,9 @@ zfs_check_snap_cb(zfs_handle_t *zhp, void *arg)
 	char name[ZFS_MAXNAMELEN];
 	int rv = 0;
 
-	(void) snprintf(name, sizeof (name),
-	    "%s@%s", zhp->zfs_name, dd->snapname);
+	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
+	    dd->snapname) >= sizeof (name))
+		return (EINVAL);
 
 	if (lzc_exists(name))
 		verify(nvlist_add_boolean(dd->nvl, name) == 0);
@@ -3534,8 +3535,9 @@ zfs_snapshot_cb(zfs_handle_t *zhp, void *arg)
 	int rv = 0;
 
 	if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) == 0) {
-		(void) snprintf(name, sizeof (name),
-		    "%s@%s", zfs_get_name(zhp), sd->sd_snapname);
+		if (snprintf(name, sizeof (name), "%s@%s", zfs_get_name(zhp),
+		    sd->sd_snapname) >= sizeof (name))
+			return (EINVAL);
 
 		fnvlist_add_boolean(sd->sd_nvl, name);
 
@@ -3889,7 +3891,6 @@ zfs_rename(zfs_handle_t *zhp, const char *target, boolean_t recursive,
 	}
 
 	if (recursive) {
-
 		parentname = zfs_strdup(zhp->zfs_hdl, zhp->zfs_name);
 		if (parentname == NULL) {
 			ret = -1;
@@ -3902,8 +3903,7 @@ zfs_rename(zfs_handle_t *zhp, const char *target, boolean_t recursive,
 			ret = -1;
 			goto error;
 		}
-
-	} else {
+	} else if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT) {
 		if ((cl = changelist_gather(zhp, ZFS_PROP_NAME, 0,
 		    force_unmount ? MS_FORCE : 0)) == NULL)
 			return (-1);
@@ -3952,23 +3952,23 @@ zfs_rename(zfs_handle_t *zhp, const char *target, boolean_t recursive,
 		 * On failure, we still want to remount any filesystems that
 		 * were previously mounted, so we don't alter the system state.
 		 */
-		if (!recursive)
+		if (cl != NULL)
 			(void) changelist_postfix(cl);
 	} else {
-		if (!recursive) {
+		if (cl != NULL) {
 			changelist_rename(cl, zfs_get_name(zhp), target);
 			ret = changelist_postfix(cl);
 		}
 	}
 
 error:
-	if (parentname) {
+	if (parentname != NULL) {
 		free(parentname);
 	}
-	if (zhrp) {
+	if (zhrp != NULL) {
 		zfs_close(zhrp);
 	}
-	if (cl) {
+	if (cl != NULL) {
 		changelist_free(cl);
 	}
 	return (ret);
@@ -4259,8 +4259,9 @@ zfs_hold_one(zfs_handle_t *zhp, void *arg)
 	char name[ZFS_MAXNAMELEN];
 	int rv = 0;
 
-	(void) snprintf(name, sizeof (name),
-	    "%s@%s", zhp->zfs_name, ha->snapname);
+	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
+	    ha->snapname) >= sizeof (name))
+		return (EINVAL);
 
 	if (lzc_exists(name))
 		fnvlist_add_string(ha->nvl, name, ha->tag);
@@ -4379,8 +4380,11 @@ zfs_release_one(zfs_handle_t *zhp, void *arg)
 	int rv = 0;
 	nvlist_t *existing_holds;
 
-	(void) snprintf(name, sizeof (name),
-	    "%s@%s", zhp->zfs_name, ha->snapname);
+	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
+	    ha->snapname) >= sizeof (name)) {
+		ha->error = EINVAL;
+		rv = EINVAL;
+	}
 
 	if (lzc_get_holds(name, &existing_holds) != 0) {
 		ha->error = ENOENT;

lib/libzfs/libzfs_import.c

@@ -97,6 +97,8 @@ typedef struct pool_list {
 	name_entry_t		*names;
 } pool_list_t;
 
+#define	DEV_BYID_PATH	"/dev/disk/by-id/"
+
 static char *
 get_devid(const char *path)
 {
@@ -121,6 +123,40 @@ get_devid(const char *path)
 	return (ret);
 }
 
+/*
+ * Wait up to timeout_ms for udev to set up the device node.  The device is
+ * considered ready when the provided path have been verified to exist and
+ * it has been allowed to settle.  At this point the device the device can
+ * be accessed reliably.  Depending on the complexity of the udev rules thisi
+ * process could take several seconds.
+ */
+int
+zpool_label_disk_wait(char *path, int timeout_ms)
+{
+	int settle_ms = 50;
+	long sleep_ms = 10;
+	hrtime_t start, settle;
+	struct stat64 statbuf;
+
+	start = gethrtime();
+	settle = 0;
+
+	do {
+		errno = 0;
+		if ((stat64(path, &statbuf) == 0) && (errno == 0)) {
+			if (settle == 0)
+				settle = gethrtime();
+			else if (NSEC2MSEC(gethrtime() - settle) >= settle_ms)
+				return (0);
+		} else if (errno != ENOENT) {
+			return (errno);
+		}
+
+		usleep(sleep_ms * MILLISEC);
+	} while (NSEC2MSEC(gethrtime() - start) < timeout_ms);
+
+	return (ENODEV);
+}
 
 /*
  * Go through and fix up any path and/or devid information for the given vdev
@@ -162,7 +198,6 @@ fix_paths(nvlist_t *nv, name_entry_t *names)
 	best = NULL;
 	for (ne = names; ne != NULL; ne = ne->ne_next) {
 		if (ne->ne_guid == guid) {
-
 			if (path == NULL) {
 				best = ne;
 				break;
@@ -186,7 +221,7 @@ fix_paths(nvlist_t *nv, name_entry_t *names)
 			}
 
 			/* Prefer paths earlier in the search order. */
-			if (best->ne_num_labels == best->ne_num_labels &&
+			if (ne->ne_num_labels == best->ne_num_labels &&
 			    ne->ne_order < best->ne_order) {
 				best = ne;
 				continue;
@@ -352,6 +387,118 @@ add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
 	return (0);
 }
 
+#ifdef HAVE_LIBBLKID
+static int
+add_path(libzfs_handle_t *hdl, pool_list_t *pools, uint64_t pool_guid,
+    uint64_t vdev_guid, const char *path, int order)
+{
+	nvlist_t *label;
+	uint64_t guid;
+	int error, fd, num_labels;
+
+	fd = open64(path, O_RDONLY);
+	if (fd < 0)
+		return (errno);
+
+	error = zpool_read_label(fd, &label, &num_labels);
+	close(fd);
+
+	if (error || label == NULL)
+		return (ENOENT);
+
+	error = nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &guid);
+	if (error || guid != pool_guid) {
+		nvlist_free(label);
+		return (EINVAL);
+	}
+
+	error = nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid);
+	if (error || guid != vdev_guid) {
+		nvlist_free(label);
+		return (EINVAL);
+	}
+
+	error = add_config(hdl, pools, path, order, num_labels, label);
+
+	return (error);
+}
+
+static int
+add_configs_from_label_impl(libzfs_handle_t *hdl, pool_list_t *pools,
+    nvlist_t *nvroot, uint64_t pool_guid, uint64_t vdev_guid)
+{
+	char udevpath[MAXPATHLEN];
+	char *path;
+	nvlist_t **child;
+	uint_t c, children;
+	uint64_t guid;
+	int error;
+
+	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
+	    &child, &children) == 0) {
+		for (c = 0; c < children; c++) {
+			error  = add_configs_from_label_impl(hdl, pools,
+			    child[c], pool_guid, vdev_guid);
+			if (error)
+				return (error);
+		}
+		return (0);
+	}
+
+	if (nvroot == NULL)
+		return (0);
+
+	error = nvlist_lookup_uint64(nvroot, ZPOOL_CONFIG_GUID, &guid);
+	if ((error != 0) || (guid != vdev_guid))
+		return (0);
+
+	error = nvlist_lookup_string(nvroot, ZPOOL_CONFIG_PATH, &path);
+	if (error == 0)
+		(void) add_path(hdl, pools, pool_guid, vdev_guid, path, 0);
+
+	error = nvlist_lookup_string(nvroot, ZPOOL_CONFIG_DEVID, &path);
+	if (error == 0) {
+		sprintf(udevpath, "%s%s", DEV_BYID_PATH, path);
+		(void) add_path(hdl, pools, pool_guid, vdev_guid, udevpath, 1);
+	}
+
+	return (0);
+}
+
+/*
+ * Given a disk label call add_config() for all known paths to the device
+ * as described by the label itself.  The paths are added in the following
+ * priority order: 'path', 'devid', 'devnode'.  As these alternate paths are
+ * added the labels are verified to make sure they refer to the same device.
+ */
+static int
+add_configs_from_label(libzfs_handle_t *hdl, pool_list_t *pools,
+    char *devname, int num_labels, nvlist_t *label)
+{
+	nvlist_t *nvroot;
+	uint64_t pool_guid;
+	uint64_t vdev_guid;
+	int error;
+
+	if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_VDEV_TREE, &nvroot) ||
+	    nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &pool_guid) ||
+	    nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &vdev_guid))
+		return (ENOENT);
+
+	/* Allow devlinks to stabilize so all paths are available. */
+	zpool_label_disk_wait(devname, DISK_LABEL_WAIT);
+
+	/* Add alternate paths as described by the label vdev_tree. */
+	(void) add_configs_from_label_impl(hdl, pools, nvroot,
+	    pool_guid, vdev_guid);
+
+	/* Add the device node /dev/sdX path as a last resort. */
+	error = add_config(hdl, pools, devname, 100, num_labels, label);
+
+	return (error);
+}
+#endif /* HAVE_LIBBLKID */
+
 /*
  * Returns true if the named pool matches the given GUID.
  */
@@ -975,9 +1122,7 @@ zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
 	blkid_cache cache;
 	blkid_dev_iterate iter;
 	blkid_dev dev;
-	const char *devname;
-	nvlist_t *config;
-	int fd, err, num_labels;
+	int err;
 
 	err = blkid_get_cache(&cache, NULL);
 	if (err != 0) {
@@ -1008,25 +1153,23 @@ zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
 	}
 
 	while (blkid_dev_next(iter, &dev) == 0) {
-		devname = blkid_dev_devname(dev);
+		nvlist_t *label;
+		char *devname;
+		int fd, num_labels;
+
+		devname = (char *) blkid_dev_devname(dev);
 		if ((fd = open64(devname, O_RDONLY)) < 0)
 			continue;
 
-		err = zpool_read_label(fd, &config, &num_labels);
+		err = zpool_read_label(fd, &label, &num_labels);
 		(void) close(fd);
 
-		if (err != 0) {
-			(void) no_memory(hdl);
-			goto err_blkid3;
-		}
+		if (err || label == NULL)
+			continue;
 
-		if (config != NULL) {
-			err = add_config(hdl, pools, devname, 0,
-			    num_labels, config);
-			if (err != 0)
-				goto err_blkid3;
-		}
+		add_configs_from_label(hdl, pools, devname, num_labels, label);
 	}
+	err = 0;
 
 err_blkid3:
 	blkid_dev_iterate_end(iter);
@@ -1194,16 +1337,33 @@ zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
 
 			if (config != NULL) {
 				boolean_t matched = B_TRUE;
+				boolean_t aux = B_FALSE;
 				char *pname;
 
-				if ((iarg->poolname != NULL) &&
+				/*
+				 * Check if it's a spare or l2cache device. If
+				 * it is, we need to skip the name and guid
+				 * check since they don't exist on aux device
+				 * label.
+				 */
+				if (iarg->poolname != NULL ||
+				    iarg->guid != 0) {
+					uint64_t state;
+					aux = nvlist_lookup_uint64(config,
+					    ZPOOL_CONFIG_POOL_STATE,
+					    &state) == 0 &&
+					    (state == POOL_STATE_SPARE ||
+					    state == POOL_STATE_L2CACHE);
+				}
+
+				if ((iarg->poolname != NULL) && !aux &&
 				    (nvlist_lookup_string(config,
 				    ZPOOL_CONFIG_POOL_NAME, &pname) == 0)) {
 
 					if (strcmp(iarg->poolname, pname))
 						matched = B_FALSE;
 
-				} else if (iarg->guid != 0) {
+				} else if (iarg->guid != 0 && !aux) {
 					uint64_t this_guid;
 
 					matched = nvlist_lookup_uint64(config,

lib/libzfs/libzfs_iter.c

@@ -204,8 +204,11 @@ zfs_iter_bookmarks(zfs_handle_t *zhp, zfs_iter_f func, void *data)
 		bmark_name = nvpair_name(pair);
 		bmark_props = fnvpair_value_nvlist(pair);
 
-		(void) snprintf(name, sizeof (name), "%s#%s", zhp->zfs_name,
-		    bmark_name);
+		if (snprintf(name, sizeof (name), "%s#%s", zhp->zfs_name,
+		    bmark_name) >= sizeof (name)) {
+			err = EINVAL;
+			goto out;
+		}
 
 		nzhp = make_bookmark_handle(zhp, name, bmark_props);
 		if (nzhp == NULL)

lib/libzfs/libzfs_mount.c

@@ -21,6 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2014 by Delphix. All rights reserved.
  */
 
 /*
@@ -363,6 +364,14 @@ zfs_add_options(zfs_handle_t *zhp, char *options, int len)
 
 	error = zfs_add_option(zhp, options, len,
 	    ZFS_PROP_ATIME, MNTOPT_ATIME, MNTOPT_NOATIME);
+	/*
+	 * don't add relatime/strictatime when atime=off, otherwise strictatime
+	 * will force atime=on
+	 */
+	if (strstr(options, MNTOPT_NOATIME) == NULL) {
+		error = zfs_add_option(zhp, options, len,
+		    ZFS_PROP_RELATIME, MNTOPT_RELATIME, MNTOPT_STRICTATIME);
+	}
 	error = error ? error : zfs_add_option(zhp, options, len,
 	    ZFS_PROP_DEVICES, MNTOPT_DEVICES, MNTOPT_NODEVICES);
 	error = error ? error : zfs_add_option(zhp, options, len,
@@ -744,13 +753,6 @@ zfs_share_proto(zfs_handle_t *zhp, zfs_share_proto_t *proto)
 	if (!zfs_is_mountable(zhp, mountpoint, sizeof (mountpoint), NULL))
 		return (0);
 
-	if ((ret = zfs_init_libshare(hdl, SA_INIT_SHARE_API)) != SA_OK) {
-		(void) zfs_error_fmt(hdl, EZFS_SHARENFSFAILED,
-		    dgettext(TEXT_DOMAIN, "cannot share '%s': %s"),
-		    zfs_get_name(zhp), sa_errorstr(ret));
-		return (-1);
-	}
-
 	for (curr_proto = proto; *curr_proto != PROTO_END; curr_proto++) {
 		/*
 		 * Return success if there are no share options.
@@ -761,6 +763,14 @@ zfs_share_proto(zfs_handle_t *zhp, zfs_share_proto_t *proto)
 		    strcmp(shareopts, "off") == 0)
 			continue;
 
+		ret = zfs_init_libshare(hdl, SA_INIT_SHARE_API);
+		if (ret != SA_OK) {
+			(void) zfs_error_fmt(hdl, EZFS_SHARENFSFAILED,
+			    dgettext(TEXT_DOMAIN, "cannot share '%s': %s"),
+			    zfs_get_name(zhp), sa_errorstr(ret));
+			return (-1);
+		}
+
 		/*
 		 * If the 'zoned' property is set, then zfs_is_mountable()
 		 * will have already bailed out if we are in the global zone.
@@ -1072,7 +1082,7 @@ libzfs_dataset_cmp(const void *a, const void *b)
 	if (gotb)
 		return (1);
 
-	return (strcmp(zfs_get_name(a), zfs_get_name(b)));
+	return (strcmp(zfs_get_name(*za), zfs_get_name(*zb)));
 }
 
 /*

lib/libzfs/libzfs_pool.c

@@ -1891,7 +1891,12 @@ zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
 			    "one or more devices are already in use\n"));
 			(void) zfs_error(hdl, EZFS_BADDEV, desc);
 			break;
-
+		case ENAMETOOLONG:
+			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+			    "new name of at least one dataset is longer than "
+			    "the maximum allowable length"));
+			(void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
+			break;
 		default:
 			(void) zpool_standard_error(hdl, error, desc);
 			zpool_explain_recover(hdl,
@@ -4094,29 +4099,6 @@ find_start_block(nvlist_t *config)
 	return (MAXOFFSET_T);
 }
 
-int
-zpool_label_disk_wait(char *path, int timeout)
-{
-	struct stat64 statbuf;
-	int i;
-
-	/*
-	 * Wait timeout miliseconds for a newly created device to be available
-	 * from the given path.  There is a small window when a /dev/ device
-	 * will exist and the udev link will not, so we must wait for the
-	 * symlink.  Depending on the udev rules this may take a few seconds.
-	 */
-	for (i = 0; i < timeout; i++) {
-		usleep(1000);
-
-		errno = 0;
-		if ((stat64(path, &statbuf) == 0) && (errno == 0))
-			return (0);
-	}
-
-	return (ENOENT);
-}
-
 int
 zpool_label_disk_check(char *path)
 {
@@ -4142,6 +4124,32 @@ zpool_label_disk_check(char *path)
 	return (0);
 }
 
+/*
+ * Generate a unique partition name for the ZFS member.  Partitions must
+ * have unique names to ensure udev will be able to create symlinks under
+ * /dev/disk/by-partlabel/ for all pool members.  The partition names are
+ * of the form <pool>-<unique-id>.
+ */
+static void
+zpool_label_name(char *label_name, int label_size)
+{
+	uint64_t id = 0;
+	int fd;
+
+	fd = open("/dev/urandom", O_RDONLY);
+	if (fd > 0) {
+		if (read(fd, &id, sizeof (id)) != sizeof (id))
+			id = 0;
+
+		close(fd);
+	}
+
+	if (id == 0)
+		id = (((uint64_t)rand()) << 32) | (uint64_t)rand();
+
+	snprintf(label_name, label_size, "zfs-%016llx", (u_longlong_t) id);
+}
+
 /*
  * Label an individual disk.  The name provided is the short name,
  * stripped of any leading /dev path.
@@ -4232,7 +4240,7 @@ zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
 	 * can get, in the absence of V_OTHER.
 	 */
 	vtoc->efi_parts[0].p_tag = V_USR;
-	(void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
+	zpool_label_name(vtoc->efi_parts[0].p_name, EFI_PART_NAME_LEN);
 
 	vtoc->efi_parts[8].p_start = slice_size + start_block;
 	vtoc->efi_parts[8].p_size = resv;
@@ -4256,12 +4264,11 @@ zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
 	(void) close(fd);
 	efi_free(vtoc);
 
-	/* Wait for the first expected partition to appear. */
-
 	(void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
 	(void) zfs_append_partition(path, MAXPATHLEN);
 
-	rval = zpool_label_disk_wait(path, 3000);
+	/* Wait to udev to signal use the device has settled. */
+	rval = zpool_label_disk_wait(path, DISK_LABEL_WAIT);
 	if (rval) {
 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to "
 		    "detect device partitions on '%s': %d"), path, rval);

lib/libzfs/libzfs_sendrecv.c

@@ -1487,9 +1487,13 @@ zfs_send(zfs_handle_t *zhp, const char *fromsnap, const char *tosnap,
 			    drr_versioninfo, DMU_COMPOUNDSTREAM);
 			DMU_SET_FEATUREFLAGS(drr.drr_u.drr_begin.
 			    drr_versioninfo, featureflags);
-			(void) snprintf(drr.drr_u.drr_begin.drr_toname,
+			if (snprintf(drr.drr_u.drr_begin.drr_toname,
 			    sizeof (drr.drr_u.drr_begin.drr_toname),
-			    "%s@%s", zhp->zfs_name, tosnap);
+			    "%s@%s", zhp->zfs_name, tosnap) >=
+			    sizeof (drr.drr_u.drr_begin.drr_toname)) {
+				err = EINVAL;
+				goto stderr_out;
+			}
 			drr.drr_payloadlen = buflen;
 			err = cksum_and_write(&drr, sizeof (drr), &zc, outfd);
 
@@ -2689,7 +2693,8 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 	    ENOENT);
 
 	if (stream_avl != NULL) {
-		char *snapname;
+		char *snapname = NULL;
+		nvlist_t *lookup = NULL;
 		nvlist_t *fs = fsavl_find(stream_avl, drrb->drr_toguid,
 		    &snapname);
 		nvlist_t *props;
@@ -2710,6 +2715,11 @@ zfs_receive_one(libzfs_handle_t *hdl, int infd, const char *tosnap,
 			nvlist_free(props);
 		if (ret != 0)
 			return (-1);
+
+		if (0 == nvlist_lookup_nvlist(fs, "snapprops", &lookup)) {
+			VERIFY(0 == nvlist_lookup_nvlist(lookup,
+			    snapname, &snapprops_nvlist));
+		}
 	}
 
 	cp = NULL;

lib/libzfs/libzfs_util.c

@@ -1352,7 +1352,8 @@ zprop_print_one_property(const char *name, zprop_get_cbdata_t *cbp,
 			continue;
 		}
 
-		if (cbp->cb_columns[i + 1] == GET_COL_NONE)
+		if (i == (ZFS_GET_NCOLS - 1) ||
+		    cbp->cb_columns[i + 1] == GET_COL_NONE)
 			(void) printf("%s", str);
 		else if (cbp->cb_scripted)
 			(void) printf("%s\t", str);

man/man5/zfs-module-parameters.5

@@ -24,6 +24,19 @@ Description of the different parameters to the ZFS module.
 .sp
 .LP
 
+.sp
+.ne 2
+.na
+\fBignore_hole_birth\fR (int)
+.ad
+.RS 12n
+When set, the hole_birth optimization will not be used, and all holes will
+always be sent on zfs send. Useful if you suspect your datasets are affected
+by a bug in hole_birth.
+.sp
+Use \fB1\fR (default) for on and \fB0\fR for off.
+.RE
+
 .sp
 .ne 2
 .na
@@ -870,7 +883,12 @@ Default value: \fB10\fR.
 Minimum asynchronous write I/Os active to each device.
 See the section "ZFS I/O SCHEDULER".
 .sp
-Default value: \fB1\fR.
+Lower values are associated with better latency on rotational media but poorer
+resilver performance. The default value of 2 was chosen as a compromise. A
+value of 3 has been shown to improve resilver performance further at a cost of
+further increasing latency.
+.sp
+Default value: \fB2\fR.
 .RE
 
 .sp

module/avl/avl.c

@@ -630,7 +630,7 @@ avl_insert_here(
 void
 avl_add(avl_tree_t *tree, void *new_node)
 {
-	avl_index_t where;
+	avl_index_t where = 0;
 
 	/*
 	 * This is unfortunate.  We want to call panic() here, even for

module/nvpair/nvpair.c

@@ -21,6 +21,7 @@
 
 /*
  * Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2015, 2016 by Delphix. All rights reserved.
  */
 
 #include <sys/stropts.h>
@@ -138,6 +139,11 @@ static int nvlist_add_common(nvlist_t *nvl, const char *name, data_type_t type,
 #define	NVPAIR2I_NVP(nvp) \
 	((i_nvp_t *)((size_t)(nvp) - offsetof(i_nvp_t, nvi_nvp)))
 
+#ifdef _KERNEL
+int nvpair_max_recursion = 20;
+#else
+int nvpair_max_recursion = 100;
+#endif
 
 int
 nv_alloc_init(nv_alloc_t *nva, const nv_alloc_ops_t *nvo, /* args */ ...)
@@ -2017,6 +2023,7 @@ typedef struct {
 	const nvs_ops_t	*nvs_ops;
 	void		*nvs_private;
 	nvpriv_t	*nvs_priv;
+	int		nvs_recursion;
 } nvstream_t;
 
 /*
@@ -2168,9 +2175,16 @@ static int
 nvs_embedded(nvstream_t *nvs, nvlist_t *embedded)
 {
 	switch (nvs->nvs_op) {
-	case NVS_OP_ENCODE:
-		return (nvs_operation(nvs, embedded, NULL));
+	case NVS_OP_ENCODE: {
+		int err;
 
+		if (nvs->nvs_recursion >= nvpair_max_recursion)
+			return (EINVAL);
+		nvs->nvs_recursion++;
+		err = nvs_operation(nvs, embedded, NULL);
+		nvs->nvs_recursion--;
+		return (err);
+	}
 	case NVS_OP_DECODE: {
 		nvpriv_t *priv;
 		int err;
@@ -2183,8 +2197,12 @@ nvs_embedded(nvstream_t *nvs, nvlist_t *embedded)
 
 		nvlist_init(embedded, embedded->nvl_nvflag, priv);
 
+		if (nvs->nvs_recursion >= nvpair_max_recursion)
+			return (EINVAL);
+		nvs->nvs_recursion++;
 		if ((err = nvs_operation(nvs, embedded, NULL)) != 0)
 			nvlist_free(embedded);
+		nvs->nvs_recursion--;
 		return (err);
 	}
 	default:
@@ -2272,6 +2290,7 @@ nvlist_common(nvlist_t *nvl, char *buf, size_t *buflen, int encoding,
 		return (EINVAL);
 
 	nvs.nvs_op = nvs_op;
+	nvs.nvs_recursion = 0;
 
 	/*
 	 * For NVS_OP_ENCODE and NVS_OP_DECODE make sure an nvlist and

module/zfs/arc.c

@@ -1451,6 +1451,13 @@ arc_buf_info(arc_buf_t *ab, arc_buf_info_t *abi, int state_index)
 	l2arc_buf_hdr_t *l2hdr = NULL;
 	arc_state_t *state = NULL;
 
+	memset(abi, 0, sizeof (arc_buf_info_t));
+
+	if (hdr == NULL)
+		return;
+
+	abi->abi_flags = hdr->b_flags;
+
 	if (HDR_HAS_L1HDR(hdr)) {
 		l1hdr = &hdr->b_l1hdr;
 		state = l1hdr->b_state;
@@ -1458,9 +1465,6 @@ arc_buf_info(arc_buf_t *ab, arc_buf_info_t *abi, int state_index)
 	if (HDR_HAS_L2HDR(hdr))
 		l2hdr = &hdr->b_l2hdr;
 
-	memset(abi, 0, sizeof (arc_buf_info_t));
-	abi->abi_flags = hdr->b_flags;
-
 	if (l1hdr) {
 		abi->abi_datacnt = l1hdr->b_datacnt;
 		abi->abi_access = l1hdr->b_arc_access;
@@ -2697,12 +2701,7 @@ arc_prune_task(void *ptr)
 	if (func != NULL)
 		func(ap->p_adjust, ap->p_private);
 
-	/* Callback unregistered concurrently with execution */
-	if (refcount_remove(&ap->p_refcnt, func) == 0) {
-		ASSERT(!list_link_active(&ap->p_node));
-		refcount_destroy(&ap->p_refcnt);
-		kmem_free(ap, sizeof (*ap));
-	}
+	refcount_remove(&ap->p_refcnt, func);
 }
 
 /*
@@ -4320,17 +4319,11 @@ top:
 
 		/*
 		 * Gracefully handle a damaged logical block size as a
-		 * checksum error by passing a dummy zio to the done callback.
+		 * checksum error.
 		 */
 		if (size > spa_maxblocksize(spa)) {
-			if (done) {
-				rzio = zio_null(pio, spa, NULL,
-				    NULL, NULL, zio_flags);
-				rzio->io_error = ECKSUM;
-				done(rzio, buf, private);
-				zio_nowait(rzio);
-			}
-			rc = ECKSUM;
+			ASSERT3P(buf, ==, NULL);
+			rc = SET_ERROR(ECKSUM);
 			goto out;
 		}
 
@@ -4568,13 +4561,19 @@ arc_add_prune_callback(arc_prune_func_t *func, void *private)
 void
 arc_remove_prune_callback(arc_prune_t *p)
 {
+	boolean_t wait = B_FALSE;
 	mutex_enter(&arc_prune_mtx);
 	list_remove(&arc_prune_list, p);
-	if (refcount_remove(&p->p_refcnt, &arc_prune_list) == 0) {
-		refcount_destroy(&p->p_refcnt);
-		kmem_free(p, sizeof (*p));
-	}
+	if (refcount_remove(&p->p_refcnt, &arc_prune_list) > 0)
+		wait = B_TRUE;
 	mutex_exit(&arc_prune_mtx);
+
+	/* wait for arc_prune_task to finish */
+	if (wait)
+		taskq_wait_outstanding(arc_prune_taskq, 0);
+	ASSERT0(refcount_count(&p->p_refcnt));
+	refcount_destroy(&p->p_refcnt);
+	kmem_free(p, sizeof (*p));
 }
 
 void
@@ -5250,7 +5249,7 @@ arc_tuning_update(void)
 		arc_c_max = zfs_arc_max;
 		arc_c = arc_c_max;
 		arc_p = (arc_c >> 1);
-		arc_meta_limit = MIN(arc_meta_limit, arc_c_max);
+		arc_meta_limit = MIN(arc_meta_limit, (3 * arc_c_max) / 4);
 	}
 
 	/* Valid range: 32M - <arc_c_max> */
@@ -5476,11 +5475,12 @@ arc_init(void)
 	 * If it has been set by a module parameter, take that.
 	 * Otherwise, use a percentage of physical memory defined by
 	 * zfs_dirty_data_max_percent (default 10%) with a cap at
-	 * zfs_dirty_data_max_max (default 25% of physical memory).
+	 * zfs_dirty_data_max_max (default 4G or 25% of physical memory).
 	 */
 	if (zfs_dirty_data_max_max == 0)
-		zfs_dirty_data_max_max = (uint64_t)physmem * PAGESIZE *
-		    zfs_dirty_data_max_max_percent / 100;
+		zfs_dirty_data_max_max = MIN(4ULL * 1024 * 1024 * 1024,
+		    (uint64_t)physmem * PAGESIZE *
+		    zfs_dirty_data_max_max_percent / 100);
 
 	if (zfs_dirty_data_max == 0) {
 		zfs_dirty_data_max = (uint64_t)physmem * PAGESIZE *

module/zfs/dbuf.c

@@ -2628,6 +2628,22 @@ dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
 
 	if (db->db_blkid == DMU_SPILL_BLKID) {
 		mutex_enter(&dn->dn_mtx);
+		if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) {
+			/*
+			 * In the previous transaction group, the bonus buffer
+			 * was entirely used to store the attributes for the
+			 * dnode which overrode the dn_spill field.  However,
+			 * when adding more attributes to the file a spill
+			 * block was required to hold the extra attributes.
+			 *
+			 * Make sure to clear the garbage left in the dn_spill
+			 * field from the previous attributes in the bonus
+			 * buffer.  Otherwise, after writing out the spill
+			 * block to the new allocated dva, it will free
+			 * the old block pointed to by the invalid dn_spill.
+			 */
+			db->db_blkptr = NULL;
+		}
 		dn->dn_phys->dn_flags |= DNODE_FLAG_SPILL_BLKPTR;
 		mutex_exit(&dn->dn_mtx);
 	}

module/zfs/dbuf_stats.c

@@ -148,7 +148,6 @@ dbuf_stats_hash_table_data(char *buf, size_t size, void *data)
 		}
 
 		mutex_enter(&db->db_mtx);
-		mutex_exit(DBUF_HASH_MUTEX(h, dsh->idx));
 
 		if (db->db_state != DB_EVICTING) {
 			length = __dbuf_stats_hash_table_data(buf, size, db);
@@ -157,7 +156,6 @@ dbuf_stats_hash_table_data(char *buf, size_t size, void *data)
 		}
 
 		mutex_exit(&db->db_mtx);
-		mutex_enter(DBUF_HASH_MUTEX(h, dsh->idx));
 	}
 	mutex_exit(DBUF_HASH_MUTEX(h, dsh->idx));
 

module/zfs/dmu.c

@@ -49,6 +49,7 @@
 #ifdef _KERNEL
 #include <sys/vmsystm.h>
 #include <sys/zfs_znode.h>
+#include <linux/kmap_compat.h>
 #endif
 
 /*
@@ -1056,6 +1057,7 @@ dmu_bio_copy(void *arg_buf, int size, struct bio *bio, size_t bio_offset)
 	char *bv_buf;
 	int tocpy, bv_len, bv_offset;
 	int offset = 0;
+	void *paddr;
 
 	bio_for_each_segment4(bv, bvp, bio, iter) {
 
@@ -1080,14 +1082,15 @@ dmu_bio_copy(void *arg_buf, int size, struct bio *bio, size_t bio_offset)
 		tocpy = MIN(bv_len, size - offset);
 		ASSERT3S(tocpy, >=, 0);
 
-		bv_buf = page_address(bvp->bv_page) + bv_offset;
-		ASSERT3P(bv_buf, !=, NULL);
+		paddr = zfs_kmap_atomic(bvp->bv_page, KM_USER0);
+		bv_buf = paddr + bv_offset;
+		ASSERT3P(paddr, !=, NULL);
 
 		if (bio_data_dir(bio) == WRITE)
 			memcpy(arg_buf + offset, bv_buf, tocpy);
 		else
 			memcpy(bv_buf, arg_buf + offset, tocpy);
-
+		zfs_kunmap_atomic(paddr, KM_USER0);
 		offset += tocpy;
 	}
 out:

module/zfs/dmu_send.c

@@ -69,7 +69,7 @@ typedef struct dump_bytes_io {
 } dump_bytes_io_t;
 
 static void
-dump_bytes_strategy(void *arg)
+dump_bytes_cb(void *arg)
 {
 	dump_bytes_io_t *dbi = (dump_bytes_io_t *)arg;
 	dmu_sendarg_t *dsp = dbi->dbi_dsp;
@@ -96,6 +96,9 @@ dump_bytes(dmu_sendarg_t *dsp, void *buf, int len)
 	dbi.dbi_buf = buf;
 	dbi.dbi_len = len;
 
+#if defined(HAVE_LARGE_STACKS)
+	dump_bytes_cb(&dbi);
+#else
 	/*
 	 * The vn_rdwr() call is performed in a taskq to ensure that there is
 	 * always enough stack space to write safely to the target filesystem.
@@ -103,7 +106,8 @@ dump_bytes(dmu_sendarg_t *dsp, void *buf, int len)
 	 * them and they are used in vdev_file.c for a similar purpose.
 	 */
 	spa_taskq_dispatch_sync(dmu_objset_spa(dsp->dsa_os), ZIO_TYPE_FREE,
-	    ZIO_TASKQ_ISSUE, dump_bytes_strategy, &dbi, TQ_SLEEP);
+	    ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
+#endif /* HAVE_LARGE_STACKS */
 
 	return (dsp->dsa_err);
 }

module/zfs/dmu_traverse.c

@@ -39,6 +39,7 @@
 #include <sys/zfeature.h>
 
 int32_t zfs_pd_bytes_max = 50 * 1024 * 1024;	/* 50MB */
+int32_t ignore_hole_birth = 1;
 
 typedef struct prefetch_data {
 	kmutex_t pd_mtx;
@@ -250,7 +251,7 @@ traverse_visitbp(traverse_data_t *td, const dnode_phys_t *dnp,
 		 *
 		 * Note that the meta-dnode cannot be reallocated.
 		 */
-		if ((!td->td_realloc_possible ||
+		if (!ignore_hole_birth && (!td->td_realloc_possible ||
 			zb->zb_object == DMU_META_DNODE_OBJECT) &&
 			td->td_hole_birth_enabled_txg <= td->td_min_txg)
 			return (0);
@@ -692,4 +693,7 @@ EXPORT_SYMBOL(traverse_pool);
 
 module_param(zfs_pd_bytes_max, int, 0644);
 MODULE_PARM_DESC(zfs_pd_bytes_max, "Max number of bytes to prefetch");
+
+module_param(ignore_hole_birth, int, 0644);
+MODULE_PARM_DESC(ignore_hole_birth, "Ignore hole_birth txg for send");
 #endif

module/zfs/dsl_dataset.c

@@ -665,6 +665,21 @@ dsl_dataset_name(dsl_dataset_t *ds, char *name)
 	}
 }
 
+int
+dsl_dataset_namelen(dsl_dataset_t *ds)
+{
+	int len;
+	VERIFY0(dsl_dataset_get_snapname(ds));
+	mutex_enter(&ds->ds_lock);
+	len = strlen(ds->ds_snapname);
+	/* add '@' if ds is a snap */
+	if (len > 0)
+		len++;
+	len += dsl_dir_namelen(ds->ds_dir);
+	mutex_exit(&ds->ds_lock);
+	return (len);
+}
+
 void
 dsl_dataset_rele(dsl_dataset_t *ds, void *tag)
 {

module/zfs/dsl_pool.c

@@ -182,12 +182,20 @@ dsl_pool_init(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
 	int err;
 	dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
 
+	/*
+	 * Initialize the caller's dsl_pool_t structure before we actually open
+	 * the meta objset.  This is done because a self-healing write zio may
+	 * be issued as part of dmu_objset_open_impl() and the spa needs its
+	 * dsl_pool_t initialized in order to handle the write.
+	 */
+	*dpp = dp;
+
 	err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
 	    &dp->dp_meta_objset);
-	if (err != 0)
+	if (err != 0) {
 		dsl_pool_close(dp);
-	else
-		*dpp = dp;
+		*dpp = NULL;
+	}
 
 	return (err);
 }

module/zfs/refcount.c

@@ -137,7 +137,7 @@ refcount_add_many(refcount_t *rc, uint64_t number, void *holder)
 }
 
 int64_t
-refcount_add(refcount_t *rc, void *holder)
+zfs_refcount_add(refcount_t *rc, void *holder)
 {
 	return (refcount_add_many(rc, 1, holder));
 }

module/zfs/spa.c

@@ -845,7 +845,7 @@ spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
 	uint_t count = ztip->zti_count;
 	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
 	char name[32];
-	uint_t i, flags = TASKQ_DYNAMIC;
+	uint_t i, flags = 0;
 	boolean_t batch = B_FALSE;
 
 	if (mode == ZTI_MODE_NULL) {
@@ -863,6 +863,7 @@ spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
 	case ZTI_MODE_FIXED:
 		ASSERT3U(value, >=, 1);
 		value = MAX(value, 1);
+		flags |= TASKQ_DYNAMIC;
 		break;
 
 	case ZTI_MODE_BATCH:
@@ -1974,6 +1975,16 @@ spa_load_verify_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 	return (0);
 }
 
+/* ARGSUSED */
+int
+verify_dataset_name_len(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
+{
+	if (dsl_dataset_namelen(ds) >= ZFS_MAX_DATASET_NAME_LEN)
+		return (SET_ERROR(ENAMETOOLONG));
+
+	return (0);
+}
+
 static int
 spa_load_verify(spa_t *spa)
 {
@@ -1988,6 +1999,14 @@ spa_load_verify(spa_t *spa)
 	if (policy.zrp_request & ZPOOL_NEVER_REWIND)
 		return (0);
 
+	dsl_pool_config_enter(spa->spa_dsl_pool, FTAG);
+	error = dmu_objset_find_dp(spa->spa_dsl_pool,
+	    spa->spa_dsl_pool->dp_root_dir_obj, verify_dataset_name_len, NULL,
+	    DS_FIND_CHILDREN);
+	dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
+	if (error != 0)
+		return (error);
+
 	rio = zio_root(spa, NULL, &sle,
 	    ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE);
 
@@ -3844,211 +3863,6 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
 	return (0);
 }
 
-#ifdef _KERNEL
-/*
- * Get the root pool information from the root disk, then import the root pool
- * during the system boot up time.
- */
-extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **);
-
-static nvlist_t *
-spa_generate_rootconf(char *devpath, char *devid, uint64_t *guid)
-{
-	nvlist_t *config;
-	nvlist_t *nvtop, *nvroot;
-	uint64_t pgid;
-
-	if (vdev_disk_read_rootlabel(devpath, devid, &config) != 0)
-		return (NULL);
-
-	/*
-	 * Add this top-level vdev to the child array.
-	 */
-	VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
-	    &nvtop) == 0);
-	VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
-	    &pgid) == 0);
-	VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, guid) == 0);
-
-	/*
-	 * Put this pool's top-level vdevs into a root vdev.
-	 */
-	VERIFY(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, KM_SLEEP) == 0);
-	VERIFY(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
-	    VDEV_TYPE_ROOT) == 0);
-	VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) == 0);
-	VERIFY(nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, pgid) == 0);
-	VERIFY(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
-	    &nvtop, 1) == 0);
-
-	/*
-	 * Replace the existing vdev_tree with the new root vdev in
-	 * this pool's configuration (remove the old, add the new).
-	 */
-	VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
-	nvlist_free(nvroot);
-	return (config);
-}
-
-/*
- * Walk the vdev tree and see if we can find a device with "better"
- * configuration. A configuration is "better" if the label on that
- * device has a more recent txg.
- */
-static void
-spa_alt_rootvdev(vdev_t *vd, vdev_t **avd, uint64_t *txg)
-{
-	int c;
-
-	for (c = 0; c < vd->vdev_children; c++)
-		spa_alt_rootvdev(vd->vdev_child[c], avd, txg);
-
-	if (vd->vdev_ops->vdev_op_leaf) {
-		nvlist_t *label;
-		uint64_t label_txg;
-
-		if (vdev_disk_read_rootlabel(vd->vdev_physpath, vd->vdev_devid,
-		    &label) != 0)
-			return;
-
-		VERIFY(nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_TXG,
-		    &label_txg) == 0);
-
-		/*
-		 * Do we have a better boot device?
-		 */
-		if (label_txg > *txg) {
-			*txg = label_txg;
-			*avd = vd;
-		}
-		nvlist_free(label);
-	}
-}
-
-/*
- * Import a root pool.
- *
- * For x86. devpath_list will consist of devid and/or physpath name of
- * the vdev (e.g. "id1,sd@SSEAGATE..." or "/pci@1f,0/ide@d/disk@0,0:a").
- * The GRUB "findroot" command will return the vdev we should boot.
- *
- * For Sparc, devpath_list consists the physpath name of the booting device
- * no matter the rootpool is a single device pool or a mirrored pool.
- * e.g.
- *	"/pci@1f,0/ide@d/disk@0,0:a"
- */
-int
-spa_import_rootpool(char *devpath, char *devid)
-{
-	spa_t *spa;
-	vdev_t *rvd, *bvd, *avd = NULL;
-	nvlist_t *config, *nvtop;
-	uint64_t guid, txg;
-	char *pname;
-	int error;
-
-	/*
-	 * Read the label from the boot device and generate a configuration.
-	 */
-	config = spa_generate_rootconf(devpath, devid, &guid);
-#if defined(_OBP) && defined(_KERNEL)
-	if (config == NULL) {
-		if (strstr(devpath, "/iscsi/ssd") != NULL) {
-			/* iscsi boot */
-			get_iscsi_bootpath_phy(devpath);
-			config = spa_generate_rootconf(devpath, devid, &guid);
-		}
-	}
-#endif
-	if (config == NULL) {
-		cmn_err(CE_NOTE, "Cannot read the pool label from '%s'",
-		    devpath);
-		return (SET_ERROR(EIO));
-	}
-
-	VERIFY(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
-	    &pname) == 0);
-	VERIFY(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, &txg) == 0);
-
-	mutex_enter(&spa_namespace_lock);
-	if ((spa = spa_lookup(pname)) != NULL) {
-		/*
-		 * Remove the existing root pool from the namespace so that we
-		 * can replace it with the correct config we just read in.
-		 */
-		spa_remove(spa);
-	}
-
-	spa = spa_add(pname, config, NULL);
-	spa->spa_is_root = B_TRUE;
-	spa->spa_import_flags = ZFS_IMPORT_VERBATIM;
-
-	/*
-	 * Build up a vdev tree based on the boot device's label config.
-	 */
-	VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
-	    &nvtop) == 0);
-	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
-	error = spa_config_parse(spa, &rvd, nvtop, NULL, 0,
-	    VDEV_ALLOC_ROOTPOOL);
-	spa_config_exit(spa, SCL_ALL, FTAG);
-	if (error) {
-		mutex_exit(&spa_namespace_lock);
-		nvlist_free(config);
-		cmn_err(CE_NOTE, "Can not parse the config for pool '%s'",
-		    pname);
-		return (error);
-	}
-
-	/*
-	 * Get the boot vdev.
-	 */
-	if ((bvd = vdev_lookup_by_guid(rvd, guid)) == NULL) {
-		cmn_err(CE_NOTE, "Can not find the boot vdev for guid %llu",
-		    (u_longlong_t)guid);
-		error = SET_ERROR(ENOENT);
-		goto out;
-	}
-
-	/*
-	 * Determine if there is a better boot device.
-	 */
-	avd = bvd;
-	spa_alt_rootvdev(rvd, &avd, &txg);
-	if (avd != bvd) {
-		cmn_err(CE_NOTE, "The boot device is 'degraded'. Please "
-		    "try booting from '%s'", avd->vdev_path);
-		error = SET_ERROR(EINVAL);
-		goto out;
-	}
-
-	/*
-	 * If the boot device is part of a spare vdev then ensure that
-	 * we're booting off the active spare.
-	 */
-	if (bvd->vdev_parent->vdev_ops == &vdev_spare_ops &&
-	    !bvd->vdev_isspare) {
-		cmn_err(CE_NOTE, "The boot device is currently spared. Please "
-		    "try booting from '%s'",
-		    bvd->vdev_parent->
-		    vdev_child[bvd->vdev_parent->vdev_children - 1]->vdev_path);
-		error = SET_ERROR(EINVAL);
-		goto out;
-	}
-
-	error = 0;
-out:
-	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
-	vdev_free(rvd);
-	spa_config_exit(spa, SCL_ALL, FTAG);
-	mutex_exit(&spa_namespace_lock);
-
-	nvlist_free(config);
-	return (error);
-}
-
-#endif
-
 /*
  * Import a non-root pool into the system.
  */
@@ -4148,12 +3962,6 @@ spa_import(char *pool, nvlist_t *config, nvlist_t *props, uint64_t flags)
 
 	VERIFY(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
 	    &nvroot) == 0);
-	if (error == 0)
-		error = spa_validate_aux(spa, nvroot, -1ULL,
-		    VDEV_ALLOC_SPARE);
-	if (error == 0)
-		error = spa_validate_aux(spa, nvroot, -1ULL,
-		    VDEV_ALLOC_L2CACHE);
 	spa_config_exit(spa, SCL_ALL, FTAG);
 
 	if (props != NULL)
@@ -6762,7 +6570,6 @@ EXPORT_SYMBOL(spa_open);
 EXPORT_SYMBOL(spa_open_rewind);
 EXPORT_SYMBOL(spa_get_stats);
 EXPORT_SYMBOL(spa_create);
-EXPORT_SYMBOL(spa_import_rootpool);
 EXPORT_SYMBOL(spa_import);
 EXPORT_SYMBOL(spa_tryimport);
 EXPORT_SYMBOL(spa_destroy);

module/zfs/spa_config.c

@@ -174,7 +174,7 @@ spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
 	VERIFY(nvlist_pack(nvl, &buf, &buflen, NV_ENCODE_XDR,
 	    KM_SLEEP) == 0);
 
-#ifdef __linux__
+#if defined(__linux__) && defined(_KERNEL)
 	/*
 	 * Write the configuration to disk.  Due to the complexity involved
 	 * in performing a rename from within the kernel the file is truncated
@@ -201,7 +201,8 @@ spa_config_write(spa_config_dirent_t *dp, nvlist_t *nvl)
 	 */
 	(void) snprintf(temp, MAXPATHLEN, "%s.tmp", dp->scd_path);
 
-	if (vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0) == 0) {
+	error = vn_open(temp, UIO_SYSSPACE, oflags, 0644, &vp, CRCREAT, 0);
+	if (error == 0) {
 		if (vn_rdwr(UIO_WRITE, vp, buf, buflen, 0, UIO_SYSSPACE,
 		    0, RLIM64_INFINITY, kcred, NULL) == 0 &&
 		    VOP_FSYNC(vp, FSYNC, kcred, NULL) == 0) {

module/zfs/vdev.c

@@ -877,11 +877,11 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 
 	ASSERT(oldc <= newc);
 
-	mspp = kmem_zalloc(newc * sizeof (*mspp), KM_SLEEP);
+	mspp = vmem_zalloc(newc * sizeof (*mspp), KM_SLEEP);
 
 	if (oldc != 0) {
 		bcopy(vd->vdev_ms, mspp, oldc * sizeof (*mspp));
-		kmem_free(vd->vdev_ms, oldc * sizeof (*mspp));
+		vmem_free(vd->vdev_ms, oldc * sizeof (*mspp));
 	}
 
 	vd->vdev_ms = mspp;
@@ -935,7 +935,7 @@ vdev_metaslab_fini(vdev_t *vd)
 			if (msp != NULL)
 				metaslab_fini(msp);
 		}
-		kmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
+		vmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
 		vd->vdev_ms = NULL;
 	}
 
@@ -1407,7 +1407,7 @@ vdev_validate(vdev_t *vd, boolean_t strict)
 		    spa_last_synced_txg(spa) : -1ULL;
 
 		if ((label = vdev_label_read_config(vd, txg)) == NULL) {
-			vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
+			vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
 			    VDEV_AUX_BAD_LABEL);
 			return (0);
 		}
@@ -1799,6 +1799,9 @@ vdev_dtl_should_excise(vdev_t *vd)
 	ASSERT0(scn->scn_phys.scn_errors);
 	ASSERT0(vd->vdev_children);
 
+	if (vd->vdev_state < VDEV_STATE_DEGRADED)
+		return (B_FALSE);
+
 	if (vd->vdev_resilver_txg == 0 ||
 	    range_tree_space(vd->vdev_dtl[DTL_MISSING]) == 0)
 		return (B_TRUE);
@@ -1894,12 +1897,15 @@ vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg, int scrub_done)
 
 		/*
 		 * If the vdev was resilvering and no longer has any
-		 * DTLs then reset its resilvering flag.
+		 * DTLs then reset its resilvering flag and dirty
+		 * the top level so that we persist the change.
 		 */
 		if (vd->vdev_resilver_txg != 0 &&
 		    range_tree_space(vd->vdev_dtl[DTL_MISSING]) == 0 &&
-		    range_tree_space(vd->vdev_dtl[DTL_OUTAGE]) == 0)
+		    range_tree_space(vd->vdev_dtl[DTL_OUTAGE]) == 0) {
 			vd->vdev_resilver_txg = 0;
+			vdev_config_dirty(vd->vdev_top);
+		}
 
 		mutex_exit(&vd->vdev_dtl_lock);
 

module/zfs/vdev_disk.c

@@ -41,10 +41,8 @@ static void *zfs_vdev_holder = VDEV_HOLDER;
  * Virtual device vector for disks.
  */
 typedef struct dio_request {
-	struct completion	dr_comp;	/* Completion for sync IO */
 	zio_t			*dr_zio;	/* Parent ZIO */
 	atomic_t		dr_ref;		/* References */
-	int			dr_wait;	/* Wait for IO */
 	int			dr_error;	/* Bio error */
 	int			dr_bio_count;	/* Count of bio's */
 	struct bio		*dr_bio[0];	/* Attached bio's */
@@ -139,7 +137,7 @@ vdev_elevator_switch(vdev_t *v, char *elevator)
 		return (0);
 
 	/* Leave existing scheduler when set to "none" */
-	if (strncmp(elevator, "none", 4) && (strlen(elevator) == 4) == 0)
+	if ((strncmp(elevator, "none", 4) == 0) && (strlen(elevator) == 4))
 		return (0);
 
 #ifdef HAVE_ELEVATOR_CHANGE
@@ -244,12 +242,12 @@ vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *max_psize,
 {
 	struct block_device *bdev = ERR_PTR(-ENXIO);
 	vdev_disk_t *vd;
-	int mode, block_size;
+	int count = 0, mode, block_size;
 
 	/* Must have a pathname and it must be absolute. */
 	if (v->vdev_path == NULL || v->vdev_path[0] != '/') {
 		v->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
-		return (EINVAL);
+		return (SET_ERROR(EINVAL));
 	}
 
 	/*
@@ -264,7 +262,7 @@ vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *max_psize,
 
 	vd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP);
 	if (vd == NULL)
-		return (ENOMEM);
+		return (SET_ERROR(ENOMEM));
 
 	/*
 	 * Devices are always opened by the path provided at configuration
@@ -279,16 +277,35 @@ vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *max_psize,
 	 * /dev/[hd]d devices which may be reordered due to probing order.
 	 * Devices in the wrong locations will be detected by the higher
 	 * level vdev validation.
+	 *
+	 * The specified paths may be briefly removed and recreated in
+	 * response to udev events.  This should be exceptionally unlikely
+	 * because the zpool command makes every effort to verify these paths
+	 * have already settled prior to reaching this point.  Therefore,
+	 * a ENOENT failure at this point is highly likely to be transient
+	 * and it is reasonable to sleep and retry before giving up.  In
+	 * practice delays have been observed to be on the order of 100ms.
 	 */
 	mode = spa_mode(v->vdev_spa);
 	if (v->vdev_wholedisk && v->vdev_expanding)
 		bdev = vdev_disk_rrpart(v->vdev_path, mode, vd);
-	if (IS_ERR(bdev))
+
+	while (IS_ERR(bdev) && count < 50) {
 		bdev = vdev_bdev_open(v->vdev_path,
 		    vdev_bdev_mode(mode), zfs_vdev_holder);
+		if (unlikely(PTR_ERR(bdev) == -ENOENT)) {
+			msleep(10);
+			count++;
+		} else if (IS_ERR(bdev)) {
+			break;
+		}
+	}
+
 	if (IS_ERR(bdev)) {
+		dprintf("failed open v->vdev_path=%s, error=%d count=%d\n",
+		    v->vdev_path, -PTR_ERR(bdev), count);
 		kmem_free(vd, sizeof (vdev_disk_t));
-		return (-PTR_ERR(bdev));
+		return (SET_ERROR(-PTR_ERR(bdev)));
 	}
 
 	v->vdev_tsd = vd;
@@ -344,7 +361,6 @@ vdev_disk_dio_alloc(int bio_count)
 	dr = kmem_zalloc(sizeof (dio_request_t) +
 	    sizeof (struct bio *) * bio_count, KM_SLEEP);
 	if (dr) {
-		init_completion(&dr->dr_comp);
 		atomic_set(&dr->dr_ref, 0);
 		dr->dr_bio_count = bio_count;
 		dr->dr_error = 0;
@@ -407,7 +423,6 @@ BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, error)
 {
 	dio_request_t *dr = bio->bi_private;
 	int rc;
-	int wait;
 
 	if (dr->dr_error == 0) {
 #ifdef HAVE_1ARG_BIO_END_IO_T
@@ -420,13 +435,8 @@ BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, error)
 #endif
 	}
 
-	wait = dr->dr_wait;
 	/* Drop reference aquired by __vdev_disk_physio */
 	rc = vdev_disk_dio_put(dr);
-
-	/* Wake up synchronous waiter this is the last outstanding bio */
-	if (wait && rc == 1)
-		complete(&dr->dr_comp);
 }
 
 static inline unsigned long
@@ -476,30 +486,43 @@ bio_map(struct bio *bio, void *bio_ptr, unsigned int bio_size)
 }
 
 static inline void
-vdev_submit_bio(int rw, struct bio *bio)
+vdev_submit_bio_impl(struct bio *bio)
+{
+#ifdef HAVE_1ARG_SUBMIT_BIO
+	submit_bio(bio);
+#else
+	submit_bio(0, bio);
+#endif
+}
+
+static inline void
+vdev_submit_bio(struct bio *bio)
 {
 #ifdef HAVE_CURRENT_BIO_TAIL
 	struct bio **bio_tail = current->bio_tail;
 	current->bio_tail = NULL;
-	submit_bio(rw, bio);
+	vdev_submit_bio_impl(bio);
 	current->bio_tail = bio_tail;
 #else
 	struct bio_list *bio_list = current->bio_list;
 	current->bio_list = NULL;
-	submit_bio(rw, bio);
+	vdev_submit_bio_impl(bio);
 	current->bio_list = bio_list;
 #endif
 }
 
 static int
 __vdev_disk_physio(struct block_device *bdev, zio_t *zio, caddr_t kbuf_ptr,
-    size_t kbuf_size, uint64_t kbuf_offset, int flags, int wait)
+    size_t kbuf_size, uint64_t kbuf_offset, int rw, int flags)
 {
 	dio_request_t *dr;
 	caddr_t bio_ptr;
 	uint64_t bio_offset;
-	int rw, bio_size, bio_count = 16;
+	int bio_size, bio_count = 16;
 	int i = 0, error = 0;
+#if defined(HAVE_BLK_QUEUE_HAVE_BLK_PLUG)
+	struct blk_plug plug;
+#endif
 
 	ASSERT3U(kbuf_offset + kbuf_size, <=, bdev->bd_inode->i_size);
 
@@ -511,9 +534,7 @@ retry:
 	if (zio && !(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD)))
 		bio_set_flags_failfast(bdev, &flags);
 
-	rw = flags;
 	dr->dr_zio = zio;
-	dr->dr_wait = wait;
 
 	/*
 	 * When the IO size exceeds the maximum bio size for the request
@@ -555,9 +576,9 @@ retry:
 
 		dr->dr_bio[i]->bi_bdev = bdev;
 		BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9;
-		dr->dr_bio[i]->bi_rw = rw;
 		dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion;
 		dr->dr_bio[i]->bi_private = dr;
+		bio_set_op_attrs(dr->dr_bio[i], rw, flags);
 
 		/* Remaining size is returned to become the new size */
 		bio_size = bio_map(dr->dr_bio[i], bio_ptr, bio_size);
@@ -572,38 +593,26 @@ retry:
 	if (zio)
 		zio->io_delay = jiffies_64;
 
+#if defined(HAVE_BLK_QUEUE_HAVE_BLK_PLUG)
+	if (dr->dr_bio_count > 1)
+		blk_start_plug(&plug);
+#endif
+
 	/* Submit all bio's associated with this dio */
 	for (i = 0; i < dr->dr_bio_count; i++)
 		if (dr->dr_bio[i])
-			vdev_submit_bio(rw, dr->dr_bio[i]);
+			vdev_submit_bio(dr->dr_bio[i]);
 
-	/*
-	 * On synchronous blocking requests we wait for all bio the completion
-	 * callbacks to run.  We will be woken when the last callback runs
-	 * for this dio.  We are responsible for putting the last dio_request
-	 * reference will in turn put back the last bio references.  The
-	 * only synchronous consumer is vdev_disk_read_rootlabel() all other
-	 * IO originating from vdev_disk_io_start() is asynchronous.
-	 */
-	if (wait) {
-		wait_for_completion(&dr->dr_comp);
-		error = dr->dr_error;
-		ASSERT3S(atomic_read(&dr->dr_ref), ==, 1);
-	}
+#if defined(HAVE_BLK_QUEUE_HAVE_BLK_PLUG)
+	if (dr->dr_bio_count > 1)
+		blk_finish_plug(&plug);
+#endif
 
 	(void) vdev_disk_dio_put(dr);
 
 	return (error);
 }
 
-int
-vdev_disk_physio(struct block_device *bdev, caddr_t kbuf,
-    size_t size, uint64_t offset, int flags)
-{
-	bio_set_flags_failfast(bdev, &flags);
-	return (__vdev_disk_physio(bdev, NULL, kbuf, size, offset, flags, 1));
-}
-
 BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, rc)
 {
 	zio_t *zio = bio->bi_private;
@@ -642,7 +651,8 @@ vdev_disk_io_flush(struct block_device *bdev, zio_t *zio)
 	bio->bi_private = zio;
 	bio->bi_bdev = bdev;
 	zio->io_delay = jiffies_64;
-	vdev_submit_bio(VDEV_WRITE_FLUSH_FUA, bio);
+	bio_set_flush(bio);
+	vdev_submit_bio(bio);
 	invalidate_bdev(bdev);
 
 	return (0);
@@ -653,8 +663,7 @@ vdev_disk_io_start(zio_t *zio)
 {
 	vdev_t *v = zio->io_vd;
 	vdev_disk_t *vd = v->vdev_tsd;
-	zio_priority_t pri = zio->io_priority;
-	int flags, error;
+	int rw, flags, error;
 
 	switch (zio->io_type) {
 	case ZIO_TYPE_IOCTL:
@@ -693,17 +702,25 @@ vdev_disk_io_start(zio_t *zio)
 		zio_execute(zio);
 		return;
 	case ZIO_TYPE_WRITE:
-		if ((pri == ZIO_PRIORITY_SYNC_WRITE) && (v->vdev_nonrot))
-			flags = WRITE_SYNC;
-		else
-			flags = WRITE;
+		rw = WRITE;
+#if defined(HAVE_BLK_QUEUE_HAVE_BIO_RW_UNPLUG)
+		flags = (1 << BIO_RW_UNPLUG);
+#elif defined(REQ_UNPLUG)
+		flags = REQ_UNPLUG;
+#else
+		flags = 0;
+#endif
 		break;
 
 	case ZIO_TYPE_READ:
-		if ((pri == ZIO_PRIORITY_SYNC_READ) && (v->vdev_nonrot))
-			flags = READ_SYNC;
-		else
-			flags = READ;
+		rw = READ;
+#if defined(HAVE_BLK_QUEUE_HAVE_BIO_RW_UNPLUG)
+		flags = (1 << BIO_RW_UNPLUG);
+#elif defined(REQ_UNPLUG)
+		flags = REQ_UNPLUG;
+#else
+		flags = 0;
+#endif
 		break;
 
 	default:
@@ -713,7 +730,7 @@ vdev_disk_io_start(zio_t *zio)
 	}
 
 	error = __vdev_disk_physio(vd->vd_bdev, zio, zio->io_data,
-	    zio->io_size, zio->io_offset, flags, 0);
+	    zio->io_size, zio->io_offset, rw, flags);
 	if (error) {
 		zio->io_error = error;
 		zio_interrupt(zio);
@@ -783,68 +800,5 @@ vdev_ops_t vdev_disk_ops = {
 	B_TRUE			/* leaf vdev */
 };
 
-/*
- * Given the root disk device devid or pathname, read the label from
- * the device, and construct a configuration nvlist.
- */
-int
-vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config)
-{
-	struct block_device *bdev;
-	vdev_label_t *label;
-	uint64_t s, size;
-	int i;
-
-	bdev = vdev_bdev_open(devpath, vdev_bdev_mode(FREAD), zfs_vdev_holder);
-	if (IS_ERR(bdev))
-		return (-PTR_ERR(bdev));
-
-	s = bdev_capacity(bdev);
-	if (s == 0) {
-		vdev_bdev_close(bdev, vdev_bdev_mode(FREAD));
-		return (EIO);
-	}
-
-	size = P2ALIGN_TYPED(s, sizeof (vdev_label_t), uint64_t);
-	label = vmem_alloc(sizeof (vdev_label_t), KM_SLEEP);
-
-	for (i = 0; i < VDEV_LABELS; i++) {
-		uint64_t offset, state, txg = 0;
-
-		/* read vdev label */
-		offset = vdev_label_offset(size, i, 0);
-		if (vdev_disk_physio(bdev, (caddr_t)label,
-		    VDEV_SKIP_SIZE + VDEV_PHYS_SIZE, offset, READ_SYNC) != 0)
-			continue;
-
-		if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
-		    sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) {
-			*config = NULL;
-			continue;
-		}
-
-		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
-		    &state) != 0 || state >= POOL_STATE_DESTROYED) {
-			nvlist_free(*config);
-			*config = NULL;
-			continue;
-		}
-
-		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
-		    &txg) != 0 || txg == 0) {
-			nvlist_free(*config);
-			*config = NULL;
-			continue;
-		}
-
-		break;
-	}
-
-	vmem_free(label, sizeof (vdev_label_t));
-	vdev_bdev_close(bdev, vdev_bdev_mode(FREAD));
-
-	return (0);
-}
-
 module_param(zfs_vdev_scheduler, charp, 0644);
 MODULE_PARM_DESC(zfs_vdev_scheduler, "I/O scheduler");

module/zfs/vdev_queue.c

@@ -24,7 +24,7 @@
  */
 
 /*
- * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -146,7 +146,7 @@ uint32_t zfs_vdev_sync_write_min_active = 10;
 uint32_t zfs_vdev_sync_write_max_active = 10;
 uint32_t zfs_vdev_async_read_min_active = 1;
 uint32_t zfs_vdev_async_read_max_active = 3;
-uint32_t zfs_vdev_async_write_min_active = 1;
+uint32_t zfs_vdev_async_write_min_active = 2;
 uint32_t zfs_vdev_async_write_max_active = 10;
 uint32_t zfs_vdev_scrub_min_active = 1;
 uint32_t zfs_vdev_scrub_max_active = 2;
@@ -249,20 +249,29 @@ static int
 vdev_queue_max_async_writes(spa_t *spa)
 {
 	int writes;
-	uint64_t dirty = spa->spa_dsl_pool->dp_dirty_total;
+	uint64_t dirty = 0;
+	dsl_pool_t *dp = spa_get_dsl(spa);
 	uint64_t min_bytes = zfs_dirty_data_max *
 	    zfs_vdev_async_write_active_min_dirty_percent / 100;
 	uint64_t max_bytes = zfs_dirty_data_max *
 	    zfs_vdev_async_write_active_max_dirty_percent / 100;
 
+	/*
+	 * Async writes may occur before the assignment of the spa's
+	 * dsl_pool_t if a self-healing zio is issued prior to the
+	 * completion of dmu_objset_open_impl().
+	 */
+	if (dp == NULL)
+		return (zfs_vdev_async_write_max_active);
+
 	/*
 	 * Sync tasks correspond to interactive user actions. To reduce the
 	 * execution time of those actions we push data out as fast as possible.
 	 */
-	if (spa_has_pending_synctask(spa)) {
+	if (spa_has_pending_synctask(spa))
 		return (zfs_vdev_async_write_max_active);
-	}
 
+	dirty = dp->dp_dirty_total;
 	if (dirty < min_bytes)
 		return (zfs_vdev_async_write_min_active);
 	if (dirty > max_bytes)
@@ -536,7 +545,7 @@ vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio)
 
 	/*
 	 * Walk backwards through sufficiently contiguous I/Os
-	 * recording the last non-option I/O.
+	 * recording the last non-optional I/O.
 	 */
 	while ((dio = AVL_PREV(t, first)) != NULL &&
 	    (dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
@@ -558,10 +567,14 @@ vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio)
 
 	/*
 	 * Walk forward through sufficiently contiguous I/Os.
+	 * The aggregation limit does not apply to optional i/os, so that
+	 * we can issue contiguous writes even if they are larger than the
+	 * aggregation limit.
 	 */
 	while ((dio = AVL_NEXT(t, last)) != NULL &&
 	    (dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags &&
-	    IO_SPAN(first, dio) <= zfs_vdev_aggregation_limit &&
+	    (IO_SPAN(first, dio) <= zfs_vdev_aggregation_limit ||
+	    (dio->io_flags & ZIO_FLAG_OPTIONAL)) &&
 	    IO_GAP(last, dio) <= maxgap) {
 		last = dio;
 		if (!(last->io_flags & ZIO_FLAG_OPTIONAL))
@@ -596,6 +609,7 @@ vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio)
 		dio = AVL_NEXT(t, last);
 		dio->io_flags &= ~ZIO_FLAG_OPTIONAL;
 	} else {
+		/* do not include the optional i/o */
 		while (last != mandatory && last != first) {
 			ASSERT(last->io_flags & ZIO_FLAG_OPTIONAL);
 			last = AVL_PREV(t, last);
@@ -607,7 +621,6 @@ vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio)
 		return (NULL);
 
 	size = IO_SPAN(first, last);
-	ASSERT3U(size, <=, zfs_vdev_aggregation_limit);
 
 	buf = zio_buf_alloc_flags(size, KM_NOSLEEP);
 	if (buf == NULL)

module/zfs/zfs_acl.c

@@ -1457,7 +1457,7 @@ zfs_aclset_common(znode_t *zp, zfs_acl_t *aclp, cred_t *cr, dmu_tx_t *tx)
 	if (ace_trivial_common(aclp, 0, zfs_ace_walk) == 0)
 		zp->z_pflags |= ZFS_ACL_TRIVIAL;
 
-	zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime, B_TRUE);
+	zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
 	return (sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
 }
 
@@ -2473,53 +2473,33 @@ zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
 {
 	uint32_t	working_mode;
 	int		error;
-	boolean_t	check_privs;
-	znode_t		*check_zp = zp;
+	int		is_attr;
+	boolean_t 	check_privs;
+	znode_t		*xzp;
+	znode_t 	*check_zp = zp;
 	mode_t		needed_bits;
 	uid_t		owner;
 
+	is_attr = ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode));
+
 	/*
 	 * If attribute then validate against base file
 	 */
-	if ((zp->z_pflags & ZFS_XATTR) && S_ISDIR(ZTOI(zp)->i_mode)) {
+	if (is_attr) {
 		uint64_t	parent;
 
-		rw_enter(&zp->z_xattr_lock, RW_READER);
-		if (zp->z_xattr_parent) {
-			check_zp = zp->z_xattr_parent;
-			rw_exit(&zp->z_xattr_lock);
+		if ((error = sa_lookup(zp->z_sa_hdl,
+		    SA_ZPL_PARENT(ZTOZSB(zp)), &parent,
+		    sizeof (parent))) != 0)
+			return (error);
 
-			/*
-			 * Verify a lookup yields the same znode.
-			 */
-			ASSERT3S(sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
-			    ZTOZSB(zp)), &parent, sizeof (parent)), ==, 0);
-			ASSERT3U(check_zp->z_id, ==, parent);
-		} else {
-			rw_exit(&zp->z_xattr_lock);
-
-			error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(
-			    ZTOZSB(zp)), &parent, sizeof (parent));
-			if (error)
-				return (error);
-
-			/*
-			 * Cache the lookup on the parent file znode as
-			 * zp->z_xattr_parent and hold a reference.  This
-			 * effectively pins the parent in memory until all
-			 * child xattr znodes have been destroyed and
-			 * release their references in zfs_inode_destroy().
-			 */
-			error = zfs_zget(ZTOZSB(zp), parent, &check_zp);
-			if (error)
-				return (error);
-
-			rw_enter(&zp->z_xattr_lock, RW_WRITER);
-			if (zp->z_xattr_parent == NULL)
-				zp->z_xattr_parent = check_zp;
-			rw_exit(&zp->z_xattr_lock);
+		if ((error = zfs_zget(ZTOZSB(zp),
+		    parent, &xzp)) != 0)	{
+			return (error);
 		}
 
+		check_zp = xzp;
+
 		/*
 		 * fixup mode to map to xattr perms
 		 */
@@ -2561,11 +2541,15 @@ zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
 
 	if ((error = zfs_zaccess_common(check_zp, mode, &working_mode,
 	    &check_privs, skipaclchk, cr)) == 0) {
+		if (is_attr)
+			iput(ZTOI(xzp));
 		return (secpolicy_vnode_access2(cr, ZTOI(zp), owner,
 		    needed_bits, needed_bits));
 	}
 
 	if (error && !check_privs) {
+		if (is_attr)
+			iput(ZTOI(xzp));
 		return (error);
 	}
 
@@ -2626,6 +2610,9 @@ zfs_zaccess(znode_t *zp, int mode, int flags, boolean_t skipaclchk, cred_t *cr)
 		    needed_bits, needed_bits);
 	}
 
+	if (is_attr)
+		iput(ZTOI(xzp));
+
 	return (error);
 }
 

module/zfs/zfs_ctldir.c

@@ -455,7 +455,7 @@ static struct inode *
 zfsctl_inode_alloc(zfs_sb_t *zsb, uint64_t id,
     const struct file_operations *fops, const struct inode_operations *ops)
 {
-	struct timespec now = current_fs_time(zsb->z_sb);
+	struct timespec now;
 	struct inode *ip;
 	znode_t *zp;
 
@@ -463,6 +463,7 @@ zfsctl_inode_alloc(zfs_sb_t *zsb, uint64_t id,
 	if (ip == NULL)
 		return (NULL);
 
+	now = current_time(ip);
 	zp = ITOZ(ip);
 	ASSERT3P(zp->z_dirlocks, ==, NULL);
 	ASSERT3P(zp->z_acl_cached, ==, NULL);
@@ -478,15 +479,12 @@ zfsctl_inode_alloc(zfs_sb_t *zsb, uint64_t id,
 	zp->z_mapcnt = 0;
 	zp->z_gen = 0;
 	zp->z_size = 0;
-	zp->z_atime[0] = 0;
-	zp->z_atime[1] = 0;
 	zp->z_links = 0;
 	zp->z_pflags = 0;
 	zp->z_uid = 0;
 	zp->z_gid = 0;
 	zp->z_mode = 0;
 	zp->z_sync_cnt = 0;
-	zp->z_is_zvol = B_FALSE;
 	zp->z_is_mapped = B_FALSE;
 	zp->z_is_ctldir = B_TRUE;
 	zp->z_is_sa = B_FALSE;
@@ -501,6 +499,9 @@ zfsctl_inode_alloc(zfs_sb_t *zsb, uint64_t id,
 	ip->i_ctime = now;
 	ip->i_fop = fops;
 	ip->i_op = ops;
+#if defined(IOP_XATTR)
+	ip->i_opflags &= ~IOP_XATTR;
+#endif
 
 	if (insert_inode_locked(ip)) {
 		unlock_new_inode(ip);
@@ -1010,16 +1011,11 @@ out:
  * best effort.  In the case where it does fail, perhaps because
  * it's in use, the unmount will fail harmlessly.
  */
-#define	SET_UNMOUNT_CMD \
-	"exec 0</dev/null " \
-	"     1>/dev/null " \
-	"     2>/dev/null; " \
-	"umount -t zfs -n %s'%s'"
-
 int
 zfsctl_snapshot_unmount(char *snapname, int flags)
 {
-	char *argv[] = { "/bin/sh", "-c", NULL, NULL };
+	char *argv[] = { "/usr/bin/env", "umount", "-t", "zfs", "-n", NULL,
+	    NULL };
 	char *envp[] = { NULL };
 	zfs_snapentry_t *se;
 	int error;
@@ -1031,12 +1027,12 @@ zfsctl_snapshot_unmount(char *snapname, int flags)
 	}
 	rw_exit(&zfs_snapshot_lock);
 
-	argv[2] = kmem_asprintf(SET_UNMOUNT_CMD,
-	    flags & MNT_FORCE ? "-f " : "", se->se_path);
-	zfsctl_snapshot_rele(se);
+	if (flags & MNT_FORCE)
+		argv[4] = "-fn";
+	argv[5] = se->se_path;
 	dprintf("unmount; path=%s\n", se->se_path);
 	error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
-	strfree(argv[2]);
+	zfsctl_snapshot_rele(se);
 
 
 	/*
@@ -1051,11 +1047,6 @@ zfsctl_snapshot_unmount(char *snapname, int flags)
 }
 
 #define	MOUNT_BUSY 0x80		/* Mount failed due to EBUSY (from mntent.h) */
-#define	SET_MOUNT_CMD \
-	"exec 0</dev/null " \
-	"     1>/dev/null " \
-	"     2>/dev/null; " \
-	"mount -t zfs -n '%s' '%s'"
 
 int
 zfsctl_snapshot_mount(struct path *path, int flags)
@@ -1066,7 +1057,8 @@ zfsctl_snapshot_mount(struct path *path, int flags)
 	zfs_sb_t *snap_zsb;
 	zfs_snapentry_t *se;
 	char *full_name, *full_path;
-	char *argv[] = { "/bin/sh", "-c", NULL, NULL };
+	char *argv[] = { "/usr/bin/env", "mount", "-t", "zfs", "-n", NULL, NULL,
+	    NULL };
 	char *envp[] = { NULL };
 	int error;
 	struct path spath;
@@ -1111,9 +1103,9 @@ zfsctl_snapshot_mount(struct path *path, int flags)
 	 * value from call_usermodehelper() will be (exitcode << 8 + signal).
 	 */
 	dprintf("mount; name=%s path=%s\n", full_name, full_path);
-	argv[2] = kmem_asprintf(SET_MOUNT_CMD, full_name, full_path);
+	argv[5] = full_name;
+	argv[6] = full_path;
 	error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
-	strfree(argv[2]);
 	if (error) {
 		if (!(error & MOUNT_BUSY << 8)) {
 			cmn_err(CE_WARN, "Unable to automount %s/%s: %d",

module/zfs/zfs_dir.c

@@ -593,7 +593,7 @@ zfs_purgedir(znode_t *dzp)
 		if (error)
 			skipped += 1;
 		dmu_tx_commit(tx);
-
+		set_nlink(ZTOI(xzp), xzp->z_links);
 		zfs_iput_async(ZTOI(xzp));
 	}
 	zap_cursor_fini(&zc);
@@ -694,6 +694,7 @@ zfs_rmnode(znode_t *zp)
 		mutex_enter(&xzp->z_lock);
 		xzp->z_unlinked = B_TRUE;	/* mark xzp for deletion */
 		xzp->z_links = 0;	/* no more links to it */
+		set_nlink(ZTOI(xzp), 0); /* this will let iput purge us */
 		VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zsb),
 		    &xzp->z_links, sizeof (xzp->z_links), tx));
 		mutex_exit(&xzp->z_lock);
@@ -759,7 +760,7 @@ zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL,
 		    ctime, sizeof (ctime));
 		zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
-		    ctime, B_TRUE);
+		    ctime);
 	}
 	error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
 	ASSERT(error == 0);
@@ -780,7 +781,7 @@ zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag)
 	    ctime, sizeof (ctime));
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
 	    &dzp->z_pflags, sizeof (dzp->z_pflags));
-	zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
+	zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime);
 	error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
 	ASSERT(error == 0);
 	mutex_exit(&dzp->z_lock);
@@ -875,8 +876,8 @@ zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
 			    NULL, &ctime, sizeof (ctime));
 			SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
 			    NULL, &zp->z_pflags, sizeof (zp->z_pflags));
-			zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
-			    B_TRUE);
+			zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
+			    ctime);
 		}
 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb),
 		    NULL, &zp->z_links, sizeof (zp->z_links));
@@ -903,7 +904,7 @@ zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
 	    NULL, mtime, sizeof (mtime));
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
 	    NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
-	zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
+	zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime);
 	error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
 	ASSERT(error == 0);
 	mutex_exit(&dzp->z_lock);

module/zfs/zfs_fm.c

@@ -457,7 +457,8 @@ update_histogram(uint64_t value_arg, uint16_t *hist, uint32_t *count)
 	/* We store the bits in big-endian (largest-first) order */
 	for (i = 0; i < 64; i++) {
 		if (value & (1ull << i)) {
-			hist[63 - i]++;
+			if (hist[63 - i] < UINT16_MAX)
+				hist[63 - i]++;
 			++bits;
 		}
 	}
@@ -615,7 +616,6 @@ annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info,
 	if (badbuf == NULL || goodbuf == NULL)
 		return (eip);
 
-	ASSERT3U(nui64s, <=, UINT16_MAX);
 	ASSERT3U(size, ==, nui64s * sizeof (uint64_t));
 	ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
 	ASSERT3U(size, <=, UINT32_MAX);

module/zfs/zfs_ioctl.c

@@ -1379,9 +1379,9 @@ get_zfs_sb(const char *dsname, zfs_sb_t **zsbp)
 
 	mutex_enter(&os->os_user_ptr_lock);
 	*zsbp = dmu_objset_get_user(os);
-	if (*zsbp && (*zsbp)->z_sb) {
-		atomic_inc(&((*zsbp)->z_sb->s_active));
-	} else {
+	/* bump s_active only when non-zero to prevent umount race */
+	if (*zsbp == NULL || (*zsbp)->z_sb == NULL ||
+	    !atomic_inc_not_zero(&((*zsbp)->z_sb->s_active))) {
 		error = SET_ERROR(ESRCH);
 	}
 	mutex_exit(&os->os_user_ptr_lock);
@@ -5817,8 +5817,11 @@ zfsdev_ioctl(struct file *filp, unsigned cmd, unsigned long arg)
 	}
 
 
-	if (error == 0 && !(flag & FKIOCTL))
+	if (error == 0 && !(flag & FKIOCTL)) {
+		cookie = spl_fstrans_mark();
 		error = vec->zvec_secpolicy(zc, innvl, CRED());
+		spl_fstrans_unmark(cookie);
+	}
 
 	if (error != 0)
 		goto out;

module/zfs/zfs_rlock.c

@@ -101,9 +101,9 @@
  * Check if a write lock can be grabbed, or wait and recheck until available.
  */
 static void
-zfs_range_lock_writer(znode_t *zp, rl_t *new)
+zfs_range_lock_writer(zfs_rlock_t *zrl, rl_t *new)
 {
-	avl_tree_t *tree = &zp->z_range_avl;
+	avl_tree_t *tree = &zrl->zr_avl;
 	rl_t *rl;
 	avl_index_t where;
 	uint64_t end_size;
@@ -112,32 +112,32 @@ zfs_range_lock_writer(znode_t *zp, rl_t *new)
 
 	for (;;) {
 		/*
-		 * Range locking is also used by zvol and uses a
-		 * dummied up znode. However, for zvol, we don't need to
-		 * append or grow blocksize, and besides we don't have
-		 * a "sa" data or zfs_sb_t - so skip that processing.
+		 * Range locking is also used by zvol. However, for zvol, we
+		 * don't need to append or grow blocksize, so skip that
+		 * processing.
 		 *
 		 * Yes, this is ugly, and would be solved by not handling
 		 * grow or append in range lock code. If that was done then
 		 * we could make the range locking code generically available
 		 * to other non-zfs consumers.
 		 */
-		if (!zp->z_is_zvol) { /* caller is ZPL */
+		if (zrl->zr_size) { /* caller is ZPL */
 			/*
 			 * If in append mode pick up the current end of file.
 			 * This is done under z_range_lock to avoid races.
 			 */
 			if (new->r_type == RL_APPEND)
-				new->r_off = zp->z_size;
+				new->r_off = *zrl->zr_size;
 
 			/*
 			 * If we need to grow the block size then grab the whole
 			 * file range. This is also done under z_range_lock to
 			 * avoid races.
 			 */
-			end_size = MAX(zp->z_size, new->r_off + len);
-			if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) ||
-			    zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) {
+			end_size = MAX(*zrl->zr_size, new->r_off + len);
+			if (end_size > *zrl->zr_blksz &&
+			    (!ISP2(*zrl->zr_blksz) ||
+			    *zrl->zr_blksz < *zrl->zr_max_blksz)) {
 				new->r_off = 0;
 				new->r_len = UINT64_MAX;
 			}
@@ -175,7 +175,7 @@ wait:
 			cv_init(&rl->r_wr_cv, NULL, CV_DEFAULT, NULL);
 			rl->r_write_wanted = B_TRUE;
 		}
-		cv_wait(&rl->r_wr_cv, &zp->z_range_lock);
+		cv_wait(&rl->r_wr_cv, &zrl->zr_mutex);
 
 		/* reset to original */
 		new->r_off = off;
@@ -353,9 +353,9 @@ zfs_range_add_reader(avl_tree_t *tree, rl_t *new, rl_t *prev, avl_index_t where)
  * Check if a reader lock can be grabbed, or wait and recheck until available.
  */
 static void
-zfs_range_lock_reader(znode_t *zp, rl_t *new)
+zfs_range_lock_reader(zfs_rlock_t *zrl, rl_t *new)
 {
-	avl_tree_t *tree = &zp->z_range_avl;
+	avl_tree_t *tree = &zrl->zr_avl;
 	rl_t *prev, *next;
 	avl_index_t where;
 	uint64_t off = new->r_off;
@@ -378,7 +378,7 @@ retry:
 				cv_init(&prev->r_rd_cv, NULL, CV_DEFAULT, NULL);
 				prev->r_read_wanted = B_TRUE;
 			}
-			cv_wait(&prev->r_rd_cv, &zp->z_range_lock);
+			cv_wait(&prev->r_rd_cv, &zrl->zr_mutex);
 			goto retry;
 		}
 		if (off + len < prev->r_off + prev->r_len)
@@ -401,7 +401,7 @@ retry:
 				cv_init(&next->r_rd_cv, NULL, CV_DEFAULT, NULL);
 				next->r_read_wanted = B_TRUE;
 			}
-			cv_wait(&next->r_rd_cv, &zp->z_range_lock);
+			cv_wait(&next->r_rd_cv, &zrl->zr_mutex);
 			goto retry;
 		}
 		if (off + len <= next->r_off + next->r_len)
@@ -423,14 +423,14 @@ got_lock:
  * previously locked as RL_WRITER).
  */
 rl_t *
-zfs_range_lock(znode_t *zp, uint64_t off, uint64_t len, rl_type_t type)
+zfs_range_lock(zfs_rlock_t *zrl, uint64_t off, uint64_t len, rl_type_t type)
 {
 	rl_t *new;
 
 	ASSERT(type == RL_READER || type == RL_WRITER || type == RL_APPEND);
 
 	new = kmem_alloc(sizeof (rl_t), KM_SLEEP);
-	new->r_zp = zp;
+	new->r_zrl = zrl;
 	new->r_off = off;
 	if (len + off < off)	/* overflow */
 		len = UINT64_MAX - off;
@@ -441,18 +441,18 @@ zfs_range_lock(znode_t *zp, uint64_t off, uint64_t len, rl_type_t type)
 	new->r_write_wanted = B_FALSE;
 	new->r_read_wanted = B_FALSE;
 
-	mutex_enter(&zp->z_range_lock);
+	mutex_enter(&zrl->zr_mutex);
 	if (type == RL_READER) {
 		/*
 		 * First check for the usual case of no locks
 		 */
-		if (avl_numnodes(&zp->z_range_avl) == 0)
-			avl_add(&zp->z_range_avl, new);
+		if (avl_numnodes(&zrl->zr_avl) == 0)
+			avl_add(&zrl->zr_avl, new);
 		else
-			zfs_range_lock_reader(zp, new);
-	} else
-		zfs_range_lock_writer(zp, new); /* RL_WRITER or RL_APPEND */
-	mutex_exit(&zp->z_range_lock);
+			zfs_range_lock_reader(zrl, new);
+	} else /* RL_WRITER or RL_APPEND */
+		zfs_range_lock_writer(zrl, new);
+	mutex_exit(&zrl->zr_mutex);
 	return (new);
 }
 
@@ -474,9 +474,9 @@ zfs_range_free(void *arg)
  * Unlock a reader lock
  */
 static void
-zfs_range_unlock_reader(znode_t *zp, rl_t *remove, list_t *free_list)
+zfs_range_unlock_reader(zfs_rlock_t *zrl, rl_t *remove, list_t *free_list)
 {
-	avl_tree_t *tree = &zp->z_range_avl;
+	avl_tree_t *tree = &zrl->zr_avl;
 	rl_t *rl, *next = NULL;
 	uint64_t len;
 
@@ -543,7 +543,7 @@ zfs_range_unlock_reader(znode_t *zp, rl_t *remove, list_t *free_list)
 void
 zfs_range_unlock(rl_t *rl)
 {
-	znode_t *zp = rl->r_zp;
+	zfs_rlock_t *zrl = rl->r_zrl;
 	list_t free_list;
 	rl_t *free_rl;
 
@@ -552,10 +552,10 @@ zfs_range_unlock(rl_t *rl)
 	ASSERT(!rl->r_proxy);
 	list_create(&free_list, sizeof (rl_t), offsetof(rl_t, rl_node));
 
-	mutex_enter(&zp->z_range_lock);
+	mutex_enter(&zrl->zr_mutex);
 	if (rl->r_type == RL_WRITER) {
 		/* writer locks can't be shared or split */
-		avl_remove(&zp->z_range_avl, rl);
+		avl_remove(&zrl->zr_avl, rl);
 		if (rl->r_write_wanted)
 			cv_broadcast(&rl->r_wr_cv);
 
@@ -568,9 +568,9 @@ zfs_range_unlock(rl_t *rl)
 		 * lock may be shared, let zfs_range_unlock_reader()
 		 * release the zp->z_range_lock lock and free the rl_t
 		 */
-		zfs_range_unlock_reader(zp, rl, &free_list);
+		zfs_range_unlock_reader(zrl, rl, &free_list);
 	}
-	mutex_exit(&zp->z_range_lock);
+	mutex_exit(&zrl->zr_mutex);
 
 	while ((free_rl = list_head(&free_list)) != NULL) {
 		list_remove(&free_list, free_rl);
@@ -588,17 +588,17 @@ zfs_range_unlock(rl_t *rl)
 void
 zfs_range_reduce(rl_t *rl, uint64_t off, uint64_t len)
 {
-	znode_t *zp = rl->r_zp;
+	zfs_rlock_t *zrl = rl->r_zrl;
 
 	/* Ensure there are no other locks */
-	ASSERT(avl_numnodes(&zp->z_range_avl) == 1);
+	ASSERT(avl_numnodes(&zrl->zr_avl) == 1);
 	ASSERT(rl->r_off == 0);
 	ASSERT(rl->r_type == RL_WRITER);
 	ASSERT(!rl->r_proxy);
 	ASSERT3U(rl->r_len, ==, UINT64_MAX);
 	ASSERT3U(rl->r_cnt, ==, 1);
 
-	mutex_enter(&zp->z_range_lock);
+	mutex_enter(&zrl->zr_mutex);
 	rl->r_off = off;
 	rl->r_len = len;
 
@@ -607,7 +607,7 @@ zfs_range_reduce(rl_t *rl, uint64_t off, uint64_t len)
 	if (rl->r_read_wanted)
 		cv_broadcast(&rl->r_rd_cv);
 
-	mutex_exit(&zp->z_range_lock);
+	mutex_exit(&zrl->zr_mutex);
 }
 
 /*
@@ -626,3 +626,10 @@ zfs_range_compare(const void *arg1, const void *arg2)
 		return (-1);
 	return (0);
 }
+
+#ifdef _KERNEL
+EXPORT_SYMBOL(zfs_range_lock);
+EXPORT_SYMBOL(zfs_range_unlock);
+EXPORT_SYMBOL(zfs_range_reduce);
+EXPORT_SYMBOL(zfs_range_compare);
+#endif

module/zfs/zfs_sa.c

@@ -277,7 +277,7 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
 	sa_bulk_attr_t *bulk, *sa_attrs;
 	zfs_acl_locator_cb_t locate = { 0 };
 	uint64_t uid, gid, mode, rdev, xattr, parent;
-	uint64_t crtime[2], mtime[2], ctime[2];
+	uint64_t crtime[2], mtime[2], ctime[2], atime[2];
 	zfs_acl_phys_t znode_acl;
 	char scanstamp[AV_SCANSTAMP_SZ];
 	boolean_t drop_lock = B_FALSE;
@@ -309,6 +309,7 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
 
 	/* First do a bulk query of the attributes that aren't cached */
 	bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, &atime, 16);
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zsb), NULL, &crtime, 16);
@@ -344,7 +345,7 @@ zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
 	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zsb), NULL,
 	    &zp->z_pflags, 8);
 	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zsb), NULL,
-	    zp->z_atime, 16);
+	    &atime, 16);
 	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zsb), NULL,
 	    &mtime, 16);
 	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zsb), NULL,

module/zfs/zfs_vfsops.c

@@ -699,20 +699,18 @@ zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp)
 
 	zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP);
 
+	/*
+	 * Optional temporary mount options, free'd in zfs_sb_free().
+	 */
+	zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc());
+
 	/*
 	 * We claim to always be readonly so we can open snapshots;
 	 * other ZPL code will prevent us from writing to snapshots.
 	 */
 	error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os);
-	if (error) {
-		kmem_free(zsb, sizeof (zfs_sb_t));
-		return (error);
-	}
-
-	/*
-	 * Optional temporary mount options, free'd in zfs_sb_free().
-	 */
-	zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc());
+	if (error)
+		goto out_zmo;
 
 	/*
 	 * Initialize the zfs-specific filesystem structure.
@@ -840,8 +838,9 @@ zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp)
 
 out:
 	dmu_objset_disown(os, zsb);
+out_zmo:
 	*zsbp = NULL;
-
+	zfs_mntopts_free(zsb->z_mntopts);
 	kmem_free(zsb, sizeof (zfs_sb_t));
 	return (error);
 }
@@ -1124,8 +1123,7 @@ zfs_root(zfs_sb_t *zsb, struct inode **ipp)
 }
 EXPORT_SYMBOL(zfs_root);
 
-#if !defined(HAVE_SPLIT_SHRINKER_CALLBACK) && !defined(HAVE_SHRINK) && \
-	defined(HAVE_D_PRUNE_ALIASES)
+#ifdef HAVE_D_PRUNE_ALIASES
 /*
  * Linux kernels older than 3.1 do not support a per-filesystem shrinker.
  * To accommodate this we must improvise and manually walk the list of znodes
@@ -1215,15 +1213,29 @@ zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
 	} else {
 			*objects = (*shrinker->scan_objects)(shrinker, &sc);
 	}
+
 #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK)
 	*objects = (*shrinker->scan_objects)(shrinker, &sc);
 #elif defined(HAVE_SHRINK)
 	*objects = (*shrinker->shrink)(shrinker, &sc);
 #elif defined(HAVE_D_PRUNE_ALIASES)
+#define	D_PRUNE_ALIASES_IS_DEFAULT
 	*objects = zfs_sb_prune_aliases(zsb, nr_to_scan);
 #else
 #error "No available dentry and inode cache pruning mechanism."
 #endif
+
+#if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT)
+#undef	D_PRUNE_ALIASES_IS_DEFAULT
+	/*
+	 * Fall back to zfs_sb_prune_aliases if the kernel's per-superblock
+	 * shrinker couldn't free anything, possibly due to the inodes being
+	 * allocated in a different memcg.
+	 */
+	if (*objects == 0)
+		*objects = zfs_sb_prune_aliases(zsb, nr_to_scan);
+#endif
+
 	ZFS_EXIT(zsb);
 
 	dprintf_ds(zsb->z_os->os_dsl_dataset,
@@ -1391,13 +1403,13 @@ zfs_domount(struct super_block *sb, zfs_mntopts_t *zmo, int silent)
 	sb->s_time_gran = 1;
 	sb->s_blocksize = recordsize;
 	sb->s_blocksize_bits = ilog2(recordsize);
-	zsb->z_bdi.ra_pages = 0;
-	sb->s_bdi = &zsb->z_bdi;
 
-	error = -zpl_bdi_setup_and_register(&zsb->z_bdi, "zfs");
+	error = -zpl_bdi_setup(sb, "zfs");
 	if (error)
 		goto out;
 
+	sb->s_bdi->ra_pages = 0;
+
 	/* Set callback operations for the file system. */
 	sb->s_op = &zpl_super_operations;
 	sb->s_xattr = zpl_xattr_handlers;
@@ -1493,7 +1505,7 @@ zfs_umount(struct super_block *sb)
 	arc_remove_prune_callback(zsb->z_arc_prune);
 	VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0);
 	os = zsb->z_os;
-	bdi_destroy(sb->s_bdi);
+	zpl_bdi_destroy(sb);
 
 	/*
 	 * z_os will be NULL if there was an error in
@@ -1601,6 +1613,14 @@ zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp)
 		ZFS_EXIT(zsb);
 		return (err);
 	}
+
+	/* Don't export xattr stuff */
+	if (zp->z_pflags & ZFS_XATTR) {
+		iput(ZTOI(zp));
+		ZFS_EXIT(zsb);
+		return (SET_ERROR(ENOENT));
+	}
+
 	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen,
 	    sizeof (uint64_t));
 	zp_gen = zp_gen & gen_mask;
@@ -1613,7 +1633,7 @@ zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp)
 		    fid_gen);
 		iput(ZTOI(zp));
 		ZFS_EXIT(zsb);
-		return (SET_ERROR(EINVAL));
+		return (SET_ERROR(ENOENT));
 	}
 
 	*ipp = ZTOI(zp);
@@ -1858,7 +1878,10 @@ zfs_init(void)
 void
 zfs_fini(void)
 {
-	taskq_wait_outstanding(system_taskq, 0);
+	/*
+	 * we don't use outstanding because zpl_posix_acl_free might add more.
+	 */
+	taskq_wait(system_taskq);
 	unregister_filesystem(&zpl_fs_type);
 	zfs_znode_fini();
 	zfsctl_fini();

module/zfs/zfs_vnops.c

@@ -332,11 +332,11 @@ update_pages(struct inode *ip, int64_t start, int len,
 	int64_t	off;
 	void *pb;
 
-	off = start & (PAGE_CACHE_SIZE-1);
-	for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) {
-		nbytes = MIN(PAGE_CACHE_SIZE - off, len);
+	off = start & (PAGE_SIZE-1);
+	for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) {
+		nbytes = MIN(PAGE_SIZE - off, len);
 
-		pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT);
+		pp = find_lock_page(mp, start >> PAGE_SHIFT);
 		if (pp) {
 			if (mapping_writably_mapped(mp))
 				flush_dcache_page(pp);
@@ -353,7 +353,7 @@ update_pages(struct inode *ip, int64_t start, int len,
 			SetPageUptodate(pp);
 			ClearPageError(pp);
 			unlock_page(pp);
-			page_cache_release(pp);
+			put_page(pp);
 		}
 
 		len -= nbytes;
@@ -384,11 +384,11 @@ mappedread(struct inode *ip, int nbytes, uio_t *uio)
 	void *pb;
 
 	start = uio->uio_loffset;
-	off = start & (PAGE_CACHE_SIZE-1);
-	for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) {
-		bytes = MIN(PAGE_CACHE_SIZE - off, len);
+	off = start & (PAGE_SIZE-1);
+	for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) {
+		bytes = MIN(PAGE_SIZE - off, len);
 
-		pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT);
+		pp = find_lock_page(mp, start >> PAGE_SHIFT);
 		if (pp) {
 			ASSERT(PageUptodate(pp));
 
@@ -401,7 +401,7 @@ mappedread(struct inode *ip, int nbytes, uio_t *uio)
 
 			mark_page_accessed(pp);
 			unlock_page(pp);
-			page_cache_release(pp);
+			put_page(pp);
 		} else {
 			error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
 			    uio, bytes);
@@ -481,7 +481,8 @@ zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 	/*
 	 * Lock the range against changes.
 	 */
-	rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
+	rl = zfs_range_lock(&zp->z_range_lock, uio->uio_loffset, uio->uio_resid,
+	    RL_READER);
 
 	/*
 	 * If we are reading past end-of-file we can skip
@@ -549,7 +550,6 @@ zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 out:
 	zfs_range_unlock(rl);
 
-	ZFS_ACCESSTIME_STAMP(zsb, zp);
 	ZFS_EXIT(zsb);
 	return (error);
 }
@@ -663,7 +663,7 @@ zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 		 * Obtain an appending range lock to guarantee file append
 		 * semantics.  We reset the write offset once we have the lock.
 		 */
-		rl = zfs_range_lock(zp, 0, n, RL_APPEND);
+		rl = zfs_range_lock(&zp->z_range_lock, 0, n, RL_APPEND);
 		woff = rl->r_off;
 		if (rl->r_len == UINT64_MAX) {
 			/*
@@ -680,7 +680,7 @@ zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 		 * this write, then this range lock will lock the entire file
 		 * so that we can re-write the block safely.
 		 */
-		rl = zfs_range_lock(zp, woff, n, RL_WRITER);
+		rl = zfs_range_lock(&zp->z_range_lock, woff, n, RL_WRITER);
 	}
 
 	if (woff >= limit) {
@@ -864,8 +864,7 @@ zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 		}
 		mutex_exit(&zp->z_acl_lock);
 
-		zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime,
-		    B_TRUE);
+		zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
 
 		/*
 		 * Update the file size (zp_size) if it has changed;
@@ -1007,7 +1006,8 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 	 * we don't have to write the data twice.
 	 */
 	if (buf != NULL) { /* immediate write */
-		zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER);
+		zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, size,
+		    RL_READER);
 		/* test for truncation needs to be done while range locked */
 		if (offset >= zp->z_size) {
 			error = SET_ERROR(ENOENT);
@@ -1028,8 +1028,8 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 			size = zp->z_blksz;
 			blkoff = ISP2(size) ? P2PHASE(offset, size) : offset;
 			offset -= blkoff;
-			zgd->zgd_rl = zfs_range_lock(zp, offset, size,
-			    RL_READER);
+			zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset,
+			    size, RL_READER);
 			if (zp->z_blksz == size)
 				break;
 			offset += blkoff;
@@ -1602,13 +1602,13 @@ top:
 	error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT);
 	if (error) {
 		zfs_dirent_unlock(dl);
-		iput(ip);
-		if (xzp)
-			iput(ZTOI(xzp));
 		if (error == ERESTART) {
 			waited = B_TRUE;
 			dmu_tx_wait(tx);
 			dmu_tx_abort(tx);
+			iput(ip);
+			if (xzp)
+				iput(ZTOI(xzp));
 			goto top;
 		}
 #ifdef HAVE_PN_UTILS
@@ -1616,6 +1616,9 @@ top:
 			pn_free(realnmp);
 #endif /* HAVE_PN_UTILS */
 		dmu_tx_abort(tx);
+		iput(ip);
+		if (xzp)
+			iput(ZTOI(xzp));
 		ZFS_EXIT(zsb);
 		return (error);
 	}
@@ -1944,14 +1947,15 @@ top:
 		rw_exit(&zp->z_parent_lock);
 		rw_exit(&zp->z_name_lock);
 		zfs_dirent_unlock(dl);
-		iput(ip);
 		if (error == ERESTART) {
 			waited = B_TRUE;
 			dmu_tx_wait(tx);
 			dmu_tx_abort(tx);
+			iput(ip);
 			goto top;
 		}
 		dmu_tx_abort(tx);
+		iput(ip);
 		ZFS_EXIT(zsb);
 		return (error);
 	}
@@ -2138,9 +2142,6 @@ update:
 	zap_cursor_fini(&zc);
 	if (error == ENOENT)
 		error = 0;
-
-	ZFS_ACCESSTIME_STAMP(zsb, zp);
-
 out:
 	ZFS_EXIT(zsb);
 
@@ -2193,11 +2194,11 @@ zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr)
 	zfs_sb_t *zsb = ITOZSB(ip);
 	int	error = 0;
 	uint64_t links;
-	uint64_t mtime[2], ctime[2];
+	uint64_t atime[2], mtime[2], ctime[2];
 	xvattr_t *xvap = (xvattr_t *)vap;	/* vap may be an xvattr_t * */
 	xoptattr_t *xoap = NULL;
 	boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
-	sa_bulk_attr_t bulk[2];
+	sa_bulk_attr_t bulk[3];
 	int count = 0;
 
 	ZFS_ENTER(zsb);
@@ -2205,6 +2206,7 @@ zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr)
 
 	zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);
 
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, &atime, 16);
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
 
@@ -2353,7 +2355,7 @@ zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr)
 		}
 	}
 
-	ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime);
+	ZFS_TIME_DECODE(&vap->va_atime, atime);
 	ZFS_TIME_DECODE(&vap->va_mtime, mtime);
 	ZFS_TIME_DECODE(&vap->va_ctime, ctime);
 
@@ -2400,7 +2402,6 @@ zfs_getattr_fast(struct inode *ip, struct kstat *sp)
 	mutex_enter(&zp->z_lock);
 
 	generic_fillattr(ip, sp);
-	ZFS_TIME_DECODE(&sp->atime, zp->z_atime);
 
 	sa_object_size(zp->z_sa_hdl, &blksize, &nblocks);
 	sp->blksize = blksize;
@@ -2464,7 +2465,7 @@ zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr)
 	uint64_t	new_mode;
 	uint64_t	new_uid, new_gid;
 	uint64_t	xattr_obj;
-	uint64_t	mtime[2], ctime[2];
+	uint64_t	mtime[2], ctime[2], atime[2];
 	znode_t		*attrzp;
 	int		need_policy = FALSE;
 	int		err, err2;
@@ -2937,10 +2938,11 @@ top:
 	}
 
 
-	if (mask & ATTR_ATIME) {
-		ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime);
+	if ((mask & ATTR_ATIME) || zp->z_atime_dirty) {
+		zp->z_atime_dirty = 0;
+		ZFS_TIME_ENCODE(&ip->i_atime, atime);
 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
-		    &zp->z_atime, sizeof (zp->z_atime));
+		    &atime, sizeof (atime));
 	}
 
 	if (mask & ATTR_MTIME) {
@@ -2955,19 +2957,17 @@ top:
 		    NULL, mtime, sizeof (mtime));
 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL,
 		    &ctime, sizeof (ctime));
-		zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime,
-		    B_TRUE);
+		zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
 	} else if (mask != 0) {
 		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL,
 		    &ctime, sizeof (ctime));
-		zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
-		    B_TRUE);
+		zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime);
 		if (attrzp) {
 			SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
 			    SA_ZPL_CTIME(zsb), NULL,
 			    &ctime, sizeof (ctime));
 			zfs_tstamp_update_setup(attrzp, STATE_CHANGED,
-			    mtime, ctime, B_TRUE);
+			    mtime, ctime);
 		}
 	}
 	/*
@@ -3029,8 +3029,6 @@ out:
 		ASSERT(err2 == 0);
 	}
 
-	if (attrzp)
-		iput(ZTOI(attrzp));
 	if (aclp)
 		zfs_acl_free(aclp);
 
@@ -3041,11 +3039,15 @@ out:
 
 	if (err) {
 		dmu_tx_abort(tx);
+		if (attrzp)
+			iput(ZTOI(attrzp));
 		if (err == ERESTART)
 			goto top;
 	} else {
 		err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
 		dmu_tx_commit(tx);
+		if (attrzp)
+			iput(ZTOI(attrzp));
 		zfs_inode_update(zp);
 	}
 
@@ -3078,7 +3080,7 @@ zfs_rename_unlock(zfs_zlock_t **zlpp)
 
 	while ((zl = *zlpp) != NULL) {
 		if (zl->zl_znode != NULL)
-			iput(ZTOI(zl->zl_znode));
+			zfs_iput_async(ZTOI(zl->zl_znode));
 		rw_exit(zl->zl_rwlock);
 		*zlpp = zl->zl_next;
 		kmem_free(zl, sizeof (*zl));
@@ -3415,16 +3417,19 @@ top:
 		if (sdzp == tdzp)
 			rw_exit(&sdzp->z_name_lock);
 
-		iput(ZTOI(szp));
-		if (tzp)
-			iput(ZTOI(tzp));
 		if (error == ERESTART) {
 			waited = B_TRUE;
 			dmu_tx_wait(tx);
 			dmu_tx_abort(tx);
+			iput(ZTOI(szp));
+			if (tzp)
+				iput(ZTOI(tzp));
 			goto top;
 		}
 		dmu_tx_abort(tx);
+		iput(ZTOI(szp));
+		if (tzp)
+			iput(ZTOI(tzp));
 		ZFS_EXIT(zsb);
 		return (error);
 	}
@@ -3688,7 +3693,6 @@ zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr)
 		error = zfs_sa_readlink(zp, uio);
 	mutex_exit(&zp->z_lock);
 
-	ZFS_ACCESSTIME_STAMP(zsb, zp);
 	ZFS_EXIT(zsb);
 	return (error);
 }
@@ -3894,8 +3898,8 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
 
 	pgoff = page_offset(pp);	/* Page byte-offset in file */
 	offset = i_size_read(ip);	/* File length in bytes */
-	pglen = MIN(PAGE_CACHE_SIZE,	/* Page length in bytes */
-	    P2ROUNDUP(offset, PAGE_CACHE_SIZE)-pgoff);
+	pglen = MIN(PAGE_SIZE,		/* Page length in bytes */
+	    P2ROUNDUP(offset, PAGE_SIZE)-pgoff);
 
 	/* Page is beyond end of file */
 	if (pgoff >= offset) {
@@ -3947,7 +3951,7 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
 	redirty_page_for_writepage(wbc, pp);
 	unlock_page(pp);
 
-	rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, pgoff, pglen, RL_WRITER);
 	lock_page(pp);
 
 	/* Page mapping changed or it was no longer dirty, we're done */
@@ -4006,7 +4010,7 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
 	}
 
 	va = kmap(pp);
-	ASSERT3U(pglen, <=, PAGE_CACHE_SIZE);
+	ASSERT3U(pglen, <=, PAGE_SIZE);
 	dmu_write(zsb->z_os, zp->z_id, pgoff, pglen, va, tx);
 	kunmap(pp);
 
@@ -4054,7 +4058,7 @@ zfs_dirty_inode(struct inode *ip, int flags)
 	dmu_tx_t	*tx;
 	uint64_t	mode, atime[2], mtime[2], ctime[2];
 	sa_bulk_attr_t	bulk[4];
-	int		error;
+	int		error = 0;
 	int		cnt = 0;
 
 	if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os))
@@ -4063,6 +4067,20 @@ zfs_dirty_inode(struct inode *ip, int flags)
 	ZFS_ENTER(zsb);
 	ZFS_VERIFY_ZP(zp);
 
+#ifdef I_DIRTY_TIME
+	/*
+	 * This is the lazytime semantic indroduced in Linux 4.0
+	 * This flag will only be called from update_time when lazytime is set.
+	 * (Note, I_DIRTY_SYNC will also set if not lazytime)
+	 * Fortunately mtime and ctime are managed within ZFS itself, so we
+	 * only need to dirty atime.
+	 */
+	if (flags == I_DIRTY_TIME) {
+		zp->z_atime_dirty = 1;
+		goto out;
+	}
+#endif
+
 	tx = dmu_tx_create(zsb->z_os);
 
 	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
@@ -4075,6 +4093,8 @@ zfs_dirty_inode(struct inode *ip, int flags)
 	}
 
 	mutex_enter(&zp->z_lock);
+	zp->z_atime_dirty = 0;
+
 	SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zsb), NULL, &mode, 8);
 	SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zsb), NULL, &atime, 16);
 	SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
@@ -4087,7 +4107,6 @@ zfs_dirty_inode(struct inode *ip, int flags)
 	mode = ip->i_mode;
 
 	zp->z_mode = mode;
-	zp->z_atime_dirty = 0;
 
 	error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx);
 	mutex_exit(&zp->z_lock);
@@ -4105,6 +4124,7 @@ zfs_inactive(struct inode *ip)
 {
 	znode_t	*zp = ITOZ(ip);
 	zfs_sb_t *zsb = ITOZSB(ip);
+	uint64_t atime[2];
 	int error;
 	int need_unlock = 0;
 
@@ -4128,9 +4148,10 @@ zfs_inactive(struct inode *ip)
 		if (error) {
 			dmu_tx_abort(tx);
 		} else {
+			ZFS_TIME_ENCODE(&ip->i_atime, atime);
 			mutex_enter(&zp->z_lock);
 			(void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zsb),
-			    (void *)&zp->z_atime, sizeof (zp->z_atime), tx);
+			    (void *)&atime, sizeof (atime), tx);
 			zp->z_atime_dirty = 0;
 			mutex_exit(&zp->z_lock);
 			dmu_tx_commit(tx);
@@ -4181,7 +4202,7 @@ zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages)
 	int err;
 
 	os = zsb->z_os;
-	io_len = nr_pages << PAGE_CACHE_SHIFT;
+	io_len = nr_pages << PAGE_SHIFT;
 	i_size = i_size_read(ip);
 	io_off = page_offset(pl[0]);
 
@@ -4192,10 +4213,10 @@ zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages)
 	 * Iterate over list of pages and read each page individually.
 	 */
 	page_idx = 0;
-	cur_pp   = pl[0];
 	for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) {
 		caddr_t va;
 
+		cur_pp = pl[page_idx++];
 		va = kmap(cur_pp);
 		err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va,
 		    DMU_READ_PREFETCH);
@@ -4206,7 +4227,6 @@ zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages)
 				err = SET_ERROR(EIO);
 			return (err);
 		}
-		cur_pp = pl[++page_idx];
 	}
 
 	return (0);
@@ -4240,9 +4260,6 @@ zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages)
 
 	err = zfs_fillpage(ip, pl, nr_pages);
 
-	if (!err)
-		ZFS_ACCESSTIME_STAMP(zsb, zp);
-
 	ZFS_EXIT(zsb);
 	return (err);
 }

module/zfs/zfs_znode.c

@@ -113,14 +113,11 @@ zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
 	mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
 	rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL);
 
-	mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
-	avl_create(&zp->z_range_avl, zfs_range_compare,
-	    sizeof (rl_t), offsetof(rl_t, r_node));
+	zfs_rlock_init(&zp->z_range_lock);
 
 	zp->z_dirlocks = NULL;
 	zp->z_acl_cached = NULL;
 	zp->z_xattr_cached = NULL;
-	zp->z_xattr_parent = NULL;
 	zp->z_moved = 0;
 	return (0);
 }
@@ -137,13 +134,11 @@ zfs_znode_cache_destructor(void *buf, void *arg)
 	rw_destroy(&zp->z_name_lock);
 	mutex_destroy(&zp->z_acl_lock);
 	rw_destroy(&zp->z_xattr_lock);
-	avl_destroy(&zp->z_range_avl);
-	mutex_destroy(&zp->z_range_lock);
+	zfs_rlock_destroy(&zp->z_range_lock);
 
 	ASSERT(zp->z_dirlocks == NULL);
 	ASSERT(zp->z_acl_cached == NULL);
 	ASSERT(zp->z_xattr_cached == NULL);
-	ASSERT(zp->z_xattr_parent == NULL);
 }
 
 static int
@@ -435,11 +430,6 @@ zfs_inode_destroy(struct inode *ip)
 		zp->z_xattr_cached = NULL;
 	}
 
-	if (zp->z_xattr_parent) {
-		zfs_iput_async(ZTOI(zp->z_xattr_parent));
-		zp->z_xattr_parent = NULL;
-	}
-
 	kmem_cache_free(znode_cache, zp);
 }
 
@@ -492,110 +482,6 @@ zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip)
 	}
 }
 
-/*
- * Construct a znode+inode and initialize.
- *
- * This does not do a call to dmu_set_user() that is
- * up to the caller to do, in case you don't want to
- * return the znode
- */
-static znode_t *
-zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
-    dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl,
-    struct inode *dip)
-{
-	znode_t	*zp;
-	struct inode *ip;
-	uint64_t mode;
-	uint64_t parent;
-	sa_bulk_attr_t bulk[9];
-	int count = 0;
-
-	ASSERT(zsb != NULL);
-
-	ip = new_inode(zsb->z_sb);
-	if (ip == NULL)
-		return (NULL);
-
-	zp = ITOZ(ip);
-	ASSERT(zp->z_dirlocks == NULL);
-	ASSERT3P(zp->z_acl_cached, ==, NULL);
-	ASSERT3P(zp->z_xattr_cached, ==, NULL);
-	ASSERT3P(zp->z_xattr_parent, ==, NULL);
-	zp->z_moved = 0;
-	zp->z_sa_hdl = NULL;
-	zp->z_unlinked = 0;
-	zp->z_atime_dirty = 0;
-	zp->z_mapcnt = 0;
-	zp->z_id = db->db_object;
-	zp->z_blksz = blksz;
-	zp->z_seq = 0x7A4653;
-	zp->z_sync_cnt = 0;
-	zp->z_is_zvol = B_FALSE;
-	zp->z_is_mapped = B_FALSE;
-	zp->z_is_ctldir = B_FALSE;
-	zp->z_is_stale = B_FALSE;
-
-	zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
-
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
-	    &zp->z_pflags, 8);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
-	    &parent, 8);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
-	    &zp->z_atime, 16);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
-
-	if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
-		if (hdl == NULL)
-			sa_handle_destroy(zp->z_sa_hdl);
-		zp->z_sa_hdl = NULL;
-		goto error;
-	}
-
-	zp->z_mode = mode;
-
-	/*
-	 * xattr znodes hold a reference on their unique parent
-	 */
-	if (dip && zp->z_pflags & ZFS_XATTR) {
-		igrab(dip);
-		zp->z_xattr_parent = ITOZ(dip);
-	}
-
-	ip->i_ino = obj;
-	zfs_inode_update(zp);
-	zfs_inode_set_ops(zsb, ip);
-
-	/*
-	 * The only way insert_inode_locked() can fail is if the ip->i_ino
-	 * number is already hashed for this super block.  This can never
-	 * happen because the inode numbers map 1:1 with the object numbers.
-	 *
-	 * The one exception is rolling back a mounted file system, but in
-	 * this case all the active inode are unhashed during the rollback.
-	 */
-	VERIFY3S(insert_inode_locked(ip), ==, 0);
-
-	mutex_enter(&zsb->z_znodes_lock);
-	list_insert_tail(&zsb->z_all_znodes, zp);
-	zsb->z_nr_znodes++;
-	membar_producer();
-	mutex_exit(&zsb->z_znodes_lock);
-
-	unlock_new_inode(ip);
-	return (zp);
-
-error:
-	iput(ip);
-	return (NULL);
-}
-
 void
 zfs_set_inode_flags(znode_t *zp, struct inode *ip)
 {
@@ -622,12 +508,13 @@ zfs_set_inode_flags(znode_t *zp, struct inode *ip)
  * inode has the correct field it should be used, and the ZFS code
  * updated to access the inode.  This can be done incrementally.
  */
-void
-zfs_inode_update(znode_t *zp)
+static void
+zfs_inode_update_impl(znode_t *zp, boolean_t new)
 {
 	zfs_sb_t	*zsb;
 	struct inode	*ip;
 	uint32_t	blksize;
+	u_longlong_t	i_blocks;
 	uint64_t	atime[2], mtime[2], ctime[2];
 
 	ASSERT(zp != NULL);
@@ -642,6 +529,8 @@ zfs_inode_update(znode_t *zp)
 	sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
 	sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
 
+	dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks);
+
 	spin_lock(&ip->i_lock);
 	ip->i_generation = zp->z_gen;
 	ip->i_uid = SUID_TO_KUID(zp->z_uid);
@@ -650,10 +539,14 @@ zfs_inode_update(znode_t *zp)
 	ip->i_mode = zp->z_mode;
 	zfs_set_inode_flags(zp, ip);
 	ip->i_blkbits = SPA_MINBLOCKSHIFT;
-	dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
-	    (u_longlong_t *)&ip->i_blocks);
+	ip->i_blocks = i_blocks;
 
-	ZFS_TIME_DECODE(&ip->i_atime, atime);
+	/*
+	 * Only read atime from SA if we are newly created inode (or rezget),
+	 * otherwise i_atime might be dirty.
+	 */
+	if (new)
+		ZFS_TIME_DECODE(&ip->i_atime, atime);
 	ZFS_TIME_DECODE(&ip->i_mtime, mtime);
 	ZFS_TIME_DECODE(&ip->i_ctime, ctime);
 
@@ -661,6 +554,112 @@ zfs_inode_update(znode_t *zp)
 	spin_unlock(&ip->i_lock);
 }
 
+static void
+zfs_inode_update_new(znode_t *zp)
+{
+	zfs_inode_update_impl(zp, B_TRUE);
+}
+
+void
+zfs_inode_update(znode_t *zp)
+{
+	zfs_inode_update_impl(zp, B_FALSE);
+}
+
+/*
+ * Construct a znode+inode and initialize.
+ *
+ * This does not do a call to dmu_set_user() that is
+ * up to the caller to do, in case you don't want to
+ * return the znode
+ */
+static znode_t *
+zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
+    dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl)
+{
+	znode_t	*zp;
+	struct inode *ip;
+	uint64_t mode;
+	uint64_t parent;
+	sa_bulk_attr_t bulk[8];
+	int count = 0;
+
+	ASSERT(zsb != NULL);
+
+	ip = new_inode(zsb->z_sb);
+	if (ip == NULL)
+		return (NULL);
+
+	zp = ITOZ(ip);
+	ASSERT(zp->z_dirlocks == NULL);
+	ASSERT3P(zp->z_acl_cached, ==, NULL);
+	ASSERT3P(zp->z_xattr_cached, ==, NULL);
+	zp->z_moved = 0;
+	zp->z_sa_hdl = NULL;
+	zp->z_unlinked = 0;
+	zp->z_atime_dirty = 0;
+	zp->z_mapcnt = 0;
+	zp->z_id = db->db_object;
+	zp->z_blksz = blksz;
+	zp->z_seq = 0x7A4653;
+	zp->z_sync_cnt = 0;
+	zp->z_is_mapped = B_FALSE;
+	zp->z_is_ctldir = B_FALSE;
+	zp->z_is_stale = B_FALSE;
+	zp->z_range_lock.zr_size = &zp->z_size;
+	zp->z_range_lock.zr_blksz = &zp->z_blksz;
+	zp->z_range_lock.zr_max_blksz = &ZTOZSB(zp)->z_max_blksz;
+
+	zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
+
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
+	    &zp->z_pflags, 8);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
+	    &parent, 8);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
+	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
+
+	if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
+		if (hdl == NULL)
+			sa_handle_destroy(zp->z_sa_hdl);
+		zp->z_sa_hdl = NULL;
+		goto error;
+	}
+
+	zp->z_mode = mode;
+
+	ip->i_ino = obj;
+	zfs_inode_update_new(zp);
+	zfs_inode_set_ops(zsb, ip);
+
+	/*
+	 * The only way insert_inode_locked() can fail is if the ip->i_ino
+	 * number is already hashed for this super block.  This can never
+	 * happen because the inode numbers map 1:1 with the object numbers.
+	 *
+	 * The one exception is rolling back a mounted file system, but in
+	 * this case all the active inode are unhashed during the rollback.
+	 */
+	VERIFY3S(insert_inode_locked(ip), ==, 0);
+
+	mutex_enter(&zsb->z_znodes_lock);
+	list_insert_tail(&zsb->z_all_znodes, zp);
+	zsb->z_nr_znodes++;
+	membar_producer();
+	mutex_exit(&zsb->z_znodes_lock);
+
+	unlock_new_inode(ip);
+	return (zp);
+
+error:
+	iput(ip);
+	return (NULL);
+}
+
 /*
  * Safely mark an inode dirty.  Inodes which are part of a read-only
  * file system or snapshot may not be dirtied.
@@ -913,8 +912,7 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
 	VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
 
 	if (!(flag & IS_ROOT_NODE)) {
-		*zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl,
-		    ZTOI(dzp));
+		*zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl);
 		VERIFY(*zpp != NULL);
 		VERIFY(dzp != NULL);
 	} else {
@@ -1124,7 +1122,7 @@ again:
 	 * bonus buffer.
 	 */
 	zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
-	    doi.doi_bonus_type, obj_num, NULL, NULL);
+	    doi.doi_bonus_type, obj_num, NULL);
 	if (zp == NULL) {
 		err = SET_ERROR(ENOENT);
 	} else {
@@ -1142,12 +1140,22 @@ zfs_rezget(znode_t *zp)
 	dmu_buf_t *db;
 	uint64_t obj_num = zp->z_id;
 	uint64_t mode;
-	sa_bulk_attr_t bulk[8];
+	sa_bulk_attr_t bulk[7];
 	int err;
 	int count = 0;
 	uint64_t gen;
 	znode_hold_t *zh;
 
+	/*
+	 * skip ctldir, otherwise they will always get invalidated. This will
+	 * cause funny behaviour for the mounted snapdirs. Especially for
+	 * Linux >= 3.18, d_invalidate will detach the mountpoint and prevent
+	 * anyone automount it again as long as someone is still using the
+	 * detached mount.
+	 */
+	if (zp->z_is_ctldir)
+		return (0);
+
 	zh = zfs_znode_hold_enter(zsb, obj_num);
 
 	mutex_enter(&zp->z_acl_lock);
@@ -1162,11 +1170,6 @@ zfs_rezget(znode_t *zp)
 		nvlist_free(zp->z_xattr_cached);
 		zp->z_xattr_cached = NULL;
 	}
-
-	if (zp->z_xattr_parent) {
-		zfs_iput_async(ZTOI(zp->z_xattr_parent));
-		zp->z_xattr_parent = NULL;
-	}
 	rw_exit(&zp->z_xattr_lock);
 
 	ASSERT(zp->z_sa_hdl == NULL);
@@ -1197,8 +1200,6 @@ zfs_rezget(znode_t *zp)
 	    &zp->z_links, sizeof (zp->z_links));
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
 	    &zp->z_pflags, sizeof (zp->z_pflags));
-	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
-	    &zp->z_atime, sizeof (zp->z_atime));
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
 	    &zp->z_uid, sizeof (zp->z_uid));
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
@@ -1222,7 +1223,8 @@ zfs_rezget(znode_t *zp)
 
 	zp->z_unlinked = (zp->z_links == 0);
 	zp->z_blksz = doi.doi_data_block_size;
-	zfs_inode_update(zp);
+	zp->z_atime_dirty = 0;
+	zfs_inode_update_new(zp);
 
 	zfs_znode_hold_exit(zsb, zh);
 
@@ -1293,78 +1295,28 @@ zfs_compare_timespec(struct timespec *t1, struct timespec *t2)
 	return (t1->tv_nsec - t2->tv_nsec);
 }
 
-/*
- *  Determine whether the znode's atime must be updated.  The logic mostly
- *  duplicates the Linux kernel's relatime_need_update() functionality.
- *  This function is only called if the underlying filesystem actually has
- *  atime updates enabled.
- */
-static inline boolean_t
-zfs_atime_need_update(znode_t *zp, timestruc_t *now)
-{
-	if (!ZTOZSB(zp)->z_relatime)
-		return (B_TRUE);
-
-	/*
-	 * In relatime mode, only update the atime if the previous atime
-	 * is earlier than either the ctime or mtime or if at least a day
-	 * has passed since the last update of atime.
-	 */
-	if (zfs_compare_timespec(&ZTOI(zp)->i_mtime, &ZTOI(zp)->i_atime) >= 0)
-		return (B_TRUE);
-
-	if (zfs_compare_timespec(&ZTOI(zp)->i_ctime, &ZTOI(zp)->i_atime) >= 0)
-		return (B_TRUE);
-
-	if ((long)now->tv_sec - ZTOI(zp)->i_atime.tv_sec >= 24*60*60)
-		return (B_TRUE);
-
-	return (B_FALSE);
-}
-
 /*
  * Prepare to update znode time stamps.
  *
  *	IN:	zp	- znode requiring timestamp update
- *		flag	- ATTR_MTIME, ATTR_CTIME, ATTR_ATIME flags
- *		have_tx	- true of caller is creating a new txg
+ *		flag	- ATTR_MTIME, ATTR_CTIME flags
  *
- *	OUT:	zp	- new atime (via underlying inode's i_atime)
+ *	OUT:	zp	- z_seq
  *		mtime	- new mtime
  *		ctime	- new ctime
  *
- * NOTE: The arguments are somewhat redundant.  The following condition
- * is always true:
- *
- *		have_tx == !(flag & ATTR_ATIME)
+ *	Note: We don't update atime here, because we rely on Linux VFS to do
+ *	atime updating.
  */
 void
 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
-    uint64_t ctime[2], boolean_t have_tx)
+    uint64_t ctime[2])
 {
 	timestruc_t	now;
 
-	ASSERT(have_tx == !(flag & ATTR_ATIME));
 	gethrestime(&now);
 
-	/*
-	 * NOTE: The following test intentionally does not update z_atime_dirty
-	 * in the case where an ATIME update has been requested but for which
-	 * the update is omitted due to relatime logic.  The rationale being
-	 * that if the flag was set somewhere else, we should leave it alone
-	 * here.
-	 */
-	if (flag & ATTR_ATIME) {
-		if (zfs_atime_need_update(zp, &now)) {
-			ZFS_TIME_ENCODE(&now, zp->z_atime);
-			ZTOI(zp)->i_atime.tv_sec = zp->z_atime[0];
-			ZTOI(zp)->i_atime.tv_nsec = zp->z_atime[1];
-			zp->z_atime_dirty = 1;
-		}
-	} else {
-		zp->z_atime_dirty = 0;
-		zp->z_seq++;
-	}
+	zp->z_seq++;
 
 	if (flag & ATTR_MTIME) {
 		ZFS_TIME_ENCODE(&now, mtime);
@@ -1437,7 +1389,7 @@ zfs_extend(znode_t *zp, uint64_t end)
 	/*
 	 * We will change zp_size, lock the whole file.
 	 */
-	rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, 0, UINT64_MAX, RL_WRITER);
 
 	/*
 	 * Nothing to do if file already at desired length.
@@ -1510,13 +1462,12 @@ zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len)
 	int64_t	off;
 	void *pb;
 
-	ASSERT((start & PAGE_CACHE_MASK) ==
-	    ((start + len - 1) & PAGE_CACHE_MASK));
+	ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK));
 
-	off = start & (PAGE_CACHE_SIZE - 1);
-	start &= PAGE_CACHE_MASK;
+	off = start & (PAGE_SIZE - 1);
+	start &= PAGE_MASK;
 
-	pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT);
+	pp = find_lock_page(mp, start >> PAGE_SHIFT);
 	if (pp) {
 		if (mapping_writably_mapped(mp))
 			flush_dcache_page(pp);
@@ -1532,7 +1483,7 @@ zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len)
 		SetPageUptodate(pp);
 		ClearPageError(pp);
 		unlock_page(pp);
-		page_cache_release(pp);
+		put_page(pp);
 	}
 }
 
@@ -1555,7 +1506,7 @@ zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
 	/*
 	 * Lock the range being freed.
 	 */
-	rl = zfs_range_lock(zp, off, len, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, off, len, RL_WRITER);
 
 	/*
 	 * Nothing to do if file already at desired length.
@@ -1579,14 +1530,14 @@ zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
 		loff_t first_page_offset, last_page_offset;
 
 		/* first possible full page in hole */
-		first_page = (off + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+		first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT;
 		/* last page of hole */
-		last_page = (off + len) >> PAGE_CACHE_SHIFT;
+		last_page = (off + len) >> PAGE_SHIFT;
 
 		/* offset of first_page */
-		first_page_offset = first_page << PAGE_CACHE_SHIFT;
+		first_page_offset = first_page << PAGE_SHIFT;
 		/* offset of last_page */
-		last_page_offset = last_page << PAGE_CACHE_SHIFT;
+		last_page_offset = last_page << PAGE_SHIFT;
 
 		/* truncate whole pages */
 		if (last_page_offset > first_page_offset) {
@@ -1637,7 +1588,7 @@ zfs_trunc(znode_t *zp, uint64_t end)
 	/*
 	 * We will change zp_size, lock the whole file.
 	 */
-	rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, 0, UINT64_MAX, RL_WRITER);
 
 	/*
 	 * Nothing to do if file already at desired length.
@@ -1737,7 +1688,7 @@ log:
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
 	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
 	    NULL, &zp->z_pflags, 8);
-	zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
+	zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
 	error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
 	ASSERT(error == 0);
 

module/zfs/zio.c

@@ -139,10 +139,10 @@ zio_init(void)
 		if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE))
 			continue;
 #endif
-		if (size <= 4 * SPA_MINBLOCKSIZE) {
+		if (size < PAGESIZE) {
 			align = SPA_MINBLOCKSIZE;
 		} else if (IS_P2ALIGNED(size, p2 >> 2)) {
-			align = MIN(p2 >> 2, PAGESIZE);
+			align = PAGESIZE;
 		}
 
 		if (align != 0) {
@@ -1415,6 +1415,31 @@ zio_execute(zio_t *zio)
 	spl_fstrans_unmark(cookie);
 }
 
+/*
+ * Used to determine if in the current context the stack is sized large
+ * enough to allow zio_execute() to be called recursively.  A minimum
+ * stack size of 16K is required to avoid needing to re-dispatch the zio.
+ */
+boolean_t
+zio_execute_stack_check(zio_t *zio)
+{
+#if !defined(HAVE_LARGE_STACKS)
+	dsl_pool_t *dp = spa_get_dsl(zio->io_spa);
+
+	/* Executing in txg_sync_thread() context. */
+	if (dp && curthread == dp->dp_tx.tx_sync_thread)
+		return (B_TRUE);
+
+	/* Pool initialization outside of zio_taskq context. */
+	if (dp && spa_is_initializing(dp->dp_spa) &&
+	    !zio_taskq_member(zio, ZIO_TASKQ_ISSUE) &&
+	    !zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH))
+		return (B_TRUE);
+#endif /* HAVE_LARGE_STACKS */
+
+	return (B_FALSE);
+}
+
 __attribute__((always_inline))
 static inline void
 __zio_execute(zio_t *zio)
@@ -1424,8 +1449,6 @@ __zio_execute(zio_t *zio)
 	while (zio->io_stage < ZIO_STAGE_DONE) {
 		enum zio_stage pipeline = zio->io_pipeline;
 		enum zio_stage stage = zio->io_stage;
-		dsl_pool_t *dp;
-		boolean_t cut;
 		int rv;
 
 		ASSERT(!MUTEX_HELD(&zio->io_lock));
@@ -1438,10 +1461,6 @@ __zio_execute(zio_t *zio)
 
 		ASSERT(stage <= ZIO_STAGE_DONE);
 
-		dp = spa_get_dsl(zio->io_spa);
-		cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
-		    zio_requeue_io_start_cut_in_line : B_FALSE;
-
 		/*
 		 * If we are in interrupt context and this pipeline stage
 		 * will grab a config lock that is held across I/O,
@@ -1453,21 +1472,19 @@ __zio_execute(zio_t *zio)
 		 */
 		if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL &&
 		    zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) {
+			boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
+			    zio_requeue_io_start_cut_in_line : B_FALSE;
 			zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
 			return;
 		}
 
 		/*
-		 * If we executing in the context of the tx_sync_thread,
-		 * or we are performing pool initialization outside of a
-		 * zio_taskq[ZIO_TASKQ_ISSUE|ZIO_TASKQ_ISSUE_HIGH] context.
-		 * Then issue the zio asynchronously to minimize stack usage
-		 * for these deep call paths.
+		 * If the current context doesn't have large enough stacks
+		 * the zio must be issued asynchronously to prevent overflow.
 		 */
-		if ((dp && curthread == dp->dp_tx.tx_sync_thread) ||
-		    (dp && spa_is_initializing(dp->dp_spa) &&
-		    !zio_taskq_member(zio, ZIO_TASKQ_ISSUE) &&
-		    !zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH))) {
+		if (zio_execute_stack_check(zio)) {
+			boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
+			    zio_requeue_io_start_cut_in_line : B_FALSE;
 			zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
 			return;
 		}
@@ -3455,7 +3472,7 @@ zbookmark_is_before(const dnode_phys_t *dnp, const zbookmark_phys_t *zb1,
 
 	if (zb1->zb_object == DMU_META_DNODE_OBJECT) {
 		uint64_t nextobj = zb1nextL0 *
-		    (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT) >> DNODE_SHIFT;
+		    (dnp->dn_datablkszsec << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
 		return (nextobj <= zb2thisobj);
 	}
 

module/zfs/zpl_ctldir.c

@@ -81,7 +81,7 @@ out:
 	return (error);
 }
 
-#if !defined(HAVE_VFS_ITERATE)
+#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
 static int
 zpl_root_readdir(struct file *filp, void *dirent, filldir_t filldir)
 {
@@ -100,16 +100,17 @@ zpl_root_readdir(struct file *filp, void *dirent, filldir_t filldir)
  */
 /* ARGSUSED */
 static int
-zpl_root_getattr(struct vfsmount *mnt, struct dentry *dentry,
-    struct kstat *stat)
+zpl_root_getattr_impl(const struct path *path, struct kstat *stat,
+    u32 request_mask, unsigned int query_flags)
 {
-	int error;
+	struct inode *ip = path->dentry->d_inode;
 
-	error = simple_getattr(mnt, dentry, stat);
-	stat->atime = CURRENT_TIME;
+	generic_fillattr(ip, stat);
+	stat->atime = current_time(ip);
 
-	return (error);
+	return (0);
 }
+ZPL_GETATTR_WRAPPER(zpl_root_getattr);
 
 static struct dentry *
 #ifdef HAVE_LOOKUP_NAMEIDATA
@@ -144,7 +145,9 @@ const struct file_operations zpl_fops_root = {
 	.open		= zpl_common_open,
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-#ifdef HAVE_VFS_ITERATE
+#ifdef HAVE_VFS_ITERATE_SHARED
+	.iterate_shared	= zpl_root_iterate,
+#elif defined(HAVE_VFS_ITERATE)
 	.iterate	= zpl_root_iterate,
 #else
 	.readdir	= zpl_root_readdir,
@@ -285,7 +288,7 @@ out:
 	return (error);
 }
 
-#if !defined(HAVE_VFS_ITERATE)
+#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
 static int
 zpl_snapdir_readdir(struct file *filp, void *dirent, filldir_t filldir)
 {
@@ -299,13 +302,17 @@ zpl_snapdir_readdir(struct file *filp, void *dirent, filldir_t filldir)
 }
 #endif /* HAVE_VFS_ITERATE */
 
-int
-zpl_snapdir_rename(struct inode *sdip, struct dentry *sdentry,
-    struct inode *tdip, struct dentry *tdentry)
+static int
+zpl_snapdir_rename2(struct inode *sdip, struct dentry *sdentry,
+    struct inode *tdip, struct dentry *tdentry, unsigned int flags)
 {
 	cred_t *cr = CRED();
 	int error;
 
+	/* We probably don't want to support renameat2(2) in ctldir */
+	if (flags)
+		return (-EINVAL);
+
 	crhold(cr);
 	error = -zfsctl_snapdir_rename(sdip, dname(sdentry),
 	    tdip, dname(tdentry), cr, 0);
@@ -315,6 +322,15 @@ zpl_snapdir_rename(struct inode *sdip, struct dentry *sdentry,
 	return (error);
 }
 
+#ifndef HAVE_RENAME_WANTS_FLAGS
+static int
+zpl_snapdir_rename(struct inode *sdip, struct dentry *sdentry,
+    struct inode *tdip, struct dentry *tdentry)
+{
+	return (zpl_snapdir_rename2(sdip, sdentry, tdip, tdentry, 0));
+}
+#endif
+
 static int
 zpl_snapdir_rmdir(struct inode *dip, struct dentry *dentry)
 {
@@ -360,21 +376,22 @@ zpl_snapdir_mkdir(struct inode *dip, struct dentry *dentry, zpl_umode_t mode)
  */
 /* ARGSUSED */
 static int
-zpl_snapdir_getattr(struct vfsmount *mnt, struct dentry *dentry,
-    struct kstat *stat)
+zpl_snapdir_getattr_impl(const struct path *path, struct kstat *stat,
+    u32 request_mask, unsigned int query_flags)
 {
-	zfs_sb_t *zsb = ITOZSB(dentry->d_inode);
-	int error;
+	struct inode *ip = path->dentry->d_inode;
+	zfs_sb_t *zsb = ITOZSB(path->dentry->d_inode);
 
 	ZFS_ENTER(zsb);
-	error = simple_getattr(mnt, dentry, stat);
+	generic_fillattr(path->dentry->d_inode, stat);
 	stat->nlink = stat->size = 2;
 	stat->ctime = stat->mtime = dmu_objset_snap_cmtime(zsb->z_os);
-	stat->atime = CURRENT_TIME;
+	stat->atime = current_time(ip);
 	ZFS_EXIT(zsb);
 
-	return (error);
+	return (0);
 }
+ZPL_GETATTR_WRAPPER(zpl_snapdir_getattr);
 
 /*
  * The '.zfs/snapshot' directory file operations.  These mainly control
@@ -385,7 +402,9 @@ const struct file_operations zpl_fops_snapdir = {
 	.open		= zpl_common_open,
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-#ifdef HAVE_VFS_ITERATE
+#ifdef HAVE_VFS_ITERATE_SHARED
+	.iterate_shared	= zpl_snapdir_iterate,
+#elif defined(HAVE_VFS_ITERATE)
 	.iterate	= zpl_snapdir_iterate,
 #else
 	.readdir	= zpl_snapdir_readdir,
@@ -401,7 +420,11 @@ const struct file_operations zpl_fops_snapdir = {
 const struct inode_operations zpl_ops_snapdir = {
 	.lookup		= zpl_snapdir_lookup,
 	.getattr	= zpl_snapdir_getattr,
+#ifdef HAVE_RENAME_WANTS_FLAGS
+	.rename		= zpl_snapdir_rename2,
+#else
 	.rename		= zpl_snapdir_rename,
+#endif
 	.rmdir		= zpl_snapdir_rmdir,
 	.mkdir		= zpl_snapdir_mkdir,
 };
@@ -472,7 +495,7 @@ out:
 	return (error);
 }
 
-#if !defined(HAVE_VFS_ITERATE)
+#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
 static int
 zpl_shares_readdir(struct file *filp, void *dirent, filldir_t filldir)
 {
@@ -488,10 +511,10 @@ zpl_shares_readdir(struct file *filp, void *dirent, filldir_t filldir)
 
 /* ARGSUSED */
 static int
-zpl_shares_getattr(struct vfsmount *mnt, struct dentry *dentry,
-    struct kstat *stat)
+zpl_shares_getattr_impl(const struct path *path, struct kstat *stat,
+    u32 request_mask, unsigned int query_flags)
 {
-	struct inode *ip = dentry->d_inode;
+	struct inode *ip = path->dentry->d_inode;
 	zfs_sb_t *zsb = ITOZSB(ip);
 	znode_t *dzp;
 	int error;
@@ -499,11 +522,11 @@ zpl_shares_getattr(struct vfsmount *mnt, struct dentry *dentry,
 	ZFS_ENTER(zsb);
 
 	if (zsb->z_shares_dir == 0) {
-		error = simple_getattr(mnt, dentry, stat);
+		generic_fillattr(path->dentry->d_inode, stat);
 		stat->nlink = stat->size = 2;
-		stat->atime = CURRENT_TIME;
+		stat->atime = current_time(ip);
 		ZFS_EXIT(zsb);
-		return (error);
+		return (0);
 	}
 
 	error = -zfs_zget(zsb, zsb->z_shares_dir, &dzp);
@@ -517,6 +540,7 @@ zpl_shares_getattr(struct vfsmount *mnt, struct dentry *dentry,
 
 	return (error);
 }
+ZPL_GETATTR_WRAPPER(zpl_shares_getattr);
 
 /*
  * The '.zfs/shares' directory file operations.
@@ -525,7 +549,9 @@ const struct file_operations zpl_fops_shares = {
 	.open		= zpl_common_open,
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-#ifdef HAVE_VFS_ITERATE
+#ifdef HAVE_VFS_ITERATE_SHARED
+	.iterate_shared	= zpl_shares_iterate,
+#elif defined(HAVE_VFS_ITERATE)
 	.iterate	= zpl_shares_iterate,
 #else
 	.readdir	= zpl_shares_readdir,

module/zfs/zpl_file.c

@@ -24,6 +24,9 @@
  */
 
 
+#ifdef CONFIG_COMPAT
+#include <linux/compat.h>
+#endif
 #include <sys/dmu_objset.h>
 #include <sys/zfs_vfsops.h>
 #include <sys/zfs_vnops.h>
@@ -90,7 +93,7 @@ zpl_iterate(struct file *filp, struct dir_context *ctx)
 	return (error);
 }
 
-#if !defined(HAVE_VFS_ITERATE)
+#if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
 static int
 zpl_readdir(struct file *filp, void *dirent, filldir_t filldir)
 {
@@ -128,12 +131,15 @@ zpl_fsync(struct file *filp, struct dentry *dentry, int datasync)
 	return (error);
 }
 
+#ifdef HAVE_FILE_AIO_FSYNC
 static int
 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
 {
 	struct file *filp = kiocb->ki_filp;
 	return (zpl_fsync(filp, filp->f_path.dentry, datasync));
 }
+#endif
+
 #elif defined(HAVE_FSYNC_WITHOUT_DENTRY)
 /*
  * Linux 2.6.35 - 3.0 API,
@@ -159,11 +165,14 @@ zpl_fsync(struct file *filp, int datasync)
 	return (error);
 }
 
+#ifdef HAVE_FILE_AIO_FSYNC
 static int
 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
 {
 	return (zpl_fsync(kiocb->ki_filp, datasync));
 }
+#endif
+
 #elif defined(HAVE_FSYNC_RANGE)
 /*
  * Linux 3.1 - 3.x API,
@@ -194,11 +203,14 @@ zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
 	return (error);
 }
 
+#ifdef HAVE_FILE_AIO_FSYNC
 static int
 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
 {
 	return (zpl_fsync(kiocb->ki_filp, kiocb->ki_pos, -1, datasync));
 }
+#endif
+
 #else
 #error "Unsupported fops->fsync() implementation"
 #endif
@@ -247,20 +259,6 @@ zpl_read_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos,
 	    flags, cr, 0));
 }
 
-static ssize_t
-zpl_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
-{
-	cred_t *cr = CRED();
-	ssize_t read;
-
-	crhold(cr);
-	read = zpl_read_common(filp->f_mapping->host, buf, len, ppos,
-	    UIO_USERSPACE, filp->f_flags, cr);
-	crfree(cr);
-
-	return (read);
-}
-
 static ssize_t
 zpl_iter_read_common(struct kiocb *kiocb, const struct iovec *iovp,
     unsigned long nr_segs, size_t count, uio_seg_t seg, size_t skip)
@@ -274,6 +272,7 @@ zpl_iter_read_common(struct kiocb *kiocb, const struct iovec *iovp,
 	    nr_segs, &kiocb->ki_pos, seg, filp->f_flags, cr, skip);
 	crfree(cr);
 
+	file_accessed(filp);
 	return (read);
 }
 
@@ -298,7 +297,14 @@ static ssize_t
 zpl_aio_read(struct kiocb *kiocb, const struct iovec *iovp,
     unsigned long nr_segs, loff_t pos)
 {
-	return (zpl_iter_read_common(kiocb, iovp, nr_segs, kiocb->ki_nbytes,
+	ssize_t ret;
+	size_t count;
+
+	ret = generic_segment_checks(iovp, &nr_segs, &count, VERIFY_WRITE);
+	if (ret)
+		return (ret);
+
+	return (zpl_iter_read_common(kiocb, iovp, nr_segs, count,
 	    UIO_USERSPACE, 0));
 }
 #endif /* HAVE_VFS_RW_ITERATE */
@@ -336,6 +342,7 @@ zpl_write_common_iovec(struct inode *ip, const struct iovec *iovp, size_t count,
 
 	return (wrote);
 }
+
 inline ssize_t
 zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos,
     uio_seg_t segment, int flags, cred_t *cr)
@@ -349,20 +356,6 @@ zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos,
 	    flags, cr, 0));
 }
 
-static ssize_t
-zpl_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
-{
-	cred_t *cr = CRED();
-	ssize_t wrote;
-
-	crhold(cr);
-	wrote = zpl_write_common(filp->f_mapping->host, buf, len, ppos,
-	    UIO_USERSPACE, filp->f_flags, cr);
-	crfree(cr);
-
-	return (wrote);
-}
-
 static ssize_t
 zpl_iter_write_common(struct kiocb *kiocb, const struct iovec *iovp,
     unsigned long nr_segs, size_t count, uio_seg_t seg, size_t skip)
@@ -383,16 +376,42 @@ zpl_iter_write_common(struct kiocb *kiocb, const struct iovec *iovp,
 static ssize_t
 zpl_iter_write(struct kiocb *kiocb, struct iov_iter *from)
 {
+	size_t count;
 	ssize_t ret;
 	uio_seg_t seg = UIO_USERSPACE;
+
+#ifndef HAVE_GENERIC_WRITE_CHECKS_KIOCB
+	struct file *file = kiocb->ki_filp;
+	struct address_space *mapping = file->f_mapping;
+	struct inode *ip = mapping->host;
+	int isblk = S_ISBLK(ip->i_mode);
+
+	count = iov_iter_count(from);
+	ret = generic_write_checks(file, &kiocb->ki_pos, &count, isblk);
+	if (ret)
+		return (ret);
+#else
+	/*
+	 * XXX - ideally this check should be in the same lock region with
+	 * write operations, so that there's no TOCTTOU race when doing
+	 * append and someone else grow the file.
+	 */
+	ret = generic_write_checks(kiocb, from);
+	if (ret <= 0)
+		return (ret);
+	count = ret;
+#endif
+
 	if (from->type & ITER_KVEC)
 		seg = UIO_SYSSPACE;
 	if (from->type & ITER_BVEC)
 		seg = UIO_BVEC;
+
 	ret = zpl_iter_write_common(kiocb, from->iov, from->nr_segs,
-	    iov_iter_count(from), seg, from->iov_offset);
+	    count, seg, from->iov_offset);
 	if (ret > 0)
 		iov_iter_advance(from, ret);
+
 	return (ret);
 }
 #else
@@ -400,7 +419,22 @@ static ssize_t
 zpl_aio_write(struct kiocb *kiocb, const struct iovec *iovp,
     unsigned long nr_segs, loff_t pos)
 {
-	return (zpl_iter_write_common(kiocb, iovp, nr_segs, kiocb->ki_nbytes,
+	struct file *file = kiocb->ki_filp;
+	struct address_space *mapping = file->f_mapping;
+	struct inode *ip = mapping->host;
+	int isblk = S_ISBLK(ip->i_mode);
+	size_t count;
+	ssize_t ret;
+
+	ret = generic_segment_checks(iovp, &nr_segs, &count, VERIFY_READ);
+	if (ret)
+		return (ret);
+
+	ret = generic_write_checks(file, &pos, &count, isblk);
+	if (ret)
+		return (ret);
+
+	return (zpl_iter_write_common(kiocb, iovp, nr_segs, count,
 	    UIO_USERSPACE, 0));
 }
 #endif /* HAVE_VFS_RW_ITERATE */
@@ -416,13 +450,13 @@ zpl_llseek(struct file *filp, loff_t offset, int whence)
 		loff_t maxbytes = ip->i_sb->s_maxbytes;
 		loff_t error;
 
-		spl_inode_lock(ip);
+		spl_inode_lock_shared(ip);
 		cookie = spl_fstrans_mark();
 		error = -zfs_holey(ip, whence, &offset);
 		spl_fstrans_unmark(cookie);
 		if (error == 0)
 			error = lseek_execute(filp, ip, offset, maxbytes);
-		spl_inode_unlock(ip);
+		spl_inode_unlock_shared(ip);
 
 		return (error);
 	}
@@ -646,8 +680,6 @@ zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
 	if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
 		return (error);
 
-	crhold(cr);
-
 	if (offset < 0 || len <= 0)
 		return (-EINVAL);
 
@@ -666,6 +698,7 @@ zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
 	bf.l_len = len;
 	bf.l_pid = 0;
 
+	crhold(cr);
 	cookie = spl_fstrans_mark();
 	error = -zfs_space(ip, F_FREESP, &bf, FWRITE, offset, cr);
 	spl_fstrans_unmark(cookie);
@@ -725,8 +758,7 @@ zpl_ioctl_getflags(struct file *filp, void __user *arg)
  * is outside of our jurisdiction.
  */
 
-#define	fchange(f0, f1, b0, b1) ((((f0) & (b0)) == (b0)) != \
-	(((b1) & (f1)) == (f1)))
+#define	fchange(f0, f1, b0, b1) (!((f0) & (b0)) != !((f1) & (b1)))
 
 static int
 zpl_ioctl_setflags(struct file *filp, void __user *arg)
@@ -798,7 +830,17 @@ zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 static long
 zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 {
-	return (zpl_ioctl(filp, cmd, arg));
+	switch (cmd) {
+	case FS_IOC32_GETFLAGS:
+		cmd = FS_IOC_GETFLAGS;
+		break;
+	case FS_IOC32_SETFLAGS:
+		cmd = FS_IOC_SETFLAGS;
+		break;
+	default:
+		return (-ENOTTY);
+	}
+	return (zpl_ioctl(filp, cmd, (unsigned long)compat_ptr(arg)));
 }
 #endif /* CONFIG_COMPAT */
 
@@ -814,18 +856,24 @@ const struct file_operations zpl_file_operations = {
 	.open		= zpl_open,
 	.release	= zpl_release,
 	.llseek		= zpl_llseek,
-	.read		= zpl_read,
-	.write		= zpl_write,
 #ifdef HAVE_VFS_RW_ITERATE
+#ifdef HAVE_NEW_SYNC_READ
+	.read		= new_sync_read,
+	.write		= new_sync_write,
+#endif
 	.read_iter	= zpl_iter_read,
 	.write_iter	= zpl_iter_write,
 #else
+	.read		= do_sync_read,
+	.write		= do_sync_write,
 	.aio_read	= zpl_aio_read,
 	.aio_write	= zpl_aio_write,
 #endif
 	.mmap		= zpl_mmap,
 	.fsync		= zpl_fsync,
+#ifdef HAVE_FILE_AIO_FSYNC
 	.aio_fsync	= zpl_aio_fsync,
+#endif
 #ifdef HAVE_FILE_FALLOCATE
 	.fallocate	= zpl_fallocate,
 #endif /* HAVE_FILE_FALLOCATE */
@@ -838,7 +886,9 @@ const struct file_operations zpl_file_operations = {
 const struct file_operations zpl_dir_file_operations = {
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-#ifdef HAVE_VFS_ITERATE
+#ifdef HAVE_VFS_ITERATE_SHARED
+	.iterate_shared	= zpl_iterate,
+#elif defined(HAVE_VFS_ITERATE)
 	.iterate	= zpl_iterate,
 #else
 	.readdir	= zpl_readdir,

module/zfs/zpl_inode.c

@@ -50,7 +50,7 @@ zpl_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
 	int zfs_flags = 0;
 	zfs_sb_t *zsb = dentry->d_sb->s_fs_info;
 
-	if (dlen(dentry) > ZFS_MAXNAMELEN)
+	if (dlen(dentry) >= ZAP_MAXNAMELEN)
 		return (ERR_PTR(-ENAMETOOLONG));
 
 	crhold(cr);
@@ -102,9 +102,13 @@ zpl_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
 		struct dentry *new_dentry;
 		struct qstr ci_name;
 
-		ci_name.name = pn.pn_buf;
-		ci_name.len = strlen(pn.pn_buf);
-		new_dentry = d_add_ci(dentry, ip, &ci_name);
+		if (strcmp(dname(dentry), pn.pn_buf) == 0) {
+			new_dentry = d_splice_alias(ip,  dentry);
+		} else {
+			ci_name.name = pn.pn_buf;
+			ci_name.len = strlen(pn.pn_buf);
+			new_dentry = d_add_ci(dentry, ip, &ci_name);
+		}
 		kmem_free(pn.pn_buf, ZFS_MAXNAMELEN);
 		return (new_dentry);
 	} else {
@@ -298,18 +302,25 @@ zpl_rmdir(struct inode * dir, struct dentry *dentry)
 }
 
 static int
-zpl_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
+zpl_getattr_impl(const struct path *path, struct kstat *stat, u32 request_mask,
+    unsigned int query_flags)
 {
 	int error;
 	fstrans_cookie_t cookie;
 
 	cookie = spl_fstrans_mark();
-	error = -zfs_getattr_fast(dentry->d_inode, stat);
+
+	/*
+	 * XXX request_mask and query_flags currently ignored.
+	 */
+
+	error = -zfs_getattr_fast(path->dentry->d_inode, stat);
 	spl_fstrans_unmark(cookie);
 	ASSERT3S(error, <=, 0);
 
 	return (error);
 }
+ZPL_GETATTR_WRAPPER(zpl_getattr);
 
 static int
 zpl_setattr(struct dentry *dentry, struct iattr *ia)
@@ -320,7 +331,7 @@ zpl_setattr(struct dentry *dentry, struct iattr *ia)
 	int error;
 	fstrans_cookie_t cookie;
 
-	error = inode_change_ok(ip, ia);
+	error = setattr_prepare(dentry, ia);
 	if (error)
 		return (error);
 
@@ -335,6 +346,9 @@ zpl_setattr(struct dentry *dentry, struct iattr *ia)
 	vap->va_mtime = ia->ia_mtime;
 	vap->va_ctime = ia->ia_ctime;
 
+	if (vap->va_mask & ATTR_ATIME)
+		ip->i_atime = ia->ia_atime;
+
 	cookie = spl_fstrans_mark();
 	error = -zfs_setattr(ip, vap, 0, cr);
 	if (!error && (ia->ia_valid & ATTR_MODE))
@@ -349,13 +363,17 @@ zpl_setattr(struct dentry *dentry, struct iattr *ia)
 }
 
 static int
-zpl_rename(struct inode *sdip, struct dentry *sdentry,
-    struct inode *tdip, struct dentry *tdentry)
+zpl_rename2(struct inode *sdip, struct dentry *sdentry,
+    struct inode *tdip, struct dentry *tdentry, unsigned int flags)
 {
 	cred_t *cr = CRED();
 	int error;
 	fstrans_cookie_t cookie;
 
+	/* We don't have renameat2(2) support */
+	if (flags)
+		return (-EINVAL);
+
 	crhold(cr);
 	cookie = spl_fstrans_mark();
 	error = -zfs_rename(sdip, dname(sdentry), tdip, dname(tdentry), cr, 0);
@@ -366,6 +384,15 @@ zpl_rename(struct inode *sdip, struct dentry *sdentry,
 	return (error);
 }
 
+#ifndef HAVE_RENAME_WANTS_FLAGS
+static int
+zpl_rename(struct inode *sdip, struct dentry *sdentry,
+    struct inode *tdip, struct dentry *tdentry)
+{
+	return (zpl_rename2(sdip, sdentry, tdip, tdentry, 0));
+}
+#endif
+
 static int
 zpl_symlink(struct inode *dir, struct dentry *dentry, const char *name)
 {
@@ -530,7 +557,7 @@ zpl_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
 		return (-EMLINK);
 
 	crhold(cr);
-	ip->i_ctime = CURRENT_TIME_SEC;
+	ip->i_ctime = current_time(ip);
 	igrab(ip); /* Use ihold() if available */
 
 	cookie = spl_fstrans_mark();
@@ -642,19 +669,13 @@ zpl_revalidate(struct dentry *dentry, unsigned int flags)
 }
 
 const struct inode_operations zpl_inode_operations = {
-	.create		= zpl_create,
-	.link		= zpl_link,
-	.unlink		= zpl_unlink,
-	.symlink	= zpl_symlink,
-	.mkdir		= zpl_mkdir,
-	.rmdir		= zpl_rmdir,
-	.mknod		= zpl_mknod,
-	.rename		= zpl_rename,
 	.setattr	= zpl_setattr,
 	.getattr	= zpl_getattr,
+#ifdef HAVE_GENERIC_SETXATTR
 	.setxattr	= generic_setxattr,
 	.getxattr	= generic_getxattr,
 	.removexattr	= generic_removexattr,
+#endif
 	.listxattr	= zpl_xattr_list,
 #ifdef HAVE_INODE_TRUNCATE_RANGE
 	.truncate_range = zpl_truncate_range,
@@ -663,6 +684,9 @@ const struct inode_operations zpl_inode_operations = {
 	.fallocate	= zpl_fallocate,
 #endif /* HAVE_INODE_FALLOCATE */
 #if defined(CONFIG_FS_POSIX_ACL)
+#if defined(HAVE_SET_ACL)
+	.set_acl	= zpl_set_acl,
+#endif
 #if defined(HAVE_GET_ACL)
 	.get_acl	= zpl_get_acl,
 #elif defined(HAVE_CHECK_ACL)
@@ -682,14 +706,23 @@ const struct inode_operations zpl_dir_inode_operations = {
 	.mkdir		= zpl_mkdir,
 	.rmdir		= zpl_rmdir,
 	.mknod		= zpl_mknod,
+#ifdef HAVE_RENAME_WANTS_FLAGS
+	.rename		= zpl_rename2,
+#else
 	.rename		= zpl_rename,
+#endif
 	.setattr	= zpl_setattr,
 	.getattr	= zpl_getattr,
+#ifdef HAVE_GENERIC_SETXATTR
 	.setxattr	= generic_setxattr,
 	.getxattr	= generic_getxattr,
 	.removexattr	= generic_removexattr,
+#endif
 	.listxattr	= zpl_xattr_list,
 #if defined(CONFIG_FS_POSIX_ACL)
+#if defined(HAVE_SET_ACL)
+	.set_acl	= zpl_set_acl,
+#endif
 #if defined(HAVE_GET_ACL)
 	.get_acl	= zpl_get_acl,
 #elif defined(HAVE_CHECK_ACL)
@@ -701,7 +734,9 @@ const struct inode_operations zpl_dir_inode_operations = {
 };
 
 const struct inode_operations zpl_symlink_inode_operations = {
+#ifdef HAVE_GENERIC_READLINK
 	.readlink	= generic_readlink,
+#endif
 #if defined(HAVE_GET_LINK_DELAYED) || defined(HAVE_GET_LINK_COOKIE)
 	.get_link	= zpl_get_link,
 #elif defined(HAVE_FOLLOW_LINK_COOKIE) || defined(HAVE_FOLLOW_LINK_NAMEIDATA)
@@ -712,20 +747,27 @@ const struct inode_operations zpl_symlink_inode_operations = {
 #endif
 	.setattr	= zpl_setattr,
 	.getattr	= zpl_getattr,
+#ifdef HAVE_GENERIC_SETXATTR
 	.setxattr	= generic_setxattr,
 	.getxattr	= generic_getxattr,
 	.removexattr	= generic_removexattr,
+#endif
 	.listxattr	= zpl_xattr_list,
 };
 
 const struct inode_operations zpl_special_inode_operations = {
 	.setattr	= zpl_setattr,
 	.getattr	= zpl_getattr,
+#ifdef HAVE_GENERIC_SETXATTR
 	.setxattr	= generic_setxattr,
 	.getxattr	= generic_getxattr,
 	.removexattr	= generic_removexattr,
+#endif
 	.listxattr	= zpl_xattr_list,
 #if defined(CONFIG_FS_POSIX_ACL)
+#if defined(HAVE_SET_ACL)
+	.set_acl	= zpl_set_acl,
+#endif
 #if defined(HAVE_GET_ACL)
 	.get_acl	= zpl_get_acl,
 #elif defined(HAVE_CHECK_ACL)

module/zfs/zpl_super.c

@@ -336,12 +336,12 @@ zpl_parse_options(char *osname, char *mntopts, zfs_mntopts_t *zmo,
 
 	if (mntopts) {
 		substring_t args[MAX_OPT_ARGS];
-		char *tmp_mntopts, *p;
+		char *tmp_mntopts, *p, *t;
 		int token;
 
-		tmp_mntopts = strdup(mntopts);
+		t = tmp_mntopts = strdup(mntopts);
 
-		while ((p = strsep(&tmp_mntopts, ",")) != NULL) {
+		while ((p = strsep(&t, ",")) != NULL) {
 			if (!*p)
 				continue;