diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c
index 19b0d61f09..d81199765c 100644
--- a/cmd/zdb/zdb.c
+++ b/cmd/zdb/zdb.c
@@ -8041,6 +8041,17 @@ dump_mos_leaks(spa_t *spa)
 		}
 	}
 
+	if (spa->spa_brt != NULL) {
+		brt_t *brt = spa->spa_brt;
+		for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
+			brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
+			if (brtvd != NULL && brtvd->bv_initiated) {
+				mos_obj_refd(brtvd->bv_mos_brtvdev);
+				mos_obj_refd(brtvd->bv_mos_entries);
+			}
+		}
+	}
+
 	/*
 	 * Visit all allocated objects and make sure they are referenced.
 	 */
diff --git a/cmd/zed/zed.d/statechange-slot_off.sh b/cmd/zed/zed.d/statechange-slot_off.sh
index 150012abe7..06acce93b8 100755
--- a/cmd/zed/zed.d/statechange-slot_off.sh
+++ b/cmd/zed/zed.d/statechange-slot_off.sh
@@ -5,7 +5,7 @@
 #
 # Bad SCSI disks can often "disappear and reappear" causing all sorts of chaos
 # as they flip between FAULTED and ONLINE.  If
-# ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT is set in zed.rc, and the disk gets
+# ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT is set in zed.rc, and the disk gets
 # FAULTED, then power down the slot via sysfs:
 #
 # /sys/class/enclosure/<enclosure>/<slot>/power_status
@@ -19,7 +19,7 @@
 # Exit codes:
 #   0: slot successfully powered off
 #   1: enclosure not available
-#   2: ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT disabled
+#   2: ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT disabled
 #   3: vdev was not FAULTED
 #   4: The enclosure sysfs path passed from ZFS does not exist
 #   5: Enclosure slot didn't actually turn off after we told it to
@@ -32,7 +32,7 @@ if [ ! -d /sys/class/enclosure ] ; then
 	exit 1
 fi
 
-if [ "${ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT}" != "1" ] ; then
+if [ "${ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT}" != "1" ] ; then
 	exit 2
 fi
 
diff --git a/cmd/zed/zed.d/zed-functions.sh b/cmd/zed/zed.d/zed-functions.sh
index 49b6b54029..3a2519633d 100644
--- a/cmd/zed/zed.d/zed-functions.sh
+++ b/cmd/zed/zed.d/zed-functions.sh
@@ -205,6 +205,10 @@ zed_notify()
     [ "${rv}" -eq 0 ] && num_success=$((num_success + 1))
     [ "${rv}" -eq 1 ] && num_failure=$((num_failure + 1))
 
+    zed_notify_ntfy "${subject}" "${pathname}"; rv=$?
+    [ "${rv}" -eq 0 ] && num_success=$((num_success + 1))
+    [ "${rv}" -eq 1 ] && num_failure=$((num_failure + 1))
+
     [ "${num_success}" -gt 0 ] && return 0
     [ "${num_failure}" -gt 0 ] && return 1
     return 2
@@ -527,6 +531,100 @@ zed_notify_pushover()
 }
 
 
+# zed_notify_ntfy (subject, pathname)
+#
+# Send a notification via Ntfy.sh <https://ntfy.sh/>.
+# The ntfy topic (ZED_NTFY_TOPIC) identifies the topic that the notification
+# will be sent to Ntfy.sh server. The ntfy url (ZED_NTFY_URL) defines the
+# self-hosted or provided hosted ntfy service location. The ntfy access token
+# <https://docs.ntfy.sh/publish/#access-tokens> (ZED_NTFY_ACCESS_TOKEN) reprsents an
+# access token that could be used if a topic is read/write protected. If a
+# topic can be written to publicaly, a ZED_NTFY_ACCESS_TOKEN is not required.
+#
+# Requires curl and sed executables to be installed in the standard PATH.
+#
+# References
+#   https://docs.ntfy.sh
+#
+# Arguments
+#   subject: notification subject
+#   pathname: pathname containing the notification message (OPTIONAL)
+#
+# Globals
+#   ZED_NTFY_TOPIC
+#   ZED_NTFY_ACCESS_TOKEN (OPTIONAL)
+#   ZED_NTFY_URL
+#
+# Return
+#   0: notification sent
+#   1: notification failed
+#   2: not configured
+#
+zed_notify_ntfy()
+{
+    local subject="$1"
+    local pathname="${2:-"/dev/null"}"
+    local msg_body
+    local msg_out
+    local msg_err
+
+    [ -n "${ZED_NTFY_TOPIC}" ] || return 2
+    local url="${ZED_NTFY_URL:-"https://ntfy.sh"}/${ZED_NTFY_TOPIC}"
+
+    if [ ! -r "${pathname}" ]; then
+        zed_log_err "ntfy cannot read \"${pathname}\""
+        return 1
+    fi
+
+    zed_check_cmd "curl" "sed" || return 1
+
+    # Read the message body in.
+    #
+    msg_body="$(cat "${pathname}")"
+
+    if [ -z "${msg_body}" ]
+    then
+        msg_body=$subject
+        subject=""
+    fi
+
+    # Send the POST request and check for errors.
+    #
+    if [ -n "${ZED_NTFY_ACCESS_TOKEN}" ]; then
+        msg_out="$( \
+        curl \
+        -u ":${ZED_NTFY_ACCESS_TOKEN}" \
+        -H "Title: ${subject}" \
+        -d "${msg_body}" \
+        -H "Priority: high" \
+        "${url}" \
+        2>/dev/null \
+        )"; rv=$?
+    else
+        msg_out="$( \
+        curl \
+        -H "Title: ${subject}" \
+        -d "${msg_body}" \
+        -H "Priority: high" \
+        "${url}" \
+        2>/dev/null \
+        )"; rv=$?
+    fi
+    if [ "${rv}" -ne 0 ]; then
+        zed_log_err "curl exit=${rv}"
+        return 1
+    fi
+    msg_err="$(echo "${msg_out}" \
+        | sed -n -e 's/.*"errors" *:.*\[\(.*\)\].*/\1/p')"
+    if [ -n "${msg_err}" ]; then
+        zed_log_err "ntfy \"${msg_err}"\"
+        return 1
+    fi
+    return 0
+}
+
+
+
 # zed_rate_limit (tag, [interval])
 #
 # Check whether an event of a given type [tag] has already occurred within the
diff --git a/cmd/zed/zed.d/zed.rc b/cmd/zed/zed.d/zed.rc
index 78dc1afc7b..bc269b155d 100644
--- a/cmd/zed/zed.d/zed.rc
+++ b/cmd/zed/zed.d/zed.rc
@@ -146,4 +146,26 @@ ZED_SYSLOG_SUBCLASS_EXCLUDE="history_event"
 # Power off the drive's slot in the enclosure if it becomes FAULTED.  This can
 # help silence misbehaving drives.  This assumes your drive enclosure fully
 # supports slot power control via sysfs.
-#ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT=1
+#ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT=1
+
+##
+# Ntfy topic
+# This defines which topic will receive the ntfy notification.
+#  <https://docs.ntfy.sh/publish/>
+# Disabled by default; uncomment to enable.
+#ZED_NTFY_TOPIC=""
+
+##
+# Ntfy access token (optional for public topics)
+# This defines an access token which can be used
+# to allow you to authenticate when sending to topics
+# <https://docs.ntfy.sh/publish/#access-tokens>
+# Disabled by default; uncomment to enable.
+#ZED_NTFY_ACCESS_TOKEN=""
+
+##
+# Ntfy Service URL
+# This defines which service the ntfy call will be directed toward
+#  <https://docs.ntfy.sh/install/>
+# https://ntfy.sh by default; uncomment to enable an alternative service url.
+#ZED_NTFY_URL="https://ntfy.sh"
diff --git a/cmd/zfs/zfs_main.c b/cmd/zfs/zfs_main.c
index 67b191d72e..3017de9ee7 100644
--- a/cmd/zfs/zfs_main.c
+++ b/cmd/zfs/zfs_main.c
@@ -3672,15 +3672,25 @@ zfs_do_list(int argc, char **argv)
 
 			for (char *tok; (tok = strsep(&optarg, ",")); ) {
 				static const char *const type_subopts[] = {
-					"filesystem", "volume",
-					"snapshot", "snap",
+					"filesystem",
+					"fs",
+					"volume",
+					"vol",
+					"snapshot",
+					"snap",
 					"bookmark",
-					"all" };
+					"all"
+				};
 				static const int type_types[] = {
-					ZFS_TYPE_FILESYSTEM, ZFS_TYPE_VOLUME,
-					ZFS_TYPE_SNAPSHOT, ZFS_TYPE_SNAPSHOT,
+					ZFS_TYPE_FILESYSTEM,
+					ZFS_TYPE_FILESYSTEM,
+					ZFS_TYPE_VOLUME,
+					ZFS_TYPE_VOLUME,
+					ZFS_TYPE_SNAPSHOT,
+					ZFS_TYPE_SNAPSHOT,
 					ZFS_TYPE_BOOKMARK,
-					ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK };
+					ZFS_TYPE_DATASET | ZFS_TYPE_BOOKMARK
+				};
 
 				for (c = 0; c < ARRAY_SIZE(type_subopts); ++c)
 					if (strcmp(tok, type_subopts[c]) == 0) {
diff --git a/cmd/zpool/zpool_main.c b/cmd/zpool/zpool_main.c
index 6687a44644..69bf9649ac 100644
--- a/cmd/zpool/zpool_main.c
+++ b/cmd/zpool/zpool_main.c
@@ -2161,6 +2161,7 @@ typedef struct status_cbdata {
 	boolean_t	cb_explain;
 	boolean_t	cb_first;
 	boolean_t	cb_dedup_stats;
+	boolean_t	cb_print_unhealthy;
 	boolean_t	cb_print_status;
 	boolean_t	cb_print_slow_ios;
 	boolean_t	cb_print_vdev_init;
@@ -2357,6 +2358,35 @@ health_str_to_color(const char *health)
 	return (NULL);
 }
 
+/*
+ * Called for each leaf vdev.  Returns 0 if the vdev is healthy.
+ * A vdev is unhealthy if any of the following are true:
+ * 1) there are read, write, or checksum errors,
+ * 2) its state is not ONLINE, or
+ * 3) slow IO reporting was requested (-s) and there are slow IOs.
+ */
+static int
+vdev_health_check_cb(void *hdl_data, nvlist_t *nv, void *data)
+{
+	status_cbdata_t *cb = data;
+	vdev_stat_t *vs;
+	uint_t vsc;
+	(void) hdl_data;
+
+	if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
+	    (uint64_t **)&vs, &vsc) != 0)
+		return (1);
+
+	if (vs->vs_checksum_errors || vs->vs_read_errors ||
+	    vs->vs_write_errors || vs->vs_state != VDEV_STATE_HEALTHY)
+		return (1);
+
+	if (cb->cb_print_slow_ios && vs->vs_slow_ios)
+		return (1);
+
+	return (0);
+}
+
 /*
  * Print out configuration state as requested by status_callback.
  */
@@ -2375,7 +2405,8 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 	const char *state;
 	const char *type;
 	const char *path = NULL;
-	const char *rcolor = NULL, *wcolor = NULL, *ccolor = NULL;
+	const char *rcolor = NULL, *wcolor = NULL, *ccolor = NULL,
+	    *scolor = NULL;
 
 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
 	    &child, &children) != 0)
@@ -2402,6 +2433,15 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 			state = gettext("AVAIL");
 	}
 
+	/*
+	 * If '-e' is specified then top-level vdevs and their children
+	 * can be pruned if all of their leaves are healthy.
+	 */
+	if (cb->cb_print_unhealthy && depth > 0 &&
+	    for_each_vdev_in_nvlist(nv, vdev_health_check_cb, cb) == 0) {
+		return;
+	}
+
 	printf_color(health_str_to_color(state),
 	    "\t%*s%-*s  %-8s", depth, "", cb->cb_namewidth - depth,
 	    name, state);
@@ -2416,6 +2456,9 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 		if (vs->vs_checksum_errors)
 			ccolor = ANSI_RED;
 
+		if (vs->vs_slow_ios)
+			scolor = ANSI_BLUE;
+
 		if (cb->cb_literal) {
 			fputc(' ', stdout);
 			printf_color(rcolor, "%5llu",
@@ -2448,9 +2491,10 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 			}
 
 			if (cb->cb_literal)
-				printf(" %5llu", (u_longlong_t)vs->vs_slow_ios);
+				printf_color(scolor, " %5llu",
+				    (u_longlong_t)vs->vs_slow_ios);
 			else
-				printf(" %5s", rbuf);
+				printf_color(scolor, " %5s", rbuf);
 		}
 		if (cb->cb_print_power) {
 			if (children == 0)  {
@@ -8999,9 +9043,11 @@ status_callback(zpool_handle_t *zhp, void *data)
 				(void) printf(gettext(
 				    "errors: No known data errors\n"));
 			} else if (!cbp->cb_verbose) {
+				color_start(ANSI_RED);
 				(void) printf(gettext("errors: %llu data "
 				    "errors, use '-v' for a list\n"),
 				    (u_longlong_t)nerr);
+				color_end();
 			} else {
 				print_error_log(zhp);
 			}
@@ -9022,6 +9068,7 @@ status_callback(zpool_handle_t *zhp, void *data)
  *              [pool] [interval [count]]
  *
  *	-c CMD	For each vdev, run command CMD
+ *	-e	Display only unhealthy vdevs
  *	-i	Display vdev initialization status.
  *	-g	Display guid for individual vdev name.
  *	-L	Follow links when resolving vdev path name.
@@ -9053,7 +9100,7 @@ zpool_do_status(int argc, char **argv)
 	};
 
 	/* check options */
-	while ((c = getopt_long(argc, argv, "c:igLpPsvxDtT:", long_options,
+	while ((c = getopt_long(argc, argv, "c:eigLpPsvxDtT:", long_options,
 	    NULL)) != -1) {
 		switch (c) {
 		case 'c':
@@ -9080,6 +9127,9 @@ zpool_do_status(int argc, char **argv)
 			}
 			cmd = optarg;
 			break;
+		case 'e':
+			cb.cb_print_unhealthy = B_TRUE;
+			break;
 		case 'i':
 			cb.cb_print_vdev_init = B_TRUE;
 			break;
diff --git a/config/kernel-mm-page-size.m4 b/config/kernel-mm-page-size.m4
new file mode 100644
index 0000000000..d5ebd92698
--- /dev/null
+++ b/config/kernel-mm-page-size.m4
@@ -0,0 +1,17 @@
+AC_DEFUN([ZFS_AC_KERNEL_SRC_MM_PAGE_SIZE], [
+	ZFS_LINUX_TEST_SRC([page_size], [
+		#include <linux/mm.h>
+	],[
+		unsigned long s;
+		s = page_size(NULL);
+	])
+])
+AC_DEFUN([ZFS_AC_KERNEL_MM_PAGE_SIZE], [
+	AC_MSG_CHECKING([whether page_size() is available])
+	ZFS_LINUX_TEST_RESULT([page_size], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_MM_PAGE_SIZE, 1, [page_size() is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
diff --git a/config/kernel-vfs-file_range.m4 b/config/kernel-vfs-file_range.m4
index cc96404d8b..8a5cbe2eee 100644
--- a/config/kernel-vfs-file_range.m4
+++ b/config/kernel-vfs-file_range.m4
@@ -16,6 +16,9 @@ dnl #
 dnl # 5.3: VFS copy_file_range() expected to do its own fallback,
 dnl #      generic_copy_file_range() added to support it
 dnl #
+dnl # 6.8: generic_copy_file_range() removed, replaced by
+dnl #      splice_copy_file_range()
+dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_COPY_FILE_RANGE], [
 	ZFS_LINUX_TEST_SRC([vfs_copy_file_range], [
 		#include <linux/fs.h>
@@ -72,6 +75,30 @@ AC_DEFUN([ZFS_AC_KERNEL_VFS_GENERIC_COPY_FILE_RANGE], [
 	])
 ])
 
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_SPLICE_COPY_FILE_RANGE], [
+	ZFS_LINUX_TEST_SRC([splice_copy_file_range], [
+		#include <linux/splice.h>
+	], [
+		struct file *src_file __attribute__ ((unused)) = NULL;
+		loff_t src_off __attribute__ ((unused)) = 0;
+		struct file *dst_file __attribute__ ((unused)) = NULL;
+		loff_t dst_off __attribute__ ((unused)) = 0;
+		size_t len __attribute__ ((unused)) = 0;
+		splice_copy_file_range(src_file, src_off, dst_file, dst_off,
+		    len);
+	])
+])
+AC_DEFUN([ZFS_AC_KERNEL_VFS_SPLICE_COPY_FILE_RANGE], [
+	AC_MSG_CHECKING([whether splice_copy_file_range() is available])
+	ZFS_LINUX_TEST_RESULT([splice_copy_file_range], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_VFS_SPLICE_COPY_FILE_RANGE, 1,
+		    [splice_copy_file_range() is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
 AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_CLONE_FILE_RANGE], [
 	ZFS_LINUX_TEST_SRC([vfs_clone_file_range], [
 		#include <linux/fs.h>
diff --git a/config/kernel.m4 b/config/kernel.m4
index e3f8645774..548905ccd0 100644
--- a/config/kernel.m4
+++ b/config/kernel.m4
@@ -118,6 +118,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_VFS_IOV_ITER
 	ZFS_AC_KERNEL_SRC_VFS_COPY_FILE_RANGE
 	ZFS_AC_KERNEL_SRC_VFS_GENERIC_COPY_FILE_RANGE
+	ZFS_AC_KERNEL_SRC_VFS_SPLICE_COPY_FILE_RANGE
 	ZFS_AC_KERNEL_SRC_VFS_REMAP_FILE_RANGE
 	ZFS_AC_KERNEL_SRC_VFS_CLONE_FILE_RANGE
 	ZFS_AC_KERNEL_SRC_VFS_DEDUPE_FILE_RANGE
@@ -166,6 +167,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_REGISTER_SYSCTL_TABLE
 	ZFS_AC_KERNEL_SRC_COPY_SPLICE_READ
 	ZFS_AC_KERNEL_SRC_SYNC_BDEV
+	ZFS_AC_KERNEL_SRC_MM_PAGE_SIZE
 	case "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_SRC_CPU_HAS_FEATURE
@@ -266,6 +268,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_VFS_IOV_ITER
 	ZFS_AC_KERNEL_VFS_COPY_FILE_RANGE
 	ZFS_AC_KERNEL_VFS_GENERIC_COPY_FILE_RANGE
+	ZFS_AC_KERNEL_VFS_SPLICE_COPY_FILE_RANGE
 	ZFS_AC_KERNEL_VFS_REMAP_FILE_RANGE
 	ZFS_AC_KERNEL_VFS_CLONE_FILE_RANGE
 	ZFS_AC_KERNEL_VFS_DEDUPE_FILE_RANGE
@@ -314,6 +317,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_REGISTER_SYSCTL_TABLE
 	ZFS_AC_KERNEL_COPY_SPLICE_READ
 	ZFS_AC_KERNEL_SYNC_BDEV
+	ZFS_AC_KERNEL_MM_PAGE_SIZE
 	case "$host_cpu" in
 		powerpc*)
 			ZFS_AC_KERNEL_CPU_HAS_FEATURE
diff --git a/include/os/freebsd/zfs/sys/zfs_vfsops_os.h b/include/os/freebsd/zfs/sys/zfs_vfsops_os.h
index 56a0ac96ac..24bb03575f 100644
--- a/include/os/freebsd/zfs/sys/zfs_vfsops_os.h
+++ b/include/os/freebsd/zfs/sys/zfs_vfsops_os.h
@@ -286,7 +286,6 @@ typedef struct zfid_long {
 
 extern uint_t zfs_fsyncer_key;
 extern int zfs_super_owner;
-extern int zfs_bclone_enabled;
 
 extern void zfs_init(void);
 extern void zfs_fini(void);
diff --git a/include/os/linux/Makefile.am b/include/os/linux/Makefile.am
index 3830d198df..51c27132b4 100644
--- a/include/os/linux/Makefile.am
+++ b/include/os/linux/Makefile.am
@@ -5,6 +5,7 @@ kernel_linux_HEADERS = \
 	%D%/kernel/linux/compiler_compat.h \
 	%D%/kernel/linux/dcache_compat.h \
 	%D%/kernel/linux/kmap_compat.h \
+	%D%/kernel/linux/mm_compat.h \
 	%D%/kernel/linux/mod_compat.h \
 	%D%/kernel/linux/page_compat.h \
 	%D%/kernel/linux/percpu_compat.h \
diff --git a/include/os/linux/kernel/linux/mm_compat.h b/include/os/linux/kernel/linux/mm_compat.h
new file mode 100644
index 0000000000..40056c68d6
--- /dev/null
+++ b/include/os/linux/kernel/linux/mm_compat.h
@@ -0,0 +1,36 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or https://opensource.org/licenses/CDDL-1.0.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright (c) 2023, 2024, Klara Inc.
+ */
+
+#ifndef _ZFS_MM_COMPAT_H
+#define	_ZFS_MM_COMPAT_H
+
+#include <linux/mm.h>
+
+/* 5.4 introduced page_size(). Older kernels can use a trivial macro instead */
+#ifndef HAVE_MM_PAGE_SIZE
+#define	page_size(p) ((unsigned long)(PAGE_SIZE << compound_order(p)))
+#endif
+
+#endif /* _ZFS_MM_COMPAT_H */
diff --git a/include/os/linux/kernel/linux/mod_compat.h b/include/os/linux/kernel/linux/mod_compat.h
index 8e20a96135..039865b703 100644
--- a/include/os/linux/kernel/linux/mod_compat.h
+++ b/include/os/linux/kernel/linux/mod_compat.h
@@ -68,6 +68,7 @@ enum scope_prefix_types {
 	zfs_trim,
 	zfs_txg,
 	zfs_vdev,
+	zfs_vdev_disk,
 	zfs_vdev_file,
 	zfs_vdev_mirror,
 	zfs_vnops,
diff --git a/include/os/linux/zfs/sys/zfs_vfsops_os.h b/include/os/linux/zfs/sys/zfs_vfsops_os.h
index 2204665502..b4d5db21f5 100644
--- a/include/os/linux/zfs/sys/zfs_vfsops_os.h
+++ b/include/os/linux/zfs/sys/zfs_vfsops_os.h
@@ -45,8 +45,6 @@ extern "C" {
 typedef struct zfsvfs zfsvfs_t;
 struct znode;
 
-extern int zfs_bclone_enabled;
-
 /*
  * This structure emulates the vfs_t from other platforms.  It's purpose
  * is to facilitate the handling of mount options and minimize structural
diff --git a/include/sys/abd.h b/include/sys/abd.h
index 750f9986c1..bee38b831b 100644
--- a/include/sys/abd.h
+++ b/include/sys/abd.h
@@ -79,6 +79,9 @@ typedef struct abd {
 
 typedef int abd_iter_func_t(void *buf, size_t len, void *priv);
 typedef int abd_iter_func2_t(void *bufa, void *bufb, size_t len, void *priv);
+#if defined(__linux__) && defined(_KERNEL)
+typedef int abd_iter_page_func_t(struct page *, size_t, size_t, void *);
+#endif
 
 extern int zfs_abd_scatter_enabled;
 
@@ -125,6 +128,10 @@ void abd_release_ownership_of_buf(abd_t *);
 int abd_iterate_func(abd_t *, size_t, size_t, abd_iter_func_t *, void *);
 int abd_iterate_func2(abd_t *, abd_t *, size_t, size_t, size_t,
     abd_iter_func2_t *, void *);
+#if defined(__linux__) && defined(_KERNEL)
+int abd_iterate_page_func(abd_t *, size_t, size_t, abd_iter_page_func_t *,
+    void *);
+#endif
 void abd_copy_off(abd_t *, abd_t *, size_t, size_t, size_t);
 void abd_copy_from_buf_off(abd_t *, const void *, size_t, size_t);
 void abd_copy_to_buf_off(void *, abd_t *, size_t, size_t);
@@ -213,6 +220,8 @@ void abd_fini(void);
 
 /*
  * Linux ABD bio functions
+ * Note: these are only needed to support vdev_classic. See comment in
+ * vdev_disk.c.
  */
 #if defined(__linux__) && defined(_KERNEL)
 unsigned int abd_bio_map_off(struct bio *, abd_t *, unsigned int, size_t);
diff --git a/include/sys/abd_impl.h b/include/sys/abd_impl.h
index 40546d4af1..f88ea25e24 100644
--- a/include/sys/abd_impl.h
+++ b/include/sys/abd_impl.h
@@ -21,6 +21,7 @@
 /*
  * Copyright (c) 2014 by Chunwei Chen. All rights reserved.
  * Copyright (c) 2016, 2019 by Delphix. All rights reserved.
+ * Copyright (c) 2023, 2024, Klara Inc.
  */
 
 #ifndef _ABD_IMPL_H
@@ -38,12 +39,30 @@ typedef enum abd_stats_op {
 	ABDSTAT_DECR  /* Decrease abdstat values */
 } abd_stats_op_t;
 
-struct scatterlist; /* forward declaration */
+/* forward declarations */
+struct scatterlist;
+struct page;
 
 struct abd_iter {
 	/* public interface */
-	void		*iter_mapaddr;	/* addr corresponding to iter_pos */
-	size_t		iter_mapsize;	/* length of data valid at mapaddr */
+	union {
+		/* for abd_iter_map()/abd_iter_unmap() */
+		struct {
+			/* addr corresponding to iter_pos */
+			void		*iter_mapaddr;
+			/* length of data valid at mapaddr */
+			size_t		iter_mapsize;
+		};
+		/* for abd_iter_page() */
+		struct {
+			/* current page */
+			struct page	*iter_page;
+			/* offset of data in page */
+			size_t		iter_page_doff;
+			/* size of data in page */
+			size_t		iter_page_dsize;
+		};
+	};
 
 	/* private */
 	abd_t		*iter_abd;	/* ABD being iterated through */
@@ -78,6 +97,7 @@ boolean_t abd_iter_at_end(struct abd_iter *);
 void abd_iter_advance(struct abd_iter *, size_t);
 void abd_iter_map(struct abd_iter *);
 void abd_iter_unmap(struct abd_iter *);
+void abd_iter_page(struct abd_iter *);
 
 /*
  * Helper macros
diff --git a/include/sys/zfs_vnops.h b/include/sys/zfs_vnops.h
index 5da103f177..e60b99bed1 100644
--- a/include/sys/zfs_vnops.h
+++ b/include/sys/zfs_vnops.h
@@ -24,8 +24,11 @@
 
 #ifndef	_SYS_FS_ZFS_VNOPS_H
 #define	_SYS_FS_ZFS_VNOPS_H
+
 #include <sys/zfs_vnops_os.h>
 
+extern int zfs_bclone_enabled;
+
 extern int zfs_fsync(znode_t *, int, cred_t *);
 extern int zfs_read(znode_t *, zfs_uio_t *, int, cred_t *);
 extern int zfs_write(znode_t *, zfs_uio_t *, int, cred_t *);
diff --git a/man/man4/zfs.4 b/man/man4/zfs.4
index 9d51348912..242f62e5e2 100644
--- a/man/man4/zfs.4
+++ b/man/man4/zfs.4
@@ -2,6 +2,7 @@
 .\" Copyright (c) 2013 by Turbo Fredriksson <turbo@bayour.com>. All rights reserved.
 .\" Copyright (c) 2019, 2021 by Delphix. All rights reserved.
 .\" Copyright (c) 2019 Datto Inc.
+.\" Copyright (c) 2023, 2024 Klara, Inc.
 .\" The contents of this file are subject to the terms of the Common Development
 .\" and Distribution License (the "License").  You may not use this file except
 .\" in compliance with the License. You can obtain a copy of the license at
@@ -15,7 +16,7 @@
 .\" own identifying information:
 .\" Portions Copyright [yyyy] [name of copyright owner]
 .\"
-.Dd July 21, 2023
+.Dd January 9, 2024
 .Dt ZFS 4
 .Os
 .
@@ -1142,6 +1143,15 @@ Enable the experimental block cloning feature.
 If this setting is 0, then even if feature@block_cloning is enabled,
 attempts to clone blocks will act as though the feature is disabled.
 .
+.It Sy zfs_bclone_wait_dirty Ns = Ns Sy 0 Ns | Ns 1 Pq int
+When set to 1 the FICLONE and FICLONERANGE ioctls wait for dirty data to be
+written to disk.
+This allows the clone operation to reliably succeed when a file is
+modified and then immediately cloned.
+For small files this may be slower than making a copy of the file.
+Therefore, this setting defaults to 0 which causes a clone operation to
+immediately fail when encountering a dirty block.
+.
 .It Sy zfs_blake3_impl Ns = Ns Sy fastest Pq string
 Select a BLAKE3 implementation.
 .Pp
@@ -1336,6 +1346,42 @@ _
 	4	Driver	No driver retries on driver errors.
 .TE
 .
+.It Sy zfs_vdev_disk_max_segs Ns = Ns Sy 0 Pq uint
+Maximum number of segments to add to a BIO (min 4).
+If this is higher than the maximum allowed by the device queue or the kernel
+itself, it will be clamped.
+Setting it to zero will cause the kernel's ideal size to be used.
+This parameter only applies on Linux.
+This parameter is ignored if
+.Sy zfs_vdev_disk_classic Ns = Ns Sy 1 .
+.
+.It Sy zfs_vdev_disk_classic Ns = Ns 0 Ns | Ns Sy 1 Pq uint
+Controls the method used to submit IO to the Linux block layer
+(default
+.Sy 1 "classic" Ns
+)
+.Pp
+If set to 1, the "classic" method is used.
+This is the method that has been in use since the earliest versions of
+ZFS-on-Linux.
+It has known issues with highly fragmented IO requests and is less efficient on
+many workloads, but it well known and well understood.
+.Pp
+If set to 0, the "new" method is used.
+This method is available since 2.2.4 and should resolve all known issues and be
+far more efficient, but has not had as much testing.
+In the 2.2.x series, this parameter defaults to 1, to use the "classic" method.
+.Pp
+It is not recommended that you change it except on advice from the OpenZFS
+developers.
+If you do change it, please also open a bug report describing why you did so,
+including the workload involved and any error messages.
+.Pp
+This parameter and the "classic" submission method will be removed in a future
+release of OpenZFS once we have total confidence in the new method.
+.Pp
+This parameter only applies on Linux, and can only be set at module load time.
+.
 .It Sy zfs_expire_snapshot Ns = Ns Sy 300 Ns s Pq int
 Time before expiring
 .Pa .zfs/snapshot .
diff --git a/man/man8/zfs-list.8 b/man/man8/zfs-list.8
index 9f6a73ab95..85bd3fbafc 100644
--- a/man/man8/zfs-list.8
+++ b/man/man8/zfs-list.8
@@ -29,7 +29,7 @@
 .\" Copyright 2018 Nexenta Systems, Inc.
 .\" Copyright 2019 Joyent, Inc.
 .\"
-.Dd March 16, 2022
+.Dd February 8, 2024
 .Dt ZFS-LIST 8
 .Os
 .
@@ -155,6 +155,15 @@ or
 For example, specifying
 .Fl t Sy snapshot
 displays only snapshots.
+.Sy fs ,
+.Sy snap ,
+or
+.Sy vol
+can be used as aliases for
+.Sy filesystem ,
+.Sy snapshot ,
+or
+.Sy volume .
 .El
 .
 .Sh EXAMPLES
diff --git a/man/man8/zpool-status.8 b/man/man8/zpool-status.8
index 56fa4aed05..24ad6e643c 100644
--- a/man/man8/zpool-status.8
+++ b/man/man8/zpool-status.8
@@ -36,7 +36,7 @@
 .Sh SYNOPSIS
 .Nm zpool
 .Cm status
-.Op Fl DigLpPstvx
+.Op Fl DeigLpPstvx
 .Op Fl T Sy u Ns | Ns Sy d
 .Op Fl c Op Ar SCRIPT1 Ns Oo , Ns Ar SCRIPT2 Oc Ns …
 .Oo Ar pool Oc Ns …
@@ -69,6 +69,8 @@ See the
 option of
 .Nm zpool Cm iostat
 for complete details.
+.It Fl e
+Only show unhealthy vdevs (not-ONLINE or with errors).
 .It Fl i
 Display vdev initialization status.
 .It Fl g
diff --git a/module/lua/ldebug.c b/module/lua/ldebug.c
index 0092474c76..23e321bb12 100644
--- a/module/lua/ldebug.c
+++ b/module/lua/ldebug.c
@@ -111,10 +111,11 @@ static const char *upvalname (Proto *p, int uv) {
 
 static const char *findvararg (CallInfo *ci, int n, StkId *pos) {
   int nparams = clLvalue(ci->func)->p->numparams;
-  if (n >= ci->u.l.base - ci->func - nparams)
+  int nvararg = cast_int(ci->u.l.base - ci->func) - nparams;
+  if (n <= -nvararg)
     return NULL;  /* no such vararg */
   else {
-    *pos = ci->func + nparams + n;
+    *pos = ci->func + nparams - n;
     return "(*vararg)";  /* generic name for any vararg */
   }
 }
@@ -126,7 +127,7 @@ static const char *findlocal (lua_State *L, CallInfo *ci, int n,
   StkId base;
   if (isLua(ci)) {
     if (n < 0)  /* access to vararg values? */
-      return findvararg(ci, -n, pos);
+      return findvararg(ci, n, pos);
     else {
       base = ci->u.l.base;
       name = luaF_getlocalname(ci_func(ci)->p, n, currentpc(ci));
diff --git a/module/os/freebsd/zfs/abd_os.c b/module/os/freebsd/zfs/abd_os.c
index 58a37df62b..3b812271f9 100644
--- a/module/os/freebsd/zfs/abd_os.c
+++ b/module/os/freebsd/zfs/abd_os.c
@@ -417,10 +417,8 @@ abd_iter_init(struct abd_iter *aiter, abd_t *abd)
 {
 	ASSERT(!abd_is_gang(abd));
 	abd_verify(abd);
+	memset(aiter, 0, sizeof (struct abd_iter));
 	aiter->iter_abd = abd;
-	aiter->iter_pos = 0;
-	aiter->iter_mapaddr = NULL;
-	aiter->iter_mapsize = 0;
 }
 
 /*
diff --git a/module/os/freebsd/zfs/zfs_vfsops.c b/module/os/freebsd/zfs/zfs_vfsops.c
index 23b8da1845..a972c720df 100644
--- a/module/os/freebsd/zfs/zfs_vfsops.c
+++ b/module/os/freebsd/zfs/zfs_vfsops.c
@@ -89,10 +89,6 @@ int zfs_debug_level;
 SYSCTL_INT(_vfs_zfs, OID_AUTO, debug, CTLFLAG_RWTUN, &zfs_debug_level, 0,
 	"Debug level");
 
-int zfs_bclone_enabled = 0;
-SYSCTL_INT(_vfs_zfs, OID_AUTO, bclone_enabled, CTLFLAG_RWTUN,
-	&zfs_bclone_enabled, 0, "Enable block cloning");
-
 struct zfs_jailparam {
 	int mount_snapshot;
 };
diff --git a/module/os/linux/zfs/abd_os.c b/module/os/linux/zfs/abd_os.c
index 24390fbbf1..d3255dcbc0 100644
--- a/module/os/linux/zfs/abd_os.c
+++ b/module/os/linux/zfs/abd_os.c
@@ -21,6 +21,7 @@
 /*
  * Copyright (c) 2014 by Chunwei Chen. All rights reserved.
  * Copyright (c) 2019 by Delphix. All rights reserved.
+ * Copyright (c) 2023, 2024, Klara Inc.
  */
 
 /*
@@ -59,7 +60,9 @@
 #include <sys/zfs_znode.h>
 #ifdef _KERNEL
 #include <linux/kmap_compat.h>
+#include <linux/mm_compat.h>
 #include <linux/scatterlist.h>
+#include <linux/version.h>
 #endif
 
 #ifdef _KERNEL
@@ -895,14 +898,9 @@ abd_iter_init(struct abd_iter *aiter, abd_t *abd)
 {
 	ASSERT(!abd_is_gang(abd));
 	abd_verify(abd);
+	memset(aiter, 0, sizeof (struct abd_iter));
 	aiter->iter_abd = abd;
-	aiter->iter_mapaddr = NULL;
-	aiter->iter_mapsize = 0;
-	aiter->iter_pos = 0;
-	if (abd_is_linear(abd)) {
-		aiter->iter_offset = 0;
-		aiter->iter_sg = NULL;
-	} else {
+	if (!abd_is_linear(abd)) {
 		aiter->iter_offset = ABD_SCATTER(abd).abd_offset;
 		aiter->iter_sg = ABD_SCATTER(abd).abd_sgl;
 	}
@@ -915,6 +913,7 @@ abd_iter_init(struct abd_iter *aiter, abd_t *abd)
 boolean_t
 abd_iter_at_end(struct abd_iter *aiter)
 {
+	ASSERT3U(aiter->iter_pos, <=, aiter->iter_abd->abd_size);
 	return (aiter->iter_pos == aiter->iter_abd->abd_size);
 }
 
@@ -926,8 +925,15 @@ abd_iter_at_end(struct abd_iter *aiter)
 void
 abd_iter_advance(struct abd_iter *aiter, size_t amount)
 {
+	/*
+	 * Ensure that last chunk is not in use. abd_iterate_*() must clear
+	 * this state (directly or abd_iter_unmap()) before advancing.
+	 */
 	ASSERT3P(aiter->iter_mapaddr, ==, NULL);
 	ASSERT0(aiter->iter_mapsize);
+	ASSERT3P(aiter->iter_page, ==, NULL);
+	ASSERT0(aiter->iter_page_doff);
+	ASSERT0(aiter->iter_page_dsize);
 
 	/* There's nothing left to advance to, so do nothing */
 	if (abd_iter_at_end(aiter))
@@ -1009,6 +1015,106 @@ abd_cache_reap_now(void)
 }
 
 #if defined(_KERNEL)
+/*
+ * Yield the next page struct and data offset and size within it, without
+ * mapping it into the address space.
+ */
+void
+abd_iter_page(struct abd_iter *aiter)
+{
+	if (abd_iter_at_end(aiter)) {
+		aiter->iter_page = NULL;
+		aiter->iter_page_doff = 0;
+		aiter->iter_page_dsize = 0;
+		return;
+	}
+
+	struct page *page;
+	size_t doff, dsize;
+
+	if (abd_is_linear(aiter->iter_abd)) {
+		ASSERT3U(aiter->iter_pos, ==, aiter->iter_offset);
+
+		/* memory address at iter_pos */
+		void *paddr = ABD_LINEAR_BUF(aiter->iter_abd) + aiter->iter_pos;
+
+		/* struct page for address */
+		page = is_vmalloc_addr(paddr) ?
+		    vmalloc_to_page(paddr) : virt_to_page(paddr);
+
+		/* offset of address within the page */
+		doff = offset_in_page(paddr);
+
+		/* total data remaining in abd from this position */
+		dsize = aiter->iter_abd->abd_size - aiter->iter_offset;
+	} else {
+		ASSERT(!abd_is_gang(aiter->iter_abd));
+
+		/* current scatter page */
+		page = sg_page(aiter->iter_sg);
+
+		/* position within page */
+		doff = aiter->iter_offset;
+
+		/* remaining data in scatterlist */
+		dsize = MIN(aiter->iter_sg->length - aiter->iter_offset,
+		    aiter->iter_abd->abd_size - aiter->iter_pos);
+	}
+	ASSERT(page);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 5, 0)
+	if (PageTail(page)) {
+		/*
+		 * This page is part of a "compound page", which is a group of
+		 * pages that can be referenced from a single struct page *.
+		 * Its organised as a "head" page, followed by a series of
+		 * "tail" pages.
+		 *
+		 * In OpenZFS, compound pages are allocated using the
+		 * __GFP_COMP flag, which we get from scatter ABDs and SPL
+		 * vmalloc slabs (ie >16K allocations). So a great many of the
+		 * IO buffers we get are going to be of this type.
+		 *
+		 * The tail pages are just regular PAGE_SIZE pages, and can be
+		 * safely used as-is. However, the head page has length
+		 * covering itself and all the tail pages. If this ABD chunk
+		 * spans multiple pages, then we can use the head page and a
+		 * >PAGE_SIZE length, which is far more efficient.
+		 *
+		 * To do this, we need to adjust the offset to be counted from
+		 * the head page. struct page for compound pages are stored
+		 * contiguously, so we can just adjust by a simple offset.
+		 *
+		 * Before kernel 4.5, compound page heads were refcounted
+		 * separately, such that moving back to the head page would
+		 * require us to take a reference to it and releasing it once
+		 * we're completely finished with it. In practice, that means
+		 * when our caller is done with the ABD, which we have no
+		 * insight into from here. Rather than contort this API to
+		 * track head page references on such ancient kernels, we just
+		 * compile this block out and use the tail pages directly. This
+		 * is slightly less efficient, but makes everything far
+		 * simpler.
+		 */
+		struct page *head = compound_head(page);
+		doff += ((page - head) * PAGESIZE);
+		page = head;
+	}
+#endif
+
+	/* final page and position within it */
+	aiter->iter_page = page;
+	aiter->iter_page_doff = doff;
+
+	/* amount of data in the chunk, up to the end of the page */
+	aiter->iter_page_dsize = MIN(dsize, page_size(page) - doff);
+}
+
+/*
+ * Note: ABD BIO functions only needed to support vdev_classic. See comments in
+ * vdev_disk.c.
+ */
+
 /*
  * bio_nr_pages for ABD.
  * @off is the offset in @abd
@@ -1163,4 +1269,5 @@ MODULE_PARM_DESC(zfs_abd_scatter_min_size,
 module_param(zfs_abd_scatter_max_order, uint, 0644);
 MODULE_PARM_DESC(zfs_abd_scatter_max_order,
 	"Maximum order allocation used for a scatter ABD.");
-#endif
+
+#endif /* _KERNEL */
diff --git a/module/os/linux/zfs/vdev_disk.c b/module/os/linux/zfs/vdev_disk.c
index b0bda5fa20..e1c19a085b 100644
--- a/module/os/linux/zfs/vdev_disk.c
+++ b/module/os/linux/zfs/vdev_disk.c
@@ -24,6 +24,7 @@
  * Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
  * LLNL-CODE-403049.
  * Copyright (c) 2012, 2019 by Delphix. All rights reserved.
+ * Copyright (c) 2023, 2024, Klara Inc.
  */
 
 #include <sys/zfs_context.h>
@@ -66,6 +67,13 @@ typedef struct vdev_disk {
 	krwlock_t			vd_lock;
 } vdev_disk_t;
 
+/*
+ * Maximum number of segments to add to a bio (min 4). If this is higher than
+ * the maximum allowed by the device queue or the kernel itself, it will be
+ * clamped. Setting it to zero will cause the kernel's ideal size to be used.
+ */
+uint_t zfs_vdev_disk_max_segs = 0;
+
 /*
  * Unique identifier for the exclusive vdev holder.
  */
@@ -83,17 +91,6 @@ static uint_t zfs_vdev_open_timeout_ms = 1000;
  */
 #define	EFI_MIN_RESV_SIZE	(16 * 1024)
 
-/*
- * Virtual device vector for disks.
- */
-typedef struct dio_request {
-	zio_t			*dr_zio;	/* Parent ZIO */
-	atomic_t		dr_ref;		/* References */
-	int			dr_error;	/* Bio error */
-	int			dr_bio_count;	/* Count of bio's */
-	struct bio		*dr_bio[];	/* Attached bio's */
-} dio_request_t;
-
 /*
  * BIO request failfast mask.
  */
@@ -457,95 +454,15 @@ vdev_disk_close(vdev_t *v)
 	if (v->vdev_reopening || vd == NULL)
 		return;
 
-	if (vd->vd_bdh != NULL) {
+	if (vd->vd_bdh != NULL)
 		vdev_blkdev_put(vd->vd_bdh, spa_mode(v->vdev_spa),
 		    zfs_vdev_holder);
-	}
 
 	rw_destroy(&vd->vd_lock);
 	kmem_free(vd, sizeof (vdev_disk_t));
 	v->vdev_tsd = NULL;
 }
 
-static dio_request_t *
-vdev_disk_dio_alloc(int bio_count)
-{
-	dio_request_t *dr = kmem_zalloc(sizeof (dio_request_t) +
-	    sizeof (struct bio *) * bio_count, KM_SLEEP);
-	atomic_set(&dr->dr_ref, 0);
-	dr->dr_bio_count = bio_count;
-	dr->dr_error = 0;
-
-	for (int i = 0; i < dr->dr_bio_count; i++)
-		dr->dr_bio[i] = NULL;
-
-	return (dr);
-}
-
-static void
-vdev_disk_dio_free(dio_request_t *dr)
-{
-	int i;
-
-	for (i = 0; i < dr->dr_bio_count; i++)
-		if (dr->dr_bio[i])
-			bio_put(dr->dr_bio[i]);
-
-	kmem_free(dr, sizeof (dio_request_t) +
-	    sizeof (struct bio *) * dr->dr_bio_count);
-}
-
-static void
-vdev_disk_dio_get(dio_request_t *dr)
-{
-	atomic_inc(&dr->dr_ref);
-}
-
-static void
-vdev_disk_dio_put(dio_request_t *dr)
-{
-	int rc = atomic_dec_return(&dr->dr_ref);
-
-	/*
-	 * Free the dio_request when the last reference is dropped and
-	 * ensure zio_interpret is called only once with the correct zio
-	 */
-	if (rc == 0) {
-		zio_t *zio = dr->dr_zio;
-		int error = dr->dr_error;
-
-		vdev_disk_dio_free(dr);
-
-		if (zio) {
-			zio->io_error = error;
-			ASSERT3S(zio->io_error, >=, 0);
-			if (zio->io_error)
-				vdev_disk_error(zio);
-
-			zio_delay_interrupt(zio);
-		}
-	}
-}
-
-BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, error)
-{
-	dio_request_t *dr = bio->bi_private;
-
-	if (dr->dr_error == 0) {
-#ifdef HAVE_1ARG_BIO_END_IO_T
-		dr->dr_error = BIO_END_IO_ERROR(bio);
-#else
-		if (error)
-			dr->dr_error = -(error);
-		else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
-			dr->dr_error = EIO;
-#endif
-	}
-
-	/* Drop reference acquired by __vdev_disk_physio */
-	vdev_disk_dio_put(dr);
-}
-
 static inline void
 vdev_submit_bio_impl(struct bio *bio)
 {
@@ -697,8 +614,462 @@ vdev_bio_alloc(struct block_device *bdev, gfp_t gfp_mask,
 	return (bio);
 }
 
+static inline uint_t
+vdev_bio_max_segs(struct block_device *bdev)
+{
+	/*
+	 * Smallest of the device max segs and the tuneable max segs. Minimum
+	 * 4, so there's room to finish split pages if they come up.
+	 */
+	const uint_t dev_max_segs = queue_max_segments(bdev_get_queue(bdev));
+	const uint_t tune_max_segs = (zfs_vdev_disk_max_segs > 0) ?
+	    MAX(4, zfs_vdev_disk_max_segs) : dev_max_segs;
+	const uint_t max_segs = MIN(tune_max_segs, dev_max_segs);
+
+#ifdef HAVE_BIO_MAX_SEGS
+	return (bio_max_segs(max_segs));
+#else
+	return (MIN(max_segs, BIO_MAX_PAGES));
+#endif
+}
+
+static inline uint_t
+vdev_bio_max_bytes(struct block_device *bdev)
+{
+	return (queue_max_sectors(bdev_get_queue(bdev)) << 9);
+}
+
+
+/*
+ * Virtual block IO object (VBIO)
+ *
+ * Linux block IO (BIO) objects have a limit on how many data segments (pages)
+ * they can hold. Depending on how they're allocated and structured, a large
+ * ZIO can require more than one BIO to be submitted to the kernel, which then
+ * all have to complete before we can return the completed ZIO back to ZFS.
+ *
+ * A VBIO is a wrapper around multiple BIOs, carrying everything needed to
+ * translate a ZIO down into the kernel block layer and back again.
+ *
+ * Note that these are only used for data ZIOs (read/write). Meta-operations
+ * (flush/trim) don't need multiple BIOs and so can just make the call
+ * directly.
+ */
+typedef struct {
+	zio_t		*vbio_zio;	/* parent zio */
+
+	struct block_device *vbio_bdev;	/* blockdev to submit bios to */
+
+	abd_t		*vbio_abd;	/* abd carrying borrowed linear buf */
+
+	uint_t		vbio_max_segs;	/* max segs per bio */
+
+	uint_t		vbio_max_bytes;	/* max bytes per bio */
+	uint_t		vbio_lbs_mask;	/* logical block size mask */
+
+	uint64_t	vbio_offset;	/* start offset of next bio */
+
+	struct bio	*vbio_bio;	/* pointer to the current bio */
+	int		vbio_flags;	/* bio flags */
+} vbio_t;
+
+static vbio_t *
+vbio_alloc(zio_t *zio, struct block_device *bdev, int flags)
+{
+	vbio_t *vbio = kmem_zalloc(sizeof (vbio_t), KM_SLEEP);
+
+	vbio->vbio_zio = zio;
+	vbio->vbio_bdev = bdev;
+	vbio->vbio_abd = NULL;
+	vbio->vbio_max_segs = vdev_bio_max_segs(bdev);
+	vbio->vbio_max_bytes = vdev_bio_max_bytes(bdev);
+	vbio->vbio_lbs_mask = ~(bdev_logical_block_size(bdev)-1);
+	vbio->vbio_offset = zio->io_offset;
+	vbio->vbio_bio = NULL;
+	vbio->vbio_flags = flags;
+
+	return (vbio);
+}
+
+BIO_END_IO_PROTO(vbio_completion, bio, error);
+
+static int
+vbio_add_page(vbio_t *vbio, struct page *page, uint_t size, uint_t offset)
+{
+	struct bio *bio = vbio->vbio_bio;
+	uint_t ssize;
+
+	while (size > 0) {
+		if (bio == NULL) {
+			/* New BIO, allocate and set up */
+			bio = vdev_bio_alloc(vbio->vbio_bdev, GFP_NOIO,
+			    vbio->vbio_max_segs);
+			VERIFY(bio);
+
+			BIO_BI_SECTOR(bio) = vbio->vbio_offset >> 9;
+			bio_set_op_attrs(bio,
+			    vbio->vbio_zio->io_type == ZIO_TYPE_WRITE ?
+			    WRITE : READ, vbio->vbio_flags);
+
+			if (vbio->vbio_bio) {
+				bio_chain(vbio->vbio_bio, bio);
+				vdev_submit_bio(vbio->vbio_bio);
+			}
+			vbio->vbio_bio = bio;
+		}
+
+		/*
+		 * Only load as much of the current page data as will fit in
+		 * the space left in the BIO, respecting lbs alignment. Older
+		 * kernels will error if we try to overfill the BIO, while
+		 * newer ones will accept it and split the BIO. This ensures
+		 * everything works on older kernels, and avoids an additional
+		 * overhead on the new.
+		 */
+		ssize = MIN(size, (vbio->vbio_max_bytes - BIO_BI_SIZE(bio)) &
+		    vbio->vbio_lbs_mask);
+		if (ssize > 0 &&
+		    bio_add_page(bio, page, ssize, offset) == ssize) {
+			/* Accepted, adjust and load any remaining. */
+			size -= ssize;
+			offset += ssize;
+			continue;
+		}
+
+		/* No room, set up for a new BIO and loop */
+		vbio->vbio_offset += BIO_BI_SIZE(bio);
+
+		/* Signal new BIO allocation wanted */
+		bio = NULL;
+	}
+
+	return (0);
+}
+
+/* Iterator callback to submit ABD pages to the vbio. */
+static int
+vbio_fill_cb(struct page *page, size_t off, size_t len, void *priv)
+{
+	vbio_t *vbio = priv;
+	return (vbio_add_page(vbio, page, len, off));
+}
+
+/* Create some BIOs, fill them with data and submit them */
+static void
+vbio_submit(vbio_t *vbio, abd_t *abd, uint64_t size)
+{
+	ASSERT(vbio->vbio_bdev);
+
+	/*
+	 * We plug so we can submit the BIOs as we go and only unplug them when
+	 * they are fully created and submitted. This is important; if we don't
+	 * plug, then the kernel may start executing earlier BIOs while we're
+	 * still creating and executing later ones, and if the device goes
+	 * away while that's happening, older kernels can get confused and
+	 * trample memory.
+	 */
+	struct blk_plug plug;
+	blk_start_plug(&plug);
+
+	(void) abd_iterate_page_func(abd, 0, size, vbio_fill_cb, vbio);
+	ASSERT(vbio->vbio_bio);
+
+	vbio->vbio_bio->bi_end_io = vbio_completion;
+	vbio->vbio_bio->bi_private = vbio;
+
+	vdev_submit_bio(vbio->vbio_bio);
+
+	blk_finish_plug(&plug);
+
+	vbio->vbio_bio = NULL;
+	vbio->vbio_bdev = NULL;
+}
+
+/* IO completion callback */
+BIO_END_IO_PROTO(vbio_completion, bio, error)
+{
+	vbio_t *vbio = bio->bi_private;
+	zio_t *zio = vbio->vbio_zio;
+
+	ASSERT(zio);
+
+	/* Capture and log any errors */
+#ifdef HAVE_1ARG_BIO_END_IO_T
+	zio->io_error = BIO_END_IO_ERROR(bio);
+#else
+	zio->io_error = 0;
+	if (error)
+		zio->io_error = -(error);
+	else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
+		zio->io_error = EIO;
+#endif
+	ASSERT3U(zio->io_error, >=, 0);
+
+	if (zio->io_error)
+		vdev_disk_error(zio);
+
+	/* Return the BIO to the kernel */
+	bio_put(bio);
+
+	/*
+	 * If we copied the ABD before issuing it, clean up and return the copy
+	 * to the ADB, with changes if appropriate.
+	 */
+	if (vbio->vbio_abd != NULL) {
+		void *buf = abd_to_buf(vbio->vbio_abd);
+		abd_free(vbio->vbio_abd);
+		vbio->vbio_abd = NULL;
+
+		if (zio->io_type == ZIO_TYPE_READ)
+			abd_return_buf_copy(zio->io_abd, buf, zio->io_size);
+		else
+			abd_return_buf(zio->io_abd, buf, zio->io_size);
+	}
+
+	/* Final cleanup */
+	kmem_free(vbio, sizeof (vbio_t));
+
+	/* All done, submit for processing */
+	zio_delay_interrupt(zio);
+}
+
+/*
+ * Iterator callback to count ABD pages and check their size & alignment.
+ *
+ * On Linux, each BIO segment can take a page pointer, and an offset+length of
+ * the data within that page. A page can be arbitrarily large ("compound"
+ * pages) but we still have to ensure the data portion is correctly sized and
+ * aligned to the logical block size, to ensure that if the kernel wants to
+ * split the BIO, the two halves will still be properly aligned.
+ */
+typedef struct {
+	uint_t  bmask;
+	uint_t  npages;
+	uint_t  end;
+} vdev_disk_check_pages_t;
+
+static int
+vdev_disk_check_pages_cb(struct page *page, size_t off, size_t len, void *priv)
+{
+	vdev_disk_check_pages_t *s = priv;
+
+	/*
+	 * If we didn't finish on a block size boundary last time, then there
+	 * would be a gap if we tried to use this ABD as-is, so abort.
+	 */
+	if (s->end != 0)
+		return (1);
+
+	/*
+	 * Note if we're taking less than a full block, so we can check it
+	 * above on the next call.
+	 */
+	s->end = len & s->bmask;
+
+	/* All blocks after the first must start on a block size boundary. */
+	if (s->npages != 0 && (off & s->bmask) != 0)
+		return (1);
+
+	s->npages++;
+	return (0);
+}
+
+/*
+ * Check if we can submit the pages in this ABD to the kernel as-is. Returns
+ * the number of pages, or 0 if it can't be submitted like this.
+ */
+static boolean_t
+vdev_disk_check_pages(abd_t *abd, uint64_t size, struct block_device *bdev)
+{
+	vdev_disk_check_pages_t s = {
+	    .bmask = bdev_logical_block_size(bdev)-1,
+	    .npages = 0,
+	    .end = 0,
+	};
+
+	if (abd_iterate_page_func(abd, 0, size, vdev_disk_check_pages_cb, &s))
+		return (B_FALSE);
+
+	return (B_TRUE);
+}
+
+static int
+vdev_disk_io_rw(zio_t *zio)
+{
+	vdev_t *v = zio->io_vd;
+	vdev_disk_t *vd = v->vdev_tsd;
+	struct block_device *bdev = BDH_BDEV(vd->vd_bdh);
+	int flags = 0;
+
+	/*
+	 * Accessing outside the block device is never allowed.
+	 */
+	if (zio->io_offset + zio->io_size > bdev->bd_inode->i_size) {
+		vdev_dbgmsg(zio->io_vd,
+		    "Illegal access %llu size %llu, device size %llu",
+		    (u_longlong_t)zio->io_offset,
+		    (u_longlong_t)zio->io_size,
+		    (u_longlong_t)i_size_read(bdev->bd_inode));
+		return (SET_ERROR(EIO));
+	}
+
+	if (!(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD)) &&
+	    v->vdev_failfast == B_TRUE) {
+		bio_set_flags_failfast(bdev, &flags, zfs_vdev_failfast_mask & 1,
+		    zfs_vdev_failfast_mask & 2, zfs_vdev_failfast_mask & 4);
+	}
+
+	/*
+	 * Check alignment of the incoming ABD. If any part of it would require
+	 * submitting a page that is not aligned to the logical block size,
+	 * then we take a copy into a linear buffer and submit that instead.
+	 * This should be impossible on a 512b LBS, and fairly rare on 4K,
+	 * usually requiring abnormally-small data blocks (eg gang blocks)
+	 * mixed into the same ABD as larger ones (eg aggregated).
+	 */
+	abd_t *abd = zio->io_abd;
+	if (!vdev_disk_check_pages(abd, zio->io_size, bdev)) {
+		void *buf;
+		if (zio->io_type == ZIO_TYPE_READ)
+			buf = abd_borrow_buf(zio->io_abd, zio->io_size);
+		else
+			buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size);
+
+		/*
+		 * Wrap the copy in an abd_t, so we can use the same iterators
+		 * to count and fill the vbio later.
+		 */
+		abd = abd_get_from_buf(buf, zio->io_size);
+
+		/*
+		 * False here would mean the borrowed copy has an invalid
+		 * alignment too, which would mean we've somehow been passed a
+		 * linear ABD with an interior page that has a non-zero offset
+		 * or a size not a multiple of PAGE_SIZE. This is not possible.
+		 * It would mean either zio_buf_alloc() or its underlying
+		 * allocators have done something extremely strange, or our
+		 * math in vdev_disk_check_pages() is wrong. In either case,
+		 * something in seriously wrong and its not safe to continue.
+		 */
+		VERIFY(vdev_disk_check_pages(abd, zio->io_size, bdev));
+	}
+
+	/* Allocate vbio, with a pointer to the borrowed ABD if necessary */
+	vbio_t *vbio = vbio_alloc(zio, bdev, flags);
+	if (abd != zio->io_abd)
+		vbio->vbio_abd = abd;
+
+	/* Fill it with data pages and submit it to the kernel */
+	vbio_submit(vbio, abd, zio->io_size);
+	return (0);
+}
+
+/* ========== */
+
+/*
+ * This is the classic, battle-tested BIO submission code. Until we're totally
+ * sure that the new code is safe and correct in all cases, this will remain
+ * available.
+ *
+ * It is enabled by setting zfs_vdev_disk_classic=1 at module load time. It is
+ * enabled (=1) by default since 2.2.4, and disabled by default (=0) on master.
+ *
+ * These functions have been renamed to vdev_classic_* to make it clear what
+ * they belong to, but their implementations are unchanged.
+ */
+
+/*
+ * Virtual device vector for disks.
+ */
+typedef struct dio_request {
+	zio_t			*dr_zio;	/* Parent ZIO */
+	atomic_t		dr_ref;		/* References */
+	int			dr_error;	/* Bio error */
+	int			dr_bio_count;	/* Count of bio's */
+	struct bio		*dr_bio[];	/* Attached bio's */
+} dio_request_t;
+
+static dio_request_t *
+vdev_classic_dio_alloc(int bio_count)
+{
+	dio_request_t *dr = kmem_zalloc(sizeof (dio_request_t) +
+	    sizeof (struct bio *) * bio_count, KM_SLEEP);
+	atomic_set(&dr->dr_ref, 0);
+	dr->dr_bio_count = bio_count;
+	dr->dr_error = 0;
+
+	for (int i = 0; i < dr->dr_bio_count; i++)
+		dr->dr_bio[i] = NULL;
+
+	return (dr);
+}
+
+static void
+vdev_classic_dio_free(dio_request_t *dr)
+{
+	int i;
+
+	for (i = 0; i < dr->dr_bio_count; i++)
+		if (dr->dr_bio[i])
+			bio_put(dr->dr_bio[i]);
+
+	kmem_free(dr, sizeof (dio_request_t) +
+	    sizeof (struct bio *) * dr->dr_bio_count);
+}
+
+static void
+vdev_classic_dio_get(dio_request_t *dr)
+{
+	atomic_inc(&dr->dr_ref);
+}
+
+static void
+vdev_classic_dio_put(dio_request_t *dr)
+{
+	int rc = atomic_dec_return(&dr->dr_ref);
+
+	/*
+	 * Free the dio_request when the last reference is dropped and
+	 * ensure zio_interpret is called only once with the correct zio
+	 */
+	if (rc == 0) {
+		zio_t *zio = dr->dr_zio;
+		int error = dr->dr_error;
+
+		vdev_classic_dio_free(dr);
+
+		if (zio) {
+			zio->io_error = error;
+			ASSERT3S(zio->io_error, >=, 0);
+			if (zio->io_error)
+				vdev_disk_error(zio);
+
+			zio_delay_interrupt(zio);
+		}
+	}
+}
+
+BIO_END_IO_PROTO(vdev_classic_physio_completion, bio, error)
+{
+	dio_request_t *dr = bio->bi_private;
+
+	if (dr->dr_error == 0) {
+#ifdef HAVE_1ARG_BIO_END_IO_T
+		dr->dr_error = BIO_END_IO_ERROR(bio);
+#else
+		if (error)
+			dr->dr_error = -(error);
+		else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
+			dr->dr_error = EIO;
+#endif
+	}
+
+	/* Drop reference acquired by vdev_classic_physio */
+	vdev_classic_dio_put(dr);
+}
+
 static inline unsigned int
-vdev_bio_max_segs(zio_t *zio, int bio_size, uint64_t abd_offset)
+vdev_classic_bio_max_segs(zio_t *zio, int bio_size, uint64_t abd_offset)
 {
 	unsigned long nr_segs = abd_nr_pages_off(zio->io_abd,
 	    bio_size, abd_offset);
@@ -711,9 +1082,16 @@ vdev_bio_max_segs(zio_t *zio, int bio_size, uint64_t abd_offset)
 }
 
 static int
-__vdev_disk_physio(struct block_device *bdev, zio_t *zio,
-    size_t io_size, uint64_t io_offset, int rw, int flags)
+vdev_classic_physio(zio_t *zio)
 {
+	vdev_t *v = zio->io_vd;
+	vdev_disk_t *vd = v->vdev_tsd;
+	struct block_device *bdev = BDH_BDEV(vd->vd_bdh);
+	size_t io_size = zio->io_size;
+	uint64_t io_offset = zio->io_offset;
+	int rw = zio->io_type == ZIO_TYPE_READ ? READ : WRITE;
+	int flags = 0;
+
 	dio_request_t *dr;
 	uint64_t abd_offset;
 	uint64_t bio_offset;
@@ -736,7 +1114,7 @@ __vdev_disk_physio(struct block_device *bdev, zio_t *zio,
 	}
 
 retry:
-	dr = vdev_disk_dio_alloc(bio_count);
+	dr = vdev_classic_dio_alloc(bio_count);
 
 	if (!(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD)) &&
 	    zio->io_vd->vdev_failfast == B_TRUE) {
@@ -771,23 +1149,23 @@ retry:
 		 * this should be rare - see the comment above.
 		 */
 		if (dr->dr_bio_count == i) {
-			vdev_disk_dio_free(dr);
+			vdev_classic_dio_free(dr);
 			bio_count *= 2;
 			goto retry;
 		}
 
-		nr_vecs = vdev_bio_max_segs(zio, bio_size, abd_offset);
+		nr_vecs = vdev_classic_bio_max_segs(zio, bio_size, abd_offset);
 		dr->dr_bio[i] = vdev_bio_alloc(bdev, GFP_NOIO, nr_vecs);
 		if (unlikely(dr->dr_bio[i] == NULL)) {
-			vdev_disk_dio_free(dr);
+			vdev_classic_dio_free(dr);
 			return (SET_ERROR(ENOMEM));
 		}
 
-		/* Matching put called by vdev_disk_physio_completion */
-		vdev_disk_dio_get(dr);
+		/* Matching put called by vdev_classic_physio_completion */
+		vdev_classic_dio_get(dr);
 
 		BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9;
-		dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion;
+		dr->dr_bio[i]->bi_end_io = vdev_classic_physio_completion;
 		dr->dr_bio[i]->bi_private = dr;
 		bio_set_op_attrs(dr->dr_bio[i], rw, flags);
 
@@ -801,7 +1179,7 @@ retry:
 	}
 
 	/* Extra reference to protect dio_request during vdev_submit_bio */
-	vdev_disk_dio_get(dr);
+	vdev_classic_dio_get(dr);
 
 	if (dr->dr_bio_count > 1)
 		blk_start_plug(&plug);
@@ -815,11 +1193,13 @@ retry:
 	if (dr->dr_bio_count > 1)
 		blk_finish_plug(&plug);
 
-	vdev_disk_dio_put(dr);
+	vdev_classic_dio_put(dr);
 
 	return (error);
 }
 
+/* ========== */
+
 BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, error)
 {
 	zio_t *zio = bio->bi_private;
@@ -928,12 +1308,14 @@ vdev_disk_io_trim(zio_t *zio)
 #endif
 }
 
+int (*vdev_disk_io_rw_fn)(zio_t *zio) = NULL;
+
 static void
 vdev_disk_io_start(zio_t *zio)
 {
 	vdev_t *v = zio->io_vd;
 	vdev_disk_t *vd = v->vdev_tsd;
-	int rw, error;
+	int error;
 
 	/*
 	 * If the vdev is closed, it's likely in the REMOVED or FAULTED state.
@@ -996,13 +1378,6 @@ vdev_disk_io_start(zio_t *zio)
 		rw_exit(&vd->vd_lock);
 		zio_execute(zio);
 		return;
-	case ZIO_TYPE_WRITE:
-		rw = WRITE;
-		break;
-
-	case ZIO_TYPE_READ:
-		rw = READ;
-		break;
 
 	case ZIO_TYPE_TRIM:
 		zio->io_error = vdev_disk_io_trim(zio);
@@ -1015,23 +1390,34 @@ vdev_disk_io_start(zio_t *zio)
 #endif
 		return;
 
+	case ZIO_TYPE_READ:
+	case ZIO_TYPE_WRITE:
+		zio->io_target_timestamp = zio_handle_io_delay(zio);
+		error = vdev_disk_io_rw_fn(zio);
+		rw_exit(&vd->vd_lock);
+		if (error) {
+			zio->io_error = error;
+			zio_interrupt(zio);
+		}
+		return;
+
 	default:
+		/*
+		 * Getting here means our parent vdev has made a very strange
+		 * request of us, and shouldn't happen. Assert here to force a
+		 * crash in dev builds, but in production return the IO
+		 * unhandled. The pool will likely suspend anyway but that's
+		 * nicer than crashing the kernel.
+		 */
+		ASSERT3S(zio->io_type, ==, -1);
+
 		rw_exit(&vd->vd_lock);
 		zio->io_error = SET_ERROR(ENOTSUP);
 		zio_interrupt(zio);
 		return;
 	}
 
-	zio->io_target_timestamp = zio_handle_io_delay(zio);
-	error = __vdev_disk_physio(BDH_BDEV(vd->vd_bdh), zio,
-	    zio->io_size, zio->io_offset, rw, 0);
-	rw_exit(&vd->vd_lock);
-
-	if (error) {
-		zio->io_error = error;
-		zio_interrupt(zio);
-		return;
-	}
+	__builtin_unreachable();
 }
 
 static void
@@ -1080,8 +1466,49 @@ vdev_disk_rele(vdev_t *vd)
 	/* XXX: Implement me as a vnode rele for the device */
 }
 
+/*
+ * BIO submission method. See comment above about vdev_classic.
+ * Set zfs_vdev_disk_classic=0 for new, =1 for classic
+ */
+static uint_t zfs_vdev_disk_classic = 1;	/* default classic */
+
+/* Set submission function from module parameter */
+static int
+vdev_disk_param_set_classic(const char *buf, zfs_kernel_param_t *kp)
+{
+	int err = param_set_uint(buf, kp);
+	if (err < 0)
+		return (SET_ERROR(err));
+
+	vdev_disk_io_rw_fn =
+	    zfs_vdev_disk_classic ? vdev_classic_physio : vdev_disk_io_rw;
+
+	printk(KERN_INFO "ZFS: forcing %s BIO submission\n",
+	    zfs_vdev_disk_classic ? "classic" : "new");
+
+	return (0);
+}
+
+/*
+ * At first use vdev use, set the submission function from the default value if
+ * it hasn't been set already.
+ */
+static int
+vdev_disk_init(spa_t *spa, nvlist_t *nv, void **tsd)
+{
+	(void) spa;
+	(void) nv;
+	(void) tsd;
+
+	if (vdev_disk_io_rw_fn == NULL)
+		vdev_disk_io_rw_fn = zfs_vdev_disk_classic ?
+		    vdev_classic_physio : vdev_disk_io_rw;
+
+	return (0);
+}
+
 vdev_ops_t vdev_disk_ops = {
-	.vdev_op_init = NULL,
+	.vdev_op_init = vdev_disk_init,
 	.vdev_op_fini = NULL,
 	.vdev_op_open = vdev_disk_open,
 	.vdev_op_close = vdev_disk_close,
@@ -1174,3 +1601,10 @@ ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, open_timeout_ms, UINT, ZMOD_RW,
 
 ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, failfast_mask, UINT, ZMOD_RW,
 	"Defines failfast mask: 1 - device, 2 - transport, 4 - driver");
+
+ZFS_MODULE_PARAM(zfs_vdev_disk, zfs_vdev_disk_, max_segs, UINT, ZMOD_RW,
+	"Maximum number of data segments to add to an IO request (min 4)");
+
+ZFS_MODULE_PARAM_CALL(zfs_vdev_disk, zfs_vdev_disk_, classic,
+    vdev_disk_param_set_classic, param_get_uint, ZMOD_RD,
+	"Use classic BIO submission method");
diff --git a/module/os/linux/zfs/zfs_vnops_os.c b/module/os/linux/zfs/zfs_vnops_os.c
index 3720256d47..d2ae1bc010 100644
--- a/module/os/linux/zfs/zfs_vnops_os.c
+++ b/module/os/linux/zfs/zfs_vnops_os.c
@@ -3821,11 +3821,8 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc,
 	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
 	zfs_sa_upgrade_txholds(tx, zp);
 
-	err = dmu_tx_assign(tx, TXG_NOWAIT);
+	err = dmu_tx_assign(tx, TXG_WAIT);
 	if (err != 0) {
-		if (err == ERESTART)
-			dmu_tx_wait(tx);
-
 		dmu_tx_abort(tx);
 #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO
 		filemap_dirty_folio(page_mapping(pp), page_folio(pp));
@@ -4277,9 +4274,4 @@ EXPORT_SYMBOL(zfs_map);
 /* CSTYLED */
 module_param(zfs_delete_blocks, ulong, 0644);
 MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async");
-
-/* CSTYLED */
-module_param(zfs_bclone_enabled, uint, 0644);
-MODULE_PARM_DESC(zfs_bclone_enabled, "Enable block cloning");
-
 #endif
diff --git a/module/os/linux/zfs/zpl_file.c b/module/os/linux/zfs/zpl_file.c
index 3caa0fc6c2..9dec52215c 100644
--- a/module/os/linux/zfs/zpl_file.c
+++ b/module/os/linux/zfs/zpl_file.c
@@ -720,23 +720,23 @@ zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
 {
 	boolean_t *for_sync = data;
 	fstrans_cookie_t cookie;
+	int ret;
 
 	ASSERT(PageLocked(pp));
 	ASSERT(!PageWriteback(pp));
 
 	cookie = spl_fstrans_mark();
-	(void) zfs_putpage(pp->mapping->host, pp, wbc, *for_sync);
+	ret = zfs_putpage(pp->mapping->host, pp, wbc, *for_sync);
 	spl_fstrans_unmark(cookie);
 
-	return (0);
+	return (ret);
 }
 
 #ifdef HAVE_WRITEPAGE_T_FOLIO
 static int
 zpl_putfolio(struct folio *pp, struct writeback_control *wbc, void *data)
 {
-	(void) zpl_putpage(&pp->page, wbc, data);
-	return (0);
+	return (zpl_putpage(&pp->page, wbc, data));
 }
 #endif
 
diff --git a/module/os/linux/zfs/zpl_file_range.c b/module/os/linux/zfs/zpl_file_range.c
index 139c51cf46..64728fdb11 100644
--- a/module/os/linux/zfs/zpl_file_range.c
+++ b/module/os/linux/zfs/zpl_file_range.c
@@ -26,13 +26,14 @@
 #include <linux/compat.h>
 #endif
 #include <linux/fs.h>
+#ifdef HAVE_VFS_SPLICE_COPY_FILE_RANGE
+#include <linux/splice.h>
+#endif
 #include <sys/file.h>
 #include <sys/zfs_znode.h>
 #include <sys/zfs_vnops.h>
 #include <sys/zfeature.h>
 
-int zfs_bclone_enabled = 0;
-
 /*
  * Clone part of a file via block cloning.
  *
@@ -40,7 +41,7 @@ int zfs_bclone_enabled = 0;
  * care of that depending on how it was called.
  */
 static ssize_t
-__zpl_clone_file_range(struct file *src_file, loff_t src_off,
+zpl_clone_file_range_impl(struct file *src_file, loff_t src_off,
     struct file *dst_file, loff_t dst_off, size_t len)
 {
 	struct inode *src_i = file_inode(src_file);
@@ -96,14 +97,15 @@ zpl_copy_file_range(struct file *src_file, loff_t src_off,
 {
 	ssize_t ret;
 
+	/* Flags is reserved for future extensions and must be zero. */
 	if (flags != 0)
 		return (-EINVAL);
 
-	/* Try to do it via zfs_clone_range() */
-	ret = __zpl_clone_file_range(src_file, src_off,
+	/* Try to do it via zfs_clone_range() and allow shortening. */
+	ret = zpl_clone_file_range_impl(src_file, src_off,
 	    dst_file, dst_off, len);
 
-#ifdef HAVE_VFS_GENERIC_COPY_FILE_RANGE
+#if defined(HAVE_VFS_GENERIC_COPY_FILE_RANGE)
 	/*
 	 * Since Linux 5.3 the filesystem driver is responsible for executing
 	 * an appropriate fallback, and a generic fallback function is provided.
@@ -112,6 +114,15 @@ zpl_copy_file_range(struct file *src_file, loff_t src_off,
 	    ret == -EAGAIN)
 		ret = generic_copy_file_range(src_file, src_off, dst_file,
 		    dst_off, len, flags);
+#elif defined(HAVE_VFS_SPLICE_COPY_FILE_RANGE)
+	/*
+	 * Since 6.8 the fallback function is called splice_copy_file_range
+	 * and has a slightly different signature.
+	 */
+	if (ret == -EOPNOTSUPP || ret == -EINVAL || ret == -EXDEV ||
+	    ret == -EAGAIN)
+		ret = splice_copy_file_range(src_file, src_off, dst_file,
+		    dst_off, len);
 #else
 	/*
 	 * Before Linux 5.3 the filesystem has to return -EOPNOTSUPP to signal
@@ -119,7 +130,7 @@ zpl_copy_file_range(struct file *src_file, loff_t src_off,
 	 */
 	if (ret == -EINVAL || ret == -EXDEV || ret == -EAGAIN)
 		ret = -EOPNOTSUPP;
-#endif /* HAVE_VFS_GENERIC_COPY_FILE_RANGE */
+#endif /* HAVE_VFS_GENERIC_COPY_FILE_RANGE || HAVE_VFS_SPLICE_COPY_FILE_RANGE */
 
 	return (ret);
 }
@@ -137,6 +148,11 @@ zpl_copy_file_range(struct file *src_file, loff_t src_off,
  * FIDEDUPERANGE is for turning a non-clone into a clone, that is, compare the
  * range in both files and if they're the same, arrange for them to be backed
  * by the same storage.
+ *
+ * REMAP_FILE_CAN_SHORTEN lets us know we can clone less than the given range
+ * if we want. It's designed for filesystems that may need to shorten the
+ * length for alignment, EOF, or any other requirement. ZFS may shorten the
+ * request when there is outstanding dirty data which hasn't been written.
  */
 loff_t
 zpl_remap_file_range(struct file *src_file, loff_t src_off,
@@ -145,24 +161,21 @@ zpl_remap_file_range(struct file *src_file, loff_t src_off,
 	if (flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_CAN_SHORTEN))
 		return (-EINVAL);
 
-	/*
-	 * REMAP_FILE_CAN_SHORTEN lets us know we can clone less than the given
-	 * range if we want. Its designed for filesystems that make data past
-	 * EOF available, and don't want it to be visible in both files. ZFS
-	 * doesn't do that, so we just turn the flag off.
-	 */
-	flags &= ~REMAP_FILE_CAN_SHORTEN;
-
+	/* No support for dedup yet */
 	if (flags & REMAP_FILE_DEDUP)
-		/* No support for dedup yet */
 		return (-EOPNOTSUPP);
 
 	/* Zero length means to clone everything to the end of the file */
 	if (len == 0)
 		len = i_size_read(file_inode(src_file)) - src_off;
 
-	return (__zpl_clone_file_range(src_file, src_off,
-	    dst_file, dst_off, len));
+	ssize_t ret = zpl_clone_file_range_impl(src_file, src_off,
+	    dst_file, dst_off, len);
+
+	if (!(flags & REMAP_FILE_CAN_SHORTEN) && ret >= 0 && ret != len)
+		ret = -EINVAL;
+
+	return (ret);
 }
 #endif /* HAVE_VFS_REMAP_FILE_RANGE */
 
@@ -179,8 +192,14 @@ zpl_clone_file_range(struct file *src_file, loff_t src_off,
 	if (len == 0)
 		len = i_size_read(file_inode(src_file)) - src_off;
 
-	return (__zpl_clone_file_range(src_file, src_off,
-	    dst_file, dst_off, len));
+	/* The entire length must be cloned or this is an error. */
+	ssize_t ret = zpl_clone_file_range_impl(src_file, src_off,
+	    dst_file, dst_off, len);
+
+	if (ret >= 0 && ret != len)
+		ret = -EINVAL;
+
+	return (ret);
 }
 #endif /* HAVE_VFS_CLONE_FILE_RANGE || HAVE_VFS_FILE_OPERATIONS_EXTEND */
 
@@ -214,8 +233,7 @@ zpl_ioctl_ficlone(struct file *dst_file, void *arg)
 
 	size_t len = i_size_read(file_inode(src_file));
 
-	ssize_t ret =
-	    __zpl_clone_file_range(src_file, 0, dst_file, 0, len);
+	ssize_t ret = zpl_clone_file_range_impl(src_file, 0, dst_file, 0, len);
 
 	fput(src_file);
 
@@ -253,7 +271,7 @@ zpl_ioctl_ficlonerange(struct file *dst_file, void __user *arg)
 	if (len == 0)
 		len = i_size_read(file_inode(src_file)) - fcr.fcr_src_offset;
 
-	ssize_t ret = __zpl_clone_file_range(src_file, fcr.fcr_src_offset,
+	ssize_t ret = zpl_clone_file_range_impl(src_file, fcr.fcr_src_offset,
 	    dst_file, fcr.fcr_dest_offset, len);
 
 	fput(src_file);
diff --git a/module/zfs/abd.c b/module/zfs/abd.c
index d982f201c9..3388e23573 100644
--- a/module/zfs/abd.c
+++ b/module/zfs/abd.c
@@ -826,6 +826,48 @@ abd_iterate_func(abd_t *abd, size_t off, size_t size,
 	return (ret);
 }
 
+#if defined(__linux__) && defined(_KERNEL)
+int
+abd_iterate_page_func(abd_t *abd, size_t off, size_t size,
+    abd_iter_page_func_t *func, void *private)
+{
+	struct abd_iter aiter;
+	int ret = 0;
+
+	if (size == 0)
+		return (0);
+
+	abd_verify(abd);
+	ASSERT3U(off + size, <=, abd->abd_size);
+
+	abd_t *c_abd = abd_init_abd_iter(abd, &aiter, off);
+
+	while (size > 0) {
+		IMPLY(abd_is_gang(abd), c_abd != NULL);
+
+		abd_iter_page(&aiter);
+
+		size_t len = MIN(aiter.iter_page_dsize, size);
+		ASSERT3U(len, >, 0);
+
+		ret = func(aiter.iter_page, aiter.iter_page_doff,
+		    len, private);
+
+		aiter.iter_page = NULL;
+		aiter.iter_page_doff = 0;
+		aiter.iter_page_dsize = 0;
+
+		if (ret != 0)
+			break;
+
+		size -= len;
+		c_abd = abd_advance_abd_iter(abd, c_abd, &aiter, len);
+	}
+
+	return (ret);
+}
+#endif
+
 struct buf_arg {
 	void *arg_buf;
 };
diff --git a/module/zfs/spa_misc.c b/module/zfs/spa_misc.c
index 72b690162d..24f038ad7f 100644
--- a/module/zfs/spa_misc.c
+++ b/module/zfs/spa_misc.c
@@ -1822,7 +1822,8 @@ spa_get_slop_space(spa_t *spa)
 	 * deduplicated data, so since it's not useful to reserve more
 	 * space with more deduplicated data, we subtract that out here.
 	 */
-	space = spa_get_dspace(spa) - spa->spa_dedup_dspace;
+	space =
+	    spa_get_dspace(spa) - spa->spa_dedup_dspace - brt_get_dspace(spa);
 	slop = MIN(space >> spa_slop_shift, spa_max_slop);
 
 	/*
diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
index e6ae574ad0..7020f88ecf 100644
--- a/module/zfs/zfs_vnops.c
+++ b/module/zfs/zfs_vnops.c
@@ -58,6 +58,26 @@
 #include <sys/zfs_vfsops.h>
 #include <sys/zfs_znode.h>
 
+/*
+ * Enable the experimental block cloning feature.  If this setting is 0, then
+ * even if feature@block_cloning is enabled, attempts to clone blocks will act
+ * as though the feature is disabled.
+ */
+int zfs_bclone_enabled = 0;
+
+/*
+ * When set zfs_clone_range() waits for dirty data to be written to disk.
+ * This allows the clone operation to reliably succeed when a file is modified
+ * and then immediately cloned. For small files this may be slower than making
+ * a copy of the file and is therefore not the default.  However, in certain
+ * scenarios this behavior may be desirable so a tunable is provided.
+ */
+static int zfs_bclone_wait_dirty = 0;
+
+/*
+ * Maximum bytes to read per chunk in zfs_read().
+ */
+static uint64_t zfs_vnops_read_chunk_size = 1024 * 1024;
 
 static ulong_t zfs_fsync_sync_cnt = 4;
 
@@ -110,7 +130,7 @@ zfs_holey_common(znode_t *zp, ulong_t cmd, loff_t *off)
 
 	/* Flush any mmap()'d data to disk */
 	if (zn_has_cached_data(zp, 0, file_sz - 1))
-		zn_flush_cached_data(zp, B_FALSE);
+		zn_flush_cached_data(zp, B_TRUE);
 
 	lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_READER);
 	error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff);
@@ -189,8 +209,6 @@ zfs_access(znode_t *zp, int mode, int flag, cred_t *cr)
 	return (error);
 }
 
-static uint64_t zfs_vnops_read_chunk_size = 1024 * 1024; /* Tunable */
-
 /*
  * Read bytes from specified file into supplied buffer.
  *
@@ -1055,6 +1073,7 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 	size_t		maxblocks, nbps;
 	uint_t		inblksz;
 	uint64_t	clear_setid_bits_txg = 0;
+	uint64_t	last_synced_txg = 0;
 
 	inoff = *inoffp;
 	outoff = *outoffp;
@@ -1174,6 +1193,10 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 		}
 	}
 
+	/* Flush any mmap()'d data to disk */
+	if (zn_has_cached_data(inzp, inoff, inoff + len - 1))
+		zn_flush_cached_data(inzp, B_TRUE);
+
 	/*
 	 * Maintain predictable lock order.
 	 */
@@ -1293,15 +1316,23 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 		}
 
 		nbps = maxblocks;
+		last_synced_txg = spa_last_synced_txg(dmu_objset_spa(inos));
 		error = dmu_read_l0_bps(inos, inzp->z_id, inoff, size, bps,
 		    &nbps);
 		if (error != 0) {
 			/*
 			 * If we are trying to clone a block that was created
-			 * in the current transaction group, error will be
-			 * EAGAIN here, which we can just return to the caller
-			 * so it can fallback if it likes.
+			 * in the current transaction group, the error will be
+			 * EAGAIN here.  Based on zfs_bclone_wait_dirty either
+			 * return a shortened range to the caller so it can
+			 * fallback, or wait for the next TXG and check again.
 			 */
+			if (error == EAGAIN && zfs_bclone_wait_dirty) {
+				txg_wait_synced(dmu_objset_pool(inos),
+				    last_synced_txg + 1);
+				continue;
+			}
+
 			break;
 		}
 
@@ -1523,3 +1554,9 @@ EXPORT_SYMBOL(zfs_clone_range_replay);
 
 ZFS_MODULE_PARAM(zfs_vnops, zfs_vnops_, read_chunk_size, U64, ZMOD_RW,
 	"Bytes to read per chunk");
+
+ZFS_MODULE_PARAM(zfs, zfs_, bclone_enabled, INT, ZMOD_RW,
+	"Enable block cloning");
+
+ZFS_MODULE_PARAM(zfs, zfs_, bclone_wait_dirty, INT, ZMOD_RW,
+	"Wait for dirty blocks when cloning");
diff --git a/tests/runfiles/common.run b/tests/runfiles/common.run
index a3550d26ab..7331244515 100644
--- a/tests/runfiles/common.run
+++ b/tests/runfiles/common.run
@@ -535,7 +535,8 @@ tags = ['functional', 'cli_root', 'zpool_split']
 tests = ['zpool_status_001_pos', 'zpool_status_002_pos',
     'zpool_status_003_pos', 'zpool_status_004_pos',
     'zpool_status_005_pos', 'zpool_status_006_pos',
-    'zpool_status_007_pos', 'zpool_status_features_001_pos']
+    'zpool_status_007_pos', 'zpool_status_008_pos',
+    'zpool_status_features_001_pos']
 tags = ['functional', 'cli_root', 'zpool_status']
 
 [tests/functional/cli_root/zpool_sync]
@@ -630,7 +631,7 @@ tests = ['compress_001_pos', 'compress_002_pos', 'compress_003_pos',
 tags = ['functional', 'compression']
 
 [tests/functional/cp_files]
-tests = ['cp_files_001_pos', 'cp_stress']
+tests = ['cp_files_001_pos', 'cp_files_002_pos', 'cp_stress']
 tags = ['functional', 'cp_files']
 
 [tests/functional/crtime]
diff --git a/tests/test-runner/bin/zts-report.py.in b/tests/test-runner/bin/zts-report.py.in
index ae4aa62754..ecc50f4871 100755
--- a/tests/test-runner/bin/zts-report.py.in
+++ b/tests/test-runner/bin/zts-report.py.in
@@ -138,7 +138,11 @@ idmap_reason = 'Idmapped mount needs kernel 5.12+'
 # copy_file_range() is not supported by all kernels
 #
 cfr_reason = 'Kernel copy_file_range support required'
-cfr_cross_reason = 'copy_file_range(2) cross-filesystem needs kernel 5.3+'
+
+if sys.platform.startswith('freebsd'):
+    cfr_cross_reason = 'copy_file_range(2) cross-filesystem needs FreeBSD 14+'
+else:
+    cfr_cross_reason = 'copy_file_range(2) cross-filesystem needs kernel 5.3+'
 
 #
 # These tests are known to fail, thus we use this list to prevent these
@@ -176,6 +180,7 @@ if sys.platform.startswith('freebsd'):
         'cli_root/zpool_wait/zpool_wait_trim_cancel': ['SKIP', trim_reason],
         'cli_root/zpool_wait/zpool_wait_trim_flag': ['SKIP', trim_reason],
         'cli_root/zfs_unshare/zfs_unshare_008_pos': ['SKIP', na_reason],
+        'cp_files/cp_files_002_pos': ['SKIP', na_reason],
         'link_count/link_count_001': ['SKIP', na_reason],
         'casenorm/mixed_create_failure': ['FAIL', 13215],
         'mmap/mmap_sync_001_pos': ['SKIP', na_reason],
@@ -267,6 +272,22 @@ if sys.platform.startswith('freebsd'):
         'pool_checkpoint/checkpoint_indirect': ['FAIL', 12623],
         'resilver/resilver_restart_001': ['FAIL', known_reason],
         'snapshot/snapshot_002_pos': ['FAIL', '14831'],
+        'bclone/bclone_crossfs_corner_cases': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_crossfs_corner_cases_limited':
+            ['SKIP', cfr_cross_reason],
+        'bclone/bclone_crossfs_data': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_crossfs_embedded': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_crossfs_hole': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_diffprops_all': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_diffprops_checksum': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_diffprops_compress': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_diffprops_copies': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_diffprops_recordsize': ['SKIP', cfr_cross_reason],
+        'bclone/bclone_prop_sync': ['SKIP', cfr_cross_reason],
+        'block_cloning/block_cloning_cross_enc_dataset':
+            ['SKIP', cfr_cross_reason],
+        'block_cloning/block_cloning_copyfilerange_cross_dataset':
+            ['SKIP', cfr_cross_reason]
     })
 elif sys.platform.startswith('linux'):
     maybe.update({
@@ -312,6 +333,7 @@ elif sys.platform.startswith('linux'):
             ['SKIP', cfr_reason],
         'cli_root/zfs_rename/zfs_rename_002_pos': ['FAIL', known_reason],
         'cli_root/zpool_reopen/zpool_reopen_003_pos': ['FAIL', known_reason],
+        'cp_files/cp_files_002_pos': ['SKIP', cfr_reason],
         'fault/auto_online_002_pos': ['FAIL', 11889],
         'fault/auto_replace_001_pos': ['FAIL', 14851],
         'fault/auto_spare_002_pos': ['FAIL', 11889],
diff --git a/tests/zfs-tests/include/libtest.shlib b/tests/zfs-tests/include/libtest.shlib
index b4d2b91dd4..dfab48d2cd 100644
--- a/tests/zfs-tests/include/libtest.shlib
+++ b/tests/zfs-tests/include/libtest.shlib
@@ -61,13 +61,8 @@ function compare_version_gte
 	[ "$(printf "$1\n$2" | sort -V | tail -n1)" = "$1" ]
 }
 
-# Linux kernel version comparison function
-#
-# $1 Linux version ("4.10", "2.6.32") or blank for installed Linux version
-#
-# Used for comparison: if [ $(linux_version) -ge $(linux_version "2.6.32") ]
-#
-function linux_version
+# Helper function used by linux_version() and freebsd_version()
+function kernel_version
 {
 	typeset ver="$1"
 
@@ -83,6 +78,24 @@ function linux_version
 	echo $((version * 100000 + major * 1000 + minor))
 }
 
+# Linux kernel version comparison function
+#
+# $1 Linux version ("4.10", "2.6.32") or blank for installed Linux version
+#
+# Used for comparison: if [ $(linux_version) -ge $(linux_version "2.6.32") ]
+function linux_version {
+	kernel_version "$1"
+}
+
+# FreeBSD version comparison function
+#
+# $1 FreeBSD version ("13.2", "14.0") or blank for installed FreeBSD version
+#
+# Used for comparison: if [ $(freebsd_version) -ge $(freebsd_version "13.2") ]
+function freebsd_version {
+	kernel_version "$1"
+}
+
 # Determine if this is a Linux test system
 #
 # Return 0 if platform Linux, 1 if otherwise
diff --git a/tests/zfs-tests/include/tunables.cfg b/tests/zfs-tests/include/tunables.cfg
index a0edad14d0..718c4cf2d8 100644
--- a/tests/zfs-tests/include/tunables.cfg
+++ b/tests/zfs-tests/include/tunables.cfg
@@ -90,7 +90,8 @@ VOL_INHIBIT_DEV			UNSUPPORTED			zvol_inhibit_dev
 VOL_MODE			vol.mode			zvol_volmode
 VOL_RECURSIVE			vol.recursive			UNSUPPORTED
 VOL_USE_BLK_MQ			UNSUPPORTED			zvol_use_blk_mq
-BCLONE_ENABLED			zfs_bclone_enabled		zfs_bclone_enabled
+BCLONE_ENABLED			bclone_enabled			zfs_bclone_enabled
+BCLONE_WAIT_DIRTY		bclone_wait_dirty		zfs_bclone_wait_dirty
 XATTR_COMPAT			xattr_compat			zfs_xattr_compat
 ZEVENT_LEN_MAX			zevent.len_max			zfs_zevent_len_max
 ZEVENT_RETAIN_MAX		zevent.retain_max		zfs_zevent_retain_max
diff --git a/tests/zfs-tests/tests/Makefile.am b/tests/zfs-tests/tests/Makefile.am
index 8bee07f480..e2824ee065 100644
--- a/tests/zfs-tests/tests/Makefile.am
+++ b/tests/zfs-tests/tests/Makefile.am
@@ -1238,6 +1238,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/cli_root/zpool_status/zpool_status_005_pos.ksh \
 	functional/cli_root/zpool_status/zpool_status_006_pos.ksh \
 	functional/cli_root/zpool_status/zpool_status_007_pos.ksh \
+	functional/cli_root/zpool_status/zpool_status_008_pos.ksh \
 	functional/cli_root/zpool_status/zpool_status_features_001_pos.ksh \
 	functional/cli_root/zpool_sync/cleanup.ksh \
 	functional/cli_root/zpool_sync/setup.ksh \
@@ -1393,6 +1394,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/compression/setup.ksh \
 	functional/cp_files/cleanup.ksh \
 	functional/cp_files/cp_files_001_pos.ksh \
+	functional/cp_files/cp_files_002_pos.ksh \
 	functional/cp_files/cp_stress.ksh \
 	functional/cp_files/setup.ksh \
 	functional/crtime/cleanup.ksh \
diff --git a/tests/zfs-tests/tests/functional/bclone/bclone_common.kshlib b/tests/zfs-tests/tests/functional/bclone/bclone_common.kshlib
index beba01c0ed..3b8eaea5bb 100644
--- a/tests/zfs-tests/tests/functional/bclone/bclone_common.kshlib
+++ b/tests/zfs-tests/tests/functional/bclone/bclone_common.kshlib
@@ -42,6 +42,12 @@ function verify_crossfs_block_cloning
 	if is_linux && [[ $(linux_version) -lt $(linux_version "5.3") ]]; then
 		log_unsupported "copy_file_range can't copy cross-filesystem before Linux 5.3"
 	fi
+
+	# Cross dataset block cloning only supported on FreeBSD 14+
+	# https://github.com/freebsd/freebsd-src/commit/969071be938c
+        if is_freebsd && [ $(freebsd_version) -lt $(freebsd_version 14.0) ] ; then
+               log_unsupported "Cloning across datasets not supported in $(uname -r)"
+        fi
 }
 
 # Unused.
diff --git a/tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh b/tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh
index 43323c207a..ad83d30291 100755
--- a/tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh
+++ b/tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh
@@ -26,12 +26,11 @@
 
 . $STF_SUITE/include/libtest.shlib
 . $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+. $STF_SUITE/tests/functional/bclone/bclone_common.kshlib
 
 verify_runnable "global"
 
-if is_linux && [[ $(linux_version) -lt $(linux_version "5.3") ]]; then
-  log_unsupported "copy_file_range can't copy cross-filesystem before Linux 5.3"
-fi
+verify_crossfs_block_cloning
 
 claim="The copy_file_range syscall can clone across datasets."
 
diff --git a/tests/zfs-tests/tests/functional/block_cloning/block_cloning_cross_enc_dataset.ksh b/tests/zfs-tests/tests/functional/block_cloning/block_cloning_cross_enc_dataset.ksh
index 34d3d26925..702e23267f 100755
--- a/tests/zfs-tests/tests/functional/block_cloning/block_cloning_cross_enc_dataset.ksh
+++ b/tests/zfs-tests/tests/functional/block_cloning/block_cloning_cross_enc_dataset.ksh
@@ -26,12 +26,11 @@
 
 . $STF_SUITE/include/libtest.shlib
 . $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+. $STF_SUITE/tests/functional/bclone/bclone_common.kshlib
 
 verify_runnable "global"
 
-if is_linux && [[ $(linux_version) -lt $(linux_version "5.3") ]]; then
-  log_unsupported "copy_file_range can't copy cross-filesystem before Linux 5.3"
-fi
+verify_crossfs_block_cloning
 
 claim="Block cloning across encrypted datasets."
 
diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_002_pos.ksh b/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_002_pos.ksh
index 3bdd7db649..d6f32cdc7a 100755
--- a/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_002_pos.ksh
+++ b/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_002_pos.ksh
@@ -51,7 +51,7 @@ else
 fi
 
 set -A args "" "-x" "-v" "-x $testpool" "-v $testpool" "-xv $testpool" \
-	"-vx $testpool"
+	"-vx $testpool" "-e $testpool" "-es $testpool"
 
 log_assert "Executing 'zpool status' with correct options succeeds"
 
@@ -64,4 +64,6 @@ while [[ $i -lt ${#args[*]} ]]; do
 	(( i = i + 1 ))
 done
 
+cleanup 
+
 log_pass "'zpool status' with correct options succeeded"
diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_003_pos.ksh b/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_003_pos.ksh
index b501aac5ad..52b22dd833 100755
--- a/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_003_pos.ksh
+++ b/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_003_pos.ksh
@@ -37,6 +37,7 @@
 # 3. Read the file
 # 4. Take a snapshot and make a clone
 # 5. Verify we see "snapshot, clone and filesystem" output in 'zpool status -v'
+#      and 'zpool status -ev'
 
 function cleanup
 {
@@ -68,6 +69,7 @@ log_must zpool status -v $TESTPOOL2
 log_must eval "zpool status -v | grep '$TESTPOOL2@snap:/10m_file'"
 log_must eval "zpool status -v | grep '$TESTPOOL2/clone/10m_file'"
 log_must eval "zpool status -v | grep '$TESTPOOL2/10m_file'"
+log_must eval "zpool status -ev | grep '$TESTPOOL2/10m_file'"
 log_mustnot eval "zpool status -v | grep '$TESTFS1'"
 
 log_pass "'zpool status -v' outputs affected filesystem, snapshot & clone"
diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_008_pos.ksh b/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_008_pos.ksh
new file mode 100755
index 0000000000..6be2ad5a74
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/cli_root/zpool_status/zpool_status_008_pos.ksh
@@ -0,0 +1,104 @@
+#!/bin/ksh -p
+
+#
+# CDDL HEADER START
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# http://www.illumos.org/license/CDDL.
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2024 by Lawrence Livermore National Security, LLC.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+#
+# DESCRIPTION:
+# Verify 'zpool status -e' only shows unhealthy devices.
+#
+# STRATEGY:
+# 1. Create zpool
+# 2. Force DEGRADE, FAULT, or inject slow IOs for vdevs
+# 3. Verify vdevs are reported correctly with -e and -s
+# 4. Verify parents are reported as DEGRADED
+# 5. Verify healthy children are not reported
+#
+
+function cleanup
+{
+	log_must set_tunable64 ZIO_SLOW_IO_MS $OLD_SLOW_IO
+	zinject -c all
+	poolexists $TESTPOOL2 && destroy_pool $TESTPOOL2
+	log_must rm -f $all_vdevs
+}
+
+log_assert "Verify 'zpool status -e'"
+
+log_onexit cleanup
+
+all_vdevs=$(echo $TESTDIR/vdev{1..6})
+log_must mkdir -p $TESTDIR
+log_must truncate -s $MINVDEVSIZE $all_vdevs
+
+OLD_SLOW_IO=$(get_tunable ZIO_SLOW_IO_MS)
+
+for raid_type in "draid2:3d:6c:1s" "raidz2"; do
+
+	log_must zpool create -f $TESTPOOL2 $raid_type $all_vdevs
+
+	# Check DEGRADED vdevs are shown.
+	log_must check_vdev_state $TESTPOOL2 $TESTDIR/vdev4 "ONLINE"
+	log_must zinject -d $TESTDIR/vdev4 -A degrade $TESTPOOL2
+	log_must eval "zpool status -e $TESTPOOL2 | grep $TESTDIR/vdev4 | grep DEGRADED"
+
+	# Check FAULTED vdevs are shown.
+	log_must check_vdev_state $TESTPOOL2 $TESTDIR/vdev5 "ONLINE"
+	log_must zinject -d $TESTDIR/vdev5 -A fault $TESTPOOL2
+	log_must eval "zpool status -e $TESTPOOL2 | grep $TESTDIR/vdev5 | grep FAULTED"
+
+	# Check no ONLINE vdevs are shown
+	log_mustnot eval "zpool status -e $TESTPOOL2 | grep ONLINE"
+
+	# Check no ONLINE slow vdevs are show.  Then mark IOs greater than
+	# 10ms slow, delay IOs 20ms to vdev6, check slow IOs.
+	log_must check_vdev_state $TESTPOOL2 $TESTDIR/vdev6 "ONLINE"
+	log_mustnot eval "zpool status -es $TESTPOOL2 | grep ONLINE"
+
+	log_must set_tunable64 ZIO_SLOW_IO_MS 10
+	log_must zinject -d $TESTDIR/vdev6 -D20:100 $TESTPOOL2
+	log_must mkfile 1048576 /$TESTPOOL2/testfile
+	sync_pool $TESTPOOL2
+	log_must set_tunable64 ZIO_SLOW_IO_MS $OLD_SLOW_IO
+
+	# Check vdev6 slow IOs are only shown when requested with -s.
+	log_mustnot eval "zpool status -e $TESTPOOL2 | grep $TESTDIR/vdev6 | grep ONLINE"
+	log_must eval "zpool status -es $TESTPOOL2 | grep $TESTDIR/vdev6 | grep ONLINE"
+
+	# Pool level and top-vdev level status must be DEGRADED.
+	log_must eval "zpool status -e $TESTPOOL2 | grep $TESTPOOL2 | grep DEGRADED"
+	log_must eval "zpool status -e $TESTPOOL2 | grep $raid_type | grep DEGRADED"
+
+	# Check that healthy vdevs[1-3] aren't shown with -e.
+	log_must check_vdev_state $TESTPOOL2 $TESTDIR/vdev1 "ONLINE"
+	log_must check_vdev_state $TESTPOOL2 $TESTDIR/vdev2 "ONLINE"
+	log_must check_vdev_state $TESTPOOL2 $TESTDIR/vdev3 "ONLINE"
+	log_mustnot eval "zpool status -es $TESTPOOL2 | grep $TESTDIR/vdev1 | grep ONLINE"
+	log_mustnot eval "zpool status -es $TESTPOOL2 | grep $TESTDIR/vdev2 | grep ONLINE"
+	log_mustnot eval "zpool status -es $TESTPOOL2 | grep $TESTDIR/vdev3 | grep ONLINE"
+
+	log_must zinject -c all
+	log_must zpool status -es $TESTPOOL2
+
+	zpool destroy $TESTPOOL2
+done
+
+log_pass "Verify zpool status -e shows only unhealthy vdevs"
diff --git a/tests/zfs-tests/tests/functional/cp_files/cleanup.ksh b/tests/zfs-tests/tests/functional/cp_files/cleanup.ksh
index 42fe70042d..c0bccab122 100755
--- a/tests/zfs-tests/tests/functional/cp_files/cleanup.ksh
+++ b/tests/zfs-tests/tests/functional/cp_files/cleanup.ksh
@@ -32,3 +32,7 @@
 . $STF_SUITE/include/libtest.shlib
 
 default_cleanup
+
+if tunable_exists BCLONE_ENABLED ; then
+	log_must restore_tunable BCLONE_ENABLED
+fi
diff --git a/tests/zfs-tests/tests/functional/cp_files/cp_files_002_pos.ksh b/tests/zfs-tests/tests/functional/cp_files/cp_files_002_pos.ksh
new file mode 100755
index 0000000000..60817449ab
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/cp_files/cp_files_002_pos.ksh
@@ -0,0 +1,161 @@
+#! /bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or https://opensource.org/licenses/CDDL-1.0.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2024 by Lawrence Livermore National Security, LLC.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/bclone/bclone_common.kshlib
+
+#
+# DESCRIPTION:
+# Verify all cp --reflink modes work with modified file.
+#
+# STRATEGY:
+# 1. Verify "cp --reflink=never|auto|always" behaves as expected.
+#    Two different modes of operation are tested.
+#
+#    a. zfs_bclone_wait_dirty=0: FICLONE and FICLONERANGE fail with EINVAL
+#       when there are dirty blocks which cannot be immediately cloned.
+#       This is the default behavior.
+#
+#    b. zfs_bclone_wait_dirty=1: FICLONE and FICLONERANGE wait for
+#       dirty blocks to be written to disk allowing the clone to succeed.
+#       The downside to this is it may be slow which depending on the
+#       situtation may defeat the point of making a clone.
+#
+
+verify_runnable "global"
+verify_block_cloning
+
+if ! is_linux; then
+	log_unsupported "cp --reflink is a GNU coreutils option"
+fi
+
+function cleanup
+{
+	datasetexists $TESTPOOL/cp-reflink && \
+	    destroy_dataset $$TESTPOOL/cp-reflink -f
+	log_must set_tunable32 BCLONE_WAIT_DIRTY 0
+}
+
+function verify_copy
+{
+	src_cksum=$(sha256digest $1)
+	dst_cksum=$(sha256digest $2)
+
+	if [[ "$src_cksum" != "$dst_cksum" ]]; then
+		log_must ls -l $CP_TESTDIR
+		log_fail "checksum mismatch ($src_cksum != $dst_cksum)"
+	fi
+}
+
+log_assert "Verify all cp --reflink modes work with modified file"
+
+log_onexit cleanup
+
+SRC_FILE=src.data
+DST_FILE=dst.data
+SRC_SIZE=$(($RANDOM % 2048))
+
+# A smaller recordsize is used merely to speed up the test.
+RECORDSIZE=4096
+
+log_must zfs create -o recordsize=$RECORDSIZE $TESTPOOL/cp-reflink
+CP_TESTDIR=$(get_prop mountpoint $TESTPOOL/cp-reflink)
+
+log_must cd $CP_TESTDIR
+
+# Never wait on dirty blocks (zfs_bclone_wait_dirty=0)
+log_must set_tunable32 BCLONE_WAIT_DIRTY 0
+
+for mode in "never" "auto" "always"; do
+	log_note "Checking 'cp --reflink=$mode'"
+
+	# Create a new file and immediately copy it.
+	log_must dd if=/dev/urandom of=$SRC_FILE bs=$RECORDSIZE count=$SRC_SIZE
+
+	if [[ "$mode" == "always" ]]; then
+		log_mustnot cp --reflink=$mode $SRC_FILE $DST_FILE
+		log_must ls -l $CP_TESTDIR
+	else
+		log_must cp --reflink=$mode $SRC_FILE $DST_FILE
+		verify_copy $SRC_FILE $DST_FILE
+	fi
+	log_must rm -f $DST_FILE
+
+	# Append to an existing file and immediately copy it.
+	sync_pool $TESTPOOL
+	log_must dd if=/dev/urandom of=$SRC_FILE bs=$RECORDSIZE seek=$SRC_SIZE \
+	    count=1 conv=notrunc
+	if [[ "$mode" == "always" ]]; then
+		log_mustnot cp --reflink=$mode $SRC_FILE $DST_FILE
+		log_must ls -l $CP_TESTDIR
+	else
+		log_must cp --reflink=$mode $SRC_FILE $DST_FILE
+		verify_copy $SRC_FILE $DST_FILE
+	fi
+	log_must rm -f $DST_FILE
+
+	# Overwrite a random range of an existing file and immediately copy it.
+	sync_pool $TESTPOOL
+	log_must dd if=/dev/urandom of=$SRC_FILE bs=$((RECORDSIZE / 2)) \
+            seek=$(($RANDOM % $SRC_SIZE)) count=$(($RANDOM % 16)) conv=notrunc
+	if [[ "$mode" == "always" ]]; then
+		log_mustnot cp --reflink=$mode $SRC_FILE $DST_FILE
+		log_must ls -l $CP_TESTDIR
+	else
+		log_must cp --reflink=$mode $SRC_FILE $DST_FILE
+		verify_copy $SRC_FILE $DST_FILE
+	fi
+	log_must rm -f $SRC_FILE $DST_FILE
+done
+
+# Wait on dirty blocks (zfs_bclone_wait_dirty=1)
+log_must set_tunable32 BCLONE_WAIT_DIRTY 1
+
+for mode in "never" "auto" "always"; do
+	log_note "Checking 'cp --reflink=$mode'"
+
+	# Create a new file and immediately copy it.
+	log_must dd if=/dev/urandom of=$SRC_FILE bs=$RECORDSIZE count=$SRC_SIZE
+	log_must cp --reflink=$mode $SRC_FILE $DST_FILE
+	verify_copy $SRC_FILE $DST_FILE
+	log_must rm -f $DST_FILE
+
+	# Append to an existing file and immediately copy it.
+	log_must dd if=/dev/urandom of=$SRC_FILE bs=$RECORDSIZE seek=$SRC_SIZE \
+	    count=1 conv=notrunc
+	log_must cp --reflink=$mode $SRC_FILE $DST_FILE
+	verify_copy $SRC_FILE $DST_FILE
+	log_must rm -f $DST_FILE
+
+	# Overwrite a random range of an existing file and immediately copy it.
+	log_must dd if=/dev/urandom of=$SRC_FILE bs=$((RECORDSIZE / 2)) \
+            seek=$(($RANDOM % $SRC_SIZE)) count=$(($RANDOM % 16)) conv=notrunc
+	log_must cp --reflink=$mode $SRC_FILE $DST_FILE
+	verify_copy $SRC_FILE $DST_FILE
+	log_must rm -f $SRC_FILE $DST_FILE
+done
+
+log_pass
diff --git a/tests/zfs-tests/tests/functional/cp_files/setup.ksh b/tests/zfs-tests/tests/functional/cp_files/setup.ksh
index b756d4e76c..4223386b36 100755
--- a/tests/zfs-tests/tests/functional/cp_files/setup.ksh
+++ b/tests/zfs-tests/tests/functional/cp_files/setup.ksh
@@ -32,4 +32,10 @@
 . $STF_SUITE/include/libtest.shlib
 
 DISK=${DISKS%% *}
+
+if tunable_exists BCLONE_ENABLED ; then
+	log_must save_tunable BCLONE_ENABLED
+	log_must set_tunable32 BCLONE_ENABLED 1
+fi
+
 default_setup $DISK