Tag 2.2.0-rc4

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

META

@@ -2,9 +2,9 @@ Meta:          1
 Name:          zfs
 Branch:        1.0
 Version:       2.2.0
-Release:       rc1
+Release:       rc4
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS
-Linux-Maximum: 6.3
+Linux-Maximum: 6.4
 Linux-Minimum: 3.10

cmd/zdb/zdb.c

@@ -79,6 +79,7 @@
 #include <sys/dsl_crypt.h>
 #include <sys/dsl_scan.h>
 #include <sys/btree.h>
+#include <sys/brt.h>
 #include <zfs_comutil.h>
 #include <sys/zstd/zstd.h>
 
@@ -5342,12 +5343,20 @@ static const char *zdb_ot_extname[] = {
 #define	ZB_TOTAL	DN_MAX_LEVELS
 #define	SPA_MAX_FOR_16M	(SPA_MAXBLOCKSHIFT+1)
 
+typedef struct zdb_brt_entry {
+	dva_t		zbre_dva;
+	uint64_t	zbre_refcount;
+	avl_node_t	zbre_node;
+} zdb_brt_entry_t;
+
 typedef struct zdb_cb {
 	zdb_blkstats_t	zcb_type[ZB_TOTAL + 1][ZDB_OT_TOTAL + 1];
 	uint64_t	zcb_removing_size;
 	uint64_t	zcb_checkpoint_size;
 	uint64_t	zcb_dedup_asize;
 	uint64_t	zcb_dedup_blocks;
+	uint64_t	zcb_clone_asize;
+	uint64_t	zcb_clone_blocks;
 	uint64_t	zcb_psize_count[SPA_MAX_FOR_16M];
 	uint64_t	zcb_lsize_count[SPA_MAX_FOR_16M];
 	uint64_t	zcb_asize_count[SPA_MAX_FOR_16M];
@@ -5368,6 +5377,8 @@ typedef struct zdb_cb {
 	int		zcb_haderrors;
 	spa_t		*zcb_spa;
 	uint32_t	**zcb_vd_obsolete_counts;
+	avl_tree_t	zcb_brt;
+	boolean_t	zcb_brt_is_active;
 } zdb_cb_t;
 
 /* test if two DVA offsets from same vdev are within the same metaslab */
@@ -5662,6 +5673,45 @@ zdb_count_block(zdb_cb_t *zcb, zilog_t *zilog, const blkptr_t *bp,
 	zcb->zcb_asize_len[bin] += BP_GET_ASIZE(bp);
 	zcb->zcb_asize_total += BP_GET_ASIZE(bp);
 
+	if (zcb->zcb_brt_is_active && brt_maybe_exists(zcb->zcb_spa, bp)) {
+		/*
+		 * Cloned blocks are special. We need to count them, so we can
+		 * later uncount them when reporting leaked space, and we must
+		 * only claim them them once.
+		 *
+		 * To do this, we keep our own in-memory BRT. For each block
+		 * we haven't seen before, we look it up in the real BRT and
+		 * if its there, we note it and its refcount then proceed as
+		 * normal. If we see the block again, we count it as a clone
+		 * and then give it no further consideration.
+		 */
+		zdb_brt_entry_t zbre_search, *zbre;
+		avl_index_t where;
+
+		zbre_search.zbre_dva = bp->blk_dva[0];
+		zbre = avl_find(&zcb->zcb_brt, &zbre_search, &where);
+		if (zbre != NULL) {
+			zcb->zcb_clone_asize += BP_GET_ASIZE(bp);
+			zcb->zcb_clone_blocks++;
+
+			zbre->zbre_refcount--;
+			if (zbre->zbre_refcount == 0) {
+				avl_remove(&zcb->zcb_brt, zbre);
+				umem_free(zbre, sizeof (zdb_brt_entry_t));
+			}
+			return;
+		}
+
+		uint64_t crefcnt = brt_entry_get_refcount(zcb->zcb_spa, bp);
+		if (crefcnt > 0) {
+			zbre = umem_zalloc(sizeof (zdb_brt_entry_t),
+			    UMEM_NOFAIL);
+			zbre->zbre_dva = bp->blk_dva[0];
+			zbre->zbre_refcount = crefcnt;
+			avl_insert(&zcb->zcb_brt, zbre, where);
+		}
+	}
+
 	if (dump_opt['L'])
 		return;
 
@@ -6664,6 +6714,20 @@ deleted_livelists_dump_mos(spa_t *spa)
 	iterate_deleted_livelists(spa, dump_livelist_cb, NULL);
 }
 
+static int
+zdb_brt_entry_compare(const void *zcn1, const void *zcn2)
+{
+	const dva_t *dva1 = &((const zdb_brt_entry_t *)zcn1)->zbre_dva;
+	const dva_t *dva2 = &((const zdb_brt_entry_t *)zcn2)->zbre_dva;
+	int cmp;
+
+	cmp = TREE_CMP(DVA_GET_VDEV(dva1), DVA_GET_VDEV(dva2));
+	if (cmp == 0)
+		cmp = TREE_CMP(DVA_GET_OFFSET(dva1), DVA_GET_OFFSET(dva2));
+
+	return (cmp);
+}
+
 static int
 dump_block_stats(spa_t *spa)
 {
@@ -6678,6 +6742,13 @@ dump_block_stats(spa_t *spa)
 
 	zcb = umem_zalloc(sizeof (zdb_cb_t), UMEM_NOFAIL);
 
+	if (spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING)) {
+		avl_create(&zcb->zcb_brt, zdb_brt_entry_compare,
+		    sizeof (zdb_brt_entry_t),
+		    offsetof(zdb_brt_entry_t, zbre_node));
+		zcb->zcb_brt_is_active = B_TRUE;
+	}
+
 	(void) printf("\nTraversing all blocks %s%s%s%s%s...\n\n",
 	    (dump_opt['c'] || !dump_opt['L']) ? "to verify " : "",
 	    (dump_opt['c'] == 1) ? "metadata " : "",
@@ -6779,7 +6850,8 @@ dump_block_stats(spa_t *spa)
 	    metaslab_class_get_alloc(spa_special_class(spa)) +
 	    metaslab_class_get_alloc(spa_dedup_class(spa)) +
 	    get_unflushed_alloc_space(spa);
-	total_found = tzb->zb_asize - zcb->zcb_dedup_asize +
+	total_found =
+	    tzb->zb_asize - zcb->zcb_dedup_asize - zcb->zcb_clone_asize +
 	    zcb->zcb_removing_size + zcb->zcb_checkpoint_size;
 
 	if (total_found == total_alloc && !dump_opt['L']) {
@@ -6820,6 +6892,9 @@ dump_block_stats(spa_t *spa)
 	    "bp deduped:", (u_longlong_t)zcb->zcb_dedup_asize,
 	    (u_longlong_t)zcb->zcb_dedup_blocks,
 	    (double)zcb->zcb_dedup_asize / tzb->zb_asize + 1.0);
+	(void) printf("\t%-16s %14llu    count: %6llu\n",
+	    "bp cloned:", (u_longlong_t)zcb->zcb_clone_asize,
+	    (u_longlong_t)zcb->zcb_clone_blocks);
 	(void) printf("\t%-16s %14llu     used: %5.2f%%\n", "Normal class:",
 	    (u_longlong_t)norm_alloc, 100.0 * norm_alloc / norm_space);
 

cmd/zed/agents/zfs_mod.c

@@ -607,8 +607,6 @@ zfs_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *data)
 		 */
 		if (nvlist_lookup_string(nvl, dp->dd_prop, &path) != 0 ||
 		    strcmp(dp->dd_compare, path) != 0) {
-			zed_log_msg(LOG_INFO, "  %s: no match (%s != vdev %s)",
-			    __func__, dp->dd_compare, path);
 			return;
 		}
 		if (dp->dd_new_vdev_guid != 0 && dp->dd_new_vdev_guid != guid) {

cmd/zed/agents/zfs_retire.c

@@ -416,6 +416,11 @@ zfs_retire_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
 		    FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, &vdev_guid) != 0)
 			return;
 
+		if (vdev_guid == 0) {
+			fmd_hdl_debug(hdl, "Got a zero GUID");
+			return;
+		}
+
 		if (spare) {
 			int nspares = find_and_remove_spares(zhdl, vdev_guid);
 			fmd_hdl_debug(hdl, "%d spares removed", nspares);

cmd/zed/zed.d/Makefile.am

@@ -16,6 +16,7 @@ dist_zedexec_SCRIPTS = \
 	%D%/scrub_finish-notify.sh \
 	%D%/statechange-led.sh \
 	%D%/statechange-notify.sh \
+	%D%/statechange-slot_off.sh \
 	%D%/trim_finish-notify.sh \
 	%D%/vdev_attach-led.sh \
 	%D%/vdev_clear-led.sh
@@ -35,6 +36,7 @@ zedconfdefaults = \
 	scrub_finish-notify.sh \
 	statechange-led.sh \
 	statechange-notify.sh \
+	statechange-slot_off.sh \
 	vdev_attach-led.sh \
 	vdev_clear-led.sh
 

cmd/zed/zed.d/statechange-slot_off.sh

@@ -0,0 +1,64 @@
+#!/bin/sh
+# shellcheck disable=SC3014,SC2154,SC2086,SC2034
+#
+# Turn off disk's enclosure slot if it becomes FAULTED.
+#
+# 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
+# FAULTED, then power down the slot via sysfs:
+#
+# /sys/class/enclosure/<enclosure>/<slot>/power_status
+#
+# We assume the user will be responsible for turning the slot back on again.
+#
+# Note that this script requires that your enclosure be supported by the
+# Linux SCSI Enclosure services (SES) driver.  The script will do nothing
+# if you have no enclosure, or if your enclosure isn't supported.
+#
+# Exit codes:
+#   0: slot successfully powered off
+#   1: enclosure not available
+#   2: ZED_POWER_OFF_ENCLOUSRE_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
+
+[ -f "${ZED_ZEDLET_DIR}/zed.rc" ] && . "${ZED_ZEDLET_DIR}/zed.rc"
+. "${ZED_ZEDLET_DIR}/zed-functions.sh"
+
+if [ ! -d /sys/class/enclosure ] ; then
+	# No JBOD enclosure or NVMe slots
+	exit 1
+fi
+
+if [ "${ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT}" != "1" ] ; then
+	exit 2
+fi
+
+if [ "$ZEVENT_VDEV_STATE_STR" != "FAULTED" ] ; then
+	exit 3
+fi
+
+if [ ! -f "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status" ] ; then
+	exit 4
+fi
+
+# Turn off the slot and wait for sysfs to report that the slot is off.
+# It can take ~400ms on some enclosures and multiple retries may be needed.
+for i in $(seq 1 20) ; do
+	echo "off" | tee "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status"
+
+	for j in $(seq 1 5) ; do
+		if [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" == "off" ] ; then
+			break 2
+		fi
+		sleep 0.1
+	done
+done
+
+if [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" != "off" ] ; then
+	exit 5
+fi
+
+zed_log_msg "powered down slot $ZEVENT_VDEV_ENC_SYSFS_PATH for $ZEVENT_VDEV_PATH"

cmd/zed/zed.d/zed.rc

@@ -142,3 +142,8 @@ ZED_SYSLOG_SUBCLASS_EXCLUDE="history_event"
 # Disabled by default, 1 to enable and 0 to disable.
 #ZED_SYSLOG_DISPLAY_GUIDS=1
 
+##
+# 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

cmd/zpool/zpool_main.c

@@ -3143,6 +3143,7 @@ do_import(nvlist_t *config, const char *newname, const char *mntopts,
     nvlist_t *props, int flags)
 {
 	int ret = 0;
+	int ms_status = 0;
 	zpool_handle_t *zhp;
 	const char *name;
 	uint64_t version;
@@ -3232,10 +3233,15 @@ do_import(nvlist_t *config, const char *newname, const char *mntopts,
 			ret = 1;
 
 	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
-	    !(flags & ZFS_IMPORT_ONLY) &&
-	    zpool_enable_datasets(zhp, mntopts, 0) != 0) {
-		zpool_close(zhp);
-		return (1);
+	    !(flags & ZFS_IMPORT_ONLY)) {
+		ms_status = zpool_enable_datasets(zhp, mntopts, 0);
+		if (ms_status == EZFS_SHAREFAILED) {
+			(void) fprintf(stderr, gettext("Import was "
+			    "successful, but unable to share some datasets"));
+		} else if (ms_status == EZFS_MOUNTFAILED) {
+			(void) fprintf(stderr, gettext("Import was "
+			    "successful, but unable to mount some datasets"));
+		}
 	}
 
 	zpool_close(zhp);
@@ -6755,6 +6761,7 @@ zpool_do_split(int argc, char **argv)
 	char *mntopts = NULL;
 	splitflags_t flags;
 	int c, ret = 0;
+	int ms_status = 0;
 	boolean_t loadkeys = B_FALSE;
 	zpool_handle_t *zhp;
 	nvlist_t *config, *props = NULL;
@@ -6891,13 +6898,18 @@ zpool_do_split(int argc, char **argv)
 			ret = 1;
 	}
 
-	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
-	    zpool_enable_datasets(zhp, mntopts, 0) != 0) {
-		ret = 1;
-		(void) fprintf(stderr, gettext("Split was successful, but "
-		    "the datasets could not all be mounted\n"));
-		(void) fprintf(stderr, gettext("Try doing '%s' with a "
-		    "different altroot\n"), "zpool import");
+	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL) {
+		ms_status = zpool_enable_datasets(zhp, mntopts, 0);
+		if (ms_status == EZFS_SHAREFAILED) {
+			(void) fprintf(stderr, gettext("Split was successful, "
+			    "datasets are mounted but sharing of some datasets "
+			    "has failed\n"));
+		} else if (ms_status == EZFS_MOUNTFAILED) {
+			(void) fprintf(stderr, gettext("Split was successful"
+			    ", but some datasets could not be mounted\n"));
+			(void) fprintf(stderr, gettext("Try doing '%s' with a "
+			    "different altroot\n"), "zpool import");
+		}
 	}
 	zpool_close(zhp);
 	nvlist_free(config);

cmd/ztest.c

@@ -2412,7 +2412,6 @@ ztest_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
 	int error;
 
 	ASSERT3P(lwb, !=, NULL);
-	ASSERT3P(zio, !=, NULL);
 	ASSERT3U(size, !=, 0);
 
 	ztest_object_lock(zd, object, RL_READER);
@@ -2446,6 +2445,7 @@ ztest_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
 		    DMU_READ_NO_PREFETCH);
 		ASSERT0(error);
 	} else {
+		ASSERT3P(zio, !=, NULL);
 		size = doi.doi_data_block_size;
 		if (ISP2(size)) {
 			offset = P2ALIGN(offset, size);

config/Shellcheck.am

@@ -4,6 +4,7 @@
 # Not following: a was not specified as input (see shellcheck -x). [SC1091]
 # Prefer putting braces around variable references even when not strictly required. [SC2250]
 # Consider invoking this command separately to avoid masking its return value (or use '|| true' to ignore). [SC2312]
+# Command appears to be unreachable. Check usage (or ignore if invoked indirectly). [SC2317]
 # In POSIX sh, 'local' is undefined. [SC2039] # older ShellCheck versions
 # In POSIX sh, 'local' is undefined. [SC3043] # newer ShellCheck versions
 
@@ -18,7 +19,7 @@ PHONY += shellcheck
 _STGT = $(subst ^,/,$(subst shellcheck-here-,,$@))
 shellcheck-here-%:
 if HAVE_SHELLCHECK
-	shellcheck --format=gcc --enable=all --exclude=SC1090,SC1091,SC2039,SC2250,SC2312,SC3043 $$([ -n "$(SHELLCHECK_SHELL)" ] && echo "--shell=$(SHELLCHECK_SHELL)") "$$([ -e "$(_STGT)" ] || echo "$(srcdir)/")$(_STGT)"
+	shellcheck --format=gcc --enable=all --exclude=SC1090,SC1091,SC2039,SC2250,SC2312,SC2317,SC3043 $$([ -n "$(SHELLCHECK_SHELL)" ] && echo "--shell=$(SHELLCHECK_SHELL)") "$$([ -e "$(_STGT)" ] || echo "$(srcdir)/")$(_STGT)"
 else
 	@echo "skipping shellcheck of" $(_STGT) "because shellcheck is not installed"
 endif

config/kernel-blkdev.m4

@@ -103,6 +103,33 @@ AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_CHECK_DISK_CHANGE], [
 	])
 ])
 
+dnl #
+dnl # 6.5.x API change
+dnl # disk_check_media_change() was added
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_DISK_CHECK_MEDIA_CHANGE], [
+	ZFS_LINUX_TEST_SRC([disk_check_media_change], [
+		#include <linux/fs.h>
+		#include <linux/blkdev.h>
+	], [
+		struct block_device *bdev = NULL;
+		bool error;
+
+		error = disk_check_media_change(bdev->bd_disk);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_DISK_CHECK_MEDIA_CHANGE], [
+	AC_MSG_CHECKING([whether disk_check_media_change() exists])
+	ZFS_LINUX_TEST_RESULT([disk_check_media_change], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_DISK_CHECK_MEDIA_CHANGE, 1,
+		    [disk_check_media_change() exists])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])
+
 dnl #
 dnl # bdev_kobj() is introduced from 5.12
 dnl #
@@ -443,6 +470,29 @@ AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_GET_ERESTARTSYS], [
 	])
 ])
 
+dnl #
+dnl # 6.5.x API change
+dnl # BLK_STS_NEXUS replaced with BLK_STS_RESV_CONFLICT
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BLK_STS_RESV_CONFLICT], [
+	ZFS_LINUX_TEST_SRC([blk_sts_resv_conflict], [
+		#include <linux/blkdev.h>
+	],[
+		blk_status_t s __attribute__ ((unused)) = BLK_STS_RESV_CONFLICT;
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BLK_STS_RESV_CONFLICT], [
+	AC_MSG_CHECKING([whether BLK_STS_RESV_CONFLICT is defined])
+		ZFS_LINUX_TEST_RESULT([blk_sts_resv_conflict], [
+			AC_DEFINE(HAVE_BLK_STS_RESV_CONFLICT, 1, [BLK_STS_RESV_CONFLICT is defined])
+			AC_MSG_RESULT(yes)
+		], [
+			AC_MSG_RESULT(no)
+		])
+	])
+])
+
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV], [
 	ZFS_AC_KERNEL_SRC_BLKDEV_GET_BY_PATH
 	ZFS_AC_KERNEL_SRC_BLKDEV_PUT
@@ -458,6 +508,8 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV], [
 	ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE
 	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_KOBJ
 	ZFS_AC_KERNEL_SRC_BLKDEV_PART_TO_DEV
+	ZFS_AC_KERNEL_SRC_BLKDEV_DISK_CHECK_MEDIA_CHANGE
+	ZFS_AC_KERNEL_SRC_BLKDEV_BLK_STS_RESV_CONFLICT
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [
@@ -476,4 +528,6 @@ AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [
 	ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE
 	ZFS_AC_KERNEL_BLKDEV_BDEV_KOBJ
 	ZFS_AC_KERNEL_BLKDEV_PART_TO_DEV
+	ZFS_AC_KERNEL_BLKDEV_DISK_CHECK_MEDIA_CHANGE
+	ZFS_AC_KERNEL_BLKDEV_BLK_STS_RESV_CONFLICT
 ])

config/kernel-vfs-extended-file_range.m4

@@ -0,0 +1,50 @@
+dnl #
+dnl # EL7 have backported copy_file_range and clone_file_range and
+dnl # added them to an "extended" file_operations struct.
+dnl #
+dnl # We're testing for both functions in one here, because they will only
+dnl # ever appear together and we don't want to match a similar method in
+dnl # some future vendor kernel.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_FILE_OPERATIONS_EXTEND], [
+	ZFS_LINUX_TEST_SRC([vfs_file_operations_extend], [
+		#include <linux/fs.h>
+
+		static ssize_t test_copy_file_range(struct file *src_file,
+		    loff_t src_off, struct file *dst_file, loff_t dst_off,
+		    size_t len, unsigned int flags) {
+			(void) src_file; (void) src_off;
+			(void) dst_file; (void) dst_off;
+			(void) len; (void) flags;
+			return (0);
+		}
+
+		static int test_clone_file_range(struct file *src_file,
+		    loff_t src_off, struct file *dst_file, loff_t dst_off,
+		    u64 len) {
+			(void) src_file; (void) src_off;
+			(void) dst_file; (void) dst_off;
+			(void) len;
+			return (0);
+		}
+
+		static const struct file_operations_extend
+		    fops __attribute__ ((unused)) = {
+			.kabi_fops = {},
+			.copy_file_range = test_copy_file_range,
+			.clone_file_range = test_clone_file_range,
+		};
+	],[])
+])
+AC_DEFUN([ZFS_AC_KERNEL_VFS_FILE_OPERATIONS_EXTEND], [
+	AC_MSG_CHECKING([whether file_operations_extend takes \
+.copy_file_range() and .clone_file_range()])
+	ZFS_LINUX_TEST_RESULT([vfs_file_operations_extend], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_FILE_OPERATIONS_EXTEND, 1,
+		    [file_operations_extend takes .copy_file_range()
+		    and .clone_file_range()])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel-vfs-file_range.m4

@@ -0,0 +1,164 @@
+dnl #
+dnl # The *_file_range APIs have a long history:
+dnl #
+dnl # 2.6.29: BTRFS_IOC_CLONE and BTRFS_IOC_CLONE_RANGE ioctl introduced
+dnl # 3.12: BTRFS_IOC_FILE_EXTENT_SAME ioctl introduced
+dnl #
+dnl # 4.5: copy_file_range() syscall introduced, added to VFS
+dnl # 4.5: BTRFS_IOC_CLONE and BTRFS_IOC_CLONE_RANGE renamed to FICLONE ands
+dnl #      FICLONERANGE, added to VFS as clone_file_range()
+dnl # 4.5: BTRFS_IOC_FILE_EXTENT_SAME renamed to FIDEDUPERANGE, added to VFS
+dnl #      as dedupe_file_range()
+dnl #
+dnl # 4.20: VFS clone_file_range() and dedupe_file_range() replaced by
+dnl #       remap_file_range()
+dnl #
+dnl # 5.3: VFS copy_file_range() expected to do its own fallback,
+dnl #      generic_copy_file_range() added to support it
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_COPY_FILE_RANGE], [
+	ZFS_LINUX_TEST_SRC([vfs_copy_file_range], [
+		#include <linux/fs.h>
+
+		static ssize_t test_copy_file_range(struct file *src_file,
+		    loff_t src_off, struct file *dst_file, loff_t dst_off,
+		    size_t len, unsigned int flags) {
+			(void) src_file; (void) src_off;
+			(void) dst_file; (void) dst_off;
+			(void) len; (void) flags;
+			return (0);
+		}
+
+		static const struct file_operations
+		    fops __attribute__ ((unused)) = {
+			.copy_file_range	= test_copy_file_range,
+		};
+	],[])
+])
+AC_DEFUN([ZFS_AC_KERNEL_VFS_COPY_FILE_RANGE], [
+	AC_MSG_CHECKING([whether fops->copy_file_range() is available])
+	ZFS_LINUX_TEST_RESULT([vfs_copy_file_range], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_COPY_FILE_RANGE, 1,
+		    [fops->copy_file_range() is available])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_GENERIC_COPY_FILE_RANGE], [
+	ZFS_LINUX_TEST_SRC([generic_copy_file_range], [
+		#include <linux/fs.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;
+		unsigned int flags __attribute__ ((unused)) = 0;
+		generic_copy_file_range(src_file, src_off, dst_file, dst_off,
+		    len, flags);
+	])
+])
+AC_DEFUN([ZFS_AC_KERNEL_VFS_GENERIC_COPY_FILE_RANGE], [
+	AC_MSG_CHECKING([whether generic_copy_file_range() is available])
+	ZFS_LINUX_TEST_RESULT_SYMBOL([generic_copy_file_range],
+	[generic_copy_file_range], [fs/read_write.c], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_VFS_GENERIC_COPY_FILE_RANGE, 1,
+		    [generic_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>
+
+		static int test_clone_file_range(struct file *src_file,
+		    loff_t src_off, struct file *dst_file, loff_t dst_off,
+		    u64 len) {
+			(void) src_file; (void) src_off;
+			(void) dst_file; (void) dst_off;
+			(void) len;
+			return (0);
+		}
+
+		static const struct file_operations
+		    fops __attribute__ ((unused)) = {
+			.clone_file_range	= test_clone_file_range,
+		};
+	],[])
+])
+AC_DEFUN([ZFS_AC_KERNEL_VFS_CLONE_FILE_RANGE], [
+	AC_MSG_CHECKING([whether fops->clone_file_range() is available])
+	ZFS_LINUX_TEST_RESULT([vfs_clone_file_range], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_CLONE_FILE_RANGE, 1,
+		    [fops->clone_file_range() is available])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_DEDUPE_FILE_RANGE], [
+	ZFS_LINUX_TEST_SRC([vfs_dedupe_file_range], [
+		#include <linux/fs.h>
+
+		static int test_dedupe_file_range(struct file *src_file,
+		    loff_t src_off, struct file *dst_file, loff_t dst_off,
+		    u64 len) {
+			(void) src_file; (void) src_off;
+			(void) dst_file; (void) dst_off;
+			(void) len;
+			return (0);
+		}
+
+		static const struct file_operations
+		    fops __attribute__ ((unused)) = {
+                .dedupe_file_range	= test_dedupe_file_range,
+		};
+	],[])
+])
+AC_DEFUN([ZFS_AC_KERNEL_VFS_DEDUPE_FILE_RANGE], [
+	AC_MSG_CHECKING([whether fops->dedupe_file_range() is available])
+	ZFS_LINUX_TEST_RESULT([vfs_dedupe_file_range], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_DEDUPE_FILE_RANGE, 1,
+		    [fops->dedupe_file_range() is available])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_REMAP_FILE_RANGE], [
+	ZFS_LINUX_TEST_SRC([vfs_remap_file_range], [
+		#include <linux/fs.h>
+
+		static loff_t test_remap_file_range(struct file *src_file,
+		    loff_t src_off, struct file *dst_file, loff_t dst_off,
+		    loff_t len, unsigned int flags) {
+			(void) src_file; (void) src_off;
+			(void) dst_file; (void) dst_off;
+			(void) len; (void) flags;
+			return (0);
+		}
+
+		static const struct file_operations
+		    fops __attribute__ ((unused)) = {
+			.remap_file_range	= test_remap_file_range,
+		};
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_VFS_REMAP_FILE_RANGE], [
+	AC_MSG_CHECKING([whether fops->remap_file_range() is available])
+	ZFS_LINUX_TEST_RESULT([vfs_remap_file_range], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_REMAP_FILE_RANGE, 1,
+		    [fops->remap_file_range() is available])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel.m4

@@ -116,6 +116,12 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_VFS_RW_ITERATE
 	ZFS_AC_KERNEL_SRC_VFS_GENERIC_WRITE_CHECKS
 	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_REMAP_FILE_RANGE
+	ZFS_AC_KERNEL_SRC_VFS_CLONE_FILE_RANGE
+	ZFS_AC_KERNEL_SRC_VFS_DEDUPE_FILE_RANGE
+	ZFS_AC_KERNEL_SRC_VFS_FILE_OPERATIONS_EXTEND
 	ZFS_AC_KERNEL_SRC_KMAP_ATOMIC_ARGS
 	ZFS_AC_KERNEL_SRC_FOLLOW_DOWN_ONE
 	ZFS_AC_KERNEL_SRC_MAKE_REQUEST_FN
@@ -249,6 +255,12 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_VFS_RW_ITERATE
 	ZFS_AC_KERNEL_VFS_GENERIC_WRITE_CHECKS
 	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_REMAP_FILE_RANGE
+	ZFS_AC_KERNEL_VFS_CLONE_FILE_RANGE
+	ZFS_AC_KERNEL_VFS_DEDUPE_FILE_RANGE
+	ZFS_AC_KERNEL_VFS_FILE_OPERATIONS_EXTEND
 	ZFS_AC_KERNEL_KMAP_ATOMIC_ARGS
 	ZFS_AC_KERNEL_FOLLOW_DOWN_ONE
 	ZFS_AC_KERNEL_MAKE_REQUEST_FN

contrib/debian/changelog

@@ -1,3 +1,9 @@
+openzfs-linux (2.2.0-0) unstable; urgency=low
+
+  * OpenZFS 2.2.0 is tagged.
+
+ -- Umer Saleem <usaleem@ixsystems.com>  Tue, 25 Jul 2023 15:00:00 +0500
+
 openzfs-linux (2.1.99-1) unstable; urgency=low
 
   * Integrate minimally modified Debian packaging from ZFS on Linux

contrib/debian/openzfs-zfs-test.install

@@ -1,10 +1,8 @@
-sbin/zinject
 sbin/ztest
 usr/bin/raidz_test
 usr/share/man/man1/raidz_test.1
 usr/share/man/man1/test-runner.1
 usr/share/man/man1/ztest.1
-usr/share/man/man8/zinject.8
 usr/share/zfs/common.sh
 usr/share/zfs/runfiles/
 usr/share/zfs/test-runner

contrib/debian/openzfs-zfsutils.install

@@ -27,6 +27,7 @@ sbin/zfs
 sbin/zfs_ids_to_path
 sbin/zgenhostid
 sbin/zhack
+sbin/zinject
 sbin/zpool
 sbin/zstream
 sbin/zstreamdump
@@ -59,7 +60,6 @@ usr/share/man/man8/zfs-get.8
 usr/share/man/man8/zfs-groupspace.8
 usr/share/man/man8/zfs-hold.8
 usr/share/man/man8/zfs-inherit.8
-usr/share/man/man8/zfs-jail.8
 usr/share/man/man8/zfs-list.8
 usr/share/man/man8/zfs-load-key.8
 usr/share/man/man8/zfs-mount-generator.8
@@ -79,7 +79,6 @@ usr/share/man/man8/zfs-set.8
 usr/share/man/man8/zfs-share.8
 usr/share/man/man8/zfs-snapshot.8
 usr/share/man/man8/zfs-unallow.8
-usr/share/man/man8/zfs-unjail.8
 usr/share/man/man8/zfs-unload-key.8
 usr/share/man/man8/zfs-unmount.8
 usr/share/man/man8/zfs-unzone.8
@@ -92,6 +91,7 @@ usr/share/man/man8/zfs_ids_to_path.8
 usr/share/man/man7/zfsconcepts.7
 usr/share/man/man7/zfsprops.7
 usr/share/man/man8/zgenhostid.8
+usr/share/man/man8/zinject.8
 usr/share/man/man8/zpool-add.8
 usr/share/man/man8/zpool-attach.8
 usr/share/man/man8/zpool-checkpoint.8

contrib/dracut/90zfs/zfs-env-bootfs.service.in

@@ -12,11 +12,12 @@ ExecStart=/bin/sh -c '
     decode_root_args || exit 0;                                                                \
     [ "$root" = "zfs:AUTO" ] && root="$(@sbindir@/zpool list -H -o bootfs | grep -m1 -vFx -)"; \
     rootflags="$(getarg rootflags=)";                                                          \
-    case ",$rootflags," in                                                                     \
-        *,zfsutil,*) ;;                                                                        \
-        ,,) rootflags=zfsutil ;;                                                               \
-        *)  rootflags="zfsutil,$rootflags" ;;                                                  \
-    esac;                                                                                      \
+    [ "$(@sbindir@/zfs get -H -o value mountpoint "$root")" = legacy ] ||                      \
+        case ",$rootflags," in                                                                 \
+            *,zfsutil,*) ;;                                                                    \
+            ,,) rootflags=zfsutil ;;                                                           \
+            *)  rootflags="zfsutil,$rootflags" ;;                                              \
+        esac;                                                                                  \
     exec systemctl set-environment BOOTFS="$root" BOOTFSFLAGS="$rootflags"'
 
 [Install]

contrib/dracut/90zfs/zfs-rollback-bootfs.service.in

@@ -2,7 +2,7 @@
 Description=Rollback bootfs just before it is mounted
 Requisite=zfs-import.target
 After=zfs-import.target dracut-pre-mount.service zfs-snapshot-bootfs.service
-Before=dracut-mount.service
+Before=dracut-mount.service sysroot.mount
 DefaultDependencies=no
 ConditionKernelCommandLine=bootfs.rollback
 ConditionEnvironment=BOOTFS

include/libzfs.h

@@ -156,6 +156,7 @@ typedef enum zfs_error {
 	EZFS_NOT_USER_NAMESPACE,	/* a file is not a user namespace */
 	EZFS_CKSUM,		/* insufficient replicas */
 	EZFS_RESUME_EXISTS,	/* Resume on existing dataset without force */
+	EZFS_SHAREFAILED,	/* filesystem share failed */
 	EZFS_UNKNOWN
 } zfs_error_t;
 

include/os/freebsd/spl/sys/vnode.h

@@ -36,7 +36,11 @@ struct xucred;
 typedef struct flock	flock64_t;
 typedef	struct vnode	vnode_t;
 typedef	struct vattr	vattr_t;
+#if __FreeBSD_version < 1400093
 typedef enum vtype vtype_t;
+#else
+#define	vtype_t __enum_uint8(vtype)
+#endif
 
 #include <sys/types.h>
 #include <sys/queue.h>

include/os/linux/kernel/linux/blkdev_compat.h

@@ -181,7 +181,11 @@ bi_status_to_errno(blk_status_t status)
 		return (ENOLINK);
 	case BLK_STS_TARGET:
 		return (EREMOTEIO);
+#ifdef HAVE_BLK_STS_RESV_CONFLICT
+	case BLK_STS_RESV_CONFLICT:
+#else
 	case BLK_STS_NEXUS:
+#endif
 		return (EBADE);
 	case BLK_STS_MEDIUM:
 		return (ENODATA);
@@ -215,7 +219,11 @@ errno_to_bi_status(int error)
 	case EREMOTEIO:
 		return (BLK_STS_TARGET);
 	case EBADE:
+#ifdef HAVE_BLK_STS_RESV_CONFLICT
+		return (BLK_STS_RESV_CONFLICT);
+#else
 		return (BLK_STS_NEXUS);
+#endif
 	case ENODATA:
 		return (BLK_STS_MEDIUM);
 	case EILSEQ:
@@ -337,6 +345,8 @@ zfs_check_media_change(struct block_device *bdev)
 	return (0);
 }
 #define	vdev_bdev_reread_part(bdev)	zfs_check_media_change(bdev)
+#elif defined(HAVE_DISK_CHECK_MEDIA_CHANGE)
+#define	vdev_bdev_reread_part(bdev)	disk_check_media_change(bdev->bd_disk)
 #else
 /*
  * This is encountered if check_disk_change() and bdev_check_media_change()

include/os/linux/kernel/linux/simd_x86.h

@@ -147,6 +147,15 @@
 #error "Toolchain needs to support the XSAVE assembler instruction"
 #endif
 
+#ifndef XFEATURE_MASK_XTILE
+/*
+ * For kernels where this doesn't exist yet, we still don't want to break
+ * by save/restoring this broken nonsense.
+ * See issue #14989 or Intel errata SPR4 for why
+ */
+#define	XFEATURE_MASK_XTILE	0x60000
+#endif
+
 #include <linux/mm.h>
 #include <linux/slab.h>
 
@@ -315,18 +324,18 @@ kfpu_begin(void)
 	uint8_t *state = zfs_kfpu_fpregs[smp_processor_id()];
 #if defined(HAVE_XSAVES)
 	if (static_cpu_has(X86_FEATURE_XSAVES)) {
-		kfpu_do_xsave("xsaves", state, ~0);
+		kfpu_do_xsave("xsaves", state, ~XFEATURE_MASK_XTILE);
 		return;
 	}
 #endif
 #if defined(HAVE_XSAVEOPT)
 	if (static_cpu_has(X86_FEATURE_XSAVEOPT)) {
-		kfpu_do_xsave("xsaveopt", state, ~0);
+		kfpu_do_xsave("xsaveopt", state, ~XFEATURE_MASK_XTILE);
 		return;
 	}
 #endif
 	if (static_cpu_has(X86_FEATURE_XSAVE)) {
-		kfpu_do_xsave("xsave", state, ~0);
+		kfpu_do_xsave("xsave", state, ~XFEATURE_MASK_XTILE);
 	} else if (static_cpu_has(X86_FEATURE_FXSR)) {
 		kfpu_save_fxsr(state);
 	} else {
@@ -376,12 +385,12 @@ kfpu_end(void)
 	uint8_t  *state = zfs_kfpu_fpregs[smp_processor_id()];
 #if defined(HAVE_XSAVES)
 	if (static_cpu_has(X86_FEATURE_XSAVES)) {
-		kfpu_do_xrstor("xrstors", state, ~0);
+		kfpu_do_xrstor("xrstors", state, ~XFEATURE_MASK_XTILE);
 		goto out;
 	}
 #endif
 	if (static_cpu_has(X86_FEATURE_XSAVE)) {
-		kfpu_do_xrstor("xrstor", state, ~0);
+		kfpu_do_xrstor("xrstor", state, ~XFEATURE_MASK_XTILE);
 	} else if (static_cpu_has(X86_FEATURE_FXSR)) {
 		kfpu_restore_fxsr(state);
 	} else {

include/os/linux/spl/sys/kmem_cache.h

@@ -198,6 +198,14 @@ extern uint64_t spl_kmem_cache_entry_size(kmem_cache_t *cache);
     spl_kmem_cache_create(name, size, align, ctor, dtor, rclm, priv, vmp, fl)
 #define	kmem_cache_set_move(skc, move)	spl_kmem_cache_set_move(skc, move)
 #define	kmem_cache_destroy(skc)		spl_kmem_cache_destroy(skc)
+/*
+ * This is necessary to be compatible with other kernel modules
+ * or in-tree filesystem that may define kmem_cache_alloc,
+ * like bcachefs does it now.
+ */
+#ifdef kmem_cache_alloc
+#undef kmem_cache_alloc
+#endif
 #define	kmem_cache_alloc(skc, flags)	spl_kmem_cache_alloc(skc, flags)
 #define	kmem_cache_free(skc, obj)	spl_kmem_cache_free(skc, obj)
 #define	kmem_cache_reap_now(skc)	spl_kmem_cache_reap_now(skc)

include/os/linux/spl/sys/types.h

@@ -38,7 +38,7 @@ typedef unsigned long		ulong_t;
 typedef unsigned long long	u_longlong_t;
 typedef long long		longlong_t;
 
-typedef unsigned long		intptr_t;
+typedef long			intptr_t;
 typedef unsigned long long	rlim64_t;
 
 typedef struct task_struct	kthread_t;

include/os/linux/zfs/sys/zpl.h

@@ -52,7 +52,11 @@ extern const struct inode_operations zpl_special_inode_operations;
 
 /* zpl_file.c */
 extern const struct address_space_operations zpl_address_space_operations;
+#ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND
+extern const struct file_operations_extend zpl_file_operations;
+#else
 extern const struct file_operations zpl_file_operations;
+#endif
 extern const struct file_operations zpl_dir_file_operations;
 
 /* zpl_super.c */
@@ -180,6 +184,55 @@ zpl_dir_emit_dots(struct file *file, zpl_dir_context_t *ctx)
 }
 #endif /* HAVE_VFS_ITERATE */
 
+
+/* zpl_file_range.c */
+
+/* handlers for file_operations of the same name */
+extern ssize_t zpl_copy_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, size_t len, unsigned int flags);
+extern loff_t zpl_remap_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, loff_t len, unsigned int flags);
+extern int zpl_clone_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, uint64_t len);
+extern int zpl_dedupe_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, uint64_t len);
+
+/* compat for FICLONE/FICLONERANGE/FIDEDUPERANGE ioctls */
+typedef struct {
+	int64_t		fcr_src_fd;
+	uint64_t	fcr_src_offset;
+	uint64_t	fcr_src_length;
+	uint64_t	fcr_dest_offset;
+} zfs_ioc_compat_file_clone_range_t;
+
+typedef struct {
+	int64_t		fdri_dest_fd;
+	uint64_t	fdri_dest_offset;
+	uint64_t	fdri_bytes_deduped;
+	int32_t		fdri_status;
+	uint32_t	fdri_reserved;
+} zfs_ioc_compat_dedupe_range_info_t;
+
+typedef struct {
+	uint64_t	fdr_src_offset;
+	uint64_t	fdr_src_length;
+	uint16_t	fdr_dest_count;
+	uint16_t	fdr_reserved1;
+	uint32_t	fdr_reserved2;
+	zfs_ioc_compat_dedupe_range_info_t	fdr_info[];
+} zfs_ioc_compat_dedupe_range_t;
+
+#define	ZFS_IOC_COMPAT_FICLONE		_IOW(0x94, 9, int)
+#define	ZFS_IOC_COMPAT_FICLONERANGE \
+    _IOW(0x94, 13, zfs_ioc_compat_file_clone_range_t)
+#define	ZFS_IOC_COMPAT_FIDEDUPERANGE \
+    _IOWR(0x94, 54, zfs_ioc_compat_dedupe_range_t)
+
+extern long zpl_ioctl_ficlone(struct file *filp, void *arg);
+extern long zpl_ioctl_ficlonerange(struct file *filp, void *arg);
+extern long zpl_ioctl_fideduperange(struct file *filp, void *arg);
+
+
 #if defined(HAVE_INODE_TIMESTAMP_TRUNCATE)
 #define	zpl_inode_timestamp_truncate(ts, ip)	timestamp_truncate(ts, ip)
 #elif defined(HAVE_INODE_TIMESPEC64_TIMES)

include/sys/bpobj.h

@@ -60,7 +60,7 @@ typedef struct bpobj {
 	kmutex_t	bpo_lock;
 	objset_t	*bpo_os;
 	uint64_t	bpo_object;
-	int		bpo_epb;
+	uint32_t	bpo_epb;
 	uint8_t		bpo_havecomp;
 	uint8_t		bpo_havesubobj;
 	uint8_t		bpo_havefreed;

include/sys/brt.h

@@ -36,6 +36,7 @@ extern "C" {
 #endif
 
 extern boolean_t brt_entry_decref(spa_t *spa, const blkptr_t *bp);
+extern uint64_t brt_entry_get_refcount(spa_t *spa, const blkptr_t *bp);
 
 extern uint64_t brt_get_dspace(spa_t *spa);
 extern uint64_t brt_get_used(spa_t *spa);

include/sys/dmu.h

@@ -572,11 +572,15 @@ int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
 int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
     uint64_t length, int read, const void *tag, int *numbufsp,
     dmu_buf_t ***dbpp);
+int dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset,
+    const void *tag, dmu_buf_t **dbp);
 int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
     const void *tag, dmu_buf_t **dbp, int flags);
 int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset,
     uint64_t length, boolean_t read, const void *tag, int *numbufsp,
     dmu_buf_t ***dbpp, uint32_t flags);
+int dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset, const void *tag,
+    dmu_buf_t **dbp);
 /*
  * Add a reference to a dmu buffer that has already been held via
  * dmu_buf_hold() in the current context.

include/sys/dmu_impl.h

@@ -247,8 +247,6 @@ typedef struct dmu_sendstatus {
 
 void dmu_object_zapify(objset_t *, uint64_t, dmu_object_type_t, dmu_tx_t *);
 void dmu_object_free_zapified(objset_t *, uint64_t, dmu_tx_t *);
-int dmu_buf_hold_noread(objset_t *, uint64_t, uint64_t,
-    const void *, dmu_buf_t **);
 
 #ifdef	__cplusplus
 }

include/sys/dmu_zfetch.h

@@ -36,8 +36,6 @@
 extern "C" {
 #endif
 
-extern uint64_t	zfetch_array_rd_sz;
-
 struct dnode;				/* so we can reference dnode */
 
 typedef struct zfetch {

include/sys/fm/fs/zfs.h

@@ -102,8 +102,6 @@ extern "C" {
 #define	FM_EREPORT_PAYLOAD_ZFS_ZIO_TIMESTAMP	"zio_timestamp"
 #define	FM_EREPORT_PAYLOAD_ZFS_ZIO_DELTA	"zio_delta"
 #define	FM_EREPORT_PAYLOAD_ZFS_PREV_STATE	"prev_state"
-#define	FM_EREPORT_PAYLOAD_ZFS_CKSUM_EXPECTED	"cksum_expected"
-#define	FM_EREPORT_PAYLOAD_ZFS_CKSUM_ACTUAL	"cksum_actual"
 #define	FM_EREPORT_PAYLOAD_ZFS_CKSUM_ALGO	"cksum_algorithm"
 #define	FM_EREPORT_PAYLOAD_ZFS_CKSUM_BYTESWAP	"cksum_byteswap"
 #define	FM_EREPORT_PAYLOAD_ZFS_BAD_OFFSET_RANGES "bad_ranges"
@@ -112,8 +110,6 @@ extern "C" {
 #define	FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_CLEARS	"bad_range_clears"
 #define	FM_EREPORT_PAYLOAD_ZFS_BAD_SET_BITS	"bad_set_bits"
 #define	FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_BITS	"bad_cleared_bits"
-#define	FM_EREPORT_PAYLOAD_ZFS_BAD_SET_HISTOGRAM "bad_set_histogram"
-#define	FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_HISTOGRAM "bad_cleared_histogram"
 #define	FM_EREPORT_PAYLOAD_ZFS_SNAPSHOT_NAME	"snapshot_name"
 #define	FM_EREPORT_PAYLOAD_ZFS_DEVICE_NAME	"device_name"
 #define	FM_EREPORT_PAYLOAD_ZFS_RAW_DEVICE_NAME	"raw_name"

include/sys/metaslab.h

@@ -80,7 +80,6 @@ uint64_t metaslab_largest_allocatable(metaslab_t *);
 #define	METASLAB_ASYNC_ALLOC		0x8
 #define	METASLAB_DONT_THROTTLE		0x10
 #define	METASLAB_MUST_RESERVE		0x20
-#define	METASLAB_FASTWRITE		0x40
 #define	METASLAB_ZIL			0x80
 
 int metaslab_alloc(spa_t *, metaslab_class_t *, uint64_t,
@@ -96,8 +95,6 @@ void metaslab_unalloc_dva(spa_t *, const dva_t *, uint64_t);
 int metaslab_claim(spa_t *, const blkptr_t *, uint64_t);
 int metaslab_claim_impl(vdev_t *, uint64_t, uint64_t, uint64_t);
 void metaslab_check_free(spa_t *, const blkptr_t *);
-void metaslab_fastwrite_mark(spa_t *, const blkptr_t *);
-void metaslab_fastwrite_unmark(spa_t *, const blkptr_t *);
 
 void metaslab_stat_init(void);
 void metaslab_stat_fini(void);

include/sys/metaslab_impl.h

@@ -313,7 +313,7 @@ struct metaslab_group {
  * Each metaslab maintains a set of in-core trees to track metaslab
  * operations.  The in-core free tree (ms_allocatable) contains the list of
  * free segments which are eligible for allocation.  As blocks are
- * allocated, the allocated segment are removed from the ms_allocatable and
+ * allocated, the allocated segments are removed from the ms_allocatable and
  * added to a per txg allocation tree (ms_allocating).  As blocks are
  * freed, they are added to the free tree (ms_freeing).  These trees
  * allow us to process all allocations and frees in syncing context
@@ -366,9 +366,9 @@ struct metaslab_group {
 struct metaslab {
 	/*
 	 * This is the main lock of the metaslab and its purpose is to
-	 * coordinate our allocations and frees [e.g metaslab_block_alloc(),
+	 * coordinate our allocations and frees [e.g., metaslab_block_alloc(),
 	 * metaslab_free_concrete(), ..etc] with our various syncing
-	 * procedures [e.g. metaslab_sync(), metaslab_sync_done(), ..etc].
+	 * procedures [e.g., metaslab_sync(), metaslab_sync_done(), ..etc].
 	 *
 	 * The lock is also used during some miscellaneous operations like
 	 * using the metaslab's histogram for the metaslab group's histogram

include/sys/spa.h

@@ -723,16 +723,10 @@ typedef enum spa_mode {
  * Send TRIM commands in-line during normal pool operation while deleting.
  *	OFF: no
  *	ON: yes
- * NB: IN_FREEBSD_BASE is defined within the FreeBSD sources.
  */
 typedef enum {
 	SPA_AUTOTRIM_OFF = 0,	/* default */
 	SPA_AUTOTRIM_ON,
-#ifdef IN_FREEBSD_BASE
-	SPA_AUTOTRIM_DEFAULT = SPA_AUTOTRIM_ON,
-#else
-	SPA_AUTOTRIM_DEFAULT = SPA_AUTOTRIM_OFF,
-#endif
 } spa_autotrim_t;
 
 /*

include/sys/spa_impl.h

@@ -250,6 +250,7 @@ struct spa {
 	uint64_t	spa_min_ashift;		/* of vdevs in normal class */
 	uint64_t	spa_max_ashift;		/* of vdevs in normal class */
 	uint64_t	spa_min_alloc;		/* of vdevs in normal class */
+	uint64_t	spa_gcd_alloc;		/* of vdevs in normal class */
 	uint64_t	spa_config_guid;	/* config pool guid */
 	uint64_t	spa_load_guid;		/* spa_load initialized guid */
 	uint64_t	spa_last_synced_guid;	/* last synced guid */

include/sys/vdev_impl.h

@@ -266,7 +266,6 @@ struct vdev {
 	metaslab_group_t *vdev_mg;	/* metaslab group		*/
 	metaslab_group_t *vdev_log_mg;	/* embedded slog metaslab group	*/
 	metaslab_t	**vdev_ms;	/* metaslab array		*/
-	uint64_t	vdev_pending_fastwrite; /* allocated fastwrites */
 	txg_list_t	vdev_ms_list;	/* per-txg dirty metaslab lists	*/
 	txg_list_t	vdev_dtl_list;	/* per-txg dirty DTL lists	*/
 	txg_node_t	vdev_txg_node;	/* per-txg dirty vdev linkage	*/
@@ -420,6 +419,7 @@ struct vdev {
 	boolean_t	vdev_copy_uberblocks;  /* post expand copy uberblocks */
 	boolean_t	vdev_resilver_deferred;  /* resilver deferred */
 	boolean_t	vdev_kobj_flag; /* kobj event record */
+	boolean_t	vdev_attaching; /* vdev attach ashift handling */
 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache and spares vdevs	*/
 	zio_t		*vdev_probe_zio; /* root of current probe	*/

include/sys/zil_impl.h

@@ -38,14 +38,22 @@ extern "C" {
 /*
  * Possible states for a given lwb structure.
  *
- * An lwb will start out in the "closed" state, and then transition to
- * the "opened" state via a call to zil_lwb_write_open(). When
- * transitioning from "closed" to "opened" the zilog's "zl_issuer_lock"
- * must be held.
+ * An lwb will start out in the "new" state, and transition to the "opened"
+ * state via a call to zil_lwb_write_open() on first itx assignment.  When
+ * transitioning from "new" to "opened" the zilog's "zl_issuer_lock" must be
+ * held.
  *
- * After the lwb is "opened", it can transition into the "issued" state
- * via zil_lwb_write_close(). Again, the zilog's "zl_issuer_lock" must
- * be held when making this transition.
+ * After the lwb is "opened", it can be assigned number of itxs and transition
+ * into the "closed" state via zil_lwb_write_close() when full or on timeout.
+ * When transitioning from "opened" to "closed" the zilog's "zl_issuer_lock"
+ * must be held.  New lwb allocation also takes "zl_lock" to protect the list.
+ *
+ * After the lwb is "closed", it can transition into the "ready" state via
+ * zil_lwb_write_issue().  "zl_lock" must be held when making this transition.
+ * Since it is done by the same thread, "zl_issuer_lock" is not needed.
+ *
+ * When lwb in "ready" state receives its block pointer, it can transition to
+ * "issued". "zl_lock" must be held when making this transition.
  *
  * After the lwb's write zio completes, it transitions into the "write
  * done" state via zil_lwb_write_done(); and then into the "flush done"
@@ -62,17 +70,20 @@ extern "C" {
  *
  * Additionally, correctness when reading an lwb's state is often
  * achieved by exploiting the fact that these state transitions occur in
- * this specific order; i.e. "closed" to "opened" to "issued" to "done".
+ * this specific order; i.e. "new" to "opened" to "closed" to "ready" to
+ * "issued" to "write_done" and finally "flush_done".
  *
- * Thus, if an lwb is in the "closed" or "opened" state, holding the
+ * Thus, if an lwb is in the "new" or "opened" state, holding the
  * "zl_issuer_lock" will prevent a concurrent thread from transitioning
- * that lwb to the "issued" state. Likewise, if an lwb is already in the
- * "issued" state, holding the "zl_lock" will prevent a concurrent
- * thread from transitioning that lwb to the "write done" state.
+ * that lwb to the "closed" state. Likewise, if an lwb is already in the
+ * "ready" state, holding the "zl_lock" will prevent a concurrent thread
+ * from transitioning that lwb to the "issued" state.
  */
 typedef enum {
-    LWB_STATE_CLOSED,
+    LWB_STATE_NEW,
     LWB_STATE_OPENED,
+    LWB_STATE_CLOSED,
+    LWB_STATE_READY,
     LWB_STATE_ISSUED,
     LWB_STATE_WRITE_DONE,
     LWB_STATE_FLUSH_DONE,
@@ -91,18 +102,21 @@ typedef enum {
 typedef struct lwb {
 	zilog_t		*lwb_zilog;	/* back pointer to log struct */
 	blkptr_t	lwb_blk;	/* on disk address of this log blk */
-	boolean_t	lwb_fastwrite;	/* is blk marked for fastwrite? */
+	boolean_t	lwb_slim;	/* log block has slim format */
 	boolean_t	lwb_slog;	/* lwb_blk is on SLOG device */
-	boolean_t	lwb_indirect;	/* do not postpone zil_lwb_commit() */
+	int		lwb_error;	/* log block allocation error */
+	int		lwb_nmax;	/* max bytes in the buffer */
 	int		lwb_nused;	/* # used bytes in buffer */
 	int		lwb_nfilled;	/* # filled bytes in buffer */
 	int		lwb_sz;		/* size of block and buffer */
 	lwb_state_t	lwb_state;	/* the state of this lwb */
 	char		*lwb_buf;	/* log write buffer */
+	zio_t		*lwb_child_zio;	/* parent zio for children */
 	zio_t		*lwb_write_zio;	/* zio for the lwb buffer */
 	zio_t		*lwb_root_zio;	/* root zio for lwb write and flushes */
 	hrtime_t	lwb_issued_timestamp; /* when was the lwb issued? */
 	uint64_t	lwb_issued_txg;	/* the txg when the write is issued */
+	uint64_t	lwb_alloc_txg;	/* the txg when lwb_blk is allocated */
 	uint64_t	lwb_max_txg;	/* highest txg in this lwb */
 	list_node_t	lwb_node;	/* zilog->zl_lwb_list linkage */
 	list_node_t	lwb_issue_node;	/* linkage of lwbs ready for issue */

include/sys/zio.h

@@ -222,7 +222,6 @@ typedef uint64_t zio_flag_t;
 #define	ZIO_FLAG_NOPWRITE	(1ULL << 28)
 #define	ZIO_FLAG_REEXECUTED	(1ULL << 29)
 #define	ZIO_FLAG_DELEGATED	(1ULL << 30)
-#define	ZIO_FLAG_FASTWRITE	(1ULL << 31)
 
 #define	ZIO_FLAG_MUSTSUCCEED		0
 #define	ZIO_FLAG_RAW	(ZIO_FLAG_RAW_COMPRESS | ZIO_FLAG_RAW_ENCRYPT)

include/sys/zio_checksum.h

@@ -94,8 +94,6 @@ typedef const struct zio_checksum_info {
 } zio_checksum_info_t;
 
 typedef struct zio_bad_cksum {
-	zio_cksum_t		zbc_expected;
-	zio_cksum_t		zbc_actual;
 	const char		*zbc_checksum_name;
 	uint8_t			zbc_byteswapped;
 	uint8_t			zbc_injected;

lib/libzfs/Makefile.am

@@ -57,7 +57,7 @@ libzfs_la_LIBADD = \
 	libzutil.la \
 	libuutil.la
 
-libzfs_la_LIBADD += -lm $(LIBCRYPTO_LIBS) $(ZLIB_LIBS) $(LIBFETCH_LIBS) $(LTLIBINTL)
+libzfs_la_LIBADD += -lrt -lm $(LIBCRYPTO_LIBS) $(ZLIB_LIBS) $(LIBFETCH_LIBS) $(LTLIBINTL)
 
 libzfs_la_LDFLAGS = -pthread
 

lib/libzfs/libzfs_mount.c

@@ -1300,7 +1300,7 @@ zpool_enable_datasets(zpool_handle_t *zhp, const char *mntopts, int flags)
 	zfs_foreach_mountpoint(zhp->zpool_hdl, cb.cb_handles, cb.cb_used,
 	    zfs_mount_one, &ms, B_TRUE);
 	if (ms.ms_mntstatus != 0)
-		ret = ms.ms_mntstatus;
+		ret = EZFS_MOUNTFAILED;
 
 	/*
 	 * Share all filesystems that need to be shared. This needs to be
@@ -1311,7 +1311,7 @@ zpool_enable_datasets(zpool_handle_t *zhp, const char *mntopts, int flags)
 	zfs_foreach_mountpoint(zhp->zpool_hdl, cb.cb_handles, cb.cb_used,
 	    zfs_share_one, &ms, B_FALSE);
 	if (ms.ms_mntstatus != 0)
-		ret = ms.ms_mntstatus;
+		ret = EZFS_SHAREFAILED;
 	else
 		zfs_commit_shares(NULL);
 

lib/libzfs/libzfs_pool.c

@@ -3926,6 +3926,12 @@ zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
 
 	switch (errno) {
 
+	case EALREADY:
+		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+		    "removal for this vdev is already in progress."));
+		(void) zfs_error(hdl, EZFS_BUSY, errbuf);
+		break;
+
 	case EINVAL:
 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 		    "invalid config; all top-level vdevs must "

lib/libzfs/libzfs_sendrecv.c

@@ -928,6 +928,39 @@ zfs_send_progress(zfs_handle_t *zhp, int fd, uint64_t *bytes_written,
 	return (0);
 }
 
+static volatile boolean_t send_progress_thread_signal_duetotimer;
+static void
+send_progress_thread_act(int sig, siginfo_t *info, void *ucontext)
+{
+	(void) sig, (void) ucontext;
+	send_progress_thread_signal_duetotimer = info->si_code == SI_TIMER;
+}
+
+struct timer_desirability {
+	timer_t timer;
+	boolean_t desired;
+};
+static void
+timer_delete_cleanup(void *timer)
+{
+	struct timer_desirability *td = timer;
+	if (td->desired)
+		timer_delete(td->timer);
+}
+
+#ifdef SIGINFO
+#define	SEND_PROGRESS_THREAD_PARENT_BLOCK_SIGINFO sigaddset(&new, SIGINFO)
+#else
+#define	SEND_PROGRESS_THREAD_PARENT_BLOCK_SIGINFO
+#endif
+#define	SEND_PROGRESS_THREAD_PARENT_BLOCK(old) { \
+	sigset_t new; \
+	sigemptyset(&new); \
+	sigaddset(&new, SIGUSR1); \
+	SEND_PROGRESS_THREAD_PARENT_BLOCK_SIGINFO; \
+	pthread_sigmask(SIG_BLOCK, &new, old); \
+}
+
 static void *
 send_progress_thread(void *arg)
 {
@@ -941,6 +974,26 @@ send_progress_thread(void *arg)
 	struct tm tm;
 	int err;
 
+	const struct sigaction signal_action =
+	    {.sa_sigaction = send_progress_thread_act, .sa_flags = SA_SIGINFO};
+	struct sigevent timer_cfg =
+	    {.sigev_notify = SIGEV_SIGNAL, .sigev_signo = SIGUSR1};
+	const struct itimerspec timer_time =
+	    {.it_value = {.tv_sec = 1}, .it_interval = {.tv_sec = 1}};
+	struct timer_desirability timer = {};
+
+	sigaction(SIGUSR1, &signal_action, NULL);
+#ifdef SIGINFO
+	sigaction(SIGINFO, &signal_action, NULL);
+#endif
+
+	if ((timer.desired = pa->pa_progress || pa->pa_astitle)) {
+		if (timer_create(CLOCK_MONOTONIC, &timer_cfg, &timer.timer))
+			return ((void *)(uintptr_t)errno);
+		(void) timer_settime(timer.timer, 0, &timer_time, NULL);
+	}
+	pthread_cleanup_push(timer_delete_cleanup, &timer);
+
 	if (!pa->pa_parsable && pa->pa_progress) {
 		(void) fprintf(stderr,
 		    "TIME       %s   %sSNAPSHOT %s\n",
@@ -953,12 +1006,12 @@ send_progress_thread(void *arg)
 	 * Print the progress from ZFS_IOC_SEND_PROGRESS every second.
 	 */
 	for (;;) {
-		(void) sleep(1);
+		pause();
 		if ((err = zfs_send_progress(zhp, pa->pa_fd, &bytes,
 		    &blocks)) != 0) {
 			if (err == EINTR || err == ENOENT)
-				return ((void *)0);
-			return ((void *)(uintptr_t)err);
+				err = 0;
+			pthread_exit(((void *)(uintptr_t)err));
 		}
 
 		(void) time(&t);
@@ -991,21 +1044,25 @@ send_progress_thread(void *arg)
 			(void) fprintf(stderr, "%02d:%02d:%02d\t%llu\t%s\n",
 			    tm.tm_hour, tm.tm_min, tm.tm_sec,
 			    (u_longlong_t)bytes, zhp->zfs_name);
-		} else if (pa->pa_progress) {
+		} else if (pa->pa_progress ||
+		    !send_progress_thread_signal_duetotimer) {
 			zfs_nicebytes(bytes, buf, sizeof (buf));
 			(void) fprintf(stderr, "%02d:%02d:%02d   %5s   %s\n",
 			    tm.tm_hour, tm.tm_min, tm.tm_sec,
 			    buf, zhp->zfs_name);
 		}
 	}
+	pthread_cleanup_pop(B_TRUE);
 }
 
 static boolean_t
-send_progress_thread_exit(libzfs_handle_t *hdl, pthread_t ptid)
+send_progress_thread_exit(
+    libzfs_handle_t *hdl, pthread_t ptid, sigset_t *oldmask)
 {
 	void *status = NULL;
 	(void) pthread_cancel(ptid);
 	(void) pthread_join(ptid, &status);
+	pthread_sigmask(SIG_SETMASK, oldmask, NULL);
 	int error = (int)(uintptr_t)status;
 	if (error != 0 && status != PTHREAD_CANCELED)
 		return (zfs_standard_error(hdl, error,
@@ -1199,7 +1256,8 @@ dump_snapshot(zfs_handle_t *zhp, void *arg)
 		 * If progress reporting is requested, spawn a new thread to
 		 * poll ZFS_IOC_SEND_PROGRESS at a regular interval.
 		 */
-		if (sdd->progress || sdd->progressastitle) {
+		sigset_t oldmask;
+		{
 			pa.pa_zhp = zhp;
 			pa.pa_fd = sdd->outfd;
 			pa.pa_parsable = sdd->parsable;
@@ -1214,13 +1272,13 @@ dump_snapshot(zfs_handle_t *zhp, void *arg)
 				zfs_close(zhp);
 				return (err);
 			}
+			SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
 		}
 
 		err = dump_ioctl(zhp, sdd->prevsnap, sdd->prevsnap_obj,
 		    fromorigin, sdd->outfd, flags, sdd->debugnv);
 
-		if ((sdd->progress || sdd->progressastitle) &&
-		    send_progress_thread_exit(zhp->zfs_hdl, tid))
+		if (send_progress_thread_exit(zhp->zfs_hdl, tid, &oldmask))
 			return (-1);
 	}
 
@@ -1562,8 +1620,9 @@ estimate_size(zfs_handle_t *zhp, const char *from, int fd, sendflags_t *flags,
 	progress_arg_t pa = { 0 };
 	int err = 0;
 	pthread_t ptid;
+	sigset_t oldmask;
 
-	if (flags->progress || flags->progressastitle) {
+	{
 		pa.pa_zhp = zhp;
 		pa.pa_fd = fd;
 		pa.pa_parsable = flags->parsable;
@@ -1577,6 +1636,7 @@ estimate_size(zfs_handle_t *zhp, const char *from, int fd, sendflags_t *flags,
 			return (zfs_error(zhp->zfs_hdl,
 			    EZFS_THREADCREATEFAILED, errbuf));
 		}
+		SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
 	}
 
 	err = lzc_send_space_resume_redacted(zhp->zfs_name, from,
@@ -1584,8 +1644,7 @@ estimate_size(zfs_handle_t *zhp, const char *from, int fd, sendflags_t *flags,
 	    redactbook, fd, &size);
 	*sizep = size;
 
-	if ((flags->progress || flags->progressastitle) &&
-	    send_progress_thread_exit(zhp->zfs_hdl, ptid))
+	if (send_progress_thread_exit(zhp->zfs_hdl, ptid, &oldmask))
 		return (-1);
 
 	if (!flags->progress && !flags->parsable)
@@ -1876,11 +1935,12 @@ zfs_send_resume_impl_cb_impl(libzfs_handle_t *hdl, sendflags_t *flags,
 	if (!flags->dryrun) {
 		progress_arg_t pa = { 0 };
 		pthread_t tid;
+		sigset_t oldmask;
 		/*
 		 * If progress reporting is requested, spawn a new thread to
 		 * poll ZFS_IOC_SEND_PROGRESS at a regular interval.
 		 */
-		if (flags->progress || flags->progressastitle) {
+		{
 			pa.pa_zhp = zhp;
 			pa.pa_fd = outfd;
 			pa.pa_parsable = flags->parsable;
@@ -1898,6 +1958,7 @@ zfs_send_resume_impl_cb_impl(libzfs_handle_t *hdl, sendflags_t *flags,
 				zfs_close(zhp);
 				return (error);
 			}
+			SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
 		}
 
 		error = lzc_send_resume_redacted(zhp->zfs_name, fromname, outfd,
@@ -1905,8 +1966,7 @@ zfs_send_resume_impl_cb_impl(libzfs_handle_t *hdl, sendflags_t *flags,
 		if (redact_book != NULL)
 			free(redact_book);
 
-		if ((flags->progressastitle || flags->progress) &&
-		    send_progress_thread_exit(hdl, tid)) {
+		if (send_progress_thread_exit(hdl, tid, &oldmask)) {
 			zfs_close(zhp);
 			return (-1);
 		}
@@ -2691,7 +2751,8 @@ zfs_send_one_cb_impl(zfs_handle_t *zhp, const char *from, int fd,
 	 * If progress reporting is requested, spawn a new thread to poll
 	 * ZFS_IOC_SEND_PROGRESS at a regular interval.
 	 */
-	if (flags->progress || flags->progressastitle) {
+	sigset_t oldmask;
+	{
 		pa.pa_zhp = zhp;
 		pa.pa_fd = fd;
 		pa.pa_parsable = flags->parsable;
@@ -2708,13 +2769,13 @@ zfs_send_one_cb_impl(zfs_handle_t *zhp, const char *from, int fd,
 			return (zfs_error(zhp->zfs_hdl,
 			    EZFS_THREADCREATEFAILED, errbuf));
 		}
+		SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
 	}
 
 	err = lzc_send_redacted(name, from, fd,
 	    lzc_flags_from_sendflags(flags), redactbook);
 
-	if ((flags->progress || flags->progressastitle) &&
-	    send_progress_thread_exit(hdl, ptid))
+	if (send_progress_thread_exit(hdl, ptid, &oldmask))
 			return (-1);
 
 	if (err == 0 && (flags->props || flags->holds || flags->backup)) {

man/Makefile.am

@@ -38,7 +38,6 @@ dist_man_MANS = \
 	%D%/man8/zfs-groupspace.8 \
 	%D%/man8/zfs-hold.8 \
 	%D%/man8/zfs-inherit.8 \
-	%D%/man8/zfs-jail.8 \
 	%D%/man8/zfs-list.8 \
 	%D%/man8/zfs-load-key.8 \
 	%D%/man8/zfs-mount.8 \
@@ -57,14 +56,11 @@ dist_man_MANS = \
 	%D%/man8/zfs-share.8 \
 	%D%/man8/zfs-snapshot.8 \
 	%D%/man8/zfs-unallow.8 \
-	%D%/man8/zfs-unjail.8 \
 	%D%/man8/zfs-unload-key.8 \
 	%D%/man8/zfs-unmount.8 \
-	%D%/man8/zfs-unzone.8 \
 	%D%/man8/zfs-upgrade.8 \
 	%D%/man8/zfs-userspace.8 \
 	%D%/man8/zfs-wait.8 \
-	%D%/man8/zfs-zone.8 \
 	%D%/man8/zfs_ids_to_path.8 \
 	%D%/man8/zgenhostid.8 \
 	%D%/man8/zinject.8 \
@@ -104,6 +100,18 @@ dist_man_MANS = \
 	%D%/man8/zstreamdump.8 \
 	%D%/man8/zpool_influxdb.8
 
+if BUILD_FREEBSD
+dist_man_MANS += \
+	%D%/man8/zfs-jail.8 \
+	%D%/man8/zfs-unjail.8
+endif
+
+if BUILD_LINUX
+dist_man_MANS += \
+	%D%/man8/zfs-unzone.8 \
+	%D%/man8/zfs-zone.8
+endif
+
 nodist_man_MANS = \
 	%D%/man8/zed.8 \
 	%D%/man8/zfs-mount-generator.8

man/man4/zfs.4

@@ -15,7 +15,7 @@
 .\" own identifying information:
 .\" Portions Copyright [yyyy] [name of copyright owner]
 .\"
-.Dd January 10, 2023
+.Dd July 21, 2023
 .Dt ZFS 4
 .Os
 .
@@ -239,6 +239,11 @@ relative to the pool.
 Make some blocks above a certain size be gang blocks.
 This option is used by the test suite to facilitate testing.
 .
+.It Sy metaslab_force_ganging_pct Ns = Ns Sy 3 Ns % Pq uint
+For blocks that could be forced to be a gang block (due to
+.Sy metaslab_force_ganging ) ,
+force this many of them to be gang blocks.
+.
 .It Sy zfs_ddt_zap_default_bs Ns = Ns Sy 15 Po 32 KiB Pc Pq int
 Default DDT ZAP data block size as a power of 2. Note that changing this after
 creating a DDT on the pool will not affect existing DDTs, only newly created
@@ -519,9 +524,6 @@ However, this is limited by
 Maximum micro ZAP size.
 A micro ZAP is upgraded to a fat ZAP, once it grows beyond the specified size.
 .
-.It Sy zfetch_array_rd_sz Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq u64
-If prefetching is enabled, disable prefetching for reads larger than this size.
-.
 .It Sy zfetch_min_distance Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq uint
 Min bytes to prefetch per stream.
 Prefetch distance starts from the demand access size and quickly grows to

man/man7/zfsprops.7

@@ -38,7 +38,7 @@
 .\" Copyright (c) 2019, Kjeld Schouten-Lebbing
 .\" Copyright (c) 2022 Hewlett Packard Enterprise Development LP.
 .\"
-.Dd April 18, 2023
+.Dd August 8, 2023
 .Dt ZFSPROPS 7
 .Os
 .
@@ -1916,13 +1916,15 @@ See
 for more information.
 Jails are a
 .Fx
-feature and are not relevant on other platforms.
-The default value is
-.Sy off .
-.It Sy zoned Ns = Ns Sy on Ns | Ns Sy off
+feature and this property is not available on other platforms.
+.It Sy zoned Ns = Ns Sy off Ns | Ns Sy on
 Controls whether the dataset is managed from a non-global zone or namespace.
-The default value is
-.Sy off .
+See
+.Xr zfs-zone 8
+for more information.
+Zoning is a
+Linux
+feature and this property is not available on other platforms.
 .El
 .Pp
 The following three properties cannot be changed after the file system is

man/man8/zfs-send.8

@@ -29,7 +29,7 @@
 .\" Copyright 2018 Nexenta Systems, Inc.
 .\" Copyright 2019 Joyent, Inc.
 .\"
-.Dd January 12, 2023
+.Dd July 27, 2023
 .Dt ZFS-SEND 8
 .Os
 .
@@ -297,6 +297,12 @@ This flag can only be used in conjunction with
 .It Fl v , -verbose
 Print verbose information about the stream package generated.
 This information includes a per-second report of how much data has been sent.
+The same report can be requested by sending
+.Dv SIGINFO
+or
+.Dv SIGUSR1 ,
+regardless of
+.Fl v .
 .Pp
 The format of the stream is committed.
 You will be able to receive your streams on future versions of ZFS.
@@ -433,6 +439,12 @@ and the verbose output goes to standard error
 .It Fl v , -verbose
 Print verbose information about the stream package generated.
 This information includes a per-second report of how much data has been sent.
+The same report can be requested by sending
+.Dv SIGINFO
+or
+.Dv SIGUSR1 ,
+regardless of
+.Fl v .
 .El
 .It Xo
 .Nm zfs
@@ -669,6 +681,10 @@ ones on the source, and are ready to be used, while the parent snapshot on the
 target contains none of the username and password data present on the source,
 because it was removed by the redacted send operation.
 .
+.Sh SIGNALS
+See
+.Fl v .
+.
 .Sh EXAMPLES
 .\" These are, respectively, examples 12, 13 from zfs.8
 .\" Make sure to update them bidirectionally

man/man8/zpool-events.8

@@ -26,7 +26,7 @@
 .\" Copyright 2017 Nexenta Systems, Inc.
 .\" Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
 .\"
-.Dd May 27, 2021
+.Dd July 11, 2023
 .Dt ZPOOL-EVENTS 8
 .Os
 .
@@ -305,10 +305,6 @@ The time when a given I/O request was submitted.
 The time required to service a given I/O request.
 .It Sy prev_state
 The previous state of the vdev.
-.It Sy cksum_expected
-The expected checksum value for the block.
-.It Sy cksum_actual
-The actual checksum value for an errant block.
 .It Sy cksum_algorithm
 Checksum algorithm used.
 See
@@ -362,23 +358,6 @@ Like
 but contains
 .Pq Ar good data No & ~( Ns Ar bad data ) ;
 that is, the bits set in the good data which are cleared in the bad data.
-.It Sy bad_set_histogram
-If this field exists, it is an array of counters.
-Each entry counts bits set in a particular bit of a big-endian uint64 type.
-The first entry counts bits
-set in the high-order bit of the first byte, the 9th byte, etc, and the last
-entry counts bits set of the low-order bit of the 8th byte, the 16th byte, etc.
-This information is useful for observing a stuck bit in a parallel data path,
-such as IDE or parallel SCSI.
-.It Sy bad_cleared_histogram
-If this field exists, it is an array of counters.
-Each entry counts bit clears in a particular bit of a big-endian uint64 type.
-The first entry counts bits
-clears of the high-order bit of the first byte, the 9th byte, etc, and the
-last entry counts clears of the low-order bit of the 8th byte, the 16th byte,
-etc.
-This information is useful for observing a stuck bit in a parallel data
-path, such as IDE or parallel SCSI.
 .El
 .
 .Sh I/O STAGES

man/man8/zpool.8

@@ -110,9 +110,10 @@ Removes ZFS label information from the specified
 .It Xo
 .Xr zpool-attach 8 Ns / Ns Xr zpool-detach 8
 .Xc
-Increases or decreases redundancy by
-.Cm attach Ns ing or
-.Cm detach Ns ing a device on an existing vdev (virtual device).
+Converts a non-redundant disk into a mirror, or increases
+the redundancy level of an existing mirror
+.Cm ( attach Ns ), or performs the inverse operation (
+.Cm detach Ns ).
 .It Xo
 .Xr zpool-add 8 Ns / Ns Xr zpool-remove 8
 .Xc
@@ -233,16 +234,16 @@ Invalid command line options were specified.
 .El
 .
 .Sh EXAMPLES
-.\" Examples 1, 2, 3, 4, 11, 12 are shared with zpool-create.8.
-.\" Examples 5, 13 are shared with zpool-add.8.
-.\" Examples 6, 15 are shared with zpool-list.8.
-.\" Examples 7 are shared with zpool-destroy.8.
-.\" Examples 8 are shared with zpool-export.8.
-.\" Examples 9 are shared with zpool-import.8.
-.\" Examples 10 are shared with zpool-upgrade.8.
-.\" Examples 14 are shared with zpool-remove.8.
-.\" Examples 16 are shared with zpool-status.8.
-.\" Examples 13, 16 are also shared with zpool-iostat.8.
+.\" Examples 1, 2, 3, 4, 12, 13 are shared with zpool-create.8.
+.\" Examples 6, 14 are shared with zpool-add.8.
+.\" Examples 7, 16 are shared with zpool-list.8.
+.\" Examples 8 are shared with zpool-destroy.8.
+.\" Examples 9 are shared with zpool-export.8.
+.\" Examples 10 are shared with zpool-import.8.
+.\" Examples 11 are shared with zpool-upgrade.8.
+.\" Examples 15 are shared with zpool-remove.8.
+.\" Examples 17 are shared with zpool-status.8.
+.\" Examples 14, 17 are also shared with zpool-iostat.8.
 .\" Make sure to update them omnidirectionally
 .Ss Example 1 : No Creating a RAID-Z Storage Pool
 The following command creates a pool with a single raidz root vdev that
@@ -264,14 +265,21 @@ While not recommended, a pool based on files can be useful for experimental
 purposes.
 .Dl # Nm zpool Cm create Ar tank Pa /path/to/file/a /path/to/file/b
 .
-.Ss Example 5 : No Adding a Mirror to a ZFS Storage Pool
+.Ss Example 5 : No Making a non-mirrored ZFS Storage Pool mirrored
+The following command converts an existing single device
+.Ar sda
+into a mirror by attaching a second device to it,
+.Ar sdb .
+.Dl # Nm zpool Cm attach Ar tank Pa sda sdb
+.
+.Ss Example 6 : No Adding a Mirror to a ZFS Storage Pool
 The following command adds two mirrored disks to the pool
 .Ar tank ,
 assuming the pool is already made up of two-way mirrors.
 The additional space is immediately available to any datasets within the pool.
 .Dl # Nm zpool Cm add Ar tank Sy mirror Pa sda sdb
 .
-.Ss Example 6 : No Listing Available ZFS Storage Pools
+.Ss Example 7 : No Listing Available ZFS Storage Pools
 The following command lists all available pools on the system.
 In this case, the pool
 .Ar zion
@@ -285,19 +293,19 @@ tank   61.5G  20.0G  41.5G         -    48%    32%  1.00x  ONLINE  -
 zion       -      -      -         -      -      -      -  FAULTED -
 .Ed
 .
-.Ss Example 7 : No Destroying a ZFS Storage Pool
+.Ss Example 8 : No Destroying a ZFS Storage Pool
 The following command destroys the pool
 .Ar tank
 and any datasets contained within:
 .Dl # Nm zpool Cm destroy Fl f Ar tank
 .
-.Ss Example 8 : No Exporting a ZFS Storage Pool
+.Ss Example 9 : No Exporting a ZFS Storage Pool
 The following command exports the devices in pool
 .Ar tank
 so that they can be relocated or later imported:
 .Dl # Nm zpool Cm export Ar tank
 .
-.Ss Example 9 : No Importing a ZFS Storage Pool
+.Ss Example 10 : No Importing a ZFS Storage Pool
 The following command displays available pools, and then imports the pool
 .Ar tank
 for use on the system.
@@ -318,7 +326,7 @@ config:
 .No # Nm zpool Cm import Ar tank
 .Ed
 .
-.Ss Example 10 : No Upgrading All ZFS Storage Pools to the Current Version
+.Ss Example 11 : No Upgrading All ZFS Storage Pools to the Current Version
 The following command upgrades all ZFS Storage pools to the current version of
 the software:
 .Bd -literal -compact -offset Ds
@@ -326,7 +334,7 @@ the software:
 This system is currently running ZFS version 2.
 .Ed
 .
-.Ss Example 11 : No Managing Hot Spares
+.Ss Example 12 : No Managing Hot Spares
 The following command creates a new pool with an available hot spare:
 .Dl # Nm zpool Cm create Ar tank Sy mirror Pa sda sdb Sy spare Pa sdc
 .Pp
@@ -341,12 +349,12 @@ The hot spare can be permanently removed from the pool using the following
 command:
 .Dl # Nm zpool Cm remove Ar tank Pa sdc
 .
-.Ss Example 12 : No Creating a ZFS Pool with Mirrored Separate Intent Logs
+.Ss Example 13 : No Creating a ZFS Pool with Mirrored Separate Intent Logs
 The following command creates a ZFS storage pool consisting of two, two-way
 mirrors and mirrored log devices:
 .Dl # Nm zpool Cm create Ar pool Sy mirror Pa sda sdb Sy mirror Pa sdc sdd Sy log mirror Pa sde sdf
 .
-.Ss Example 13 : No Adding Cache Devices to a ZFS Pool
+.Ss Example 14 : No Adding Cache Devices to a ZFS Pool
 The following command adds two disks for use as cache devices to a ZFS storage
 pool:
 .Dl # Nm zpool Cm add Ar pool Sy cache Pa sdc sdd
@@ -359,7 +367,7 @@ Capacity and reads can be monitored using the
 subcommand as follows:
 .Dl # Nm zpool Cm iostat Fl v Ar pool 5
 .
-.Ss Example 14 : No Removing a Mirrored top-level (Log or Data) Device
+.Ss Example 15 : No Removing a Mirrored top-level (Log or Data) Device
 The following commands remove the mirrored log device
 .Sy mirror-2
 and mirrored top-level data device
@@ -394,7 +402,7 @@ The command to remove the mirrored data
 .Ar mirror-1 No is :
 .Dl # Nm zpool Cm remove Ar tank mirror-1
 .
-.Ss Example 15 : No Displaying expanded space on a device
+.Ss Example 16 : No Displaying expanded space on a device
 The following command displays the detailed information for the pool
 .Ar data .
 This pool is comprised of a single raidz vdev where one of its devices
@@ -411,7 +419,7 @@ data        23.9G  14.6G  9.30G         -    48%    61%  1.00x  ONLINE  -
     sdc         -      -      -         -      -
 .Ed
 .
-.Ss Example 16 : No Adding output columns
+.Ss Example 17 : No Adding output columns
 Additional columns can be added to the
 .Nm zpool Cm status No and Nm zpool Cm iostat No output with Fl c .
 .Bd -literal -compact -offset Ds

module/Kbuild.in

@@ -461,6 +461,7 @@ ZFS_OBJS_OS := \
 	zpl_ctldir.o \
 	zpl_export.o \
 	zpl_file.o \
+	zpl_file_range.o \
 	zpl_inode.o \
 	zpl_super.o \
 	zpl_xattr.o \

module/os/freebsd/zfs/zfs_vnops_os.c

@@ -6263,7 +6263,8 @@ zfs_freebsd_copy_file_range(struct vop_copy_file_range_args *ap)
 			goto bad_write_fallback;
 		}
 	} else {
-#if __FreeBSD_version >= 1400086
+#if (__FreeBSD_version >= 1302506 && __FreeBSD_version < 1400000) || \
+	__FreeBSD_version >= 1400086
 		vn_lock_pair(invp, false, LK_EXCLUSIVE, outvp, false,
 		    LK_EXCLUSIVE);
 #else
@@ -6289,7 +6290,8 @@ zfs_freebsd_copy_file_range(struct vop_copy_file_range_args *ap)
 
 	error = zfs_clone_range(VTOZ(invp), ap->a_inoffp, VTOZ(outvp),
 	    ap->a_outoffp, &len, ap->a_outcred);
-	if (error == EXDEV)
+	if (error == EXDEV || error == EAGAIN || error == EINVAL ||
+	    error == EOPNOTSUPP)
 		goto bad_locked_fallback;
 	*ap->a_lenp = (size_t)len;
 out_locked:

module/os/linux/zfs/qat_compress.c

@@ -193,7 +193,9 @@ qat_dc_init(void)
 		sd.huffType = CPA_DC_HT_FULL_DYNAMIC;
 		sd.sessDirection = CPA_DC_DIR_COMBINED;
 		sd.sessState = CPA_DC_STATELESS;
+#if (CPA_DC_API_VERSION_NUM_MAJOR == 1 && CPA_DC_API_VERSION_NUM_MINOR < 6)
 		sd.deflateWindowSize = 7;
+#endif
 		sd.checksum = CPA_DC_ADLER32;
 		status = cpaDcGetSessionSize(dc_inst_handles[i],
 		    &sd, &sess_size, &ctx_size);

module/os/linux/zfs/zfs_ctldir.c

@@ -478,17 +478,19 @@ zfsctl_is_snapdir(struct inode *ip)
  */
 static struct inode *
 zfsctl_inode_alloc(zfsvfs_t *zfsvfs, uint64_t id,
-    const struct file_operations *fops, const struct inode_operations *ops)
+    const struct file_operations *fops, const struct inode_operations *ops,
+    uint64_t creation)
 {
-	inode_timespec_t now;
 	struct inode *ip;
 	znode_t *zp;
+	inode_timespec_t now = {.tv_sec = creation};
 
 	ip = new_inode(zfsvfs->z_sb);
 	if (ip == NULL)
 		return (NULL);
 
-	now = current_time(ip);
+	if (!creation)
+		now = current_time(ip);
 	zp = ITOZ(ip);
 	ASSERT3P(zp->z_dirlocks, ==, NULL);
 	ASSERT3P(zp->z_acl_cached, ==, NULL);
@@ -552,14 +554,28 @@ zfsctl_inode_lookup(zfsvfs_t *zfsvfs, uint64_t id,
     const struct file_operations *fops, const struct inode_operations *ops)
 {
 	struct inode *ip = NULL;
+	uint64_t creation = 0;
+	dsl_dataset_t *snap_ds;
+	dsl_pool_t *pool;
 
 	while (ip == NULL) {
 		ip = ilookup(zfsvfs->z_sb, (unsigned long)id);
 		if (ip)
 			break;
 
+		if (id <= ZFSCTL_INO_SNAPDIRS && !creation) {
+			pool = dmu_objset_pool(zfsvfs->z_os);
+			dsl_pool_config_enter(pool, FTAG);
+			if (!dsl_dataset_hold_obj(pool,
+			    ZFSCTL_INO_SNAPDIRS - id, FTAG, &snap_ds)) {
+				creation = dsl_get_creation(snap_ds);
+				dsl_dataset_rele(snap_ds, FTAG);
+			}
+			dsl_pool_config_exit(pool, FTAG);
+		}
+
 		/* May fail due to concurrent zfsctl_inode_alloc() */
-		ip = zfsctl_inode_alloc(zfsvfs, id, fops, ops);
+		ip = zfsctl_inode_alloc(zfsvfs, id, fops, ops, creation);
 	}
 
 	return (ip);
@@ -581,7 +597,7 @@ zfsctl_create(zfsvfs_t *zfsvfs)
 	ASSERT(zfsvfs->z_ctldir == NULL);
 
 	zfsvfs->z_ctldir = zfsctl_inode_alloc(zfsvfs, ZFSCTL_INO_ROOT,
-	    &zpl_fops_root, &zpl_ops_root);
+	    &zpl_fops_root, &zpl_ops_root, 0);
 	if (zfsvfs->z_ctldir == NULL)
 		return (SET_ERROR(ENOENT));
 

module/os/linux/zfs/zfs_vfsops.c

@@ -1662,6 +1662,7 @@ zfs_umount(struct super_block *sb)
 	}
 
 	zfsvfs_free(zfsvfs);
+	sb->s_fs_info = NULL;
 	return (0);
 }
 
@@ -2091,6 +2092,9 @@ zfs_init(void)
 	zfs_znode_init();
 	dmu_objset_register_type(DMU_OST_ZFS, zpl_get_file_info);
 	register_filesystem(&zpl_fs_type);
+#ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND
+	register_fo_extend(&zpl_file_operations);
+#endif
 }
 
 void
@@ -2101,6 +2105,9 @@ zfs_fini(void)
 	 */
 	taskq_wait(system_delay_taskq);
 	taskq_wait(system_taskq);
+#ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND
+	unregister_fo_extend(&zpl_file_operations);
+#endif
 	unregister_filesystem(&zpl_fs_type);
 	zfs_znode_fini();
 	zfsctl_fini();

module/os/linux/zfs/zfs_znode.c

@@ -415,7 +415,11 @@ zfs_inode_set_ops(zfsvfs_t *zfsvfs, struct inode *ip)
 	switch (ip->i_mode & S_IFMT) {
 	case S_IFREG:
 		ip->i_op = &zpl_inode_operations;
+#ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND
+		ip->i_fop = &zpl_file_operations.kabi_fops;
+#else
 		ip->i_fop = &zpl_file_operations;
+#endif
 		ip->i_mapping->a_ops = &zpl_address_space_operations;
 		break;
 
@@ -455,7 +459,11 @@ zfs_inode_set_ops(zfsvfs_t *zfsvfs, struct inode *ip)
 		/* Assume the inode is a file and attempt to continue */
 		ip->i_mode = S_IFREG | 0644;
 		ip->i_op = &zpl_inode_operations;
+#ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND
+		ip->i_fop = &zpl_file_operations.kabi_fops;
+#else
 		ip->i_fop = &zpl_file_operations;
+#endif
 		ip->i_mapping->a_ops = &zpl_address_space_operations;
 		break;
 	}

module/os/linux/zfs/zpl_file.c

@@ -1257,6 +1257,12 @@ zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 		return (zpl_ioctl_getdosflags(filp, (void *)arg));
 	case ZFS_IOC_SETDOSFLAGS:
 		return (zpl_ioctl_setdosflags(filp, (void *)arg));
+	case ZFS_IOC_COMPAT_FICLONE:
+		return (zpl_ioctl_ficlone(filp, (void *)arg));
+	case ZFS_IOC_COMPAT_FICLONERANGE:
+		return (zpl_ioctl_ficlonerange(filp, (void *)arg));
+	case ZFS_IOC_COMPAT_FIDEDUPERANGE:
+		return (zpl_ioctl_fideduperange(filp, (void *)arg));
 	default:
 		return (-ENOTTY);
 	}
@@ -1283,7 +1289,6 @@ zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 }
 #endif /* CONFIG_COMPAT */
 
-
 const struct address_space_operations zpl_address_space_operations = {
 #ifdef HAVE_VFS_READPAGES
 	.readpages	= zpl_readpages,
@@ -1306,7 +1311,12 @@ const struct address_space_operations zpl_address_space_operations = {
 #endif
 };
 
+#ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND
+const struct file_operations_extend zpl_file_operations = {
+	.kabi_fops = {
+#else
 const struct file_operations zpl_file_operations = {
+#endif
 	.open		= zpl_open,
 	.release	= zpl_release,
 	.llseek		= zpl_llseek,
@@ -1333,6 +1343,18 @@ const struct file_operations zpl_file_operations = {
 	.aio_fsync	= zpl_aio_fsync,
 #endif
 	.fallocate	= zpl_fallocate,
+#ifdef HAVE_VFS_COPY_FILE_RANGE
+	.copy_file_range	= zpl_copy_file_range,
+#endif
+#ifdef HAVE_VFS_CLONE_FILE_RANGE
+	.clone_file_range	= zpl_clone_file_range,
+#endif
+#ifdef HAVE_VFS_REMAP_FILE_RANGE
+	.remap_file_range	= zpl_remap_file_range,
+#endif
+#ifdef HAVE_VFS_DEDUPE_FILE_RANGE
+	.dedupe_file_range	= zpl_dedupe_file_range,
+#endif
 #ifdef HAVE_FILE_FADVISE
 	.fadvise	= zpl_fadvise,
 #endif
@@ -1340,6 +1362,11 @@ const struct file_operations zpl_file_operations = {
 #ifdef CONFIG_COMPAT
 	.compat_ioctl	= zpl_compat_ioctl,
 #endif
+#ifdef HAVE_VFS_FILE_OPERATIONS_EXTEND
+	}, /* kabi_fops */
+	.copy_file_range	= zpl_copy_file_range,
+	.clone_file_range	= zpl_clone_file_range,
+#endif
 };
 
 const struct file_operations zpl_dir_file_operations = {

module/os/linux/zfs/zpl_file_range.c

@@ -0,0 +1,272 @@
+/*
+ * 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, Klara Inc.
+ */
+
+#ifdef CONFIG_COMPAT
+#include <linux/compat.h>
+#endif
+#include <linux/fs.h>
+#include <sys/file.h>
+#include <sys/zfs_znode.h>
+#include <sys/zfs_vnops.h>
+#include <sys/zfeature.h>
+
+/*
+ * Clone part of a file via block cloning.
+ *
+ * Note that we are not required to update file offsets; the kernel will take
+ * care of that depending on how it was called.
+ */
+static ssize_t
+__zpl_clone_file_range(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);
+	struct inode *dst_i = file_inode(dst_file);
+	uint64_t src_off_o = (uint64_t)src_off;
+	uint64_t dst_off_o = (uint64_t)dst_off;
+	uint64_t len_o = (uint64_t)len;
+	cred_t *cr = CRED();
+	fstrans_cookie_t cookie;
+	int err;
+
+	if (!spa_feature_is_enabled(
+	    dmu_objset_spa(ITOZSB(dst_i)->z_os), SPA_FEATURE_BLOCK_CLONING))
+		return (-EOPNOTSUPP);
+
+	if (src_i != dst_i)
+		spl_inode_lock_shared(src_i);
+	spl_inode_lock(dst_i);
+
+	crhold(cr);
+	cookie = spl_fstrans_mark();
+
+	err = -zfs_clone_range(ITOZ(src_i), &src_off_o, ITOZ(dst_i),
+	    &dst_off_o, &len_o, cr);
+
+	spl_fstrans_unmark(cookie);
+	crfree(cr);
+
+	spl_inode_unlock(dst_i);
+	if (src_i != dst_i)
+		spl_inode_unlock_shared(src_i);
+
+	if (err < 0)
+		return (err);
+
+	return ((ssize_t)len_o);
+}
+
+#if defined(HAVE_VFS_COPY_FILE_RANGE) || \
+    defined(HAVE_VFS_FILE_OPERATIONS_EXTEND)
+/*
+ * Entry point for copy_file_range(). Copy len bytes from src_off in src_file
+ * to dst_off in dst_file. We are permitted to do this however we like, so we
+ * try to just clone the blocks, and if we can't support it, fall back to the
+ * kernel's generic byte copy function.
+ */
+ssize_t
+zpl_copy_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, size_t len, unsigned int flags)
+{
+	ssize_t ret;
+
+	if (flags != 0)
+		return (-EINVAL);
+
+	/* Try to do it via zfs_clone_range() */
+	ret = __zpl_clone_file_range(src_file, src_off,
+	    dst_file, dst_off, len);
+
+#ifdef 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.
+	 */
+	if (ret == -EOPNOTSUPP || ret == -EINVAL || ret == -EXDEV ||
+	    ret == -EAGAIN)
+		ret = generic_copy_file_range(src_file, src_off, dst_file,
+		    dst_off, len, flags);
+#else
+	/*
+	 * Before Linux 5.3 the filesystem has to return -EOPNOTSUPP to signal
+	 * to the kernel that it should fallback to a content copy.
+	 */
+	if (ret == -EINVAL || ret == -EXDEV || ret == -EAGAIN)
+		ret = -EOPNOTSUPP;
+#endif /* HAVE_VFS_GENERIC_COPY_FILE_RANGE */
+
+	return (ret);
+}
+#endif /* HAVE_VFS_COPY_FILE_RANGE || HAVE_VFS_FILE_OPERATIONS_EXTEND */
+
+#ifdef HAVE_VFS_REMAP_FILE_RANGE
+/*
+ * Entry point for FICLONE/FICLONERANGE/FIDEDUPERANGE.
+ *
+ * FICLONE and FICLONERANGE are basically the same as copy_file_range(), except
+ * that they must clone - they cannot fall back to copying. FICLONE is exactly
+ * FICLONERANGE, for the entire file. We don't need to try to tell them apart;
+ * the kernel will sort that out for us.
+ *
+ * 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.
+ */
+loff_t
+zpl_remap_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, loff_t len, unsigned int flags)
+{
+	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;
+
+	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));
+}
+#endif /* HAVE_VFS_REMAP_FILE_RANGE */
+
+#if defined(HAVE_VFS_CLONE_FILE_RANGE) || \
+    defined(HAVE_VFS_FILE_OPERATIONS_EXTEND)
+/*
+ * Entry point for FICLONE and FICLONERANGE, before Linux 4.20.
+ */
+int
+zpl_clone_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, uint64_t len)
+{
+	/* 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));
+}
+#endif /* HAVE_VFS_CLONE_FILE_RANGE || HAVE_VFS_FILE_OPERATIONS_EXTEND */
+
+#ifdef HAVE_VFS_DEDUPE_FILE_RANGE
+/*
+ * Entry point for FIDEDUPERANGE, before Linux 4.20.
+ */
+int
+zpl_dedupe_file_range(struct file *src_file, loff_t src_off,
+    struct file *dst_file, loff_t dst_off, uint64_t len)
+{
+	/* No support for dedup yet */
+	return (-EOPNOTSUPP);
+}
+#endif /* HAVE_VFS_DEDUPE_FILE_RANGE */
+
+/* Entry point for FICLONE, before Linux 4.5. */
+long
+zpl_ioctl_ficlone(struct file *dst_file, void *arg)
+{
+	unsigned long sfd = (unsigned long)arg;
+
+	struct file *src_file = fget(sfd);
+	if (src_file == NULL)
+		return (-EBADF);
+
+	if (dst_file->f_op != src_file->f_op)
+		return (-EXDEV);
+
+	size_t len = i_size_read(file_inode(src_file));
+
+	ssize_t ret =
+	    __zpl_clone_file_range(src_file, 0, dst_file, 0, len);
+
+	fput(src_file);
+
+	if (ret < 0) {
+		if (ret == -EOPNOTSUPP)
+			return (-ENOTTY);
+		return (ret);
+	}
+
+	if (ret != len)
+		return (-EINVAL);
+
+	return (0);
+}
+
+/* Entry point for FICLONERANGE, before Linux 4.5. */
+long
+zpl_ioctl_ficlonerange(struct file *dst_file, void __user *arg)
+{
+	zfs_ioc_compat_file_clone_range_t fcr;
+
+	if (copy_from_user(&fcr, arg, sizeof (fcr)))
+		return (-EFAULT);
+
+	struct file *src_file = fget(fcr.fcr_src_fd);
+	if (src_file == NULL)
+		return (-EBADF);
+
+	if (dst_file->f_op != src_file->f_op)
+		return (-EXDEV);
+
+	size_t len = fcr.fcr_src_length;
+	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,
+	    dst_file, fcr.fcr_dest_offset, len);
+
+	fput(src_file);
+
+	if (ret < 0) {
+		if (ret == -EOPNOTSUPP)
+			return (-ENOTTY);
+		return (ret);
+	}
+
+	if (ret != len)
+		return (-EINVAL);
+
+	return (0);
+}
+
+/* Entry point for FIDEDUPERANGE, before Linux 4.5. */
+long
+zpl_ioctl_fideduperange(struct file *filp, void *arg)
+{
+	(void) arg;
+
+	/* No support for dedup yet */
+	return (-ENOTTY);
+}

module/os/linux/zfs/zpl_super.c

@@ -277,8 +277,6 @@ zpl_test_super(struct super_block *s, void *data)
 {
 	zfsvfs_t *zfsvfs = s->s_fs_info;
 	objset_t *os = data;
-	int match;
-
 	/*
 	 * If the os doesn't match the z_os in the super_block, assume it is
 	 * not a match. Matching would imply a multimount of a dataset. It is
@@ -286,19 +284,7 @@ zpl_test_super(struct super_block *s, void *data)
 	 * that changes the z_os, e.g., rollback, where the match will be
 	 * missed, but in that case the user will get an EBUSY.
 	 */
-	if (zfsvfs == NULL || os != zfsvfs->z_os)
-		return (0);
-
-	/*
-	 * If they do match, recheck with the lock held to prevent mounting the
-	 * wrong dataset since z_os can be stale when the teardown lock is held.
-	 */
-	if (zpl_enter(zfsvfs, FTAG) != 0)
-		return (0);
-	match = (os == zfsvfs->z_os);
-	zpl_exit(zfsvfs, FTAG);
-
-	return (match);
+	return (zfsvfs != NULL && os == zfsvfs->z_os);
 }
 
 static struct super_block *
@@ -324,12 +310,35 @@ zpl_mount_impl(struct file_system_type *fs_type, int flags, zfs_mnt_t *zm)
 
 	s = sget(fs_type, zpl_test_super, set_anon_super, flags, os);
 
+	/*
+	 * Recheck with the lock held to prevent mounting the wrong dataset
+	 * since z_os can be stale when the teardown lock is held.
+	 *
+	 * We can't do this in zpl_test_super in since it's under spinlock and
+	 * also s_umount lock is not held there so it would race with
+	 * zfs_umount and zfsvfs can be freed.
+	 */
+	if (!IS_ERR(s) && s->s_fs_info != NULL) {
+		zfsvfs_t *zfsvfs = s->s_fs_info;
+		if (zpl_enter(zfsvfs, FTAG) == 0) {
+			if (os != zfsvfs->z_os)
+				err = -SET_ERROR(EBUSY);
+			zpl_exit(zfsvfs, FTAG);
+		} else {
+			err = -SET_ERROR(EBUSY);
+		}
+	}
 	dsl_dataset_long_rele(dmu_objset_ds(os), FTAG);
 	dsl_dataset_rele(dmu_objset_ds(os), FTAG);
 
 	if (IS_ERR(s))
 		return (ERR_CAST(s));
 
+	if (err) {
+		deactivate_locked_super(s);
+		return (ERR_PTR(err));
+	}
+
 	if (s->s_root == NULL) {
 		err = zpl_fill_super(s, zm, flags & SB_SILENT ? 1 : 0);
 		if (err) {

module/zcommon/zpool_prop.c

@@ -160,7 +160,7 @@ zpool_prop_init(void)
 	    "wait | continue | panic", "FAILMODE", failuremode_table,
 	    sfeatures);
 	zprop_register_index(ZPOOL_PROP_AUTOTRIM, "autotrim",
-	    SPA_AUTOTRIM_DEFAULT, PROP_DEFAULT, ZFS_TYPE_POOL,
+	    SPA_AUTOTRIM_OFF, PROP_DEFAULT, ZFS_TYPE_POOL,
 	    "on | off", "AUTOTRIM", boolean_table, sfeatures);
 
 	/* hidden properties */

module/zfs/bpobj.c

@@ -284,7 +284,17 @@ bpobj_iterate_blkptrs(bpobj_info_t *bpi, bpobj_itor_t func, void *arg,
 	dmu_buf_t *dbuf = NULL;
 	bpobj_t *bpo = bpi->bpi_bpo;
 
-	for (int64_t i = bpo->bpo_phys->bpo_num_blkptrs - 1; i >= start; i--) {
+	int64_t i = bpo->bpo_phys->bpo_num_blkptrs - 1;
+	uint64_t pe = P2ALIGN_TYPED(i, bpo->bpo_epb, uint64_t) *
+	    sizeof (blkptr_t);
+	uint64_t ps = start * sizeof (blkptr_t);
+	uint64_t pb = MAX((pe > dmu_prefetch_max) ? pe - dmu_prefetch_max : 0,
+	    ps);
+	if (pe > pb) {
+		dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0, pb, pe - pb,
+		    ZIO_PRIORITY_ASYNC_READ);
+	}
+	for (; i >= start; i--) {
 		uint64_t offset = i * sizeof (blkptr_t);
 		uint64_t blkoff = P2PHASE(i, bpo->bpo_epb);
 
@@ -292,9 +302,16 @@ bpobj_iterate_blkptrs(bpobj_info_t *bpi, bpobj_itor_t func, void *arg,
 			if (dbuf)
 				dmu_buf_rele(dbuf, FTAG);
 			err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
-			    offset, FTAG, &dbuf, 0);
+			    offset, FTAG, &dbuf, DMU_READ_NO_PREFETCH);
 			if (err)
 				break;
+			pe = pb;
+			pb = MAX((dbuf->db_offset > dmu_prefetch_max) ?
+			    dbuf->db_offset - dmu_prefetch_max : 0, ps);
+			if (pe > pb) {
+				dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0,
+				    pb, pe - pb, ZIO_PRIORITY_ASYNC_READ);
+			}
 		}
 
 		ASSERT3U(offset, >=, dbuf->db_offset);
@@ -466,22 +483,30 @@ bpobj_iterate_impl(bpobj_t *initial_bpo, bpobj_itor_t func, void *arg,
 			int64_t i = bpi->bpi_unprocessed_subobjs - 1;
 			uint64_t offset = i * sizeof (uint64_t);
 
-			uint64_t obj_from_sublist;
+			uint64_t subobj;
 			err = dmu_read(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
-			    offset, sizeof (uint64_t), &obj_from_sublist,
-			    DMU_READ_PREFETCH);
+			    offset, sizeof (uint64_t), &subobj,
+			    DMU_READ_NO_PREFETCH);
 			if (err)
 				break;
-			bpobj_t *sublist = kmem_alloc(sizeof (bpobj_t),
+
+			bpobj_t *subbpo = kmem_alloc(sizeof (bpobj_t),
 			    KM_SLEEP);
-
-			err = bpobj_open(sublist, bpo->bpo_os,
-			    obj_from_sublist);
-			if (err)
+			err = bpobj_open(subbpo, bpo->bpo_os, subobj);
+			if (err) {
+				kmem_free(subbpo, sizeof (bpobj_t));
 				break;
+			}
 
-			list_insert_head(&stack, bpi_alloc(sublist, bpi, i));
-			mutex_enter(&sublist->bpo_lock);
+			if (subbpo->bpo_havesubobj &&
+			    subbpo->bpo_phys->bpo_subobjs != 0) {
+				dmu_prefetch(subbpo->bpo_os,
+				    subbpo->bpo_phys->bpo_subobjs, 0, 0, 0,
+				    ZIO_PRIORITY_ASYNC_READ);
+			}
+
+			list_insert_head(&stack, bpi_alloc(subbpo, bpi, i));
+			mutex_enter(&subbpo->bpo_lock);
 			bpi->bpi_unprocessed_subobjs--;
 		}
 	}

module/zfs/brt.c

@@ -174,7 +174,7 @@
  *	                size_t len, unsigned int flags);
  *
  * Even though offsets and length represent bytes, they have to be
- * block-aligned or we will return the EXDEV error so the upper layer can
+ * block-aligned or we will return an error so the upper layer can
  * fallback to the generic mechanism that will just copy the data.
  * Using copy_file_range(2) will call OS-independent zfs_clone_range() function.
  * This function was implemented based on zfs_write(), but instead of writing
@@ -192,9 +192,9 @@
  * Some special cases to consider and how we address them:
  * - The block we want to clone may have been created within the same
  *   transaction group that we are trying to clone. Such block has no BP
- *   allocated yet, so cannot be immediately cloned. We return EXDEV.
+ *   allocated yet, so cannot be immediately cloned. We return EAGAIN.
  * - The block we want to clone may have been modified within the same
- *   transaction group. We return EXDEV.
+ *   transaction group. We return EAGAIN.
  * - A block may be cloned multiple times during one transaction group (that's
  *   why pending list is actually a tree and not an append-only list - this
  *   way we can figure out faster if this block is cloned for the first time
@@ -680,7 +680,7 @@ brt_vdev_realloc(brt_t *brt, brt_vdev_t *brtvd)
 	size = (vdev_get_min_asize(vd) - 1) / brt->brt_rangesize + 1;
 	spa_config_exit(brt->brt_spa, SCL_VDEV, FTAG);
 
-	entcount = kmem_zalloc(sizeof (entcount[0]) * size, KM_SLEEP);
+	entcount = vmem_zalloc(sizeof (entcount[0]) * size, KM_SLEEP);
 	nblocks = BRT_RANGESIZE_TO_NBLOCKS(size);
 	bitmap = kmem_zalloc(BT_SIZEOFMAP(nblocks), KM_SLEEP);
 
@@ -709,7 +709,7 @@ brt_vdev_realloc(brt_t *brt, brt_vdev_t *brtvd)
 		    sizeof (entcount[0]) * MIN(size, brtvd->bv_size));
 		memcpy(bitmap, brtvd->bv_bitmap, MIN(BT_SIZEOFMAP(nblocks),
 		    BT_SIZEOFMAP(brtvd->bv_nblocks)));
-		kmem_free(brtvd->bv_entcount,
+		vmem_free(brtvd->bv_entcount,
 		    sizeof (entcount[0]) * brtvd->bv_size);
 		kmem_free(brtvd->bv_bitmap, BT_SIZEOFMAP(brtvd->bv_nblocks));
 	}
@@ -792,7 +792,7 @@ brt_vdev_dealloc(brt_t *brt, brt_vdev_t *brtvd)
 	ASSERT(RW_WRITE_HELD(&brt->brt_lock));
 	ASSERT(brtvd->bv_initiated);
 
-	kmem_free(brtvd->bv_entcount, sizeof (uint16_t) * brtvd->bv_size);
+	vmem_free(brtvd->bv_entcount, sizeof (uint16_t) * brtvd->bv_size);
 	brtvd->bv_entcount = NULL;
 	kmem_free(brtvd->bv_bitmap, BT_SIZEOFMAP(brtvd->bv_nblocks));
 	brtvd->bv_bitmap = NULL;
@@ -1544,6 +1544,37 @@ out:
 	return (B_FALSE);
 }
 
+uint64_t
+brt_entry_get_refcount(spa_t *spa, const blkptr_t *bp)
+{
+	brt_t *brt = spa->spa_brt;
+	brt_vdev_t *brtvd;
+	brt_entry_t bre_search, *bre;
+	uint64_t vdevid, refcnt;
+	int error;
+
+	brt_entry_fill(bp, &bre_search, &vdevid);
+
+	brt_rlock(brt);
+
+	brtvd = brt_vdev(brt, vdevid);
+	ASSERT(brtvd != NULL);
+
+	bre = avl_find(&brtvd->bv_tree, &bre_search, NULL);
+	if (bre == NULL) {
+		error = brt_entry_lookup(brt, brtvd, &bre_search);
+		ASSERT(error == 0 || error == ENOENT);
+		if (error == ENOENT)
+			refcnt = 0;
+		else
+			refcnt = bre_search.bre_refcount;
+	} else
+		refcnt = bre->bre_refcount;
+
+	brt_unlock(brt);
+	return (refcnt);
+}
+
 static void
 brt_prefetch(brt_t *brt, const blkptr_t *bp)
 {

module/zfs/dbuf.c

@@ -2701,7 +2701,7 @@ dmu_buf_will_clone(dmu_buf_t *db_fake, dmu_tx_t *tx)
 	 */
 	mutex_enter(&db->db_mtx);
 	VERIFY(!dbuf_undirty(db, tx));
-	ASSERT(list_head(&db->db_dirty_records) == NULL);
+	ASSERT3P(dbuf_find_dirty_eq(db, tx->tx_txg), ==, NULL);
 	if (db->db_buf != NULL) {
 		arc_buf_destroy(db->db_buf, db);
 		db->db_buf = NULL;
@@ -4457,6 +4457,15 @@ dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx)
 	} else if (db->db_state == DB_FILL) {
 		/* This buffer was freed and is now being re-filled */
 		ASSERT(db->db.db_data != dr->dt.dl.dr_data);
+	} else if (db->db_state == DB_READ) {
+		/*
+		 * This buffer has a clone we need to write, and an in-flight
+		 * read on the BP we're about to clone. Its safe to issue the
+		 * write here because the read has already been issued and the
+		 * contents won't change.
+		 */
+		ASSERT(dr->dt.dl.dr_brtwrite &&
+		    dr->dt.dl.dr_override_state == DR_OVERRIDDEN);
 	} else {
 		ASSERT(db->db_state == DB_CACHED || db->db_state == DB_NOFILL);
 	}

module/zfs/dmu.c

@@ -89,7 +89,11 @@ static int zfs_dmu_offset_next_sync = 1;
  * helps to limit the amount of memory that can be used by prefetching.
  * Larger objects should be prefetched a bit at a time.
  */
+#ifdef _ILP32
+uint_t dmu_prefetch_max = 8 * 1024 * 1024;
+#else
 uint_t dmu_prefetch_max = 8 * SPA_MAXBLOCKSIZE;
+#endif
 
 const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
 	{DMU_BSWAP_UINT8,  TRUE,  FALSE, FALSE, "unallocated"		},
@@ -161,7 +165,7 @@ dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = {
 	{	zfs_acl_byteswap,	"acl"		}
 };
 
-static int
+int
 dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset,
     const void *tag, dmu_buf_t **dbp)
 {
@@ -181,6 +185,7 @@ dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset,
 	*dbp = &db->db;
 	return (0);
 }
+
 int
 dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset,
     const void *tag, dmu_buf_t **dbp)
@@ -552,8 +557,7 @@ dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length,
 		zio = zio_root(dn->dn_objset->os_spa, NULL, NULL,
 		    ZIO_FLAG_CANFAIL);
 	blkid = dbuf_whichblock(dn, 0, offset);
-	if ((flags & DMU_READ_NO_PREFETCH) == 0 &&
-	    length <= zfetch_array_rd_sz) {
+	if ((flags & DMU_READ_NO_PREFETCH) == 0) {
 		/*
 		 * Prepare the zfetch before initiating the demand reads, so
 		 * that if multiple threads block on same indirect block, we
@@ -1650,10 +1654,22 @@ dmu_sync_late_arrival(zio_t *pio, objset_t *os, dmu_sync_cb_t *done, zgd_t *zgd,
 {
 	dmu_sync_arg_t *dsa;
 	dmu_tx_t *tx;
+	int error;
+
+	error = dbuf_read((dmu_buf_impl_t *)zgd->zgd_db, NULL,
+	    DB_RF_CANFAIL | DB_RF_NOPREFETCH);
+	if (error != 0)
+		return (error);
 
 	tx = dmu_tx_create(os);
 	dmu_tx_hold_space(tx, zgd->zgd_db->db_size);
-	if (dmu_tx_assign(tx, TXG_WAIT) != 0) {
+	/*
+	 * This transaction does not produce any dirty data or log blocks, so
+	 * it should not be throttled.  All other cases wait for TXG sync, by
+	 * which time the log block we are writing will be obsolete, so we can
+	 * skip waiting and just return error here instead.
+	 */
+	if (dmu_tx_assign(tx, TXG_NOWAIT | TXG_NOTHROTTLE) != 0) {
 		dmu_tx_abort(tx);
 		/* Make zl_get_data do txg_waited_synced() */
 		return (SET_ERROR(EIO));

module/zfs/dmu_recv.c

@@ -1795,17 +1795,19 @@ receive_handle_existing_object(const struct receive_writer_arg *rwa,
 	}
 
 	/*
-	 * The dmu does not currently support decreasing nlevels
-	 * or changing the number of dnode slots on an object. For
-	 * non-raw sends, this does not matter and the new object
-	 * can just use the previous one's nlevels. For raw sends,
-	 * however, the structure of the received dnode (including
-	 * nlevels and dnode slots) must match that of the send
-	 * side. Therefore, instead of using dmu_object_reclaim(),
-	 * we must free the object completely and call
-	 * dmu_object_claim_dnsize() instead.
+	 * The dmu does not currently support decreasing nlevels or changing
+	 * indirect block size if there is already one, same as changing the
+	 * number of of dnode slots on an object.  For non-raw sends this
+	 * does not matter and the new object can just use the previous one's
+	 * parameters.  For raw sends, however, the structure of the received
+	 * dnode (including indirects and dnode slots) must match that of the
+	 * send side.  Therefore, instead of using dmu_object_reclaim(), we
+	 * must free the object completely and call dmu_object_claim_dnsize()
+	 * instead.
 	 */
-	if ((rwa->raw && drro->drr_nlevels < doi->doi_indirection) ||
+	if ((rwa->raw && ((doi->doi_indirection > 1 &&
+	    indblksz != doi->doi_metadata_block_size) ||
+	    drro->drr_nlevels < doi->doi_indirection)) ||
 	    dn_slots != doi->doi_dnodesize >> DNODE_SHIFT) {
 		err = dmu_free_long_object(rwa->os, drro->drr_object);
 		if (err != 0)

module/zfs/dmu_zfetch.c

@@ -52,14 +52,19 @@ static unsigned int	zfetch_max_streams = 8;
 static unsigned int	zfetch_min_sec_reap = 1;
 /* max time before stream delete */
 static unsigned int	zfetch_max_sec_reap = 2;
+#ifdef _ILP32
+/* min bytes to prefetch per stream (default 2MB) */
+static unsigned int	zfetch_min_distance = 2 * 1024 * 1024;
+/* max bytes to prefetch per stream (default 8MB) */
+unsigned int	zfetch_max_distance = 8 * 1024 * 1024;
+#else
 /* min bytes to prefetch per stream (default 4MB) */
 static unsigned int	zfetch_min_distance = 4 * 1024 * 1024;
 /* max bytes to prefetch per stream (default 64MB) */
 unsigned int	zfetch_max_distance = 64 * 1024 * 1024;
+#endif
 /* max bytes to prefetch indirects for per stream (default 64MB) */
 unsigned int	zfetch_max_idistance = 64 * 1024 * 1024;
-/* max number of bytes in an array_read in which we allow prefetching (1MB) */
-uint64_t	zfetch_array_rd_sz = 1024 * 1024;
 
 typedef struct zfetch_stats {
 	kstat_named_t zfetchstat_hits;
@@ -580,6 +585,3 @@ ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_distance, UINT, ZMOD_RW,
 
 ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_idistance, UINT, ZMOD_RW,
 	"Max bytes to prefetch indirects for per stream");
-
-ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, array_rd_sz, U64, ZMOD_RW,
-	"Number of bytes in a array_read");

module/zfs/dnode.c

@@ -1882,7 +1882,7 @@ dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx)
 	if (ibs == dn->dn_indblkshift)
 		ibs = 0;
 
-	if (size >> SPA_MINBLOCKSHIFT == dn->dn_datablkszsec && ibs == 0)
+	if (size == dn->dn_datablksz && ibs == 0)
 		return (0);
 
 	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
@@ -1905,24 +1905,25 @@ dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx)
 	if (ibs && dn->dn_nlevels != 1)
 		goto fail;
 
-	/* resize the old block */
-	err = dbuf_hold_impl(dn, 0, 0, TRUE, FALSE, FTAG, &db);
-	if (err == 0) {
-		dbuf_new_size(db, size, tx);
-	} else if (err != ENOENT) {
-		goto fail;
-	}
-
-	dnode_setdblksz(dn, size);
 	dnode_setdirty(dn, tx);
-	dn->dn_next_blksz[tx->tx_txg&TXG_MASK] = size;
+	if (size != dn->dn_datablksz) {
+		/* resize the old block */
+		err = dbuf_hold_impl(dn, 0, 0, TRUE, FALSE, FTAG, &db);
+		if (err == 0) {
+			dbuf_new_size(db, size, tx);
+		} else if (err != ENOENT) {
+			goto fail;
+		}
+
+		dnode_setdblksz(dn, size);
+		dn->dn_next_blksz[tx->tx_txg & TXG_MASK] = size;
+		if (db)
+			dbuf_rele(db, FTAG);
+	}
 	if (ibs) {
 		dn->dn_indblkshift = ibs;
-		dn->dn_next_indblkshift[tx->tx_txg&TXG_MASK] = ibs;
+		dn->dn_next_indblkshift[tx->tx_txg & TXG_MASK] = ibs;
 	}
-	/* release after we have fixed the blocksize in the dnode */
-	if (db)
-		dbuf_rele(db, FTAG);
 
 	rw_exit(&dn->dn_struct_rwlock);
 	return (0);

module/zfs/dsl_deadlist.c

@@ -892,9 +892,9 @@ dsl_deadlist_merge(dsl_deadlist_t *dl, uint64_t obj, dmu_tx_t *tx)
 	for (zap_cursor_init(&zc, dl->dl_os, obj);
 	    (error = zap_cursor_retrieve(&zc, za)) == 0;
 	    zap_cursor_advance(&zc)) {
-		uint64_t mintxg = zfs_strtonum(za->za_name, NULL);
-		dsl_deadlist_insert_bpobj(dl, za->za_first_integer, mintxg, tx);
-		VERIFY0(zap_remove_int(dl->dl_os, obj, mintxg, tx));
+		dsl_deadlist_insert_bpobj(dl, za->za_first_integer,
+		    zfs_strtonum(za->za_name, NULL), tx);
+		VERIFY0(zap_remove(dl->dl_os, obj, za->za_name, tx));
 		if (perror == 0) {
 			dsl_deadlist_prefetch_bpobj(dl, pza->za_first_integer,
 			    zfs_strtonum(pza->za_name, NULL));

module/zfs/dsl_scan.c

@@ -2015,6 +2015,11 @@ dsl_scan_prefetch_cb(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp,
 		    zb->zb_objset, DMU_META_DNODE_OBJECT);
 
 		if (OBJSET_BUF_HAS_USERUSED(buf)) {
+			if (OBJSET_BUF_HAS_PROJECTUSED(buf)) {
+				dsl_scan_prefetch_dnode(scn,
+				    &osp->os_projectused_dnode, zb->zb_objset,
+				    DMU_PROJECTUSED_OBJECT);
+			}
 			dsl_scan_prefetch_dnode(scn,
 			    &osp->os_groupused_dnode, zb->zb_objset,
 			    DMU_GROUPUSED_OBJECT);
@@ -2075,10 +2080,16 @@ dsl_scan_prefetch_thread(void *arg)
 			zio_flags |= ZIO_FLAG_RAW;
 		}
 
+		/* We don't need data L1 buffer since we do not prefetch L0. */
+		blkptr_t *bp = &spic->spic_bp;
+		if (BP_GET_LEVEL(bp) == 1 && BP_GET_TYPE(bp) != DMU_OT_DNODE &&
+		    BP_GET_TYPE(bp) != DMU_OT_OBJSET)
+			flags |= ARC_FLAG_NO_BUF;
+
 		/* issue the prefetch asynchronously */
-		(void) arc_read(scn->scn_zio_root, scn->scn_dp->dp_spa,
-		    &spic->spic_bp, dsl_scan_prefetch_cb, spic->spic_spc,
-		    ZIO_PRIORITY_SCRUB, zio_flags, &flags, &spic->spic_zb);
+		(void) arc_read(scn->scn_zio_root, spa, bp,
+		    dsl_scan_prefetch_cb, spic->spic_spc, ZIO_PRIORITY_SCRUB,
+		    zio_flags, &flags, &spic->spic_zb);
 
 		kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
 	}

module/zfs/metaslab.c

@@ -58,6 +58,11 @@ static uint64_t metaslab_aliquot = 1024 * 1024;
  */
 uint64_t metaslab_force_ganging = SPA_MAXBLOCKSIZE + 1;
 
+/*
+ * Of blocks of size >= metaslab_force_ganging, actually gang them this often.
+ */
+uint_t metaslab_force_ganging_pct = 3;
+
 /*
  * In pools where the log space map feature is not enabled we touch
  * multiple metaslabs (and their respective space maps) with each
@@ -1287,7 +1292,7 @@ metaslab_group_allocatable(metaslab_group_t *mg, metaslab_group_t *rotor,
 
 		/*
 		 * If this metaslab group is below its qmax or it's
-		 * the only allocatable metasable group, then attempt
+		 * the only allocatable metaslab group, then attempt
 		 * to allocate from it.
 		 */
 		if (qdepth < qmax || mc->mc_alloc_groups == 1)
@@ -5096,7 +5101,7 @@ metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
     zio_alloc_list_t *zal, int allocator)
 {
 	metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
-	metaslab_group_t *mg, *fast_mg, *rotor;
+	metaslab_group_t *mg, *rotor;
 	vdev_t *vd;
 	boolean_t try_hard = B_FALSE;
 
@@ -5109,7 +5114,9 @@ metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
 	 * damage can result in extremely long reconstruction times.  This
 	 * will also test spilling from special to normal.
 	 */
-	if (psize >= metaslab_force_ganging && (random_in_range(100) < 3)) {
+	if (psize >= metaslab_force_ganging &&
+	    metaslab_force_ganging_pct > 0 &&
+	    (random_in_range(100) < MIN(metaslab_force_ganging_pct, 100))) {
 		metaslab_trace_add(zal, NULL, NULL, psize, d, TRACE_FORCE_GANG,
 		    allocator);
 		return (SET_ERROR(ENOSPC));
@@ -5157,15 +5164,6 @@ metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
 	} else if (d != 0) {
 		vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
 		mg = vd->vdev_mg->mg_next;
-	} else if (flags & METASLAB_FASTWRITE) {
-		mg = fast_mg = mca->mca_rotor;
-
-		do {
-			if (fast_mg->mg_vd->vdev_pending_fastwrite <
-			    mg->mg_vd->vdev_pending_fastwrite)
-				mg = fast_mg;
-		} while ((fast_mg = fast_mg->mg_next) != mca->mca_rotor);
-
 	} else {
 		ASSERT(mca->mca_rotor != NULL);
 		mg = mca->mca_rotor;
@@ -5290,7 +5288,7 @@ top:
 				mg->mg_bias = 0;
 			}
 
-			if ((flags & METASLAB_FASTWRITE) ||
+			if ((flags & METASLAB_ZIL) ||
 			    atomic_add_64_nv(&mca->mca_aliquot, asize) >=
 			    mg->mg_aliquot + mg->mg_bias) {
 				mca->mca_rotor = mg->mg_next;
@@ -5303,11 +5301,6 @@ top:
 			    ((flags & METASLAB_GANG_HEADER) ? 1 : 0));
 			DVA_SET_ASIZE(&dva[d], asize);
 
-			if (flags & METASLAB_FASTWRITE) {
-				atomic_add_64(&vd->vdev_pending_fastwrite,
-				    psize);
-			}
-
 			return (0);
 		}
 next:
@@ -5943,55 +5936,6 @@ metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg)
 	return (error);
 }
 
-void
-metaslab_fastwrite_mark(spa_t *spa, const blkptr_t *bp)
-{
-	const dva_t *dva = bp->blk_dva;
-	int ndvas = BP_GET_NDVAS(bp);
-	uint64_t psize = BP_GET_PSIZE(bp);
-	int d;
-	vdev_t *vd;
-
-	ASSERT(!BP_IS_HOLE(bp));
-	ASSERT(!BP_IS_EMBEDDED(bp));
-	ASSERT(psize > 0);
-
-	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
-
-	for (d = 0; d < ndvas; d++) {
-		if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
-			continue;
-		atomic_add_64(&vd->vdev_pending_fastwrite, psize);
-	}
-
-	spa_config_exit(spa, SCL_VDEV, FTAG);
-}
-
-void
-metaslab_fastwrite_unmark(spa_t *spa, const blkptr_t *bp)
-{
-	const dva_t *dva = bp->blk_dva;
-	int ndvas = BP_GET_NDVAS(bp);
-	uint64_t psize = BP_GET_PSIZE(bp);
-	int d;
-	vdev_t *vd;
-
-	ASSERT(!BP_IS_HOLE(bp));
-	ASSERT(!BP_IS_EMBEDDED(bp));
-	ASSERT(psize > 0);
-
-	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
-
-	for (d = 0; d < ndvas; d++) {
-		if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
-			continue;
-		ASSERT3U(vd->vdev_pending_fastwrite, >=, psize);
-		atomic_sub_64(&vd->vdev_pending_fastwrite, psize);
-	}
-
-	spa_config_exit(spa, SCL_VDEV, FTAG);
-}
-
 static void
 metaslab_check_free_impl_cb(uint64_t inner, vdev_t *vd, uint64_t offset,
     uint64_t size, void *arg)
@@ -6266,7 +6210,10 @@ ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, switch_threshold, INT, ZMOD_RW,
 	"Segment-based metaslab selection maximum buckets before switching");
 
 ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, force_ganging, U64, ZMOD_RW,
-	"Blocks larger than this size are forced to be gang blocks");
+	"Blocks larger than this size are sometimes forced to be gang blocks");
+
+ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, force_ganging_pct, UINT, ZMOD_RW,
+	"Percentage of large blocks that will be forced to be gang blocks");
 
 ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_max_search, UINT, ZMOD_RW,
 	"Max distance (bytes) to search forward before using size tree");

module/zfs/spa_misc.c

@@ -772,6 +772,7 @@ spa_add(const char *name, nvlist_t *config, const char *altroot)
 	spa->spa_min_ashift = INT_MAX;
 	spa->spa_max_ashift = 0;
 	spa->spa_min_alloc = INT_MAX;
+	spa->spa_gcd_alloc = INT_MAX;
 
 	/* Reset cached value */
 	spa->spa_dedup_dspace = ~0ULL;

module/zfs/vdev.c

@@ -889,9 +889,15 @@ vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *nv, vdev_t *parent, uint_t id,
 	    &vd->vdev_not_present);
 
 	/*
-	 * Get the alignment requirement.
+	 * Get the alignment requirement. Ignore pool ashift for vdev
+	 * attach case.
 	 */
-	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT, &vd->vdev_ashift);
+	if (alloctype != VDEV_ALLOC_ATTACH) {
+		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT,
+		    &vd->vdev_ashift);
+	} else {
+		vd->vdev_attaching = B_TRUE;
+	}
 
 	/*
 	 * Retrieve the vdev creation time.
@@ -1186,7 +1192,6 @@ vdev_top_transfer(vdev_t *svd, vdev_t *tvd)
 
 	ASSERT(tvd == tvd->vdev_top);
 
-	tvd->vdev_pending_fastwrite = svd->vdev_pending_fastwrite;
 	tvd->vdev_ms_array = svd->vdev_ms_array;
 	tvd->vdev_ms_shift = svd->vdev_ms_shift;
 	tvd->vdev_ms_count = svd->vdev_ms_count;
@@ -1393,6 +1398,36 @@ vdev_remove_parent(vdev_t *cvd)
 	vdev_free(mvd);
 }
 
+/*
+ * Choose GCD for spa_gcd_alloc.
+ */
+static uint64_t
+vdev_gcd(uint64_t a, uint64_t b)
+{
+	while (b != 0) {
+		uint64_t t = b;
+		b = a % b;
+		a = t;
+	}
+	return (a);
+}
+
+/*
+ * Set spa_min_alloc and spa_gcd_alloc.
+ */
+static void
+vdev_spa_set_alloc(spa_t *spa, uint64_t min_alloc)
+{
+	if (min_alloc < spa->spa_min_alloc)
+		spa->spa_min_alloc = min_alloc;
+	if (spa->spa_gcd_alloc == INT_MAX) {
+		spa->spa_gcd_alloc = min_alloc;
+	} else {
+		spa->spa_gcd_alloc = vdev_gcd(min_alloc,
+		    spa->spa_gcd_alloc);
+	}
+}
+
 void
 vdev_metaslab_group_create(vdev_t *vd)
 {
@@ -1445,8 +1480,7 @@ vdev_metaslab_group_create(vdev_t *vd)
 				spa->spa_min_ashift = vd->vdev_ashift;
 
 			uint64_t min_alloc = vdev_get_min_alloc(vd);
-			if (min_alloc < spa->spa_min_alloc)
-				spa->spa_min_alloc = min_alloc;
+			vdev_spa_set_alloc(spa, min_alloc);
 		}
 	}
 }
@@ -1620,7 +1654,6 @@ vdev_metaslab_fini(vdev_t *vd)
 		}
 	}
 	ASSERT0(vd->vdev_ms_count);
-	ASSERT3U(vd->vdev_pending_fastwrite, ==, 0);
 }
 
 typedef struct vdev_probe_stats {
@@ -2144,9 +2177,9 @@ vdev_open(vdev_t *vd)
 				return (SET_ERROR(EDOM));
 			}
 
-			if (vd->vdev_top == vd) {
+			if (vd->vdev_top == vd && vd->vdev_attaching == B_FALSE)
 				vdev_ashift_optimize(vd);
-			}
+			vd->vdev_attaching = B_FALSE;
 		}
 		if (vd->vdev_ashift != 0 && (vd->vdev_ashift < ASHIFT_MIN ||
 		    vd->vdev_ashift > ASHIFT_MAX)) {
@@ -2207,8 +2240,7 @@ vdev_open(vdev_t *vd)
 	if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
 	    vd->vdev_islog == 0 && vd->vdev_aux == NULL) {
 		uint64_t min_alloc = vdev_get_min_alloc(vd);
-		if (min_alloc < spa->spa_min_alloc)
-			spa->spa_min_alloc = min_alloc;
+		vdev_spa_set_alloc(spa, min_alloc);
 	}
 
 	/*
@@ -5688,6 +5720,7 @@ vdev_props_set_sync(void *arg, dmu_tx_t *tx)
 	objset_t *mos = spa->spa_meta_objset;
 	nvpair_t *elem = NULL;
 	uint64_t vdev_guid;
+	uint64_t objid;
 	nvlist_t *nvprops;
 
 	vdev_guid = fnvlist_lookup_uint64(nvp, ZPOOL_VDEV_PROPS_SET_VDEV);
@@ -5698,31 +5731,28 @@ vdev_props_set_sync(void *arg, dmu_tx_t *tx)
 	if (vd == NULL)
 		return;
 
+	/*
+	 * Set vdev property values in the vdev props mos object.
+	 */
+	if (vd->vdev_root_zap != 0) {
+		objid = vd->vdev_root_zap;
+	} else if (vd->vdev_top_zap != 0) {
+		objid = vd->vdev_top_zap;
+	} else if (vd->vdev_leaf_zap != 0) {
+		objid = vd->vdev_leaf_zap;
+	} else {
+		panic("unexpected vdev type");
+	}
+
 	mutex_enter(&spa->spa_props_lock);
 
 	while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
-		uint64_t intval, objid = 0;
+		uint64_t intval;
 		const char *strval;
 		vdev_prop_t prop;
 		const char *propname = nvpair_name(elem);
 		zprop_type_t proptype;
 
-		/*
-		 * Set vdev property values in the vdev props mos object.
-		 */
-		if (vd->vdev_root_zap != 0) {
-			objid = vd->vdev_root_zap;
-		} else if (vd->vdev_top_zap != 0) {
-			objid = vd->vdev_top_zap;
-		} else if (vd->vdev_leaf_zap != 0) {
-			objid = vd->vdev_leaf_zap;
-		} else {
-			/*
-			 * XXX: implement vdev_props_set_check()
-			 */
-			panic("vdev not root/top/leaf");
-		}
-
 		switch (prop = vdev_name_to_prop(propname)) {
 		case VDEV_PROP_USERPROP:
 			if (vdev_prop_user(propname)) {
@@ -5791,6 +5821,12 @@ vdev_prop_set(vdev_t *vd, nvlist_t *innvl, nvlist_t *outnvl)
 
 	ASSERT(vd != NULL);
 
+	/* Check that vdev has a zap we can use */
+	if (vd->vdev_root_zap == 0 &&
+	    vd->vdev_top_zap == 0 &&
+	    vd->vdev_leaf_zap == 0)
+		return (SET_ERROR(EINVAL));
+
 	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
 	    &vdev_guid) != 0)
 		return (SET_ERROR(EINVAL));

module/zfs/vdev_indirect.c

@@ -1398,7 +1398,7 @@ vdev_indirect_checksum_error(zio_t *zio,
 	vd->vdev_stat.vs_checksum_errors++;
 	mutex_exit(&vd->vdev_stat_lock);
 
-	zio_bad_cksum_t zbc = {{{ 0 }}};
+	zio_bad_cksum_t zbc = { 0 };
 	abd_t *bad_abd = ic->ic_data;
 	abd_t *good_abd = is->is_good_child->ic_data;
 	(void) zfs_ereport_post_checksum(zio->io_spa, vd, NULL, zio,

module/zfs/vdev_raidz.c

@@ -1785,7 +1785,7 @@ vdev_raidz_checksum_error(zio_t *zio, raidz_col_t *rc, abd_t *bad_data)
 static int
 raidz_checksum_verify(zio_t *zio)
 {
-	zio_bad_cksum_t zbc = {{{0}}};
+	zio_bad_cksum_t zbc = {0};
 	raidz_map_t *rm = zio->io_vsd;
 
 	int ret = zio_checksum_error(zio, &zbc);

module/zfs/zfs_fm.c

@@ -754,10 +754,6 @@ zfs_ereport_start(nvlist_t **ereport_out, nvlist_t **detector_out,
 #define	MAX_RANGES		16
 
 typedef struct zfs_ecksum_info {
-	/* histograms of set and cleared bits by bit number in a 64-bit word */
-	uint8_t zei_histogram_set[sizeof (uint64_t) * NBBY];
-	uint8_t zei_histogram_cleared[sizeof (uint64_t) * NBBY];
-
 	/* inline arrays of bits set and cleared. */
 	uint64_t zei_bits_set[ZFM_MAX_INLINE];
 	uint64_t zei_bits_cleared[ZFM_MAX_INLINE];
@@ -781,7 +777,7 @@ typedef struct zfs_ecksum_info {
 } zfs_ecksum_info_t;
 
 static void
-update_histogram(uint64_t value_arg, uint8_t *hist, uint32_t *count)
+update_bad_bits(uint64_t value_arg, uint32_t *count)
 {
 	size_t i;
 	size_t bits = 0;
@@ -789,10 +785,8 @@ update_histogram(uint64_t value_arg, uint8_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 (value & (1ull << i))
 			++bits;
-		}
 	}
 	/* update the count of bits changed */
 	*count += bits;
@@ -920,14 +914,6 @@ annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info,
 
 	if (info != NULL && info->zbc_has_cksum) {
 		fm_payload_set(ereport,
-		    FM_EREPORT_PAYLOAD_ZFS_CKSUM_EXPECTED,
-		    DATA_TYPE_UINT64_ARRAY,
-		    sizeof (info->zbc_expected) / sizeof (uint64_t),
-		    (uint64_t *)&info->zbc_expected,
-		    FM_EREPORT_PAYLOAD_ZFS_CKSUM_ACTUAL,
-		    DATA_TYPE_UINT64_ARRAY,
-		    sizeof (info->zbc_actual) / sizeof (uint64_t),
-		    (uint64_t *)&info->zbc_actual,
 		    FM_EREPORT_PAYLOAD_ZFS_CKSUM_ALGO,
 		    DATA_TYPE_STRING,
 		    info->zbc_checksum_name,
@@ -1010,10 +996,8 @@ annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info,
 				offset++;
 			}
 
-			update_histogram(set, eip->zei_histogram_set,
-			    &eip->zei_range_sets[range]);
-			update_histogram(cleared, eip->zei_histogram_cleared,
-			    &eip->zei_range_clears[range]);
+			update_bad_bits(set, &eip->zei_range_sets[range]);
+			update_bad_bits(cleared, &eip->zei_range_clears[range]);
 		}
 
 		/* convert to byte offsets */
@@ -1049,15 +1033,6 @@ annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info,
 		    DATA_TYPE_UINT8_ARRAY,
 		    inline_size, (uint8_t *)eip->zei_bits_cleared,
 		    NULL);
-	} else {
-		fm_payload_set(ereport,
-		    FM_EREPORT_PAYLOAD_ZFS_BAD_SET_HISTOGRAM,
-		    DATA_TYPE_UINT8_ARRAY,
-		    NBBY * sizeof (uint64_t), eip->zei_histogram_set,
-		    FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_HISTOGRAM,
-		    DATA_TYPE_UINT8_ARRAY,
-		    NBBY * sizeof (uint64_t), eip->zei_histogram_cleared,
-		    NULL);
 	}
 	return (eip);
 }

module/zfs/zfs_vnops.c

@@ -839,7 +839,6 @@ zfs_get_data(void *arg, uint64_t gen, lr_write_t *lr, char *buf,
 	uint64_t zp_gen;
 
 	ASSERT3P(lwb, !=, NULL);
-	ASSERT3P(zio, !=, NULL);
 	ASSERT3U(size, !=, 0);
 
 	/*
@@ -889,6 +888,7 @@ zfs_get_data(void *arg, uint64_t gen, lr_write_t *lr, char *buf,
 		}
 		ASSERT(error == 0 || error == ENOENT);
 	} else { /* indirect write */
+		ASSERT3P(zio, !=, NULL);
 		/*
 		 * Have to lock the whole block to ensure when it's
 		 * written out and its checksum is being calculated
@@ -917,8 +917,8 @@ zfs_get_data(void *arg, uint64_t gen, lr_write_t *lr, char *buf,
 		}
 #endif
 		if (error == 0)
-			error = dmu_buf_hold(os, object, offset, zgd, &db,
-			    DMU_READ_NO_PREFETCH);
+			error = dmu_buf_hold_noread(os, object, offset, zgd,
+			    &db);
 
 		if (error == 0) {
 			blkptr_t *bp = &lr->lr_blkptr;
@@ -1028,6 +1028,10 @@ zfs_exit_two(zfsvfs_t *zfsvfs1, zfsvfs_t *zfsvfs2, const char *tag)
  *
  * On success, the function return the number of bytes copied in *lenp.
  * Note, it doesn't return how much bytes are left to be copied.
+ * On errors which are caused by any file system limitations or
+ * brt limitations `EINVAL` is returned. In the most cases a user
+ * requested bad parameters, it could be possible to clone the file but
+ * some parameters don't match the requirements.
  */
 int
 zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
@@ -1078,6 +1082,16 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 		return (SET_ERROR(EXDEV));
 	}
 
+	/*
+	 * outos and inos belongs to the same storage pool.
+	 * see a few lines above, only one check.
+	 */
+	if (!spa_feature_is_enabled(dmu_objset_spa(outos),
+	    SPA_FEATURE_BLOCK_CLONING)) {
+		zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
+		return (SET_ERROR(EOPNOTSUPP));
+	}
+
 	ASSERT(!outzfsvfs->z_replay);
 
 	error = zfs_verify_zp(inzp);
@@ -1088,12 +1102,6 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 		return (error);
 	}
 
-	if (!spa_feature_is_enabled(dmu_objset_spa(outos),
-	    SPA_FEATURE_BLOCK_CLONING)) {
-		zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
-		return (SET_ERROR(EXDEV));
-	}
-
 	/*
 	 * We don't copy source file's flags that's why we don't allow to clone
 	 * files that are in quarantine.
@@ -1167,7 +1175,7 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 	 * We cannot clone into files with different block size.
 	 */
 	if (inblksz != outzp->z_blksz && outzp->z_size > inblksz) {
-		error = SET_ERROR(EXDEV);
+		error = SET_ERROR(EINVAL);
 		goto unlock;
 	}
 
@@ -1175,7 +1183,7 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 	 * Offsets and len must be at block boundries.
 	 */
 	if ((inoff % inblksz) != 0 || (outoff % inblksz) != 0) {
-		error = SET_ERROR(EXDEV);
+		error = SET_ERROR(EINVAL);
 		goto unlock;
 	}
 	/*
@@ -1183,7 +1191,7 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 	 */
 	if ((len % inblksz) != 0 &&
 	    (len < inzp->z_size - inoff || len < outzp->z_size - outoff)) {
-		error = SET_ERROR(EXDEV);
+		error = SET_ERROR(EINVAL);
 		goto unlock;
 	}
 
@@ -1212,7 +1220,7 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 	gid = KGID_TO_SGID(ZTOGID(outzp));
 	projid = outzp->z_projid;
 
-	bps = kmem_alloc(sizeof (bps[0]) * maxblocks, KM_SLEEP);
+	bps = vmem_alloc(sizeof (bps[0]) * maxblocks, KM_SLEEP);
 
 	/*
 	 * Clone the file in reasonable size chunks.  Each chunk is cloned
@@ -1238,13 +1246,11 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 		    &nbps);
 		if (error != 0) {
 			/*
-			 * If we are tyring to clone a block that was created
-			 * in the current transaction group. Return an error,
-			 * so the caller can fallback to just copying the data.
+			 * 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.
 			 */
-			if (error == EAGAIN) {
-				error = SET_ERROR(EXDEV);
-			}
 			break;
 		}
 		/*
@@ -1330,7 +1336,7 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 		done += size;
 	}
 
-	kmem_free(bps, sizeof (bps[0]) * maxblocks);
+	vmem_free(bps, sizeof (bps[0]) * maxblocks);
 	zfs_znode_update_vfs(outzp);
 
 unlock:

module/zfs/zio.c

@@ -1596,6 +1596,19 @@ zio_shrink(zio_t *zio, uint64_t size)
 	}
 }
 
+/*
+ * Round provided allocation size up to a value that can be allocated
+ * by at least some vdev(s) in the pool with minimum or no additional
+ * padding and without extra space usage on others
+ */
+static uint64_t
+zio_roundup_alloc_size(spa_t *spa, uint64_t size)
+{
+	if (size > spa->spa_min_alloc)
+		return (roundup(size, spa->spa_gcd_alloc));
+	return (spa->spa_min_alloc);
+}
+
 /*
  * ==========================================================================
  * Prepare to read and write logical blocks
@@ -1762,8 +1775,9 @@ zio_write_compress(zio_t *zio)
 			compress = ZIO_COMPRESS_OFF;
 
 		/* Make sure someone doesn't change their mind on overwrites */
-		ASSERT(BP_IS_EMBEDDED(bp) || MIN(zp->zp_copies + BP_IS_GANG(bp),
-		    spa_max_replication(spa)) == BP_GET_NDVAS(bp));
+		ASSERT(BP_IS_EMBEDDED(bp) || BP_IS_GANG(bp) ||
+		    MIN(zp->zp_copies, spa_max_replication(spa))
+		    == BP_GET_NDVAS(bp));
 	}
 
 	/* If it's a compressed write that is not raw, compress the buffer. */
@@ -1802,9 +1816,8 @@ zio_write_compress(zio_t *zio)
 			 * in that we charge for the padding used to fill out
 			 * the last sector.
 			 */
-			ASSERT3U(spa->spa_min_alloc, >=, SPA_MINBLOCKSHIFT);
-			size_t rounded = (size_t)roundup(psize,
-			    spa->spa_min_alloc);
+			size_t rounded = (size_t)zio_roundup_alloc_size(spa,
+			    psize);
 			if (rounded >= lsize) {
 				compress = ZIO_COMPRESS_OFF;
 				zio_buf_free(cbuf, lsize);
@@ -1847,8 +1860,8 @@ zio_write_compress(zio_t *zio)
 		 * take this codepath because it will change the on-disk block
 		 * and decryption will fail.
 		 */
-		size_t rounded = MIN((size_t)roundup(psize,
-		    spa->spa_min_alloc), lsize);
+		size_t rounded = MIN((size_t)zio_roundup_alloc_size(spa, psize),
+		    lsize);
 
 		if (rounded != psize) {
 			abd_t *cdata = abd_alloc_linear(rounded, B_TRUE);
@@ -3012,11 +3025,6 @@ zio_write_gang_block(zio_t *pio, metaslab_class_t *mc)
 	 */
 	pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
 
-	/*
-	 * We didn't allocate this bp, so make sure it doesn't get unmarked.
-	 */
-	pio->io_flags &= ~ZIO_FLAG_FASTWRITE;
-
 	zio_nowait(zio);
 
 	return (pio);
@@ -3604,7 +3612,6 @@ zio_dva_allocate(zio_t *zio)
 	ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa));
 	ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));
 
-	flags |= (zio->io_flags & ZIO_FLAG_FASTWRITE) ? METASLAB_FASTWRITE : 0;
 	if (zio->io_flags & ZIO_FLAG_NODATA)
 		flags |= METASLAB_DONT_THROTTLE;
 	if (zio->io_flags & ZIO_FLAG_GANG_CHILD)
@@ -3764,7 +3771,7 @@ zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
 	 * of, so we just hash the objset ID to pick the allocator to get
 	 * some parallelism.
 	 */
-	int flags = METASLAB_FASTWRITE | METASLAB_ZIL;
+	int flags = METASLAB_ZIL;
 	int allocator = (uint_t)cityhash4(0, 0, 0,
 	    os->os_dsl_dataset->ds_object) % spa->spa_alloc_count;
 	error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1,
@@ -4460,8 +4467,8 @@ zio_ready(zio_t *zio)
 	zio_t *pio, *pio_next;
 	zio_link_t *zl = NULL;
 
-	if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT | ZIO_CHILD_DDT_BIT,
-	    ZIO_WAIT_READY)) {
+	if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT |
+	    ZIO_CHILD_GANG_BIT | ZIO_CHILD_DDT_BIT, ZIO_WAIT_READY)) {
 		return (NULL);
 	}
 
@@ -4919,12 +4926,6 @@ zio_done(zio_t *zio)
 		zfs_ereport_free_checksum(zcr);
 	}
 
-	if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp &&
-	    !BP_IS_HOLE(zio->io_bp) && !BP_IS_EMBEDDED(zio->io_bp) &&
-	    !(zio->io_flags & ZIO_FLAG_NOPWRITE)) {
-		metaslab_fastwrite_unmark(zio->io_spa, zio->io_bp);
-	}
-
 	/*
 	 * It is the responsibility of the done callback to ensure that this
 	 * particular zio is no longer discoverable for adoption, and as

module/zfs/zio_checksum.c

@@ -515,8 +515,6 @@ zio_checksum_error_impl(spa_t *spa, const blkptr_t *bp,
 	}
 
 	if (info != NULL) {
-		info->zbc_expected = expected_cksum;
-		info->zbc_actual = actual_cksum;
 		info->zbc_checksum_name = ci->ci_name;
 		info->zbc_byteswapped = byteswap;
 		info->zbc_injected = 0;

module/zfs/zvol.c

@@ -698,7 +698,6 @@ zvol_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
 	int error;
 
 	ASSERT3P(lwb, !=, NULL);
-	ASSERT3P(zio, !=, NULL);
 	ASSERT3U(size, !=, 0);
 
 	zgd = kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
@@ -717,6 +716,7 @@ zvol_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
 		error = dmu_read_by_dnode(zv->zv_dn, offset, size, buf,
 		    DMU_READ_NO_PREFETCH);
 	} else { /* indirect write */
+		ASSERT3P(zio, !=, NULL);
 		/*
 		 * Have to lock the whole block to ensure when it's written out
 		 * and its checksum is being calculated that no one can change
@@ -727,8 +727,8 @@ zvol_get_data(void *arg, uint64_t arg2, lr_write_t *lr, char *buf,
 		offset = P2ALIGN_TYPED(offset, size, uint64_t);
 		zgd->zgd_lr = zfs_rangelock_enter(&zv->zv_rangelock, offset,
 		    size, RL_READER);
-		error = dmu_buf_hold_by_dnode(zv->zv_dn, offset, zgd, &db,
-		    DMU_READ_NO_PREFETCH);
+		error = dmu_buf_hold_noread_by_dnode(zv->zv_dn, offset, zgd,
+		    &db);
 		if (error == 0) {
 			blkptr_t *bp = &lr->lr_blkptr;
 

tests/runfiles/linux.run

@@ -34,6 +34,17 @@ tags = ['functional', 'acl', 'posix-sa']
 tests = ['atime_003_pos', 'root_relatime_on']
 tags = ['functional', 'atime']
 
+[tests/functional/block_cloning:Linux]
+tests = ['block_cloning_copyfilerange', 'block_cloning_copyfilerange_partial',
+    'block_cloning_copyfilerange_fallback',
+    'block_cloning_ficlone', 'block_cloning_ficlonerange',
+    'block_cloning_ficlonerange_partial',
+    'block_cloning_disabled_copyfilerange', 'block_cloning_disabled_ficlone',
+    'block_cloning_disabled_ficlonerange',
+    'block_cloning_copyfilerange_cross_dataset',
+    'block_cloning_copyfilerange_fallback_same_txg']
+tags = ['functional', 'block_cloning']
+
 [tests/functional/chattr:Linux]
 tests = ['chattr_001_pos', 'chattr_002_neg']
 tags = ['functional', 'chattr']

tests/test-runner/bin/zts-report.py.in

@@ -134,6 +134,12 @@ ci_reason = 'CI runner doesn\'t have all requirements'
 #
 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+'
+
 #
 # These tests are known to fail, thus we use this list to prevent these
 # failures from failing the job as a whole; only unexpected failures
@@ -288,6 +294,18 @@ elif sys.platform.startswith('linux'):
         'idmap_mount/idmap_mount_003': ['SKIP', idmap_reason],
         'idmap_mount/idmap_mount_004': ['SKIP', idmap_reason],
         'idmap_mount/idmap_mount_005': ['SKIP', idmap_reason],
+        'block_cloning/block_cloning_disabled_copyfilerange':
+            ['SKIP', cfr_reason],
+        'block_cloning/block_cloning_copyfilerange':
+            ['SKIP', cfr_reason],
+        'block_cloning/block_cloning_copyfilerange_partial':
+            ['SKIP', cfr_reason],
+        'block_cloning/block_cloning_copyfilerange_fallback':
+            ['SKIP', cfr_reason],
+        'block_cloning/block_cloning_copyfilerange_cross_dataset':
+            ['SKIP', cfr_cross_reason],
+        'block_cloning/block_cloning_copyfilerange_fallback_same_txg':
+            ['SKIP', cfr_cross_reason],
     })
 
 

tests/zfs-tests/cmd/.gitignore

@@ -1,6 +1,7 @@
 /badsend
 /btree_test
 /chg_usr_exec
+/clonefile
 /devname2devid
 /dir_rd_update
 /draid

tests/zfs-tests/cmd/Makefile.am

@@ -119,6 +119,7 @@ scripts_zfs_tests_bin_PROGRAMS += %D%/renameat2
 scripts_zfs_tests_bin_PROGRAMS += %D%/xattrtest
 scripts_zfs_tests_bin_PROGRAMS += %D%/zed_fd_spill-zedlet
 scripts_zfs_tests_bin_PROGRAMS += %D%/idmap_util
+scripts_zfs_tests_bin_PROGRAMS += %D%/clonefile
 
 %C%_idmap_util_LDADD = libspl.la
 

tests/zfs-tests/cmd/clonefile.c

@@ -0,0 +1,333 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright (c) 2023, Rob Norris <robn@despairlabs.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+/*
+ * This program is to test the availability and behaviour of copy_file_range,
+ * FICLONE, FICLONERANGE and FIDEDUPERANGE in the Linux kernel. It should
+ * compile and run even if these features aren't exposed through the libc.
+ */
+
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <stdlib.h>
+#include <limits.h>
+#include <stdio.h>
+#include <string.h>
+#include <errno.h>
+
+#ifndef __NR_copy_file_range
+#if defined(__x86_64__)
+#define	__NR_copy_file_range (326)
+#elif defined(__i386__)
+#define	__NR_copy_file_range (377)
+#elif defined(__s390__)
+#define	__NR_copy_file_range (375)
+#elif defined(__arm__)
+#define	__NR_copy_file_range (391)
+#elif defined(__aarch64__)
+#define	__NR_copy_file_range (285)
+#elif defined(__powerpc__)
+#define	__NR_copy_file_range (379)
+#else
+#error "no definition of __NR_copy_file_range for this platform"
+#endif
+#endif /* __NR_copy_file_range */
+
+ssize_t
+copy_file_range(int, loff_t *, int, loff_t *, size_t, unsigned int)
+    __attribute__((weak));
+
+static inline ssize_t
+cf_copy_file_range(int sfd, loff_t *soff, int dfd, loff_t *doff,
+    size_t len, unsigned int flags)
+{
+	if (copy_file_range)
+		return (copy_file_range(sfd, soff, dfd, doff, len, flags));
+	return (
+	    syscall(__NR_copy_file_range, sfd, soff, dfd, doff, len, flags));
+}
+
+/* Define missing FICLONE */
+#ifdef FICLONE
+#define	CF_FICLONE	FICLONE
+#else
+#define	CF_FICLONE	_IOW(0x94, 9, int)
+#endif
+
+/* Define missing FICLONERANGE and support structs */
+#ifdef FICLONERANGE
+#define	CF_FICLONERANGE	FICLONERANGE
+typedef struct file_clone_range cf_file_clone_range_t;
+#else
+typedef struct {
+	int64_t		src_fd;
+	uint64_t	src_offset;
+	uint64_t	src_length;
+	uint64_t	dest_offset;
+} cf_file_clone_range_t;
+#define	CF_FICLONERANGE	_IOW(0x94, 13, cf_file_clone_range_t)
+#endif
+
+/* Define missing FIDEDUPERANGE and support structs */
+#ifdef FIDEDUPERANGE
+#define	CF_FIDEDUPERANGE		FIDEDUPERANGE
+#define	CF_FILE_DEDUPE_RANGE_SAME	FILE_DEDUPE_RANGE_SAME
+#define	CF_FILE_DEDUPE_RANGE_DIFFERS	FILE_DEDUPE_RANGE_DIFFERS
+typedef struct file_dedupe_range_info	cf_file_dedupe_range_info_t;
+typedef struct file_dedupe_range	cf_file_dedupe_range_t;
+#else
+typedef struct {
+	int64_t dest_fd;
+	uint64_t dest_offset;
+	uint64_t bytes_deduped;
+	int32_t status;
+	uint32_t reserved;
+} cf_file_dedupe_range_info_t;
+typedef struct {
+	uint64_t src_offset;
+	uint64_t src_length;
+	uint16_t dest_count;
+	uint16_t reserved1;
+	uint32_t reserved2;
+	cf_file_dedupe_range_info_t info[0];
+} cf_file_dedupe_range_t;
+#define	CF_FIDEDUPERANGE		_IOWR(0x94, 54, cf_file_dedupe_range_t)
+#define	CF_FILE_DEDUPE_RANGE_SAME	(0)
+#define	CF_FILE_DEDUPE_RANGE_DIFFERS	(1)
+#endif
+
+typedef enum {
+	CF_MODE_NONE,
+	CF_MODE_CLONE,
+	CF_MODE_CLONERANGE,
+	CF_MODE_COPYFILERANGE,
+	CF_MODE_DEDUPERANGE,
+} cf_mode_t;
+
+static int
+usage(void)
+{
+	printf(
+	    "usage:\n"
+	    "  FICLONE:\n"
+	    "    clonefile -c <src> <dst>\n"
+	    "  FICLONERANGE:\n"
+	    "    clonefile -r <src> <dst> <soff> <doff> <len>\n"
+	    "  copy_file_range:\n"
+	    "    clonefile -f <src> <dst> <soff> <doff> <len>\n"
+	    "  FIDEDUPERANGE:\n"
+	    "    clonefile -d <src> <dst> <soff> <doff> <len>\n");
+	return (1);
+}
+
+int do_clone(int sfd, int dfd);
+int do_clonerange(int sfd, int dfd, loff_t soff, loff_t doff, size_t len);
+int do_copyfilerange(int sfd, int dfd, loff_t soff, loff_t doff, size_t len);
+int do_deduperange(int sfd, int dfd, loff_t soff, loff_t doff, size_t len);
+
+int quiet = 0;
+
+int
+main(int argc, char **argv)
+{
+	cf_mode_t mode = CF_MODE_NONE;
+
+	char c;
+	while ((c = getopt(argc, argv, "crfdq")) != -1) {
+		switch (c) {
+			case 'c':
+				mode = CF_MODE_CLONE;
+				break;
+			case 'r':
+				mode = CF_MODE_CLONERANGE;
+				break;
+			case 'f':
+				mode = CF_MODE_COPYFILERANGE;
+				break;
+			case 'd':
+				mode = CF_MODE_DEDUPERANGE;
+				break;
+			case 'q':
+				quiet = 1;
+				break;
+		}
+	}
+
+	if (mode == CF_MODE_NONE || (argc-optind) < 2 ||
+	    (mode != CF_MODE_CLONE && (argc-optind) < 5))
+		return (usage());
+
+	loff_t soff = 0, doff = 0;
+	size_t len = 0;
+	if (mode != CF_MODE_CLONE) {
+		soff = strtoull(argv[optind+2], NULL, 10);
+		if (soff == ULLONG_MAX) {
+			fprintf(stderr, "invalid source offset");
+			return (1);
+		}
+		doff = strtoull(argv[optind+3], NULL, 10);
+		if (doff == ULLONG_MAX) {
+			fprintf(stderr, "invalid dest offset");
+			return (1);
+		}
+		len = strtoull(argv[optind+4], NULL, 10);
+		if (len == ULLONG_MAX) {
+			fprintf(stderr, "invalid length");
+			return (1);
+		}
+	}
+
+	int sfd = open(argv[optind], O_RDONLY);
+	if (sfd < 0) {
+		fprintf(stderr, "open: %s: %s\n",
+		    argv[optind], strerror(errno));
+		return (1);
+	}
+
+	int dfd = open(argv[optind+1], O_WRONLY|O_CREAT,
+	    S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH);
+	if (dfd < 0) {
+		fprintf(stderr, "open: %s: %s\n",
+		    argv[optind+1], strerror(errno));
+		close(sfd);
+		return (1);
+	}
+
+	int err;
+	switch (mode) {
+		case CF_MODE_CLONE:
+			err = do_clone(sfd, dfd);
+			break;
+		case CF_MODE_CLONERANGE:
+			err = do_clonerange(sfd, dfd, soff, doff, len);
+			break;
+		case CF_MODE_COPYFILERANGE:
+			err = do_copyfilerange(sfd, dfd, soff, doff, len);
+			break;
+		case CF_MODE_DEDUPERANGE:
+			err = do_deduperange(sfd, dfd, soff, doff, len);
+			break;
+		default:
+			abort();
+	}
+
+	off_t spos = lseek(sfd, 0, SEEK_CUR);
+	off_t slen = lseek(sfd, 0, SEEK_END);
+	off_t dpos = lseek(dfd, 0, SEEK_CUR);
+	off_t dlen = lseek(dfd, 0, SEEK_END);
+
+	fprintf(stderr, "file offsets: src=%lu/%lu; dst=%lu/%lu\n", spos, slen,
+	    dpos, dlen);
+
+	close(dfd);
+	close(sfd);
+
+	return (err == 0 ? 0 : 1);
+}
+
+int
+do_clone(int sfd, int dfd)
+{
+	fprintf(stderr, "using FICLONE\n");
+	int err = ioctl(dfd, CF_FICLONE, sfd);
+	if (err < 0) {
+		fprintf(stderr, "ioctl(FICLONE): %s\n", strerror(errno));
+		return (err);
+	}
+	return (0);
+}
+
+int
+do_clonerange(int sfd, int dfd, loff_t soff, loff_t doff, size_t len)
+{
+	fprintf(stderr, "using FICLONERANGE\n");
+	cf_file_clone_range_t fcr = {
+		.src_fd = sfd,
+		.src_offset = soff,
+		.src_length = len,
+		.dest_offset = doff,
+	};
+	int err = ioctl(dfd, CF_FICLONERANGE, &fcr);
+	if (err < 0) {
+		fprintf(stderr, "ioctl(FICLONERANGE): %s\n", strerror(errno));
+		return (err);
+	}
+	return (0);
+}
+
+int
+do_copyfilerange(int sfd, int dfd, loff_t soff, loff_t doff, size_t len)
+{
+	fprintf(stderr, "using copy_file_range\n");
+	ssize_t copied = cf_copy_file_range(sfd, &soff, dfd, &doff, len, 0);
+	if (copied < 0) {
+		fprintf(stderr, "copy_file_range: %s\n", strerror(errno));
+		return (1);
+	}
+	if (copied != len) {
+		fprintf(stderr, "copy_file_range: copied less than requested: "
+		    "requested=%lu; copied=%lu\n", len, copied);
+		return (1);
+	}
+	return (0);
+}
+
+int
+do_deduperange(int sfd, int dfd, loff_t soff, loff_t doff, size_t len)
+{
+	fprintf(stderr, "using FIDEDUPERANGE\n");
+
+	char buf[sizeof (cf_file_dedupe_range_t)+
+	    sizeof (cf_file_dedupe_range_info_t)] = {0};
+	cf_file_dedupe_range_t *fdr = (cf_file_dedupe_range_t *)&buf[0];
+	cf_file_dedupe_range_info_t *fdri =
+	    (cf_file_dedupe_range_info_t *)
+	    &buf[sizeof (cf_file_dedupe_range_t)];
+
+	fdr->src_offset = soff;
+	fdr->src_length = len;
+	fdr->dest_count = 1;
+
+	fdri->dest_fd = dfd;
+	fdri->dest_offset = doff;
+
+	int err = ioctl(sfd, CF_FIDEDUPERANGE, fdr);
+	if (err != 0)
+		fprintf(stderr, "ioctl(FIDEDUPERANGE): %s\n", strerror(errno));
+
+	if (fdri->status < 0) {
+		fprintf(stderr, "dedup failed: %s\n", strerror(-fdri->status));
+		err = -1;
+	} else if (fdri->status == CF_FILE_DEDUPE_RANGE_DIFFERS) {
+		fprintf(stderr, "dedup failed: range differs\n");
+		err = -1;
+	}
+
+	return (err);
+}

tests/zfs-tests/cmd/readmmap.c

@@ -44,6 +44,7 @@
 #include <fcntl.h>
 #include <errno.h>
 #include <sys/mman.h>
+#include <sys/types.h>
 #include <time.h>
 
 int

tests/zfs-tests/include/commands.cfg

@@ -182,6 +182,7 @@ export ZFS_FILES='zdb
 export ZFSTEST_FILES='badsend
     btree_test
     chg_usr_exec
+    clonefile
     devname2devid
     dir_rd_update
     draid

tests/zfs-tests/tests/Makefile.am

@@ -90,6 +90,7 @@ nobase_dist_datadir_zfs_tests_tests_DATA += \
 	functional/alloc_class/alloc_class.kshlib \
 	functional/atime/atime.cfg \
 	functional/atime/atime_common.kshlib \
+	functional/block_cloning/block_cloning.kshlib \
 	functional/cache/cache.cfg \
 	functional/cache/cache.kshlib \
 	functional/cachefile/cachefile.cfg \
@@ -437,6 +438,19 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/atime/root_atime_on.ksh \
 	functional/atime/root_relatime_on.ksh \
 	functional/atime/setup.ksh \
+	functional/block_cloning/cleanup.ksh \
+	functional/block_cloning/setup.ksh \
+	functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh \
+	functional/block_cloning/block_cloning_copyfilerange_fallback.ksh \
+	functional/block_cloning/block_cloning_copyfilerange_fallback_same_txg.ksh \
+	functional/block_cloning/block_cloning_copyfilerange.ksh \
+	functional/block_cloning/block_cloning_copyfilerange_partial.ksh \
+	functional/block_cloning/block_cloning_disabled_copyfilerange.ksh \
+	functional/block_cloning/block_cloning_disabled_ficlone.ksh \
+	functional/block_cloning/block_cloning_disabled_ficlonerange.ksh \
+	functional/block_cloning/block_cloning_ficlone.ksh \
+	functional/block_cloning/block_cloning_ficlonerange.ksh \
+	functional/block_cloning/block_cloning_ficlonerange_partial.ksh \
 	functional/bootfs/bootfs_001_pos.ksh \
 	functional/bootfs/bootfs_002_neg.ksh \
 	functional/bootfs/bootfs_003_pos.ksh \

tests/zfs-tests/tests/functional/block_cloning/block_cloning.kshlib

@@ -0,0 +1,54 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+function have_same_content
+{
+	typeset hash1=$(cat $1 | md5sum)
+	typeset hash2=$(cat $2 | md5sum)
+
+	log_must [ "$hash1" = "$hash2" ]
+}
+
+#
+# get_same_blocks dataset1 path/to/file1 dataset2 path/to/file2
+#
+# Returns a space-separated list of the indexes (starting at 0) of the L0
+# blocks that are shared between both files (by first DVA and checksum).
+# Assumes that the two files have the same content, use have_same_content to
+# confirm that.
+#
+function get_same_blocks
+{
+	typeset zdbout=${TMPDIR:-$TEST_BASE_DIR}/zdbout.$$
+	zdb -vvvvv $1 -O $2 | \
+	    awk '/ L0 / { print l++ " " $3 " " $7 }' > $zdbout.a
+	zdb -vvvvv $3 -O $4 | \
+	    awk '/ L0 / { print l++ " " $3 " " $7 }' > $zdbout.b
+	echo $(sort $zdbout.a $zdbout.b | uniq -d | cut -f1 -d' ')
+}
+

tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange.ksh

@@ -0,0 +1,60 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+if [[ $(linux_version) -lt $(linux_version "4.5") ]]; then
+  log_unsupported "copy_file_range not available before Linux 4.5"
+fi
+
+claim="The copy_file_range syscall can clone whole files."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must clonefile -f /$TESTPOOL/file1 /$TESTPOOL/file2 0 0 524288
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "0 1 2 3" ]
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh

@@ -0,0 +1,65 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+if [[ $(linux_version) -lt $(linux_version "5.3") ]]; then
+  log_unsupported "copy_file_range can't copy cross-filesystem before Linux 5.3"
+fi
+
+claim="The copy_file_range syscall can clone across datasets."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must zfs create $TESTPOOL/$TESTFS1
+log_must zfs create $TESTPOOL/$TESTFS2
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS1/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must \
+    clonefile -f /$TESTPOOL/$TESTFS1/file1 /$TESTPOOL/$TESTFS2/file2 0 0 524288
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/$TESTFS1/file1 /$TESTPOOL/$TESTFS2/file2
+
+typeset blocks=$(get_same_blocks \
+  $TESTPOOL/$TESTFS1 file1 $TESTPOOL/$TESTFS2 file2)
+log_must [ "$blocks" = "0 1 2 3" ]
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_fallback.ksh

@@ -0,0 +1,86 @@
+#!/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) 2023, Klara Inc.
+# Copyright (c) 2023, Rob Norris <robn@despairlabs.com>
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+if [[ $(linux_version) -lt $(linux_version "4.5") ]]; then
+  log_unsupported "copy_file_range not available before Linux 4.5"
+fi
+
+claim="copy_file_range will fall back to copy when cloning not possible."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+
+log_note "Copying entire file with copy_file_range"
+
+log_must clonefile -f /$TESTPOOL/file /$TESTPOOL/clone 0 0 524288
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file /$TESTPOOL/clone
+
+typeset blocks=$(get_same_blocks $TESTPOOL file $TESTPOOL clone)
+log_must [ "$blocks" = "0 1 2 3" ]
+
+
+log_note "Copying within a block with copy_file_range"
+
+log_must clonefile -f /$TESTPOOL/file /$TESTPOOL/clone 32768 32768 65536
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file /$TESTPOOL/clone
+
+typeset blocks=$(get_same_blocks $TESTPOOL file $TESTPOOL clone)
+log_must [ "$blocks" = "1 2 3" ]
+
+
+log_note "Copying across a block with copy_file_range"
+
+log_must clonefile -f /$TESTPOOL/file /$TESTPOOL/clone 327680 327680 131072
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file /$TESTPOOL/clone
+
+typeset blocks=$(get_same_blocks $TESTPOOL file $TESTPOOL clone)
+log_must [ "$blocks" = "1" ]
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_fallback_same_txg.ksh

@@ -0,0 +1,66 @@
+#!/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) 2023, Klara Inc.
+# Copyright (c) 2023, Rob Norris <robn@despairlabs.com>
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+if [[ $(linux_version) -lt $(linux_version "4.5") ]]; then
+  log_unsupported "copy_file_range not available before Linux 4.5"
+fi
+
+claim="copy_file_range will fall back to copy when cloning on same txg"
+
+log_assert $claim
+
+typeset timeout=$(get_tunable TXG_TIMEOUT)
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+	set_tunable64 TXG_TIMEOUT $timeout
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must set_tunable64 TXG_TIMEOUT 5000
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file bs=128K count=4
+log_must clonefile -f /$TESTPOOL/file /$TESTPOOL/clone 0 0 524288
+
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file /$TESTPOOL/clone
+
+typeset blocks=$(get_same_blocks $TESTPOOL file $TESTPOOL clone)
+log_must [ "$blocks" = "" ]
+
+log_pass $claim
+

tests/zfs-tests/tests/functional/block_cloning/block_cloning_copyfilerange_partial.ksh

@@ -0,0 +1,68 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+if [[ $(linux_version) -lt $(linux_version "4.5") ]]; then
+  log_unsupported "copy_file_range not available before Linux 4.5"
+fi
+
+claim="The copy_file_range syscall can clone parts of a file."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must dd if=/$TESTPOOL/file1 of=/$TESTPOOL/file2 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "" ]
+
+log_must clonefile -f /$TESTPOOL/file1 /$TESTPOOL/file2 131072 131072 262144
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "1 2" ]
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_disabled_copyfilerange.ksh

@@ -0,0 +1,60 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+if [[ $(linux_version) -lt $(linux_version "4.5") ]]; then
+  log_unsupported "copy_file_range not available before Linux 4.5"
+fi
+
+claim="The copy_file_range syscall copies files when block cloning is disabled."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=disabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must clonefile -f /$TESTPOOL/file1 /$TESTPOOL/file2 0 0 524288
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "" ]
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_disabled_ficlone.ksh

@@ -0,0 +1,50 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+claim="The FICLONE ioctl fails when block cloning is disabled."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=disabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_mustnot clonefile -c /$TESTPOOL/file1 /$TESTPOOL/file2
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_disabled_ficlonerange.ksh

@@ -0,0 +1,50 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+claim="The FICLONERANGE ioctl fails when block cloning is disabled."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=disabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_mustnot clonefile -r /$TESTPOOL/file1 /$TESTPOOL/file2 0 0 524288
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_ficlone.ksh

@@ -0,0 +1,56 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+claim="The FICLONE ioctl can clone files."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must clonefile -c /$TESTPOOL/file1 /$TESTPOOL/file2
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "0 1 2 3" ]
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_ficlonerange.ksh

@@ -0,0 +1,56 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+claim="The FICLONERANGE ioctl can clone whole files."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must clonefile -r /$TESTPOOL/file1 /$TESTPOOL/file2 0 0 524288
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "0 1 2 3" ]
+
+log_pass $claim

tests/zfs-tests/tests/functional/block_cloning/block_cloning_ficlonerange_partial.ksh

@@ -0,0 +1,64 @@
+#!/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) 2023, Klara Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+verify_runnable "global"
+
+claim="The FICLONERANGE ioctl can clone parts of a file."
+
+log_assert $claim
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+}
+
+log_onexit cleanup
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $DISKS
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/file1 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must dd if=/$TESTPOOL/file1 of=/$TESTPOOL/file2 bs=128K count=4
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "" ]
+
+log_must clonefile -r /$TESTPOOL/file1 /$TESTPOOL/file2 131072 131072 262144
+log_must sync_pool $TESTPOOL
+
+log_must have_same_content /$TESTPOOL/file1 /$TESTPOOL/file2
+
+typeset blocks=$(get_same_blocks $TESTPOOL file1 $TESTPOOL file2)
+log_must [ "$blocks" = "1 2" ]
+
+log_pass $claim