Tag 2.2.0-rc2

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

META

@@ -2,7 +2,7 @@ Meta:          1
 Name:          zfs
 Branch:        1.0
 Version:       2.2.0
-Release:       rc1
+Release:       rc2
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS

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);

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
 ])

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/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/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/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/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/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

@@ -420,6 +420,7 @@ 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/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;

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/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

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

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);
 }
 

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 (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);
 }
 
 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,
+			    offset, sizeof (uint64_t), &subobj,
-			    DMU_READ_PREFETCH);
+			    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(subbpo, bpo->bpo_os, subobj);
-			err = bpobj_open(sublist, bpo->bpo_os,
+			if (err) {
-			    obj_from_sublist);
+				kmem_free(subbpo, sizeof (bpobj_t));
-			if (err)
 				break;
+			}
 
-			list_insert_head(&stack, bpi_alloc(sublist, bpi, i));
+			if (subbpo->bpo_havesubobj &&
-			mutex_enter(&sublist->bpo_lock);
+			    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/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"		},
@@ -552,8 +556,7 @@ 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 &&
+	if ((flags & DMU_READ_NO_PREFETCH) == 0) {
-	    length <= zfetch_array_rd_sz) {
 		/*
 		 * Prepare the zfetch before initiating the demand reads, so
 		 * that if multiple threads block on same indirect block, we

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
+	 * The dmu does not currently support decreasing nlevels or changing
-	 * or changing the number of dnode slots on an object. For
+	 * indirect block size if there is already one, same as changing the
-	 * non-raw sends, this does not matter and the new object
+	 * number of of dnode slots on an object.  For non-raw sends this
-	 * can just use the previous one's nlevels. For raw sends,
+	 * does not matter and the new object can just use the previous one's
-	 * however, the structure of the received dnode (including
+	 * parameters.  For raw sends, however, the structure of the received
-	 * nlevels and dnode slots) must match that of the send
+	 * dnode (including indirects and dnode slots) must match that of the
-	 * side. Therefore, instead of using dmu_object_reclaim(),
+	 * send side.  Therefore, instead of using dmu_object_reclaim(), we
-	 * we must free the object completely and call
+	 * must free the object completely and call dmu_object_claim_dnsize()
-	 * dmu_object_claim_dnsize() instead.
+	 * 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,
-		dsl_deadlist_insert_bpobj(dl, za->za_first_integer, mintxg, tx);
+		    zfs_strtonum(za->za_name, NULL), tx);
-		VERIFY0(zap_remove_int(dl->dl_os, obj, mintxg, 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,
+		(void) arc_read(scn->scn_zio_root, spa, bp,
-		    &spic->spic_bp, dsl_scan_prefetch_cb, spic->spic_spc,
+		    dsl_scan_prefetch_cb, spic->spic_spc, ZIO_PRIORITY_SCRUB,
-		    ZIO_PRIORITY_SCRUB, zio_flags, &flags, &spic->spic_zb);
+		    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
@@ -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));
@@ -6266,7 +6273,10 @@ 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.
@@ -1393,6 +1399,36 @@ 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 +1481,7 @@ 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)
+			vdev_spa_set_alloc(spa, min_alloc);
-				spa->spa_min_alloc = min_alloc;
 		}
 	}
 }
@@ -2144,9 +2179,9 @@ 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 +2242,7 @@ 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)
+		vdev_spa_set_alloc(spa, min_alloc);
-			spa->spa_min_alloc = min_alloc;
 	}
 
 	/*
@@ -5688,6 +5722,7 @@ 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 +5733,28 @@ 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 +5823,12 @@ 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)) {
+		if (value & (1ull << i))
-			hist[63 - 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,
+			update_bad_bits(set, &eip->zei_range_sets[range]);
-			    &eip->zei_range_sets[range]);
+			update_bad_bits(cleared, &eip->zei_range_clears[range]);
-			update_histogram(cleared, eip->zei_histogram_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/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
@@ -1802,9 +1815,8 @@ 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)zio_roundup_alloc_size(spa,
-			size_t rounded = (size_t)roundup(psize,
+			    psize);
-			    spa->spa_min_alloc);
 			if (rounded >= lsize) {
 				compress = ZIO_COMPRESS_OFF;
 				zio_buf_free(cbuf, lsize);
@@ -1847,8 +1859,8 @@ 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,
+		size_t rounded = MIN((size_t)zio_roundup_alloc_size(spa, psize),
-		    spa->spa_min_alloc), lsize);
+		    lsize);
 
 		if (rounded != psize) {
 			abd_t *cdata = abd_alloc_linear(rounded, B_TRUE);

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;

tests/zfs-tests/tests/functional/cli_root/zpool_attach/attach-o_ashift.ksh

@@ -35,7 +35,7 @@
 #
 # STRATEGY:
 #	1. Create various pools with different ashift values.
-#	2. Verify 'attach -o ashift=<n>' works only with allowed values.
+#	2. Verify 'attach' works.
 #
 
 verify_runnable "global"
@@ -66,26 +66,14 @@ log_must set_tunable32 VDEV_FILE_PHYSICAL_ASHIFT 16
 typeset ashifts=("9" "10" "11" "12" "13" "14" "15" "16")
 for ashift in ${ashifts[@]}
 do
-	for cmdval in ${ashifts[@]}
+	log_must zpool create -o ashift=$ashift $TESTPOOL1 $disk1
-	do
+	log_must verify_ashift $disk1 $ashift
-		log_must zpool create -o ashift=$ashift $TESTPOOL1 $disk1
+	log_must zpool attach $TESTPOOL1 $disk1 $disk2
-		log_must verify_ashift $disk1 $ashift
+	log_must verify_ashift $disk2 $ashift
-
+	# clean things for the next run
-		# ashift_of(attached_disk) <= ashift_of(existing_vdev)
+	log_must zpool destroy $TESTPOOL1
-		if [[ $cmdval -le $ashift ]]
+	log_must zpool labelclear $disk1
-		then
+	log_must zpool labelclear $disk2
-			log_must zpool attach -o ashift=$cmdval $TESTPOOL1 \
-			    $disk1 $disk2
-			log_must verify_ashift $disk2 $ashift
-		else
-			log_mustnot zpool attach -o ashift=$cmdval $TESTPOOL1 \
-			    $disk1 $disk2
-		fi
-		# clean things for the next run
-		log_must zpool destroy $TESTPOOL1
-		log_must zpool labelclear $disk1
-		log_must zpool labelclear $disk2
-	done
 done
 
 typeset badvals=("off" "on" "1" "8" "17" "1b" "ff" "-")

tests/zfs-tests/tests/functional/cli_root/zpool_replace/replace-o_ashift.ksh

@@ -35,7 +35,7 @@
 #
 # STRATEGY:
 #	1. Create various pools with different ashift values.
-#	2. Verify 'replace -o ashift=<n>' works only with allowed values.
+#	2. Verify 'replace' works.
 #
 
 verify_runnable "global"
@@ -66,26 +66,16 @@ log_must set_tunable32 VDEV_FILE_PHYSICAL_ASHIFT 16
 typeset ashifts=("9" "10" "11" "12" "13" "14" "15" "16")
 for ashift in ${ashifts[@]}
 do
-	for cmdval in ${ashifts[@]}
+	log_must zpool create -o ashift=$ashift $TESTPOOL1 $disk1
-	do
+	log_must verify_ashift $disk1 $ashift
-		log_must zpool create -o ashift=$ashift $TESTPOOL1 $disk1
+	# ashift_of(replacing_disk) <= ashift_of(existing_vdev)
-		log_must verify_ashift $disk1 $ashift
+	log_must zpool replace $TESTPOOL1 $disk1 $disk2
-		# ashift_of(replacing_disk) <= ashift_of(existing_vdev)
+	log_must verify_ashift $disk2 $ashift
-		if [[ $cmdval -le $ashift ]]
+	wait_replacing $TESTPOOL1
-		then
+	# clean things for the next run
-			log_must zpool replace -o ashift=$cmdval $TESTPOOL1 \
+	log_must zpool destroy $TESTPOOL1
-			    $disk1 $disk2
+	log_must zpool labelclear $disk1
-			log_must verify_ashift $disk2 $ashift
+	log_must zpool labelclear $disk2
-			wait_replacing $TESTPOOL1
-		else
-			log_mustnot zpool replace -o ashift=$cmdval $TESTPOOL1 \
-			    $disk1 $disk2
-		fi
-		# clean things for the next run
-		log_must zpool destroy $TESTPOOL1
-		log_must zpool labelclear $disk1
-		log_must zpool labelclear $disk2
-	done
 done
 
 typeset badvals=("off" "on" "1" "8" "17" "1b" "ff" "-")

tests/zfs-tests/tests/functional/cli_root/zpool_replace/replace_prop_ashift.ksh

@@ -34,10 +34,8 @@
 #
 # STRATEGY:
 #	1. Create a pool with default values.
-#	2. Verify 'zpool replace' uses the ashift pool property value when
+#	2. Override the pool ashift property.
-#	   replacing an existing device.
+#	3. Verify 'zpool replace' works.
-#	3. Verify the default ashift value can still be overridden by manually
-#	   specifying '-o ashift=<n>' from the command line.
 #
 
 verify_runnable "global"
@@ -72,21 +70,9 @@ do
 	do
 		log_must zpool create -o ashift=$ashift $TESTPOOL1 $disk1
 		log_must zpool set ashift=$pprop $TESTPOOL1
-		# ashift_of(replacing_disk) <= ashift_of(existing_vdev)
+		log_must zpool replace $TESTPOOL1 $disk1 $disk2
-		if [[ $pprop -le $ashift ]]
+		wait_replacing $TESTPOOL1
-		then
+		log_must verify_ashift $disk2 $ashift
-			log_must zpool replace $TESTPOOL1 $disk1 $disk2
-			wait_replacing $TESTPOOL1
-			log_must verify_ashift $disk2 $ashift
-		else
-			# cannot replace if pool prop ashift > vdev ashift
-			log_mustnot zpool replace $TESTPOOL1 $disk1 $disk2
-			# verify we can override the pool prop value manually
-			log_must zpool replace -o ashift=$ashift $TESTPOOL1 \
-			    $disk1 $disk2
-			wait_replacing $TESTPOOL1
-			log_must verify_ashift $disk2 $ashift
-		fi
 		# clean things for the next run
 		log_must zpool destroy $TESTPOOL1
 		log_must zpool labelclear $disk1