Tag zfs-2.1.6

META file and changelog updated.

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

META

@@ -1,10 +1,10 @@
 Meta:          1
 Name:          zfs
 Branch:        1.0
-Version:       2.1.4
+Version:       2.1.6
 Release:       1
 Release-Tags:  relext
 License:       CDDL
 Author:        OpenZFS
-Linux-Maximum: 5.17
+Linux-Maximum: 5.19
 Linux-Minimum: 3.10

Makefile.am

@@ -103,7 +103,7 @@ endif
 endif
 
 PHONY += codecheck
-codecheck: cstyle shellcheck checkbashisms flake8 mancheck testscheck vcscheck
+codecheck: cstyle shellcheck checkbashisms flake8 mancheck testscheck vcscheck zstdcheck
 
 PHONY += checkstyle
 checkstyle: codecheck commitcheck
@@ -114,14 +114,20 @@ commitcheck:
 		${top_srcdir}/scripts/commitcheck.sh; \
 	fi
 
+if HAVE_PARALLEL
+cstyle_line = -print0 | parallel -X0 ${top_srcdir}/scripts/cstyle.pl -cpP {}
+else
+cstyle_line = -exec ${top_srcdir}/scripts/cstyle.pl -cpP {} +
+endif
 PHONY += cstyle
 cstyle:
 	@find ${top_srcdir} -name build -prune \
 		-o -type f -name '*.[hc]' \
 		! -name 'zfs_config.*' ! -name '*.mod.c' \
 		! -name 'opt_global.h' ! -name '*_if*.h' \
+		! -name 'zstd_compat_wrapper.h' \
 		! -path './module/zstd/lib/*' \
-		-exec ${top_srcdir}/scripts/cstyle.pl -cpP {} \+
+		$(cstyle_line)
 
 filter_executable = -exec test -x '{}' \; -print
 
@@ -173,6 +179,10 @@ vcscheck:
 		awk '{c++; print} END {if(c>0) exit 1}' ; \
 	fi
 
+PHONY += zstdcheck
+zstdcheck:
+	@$(MAKE) -C module/zstd checksymbols
+
 PHONY += lint
 lint: cppcheck paxcheck
 

cmd/arcstat/arcstat.in

@@ -271,7 +271,7 @@ def print_values():
     if pretty_print:
         fmt = lambda col: prettynum(cols[col][0], cols[col][1], v[col])
     else:
-        fmt = lambda col: v[col]
+        fmt = lambda col: str(v[col])
 
     sys.stdout.write(sep.join(fmt(col) for col in hdr))
     sys.stdout.write("\n")

cmd/zdb/zdb.c

@@ -110,8 +110,9 @@ extern int zfs_recover;
 extern unsigned long zfs_arc_meta_min, zfs_arc_meta_limit;
 extern int zfs_vdev_async_read_max_active;
 extern boolean_t spa_load_verify_dryrun;
+extern boolean_t spa_mode_readable_spacemaps;
 extern int zfs_reconstruct_indirect_combinations_max;
-extern int zfs_btree_verify_intensity;
+extern uint_t zfs_btree_verify_intensity;
 
 static const char cmdname[] = "zdb";
 uint8_t dump_opt[256];
@@ -3124,13 +3125,18 @@ dump_znode_symlink(sa_handle_t *hdl)
 {
 	int sa_symlink_size = 0;
 	char linktarget[MAXPATHLEN];
-	linktarget[0] = '\0';
 	int error;
 
 	error = sa_size(hdl, sa_attr_table[ZPL_SYMLINK], &sa_symlink_size);
 	if (error || sa_symlink_size == 0) {
 		return;
 	}
+	if (sa_symlink_size >= sizeof (linktarget)) {
+		(void) printf("symlink size %d is too large\n",
+		    sa_symlink_size);
+		return;
+	}
+	linktarget[sa_symlink_size] = '\0';
 	if (sa_lookup(hdl, sa_attr_table[ZPL_SYMLINK],
 	    &linktarget, sa_symlink_size) == 0)
 		(void) printf("\ttarget	%s\n", linktarget);
@@ -8266,6 +8272,23 @@ zdb_embedded_block(char *thing)
 	free(buf);
 }
 
+/* check for valid hex or decimal numeric string */
+static boolean_t
+zdb_numeric(char *str)
+{
+	int i = 0;
+
+	if (strlen(str) == 0)
+		return (B_FALSE);
+	if (strncmp(str, "0x", 2) == 0 || strncmp(str, "0X", 2) == 0)
+		i = 2;
+	for (; i < strlen(str); i++) {
+		if (!isxdigit(str[i]))
+			return (B_FALSE);
+	}
+	return (B_TRUE);
+}
+
 int
 main(int argc, char **argv)
 {
@@ -8311,7 +8334,7 @@ main(int argc, char **argv)
 	zfs_btree_verify_intensity = 3;
 
 	while ((c = getopt(argc, argv,
-	    "AbcCdDeEFGhiI:klLmMo:Op:PqrRsSt:uU:vVx:XYyZ")) != -1) {
+	    "AbcCdDeEFGhiI:klLmMNo:Op:PqrRsSt:uU:vVx:XYyZ")) != -1) {
 		switch (c) {
 		case 'b':
 		case 'c':
@@ -8325,6 +8348,7 @@ main(int argc, char **argv)
 		case 'l':
 		case 'm':
 		case 'M':
+		case 'N':
 		case 'O':
 		case 'r':
 		case 'R':
@@ -8416,31 +8440,6 @@ main(int argc, char **argv)
 		(void) fprintf(stderr, "-p option requires use of -e\n");
 		usage();
 	}
-	if (dump_opt['d'] || dump_opt['r']) {
-		/* <pool>[/<dataset | objset id> is accepted */
-		if (argv[2] && (objset_str = strchr(argv[2], '/')) != NULL &&
-		    objset_str++ != NULL) {
-			char *endptr;
-			errno = 0;
-			objset_id = strtoull(objset_str, &endptr, 0);
-			/* dataset 0 is the same as opening the pool */
-			if (errno == 0 && endptr != objset_str &&
-			    objset_id != 0) {
-				target_is_spa = B_FALSE;
-				dataset_lookup = B_TRUE;
-			} else if (objset_id != 0) {
-				printf("failed to open objset %s "
-				    "%llu %s", objset_str,
-				    (u_longlong_t)objset_id,
-				    strerror(errno));
-				exit(1);
-			}
-			/* normal dataset name not an objset ID */
-			if (endptr == objset_str) {
-				objset_id = -1;
-			}
-		}
-	}
 
 #if defined(_LP64)
 	/*
@@ -8469,13 +8468,18 @@ main(int argc, char **argv)
 	 */
 	spa_load_verify_dryrun = B_TRUE;
 
+	/*
+	 * ZDB should have ability to read spacemaps.
+	 */
+	spa_mode_readable_spacemaps = B_TRUE;
+
 	kernel_init(SPA_MODE_READ);
 
 	if (dump_all)
 		verbose = MAX(verbose, 1);
 
 	for (c = 0; c < 256; c++) {
-		if (dump_all && strchr("AeEFklLOPrRSXy", c) == NULL)
+		if (dump_all && strchr("AeEFklLNOPrRSXy", c) == NULL)
 			dump_opt[c] = 1;
 		if (dump_opt[c])
 			dump_opt[c] += verbose;
@@ -8514,6 +8518,7 @@ main(int argc, char **argv)
 		return (dump_path(argv[0], argv[1], NULL));
 	}
 	if (dump_opt['r']) {
+		target_is_spa = B_FALSE;
 		if (argc != 3)
 			usage();
 		dump_opt['v'] = verbose;
@@ -8524,6 +8529,10 @@ main(int argc, char **argv)
 		rewind = ZPOOL_DO_REWIND |
 		    (dump_opt['X'] ? ZPOOL_EXTREME_REWIND : 0);
 
+	/* -N implies -d */
+	if (dump_opt['N'] && dump_opt['d'] == 0)
+		dump_opt['d'] = dump_opt['N'];
+
 	if (nvlist_alloc(&policy, NV_UNIQUE_NAME_TYPE, 0) != 0 ||
 	    nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, max_txg) != 0 ||
 	    nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY, rewind) != 0)
@@ -8542,6 +8551,34 @@ main(int argc, char **argv)
 		targetlen = strlen(target);
 		if (targetlen && target[targetlen - 1] == '/')
 			target[targetlen - 1] = '\0';
+		/*
+		 * See if an objset ID was supplied (-d <pool>/<objset ID>).
+		 * To disambiguate tank/100, consider the 100 as objsetID
+		 * if -N was given, otherwise 100 is an objsetID iff
+		 * tank/100 as a named dataset fails on lookup.
+		 */
+		objset_str = strchr(target, '/');
+		if (objset_str && strlen(objset_str) > 1 &&
+		    zdb_numeric(objset_str + 1)) {
+			char *endptr;
+			errno = 0;
+			objset_str++;
+			objset_id = strtoull(objset_str, &endptr, 0);
+			/* dataset 0 is the same as opening the pool */
+			if (errno == 0 && endptr != objset_str &&
+			    objset_id != 0) {
+				if (dump_opt['N'])
+					dataset_lookup = B_TRUE;
+			}
+			/* normal dataset name not an objset ID */
+			if (endptr == objset_str) {
+				objset_id = -1;
+			}
+		} else if (objset_str && !zdb_numeric(objset_str + 1) &&
+		    dump_opt['N']) {
+			printf("Supply a numeric objset ID with -N\n");
+			exit(1);
+		}
 	} else {
 		target_pool = target;
 	}
@@ -8659,13 +8696,27 @@ main(int argc, char **argv)
 			}
 			return (error);
 		} else {
+			target_pool = strdup(target);
+			if (strpbrk(target, "/@") != NULL)
+				*strpbrk(target_pool, "/@") = '\0';
+
 			zdb_set_skip_mmp(target);
+			/*
+			 * If -N was supplied, the user has indicated that
+			 * zdb -d <pool>/<objsetID> is in effect.  Otherwise
+			 * we first assume that the dataset string is the
+			 * dataset name.  If dmu_objset_hold fails with the
+			 * dataset string, and we have an objset_id, retry the
+			 * lookup with the objsetID.
+			 */
+			boolean_t retry = B_TRUE;
+retry_lookup:
 			if (dataset_lookup == B_TRUE) {
 				/*
 				 * Use the supplied id to get the name
 				 * for open_objset.
 				 */
-				error = spa_open(target, &spa, FTAG);
+				error = spa_open(target_pool, &spa, FTAG);
 				if (error == 0) {
 					error = name_from_objset_id(spa,
 					    objset_id, dsname);
@@ -8674,10 +8725,23 @@ main(int argc, char **argv)
 						target = dsname;
 				}
 			}
-			if (error == 0)
+			if (error == 0) {
+				if (objset_id > 0 && retry) {
+					int err = dmu_objset_hold(target, FTAG,
+					    &os);
+					if (err) {
+						dataset_lookup = B_TRUE;
+						retry = B_FALSE;
+						goto retry_lookup;
+					} else {
+						dmu_objset_rele(os, FTAG);
+					}
+				}
 				error = open_objset(target, FTAG, &os);
+			}
 			if (error == 0)
 				spa = dmu_objset_spa(os);
+			free(target_pool);
 		}
 	}
 	nvlist_free(policy);

cmd/zed/agents/zfs_diagnosis.c

@@ -35,6 +35,7 @@
 #include <sys/fs/zfs.h>
 #include <sys/fm/protocol.h>
 #include <sys/fm/fs/zfs.h>
+#include <sys/zio.h>
 
 #include "zfs_agents.h"
 #include "fmd_api.h"
@@ -773,6 +774,8 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 	    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_PROBE_FAILURE))) {
 		char *failmode = NULL;
 		boolean_t checkremove = B_FALSE;
+		uint32_t pri = 0;
+		int32_t flags = 0;
 
 		/*
 		 * If this is a checksum or I/O error, then toss it into the
@@ -795,6 +798,23 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class)
 				checkremove = B_TRUE;
 		} else if (fmd_nvl_class_match(hdl, nvl,
 		    ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_CHECKSUM))) {
+			/*
+			 * We ignore ereports for checksum errors generated by
+			 * scrub/resilver I/O to avoid potentially further
+			 * degrading the pool while it's being repaired.
+			 */
+			if (((nvlist_lookup_uint32(nvl,
+			    FM_EREPORT_PAYLOAD_ZFS_ZIO_PRIORITY, &pri) == 0) &&
+			    (pri == ZIO_PRIORITY_SCRUB ||
+			    pri == ZIO_PRIORITY_REBUILD)) ||
+			    ((nvlist_lookup_int32(nvl,
+			    FM_EREPORT_PAYLOAD_ZFS_ZIO_FLAGS, &flags) == 0) &&
+			    (flags & (ZIO_FLAG_SCRUB | ZIO_FLAG_RESILVER)))) {
+				fmd_hdl_debug(hdl, "ignoring '%s' for "
+				    "scrub/resilver I/O", class);
+				return;
+			}
+
 			if (zcp->zc_data.zc_serd_checksum[0] == '\0') {
 				zfs_serd_name(zcp->zc_data.zc_serd_checksum,
 				    pool_guid, vdev_guid, "checksum");

cmd/zed/agents/zfs_mod.c

@@ -894,14 +894,90 @@ zfs_deliver_check(nvlist_t *nvl)
 	return (0);
 }
 
+/*
+ * Given a path to a vdev, lookup the vdev's physical size from its
+ * config nvlist.
+ *
+ * Returns the vdev's physical size in bytes on success, 0 on error.
+ */
+static uint64_t
+vdev_size_from_config(zpool_handle_t *zhp, const char *vdev_path)
+{
+	nvlist_t *nvl = NULL;
+	boolean_t avail_spare, l2cache, log;
+	vdev_stat_t *vs = NULL;
+	uint_t c;
+
+	nvl = zpool_find_vdev(zhp, vdev_path, &avail_spare, &l2cache, &log);
+	if (!nvl)
+		return (0);
+
+	verify(nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_VDEV_STATS,
+	    (uint64_t **)&vs, &c) == 0);
+	if (!vs) {
+		zed_log_msg(LOG_INFO, "%s: no nvlist for '%s'", __func__,
+		    vdev_path);
+		return (0);
+	}
+
+	return (vs->vs_pspace);
+}
+
+/*
+ * Given a path to a vdev, lookup if the vdev is a "whole disk" in the
+ * config nvlist.  "whole disk" means that ZFS was passed a whole disk
+ * at pool creation time, which it partitioned up and has full control over.
+ * Thus a partition with wholedisk=1 set tells us that zfs created the
+ * partition at creation time.  A partition without whole disk set would have
+ * been created by externally (like with fdisk) and passed to ZFS.
+ *
+ * Returns the whole disk value (either 0 or 1).
+ */
+static uint64_t
+vdev_whole_disk_from_config(zpool_handle_t *zhp, const char *vdev_path)
+{
+	nvlist_t *nvl = NULL;
+	boolean_t avail_spare, l2cache, log;
+	uint64_t wholedisk;
+
+	nvl = zpool_find_vdev(zhp, vdev_path, &avail_spare, &l2cache, &log);
+	if (!nvl)
+		return (0);
+
+	verify(nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_WHOLE_DISK,
+	    &wholedisk) == 0);
+
+	return (wholedisk);
+}
+
+/*
+ * If the device size grew more than 1% then return true.
+ */
+#define	DEVICE_GREW(oldsize, newsize) \
+		    ((newsize > oldsize) && \
+		    ((newsize / (newsize - oldsize)) <= 100))
+
 static int
 zfsdle_vdev_online(zpool_handle_t *zhp, void *data)
 {
-	char *devname = data;
 	boolean_t avail_spare, l2cache;
+	nvlist_t *udev_nvl = data;
 	nvlist_t *tgt;
 	int error;
 
+	char *tmp_devname, devname[MAXPATHLEN];
+	uint64_t guid;
+
+	if (nvlist_lookup_uint64(udev_nvl, ZFS_EV_VDEV_GUID, &guid) == 0) {
+		sprintf(devname, "%llu", (u_longlong_t)guid);
+	} else if (nvlist_lookup_string(udev_nvl, DEV_PHYS_PATH,
+	    &tmp_devname) == 0) {
+		strlcpy(devname, tmp_devname, MAXPATHLEN);
+		zfs_append_partition(devname, MAXPATHLEN);
+	} else {
+		zed_log_msg(LOG_INFO, "%s: no guid or physpath", __func__);
+	}
+
 	zed_log_msg(LOG_INFO, "zfsdle_vdev_online: searching for '%s' in '%s'",
 	    devname, zpool_get_name(zhp));
 
@@ -953,12 +1029,75 @@ zfsdle_vdev_online(zpool_handle_t *zhp, void *data)
 			vdev_state_t newstate;
 
 			if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL) {
-				error = zpool_vdev_online(zhp, fullpath, 0,
-				    &newstate);
-				zed_log_msg(LOG_INFO, "zfsdle_vdev_online: "
-				    "setting device '%s' to ONLINE state "
-				    "in pool '%s': %d", fullpath,
-				    zpool_get_name(zhp), error);
+				/*
+				 * If this disk size has not changed, then
+				 * there's no need to do an autoexpand.  To
+				 * check we look at the disk's size in its
+				 * config, and compare it to the disk size
+				 * that udev is reporting.
+				 */
+				uint64_t udev_size = 0, conf_size = 0,
+				    wholedisk = 0, udev_parent_size = 0;
+
+				/*
+				 * Get the size of our disk that udev is
+				 * reporting.
+				 */
+				if (nvlist_lookup_uint64(udev_nvl, DEV_SIZE,
+				    &udev_size) != 0) {
+					udev_size = 0;
+				}
+
+				/*
+				 * Get the size of our disk's parent device
+				 * from udev (where sda1's parent is sda).
+				 */
+				if (nvlist_lookup_uint64(udev_nvl,
+				    DEV_PARENT_SIZE, &udev_parent_size) != 0) {
+					udev_parent_size = 0;
+				}
+
+				conf_size = vdev_size_from_config(zhp,
+				    fullpath);
+
+				wholedisk = vdev_whole_disk_from_config(zhp,
+				    fullpath);
+
+				/*
+				 * Only attempt an autoexpand if the vdev size
+				 * changed.  There are two different cases
+				 * to consider.
+				 *
+				 * 1. wholedisk=1
+				 * If you do a 'zpool create' on a whole disk
+				 * (like /dev/sda), then zfs will create
+				 * partitions on the disk (like /dev/sda1).  In
+				 * that case, wholedisk=1 will be set in the
+				 * partition's nvlist config.  So zed will need
+				 * to see if your parent device (/dev/sda)
+				 * expanded in size, and if so, then attempt
+				 * the autoexpand.
+				 *
+				 * 2. wholedisk=0
+				 * If you do a 'zpool create' on an existing
+				 * partition, or a device that doesn't allow
+				 * partitions, then wholedisk=0, and you will
+				 * simply need to check if the device itself
+				 * expanded in size.
+				 */
+				if (DEVICE_GREW(conf_size, udev_size) ||
+				    (wholedisk && DEVICE_GREW(conf_size,
+				    udev_parent_size))) {
+					error = zpool_vdev_online(zhp, fullpath,
+					    0, &newstate);
+
+					zed_log_msg(LOG_INFO,
+					    "%s: autoexpanding '%s' from %llu"
+					    " to %llu bytes in pool '%s': %d",
+					    __func__, fullpath, conf_size,
+					    MAX(udev_size, udev_parent_size),
+					    zpool_get_name(zhp), error);
+				}
 			}
 		}
 		zpool_close(zhp);
@@ -989,7 +1128,7 @@ zfs_deliver_dle(nvlist_t *nvl)
 		zed_log_msg(LOG_INFO, "zfs_deliver_dle: no guid or physpath");
 	}
 
-	if (zpool_iter(g_zfshdl, zfsdle_vdev_online, name) != 1) {
+	if (zpool_iter(g_zfshdl, zfsdle_vdev_online, nvl) != 1) {
 		zed_log_msg(LOG_INFO, "zfs_deliver_dle: device '%s' not "
 		    "found", name);
 		return (1);

cmd/zed/zed.d/zed-functions.sh

@@ -224,6 +224,8 @@ zed_notify()
 # ZED_EMAIL_OPTS.  This undergoes the following keyword substitutions:
 # - @ADDRESS@ is replaced with the space-delimited recipient email address(es)
 # - @SUBJECT@ is replaced with the notification subject
+#   If @SUBJECT@ was omited here, a "Subject: ..." header will be added to notification
+#
 #
 # Arguments
 #   subject: notification subject
@@ -241,7 +243,7 @@ zed_notify()
 #
 zed_notify_email()
 {
-    local subject="$1"
+    local subject="${1:-"ZED notification"}"
     local pathname="${2:-"/dev/null"}"
 
     : "${ZED_EMAIL_PROG:="mail"}"
@@ -262,12 +264,23 @@ zed_notify_email()
         return 1
     fi
 
-    ZED_EMAIL_OPTS="$(echo "${ZED_EMAIL_OPTS}" \
+    # construct cmdline options
+    ZED_EMAIL_OPTS_PARSED="$(echo "${ZED_EMAIL_OPTS}" \
         | sed   -e "s/@ADDRESS@/${ZED_EMAIL_ADDR}/g" \
                 -e "s/@SUBJECT@/${subject}/g")"
 
-    # shellcheck disable=SC2086
-    eval ${ZED_EMAIL_PROG} ${ZED_EMAIL_OPTS} < "${pathname}" >/dev/null 2>&1
+    # pipe message to email prog
+    # shellcheck disable=SC2086,SC2248
+    {
+        # no subject passed as option?
+        if [ "${ZED_EMAIL_OPTS%@SUBJECT@*}" = "${ZED_EMAIL_OPTS}" ] ; then
+            # inject subject header
+            printf "Subject: %s\n" "${subject}"
+        fi
+        # output message
+        cat "${pathname}"
+    } |
+    eval ${ZED_EMAIL_PROG} ${ZED_EMAIL_OPTS_PARSED} >/dev/null 2>&1
     rv=$?
     if [ "${rv}" -ne 0 ]; then
         zed_log_err "${ZED_EMAIL_PROG##*/} exit=${rv}"

cmd/zed/zed.d/zed.rc

@@ -30,6 +30,7 @@ ZED_EMAIL_ADDR="root"
 # The string @SUBJECT@ will be replaced with the notification subject;
 #   this should be protected with quotes to prevent word-splitting.
 # Email will only be sent if ZED_EMAIL_ADDR is defined.
+# If @SUBJECT@ was omited here, a "Subject: ..." header will be added to notification
 #
 #ZED_EMAIL_OPTS="-s '@SUBJECT@' @ADDRESS@"
 

cmd/zed/zed_disk_event.c

@@ -78,6 +78,8 @@ zed_udev_event(const char *class, const char *subclass, nvlist_t *nvl)
 		zed_log_msg(LOG_INFO, "\t%s: %s", DEV_PHYS_PATH, strval);
 	if (nvlist_lookup_uint64(nvl, DEV_SIZE, &numval) == 0)
 		zed_log_msg(LOG_INFO, "\t%s: %llu", DEV_SIZE, numval);
+	if (nvlist_lookup_uint64(nvl, DEV_PARENT_SIZE, &numval) == 0)
+		zed_log_msg(LOG_INFO, "\t%s: %llu", DEV_PARENT_SIZE, numval);
 	if (nvlist_lookup_uint64(nvl, ZFS_EV_POOL_GUID, &numval) == 0)
 		zed_log_msg(LOG_INFO, "\t%s: %llu", ZFS_EV_POOL_GUID, numval);
 	if (nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID, &numval) == 0)
@@ -130,6 +132,20 @@ dev_event_nvlist(struct udev_device *dev)
 
 		numval *= strtoull(value, NULL, 10);
 		(void) nvlist_add_uint64(nvl, DEV_SIZE, numval);
+
+		/*
+		 * If the device has a parent, then get the parent block
+		 * device's size as well.  For example, /dev/sda1's parent
+		 * is /dev/sda.
+		 */
+		struct udev_device *parent_dev = udev_device_get_parent(dev);
+		if ((value = udev_device_get_sysattr_value(parent_dev, "size"))
+		    != NULL) {
+			uint64_t numval = DEV_BSIZE;
+
+			numval *= strtoull(value, NULL, 10);
+			(void) nvlist_add_uint64(nvl, DEV_PARENT_SIZE, numval);
+		}
 	}
 
 	/*
@@ -169,7 +185,7 @@ zed_udev_monitor(void *arg)
 	while (1) {
 		struct udev_device *dev;
 		const char *action, *type, *part, *sectors;
-		const char *bus, *uuid;
+		const char *bus, *uuid, *devpath;
 		const char *class, *subclass;
 		nvlist_t *nvl;
 		boolean_t is_zfs = B_FALSE;
@@ -208,6 +224,12 @@ zed_udev_monitor(void *arg)
 		 * if this is a disk and it is partitioned, then the
 		 * zfs label will reside in a DEVTYPE=partition and
 		 * we can skip passing this event
+		 *
+		 * Special case: Blank disks are sometimes reported with
+		 * an erroneous 'atari' partition, and should not be
+		 * excluded from being used as an autoreplace disk:
+		 *
+		 * https://github.com/openzfs/zfs/issues/13497
 		 */
 		type = udev_device_get_property_value(dev, "DEVTYPE");
 		part = udev_device_get_property_value(dev,
@@ -215,14 +237,23 @@ zed_udev_monitor(void *arg)
 		if (type != NULL && type[0] != '\0' &&
 		    strcmp(type, "disk") == 0 &&
 		    part != NULL && part[0] != '\0') {
-			zed_log_msg(LOG_INFO,
-			    "%s: skip %s since it has a %s partition already",
-			    __func__,
-			    udev_device_get_property_value(dev, "DEVNAME"),
-			    part);
-			/* skip and wait for partition event */
-			udev_device_unref(dev);
-			continue;
+			const char *devname =
+			    udev_device_get_property_value(dev, "DEVNAME");
+
+			if (strcmp(part, "atari") == 0) {
+				zed_log_msg(LOG_INFO,
+				    "%s: %s is reporting an atari partition, "
+				    "but we're going to assume it's a false "
+				    "positive and still use it (issue #13497)",
+				    __func__, devname);
+			} else {
+				zed_log_msg(LOG_INFO,
+				    "%s: skip %s since it has a %s partition "
+				    "already", __func__, devname, part);
+				/* skip and wait for partition event */
+				udev_device_unref(dev);
+				continue;
+			}
 		}
 
 		/*
@@ -248,10 +279,19 @@ zed_udev_monitor(void *arg)
 		 * device id string is required in the message schema
 		 * for matching with vdevs. Preflight here for expected
 		 * udev information.
+		 *
+		 * Special case:
+		 * NVMe devices don't have ID_BUS set (at least on RHEL 7-8),
+		 * but they are valid for autoreplace.  Add a special case for
+		 * them by searching for "/nvme/" in the udev DEVPATH:
+		 *
+		 * DEVPATH=/devices/pci0000:00/0000:00:1e.0/nvme/nvme2/nvme2n1
 		 */
 		bus = udev_device_get_property_value(dev, "ID_BUS");
 		uuid = udev_device_get_property_value(dev, "DM_UUID");
-		if (!is_zfs && (bus == NULL && uuid == NULL)) {
+		devpath = udev_device_get_devpath(dev);
+		if (!is_zfs && (bus == NULL && uuid == NULL &&
+		    strstr(devpath, "/nvme/") == NULL)) {
 			zed_log_msg(LOG_INFO, "zed_udev_monitor: %s no devid "
 			    "source", udev_device_get_devnode(dev));
 			udev_device_unref(dev);

cmd/zfs/zfs_main.c

@@ -2480,7 +2480,7 @@ upgrade_set_callback(zfs_handle_t *zhp, void *data)
 
 	/* upgrade */
 	if (version < cb->cb_version) {
-		char verstr[16];
+		char verstr[24];
 		(void) snprintf(verstr, sizeof (verstr),
 		    "%llu", (u_longlong_t)cb->cb_version);
 		if (cb->cb_lastfs[0] && !same_pool(zhp, cb->cb_lastfs)) {
@@ -6593,7 +6593,7 @@ zfs_do_holds(int argc, char **argv)
 		/*
 		 *  1. collect holds data, set format options
 		 */
-		ret = zfs_for_each(argc, argv, flags, types, NULL, NULL, limit,
+		ret = zfs_for_each(1, argv + i, flags, types, NULL, NULL, limit,
 		    holds_callback, &cb);
 		if (ret != 0)
 			++errors;
@@ -7672,7 +7672,7 @@ zfs_do_diff(int argc, char **argv)
 	int c;
 	struct sigaction sa;
 
-	while ((c = getopt(argc, argv, "FHt")) != -1) {
+	while ((c = getopt(argc, argv, "FHth")) != -1) {
 		switch (c) {
 		case 'F':
 			flags |= ZFS_DIFF_CLASSIFY;
@@ -7683,6 +7683,9 @@ zfs_do_diff(int argc, char **argv)
 		case 't':
 			flags |= ZFS_DIFF_TIMESTAMP;
 			break;
+		case 'h':
+			flags |= ZFS_DIFF_NO_MANGLE;
+			break;
 		default:
 			(void) fprintf(stderr,
 			    gettext("invalid option '%c'\n"), optopt);

cmd/zfs/zfs_project.c

@@ -207,7 +207,6 @@ static int
 zfs_project_handle_dir(const char *name, zfs_project_control_t *zpc,
     list_t *head)
 {
-	char fullname[PATH_MAX];
 	struct dirent *ent;
 	DIR *dir;
 	int ret = 0;
@@ -227,21 +226,28 @@ zfs_project_handle_dir(const char *name, zfs_project_control_t *zpc,
 	zpc->zpc_ignore_noent = B_TRUE;
 	errno = 0;
 	while (!ret && (ent = readdir(dir)) != NULL) {
+		char *fullname;
+
 		/* skip "." and ".." */
 		if (strcmp(ent->d_name, ".") == 0 ||
 		    strcmp(ent->d_name, "..") == 0)
 			continue;
 
-		if (strlen(ent->d_name) + strlen(name) >=
-		    sizeof (fullname) + 1) {
+		if (strlen(ent->d_name) + strlen(name) + 1 >= PATH_MAX) {
 			errno = ENAMETOOLONG;
 			break;
 		}
 
-		sprintf(fullname, "%s/%s", name, ent->d_name);
+		if (asprintf(&fullname, "%s/%s", name, ent->d_name) == -1) {
+			errno = ENOMEM;
+			break;
+		}
+
 		ret = zfs_project_handle_one(fullname, zpc);
 		if (!ret && zpc->zpc_recursive && ent->d_type == DT_DIR)
 			zfs_project_item_alloc(head, fullname);
+
+		free(fullname);
 	}
 
 	if (errno && !ret) {

cmd/zpool/zpool_main.c

@@ -2438,7 +2438,14 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 	(void) nvlist_lookup_uint64_array(root, ZPOOL_CONFIG_SCAN_STATS,
 	    (uint64_t **)&ps, &c);
 
-	if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0) {
+	/*
+	 * If you force fault a drive that's resilvering, its scan stats can
+	 * get frozen in time, giving the false impression that it's
+	 * being resilvered.  That's why we check the state to see if the vdev
+	 * is healthy before reporting "resilvering" or "repairing".
+	 */
+	if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0 &&
+	    vs->vs_state == VDEV_STATE_HEALTHY) {
 		if (vs->vs_scan_processed != 0) {
 			(void) printf(gettext("  (%s)"),
 			    (ps->pss_func == POOL_SCAN_RESILVER) ?
@@ -2450,7 +2457,7 @@ print_status_config(zpool_handle_t *zhp, status_cbdata_t *cb, const char *name,
 
 	/* The top-level vdevs have the rebuild stats */
 	if (vrs != NULL && vrs->vrs_state == VDEV_REBUILD_ACTIVE &&
-	    children == 0) {
+	    children == 0 && vs->vs_state == VDEV_STATE_HEALTHY) {
 		if (vs->vs_rebuild_processed != 0) {
 			(void) printf(gettext("  (resilvering)"));
 		}
@@ -5458,8 +5465,8 @@ get_namewidth_iostat(zpool_handle_t *zhp, void *data)
 	 * get_namewidth() returns the maximum width of any name in that column
 	 * for any pool/vdev/device line that will be output.
 	 */
-	width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_name_flags,
-	    cb->cb_verbose);
+	width = get_namewidth(zhp, cb->cb_namewidth,
+	    cb->cb_name_flags | VDEV_NAME_TYPE_ID, cb->cb_verbose);
 
 	/*
 	 * The width we are calculating is the width of the header and also the
@@ -6035,6 +6042,7 @@ print_one_column(zpool_prop_t prop, uint64_t value, const char *str,
 	size_t width = zprop_width(prop, &fixed, ZFS_TYPE_POOL);
 
 	switch (prop) {
+	case ZPOOL_PROP_SIZE:
 	case ZPOOL_PROP_EXPANDSZ:
 	case ZPOOL_PROP_CHECKPOINT:
 	case ZPOOL_PROP_DEDUPRATIO:
@@ -6130,8 +6138,12 @@ print_list_stats(zpool_handle_t *zhp, const char *name, nvlist_t *nv,
 		 * 'toplevel' boolean value is passed to the print_one_column()
 		 * to indicate that the value is valid.
 		 */
-		print_one_column(ZPOOL_PROP_SIZE, vs->vs_space, NULL, scripted,
-		    toplevel, format);
+		if (vs->vs_pspace)
+			print_one_column(ZPOOL_PROP_SIZE, vs->vs_pspace, NULL,
+			    scripted, B_TRUE, format);
+		else
+			print_one_column(ZPOOL_PROP_SIZE, vs->vs_space, NULL,
+			    scripted, toplevel, format);
 		print_one_column(ZPOOL_PROP_ALLOCATED, vs->vs_alloc, NULL,
 		    scripted, toplevel, format);
 		print_one_column(ZPOOL_PROP_FREE, vs->vs_space - vs->vs_alloc,
@@ -6282,8 +6294,8 @@ get_namewidth_list(zpool_handle_t *zhp, void *data)
 	list_cbdata_t *cb = data;
 	int width;
 
-	width = get_namewidth(zhp, cb->cb_namewidth, cb->cb_name_flags,
-	    cb->cb_verbose);
+	width = get_namewidth(zhp, cb->cb_namewidth,
+	    cb->cb_name_flags | VDEV_NAME_TYPE_ID, cb->cb_verbose);
 
 	if (width < 9)
 		width = 9;

cmd/zvol_wait/zvol_wait

@@ -28,15 +28,17 @@ filter_out_deleted_zvols() {
 list_zvols() {
 	read -r default_volmode < /sys/module/zfs/parameters/zvol_volmode
 	zfs list -t volume -H -o \
-	    name,volmode,receive_resume_token,redact_snaps |
-	    while IFS="	" read -r name volmode token redacted; do # IFS=\t here!
+	    name,volmode,receive_resume_token,redact_snaps,keystatus |
+	    while IFS="	" read -r name volmode token redacted keystatus; do # IFS=\t here!
 
-		# /dev links are not created for zvols with volmode = "none"
-		# or for redacted zvols.
+		# /dev links are not created for zvols with volmode = "none",
+		# redacted zvols, or encrypted zvols for which the key has not
+		# been loaded.
 		[ "$volmode" = "none" ] && continue
 		[ "$volmode" = "default" ] && [ "$default_volmode" = "3" ] &&
 		    continue
 		[ "$redacted" = "-" ] || continue
+		[ "$keystatus" = "unavailable" ] && continue
 
 		# We also ignore partially received zvols if it is
 		# not an incremental receive, as those won't even have a block

config/always-compiler-options.m4

@@ -88,7 +88,7 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_TRUNCATION], [
 ])
 
 dnl #
-dnl # Check if gcc supports -Wno-format-truncation option.
+dnl # Check if gcc supports -Wno-format-zero-length option.
 dnl #
 AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_ZERO_LENGTH], [
 	AC_MSG_CHECKING([whether $CC supports -Wno-format-zero-length])
@@ -108,57 +108,30 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_ZERO_LENGTH], [
 	AC_SUBST([NO_FORMAT_ZERO_LENGTH])
 ])
 
-
 dnl #
-dnl # Check if gcc supports -Wno-bool-compare option.
+dnl # Check if gcc supports -Wno-clobbered option.
 dnl #
-dnl # We actually invoke gcc with the -Wbool-compare option
+dnl # We actually invoke gcc with the -Wclobbered option
 dnl # and infer the 'no-' version does or doesn't exist based upon
 dnl # the results.  This is required because when checking any of
 dnl # no- prefixed options gcc always returns success.
 dnl #
-AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_BOOL_COMPARE], [
-	AC_MSG_CHECKING([whether $CC supports -Wno-bool-compare])
+AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_CLOBBERED], [
+	AC_MSG_CHECKING([whether $CC supports -Wno-clobbered])
 
 	saved_flags="$CFLAGS"
-	CFLAGS="$CFLAGS -Werror -Wbool-compare"
+	CFLAGS="$CFLAGS -Werror -Wclobbered"
 
 	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])], [
-		NO_BOOL_COMPARE=-Wno-bool-compare
+		NO_CLOBBERED=-Wno-clobbered
 		AC_MSG_RESULT([yes])
 	], [
-		NO_BOOL_COMPARE=
+		NO_CLOBBERED=
 		AC_MSG_RESULT([no])
 	])
 
 	CFLAGS="$saved_flags"
-	AC_SUBST([NO_BOOL_COMPARE])
-])
-
-dnl #
-dnl # Check if gcc supports -Wno-unused-but-set-variable option.
-dnl #
-dnl # We actually invoke gcc with the -Wunused-but-set-variable option
-dnl # and infer the 'no-' version does or doesn't exist based upon
-dnl # the results.  This is required because when checking any of
-dnl # no- prefixed options gcc always returns success.
-dnl #
-AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_NO_UNUSED_BUT_SET_VARIABLE], [
-	AC_MSG_CHECKING([whether $CC supports -Wno-unused-but-set-variable])
-
-	saved_flags="$CFLAGS"
-	CFLAGS="$CFLAGS -Werror -Wunused-but-set-variable"
-
-	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])], [
-		NO_UNUSED_BUT_SET_VARIABLE=-Wno-unused-but-set-variable
-		AC_MSG_RESULT([yes])
-	], [
-		NO_UNUSED_BUT_SET_VARIABLE=
-		AC_MSG_RESULT([no])
-	])
-
-	CFLAGS="$saved_flags"
-	AC_SUBST([NO_UNUSED_BUT_SET_VARIABLE])
+	AC_SUBST([NO_CLOBBERED])
 ])
 
 dnl #
@@ -184,6 +157,29 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_IMPLICIT_FALLTHROUGH], [
 	AC_SUBST([IMPLICIT_FALLTHROUGH])
 ])
 
+dnl #
+dnl # Check if cc supports -Winfinite-recursion option.
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_CC_INFINITE_RECURSION], [
+	AC_MSG_CHECKING([whether $CC supports -Winfinite-recursion])
+
+	saved_flags="$CFLAGS"
+	CFLAGS="$CFLAGS -Werror -Winfinite-recursion"
+
+	AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], [])], [
+		INFINITE_RECURSION=-Winfinite-recursion
+		AC_DEFINE([HAVE_INFINITE_RECURSION], 1,
+			[Define if compiler supports -Winfinite-recursion])
+		AC_MSG_RESULT([yes])
+	], [
+		INFINITE_RECURSION=
+		AC_MSG_RESULT([no])
+	])
+
+	CFLAGS="$saved_flags"
+	AC_SUBST([INFINITE_RECURSION])
+])
+
 dnl #
 dnl # Check if gcc supports -fno-omit-frame-pointer option.
 dnl #

config/always-parallel.m4

@@ -0,0 +1,8 @@
+dnl #
+dnl # Check if GNU parallel is available.
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_PARALLEL], [
+	AC_CHECK_PROG([PARALLEL], [parallel], [yes])
+
+	AM_CONDITIONAL([HAVE_PARALLEL], [test "x$PARALLEL" = "xyes"])
+])

config/ax_python_devel.m4

@@ -224,7 +224,7 @@ EOD`
 				ac_python_version=$PYTHON_VERSION
 			else
 				ac_python_version=`$PYTHON -c "import sys; \
-					print (sys.version[[:3]])"`
+					print ('.'.join(sys.version.split('.')[[:2]]))"`
 			fi
 		fi
 

config/kernel-add-disk.m4

@@ -3,16 +3,15 @@ dnl # 5.16 API change
 dnl # add_disk grew a must-check return code
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_ADD_DISK], [
-
 	ZFS_LINUX_TEST_SRC([add_disk_ret], [
-		#include <linux/genhd.h>
+		#include <linux/blkdev.h>
 	], [
 		struct gendisk *disk = NULL;
 		int err = add_disk(disk);
 		err = err;
 	])
-
 ])
+
 AC_DEFUN([ZFS_AC_KERNEL_ADD_DISK], [
 	AC_MSG_CHECKING([whether add_disk() returns int])
 	ZFS_LINUX_TEST_RESULT([add_disk_ret],

config/kernel-bio.m4

@@ -464,7 +464,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_CGROUP_HEADER], [
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLK_CGROUP_HEADER], [
-	AC_MSG_CHECKING([for existence of linux/blk-cgroup.h])
+	AC_MSG_CHECKING([whether linux/blk-cgroup.h exists])
 	ZFS_LINUX_TEST_RESULT([blk_cgroup_header],[
 		AC_MSG_RESULT(yes)
 		AC_DEFINE(HAVE_LINUX_BLK_CGROUP_HEADER, 1,
@@ -474,6 +474,41 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_CGROUP_HEADER], [
 	])
 ])
 
+dnl #
+dnl # Linux 5.18 API
+dnl #
+dnl # In 07888c665b405b1cd3577ddebfeb74f4717a84c4 ("block: pass a block_device and opf to bio_alloc")
+dnl #   bio_alloc(gfp_t gfp_mask, unsigned short nr_iovecs)
+dnl # became
+dnl #   bio_alloc(struct block_device *bdev, unsigned short nr_vecs, unsigned int opf, gfp_t gfp_mask)
+dnl # however
+dnl # > NULL/0 can be passed, both for the
+dnl # > passthrough case on a raw request_queue and to temporarily avoid
+dnl # > refactoring some nasty code.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BIO_ALLOC_4ARG], [
+	ZFS_LINUX_TEST_SRC([bio_alloc_4arg], [
+		#include <linux/bio.h>
+	],[
+		gfp_t gfp_mask = 0;
+		unsigned short nr_iovecs = 0;
+		struct block_device *bdev = NULL;
+		unsigned int opf = 0;
+
+		struct bio *__attribute__((unused)) allocated = bio_alloc(bdev, nr_iovecs, opf, gfp_mask);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BIO_ALLOC_4ARG], [
+	AC_MSG_CHECKING([whether bio_alloc() wants 4 args])
+	ZFS_LINUX_TEST_RESULT([bio_alloc_4arg],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE([HAVE_BIO_ALLOC_4ARG], 1, [bio_alloc() takes 4 arguments])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BIO], [
 	ZFS_AC_KERNEL_SRC_REQ
 	ZFS_AC_KERNEL_SRC_BIO_OPS
@@ -488,6 +523,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BIO], [
 	ZFS_AC_KERNEL_SRC_BDEV_SUBMIT_BIO_RETURNS_VOID
 	ZFS_AC_KERNEL_SRC_BIO_SET_DEV_MACRO
 	ZFS_AC_KERNEL_SRC_BLK_CGROUP_HEADER
+	ZFS_AC_KERNEL_SRC_BIO_ALLOC_4ARG
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BIO], [
@@ -512,4 +548,5 @@ AC_DEFUN([ZFS_AC_KERNEL_BIO], [
 	ZFS_AC_KERNEL_BIO_BDEV_DISK
 	ZFS_AC_KERNEL_BDEV_SUBMIT_BIO_RETURNS_VOID
 	ZFS_AC_KERNEL_BLK_CGROUP_HEADER
+	ZFS_AC_KERNEL_BIO_ALLOC_4ARG
 ])

config/kernel-blk-queue.m4

@@ -74,6 +74,8 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_UPDATE_READAHEAD], [
 		AC_DEFINE(HAVE_BLK_QUEUE_UPDATE_READAHEAD, 1,
 		    [blk_queue_update_readahead() exists])
 	],[
+		AC_MSG_RESULT(no)
+
 		AC_MSG_CHECKING([whether disk_update_readahead() exists])
 		ZFS_LINUX_TEST_RESULT([disk_update_readahead], [
 			AC_MSG_RESULT(yes)
@@ -86,69 +88,111 @@ AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_UPDATE_READAHEAD], [
 ])
 
 dnl #
-dnl # 2.6.32 API,
-dnl #   blk_queue_discard()
+dnl # 5.19: bdev_max_discard_sectors() available
+dnl # 2.6.32: blk_queue_discard() available
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_DISCARD], [
+	ZFS_LINUX_TEST_SRC([bdev_max_discard_sectors], [
+		#include <linux/blkdev.h>
+	],[
+		struct block_device *bdev __attribute__ ((unused)) = NULL;
+		unsigned int error __attribute__ ((unused));
+
+		error = bdev_max_discard_sectors(bdev);
+	])
+
 	ZFS_LINUX_TEST_SRC([blk_queue_discard], [
 		#include <linux/blkdev.h>
 	],[
-		struct request_queue *q __attribute__ ((unused)) = NULL;
+		struct request_queue r;
+		struct request_queue *q = &r;
 		int value __attribute__ ((unused));
+		memset(q, 0, sizeof(r));
 		value = blk_queue_discard(q);
 	])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_DISCARD], [
-	AC_MSG_CHECKING([whether blk_queue_discard() is available])
-	ZFS_LINUX_TEST_RESULT([blk_queue_discard], [
+	AC_MSG_CHECKING([whether bdev_max_discard_sectors() is available])
+	ZFS_LINUX_TEST_RESULT([bdev_max_discard_sectors], [
 		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BDEV_MAX_DISCARD_SECTORS, 1,
+		    [bdev_max_discard_sectors() is available])
 	],[
-		ZFS_LINUX_TEST_ERROR([blk_queue_discard])
+		AC_MSG_RESULT(no)
+
+		AC_MSG_CHECKING([whether blk_queue_discard() is available])
+		ZFS_LINUX_TEST_RESULT([blk_queue_discard], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_BLK_QUEUE_DISCARD, 1,
+			    [blk_queue_discard() is available])
+		],[
+			ZFS_LINUX_TEST_ERROR([blk_queue_discard])
+		])
 	])
 ])
 
 dnl #
-dnl # 4.8 API,
-dnl #   blk_queue_secure_erase()
-dnl #
-dnl # 2.6.36 - 4.7 API,
-dnl #   blk_queue_secdiscard()
+dnl # 5.19: bdev_max_secure_erase_sectors() available
+dnl # 4.8: blk_queue_secure_erase() available
+dnl # 2.6.36: blk_queue_secdiscard() available
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_SECURE_ERASE], [
+	ZFS_LINUX_TEST_SRC([bdev_max_secure_erase_sectors], [
+		#include <linux/blkdev.h>
+	],[
+		struct block_device *bdev __attribute__ ((unused)) = NULL;
+		unsigned int error __attribute__ ((unused));
+
+		error = bdev_max_secure_erase_sectors(bdev);
+	])
+
 	ZFS_LINUX_TEST_SRC([blk_queue_secure_erase], [
 		#include <linux/blkdev.h>
 	],[
-		struct request_queue *q __attribute__ ((unused)) = NULL;
+		struct request_queue r;
+		struct request_queue *q = &r;
 		int value __attribute__ ((unused));
+		memset(q, 0, sizeof(r));
 		value = blk_queue_secure_erase(q);
 	])
 
 	ZFS_LINUX_TEST_SRC([blk_queue_secdiscard], [
 		#include <linux/blkdev.h>
 	],[
-		struct request_queue *q __attribute__ ((unused)) = NULL;
+		struct request_queue r;
+		struct request_queue *q = &r;
 		int value __attribute__ ((unused));
+		memset(q, 0, sizeof(r));
 		value = blk_queue_secdiscard(q);
 	])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_SECURE_ERASE], [
-	AC_MSG_CHECKING([whether blk_queue_secure_erase() is available])
-	ZFS_LINUX_TEST_RESULT([blk_queue_secure_erase], [
+	AC_MSG_CHECKING([whether bdev_max_secure_erase_sectors() is available])
+	ZFS_LINUX_TEST_RESULT([bdev_max_secure_erase_sectors], [
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_BLK_QUEUE_SECURE_ERASE, 1,
-		    [blk_queue_secure_erase() is available])
+		AC_DEFINE(HAVE_BDEV_MAX_SECURE_ERASE_SECTORS, 1,
+		    [bdev_max_secure_erase_sectors() is available])
 	],[
 		AC_MSG_RESULT(no)
 
-		AC_MSG_CHECKING([whether blk_queue_secdiscard() is available])
-		ZFS_LINUX_TEST_RESULT([blk_queue_secdiscard], [
+		AC_MSG_CHECKING([whether blk_queue_secure_erase() is available])
+		ZFS_LINUX_TEST_RESULT([blk_queue_secure_erase], [
 			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_BLK_QUEUE_SECDISCARD, 1,
-			    [blk_queue_secdiscard() is available])
+			AC_DEFINE(HAVE_BLK_QUEUE_SECURE_ERASE, 1,
+			    [blk_queue_secure_erase() is available])
 		],[
-			ZFS_LINUX_TEST_ERROR([blk_queue_secure_erase])
+			AC_MSG_RESULT(no)
+
+			AC_MSG_CHECKING([whether blk_queue_secdiscard() is available])
+			ZFS_LINUX_TEST_RESULT([blk_queue_secdiscard], [
+				AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_BLK_QUEUE_SECDISCARD, 1,
+				    [blk_queue_secdiscard() is available])
+			],[
+				ZFS_LINUX_TEST_ERROR([blk_queue_secure_erase])
+			])
 		])
 	])
 ])
@@ -215,17 +259,17 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_FLUSH], [
 	ZFS_LINUX_TEST_SRC([blk_queue_flush], [
 		#include <linux/blkdev.h>
 	], [
-		struct request_queue *q = NULL;
+		struct request_queue *q __attribute__ ((unused)) = NULL;
 		(void) blk_queue_flush(q, REQ_FLUSH);
-	], [$NO_UNUSED_BUT_SET_VARIABLE], [ZFS_META_LICENSE])
+	], [], [ZFS_META_LICENSE])
 
 	ZFS_LINUX_TEST_SRC([blk_queue_write_cache], [
 		#include <linux/kernel.h>
 		#include <linux/blkdev.h>
 	], [
-		struct request_queue *q = NULL;
+		struct request_queue *q __attribute__ ((unused)) = NULL;
 		blk_queue_write_cache(q, true, true);
-	], [$NO_UNUSED_BUT_SET_VARIABLE], [ZFS_META_LICENSE])
+	], [], [ZFS_META_LICENSE])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_FLUSH], [
@@ -278,9 +322,9 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_MAX_HW_SECTORS], [
 	ZFS_LINUX_TEST_SRC([blk_queue_max_hw_sectors], [
 		#include <linux/blkdev.h>
 	], [
-		struct request_queue *q = NULL;
+		struct request_queue *q __attribute__ ((unused)) = NULL;
 		(void) blk_queue_max_hw_sectors(q, BLK_SAFE_MAX_SECTORS);
-	], [$NO_UNUSED_BUT_SET_VARIABLE])
+	], [])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_MAX_HW_SECTORS], [
@@ -301,9 +345,9 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLK_QUEUE_MAX_SEGMENTS], [
 	ZFS_LINUX_TEST_SRC([blk_queue_max_segments], [
 		#include <linux/blkdev.h>
 	], [
-		struct request_queue *q = NULL;
+		struct request_queue *q __attribute__ ((unused)) = NULL;
 		(void) blk_queue_max_segments(q, BLK_MAX_SEGMENTS);
-	], [$NO_UNUSED_BUT_SET_VARIABLE])
+	], [])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLK_QUEUE_MAX_SEGMENTS], [

config/kernel-blkdev.m4

@@ -294,6 +294,83 @@ AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEV_WHOLE], [
 	])
 ])
 
+dnl #
+dnl # 5.20 API change,
+dnl # Removed bdevname(), snprintf(.., %pg) should be used.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_BDEVNAME], [
+	ZFS_LINUX_TEST_SRC([bdevname], [
+		#include <linux/fs.h>
+		#include <linux/blkdev.h>
+	], [
+		struct block_device *bdev __attribute__ ((unused)) = NULL;
+		char path[BDEVNAME_SIZE];
+
+		(void) bdevname(bdev, path);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_BDEVNAME], [
+	AC_MSG_CHECKING([whether bdevname() exists])
+	ZFS_LINUX_TEST_RESULT([bdevname], [
+		AC_DEFINE(HAVE_BDEVNAME, 1, [bdevname() is available])
+		AC_MSG_RESULT(yes)
+	], [
+		AC_MSG_RESULT(no)
+	])
+])
+
+dnl #
+dnl # 5.19 API: blkdev_issue_secure_erase()
+dnl # 3.10 API: blkdev_issue_discard(..., BLKDEV_DISCARD_SECURE)
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE], [
+	ZFS_LINUX_TEST_SRC([blkdev_issue_secure_erase], [
+		#include <linux/blkdev.h>
+	],[
+		struct block_device *bdev = NULL;
+		sector_t sector = 0;
+		sector_t nr_sects = 0;
+		int error __attribute__ ((unused));
+
+		error = blkdev_issue_secure_erase(bdev,
+		    sector, nr_sects, GFP_KERNEL);
+	])
+
+	ZFS_LINUX_TEST_SRC([blkdev_issue_discard_flags], [
+		#include <linux/blkdev.h>
+	],[
+		struct block_device *bdev = NULL;
+		sector_t sector = 0;
+		sector_t nr_sects = 0;
+		unsigned long flags = 0;
+		int error __attribute__ ((unused));
+
+		error = blkdev_issue_discard(bdev,
+		    sector, nr_sects, GFP_KERNEL, flags);
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE], [
+	AC_MSG_CHECKING([whether blkdev_issue_secure_erase() is available])
+	ZFS_LINUX_TEST_RESULT([blkdev_issue_secure_erase], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_BLKDEV_ISSUE_SECURE_ERASE, 1,
+		    [blkdev_issue_secure_erase() is available])
+	],[
+		AC_MSG_RESULT(no)
+
+		AC_MSG_CHECKING([whether blkdev_issue_discard() is available])
+		ZFS_LINUX_TEST_RESULT([blkdev_issue_discard_flags], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(HAVE_BLKDEV_ISSUE_DISCARD, 1,
+			    [blkdev_issue_discard() is available])
+		],[
+			ZFS_LINUX_TEST_ERROR([blkdev_issue_discard()])
+		])
+	])
+])
+
 dnl #
 dnl # 5.13 API change
 dnl # blkdev_get_by_path() no longer handles ERESTARTSYS
@@ -326,6 +403,8 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLKDEV], [
 	ZFS_AC_KERNEL_SRC_BLKDEV_CHECK_DISK_CHANGE
 	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_CHECK_MEDIA_CHANGE
 	ZFS_AC_KERNEL_SRC_BLKDEV_BDEV_WHOLE
+	ZFS_AC_KERNEL_SRC_BLKDEV_BDEVNAME
+	ZFS_AC_KERNEL_SRC_BLKDEV_ISSUE_SECURE_ERASE
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [
@@ -339,5 +418,7 @@ AC_DEFUN([ZFS_AC_KERNEL_BLKDEV], [
 	ZFS_AC_KERNEL_BLKDEV_CHECK_DISK_CHANGE
 	ZFS_AC_KERNEL_BLKDEV_BDEV_CHECK_MEDIA_CHANGE
 	ZFS_AC_KERNEL_BLKDEV_BDEV_WHOLE
+	ZFS_AC_KERNEL_BLKDEV_BDEVNAME
 	ZFS_AC_KERNEL_BLKDEV_GET_ERESTARTSYS
+	ZFS_AC_KERNEL_BLKDEV_ISSUE_SECURE_ERASE
 ])

config/kernel-block-device-operations.m4

@@ -6,13 +6,16 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS], [
 		#include <linux/blkdev.h>
 
 		unsigned int blk_check_events(struct gendisk *disk,
-		    unsigned int clearing) { return (0); }
+		    unsigned int clearing) {
+			(void) disk, (void) clearing;
+			return (0);
+		}
 
 		static const struct block_device_operations
 		    bops __attribute__ ((unused)) = {
 			.check_events	= blk_check_events,
 		};
-	], [], [$NO_UNUSED_BUT_SET_VARIABLE])
+	], [], [])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS], [
@@ -31,7 +34,10 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID], [
 	ZFS_LINUX_TEST_SRC([block_device_operations_release_void], [
 		#include <linux/blkdev.h>
 
-		void blk_release(struct gendisk *g, fmode_t mode) { return; }
+		void blk_release(struct gendisk *g, fmode_t mode) {
+			(void) g, (void) mode;
+			return;
+		}
 
 		static const struct block_device_operations
 		    bops __attribute__ ((unused)) = {
@@ -40,7 +46,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID], [
 			.ioctl		= NULL,
 			.compat_ioctl	= NULL,
 		};
-	], [], [$NO_UNUSED_BUT_SET_VARIABLE])
+	], [], [])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID], [
@@ -61,6 +67,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [
 		#include <linux/blkdev.h>
 
 		int blk_revalidate_disk(struct gendisk *disk) {
+			(void) disk;
 			return(0);
 		}
 
@@ -68,7 +75,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [
 		    bops __attribute__ ((unused)) = {
 			.revalidate_disk	= blk_revalidate_disk,
 		};
-	], [], [$NO_UNUSED_BUT_SET_VARIABLE])
+	], [], [])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [

config/kernel-config-defined.m4

@@ -19,19 +19,48 @@ AC_DEFUN([ZFS_AC_KERNEL_CONFIG_DEFINED], [
 		])
 	])
 
+	ZFS_AC_KERNEL_SRC_CONFIG_MODULES
+	ZFS_AC_KERNEL_SRC_CONFIG_BLOCK
 	ZFS_AC_KERNEL_SRC_CONFIG_DEBUG_LOCK_ALLOC
 	ZFS_AC_KERNEL_SRC_CONFIG_TRIM_UNUSED_KSYMS
-	ZFS_AC_KERNEL_SRC_CONFIG_ZLIB_INFLATE
 	ZFS_AC_KERNEL_SRC_CONFIG_ZLIB_DEFLATE
+	ZFS_AC_KERNEL_SRC_CONFIG_ZLIB_INFLATE
 
 	AC_MSG_CHECKING([for kernel config option compatibility])
 	ZFS_LINUX_TEST_COMPILE_ALL([config])
 	AC_MSG_RESULT([done])
 
+	ZFS_AC_KERNEL_CONFIG_MODULES
+	ZFS_AC_KERNEL_CONFIG_BLOCK
 	ZFS_AC_KERNEL_CONFIG_DEBUG_LOCK_ALLOC
 	ZFS_AC_KERNEL_CONFIG_TRIM_UNUSED_KSYMS
-	ZFS_AC_KERNEL_CONFIG_ZLIB_INFLATE
 	ZFS_AC_KERNEL_CONFIG_ZLIB_DEFLATE
+	ZFS_AC_KERNEL_CONFIG_ZLIB_INFLATE
+])
+
+dnl #
+dnl # Check CONFIG_BLOCK
+dnl #
+dnl # Verify the kernel has CONFIG_BLOCK support enabled.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_CONFIG_BLOCK], [
+	ZFS_LINUX_TEST_SRC([config_block], [
+		#if !defined(CONFIG_BLOCK)
+		#error CONFIG_BLOCK not defined
+		#endif
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_CONFIG_BLOCK], [
+	AC_MSG_CHECKING([whether CONFIG_BLOCK is defined])
+	ZFS_LINUX_TEST_RESULT([config_block], [
+		AC_MSG_RESULT([yes])
+	],[
+		AC_MSG_RESULT([no])
+		AC_MSG_ERROR([
+	*** This kernel does not include the required block device support.
+	*** Rebuild the kernel with CONFIG_BLOCK=y set.])
+	])
 ])
 
 dnl #
@@ -72,6 +101,61 @@ AC_DEFUN([ZFS_AC_KERNEL_CONFIG_DEBUG_LOCK_ALLOC], [
 	])
 ])
 
+dnl #
+dnl # Check CONFIG_MODULES
+dnl #
+dnl # Verify the kernel has CONFIG_MODULES support enabled.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_CONFIG_MODULES], [
+	ZFS_LINUX_TEST_SRC([config_modules], [
+		#if !defined(CONFIG_MODULES)
+		#error CONFIG_MODULES not defined
+		#endif
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_CONFIG_MODULES], [
+	AC_MSG_CHECKING([whether CONFIG_MODULES is defined])
+	AS_IF([test "x$enable_linux_builtin" != xyes], [
+		ZFS_LINUX_TEST_RESULT([config_modules], [
+			AC_MSG_RESULT([yes])
+		],[
+			AC_MSG_RESULT([no])
+			AC_MSG_ERROR([
+		*** This kernel does not include the required loadable module
+		*** support!
+		***
+		*** To build OpenZFS as a loadable Linux kernel module
+		*** enable loadable module support by setting
+		*** `CONFIG_MODULES=y` in the kernel configuration and run
+		*** `make modules_prepare` in the Linux source tree.
+		***
+		*** If you don't intend to enable loadable kernel module
+		*** support, please compile OpenZFS as a Linux kernel built-in.
+		***
+		*** Prepare the Linux source tree by running `make prepare`,
+		*** use the OpenZFS `--enable-linux-builtin` configure option,
+		*** copy the OpenZFS sources into the Linux source tree using
+		*** `./copy-builtin <linux source directory>`,
+		*** set `CONFIG_ZFS=y` in the kernel configuration and compile
+		*** kernel as usual.
+			])
+		])
+	], [
+		ZFS_LINUX_TRY_COMPILE([], [], [
+			AC_MSG_RESULT([not needed])
+		],[
+			AC_MSG_RESULT([error])
+			AC_MSG_ERROR([
+		*** This kernel is unable to compile object files.
+		***
+		*** Please make sure you prepared the Linux source tree
+		*** by running `make prepare` there.
+			])
+		])
+	])
+])
+
 dnl #
 dnl # Check CONFIG_TRIM_UNUSED_KSYMS
 dnl #

config/kernel-copy-from-user-inatomic.m4

@@ -0,0 +1,29 @@
+dnl #
+dnl # On certain architectures `__copy_from_user_inatomic`
+dnl # is a GPL exported variable and cannot be used by OpenZFS.
+dnl #
+
+dnl #
+dnl # Checking if `__copy_from_user_inatomic` is available.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC___COPY_FROM_USER_INATOMIC], [
+	ZFS_LINUX_TEST_SRC([__copy_from_user_inatomic], [
+		#include <linux/uaccess.h>
+	], [
+		int result __attribute__ ((unused)) = __copy_from_user_inatomic(NULL, NULL, 0);
+	], [], [ZFS_META_LICENSE])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL___COPY_FROM_USER_INATOMIC], [
+	AC_MSG_CHECKING([whether __copy_from_user_inatomic is available])
+	ZFS_LINUX_TEST_RESULT([__copy_from_user_inatomic_license], [
+		AC_MSG_RESULT(yes)
+	], [
+		AC_MSG_RESULT(no)
+		AC_MSG_ERROR([
+	*** The `__copy_from_user_inatomic()` Linux kernel function is
+	*** incompatible with the CDDL license and will prevent the module
+	*** linking stage from succeeding.  OpenZFS cannot be compiled.
+		])
+	])
+])

config/kernel-fpu.m4

@@ -2,6 +2,9 @@ dnl #
 dnl # Handle differences in kernel FPU code.
 dnl #
 dnl # Kernel
+dnl # 5.19:	The asm/fpu/internal.h header was removed, it has been
+dnl #		effectively empty since the 5.16 kernel.
+dnl #
 dnl # 5.16:	XCR code put into asm/fpu/xcr.h
 dnl #		HAVE_KERNEL_FPU_XCR_HEADER
 dnl #
@@ -33,21 +36,31 @@ AC_DEFUN([ZFS_AC_KERNEL_FPU_HEADER], [
 	],[
 		AC_DEFINE(HAVE_KERNEL_FPU_API_HEADER, 1,
 		    [kernel has asm/fpu/api.h])
-		AC_MSG_RESULT(asm/fpu/api.h)
-		AC_MSG_CHECKING([whether fpu/xcr header is available])
+		fpu_headers="asm/fpu/api.h"
+
 		ZFS_LINUX_TRY_COMPILE([
 			#include <linux/module.h>
 			#include <asm/fpu/xcr.h>
 		],[
 		],[
 			AC_DEFINE(HAVE_KERNEL_FPU_XCR_HEADER, 1,
-				[kernel has asm/fpu/xcr.h])
-			AC_MSG_RESULT(asm/fpu/xcr.h)
-		],[
-			AC_MSG_RESULT(no asm/fpu/xcr.h)
+			    [kernel has asm/fpu/xcr.h])
+			fpu_headers="$fpu_headers asm/fpu/xcr.h"
 		])
+
+		ZFS_LINUX_TRY_COMPILE([
+			#include <linux/module.h>
+			#include <asm/fpu/internal.h>
+		],[
+		],[
+			AC_DEFINE(HAVE_KERNEL_FPU_INTERNAL_HEADER, 1,
+			    [kernel has asm/fpu/internal.h])
+			fpu_headers="$fpu_headers asm/fpu/internal.h"
+		])
+
+		AC_MSG_RESULT([$fpu_headers])
 	],[
-		AC_MSG_RESULT(i387.h & xcr.h)
+		AC_MSG_RESULT([i387.h & xcr.h])
 	])
 ])
 
@@ -93,7 +106,9 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_FPU], [
 		#include <linux/types.h>
 		#ifdef HAVE_KERNEL_FPU_API_HEADER
 		#include <asm/fpu/api.h>
+		#ifdef HAVE_KERNEL_FPU_INTERNAL_HEADER
 		#include <asm/fpu/internal.h>
+		#endif
 		#else
 		#include <asm/i387.h>
 		#include <asm/xcr.h>
@@ -130,7 +145,9 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_FPU], [
 		#include <linux/types.h>
 		#ifdef HAVE_KERNEL_FPU_API_HEADER
 		#include <asm/fpu/api.h>
+		#ifdef HAVE_KERNEL_FPU_INTERNAL_HEADER
 		#include <asm/fpu/internal.h>
+		#endif
 		#else
 		#include <asm/i387.h>
 		#include <asm/xcr.h>

config/kernel-generic_io_acct.m4

@@ -2,6 +2,19 @@ dnl #
 dnl # Check for generic io accounting interface.
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_GENERIC_IO_ACCT], [
+	ZFS_LINUX_TEST_SRC([bdev_io_acct], [
+		#include <linux/blkdev.h>
+	], [
+		struct block_device *bdev = NULL;
+		struct bio *bio = NULL;
+		unsigned long passed_time = 0;
+		unsigned long start_time;
+
+		start_time = bdev_start_io_acct(bdev, bio_sectors(bio),
+		    bio_op(bio), passed_time);
+		bdev_end_io_acct(bdev, bio_op(bio), start_time);
+	])
+
 	ZFS_LINUX_TEST_SRC([disk_io_acct], [
 		#include <linux/blkdev.h>
 	], [
@@ -50,61 +63,75 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_GENERIC_IO_ACCT], [
 
 AC_DEFUN([ZFS_AC_KERNEL_GENERIC_IO_ACCT], [
 	dnl #
-	dnl # 5.12 API,
+	dnl # 5.19 API,
 	dnl #
-	dnl # bio_start_io_acct() and bio_end_io_acct() became GPL-exported
-	dnl # so use disk_start_io_acct() and disk_end_io_acct() instead
+	dnl # disk_start_io_acct() and disk_end_io_acct() have been replaced by
+	dnl # bdev_start_io_acct() and bdev_end_io_acct().
 	dnl #
-	AC_MSG_CHECKING([whether generic disk_*_io_acct() are available])
-	ZFS_LINUX_TEST_RESULT([disk_io_acct], [
+	AC_MSG_CHECKING([whether generic bdev_*_io_acct() are available])
+	ZFS_LINUX_TEST_RESULT([bdev_io_acct], [
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_DISK_IO_ACCT, 1, [disk_*_io_acct() available])
+		AC_DEFINE(HAVE_BDEV_IO_ACCT, 1, [bdev_*_io_acct() available])
 	], [
 		AC_MSG_RESULT(no)
 
 		dnl #
-		dnl # 5.7 API,
+		dnl # 5.12 API,
 		dnl #
-		dnl # Added bio_start_io_acct() and bio_end_io_acct() helpers.
+		dnl # bio_start_io_acct() and bio_end_io_acct() became GPL-exported
+		dnl # so use disk_start_io_acct() and disk_end_io_acct() instead
 		dnl #
-		AC_MSG_CHECKING([whether generic bio_*_io_acct() are available])
-		ZFS_LINUX_TEST_RESULT([bio_io_acct], [
+		AC_MSG_CHECKING([whether generic disk_*_io_acct() are available])
+		ZFS_LINUX_TEST_RESULT([disk_io_acct], [
 			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_BIO_IO_ACCT, 1, [bio_*_io_acct() available])
+			AC_DEFINE(HAVE_DISK_IO_ACCT, 1, [disk_*_io_acct() available])
 		], [
 			AC_MSG_RESULT(no)
 
 			dnl #
-			dnl # 4.14 API,
+			dnl # 5.7 API,
 			dnl #
-			dnl # generic_start_io_acct/generic_end_io_acct now require
-			dnl # request_queue to be provided. No functional changes,
-			dnl # but preparation for inflight accounting.
+			dnl # Added bio_start_io_acct() and bio_end_io_acct() helpers.
 			dnl #
-			AC_MSG_CHECKING([whether generic_*_io_acct wants 4 args])
-			ZFS_LINUX_TEST_RESULT_SYMBOL([generic_acct_4args],
-			    [generic_start_io_acct], [block/bio.c], [
+			AC_MSG_CHECKING([whether generic bio_*_io_acct() are available])
+			ZFS_LINUX_TEST_RESULT([bio_io_acct], [
 				AC_MSG_RESULT(yes)
-				AC_DEFINE(HAVE_GENERIC_IO_ACCT_4ARG, 1,
-				    [generic_*_io_acct() 4 arg available])
+				AC_DEFINE(HAVE_BIO_IO_ACCT, 1, [bio_*_io_acct() available])
 			], [
 				AC_MSG_RESULT(no)
 
 				dnl #
-				dnl # 3.19 API addition
+				dnl # 4.14 API,
 				dnl #
-				dnl # torvalds/linux@394ffa50 allows us to increment
-				dnl # iostat counters without generic_make_request().
+				dnl # generic_start_io_acct/generic_end_io_acct now require
+				dnl # request_queue to be provided. No functional changes,
+				dnl # but preparation for inflight accounting.
 				dnl #
-				AC_MSG_CHECKING(
-				    [whether generic_*_io_acct wants 3 args])
-				ZFS_LINUX_TEST_RESULT_SYMBOL([generic_acct_3args],
+				AC_MSG_CHECKING([whether generic_*_io_acct wants 4 args])
+				ZFS_LINUX_TEST_RESULT_SYMBOL([generic_acct_4args],
 				    [generic_start_io_acct], [block/bio.c], [
 					AC_MSG_RESULT(yes)
-					AC_DEFINE(HAVE_GENERIC_IO_ACCT_3ARG, 1,
-					    [generic_*_io_acct() 3 arg available])
+					AC_DEFINE(HAVE_GENERIC_IO_ACCT_4ARG, 1,
+					    [generic_*_io_acct() 4 arg available])
 				], [
 					AC_MSG_RESULT(no)
+
+					dnl #
+					dnl # 3.19 API addition
+					dnl #
+					dnl # torvalds/linux@394ffa50 allows us to increment
+					dnl # iostat counters without generic_make_request().
+					dnl #
+					AC_MSG_CHECKING(
+					    [whether generic_*_io_acct wants 3 args])
+					ZFS_LINUX_TEST_RESULT_SYMBOL([generic_acct_3args],
+					    [generic_start_io_acct], [block/bio.c], [
+						AC_MSG_RESULT(yes)
+						AC_DEFINE(HAVE_GENERIC_IO_ACCT_3ARG, 1,
+						    [generic_*_io_acct() 3 arg available])
+					], [
+						AC_MSG_RESULT(no)
+					])
 				])
 			])
 		])

config/kernel-genhd-flags.m4

@@ -0,0 +1,58 @@
+dnl #
+dnl # 5.17 API change,
+dnl #
+dnl # GENHD_FL_EXT_DEVT flag removed
+dnl # GENHD_FL_NO_PART_SCAN renamed GENHD_FL_NO_PART
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_GENHD_FLAGS], [
+
+	ZFS_LINUX_TEST_SRC([genhd_fl_ext_devt], [
+		#include <linux/blkdev.h>
+	], [
+		int flags __attribute__ ((unused)) = GENHD_FL_EXT_DEVT;
+	])
+
+	ZFS_LINUX_TEST_SRC([genhd_fl_no_part], [
+		#include <linux/blkdev.h>
+	], [
+		int flags __attribute__ ((unused)) = GENHD_FL_NO_PART;
+	])
+
+	ZFS_LINUX_TEST_SRC([genhd_fl_no_part_scan], [
+		#include <linux/blkdev.h>
+	], [
+		int flags __attribute__ ((unused)) = GENHD_FL_NO_PART_SCAN;
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_GENHD_FLAGS], [
+
+	AC_MSG_CHECKING([whether GENHD_FL_EXT_DEVT flag is available])
+	ZFS_LINUX_TEST_RESULT([genhd_fl_ext_devt], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(ZFS_GENHD_FL_EXT_DEVT, GENHD_FL_EXT_DEVT,
+		    [GENHD_FL_EXT_DEVT flag is available])
+	], [
+		AC_MSG_RESULT(no)
+		AC_DEFINE(ZFS_GENHD_FL_EXT_DEVT, 0,
+		    [GENHD_FL_EXT_DEVT flag is not available])
+	])
+
+	AC_MSG_CHECKING([whether GENHD_FL_NO_PART flag is available])
+	ZFS_LINUX_TEST_RESULT([genhd_fl_no_part], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(ZFS_GENHD_FL_NO_PART, GENHD_FL_NO_PART,
+		    [GENHD_FL_NO_PART flag is available])
+	], [
+		AC_MSG_RESULT(no)
+
+		AC_MSG_CHECKING([whether GENHD_FL_NO_PART_SCAN flag is available])
+		ZFS_LINUX_TEST_RESULT([genhd_fl_no_part_scan], [
+			AC_MSG_RESULT(yes)
+			AC_DEFINE(ZFS_GENHD_FL_NO_PART, GENHD_FL_NO_PART_SCAN,
+			    [GENHD_FL_NO_PART_SCAN flag is available])
+		], [
+			ZFS_LINUX_TEST_ERROR([GENHD_FL_NO_PART|GENHD_FL_NO_PART_SCAN])
+		])
+	])
+])

config/kernel-get-disk-ro.m4

@@ -5,9 +5,9 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_GET_DISK_RO], [
 	ZFS_LINUX_TEST_SRC([get_disk_ro], [
 		#include <linux/blkdev.h>
 	],[
-		struct gendisk *disk = NULL;
+		struct gendisk *disk __attribute__ ((unused)) = NULL;
 		(void) get_disk_ro(disk);
-	], [$NO_UNUSED_BUT_SET_VARIABLE])
+	], [])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_GET_DISK_RO], [

config/kernel-group-info.m4

@@ -6,8 +6,8 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_GROUP_INFO_GID], [
 	ZFS_LINUX_TEST_SRC([group_info_gid], [
 		#include <linux/cred.h>
 	],[
-		struct group_info *gi = groups_alloc(1);
-		gi->gid[0] = KGIDT_INIT(0);
+		struct group_info gi __attribute__ ((unused)) = {};
+		gi.gid[0] = KGIDT_INIT(0);
 	])
 ])
 

config/kernel-make-request-fn.m4

@@ -49,6 +49,13 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_MAKE_REQUEST_FN], [
 		struct gendisk *disk  __attribute__ ((unused));
 		disk = blk_alloc_disk(NUMA_NO_NODE);
 	])
+
+	ZFS_LINUX_TEST_SRC([blk_cleanup_disk], [
+		#include <linux/blkdev.h>
+	],[
+		struct gendisk *disk  __attribute__ ((unused));
+		blk_cleanup_disk(disk);
+	])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_MAKE_REQUEST_FN], [
@@ -73,6 +80,19 @@ AC_DEFUN([ZFS_AC_KERNEL_MAKE_REQUEST_FN], [
 		ZFS_LINUX_TEST_RESULT([blk_alloc_disk], [
 			AC_MSG_RESULT(yes)
 			AC_DEFINE([HAVE_BLK_ALLOC_DISK], 1, [blk_alloc_disk() exists])
+
+			dnl #
+			dnl # 5.20 API change,
+			dnl # Removed blk_cleanup_disk(), put_disk() should be used.
+			dnl #
+			AC_MSG_CHECKING([whether blk_cleanup_disk() exists])
+			ZFS_LINUX_TEST_RESULT([blk_cleanup_disk], [
+				AC_MSG_RESULT(yes)
+				AC_DEFINE([HAVE_BLK_CLEANUP_DISK], 1,
+				    [blk_cleanup_disk() exists])
+			], [
+				AC_MSG_RESULT(no)
+			])
 		], [
 			AC_MSG_RESULT(no)
 		])

config/kernel-mkdir.m4

@@ -53,6 +53,8 @@ AC_DEFUN([ZFS_AC_KERNEL_MKDIR], [
 		AC_DEFINE(HAVE_IOPS_MKDIR_USERNS, 1,
 		    [iops->mkdir() takes struct user_namespace*])
 	],[
+		AC_MSG_RESULT(no)
+
 		AC_MSG_CHECKING([whether iops->mkdir() takes umode_t])
 		ZFS_LINUX_TEST_RESULT([inode_operations_mkdir], [
 			AC_MSG_RESULT(yes)

config/kernel-pagemap-folio_wait_bit.m4

@@ -15,7 +15,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_PAGEMAP_FOLIO_WAIT_BIT], [
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_PAGEMAP_FOLIO_WAIT_BIT], [
-	AC_MSG_CHECKING([folio_wait_bit() exists])
+	AC_MSG_CHECKING([whether folio_wait_bit() exists])
 	ZFS_LINUX_TEST_RESULT([pagemap_has_folio_wait_bit], [
 		AC_MSG_RESULT([yes])
 		AC_DEFINE(HAVE_PAGEMAP_FOLIO_WAIT_BIT, 1,

config/kernel-readpages.m4

@@ -0,0 +1,25 @@
+dnl #
+dnl # Linux 5.18 removes address_space_operations ->readpages in favour of
+dnl # ->readahead
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_READPAGES], [
+	ZFS_LINUX_TEST_SRC([vfs_has_readpages], [
+		#include <linux/fs.h>
+
+		static const struct address_space_operations
+		    aops __attribute__ ((unused)) = {
+			.readpages = NULL,
+		};
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_VFS_READPAGES], [
+	AC_MSG_CHECKING([whether aops->readpages exists])
+	ZFS_LINUX_TEST_RESULT([vfs_has_readpages], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_READPAGES, 1,
+			[address_space_operations->readpages exists])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel-revalidate-disk-size.m4

@@ -8,14 +8,14 @@ dnl #
 AC_DEFUN([ZFS_AC_KERNEL_SRC_REVALIDATE_DISK], [
 
 	ZFS_LINUX_TEST_SRC([revalidate_disk_size], [
-		#include <linux/genhd.h>
+		#include <linux/blkdev.h>
 	], [
 		struct gendisk *disk = NULL;
 		(void) revalidate_disk_size(disk, false);
 	])
 
 	ZFS_LINUX_TEST_SRC([revalidate_disk], [
-		#include <linux/genhd.h>
+		#include <linux/blkdev.h>
 	], [
 		struct gendisk *disk = NULL;
 		(void) revalidate_disk(disk);

config/kernel-shrink.m4

@@ -54,6 +54,21 @@ AC_DEFUN([ZFS_AC_KERNEL_SHRINK_CONTROL_HAS_NID], [
 	])
 ])
 
+AC_DEFUN([ZFS_AC_KERNEL_SRC_REGISTER_SHRINKER_VARARG], [
+	ZFS_LINUX_TEST_SRC([register_shrinker_vararg], [
+		#include <linux/mm.h>
+		unsigned long shrinker_cb(struct shrinker *shrink,
+		    struct shrink_control *sc) { return 0; }
+	],[
+		struct shrinker cache_shrinker = {
+			.count_objects = shrinker_cb,
+			.scan_objects = shrinker_cb,
+			.seeks = DEFAULT_SEEKS,
+		};
+		register_shrinker(&cache_shrinker, "vararg-reg-shrink-test");
+	])
+])
+
 AC_DEFUN([ZFS_AC_KERNEL_SRC_SHRINKER_CALLBACK], [
 	ZFS_LINUX_TEST_SRC([shrinker_cb_shrink_control], [
 		#include <linux/mm.h>
@@ -83,29 +98,50 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_SHRINKER_CALLBACK], [
 
 AC_DEFUN([ZFS_AC_KERNEL_SHRINKER_CALLBACK],[
 	dnl #
-	dnl # 3.0 - 3.11 API change
-	dnl # ->shrink(struct shrinker *, struct shrink_control *sc)
+	dnl # 6.0 API change
+	dnl # register_shrinker() becomes a var-arg function that takes
+	dnl # a printf-style format string as args > 0
 	dnl #
-	AC_MSG_CHECKING([whether new 2-argument shrinker exists])
-	ZFS_LINUX_TEST_RESULT([shrinker_cb_shrink_control], [
+	AC_MSG_CHECKING([whether new var-arg register_shrinker() exists])
+	ZFS_LINUX_TEST_RESULT([register_shrinker_vararg], [
 		AC_MSG_RESULT(yes)
-		AC_DEFINE(HAVE_SINGLE_SHRINKER_CALLBACK, 1,
-		    [new shrinker callback wants 2 args])
+		AC_DEFINE(HAVE_REGISTER_SHRINKER_VARARG, 1,
+		    [register_shrinker is vararg])
+
+		dnl # We assume that the split shrinker callback exists if the
+		dnl # vararg register_shrinker() exists, because the latter is
+		dnl # a much more recent addition, and the macro test for the
+		dnl # var-arg version only works if the callback is split
+		AC_DEFINE(HAVE_SPLIT_SHRINKER_CALLBACK, 1,
+			[cs->count_objects exists])
 	],[
 		AC_MSG_RESULT(no)
-
 		dnl #
-		dnl # 3.12 API change,
-		dnl # ->shrink() is logically split in to
-		dnl # ->count_objects() and ->scan_objects()
+		dnl # 3.0 - 3.11 API change
+		dnl # cs->shrink(struct shrinker *, struct shrink_control *sc)
 		dnl #
-		AC_MSG_CHECKING([whether ->count_objects callback exists])
-		ZFS_LINUX_TEST_RESULT([shrinker_cb_shrink_control_split], [
+		AC_MSG_CHECKING([whether new 2-argument shrinker exists])
+		ZFS_LINUX_TEST_RESULT([shrinker_cb_shrink_control], [
 			AC_MSG_RESULT(yes)
-			AC_DEFINE(HAVE_SPLIT_SHRINKER_CALLBACK, 1,
-			    [->count_objects exists])
+			AC_DEFINE(HAVE_SINGLE_SHRINKER_CALLBACK, 1,
+				[new shrinker callback wants 2 args])
 		],[
-			ZFS_LINUX_TEST_ERROR([shrinker])
+			AC_MSG_RESULT(no)
+
+			dnl #
+			dnl # 3.12 API change,
+			dnl # cs->shrink() is logically split in to
+			dnl # cs->count_objects() and cs->scan_objects()
+			dnl #
+			AC_MSG_CHECKING([if cs->count_objects callback exists])
+			ZFS_LINUX_TEST_RESULT(
+				[shrinker_cb_shrink_control_split],[
+					AC_MSG_RESULT(yes)
+					AC_DEFINE(HAVE_SPLIT_SHRINKER_CALLBACK, 1,
+						[cs->count_objects exists])
+			],[
+					ZFS_LINUX_TEST_ERROR([shrinker])
+			])
 		])
 	])
 ])
@@ -141,6 +177,7 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_SHRINKER], [
 	ZFS_AC_KERNEL_SRC_SHRINK_CONTROL_HAS_NID
 	ZFS_AC_KERNEL_SRC_SHRINKER_CALLBACK
 	ZFS_AC_KERNEL_SRC_SHRINK_CONTROL_STRUCT
+	ZFS_AC_KERNEL_SRC_REGISTER_SHRINKER_VARARG
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_SHRINKER], [

config/kernel-sysfs.m4

@@ -0,0 +1,37 @@
+dnl #
+dnl # Linux 5.2/5.18 API
+dnl #
+dnl # In cdb4f26a63c391317e335e6e683a614358e70aeb ("kobject: kobj_type: remove default_attrs")
+dnl # 	struct kobj_type.default_attrs
+dnl # was finally removed in favour of
+dnl # 	struct kobj_type.default_groups
+dnl #
+dnl # This was added in aa30f47cf666111f6bbfd15f290a27e8a7b9d854 ("kobject: Add support for default attribute groups to kobj_type"),
+dnl # if both are present (5.2-5.17), we prefer default_groups; they're otherwise equivalent
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_SYSFS_DEFAULT_GROUPS], [
+	ZFS_LINUX_TEST_SRC([sysfs_default_groups], [
+		#include <linux/kobject.h>
+	],[
+		struct kobj_type __attribute__ ((unused)) kt = {
+			.default_groups = (const struct attribute_group **)NULL };
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SYSFS_DEFAULT_GROUPS], [
+	AC_MSG_CHECKING([whether struct kobj_type.default_groups exists])
+	ZFS_LINUX_TEST_RESULT([sysfs_default_groups],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE([HAVE_SYSFS_DEFAULT_GROUPS], 1, [struct kobj_type has default_groups])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SRC_SYSFS], [
+	ZFS_AC_KERNEL_SRC_SYSFS_DEFAULT_GROUPS
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_SYSFS], [
+	ZFS_AC_KERNEL_SYSFS_DEFAULT_GROUPS
+])

config/kernel-vfs-filemap_dirty_folio.m4

@@ -0,0 +1,30 @@
+dnl #
+dnl # Linux 5.18 uses filemap_dirty_folio in lieu of
+dnl # ___set_page_dirty_nobuffers
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_FILEMAP_DIRTY_FOLIO], [
+	ZFS_LINUX_TEST_SRC([vfs_has_filemap_dirty_folio], [
+		#include <linux/pagemap.h>
+		#include <linux/writeback.h>
+
+		static const struct address_space_operations
+		    aops __attribute__ ((unused)) = {
+			.dirty_folio	= filemap_dirty_folio,
+		};
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_VFS_FILEMAP_DIRTY_FOLIO], [
+	dnl #
+	dnl # Linux 5.18 uses filemap_dirty_folio in lieu of
+	dnl # ___set_page_dirty_nobuffers
+	dnl #
+	AC_MSG_CHECKING([whether filemap_dirty_folio exists])
+	ZFS_LINUX_TEST_RESULT([vfs_has_filemap_dirty_folio], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_FILEMAP_DIRTY_FOLIO, 1,
+			[filemap_dirty_folio exists])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel-vfs-iov_iter.m4

@@ -134,6 +134,8 @@ AC_DEFUN([ZFS_AC_KERNEL_VFS_IOV_ITER], [
 		AC_DEFINE(HAVE_IOV_ITER_FAULT_IN_READABLE, 1,
 		    [iov_iter_fault_in_readable() is available])
 	],[
+		AC_MSG_RESULT(no)
+
 		AC_MSG_CHECKING([whether fault_in_iov_iter_readable() is available])
 		ZFS_LINUX_TEST_RESULT([fault_in_iov_iter_readable], [
 			AC_MSG_RESULT(yes)

config/kernel-vfs-read_folio.m4

@@ -0,0 +1,32 @@
+dnl #
+dnl # Linux 5.19 uses read_folio in lieu of readpage
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_VFS_READ_FOLIO], [
+	ZFS_LINUX_TEST_SRC([vfs_has_read_folio], [
+		#include <linux/fs.h>
+
+		static int
+		test_read_folio(struct file *file, struct folio *folio) {
+			(void) file; (void) folio;
+			return (0);
+		}
+
+		static const struct address_space_operations
+		    aops __attribute__ ((unused)) = {
+			.read_folio	= test_read_folio,
+		};
+	],[])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_VFS_READ_FOLIO], [
+	dnl #
+	dnl # Linux 5.19 uses read_folio in lieu of readpage
+	dnl #
+	AC_MSG_CHECKING([whether read_folio exists])
+	ZFS_LINUX_TEST_RESULT([vfs_has_read_folio], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE(HAVE_VFS_READ_FOLIO, 1, [read_folio exists])
+	],[
+		AC_MSG_RESULT([no])
+	])
+])

config/kernel-vfs-set_page_dirty.m4

@@ -23,7 +23,7 @@ AC_DEFUN([ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS], [
 	dnl # Linux 5.14 change requires set_page_dirty() to be assigned
 	dnl # in address_space_operations()
 	dnl #
-	AC_MSG_CHECKING([__set_page_dirty_nobuffers exists])
+	AC_MSG_CHECKING([whether __set_page_dirty_nobuffers exists])
 	ZFS_LINUX_TEST_RESULT([vfs_has_set_page_dirty_nobuffers], [
 		AC_MSG_RESULT([yes])
 		AC_DEFINE(HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS, 1,

config/kernel-xattr-handler.m4

@@ -100,6 +100,19 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_XATTR_HANDLER_GET], [
 			.get = get,
 		};
 	],[])
+
+	ZFS_LINUX_TEST_SRC([xattr_handler_get_dentry_inode_flags], [
+		#include <linux/xattr.h>
+
+		int get(const struct xattr_handler *handler,
+		    struct dentry *dentry, struct inode *inode,
+		    const char *name, void *buffer,
+		    size_t size, int flags) { return 0; }
+		static const struct xattr_handler
+		    xops __attribute__ ((unused)) = {
+			.get = get,
+		};
+	],[])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_GET], [
@@ -142,7 +155,21 @@ AC_DEFUN([ZFS_AC_KERNEL_XATTR_HANDLER_GET], [
 				AC_DEFINE(HAVE_XATTR_GET_DENTRY, 1,
 				    [xattr_handler->get() wants dentry])
 			],[
-				ZFS_LINUX_TEST_ERROR([xattr get()])
+				dnl #
+				dnl # Android API change,
+				dnl # The xattr_handler->get() callback was
+				dnl # changed to take dentry, inode and flags.
+				dnl #
+				AC_MSG_RESULT(no)
+				AC_MSG_CHECKING(
+				    [whether xattr_handler->get() wants dentry and inode and flags])
+				ZFS_LINUX_TEST_RESULT([xattr_handler_get_dentry_inode_flags], [
+					AC_MSG_RESULT(yes)
+					AC_DEFINE(HAVE_XATTR_GET_DENTRY_INODE_FLAGS, 1,
+					    [xattr_handler->get() wants dentry and inode and flags])
+				],[
+					ZFS_LINUX_TEST_ERROR([xattr get()])
+				])
 			])
 		])
 	])

config/kernel-zero_page.m4

@@ -0,0 +1,27 @@
+dnl #
+dnl # ZERO_PAGE() is an alias for emtpy_zero_page. On certain architectures
+dnl # this is a GPL exported variable.
+dnl #
+
+dnl #
+dnl # Checking if ZERO_PAGE is exported GPL-only
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_SRC_ZERO_PAGE], [
+	ZFS_LINUX_TEST_SRC([zero_page], [
+		#include <asm/pgtable.h>
+	], [
+		struct page *p __attribute__ ((unused));
+		p = ZERO_PAGE(0);
+	], [], [ZFS_META_LICENSE])
+])
+
+AC_DEFUN([ZFS_AC_KERNEL_ZERO_PAGE], [
+	AC_MSG_CHECKING([whether ZERO_PAGE() is GPL-only])
+	ZFS_LINUX_TEST_RESULT([zero_page_license], [
+		AC_MSG_RESULT(no)
+	], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_ZERO_PAGE_GPL_ONLY, 1,
+		    [ZERO_PAGE() is GPL-only])
+	])
+])

config/kernel.m4

@@ -8,8 +8,8 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 		ZFS_AC_QAT
 
 		dnl # Sanity checks for module building and CONFIG_* defines
-		ZFS_AC_KERNEL_TEST_MODULE
 		ZFS_AC_KERNEL_CONFIG_DEFINED
+		ZFS_AC_MODULE_SYMVERS
 
 		dnl # Sequential ZFS_LINUX_TRY_COMPILE tests
 		ZFS_AC_KERNEL_FPU_HEADER
@@ -61,6 +61,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_BIO
 	ZFS_AC_KERNEL_SRC_BLKDEV
 	ZFS_AC_KERNEL_SRC_BLK_QUEUE
+	ZFS_AC_KERNEL_SRC_GENHD_FLAGS
 	ZFS_AC_KERNEL_SRC_REVALIDATE_DISK
 	ZFS_AC_KERNEL_SRC_GET_DISK_RO
 	ZFS_AC_KERNEL_SRC_GENERIC_READLINK_GLOBAL
@@ -99,10 +100,14 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_SET_NLINK
 	ZFS_AC_KERNEL_SRC_SGET
 	ZFS_AC_KERNEL_SRC_LSEEK_EXECUTE
+	ZFS_AC_KERNEL_SRC_VFS_FILEMAP_DIRTY_FOLIO
+	ZFS_AC_KERNEL_SRC_VFS_READ_FOLIO
 	ZFS_AC_KERNEL_SRC_VFS_GETATTR
 	ZFS_AC_KERNEL_SRC_VFS_FSYNC_2ARGS
 	ZFS_AC_KERNEL_SRC_VFS_ITERATE
 	ZFS_AC_KERNEL_SRC_VFS_DIRECT_IO
+	ZFS_AC_KERNEL_SRC_VFS_READPAGES
+	ZFS_AC_KERNEL_SRC_VFS_SET_PAGE_DIRTY_NOBUFFERS
 	ZFS_AC_KERNEL_SRC_VFS_RW_ITERATE
 	ZFS_AC_KERNEL_SRC_VFS_GENERIC_WRITE_CHECKS
 	ZFS_AC_KERNEL_SRC_VFS_IOV_ITER
@@ -131,12 +136,14 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_BIO_MAX_SEGS
 	ZFS_AC_KERNEL_SRC_SIGNAL_STOP
 	ZFS_AC_KERNEL_SRC_SIGINFO
+	ZFS_AC_KERNEL_SRC_SYSFS
 	ZFS_AC_KERNEL_SRC_SET_SPECIAL_STATE
-	ZFS_AC_KERNEL_SRC_VFS_SET_PAGE_DIRTY_NOBUFFERS
 	ZFS_AC_KERNEL_SRC_STANDALONE_LINUX_STDARG
 	ZFS_AC_KERNEL_SRC_PAGEMAP_FOLIO_WAIT_BIT
 	ZFS_AC_KERNEL_SRC_ADD_DISK
 	ZFS_AC_KERNEL_SRC_KTHREAD
+	ZFS_AC_KERNEL_SRC_ZERO_PAGE
+	ZFS_AC_KERNEL_SRC___COPY_FROM_USER_INATOMIC
 
 	AC_MSG_CHECKING([for available kernel interfaces])
 	ZFS_LINUX_TEST_COMPILE_ALL([kabi])
@@ -171,6 +178,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_BIO
 	ZFS_AC_KERNEL_BLKDEV
 	ZFS_AC_KERNEL_BLK_QUEUE
+	ZFS_AC_KERNEL_GENHD_FLAGS
 	ZFS_AC_KERNEL_REVALIDATE_DISK
 	ZFS_AC_KERNEL_GET_DISK_RO
 	ZFS_AC_KERNEL_GENERIC_READLINK_GLOBAL
@@ -209,10 +217,14 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_SET_NLINK
 	ZFS_AC_KERNEL_SGET
 	ZFS_AC_KERNEL_LSEEK_EXECUTE
+	ZFS_AC_KERNEL_VFS_FILEMAP_DIRTY_FOLIO
+	ZFS_AC_KERNEL_VFS_READ_FOLIO
 	ZFS_AC_KERNEL_VFS_GETATTR
 	ZFS_AC_KERNEL_VFS_FSYNC_2ARGS
 	ZFS_AC_KERNEL_VFS_ITERATE
 	ZFS_AC_KERNEL_VFS_DIRECT_IO
+	ZFS_AC_KERNEL_VFS_READPAGES
+	ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS
 	ZFS_AC_KERNEL_VFS_RW_ITERATE
 	ZFS_AC_KERNEL_VFS_GENERIC_WRITE_CHECKS
 	ZFS_AC_KERNEL_VFS_IOV_ITER
@@ -241,12 +253,14 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_BIO_MAX_SEGS
 	ZFS_AC_KERNEL_SIGNAL_STOP
 	ZFS_AC_KERNEL_SIGINFO
+	ZFS_AC_KERNEL_SYSFS
 	ZFS_AC_KERNEL_SET_SPECIAL_STATE
-	ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS
 	ZFS_AC_KERNEL_STANDALONE_LINUX_STDARG
 	ZFS_AC_KERNEL_PAGEMAP_FOLIO_WAIT_BIT
 	ZFS_AC_KERNEL_ADD_DISK
 	ZFS_AC_KERNEL_KTHREAD
+	ZFS_AC_KERNEL_ZERO_PAGE
+	ZFS_AC_KERNEL___COPY_FROM_USER_INATOMIC
 ])
 
 dnl #
@@ -435,8 +449,6 @@ AC_DEFUN([ZFS_AC_KERNEL], [
 	AC_SUBST(LINUX)
 	AC_SUBST(LINUX_OBJ)
 	AC_SUBST(LINUX_VERSION)
-
-	ZFS_AC_MODULE_SYMVERS
 ])
 
 dnl #
@@ -531,27 +543,6 @@ AC_DEFUN([ZFS_AC_QAT], [
 	])
 ])
 
-dnl #
-dnl # Basic toolchain sanity check.
-dnl #
-AC_DEFUN([ZFS_AC_KERNEL_TEST_MODULE], [
-	AC_MSG_CHECKING([whether modules can be built])
-	ZFS_LINUX_TRY_COMPILE([], [], [
-		AC_MSG_RESULT([yes])
-	],[
-		AC_MSG_RESULT([no])
-		if test "x$enable_linux_builtin" != xyes; then
-			AC_MSG_ERROR([
-	*** Unable to build an empty module.
-			])
-		else
-			AC_MSG_ERROR([
-	*** Unable to build an empty module.
-	*** Please run 'make scripts' inside the kernel source tree.])
-		fi
-	])
-])
-
 dnl #
 dnl # ZFS_LINUX_CONFTEST_H
 dnl #
@@ -654,8 +645,10 @@ AC_DEFUN([ZFS_LINUX_COMPILE], [
 		build kernel modules with LLVM/CLANG toolchain])
 	AC_TRY_COMMAND([
 	    KBUILD_MODPOST_NOFINAL="$5" KBUILD_MODPOST_WARN="$6"
-	    make modules -k -j$TEST_JOBS ${KERNEL_CC:+CC=$KERNEL_CC} ${KERNEL_LD:+LD=$KERNEL_LD} ${KERNEL_LLVM:+LLVM=$KERNEL_LLVM} -C $LINUX_OBJ $ARCH_UM
-	    M=$PWD/$1 >$1/build.log 2>&1])
+	    make modules -k -j$TEST_JOBS ${KERNEL_CC:+CC=$KERNEL_CC}
+	    ${KERNEL_LD:+LD=$KERNEL_LD} ${KERNEL_LLVM:+LLVM=$KERNEL_LLVM}
+	    CONFIG_MODULES=y CFLAGS_MODULE=-DCONFIG_MODULES
+	    -C $LINUX_OBJ $ARCH_UM M=$PWD/$1 >$1/build.log 2>&1])
 	AS_IF([AC_TRY_COMMAND([$2])], [$3], [$4])
 ])
 

config/zfs-build.m4

@@ -209,8 +209,8 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS], [
 	AX_COUNT_CPUS([])
 	AC_SUBST(CPU_COUNT)
 
-	ZFS_AC_CONFIG_ALWAYS_CC_NO_UNUSED_BUT_SET_VARIABLE
-	ZFS_AC_CONFIG_ALWAYS_CC_NO_BOOL_COMPARE
+	ZFS_AC_CONFIG_ALWAYS_CC_NO_CLOBBERED
+	ZFS_AC_CONFIG_ALWAYS_CC_INFINITE_RECURSION
 	ZFS_AC_CONFIG_ALWAYS_CC_IMPLICIT_FALLTHROUGH
 	ZFS_AC_CONFIG_ALWAYS_CC_FRAME_LARGER_THAN
 	ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_TRUNCATION
@@ -226,6 +226,7 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS], [
 	ZFS_AC_CONFIG_ALWAYS_SED
 	ZFS_AC_CONFIG_ALWAYS_CPPCHECK
 	ZFS_AC_CONFIG_ALWAYS_SHELLCHECK
+	ZFS_AC_CONFIG_ALWAYS_PARALLEL
 ])
 
 AC_DEFUN([ZFS_AC_CONFIG], [
@@ -323,6 +324,10 @@ AC_DEFUN([ZFS_AC_RPM], [
 	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "$(DEBUG_KMEM_TRACKING_ZFS) 1"'
 	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "$(ASAN_ZFS) 1"'
 
+	AS_IF([test "x$enable_debuginfo" = xyes], [
+		RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "__strip /bin/true"'
+	])
+
 	RPM_DEFINE_UTIL=' --define "_initconfdir $(initconfdir)"'
 
 	dnl # Make the next three RPM_DEFINE_UTIL additions conditional, since

contrib/dracut/90zfs/export-zfs.sh.in

@@ -1,14 +1,12 @@
 #!/bin/sh
 
-. /lib/dracut-zfs-lib.sh
-
 _do_zpool_export() {
 	ret=0
 	errs=""
 	final="${1}"
 
 	info "ZFS: Exporting ZFS storage pools..."
-	errs=$(export_all -F 2>&1)
+	errs=$(zpool export -aF 2>&1)
 	ret=$?
 	[ -z "${errs}" ] || echo "${errs}" | vwarn
 	if [ "x${ret}" != "x0" ]; then

contrib/dracut/90zfs/module-setup.sh.in

@@ -6,8 +6,8 @@ check() {
 	[ "${1}" = "-d" ] && return 0
 
 	# Verify the zfs tool chain
-	for tool in "@sbindir@/zgenhostid" "@sbindir@/zpool" "@sbindir@/zfs" "@mounthelperdir@/mount.zfs" ; do
-		test -x "$tool" || return 1
+	for tool in "zgenhostid" "zpool" "zfs" "mount.zfs"; do
+		command -v "${tool}" >/dev/null || return 1
 	done
 
 	return 0
@@ -19,125 +19,85 @@ depends() {
 }
 
 installkernel() {
-	instmods zfs
-	instmods zcommon
-	instmods znvpair
-	instmods zavl
-	instmods zunicode
-	instmods zlua
-	instmods icp
-	instmods spl
-	instmods zlib_deflate
-	instmods zlib_inflate
+	instmods -c zfs
 }
 
 install() {
-	inst_rules @udevruledir@/90-zfs.rules
-	inst_rules @udevruledir@/69-vdev.rules
-	inst_rules @udevruledir@/60-zvol.rules
-	dracut_install hostid
-	dracut_install grep
-	dracut_install @sbindir@/zgenhostid
-	dracut_install @sbindir@/zfs
-	dracut_install @sbindir@/zpool
-	# Workaround for https://github.com/openzfs/zfs/issues/4749 by
-	# ensuring libgcc_s.so(.1) is included
-	if ldd @sbindir@/zpool | grep -qF 'libgcc_s.so'; then
-		# Dracut will have already tracked and included it
-		:;
-	elif command -v gcc-config >/dev/null 2>&1; then
-		# On systems with gcc-config (Gentoo, Funtoo, etc.):
-		# Use the current profile to resolve the appropriate path
-		s="$(gcc-config -c)"
-		dracut_install "/usr/lib/gcc/${s%-*}/${s##*-}/libgcc_s.so"*
-	elif [ "$(echo /usr/lib/libgcc_s.so*)" != "/usr/lib/libgcc_s.so*" ]; then
-		# Try a simple path first
-		dracut_install /usr/lib/libgcc_s.so*
-	elif [ "$(echo /lib*/libgcc_s.so*)" != "/lib*/libgcc_s.so*" ]; then
-		# SUSE
-		dracut_install /lib*/libgcc_s.so*
-	else
-		# Fallback: Guess the path and include all matches
-		dracut_install /usr/lib*/gcc/**/libgcc_s.so*
+	inst_rules 90-zfs.rules 69-vdev.rules 60-zvol.rules
+
+	inst_multiple \
+		zgenhostid \
+		zfs \
+		zpool \
+		mount.zfs \
+		hostid \
+		grep \
+		awk \
+		tr \
+		cut \
+		head ||
+		{ dfatal "Failed to install essential binaries"; exit 1; }
+
+	# Adapted from https://github.com/zbm-dev/zfsbootmenu
+	if ! ldd "$(command -v zpool)" | grep -qF 'libgcc_s.so'; then
+		# On systems with gcc-config (Gentoo, Funtoo, etc.), use it to find libgcc_s
+		if command -v gcc-config >/dev/null; then
+			inst_simple "/usr/lib/gcc/$(s=$(gcc-config -c); echo "${s%-*}/${s##*-}")/libgcc_s.so.1" ||
+				{ dfatal "Unable to install libgcc_s.so"; exit 1; }
+			# Otherwise, use dracut's library installation function to find the right one
+		elif ! inst_libdir_file "libgcc_s.so*"; then
+			# If all else fails, just try looking for some gcc arch directory
+			inst_simple /usr/lib/gcc/*/*/libgcc_s.so* ||
+				{ dfatal "Unable to install libgcc_s.so"; exit 1; }
+		fi
 	fi
-	# shellcheck disable=SC2050
-	if [ @LIBFETCH_DYNAMIC@ -gt 0 ]; then
-		for d in $libdirs; do
-			[ -e "$d/@LIBFETCH_SONAME@" ] && dracut_install "$d/@LIBFETCH_SONAME@"
-		done
-	fi
-	dracut_install @mounthelperdir@/mount.zfs
-	dracut_install @udevdir@/vdev_id
-	dracut_install awk
-	dracut_install cut
-	dracut_install tr
-	dracut_install head
-	dracut_install @udevdir@/zvol_id
+
 	inst_hook cmdline 95 "${moddir}/parse-zfs.sh"
-	if [ -n "$systemdutildir" ] ; then
-		inst_script "${moddir}/zfs-generator.sh" "$systemdutildir"/system-generators/dracut-zfs-generator
+	if [ -n "${systemdutildir}" ]; then
+		inst_script "${moddir}/zfs-generator.sh" "${systemdutildir}/system-generators/dracut-zfs-generator"
 	fi
 	inst_hook pre-mount 90 "${moddir}/zfs-load-key.sh"
 	inst_hook mount 98 "${moddir}/mount-zfs.sh"
 	inst_hook cleanup 99 "${moddir}/zfs-needshutdown.sh"
 	inst_hook shutdown 20 "${moddir}/export-zfs.sh"
 
-	inst_simple "${moddir}/zfs-lib.sh" "/lib/dracut-zfs-lib.sh"
-	if [ -e @sysconfdir@/zfs/zpool.cache ]; then
-		inst @sysconfdir@/zfs/zpool.cache
-		type mark_hostonly >/dev/null 2>&1 && mark_hostonly @sysconfdir@/zfs/zpool.cache
-	fi
+	inst_script "${moddir}/zfs-lib.sh" "/lib/dracut-zfs-lib.sh"
 
-	if [ -e @sysconfdir@/zfs/vdev_id.conf ]; then
-		inst @sysconfdir@/zfs/vdev_id.conf
-		type mark_hostonly >/dev/null 2>&1 && mark_hostonly @sysconfdir@/zfs/vdev_id.conf
-	fi
+	# -H ensures they are marked host-only
+	# -o ensures there is no error upon absence of these files
+	inst_multiple -o -H \
+		"@sysconfdir@/zfs/zpool.cache" \
+		"@sysconfdir@/zfs/vdev_id.conf"
 
 	# Synchronize initramfs and system hostid
-	if [ -f @sysconfdir@/hostid ]; then
-		inst @sysconfdir@/hostid
-		type mark_hostonly >/dev/null 2>&1 && mark_hostonly @sysconfdir@/hostid
-	elif HOSTID="$(hostid 2>/dev/null)" && [ "${HOSTID}" != "00000000" ]; then
-		zgenhostid -o "${initdir}@sysconfdir@/hostid" "${HOSTID}"
-		type mark_hostonly >/dev/null 2>&1 && mark_hostonly @sysconfdir@/hostid
+	if ! inst_simple -H @sysconfdir@/hostid; then
+		if HOSTID="$(hostid 2>/dev/null)" && [ "${HOSTID}" != "00000000" ]; then
+			zgenhostid -o "${initdir}@sysconfdir@/hostid" "${HOSTID}"
+			mark_hostonly @sysconfdir@/hostid
+		fi
 	fi
 
 	if dracut_module_included "systemd"; then
-		mkdir -p "${initdir}/$systemdsystemunitdir/zfs-import.target.wants"
-		for _service in "zfs-import-scan.service" "zfs-import-cache.service" ; do
-			dracut_install "@systemdunitdir@/$_service"
-			if ! [ -L "${initdir}/$systemdsystemunitdir/zfs-import.target.wants/$_service" ]; then
-				ln -sf ../$_service "${initdir}/$systemdsystemunitdir/zfs-import.target.wants/$_service"
-				type mark_hostonly >/dev/null 2>&1 && mark_hostonly "@systemdunitdir@/$_service"
-			fi
+		inst_simple "${systemdsystemunitdir}/zfs-import.target"
+		systemctl -q --root "${initdir}" add-wants initrd.target zfs-import.target
+
+		inst_simple "${moddir}/zfs-env-bootfs.service" "${systemdsystemunitdir}/zfs-env-bootfs.service"
+		systemctl -q --root "${initdir}" add-wants zfs-import.target zfs-env-bootfs.service
+
+		for _service in \
+			"zfs-import-scan.service" \
+			"zfs-import-cache.service"; do
+			inst_simple "${systemdsystemunitdir}/${_service}"
+			systemctl -q --root "${initdir}" add-wants zfs-import.target "${_service}"
 		done
 
-		inst "${moddir}"/zfs-env-bootfs.service "${systemdsystemunitdir}"/zfs-env-bootfs.service
-		ln -s ../zfs-env-bootfs.service "${initdir}/${systemdsystemunitdir}/zfs-import.target.wants"/zfs-env-bootfs.service
-		type mark_hostonly >/dev/null 2>&1 && mark_hostonly @systemdunitdir@/zfs-env-bootfs.service
-
-		dracut_install systemd-ask-password
-		dracut_install systemd-tty-ask-password-agent
-
-		mkdir -p "${initdir}/$systemdsystemunitdir/initrd.target.wants"
-		dracut_install @systemdunitdir@/zfs-import.target
-		if ! [ -L "${initdir}/$systemdsystemunitdir/initrd.target.wants"/zfs-import.target ]; then
-			ln -s ../zfs-import.target "${initdir}/$systemdsystemunitdir/initrd.target.wants"/zfs-import.target
-			type mark_hostonly >/dev/null 2>&1 && mark_hostonly @systemdunitdir@/zfs-import.target
-		fi
-
-		for _service in zfs-snapshot-bootfs.service zfs-rollback-bootfs.service ; do
-			inst "${moddir}/$_service" "${systemdsystemunitdir}/$_service"
-			if ! [ -L "${initdir}/$systemdsystemunitdir/initrd.target.wants/$_service" ]; then
-				ln -s "../$_service" "${initdir}/$systemdsystemunitdir/initrd.target.wants/$_service"
-			fi
+		for _service in \
+			"zfs-snapshot-bootfs.service" \
+			"zfs-rollback-bootfs.service"; do
+			inst_simple "${moddir}/${_service}" "${systemdsystemunitdir}/${_service}"
+			systemctl -q --root "${initdir}" add-wants initrd.target "${_service}"
 		done
 
-		# There isn't a pkg-config variable for this,
-		# and dracut doesn't automatically resolve anything this'd be next to
-		local systemdsystemenvironmentgeneratordir
-		systemdsystemenvironmentgeneratordir="$(pkg-config --variable=prefix systemd || echo "/usr")/lib/systemd/system-environment-generators"
-		mkdir -p "${initdir}/${systemdsystemenvironmentgeneratordir}"
-		inst "${moddir}"/import-opts-generator.sh "${systemdsystemenvironmentgeneratordir}"/zfs-import-opts.sh
+		inst_simple "${moddir}/import-opts-generator.sh" "${systemdutildir}/system-environment-generators/zfs-import-opts.sh"
 	fi
 }

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

@@ -3,48 +3,73 @@
 
 . /lib/dracut-zfs-lib.sh
 
-ZFS_DATASET=""
-ZFS_POOL=""
-
-case "${root}" in
-	zfs:*) ;;
-	*) return ;;
-esac
+decode_root_args || return 0
 
 GENERATOR_FILE=/run/systemd/generator/sysroot.mount
 GENERATOR_EXTENSION=/run/systemd/generator/sysroot.mount.d/zfs-enhancement.conf
 
-if [ -e "$GENERATOR_FILE" ] && [ -e "$GENERATOR_EXTENSION" ] ; then
-	# If the ZFS sysroot.mount flag exists, the initial RAM disk configured
-	# it to mount ZFS on root.  In that case, we bail early.  This flag
-	# file gets created by the zfs-generator program upon successful run.
-	info "ZFS: There is a sysroot.mount and zfs-generator has extended it."
-	info "ZFS: Delegating root mount to sysroot.mount."
-	# Let us tell the initrd to run on shutdown.
-	# We have a shutdown hook to run
-	# because we imported the pool.
+if [ -e "$GENERATOR_FILE" ] && [ -e "$GENERATOR_EXTENSION" ]; then
+	# We're under systemd and dracut-zfs-generator ran to completion.
+	info "ZFS: Delegating root mount to sysroot.mount at al."
+
 	# We now prevent Dracut from running this thing again.
-	for zfsmounthook in "$hookdir"/mount/*zfs* ; do
-		if [ -f "$zfsmounthook" ] ; then
-			rm -f "$zfsmounthook"
-		fi
-	done
+	rm -f "$hookdir"/mount/*zfs*
 	return
 fi
+
 info "ZFS: No sysroot.mount exists or zfs-generator did not extend it."
 info "ZFS: Mounting root with the traditional mount-zfs.sh instead."
 
+# ask_for_password tries prompt cmd
+#
+# Wraps around plymouth ask-for-password and adds fallback to tty password ask
+# if plymouth is not present.
+ask_for_password() {
+    tries="$1"
+    prompt="$2"
+    cmd="$3"
+
+    {
+        flock -s 9
+
+        # Prompt for password with plymouth, if installed and running.
+        if plymouth --ping 2>/dev/null; then
+            plymouth ask-for-password \
+                --prompt "$prompt" --number-of-tries="$tries" | \
+                eval "$cmd"
+            ret=$?
+        else
+            i=1
+            while [ "$i" -le "$tries" ]; do
+                printf "%s [%i/%i]:" "$prompt" "$i" "$tries" >&2
+                eval "$cmd" && ret=0 && break
+                ret=$?
+                i=$((i+1))
+                printf '\n' >&2
+            done
+            unset i
+        fi
+    } 9>/.console_lock
+
+    [ "$ret" -ne 0 ] && echo "Wrong password" >&2
+    return "$ret"
+}
+
+
 # Delay until all required block devices are present.
 modprobe zfs 2>/dev/null
 udevadm settle
 
+ZFS_DATASET=
+ZFS_POOL=
+
 if [ "${root}" = "zfs:AUTO" ] ; then
-	if ! ZFS_DATASET="$(find_bootfs)" ; then
+	if ! ZFS_DATASET="$(zpool get -Ho value bootfs | grep -m1 -vFx -)"; then
 		# shellcheck disable=SC2086
 		zpool import -N -a ${ZPOOL_IMPORT_OPTS}
-		if ! ZFS_DATASET="$(find_bootfs)" ; then
+		if ! ZFS_DATASET="$(zpool get -Ho value bootfs | grep -m1 -vFx -)"; then
 			warn "ZFS: No bootfs attribute found in importable pools."
-			export_all -F
+			zpool export -aF
 
 			rootok=0
 			return 1
@@ -53,34 +78,43 @@ if [ "${root}" = "zfs:AUTO" ] ; then
 	info "ZFS: Using ${ZFS_DATASET} as root."
 fi
 
-ZFS_DATASET="${ZFS_DATASET:-${root#zfs:}}"
+ZFS_DATASET="${ZFS_DATASET:-${root}}"
 ZFS_POOL="${ZFS_DATASET%%/*}"
 
-if import_pool "${ZFS_POOL}" ; then
-	# Load keys if we can or if we need to
-	if [ "$(zpool list -H -o feature@encryption "${ZFS_POOL}")" = 'active' ]; then
-		# if the root dataset has encryption enabled
-		ENCRYPTIONROOT="$(zfs get -H -o value encryptionroot "${ZFS_DATASET}")"
-		if ! [ "${ENCRYPTIONROOT}" = "-" ]; then
-			KEYSTATUS="$(zfs get -H -o value keystatus "${ENCRYPTIONROOT}")"
-			# if the key needs to be loaded
-			if [ "$KEYSTATUS" = "unavailable" ]; then
-				# decrypt them
-				ask_for_password \
-					--tries 5 \
-					--prompt "Encrypted ZFS password for ${ENCRYPTIONROOT}: " \
-					--cmd "zfs load-key '${ENCRYPTIONROOT}'"
-			fi
+
+if ! zpool get -Ho name "${ZFS_POOL}" > /dev/null 2>&1; then
+    info "ZFS: Importing pool ${ZFS_POOL}..."
+    # shellcheck disable=SC2086
+    if ! zpool import -N ${ZPOOL_IMPORT_OPTS} "${ZFS_POOL}"; then
+        warn "ZFS: Unable to import pool ${ZFS_POOL}"
+        rootok=0
+        return 1
+    fi
+fi
+
+# Load keys if we can or if we need to
+# TODO: for_relevant_root_children like in zfs-load-key.sh.in
+if [ "$(zpool get -Ho value feature@encryption "${ZFS_POOL}")" = 'active' ]; then
+	# if the root dataset has encryption enabled
+	ENCRYPTIONROOT="$(zfs get -Ho value encryptionroot "${ZFS_DATASET}")"
+	if ! [ "${ENCRYPTIONROOT}" = "-" ]; then
+		KEYSTATUS="$(zfs get -Ho value keystatus "${ENCRYPTIONROOT}")"
+		# if the key needs to be loaded
+		if [ "$KEYSTATUS" = "unavailable" ]; then
+			# decrypt them
+			ask_for_password \
+				5 \
+				"Encrypted ZFS password for ${ENCRYPTIONROOT}: " \
+				"zfs load-key '${ENCRYPTIONROOT}'"
 		fi
 	fi
-	# Let us tell the initrd to run on shutdown.
-	# We have a shutdown hook to run
-	# because we imported the pool.
-	info "ZFS: Mounting dataset ${ZFS_DATASET}..."
-	if mount_dataset "${ZFS_DATASET}" ; then
-		ROOTFS_MOUNTED=yes
-		return 0
-	fi
 fi
 
-rootok=0
+# Let us tell the initrd to run on shutdown.
+# We have a shutdown hook to run
+# because we imported the pool.
+info "ZFS: Mounting dataset ${ZFS_DATASET}..."
+if ! mount_dataset "${ZFS_DATASET}"; then
+  rootok=0
+  return 1
+fi

contrib/dracut/90zfs/parse-zfs.sh.in

@@ -1,7 +1,8 @@
 #!/bin/sh
 # shellcheck disable=SC2034,SC2154
 
-. /lib/dracut-lib.sh
+# shellcheck source=zfs-lib.sh.in
+. /lib/dracut-zfs-lib.sh
 
 # Let the command line override our host id.
 spl_hostid=$(getarg spl_hostid=)
@@ -15,49 +16,20 @@ else
 	warn "ZFS: Pools may not import correctly."
 fi
 
-wait_for_zfs=0
-case "${root}" in
-	""|zfs|zfs:)
-		# We'll take root unset, root=zfs, or root=zfs:
-		# No root set, so we want to read the bootfs attribute.  We
-		# can't do that until udev settles so we'll set dummy values
-		# and hope for the best later on.
-		root="zfs:AUTO"
-		rootok=1
-		wait_for_zfs=1
+if decode_root_args; then
+	if [ "$root" = "zfs:AUTO" ]; then
+		info "ZFS: Boot dataset autodetected from bootfs=."
+	else
+		info "ZFS: Boot dataset is ${root}."
+	fi
 
-		info "ZFS: Enabling autodetection of bootfs after udev settles."
-		;;
-
-	ZFS=*|zfs:*|FILESYSTEM=*)
-		# root is explicit ZFS root.  Parse it now.  We can handle
-		# a root=... param in any of the following formats:
-		# root=ZFS=rpool/ROOT
-		# root=zfs:rpool/ROOT
-		# root=zfs:FILESYSTEM=rpool/ROOT
-		# root=FILESYSTEM=rpool/ROOT
-		# root=ZFS=pool+with+space/ROOT+WITH+SPACE (translates to root=ZFS=pool with space/ROOT WITH SPACE)
-
-		# Strip down to just the pool/fs
-		root="${root#zfs:}"
-		root="${root#FILESYSTEM=}"
-		root="zfs:${root#ZFS=}"
-		# switch + with spaces because kernel cmdline does not allow us to quote parameters
-		root=$(echo "$root" | tr '+' ' ')
-		rootok=1
-		wait_for_zfs=1
-
-		info "ZFS: Set ${root} as bootfs."
-		;;
-esac
-
-# Make sure Dracut is happy that we have a root and will wait for ZFS
-# modules to settle before mounting.
-if [ ${wait_for_zfs} -eq 1 ]; then
-	ln -s /dev/null /dev/root 2>/dev/null
-	initqueuedir="${hookdir}/initqueue/finished"
-	test -d "${initqueuedir}" || {
-		initqueuedir="${hookdir}/initqueue-finished"
-	}
-	echo '[ -e /dev/zfs ]' > "${initqueuedir}/zfs.sh"
+	rootok=1
+	# Make sure Dracut is happy that we have a root and will wait for ZFS
+	# modules to settle before mounting.
+	if [ -n "${wait_for_zfs}" ]; then
+		ln -s null /dev/root
+		echo '[ -e /dev/zfs ]' > "${hookdir}/initqueue/finished/zfs.sh"
+	fi
+else
+	info "ZFS: no ZFS-on-root."
 fi

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

@@ -8,7 +8,7 @@ Before=zfs-import.target
 
 [Service]
 Type=oneshot
-ExecStart=/bin/sh -c "exec systemctl set-environment BOOTFS=$(@sbindir@/zpool list -H -o bootfs | grep -m1 -v '^-$')"
+ExecStart=/bin/sh -c "exec systemctl set-environment BOOTFS=$(@sbindir@/zpool list -H -o bootfs | grep -m1 -vFx -)"
 
 [Install]
 WantedBy=zfs-import.target

contrib/dracut/90zfs/zfs-generator.sh.in

@@ -1,5 +1,5 @@
 #!/bin/sh
-# shellcheck disable=SC2016,SC1004
+# shellcheck disable=SC2016,SC1004,SC2154
 
 grep -wq debug /proc/cmdline && debug=1
 [ -n "$debug" ] && echo "zfs-generator: starting" >> /dev/kmsg
@@ -10,37 +10,17 @@ GENERATOR_DIR="$1"
     exit 1
 }
 
-[ -f /lib/dracut-lib.sh ] && dracutlib=/lib/dracut-lib.sh
-[ -f /usr/lib/dracut/modules.d/99base/dracut-lib.sh ] && dracutlib=/usr/lib/dracut/modules.d/99base/dracut-lib.sh
-command -v getarg >/dev/null 2>&1 || {
-    [ -n "$debug" ] && echo "zfs-generator: loading Dracut library from $dracutlib" >> /dev/kmsg
-    . "$dracutlib"
-}
-
+# shellcheck source=zfs-lib.sh.in
 . /lib/dracut-zfs-lib.sh
+decode_root_args || exit 0
 
-[ -z "$root" ]       && root=$(getarg root=)
-[ -z "$rootfstype" ] && rootfstype=$(getarg rootfstype=)
-[ -z "$rootflags" ]  && rootflags=$(getarg rootflags=)
-
-# If root is not ZFS= or zfs: or rootfstype is not zfs
-# then we are not supposed to handle it.
-[ "${root##zfs:}" = "${root}" ] &&
-	[ "${root##ZFS=}" = "${root}" ] &&
-	[ "$rootfstype" != "zfs" ] &&
-	exit 0
-
+[ -z "${rootflags}" ] && rootflags=$(getarg rootflags=)
 case ",${rootflags}," in
 	*,zfsutil,*) ;;
 	,,)	rootflags=zfsutil ;;
 	*)	rootflags="zfsutil,${rootflags}" ;;
 esac
 
-if [ "${root}" != "zfs:AUTO" ]; then
-  root="${root##zfs:}"
-  root="${root##ZFS=}"
-fi
-
 [ -n "$debug" ] && echo "zfs-generator: writing extension for sysroot.mount to $GENERATOR_DIR/sysroot.mount.d/zfs-enhancement.conf" >> /dev/kmsg
 
 
@@ -89,7 +69,7 @@ else
   _zfs_generator_cb() {
       dset="${1}"
       mpnt="${2}"
-      unit="sysroot$(echo "$mpnt" | tr '/' '-').mount"
+      unit="$(systemd-escape --suffix=mount -p "/sysroot${mpnt}")"
 
       {
           echo "[Unit]"

contrib/dracut/90zfs/zfs-lib.sh.in

@@ -1,74 +1,16 @@
 #!/bin/sh
+# shellcheck disable=SC2034
 
-command -v getarg >/dev/null || . /lib/dracut-lib.sh
-command -v getargbool >/dev/null || {
-    # Compatibility with older Dracut versions.
-    # With apologies to the Dracut developers.
-    getargbool() {
-        _default="$1"; shift
-        ! _b=$(getarg "$@") && [ -z "$_b" ] && _b="$_default"
-        if [ -n "$_b" ]; then
-            [ "$_b" = "0" ] && return 1
-            [ "$_b" = "no" ] && return 1
-            [ "$_b" = "off" ] && return 1
-        fi
-        return 0
-    }
-}
+command -v getarg >/dev/null || . /lib/dracut-lib.sh || . /usr/lib/dracut/modules.d/99base/dracut-lib.sh
 
-OLDIFS="${IFS}"
-NEWLINE="
-"
 TAB="	"
 
-ZPOOL_IMPORT_OPTS=""
-if getargbool 0 zfs_force -y zfs.force -y zfsforce ; then
+ZPOOL_IMPORT_OPTS=
+if getargbool 0 zfs_force -y zfs.force -y zfsforce; then
     warn "ZFS: Will force-import pools if necessary."
-    ZPOOL_IMPORT_OPTS="${ZPOOL_IMPORT_OPTS} -f"
+    ZPOOL_IMPORT_OPTS=-f
 fi
 
-# find_bootfs
-#   returns the first dataset with the bootfs attribute.
-find_bootfs() {
-    IFS="${NEWLINE}"
-    for dataset in $(zpool list -H -o bootfs); do
-        case "${dataset}" in
-            "" | "-")
-                continue
-                ;;
-            "no pools available")
-                IFS="${OLDIFS}"
-                return 1
-                ;;
-            *)
-                IFS="${OLDIFS}"
-                echo "${dataset}"
-                return 0
-                ;;
-        esac
-    done
-
-    IFS="${OLDIFS}"
-    return 1
-}
-
-# import_pool POOL
-#   imports the given zfs pool if it isn't imported already.
-import_pool() {
-    pool="${1}"
-
-    if ! zpool list -H "${pool}" > /dev/null 2>&1; then
-        info "ZFS: Importing pool ${pool}..."
-        # shellcheck disable=SC2086
-        if ! zpool import -N ${ZPOOL_IMPORT_OPTS} "${pool}" ; then
-            warn "ZFS: Unable to import pool ${pool}"
-            return 1
-        fi
-    fi
-
-    return 0
-}
-
 _mount_dataset_cb() {
     mount -o zfsutil -t zfs "${1}" "${NEWROOT}${2}"
 }
@@ -121,87 +63,57 @@ for_relevant_root_children() {
         )
 }
 
-# export_all OPTS
-#   exports all imported zfs pools.
-export_all() {
-    ret=0
-
-    IFS="${NEWLINE}"
-    for pool in $(zpool list -H -o name) ; do
-        if zpool list -H "${pool}" > /dev/null 2>&1; then
-            zpool export "${pool}" "$@" || ret=$?
-        fi
-    done
-    IFS="${OLDIFS}"
-
-    return ${ret}
-}
-
-# ask_for_password
+# Parse root=, rootfstype=, return them decoded and normalised to zfs:AUTO for auto, plain dset for explicit
 #
-# Wraps around plymouth ask-for-password and adds fallback to tty password ask
-# if plymouth is not present.
+# True if ZFS-on-root, false if we shouldn't
 #
-# --cmd command
-#   Command to execute. Required.
-# --prompt prompt
-#   Password prompt. Note that function already adds ':' at the end.
-#   Recommended.
-# --tries n
-#   How many times repeat command on its failure.  Default is 3.
-# --ply-[cmd|prompt|tries]
-#   Command/prompt/tries specific for plymouth password ask only.
-# --tty-[cmd|prompt|tries]
-#   Command/prompt/tries specific for tty password ask only.
-# --tty-echo-off
-#   Turn off input echo before tty command is executed and turn on after.
-#   It's useful when password is read from stdin.
-ask_for_password() {
-    ply_tries=3
-    tty_tries=3
-    while [ "$#" -gt 0 ]; do
-        case "$1" in
-            --cmd) ply_cmd="$2"; tty_cmd="$2"; shift;;
-            --ply-cmd) ply_cmd="$2"; shift;;
-            --tty-cmd) tty_cmd="$2"; shift;;
-            --prompt) ply_prompt="$2"; tty_prompt="$2"; shift;;
-            --ply-prompt) ply_prompt="$2"; shift;;
-            --tty-prompt) tty_prompt="$2"; shift;;
-            --tries) ply_tries="$2"; tty_tries="$2"; shift;;
-            --ply-tries) ply_tries="$2"; shift;;
-            --tty-tries) tty_tries="$2"; shift;;
-            --tty-echo-off) tty_echo_off=yes;;
+# Supported values:
+#   root=
+#   root=zfs
+#   root=zfs:
+#   root=zfs:AUTO
+#
+#   root=ZFS=data/set
+#   root=zfs:data/set
+#   root=zfs:ZFS=data/set (as a side-effect; allowed but undocumented)
+#
+#   rootfstype=zfs AND root=data/set <=> root=data/set
+#   rootfstype=zfs AND root=         <=> root=zfs:AUTO
+#
+# '+'es in explicit dataset decoded to ' 's.
+decode_root_args() {
+    if [ -n "$rootfstype" ]; then
+        [ "$rootfstype" = zfs ]
+        return
+    fi
+
+    root=$(getarg root=)
+    rootfstype=$(getarg rootfstype=)
+
+    # shellcheck disable=SC2249
+    case "$root" in
+        ""|zfs|zfs:|zfs:AUTO)
+            root=zfs:AUTO
+            rootfstype=zfs
+            return 0
+            ;;
+
+        ZFS=*|zfs:*)
+            root="${root#zfs:}"
+            root="${root#ZFS=}"
+            root=$(echo "$root" | tr '+' ' ')
+            rootfstype=zfs
+            return 0
+            ;;
+    esac
+
+    if [ "$rootfstype" = "zfs" ]; then
+        case "$root" in
+            "") root=zfs:AUTO ;;
+            *)  root=$(echo "$root" | tr '+' ' ') ;;
         esac
-        shift
-    done
+        return 0
+    fi
 
-    { flock -s 9;
-        # Prompt for password with plymouth, if installed and running.
-        if plymouth --ping 2>/dev/null; then
-            plymouth ask-for-password \
-                --prompt "$ply_prompt" --number-of-tries="$ply_tries" | \
-                eval "$ply_cmd"
-            ret=$?
-        else
-            if [ "$tty_echo_off" = yes ]; then
-                stty_orig="$(stty -g)"
-                stty -echo
-            fi
-
-            i=1
-            while [ "$i" -le "$tty_tries" ]; do
-                [ -n "$tty_prompt" ] && \
-                    printf "%s [%i/%i]:" "$tty_prompt" "$i" "$tty_tries" >&2
-                eval "$tty_cmd" && ret=0 && break
-                ret=$?
-                i=$((i+1))
-                [ -n "$tty_prompt" ] && printf '\n' >&2
-            done
-            unset i
-            [ "$tty_echo_off" = yes ] && stty "$stty_orig"
-        fi
-    } 9>/.console_lock
-
-    [ $ret -ne 0 ] && echo "Wrong password" >&2
-    return $ret
+    return 1
 }

contrib/dracut/90zfs/zfs-load-key.sh.in

@@ -4,70 +4,61 @@
 # only run this on systemd systems, we handle the decrypt in mount-zfs.sh in the mount hook otherwise
 [ -e /bin/systemctl ] || [ -e /usr/bin/systemctl ] || return 0
 
-# This script only gets executed on systemd systems, see mount-zfs.sh for non-systemd systems
+# shellcheck source=zfs-lib.sh.in
+. /lib/dracut-zfs-lib.sh
 
-# import the libs now that we know the pool imported
-[ -f /lib/dracut-lib.sh ] && dracutlib=/lib/dracut-lib.sh
-[ -f /usr/lib/dracut/modules.d/99base/dracut-lib.sh ] && dracutlib=/usr/lib/dracut/modules.d/99base/dracut-lib.sh
-# shellcheck source=./lib-zfs.sh.in
-. "$dracutlib"
-
-# load the kernel command line vars
-[ -z "$root" ] && root="$(getarg root=)"
-# If root is not ZFS= or zfs: or rootfstype is not zfs then we are not supposed to handle it.
-[ "${root##zfs:}" = "${root}" ] && [ "${root##ZFS=}" = "${root}" ] && [ "$rootfstype" != "zfs" ] && exit 0
+decode_root_args || return 0
 
 # There is a race between the zpool import and the pre-mount hooks, so we wait for a pool to be imported
-while [ "$(zpool list -H)" = "" ]; do
-    systemctl is-failed --quiet zfs-import-cache.service zfs-import-scan.service && exit 1
+while ! systemctl is-active --quiet zfs-import.target; do
+    systemctl is-failed --quiet zfs-import-cache.service zfs-import-scan.service && return 1
     sleep 0.1s
 done
 
-# run this after import as zfs-import-cache/scan service is confirmed good
-# we do not overwrite the ${root} variable, but create a new one, BOOTFS, to hold the dataset
-if [ "${root}" = "zfs:AUTO" ] ; then
-    BOOTFS="$(zpool list -H -o bootfs | awk '$1 != "-" {print; exit}')"
-else
-    BOOTFS="${root##zfs:}"
-    BOOTFS="${BOOTFS##ZFS=}"
+BOOTFS="$root"
+if [ "$BOOTFS" = "zfs:AUTO" ]; then
+    BOOTFS="$(zpool get -Ho value bootfs | grep -m1 -vFx -)"
 fi
 
-# if pool encryption is active and the zfs command understands '-o encryption'
-if [ "$(zpool list -H -o feature@encryption "${BOOTFS%%/*}")" = 'active' ]; then
-    # if the root dataset has encryption enabled
-    ENCRYPTIONROOT="$(zfs get -H -o value encryptionroot "${BOOTFS}")"
-    if ! [ "${ENCRYPTIONROOT}" = "-" ]; then
-        KEYSTATUS="$(zfs get -H -o value keystatus "${ENCRYPTIONROOT}")"
-        # continue only if the key needs to be loaded
-        [ "$KEYSTATUS" = "unavailable" ] || exit 0
+[ "$(zpool get -Ho value feature@encryption "${BOOTFS%%/*}")" = 'active' ] || return 0
 
-        KEYLOCATION="$(zfs get -H -o value keylocation "${ENCRYPTIONROOT}")"
-        case "${KEYLOCATION%%://*}" in
-            prompt)
-                for _ in 1 2 3; do
-                    systemd-ask-password --no-tty "Encrypted ZFS password for ${BOOTFS}" | zfs load-key "${ENCRYPTIONROOT}" && break
+_load_key_cb() {
+    dataset="$1"
+
+    ENCRYPTIONROOT="$(zfs get -Ho value encryptionroot "${dataset}")"
+    [ "${ENCRYPTIONROOT}" = "-" ] && return 0
+
+    [ "$(zfs get -Ho value keystatus "${ENCRYPTIONROOT}")" = "unavailable" ] || return 0
+
+    KEYLOCATION="$(zfs get -Ho value keylocation "${ENCRYPTIONROOT}")"
+    case "${KEYLOCATION%%://*}" in
+        prompt)
+            for _ in 1 2 3; do
+                systemd-ask-password --no-tty "Encrypted ZFS password for ${dataset}" | zfs load-key "${ENCRYPTIONROOT}" && break
+            done
+            ;;
+        http*)
+            systemctl start network-online.target
+            zfs load-key "${ENCRYPTIONROOT}"
+            ;;
+        file)
+            KEYFILE="${KEYLOCATION#file://}"
+            [ -r "${KEYFILE}" ] || udevadm settle
+            [ -r "${KEYFILE}" ] || {
+                info "ZFS: Waiting for key ${KEYFILE} for ${ENCRYPTIONROOT}..."
+                for _ in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20; do
+                    sleep 0.5s
+                    [ -r "${KEYFILE}" ] && break
                 done
-                ;;
-            http*)
-                systemctl start network-online.target
-                zfs load-key "${ENCRYPTIONROOT}"
-                ;;
-            file)
-                KEYFILE="${KEYLOCATION#file://}"
-                [ -r "${KEYFILE}" ] || udevadm settle
-                [ -r "${KEYFILE}" ] || {
-                    info "Waiting for key ${KEYFILE} for ${ENCRYPTIONROOT}..."
-                    for _ in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20; do
-                        sleep 0.5s
-                        [ -r "${KEYFILE}" ] && break
-                    done
-                }
-                [ -r "${KEYFILE}" ] || warn "Key ${KEYFILE} for ${ENCRYPTIONROOT} hasn't appeared. Trying anyway."
-                zfs load-key "${ENCRYPTIONROOT}"
-                ;;
-            *)
-                zfs load-key "${ENCRYPTIONROOT}"
-                ;;
-        esac
-    fi
-fi
+            }
+            [ -r "${KEYFILE}" ] || warn "ZFS: Key ${KEYFILE} for ${ENCRYPTIONROOT} hasn't appeared. Trying anyway."
+            zfs load-key "${ENCRYPTIONROOT}"
+            ;;
+        *)
+            zfs load-key "${ENCRYPTIONROOT}"
+            ;;
+    esac
+}
+
+_load_key_cb "$BOOTFS"
+for_relevant_root_children "$BOOTFS" _load_key_cb

contrib/dracut/90zfs/zfs-needshutdown.sh.in

@@ -2,7 +2,7 @@
 
 command -v getarg >/dev/null 2>&1 || . /lib/dracut-lib.sh
 
-if zpool list 2>&1 | grep -q 'no pools available' ; then
+if [ -z "$(zpool get -Ho value name)" ]; then
     info "ZFS: No active pools, no need to export anything."
 else
     info "ZFS: There is an active pool, will export it."

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

@@ -1,14 +1,12 @@
 [Unit]
 Description=Rollback bootfs just before it is mounted
 Requisite=zfs-import.target
-After=zfs-import.target zfs-snapshot-bootfs.service
+After=zfs-import.target dracut-pre-mount.service zfs-snapshot-bootfs.service
 Before=dracut-mount.service
 DefaultDependencies=no
 ConditionKernelCommandLine=bootfs.rollback
 
 [Service]
-# ${BOOTFS} should have been set by zfs-env-bootfs.service
 Type=oneshot
-ExecStartPre=/bin/sh -c 'test -n "${BOOTFS}"'
-ExecStart=/bin/sh -c '. /lib/dracut-lib.sh; SNAPNAME="$(getarg bootfs.rollback)"; exec @sbindir@/zfs rollback -Rf "${BOOTFS}@${SNAPNAME:-%v}"'
+ExecStart=/bin/sh -c '. /lib/dracut-zfs-lib.sh; decode_root_args || exit; [ "$root" = "zfs:AUTO" ] && root="$BOOTFS" SNAPNAME="$(getarg bootfs.rollback)"; exec @sbindir@/zfs rollback -Rf "$root@${SNAPNAME:-%v}"'
 RemainAfterExit=yes

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

@@ -1,14 +1,12 @@
 [Unit]
 Description=Snapshot bootfs just before it is mounted
 Requisite=zfs-import.target
-After=zfs-import.target
+After=zfs-import.target dracut-pre-mount.service
 Before=dracut-mount.service
 DefaultDependencies=no
 ConditionKernelCommandLine=bootfs.snapshot
 
 [Service]
-# ${BOOTFS} should have been set by zfs-env-bootfs.service
 Type=oneshot
-ExecStartPre=/bin/sh -c 'test -n "${BOOTFS}"'
-ExecStart=-/bin/sh -c '. /lib/dracut-lib.sh; SNAPNAME="$(getarg bootfs.snapshot)"; exec @sbindir@/zfs snapshot "${BOOTFS}@${SNAPNAME:-%v}"'
+-ExecStart=/bin/sh -c '. /lib/dracut-zfs-lib.sh; decode_root_args || exit; [ "$root" = "zfs:AUTO" ] && root="$BOOTFS" SNAPNAME="$(getarg bootfs.snapshot)"; exec @sbindir@/zfs snapshot "$root@${SNAPNAME:-%v}"'
 RemainAfterExit=yes

contrib/dracut/README.dracut.markdown

@@ -1,225 +1,50 @@
-How to setup a zfs root filesystem using dracut
------------------------------------------------
+## Basic setup
+1. Install `zfs-dracut`
+2. Set `mountpoint=/` for your root dataset (for compatibility, `legacy` also works, but is not recommended for new installations):
+    ```sh
+    zfs set mountpoint=/ pool/dataset
+    ```
+3. Either (a) set `bootfs=` on the pool to the dataset:
+    ```sh
+    zpool set bootfs=pool/dataset pool
+    ```
+4. Or (b) append `root=zfs:pool/dataset` to your kernel cmdline.
+5. Re-generate your initrd and update it in your boot bundle
 
-1) Install the zfs-dracut package.  This package adds a zfs dracut module
-to the /usr/share/dracut/modules.d/ directory which allows dracut to
-create an initramfs which is zfs aware.
+Encrypted datasets have keys loaded automatically or prompted for.
 
-2) Set the bootfs property for the bootable dataset in the pool.  Then set
-the dataset mountpoint property to '/'.
+If the root dataset contains children with `mountpoint=`s of `/etc`, `/bin`, `/lib*`, or `/usr`, they're mounted too.
 
-    $ zpool set bootfs=pool/dataset pool
-    $ zfs set mountpoint=/ pool/dataset
+For complete documentation, see `dracut.zfs(7)`.
 
-Alternately, legacy mountpoints can be used by setting the 'root=' option
-on the kernel line of your grub.conf/menu.lst configuration file.  Then
-set the dataset mountpoint property to 'legacy'.
+## cmdline
+1. `root=`                    | Root dataset is…                                         |
+   ---------------------------|----------------------------------------------------------|
+   *(empty)*                  | the first `bootfs=` after `zpool import -aN`             |
+   `zfs:AUTO`, `zfs:`, `zfs`  | *(as above, but overriding other autoselection methods)* |
+   `ZFS=pool/dataset`         | `pool/dataset`                                           |
+   `zfs:pool/dataset`         | *(as above)*                                             |
 
-    $ grub.conf/menu.lst: kernel ... root=ZFS=pool/dataset
-    $ zfs set mountpoint=legacy pool/dataset
+   All `+`es are replaced with spaces (i.e. to boot from `root pool/data set`, pass `root=zfs:root+pool/data+set`).
 
-3) To set zfs module options put them in /etc/modprobe.d/zfs.conf file.
-The complete list of zfs module options is available by running the
-_modinfo zfs_ command.  Commonly set options include: zfs_arc_min,
-zfs_arc_max, zfs_prefetch_disable, and zfs_vdev_max_pending.
+   The dataset can be at any depth, including being the pool's root dataset (i.e. `root=zfs:pool`).
 
-4) Finally, create your new initramfs by running dracut.
+   `rootfstype=zfs` is equivalent to `root=zfs:AUTO`, `rootfstype=zfs root=pool/dataset` is equivalent to `root=zfs:pool/dataset`.
 
-    $ dracut --force /path/to/initramfs kernel_version
+2. `spl_hostid`: passed to `zgenhostid -f`, useful to override the `/etc/hostid` file baked into the initrd.
 
-Kernel Command Line
--------------------
+3. `bootfs.snapshot`, `bootfs.snapshot=snapshot-name`: enables `zfs-snapshot-bootfs.service`,
+   which creates a snapshot `$root_dataset@$(uname -r)` (or, in the second form, `$root_dataset@snapshot-name`)
+   after pool import but before the rootfs is mounted.
+   Failure to create the snapshot is noted, but booting continues.
 
-The initramfs' behavior is influenced by the following kernel command line
-parameters passed in from the boot loader:
+4. `bootfs.rollback`, `bootfs.rollback=snapshot-name`: enables `zfs-snapshot-bootfs.service`,
+   which `-Rf` rolls back to `$root_dataset@$(uname -r)` (or, in the second form, `$root_dataset@snapshot-name`)
+   after pool import but before the rootfs is mounted.
+   Failure to roll back will fall down to the rescue shell.
+   This has obvious potential for data loss: make sure your persistent data is not below the rootfs and you don't care about any intermediate snapshots.
 
-* `root=...`: If not set, importable pools are searched for a bootfs
-attribute.  If an explicitly set root is desired, you may use
-`root=ZFS:pool/dataset`
+5. If both `bootfs.snapshot` and `bootfs.rollback` are set, `bootfs.rollback` is ordered *after* `bootfs.snapshot`.
 
-* `zfs_force=0`: If set to 1, the initramfs will run `zpool import -f` when
-attempting to import pools if the required pool isn't automatically imported
-by the zfs module.  This can save you a trip to a bootcd if hostid has
-changed, but is dangerous and can lead to zpool corruption, particularly in
-cases where storage is on a shared fabric such as iSCSI where multiple hosts
-can access storage devices concurrently.  _Please understand the implications
-of force-importing a pool before enabling this option!_
-
-* `spl_hostid`: By default, the hostid used by the SPL module is read from
-/etc/hostid inside the initramfs.  This file is placed there from the host
-system when the initramfs is built which effectively ties the ramdisk to the
-host which builds it.  If a different hostid is desired, one may be set in
-this attribute and will override any file present in the ramdisk.  The
-format should be hex exactly as found in the `/etc/hostid` file, IE
-`spl_hostid=0x00bab10c`.
-
-Note that changing the hostid between boots will most likely lead to an
-un-importable pool since the last importing hostid won't match.  In order
-to recover from this, you may use the `zfs_force` option or boot from a
-different filesystem and `zpool import -f` then `zpool export` the pool
-before rebooting with the new hostid.
-
-* `bootfs.snapshot`: If listed, enables the zfs-snapshot-bootfs service on a Dracut system. The zfs-snapshot-bootfs service simply runs `zfs snapshot $BOOTFS@%v` after the pool has been imported but before the bootfs is mounted. `$BOOTFS` is substituted with the value of the bootfs setting on the pool. `%v` is substituted with the version string of the kernel currently being booted (e.g. 5.6.6-200.fc31.x86\_64). Failure to create the snapshot (e.g. because one with the same name already exists) will be logged, but will not otherwise interrupt the boot process.
-
-    It is safe to leave the bootfs.snapshot flag set persistently on your kernel command line so that a new snapshot of your bootfs will be created on every kernel update. If you leave bootfs.snapshot set persistently on your kernel command line, you may find the below script helpful for automatically removing old snapshots of the bootfs along with their associated kernel.
-
-        #!/usr/bin/sh
-
-        if [[ "$1" == "remove" ]] && grep -q "\bbootfs.snapshot\b" /proc/cmdline; then
-           zfs destroy $(findmnt -n -o source /)@$2 &> /dev/null
-        fi
-
-        exit 0
-
-    To use the above script place it in a plain text file named /etc/kernel/install.d/99-zfs-cleanup.install and mark it executable with the following command:
-
-        $ chmod +x /etc/kernel/install.d/99-zfs-cleanup.install
-
-    On Red Hat based systems, you can change the value of `installonly_limit` in /etc/dnf/dnf.conf to adjust the number of kernels and their associated snapshots that are kept.
-
-* `bootfs.snapshot=<snapname>`: Is identical to the bootfs.snapshot parameter explained above except that the value substituted for \<snapname\> will be used when creating the snapshot instead of the version string of the kernel currently being booted. 
-
-* `bootfs.rollback`: If listed, enables the zfs-rollback-bootfs service on a Dracut system. The zfs-rollback-bootfs service simply runs `zfs rollback -Rf $BOOTFS@%v` after the pool has been imported but before the bootfs is mounted. If the rollback operation fails, the boot process will be interrupted with a Dracut rescue shell. __Use this parameter with caution. Intermediate snapshots of the bootfs will be destroyed!__ TIP: Keep your user data (e.g. /home) on separate file systems (it can be in the same pool though).
-
-* `bootfs.rollback=<snapname>`: Is identical to the bootfs.rollback parameter explained above except that the value substituted for \<snapname\> will be used when rolling back the bootfs instead of the version string of the kernel currently being booted. If you use this form, choose a snapshot that is new enough to contain the needed kernel modules under /lib/modules or use a kernel that has all the needed modules built-in.
-
-How it Works
-============
-
-The Dracut module consists of the following files (less Makefile's):
-
-* `module-setup.sh`: Script run by the initramfs builder to create the
-ramdisk.  Contains instructions on which files are required by the modules
-and z* programs.  Also triggers inclusion of `/etc/hostid` and the zpool
-cache.  This file is not included in the initramfs.
-
-* `90-zfs.rules`: udev rules which trigger loading of the ZFS modules at boot.
-
-* `zfs-lib.sh`: Utility functions used by the other files.
-
-* `parse-zfs.sh`: Run early in the initramfs boot process to parse kernel
-command line and determine if ZFS is the active root filesystem.
-
-* `mount-zfs.sh`: Run later in initramfs boot process after udev has settled
-to mount the root dataset.
-
-* `export-zfs.sh`: Run on shutdown after dracut has restored the initramfs
-and pivoted to it, allowing for a clean unmount and export of the ZFS root.
-
-`zfs-lib.sh`
-------------
-
-This file provides a few handy functions for working with ZFS. Those
-functions are used by the `mount-zfs.sh` and `export-zfs.sh` files.
-However, they could be used by any other file as well, as long as the file
-sources `/lib/dracut-zfs-lib.sh`.
-
-`module-setup.sh`
------------------
-
-This file is run by the Dracut script within the live system, not at boot
-time.  It's not included in the final initramfs.  Functions in this script
-describe which files are needed by ZFS at boot time.
-
-Currently all the various z* and spl modules are included, a dependency is
-asserted on udev-rules, and the various zfs, zpool, etc. helpers are included.
-Dracut provides library functions which automatically gather the shared libs
-necessary to run each of these binaries, so statically built binaries are
-not required.
-
-The zpool and zvol udev rules files are copied from where they are
-installed by the ZFS build.  __PACKAGERS TAKE NOTE__: If you move
-`/etc/udev/rules/60-z*.rules`, you'll need to update this file to match.
-
-Currently this file also includes `/etc/hostid` and `/etc/zfs/zpool.cache`
-which means the generated ramdisk is specific to the host system which built
-it.  If a generic initramfs is required, it may be preferable to omit these
-files and specify the `spl_hostid` from the boot loader instead.
-
-`parse-zfs.sh`
---------------
-
-Run during the cmdline phase of the initramfs boot process, this script
-performs some basic sanity checks on kernel command line parameters to
-determine if booting from ZFS is likely to be what is desired.  Dracut
-requires this script to adjust the `root` variable if required and to set
-`rootok=1` if a mountable root filesystem is available.  Unfortunately this
-script must run before udev is settled and kernel modules are known to be
-loaded, so accessing the zpool and zfs commands is unsafe.
-
-If the root=ZFS... parameter is set on the command line, then it's at least
-certain that ZFS is what is desired, though this script is unable to
-determine if ZFS is in fact available.  This script will alter the `root`
-parameter to replace several historical forms of specifying the pool and
-dataset name with the canonical form of `zfs:pool/dataset`.
-
-If no root= parameter is set, the best this script can do is guess that
-ZFS is desired.  At present, no other known filesystems will work with no
-root= parameter, though this might possibly interfere with using the
-compiled-in default root in the kernel image.  It's considered unlikely
-that would ever be the case when an initramfs is in use, so this script
-sets `root=zfs:AUTO` and hopes for the best.
-
-Once the root=... (or lack thereof) parameter is parsed, a dummy symlink
-is created from `/dev/root` -> `/dev/null` to satisfy parts of the Dracut
-process which check for presence of a single root device node.
-
-Finally, an initqueue/finished hook is registered which causes the initqueue
-phase of Dracut to wait for `/dev/zfs` to become available before attempting
-to mount anything.
-
-`mount-zfs.sh`
---------------
-
-This script is run after udev has settled and all tasks in the initqueue
-have succeeded.  This ensures that `/dev/zfs` is available and that the
-various ZFS modules are successfully loaded.  As it is now safe to call
-zpool and friends, we can proceed to find the bootfs attribute if necessary.
-
-If the root parameter was explicitly set on the command line, no parsing is
-necessary.  The list of imported pools is checked to see if the desired pool
-is already imported.  If it's not, and attempt is made to import the pool
-explicitly, though no force is attempted.  Finally the specified dataset
-is mounted on `$NEWROOT`, first using the `-o zfsutil` option to handle
-non-legacy mounts, then if that fails, without zfsutil to handle legacy
-mount points.
-
-If no root parameter was specified, this script attempts to find a pool with
-its bootfs attribute set.  First, already-imported pools are scanned and if
-an appropriate pool is found, no additional pools are imported.  If no pool
-with bootfs is found, any additional pools in the system are imported with
-`zpool import -N -a`, and the scan for bootfs is tried again.  If no bootfs
-is found with all pools imported, all pools are re-exported, and boot fails.
-Assuming a bootfs is found, an attempt is made to mount it to `$NEWROOT`,
-first with, then without the zfsutil option as above.
-
-Ordinarily pools are imported _without_ the force option which may cause
-boot to fail if the hostid has changed or a pool has been physically moved
-between servers.  The `zfs_force` kernel parameter is provided which when
-set to `1` causes `zpool import` to be run with the `-f` flag.  Forcing pool
-import can lead to serious data corruption and loss of pools, so this option
-should be used with extreme caution.  Note that even with this flag set, if
-the required zpool was auto-imported by the kernel module, no additional
-`zpool import` commands are run, so nothing is forced.
-
-`export-zfs.sh`
----------------
-
-Normally the zpool containing the root dataset cannot be exported on
-shutdown as it is still in use by the init process. To work around this,
-Dracut is able to restore the initramfs on shutdown and pivot to it.
-All remaining process are then running from a ramdisk, allowing for a
-clean unmount and export of the ZFS root. The theory of operation is
-described in detail in the [Dracut manual](https://www.kernel.org/pub/linux/utils/boot/dracut/dracut.html#_dracut_on_shutdown).
-
-This script will try to export all remaining zpools after Dracut has
-pivoted to the initramfs. If an initial regular export is not successful,
-Dracut will call this script once more with the `final` option,
-in which case a forceful export is attempted.
-
-Other Dracut modules include similar shutdown scripts and Dracut
-invokes these scripts round-robin until they succeed. In particular,
-the `90dm` module installs a script which tries to close and remove
-all device mapper targets. Thus, if there are ZVOLs containing
-dm-crypt volumes or if the zpool itself is backed by a dm-crypt
-volume, the shutdown scripts will try to untangle this.
+6. `zfs_force`, `zfs.force`, `zfsforce`: add `-f` to all `zpool import` invocations.
+   May be useful. Use with caution.

contrib/pyzfs/libzfs_core/_constants.py

@@ -104,7 +104,7 @@ zfs_errno = enum_with_offset(1024, [
 )
 # compat before we used the enum helper for these values
 ZFS_ERR_CHECKPOINT_EXISTS = zfs_errno.ZFS_ERR_CHECKPOINT_EXISTS
-assert(ZFS_ERR_CHECKPOINT_EXISTS == 1024)
+assert (ZFS_ERR_CHECKPOINT_EXISTS == 1024)
 ZFS_ERR_DISCARDING_CHECKPOINT = zfs_errno.ZFS_ERR_DISCARDING_CHECKPOINT
 ZFS_ERR_NO_CHECKPOINT = zfs_errno.ZFS_ERR_NO_CHECKPOINT
 ZFS_ERR_DEVRM_IN_PROGRESS = zfs_errno.ZFS_ERR_DEVRM_IN_PROGRESS

etc/default/Makefile.am

@@ -1,9 +1,10 @@
 include $(top_srcdir)/config/Substfiles.am
 include $(top_srcdir)/config/Shellcheck.am
 
-initconf_SCRIPTS = zfs
+initconf_DATA = zfs
 
-SUBSTFILES += $(initconf_SCRIPTS)
+SUBSTFILES += $(initconf_DATA)
 
+SHELLCHECKSCRIPTS = $(initconf_DATA)
 SHELLCHECK_SHELL = sh
 SHELLCHECK_IGNORE = ,SC2034

etc/systemd/system-generators/zfs-mount-generator.c

@@ -27,9 +27,6 @@
 #include <sys/types.h>
 #include <sys/time.h>
 #include <sys/stat.h>
-#include <sys/wait.h>
-#include <sys/mman.h>
-#include <semaphore.h>
 #include <stdbool.h>
 #include <unistd.h>
 #include <fcntl.h>
@@ -44,25 +41,16 @@
 #include <errno.h>
 #include <libzfs.h>
 
+/*
+ * For debugging only.
+ *
+ * Free statics with trivial life-times,
+ * but saved line filenames are replaced with a static string.
+ */
+#define	FREE_STATICS false
+
+#define	nitems(arr) (sizeof (arr) / sizeof (*arr))
 #define	STRCMP ((int(*)(const void *, const void *))&strcmp)
-#define	PID_T_CMP ((int(*)(const void *, const void *))&pid_t_cmp)
-
-static int
-pid_t_cmp(const pid_t *lhs, const pid_t *rhs)
-{
-	/*
-	 * This is always valid, quoth sys_types.h(7posix):
-	 * > blksize_t, pid_t, and ssize_t shall be signed integer types.
-	 */
-	return (*lhs - *rhs);
-}
-
-#define	EXIT_ENOMEM() \
-	do { \
-		fprintf(stderr, PROGNAME "[%d]: " \
-		    "not enough memory (L%d)!\n", getpid(), __LINE__); \
-		_exit(1); \
-	} while (0)
 
 
 #define	PROGNAME "zfs-mount-generator"
@@ -80,20 +68,11 @@ pid_t_cmp(const pid_t *lhs, const pid_t *rhs)
 #define	URI_REGEX_S "^\\([A-Za-z][A-Za-z0-9+.\\-]*\\):\\/\\/\\(.*\\)$"
 static regex_t uri_regex;
 
-static char *argv0;
-
 static const char *destdir = "/tmp";
 static int destdir_fd = -1;
 
 static void *known_pools = NULL; /* tsearch() of C strings */
-static struct {
-	sem_t noauto_not_on_sem;
-
-	sem_t noauto_names_sem;
-	size_t noauto_names_len;
-	size_t noauto_names_max;
-	char noauto_names[][NAME_MAX];
-} *noauto_files;
+static void *noauto_files = NULL; /* tsearch() of C strings */
 
 
 static char *
@@ -103,8 +82,12 @@ systemd_escape(const char *input, const char *prepend, const char *append)
 	size_t applen = strlen(append);
 	size_t prelen = strlen(prepend);
 	char *ret = malloc(4 * len + prelen + applen + 1);
-	if (!ret)
-		EXIT_ENOMEM();
+	if (!ret) {
+		fprintf(stderr, PROGNAME "[%d]: "
+		    "out of memory to escape \"%s%s%s\"!\n",
+		    getpid(), prepend, input, append);
+		return (NULL);
+	}
 
 	memcpy(ret, prepend, prelen);
 	char *out = ret + prelen;
@@ -166,8 +149,12 @@ systemd_escape_path(char *input, const char *prepend, const char *append)
 {
 	if (strcmp(input, "/") == 0) {
 		char *ret;
-		if (asprintf(&ret, "%s-%s", prepend, append) == -1)
-			EXIT_ENOMEM();
+		if (asprintf(&ret, "%s-%s", prepend, append) == -1) {
+			fprintf(stderr, PROGNAME "[%d]: "
+			    "out of memory to escape \"%s%s%s\"!\n",
+			    getpid(), prepend, input, append);
+			ret = NULL;
+		}
 		return (ret);
 	} else {
 		/*
@@ -209,6 +196,10 @@ fopenat(int dirfd, const char *pathname, int flags,
 static int
 line_worker(char *line, const char *cachefile)
 {
+	int ret = 0;
+	void *tofree_all[8];
+	void **tofree = tofree_all;
+
 	char *toktmp;
 	/* BEGIN CSTYLED */
 	const char *dataset                     = strtok_r(line, "\t", &toktmp);
@@ -240,11 +231,9 @@ line_worker(char *line, const char *cachefile)
 	if (p_nbmand == NULL) {
 		fprintf(stderr, PROGNAME "[%d]: %s: not enough tokens!\n",
 		    getpid(), dataset);
-		return (1);
+		goto err;
 	}
 
-	strncpy(argv0, dataset, strlen(argv0));
-
 	/* Minimal pre-requisites to mount a ZFS dataset */
 	const char *after = "zfs-import.target";
 	const char *wants = "zfs-import.target";
@@ -280,28 +269,31 @@ line_worker(char *line, const char *cachefile)
 
 
 	if (strcmp(p_encroot, "-") != 0) {
-		char *keyloadunit =
+		char *keyloadunit = *(tofree++) =
 		    systemd_escape(p_encroot, "zfs-load-key@", ".service");
+		if (keyloadunit == NULL)
+			goto err;
 
 		if (strcmp(dataset, p_encroot) == 0) {
 			const char *keymountdep = NULL;
 			bool is_prompt = false;
+			bool need_network = false;
 
 			regmatch_t uri_matches[3];
 			if (regexec(&uri_regex, p_keyloc,
-			    sizeof (uri_matches) / sizeof (*uri_matches),
-			    uri_matches, 0) == 0) {
+			    nitems(uri_matches), uri_matches, 0) == 0) {
+				p_keyloc[uri_matches[1].rm_eo] = '\0';
 				p_keyloc[uri_matches[2].rm_eo] = '\0';
+				const char *scheme =
+				    &p_keyloc[uri_matches[1].rm_so];
 				const char *path =
 				    &p_keyloc[uri_matches[2].rm_so];
 
-				/*
-				 * Assumes all URI keylocations need
-				 * the mount for their path;
-				 * http://, for example, wouldn't
-				 * (but it'd need network-online.target et al.)
-				 */
-				keymountdep = path;
+				if (strcmp(scheme, "https") == 0 ||
+				    strcmp(scheme, "http") == 0)
+					need_network = true;
+				else
+					keymountdep = path;
 			} else {
 				if (strcmp(p_keyloc, "prompt") != 0)
 					fprintf(stderr, PROGNAME "[%d]: %s: "
@@ -321,7 +313,7 @@ line_worker(char *line, const char *cachefile)
 				    "couldn't open %s under %s: %s\n",
 				    getpid(), dataset, keyloadunit, destdir,
 				    strerror(errno));
-				return (1);
+				goto err;
 			}
 
 			fprintf(keyloadunit_f,
@@ -335,20 +327,22 @@ line_worker(char *line, const char *cachefile)
 			    "After=%s\n",
 			    dataset, cachefile, wants, after);
 
+			if (need_network)
+				fprintf(keyloadunit_f,
+				    "Wants=network-online.target\n"
+				    "After=network-online.target\n");
+
 			if (p_systemd_requires)
 				fprintf(keyloadunit_f,
 				    "Requires=%s\n", p_systemd_requires);
 
-			if (p_systemd_requiresmountsfor || keymountdep) {
-				fprintf(keyloadunit_f, "RequiresMountsFor=");
-				if (p_systemd_requiresmountsfor)
-					fprintf(keyloadunit_f,
-					    "%s ", p_systemd_requiresmountsfor);
-				if (keymountdep)
-					fprintf(keyloadunit_f,
-					    "'%s'", keymountdep);
-				fprintf(keyloadunit_f, "\n");
-			}
+			if (p_systemd_requiresmountsfor)
+				fprintf(keyloadunit_f,
+				    "RequiresMountsFor=%s\n",
+				    p_systemd_requiresmountsfor);
+			if (keymountdep)
+				fprintf(keyloadunit_f,
+				    "RequiresMountsFor='%s'\n", keymountdep);
 
 			/* BEGIN CSTYLED */
 			fprintf(keyloadunit_f,
@@ -393,9 +387,13 @@ line_worker(char *line, const char *cachefile)
 		if (after[0] == '\0')
 			after = keyloadunit;
 		else if (asprintf(&toktmp, "%s %s", after, keyloadunit) != -1)
-			after = toktmp;
-		else
-			EXIT_ENOMEM();
+			after = *(tofree++) = toktmp;
+		else {
+			fprintf(stderr, PROGNAME "[%d]: %s: "
+			    "out of memory to generate after=\"%s %s\"!\n",
+			    getpid(), dataset, after, keyloadunit);
+			goto err;
+		}
 	}
 
 
@@ -404,12 +402,12 @@ line_worker(char *line, const char *cachefile)
 	    strcmp(p_systemd_ignore, "off") == 0) {
 		/* ok */
 	} else if (strcmp(p_systemd_ignore, "on") == 0)
-		return (0);
+		goto end;
 	else {
 		fprintf(stderr, PROGNAME "[%d]: %s: "
 		    "invalid org.openzfs.systemd:ignore=%s\n",
 		    getpid(), dataset, p_systemd_ignore);
-		return (1);
+		goto err;
 	}
 
 	/* Check for canmount */
@@ -418,21 +416,21 @@ line_worker(char *line, const char *cachefile)
 	} else if (strcmp(p_canmount, "noauto") == 0)
 		noauto = true;
 	else if (strcmp(p_canmount, "off") == 0)
-		return (0);
+		goto end;
 	else {
 		fprintf(stderr, PROGNAME "[%d]: %s: invalid canmount=%s\n",
 		    getpid(), dataset, p_canmount);
-		return (1);
+		goto err;
 	}
 
 	/* Check for legacy and blank mountpoints */
 	if (strcmp(p_mountpoint, "legacy") == 0 ||
 	    strcmp(p_mountpoint, "none") == 0)
-		return (0);
+		goto end;
 	else if (p_mountpoint[0] != '/') {
 		fprintf(stderr, PROGNAME "[%d]: %s: invalid mountpoint=%s\n",
 		    getpid(), dataset, p_mountpoint);
-		return (1);
+		goto err;
 	}
 
 	/* Escape the mountpoint per systemd policy */
@@ -442,7 +440,7 @@ line_worker(char *line, const char *cachefile)
 		fprintf(stderr,
 		    PROGNAME "[%d]: %s: abnormal simplified mountpoint: %s\n",
 		    getpid(), dataset, p_mountpoint);
-		return (1);
+		goto err;
 	}
 
 
@@ -552,8 +550,7 @@ line_worker(char *line, const char *cachefile)
 	 * 	files if we're sure they were created by us. (see 5.)
 	 * 2.	We handle files differently based on canmount.
 	 * 	Units with canmount=on always have precedence over noauto.
-	 * 	This is enforced by the noauto_not_on_sem semaphore,
-	 * 	which is only unlocked when the last canmount=on process exits.
+	 * 	This is enforced by processing these units before all others.
 	 * 	It is important to use p_canmount and not noauto here,
 	 * 	since we categorise by canmount while other properties,
 	 * 	e.g. org.openzfs.systemd:wanted-by, also modify noauto.
@@ -561,7 +558,7 @@ line_worker(char *line, const char *cachefile)
 	 * 	Additionally, we use noauto_files to track the unit file names
 	 * 	(which are the systemd-escaped mountpoints) of all (exclusively)
 	 * 	noauto datasets that had a file created.
-	 * 4.	If the file to be created is found in the tracking array,
+	 * 4.	If the file to be created is found in the tracking tree,
 	 * 	we do NOT create it.
 	 * 5.	If a file exists for a noauto dataset,
 	 * 	we check whether the file name is in the array.
@@ -571,29 +568,14 @@ line_worker(char *line, const char *cachefile)
 	 * 	further noauto datasets creating a file for this path again.
 	 */
 
-	{
-		sem_t *our_sem = (strcmp(p_canmount, "on") == 0) ?
-		    &noauto_files->noauto_names_sem :
-		    &noauto_files->noauto_not_on_sem;
-		while (sem_wait(our_sem) == -1 && errno == EINTR)
-			;
-	}
-
 	struct stat stbuf;
 	bool already_exists = fstatat(destdir_fd, mountfile, &stbuf, 0) == 0;
+	bool is_known = tfind(mountfile, &noauto_files, STRCMP) != NULL;
 
-	bool is_known = false;
-	for (size_t i = 0; i < noauto_files->noauto_names_len; ++i) {
-		if (strncmp(
-		    noauto_files->noauto_names[i], mountfile, NAME_MAX) == 0) {
-			is_known = true;
-			break;
-		}
-	}
-
+	*(tofree++) = (void *)mountfile;
 	if (already_exists) {
 		if (is_known) {
-			/* If it's in $noauto_files, we must be noauto too */
+			/* If it's in noauto_files, we must be noauto too */
 
 			/* See 5 */
 			errno = 0;
@@ -614,43 +596,31 @@ line_worker(char *line, const char *cachefile)
 		}
 
 		/* File exists: skip current dataset */
-		if (strcmp(p_canmount, "on") == 0)
-			sem_post(&noauto_files->noauto_names_sem);
-		return (0);
+		goto end;
 	} else {
 		if (is_known) {
 			/* See 4 */
-			if (strcmp(p_canmount, "on") == 0)
-				sem_post(&noauto_files->noauto_names_sem);
-			return (0);
+			goto end;
 		} else if (strcmp(p_canmount, "noauto") == 0) {
-			if (noauto_files->noauto_names_len ==
-			    noauto_files->noauto_names_max)
+			if (tsearch(mountfile, &noauto_files, STRCMP) == NULL)
 				fprintf(stderr, PROGNAME "[%d]: %s: "
-				    "noauto dataset limit (%zu) reached! "
-				    "Not tracking %s. Please report this to "
-				    "https://github.com/openzfs/zfs\n",
-				    getpid(), dataset,
-				    noauto_files->noauto_names_max, mountfile);
-			else {
-				strncpy(noauto_files->noauto_names[
-				    noauto_files->noauto_names_len],
-				    mountfile, NAME_MAX);
-				++noauto_files->noauto_names_len;
-			}
+				    "out of memory for noauto datasets! "
+				    "Not tracking %s.\n",
+				    getpid(), dataset, mountfile);
+			else
+				/* mountfile escaped to noauto_files */
+				*(--tofree) = NULL;
 		}
 	}
 
 
 	FILE *mountfile_f = fopenat(destdir_fd, mountfile,
 	    O_WRONLY | O_CREAT | O_TRUNC | O_CLOEXEC, "w", 0644);
-	if (strcmp(p_canmount, "on") == 0)
-		sem_post(&noauto_files->noauto_names_sem);
 	if (!mountfile_f) {
 		fprintf(stderr,
 		    PROGNAME "[%d]: %s: couldn't open %s under %s: %s\n",
 		    getpid(), dataset, mountfile, destdir, strerror(errno));
-		return (1);
+		goto err;
 	}
 
 	fprintf(mountfile_f,
@@ -699,12 +669,17 @@ line_worker(char *line, const char *cachefile)
 	(void) fclose(mountfile_f);
 
 	if (!requiredby && !wantedby)
-		return (0);
+		goto end;
 
 	/* Finally, create the appropriate dependencies */
 	char *linktgt;
-	if (asprintf(&linktgt, "../%s", mountfile) == -1)
-		EXIT_ENOMEM();
+	if (asprintf(&linktgt, "../%s", mountfile) == -1) {
+		fprintf(stderr, PROGNAME "[%d]: %s: "
+		    "out of memory for dependents of %s!\n",
+		    getpid(), dataset, mountfile);
+		goto err;
+	}
+	*(tofree++) = linktgt;
 
 	char *dependencies[][2] = {
 		{"wants", wantedby},
@@ -719,8 +694,14 @@ line_worker(char *line, const char *cachefile)
 		    reqby;
 		    reqby = strtok_r(NULL, " ", &toktmp)) {
 			char *depdir;
-			if (asprintf(&depdir, "%s.%s", reqby, (*dep)[0]) == -1)
-				EXIT_ENOMEM();
+			if (asprintf(
+			    &depdir, "%s.%s", reqby, (*dep)[0]) == -1) {
+				fprintf(stderr, PROGNAME "[%d]: %s: "
+				    "out of memory for dependent dir name "
+				    "\"%s.%s\"!\n",
+				    getpid(), dataset, reqby, (*dep)[0]);
+				continue;
+			}
 
 			(void) mkdirat(destdir_fd, depdir, 0755);
 			int depdir_fd = openat(destdir_fd, depdir,
@@ -746,7 +727,24 @@ line_worker(char *line, const char *cachefile)
 		}
 	}
 
-	return (0);
+end:
+	if (tofree >= tofree_all + nitems(tofree_all)) {
+		/*
+		 * This won't happen as-is:
+		 * we've got 8 slots and allocate 4 things at most.
+		 */
+		fprintf(stderr,
+		    PROGNAME "[%d]: %s: need to free %zu > %zu!\n",
+		    getpid(), dataset, tofree - tofree_all, nitems(tofree_all));
+		ret = tofree - tofree_all;
+	}
+
+	while (tofree-- != tofree_all)
+		free(*tofree);
+	return (ret);
+err:
+	ret = 1;
+	goto end;
 }
 
 
@@ -780,12 +778,11 @@ main(int argc, char **argv)
 		if (kmfd >= 0) {
 			(void) dup2(kmfd, STDERR_FILENO);
 			(void) close(kmfd);
+
+			setlinebuf(stderr);
 		}
 	}
 
-	uint8_t debug = 0;
-
-	argv0 = argv[0];
 	switch (argc) {
 	case 1:
 		/* Use default */
@@ -844,33 +841,9 @@ main(int argc, char **argv)
 		}
 	}
 
-	{
-		/*
-		 * We could just get a gigabyte here and Not Care,
-		 * but if vm.overcommit_memory=2, then MAP_NORESERVE is ignored
-		 * and we'd try (and likely fail) to rip it out of swap
-		 */
-		noauto_files = mmap(NULL, 4 * 1024 * 1024,
-		    PROT_READ | PROT_WRITE,
-		    MAP_SHARED | MAP_ANONYMOUS | MAP_NORESERVE, -1, 0);
-		if (noauto_files == MAP_FAILED) {
-			fprintf(stderr,
-			    PROGNAME "[%d]: couldn't allocate IPC region: %s\n",
-			    getpid(), strerror(errno));
-			_exit(1);
-		}
-
-		sem_init(&noauto_files->noauto_not_on_sem, true, 0);
-		sem_init(&noauto_files->noauto_names_sem, true, 1);
-		noauto_files->noauto_names_len = 0;
-		/* Works out to 16447ish, *well* enough */
-		noauto_files->noauto_names_max =
-		    (4 * 1024 * 1024 - sizeof (*noauto_files)) / NAME_MAX;
-	}
-
+	bool debug = false;
 	char *line = NULL;
 	size_t linelen = 0;
-	struct timespec time_start = {};
 	{
 		const char *dbgenv = getenv("ZFS_DEBUG");
 		if (dbgenv)
@@ -879,7 +852,7 @@ main(int argc, char **argv)
 			FILE *cmdline = fopen("/proc/cmdline", "re");
 			if (cmdline != NULL) {
 				if (getline(&line, &linelen, cmdline) >= 0)
-					debug = strstr(line, "debug") ? 2 : 0;
+					debug = strstr(line, "debug");
 				(void) fclose(cmdline);
 			}
 		}
@@ -888,19 +861,17 @@ main(int argc, char **argv)
 			dup2(STDERR_FILENO, STDOUT_FILENO);
 	}
 
-	size_t forked_canmount_on = 0;
-	size_t forked_canmount_not_on = 0;
-	size_t canmount_on_pids_len = 128;
-	pid_t *canmount_on_pids =
-	    malloc(canmount_on_pids_len * sizeof (*canmount_on_pids));
-	if (canmount_on_pids == NULL)
-		canmount_on_pids_len = 0;
-
+	struct timespec time_start = {};
 	if (debug)
 		clock_gettime(CLOCK_MONOTONIC_RAW, &time_start);
 
-	ssize_t read;
-	pid_t pid;
+	struct line {
+		char *line;
+		const char *fname;
+		struct line *next;
+	} *lines_canmount_not_on = NULL;
+
+	int ret = 0;
 	struct dirent *cachent;
 	while ((cachent = readdir(fslist_dir)) != NULL) {
 		if (strcmp(cachent->d_name, ".") == 0 ||
@@ -916,129 +887,67 @@ main(int argc, char **argv)
 			continue;
 		}
 
+		const char *filename = FREE_STATICS ? "(elided)" : NULL;
+
+		ssize_t read;
 		while ((read = getline(&line, &linelen, cachefile)) >= 0) {
 			line[read - 1] = '\0'; /* newline */
 
-			switch (pid = fork()) {
-			case -1:
-				fprintf(stderr,
-				    PROGNAME "[%d]: couldn't fork for %s: %s\n",
-				    getpid(), line, strerror(errno));
-				break;
-			case 0: /* child */
-				_exit(line_worker(line, cachent->d_name));
-			default: { /* parent */
-				char *tmp;
-				char *dset = strtok_r(line, "\t", &tmp);
-				strtok_r(NULL, "\t", &tmp);
-				char *canmount = strtok_r(NULL, "\t", &tmp);
-				bool canmount_on =
-				    canmount && strncmp(canmount, "on", 2) == 0;
+			char *canmount = line;
+			canmount += strcspn(canmount, "\t");
+			canmount += strspn(canmount, "\t");
+			canmount += strcspn(canmount, "\t");
+			canmount += strspn(canmount, "\t");
+			bool canmount_on = strncmp(canmount, "on", 2) == 0;
 
-				if (debug >= 2)
-					printf(PROGNAME ": forked %d, "
-					    "canmount_on=%d, dataset=%s\n",
-					    (int)pid, canmount_on, dset);
+			if (canmount_on)
+				ret |= line_worker(line, cachent->d_name);
+			else {
+				if (filename == NULL)
+					filename =
+					    strdup(cachent->d_name) ?: "(?)";
 
-				if (canmount_on &&
-				    forked_canmount_on ==
-				    canmount_on_pids_len) {
-					size_t new_len =
-					    (canmount_on_pids_len ?: 16) * 2;
-					void *new_pidlist =
-					    realloc(canmount_on_pids,
-					    new_len *
-					    sizeof (*canmount_on_pids));
-					if (!new_pidlist) {
-						fprintf(stderr,
-						    PROGNAME "[%d]: "
-						    "out of memory! "
-						    "Mount ordering may be "
-						    "affected.\n", getpid());
-						continue;
-					}
-
-					canmount_on_pids = new_pidlist;
-					canmount_on_pids_len = new_len;
+				struct line *l = calloc(1, sizeof (*l));
+				char *nl = strdup(line);
+				if (l == NULL || nl == NULL) {
+					fprintf(stderr, PROGNAME "[%d]: "
+					    "out of memory for \"%s\" in %s\n",
+					    getpid(), line, cachent->d_name);
+					free(l);
+					free(nl);
+					continue;
 				}
-
-				if (canmount_on) {
-					canmount_on_pids[forked_canmount_on] =
-					    pid;
-					++forked_canmount_on;
-				} else
-					++forked_canmount_not_on;
-				break;
-			}
+				l->line = nl;
+				l->fname = filename;
+				l->next = lines_canmount_not_on;
+				lines_canmount_not_on = l;
 			}
 		}
 
-		(void) fclose(cachefile);
+		fclose(cachefile);
 	}
 	free(line);
 
-	if (forked_canmount_on == 0) {
-		/* No canmount=on processes to finish, so don't deadlock here */
-		for (size_t i = 0; i < forked_canmount_not_on; ++i)
-			sem_post(&noauto_files->noauto_not_on_sem);
-	} else {
-		/* Likely a no-op, since we got these from a narrow fork loop */
-		qsort(canmount_on_pids, forked_canmount_on,
-		    sizeof (*canmount_on_pids), PID_T_CMP);
-	}
+	while (lines_canmount_not_on) {
+		struct line *l = lines_canmount_not_on;
+		lines_canmount_not_on = l->next;
 
-	int status, ret = 0;
-	struct rusage usage;
-	size_t forked_canmount_on_max = forked_canmount_on;
-	while ((pid = wait4(-1, &status, 0, &usage)) != -1) {
-		ret |= WEXITSTATUS(status) | WTERMSIG(status);
-
-		if (forked_canmount_on != 0) {
-			if (bsearch(&pid, canmount_on_pids,
-			    forked_canmount_on_max, sizeof (*canmount_on_pids),
-			    PID_T_CMP))
-				--forked_canmount_on;
-
-			if (forked_canmount_on == 0) {
-				/*
-				 * All canmount=on processes have finished,
-				 * let all the lower-priority ones finish now
-				 */
-				for (size_t i = 0;
-				    i < forked_canmount_not_on; ++i)
-					sem_post(
-					    &noauto_files->noauto_not_on_sem);
-			}
+		ret |= line_worker(l->line, l->fname);
+		if (FREE_STATICS) {
+			free(l->line);
+			free(l);
 		}
-
-		if (debug >= 2)
-			printf(PROGNAME ": %d done, user=%llu.%06us, "
-			    "system=%llu.%06us, maxrss=%ldB, ex=0x%x\n",
-			    (int)pid,
-			    (unsigned long long) usage.ru_utime.tv_sec,
-			    (unsigned int) usage.ru_utime.tv_usec,
-			    (unsigned long long) usage.ru_stime.tv_sec,
-			    (unsigned int) usage.ru_stime.tv_usec,
-			    usage.ru_maxrss * 1024, status);
 	}
 
 	if (debug) {
 		struct timespec time_end = {};
 		clock_gettime(CLOCK_MONOTONIC_RAW, &time_end);
 
+		struct rusage usage;
 		getrusage(RUSAGE_SELF, &usage);
 		printf(
 		    "\n"
-		    PROGNAME ": self    : "
-		    "user=%llu.%06us, system=%llu.%06us, maxrss=%ldB\n",
-		    (unsigned long long) usage.ru_utime.tv_sec,
-		    (unsigned int) usage.ru_utime.tv_usec,
-		    (unsigned long long) usage.ru_stime.tv_sec,
-		    (unsigned int) usage.ru_stime.tv_usec,
-		    usage.ru_maxrss * 1024);
-
-		getrusage(RUSAGE_CHILDREN, &usage);
-		printf(PROGNAME ": children: "
+		    PROGNAME ": "
 		    "user=%llu.%06us, system=%llu.%06us, maxrss=%ldB\n",
 		    (unsigned long long) usage.ru_utime.tv_sec,
 		    (unsigned int) usage.ru_utime.tv_usec,
@@ -1068,7 +977,7 @@ main(int argc, char **argv)
 		    time_init.tv_nsec / 1000000000;
 		time_init.tv_nsec %= 1000000000;
 
-		printf(PROGNAME ": wall    : "
+		printf(PROGNAME ": "
 		    "total=%llu.%09llus = "
 		    "init=%llu.%09llus + real=%llu.%09llus\n",
 		    (unsigned long long) time_init.tv_sec,
@@ -1077,7 +986,15 @@ main(int argc, char **argv)
 		    (unsigned long long) time_start.tv_nsec,
 		    (unsigned long long) time_end.tv_sec,
 		    (unsigned long long) time_end.tv_nsec);
+
+		fflush(stdout);
 	}
 
+	if (FREE_STATICS) {
+		closedir(fslist_dir);
+		tdestroy(noauto_files, free);
+		tdestroy(known_pools, free);
+		regfree(&uri_regex);
+	}
 	_exit(ret);
 }

etc/zfs/Makefile.am

@@ -10,9 +10,10 @@ dist_pkgsysconf_DATA = \
 	vdev_id.conf.multipath.example \
 	vdev_id.conf.scsi.example
 
-pkgsysconf_SCRIPTS = \
+pkgsysconf_DATA = \
 	zfs-functions
 
-SUBSTFILES += $(pkgsysconf_SCRIPTS)
+SUBSTFILES += $(pkgsysconf_DATA)
 
+SHELLCHECKSCRIPTS = $(pkgsysconf_DATA)
 SHELLCHECK_SHELL = dash # local variables

include/libzfs.h

@@ -795,9 +795,10 @@ extern int zfs_receive(libzfs_handle_t *, const char *, nvlist_t *,
     recvflags_t *, int, avl_tree_t *);
 
 typedef enum diff_flags {
-	ZFS_DIFF_PARSEABLE = 0x1,
-	ZFS_DIFF_TIMESTAMP = 0x2,
-	ZFS_DIFF_CLASSIFY = 0x4
+	ZFS_DIFF_PARSEABLE = 1 << 0,
+	ZFS_DIFF_TIMESTAMP = 1 << 1,
+	ZFS_DIFF_CLASSIFY = 1 << 2,
+	ZFS_DIFF_NO_MANGLE = 1 << 3
 } diff_flags_t;
 
 extern int zfs_show_diffs(zfs_handle_t *, int, const char *, const char *,

include/libzfs_impl.h

@@ -234,6 +234,7 @@ typedef struct differ_info {
 	boolean_t scripted;
 	boolean_t classify;
 	boolean_t timestamped;
+	boolean_t no_mangle;
 	uint64_t shares;
 	int zerr;
 	int cleanupfd;

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

@@ -52,7 +52,7 @@
 
 #define	ZFS_MODULE_PARAM_CALL_IMPL(parent, name, perm, args, desc) \
     SYSCTL_DECL(parent); \
-    SYSCTL_PROC(parent, OID_AUTO, name, perm | args, desc)
+    SYSCTL_PROC(parent, OID_AUTO, name, CTLFLAG_MPSAFE | perm | args, desc)
 
 #define	ZFS_MODULE_PARAM_CALL(scope_prefix, name_prefix, name, func, _, perm, desc) \
     ZFS_MODULE_PARAM_CALL_IMPL(_vfs_ ## scope_prefix, name, perm, func ## _args(name_prefix ## name), desc)

include/os/freebsd/zfs/sys/zfs_znode_impl.h

@@ -123,25 +123,29 @@ extern minor_t zfsdev_minor_alloc(void);
 #define	zn_rlimit_fsize(zp, uio) \
     vn_rlimit_fsize(ZTOV(zp), GET_UIO_STRUCT(uio), zfs_uio_td(uio))
 
+#define	ZFS_ENTER_ERROR(zfsvfs, error) do {			\
+	ZFS_TEARDOWN_ENTER_READ((zfsvfs), FTAG);		\
+	if (__predict_false((zfsvfs)->z_unmounted)) {		\
+		ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG);		\
+		return (error);					\
+	}							\
+} while (0)
+
 /* Called on entry to each ZFS vnode and vfs operation  */
-#define	ZFS_ENTER(zfsvfs) \
-	{ \
-		ZFS_TEARDOWN_ENTER_READ((zfsvfs), FTAG); \
-		if (__predict_false((zfsvfs)->z_unmounted)) { \
-			ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG); \
-			return (EIO); \
-		} \
-	}
+#define	ZFS_ENTER(zfsvfs)	ZFS_ENTER_ERROR(zfsvfs, EIO)
 
 /* Must be called before exiting the vop */
-#define	ZFS_EXIT(zfsvfs) ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG)
+#define	ZFS_EXIT(zfsvfs)	ZFS_TEARDOWN_EXIT_READ(zfsvfs, FTAG)
+
+#define	ZFS_VERIFY_ZP_ERROR(zp, error) do {			\
+	if (__predict_false((zp)->z_sa_hdl == NULL)) {		\
+		ZFS_EXIT((zp)->z_zfsvfs);			\
+		return (error);					\
+	}							\
+} while (0)
 
 /* Verifies the znode is valid */
-#define	ZFS_VERIFY_ZP(zp) \
-	if (__predict_false((zp)->z_sa_hdl == NULL)) { \
-		ZFS_EXIT((zp)->z_zfsvfs); \
-		return (EIO); \
-	} \
+#define	ZFS_VERIFY_ZP(zp)	ZFS_VERIFY_ZP_ERROR(zp, EIO)
 
 /*
  * Macros for dealing with dmu_buf_hold

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

@@ -495,21 +495,45 @@ blk_queue_discard_granularity(struct request_queue *q, unsigned int dg)
 }
 
 /*
+ * 5.19 API,
+ *   bdev_max_discard_sectors()
+ *
+ * 2.6.32 API,
+ *   blk_queue_discard()
+ */
+static inline boolean_t
+bdev_discard_supported(struct block_device *bdev)
+{
+#if defined(HAVE_BDEV_MAX_DISCARD_SECTORS)
+	return (!!bdev_max_discard_sectors(bdev));
+#elif defined(HAVE_BLK_QUEUE_DISCARD)
+	return (!!blk_queue_discard(bdev_get_queue(bdev)));
+#else
+#error "Unsupported kernel"
+#endif
+}
+
+/*
+ * 5.19 API,
+ *   bdev_max_secure_erase_sectors()
+ *
  * 4.8 API,
  *   blk_queue_secure_erase()
  *
  * 2.6.36 - 4.7 API,
  *   blk_queue_secdiscard()
  */
-static inline int
-blk_queue_discard_secure(struct request_queue *q)
+static inline boolean_t
+bdev_secure_discard_supported(struct block_device *bdev)
 {
-#if defined(HAVE_BLK_QUEUE_SECURE_ERASE)
-	return (blk_queue_secure_erase(q));
+#if defined(HAVE_BDEV_MAX_SECURE_ERASE_SECTORS)
+	return (!!bdev_max_secure_erase_sectors(bdev));
+#elif defined(HAVE_BLK_QUEUE_SECURE_ERASE)
+	return (!!blk_queue_secure_erase(bdev_get_queue(bdev)));
 #elif defined(HAVE_BLK_QUEUE_SECDISCARD)
-	return (blk_queue_secdiscard(q));
+	return (!!blk_queue_secdiscard(bdev_get_queue(bdev)));
 #else
-	return (0);
+#error "Unsupported kernel"
 #endif
 }
 
@@ -527,7 +551,10 @@ blk_generic_start_io_acct(struct request_queue *q __attribute__((unused)),
     struct gendisk *disk __attribute__((unused)),
     int rw __attribute__((unused)), struct bio *bio)
 {
-#if defined(HAVE_DISK_IO_ACCT)
+#if defined(HAVE_BDEV_IO_ACCT)
+	return (bdev_start_io_acct(bio->bi_bdev, bio_sectors(bio),
+	    bio_op(bio), jiffies));
+#elif defined(HAVE_DISK_IO_ACCT)
 	return (disk_start_io_acct(disk, bio_sectors(bio), bio_op(bio)));
 #elif defined(HAVE_BIO_IO_ACCT)
 	return (bio_start_io_acct(bio));
@@ -550,7 +577,9 @@ blk_generic_end_io_acct(struct request_queue *q __attribute__((unused)),
     struct gendisk *disk __attribute__((unused)),
     int rw __attribute__((unused)), struct bio *bio, unsigned long start_time)
 {
-#if defined(HAVE_DISK_IO_ACCT)
+#if defined(HAVE_BDEV_IO_ACCT)
+	bdev_end_io_acct(bio->bi_bdev, bio_op(bio), start_time);
+#elif defined(HAVE_DISK_IO_ACCT)
 	disk_end_io_acct(disk, bio_op(bio), start_time);
 #elif defined(HAVE_BIO_IO_ACCT)
 	bio_end_io_acct(bio, start_time);

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

@@ -87,7 +87,9 @@
 
 #if defined(HAVE_KERNEL_FPU_API_HEADER)
 #include <asm/fpu/api.h>
+#if defined(HAVE_KERNEL_FPU_INTERNAL_HEADER)
 #include <asm/fpu/internal.h>
+#endif
 #if defined(HAVE_KERNEL_FPU_XCR_HEADER)
 #include <asm/fpu/xcr.h>
 #endif

include/os/linux/kernel/linux/xattr_compat.h

@@ -115,6 +115,20 @@ fn(struct dentry *dentry, const char *name, void *buffer, size_t size,	\
 {									\
 	return (__ ## fn(dentry->d_inode, name, buffer, size));		\
 }
+/*
+ * Android API change,
+ * The xattr_handler->get() callback was changed to take a dentry and inode
+ * and flags, because the dentry might not be attached to an inode yet.
+ */
+#elif defined(HAVE_XATTR_GET_DENTRY_INODE_FLAGS)
+#define	ZPL_XATTR_GET_WRAPPER(fn)					\
+static int								\
+fn(const struct xattr_handler *handler, struct dentry *dentry,		\
+    struct inode *inode, const char *name, void *buffer,		\
+    size_t size, int flags)						\
+{									\
+	return (__ ## fn(inode, name, buffer, size));			\
+}
 #else
 #error "Unsupported kernel"
 #endif

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

@@ -64,7 +64,11 @@
  * }
  */
 
+#ifdef HAVE_REGISTER_SHRINKER_VARARG
+#define	spl_register_shrinker(x)	register_shrinker(x, "zfs-arc-shrinker")
+#else
 #define	spl_register_shrinker(x)	register_shrinker(x)
+#endif
 #define	spl_unregister_shrinker(x)	unregister_shrinker(x)
 
 /*

include/sys/abd.h

@@ -91,6 +91,7 @@ abd_t *abd_alloc_linear(size_t, boolean_t);
 abd_t *abd_alloc_gang(void);
 abd_t *abd_alloc_for_io(size_t, boolean_t);
 abd_t *abd_alloc_sametype(abd_t *, size_t);
+boolean_t abd_size_alloc_linear(size_t);
 void abd_gang_add(abd_t *, abd_t *, boolean_t);
 void abd_free(abd_t *);
 abd_t *abd_get_offset(abd_t *, size_t);

include/sys/abd_impl.h

@@ -68,7 +68,6 @@ abd_t *abd_get_offset_scatter(abd_t *, abd_t *, size_t, size_t);
 void abd_free_struct_impl(abd_t *);
 void abd_alloc_chunks(abd_t *, size_t);
 void abd_free_chunks(abd_t *);
-boolean_t abd_size_alloc_linear(size_t);
 void abd_update_scatter_stats(abd_t *, abd_stats_op_t);
 void abd_update_linear_stats(abd_t *, abd_stats_op_t);
 void abd_verify_scatter(abd_t *);

include/sys/arc.h

@@ -85,6 +85,7 @@ typedef void arc_prune_func_t(int64_t bytes, void *priv);
 
 /* Shared module parameters */
 extern int zfs_arc_average_blocksize;
+extern int l2arc_exclude_special;
 
 /* generic arc_done_func_t's which you can use */
 arc_read_done_func_t arc_bcopy_func;

include/sys/bqueue.h

@@ -30,22 +30,22 @@ typedef struct bqueue {
 	kmutex_t bq_lock;
 	kcondvar_t bq_add_cv;
 	kcondvar_t bq_pop_cv;
-	uint64_t bq_size;
-	uint64_t bq_maxsize;
-	uint64_t bq_fill_fraction;
+	size_t bq_size;
+	size_t bq_maxsize;
+	uint_t bq_fill_fraction;
 	size_t bq_node_offset;
 } bqueue_t;
 
 typedef struct bqueue_node {
 	list_node_t bqn_node;
-	uint64_t bqn_size;
+	size_t bqn_size;
 } bqueue_node_t;
 
 
-int bqueue_init(bqueue_t *, uint64_t, uint64_t, size_t);
+int bqueue_init(bqueue_t *, uint_t, size_t, size_t);
 void bqueue_destroy(bqueue_t *);
-void bqueue_enqueue(bqueue_t *, void *, uint64_t);
-void bqueue_enqueue_flush(bqueue_t *, void *, uint64_t);
+void bqueue_enqueue(bqueue_t *, void *, size_t);
+void bqueue_enqueue_flush(bqueue_t *, void *, size_t);
 void *bqueue_dequeue(bqueue_t *);
 boolean_t bqueue_empty(bqueue_t *);
 

include/sys/btree.h

@@ -72,7 +72,11 @@ extern kmem_cache_t *zfs_btree_leaf_cache;
 
 typedef struct zfs_btree_hdr {
 	struct zfs_btree_core	*bth_parent;
-	boolean_t		bth_core;
+	/*
+	 * Set to -1 to indicate core nodes. Other values represent first
+	 * valid element offset for leaf nodes.
+	 */
+	uint32_t		bth_first;
 	/*
 	 * For both leaf and core nodes, represents the number of elements in
 	 * the node. For core nodes, they will have bth_count + 1 children.
@@ -91,9 +95,12 @@ typedef struct zfs_btree_leaf {
 	uint8_t		btl_elems[];
 } zfs_btree_leaf_t;
 
+#define	BTREE_LEAF_ESIZE	(BTREE_LEAF_SIZE - \
+    offsetof(zfs_btree_leaf_t, btl_elems))
+
 typedef struct zfs_btree_index {
 	zfs_btree_hdr_t	*bti_node;
-	uint64_t	bti_offset;
+	uint32_t	bti_offset;
 	/*
 	 * True if the location is before the list offset, false if it's at
 	 * the listed offset.
@@ -105,6 +112,7 @@ typedef struct btree {
 	zfs_btree_hdr_t		*bt_root;
 	int64_t			bt_height;
 	size_t			bt_elem_size;
+	uint32_t		bt_leaf_cap;
 	uint64_t		bt_num_elems;
 	uint64_t		bt_num_nodes;
 	zfs_btree_leaf_t	*bt_bulk; // non-null if bulk loading

include/sys/crypto/icp.h

@@ -32,9 +32,6 @@ int aes_mod_fini(void);
 int edonr_mod_init(void);
 int edonr_mod_fini(void);
 
-int sha1_mod_init(void);
-int sha1_mod_fini(void);
-
 int sha2_mod_init(void);
 int sha2_mod_fini(void);
 

include/sys/dbuf.h

@@ -330,7 +330,7 @@ typedef struct dbuf_hash_table {
 	kmutex_t hash_mutexes[DBUF_MUTEXES] ____cacheline_aligned;
 } dbuf_hash_table_t;
 
-typedef void (*dbuf_prefetch_fn)(void *, boolean_t);
+typedef void (*dbuf_prefetch_fn)(void *, uint64_t, uint64_t, boolean_t);
 
 uint64_t dbuf_whichblock(const struct dnode *di, const int64_t level,
     const uint64_t offset);
@@ -442,16 +442,7 @@ dbuf_find_dirty_eq(dmu_buf_impl_t *db, uint64_t txg)
 	(dbuf_is_metadata(_db) &&					\
 	((_db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))
 
-#define	DBUF_IS_L2CACHEABLE(_db)					\
-	((_db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL ||	\
-	(dbuf_is_metadata(_db) &&					\
-	((_db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
-
-#define	DNODE_LEVEL_IS_L2CACHEABLE(_dn, _level)				\
-	((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_ALL ||	\
-	(((_level) > 0 ||						\
-	DMU_OT_IS_METADATA((_dn)->dn_handle->dnh_dnode->dn_type)) &&	\
-	((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
+boolean_t dbuf_is_l2cacheable(dmu_buf_impl_t *db);
 
 #ifdef ZFS_DEBUG
 

include/sys/dmu.h

@@ -136,7 +136,7 @@ typedef enum dmu_object_byteswap {
 #endif
 
 #define	DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \
-	((ot) & DMU_OT_METADATA) : \
+	(((ot) & DMU_OT_METADATA) != 0) : \
 	DMU_OT_IS_METADATA_IMPL(ot))
 
 #define	DMU_OT_IS_DDT(ot) \
@@ -147,7 +147,7 @@ typedef enum dmu_object_byteswap {
 	((ot) == DMU_OT_PLAIN_FILE_CONTENTS || (ot) == DMU_OT_UINT64_OTHER)
 
 #define	DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
-	((ot) & DMU_OT_ENCRYPTED) : \
+	(((ot) & DMU_OT_ENCRYPTED) != 0) : \
 	DMU_OT_IS_ENCRYPTED_IMPL(ot))
 
 /*
@@ -1067,6 +1067,8 @@ int dmu_diff(const char *tosnap_name, const char *fromsnap_name,
 #define	ZFS_CRC64_POLY	0xC96C5795D7870F42ULL	/* ECMA-182, reflected form */
 extern uint64_t zfs_crc64_table[256];
 
+extern int dmu_prefetch_max;
+
 #ifdef	__cplusplus
 }
 #endif

include/sys/dmu_objset.h

@@ -200,10 +200,6 @@ struct objset {
 #define	DMU_GROUPUSED_DNODE(os)	((os)->os_groupused_dnode.dnh_dnode)
 #define	DMU_PROJECTUSED_DNODE(os) ((os)->os_projectused_dnode.dnh_dnode)
 
-#define	DMU_OS_IS_L2CACHEABLE(os)				\
-	((os)->os_secondary_cache == ZFS_CACHE_ALL ||		\
-	(os)->os_secondary_cache == ZFS_CACHE_METADATA)
-
 /* called from zpl */
 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
 int dmu_objset_hold_flags(const char *name, boolean_t decrypt, void *tag,

include/sys/dmu_tx.h

@@ -125,6 +125,7 @@ typedef struct dmu_tx_stats {
 	kstat_named_t dmu_tx_dirty_delay;
 	kstat_named_t dmu_tx_dirty_over_max;
 	kstat_named_t dmu_tx_dirty_frees_delay;
+	kstat_named_t dmu_tx_wrlog_delay;
 	kstat_named_t dmu_tx_quota;
 } dmu_tx_stats_t;
 

include/sys/dmu_zfetch.h

@@ -49,20 +49,18 @@ typedef struct zfetch {
 
 typedef struct zstream {
 	uint64_t	zs_blkid;	/* expect next access at this blkid */
-	uint64_t	zs_pf_blkid1;	/* first block to prefetch */
-	uint64_t	zs_pf_blkid;	/* block to prefetch up to */
-
-	/*
-	 * We will next prefetch the L1 indirect block of this level-0
-	 * block id.
-	 */
-	uint64_t	zs_ipf_blkid1;	/* first block to prefetch */
-	uint64_t	zs_ipf_blkid;	/* block to prefetch up to */
+	unsigned int	zs_pf_dist;	/* data prefetch distance in bytes */
+	unsigned int	zs_ipf_dist;	/* L1 prefetch distance in bytes */
+	uint64_t	zs_pf_start;	/* first data block to prefetch */
+	uint64_t	zs_pf_end;	/* data block to prefetch up to */
+	uint64_t	zs_ipf_start;	/* first data block to prefetch L1 */
+	uint64_t	zs_ipf_end;	/* data block to prefetch L1 up to */
 
 	list_node_t	zs_node;	/* link for zf_stream */
 	hrtime_t	zs_atime;	/* time last prefetch issued */
 	zfetch_t	*zs_fetch;	/* parent fetch */
 	boolean_t	zs_missed;	/* stream saw cache misses */
+	boolean_t	zs_more;	/* need more distant prefetch */
 	zfs_refcount_t	zs_callers;	/* number of pending callers */
 	/*
 	 * Number of stream references: dnode, callers and pending blocks.

include/sys/dnode.h

@@ -616,7 +616,7 @@ _NOTE(CONSTCOND) } while (0)
 #else
 
 #define	dprintf_dnode(db, fmt, ...)
-#define	DNODE_VERIFY(dn)
+#define	DNODE_VERIFY(dn)		((void) sizeof ((uintptr_t)(dn)))
 #define	FREE_VERIFY(db, start, end, tx)
 
 #endif

include/sys/dsl_pool.h

@@ -40,6 +40,7 @@
 #include <sys/rrwlock.h>
 #include <sys/dsl_synctask.h>
 #include <sys/mmp.h>
+#include <sys/aggsum.h>
 
 #ifdef	__cplusplus
 extern "C" {
@@ -58,6 +59,7 @@ struct dsl_deadlist;
 
 extern unsigned long zfs_dirty_data_max;
 extern unsigned long zfs_dirty_data_max_max;
+extern unsigned long zfs_wrlog_data_max;
 extern int zfs_dirty_data_sync_percent;
 extern int zfs_dirty_data_max_percent;
 extern int zfs_dirty_data_max_max_percent;
@@ -82,7 +84,6 @@ typedef struct zfs_blkstat {
 
 typedef struct zfs_all_blkstats {
 	zfs_blkstat_t	zab_type[DN_MAX_LEVELS + 1][DMU_OT_TOTAL + 1];
-	kmutex_t	zab_lock;
 } zfs_all_blkstats_t;
 
 
@@ -119,6 +120,9 @@ typedef struct dsl_pool {
 	uint64_t dp_mos_compressed_delta;
 	uint64_t dp_mos_uncompressed_delta;
 
+	aggsum_t dp_wrlog_pertxg[TXG_SIZE];
+	aggsum_t dp_wrlog_total;
+
 	/*
 	 * Time of most recently scheduled (furthest in the future)
 	 * wakeup for delayed transactions.
@@ -159,6 +163,8 @@ uint64_t dsl_pool_adjustedsize(dsl_pool_t *dp, zfs_space_check_t slop_policy);
 uint64_t dsl_pool_unreserved_space(dsl_pool_t *dp,
     zfs_space_check_t slop_policy);
 uint64_t dsl_pool_deferred_space(dsl_pool_t *dp);
+void dsl_pool_wrlog_count(dsl_pool_t *dp, int64_t size, uint64_t txg);
+boolean_t dsl_pool_need_wrlog_delay(dsl_pool_t *dp);
 void dsl_pool_dirty_space(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx);
 void dsl_pool_undirty_space(dsl_pool_t *dp, int64_t space, uint64_t txg);
 void dsl_free(dsl_pool_t *dp, uint64_t txg, const blkptr_t *bpp);

include/sys/dsl_scan.h

@@ -155,7 +155,7 @@ typedef struct dsl_scan {
 	dsl_scan_phys_t scn_phys;	/* on disk representation of scan */
 	dsl_scan_phys_t scn_phys_cached;
 	avl_tree_t scn_queue;		/* queue of datasets to scan */
-	uint64_t scn_bytes_pending;	/* outstanding data to issue */
+	uint64_t scn_queues_pending;	/* outstanding data to issue */
 } dsl_scan_t;
 
 typedef struct dsl_scan_io_queue dsl_scan_io_queue_t;

include/sys/fs/zfs.h

@@ -757,6 +757,7 @@ typedef struct zpool_load_policy {
 
 /* Rewind data discovered */
 #define	ZPOOL_CONFIG_LOAD_TIME		"rewind_txg_ts"
+#define	ZPOOL_CONFIG_LOAD_META_ERRORS	"verify_meta_errors"
 #define	ZPOOL_CONFIG_LOAD_DATA_ERRORS	"verify_data_errors"
 #define	ZPOOL_CONFIG_REWIND_TIME	"seconds_of_rewind"
 
@@ -1101,6 +1102,7 @@ typedef struct vdev_stat {
 	uint64_t	vs_configured_ashift;   /* TLV vdev_ashift */
 	uint64_t	vs_logical_ashift;	/* vdev_logical_ashift  */
 	uint64_t	vs_physical_ashift;	/* vdev_physical_ashift */
+	uint64_t	vs_pspace;		/* physical capacity */
 } vdev_stat_t;
 
 /* BEGIN CSTYLED */
@@ -1613,6 +1615,47 @@ typedef enum {
 #define	ZFS_EV_HIST_DSID	"history_dsid"
 #define	ZFS_EV_RESILVER_TYPE	"resilver_type"
 
+
+/*
+ * We currently support block sizes from 512 bytes to 16MB.
+ * The benefits of larger blocks, and thus larger IO, need to be weighed
+ * against the cost of COWing a giant block to modify one byte, and the
+ * large latency of reading or writing a large block.
+ *
+ * The recordsize property can not be set larger than zfs_max_recordsize
+ * (default 16MB on 64-bit and 1MB on 32-bit). See the comment near
+ * zfs_max_recordsize in dsl_dataset.c for details.
+ *
+ * Note that although the LSIZE field of the blkptr_t can store sizes up
+ * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
+ * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
+ */
+#define	SPA_MINBLOCKSHIFT	9
+#define	SPA_OLD_MAXBLOCKSHIFT	17
+#define	SPA_MAXBLOCKSHIFT	24
+#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
+#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
+#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
+
+
+/* supported encryption algorithms */
+enum zio_encrypt {
+	ZIO_CRYPT_INHERIT = 0,
+	ZIO_CRYPT_ON,
+	ZIO_CRYPT_OFF,
+	ZIO_CRYPT_AES_128_CCM,
+	ZIO_CRYPT_AES_192_CCM,
+	ZIO_CRYPT_AES_256_CCM,
+	ZIO_CRYPT_AES_128_GCM,
+	ZIO_CRYPT_AES_192_GCM,
+	ZIO_CRYPT_AES_256_GCM,
+	ZIO_CRYPT_FUNCTIONS
+};
+
+#define	ZIO_CRYPT_ON_VALUE	ZIO_CRYPT_AES_256_GCM
+#define	ZIO_CRYPT_DEFAULT	ZIO_CRYPT_OFF
+
+
 #ifdef	__cplusplus
 }
 #endif

include/sys/metaslab.h

@@ -49,11 +49,14 @@ int metaslab_init(metaslab_group_t *, uint64_t, uint64_t, uint64_t,
     metaslab_t **);
 void metaslab_fini(metaslab_t *);
 
+void metaslab_set_unflushed_dirty(metaslab_t *, boolean_t);
 void metaslab_set_unflushed_txg(metaslab_t *, uint64_t, dmu_tx_t *);
 void metaslab_set_estimated_condensed_size(metaslab_t *, uint64_t, dmu_tx_t *);
+boolean_t metaslab_unflushed_dirty(metaslab_t *);
 uint64_t metaslab_unflushed_txg(metaslab_t *);
 uint64_t metaslab_estimated_condensed_size(metaslab_t *);
 int metaslab_sort_by_flushed(const void *, const void *);
+void metaslab_unflushed_bump(metaslab_t *, dmu_tx_t *, boolean_t);
 uint64_t metaslab_unflushed_changes_memused(metaslab_t *);
 
 int metaslab_load(metaslab_t *);

include/sys/metaslab_impl.h

@@ -553,6 +553,7 @@ struct metaslab {
 	 * log space maps.
 	 */
 	uint64_t	ms_unflushed_txg;
+	boolean_t	ms_unflushed_dirty;
 
 	/* updated every time we are done syncing the metaslab's space map */
 	uint64_t	ms_synced_length;

include/sys/range_tree.h

@@ -63,12 +63,8 @@ typedef struct range_tree {
 	 */
 	uint8_t		rt_shift;
 	uint64_t	rt_start;
-	range_tree_ops_t *rt_ops;
-
-	/* rt_btree_compare should only be set if rt_arg is a b-tree */
+	const range_tree_ops_t *rt_ops;
 	void		*rt_arg;
-	int (*rt_btree_compare)(const void *, const void *);
-
 	uint64_t	rt_gap;		/* allowable inter-segment gap */
 
 	/*
@@ -278,11 +274,11 @@ rs_set_fill(range_seg_t *rs, range_tree_t *rt, uint64_t fill)
 
 typedef void range_tree_func_t(void *arg, uint64_t start, uint64_t size);
 
-range_tree_t *range_tree_create_impl(range_tree_ops_t *ops,
+range_tree_t *range_tree_create_gap(const range_tree_ops_t *ops,
     range_seg_type_t type, void *arg, uint64_t start, uint64_t shift,
-    int (*zfs_btree_compare) (const void *, const void *), uint64_t gap);
-range_tree_t *range_tree_create(range_tree_ops_t *ops, range_seg_type_t type,
-    void *arg, uint64_t start, uint64_t shift);
+    uint64_t gap);
+range_tree_t *range_tree_create(const range_tree_ops_t *ops,
+    range_seg_type_t type, void *arg, uint64_t start, uint64_t shift);
 void range_tree_destroy(range_tree_t *rt);
 boolean_t range_tree_contains(range_tree_t *rt, uint64_t start, uint64_t size);
 range_seg_t *range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size);
@@ -316,13 +312,6 @@ void range_tree_remove_xor_add_segment(uint64_t start, uint64_t end,
 void range_tree_remove_xor_add(range_tree_t *rt, range_tree_t *removefrom,
     range_tree_t *addto);
 
-void rt_btree_create(range_tree_t *rt, void *arg);
-void rt_btree_destroy(range_tree_t *rt, void *arg);
-void rt_btree_add(range_tree_t *rt, range_seg_t *rs, void *arg);
-void rt_btree_remove(range_tree_t *rt, range_seg_t *rs, void *arg);
-void rt_btree_vacate(range_tree_t *rt, void *arg);
-extern range_tree_ops_t rt_btree_ops;
-
 #ifdef	__cplusplus
 }
 #endif

include/sys/spa.h

@@ -72,27 +72,6 @@ struct dsl_pool;
 struct dsl_dataset;
 struct dsl_crypto_params;
 
-/*
- * We currently support block sizes from 512 bytes to 16MB.
- * The benefits of larger blocks, and thus larger IO, need to be weighed
- * against the cost of COWing a giant block to modify one byte, and the
- * large latency of reading or writing a large block.
- *
- * Note that although blocks up to 16MB are supported, the recordsize
- * property can not be set larger than zfs_max_recordsize (default 1MB).
- * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
- *
- * Note that although the LSIZE field of the blkptr_t can store sizes up
- * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
- * 32MB - 512 bytes.  Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
- */
-#define	SPA_MINBLOCKSHIFT	9
-#define	SPA_OLD_MAXBLOCKSHIFT	17
-#define	SPA_MAXBLOCKSHIFT	24
-#define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
-#define	SPA_OLD_MAXBLOCKSIZE	(1ULL << SPA_OLD_MAXBLOCKSHIFT)
-#define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
-
 /*
  * Alignment Shift (ashift) is an immutable, internal top-level vdev property
  * which can only be set at vdev creation time. Physical writes are always done

include/sys/spa_impl.h

@@ -146,9 +146,9 @@ typedef struct spa_config_lock {
 	kmutex_t	scl_lock;
 	kthread_t	*scl_writer;
 	int		scl_write_wanted;
+	int		scl_count;
 	kcondvar_t	scl_cv;
-	zfs_refcount_t	scl_count;
-} spa_config_lock_t;
+} ____cacheline_aligned spa_config_lock_t;
 
 typedef struct spa_config_dirent {
 	list_node_t	scd_link;
@@ -370,6 +370,7 @@ struct spa {
 	boolean_t	spa_is_root;		/* pool is root */
 	int		spa_minref;		/* num refs when first opened */
 	spa_mode_t	spa_mode;		/* SPA_MODE_{READ|WRITE} */
+	boolean_t	spa_read_spacemaps;	/* spacemaps available if ro */
 	spa_log_state_t spa_log_state;		/* log state */
 	uint64_t	spa_autoexpand;		/* lun expansion on/off */
 	ddt_t		*spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */

include/sys/spa_log_spacemap.h

@@ -30,7 +30,10 @@
 
 typedef struct log_summary_entry {
 	uint64_t lse_start;	/* start TXG */
+	uint64_t lse_end;	/* last TXG */
+	uint64_t lse_txgcount;	/* # of TXGs */
 	uint64_t lse_mscount;	/* # of metaslabs needed to be flushed */
+	uint64_t lse_msdcount;	/* # of dirty metaslabs needed to be flushed */
 	uint64_t lse_blkcount;	/* blocks held by this entry  */
 	list_node_t lse_node;
 } log_summary_entry_t;
@@ -50,6 +53,7 @@ typedef struct spa_log_sm {
 	uint64_t sls_nblocks;	/* number of blocks in this log */
 	uint64_t sls_mscount;	/* # of metaslabs flushed in the log's txg */
 	avl_node_t sls_node;	/* node in spa_sm_logs_by_txg */
+	space_map_t *sls_sm;	/* space map pointer, if open */
 } spa_log_sm_t;
 
 int spa_ld_log_spacemaps(spa_t *);
@@ -68,8 +72,9 @@ uint64_t spa_log_sm_memused(spa_t *);
 void spa_log_sm_decrement_mscount(spa_t *, uint64_t);
 void spa_log_sm_increment_current_mscount(spa_t *);
 
-void spa_log_summary_add_flushed_metaslab(spa_t *);
-void spa_log_summary_decrement_mscount(spa_t *, uint64_t);
+void spa_log_summary_add_flushed_metaslab(spa_t *, boolean_t);
+void spa_log_summary_dirty_flushed_metaslab(spa_t *, uint64_t);
+void spa_log_summary_decrement_mscount(spa_t *, uint64_t, boolean_t);
 void spa_log_summary_decrement_blkcount(spa_t *, uint64_t);
 
 boolean_t spa_flush_all_logs_requested(spa_t *);

include/sys/sysevent/dev.h

@@ -244,6 +244,9 @@ extern "C" {
 #define	DEV_PATH		"path"
 #define	DEV_IS_PART		"is_slice"
 #define	DEV_SIZE		"dev_size"
+
+/* Size of the whole parent block device (if dev is a partition) */
+#define	DEV_PARENT_SIZE		"dev_parent_size"
 #endif /* __linux__ */
 
 #define	EV_V1			1

include/sys/txg.h

@@ -78,7 +78,7 @@ extern void txg_register_callbacks(txg_handle_t *txghp, list_t *tx_callbacks);
 
 extern void txg_delay(struct dsl_pool *dp, uint64_t txg, hrtime_t delta,
     hrtime_t resolution);
-extern void txg_kick(struct dsl_pool *dp);
+extern void txg_kick(struct dsl_pool *dp, uint64_t txg);
 
 /*
  * Wait until the given transaction group has finished syncing.

include/sys/vdev_impl.h

@@ -642,6 +642,7 @@ extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise);
  */
 int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj);
 void vdev_metaslab_group_create(vdev_t *vd);
+uint64_t vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b);
 
 /*
  * Vdev ashift optimization tunables

include/sys/zap.h

@@ -110,7 +110,12 @@ typedef enum zap_flags {
 	 * already randomly distributed.
 	 */
 	ZAP_FLAG_PRE_HASHED_KEY = 1 << 2,
+#if defined(__linux__) && defined(_KERNEL)
+} zfs_zap_flags_t;
+#define	zap_flags_t	zfs_zap_flags_t
+#else
 } zap_flags_t;
+#endif
 
 /*
  * Create a new zapobj with no attributes and return its object number.

include/sys/zil.h

@@ -221,6 +221,15 @@ typedef struct {
 	uint64_t	lr_foid;	/* object id */
 } lr_ooo_t;
 
+/*
+ * Additional lr_attr_t fields.
+ */
+typedef struct {
+	uint64_t	lr_attr_attrs;		/* all of the attributes */
+	uint64_t	lr_attr_crtime[2];	/* create time */
+	uint8_t		lr_attr_scanstamp[32];
+} lr_attr_end_t;
+
 /*
  * Handle option extended vattr attributes.
  *
@@ -231,7 +240,7 @@ typedef struct {
 typedef struct {
 	uint32_t	lr_attr_masksize; /* number of elements in array */
 	uint32_t	lr_attr_bitmap; /* First entry of array */
-	/* remainder of array and any additional fields */
+	/* remainder of array and additional lr_attr_end_t fields */
 } lr_attr_t;
 
 /*

include/sys/zio.h

@@ -108,23 +108,6 @@ enum zio_checksum {
 
 #define	ZIO_DEDUPCHECKSUM	ZIO_CHECKSUM_SHA256
 
-/* supported encryption algorithms */
-enum zio_encrypt {
-	ZIO_CRYPT_INHERIT = 0,
-	ZIO_CRYPT_ON,
-	ZIO_CRYPT_OFF,
-	ZIO_CRYPT_AES_128_CCM,
-	ZIO_CRYPT_AES_192_CCM,
-	ZIO_CRYPT_AES_256_CCM,
-	ZIO_CRYPT_AES_128_GCM,
-	ZIO_CRYPT_AES_192_GCM,
-	ZIO_CRYPT_AES_256_GCM,
-	ZIO_CRYPT_FUNCTIONS
-};
-
-#define	ZIO_CRYPT_ON_VALUE	ZIO_CRYPT_AES_256_GCM
-#define	ZIO_CRYPT_DEFAULT	ZIO_CRYPT_OFF
-
 /* macros defining encryption lengths */
 #define	ZIO_OBJSET_MAC_LEN		32
 #define	ZIO_DATA_IV_LEN			12
@@ -699,6 +682,8 @@ extern void spa_handle_ignored_writes(spa_t *spa);
 /* zbookmark_phys functions */
 boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
     const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
+boolean_t zbookmark_subtree_tbd(const struct dnode_phys *dnp,
+    const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
 int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
     uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
 

lib/libavl/Makefile.am

@@ -5,6 +5,9 @@ VPATH = $(top_srcdir)/module/avl/
 # Includes kernel code, generate warnings for large stack frames
 AM_CFLAGS += $(FRAME_LARGER_THAN)
 
+# See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=54020
+AM_CFLAGS += -no-suppress
+
 noinst_LTLIBRARIES = libavl.la
 
 KERNEL_C = \

lib/libefi/Makefile.am

@@ -2,6 +2,9 @@ include $(top_srcdir)/config/Rules.am
 
 AM_CFLAGS += $(LIBUUID_CFLAGS) $(ZLIB_CFLAGS)
 
+# See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=54020
+AM_CFLAGS += -no-suppress
+
 noinst_LTLIBRARIES = libefi.la
 
 USER_C = \

lib/libicp/Makefile.am

@@ -6,6 +6,8 @@ VPATH = \
 
 # Includes kernel code, generate warnings for large stack frames
 AM_CFLAGS += $(FRAME_LARGER_THAN)
+# See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=54020
+AM_CFLAGS += -no-suppress
 
 noinst_LTLIBRARIES = libicp.la
 
@@ -17,7 +19,6 @@ ASM_SOURCES_AS = \
 	asm-x86_64/modes/gcm_pclmulqdq.S \
 	asm-x86_64/modes/aesni-gcm-x86_64.S \
 	asm-x86_64/modes/ghash-x86_64.S \
-	asm-x86_64/sha1/sha1-x86_64.S \
 	asm-x86_64/sha2/sha256_impl.S \
 	asm-x86_64/sha2/sha512_impl.S
 else
@@ -46,7 +47,6 @@ KERNEL_C = \
 	algs/modes/ctr.c \
 	algs/modes/ccm.c \
 	algs/modes/ecb.c \
-	algs/sha1/sha1.c \
 	algs/sha2/sha2.c \
 	algs/skein/skein.c \
 	algs/skein/skein_block.c \
@@ -54,7 +54,6 @@ KERNEL_C = \
 	illumos-crypto.c \
 	io/aes.c \
 	io/edonr_mod.c \
-	io/sha1_mod.c \
 	io/sha2_mod.c \
 	io/skein_mod.c \
 	os/modhash.c \

lib/libnvpair/Makefile.am

@@ -8,6 +8,9 @@ VPATH = \
 # and required CFLAGS for libtirpc
 AM_CFLAGS += $(FRAME_LARGER_THAN) $(LIBTIRPC_CFLAGS)
 
+# See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=54020
+AM_CFLAGS += -no-suppress
+
 lib_LTLIBRARIES = libnvpair.la
 
 include $(top_srcdir)/config/Abigail.am

lib/libshare/Makefile.am

@@ -2,6 +2,9 @@ include $(top_srcdir)/config/Rules.am
 
 DEFAULT_INCLUDES += -I$(srcdir)
 
+# See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=54020
+AM_CFLAGS += -no-suppress
+
 noinst_LTLIBRARIES = libshare.la
 
 USER_C = \

lib/libspl/Makefile.am

@@ -2,6 +2,9 @@ include $(top_srcdir)/config/Rules.am
 
 SUBDIRS = include
 
+# See https://debbugs.gnu.org/cgi/bugreport.cgi?bug=54020
+AM_CFLAGS += -no-suppress
+
 noinst_LTLIBRARIES = libspl_assert.la libspl.la
 
 libspl_assert_la_SOURCES = \