From 5a21214be84e4b00f419220e4da51f16cee13cad Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Fri, 10 Dec 2021 00:02:52 +0100
Subject: [PATCH 01/19] zfs, libzfs: diff: accept -h/ZFS_DIFF_NO_MANGLE,
 disabling path escaping
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Rich Ercolani <rincebrain@gmail.com>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Upstream-commit: 344bbc82e7054f61d5e7b3610b119820285fd2cb
Closes #12829
---
 cmd/zfs/zfs_main.c                            |  5 +-
 include/libzfs.h                              |  7 +--
 include/libzfs_impl.h                         |  1 +
 lib/libzfs/libzfs_diff.c                      | 10 +++-
 man/man8/zfs-diff.8                           |  6 ++-
 tests/runfiles/common.run                     |  2 +-
 .../functional/cli_root/zfs_diff/Makefile.am  |  1 +
 .../cli_root/zfs_diff/zfs_diff_cliargs.ksh    |  4 +-
 .../cli_root/zfs_diff/zfs_diff_mangle.ksh     | 48 +++++++++++++++++++
 9 files changed, 74 insertions(+), 10 deletions(-)
 create mode 100755 tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_mangle.ksh

diff --git a/cmd/zfs/zfs_main.c b/cmd/zfs/zfs_main.c
index d0bb73a72513..d648aef7e71c 100644
--- a/cmd/zfs/zfs_main.c
+++ b/cmd/zfs/zfs_main.c
@@ -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);
diff --git a/include/libzfs.h b/include/libzfs.h
index eeb4daae723b..d55e3f2e7384 100644
--- a/include/libzfs.h
+++ b/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 *,
diff --git a/include/libzfs_impl.h b/include/libzfs_impl.h
index 96b11dad137c..043ff9cd77e4 100644
--- a/include/libzfs_impl.h
+++ b/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;
diff --git a/lib/libzfs/libzfs_diff.c b/lib/libzfs/libzfs_diff.c
index d46e23a2fc0e..b721a9fd9ec5 100644
--- a/lib/libzfs/libzfs_diff.c
+++ b/lib/libzfs/libzfs_diff.c
@@ -176,8 +176,13 @@ print_what(FILE *fp, mode_t what)
 static void
 print_cmn(FILE *fp, differ_info_t *di, const char *file)
 {
-	stream_bytes(fp, di->dsmnt);
-	stream_bytes(fp, file);
+	if (!di->no_mangle) {
+		stream_bytes(fp, di->dsmnt);
+		stream_bytes(fp, file);
+	} else {
+		(void) fputs(di->dsmnt, fp);
+		(void) fputs(file, fp);
+	}
 }
 
 static void
@@ -752,6 +757,7 @@ zfs_show_diffs(zfs_handle_t *zhp, int outfd, const char *fromsnap,
 	di.scripted = (flags & ZFS_DIFF_PARSEABLE);
 	di.classify = (flags & ZFS_DIFF_CLASSIFY);
 	di.timestamped = (flags & ZFS_DIFF_TIMESTAMP);
+	di.no_mangle = (flags & ZFS_DIFF_NO_MANGLE);
 
 	di.outputfd = outfd;
 	di.datafd = pipefd[0];
diff --git a/man/man8/zfs-diff.8 b/man/man8/zfs-diff.8
index 49443bf47d17..a347f325203e 100644
--- a/man/man8/zfs-diff.8
+++ b/man/man8/zfs-diff.8
@@ -39,7 +39,7 @@
 .Sh SYNOPSIS
 .Nm zfs
 .Cm diff
-.Op Fl FHt
+.Op Fl FHth
 .Ar snapshot Ar snapshot Ns | Ns Ar filesystem
 .
 .Sh DESCRIPTION
@@ -92,6 +92,10 @@ Give more parsable tab-separated output, without header lines and without
 arrows.
 .It Fl t
 Display the path's inode change time as the first column of output.
+.It Fl h
+Do not
+.Sy \e0 Ns Ar ooo Ns -escape
+non-ASCII paths.
 .El
 .
 .Sh SEE ALSO
diff --git a/tests/runfiles/common.run b/tests/runfiles/common.run
index aefcd98436d7..19919a00afb3 100644
--- a/tests/runfiles/common.run
+++ b/tests/runfiles/common.run
@@ -178,7 +178,7 @@ tags = ['functional', 'cli_root', 'zfs_destroy']
 
 [tests/functional/cli_root/zfs_diff]
 tests = ['zfs_diff_changes', 'zfs_diff_cliargs', 'zfs_diff_timestamp',
-    'zfs_diff_types', 'zfs_diff_encrypted']
+    'zfs_diff_types', 'zfs_diff_encrypted', 'zfs_diff_mangle']
 tags = ['functional', 'cli_root', 'zfs_diff']
 
 [tests/functional/cli_root/zfs_get]
diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_diff/Makefile.am b/tests/zfs-tests/tests/functional/cli_root/zfs_diff/Makefile.am
index db90e058559d..bfb01dcb8f86 100644
--- a/tests/zfs-tests/tests/functional/cli_root/zfs_diff/Makefile.am
+++ b/tests/zfs-tests/tests/functional/cli_root/zfs_diff/Makefile.am
@@ -8,6 +8,7 @@ dist_pkgdata_SCRIPTS = \
 	zfs_diff_changes.ksh \
 	zfs_diff_cliargs.ksh \
 	zfs_diff_encrypted.ksh \
+	zfs_diff_mangle.ksh \
 	zfs_diff_timestamp.ksh \
 	zfs_diff_types.ksh
 
diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_cliargs.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_cliargs.ksh
index 7063bbe9ce6a..67eb18fa4a5d 100755
--- a/tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_cliargs.ksh
+++ b/tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_cliargs.ksh
@@ -39,8 +39,8 @@ function cleanup
 log_assert "'zfs diff' should only work with supported options."
 log_onexit cleanup
 
-typeset goodopts=("" "-F" "-H" "-t" "-FH" "-Ft" "-Ht" "-FHt")
-typeset badopts=("-f" "-h" "-h" "-T" "-Fx" "-Ho" "-tT" "-")
+typeset goodopts=("" "-h" "-t" "-th" "-H" "-Hh" "-Ht" "-Hth" "-F" "-Fh" "-Ft" "-Fth" "-FH" "-FHh" "-FHt" "-FHth")
+typeset badopts=("-f" "-T" "-Fx" "-Ho" "-tT" "-")
 
 DATASET="$TESTPOOL/$TESTFS"
 TESTSNAP1="$DATASET@snap1"
diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_mangle.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_mangle.ksh
new file mode 100755
index 000000000000..ffce9f06848f
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/cli_root/zfs_diff/zfs_diff_mangle.ksh
@@ -0,0 +1,48 @@
+#!/bin/ksh -p
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# http://www.illumos.org/license/CDDL.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+#
+# DESCRIPTION:
+# 'zfs diff' escapes filenames as expected, 'zfs diff -h' doesn't
+#
+# STRATEGY:
+# 1. Prepare a dataset
+# 2. Create some files
+# 3. verify 'zfs diff' mangles them and 'zfs diff -h' doesn't
+#
+
+verify_runnable "both"
+
+function cleanup
+{
+	log_must zfs destroy -r "$DATASET"
+}
+
+log_assert "'zfs diff' mangles filenames, 'zfs diff -h' doesn't"
+log_onexit cleanup
+
+DATASET="$TESTPOOL/$TESTFS/fs"
+TESTSNAP1="$DATASET@snap1"
+
+# 1. Prepare a dataset
+log_must zfs create "$DATASET"
+MNTPOINT="$(get_prop mountpoint "$DATASET")"
+log_must zfs snapshot "$TESTSNAP1"
+
+printf '%c\t'"$MNTPOINT/"'%s\n' M '' + 'śmieszny żupan'                       + 'достопримечательности'                                                                                                                                                                                              | sort > "$MNTPOINT/śmieszny żupan"
+printf '%c\t'"$MNTPOINT/"'%s\n' M '' + '\0305\0233mieszny\0040\0305\0274upan' + '\0320\0264\0320\0276\0321\0201\0321\0202\0320\0276\0320\0277\0321\0200\0320\0270\0320\0274\0320\0265\0321\0207\0320\0260\0321\0202\0320\0265\0320\0273\0321\0214\0320\0275\0320\0276\0321\0201\0321\0202\0320\0270' | sort > "$MNTPOINT/достопримечательности"
+log_must diff -u <(zfs diff -h "$TESTSNAP1" | grep -vF '<xattrdir>' | sort) "$MNTPOINT/śmieszny żupan"
+log_must diff -u <(zfs diff    "$TESTSNAP1" | grep -vF '<xattrdir>' | sort) "$MNTPOINT/достопримечательности"
+
+log_pass "'zfs diff' mangles filenames, 'zfs diff -h' doesn't"

From 7fbb90feeab7b8a61875e9ba1ec136125e8177cc Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Thu, 9 Dec 2021 23:42:02 +0100
Subject: [PATCH 02/19] libzfs: diff: stream_bytes: use fputc, %hho formats
 chars
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Rich Ercolani <rincebrain@gmail.com>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Upstream-commit: a72129edcb7353f5301096ea9541e8e345f052d4
Closes #12829
---
 lib/libzfs/libzfs_diff.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/lib/libzfs/libzfs_diff.c b/lib/libzfs/libzfs_diff.c
index b721a9fd9ec5..57a7c1556599 100644
--- a/lib/libzfs/libzfs_diff.c
+++ b/lib/libzfs/libzfs_diff.c
@@ -122,9 +122,9 @@ stream_bytes(FILE *fp, const char *string)
 
 	while ((c = *string++) != '\0') {
 		if (c > ' ' && c != '\\' && c < '\177') {
-			(void) fprintf(fp, "%c", c);
+			(void) fputc(c, fp);
 		} else {
-			(void) fprintf(fp, "\\%04o", (uint8_t)c);
+			(void) fprintf(fp, "\\%04hho", (uint8_t)c);
 		}
 	}
 }

From fe6f2651f55de3bf68ac4729386b5e85aa23a447 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Tue, 30 Nov 2021 17:29:50 +0100
Subject: [PATCH 03/19] etc/systemd/zfs-mount-generator: serialise, handle
 keylocation=http[s]://
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* etc/systemd/zfs-mount-generator: serialise

The wins for a relatively normal workload are rather slim:
	real	0.02119s/0.00985s=2.15029x
	user	0.02130s/0.00346s=6.15560x
	sys	0.03858s/0.00643s=6.00062x

	wall-total	0.014518s/0.005925s=2.45009x
	wall-init	0.014518s/0.002457s=5.90684x
	wall-real	0.014518s/0.003467s=4.18668x

But this is a big win on machines with a lot of datasets and expensive
forks.

For example, the gain on a VM on my work laptop with 900+ legacy-mount
Docker datasets, the original gains from the C rewrite were
only five-fold:
	real    0.516s/0.102s=5.05882x
	user    0.237s/0.143s=1.65734x
	sys     0.287s/0.100s=2.87x

And this serial variant gains this back there as well:
	real    0.102s/0.008s=12.75x
	user    0.143s/0.007s=20.42857
	sys     0.100s/0.001s=100x

	wall-total	0.09717s/0.00319s=30.40255x
	wall-init	0.00203s/0.00200s=1.015941x
	wall-real	0.09513s/0.00118s=80.02043x

For a total of
	real    0.516s/0.008s=64.5x
	user    0.237s/0.007s=33.85714x
	sys     0.287s/0.001s=287x

Suggested-by: Richard Laager <rlaager@wiktel.com>

* etc/systemd/zfs-mount-generator: pull in network for keylocation=https

Also simplify RequiresMountsFor= handling
Ref: #11956

Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Upstream-commit: 4325de09cd2993837bc32a83d61872b57e58298e
Closes #12138
---
 .../system-generators/zfs-mount-generator.c   | 437 +++++++-----------
 man/man8/zfs-mount-generator.8.in             |  19 +-
 2 files changed, 181 insertions(+), 275 deletions(-)

diff --git a/etc/systemd/system-generators/zfs-mount-generator.c b/etc/systemd/system-generators/zfs-mount-generator.c
index b806339deb2f..f4c6c26a0b34 100644
--- a/etc/systemd/system-generators/zfs-mount-generator.c
+++ b/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>
 
-#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);
-}
+/*
+ * For debugging only.
+ *
+ * Free statics with trivial life-times,
+ * but saved line filenames are replaced with a static string.
+ */
+#define	FREE_STATICS false
 
-#define	EXIT_ENOMEM() \
-	do { \
-		fprintf(stderr, PROGNAME "[%d]: " \
-		    "not enough memory (L%d)!\n", getpid(), __LINE__); \
-		_exit(1); \
-	} while (0)
+#define	nitems(arr) (sizeof (arr) / sizeof (*arr))
+#define	STRCMP ((int(*)(const void *, const void *))&strcmp)
 
 
 #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;
-
-				if (debug >= 2)
-					printf(PROGNAME ": forked %d, "
-					    "canmount_on=%d, dataset=%s\n",
-					    (int)pid, canmount_on, dset);
-
-				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;
-				}
+			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 (canmount_on) {
-					canmount_on_pids[forked_canmount_on] =
-					    pid;
-					++forked_canmount_on;
-				} else
-					++forked_canmount_not_on;
-				break;
-			}
+			if (canmount_on)
+				ret |= line_worker(line, cachent->d_name);
+			else {
+				if (filename == NULL)
+					filename =
+					    strdup(cachent->d_name) ?: "(?)";
+
+				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;
+				}
+				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);
 }
diff --git a/man/man8/zfs-mount-generator.8.in b/man/man8/zfs-mount-generator.8.in
index 7aa332ba8174..ae8937038e67 100644
--- a/man/man8/zfs-mount-generator.8.in
+++ b/man/man8/zfs-mount-generator.8.in
@@ -142,22 +142,11 @@ ZEDLET, if enabled
 .Pq see Xr zed 8 .
 .
 .Sh ENVIRONMENT
-The
+If the
 .Sy ZFS_DEBUG
-environment variable can either be
-.Sy 0
-(default),
-.Sy 1
-(print summary accounting information at the end), or at least
-.Sy 2
-(print accounting information for each subprocess as it finishes).
-.
-If not present,
-.Pa /proc/cmdline
-is additionally checked for
-.Qq debug ,
-in which case the debug level is set to
-.Sy 2 .
+environment variable is nonzero
+.Pq or unset and Pa /proc/cmdline No contains Qq Sy debug ,
+print summary accounting information at the end.
 .
 .Sh EXAMPLES
 To begin, enable tracking for the pool:

From 9f6943504aec36f897f814fb7ae5987425436b11 Mon Sep 17 00:00:00 2001
From: Brian Behlendorf <behlendorf1@llnl.gov>
Date: Tue, 30 Nov 2021 10:38:09 -0800
Subject: [PATCH 04/19] Default to zfs_dmu_offset_next_sync=1

Strict hole reporting was previously disabled by default as a
performance optimization.  However, this has lead to confusion
over the expected behavior and a variety of workarounds being
adopted by consumers of ZFS.  Change the default behavior to
always report holes and force the TXG sync.

Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Upstream-commit: 05b3eb6d232009db247882a39d518e7282630753
Ref: #13261
Closes #12746
---
 man/man4/zfs.4   |  8 ++++----
 module/zfs/dmu.c | 12 ++++++++----
 2 files changed, 12 insertions(+), 8 deletions(-)

diff --git a/man/man4/zfs.4 b/man/man4/zfs.4
index 657afc6169b6..3eeed8f439fa 100644
--- a/man/man4/zfs.4
+++ b/man/man4/zfs.4
@@ -1581,12 +1581,12 @@ Allow no-operation writes.
 The occurrence of nopwrites will further depend on other pool properties
 .Pq i.a. the checksumming and compression algorithms .
 .
-.It Sy zfs_dmu_offset_next_sync Ns = Ns Sy 0 Ns | Ns 1 Pq int
+.It Sy zfs_dmu_offset_next_sync Ns = Ns Sy 1 Ns | Ns 0 Pq int
 Enable forcing TXG sync to find holes.
-When enabled forces ZFS to act like prior versions when
+When enabled forces ZFS to sync data when
 .Sy SEEK_HOLE No or Sy SEEK_DATA
-flags are used, which, when a dnode is dirty,
-causes TXGs to be synced so that this data can be found.
+flags are used allowing holes in a file to be accurately reported.
+When disabled holes will not be reported in recently dirtied files.
 .
 .It Sy zfs_pd_bytes_max Ns = Ns Sy 52428800 Ns B Po 50MB Pc Pq int
 The number of bytes which should be prefetched during a pool traversal, like
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
index 0c528f68ccf4..4e7127bd1bab 100644
--- a/module/zfs/dmu.c
+++ b/module/zfs/dmu.c
@@ -73,9 +73,13 @@ int zfs_nopwrite_enabled = 1;
 unsigned long zfs_per_txg_dirty_frees_percent = 5;
 
 /*
- * Enable/disable forcing txg sync when dirty in dmu_offset_next.
+ * Enable/disable forcing txg sync when dirty checking for holes with lseek().
+ * By default this is enabled to ensure accurate hole reporting, it can result
+ * in a significant performance penalty for lseek(SEEK_HOLE) heavy workloads.
+ * Disabling this option will result in holes never being reported in dirty
+ * files which is always safe.
  */
-int zfs_dmu_offset_next_sync = 0;
+int zfs_dmu_offset_next_sync = 1;
 
 /*
  * Limit the amount we can prefetch with one call to this amount.  This
@@ -2110,8 +2114,8 @@ dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off)
 		 * If the zfs_dmu_offset_next_sync module option is enabled
 		 * then strict hole reporting has been requested.  Dirty
 		 * dnodes must be synced to disk to accurately report all
-		 * holes.  When disabled (the default) dirty dnodes are
-		 * reported to not have any holes which is always safe.
+		 * holes.  When disabled dirty dnodes are reported to not
+		 * have any holes which is always safe.
 		 *
 		 * When called by zfs_holey_common() the zp->z_rangelock
 		 * is held to prevent zfs_write() and mmap writeback from

From 5a9994f5ae69448ad604e8184bd2ee8e437ffbca Mon Sep 17 00:00:00 2001
From: Brian Behlendorf <behlendorf1@llnl.gov>
Date: Mon, 11 Oct 2021 10:54:39 -0700
Subject: [PATCH 05/19] Export minimal zfs_refcount interfaces

Lustre makes light use of the zfs_refcount interfaces which
isn't a problem when using a non-debug build of OpenZFS. However,
when debugging is enabled the required symbols are not exported.

Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #12613
---
 module/zfs/refcount.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/module/zfs/refcount.c b/module/zfs/refcount.c
index 354e021d9d26..35a379dded69 100644
--- a/module/zfs/refcount.c
+++ b/module/zfs/refcount.c
@@ -318,6 +318,14 @@ zfs_refcount_not_held(zfs_refcount_t *rc, const void *holder)
 	return (B_TRUE);
 }
 
+EXPORT_SYMBOL(zfs_refcount_create);
+EXPORT_SYMBOL(zfs_refcount_destroy);
+EXPORT_SYMBOL(zfs_refcount_is_zero);
+EXPORT_SYMBOL(zfs_refcount_count);
+EXPORT_SYMBOL(zfs_refcount_add);
+EXPORT_SYMBOL(zfs_refcount_remove);
+EXPORT_SYMBOL(zfs_refcount_held);
+
 /* BEGIN CSTYLED */
 ZFS_MODULE_PARAM(zfs, ,reference_tracking_enable, INT, ZMOD_RW,
 	"Track reference holders to refcount_t objects");

From a5a28723bdd09903c55c4a6aa471e361392ff1b2 Mon Sep 17 00:00:00 2001
From: Ryan Moeller <freqlabs@FreeBSD.org>
Date: Sat, 2 Apr 2022 15:10:55 -0400
Subject: [PATCH 06/19] FreeBSD: Use NDFREE_PNBUF if available

NDF_ONLY_PNBUF has been removed from FreeBSD in favor of NDFREE_PNBUF.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ryan Moeller <freqlabs@FreeBSD.org>
Closes #13277
---
 module/os/freebsd/zfs/zfs_vnops_os.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/module/os/freebsd/zfs/zfs_vnops_os.c b/module/os/freebsd/zfs/zfs_vnops_os.c
index b2cc3d063f9c..9e0de0f6f38a 100644
--- a/module/os/freebsd/zfs/zfs_vnops_os.c
+++ b/module/os/freebsd/zfs/zfs_vnops_os.c
@@ -97,6 +97,10 @@
 
 VFS_SMR_DECLARE;
 
+#if __FreeBSD_version < 1300103
+#define	NDFREE_PNBUF(ndp)	NDFREE((ndp), NDF_ONLY_PNBUF)
+#endif
+
 #if __FreeBSD_version >= 1300047
 #define	vm_page_wire_lock(pp)
 #define	vm_page_wire_unlock(pp)
@@ -5357,7 +5361,7 @@ zfs_getextattr_dir(struct vop_getextattr_args *ap, const char *attrname)
 #endif
 	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_INVFS, ap->a_cred, NULL);
 	vp = nd.ni_vp;
-	NDFREE(&nd, NDF_ONLY_PNBUF);
+	NDFREE_PNBUF(&nd);
 	if (error != 0)
 		return (error);
 
@@ -5475,12 +5479,12 @@ zfs_deleteextattr_dir(struct vop_deleteextattr_args *ap, const char *attrname)
 	error = namei(&nd);
 	vp = nd.ni_vp;
 	if (error != 0) {
-		NDFREE(&nd, NDF_ONLY_PNBUF);
+		NDFREE_PNBUF(&nd);
 		return (error);
 	}
 
 	error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
-	NDFREE(&nd, NDF_ONLY_PNBUF);
+	NDFREE_PNBUF(&nd);
 
 	vput(nd.ni_dvp);
 	if (vp == nd.ni_dvp)
@@ -5605,7 +5609,7 @@ zfs_setextattr_dir(struct vop_setextattr_args *ap, const char *attrname)
 	error = vn_open_cred(&nd, &flags, 0600, VN_OPEN_INVFS, ap->a_cred,
 	    NULL);
 	vp = nd.ni_vp;
-	NDFREE(&nd, NDF_ONLY_PNBUF);
+	NDFREE_PNBUF(&nd);
 	if (error != 0)
 		return (error);
 
@@ -5760,7 +5764,7 @@ zfs_listextattr_dir(struct vop_listextattr_args *ap, const char *attrprefix)
 #endif
 	error = namei(&nd);
 	vp = nd.ni_vp;
-	NDFREE(&nd, NDF_ONLY_PNBUF);
+	NDFREE_PNBUF(&nd);
 	if (error != 0)
 		return (error);
 

From 215a8255a97f32cb26881d6fb4f883b7608329d0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Thu, 24 Mar 2022 15:22:53 +0100
Subject: [PATCH 07/19] Linux 5.18 compat: 4-argument bio_alloc()
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

  bio_alloc(gfp_t gfp_mask, unsigned short nr_iovecs)

became

  bio_alloc(struct block_device *bdev, unsigned short nr_vecs,
            unsigned int opf, gfp_t gfp_mask)
passing NULL/0 continues previous behaviour

Upstream-commit: 07888c665b405b1cd3577ddebfeb74f4717a84c4 ("block:
 pass a block_device and opf to bio_alloc")

Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Closes #13251
---
 config/kernel-bio.m4            | 37 +++++++++++++++++++++++++++++++++
 module/os/linux/zfs/vdev_disk.c |  4 ++++
 2 files changed, 41 insertions(+)

diff --git a/config/kernel-bio.m4 b/config/kernel-bio.m4
index d088d7023cb0..b5d25448155f 100644
--- a/config/kernel-bio.m4
+++ b/config/kernel-bio.m4
@@ -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([for 4-argument bio_alloc()])
+	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
 ])
diff --git a/module/os/linux/zfs/vdev_disk.c b/module/os/linux/zfs/vdev_disk.c
index 581a790865b7..23bb7c11d2ec 100644
--- a/module/os/linux/zfs/vdev_disk.c
+++ b/module/os/linux/zfs/vdev_disk.c
@@ -563,6 +563,10 @@ vdev_submit_bio(struct bio *bio)
 	current->bio_list = bio_list;
 }
 
+#ifdef HAVE_BIO_ALLOC_4ARG
+#define	bio_alloc(gfp_mask, nr_iovecs) bio_alloc(NULL, nr_iovecs, 0, gfp_mask)
+#endif
+
 static int
 __vdev_disk_physio(struct block_device *bdev, zio_t *zio,
     size_t io_size, uint64_t io_offset, int rw, int flags)

From 9f7f70450791822518e6e9dc515a4ded47e74c51 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Thu, 24 Mar 2022 16:33:40 +0100
Subject: [PATCH 08/19] Linux 5.18 compat: replace genhd.h with blkdev.h
 includes
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

blkdev.h includes genhd.h since dawn of upstream git, so this is
globally safe

Upstream-commit: 322cbb50de711814c42fb088f6d31901502c711a ("block:
 remove genhd.h")

Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Closes #13251
---
 config/kernel-add-disk.m4             | 5 ++---
 config/kernel-revalidate-disk-size.m4 | 4 ++--
 module/os/linux/zfs/zvol_os.c         | 2 +-
 3 files changed, 5 insertions(+), 6 deletions(-)

diff --git a/config/kernel-add-disk.m4 b/config/kernel-add-disk.m4
index 5d1779eb4328..44a8a5fd25b6 100644
--- a/config/kernel-add-disk.m4
+++ b/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],
diff --git a/config/kernel-revalidate-disk-size.m4 b/config/kernel-revalidate-disk-size.m4
index a7d0cb3cdab4..13cb92a174e3 100644
--- a/config/kernel-revalidate-disk-size.m4
+++ b/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);
diff --git a/module/os/linux/zfs/zvol_os.c b/module/os/linux/zfs/zvol_os.c
index 7a979eb91e7f..b1321df8da4b 100644
--- a/module/os/linux/zfs/zvol_os.c
+++ b/module/os/linux/zfs/zvol_os.c
@@ -906,7 +906,7 @@ zvol_alloc(dev_t dev, const char *name)
 	if (volmode == ZFS_VOLMODE_DEV) {
 		/*
 		 * ZFS_VOLMODE_DEV disable partitioning on ZVOL devices: set
-		 * gendisk->minors = 1 as noted in include/linux/genhd.h.
+		 * gendisk->minors = 1 as noted in include/linux/blkdev.h.
 		 * Also disable extended partition numbers (GENHD_FL_EXT_DEVT)
 		 * and suppresses partition scanning (GENHD_FL_NO_PART_SCAN)
 		 * setting gendisk->flags accordingly.

From 10a9f5fc47acc6c9d3475023001264ca5fb741a6 Mon Sep 17 00:00:00 2001
From: Riccardo Schirone <rschirone91@gmail.com>
Date: Fri, 1 Apr 2022 23:47:36 +0200
Subject: [PATCH 09/19] Linux 5.18 compat: blkg_tryget is moved to private
 headers

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Riccardo Schirone <rschirone91@gmail.com>
Closes #13278
---
 module/os/linux/zfs/vdev_disk.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/module/os/linux/zfs/vdev_disk.c b/module/os/linux/zfs/vdev_disk.c
index 23bb7c11d2ec..2708535b3b88 100644
--- a/module/os/linux/zfs/vdev_disk.c
+++ b/module/os/linux/zfs/vdev_disk.c
@@ -467,8 +467,11 @@ vdev_submit_bio_impl(struct bio *bio)
  * blkg_tryget() to use rcu_read_lock() instead of rcu_read_lock_sched().
  * As a side effect the function was converted to GPL-only.  Define our
  * own version when needed which uses rcu_read_lock_sched().
+ *
+ * The Linux 5.17 kernel split linux/blk-cgroup.h into a private and a public
+ * part, moving blkg_tryget into the private one. Define our own version.
  */
-#if defined(HAVE_BLKG_TRYGET_GPL_ONLY)
+#if defined(HAVE_BLKG_TRYGET_GPL_ONLY) || !defined(HAVE_BLKG_TRYGET)
 static inline bool
 vdev_blkg_tryget(struct blkcg_gq *blkg)
 {
@@ -493,7 +496,7 @@ vdev_blkg_tryget(struct blkcg_gq *blkg)
 
 	return (rc);
 }
-#elif defined(HAVE_BLKG_TRYGET)
+#else
 #define	vdev_blkg_tryget(bg)	blkg_tryget(bg)
 #endif
 #ifdef HAVE_BIO_SET_DEV_MACRO

From 35ddd8ee2e5fdbaa5b0d94c063bbc1a79ab12cb1 Mon Sep 17 00:00:00 2001
From: Riccardo Schirone <rschirone91@gmail.com>
Date: Sat, 2 Apr 2022 01:15:25 +0200
Subject: [PATCH 10/19] Linux 5.18 compat: use
 address_space_operations->readahead

->readpages was removed and replaced by ->readahead. Define
zpl_readahead for kernels that don't have ->readpages.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Riccardo Schirone <rschirone91@gmail.com>
Closes #13278
---
 config/kernel-readpages.m4     | 25 +++++++++++++++++++++++++
 config/kernel.m4               |  2 ++
 module/os/linux/zfs/zpl_file.c | 21 +++++++++++++++++++++
 3 files changed, 48 insertions(+)
 create mode 100644 config/kernel-readpages.m4

diff --git a/config/kernel-readpages.m4 b/config/kernel-readpages.m4
new file mode 100644
index 000000000000..0bf67ffd933d
--- /dev/null
+++ b/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([address_space_operations->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])
+	])
+])
diff --git a/config/kernel.m4 b/config/kernel.m4
index 3122e9dbaa94..7e546b58a77c 100644
--- a/config/kernel.m4
+++ b/config/kernel.m4
@@ -132,6 +132,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
 	ZFS_AC_KERNEL_SRC_SIGNAL_STOP
 	ZFS_AC_KERNEL_SRC_SIGINFO
 	ZFS_AC_KERNEL_SRC_SET_SPECIAL_STATE
+	ZFS_AC_KERNEL_SRC_VFS_READPAGES
 	ZFS_AC_KERNEL_SRC_VFS_SET_PAGE_DIRTY_NOBUFFERS
 	ZFS_AC_KERNEL_SRC_STANDALONE_LINUX_STDARG
 	ZFS_AC_KERNEL_SRC_PAGEMAP_FOLIO_WAIT_BIT
@@ -242,6 +243,7 @@ AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
 	ZFS_AC_KERNEL_SIGNAL_STOP
 	ZFS_AC_KERNEL_SIGINFO
 	ZFS_AC_KERNEL_SET_SPECIAL_STATE
+	ZFS_AC_KERNEL_VFS_READPAGES
 	ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS
 	ZFS_AC_KERNEL_STANDALONE_LINUX_STDARG
 	ZFS_AC_KERNEL_PAGEMAP_FOLIO_WAIT_BIT
diff --git a/module/os/linux/zfs/zpl_file.c b/module/os/linux/zfs/zpl_file.c
index f1241c443797..4965815ece17 100644
--- a/module/os/linux/zfs/zpl_file.c
+++ b/module/os/linux/zfs/zpl_file.c
@@ -647,12 +647,29 @@ zpl_readpage_filler(void *data, struct page *pp)
  * paging.  For simplicity, the code relies on read_cache_pages() to
  * correctly lock each page for IO and call zpl_readpage().
  */
+#ifdef HAVE_VFS_READPAGES
 static int
 zpl_readpages(struct file *filp, struct address_space *mapping,
     struct list_head *pages, unsigned nr_pages)
 {
 	return (read_cache_pages(mapping, pages, zpl_readpage_filler, NULL));
 }
+#else
+static void
+zpl_readahead(struct readahead_control *ractl)
+{
+	struct page *page;
+
+	while ((page = readahead_page(ractl)) != NULL) {
+		int ret;
+
+		ret = zpl_readpage_filler(NULL, page);
+		put_page(page);
+		if (ret)
+			break;
+	}
+}
+#endif
 
 static int
 zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
@@ -1027,7 +1044,11 @@ zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 
 
 const struct address_space_operations zpl_address_space_operations = {
+#ifdef HAVE_VFS_READPAGES
 	.readpages	= zpl_readpages,
+#else
+	.readahead	= zpl_readahead,
+#endif
 	.readpage	= zpl_readpage,
 	.writepage	= zpl_writepage,
 	.writepages	= zpl_writepages,

From 1f4c79b1ce0b27b747149b2820fa5f6f051b7d4b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Mon, 4 Apr 2022 14:25:01 +0200
Subject: [PATCH 11/19] libzfs: sendrecv: always cancel progress thread in
 zfs_send_one()
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This is in line with all the other uses of the progress thread

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Closes #11560
Closes #13284
---
 lib/libzfs/libzfs_sendrecv.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/lib/libzfs/libzfs_sendrecv.c b/lib/libzfs/libzfs_sendrecv.c
index b67c9b30c84a..60050e22a31a 100644
--- a/lib/libzfs/libzfs_sendrecv.c
+++ b/lib/libzfs/libzfs_sendrecv.c
@@ -2522,8 +2522,7 @@ zfs_send_one(zfs_handle_t *zhp, const char *from, int fd, sendflags_t *flags,
 
 	if (flags->progress) {
 		void *status = NULL;
-		if (err != 0)
-			(void) pthread_cancel(ptid);
+		(void) pthread_cancel(ptid);
 		(void) pthread_join(ptid, &status);
 		int error = (int)(uintptr_t)status;
 		if (error != 0 && status != PTHREAD_CANCELED)

From ff23ef0c99a71e8697b4a39886b0f62b4ecf935d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Thu, 7 Apr 2022 04:32:27 +0200
Subject: [PATCH 12/19] libzfs: import: zpool_clear_label: actually fail if
 clearing l2arc header fails
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Found with -Wunused-but-set-variable on Clang trunk

Upstream-commit: a4e0cee1780cbd8f2cb9a263a0ed8d91dbe68b4a
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Closes #13304
---
 lib/libzfs/libzfs_import.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/lib/libzfs/libzfs_import.c b/lib/libzfs/libzfs_import.c
index 64fa31c67d0f..ddaa5de5db11 100644
--- a/lib/libzfs/libzfs_import.c
+++ b/lib/libzfs/libzfs_import.c
@@ -146,10 +146,10 @@ zpool_clear_label(int fd)
 	struct stat64 statbuf;
 	int l;
 	vdev_label_t *label;
-	l2arc_dev_hdr_phys_t *l2dhdr;
+	l2arc_dev_hdr_phys_t *l2dhdr = NULL;
 	uint64_t size;
-	int labels_cleared = 0, header_cleared = 0;
-	boolean_t clear_l2arc_header = B_FALSE;
+	int labels_cleared = 0;
+	boolean_t clear_l2arc_header = B_FALSE, header_cleared = B_FALSE;
 
 	if (fstat64_blk(fd, &statbuf) == -1)
 		return (0);
@@ -219,13 +219,10 @@ zpool_clear_label(int fd)
 	}
 
 	/* Clear the L2ARC header. */
-	if (clear_l2arc_header) {
-		memset(l2dhdr, 0, sizeof (l2arc_dev_hdr_phys_t));
-		if (pwrite64(fd, l2dhdr, sizeof (l2arc_dev_hdr_phys_t),
-		    VDEV_LABEL_START_SIZE) == sizeof (l2arc_dev_hdr_phys_t)) {
-			header_cleared++;
-		}
-	}
+	if (clear_l2arc_header &&
+	    pwrite64(fd, l2dhdr, sizeof (l2arc_dev_hdr_phys_t),
+	    VDEV_LABEL_START_SIZE) == sizeof (l2arc_dev_hdr_phys_t))
+			header_cleared = B_TRUE;
 
 	free(label);
 	free(l2dhdr);
@@ -233,6 +230,9 @@ zpool_clear_label(int fd)
 	if (labels_cleared == 0)
 		return (-1);
 
+	if (clear_l2arc_header && !header_cleared)
+		return (-1);
+
 	return (0);
 }
 

From e9cd90f6e5953df79ae5db84e3b23f339d6098f7 Mon Sep 17 00:00:00 2001
From: Mark Johnston <markj@FreeBSD.org>
Date: Thu, 7 Apr 2022 17:11:00 -0400
Subject: [PATCH 13/19] FreeBSD: Parameterize ZFS_ENTER/ZFS_VERIFY_VP with an
 error code

For legacy reasons, a couple of VOPs have to return error numbers that
don't come from the usual errno namespace.  To handle the cases where
ZFS_ENTER or ZFS_VERIFY_ZP fail, we need to be able to override the
default error return value of EIO.  Extend the macros to permit this.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes #13311
---
 include/os/freebsd/zfs/sys/zfs_znode_impl.h | 32 ++++++++++++---------
 module/os/freebsd/zfs/zfs_vnops_os.c        |  2 +-
 2 files changed, 19 insertions(+), 15 deletions(-)

diff --git a/include/os/freebsd/zfs/sys/zfs_znode_impl.h b/include/os/freebsd/zfs/sys/zfs_znode_impl.h
index edb28d041a0c..3d93525b45ab 100644
--- a/include/os/freebsd/zfs/sys/zfs_znode_impl.h
+++ b/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
diff --git a/module/os/freebsd/zfs/zfs_vnops_os.c b/module/os/freebsd/zfs/zfs_vnops_os.c
index 9e0de0f6f38a..75a4fd183310 100644
--- a/module/os/freebsd/zfs/zfs_vnops_os.c
+++ b/module/os/freebsd/zfs/zfs_vnops_os.c
@@ -5434,7 +5434,7 @@ zfs_getextattr(struct vop_getextattr_args *ap)
 
 	error = ENOENT;
 	ZFS_ENTER(zfsvfs);
-	ZFS_VERIFY_ZP(zp)
+	ZFS_VERIFY_ZP(zp);
 	rw_enter(&zp->z_xattr_lock, RW_READER);
 	if (zfsvfs->z_use_sa && zp->z_is_sa)
 		error = zfs_getextattr_sa(ap, attrname);

From b7546f92eaeb295cf0302662773792a2a8d359a5 Mon Sep 17 00:00:00 2001
From: Mark Johnston <markj@FreeBSD.org>
Date: Thu, 7 Apr 2022 17:13:18 -0400
Subject: [PATCH 14/19] FreeBSD: Return Mach error codes from
 VOP_(GET|PUT)PAGES

FreeBSD's memory management system uses its own error numbers and gets
confused when these VOPs return EIO.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Reported-by: Peter Holm <pho@FreeBSD.org>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes #13311
---
 module/os/freebsd/zfs/zfs_vnops_os.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/module/os/freebsd/zfs/zfs_vnops_os.c b/module/os/freebsd/zfs/zfs_vnops_os.c
index 75a4fd183310..d273e70981db 100644
--- a/module/os/freebsd/zfs/zfs_vnops_os.c
+++ b/module/os/freebsd/zfs/zfs_vnops_os.c
@@ -4062,8 +4062,8 @@ zfs_getpages(struct vnode *vp, vm_page_t *ma, int count, int *rbehind,
 	int pgsin_b, pgsin_a;
 	int error;
 
-	ZFS_ENTER(zfsvfs);
-	ZFS_VERIFY_ZP(zp);
+	ZFS_ENTER_ERROR(zfsvfs, zfs_vm_pagerret_error);
+	ZFS_VERIFY_ZP_ERROR(zp, zfs_vm_pagerret_error);
 
 	start = IDX_TO_OFF(ma[0]->pindex);
 	end = IDX_TO_OFF(ma[count - 1]->pindex + 1);
@@ -4187,19 +4187,18 @@ zfs_putpages(struct vnode *vp, vm_page_t *ma, size_t len, int flags,
 	int		err;
 	int		i;
 
-	ZFS_ENTER(zfsvfs);
-	ZFS_VERIFY_ZP(zp);
-
 	object = vp->v_object;
-	pcount = btoc(len);
-	ncount = pcount;
-
 	KASSERT(ma[0]->object == object, ("mismatching object"));
 	KASSERT(len > 0 && (len & PAGE_MASK) == 0, ("unexpected length"));
 
+	pcount = btoc(len);
+	ncount = pcount;
 	for (i = 0; i < pcount; i++)
 		rtvals[i] = zfs_vm_pagerret_error;
 
+	ZFS_ENTER_ERROR(zfsvfs, zfs_vm_pagerret_error);
+	ZFS_VERIFY_ZP_ERROR(zp, zfs_vm_pagerret_error);
+
 	off = IDX_TO_OFF(ma[0]->pindex);
 	blksz = zp->z_blksz;
 	lo_off = rounddown(off, blksz);

From aa1c3c1d1d829a877e56c8da6390d40f7656e8bb Mon Sep 17 00:00:00 2001
From: Brian Behlendorf <behlendorf1@llnl.gov>
Date: Tue, 19 Apr 2022 10:38:04 -0700
Subject: [PATCH 15/19] Linux 5.17 compat: GENHD_FL_EXT_DEVT /
 GENHD_FL_NO_PART_SCAN

As of the 5.17 kernel the GENHD_FL_EXT_DEVT flag has been removed
and the GENHD_FL_NO_PART_SCAN flag renamed GENHD_FL_NO_PART. Update
zvol_alloc() to set GENHD_FL_NO_PART for the newer kernels which
is sufficient.  The behavior for prior kernels remains unchanged.

1ebe2e5f ("block: remove GENHD_FL_EXT_DEVT")
46e7eac6 ("block: rename GENHD_FL_NO_PART_SCAN to GENHD_FL_NO_PART")

Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #13294
Closes #13297
---
 config/kernel-genhd-flags.m4  | 58 +++++++++++++++++++++++++++++++++++
 config/kernel.m4              |  2 ++
 module/os/linux/zfs/zvol_os.c | 21 +++++--------
 3 files changed, 68 insertions(+), 13 deletions(-)
 create mode 100644 config/kernel-genhd-flags.m4

diff --git a/config/kernel-genhd-flags.m4 b/config/kernel-genhd-flags.m4
new file mode 100644
index 000000000000..af6a8a086bc9
--- /dev/null
+++ b/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])
+		])
+	])
+])
diff --git a/config/kernel.m4 b/config/kernel.m4
index 7e546b58a77c..0ef7c7d0caa7 100644
--- a/config/kernel.m4
+++ b/config/kernel.m4
@@ -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
@@ -172,6 +173,7 @@ 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
diff --git a/module/os/linux/zfs/zvol_os.c b/module/os/linux/zfs/zvol_os.c
index b1321df8da4b..5c509a06a4fb 100644
--- a/module/os/linux/zfs/zvol_os.c
+++ b/module/os/linux/zfs/zvol_os.c
@@ -903,22 +903,17 @@ zvol_alloc(dev_t dev, const char *name)
 	zso->zvo_disk->major = zvol_major;
 	zso->zvo_disk->events = DISK_EVENT_MEDIA_CHANGE;
 
+	/*
+	 * Setting ZFS_VOLMODE_DEV disables partitioning on ZVOL devices.
+	 * This is accomplished by limiting the number of minors for the
+	 * device to one and explicitly disabling partition scanning.
+	 */
 	if (volmode == ZFS_VOLMODE_DEV) {
-		/*
-		 * ZFS_VOLMODE_DEV disable partitioning on ZVOL devices: set
-		 * gendisk->minors = 1 as noted in include/linux/blkdev.h.
-		 * Also disable extended partition numbers (GENHD_FL_EXT_DEVT)
-		 * and suppresses partition scanning (GENHD_FL_NO_PART_SCAN)
-		 * setting gendisk->flags accordingly.
-		 */
 		zso->zvo_disk->minors = 1;
-#if defined(GENHD_FL_EXT_DEVT)
-		zso->zvo_disk->flags &= ~GENHD_FL_EXT_DEVT;
-#endif
-#if defined(GENHD_FL_NO_PART_SCAN)
-		zso->zvo_disk->flags |= GENHD_FL_NO_PART_SCAN;
-#endif
+		zso->zvo_disk->flags &= ~ZFS_GENHD_FL_EXT_DEVT;
+		zso->zvo_disk->flags |= ZFS_GENHD_FL_NO_PART;
 	}
+
 	zso->zvo_disk->first_minor = (dev & MINORMASK);
 	zso->zvo_disk->fops = &zvol_ops;
 	zso->zvo_disk->private_data = zv;

From 361dc138b1bdd6b7ec32fbd74b775c55c595f45f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Sat, 16 Apr 2022 16:07:04 +0200
Subject: [PATCH 16/19] Document zfs inherit -S's interaction with
 noninheritable properties
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Damian Szuberski <szuberskidamian@gmail.com>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Upstream-commit: 92295af8004b07a58e0860d66dd565143486a757
Closes #11894
Closes #13335
---
 man/man8/zfs-set.8 | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/man/man8/zfs-set.8 b/man/man8/zfs-set.8
index a3588cc26638..ccd90f0917ea 100644
--- a/man/man8/zfs-set.8
+++ b/man/man8/zfs-set.8
@@ -170,8 +170,9 @@ inherited.
 .It Fl r
 Recursively inherit the given property for all children.
 .It Fl S
-Revert the property to the received value if one exists; otherwise operate as
-if the
+Revert the property to the received value, if one exists;
+otherwise, for non-inheritable properties, to the default;
+otherwise, operate as if the
 .Fl S
 option was not specified.
 .El

From c220771a47e4206fb43e6849957657c9504b1b14 Mon Sep 17 00:00:00 2001
From: Rich Ercolani <214141+rincebrain@users.noreply.github.com>
Date: Wed, 20 Apr 2022 19:07:03 -0400
Subject: [PATCH 17/19] Corrected oversight in ZERO_RANGE behavior

It turns out, no, in fact, ZERO_RANGE and PUNCH_HOLE do
have differing semantics in some ways - in particular,
one requires KEEP_SIZE, and the other does not.

Also added a zero-range test to catch this, corrected a flaw
that made the punch-hole test succeed vacuously, and a typo
in file_write.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Rich Ercolani <rincebrain@gmail.com>
Closes #13329
Closes #13338
---
 module/os/linux/zfs/zpl_file.c                |  10 +-
 tests/runfiles/linux.run                      |   2 +-
 tests/zfs-tests/cmd/file_write/file_write.c   |   2 +-
 tests/zfs-tests/include/libtest.shlib         |  16 +++
 .../tests/functional/fallocate/Makefile.am    |   3 +-
 .../fallocate/fallocate_punch-hole.ksh        |  35 ++++--
 .../fallocate/fallocate_zero-range.ksh        | 119 ++++++++++++++++++
 .../tests/functional/fallocate/setup.ksh      |   5 +-
 8 files changed, 171 insertions(+), 21 deletions(-)
 create mode 100755 tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh

diff --git a/module/os/linux/zfs/zpl_file.c b/module/os/linux/zfs/zpl_file.c
index 4965815ece17..1ed41b9c4c20 100644
--- a/module/os/linux/zfs/zpl_file.c
+++ b/module/os/linux/zfs/zpl_file.c
@@ -781,11 +781,13 @@ zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
 	if (mode & (test_mode)) {
 		flock64_t bf;
 
-		if (offset > olen)
-			goto out_unmark;
+		if (mode & FALLOC_FL_KEEP_SIZE) {
+			if (offset > olen)
+				goto out_unmark;
 
-		if (offset + len > olen)
-			len = olen - offset;
+			if (offset + len > olen)
+				len = olen - offset;
+		}
 		bf.l_type = F_WRLCK;
 		bf.l_whence = SEEK_SET;
 		bf.l_start = offset;
diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run
index c01e1e3c4d53..94c1cbbc3f9f 100644
--- a/tests/runfiles/linux.run
+++ b/tests/runfiles/linux.run
@@ -94,7 +94,7 @@ tests = ['events_001_pos', 'events_002_pos', 'zed_rc_filter', 'zed_fd_spill']
 tags = ['functional', 'events']
 
 [tests/functional/fallocate:Linux]
-tests = ['fallocate_prealloc']
+tests = ['fallocate_prealloc', 'fallocate_zero-range']
 tags = ['functional', 'fallocate']
 
 [tests/functional/fault:Linux]
diff --git a/tests/zfs-tests/cmd/file_write/file_write.c b/tests/zfs-tests/cmd/file_write/file_write.c
index 60893c34fbc9..9d2e71b679eb 100644
--- a/tests/zfs-tests/cmd/file_write/file_write.c
+++ b/tests/zfs-tests/cmd/file_write/file_write.c
@@ -251,7 +251,7 @@ usage(char *prog)
 	    "\t[-s offset] [-c write_count] [-d data]\n\n"
 	    "Where [data] equal to zero causes chars "
 	    "0->%d to be repeated throughout, or [data]\n"
-	    "equal to 'R' for psudorandom data.\n",
+	    "equal to 'R' for pseudorandom data.\n",
 	    prog, DATA_RANGE);
 
 	exit(1);
diff --git a/tests/zfs-tests/include/libtest.shlib b/tests/zfs-tests/include/libtest.shlib
index dd43b02a6868..94ab7ffd20f7 100644
--- a/tests/zfs-tests/include/libtest.shlib
+++ b/tests/zfs-tests/include/libtest.shlib
@@ -4236,6 +4236,22 @@ function punch_hole # offset length file
 	esac
 }
 
+function zero_range # offset length file
+{
+	typeset offset=$1
+	typeset length=$2
+	typeset file=$3
+
+	case "$UNAME" in
+	Linux)
+		fallocate --zero-range --offset $offset --length $length "$file"
+		;;
+	*)
+		false
+		;;
+	esac
+}
+
 #
 # Wait for the specified arcstat to reach non-zero quiescence.
 # If echo is 1 echo the value after reaching quiescence, otherwise
diff --git a/tests/zfs-tests/tests/functional/fallocate/Makefile.am b/tests/zfs-tests/tests/functional/fallocate/Makefile.am
index 5ff366d2482c..86364d7895dd 100644
--- a/tests/zfs-tests/tests/functional/fallocate/Makefile.am
+++ b/tests/zfs-tests/tests/functional/fallocate/Makefile.am
@@ -3,4 +3,5 @@ dist_pkgdata_SCRIPTS = \
 	setup.ksh \
 	cleanup.ksh \
 	fallocate_prealloc.ksh \
-	fallocate_punch-hole.ksh
+	fallocate_punch-hole.ksh \
+	fallocate_zero-range.ksh
diff --git a/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh b/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh
index ed83561bd556..92f4552f5bd7 100755
--- a/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh
+++ b/tests/zfs-tests/tests/functional/fallocate/fallocate_punch-hole.ksh
@@ -60,13 +60,17 @@ function cleanup
 	[[ -e $TESTDIR ]] && log_must rm -f $FILE
 }
 
-function check_disk_size
+function check_reported_size
 {
 	typeset expected_size=$1
 
-	disk_size=$(du $TESTDIR/file | awk '{print $1}')
-	if [ $disk_size -ne $expected_size ]; then
-		log_fail "Incorrect size: $disk_size != $expected_size"
+	if ! [ -e "${FILE}" ]; then
+		log_fail "$FILE does not exist"
+	fi
+		
+	reported_size=$(du "${FILE}" | awk '{print $1}')
+	if [ "$reported_size" != "$expected_size" ]; then
+		log_fail "Incorrect reported size: $reported_size != $expected_size"
 	fi
 }
 
@@ -74,9 +78,9 @@ function check_apparent_size
 {
 	typeset expected_size=$1
 
-	apparent_size=$(stat_size)
-	if [ $apparent_size -ne $expected_size ]; then
-		log_fail "Incorrect size: $apparent_size != $expected_size"
+	apparent_size=$(stat_size "${FILE}")
+	if [ "$apparent_size" != "$expected_size" ]; then
+		log_fail "Incorrect apparent size: $apparent_size != $expected_size"
 	fi
 }
 
@@ -86,25 +90,30 @@ log_onexit cleanup
 
 # Create a dense file and check it is the correct size.
 log_must file_write -o create -f $FILE -b $BLKSZ -c 8
-log_must check_disk_size  $((131072 * 8))
+sync_pool $TESTPOOL
+log_must check_reported_size 1027
 
 # Punch a hole for the first full block.
 log_must punch_hole 0 $BLKSZ $FILE
-log_must check_disk_size  $((131072 * 7))
+sync_pool $TESTPOOL
+log_must check_reported_size 899
 
 # Partially punch a hole in the second block.
 log_must punch_hole $BLKSZ $((BLKSZ / 2)) $FILE
-log_must check_disk_size  $((131072 * 7))
+sync_pool $TESTPOOL
+log_must check_reported_size 899
 
-# Punch a hole which overlaps the third and forth block.
+# Punch a hole which overlaps the third and fourth block.
 log_must punch_hole $(((BLKSZ * 2) + (BLKSZ / 2))) $((BLKSZ)) $FILE
-log_must check_disk_size  $((131072 * 7))
+sync_pool $TESTPOOL
+log_must check_reported_size 899
 
 # Punch a hole from the fifth block past the end of file.  The apparent
 # file size should not change since --keep-size is implied.
 apparent_size=$(stat_size $FILE)
 log_must punch_hole $((BLKSZ * 4)) $((BLKSZ * 10)) $FILE
-log_must check_disk_size  $((131072 * 4))
+sync_pool $TESTPOOL
+log_must check_reported_size 387
 log_must check_apparent_size $apparent_size
 
 log_pass "Ensure holes can be punched in files making them sparse"
diff --git a/tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh b/tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh
new file mode 100755
index 000000000000..e907b0f5d4c4
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/fallocate/fallocate_zero-range.ksh
@@ -0,0 +1,119 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or http://www.opensolaris.org/os/licensing.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2020 by Lawrence Livermore National Security, LLC.
+# Copyright (c) 2021 by The FreeBSD Foundation.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+#
+# DESCRIPTION:
+# Test FALLOC_FL_ZERO_RANGE functionality
+#
+# STRATEGY:
+# 1. Create a dense file
+# 2. Zero various ranges in the file and verify the result.
+#
+
+verify_runnable "global"
+
+if is_freebsd; then
+	log_unsupported "FreeBSD does not implement an analogue to ZERO_RANGE."
+fi
+
+FILE=$TESTDIR/$TESTFILE0
+BLKSZ=$(get_prop recordsize $TESTPOOL)
+
+function cleanup
+{
+	[[ -e $TESTDIR ]] && log_must rm -f $FILE
+}
+
+# Helpfully, this function expects kilobytes, and check_apparent_size expects bytes.
+function check_reported_size
+{
+	typeset expected_size=$1
+
+	if ! [ -e "${FILE}" ]; then
+		log_fail "$FILE does not exist"
+	fi
+		
+	reported_size=$(du "${FILE}" | awk '{print $1}')
+	if [ "$reported_size" != "$expected_size" ]; then
+		log_fail "Incorrect reported size: $reported_size != $expected_size"
+	fi
+}
+
+function check_apparent_size
+{
+	typeset expected_size=$1
+
+	apparent_size=$(stat_size "${FILE}")
+	if [ "$apparent_size" != "$expected_size" ]; then
+		log_fail "Incorrect apparent size: $apparent_size != $expected_size"
+	fi
+}
+
+log_assert "Ensure ranges can be zeroed in files"
+
+log_onexit cleanup
+
+# Create a dense file and check it is the correct size.
+log_must file_write -o create -f $FILE -b $BLKSZ -c 8
+sync_pool $TESTPOOL
+log_must check_reported_size 1027
+
+# Zero a range covering the first full block.
+log_must zero_range 0 $BLKSZ $FILE
+sync_pool $TESTPOOL
+log_must check_reported_size 899
+
+# Partially zero a range in the second block.
+log_must zero_range $BLKSZ $((BLKSZ / 2)) $FILE
+sync_pool $TESTPOOL
+log_must check_reported_size 899
+
+# Zero range which overlaps the third and fourth block.
+log_must zero_range $(((BLKSZ * 2) + (BLKSZ / 2))) $((BLKSZ)) $FILE
+sync_pool $TESTPOOL
+log_must check_reported_size 899
+
+# Zero range from the fifth block past the end of file, with --keep-size.
+# The apparent file size must not change, since we did specify --keep-size.
+apparent_size=$(stat_size $FILE)
+log_must fallocate --keep-size --zero-range --offset $((BLKSZ * 4)) --length $((BLKSZ * 10)) "$FILE"
+sync_pool $TESTPOOL
+log_must check_reported_size 387
+log_must check_apparent_size $apparent_size
+
+# Zero range from the fifth block past the end of file.  The apparent
+# file size should change since --keep-size is not implied, unlike
+# with PUNCH_HOLE.
+apparent_size=$(stat_size $FILE)
+log_must zero_range $((BLKSZ * 4)) $((BLKSZ * 10)) $FILE
+sync_pool $TESTPOOL
+log_must check_reported_size 387
+log_must check_apparent_size $((BLKSZ * 14))
+
+log_pass "Ensure ranges can be zeroed in files"
diff --git a/tests/zfs-tests/tests/functional/fallocate/setup.ksh b/tests/zfs-tests/tests/functional/fallocate/setup.ksh
index 32334d396865..586ac026aa43 100755
--- a/tests/zfs-tests/tests/functional/fallocate/setup.ksh
+++ b/tests/zfs-tests/tests/functional/fallocate/setup.ksh
@@ -26,4 +26,7 @@
 . $STF_SUITE/include/libtest.shlib
 
 DISK=${DISKS%% *}
-default_setup $DISK
+default_setup_noexit $DISK
+log_must zfs set compression=off $TESTPOOL
+log_pass
+

From 972637dc06a04432dc58e240b8ef3e9f538b98bb Mon Sep 17 00:00:00 2001
From: Alexander Motin <mav@FreeBSD.org>
Date: Mon, 18 Apr 2022 20:25:49 -0400
Subject: [PATCH 18/19] FreeBSD: Fix translation from ABD to physical pages.

In hypothetical case of non-linear ABD with single segment, multiple
to page size but not aligned to it, vdev_geom_fill_unmap_cb() could
fill one page less into bio_ma array.

I am not sure it is expoitable, but better to be safe than sorry.

Reported-by: Mark Johnston <markj@FreeBSD.org>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
(cherry picked from commit 5352f85cddce44e82fb1c4caec3b333e3666d7fd)
---
 module/os/freebsd/zfs/vdev_geom.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/module/os/freebsd/zfs/vdev_geom.c b/module/os/freebsd/zfs/vdev_geom.c
index 2ef4811a8a4e..5447eb922062 100644
--- a/module/os/freebsd/zfs/vdev_geom.c
+++ b/module/os/freebsd/zfs/vdev_geom.c
@@ -1132,8 +1132,12 @@ vdev_geom_fill_unmap_cb(void *buf, size_t len, void *priv)
 	vm_offset_t addr = (vm_offset_t)buf;
 	vm_offset_t end = addr + len;
 
-	if (bp->bio_ma_n == 0)
+	if (bp->bio_ma_n == 0) {
 		bp->bio_ma_offset = addr & PAGE_MASK;
+		addr &= ~PAGE_MASK;
+	} else {
+		ASSERT0(P2PHASE(addr, PAGE_SIZE));
+	}
 	do {
 		bp->bio_ma[bp->bio_ma_n++] =
 		    PHYS_TO_VM_PAGE(pmap_kextract(addr));

From 642426095ae165ce64896c11ec52cab3ec3073ac Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=D0=BD=D0=B0=D0=B1?= <nabijaczleweli@nabijaczleweli.xyz>
Date: Thu, 21 Apr 2022 16:27:15 +0200
Subject: [PATCH 19/19] Linux 5.18 compat: kobj_type.default_attrs replaced
 with default_groups
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Upstream-commit: cdb4f26a63c391317e335e6e683a614358e70aeb ("kobject:
 kobj_type: remove default_attrs")
Upstream-commit: 0cdda2edb3f312d56518934811960ad57564c1bb
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Closes #13357
---
 config/kernel-sysfs.m4          | 37 +++++++++++++++++++++++++++++++
 config/kernel.m4                |  2 ++
 module/os/linux/zfs/zfs_sysfs.c | 39 +++++++++++++++++++--------------
 3 files changed, 61 insertions(+), 17 deletions(-)
 create mode 100644 config/kernel-sysfs.m4

diff --git a/config/kernel-sysfs.m4 b/config/kernel-sysfs.m4
new file mode 100644
index 000000000000..668def5fe6bf
--- /dev/null
+++ b/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([for struct kobj_type.default_groups])
+	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
+])
diff --git a/config/kernel.m4 b/config/kernel.m4
index 0ef7c7d0caa7..bb3fe2ebdd14 100644
--- a/config/kernel.m4
+++ b/config/kernel.m4
@@ -132,6 +132,7 @@ 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_READPAGES
 	ZFS_AC_KERNEL_SRC_VFS_SET_PAGE_DIRTY_NOBUFFERS
@@ -244,6 +245,7 @@ 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_READPAGES
 	ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS
diff --git a/module/os/linux/zfs/zfs_sysfs.c b/module/os/linux/zfs/zfs_sysfs.c
index fb7c68987360..e73b34a2f37a 100644
--- a/module/os/linux/zfs/zfs_sysfs.c
+++ b/module/os/linux/zfs/zfs_sysfs.c
@@ -65,16 +65,15 @@
 /*
  * A zfs_mod_kobj_t represents a zfs kobject under '/sys/module/zfs'
  */
-struct zfs_mod_kobj;
 typedef struct zfs_mod_kobj zfs_mod_kobj_t;
-
 struct zfs_mod_kobj {
 	struct kobject		zko_kobj;
 	struct kobj_type	zko_kobj_type;
 	struct sysfs_ops	zko_sysfs_ops;
 	size_t			zko_attr_count;
 	struct attribute	*zko_attr_list;		/* allocated */
-	struct attribute	**zko_default_attrs;	/* allocated */
+	struct attribute_group	zko_default_group;	/* .attrs allocated */
+	const struct attribute_group	*zko_default_groups[2];
 	size_t			zko_child_count;
 	zfs_mod_kobj_t		*zko_children;		/* allocated */
 };
@@ -126,10 +125,10 @@ zfs_kobj_release(struct kobject *kobj)
 		zkobj->zko_attr_list = NULL;
 	}
 
-	if (zkobj->zko_default_attrs != NULL) {
-		kmem_free(zkobj->zko_default_attrs,
+	if (zkobj->zko_default_group.attrs != NULL) {
+		kmem_free(zkobj->zko_default_group.attrs,
 		    DEFAULT_ATTR_SIZE(zkobj->zko_attr_count));
-		zkobj->zko_default_attrs = NULL;
+		zkobj->zko_default_group.attrs = NULL;
 	}
 
 	if (zkobj->zko_child_count != 0) {
@@ -153,11 +152,12 @@ zfs_kobj_add_attr(zfs_mod_kobj_t *zkobj, int attr_num, const char *attr_name)
 {
 	VERIFY3U(attr_num, <, zkobj->zko_attr_count);
 	ASSERT(zkobj->zko_attr_list);
-	ASSERT(zkobj->zko_default_attrs);
+	ASSERT(zkobj->zko_default_group.attrs);
 
 	zkobj->zko_attr_list[attr_num].name = attr_name;
 	zkobj->zko_attr_list[attr_num].mode = 0444;
-	zkobj->zko_default_attrs[attr_num] = &zkobj->zko_attr_list[attr_num];
+	zkobj->zko_default_group.attrs[attr_num] =
+	    &zkobj->zko_attr_list[attr_num];
 	sysfs_attr_init(&zkobj->zko_attr_list[attr_num]);
 }
 
@@ -175,9 +175,9 @@ zfs_kobj_init(zfs_mod_kobj_t *zkobj, int attr_cnt, int child_cnt,
 			return (ENOMEM);
 	}
 	/* this will always have at least one slot for NULL termination */
-	zkobj->zko_default_attrs = kmem_zalloc(DEFAULT_ATTR_SIZE(attr_cnt),
-	    KM_SLEEP);
-	if (zkobj->zko_default_attrs == NULL) {
+	zkobj->zko_default_group.attrs =
+	    kmem_zalloc(DEFAULT_ATTR_SIZE(attr_cnt), KM_SLEEP);
+	if (zkobj->zko_default_group.attrs == NULL) {
 		if (zkobj->zko_attr_list != NULL) {
 			kmem_free(zkobj->zko_attr_list,
 			    ATTR_TABLE_SIZE(attr_cnt));
@@ -185,14 +185,19 @@ zfs_kobj_init(zfs_mod_kobj_t *zkobj, int attr_cnt, int child_cnt,
 		return (ENOMEM);
 	}
 	zkobj->zko_attr_count = attr_cnt;
-	zkobj->zko_kobj_type.default_attrs = zkobj->zko_default_attrs;
+	zkobj->zko_default_groups[0] = &zkobj->zko_default_group;
+#ifdef HAVE_SYSFS_DEFAULT_GROUPS
+	zkobj->zko_kobj_type.default_groups = zkobj->zko_default_groups;
+#else
+	zkobj->zko_kobj_type.default_attrs = zkobj->zko_default_group.attrs;
+#endif
 
 	if (child_cnt > 0) {
 		zkobj->zko_children = kmem_zalloc(CHILD_TABLE_SIZE(child_cnt),
 		    KM_SLEEP);
 		if (zkobj->zko_children == NULL) {
-			if (zkobj->zko_default_attrs != NULL) {
-				kmem_free(zkobj->zko_default_attrs,
+			if (zkobj->zko_default_group.attrs != NULL) {
+				kmem_free(zkobj->zko_default_group.attrs,
 				    DEFAULT_ATTR_SIZE(attr_cnt));
 			}
 			if (zkobj->zko_attr_list != NULL) {
@@ -214,9 +219,9 @@ zfs_kobj_init(zfs_mod_kobj_t *zkobj, int attr_cnt, int child_cnt,
 static int
 zfs_kobj_add(zfs_mod_kobj_t *zkobj, struct kobject *parent, const char *name)
 {
-	/* zko_default_attrs must be NULL terminated */
-	ASSERT(zkobj->zko_default_attrs != NULL);
-	ASSERT(zkobj->zko_default_attrs[zkobj->zko_attr_count] == NULL);
+	/* zko_default_group.attrs must be NULL terminated */
+	ASSERT(zkobj->zko_default_group.attrs != NULL);
+	ASSERT(zkobj->zko_default_group.attrs[zkobj->zko_attr_count] == NULL);
 
 	kobject_init(&zkobj->zko_kobj, &zkobj->zko_kobj_type);
 	return (kobject_add(&zkobj->zko_kobj, parent, name));