Add Linux posix_fadvise support

The purpose of this PR is to accepts fadvise ioctl from userland
to do read-ahead by demand.

It could dramatically improve sequential read performance especially
when primarycache is set to metadata or zfs_prefetch_disable is 1.

If the file is mmaped, generic_fadvise is also called for page cache
read-ahead besides dmu_prefetch.

Only POSIX_FADV_WILLNEED and POSIX_FADV_SEQUENTIAL are supported in
this PR currently.

Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Finix Yan <[email protected]>
Closes #13694

10 files changed, 245 insertions, 2 deletions

diff --git a/tests/zfs-tests/cmd/.gitignore b/tests/zfs-tests/cmd/.gitignore
index 20d138253..1fd54c1dd 100644
--- a/tests/zfs-tests/cmd/.gitignore
+++ b/tests/zfs-tests/cmd/.gitignore
@@ -4,6 +4,7 @@
 /devname2devid
 /dir_rd_update
 /draid
+/file_fadvise
 /file_append
 /file_check
 /file_trunc
diff --git a/tests/zfs-tests/cmd/Makefile.am b/tests/zfs-tests/cmd/Makefile.am
index 3c8faf5af..c19c870cf 100644
--- a/tests/zfs-tests/cmd/Makefile.am
+++ b/tests/zfs-tests/cmd/Makefile.am
@@ -128,4 +128,7 @@ scripts_zfs_tests_bin_PROGRAMS  += %D%/read_dos_attributes %D%/write_dos_attribu
 
 scripts_zfs_tests_bin_PROGRAMS += %D%/randfree_file
 %C%_randfree_file_SOURCES       = %D%/file/randfree_file.c
+
+scripts_zfs_tests_bin_PROGRAMS += %D%/file_fadvise
+%C%_file_fadvise_SOURCES  = %D%/file/file_fadvise.c
 endif
diff --git a/tests/zfs-tests/cmd/file/file_fadvise.c b/tests/zfs-tests/cmd/file/file_fadvise.c
new file mode 100644
index 000000000..e1afb6d0a
--- /dev/null
+++ b/tests/zfs-tests/cmd/file/file_fadvise.c
@@ -0,0 +1,97 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or https://opensource.org/licenses/CDDL-1.0.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
+ * Use is subject to license terms.
+ */
+
+/*
+ * Copyright (c) 2022 by Information2 Software, Inc. All rights reserved.
+ */
+
+#include "file_common.h"
+#include <sys/types.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <string.h>
+
+/*
+ * Call fadvise to prefetch data
+ */
+static const char *execname = "file_fadvise";
+
+static void
+usage(void)
+{
+	(void) fprintf(stderr,
+	    "usage: %s -f filename -a advise \n", execname);
+}
+
+int
+main(int argc, char *argv[])
+{
+	char *filename = NULL;
+	int advise = 0;
+	int fd, ch;
+	int	err = 0;
+
+	while ((ch = getopt(argc, argv, "a:f:")) != EOF) {
+		switch (ch) {
+		case 'a':
+			advise = atoll(optarg);
+			break;
+		case 'f':
+			filename = optarg;
+			break;
+		case '?':
+			(void) printf("unknown arg %c\n", optopt);
+			usage();
+			break;
+		}
+	}
+
+	if (!filename) {
+		(void) printf("Filename not specified (-f <file>)\n");
+		err++;
+	}
+
+	if (advise < POSIX_FADV_NORMAL || advise > POSIX_FADV_NOREUSE) {
+		(void) printf("advise is invalid\n");
+		err++;
+	}
+
+	if (err) {
+		usage(); /* no return */
+		return (1);
+	}
+
+	if ((fd = open(filename, O_RDWR, 0666)) < 0) {
+		perror("open");
+		return (1);
+	}
+
+	posix_fadvise(fd, 0, 0, advise);
+
+	close(fd);
+
+	return (0);
+}
diff --git a/tests/zfs-tests/include/commands.cfg b/tests/zfs-tests/include/commands.cfg
index 409856221..c05b91832 100644
--- a/tests/zfs-tests/include/commands.cfg
+++ b/tests/zfs-tests/include/commands.cfg
@@ -184,6 +184,7 @@ export ZFSTEST_FILES='badsend
     devname2devid
     dir_rd_update
     draid
+    file_fadvise
     file_append
     file_check
     file_trunc
diff --git a/tests/zfs-tests/tests/Makefile.am b/tests/zfs-tests/tests/Makefile.am
index 89b2ca866..d53316643 100644
--- a/tests/zfs-tests/tests/Makefile.am
+++ b/tests/zfs-tests/tests/Makefile.am
@@ -1370,6 +1370,9 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/exec/exec_001_pos.ksh \
 	functional/exec/exec_002_neg.ksh \
 	functional/exec/setup.ksh \
+	functional/fadvise/cleanup.ksh \
+	functional/fadvise/fadvise_sequential.ksh \
+	functional/fadvise/setup.ksh \
 	functional/fallocate/cleanup.ksh \
 	functional/fallocate/fallocate_prealloc.ksh \
 	functional/fallocate/fallocate_punch-hole.ksh \
diff --git a/tests/zfs-tests/tests/functional/checksum/filetest_002_pos.ksh b/tests/zfs-tests/tests/functional/checksum/filetest_002_pos.ksh
index a0be1c205..23e7aa577 100755
--- a/tests/zfs-tests/tests/functional/checksum/filetest_002_pos.ksh
+++ b/tests/zfs-tests/tests/functional/checksum/filetest_002_pos.ksh
@@ -76,7 +76,7 @@ while [[ $j -lt ${#CHECKSUM_TYPES[*]} ]]; do
 	log_must zpool export $TESTPOOL
 	log_must zpool import $TESTPOOL
 
-	log_mustnot eval "cat $TESTDIR/test_$type >/dev/null"
+	log_mustnot eval "dd if=$TESTDIR/test_$type of=/dev/null bs=$WRITESZ count=$NWRITES"
 
 	cksum=$(zpool status -P -v $TESTPOOL | grep "$firstvdev" | \
 	    awk '{print $5}')
diff --git a/tests/zfs-tests/tests/functional/fadvise/cleanup.ksh b/tests/zfs-tests/tests/functional/fadvise/cleanup.ksh
new file mode 100755
index 000000000..8b5b43a74
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/fadvise/cleanup.ksh
@@ -0,0 +1,28 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or https://opensource.org/licenses/CDDL-1.0.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+
+#
+# Portions Copyright (c) 2022 Information2 Software, Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+default_cleanup
diff --git a/tests/zfs-tests/tests/functional/fadvise/fadvise_sequential.ksh b/tests/zfs-tests/tests/functional/fadvise/fadvise_sequential.ksh
new file mode 100755
index 000000000..7b7d1d379
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/fadvise/fadvise_sequential.ksh
@@ -0,0 +1,80 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or https://opensource.org/licenses/CDDL-1.0.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+#
+# Portions Copyright (c) 2022 Information2 Software, Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/include/math.shlib
+
+#
+# DESCRIPTION:
+# Test posix_fadvise.
+#
+# STRATEGY:
+# 1. Set primarycache to metadata in order to disable prefetch
+# 2. Write some data to file 
+# 3. get data_size field from arcstat
+# 4. call file_fadvise with POSIX_FADV_SEQUENTIAL
+# 5. get data_size field from arcstat again
+# 6. latter data_size should be bigger than former one
+#
+
+# NOTE: if HAVE_FILE_FADVISE is not defined former data_size
+# should less or eaqul to latter one
+
+verify_runnable "global"
+
+FILE=$TESTDIR/$TESTFILE0
+BLKSZ=$(get_prop recordsize $TESTPOOL)
+
+function cleanup
+{
+	log_must zfs set primarycache=all $TESTPOOL
+	[[ -e $TESTDIR ]] && log_must rm -Rf $TESTDIR/*
+}
+
+getstat() {
+	awk -v c="$1" '$1 == c {print $3; exit}' /proc/spl/kstat/zfs/arcstats
+}
+
+log_assert "Ensure fadvise prefetch data"
+
+log_onexit cleanup
+
+log_must zfs set primarycache=metadata $TESTPOOL
+
+log_must file_write -o create -f $FILE -b $BLKSZ -c 1000
+sync_pool $TESTPOOL
+
+data_size1=$(getstat data_size)
+
+log_must file_fadvise -f $FILE -a 2
+sleep 10
+
+data_size2=$(getstat data_size)
+log_note "original data_size is $data_size1, final data_size is $data_size2"
+
+log_must [ $data_size1 -le $data_size2 ]
+
+log_pass "Ensure data could be prefetched"
diff --git a/tests/zfs-tests/tests/functional/fadvise/setup.ksh b/tests/zfs-tests/tests/functional/fadvise/setup.ksh
new file mode 100755
index 000000000..8ddd73307
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/fadvise/setup.ksh
@@ -0,0 +1,30 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or https://opensource.org/licenses/CDDL-1.0.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+
+#
+# Portions Copyright (c) 2022 Information2 Software, Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+DISK=${DISKS%% *}
+default_setup_noexit $DISK
+log_pass
diff --git a/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh b/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh
index e9517bad7..bd32be9a4 100755
--- a/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh
+++ b/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh
@@ -73,7 +73,7 @@ for type in "mirror" "raidz" "raidz2"; do
 
 	# 4. Inject CHECKSUM ERRORS on read with a zinject error handler
 	log_must zinject -d $FAULT_FILE -e corrupt -f 50 -T read $TESTPOOL
-	log_must cp $TESTFILE /dev/null
+	log_must dd if=$TESTFILE of=/dev/null bs=1M count=64
 
 	# 5. Verify the ZED kicks in a hot spare and expected pool/device status
 	log_note "Wait for ZED to auto-spare"