Fix cloning into mmaped and cached file.

If the destination file is mmaped and the mmaped region was already
read, so it is cached, we need to update mmaped pages after successful
clone using update_pages().

Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Pointed out by: Ka Ho Ng <khng@freebsd.org>
Signed-off-by: Pawel Jakub Dawidek <pawel@dawidek.net>
Closes #15772

module/zfs/zfs_vnops.c

@@ -1349,6 +1349,10 @@ zfs_clone_range(znode_t *inzp, uint64_t *inoffp, znode_t *outzp,
 			break;
 		}
 
+		if (zn_has_cached_data(outzp, outoff, outoff + size - 1)) {
+			update_pages(outzp, outoff, size, outos);
+		}
+
 		zfs_clear_setid_bits_if_necessary(outzfsvfs, outzp, cr,
 		    &clear_setid_bits_txg, tx);
 

tests/runfiles/common.run

@@ -72,7 +72,9 @@ tags = ['functional', 'bclone']
 timeout = 7200
 
 [tests/functional/block_cloning]
-tests = ['block_cloning_copyfilerange', 'block_cloning_copyfilerange_partial',
+tests = ['block_cloning_clone_mmap_cached',
+    'block_cloning_copyfilerange',
+    'block_cloning_copyfilerange_partial',
     'block_cloning_copyfilerange_fallback',
     'block_cloning_disabled_copyfilerange',
     'block_cloning_copyfilerange_cross_dataset',

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

@@ -287,6 +287,7 @@ elif sys.platform.startswith('linux'):
         'bclone/bclone_samefs_data': ['SKIP', cfr_reason],
         'bclone/bclone_samefs_embedded': ['SKIP', cfr_reason],
         'bclone/bclone_samefs_hole': ['SKIP', cfr_reason],
+        'block_cloning/block_cloning_clone_mmap_cached': ['SKIP', cfr_reason],
         'block_cloning/block_cloning_clone_mmap_write':
             ['SKIP', cfr_reason],
         'block_cloning/block_cloning_copyfilerange':

tests/zfs-tests/cmd/.gitignore

@@ -2,6 +2,7 @@
 /btree_test
 /chg_usr_exec
 /clonefile
+/clone_mmap_cached
 /clone_mmap_write
 /devname2devid
 /dir_rd_update

tests/zfs-tests/cmd/Makefile.am

@@ -3,6 +3,7 @@ scripts_zfs_tests_bindir = $(datadir)/$(PACKAGE)/zfs-tests/bin
 
 scripts_zfs_tests_bin_PROGRAMS  = %D%/chg_usr_exec
 scripts_zfs_tests_bin_PROGRAMS += %D%/clonefile
+scripts_zfs_tests_bin_PROGRAMS += %D%/clone_mmap_cached
 scripts_zfs_tests_bin_PROGRAMS += %D%/clone_mmap_write
 scripts_zfs_tests_bin_PROGRAMS += %D%/cp_files
 scripts_zfs_tests_bin_PROGRAMS += %D%/ctime

tests/zfs-tests/cmd/clone_mmap_cached.c

@@ -0,0 +1,146 @@
+/*
+ * CDDL HEADER START
+ *
+ * The contents of this file are subject to the terms of the
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
+ *
+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+ * or https://opensource.org/licenses/CDDL-1.0.
+ * See the License for the specific language governing permissions
+ * and limitations under the License.
+ *
+ * When distributing Covered Code, include this CDDL HEADER in each
+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+ * If applicable, add the following below this CDDL HEADER, with the
+ * fields enclosed by brackets "[]" replaced with your own identifying
+ * information: Portions Copyright [yyyy] [name of copyright owner]
+ *
+ * CDDL HEADER END
+ */
+
+/*
+ * Copyright (c) 2024 by Pawel Jakub Dawidek
+ */
+
+#include <sys/mman.h>
+#include <sys/stat.h>
+
+#include <assert.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#ifdef __FreeBSD__
+#define	loff_t	off_t
+#endif
+
+ssize_t
+copy_file_range(int, loff_t *, int, loff_t *, size_t, unsigned int)
+    __attribute__((weak));
+
+static void *
+mmap_file(int fd, size_t size)
+{
+	void *p;
+
+	p = mmap(NULL, size, PROT_READ, MAP_SHARED, fd, 0);
+	if (p == MAP_FAILED) {
+		(void) fprintf(stderr, "mmap failed: %s\n", strerror(errno));
+		exit(2);
+	}
+
+	return (p);
+}
+
+static void
+usage(const char *progname)
+{
+
+	/*
+	 * -i cache input before copy_file_range(2).
+	 * -o cache input before copy_file_range(2).
+	 */
+	(void) fprintf(stderr, "usage: %s [-io] <input> <output>\n", progname);
+	exit(3);
+}
+
+int
+main(int argc, char *argv[])
+{
+	int dfd, sfd;
+	size_t dsize, ssize;
+	void *dmem, *smem, *ptr;
+	off_t doff, soff;
+	struct stat sb;
+	bool cache_input, cache_output;
+	const char *progname;
+	int c;
+
+	progname = argv[0];
+	cache_input = cache_output = false;
+
+	while ((c = getopt(argc, argv, "io")) != -1) {
+		switch (c) {
+		case 'i':
+			cache_input = true;
+			break;
+		case 'o':
+			cache_output = true;
+			break;
+		default:
+			usage(progname);
+		}
+	}
+	argc -= optind;
+	argv += optind;
+
+	if (argc != 2) {
+		usage(progname);
+	}
+
+	sfd = open(argv[0], O_RDONLY);
+	if (fstat(sfd, &sb) == -1) {
+		(void) fprintf(stderr, "fstat failed: %s\n", strerror(errno));
+		exit(2);
+	}
+	ssize = sb.st_size;
+	smem = mmap_file(sfd, ssize);
+
+	dfd = open(argv[1], O_RDWR);
+	if (fstat(dfd, &sb) == -1) {
+		(void) fprintf(stderr, "fstat failed: %s\n", strerror(errno));
+		exit(2);
+	}
+	dsize = sb.st_size;
+	dmem = mmap_file(dfd, dsize);
+
+	/*
+	 * Hopefully it won't be compiled out.
+	 */
+	if (cache_input) {
+		ptr = malloc(ssize);
+		assert(ptr != NULL);
+		memcpy(ptr, smem, ssize);
+		free(ptr);
+	}
+	if (cache_output) {
+		ptr = malloc(ssize);
+		assert(ptr != NULL);
+		memcpy(ptr, dmem, dsize);
+		free(ptr);
+	}
+
+	soff = doff = 0;
+	if (copy_file_range(sfd, &soff, dfd, &doff, ssize, 0) < 0) {
+		(void) fprintf(stderr, "copy_file_range failed: %s\n",
+		    strerror(errno));
+		exit(2);
+	}
+
+	exit(memcmp(smem, dmem, ssize) == 0 ? 0 : 1);
+}

tests/zfs-tests/include/commands.cfg

@@ -185,6 +185,7 @@ export ZFSTEST_FILES='badsend
     btree_test
     chg_usr_exec
     clonefile
+    clone_mmap_cached
     clone_mmap_write
     devname2devid
     dir_rd_update

tests/zfs-tests/tests/Makefile.am

@@ -461,6 +461,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/bclone/setup.ksh \
 	functional/block_cloning/cleanup.ksh \
 	functional/block_cloning/setup.ksh \
+	functional/block_cloning/block_cloning_clone_mmap_cached.ksh \
 	functional/block_cloning/block_cloning_clone_mmap_write.ksh \
 	functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh \
 	functional/block_cloning/block_cloning_copyfilerange_fallback.ksh \

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

@@ -0,0 +1,86 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or https://opensource.org/licenses/CDDL-1.0.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/block_cloning/block_cloning.kshlib
+
+#
+# DESCRIPTION:
+#	When the destination file is mmaped and is already cached we need to
+#	update mmaped pages after successful clone.
+#
+# STRATEGY:
+#	1. Create a pool.
+#	2. Create a two test files with random content.
+#	3. mmap the files, read them and clone from one to the other using
+#	   clone_mmap_cached.
+#	4. clone_mmap_cached also verifies if the content of the destination
+#	   file was updated while reading it from mmaped memory.
+#
+
+verify_runnable "global"
+
+if is_linux && [[ $(linux_version) -lt $(linux_version "4.5") ]]; then
+  log_unsupported "copy_file_range not available before Linux 4.5"
+fi
+
+VDIR=$TEST_BASE_DIR/disk-bclone
+VDEV="$VDIR/a"
+
+function cleanup
+{
+	datasetexists $TESTPOOL && destroy_pool $TESTPOOL
+	rm -rf $VDIR
+}
+
+log_onexit cleanup
+
+log_assert "Test for clone into mmaped and cached file"
+
+log_must rm -rf $VDIR
+log_must mkdir -p $VDIR
+log_must truncate -s 1G $VDEV
+
+log_must zpool create -o feature@block_cloning=enabled $TESTPOOL $VDEV
+log_must zfs create $TESTPOOL/$TESTFS
+
+for opts in "--" "-i" "-o" "-io"
+do
+	log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/src bs=1M count=1
+	log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/dst bs=1M count=1
+
+	# Clear cache.
+	log_must zpool export $TESTPOOL
+	log_must zpool import -d $VDIR $TESTPOOL
+
+	log_must clone_mmap_cached $opts /$TESTPOOL/$TESTFS/src /$TESTPOOL/$TESTFS/dst
+
+	sync_pool $TESTPOOL
+	log_must sync
+
+	log_must have_same_content /$TESTPOOL/$TESTFS/src /$TESTPOOL/$TESTFS/dst
+	blocks=$(get_same_blocks $TESTPOOL/$TESTFS src $TESTPOOL/$TESTFS dst)
+	# FreeBSD's seq(1) leaves a trailing space, remove it with sed(1).
+	log_must [ "$blocks" = "$(seq -s " " 0 7 | sed 's/ $//')" ]
+done
+
+log_pass "Clone properly updates mmapped and cached pages"

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

@@ -30,6 +30,9 @@
 if ! command -v clonefile > /dev/null ; then
   log_unsupported "clonefile program required to test block cloning"
 fi
+if ! command -v clone_mmap_cached > /dev/null ; then
+  log_unsupported "clone_mmap_cached program required to test block cloning"
+fi
 
 verify_runnable "global"