add a new test for racing AIO COW writes

This can be used to trigger an assert in the current XFS code because it can't handle the case where there are COW extents on a file, but none at or below the range converted by the AIO completion handler.

Note that it doesn't trigger the assert 100% but fairly reliably.

Signed-off-by Christoph Hellwig <[email protected]>

src/aio-dio-regress/aio-dio-cow-race.c new file +119 -0

new file mode 100755
index 0000000..c7f8e89
@@ -0,0 +1,119 @@
+/*
+ * Copyright (c) Christoph Hellwig.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <errno.h>
+#include <libgen.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <ctype.h>
+#include <sys/ioctl.h>
+#include <linux/fs.h>
+
+#include <libaio.h>
+
+#ifndef FICLONE
+#define FICLONE		_IOW(0x94, 9, int)
+#endif
+
+#define IO_PATTERN	0xab
+
+int main(int argc, char *argv[])
+{
+	struct io_context *ctx = NULL;
+	struct io_event evs[4];
+	struct iocb iocb1, iocb2, iocb3, iocb4;
+	struct iocb *iocbs[] = { &iocb1, &iocb2, &iocb3, &iocb4 };
+	void *buf;
+	int fd, clone, err = 0;
+	unsigned long buf_size = getpagesize() * 2;
+	char *filename, *clonename;
+
+	if (argc != 3) {
+		fprintf(stderr, "usage: %s filename clonename\n", argv[0]);
+		exit(1);
+	}
+
+	err = posix_memalign(&buf, getpagesize(), buf_size);
+	if (err) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "posix_memalign");
+		return 1;
+	}
+	memset(buf, IO_PATTERN, buf_size);
+
+	filename = argv[1];
+	fd = open(filename, O_DIRECT | O_CREAT | O_TRUNC | O_RDWR, 0600);
+	if (fd == -1) {
+		perror("open");
+		return 1;
+	}
+
+	if (write(fd, buf, buf_size) != buf_size) {
+		perror("write");
+		return 1;
+	}
+
+	clonename = argv[2];
+	clone = open(clonename, O_DIRECT | O_CREAT | O_TRUNC | O_RDWR, 0600);
+	if (clone == -1) {
+		perror("open clone");
+		return 1;
+	}
+
+	if (ioctl(clone, FICLONE, fd)) {
+		perror("FICLONE");
+		return 1;
+	}
+
+	err = io_setup(4, &ctx);
+	if (err) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err),
+			"io_setup");
+		return 1;
+	}
+
+	/*
+	 * Test overlapping aio writes, where an earlier one clears the whole
+	 * range of a later aio, thus leaving nothing to do for the I/O
+	 * completion to do.  To make things harder also make sure there is
+	 * other outstanding COW I/O.
+	 */
+	io_prep_pwrite(&iocb1, fd, buf, buf_size, 0);
+	io_prep_pwrite(&iocb2, fd, buf, buf_size / 2, 0);
+	io_prep_pwrite(&iocb3, fd, buf, buf_size / 2, buf_size);
+	io_prep_pwrite(&iocb4, fd, buf, buf_size, 0);
+
+	err = io_submit(ctx, 4, iocbs);
+	if (err != 4) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "io_submit");
+		return 1;
+	}
+
+	err = io_getevents(ctx, 4, 4, evs, NULL);
+	if (err != 4) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "io_getevents");
+		return 1;
+	}
+
+	return 0;
+}

tests/generic/463 new file +60 -0

new file mode 100755
index 0000000..b3afbb0
@@ -0,0 +1,60 @@
+#! /bin/bash
+# FS QA Test 463
+#
+# Test racy COW AIO write completions.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2017 Christoph Hellwig.  All Rights Reserved.
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License as
+# published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it would be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write the Free Software Foundation,
+# Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+#-----------------------------------------------------------------------
+#
+
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+	rm -f $TEST_DIR/file $TEST_DIR/clone
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/reflink
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+_supported_fs generic
+_supported_os Linux
+
+_require_test
+_require_test_reflink
+_require_aio
+_require_odirect
+
+$here/src/aio-dio-regress/aio-dio-cow-race $TEST_DIR/file $TEST_DIR/clone
+
+echo "Silence is golden"
+status=0
+exit

tests/generic/group +0 -0

index f2a6cda..014d6ea 100644

--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

This can be used to trigger an assert in the current XFS code because it can't handle the case where there are COW extents on a file, but none at or below the range converted by the AIO completion handler.

Signed-off-by Christoph Hellwig <[email protected]>

src/aio-dio-regress/aio-dio-cow-race.c new file +119 -0

new file mode 100755
index 0000000..c7f8e89
@@ -0,0 +1,119 @@
+/*
+ * Copyright (c) Christoph Hellwig.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <errno.h>
+#include <libgen.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <ctype.h>
+#include <sys/ioctl.h>
+#include <linux/fs.h>
+
+#include <libaio.h>
+
+#ifndef FICLONE
+#define FICLONE		_IOW(0x94, 9, int)
+#endif
+
+#define IO_PATTERN	0xab
+
+int main(int argc, char *argv[])
+{
+	struct io_context *ctx = NULL;
+	struct io_event evs[4];
+	struct iocb iocb1, iocb2, iocb3, iocb4;
+	struct iocb *iocbs[] = { &iocb1, &iocb2, &iocb3, &iocb4 };
+	void *buf;
+	int fd, clone, err = 0;
+	unsigned long buf_size = getpagesize() * 2;
+	char *filename, *clonename;
+
+	if (argc != 3) {
+		fprintf(stderr, "usage: %s filename clonename\n", argv[0]);
+		exit(1);
+	}
+
+	err = posix_memalign(&buf, getpagesize(), buf_size);
+	if (err) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "posix_memalign");
+		return 1;
+	}
+	memset(buf, IO_PATTERN, buf_size);
+
+	filename = argv[1];
+	fd = open(filename, O_DIRECT | O_CREAT | O_TRUNC | O_RDWR, 0600);
+	if (fd == -1) {
+		perror("open");
+		return 1;
+	}
+
+	if (write(fd, buf, buf_size) != buf_size) {
+		perror("write");
+		return 1;
+	}
+
+	clonename = argv[2];
+	clone = open(clonename, O_DIRECT | O_CREAT | O_TRUNC | O_RDWR, 0600);
+	if (clone == -1) {
+		perror("open clone");
+		return 1;
+	}
+
+	if (ioctl(clone, FICLONE, fd)) {
+		perror("FICLONE");
+		return 1;
+	}
+
+	err = io_setup(4, &ctx);
+	if (err) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err),
+			"io_setup");
+		return 1;
+	}
+
+	/*
+	 * Test overlapping aio writes, where an earlier one clears the whole
+	 * range of a later aio, thus leaving nothing to do for the I/O
+	 * completion to do.  To make things harder also make sure there is
+	 * other outstanding COW I/O.
+	 */
+	io_prep_pwrite(&iocb1, fd, buf, buf_size, 0);
+	io_prep_pwrite(&iocb2, fd, buf, buf_size / 2, 0);
+	io_prep_pwrite(&iocb3, fd, buf, buf_size / 2, buf_size);
+	io_prep_pwrite(&iocb4, fd, buf, buf_size, 0);
+
+	err = io_submit(ctx, 4, iocbs);
+	if (err != 4) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "io_submit");
+		return 1;
+	}
+
+	err = io_getevents(ctx, 4, 4, evs, NULL);
+	if (err != 4) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "io_getevents");
+		return 1;
+	}
+
+	return 0;
+}

tests/generic/463 new file +60 -0

new file mode 100755
index 0000000..b3afbb0
@@ -0,0 +1,60 @@
+#! /bin/bash
+# FS QA Test 463
+#
+# Test racy COW AIO write completions.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2017 Christoph Hellwig.  All Rights Reserved.
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License as
+# published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it would be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write the Free Software Foundation,
+# Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+#-----------------------------------------------------------------------
+#
+
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+	rm -f $TEST_DIR/file $TEST_DIR/clone
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/reflink
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+_supported_fs generic
+_supported_os Linux
+
+_require_test
+_require_test_reflink
+_require_aio
+_require_odirect
+
+$here/src/aio-dio-regress/aio-dio-cow-race $TEST_DIR/file $TEST_DIR/clone
+
+echo "Silence is golden"
+status=0
+exit

tests/generic/group +1 -1

index f2a6cda..014d6ea 100644
@@ -465,3 +465,4 @@
 460 auto quick rw
 461 auto shutdown stress
 462 auto quick dax
+463 auto quick clone dangerous
--

To unsubscribe from this list: send the line "unsubscribe linux-xfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

This can be used to trigger an assert in the current XFS code because it can't handle the case where there are COW extents on a file, but none at or below the range converted by the AIO completion handler.

Signed-off-by Christoph Hellwig <[email protected]>

src/aio-dio-regress/aio-dio-cow-race.c new file +79 -0

@@ -0,0 +1,119 @@
+/*
+ * Copyright (c) Christoph Hellwig.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <errno.h>
+#include <libgen.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <ctype.h>
+#include <sys/ioctl.h>
+#include <linux/fs.h>
+
+#include <libaio.h>
+
+#ifndef FICLONE
+#define FICLONE		_IOW(0x94, 9, int)
+#endif
+
+#define IO_PATTERN	0xab
+
+int main(int argc, char *argv[])
+{
+	struct io_context *ctx = NULL;
+	struct io_event evs[4];
+	struct iocb iocb1, iocb2, iocb3, iocb4;
+	struct iocb *iocbs[] = { &iocb1, &iocb2, &iocb3, &iocb4 };
+	void *buf;
+	int fd, clone, err = 0;
+	unsigned long buf_size = getpagesize() * 2;
+	char *filename, *clonename;
+
+	if (argc != 3) {
+		fprintf(stderr, "usage: %s filename clonename\n", argv[0]);
+		exit(1);
+	}
+
+	err = posix_memalign(&buf, getpagesize(), buf_size);
+	if (err) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "posix_memalign");
+		return 1;
+	}
+	memset(buf, IO_PATTERN, buf_size);
+
+	filename = argv[1];
+	fd = open(filename, O_DIRECT | O_CREAT | O_TRUNC | O_RDWR, 0600);
+	if (fd == -1) {
+		perror("open");
+		return 1;
+	}
+
+	if (write(fd, buf, buf_size) != buf_size) {
+		perror("write");
+		return 1;
+	}
+
+	clonename = argv[2];
+	clone = open(clonename, O_DIRECT | O_CREAT | O_TRUNC | O_RDWR, 0600);
+	if (clone == -1) {
+		perror("open clone");
+		return 1;
+	}
+

src/aio-dio-regress/aio-dio-cow-race.c new file +30 -0

+		return 1;
+	}
+
+	/*
+	 * Test overlapping aio writes, where an earlier one clears the whole
+	 * range of a later aio, thus leaving nothing to do for the I/O
+	 * completion to do.  To make things harder also make sure there is
+	 * other outstanding COW I/O.
+	 */
+	io_prep_pwrite(&iocb1, fd, buf, buf_size, 0);
+	io_prep_pwrite(&iocb2, fd, buf, buf_size / 2, 0);
+	io_prep_pwrite(&iocb3, fd, buf, buf_size / 2, buf_size);
+	io_prep_pwrite(&iocb4, fd, buf, buf_size, 0);
+
+	err = io_submit(ctx, 4, iocbs);
+	if (err != 4) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "io_submit");
+		return 1;
+	}
+
+	err = io_getevents(ctx, 4, 4, evs, NULL);
+	if (err != 4) {
+		fprintf(stderr, "error %s during %s\n",
+			strerror(err), "io_getevents");
+		return 1;
+	}
+
+	return 0;
+}

tests/generic/463 new file +44 -0

new file mode 100755
index 0000000..b3afbb0
@@ -0,0 +1,60 @@
+#! /bin/bash
+# FS QA Test 463
+#
+# Test racy COW AIO write completions.
+#
+#-----------------------------------------------------------------------
+# Copyright (c) 2017 Christoph Hellwig.  All Rights Reserved.
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License as
+# published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it would be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write the Free Software Foundation,
+# Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+#-----------------------------------------------------------------------
+#
+
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+	rm -f $TEST_DIR/file $TEST_DIR/clone
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+. ./common/reflink
+

tests/generic/463 new file +2 -0

+status=0
+exit

tests/generic/group +1 -1

index f2a6cda..014d6ea 100644
@@ -465,3 +465,4 @@
 460 auto quick rw
 461 auto shutdown stress
 462 auto quick dax
+463 auto quick clone dangerous
--

To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Quoted 4 lines · Christoph Hellwig

This can be used to trigger an assert in the current XFS code because it can't handle the case where there are COW extents on a file, but none at or below the range converted by the AIO completion handler.

Quoted 1 line

Signed-off-by Christoph Hellwig <[email protected]>

I fixed several minor issues on commit (see inline comments).

I haven't reproduced the assert failure yet on my test vm, but as the commit log in v1 patch, it's not 100% reproducible, maybe I'm just lucky (or not lucky :)