Re: [PATCH v5 1/3] generic: mmap write readonly DAX file

[Eryu Guan <eguan@redhat.com>]

On Mon, Sep 25, 2017 at 04:40:45PM +0800, Xiong Zhou wrote:
> Regression case that one can write to read-only
> file in a DAX mountpoint.
> 
> Signed-off-by: Xiong Zhou <xzhou@redhat.com>
> ---
>  .gitignore            |  1 +
>  src/Makefile          |  2 +-
>  src/t_mmap_write_ro.c | 77 ++++++++++++++++++++++++++++++++++++++++++
>  tests/generic/461     | 93 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/generic/461.out |  2 ++
>  tests/generic/group   |  1 +
>  6 files changed, 175 insertions(+), 1 deletion(-)
>  create mode 100644 src/t_mmap_write_ro.c
>  create mode 100755 tests/generic/461
>  create mode 100644 tests/generic/461.out
> 
> diff --git a/.gitignore b/.gitignore
> index 63928b4..864d899 100644
> --- a/.gitignore
> +++ b/.gitignore
> @@ -155,6 +155,7 @@
>  /src/t_mmap_cow_race
>  /src/t_mmap_fallocate
>  /src/log-writes/replay-log
> +/src/t_mmap_write_ro

Better to make these entries sorted, perhaps in a separate patch.

>  
>  # dmapi/ binaries
>  /dmapi/src/common/cmd/read_invis
> diff --git a/src/Makefile b/src/Makefile
> index 7d1306b..3eb25b1 100644
> --- a/src/Makefile
> +++ b/src/Makefile
> @@ -13,7 +13,7 @@ TARGETS = dirstress fill fill2 getpagesize holes lstat64 \
>  	multi_open_unlink dmiperf unwritten_sync genhashnames t_holes \
>  	t_mmap_writev t_truncate_cmtime dirhash_collide t_rename_overwrite \
>  	holetest t_truncate_self t_mmap_dio af_unix t_mmap_stale_pmd \
> -	t_mmap_cow_race t_mmap_fallocate fsync-err
> +	t_mmap_cow_race t_mmap_fallocate fsync-err t_mmap_write_ro
>  
>  LINUX_TARGETS = xfsctl bstat t_mtab getdevicesize preallo_rw_pattern_reader \
>  	preallo_rw_pattern_writer ftrunc trunc fs_perms testx looptest \
> diff --git a/src/t_mmap_write_ro.c b/src/t_mmap_write_ro.c
> new file mode 100644
> index 0000000..55cb3d1
> --- /dev/null
> +++ b/src/t_mmap_write_ro.c
> @@ -0,0 +1,77 @@
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <fcntl.h>
> +#include <unistd.h>
> +#include <libgen.h>
> +#include <errno.h>
> +#include <sys/mman.h>
> +
> +void
> +err_exit(char *op)
> +{
> +	fprintf(stderr, "%s: %s\n", op, strerror(errno));
> +	exit(1);
> +}
> +
> +int
> +main(int argc, char **argv)
> +{
> +	int fd, pfd, ret;
> +	char *buf;
> +	/* gcc -O2 will optimize foo's storage, preventing

Nitpick, let "/*" stays in the line alone for multi-line comments.
	/*
	 * gcc -O2...
	 * ...
	 */

> +	 * reproduce this issue.
> +	 * foo is never actually used after fault in value stored.
> +	 */
> +	volatile char foo __attribute__((__unused__));
> +	int pagesize = getpagesize();
> +
> +	if (argc < 3) {
> +		printf("Usage: %s <file> <pmem file>\n", basename(argv[0]));
> +		exit(0);
> +	}
> +
> +	fd = open(argv[1], O_RDONLY|O_DIRECT);

Is O_DIRECT really required? If so, some comments would be good.

> +	if (fd < 0)
> +		err_exit("open");
> +
> +	pfd = open(argv[2], O_RDONLY);
> +	if (pfd < 0)
> +		err_exit("pmem open");
> +
> +	buf = mmap(NULL, pagesize, PROT_READ, MAP_SHARED, pfd, 0);
> +	if (buf == MAP_FAILED)
> +		err_exit("mmap");
> +
> +	/*
> +	 * Read from the DAX mmap to populate the first page in the
> +	 * address_space with a read-only mapping.
> +	 */
> +	foo = *buf;
> +
> +	/*
> +	 * Now write to the DAX mmap.  This *should* fail, but if the bug is
> +	 * present in __get_user_pages_fast(), it will succeed.
> +	 */
> +	ret = read(fd, buf, pagesize);
> +	if (ret != pagesize)
> +		err_exit("read");
> +
> +	ret = msync(buf, pagesize, MS_SYNC);
> +	if (ret != 0)
> +		err_exit("msync");
> +
> +	ret = munmap(buf, pagesize);
> +	if (ret < 0)
> +		err_exit("munmap");
> +
> +	ret = close(fd);
> +	if (ret < 0)
> +		err_exit("clsoe fd");
> +
> +	ret = close(pfd);
> +	if (ret < 0)
> +		err_exit("close pfd");
> +
> +	exit(0);
> +}
> diff --git a/tests/generic/461 b/tests/generic/461
> new file mode 100755
> index 0000000..a6df2b6
> --- /dev/null
> +++ b/tests/generic/461
> @@ -0,0 +1,93 @@
> +#! /bin/bash
> +# FS QA Test 461
> +#
> +# This is a regression test for kernel commit
> +#  ef947b2 x86, mm: fix gup_pte_range() vs DAX mappings
> +# created by Jeffrey Moyer <jmoyer@redhat.com>
> +#
> +# This is reproducible only when testing on pmem device
> +# which is configured in "memory mode", not in "raw mode".
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2017 Red Hat.  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 -rf $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_supported_os Linux
> +_require_test
> +_require_scratch_dax
> +_require_test_program "t_mmap_write_ro"
> +_require_user

Hmm, not sure why we need the test run by unprivileged user just from
the test.

> +
> +# real QA test starts here
> +
> +_scratch_mkfs >>$seqres.full 2>&1
> +_scratch_mount "-o dax"
> +
> +# remount TEST_DEV wo/ dax
> +export MOUNT_OPTIONS=""

This is not needed.

> +export TEST_FS_MOUNT_OPTS=""
> +_test_cycle_mount
> +
> +# prepare a 4k read-only DAX file, save its md5sum
> +$XFS_IO_PROG -f -c "pwrite -S 0xFFFF 0 4096" \

"pwrite -S 0xFF" should be sufficient.

> +  $SCRATCH_MNT/readonlyfile >> $seqres.full 2>&1
> +chmod 0644  $SCRATCH_MNT/readonlyfile
> +md5_1="$(_md5_checksum $SCRATCH_MNT/readonlyfile)"
> +
> +# prepare another larger non-DAX file
> +$XFS_IO_PROG -f -c "pwrite -S 0x0000 0 40960" \
> +  $TEST_DIR/${seq}.largefile >> $seqres.full 2>&1
> +# allow qa_user access
> +chown $qa_user $TEST_DIR/${seq}.largefile
> +
> +# run test programme, read another larger file writing into
> +# the read-only file with mmap, which should fail.
> +_user_do "src/t_mmap_write_ro $TEST_DIR/${seq}.largefile \
> +  $SCRATCH_MNT/readonlyfile"
> +
> +# read-only file should not get updated, md5sum again.
> +md5_2="$(_md5_checksum $SCRATCH_MNT/readonlyfile)"
> +
> +[ "$md5_1" != "$md5_2" ] && echo "read only file changed"
> +
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/generic/461.out b/tests/generic/461.out
> new file mode 100644
> index 0000000..8b1e29c
> --- /dev/null
> +++ b/tests/generic/461.out
> @@ -0,0 +1,2 @@
> +QA output created by 461
> +read: Bad address
> diff --git a/tests/generic/group b/tests/generic/group
> index f922b49..1575a2d 100644
> --- a/tests/generic/group
> +++ b/tests/generic/group
> @@ -463,3 +463,4 @@
>  458 auto quick clone
>  459 auto dangerous
>  460 auto quick rw
> +461 auto quick

Time for a new 'dax' group? Perhaps add 'dax' group to other existing
tests too in a separate patch.

Thanks,
Eryu