]> www.infradead.org Git - users/hch/xfstests-dev.git/commitdiff
btrfs/219: fix _cleanup() to successful release the loop-device
authorAnand Jain <anand.jain@oracle.com>
Thu, 2 Nov 2023 11:28:20 +0000 (19:28 +0800)
committerZorro Lang <zlang@kernel.org>
Thu, 16 Nov 2023 02:42:39 +0000 (10:42 +0800)
When we fail with the message 'We were allowed to mount when we should
have failed,' it will fail to clean up the loop devices, making it
difficult to run further test cases or the same test case again.

So we need a 2nd loop device local variable to release it. Let's
reorganize the local variables to clean them up in the _cleanup() function.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Zorro Lang <zlang@kernel.org>
tests/btrfs/219

index b747ce34fcc46e3a9b3b49df2376700657c89a64..35824df2baaa884739110effc3814634ae4d8e1f 100755 (executable)
@@ -19,14 +19,19 @@ _cleanup()
 {
        cd /
        rm -f $tmp.*
-       if [ ! -z "$loop_mnt" ]; then
-               $UMOUNT_PROG $loop_mnt
-               rm -rf $loop_mnt
-       fi
-       [ ! -z "$loop_mnt1" ] && rm -rf $loop_mnt1
-       [ ! -z "$fs_img1" ] && rm -rf $fs_img1
-       [ ! -z "$fs_img2" ] && rm -rf $fs_img2
-       [ ! -z "$loop_dev" ] && _destroy_loop_device $loop_dev
+
+       # The variables are set before the test case can fail.
+       $UMOUNT_PROG ${loop_mnt1} &> /dev/null
+       $UMOUNT_PROG ${loop_mnt2} &> /dev/null
+       rm -rf $loop_mnt1
+       rm -rf $loop_mnt2
+
+       [ ! -z $loop_dev1 ] && _destroy_loop_device $loop_dev1
+       [ ! -z $loop_dev1 ] && _destroy_loop_device $loop_dev2
+
+       rm -f $fs_img1
+       rm -f $fs_img2
+
        _btrfs_rescan_devices
 }
 
@@ -36,56 +41,60 @@ _cleanup()
 # real QA test starts here
 
 _supported_fs btrfs
+
+loop_mnt1=$TEST_DIR/$seq/mnt1
+loop_mnt2=$TEST_DIR/$seq/mnt2
+fs_img1=$TEST_DIR/$seq/img1
+fs_img2=$TEST_DIR/$seq/img2
+loop_dev1=""
+loop_dev2=""
+
 _require_test
 _require_loop
 _require_btrfs_forget_or_module_loadable
 _fixed_by_kernel_commit 5f58d783fd78 \
        "btrfs: free device in btrfs_close_devices for a single device filesystem"
 
-loop_mnt=$TEST_DIR/$seq.mnt
-loop_mnt1=$TEST_DIR/$seq.mnt1
-fs_img1=$TEST_DIR/$seq.img1
-fs_img2=$TEST_DIR/$seq.img2
-
-mkdir $loop_mnt
-mkdir $loop_mnt1
+mkdir -p $loop_mnt1
+mkdir -p $loop_mnt2
 
 $XFS_IO_PROG -f -c "truncate 256m" $fs_img1 >>$seqres.full 2>&1
 
 _mkfs_dev $fs_img1 >>$seqres.full 2>&1
 cp $fs_img1 $fs_img2
 
+loop_dev1=`_create_loop_device $fs_img1`
+loop_dev2=`_create_loop_device $fs_img2`
+
 # Normal single device case, should pass just fine
-_mount -o loop $fs_img1 $loop_mnt > /dev/null  2>&1 || \
+_mount $loop_dev1 $loop_mnt1 > /dev/null  2>&1 || \
        _fail "Couldn't do initial mount"
-$UMOUNT_PROG $loop_mnt
+$UMOUNT_PROG $loop_mnt1
 
 _btrfs_forget_or_module_reload
 
 # Now mount the new version again to get the higher generation cached, umount
 # and try to mount the old version.  Mount the new version again just for good
 # measure.
-loop_dev=`_create_loop_device $fs_img1`
-
-_mount $loop_dev $loop_mnt > /dev/null 2>&1 || \
+_mount $loop_dev1 $loop_mnt1 > /dev/null 2>&1 || \
        _fail "Failed to mount the second time"
-$UMOUNT_PROG $loop_mnt
+$UMOUNT_PROG $loop_mnt1
 
-_mount -o loop $fs_img2 $loop_mnt > /dev/null 2>&1 || \
+_mount $loop_dev2 $loop_mnt2 > /dev/null 2>&1 || \
        _fail "We couldn't mount the old generation"
-$UMOUNT_PROG $loop_mnt
+$UMOUNT_PROG $loop_mnt2
 
-_mount $loop_dev $loop_mnt > /dev/null 2>&1 || \
+_mount $loop_dev1 $loop_mnt1 > /dev/null 2>&1 || \
        _fail "Failed to mount the second time"
-$UMOUNT_PROG $loop_mnt
+$UMOUNT_PROG $loop_mnt1
 
 # Now we definitely can't mount them at the same time, because we're still tied
 # to the limitation of one fs_devices per fsid.
 _btrfs_forget_or_module_reload
 
-_mount $loop_dev $loop_mnt > /dev/null 2>&1 || \
+_mount $loop_dev1 $loop_mnt1 > /dev/null 2>&1 || \
        _fail "Failed to mount the third time"
-_mount -o loop $fs_img2 $loop_mnt1 > /dev/null 2>&1 && \
+_mount $loop_dev2 $loop_mnt2 > /dev/null 2>&1 && \
        _fail "We were allowed to mount when we should have failed"
 
 _btrfs_rescan_devices