diff mbox series

ext4/309: add test for ext4_dir_entry2 wipe

Message ID 20210504193425.1642352-1-leah.rumancik@gmail.com (mailing list archive)
State New, archived
Headers show
Series ext4/309: add test for ext4_dir_entry2 wipe | expand

Commit Message

Leah Rumancik May 4, 2021, 7:34 p.m. UTC
From: Leah Rumancik <lrumancik@google.com>

Check wiping of dir entry data upon removing a file, converting to an
htree, and splitting htree nodes.

Tests commit 6c0912739699d8e4b6a87086401bf3ad3c59502d ("ext4: wipe
ext4_dir_entry2 upon file deletion").

Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com>
---
 tests/ext4/309     | 193 +++++++++++++++++++++++++++++++++++++++++++++
 tests/ext4/309.out |   5 ++
 tests/ext4/group   |   1 +
 3 files changed, 199 insertions(+)
 create mode 100755 tests/ext4/309
 create mode 100644 tests/ext4/309.out

Comments

Eryu Guan May 16, 2021, 4:11 p.m. UTC | #1
On Tue, May 04, 2021 at 07:34:25PM +0000, Leah Rumancik wrote:
> From: Leah Rumancik <lrumancik@google.com>
> 
> Check wiping of dir entry data upon removing a file, converting to an
> htree, and splitting htree nodes.
> 
> Tests commit 6c0912739699d8e4b6a87086401bf3ad3c59502d ("ext4: wipe
> ext4_dir_entry2 upon file deletion").
> 
> Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com>
> ---
>  tests/ext4/309     | 193 +++++++++++++++++++++++++++++++++++++++++++++
>  tests/ext4/309.out |   5 ++
>  tests/ext4/group   |   1 +
>  3 files changed, 199 insertions(+)
>  create mode 100755 tests/ext4/309
>  create mode 100644 tests/ext4/309.out
> 
> diff --git a/tests/ext4/309 b/tests/ext4/309
> new file mode 100755
> index 00000000..c46db688
> --- /dev/null
> +++ b/tests/ext4/309
> @@ -0,0 +1,193 @@
> +#!/bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2021 Google, Inc. All Rights Reserved.
> +#
> +# FS QA Test No. 309
> +#
> +# Test wiping of ext4_dir_entry2 data upon file removal, conversion
> +# to htree, and splitting of htree nodes
> +#
> +seq=`basename $0`
> +seqres=$RESULT_DIR/$seq
> +echo "QA output created by $seq"
> +
> +status=1       # failure is the default!
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# real QA test starts here
> +_supported_fs ext4
> +
> +_require_scratch
> +_require_dumpe2fs "$DUMPE2FS_PROG" dumpe2fs
> +_require_command "$DEBUGFS_PROG" debugfs
> +
> +
> +SCRATCH_DIR="${SCRATCH_MNT}/scratch_dir"

SCRATCH_DIR could be easily confused with SCRATCH_MNT. 'testdir' with
lower case variable name should be fine.

> +
> +# get block number filename's dir ent
> +# argument 1: filename
> +get_block() {
> +  echo $($DEBUGFS_PROG $SCRATCH_DEV -R "dirsearch /scratch_dir $1" 2>> $seqres.full | grep -o -m 1 "phys [0-9]\+" | cut -c 6-)

Please use tab for indention.

Thanks,
Eryu

> +}
> +
> +# get offset of filename's dirent within the block
> +# argument 1: filename
> +get_offset() {
> +  echo $($DEBUGFS_PROG $SCRATCH_DEV -R "dirsearch /scratch_dir $1" 2>> $seqres.full | grep -o -m 1 "offset [0-9]\+" | cut -c 8-)
> +}
> +
> +# get record length of dir ent at specified block and offset
> +# argument 1: block
> +# argument 2: offset
> +get_reclen() {
> +  echo $(od $SCRATCH_DEV --skip-bytes=$(($1 * $blocksize + $2 + 4)) --read-bytes=2  -d -An  --endian=little | tr -d ' \t\n\r')
> +}
> +
> +# reads portion of dirent that should be zero'd out (starting at offset of name_len = 6)
> +# and trims 0s and whitespace
> +# argument 1: block num containing dir ent
> +# argument 2: offset of dir ent within block
> +# argument 3: rec len of dir ent
> +read_dir_ent() {
> + echo $(od $SCRATCH_DEV --skip-bytes=$(($1 * $blocksize + $2 + 6)) --read-bytes=$(($3 - 6)) -d -An -v | sed -e 's/[0 \t\n\r]//g')
> +}
> +
> +# forces node split on test directory
> +# can be used to convert to htree and to split node on existing htree
> +# looks for jump in directory size as indicator of node split
> +induce_node_split() {
> +  _scratch_mount >> $seqres.full 2>&1
> +  dir_size="$(stat --printf="%s" $SCRATCH_DIR)"
> +  while [[ "$(stat --printf="%s" $SCRATCH_DIR)" == "$dir_size" ]]; do
> +    file_num=$(($file_num + 1))
> +    touch $SCRATCH_DIR/test"$(printf "%04d" $file_num)"
> +  done
> +  _scratch_unmount >> $seqres.full 2>&1
> +}
> +
> +#
> +# TEST 1: dir entry fields wiped upon file removal
> +#
> +
> +test_file1="test0001"
> +test_file2="test0002"
> +test_file3="test0003"
> +
> +_scratch_mkfs_sized $((128 * 1024 * 1024)) >> $seqres.full 2>&1
> +blocksize=$($DUMPE2FS_PROG -h $SCRATCH_DEV 2>> $seqres.full | grep "Block size:" | cut -d ":" -f2 | tr -d " \t\n\r")

How about doing "_get_block_size $SCRATCH_MNT" after _scratch_mount ?

> +
> +# create scratch dir for testing
> +# create some files with no name a substr of another name so we can grep later
> +_scratch_mount >> $seqres.full 2>&1
> +mkdir $SCRATCH_DIR
> +file_num=1
> +for file_num in {1..10}; do
> +  touch $SCRATCH_DIR/test"$(printf "%04d" $file_num)"
> +done
> +_scratch_unmount >> $seqres.full 2>&1
> +
> +# get block, offset, and rec_len of two test files
> +block1=$(get_block $test_file1)
> +offset1=$(get_offset $test_file1)
> +rec_len1=$(get_reclen $block1 $offset1)
> +
> +block2=$(get_block $test_file2)
> +offset2=$(get_offset $test_file2)
> +rec_len2=$(get_reclen $block2 $offset2)
> +
> +_scratch_mount >> $seqres.full 2>&1
> +rm $SCRATCH_DIR/$test_file1
> +_scratch_unmount >> $seqres.full 2>&1
> +
> +# read name_len field to end of dir entry
> +check1=$(read_dir_ent $block1 $offset1 $rec_len1)
> +check2=$(read_dir_ent $block2 $offset2 $rec_len2)
> +
> +# if check is empty, bytes read was all 0's, file data wiped
> +# at this point, check1 should be empty, but check 2 should not be
> +if [ -z "$check1" ] && [ ! -z "$check2" ]; then
> +  echo "Test 1 part 1 passed."
> +else
> +  _fail "ERROR (test 1 part 1): metadata not wiped upon removing test file 1"
> +fi
> +
> +_scratch_mount >> $seqres.full 2>&1
> +rm $SCRATCH_DIR/$test_file2
> +_scratch_unmount >> $seqres.full 2>&1
> +
> +check2=$(read_dir_ent $block2 $offset2 $rec_len2)
> +
> +# at this point, both should be wiped
> +[ -z "$check2" ] && echo "Test 1 part 2 passed." || _fail "ERROR (test 1 part 2): metadata not wiped upon removing test file 2"
> +
> +#
> +# TEST 2: old dir entry fields wiped when directory converted to htree
> +#
> +
> +# get original location
> +block1=$(get_block $test_file3)
> +offset1=$(get_offset $test_file3)
> +rec_len1=$(get_reclen $block1 $offset1)
> +
> +# sanity check, ensures not an htree yet
> +check_htree=$($DEBUGFS_PROG $SCRATCH_DEV -R "htree_dump /scratch_dir" 2>&1)
> +if [[ "$check_htree" != *"htree_dump: Not a hash-indexed directory"* ]]; then
> +  _fail "ERROR (test 2): already an htree"
> +fi
> +
> +# force conversion to htree
> +induce_node_split
> +
> +# ensure it is now an htree
> +check_htree=$($DEBUGFS_PROG $SCRATCH_DEV -R "htree_dump /scratch_dir" 2>&1)
> +if [[ "$check_htree" == *"htree_dump: Not a hash-indexed directory"* ]]; then
> +  _fail "ERROR (test 2): directory was not converted to an htree after creation of many files"
> +fi
> +
> +# check that old data was wiped
> +# (this location is not immediately reused by ext4)
> +check1=$(read_dir_ent $block1 $offset1 $rec_len1)
> +
> +# at this point, check1 should be empty meaning data was wiped
> +[ -z "$check1" ] &&  echo "Test 2 passed." || _fail "ERROR (test 2): file metadata not wiped during conversion to htree"
> +
> +#
> +# TEST 3: old dir entries wiped when moved to another block during split_node
> +#
> +
> +# force splitting of a node
> +induce_node_split
> +# use debugfs to get names of two files from block 3
> +hdump=$($DEBUGFS_PROG $SCRATCH_DEV -R "htree_dump /scratch_dir" 2>> $seqres.full)
> +
> +# get line number of "Reading directory block 3"
> +block3_line=$(echo "$hdump" | awk '/Reading directory block 3/{ print NR; exit }')
> +
> +[ -z "$block3_line" ] && echo "ERROR (test 3): could not find block number 3 after node split"
> +
> +test_file1=$(echo "$hdump" | sed -n "$(($block3_line + 1))"p | cut -d ' ' -f4)
> +test_file2=$(echo "$hdump" | sed -n "$(($block3_line + 2))"p | cut -d ' ' -f4)
> +
> +# check these filenames don't exist in block 1 or 2
> +# get block numbers of first two blocks
> +block1=$(echo "$hdump" | grep -o -m 1 "Reading directory block 1, phys [0-9]\+" | cut -c 33-)
> +block2=$(echo "$hdump" | grep -o -m 1 "Reading directory block 2, phys [0-9]\+" | cut -c 33-)
> +
> +# search all of both these blocks for these file names
> +check1=$(od $SCRATCH_DEV --skip-bytes=$(($block1 * $blocksize)) --read-bytes=$blocksize -c -An -v | tr -d '\\ \t\n\r\v' | grep -e $test_file1 -e $test_file2)
> +check2=$(od $SCRATCH_DEV --skip-bytes=$(($block2 * $blocksize)) --read-bytes=$blocksize -c -An -v | tr -d '\\ \t\n\r\v' | grep -e $test_file1 -e $test_file2)
> +
> +if [ -z "$check1" ] && [ -z "$check2" ]; then
> +  echo "Test 3 passed."
> +else
> +  _fail "ERROR (test 3): file name not wiped during node split"
> +fi
> +
> +status=0
> +exit
> diff --git a/tests/ext4/309.out b/tests/ext4/309.out
> new file mode 100644
> index 00000000..e5febaac
> --- /dev/null
> +++ b/tests/ext4/309.out
> @@ -0,0 +1,5 @@
> +QA output created by 309
> +Test 1 part 1 passed.
> +Test 1 part 2 passed.
> +Test 2 passed.
> +Test 3 passed.
> diff --git a/tests/ext4/group b/tests/ext4/group
> index ceda2ba6..e7ad3c24 100644
> --- a/tests/ext4/group
> +++ b/tests/ext4/group
> @@ -59,3 +59,4 @@
>  306 auto rw resize quick
>  307 auto ioctl rw defrag
>  308 auto ioctl rw prealloc quick defrag
> +309 auto quick dir
> -- 
> 2.31.1.527.g47e6f16901-goog
Leah Rumancik May 17, 2021, 1:41 p.m. UTC | #2
On Mon, May 17, 2021 at 12:11:08AM +0800, Eryu Guan wrote:
> On Tue, May 04, 2021 at 07:34:25PM +0000, Leah Rumancik wrote:
> > From: Leah Rumancik <lrumancik@google.com>
> > 
> > Check wiping of dir entry data upon removing a file, converting to an
> > htree, and splitting htree nodes.
> > 
> > Tests commit 6c0912739699d8e4b6a87086401bf3ad3c59502d ("ext4: wipe
> > ext4_dir_entry2 upon file deletion").
> > 
> > Signed-off-by: Leah Rumancik <leah.rumancik@gmail.com>
> > ---
> >  tests/ext4/309     | 193 +++++++++++++++++++++++++++++++++++++++++++++
> >  tests/ext4/309.out |   5 ++
> >  tests/ext4/group   |   1 +
> >  3 files changed, 199 insertions(+)
> >  create mode 100755 tests/ext4/309
> >  create mode 100644 tests/ext4/309.out
> > 
> > diff --git a/tests/ext4/309 b/tests/ext4/309
> > new file mode 100755
> > index 00000000..c46db688
> > --- /dev/null
> > +++ b/tests/ext4/309
> > @@ -0,0 +1,193 @@
> > +#!/bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (c) 2021 Google, Inc. All Rights Reserved.
> > +#
> > +# FS QA Test No. 309
> > +#
> > +# Test wiping of ext4_dir_entry2 data upon file removal, conversion
> > +# to htree, and splitting of htree nodes
> > +#
> > +seq=`basename $0`
> > +seqres=$RESULT_DIR/$seq
> > +echo "QA output created by $seq"
> > +
> > +status=1       # failure is the default!
> > +
> > +# get standard environment, filters and checks
> > +. ./common/rc
> > +. ./common/filter
> > +
> > +# remove previous $seqres.full before test
> > +rm -f $seqres.full
> > +
> > +# real QA test starts here
> > +_supported_fs ext4
> > +
> > +_require_scratch
> > +_require_dumpe2fs "$DUMPE2FS_PROG" dumpe2fs
> > +_require_command "$DEBUGFS_PROG" debugfs
> > +
> > +
> > +SCRATCH_DIR="${SCRATCH_MNT}/scratch_dir"
> 
> SCRATCH_DIR could be easily confused with SCRATCH_MNT. 'testdir' with
> lower case variable name should be fine.
> 
> > +
> > +# get block number filename's dir ent
> > +# argument 1: filename
> > +get_block() {
> > +  echo $($DEBUGFS_PROG $SCRATCH_DEV -R "dirsearch /scratch_dir $1" 2>> $seqres.full | grep -o -m 1 "phys [0-9]\+" | cut -c 6-)
> 
> Please use tab for indention.
> 
> Thanks,
> Eryu
> 
> > +_scratch_mkfs_sized $((128 * 1024 * 1024)) >> $seqres.full 2>&1
> > +blocksize=$($DUMPE2FS_PROG -h $SCRATCH_DEV 2>> $seqres.full | grep "Block size:" | cut -d ":" -f2 | tr -d " \t\n\r")
> 
> How about doing "_get_block_size $SCRATCH_MNT" after _scratch_mount ?
>

Sure, this all sounds good. I'll send out an updated version.

Thanks,
Leah
diff mbox series

Patch

diff --git a/tests/ext4/309 b/tests/ext4/309
new file mode 100755
index 00000000..c46db688
--- /dev/null
+++ b/tests/ext4/309
@@ -0,0 +1,193 @@ 
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2021 Google, Inc. All Rights Reserved.
+#
+# FS QA Test No. 309
+#
+# Test wiping of ext4_dir_entry2 data upon file removal, conversion
+# to htree, and splitting of htree nodes
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+status=1       # failure is the default!
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+_supported_fs ext4
+
+_require_scratch
+_require_dumpe2fs "$DUMPE2FS_PROG" dumpe2fs
+_require_command "$DEBUGFS_PROG" debugfs
+
+
+SCRATCH_DIR="${SCRATCH_MNT}/scratch_dir"
+
+# get block number filename's dir ent
+# argument 1: filename
+get_block() {
+  echo $($DEBUGFS_PROG $SCRATCH_DEV -R "dirsearch /scratch_dir $1" 2>> $seqres.full | grep -o -m 1 "phys [0-9]\+" | cut -c 6-)
+}
+
+# get offset of filename's dirent within the block
+# argument 1: filename
+get_offset() {
+  echo $($DEBUGFS_PROG $SCRATCH_DEV -R "dirsearch /scratch_dir $1" 2>> $seqres.full | grep -o -m 1 "offset [0-9]\+" | cut -c 8-)
+}
+
+# get record length of dir ent at specified block and offset
+# argument 1: block
+# argument 2: offset
+get_reclen() {
+  echo $(od $SCRATCH_DEV --skip-bytes=$(($1 * $blocksize + $2 + 4)) --read-bytes=2  -d -An  --endian=little | tr -d ' \t\n\r')
+}
+
+# reads portion of dirent that should be zero'd out (starting at offset of name_len = 6)
+# and trims 0s and whitespace
+# argument 1: block num containing dir ent
+# argument 2: offset of dir ent within block
+# argument 3: rec len of dir ent
+read_dir_ent() {
+ echo $(od $SCRATCH_DEV --skip-bytes=$(($1 * $blocksize + $2 + 6)) --read-bytes=$(($3 - 6)) -d -An -v | sed -e 's/[0 \t\n\r]//g')
+}
+
+# forces node split on test directory
+# can be used to convert to htree and to split node on existing htree
+# looks for jump in directory size as indicator of node split
+induce_node_split() {
+  _scratch_mount >> $seqres.full 2>&1
+  dir_size="$(stat --printf="%s" $SCRATCH_DIR)"
+  while [[ "$(stat --printf="%s" $SCRATCH_DIR)" == "$dir_size" ]]; do
+    file_num=$(($file_num + 1))
+    touch $SCRATCH_DIR/test"$(printf "%04d" $file_num)"
+  done
+  _scratch_unmount >> $seqres.full 2>&1
+}
+
+#
+# TEST 1: dir entry fields wiped upon file removal
+#
+
+test_file1="test0001"
+test_file2="test0002"
+test_file3="test0003"
+
+_scratch_mkfs_sized $((128 * 1024 * 1024)) >> $seqres.full 2>&1
+blocksize=$($DUMPE2FS_PROG -h $SCRATCH_DEV 2>> $seqres.full | grep "Block size:" | cut -d ":" -f2 | tr -d " \t\n\r")
+
+# create scratch dir for testing
+# create some files with no name a substr of another name so we can grep later
+_scratch_mount >> $seqres.full 2>&1
+mkdir $SCRATCH_DIR
+file_num=1
+for file_num in {1..10}; do
+  touch $SCRATCH_DIR/test"$(printf "%04d" $file_num)"
+done
+_scratch_unmount >> $seqres.full 2>&1
+
+# get block, offset, and rec_len of two test files
+block1=$(get_block $test_file1)
+offset1=$(get_offset $test_file1)
+rec_len1=$(get_reclen $block1 $offset1)
+
+block2=$(get_block $test_file2)
+offset2=$(get_offset $test_file2)
+rec_len2=$(get_reclen $block2 $offset2)
+
+_scratch_mount >> $seqres.full 2>&1
+rm $SCRATCH_DIR/$test_file1
+_scratch_unmount >> $seqres.full 2>&1
+
+# read name_len field to end of dir entry
+check1=$(read_dir_ent $block1 $offset1 $rec_len1)
+check2=$(read_dir_ent $block2 $offset2 $rec_len2)
+
+# if check is empty, bytes read was all 0's, file data wiped
+# at this point, check1 should be empty, but check 2 should not be
+if [ -z "$check1" ] && [ ! -z "$check2" ]; then
+  echo "Test 1 part 1 passed."
+else
+  _fail "ERROR (test 1 part 1): metadata not wiped upon removing test file 1"
+fi
+
+_scratch_mount >> $seqres.full 2>&1
+rm $SCRATCH_DIR/$test_file2
+_scratch_unmount >> $seqres.full 2>&1
+
+check2=$(read_dir_ent $block2 $offset2 $rec_len2)
+
+# at this point, both should be wiped
+[ -z "$check2" ] && echo "Test 1 part 2 passed." || _fail "ERROR (test 1 part 2): metadata not wiped upon removing test file 2"
+
+#
+# TEST 2: old dir entry fields wiped when directory converted to htree
+#
+
+# get original location
+block1=$(get_block $test_file3)
+offset1=$(get_offset $test_file3)
+rec_len1=$(get_reclen $block1 $offset1)
+
+# sanity check, ensures not an htree yet
+check_htree=$($DEBUGFS_PROG $SCRATCH_DEV -R "htree_dump /scratch_dir" 2>&1)
+if [[ "$check_htree" != *"htree_dump: Not a hash-indexed directory"* ]]; then
+  _fail "ERROR (test 2): already an htree"
+fi
+
+# force conversion to htree
+induce_node_split
+
+# ensure it is now an htree
+check_htree=$($DEBUGFS_PROG $SCRATCH_DEV -R "htree_dump /scratch_dir" 2>&1)
+if [[ "$check_htree" == *"htree_dump: Not a hash-indexed directory"* ]]; then
+  _fail "ERROR (test 2): directory was not converted to an htree after creation of many files"
+fi
+
+# check that old data was wiped
+# (this location is not immediately reused by ext4)
+check1=$(read_dir_ent $block1 $offset1 $rec_len1)
+
+# at this point, check1 should be empty meaning data was wiped
+[ -z "$check1" ] &&  echo "Test 2 passed." || _fail "ERROR (test 2): file metadata not wiped during conversion to htree"
+
+#
+# TEST 3: old dir entries wiped when moved to another block during split_node
+#
+
+# force splitting of a node
+induce_node_split
+# use debugfs to get names of two files from block 3
+hdump=$($DEBUGFS_PROG $SCRATCH_DEV -R "htree_dump /scratch_dir" 2>> $seqres.full)
+
+# get line number of "Reading directory block 3"
+block3_line=$(echo "$hdump" | awk '/Reading directory block 3/{ print NR; exit }')
+
+[ -z "$block3_line" ] && echo "ERROR (test 3): could not find block number 3 after node split"
+
+test_file1=$(echo "$hdump" | sed -n "$(($block3_line + 1))"p | cut -d ' ' -f4)
+test_file2=$(echo "$hdump" | sed -n "$(($block3_line + 2))"p | cut -d ' ' -f4)
+
+# check these filenames don't exist in block 1 or 2
+# get block numbers of first two blocks
+block1=$(echo "$hdump" | grep -o -m 1 "Reading directory block 1, phys [0-9]\+" | cut -c 33-)
+block2=$(echo "$hdump" | grep -o -m 1 "Reading directory block 2, phys [0-9]\+" | cut -c 33-)
+
+# search all of both these blocks for these file names
+check1=$(od $SCRATCH_DEV --skip-bytes=$(($block1 * $blocksize)) --read-bytes=$blocksize -c -An -v | tr -d '\\ \t\n\r\v' | grep -e $test_file1 -e $test_file2)
+check2=$(od $SCRATCH_DEV --skip-bytes=$(($block2 * $blocksize)) --read-bytes=$blocksize -c -An -v | tr -d '\\ \t\n\r\v' | grep -e $test_file1 -e $test_file2)
+
+if [ -z "$check1" ] && [ -z "$check2" ]; then
+  echo "Test 3 passed."
+else
+  _fail "ERROR (test 3): file name not wiped during node split"
+fi
+
+status=0
+exit
diff --git a/tests/ext4/309.out b/tests/ext4/309.out
new file mode 100644
index 00000000..e5febaac
--- /dev/null
+++ b/tests/ext4/309.out
@@ -0,0 +1,5 @@ 
+QA output created by 309
+Test 1 part 1 passed.
+Test 1 part 2 passed.
+Test 2 passed.
+Test 3 passed.
diff --git a/tests/ext4/group b/tests/ext4/group
index ceda2ba6..e7ad3c24 100644
--- a/tests/ext4/group
+++ b/tests/ext4/group
@@ -59,3 +59,4 @@ 
 306 auto rw resize quick
 307 auto ioctl rw defrag
 308 auto ioctl rw prealloc quick defrag
+309 auto quick dir