Hello,
This patch series improves e2fsprogs for case-insensitive filesystems.
First, it allows tune2fs to enable the 'casefold' feature on existing
filesystems.
Then, it improves e2fsck by allowing it to:
- fix entries containing invalid UTF-8 characters
- detect duplicated entries
By default, invalid filenames are only checked when strict mode is enabled.
A new option is therefore added to allow the user to force this verification.
This series has been tested by running xfstests, and by manually corrupting
the test filesystem using debugfs as well.
Best regards,
Arnaud
---
Changes in v2:
- added missing comment in e2fsck/pass1.c
- added a new problem code dedicated to bad encoded file names
- reworked a test in e2fsck/pass2.c
Arnaud Ferraris (1):
e2fsck: add new problem for casefolded name check
Gabriel Krisman Bertazi (11):
tune2fs: Allow enabling casefold feature after fs creation
tune2fs: Fix casefold+encrypt error message
ext2fs: Add method to validate casefolded strings
ext2fs: Implement faster CI comparison of strings
e2fsck: Fix entries with invalid encoded characters
e2fsck: Support casefold directories when rehashing
dict: Support comparison with context
e2fsck: Detect duplicated casefolded direntries for rehash
e2fsck: Add option to force encoded filename verification
e2fsck.8.in: Document check_encoding extended option
tests: f_bad_fname: Test fixes of invalid filenames and duplicates
e2fsck/e2fsck.8.in | 4 ++
e2fsck/e2fsck.c | 4 ++
e2fsck/e2fsck.h | 2 +
e2fsck/pass1.c | 18 ++++++++
e2fsck/pass1b.c | 2 +-
e2fsck/pass2.c | 76 +++++++++++++++++++++++++++++---
e2fsck/problem.c | 5 +++
e2fsck/problem.h | 3 ++
e2fsck/rehash.c | 88 ++++++++++++++++++++++++++++++-------
e2fsck/unix.c | 4 ++
lib/ext2fs/ext2fs.h | 6 +++
lib/ext2fs/ext2fsP.h | 6 +++
lib/ext2fs/nls_utf8.c | 62 ++++++++++++++++++++++++++
lib/support/dict.c | 22 +++++++---
lib/support/dict.h | 4 +-
lib/support/mkquota.c | 2 +-
misc/tune2fs.c | 18 +++++++-
tests/f_bad_fname/expect.1 | 22 ++++++++++
tests/f_bad_fname/expect.2 | 7 +++
tests/f_bad_fname/image.gz | Bin 0 -> 802 bytes
tests/f_bad_fname/name | 1 +
21 files changed, 322 insertions(+), 34 deletions(-)
create mode 100644 tests/f_bad_fname/expect.1
create mode 100644 tests/f_bad_fname/expect.2
create mode 100644 tests/f_bad_fname/image.gz
create mode 100644 tests/f_bad_fname/name
--
2.29.2
From: Gabriel Krisman Bertazi <[email protected]>
This is exported to be used by fsck.
Signed-off-by: Gabriel Krisman Bertazi <[email protected]>
Signed-off-by: Arnaud Ferraris <[email protected]>
---
lib/ext2fs/ext2fs.h | 2 ++
lib/ext2fs/ext2fsP.h | 2 ++
lib/ext2fs/nls_utf8.c | 29 +++++++++++++++++++++++++++++
3 files changed, 33 insertions(+)
diff --git a/lib/ext2fs/ext2fs.h b/lib/ext2fs/ext2fs.h
index 69c8a3ff..4065cb70 100644
--- a/lib/ext2fs/ext2fs.h
+++ b/lib/ext2fs/ext2fs.h
@@ -1613,6 +1613,8 @@ extern errcode_t ext2fs_new_dir_inline_data(ext2_filsys fs, ext2_ino_t dir_ino,
/* nls_utf8.c */
extern const struct ext2fs_nls_table *ext2fs_load_nls_table(int encoding);
+extern int ext2fs_check_encoded_name(const struct ext2fs_nls_table *table,
+ char *s, size_t len, char **pos);
/* mkdir.c */
extern errcode_t ext2fs_mkdir(ext2_filsys fs, ext2_ino_t parent, ext2_ino_t inum,
diff --git a/lib/ext2fs/ext2fsP.h b/lib/ext2fs/ext2fsP.h
index ad8b7d52..30564ded 100644
--- a/lib/ext2fs/ext2fsP.h
+++ b/lib/ext2fs/ext2fsP.h
@@ -104,6 +104,8 @@ struct ext2fs_nls_ops {
int (*casefold)(const struct ext2fs_nls_table *charset,
const unsigned char *str, size_t len,
unsigned char *dest, size_t dlen);
+ int (*validate)(const struct ext2fs_nls_table *table,
+ char *s, size_t len, char **pos);
};
/* Function prototypes */
diff --git a/lib/ext2fs/nls_utf8.c b/lib/ext2fs/nls_utf8.c
index e4c4e7a3..903c65ba 100644
--- a/lib/ext2fs/nls_utf8.c
+++ b/lib/ext2fs/nls_utf8.c
@@ -920,8 +920,31 @@ invalid_seq:
return -EINVAL;
}
+
+static int utf8_validate(const struct ext2fs_nls_table *table,
+ char *s, size_t len, char **pos)
+{
+ const struct utf8data *data = utf8nfdicf(table->version);
+ utf8leaf_t *leaf;
+ unsigned char hangul[UTF8HANGULLEAF];
+
+ if (!data)
+ return -1;
+ while (len && *s) {
+ leaf = utf8nlookup(data, hangul, s, len);
+ if (!leaf) {
+ *pos = s;
+ return 1;
+ }
+ len -= utf8clen(s);
+ s += utf8clen(s);
+ }
+ return 0;
+}
+
static const struct ext2fs_nls_ops utf8_ops = {
.casefold = utf8_casefold,
+ .validate = utf8_validate,
};
static const struct ext2fs_nls_table nls_utf8 = {
@@ -936,3 +959,9 @@ const struct ext2fs_nls_table *ext2fs_load_nls_table(int encoding)
return NULL;
}
+
+int ext2fs_check_encoded_name(const struct ext2fs_nls_table *table,
+ char *name, size_t len, char **pos)
+{
+ return table->ops->validate(table, name, len, pos);
+}
--
2.29.2
From: Gabriel Krisman Bertazi <[email protected]>
On strict mode, invalid Unicode sequences are not permited. This patch
adds a verification step to pass2 to detect and modify the entries with
the same replacement char used for non-encoding directories '.'.
After the encoding test, we still want to check the name for usual
problems, '\0', '/' in the middle of the sequence.
Signed-off-by: Gabriel Krisman Bertazi <[email protected]>
Signed-off-by: Arnaud Ferraris <[email protected]>
---
Changes in v2:
- added missing comment
- uses the problem code introduced by the previous patch
- reworked a test to ease future support of encrypted+casefolded
directories
e2fsck/e2fsck.c | 4 ++++
e2fsck/e2fsck.h | 1 +
e2fsck/pass1.c | 18 +++++++++++++++++
e2fsck/pass2.c | 51 ++++++++++++++++++++++++++++++++++++++++++++-----
4 files changed, 69 insertions(+), 5 deletions(-)
diff --git a/e2fsck/e2fsck.c b/e2fsck/e2fsck.c
index d8be566f..dc4b45e2 100644
--- a/e2fsck/e2fsck.c
+++ b/e2fsck/e2fsck.c
@@ -75,6 +75,10 @@ errcode_t e2fsck_reset_context(e2fsck_t ctx)
ext2fs_free_block_bitmap(ctx->block_found_map);
ctx->block_found_map = 0;
}
+ if (ctx->inode_casefold_map) {
+ ext2fs_free_block_bitmap(ctx->inode_casefold_map);
+ ctx->inode_casefold_map = 0;
+ }
if (ctx->inode_link_info) {
ext2fs_free_icount(ctx->inode_link_info);
ctx->inode_link_info = 0;
diff --git a/e2fsck/e2fsck.h b/e2fsck/e2fsck.h
index 85f953b2..dcaab0a1 100644
--- a/e2fsck/e2fsck.h
+++ b/e2fsck/e2fsck.h
@@ -262,6 +262,7 @@ struct e2fsck_struct {
ext2fs_inode_bitmap inode_bb_map; /* Inodes which are in bad blocks */
ext2fs_inode_bitmap inode_imagic_map; /* AFS inodes */
ext2fs_inode_bitmap inode_reg_map; /* Inodes which are regular files*/
+ ext2fs_inode_bitmap inode_casefold_map; /* Inodes which are casefolded */
ext2fs_block_bitmap block_found_map; /* Blocks which are in use */
ext2fs_block_bitmap block_dup_map; /* Blks referenced more than once */
diff --git a/e2fsck/pass1.c b/e2fsck/pass1.c
index 8eecd958..6909fed5 100644
--- a/e2fsck/pass1.c
+++ b/e2fsck/pass1.c
@@ -23,6 +23,7 @@
* - A bitmap of which inodes have bad fields. (inode_bad_map)
* - A bitmap of which inodes are in bad blocks. (inode_bb_map)
* - A bitmap of which inodes are imagic inodes. (inode_imagic_map)
+ * - A bitmap of which inodes are casefolded. (inode_casefold_map)
* - A bitmap of which blocks are in use. (block_found_map)
* - A bitmap of which blocks are in use by two inodes (block_dup_map)
* - The data blocks of the directory inodes. (dir_map)
@@ -1260,6 +1261,20 @@ void e2fsck_pass1(e2fsck_t ctx)
ctx->flags |= E2F_FLAG_ABORT;
return;
}
+ if (casefold_fs) {
+ pctx.errcode =
+ e2fsck_allocate_inode_bitmap(fs,
+ _("inode casefold map"),
+ EXT2FS_BMAP64_RBTREE,
+ "inode_casefold_map",
+ &ctx->inode_casefold_map);
+ if (pctx.errcode) {
+ pctx.num = 1;
+ fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
+ ctx->flags |= E2F_FLAG_ABORT;
+ return;
+ }
+ }
pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
&ctx->inode_link_info);
if (pctx.errcode) {
@@ -1870,6 +1885,9 @@ void e2fsck_pass1(e2fsck_t ctx)
add_encrypted_file(ctx, &pctx) < 0)
goto clear_inode;
+ if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
+ ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
+
if (LINUX_S_ISDIR(inode->i_mode)) {
ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
e2fsck_add_dir_info(ctx, ino, 0);
diff --git a/e2fsck/pass2.c b/e2fsck/pass2.c
index 4dbc44ea..b9402b24 100644
--- a/e2fsck/pass2.c
+++ b/e2fsck/pass2.c
@@ -36,11 +36,13 @@
* - The inode_bad_map bitmap
* - The inode_dir_map bitmap
* - The encrypted_file_info
+ * - The inode_casefold_map bitmap
*
* Pass 2 frees the following data structures
* - The inode_bad_map bitmap
* - The inode_reg_map bitmap
* - The encrypted_file_info
+ * - The inode_casefold_map bitmap
*/
#define _GNU_SOURCE 1 /* get strnlen() */
@@ -287,6 +289,10 @@ void e2fsck_pass2(e2fsck_t ctx)
ext2fs_free_inode_bitmap(ctx->inode_reg_map);
ctx->inode_reg_map = 0;
}
+ if (ctx->inode_casefold_map) {
+ ext2fs_free_inode_bitmap(ctx->inode_casefold_map);
+ ctx->inode_casefold_map = 0;
+ }
destroy_encrypted_file_info(ctx);
clear_problem_context(&pctx);
@@ -515,6 +521,30 @@ static int encrypted_check_name(e2fsck_t ctx,
return 0;
}
+static int encoded_check_name(e2fsck_t ctx,
+ struct ext2_dir_entry *dirent,
+ struct problem_context *pctx)
+{
+ const struct ext2fs_nls_table *tbl = ctx->fs->encoding;
+ int ret;
+ int len = ext2fs_dirent_name_len(dirent);
+ char *pos, *end;
+
+ ret = ext2fs_check_encoded_name(tbl, dirent->name, len, &pos);
+ if (ret < 0) {
+ fatal_error(ctx, _("NLS is broken."));
+ } else if(ret > 0) {
+ ret = fix_problem(ctx, PR_2_BAD_CASEFOLDED_NAME, pctx);
+ if (ret) {
+ end = &dirent->name[len];
+ for (; *pos && pos != end; pos++)
+ *pos = '.';
+ }
+ }
+
+ return (ret || check_name(ctx, dirent, pctx));
+}
+
/*
* Check the directory filetype (if present)
*/
@@ -998,11 +1028,18 @@ static int check_dir_block(ext2_filsys fs,
size_t max_block_size;
int hash_flags = 0;
static char *eop_read_dirblock = NULL;
+ int cf_dir = 0;
cd = (struct check_dir_struct *) priv_data;
ibuf = buf = cd->buf;
ctx = cd->ctx;
+ /* We only want filename encoding verification on strict
+ * mode. */
+ if (ext2fs_test_inode_bitmap2(ctx->inode_casefold_map, ino) &&
+ (ctx->fs->super->s_encoding_flags & EXT4_ENC_STRICT_MODE_FL))
+ cf_dir = 1;
+
if (ctx->flags & E2F_FLAG_RUN_RETURN)
return DIRENT_ABORT;
@@ -1483,11 +1520,7 @@ skip_checksum:
if (check_filetype(ctx, dirent, ino, &cd->pctx))
dir_modified++;
- if (dir_encpolicy_id == NO_ENCRYPTION_POLICY) {
- /* Unencrypted directory */
- if (check_name(ctx, dirent, &cd->pctx))
- dir_modified++;
- } else {
+ if (dir_encpolicy_id != NO_ENCRYPTION_POLICY) {
/* Encrypted directory */
if (dot_state > 1 &&
check_encrypted_dirent(ctx, dirent,
@@ -1497,6 +1530,14 @@ skip_checksum:
dir_modified++;
goto next;
}
+ } else if (cf_dir) {
+ /* Casefolded directory */
+ if (encoded_check_name(ctx, dirent, &cd->pctx))
+ dir_modified++;
+ } else {
+ /* Unencrypted and uncasefolded directory */
+ if (check_name(ctx, dirent, &cd->pctx))
+ dir_modified++;
}
if (dx_db) {
--
2.29.2
From: Gabriel Krisman Bertazi <[email protected]>
Signed-off-by: Gabriel Krisman Bertazi <[email protected]>
Signed-off-by: Arnaud Ferraris <[email protected]>
---
tests/f_bad_fname/expect.1 | 22 ++++++++++++++++++++++
tests/f_bad_fname/expect.2 | 7 +++++++
tests/f_bad_fname/image.gz | Bin 0 -> 802 bytes
tests/f_bad_fname/name | 1 +
4 files changed, 30 insertions(+)
create mode 100644 tests/f_bad_fname/expect.1
create mode 100644 tests/f_bad_fname/expect.2
create mode 100644 tests/f_bad_fname/image.gz
create mode 100644 tests/f_bad_fname/name
diff --git a/tests/f_bad_fname/expect.1 b/tests/f_bad_fname/expect.1
new file mode 100644
index 00000000..1d860b22
--- /dev/null
+++ b/tests/f_bad_fname/expect.1
@@ -0,0 +1,22 @@
+Pass 1: Checking inodes, blocks, and sizes
+Pass 2: Checking directory structure
+Entry 'AM-^?' in /ci_dir (12) has illegal characters in its name.
+Fix? yes
+
+Entry 'AM-~' in /ci_dir (12) has illegal characters in its name.
+Fix? yes
+
+Duplicate entry 'A.' found.
+ Marking /ci_dir (12) to be rebuilt.
+
+Pass 3: Checking directory connectivity
+Pass 3A: Optimizing directories
+Entry 'A.' in /ci_dir (12) has a non-unique filename.
+Rename to A.~0? yes
+
+Pass 4: Checking reference counts
+Pass 5: Checking group summary information
+
+test_filesys: ***** FILE SYSTEM WAS MODIFIED *****
+test_filesys: 14/16 files (0.0% non-contiguous), 22/100 blocks
+Exit status is 1
diff --git a/tests/f_bad_fname/expect.2 b/tests/f_bad_fname/expect.2
new file mode 100644
index 00000000..13de1c08
--- /dev/null
+++ b/tests/f_bad_fname/expect.2
@@ -0,0 +1,7 @@
+Pass 1: Checking inodes, blocks, and sizes
+Pass 2: Checking directory structure
+Pass 3: Checking directory connectivity
+Pass 4: Checking reference counts
+Pass 5: Checking group summary information
+test_filesys: 14/16 files (0.0% non-contiguous), 22/100 blocks
+Exit status is 0
diff --git a/tests/f_bad_fname/image.gz b/tests/f_bad_fname/image.gz
new file mode 100644
index 0000000000000000000000000000000000000000..a8b3fc6b8397a7859d9697c462f24f498bb57fd8
GIT binary patch
literal 802
zcmb2|=HU3ZwK|T0IWspgJ(c0@9p4PuP!Wa)#-G*9mUQmd6)h1gP)&N{wkF^Lhf?9g
zMNtKcnpXpOe4{cJM+7ff`jtKW--4vV=~{YsIv=@RXj&kBGDu*_q5yNH8;i;k=a=78
z@%4!p{B((@Y#;x**}v1?o!R+$Msd2?XQhT^yX=m-bnO%AUveTi<+L?Vc;L6)A5Olh
zkgUC!Xa6hu>if4lZ+<+wuKRbcPoc`uZ6eD**?bcdtk2o`_gzv|PMx0hpO>@$rmu^(
z?AyKU``I-=Pp&oC_H)|fqtg6-`@Wz0F1_p<&)<1R{~GSt@b*UM($5jzk<&|Eqb!S5
zC;dI25?e2O?evSo@-lf#pZaR%b#MCjIm7kh&2Lv9-_JR}c*edv`_`Z5uit04m}T9R
zvg*S}s~^PYF7Q0{{O`dNpFUM?$$I|tgvVE&{hDl_R{mvAe|KZ^|CzPI7iX&P%Zr_C
zZ@M!(>662efM5S2S5LLK+`4<&P40-^kkj^ie!o3`-2B-7*Z-%7h5kQqb@iJ6?SA`q
zEzOyqebrVf!FKA``uEYh>$GmalAq=J*SUCQeTLpw{<BN}E!=s<-emf(_Gdx=FT?;P
zy@8Ut>+2WHWLy0EcfE6{wREm~%<NmA*IwBd@H(K;H0a*{^XGG{ukTy`&U`}Su8{AQ
z@BceZp7Q<wb+ftkJKq&Dpn&P+Z;rhAk+dfKoOr+6AFEkXu17w<Ki@Zh-s(lSqDt<$
z_q}+p!+tI;b$j=wtnKr7?5{m$tlIMVebwjxKa=<Ve7|$+y+8WCt}>q26LNIByx0HF
zdH5@Ouk^p<-#3bF)m~g&Irr<c^Y&}Z{~y2aS`&EAK6>}r`pP?ZBEOZ}g`cy3`d{tq
z*BSM{UoQRmzSZQvY*fe5KXYH}*Z*qVr}y*p&-cgVKG&N_7JkaOdT;h8y<<UbLG{Z`
z3(CHJe*4V2-2Qg%U+dGK<aa&!b6u~#rZUZa`&H|C5pNk8Q9MX0=fSHl{Bc6hCNNB3
GWB>rT_?(Xb
literal 0
HcmV?d00001
diff --git a/tests/f_bad_fname/name b/tests/f_bad_fname/name
new file mode 100644
index 00000000..675165a6
--- /dev/null
+++ b/tests/f_bad_fname/name
@@ -0,0 +1 @@
+Case-insensitive directory with broken file names
--
2.29.2
---
Changes in v2:
- added in this version
e2fsck/problem.c | 5 +++++
e2fsck/problem.h | 3 +++
2 files changed, 8 insertions(+)
diff --git a/e2fsck/problem.c b/e2fsck/problem.c
index e79c853b..2b596303 100644
--- a/e2fsck/problem.c
+++ b/e2fsck/problem.c
@@ -1805,6 +1805,11 @@ static struct e2fsck_problem problem_table[] = {
N_("Encrypted @E references @i %Di, which has a different encryption policy.\n"),
PROMPT_CLEAR, 0, 0, 0, 0 },
+ /* Casefolded directory entry has illegal characters in its name */
+ { PR_2_BAD_CASEFOLDED_NAME,
+ N_("@E has illegal UTF-8 characters in its name.\n"),
+ PROMPT_FIX, 0, 0, 0, 0 },
+
/* Pass 3 errors */
/* Pass 3: Checking directory connectivity */
diff --git a/e2fsck/problem.h b/e2fsck/problem.h
index 4185e517..a8806fd4 100644
--- a/e2fsck/problem.h
+++ b/e2fsck/problem.h
@@ -1028,6 +1028,9 @@ struct problem_context {
/* Encrypted directory contains file with different encryption policy */
#define PR_2_INCONSISTENT_ENCRYPTION_POLICY 0x020052
+/* Casefolded directory entry has illegal characters in its name */
+#define PR_2_BAD_CASEFOLDED_NAME 0x0200053
+
/*
* Pass 3 errors
*/
--
2.29.2
From: Gabriel Krisman Bertazi <[email protected]>
Instead of calling casefold two times and memcmp the result, which
require allocating a temporary buffer for the casefolded version, add a
strcasecmp-like method to perform the comparison of each code-point
during the casefold itself.
This method is exposed because it needs to be used directly by fsck.
Signed-off-by: Gabriel Krisman Bertazi <[email protected]>
Signed-off-by: Arnaud Ferraris <[email protected]>
---
lib/ext2fs/ext2fs.h | 4 ++++
lib/ext2fs/ext2fsP.h | 4 ++++
lib/ext2fs/nls_utf8.c | 33 +++++++++++++++++++++++++++++++++
3 files changed, 41 insertions(+)
diff --git a/lib/ext2fs/ext2fs.h b/lib/ext2fs/ext2fs.h
index 4065cb70..9e96ca5c 100644
--- a/lib/ext2fs/ext2fs.h
+++ b/lib/ext2fs/ext2fs.h
@@ -1615,6 +1615,10 @@ extern errcode_t ext2fs_new_dir_inline_data(ext2_filsys fs, ext2_ino_t dir_ino,
extern const struct ext2fs_nls_table *ext2fs_load_nls_table(int encoding);
extern int ext2fs_check_encoded_name(const struct ext2fs_nls_table *table,
char *s, size_t len, char **pos);
+extern int ext2fs_casefold_cmp(const struct ext2fs_nls_table *table,
+ const unsigned char *str1, size_t len1,
+ const unsigned char *str2, size_t len2);
+
/* mkdir.c */
extern errcode_t ext2fs_mkdir(ext2_filsys fs, ext2_ino_t parent, ext2_ino_t inum,
diff --git a/lib/ext2fs/ext2fsP.h b/lib/ext2fs/ext2fsP.h
index 30564ded..99239be0 100644
--- a/lib/ext2fs/ext2fsP.h
+++ b/lib/ext2fs/ext2fsP.h
@@ -106,6 +106,10 @@ struct ext2fs_nls_ops {
unsigned char *dest, size_t dlen);
int (*validate)(const struct ext2fs_nls_table *table,
char *s, size_t len, char **pos);
+ int (*casefold_cmp)(const struct ext2fs_nls_table *table,
+ const unsigned char *str1, size_t len1,
+ const unsigned char *str2, size_t len2);
+
};
/* Function prototypes */
diff --git a/lib/ext2fs/nls_utf8.c b/lib/ext2fs/nls_utf8.c
index 903c65ba..1c444ca2 100644
--- a/lib/ext2fs/nls_utf8.c
+++ b/lib/ext2fs/nls_utf8.c
@@ -942,9 +942,36 @@ static int utf8_validate(const struct ext2fs_nls_table *table,
return 0;
}
+static int utf8_casefold_cmp(const struct ext2fs_nls_table *table,
+ const unsigned char *str1, size_t len1,
+ const unsigned char *str2, size_t len2)
+{
+ const struct utf8data *data = utf8nfdicf(table->version);
+ int c1, c2;
+ struct utf8cursor cur1, cur2;
+
+ if (utf8ncursor(&cur1, data, (const char *) str1, len1) < 0)
+ return -1;
+ if (utf8ncursor(&cur2, data, (const char *) str2, len2) < 0)
+ return -1;
+
+ do {
+ c1 = utf8byte(&cur1);
+ c2 = utf8byte(&cur2);
+
+ if (c1 < 0 || c2 < 0)
+ return -1;
+ if (c1 != c2)
+ return c1 - c2;
+ } while (c1);
+
+ return 0;
+}
+
static const struct ext2fs_nls_ops utf8_ops = {
.casefold = utf8_casefold,
.validate = utf8_validate,
+ .casefold_cmp = utf8_casefold_cmp,
};
static const struct ext2fs_nls_table nls_utf8 = {
@@ -965,3 +992,9 @@ int ext2fs_check_encoded_name(const struct ext2fs_nls_table *table,
{
return table->ops->validate(table, name, len, pos);
}
+int ext2fs_casefold_cmp(const struct ext2fs_nls_table *table,
+ const unsigned char *str1, size_t len1,
+ const unsigned char *str2, size_t len2)
+{
+ return table->ops->casefold_cmp(table, str1, len1, str2, len2);
+}
--
2.29.2
Arnaud Ferraris <[email protected]> writes:
> ---
> Changes in v2:
> - added in this version
>
> e2fsck/problem.c | 5 +++++
> e2fsck/problem.h | 3 +++
> 2 files changed, 8 insertions(+)
>
> diff --git a/e2fsck/problem.c b/e2fsck/problem.c
> index e79c853b..2b596303 100644
> --- a/e2fsck/problem.c
> +++ b/e2fsck/problem.c
> @@ -1805,6 +1805,11 @@ static struct e2fsck_problem problem_table[] = {
> N_("Encrypted @E references @i %Di, which has a different encryption policy.\n"),
> PROMPT_CLEAR, 0, 0, 0, 0 },
>
> + /* Casefolded directory entry has illegal characters in its name */
> + { PR_2_BAD_CASEFOLDED_NAME,
> + N_("@E has illegal UTF-8 characters in its name.\n"),
> + PROMPT_FIX, 0, 0, 0, 0 },
> +
> /* Pass 3 errors */
>
> /* Pass 3: Checking directory connectivity */
> diff --git a/e2fsck/problem.h b/e2fsck/problem.h
> index 4185e517..a8806fd4 100644
> --- a/e2fsck/problem.h
> +++ b/e2fsck/problem.h
> @@ -1028,6 +1028,9 @@ struct problem_context {
> /* Encrypted directory contains file with different encryption policy */
> #define PR_2_INCONSISTENT_ENCRYPTION_POLICY 0x020052
>
> +/* Casefolded directory entry has illegal characters in its name */
> +#define PR_2_BAD_CASEFOLDED_NAME 0x0200053
This should be 0x020053 (yours has an extra 0)
> +
> /*
> * Pass 3 errors
> */
--
Gabriel Krisman Bertazi
Arnaud Ferraris <[email protected]> writes:
> ---
> Changes in v2:
> - added in this version
>
> e2fsck/problem.c | 5 +++++
> e2fsck/problem.h | 3 +++
> 2 files changed, 8 insertions(+)
>
> diff --git a/e2fsck/problem.c b/e2fsck/problem.c
> index e79c853b..2b596303 100644
> --- a/e2fsck/problem.c
> +++ b/e2fsck/problem.c
> @@ -1805,6 +1805,11 @@ static struct e2fsck_problem problem_table[] = {
> N_("Encrypted @E references @i %Di, which has a different encryption policy.\n"),
> PROMPT_CLEAR, 0, 0, 0, 0 },
>
> + /* Casefolded directory entry has illegal characters in its name */
> + { PR_2_BAD_CASEFOLDED_NAME,
> + N_("@E has illegal UTF-8 characters in its name.\n"),
> + PROMPT_FIX, 0, 0, 0, 0 },
> +
> /* Pass 3 errors */
>
> /* Pass 3: Checking directory connectivity */
> diff --git a/e2fsck/problem.h b/e2fsck/problem.h
> index 4185e517..a8806fd4 100644
> --- a/e2fsck/problem.h
> +++ b/e2fsck/problem.h
> @@ -1028,6 +1028,9 @@ struct problem_context {
> /* Encrypted directory contains file with different encryption policy */
> #define PR_2_INCONSISTENT_ENCRYPTION_POLICY 0x020052
>
> +/* Casefolded directory entry has illegal characters in its name */
> +#define PR_2_BAD_CASEFOLDED_NAME 0x0200053
Also, PR_2_BAD_ENCODED_NAME makes more sense than CASEFOLDED. The
name is encoded in utf-8 but not casefolded on-disk.
> +
> /*
> * Pass 3 errors
> */
--
Gabriel Krisman Bertazi
Arnaud Ferraris <[email protected]> writes:
> From: Gabriel Krisman Bertazi <[email protected]>
>
> On strict mode, invalid Unicode sequences are not permited. This patch
> adds a verification step to pass2 to detect and modify the entries with
> the same replacement char used for non-encoding directories '.'.
>
> After the encoding test, we still want to check the name for usual
> problems, '\0', '/' in the middle of the sequence.
>
> Signed-off-by: Gabriel Krisman Bertazi <[email protected]>
> Signed-off-by: Arnaud Ferraris <[email protected]>
> ---
> Changes in v2:
> - added missing comment
> - uses the problem code introduced by the previous patch
> - reworked a test to ease future support of encrypted+casefolded
> directories
>
> e2fsck/e2fsck.c | 4 ++++
> e2fsck/e2fsck.h | 1 +
> e2fsck/pass1.c | 18 +++++++++++++++++
> e2fsck/pass2.c | 51 ++++++++++++++++++++++++++++++++++++++++++++-----
> 4 files changed, 69 insertions(+), 5 deletions(-)
>
> diff --git a/e2fsck/e2fsck.c b/e2fsck/e2fsck.c
> index d8be566f..dc4b45e2 100644
> --- a/e2fsck/e2fsck.c
> +++ b/e2fsck/e2fsck.c
> @@ -75,6 +75,10 @@ errcode_t e2fsck_reset_context(e2fsck_t ctx)
> ext2fs_free_block_bitmap(ctx->block_found_map);
> ctx->block_found_map = 0;
> }
> + if (ctx->inode_casefold_map) {
> + ext2fs_free_block_bitmap(ctx->inode_casefold_map);
> + ctx->inode_casefold_map = 0;
> + }
> if (ctx->inode_link_info) {
> ext2fs_free_icount(ctx->inode_link_info);
> ctx->inode_link_info = 0;
> diff --git a/e2fsck/e2fsck.h b/e2fsck/e2fsck.h
> index 85f953b2..dcaab0a1 100644
> --- a/e2fsck/e2fsck.h
> +++ b/e2fsck/e2fsck.h
> @@ -262,6 +262,7 @@ struct e2fsck_struct {
> ext2fs_inode_bitmap inode_bb_map; /* Inodes which are in bad blocks */
> ext2fs_inode_bitmap inode_imagic_map; /* AFS inodes */
> ext2fs_inode_bitmap inode_reg_map; /* Inodes which are regular files*/
> + ext2fs_inode_bitmap inode_casefold_map; /* Inodes which are casefolded */
>
> ext2fs_block_bitmap block_found_map; /* Blocks which are in use */
> ext2fs_block_bitmap block_dup_map; /* Blks referenced more than once */
> diff --git a/e2fsck/pass1.c b/e2fsck/pass1.c
> index 8eecd958..6909fed5 100644
> --- a/e2fsck/pass1.c
> +++ b/e2fsck/pass1.c
> @@ -23,6 +23,7 @@
> * - A bitmap of which inodes have bad fields. (inode_bad_map)
> * - A bitmap of which inodes are in bad blocks. (inode_bb_map)
> * - A bitmap of which inodes are imagic inodes. (inode_imagic_map)
> + * - A bitmap of which inodes are casefolded. (inode_casefold_map)
> * - A bitmap of which blocks are in use. (block_found_map)
> * - A bitmap of which blocks are in use by two inodes (block_dup_map)
> * - The data blocks of the directory inodes. (dir_map)
> @@ -1260,6 +1261,20 @@ void e2fsck_pass1(e2fsck_t ctx)
> ctx->flags |= E2F_FLAG_ABORT;
> return;
> }
> + if (casefold_fs) {
> + pctx.errcode =
> + e2fsck_allocate_inode_bitmap(fs,
> + _("inode casefold map"),
> + EXT2FS_BMAP64_RBTREE,
> + "inode_casefold_map",
> + &ctx->inode_casefold_map);
> + if (pctx.errcode) {
> + pctx.num = 1;
> + fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
> + ctx->flags |= E2F_FLAG_ABORT;
> + return;
> + }
> + }
> pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
> &ctx->inode_link_info);
> if (pctx.errcode) {
> @@ -1870,6 +1885,9 @@ void e2fsck_pass1(e2fsck_t ctx)
> add_encrypted_file(ctx, &pctx) < 0)
> goto clear_inode;
>
> + if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
> + ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
> +
> if (LINUX_S_ISDIR(inode->i_mode)) {
> ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
> e2fsck_add_dir_info(ctx, ino, 0);
> diff --git a/e2fsck/pass2.c b/e2fsck/pass2.c
> index 4dbc44ea..b9402b24 100644
> --- a/e2fsck/pass2.c
> +++ b/e2fsck/pass2.c
> @@ -36,11 +36,13 @@
> * - The inode_bad_map bitmap
> * - The inode_dir_map bitmap
> * - The encrypted_file_info
> + * - The inode_casefold_map bitmap
> *
> * Pass 2 frees the following data structures
> * - The inode_bad_map bitmap
> * - The inode_reg_map bitmap
> * - The encrypted_file_info
> + * - The inode_casefold_map bitmap
> */
>
> #define _GNU_SOURCE 1 /* get strnlen() */
> @@ -287,6 +289,10 @@ void e2fsck_pass2(e2fsck_t ctx)
> ext2fs_free_inode_bitmap(ctx->inode_reg_map);
> ctx->inode_reg_map = 0;
> }
> + if (ctx->inode_casefold_map) {
> + ext2fs_free_inode_bitmap(ctx->inode_casefold_map);
> + ctx->inode_casefold_map = 0;
> + }
> destroy_encrypted_file_info(ctx);
>
> clear_problem_context(&pctx);
> @@ -515,6 +521,30 @@ static int encrypted_check_name(e2fsck_t ctx,
> return 0;
> }
>
> +static int encoded_check_name(e2fsck_t ctx,
> + struct ext2_dir_entry *dirent,
> + struct problem_context *pctx)
> +{
> + const struct ext2fs_nls_table *tbl = ctx->fs->encoding;
> + int ret;
> + int len = ext2fs_dirent_name_len(dirent);
> + char *pos, *end;
> +
> + ret = ext2fs_check_encoded_name(tbl, dirent->name, len, &pos);
> + if (ret < 0) {
> + fatal_error(ctx, _("NLS is broken."));
> + } else if(ret > 0) {
> + ret = fix_problem(ctx, PR_2_BAD_CASEFOLDED_NAME, pctx);
> + if (ret) {
> + end = &dirent->name[len];
> + for (; *pos && pos != end; pos++)
> + *pos = '.';
> + }
> + }
> +
> + return (ret || check_name(ctx, dirent, pctx));
> +}
> +
> /*
> * Check the directory filetype (if present)
> */
> @@ -998,11 +1028,18 @@ static int check_dir_block(ext2_filsys fs,
> size_t max_block_size;
> int hash_flags = 0;
> static char *eop_read_dirblock = NULL;
> + int cf_dir = 0;
>
> cd = (struct check_dir_struct *) priv_data;
> ibuf = buf = cd->buf;
> ctx = cd->ctx;
>
> + /* We only want filename encoding verification on strict
> + * mode. */
> + if (ext2fs_test_inode_bitmap2(ctx->inode_casefold_map, ino) &&
> + (ctx->fs->super->s_encoding_flags & EXT4_ENC_STRICT_MODE_FL))
> + cf_dir = 1;
> +
> if (ctx->flags & E2F_FLAG_RUN_RETURN)
> return DIRENT_ABORT;
>
> @@ -1483,11 +1520,7 @@ skip_checksum:
> if (check_filetype(ctx, dirent, ino, &cd->pctx))
> dir_modified++;
>
> - if (dir_encpolicy_id == NO_ENCRYPTION_POLICY) {
> - /* Unencrypted directory */
> - if (check_name(ctx, dirent, &cd->pctx))
> - dir_modified++;
> - } else {
> + if (dir_encpolicy_id != NO_ENCRYPTION_POLICY) {
> /* Encrypted directory */
> if (dot_state > 1 &&
> check_encrypted_dirent(ctx, dirent,
> @@ -1497,6 +1530,14 @@ skip_checksum:
> dir_modified++;
> goto next;
> }
> + } else if (cf_dir) {
> + /* Casefolded directory */
> + if (encoded_check_name(ctx, dirent, &cd->pctx))
> + dir_modified++;
> + } else {
> + /* Unencrypted and uncasefolded directory */
> + if (check_name(ctx, dirent, &cd->pctx))
> + dir_modified++;
> }
This won't do for encrypted+casefolded directories, right?
>
> if (dx_db) {
--
Gabriel Krisman Bertazi
Hi Gabriel,
Le 10/12/2020 à 21:51, Gabriel Krisman Bertazi a écrit :
> Arnaud Ferraris <[email protected]> writes:
>
>> From: Gabriel Krisman Bertazi <[email protected]>
>>
>> @@ -1483,11 +1520,7 @@ skip_checksum:
>> if (check_filetype(ctx, dirent, ino, &cd->pctx))
>> dir_modified++;
>>
>> - if (dir_encpolicy_id == NO_ENCRYPTION_POLICY) {
>> - /* Unencrypted directory */
>> - if (check_name(ctx, dirent, &cd->pctx))
>> - dir_modified++;
>> - } else {
>> + if (dir_encpolicy_id != NO_ENCRYPTION_POLICY) {
>> /* Encrypted directory */
>> if (dot_state > 1 &&
>> check_encrypted_dirent(ctx, dirent,
>> @@ -1497,6 +1530,14 @@ skip_checksum:
>> dir_modified++;
>> goto next;
>> }
>> + } else if (cf_dir) {
>> + /* Casefolded directory */
>> + if (encoded_check_name(ctx, dirent, &cd->pctx))
>> + dir_modified++;
>> + } else {
>> + /* Unencrypted and uncasefolded directory */
>> + if (check_name(ctx, dirent, &cd->pctx))
>> + dir_modified++;
>> }
>
> This won't do for encrypted+casefolded directories, right?
Indeed, as encrypted+casefolded isn't supported right now, it's just a
re-arrangement to ease future support, as suggested by Eric.
Arnaud
>
>>
>> if (dx_db) {
>
On Thu, Dec 10, 2020 at 04:03:44PM +0100, Arnaud Ferraris wrote:
> From: Gabriel Krisman Bertazi <[email protected]>
>
> This is exported to be used by fsck.
>
> Signed-off-by: Gabriel Krisman Bertazi <[email protected]>
> Signed-off-by: Arnaud Ferraris <[email protected]>
Thanks, applied.
- Ted
On Thu, Dec 10, 2020 at 04:03:45PM +0100, Arnaud Ferraris wrote:
> From: Gabriel Krisman Bertazi <[email protected]>
>
> Instead of calling casefold two times and memcmp the result, which
> require allocating a temporary buffer for the casefolded version, add a
> strcasecmp-like method to perform the comparison of each code-point
> during the casefold itself.
>
> This method is exposed because it needs to be used directly by fsck.
>
> Signed-off-by: Gabriel Krisman Bertazi <[email protected]>
> Signed-off-by: Arnaud Ferraris <[email protected]>
Thanks, applied.
- Ted
Oops, I started applying the wrong version. I'm aborting and
restarting with the v3 of this patch series.
- Ted