These patches are all on top of fscrypt's developement branch
Ext4 and F2FS currently both support casefolding and encryption, but not at
the same time. These patches aim to rectify that.
Since directory names are stored case preserved, we cannot just take the hash
of the ciphertext. Instead we use the siphash of the casefolded name. With this
we no longer have a direct path from an encrypted name to the hash without the
key. To deal with this, fscrypt now always includes the hash in the name it
presents when the key is not present. There is a pre-existing bug where you can
change parts of the hash and still match the name so long as the disruption to
the hash does not happen to affect lookup on that filesystem. I'm not sure how
to fix that without making ext4 lookups slower in the more common case.
I moved the identical dcache operations for ext4 and f2fs into the VFS, as any
filesystem that uses casefolding will need the same code. This will also allow
further optimizations to that path, although my current changes don't take
advantage of that yet.
For Ext4, this also means that we need to store the hash on disk. We only do so
for encrypted and casefolded directories to avoid on disk format changes.
Previously encryption and casefolding could not live on the same filesystem,
and we're relaxing that requirement. F2fs is a bit more straightforward since
it already stores hashes on disk.
I've updated the related tools with just enough to enable the feature. I still
need to adjust ext4's fsck's, although without access to the keys,
neither fsck will be able to verify the hashes of casefolded and encrypted names.
v3 changes:
fscrypt patch only creates hash key if it will be needed.
Rebased on top of fscrypt branch, reconstified match functions in ext4/f2fs
v2 changes:
fscrypt moved to separate thread to rebase on fscrypt dev branch
addressed feedback, plus some minor fixes
Daniel Rosenberg (9):
fscrypt: Add siphash and hash key for policy v2
fscrypt: Don't allow v1 policies with casefolding
fscrypt: Change format of no-key token
fscrypt: Only create hash key when needed
vfs: Fold casefolding into vfs
f2fs: Handle casefolding with Encryption
ext4: Use struct super_blocks' casefold data
ext4: Hande casefolding with encryption
ext4: Optimize match for casefolded encrypted dirs
Documentation/filesystems/ext4/directory.rst | 27 ++
fs/crypto/Kconfig | 1 +
fs/crypto/fname.c | 232 ++++++++++---
fs/crypto/fscrypt_private.h | 9 +
fs/crypto/keysetup.c | 35 +-
fs/crypto/policy.c | 53 +++
fs/dcache.c | 28 ++
fs/ext4/dir.c | 75 +----
fs/ext4/ext4.h | 85 +++--
fs/ext4/hash.c | 26 +-
fs/ext4/ialloc.c | 5 +-
fs/ext4/inline.c | 41 +--
fs/ext4/namei.c | 324 ++++++++++++-------
fs/ext4/super.c | 21 +-
fs/f2fs/dir.c | 112 +++----
fs/f2fs/f2fs.h | 12 +-
fs/f2fs/hash.c | 25 +-
fs/f2fs/inline.c | 9 +-
fs/f2fs/super.c | 17 +-
fs/f2fs/sysfs.c | 8 +-
fs/inode.c | 3 +-
fs/namei.c | 41 ++-
include/linux/fs.h | 10 +
include/linux/fscrypt.h | 95 ++----
include/linux/unicode.h | 14 +
25 files changed, 835 insertions(+), 473 deletions(-)
--
2.25.0.341.g760bfbb309-goog
With encryption and casefolding, we cannot simply take the hash of the
ciphertext because of case insensitivity, and we can't take the hash of
the unencrypted name since that would leak information about the
encrypted name. Instead we can use siphash to compute a keyed hash of
the file names.
When a v2 policy is used on a directory, we derive a key for use with
siphash.
Signed-off-by: Daniel Rosenberg <[email protected]>
---
fs/crypto/fname.c | 22 ++++++++++++++++++++++
fs/crypto/fscrypt_private.h | 9 +++++++++
fs/crypto/keysetup.c | 35 +++++++++++++++++++++++++----------
include/linux/fscrypt.h | 9 +++++++++
4 files changed, 65 insertions(+), 10 deletions(-)
diff --git a/fs/crypto/fname.c b/fs/crypto/fname.c
index 3fd27e14ebdd6..371e8f01d1c8e 100644
--- a/fs/crypto/fname.c
+++ b/fs/crypto/fname.c
@@ -402,6 +402,28 @@ int fscrypt_setup_filename(struct inode *dir, const struct qstr *iname,
}
EXPORT_SYMBOL(fscrypt_setup_filename);
+/**
+ * fscrypt_fname_siphash() - Calculate the siphash for a file name
+ * @dir: the parent directory
+ * @name: the name of the file to get the siphash of
+ *
+ * Given a user-provided filename @name, this function calculates the siphash of
+ * that name using the directory's hash key.
+ *
+ * This assumes the directory uses a v2 policy, and the key is available.
+ *
+ * Return: the siphash of @name using the hash key of @dir
+ */
+u64 fscrypt_fname_siphash(const struct inode *dir, const struct qstr *name)
+{
+ struct fscrypt_info *ci = dir->i_crypt_info;
+
+ WARN_ON(!ci->ci_hash_key_initialized);
+
+ return siphash(name->name, name->len, &ci->ci_hash_key);
+}
+EXPORT_SYMBOL(fscrypt_fname_siphash);
+
/*
* Validate dentries in encrypted directories to make sure we aren't potentially
* caching stale dentries after a key has been added.
diff --git a/fs/crypto/fscrypt_private.h b/fs/crypto/fscrypt_private.h
index b22e8decebedd..8b37a5eebb574 100644
--- a/fs/crypto/fscrypt_private.h
+++ b/fs/crypto/fscrypt_private.h
@@ -12,6 +12,7 @@
#define _FSCRYPT_PRIVATE_H
#include <linux/fscrypt.h>
+#include <linux/siphash.h>
#include <crypto/hash.h>
#define CONST_STRLEN(str) (sizeof(str) - 1)
@@ -188,6 +189,13 @@ struct fscrypt_info {
*/
struct fscrypt_direct_key *ci_direct_key;
+ /*
+ * With v2 policies, this can be used with siphash
+ * When the key has been set, ci_hash_key_initialized is set to true
+ */
+ siphash_key_t ci_hash_key;
+ bool ci_hash_key_initialized;
+
/* The encryption policy used by this inode */
union fscrypt_policy ci_policy;
@@ -262,6 +270,7 @@ extern int fscrypt_init_hkdf(struct fscrypt_hkdf *hkdf, const u8 *master_key,
#define HKDF_CONTEXT_PER_FILE_KEY 2
#define HKDF_CONTEXT_DIRECT_KEY 3
#define HKDF_CONTEXT_IV_INO_LBLK_64_KEY 4
+#define HKDF_CONTEXT_FNAME_HASH_KEY 5
extern int fscrypt_hkdf_expand(const struct fscrypt_hkdf *hkdf, u8 context,
const u8 *info, unsigned int infolen,
diff --git a/fs/crypto/keysetup.c b/fs/crypto/keysetup.c
index 96074054bdbc8..7445ab76e0b32 100644
--- a/fs/crypto/keysetup.c
+++ b/fs/crypto/keysetup.c
@@ -189,7 +189,7 @@ static int fscrypt_setup_v2_file_key(struct fscrypt_info *ci,
* This ensures that the master key is consistently used only
* for HKDF, avoiding key reuse issues.
*/
- return setup_per_mode_key(ci, mk, mk->mk_direct_tfms,
+ err = setup_per_mode_key(ci, mk, mk->mk_direct_tfms,
HKDF_CONTEXT_DIRECT_KEY, false);
} else if (ci->ci_policy.v2.flags &
FSCRYPT_POLICY_FLAG_IV_INO_LBLK_64) {
@@ -199,21 +199,36 @@ static int fscrypt_setup_v2_file_key(struct fscrypt_info *ci,
* the IVs. This format is optimized for use with inline
* encryption hardware compliant with the UFS or eMMC standards.
*/
- return setup_per_mode_key(ci, mk, mk->mk_iv_ino_lblk_64_tfms,
+ err = setup_per_mode_key(ci, mk, mk->mk_iv_ino_lblk_64_tfms,
HKDF_CONTEXT_IV_INO_LBLK_64_KEY,
true);
+ } else {
+ err = fscrypt_hkdf_expand(&mk->mk_secret.hkdf,
+ HKDF_CONTEXT_PER_FILE_KEY,
+ ci->ci_nonce,
+ FS_KEY_DERIVATION_NONCE_SIZE,
+ derived_key, ci->ci_mode->keysize);
+ if (err)
+ return err;
+
+ err = fscrypt_set_derived_key(ci, derived_key);
+ memzero_explicit(derived_key, ci->ci_mode->keysize);
}
-
- err = fscrypt_hkdf_expand(&mk->mk_secret.hkdf,
- HKDF_CONTEXT_PER_FILE_KEY,
- ci->ci_nonce, FS_KEY_DERIVATION_NONCE_SIZE,
- derived_key, ci->ci_mode->keysize);
if (err)
return err;
- err = fscrypt_set_derived_key(ci, derived_key);
- memzero_explicit(derived_key, ci->ci_mode->keysize);
- return err;
+ if (S_ISDIR(ci->ci_inode->i_mode)) {
+ err = fscrypt_hkdf_expand(&mk->mk_secret.hkdf,
+ HKDF_CONTEXT_FNAME_HASH_KEY,
+ ci->ci_nonce,
+ FS_KEY_DERIVATION_NONCE_SIZE,
+ (u8 *)&ci->ci_hash_key,
+ sizeof(ci->ci_hash_key));
+ if (err)
+ return err;
+ ci->ci_hash_key_initialized = true;
+ }
+ return 0;
}
/*
diff --git a/include/linux/fscrypt.h b/include/linux/fscrypt.h
index 6fe8d0f96a4ac..1dfbed855beeb 100644
--- a/include/linux/fscrypt.h
+++ b/include/linux/fscrypt.h
@@ -172,6 +172,8 @@ extern int fscrypt_fname_disk_to_usr(const struct inode *inode,
u32 hash, u32 minor_hash,
const struct fscrypt_str *iname,
struct fscrypt_str *oname);
+extern u64 fscrypt_fname_siphash(const struct inode *dir,
+ const struct qstr *name);
#define FSCRYPT_FNAME_MAX_UNDIGESTED_SIZE 32
@@ -468,6 +470,13 @@ static inline int fscrypt_fname_disk_to_usr(const struct inode *inode,
return -EOPNOTSUPP;
}
+static inline u64 fscrypt_fname_siphash(const struct inode *dir,
+ const struct qstr *name)
+{
+ WARN_ON_ONCE(1);
+ return 0;
+}
+
static inline bool fscrypt_match_name(const struct fscrypt_name *fname,
const u8 *de_name, u32 de_name_len)
{
--
2.25.0.341.g760bfbb309-goog
This expands f2fs's casefolding support to include encrypted
directories. For encrypted directories, we use the siphash of the
casefolded name. This ensures there is no direct way to go from an
unencrypted name to the stored hash on disk without knowledge of the
encryption policy keys.
Additionally, we switch to using the vfs layer's casefolding support
instead of storing this information inside of f2fs's private data.
Signed-off-by: Daniel Rosenberg <[email protected]>
---
fs/f2fs/dir.c | 112 ++++++++++++++++++-----------------------------
fs/f2fs/f2fs.h | 12 ++---
fs/f2fs/hash.c | 25 +++++++----
fs/f2fs/inline.c | 9 ++--
fs/f2fs/super.c | 17 +++----
fs/f2fs/sysfs.c | 8 ++--
6 files changed, 78 insertions(+), 105 deletions(-)
diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index d9ad842945df5..e0822c319887c 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -112,30 +112,49 @@ static struct f2fs_dir_entry *find_in_block(struct inode *dir,
* doesn't match or less than zero on error.
*/
int f2fs_ci_compare(const struct inode *parent, const struct qstr *name,
- const struct qstr *entry, bool quick)
+ unsigned char *name2, size_t len, bool quick)
{
const struct f2fs_sb_info *sbi = F2FS_SB(parent->i_sb);
- const struct unicode_map *um = sbi->s_encoding;
+ const struct unicode_map *um = sbi->sb->s_encoding;
+ const struct fscrypt_str crypt_entry = FSTR_INIT(name2, len);
+ struct fscrypt_str decrypted_entry;
+ struct qstr decrypted;
+ struct qstr entry = QSTR_INIT(name2, len);
int ret;
+ decrypted_entry.name = NULL;
+
+ if (IS_ENCRYPTED(parent) && fscrypt_has_encryption_key(parent)) {
+ decrypted_entry.name = kmalloc(len, GFP_ATOMIC);
+ decrypted.name = decrypted_entry.name;
+ decrypted_entry.len = len;
+ decrypted.len = len;
+ if (!decrypted.name)
+ return -ENOMEM;
+ fscrypt_fname_disk_to_usr(parent, 0, 0, &crypt_entry,
+ &decrypted_entry);
+ }
+
if (quick)
- ret = utf8_strncasecmp_folded(um, name, entry);
+ ret = utf8_strncasecmp_folded(um, name, decrypted_entry.name ?
+ &decrypted : &entry);
else
- ret = utf8_strncasecmp(um, name, entry);
-
+ ret = utf8_strncasecmp(um, name, decrypted_entry.name ?
+ &decrypted : &entry);
if (ret < 0) {
/* Handle invalid character sequence as either an error
* or as an opaque byte sequence.
*/
- if (f2fs_has_strict_mode(sbi))
+ if (sb_has_enc_strict_mode(sbi->sb))
return -EINVAL;
- if (name->len != entry->len)
+ if (name->len != len)
return 1;
- return !!memcmp(name->name, entry->name, name->len);
+ ret = !!memcmp(name->name,
+ decrypted_entry.name ?: name2, name->len);
}
-
+ kfree(decrypted_entry.name);
return ret;
}
@@ -154,7 +173,7 @@ static void f2fs_fname_setup_ci_filename(struct inode *dir,
if (!cf_name->name)
return;
- cf_name->len = utf8_casefold(sbi->s_encoding,
+ cf_name->len = utf8_casefold(dir->i_sb->s_encoding,
iname, cf_name->name,
F2FS_NAME_LEN);
if ((int)cf_name->len <= 0) {
@@ -173,24 +192,24 @@ static inline bool f2fs_match_name(struct f2fs_dentry_ptr *d,
{
#ifdef CONFIG_UNICODE
struct inode *parent = d->inode;
- struct f2fs_sb_info *sbi = F2FS_I_SB(parent);
- struct qstr entry;
+ unsigned char *name;
+ int len;
#endif
if (de->hash_code != namehash)
return false;
#ifdef CONFIG_UNICODE
- entry.name = d->filename[bit_pos];
- entry.len = de->name_len;
+ name = d->filename[bit_pos];
+ len = de->name_len;
- if (sbi->s_encoding && IS_CASEFOLDED(parent)) {
+ if (needs_casefold(parent)) {
if (cf_str->name) {
struct qstr cf = {.name = cf_str->name,
.len = cf_str->len};
- return !f2fs_ci_compare(parent, &cf, &entry, true);
+ return !f2fs_ci_compare(parent, &cf, name, len, true);
}
- return !f2fs_ci_compare(parent, fname->usr_fname, &entry,
+ return !f2fs_ci_compare(parent, fname->usr_fname, name, len,
false);
}
#endif
@@ -357,8 +376,8 @@ struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
int err;
#ifdef CONFIG_UNICODE
- if (f2fs_has_strict_mode(F2FS_I_SB(dir)) && IS_CASEFOLDED(dir) &&
- utf8_validate(F2FS_I_SB(dir)->s_encoding, child)) {
+ if (sb_has_enc_strict_mode(dir->i_sb) && IS_CASEFOLDED(dir) &&
+ utf8_validate(dir->i_sb->s_encoding, child)) {
*res_page = ERR_PTR(-EINVAL);
return NULL;
}
@@ -602,13 +621,13 @@ void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
int f2fs_add_regular_entry(struct inode *dir, const struct qstr *new_name,
const struct qstr *orig_name,
+ f2fs_hash_t dentry_hash,
struct inode *inode, nid_t ino, umode_t mode)
{
unsigned int bit_pos;
unsigned int level;
unsigned int current_depth;
unsigned long bidx, block;
- f2fs_hash_t dentry_hash;
unsigned int nbucket, nblock;
struct page *dentry_page = NULL;
struct f2fs_dentry_block *dentry_blk = NULL;
@@ -618,7 +637,6 @@ int f2fs_add_regular_entry(struct inode *dir, const struct qstr *new_name,
level = 0;
slots = GET_DENTRY_SLOTS(new_name->len);
- dentry_hash = f2fs_dentry_hash(dir, new_name, NULL);
current_depth = F2FS_I(dir)->i_current_depth;
if (F2FS_I(dir)->chash == dentry_hash) {
@@ -704,17 +722,19 @@ int f2fs_add_dentry(struct inode *dir, struct fscrypt_name *fname,
struct inode *inode, nid_t ino, umode_t mode)
{
struct qstr new_name;
+ f2fs_hash_t dentry_hash;
int err = -EAGAIN;
new_name.name = fname_name(fname);
new_name.len = fname_len(fname);
if (f2fs_has_inline_dentry(dir))
- err = f2fs_add_inline_entry(dir, &new_name, fname->usr_fname,
+ err = f2fs_add_inline_entry(dir, &new_name, fname,
inode, ino, mode);
+ dentry_hash = f2fs_dentry_hash(dir, &new_name, fname);
if (err == -EAGAIN)
err = f2fs_add_regular_entry(dir, &new_name, fname->usr_fname,
- inode, ino, mode);
+ dentry_hash, inode, ino, mode);
f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
return err;
@@ -1064,49 +1084,3 @@ const struct file_operations f2fs_dir_operations = {
#endif
};
-#ifdef CONFIG_UNICODE
-static int f2fs_d_compare(const struct dentry *dentry, unsigned int len,
- const char *str, const struct qstr *name)
-{
- struct qstr qstr = {.name = str, .len = len };
-
- if (!IS_CASEFOLDED(dentry->d_parent->d_inode)) {
- if (len != name->len)
- return -1;
- return memcmp(str, name, len);
- }
-
- return f2fs_ci_compare(dentry->d_parent->d_inode, name, &qstr, false);
-}
-
-static int f2fs_d_hash(const struct dentry *dentry, struct qstr *str)
-{
- struct f2fs_sb_info *sbi = F2FS_SB(dentry->d_sb);
- const struct unicode_map *um = sbi->s_encoding;
- unsigned char *norm;
- int len, ret = 0;
-
- if (!IS_CASEFOLDED(dentry->d_inode))
- return 0;
-
- norm = f2fs_kmalloc(sbi, PATH_MAX, GFP_ATOMIC);
- if (!norm)
- return -ENOMEM;
-
- len = utf8_casefold(um, str, norm, PATH_MAX);
- if (len < 0) {
- if (f2fs_has_strict_mode(sbi))
- ret = -EINVAL;
- goto out;
- }
- str->hash = full_name_hash(dentry, norm, len);
-out:
- kvfree(norm);
- return ret;
-}
-
-const struct dentry_operations f2fs_dentry_ops = {
- .d_hash = f2fs_d_hash,
- .d_compare = f2fs_d_compare,
-};
-#endif
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 5a888a063c7f1..9387d0b894264 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1184,10 +1184,6 @@ struct f2fs_sb_info {
int valid_super_block; /* valid super block no */
unsigned long s_flag; /* flags for sbi */
struct mutex writepages; /* mutex for writepages() */
-#ifdef CONFIG_UNICODE
- struct unicode_map *s_encoding;
- __u16 s_encoding_flags;
-#endif
#ifdef CONFIG_BLK_DEV_ZONED
unsigned int blocks_per_blkz; /* F2FS blocks per zone */
@@ -2971,7 +2967,7 @@ struct dentry *f2fs_get_parent(struct dentry *child);
extern int f2fs_ci_compare(const struct inode *parent,
const struct qstr *name,
- const struct qstr *entry,
+ unsigned char *name2, size_t len,
bool quick);
/*
@@ -3005,7 +3001,7 @@ void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
const struct qstr *name, f2fs_hash_t name_hash,
unsigned int bit_pos);
int f2fs_add_regular_entry(struct inode *dir, const struct qstr *new_name,
- const struct qstr *orig_name,
+ const struct qstr *orig_name, f2fs_hash_t dentry_hash,
struct inode *inode, nid_t ino, umode_t mode);
int f2fs_add_dentry(struct inode *dir, struct fscrypt_name *fname,
struct inode *inode, nid_t ino, umode_t mode);
@@ -3038,7 +3034,7 @@ int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi);
* hash.c
*/
f2fs_hash_t f2fs_dentry_hash(const struct inode *dir,
- const struct qstr *name_info, struct fscrypt_name *fname);
+ const struct qstr *name_info, const struct fscrypt_name *fname);
/*
* node.c
@@ -3517,7 +3513,7 @@ struct f2fs_dir_entry *f2fs_find_in_inline_dir(struct inode *dir,
int f2fs_make_empty_inline_dir(struct inode *inode, struct inode *parent,
struct page *ipage);
int f2fs_add_inline_entry(struct inode *dir, const struct qstr *new_name,
- const struct qstr *orig_name,
+ const struct fscrypt_name *fname,
struct inode *inode, nid_t ino, umode_t mode);
void f2fs_delete_inline_entry(struct f2fs_dir_entry *dentry,
struct page *page, struct inode *dir,
diff --git a/fs/f2fs/hash.c b/fs/f2fs/hash.c
index 5bc4dcd8fc03f..954d03dee4506 100644
--- a/fs/f2fs/hash.c
+++ b/fs/f2fs/hash.c
@@ -68,8 +68,9 @@ static void str2hashbuf(const unsigned char *msg, size_t len,
*buf++ = pad;
}
-static f2fs_hash_t __f2fs_dentry_hash(const struct qstr *name_info,
- struct fscrypt_name *fname)
+static f2fs_hash_t __f2fs_dentry_hash(const struct inode *dir,
+ const struct qstr *name_info,
+ const struct fscrypt_name *fname)
{
__u32 hash;
f2fs_hash_t f2fs_hash;
@@ -85,6 +86,11 @@ static f2fs_hash_t __f2fs_dentry_hash(const struct qstr *name_info,
if (is_dot_dotdot(name_info))
return 0;
+ if (IS_CASEFOLDED(dir) && IS_ENCRYPTED(dir)) {
+ f2fs_hash = fscrypt_fname_siphash(dir, name_info);
+ return f2fs_hash;
+ }
+
/* Initialize the default seed for the hash checksum functions */
buf[0] = 0x67452301;
buf[1] = 0xefcdab89;
@@ -106,35 +112,38 @@ static f2fs_hash_t __f2fs_dentry_hash(const struct qstr *name_info,
}
f2fs_hash_t f2fs_dentry_hash(const struct inode *dir,
- const struct qstr *name_info, struct fscrypt_name *fname)
+ const struct qstr *name_info, const struct fscrypt_name *fname)
{
#ifdef CONFIG_UNICODE
struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
- const struct unicode_map *um = sbi->s_encoding;
+ const struct unicode_map *um = sbi->sb->s_encoding;
int r, dlen;
unsigned char *buff;
struct qstr folded;
+ const struct qstr *name = fname ? fname->usr_fname : name_info;
if (!name_info->len || !IS_CASEFOLDED(dir))
goto opaque_seq;
+ if (IS_ENCRYPTED(dir) && !fscrypt_has_encryption_key(dir))
+ goto opaque_seq;
+
buff = f2fs_kzalloc(sbi, sizeof(char) * PATH_MAX, GFP_KERNEL);
if (!buff)
return -ENOMEM;
-
- dlen = utf8_casefold(um, name_info, buff, PATH_MAX);
+ dlen = utf8_casefold(um, name, buff, PATH_MAX);
if (dlen < 0) {
kvfree(buff);
goto opaque_seq;
}
folded.name = buff;
folded.len = dlen;
- r = __f2fs_dentry_hash(&folded, fname);
+ r = __f2fs_dentry_hash(dir, &folded, fname);
kvfree(buff);
return r;
opaque_seq:
#endif
- return __f2fs_dentry_hash(name_info, fname);
+ return __f2fs_dentry_hash(dir, name_info, fname);
}
diff --git a/fs/f2fs/inline.c b/fs/f2fs/inline.c
index 896db0416f0e6..3c37720941539 100644
--- a/fs/f2fs/inline.c
+++ b/fs/f2fs/inline.c
@@ -465,8 +465,8 @@ static int f2fs_add_inline_entries(struct inode *dir, void *inline_dentry)
ino = le32_to_cpu(de->ino);
fake_mode = f2fs_get_de_type(de) << S_SHIFT;
- err = f2fs_add_regular_entry(dir, &new_name, NULL, NULL,
- ino, fake_mode);
+ err = f2fs_add_regular_entry(dir, &new_name, NULL,
+ de->hash_code, NULL, ino, fake_mode);
if (err)
goto punch_dentry_pages;
@@ -540,7 +540,7 @@ static int f2fs_convert_inline_dir(struct inode *dir, struct page *ipage,
}
int f2fs_add_inline_entry(struct inode *dir, const struct qstr *new_name,
- const struct qstr *orig_name,
+ const struct fscrypt_name *fname,
struct inode *inode, nid_t ino, umode_t mode)
{
struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
@@ -551,6 +551,7 @@ int f2fs_add_inline_entry(struct inode *dir, const struct qstr *new_name,
struct f2fs_dentry_ptr d;
int slots = GET_DENTRY_SLOTS(new_name->len);
struct page *page = NULL;
+ const struct qstr *orig_name = fname->usr_fname;
int err = 0;
ipage = f2fs_get_node_page(sbi, dir->i_ino);
@@ -581,7 +582,7 @@ int f2fs_add_inline_entry(struct inode *dir, const struct qstr *new_name,
f2fs_wait_on_page_writeback(ipage, NODE, true, true);
- name_hash = f2fs_dentry_hash(dir, new_name, NULL);
+ name_hash = f2fs_dentry_hash(dir, new_name, fname);
f2fs_update_dentry(ino, mode, &d, new_name, name_hash, bit_pos);
set_page_dirty(ipage);
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 5111e1ffe58ab..5e4e76332c4ca 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1144,7 +1144,7 @@ static void f2fs_put_super(struct super_block *sb)
for (i = 0; i < NR_PAGE_TYPE; i++)
kvfree(sbi->write_io[i]);
#ifdef CONFIG_UNICODE
- utf8_unload(sbi->s_encoding);
+ utf8_unload(sbi->sb->s_encoding);
#endif
kvfree(sbi);
}
@@ -3136,17 +3136,11 @@ static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
{
#ifdef CONFIG_UNICODE
- if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
+ if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
const struct f2fs_sb_encodings *encoding_info;
struct unicode_map *encoding;
__u16 encoding_flags;
- if (f2fs_sb_has_encrypt(sbi)) {
- f2fs_err(sbi,
- "Can't mount with encoding and encryption");
- return -EINVAL;
- }
-
if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
&encoding_flags)) {
f2fs_err(sbi,
@@ -3167,9 +3161,8 @@ static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
"%s-%s with flags 0x%hx", encoding_info->name,
encoding_info->version?:"\b", encoding_flags);
- sbi->s_encoding = encoding;
- sbi->s_encoding_flags = encoding_flags;
- sbi->sb->s_d_op = &f2fs_dentry_ops;
+ sbi->sb->s_encoding = encoding;
+ sbi->sb->s_encoding_flags = encoding_flags;
}
#else
if (f2fs_sb_has_casefold(sbi)) {
@@ -3637,7 +3630,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
kvfree(sbi->write_io[i]);
#ifdef CONFIG_UNICODE
- utf8_unload(sbi->s_encoding);
+ utf8_unload(sbi->sb->s_encoding);
#endif
free_options:
#ifdef CONFIG_QUOTA
diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c
index 70945ceb9c0ca..7fd37c8c9733a 100644
--- a/fs/f2fs/sysfs.c
+++ b/fs/f2fs/sysfs.c
@@ -88,10 +88,10 @@ static ssize_t encoding_show(struct f2fs_attr *a,
#ifdef CONFIG_UNICODE
if (f2fs_sb_has_casefold(sbi))
return snprintf(buf, PAGE_SIZE, "%s (%d.%d.%d)\n",
- sbi->s_encoding->charset,
- (sbi->s_encoding->version >> 16) & 0xff,
- (sbi->s_encoding->version >> 8) & 0xff,
- sbi->s_encoding->version & 0xff);
+ sbi->sb->s_encoding->charset,
+ (sbi->sb->s_encoding->version >> 16) & 0xff,
+ (sbi->sb->s_encoding->version >> 8) & 0xff,
+ sbi->sb->s_encoding->version & 0xff);
#endif
return snprintf(buf, PAGE_SIZE, "(none)");
}
--
2.25.0.341.g760bfbb309-goog
Matching names with casefolded encrypting directories requires
decrypting entries to confirm case since we are case preserving. We can
avoid needing to decrypt if our hash values don't match.
Signed-off-by: Daniel Rosenberg <[email protected]>
---
fs/ext4/ext4.h | 17 ++++++++-------
fs/ext4/namei.c | 55 ++++++++++++++++++++++++++-----------------------
2 files changed, 38 insertions(+), 34 deletions(-)
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index a70b4db05b745..6755eb30a89b7 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -2393,9 +2393,9 @@ extern unsigned ext4_free_clusters_after_init(struct super_block *sb,
ext4_fsblk_t ext4_inode_to_goal_block(struct inode *);
#ifdef CONFIG_UNICODE
-extern void ext4_fname_setup_ci_filename(struct inode *dir,
+extern int ext4_fname_setup_ci_filename(struct inode *dir,
const struct qstr *iname,
- struct fscrypt_str *fname);
+ struct ext4_filename *fname);
#endif
#ifdef CONFIG_FS_ENCRYPTION
@@ -2426,9 +2426,9 @@ static inline int ext4_fname_setup_filename(struct inode *dir,
ext4_fname_from_fscrypt_name(fname, &name);
#ifdef CONFIG_UNICODE
- ext4_fname_setup_ci_filename(dir, iname, &fname->cf_name);
+ err = ext4_fname_setup_ci_filename(dir, iname, fname);
#endif
- return 0;
+ return err;
}
static inline int ext4_fname_prepare_lookup(struct inode *dir,
@@ -2445,9 +2445,9 @@ static inline int ext4_fname_prepare_lookup(struct inode *dir,
ext4_fname_from_fscrypt_name(fname, &name);
#ifdef CONFIG_UNICODE
- ext4_fname_setup_ci_filename(dir, &dentry->d_name, &fname->cf_name);
+ err = ext4_fname_setup_ci_filename(dir, &dentry->d_name, fname);
#endif
- return 0;
+ return err;
}
static inline void ext4_fname_free_filename(struct ext4_filename *fname)
@@ -2472,15 +2472,16 @@ static inline int ext4_fname_setup_filename(struct inode *dir,
int lookup,
struct ext4_filename *fname)
{
+ int err = 0;
fname->usr_fname = iname;
fname->disk_name.name = (unsigned char *) iname->name;
fname->disk_name.len = iname->len;
#ifdef CONFIG_UNICODE
- ext4_fname_setup_ci_filename(dir, iname, &fname->cf_name);
+ err = ext4_fname_setup_ci_filename(dir, iname, fname);
#endif
- return 0;
+ return err;
}
static inline int ext4_fname_prepare_lookup(struct inode *dir,
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index a5ee76a14e3b7..6b63271978a0b 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -784,7 +784,9 @@ dx_probe(struct ext4_filename *fname, struct inode *dir,
if (hinfo->hash_version <= DX_HASH_TEA)
hinfo->hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned;
hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed;
- if (fname && fname_name(fname))
+ /* hash is already computed for encrypted casefolded directory */
+ if (fname && fname_name(fname) &&
+ !(IS_ENCRYPTED(dir) && IS_CASEFOLDED(dir)))
ext4fs_dirhash(dir, fname_name(fname), fname_len(fname), hinfo);
hash = hinfo->hash;
@@ -1352,19 +1354,21 @@ int ext4_ci_compare(const struct inode *parent, const struct qstr *name,
return ret;
}
-void ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
- struct fscrypt_str *cf_name)
+int ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
+ struct ext4_filename *name)
{
+ struct fscrypt_str *cf_name = &name->cf_name;
+ struct dx_hash_info *hinfo = &name->hinfo;
int len;
if (!needs_casefold(dir)) {
cf_name->name = NULL;
- return;
+ return 0;
}
cf_name->name = kmalloc(EXT4_NAME_LEN, GFP_NOFS);
if (!cf_name->name)
- return;
+ return -ENOMEM;
len = utf8_casefold(dir->i_sb->s_encoding,
iname, cf_name->name,
@@ -1372,10 +1376,18 @@ void ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
if (len <= 0) {
kfree(cf_name->name);
cf_name->name = NULL;
- return;
}
cf_name->len = (unsigned) len;
+ if (!IS_ENCRYPTED(dir))
+ return 0;
+ hinfo->hash_version = DX_HASH_SIPHASH;
+ hinfo->seed = NULL;
+ if (cf_name->name)
+ ext4fs_dirhash(dir, cf_name->name, cf_name->len, hinfo);
+ else
+ ext4fs_dirhash(dir, iname->name, iname->len, hinfo);
+ return 0;
}
#endif
@@ -1405,16 +1417,12 @@ static bool ext4_match(struct inode *parent,
struct qstr cf = {.name = fname->cf_name.name,
.len = fname->cf_name.len};
if (IS_ENCRYPTED(parent)) {
- struct dx_hash_info hinfo;
-
- hinfo.hash_version = DX_HASH_SIPHASH;
- hinfo.seed = NULL;
- ext4fs_dirhash(parent, fname->cf_name.name,
- fname_len(fname), &hinfo);
- if (hinfo.hash != EXT4_DIRENT_HASH(de) ||
- hinfo.minor_hash !=
- EXT4_DIRENT_MINOR_HASH(de))
+ if (fname->hinfo.hash != EXT4_DIRENT_HASH(de) ||
+ fname->hinfo.minor_hash !=
+ EXT4_DIRENT_MINOR_HASH(de)) {
+
return 0;
+ }
}
return !ext4_ci_compare(parent, &cf, de->name,
de->name_len, true);
@@ -2036,15 +2044,11 @@ void ext4_insert_dentry(struct inode *dir,
de->name_len = fname_len(fname);
memcpy(de->name, fname_name(fname), fname_len(fname));
if (ext4_hash_in_dirent(dir)) {
- struct dx_hash_info hinfo;
+ struct dx_hash_info *hinfo = &fname->hinfo;
- hinfo.hash_version = DX_HASH_SIPHASH;
- hinfo.seed = NULL;
- ext4fs_dirhash(dir, fname_usr_name(fname),
- fname_len(fname), &hinfo);
- EXT4_EXTENDED_DIRENT(de)->hash = cpu_to_le32(hinfo.hash);
+ EXT4_EXTENDED_DIRENT(de)->hash = cpu_to_le32(hinfo->hash);
EXT4_EXTENDED_DIRENT(de)->minor_hash =
- cpu_to_le32(hinfo.minor_hash);
+ cpu_to_le32(hinfo->minor_hash);
}
}
@@ -2195,10 +2199,9 @@ static int make_indexed_dir(handle_t *handle, struct ext4_filename *fname,
if (fname->hinfo.hash_version <= DX_HASH_TEA)
fname->hinfo.hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned;
fname->hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed;
- if (ext4_hash_in_dirent(dir))
- ext4fs_dirhash(dir, fname_usr_name(fname),
- fname_len(fname), &fname->hinfo);
- else
+
+ /* casefolded encrypted hashes are computed on fname setup */
+ if (!ext4_hash_in_dirent(dir))
ext4fs_dirhash(dir, fname_name(fname),
fname_len(fname), &fname->hinfo);
--
2.25.0.341.g760bfbb309-goog
Switch over to using the struct entries added to the VFS
Signed-off-by: Daniel Rosenberg <[email protected]>
---
fs/ext4/dir.c | 47 -----------------------------------------------
fs/ext4/ext4.h | 4 ----
fs/ext4/hash.c | 2 +-
fs/ext4/namei.c | 20 ++++++++------------
fs/ext4/super.c | 15 +++++----------
5 files changed, 14 insertions(+), 74 deletions(-)
diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c
index 4e093277c8bfb..e0a9b3f0682dd 100644
--- a/fs/ext4/dir.c
+++ b/fs/ext4/dir.c
@@ -667,50 +667,3 @@ const struct file_operations ext4_dir_operations = {
.release = ext4_release_dir,
};
-#ifdef CONFIG_UNICODE
-static int ext4_d_compare(const struct dentry *dentry, unsigned int len,
- const char *str, const struct qstr *name)
-{
- struct qstr qstr = {.name = str, .len = len };
- struct inode *inode = dentry->d_parent->d_inode;
-
- if (!IS_CASEFOLDED(inode) || !EXT4_SB(inode->i_sb)->s_encoding) {
- if (len != name->len)
- return -1;
- return memcmp(str, name->name, len);
- }
-
- return ext4_ci_compare(inode, name, &qstr, false);
-}
-
-static int ext4_d_hash(const struct dentry *dentry, struct qstr *str)
-{
- const struct ext4_sb_info *sbi = EXT4_SB(dentry->d_sb);
- const struct unicode_map *um = sbi->s_encoding;
- unsigned char *norm;
- int len, ret = 0;
-
- if (!IS_CASEFOLDED(dentry->d_inode) || !um)
- return 0;
-
- norm = kmalloc(PATH_MAX, GFP_ATOMIC);
- if (!norm)
- return -ENOMEM;
-
- len = utf8_casefold(um, str, norm, PATH_MAX);
- if (len < 0) {
- if (ext4_has_strict_mode(sbi))
- ret = -EINVAL;
- goto out;
- }
- str->hash = full_name_hash(dentry, norm, len);
-out:
- kfree(norm);
- return ret;
-}
-
-const struct dentry_operations ext4_dentry_ops = {
- .d_hash = ext4_d_hash,
- .d_compare = ext4_d_compare,
-};
-#endif
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index f8578caba40d5..3162ef2e53d46 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1430,10 +1430,6 @@ struct ext4_sb_info {
struct kobject s_kobj;
struct completion s_kobj_unregister;
struct super_block *s_sb;
-#ifdef CONFIG_UNICODE
- struct unicode_map *s_encoding;
- __u16 s_encoding_flags;
-#endif
/* Journaling */
struct journal_s *s_journal;
diff --git a/fs/ext4/hash.c b/fs/ext4/hash.c
index 3e133793a5a34..143b0073b3f46 100644
--- a/fs/ext4/hash.c
+++ b/fs/ext4/hash.c
@@ -275,7 +275,7 @@ int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
struct dx_hash_info *hinfo)
{
#ifdef CONFIG_UNICODE
- const struct unicode_map *um = EXT4_SB(dir->i_sb)->s_encoding;
+ const struct unicode_map *um = dir->i_sb->s_encoding;
int r, dlen;
unsigned char *buff;
struct qstr qstr = {.name = name, .len = len };
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 1cb42d9407847..7f4e625ab2f9b 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -1282,8 +1282,8 @@ static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block)
int ext4_ci_compare(const struct inode *parent, const struct qstr *name,
const struct qstr *entry, bool quick)
{
- const struct ext4_sb_info *sbi = EXT4_SB(parent->i_sb);
- const struct unicode_map *um = sbi->s_encoding;
+ const struct super_block *sb = parent->i_sb;
+ const struct unicode_map *um = sb->s_encoding;
int ret;
if (quick)
@@ -1295,7 +1295,7 @@ int ext4_ci_compare(const struct inode *parent, const struct qstr *name,
/* Handle invalid character sequence as either an error
* or as an opaque byte sequence.
*/
- if (ext4_has_strict_mode(sbi))
+ if (sb_has_enc_strict_mode(sb))
return -EINVAL;
if (name->len != entry->len)
@@ -1312,7 +1312,7 @@ void ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
{
int len;
- if (!IS_CASEFOLDED(dir) || !EXT4_SB(dir->i_sb)->s_encoding) {
+ if (!needs_casefold(dir)) {
cf_name->name = NULL;
return;
}
@@ -1321,7 +1321,7 @@ void ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
if (!cf_name->name)
return;
- len = utf8_casefold(EXT4_SB(dir->i_sb)->s_encoding,
+ len = utf8_casefold(dir->i_sb->s_encoding,
iname, cf_name->name,
EXT4_NAME_LEN);
if (len <= 0) {
@@ -1358,7 +1358,7 @@ static inline bool ext4_match(const struct inode *parent,
#endif
#ifdef CONFIG_UNICODE
- if (EXT4_SB(parent->i_sb)->s_encoding && IS_CASEFOLDED(parent)) {
+ if (needs_casefold(parent)) {
if (fname->cf_name.name) {
struct qstr cf = {.name = fname->cf_name.name,
.len = fname->cf_name.len};
@@ -2164,9 +2164,6 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
struct buffer_head *bh = NULL;
struct ext4_dir_entry_2 *de;
struct super_block *sb;
-#ifdef CONFIG_UNICODE
- struct ext4_sb_info *sbi;
-#endif
struct ext4_filename fname;
int retval;
int dx_fallback=0;
@@ -2183,9 +2180,8 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
return -EINVAL;
#ifdef CONFIG_UNICODE
- sbi = EXT4_SB(sb);
- if (ext4_has_strict_mode(sbi) && IS_CASEFOLDED(dir) &&
- sbi->s_encoding && utf8_validate(sbi->s_encoding, &dentry->d_name))
+ if (sb_has_enc_strict_mode(sb) && IS_CASEFOLDED(dir) &&
+ sb->s_encoding && utf8_validate(sb->s_encoding, &dentry->d_name))
return -EINVAL;
#endif
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 2937a8873fe13..11584bdc3e237 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -1056,7 +1056,7 @@ static void ext4_put_super(struct super_block *sb)
kfree(sbi->s_blockgroup_lock);
fs_put_dax(sbi->s_daxdev);
#ifdef CONFIG_UNICODE
- utf8_unload(sbi->s_encoding);
+ utf8_unload(sb->s_encoding);
#endif
kfree(sbi);
}
@@ -3850,7 +3850,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
goto failed_mount;
#ifdef CONFIG_UNICODE
- if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
+ if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
const struct ext4_sb_encodings *encoding_info;
struct unicode_map *encoding;
__u16 encoding_flags;
@@ -3881,8 +3881,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
"%s-%s with flags 0x%hx", encoding_info->name,
encoding_info->version?:"\b", encoding_flags);
- sbi->s_encoding = encoding;
- sbi->s_encoding_flags = encoding_flags;
+ sb->s_encoding = encoding;
+ sb->s_encoding_flags = encoding_flags;
}
#endif
@@ -4497,11 +4497,6 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
goto failed_mount4;
}
-#ifdef CONFIG_UNICODE
- if (sbi->s_encoding)
- sb->s_d_op = &ext4_dentry_ops;
-#endif
-
sb->s_root = d_make_root(root);
if (!sb->s_root) {
ext4_msg(sb, KERN_ERR, "get root dentry failed");
@@ -4684,7 +4679,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
crypto_free_shash(sbi->s_chksum_driver);
#ifdef CONFIG_UNICODE
- utf8_unload(sbi->s_encoding);
+ utf8_unload(sb->s_encoding);
#endif
#ifdef CONFIG_QUOTA
--
2.25.0.341.g760bfbb309-goog
Ext4 and F2fs are both using casefolding, and they, along with any other
filesystem that adds the feature, will be using identical dentry_ops.
Additionally, those dentry ops interfere with the dentry_ops required
for fscrypt once we add support for casefolding and encryption.
Moving this into the vfs removes code duplication as well as the
complication with encryption.
Currently this is pretty close to just moving the existing f2fs/ext4
code up a level into the vfs, although there is a lot of room for
improvement now.
Signed-off-by: Daniel Rosenberg <[email protected]>
---
fs/dcache.c | 28 ++++++++++++++++++++++++++++
fs/namei.c | 41 ++++++++++++++++++++++++++++++++++++++---
include/linux/fs.h | 10 ++++++++++
include/linux/unicode.h | 14 ++++++++++++++
4 files changed, 90 insertions(+), 3 deletions(-)
diff --git a/fs/dcache.c b/fs/dcache.c
index b280e07e162b1..a8bbb7f4fad30 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -32,6 +32,7 @@
#include <linux/bit_spinlock.h>
#include <linux/rculist_bl.h>
#include <linux/list_lru.h>
+#include <linux/unicode.h>
#include "internal.h"
#include "mount.h"
@@ -247,7 +248,19 @@ static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *c
* be no NUL in the ct/tcount data)
*/
const unsigned char *cs = READ_ONCE(dentry->d_name.name);
+#ifdef CONFIG_UNICODE
+ struct inode *parent = dentry->d_parent->d_inode;
+ if (unlikely(needs_casefold(parent))) {
+ const struct qstr n1 = QSTR_INIT(cs, tcount);
+ const struct qstr n2 = QSTR_INIT(ct, tcount);
+ int result = utf8_strncasecmp(dentry->d_sb->s_encoding,
+ &n1, &n2);
+
+ if (result >= 0 || sb_has_enc_strict_mode(dentry->d_sb))
+ return result;
+ }
+#endif
return dentry_string_cmp(cs, ct, tcount);
}
@@ -2406,7 +2419,22 @@ struct dentry *d_hash_and_lookup(struct dentry *dir, struct qstr *name)
* calculate the standard hash first, as the d_op->d_hash()
* routine may choose to leave the hash value unchanged.
*/
+#ifdef CONFIG_UNICODE
+ unsigned char *hname = NULL;
+ int hlen = name->len;
+
+ if (IS_CASEFOLDED(dir->d_inode)) {
+ hname = kmalloc(PATH_MAX, GFP_ATOMIC);
+ if (!hname)
+ return ERR_PTR(-ENOMEM);
+ hlen = utf8_casefold(dir->d_sb->s_encoding,
+ name, hname, PATH_MAX);
+ }
+ name->hash = full_name_hash(dir, hname ?: name->name, hlen);
+ kfree(hname);
+#else
name->hash = full_name_hash(dir, name->name, name->len);
+#endif
if (dir->d_flags & DCACHE_OP_HASH) {
int err = dir->d_op->d_hash(dir, name);
if (unlikely(err < 0))
diff --git a/fs/namei.c b/fs/namei.c
index d6c91d1e88cb3..f8e65c9f31444 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -39,6 +39,7 @@
#include <linux/bitops.h>
#include <linux/init_task.h>
#include <linux/uaccess.h>
+#include <linux/unicode.h>
#include "internal.h"
#include "mount.h"
@@ -2055,6 +2056,10 @@ static inline u64 hash_name(const void *salt, const char *name)
static int link_path_walk(const char *name, struct nameidata *nd)
{
int err;
+#ifdef CONFIG_UNICODE
+ char *hname = NULL;
+ int hlen = 0;
+#endif
if (IS_ERR(name))
return PTR_ERR(name);
@@ -2071,9 +2076,22 @@ static int link_path_walk(const char *name, struct nameidata *nd)
err = may_lookup(nd);
if (err)
return err;
-
+#ifdef CONFIG_UNICODE
+ if (needs_casefold(nd->path.dentry->d_inode)) {
+ struct qstr str = QSTR_INIT(name, PATH_MAX);
+
+ hname = kmalloc(PATH_MAX, GFP_ATOMIC);
+ if (!hname)
+ return -ENOMEM;
+ hlen = utf8_casefold(nd->path.dentry->d_sb->s_encoding,
+ &str, hname, PATH_MAX);
+ }
+ hash_len = hash_name(nd->path.dentry, hname ?: name);
+ kfree(hname);
+ hname = NULL;
+#else
hash_len = hash_name(nd->path.dentry, name);
-
+#endif
type = LAST_NORM;
if (name[0] == '.') switch (hashlen_len(hash_len)) {
case 2:
@@ -2445,9 +2463,26 @@ EXPORT_SYMBOL(vfs_path_lookup);
static int lookup_one_len_common(const char *name, struct dentry *base,
int len, struct qstr *this)
{
+#ifdef CONFIG_UNICODE
+ char *hname = NULL;
+ int hlen = len;
+
+ if (needs_casefold(base->d_inode)) {
+ struct qstr str = QSTR_INIT(name, len);
+
+ hname = kmalloc(PATH_MAX, GFP_ATOMIC);
+ if (!hname)
+ return -ENOMEM;
+ hlen = utf8_casefold(base->d_sb->s_encoding,
+ &str, hname, PATH_MAX);
+ }
+ this->hash = full_name_hash(base, hname ?: name, hlen);
+ kfree(hname);
+#else
+ this->hash = full_name_hash(base, name, len);
+#endif
this->name = name;
this->len = len;
- this->hash = full_name_hash(base, name, len);
if (!len)
return -EACCES;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 98e0349adb526..9a7092449e94f 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1382,6 +1382,12 @@ extern int send_sigurg(struct fown_struct *fown);
#define SB_ACTIVE (1<<30)
#define SB_NOUSER (1<<31)
+/* These flags relate to encoding and casefolding */
+#define SB_ENC_STRICT_MODE_FL (1 << 0)
+
+#define sb_has_enc_strict_mode(sb) \
+ (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
+
/*
* Umount options
*/
@@ -1449,6 +1455,10 @@ struct super_block {
#endif
#ifdef CONFIG_FS_VERITY
const struct fsverity_operations *s_vop;
+#endif
+#ifdef CONFIG_UNICODE
+ struct unicode_map *s_encoding;
+ __u16 s_encoding_flags;
#endif
struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
struct list_head s_mounts; /* list of mounts; _not_ for fs use */
diff --git a/include/linux/unicode.h b/include/linux/unicode.h
index 990aa97d80496..182352f3cc30f 100644
--- a/include/linux/unicode.h
+++ b/include/linux/unicode.h
@@ -4,6 +4,8 @@
#include <linux/init.h>
#include <linux/dcache.h>
+#include <linux/fscrypt.h>
+#include <linux/fs.h>
struct unicode_map {
const char *charset;
@@ -30,4 +32,16 @@ int utf8_casefold(const struct unicode_map *um, const struct qstr *str,
struct unicode_map *utf8_load(const char *version);
void utf8_unload(struct unicode_map *um);
+#ifdef CONFIG_UNICODE
+static inline bool needs_casefold(const struct inode *dir)
+{
+ return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding &&
+ (!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir));
+}
+#else
+static inline bool needs_casefold(const struct inode *dir)
+{
+ return 0;
+}
+#endif
#endif /* _LINUX_UNICODE_H */
--
2.25.0.341.g760bfbb309-goog
If a directory isn't casefolded, it doesn't need the hash key. Skip
deriving it unless we enable it later.
Signed-off-by: Daniel Rosenberg <[email protected]>
---
fs/crypto/keysetup.c | 2 +-
fs/crypto/policy.c | 25 +++++++++++++++++++++++++
2 files changed, 26 insertions(+), 1 deletion(-)
diff --git a/fs/crypto/keysetup.c b/fs/crypto/keysetup.c
index 7445ab76e0b32..c0db9e8d31f15 100644
--- a/fs/crypto/keysetup.c
+++ b/fs/crypto/keysetup.c
@@ -217,7 +217,7 @@ static int fscrypt_setup_v2_file_key(struct fscrypt_info *ci,
if (err)
return err;
- if (S_ISDIR(ci->ci_inode->i_mode)) {
+ if (S_ISDIR(ci->ci_inode->i_mode) && IS_CASEFOLDED(ci->ci_inode)) {
err = fscrypt_hkdf_expand(&mk->mk_secret.hkdf,
HKDF_CONTEXT_FNAME_HASH_KEY,
ci->ci_nonce,
diff --git a/fs/crypto/policy.c b/fs/crypto/policy.c
index 2cd9a940d8f46..632ca355e1184 100644
--- a/fs/crypto/policy.c
+++ b/fs/crypto/policy.c
@@ -10,6 +10,7 @@
* Modified by Eric Biggers, 2019 for v2 policy support.
*/
+#include <linux/key-type.h>
#include <linux/random.h>
#include <linux/string.h>
#include <linux/mount.h>
@@ -591,6 +592,8 @@ int fscrypt_ioc_setflags_prepare(struct inode *inode,
unsigned int flags)
{
union fscrypt_policy policy;
+ struct fscrypt_info *ci;
+ struct fscrypt_master_key *mk;
int err;
/*
@@ -603,6 +606,28 @@ int fscrypt_ioc_setflags_prepare(struct inode *inode,
return err;
if (policy.version != FSCRYPT_POLICY_V2)
return -EINVAL;
+ err = fscrypt_require_key(inode);
+ if (err)
+ return err;
+ ci = inode->i_crypt_info;
+ if (ci->ci_hash_key_initialized)
+ return 0;
+ mk = ci->ci_master_key->payload.data[0];
+ down_read(&mk->mk_secret_sem);
+ if (!is_master_key_secret_present(&mk->mk_secret)) {
+ err = -ENOKEY;
+ } else {
+ err = fscrypt_hkdf_expand(&mk->mk_secret.hkdf,
+ HKDF_CONTEXT_FNAME_HASH_KEY,
+ ci->ci_nonce,
+ FS_KEY_DERIVATION_NONCE_SIZE,
+ (u8 *)&ci->ci_hash_key,
+ sizeof(ci->ci_hash_key));
+ }
+ up_read(&mk->mk_secret_sem);
+ if (err)
+ return err;
+ ci->ci_hash_key_initialized = true;
}
return 0;
--
2.25.0.341.g760bfbb309-goog
Casefolding currently requires a derived key for computing the siphash.
This is available for v2 policies, but not v1, so we disallow it for v1.
Signed-off-by: Daniel Rosenberg <[email protected]>
---
fs/crypto/policy.c | 28 ++++++++++++++++++++++++++++
fs/inode.c | 3 ++-
include/linux/fscrypt.h | 11 +++++++++++
3 files changed, 41 insertions(+), 1 deletion(-)
diff --git a/fs/crypto/policy.c b/fs/crypto/policy.c
index f1cff83c151ac..2cd9a940d8f46 100644
--- a/fs/crypto/policy.c
+++ b/fs/crypto/policy.c
@@ -124,6 +124,12 @@ static bool fscrypt_supported_v1_policy(const struct fscrypt_policy_v1 *policy,
policy->filenames_encryption_mode))
return false;
+ if (IS_CASEFOLDED(inode)) {
+ fscrypt_warn(inode,
+ "v1 policy does not support casefolded directories");
+ return false;
+ }
+
return true;
}
@@ -579,3 +585,25 @@ int fscrypt_inherit_context(struct inode *parent, struct inode *child,
return preload ? fscrypt_get_encryption_info(child): 0;
}
EXPORT_SYMBOL(fscrypt_inherit_context);
+
+int fscrypt_ioc_setflags_prepare(struct inode *inode,
+ unsigned int oldflags,
+ unsigned int flags)
+{
+ union fscrypt_policy policy;
+ int err;
+
+ /*
+ * When a directory is encrypted, the CASEFOLD flag can only be turned
+ * on if the fscrypt policy supports it.
+ */
+ if (IS_ENCRYPTED(inode) && (flags & ~oldflags & FS_CASEFOLD_FL)) {
+ err = fscrypt_get_policy(inode, &policy);
+ if (err)
+ return err;
+ if (policy.version != FSCRYPT_POLICY_V2)
+ return -EINVAL;
+ }
+
+ return 0;
+}
diff --git a/fs/inode.c b/fs/inode.c
index 96d62d97694ef..8f6267858d0c1 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -20,6 +20,7 @@
#include <linux/ratelimit.h>
#include <linux/list_lru.h>
#include <linux/iversion.h>
+#include <linux/fscrypt.h>
#include <trace/events/writeback.h>
#include "internal.h"
@@ -2252,7 +2253,7 @@ int vfs_ioc_setflags_prepare(struct inode *inode, unsigned int oldflags,
!capable(CAP_LINUX_IMMUTABLE))
return -EPERM;
- return 0;
+ return fscrypt_ioc_setflags_prepare(inode, oldflags, flags);
}
EXPORT_SYMBOL(vfs_ioc_setflags_prepare);
diff --git a/include/linux/fscrypt.h b/include/linux/fscrypt.h
index 1dfbed855beeb..2c292f19c6b94 100644
--- a/include/linux/fscrypt.h
+++ b/include/linux/fscrypt.h
@@ -142,6 +142,10 @@ extern int fscrypt_ioctl_get_policy_ex(struct file *, void __user *);
extern int fscrypt_has_permitted_context(struct inode *, struct inode *);
extern int fscrypt_inherit_context(struct inode *, struct inode *,
void *, bool);
+extern int fscrypt_ioc_setflags_prepare(struct inode *inode,
+ unsigned int oldflags,
+ unsigned int flags);
+
/* keyring.c */
extern void fscrypt_sb_free(struct super_block *sb);
extern int fscrypt_ioctl_add_key(struct file *filp, void __user *arg);
@@ -383,6 +387,13 @@ static inline int fscrypt_inherit_context(struct inode *parent,
return -EOPNOTSUPP;
}
+static inline int fscrypt_ioc_setflags_prepare(struct inode *inode,
+ unsigned int oldflags,
+ unsigned int flags)
+{
+ return 0;
+}
+
/* keyring.c */
static inline void fscrypt_sb_free(struct super_block *sb)
{
--
2.25.0.341.g760bfbb309-goog
On Fri, Jan 17, 2020 at 01:42:42PM -0800, Daniel Rosenberg wrote:
> Ext4 and F2fs are both using casefolding, and they, along with any other
> filesystem that adds the feature, will be using identical dentry_ops.
> Additionally, those dentry ops interfere with the dentry_ops required
> for fscrypt once we add support for casefolding and encryption.
> Moving this into the vfs removes code duplication as well as the
> complication with encryption.
>
> Currently this is pretty close to just moving the existing f2fs/ext4
> code up a level into the vfs,
... buggering the filesystems (and boxen) that never planned to use
that garbage.
> @@ -247,7 +248,19 @@ static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *c
> * be no NUL in the ct/tcount data)
> */
> const unsigned char *cs = READ_ONCE(dentry->d_name.name);
> +#ifdef CONFIG_UNICODE
> + struct inode *parent = dentry->d_parent->d_inode;
What happens if dentry gets moved under you? And that's not mentioning the joy
of extra cachelines to shit the cache with. For every sodding dentry in the
hashchain you are walking.
> + if (unlikely(needs_casefold(parent))) {
> + const struct qstr n1 = QSTR_INIT(cs, tcount);
> + const struct qstr n2 = QSTR_INIT(ct, tcount);
> + int result = utf8_strncasecmp(dentry->d_sb->s_encoding,
> + &n1, &n2);
Is that safe in face of renames? We are *NOT* guaranteed ->d_lock here;
->d_name can change under you just fine. False negatives are OK, but
there's a lot more ways for the things to go wrong.
> static int link_path_walk(const char *name, struct nameidata *nd)
> {
> +#ifdef CONFIG_UNICODE
> + if (needs_casefold(nd->path.dentry->d_inode)) {
> + struct qstr str = QSTR_INIT(name, PATH_MAX);
> +
> + hname = kmalloc(PATH_MAX, GFP_ATOMIC);
> + if (!hname)
> + return -ENOMEM;
> + hlen = utf8_casefold(nd->path.dentry->d_sb->s_encoding,
> + &str, hname, PATH_MAX);
> + }
> + hash_len = hash_name(nd->path.dentry, hname ?: name);
> + kfree(hname);
> + hname = NULL;
> +#else
> hash_len = hash_name(nd->path.dentry, name);
> -
> +#endif
Are you serious?
1) who said that ->d_inode is stable here? If we are in RCU mode,
it won't be.
2) page-sized kmalloc/kfree *ON* *COMPONENT* *AFTER* *COMPONENT*?
> +static inline bool needs_casefold(const struct inode *dir)
> +{
> + return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding &&
> + (!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir));
... and again, you are pulling in a lot of cachelines.
<understatement> IMO the whole thing is not a good idea. </understatement>
On Fri, Jan 17, 2020 at 01:42:37PM -0800, Daniel Rosenberg wrote:
> These patches are all on top of fscrypt's developement branch
>
> Ext4 and F2FS currently both support casefolding and encryption, but not at
> the same time. These patches aim to rectify that.
>
> Since directory names are stored case preserved, we cannot just take the hash
> of the ciphertext. Instead we use the siphash of the casefolded name. With this
> we no longer have a direct path from an encrypted name to the hash without the
> key. To deal with this, fscrypt now always includes the hash in the name it
> presents when the key is not present. There is a pre-existing bug where you can
> change parts of the hash and still match the name so long as the disruption to
> the hash does not happen to affect lookup on that filesystem. I'm not sure how
> to fix that without making ext4 lookups slower in the more common case.
>
> I moved the identical dcache operations for ext4 and f2fs into the VFS, as any
> filesystem that uses casefolding will need the same code. This will also allow
> further optimizations to that path, although my current changes don't take
> advantage of that yet.
>
> For Ext4, this also means that we need to store the hash on disk. We only do so
> for encrypted and casefolded directories to avoid on disk format changes.
> Previously encryption and casefolding could not live on the same filesystem,
> and we're relaxing that requirement. F2fs is a bit more straightforward since
> it already stores hashes on disk.
>
> I've updated the related tools with just enough to enable the feature. I still
> need to adjust ext4's fsck's, although without access to the keys,
> neither fsck will be able to verify the hashes of casefolded and encrypted names.
>
> v3 changes:
> fscrypt patch only creates hash key if it will be needed.
> Rebased on top of fscrypt branch, reconstified match functions in ext4/f2fs
>
> v2 changes:
> fscrypt moved to separate thread to rebase on fscrypt dev branch
> addressed feedback, plus some minor fixes
>
>
> Daniel Rosenberg (9):
> fscrypt: Add siphash and hash key for policy v2
> fscrypt: Don't allow v1 policies with casefolding
> fscrypt: Change format of no-key token
> fscrypt: Only create hash key when needed
> vfs: Fold casefolding into vfs
> f2fs: Handle casefolding with Encryption
> ext4: Use struct super_blocks' casefold data
> ext4: Hande casefolding with encryption
> ext4: Optimize match for casefolded encrypted dirs
Thanks for the new version of this patchset, Daniel!
I'd like to apply the first four patches (the fs/crypto/ part, to prepare for
the new dirhash method) for 5.6, to get ready for the actual
encrypted+casefolded support in filesystems later.
But we don't have much time left before the merge window, the more I look at the
patches I'm still not very happy with them. E.g., some comments I made haven't
been addressed, it's missing updates to the documentation, and some of the code
comments and commit messages are still confusing. For one, there's still some
ambiguity between the dirhash and the SHA-256 hash, and it's not really
explained why the patch introduces the SHA-256 stuff, which actually has nothing
to do with encrypted+casefold (other than it was a good opportunity to do it as
the nokey name format had to be changed for encrypted+casefold anyway).
I also found a bug where the return value of base64_decode() isn't being checked
properly. We should also keep fscrypt_match_name() simpler by setting disk_name
for short names, like we were before. There are also some places that count the
padding in struct fscrypt_nokey_name and some that don't, which is confusing.
We also no longer need to call fscrypt_get_policy() during setflags, as we call
fscrypt_require_key() now anyway. And there's now some ambiguity about what's
meant by a "per-file key", since now there will be 2 types of per-file keys.
So I hope you don't mind, but to move things along I've had a go at cleaning up
the fscrypt patches, and I've sent out an updated version of them. Can you
please take a look when you have a chance?:
https://lkml.kernel.org/linux-fscrypt/[email protected]/T/#u
Thanks!
- Eric
On Sun, Jan 19, 2020 at 08:52:16PM -0800, Eric Biggers wrote:
> On Fri, Jan 17, 2020 at 01:42:37PM -0800, Daniel Rosenberg wrote:
> > These patches are all on top of fscrypt's developement branch
> >
> > Ext4 and F2FS currently both support casefolding and encryption, but not at
> > the same time. These patches aim to rectify that.
> >
> > Since directory names are stored case preserved, we cannot just take the hash
> > of the ciphertext. Instead we use the siphash of the casefolded name. With this
> > we no longer have a direct path from an encrypted name to the hash without the
> > key. To deal with this, fscrypt now always includes the hash in the name it
> > presents when the key is not present. There is a pre-existing bug where you can
> > change parts of the hash and still match the name so long as the disruption to
> > the hash does not happen to affect lookup on that filesystem. I'm not sure how
> > to fix that without making ext4 lookups slower in the more common case.
> >
> > I moved the identical dcache operations for ext4 and f2fs into the VFS, as any
> > filesystem that uses casefolding will need the same code. This will also allow
> > further optimizations to that path, although my current changes don't take
> > advantage of that yet.
> >
> > For Ext4, this also means that we need to store the hash on disk. We only do so
> > for encrypted and casefolded directories to avoid on disk format changes.
> > Previously encryption and casefolding could not live on the same filesystem,
> > and we're relaxing that requirement. F2fs is a bit more straightforward since
> > it already stores hashes on disk.
> >
> > I've updated the related tools with just enough to enable the feature. I still
> > need to adjust ext4's fsck's, although without access to the keys,
> > neither fsck will be able to verify the hashes of casefolded and encrypted names.
> >
> > v3 changes:
> > fscrypt patch only creates hash key if it will be needed.
> > Rebased on top of fscrypt branch, reconstified match functions in ext4/f2fs
> >
> > v2 changes:
> > fscrypt moved to separate thread to rebase on fscrypt dev branch
> > addressed feedback, plus some minor fixes
> >
> >
> > Daniel Rosenberg (9):
> > fscrypt: Add siphash and hash key for policy v2
> > fscrypt: Don't allow v1 policies with casefolding
> > fscrypt: Change format of no-key token
> > fscrypt: Only create hash key when needed
> > vfs: Fold casefolding into vfs
> > f2fs: Handle casefolding with Encryption
> > ext4: Use struct super_blocks' casefold data
> > ext4: Hande casefolding with encryption
> > ext4: Optimize match for casefolded encrypted dirs
>
> Thanks for the new version of this patchset, Daniel!
>
> I'd like to apply the first four patches (the fs/crypto/ part, to prepare for
> the new dirhash method) for 5.6, to get ready for the actual
> encrypted+casefolded support in filesystems later.
>
> But we don't have much time left before the merge window, the more I look at the
> patches I'm still not very happy with them. E.g., some comments I made haven't
> been addressed, it's missing updates to the documentation, and some of the code
> comments and commit messages are still confusing. For one, there's still some
> ambiguity between the dirhash and the SHA-256 hash, and it's not really
> explained why the patch introduces the SHA-256 stuff, which actually has nothing
> to do with encrypted+casefold (other than it was a good opportunity to do it as
> the nokey name format had to be changed for encrypted+casefold anyway).
>
> I also found a bug where the return value of base64_decode() isn't being checked
> properly. We should also keep fscrypt_match_name() simpler by setting disk_name
> for short names, like we were before. There are also some places that count the
> padding in struct fscrypt_nokey_name and some that don't, which is confusing.
> We also no longer need to call fscrypt_get_policy() during setflags, as we call
> fscrypt_require_key() now anyway. And there's now some ambiguity about what's
> meant by a "per-file key", since now there will be 2 types of per-file keys.
>
> So I hope you don't mind, but to move things along I've had a go at cleaning up
> the fscrypt patches, and I've sent out an updated version of them. Can you
> please take a look when you have a chance?:
> https://lkml.kernel.org/linux-fscrypt/[email protected]/T/#u
>
The new fscrypt no-key name format also breaks UBIFS encryption. So we'll need
a couple UBIFS fixes too. I'll send out a new series that includes them.
- Eric
> ... buggering the filesystems (and boxen) that never planned to use
> that garbage.
>
I'm planning to rework this as dentry ops again. Your other comments
point out some issues that also exist in the old dentry_operations, so
that's a good opportunity to fix those up. How do you feel about just
having the two entries in struct super_block? With them there, I can
add the dentry_operations to fs/unicode where they won't bother anyone
else, while not making every filesystem that uses it have to carry
near identical code.
>
> Are you serious?
> 1) who said that ->d_inode is stable here? If we are in RCU mode,
> it won't be.
> 2) page-sized kmalloc/kfree *ON* *COMPONENT* *AFTER* *COMPONENT*?
>
#2 is the part that made me the saddest in the patch. I'm planning to
move this to the unicode subsystem so it can just walk through the
name as it computes the hash without needing any allocation.
>
> ... and again, you are pulling in a lot of cachelines.
>
I probably should've just given it a DCACHE flag, like what fscrypt is
using. A simple flag there would've done everything that I'm doing
without making the cache super sad and making any attempts at making
it actually work with RCU much simpler.
> <understatement> IMO the whole thing is not a good idea. </understatement>