Fscrypt support for casefolded encryption

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        | 32 +++++++++++++++++++++++---------
 include/linux/fscrypt.h     |  9 +++++++++
 4 files changed, 63 insertions(+), 9 deletions(-)

fs/crypto/fname.c +22 -0

index 3fd27e14ebdd..371e8f01d1c8 100644
@@ -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.

fs/crypto/fscrypt_private.h +9 -0

index b22e8decebed..8b37a5eebb57 100644
@@ -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,

fs/crypto/keysetup.c +16 -9

index 96074054bdbc..c1bd897c9310 100644
@@ -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) {

@@ -199,20 +199,34 @@ 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);
+	if (S_ISDIR(ci->ci_inode->i_mode)) {
+		err = fscrypt_hkdf_expand(&mk->mk_secret.hkdf,
+					  HKDF_CONTEXT_FNAME_HASH_KEY,

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/keysetup.c    |  7 ++++---
 fs/crypto/policy.c      | 39 +++++++++++++++++++++++++++++++++++++++
 fs/inode.c              |  7 +++++++
 include/linux/fscrypt.h | 11 +++++++++++
 4 files changed, 61 insertions(+), 3 deletions(-)

fs/crypto/keysetup.c +0 -0

index c1bd897c9310..7445ab76e0b3 100644
@@ -224,10 +224,11 @@ static int fscrypt_setup_v2_file_key(struct fscrypt_info *ci,
 					  FS_KEY_DERIVATION_NONCE_SIZE,
 					  (u8 *)&ci->ci_hash_key,

fs/crypto/keysetup.c +0 -0

 /*

fs/crypto/policy.c +29 -0

index f1cff83c151a..9e937cfa732c 100644
@@ -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 +585,36 @@ int fscrypt_inherit_context(struct inode *parent, struct inode *child,
 	return preload ? fscrypt_get_encryption_info(child): 0;
 }
 EXPORT_SYMBOL(fscrypt_inherit_context);
+
+static int fscrypt_set_casefolding_allowed(struct inode *inode)
+{
+	union fscrypt_policy policy;
+	int err = fscrypt_get_policy(inode, &policy);
+
+	if (err)
+		return err;
+
+	if (policy.version != FSCRYPT_POLICY_V2)
+		return -EINVAL;
+
+	return 0;
+}
+
+int fscrypt_ioc_setflags_prepare(struct inode *inode,
+				 unsigned int oldflags,
+				 unsigned int flags)
+{
+	int err;
+
+	/*
+	 * When a directory is encrypted, the CASEFOLD flag can only be turned

fs/crypto/policy.c +2 -0

@@ -2242,6 +2243,8 @@ EXPORT_SYMBOL(current_time);
 int vfs_ioc_setflags_prepare(struct inode *inode, unsigned int oldflags,
 			     unsigned int flags)
 {
+	int err;
+
 	/*
 	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
 	 * the relevant capability.

fs/crypto/fname.c +1 -2

@@ -307,8 +372,7 @@ EXPORT_SYMBOL(fscrypt_fname_disk_to_usr);
  * get the disk_name.
  *
  * Else, for keyless @lookup operations, @iname is the presented ciphertext, so
- * we decode it to get either the ciphertext disk_name (for short names) or the
- * fscrypt_digested_name (for long names).  Non-@lookup operations will be
+ * we decode it to get the fscrypt_nokey_name. Non-@lookup operations will be
  * impossible in this case, so we fail them with ENOKEY.
  *
  * If successful, fscrypt_free_filename() must be called later to clean up.

fs/crypto/fname.c +19 -27

 	int ret;
-	int digested;
 
 	memset(fname, 0, sizeof(struct fscrypt_name));
 	fname->usr_fname = iname;

@@ -359,41 +423,29 @@ int fscrypt_setup_filename(struct inode *dir, const struct qstr *iname,
 	 * We don't have the key and we are doing a lookup; decode the
 	 * user-supplied name
 	 */
-	if (iname->name[0] == '_') {
-		if (iname->len !=
-		    1 + BASE64_CHARS(sizeof(struct fscrypt_digested_name)))
-			return -ENOENT;
-		digested = 1;
-	} else {
-		if (iname->len >
-		    BASE64_CHARS(FSCRYPT_FNAME_MAX_UNDIGESTED_SIZE))
-			return -ENOENT;
-		digested = 0;
-	}
 
 	fname->crypto_buf.name =
-		kmalloc(max_t(size_t, FSCRYPT_FNAME_MAX_UNDIGESTED_SIZE,
-			      sizeof(struct fscrypt_digested_name)),
-			GFP_KERNEL);
+			kmalloc(sizeof(struct fscrypt_nokey_name), GFP_KERNEL);
 	if (fname->crypto_buf.name == NULL)
 		return -ENOMEM;
 
-	ret = base64_decode(iname->name + digested, iname->len - digested,
-			    fname->crypto_buf.name);
-	if (ret < 0) {
+	if (iname->len > BASE64_CHARS(sizeof(struct fscrypt_nokey_name))) {
 		ret = -ENOENT;
 		goto errout;
 	}
-	fname->crypto_buf.len = ret;
-	if (digested) {
-		const struct fscrypt_digested_name *n =
-			(const void *)fname->crypto_buf.name;
-		fname->hash = n->hash;
-		fname->minor_hash = n->minor_hash;
-	} else {
-		fname->disk_name.name = fname->crypto_buf.name;
-		fname->disk_name.len = fname->crypto_buf.len;
+	ret = base64_decode(iname->name, iname->len, fname->crypto_buf.name);
+	if ((int)ret < offsetof(struct fscrypt_nokey_name, bytes[1]) ||
+	    (ret > offsetof(struct fscrypt_nokey_name, sha256) &&
+	     ret != offsetofend(struct fscrypt_nokey_name, sha256))) {
+		ret = -ENOENT;
+		goto errout;
 	}
+
+	nokey_name = (void *)fname->crypto_buf.name;
+	fname->crypto_buf.len = ret;
+
+	fname->hash = nokey_name->dirtree_hash[0];
+	fname->minor_hash = nokey_name->dirtree_hash[1];
 	return 0;
 
 errout:

@@ -402,6 +454,62 @@ int fscrypt_setup_filename(struct inode *dir, const struct qstr *iname,
 }
 EXPORT_SYMBOL(fscrypt_setup_filename);
 
+/**
+ * fscrypt_match_name() - test whether the given name matches a directory entry
+ * @fname: the name being searched for
+ * @de_name: the name from the directory entry
+ * @de_name_len: the length of @de_name in bytes

fs/crypto/fname.c +8 -0

+	if (unlikely(!fname->disk_name.name)) {
+		const struct fscrypt_nokey_name *n =
+			(const void *)fname->crypto_buf.name;
+
+		if (fname->crypto_buf.len ==
+			    offsetofend(struct fscrypt_nokey_name, sha256)) {
+			u8 sha256[SHA256_DIGEST_SIZE];
+

fs/crypto/fname.c +19 -0

+			u32 len = fname->crypto_buf.len -
+				offsetof(struct fscrypt_nokey_name, bytes);
+
+			if (de_name_len != len)
+				return false;
+
+			if (memcmp(de_name, n->bytes, len) != 0)
+				return false;
+		}
+
+		return true;
+	}
+
+	if (de_name_len != fname->disk_name.len)
+		return false;
+	return !memcmp(de_name, fname->disk_name.name, fname->disk_name.len);
+}
+EXPORT_SYMBOL(fscrypt_match_name);
+
 /**
  * fscrypt_fname_siphash() - Calculate the siphash for a file name
  * @dir: the parent directory

include/linux/fscrypt.h +0 -65

index 2c292f19c6b9..14a727759a81 100644
@@ -179,79 +179,8 @@ extern int fscrypt_fname_disk_to_usr(const struct inode *inode,
 extern u64 fscrypt_fname_siphash(const struct inode *dir,
 				 const struct qstr *name);
 
-#define FSCRYPT_FNAME_MAX_UNDIGESTED_SIZE	32
-
-/* Extracts the second-to-last ciphertext block; see explanation below */
-#define FSCRYPT_FNAME_DIGEST(name, len)	\
-	((name) + round_down((len) - FS_CRYPTO_BLOCK_SIZE - 1, \
-			     FS_CRYPTO_BLOCK_SIZE))
-
-#define FSCRYPT_FNAME_DIGEST_SIZE	FS_CRYPTO_BLOCK_SIZE
-
-/**
- * fscrypt_digested_name - alternate identifier for an on-disk filename
- *
- * When userspace lists an encrypted directory without access to the key,
- * filenames whose ciphertext is longer than FSCRYPT_FNAME_MAX_UNDIGESTED_SIZE
- * bytes are shown in this abbreviated form (base64-encoded) rather than as the
- * full ciphertext (base64-encoded).  This is necessary to allow supporting
- * filenames up to NAME_MAX bytes, since base64 encoding expands the length.
- *
- * To make it possible for filesystems to still find the correct directory entry
- * despite not knowing the full on-disk name, we encode any filesystem-specific
- * 'hash' and/or 'minor_hash' which the filesystem may need for its lookups,
- * followed by the second-to-last ciphertext block of the filename.  Due to the
- * use of the CBC-CTS encryption mode, the second-to-last ciphertext block
- * depends on the full plaintext.  (Note that ciphertext stealing causes the
- * last two blocks to appear "flipped".)  This makes accidental collisions very
- * unlikely: just a 1 in 2^128 chance for two filenames to collide even if they
- * share the same filesystem-specific hashes.
- *
- * However, this scheme isn't immune to intentional collisions, which can be
- * created by anyone able to create arbitrary plaintext filenames and view them
- * without the key.  Making the "digest" be a real cryptographic hash like
- * SHA-256 over the full ciphertext would prevent this, although it would be
- * less efficient and harder to implement, especially since the filesystem would
- * need to calculate it for each directory entry examined during a search.
- */
-struct fscrypt_digested_name {
-	u32 hash;
-	u32 minor_hash;
-	u8 digest[FSCRYPT_FNAME_DIGEST_SIZE];
-};
-
-/**
- * fscrypt_match_name() - test whether the given name matches a directory entry
- * @fname: the name being searched for
- * @de_name: the name from the directory entry
- * @de_name_len: the length of @de_name in bytes
- *
- * Normally @fname->disk_name will be set, and in that case we simply compare
- * that to the name stored in the directory entry.  The only exception is that
- * if we don't have the key for an encrypted directory and a filename in it is
- * very long, then we won't have the full disk_name and we'll instead need to
- * match against the fscrypt_digested_name.
- *
- * Return: %true if the name matches, otherwise %false.
- */
-static inline bool fscrypt_match_name(const struct fscrypt_name *fname,
-				      const u8 *de_name, u32 de_name_len)
-{
-	if (unlikely(!fname->disk_name.name)) {
-		const struct fscrypt_digested_name *n =
-			(const void *)fname->crypto_buf.name;
-		if (WARN_ON_ONCE(fname->usr_fname->name[0] != '_'))
-			return false;
-		if (de_name_len <= FSCRYPT_FNAME_MAX_UNDIGESTED_SIZE)
-			return false;

These patches are to prepare fscrypt to support casefolding and encryption at the same time. Other patches will add those to the vfs, ext4 and f2fs. These patches are against fscrypt/master

Daniel Rosenberg (3):
fscrypt: Add siphash and hash key for policy v2
fscrypt: Don't allow v1 policies with casefolding
fscrypt: Change format of no-key token