diff mbox series

[f2fs-dev,v3,02/10] fscrypt: Share code between functions that prepare lookup

Message ID 20240119184742.31088-3-krisman@suse.de (mailing list archive)
State Superseded
Headers show
Series Set casefold/fscrypt dentry operations through sb->s_d_op | expand

Commit Message

Gabriel Krisman Bertazi Jan. 19, 2024, 6:47 p.m. UTC
Right now, we have two functions that can be called by the filesystem
during lookup to set up fscrypt internal state for the dentry and
inode under lookup:

  1) fscrypt_prepare_lookup_dentry_partial: used only by ceph. It sets
  encryption information in the inode and sets the dentry flag if the
  key is not available.

  2) fscrypt_prepare_lookup: used by everything else. Does all the
  above, plus also sets struct fname.

This patch refactors the code to implement the later using the former.
This way, we'll have a single place where we set DCACHE_NOKEY_NAME, in
preparation to add more churn to that condition in the following patch.

To make the patch simpler, we now call fscrypt_get_encryption_info twice
for fscrypt_prepare_lookup, once inside fscrypt_setup_filename and once
inside fscrypt_prepare_lookup_dentry.  It seems safe to do, and
considering it will bail early in the second lookup and most lookups
should go to the dcache anyway, it doesn't seem problematic for
performance.  In addition, we add a function call for the unencrypted
case, also during lookup.

Apart from the above, it should have no behavior change.

I tested ext4/f2fs manually and with fstests, which yielded no regressions.

Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
---
 fs/ceph/dir.c           |  2 +-
 fs/ceph/file.c          |  2 +-
 fs/crypto/hooks.c       | 53 ++++++++++++++++++-----------------------
 include/linux/fscrypt.h | 40 +++++++++++++++++--------------
 4 files changed, 47 insertions(+), 50 deletions(-)

Comments

Eric Biggers Jan. 25, 2024, 3:05 a.m. UTC | #1
On Fri, Jan 19, 2024 at 03:47:34PM -0300, Gabriel Krisman Bertazi wrote:
> To make the patch simpler, we now call fscrypt_get_encryption_info twice
> for fscrypt_prepare_lookup, once inside fscrypt_setup_filename and once
> inside fscrypt_prepare_lookup_dentry.  It seems safe to do, and
> considering it will bail early in the second lookup and most lookups
> should go to the dcache anyway, it doesn't seem problematic for
> performance.  In addition, we add a function call for the unencrypted
> case, also during lookup.

Unfortunately I don't think it's correct.  This is basically undoing my fix
b01531db6cec ("fscrypt: fix race where ->lookup() marks plaintext dentry as
ciphertext") from several years ago.

When a lookup is done, the filesystem needs to either treat the name being
looked up as a no-key name *or* as a regular name, depending on whether the
directory's key is present.  We shouldn't enable race conditions where, due to
the key being concurrently added, the name is treated as a no-key name for
filename matching purposes but a regular name for dentry validation purposes.
That can result in an anomaly where a file that exists ends up with a negative
dentry that doesn't get invalidated.

Basically, the boolean fscrypt_name::is_nokey_name that's produced by
fscrypt_setup_filename() should continue to be propagated to DCACHE_NOKEY_NAME.

- Eric
Gabriel Krisman Bertazi Jan. 25, 2024, 8:18 p.m. UTC | #2
Eric Biggers <ebiggers@kernel.org> writes:

> On Fri, Jan 19, 2024 at 03:47:34PM -0300, Gabriel Krisman Bertazi wrote:
>> To make the patch simpler, we now call fscrypt_get_encryption_info twice
>> for fscrypt_prepare_lookup, once inside fscrypt_setup_filename and once
>> inside fscrypt_prepare_lookup_dentry.  It seems safe to do, and
>> considering it will bail early in the second lookup and most lookups
>> should go to the dcache anyway, it doesn't seem problematic for
>> performance.  In addition, we add a function call for the unencrypted
>> case, also during lookup.
>
> Unfortunately I don't think it's correct.  This is basically undoing my fix
> b01531db6cec ("fscrypt: fix race where ->lookup() marks plaintext dentry as
> ciphertext") from several years ago.
>
> When a lookup is done, the filesystem needs to either treat the name being
> looked up as a no-key name *or* as a regular name, depending on whether the
> directory's key is present.  We shouldn't enable race conditions where, due to
> the key being concurrently added, the name is treated as a no-key name for
> filename matching purposes but a regular name for dentry validation purposes.
> That can result in an anomaly where a file that exists ends up with a negative
> dentry that doesn't get invalidated.
>
> Basically, the boolean fscrypt_name::is_nokey_name that's produced by
> fscrypt_setup_filename() should continue to be propagated to DCACHE_NOKEY_NAME.

I see your point.  I'll drop this patch and replace it with a patch that
just merges the DCACHE_NOKEY_NAME configuration.  Sadly, we gotta keep
the two variants I think.

thanks for the review
diff mbox series

Patch

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
index 91709934c8b1..835421e2845d 100644
--- a/fs/ceph/dir.c
+++ b/fs/ceph/dir.c
@@ -813,7 +813,7 @@  static struct dentry *ceph_lookup(struct inode *dir, struct dentry *dentry,
 	if (IS_ENCRYPTED(dir)) {
 		bool had_key = fscrypt_has_encryption_key(dir);
 
-		err = fscrypt_prepare_lookup_partial(dir, dentry);
+		err = fscrypt_prepare_lookup_dentry(dir, dentry);
 		if (err < 0)
 			return ERR_PTR(err);
 
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index 3b5aae29e944..5c9206670533 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -812,7 +812,7 @@  int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
 	ihold(dir);
 	if (IS_ENCRYPTED(dir)) {
 		set_bit(CEPH_MDS_R_FSCRYPT_FILE, &req->r_req_flags);
-		err = fscrypt_prepare_lookup_partial(dir, dentry);
+		err = fscrypt_prepare_lookup_dentry(dir, dentry);
 		if (err < 0)
 			goto out_req;
 	}
diff --git a/fs/crypto/hooks.c b/fs/crypto/hooks.c
index 52504dd478d3..41df986d1230 100644
--- a/fs/crypto/hooks.c
+++ b/fs/crypto/hooks.c
@@ -94,52 +94,45 @@  int __fscrypt_prepare_rename(struct inode *old_dir, struct dentry *old_dentry,
 }
 EXPORT_SYMBOL_GPL(__fscrypt_prepare_rename);
 
-int __fscrypt_prepare_lookup(struct inode *dir, struct dentry *dentry,
-			     struct fscrypt_name *fname)
-{
-	int err = fscrypt_setup_filename(dir, &dentry->d_name, 1, fname);
-
-	if (err && err != -ENOENT)
-		return err;
-
-	if (fname->is_nokey_name) {
-		spin_lock(&dentry->d_lock);
-		dentry->d_flags |= DCACHE_NOKEY_NAME;
-		spin_unlock(&dentry->d_lock);
-	}
-	return err;
-}
-EXPORT_SYMBOL_GPL(__fscrypt_prepare_lookup);
-
 /**
- * fscrypt_prepare_lookup_partial() - prepare lookup without filename setup
+ * fscrypt_prepare_lookup_dentry() - prepare lookup without filename setup
  * @dir: the encrypted directory being searched
  * @dentry: the dentry being looked up in @dir
  *
- * This function should be used by the ->lookup and ->atomic_open methods of
- * filesystems that handle filename encryption and no-key name encoding
- * themselves and thus can't use fscrypt_prepare_lookup().  Like
- * fscrypt_prepare_lookup(), this will try to set up the directory's encryption
- * key and will set DCACHE_NOKEY_NAME on the dentry if the key is unavailable.
- * However, this function doesn't set up a struct fscrypt_name for the filename.
+ * This function should be used by the ->lookup and ->atomic_open
+ * methods of filesystems that handle filename encryption and no-key
+ * name encoding themselves and thus can't use fscrypt_prepare_lookup()
+ * directly.  This will try to set up the directory's encryption key and
+ * will set DCACHE_NOKEY_NAME on the dentry if the key is unavailable.
+ * However, this function doesn't set up a struct fscrypt_name for the
+ * filename.
  *
  * Return: 0 on success; -errno on error.  Note that the encryption key being
  *	   unavailable is not considered an error.  It is also not an error if
  *	   the encryption policy is unsupported by this kernel; that is treated
  *	   like the key being unavailable, so that files can still be deleted.
  */
-int fscrypt_prepare_lookup_partial(struct inode *dir, struct dentry *dentry)
+int fscrypt_prepare_lookup_dentry(struct inode *dir,
+				  struct dentry *dentry)
 {
-	int err = fscrypt_get_encryption_info(dir, true);
+	bool nokey_name = false;
+	int err = 0;
 
-	if (!err && !fscrypt_has_encryption_key(dir)) {
-		spin_lock(&dentry->d_lock);
+	if (IS_ENCRYPTED(dir)) {
+		err = fscrypt_get_encryption_info(dir, true);
+		if (!err && !fscrypt_has_encryption_key(dir))
+			nokey_name = true;
+	}
+
+	spin_lock(&dentry->d_lock);
+	if (nokey_name) {
 		dentry->d_flags |= DCACHE_NOKEY_NAME;
-		spin_unlock(&dentry->d_lock);
 	}
+	spin_unlock(&dentry->d_lock);
+
 	return err;
 }
-EXPORT_SYMBOL_GPL(fscrypt_prepare_lookup_partial);
+EXPORT_SYMBOL_GPL(fscrypt_prepare_lookup_dentry);
 
 int __fscrypt_prepare_readdir(struct inode *dir)
 {
diff --git a/include/linux/fscrypt.h b/include/linux/fscrypt.h
index 12f9e455d569..3801c5c94fb6 100644
--- a/include/linux/fscrypt.h
+++ b/include/linux/fscrypt.h
@@ -382,9 +382,7 @@  int __fscrypt_prepare_link(struct inode *inode, struct inode *dir,
 int __fscrypt_prepare_rename(struct inode *old_dir, struct dentry *old_dentry,
 			     struct inode *new_dir, struct dentry *new_dentry,
 			     unsigned int flags);
-int __fscrypt_prepare_lookup(struct inode *dir, struct dentry *dentry,
-			     struct fscrypt_name *fname);
-int fscrypt_prepare_lookup_partial(struct inode *dir, struct dentry *dentry);
+int fscrypt_prepare_lookup_dentry(struct inode *dir, struct dentry *dentry);
 int __fscrypt_prepare_readdir(struct inode *dir);
 int __fscrypt_prepare_setattr(struct dentry *dentry, struct iattr *attr);
 int fscrypt_prepare_setflags(struct inode *inode,
@@ -704,14 +702,7 @@  static inline int __fscrypt_prepare_rename(struct inode *old_dir,
 	return -EOPNOTSUPP;
 }
 
-static inline int __fscrypt_prepare_lookup(struct inode *dir,
-					   struct dentry *dentry,
-					   struct fscrypt_name *fname)
-{
-	return -EOPNOTSUPP;
-}
-
-static inline int fscrypt_prepare_lookup_partial(struct inode *dir,
+static inline int fscrypt_prepare_lookup_dentry(struct inode *dir,
 						 struct dentry *dentry)
 {
 	return -EOPNOTSUPP;
@@ -975,14 +966,27 @@  static inline int fscrypt_prepare_lookup(struct inode *dir,
 					 struct dentry *dentry,
 					 struct fscrypt_name *fname)
 {
-	if (IS_ENCRYPTED(dir))
-		return __fscrypt_prepare_lookup(dir, dentry, fname);
+	int ret, err = 0;
+
+	if (IS_ENCRYPTED(dir)) {
+		err = fscrypt_setup_filename(dir, &dentry->d_name, 1, fname);
+		if (err && err != -ENOENT)
+			return err;
+	} else {
+		memset(fname, 0, sizeof(*fname));
+		fname->usr_fname = &dentry->d_name;
+		fname->disk_name.name = (unsigned char *)dentry->d_name.name;
+		fname->disk_name.len = dentry->d_name.len;
+	}
 
-	memset(fname, 0, sizeof(*fname));
-	fname->usr_fname = &dentry->d_name;
-	fname->disk_name.name = (unsigned char *)dentry->d_name.name;
-	fname->disk_name.len = dentry->d_name.len;
-	return 0;
+	/*
+	 * fscrypt_prepare_lookup_dentry will succeed even if the
+	 * directory key is unavailable but the filename isn't a valid
+	 * no-key name (-ENOENT). Thus, propagate the previous
+	 * error, if any.
+	 */
+	ret = fscrypt_prepare_lookup_dentry(dir, dentry);
+	return err ? err : ret;
 }
 
 /**