Merge patch series "vfs: recall-only directory delegations for knfsd"

Jeff Layton <jlayton@kernel.org> says:

At the fall NFS Bakeathon last week, the NFS client and server
maintainers had a discussion about how to merge support for directory
delegations. We decided to start with just merging support for simple,
recallable-only directory delegation support, for a number of reasons:

1/ RFC8881 has some gaps in coverage that we are hoping to have
addressed in RFC8881bis. In particular, it's written such that CB_NOTIFY
callbacks require directory position information. That will be hard to
do properly under Linux, so we're planning to extend the spec to allow
that information to be omitted.

2/ client-side support for CB_NOTIFY still lags a bit. The client side
is tricky, as it involves heuristics about when to request a delegation.

3/ we have some early indication that simple, recallable-only
delegations can help performance in some cases. Anna mentioned seeing a
multi-minute speedup in xfstests runs with them enabled. This needs more
investigation, but it's promising and seems like enough justification to
merge support.

This patchset is quite similar to the set I initially posted back in
early 2024. We've merged some GET_DIR_DELEGATION handling patches
since then, but the VFS layer support is basically the same.

One thing that I want to make clear is that with this patchset, userspace
can request a read lease on a directory that will be recalled on
conflicting accesses. I saw no reason to prevent this, and I think it may
be something useful for applications like Samba.

As always, users can disable leases altogether via the fs.leases-enable
sysctl if this is an issue, but I wanted to point this out in case
anyone sees footguns here.

* patches from https://patch.msgid.link/20251111-dir-deleg-ro-v6-0-52f3feebb2f2@kernel.org:
  vfs: expose delegation support to userland
  nfsd: wire up GET_DIR_DELEGATION handling
  nfsd: allow DELEGRETURN on directories
  nfsd: allow filecache to hold S_IFDIR files
  filelock: lift the ban on directory leases in generic_setlease
  vfs: make vfs_symlink break delegations on parent dir
  vfs: make vfs_mknod break delegations on parent directory
  vfs: make vfs_create break delegations on parent directory
  vfs: clean up argument list for vfs_create()
  vfs: break parent dir delegations in open(..., O_CREAT) codepath
  vfs: allow rmdir to wait for delegation break on parent
  vfs: allow mkdir to wait for delegation break on parent
  vfs: add try_break_deleg calls for parents to vfs_{link,rename,unlink}
  filelock: push the S_ISREG check down to ->setlease handlers
  filelock: add struct delegated_inode
  filelock: rework the __break_lease API to use flags
  filelock: make lease_alloc() take a flags argument

Link: https://patch.msgid.link/20251111-dir-deleg-ro-v6-0-52f3feebb2f2@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>

3 files changed, 91 insertions, 35 deletions

diff --git a/include/linux/filelock.h b/include/linux/filelock.h
index c2ce8ba05d06..54b824c05299 100644
--- a/include/linux/filelock.h
+++ b/include/linux/filelock.h
@@ -159,6 +159,8 @@ int fcntl_setlk64(unsigned int, struct file *, unsigned int,
 
 int fcntl_setlease(unsigned int fd, struct file *filp, int arg);
 int fcntl_getlease(struct file *filp);
+int fcntl_setdeleg(unsigned int fd, struct file *filp, struct delegation *deleg);
+int fcntl_getdeleg(struct file *filp, struct delegation *deleg);
 
 static inline bool lock_is_unlock(struct file_lock *fl)
 {
@@ -212,7 +214,14 @@ int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl);
 void locks_init_lease(struct file_lease *);
 void locks_free_lease(struct file_lease *fl);
 struct file_lease *locks_alloc_lease(void);
-int __break_lease(struct inode *inode, unsigned int flags, unsigned int type);
+
+#define LEASE_BREAK_LEASE		BIT(0)	// break leases and delegations
+#define LEASE_BREAK_DELEG		BIT(1)	// break delegations only
+#define LEASE_BREAK_LAYOUT		BIT(2)	// break layouts only
+#define LEASE_BREAK_NONBLOCK		BIT(3)	// non-blocking break
+#define LEASE_BREAK_OPEN_RDONLY		BIT(4)	// readonly open event
+
+int __break_lease(struct inode *inode, unsigned int flags);
 void lease_get_mtime(struct inode *, struct timespec64 *time);
 int generic_setlease(struct file *, int, struct file_lease **, void **priv);
 int kernel_setlease(struct file *, int, struct file_lease **, void **);
@@ -271,6 +280,16 @@ static inline int fcntl_getlease(struct file *filp)
 	return F_UNLCK;
 }
 
+static inline int fcntl_setdeleg(unsigned int fd, struct file *filp, struct delegation *deleg)
+{
+	return -EINVAL;
+}
+
+static inline int fcntl_getdeleg(struct file *filp, struct delegation *deleg)
+{
+	return -EINVAL;
+}
+
 static inline bool lock_is_unlock(struct file_lock *fl)
 {
 	return false;
@@ -367,7 +386,7 @@ static inline int locks_lock_inode_wait(struct inode *inode, struct file_lock *f
 	return -ENOLCK;
 }
 
-static inline int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
+static inline int __break_lease(struct inode *inode, unsigned int flags)
 {
 	return 0;
 }
@@ -428,6 +447,17 @@ static inline int locks_lock_file_wait(struct file *filp, struct file_lock *fl)
 }
 
 #ifdef CONFIG_FILE_LOCKING
+static inline unsigned int openmode_to_lease_flags(unsigned int mode)
+{
+	unsigned int flags = 0;
+
+	if ((mode & O_ACCMODE) == O_RDONLY)
+		flags |= LEASE_BREAK_OPEN_RDONLY;
+	if (mode & O_NONBLOCK)
+		flags |= LEASE_BREAK_NONBLOCK;
+	return flags;
+}
+
 static inline int break_lease(struct inode *inode, unsigned int mode)
 {
 	struct file_lock_context *flctx;
@@ -443,11 +473,11 @@ static inline int break_lease(struct inode *inode, unsigned int mode)
 		return 0;
 	smp_mb();
 	if (!list_empty_careful(&flctx->flc_lease))
-		return __break_lease(inode, mode, FL_LEASE);
+		return __break_lease(inode, LEASE_BREAK_LEASE | openmode_to_lease_flags(mode));
 	return 0;
 }
 
-static inline int break_deleg(struct inode *inode, unsigned int mode)
+static inline int break_deleg(struct inode *inode, unsigned int flags)
 {
 	struct file_lock_context *flctx;
 
@@ -461,60 +491,84 @@ static inline int break_deleg(struct inode *inode, unsigned int mode)
 	if (!flctx)
 		return 0;
 	smp_mb();
-	if (!list_empty_careful(&flctx->flc_lease))
-		return __break_lease(inode, mode, FL_DELEG);
+	if (!list_empty_careful(&flctx->flc_lease)) {
+		flags |= LEASE_BREAK_DELEG;
+		return __break_lease(inode, flags);
+	}
 	return 0;
 }
 
-static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
+struct delegated_inode {
+	struct inode *di_inode;
+};
+
+static inline bool is_delegated(struct delegated_inode *di)
+{
+	return di->di_inode;
+}
+
+static inline int try_break_deleg(struct inode *inode,
+				  struct delegated_inode *di)
 {
 	int ret;
 
-	ret = break_deleg(inode, O_WRONLY|O_NONBLOCK);
-	if (ret == -EWOULDBLOCK && delegated_inode) {
-		*delegated_inode = inode;
+	ret = break_deleg(inode, LEASE_BREAK_NONBLOCK);
+	if (ret == -EWOULDBLOCK && di) {
+		di->di_inode = inode;
 		ihold(inode);
 	}
 	return ret;
 }
 
-static inline int break_deleg_wait(struct inode **delegated_inode)
+static inline int break_deleg_wait(struct delegated_inode *di)
 {
 	int ret;
 
-	ret = break_deleg(*delegated_inode, O_WRONLY);
-	iput(*delegated_inode);
-	*delegated_inode = NULL;
+	ret = break_deleg(di->di_inode, 0);
+	iput(di->di_inode);
+	di->di_inode = NULL;
 	return ret;
 }
 
 static inline int break_layout(struct inode *inode, bool wait)
 {
 	smp_mb();
-	if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
-		return __break_lease(inode,
-				wait ? O_WRONLY : O_WRONLY | O_NONBLOCK,
-				FL_LAYOUT);
+	if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease)) {
+		unsigned int flags = LEASE_BREAK_LAYOUT;
+
+		if (!wait)
+			flags |= LEASE_BREAK_NONBLOCK;
+
+		return __break_lease(inode, flags);
+	}
 	return 0;
 }
 
 #else /* !CONFIG_FILE_LOCKING */
-static inline int break_lease(struct inode *inode, unsigned int mode)
+struct delegated_inode { };
+
+static inline bool is_delegated(struct delegated_inode *di)
+{
+	return false;
+}
+
+static inline int break_lease(struct inode *inode, bool wait)
 {
 	return 0;
 }
 
-static inline int break_deleg(struct inode *inode, unsigned int mode)
+static inline int break_deleg(struct inode *inode, unsigned int flags)
 {
 	return 0;
 }
 
-static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
+static inline int try_break_deleg(struct inode *inode,
+				  struct delegated_inode *delegated_inode)
 {
 	return 0;
 }
 
-static inline int break_deleg_wait(struct inode **delegated_inode)
+static inline int break_deleg_wait(struct delegated_inode *delegated_inode)
 {
 	BUG();
 	return 0;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index c895146c1444..64323e618724 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -80,6 +80,7 @@ struct fs_context;
 struct fs_parameter_spec;
 struct file_kattr;
 struct iomap_ops;
+struct delegated_inode;
 
 extern void __init inode_init(void);
 extern void __init inode_init_early(void);
@@ -2110,19 +2111,20 @@ bool inode_owner_or_capable(struct mnt_idmap *idmap,
 /*
  * VFS helper functions..
  */
-int vfs_create(struct mnt_idmap *, struct inode *,
-	       struct dentry *, umode_t, bool);
+int vfs_create(struct mnt_idmap *, struct dentry *, umode_t,
+	       struct delegated_inode *);
 struct dentry *vfs_mkdir(struct mnt_idmap *, struct inode *,
-			 struct dentry *, umode_t);
+			 struct dentry *, umode_t, struct delegated_inode *);
 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *,
-              umode_t, dev_t);
+	      umode_t, dev_t, struct delegated_inode *);
 int vfs_symlink(struct mnt_idmap *, struct inode *,
-		struct dentry *, const char *);
+		struct dentry *, const char *, struct delegated_inode *);
 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *,
-	     struct dentry *, struct inode **);
-int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *);
+	     struct dentry *, struct delegated_inode *);
+int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *,
+	      struct delegated_inode *);
 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *,
-	       struct inode **);
+	       struct delegated_inode *);
 
 /**
  * struct renamedata - contains all information required for renaming
@@ -2140,7 +2142,7 @@ struct renamedata {
 	struct dentry *old_dentry;
 	struct dentry *new_parent;
 	struct dentry *new_dentry;
-	struct inode **delegated_inode;
+	struct delegated_inode *delegated_inode;
 	unsigned int flags;
 } __randomize_layout;
 
@@ -2150,7 +2152,7 @@ static inline int vfs_whiteout(struct mnt_idmap *idmap,
 			       struct inode *dir, struct dentry *dentry)
 {
 	return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE,
-			 WHITEOUT_DEV);
+			 WHITEOUT_DEV, NULL);
 }
 
 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap,
@@ -3071,7 +3073,7 @@ static inline int bmap(struct inode *inode,  sector_t *block)
 #endif
 
 int notify_change(struct mnt_idmap *, struct dentry *,
-		  struct iattr *, struct inode **);
+		  struct iattr *, struct delegated_inode *);
 int inode_permission(struct mnt_idmap *, struct inode *, int);
 int generic_permission(struct mnt_idmap *, struct inode *, int);
 static inline int file_permission(struct file *file, int mask)
diff --git a/include/linux/xattr.h b/include/linux/xattr.h
index 86b0d47984a1..64e9afe7d647 100644
--- a/include/linux/xattr.h
+++ b/include/linux/xattr.h
@@ -85,12 +85,12 @@ int __vfs_setxattr_noperm(struct mnt_idmap *, struct dentry *,
 			  const char *, const void *, size_t, int);
 int __vfs_setxattr_locked(struct mnt_idmap *, struct dentry *,
 			  const char *, const void *, size_t, int,
-			  struct inode **);
+			  struct delegated_inode *);
 int vfs_setxattr(struct mnt_idmap *, struct dentry *, const char *,
 		 const void *, size_t, int);
 int __vfs_removexattr(struct mnt_idmap *, struct dentry *, const char *);
 int __vfs_removexattr_locked(struct mnt_idmap *, struct dentry *,
-			     const char *, struct inode **);
+			     const char *, struct delegated_inode *);
 int vfs_removexattr(struct mnt_idmap *, struct dentry *, const char *);
 
 ssize_t generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size);