[PATCH 0/2] fsnotify hooks consolidation

[PATCH 0/2] fsnotify hooks consolidation

[Amir Goldstein]

Jan,

I've found this opportunity to reduce the amount of sprinkled
fsnotify hooks.

There are still a few fsnotify hooks sprinkled in some pseudo fs,
but all those removed by this series are obvious boiler plate code
and for most of these fs, this removes all of the custom hooks.

I could send a series to convert each fs with its own patch,
but to me that seems a bit unnecessary.

WDYT?
Amir.

Amir Goldstein (2):
  fsnotify: make fsnotify_create() agnostic to file/dir
  fs: use simple_end_creating helper to consolidate fsnotify hooks

 drivers/android/binder/rust_binderfs.c | 11 +++--------
 drivers/android/binderfs.c             | 10 +++-------
 fs/debugfs/inode.c                     |  5 +----
 fs/libfs.c                             | 14 ++++++++++++++
 fs/nfsd/nfsctl.c                       | 11 +++--------
 fs/tracefs/event_inode.c               |  2 --
 fs/tracefs/inode.c                     |  5 +----
 include/linux/fs.h                     |  1 +
 include/linux/fsnotify.h               |  8 +++++++-
 net/sunrpc/rpc_pipe.c                  | 10 +++-------
 10 files changed, 36 insertions(+), 41 deletions(-)

-- 
2.53.0

[PATCH 1/2] fsnotify: make fsnotify_create() agnostic to file/dir

[Amir Goldstein]

Like fsnotify_delete(), let fsnotify_create() handle creation of both
dir and non-dir objects, if d_inode is available at creation time.

Unlike fsnotify_rmdir(), we do not call fsnotify_create() from
fsnotify_mkdir(), because of the case where d_inode is instantiated
lazily (e.g. kernfs).

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
---
 include/linux/fsnotify.h | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/include/linux/fsnotify.h b/include/linux/fsnotify.h
index 079c18bcdbde6..13156d165d845 100644
--- a/include/linux/fsnotify.h
+++ b/include/linux/fsnotify.h
@@ -318,9 +318,15 @@ static inline void fsnotify_inoderemove(struct inode *inode)
  */
 static inline void fsnotify_create(struct inode *dir, struct dentry *dentry)
 {
+	struct inode *inode = d_inode(dentry);
+	__u32 mask = FS_CREATE;
+
+	if (inode && S_ISDIR(inode->i_mode))
+		mask |= FS_ISDIR;
+
 	audit_inode_child(dir, dentry, AUDIT_TYPE_CHILD_CREATE);
 
-	fsnotify_dirent(dir, dentry, FS_CREATE);
+	fsnotify_dirent(dir, dentry, mask);
 }
 
 /*
-- 
2.53.0

[PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[Amir Goldstein]

Add simple_end_creating() helper which combines fsnotify_create/mkdir()
hook and simple_done_creating().

Use the new helper to consolidate this pattern in several pseudo fs
which had open coded fsnotify_create/mkdir() hooks:
binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.

For those filesystems, the paired fsnotify_delete() hook is already
inside the library helper simple_recursive_removal().

Note that in debugfs_create_symlink(), the fsnotify hook was missing,
so the missing hook is fixed by this change.

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
---
 drivers/android/binder/rust_binderfs.c | 11 +++--------
 drivers/android/binderfs.c             | 10 +++-------
 fs/debugfs/inode.c                     |  5 +----
 fs/libfs.c                             | 14 ++++++++++++++
 fs/nfsd/nfsctl.c                       | 11 +++--------
 fs/tracefs/event_inode.c               |  2 --
 fs/tracefs/inode.c                     |  5 +----
 include/linux/fs.h                     |  1 +
 net/sunrpc/rpc_pipe.c                  | 10 +++-------
 9 files changed, 29 insertions(+), 40 deletions(-)

diff --git a/drivers/android/binder/rust_binderfs.c b/drivers/android/binder/rust_binderfs.c
index ade1c4d924992..d69490c6b9e09 100644
--- a/drivers/android/binder/rust_binderfs.c
+++ b/drivers/android/binder/rust_binderfs.c
@@ -3,7 +3,6 @@
 #include <linux/compiler_types.h>
 #include <linux/errno.h>
 #include <linux/fs.h>
-#include <linux/fsnotify.h>
 #include <linux/gfp.h>
 #include <linux/idr.h>
 #include <linux/init.h>
@@ -186,9 +185,7 @@ static int binderfs_binder_device_create(struct inode *ref_inode,
 
 	inode->i_private = device;
 	d_make_persistent(dentry, inode);
-
-	fsnotify_create(root->d_inode, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 
 	return 0;
 
@@ -481,8 +478,7 @@ static struct dentry *rust_binderfs_create_file(struct dentry *parent, const cha
 	}
 
 	d_make_persistent(dentry, new_inode);
-	fsnotify_create(parent->d_inode, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return dentry;
 }
 
@@ -522,8 +518,7 @@ static struct dentry *binderfs_create_dir(struct dentry *parent,
 	inc_nlink(parent->d_inode);
 	set_nlink(new_inode, 2);
 	d_make_persistent(dentry, new_inode);
-	fsnotify_mkdir(parent->d_inode, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return dentry;
 }
 
diff --git a/drivers/android/binderfs.c b/drivers/android/binderfs.c
index 361d69f756f50..60277f0bb679f 100644
--- a/drivers/android/binderfs.c
+++ b/drivers/android/binderfs.c
@@ -3,7 +3,6 @@
 #include <linux/compiler_types.h>
 #include <linux/errno.h>
 #include <linux/fs.h>
-#include <linux/fsnotify.h>
 #include <linux/gfp.h>
 #include <linux/idr.h>
 #include <linux/init.h>
@@ -190,8 +189,7 @@ static int binderfs_binder_device_create(struct inode *ref_inode,
 	}
 	inode->i_private = device;
 	d_make_persistent(dentry, inode);
-	fsnotify_create(root->d_inode, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 
 	binder_add_device(device);
 
@@ -490,8 +488,7 @@ struct dentry *binderfs_create_file(struct dentry *parent, const char *name,
 	new_inode->i_fop = fops;
 	new_inode->i_private = data;
 	d_make_persistent(dentry, new_inode);
-	fsnotify_create(parent_inode, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return dentry; // borrowed
 }
 
@@ -521,8 +518,7 @@ static struct dentry *binderfs_create_dir(struct dentry *parent,
 	set_nlink(new_inode, 2);
 	d_make_persistent(dentry, new_inode);
 	inc_nlink(parent_inode);
-	fsnotify_mkdir(parent_inode, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return dentry;
 }
 
diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
index 4598142355b97..f51b008d42cbf 100644
--- a/fs/debugfs/inode.c
+++ b/fs/debugfs/inode.c
@@ -409,7 +409,7 @@ static struct dentry *debugfs_failed_creating(struct dentry *dentry)
 
 static struct dentry *debugfs_end_creating(struct dentry *dentry)
 {
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return dentry; // borrowed
 }
 
@@ -448,7 +448,6 @@ static struct dentry *__debugfs_create_file(const char *name, umode_t mode,
 	DEBUGFS_I(inode)->aux = (void *)aux;
 
 	d_make_persistent(dentry, inode);
-	fsnotify_create(d_inode(dentry->d_parent), dentry);
 	return debugfs_end_creating(dentry);
 }
 
@@ -590,7 +589,6 @@ struct dentry *debugfs_create_dir(const char *name, struct dentry *parent)
 	inc_nlink(inode);
 	d_make_persistent(dentry, inode);
 	inc_nlink(d_inode(dentry->d_parent));
-	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
 	return debugfs_end_creating(dentry);
 }
 EXPORT_SYMBOL_GPL(debugfs_create_dir);
@@ -632,7 +630,6 @@ struct dentry *debugfs_create_automount(const char *name,
 	inc_nlink(inode);
 	d_make_persistent(dentry, inode);
 	inc_nlink(d_inode(dentry->d_parent));
-	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
 	return debugfs_end_creating(dentry);
 }
 EXPORT_SYMBOL(debugfs_create_automount);
diff --git a/fs/libfs.c b/fs/libfs.c
index 74134ba2e8d1e..2f6dcae40ce7f 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -2322,3 +2322,17 @@ void simple_done_creating(struct dentry *child)
 	dput(child);
 }
 EXPORT_SYMBOL(simple_done_creating);
+
+/*
+ * Like simple_done_creating(), but also fires fsnotify_create() for the
+ * new dentry.  Call on the success path after d_make_persistent().
+ * Use simple_done_creating() on the failure path where the child dentry is
+ * still negative.
+ */
+void simple_end_creating(struct dentry *child)
+{
+	fsnotify_create(child->d_parent->d_inode, child);
+	inode_unlock(child->d_parent->d_inode);
+	dput(child);
+}
+EXPORT_SYMBOL(simple_end_creating);
diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index e9acd2cd602cb..6e600d52b66d0 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -17,7 +17,6 @@
 #include <linux/sunrpc/rpc_pipe_fs.h>
 #include <linux/sunrpc/svc.h>
 #include <linux/module.h>
-#include <linux/fsnotify.h>
 #include <linux/nfslocalio.h>
 
 #include "idmap.h"
@@ -1146,8 +1145,7 @@ static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *nc
 	}
 	d_make_persistent(dentry, inode);
 	inc_nlink(dir);
-	fsnotify_mkdir(dir, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return dentry;	// borrowed
 }
 
@@ -1178,8 +1176,7 @@ static void _nfsd_symlink(struct dentry *parent, const char *name,
 	inode->i_size = strlen(content);
 
 	d_make_persistent(dentry, inode);
-	fsnotify_create(dir, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 }
 #else
 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
@@ -1219,7 +1216,6 @@ static int nfsdfs_create_files(struct dentry *root,
 				struct nfsdfs_client *ncl,
 				struct dentry **fdentries)
 {
-	struct inode *dir = d_inode(root);
 	struct dentry *dentry;
 
 	for (int i = 0; files->name && files->name[0]; i++, files++) {
@@ -1236,10 +1232,9 @@ static int nfsdfs_create_files(struct dentry *root,
 		inode->i_fop = files->ops;
 		inode->i_private = ncl;
 		d_make_persistent(dentry, inode);
-		fsnotify_create(dir, dentry);
 		if (fdentries)
 			fdentries[i] = dentry; // borrowed
-		simple_done_creating(dentry);
+		simple_end_creating(dentry);
 	}
 	return 0;
 }
diff --git a/fs/tracefs/event_inode.c b/fs/tracefs/event_inode.c
index 8e5ac464b3284..059ea7560b9b4 100644
--- a/fs/tracefs/event_inode.c
+++ b/fs/tracefs/event_inode.c
@@ -14,7 +14,6 @@
  *  inodes/dentries in a just-in-time (JIT) manner. The eventfs will clean up
  *  and delete the inodes/dentries when they are no longer referenced.
  */
-#include <linux/fsnotify.h>
 #include <linux/fs.h>
 #include <linux/namei.h>
 #include <linux/workqueue.h>
@@ -826,7 +825,6 @@ struct eventfs_inode *eventfs_create_events_dir(const char *name, struct dentry
 	d_make_persistent(dentry, inode);
 	/* The dentry of the "events" parent does keep track though */
 	inc_nlink(dentry->d_parent->d_inode);
-	fsnotify_mkdir(dentry->d_parent->d_inode, dentry);
 	tracefs_end_creating(dentry);
 
 	return ei;
diff --git a/fs/tracefs/inode.c b/fs/tracefs/inode.c
index 51c00c8fa1755..90a15c064e309 100644
--- a/fs/tracefs/inode.c
+++ b/fs/tracefs/inode.c
@@ -16,7 +16,6 @@
 #include <linux/kobject.h>
 #include <linux/namei.h>
 #include <linux/tracefs.h>
-#include <linux/fsnotify.h>
 #include <linux/security.h>
 #include <linux/seq_file.h>
 #include <linux/magic.h>
@@ -578,7 +577,7 @@ struct dentry *tracefs_failed_creating(struct dentry *dentry)
 
 struct dentry *tracefs_end_creating(struct dentry *dentry)
 {
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return dentry;	// borrowed
 }
 
@@ -661,7 +660,6 @@ struct dentry *tracefs_create_file(const char *name, umode_t mode,
 	inode->i_uid = d_inode(dentry->d_parent)->i_uid;
 	inode->i_gid = d_inode(dentry->d_parent)->i_gid;
 	d_make_persistent(dentry, inode);
-	fsnotify_create(d_inode(dentry->d_parent), dentry);
 	return tracefs_end_creating(dentry);
 }
 
@@ -693,7 +691,6 @@ static struct dentry *__create_dir(const char *name, struct dentry *parent,
 	inc_nlink(inode);
 	d_make_persistent(dentry, inode);
 	inc_nlink(d_inode(dentry->d_parent));
-	fsnotify_mkdir(d_inode(dentry->d_parent), dentry);
 	return tracefs_end_creating(dentry);
 }
 
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 8b3dd145b25ec..8e5080be34308 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -3271,6 +3271,7 @@ extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int
 extern void simple_release_fs(struct vfsmount **mount, int *count);
 struct dentry *simple_start_creating(struct dentry *, const char *);
 void simple_done_creating(struct dentry *);
+void simple_end_creating(struct dentry *);
 
 extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
 			loff_t *ppos, const void *from, size_t available);
diff --git a/net/sunrpc/rpc_pipe.c b/net/sunrpc/rpc_pipe.c
index 9d349cfbc4831..84d0ba37d9d6f 100644
--- a/net/sunrpc/rpc_pipe.c
+++ b/net/sunrpc/rpc_pipe.c
@@ -16,7 +16,6 @@
 #include <linux/mount.h>
 #include <linux/fs_context.h>
 #include <linux/namei.h>
-#include <linux/fsnotify.h>
 #include <linux/kernel.h>
 #include <linux/rcupdate.h>
 #include <linux/utsname.h>
@@ -544,8 +543,7 @@ static int rpc_new_file(struct dentry *parent,
 		inode->i_fop = i_fop;
 	rpc_inode_setowner(inode, private);
 	d_make_persistent(dentry, inode);
-	fsnotify_create(dir, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return 0;
 }
 
@@ -569,8 +567,7 @@ static struct dentry *rpc_new_dir(struct dentry *parent,
 	inode->i_ino = iunique(dir->i_sb, 100);
 	inc_nlink(dir);
 	d_make_persistent(dentry, inode);
-	fsnotify_mkdir(dir, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 
 	return dentry; // borrowed
 }
@@ -667,8 +664,7 @@ int rpc_mkpipe_dentry(struct dentry *parent, const char *name,
 	rpc_inode_setowner(inode, private);
 	pipe->dentry = dentry; // borrowed
 	d_make_persistent(dentry, inode);
-	fsnotify_create(dir, dentry);
-	simple_done_creating(dentry);
+	simple_end_creating(dentry);
 	return 0;
 
 failed:
-- 
2.53.0

Re: [PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[Chuck Lever]

On 3/2/26 1:37 PM, Amir Goldstein wrote:
> Add simple_end_creating() helper which combines fsnotify_create/mkdir()
> hook and simple_done_creating().
> 
> Use the new helper to consolidate this pattern in several pseudo fs
> which had open coded fsnotify_create/mkdir() hooks:
> binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.
> 
> For those filesystems, the paired fsnotify_delete() hook is already
> inside the library helper simple_recursive_removal().
> 
> Note that in debugfs_create_symlink(), the fsnotify hook was missing,
> so the missing hook is fixed by this change.
> 
> Signed-off-by: Amir Goldstein <amir73il@gmail.com>

> diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> index e9acd2cd602cb..6e600d52b66d0 100644
> --- a/fs/nfsd/nfsctl.c
> +++ b/fs/nfsd/nfsctl.c
> @@ -17,7 +17,6 @@
>  #include <linux/sunrpc/rpc_pipe_fs.h>
>  #include <linux/sunrpc/svc.h>
>  #include <linux/module.h>
> -#include <linux/fsnotify.h>
>  #include <linux/nfslocalio.h>
>  
>  #include "idmap.h"
> @@ -1146,8 +1145,7 @@ static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *nc
>  	}
>  	d_make_persistent(dentry, inode);
>  	inc_nlink(dir);
> -	fsnotify_mkdir(dir, dentry);
> -	simple_done_creating(dentry);
> +	simple_end_creating(dentry);
>  	return dentry;	// borrowed
>  }
>  
> @@ -1178,8 +1176,7 @@ static void _nfsd_symlink(struct dentry *parent, const char *name,
>  	inode->i_size = strlen(content);
>  
>  	d_make_persistent(dentry, inode);
> -	fsnotify_create(dir, dentry);
> -	simple_done_creating(dentry);
> +	simple_end_creating(dentry);
>  }
>  #else
>  static inline void _nfsd_symlink(struct dentry *parent, const char *name,
> @@ -1219,7 +1216,6 @@ static int nfsdfs_create_files(struct dentry *root,
>  				struct nfsdfs_client *ncl,
>  				struct dentry **fdentries)
>  {
> -	struct inode *dir = d_inode(root);
>  	struct dentry *dentry;
>  
>  	for (int i = 0; files->name && files->name[0]; i++, files++) {
> @@ -1236,10 +1232,9 @@ static int nfsdfs_create_files(struct dentry *root,
>  		inode->i_fop = files->ops;
>  		inode->i_private = ncl;
>  		d_make_persistent(dentry, inode);
> -		fsnotify_create(dir, dentry);
>  		if (fdentries)
>  			fdentries[i] = dentry; // borrowed
> -		simple_done_creating(dentry);
> +		simple_end_creating(dentry);
>  	}
>  	return 0;
>  }

For the NFSD hunks:

Acked-by: Chuck Lever <chuck.lever@oracle.com>


-- 
Chuck Lever

Re: [PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[Amir Goldstein]

On Mon, Mar 2, 2026 at 11:28 PM Chuck Lever <chuck.lever@oracle.com> wrote:
>
> On 3/2/26 1:37 PM, Amir Goldstein wrote:
> > Add simple_end_creating() helper which combines fsnotify_create/mkdir()
> > hook and simple_done_creating().
> >
> > Use the new helper to consolidate this pattern in several pseudo fs
> > which had open coded fsnotify_create/mkdir() hooks:
> > binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.
> >
> > For those filesystems, the paired fsnotify_delete() hook is already
> > inside the library helper simple_recursive_removal().
> >
> > Note that in debugfs_create_symlink(), the fsnotify hook was missing,
> > so the missing hook is fixed by this change.
> >
> > Signed-off-by: Amir Goldstein <amir73il@gmail.com>
>
> > diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> > index e9acd2cd602cb..6e600d52b66d0 100644
> > --- a/fs/nfsd/nfsctl.c
> > +++ b/fs/nfsd/nfsctl.c
> > @@ -17,7 +17,6 @@
> >  #include <linux/sunrpc/rpc_pipe_fs.h>
> >  #include <linux/sunrpc/svc.h>
> >  #include <linux/module.h>
> > -#include <linux/fsnotify.h>
> >  #include <linux/nfslocalio.h>
> >
> >  #include "idmap.h"
> > @@ -1146,8 +1145,7 @@ static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *nc
> >       }
> >       d_make_persistent(dentry, inode);
> >       inc_nlink(dir);
> > -     fsnotify_mkdir(dir, dentry);
> > -     simple_done_creating(dentry);
> > +     simple_end_creating(dentry);
> >       return dentry;  // borrowed
> >  }
> >
> > @@ -1178,8 +1176,7 @@ static void _nfsd_symlink(struct dentry *parent, const char *name,
> >       inode->i_size = strlen(content);
> >
> >       d_make_persistent(dentry, inode);
> > -     fsnotify_create(dir, dentry);
> > -     simple_done_creating(dentry);
> > +     simple_end_creating(dentry);
> >  }
> >  #else
> >  static inline void _nfsd_symlink(struct dentry *parent, const char *name,
> > @@ -1219,7 +1216,6 @@ static int nfsdfs_create_files(struct dentry *root,
> >                               struct nfsdfs_client *ncl,
> >                               struct dentry **fdentries)
> >  {
> > -     struct inode *dir = d_inode(root);
> >       struct dentry *dentry;
> >
> >       for (int i = 0; files->name && files->name[0]; i++, files++) {
> > @@ -1236,10 +1232,9 @@ static int nfsdfs_create_files(struct dentry *root,
> >               inode->i_fop = files->ops;
> >               inode->i_private = ncl;
> >               d_make_persistent(dentry, inode);
> > -             fsnotify_create(dir, dentry);
> >               if (fdentries)
> >                       fdentries[i] = dentry; // borrowed
> > -             simple_done_creating(dentry);
> > +             simple_end_creating(dentry);
> >       }
> >       return 0;
> >  }
>
> For the NFSD hunks:
>
> Acked-by: Chuck Lever <chuck.lever@oracle.com>

FWIW, you are technically also CCed for the sunrpc hunk ;)

BTW, forgot to CC Neil and mention this patch:
https://lore.kernel.org/linux-fsdevel/20260224222542.3458677-5-neilb@ownmail.net/

Since simple_done_creating() starts using end_creating()
so should simple_end_creating().
I will change that in v2 after waiting for more feedback on v1.

I don't want to get into naming discussion - I will just say that I wanted
to avoid renaming simple_done_creating() to avoid unneeded churn.

Thanks,
Amir.

Re: [PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[NeilBrown]

On Tue, 03 Mar 2026, Amir Goldstein wrote:
> On Mon, Mar 2, 2026 at 11:28 PM Chuck Lever <chuck.lever@oracle.com> wrote:
> >
> > On 3/2/26 1:37 PM, Amir Goldstein wrote:
> > > Add simple_end_creating() helper which combines fsnotify_create/mkdir()
> > > hook and simple_done_creating().
> > >
> > > Use the new helper to consolidate this pattern in several pseudo fs
> > > which had open coded fsnotify_create/mkdir() hooks:
> > > binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.
> > >
> > > For those filesystems, the paired fsnotify_delete() hook is already
> > > inside the library helper simple_recursive_removal().
> > >
> > > Note that in debugfs_create_symlink(), the fsnotify hook was missing,
> > > so the missing hook is fixed by this change.
> > >
> > > Signed-off-by: Amir Goldstein <amir73il@gmail.com>
> >
> > > diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> > > index e9acd2cd602cb..6e600d52b66d0 100644
> > > --- a/fs/nfsd/nfsctl.c
> > > +++ b/fs/nfsd/nfsctl.c
> > > @@ -17,7 +17,6 @@
> > >  #include <linux/sunrpc/rpc_pipe_fs.h>
> > >  #include <linux/sunrpc/svc.h>
> > >  #include <linux/module.h>
> > > -#include <linux/fsnotify.h>
> > >  #include <linux/nfslocalio.h>
> > >
> > >  #include "idmap.h"
> > > @@ -1146,8 +1145,7 @@ static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *nc
> > >       }
> > >       d_make_persistent(dentry, inode);
> > >       inc_nlink(dir);
> > > -     fsnotify_mkdir(dir, dentry);
> > > -     simple_done_creating(dentry);
> > > +     simple_end_creating(dentry);
> > >       return dentry;  // borrowed
> > >  }
> > >
> > > @@ -1178,8 +1176,7 @@ static void _nfsd_symlink(struct dentry *parent, const char *name,
> > >       inode->i_size = strlen(content);
> > >
> > >       d_make_persistent(dentry, inode);
> > > -     fsnotify_create(dir, dentry);
> > > -     simple_done_creating(dentry);
> > > +     simple_end_creating(dentry);
> > >  }
> > >  #else
> > >  static inline void _nfsd_symlink(struct dentry *parent, const char *name,
> > > @@ -1219,7 +1216,6 @@ static int nfsdfs_create_files(struct dentry *root,
> > >                               struct nfsdfs_client *ncl,
> > >                               struct dentry **fdentries)
> > >  {
> > > -     struct inode *dir = d_inode(root);
> > >       struct dentry *dentry;
> > >
> > >       for (int i = 0; files->name && files->name[0]; i++, files++) {
> > > @@ -1236,10 +1232,9 @@ static int nfsdfs_create_files(struct dentry *root,
> > >               inode->i_fop = files->ops;
> > >               inode->i_private = ncl;
> > >               d_make_persistent(dentry, inode);
> > > -             fsnotify_create(dir, dentry);
> > >               if (fdentries)
> > >                       fdentries[i] = dentry; // borrowed
> > > -             simple_done_creating(dentry);
> > > +             simple_end_creating(dentry);
> > >       }
> > >       return 0;
> > >  }
> >
> > For the NFSD hunks:
> >
> > Acked-by: Chuck Lever <chuck.lever@oracle.com>
> 
> FWIW, you are technically also CCed for the sunrpc hunk ;)
> 
> BTW, forgot to CC Neil and mention this patch:
> https://lore.kernel.org/linux-fsdevel/20260224222542.3458677-5-neilb@ownmail.net/
> 
> Since simple_done_creating() starts using end_creating()
> so should simple_end_creating().
> I will change that in v2 after waiting for more feedback on v1.
> 
> I don't want to get into naming discussion - I will just say that I wanted
> to avoid renaming simple_done_creating() to avoid unneeded churn.
> 
> Thanks,
> Amir.
> 

Thanks for the Cc.

Would there be a problem with doing the fs-notify for *every* caller of
simple_done_creating?
It does make sense to include the notify in the common code, but having
two different interfaces that only differ in the notification (and don't
contain some form of "notify" in their name) does seem like a recipe for
confusion.

Thanks,
NeilBrown

Re: [PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[Amir Goldstein]

On Tue, Mar 3, 2026 at 11:28 AM NeilBrown <neilb@ownmail.net> wrote:
>
> On Tue, 03 Mar 2026, Amir Goldstein wrote:
> > On Mon, Mar 2, 2026 at 11:28 PM Chuck Lever <chuck.lever@oracle.com> wrote:
> > >
> > > On 3/2/26 1:37 PM, Amir Goldstein wrote:
> > > > Add simple_end_creating() helper which combines fsnotify_create/mkdir()
> > > > hook and simple_done_creating().
> > > >
> > > > Use the new helper to consolidate this pattern in several pseudo fs
> > > > which had open coded fsnotify_create/mkdir() hooks:
> > > > binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.
> > > >
> > > > For those filesystems, the paired fsnotify_delete() hook is already
> > > > inside the library helper simple_recursive_removal().
> > > >
> > > > Note that in debugfs_create_symlink(), the fsnotify hook was missing,
> > > > so the missing hook is fixed by this change.
> > > >
> > > > Signed-off-by: Amir Goldstein <amir73il@gmail.com>
> > >
> > > > diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> > > > index e9acd2cd602cb..6e600d52b66d0 100644
> > > > --- a/fs/nfsd/nfsctl.c
> > > > +++ b/fs/nfsd/nfsctl.c
> > > > @@ -17,7 +17,6 @@
> > > >  #include <linux/sunrpc/rpc_pipe_fs.h>
> > > >  #include <linux/sunrpc/svc.h>
> > > >  #include <linux/module.h>
> > > > -#include <linux/fsnotify.h>
> > > >  #include <linux/nfslocalio.h>
> > > >
> > > >  #include "idmap.h"
> > > > @@ -1146,8 +1145,7 @@ static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *nc
> > > >       }
> > > >       d_make_persistent(dentry, inode);
> > > >       inc_nlink(dir);
> > > > -     fsnotify_mkdir(dir, dentry);
> > > > -     simple_done_creating(dentry);
> > > > +     simple_end_creating(dentry);
> > > >       return dentry;  // borrowed
> > > >  }
> > > >
> > > > @@ -1178,8 +1176,7 @@ static void _nfsd_symlink(struct dentry *parent, const char *name,
> > > >       inode->i_size = strlen(content);
> > > >
> > > >       d_make_persistent(dentry, inode);
> > > > -     fsnotify_create(dir, dentry);
> > > > -     simple_done_creating(dentry);
> > > > +     simple_end_creating(dentry);
> > > >  }
> > > >  #else
> > > >  static inline void _nfsd_symlink(struct dentry *parent, const char *name,
> > > > @@ -1219,7 +1216,6 @@ static int nfsdfs_create_files(struct dentry *root,
> > > >                               struct nfsdfs_client *ncl,
> > > >                               struct dentry **fdentries)
> > > >  {
> > > > -     struct inode *dir = d_inode(root);
> > > >       struct dentry *dentry;
> > > >
> > > >       for (int i = 0; files->name && files->name[0]; i++, files++) {
> > > > @@ -1236,10 +1232,9 @@ static int nfsdfs_create_files(struct dentry *root,
> > > >               inode->i_fop = files->ops;
> > > >               inode->i_private = ncl;
> > > >               d_make_persistent(dentry, inode);
> > > > -             fsnotify_create(dir, dentry);
> > > >               if (fdentries)
> > > >                       fdentries[i] = dentry; // borrowed
> > > > -             simple_done_creating(dentry);
> > > > +             simple_end_creating(dentry);
> > > >       }
> > > >       return 0;
> > > >  }
> > >
> > > For the NFSD hunks:
> > >
> > > Acked-by: Chuck Lever <chuck.lever@oracle.com>
> >
> > FWIW, you are technically also CCed for the sunrpc hunk ;)
> >
> > BTW, forgot to CC Neil and mention this patch:
> > https://lore.kernel.org/linux-fsdevel/20260224222542.3458677-5-neilb@ownmail.net/
> >
> > Since simple_done_creating() starts using end_creating()
> > so should simple_end_creating().
> > I will change that in v2 after waiting for more feedback on v1.
> >
> > I don't want to get into naming discussion - I will just say that I wanted
> > to avoid renaming simple_done_creating() to avoid unneeded churn.
> >
> > Thanks,
> > Amir.
> >
>
> Thanks for the Cc.
>
> Would there be a problem with doing the fs-notify for *every* caller of
> simple_done_creating?

1. simple_done_creating() is also called for the failed case, where the
fsnotify hook is not desired. I am not sure if failure is always equivalent
to negative child dentry.

2. I assume you meant the simple_done_creating() calls in other fs that
do not have fsnotify hooks in current code. This is a valid point.
I am hesitant to add the FS_CREATE events for *all* the pseudo fs
dentry creations.
Specifically, I think we need not send the events for init/populate of fs.
It's worth nothing that some of those fs do already send FS_DELETE
events via simple_recursive_removal().

> It does make sense to include the notify in the common code, but having
> two different interfaces that only differ in the notification (and don't
> contain some form of "notify" in their name) does seem like a recipe for
> confusion.

Right. I could make it simple_end_creating_notify().

Now where does this sound familiar from?

Oh yeah, my old fsnotify path hooks series [1] :)

Cheers,
Amir.

[1]  https://lore.kernel.org/linux-fsdevel/CAOQ4uxi-UhF=6eaxhybvdBX-L5qYx_uEuu-eCiiUzJPvz2U8aw@mail.gmail.com/

Re: [PATCH 0/2] fsnotify hooks consolidation

[Jeff Layton]

On Mon, 2026-03-02 at 19:37 +0100, Amir Goldstein wrote:
> Jan,
> 
> I've found this opportunity to reduce the amount of sprinkled
> fsnotify hooks.
> 
> There are still a few fsnotify hooks sprinkled in some pseudo fs,
> but all those removed by this series are obvious boiler plate code
> and for most of these fs, this removes all of the custom hooks.
> 
> I could send a series to convert each fs with its own patch,
> but to me that seems a bit unnecessary.
> 
> WDYT?
> Amir.
> 
> Amir Goldstein (2):
>   fsnotify: make fsnotify_create() agnostic to file/dir
>   fs: use simple_end_creating helper to consolidate fsnotify hooks
> 
>  drivers/android/binder/rust_binderfs.c | 11 +++--------
>  drivers/android/binderfs.c             | 10 +++-------
>  fs/debugfs/inode.c                     |  5 +----
>  fs/libfs.c                             | 14 ++++++++++++++
>  fs/nfsd/nfsctl.c                       | 11 +++--------
>  fs/tracefs/event_inode.c               |  2 --
>  fs/tracefs/inode.c                     |  5 +----
>  include/linux/fs.h                     |  1 +
>  include/linux/fsnotify.h               |  8 +++++++-
>  net/sunrpc/rpc_pipe.c                  | 10 +++-------
>  10 files changed, 36 insertions(+), 41 deletions(-)

Conceptually, this all seems fine.

My only gripe is that there is nothing mnemonic about the names
simple_done_creating() and simple_end_creating(). Since the only
difference are the notification bits, maybe
simple_done_creating_notify() is a better name?

Reviewed-by: Jeff Layton <jlayton@kernel.org>

Re: [PATCH 0/2] fsnotify hooks consolidation

[Amir Goldstein]

On Tue, Mar 3, 2026 at 1:45 PM Jeff Layton <jlayton@kernel.org> wrote:
>
> On Mon, 2026-03-02 at 19:37 +0100, Amir Goldstein wrote:
> > Jan,
> >
> > I've found this opportunity to reduce the amount of sprinkled
> > fsnotify hooks.
> >
> > There are still a few fsnotify hooks sprinkled in some pseudo fs,
> > but all those removed by this series are obvious boiler plate code
> > and for most of these fs, this removes all of the custom hooks.
> >
> > I could send a series to convert each fs with its own patch,
> > but to me that seems a bit unnecessary.
> >
> > WDYT?
> > Amir.
> >
> > Amir Goldstein (2):
> >   fsnotify: make fsnotify_create() agnostic to file/dir
> >   fs: use simple_end_creating helper to consolidate fsnotify hooks
> >
> >  drivers/android/binder/rust_binderfs.c | 11 +++--------
> >  drivers/android/binderfs.c             | 10 +++-------
> >  fs/debugfs/inode.c                     |  5 +----
> >  fs/libfs.c                             | 14 ++++++++++++++
> >  fs/nfsd/nfsctl.c                       | 11 +++--------
> >  fs/tracefs/event_inode.c               |  2 --
> >  fs/tracefs/inode.c                     |  5 +----
> >  include/linux/fs.h                     |  1 +
> >  include/linux/fsnotify.h               |  8 +++++++-
> >  net/sunrpc/rpc_pipe.c                  | 10 +++-------
> >  10 files changed, 36 insertions(+), 41 deletions(-)
>
> Conceptually, this all seems fine.
>
> My only gripe is that there is nothing mnemonic about the names
> simple_done_creating() and simple_end_creating(). Since the only
> difference are the notification bits, maybe
> simple_done_creating_notify() is a better name?

I will go with simple_end_creating_notify() because what it does is:

void simple_end_creating_notify(struct dentry *child)
{
        fsnotify_create(child->d_parent->d_inode, child);
        end_creating(child);
}

>
> Reviewed-by: Jeff Layton <jlayton@kernel.org>

Thanks,
Amir.

Re: [PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[Chuck Lever]

On 3/3/26 5:05 AM, Amir Goldstein wrote:
> On Mon, Mar 2, 2026 at 11:28 PM Chuck Lever <chuck.lever@oracle.com> wrote:
>>
>> On 3/2/26 1:37 PM, Amir Goldstein wrote:
>>> Add simple_end_creating() helper which combines fsnotify_create/mkdir()
>>> hook and simple_done_creating().
>>>
>>> Use the new helper to consolidate this pattern in several pseudo fs
>>> which had open coded fsnotify_create/mkdir() hooks:
>>> binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.
>>>
>>> For those filesystems, the paired fsnotify_delete() hook is already
>>> inside the library helper simple_recursive_removal().
>>>
>>> Note that in debugfs_create_symlink(), the fsnotify hook was missing,
>>> so the missing hook is fixed by this change.
>>>
>>> Signed-off-by: Amir Goldstein <amir73il@gmail.com>
>>
>>> diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
>>> index e9acd2cd602cb..6e600d52b66d0 100644
>>> --- a/fs/nfsd/nfsctl.c
>>> +++ b/fs/nfsd/nfsctl.c
>>> @@ -17,7 +17,6 @@
>>>  #include <linux/sunrpc/rpc_pipe_fs.h>
>>>  #include <linux/sunrpc/svc.h>
>>>  #include <linux/module.h>
>>> -#include <linux/fsnotify.h>
>>>  #include <linux/nfslocalio.h>
>>>
>>>  #include "idmap.h"
>>> @@ -1146,8 +1145,7 @@ static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *nc
>>>       }
>>>       d_make_persistent(dentry, inode);
>>>       inc_nlink(dir);
>>> -     fsnotify_mkdir(dir, dentry);
>>> -     simple_done_creating(dentry);
>>> +     simple_end_creating(dentry);
>>>       return dentry;  // borrowed
>>>  }
>>>
>>> @@ -1178,8 +1176,7 @@ static void _nfsd_symlink(struct dentry *parent, const char *name,
>>>       inode->i_size = strlen(content);
>>>
>>>       d_make_persistent(dentry, inode);
>>> -     fsnotify_create(dir, dentry);
>>> -     simple_done_creating(dentry);
>>> +     simple_end_creating(dentry);
>>>  }
>>>  #else
>>>  static inline void _nfsd_symlink(struct dentry *parent, const char *name,
>>> @@ -1219,7 +1216,6 @@ static int nfsdfs_create_files(struct dentry *root,
>>>                               struct nfsdfs_client *ncl,
>>>                               struct dentry **fdentries)
>>>  {
>>> -     struct inode *dir = d_inode(root);
>>>       struct dentry *dentry;
>>>
>>>       for (int i = 0; files->name && files->name[0]; i++, files++) {
>>> @@ -1236,10 +1232,9 @@ static int nfsdfs_create_files(struct dentry *root,
>>>               inode->i_fop = files->ops;
>>>               inode->i_private = ncl;
>>>               d_make_persistent(dentry, inode);
>>> -             fsnotify_create(dir, dentry);
>>>               if (fdentries)
>>>                       fdentries[i] = dentry; // borrowed
>>> -             simple_done_creating(dentry);
>>> +             simple_end_creating(dentry);
>>>       }
>>>       return 0;
>>>  }
>>
>> For the NFSD hunks:
>>
>> Acked-by: Chuck Lever <chuck.lever@oracle.com>
> 
> FWIW, you are technically also CCed for the sunrpc hunk ;)
rpc_pipe.c is client-side. On the next round, Cc: Trond, Anna, and
linux-nfs@ .



-- 
Chuck Lever

Re: [PATCH 0/2] fsnotify hooks consolidation

[Christoph Hellwig]

On Tue, Mar 03, 2026 at 02:18:47PM +0100, Amir Goldstein wrote:
> > simple_done_creating_notify() is a better name?
> 
> I will go with simple_end_creating_notify() because what it does is:

Shouldn't the notify case be the default one and the nonotify one
stand out with a prefix like _nonotify?  I.e., steer people to the
more useful one by default.

Re: [PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[Alice Ryhl]

On Mon, Mar 02, 2026 at 07:37:41PM +0100, Amir Goldstein wrote:
> Add simple_end_creating() helper which combines fsnotify_create/mkdir()
> hook and simple_done_creating().
> 
> Use the new helper to consolidate this pattern in several pseudo fs
> which had open coded fsnotify_create/mkdir() hooks:
> binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.
> 
> For those filesystems, the paired fsnotify_delete() hook is already
> inside the library helper simple_recursive_removal().
> 
> Note that in debugfs_create_symlink(), the fsnotify hook was missing,
> so the missing hook is fixed by this change.
> 
> Signed-off-by: Amir Goldstein <amir73il@gmail.com>

Acked-by: Alice Ryhl <aliceryhl@google.com> # binderfs + rust_binderfs

Re: [PATCH 0/2] fsnotify hooks consolidation

[Amir Goldstein]

On Tue, Mar 3, 2026 at 4:00 PM Christoph Hellwig <hch@infradead.org> wrote:
>
> On Tue, Mar 03, 2026 at 02:18:47PM +0100, Amir Goldstein wrote:
> > > simple_done_creating_notify() is a better name?
> >
> > I will go with simple_end_creating_notify() because what it does is:
>
> Shouldn't the notify case be the default one and the nonotify one
> stand out with a prefix like _nonotify?  I.e., steer people to the
> more useful one by default.
>

It may sound like that, but in reality, it is quite hard to differentiate
in pseduo fs code between creations that are auto initialized
at mount time and creations that are user triggerred.

Only the user triggerred ones should be notified and also we only
have a handful of pseduo fs who opted in so far, so I think in this case
staying with opt in is the right way, but open to hear what folks think.

Thanks,
Amir.

Re: [PATCH 0/2] fsnotify hooks consolidation

[Christoph Hellwig]

On Tue, Mar 03, 2026 at 04:05:09PM +0100, Amir Goldstein wrote:
> It may sound like that, but in reality, it is quite hard to differentiate
> in pseduo fs code between creations that are auto initialized
> at mount time and creations that are user triggerred.
> 
> Only the user triggerred ones should be notified and also we only
> have a handful of pseduo fs who opted in so far, so I think in this case
> staying with opt in is the right way, but open to hear what folks think.

Then maybe encode that in the name with _user and _kernel postfix, and
also document it very well?

Re: [PATCH 0/2] fsnotify hooks consolidation

[Amir Goldstein]

On Tue, Mar 3, 2026 at 4:07 PM Christoph Hellwig <hch@infradead.org> wrote:
>
> On Tue, Mar 03, 2026 at 04:05:09PM +0100, Amir Goldstein wrote:
> > It may sound like that, but in reality, it is quite hard to differentiate
> > in pseduo fs code between creations that are auto initialized
> > at mount time and creations that are user triggerred.
> >
> > Only the user triggerred ones should be notified and also we only
> > have a handful of pseduo fs who opted in so far, so I think in this case
> > staying with opt in is the right way, but open to hear what folks think.
>
> Then maybe encode that in the name with _user and _kernel postfix, and
> also document it very well?
>

Sure but _user vs. _kernel is maybe not the best terminology
it is alreayd the case that dentries could be created from _user API
via vfs syscalls.

The code changed by this patch is specifically code that creates
entries not via vfs syscalls, so vfs code refers to these changes
and _kernel changes.

However, they are trigerred by user writing something to some file etc.

Calling the helper by what it does keeps things simple -
if author wants a notification, author should use this helper.

I honestly don't know why this fs opted for notifications
and other fs did not - I assume it's just based on actual real
world use cases, so I'd rather not change those decisions
with this mechanical cleanup task or with an "alluring" helper name.

Thanks,
Amir.

Re: [PATCH 2/2] fs: use simple_end_creating helper to consolidate fsnotify hooks

[NeilBrown]

On Tue, 03 Mar 2026, Amir Goldstein wrote:
> On Tue, Mar 3, 2026 at 11:28 AM NeilBrown <neilb@ownmail.net> wrote:
> >
> > On Tue, 03 Mar 2026, Amir Goldstein wrote:
> > > On Mon, Mar 2, 2026 at 11:28 PM Chuck Lever <chuck.lever@oracle.com> wrote:
> > > >
> > > > On 3/2/26 1:37 PM, Amir Goldstein wrote:
> > > > > Add simple_end_creating() helper which combines fsnotify_create/mkdir()
> > > > > hook and simple_done_creating().
> > > > >
> > > > > Use the new helper to consolidate this pattern in several pseudo fs
> > > > > which had open coded fsnotify_create/mkdir() hooks:
> > > > > binderfs, debugfs, nfsctl, tracefs, rpc_pipefs.
> > > > >
> > > > > For those filesystems, the paired fsnotify_delete() hook is already
> > > > > inside the library helper simple_recursive_removal().
> > > > >
> > > > > Note that in debugfs_create_symlink(), the fsnotify hook was missing,
> > > > > so the missing hook is fixed by this change.
> > > > >
> > > > > Signed-off-by: Amir Goldstein <amir73il@gmail.com>
> > > >
> > > > > diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> > > > > index e9acd2cd602cb..6e600d52b66d0 100644
> > > > > --- a/fs/nfsd/nfsctl.c
> > > > > +++ b/fs/nfsd/nfsctl.c
> > > > > @@ -17,7 +17,6 @@
> > > > >  #include <linux/sunrpc/rpc_pipe_fs.h>
> > > > >  #include <linux/sunrpc/svc.h>
> > > > >  #include <linux/module.h>
> > > > > -#include <linux/fsnotify.h>
> > > > >  #include <linux/nfslocalio.h>
> > > > >
> > > > >  #include "idmap.h"
> > > > > @@ -1146,8 +1145,7 @@ static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *nc
> > > > >       }
> > > > >       d_make_persistent(dentry, inode);
> > > > >       inc_nlink(dir);
> > > > > -     fsnotify_mkdir(dir, dentry);
> > > > > -     simple_done_creating(dentry);
> > > > > +     simple_end_creating(dentry);
> > > > >       return dentry;  // borrowed
> > > > >  }
> > > > >
> > > > > @@ -1178,8 +1176,7 @@ static void _nfsd_symlink(struct dentry *parent, const char *name,
> > > > >       inode->i_size = strlen(content);
> > > > >
> > > > >       d_make_persistent(dentry, inode);
> > > > > -     fsnotify_create(dir, dentry);
> > > > > -     simple_done_creating(dentry);
> > > > > +     simple_end_creating(dentry);
> > > > >  }
> > > > >  #else
> > > > >  static inline void _nfsd_symlink(struct dentry *parent, const char *name,
> > > > > @@ -1219,7 +1216,6 @@ static int nfsdfs_create_files(struct dentry *root,
> > > > >                               struct nfsdfs_client *ncl,
> > > > >                               struct dentry **fdentries)
> > > > >  {
> > > > > -     struct inode *dir = d_inode(root);
> > > > >       struct dentry *dentry;
> > > > >
> > > > >       for (int i = 0; files->name && files->name[0]; i++, files++) {
> > > > > @@ -1236,10 +1232,9 @@ static int nfsdfs_create_files(struct dentry *root,
> > > > >               inode->i_fop = files->ops;
> > > > >               inode->i_private = ncl;
> > > > >               d_make_persistent(dentry, inode);
> > > > > -             fsnotify_create(dir, dentry);
> > > > >               if (fdentries)
> > > > >                       fdentries[i] = dentry; // borrowed
> > > > > -             simple_done_creating(dentry);
> > > > > +             simple_end_creating(dentry);
> > > > >       }
> > > > >       return 0;
> > > > >  }
> > > >
> > > > For the NFSD hunks:
> > > >
> > > > Acked-by: Chuck Lever <chuck.lever@oracle.com>
> > >
> > > FWIW, you are technically also CCed for the sunrpc hunk ;)
> > >
> > > BTW, forgot to CC Neil and mention this patch:
> > > https://lore.kernel.org/linux-fsdevel/20260224222542.3458677-5-neilb@ownmail.net/
> > >
> > > Since simple_done_creating() starts using end_creating()
> > > so should simple_end_creating().
> > > I will change that in v2 after waiting for more feedback on v1.
> > >
> > > I don't want to get into naming discussion - I will just say that I wanted
> > > to avoid renaming simple_done_creating() to avoid unneeded churn.
> > >
> > > Thanks,
> > > Amir.
> > >
> >
> > Thanks for the Cc.
> >
> > Would there be a problem with doing the fs-notify for *every* caller of
> > simple_done_creating?
> 
> 1. simple_done_creating() is also called for the failed case, where the
> fsnotify hook is not desired. I am not sure if failure is always equivalent
> to negative child dentry.

If the dentry is still hashed and positive, then it is hard to see that
could represent a failure.  If it is unhashed or negative, then ...  it
could possibly be a success as some filesystems create the object but
leave it to a subsequent ->lookup to add it to the dcache.  I doubt such
filesystems would be using simple_* though.

> 
> 2. I assume you meant the simple_done_creating() calls in other fs that
> do not have fsnotify hooks in current code. This is a valid point.
> I am hesitant to add the FS_CREATE events for *all* the pseudo fs
> dentry creations.
> Specifically, I think we need not send the events for init/populate of fs.

fsnotify_create hooks are effectively no-op if no "marks" have be
registered on sb or parent inode.  During init/populate there would be
no marks.

I think filesystems should have a good reason for opting out of
fsnotify, and it should be harder to opt-out than to opt-in.

> It's worth nothing that some of those fs do already send FS_DELETE
> events via simple_recursive_removal().
> \x16
> > It does make sense to include the notify in the common code, but having
> > two different interfaces that only differ in the notification (and don't
> > contain some form of "notify" in their name) does seem like a recipe for
> > confusion.
> 
> Right. I could make it simple_end_creating_notify().
> 
> Now where does this sound familiar from?
> 
> Oh yeah, my old fsnotify path hooks series [1] :)

Everything old is new again :-)

Thanks,
NeilBrown

> 
> Cheers,
> Amir.
> 
> [1]  https://lore.kernel.org/linux-fsdevel/CAOQ4uxi-UhF=6eaxhybvdBX-L5qYx_uEuu-eCiiUzJPvz2U8aw@mail.gmail.com/
>