tracing: Have mkdir and rmdir be part of tracefs

The tracing "instances" directory can create sub tracing buffers
with mkdir, and remove them with rmdir. As a mkdir will also create
all the files and directories that control the sub buffer the inode
mutexes need to be released before this is done, to avoid deadlocks.
It is better to let the tracing system unlock the inode mutexes before
calling the functions that create the files within the new directory
(or deletes the files from the one being destroyed).

Now that tracing has been converted over to tracefs, the tracefs file
system can be modified to accommodate this feature. It still releases
the locks, but the filesystem itself can take care of the ugly
business and let the user just do what it needs.

The tracing system now attaches a descriptor to the directory dentry
that can have userspace create or remove sub directories. If this
descriptor does not exist for a dentry, then that dentry can not be
used to create other directories. This descriptor holds a mkdir and
rmdir method that only takes a character string as an argument.

The tracefs file system will first make a copy of the dentry name
before releasing the locks. Then it will pass the copied name to the
methods. It is up to the tracing system that supplied the methods to
handle races with duplicate names and such as all the inode mutexes
would be released when the functions are called.

Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

fs/tracefs/inode.c

@@ -50,6 +50,84 @@ static const struct file_operations tracefs_file_operations = {
 	.llseek =	noop_llseek,
 };
 
+static struct tracefs_dir_ops {
+	int (*mkdir)(const char *name);
+	int (*rmdir)(const char *name);
+} tracefs_ops;
+
+static char *get_dname(struct dentry *dentry)
+{
+	const char *dname;
+	char *name;
+	int len = dentry->d_name.len;
+
+	dname = dentry->d_name.name;
+	name = kmalloc(len + 1, GFP_KERNEL);
+	if (!name)
+		return NULL;
+	memcpy(name, dname, len);
+	name[len] = 0;
+	return name;
+}
+
+static int tracefs_syscall_mkdir(struct inode *inode, struct dentry *dentry, umode_t mode)
+{
+	char *name;
+	int ret;
+
+	name = get_dname(dentry);
+	if (!name)
+		return -ENOMEM;
+
+	/*
+	 * The mkdir call can call the generic functions that create
+	 * the files within the tracefs system. It is up to the individual
+	 * mkdir routine to handle races.
+	 */
+	mutex_unlock(&inode->i_mutex);
+	ret = tracefs_ops.mkdir(name);
+	mutex_lock(&inode->i_mutex);
+
+	kfree(name);
+
+	return ret;
+}
+
+static int tracefs_syscall_rmdir(struct inode *inode, struct dentry *dentry)
+{
+	char *name;
+	int ret;
+
+	name = get_dname(dentry);
+	if (!name)
+		return -ENOMEM;
+
+	/*
+	 * The rmdir call can call the generic functions that create
+	 * the files within the tracefs system. It is up to the individual
+	 * rmdir routine to handle races.
+	 * This time we need to unlock not only the parent (inode) but
+	 * also the directory that is being deleted.
+	 */
+	mutex_unlock(&inode->i_mutex);
+	mutex_unlock(&dentry->d_inode->i_mutex);
+
+	ret = tracefs_ops.rmdir(name);
+
+	mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
+	mutex_lock(&dentry->d_inode->i_mutex);
+
+	kfree(name);
+
+	return ret;
+}
+
+static const struct inode_operations tracefs_dir_inode_operations = {
+	.lookup		= simple_lookup,
+	.mkdir		= tracefs_syscall_mkdir,
+	.rmdir		= tracefs_syscall_rmdir,
+};
+
 static struct inode *tracefs_get_inode(struct super_block *sb)
 {
 	struct inode *inode = new_inode(sb);
@@ -334,6 +412,31 @@ struct dentry *tracefs_create_file(const char *name, umode_t mode,
 	return end_creating(dentry);
 }
 
+static struct dentry *__create_dir(const char *name, struct dentry *parent,
+				   const struct inode_operations *ops)
+{
+	struct dentry *dentry = start_creating(name, parent);
+	struct inode *inode;
+
+	if (IS_ERR(dentry))
+		return NULL;
+
+	inode = tracefs_get_inode(dentry->d_sb);
+	if (unlikely(!inode))
+		return failed_creating(dentry);
+
+	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUGO | S_IXUGO;
+	inode->i_op = ops;
+	inode->i_fop = &simple_dir_operations;
+
+	/* directory inodes start off with i_nlink == 2 (for "." entry) */
+	inc_nlink(inode);
+	d_instantiate(dentry, inode);
+	inc_nlink(dentry->d_parent->d_inode);
+	fsnotify_mkdir(dentry->d_parent->d_inode, dentry);
+	return end_creating(dentry);
+}
+
 /**
  * tracefs_create_dir - create a directory in the tracefs filesystem
  * @name: a pointer to a string containing the name of the directory to
@@ -353,26 +456,44 @@ struct dentry *tracefs_create_file(const char *name, umode_t mode,
  */
 struct dentry *tracefs_create_dir(const char *name, struct dentry *parent)
 {
-	struct dentry *dentry = start_creating(name, parent);
+	return __create_dir(name, parent, &simple_dir_inode_operations);
-	struct inode *inode;
+}
 
-	if (IS_ERR(dentry))
+/**
+ * tracefs_create_instance_dir - create the tracing instances directory
+ * @name: The name of the instances directory to create
+ * @parent: The parent directory that the instances directory will exist
+ * @mkdir: The function to call when a mkdir is performed.
+ * @rmdir: The function to call when a rmdir is performed.
+ *
+ * Only one instances directory is allowed.
+ *
+ * The instances directory is special as it allows for mkdir and rmdir to
+ * to be done by userspace. When a mkdir or rmdir is performed, the inode
+ * locks are released and the methhods passed in (@mkdir and @rmdir) are
+ * called without locks and with the name of the directory being created
+ * within the instances directory.
+ *
+ * Returns the dentry of the instances directory.
+ */
+struct dentry *tracefs_create_instance_dir(const char *name, struct dentry *parent,
+					  int (*mkdir)(const char *name),
+					  int (*rmdir)(const char *name))
+{
+	struct dentry *dentry;
+
+	/* Only allow one instance of the instances directory. */
+	if (WARN_ON(tracefs_ops.mkdir || tracefs_ops.rmdir))
 		return NULL;
 
-	inode = tracefs_get_inode(dentry->d_sb);
+	dentry = __create_dir(name, parent, &tracefs_dir_inode_operations);
-	if (unlikely(!inode))
+	if (!dentry)
-		return failed_creating(dentry);
+		return NULL;
 
-	inode->i_mode = S_IFDIR | S_IRWXU | S_IRUGO | S_IXUGO;
+	tracefs_ops.mkdir = mkdir;
-	inode->i_op = &simple_dir_inode_operations;
+	tracefs_ops.rmdir = rmdir;
-	inode->i_fop = &simple_dir_operations;
 
-	/* directory inodes start off with i_nlink == 2 (for "." entry) */
+	return dentry;
-	inc_nlink(inode);
-	d_instantiate(dentry, inode);
-	inc_nlink(dentry->d_parent->d_inode);
-	fsnotify_mkdir(dentry->d_parent->d_inode, dentry);
-	return end_creating(dentry);
 }
 
 static inline int tracefs_positive(struct dentry *dentry)

include/linux/tracefs.h

@@ -34,6 +34,10 @@ struct dentry *tracefs_create_dir(const char *name, struct dentry *parent);
 void tracefs_remove(struct dentry *dentry);
 void tracefs_remove_recursive(struct dentry *dentry);
 
+struct dentry *tracefs_create_instance_dir(const char *name, struct dentry *parent,
+					   int (*mkdir)(const char *name),
+					   int (*rmdir)(const char *name));
+
 bool tracefs_initialized(void);
 
 #endif /* CONFIG_TRACING */

kernel/trace/trace.c

@@ -6292,7 +6292,7 @@ static void free_trace_buffers(struct trace_array *tr)
 #endif
 }
 
-static int new_instance_create(const char *name)
+static int instance_mkdir(const char *name)
 {
 	struct trace_array *tr;
 	int ret;
@@ -6362,7 +6362,7 @@ static int new_instance_create(const char *name)
 
 }
 
-static int instance_delete(const char *name)
+static int instance_rmdir(const char *name)
 {
 	struct trace_array *tr;
 	int found = 0;
@@ -6403,78 +6403,13 @@ static int instance_delete(const char *name)
 	return ret;
 }
 
-static int instance_mkdir (struct inode *inode, struct dentry *dentry, umode_t mode)
-{
-	struct dentry *parent;
-	int ret;
-
-	/* Paranoid: Make sure the parent is the "instances" directory */
-	parent = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
-	if (WARN_ON_ONCE(parent != trace_instance_dir))
-		return -ENOENT;
-
-	/*
-	 * The inode mutex is locked, but tracefs_create_dir() will also
-	 * take the mutex. As the instances directory can not be destroyed
-	 * or changed in any other way, it is safe to unlock it, and
-	 * let the dentry try. If two users try to make the same dir at
-	 * the same time, then the new_instance_create() will determine the
-	 * winner.
-	 */
-	mutex_unlock(&inode->i_mutex);
-
-	ret = new_instance_create(dentry->d_iname);
-
-	mutex_lock(&inode->i_mutex);
-
-	return ret;
-}
-
-static int instance_rmdir(struct inode *inode, struct dentry *dentry)
-{
-	struct dentry *parent;
-	int ret;
-
-	/* Paranoid: Make sure the parent is the "instances" directory */
-	parent = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
-	if (WARN_ON_ONCE(parent != trace_instance_dir))
-		return -ENOENT;
-
-	/* The caller did a dget() on dentry */
-	mutex_unlock(&dentry->d_inode->i_mutex);
-
-	/*
-	 * The inode mutex is locked, but tracefs_create_dir() will also
-	 * take the mutex. As the instances directory can not be destroyed
-	 * or changed in any other way, it is safe to unlock it, and
-	 * let the dentry try. If two users try to make the same dir at
-	 * the same time, then the instance_delete() will determine the
-	 * winner.
-	 */
-	mutex_unlock(&inode->i_mutex);
-
-	ret = instance_delete(dentry->d_iname);
-
-	mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
-	mutex_lock(&dentry->d_inode->i_mutex);
-
-	return ret;
-}
-
-static const struct inode_operations instance_dir_inode_operations = {
-	.lookup		= simple_lookup,
-	.mkdir		= instance_mkdir,
-	.rmdir		= instance_rmdir,
-};
-
 static __init void create_trace_instances(struct dentry *d_tracer)
 {
-	trace_instance_dir = tracefs_create_dir("instances", d_tracer);
+	trace_instance_dir = tracefs_create_instance_dir("instances", d_tracer,
+							 instance_mkdir,
+							 instance_rmdir);
 	if (WARN_ON(!trace_instance_dir))
 		return;
-
-	/* Hijack the dir inode operations, to allow mkdir */
-	trace_instance_dir->d_inode->i_op = &instance_dir_inode_operations;
 }
 
 static void