[linux-2.6-block.git] / kernel / bpf / inode.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * Minimal file system backend for holding eBPF maps and programs,
 * used by bpf(2) object pinning.
 *
 * Authors:
 *
 *	Daniel Borkmann <daniel@iogearbox.net>
 */

#include <linux/init.h>
#include <linux/magic.h>
#include <linux/major.h>
#include <linux/mount.h>
#include <linux/namei.h>
#include <linux/fs.h>
#include <linux/kdev_t.h>
#include <linux/parser.h>
#include <linux/filter.h>
#include <linux/bpf.h>
#include <linux/bpf_trace.h>

enum bpf_type {
	BPF_TYPE_UNSPEC	= 0,
	BPF_TYPE_PROG,
	BPF_TYPE_MAP,
};

static void *bpf_any_get(void *raw, enum bpf_type type)
{
	switch (type) {
	case BPF_TYPE_PROG:
		raw = bpf_prog_inc(raw);
		break;
	case BPF_TYPE_MAP:
		raw = bpf_map_inc(raw, true);
		break;
	default:
		WARN_ON_ONCE(1);
		break;
	}

	return raw;
}

static void bpf_any_put(void *raw, enum bpf_type type)
{
	switch (type) {
	case BPF_TYPE_PROG:
		bpf_prog_put(raw);
		break;
	case BPF_TYPE_MAP:
		bpf_map_put_with_uref(raw);
		break;
	default:
		WARN_ON_ONCE(1);
		break;
	}
}

static void *bpf_fd_probe_obj(u32 ufd, enum bpf_type *type)
{
	void *raw;

	*type = BPF_TYPE_MAP;
	raw = bpf_map_get_with_uref(ufd);
	if (IS_ERR(raw)) {
		*type = BPF_TYPE_PROG;
		raw = bpf_prog_get(ufd);
	}

	return raw;
}

static const struct inode_operations bpf_dir_iops;

static const struct inode_operations bpf_prog_iops = { };
static const struct inode_operations bpf_map_iops  = { };

static struct inode *bpf_get_inode(struct super_block *sb,
				   const struct inode *dir,
				   umode_t mode)
{
	struct inode *inode;

	switch (mode & S_IFMT) {
	case S_IFDIR:
	case S_IFREG:
	case S_IFLNK:
		break;
	default:
		return ERR_PTR(-EINVAL);
	}

	inode = new_inode(sb);
	if (!inode)
		return ERR_PTR(-ENOSPC);

	inode->i_ino = get_next_ino();
	inode->i_atime = current_time(inode);
	inode->i_mtime = inode->i_atime;
	inode->i_ctime = inode->i_atime;

	inode_init_owner(inode, dir, mode);

	return inode;
}

static int bpf_inode_type(const struct inode *inode, enum bpf_type *type)
{
	*type = BPF_TYPE_UNSPEC;
	if (inode->i_op == &bpf_prog_iops)
		*type = BPF_TYPE_PROG;
	else if (inode->i_op == &bpf_map_iops)
		*type = BPF_TYPE_MAP;
	else
		return -EACCES;

	return 0;
}

static void bpf_dentry_finalize(struct dentry *dentry, struct inode *inode,
				struct inode *dir)
{
	d_instantiate(dentry, inode);
	dget(dentry);

	dir->i_mtime = current_time(dir);
	dir->i_ctime = dir->i_mtime;
}

static int bpf_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
{
	struct inode *inode;

	inode = bpf_get_inode(dir->i_sb, dir, mode | S_IFDIR);
	if (IS_ERR(inode))
		return PTR_ERR(inode);

	inode->i_op = &bpf_dir_iops;
	inode->i_fop = &simple_dir_operations;

	inc_nlink(inode);
	inc_nlink(dir);

	bpf_dentry_finalize(dentry, inode, dir);
	return 0;
}

struct map_iter {
	void *key;
	bool done;
};

static struct map_iter *map_iter(struct seq_file *m)
{
	return m->private;
}

static struct bpf_map *seq_file_to_map(struct seq_file *m)
{
	return file_inode(m->file)->i_private;
}

static void map_iter_free(struct map_iter *iter)
{
	if (iter) {
		kfree(iter->key);
		kfree(iter);
	}
}

static struct map_iter *map_iter_alloc(struct bpf_map *map)
{
	struct map_iter *iter;

	iter = kzalloc(sizeof(*iter), GFP_KERNEL | __GFP_NOWARN);
	if (!iter)
		goto error;

	iter->key = kzalloc(map->key_size, GFP_KERNEL | __GFP_NOWARN);
	if (!iter->key)
		goto error;

	return iter;

error:
	map_iter_free(iter);
	return NULL;
}

static void *map_seq_next(struct seq_file *m, void *v, loff_t *pos)
{
	struct bpf_map *map = seq_file_to_map(m);
	void *key = map_iter(m)->key;
	void *prev_key;

	if (map_iter(m)->done)
		return NULL;

	if (unlikely(v == SEQ_START_TOKEN))
		prev_key = NULL;
	else
		prev_key = key;

	if (map->ops->map_get_next_key(map, prev_key, key)) {
		map_iter(m)->done = true;
		return NULL;
	}

	++(*pos);
	return key;
}

static void *map_seq_start(struct seq_file *m, loff_t *pos)
{
	if (map_iter(m)->done)
		return NULL;

	return *pos ? map_iter(m)->key : SEQ_START_TOKEN;
}

static void map_seq_stop(struct seq_file *m, void *v)
{
}

static int map_seq_show(struct seq_file *m, void *v)
{
	struct bpf_map *map = seq_file_to_map(m);
	void *key = map_iter(m)->key;

	if (unlikely(v == SEQ_START_TOKEN)) {
		seq_puts(m, "# WARNING!! The output is for debug purpose only\n");
		seq_puts(m, "# WARNING!! The output format will change\n");
	} else {
		map->ops->map_seq_show_elem(map, key, m);
	}

	return 0;
}

static const struct seq_operations bpffs_map_seq_ops = {
	.start	= map_seq_start,
	.next	= map_seq_next,
	.show	= map_seq_show,
	.stop	= map_seq_stop,
};

static int bpffs_map_open(struct inode *inode, struct file *file)
{
	struct bpf_map *map = inode->i_private;
	struct map_iter *iter;
	struct seq_file *m;
	int err;

	iter = map_iter_alloc(map);
	if (!iter)
		return -ENOMEM;

	err = seq_open(file, &bpffs_map_seq_ops);
	if (err) {
		map_iter_free(iter);
		return err;
	}

	m = file->private_data;
	m->private = iter;

	return 0;
}

static int bpffs_map_release(struct inode *inode, struct file *file)
{
	struct seq_file *m = file->private_data;

	map_iter_free(map_iter(m));

	return seq_release(inode, file);
}

/* bpffs_map_fops should only implement the basic
 * read operation for a BPF map.  The purpose is to
 * provide a simple user intuitive way to do
 * "cat bpffs/pathto/a-pinned-map".
 *
 * Other operations (e.g. write, lookup...) should be realized by
 * the userspace tools (e.g. bpftool) through the
 * BPF_OBJ_GET_INFO_BY_FD and the map's lookup/update
 * interface.
 */
static const struct file_operations bpffs_map_fops = {
	.open		= bpffs_map_open,
	.read		= seq_read,
	.release	= bpffs_map_release,
};

static int bpffs_obj_open(struct inode *inode, struct file *file)
{
	return -EIO;
}

static const struct file_operations bpffs_obj_fops = {
	.open		= bpffs_obj_open,
};

static int bpf_mkobj_ops(struct dentry *dentry, umode_t mode, void *raw,
			 const struct inode_operations *iops,
			 const struct file_operations *fops)
{
	struct inode *dir = dentry->d_parent->d_inode;
	struct inode *inode = bpf_get_inode(dir->i_sb, dir, mode);
	if (IS_ERR(inode))
		return PTR_ERR(inode);

	inode->i_op = iops;
	inode->i_fop = fops;
	inode->i_private = raw;

	bpf_dentry_finalize(dentry, inode, dir);
	return 0;
}

static int bpf_mkprog(struct dentry *dentry, umode_t mode, void *arg)
{
	return bpf_mkobj_ops(dentry, mode, arg, &bpf_prog_iops,
			     &bpffs_obj_fops);
}

static int bpf_mkmap(struct dentry *dentry, umode_t mode, void *arg)
{
	struct bpf_map *map = arg;

	return bpf_mkobj_ops(dentry, mode, arg, &bpf_map_iops,
			     bpf_map_support_seq_show(map) ?
			     &bpffs_map_fops : &bpffs_obj_fops);
}

static struct dentry *
bpf_lookup(struct inode *dir, struct dentry *dentry, unsigned flags)
{
	/* Dots in names (e.g. "/sys/fs/bpf/foo.bar") are reserved for future
	 * extensions.
	 */
	if (strchr(dentry->d_name.name, '.'))
		return ERR_PTR(-EPERM);

	return simple_lookup(dir, dentry, flags);
}

static int bpf_symlink(struct inode *dir, struct dentry *dentry,
		       const char *target)
{
	char *link = kstrdup(target, GFP_USER | __GFP_NOWARN);
	struct inode *inode;

	if (!link)
		return -ENOMEM;

	inode = bpf_get_inode(dir->i_sb, dir, S_IRWXUGO | S_IFLNK);
	if (IS_ERR(inode)) {
		kfree(link);
		return PTR_ERR(inode);
	}

	inode->i_op = &simple_symlink_inode_operations;
	inode->i_link = link;

	bpf_dentry_finalize(dentry, inode, dir);
	return 0;
}

static const struct inode_operations bpf_dir_iops = {
	.lookup		= bpf_lookup,
	.mkdir		= bpf_mkdir,
	.symlink	= bpf_symlink,
	.rmdir		= simple_rmdir,
	.rename		= simple_rename,
	.link		= simple_link,
	.unlink		= simple_unlink,
};

static int bpf_obj_do_pin(const struct filename *pathname, void *raw,
			  enum bpf_type type)
{
	struct dentry *dentry;
	struct inode *dir;
	struct path path;
	umode_t mode;
	int ret;

	dentry = kern_path_create(AT_FDCWD, pathname->name, &path, 0);
	if (IS_ERR(dentry))
		return PTR_ERR(dentry);

	mode = S_IFREG | ((S_IRUSR | S_IWUSR) & ~current_umask());

	ret = security_path_mknod(&path, dentry, mode, 0);
	if (ret)
		goto out;

	dir = d_inode(path.dentry);
	if (dir->i_op != &bpf_dir_iops) {
		ret = -EPERM;
		goto out;
	}

	switch (type) {
	case BPF_TYPE_PROG:
		ret = vfs_mkobj(dentry, mode, bpf_mkprog, raw);
		break;
	case BPF_TYPE_MAP:
		ret = vfs_mkobj(dentry, mode, bpf_mkmap, raw);
		break;
	default:
		ret = -EPERM;
	}
out:
	done_path_create(&path, dentry);
	return ret;
}

int bpf_obj_pin_user(u32 ufd, const char __user *pathname)
{
	struct filename *pname;
	enum bpf_type type;
	void *raw;
	int ret;

	pname = getname(pathname);
	if (IS_ERR(pname))
		return PTR_ERR(pname);

	raw = bpf_fd_probe_obj(ufd, &type);
	if (IS_ERR(raw)) {
		ret = PTR_ERR(raw);
		goto out;
	}

	ret = bpf_obj_do_pin(pname, raw, type);
	if (ret != 0)
		bpf_any_put(raw, type);
out:
	putname(pname);
	return ret;
}

static void *bpf_obj_do_get(const struct filename *pathname,
			    enum bpf_type *type, int flags)
{
	struct inode *inode;
	struct path path;
	void *raw;
	int ret;

	ret = kern_path(pathname->name, LOOKUP_FOLLOW, &path);
	if (ret)
		return ERR_PTR(ret);

	inode = d_backing_inode(path.dentry);
	ret = inode_permission(inode, ACC_MODE(flags));
	if (ret)
		goto out;

	ret = bpf_inode_type(inode, type);
	if (ret)
		goto out;

	raw = bpf_any_get(inode->i_private, *type);
	if (!IS_ERR(raw))
		touch_atime(&path);

	path_put(&path);
	return raw;
out:
	path_put(&path);
	return ERR_PTR(ret);
}

int bpf_obj_get_user(const char __user *pathname, int flags)
{
	enum bpf_type type = BPF_TYPE_UNSPEC;
	struct filename *pname;
	int ret = -ENOENT;
	int f_flags;
	void *raw;

	f_flags = bpf_get_file_flag(flags);
	if (f_flags < 0)
		return f_flags;

	pname = getname(pathname);
	if (IS_ERR(pname))
		return PTR_ERR(pname);

	raw = bpf_obj_do_get(pname, &type, f_flags);
	if (IS_ERR(raw)) {
		ret = PTR_ERR(raw);
		goto out;
	}

	if (type == BPF_TYPE_PROG)
		ret = bpf_prog_new_fd(raw);
	else if (type == BPF_TYPE_MAP)
		ret = bpf_map_new_fd(raw, f_flags);
	else
		goto out;

	if (ret < 0)
		bpf_any_put(raw, type);
out:
	putname(pname);
	return ret;
}

static struct bpf_prog *__get_prog_inode(struct inode *inode, enum bpf_prog_type type)
{
	struct bpf_prog *prog;
	int ret = inode_permission(inode, MAY_READ);
	if (ret)
		return ERR_PTR(ret);

	if (inode->i_op == &bpf_map_iops)
		return ERR_PTR(-EINVAL);
	if (inode->i_op != &bpf_prog_iops)
		return ERR_PTR(-EACCES);

	prog = inode->i_private;

	ret = security_bpf_prog(prog);
	if (ret < 0)
		return ERR_PTR(ret);

	if (!bpf_prog_get_ok(prog, &type, false))
		return ERR_PTR(-EINVAL);

	return bpf_prog_inc(prog);
}

struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type)
{
	struct bpf_prog *prog;
	struct path path;
	int ret = kern_path(name, LOOKUP_FOLLOW, &path);
	if (ret)
		return ERR_PTR(ret);
	prog = __get_prog_inode(d_backing_inode(path.dentry), type);
	if (!IS_ERR(prog))
		touch_atime(&path);
	path_put(&path);
	return prog;
}
EXPORT_SYMBOL(bpf_prog_get_type_path);

/*
 * Display the mount options in /proc/mounts.
 */
static int bpf_show_options(struct seq_file *m, struct dentry *root)
{
	umode_t mode = d_inode(root)->i_mode & S_IALLUGO & ~S_ISVTX;

	if (mode != S_IRWXUGO)
		seq_printf(m, ",mode=%o", mode);
	return 0;
}

static void bpf_free_inode(struct inode *inode)
{
	enum bpf_type type;

	if (S_ISLNK(inode->i_mode))
		kfree(inode->i_link);
	if (!bpf_inode_type(inode, &type))
		bpf_any_put(inode->i_private, type);
	free_inode_nonrcu(inode);
}

static const struct super_operations bpf_super_ops = {
	.statfs		= simple_statfs,
	.drop_inode	= generic_delete_inode,
	.show_options	= bpf_show_options,
	.free_inode	= bpf_free_inode,
};

enum {
	OPT_MODE,
	OPT_ERR,
};

static const match_table_t bpf_mount_tokens = {
	{ OPT_MODE, "mode=%o" },
	{ OPT_ERR, NULL },
};

struct bpf_mount_opts {
	umode_t mode;
};

static int bpf_parse_options(char *data, struct bpf_mount_opts *opts)
{
	substring_t args[MAX_OPT_ARGS];
	int option, token;
	char *ptr;

	opts->mode = S_IRWXUGO;

	while ((ptr = strsep(&data, ",")) != NULL) {
		if (!*ptr)
			continue;

		token = match_token(ptr, bpf_mount_tokens, args);
		switch (token) {
		case OPT_MODE:
			if (match_octal(&args[0], &option))
				return -EINVAL;
			opts->mode = option & S_IALLUGO;
			break;
		/* We might like to report bad mount options here, but
		 * traditionally we've ignored all mount options, so we'd
		 * better continue to ignore non-existing options for bpf.
		 */
		}
	}

	return 0;
}

static int bpf_fill_super(struct super_block *sb, void *data, int silent)
{
	static const struct tree_descr bpf_rfiles[] = { { "" } };
	struct bpf_mount_opts opts;
	struct inode *inode;
	int ret;

	ret = bpf_parse_options(data, &opts);
	if (ret)
		return ret;

	ret = simple_fill_super(sb, BPF_FS_MAGIC, bpf_rfiles);
	if (ret)
		return ret;

	sb->s_op = &bpf_super_ops;

	inode = sb->s_root->d_inode;
	inode->i_op = &bpf_dir_iops;
	inode->i_mode &= ~S_IALLUGO;
	inode->i_mode |= S_ISVTX | opts.mode;

	return 0;
}

static struct dentry *bpf_mount(struct file_system_type *type, int flags,
				const char *dev_name, void *data)
{
	return mount_nodev(type, flags, data, bpf_fill_super);
}

static struct file_system_type bpf_fs_type = {
	.owner		= THIS_MODULE,
	.name		= "bpf",
	.mount		= bpf_mount,
	.kill_sb	= kill_litter_super,
};

static int __init bpf_init(void)
{
	int ret;

	ret = sysfs_create_mount_point(fs_kobj, "bpf");
	if (ret)
		return ret;

	ret = register_filesystem(&bpf_fs_type);
	if (ret)
		sysfs_remove_mount_point(fs_kobj, "bpf");

	return ret;
}
fs_initcall(bpf_init);
Commit	Line	Data
d2912cb1	1	// SPDX-License-Identifier: GPL-2.0-only
b2197755 DB	2	/*
	3	* Minimal file system backend for holding eBPF maps and programs,
	4	* used by bpf(2) object pinning.
	5	*
	6	* Authors:
	7	*
	8	* Daniel Borkmann <daniel@iogearbox.net>
b2197755 DB	9	*/
b2197755 DB	10
a536a6e1	11	#include <linux/init.h>
b2197755 DB	12	#include <linux/magic.h>
	13	#include <linux/major.h>
	14	#include <linux/mount.h>
	15	#include <linux/namei.h>
	16	#include <linux/fs.h>
	17	#include <linux/kdev_t.h>
a3af5f80	18	#include <linux/parser.h>
b2197755 DB	19	#include <linux/filter.h>
b2197755 DB	20	#include <linux/bpf.h>
a67edbf4	21	#include <linux/bpf_trace.h>
b2197755 DB	22
	23	enum bpf_type {
	24	BPF_TYPE_UNSPEC = 0,
	25	BPF_TYPE_PROG,
	26	BPF_TYPE_MAP,
	27	};
	28
	29	static void bpf_any_get(void raw, enum bpf_type type)
	30	{
	31	switch (type) {
	32	case BPF_TYPE_PROG:
92117d84	33	raw = bpf_prog_inc(raw);
b2197755 DB	34	break;
b2197755 DB	35	case BPF_TYPE_MAP:
92117d84	36	raw = bpf_map_inc(raw, true);
b2197755 DB	37	break;
	38	default:
	39	WARN_ON_ONCE(1);
	40	break;
	41	}
	42
	43	return raw;
	44	}
	45
	46	static void bpf_any_put(void *raw, enum bpf_type type)
	47	{
	48	switch (type) {
	49	case BPF_TYPE_PROG:
	50	bpf_prog_put(raw);
	51	break;
	52	case BPF_TYPE_MAP:
c9da161c	53	bpf_map_put_with_uref(raw);
b2197755 DB	54	break;
	55	default:
	56	WARN_ON_ONCE(1);
	57	break;
	58	}
	59	}
	60
	61	static void bpf_fd_probe_obj(u32 ufd, enum bpf_type type)
	62	{
	63	void *raw;
	64
	65	*type = BPF_TYPE_MAP;
c9da161c	66	raw = bpf_map_get_with_uref(ufd);
b2197755 DB	67	if (IS_ERR(raw)) {
	68	*type = BPF_TYPE_PROG;
	69	raw = bpf_prog_get(ufd);
	70	}
	71
	72	return raw;
	73	}
	74
	75	static const struct inode_operations bpf_dir_iops;
	76
	77	static const struct inode_operations bpf_prog_iops = { };
	78	static const struct inode_operations bpf_map_iops = { };
	79
	80	static struct inode bpf_get_inode(struct super_block sb,
	81	const struct inode *dir,
	82	umode_t mode)
	83	{
	84	struct inode *inode;
	85
	86	switch (mode & S_IFMT) {
	87	case S_IFDIR:
	88	case S_IFREG:
0f98621b	89	case S_IFLNK:
b2197755 DB	90	break;
	91	default:
	92	return ERR_PTR(-EINVAL);
	93	}
	94
	95	inode = new_inode(sb);
	96	if (!inode)
	97	return ERR_PTR(-ENOSPC);
	98
	99	inode->i_ino = get_next_ino();
078cd827	100	inode->i_atime = current_time(inode);
b2197755 DB	101	inode->i_mtime = inode->i_atime;
	102	inode->i_ctime = inode->i_atime;
	103
	104	inode_init_owner(inode, dir, mode);
	105
	106	return inode;
	107	}
	108
	109	static int bpf_inode_type(const struct inode inode, enum bpf_type type)
	110	{
	111	*type = BPF_TYPE_UNSPEC;
	112	if (inode->i_op == &bpf_prog_iops)
	113	*type = BPF_TYPE_PROG;
	114	else if (inode->i_op == &bpf_map_iops)
	115	*type = BPF_TYPE_MAP;
	116	else
	117	return -EACCES;
	118
	119	return 0;
	120	}
	121
0f98621b DB	122	static void bpf_dentry_finalize(struct dentry dentry, struct inode inode,
	123	struct inode *dir)
	124	{
	125	d_instantiate(dentry, inode);
	126	dget(dentry);
	127
	128	dir->i_mtime = current_time(dir);
	129	dir->i_ctime = dir->i_mtime;
	130	}
	131
b2197755 DB	132	static int bpf_mkdir(struct inode dir, struct dentry dentry, umode_t mode)
	133	{
	134	struct inode *inode;
	135
b2197755 DB	136	inode = bpf_get_inode(dir->i_sb, dir, mode \| S_IFDIR);
	137	if (IS_ERR(inode))
	138	return PTR_ERR(inode);
	139
	140	inode->i_op = &bpf_dir_iops;
	141	inode->i_fop = &simple_dir_operations;
	142
	143	inc_nlink(inode);
	144	inc_nlink(dir);
	145
0f98621b	146	bpf_dentry_finalize(dentry, inode, dir);
b2197755 DB	147	return 0;
	148	}
	149
a26ca7c9 MKL	150	struct map_iter {
	151	void *key;
	152	bool done;
	153	};
	154
	155	static struct map_iter map_iter(struct seq_file m)
	156	{
	157	return m->private;
	158	}
	159
	160	static struct bpf_map seq_file_to_map(struct seq_file m)
	161	{
	162	return file_inode(m->file)->i_private;
	163	}
	164
	165	static void map_iter_free(struct map_iter *iter)
	166	{
	167	if (iter) {
	168	kfree(iter->key);
	169	kfree(iter);
	170	}
	171	}
	172
	173	static struct map_iter map_iter_alloc(struct bpf_map map)
	174	{
	175	struct map_iter *iter;
	176
	177	iter = kzalloc(sizeof(*iter), GFP_KERNEL \| __GFP_NOWARN);
	178	if (!iter)
	179	goto error;
	180
	181	iter->key = kzalloc(map->key_size, GFP_KERNEL \| __GFP_NOWARN);
	182	if (!iter->key)
	183	goto error;
	184
	185	return iter;
	186
	187	error:
	188	map_iter_free(iter);
	189	return NULL;
	190	}
	191
	192	static void map_seq_next(struct seq_file m, void v, loff_t pos)
	193	{
	194	struct bpf_map *map = seq_file_to_map(m);
	195	void *key = map_iter(m)->key;
dc1508a5	196	void *prev_key;
a26ca7c9 MKL	197
	198	if (map_iter(m)->done)
	199	return NULL;
	200
	201	if (unlikely(v == SEQ_START_TOKEN))
dc1508a5 YS	202	prev_key = NULL;
	203	else
	204	prev_key = key;
a26ca7c9	205
dc1508a5	206	if (map->ops->map_get_next_key(map, prev_key, key)) {
a26ca7c9 MKL	207	map_iter(m)->done = true;
	208	return NULL;
	209	}
	210
a26ca7c9 MKL	211	++(*pos);
	212	return key;
	213	}
	214
	215	static void map_seq_start(struct seq_file m, loff_t *pos)
	216	{
	217	if (map_iter(m)->done)
	218	return NULL;
	219
	220	return *pos ? map_iter(m)->key : SEQ_START_TOKEN;
	221	}
	222
	223	static void map_seq_stop(struct seq_file m, void v)
	224	{
	225	}
	226
	227	static int map_seq_show(struct seq_file m, void v)
	228	{
	229	struct bpf_map *map = seq_file_to_map(m);
	230	void *key = map_iter(m)->key;
	231
	232	if (unlikely(v == SEQ_START_TOKEN)) {
	233	seq_puts(m, "# WARNING!! The output is for debug purpose only\n");
	234	seq_puts(m, "# WARNING!! The output format will change\n");
	235	} else {
	236	map->ops->map_seq_show_elem(map, key, m);
	237	}
	238
	239	return 0;
	240	}
	241
	242	static const struct seq_operations bpffs_map_seq_ops = {
	243	.start = map_seq_start,
	244	.next = map_seq_next,
	245	.show = map_seq_show,
	246	.stop = map_seq_stop,
	247	};
	248
	249	static int bpffs_map_open(struct inode inode, struct file file)
	250	{
	251	struct bpf_map *map = inode->i_private;
	252	struct map_iter *iter;
	253	struct seq_file *m;
	254	int err;
	255
	256	iter = map_iter_alloc(map);
	257	if (!iter)
	258	return -ENOMEM;
	259
	260	err = seq_open(file, &bpffs_map_seq_ops);
	261	if (err) {
	262	map_iter_free(iter);
	263	return err;
	264	}
	265
	266	m = file->private_data;
	267	m->private = iter;
	268
	269	return 0;
	270	}
	271
	272	static int bpffs_map_release(struct inode inode, struct file file)
	273	{
	274	struct seq_file *m = file->private_data;
275
276	map_iter_free(map_iter(m));
277
278	return seq_release(inode, file);
279	}
280
281	/* bpffs_map_fops should only implement the basic
282	* read operation for a BPF map. The purpose is to
283	* provide a simple user intuitive way to do
284	* "cat bpffs/pathto/a-pinned-map".
285	*
286	* Other operations (e.g. write, lookup...) should be realized by
287	* the userspace tools (e.g. bpftool) through the
288	* BPF_OBJ_GET_INFO_BY_FD and the map's lookup/update
289	* interface.
290	*/
291	static const struct file_operations bpffs_map_fops = {
292	.open = bpffs_map_open,
293	.read = seq_read,
294	.release = bpffs_map_release,
295	};
296
b1655857 DB	297	static int bpffs_obj_open(struct inode inode, struct file file)
	298	{
	299	return -EIO;
	300	}
	301
	302	static const struct file_operations bpffs_obj_fops = {
	303	.open = bpffs_obj_open,
	304	};
	305
a4a0683f	306	static int bpf_mkobj_ops(struct dentry dentry, umode_t mode, void raw,
a26ca7c9 MKL	307	const struct inode_operations *iops,
a26ca7c9 MKL	308	const struct file_operations *fops)
b2197755	309	{
a4a0683f AV	310	struct inode *dir = dentry->d_parent->d_inode;
a4a0683f AV	311	struct inode *inode = bpf_get_inode(dir->i_sb, dir, mode);
b2197755 DB	312	if (IS_ERR(inode))
	313	return PTR_ERR(inode);
	314
	315	inode->i_op = iops;
a26ca7c9	316	inode->i_fop = fops;
a4a0683f	317	inode->i_private = raw;
b2197755	318
0f98621b	319	bpf_dentry_finalize(dentry, inode, dir);
b2197755 DB	320	return 0;
	321	}
	322
a4a0683f	323	static int bpf_mkprog(struct dentry dentry, umode_t mode, void arg)
b2197755	324	{
b1655857 DB	325	return bpf_mkobj_ops(dentry, mode, arg, &bpf_prog_iops,
b1655857 DB	326	&bpffs_obj_fops);
a4a0683f	327	}
b2197755	328
a4a0683f AV	329	static int bpf_mkmap(struct dentry dentry, umode_t mode, void arg)
a4a0683f AV	330	{
a26ca7c9 MKL	331	struct bpf_map *map = arg;
	332
	333	return bpf_mkobj_ops(dentry, mode, arg, &bpf_map_iops,
e8d2bec0 DB	334	bpf_map_support_seq_show(map) ?
e8d2bec0 DB	335	&bpffs_map_fops : &bpffs_obj_fops);
b2197755 DB	336	}
b2197755 DB	337
0c93b7d8 AV	338	static struct dentry *
0c93b7d8 AV	339	bpf_lookup(struct inode dir, struct dentry dentry, unsigned flags)
bb35a6ef	340	{
6d8cb045 QM	341	/* Dots in names (e.g. "/sys/fs/bpf/foo.bar") are reserved for future
	342	* extensions.
	343	*/
0c93b7d8 AV	344	if (strchr(dentry->d_name.name, '.'))
0c93b7d8 AV	345	return ERR_PTR(-EPERM);
0f98621b	346
0c93b7d8	347	return simple_lookup(dir, dentry, flags);
bb35a6ef DB	348	}
bb35a6ef DB	349
0f98621b DB	350	static int bpf_symlink(struct inode dir, struct dentry dentry,
	351	const char *target)
	352	{
	353	char *link = kstrdup(target, GFP_USER \| __GFP_NOWARN);
	354	struct inode *inode;
	355
	356	if (!link)
	357	return -ENOMEM;
	358
	359	inode = bpf_get_inode(dir->i_sb, dir, S_IRWXUGO \| S_IFLNK);
	360	if (IS_ERR(inode)) {
	361	kfree(link);
	362	return PTR_ERR(inode);
	363	}
	364
	365	inode->i_op = &simple_symlink_inode_operations;
	366	inode->i_link = link;
	367
	368	bpf_dentry_finalize(dentry, inode, dir);
	369	return 0;
	370	}
	371
b2197755	372	static const struct inode_operations bpf_dir_iops = {
0c93b7d8	373	.lookup = bpf_lookup,
b2197755	374	.mkdir = bpf_mkdir,
0f98621b	375	.symlink = bpf_symlink,
b2197755	376	.rmdir = simple_rmdir,
0c93b7d8 AV	377	.rename = simple_rename,
0c93b7d8 AV	378	.link = simple_link,
b2197755 DB	379	.unlink = simple_unlink,
	380	};
	381
	382	static int bpf_obj_do_pin(const struct filename pathname, void raw,
	383	enum bpf_type type)
	384	{
	385	struct dentry *dentry;
	386	struct inode *dir;
	387	struct path path;
	388	umode_t mode;
b2197755 DB	389	int ret;
	390
	391	dentry = kern_path_create(AT_FDCWD, pathname->name, &path, 0);
	392	if (IS_ERR(dentry))
	393	return PTR_ERR(dentry);
	394
	395	mode = S_IFREG \| ((S_IRUSR \| S_IWUSR) & ~current_umask());
b2197755	396
a4a0683f	397	ret = security_path_mknod(&path, dentry, mode, 0);
b2197755 DB	398	if (ret)
	399	goto out;
	400
	401	dir = d_inode(path.dentry);
	402	if (dir->i_op != &bpf_dir_iops) {
	403	ret = -EPERM;
	404	goto out;
	405	}
	406
a4a0683f AV	407	switch (type) {
	408	case BPF_TYPE_PROG:
	409	ret = vfs_mkobj(dentry, mode, bpf_mkprog, raw);
	410	break;
	411	case BPF_TYPE_MAP:
	412	ret = vfs_mkobj(dentry, mode, bpf_mkmap, raw);
	413	break;
	414	default:
	415	ret = -EPERM;
	416	}
b2197755 DB	417	out:
	418	done_path_create(&path, dentry);
	419	return ret;
	420	}
	421
	422	int bpf_obj_pin_user(u32 ufd, const char __user *pathname)
	423	{
	424	struct filename *pname;
	425	enum bpf_type type;
	426	void *raw;
	427	int ret;
	428
	429	pname = getname(pathname);
	430	if (IS_ERR(pname))
	431	return PTR_ERR(pname);
	432
	433	raw = bpf_fd_probe_obj(ufd, &type);
	434	if (IS_ERR(raw)) {
	435	ret = PTR_ERR(raw);
	436	goto out;
	437	}
	438
	439	ret = bpf_obj_do_pin(pname, raw, type);
	440	if (ret != 0)
	441	bpf_any_put(raw, type);
	442	out:
	443	putname(pname);
	444	return ret;
	445	}
	446
	447	static void bpf_obj_do_get(const struct filename pathname,
6e71b04a	448	enum bpf_type *type, int flags)
b2197755 DB	449	{
	450	struct inode *inode;
	451	struct path path;
	452	void *raw;
	453	int ret;
	454
	455	ret = kern_path(pathname->name, LOOKUP_FOLLOW, &path);
	456	if (ret)
	457	return ERR_PTR(ret);
	458
	459	inode = d_backing_inode(path.dentry);
6e71b04a	460	ret = inode_permission(inode, ACC_MODE(flags));
b2197755 DB	461	if (ret)
	462	goto out;
	463
	464	ret = bpf_inode_type(inode, type);
	465	if (ret)
	466	goto out;
	467
	468	raw = bpf_any_get(inode->i_private, *type);
92117d84 AS	469	if (!IS_ERR(raw))
92117d84 AS	470	touch_atime(&path);
b2197755 DB	471
	472	path_put(&path);
	473	return raw;
	474	out:
	475	path_put(&path);
	476	return ERR_PTR(ret);
	477	}
	478
6e71b04a	479	int bpf_obj_get_user(const char __user *pathname, int flags)
b2197755 DB	480	{
	481	enum bpf_type type = BPF_TYPE_UNSPEC;
	482	struct filename *pname;
	483	int ret = -ENOENT;
6e71b04a	484	int f_flags;
b2197755 DB	485	void *raw;
b2197755 DB	486
6e71b04a CF	487	f_flags = bpf_get_file_flag(flags);
	488	if (f_flags < 0)
	489	return f_flags;
	490
b2197755 DB	491	pname = getname(pathname);
	492	if (IS_ERR(pname))
	493	return PTR_ERR(pname);
	494
6e71b04a	495	raw = bpf_obj_do_get(pname, &type, f_flags);
b2197755 DB	496	if (IS_ERR(raw)) {
	497	ret = PTR_ERR(raw);
	498	goto out;
	499	}
	500
	501	if (type == BPF_TYPE_PROG)
	502	ret = bpf_prog_new_fd(raw);
	503	else if (type == BPF_TYPE_MAP)
6e71b04a	504	ret = bpf_map_new_fd(raw, f_flags);
b2197755 DB	505	else
	506	goto out;
	507
4d220ed0	508	if (ret < 0)
b2197755 DB	509	bpf_any_put(raw, type);
	510	out:
	511	putname(pname);
	512	return ret;
	513	}
040ee692 AV	514
	515	static struct bpf_prog __get_prog_inode(struct inode inode, enum bpf_prog_type type)
	516	{
	517	struct bpf_prog *prog;
e547ff3f	518	int ret = inode_permission(inode, MAY_READ);
040ee692 AV	519	if (ret)
	520	return ERR_PTR(ret);
	521
	522	if (inode->i_op == &bpf_map_iops)
	523	return ERR_PTR(-EINVAL);
	524	if (inode->i_op != &bpf_prog_iops)
	525	return ERR_PTR(-EACCES);
	526
	527	prog = inode->i_private;
	528
	529	ret = security_bpf_prog(prog);
	530	if (ret < 0)
	531	return ERR_PTR(ret);
	532
	533	if (!bpf_prog_get_ok(prog, &type, false))
	534	return ERR_PTR(-EINVAL);
	535
	536	return bpf_prog_inc(prog);
	537	}
	538
	539	struct bpf_prog bpf_prog_get_type_path(const char name, enum bpf_prog_type type)
	540	{
	541	struct bpf_prog *prog;
	542	struct path path;
	543	int ret = kern_path(name, LOOKUP_FOLLOW, &path);
	544	if (ret)
	545	return ERR_PTR(ret);
	546	prog = __get_prog_inode(d_backing_inode(path.dentry), type);
	547	if (!IS_ERR(prog))
	548	touch_atime(&path);
	549	path_put(&path);
	550	return prog;
	551	}
	552	EXPORT_SYMBOL(bpf_prog_get_type_path);
b2197755	553
4cc7c186 DH	554	/*
	555	* Display the mount options in /proc/mounts.
	556	*/
	557	static int bpf_show_options(struct seq_file m, struct dentry root)
	558	{
	559	umode_t mode = d_inode(root)->i_mode & S_IALLUGO & ~S_ISVTX;
	560
	561	if (mode != S_IRWXUGO)
	562	seq_printf(m, ",mode=%o", mode);
	563	return 0;
	564	}
	565
524845ff	566	static void bpf_free_inode(struct inode *inode)
1da6c4d9	567	{
1da6c4d9 DB	568	enum bpf_type type;
	569
	570	if (S_ISLNK(inode->i_mode))
	571	kfree(inode->i_link);
	572	if (!bpf_inode_type(inode, &type))
	573	bpf_any_put(inode->i_private, type);
	574	free_inode_nonrcu(inode);
	575	}
	576
b2197755 DB	577	static const struct super_operations bpf_super_ops = {
	578	.statfs = simple_statfs,
	579	.drop_inode = generic_delete_inode,
4cc7c186	580	.show_options = bpf_show_options,
524845ff	581	.free_inode = bpf_free_inode,
b2197755 DB	582	};
b2197755 DB	583
a3af5f80 DB	584	enum {
	585	OPT_MODE,
	586	OPT_ERR,
	587	};
	588
	589	static const match_table_t bpf_mount_tokens = {
	590	{ OPT_MODE, "mode=%o" },
	591	{ OPT_ERR, NULL },
	592	};
	593
	594	struct bpf_mount_opts {
	595	umode_t mode;
	596	};
	597
	598	static int bpf_parse_options(char data, struct bpf_mount_opts opts)
	599	{
	600	substring_t args[MAX_OPT_ARGS];
	601	int option, token;
	602	char *ptr;
	603
	604	opts->mode = S_IRWXUGO;
	605
	606	while ((ptr = strsep(&data, ",")) != NULL) {
	607	if (!*ptr)
	608	continue;
	609
	610	token = match_token(ptr, bpf_mount_tokens, args);
	611	switch (token) {
	612	case OPT_MODE:
	613	if (match_octal(&args[0], &option))
	614	return -EINVAL;
	615	opts->mode = option & S_IALLUGO;
	616	break;
	617	/* We might like to report bad mount options here, but
	618	* traditionally we've ignored all mount options, so we'd
	619	* better continue to ignore non-existing options for bpf.
	620	*/
	621	}
	622	}
	623
	624	return 0;
	625	}
	626
b2197755 DB	627	static int bpf_fill_super(struct super_block sb, void data, int silent)
b2197755 DB	628	{
cda37124	629	static const struct tree_descr bpf_rfiles[] = { { "" } };
a3af5f80	630	struct bpf_mount_opts opts;
b2197755 DB	631	struct inode *inode;
	632	int ret;
	633
a3af5f80 DB	634	ret = bpf_parse_options(data, &opts);
	635	if (ret)
	636	return ret;
	637
b2197755 DB	638	ret = simple_fill_super(sb, BPF_FS_MAGIC, bpf_rfiles);
	639	if (ret)
	640	return ret;
	641
	642	sb->s_op = &bpf_super_ops;
	643
	644	inode = sb->s_root->d_inode;
	645	inode->i_op = &bpf_dir_iops;
	646	inode->i_mode &= ~S_IALLUGO;
a3af5f80	647	inode->i_mode \|= S_ISVTX \| opts.mode;
b2197755 DB	648
	649	return 0;
	650	}
	651
	652	static struct dentry bpf_mount(struct file_system_type type, int flags,
	653	const char dev_name, void data)
	654	{
e27f4a94	655	return mount_nodev(type, flags, data, bpf_fill_super);
b2197755 DB	656	}
	657
	658	static struct file_system_type bpf_fs_type = {
	659	.owner = THIS_MODULE,
	660	.name = "bpf",
	661	.mount = bpf_mount,
	662	.kill_sb = kill_litter_super,
b2197755 DB	663	};
b2197755 DB	664
b2197755 DB	665	static int __init bpf_init(void)
	666	{
	667	int ret;
	668
	669	ret = sysfs_create_mount_point(fs_kobj, "bpf");
	670	if (ret)
	671	return ret;
	672
	673	ret = register_filesystem(&bpf_fs_type);
	674	if (ret)
	675	sysfs_remove_mount_point(fs_kobj, "bpf");
	676
	677	return ret;
	678	}
	679	fs_initcall(bpf_init);