[linux.git] / mm / memfd.c

/*
 * memfd_create system call and file sealing support
 *
 * Code was originally included in shmem.c, and broken out to facilitate
 * use by hugetlbfs as well as tmpfs.
 *
 * This file is released under the GPL.
 */

#include <linux/fs.h>
#include <linux/vfs.h>
#include <linux/pagemap.h>
#include <linux/file.h>
#include <linux/mm.h>
#include <linux/sched/signal.h>
#include <linux/khugepaged.h>
#include <linux/syscalls.h>
#include <linux/hugetlb.h>
#include <linux/shmem_fs.h>
#include <linux/memfd.h>
#include <uapi/linux/memfd.h>

/*
 * We need a tag: a new tag would expand every radix_tree_node by 8 bytes,
 * so reuse a tag which we firmly believe is never set or cleared on tmpfs
 * or hugetlbfs because they are memory only filesystems.
 */
#define MEMFD_TAG_PINNED        PAGECACHE_TAG_TOWRITE
#define LAST_SCAN               4       /* about 150ms max */

static void memfd_tag_pins(struct address_space *mapping)
{
	struct radix_tree_iter iter;
	void __rcu **slot;
	pgoff_t start;
	struct page *page;

	lru_add_drain();
	start = 0;
	rcu_read_lock();

	radix_tree_for_each_slot(slot, &mapping->i_pages, &iter, start) {
		page = radix_tree_deref_slot(slot);
		if (!page || radix_tree_exception(page)) {
			if (radix_tree_deref_retry(page)) {
				slot = radix_tree_iter_retry(&iter);
				continue;
			}
		} else if (page_count(page) - page_mapcount(page) > 1) {
			xa_lock_irq(&mapping->i_pages);
			radix_tree_tag_set(&mapping->i_pages, iter.index,
					   MEMFD_TAG_PINNED);
			xa_unlock_irq(&mapping->i_pages);
		}

		if (need_resched()) {
			slot = radix_tree_iter_resume(slot, &iter);
			cond_resched_rcu();
		}
	}
	rcu_read_unlock();
}

/*
 * Setting SEAL_WRITE requires us to verify there's no pending writer. However,
 * via get_user_pages(), drivers might have some pending I/O without any active
 * user-space mappings (eg., direct-IO, AIO). Therefore, we look at all pages
 * and see whether it has an elevated ref-count. If so, we tag them and wait for
 * them to be dropped.
 * The caller must guarantee that no new user will acquire writable references
 * to those pages to avoid races.
 */
static int memfd_wait_for_pins(struct address_space *mapping)
{
	struct radix_tree_iter iter;
	void __rcu **slot;
	pgoff_t start;
	struct page *page;
	int error, scan;

	memfd_tag_pins(mapping);

	error = 0;
	for (scan = 0; scan <= LAST_SCAN; scan++) {
		if (!radix_tree_tagged(&mapping->i_pages, MEMFD_TAG_PINNED))
			break;

		if (!scan)
			lru_add_drain_all();
		else if (schedule_timeout_killable((HZ << scan) / 200))
			scan = LAST_SCAN;

		start = 0;
		rcu_read_lock();
		radix_tree_for_each_tagged(slot, &mapping->i_pages, &iter,
					   start, MEMFD_TAG_PINNED) {

			page = radix_tree_deref_slot(slot);
			if (radix_tree_exception(page)) {
				if (radix_tree_deref_retry(page)) {
					slot = radix_tree_iter_retry(&iter);
					continue;
				}

				page = NULL;
			}

			if (page &&
			    page_count(page) - page_mapcount(page) != 1) {
				if (scan < LAST_SCAN)
					goto continue_resched;

				/*
				 * On the last scan, we clean up all those tags
				 * we inserted; but make a note that we still
				 * found pages pinned.
				 */
				error = -EBUSY;
			}

			xa_lock_irq(&mapping->i_pages);
			radix_tree_tag_clear(&mapping->i_pages,
					     iter.index, MEMFD_TAG_PINNED);
			xa_unlock_irq(&mapping->i_pages);
continue_resched:
			if (need_resched()) {
				slot = radix_tree_iter_resume(slot, &iter);
				cond_resched_rcu();
			}
		}
		rcu_read_unlock();
	}

	return error;
}

static unsigned int *memfd_file_seals_ptr(struct file *file)
{
	if (shmem_file(file))
		return &SHMEM_I(file_inode(file))->seals;

#ifdef CONFIG_HUGETLBFS
	if (is_file_hugepages(file))
		return &HUGETLBFS_I(file_inode(file))->seals;
#endif

	return NULL;
}

#define F_ALL_SEALS (F_SEAL_SEAL | \
		     F_SEAL_SHRINK | \
		     F_SEAL_GROW | \
		     F_SEAL_WRITE)

static int memfd_add_seals(struct file *file, unsigned int seals)
{
	struct inode *inode = file_inode(file);
	unsigned int *file_seals;
	int error;

	/*
	 * SEALING
	 * Sealing allows multiple parties to share a tmpfs or hugetlbfs file
	 * but restrict access to a specific subset of file operations. Seals
	 * can only be added, but never removed. This way, mutually untrusted
	 * parties can share common memory regions with a well-defined policy.
	 * A malicious peer can thus never perform unwanted operations on a
	 * shared object.
	 *
	 * Seals are only supported on special tmpfs or hugetlbfs files and
	 * always affect the whole underlying inode. Once a seal is set, it
	 * may prevent some kinds of access to the file. Currently, the
	 * following seals are defined:
	 *   SEAL_SEAL: Prevent further seals from being set on this file
	 *   SEAL_SHRINK: Prevent the file from shrinking
	 *   SEAL_GROW: Prevent the file from growing
	 *   SEAL_WRITE: Prevent write access to the file
	 *
	 * As we don't require any trust relationship between two parties, we
	 * must prevent seals from being removed. Therefore, sealing a file
	 * only adds a given set of seals to the file, it never touches
	 * existing seals. Furthermore, the "setting seals"-operation can be
	 * sealed itself, which basically prevents any further seal from being
	 * added.
	 *
	 * Semantics of sealing are only defined on volatile files. Only
	 * anonymous tmpfs and hugetlbfs files support sealing. More
	 * importantly, seals are never written to disk. Therefore, there's
	 * no plan to support it on other file types.
	 */

	if (!(file->f_mode & FMODE_WRITE))
		return -EPERM;
	if (seals & ~(unsigned int)F_ALL_SEALS)
		return -EINVAL;

	inode_lock(inode);

	file_seals = memfd_file_seals_ptr(file);
	if (!file_seals) {
		error = -EINVAL;
		goto unlock;
	}

	if (*file_seals & F_SEAL_SEAL) {
		error = -EPERM;
		goto unlock;
	}

	if ((seals & F_SEAL_WRITE) && !(*file_seals & F_SEAL_WRITE)) {
		error = mapping_deny_writable(file->f_mapping);
		if (error)
			goto unlock;

		error = memfd_wait_for_pins(file->f_mapping);
		if (error) {
			mapping_allow_writable(file->f_mapping);
			goto unlock;
		}
	}

	*file_seals |= seals;
	error = 0;

unlock:
	inode_unlock(inode);
	return error;
}

static int memfd_get_seals(struct file *file)
{
	unsigned int *seals = memfd_file_seals_ptr(file);

	return seals ? *seals : -EINVAL;
}

long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
{
	long error;

	switch (cmd) {
	case F_ADD_SEALS:
		/* disallow upper 32bit */
		if (arg > UINT_MAX)
			return -EINVAL;

		error = memfd_add_seals(file, arg);
		break;
	case F_GET_SEALS:
		error = memfd_get_seals(file);
		break;
	default:
		error = -EINVAL;
		break;
	}

	return error;
}

#define MFD_NAME_PREFIX "memfd:"
#define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
#define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)

#define MFD_ALL_FLAGS (MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB)

SYSCALL_DEFINE2(memfd_create,
		const char __user *, uname,
		unsigned int, flags)
{
	unsigned int *file_seals;
	struct file *file;
	int fd, error;
	char *name;
	long len;

	if (!(flags & MFD_HUGETLB)) {
		if (flags & ~(unsigned int)MFD_ALL_FLAGS)
			return -EINVAL;
	} else {
		/* Allow huge page size encoding in flags. */
		if (flags & ~(unsigned int)(MFD_ALL_FLAGS |
				(MFD_HUGE_MASK << MFD_HUGE_SHIFT)))
			return -EINVAL;
	}

	/* length includes terminating zero */
	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
	if (len <= 0)
		return -EFAULT;
	if (len > MFD_NAME_MAX_LEN + 1)
		return -EINVAL;

	name = kmalloc(len + MFD_NAME_PREFIX_LEN, GFP_KERNEL);
	if (!name)
		return -ENOMEM;

	strcpy(name, MFD_NAME_PREFIX);
	if (copy_from_user(&name[MFD_NAME_PREFIX_LEN], uname, len)) {
		error = -EFAULT;
		goto err_name;
	}

	/* terminating-zero may have changed after strnlen_user() returned */
	if (name[len + MFD_NAME_PREFIX_LEN - 1]) {
		error = -EFAULT;
		goto err_name;
	}

	fd = get_unused_fd_flags((flags & MFD_CLOEXEC) ? O_CLOEXEC : 0);
	if (fd < 0) {
		error = fd;
		goto err_name;
	}

	if (flags & MFD_HUGETLB) {
		struct user_struct *user = NULL;

		file = hugetlb_file_setup(name, 0, VM_NORESERVE, &user,
					HUGETLB_ANONHUGE_INODE,
					(flags >> MFD_HUGE_SHIFT) &
					MFD_HUGE_MASK);
	} else
		file = shmem_file_setup(name, 0, VM_NORESERVE);
	if (IS_ERR(file)) {
		error = PTR_ERR(file);
		goto err_fd;
	}
	file->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE;
	file->f_flags |= O_RDWR | O_LARGEFILE;

	if (flags & MFD_ALLOW_SEALING) {
		file_seals = memfd_file_seals_ptr(file);
		*file_seals &= ~F_SEAL_SEAL;
	}

	fd_install(fd, file);
	kfree(name);
	return fd;

err_fd:
	put_unused_fd(fd);
err_name:
	kfree(name);
	return error;
}
Commit	Line	Data
5d752600 MK	1	/*
	2	* memfd_create system call and file sealing support
	3	*
	4	* Code was originally included in shmem.c, and broken out to facilitate
	5	* use by hugetlbfs as well as tmpfs.
	6	*
	7	* This file is released under the GPL.
	8	*/
	9
	10	#include <linux/fs.h>
	11	#include <linux/vfs.h>
	12	#include <linux/pagemap.h>
	13	#include <linux/file.h>
	14	#include <linux/mm.h>
	15	#include <linux/sched/signal.h>
	16	#include <linux/khugepaged.h>
	17	#include <linux/syscalls.h>
	18	#include <linux/hugetlb.h>
	19	#include <linux/shmem_fs.h>
	20	#include <linux/memfd.h>
	21	#include <uapi/linux/memfd.h>
	22
	23	/*
	24	* We need a tag: a new tag would expand every radix_tree_node by 8 bytes,
	25	* so reuse a tag which we firmly believe is never set or cleared on tmpfs
	26	* or hugetlbfs because they are memory only filesystems.
	27	*/
	28	#define MEMFD_TAG_PINNED PAGECACHE_TAG_TOWRITE
	29	#define LAST_SCAN 4 /* about 150ms max */
	30
	31	static void memfd_tag_pins(struct address_space *mapping)
	32	{
	33	struct radix_tree_iter iter;
	34	void __rcu **slot;
	35	pgoff_t start;
	36	struct page *page;
	37
	38	lru_add_drain();
	39	start = 0;
	40	rcu_read_lock();
	41
	42	radix_tree_for_each_slot(slot, &mapping->i_pages, &iter, start) {
	43	page = radix_tree_deref_slot(slot);
	44	if (!page \|\| radix_tree_exception(page)) {
	45	if (radix_tree_deref_retry(page)) {
	46	slot = radix_tree_iter_retry(&iter);
	47	continue;
	48	}
	49	} else if (page_count(page) - page_mapcount(page) > 1) {
	50	xa_lock_irq(&mapping->i_pages);
	51	radix_tree_tag_set(&mapping->i_pages, iter.index,
	52	MEMFD_TAG_PINNED);
	53	xa_unlock_irq(&mapping->i_pages);
	54	}
	55
	56	if (need_resched()) {
	57	slot = radix_tree_iter_resume(slot, &iter);
	58	cond_resched_rcu();
	59	}
	60	}
	61	rcu_read_unlock();
	62	}
	63
	64	/*
65	* Setting SEAL_WRITE requires us to verify there's no pending writer. However,
66	* via get_user_pages(), drivers might have some pending I/O without any active
67	* user-space mappings (eg., direct-IO, AIO). Therefore, we look at all pages
68	* and see whether it has an elevated ref-count. If so, we tag them and wait for
69	* them to be dropped.
70	* The caller must guarantee that no new user will acquire writable references
71	* to those pages to avoid races.
72	*/
73	static int memfd_wait_for_pins(struct address_space *mapping)
74	{
75	struct radix_tree_iter iter;
76	void __rcu **slot;
77	pgoff_t start;
78	struct page *page;
79	int error, scan;
80
81	memfd_tag_pins(mapping);
82
83	error = 0;
84	for (scan = 0; scan <= LAST_SCAN; scan++) {
85	if (!radix_tree_tagged(&mapping->i_pages, MEMFD_TAG_PINNED))
86	break;
87
88	if (!scan)
89	lru_add_drain_all();
90	else if (schedule_timeout_killable((HZ << scan) / 200))
91	scan = LAST_SCAN;
92
93	start = 0;
94	rcu_read_lock();
95	radix_tree_for_each_tagged(slot, &mapping->i_pages, &iter,
96	start, MEMFD_TAG_PINNED) {
97
98	page = radix_tree_deref_slot(slot);
99	if (radix_tree_exception(page)) {
100	if (radix_tree_deref_retry(page)) {
101	slot = radix_tree_iter_retry(&iter);
102	continue;
103	}
104
105	page = NULL;
106	}
107
108	if (page &&
109	page_count(page) - page_mapcount(page) != 1) {
110	if (scan < LAST_SCAN)
111	goto continue_resched;
112
113	/*
114	* On the last scan, we clean up all those tags
115	* we inserted; but make a note that we still
116	* found pages pinned.
117	*/
118	error = -EBUSY;
119	}
120
121	xa_lock_irq(&mapping->i_pages);
122	radix_tree_tag_clear(&mapping->i_pages,
123	iter.index, MEMFD_TAG_PINNED);
124	xa_unlock_irq(&mapping->i_pages);
125	continue_resched:
126	if (need_resched()) {
127	slot = radix_tree_iter_resume(slot, &iter);
128	cond_resched_rcu();
129	}
130	}
131	rcu_read_unlock();
132	}
133
134	return error;
135	}
136
137	static unsigned int memfd_file_seals_ptr(struct file file)
138	{
139	if (shmem_file(file))
140	return &SHMEM_I(file_inode(file))->seals;
141
142	#ifdef CONFIG_HUGETLBFS
143	if (is_file_hugepages(file))
144	return &HUGETLBFS_I(file_inode(file))->seals;
145	#endif
146
147	return NULL;
148	}
149
150	#define F_ALL_SEALS (F_SEAL_SEAL \| \
151	F_SEAL_SHRINK \| \
152	F_SEAL_GROW \| \
153	F_SEAL_WRITE)
154
155	static int memfd_add_seals(struct file *file, unsigned int seals)
156	{
157	struct inode *inode = file_inode(file);
158	unsigned int *file_seals;
159	int error;
160
161	/*
162	* SEALING
163	* Sealing allows multiple parties to share a tmpfs or hugetlbfs file
164	* but restrict access to a specific subset of file operations. Seals
165	* can only be added, but never removed. This way, mutually untrusted
166	* parties can share common memory regions with a well-defined policy.
167	* A malicious peer can thus never perform unwanted operations on a
168	* shared object.
169	*
170	* Seals are only supported on special tmpfs or hugetlbfs files and
171	* always affect the whole underlying inode. Once a seal is set, it
172	* may prevent some kinds of access to the file. Currently, the
173	* following seals are defined:
174	* SEAL_SEAL: Prevent further seals from being set on this file
175	* SEAL_SHRINK: Prevent the file from shrinking
176	* SEAL_GROW: Prevent the file from growing
177	* SEAL_WRITE: Prevent write access to the file
178	*
179	* As we don't require any trust relationship between two parties, we
180	* must prevent seals from being removed. Therefore, sealing a file
181	* only adds a given set of seals to the file, it never touches
182	* existing seals. Furthermore, the "setting seals"-operation can be
183	* sealed itself, which basically prevents any further seal from being
184	* added.
185	*
186	* Semantics of sealing are only defined on volatile files. Only
187	* anonymous tmpfs and hugetlbfs files support sealing. More
188	* importantly, seals are never written to disk. Therefore, there's
189	* no plan to support it on other file types.
190	*/
191
192	if (!(file->f_mode & FMODE_WRITE))
193	return -EPERM;
194	if (seals & ~(unsigned int)F_ALL_SEALS)
195	return -EINVAL;
196
197	inode_lock(inode);
198
199	file_seals = memfd_file_seals_ptr(file);
200	if (!file_seals) {
201	error = -EINVAL;
202	goto unlock;
203	}
204
205	if (*file_seals & F_SEAL_SEAL) {
206	error = -EPERM;
207	goto unlock;
208	}
209
210	if ((seals & F_SEAL_WRITE) && !(*file_seals & F_SEAL_WRITE)) {
211	error = mapping_deny_writable(file->f_mapping);
212	if (error)
213	goto unlock;
214
215	error = memfd_wait_for_pins(file->f_mapping);
216	if (error) {
217	mapping_allow_writable(file->f_mapping);
218	goto unlock;
219	}
220	}
221
222	*file_seals \|= seals;
223	error = 0;
224
225	unlock:
226	inode_unlock(inode);
227	return error;
228	}
229
230	static int memfd_get_seals(struct file *file)
231	{
232	unsigned int *seals = memfd_file_seals_ptr(file);
233
234	return seals ? *seals : -EINVAL;
235	}
236
237	long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
238	{
239	long error;
240
241	switch (cmd) {
242	case F_ADD_SEALS:
243	/* disallow upper 32bit */
244	if (arg > UINT_MAX)
245	return -EINVAL;
246
247	error = memfd_add_seals(file, arg);
248	break;
249	case F_GET_SEALS:
250	error = memfd_get_seals(file);
251	break;
252	default:
253	error = -EINVAL;
254	break;
255	}
256
257	return error;
258	}
259
260	#define MFD_NAME_PREFIX "memfd:"
261	#define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
262	#define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
263
264	#define MFD_ALL_FLAGS (MFD_CLOEXEC \| MFD_ALLOW_SEALING \| MFD_HUGETLB)
265
266	SYSCALL_DEFINE2(memfd_create,
267	const char __user *, uname,
268	unsigned int, flags)
269	{
270	unsigned int *file_seals;
271	struct file *file;
272	int fd, error;
273	char *name;
274	long len;
275
276	if (!(flags & MFD_HUGETLB)) {
277	if (flags & ~(unsigned int)MFD_ALL_FLAGS)
278	return -EINVAL;
279	} else {
280	/* Allow huge page size encoding in flags. */
281	if (flags & ~(unsigned int)(MFD_ALL_FLAGS \|
282	(MFD_HUGE_MASK << MFD_HUGE_SHIFT)))
283	return -EINVAL;
284	}
285
286	/* length includes terminating zero */
287	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
288	if (len <= 0)
289	return -EFAULT;
290	if (len > MFD_NAME_MAX_LEN + 1)
291	return -EINVAL;
292
293	name = kmalloc(len + MFD_NAME_PREFIX_LEN, GFP_KERNEL);
294	if (!name)
295	return -ENOMEM;
296
297	strcpy(name, MFD_NAME_PREFIX);
298	if (copy_from_user(&name[MFD_NAME_PREFIX_LEN], uname, len)) {
299	error = -EFAULT;
300	goto err_name;
301	}
302
303	/* terminating-zero may have changed after strnlen_user() returned */
304	if (name[len + MFD_NAME_PREFIX_LEN - 1]) {
305	error = -EFAULT;
306	goto err_name;
307	}
308
309	fd = get_unused_fd_flags((flags & MFD_CLOEXEC) ? O_CLOEXEC : 0);
310	if (fd < 0) {
311	error = fd;
312	goto err_name;
313	}
314
315	if (flags & MFD_HUGETLB) {
316	struct user_struct *user = NULL;
317
318	file = hugetlb_file_setup(name, 0, VM_NORESERVE, &user,
319	HUGETLB_ANONHUGE_INODE,
320	(flags >> MFD_HUGE_SHIFT) &
321	MFD_HUGE_MASK);
322	} else
323	file = shmem_file_setup(name, 0, VM_NORESERVE);
324	if (IS_ERR(file)) {
325	error = PTR_ERR(file);
326	goto err_fd;
327	}
328	file->f_mode \|= FMODE_LSEEK \| FMODE_PREAD \| FMODE_PWRITE;
329	file->f_flags \|= O_RDWR \| O_LARGEFILE;
330
331	if (flags & MFD_ALLOW_SEALING) {
332	file_seals = memfd_file_seals_ptr(file);
333	*file_seals &= ~F_SEAL_SEAL;
334	}
335
336	fd_install(fd, file);
337	kfree(name);
338	return fd;
339
340	err_fd:
341	put_unused_fd(fd);
342	err_name:
343	kfree(name);
344	return error;
345	}