[linux-2.6-block.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 xa_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)
{
	XA_STATE(xas, &mapping->i_pages, 0);
	struct page *page;
	int error, scan;

	memfd_tag_pins(mapping);

	error = 0;
	for (scan = 0; scan <= LAST_SCAN; scan++) {
		unsigned int tagged = 0;

		if (!xas_marked(&xas, MEMFD_TAG_PINNED))
			break;

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

		xas_set(&xas, 0);
		xas_lock_irq(&xas);
		xas_for_each_marked(&xas, page, ULONG_MAX, MEMFD_TAG_PINNED) {
			bool clear = true;
			if (xa_is_value(page))
				continue;
			if (page_count(page) - page_mapcount(page) != 1) {
				/*
				 * On the last scan, we clean up all those tags
				 * we inserted; but make a note that we still
				 * found pages pinned.
				 */
				if (scan == LAST_SCAN)
					error = -EBUSY;
				else
					clear = false;
			}
			if (clear)
				xas_clear_mark(&xas, MEMFD_TAG_PINNED);
			if (++tagged % XA_CHECK_SCHED)
				continue;

			xas_pause(&xas);
			xas_unlock_irq(&xas);
			cond_resched();
			xas_lock_irq(&xas);
		}
		xas_unlock_irq(&xas);
	}

	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_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	/*
2313216f	24	* We need a tag: a new tag would expand every xa_node by 8 bytes,
5d752600 MK	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	{
2313216f	75	XA_STATE(xas, &mapping->i_pages, 0);
5d752600 MK	76	struct page *page;
	77	int error, scan;
	78
	79	memfd_tag_pins(mapping);
	80
	81	error = 0;
	82	for (scan = 0; scan <= LAST_SCAN; scan++) {
2313216f MW	83	unsigned int tagged = 0;
	84
	85	if (!xas_marked(&xas, MEMFD_TAG_PINNED))
5d752600 MK	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
2313216f MW	93	xas_set(&xas, 0);
	94	xas_lock_irq(&xas);
	95	xas_for_each_marked(&xas, page, ULONG_MAX, MEMFD_TAG_PINNED) {
	96	bool clear = true;
	97	if (xa_is_value(page))
	98	continue;
	99	if (page_count(page) - page_mapcount(page) != 1) {
5d752600 MK	100	/*
	101	* On the last scan, we clean up all those tags
	102	* we inserted; but make a note that we still
	103	* found pages pinned.
	104	*/
2313216f MW	105	if (scan == LAST_SCAN)
	106	error = -EBUSY;
	107	else
	108	clear = false;
5d752600	109	}
2313216f MW	110	if (clear)
	111	xas_clear_mark(&xas, MEMFD_TAG_PINNED);
	112	if (++tagged % XA_CHECK_SCHED)
	113	continue;
5d752600	114
2313216f MW	115	xas_pause(&xas);
	116	xas_unlock_irq(&xas);
	117	cond_resched();
	118	xas_lock_irq(&xas);
5d752600	119	}
2313216f	120	xas_unlock_irq(&xas);
5d752600 MK	121	}
	122
	123	return error;
	124	}
	125
	126	static unsigned int memfd_file_seals_ptr(struct file file)
	127	{
	128	if (shmem_file(file))
	129	return &SHMEM_I(file_inode(file))->seals;
	130
	131	#ifdef CONFIG_HUGETLBFS
	132	if (is_file_hugepages(file))
	133	return &HUGETLBFS_I(file_inode(file))->seals;
	134	#endif
	135
	136	return NULL;
	137	}
	138
	139	#define F_ALL_SEALS (F_SEAL_SEAL \| \
	140	F_SEAL_SHRINK \| \
	141	F_SEAL_GROW \| \
	142	F_SEAL_WRITE)
	143
	144	static int memfd_add_seals(struct file *file, unsigned int seals)
	145	{
	146	struct inode *inode = file_inode(file);
	147	unsigned int *file_seals;
	148	int error;
	149
	150	/*
	151	* SEALING
	152	* Sealing allows multiple parties to share a tmpfs or hugetlbfs file
	153	* but restrict access to a specific subset of file operations. Seals
	154	* can only be added, but never removed. This way, mutually untrusted
	155	* parties can share common memory regions with a well-defined policy.
	156	* A malicious peer can thus never perform unwanted operations on a
	157	* shared object.
	158	*
	159	* Seals are only supported on special tmpfs or hugetlbfs files and
	160	* always affect the whole underlying inode. Once a seal is set, it
	161	* may prevent some kinds of access to the file. Currently, the
	162	* following seals are defined:
	163	* SEAL_SEAL: Prevent further seals from being set on this file
	164	* SEAL_SHRINK: Prevent the file from shrinking
	165	* SEAL_GROW: Prevent the file from growing
	166	* SEAL_WRITE: Prevent write access to the file
	167	*
	168	* As we don't require any trust relationship between two parties, we
	169	* must prevent seals from being removed. Therefore, sealing a file
	170	* only adds a given set of seals to the file, it never touches
	171	* existing seals. Furthermore, the "setting seals"-operation can be
	172	* sealed itself, which basically prevents any further seal from being
	173	* added.
	174	*
	175	* Semantics of sealing are only defined on volatile files. Only
	176	* anonymous tmpfs and hugetlbfs files support sealing. More
	177	* importantly, seals are never written to disk. Therefore, there's
	178	* no plan to support it on other file types.
	179	*/
	180
	181	if (!(file->f_mode & FMODE_WRITE))
	182	return -EPERM;
	183	if (seals & ~(unsigned int)F_ALL_SEALS)
	184	return -EINVAL;
185
186	inode_lock(inode);
187
188	file_seals = memfd_file_seals_ptr(file);
189	if (!file_seals) {
190	error = -EINVAL;
191	goto unlock;
192	}
193
194	if (*file_seals & F_SEAL_SEAL) {
195	error = -EPERM;
196	goto unlock;
197	}
198
199	if ((seals & F_SEAL_WRITE) && !(*file_seals & F_SEAL_WRITE)) {
200	error = mapping_deny_writable(file->f_mapping);
201	if (error)
202	goto unlock;
203
204	error = memfd_wait_for_pins(file->f_mapping);
205	if (error) {
206	mapping_allow_writable(file->f_mapping);
207	goto unlock;
208	}
209	}
210
211	*file_seals \|= seals;
212	error = 0;
213
214	unlock:
215	inode_unlock(inode);
216	return error;
217	}
218
219	static int memfd_get_seals(struct file *file)
220	{
221	unsigned int *seals = memfd_file_seals_ptr(file);
222
223	return seals ? *seals : -EINVAL;
224	}
225
226	long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
227	{
228	long error;
229
230	switch (cmd) {
231	case F_ADD_SEALS:
232	/* disallow upper 32bit */
233	if (arg > UINT_MAX)
234	return -EINVAL;
235
236	error = memfd_add_seals(file, arg);
237	break;
238	case F_GET_SEALS:
239	error = memfd_get_seals(file);
240	break;
241	default:
242	error = -EINVAL;
243	break;
244	}
245
246	return error;
247	}
248
249	#define MFD_NAME_PREFIX "memfd:"
250	#define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
251	#define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
252
253	#define MFD_ALL_FLAGS (MFD_CLOEXEC \| MFD_ALLOW_SEALING \| MFD_HUGETLB)
254
255	SYSCALL_DEFINE2(memfd_create,
256	const char __user *, uname,
257	unsigned int, flags)
258	{
259	unsigned int *file_seals;
260	struct file *file;
261	int fd, error;
262	char *name;
263	long len;
264
265	if (!(flags & MFD_HUGETLB)) {
266	if (flags & ~(unsigned int)MFD_ALL_FLAGS)
267	return -EINVAL;
268	} else {
269	/* Allow huge page size encoding in flags. */
270	if (flags & ~(unsigned int)(MFD_ALL_FLAGS \|
271	(MFD_HUGE_MASK << MFD_HUGE_SHIFT)))
272	return -EINVAL;
273	}
274
275	/* length includes terminating zero */
276	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
277	if (len <= 0)
278	return -EFAULT;
279	if (len > MFD_NAME_MAX_LEN + 1)
280	return -EINVAL;
281
282	name = kmalloc(len + MFD_NAME_PREFIX_LEN, GFP_KERNEL);
283	if (!name)
284	return -ENOMEM;
285
286	strcpy(name, MFD_NAME_PREFIX);
287	if (copy_from_user(&name[MFD_NAME_PREFIX_LEN], uname, len)) {
288	error = -EFAULT;
289	goto err_name;
290	}
291
292	/* terminating-zero may have changed after strnlen_user() returned */
293	if (name[len + MFD_NAME_PREFIX_LEN - 1]) {
294	error = -EFAULT;
295	goto err_name;
296	}
297
298	fd = get_unused_fd_flags((flags & MFD_CLOEXEC) ? O_CLOEXEC : 0);
299	if (fd < 0) {
300	error = fd;
301	goto err_name;
302	}
303
304	if (flags & MFD_HUGETLB) {
305	struct user_struct *user = NULL;
306
307	file = hugetlb_file_setup(name, 0, VM_NORESERVE, &user,
308	HUGETLB_ANONHUGE_INODE,
309	(flags >> MFD_HUGE_SHIFT) &
310	MFD_HUGE_MASK);
311	} else
312	file = shmem_file_setup(name, 0, VM_NORESERVE);
313	if (IS_ERR(file)) {
314	error = PTR_ERR(file);
315	goto err_fd;
316	}
317	file->f_mode \|= FMODE_LSEEK \| FMODE_PREAD \| FMODE_PWRITE;
c9c554f2	318	file->f_flags \|= O_LARGEFILE;
5d752600 MK	319
	320	if (flags & MFD_ALLOW_SEALING) {
	321	file_seals = memfd_file_seals_ptr(file);
	322	*file_seals &= ~F_SEAL_SEAL;
	323	}
	324
	325	fd_install(fd, file);
	326	kfree(name);
	327	return fd;
	328
	329	err_fd:
	330	put_unused_fd(fd);
	331	err_name:
	332	kfree(name);
	333	return error;
	334	}