[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 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 xa_state *xas)
{
	struct page *page;
	unsigned int tagged = 0;

	lru_add_drain();

	xas_lock_irq(xas);
	xas_for_each(xas, page, ULONG_MAX) {
		if (xa_is_value(page))
			continue;
		if (page_count(page) - page_mapcount(page) > 1)
			xas_set_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);
}

/*
 * 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(&xas);

	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
ef3038a5	31	static void memfd_tag_pins(struct xa_state *xas)
5d752600	32	{
5d752600	33	struct page *page;
ef3038a5	34	unsigned int tagged = 0;
5d752600 MK	35
5d752600 MK	36	lru_add_drain();
5d752600	37
ef3038a5 MW	38	xas_lock_irq(xas);
	39	xas_for_each(xas, page, ULONG_MAX) {
	40	if (xa_is_value(page))
	41	continue;
	42	if (page_count(page) - page_mapcount(page) > 1)
	43	xas_set_mark(xas, MEMFD_TAG_PINNED);
5d752600	44
ef3038a5 MW	45	if (++tagged % XA_CHECK_SCHED)
	46	continue;
	47
	48	xas_pause(xas);
	49	xas_unlock_irq(xas);
	50	cond_resched();
	51	xas_lock_irq(xas);
5d752600	52	}
ef3038a5	53	xas_unlock_irq(xas);
5d752600 MK	54	}
	55
	56	/*
	57	* Setting SEAL_WRITE requires us to verify there's no pending writer. However,
	58	* via get_user_pages(), drivers might have some pending I/O without any active
	59	* user-space mappings (eg., direct-IO, AIO). Therefore, we look at all pages
	60	* and see whether it has an elevated ref-count. If so, we tag them and wait for
	61	* them to be dropped.
	62	* The caller must guarantee that no new user will acquire writable references
	63	* to those pages to avoid races.
	64	*/
	65	static int memfd_wait_for_pins(struct address_space *mapping)
	66	{
2313216f	67	XA_STATE(xas, &mapping->i_pages, 0);
5d752600 MK	68	struct page *page;
	69	int error, scan;
	70
ef3038a5	71	memfd_tag_pins(&xas);
5d752600 MK	72
	73	error = 0;
	74	for (scan = 0; scan <= LAST_SCAN; scan++) {
2313216f MW	75	unsigned int tagged = 0;
	76
	77	if (!xas_marked(&xas, MEMFD_TAG_PINNED))
5d752600 MK	78	break;
	79
	80	if (!scan)
	81	lru_add_drain_all();
	82	else if (schedule_timeout_killable((HZ << scan) / 200))
	83	scan = LAST_SCAN;
	84
2313216f MW	85	xas_set(&xas, 0);
	86	xas_lock_irq(&xas);
	87	xas_for_each_marked(&xas, page, ULONG_MAX, MEMFD_TAG_PINNED) {
	88	bool clear = true;
	89	if (xa_is_value(page))
	90	continue;
	91	if (page_count(page) - page_mapcount(page) != 1) {
5d752600 MK	92	/*
	93	* On the last scan, we clean up all those tags
	94	* we inserted; but make a note that we still
	95	* found pages pinned.
	96	*/
2313216f MW	97	if (scan == LAST_SCAN)
	98	error = -EBUSY;
	99	else
	100	clear = false;
5d752600	101	}
2313216f MW	102	if (clear)
	103	xas_clear_mark(&xas, MEMFD_TAG_PINNED);
	104	if (++tagged % XA_CHECK_SCHED)
	105	continue;
5d752600	106
2313216f MW	107	xas_pause(&xas);
	108	xas_unlock_irq(&xas);
	109	cond_resched();
	110	xas_lock_irq(&xas);
5d752600	111	}
2313216f	112	xas_unlock_irq(&xas);
5d752600 MK	113	}
	114
	115	return error;
	116	}
	117
	118	static unsigned int memfd_file_seals_ptr(struct file file)
	119	{
	120	if (shmem_file(file))
	121	return &SHMEM_I(file_inode(file))->seals;
	122
	123	#ifdef CONFIG_HUGETLBFS
	124	if (is_file_hugepages(file))
	125	return &HUGETLBFS_I(file_inode(file))->seals;
	126	#endif
	127
	128	return NULL;
	129	}
	130
	131	#define F_ALL_SEALS (F_SEAL_SEAL \| \
	132	F_SEAL_SHRINK \| \
	133	F_SEAL_GROW \| \
	134	F_SEAL_WRITE)
	135
	136	static int memfd_add_seals(struct file *file, unsigned int seals)
	137	{
	138	struct inode *inode = file_inode(file);
	139	unsigned int *file_seals;
	140	int error;
	141
	142	/*
	143	* SEALING
	144	* Sealing allows multiple parties to share a tmpfs or hugetlbfs file
	145	* but restrict access to a specific subset of file operations. Seals
	146	* can only be added, but never removed. This way, mutually untrusted
	147	* parties can share common memory regions with a well-defined policy.
	148	* A malicious peer can thus never perform unwanted operations on a
	149	* shared object.
	150	*
	151	* Seals are only supported on special tmpfs or hugetlbfs files and
	152	* always affect the whole underlying inode. Once a seal is set, it
	153	* may prevent some kinds of access to the file. Currently, the
	154	* following seals are defined:
	155	* SEAL_SEAL: Prevent further seals from being set on this file
	156	* SEAL_SHRINK: Prevent the file from shrinking
	157	* SEAL_GROW: Prevent the file from growing
	158	* SEAL_WRITE: Prevent write access to the file
	159	*
	160	* As we don't require any trust relationship between two parties, we
	161	* must prevent seals from being removed. Therefore, sealing a file
	162	* only adds a given set of seals to the file, it never touches
	163	* existing seals. Furthermore, the "setting seals"-operation can be
	164	* sealed itself, which basically prevents any further seal from being
	165	* added.
	166	*
	167	* Semantics of sealing are only defined on volatile files. Only
	168	* anonymous tmpfs and hugetlbfs files support sealing. More
	169	* importantly, seals are never written to disk. Therefore, there's
	170	* no plan to support it on other file types.
	171	*/
	172
	173	if (!(file->f_mode & FMODE_WRITE))
	174	return -EPERM;
	175	if (seals & ~(unsigned int)F_ALL_SEALS)
	176	return -EINVAL;
177
178	inode_lock(inode);
179
180	file_seals = memfd_file_seals_ptr(file);
181	if (!file_seals) {
182	error = -EINVAL;
183	goto unlock;
184	}
185
186	if (*file_seals & F_SEAL_SEAL) {
187	error = -EPERM;
188	goto unlock;
189	}
190
191	if ((seals & F_SEAL_WRITE) && !(*file_seals & F_SEAL_WRITE)) {
192	error = mapping_deny_writable(file->f_mapping);
193	if (error)
194	goto unlock;
195
196	error = memfd_wait_for_pins(file->f_mapping);
197	if (error) {
198	mapping_allow_writable(file->f_mapping);
199	goto unlock;
200	}
201	}
202
203	*file_seals \|= seals;
204	error = 0;
205
206	unlock:
207	inode_unlock(inode);
208	return error;
209	}
210
211	static int memfd_get_seals(struct file *file)
212	{
213	unsigned int *seals = memfd_file_seals_ptr(file);
214
215	return seals ? *seals : -EINVAL;
216	}
217
218	long memfd_fcntl(struct file *file, unsigned int cmd, unsigned long arg)
219	{
220	long error;
221
222	switch (cmd) {
223	case F_ADD_SEALS:
224	/* disallow upper 32bit */
225	if (arg > UINT_MAX)
226	return -EINVAL;
227
228	error = memfd_add_seals(file, arg);
229	break;
230	case F_GET_SEALS:
231	error = memfd_get_seals(file);
232	break;
233	default:
234	error = -EINVAL;
235	break;
236	}
237
238	return error;
239	}
240
241	#define MFD_NAME_PREFIX "memfd:"
242	#define MFD_NAME_PREFIX_LEN (sizeof(MFD_NAME_PREFIX) - 1)
243	#define MFD_NAME_MAX_LEN (NAME_MAX - MFD_NAME_PREFIX_LEN)
244
245	#define MFD_ALL_FLAGS (MFD_CLOEXEC \| MFD_ALLOW_SEALING \| MFD_HUGETLB)
246
247	SYSCALL_DEFINE2(memfd_create,
248	const char __user *, uname,
249	unsigned int, flags)
250	{
251	unsigned int *file_seals;
252	struct file *file;
253	int fd, error;
254	char *name;
255	long len;
256
257	if (!(flags & MFD_HUGETLB)) {
258	if (flags & ~(unsigned int)MFD_ALL_FLAGS)
259	return -EINVAL;
260	} else {
261	/* Allow huge page size encoding in flags. */
262	if (flags & ~(unsigned int)(MFD_ALL_FLAGS \|
263	(MFD_HUGE_MASK << MFD_HUGE_SHIFT)))
264	return -EINVAL;
265	}
266
267	/* length includes terminating zero */
268	len = strnlen_user(uname, MFD_NAME_MAX_LEN + 1);
269	if (len <= 0)
270	return -EFAULT;
271	if (len > MFD_NAME_MAX_LEN + 1)
272	return -EINVAL;
273
274	name = kmalloc(len + MFD_NAME_PREFIX_LEN, GFP_KERNEL);
275	if (!name)
276	return -ENOMEM;
277
278	strcpy(name, MFD_NAME_PREFIX);
279	if (copy_from_user(&name[MFD_NAME_PREFIX_LEN], uname, len)) {
280	error = -EFAULT;
281	goto err_name;
282	}
283
284	/* terminating-zero may have changed after strnlen_user() returned */
285	if (name[len + MFD_NAME_PREFIX_LEN - 1]) {
286	error = -EFAULT;
287	goto err_name;
288	}
289
290	fd = get_unused_fd_flags((flags & MFD_CLOEXEC) ? O_CLOEXEC : 0);
291	if (fd < 0) {
292	error = fd;
293	goto err_name;
294	}
295
296	if (flags & MFD_HUGETLB) {
297	struct user_struct *user = NULL;
298
299	file = hugetlb_file_setup(name, 0, VM_NORESERVE, &user,
300	HUGETLB_ANONHUGE_INODE,
301	(flags >> MFD_HUGE_SHIFT) &
302	MFD_HUGE_MASK);
303	} else
304	file = shmem_file_setup(name, 0, VM_NORESERVE);
305	if (IS_ERR(file)) {
306	error = PTR_ERR(file);
307	goto err_fd;
308	}
309	file->f_mode \|= FMODE_LSEEK \| FMODE_PREAD \| FMODE_PWRITE;
c9c554f2	310	file->f_flags \|= O_LARGEFILE;
5d752600 MK	311
	312	if (flags & MFD_ALLOW_SEALING) {
	313	file_seals = memfd_file_seals_ptr(file);
	314	*file_seals &= ~F_SEAL_SEAL;
	315	}
	316
	317	fd_install(fd, file);
	318	kfree(name);
	319	return fd;
	320
	321	err_fd:
	322	put_unused_fd(fd);
	323	err_name:
	324	kfree(name);
	325	return error;
	326	}