[linux.git] / fs / xfs / xfs_pnfs.c

/*
 * Copyright (c) 2014 Christoph Hellwig.
 */
#include "xfs.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_sb.h"
#include "xfs_mount.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_log.h"
#include "xfs_bmap.h"
#include "xfs_bmap_util.h"
#include "xfs_error.h"
#include "xfs_iomap.h"
#include "xfs_shared.h"
#include "xfs_bit.h"
#include "xfs_pnfs.h"

/*
 * Ensure that we do not have any outstanding pNFS layouts that can be used by
 * clients to directly read from or write to this inode.  This must be called
 * before every operation that can remove blocks from the extent map.
 * Additionally we call it during the write operation, where aren't concerned
 * about exposing unallocated blocks but just want to provide basic
 * synchronization between a local writer and pNFS clients.  mmap writes would
 * also benefit from this sort of synchronization, but due to the tricky locking
 * rules in the page fault path we don't bother.
 */
int
xfs_break_layouts(
	struct inode		*inode,
	uint			*iolock,
	bool			with_imutex)
{
	struct xfs_inode	*ip = XFS_I(inode);
	int			error;

	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL));

	while ((error = break_layout(inode, false) == -EWOULDBLOCK)) {
		xfs_iunlock(ip, *iolock);
		if (with_imutex && (*iolock & XFS_IOLOCK_EXCL))
			inode_unlock(inode);
		error = break_layout(inode, true);
		*iolock = XFS_IOLOCK_EXCL;
		if (with_imutex)
			inode_lock(inode);
		xfs_ilock(ip, *iolock);
	}

	return error;
}

/*
 * Get a unique ID including its location so that the client can identify
 * the exported device.
 */
int
xfs_fs_get_uuid(
	struct super_block	*sb,
	u8			*buf,
	u32			*len,
	u64			*offset)
{
	struct xfs_mount	*mp = XFS_M(sb);

	printk_once(KERN_NOTICE
"XFS (%s): using experimental pNFS feature, use at your own risk!\n",
		mp->m_fsname);

	if (*len < sizeof(uuid_t))
		return -EINVAL;

	memcpy(buf, &mp->m_sb.sb_uuid, sizeof(uuid_t));
	*len = sizeof(uuid_t);
	*offset = offsetof(struct xfs_dsb, sb_uuid);
	return 0;
}

static void
xfs_bmbt_to_iomap(
	struct xfs_inode	*ip,
	struct iomap		*iomap,
	struct xfs_bmbt_irec	*imap)
{
	struct xfs_mount	*mp = ip->i_mount;

	if (imap->br_startblock == HOLESTARTBLOCK) {
		iomap->blkno = IOMAP_NULL_BLOCK;
		iomap->type = IOMAP_HOLE;
	} else if (imap->br_startblock == DELAYSTARTBLOCK) {
		iomap->blkno = IOMAP_NULL_BLOCK;
		iomap->type = IOMAP_DELALLOC;
	} else {
		iomap->blkno =
			XFS_FSB_TO_DADDR(ip->i_mount, imap->br_startblock);
		if (imap->br_state == XFS_EXT_UNWRITTEN)
			iomap->type = IOMAP_UNWRITTEN;
		else
			iomap->type = IOMAP_MAPPED;
	}
	iomap->offset = XFS_FSB_TO_B(mp, imap->br_startoff);
	iomap->length = XFS_FSB_TO_B(mp, imap->br_blockcount);
}

/*
 * Get a layout for the pNFS client.
 */
int
xfs_fs_map_blocks(
	struct inode		*inode,
	loff_t			offset,
	u64			length,
	struct iomap		*iomap,
	bool			write,
	u32			*device_generation)
{
	struct xfs_inode	*ip = XFS_I(inode);
	struct xfs_mount	*mp = ip->i_mount;
	struct xfs_bmbt_irec	imap;
	xfs_fileoff_t		offset_fsb, end_fsb;
	loff_t			limit;
	int			bmapi_flags = XFS_BMAPI_ENTIRE;
	int			nimaps = 1;
	uint			lock_flags;
	int			error = 0;

	if (XFS_FORCED_SHUTDOWN(mp))
		return -EIO;

	/*
	 * We can't export inodes residing on the realtime device.  The realtime
	 * device doesn't have a UUID to identify it, so the client has no way
	 * to find it.
	 */
	if (XFS_IS_REALTIME_INODE(ip))
		return -ENXIO;

	/*
	 * Lock out any other I/O before we flush and invalidate the pagecache,
	 * and then hand out a layout to the remote system.  This is very
	 * similar to direct I/O, except that the synchronization is much more
	 * complicated.  See the comment near xfs_break_layouts for a detailed
	 * explanation.
	 */
	xfs_ilock(ip, XFS_IOLOCK_EXCL);

	error = -EINVAL;
	limit = mp->m_super->s_maxbytes;
	if (!write)
		limit = max(limit, round_up(i_size_read(inode),
				     inode->i_sb->s_blocksize));
	if (offset > limit)
		goto out_unlock;
	if (offset > limit - length)
		length = limit - offset;

	error = filemap_write_and_wait(inode->i_mapping);
	if (error)
		goto out_unlock;
	error = invalidate_inode_pages2(inode->i_mapping);
	if (WARN_ON_ONCE(error))
		return error;

	end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + length);
	offset_fsb = XFS_B_TO_FSBT(mp, offset);

	lock_flags = xfs_ilock_data_map_shared(ip);
	error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb,
				&imap, &nimaps, bmapi_flags);
	xfs_iunlock(ip, lock_flags);

	if (error)
		goto out_unlock;

	if (write) {
		enum xfs_prealloc_flags	flags = 0;

		ASSERT(imap.br_startblock != DELAYSTARTBLOCK);

		if (!nimaps || imap.br_startblock == HOLESTARTBLOCK) {
			/*
			 * xfs_iomap_write_direct() expects to take ownership of
			 * the shared ilock.
			 */
			xfs_ilock(ip, XFS_ILOCK_SHARED);
			error = xfs_iomap_write_direct(ip, offset, length,
						       &imap, nimaps);
			if (error)
				goto out_unlock;

			/*
			 * Ensure the next transaction is committed
			 * synchronously so that the blocks allocated and
			 * handed out to the client are guaranteed to be
			 * present even after a server crash.
			 */
			flags |= XFS_PREALLOC_SET | XFS_PREALLOC_SYNC;
		}

		error = xfs_update_prealloc_flags(ip, flags);
		if (error)
			goto out_unlock;
	}
	xfs_iunlock(ip, XFS_IOLOCK_EXCL);

	xfs_bmbt_to_iomap(ip, iomap, &imap);
	*device_generation = mp->m_generation;
	return error;
out_unlock:
	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	return error;
}

/*
 * Ensure the size update falls into a valid allocated block.
 */
static int
xfs_pnfs_validate_isize(
	struct xfs_inode	*ip,
	xfs_off_t		isize)
{
	struct xfs_bmbt_irec	imap;
	int			nimaps = 1;
	int			error = 0;

	xfs_ilock(ip, XFS_ILOCK_SHARED);
	error = xfs_bmapi_read(ip, XFS_B_TO_FSBT(ip->i_mount, isize - 1), 1,
				&imap, &nimaps, 0);
	xfs_iunlock(ip, XFS_ILOCK_SHARED);
	if (error)
		return error;

	if (imap.br_startblock == HOLESTARTBLOCK ||
	    imap.br_startblock == DELAYSTARTBLOCK ||
	    imap.br_state == XFS_EXT_UNWRITTEN)
		return -EIO;
	return 0;
}

/*
 * Make sure the blocks described by maps are stable on disk.  This includes
 * converting any unwritten extents, flushing the disk cache and updating the
 * time stamps.
 *
 * Note that we rely on the caller to always send us a timestamp update so that
 * we always commit a transaction here.  If that stops being true we will have
 * to manually flush the cache here similar to what the fsync code path does
 * for datasyncs on files that have no dirty metadata.
 */
int
xfs_fs_commit_blocks(
	struct inode		*inode,
	struct iomap		*maps,
	int			nr_maps,
	struct iattr		*iattr)
{
	struct xfs_inode	*ip = XFS_I(inode);
	struct xfs_mount	*mp = ip->i_mount;
	struct xfs_trans	*tp;
	bool			update_isize = false;
	int			error, i;
	loff_t			size;

	ASSERT(iattr->ia_valid & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME));

	xfs_ilock(ip, XFS_IOLOCK_EXCL);

	size = i_size_read(inode);
	if ((iattr->ia_valid & ATTR_SIZE) && iattr->ia_size > size) {
		update_isize = true;
		size = iattr->ia_size;
	}

	for (i = 0; i < nr_maps; i++) {
		u64 start, length, end;

		start = maps[i].offset;
		if (start > size)
			continue;

		end = start + maps[i].length;
		if (end > size)
			end = size;

		length = end - start;
		if (!length)
			continue;
	
		/*
		 * Make sure reads through the pagecache see the new data.
		 */
		error = invalidate_inode_pages2_range(inode->i_mapping,
					start >> PAGE_CACHE_SHIFT,
					(end - 1) >> PAGE_CACHE_SHIFT);
		WARN_ON_ONCE(error);

		error = xfs_iomap_write_unwritten(ip, start, length);
		if (error)
			goto out_drop_iolock;
	}

	if (update_isize) {
		error = xfs_pnfs_validate_isize(ip, size);
		if (error)
			goto out_drop_iolock;
	}

	tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
	if (error) {
		xfs_trans_cancel(tp);
		goto out_drop_iolock;
	}

	xfs_ilock(ip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);

	xfs_setattr_time(ip, iattr);
	if (update_isize) {
		i_size_write(inode, iattr->ia_size);
		ip->i_d.di_size = iattr->ia_size;
	}

	xfs_trans_set_sync(tp);
	error = xfs_trans_commit(tp);

out_drop_iolock:
	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	return error;
}
Commit	Line	Data
52785112 CH	1	/*
	2	* Copyright (c) 2014 Christoph Hellwig.
	3	*/
	4	#include "xfs.h"
	5	#include "xfs_format.h"
	6	#include "xfs_log_format.h"
	7	#include "xfs_trans_resv.h"
	8	#include "xfs_sb.h"
	9	#include "xfs_mount.h"
	10	#include "xfs_inode.h"
	11	#include "xfs_trans.h"
	12	#include "xfs_log.h"
	13	#include "xfs_bmap.h"
	14	#include "xfs_bmap_util.h"
	15	#include "xfs_error.h"
	16	#include "xfs_iomap.h"
	17	#include "xfs_shared.h"
	18	#include "xfs_bit.h"
	19	#include "xfs_pnfs.h"
	20
781355c6 CH	21	/*
	22	* Ensure that we do not have any outstanding pNFS layouts that can be used by
	23	* clients to directly read from or write to this inode. This must be called
	24	* before every operation that can remove blocks from the extent map.
	25	* Additionally we call it during the write operation, where aren't concerned
	26	* about exposing unallocated blocks but just want to provide basic
	27	* synchronization between a local writer and pNFS clients. mmap writes would
	28	* also benefit from this sort of synchronization, but due to the tricky locking
	29	* rules in the page fault path we don't bother.
	30	*/
	31	int
	32	xfs_break_layouts(
	33	struct inode *inode,
21c3ea18 CH	34	uint *iolock,
21c3ea18 CH	35	bool with_imutex)
781355c6 CH	36	{
	37	struct xfs_inode *ip = XFS_I(inode);
	38	int error;
	39
	40	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_SHARED\|XFS_IOLOCK_EXCL));
	41
	42	while ((error = break_layout(inode, false) == -EWOULDBLOCK)) {
	43	xfs_iunlock(ip, *iolock);
21c3ea18	44	if (with_imutex && (*iolock & XFS_IOLOCK_EXCL))
5955102c	45	inode_unlock(inode);
781355c6 CH	46	error = break_layout(inode, true);
781355c6 CH	47	*iolock = XFS_IOLOCK_EXCL;
21c3ea18	48	if (with_imutex)
5955102c	49	inode_lock(inode);
781355c6 CH	50	xfs_ilock(ip, *iolock);
	51	}
	52
	53	return error;
	54	}
	55
52785112 CH	56	/*
	57	* Get a unique ID including its location so that the client can identify
	58	* the exported device.
	59	*/
	60	int
	61	xfs_fs_get_uuid(
	62	struct super_block *sb,
	63	u8 *buf,
	64	u32 *len,
	65	u64 *offset)
	66	{
	67	struct xfs_mount *mp = XFS_M(sb);
	68
	69	printk_once(KERN_NOTICE
	70	"XFS (%s): using experimental pNFS feature, use at your own risk!\n",
	71	mp->m_fsname);
	72
	73	if (*len < sizeof(uuid_t))
	74	return -EINVAL;
	75
	76	memcpy(buf, &mp->m_sb.sb_uuid, sizeof(uuid_t));
	77	*len = sizeof(uuid_t);
	78	*offset = offsetof(struct xfs_dsb, sb_uuid);
	79	return 0;
	80	}
	81
	82	static void
	83	xfs_bmbt_to_iomap(
	84	struct xfs_inode *ip,
	85	struct iomap *iomap,
	86	struct xfs_bmbt_irec *imap)
	87	{
	88	struct xfs_mount *mp = ip->i_mount;
	89
	90	if (imap->br_startblock == HOLESTARTBLOCK) {
	91	iomap->blkno = IOMAP_NULL_BLOCK;
	92	iomap->type = IOMAP_HOLE;
	93	} else if (imap->br_startblock == DELAYSTARTBLOCK) {
	94	iomap->blkno = IOMAP_NULL_BLOCK;
	95	iomap->type = IOMAP_DELALLOC;
	96	} else {
	97	iomap->blkno =
	98	XFS_FSB_TO_DADDR(ip->i_mount, imap->br_startblock);
	99	if (imap->br_state == XFS_EXT_UNWRITTEN)
	100	iomap->type = IOMAP_UNWRITTEN;
	101	else
	102	iomap->type = IOMAP_MAPPED;
	103	}
	104	iomap->offset = XFS_FSB_TO_B(mp, imap->br_startoff);
	105	iomap->length = XFS_FSB_TO_B(mp, imap->br_blockcount);
	106	}
	107
	108	/*
	109	* Get a layout for the pNFS client.
	110	*/
	111	int
	112	xfs_fs_map_blocks(
	113	struct inode *inode,
	114	loff_t offset,
	115	u64 length,
	116	struct iomap *iomap,
	117	bool write,
	118	u32 *device_generation)
	119	{
120	struct xfs_inode *ip = XFS_I(inode);
121	struct xfs_mount *mp = ip->i_mount;
122	struct xfs_bmbt_irec imap;
123	xfs_fileoff_t offset_fsb, end_fsb;
124	loff_t limit;
125	int bmapi_flags = XFS_BMAPI_ENTIRE;
126	int nimaps = 1;
127	uint lock_flags;
128	int error = 0;
129
130	if (XFS_FORCED_SHUTDOWN(mp))
131	return -EIO;
132
133	/*
134	* We can't export inodes residing on the realtime device. The realtime
135	* device doesn't have a UUID to identify it, so the client has no way
136	* to find it.
137	*/
138	if (XFS_IS_REALTIME_INODE(ip))
139	return -ENXIO;
140
141	/*
142	* Lock out any other I/O before we flush and invalidate the pagecache,
143	* and then hand out a layout to the remote system. This is very
144	* similar to direct I/O, except that the synchronization is much more
145	* complicated. See the comment near xfs_break_layouts for a detailed
146	* explanation.
147	*/
148	xfs_ilock(ip, XFS_IOLOCK_EXCL);
149
150	error = -EINVAL;
151	limit = mp->m_super->s_maxbytes;
152	if (!write)
153	limit = max(limit, round_up(i_size_read(inode),
154	inode->i_sb->s_blocksize));
155	if (offset > limit)
156	goto out_unlock;
157	if (offset > limit - length)
158	length = limit - offset;
159
160	error = filemap_write_and_wait(inode->i_mapping);
161	if (error)
162	goto out_unlock;
163	error = invalidate_inode_pages2(inode->i_mapping);
164	if (WARN_ON_ONCE(error))
165	return error;
166
167	end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + length);
168	offset_fsb = XFS_B_TO_FSBT(mp, offset);
169
170	lock_flags = xfs_ilock_data_map_shared(ip);
171	error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb,
172	&imap, &nimaps, bmapi_flags);
173	xfs_iunlock(ip, lock_flags);
174
175	if (error)
176	goto out_unlock;
177
178	if (write) {
179	enum xfs_prealloc_flags flags = 0;
180
181	ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
182
183	if (!nimaps \|\| imap.br_startblock == HOLESTARTBLOCK) {
009c6e87 BF	184	/*
	185	* xfs_iomap_write_direct() expects to take ownership of
	186	* the shared ilock.
	187	*/
	188	xfs_ilock(ip, XFS_ILOCK_SHARED);
52785112 CH	189	error = xfs_iomap_write_direct(ip, offset, length,
	190	&imap, nimaps);
	191	if (error)
	192	goto out_unlock;
	193
	194	/*
	195	* Ensure the next transaction is committed
	196	* synchronously so that the blocks allocated and
	197	* handed out to the client are guaranteed to be
	198	* present even after a server crash.
	199	*/
	200	flags \|= XFS_PREALLOC_SET \| XFS_PREALLOC_SYNC;
	201	}
	202
	203	error = xfs_update_prealloc_flags(ip, flags);
	204	if (error)
	205	goto out_unlock;
	206	}
	207	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	208
	209	xfs_bmbt_to_iomap(ip, iomap, &imap);
	210	*device_generation = mp->m_generation;
	211	return error;
	212	out_unlock:
	213	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	214	return error;
	215	}
	216
	217	/*
	218	* Ensure the size update falls into a valid allocated block.
	219	*/
	220	static int
	221	xfs_pnfs_validate_isize(
	222	struct xfs_inode *ip,
	223	xfs_off_t isize)
	224	{
	225	struct xfs_bmbt_irec imap;
	226	int nimaps = 1;
	227	int error = 0;
	228
	229	xfs_ilock(ip, XFS_ILOCK_SHARED);
	230	error = xfs_bmapi_read(ip, XFS_B_TO_FSBT(ip->i_mount, isize - 1), 1,
	231	&imap, &nimaps, 0);
	232	xfs_iunlock(ip, XFS_ILOCK_SHARED);
	233	if (error)
	234	return error;
	235
	236	if (imap.br_startblock == HOLESTARTBLOCK \|\|
	237	imap.br_startblock == DELAYSTARTBLOCK \|\|
	238	imap.br_state == XFS_EXT_UNWRITTEN)
	239	return -EIO;
	240	return 0;
	241	}
	242
	243	/*
	244	* Make sure the blocks described by maps are stable on disk. This includes
	245	* converting any unwritten extents, flushing the disk cache and updating the
	246	* time stamps.
	247	*
	248	* Note that we rely on the caller to always send us a timestamp update so that
	249	* we always commit a transaction here. If that stops being true we will have
	250	* to manually flush the cache here similar to what the fsync code path does
	251	* for datasyncs on files that have no dirty metadata.
	252	*/
253	int
254	xfs_fs_commit_blocks(
255	struct inode *inode,
256	struct iomap *maps,
257	int nr_maps,
258	struct iattr *iattr)
259	{
260	struct xfs_inode *ip = XFS_I(inode);
261	struct xfs_mount *mp = ip->i_mount;
262	struct xfs_trans *tp;
263	bool update_isize = false;
264	int error, i;
265	loff_t size;
266
267	ASSERT(iattr->ia_valid & (ATTR_ATIME\|ATTR_CTIME\|ATTR_MTIME));
268
269	xfs_ilock(ip, XFS_IOLOCK_EXCL);
270
271	size = i_size_read(inode);
272	if ((iattr->ia_valid & ATTR_SIZE) && iattr->ia_size > size) {
273	update_isize = true;
274	size = iattr->ia_size;
275	}
276
277	for (i = 0; i < nr_maps; i++) {
278	u64 start, length, end;
279
280	start = maps[i].offset;
281	if (start > size)
282	continue;
283
284	end = start + maps[i].length;
285	if (end > size)
286	end = size;
287
288	length = end - start;
289	if (!length)
290	continue;
291
292	/*
293	* Make sure reads through the pagecache see the new data.
294	*/
295	error = invalidate_inode_pages2_range(inode->i_mapping,
296	start >> PAGE_CACHE_SHIFT,
297	(end - 1) >> PAGE_CACHE_SHIFT);
298	WARN_ON_ONCE(error);
299
300	error = xfs_iomap_write_unwritten(ip, start, length);
301	if (error)
302	goto out_drop_iolock;
303	}
304
305	if (update_isize) {
306	error = xfs_pnfs_validate_isize(ip, size);
307	if (error)
308	goto out_drop_iolock;
309	}
310
311	tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
312	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
83d5f018	313	if (error) {
4906e215	314	xfs_trans_cancel(tp);
52785112	315	goto out_drop_iolock;
83d5f018	316	}
52785112 CH	317
	318	xfs_ilock(ip, XFS_ILOCK_EXCL);
	319	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	320	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	321
	322	xfs_setattr_time(ip, iattr);
	323	if (update_isize) {
	324	i_size_write(inode, iattr->ia_size);
	325	ip->i_d.di_size = iattr->ia_size;
	326	}
	327
	328	xfs_trans_set_sync(tp);
70393313	329	error = xfs_trans_commit(tp);
52785112 CH	330
	331	out_drop_iolock:
	332	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	333	return error;
	334	}