[linux-2.6-block.git] / fs / xfs / xfs_pnfs.c

/*
 * Copyright (c) 2014 Christoph Hellwig.
 */
#include <linux/iomap.h>
#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;
}

/*
 * 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_SHIFT,
					(end - 1) >> PAGE_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;
	}

	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
	if (error)
		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	*/
199a31c6	4	#include <linux/iomap.h>
52785112 CH	5	#include "xfs.h"
	6	#include "xfs_format.h"
	7	#include "xfs_log_format.h"
	8	#include "xfs_trans_resv.h"
	9	#include "xfs_sb.h"
	10	#include "xfs_mount.h"
	11	#include "xfs_inode.h"
	12	#include "xfs_trans.h"
	13	#include "xfs_log.h"
	14	#include "xfs_bmap.h"
	15	#include "xfs_bmap_util.h"
	16	#include "xfs_error.h"
	17	#include "xfs_iomap.h"
	18	#include "xfs_shared.h"
	19	#include "xfs_bit.h"
	20	#include "xfs_pnfs.h"
	21
781355c6 CH	22	/*
	23	* Ensure that we do not have any outstanding pNFS layouts that can be used by
	24	* clients to directly read from or write to this inode. This must be called
	25	* before every operation that can remove blocks from the extent map.
	26	* Additionally we call it during the write operation, where aren't concerned
	27	* about exposing unallocated blocks but just want to provide basic
	28	* synchronization between a local writer and pNFS clients. mmap writes would
	29	* also benefit from this sort of synchronization, but due to the tricky locking
	30	* rules in the page fault path we don't bother.
	31	*/
	32	int
	33	xfs_break_layouts(
	34	struct inode *inode,
21c3ea18 CH	35	uint *iolock,
21c3ea18 CH	36	bool with_imutex)
781355c6 CH	37	{
	38	struct xfs_inode *ip = XFS_I(inode);
	39	int error;
	40
	41	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_SHARED\|XFS_IOLOCK_EXCL));
	42
	43	while ((error = break_layout(inode, false) == -EWOULDBLOCK)) {
	44	xfs_iunlock(ip, *iolock);
21c3ea18	45	if (with_imutex && (*iolock & XFS_IOLOCK_EXCL))
5955102c	46	inode_unlock(inode);
781355c6 CH	47	error = break_layout(inode, true);
781355c6 CH	48	*iolock = XFS_IOLOCK_EXCL;
21c3ea18	49	if (with_imutex)
5955102c	50	inode_lock(inode);
781355c6 CH	51	xfs_ilock(ip, *iolock);
	52	}
	53
	54	return error;
	55	}
	56
52785112 CH	57	/*
	58	* Get a unique ID including its location so that the client can identify
	59	* the exported device.
	60	*/
	61	int
	62	xfs_fs_get_uuid(
	63	struct super_block *sb,
	64	u8 *buf,
	65	u32 *len,
	66	u64 *offset)
	67	{
	68	struct xfs_mount *mp = XFS_M(sb);
	69
	70	printk_once(KERN_NOTICE
	71	"XFS (%s): using experimental pNFS feature, use at your own risk!\n",
	72	mp->m_fsname);
	73
	74	if (*len < sizeof(uuid_t))
	75	return -EINVAL;
	76
	77	memcpy(buf, &mp->m_sb.sb_uuid, sizeof(uuid_t));
	78	*len = sizeof(uuid_t);
	79	*offset = offsetof(struct xfs_dsb, sb_uuid);
	80	return 0;
	81	}
	82
52785112 CH	83	/*
	84	* Get a layout for the pNFS client.
	85	*/
	86	int
	87	xfs_fs_map_blocks(
	88	struct inode *inode,
	89	loff_t offset,
	90	u64 length,
	91	struct iomap *iomap,
	92	bool write,
	93	u32 *device_generation)
	94	{
	95	struct xfs_inode *ip = XFS_I(inode);
	96	struct xfs_mount *mp = ip->i_mount;
	97	struct xfs_bmbt_irec imap;
	98	xfs_fileoff_t offset_fsb, end_fsb;
	99	loff_t limit;
	100	int bmapi_flags = XFS_BMAPI_ENTIRE;
	101	int nimaps = 1;
	102	uint lock_flags;
	103	int error = 0;
	104
	105	if (XFS_FORCED_SHUTDOWN(mp))
	106	return -EIO;
	107
	108	/*
	109	* We can't export inodes residing on the realtime device. The realtime
	110	* device doesn't have a UUID to identify it, so the client has no way
	111	* to find it.
	112	*/
	113	if (XFS_IS_REALTIME_INODE(ip))
	114	return -ENXIO;
	115
	116	/*
	117	* Lock out any other I/O before we flush and invalidate the pagecache,
	118	* and then hand out a layout to the remote system. This is very
	119	* similar to direct I/O, except that the synchronization is much more
	120	* complicated. See the comment near xfs_break_layouts for a detailed
	121	* explanation.
	122	*/
	123	xfs_ilock(ip, XFS_IOLOCK_EXCL);
	124
	125	error = -EINVAL;
	126	limit = mp->m_super->s_maxbytes;
	127	if (!write)
	128	limit = max(limit, round_up(i_size_read(inode),
	129	inode->i_sb->s_blocksize));
	130	if (offset > limit)
	131	goto out_unlock;
	132	if (offset > limit - length)
	133	length = limit - offset;
	134
	135	error = filemap_write_and_wait(inode->i_mapping);
	136	if (error)
	137	goto out_unlock;
	138	error = invalidate_inode_pages2(inode->i_mapping);
	139	if (WARN_ON_ONCE(error))
	140	return error;
	141
	142	end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + length);
	143	offset_fsb = XFS_B_TO_FSBT(mp, offset);
	144
	145	lock_flags = xfs_ilock_data_map_shared(ip);
	146	error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb,
147	&imap, &nimaps, bmapi_flags);
148	xfs_iunlock(ip, lock_flags);
149
150	if (error)
151	goto out_unlock;
152
153	if (write) {
154	enum xfs_prealloc_flags flags = 0;
155
156	ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
157
158	if (!nimaps \|\| imap.br_startblock == HOLESTARTBLOCK) {
009c6e87 BF	159	/*
	160	* xfs_iomap_write_direct() expects to take ownership of
	161	* the shared ilock.
	162	*/
	163	xfs_ilock(ip, XFS_ILOCK_SHARED);
52785112 CH	164	error = xfs_iomap_write_direct(ip, offset, length,
	165	&imap, nimaps);
	166	if (error)
	167	goto out_unlock;
	168
	169	/*
	170	* Ensure the next transaction is committed
	171	* synchronously so that the blocks allocated and
	172	* handed out to the client are guaranteed to be
	173	* present even after a server crash.
	174	*/
	175	flags \|= XFS_PREALLOC_SET \| XFS_PREALLOC_SYNC;
	176	}
	177
	178	error = xfs_update_prealloc_flags(ip, flags);
	179	if (error)
	180	goto out_unlock;
	181	}
	182	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	183
	184	xfs_bmbt_to_iomap(ip, iomap, &imap);
	185	*device_generation = mp->m_generation;
	186	return error;
	187	out_unlock:
	188	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	189	return error;
	190	}
	191
	192	/*
	193	* Ensure the size update falls into a valid allocated block.
	194	*/
	195	static int
	196	xfs_pnfs_validate_isize(
	197	struct xfs_inode *ip,
	198	xfs_off_t isize)
	199	{
	200	struct xfs_bmbt_irec imap;
	201	int nimaps = 1;
	202	int error = 0;
	203
	204	xfs_ilock(ip, XFS_ILOCK_SHARED);
	205	error = xfs_bmapi_read(ip, XFS_B_TO_FSBT(ip->i_mount, isize - 1), 1,
	206	&imap, &nimaps, 0);
	207	xfs_iunlock(ip, XFS_ILOCK_SHARED);
	208	if (error)
	209	return error;
	210
	211	if (imap.br_startblock == HOLESTARTBLOCK \|\|
	212	imap.br_startblock == DELAYSTARTBLOCK \|\|
	213	imap.br_state == XFS_EXT_UNWRITTEN)
	214	return -EIO;
	215	return 0;
	216	}
	217
	218	/*
	219	* Make sure the blocks described by maps are stable on disk. This includes
	220	* converting any unwritten extents, flushing the disk cache and updating the
	221	* time stamps.
	222	*
	223	* Note that we rely on the caller to always send us a timestamp update so that
	224	* we always commit a transaction here. If that stops being true we will have
	225	* to manually flush the cache here similar to what the fsync code path does
	226	* for datasyncs on files that have no dirty metadata.
	227	*/
228	int
229	xfs_fs_commit_blocks(
230	struct inode *inode,
231	struct iomap *maps,
232	int nr_maps,
233	struct iattr *iattr)
234	{
235	struct xfs_inode *ip = XFS_I(inode);
236	struct xfs_mount *mp = ip->i_mount;
237	struct xfs_trans *tp;
238	bool update_isize = false;
239	int error, i;
240	loff_t size;
241
242	ASSERT(iattr->ia_valid & (ATTR_ATIME\|ATTR_CTIME\|ATTR_MTIME));
243
244	xfs_ilock(ip, XFS_IOLOCK_EXCL);
245
246	size = i_size_read(inode);
247	if ((iattr->ia_valid & ATTR_SIZE) && iattr->ia_size > size) {
248	update_isize = true;
249	size = iattr->ia_size;
250	}
251
252	for (i = 0; i < nr_maps; i++) {
253	u64 start, length, end;
254
255	start = maps[i].offset;
256	if (start > size)
257	continue;
258
259	end = start + maps[i].length;
260	if (end > size)
261	end = size;
262
263	length = end - start;
264	if (!length)
265	continue;
266
267	/*
268	* Make sure reads through the pagecache see the new data.
269	*/
270	error = invalidate_inode_pages2_range(inode->i_mapping,
09cbfeaf KS	271	start >> PAGE_SHIFT,
09cbfeaf KS	272	(end - 1) >> PAGE_SHIFT);
52785112 CH	273	WARN_ON_ONCE(error);
	274
	275	error = xfs_iomap_write_unwritten(ip, start, length);
	276	if (error)
	277	goto out_drop_iolock;
	278	}
	279
	280	if (update_isize) {
	281	error = xfs_pnfs_validate_isize(ip, size);
	282	if (error)
	283	goto out_drop_iolock;
	284	}
	285
253f4911 CH	286	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
253f4911 CH	287	if (error)
52785112 CH	288	goto out_drop_iolock;
	289
	290	xfs_ilock(ip, XFS_ILOCK_EXCL);
	291	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	292	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	293
	294	xfs_setattr_time(ip, iattr);
	295	if (update_isize) {
	296	i_size_write(inode, iattr->ia_size);
	297	ip->i_d.di_size = iattr->ia_size;
	298	}
	299
	300	xfs_trans_set_sync(tp);
70393313	301	error = xfs_trans_commit(tp);
52785112 CH	302
	303	out_drop_iolock:
	304	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	305	return error;
	306	}