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

// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (c) 2014 Christoph Hellwig.
 */
#include "xfs.h"
#include "xfs_shared.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_mount.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_bmap.h"
#include "xfs_iomap.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_leased_layouts(
	struct inode		*inode,
	uint			*iolock,
	bool			*did_unlock)
{
	struct xfs_inode	*ip = XFS_I(inode);
	int			error;

	while ((error = break_layout(inode, false)) == -EWOULDBLOCK) {
		xfs_iunlock(ip, *iolock);
		*did_unlock = true;
		error = break_layout(inode, true);
		*iolock &= ~XFS_IOLOCK_SHARED;
		*iolock |= XFS_IOLOCK_EXCL;
		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;

	/*
	 * The pNFS block layout spec actually supports reflink like
	 * functionality, but the Linux pNFS server doesn't implement it yet.
	 */
	if (xfs_is_reflink_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_leased_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);

	error = xfs_bmbt_to_iomap(ip, iomap, &imap, false);
	*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, false);
		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
b2441318	1	// SPDX-License-Identifier: GPL-2.0
52785112 CH	2	/*
	3	* Copyright (c) 2014 Christoph Hellwig.
	4	*/
	5	#include "xfs.h"
5467b34b	6	#include "xfs_shared.h"
52785112 CH	7	#include "xfs_format.h"
	8	#include "xfs_log_format.h"
	9	#include "xfs_trans_resv.h"
52785112 CH	10	#include "xfs_mount.h"
	11	#include "xfs_inode.h"
	12	#include "xfs_trans.h"
52785112	13	#include "xfs_bmap.h"
52785112	14	#include "xfs_iomap.h"
52785112	15
781355c6 CH	16	/*
	17	* Ensure that we do not have any outstanding pNFS layouts that can be used by
	18	* clients to directly read from or write to this inode. This must be called
	19	* before every operation that can remove blocks from the extent map.
	20	* Additionally we call it during the write operation, where aren't concerned
	21	* about exposing unallocated blocks but just want to provide basic
	22	* synchronization between a local writer and pNFS clients. mmap writes would
	23	* also benefit from this sort of synchronization, but due to the tricky locking
	24	* rules in the page fault path we don't bother.
	25	*/
	26	int
69eb5fa1	27	xfs_break_leased_layouts(
781355c6	28	struct inode *inode,
69eb5fa1 DW	29	uint *iolock,
69eb5fa1 DW	30	bool *did_unlock)
781355c6 CH	31	{
	32	struct xfs_inode *ip = XFS_I(inode);
	33	int error;
	34
b6827160	35	while ((error = break_layout(inode, false)) == -EWOULDBLOCK) {
781355c6	36	xfs_iunlock(ip, *iolock);
69eb5fa1	37	*did_unlock = true;
781355c6	38	error = break_layout(inode, true);
c63a8eae DW	39	*iolock &= ~XFS_IOLOCK_SHARED;
c63a8eae DW	40	*iolock \|= XFS_IOLOCK_EXCL;
781355c6 CH	41	xfs_ilock(ip, *iolock);
	42	}
	43
	44	return error;
	45	}
	46
52785112 CH	47	/*
	48	* Get a unique ID including its location so that the client can identify
	49	* the exported device.
	50	*/
	51	int
	52	xfs_fs_get_uuid(
	53	struct super_block *sb,
	54	u8 *buf,
	55	u32 *len,
	56	u64 *offset)
	57	{
	58	struct xfs_mount *mp = XFS_M(sb);
	59
	60	printk_once(KERN_NOTICE
	61	"XFS (%s): using experimental pNFS feature, use at your own risk!\n",
	62	mp->m_fsname);
	63
	64	if (*len < sizeof(uuid_t))
	65	return -EINVAL;
	66
	67	memcpy(buf, &mp->m_sb.sb_uuid, sizeof(uuid_t));
	68	*len = sizeof(uuid_t);
	69	*offset = offsetof(struct xfs_dsb, sb_uuid);
	70	return 0;
	71	}
	72
52785112 CH	73	/*
	74	* Get a layout for the pNFS client.
	75	*/
	76	int
	77	xfs_fs_map_blocks(
	78	struct inode *inode,
	79	loff_t offset,
	80	u64 length,
	81	struct iomap *iomap,
	82	bool write,
	83	u32 *device_generation)
	84	{
	85	struct xfs_inode *ip = XFS_I(inode);
	86	struct xfs_mount *mp = ip->i_mount;
	87	struct xfs_bmbt_irec imap;
	88	xfs_fileoff_t offset_fsb, end_fsb;
	89	loff_t limit;
	90	int bmapi_flags = XFS_BMAPI_ENTIRE;
	91	int nimaps = 1;
	92	uint lock_flags;
	93	int error = 0;
	94
	95	if (XFS_FORCED_SHUTDOWN(mp))
	96	return -EIO;
	97
	98	/*
	99	* We can't export inodes residing on the realtime device. The realtime
	100	* device doesn't have a UUID to identify it, so the client has no way
	101	* to find it.
	102	*/
	103	if (XFS_IS_REALTIME_INODE(ip))
	104	return -ENXIO;
	105
46eeb521 DW	106	/*
	107	* The pNFS block layout spec actually supports reflink like
	108	* functionality, but the Linux pNFS server doesn't implement it yet.
	109	*/
	110	if (xfs_is_reflink_inode(ip))
	111	return -ENXIO;
	112
52785112 CH	113	/*
	114	* Lock out any other I/O before we flush and invalidate the pagecache,
	115	* and then hand out a layout to the remote system. This is very
	116	* similar to direct I/O, except that the synchronization is much more
69eb5fa1 DW	117	* complicated. See the comment near xfs_break_leased_layouts
69eb5fa1 DW	118	* for a detailed explanation.
52785112 CH	119	*/
	120	xfs_ilock(ip, XFS_IOLOCK_EXCL);
	121
	122	error = -EINVAL;
	123	limit = mp->m_super->s_maxbytes;
	124	if (!write)
	125	limit = max(limit, round_up(i_size_read(inode),
	126	inode->i_sb->s_blocksize));
	127	if (offset > limit)
	128	goto out_unlock;
	129	if (offset > limit - length)
	130	length = limit - offset;
	131
	132	error = filemap_write_and_wait(inode->i_mapping);
	133	if (error)
	134	goto out_unlock;
	135	error = invalidate_inode_pages2(inode->i_mapping);
	136	if (WARN_ON_ONCE(error))
	137	return error;
	138
	139	end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + length);
	140	offset_fsb = XFS_B_TO_FSBT(mp, offset);
	141
	142	lock_flags = xfs_ilock_data_map_shared(ip);
	143	error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb,
	144	&imap, &nimaps, bmapi_flags);
	145	xfs_iunlock(ip, lock_flags);
	146
	147	if (error)
	148	goto out_unlock;
	149
	150	if (write) {
	151	enum xfs_prealloc_flags flags = 0;
	152
	153	ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
	154
	155	if (!nimaps \|\| imap.br_startblock == HOLESTARTBLOCK) {
009c6e87 BF	156	/*
	157	* xfs_iomap_write_direct() expects to take ownership of
	158	* the shared ilock.
	159	*/
	160	xfs_ilock(ip, XFS_ILOCK_SHARED);
52785112 CH	161	error = xfs_iomap_write_direct(ip, offset, length,
	162	&imap, nimaps);
	163	if (error)
	164	goto out_unlock;
	165
	166	/*
	167	* Ensure the next transaction is committed
	168	* synchronously so that the blocks allocated and
	169	* handed out to the client are guaranteed to be
	170	* present even after a server crash.
	171	*/
	172	flags \|= XFS_PREALLOC_SET \| XFS_PREALLOC_SYNC;
	173	}
	174
	175	error = xfs_update_prealloc_flags(ip, flags);
	176	if (error)
	177	goto out_unlock;
	178	}
	179	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	180
16be1433	181	error = xfs_bmbt_to_iomap(ip, iomap, &imap, false);
52785112 CH	182	*device_generation = mp->m_generation;
	183	return error;
	184	out_unlock:
	185	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	186	return error;
	187	}
	188
	189	/*
	190	* Ensure the size update falls into a valid allocated block.
	191	*/
	192	static int
	193	xfs_pnfs_validate_isize(
	194	struct xfs_inode *ip,
	195	xfs_off_t isize)
	196	{
	197	struct xfs_bmbt_irec imap;
	198	int nimaps = 1;
	199	int error = 0;
	200
	201	xfs_ilock(ip, XFS_ILOCK_SHARED);
	202	error = xfs_bmapi_read(ip, XFS_B_TO_FSBT(ip->i_mount, isize - 1), 1,
	203	&imap, &nimaps, 0);
	204	xfs_iunlock(ip, XFS_ILOCK_SHARED);
	205	if (error)
	206	return error;
	207
	208	if (imap.br_startblock == HOLESTARTBLOCK \|\|
	209	imap.br_startblock == DELAYSTARTBLOCK \|\|
	210	imap.br_state == XFS_EXT_UNWRITTEN)
	211	return -EIO;
	212	return 0;
	213	}
	214
	215	/*
	216	* Make sure the blocks described by maps are stable on disk. This includes
	217	* converting any unwritten extents, flushing the disk cache and updating the
	218	* time stamps.
	219	*
	220	* Note that we rely on the caller to always send us a timestamp update so that
	221	* we always commit a transaction here. If that stops being true we will have
	222	* to manually flush the cache here similar to what the fsync code path does
	223	* for datasyncs on files that have no dirty metadata.
	224	*/
	225	int
	226	xfs_fs_commit_blocks(
	227	struct inode *inode,
	228	struct iomap *maps,
	229	int nr_maps,
	230	struct iattr *iattr)
	231	{
	232	struct xfs_inode *ip = XFS_I(inode);
	233	struct xfs_mount *mp = ip->i_mount;
	234	struct xfs_trans *tp;
	235	bool update_isize = false;
	236	int error, i;
	237	loff_t size;
	238
	239	ASSERT(iattr->ia_valid & (ATTR_ATIME\|ATTR_CTIME\|ATTR_MTIME));
	240
	241	xfs_ilock(ip, XFS_IOLOCK_EXCL);
	242
	243	size = i_size_read(inode);
	244	if ((iattr->ia_valid & ATTR_SIZE) && iattr->ia_size > size) {
	245	update_isize = true;
246	size = iattr->ia_size;
247	}
248
249	for (i = 0; i < nr_maps; i++) {
250	u64 start, length, end;
251
252	start = maps[i].offset;
253	if (start > size)
254	continue;
255
256	end = start + maps[i].length;
257	if (end > size)
258	end = size;
259
260	length = end - start;
261	if (!length)
262	continue;
263
264	/*
265	* Make sure reads through the pagecache see the new data.
266	*/
267	error = invalidate_inode_pages2_range(inode->i_mapping,
09cbfeaf KS	268	start >> PAGE_SHIFT,
09cbfeaf KS	269	(end - 1) >> PAGE_SHIFT);
52785112 CH	270	WARN_ON_ONCE(error);
52785112 CH	271
ee70daab	272	error = xfs_iomap_write_unwritten(ip, start, length, false);
52785112 CH	273	if (error)
	274	goto out_drop_iolock;
	275	}
	276
	277	if (update_isize) {
	278	error = xfs_pnfs_validate_isize(ip, size);
	279	if (error)
	280	goto out_drop_iolock;
	281	}
	282
253f4911 CH	283	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
253f4911 CH	284	if (error)
52785112 CH	285	goto out_drop_iolock;
	286
	287	xfs_ilock(ip, XFS_ILOCK_EXCL);
	288	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	289	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	290
	291	xfs_setattr_time(ip, iattr);
	292	if (update_isize) {
	293	i_size_write(inode, iattr->ia_size);
	294	ip->i_d.di_size = iattr->ia_size;
	295	}
	296
	297	xfs_trans_set_sync(tp);
70393313	298	error = xfs_trans_commit(tp);
52785112 CH	299
	300	out_drop_iolock:
	301	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	302	return error;
	303	}