[linux-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"
#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_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_super->s_id);

	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);

	ASSERT(!nimaps || imap.br_startblock != DELAYSTARTBLOCK);

	if (!error && write &&
	    (!nimaps || imap.br_startblock == HOLESTARTBLOCK)) {
		if (offset + length > XFS_ISIZE(ip))
			end_fsb = xfs_iomap_eof_align_last_fsb(ip, end_fsb);
		else if (nimaps && imap.br_startblock == HOLESTARTBLOCK)
			end_fsb = min(end_fsb, imap.br_startoff +
					       imap.br_blockcount);
		xfs_iunlock(ip, lock_flags);

		error = xfs_iomap_write_direct(ip, offset_fsb,
				end_fsb - offset_fsb, &imap);
		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.
		 */
		error = xfs_update_prealloc_flags(ip,
				XFS_PREALLOC_SET | XFS_PREALLOC_SYNC);
		if (error)
			goto out_unlock;
	} else {
		xfs_iunlock(ip, lock_flags);
	}
	xfs_iunlock(ip, XFS_IOLOCK_EXCL);

	error = xfs_bmbt_to_iomap(ip, iomap, &imap, 0);
	*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"
1aa63006	15	#include "xfs_pnfs.h"
52785112	16
781355c6 CH	17	/*
	18	* Ensure that we do not have any outstanding pNFS layouts that can be used by
	19	* clients to directly read from or write to this inode. This must be called
	20	* before every operation that can remove blocks from the extent map.
	21	* Additionally we call it during the write operation, where aren't concerned
	22	* about exposing unallocated blocks but just want to provide basic
	23	* synchronization between a local writer and pNFS clients. mmap writes would
	24	* also benefit from this sort of synchronization, but due to the tricky locking
	25	* rules in the page fault path we don't bother.
	26	*/
	27	int
69eb5fa1	28	xfs_break_leased_layouts(
781355c6	29	struct inode *inode,
69eb5fa1 DW	30	uint *iolock,
69eb5fa1 DW	31	bool *did_unlock)
781355c6 CH	32	{
	33	struct xfs_inode *ip = XFS_I(inode);
	34	int error;
	35
b6827160	36	while ((error = break_layout(inode, false)) == -EWOULDBLOCK) {
781355c6	37	xfs_iunlock(ip, *iolock);
69eb5fa1	38	*did_unlock = true;
781355c6	39	error = break_layout(inode, true);
c63a8eae DW	40	*iolock &= ~XFS_IOLOCK_SHARED;
c63a8eae DW	41	*iolock \|= XFS_IOLOCK_EXCL;
781355c6 CH	42	xfs_ilock(ip, *iolock);
	43	}
	44
	45	return error;
	46	}
	47
52785112 CH	48	/*
	49	* Get a unique ID including its location so that the client can identify
	50	* the exported device.
	51	*/
	52	int
	53	xfs_fs_get_uuid(
	54	struct super_block *sb,
	55	u8 *buf,
	56	u32 *len,
	57	u64 *offset)
	58	{
	59	struct xfs_mount *mp = XFS_M(sb);
	60
	61	printk_once(KERN_NOTICE
	62	"XFS (%s): using experimental pNFS feature, use at your own risk!\n",
e1d3d218	63	mp->m_super->s_id);
52785112 CH	64
	65	if (*len < sizeof(uuid_t))
	66	return -EINVAL;
	67
	68	memcpy(buf, &mp->m_sb.sb_uuid, sizeof(uuid_t));
	69	*len = sizeof(uuid_t);
	70	*offset = offsetof(struct xfs_dsb, sb_uuid);
	71	return 0;
	72	}
	73
52785112 CH	74	/*
	75	* Get a layout for the pNFS client.
	76	*/
	77	int
	78	xfs_fs_map_blocks(
	79	struct inode *inode,
	80	loff_t offset,
	81	u64 length,
	82	struct iomap *iomap,
	83	bool write,
	84	u32 *device_generation)
	85	{
	86	struct xfs_inode *ip = XFS_I(inode);
	87	struct xfs_mount *mp = ip->i_mount;
	88	struct xfs_bmbt_irec imap;
	89	xfs_fileoff_t offset_fsb, end_fsb;
	90	loff_t limit;
	91	int bmapi_flags = XFS_BMAPI_ENTIRE;
	92	int nimaps = 1;
	93	uint lock_flags;
	94	int error = 0;
	95
	96	if (XFS_FORCED_SHUTDOWN(mp))
	97	return -EIO;
	98
	99	/*
	100	* We can't export inodes residing on the realtime device. The realtime
	101	* device doesn't have a UUID to identify it, so the client has no way
	102	* to find it.
	103	*/
	104	if (XFS_IS_REALTIME_INODE(ip))
	105	return -ENXIO;
	106
46eeb521 DW	107	/*
	108	* The pNFS block layout spec actually supports reflink like
	109	* functionality, but the Linux pNFS server doesn't implement it yet.
	110	*/
	111	if (xfs_is_reflink_inode(ip))
	112	return -ENXIO;
	113
52785112 CH	114	/*
	115	* Lock out any other I/O before we flush and invalidate the pagecache,
	116	* and then hand out a layout to the remote system. This is very
	117	* similar to direct I/O, except that the synchronization is much more
69eb5fa1 DW	118	* complicated. See the comment near xfs_break_leased_layouts
69eb5fa1 DW	119	* for a detailed explanation.
52785112 CH	120	*/
	121	xfs_ilock(ip, XFS_IOLOCK_EXCL);
	122
	123	error = -EINVAL;
	124	limit = mp->m_super->s_maxbytes;
	125	if (!write)
	126	limit = max(limit, round_up(i_size_read(inode),
	127	inode->i_sb->s_blocksize));
	128	if (offset > limit)
	129	goto out_unlock;
	130	if (offset > limit - length)
	131	length = limit - offset;
	132
	133	error = filemap_write_and_wait(inode->i_mapping);
	134	if (error)
	135	goto out_unlock;
	136	error = invalidate_inode_pages2(inode->i_mapping);
	137	if (WARN_ON_ONCE(error))
	138	return error;
	139
	140	end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + length);
	141	offset_fsb = XFS_B_TO_FSBT(mp, offset);
	142
	143	lock_flags = xfs_ilock_data_map_shared(ip);
	144	error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb,
	145	&imap, &nimaps, bmapi_flags);
52785112	146
88cdb714 CH	147	ASSERT(!nimaps \|\| imap.br_startblock != DELAYSTARTBLOCK);
88cdb714 CH	148
e696663a CH	149	if (!error && write &&
	150	(!nimaps \|\| imap.br_startblock == HOLESTARTBLOCK)) {
	151	if (offset + length > XFS_ISIZE(ip))
	152	end_fsb = xfs_iomap_eof_align_last_fsb(ip, end_fsb);
	153	else if (nimaps && imap.br_startblock == HOLESTARTBLOCK)
	154	end_fsb = min(end_fsb, imap.br_startoff +
	155	imap.br_blockcount);
	156	xfs_iunlock(ip, lock_flags);
	157
	158	error = xfs_iomap_write_direct(ip, offset_fsb,
	159	end_fsb - offset_fsb, &imap);
307cdb54 CH	160	if (error)
	161	goto out_unlock;
	162
	163	/*
	164	* Ensure the next transaction is committed synchronously so
	165	* that the blocks allocated and handed out to the client are
	166	* guaranteed to be present even after a server crash.
	167	*/
	168	error = xfs_update_prealloc_flags(ip,
	169	XFS_PREALLOC_SET \| XFS_PREALLOC_SYNC);
52785112 CH	170	if (error)
52785112 CH	171	goto out_unlock;
e696663a CH	172	} else {
e696663a CH	173	xfs_iunlock(ip, lock_flags);
52785112 CH	174	}
	175	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	176
2492a606	177	error = xfs_bmbt_to_iomap(ip, iomap, &imap, 0);
52785112 CH	178	*device_generation = mp->m_generation;
	179	return error;
	180	out_unlock:
	181	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	182	return error;
	183	}
	184
	185	/*
	186	* Ensure the size update falls into a valid allocated block.
	187	*/
	188	static int
	189	xfs_pnfs_validate_isize(
	190	struct xfs_inode *ip,
	191	xfs_off_t isize)
	192	{
	193	struct xfs_bmbt_irec imap;
	194	int nimaps = 1;
	195	int error = 0;
	196
	197	xfs_ilock(ip, XFS_ILOCK_SHARED);
	198	error = xfs_bmapi_read(ip, XFS_B_TO_FSBT(ip->i_mount, isize - 1), 1,
	199	&imap, &nimaps, 0);
	200	xfs_iunlock(ip, XFS_ILOCK_SHARED);
	201	if (error)
	202	return error;
	203
	204	if (imap.br_startblock == HOLESTARTBLOCK \|\|
	205	imap.br_startblock == DELAYSTARTBLOCK \|\|
	206	imap.br_state == XFS_EXT_UNWRITTEN)
	207	return -EIO;
	208	return 0;
	209	}
	210
	211	/*
	212	* Make sure the blocks described by maps are stable on disk. This includes
	213	* converting any unwritten extents, flushing the disk cache and updating the
	214	* time stamps.
	215	*
	216	* Note that we rely on the caller to always send us a timestamp update so that
	217	* we always commit a transaction here. If that stops being true we will have
	218	* to manually flush the cache here similar to what the fsync code path does
	219	* for datasyncs on files that have no dirty metadata.
	220	*/
	221	int
	222	xfs_fs_commit_blocks(
	223	struct inode *inode,
	224	struct iomap *maps,
	225	int nr_maps,
	226	struct iattr *iattr)
	227	{
	228	struct xfs_inode *ip = XFS_I(inode);
	229	struct xfs_mount *mp = ip->i_mount;
	230	struct xfs_trans *tp;
	231	bool update_isize = false;
	232	int error, i;
	233	loff_t size;
	234
	235	ASSERT(iattr->ia_valid & (ATTR_ATIME\|ATTR_CTIME\|ATTR_MTIME));
	236
	237	xfs_ilock(ip, XFS_IOLOCK_EXCL);
	238
	239	size = i_size_read(inode);
	240	if ((iattr->ia_valid & ATTR_SIZE) && iattr->ia_size > size) {
	241	update_isize = true;
242	size = iattr->ia_size;
243	}
244
245	for (i = 0; i < nr_maps; i++) {
246	u64 start, length, end;
247
248	start = maps[i].offset;
249	if (start > size)
250	continue;
251
252	end = start + maps[i].length;
253	if (end > size)
254	end = size;
255
256	length = end - start;
257	if (!length)
258	continue;
259
260	/*
261	* Make sure reads through the pagecache see the new data.
262	*/
263	error = invalidate_inode_pages2_range(inode->i_mapping,
09cbfeaf KS	264	start >> PAGE_SHIFT,
09cbfeaf KS	265	(end - 1) >> PAGE_SHIFT);
52785112 CH	266	WARN_ON_ONCE(error);
52785112 CH	267
ee70daab	268	error = xfs_iomap_write_unwritten(ip, start, length, false);
52785112 CH	269	if (error)
	270	goto out_drop_iolock;
	271	}
	272
	273	if (update_isize) {
	274	error = xfs_pnfs_validate_isize(ip, size);
	275	if (error)
	276	goto out_drop_iolock;
	277	}
	278
253f4911 CH	279	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
253f4911 CH	280	if (error)
52785112 CH	281	goto out_drop_iolock;
	282
	283	xfs_ilock(ip, XFS_ILOCK_EXCL);
	284	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	285	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	286
	287	xfs_setattr_time(ip, iattr);
	288	if (update_isize) {
	289	i_size_write(inode, iattr->ia_size);
	290	ip->i_d.di_size = iattr->ia_size;
	291	}
	292
	293	xfs_trans_set_sync(tp);
70393313	294	error = xfs_trans_commit(tp);
52785112 CH	295
	296	out_drop_iolock:
	297	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	298	return error;
	299	}