Merge tag 'xfs-pnfs-for-linus-3.20-rc1' of git://git.kernel.org/pub/scm/linux/ke…

…rnel/git/dgc/linux-xfs Pull xfs pnfs block layout support from Dave Chinner: "This contains the changes to XFS needed to support the PNFS block layout server that you pulled in through Bruce's NFS server tree merge. I originally thought that I'd need to merge changes into the NFS server side, but Bruce had already picked them up and so this is purely changes to the fs/xfs/ codebase. Summary: This update contains the implementation of the PNFS server export methods that enable use of XFS filesystems as a block layout target" * tag 'xfs-pnfs-for-linus-3.20-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs: xfs: recall pNFS layouts on conflicting access xfs: implement pNFS export operations

Merge tag 'xfs-pnfs-for-linus-3.20-rc1' of git://git.kernel.org/pub/scm/linux/ke…
…rnel/git/dgc/linux-xfs Pull xfs pnfs block layout support from Dave Chinner: "This contains the changes to XFS needed to support the PNFS block layout server that you pulled in through Bruce's NFS server tree merge. I originally thought that I'd need to merge changes into the NFS server side, but Bruce had already picked them up and so this is purely changes to the fs/xfs/ codebase. Summary: This update contains the implementation of the PNFS server export methods that enable use of XFS filesystems as a block layout target" * tag 'xfs-pnfs-for-linus-3.20-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs: xfs: recall pNFS layouts on conflicting access xfs: implement pNFS export operations
Linus Torvalds
2 parents 24a52e412e 781355c6e5
Showing 10 changed files Side-by-side Diff
fs/xfs/Makefile
fs/xfs/xfs_export.c
fs/xfs/xfs_file.c
fs/xfs/xfs_fsops.c
fs/xfs/xfs_ioctl.c
fs/xfs/xfs_iops.c
fs/xfs/xfs_iops.h
fs/xfs/xfs_mount.h
fs/xfs/xfs_pnfs.c
fs/xfs/xfs_pnfs.h
@@ -121,4 +121,5 @@
 xfs-$(CONFIG_PROC_FS)		+= xfs_stats.o
 xfs-$(CONFIG_SYSCTL)		+= xfs_sysctl.o
 xfs-$(CONFIG_COMPAT)		+= xfs_ioctl32.o
+xfs-$(CONFIG_NFSD_PNFS)		+= xfs_pnfs.o
@@ -30,6 +30,7 @@
 #include "xfs_trace.h"
 #include "xfs_icache.h"
 #include "xfs_log.h"
+#include "xfs_pnfs.h"
  
 /*
  * Note that we only accept fileids which are long enough rather than allow
@@ -245,5 +246,10 @@
 	.fh_to_parent		= xfs_fs_fh_to_parent,
 	.get_parent		= xfs_fs_get_parent,
 	.commit_metadata	= xfs_fs_nfs_commit_metadata,
+#ifdef CONFIG_NFSD_PNFS
+	.get_uuid		= xfs_fs_get_uuid,
+	.map_blocks		= xfs_fs_map_blocks,
+	.commit_blocks		= xfs_fs_commit_blocks,
+#endif
 };
@@ -36,6 +36,7 @@
 #include "xfs_trace.h"
 #include "xfs_log.h"
 #include "xfs_icache.h"
+#include "xfs_pnfs.h"
  
 #include <linux/aio.h>
 #include <linux/dcache.h>
@@ -554,6 +555,10 @@
 	if (error)
 		return error;
  
+	error = xfs_break_layouts(inode, iolock);
+	if (error)
+		return error;
+
 	/*
 	 * If the offset is beyond the size of the file, we need to zero any
 	 * blocks that fall between the existing EOF and the start of this
@@ -822,6 +827,7 @@
 	struct xfs_inode	*ip = XFS_I(inode);
 	long			error;
 	enum xfs_prealloc_flags	flags = 0;
+	uint			iolock = XFS_IOLOCK_EXCL;
 	loff_t			new_size = 0;
  
 	if (!S_ISREG(inode->i_mode))
@@ -830,7 +836,11 @@
 		     FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE))
 		return -EOPNOTSUPP;
  
-	xfs_ilock(ip, XFS_IOLOCK_EXCL);
+	xfs_ilock(ip, iolock);
+	error = xfs_break_layouts(inode, &iolock);
+	if (error)
+		goto out_unlock;
+
 	if (mode & FALLOC_FL_PUNCH_HOLE) {
 		error = xfs_free_file_space(ip, offset, len);
 		if (error)
@@ -894,7 +904,7 @@
 	}
  
 out_unlock:
-	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
+	xfs_iunlock(ip, iolock);
 	return error;
 }
  
@@ -602,6 +602,12 @@
 	if (!mutex_trylock(&mp->m_growlock))
 		return -EWOULDBLOCK;
 	error = xfs_growfs_data_private(mp, in);
+	/*
+	 * Increment the generation unconditionally, the error could be from
+	 * updating the secondary superblocks, in which case the new size
+	 * is live already.
+	 */
+	mp->m_generation++;
 	mutex_unlock(&mp->m_growlock);
 	return error;
 }
@@ -39,6 +39,7 @@
 #include "xfs_icache.h"
 #include "xfs_symlink.h"
 #include "xfs_trans.h"
+#include "xfs_pnfs.h"
  
 #include <linux/capability.h>
 #include <linux/dcache.h>
@@ -608,6 +609,7 @@
 {
 	struct iattr		iattr;
 	enum xfs_prealloc_flags	flags = 0;
+	uint			iolock = XFS_IOLOCK_EXCL;
 	int			error;
  
 	/*
@@ -636,7 +638,10 @@
 	if (error)
 		return error;
  
-	xfs_ilock(ip, XFS_IOLOCK_EXCL);
+	xfs_ilock(ip, iolock);
+	error = xfs_break_layouts(inode, &iolock);
+	if (error)
+		goto out_unlock;
  
 	switch (bf->l_whence) {
 	case 0: /*SEEK_SET*/
@@ -725,7 +730,7 @@
 	error = xfs_update_prealloc_flags(ip, flags);
  
 out_unlock:
-	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
+	xfs_iunlock(ip, iolock);
 	mnt_drop_write_file(filp);
 	return error;
 }
@@ -37,6 +37,7 @@
 #include "xfs_da_btree.h"
 #include "xfs_dir2.h"
 #include "xfs_trans_space.h"
+#include "xfs_pnfs.h"
  
 #include <linux/capability.h>
 #include <linux/xattr.h>
@@ -505,7 +506,7 @@
 	inode->i_mode |= mode & ~S_IFMT;
 }
  
-static void
+void
 xfs_setattr_time(
 	struct xfs_inode	*ip,
 	struct iattr		*iattr)
@@ -979,9 +980,13 @@
 	int			error;
  
 	if (iattr->ia_valid & ATTR_SIZE) {
-		xfs_ilock(ip, XFS_IOLOCK_EXCL);
-		error = xfs_setattr_size(ip, iattr);
-		xfs_iunlock(ip, XFS_IOLOCK_EXCL);
+		uint		iolock = XFS_IOLOCK_EXCL;
+
+		xfs_ilock(ip, iolock);
+		error = xfs_break_layouts(dentry->d_inode, &iolock);
+		if (!error)
+			error = xfs_setattr_size(ip, iattr);
+		xfs_iunlock(ip, iolock);
 	} else {
 		error = xfs_setattr_nonsize(ip, iattr, 0);
 	}
@@ -32,6 +32,7 @@
  */
 #define XFS_ATTR_NOACL		0x01	/* Don't call posix_acl_chmod */
  
+extern void xfs_setattr_time(struct xfs_inode *ip, struct iattr *iattr);
 extern int xfs_setattr_nonsize(struct xfs_inode *ip, struct iattr *vap,
 			       int flags);
 extern int xfs_setattr_size(struct xfs_inode *ip, struct iattr *vap);
@@ -174,6 +174,17 @@
 	struct workqueue_struct	*m_reclaim_workqueue;
 	struct workqueue_struct	*m_log_workqueue;
 	struct workqueue_struct *m_eofblocks_workqueue;
+
+	/*
+	 * Generation of the filesysyem layout.  This is incremented by each
+	 * growfs, and used by the pNFS server to ensure the client updates
+	 * its view of the block device once it gets a layout that might
+	 * reference the newly added blocks.  Does not need to be persistent
+	 * as long as we only allow file system size increments, but if we
+	 * ever support shrinks it would have to be persisted in addition
+	 * to various other kinds of pain inflicted on the pNFS server.
+	 */
+	__uint32_t		m_generation;
 } xfs_mount_t;
  
 /*
+/*
+ * 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)
+{
+	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);
+		error = break_layout(inode, true);
+		*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;
+}
+
+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) {
+			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)
+		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, 0);
+
+out_drop_iolock:
+	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
+	return error;
+}
+#ifndef _XFS_PNFS_H
+#define _XFS_PNFS_H 1
+
+#ifdef CONFIG_NFSD_PNFS
+int xfs_fs_get_uuid(struct super_block *sb, u8 *buf, u32 *len, u64 *offset);
+int xfs_fs_map_blocks(struct inode *inode, loff_t offset, u64 length,
+		struct iomap *iomap, bool write, u32 *device_generation);
+int xfs_fs_commit_blocks(struct inode *inode, struct iomap *maps, int nr_maps,
+		struct iattr *iattr);
+
+int xfs_break_layouts(struct inode *inode, uint *iolock);
+#else
+static inline int xfs_break_layouts(struct inode *inode, uint *iolock)
+{
+	return 0;
+}
+#endif /* CONFIG_NFSD_PNFS */
+#endif /* _XFS_PNFS_H */
...	...	@@ -121,4 +121,5 @@
121	121	xfs-$(CONFIG_PROC_FS) += xfs_stats.o
122	122	xfs-$(CONFIG_SYSCTL) += xfs_sysctl.o
123	123	xfs-$(CONFIG_COMPAT) += xfs_ioctl32.o
	124	+xfs-$(CONFIG_NFSD_PNFS) += xfs_pnfs.o
...	...	@@ -30,6 +30,7 @@
30	30	#include "xfs_trace.h"
31	31	#include "xfs_icache.h"
32	32	#include "xfs_log.h"
	33	+#include "xfs_pnfs.h"
33	34
34	35	/*
35	36	* Note that we only accept fileids which are long enough rather than allow
...	...	@@ -245,5 +246,10 @@
245	246	.fh_to_parent = xfs_fs_fh_to_parent,
246	247	.get_parent = xfs_fs_get_parent,
247	248	.commit_metadata = xfs_fs_nfs_commit_metadata,
	249	+#ifdef CONFIG_NFSD_PNFS
	250	+ .get_uuid = xfs_fs_get_uuid,
	251	+ .map_blocks = xfs_fs_map_blocks,
	252	+ .commit_blocks = xfs_fs_commit_blocks,
	253	+#endif
248	254	};
...	...	@@ -36,6 +36,7 @@
36	36	#include "xfs_trace.h"
37	37	#include "xfs_log.h"
38	38	#include "xfs_icache.h"
	39	+#include "xfs_pnfs.h"
39	40
40	41	#include <linux/aio.h>
41	42	#include <linux/dcache.h>
...	...	@@ -554,6 +555,10 @@
554	555	if (error)
555	556	return error;
556	557
	558	+ error = xfs_break_layouts(inode, iolock);
	559	+ if (error)
	560	+ return error;
	561	+
557	562	/*
558	563	* If the offset is beyond the size of the file, we need to zero any
559	564	* blocks that fall between the existing EOF and the start of this
...	...	@@ -822,6 +827,7 @@
822	827	struct xfs_inode *ip = XFS_I(inode);
823	828	long error;
824	829	enum xfs_prealloc_flags flags = 0;
	830	+ uint iolock = XFS_IOLOCK_EXCL;
825	831	loff_t new_size = 0;
826	832
827	833	if (!S_ISREG(inode->i_mode))
...	...	@@ -830,7 +836,11 @@
830	836	FALLOC_FL_COLLAPSE_RANGE \| FALLOC_FL_ZERO_RANGE))
831	837	return -EOPNOTSUPP;
832	838
833		- xfs_ilock(ip, XFS_IOLOCK_EXCL);
	839	+ xfs_ilock(ip, iolock);
	840	+ error = xfs_break_layouts(inode, &iolock);
	841	+ if (error)
	842	+ goto out_unlock;
	843	+
834	844	if (mode & FALLOC_FL_PUNCH_HOLE) {
835	845	error = xfs_free_file_space(ip, offset, len);
836	846	if (error)
...	...	@@ -894,7 +904,7 @@
894	904	}
895	905
896	906	out_unlock:
897		- xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	907	+ xfs_iunlock(ip, iolock);
898	908	return error;
899	909	}
900	910
...	...	@@ -602,6 +602,12 @@
602	602	if (!mutex_trylock(&mp->m_growlock))
603	603	return -EWOULDBLOCK;
604	604	error = xfs_growfs_data_private(mp, in);
	605	+ /*
	606	+ * Increment the generation unconditionally, the error could be from
	607	+ * updating the secondary superblocks, in which case the new size
	608	+ * is live already.
	609	+ */
	610	+ mp->m_generation++;
605	611	mutex_unlock(&mp->m_growlock);
606	612	return error;
607	613	}
...	...	@@ -39,6 +39,7 @@
39	39	#include "xfs_icache.h"
40	40	#include "xfs_symlink.h"
41	41	#include "xfs_trans.h"
	42	+#include "xfs_pnfs.h"
42	43
43	44	#include <linux/capability.h>
44	45	#include <linux/dcache.h>
...	...	@@ -608,6 +609,7 @@
608	609	{
609	610	struct iattr iattr;
610	611	enum xfs_prealloc_flags flags = 0;
	612	+ uint iolock = XFS_IOLOCK_EXCL;
611	613	int error;
612	614
613	615	/*
...	...	@@ -636,7 +638,10 @@
636	638	if (error)
637	639	return error;
638	640
639		- xfs_ilock(ip, XFS_IOLOCK_EXCL);
	641	+ xfs_ilock(ip, iolock);
	642	+ error = xfs_break_layouts(inode, &iolock);
	643	+ if (error)
	644	+ goto out_unlock;
640	645
641	646	switch (bf->l_whence) {
642	647	case 0: /SEEK_SET/
...	...	@@ -725,7 +730,7 @@
725	730	error = xfs_update_prealloc_flags(ip, flags);
726	731
727	732	out_unlock:
728		- xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	733	+ xfs_iunlock(ip, iolock);
729	734	mnt_drop_write_file(filp);
730	735	return error;
731	736	}
...	...	@@ -37,6 +37,7 @@
37	37	#include "xfs_da_btree.h"
38	38	#include "xfs_dir2.h"
39	39	#include "xfs_trans_space.h"
	40	+#include "xfs_pnfs.h"
40	41
41	42	#include <linux/capability.h>
42	43	#include <linux/xattr.h>
...	...	@@ -505,7 +506,7 @@
505	506	inode->i_mode \|= mode & ~S_IFMT;
506	507	}
507	508
508		-static void
	509	+void
509	510	xfs_setattr_time(
510	511	struct xfs_inode *ip,
511	512	struct iattr *iattr)
...	...	@@ -979,9 +980,13 @@
979	980	int error;
980	981
981	982	if (iattr->ia_valid & ATTR_SIZE) {
982		- xfs_ilock(ip, XFS_IOLOCK_EXCL);
983		- error = xfs_setattr_size(ip, iattr);
984		- xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	983	+ uint iolock = XFS_IOLOCK_EXCL;
	984	+
	985	+ xfs_ilock(ip, iolock);
	986	+ error = xfs_break_layouts(dentry->d_inode, &iolock);
	987	+ if (!error)
	988	+ error = xfs_setattr_size(ip, iattr);
	989	+ xfs_iunlock(ip, iolock);
985	990	} else {
986	991	error = xfs_setattr_nonsize(ip, iattr, 0);
987	992	}
...	...	@@ -32,6 +32,7 @@
32	32	*/
33	33	#define XFS_ATTR_NOACL 0x01 /* Don't call posix_acl_chmod */
34	34
	35	+extern void xfs_setattr_time(struct xfs_inode ip, struct iattr iattr);
35	36	extern int xfs_setattr_nonsize(struct xfs_inode ip, struct iattr vap,
36	37	int flags);
37	38	extern int xfs_setattr_size(struct xfs_inode ip, struct iattr vap);
...	...	@@ -174,6 +174,17 @@
174	174	struct workqueue_struct *m_reclaim_workqueue;
175	175	struct workqueue_struct *m_log_workqueue;
176	176	struct workqueue_struct *m_eofblocks_workqueue;
	177	+
	178	+ /*
	179	+ * Generation of the filesysyem layout. This is incremented by each
	180	+ * growfs, and used by the pNFS server to ensure the client updates
	181	+ * its view of the block device once it gets a layout that might
	182	+ * reference the newly added blocks. Does not need to be persistent
	183	+ * as long as we only allow file system size increments, but if we
	184	+ * ever support shrinks it would have to be persisted in addition
	185	+ * to various other kinds of pain inflicted on the pNFS server.
	186	+ */
	187	+ __uint32_t m_generation;
177	188	} xfs_mount_t;
178	189
179	190	/*
	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	+
	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,
	34	+ uint *iolock)
	35	+{
	36	+ struct xfs_inode *ip = XFS_I(inode);
	37	+ int error;
	38	+
	39	+ ASSERT(xfs_isilocked(ip, XFS_IOLOCK_SHARED\|XFS_IOLOCK_EXCL));
	40	+
	41	+ while ((error = break_layout(inode, false) == -EWOULDBLOCK)) {
	42	+ xfs_iunlock(ip, *iolock);
	43	+ error = break_layout(inode, true);
	44	+ *iolock = XFS_IOLOCK_EXCL;
	45	+ xfs_ilock(ip, *iolock);
	46	+ }
	47	+
	48	+ return error;
	49	+}
	50	+
	51	+/*
	52	+ * Get a unique ID including its location so that the client can identify
	53	+ * the exported device.
	54	+ */
	55	+int
	56	+xfs_fs_get_uuid(
	57	+ struct super_block *sb,
	58	+ u8 *buf,
	59	+ u32 *len,
	60	+ u64 *offset)
	61	+{
	62	+ struct xfs_mount *mp = XFS_M(sb);
	63	+
	64	+ printk_once(KERN_NOTICE
	65	+"XFS (%s): using experimental pNFS feature, use at your own risk!\n",
	66	+ mp->m_fsname);
	67	+
	68	+ if (*len < sizeof(uuid_t))
	69	+ return -EINVAL;
	70	+
	71	+ memcpy(buf, &mp->m_sb.sb_uuid, sizeof(uuid_t));
	72	+ *len = sizeof(uuid_t);
	73	+ *offset = offsetof(struct xfs_dsb, sb_uuid);
	74	+ return 0;
	75	+}
	76	+
	77	+static void
	78	+xfs_bmbt_to_iomap(
	79	+ struct xfs_inode *ip,
	80	+ struct iomap *iomap,
	81	+ struct xfs_bmbt_irec *imap)
	82	+{
	83	+ struct xfs_mount *mp = ip->i_mount;
	84	+
	85	+ if (imap->br_startblock == HOLESTARTBLOCK) {
	86	+ iomap->blkno = IOMAP_NULL_BLOCK;
	87	+ iomap->type = IOMAP_HOLE;
	88	+ } else if (imap->br_startblock == DELAYSTARTBLOCK) {
	89	+ iomap->blkno = IOMAP_NULL_BLOCK;
	90	+ iomap->type = IOMAP_DELALLOC;
	91	+ } else {
	92	+ iomap->blkno =
	93	+ XFS_FSB_TO_DADDR(ip->i_mount, imap->br_startblock);
	94	+ if (imap->br_state == XFS_EXT_UNWRITTEN)
	95	+ iomap->type = IOMAP_UNWRITTEN;
	96	+ else
	97	+ iomap->type = IOMAP_MAPPED;
	98	+ }
	99	+ iomap->offset = XFS_FSB_TO_B(mp, imap->br_startoff);
	100	+ iomap->length = XFS_FSB_TO_B(mp, imap->br_blockcount);
	101	+}
	102	+
	103	+/*
	104	+ * Get a layout for the pNFS client.
	105	+ */
	106	+int
	107	+xfs_fs_map_blocks(
	108	+ struct inode *inode,
	109	+ loff_t offset,
	110	+ u64 length,
	111	+ struct iomap *iomap,
	112	+ bool write,
	113	+ u32 *device_generation)
	114	+{
	115	+ struct xfs_inode *ip = XFS_I(inode);
	116	+ struct xfs_mount *mp = ip->i_mount;
	117	+ struct xfs_bmbt_irec imap;
	118	+ xfs_fileoff_t offset_fsb, end_fsb;
	119	+ loff_t limit;
	120	+ int bmapi_flags = XFS_BMAPI_ENTIRE;
	121	+ int nimaps = 1;
	122	+ uint lock_flags;
	123	+ int error = 0;
	124	+
	125	+ if (XFS_FORCED_SHUTDOWN(mp))
	126	+ return -EIO;
	127	+
	128	+ /*
	129	+ * We can't export inodes residing on the realtime device. The realtime
	130	+ * device doesn't have a UUID to identify it, so the client has no way
	131	+ * to find it.
	132	+ */
	133	+ if (XFS_IS_REALTIME_INODE(ip))
	134	+ return -ENXIO;
	135	+
	136	+ /*
	137	+ * Lock out any other I/O before we flush and invalidate the pagecache,
	138	+ * and then hand out a layout to the remote system. This is very
	139	+ * similar to direct I/O, except that the synchronization is much more
	140	+ * complicated. See the comment near xfs_break_layouts for a detailed
	141	+ * explanation.
	142	+ */
	143	+ xfs_ilock(ip, XFS_IOLOCK_EXCL);
	144	+
	145	+ error = -EINVAL;
	146	+ limit = mp->m_super->s_maxbytes;
	147	+ if (!write)
	148	+ limit = max(limit, round_up(i_size_read(inode),
	149	+ inode->i_sb->s_blocksize));
	150	+ if (offset > limit)
	151	+ goto out_unlock;
	152	+ if (offset > limit - length)
	153	+ length = limit - offset;
	154	+
	155	+ error = filemap_write_and_wait(inode->i_mapping);
	156	+ if (error)
	157	+ goto out_unlock;
	158	+ error = invalidate_inode_pages2(inode->i_mapping);
	159	+ if (WARN_ON_ONCE(error))
	160	+ return error;
	161	+
	162	+ end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + length);
	163	+ offset_fsb = XFS_B_TO_FSBT(mp, offset);
	164	+
	165	+ lock_flags = xfs_ilock_data_map_shared(ip);
	166	+ error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb,
	167	+ &imap, &nimaps, bmapi_flags);
	168	+ xfs_iunlock(ip, lock_flags);
	169	+
	170	+ if (error)
	171	+ goto out_unlock;
	172	+
	173	+ if (write) {
	174	+ enum xfs_prealloc_flags flags = 0;
	175	+
	176	+ ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
	177	+
	178	+ if (!nimaps \|\| imap.br_startblock == HOLESTARTBLOCK) {
	179	+ error = xfs_iomap_write_direct(ip, offset, length,
	180	+ &imap, nimaps);
	181	+ if (error)
	182	+ goto out_unlock;
	183	+
	184	+ /*
	185	+ * Ensure the next transaction is committed
	186	+ * synchronously so that the blocks allocated and
	187	+ * handed out to the client are guaranteed to be
	188	+ * present even after a server crash.
	189	+ */
	190	+ flags \|= XFS_PREALLOC_SET \| XFS_PREALLOC_SYNC;
	191	+ }
	192	+
	193	+ error = xfs_update_prealloc_flags(ip, flags);
	194	+ if (error)
	195	+ goto out_unlock;
	196	+ }
	197	+ xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	198	+
	199	+ xfs_bmbt_to_iomap(ip, iomap, &imap);
	200	+ *device_generation = mp->m_generation;
	201	+ return error;
	202	+out_unlock:
	203	+ xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	204	+ return error;
	205	+}
	206	+
	207	+/*
	208	+ * Ensure the size update falls into a valid allocated block.
	209	+ */
	210	+static int
	211	+xfs_pnfs_validate_isize(
	212	+ struct xfs_inode *ip,
	213	+ xfs_off_t isize)
	214	+{
	215	+ struct xfs_bmbt_irec imap;
	216	+ int nimaps = 1;
	217	+ int error = 0;
	218	+
	219	+ xfs_ilock(ip, XFS_ILOCK_SHARED);
	220	+ error = xfs_bmapi_read(ip, XFS_B_TO_FSBT(ip->i_mount, isize - 1), 1,
	221	+ &imap, &nimaps, 0);
	222	+ xfs_iunlock(ip, XFS_ILOCK_SHARED);
	223	+ if (error)
	224	+ return error;
	225	+
	226	+ if (imap.br_startblock == HOLESTARTBLOCK \|\|
	227	+ imap.br_startblock == DELAYSTARTBLOCK \|\|
	228	+ imap.br_state == XFS_EXT_UNWRITTEN)
	229	+ return -EIO;
	230	+ return 0;
	231	+}
	232	+
	233	+/*
	234	+ * Make sure the blocks described by maps are stable on disk. This includes
	235	+ * converting any unwritten extents, flushing the disk cache and updating the
	236	+ * time stamps.
	237	+ *
	238	+ * Note that we rely on the caller to always send us a timestamp update so that
	239	+ * we always commit a transaction here. If that stops being true we will have
	240	+ * to manually flush the cache here similar to what the fsync code path does
	241	+ * for datasyncs on files that have no dirty metadata.
	242	+ */
	243	+int
	244	+xfs_fs_commit_blocks(
	245	+ struct inode *inode,
	246	+ struct iomap *maps,
	247	+ int nr_maps,
	248	+ struct iattr *iattr)
	249	+{
	250	+ struct xfs_inode *ip = XFS_I(inode);
	251	+ struct xfs_mount *mp = ip->i_mount;
	252	+ struct xfs_trans *tp;
	253	+ bool update_isize = false;
	254	+ int error, i;
	255	+ loff_t size;
	256	+
	257	+ ASSERT(iattr->ia_valid & (ATTR_ATIME\|ATTR_CTIME\|ATTR_MTIME));
	258	+
	259	+ xfs_ilock(ip, XFS_IOLOCK_EXCL);
	260	+
	261	+ size = i_size_read(inode);
	262	+ if ((iattr->ia_valid & ATTR_SIZE) && iattr->ia_size > size) {
	263	+ update_isize = true;
	264	+ size = iattr->ia_size;
	265	+ }
	266	+
	267	+ for (i = 0; i < nr_maps; i++) {
	268	+ u64 start, length, end;
	269	+
	270	+ start = maps[i].offset;
	271	+ if (start > size)
	272	+ continue;
	273	+
	274	+ end = start + maps[i].length;
	275	+ if (end > size)
	276	+ end = size;
	277	+
	278	+ length = end - start;
	279	+ if (!length)
	280	+ continue;
	281	+
	282	+ /*
	283	+ * Make sure reads through the pagecache see the new data.
	284	+ */
	285	+ error = invalidate_inode_pages2_range(inode->i_mapping,
	286	+ start >> PAGE_CACHE_SHIFT,
	287	+ (end - 1) >> PAGE_CACHE_SHIFT);
	288	+ WARN_ON_ONCE(error);
	289	+
	290	+ error = xfs_iomap_write_unwritten(ip, start, length);
	291	+ if (error)
	292	+ goto out_drop_iolock;
	293	+ }
	294	+
	295	+ if (update_isize) {
	296	+ error = xfs_pnfs_validate_isize(ip, size);
	297	+ if (error)
	298	+ goto out_drop_iolock;
	299	+ }
	300	+
	301	+ tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
	302	+ error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
	303	+ if (error)
	304	+ goto out_drop_iolock;
	305	+
	306	+ xfs_ilock(ip, XFS_ILOCK_EXCL);
	307	+ xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	308	+ xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	309	+
	310	+ xfs_setattr_time(ip, iattr);
	311	+ if (update_isize) {
	312	+ i_size_write(inode, iattr->ia_size);
	313	+ ip->i_d.di_size = iattr->ia_size;
	314	+ }
	315	+
	316	+ xfs_trans_set_sync(tp);
	317	+ error = xfs_trans_commit(tp, 0);
	318	+
	319	+out_drop_iolock:
	320	+ xfs_iunlock(ip, XFS_IOLOCK_EXCL);
	321	+ return error;
	322	+}
	1	+#ifndef _XFS_PNFS_H
	2	+#define _XFS_PNFS_H 1
	3	+
	4	+#ifdef CONFIG_NFSD_PNFS
	5	+int xfs_fs_get_uuid(struct super_block sb, u8 buf, u32 len, u64 offset);
	6	+int xfs_fs_map_blocks(struct inode *inode, loff_t offset, u64 length,
	7	+ struct iomap iomap, bool write, u32 device_generation);
	8	+int xfs_fs_commit_blocks(struct inode inode, struct iomap maps, int nr_maps,
	9	+ struct iattr *iattr);
	10	+
	11	+int xfs_break_layouts(struct inode inode, uint iolock);
	12	+#else
	13	+static inline int xfs_break_layouts(struct inode inode, uint iolock)
	14	+{
	15	+ return 0;
	16	+}
	17	+#endif /* CONFIG_NFSD_PNFS */
	18	+#endif /* _XFS_PNFS_H */