xfs: Fast zeroing of allocated space

index 1f279b0..24f0b0a 100644
@@ -32,10 +32,9 @@ xfs_tosspages(
 	xfs_off_t	last,
 	int		fiopt)
 {
-	struct address_space *mapping = VFS_I(ip)->i_mapping;
-
-	if (mapping->nrpages)
-		truncate_inode_pages(mapping, first);
+	/* can't toss partial tail pages, so mask them out */
+	last &= ~(PAGE_SIZE - 1);
+	truncate_inode_pages_range(VFS_I(ip)->i_mapping, first, last);
 }
 
 int

@@ -50,12 +49,11 @@ xfs_flushinval_pages(
 
 	trace_xfs_pagecache_inval(ip, first, last);
 
-	if (mapping->nrpages) {
-		xfs_iflags_clear(ip, XFS_ITRUNCATED);
-		ret = filemap_write_and_wait(mapping);
-		if (!ret)
-			truncate_inode_pages(mapping, first);
-	}
+	xfs_iflags_clear(ip, XFS_ITRUNCATED);
+	ret = filemap_write_and_wait_range(mapping, first,
+				last == -1 ? LLONG_MAX : last);
+	if (!ret)
+		truncate_inode_pages_range(mapping, first, last);
 	return -ret;
 }
 

@@ -71,10 +69,9 @@ xfs_flush_pages(
 	int		ret = 0;
 	int		ret2;
 
-	if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
-		xfs_iflags_clear(ip, XFS_ITRUNCATED);
-		ret = -filemap_fdatawrite(mapping);
-	}
+	xfs_iflags_clear(ip, XFS_ITRUNCATED);
+	ret = -filemap_fdatawrite_range(mapping, first,
+				last == -1 ? LLONG_MAX : last);
 	if (flags & XBF_ASYNC)
 		return ret;
 	ret2 = xfs_wait_on_pages(ip, first, last);

@@ -91,7 +88,9 @@ xfs_wait_on_pages(
 {
 	struct address_space *mapping = VFS_I(ip)->i_mapping;
 
-	if (mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK))
-		return -filemap_fdatawait(mapping);
+	if (mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) {
+		return -filemap_fdatawait_range(mapping, first,
+					last == -1 ? ip->i_size - 1 : last);
+	}
 	return 0;
 }

index 237f5ff..a9c4810 100644
@@ -1292,7 +1292,8 @@ xfs_file_ioctl(
 	case XFS_IOC_ALLOCSP64:
 	case XFS_IOC_FREESP64:
 	case XFS_IOC_RESVSP64:
-	case XFS_IOC_UNRESVSP64: {
+	case XFS_IOC_UNRESVSP64:
+	case XFS_IOC_ZERO_RANGE: {
 		xfs_flock64_t		bf;
 
 		if (copy_from_user(&bf, arg, sizeof(bf)))

index 6c83f7f..464bcc7 100644
@@ -574,6 +574,7 @@ xfs_file_compat_ioctl(
 	case XFS_IOC_FSGEOMETRY_V1:
 	case XFS_IOC_FSGROWFSDATA:
 	case XFS_IOC_FSGROWFSRT:
+	case XFS_IOC_ZERO_RANGE:
 		return xfs_file_ioctl(filp, cmd, p);
 #else
 	case XFS_IOC_ALLOCSP_32:

index 23f14e5..d36770f 100644
@@ -4744,8 +4744,12 @@ xfs_bmapi(
 		 * Check if writing previously allocated but
 		 * unwritten extents.
 		 */
-		if (wr && mval->br_state == XFS_EXT_UNWRITTEN &&
-		    ((flags & (XFS_BMAPI_PREALLOC|XFS_BMAPI_DELAY)) == 0)) {
+		if (wr &&
+		    ((mval->br_state == XFS_EXT_UNWRITTEN &&
+		      ((flags & (XFS_BMAPI_PREALLOC|XFS_BMAPI_DELAY)) == 0)) ||
+		     (mval->br_state == XFS_EXT_NORM &&
+		      ((flags & (XFS_BMAPI_PREALLOC|XFS_BMAPI_CONVERT)) ==
+				(XFS_BMAPI_PREALLOC|XFS_BMAPI_CONVERT))))) {
 			/*
 			 * Modify (by adding) the state flag, if writing.
 			 */

@@ -4757,7 +4761,9 @@ xfs_bmapi(
 					*firstblock;
 				cur->bc_private.b.flist = flist;
 			}
-			mval->br_state = XFS_EXT_NORM;
+			mval->br_state = (mval->br_state == XFS_EXT_UNWRITTEN)
+						? XFS_EXT_NORM
+						: XFS_EXT_UNWRITTEN;
 			error = xfs_bmap_add_extent(ip, lastx, &cur, mval,
 				firstblock, flist, &tmp_logflags,
 				whichfork, (flags & XFS_BMAPI_RSVBLOCKS));

index b13569a..71ec9b6 100644
@@ -74,9 +74,12 @@ typedef	struct xfs_bmap_free
 #define	XFS_BMAPI_IGSTATE	0x080	/* Ignore state - */
 					/* combine contig. space */
 #define	XFS_BMAPI_CONTIG	0x100	/* must allocate only one extent */
-#define XFS_BMAPI_CONVERT	0x200	/* unwritten extent conversion - */
-					/* need write cache flushing and no */
-					/* additional allocation alignments */
+/*
+ * unwritten extent conversion - this needs write cache flushing and no additional
+ * allocation alignments. When specified with XFS_BMAPI_PREALLOC it converts
+ * from written to unwritten, otherwise convert from unwritten to written.
+ */
+#define XFS_BMAPI_CONVERT	0x200
 
 #define XFS_BMAPI_FLAGS \
 	{ XFS_BMAPI_WRITE,	"WRITE" }, \

index 7cf7220..6af6518 100644
@@ -446,6 +446,7 @@ typedef struct xfs_handle {
 /*	XFS_IOC_SETBIOSIZE ---- deprecated 46	   */
 /*	XFS_IOC_GETBIOSIZE ---- deprecated 47	   */
 #define XFS_IOC_GETBMAPX	_IOWR('X', 56, struct getbmap)
+#define XFS_IOC_ZERO_RANGE	_IOW ('X', 57, struct xfs_flock64)
 
 /*
  * ioctl commands that replace IRIX syssgi()'s

index 3ac137d..9e0ebd3 100644
@@ -2270,7 +2270,7 @@ xfs_alloc_file_space(
 	count = len;
 	imapp = &imaps[0];
 	nimaps = 1;
-	bmapi_flag = XFS_BMAPI_WRITE | (alloc_type ? XFS_BMAPI_PREALLOC : 0);
+	bmapi_flag = XFS_BMAPI_WRITE | alloc_type;
 	startoffset_fsb	= XFS_B_TO_FSBT(mp, offset);
 	allocatesize_fsb = XFS_B_TO_FSB(mp, count);
 

@@ -2702,6 +2702,7 @@ xfs_change_file_space(
 	xfs_off_t	llen;
 	xfs_trans_t	*tp;
 	struct iattr	iattr;
+	int		prealloc_type;
 
 	if (!S_ISREG(ip->i_d.di_mode))
 		return XFS_ERROR(EINVAL);

@@ -2744,12 +2745,17 @@ xfs_change_file_space(
 	 * size to be changed.
 	 */
 	setprealloc = clrprealloc = 0;
+	prealloc_type = XFS_BMAPI_PREALLOC;
 
 	switch (cmd) {
+	case XFS_IOC_ZERO_RANGE:
+		prealloc_type |= XFS_BMAPI_CONVERT;
+		xfs_tosspages(ip, startoffset, startoffset + bf->l_len, 0);
+		/* FALLTHRU */
 	case XFS_IOC_RESVSP:
 	case XFS_IOC_RESVSP64:
 		error = xfs_alloc_file_space(ip, startoffset, bf->l_len,
-								1, attr_flags);
+						prealloc_type, attr_flags);
 		if (error)
 			return error;
 		setprealloc = 1;