Merge git://oss.sgi.com:8090/oss/git/xfs-2.6

fs/xfs/Makefile

@@ -1,150 +1 @@
-#
-# Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
-#
-# This program is free software; you can redistribute it and/or modify it
-# under the terms of version 2 of the GNU General Public License as
-# published by the Free Software Foundation.
-#
-# This program is distributed in the hope that it would be useful, but
-# WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
-#
-# Further, this software is distributed without any warranty that it is
-# free of the rightful claim of any third person regarding infringement
-# or the like.  Any license provided herein, whether implied or
-# otherwise, applies only to this software file.  Patent licenses, if
-# any, provided herein do not apply to combinations of this program with
-# other software, or any other product whatsoever.
-#
-# You should have received a copy of the GNU General Public License along
-# with this program; if not, write the Free Software Foundation, Inc., 59
-# Temple Place - Suite 330, Boston MA 02111-1307, USA.
-#
-# Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
-# Mountain View, CA  94043, or:
-#
-# http://www.sgi.com
-#
-# For further information regarding this notice, see:
-#
-# http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
-#
-
-EXTRA_CFLAGS +=	 -Ifs/xfs -Ifs/xfs/linux-2.6 -funsigned-char
-
-ifeq ($(CONFIG_XFS_DEBUG),y)
-	EXTRA_CFLAGS += -g -DSTATIC="" -DDEBUG
-	EXTRA_CFLAGS += -DPAGEBUF_LOCK_TRACKING
-endif
-ifeq ($(CONFIG_XFS_TRACE),y)
-	EXTRA_CFLAGS += -DXFS_ALLOC_TRACE
-	EXTRA_CFLAGS += -DXFS_ATTR_TRACE
-	EXTRA_CFLAGS += -DXFS_BLI_TRACE
-	EXTRA_CFLAGS += -DXFS_BMAP_TRACE
-	EXTRA_CFLAGS += -DXFS_BMBT_TRACE
-	EXTRA_CFLAGS += -DXFS_DIR_TRACE
-	EXTRA_CFLAGS += -DXFS_DIR2_TRACE
-	EXTRA_CFLAGS += -DXFS_DQUOT_TRACE
-	EXTRA_CFLAGS += -DXFS_ILOCK_TRACE
-	EXTRA_CFLAGS += -DXFS_LOG_TRACE
-	EXTRA_CFLAGS += -DXFS_RW_TRACE
-	EXTRA_CFLAGS += -DPAGEBUF_TRACE
-	EXTRA_CFLAGS += -DXFS_VNODE_TRACE
-endif
-
-obj-$(CONFIG_XFS_FS)		+= xfs.o
-
-xfs-$(CONFIG_XFS_QUOTA)		+= $(addprefix quota/, \
-				   xfs_dquot.o \
-				   xfs_dquot_item.o \
-				   xfs_trans_dquot.o \
-				   xfs_qm_syscalls.o \
-				   xfs_qm_bhv.o \
-				   xfs_qm.o)
-ifeq ($(CONFIG_XFS_QUOTA),y)
-xfs-$(CONFIG_PROC_FS)		+= quota/xfs_qm_stats.o
-endif
-
-xfs-$(CONFIG_XFS_RT)		+= xfs_rtalloc.o
-xfs-$(CONFIG_XFS_POSIX_ACL)	+= xfs_acl.o
-xfs-$(CONFIG_PROC_FS)		+= linux-2.6/xfs_stats.o
-xfs-$(CONFIG_SYSCTL)		+= linux-2.6/xfs_sysctl.o
-xfs-$(CONFIG_COMPAT)		+= linux-2.6/xfs_ioctl32.o
-xfs-$(CONFIG_XFS_EXPORT)	+= linux-2.6/xfs_export.o
-
-
-xfs-y				+= xfs_alloc.o \
-				   xfs_alloc_btree.o \
-				   xfs_attr.o \
-				   xfs_attr_leaf.o \
-				   xfs_behavior.o \
-				   xfs_bit.o \
-				   xfs_bmap.o \
-				   xfs_bmap_btree.o \
-				   xfs_btree.o \
-				   xfs_buf_item.o \
-				   xfs_da_btree.o \
-				   xfs_dir.o \
-				   xfs_dir2.o \
-				   xfs_dir2_block.o \
-				   xfs_dir2_data.o \
-				   xfs_dir2_leaf.o \
-				   xfs_dir2_node.o \
-				   xfs_dir2_sf.o \
-				   xfs_dir_leaf.o \
-				   xfs_error.o \
-				   xfs_extfree_item.o \
-				   xfs_fsops.o \
-				   xfs_ialloc.o \
-				   xfs_ialloc_btree.o \
-				   xfs_iget.o \
-				   xfs_inode.o \
-				   xfs_inode_item.o \
-				   xfs_iocore.o \
-				   xfs_iomap.o \
-				   xfs_itable.o \
-				   xfs_dfrag.o \
-				   xfs_log.o \
-				   xfs_log_recover.o \
-				   xfs_macros.o \
-				   xfs_mount.o \
-				   xfs_rename.o \
-				   xfs_trans.o \
-				   xfs_trans_ail.o \
-				   xfs_trans_buf.o \
-				   xfs_trans_extfree.o \
-				   xfs_trans_inode.o \
-				   xfs_trans_item.o \
-				   xfs_utils.o \
-				   xfs_vfsops.o \
-				   xfs_vnodeops.o \
-				   xfs_rw.o \
-				   xfs_dmops.o \
-				   xfs_qmops.o
-
-xfs-$(CONFIG_XFS_TRACE)		+= xfs_dir2_trace.o
-
-# Objects in linux-2.6/
-xfs-y				+= $(addprefix linux-2.6/, \
-				   kmem.o \
-				   xfs_aops.o \
-				   xfs_buf.o \
-				   xfs_file.o \
-				   xfs_fs_subr.o \
-				   xfs_globals.o \
-				   xfs_ioctl.o \
-				   xfs_iops.o \
-				   xfs_lrw.o \
-				   xfs_super.o \
-				   xfs_vfs.o \
-				   xfs_vnode.o)
-
-# Objects in support/
-xfs-y				+= $(addprefix support/, \
-				   debug.o \
-				   move.o \
-				   qsort.o \
-				   uuid.o)
-
-xfs-$(CONFIG_XFS_TRACE)		+= support/ktrace.o
-
+include $(TOPDIR)/fs/xfs/Makefile-linux-$(VERSION).$(PATCHLEVEL)

fs/xfs/Makefile-linux-2.6

@@ -0,0 +1,141 @@
+#
+# Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+#
+# This program is free software; you can redistribute it and/or modify it
+# under the terms of version 2 of the GNU General Public License as
+# published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it would be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+#
+# Further, this software is distributed without any warranty that it is
+# free of the rightful claim of any third person regarding infringement
+# or the like.  Any license provided herein, whether implied or
+# otherwise, applies only to this software file.  Patent licenses, if
+# any, provided herein do not apply to combinations of this program with
+# other software, or any other product whatsoever.
+#
+# You should have received a copy of the GNU General Public License along
+# with this program; if not, write the Free Software Foundation, Inc., 59
+# Temple Place - Suite 330, Boston MA 02111-1307, USA.
+#
+# Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+# Mountain View, CA  94043, or:
+#
+# http://www.sgi.com
+#
+# For further information regarding this notice, see:
+#
+# http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+#
+
+EXTRA_CFLAGS +=	 -Ifs/xfs -Ifs/xfs/linux-2.6 -funsigned-char
+
+XFS_LINUX := linux-2.6
+
+ifeq ($(CONFIG_XFS_DEBUG),y)
+	EXTRA_CFLAGS += -g -DSTATIC="" -DDEBUG
+	EXTRA_CFLAGS += -DPAGEBUF_LOCK_TRACKING
+endif
+ifeq ($(CONFIG_XFS_TRACE),y)
+	EXTRA_CFLAGS += -DXFS_ALLOC_TRACE
+	EXTRA_CFLAGS += -DXFS_ATTR_TRACE
+	EXTRA_CFLAGS += -DXFS_BLI_TRACE
+	EXTRA_CFLAGS += -DXFS_BMAP_TRACE
+	EXTRA_CFLAGS += -DXFS_BMBT_TRACE
+	EXTRA_CFLAGS += -DXFS_DIR_TRACE
+	EXTRA_CFLAGS += -DXFS_DIR2_TRACE
+	EXTRA_CFLAGS += -DXFS_DQUOT_TRACE
+	EXTRA_CFLAGS += -DXFS_ILOCK_TRACE
+	EXTRA_CFLAGS += -DXFS_LOG_TRACE
+	EXTRA_CFLAGS += -DXFS_RW_TRACE
+	EXTRA_CFLAGS += -DPAGEBUF_TRACE
+	EXTRA_CFLAGS += -DXFS_VNODE_TRACE
+endif
+
+obj-$(CONFIG_XFS_FS)		+= xfs.o
+obj-$(CONFIG_XFS_QUOTA)		+= quota/
+
+xfs-$(CONFIG_XFS_RT)		+= xfs_rtalloc.o
+xfs-$(CONFIG_XFS_POSIX_ACL)	+= xfs_acl.o
+xfs-$(CONFIG_PROC_FS)		+= $(XFS_LINUX)/xfs_stats.o
+xfs-$(CONFIG_SYSCTL)		+= $(XFS_LINUX)/xfs_sysctl.o
+xfs-$(CONFIG_COMPAT)		+= $(XFS_LINUX)/xfs_ioctl32.o
+xfs-$(CONFIG_XFS_EXPORT)	+= $(XFS_LINUX)/xfs_export.o
+
+
+xfs-y				+= xfs_alloc.o \
+				   xfs_alloc_btree.o \
+				   xfs_attr.o \
+				   xfs_attr_leaf.o \
+				   xfs_behavior.o \
+				   xfs_bit.o \
+				   xfs_bmap.o \
+				   xfs_bmap_btree.o \
+				   xfs_btree.o \
+				   xfs_buf_item.o \
+				   xfs_da_btree.o \
+				   xfs_dir.o \
+				   xfs_dir2.o \
+				   xfs_dir2_block.o \
+				   xfs_dir2_data.o \
+				   xfs_dir2_leaf.o \
+				   xfs_dir2_node.o \
+				   xfs_dir2_sf.o \
+				   xfs_dir_leaf.o \
+				   xfs_error.o \
+				   xfs_extfree_item.o \
+				   xfs_fsops.o \
+				   xfs_ialloc.o \
+				   xfs_ialloc_btree.o \
+				   xfs_iget.o \
+				   xfs_inode.o \
+				   xfs_inode_item.o \
+				   xfs_iocore.o \
+				   xfs_iomap.o \
+				   xfs_itable.o \
+				   xfs_dfrag.o \
+				   xfs_log.o \
+				   xfs_log_recover.o \
+				   xfs_macros.o \
+				   xfs_mount.o \
+				   xfs_rename.o \
+				   xfs_trans.o \
+				   xfs_trans_ail.o \
+				   xfs_trans_buf.o \
+				   xfs_trans_extfree.o \
+				   xfs_trans_inode.o \
+				   xfs_trans_item.o \
+				   xfs_utils.o \
+				   xfs_vfsops.o \
+				   xfs_vnodeops.o \
+				   xfs_rw.o \
+				   xfs_dmops.o \
+				   xfs_qmops.o
+
+xfs-$(CONFIG_XFS_TRACE)		+= xfs_dir2_trace.o
+
+# Objects in linux/
+xfs-y				+= $(addprefix $(XFS_LINUX)/, \
+				   kmem.o \
+				   xfs_aops.o \
+				   xfs_buf.o \
+				   xfs_file.o \
+				   xfs_fs_subr.o \
+				   xfs_globals.o \
+				   xfs_ioctl.o \
+				   xfs_iops.o \
+				   xfs_lrw.o \
+				   xfs_super.o \
+				   xfs_vfs.o \
+				   xfs_vnode.o)
+
+# Objects in support/
+xfs-y				+= $(addprefix support/, \
+				   debug.o \
+				   move.o \
+				   uuid.o)
+
+xfs-$(CONFIG_XFS_TRACE)		+= support/ktrace.o
+

fs/xfs/linux-2.6/kmem.c

@@ -45,11 +45,11 @@
 
 
 void *
-kmem_alloc(size_t size, int flags)
+kmem_alloc(size_t size, unsigned int __nocast flags)
 {
-	int	retries = 0;
-	int	lflags = kmem_flags_convert(flags);
-	void	*ptr;
+	int		retries = 0;
+	unsigned int	lflags = kmem_flags_convert(flags);
+	void		*ptr;
 
 	do {
 		if (size < MAX_SLAB_SIZE || retries > MAX_VMALLOCS)
@@ -67,7 +67,7 @@ kmem_alloc(size_t size, int flags)
 }
 
 void *
-kmem_zalloc(size_t size, int flags)
+kmem_zalloc(size_t size, unsigned int __nocast flags)
 {
 	void	*ptr;
 
@@ -89,7 +89,8 @@ kmem_free(void *ptr, size_t size)
 }
 
 void *
-kmem_realloc(void *ptr, size_t newsize, size_t oldsize, int flags)
+kmem_realloc(void *ptr, size_t newsize, size_t oldsize,
+	     unsigned int __nocast flags)
 {
 	void	*new;
 
@@ -104,11 +105,11 @@ kmem_realloc(void *ptr, size_t newsize, size_t oldsize, int flags)
 }
 
 void *
-kmem_zone_alloc(kmem_zone_t *zone, int flags)
+kmem_zone_alloc(kmem_zone_t *zone, unsigned int __nocast flags)
 {
-	int	retries = 0;
-	int	lflags = kmem_flags_convert(flags);
-	void	*ptr;
+	int		retries = 0;
+	unsigned int	lflags = kmem_flags_convert(flags);
+	void		*ptr;
 
 	do {
 		ptr = kmem_cache_alloc(zone, lflags);
@@ -123,7 +124,7 @@ kmem_zone_alloc(kmem_zone_t *zone, int flags)
 }
 
 void *
-kmem_zone_zalloc(kmem_zone_t *zone, int flags)
+kmem_zone_zalloc(kmem_zone_t *zone, unsigned int __nocast flags)
 {
 	void	*ptr;
 

fs/xfs/linux-2.6/kmem.h

@@ -39,10 +39,10 @@
 /*
  * memory management routines
  */
-#define KM_SLEEP	0x0001
-#define KM_NOSLEEP	0x0002
-#define KM_NOFS		0x0004
-#define KM_MAYFAIL	0x0008
+#define KM_SLEEP	0x0001u
+#define KM_NOSLEEP	0x0002u
+#define KM_NOFS		0x0004u
+#define KM_MAYFAIL	0x0008u
 
 #define	kmem_zone	kmem_cache_s
 #define kmem_zone_t	kmem_cache_t
@@ -81,9 +81,9 @@ typedef unsigned long xfs_pflags_t;
 	*(NSTATEP) = *(OSTATEP);	\
 } while (0)
 
-static __inline unsigned int kmem_flags_convert(int flags)
+static __inline unsigned int kmem_flags_convert(unsigned int __nocast flags)
 {
-	int	lflags = __GFP_NOWARN;	/* we'll report problems, if need be */
+	unsigned int	lflags = __GFP_NOWARN;	/* we'll report problems, if need be */
 
 #ifdef DEBUG
 	if (unlikely(flags & ~(KM_SLEEP|KM_NOSLEEP|KM_NOFS|KM_MAYFAIL))) {
@@ -125,12 +125,13 @@ kmem_zone_destroy(kmem_zone_t *zone)
 		BUG();
 }
 
-extern void	    *kmem_zone_zalloc(kmem_zone_t *, int);
-extern void	    *kmem_zone_alloc(kmem_zone_t *, int);
+extern void	    *kmem_zone_zalloc(kmem_zone_t *, unsigned int __nocast);
+extern void	    *kmem_zone_alloc(kmem_zone_t *, unsigned int __nocast);
 
-extern void	    *kmem_alloc(size_t, int);
-extern void	    *kmem_realloc(void *, size_t, size_t, int);
-extern void	    *kmem_zalloc(size_t, int);
+extern void	    *kmem_alloc(size_t, unsigned int __nocast);
+extern void	    *kmem_realloc(void *, size_t, size_t,
+				  unsigned int __nocast);
+extern void	    *kmem_zalloc(size_t, unsigned int __nocast);
 extern void         kmem_free(void *, size_t);
 
 typedef struct shrinker *kmem_shaker_t;

fs/xfs/linux-2.6/spin.h

@@ -45,6 +45,9 @@
 typedef spinlock_t lock_t;
 
 #define SPLDECL(s)			unsigned long s
+#ifndef DEFINE_SPINLOCK
+#define DEFINE_SPINLOCK(s)		spinlock_t s = SPIN_LOCK_UNLOCKED
+#endif
 
 #define spinlock_init(lock, name)	spin_lock_init(lock)
 #define	spinlock_destroy(lock)

fs/xfs/linux-2.6/xfs_aops.c

@@ -104,66 +104,114 @@ xfs_page_trace(
 #define xfs_page_trace(tag, inode, page, mask)
 #endif
 
-void
-linvfs_unwritten_done(
-	struct buffer_head	*bh,
-	int			uptodate)
+/*
+ * Schedule IO completion handling on a xfsdatad if this was
+ * the final hold on this ioend.
+ */
+STATIC void
+xfs_finish_ioend(
+	xfs_ioend_t		*ioend)
 {
-	xfs_buf_t		*pb = (xfs_buf_t *)bh->b_private;
+	if (atomic_dec_and_test(&ioend->io_remaining))
+		queue_work(xfsdatad_workqueue, &ioend->io_work);
+}
 
-	ASSERT(buffer_unwritten(bh));
-	bh->b_end_io = NULL;
-	clear_buffer_unwritten(bh);
-	if (!uptodate)
-		pagebuf_ioerror(pb, EIO);
-	if (atomic_dec_and_test(&pb->pb_io_remaining) == 1) {
-		pagebuf_iodone(pb, 1, 1);
-	}
-	end_buffer_async_write(bh, uptodate);
+STATIC void
+xfs_destroy_ioend(
+	xfs_ioend_t		*ioend)
+{
+	vn_iowake(ioend->io_vnode);
+	mempool_free(ioend, xfs_ioend_pool);
 }
 
 /*
  * Issue transactions to convert a buffer range from unwritten
- * to written extents (buffered IO).
+ * to written extents.
  */
 STATIC void
-linvfs_unwritten_convert(
-	xfs_buf_t	*bp)
+xfs_end_bio_unwritten(
+	void			*data)
 {
-	vnode_t		*vp = XFS_BUF_FSPRIVATE(bp, vnode_t *);
-	int		error;
+	xfs_ioend_t		*ioend = data;
+	vnode_t			*vp = ioend->io_vnode;
+	xfs_off_t		offset = ioend->io_offset;
+	size_t			size = ioend->io_size;
+	struct buffer_head	*bh, *next;
+	int			error;
+
+	if (ioend->io_uptodate)
+		VOP_BMAP(vp, offset, size, BMAPI_UNWRITTEN, NULL, NULL, error);
+
+	/* ioend->io_buffer_head is only non-NULL for buffered I/O */
+	for (bh = ioend->io_buffer_head; bh; bh = next) {
+		next = bh->b_private;
+
+		bh->b_end_io = NULL;
+		clear_buffer_unwritten(bh);
+		end_buffer_async_write(bh, ioend->io_uptodate);
+	}
 
-	BUG_ON(atomic_read(&bp->pb_hold) < 1);
-	VOP_BMAP(vp, XFS_BUF_OFFSET(bp), XFS_BUF_SIZE(bp),
-			BMAPI_UNWRITTEN, NULL, NULL, error);
-	XFS_BUF_SET_FSPRIVATE(bp, NULL);
-	XFS_BUF_CLR_IODONE_FUNC(bp);
-	XFS_BUF_UNDATAIO(bp);
-	iput(LINVFS_GET_IP(vp));
-	pagebuf_iodone(bp, 0, 0);
+	xfs_destroy_ioend(ioend);
 }
 
 /*
- * Issue transactions to convert a buffer range from unwritten
- * to written extents (direct IO).
+ * Allocate and initialise an IO completion structure.
+ * We need to track unwritten extent write completion here initially.
+ * We'll need to extend this for updating the ondisk inode size later
+ * (vs. incore size).
  */
-STATIC void
-linvfs_unwritten_convert_direct(
-	struct kiocb	*iocb,
-	loff_t		offset,
-	ssize_t		size,
-	void		*private)
+STATIC xfs_ioend_t *
+xfs_alloc_ioend(
+	struct inode		*inode)
 {
-	struct inode	*inode = iocb->ki_filp->f_dentry->d_inode;
-	ASSERT(!private || inode == (struct inode *)private);
+	xfs_ioend_t		*ioend;
 
-	/* private indicates an unwritten extent lay beneath this IO */
-	if (private && size > 0) {
-		vnode_t	*vp = LINVFS_GET_VP(inode);
-		int	error;
+	ioend = mempool_alloc(xfs_ioend_pool, GFP_NOFS);
 
-		VOP_BMAP(vp, offset, size, BMAPI_UNWRITTEN, NULL, NULL, error);
-	}
+	/*
+	 * Set the count to 1 initially, which will prevent an I/O
+	 * completion callback from happening before we have started
+	 * all the I/O from calling the completion routine too early.
+	 */
+	atomic_set(&ioend->io_remaining, 1);
+	ioend->io_uptodate = 1; /* cleared if any I/O fails */
+	ioend->io_vnode = LINVFS_GET_VP(inode);
+	ioend->io_buffer_head = NULL;
+	atomic_inc(&ioend->io_vnode->v_iocount);
+	ioend->io_offset = 0;
+	ioend->io_size = 0;
+
+	INIT_WORK(&ioend->io_work, xfs_end_bio_unwritten, ioend);
+
+	return ioend;
+}
+
+void
+linvfs_unwritten_done(
+	struct buffer_head	*bh,
+	int			uptodate)
+{
+	xfs_ioend_t		*ioend = bh->b_private;
+	static spinlock_t	unwritten_done_lock = SPIN_LOCK_UNLOCKED;
+	unsigned long		flags;
+
+	ASSERT(buffer_unwritten(bh));
+	bh->b_end_io = NULL;
+
+	if (!uptodate)
+		ioend->io_uptodate = 0;
+
+	/*
+	 * Deep magic here.  We reuse b_private in the buffer_heads to build
+	 * a chain for completing the I/O from user context after we've issued
+	 * a transaction to convert the unwritten extent.
+	 */
+	spin_lock_irqsave(&unwritten_done_lock, flags);
+	bh->b_private = ioend->io_buffer_head;
+	ioend->io_buffer_head = bh;
+	spin_unlock_irqrestore(&unwritten_done_lock, flags);
+
+	xfs_finish_ioend(ioend);
 }
 
 STATIC int
@@ -255,7 +303,7 @@ xfs_probe_unwritten_page(
 	struct address_space	*mapping,
 	pgoff_t			index,
 	xfs_iomap_t		*iomapp,
-	xfs_buf_t		*pb,
+	xfs_ioend_t		*ioend,
 	unsigned long		max_offset,
 	unsigned long		*fsbs,
 	unsigned int            bbits)
@@ -283,7 +331,7 @@ xfs_probe_unwritten_page(
 				break;
 			xfs_map_at_offset(page, bh, p_offset, bbits, iomapp);
 			set_buffer_unwritten_io(bh);
-			bh->b_private = pb;
+			bh->b_private = ioend;
 			p_offset += bh->b_size;
 			(*fsbs)++;
 		} while ((bh = bh->b_this_page) != head);
@@ -434,34 +482,15 @@ xfs_map_unwritten(
 {
 	struct buffer_head	*bh = curr;
 	xfs_iomap_t		*tmp;
-	xfs_buf_t		*pb;
-	loff_t			offset, size;
+	xfs_ioend_t		*ioend;
+	loff_t			offset;
 	unsigned long		nblocks = 0;
 
 	offset = start_page->index;
 	offset <<= PAGE_CACHE_SHIFT;
 	offset += p_offset;
 
-	/* get an "empty" pagebuf to manage IO completion
-	 * Proper values will be set before returning */
-	pb = pagebuf_lookup(iomapp->iomap_target, 0, 0, 0);
-	if (!pb)
-		return -EAGAIN;
-
-	/* Take a reference to the inode to prevent it from
-	 * being reclaimed while we have outstanding unwritten
-	 * extent IO on it.
-	 */
-	if ((igrab(inode)) != inode) {
-		pagebuf_free(pb);
-		return -EAGAIN;
-	}
-
-	/* Set the count to 1 initially, this will stop an I/O
-	 * completion callout which happens before we have started
-	 * all the I/O from calling pagebuf_iodone too early.
-	 */
-	atomic_set(&pb->pb_io_remaining, 1);
+	ioend = xfs_alloc_ioend(inode);
 
 	/* First map forwards in the page consecutive buffers
 	 * covering this unwritten extent
@@ -474,12 +503,12 @@ xfs_map_unwritten(
 			break;
 		xfs_map_at_offset(start_page, bh, p_offset, block_bits, iomapp);
 		set_buffer_unwritten_io(bh);
-		bh->b_private = pb;
+		bh->b_private = ioend;
 		p_offset += bh->b_size;
 		nblocks++;
 	} while ((bh = bh->b_this_page) != head);
 
-	atomic_add(nblocks, &pb->pb_io_remaining);
+	atomic_add(nblocks, &ioend->io_remaining);
 
 	/* If we reached the end of the page, map forwards in any
 	 * following pages which are also covered by this extent.
@@ -496,13 +525,13 @@ xfs_map_unwritten(
 		tloff = min(tlast, tloff);
 		for (tindex = start_page->index + 1; tindex < tloff; tindex++) {
 			page = xfs_probe_unwritten_page(mapping,
-						tindex, iomapp, pb,
+						tindex, iomapp, ioend,
 						PAGE_CACHE_SIZE, &bs, bbits);
 			if (!page)
 				break;
 			nblocks += bs;
-			atomic_add(bs, &pb->pb_io_remaining);
-			xfs_convert_page(inode, page, iomapp, wbc, pb,
+			atomic_add(bs, &ioend->io_remaining);
+			xfs_convert_page(inode, page, iomapp, wbc, ioend,
 							startio, all_bh);
 			/* stop if converting the next page might add
 			 * enough blocks that the corresponding byte
@@ -514,12 +543,12 @@ xfs_map_unwritten(
 		if (tindex == tlast &&
 		    (pg_offset = (i_size_read(inode) & (PAGE_CACHE_SIZE - 1)))) {
 			page = xfs_probe_unwritten_page(mapping,
-							tindex, iomapp, pb,
+							tindex, iomapp, ioend,
 							pg_offset, &bs, bbits);
 			if (page) {
 				nblocks += bs;
-				atomic_add(bs, &pb->pb_io_remaining);
-				xfs_convert_page(inode, page, iomapp, wbc, pb,
+				atomic_add(bs, &ioend->io_remaining);
+				xfs_convert_page(inode, page, iomapp, wbc, ioend,
 							startio, all_bh);
 				if (nblocks >= ((ULONG_MAX - PAGE_SIZE) >> block_bits))
 					goto enough;
@@ -528,21 +557,9 @@ xfs_map_unwritten(
 	}
 
 enough:
-	size = nblocks;		/* NB: using 64bit number here */
-	size <<= block_bits;	/* convert fsb's to byte range */
-
-	XFS_BUF_DATAIO(pb);
-	XFS_BUF_ASYNC(pb);
-	XFS_BUF_SET_SIZE(pb, size);
-	XFS_BUF_SET_COUNT(pb, size);
-	XFS_BUF_SET_OFFSET(pb, offset);
-	XFS_BUF_SET_FSPRIVATE(pb, LINVFS_GET_VP(inode));
-	XFS_BUF_SET_IODONE_FUNC(pb, linvfs_unwritten_convert);
-
-	if (atomic_dec_and_test(&pb->pb_io_remaining) == 1) {
-		pagebuf_iodone(pb, 1, 1);
-	}
-
+	ioend->io_size = (xfs_off_t)nblocks << block_bits;
+	ioend->io_offset = offset;
+	xfs_finish_ioend(ioend);
 	return 0;
 }
 
@@ -787,7 +804,7 @@ xfs_page_state_convert(
 				continue;
 			if (!iomp) {
 				err = xfs_map_blocks(inode, offset, len, &iomap,
-						BMAPI_READ|BMAPI_IGNSTATE);
+						BMAPI_WRITE|BMAPI_IGNSTATE);
 				if (err) {
 					goto error;
 				}
@@ -1028,6 +1045,44 @@ linvfs_get_blocks_direct(
 					create, 1, BMAPI_WRITE|BMAPI_DIRECT);
 }
 
+STATIC void
+linvfs_end_io_direct(
+	struct kiocb	*iocb,
+	loff_t		offset,
+	ssize_t		size,
+	void		*private)
+{
+	xfs_ioend_t	*ioend = iocb->private;
+
+	/*
+	 * Non-NULL private data means we need to issue a transaction to
+	 * convert a range from unwritten to written extents.  This needs
+	 * to happen from process contect but aio+dio I/O completion
+	 * happens from irq context so we need to defer it to a workqueue.
+	 * This is not nessecary for synchronous direct I/O, but we do
+	 * it anyway to keep the code uniform and simpler.
+	 *
+	 * The core direct I/O code might be changed to always call the
+	 * completion handler in the future, in which case all this can
+	 * go away.
+	 */
+	if (private && size > 0) {
+		ioend->io_offset = offset;
+		ioend->io_size = size;
+		xfs_finish_ioend(ioend);
+	} else {
+		ASSERT(size >= 0);
+		xfs_destroy_ioend(ioend);
+	}
+
+	/*
+	 * blockdev_direct_IO can return an error even afer the I/O
+	 * completion handler was called.  Thus we need to protect
+	 * against double-freeing.
+	 */
+	iocb->private = NULL;
+}
+
 STATIC ssize_t
 linvfs_direct_IO(
 	int			rw,
@@ -1042,16 +1097,23 @@ linvfs_direct_IO(
 	xfs_iomap_t	iomap;
 	int		maps = 1;
 	int		error;
+	ssize_t		ret;
 
 	VOP_BMAP(vp, offset, 0, BMAPI_DEVICE, &iomap, &maps, error);
 	if (error)
 		return -error;
 
-	return blockdev_direct_IO_own_locking(rw, iocb, inode,
+	iocb->private = xfs_alloc_ioend(inode);
+
+	ret = blockdev_direct_IO_own_locking(rw, iocb, inode,
 		iomap.iomap_target->pbr_bdev,
 		iov, offset, nr_segs,
 		linvfs_get_blocks_direct,
-		linvfs_unwritten_convert_direct);
+		linvfs_end_io_direct);
+
+	if (unlikely(ret <= 0 && iocb->private))
+		xfs_destroy_ioend(iocb->private);
+	return ret;
 }
 
 
@@ -1202,6 +1264,16 @@ out_unlock:
 	return error;
 }
 
+STATIC int
+linvfs_invalidate_page(
+	struct page		*page,
+	unsigned long		offset)
+{
+	xfs_page_trace(XFS_INVALIDPAGE_ENTER,
+			page->mapping->host, page, offset);
+	return block_invalidatepage(page, offset);
+}
+
 /*
  * Called to move a page into cleanable state - and from there
  * to be released. Possibly the page is already clean. We always
@@ -1279,6 +1351,7 @@ struct address_space_operations linvfs_aops = {
 	.writepage		= linvfs_writepage,
 	.sync_page		= block_sync_page,
 	.releasepage		= linvfs_release_page,
+	.invalidatepage		= linvfs_invalidate_page,
 	.prepare_write		= linvfs_prepare_write,
 	.commit_write		= generic_commit_write,
 	.bmap			= linvfs_bmap,

fs/xfs/linux-2.6/xfs_aops.h

@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) 2005 Silicon Graphics, Inc.  All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * Further, this software is distributed without any warranty that it is
+ * free of the rightful claim of any third person regarding infringement
+ * or the like.  Any license provided herein, whether implied or
+ * otherwise, applies only to this software file.  Patent licenses, if
+ * any, provided herein do not apply to combinations of this program with
+ * other software, or any other product whatsoever.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write the Free Software Foundation, Inc., 59
+ * Temple Place - Suite 330, Boston MA 02111-1307, USA.
+ *
+ * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+ * Mountain View, CA  94043, or:
+ *
+ * http://www.sgi.com
+ *
+ * For further information regarding this notice, see:
+ *
+ * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ */
+#ifndef __XFS_AOPS_H__
+#define __XFS_AOPS_H__
+
+extern struct workqueue_struct *xfsdatad_workqueue;
+extern mempool_t *xfs_ioend_pool;
+
+typedef void (*xfs_ioend_func_t)(void *);
+
+typedef struct xfs_ioend {
+	unsigned int		io_uptodate;	/* I/O status register */
+	atomic_t		io_remaining;	/* hold count */
+	struct vnode		*io_vnode;	/* file being written to */
+	struct buffer_head	*io_buffer_head;/* buffer linked list head */
+	size_t			io_size;	/* size of the extent */
+	xfs_off_t		io_offset;	/* offset in the file */
+	struct work_struct	io_work;	/* xfsdatad work queue */
+} xfs_ioend_t;
+
+#endif /* __XFS_IOPS_H__ */

fs/xfs/linux-2.6/xfs_buf.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -54,6 +54,7 @@
 #include <linux/percpu.h>
 #include <linux/blkdev.h>
 #include <linux/hash.h>
+#include <linux/kthread.h>
 
 #include "xfs_linux.h"
 
@@ -67,7 +68,7 @@ STATIC int xfsbufd_wakeup(int, unsigned int);
 STATIC void pagebuf_delwri_queue(xfs_buf_t *, int);
 
 STATIC struct workqueue_struct *xfslogd_workqueue;
-STATIC struct workqueue_struct *xfsdatad_workqueue;
+struct workqueue_struct *xfsdatad_workqueue;
 
 /*
  * Pagebuf debugging
@@ -590,8 +591,10 @@ found:
 		PB_SET_OWNER(pb);
 	}
 
-	if (pb->pb_flags & PBF_STALE)
+	if (pb->pb_flags & PBF_STALE) {
+		ASSERT((pb->pb_flags & _PBF_DELWRI_Q) == 0);
 		pb->pb_flags &= PBF_MAPPED;
+	}
 	PB_TRACE(pb, "got_lock", 0);
 	XFS_STATS_INC(pb_get_locked);
 	return (pb);
@@ -699,25 +702,6 @@ xfs_buf_read_flags(
 	return NULL;
 }
 
-/*
- * Create a skeletal pagebuf (no pages associated with it).
- */
-xfs_buf_t *
-pagebuf_lookup(
-	xfs_buftarg_t		*target,
-	loff_t			ioff,
-	size_t			isize,
-	page_buf_flags_t	flags)
-{
-	xfs_buf_t		*pb;
-
-	pb = pagebuf_allocate(flags);
-	if (pb) {
-		_pagebuf_initialize(pb, target, ioff, isize, flags);
-	}
-	return pb;
-}
-
 /*
  * If we are not low on memory then do the readahead in a deadlock
  * safe manner.
@@ -913,22 +897,23 @@ pagebuf_rele(
 			do_free = 0;
 		}
 
-		if (pb->pb_flags & PBF_DELWRI) {
-			pb->pb_flags |= PBF_ASYNC;
-			atomic_inc(&pb->pb_hold);
-			pagebuf_delwri_queue(pb, 0);
-			do_free = 0;
-		} else if (pb->pb_flags & PBF_FS_MANAGED) {
+		if (pb->pb_flags & PBF_FS_MANAGED) {
 			do_free = 0;
 		}
 
 		if (do_free) {
+			ASSERT((pb->pb_flags & (PBF_DELWRI|_PBF_DELWRI_Q)) == 0);
 			list_del_init(&pb->pb_hash_list);
 			spin_unlock(&hash->bh_lock);
 			pagebuf_free(pb);
 		} else {
 			spin_unlock(&hash->bh_lock);
 		}
+	} else {
+		/*
+		 * Catch reference count leaks
+		 */
+		ASSERT(atomic_read(&pb->pb_hold) >= 0);
 	}
 }
 
@@ -1006,13 +991,24 @@ pagebuf_lock(
  *	pagebuf_unlock
  *
  *	pagebuf_unlock releases the lock on the buffer object created by
- *	pagebuf_lock or pagebuf_cond_lock (not any
- *	pinning of underlying pages created by pagebuf_pin).
+ *	pagebuf_lock or pagebuf_cond_lock (not any pinning of underlying pages
+ *	created by pagebuf_pin).
+ *
+ *	If the buffer is marked delwri but is not queued, do so before we
+ *	unlock the buffer as we need to set flags correctly. We also need to
+ *	take a reference for the delwri queue because the unlocker is going to
+ *	drop their's and they don't know we just queued it.
  */
 void
 pagebuf_unlock(				/* unlock buffer		*/
 	xfs_buf_t		*pb)	/* buffer to unlock		*/
 {
+	if ((pb->pb_flags & (PBF_DELWRI|_PBF_DELWRI_Q)) == PBF_DELWRI) {
+		atomic_inc(&pb->pb_hold);
+		pb->pb_flags |= PBF_ASYNC;
+		pagebuf_delwri_queue(pb, 0);
+	}
+
 	PB_CLEAR_OWNER(pb);
 	up(&pb->pb_sema);
 	PB_TRACE(pb, "unlock", 0);
@@ -1249,8 +1245,8 @@ bio_end_io_pagebuf(
 	int			error)
 {
 	xfs_buf_t		*pb = (xfs_buf_t *)bio->bi_private;
-	unsigned int		i, blocksize = pb->pb_target->pbr_bsize;
-	struct bio_vec		*bvec = bio->bi_io_vec;
+	unsigned int		blocksize = pb->pb_target->pbr_bsize;
+	struct bio_vec		*bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
 
 	if (bio->bi_size)
 		return 1;
@@ -1258,10 +1254,12 @@ bio_end_io_pagebuf(
 	if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
 		pb->pb_error = EIO;
 
-	for (i = 0; i < bio->bi_vcnt; i++, bvec++) {
+	do {
 		struct page	*page = bvec->bv_page;
 
-		if (pb->pb_error) {
+		if (unlikely(pb->pb_error)) {
+			if (pb->pb_flags & PBF_READ)
+				ClearPageUptodate(page);
 			SetPageError(page);
 		} else if (blocksize == PAGE_CACHE_SIZE) {
 			SetPageUptodate(page);
@@ -1270,10 +1268,13 @@ bio_end_io_pagebuf(
 			set_page_region(page, bvec->bv_offset, bvec->bv_len);
 		}
 
+		if (--bvec >= bio->bi_io_vec)
+			prefetchw(&bvec->bv_page->flags);
+
 		if (_pagebuf_iolocked(pb)) {
 			unlock_page(page);
 		}
-	}
+	} while (bvec >= bio->bi_io_vec);
 
 	_pagebuf_iodone(pb, 1);
 	bio_put(bio);
@@ -1511,6 +1512,11 @@ again:
 			ASSERT(btp == bp->pb_target);
 			if (!(bp->pb_flags & PBF_FS_MANAGED)) {
 				spin_unlock(&hash->bh_lock);
+				/*
+				 * Catch superblock reference count leaks
+				 * immediately
+				 */
+				BUG_ON(bp->pb_bn == 0);
 				delay(100);
 				goto again;
 			}
@@ -1686,17 +1692,20 @@ pagebuf_delwri_queue(
 	int			unlock)
 {
 	PB_TRACE(pb, "delwri_q", (long)unlock);
-	ASSERT(pb->pb_flags & PBF_DELWRI);
+	ASSERT((pb->pb_flags & (PBF_DELWRI|PBF_ASYNC)) ==
+					(PBF_DELWRI|PBF_ASYNC));
 
 	spin_lock(&pbd_delwrite_lock);
 	/* If already in the queue, dequeue and place at tail */
 	if (!list_empty(&pb->pb_list)) {
+		ASSERT(pb->pb_flags & _PBF_DELWRI_Q);
 		if (unlock) {
 			atomic_dec(&pb->pb_hold);
 		}
 		list_del(&pb->pb_list);
 	}
 
+	pb->pb_flags |= _PBF_DELWRI_Q;
 	list_add_tail(&pb->pb_list, &pbd_delwrite_queue);
 	pb->pb_queuetime = jiffies;
 	spin_unlock(&pbd_delwrite_lock);
@@ -1713,10 +1722,11 @@ pagebuf_delwri_dequeue(
 
 	spin_lock(&pbd_delwrite_lock);
 	if ((pb->pb_flags & PBF_DELWRI) && !list_empty(&pb->pb_list)) {
+		ASSERT(pb->pb_flags & _PBF_DELWRI_Q);
 		list_del_init(&pb->pb_list);
 		dequeued = 1;
 	}
-	pb->pb_flags &= ~PBF_DELWRI;
+	pb->pb_flags &= ~(PBF_DELWRI|_PBF_DELWRI_Q);
 	spin_unlock(&pbd_delwrite_lock);
 
 	if (dequeued)
@@ -1733,9 +1743,7 @@ pagebuf_runall_queues(
 }
 
 /* Defines for pagebuf daemon */
-STATIC DECLARE_COMPLETION(xfsbufd_done);
 STATIC struct task_struct *xfsbufd_task;
-STATIC int xfsbufd_active;
 STATIC int xfsbufd_force_flush;
 STATIC int xfsbufd_force_sleep;
 
@@ -1761,14 +1769,8 @@ xfsbufd(
 	xfs_buftarg_t		*target;
 	xfs_buf_t		*pb, *n;
 
-	/*  Set up the thread  */
-	daemonize("xfsbufd");
 	current->flags |= PF_MEMALLOC;
 
-	xfsbufd_task = current;
-	xfsbufd_active = 1;
-	barrier();
-
 	INIT_LIST_HEAD(&tmp);
 	do {
 		if (unlikely(freezing(current))) {
@@ -1795,7 +1797,7 @@ xfsbufd(
 					break;
 				}
 
-				pb->pb_flags &= ~PBF_DELWRI;
+				pb->pb_flags &= ~(PBF_DELWRI|_PBF_DELWRI_Q);
 				pb->pb_flags |= PBF_WRITE;
 				list_move(&pb->pb_list, &tmp);
 			}
@@ -1816,9 +1818,9 @@ xfsbufd(
 			purge_addresses();
 
 		xfsbufd_force_flush = 0;
-	} while (xfsbufd_active);
+	} while (!kthread_should_stop());
 
-	complete_and_exit(&xfsbufd_done, 0);
+	return 0;
 }
 
 /*
@@ -1845,15 +1847,13 @@ xfs_flush_buftarg(
 		if (pb->pb_target != target)
 			continue;
 
-		ASSERT(pb->pb_flags & PBF_DELWRI);
+		ASSERT(pb->pb_flags & (PBF_DELWRI|_PBF_DELWRI_Q));
 		PB_TRACE(pb, "walkq2", (long)pagebuf_ispin(pb));
 		if (pagebuf_ispin(pb)) {
 			pincount++;
 			continue;
 		}
 
-		pb->pb_flags &= ~PBF_DELWRI;
-		pb->pb_flags |= PBF_WRITE;
 		list_move(&pb->pb_list, &tmp);
 	}
 	spin_unlock(&pbd_delwrite_lock);
@@ -1862,12 +1862,14 @@ xfs_flush_buftarg(
 	 * Dropped the delayed write list lock, now walk the temporary list
 	 */
 	list_for_each_entry_safe(pb, n, &tmp, pb_list) {
+		pagebuf_lock(pb);
+		pb->pb_flags &= ~(PBF_DELWRI|_PBF_DELWRI_Q);
+		pb->pb_flags |= PBF_WRITE;
 		if (wait)
 			pb->pb_flags &= ~PBF_ASYNC;
 		else
 			list_del_init(&pb->pb_list);
 
-		pagebuf_lock(pb);
 		pagebuf_iostrategy(pb);
 	}
 
@@ -1901,9 +1903,11 @@ xfs_buf_daemons_start(void)
 	if (!xfsdatad_workqueue)
 		goto out_destroy_xfslogd_workqueue;
 
-	error = kernel_thread(xfsbufd, NULL, CLONE_FS|CLONE_FILES);
-	if (error < 0)
+	xfsbufd_task = kthread_run(xfsbufd, NULL, "xfsbufd");
+	if (IS_ERR(xfsbufd_task)) {
+		error = PTR_ERR(xfsbufd_task);
 		goto out_destroy_xfsdatad_workqueue;
+	}
 	return 0;
 
  out_destroy_xfsdatad_workqueue:
@@ -1920,10 +1924,7 @@ xfs_buf_daemons_start(void)
 STATIC void
 xfs_buf_daemons_stop(void)
 {
-	xfsbufd_active = 0;
-	barrier();
-	wait_for_completion(&xfsbufd_done);
-
+	kthread_stop(xfsbufd_task);
 	destroy_workqueue(xfslogd_workqueue);
 	destroy_workqueue(xfsdatad_workqueue);
 }

fs/xfs/linux-2.6/xfs_buf.h

@@ -89,6 +89,7 @@ typedef enum page_buf_flags_e {		/* pb_flags values */
 	_PBF_PAGE_CACHE = (1 << 17),/* backed by pagecache		   */
 	_PBF_KMEM_ALLOC = (1 << 18),/* backed by kmem_alloc()		   */
 	_PBF_RUN_QUEUES = (1 << 19),/* run block device task queue	   */
+	_PBF_DELWRI_Q = (1 << 21),   /* buffer on delwri queue		   */
 } page_buf_flags_t;
 
 #define PBF_UPDATE (PBF_READ | PBF_WRITE)
@@ -206,13 +207,6 @@ extern xfs_buf_t *xfs_buf_read_flags(	/* allocate and read a buffer	*/
 #define xfs_buf_read(target, blkno, len, flags) \
 	xfs_buf_read_flags((target), (blkno), (len), PBF_LOCK | PBF_MAPPED)
 
-extern xfs_buf_t *pagebuf_lookup(
-		xfs_buftarg_t *,
-		loff_t,			/* starting offset of range	*/
-		size_t,			/* length of range		*/
-		page_buf_flags_t);	/* PBF_READ, PBF_WRITE,		*/
-					/* PBF_FORCEIO, 		*/
-
 extern xfs_buf_t *pagebuf_get_empty(	/* allocate pagebuf struct with	*/
 					/*  no memory or disk address	*/
 		size_t len,
@@ -344,8 +338,6 @@ extern void pagebuf_trace(
 
 
 
-
-
 /* These are just for xfs_syncsub... it sets an internal variable
  * then passes it to VOP_FLUSH_PAGES or adds the flags to a newly gotten buf_t
  */
@@ -452,7 +444,7 @@ extern void pagebuf_trace(
 
 #define XFS_BUF_PTR(bp)		(xfs_caddr_t)((bp)->pb_addr)
 
-extern inline xfs_caddr_t xfs_buf_offset(xfs_buf_t *bp, size_t offset)
+static inline xfs_caddr_t xfs_buf_offset(xfs_buf_t *bp, size_t offset)
 {
 	if (bp->pb_flags & PBF_MAPPED)
 		return XFS_BUF_PTR(bp) + offset;

fs/xfs/linux-2.6/xfs_file.c

@@ -311,6 +311,31 @@ linvfs_fsync(
 
 #define nextdp(dp)      ((struct xfs_dirent *)((char *)(dp) + (dp)->d_reclen))
 
+#ifdef CONFIG_XFS_DMAPI
+
+STATIC struct page *
+linvfs_filemap_nopage(
+	struct vm_area_struct	*area,
+	unsigned long		address,
+	int			*type)
+{
+	struct inode	*inode = area->vm_file->f_dentry->d_inode;
+	vnode_t		*vp = LINVFS_GET_VP(inode);
+	xfs_mount_t	*mp = XFS_VFSTOM(vp->v_vfsp);
+	int		error;
+
+	ASSERT_ALWAYS(vp->v_vfsp->vfs_flag & VFS_DMI);
+
+	error = XFS_SEND_MMAP(mp, area, 0);
+	if (error)
+		return NULL;
+
+	return filemap_nopage(area, address, type);
+}
+
+#endif /* CONFIG_XFS_DMAPI */
+
+
 STATIC int
 linvfs_readdir(
 	struct file	*filp,
@@ -390,14 +415,6 @@ done:
 	return -error;
 }
 
-#ifdef CONFIG_XFS_DMAPI
-STATIC void
-linvfs_mmap_close(
-	struct vm_area_struct	*vma)
-{
-	xfs_dm_mm_put(vma);
-}
-#endif /* CONFIG_XFS_DMAPI */
 
 STATIC int
 linvfs_file_mmap(
@@ -411,16 +428,11 @@ linvfs_file_mmap(
 
 	vma->vm_ops = &linvfs_file_vm_ops;
 
-	if (vp->v_vfsp->vfs_flag & VFS_DMI) {
-		xfs_mount_t	*mp = XFS_VFSTOM(vp->v_vfsp);
-
-		error = -XFS_SEND_MMAP(mp, vma, 0);
-		if (error)
-			return error;
 #ifdef CONFIG_XFS_DMAPI
+	if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 		vma->vm_ops = &linvfs_dmapi_file_vm_ops;
-#endif
 	}
+#endif /* CONFIG_XFS_DMAPI */
 
 	VOP_SETATTR(vp, &va, XFS_AT_UPDATIME, NULL, error);
 	if (!error)
@@ -474,6 +486,7 @@ linvfs_ioctl_invis(
 	return error;
 }
 
+#ifdef CONFIG_XFS_DMAPI
 #ifdef HAVE_VMOP_MPROTECT
 STATIC int
 linvfs_mprotect(
@@ -494,6 +507,7 @@ linvfs_mprotect(
 	return error;
 }
 #endif /* HAVE_VMOP_MPROTECT */
+#endif /* CONFIG_XFS_DMAPI */
 
 #ifdef HAVE_FOP_OPEN_EXEC
 /* If the user is attempting to execute a file that is offline then
@@ -528,49 +542,10 @@ open_exec_out:
 }
 #endif /* HAVE_FOP_OPEN_EXEC */
 
-/*
- * Temporary workaround to the AIO direct IO write problem.
- * This code can go and we can revert to do_sync_write once
- * the writepage(s) rework is merged.
- */
-STATIC ssize_t
-linvfs_write(
-	struct file	*filp,
-	const char	__user *buf,
-	size_t		len,
-	loff_t		*ppos)
-{
-	struct kiocb	kiocb;
-	ssize_t		ret;
-
-	init_sync_kiocb(&kiocb, filp);
-	kiocb.ki_pos = *ppos;
-	ret = __linvfs_write(&kiocb, buf, 0, len, kiocb.ki_pos);
-	*ppos = kiocb.ki_pos;
-	return ret;
-}
-STATIC ssize_t
-linvfs_write_invis(
-	struct file	*filp,
-	const char	__user *buf,
-	size_t		len,
-	loff_t		*ppos)
-{
-	struct kiocb	kiocb;
-	ssize_t		ret;
-
-	init_sync_kiocb(&kiocb, filp);
-	kiocb.ki_pos = *ppos;
-	ret = __linvfs_write(&kiocb, buf, IO_INVIS, len, kiocb.ki_pos);
-	*ppos = kiocb.ki_pos;
-	return ret;
-}
-
-
 struct file_operations linvfs_file_operations = {
 	.llseek		= generic_file_llseek,
 	.read		= do_sync_read,
-	.write		= linvfs_write,
+	.write		= do_sync_write,
 	.readv		= linvfs_readv,
 	.writev		= linvfs_writev,
 	.aio_read	= linvfs_aio_read,
@@ -592,7 +567,7 @@ struct file_operations linvfs_file_operations = {
 struct file_operations linvfs_invis_file_operations = {
 	.llseek		= generic_file_llseek,
 	.read		= do_sync_read,
-	.write		= linvfs_write_invis,
+	.write		= do_sync_write,
 	.readv		= linvfs_readv_invis,
 	.writev		= linvfs_writev_invis,
 	.aio_read	= linvfs_aio_read_invis,
@@ -626,8 +601,7 @@ static struct vm_operations_struct linvfs_file_vm_ops = {
 
 #ifdef CONFIG_XFS_DMAPI
 static struct vm_operations_struct linvfs_dmapi_file_vm_ops = {
-	.close		= linvfs_mmap_close,
-	.nopage		= filemap_nopage,
+	.nopage		= linvfs_filemap_nopage,
 	.populate	= filemap_populate,
 #ifdef HAVE_VMOP_MPROTECT
 	.mprotect	= linvfs_mprotect,

fs/xfs/linux-2.6/xfs_ioctl.c

@@ -141,13 +141,19 @@ xfs_find_handle(
 		return -XFS_ERROR(EINVAL);
 	}
 
-	/* we need the vnode */
-	vp = LINVFS_GET_VP(inode);
-	if (vp->v_type != VREG && vp->v_type != VDIR && vp->v_type != VLNK) {
+	switch (inode->i_mode & S_IFMT) {
+	case S_IFREG:
+	case S_IFDIR:
+	case S_IFLNK:
+		break;
+	default:
 		iput(inode);
 		return -XFS_ERROR(EBADF);
 	}
 
+	/* we need the vnode */
+	vp = LINVFS_GET_VP(inode);
+
 	/* now we can grab the fsid */
 	memcpy(&handle.ha_fsid, vp->v_vfsp->vfs_altfsid, sizeof(xfs_fsid_t));
 	hsize = sizeof(xfs_fsid_t);
@@ -386,7 +392,7 @@ xfs_readlink_by_handle(
 		return -error;
 
 	/* Restrict this handle operation to symlinks only. */
-	if (vp->v_type != VLNK) {
+	if (!S_ISLNK(inode->i_mode)) {
 		VN_RELE(vp);
 		return -XFS_ERROR(EINVAL);
 	}
@@ -982,10 +988,10 @@ xfs_ioc_space(
 	if (vp->v_inode.i_flags & (S_IMMUTABLE|S_APPEND))
 		return -XFS_ERROR(EPERM);
 
-	if (!(filp->f_flags & FMODE_WRITE))
+	if (!(filp->f_mode & FMODE_WRITE))
 		return -XFS_ERROR(EBADF);
 
-	if (vp->v_type != VREG)
+	if (!VN_ISREG(vp))
 		return -XFS_ERROR(EINVAL);
 
 	if (copy_from_user(&bf, arg, sizeof(bf)))

fs/xfs/linux-2.6/xfs_ioctl32.c

@@ -47,8 +47,52 @@
 #include "xfs_vnode.h"
 #include "xfs_dfrag.h"
 
+#define  _NATIVE_IOC(cmd, type) \
+	  _IOC(_IOC_DIR(cmd), _IOC_TYPE(cmd), _IOC_NR(cmd), sizeof(type))
+
 #if defined(CONFIG_IA64) || defined(CONFIG_X86_64)
 #define BROKEN_X86_ALIGNMENT
+/* on ia32 l_start is on a 32-bit boundary */
+typedef struct xfs_flock64_32 {
+	__s16		l_type;
+	__s16		l_whence;
+	__s64		l_start	__attribute__((packed));
+			/* len == 0 means until end of file */
+	__s64		l_len __attribute__((packed));
+	__s32		l_sysid;
+	__u32		l_pid;
+	__s32		l_pad[4];	/* reserve area */
+} xfs_flock64_32_t;
+
+#define XFS_IOC_ALLOCSP_32	_IOW ('X', 10, struct xfs_flock64_32)
+#define XFS_IOC_FREESP_32	_IOW ('X', 11, struct xfs_flock64_32)
+#define XFS_IOC_ALLOCSP64_32	_IOW ('X', 36, struct xfs_flock64_32)
+#define XFS_IOC_FREESP64_32	_IOW ('X', 37, struct xfs_flock64_32)
+#define XFS_IOC_RESVSP_32	_IOW ('X', 40, struct xfs_flock64_32)
+#define XFS_IOC_UNRESVSP_32	_IOW ('X', 41, struct xfs_flock64_32)
+#define XFS_IOC_RESVSP64_32	_IOW ('X', 42, struct xfs_flock64_32)
+#define XFS_IOC_UNRESVSP64_32	_IOW ('X', 43, struct xfs_flock64_32)
+
+/* just account for different alignment */
+STATIC unsigned long
+xfs_ioctl32_flock(
+	unsigned long		arg)
+{
+	xfs_flock64_32_t	__user *p32 = (void __user *)arg;
+	xfs_flock64_t		__user *p = compat_alloc_user_space(sizeof(*p));
+
+	if (copy_in_user(&p->l_type,	&p32->l_type,	sizeof(s16)) ||
+	    copy_in_user(&p->l_whence,	&p32->l_whence, sizeof(s16)) ||
+	    copy_in_user(&p->l_start,	&p32->l_start,	sizeof(s64)) ||
+	    copy_in_user(&p->l_len,	&p32->l_len,	sizeof(s64)) ||
+	    copy_in_user(&p->l_sysid,	&p32->l_sysid,	sizeof(s32)) ||
+	    copy_in_user(&p->l_pid,	&p32->l_pid,	sizeof(u32)) ||
+	    copy_in_user(&p->l_pad,	&p32->l_pad,	4*sizeof(u32)))
+		return -EFAULT;
+	
+	return (unsigned long)p;
+}
+
 #else
 
 typedef struct xfs_fsop_bulkreq32 {
@@ -103,7 +147,6 @@ __linvfs_compat_ioctl(int mode, struct file *f, unsigned cmd, unsigned long arg)
 /* not handled
 	case XFS_IOC_FD_TO_HANDLE:
 	case XFS_IOC_PATH_TO_HANDLE:
-	case XFS_IOC_PATH_TO_HANDLE:
 	case XFS_IOC_PATH_TO_FSHANDLE:
 	case XFS_IOC_OPEN_BY_HANDLE:
 	case XFS_IOC_FSSETDM_BY_HANDLE:
@@ -124,8 +167,21 @@ __linvfs_compat_ioctl(int mode, struct file *f, unsigned cmd, unsigned long arg)
 	case XFS_IOC_ERROR_CLEARALL:
 		break;
 
-#ifndef BROKEN_X86_ALIGNMENT
-	/* xfs_flock_t and xfs_bstat_t have wrong u32 vs u64 alignment */
+#ifdef BROKEN_X86_ALIGNMENT
+	/* xfs_flock_t has wrong u32 vs u64 alignment */
+	case XFS_IOC_ALLOCSP_32:
+	case XFS_IOC_FREESP_32:
+	case XFS_IOC_ALLOCSP64_32:
+	case XFS_IOC_FREESP64_32:
+	case XFS_IOC_RESVSP_32:
+	case XFS_IOC_UNRESVSP_32:
+	case XFS_IOC_RESVSP64_32:
+	case XFS_IOC_UNRESVSP64_32:
+		arg = xfs_ioctl32_flock(arg);
+		cmd = _NATIVE_IOC(cmd, struct xfs_flock64);
+		break;
+
+#else /* These are handled fine if no alignment issues */
 	case XFS_IOC_ALLOCSP:
 	case XFS_IOC_FREESP:
 	case XFS_IOC_RESVSP:
@@ -134,6 +190,9 @@ __linvfs_compat_ioctl(int mode, struct file *f, unsigned cmd, unsigned long arg)
 	case XFS_IOC_FREESP64:
 	case XFS_IOC_RESVSP64:
 	case XFS_IOC_UNRESVSP64:
+		break;
+
+	/* xfs_bstat_t still has wrong u32 vs u64 alignment */
 	case XFS_IOC_SWAPEXT:
 		break;
 

fs/xfs/linux-2.6/xfs_iops.c

@@ -140,7 +140,6 @@ linvfs_mknod(
 
 	memset(&va, 0, sizeof(va));
 	va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
-	va.va_type = IFTOVT(mode);
 	va.va_mode = mode;
 
 	switch (mode & S_IFMT) {
@@ -308,14 +307,13 @@ linvfs_symlink(
 	cvp = NULL;
 
 	memset(&va, 0, sizeof(va));
-	va.va_type = VLNK;
-	va.va_mode = irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO;
+	va.va_mode = S_IFLNK |
+		(irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO);
 	va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
 
 	error = 0;
 	VOP_SYMLINK(dvp, dentry, &va, (char *)symname, &cvp, NULL, error);
 	if (!error && cvp) {
-		ASSERT(cvp->v_type == VLNK);
 		ip = LINVFS_GET_IP(cvp);
 		d_instantiate(dentry, ip);
 		validate_fields(dir);
@@ -425,9 +423,14 @@ linvfs_follow_link(
 	return NULL;
 }
 
-static void linvfs_put_link(struct dentry *dentry, struct nameidata *nd, void *p)
+STATIC void
+linvfs_put_link(
+	struct dentry	*dentry,
+	struct nameidata *nd,
+	void		*p)
 {
-	char *s = nd_get_link(nd);
+	char		*s = nd_get_link(nd);
+
 	if (!IS_ERR(s))
 		kfree(s);
 }

fs/xfs/linux-2.6/xfs_linux.h

@@ -64,7 +64,6 @@
 #include <sema.h>
 #include <time.h>
 
-#include <support/qsort.h>
 #include <support/ktrace.h>
 #include <support/debug.h>
 #include <support/move.h>
@@ -104,6 +103,7 @@
 #include <xfs_stats.h>
 #include <xfs_sysctl.h>
 #include <xfs_iops.h>
+#include <xfs_aops.h>
 #include <xfs_super.h>
 #include <xfs_globals.h>
 #include <xfs_fs_subr.h>
@@ -254,11 +254,18 @@ static inline void set_buffer_unwritten_io(struct buffer_head *bh)
 #define MAX(a,b)	(max(a,b))
 #define howmany(x, y)	(((x)+((y)-1))/(y))
 #define roundup(x, y)	((((x)+((y)-1))/(y))*(y))
+#define qsort(a,n,s,fn)	sort(a,n,s,fn,NULL)
 
+/*
+ * Various platform dependent calls that don't fit anywhere else
+ */
 #define xfs_stack_trace()	dump_stack()
-
 #define xfs_itruncate_data(ip, off)	\
 	(-vmtruncate(LINVFS_GET_IP(XFS_ITOV(ip)), (off)))
+#define xfs_statvfs_fsid(statp, mp)	\
+	({ u64 id = huge_encode_dev((mp)->m_dev);	\
+	   __kernel_fsid_t *fsid = &(statp)->f_fsid;	\
+	(fsid->val[0] = (u32)id, fsid->val[1] = (u32)(id >> 32)); })
 
 
 /* Move the kernel do_div definition off to one side */
@@ -371,6 +378,4 @@ static inline __uint64_t roundup_64(__uint64_t x, __uint32_t y)
 	return(x * y);
 }
 
-#define qsort(a, n, s, cmp) sort(a, n, s, cmp, NULL)
-
 #endif /* __XFS_LINUX__ */

fs/xfs/linux-2.6/xfs_lrw.c

@@ -660,9 +660,6 @@ xfs_write(
 			(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
 				mp->m_rtdev_targp : mp->m_ddev_targp;
 
-		if (ioflags & IO_ISAIO)
-			return XFS_ERROR(-ENOSYS);
-
 		if ((pos & target->pbr_smask) || (count & target->pbr_smask))
 			return XFS_ERROR(-EINVAL);
 

fs/xfs/linux-2.6/xfs_lrw.h

@@ -70,9 +70,10 @@ struct xfs_iomap;
 #define	XFS_SENDFILE_ENTER	21
 #define	XFS_WRITEPAGE_ENTER	22
 #define	XFS_RELEASEPAGE_ENTER	23
-#define	XFS_IOMAP_ALLOC_ENTER	24
-#define	XFS_IOMAP_ALLOC_MAP	25
-#define	XFS_IOMAP_UNWRITTEN	26
+#define	XFS_INVALIDPAGE_ENTER	24
+#define	XFS_IOMAP_ALLOC_ENTER	25
+#define	XFS_IOMAP_ALLOC_MAP	26
+#define	XFS_IOMAP_UNWRITTEN	27
 extern void xfs_rw_enter_trace(int, struct xfs_iocore *,
 				void *, size_t, loff_t, int);
 extern void xfs_inval_cached_trace(struct xfs_iocore *,

fs/xfs/linux-2.6/xfs_super.c

@@ -70,11 +70,15 @@
 #include <linux/namei.h>
 #include <linux/init.h>
 #include <linux/mount.h>
+#include <linux/mempool.h>
 #include <linux/writeback.h>
+#include <linux/kthread.h>
 
 STATIC struct quotactl_ops linvfs_qops;
 STATIC struct super_operations linvfs_sops;
-STATIC kmem_zone_t *linvfs_inode_zone;
+STATIC kmem_zone_t *xfs_vnode_zone;
+STATIC kmem_zone_t *xfs_ioend_zone;
+mempool_t *xfs_ioend_pool;
 
 STATIC struct xfs_mount_args *
 xfs_args_allocate(
@@ -138,24 +142,25 @@ STATIC __inline__ void
 xfs_set_inodeops(
 	struct inode		*inode)
 {
-	vnode_t			*vp = LINVFS_GET_VP(inode);
-
-	if (vp->v_type == VNON) {
-		vn_mark_bad(vp);
-	} else if (S_ISREG(inode->i_mode)) {
+	switch (inode->i_mode & S_IFMT) {
+	case S_IFREG:
 		inode->i_op = &linvfs_file_inode_operations;
 		inode->i_fop = &linvfs_file_operations;
 		inode->i_mapping->a_ops = &linvfs_aops;
-	} else if (S_ISDIR(inode->i_mode)) {
+		break;
+	case S_IFDIR:
 		inode->i_op = &linvfs_dir_inode_operations;
 		inode->i_fop = &linvfs_dir_operations;
-	} else if (S_ISLNK(inode->i_mode)) {
+		break;
+	case S_IFLNK:
 		inode->i_op = &linvfs_symlink_inode_operations;
 		if (inode->i_blocks)
 			inode->i_mapping->a_ops = &linvfs_aops;
-	} else {
+		break;
+	default:
 		inode->i_op = &linvfs_file_inode_operations;
 		init_special_inode(inode, inode->i_mode, inode->i_rdev);
+		break;
 	}
 }
 
@@ -167,16 +172,23 @@ xfs_revalidate_inode(
 {
 	struct inode		*inode = LINVFS_GET_IP(vp);
 
-	inode->i_mode	= (ip->i_d.di_mode & MODEMASK) | VTTOIF(vp->v_type);
+	inode->i_mode	= ip->i_d.di_mode;
 	inode->i_nlink	= ip->i_d.di_nlink;
 	inode->i_uid	= ip->i_d.di_uid;
 	inode->i_gid	= ip->i_d.di_gid;
-	if (((1 << vp->v_type) & ((1<<VBLK) | (1<<VCHR))) == 0) {
+
+	switch (inode->i_mode & S_IFMT) {
+	case S_IFBLK:
+	case S_IFCHR:
+		inode->i_rdev =
+			MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
+			      sysv_minor(ip->i_df.if_u2.if_rdev));
+		break;
+	default:
 		inode->i_rdev = 0;
-	} else {
-		xfs_dev_t dev = ip->i_df.if_u2.if_rdev;
-		inode->i_rdev = MKDEV(sysv_major(dev) & 0x1ff, sysv_minor(dev));
+		break;
 	}
+
 	inode->i_blksize = PAGE_CACHE_SIZE;
 	inode->i_generation = ip->i_d.di_gen;
 	i_size_write(inode, ip->i_d.di_size);
@@ -231,7 +243,6 @@ xfs_initialize_vnode(
 	 * finish our work.
 	 */
 	if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) {
-		vp->v_type = IFTOVT(ip->i_d.di_mode);
 		xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip);
 		xfs_set_inodeops(inode);
 	
@@ -274,8 +285,7 @@ linvfs_alloc_inode(
 {
 	vnode_t			*vp;
 
-	vp = (vnode_t *)kmem_cache_alloc(linvfs_inode_zone, 
-                kmem_flags_convert(KM_SLEEP));
+	vp = kmem_cache_alloc(xfs_vnode_zone, kmem_flags_convert(KM_SLEEP));
 	if (!vp)
 		return NULL;
 	return LINVFS_GET_IP(vp);
@@ -285,11 +295,11 @@ STATIC void
 linvfs_destroy_inode(
 	struct inode		*inode)
 {
-	kmem_cache_free(linvfs_inode_zone, LINVFS_GET_VP(inode));
+	kmem_zone_free(xfs_vnode_zone, LINVFS_GET_VP(inode));
 }
 
 STATIC void
-init_once(
+linvfs_inode_init_once(
 	void			*data,
 	kmem_cache_t		*cachep,
 	unsigned long		flags)
@@ -302,21 +312,41 @@ init_once(
 }
 
 STATIC int
-init_inodecache( void )
+linvfs_init_zones(void)
 {
-	linvfs_inode_zone = kmem_cache_create("linvfs_icache",
+	xfs_vnode_zone = kmem_cache_create("xfs_vnode",
 				sizeof(vnode_t), 0, SLAB_RECLAIM_ACCOUNT,
-				init_once, NULL);
-	if (linvfs_inode_zone == NULL)
-		return -ENOMEM;
+				linvfs_inode_init_once, NULL);
+	if (!xfs_vnode_zone)
+		goto out;
+
+	xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
+	if (!xfs_ioend_zone)
+		goto out_destroy_vnode_zone;
+
+	xfs_ioend_pool = mempool_create(4 * MAX_BUF_PER_PAGE,
+			mempool_alloc_slab, mempool_free_slab,
+			xfs_ioend_zone);
+	if (!xfs_ioend_pool)
+		goto out_free_ioend_zone;
+
 	return 0;
+
+
+ out_free_ioend_zone:
+	kmem_zone_destroy(xfs_ioend_zone);
+ out_destroy_vnode_zone:
+	kmem_zone_destroy(xfs_vnode_zone);
+ out:
+	return -ENOMEM;
 }
 
 STATIC void
-destroy_inodecache( void )
+linvfs_destroy_zones(void)
 {
-	if (kmem_cache_destroy(linvfs_inode_zone))
-		printk(KERN_WARNING "%s: cache still in use!\n", __FUNCTION__);
+	mempool_destroy(xfs_ioend_pool);
+	kmem_zone_destroy(xfs_vnode_zone);
+	kmem_zone_destroy(xfs_ioend_zone);
 }
 
 /*
@@ -354,17 +384,38 @@ linvfs_clear_inode(
 	struct inode		*inode)
 {
 	vnode_t			*vp = LINVFS_GET_VP(inode);
+	int			error, cache;
 
-	if (vp) {
-		vn_rele(vp);
-		vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address);
-		/*
-		 * Do all our cleanup, and remove this vnode.
-		 */
-		vn_remove(vp);
+	vn_trace_entry(vp, "clear_inode", (inst_t *)__return_address);
+
+	XFS_STATS_INC(vn_rele);
+	XFS_STATS_INC(vn_remove);
+	XFS_STATS_INC(vn_reclaim);
+	XFS_STATS_DEC(vn_active);
+
+	/*
+	 * This can happen because xfs_iget_core calls xfs_idestroy if we
+	 * find an inode with di_mode == 0 but without IGET_CREATE set.
+	 */
+	if (vp->v_fbhv)
+		VOP_INACTIVE(vp, NULL, cache);
+
+	VN_LOCK(vp);
+	vp->v_flag &= ~VMODIFIED;
+	VN_UNLOCK(vp, 0);
+
+	if (vp->v_fbhv) {
+		VOP_RECLAIM(vp, error);
+		if (error)
+			panic("vn_purge: cannot reclaim");
 	}
-}
 
+	ASSERT(vp->v_fbhv == NULL);
+
+#ifdef XFS_VNODE_TRACE
+	ktrace_free(vp->v_trace);
+#endif
+}
 
 /*
  * Enqueue a work item to be picked up by the vfs xfssyncd thread.
@@ -466,25 +517,16 @@ xfssyncd(
 {
 	long			timeleft;
 	vfs_t			*vfsp = (vfs_t *) arg;
-	struct list_head	tmp;
 	struct vfs_sync_work	*work, *n;
+	LIST_HEAD		(tmp);
 
-	daemonize("xfssyncd");
-
-	vfsp->vfs_sync_work.w_vfs = vfsp;
-	vfsp->vfs_sync_work.w_syncer = vfs_sync_worker;
-	vfsp->vfs_sync_task = current;
-	wmb();
-	wake_up(&vfsp->vfs_wait_sync_task);
-
-	INIT_LIST_HEAD(&tmp);
 	timeleft = (xfs_syncd_centisecs * HZ) / 100;
 	for (;;) {
 		set_current_state(TASK_INTERRUPTIBLE);
 		timeleft = schedule_timeout(timeleft);
 		/* swsusp */
 		try_to_freeze();
-		if (vfsp->vfs_flag & VFS_UMOUNT)
+		if (kthread_should_stop())
 			break;
 
 		spin_lock(&vfsp->vfs_sync_lock);
@@ -513,10 +555,6 @@ xfssyncd(
 		}
 	}
 
-	vfsp->vfs_sync_task = NULL;
-	wmb();
-	wake_up(&vfsp->vfs_wait_sync_task);
-
 	return 0;
 }
 
@@ -524,13 +562,11 @@ STATIC int
 linvfs_start_syncd(
 	vfs_t			*vfsp)
 {
-	int			pid;
-
-	pid = kernel_thread(xfssyncd, (void *) vfsp,
-			CLONE_VM | CLONE_FS | CLONE_FILES);
-	if (pid < 0)
-		return -pid;
-	wait_event(vfsp->vfs_wait_sync_task, vfsp->vfs_sync_task);
+	vfsp->vfs_sync_work.w_syncer = vfs_sync_worker;
+	vfsp->vfs_sync_work.w_vfs = vfsp;
+	vfsp->vfs_sync_task = kthread_run(xfssyncd, vfsp, "xfssyncd");
+	if (IS_ERR(vfsp->vfs_sync_task))
+		return -PTR_ERR(vfsp->vfs_sync_task);
 	return 0;
 }
 
@@ -538,11 +574,7 @@ STATIC void
 linvfs_stop_syncd(
 	vfs_t			*vfsp)
 {
-	vfsp->vfs_flag |= VFS_UMOUNT;
-	wmb();
-
-	wake_up_process(vfsp->vfs_sync_task);
-	wait_event(vfsp->vfs_wait_sync_task, !vfsp->vfs_sync_task);
+	kthread_stop(vfsp->vfs_sync_task);
 }
 
 STATIC void
@@ -866,9 +898,9 @@ init_xfs_fs( void )
 
 	ktrace_init(64);
 
-	error = init_inodecache();
+	error = linvfs_init_zones();
 	if (error < 0)
-		goto undo_inodecache;
+		goto undo_zones;
 
 	error = pagebuf_init();
 	if (error < 0)
@@ -889,9 +921,9 @@ undo_register:
 	pagebuf_terminate();
 
 undo_pagebuf:
-	destroy_inodecache();
+	linvfs_destroy_zones();
 
-undo_inodecache:
+undo_zones:
 	return error;
 }
 
@@ -903,7 +935,7 @@ exit_xfs_fs( void )
 	unregister_filesystem(&xfs_fs_type);
 	xfs_cleanup();
 	pagebuf_terminate();
-	destroy_inodecache();
+	linvfs_destroy_zones();
 	ktrace_uninit();
 }
 

fs/xfs/linux-2.6/xfs_vfs.c

@@ -251,7 +251,6 @@ vfs_allocate( void )
 	bhv_head_init(VFS_BHVHEAD(vfsp), "vfs");
 	INIT_LIST_HEAD(&vfsp->vfs_sync_list);
 	spin_lock_init(&vfsp->vfs_sync_lock);
-	init_waitqueue_head(&vfsp->vfs_wait_sync_task);
 	init_waitqueue_head(&vfsp->vfs_wait_single_sync_task);
 	return vfsp;
 }

fs/xfs/linux-2.6/xfs_vfs.h

@@ -65,7 +65,6 @@ typedef struct vfs {
 	spinlock_t		vfs_sync_lock;	/* work item list lock */
 	int 			vfs_sync_seq;	/* sync thread generation no. */
 	wait_queue_head_t	vfs_wait_single_sync_task;
-	wait_queue_head_t	vfs_wait_sync_task;
 } vfs_t;
 
 #define vfs_fbhv		vfs_bh.bh_first	/* 1st on vfs behavior chain */
@@ -96,7 +95,6 @@ typedef enum {
 #define VFS_RDONLY		0x0001	/* read-only vfs */
 #define VFS_GRPID		0x0002	/* group-ID assigned from directory */
 #define VFS_DMI			0x0004	/* filesystem has the DMI enabled */
-#define VFS_UMOUNT		0x0008	/* unmount in progress */
 #define VFS_END			0x0008	/* max flag */
 
 #define SYNC_ATTR		0x0001	/* sync attributes */

fs/xfs/linux-2.6/xfs_vnode.c

@@ -42,93 +42,33 @@ DEFINE_SPINLOCK(vnumber_lock);
  */
 #define NVSYNC                  37
 #define vptosync(v)             (&vsync[((unsigned long)v) % NVSYNC])
-sv_t vsync[NVSYNC];
-
-/*
- * Translate stat(2) file types to vnode types and vice versa.
- * Aware of numeric order of S_IFMT and vnode type values.
- */
-enum vtype iftovt_tab[] = {
-	VNON, VFIFO, VCHR, VNON, VDIR, VNON, VBLK, VNON,
-	VREG, VNON, VLNK, VNON, VSOCK, VNON, VNON, VNON
-};
-
-u_short vttoif_tab[] = {
-	0, S_IFREG, S_IFDIR, S_IFBLK, S_IFCHR, S_IFLNK, S_IFIFO, 0, S_IFSOCK
-};
+STATIC wait_queue_head_t vsync[NVSYNC];
 
 
 void
 vn_init(void)
 {
-	register sv_t *svp;
-	register int i;
+	int i;
 
-	for (svp = vsync, i = 0; i < NVSYNC; i++, svp++)
-		init_sv(svp, SV_DEFAULT, "vsy", i);
+	for (i = 0; i < NVSYNC; i++)
+		init_waitqueue_head(&vsync[i]);
 }
 
-/*
- * Clean a vnode of filesystem-specific data and prepare it for reuse.
- */
-STATIC int
-vn_reclaim(
+void
+vn_iowait(
 	struct vnode	*vp)
 {
-	int		error;
+	wait_queue_head_t *wq = vptosync(vp);
 
-	XFS_STATS_INC(vn_reclaim);
-	vn_trace_entry(vp, "vn_reclaim", (inst_t *)__return_address);
-
-	/*
-	 * Only make the VOP_RECLAIM call if there are behaviors
-	 * to call.
-	 */
-	if (vp->v_fbhv) {
-		VOP_RECLAIM(vp, error);
-		if (error)
-			return -error;
-	}
-	ASSERT(vp->v_fbhv == NULL);
-
-	VN_LOCK(vp);
-	vp->v_flag &= (VRECLM|VWAIT);
-	VN_UNLOCK(vp, 0);
-
-	vp->v_type = VNON;
-	vp->v_fbhv = NULL;
-
-#ifdef XFS_VNODE_TRACE
-	ktrace_free(vp->v_trace);
-	vp->v_trace = NULL;
-#endif
-
-	return 0;
-}
-
-STATIC void
-vn_wakeup(
-	struct vnode	*vp)
-{
-	VN_LOCK(vp);
-	if (vp->v_flag & VWAIT)
-		sv_broadcast(vptosync(vp));
-	vp->v_flag &= ~(VRECLM|VWAIT|VMODIFIED);
-	VN_UNLOCK(vp, 0);
+	wait_event(*wq, (atomic_read(&vp->v_iocount) == 0));
 }
 
-int
-vn_wait(
+void
+vn_iowake(
 	struct vnode	*vp)
 {
-	VN_LOCK(vp);
-	if (vp->v_flag & (VINACT | VRECLM)) {
-		vp->v_flag |= VWAIT;
-		sv_wait(vptosync(vp), PINOD, &vp->v_lock, 0);
-		return 1;
-	}
-	VN_UNLOCK(vp, 0);
-	return 0;
+	if (atomic_dec_and_test(&vp->v_iocount))
+		wake_up(vptosync(vp));
 }
 
 struct vnode *
@@ -154,6 +94,8 @@ vn_initialize(
 	/* Initialize the first behavior and the behavior chain head. */
 	vn_bhv_head_init(VN_BHV_HEAD(vp), "vnode");
 
+	atomic_set(&vp->v_iocount, 0);
+
 #ifdef	XFS_VNODE_TRACE
 	vp->v_trace = ktrace_alloc(VNODE_TRACE_SIZE, KM_SLEEP);
 #endif	/* XFS_VNODE_TRACE */
@@ -162,30 +104,6 @@ vn_initialize(
 	return vp;
 }
 
-/*
- * Get a reference on a vnode.
- */
-vnode_t *
-vn_get(
-	struct vnode	*vp,
-	vmap_t		*vmap)
-{
-	struct inode	*inode;
-
-	XFS_STATS_INC(vn_get);
-	inode = LINVFS_GET_IP(vp);
-	if (inode->i_state & I_FREEING)
-		return NULL;
-
-	inode = ilookup(vmap->v_vfsp->vfs_super, vmap->v_ino);
-	if (!inode)	/* Inode not present */
-		return NULL;
-
-	vn_trace_exit(vp, "vn_get", (inst_t *)__return_address);
-
-	return vp;
-}
-
 /*
  * Revalidate the Linux inode from the vattr.
  * Note: i_size _not_ updated; we must hold the inode
@@ -198,7 +116,7 @@ vn_revalidate_core(
 {
 	struct inode	*inode = LINVFS_GET_IP(vp);
 
-	inode->i_mode	    = VTTOIF(vap->va_type) | vap->va_mode;
+	inode->i_mode	    = vap->va_mode;
 	inode->i_nlink	    = vap->va_nlink;
 	inode->i_uid	    = vap->va_uid;
 	inode->i_gid	    = vap->va_gid;
@@ -246,71 +164,6 @@ vn_revalidate(
 	return -error;
 }
 
-/*
- * purge a vnode from the cache
- * At this point the vnode is guaranteed to have no references (vn_count == 0)
- * The caller has to make sure that there are no ways someone could
- * get a handle (via vn_get) on the vnode (usually done via a mount/vfs lock).
- */
-void
-vn_purge(
-	struct vnode	*vp,
-	vmap_t		*vmap)
-{
-	vn_trace_entry(vp, "vn_purge", (inst_t *)__return_address);
-
-again:
-	/*
-	 * Check whether vp has already been reclaimed since our caller
-	 * sampled its version while holding a filesystem cache lock that
-	 * its VOP_RECLAIM function acquires.
-	 */
-	VN_LOCK(vp);
-	if (vp->v_number != vmap->v_number) {
-		VN_UNLOCK(vp, 0);
-		return;
-	}
-
-	/*
-	 * If vp is being reclaimed or inactivated, wait until it is inert,
-	 * then proceed.  Can't assume that vnode is actually reclaimed
-	 * just because the reclaimed flag is asserted -- a vn_alloc
-	 * reclaim can fail.
-	 */
-	if (vp->v_flag & (VINACT | VRECLM)) {
-		ASSERT(vn_count(vp) == 0);
-		vp->v_flag |= VWAIT;
-		sv_wait(vptosync(vp), PINOD, &vp->v_lock, 0);
-		goto again;
-	}
-
-	/*
-	 * Another process could have raced in and gotten this vnode...
-	 */
-	if (vn_count(vp) > 0) {
-		VN_UNLOCK(vp, 0);
-		return;
-	}
-
-	XFS_STATS_DEC(vn_active);
-	vp->v_flag |= VRECLM;
-	VN_UNLOCK(vp, 0);
-
-	/*
-	 * Call VOP_RECLAIM and clean vp. The FSYNC_INVAL flag tells
-	 * vp's filesystem to flush and invalidate all cached resources.
-	 * When vn_reclaim returns, vp should have no private data,
-	 * either in a system cache or attached to v_data.
-	 */
-	if (vn_reclaim(vp) != 0)
-		panic("vn_purge: cannot reclaim");
-
-	/*
-	 * Wakeup anyone waiting for vp to be reclaimed.
-	 */
-	vn_wakeup(vp);
-}
-
 /*
  * Add a reference to a referenced vnode.
  */
@@ -330,80 +183,6 @@ vn_hold(
 	return vp;
 }
 
-/*
- *  Call VOP_INACTIVE on last reference.
- */
-void
-vn_rele(
-	struct vnode	*vp)
-{
-	int		vcnt;
-	int		cache;
-
-	XFS_STATS_INC(vn_rele);
-
-	VN_LOCK(vp);
-
-	vn_trace_entry(vp, "vn_rele", (inst_t *)__return_address);
-	vcnt = vn_count(vp);
-
-	/*
-	 * Since we always get called from put_inode we know
-	 * that i_count won't be decremented after we
-	 * return.
-	 */
-	if (!vcnt) {
-		/*
-		 * As soon as we turn this on, noone can find us in vn_get
-		 * until we turn off VINACT or VRECLM
-		 */
-		vp->v_flag |= VINACT;
-		VN_UNLOCK(vp, 0);
-
-		/*
-		 * Do not make the VOP_INACTIVE call if there
-		 * are no behaviors attached to the vnode to call.
-		 */
-		if (vp->v_fbhv)
-			VOP_INACTIVE(vp, NULL, cache);
-
-		VN_LOCK(vp);
-		if (vp->v_flag & VWAIT)
-			sv_broadcast(vptosync(vp));
-
-		vp->v_flag &= ~(VINACT|VWAIT|VRECLM|VMODIFIED);
-	}
-
-	VN_UNLOCK(vp, 0);
-
-	vn_trace_exit(vp, "vn_rele", (inst_t *)__return_address);
-}
-
-/*
- * Finish the removal of a vnode.
- */
-void
-vn_remove(
-	struct vnode	*vp)
-{
-	vmap_t		vmap;
-
-	/* Make sure we don't do this to the same vnode twice */
-	if (!(vp->v_fbhv))
-		return;
-
-	XFS_STATS_INC(vn_remove);
-	vn_trace_exit(vp, "vn_remove", (inst_t *)__return_address);
-
-	/*
-	 * After the following purge the vnode
-	 * will no longer exist.
-	 */
-	VMAP(vp, vmap);
-	vn_purge(vp, &vmap);
-}
-
-
 #ifdef	XFS_VNODE_TRACE
 
 #define KTRACE_ENTER(vp, vk, s, line, ra)			\

fs/xfs/linux-2.6/xfs_vnode.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -65,10 +65,6 @@ struct vattr;
 struct xfs_iomap;
 struct attrlist_cursor_kern;
 
-/*
- * Vnode types.  VNON means no type.
- */
-enum vtype	{ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VFIFO, VBAD, VSOCK };
 
 typedef xfs_ino_t vnumber_t;
 typedef struct dentry vname_t;
@@ -77,15 +73,14 @@ typedef bhv_head_t vn_bhv_head_t;
 /*
  * MP locking protocols:
  *	v_flag, v_vfsp				VN_LOCK/VN_UNLOCK
- *	v_type					read-only or fs-dependent
  */
 typedef struct vnode {
 	__u32		v_flag;			/* vnode flags (see below) */
-	enum vtype	v_type;			/* vnode type */
 	struct vfs	*v_vfsp;		/* ptr to containing VFS */
 	vnumber_t	v_number;		/* in-core vnode number */
 	vn_bhv_head_t	v_bh;			/* behavior head */
 	spinlock_t	v_lock;			/* VN_LOCK/VN_UNLOCK */
+	atomic_t	v_iocount;		/* outstanding I/O count */
 #ifdef XFS_VNODE_TRACE
 	struct ktrace	*v_trace;		/* trace header structure    */
 #endif
@@ -93,6 +88,12 @@ typedef struct vnode {
 	/* inode MUST be last */
 } vnode_t;
 
+#define VN_ISLNK(vp)	S_ISLNK((vp)->v_inode.i_mode)
+#define VN_ISREG(vp)	S_ISREG((vp)->v_inode.i_mode)
+#define VN_ISDIR(vp)	S_ISDIR((vp)->v_inode.i_mode)
+#define VN_ISCHR(vp)	S_ISCHR((vp)->v_inode.i_mode)
+#define VN_ISBLK(vp)	S_ISBLK((vp)->v_inode.i_mode)
+
 #define v_fbhv			v_bh.bh_first	       /* first behavior */
 #define v_fops			v_bh.bh_first->bd_ops  /* first behavior ops */
 
@@ -132,23 +133,9 @@ typedef enum {
 #define LINVFS_GET_VP(inode)	((vnode_t *)list_entry(inode, vnode_t, v_inode))
 #define LINVFS_GET_IP(vp)	(&(vp)->v_inode)
 
-/*
- * Convert between vnode types and inode formats (since POSIX.1
- * defines mode word of stat structure in terms of inode formats).
- */
-extern enum vtype	iftovt_tab[];
-extern u_short		vttoif_tab[];
-#define IFTOVT(mode)	(iftovt_tab[((mode) & S_IFMT) >> 12])
-#define VTTOIF(indx)	(vttoif_tab[(int)(indx)])
-#define MAKEIMODE(indx, mode)	(int)(VTTOIF(indx) | (mode))
-
-
 /*
  * Vnode flags.
  */
-#define VINACT		       0x1	/* vnode is being inactivated	*/
-#define VRECLM		       0x2	/* vnode is being reclaimed	*/
-#define VWAIT		       0x4	/* waiting for VINACT/VRECLM to end */
 #define VMODIFIED	       0x8	/* XFS inode state possibly differs */
 					/* to the Linux inode state.	*/
 
@@ -408,7 +395,6 @@ typedef struct vnodeops {
  */
 typedef struct vattr {
 	int		va_mask;	/* bit-mask of attributes present */
-	enum vtype	va_type;	/* vnode type (for create) */
 	mode_t		va_mode;	/* file access mode and type */
 	xfs_nlink_t	va_nlink;	/* number of references to file */
 	uid_t		va_uid;		/* owner user id */
@@ -498,26 +484,11 @@ typedef struct vattr {
  * Check whether mandatory file locking is enabled.
  */
 #define MANDLOCK(vp, mode)	\
-	((vp)->v_type == VREG && ((mode) & (VSGID|(VEXEC>>3))) == VSGID)
+	(VN_ISREG(vp) && ((mode) & (VSGID|(VEXEC>>3))) == VSGID)
 
 extern void	vn_init(void);
-extern int	vn_wait(struct vnode *);
 extern vnode_t	*vn_initialize(struct inode *);
 
-/*
- * Acquiring and invalidating vnodes:
- *
- *	if (vn_get(vp, version, 0))
- *		...;
- *	vn_purge(vp, version);
- *
- * vn_get and vn_purge must be called with vmap_t arguments, sampled
- * while a lock that the vnode's VOP_RECLAIM function acquires is
- * held, to ensure that the vnode sampled with the lock held isn't
- * recycled (VOP_RECLAIMed) or deallocated between the release of the lock
- * and the subsequent vn_get or vn_purge.
- */
-
 /*
  * vnode_map structures _must_ match vn_epoch and vnode structure sizes.
  */
@@ -531,11 +502,11 @@ typedef struct vnode_map {
 			 (vmap).v_number = (vp)->v_number,	\
 			 (vmap).v_ino	 = (vp)->v_inode.i_ino; }
 
-extern void	vn_purge(struct vnode *, vmap_t *);
-extern vnode_t	*vn_get(struct vnode *, vmap_t *);
 extern int	vn_revalidate(struct vnode *);
 extern void	vn_revalidate_core(struct vnode *, vattr_t *);
-extern void	vn_remove(struct vnode *);
+
+extern void	vn_iowait(struct vnode *vp);
+extern void	vn_iowake(struct vnode *vp);
 
 static inline int vn_count(struct vnode *vp)
 {
@@ -546,7 +517,6 @@ static inline int vn_count(struct vnode *vp)
  * Vnode reference counting functions (and macros for compatibility).
  */
 extern vnode_t	*vn_hold(struct vnode *);
-extern void	vn_rele(struct vnode *);
 
 #if defined(XFS_VNODE_TRACE)
 #define VN_HOLD(vp)		\
@@ -560,6 +530,12 @@ extern void	vn_rele(struct vnode *);
 #define VN_RELE(vp)		(iput(LINVFS_GET_IP(vp)))
 #endif
 
+static inline struct vnode *vn_grab(struct vnode *vp)
+{
+	struct inode *inode = igrab(LINVFS_GET_IP(vp));
+	return inode ? LINVFS_GET_VP(inode) : NULL;
+}
+
 /*
  * Vname handling macros.
  */

fs/xfs/quota/Makefile

@@ -0,0 +1 @@
+include $(TOPDIR)/fs/xfs/quota/Makefile-linux-$(VERSION).$(PATCHLEVEL)

fs/xfs/quota/Makefile-linux-2.6

@@ -0,0 +1,53 @@
+#
+# Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+#
+# This program is free software; you can redistribute it and/or modify it
+# under the terms of version 2 of the GNU General Public License as
+# published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it would be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+#
+# Further, this software is distributed without any warranty that it is
+# free of the rightful claim of any third person regarding infringement
+# or the like.	Any license provided herein, whether implied or
+# otherwise, applies only to this software file.  Patent licenses, if
+# any, provided herein do not apply to combinations of this program with
+# other software, or any other product whatsoever.
+#
+# You should have received a copy of the GNU General Public License along
+# with this program; if not, write the Free Software Foundation, Inc., 59
+# Temple Place - Suite 330, Boston MA 02111-1307, USA.
+#
+# Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
+# Mountain View, CA  94043, or:
+#
+# http://www.sgi.com
+#
+# For further information regarding this notice, see:
+#
+# http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+#
+
+EXTRA_CFLAGS += -I $(TOPDIR)/fs/xfs -I $(TOPDIR)/fs/xfs/linux-2.6
+
+ifeq ($(CONFIG_XFS_DEBUG),y)
+	EXTRA_CFLAGS += -g -DDEBUG
+	#EXTRA_CFLAGS += -DQUOTADEBUG
+endif
+ifeq ($(CONFIG_XFS_TRACE),y)
+	EXTRA_CFLAGS += -DXFS_DQUOT_TRACE
+	EXTRA_CFLAGS += -DXFS_VNODE_TRACE
+endif
+
+obj-$(CONFIG_XFS_QUOTA)		+= xfs_quota.o
+
+xfs_quota-y			+= xfs_dquot.o \
+				   xfs_dquot_item.o \
+				   xfs_trans_dquot.o \
+				   xfs_qm_syscalls.o \
+				   xfs_qm_bhv.o \
+				   xfs_qm.o
+
+xfs_quota-$(CONFIG_PROC_FS)	+= xfs_qm_stats.o

fs/xfs/quota/xfs_dquot.c

@@ -421,7 +421,7 @@ xfs_qm_init_dquot_blk(
  */
 STATIC int
 xfs_qm_dqalloc(
-	xfs_trans_t	*tp,
+	xfs_trans_t	**tpp,
 	xfs_mount_t	*mp,
 	xfs_dquot_t	*dqp,
 	xfs_inode_t	*quotip,
@@ -433,6 +433,7 @@ xfs_qm_dqalloc(
 	xfs_bmbt_irec_t map;
 	int		nmaps, error, committed;
 	xfs_buf_t	*bp;
+	xfs_trans_t	*tp = *tpp;
 
 	ASSERT(tp != NULL);
 	xfs_dqtrace_entry(dqp, "DQALLOC");
@@ -492,10 +493,32 @@ xfs_qm_dqalloc(
 	xfs_qm_init_dquot_blk(tp, mp, INT_GET(dqp->q_core.d_id, ARCH_CONVERT),
 			      dqp->dq_flags & XFS_DQ_ALLTYPES, bp);
 
-	if ((error = xfs_bmap_finish(&tp, &flist, firstblock, &committed))) {
+	/*
+	 * xfs_bmap_finish() may commit the current transaction and
+	 * start a second transaction if the freelist is not empty.
+	 *
+	 * Since we still want to modify this buffer, we need to
+	 * ensure that the buffer is not released on commit of
+	 * the first transaction and ensure the buffer is added to the
+	 * second transaction.
+	 *
+	 * If there is only one transaction then don't stop the buffer
+	 * from being released when it commits later on.
+	 */
+
+	xfs_trans_bhold(tp, bp);
+
+	if ((error = xfs_bmap_finish(tpp, &flist, firstblock, &committed))) {
 		goto error1;
 	}
 
+	if (committed) {
+		tp = *tpp;
+		xfs_trans_bjoin(tp, bp);
+	} else {
+		xfs_trans_bhold_release(tp, bp);
+	}
+
 	*O_bpp = bp;
 	return 0;
 
@@ -514,7 +537,7 @@ xfs_qm_dqalloc(
  */
 STATIC int
 xfs_qm_dqtobp(
-	xfs_trans_t		*tp,
+	xfs_trans_t		**tpp,
 	xfs_dquot_t		*dqp,
 	xfs_disk_dquot_t	**O_ddpp,
 	xfs_buf_t		**O_bpp,
@@ -528,6 +551,7 @@ xfs_qm_dqtobp(
 	xfs_disk_dquot_t *ddq;
 	xfs_dqid_t	id;
 	boolean_t	newdquot;
+	xfs_trans_t	*tp = (tpp ? *tpp : NULL);
 
 	mp = dqp->q_mount;
 	id = INT_GET(dqp->q_core.d_id, ARCH_CONVERT);
@@ -579,9 +603,10 @@ xfs_qm_dqtobp(
 				return (ENOENT);
 
 			ASSERT(tp);
-			if ((error = xfs_qm_dqalloc(tp, mp, dqp, quotip,
+			if ((error = xfs_qm_dqalloc(tpp, mp, dqp, quotip,
 						dqp->q_fileoffset, &bp)))
 				return (error);
+			tp = *tpp;
 			newdquot = B_TRUE;
 		} else {
 			/*
@@ -645,7 +670,7 @@ xfs_qm_dqtobp(
 /* ARGSUSED */
 STATIC int
 xfs_qm_dqread(
-	xfs_trans_t	*tp,
+	xfs_trans_t	**tpp,
 	xfs_dqid_t	id,
 	xfs_dquot_t	*dqp,	/* dquot to get filled in */
 	uint		flags)
@@ -653,15 +678,19 @@ xfs_qm_dqread(
 	xfs_disk_dquot_t *ddqp;
 	xfs_buf_t	 *bp;
 	int		 error;
+	xfs_trans_t	 *tp;
+
+	ASSERT(tpp);
 
 	/*
 	 * get a pointer to the on-disk dquot and the buffer containing it
 	 * dqp already knows its own type (GROUP/USER).
 	 */
 	xfs_dqtrace_entry(dqp, "DQREAD");
-	if ((error = xfs_qm_dqtobp(tp, dqp, &ddqp, &bp, flags))) {
+	if ((error = xfs_qm_dqtobp(tpp, dqp, &ddqp, &bp, flags))) {
 		return (error);
 	}
+	tp = *tpp;
 
 	/* copy everything from disk dquot to the incore dquot */
 	memcpy(&dqp->q_core, ddqp, sizeof(xfs_disk_dquot_t));
@@ -740,7 +769,7 @@ xfs_qm_idtodq(
 	 * Read it from disk; xfs_dqread() takes care of
 	 * all the necessary initialization of dquot's fields (locks, etc)
 	 */
-	if ((error = xfs_qm_dqread(tp, id, dqp, flags))) {
+	if ((error = xfs_qm_dqread(&tp, id, dqp, flags))) {
 		/*
 		 * This can happen if quotas got turned off (ESRCH),
 		 * or if the dquot didn't exist on disk and we ask to

fs/xfs/quota/xfs_dquot.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -113,20 +113,6 @@ typedef struct xfs_dquot {
 
 #define XFS_DQHOLD(dqp)		((dqp)->q_nrefs++)
 
-/*
- * Quota Accounting/Enforcement flags
- */
-#define XFS_ALL_QUOTA_ACCT	\
-		(XFS_UQUOTA_ACCT | XFS_GQUOTA_ACCT | XFS_PQUOTA_ACCT)
-#define XFS_ALL_QUOTA_ENFD	(XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD)
-#define XFS_ALL_QUOTA_CHKD	(XFS_UQUOTA_CHKD | XFS_OQUOTA_CHKD)
-
-#define XFS_IS_QUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_ALL_QUOTA_ACCT)
-#define XFS_IS_QUOTA_ENFORCED(mp)	((mp)->m_qflags & XFS_ALL_QUOTA_ENFD)
-#define XFS_IS_UQUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_UQUOTA_ACCT)
-#define XFS_IS_PQUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_PQUOTA_ACCT)
-#define XFS_IS_GQUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_GQUOTA_ACCT)
-
 #ifdef DEBUG
 static inline int
 XFS_DQ_IS_LOCKED(xfs_dquot_t *dqp)

fs/xfs/quota/xfs_dquot_item.c

@@ -509,6 +509,7 @@ xfs_qm_qoff_logitem_format(xfs_qoff_logitem_t	*qf,
 
 	log_vector->i_addr = (xfs_caddr_t)&(qf->qql_format);
 	log_vector->i_len = sizeof(xfs_qoff_logitem_t);
+	XLOG_VEC_SET_TYPE(log_vector, XLOG_REG_TYPE_QUOTAOFF);
 	qf->qql_format.qf_size = 1;
 }
 

fs/xfs/quota/xfs_qm.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -365,16 +365,6 @@ xfs_qm_mount_quotas(
 	int		error = 0;
 	uint		sbf;
 
-	/*
-	 * If a file system had quotas running earlier, but decided to
-	 * mount without -o uquota/pquota/gquota options, revoke the
-	 * quotachecked license, and bail out.
-	 */
-	if (! XFS_IS_QUOTA_ON(mp) &&
-	    (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_ACCT)) {
-		mp->m_qflags = 0;
-		goto write_changes;
-	}
 
 	/*
 	 * If quotas on realtime volumes is not supported, we disable
@@ -388,11 +378,8 @@ xfs_qm_mount_quotas(
 		goto write_changes;
 	}
 
-#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
-	cmn_err(CE_NOTE, "Attempting to turn on disk quotas.");
-#endif
-
 	ASSERT(XFS_IS_QUOTA_RUNNING(mp));
+
 	/*
 	 * Allocate the quotainfo structure inside the mount struct, and
 	 * create quotainode(s), and change/rev superblock if necessary.
@@ -410,19 +397,14 @@ xfs_qm_mount_quotas(
 	 */
 	if (XFS_QM_NEED_QUOTACHECK(mp) &&
 		!(mfsi_flags & XFS_MFSI_NO_QUOTACHECK)) {
-#ifdef DEBUG
-		cmn_err(CE_NOTE, "Doing a quotacheck. Please wait.");
-#endif
 		if ((error = xfs_qm_quotacheck(mp))) {
 			/* Quotacheck has failed and quotas have
 			 * been disabled.
 			 */
 			return XFS_ERROR(error);
 		}
-#ifdef DEBUG
-		cmn_err(CE_NOTE, "Done quotacheck.");
-#endif
 	}
+
  write_changes:
 	/*
 	 * We actually don't have to acquire the SB_LOCK at all.
@@ -2010,7 +1992,7 @@ xfs_qm_quotacheck(
 		ASSERT(mp->m_quotainfo != NULL);
 		ASSERT(xfs_Gqm != NULL);
 		xfs_qm_destroy_quotainfo(mp);
-		xfs_mount_reset_sbqflags(mp);
+		(void)xfs_mount_reset_sbqflags(mp);
 	} else {
 		cmn_err(CE_NOTE, "XFS quotacheck %s: Done.", mp->m_fsname);
 	}

fs/xfs/quota/xfs_qm.h

@@ -184,8 +184,6 @@ typedef struct xfs_dquot_acct {
 #define XFS_QM_HOLD(xqm)	((xqm)->qm_nrefs++)
 #define XFS_QM_RELE(xqm)	((xqm)->qm_nrefs--)
 
-extern void		xfs_mount_reset_sbqflags(xfs_mount_t *);
-
 extern void		xfs_qm_destroy_quotainfo(xfs_mount_t *);
 extern int		xfs_qm_mount_quotas(xfs_mount_t *, int);
 extern void		xfs_qm_mount_quotainit(xfs_mount_t *, uint);

fs/xfs/quota/xfs_qm_bhv.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -229,48 +229,6 @@ xfs_qm_syncall(
 	return error;
 }
 
-/*
- * Clear the quotaflags in memory and in the superblock.
- */
-void
-xfs_mount_reset_sbqflags(
-	xfs_mount_t		*mp)
-{
-	xfs_trans_t		*tp;
-	unsigned long		s;
-
-	mp->m_qflags = 0;
-	/*
-	 * It is OK to look at sb_qflags here in mount path,
-	 * without SB_LOCK.
-	 */
-	if (mp->m_sb.sb_qflags == 0)
-		return;
-	s = XFS_SB_LOCK(mp);
-	mp->m_sb.sb_qflags = 0;
-	XFS_SB_UNLOCK(mp, s);
-
-	/*
-	 * if the fs is readonly, let the incore superblock run
-	 * with quotas off but don't flush the update out to disk
-	 */
-	if (XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY)
-		return;
-#ifdef QUOTADEBUG
-	xfs_fs_cmn_err(CE_NOTE, mp, "Writing superblock quota changes");
-#endif
-	tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SBCHANGE);
-	if (xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
-				      XFS_DEFAULT_LOG_COUNT)) {
-		xfs_trans_cancel(tp, 0);
-		xfs_fs_cmn_err(CE_ALERT, mp,
-			"xfs_mount_reset_sbqflags: Superblock update failed!");
-		return;
-	}
-	xfs_mod_sb(tp, XFS_SB_QFLAGS);
-	xfs_trans_commit(tp, 0, NULL);
-}
-
 STATIC int
 xfs_qm_newmount(
 	xfs_mount_t	*mp,

fs/xfs/quota/xfs_qm_syscalls.c

@@ -1053,7 +1053,6 @@ xfs_qm_dqrele_all_inodes(
 	struct xfs_mount *mp,
 	uint		 flags)
 {
-	vmap_t		vmap;
 	xfs_inode_t	*ip, *topino;
 	uint		ireclaims;
 	vnode_t		*vp;
@@ -1061,8 +1060,8 @@ xfs_qm_dqrele_all_inodes(
 
 	ASSERT(mp->m_quotainfo);
 
-again:
 	XFS_MOUNT_ILOCK(mp);
+again:
 	ip = mp->m_inodes;
 	if (ip == NULL) {
 		XFS_MOUNT_IUNLOCK(mp);
@@ -1090,18 +1089,14 @@ again:
 		}
 		vnode_refd = B_FALSE;
 		if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
-			/*
-			 * Sample vp mapping while holding the mplock, lest
-			 * we come across a non-existent vnode.
-			 */
-			VMAP(vp, vmap);
 			ireclaims = mp->m_ireclaims;
 			topino = mp->m_inodes;
-			XFS_MOUNT_IUNLOCK(mp);
+			vp = vn_grab(vp);
+			if (!vp)
+				goto again;
 
+			XFS_MOUNT_IUNLOCK(mp);
 			/* XXX restart limit ? */
-			if ( ! (vp = vn_get(vp, &vmap)))
-				goto again;
 			xfs_ilock(ip, XFS_ILOCK_EXCL);
 			vnode_refd = B_TRUE;
 		} else {
@@ -1137,7 +1132,6 @@ again:
 		 */
 		if (topino != mp->m_inodes || mp->m_ireclaims != ireclaims) {
 			/* XXX use a sentinel */
-			XFS_MOUNT_IUNLOCK(mp);
 			goto again;
 		}
 		ip = ip->i_mnext;

fs/xfs/support/debug.c

@@ -31,6 +31,7 @@
  */
 
 #include "debug.h"
+#include "spin.h"
 
 #include <asm/page.h>
 #include <linux/sched.h>

fs/xfs/xfs_acl.c

@@ -85,7 +85,7 @@ xfs_acl_vhasacl_default(
 {
 	int		error;
 
-	if (vp->v_type != VDIR)
+	if (!VN_ISDIR(vp))
 		return 0;
 	xfs_acl_get_attr(vp, NULL, _ACL_TYPE_DEFAULT, ATTR_KERNOVAL, &error);
 	return (error == 0);
@@ -389,7 +389,7 @@ xfs_acl_allow_set(
 
 	if (vp->v_inode.i_flags & (S_IMMUTABLE|S_APPEND))
 		return EPERM;
-	if (kind == _ACL_TYPE_DEFAULT && vp->v_type != VDIR)
+	if (kind == _ACL_TYPE_DEFAULT && !VN_ISDIR(vp))
 		return ENOTDIR;
 	if (vp->v_vfsp->vfs_flag & VFS_RDONLY)
 		return EROFS;
@@ -750,7 +750,7 @@ xfs_acl_inherit(
 	 * If the new file is a directory, its default ACL is a copy of
 	 * the containing directory's default ACL.
 	 */
-	if (vp->v_type == VDIR)
+	if (VN_ISDIR(vp))
 		xfs_acl_set_attr(vp, pdaclp, _ACL_TYPE_DEFAULT, &error);
 	if (!error && !basicperms)
 		xfs_acl_set_attr(vp, cacl, _ACL_TYPE_ACCESS, &error);

fs/xfs/xfs_bmap.c

@@ -4754,10 +4754,20 @@ xfs_bmapi(
 					error = xfs_mod_incore_sb(mp,
 							XFS_SBS_FDBLOCKS,
 							-(alen), rsvd);
-				if (!error)
+				if (!error) {
 					error = xfs_mod_incore_sb(mp,
 							XFS_SBS_FDBLOCKS,
 							-(indlen), rsvd);
+					if (error && rt) {
+						xfs_mod_incore_sb(ip->i_mount,
+							XFS_SBS_FREXTENTS,
+							extsz, rsvd);
+					} else if (error) {
+						xfs_mod_incore_sb(ip->i_mount,
+							XFS_SBS_FDBLOCKS,
+							alen, rsvd);
+					}
+				}
 
 				if (error) {
 					if (XFS_IS_QUOTA_ON(ip->i_mount))

fs/xfs/xfs_buf_item.c

@@ -274,6 +274,7 @@ xfs_buf_item_format(
 		       ((bip->bli_format.blf_map_size - 1) * sizeof(uint)));
 	vecp->i_addr = (xfs_caddr_t)&bip->bli_format;
 	vecp->i_len = base_size;
+	XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_BFORMAT);
 	vecp++;
 	nvecs = 1;
 
@@ -320,12 +321,14 @@ xfs_buf_item_format(
 			buffer_offset = first_bit * XFS_BLI_CHUNK;
 			vecp->i_addr = xfs_buf_offset(bp, buffer_offset);
 			vecp->i_len = nbits * XFS_BLI_CHUNK;
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_BCHUNK);
 			nvecs++;
 			break;
 		} else if (next_bit != last_bit + 1) {
 			buffer_offset = first_bit * XFS_BLI_CHUNK;
 			vecp->i_addr = xfs_buf_offset(bp, buffer_offset);
 			vecp->i_len = nbits * XFS_BLI_CHUNK;
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_BCHUNK);
 			nvecs++;
 			vecp++;
 			first_bit = next_bit;
@@ -337,6 +340,7 @@ xfs_buf_item_format(
 			buffer_offset = first_bit * XFS_BLI_CHUNK;
 			vecp->i_addr = xfs_buf_offset(bp, buffer_offset);
 			vecp->i_len = nbits * XFS_BLI_CHUNK;
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_BCHUNK);
 /* You would think we need to bump the nvecs here too, but we do not
  * this number is used by recovery, and it gets confused by the boundary
  * split here

fs/xfs/xfs_dmapi.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2004 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as

fs/xfs/xfs_extfree_item.c

@@ -107,6 +107,7 @@ xfs_efi_item_format(xfs_efi_log_item_t	*efip,
 
 	log_vector->i_addr = (xfs_caddr_t)&(efip->efi_format);
 	log_vector->i_len = size;
+	XLOG_VEC_SET_TYPE(log_vector, XLOG_REG_TYPE_EFI_FORMAT);
 	ASSERT(size >= sizeof(xfs_efi_log_format_t));
 }
 
@@ -426,6 +427,7 @@ xfs_efd_item_format(xfs_efd_log_item_t	*efdp,
 
 	log_vector->i_addr = (xfs_caddr_t)&(efdp->efd_format);
 	log_vector->i_len = size;
+	XLOG_VEC_SET_TYPE(log_vector, XLOG_REG_TYPE_EFD_FORMAT);
 	ASSERT(size >= sizeof(xfs_efd_log_format_t));
 }
 

fs/xfs/xfs_iget.c

@@ -30,6 +30,8 @@
  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
  */
 
+#include <linux/delay.h>
+
 #include "xfs.h"
 
 #include "xfs_macros.h"
@@ -505,17 +507,15 @@ xfs_iget(
 	vnode_t		*vp = NULL;
 	int		error;
 
-retry:
 	XFS_STATS_INC(xs_ig_attempts);
 
+retry:
 	if ((inode = iget_locked(XFS_MTOVFS(mp)->vfs_super, ino))) {
 		bhv_desc_t	*bdp;
 		xfs_inode_t	*ip;
-		int		newnode;
 
 		vp = LINVFS_GET_VP(inode);
 		if (inode->i_state & I_NEW) {
-inode_allocate:
 			vn_initialize(inode);
 			error = xfs_iget_core(vp, mp, tp, ino, flags,
 					lock_flags, ipp, bno);
@@ -526,32 +526,25 @@ inode_allocate:
 				iput(inode);
 			}
 		} else {
-			/* These are true if the inode is in inactive or
-			 * reclaim. The linux inode is about to go away,
-			 * wait for that path to finish, and try again.
+			/*
+			 * If the inode is not fully constructed due to
+			 * filehandle mistmatches wait for the inode to go
+			 * away and try again.
+			 *
+			 * iget_locked will call __wait_on_freeing_inode
+			 * to wait for the inode to go away.
 			 */
-			if (vp->v_flag & (VINACT | VRECLM)) {
-				vn_wait(vp);
+			if (is_bad_inode(inode) ||
+			    ((bdp = vn_bhv_lookup(VN_BHV_HEAD(vp),
+						  &xfs_vnodeops)) == NULL)) {
 				iput(inode);
+				delay(1);
 				goto retry;
 			}
 
-			if (is_bad_inode(inode)) {
-				iput(inode);
-				return EIO;
-			}
-
-			bdp = vn_bhv_lookup(VN_BHV_HEAD(vp), &xfs_vnodeops);
-			if (bdp == NULL) {
-				XFS_STATS_INC(xs_ig_dup);
-				goto inode_allocate;
-			}
 			ip = XFS_BHVTOI(bdp);
 			if (lock_flags != 0)
 				xfs_ilock(ip, lock_flags);
-			newnode = (ip->i_d.di_mode == 0);
-			if (newnode)
-				xfs_iocore_inode_reinit(ip);
 			XFS_STATS_INC(xs_ig_found);
 			*ipp = ip;
 			error = 0;

fs/xfs/xfs_inode.c

@@ -1128,7 +1128,6 @@ xfs_ialloc(
 	ASSERT(ip != NULL);
 
 	vp = XFS_ITOV(ip);
-	vp->v_type = IFTOVT(mode);
 	ip->i_d.di_mode = (__uint16_t)mode;
 	ip->i_d.di_onlink = 0;
 	ip->i_d.di_nlink = nlink;
@@ -1250,7 +1249,7 @@ xfs_ialloc(
 	 */
 	xfs_trans_log_inode(tp, ip, flags);
 
-	/* now that we have a v_type we can set Linux inode ops (& unlock) */
+	/* now that we have an i_mode  we can set Linux inode ops (& unlock) */
 	VFS_INIT_VNODE(XFS_MTOVFS(tp->t_mountp), vp, XFS_ITOBHV(ip), 1);
 
 	*ipp = ip;

fs/xfs/xfs_inode_item.c

@@ -248,6 +248,7 @@ xfs_inode_item_format(
 
 	vecp->i_addr = (xfs_caddr_t)&iip->ili_format;
 	vecp->i_len  = sizeof(xfs_inode_log_format_t);
+	XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_IFORMAT);
 	vecp++;
 	nvecs	     = 1;
 
@@ -292,6 +293,7 @@ xfs_inode_item_format(
 
 	vecp->i_addr = (xfs_caddr_t)&ip->i_d;
 	vecp->i_len  = sizeof(xfs_dinode_core_t);
+	XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_ICORE);
 	vecp++;
 	nvecs++;
 	iip->ili_format.ilf_fields |= XFS_ILOG_CORE;
@@ -349,6 +351,7 @@ xfs_inode_item_format(
 				vecp->i_addr =
 					(char *)(ip->i_df.if_u1.if_extents);
 				vecp->i_len = ip->i_df.if_bytes;
+				XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_IEXT);
 			} else
 #endif
 			{
@@ -367,6 +370,7 @@ xfs_inode_item_format(
 				vecp->i_addr = (xfs_caddr_t)ext_buffer;
 				vecp->i_len = xfs_iextents_copy(ip, ext_buffer,
 						XFS_DATA_FORK);
+				XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_IEXT);
 			}
 			ASSERT(vecp->i_len <= ip->i_df.if_bytes);
 			iip->ili_format.ilf_dsize = vecp->i_len;
@@ -384,6 +388,7 @@ xfs_inode_item_format(
 			ASSERT(ip->i_df.if_broot != NULL);
 			vecp->i_addr = (xfs_caddr_t)ip->i_df.if_broot;
 			vecp->i_len = ip->i_df.if_broot_bytes;
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_IBROOT);
 			vecp++;
 			nvecs++;
 			iip->ili_format.ilf_dsize = ip->i_df.if_broot_bytes;
@@ -409,6 +414,7 @@ xfs_inode_item_format(
 			ASSERT((ip->i_df.if_real_bytes == 0) ||
 			       (ip->i_df.if_real_bytes == data_bytes));
 			vecp->i_len = (int)data_bytes;
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_ILOCAL);
 			vecp++;
 			nvecs++;
 			iip->ili_format.ilf_dsize = (unsigned)data_bytes;
@@ -486,6 +492,7 @@ xfs_inode_item_format(
 			vecp->i_len = xfs_iextents_copy(ip, ext_buffer,
 					XFS_ATTR_FORK);
 #endif
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_IATTR_EXT);
 			iip->ili_format.ilf_asize = vecp->i_len;
 			vecp++;
 			nvecs++;
@@ -500,6 +507,7 @@ xfs_inode_item_format(
 			ASSERT(ip->i_afp->if_broot != NULL);
 			vecp->i_addr = (xfs_caddr_t)ip->i_afp->if_broot;
 			vecp->i_len = ip->i_afp->if_broot_bytes;
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_IATTR_BROOT);
 			vecp++;
 			nvecs++;
 			iip->ili_format.ilf_asize = ip->i_afp->if_broot_bytes;
@@ -523,6 +531,7 @@ xfs_inode_item_format(
 			ASSERT((ip->i_afp->if_real_bytes == 0) ||
 			       (ip->i_afp->if_real_bytes == data_bytes));
 			vecp->i_len = (int)data_bytes;
+			XLOG_VEC_SET_TYPE(vecp, XLOG_REG_TYPE_IATTR_LOCAL);
 			vecp++;
 			nvecs++;
 			iip->ili_format.ilf_asize = (unsigned)data_bytes;

fs/xfs/xfs_iomap.c

@@ -226,13 +226,12 @@ xfs_iomap(
 		xfs_iomap_enter_trace(XFS_IOMAP_READ_ENTER, io, offset, count);
 		lockmode = XFS_LCK_MAP_SHARED(mp, io);
 		bmapi_flags = XFS_BMAPI_ENTIRE;
-		if (flags & BMAPI_IGNSTATE)
-			bmapi_flags |= XFS_BMAPI_IGSTATE;
 		break;
 	case BMAPI_WRITE:
 		xfs_iomap_enter_trace(XFS_IOMAP_WRITE_ENTER, io, offset, count);
 		lockmode = XFS_ILOCK_EXCL|XFS_EXTSIZE_WR;
-		bmapi_flags = 0;
+		if (flags & BMAPI_IGNSTATE)
+			bmapi_flags |= XFS_BMAPI_IGSTATE|XFS_BMAPI_ENTIRE;
 		XFS_ILOCK(mp, io, lockmode);
 		break;
 	case BMAPI_ALLOCATE:
@@ -391,9 +390,9 @@ xfs_iomap_write_direct(
 	xfs_bmbt_irec_t imap[XFS_WRITE_IMAPS], *imapp;
 	xfs_bmap_free_t free_list;
 	int		aeof;
-	xfs_filblks_t	datablocks, qblocks, resblks;
+	xfs_filblks_t	qblocks, resblks;
 	int		committed;
-	int		numrtextents;
+	int		resrtextents;
 
 	/*
 	 * Make sure that the dquots are there. This doesn't hold
@@ -434,14 +433,14 @@ xfs_iomap_write_direct(
 
 		if (!(extsz = ip->i_d.di_extsize))
 			extsz = mp->m_sb.sb_rextsize;
-		numrtextents = qblocks = (count_fsb + extsz - 1);
-		do_div(numrtextents, mp->m_sb.sb_rextsize);
+		resrtextents = qblocks = (count_fsb + extsz - 1);
+		do_div(resrtextents, mp->m_sb.sb_rextsize);
+		resblks = XFS_DIOSTRAT_SPACE_RES(mp, 0);
 		quota_flag = XFS_QMOPT_RES_RTBLKS;
-		datablocks = 0;
 	} else {
-		datablocks = qblocks = count_fsb;
+		resrtextents = 0;
+		resblks = qblocks = XFS_DIOSTRAT_SPACE_RES(mp, count_fsb);
 		quota_flag = XFS_QMOPT_RES_REGBLKS;
-		numrtextents = 0;
 	}
 
 	/*
@@ -449,9 +448,8 @@ xfs_iomap_write_direct(
 	 */
 	xfs_iunlock(ip, XFS_ILOCK_EXCL);
 	tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT);
-	resblks = XFS_DIOSTRAT_SPACE_RES(mp, datablocks);
 	error = xfs_trans_reserve(tp, resblks,
-			XFS_WRITE_LOG_RES(mp), numrtextents,
+			XFS_WRITE_LOG_RES(mp), resrtextents,
 			XFS_TRANS_PERM_LOG_RES,
 			XFS_WRITE_LOG_COUNT);
 

fs/xfs/xfs_log.c

@@ -159,11 +159,15 @@ xfs_buftarg_t *xlog_target;
 void
 xlog_trace_loggrant(xlog_t *log, xlog_ticket_t *tic, xfs_caddr_t string)
 {
-	if (! log->l_grant_trace) {
-		log->l_grant_trace = ktrace_alloc(1024, KM_NOSLEEP);
-		if (! log->l_grant_trace)
+	unsigned long cnts;
+
+	if (!log->l_grant_trace) {
+		log->l_grant_trace = ktrace_alloc(2048, KM_NOSLEEP);
+		if (!log->l_grant_trace)
 			return;
 	}
+	/* ticket counts are 1 byte each */
+	cnts = ((unsigned long)tic->t_ocnt) | ((unsigned long)tic->t_cnt) << 8;
 
 	ktrace_enter(log->l_grant_trace,
 		     (void *)tic,
@@ -178,10 +182,10 @@ xlog_trace_loggrant(xlog_t *log, xlog_ticket_t *tic, xfs_caddr_t string)
 		     (void *)((unsigned long)CYCLE_LSN(log->l_tail_lsn)),
 		     (void *)((unsigned long)BLOCK_LSN(log->l_tail_lsn)),
 		     (void *)string,
-		     (void *)((unsigned long)13),
-		     (void *)((unsigned long)14),
-		     (void *)((unsigned long)15),
-		     (void *)((unsigned long)16));
+		     (void *)((unsigned long)tic->t_trans_type),
+		     (void *)cnts,
+		     (void *)((unsigned long)tic->t_curr_res),
+		     (void *)((unsigned long)tic->t_unit_res));
 }
 
 void
@@ -274,9 +278,11 @@ xfs_log_done(xfs_mount_t	*mp,
 		 * Release ticket if not permanent reservation or a specifc
 		 * request has been made to release a permanent reservation.
 		 */
+		xlog_trace_loggrant(log, ticket, "xfs_log_done: (non-permanent)");
 		xlog_ungrant_log_space(log, ticket);
 		xlog_state_put_ticket(log, ticket);
 	} else {
+		xlog_trace_loggrant(log, ticket, "xfs_log_done: (permanent)");
 		xlog_regrant_reserve_log_space(log, ticket);
 	}
 
@@ -399,7 +405,8 @@ xfs_log_reserve(xfs_mount_t	 *mp,
 		int		 cnt,
 		xfs_log_ticket_t *ticket,
 		__uint8_t	 client,
-		uint		 flags)
+		uint		 flags,
+		uint		 t_type)
 {
 	xlog_t		*log = mp->m_log;
 	xlog_ticket_t	*internal_ticket;
@@ -421,13 +428,19 @@ xfs_log_reserve(xfs_mount_t	 *mp,
 	if (*ticket != NULL) {
 		ASSERT(flags & XFS_LOG_PERM_RESERV);
 		internal_ticket = (xlog_ticket_t *)*ticket;
+		xlog_trace_loggrant(log, internal_ticket, "xfs_log_reserve: existing ticket (permanent trans)");
 		xlog_grant_push_ail(mp, internal_ticket->t_unit_res);
 		retval = xlog_regrant_write_log_space(log, internal_ticket);
 	} else {
 		/* may sleep if need to allocate more tickets */
 		internal_ticket = xlog_ticket_get(log, unit_bytes, cnt,
 						  client, flags);
+		internal_ticket->t_trans_type = t_type;
 		*ticket = internal_ticket;
+		xlog_trace_loggrant(log, internal_ticket, 
+			(internal_ticket->t_flags & XLOG_TIC_PERM_RESERV) ?
+			"xfs_log_reserve: create new ticket (permanent trans)" :
+			"xfs_log_reserve: create new ticket");
 		xlog_grant_push_ail(mp,
 				    (internal_ticket->t_unit_res *
 				     internal_ticket->t_cnt));
@@ -601,8 +614,9 @@ xfs_log_unmount_write(xfs_mount_t *mp)
 	if (! (XLOG_FORCED_SHUTDOWN(log))) {
 		reg[0].i_addr = (void*)&magic;
 		reg[0].i_len  = sizeof(magic);
+		XLOG_VEC_SET_TYPE(&reg[0], XLOG_REG_TYPE_UNMOUNT);
 
-		error = xfs_log_reserve(mp, 600, 1, &tic, XFS_LOG, 0);
+		error = xfs_log_reserve(mp, 600, 1, &tic, XFS_LOG, 0, 0);
 		if (!error) {
 			/* remove inited flag */
 			((xlog_ticket_t *)tic)->t_flags = 0;
@@ -1272,6 +1286,7 @@ xlog_commit_record(xfs_mount_t  *mp,
 
 	reg[0].i_addr = NULL;
 	reg[0].i_len = 0;
+	XLOG_VEC_SET_TYPE(&reg[0], XLOG_REG_TYPE_COMMIT);
 
 	ASSERT_ALWAYS(iclog);
 	if ((error = xlog_write(mp, reg, 1, ticket, commitlsnp,
@@ -1604,6 +1619,117 @@ xlog_state_finish_copy(xlog_t		*log,
 
 
 
+/*
+ * print out info relating to regions written which consume
+ * the reservation
+ */
+#if defined(XFS_LOG_RES_DEBUG)
+STATIC void
+xlog_print_tic_res(xfs_mount_t *mp, xlog_ticket_t *ticket)
+{
+	uint i;
+	uint ophdr_spc = ticket->t_res_num_ophdrs * (uint)sizeof(xlog_op_header_t);
+
+	/* match with XLOG_REG_TYPE_* in xfs_log.h */
+	static char *res_type_str[XLOG_REG_TYPE_MAX] = {
+	    "bformat",
+	    "bchunk",
+	    "efi_format",
+	    "efd_format",
+	    "iformat",
+	    "icore",
+	    "iext",
+	    "ibroot",
+	    "ilocal",
+	    "iattr_ext",
+	    "iattr_broot",
+	    "iattr_local",
+	    "qformat",
+	    "dquot",
+	    "quotaoff",
+	    "LR header",
+	    "unmount",
+	    "commit",
+	    "trans header"
+	};
+	static char *trans_type_str[XFS_TRANS_TYPE_MAX] = {
+	    "SETATTR_NOT_SIZE",
+	    "SETATTR_SIZE",
+	    "INACTIVE",
+	    "CREATE",
+	    "CREATE_TRUNC",
+	    "TRUNCATE_FILE",
+	    "REMOVE",
+	    "LINK",
+	    "RENAME",
+	    "MKDIR",
+	    "RMDIR",
+	    "SYMLINK",
+	    "SET_DMATTRS",
+	    "GROWFS",
+	    "STRAT_WRITE",
+	    "DIOSTRAT",
+	    "WRITE_SYNC",
+	    "WRITEID",
+	    "ADDAFORK",
+	    "ATTRINVAL",
+	    "ATRUNCATE",
+	    "ATTR_SET",
+	    "ATTR_RM",
+	    "ATTR_FLAG",
+	    "CLEAR_AGI_BUCKET",
+	    "QM_SBCHANGE",
+	    "DUMMY1",
+	    "DUMMY2",
+	    "QM_QUOTAOFF",
+	    "QM_DQALLOC",
+	    "QM_SETQLIM",
+	    "QM_DQCLUSTER",
+	    "QM_QINOCREATE",
+	    "QM_QUOTAOFF_END",
+	    "SB_UNIT",
+	    "FSYNC_TS",
+	    "GROWFSRT_ALLOC",
+	    "GROWFSRT_ZERO",
+	    "GROWFSRT_FREE",
+	    "SWAPEXT"
+	};
+
+	xfs_fs_cmn_err(CE_WARN, mp,
+			"xfs_log_write: reservation summary:\n"
+			"  trans type  = %s (%u)\n"
+			"  unit res    = %d bytes\n"
+			"  current res = %d bytes\n"
+			"  total reg   = %u bytes (o/flow = %u bytes)\n"
+			"  ophdrs      = %u (ophdr space = %u bytes)\n"
+			"  ophdr + reg = %u bytes\n"
+			"  num regions = %u\n",
+			((ticket->t_trans_type <= 0 ||
+			  ticket->t_trans_type > XFS_TRANS_TYPE_MAX) ?
+			  "bad-trans-type" : trans_type_str[ticket->t_trans_type-1]),
+			ticket->t_trans_type,
+			ticket->t_unit_res,
+			ticket->t_curr_res,
+			ticket->t_res_arr_sum, ticket->t_res_o_flow,
+			ticket->t_res_num_ophdrs, ophdr_spc,
+			ticket->t_res_arr_sum + 
+			  ticket->t_res_o_flow + ophdr_spc,
+			ticket->t_res_num);
+
+	for (i = 0; i < ticket->t_res_num; i++) {
+	   	uint r_type = ticket->t_res_arr[i].r_type; 
+		cmn_err(CE_WARN,
+			    "region[%u]: %s - %u bytes\n",
+			    i, 
+			    ((r_type <= 0 || r_type > XLOG_REG_TYPE_MAX) ?
+			    "bad-rtype" : res_type_str[r_type-1]),
+			    ticket->t_res_arr[i].r_len);
+	}
+}
+#else
+#define xlog_print_tic_res(mp, ticket)
+#endif
+
 /*
  * Write some region out to in-core log
  *
@@ -1677,16 +1803,21 @@ xlog_write(xfs_mount_t *	mp,
      * xlog_op_header_t and may need to be double word aligned.
      */
     len = 0;
-    if (ticket->t_flags & XLOG_TIC_INITED)     /* acct for start rec of xact */
+    if (ticket->t_flags & XLOG_TIC_INITED) {    /* acct for start rec of xact */
 	len += sizeof(xlog_op_header_t);
+	XLOG_TIC_ADD_OPHDR(ticket);
+    }
 
     for (index = 0; index < nentries; index++) {
 	len += sizeof(xlog_op_header_t);	    /* each region gets >= 1 */
+	XLOG_TIC_ADD_OPHDR(ticket);
 	len += reg[index].i_len;
+	XLOG_TIC_ADD_REGION(ticket, reg[index].i_len, reg[index].i_type);
     }
     contwr = *start_lsn = 0;
 
     if (ticket->t_curr_res < len) {
+	xlog_print_tic_res(mp, ticket);
 #ifdef DEBUG
 	xlog_panic(
 		"xfs_log_write: reservation ran out. Need to up reservation");
@@ -1790,6 +1921,7 @@ xlog_write(xfs_mount_t *	mp,
 		len += sizeof(xlog_op_header_t); /* from splitting of region */
 		/* account for new log op header */
 		ticket->t_curr_res -= sizeof(xlog_op_header_t);
+		XLOG_TIC_ADD_OPHDR(ticket);
 	    }
 	    xlog_verify_dest_ptr(log, ptr);
 
@@ -2282,6 +2414,9 @@ restart:
 	 */
 	if (log_offset == 0) {
 		ticket->t_curr_res -= log->l_iclog_hsize;
+		XLOG_TIC_ADD_REGION(ticket,
+				    log->l_iclog_hsize,
+				    XLOG_REG_TYPE_LRHEADER);
 		INT_SET(head->h_cycle, ARCH_CONVERT, log->l_curr_cycle);
 		ASSIGN_LSN(head->h_lsn, log);
 		ASSERT(log->l_curr_block >= 0);
@@ -2468,6 +2603,7 @@ xlog_regrant_write_log_space(xlog_t	   *log,
 #endif
 
 	tic->t_curr_res = tic->t_unit_res;
+	XLOG_TIC_RESET_RES(tic);
 
 	if (tic->t_cnt > 0)
 		return (0);
@@ -2608,6 +2744,7 @@ xlog_regrant_reserve_log_space(xlog_t	     *log,
 	XLOG_GRANT_SUB_SPACE(log, ticket->t_curr_res, 'w');
 	XLOG_GRANT_SUB_SPACE(log, ticket->t_curr_res, 'r');
 	ticket->t_curr_res = ticket->t_unit_res;
+	XLOG_TIC_RESET_RES(ticket);
 	xlog_trace_loggrant(log, ticket,
 			    "xlog_regrant_reserve_log_space: sub current res");
 	xlog_verify_grant_head(log, 1);
@@ -2624,6 +2761,7 @@ xlog_regrant_reserve_log_space(xlog_t	     *log,
 	xlog_verify_grant_head(log, 0);
 	GRANT_UNLOCK(log, s);
 	ticket->t_curr_res = ticket->t_unit_res;
+	XLOG_TIC_RESET_RES(ticket);
 }	/* xlog_regrant_reserve_log_space */
 
 
@@ -3179,29 +3317,57 @@ xlog_ticket_get(xlog_t		*log,
 	 * and their unit amount is the total amount of space required.
 	 *
 	 * The following lines of code account for non-transaction data
-	 * which occupy space in the on-disk log. 
+	 * which occupy space in the on-disk log.
+	 *
+	 * Normal form of a transaction is:
+	 * <oph><trans-hdr><start-oph><reg1-oph><reg1><reg2-oph>...<commit-oph>
+	 * and then there are LR hdrs, split-recs and roundoff at end of syncs.
+	 *
+	 * We need to account for all the leadup data and trailer data
+	 * around the transaction data.
+	 * And then we need to account for the worst case in terms of using
+	 * more space.
+	 * The worst case will happen if:
+	 * - the placement of the transaction happens to be such that the
+	 *   roundoff is at its maximum
+	 * - the transaction data is synced before the commit record is synced
+	 *   i.e. <transaction-data><roundoff> | <commit-rec><roundoff>
+	 *   Therefore the commit record is in its own Log Record.
+	 *   This can happen as the commit record is called with its
+	 *   own region to xlog_write().
+	 *   This then means that in the worst case, roundoff can happen for
+	 *   the commit-rec as well.
+	 *   The commit-rec is smaller than padding in this scenario and so it is
+	 *   not added separately.
 	 */
 
+	/* for trans header */
+	unit_bytes += sizeof(xlog_op_header_t);
+	unit_bytes += sizeof(xfs_trans_header_t);
+
 	/* for start-rec */
-	unit_bytes += sizeof(xlog_op_header_t); 
+	unit_bytes += sizeof(xlog_op_header_t);
+
+	/* for LR headers */
+	num_headers = ((unit_bytes + log->l_iclog_size-1) >> log->l_iclog_size_log);
+	unit_bytes += log->l_iclog_hsize * num_headers;
+
+	/* for commit-rec LR header - note: padding will subsume the ophdr */
+	unit_bytes += log->l_iclog_hsize;
+
+	/* for split-recs - ophdrs added when data split over LRs */
+	unit_bytes += sizeof(xlog_op_header_t) * num_headers;
 
-	/* for padding */
+	/* for roundoff padding for transaction data and one for commit record */
 	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) &&
-		log->l_mp->m_sb.sb_logsunit > 1) {
+	    log->l_mp->m_sb.sb_logsunit > 1) {
 		/* log su roundoff */
-		unit_bytes += log->l_mp->m_sb.sb_logsunit;  
+		unit_bytes += 2*log->l_mp->m_sb.sb_logsunit;
 	} else {
 		/* BB roundoff */
-		unit_bytes += BBSIZE;
+		unit_bytes += 2*BBSIZE;
         }
 
-	/* for commit-rec */
-	unit_bytes += sizeof(xlog_op_header_t);
- 
-	/* for LR headers */
-	num_headers = ((unit_bytes + log->l_iclog_size-1) >> log->l_iclog_size_log);
-	unit_bytes += log->l_iclog_hsize * num_headers;
-
 	tic->t_unit_res		= unit_bytes;
 	tic->t_curr_res		= unit_bytes;
 	tic->t_cnt		= cnt;
@@ -3209,10 +3375,13 @@ xlog_ticket_get(xlog_t		*log,
 	tic->t_tid		= (xlog_tid_t)((__psint_t)tic & 0xffffffff);
 	tic->t_clientid		= client;
 	tic->t_flags		= XLOG_TIC_INITED;
+	tic->t_trans_type	= 0;
 	if (xflags & XFS_LOG_PERM_RESERV)
 		tic->t_flags |= XLOG_TIC_PERM_RESERV;
 	sv_init(&(tic->t_sema), SV_DEFAULT, "logtick");
 
+	XLOG_TIC_RESET_RES(tic);
+
 	return tic;
 }	/* xlog_ticket_get */
 

fs/xfs/xfs_log.h

@@ -114,9 +114,44 @@ xfs_lsn_t	_lsn_cmp(xfs_lsn_t lsn1, xfs_lsn_t lsn2)
 #define XFS_VOLUME		0x2
 #define XFS_LOG			0xaa
 
+
+/* Region types for iovec's i_type */
+#if defined(XFS_LOG_RES_DEBUG)
+#define XLOG_REG_TYPE_BFORMAT		1
+#define XLOG_REG_TYPE_BCHUNK		2
+#define XLOG_REG_TYPE_EFI_FORMAT	3
+#define XLOG_REG_TYPE_EFD_FORMAT	4
+#define XLOG_REG_TYPE_IFORMAT		5
+#define XLOG_REG_TYPE_ICORE		6
+#define XLOG_REG_TYPE_IEXT		7
+#define XLOG_REG_TYPE_IBROOT		8
+#define XLOG_REG_TYPE_ILOCAL		9
+#define XLOG_REG_TYPE_IATTR_EXT		10
+#define XLOG_REG_TYPE_IATTR_BROOT	11
+#define XLOG_REG_TYPE_IATTR_LOCAL	12
+#define XLOG_REG_TYPE_QFORMAT		13
+#define XLOG_REG_TYPE_DQUOT		14
+#define XLOG_REG_TYPE_QUOTAOFF		15
+#define XLOG_REG_TYPE_LRHEADER		16
+#define XLOG_REG_TYPE_UNMOUNT		17
+#define XLOG_REG_TYPE_COMMIT		18
+#define XLOG_REG_TYPE_TRANSHDR		19
+#define XLOG_REG_TYPE_MAX		19
+#endif
+
+#if defined(XFS_LOG_RES_DEBUG)
+#define XLOG_VEC_SET_TYPE(vecp, t) ((vecp)->i_type = (t))
+#else
+#define XLOG_VEC_SET_TYPE(vecp, t)
+#endif
+
+
 typedef struct xfs_log_iovec {
 	xfs_caddr_t		i_addr;		/* beginning address of region */
 	int		i_len;		/* length in bytes of region */
+#if defined(XFS_LOG_RES_DEBUG)
+ 	uint		i_type;		/* type of region */
+#endif
 } xfs_log_iovec_t;
 
 typedef void* xfs_log_ticket_t;
@@ -159,7 +194,8 @@ int	  xfs_log_reserve(struct xfs_mount *mp,
 			  int		   count,
 			  xfs_log_ticket_t *ticket,
 			  __uint8_t	   clientid,
-			  uint		   flags);
+			  uint		   flags,
+			  uint		   t_type);
 int	  xfs_log_write(struct xfs_mount *mp,
 			xfs_log_iovec_t  region[],
 			int		 nentries,

fs/xfs/xfs_log_priv.h

@@ -335,18 +335,66 @@ typedef __uint32_t xlog_tid_t;
 
 #define XLOG_COVER_OPS		5
 
+
+/* Ticket reservation region accounting */ 
+#if defined(XFS_LOG_RES_DEBUG)
+#define XLOG_TIC_LEN_MAX	15
+#define XLOG_TIC_RESET_RES(t) ((t)->t_res_num = \
+				(t)->t_res_arr_sum = (t)->t_res_num_ophdrs = 0)
+#define XLOG_TIC_ADD_OPHDR(t) ((t)->t_res_num_ophdrs++)
+#define XLOG_TIC_ADD_REGION(t, len, type)				\
+	do {								\
+		if ((t)->t_res_num == XLOG_TIC_LEN_MAX) { 		\
+			/* add to overflow and start again */		\
+			(t)->t_res_o_flow += (t)->t_res_arr_sum;	\
+			(t)->t_res_num = 0;				\
+			(t)->t_res_arr_sum = 0;				\
+		}							\
+		(t)->t_res_arr[(t)->t_res_num].r_len = (len);		\
+		(t)->t_res_arr[(t)->t_res_num].r_type = (type);		\
+		(t)->t_res_arr_sum += (len);				\
+		(t)->t_res_num++;					\
+	} while (0)
+
+/*
+ * Reservation region
+ * As would be stored in xfs_log_iovec but without the i_addr which
+ * we don't care about.
+ */
+typedef struct xlog_res {
+	uint	r_len;
+	uint	r_type;
+} xlog_res_t;
+#else
+#define XLOG_TIC_RESET_RES(t)
+#define XLOG_TIC_ADD_OPHDR(t)
+#define XLOG_TIC_ADD_REGION(t, len, type)
+#endif
+
+
 typedef struct xlog_ticket {
-	sv_t		   t_sema;	 /* sleep on this semaphore	 :20 */
-	struct xlog_ticket *t_next;	 /*			         : 4 */
-	struct xlog_ticket *t_prev;	 /*				 : 4 */
-	xlog_tid_t	   t_tid;	 /* transaction identifier	 : 4 */
-	int		   t_curr_res;	 /* current reservation in bytes : 4 */
-	int		   t_unit_res;	 /* unit reservation in bytes    : 4 */
-	__uint8_t	   t_ocnt;	 /* original count		 : 1 */
-	__uint8_t	   t_cnt;	 /* current count		 : 1 */
-	__uint8_t	   t_clientid;	 /* who does this belong to;	 : 1 */
-	__uint8_t	   t_flags;	 /* properties of reservation	 : 1 */
+	sv_t		   t_sema;	 /* sleep on this semaphore      : 20 */
+ 	struct xlog_ticket *t_next;	 /*			         :4|8 */
+	struct xlog_ticket *t_prev;	 /*				 :4|8 */
+	xlog_tid_t	   t_tid;	 /* transaction identifier	 : 4  */
+	int		   t_curr_res;	 /* current reservation in bytes : 4  */
+	int		   t_unit_res;	 /* unit reservation in bytes    : 4  */
+	char		   t_ocnt;	 /* original count		 : 1  */
+	char		   t_cnt;	 /* current count		 : 1  */
+	char		   t_clientid;	 /* who does this belong to;	 : 1  */
+	char		   t_flags;	 /* properties of reservation	 : 1  */
+	uint		   t_trans_type; /* transaction type             : 4  */
+
+#if defined (XFS_LOG_RES_DEBUG)
+        /* reservation array fields */
+	uint		   t_res_num;                    /* num in array : 4 */
+	xlog_res_t	   t_res_arr[XLOG_TIC_LEN_MAX];  /* array of res : X */ 
+	uint		   t_res_num_ophdrs;		 /* num op hdrs  : 4 */
+	uint		   t_res_arr_sum;		 /* array sum    : 4 */
+	uint		   t_res_o_flow;		 /* sum overflow : 4 */
+#endif
 } xlog_ticket_t;
+
 #endif
 
 

fs/xfs/xfs_log_recover.c

@@ -1387,7 +1387,7 @@ xlog_recover_add_to_cont_trans(
 	old_ptr = item->ri_buf[item->ri_cnt-1].i_addr;
 	old_len = item->ri_buf[item->ri_cnt-1].i_len;
 
-	ptr = kmem_realloc(old_ptr, len+old_len, old_len, 0);
+	ptr = kmem_realloc(old_ptr, len+old_len, old_len, 0u);
 	memcpy(&ptr[old_len], dp, len); /* d, s, l */
 	item->ri_buf[item->ri_cnt-1].i_len += len;
 	item->ri_buf[item->ri_cnt-1].i_addr = ptr;

fs/xfs/xfs_qmops.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -42,7 +42,8 @@
 #include "xfs_dir2.h"
 #include "xfs_dmapi.h"
 #include "xfs_mount.h"
-
+#include "xfs_quota.h"
+#include "xfs_error.h"
 
 STATIC struct xfs_dquot *
 xfs_dqvopchown_default(
@@ -54,8 +55,79 @@ xfs_dqvopchown_default(
 	return NULL;
 }
 
+/*
+ * Clear the quotaflags in memory and in the superblock.
+ */
+int
+xfs_mount_reset_sbqflags(xfs_mount_t *mp)
+{
+	int			error;
+	xfs_trans_t		*tp;
+	unsigned long		s;
+
+	mp->m_qflags = 0;
+	/*
+	 * It is OK to look at sb_qflags here in mount path,
+	 * without SB_LOCK.
+	 */
+	if (mp->m_sb.sb_qflags == 0)
+		return 0;
+	s = XFS_SB_LOCK(mp);
+	mp->m_sb.sb_qflags = 0;
+	XFS_SB_UNLOCK(mp, s);
+
+	/*
+	 * if the fs is readonly, let the incore superblock run
+	 * with quotas off but don't flush the update out to disk
+	 */
+	if (XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY)
+		return 0;
+#ifdef QUOTADEBUG
+	xfs_fs_cmn_err(CE_NOTE, mp, "Writing superblock quota changes");
+#endif
+	tp = xfs_trans_alloc(mp, XFS_TRANS_QM_SBCHANGE);
+	if ((error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
+				      XFS_DEFAULT_LOG_COUNT))) {
+		xfs_trans_cancel(tp, 0);
+		xfs_fs_cmn_err(CE_ALERT, mp,
+			"xfs_mount_reset_sbqflags: Superblock update failed!");
+		return error;
+	}
+	xfs_mod_sb(tp, XFS_SB_QFLAGS);
+	error = xfs_trans_commit(tp, 0, NULL);
+	return error;
+}
+
+STATIC int
+xfs_noquota_init(
+	xfs_mount_t	*mp,
+	uint		*needquotamount,
+	uint		*quotaflags)
+{
+	int		error = 0;
+
+	*quotaflags = 0;
+	*needquotamount = B_FALSE;
+
+	ASSERT(!XFS_IS_QUOTA_ON(mp));
+
+	/*
+	 * If a file system had quotas running earlier, but decided to
+	 * mount without -o uquota/pquota/gquota options, revoke the
+	 * quotachecked license.
+	 */
+	if (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_ACCT) {
+		cmn_err(CE_NOTE,
+                        "XFS resetting qflags for filesystem %s",
+                        mp->m_fsname);
+
+		error = xfs_mount_reset_sbqflags(mp);
+	}
+	return error;
+}
+
 xfs_qmops_t	xfs_qmcore_stub = {
-	.xfs_qminit		= (xfs_qminit_t) fs_noerr,
+	.xfs_qminit		= (xfs_qminit_t) xfs_noquota_init,
 	.xfs_qmdone		= (xfs_qmdone_t) fs_noerr,
 	.xfs_qmmount		= (xfs_qmmount_t) fs_noerr,
 	.xfs_qmunmount		= (xfs_qmunmount_t) fs_noerr,

fs/xfs/xfs_quota.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2 of the GNU General Public License as
@@ -159,6 +159,20 @@ typedef struct xfs_qoff_logformat {
 #define XFS_OQUOTA_CHKD	0x0020  /* quotacheck run on other (grp/prj) quotas */
 #define XFS_GQUOTA_ACCT	0x0040  /* group quota accounting ON */
 
+/*
+ * Quota Accounting/Enforcement flags
+ */
+#define XFS_ALL_QUOTA_ACCT	\
+		(XFS_UQUOTA_ACCT | XFS_GQUOTA_ACCT | XFS_PQUOTA_ACCT)
+#define XFS_ALL_QUOTA_ENFD	(XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD)
+#define XFS_ALL_QUOTA_CHKD	(XFS_UQUOTA_CHKD | XFS_OQUOTA_CHKD)
+
+#define XFS_IS_QUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_ALL_QUOTA_ACCT)
+#define XFS_IS_QUOTA_ENFORCED(mp)	((mp)->m_qflags & XFS_ALL_QUOTA_ENFD)
+#define XFS_IS_UQUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_UQUOTA_ACCT)
+#define XFS_IS_PQUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_PQUOTA_ACCT)
+#define XFS_IS_GQUOTA_RUNNING(mp)	((mp)->m_qflags & XFS_GQUOTA_ACCT)
+
 /*
  * Incore only flags for quotaoff - these bits get cleared when quota(s)
  * are in the process of getting turned off. These flags are in m_qflags but
@@ -362,6 +376,7 @@ typedef struct xfs_dqtrxops {
 				f | XFS_QMOPT_RES_REGBLKS)
 
 extern int xfs_qm_dqcheck(xfs_disk_dquot_t *, xfs_dqid_t, uint, uint, char *);
+extern int xfs_mount_reset_sbqflags(struct xfs_mount *);
 
 extern struct bhv_vfsops xfs_qmops;
 

fs/xfs/xfs_trans.c

@@ -276,7 +276,7 @@ xfs_trans_reserve(
 
 		error = xfs_log_reserve(tp->t_mountp, logspace, logcount,
 					&tp->t_ticket,
-					XFS_TRANSACTION, log_flags);
+					XFS_TRANSACTION, log_flags, tp->t_type);
 		if (error) {
 			goto undo_blocks;
 		}
@@ -1032,6 +1032,7 @@ xfs_trans_fill_vecs(
 	tp->t_header.th_num_items = nitems;
 	log_vector->i_addr = (xfs_caddr_t)&tp->t_header;
 	log_vector->i_len = sizeof(xfs_trans_header_t);
+	XLOG_VEC_SET_TYPE(log_vector, XLOG_REG_TYPE_TRANSHDR);
 }
 
 

fs/xfs/xfs_trans.h

@@ -112,6 +112,7 @@ typedef struct xfs_trans_header {
 #define	XFS_TRANS_GROWFSRT_ZERO		38
 #define	XFS_TRANS_GROWFSRT_FREE		39
 #define	XFS_TRANS_SWAPEXT		40
+#define	XFS_TRANS_TYPE_MAX		40
 /* new transaction types need to be reflected in xfs_logprint(8) */
 
 
@@ -998,6 +999,7 @@ struct xfs_buf	*xfs_trans_getsb(xfs_trans_t *, struct xfs_mount *, int);
 void		xfs_trans_brelse(xfs_trans_t *, struct xfs_buf *);
 void		xfs_trans_bjoin(xfs_trans_t *, struct xfs_buf *);
 void		xfs_trans_bhold(xfs_trans_t *, struct xfs_buf *);
+void		xfs_trans_bhold_release(xfs_trans_t *, struct xfs_buf *);
 void		xfs_trans_binval(xfs_trans_t *, struct xfs_buf *);
 void		xfs_trans_inode_buf(xfs_trans_t *, struct xfs_buf *);
 void		xfs_trans_inode_buf(xfs_trans_t *, struct xfs_buf *);

fs/xfs/xfs_trans_ail.c

@@ -379,8 +379,8 @@ xfs_trans_delete_ail(
 		else {
 			xfs_cmn_err(XFS_PTAG_AILDELETE, CE_ALERT, mp,
 				"xfs_trans_delete_ail: attempting to delete a log item that is not in the AIL");
-			xfs_force_shutdown(mp, XFS_CORRUPT_INCORE);
 			AIL_UNLOCK(mp, s);
+			xfs_force_shutdown(mp, XFS_CORRUPT_INCORE);
 		}
 	}
 }

fs/xfs/xfs_trans_buf.c

@@ -713,6 +713,29 @@ xfs_trans_bhold(xfs_trans_t	*tp,
 	xfs_buf_item_trace("BHOLD", bip);
 }
 
+/*
+ * Cancel the previous buffer hold request made on this buffer
+ * for this transaction.
+ */
+void
+xfs_trans_bhold_release(xfs_trans_t	*tp,
+			xfs_buf_t	*bp)
+{
+	xfs_buf_log_item_t	*bip;
+
+	ASSERT(XFS_BUF_ISBUSY(bp));
+	ASSERT(XFS_BUF_FSPRIVATE2(bp, xfs_trans_t *) == tp);
+	ASSERT(XFS_BUF_FSPRIVATE(bp, void *) != NULL);
+
+	bip = XFS_BUF_FSPRIVATE(bp, xfs_buf_log_item_t *);
+	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
+	ASSERT(!(bip->bli_format.blf_flags & XFS_BLI_CANCEL));
+	ASSERT(atomic_read(&bip->bli_refcount) > 0);
+	ASSERT(bip->bli_flags & XFS_BLI_HOLD);
+	bip->bli_flags &= ~XFS_BLI_HOLD;
+	xfs_buf_item_trace("BHOLD RELEASE", bip);
+}
+
 /*
  * This is called to mark bytes first through last inclusive of the given
  * buffer as needing to be logged when the transaction is committed.

fs/xfs/xfs_vfsops.c

@@ -795,7 +795,6 @@ xfs_statvfs(
 	xfs_mount_t	*mp;
 	xfs_sb_t	*sbp;
 	unsigned long	s;
-	u64 id;
 
 	mp = XFS_BHVTOM(bdp);
 	sbp = &(mp->m_sb);
@@ -823,9 +822,7 @@ xfs_statvfs(
 	statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
 	XFS_SB_UNLOCK(mp, s);
 
-	id = huge_encode_dev(mp->m_dev);
-	statp->f_fsid.val[0] = (u32)id;
-	statp->f_fsid.val[1] = (u32)(id >> 32);
+	xfs_statvfs_fsid(statp, mp);
 	statp->f_namelen = MAXNAMELEN - 1;
 
 	return 0;
@@ -906,7 +903,6 @@ xfs_sync_inodes(
 	xfs_inode_t	*ip_next;
 	xfs_buf_t	*bp;
 	vnode_t		*vp = NULL;
-	vmap_t		vmap;
 	int		error;
 	int		last_error;
 	uint64_t	fflag;
@@ -1101,48 +1097,21 @@ xfs_sync_inodes(
 		 * lock in xfs_ireclaim() after the inode is pulled from
 		 * the mount list will sleep until we release it here.
 		 * This keeps the vnode from being freed while we reference
-		 * it.  It is also cheaper and simpler than actually doing
-		 * a vn_get() for every inode we touch here.
+		 * it.
 		 */
 		if (xfs_ilock_nowait(ip, lock_flags) == 0) {
-
 			if ((flags & SYNC_BDFLUSH) || (vp == NULL)) {
 				ip = ip->i_mnext;
 				continue;
 			}
 
-			/*
-			 * We need to unlock the inode list lock in order
-			 * to lock the inode. Insert a marker record into
-			 * the inode list to remember our position, dropping
-			 * the lock is now done inside the IPOINTER_INSERT
-			 * macro.
-			 *
-			 * We also use the inode list lock to protect us
-			 * in taking a snapshot of the vnode version number
-			 * for use in calling vn_get().
-			 */
-			VMAP(vp, vmap);
-			IPOINTER_INSERT(ip, mp);
-
-			vp = vn_get(vp, &vmap);
+			vp = vn_grab(vp);
 			if (vp == NULL) {
-				/*
-				 * The vnode was reclaimed once we let go
-				 * of the inode list lock.  Skip to the
-				 * next list entry. Remove the marker.
-				 */
-
-				XFS_MOUNT_ILOCK(mp);
-
-				mount_locked = B_TRUE;
-				vnode_refed  = B_FALSE;
-
-				IPOINTER_REMOVE(ip, mp);
-
+				ip = ip->i_mnext;
 				continue;
 			}
 
+			IPOINTER_INSERT(ip, mp);
 			xfs_ilock(ip, lock_flags);
 
 			ASSERT(vp == XFS_ITOV(ip));
@@ -1533,7 +1502,10 @@ xfs_syncsub(
 	 * eventually kicked out of the cache.
 	 */
 	if (flags & SYNC_REFCACHE) {
-		xfs_refcache_purge_some(mp);
+		if (flags & SYNC_WAIT)
+			xfs_refcache_purge_mp(mp);
+		else
+			xfs_refcache_purge_some(mp);
 	}
 
 	/*
@@ -1649,6 +1621,10 @@ xfs_vget(
 #define MNTOPT_SWIDTH	"swidth"	/* data volume stripe width */
 #define MNTOPT_NOUUID	"nouuid"	/* ignore filesystem UUID */
 #define MNTOPT_MTPT	"mtpt"		/* filesystem mount point */
+#define MNTOPT_GRPID	"grpid"		/* group-ID from parent directory */
+#define MNTOPT_NOGRPID	"nogrpid"	/* group-ID from current process */
+#define MNTOPT_BSDGROUPS    "bsdgroups"    /* group-ID from parent directory */
+#define MNTOPT_SYSVGROUPS   "sysvgroups"   /* group-ID from current process */
 #define MNTOPT_ALLOCSIZE    "allocsize"    /* preferred allocation size */
 #define MNTOPT_IHASHSIZE    "ihashsize"    /* size of inode hash table */
 #define MNTOPT_NORECOVERY   "norecovery"   /* don't run XFS recovery */
@@ -1769,6 +1745,12 @@ xfs_parseargs(
 			}
 			args->flags |= XFSMNT_IHASHSIZE;
 			args->ihashsize = simple_strtoul(value, &eov, 10);
+		} else if (!strcmp(this_char, MNTOPT_GRPID) ||
+			   !strcmp(this_char, MNTOPT_BSDGROUPS)) {
+			vfsp->vfs_flag |= VFS_GRPID;
+		} else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
+			   !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
+			vfsp->vfs_flag &= ~VFS_GRPID;
 		} else if (!strcmp(this_char, MNTOPT_WSYNC)) {
 			args->flags |= XFSMNT_WSYNC;
 		} else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
@@ -1890,6 +1872,7 @@ xfs_showargs(
 	};
 	struct proc_xfs_info	*xfs_infop;
 	struct xfs_mount	*mp = XFS_BHVTOM(bhv);
+	struct vfs		*vfsp = XFS_MTOVFS(mp);
 
 	for (xfs_infop = xfs_info; xfs_infop->flag; xfs_infop++) {
 		if (mp->m_flags & xfs_infop->flag)
@@ -1926,7 +1909,10 @@ xfs_showargs(
 
 	if (!(mp->m_flags & XFS_MOUNT_32BITINOOPT))
 		seq_printf(m, "," MNTOPT_64BITINODE);
-	
+
+	if (vfsp->vfs_flag & VFS_GRPID)
+		seq_printf(m, "," MNTOPT_GRPID);
+
 	return 0;
 }
 

fs/xfs/xfs_vnodeops.c

@@ -104,7 +104,7 @@ xfs_open(
 	 * If it's a directory with any blocks, read-ahead block 0
 	 * as we're almost certain to have the next operation be a read there.
 	 */
-	if (vp->v_type == VDIR && ip->i_d.di_nextents > 0) {
+	if (VN_ISDIR(vp) && ip->i_d.di_nextents > 0) {
 		mode = xfs_ilock_map_shared(ip);
 		if (ip->i_d.di_nextents > 0)
 			(void)xfs_da_reada_buf(NULL, ip, 0, XFS_DATA_FORK);
@@ -163,18 +163,21 @@ xfs_getattr(
 	/*
 	 * Copy from in-core inode.
 	 */
-	vap->va_type = vp->v_type;
-	vap->va_mode = ip->i_d.di_mode & MODEMASK;
+	vap->va_mode = ip->i_d.di_mode;
 	vap->va_uid = ip->i_d.di_uid;
 	vap->va_gid = ip->i_d.di_gid;
 	vap->va_projid = ip->i_d.di_projid;
 
 	/*
 	 * Check vnode type block/char vs. everything else.
-	 * Do it with bitmask because that's faster than looking
-	 * for multiple values individually.
 	 */
-	if (((1 << vp->v_type) & ((1<<VBLK) | (1<<VCHR))) == 0) {
+	switch (ip->i_d.di_mode & S_IFMT) {
+	case S_IFBLK:
+	case S_IFCHR:
+		vap->va_rdev = ip->i_df.if_u2.if_rdev;
+		vap->va_blocksize = BLKDEV_IOSIZE;
+		break;
+	default:
 		vap->va_rdev = 0;
 
 		if (!(ip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
@@ -224,9 +227,7 @@ xfs_getattr(
 				(ip->i_d.di_extsize << mp->m_sb.sb_blocklog) :
 				(mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog);
 		}
-	} else {
-		vap->va_rdev = ip->i_df.if_u2.if_rdev;
-		vap->va_blocksize = BLKDEV_IOSIZE;
+		break;
 	}
 
 	vap->va_atime.tv_sec = ip->i_d.di_atime.t_sec;
@@ -468,7 +469,7 @@ xfs_setattr(
 				m |= S_ISGID;
 #if 0
 			/* Linux allows this, Irix doesn't. */
-			if ((vap->va_mode & S_ISVTX) && vp->v_type != VDIR)
+			if ((vap->va_mode & S_ISVTX) && !VN_ISDIR(vp))
 				m |= S_ISVTX;
 #endif
 			if (m && !capable(CAP_FSETID))
@@ -546,10 +547,10 @@ xfs_setattr(
 			goto error_return;
 		}
 
-		if (vp->v_type == VDIR) {
+		if (VN_ISDIR(vp)) {
 			code = XFS_ERROR(EISDIR);
 			goto error_return;
-		} else if (vp->v_type != VREG) {
+		} else if (!VN_ISREG(vp)) {
 			code = XFS_ERROR(EINVAL);
 			goto error_return;
 		}
@@ -1567,7 +1568,7 @@ xfs_release(
 	vp = BHV_TO_VNODE(bdp);
 	ip = XFS_BHVTOI(bdp);
 
-	if ((vp->v_type != VREG) || (ip->i_d.di_mode == 0)) {
+	if (!VN_ISREG(vp) || (ip->i_d.di_mode == 0)) {
 		return 0;
 	}
 
@@ -1895,7 +1896,7 @@ xfs_create(
 	dp = XFS_BHVTOI(dir_bdp);
 	mp = dp->i_mount;
 
-	dm_di_mode = vap->va_mode|VTTOIF(vap->va_type);
+	dm_di_mode = vap->va_mode;
 	namelen = VNAMELEN(dentry);
 
 	if (DM_EVENT_ENABLED(dir_vp->v_vfsp, dp, DM_EVENT_CREATE)) {
@@ -1973,8 +1974,7 @@ xfs_create(
 	    (error = XFS_DIR_CANENTER(mp, tp, dp, name, namelen)))
 		goto error_return;
 	rdev = (vap->va_mask & XFS_AT_RDEV) ? vap->va_rdev : 0;
-	error = xfs_dir_ialloc(&tp, dp,
-			MAKEIMODE(vap->va_type,vap->va_mode), 1,
+	error = xfs_dir_ialloc(&tp, dp, vap->va_mode, 1,
 			rdev, credp, prid, resblks > 0,
 			&ip, &committed);
 	if (error) {
@@ -2620,7 +2620,7 @@ xfs_link(
 	vn_trace_entry(src_vp, __FUNCTION__, (inst_t *)__return_address);
 
 	target_namelen = VNAMELEN(dentry);
-	if (src_vp->v_type == VDIR)
+	if (VN_ISDIR(src_vp))
 		return XFS_ERROR(EPERM);
 
 	src_bdp = vn_bhv_lookup_unlocked(VN_BHV_HEAD(src_vp), &xfs_vnodeops);
@@ -2805,7 +2805,7 @@ xfs_mkdir(
 
 	tp = NULL;
 	dp_joined_to_trans = B_FALSE;
-	dm_di_mode = vap->va_mode|VTTOIF(vap->va_type);
+	dm_di_mode = vap->va_mode;
 
 	if (DM_EVENT_ENABLED(dir_vp->v_vfsp, dp, DM_EVENT_CREATE)) {
 		error = XFS_SEND_NAMESP(mp, DM_EVENT_CREATE,
@@ -2879,8 +2879,7 @@ xfs_mkdir(
 	/*
 	 * create the directory inode.
 	 */
-	error = xfs_dir_ialloc(&tp, dp,
-			MAKEIMODE(vap->va_type,vap->va_mode), 2,
+	error = xfs_dir_ialloc(&tp, dp, vap->va_mode, 2,
 			0, credp, prid, resblks > 0,
 		&cdp, NULL);
 	if (error) {
@@ -3650,7 +3649,7 @@ xfs_rwlock(
 	vnode_t		*vp;
 
 	vp = BHV_TO_VNODE(bdp);
-	if (vp->v_type == VDIR)
+	if (VN_ISDIR(vp))
 		return 1;
 	ip = XFS_BHVTOI(bdp);
 	if (locktype == VRWLOCK_WRITE) {
@@ -3681,7 +3680,7 @@ xfs_rwunlock(
 	vnode_t		*vp;
 
 	vp = BHV_TO_VNODE(bdp);
-	if (vp->v_type == VDIR)
+	if (VN_ISDIR(vp))
 		return;
 	ip = XFS_BHVTOI(bdp);
 	if (locktype == VRWLOCK_WRITE) {
@@ -3847,51 +3846,10 @@ xfs_reclaim(
 		return 0;
 	}
 
-	if ((ip->i_d.di_mode & S_IFMT) == S_IFREG) {
-		if (ip->i_d.di_size > 0) {
-			/*
-			 * Flush and invalidate any data left around that is
-			 * a part of this file.
-			 *
-			 * Get the inode's i/o lock so that buffers are pushed
-			 * out while holding the proper lock.  We can't hold
-			 * the inode lock here since flushing out buffers may
-			 * cause us to try to get the lock in xfs_strategy().
-			 *
-			 * We don't have to call remapf() here, because there
-			 * cannot be any mapped file references to this vnode
-			 * since it is being reclaimed.
-			 */
-			xfs_ilock(ip, XFS_IOLOCK_EXCL);
-
-			/*
-			 * If we hit an IO error, we need to make sure that the
-			 * buffer and page caches of file data for
-			 * the file are tossed away. We don't want to use
-			 * VOP_FLUSHINVAL_PAGES here because we don't want dirty
-			 * pages to stay attached to the vnode, but be
-			 * marked P_BAD. pdflush/vnode_pagebad
-			 * hates that.
-			 */
-			if (!XFS_FORCED_SHUTDOWN(ip->i_mount)) {
-				VOP_FLUSHINVAL_PAGES(vp, 0, -1, FI_NONE);
-			} else {
-				VOP_TOSS_PAGES(vp, 0, -1, FI_NONE);
-			}
+	vn_iowait(vp);
 
-			ASSERT(VN_CACHED(vp) == 0);
-			ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) ||
-			       ip->i_delayed_blks == 0);
-			xfs_iunlock(ip, XFS_IOLOCK_EXCL);
-		} else if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
-			/*
-			 * di_size field may not be quite accurate if we're
-			 * shutting down.
-			 */
-			VOP_TOSS_PAGES(vp, 0, -1, FI_NONE);
-			ASSERT(VN_CACHED(vp) == 0);
-		}
-	}
+	ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
+	ASSERT(VN_CACHED(vp) == 0);
 
 	/* If we have nothing to flush with this inode then complete the
 	 * teardown now, otherwise break the link between the xfs inode
@@ -4567,7 +4525,7 @@ xfs_change_file_space(
 	/*
 	 * must be a regular file and have write permission
 	 */
-	if (vp->v_type != VREG)
+	if (!VN_ISREG(vp))
 		return XFS_ERROR(EINVAL);
 
 	xfs_ilock(ip, XFS_ILOCK_SHARED);