Merge ssh://lord@kernel.bkbits.net/xfs-2.5

into jen.americas.sgi.com:/src/lord/bitkeeper/xfs-2.5

fs/xfs/linux/xfs_aops.c

@@ -803,7 +803,7 @@ page_state_convert(
 		bh = bh->b_this_page;
 	} while (offset < end_offset);
 
-	if (uptodate)
+	if (uptodate && bh == head)
 		SetPageUptodate(page);
 
 	if (startio)

fs/xfs/linux/xfs_file.c

@@ -262,7 +262,11 @@ linvfs_readdir(
 	uio.uio_iov = &iov;
 	uio.uio_fmode = filp->f_mode;
 	uio.uio_segflg = UIO_SYSSPACE;
-	curr_offset = uio.uio_offset = filp->f_pos;
+	curr_offset = filp->f_pos;
+	if (filp->f_pos != 0x7fffffff)
+		uio.uio_offset = filp->f_pos;
+	else
+		uio.uio_offset = 0xffffffff;
 
 	while (!eof) {
 		uio.uio_resid = iov.iov_len = rlen;
@@ -283,13 +287,13 @@ linvfs_readdir(
 			namelen = strlen(dbp->d_name);
 
 			if (filldir(dirent, dbp->d_name, namelen,
-					(loff_t) curr_offset,
+					(loff_t) curr_offset & 0x7fffffff,
 					(ino_t) dbp->d_ino,
 					DT_UNKNOWN)) {
 				goto done;
 			}
 			size -= dbp->d_reclen;
-			curr_offset = (loff_t)dbp->d_off & 0x7fffffff;
+			curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
 			dbp = nextdp(dbp);
 		}
 	}

fs/xfs/linux/xfs_super.c

@@ -506,6 +506,24 @@ linvfs_write_super(
 	sb->s_dirt = 0;
 }
 
+STATIC int
+linvfs_sync_super(
+	struct super_block	*sb,
+	int			wait)
+{
+	vfs_t		*vfsp = LINVFS_GET_VFS(sb);
+	int		error;
+	int		flags = SYNC_FSDATA;
+
+	if (wait)
+		flags |= SYNC_WAIT;
+
+	VFS_SYNC(vfsp, flags, NULL, error);
+	sb->s_dirt = 0;
+
+	return -error;
+}
+
 STATIC int
 linvfs_statfs(
 	struct super_block	*sb,
@@ -798,6 +816,7 @@ STATIC struct super_operations linvfs_sops = {
 	.clear_inode		= linvfs_clear_inode,
 	.put_super		= linvfs_put_super,
 	.write_super		= linvfs_write_super,
+	.sync_fs		= linvfs_sync_super,
 	.write_super_lockfs	= linvfs_freeze_fs,
 	.unlockfs		= linvfs_unfreeze_fs,
 	.statfs			= linvfs_statfs,

fs/xfs/xfs_dinode.h

@@ -72,7 +72,8 @@ typedef struct xfs_dinode_core
 	__uint32_t	di_gid;		/* owner's group id */
 	__uint32_t	di_nlink;	/* number of links to file */
 	__uint16_t	di_projid;	/* owner's project id */
-	__uint8_t	di_pad[10];	/* unused, zeroed space */
+	__uint8_t	di_pad[8];	/* unused, zeroed space */
+	__uint16_t	di_flushiter;	/* incremented on flush */
 	xfs_timestamp_t	di_atime;	/* time last accessed */
 	xfs_timestamp_t	di_mtime;	/* time last modified */
 	xfs_timestamp_t	di_ctime;	/* time created/inode modified */
@@ -89,6 +90,8 @@ typedef struct xfs_dinode_core
 	__uint32_t	di_gen;		/* generation number */
 } xfs_dinode_core_t;
 
+#define DI_MAX_FLUSH 0xffff
+
 typedef struct xfs_dinode
 {
 	xfs_dinode_core_t	di_core;

fs/xfs/xfs_dir2_leaf.h

@@ -64,7 +64,7 @@ struct xfs_trans;
  * Offset in data space of a data entry.
  */
 typedef	__uint32_t	xfs_dir2_dataptr_t;
-#define	XFS_DIR2_MAX_DATAPTR	((xfs_dir2_dataptr_t)0x7fffffff)
+#define	XFS_DIR2_MAX_DATAPTR	((xfs_dir2_dataptr_t)0xffffffff)
 #define	XFS_DIR2_NULL_DATAPTR	((xfs_dir2_dataptr_t)0)
 
 /*

fs/xfs/xfs_dir_leaf.c

@@ -560,14 +560,7 @@ xfs_dir_shortform_getdents(xfs_inode_t *dp, uio_t *uio, int *eofp,
 		 */
 		if (sbp->seqno == 0 || sbp == sbuf)
 			lastresid = uio->uio_resid;
-		/*
-		 * NOTE! Linux "filldir" semantics require that the
-		 *	 offset "cookie" be for this entry, not the
-		 *	 next; all the actual shuffling to make it
-		 *	 "look right" to the user is done in filldir.
-		 */
-		XFS_PUT_COOKIE(p.cook, mp, 0, sbp->seqno, sbp->hash);
-
+		XFS_PUT_COOKIE(p.cook, mp, 0, sbp[1].seqno, sbp[1].hash);
 #if XFS_BIG_FILESYSTEMS
 		p.ino = sbp->ino + mp->m_inoadd;
 #else
@@ -575,9 +568,7 @@ xfs_dir_shortform_getdents(xfs_inode_t *dp, uio_t *uio, int *eofp,
 #endif
 		p.name = sbp->name;
 		p.namelen = sbp->namelen;
-
 		retval = p.put(&p);
-
 		if (!p.done) {
 			uio->uio_offset =
 				XFS_DA_MAKE_COOKIE(mp, 0, 0, sbp->hash);
@@ -586,20 +577,12 @@ xfs_dir_shortform_getdents(xfs_inode_t *dp, uio_t *uio, int *eofp,
 			xfs_dir_trace_g_du("sf: E-O-B", dp, uio);
 			return retval;
 		}
-
 		sbp++;
 	}
-
 	kmem_free(sbuf, sbsize);
-
-	XFS_PUT_COOKIE(p.cook, mp, 0, 0, XFS_DA_MAXHASH);
-
 	uio->uio_offset = p.cook.o;
-
 	*eofp = 1;
-
 	xfs_dir_trace_g_du("sf: E-O-F", dp, uio);
-
 	return 0;
 }
 
@@ -2070,16 +2053,6 @@ xfs_dir_leaf_getdents_int(
 			return XFS_ERROR(EFSCORRUPTED);
 		}
 
-		thishash = INT_GET(entry->hashval, ARCH_CONVERT);
-
-		/*
-		 * NOTE! Linux "filldir" semantics require that the
-		 *	 offset "cookie" be for this entry, not the
-		 *	 next; all the actual shuffling to make it
-		 *	 "look right" to the user is done in filldir.
-		 */
-		XFS_PUT_COOKIE(p.cook, mp, bno, entno, thishash);
-
 		xfs_dir_trace_g_duc("leaf: middle cookie  ",
 						   dp, uio, p.cook.o);
 
@@ -2090,17 +2063,19 @@ xfs_dir_leaf_getdents_int(
 				nextentno = entno + 1;
 			else
 				nextentno = 0;
+			XFS_PUT_COOKIE(p.cook, mp, bno, nextentno, nexthash);
+			xfs_dir_trace_g_duc("leaf: middle cookie  ",
+						   dp, uio, p.cook.o);
 
-		} else if (INT_GET(leaf->hdr.info.forw, ARCH_CONVERT)) {
+		} else if ((thishash = INT_GET(leaf->hdr.info.forw,
+							ARCH_CONVERT))) {
 			xfs_dabuf_t *bp2;
 			xfs_dir_leafblock_t *leaf2;
 
 			ASSERT(nextda != -1);
 
-			retval = xfs_da_read_buf(dp->i_transp, dp,
-						 INT_GET(leaf->hdr.info.forw,
-						 ARCH_CONVERT), nextda,
-						 &bp2, XFS_DATA_FORK);
+			retval = xfs_da_read_buf(dp->i_transp, dp, thishash,
+						 nextda, &bp2, XFS_DATA_FORK);
 			if (retval)
 				return(retval);
 
@@ -2124,13 +2099,13 @@ xfs_dir_leaf_getdents_int(
 			nexthash = INT_GET(leaf2->entries[0].hashval,
 								ARCH_CONVERT);
 			nextentno = -1;
-
+			XFS_PUT_COOKIE(p.cook, mp, thishash, 0, nexthash);
 			xfs_da_brelse(dp->i_transp, bp2);
 			xfs_dir_trace_g_duc("leaf: next blk cookie",
 						   dp, uio, p.cook.o);
 		} else {
 			nextentno = -1;
-			nexthash  = XFS_DA_MAXHASH;
+			XFS_PUT_COOKIE(p.cook, mp, 0, 0, XFS_DA_MAXHASH);
 		}
 
 		/*
@@ -2147,7 +2122,8 @@ xfs_dir_leaf_getdents_int(
 		 * provided is big enough to handle it (see pv763517).
 		 */
 #if (BITS_PER_LONG == 32)
-		if (INT_GET(entry->hashval, ARCH_CONVERT) != lasthash) {
+		if ((thishash = INT_GET(entry->hashval, ARCH_CONVERT))
+								!= lasthash) {
 			XFS_PUT_COOKIE(lastoffset, mp, bno, entno, thishash);
 			lastresid = uio->uio_resid;
 			lasthash = thishash;
@@ -2156,6 +2132,7 @@ xfs_dir_leaf_getdents_int(
 						   dp, uio, p.cook.o);
 		}
 #else
+		thishash = INT_GET(entry->hashval, ARCH_CONVERT);
 		XFS_PUT_COOKIE(lastoffset, mp, bno, entno, thishash);
 		lastresid = uio->uio_resid;
 #endif /* BITS_PER_LONG == 32 */
@@ -2187,8 +2164,6 @@ xfs_dir_leaf_getdents_int(
 		}
 	}
 
-	XFS_PUT_COOKIE(p.cook, mp, 0, 0, nexthash);
-
 	uio->uio_offset = p.cook.o;
 
 	*eobp = 0;

fs/xfs/xfs_inode.c

@@ -825,6 +825,8 @@ xfs_xlate_dinode_core(
 			sizeof(buf_core->di_pad));
 	}
 
+	INT_XLATE(buf_core->di_flushiter, mem_core->di_flushiter, dir, arch);
+
 	INT_XLATE(buf_core->di_atime.t_sec, mem_core->di_atime.t_sec,
 			dir, arch);
 	INT_XLATE(buf_core->di_atime.t_nsec, mem_core->di_atime.t_nsec,
@@ -956,6 +958,7 @@ xfs_iread(
 		ip->i_d.di_magic = INT_GET(dip->di_core.di_magic, ARCH_CONVERT);
 		ip->i_d.di_version = INT_GET(dip->di_core.di_version, ARCH_CONVERT);
 		ip->i_d.di_gen = INT_GET(dip->di_core.di_gen, ARCH_CONVERT);
+		ip->i_d.di_flushiter = INT_GET(dip->di_core.di_flushiter, ARCH_CONVERT);
 		/*
 		 * Make sure to pull in the mode here as well in
 		 * case the inode is released without being used.
@@ -3232,6 +3235,13 @@ xfs_iflush_int(
 			ip->i_ino, ip->i_d.di_forkoff, ip);
 		goto corrupt_out;
 	}
+	/*
+	 * bump the flush iteration count, used to detect flushes which
+	 * postdate a log record during recovery.
+	 */
+
+	ip->i_d.di_flushiter++;
+
 	/*
 	 * Copy the dirty parts of the inode into the on-disk
 	 * inode.  We always copy out the core of the inode,
@@ -3241,6 +3251,10 @@ xfs_iflush_int(
 	xfs_xlate_dinode_core((xfs_caddr_t)&(dip->di_core), &(ip->i_d),
 		-1, ARCH_CONVERT);
 
+	/* Wrap, we never let the log put out DI_MAX_FLUSH */
+	if (ip->i_d.di_flushiter == DI_MAX_FLUSH)
+		ip->i_d.di_flushiter = 0;
+
 	/*
 	 * If this is really an old format inode and the superblock version
 	 * has not been updated to support only new format inodes, then

fs/xfs/xfs_log_recover.c

@@ -2326,6 +2326,26 @@ xlog_recover_do_inode_trans(
 				 XFS_ERRLEVEL_LOW, mp);
 		return XFS_ERROR(EFSCORRUPTED);
 	}
+
+	/* Skip replay when the on disk inode is newer than the log one */
+	if (dicp->di_flushiter <
+	    INT_GET(dip->di_core.di_flushiter, ARCH_CONVERT)) {
+		/*
+		 * Deal with the wrap case, DI_MAX_FLUSH is less
+		 * than smaller numbers
+		 */
+		if ((INT_GET(dip->di_core.di_flushiter, ARCH_CONVERT)
+							== DI_MAX_FLUSH) &&
+		    (dicp->di_flushiter < (DI_MAX_FLUSH>>1))) {
+			/* do nothing */
+		} else {
+			xfs_buf_relse(bp);
+			return 0;
+		}
+	}
+	/* Take the opportunity to reset the flush iteration count */
+	dicp->di_flushiter = 0;
+
 	if (unlikely((dicp->di_mode & IFMT) == IFREG)) {
 		if ((dicp->di_format != XFS_DINODE_FMT_EXTENTS) &&
 		    (dicp->di_format != XFS_DINODE_FMT_BTREE)) {

fs/xfs/xfs_vfsops.c

@@ -1506,7 +1506,7 @@ xfs_syncsub(
 		xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 		error = xfs_trans_commit(tp, 0, NULL);
 		xfs_iunlock(ip, XFS_ILOCK_EXCL);
-		xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
+		xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
 	}
 
 	/*

fs/xfs/xfs_vnodeops.c

@@ -3730,9 +3730,7 @@ xfs_inode_flush(
 	int		flags)
 {
 	xfs_inode_t	*ip;
-	xfs_dinode_t	*dip;
 	xfs_mount_t	*mp;
-	xfs_buf_t	*bp;
 	int		error = 0;
 
 	ip = XFS_BHVTOI(bdp);
@@ -3778,44 +3776,30 @@ xfs_inode_flush(
 	 * now, they get caught later by xfs_sync.
 	 */
 	if (flags & FLUSH_INODE) {
-		if (xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) {
-			if ((xfs_ipincount(ip) == 0) && xfs_iflock_nowait(ip)) {
-				int	flush_flags;
+		int	flush_flags;
 
-#if 0
-				/* not turning this on until some
-				 * performance analysis is done
-				 */
-				if (flags & FLUSH_SYNC)
-					flush_flags = XFS_IFLUSH_SYNC;
-				else
-#endif
-				flush_flags = XFS_IFLUSH_DELWRI_ELSE_ASYNC;
+		if (xfs_ipincount(ip))
+			return EAGAIN;
 
-				xfs_ifunlock(ip);
+		if (flags & FLUSH_SYNC) {
+			xfs_ilock(ip, XFS_ILOCK_SHARED);
+			xfs_iflock(ip);
+		} else if (xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) {
+			if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip)) {
 				xfs_iunlock(ip, XFS_ILOCK_SHARED);
-				error = xfs_itobp(mp, NULL, ip, &dip, &bp, 0);
-				if (error)
-					return error;
-				xfs_buf_relse(bp);
-
-				if (xfs_ilock_nowait(ip, XFS_ILOCK_SHARED) == 0)
-					return EAGAIN;
-
-				if (xfs_ipincount(ip) ||
-				    !xfs_iflock_nowait(ip)) {
-					xfs_iunlock(ip, XFS_ILOCK_SHARED);
-					return EAGAIN;
-				}
-
-				error = xfs_iflush(ip, flush_flags);
-			} else {
-				error = EAGAIN;
+				return EAGAIN;
 			}
-			xfs_iunlock(ip, XFS_ILOCK_SHARED);
 		} else {
-			error = EAGAIN;
+			return EAGAIN;
 		}
+
+		if (flags & FLUSH_SYNC)
+			flush_flags = XFS_IFLUSH_SYNC;
+		else
+			flush_flags = XFS_IFLUSH_ASYNC;
+
+		error = xfs_iflush(ip, flush_flags);
+		xfs_iunlock(ip, XFS_ILOCK_SHARED);
 	}
 
 	return error;

fs/xfs/xfsidbg.c

@@ -3033,11 +3033,12 @@ xfs_prdinode_core(xfs_dinode_core_t *dip, int convert)
 		INT_GET(dip->di_format, convert),
 		xfs_fmtformat(
 		    (xfs_dinode_fmt_t)INT_GET(dip->di_format, convert)));
-	kdb_printf("nlink 0x%x uid 0x%x gid 0x%x projid 0x%x\n",
+	kdb_printf("nlink %d uid %d gid %d projid %d flushiter %u\n",
 		INT_GET(dip->di_nlink, convert),
 		INT_GET(dip->di_uid, convert),
 		INT_GET(dip->di_gid, convert),
-		(uint)INT_GET(dip->di_projid, convert));
+		(uint)INT_GET(dip->di_projid, convert),
+		(uint)INT_GET(dip->di_flushiter, convert));
 	kdb_printf("atime 0x%x:%x mtime 0x%x:%x ctime 0x%x:%x\n",
 		INT_GET(dip->di_atime.t_sec, convert),
 		INT_GET(dip->di_atime.t_nsec, convert),