Merge http://oss.sgi.com:8090/xfs-linux-2.6

into ppc970.osdl.org:/home/torvalds/v2.6/linux

fs/xfs/linux-2.6/xfs_aops.c

@@ -550,6 +550,7 @@ xfs_map_unwritten(
 STATIC void
 xfs_submit_page(
 	struct page		*page,
+	struct writeback_control *wbc,
 	struct buffer_head	*bh_arr[],
 	int			cnt)
 {
@@ -573,8 +574,10 @@ xfs_submit_page(
 
 		for (i = 0; i < cnt; i++)
 			submit_bh(WRITE, bh_arr[i]);
-	} else
+	} else {
 		end_page_writeback(page);
+		wbc->pages_skipped++;	/* We didn't write this page */
+	}
 }
 
 /*
@@ -652,7 +655,7 @@ xfs_convert_page(
 
 	if (startio) {
 		wbc->nr_to_write--;
-		xfs_submit_page(page, bh_arr, index);
+		xfs_submit_page(page, wbc, bh_arr, index);
 	} else {
 		unlock_page(page);
 	}
@@ -864,7 +867,7 @@ xfs_page_state_convert(
 		SetPageUptodate(page);
 
 	if (startio)
-		xfs_submit_page(page, bh_arr, cnt);
+		xfs_submit_page(page, wbc, bh_arr, cnt);
 
 	if (iomp) {
 		tlast = (iomp->iomap_offset + iomp->iomap_bsize - 1) >>

fs/xfs/linux-2.6/xfs_file.c

@@ -409,6 +409,8 @@ linvfs_file_mmap(
 	vma->vm_ops = &linvfs_file_vm_ops;
 
 	VOP_SETATTR(vp, &va, XFS_AT_UPDATIME, NULL, error);
+	if (!error)
+		vn_revalidate(vp);	/* update Linux inode flags */
 	return 0;
 }
 

fs/xfs/linux-2.6/xfs_globals.c

@@ -63,6 +63,7 @@ xfs_param_t xfs_params = {
 	.inherit_noatim = {	0,		1,		1	},
 	.xfs_buf_timer	= {	100/2,		1*100,		30*100	},
 	.xfs_buf_age	= {	1*100,		15*100,		7200*100},
+	.inherit_nosym	= {	0,		0,		1	},
 };
 
 /*

fs/xfs/linux-2.6/xfs_iops.c

@@ -174,8 +174,9 @@ linvfs_mknod(
 				 */
 				teardown.d_inode = ip = LINVFS_GET_IP(vp);
 				teardown.d_name = dentry->d_name;
-				remove_inode_hash(ip);
-				make_bad_inode(ip);
+
+				vn_mark_bad(vp);
+				
 				if (S_ISDIR(mode))
 					VOP_RMDIR(dvp, &teardown, NULL, err2);
 				else
@@ -225,26 +226,21 @@ linvfs_lookup(
 	struct dentry	*dentry,
 	struct nameidata *nd)
 {
-	struct inode	*ip = NULL;
-	vnode_t		*vp, *cvp = NULL;
+	struct vnode	*vp = LINVFS_GET_VP(dir), *cvp;
 	int		error;
 
 	if (dentry->d_name.len >= MAXNAMELEN)
 		return ERR_PTR(-ENAMETOOLONG);
 
-	vp = LINVFS_GET_VP(dir);
 	VOP_LOOKUP(vp, dentry, &cvp, 0, NULL, NULL, error);
-	if (!error) {
-		ASSERT(cvp);
-		ip = LINVFS_GET_IP(cvp);
-		if (!ip) {
-			VN_RELE(cvp);
-			return ERR_PTR(-EACCES);
-		}
-	}
-	if (error && (error != ENOENT))
+	if (error) {
+		if (unlikely(error != ENOENT))
 			return ERR_PTR(-error);
-	return d_splice_alias(ip, dentry);
+		d_add(dentry, NULL);
+		return NULL;
+	}
+
+	return d_splice_alias(LINVFS_GET_IP(cvp), dentry);
 }
 
 STATIC int
@@ -304,7 +300,7 @@ linvfs_symlink(
 {
 	struct inode	*ip;
 	vattr_t		va;
-	vnode_t		*dvp;	/* directory containing name to remove */
+	vnode_t		*dvp;	/* directory containing name of symlink */
 	vnode_t		*cvp;	/* used to lookup symlink to put in dentry */
 	int		error;
 

fs/xfs/linux-2.6/xfs_linux.h

@@ -85,6 +85,7 @@
 #include <linux/vfs.h>
 #include <linux/seq_file.h>
 #include <linux/init.h>
+#include <linux/list.h>
 #include <linux/proc_fs.h>
 #include <linux/version.h>
 
@@ -140,6 +141,7 @@ static inline void set_buffer_unwritten_io(struct buffer_head *bh)
 #define xfs_inherit_noatime	xfs_params.inherit_noatim.val
 #define xfs_buf_timer_centisecs	xfs_params.xfs_buf_timer.val
 #define xfs_buf_age_centisecs	xfs_params.xfs_buf_age.val
+#define xfs_inherit_nosymlinks	xfs_params.inherit_nosym.val
 
 #define current_cpu()		smp_processor_id()
 #define current_pid()		(current->pid)

fs/xfs/linux-2.6/xfs_lrw.c

@@ -296,11 +296,6 @@ xfs_read(
 		return -EIO;
 	}
 
-	/* OK so we are holding the I/O lock for the duration
-	 * of the submission, then what happens if the I/O
-	 * does not really happen here, but is scheduled 
-	 * later?
-	 */
 	xfs_ilock(ip, XFS_IOLOCK_SHARED);
 
 	if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ) &&
@@ -321,6 +316,7 @@ xfs_read(
 	ret = __generic_file_aio_read(iocb, iovp, segs, offset);
 	if (ret == -EIOCBQUEUED)
 		ret = wait_on_sync_kiocb(iocb);
+
 	xfs_iunlock(ip, XFS_IOLOCK_SHARED);
 
 	if (ret > 0)
@@ -380,12 +376,17 @@ xfs_sendfile(
 		}
 	}
 	xfs_rw_enter_trace(XFS_SENDFILE_ENTER, &ip->i_iocore,
-			   (void*)(unsigned long)target, count, *offset, ioflags);
+		   (void *)(unsigned long)target, count, *offset, ioflags);
 	ret = generic_file_sendfile(filp, offset, count, actor, target);
+
 	xfs_iunlock(ip, XFS_IOLOCK_SHARED);
 
+	if (ret > 0)
 		XFS_STATS_ADD(xs_read_bytes, ret);
+
+	if (likely(!(ioflags & IO_INVIS)))
 		xfs_ichgtime(ip, XFS_ICHGTIME_ACC);
+
 	return ret;
 }
 

fs/xfs/linux-2.6/xfs_super.c

@@ -141,7 +141,7 @@ xfs_set_inodeops(
 	vnode_t			*vp = LINVFS_GET_VP(inode);
 
 	if (vp->v_type == VNON) {
-		make_bad_inode(inode);
+		vn_mark_bad(vp);
 	} else if (S_ISREG(inode->i_mode)) {
 		inode->i_op = &linvfs_file_inode_operations;
 		inode->i_fop = &linvfs_file_operations;
@@ -223,42 +223,21 @@ xfs_initialize_vnode(
 		bhv_insert(VN_BHV_HEAD(vp), inode_bhv);
 	}
 
-	vp->v_type = IFTOVT(ip->i_d.di_mode);
-
-	/* Have we been called during the new inode create process,
-	 * in which case we are too early to fill in the Linux inode.
+	/*
+	 * We need to set the ops vectors, and unlock the inode, but if
+	 * we have been called during the new inode create process, it is
+	 * too early to fill in the Linux inode.  We will get called a
+	 * second time once the inode is properly set up, and then we can
+	 * finish our work.
 	 */
-	if (vp->v_type == VNON)
-		return;
-
+	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);
-
-	/* For new inodes we need to set the ops vectors,
-	 * and unlock the inode.
-	 */
-	if (unlock && (inode->i_state & I_NEW)) {
 		xfs_set_inodeops(inode);
 		unlock_new_inode(inode);
 	}
 }
 
-void
-xfs_flush_inode(
-	xfs_inode_t	*ip)
-{
-	struct inode	*inode = LINVFS_GET_IP(XFS_ITOV(ip));
-
-	filemap_flush(inode->i_mapping);
-}
-
-void
-xfs_flush_device(
-	xfs_inode_t	*ip)
-{
-	sync_blockdev(XFS_ITOV(ip)->v_vfsp->vfs_super->s_bdev);
-	xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
-}
-
 int
 xfs_blkdev_get(
 	xfs_mount_t		*mp,
@@ -312,7 +291,6 @@ xfs_inode_shake(
 {
 	int		pages;
 
-	
 	pages = kmem_zone_shrink(linvfs_inode_zone);
 	pages += kmem_zone_shrink(xfs_inode_zone);
 	return pages;
@@ -337,7 +315,6 @@ init_inodecache( void )
 	linvfs_inode_zone = kmem_cache_create("linvfs_icache",
 				sizeof(vnode_t), 0, SLAB_RECLAIM_ACCOUNT,
 				init_once, NULL);
-
 	if (linvfs_inode_zone == NULL)
 		return -ENOMEM;
 	return 0;
@@ -391,36 +368,146 @@ linvfs_clear_inode(
 }
 
 
+/*
+ * Enqueue a work item to be picked up by the vfs xfssyncd thread.
+ * Doing this has two advantages:
+ * - It saves on stack space, which is tight in certain situations
+ * - It can be used (with care) as a mechanism to avoid deadlocks.
+ * Flushing while allocating in a full filesystem requires both.
+ */
+STATIC void
+xfs_syncd_queue_work(
+	struct vfs	*vfs,
+	void		*data,
+	void		(*syncer)(vfs_t *, void *))
+{
+	vfs_sync_work_t	*work;
+
+	work = kmem_alloc(sizeof(struct vfs_sync_work), KM_SLEEP);
+	INIT_LIST_HEAD(&work->w_list);
+	work->w_syncer = syncer;
+	work->w_data = data;
+	work->w_vfs = vfs;
+	spin_lock(&vfs->vfs_sync_lock);
+	list_add_tail(&work->w_list, &vfs->vfs_sync_list);
+	spin_unlock(&vfs->vfs_sync_lock);
+	wake_up_process(vfs->vfs_sync_task);
+}
+
+/*
+ * Flush delayed allocate data, attempting to free up reserved space
+ * from existing allocations.  At this point a new allocation attempt
+ * has failed with ENOSPC and we are in the process of scratching our
+ * heads, looking about for more room...
+ */
+STATIC void
+xfs_flush_inode_work(
+	vfs_t		*vfs,
+	void		*inode)
+{
+	filemap_flush(((struct inode *)inode)->i_mapping);
+	iput((struct inode *)inode);
+}
+
+void
+xfs_flush_inode(
+	xfs_inode_t	*ip)
+{
+	struct inode	*inode = LINVFS_GET_IP(XFS_ITOV(ip));
+	struct vfs	*vfs = XFS_MTOVFS(ip->i_mount);
+
+	igrab(inode);
+	xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work);
+	delay(HZ/2);
+}
+
+/*
+ * This is the "bigger hammer" version of xfs_flush_inode_work...
+ * (IOW, "If at first you don't succeed, use a Bigger Hammer").
+ */
+STATIC void
+xfs_flush_device_work(
+	vfs_t		*vfs,
+	void		*inode)
+{
+	sync_blockdev(vfs->vfs_super->s_bdev);
+	iput((struct inode *)inode);
+}
+
+void
+xfs_flush_device(
+	xfs_inode_t	*ip)
+{
+	struct inode	*inode = LINVFS_GET_IP(XFS_ITOV(ip));
+	struct vfs	*vfs = XFS_MTOVFS(ip->i_mount);
+
+	igrab(inode);
+	xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work);
+	delay(HZ/2);
+	xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
+}
+
 #define SYNCD_FLAGS	(SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR)
+STATIC void
+vfs_sync_worker(
+	vfs_t		*vfsp,
+	void		*unused)
+{
+	int		error;
+
+	if (!(vfsp->vfs_flag & VFS_RDONLY))
+		VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error);
+	vfsp->vfs_sync_seq++;
+	wmb();
+	wake_up(&vfsp->vfs_wait_single_sync_task);
+}
 
 STATIC int
 xfssyncd(
 	void			*arg)
 {
+	long			timeleft;
 	vfs_t			*vfsp = (vfs_t *) arg;
-	int			error;
+	struct list_head	tmp;
+	struct vfs_sync_work	*work, *n;
 
 	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);
-		schedule_timeout((xfs_syncd_centisecs * HZ) / 100);
+		timeleft = schedule_timeout(timeleft);
 		/* swsusp */
 		if (current->flags & PF_FREEZE)
 			refrigerator(PF_FREEZE);
 		if (vfsp->vfs_flag & VFS_UMOUNT)
 			break;
-		if (vfsp->vfs_flag & VFS_RDONLY)
-			continue;
-		VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error);
 
-		vfsp->vfs_sync_seq++;
-		wmb();
-		wake_up(&vfsp->vfs_wait_single_sync_task);
+		spin_lock(&vfsp->vfs_sync_lock);
+		if (!timeleft) {
+			timeleft = (xfs_syncd_centisecs * HZ) / 100;
+			INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list);
+			list_add_tail(&vfsp->vfs_sync_work.w_list,
+					&vfsp->vfs_sync_list);
+		}
+		list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list)
+			list_move(&work->w_list, &tmp);
+		spin_unlock(&vfsp->vfs_sync_lock);
+
+		list_for_each_entry_safe(work, n, &tmp, w_list) {
+			(*work->w_syncer)(vfsp, work->w_data);
+			list_del(&work->w_list);
+			if (work == &vfsp->vfs_sync_work)
+				continue;
+			kmem_free(work, sizeof(struct vfs_sync_work));
+		}
 	}
 
 	vfsp->vfs_sync_task = NULL;
@@ -570,7 +657,6 @@ linvfs_get_parent(
 	int			error;
 	vnode_t			*vp, *cvp;
 	struct dentry		*parent;
-	struct inode		*ip = NULL;
 	struct dentry		dotdot;
 
 	dotdot.d_name.name = "..";
@@ -580,21 +666,13 @@ linvfs_get_parent(
 	cvp = NULL;
 	vp = LINVFS_GET_VP(child->d_inode);
 	VOP_LOOKUP(vp, &dotdot, &cvp, 0, NULL, NULL, error);
-
-	if (!error) {
-		ASSERT(cvp);
-		ip = LINVFS_GET_IP(cvp);
-		if (!ip) {
-			VN_RELE(cvp);
-			return ERR_PTR(-EACCES);
-		}
-	}
-	if (error)
+	if (unlikely(error))
 		return ERR_PTR(-error);
-	parent = d_alloc_anon(ip);
-	if (!parent) {
+
+	parent = d_alloc_anon(LINVFS_GET_IP(cvp));
+	if (unlikely(!parent)) {
 		VN_RELE(cvp);
-		parent = ERR_PTR(-ENOMEM);
+		return ERR_PTR(-ENOMEM);
 	}
 	return parent;
 }

fs/xfs/linux-2.6/xfs_sysctl.c

@@ -129,6 +129,11 @@ STATIC ctl_table xfs_table[] = {
 	&sysctl_intvec, NULL,
 	&xfs_params.xfs_buf_age.min, &xfs_params.xfs_buf_age.max},
 
+	{XFS_INHERIT_NOSYM, "inherit_nosymlinks", &xfs_params.inherit_nosym.val,
+	sizeof(int), 0644, NULL, &proc_dointvec_minmax,
+	&sysctl_intvec, NULL,
+	&xfs_params.inherit_nosym.min, &xfs_params.inherit_nosym.max},
+
 	/* please keep this the last entry */
 #ifdef CONFIG_PROC_FS
 	{XFS_STATS_CLEAR, "stats_clear", &xfs_params.stats_clear.val,

fs/xfs/linux-2.6/xfs_sysctl.h

@@ -59,6 +59,7 @@ typedef struct xfs_param {
 	xfs_sysctl_val_t inherit_noatim;/* Inherit the "noatime" inode flag. */
 	xfs_sysctl_val_t xfs_buf_timer;	/* Interval between xfsbufd wakeups. */
 	xfs_sysctl_val_t xfs_buf_age;	/* Metadata buffer age before flush. */
+	xfs_sysctl_val_t inherit_nosym;	/* Inherit the "nosymlinks" flag. */
 } xfs_param_t;
 
 /*
@@ -95,6 +96,7 @@ enum {
 	XFS_BUF_TIMER = 16,
 	XFS_BUF_AGE = 17,
 	/* XFS_IO_BYPASS = 18 */
+	XFS_INHERIT_NOSYM = 19,
 };
 
 extern xfs_param_t	xfs_params;

fs/xfs/linux-2.6/xfs_vfs.c

@@ -249,6 +249,8 @@ vfs_allocate( void )
 
 	vfsp = kmem_zalloc(sizeof(vfs_t), KM_SLEEP);
 	bhv_head_init(VFS_BHVHEAD(vfsp), "vfs");
+	INIT_LIST_HEAD(&vfsp->vfs_sync_list);
+	vfsp->vfs_sync_lock = SPIN_LOCK_UNLOCKED;
 	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

@@ -36,6 +36,7 @@
 #include "xfs_fs.h"
 
 struct fid;
+struct vfs;
 struct cred;
 struct vnode;
 struct kstatfs;
@@ -45,14 +46,24 @@ struct xfs_mount_args;
 
 typedef struct kstatfs xfs_statfs_t;
 
+typedef struct vfs_sync_work {
+	struct list_head	w_list;
+	struct vfs		*w_vfs;
+	void			*w_data;	/* syncer routine argument */
+	void			(*w_syncer)(struct vfs *, void *);
+} vfs_sync_work_t;
+
 typedef struct vfs {
 	u_int			vfs_flag;	/* flags */
 	xfs_fsid_t		vfs_fsid;	/* file system ID */
 	xfs_fsid_t		*vfs_altfsid;	/* An ID fixed for life of FS */
 	bhv_head_t		vfs_bh;		/* head of vfs behavior chain */
-	struct super_block	*vfs_super;	/* Linux superblock structure */
-	struct task_struct	*vfs_sync_task;	/* xfssyncd process */
-	int 			vfs_sync_seq;	/* xfssyncd generation number */
+	struct super_block	*vfs_super;	/* generic superblock pointer */
+	struct task_struct	*vfs_sync_task;	/* generalised sync thread */
+	vfs_sync_work_t		vfs_sync_work;	/* work item for VFS_SYNC */
+	struct list_head	vfs_sync_list;	/* sync thread work item list */
+	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;

fs/xfs/linux-2.6/xfs_vnode.h

@@ -594,6 +594,19 @@ static __inline__ void vn_flagclr(struct vnode *vp, uint flag)
 #define VN_ATIMESET(vp, tvp)	(LINVFS_GET_IP(vp)->i_atime = *(tvp))
 #define VN_CTIMESET(vp, tvp)	(LINVFS_GET_IP(vp)->i_ctime = *(tvp))
 
+/*
+ * Dealing with bad inodes
+ */
+static inline void vn_mark_bad(struct vnode *vp)
+{
+	make_bad_inode(LINVFS_GET_IP(vp));
+}
+
+static inline int VN_BAD(struct vnode *vp)
+{
+	return is_bad_inode(LINVFS_GET_IP(vp));
+}
+
 /*
  * Some useful predicates.
  */

fs/xfs/xfs_attr.c

@@ -121,23 +121,15 @@ xfs_attr_fetch(xfs_inode_t *ip, char *name, int namelen,
 	xfs_da_args_t   args;
 	int             error;
 
-	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
-		return(EIO);
-
 	if ((XFS_IFORK_Q(ip) == 0) ||
 	    (ip->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
 	     ip->i_d.di_anextents == 0))
 		return(ENOATTR);
 
-	if (!(flags & ATTR_KERNACCESS)) {
-		xfs_ilock(ip, XFS_ILOCK_SHARED);
-
-		if (!(flags & ATTR_SECURE) &&
-		    ((error = xfs_iaccess(ip, S_IRUSR, cred)))) {
-			xfs_iunlock(ip, XFS_ILOCK_SHARED);
+	if (!(flags & (ATTR_KERNACCESS|ATTR_SECURE))) {
+		if ((error = xfs_iaccess(ip, S_IRUSR, cred)))
 			return(XFS_ERROR(error));
 	}
-	}
 
 	/*
 	 * Fill in the arg structure for this request.
@@ -167,9 +159,6 @@ xfs_attr_fetch(xfs_inode_t *ip, char *name, int namelen,
 		error = xfs_attr_node_get(&args);
 	}
 
-	if (!(flags & ATTR_KERNACCESS))
-		xfs_iunlock(ip, XFS_ILOCK_SHARED);
-
 	/*
 	 * Return the number of bytes in the value to the caller.
 	 */
@@ -185,7 +174,7 @@ xfs_attr_get(bhv_desc_t *bdp, char *name, char *value, int *valuelenp,
 	     int flags, struct cred *cred)
 {
 	xfs_inode_t	*ip = XFS_BHVTOI(bdp);
-	int		namelen;
+	int		error, namelen;
 
 	XFS_STATS_INC(xs_attr_get);
 
@@ -195,7 +184,13 @@ xfs_attr_get(bhv_desc_t *bdp, char *name, char *value, int *valuelenp,
 	if (namelen >= MAXNAMELEN)
 		return(EFAULT);		/* match IRIX behaviour */
 
-	return xfs_attr_fetch(ip, name, namelen, value, valuelenp, flags, cred);
+	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
+		return(EIO);
+
+	xfs_ilock(ip, XFS_ILOCK_SHARED);
+	error = xfs_attr_fetch(ip, name, namelen, value, valuelenp, flags, cred);
+	xfs_iunlock(ip, XFS_ILOCK_SHARED);
+	return(error);
 }
 
 /*ARGSUSED*/
@@ -718,16 +713,15 @@ xfs_attr_inactive(xfs_inode_t *dp)
 	mp = dp->i_mount;
 	ASSERT(! XFS_NOT_DQATTACHED(mp, dp));
 
-	/* XXXsup - why on earth are we taking ILOCK_EXCL here??? */
-	xfs_ilock(dp, XFS_ILOCK_EXCL);
+	xfs_ilock(dp, XFS_ILOCK_SHARED);
 	if ((XFS_IFORK_Q(dp) == 0) ||
 	    (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) ||
 	    (dp->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS &&
 	     dp->i_d.di_anextents == 0)) {
-		xfs_iunlock(dp, XFS_ILOCK_EXCL);
+		xfs_iunlock(dp, XFS_ILOCK_SHARED);
 		return(0);
 	}
-	xfs_iunlock(dp, XFS_ILOCK_EXCL);
+	xfs_iunlock(dp, XFS_ILOCK_SHARED);
 
 	/*
 	 * Start our first transaction of the day.

fs/xfs/xfs_dinode.h

@@ -457,6 +457,8 @@ xfs_dinode_t *xfs_buf_to_dinode(struct xfs_buf *bp);
 #define XFS_DIFLAG_NOATIME_BIT   6	/* do not update atime */
 #define XFS_DIFLAG_NODUMP_BIT    7	/* do not dump */
 #define XFS_DIFLAG_RTINHERIT_BIT 8	/* create with realtime bit set */
+#define XFS_DIFLAG_PROJINHERIT_BIT  9	/* create with parents projid */
+#define XFS_DIFLAG_NOSYMLINKS_BIT  10	/* disallow symlink creation */
 #define XFS_DIFLAG_REALTIME      (1 << XFS_DIFLAG_REALTIME_BIT)
 #define XFS_DIFLAG_PREALLOC      (1 << XFS_DIFLAG_PREALLOC_BIT)
 #define XFS_DIFLAG_NEWRTBM       (1 << XFS_DIFLAG_NEWRTBM_BIT)
@@ -466,10 +468,13 @@ xfs_dinode_t *xfs_buf_to_dinode(struct xfs_buf *bp);
 #define XFS_DIFLAG_NOATIME       (1 << XFS_DIFLAG_NOATIME_BIT)
 #define XFS_DIFLAG_NODUMP        (1 << XFS_DIFLAG_NODUMP_BIT)
 #define XFS_DIFLAG_RTINHERIT     (1 << XFS_DIFLAG_RTINHERIT_BIT)
+#define XFS_DIFLAG_PROJINHERIT   (1 << XFS_DIFLAG_PROJINHERIT_BIT)
+#define XFS_DIFLAG_NOSYMLINKS    (1 << XFS_DIFLAG_NOSYMLINKS_BIT)
 
 #define XFS_DIFLAG_ANY \
 	(XFS_DIFLAG_REALTIME | XFS_DIFLAG_PREALLOC | XFS_DIFLAG_NEWRTBM | \
 	 XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND | XFS_DIFLAG_SYNC | \
-	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT)
+	 XFS_DIFLAG_NOATIME | XFS_DIFLAG_NODUMP | XFS_DIFLAG_RTINHERIT | \
+	 XFS_DIFLAG_PROJINHERIT | XFS_DIFLAG_NOSYMLINKS)
 
 #endif	/* __XFS_DINODE_H__ */

fs/xfs/xfs_fs.h

@@ -77,6 +77,8 @@ struct fsxattr {
 #define XFS_XFLAG_NOATIME	0x00000040	/* do not update access time */
 #define XFS_XFLAG_NODUMP	0x00000080	/* do not include in backups */
 #define XFS_XFLAG_RTINHERIT	0x00000100	/* create with rt bit set */
+#define XFS_XFLAG_PROJINHERIT	0x00000200	/* create with parents projid */
+#define XFS_XFLAG_NOSYMLINKS	0x00000400	/* disallow symlink creation */
 #define XFS_XFLAG_HASATTR	0x80000000	/* no DIFLAG for this	*/
 
 /*

fs/xfs/xfs_iget.c

@@ -457,7 +457,7 @@ xfs_iget(
 			error = xfs_iget_core(vp, mp, tp, ino,
 						lock_flags, ipp, bno);
 			if (error) {
-				make_bad_inode(inode);
+				vn_mark_bad(vp);
 				if (inode->i_state & I_NEW)
 					unlock_new_inode(inode);
 				iput(inode);
@@ -576,11 +576,8 @@ xfs_iput_new(xfs_inode_t	*ip,
 
 	vn_trace_entry(vp, "xfs_iput_new", (inst_t *)__return_address);
 
-	/* We shouldn't get here without this being true, but just in case */
-	if (inode->i_state & I_NEW) {
-		make_bad_inode(inode);
+	if (inode->i_state & I_NEW)
 		unlock_new_inode(inode);
-	}
 	if (lock_flags)
 		xfs_iunlock(ip, lock_flags);
 	VN_RELE(vp);

fs/xfs/xfs_inode.c

@@ -881,6 +881,10 @@ xfs_dic2xflags(
 			flags |= XFS_XFLAG_NODUMP;
 		if (di_flags & XFS_DIFLAG_RTINHERIT)
 			flags |= XFS_XFLAG_RTINHERIT;
+		if (di_flags & XFS_DIFLAG_PROJINHERIT)
+			flags |= XFS_XFLAG_PROJINHERIT;
+		if (di_flags & XFS_DIFLAG_NOSYMLINKS)
+			flags |= XFS_XFLAG_NOSYMLINKS;
 	}
 	return flags;
 }
@@ -1257,6 +1261,9 @@ xfs_ialloc(
 			if ((pip->i_d.di_flags & XFS_DIFLAG_SYNC) &&
 			    xfs_inherit_sync)
 				ip->i_d.di_flags |= XFS_DIFLAG_SYNC;
+			if ((pip->i_d.di_flags & XFS_DIFLAG_NOSYMLINKS) &&
+			    xfs_inherit_nosymlinks)
+				ip->i_d.di_flags |= XFS_DIFLAG_NOSYMLINKS;
 		}
 		/* FALLTHROUGH */
 	case S_IFLNK:

fs/xfs/xfs_rtalloc.c

@@ -1966,7 +1966,8 @@ xfs_growfs_rt(
 	/*
 	 * Calculate new parameters.  These are the final values to be reached.
 	 */
-	nrextents = do_div(nrblocks, in->extsize);
+	nrextents = nrblocks;
+	do_div(nrextents, in->extsize);
 	nrbmblocks = roundup_64(nrextents, NBBY * sbp->sb_blocksize);
 	nrextslog = xfs_highbit32(nrextents);
 	nrsumlevels = nrextslog + 1;
@@ -2021,7 +2022,8 @@ xfs_growfs_rt(
 			XFS_RTMIN(nrblocks,
 				  nsbp->sb_rbmblocks * NBBY *
 				  nsbp->sb_blocksize * nsbp->sb_rextsize);
-		nsbp->sb_rextents = do_div(nsbp->sb_rblocks, nsbp->sb_rextsize);
+		nsbp->sb_rextents = nsbp->sb_rblocks;
+		do_div(nsbp->sb_rextents, nsbp->sb_rextsize);
 		nsbp->sb_rextslog = xfs_highbit32(nsbp->sb_rextents);
 		nrsumlevels = nmp->m_rsumlevels = nsbp->sb_rextslog + 1;
 		nrsumsize =

fs/xfs/xfs_vfsops.c

@@ -1055,6 +1055,11 @@ xfs_sync_inodes(
 			continue;
 		}
 
+		if (VN_BAD(vp)) {
+			ip = ip->i_mnext;
+			continue;
+		}
+
 		if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
 			XFS_MOUNT_IUNLOCK(mp);
 			kmem_free(ipointer, sizeof(xfs_iptr_t));
@@ -1582,31 +1587,35 @@ xfs_vget(
 	vnode_t		**vpp,
 	fid_t		*fidp)
 {
-	xfs_fid_t	*xfid;
+	xfs_mount_t	*mp = XFS_BHVTOM(bdp);
+	xfs_fid_t	*xfid = (struct xfs_fid *)fidp;
 	xfs_inode_t	*ip;
 	int		error;
 	xfs_ino_t	ino;
 	unsigned int	igen;
-	xfs_mount_t	*mp;
 
-	xfid  = (struct xfs_fid *)fidp;
-	if (xfid->xfs_fid_len == sizeof(*xfid) - sizeof(xfid->xfs_fid_len)) {
+	/*
+	 * Invalid.  Since handles can be created in user space and passed in
+	 * via gethandle(), this is not cause for a panic.
+	 */
+	if (xfid->xfs_fid_len != sizeof(*xfid) - sizeof(xfid->xfs_fid_len))
+		return XFS_ERROR(EINVAL);
+
 	ino  = xfid->xfs_fid_ino;
 	igen = xfid->xfs_fid_gen;
-	} else {
+
 	/*
-		 * Invalid.  Since handles can be created in user space
-		 * and passed in via gethandle(), this is not cause for
-		 * a panic.
+	 * NFS can sometimes send requests for ino 0.  Fail them gracefully.
 	 */
-		return XFS_ERROR(EINVAL);
-	}
-	mp = XFS_BHVTOM(bdp);
+	if (ino == 0)
+		return XFS_ERROR(ESTALE);
+
 	error = xfs_iget(mp, NULL, ino, XFS_ILOCK_SHARED, &ip, 0);
 	if (error) {
 		*vpp = NULL;
 		return error;
 	}
+
 	if (ip == NULL) {
 		*vpp = NULL;
 		return XFS_ERROR(EIO);

fs/xfs/xfs_vnodeops.c

@@ -826,29 +826,34 @@ xfs_setattr(
 				mp->m_sb.sb_blocklog;
 		}
 		if (mask & XFS_AT_XFLAGS) {
+			uint	di_flags;
+
 			/* can't set PREALLOC this way, just preserve it */
-			ip->i_d.di_flags =
-				(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
-			if (vap->va_xflags & XFS_XFLAG_REALTIME &&
-			    (ip->i_d.di_mode & S_IFMT) == S_IFREG) {
-				ip->i_d.di_flags |= XFS_DIFLAG_REALTIME;
-				ip->i_iocore.io_flags |= XFS_IOCORE_RT;
-			} else {
-				ip->i_iocore.io_flags &= ~XFS_IOCORE_RT;
-			}
+			di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
 			if (vap->va_xflags & XFS_XFLAG_IMMUTABLE)
-				ip->i_d.di_flags |= XFS_DIFLAG_IMMUTABLE;
+				di_flags |= XFS_DIFLAG_IMMUTABLE;
 			if (vap->va_xflags & XFS_XFLAG_APPEND)
-				ip->i_d.di_flags |= XFS_DIFLAG_APPEND;
+				di_flags |= XFS_DIFLAG_APPEND;
 			if (vap->va_xflags & XFS_XFLAG_SYNC)
-				ip->i_d.di_flags |= XFS_DIFLAG_SYNC;
+				di_flags |= XFS_DIFLAG_SYNC;
 			if (vap->va_xflags & XFS_XFLAG_NOATIME)
-				ip->i_d.di_flags |= XFS_DIFLAG_NOATIME;
+				di_flags |= XFS_DIFLAG_NOATIME;
 			if (vap->va_xflags & XFS_XFLAG_NODUMP)
-				ip->i_d.di_flags |= XFS_DIFLAG_NODUMP;
-			if ((vap->va_xflags & XFS_XFLAG_RTINHERIT) &&
-			    (ip->i_d.di_mode & S_IFMT) == S_IFDIR)
-				ip->i_d.di_flags |= XFS_DIFLAG_RTINHERIT;
+				di_flags |= XFS_DIFLAG_NODUMP;
+			if ((ip->i_d.di_mode & S_IFMT) == S_IFDIR) {
+				if (vap->va_xflags & XFS_XFLAG_REALTIME) {
+					ip->i_iocore.io_flags |= XFS_IOCORE_RT;
+					di_flags |= XFS_DIFLAG_REALTIME;
+				}
+				if (vap->va_xflags & XFS_XFLAG_RTINHERIT)
+					di_flags |= XFS_DIFLAG_RTINHERIT;
+				if (vap->va_xflags & XFS_XFLAG_NOSYMLINKS)
+					di_flags |= XFS_DIFLAG_NOSYMLINKS;
+			} else {
+				if (!(vap->va_xflags & XFS_XFLAG_REALTIME))
+					ip->i_iocore.io_flags &= ~XFS_IOCORE_RT;
+			}
+			ip->i_d.di_flags = di_flags;
 		}
 		xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 		timeflags |= XFS_ICHGTIME_CHG;
@@ -1606,7 +1611,7 @@ xfs_inactive(
 	 * If the inode is already free, then there can be nothing
 	 * to clean up here.
 	 */
-	if (ip->i_d.di_mode == 0) {
+	if (ip->i_d.di_mode == 0 || VN_BAD(vp)) {
 		ASSERT(ip->i_df.if_real_bytes == 0);
 		ASSERT(ip->i_df.if_broot_bytes == 0);
 		return VN_INACTIVE_CACHE;
@@ -3389,6 +3394,14 @@ xfs_symlink(
 
 	xfs_ilock(dp, XFS_ILOCK_EXCL);
 
+	/*
+	 * Check whether the directory allows new symlinks or not.
+	 */
+	if (dp->i_d.di_flags & XFS_DIFLAG_NOSYMLINKS) {
+		error = XFS_ERROR(EPERM);
+		goto error_return;
+	}
+
 	/*
 	 * Reserve disk quota : blocks and inode.
 	 */
@@ -3795,11 +3808,17 @@ xfs_reclaim(
 	vnode_t		*vp;
 
 	vp = BHV_TO_VNODE(bdp);
+	ip = XFS_BHVTOI(bdp);
 
 	vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address);
 
 	ASSERT(!VN_MAPPED(vp));
-	ip = XFS_BHVTOI(bdp);
+
+	/* bad inode, get out here ASAP */
+	if (VN_BAD(vp)) {
+		xfs_ireclaim(ip);
+		return 0;
+	}
 
 	if ((ip->i_d.di_mode & S_IFMT) == S_IFREG) {
 		if (ip->i_d.di_size > 0) {
@@ -3877,8 +3896,12 @@ xfs_finish_reclaim(
 	int		sync_mode)
 {
 	xfs_ihash_t	*ih = ip->i_hash;
+	vnode_t		*vp = XFS_ITOV_NULL(ip);
 	int		error;
 
+	if (vp && VN_BAD(vp))
+		return 0;
+
 	/* The hash lock here protects a thread in xfs_iget_core from
 	 * racing with us on linking the inode back with a vnode.
 	 * Once we have the XFS_IRECLAIM flag set it will not touch
@@ -3886,8 +3909,7 @@ xfs_finish_reclaim(
 	 */
 	write_lock(&ih->ih_lock);
 	if ((ip->i_flags & XFS_IRECLAIM) ||
-	    (!(ip->i_flags & XFS_IRECLAIMABLE) &&
-	      (XFS_ITOV_NULL(ip) == NULL))) {
+	    (!(ip->i_flags & XFS_IRECLAIMABLE) && vp == NULL)) {
 		write_unlock(&ih->ih_lock);
 		if (locked) {
 			xfs_ifunlock(ip);
@@ -3954,15 +3976,13 @@ int
 xfs_finish_reclaim_all(xfs_mount_t *mp, int noblock)
 {
 	int		purged;
-	struct list_head	*curr, *next;
-	xfs_inode_t	*ip;
+	xfs_inode_t	*ip, *n;
 	int		done = 0;
 
 	while (!done) {
 		purged = 0;
 		XFS_MOUNT_ILOCK(mp);
-		list_for_each_safe(curr, next, &mp->m_del_inodes) {
-			ip = list_entry(curr, xfs_inode_t, i_reclaim);
+		list_for_each_entry_safe(ip, n, &mp->m_del_inodes, i_reclaim) {
 			if (noblock) {
 				if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0)
 					continue;

include/linux/dqblk_xfs.h

 /*
- * Copyright (c) 1995-2001 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 1995-2001,2004 Silicon Graphics, Inc.  All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or modify it
  * under the terms of version 2.1 of the GNU Lesser 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.
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
  *
- * 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 to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307
+ * USA
  *
- * You should have received a copy of the GNU Lesser 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/
+ * Contact information: Silicon Graphics, Inc., 1500 Crittenden Lane,
+ * Mountain View, CA 94043, USA, or: http://www.sgi.com
  */
 #ifndef _LINUX_DQBLK_XFS_H
 #define _LINUX_DQBLK_XFS_H
@@ -40,12 +28,12 @@
  */
 
 #define XQM_CMD(x)	(('X'<<8)+(x))	/* note: forms first QCMD argument */
-#define Q_XQUOTAON	XQM_CMD(0x1)	/* enable accounting/enforcement */
-#define Q_XQUOTAOFF	XQM_CMD(0x2)	/* disable accounting/enforcement */
-#define Q_XGETQUOTA	XQM_CMD(0x3)	/* get disk limits and usage */
-#define Q_XSETQLIM	XQM_CMD(0x4)	/* set disk limits */
-#define Q_XGETQSTAT	XQM_CMD(0x5)	/* get quota subsystem status */
-#define Q_XQUOTARM	XQM_CMD(0x6)	/* free disk space used by dquots */
+#define Q_XQUOTAON	XQM_CMD(1)	/* enable accounting/enforcement */
+#define Q_XQUOTAOFF	XQM_CMD(2)	/* disable accounting/enforcement */
+#define Q_XGETQUOTA	XQM_CMD(3)	/* get disk limits and usage */
+#define Q_XSETQLIM	XQM_CMD(4)	/* set disk limits */
+#define Q_XGETQSTAT	XQM_CMD(5)	/* get quota subsystem status */
+#define Q_XQUOTARM	XQM_CMD(6)	/* free disk space used by dquots */
 
 /*
  * fs_disk_quota structure:
@@ -104,6 +92,19 @@ typedef struct fs_disk_quota {
 #define FS_DQ_RTBTIMER 	(1<<8)
 #define FS_DQ_TIMER_MASK	(FS_DQ_BTIMER | FS_DQ_ITIMER | FS_DQ_RTBTIMER)
 
+/*
+ * Warning counts are set in both super user's dquot and others. For others,
+ * warnings are set/cleared by the administrators (or automatically by going
+ * below the soft limit).  Superusers warning values set the warning limits
+ * for the rest.  In case these values are zero, the DQ_{F,B}WARNLIMIT values
+ * defined below are used. 
+ * These values also apply only to the d_fieldmask field for Q_XSETQLIM.
+ */
+#define FS_DQ_BWARNS	(1<<9)
+#define FS_DQ_IWARNS	(1<<10)
+#define FS_DQ_RTBWARNS	(1<<11)
+#define FS_DQ_WARNS_MASK	(FS_DQ_BWARNS | FS_DQ_IWARNS | FS_DQ_RTBWARNS)
+
 /*
  * Various flags related to quotactl(2).  Only relevant to XFS filesystems.
  */
@@ -111,9 +112,11 @@ typedef struct fs_disk_quota {
 #define XFS_QUOTA_UDQ_ENFD	(1<<1)  /* user quota limits enforcement */
 #define XFS_QUOTA_GDQ_ACCT	(1<<2)  /* group quota accounting */
 #define XFS_QUOTA_GDQ_ENFD	(1<<3)  /* group quota limits enforcement */
+#define XFS_QUOTA_PDQ_ACCT	(1<<4)  /* project quota accounting */
+#define XFS_QUOTA_PDQ_ENFD	(1<<5)  /* project quota limits enforcement */
 
 #define XFS_USER_QUOTA		(1<<0)	/* user quota type */
-#define XFS_PROJ_QUOTA		(1<<1)	/* (IRIX) project quota type */
+#define XFS_PROJ_QUOTA		(1<<1)	/* project quota type */
 #define XFS_GROUP_QUOTA		(1<<2)	/* group quota type */
 
 /*