Merge with /usr/src/ntfs-2.6.git

20 files changed, 264 insertions, 114 deletions

diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 554e4a18c152..d3ff78354638 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -49,7 +49,7 @@ ifeq ($(CONFIG_XFS_TRACE),y)
         EXTRA_CFLAGS += -DXFS_LOG_TRACE
         EXTRA_CFLAGS += -DXFS_RW_TRACE
         EXTRA_CFLAGS += -DPAGEBUF_TRACE
-        # EXTRA_CFLAGS += -DXFS_VNODE_TRACE
+        EXTRA_CFLAGS += -DXFS_VNODE_TRACE
 endif
 
 obj-$(CONFIG_XFS_FS)            += xfs.o
diff --git a/fs/xfs/linux-2.6/xfs_aops.c b/fs/xfs/linux-2.6/xfs_aops.c
index 76a84758073a..9278e9aba9ba 100644
--- a/fs/xfs/linux-2.6/xfs_aops.c
+++ b/fs/xfs/linux-2.6/xfs_aops.c
@@ -558,7 +558,8 @@ xfs_submit_page(
         int                     i;
 
         BUG_ON(PageWriteback(page));
-        set_page_writeback(page);
+        if (bh_count)
+                set_page_writeback(page);
         if (clear_dirty)
                 clear_page_dirty(page);
         unlock_page(page);
@@ -578,9 +579,6 @@ xfs_submit_page(
 
                 if (probed_page && clear_dirty)
                         wbc->nr_to_write--;     /* Wrote an "extra" page */
-        } else {
-                end_page_writeback(page);
-                wbc->pages_skipped++;   /* We didn't write this page */
         }
 }
 
@@ -602,21 +600,26 @@ xfs_convert_page(
 {
         struct buffer_head      *bh_arr[MAX_BUF_PER_PAGE], *bh, *head;
         xfs_iomap_t             *mp = iomapp, *tmp;
-        unsigned long           end, offset;
-        pgoff_t                 end_index;
-        int                     i = 0, index = 0;
+        unsigned long           offset, end_offset;
+        int                     index = 0;
         int                     bbits = inode->i_blkbits;
+        int                     len, page_dirty;
 
-        end_index = i_size_read(inode) >> PAGE_CACHE_SHIFT;
-        if (page->index < end_index) {
-                end = PAGE_CACHE_SIZE;
-        } else {
-                end = i_size_read(inode) & (PAGE_CACHE_SIZE-1);
-        }
+        end_offset = (i_size_read(inode) & (PAGE_CACHE_SIZE - 1));
+
+        /*
+         * page_dirty is initially a count of buffers on the page before
+         * EOF and is decrememted as we move each into a cleanable state.
+         */
+        len = 1 << inode->i_blkbits;
+        end_offset = max(end_offset, PAGE_CACHE_SIZE);
+        end_offset = roundup(end_offset, len);
+        page_dirty = end_offset / len;
+
+        offset = 0;
         bh = head = page_buffers(page);
         do {
-                offset = i << bbits;
-                if (offset >= end)
+                if (offset >= end_offset)
                         break;
                 if (!(PageUptodate(page) || buffer_uptodate(bh)))
                         continue;
@@ -625,6 +628,7 @@ xfs_convert_page(
                         if (startio) {
                                 lock_buffer(bh);
                                 bh_arr[index++] = bh;
+                                page_dirty--;
                         }
                         continue;
                 }
@@ -657,10 +661,11 @@ xfs_convert_page(
                         unlock_buffer(bh);
                         mark_buffer_dirty(bh);
                 }
-        } while (i++, (bh = bh->b_this_page) != head);
+                page_dirty--;
+        } while (offset += len, (bh = bh->b_this_page) != head);
 
-        if (startio) {
-                xfs_submit_page(page, wbc, bh_arr, index, 1, index == i);
+        if (startio && index) {
+                xfs_submit_page(page, wbc, bh_arr, index, 1, !page_dirty);
         } else {
                 unlock_page(page);
         }
@@ -725,8 +730,11 @@ xfs_page_state_convert(
         __uint64_t              end_offset;
         pgoff_t                 end_index, last_index, tlast;
         int                     len, err, i, cnt = 0, uptodate = 1;
-        int                     flags = startio ? 0 : BMAPI_TRYLOCK;
-        int                     page_dirty, delalloc = 0;
+        int                     flags;
+        int                     page_dirty;
+
+        /* wait for other IO threads? */
+        flags = (startio && wbc->sync_mode != WB_SYNC_NONE) ? 0 : BMAPI_TRYLOCK;
 
         /* Is this page beyond the end of the file? */
         offset = i_size_read(inode);
@@ -740,19 +748,22 @@ xfs_page_state_convert(
                 }
         }
 
-        offset = (loff_t)page->index << PAGE_CACHE_SHIFT;
         end_offset = min_t(unsigned long long,
-                        offset + PAGE_CACHE_SIZE, i_size_read(inode));
-
-        bh = head = page_buffers(page);
-        iomp = NULL;
+                        (loff_t)(page->index + 1) << PAGE_CACHE_SHIFT, offset);
+        offset = (loff_t)page->index << PAGE_CACHE_SHIFT;
 
         /*
-         * page_dirty is initially a count of buffers on the page and
-         * is decrememted as we move each into a cleanable state.
+         * page_dirty is initially a count of buffers on the page before
+         * EOF and is decrememted as we move each into a cleanable state.
          */
-        len = bh->b_size;
-        page_dirty = PAGE_CACHE_SIZE / len;
+        len = 1 << inode->i_blkbits;
+        p_offset = max(p_offset, PAGE_CACHE_SIZE);
+        p_offset = roundup(p_offset, len);
+        page_dirty = p_offset / len;
+
+        iomp = NULL;
+        p_offset = 0;
+        bh = head = page_buffers(page);
 
         do {
                 if (offset >= end_offset)
@@ -804,7 +815,6 @@ xfs_page_state_convert(
                  */
                 } else if (buffer_delay(bh)) {
                         if (!iomp) {
-                                delalloc = 1;
                                 err = xfs_map_blocks(inode, offset, len, &iomap,
                                                 BMAPI_ALLOCATE | flags);
                                 if (err) {
@@ -875,14 +885,15 @@ xfs_page_state_convert(
         if (uptodate && bh == head)
                 SetPageUptodate(page);
 
-        if (startio)
-                xfs_submit_page(page, wbc, bh_arr, cnt, 0, 1);
+        if (startio) {
+                WARN_ON(page_dirty);
+                xfs_submit_page(page, wbc, bh_arr, cnt, 0, !page_dirty);
+        }
 
         if (iomp) {
-                tlast = (iomp->iomap_offset + iomp->iomap_bsize - 1) >>
+                offset = (iomp->iomap_offset + iomp->iomap_bsize - 1) >>
                                         PAGE_CACHE_SHIFT;
-                if (delalloc && (tlast > last_index))
-                        tlast = last_index;
+                tlast = min_t(pgoff_t, offset, last_index);
                 xfs_cluster_write(inode, page->index + 1, iomp, wbc,
                                         startio, unmapped, tlast);
         }
diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c
index 23e0eb67fc25..997963e53622 100644
--- a/fs/xfs/linux-2.6/xfs_buf.c
+++ b/fs/xfs/linux-2.6/xfs_buf.c
@@ -1746,13 +1746,15 @@ STATIC DECLARE_COMPLETION(pagebuf_daemon_done);
 STATIC struct task_struct *pagebuf_daemon_task;
 STATIC int pagebuf_daemon_active;
 STATIC int force_flush;
-
+STATIC int force_sleep;
 
 STATIC int
 pagebuf_daemon_wakeup(
         int                     priority,
         unsigned int            mask)
 {
+        if (force_sleep)
+                return 0;
         force_flush = 1;
         barrier();
         wake_up_process(pagebuf_daemon_task);
@@ -1778,7 +1780,12 @@ pagebuf_daemon(
 
         INIT_LIST_HEAD(&tmp);
         do {
-                try_to_freeze(PF_FREEZE);
+                if (unlikely(current->flags & PF_FREEZE)) {
+                        force_sleep = 1;
+                        refrigerator(PF_FREEZE);
+                } else {
+                        force_sleep = 0;
+                }
 
                 set_current_state(TASK_INTERRUPTIBLE);
                 schedule_timeout((xfs_buf_timer_centisecs * HZ) / 100);
diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c
index 9f057a4a5b06..d0d412afd261 100644
--- a/fs/xfs/linux-2.6/xfs_file.c
+++ b/fs/xfs/linux-2.6/xfs_file.c
@@ -515,10 +515,49 @@ 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          = do_sync_write,
+        .write          = linvfs_write,
         .readv          = linvfs_readv,
         .writev         = linvfs_writev,
         .aio_read       = linvfs_aio_read,
@@ -540,7 +579,7 @@ struct file_operations linvfs_file_operations = {
 struct file_operations linvfs_invis_file_operations = {
         .llseek         = generic_file_llseek,
         .read           = do_sync_read,
-        .write          = do_sync_write,
+        .write          = linvfs_write_invis,
         .readv          = linvfs_readv_invis,
         .writev         = linvfs_writev_invis,
         .aio_read       = linvfs_aio_read_invis,
diff --git a/fs/xfs/linux-2.6/xfs_lrw.c b/fs/xfs/linux-2.6/xfs_lrw.c
index ff145fd0d1a4..aa9daaea6c34 100644
--- a/fs/xfs/linux-2.6/xfs_lrw.c
+++ b/fs/xfs/linux-2.6/xfs_lrw.c
@@ -683,6 +683,9 @@ 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);
 
diff --git a/fs/xfs/linux-2.6/xfs_vnode.c b/fs/xfs/linux-2.6/xfs_vnode.c
index 849c61c74f3c..a832d165f24f 100644
--- a/fs/xfs/linux-2.6/xfs_vnode.c
+++ b/fs/xfs/linux-2.6/xfs_vnode.c
@@ -156,7 +156,6 @@ vn_initialize(
 
 #ifdef  XFS_VNODE_TRACE
         vp->v_trace = ktrace_alloc(VNODE_TRACE_SIZE, KM_SLEEP);
-        printk("Allocated VNODE_TRACE at 0x%p\n", vp->v_trace);
 #endif  /* XFS_VNODE_TRACE */
 
         vn_trace_exit(vp, "vn_initialize", (inst_t *)__return_address);
@@ -424,13 +423,13 @@ vn_remove(
  * Vnode tracing code.
  */
 void
-vn_trace_entry(vnode_t *vp, char *func, inst_t *ra)
+vn_trace_entry(vnode_t *vp, const char *func, inst_t *ra)
 {
         KTRACE_ENTER(vp, VNODE_KTRACE_ENTRY, func, 0, ra);
 }
 
 void
-vn_trace_exit(vnode_t *vp, char *func, inst_t *ra)
+vn_trace_exit(vnode_t *vp, const char *func, inst_t *ra)
 {
         KTRACE_ENTER(vp, VNODE_KTRACE_EXIT, func, 0, ra);
 }
diff --git a/fs/xfs/linux-2.6/xfs_vnode.h b/fs/xfs/linux-2.6/xfs_vnode.h
index da76c1f1e11c..00466c3194ac 100644
--- a/fs/xfs/linux-2.6/xfs_vnode.h
+++ b/fs/xfs/linux-2.6/xfs_vnode.h
@@ -86,10 +86,11 @@ typedef struct vnode {
         vnumber_t       v_number;               /* in-core vnode number */
         vn_bhv_head_t   v_bh;                   /* behavior head */
         spinlock_t      v_lock;                 /* VN_LOCK/VN_UNLOCK */
-        struct inode    v_inode;                /* Linux inode */
 #ifdef XFS_VNODE_TRACE
         struct ktrace   *v_trace;               /* trace header structure    */
 #endif
+        struct inode    v_inode;                /* Linux inode */
+        /* inode MUST be last */
 } vnode_t;
 
 #define v_fbhv                  v_bh.bh_first          /* first behavior */
@@ -409,7 +410,7 @@ 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 */
-        nlink_t         va_nlink;       /* number of references to file */
+        xfs_nlink_t     va_nlink;       /* number of references to file */
         uid_t           va_uid;         /* owner user id */
         gid_t           va_gid;         /* owner group id */
         xfs_ino_t       va_nodeid;      /* file id */
@@ -625,6 +626,7 @@ static inline int VN_BAD(struct vnode *vp)
 #define ATTR_DMI        0x08    /* invocation from a DMI function */
 #define ATTR_LAZY       0x80    /* set/get attributes lazily */
 #define ATTR_NONBLOCK   0x100   /* return EAGAIN if operation would block */
+#define ATTR_NOLOCK     0x200   /* Don't grab any conflicting locks */
 
 /*
  * Flags to VOP_FSYNC and VOP_RECLAIM.
@@ -646,8 +648,8 @@ static inline int VN_BAD(struct vnode *vp)
 #define VNODE_KTRACE_REF        4
 #define VNODE_KTRACE_RELE       5
 
-extern void vn_trace_entry(struct vnode *, char *, inst_t *);
-extern void vn_trace_exit(struct vnode *, char *, inst_t *);
+extern void vn_trace_entry(struct vnode *, const char *, inst_t *);
+extern void vn_trace_exit(struct vnode *, const char *, inst_t *);
 extern void vn_trace_hold(struct vnode *, char *, int, inst_t *);
 extern void vn_trace_ref(struct vnode *, char *, int, inst_t *);
 extern void vn_trace_rele(struct vnode *, char *, int, inst_t *);
diff --git a/fs/xfs/xfs_dfrag.c b/fs/xfs/xfs_dfrag.c
index 08d551a17347..63abdc2ac7f4 100644
--- a/fs/xfs/xfs_dfrag.c
+++ b/fs/xfs/xfs_dfrag.c
@@ -182,7 +182,7 @@ xfs_swapext(
 
         if (VN_CACHED(tvp) != 0)
                 xfs_inval_cached_pages(XFS_ITOV(tip), &(tip->i_iocore),
-                                                (loff_t)0, 0, 0);
+                                                (xfs_off_t)0, 0, 0);
 
         /* Verify O_DIRECT for ftmp */
         if (VN_CACHED(tvp) != 0) {
diff --git a/fs/xfs/xfs_iget.c b/fs/xfs/xfs_iget.c
index 3a0ba1dfd0e8..d3da00045f26 100644
--- a/fs/xfs/xfs_iget.c
+++ b/fs/xfs/xfs_iget.c
@@ -136,6 +136,40 @@ xfs_chash_free(xfs_mount_t *mp)
 }
 
 /*
+ * Try to move an inode to the front of its hash list if possible
+ * (and if its not there already).  Called right after obtaining
+ * the list version number and then dropping the read_lock on the
+ * hash list in question (which is done right after looking up the
+ * inode in question...).
+ */
+STATIC void
+xfs_ihash_promote(
+        xfs_ihash_t     *ih,
+        xfs_inode_t     *ip,
+        ulong           version)
+{
+        xfs_inode_t     *iq;
+
+        if ((ip->i_prevp != &ih->ih_next) && write_trylock(&ih->ih_lock)) {
+                if (likely(version == ih->ih_version)) {
+                        /* remove from list */
+                        if ((iq = ip->i_next)) {
+                                iq->i_prevp = ip->i_prevp;
+                        }
+                        *ip->i_prevp = iq;
+
+                        /* insert at list head */
+                        iq = ih->ih_next;
+                        iq->i_prevp = &ip->i_next;
+                        ip->i_next = iq;
+                        ip->i_prevp = &ih->ih_next;
+                        ih->ih_next = ip;
+                }
+                write_unlock(&ih->ih_lock);
+        }
+}
+
+/*
  * Look up an inode by number in the given file system.
  * The inode is looked up in the hash table for the file system
  * represented by the mount point parameter mp.  Each bucket of
@@ -229,7 +263,9 @@ again:
                                 XFS_STATS_INC(xs_ig_found);
 
                                 ip->i_flags &= ~XFS_IRECLAIMABLE;
+                                version = ih->ih_version;
                                 read_unlock(&ih->ih_lock);
+                                xfs_ihash_promote(ih, ip, version);
 
                                 XFS_MOUNT_ILOCK(mp);
                                 list_del_init(&ip->i_reclaim);
@@ -259,8 +295,15 @@ again:
                                                 inode_vp, vp);
                         }
 
+                        /*
+                         * Inode cache hit: if ip is not at the front of
+                         * its hash chain, move it there now.
+                         * Do this with the lock held for update, but
+                         * do statistics after releasing the lock.
+                         */
+                        version = ih->ih_version;
                         read_unlock(&ih->ih_lock);
-
+                        xfs_ihash_promote(ih, ip, version);
                         XFS_STATS_INC(xs_ig_found);
 
 finish_inode:
@@ -547,6 +590,7 @@ xfs_inode_incore(xfs_mount_t	*mp,
 {
         xfs_ihash_t     *ih;
         xfs_inode_t     *ip;
+        ulong           version;
 
         ih = XFS_IHASH(mp, ino);
         read_lock(&ih->ih_lock);
@@ -554,11 +598,15 @@ xfs_inode_incore(xfs_mount_t	*mp,
                 if (ip->i_ino == ino) {
                         /*
                          * If we find it and tp matches, return it.
+                         * Also move it to the front of the hash list
+                         * if we find it and it is not already there.
                          * Otherwise break from the loop and return
                          * NULL.
                          */
                         if (ip->i_transp == tp) {
+                                version = ih->ih_version;
                                 read_unlock(&ih->ih_lock);
+                                xfs_ihash_promote(ih, ip, version);
                                 return (ip);
                         }
                         break;
@@ -685,6 +733,7 @@ xfs_iextract(
                 iq->i_prevp = ip->i_prevp;
         }
         *ip->i_prevp = iq;
+        ih->ih_version++;
         write_unlock(&ih->ih_lock);
 
         /*
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 43c632ab86ad..bc8c8c7f9039 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -1130,7 +1130,7 @@ xfs_ialloc(
         xfs_trans_t     *tp,
         xfs_inode_t     *pip,
         mode_t          mode,
-        nlink_t         nlink,
+        xfs_nlink_t     nlink,
         xfs_dev_t       rdev,
         cred_t          *cr,
         xfs_prid_t      prid,
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index a53b1ccf6070..37e1c316f3b6 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -495,9 +495,9 @@ int		xfs_itobp(struct xfs_mount *, struct xfs_trans *,
 int             xfs_iread(struct xfs_mount *, struct xfs_trans *, xfs_ino_t,
                           xfs_inode_t **, xfs_daddr_t);
 int             xfs_iread_extents(struct xfs_trans *, xfs_inode_t *, int);
-int             xfs_ialloc(struct xfs_trans *, xfs_inode_t *, mode_t, nlink_t,
-                           xfs_dev_t, struct cred *, xfs_prid_t, int,
-                           struct xfs_buf **, boolean_t *, xfs_inode_t **);
+int             xfs_ialloc(struct xfs_trans *, xfs_inode_t *, mode_t,
+                           xfs_nlink_t, xfs_dev_t, struct cred *, xfs_prid_t,
+                           int, struct xfs_buf **, boolean_t *, xfs_inode_t **);
 void            xfs_xlate_dinode_core(xfs_caddr_t, struct xfs_dinode_core *,
                                         int);
 uint            xfs_ip2xflags(struct xfs_inode *);
diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
index 3826e8f0e28a..991f8a61f7c4 100644
--- a/fs/xfs/xfs_iomap.c
+++ b/fs/xfs/xfs_iomap.c
@@ -308,7 +308,8 @@ phase2:
                         break;
                 }
 
-                error = XFS_IOMAP_WRITE_ALLOCATE(mp, io, &imap, &nimaps);
+                error = XFS_IOMAP_WRITE_ALLOCATE(mp, io, offset, count,
+                                                 &imap, &nimaps);
                 break;
         case BMAPI_UNWRITTEN:
                 lockmode = 0;
@@ -365,7 +366,7 @@ xfs_flush_space(
 int
 xfs_iomap_write_direct(
         xfs_inode_t     *ip,
-        loff_t          offset,
+        xfs_off_t       offset,
         size_t          count,
         int             flags,
         xfs_bmbt_irec_t *ret_imap,
@@ -541,7 +542,7 @@ error_out:
 int
 xfs_iomap_write_delay(
         xfs_inode_t     *ip,
-        loff_t          offset,
+        xfs_off_t       offset,
         size_t          count,
         int             ioflag,
         xfs_bmbt_irec_t *ret_imap,
@@ -746,6 +747,8 @@ write_map:
 int
 xfs_iomap_write_allocate(
         xfs_inode_t     *ip,
+        xfs_off_t       offset,
+        size_t          count,
         xfs_bmbt_irec_t *map,
         int             *retmap)
 {
@@ -770,9 +773,9 @@ xfs_iomap_write_allocate(
         if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
                 return XFS_ERROR(error);
 
-        offset_fsb = map->br_startoff;
+        offset_fsb = XFS_B_TO_FSBT(mp, offset);
         count_fsb = map->br_blockcount;
-        map_start_fsb = offset_fsb;
+        map_start_fsb = map->br_startoff;
 
         XFS_STATS_ADD(xs_xstrat_bytes, XFS_FSB_TO_B(mp, count_fsb));
 
@@ -868,9 +871,9 @@ xfs_iomap_write_allocate(
                                         imap[i].br_startoff,
                                         imap[i].br_blockcount,imap[i].br_state);
                         }
-                        if ((map->br_startoff >= imap[i].br_startoff) &&
-                            (map->br_startoff < (imap[i].br_startoff +
-                                                 imap[i].br_blockcount))) {
+                        if ((offset_fsb >= imap[i].br_startoff) &&
+                            (offset_fsb < (imap[i].br_startoff +
+                                           imap[i].br_blockcount))) {
                                 *map = imap[i];
                                 *retmap = 1;
                                 XFS_STATS_INC(xs_xstrat_quick);
@@ -883,9 +886,8 @@ xfs_iomap_write_allocate(
                  * file, just surrounding data, try again.
                  */
                 nimaps--;
-                offset_fsb = imap[nimaps].br_startoff +
-                             imap[nimaps].br_blockcount;
-                map_start_fsb = offset_fsb;
+                map_start_fsb = imap[nimaps].br_startoff +
+                                imap[nimaps].br_blockcount;
         }
 
 trans_cancel:
@@ -899,7 +901,7 @@ error0:
 int
 xfs_iomap_write_unwritten(
         xfs_inode_t     *ip,
-        loff_t          offset,
+        xfs_off_t       offset,
         size_t          count)
 {
         xfs_mount_t     *mp = ip->i_mount;
diff --git a/fs/xfs/xfs_iomap.h b/fs/xfs/xfs_iomap.h
index 31c91087cb33..4daaa5212102 100644
--- a/fs/xfs/xfs_iomap.h
+++ b/fs/xfs/xfs_iomap.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003,2004 Silicon Graphics, Inc.  All Rights Reserved.
+ * Copyright (c) 2003-2005 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
@@ -29,9 +29,6 @@
  *
  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
  */
-
-
-
 #ifndef __XFS_IOMAP_H__
 #define __XFS_IOMAP_H__
 
@@ -56,7 +53,7 @@ typedef enum {
         BMAPI_UNWRITTEN  = (1 << 3),    /* unwritten extents to real extents */
         /* modifiers */
         BMAPI_IGNSTATE = (1 << 4),      /* ignore unwritten state on read */
-        BMAPI_DIRECT = (1 << 5),                /* direct instead of buffered write */
+        BMAPI_DIRECT = (1 << 5),        /* direct instead of buffered write */
         BMAPI_MMAP = (1 << 6),          /* allocate for mmap write */
         BMAPI_SYNC = (1 << 7),          /* sync write to flush delalloc space */
         BMAPI_TRYLOCK = (1 << 8),       /* non-blocking request */
@@ -67,13 +64,13 @@ typedef enum {
 /*
  * xfs_iomap_t:  File system I/O map
  *
- * The iomap_bn field is expressed in 512-byte blocks, and is where the 
+ * The iomap_bn field is expressed in 512-byte blocks, and is where the
  * mapping starts on disk.
  *
  * The iomap_offset, iomap_bsize and iomap_delta fields are in bytes.
  * iomap_offset is the offset of the mapping in the file itself.
- * iomap_bsize is the size of the mapping,  iomap_delta is the 
- * desired data's offset into the mapping, given the offset supplied 
+ * iomap_bsize is the size of the mapping,  iomap_delta is the
+ * desired data's offset into the mapping, given the offset supplied
  * to the file I/O map routine.
  *
  * When a request is made to read beyond the logical end of the object,
@@ -84,8 +81,8 @@ typedef enum {
 typedef struct xfs_iomap {
         xfs_daddr_t             iomap_bn;       /* first 512b blk of mapping */
         xfs_buftarg_t           *iomap_target;
-        loff_t                  iomap_offset;   /* offset of mapping, bytes */
-        loff_t                  iomap_bsize;    /* size of mapping, bytes */
+        xfs_off_t               iomap_offset;   /* offset of mapping, bytes */
+        xfs_off_t               iomap_bsize;    /* size of mapping, bytes */
         size_t                  iomap_delta;    /* offset into mapping, bytes */
         iomap_flags_t           iomap_flags;
 } xfs_iomap_t;
@@ -96,12 +93,12 @@ struct xfs_bmbt_irec;
 
 extern int xfs_iomap(struct xfs_iocore *, xfs_off_t, ssize_t, int,
                      struct xfs_iomap *, int *);
-extern int xfs_iomap_write_direct(struct xfs_inode *, loff_t, size_t,
+extern int xfs_iomap_write_direct(struct xfs_inode *, xfs_off_t, size_t,
                                   int, struct xfs_bmbt_irec *, int *, int);
-extern int xfs_iomap_write_delay(struct xfs_inode *, loff_t, size_t, int,
+extern int xfs_iomap_write_delay(struct xfs_inode *, xfs_off_t, size_t, int,
                                  struct xfs_bmbt_irec *, int *);
-extern int xfs_iomap_write_allocate(struct xfs_inode *,
+extern int xfs_iomap_write_allocate(struct xfs_inode *, xfs_off_t, size_t,
                                 struct xfs_bmbt_irec *, int *);
-extern int xfs_iomap_write_unwritten(struct xfs_inode *, loff_t, size_t);
+extern int xfs_iomap_write_unwritten(struct xfs_inode *, xfs_off_t, size_t);
 
 #endif /* __XFS_IOMAP_H__*/
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index b57423caef9b..2ec967d93e5a 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -301,6 +301,15 @@ xfs_mount_validate_sb(
         }
 
         /*
+         * Version 1 directory format has never worked on Linux.
+         */
+        if (unlikely(!XFS_SB_VERSION_HASDIRV2(sbp))) {
+                cmn_err(CE_WARN,
+        "XFS: Attempted to mount file system using version 1 directory format");
+                return XFS_ERROR(ENOSYS);
+        }
+
+        /*
          * Until this is fixed only page-sized or smaller data blocks work.
          */
         if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 5fc6201dd8e2..30dd08fb9f57 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -210,15 +210,16 @@ typedef int		(*xfs_bmapi_t)(struct xfs_trans *, void *,
                                 struct xfs_bmap_free *);
 typedef int             (*xfs_bmap_eof_t)(void *, xfs_fileoff_t, int, int *);
 typedef int             (*xfs_iomap_write_direct_t)(
-                                void *, loff_t, size_t, int,
+                                void *, xfs_off_t, size_t, int,
                                 struct xfs_bmbt_irec *, int *, int);
 typedef int             (*xfs_iomap_write_delay_t)(
-                                void *, loff_t, size_t, int,
+                                void *, xfs_off_t, size_t, int,
                                 struct xfs_bmbt_irec *, int *);
 typedef int             (*xfs_iomap_write_allocate_t)(
-                                void *, struct xfs_bmbt_irec *, int *);
+                                void *, xfs_off_t, size_t,
+                                struct xfs_bmbt_irec *, int *);
 typedef int             (*xfs_iomap_write_unwritten_t)(
-                                void *, loff_t, size_t);
+                                void *, xfs_off_t, size_t);
 typedef uint            (*xfs_lck_map_shared_t)(void *);
 typedef void            (*xfs_lock_t)(void *, uint);
 typedef void            (*xfs_lock_demote_t)(void *, uint);
@@ -258,9 +259,9 @@ typedef struct xfs_ioops {
 #define XFS_IOMAP_WRITE_DELAY(mp, io, offset, count, flags, mval, nmap) \
         (*(mp)->m_io_ops.xfs_iomap_write_delay) \
                 ((io)->io_obj, offset, count, flags, mval, nmap)
-#define XFS_IOMAP_WRITE_ALLOCATE(mp, io, mval, nmap) \
+#define XFS_IOMAP_WRITE_ALLOCATE(mp, io, offset, count, mval, nmap) \
         (*(mp)->m_io_ops.xfs_iomap_write_allocate) \
-                ((io)->io_obj, mval, nmap)
+                ((io)->io_obj, offset, count, mval, nmap)
 #define XFS_IOMAP_WRITE_UNWRITTEN(mp, io, offset, count) \
         (*(mp)->m_io_ops.xfs_iomap_write_unwritten) \
                 ((io)->io_obj, offset, count)
@@ -428,10 +429,10 @@ typedef struct xfs_mount {
 #define XFS_WRITEIO_LOG_LARGE   16
 
 /*
- * Max and min values for UIO and mount-option defined I/O sizes;
- * min value can't be less than a page.  Currently unused.
+ * Max and min values for mount-option defined I/O
+ * preallocation sizes.
  */
-#define XFS_MAX_IO_LOG          16      /* 64K */
+#define XFS_MAX_IO_LOG          30      /* 1G */
 #define XFS_MIN_IO_LOG          PAGE_SHIFT
 
 /*
diff --git a/fs/xfs/xfs_types.h b/fs/xfs/xfs_types.h
index 04609d27ea51..e4bf711e48ff 100644
--- a/fs/xfs/xfs_types.h
+++ b/fs/xfs/xfs_types.h
@@ -63,6 +63,7 @@ typedef __u64			xfs_ino_t;	/* <inode> type */
 typedef __s64                   xfs_daddr_t;    /* <disk address> type */
 typedef char *                  xfs_caddr_t;    /* <core address> type */
 typedef __u32                   xfs_dev_t;
+typedef __u32                   xfs_nlink_t;
 
 /* __psint_t is the same size as a pointer */
 #if (BITS_PER_LONG == 32)
diff --git a/fs/xfs/xfs_utils.c b/fs/xfs/xfs_utils.c
index 816b945fa0ea..d1f8146a06ea 100644
--- a/fs/xfs/xfs_utils.c
+++ b/fs/xfs/xfs_utils.c
@@ -147,7 +147,7 @@ xfs_dir_ialloc(
         xfs_inode_t     *dp,            /* directory within whose allocate
                                            the inode. */
         mode_t          mode,
-        nlink_t         nlink,
+        xfs_nlink_t     nlink,
         xfs_dev_t       rdev,
         cred_t          *credp,
         prid_t          prid,           /* project id */
diff --git a/fs/xfs/xfs_utils.h b/fs/xfs/xfs_utils.h
index e1ed6a588000..01d98b4b7af7 100644
--- a/fs/xfs/xfs_utils.h
+++ b/fs/xfs/xfs_utils.h
@@ -42,7 +42,7 @@ extern int xfs_get_dir_entry (vname_t *, xfs_inode_t **);
 extern int xfs_dir_lookup_int (bhv_desc_t *, uint, vname_t *, xfs_ino_t *,
                                 xfs_inode_t **);
 extern int xfs_truncate_file (xfs_mount_t *, xfs_inode_t *);
-extern int xfs_dir_ialloc (xfs_trans_t **, xfs_inode_t *, mode_t, nlink_t,
+extern int xfs_dir_ialloc (xfs_trans_t **, xfs_inode_t *, mode_t, xfs_nlink_t,
                                 xfs_dev_t, cred_t *, prid_t, int,
                                 xfs_inode_t **, int *);
 extern int xfs_droplink (xfs_trans_t *, xfs_inode_t *);
diff --git a/fs/xfs/xfs_vfsops.c b/fs/xfs/xfs_vfsops.c
index 00aae9c6a904..b53736650100 100644
--- a/fs/xfs/xfs_vfsops.c
+++ b/fs/xfs/xfs_vfsops.c
@@ -1649,6 +1649,7 @@ 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_ALLOCSIZE    "allocsize"    /* preferred allocation size */
 #define MNTOPT_IHASHSIZE    "ihashsize"    /* size of inode hash table */
 #define MNTOPT_NORECOVERY   "norecovery"   /* don't run XFS recovery */
 #define MNTOPT_NOLOGFLUSH   "nologflush"   /* don't hard flush on log writes */
@@ -1657,6 +1658,28 @@ xfs_vget(
 #define MNTOPT_IKEEP    "ikeep"         /* do not free empty inode clusters */
 #define MNTOPT_NOIKEEP  "noikeep"       /* free empty inode clusters */
 
+STATIC unsigned long
+suffix_strtoul(const char *cp, char **endp, unsigned int base)
+{
+        int     last, shift_left_factor = 0;
+        char    *value = (char *)cp;
+
+        last = strlen(value) - 1;
+        if (value[last] == 'K' || value[last] == 'k') {
+                shift_left_factor = 10;
+                value[last] = '\0';
+        }
+        if (value[last] == 'M' || value[last] == 'm') {
+                shift_left_factor = 20;
+                value[last] = '\0';
+        }
+        if (value[last] == 'G' || value[last] == 'g') {
+                shift_left_factor = 30;
+                value[last] = '\0';
+        }
+
+        return simple_strtoul(cp, endp, base) << shift_left_factor;
+}
 
 int
 xfs_parseargs(
@@ -1688,60 +1711,60 @@ xfs_parseargs(
                 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_LOGBUFS);
+                                        this_char);
                                 return EINVAL;
                         }
                         args->logbufs = simple_strtoul(value, &eov, 10);
                 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
-                        int     last, in_kilobytes = 0;
-
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_LOGBSIZE);
+                                        this_char);
                                 return EINVAL;
                         }
-                        last = strlen(value) - 1;
-                        if (value[last] == 'K' || value[last] == 'k') {
-                                in_kilobytes = 1;
-                                value[last] = '\0';
-                        }
-                        args->logbufsize = simple_strtoul(value, &eov, 10);
-                        if (in_kilobytes)
-                                args->logbufsize <<= 10;
+                        args->logbufsize = suffix_strtoul(value, &eov, 10);
                 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_LOGDEV);
+                                        this_char);
                                 return EINVAL;
                         }
                         strncpy(args->logname, value, MAXNAMELEN);
                 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_MTPT);
+                                        this_char);
                                 return EINVAL;
                         }
                         strncpy(args->mtpt, value, MAXNAMELEN);
                 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_RTDEV);
+                                        this_char);
                                 return EINVAL;
                         }
                         strncpy(args->rtname, value, MAXNAMELEN);
                 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_BIOSIZE); 
+                                        this_char);
                                 return EINVAL;
                         }
                         iosize = simple_strtoul(value, &eov, 10);
                         args->flags |= XFSMNT_IOSIZE;
                         args->iosizelog = (uint8_t) iosize;
+                } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
+                        if (!value || !*value) {
+                                printk("XFS: %s option requires an argument\n",
+                                        this_char);
+                                return EINVAL;
+                        }
+                        iosize = suffix_strtoul(value, &eov, 10);
+                        args->flags |= XFSMNT_IOSIZE;
+                        args->iosizelog = ffs(iosize) - 1;
                 } else if (!strcmp(this_char, MNTOPT_IHASHSIZE)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        this_char); 
+                                        this_char);
                                 return EINVAL;
                         }
                         args->flags |= XFSMNT_IHASHSIZE;
@@ -1756,7 +1779,7 @@ xfs_parseargs(
                         args->flags |= XFSMNT_INO64;
 #if !XFS_BIG_INUMS
                         printk("XFS: %s option not allowed on this system\n",
-                                MNTOPT_INO64);
+                                this_char);
                         return EINVAL;
 #endif
                 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
@@ -1766,14 +1789,14 @@ xfs_parseargs(
                 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_SUNIT);
+                                        this_char);
                                 return EINVAL;
                         }
                         dsunit = simple_strtoul(value, &eov, 10);
                 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
                         if (!value || !*value) {
                                 printk("XFS: %s option requires an argument\n",
-                                        MNTOPT_SWIDTH);
+                                        this_char);
                                 return EINVAL;
                         }
                         dswidth = simple_strtoul(value, &eov, 10);
@@ -1781,7 +1804,7 @@ xfs_parseargs(
                         args->flags &= ~XFSMNT_32BITINODES;
 #if !XFS_BIG_INUMS
                         printk("XFS: %s option not allowed on this system\n",
-                                MNTOPT_64BITINODE);
+                                this_char);
                         return EINVAL;
 #endif
                 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
@@ -1877,7 +1900,7 @@ xfs_showargs(
                 seq_printf(m, "," MNTOPT_IHASHSIZE "=%d", mp->m_ihsize);
 
         if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
-                seq_printf(m, "," MNTOPT_BIOSIZE "=%d", mp->m_writeio_log);
+                seq_printf(m, "," MNTOPT_ALLOCSIZE "=%d", 1<<mp->m_writeio_log);
 
         if (mp->m_logbufs > 0)
                 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c
index 70092963ca9e..25a526629b12 100644
--- a/fs/xfs/xfs_vnodeops.c
+++ b/fs/xfs/xfs_vnodeops.c
@@ -305,7 +305,7 @@ xfs_setattr(
         int                     mandlock_before, mandlock_after;
         struct xfs_dquot        *udqp, *gdqp, *olddquot1, *olddquot2;
         int                     file_owner;
-        int                     need_iolock = (flags & ATTR_DMI) == 0;
+        int                     need_iolock = 1;
 
         vp = BHV_TO_VNODE(bdp);
         vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address);
@@ -384,6 +384,9 @@ xfs_setattr(
          */
         tp = NULL;
         lock_flags = XFS_ILOCK_EXCL;
+        ASSERT(flags & ATTR_NOLOCK ? flags & ATTR_DMI : 1);
+        if (flags & ATTR_NOLOCK)
+                need_iolock = 0;
         if (!(mask & XFS_AT_SIZE)) {
                 if ((mask != (XFS_AT_CTIME|XFS_AT_ATIME|XFS_AT_MTIME)) ||
                     (mp->m_flags & XFS_MOUNT_WSYNC)) {
@@ -4320,7 +4323,7 @@ xfs_free_file_space(
         int                     rt;
         xfs_fileoff_t           startoffset_fsb;
         xfs_trans_t             *tp;
-        int                     need_iolock = (attr_flags & ATTR_DMI) == 0;
+        int                     need_iolock = 1;
 
         vn_trace_entry(XFS_ITOV(ip), __FUNCTION__, (inst_t *)__return_address);
         mp = ip->i_mount;
@@ -4348,8 +4351,12 @@ xfs_free_file_space(
                         return(error);
         }
 
+        ASSERT(attr_flags & ATTR_NOLOCK ? attr_flags & ATTR_DMI : 1);
+        if (attr_flags & ATTR_NOLOCK)
+                need_iolock = 0;
         if (need_iolock)
                 xfs_ilock(ip, XFS_IOLOCK_EXCL);
+
         rounding = MAX((__uint8_t)(1 << mp->m_sb.sb_blocklog),
                         (__uint8_t)NBPP);
         ilen = len + (offset & (rounding - 1));