6 files changed, 107 insertions, 46 deletions

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 *);