writeback: split inode_wb_list_lock into bdi_writeback.list_lock
authorChristoph Hellwig <hch@infradead.org>
Fri, 22 Apr 2011 00:19:44 +0000 (18:19 -0600)
committerWu Fengguang <fengguang.wu@intel.com>
Wed, 8 Jun 2011 00:25:21 +0000 (08:25 +0800)
Split the global inode_wb_list_lock into a per-bdi_writeback list_lock,
as it's currently the most contended lock in the system for metadata
heavy workloads.  It won't help for single-filesystem workloads for
which we'll need the I/O-less balance_dirty_pages, but at least we
can dedicate a cpu to spinning on each bdi now for larger systems.

Based on earlier patches from Nick Piggin and Dave Chinner.

It reduces lock contentions to 1/4 in this test case:
10 HDD JBOD, 100 dd on each disk, XFS, 6GB ram

lock_stat version 0.3
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
                              class name    con-bounces    contentions   waittime-min   waittime-max waittime-total    acq-bounces   acquisitions   holdtime-min   holdtime-max holdtime-total
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
vanilla 2.6.39-rc3:
                      inode_wb_list_lock:         42590          44433           0.12         147.74      144127.35         252274         886792           0.08         121.34      917211.23
                      ------------------
                      inode_wb_list_lock              2          [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85
                      inode_wb_list_lock             34          [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49
                      inode_wb_list_lock          12893          [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0
                      inode_wb_list_lock          10702          [<ffffffff8115afef>] writeback_single_inode+0x16d/0x20a
                      ------------------
                      inode_wb_list_lock              2          [<ffffffff81165da5>] bdev_inode_switch_bdi+0x29/0x85
                      inode_wb_list_lock             19          [<ffffffff8115bd0b>] inode_wb_list_del+0x22/0x49
                      inode_wb_list_lock           5550          [<ffffffff8115bb53>] __mark_inode_dirty+0x170/0x1d0
                      inode_wb_list_lock           8511          [<ffffffff8115b4ad>] writeback_sb_inodes+0x10f/0x157

2.6.39-rc3 + patch:
                &(&wb->list_lock)->rlock:         11383          11657           0.14         151.69       40429.51          90825         527918           0.11         145.90      556843.37
                ------------------------
                &(&wb->list_lock)->rlock             10          [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86
                &(&wb->list_lock)->rlock           1493          [<ffffffff8115b1ed>] writeback_inodes_wb+0x3d/0x150
                &(&wb->list_lock)->rlock           3652          [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f
                &(&wb->list_lock)->rlock           1412          [<ffffffff8115a38e>] writeback_single_inode+0x17f/0x223
                ------------------------
                &(&wb->list_lock)->rlock              3          [<ffffffff8110b5af>] bdi_lock_two+0x46/0x4b
                &(&wb->list_lock)->rlock              6          [<ffffffff8115b189>] inode_wb_list_del+0x5f/0x86
                &(&wb->list_lock)->rlock           2061          [<ffffffff8115af97>] __mark_inode_dirty+0x173/0x1cf
                &(&wb->list_lock)->rlock           2629          [<ffffffff8115a8e9>] writeback_sb_inodes+0x123/0x16f

hughd@google.com: fix recursive lock when bdi_lock_two() is called with new the same as old
akpm@linux-foundation.org: cleanup bdev_inode_switch_bdi() comment

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
fs/block_dev.c
fs/fs-writeback.c
fs/inode.c
include/linux/backing-dev.h
include/linux/writeback.h
mm/backing-dev.c
mm/filemap.c
mm/rmap.c

index 1a2421f..3c9a03e 100644 (file)
@@ -44,24 +44,28 @@ inline struct block_device *I_BDEV(struct inode *inode)
 {
        return &BDEV_I(inode)->bdev;
 }
-
 EXPORT_SYMBOL(I_BDEV);
 
 /*
- * move the inode from it's current bdi to the a new bdi. if the inode is dirty
- * we need to move it onto the dirty list of @dst so that the inode is always
- * on the right list.
+ * Move the inode from its current bdi to a new bdi. If the inode is dirty we
+ * need to move it onto the dirty list of @dst so that the inode is always on
+ * the right list.
  */
 static void bdev_inode_switch_bdi(struct inode *inode,
                        struct backing_dev_info *dst)
 {
-       spin_lock(&inode_wb_list_lock);
+       struct backing_dev_info *old = inode->i_data.backing_dev_info;
+
+       if (unlikely(dst == old))               /* deadlock avoidance */
+               return;
+       bdi_lock_two(&old->wb, &dst->wb);
        spin_lock(&inode->i_lock);
        inode->i_data.backing_dev_info = dst;
        if (inode->i_state & I_DIRTY)
                list_move(&inode->i_wb_list, &dst->wb.b_dirty);
        spin_unlock(&inode->i_lock);
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&old->wb.list_lock);
+       spin_unlock(&dst->wb.list_lock);
 }
 
 static sector_t max_block(struct block_device *bdev)
index 664acdb..36a3091 100644 (file)
@@ -181,12 +181,13 @@ void bdi_start_background_writeback(struct backing_dev_info *bdi)
  */
 void inode_wb_list_del(struct inode *inode)
 {
-       spin_lock(&inode_wb_list_lock);
+       struct backing_dev_info *bdi = inode_to_bdi(inode);
+
+       spin_lock(&bdi->wb.list_lock);
        list_del_init(&inode->i_wb_list);
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&bdi->wb.list_lock);
 }
 
-
 /*
  * Redirty an inode: set its when-it-was dirtied timestamp and move it to the
  * furthest end of its superblock's dirty-inode list.
@@ -196,11 +197,9 @@ void inode_wb_list_del(struct inode *inode)
  * the case then the inode must have been redirtied while it was being written
  * out and we don't reset its dirtied_when.
  */
-static void redirty_tail(struct inode *inode)
+static void redirty_tail(struct inode *inode, struct bdi_writeback *wb)
 {
-       struct bdi_writeback *wb = &inode_to_bdi(inode)->wb;
-
-       assert_spin_locked(&inode_wb_list_lock);
+       assert_spin_locked(&wb->list_lock);
        if (!list_empty(&wb->b_dirty)) {
                struct inode *tail;
 
@@ -214,11 +213,9 @@ static void redirty_tail(struct inode *inode)
 /*
  * requeue inode for re-scanning after bdi->b_io list is exhausted.
  */
-static void requeue_io(struct inode *inode)
+static void requeue_io(struct inode *inode, struct bdi_writeback *wb)
 {
-       struct bdi_writeback *wb = &inode_to_bdi(inode)->wb;
-
-       assert_spin_locked(&inode_wb_list_lock);
+       assert_spin_locked(&wb->list_lock);
        list_move(&inode->i_wb_list, &wb->b_more_io);
 }
 
@@ -226,7 +223,7 @@ static void inode_sync_complete(struct inode *inode)
 {
        /*
         * Prevent speculative execution through
-        * spin_unlock(&inode_wb_list_lock);
+        * spin_unlock(&wb->list_lock);
         */
 
        smp_mb();
@@ -302,7 +299,7 @@ static void move_expired_inodes(struct list_head *delaying_queue,
  */
 static void queue_io(struct bdi_writeback *wb, unsigned long *older_than_this)
 {
-       assert_spin_locked(&inode_wb_list_lock);
+       assert_spin_locked(&wb->list_lock);
        list_splice_init(&wb->b_more_io, &wb->b_io);
        move_expired_inodes(&wb->b_dirty, &wb->b_io, older_than_this);
 }
@@ -317,7 +314,8 @@ static int write_inode(struct inode *inode, struct writeback_control *wbc)
 /*
  * Wait for writeback on an inode to complete.
  */
-static void inode_wait_for_writeback(struct inode *inode)
+static void inode_wait_for_writeback(struct inode *inode,
+                                    struct bdi_writeback *wb)
 {
        DEFINE_WAIT_BIT(wq, &inode->i_state, __I_SYNC);
        wait_queue_head_t *wqh;
@@ -325,15 +323,15 @@ static void inode_wait_for_writeback(struct inode *inode)
        wqh = bit_waitqueue(&inode->i_state, __I_SYNC);
        while (inode->i_state & I_SYNC) {
                spin_unlock(&inode->i_lock);
-               spin_unlock(&inode_wb_list_lock);
+               spin_unlock(&wb->list_lock);
                __wait_on_bit(wqh, &wq, inode_wait, TASK_UNINTERRUPTIBLE);
-               spin_lock(&inode_wb_list_lock);
+               spin_lock(&wb->list_lock);
                spin_lock(&inode->i_lock);
        }
 }
 
 /*
- * Write out an inode's dirty pages.  Called under inode_wb_list_lock and
+ * Write out an inode's dirty pages.  Called under wb->list_lock and
  * inode->i_lock.  Either the caller has an active reference on the inode or
  * the inode has I_WILL_FREE set.
  *
@@ -344,13 +342,14 @@ static void inode_wait_for_writeback(struct inode *inode)
  * livelocks, etc.
  */
 static int
-writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
+writeback_single_inode(struct inode *inode, struct bdi_writeback *wb,
+                      struct writeback_control *wbc)
 {
        struct address_space *mapping = inode->i_mapping;
        unsigned dirty;
        int ret;
 
-       assert_spin_locked(&inode_wb_list_lock);
+       assert_spin_locked(&wb->list_lock);
        assert_spin_locked(&inode->i_lock);
 
        if (!atomic_read(&inode->i_count))
@@ -368,14 +367,14 @@ writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
                 * completed a full scan of b_io.
                 */
                if (wbc->sync_mode != WB_SYNC_ALL) {
-                       requeue_io(inode);
+                       requeue_io(inode, wb);
                        return 0;
                }
 
                /*
                 * It's a data-integrity sync.  We must wait.
                 */
-               inode_wait_for_writeback(inode);
+               inode_wait_for_writeback(inode, wb);
        }
 
        BUG_ON(inode->i_state & I_SYNC);
@@ -384,7 +383,7 @@ writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
        inode->i_state |= I_SYNC;
        inode->i_state &= ~I_DIRTY_PAGES;
        spin_unlock(&inode->i_lock);
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&wb->list_lock);
 
        ret = do_writepages(mapping, wbc);
 
@@ -415,7 +414,7 @@ writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
                        ret = err;
        }
 
-       spin_lock(&inode_wb_list_lock);
+       spin_lock(&wb->list_lock);
        spin_lock(&inode->i_lock);
        inode->i_state &= ~I_SYNC;
        if (!(inode->i_state & I_FREEING)) {
@@ -438,7 +437,7 @@ writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
                                /*
                                 * slice used up: queue for next turn
                                 */
-                               requeue_io(inode);
+                               requeue_io(inode, wb);
                        } else {
                                /*
                                 * Writeback blocked by something other than
@@ -447,7 +446,7 @@ writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
                                 * retrying writeback of the dirty page/inode
                                 * that cannot be performed immediately.
                                 */
-                               redirty_tail(inode);
+                               redirty_tail(inode, wb);
                        }
                } else if (inode->i_state & I_DIRTY) {
                        /*
@@ -456,7 +455,7 @@ writeback_single_inode(struct inode *inode, struct writeback_control *wbc)
                         * submission or metadata updates after data IO
                         * completion.
                         */
-                       redirty_tail(inode);
+                       redirty_tail(inode, wb);
                } else {
                        /*
                         * The inode is clean.  At this point we either have
@@ -521,7 +520,7 @@ static int writeback_sb_inodes(struct super_block *sb, struct bdi_writeback *wb,
                                 * superblock, move all inodes not belonging
                                 * to it back onto the dirty list.
                                 */
-                               redirty_tail(inode);
+                               redirty_tail(inode, wb);
                                continue;
                        }
 
@@ -541,7 +540,7 @@ static int writeback_sb_inodes(struct super_block *sb, struct bdi_writeback *wb,
                spin_lock(&inode->i_lock);
                if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
                        spin_unlock(&inode->i_lock);
-                       requeue_io(inode);
+                       requeue_io(inode, wb);
                        continue;
                }
 
@@ -557,19 +556,19 @@ static int writeback_sb_inodes(struct super_block *sb, struct bdi_writeback *wb,
                __iget(inode);
 
                pages_skipped = wbc->pages_skipped;
-               writeback_single_inode(inode, wbc);
+               writeback_single_inode(inode, wb, wbc);
                if (wbc->pages_skipped != pages_skipped) {
                        /*
                         * writeback is not making progress due to locked
                         * buffers.  Skip this inode for now.
                         */
-                       redirty_tail(inode);
+                       redirty_tail(inode, wb);
                }
                spin_unlock(&inode->i_lock);
-               spin_unlock(&inode_wb_list_lock);
+               spin_unlock(&wb->list_lock);
                iput(inode);
                cond_resched();
-               spin_lock(&inode_wb_list_lock);
+               spin_lock(&wb->list_lock);
                if (wbc->nr_to_write <= 0) {
                        wbc->more_io = 1;
                        return 1;
@@ -588,7 +587,7 @@ void writeback_inodes_wb(struct bdi_writeback *wb,
 
        if (!wbc->wb_start)
                wbc->wb_start = jiffies; /* livelock avoidance */
-       spin_lock(&inode_wb_list_lock);
+       spin_lock(&wb->list_lock);
 
        if (list_empty(&wb->b_io))
                queue_io(wb, wbc->older_than_this);
@@ -598,7 +597,7 @@ void writeback_inodes_wb(struct bdi_writeback *wb,
                struct super_block *sb = inode->i_sb;
 
                if (!pin_sb_for_writeback(sb)) {
-                       requeue_io(inode);
+                       requeue_io(inode, wb);
                        continue;
                }
                ret = writeback_sb_inodes(sb, wb, wbc, false);
@@ -607,7 +606,7 @@ void writeback_inodes_wb(struct bdi_writeback *wb,
                if (ret)
                        break;
        }
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&wb->list_lock);
        /* Leave any unwritten inodes on b_io */
 }
 
@@ -616,11 +615,11 @@ static void __writeback_inodes_sb(struct super_block *sb,
 {
        WARN_ON(!rwsem_is_locked(&sb->s_umount));
 
-       spin_lock(&inode_wb_list_lock);
+       spin_lock(&wb->list_lock);
        if (list_empty(&wb->b_io))
                queue_io(wb, wbc->older_than_this);
        writeback_sb_inodes(sb, wb, wbc, true);
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&wb->list_lock);
 }
 
 /*
@@ -762,15 +761,15 @@ static long wb_writeback(struct bdi_writeback *wb,
                 * become available for writeback. Otherwise
                 * we'll just busyloop.
                 */
-               spin_lock(&inode_wb_list_lock);
+               spin_lock(&wb->list_lock);
                if (!list_empty(&wb->b_more_io))  {
                        inode = wb_inode(wb->b_more_io.prev);
                        trace_wbc_writeback_wait(&wbc, wb->bdi);
                        spin_lock(&inode->i_lock);
-                       inode_wait_for_writeback(inode);
+                       inode_wait_for_writeback(inode, wb);
                        spin_unlock(&inode->i_lock);
                }
-               spin_unlock(&inode_wb_list_lock);
+               spin_unlock(&wb->list_lock);
        }
 
        return wrote;
@@ -1104,10 +1103,10 @@ void __mark_inode_dirty(struct inode *inode, int flags)
                        }
 
                        spin_unlock(&inode->i_lock);
-                       spin_lock(&inode_wb_list_lock);
+                       spin_lock(&bdi->wb.list_lock);
                        inode->dirtied_when = jiffies;
                        list_move(&inode->i_wb_list, &bdi->wb.b_dirty);
-                       spin_unlock(&inode_wb_list_lock);
+                       spin_unlock(&bdi->wb.list_lock);
 
                        if (wakeup_bdi)
                                bdi_wakeup_thread_delayed(bdi);
@@ -1309,6 +1308,7 @@ EXPORT_SYMBOL(sync_inodes_sb);
  */
 int write_inode_now(struct inode *inode, int sync)
 {
+       struct bdi_writeback *wb = &inode_to_bdi(inode)->wb;
        int ret;
        struct writeback_control wbc = {
                .nr_to_write = LONG_MAX,
@@ -1321,11 +1321,11 @@ int write_inode_now(struct inode *inode, int sync)
                wbc.nr_to_write = 0;
 
        might_sleep();
-       spin_lock(&inode_wb_list_lock);
+       spin_lock(&wb->list_lock);
        spin_lock(&inode->i_lock);
-       ret = writeback_single_inode(inode, &wbc);
+       ret = writeback_single_inode(inode, wb, &wbc);
        spin_unlock(&inode->i_lock);
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&wb->list_lock);
        if (sync)
                inode_sync_wait(inode);
        return ret;
@@ -1345,13 +1345,14 @@ EXPORT_SYMBOL(write_inode_now);
  */
 int sync_inode(struct inode *inode, struct writeback_control *wbc)
 {
+       struct bdi_writeback *wb = &inode_to_bdi(inode)->wb;
        int ret;
 
-       spin_lock(&inode_wb_list_lock);
+       spin_lock(&wb->list_lock);
        spin_lock(&inode->i_lock);
-       ret = writeback_single_inode(inode, wbc);
+       ret = writeback_single_inode(inode, wb, wbc);
        spin_unlock(&inode->i_lock);
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&wb->list_lock);
        return ret;
 }
 EXPORT_SYMBOL(sync_inode);
index 0f7e88a..4be128c 100644 (file)
@@ -37,7 +37,7 @@
  *   inode_lru, inode->i_lru
  * inode_sb_list_lock protects:
  *   sb->s_inodes, inode->i_sb_list
- * inode_wb_list_lock protects:
+ * bdi->wb.list_lock protects:
  *   bdi->wb.b_{dirty,io,more_io}, inode->i_wb_list
  * inode_hash_lock protects:
  *   inode_hashtable, inode->i_hash
@@ -48,7 +48,7 @@
  *   inode->i_lock
  *     inode_lru_lock
  *
- * inode_wb_list_lock
+ * bdi->wb.list_lock
  *   inode->i_lock
  *
  * inode_hash_lock
@@ -68,7 +68,6 @@ static LIST_HEAD(inode_lru);
 static DEFINE_SPINLOCK(inode_lru_lock);
 
 __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_sb_list_lock);
-__cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_wb_list_lock);
 
 /*
  * iprune_sem provides exclusion between the icache shrinking and the
index 96f4094..47feb2c 100644 (file)
@@ -57,6 +57,7 @@ struct bdi_writeback {
        struct list_head b_dirty;       /* dirty inodes */
        struct list_head b_io;          /* parked for writeback */
        struct list_head b_more_io;     /* parked for more writeback */
+       spinlock_t list_lock;           /* protects the b_* lists */
 };
 
 struct backing_dev_info {
@@ -106,6 +107,7 @@ int bdi_writeback_thread(void *data);
 int bdi_has_dirty_io(struct backing_dev_info *bdi);
 void bdi_arm_supers_timer(void);
 void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi);
+void bdi_lock_two(struct bdi_writeback *wb1, struct bdi_writeback *wb2);
 
 extern spinlock_t bdi_lock;
 extern struct list_head bdi_list;
index 7df9026..c2d957f 100644 (file)
@@ -9,8 +9,6 @@
 
 struct backing_dev_info;
 
-extern spinlock_t inode_wb_list_lock;
-
 /*
  * fs/fs-writeback.c
  */
index f032e6e..5f6553e 100644 (file)
@@ -45,6 +45,17 @@ static struct timer_list sync_supers_timer;
 static int bdi_sync_supers(void *);
 static void sync_supers_timer_fn(unsigned long);
 
+void bdi_lock_two(struct bdi_writeback *wb1, struct bdi_writeback *wb2)
+{
+       if (wb1 < wb2) {
+               spin_lock(&wb1->list_lock);
+               spin_lock_nested(&wb2->list_lock, 1);
+       } else {
+               spin_lock(&wb2->list_lock);
+               spin_lock_nested(&wb1->list_lock, 1);
+       }
+}
+
 #ifdef CONFIG_DEBUG_FS
 #include <linux/debugfs.h>
 #include <linux/seq_file.h>
@@ -67,14 +78,14 @@ static int bdi_debug_stats_show(struct seq_file *m, void *v)
        struct inode *inode;
 
        nr_dirty = nr_io = nr_more_io = 0;
-       spin_lock(&inode_wb_list_lock);
+       spin_lock(&wb->list_lock);
        list_for_each_entry(inode, &wb->b_dirty, i_wb_list)
                nr_dirty++;
        list_for_each_entry(inode, &wb->b_io, i_wb_list)
                nr_io++;
        list_for_each_entry(inode, &wb->b_more_io, i_wb_list)
                nr_more_io++;
-       spin_unlock(&inode_wb_list_lock);
+       spin_unlock(&wb->list_lock);
 
        global_dirty_limits(&background_thresh, &dirty_thresh);
        bdi_thresh = bdi_dirty_limit(bdi, dirty_thresh);
@@ -628,6 +639,7 @@ static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
        INIT_LIST_HEAD(&wb->b_dirty);
        INIT_LIST_HEAD(&wb->b_io);
        INIT_LIST_HEAD(&wb->b_more_io);
+       spin_lock_init(&wb->list_lock);
        setup_timer(&wb->wakeup_timer, wakeup_timer_fn, (unsigned long)bdi);
 }
 
@@ -676,11 +688,12 @@ void bdi_destroy(struct backing_dev_info *bdi)
        if (bdi_has_dirty_io(bdi)) {
                struct bdi_writeback *dst = &default_backing_dev_info.wb;
 
-               spin_lock(&inode_wb_list_lock);
+               bdi_lock_two(&bdi->wb, dst);
                list_splice(&bdi->wb.b_dirty, &dst->b_dirty);
                list_splice(&bdi->wb.b_io, &dst->b_io);
                list_splice(&bdi->wb.b_more_io, &dst->b_more_io);
-               spin_unlock(&inode_wb_list_lock);
+               spin_unlock(&bdi->wb.list_lock);
+               spin_unlock(&dst->list_lock);
        }
 
        bdi_unregister(bdi);
index d7b1057..1e492c3 100644 (file)
@@ -81,7 +81,7 @@
  *  ->i_mutex
  *    ->i_alloc_sem             (various)
  *
- *  inode_wb_list_lock
+ *  bdi->wb.list_lock
  *    sb_lock                  (fs/fs-writeback.c)
  *    ->mapping->tree_lock     (__sync_single_inode)
  *
@@ -99,9 +99,9 @@
  *    ->zone.lru_lock          (check_pte_range->isolate_lru_page)
  *    ->private_lock           (page_remove_rmap->set_page_dirty)
  *    ->tree_lock              (page_remove_rmap->set_page_dirty)
- *    inode_wb_list_lock       (page_remove_rmap->set_page_dirty)
+ *    bdi.wb->list_lock                (page_remove_rmap->set_page_dirty)
  *    ->inode->i_lock          (page_remove_rmap->set_page_dirty)
- *    inode_wb_list_lock       (zap_pte_range->set_page_dirty)
+ *    bdi.wb->list_lock                (zap_pte_range->set_page_dirty)
  *    ->inode->i_lock          (zap_pte_range->set_page_dirty)
  *    ->private_lock           (zap_pte_range->__set_page_dirty_buffers)
  *
index 0eb463e..d04e36a 100644 (file)
--- a/mm/rmap.c
+++ b/mm/rmap.c
  *               mmlist_lock (in mmput, drain_mmlist and others)
  *               mapping->private_lock (in __set_page_dirty_buffers)
  *               inode->i_lock (in set_page_dirty's __mark_inode_dirty)
- *               inode_wb_list_lock (in set_page_dirty's __mark_inode_dirty)
+ *               bdi.wb->list_lock (in set_page_dirty's __mark_inode_dirty)
  *                 sb_lock (within inode_lock in fs/fs-writeback.c)
  *                 mapping->tree_lock (widely used, in set_page_dirty,
  *                           in arch-dependent flush_dcache_mmap_lock,
- *                           within inode_wb_list_lock in __sync_single_inode)
+ *                           within bdi.wb->list_lock in __sync_single_inode)
  *
  * (code doesn't rely on that order so it could be switched around)
  * ->tasklist_lock