writeback: move backing_dev_info->bdi_stat[] into bdi_writeback
authorTejun Heo <tj@kernel.org>
Fri, 22 May 2015 21:13:27 +0000 (17:13 -0400)
committerJens Axboe <axboe@fb.com>
Tue, 2 Jun 2015 14:33:34 +0000 (08:33 -0600)
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear.  For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi.  To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.

This patch moves bdi->bdi_stat[] into wb.

* enum bdi_stat_item is renamed to wb_stat_item and the prefix of all
  enums is changed from BDI_ to WB_.

* BDI_STAT_BATCH() -> WB_STAT_BATCH()

* [__]{add|inc|dec|sum}_wb_stat(bdi, ...) -> [__]{add|inc}_wb_stat(wb, ...)

* bdi_stat[_error]() -> wb_stat[_error]()

* bdi_writeout_inc() -> wb_writeout_inc()

* stat init is moved to bdi_wb_init() and bdi_wb_exit() is added and
  frees stat.

* As there's still only one bdi_writeback per backing_dev_info, all
  uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
  introducing no behavior changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
fs/fs-writeback.c
fs/fuse/file.c
fs/nfs/internal.h
fs/nfs/write.c
include/linux/backing-dev.h
mm/backing-dev.c
mm/page-writeback.c

index 983312cea24546077d8692b1961515bf82852def..8873ecd1578c97eb6dd7b7e5625af12b60441452 100644 (file)
@@ -840,7 +840,7 @@ static bool over_bground_thresh(struct backing_dev_info *bdi)
            global_page_state(NR_UNSTABLE_NFS) > background_thresh)
                return true;
 
-       if (bdi_stat(bdi, BDI_RECLAIMABLE) >
+       if (wb_stat(&bdi->wb, WB_RECLAIMABLE) >
                                bdi_dirty_limit(bdi, background_thresh))
                return true;
 
index 5ef05b5c4cff86e9353a0594f0faadc1ab8612f5..8c5e2fa68835a07216d250dc9536aafaa1063dc2 100644 (file)
@@ -1445,9 +1445,9 @@ static void fuse_writepage_finish(struct fuse_conn *fc, struct fuse_req *req)
 
        list_del(&req->writepages_entry);
        for (i = 0; i < req->num_pages; i++) {
-               dec_bdi_stat(bdi, BDI_WRITEBACK);
+               dec_wb_stat(&bdi->wb, WB_WRITEBACK);
                dec_zone_page_state(req->pages[i], NR_WRITEBACK_TEMP);
-               bdi_writeout_inc(bdi);
+               wb_writeout_inc(&bdi->wb);
        }
        wake_up(&fi->page_waitq);
 }
@@ -1634,7 +1634,7 @@ static int fuse_writepage_locked(struct page *page)
        req->end = fuse_writepage_end;
        req->inode = inode;
 
-       inc_bdi_stat(inode_to_bdi(inode), BDI_WRITEBACK);
+       inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK);
        inc_zone_page_state(tmp_page, NR_WRITEBACK_TEMP);
 
        spin_lock(&fc->lock);
@@ -1749,9 +1749,9 @@ static bool fuse_writepage_in_flight(struct fuse_req *new_req,
                copy_highpage(old_req->pages[0], page);
                spin_unlock(&fc->lock);
 
-               dec_bdi_stat(bdi, BDI_WRITEBACK);
+               dec_wb_stat(&bdi->wb, WB_WRITEBACK);
                dec_zone_page_state(page, NR_WRITEBACK_TEMP);
-               bdi_writeout_inc(bdi);
+               wb_writeout_inc(&bdi->wb);
                fuse_writepage_free(fc, new_req);
                fuse_request_free(new_req);
                goto out;
@@ -1848,7 +1848,7 @@ static int fuse_writepages_fill(struct page *page,
        req->page_descs[req->num_pages].offset = 0;
        req->page_descs[req->num_pages].length = PAGE_SIZE;
 
-       inc_bdi_stat(inode_to_bdi(inode), BDI_WRITEBACK);
+       inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK);
        inc_zone_page_state(tmp_page, NR_WRITEBACK_TEMP);
 
        err = 0;
index 9e6475bc5ba22be7293c687d78dda1a626d3024b..7e3c4604bea8a6e6e92906b6c2096adb6df5ca3f 100644 (file)
@@ -607,7 +607,7 @@ void nfs_mark_page_unstable(struct page *page)
        struct inode *inode = page_file_mapping(page)->host;
 
        inc_zone_page_state(page, NR_UNSTABLE_NFS);
-       inc_bdi_stat(inode_to_bdi(inode), BDI_RECLAIMABLE);
+       inc_wb_stat(&inode_to_bdi(inode)->wb, WB_RECLAIMABLE);
         __mark_inode_dirty(inode, I_DIRTY_DATASYNC);
 }
 
index d12a4be613a5ced58599f8095822f4659b32f9a4..94c7ce01dfb1b27403ca706ab73fe48445163570 100644 (file)
@@ -853,7 +853,8 @@ static void
 nfs_clear_page_commit(struct page *page)
 {
        dec_zone_page_state(page, NR_UNSTABLE_NFS);
-       dec_bdi_stat(inode_to_bdi(page_file_mapping(page)->host), BDI_RECLAIMABLE);
+       dec_wb_stat(&inode_to_bdi(page_file_mapping(page)->host)->wb,
+                   WB_RECLAIMABLE);
 }
 
 /* Called holding inode (/cinfo) lock */
index eb14f988a63ed6667e8b72bac76bcb294dcd34c5..fe7a907a4e16fda739d2d63413a95f5bd2172eff 100644 (file)
@@ -36,15 +36,15 @@ enum wb_state {
 
 typedef int (congested_fn)(void *, int);
 
-enum bdi_stat_item {
-       BDI_RECLAIMABLE,
-       BDI_WRITEBACK,
-       BDI_DIRTIED,
-       BDI_WRITTEN,
-       NR_BDI_STAT_ITEMS
+enum wb_stat_item {
+       WB_RECLAIMABLE,
+       WB_WRITEBACK,
+       WB_DIRTIED,
+       WB_WRITTEN,
+       NR_WB_STAT_ITEMS
 };
 
-#define BDI_STAT_BATCH (8*(1+ilog2(nr_cpu_ids)))
+#define WB_STAT_BATCH (8*(1+ilog2(nr_cpu_ids)))
 
 struct bdi_writeback {
        struct backing_dev_info *bdi;   /* our parent bdi */
@@ -58,6 +58,8 @@ struct bdi_writeback {
        struct list_head b_more_io;     /* parked for more writeback */
        struct list_head b_dirty_time;  /* time stamps are dirty */
        spinlock_t list_lock;           /* protects the b_* lists */
+
+       struct percpu_counter stat[NR_WB_STAT_ITEMS];
 };
 
 struct backing_dev_info {
@@ -69,8 +71,6 @@ struct backing_dev_info {
 
        char *name;
 
-       struct percpu_counter bdi_stat[NR_BDI_STAT_ITEMS];
-
        unsigned long bw_time_stamp;    /* last time write bw is updated */
        unsigned long dirtied_stamp;
        unsigned long written_stamp;    /* pages written at bw_time_stamp */
@@ -137,78 +137,74 @@ static inline int wb_has_dirty_io(struct bdi_writeback *wb)
               !list_empty(&wb->b_more_io);
 }
 
-static inline void __add_bdi_stat(struct backing_dev_info *bdi,
-               enum bdi_stat_item item, s64 amount)
+static inline void __add_wb_stat(struct bdi_writeback *wb,
+                                enum wb_stat_item item, s64 amount)
 {
-       __percpu_counter_add(&bdi->bdi_stat[item], amount, BDI_STAT_BATCH);
+       __percpu_counter_add(&wb->stat[item], amount, WB_STAT_BATCH);
 }
 
-static inline void __inc_bdi_stat(struct backing_dev_info *bdi,
-               enum bdi_stat_item item)
+static inline void __inc_wb_stat(struct bdi_writeback *wb,
+                                enum wb_stat_item item)
 {
-       __add_bdi_stat(bdi, item, 1);
+       __add_wb_stat(wb, item, 1);
 }
 
-static inline void inc_bdi_stat(struct backing_dev_info *bdi,
-               enum bdi_stat_item item)
+static inline void inc_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
 {
        unsigned long flags;
 
        local_irq_save(flags);
-       __inc_bdi_stat(bdi, item);
+       __inc_wb_stat(wb, item);
        local_irq_restore(flags);
 }
 
-static inline void __dec_bdi_stat(struct backing_dev_info *bdi,
-               enum bdi_stat_item item)
+static inline void __dec_wb_stat(struct bdi_writeback *wb,
+                                enum wb_stat_item item)
 {
-       __add_bdi_stat(bdi, item, -1);
+       __add_wb_stat(wb, item, -1);
 }
 
-static inline void dec_bdi_stat(struct backing_dev_info *bdi,
-               enum bdi_stat_item item)
+static inline void dec_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
 {
        unsigned long flags;
 
        local_irq_save(flags);
-       __dec_bdi_stat(bdi, item);
+       __dec_wb_stat(wb, item);
        local_irq_restore(flags);
 }
 
-static inline s64 bdi_stat(struct backing_dev_info *bdi,
-               enum bdi_stat_item item)
+static inline s64 wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
 {
-       return percpu_counter_read_positive(&bdi->bdi_stat[item]);
+       return percpu_counter_read_positive(&wb->stat[item]);
 }
 
-static inline s64 __bdi_stat_sum(struct backing_dev_info *bdi,
-               enum bdi_stat_item item)
+static inline s64 __wb_stat_sum(struct bdi_writeback *wb,
+                               enum wb_stat_item item)
 {
-       return percpu_counter_sum_positive(&bdi->bdi_stat[item]);
+       return percpu_counter_sum_positive(&wb->stat[item]);
 }
 
-static inline s64 bdi_stat_sum(struct backing_dev_info *bdi,
-               enum bdi_stat_item item)
+static inline s64 wb_stat_sum(struct bdi_writeback *wb, enum wb_stat_item item)
 {
        s64 sum;
        unsigned long flags;
 
        local_irq_save(flags);
-       sum = __bdi_stat_sum(bdi, item);
+       sum = __wb_stat_sum(wb, item);
        local_irq_restore(flags);
 
        return sum;
 }
 
-extern void bdi_writeout_inc(struct backing_dev_info *bdi);
+extern void wb_writeout_inc(struct bdi_writeback *wb);
 
 /*
  * maximal error of a stat counter.
  */
-static inline unsigned long bdi_stat_error(struct backing_dev_info *bdi)
+static inline unsigned long wb_stat_error(struct bdi_writeback *wb)
 {
 #ifdef CONFIG_SMP
-       return nr_cpu_ids * BDI_STAT_BATCH;
+       return nr_cpu_ids * WB_STAT_BATCH;
 #else
        return 1;
 #endif
index b23cf0ea5912b4475ff2923c105fda0808fec9a7..7b1d1917b6583ebf7ed65b4fd4e530162259f2f7 100644 (file)
@@ -84,13 +84,13 @@ static int bdi_debug_stats_show(struct seq_file *m, void *v)
                   "b_dirty_time:       %10lu\n"
                   "bdi_list:           %10u\n"
                   "state:              %10lx\n",
-                  (unsigned long) K(bdi_stat(bdi, BDI_WRITEBACK)),
-                  (unsigned long) K(bdi_stat(bdi, BDI_RECLAIMABLE)),
+                  (unsigned long) K(wb_stat(wb, WB_WRITEBACK)),
+                  (unsigned long) K(wb_stat(wb, WB_RECLAIMABLE)),
                   K(bdi_thresh),
                   K(dirty_thresh),
                   K(background_thresh),
-                  (unsigned long) K(bdi_stat(bdi, BDI_DIRTIED)),
-                  (unsigned long) K(bdi_stat(bdi, BDI_WRITTEN)),
+                  (unsigned long) K(wb_stat(wb, WB_DIRTIED)),
+                  (unsigned long) K(wb_stat(wb, WB_WRITTEN)),
                   (unsigned long) K(bdi->write_bandwidth),
                   nr_dirty,
                   nr_io,
@@ -376,8 +376,10 @@ void bdi_unregister(struct backing_dev_info *bdi)
 }
 EXPORT_SYMBOL(bdi_unregister);
 
-static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
+static int bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
 {
+       int i, err;
+
        memset(wb, 0, sizeof(*wb));
 
        wb->bdi = bdi;
@@ -388,6 +390,27 @@ static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
        INIT_LIST_HEAD(&wb->b_dirty_time);
        spin_lock_init(&wb->list_lock);
        INIT_DELAYED_WORK(&wb->dwork, bdi_writeback_workfn);
+
+       for (i = 0; i < NR_WB_STAT_ITEMS; i++) {
+               err = percpu_counter_init(&wb->stat[i], 0, GFP_KERNEL);
+               if (err) {
+                       while (--i)
+                               percpu_counter_destroy(&wb->stat[i]);
+                       return err;
+               }
+       }
+
+       return 0;
+}
+
+static void bdi_wb_exit(struct bdi_writeback *wb)
+{
+       int i;
+
+       WARN_ON(delayed_work_pending(&wb->dwork));
+
+       for (i = 0; i < NR_WB_STAT_ITEMS; i++)
+               percpu_counter_destroy(&wb->stat[i]);
 }
 
 /*
@@ -397,7 +420,7 @@ static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
 
 int bdi_init(struct backing_dev_info *bdi)
 {
-       int i, err;
+       int err;
 
        bdi->dev = NULL;
 
@@ -408,13 +431,9 @@ int bdi_init(struct backing_dev_info *bdi)
        INIT_LIST_HEAD(&bdi->bdi_list);
        INIT_LIST_HEAD(&bdi->work_list);
 
-       bdi_wb_init(&bdi->wb, bdi);
-
-       for (i = 0; i < NR_BDI_STAT_ITEMS; i++) {
-               err = percpu_counter_init(&bdi->bdi_stat[i], 0, GFP_KERNEL);
-               if (err)
-                       goto err;
-       }
+       err = bdi_wb_init(&bdi->wb, bdi);
+       if (err)
+               return err;
 
        bdi->dirty_exceeded = 0;
 
@@ -427,25 +446,20 @@ int bdi_init(struct backing_dev_info *bdi)
        bdi->avg_write_bandwidth = INIT_BW;
 
        err = fprop_local_init_percpu(&bdi->completions, GFP_KERNEL);
-
        if (err) {
-err:
-               while (i--)
-                       percpu_counter_destroy(&bdi->bdi_stat[i]);
+               bdi_wb_exit(&bdi->wb);
+               return err;
        }
 
-       return err;
+       return 0;
 }
 EXPORT_SYMBOL(bdi_init);
 
 void bdi_destroy(struct backing_dev_info *bdi)
 {
-       int i;
-
        bdi_wb_shutdown(bdi);
 
        WARN_ON(!list_empty(&bdi->work_list));
-       WARN_ON(delayed_work_pending(&bdi->wb.dwork));
 
        if (bdi->dev) {
                bdi_debug_unregister(bdi);
@@ -453,8 +467,8 @@ void bdi_destroy(struct backing_dev_info *bdi)
                bdi->dev = NULL;
        }
 
-       for (i = 0; i < NR_BDI_STAT_ITEMS; i++)
-               percpu_counter_destroy(&bdi->bdi_stat[i]);
+       bdi_wb_exit(&bdi->wb);
+
        fprop_local_destroy_percpu(&bdi->completions);
 }
 EXPORT_SYMBOL(bdi_destroy);
index bdeecad004895ccfdbbd5c871a65182795534a92..dc673a03541395adbd7d797c3556a2ff17e937ed 100644 (file)
@@ -396,11 +396,11 @@ static unsigned long wp_next_time(unsigned long cur_time)
  * Increment the BDI's writeout completion count and the global writeout
  * completion count. Called from test_clear_page_writeback().
  */
-static inline void __bdi_writeout_inc(struct backing_dev_info *bdi)
+static inline void __wb_writeout_inc(struct bdi_writeback *wb)
 {
-       __inc_bdi_stat(bdi, BDI_WRITTEN);
-       __fprop_inc_percpu_max(&writeout_completions, &bdi->completions,
-                              bdi->max_prop_frac);
+       __inc_wb_stat(wb, WB_WRITTEN);
+       __fprop_inc_percpu_max(&writeout_completions, &wb->bdi->completions,
+                              wb->bdi->max_prop_frac);
        /* First event after period switching was turned off? */
        if (!unlikely(writeout_period_time)) {
                /*
@@ -414,15 +414,15 @@ static inline void __bdi_writeout_inc(struct backing_dev_info *bdi)
        }
 }
 
-void bdi_writeout_inc(struct backing_dev_info *bdi)
+void wb_writeout_inc(struct bdi_writeback *wb)
 {
        unsigned long flags;
 
        local_irq_save(flags);
-       __bdi_writeout_inc(bdi);
+       __wb_writeout_inc(wb);
        local_irq_restore(flags);
 }
-EXPORT_SYMBOL_GPL(bdi_writeout_inc);
+EXPORT_SYMBOL_GPL(wb_writeout_inc);
 
 /*
  * Obtain an accurate fraction of the BDI's portion.
@@ -1130,8 +1130,8 @@ void __bdi_update_bandwidth(struct backing_dev_info *bdi,
        if (elapsed < BANDWIDTH_INTERVAL)
                return;
 
-       dirtied = percpu_counter_read(&bdi->bdi_stat[BDI_DIRTIED]);
-       written = percpu_counter_read(&bdi->bdi_stat[BDI_WRITTEN]);
+       dirtied = percpu_counter_read(&bdi->wb.stat[WB_DIRTIED]);
+       written = percpu_counter_read(&bdi->wb.stat[WB_WRITTEN]);
 
        /*
         * Skip quiet periods when disk bandwidth is under-utilized.
@@ -1288,7 +1288,8 @@ static inline void bdi_dirty_limits(struct backing_dev_info *bdi,
                                    unsigned long *bdi_thresh,
                                    unsigned long *bdi_bg_thresh)
 {
-       unsigned long bdi_reclaimable;
+       struct bdi_writeback *wb = &bdi->wb;
+       unsigned long wb_reclaimable;
 
        /*
         * bdi_thresh is not treated as some limiting factor as
@@ -1320,14 +1321,12 @@ static inline void bdi_dirty_limits(struct backing_dev_info *bdi,
         * actually dirty; with m+n sitting in the percpu
         * deltas.
         */
-       if (*bdi_thresh < 2 * bdi_stat_error(bdi)) {
-               bdi_reclaimable = bdi_stat_sum(bdi, BDI_RECLAIMABLE);
-               *bdi_dirty = bdi_reclaimable +
-                       bdi_stat_sum(bdi, BDI_WRITEBACK);
+       if (*bdi_thresh < 2 * wb_stat_error(wb)) {
+               wb_reclaimable = wb_stat_sum(wb, WB_RECLAIMABLE);
+               *bdi_dirty = wb_reclaimable + wb_stat_sum(wb, WB_WRITEBACK);
        } else {
-               bdi_reclaimable = bdi_stat(bdi, BDI_RECLAIMABLE);
-               *bdi_dirty = bdi_reclaimable +
-                       bdi_stat(bdi, BDI_WRITEBACK);
+               wb_reclaimable = wb_stat(wb, WB_RECLAIMABLE);
+               *bdi_dirty = wb_reclaimable + wb_stat(wb, WB_WRITEBACK);
        }
 }
 
@@ -1514,9 +1513,9 @@ pause:
                 * In theory 1 page is enough to keep the comsumer-producer
                 * pipe going: the flusher cleans 1 page => the task dirties 1
                 * more page. However bdi_dirty has accounting errors.  So use
-                * the larger and more IO friendly bdi_stat_error.
+                * the larger and more IO friendly wb_stat_error.
                 */
-               if (bdi_dirty <= bdi_stat_error(bdi))
+               if (bdi_dirty <= wb_stat_error(&bdi->wb))
                        break;
 
                if (fatal_signal_pending(current))
@@ -2106,8 +2105,8 @@ void account_page_dirtied(struct page *page, struct address_space *mapping,
                mem_cgroup_inc_page_stat(memcg, MEM_CGROUP_STAT_DIRTY);
                __inc_zone_page_state(page, NR_FILE_DIRTY);
                __inc_zone_page_state(page, NR_DIRTIED);
-               __inc_bdi_stat(bdi, BDI_RECLAIMABLE);
-               __inc_bdi_stat(bdi, BDI_DIRTIED);
+               __inc_wb_stat(&bdi->wb, WB_RECLAIMABLE);
+               __inc_wb_stat(&bdi->wb, WB_DIRTIED);
                task_io_account_write(PAGE_CACHE_SIZE);
                current->nr_dirtied++;
                this_cpu_inc(bdp_ratelimits);
@@ -2126,7 +2125,7 @@ void account_page_cleaned(struct page *page, struct address_space *mapping,
        if (mapping_cap_account_dirty(mapping)) {
                mem_cgroup_dec_page_stat(memcg, MEM_CGROUP_STAT_DIRTY);
                dec_zone_page_state(page, NR_FILE_DIRTY);
-               dec_bdi_stat(inode_to_bdi(mapping->host), BDI_RECLAIMABLE);
+               dec_wb_stat(&inode_to_bdi(mapping->host)->wb, WB_RECLAIMABLE);
                task_io_account_cancelled_write(PAGE_CACHE_SIZE);
        }
 }
@@ -2190,7 +2189,7 @@ void account_page_redirty(struct page *page)
        if (mapping && mapping_cap_account_dirty(mapping)) {
                current->nr_dirtied--;
                dec_zone_page_state(page, NR_DIRTIED);
-               dec_bdi_stat(inode_to_bdi(mapping->host), BDI_DIRTIED);
+               dec_wb_stat(&inode_to_bdi(mapping->host)->wb, WB_DIRTIED);
        }
 }
 EXPORT_SYMBOL(account_page_redirty);
@@ -2369,8 +2368,8 @@ int clear_page_dirty_for_io(struct page *page)
                if (TestClearPageDirty(page)) {
                        mem_cgroup_dec_page_stat(memcg, MEM_CGROUP_STAT_DIRTY);
                        dec_zone_page_state(page, NR_FILE_DIRTY);
-                       dec_bdi_stat(inode_to_bdi(mapping->host),
-                                       BDI_RECLAIMABLE);
+                       dec_wb_stat(&inode_to_bdi(mapping->host)->wb,
+                                   WB_RECLAIMABLE);
                        ret = 1;
                }
                mem_cgroup_end_page_stat(memcg);
@@ -2398,8 +2397,8 @@ int test_clear_page_writeback(struct page *page)
                                                page_index(page),
                                                PAGECACHE_TAG_WRITEBACK);
                        if (bdi_cap_account_writeback(bdi)) {
-                               __dec_bdi_stat(bdi, BDI_WRITEBACK);
-                               __bdi_writeout_inc(bdi);
+                               __dec_wb_stat(&bdi->wb, WB_WRITEBACK);
+                               __wb_writeout_inc(&bdi->wb);
                        }
                }
                spin_unlock_irqrestore(&mapping->tree_lock, flags);
@@ -2433,7 +2432,7 @@ int __test_set_page_writeback(struct page *page, bool keep_write)
                                                page_index(page),
                                                PAGECACHE_TAG_WRITEBACK);
                        if (bdi_cap_account_writeback(bdi))
-                               __inc_bdi_stat(bdi, BDI_WRITEBACK);
+                               __inc_wb_stat(&bdi->wb, WB_WRITEBACK);
                }
                if (!PageDirty(page))
                        radix_tree_tag_clear(&mapping->page_tree,