backing-dev: ensure wakeup_timer is deleted
[linux-2.6.git] / mm / backing-dev.c
index f50a281..253b071 100644 (file)
 
 #include <linux/wait.h>
 #include <linux/backing-dev.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
 #include <linux/fs.h>
+#include <linux/pagemap.h>
+#include <linux/mm.h>
 #include <linux/sched.h>
 #include <linux/module.h>
+#include <linux/writeback.h>
+#include <linux/device.h>
+#include <trace/events/writeback.h>
+
+static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
+
+struct backing_dev_info default_backing_dev_info = {
+       .name           = "default",
+       .ra_pages       = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE,
+       .state          = 0,
+       .capabilities   = BDI_CAP_MAP_COPY,
+};
+EXPORT_SYMBOL_GPL(default_backing_dev_info);
+
+struct backing_dev_info noop_backing_dev_info = {
+       .name           = "noop",
+       .capabilities   = BDI_CAP_NO_ACCT_AND_WRITEBACK,
+};
+EXPORT_SYMBOL_GPL(noop_backing_dev_info);
+
+static struct class *bdi_class;
+
+/*
+ * bdi_lock protects updates to bdi_list and bdi_pending_list, as well as
+ * reader side protection for bdi_pending_list. bdi_list has RCU reader side
+ * locking.
+ */
+DEFINE_SPINLOCK(bdi_lock);
+LIST_HEAD(bdi_list);
+LIST_HEAD(bdi_pending_list);
+
+static struct task_struct *sync_supers_tsk;
+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>
+
+static struct dentry *bdi_debug_root;
+
+static void bdi_debug_init(void)
+{
+       bdi_debug_root = debugfs_create_dir("bdi", NULL);
+}
+
+static int bdi_debug_stats_show(struct seq_file *m, void *v)
+{
+       struct backing_dev_info *bdi = m->private;
+       struct bdi_writeback *wb = &bdi->wb;
+       unsigned long background_thresh;
+       unsigned long dirty_thresh;
+       unsigned long bdi_thresh;
+       unsigned long nr_dirty, nr_io, nr_more_io;
+       struct inode *inode;
+
+       nr_dirty = nr_io = nr_more_io = 0;
+       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(&wb->list_lock);
+
+       global_dirty_limits(&background_thresh, &dirty_thresh);
+       bdi_thresh = bdi_dirty_limit(bdi, dirty_thresh);
+
+#define K(x) ((x) << (PAGE_SHIFT - 10))
+       seq_printf(m,
+                  "BdiWriteback:       %10lu kB\n"
+                  "BdiReclaimable:     %10lu kB\n"
+                  "BdiDirtyThresh:     %10lu kB\n"
+                  "DirtyThresh:        %10lu kB\n"
+                  "BackgroundThresh:   %10lu kB\n"
+                  "BdiWritten:         %10lu kB\n"
+                  "BdiWriteBandwidth:  %10lu kBps\n"
+                  "b_dirty:            %10lu\n"
+                  "b_io:               %10lu\n"
+                  "b_more_io:          %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)),
+                  K(bdi_thresh),
+                  K(dirty_thresh),
+                  K(background_thresh),
+                  (unsigned long) K(bdi_stat(bdi, BDI_WRITTEN)),
+                  (unsigned long) K(bdi->write_bandwidth),
+                  nr_dirty,
+                  nr_io,
+                  nr_more_io,
+                  !list_empty(&bdi->bdi_list), bdi->state);
+#undef K
+
+       return 0;
+}
+
+static int bdi_debug_stats_open(struct inode *inode, struct file *file)
+{
+       return single_open(file, bdi_debug_stats_show, inode->i_private);
+}
+
+static const struct file_operations bdi_debug_stats_fops = {
+       .open           = bdi_debug_stats_open,
+       .read           = seq_read,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+
+static void bdi_debug_register(struct backing_dev_info *bdi, const char *name)
+{
+       bdi->debug_dir = debugfs_create_dir(name, bdi_debug_root);
+       bdi->debug_stats = debugfs_create_file("stats", 0444, bdi->debug_dir,
+                                              bdi, &bdi_debug_stats_fops);
+}
+
+static void bdi_debug_unregister(struct backing_dev_info *bdi)
+{
+       debugfs_remove(bdi->debug_stats);
+       debugfs_remove(bdi->debug_dir);
+}
+#else
+static inline void bdi_debug_init(void)
+{
+}
+static inline void bdi_debug_register(struct backing_dev_info *bdi,
+                                     const char *name)
+{
+}
+static inline void bdi_debug_unregister(struct backing_dev_info *bdi)
+{
+}
+#endif
+
+static ssize_t read_ahead_kb_store(struct device *dev,
+                                 struct device_attribute *attr,
+                                 const char *buf, size_t count)
+{
+       struct backing_dev_info *bdi = dev_get_drvdata(dev);
+       char *end;
+       unsigned long read_ahead_kb;
+       ssize_t ret = -EINVAL;
+
+       read_ahead_kb = simple_strtoul(buf, &end, 10);
+       if (*buf && (end[0] == '\0' || (end[0] == '\n' && end[1] == '\0'))) {
+               bdi->ra_pages = read_ahead_kb >> (PAGE_SHIFT - 10);
+               ret = count;
+       }
+       return ret;
+}
+
+#define K(pages) ((pages) << (PAGE_SHIFT - 10))
+
+#define BDI_SHOW(name, expr)                                           \
+static ssize_t name##_show(struct device *dev,                         \
+                          struct device_attribute *attr, char *page)   \
+{                                                                      \
+       struct backing_dev_info *bdi = dev_get_drvdata(dev);            \
+                                                                       \
+       return snprintf(page, PAGE_SIZE-1, "%lld\n", (long long)expr);  \
+}
+
+BDI_SHOW(read_ahead_kb, K(bdi->ra_pages))
+
+static ssize_t min_ratio_store(struct device *dev,
+               struct device_attribute *attr, const char *buf, size_t count)
+{
+       struct backing_dev_info *bdi = dev_get_drvdata(dev);
+       char *end;
+       unsigned int ratio;
+       ssize_t ret = -EINVAL;
+
+       ratio = simple_strtoul(buf, &end, 10);
+       if (*buf && (end[0] == '\0' || (end[0] == '\n' && end[1] == '\0'))) {
+               ret = bdi_set_min_ratio(bdi, ratio);
+               if (!ret)
+                       ret = count;
+       }
+       return ret;
+}
+BDI_SHOW(min_ratio, bdi->min_ratio)
+
+static ssize_t max_ratio_store(struct device *dev,
+               struct device_attribute *attr, const char *buf, size_t count)
+{
+       struct backing_dev_info *bdi = dev_get_drvdata(dev);
+       char *end;
+       unsigned int ratio;
+       ssize_t ret = -EINVAL;
+
+       ratio = simple_strtoul(buf, &end, 10);
+       if (*buf && (end[0] == '\0' || (end[0] == '\n' && end[1] == '\0'))) {
+               ret = bdi_set_max_ratio(bdi, ratio);
+               if (!ret)
+                       ret = count;
+       }
+       return ret;
+}
+BDI_SHOW(max_ratio, bdi->max_ratio)
+
+#define __ATTR_RW(attr) __ATTR(attr, 0644, attr##_show, attr##_store)
+
+static struct device_attribute bdi_dev_attrs[] = {
+       __ATTR_RW(read_ahead_kb),
+       __ATTR_RW(min_ratio),
+       __ATTR_RW(max_ratio),
+       __ATTR_NULL,
+};
+
+static __init int bdi_class_init(void)
+{
+       bdi_class = class_create(THIS_MODULE, "bdi");
+       if (IS_ERR(bdi_class))
+               return PTR_ERR(bdi_class);
+
+       bdi_class->dev_attrs = bdi_dev_attrs;
+       bdi_debug_init();
+       return 0;
+}
+postcore_initcall(bdi_class_init);
+
+static int __init default_bdi_init(void)
+{
+       int err;
+
+       sync_supers_tsk = kthread_run(bdi_sync_supers, NULL, "sync_supers");
+       BUG_ON(IS_ERR(sync_supers_tsk));
+
+       setup_timer(&sync_supers_timer, sync_supers_timer_fn, 0);
+       bdi_arm_supers_timer();
+
+       err = bdi_init(&default_backing_dev_info);
+       if (!err)
+               bdi_register(&default_backing_dev_info, NULL, "default");
+       err = bdi_init(&noop_backing_dev_info);
+
+       return err;
+}
+subsys_initcall(default_bdi_init);
+
+int bdi_has_dirty_io(struct backing_dev_info *bdi)
+{
+       return wb_has_dirty_io(&bdi->wb);
+}
+
+/*
+ * kupdated() used to do this. We cannot do it from the bdi_forker_thread()
+ * or we risk deadlocking on ->s_umount. The longer term solution would be
+ * to implement sync_supers_bdi() or similar and simply do it from the
+ * bdi writeback thread individually.
+ */
+static int bdi_sync_supers(void *unused)
+{
+       set_user_nice(current, 0);
+
+       while (!kthread_should_stop()) {
+               set_current_state(TASK_INTERRUPTIBLE);
+               schedule();
+
+               /*
+                * Do this periodically, like kupdated() did before.
+                */
+               sync_supers();
+       }
+
+       return 0;
+}
+
+void bdi_arm_supers_timer(void)
+{
+       unsigned long next;
+
+       if (!dirty_writeback_interval)
+               return;
+
+       next = msecs_to_jiffies(dirty_writeback_interval * 10) + jiffies;
+       mod_timer(&sync_supers_timer, round_jiffies_up(next));
+}
+
+static void sync_supers_timer_fn(unsigned long unused)
+{
+       wake_up_process(sync_supers_tsk);
+       bdi_arm_supers_timer();
+}
+
+static void wakeup_timer_fn(unsigned long data)
+{
+       struct backing_dev_info *bdi = (struct backing_dev_info *)data;
+
+       spin_lock_bh(&bdi->wb_lock);
+       if (bdi->wb.task) {
+               trace_writeback_wake_thread(bdi);
+               wake_up_process(bdi->wb.task);
+       } else {
+               /*
+                * When bdi tasks are inactive for long time, they are killed.
+                * In this case we have to wake-up the forker thread which
+                * should create and run the bdi thread.
+                */
+               trace_writeback_wake_forker_thread(bdi);
+               wake_up_process(default_backing_dev_info.wb.task);
+       }
+       spin_unlock_bh(&bdi->wb_lock);
+}
+
+/*
+ * This function is used when the first inode for this bdi is marked dirty. It
+ * wakes-up the corresponding bdi thread which should then take care of the
+ * periodic background write-out of dirty inodes. Since the write-out would
+ * starts only 'dirty_writeback_interval' centisecs from now anyway, we just
+ * set up a timer which wakes the bdi thread up later.
+ *
+ * Note, we wouldn't bother setting up the timer, but this function is on the
+ * fast-path (used by '__mark_inode_dirty()'), so we save few context switches
+ * by delaying the wake-up.
+ */
+void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi)
+{
+       unsigned long timeout;
+
+       timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
+       mod_timer(&bdi->wb.wakeup_timer, jiffies + timeout);
+}
+
+/*
+ * Calculate the longest interval (jiffies) bdi threads are allowed to be
+ * inactive.
+ */
+static unsigned long bdi_longest_inactive(void)
+{
+       unsigned long interval;
+
+       interval = msecs_to_jiffies(dirty_writeback_interval * 10);
+       return max(5UL * 60 * HZ, interval);
+}
+
+/*
+ * Clear pending bit and wakeup anybody waiting for flusher thread creation or
+ * shutdown
+ */
+static void bdi_clear_pending(struct backing_dev_info *bdi)
+{
+       clear_bit(BDI_pending, &bdi->state);
+       smp_mb__after_clear_bit();
+       wake_up_bit(&bdi->state, BDI_pending);
+}
+
+static int bdi_forker_thread(void *ptr)
+{
+       struct bdi_writeback *me = ptr;
+
+       current->flags |= PF_SWAPWRITE;
+       set_freezable();
+
+       /*
+        * Our parent may run at a different priority, just set us to normal
+        */
+       set_user_nice(current, 0);
+
+       for (;;) {
+               struct task_struct *task = NULL;
+               struct backing_dev_info *bdi;
+               enum {
+                       NO_ACTION,   /* Nothing to do */
+                       FORK_THREAD, /* Fork bdi thread */
+                       KILL_THREAD, /* Kill inactive bdi thread */
+               } action = NO_ACTION;
+
+               /*
+                * Temporary measure, we want to make sure we don't see
+                * dirty data on the default backing_dev_info
+                */
+               if (wb_has_dirty_io(me) || !list_empty(&me->bdi->work_list)) {
+                       del_timer(&me->wakeup_timer);
+                       wb_do_writeback(me, 0);
+               }
+
+               spin_lock_bh(&bdi_lock);
+               /*
+                * In the following loop we are going to check whether we have
+                * some work to do without any synchronization with tasks
+                * waking us up to do work for them. So we have to set task
+                * state already here so that we don't miss wakeups coming
+                * after we verify some condition.
+                */
+               set_current_state(TASK_INTERRUPTIBLE);
+
+               list_for_each_entry(bdi, &bdi_list, bdi_list) {
+                       bool have_dirty_io;
+
+                       if (!bdi_cap_writeback_dirty(bdi) ||
+                            bdi_cap_flush_forker(bdi))
+                               continue;
+
+                       WARN(!test_bit(BDI_registered, &bdi->state),
+                            "bdi %p/%s is not registered!\n", bdi, bdi->name);
+
+                       have_dirty_io = !list_empty(&bdi->work_list) ||
+                                       wb_has_dirty_io(&bdi->wb);
+
+                       /*
+                        * If the bdi has work to do, but the thread does not
+                        * exist - create it.
+                        */
+                       if (!bdi->wb.task && have_dirty_io) {
+                               /*
+                                * Set the pending bit - if someone will try to
+                                * unregister this bdi - it'll wait on this bit.
+                                */
+                               set_bit(BDI_pending, &bdi->state);
+                               action = FORK_THREAD;
+                               break;
+                       }
+
+                       spin_lock(&bdi->wb_lock);
+
+                       /*
+                        * If there is no work to do and the bdi thread was
+                        * inactive long enough - kill it. The wb_lock is taken
+                        * to make sure no-one adds more work to this bdi and
+                        * wakes the bdi thread up.
+                        */
+                       if (bdi->wb.task && !have_dirty_io &&
+                           time_after(jiffies, bdi->wb.last_active +
+                                               bdi_longest_inactive())) {
+                               task = bdi->wb.task;
+                               bdi->wb.task = NULL;
+                               spin_unlock(&bdi->wb_lock);
+                               set_bit(BDI_pending, &bdi->state);
+                               action = KILL_THREAD;
+                               break;
+                       }
+                       spin_unlock(&bdi->wb_lock);
+               }
+               spin_unlock_bh(&bdi_lock);
+
+               /* Keep working if default bdi still has things to do */
+               if (!list_empty(&me->bdi->work_list))
+                       __set_current_state(TASK_RUNNING);
+
+               switch (action) {
+               case FORK_THREAD:
+                       __set_current_state(TASK_RUNNING);
+                       task = kthread_create(bdi_writeback_thread, &bdi->wb,
+                                             "flush-%s", dev_name(bdi->dev));
+                       if (IS_ERR(task)) {
+                               /*
+                                * If thread creation fails, force writeout of
+                                * the bdi from the thread. Hopefully 1024 is
+                                * large enough for efficient IO.
+                                */
+                               writeback_inodes_wb(&bdi->wb, 1024);
+                       } else {
+                               /*
+                                * The spinlock makes sure we do not lose
+                                * wake-ups when racing with 'bdi_queue_work()'.
+                                * And as soon as the bdi thread is visible, we
+                                * can start it.
+                                */
+                               spin_lock_bh(&bdi->wb_lock);
+                               bdi->wb.task = task;
+                               spin_unlock_bh(&bdi->wb_lock);
+                               wake_up_process(task);
+                       }
+                       bdi_clear_pending(bdi);
+                       break;
+
+               case KILL_THREAD:
+                       __set_current_state(TASK_RUNNING);
+                       kthread_stop(task);
+                       bdi_clear_pending(bdi);
+                       break;
+
+               case NO_ACTION:
+                       if (!wb_has_dirty_io(me) || !dirty_writeback_interval)
+                               /*
+                                * There are no dirty data. The only thing we
+                                * should now care about is checking for
+                                * inactive bdi threads and killing them. Thus,
+                                * let's sleep for longer time, save energy and
+                                * be friendly for battery-driven devices.
+                                */
+                               schedule_timeout(bdi_longest_inactive());
+                       else
+                               schedule_timeout(msecs_to_jiffies(dirty_writeback_interval * 10));
+                       try_to_freeze();
+                       break;
+               }
+       }
+
+       return 0;
+}
+
+/*
+ * Remove bdi from bdi_list, and ensure that it is no longer visible
+ */
+static void bdi_remove_from_list(struct backing_dev_info *bdi)
+{
+       spin_lock_bh(&bdi_lock);
+       list_del_rcu(&bdi->bdi_list);
+       spin_unlock_bh(&bdi_lock);
+
+       synchronize_rcu_expedited();
+}
+
+int bdi_register(struct backing_dev_info *bdi, struct device *parent,
+               const char *fmt, ...)
+{
+       va_list args;
+       struct device *dev;
+
+       if (bdi->dev)   /* The driver needs to use separate queues per device */
+               return 0;
+
+       va_start(args, fmt);
+       dev = device_create_vargs(bdi_class, parent, MKDEV(0, 0), bdi, fmt, args);
+       va_end(args);
+       if (IS_ERR(dev))
+               return PTR_ERR(dev);
+
+       bdi->dev = dev;
+
+       /*
+        * Just start the forker thread for our default backing_dev_info,
+        * and add other bdi's to the list. They will get a thread created
+        * on-demand when they need it.
+        */
+       if (bdi_cap_flush_forker(bdi)) {
+               struct bdi_writeback *wb = &bdi->wb;
+
+               wb->task = kthread_run(bdi_forker_thread, wb, "bdi-%s",
+                                               dev_name(dev));
+               if (IS_ERR(wb->task))
+                       return PTR_ERR(wb->task);
+       }
+
+       bdi_debug_register(bdi, dev_name(dev));
+       set_bit(BDI_registered, &bdi->state);
+
+       spin_lock_bh(&bdi_lock);
+       list_add_tail_rcu(&bdi->bdi_list, &bdi_list);
+       spin_unlock_bh(&bdi_lock);
+
+       trace_writeback_bdi_register(bdi);
+       return 0;
+}
+EXPORT_SYMBOL(bdi_register);
+
+int bdi_register_dev(struct backing_dev_info *bdi, dev_t dev)
+{
+       return bdi_register(bdi, NULL, "%u:%u", MAJOR(dev), MINOR(dev));
+}
+EXPORT_SYMBOL(bdi_register_dev);
+
+/*
+ * Remove bdi from the global list and shutdown any threads we have running
+ */
+static void bdi_wb_shutdown(struct backing_dev_info *bdi)
+{
+       if (!bdi_cap_writeback_dirty(bdi))
+               return;
+
+       /*
+        * Make sure nobody finds us on the bdi_list anymore
+        */
+       bdi_remove_from_list(bdi);
+
+       /*
+        * If setup is pending, wait for that to complete first
+        */
+       wait_on_bit(&bdi->state, BDI_pending, bdi_sched_wait,
+                       TASK_UNINTERRUPTIBLE);
+
+       /*
+        * Finally, kill the kernel thread. We don't need to be RCU
+        * safe anymore, since the bdi is gone from visibility. Force
+        * unfreeze of the thread before calling kthread_stop(), otherwise
+        * it would never exet if it is currently stuck in the refrigerator.
+        */
+       if (bdi->wb.task) {
+               thaw_process(bdi->wb.task);
+               kthread_stop(bdi->wb.task);
+       }
+}
+
+/*
+ * This bdi is going away now, make sure that no super_blocks point to it
+ */
+static void bdi_prune_sb(struct backing_dev_info *bdi)
+{
+       struct super_block *sb;
+
+       spin_lock(&sb_lock);
+       list_for_each_entry(sb, &super_blocks, s_list) {
+               if (sb->s_bdi == bdi)
+                       sb->s_bdi = &default_backing_dev_info;
+       }
+       spin_unlock(&sb_lock);
+}
+
+void bdi_unregister(struct backing_dev_info *bdi)
+{
+       if (bdi->dev) {
+               bdi_set_min_ratio(bdi, 0);
+               trace_writeback_bdi_unregister(bdi);
+               bdi_prune_sb(bdi);
+               del_timer_sync(&bdi->wb.wakeup_timer);
+
+               if (!bdi_cap_flush_forker(bdi))
+                       bdi_wb_shutdown(bdi);
+               bdi_debug_unregister(bdi);
+               device_unregister(bdi->dev);
+               bdi->dev = NULL;
+       }
+}
+EXPORT_SYMBOL(bdi_unregister);
+
+static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
+{
+       memset(wb, 0, sizeof(*wb));
+
+       wb->bdi = bdi;
+       wb->last_old_flush = jiffies;
+       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);
+}
+
+/*
+ * Initial write bandwidth: 100 MB/s
+ */
+#define INIT_BW                (100 << (20 - PAGE_SHIFT))
+
+int bdi_init(struct backing_dev_info *bdi)
+{
+       int i, err;
+
+       bdi->dev = NULL;
+
+       bdi->min_ratio = 0;
+       bdi->max_ratio = 100;
+       bdi->max_prop_frac = PROP_FRAC_BASE;
+       spin_lock_init(&bdi->wb_lock);
+       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);
+               if (err)
+                       goto err;
+       }
+
+       bdi->dirty_exceeded = 0;
+
+       bdi->bw_time_stamp = jiffies;
+       bdi->written_stamp = 0;
+
+       bdi->write_bandwidth = INIT_BW;
+       bdi->avg_write_bandwidth = INIT_BW;
+
+       err = prop_local_init_percpu(&bdi->completions);
+
+       if (err) {
+err:
+               while (i--)
+                       percpu_counter_destroy(&bdi->bdi_stat[i]);
+       }
+
+       return err;
+}
+EXPORT_SYMBOL(bdi_init);
+
+void bdi_destroy(struct backing_dev_info *bdi)
+{
+       int i;
+
+       /*
+        * Splice our entries to the default_backing_dev_info, if this
+        * bdi disappears
+        */
+       if (bdi_has_dirty_io(bdi)) {
+               struct bdi_writeback *dst = &default_backing_dev_info.wb;
+
+               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(&bdi->wb.list_lock);
+               spin_unlock(&dst->list_lock);
+       }
+
+       bdi_unregister(bdi);
+
+       /*
+        * If bdi_unregister() had already been called earlier, the
+        * wakeup_timer could still be armed because bdi_prune_sb()
+        * can race with the bdi_wakeup_thread_delayed() calls from
+        * __mark_inode_dirty().
+        */
+       del_timer_sync(&bdi->wb.wakeup_timer);
+
+       for (i = 0; i < NR_BDI_STAT_ITEMS; i++)
+               percpu_counter_destroy(&bdi->bdi_stat[i]);
+
+       prop_local_destroy_percpu(&bdi->completions);
+}
+EXPORT_SYMBOL(bdi_destroy);
+
+/*
+ * For use from filesystems to quickly init and register a bdi associated
+ * with dirty writeback
+ */
+int bdi_setup_and_register(struct backing_dev_info *bdi, char *name,
+                          unsigned int cap)
+{
+       char tmp[32];
+       int err;
+
+       bdi->name = name;
+       bdi->capabilities = cap;
+       err = bdi_init(bdi);
+       if (err)
+               return err;
+
+       sprintf(tmp, "%.28s%s", name, "-%d");
+       err = bdi_register(bdi, NULL, tmp, atomic_long_inc_return(&bdi_seq));
+       if (err) {
+               bdi_destroy(bdi);
+               return err;
+       }
+
+       return 0;
+}
+EXPORT_SYMBOL(bdi_setup_and_register);
 
 static wait_queue_head_t congestion_wqh[2] = {
                __WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[0]),
                __WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[1])
        };
+static atomic_t nr_bdi_congested[2];
 
-
-void clear_bdi_congested(struct backing_dev_info *bdi, int rw)
+void clear_bdi_congested(struct backing_dev_info *bdi, int sync)
 {
        enum bdi_state bit;
-       wait_queue_head_t *wqh = &congestion_wqh[rw];
+       wait_queue_head_t *wqh = &congestion_wqh[sync];
 
-       bit = (rw == WRITE) ? BDI_write_congested : BDI_read_congested;
-       clear_bit(bit, &bdi->state);
+       bit = sync ? BDI_sync_congested : BDI_async_congested;
+       if (test_and_clear_bit(bit, &bdi->state))
+               atomic_dec(&nr_bdi_congested[sync]);
        smp_mb__after_clear_bit();
        if (waitqueue_active(wqh))
                wake_up(wqh);
 }
 EXPORT_SYMBOL(clear_bdi_congested);
 
-void set_bdi_congested(struct backing_dev_info *bdi, int rw)
+void set_bdi_congested(struct backing_dev_info *bdi, int sync)
 {
        enum bdi_state bit;
 
-       bit = (rw == WRITE) ? BDI_write_congested : BDI_read_congested;
-       set_bit(bit, &bdi->state);
+       bit = sync ? BDI_sync_congested : BDI_async_congested;
+       if (!test_and_set_bit(bit, &bdi->state))
+               atomic_inc(&nr_bdi_congested[sync]);
 }
 EXPORT_SYMBOL(set_bdi_congested);
 
 /**
  * congestion_wait - wait for a backing_dev to become uncongested
- * @rw: READ or WRITE
+ * @sync: SYNC or ASYNC IO
  * @timeout: timeout in jiffies
  *
  * Waits for up to @timeout jiffies for a backing_dev (any backing_dev) to exit
  * write congestion.  If no backing_devs are congested then just wait for the
  * next write to be completed.
  */
-long congestion_wait(int rw, long timeout)
+long congestion_wait(int sync, long timeout)
 {
        long ret;
+       unsigned long start = jiffies;
        DEFINE_WAIT(wait);
-       wait_queue_head_t *wqh = &congestion_wqh[rw];
+       wait_queue_head_t *wqh = &congestion_wqh[sync];
 
        prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
        ret = io_schedule_timeout(timeout);
        finish_wait(wqh, &wait);
+
+       trace_writeback_congestion_wait(jiffies_to_usecs(timeout),
+                                       jiffies_to_usecs(jiffies - start));
+
        return ret;
 }
 EXPORT_SYMBOL(congestion_wait);
 
 /**
- * congestion_end - wake up sleepers on a congested backing_dev_info
- * @rw: READ or WRITE
+ * wait_iff_congested - Conditionally wait for a backing_dev to become uncongested or a zone to complete writes
+ * @zone: A zone to check if it is heavily congested
+ * @sync: SYNC or ASYNC IO
+ * @timeout: timeout in jiffies
+ *
+ * In the event of a congested backing_dev (any backing_dev) and the given
+ * @zone has experienced recent congestion, this waits for up to @timeout
+ * jiffies for either a BDI to exit congestion of the given @sync queue
+ * or a write to complete.
+ *
+ * In the absence of zone congestion, cond_resched() is called to yield
+ * the processor if necessary but otherwise does not sleep.
+ *
+ * The return value is 0 if the sleep is for the full timeout. Otherwise,
+ * it is the number of jiffies that were still remaining when the function
+ * returned. return_value == timeout implies the function did not sleep.
  */
-void congestion_end(int rw)
+long wait_iff_congested(struct zone *zone, int sync, long timeout)
 {
-       wait_queue_head_t *wqh = &congestion_wqh[rw];
+       long ret;
+       unsigned long start = jiffies;
+       DEFINE_WAIT(wait);
+       wait_queue_head_t *wqh = &congestion_wqh[sync];
 
-       if (waitqueue_active(wqh))
-               wake_up(wqh);
+       /*
+        * If there is no congestion, or heavy congestion is not being
+        * encountered in the current zone, yield if necessary instead
+        * of sleeping on the congestion queue
+        */
+       if (atomic_read(&nr_bdi_congested[sync]) == 0 ||
+                       !zone_is_reclaim_congested(zone)) {
+               cond_resched();
+
+               /* In case we scheduled, work out time remaining */
+               ret = timeout - (jiffies - start);
+               if (ret < 0)
+                       ret = 0;
+
+               goto out;
+       }
+
+       /* Sleep until uncongested or a write happens */
+       prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
+       ret = io_schedule_timeout(timeout);
+       finish_wait(wqh, &wait);
+
+out:
+       trace_writeback_wait_iff_congested(jiffies_to_usecs(timeout),
+                                       jiffies_to_usecs(jiffies - start));
+
+       return ret;
 }
-EXPORT_SYMBOL(congestion_end);
+EXPORT_SYMBOL(wait_iff_congested);