writeback: move backing_dev_info->state into bdi_writeback
authorTejun Heo <tj@kernel.org>
Fri, 22 May 2015 21:13:26 +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->state into wb.

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

* Explicit zeroing of bdi->state is removed without adding zeoring of
  wb->state as the whole data structure is zeroed on init anyway.

* As there's still only one bdi_writeback per backing_dev_info, all
  uses of bdi->state are mechanically replaced with bdi->wb.state
  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: drbd-dev@lists.linbit.com
Cc: Neil Brown <neilb@suse.de>
Cc: Alasdair Kergon <agk@redhat.com>
Cc: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
block/blk-core.c
drivers/block/drbd/drbd_main.c
drivers/md/dm.c
drivers/md/raid1.c
drivers/md/raid10.c
fs/fs-writeback.c
include/linux/backing-dev.h
mm/backing-dev.c

index 4afac14d44993a7cec1c39a3f4185192740d2d47..c3ba9c3b08d4830686ea3808d73b83ee7d5f6cd8 100644 (file)
@@ -621,7 +621,6 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
 
        q->backing_dev_info.ra_pages =
                        (VM_MAX_READAHEAD * 1024) / PAGE_CACHE_SIZE;
-       q->backing_dev_info.state = 0;
        q->backing_dev_info.capabilities = 0;
        q->backing_dev_info.name = "block";
        q->node = node_id;
index 81fde9ef7f8e0e21e0d8bba275e14bdde5f90884..a1518539b8580420ea1d278c1daeeeb3e3721ae8 100644 (file)
@@ -2359,7 +2359,7 @@ static void drbd_cleanup(void)
  * @congested_data:    User data
  * @bdi_bits:          Bits the BDI flusher thread is currently interested in
  *
- * Returns 1<<BDI_async_congested and/or 1<<BDI_sync_congested if we are congested.
+ * Returns 1<<WB_async_congested and/or 1<<WB_sync_congested if we are congested.
  */
 static int drbd_congested(void *congested_data, int bdi_bits)
 {
@@ -2376,14 +2376,14 @@ static int drbd_congested(void *congested_data, int bdi_bits)
        }
 
        if (test_bit(CALLBACK_PENDING, &first_peer_device(device)->connection->flags)) {
-               r |= (1 << BDI_async_congested);
+               r |= (1 << WB_async_congested);
                /* Without good local data, we would need to read from remote,
                 * and that would need the worker thread as well, which is
                 * currently blocked waiting for that usermode helper to
                 * finish.
                 */
                if (!get_ldev_if_state(device, D_UP_TO_DATE))
-                       r |= (1 << BDI_sync_congested);
+                       r |= (1 << WB_sync_congested);
                else
                        put_ldev(device);
                r &= bdi_bits;
@@ -2399,9 +2399,9 @@ static int drbd_congested(void *congested_data, int bdi_bits)
                        reason = 'b';
        }
 
-       if (bdi_bits & (1 << BDI_async_congested) &&
+       if (bdi_bits & (1 << WB_async_congested) &&
            test_bit(NET_CONGESTED, &first_peer_device(device)->connection->flags)) {
-               r |= (1 << BDI_async_congested);
+               r |= (1 << WB_async_congested);
                reason = reason == 'b' ? 'a' : 'n';
        }
 
index 38837f8ea3279c722cc0a929085e845357825149..2161ed9329c41e95969415a23ec76af87a3476b1 100644 (file)
@@ -2074,7 +2074,7 @@ static int dm_any_congested(void *congested_data, int bdi_bits)
                         * the query about congestion status of request_queue
                         */
                        if (dm_request_based(md))
-                               r = md->queue->backing_dev_info.state &
+                               r = md->queue->backing_dev_info.wb.state &
                                    bdi_bits;
                        else
                                r = dm_table_any_congested(map, bdi_bits);
index 9157a29c8dbf133c713d67bdaaea45aaf650a255..f80f1af61ce70bce15a72d242d87c1c34da49a79 100644 (file)
@@ -745,7 +745,7 @@ static int raid1_congested(struct mddev *mddev, int bits)
        struct r1conf *conf = mddev->private;
        int i, ret = 0;
 
-       if ((bits & (1 << BDI_async_congested)) &&
+       if ((bits & (1 << WB_async_congested)) &&
            conf->pending_count >= max_queued_requests)
                return 1;
 
@@ -760,7 +760,7 @@ static int raid1_congested(struct mddev *mddev, int bits)
                        /* Note the '|| 1' - when read_balance prefers
                         * non-congested targets, it can be removed
                         */
-                       if ((bits & (1<<BDI_async_congested)) || 1)
+                       if ((bits & (1 << WB_async_congested)) || 1)
                                ret |= bdi_congested(&q->backing_dev_info, bits);
                        else
                                ret &= bdi_congested(&q->backing_dev_info, bits);
index e793ab6b35705e0ed1ad6904ebe9353b6dbf6fd6..fca825718f29a2ba6d5d39789ba6593699b3aec9 100644 (file)
@@ -914,7 +914,7 @@ static int raid10_congested(struct mddev *mddev, int bits)
        struct r10conf *conf = mddev->private;
        int i, ret = 0;
 
-       if ((bits & (1 << BDI_async_congested)) &&
+       if ((bits & (1 << WB_async_congested)) &&
            conf->pending_count >= max_queued_requests)
                return 1;
 
index 32a8bbd7a9ad1121f9b100da08f80500736bcebf..983312cea24546077d8692b1961515bf82852def 100644 (file)
@@ -74,7 +74,7 @@ unsigned int dirtytime_expire_interval = 12 * 60 * 60;
  */
 int writeback_in_progress(struct backing_dev_info *bdi)
 {
-       return test_bit(BDI_writeback_running, &bdi->state);
+       return test_bit(WB_writeback_running, &bdi->wb.state);
 }
 EXPORT_SYMBOL(writeback_in_progress);
 
@@ -112,7 +112,7 @@ EXPORT_TRACEPOINT_SYMBOL_GPL(wbc_writepage);
 static void bdi_wakeup_thread(struct backing_dev_info *bdi)
 {
        spin_lock_bh(&bdi->wb_lock);
-       if (test_bit(BDI_registered, &bdi->state))
+       if (test_bit(WB_registered, &bdi->wb.state))
                mod_delayed_work(bdi_wq, &bdi->wb.dwork, 0);
        spin_unlock_bh(&bdi->wb_lock);
 }
@@ -123,7 +123,7 @@ static void bdi_queue_work(struct backing_dev_info *bdi,
        trace_writeback_queue(bdi, work);
 
        spin_lock_bh(&bdi->wb_lock);
-       if (!test_bit(BDI_registered, &bdi->state)) {
+       if (!test_bit(WB_registered, &bdi->wb.state)) {
                if (work->done)
                        complete(work->done);
                goto out_unlock;
@@ -1057,7 +1057,7 @@ static long wb_do_writeback(struct bdi_writeback *wb)
        struct wb_writeback_work *work;
        long wrote = 0;
 
-       set_bit(BDI_writeback_running, &wb->bdi->state);
+       set_bit(WB_writeback_running, &wb->state);
        while ((work = get_next_work_item(bdi)) != NULL) {
 
                trace_writeback_exec(bdi, work);
@@ -1079,7 +1079,7 @@ static long wb_do_writeback(struct bdi_writeback *wb)
         */
        wrote += wb_check_old_data_flush(wb);
        wrote += wb_check_background_flush(wb);
-       clear_bit(BDI_writeback_running, &wb->bdi->state);
+       clear_bit(WB_writeback_running, &wb->state);
 
        return wrote;
 }
@@ -1099,7 +1099,7 @@ void bdi_writeback_workfn(struct work_struct *work)
        current->flags |= PF_SWAPWRITE;
 
        if (likely(!current_is_workqueue_rescuer() ||
-                  !test_bit(BDI_registered, &bdi->state))) {
+                  !test_bit(WB_registered, &wb->state))) {
                /*
                 * The normal path.  Keep writing back @bdi until its
                 * work_list is empty.  Note that this path is also taken
@@ -1323,7 +1323,7 @@ void __mark_inode_dirty(struct inode *inode, int flags)
                        spin_unlock(&inode->i_lock);
                        spin_lock(&bdi->wb.list_lock);
                        if (bdi_cap_writeback_dirty(bdi)) {
-                               WARN(!test_bit(BDI_registered, &bdi->state),
+                               WARN(!test_bit(WB_registered, &bdi->wb.state),
                                     "bdi-%s not registered\n", bdi->name);
 
                                /*
index aff923ae8c4b963272563759b9ac52ad55778bd0..eb14f988a63ed6667e8b72bac76bcb294dcd34c5 100644 (file)
@@ -25,13 +25,13 @@ struct device;
 struct dentry;
 
 /*
- * Bits in backing_dev_info.state
+ * Bits in bdi_writeback.state
  */
-enum bdi_state {
-       BDI_async_congested,    /* The async (write) queue is getting full */
-       BDI_sync_congested,     /* The sync queue is getting full */
-       BDI_registered,         /* bdi_register() was done */
-       BDI_writeback_running,  /* Writeback is in progress */
+enum wb_state {
+       WB_async_congested,     /* The async (write) queue is getting full */
+       WB_sync_congested,      /* The sync queue is getting full */
+       WB_registered,          /* bdi_register() was done */
+       WB_writeback_running,   /* Writeback is in progress */
 };
 
 typedef int (congested_fn)(void *, int);
@@ -49,6 +49,7 @@ enum bdi_stat_item {
 struct bdi_writeback {
        struct backing_dev_info *bdi;   /* our parent bdi */
 
+       unsigned long state;            /* Always use atomic bitops on this */
        unsigned long last_old_flush;   /* last old data flush */
 
        struct delayed_work dwork;      /* work item used for writeback */
@@ -62,7 +63,6 @@ struct bdi_writeback {
 struct backing_dev_info {
        struct list_head bdi_list;
        unsigned long ra_pages; /* max readahead in PAGE_CACHE_SIZE units */
-       unsigned long state;    /* Always use atomic bitops on this */
        unsigned int capabilities; /* Device capabilities */
        congested_fn *congested_fn; /* Function pointer if device is md/dm */
        void *congested_data;   /* Pointer to aux data for congested func */
@@ -250,23 +250,23 @@ static inline int bdi_congested(struct backing_dev_info *bdi, int bdi_bits)
 {
        if (bdi->congested_fn)
                return bdi->congested_fn(bdi->congested_data, bdi_bits);
-       return (bdi->state & bdi_bits);
+       return (bdi->wb.state & bdi_bits);
 }
 
 static inline int bdi_read_congested(struct backing_dev_info *bdi)
 {
-       return bdi_congested(bdi, 1 << BDI_sync_congested);
+       return bdi_congested(bdi, 1 << WB_sync_congested);
 }
 
 static inline int bdi_write_congested(struct backing_dev_info *bdi)
 {
-       return bdi_congested(bdi, 1 << BDI_async_congested);
+       return bdi_congested(bdi, 1 << WB_async_congested);
 }
 
 static inline int bdi_rw_congested(struct backing_dev_info *bdi)
 {
-       return bdi_congested(bdi, (1 << BDI_sync_congested) |
-                                 (1 << BDI_async_congested));
+       return bdi_congested(bdi, (1 << WB_sync_congested) |
+                                 (1 << WB_async_congested));
 }
 
 enum {
index 6dc4580df2af040b10bc10a5f9c423becc3ff47e..b23cf0ea5912b4475ff2923c105fda0808fec9a7 100644 (file)
@@ -96,7 +96,7 @@ static int bdi_debug_stats_show(struct seq_file *m, void *v)
                   nr_io,
                   nr_more_io,
                   nr_dirty_time,
-                  !list_empty(&bdi->bdi_list), bdi->state);
+                  !list_empty(&bdi->bdi_list), bdi->wb.state);
 #undef K
 
        return 0;
@@ -280,7 +280,7 @@ void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi)
 
        timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
        spin_lock_bh(&bdi->wb_lock);
-       if (test_bit(BDI_registered, &bdi->state))
+       if (test_bit(WB_registered, &bdi->wb.state))
                queue_delayed_work(bdi_wq, &bdi->wb.dwork, timeout);
        spin_unlock_bh(&bdi->wb_lock);
 }
@@ -315,7 +315,7 @@ int bdi_register(struct backing_dev_info *bdi, struct device *parent,
        bdi->dev = dev;
 
        bdi_debug_register(bdi, dev_name(dev));
-       set_bit(BDI_registered, &bdi->state);
+       set_bit(WB_registered, &bdi->wb.state);
 
        spin_lock_bh(&bdi_lock);
        list_add_tail_rcu(&bdi->bdi_list, &bdi_list);
@@ -339,7 +339,7 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
 {
        /* Make sure nobody queues further work */
        spin_lock_bh(&bdi->wb_lock);
-       if (!test_and_clear_bit(BDI_registered, &bdi->state)) {
+       if (!test_and_clear_bit(WB_registered, &bdi->wb.state)) {
                spin_unlock_bh(&bdi->wb_lock);
                return;
        }
@@ -492,11 +492,11 @@ static atomic_t nr_bdi_congested[2];
 
 void clear_bdi_congested(struct backing_dev_info *bdi, int sync)
 {
-       enum bdi_state bit;
+       enum wb_state bit;
        wait_queue_head_t *wqh = &congestion_wqh[sync];
 
-       bit = sync ? BDI_sync_congested : BDI_async_congested;
-       if (test_and_clear_bit(bit, &bdi->state))
+       bit = sync ? WB_sync_congested : WB_async_congested;
+       if (test_and_clear_bit(bit, &bdi->wb.state))
                atomic_dec(&nr_bdi_congested[sync]);
        smp_mb__after_atomic();
        if (waitqueue_active(wqh))
@@ -506,10 +506,10 @@ EXPORT_SYMBOL(clear_bdi_congested);
 
 void set_bdi_congested(struct backing_dev_info *bdi, int sync)
 {
-       enum bdi_state bit;
+       enum wb_state bit;
 
-       bit = sync ? BDI_sync_congested : BDI_async_congested;
-       if (!test_and_set_bit(bit, &bdi->state))
+       bit = sync ? WB_sync_congested : WB_async_congested;
+       if (!test_and_set_bit(bit, &bdi->wb.state))
                atomic_inc(&nr_bdi_congested[sync]);
 }
 EXPORT_SYMBOL(set_bdi_congested);