1 files changed, 120 insertions, 13 deletions

diff --git a/drivers/md/bcache/writeback.c b/drivers/md/bcache/writeback.c
index 2714ed3991d1..22cbff551628 100644
--- a/drivers/md/bcache/writeback.c
+++ b/drivers/md/bcache/writeback.c
@@ -9,6 +9,9 @@
 #include "bcache.h"
 #include "btree.h"
 #include "debug.h"
+#include "writeback.h"
+
+#include <trace/events/bcache.h>
 
 static struct workqueue_struct *dirty_wq;
 
@@ -36,7 +39,7 @@ static void __update_writeback_rate(struct cached_dev *dc)
 
         int change = 0;
         int64_t error;
-        int64_t dirty = atomic_long_read(&dc->disk.sectors_dirty);
+        int64_t dirty = bcache_dev_sectors_dirty(&dc->disk);
         int64_t derivative = dirty - dc->disk.sectors_dirty_last;
 
         dc->disk.sectors_dirty_last = dirty;
@@ -105,6 +108,31 @@ static bool dirty_pred(struct keybuf *buf, struct bkey *k)
         return KEY_DIRTY(k);
 }
 
+static bool dirty_full_stripe_pred(struct keybuf *buf, struct bkey *k)
+{
+        uint64_t stripe;
+        unsigned nr_sectors = KEY_SIZE(k);
+        struct cached_dev *dc = container_of(buf, struct cached_dev,
+                                             writeback_keys);
+        unsigned stripe_size = 1 << dc->disk.stripe_size_bits;
+
+        if (!KEY_DIRTY(k))
+                return false;
+
+        stripe = KEY_START(k) >> dc->disk.stripe_size_bits;
+        while (1) {
+                if (atomic_read(dc->disk.stripe_sectors_dirty + stripe) !=
+                    stripe_size)
+                        return false;
+
+                if (nr_sectors <= stripe_size)
+                        return true;
+
+                nr_sectors -= stripe_size;
+                stripe++;
+        }
+}
+
 static void dirty_init(struct keybuf_key *w)
 {
         struct dirty_io *io = w->private;
@@ -149,7 +177,22 @@ static void refill_dirty(struct closure *cl)
                 searched_from_start = true;
         }
 
-        bch_refill_keybuf(dc->disk.c, buf, &end);
+        if (dc->partial_stripes_expensive) {
+                uint64_t i;
+
+                for (i = 0; i < dc->disk.nr_stripes; i++)
+                        if (atomic_read(dc->disk.stripe_sectors_dirty + i) ==
+                            1 << dc->disk.stripe_size_bits)
+                                goto full_stripes;
+
+                goto normal_refill;
+full_stripes:
+                bch_refill_keybuf(dc->disk.c, buf, &end,
+                                  dirty_full_stripe_pred);
+        } else {
+normal_refill:
+                bch_refill_keybuf(dc->disk.c, buf, &end, dirty_pred);
+        }
 
         if (bkey_cmp(&buf->last_scanned, &end) >= 0 && searched_from_start) {
                 /* Searched the entire btree  - delay awhile */
@@ -181,10 +224,8 @@ void bch_writeback_queue(struct cached_dev *dc)
         }
 }
 
-void bch_writeback_add(struct cached_dev *dc, unsigned sectors)
+void bch_writeback_add(struct cached_dev *dc)
 {
-        atomic_long_add(sectors, &dc->disk.sectors_dirty);
-
         if (!atomic_read(&dc->has_dirty) &&
             !atomic_xchg(&dc->has_dirty, 1)) {
                 atomic_inc(&dc->count);
@@ -203,6 +244,34 @@ void bch_writeback_add(struct cached_dev *dc, unsigned sectors)
         }
 }
 
+void bcache_dev_sectors_dirty_add(struct cache_set *c, unsigned inode,
+                                  uint64_t offset, int nr_sectors)
+{
+        struct bcache_device *d = c->devices[inode];
+        unsigned stripe_size, stripe_offset;
+        uint64_t stripe;
+
+        if (!d)
+                return;
+
+        stripe_size = 1 << d->stripe_size_bits;
+        stripe = offset >> d->stripe_size_bits;
+        stripe_offset = offset & (stripe_size - 1);
+
+        while (nr_sectors) {
+                int s = min_t(unsigned, abs(nr_sectors),
+                              stripe_size - stripe_offset);
+
+                if (nr_sectors < 0)
+                        s = -s;
+
+                atomic_add(s, d->stripe_sectors_dirty + stripe);
+                nr_sectors -= s;
+                stripe_offset = 0;
+                stripe++;
+        }
+}
+
 /* Background writeback - IO loop */
 
 static void dirty_io_destructor(struct closure *cl)
@@ -216,9 +285,10 @@ static void write_dirty_finish(struct closure *cl)
         struct dirty_io *io = container_of(cl, struct dirty_io, cl);
         struct keybuf_key *w = io->bio.bi_private;
         struct cached_dev *dc = io->dc;
-        struct bio_vec *bv = bio_iovec_idx(&io->bio, io->bio.bi_vcnt);
+        struct bio_vec *bv;
+        int i;
 
-        while (bv-- != io->bio.bi_io_vec)
+        bio_for_each_segment_all(bv, &io->bio, i)
                 __free_page(bv->bv_page);
 
         /* This is kind of a dumb way of signalling errors. */
@@ -236,10 +306,12 @@ static void write_dirty_finish(struct closure *cl)
                 for (i = 0; i < KEY_PTRS(&w->key); i++)
                         atomic_inc(&PTR_BUCKET(dc->disk.c, &w->key, i)->pin);
 
-                pr_debug("clearing %s", pkey(&w->key));
                 bch_btree_insert(&op, dc->disk.c);
                 closure_sync(&op.cl);
 
+                if (op.insert_collision)
+                        trace_bcache_writeback_collision(&w->key);
+
                 atomic_long_inc(op.insert_collision
                                 ? &dc->disk.c->writeback_keys_failed
                                 : &dc->disk.c->writeback_keys_done);
@@ -275,7 +347,6 @@ static void write_dirty(struct closure *cl)
         io->bio.bi_bdev         = io->dc->bdev;
         io->bio.bi_end_io       = dirty_endio;
 
-        trace_bcache_write_dirty(&io->bio);
         closure_bio_submit(&io->bio, cl, &io->dc->disk);
 
         continue_at(cl, write_dirty_finish, dirty_wq);
@@ -296,7 +367,6 @@ static void read_dirty_submit(struct closure *cl)
 {
         struct dirty_io *io = container_of(cl, struct dirty_io, cl);
 
-        trace_bcache_read_dirty(&io->bio);
         closure_bio_submit(&io->bio, cl, &io->dc->disk);
 
         continue_at(cl, write_dirty, dirty_wq);
@@ -349,10 +419,10 @@ static void read_dirty(struct closure *cl)
                 io->bio.bi_rw           = READ;
                 io->bio.bi_end_io       = read_dirty_endio;
 
-                if (bch_bio_alloc_pages(&io->bio, GFP_KERNEL))
+                if (bio_alloc_pages(&io->bio, GFP_KERNEL))
                         goto err_free;
 
-                pr_debug("%s", pkey(&w->key));
+                trace_bcache_writeback(&w->key);
 
                 closure_call(&io->cl, read_dirty_submit, NULL, &dc->disk.cl);
 
@@ -375,12 +445,49 @@ err:
         refill_dirty(cl);
 }
 
+/* Init */
+
+static int bch_btree_sectors_dirty_init(struct btree *b, struct btree_op *op,
+                                        struct cached_dev *dc)
+{
+        struct bkey *k;
+        struct btree_iter iter;
+
+        bch_btree_iter_init(b, &iter, &KEY(dc->disk.id, 0, 0));
+        while ((k = bch_btree_iter_next_filter(&iter, b, bch_ptr_bad)))
+                if (!b->level) {
+                        if (KEY_INODE(k) > dc->disk.id)
+                                break;
+
+                        if (KEY_DIRTY(k))
+                                bcache_dev_sectors_dirty_add(b->c, dc->disk.id,
+                                                             KEY_START(k),
+                                                             KEY_SIZE(k));
+                } else {
+                        btree(sectors_dirty_init, k, b, op, dc);
+                        if (KEY_INODE(k) > dc->disk.id)
+                                break;
+
+                        cond_resched();
+                }
+
+        return 0;
+}
+
+void bch_sectors_dirty_init(struct cached_dev *dc)
+{
+        struct btree_op op;
+
+        bch_btree_op_init_stack(&op);
+        btree_root(sectors_dirty_init, dc->disk.c, &op, dc);
+}
+
 void bch_cached_dev_writeback_init(struct cached_dev *dc)
 {
         closure_init_unlocked(&dc->writeback);
         init_rwsem(&dc->writeback_lock);
 
-        bch_keybuf_init(&dc->writeback_keys, dirty_pred);
+        bch_keybuf_init(&dc->writeback_keys);
 
         dc->writeback_metadata          = true;
         dc->writeback_running           = true;