static void bch_data_insert_keys(struct closure *cl)
 {
-       struct search *s = container_of(cl, struct search, btree);
+       struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
        atomic_t *journal_ref = NULL;
-       struct bkey *replace_key = s->replace ? &s->replace_key : NULL;
+       struct bkey *replace_key = op->replace ? &op->replace_key : NULL;
        int ret;
 
        /*
                closure_sync(&s->cl);
 #endif
 
-       if (s->write)
-               journal_ref = bch_journal(s->c, &s->insert_keys,
-                                         s->flush_journal
-                                         ? &s->cl : NULL);
+       if (!op->replace)
+               journal_ref = bch_journal(op->c, &op->insert_keys,
+                                         op->flush_journal ? cl : NULL);
 
-       ret = bch_btree_insert(s->c, &s->insert_keys,
+       ret = bch_btree_insert(op->c, &op->insert_keys,
                               journal_ref, replace_key);
        if (ret == -ESRCH) {
-               s->insert_collision = true;
+               op->replace_collision = true;
        } else if (ret) {
-               s->error                = -ENOMEM;
-               s->insert_data_done     = true;
+               op->error               = -ENOMEM;
+               op->insert_data_done    = true;
        }
 
        if (journal_ref)
                atomic_dec_bug(journal_ref);
 
-       if (!s->insert_data_done)
+       if (!op->insert_data_done)
                continue_at(cl, bch_data_insert_start, bcache_wq);
 
-       bch_keylist_free(&s->insert_keys);
+       bch_keylist_free(&op->insert_keys);
        closure_return(cl);
 }
 
  *
  * If s->writeback is true, will not fail.
  */
-static bool bch_alloc_sectors(struct bkey *k, unsigned sectors,
-                             struct search *s)
+static bool bch_alloc_sectors(struct data_insert_op *op,
+                             struct bkey *k, unsigned sectors)
 {
-       struct cache_set *c = s->c;
+       struct cache_set *c = op->c;
        struct open_bucket *b;
        BKEY_PADDED(key) alloc;
        unsigned i;
        bkey_init(&alloc.key);
        spin_lock(&c->data_bucket_lock);
 
-       while (!(b = pick_data_bucket(c, k, s->task, &alloc.key))) {
-               unsigned watermark = s->write_prio
+       while (!(b = pick_data_bucket(c, k, op->task, &alloc.key))) {
+               unsigned watermark = op->write_prio
                        ? WATERMARK_MOVINGGC
                        : WATERMARK_NONE;
 
                spin_unlock(&c->data_bucket_lock);
 
                if (bch_bucket_alloc_set(c, watermark, &alloc.key,
-                                        1, s->writeback))
+                                        1, op->writeback))
                        return false;
 
                spin_lock(&c->data_bucket_lock);
         */
        list_move_tail(&b->list, &c->data_buckets);
        bkey_copy_key(&b->key, k);
-       b->last = s->task;
+       b->last = op->task;
 
        b->sectors_free -= sectors;
 
 
 static void bch_data_invalidate(struct closure *cl)
 {
-       struct search *s = container_of(cl, struct search, btree);
-       struct bio *bio = s->cache_bio;
+       struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+       struct bio *bio = op->bio;
 
        pr_debug("invalidating %i sectors from %llu",
                 bio_sectors(bio), (uint64_t) bio->bi_sector);
        while (bio_sectors(bio)) {
                unsigned len = min(bio_sectors(bio), 1U << 14);
 
-               if (bch_keylist_realloc(&s->insert_keys, 0, s->c))
+               if (bch_keylist_realloc(&op->insert_keys, 0, op->c))
                        goto out;
 
                bio->bi_sector  += len;
                bio->bi_size    -= len << 9;
 
-               bch_keylist_add(&s->insert_keys,
-                               &KEY(s->inode, bio->bi_sector, len));
+               bch_keylist_add(&op->insert_keys,
+                               &KEY(op->inode, bio->bi_sector, len));
        }
 
-       s->insert_data_done = true;
+       op->insert_data_done = true;
        bio_put(bio);
 out:
        continue_at(cl, bch_data_insert_keys, bcache_wq);
 
 static void bch_data_insert_error(struct closure *cl)
 {
-       struct search *s = container_of(cl, struct search, btree);
+       struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
 
        /*
         * Our data write just errored, which means we've got a bunch of keys to
         * from the keys we'll accomplish just that.
         */
 
-       struct bkey *src = s->insert_keys.keys, *dst = s->insert_keys.keys;
+       struct bkey *src = op->insert_keys.keys, *dst = op->insert_keys.keys;
 
-       while (src != s->insert_keys.top) {
+       while (src != op->insert_keys.top) {
                struct bkey *n = bkey_next(src);
 
                SET_KEY_PTRS(src, 0);
                src = n;
        }
 
-       s->insert_keys.top = dst;
+       op->insert_keys.top = dst;
 
        bch_data_insert_keys(cl);
 }
 static void bch_data_insert_endio(struct bio *bio, int error)
 {
        struct closure *cl = bio->bi_private;
-       struct search *s = container_of(cl, struct search, btree);
+       struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
 
        if (error) {
                /* TODO: We could try to recover from this. */
-               if (s->writeback)
-                       s->error = error;
-               else if (s->write)
+               if (op->writeback)
+                       op->error = error;
+               else if (!op->replace)
                        set_closure_fn(cl, bch_data_insert_error, bcache_wq);
                else
                        set_closure_fn(cl, NULL, NULL);
        }
 
-       bch_bbio_endio(s->c, bio, error, "writing data to cache");
+       bch_bbio_endio(op->c, bio, error, "writing data to cache");
 }
 
 static void bch_data_insert_start(struct closure *cl)
 {
-       struct search *s = container_of(cl, struct search, btree);
-       struct bio *bio = s->cache_bio, *n;
+       struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+       struct bio *bio = op->bio, *n;
 
-       if (s->bypass)
+       if (op->bypass)
                return bch_data_invalidate(cl);
 
-       if (atomic_sub_return(bio_sectors(bio), &s->c->sectors_to_gc) < 0) {
-               set_gc_sectors(s->c);
-               wake_up_gc(s->c);
+       if (atomic_sub_return(bio_sectors(bio), &op->c->sectors_to_gc) < 0) {
+               set_gc_sectors(op->c);
+               wake_up_gc(op->c);
        }
 
        /*
        do {
                unsigned i;
                struct bkey *k;
-               struct bio_set *split = s->d
-                       ? s->d->bio_split : s->c->bio_split;
+               struct bio_set *split = op->c->bio_split;
 
                /* 1 for the device pointer and 1 for the chksum */
-               if (bch_keylist_realloc(&s->insert_keys,
-                                       1 + (s->csum ? 1 : 0),
-                                       s->c))
+               if (bch_keylist_realloc(&op->insert_keys,
+                                       1 + (op->csum ? 1 : 0),
+                                       op->c))
                        continue_at(cl, bch_data_insert_keys, bcache_wq);
 
-               k = s->insert_keys.top;
+               k = op->insert_keys.top;
                bkey_init(k);
-               SET_KEY_INODE(k, s->inode);
+               SET_KEY_INODE(k, op->inode);
                SET_KEY_OFFSET(k, bio->bi_sector);
 
-               if (!bch_alloc_sectors(k, bio_sectors(bio), s))
+               if (!bch_alloc_sectors(op, k, bio_sectors(bio)))
                        goto err;
 
                n = bch_bio_split(bio, KEY_SIZE(k), GFP_NOIO, split);
                n->bi_end_io    = bch_data_insert_endio;
                n->bi_private   = cl;
 
-               if (s->writeback) {
+               if (op->writeback) {
                        SET_KEY_DIRTY(k, true);
 
                        for (i = 0; i < KEY_PTRS(k); i++)
-                               SET_GC_MARK(PTR_BUCKET(s->c, k, i),
+                               SET_GC_MARK(PTR_BUCKET(op->c, k, i),
                                            GC_MARK_DIRTY);
                }
 
-               SET_KEY_CSUM(k, s->csum);
+               SET_KEY_CSUM(k, op->csum);
                if (KEY_CSUM(k))
                        bio_csum(n, k);
 
                trace_bcache_cache_insert(k);
-               bch_keylist_push(&s->insert_keys);
+               bch_keylist_push(&op->insert_keys);
 
                n->bi_rw |= REQ_WRITE;
-               bch_submit_bbio(n, s->c, k, 0);
+               bch_submit_bbio(n, op->c, k, 0);
        } while (n != bio);
 
-       s->insert_data_done = true;
+       op->insert_data_done = true;
        continue_at(cl, bch_data_insert_keys, bcache_wq);
 err:
        /* bch_alloc_sectors() blocks if s->writeback = true */
-       BUG_ON(s->writeback);
+       BUG_ON(op->writeback);
 
        /*
         * But if it's not a writeback write we'd rather just bail out if
         * we might be starving btree writes for gc or something.
         */
 
-       if (s->write) {
+       if (!op->replace) {
                /*
                 * Writethrough write: We can't complete the write until we've
                 * updated the index. But we don't want to delay the write while
                 * we wait for buckets to be freed up, so just invalidate the
                 * rest of the write.
                 */
-               s->bypass = true;
+               op->bypass = true;
                return bch_data_invalidate(cl);
        } else {
                /*
                 * From a cache miss, we can just insert the keys for the data
                 * we have written or bail out if we didn't do anything.
                 */
-               s->insert_data_done = true;
+               op->insert_data_done = true;
                bio_put(bio);
 
-               if (!bch_keylist_empty(&s->insert_keys))
+               if (!bch_keylist_empty(&op->insert_keys))
                        continue_at(cl, bch_data_insert_keys, bcache_wq);
                else
                        closure_return(cl);
  */
 void bch_data_insert(struct closure *cl)
 {
-       struct search *s = container_of(cl, struct search, btree);
+       struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
 
-       bch_keylist_init(&s->insert_keys);
-       bio_get(s->cache_bio);
+       trace_bcache_write(op->bio, op->writeback, op->bypass);
+
+       bch_keylist_init(&op->insert_keys);
+       bio_get(op->bio);
        bch_data_insert_start(cl);
 }
 
+/* Congested? */
+
+unsigned bch_get_congested(struct cache_set *c)
+{
+       int i;
+       long rand;
+
+       if (!c->congested_read_threshold_us &&
+           !c->congested_write_threshold_us)
+               return 0;
+
+       i = (local_clock_us() - c->congested_last_us) / 1024;
+       if (i < 0)
+               return 0;
+
+       i += atomic_read(&c->congested);
+       if (i >= 0)
+               return 0;
+
+       i += CONGESTED_MAX;
+
+       if (i > 0)
+               i = fract_exp_two(i, 6);
+
+       rand = get_random_int();
+       i -= bitmap_weight(&rand, BITS_PER_LONG);
+
+       return i > 0 ? i : 1;
+}
+
+static void add_sequential(struct task_struct *t)
+{
+       ewma_add(t->sequential_io_avg,
+                t->sequential_io, 8, 0);
+
+       t->sequential_io = 0;
+}
+
+static struct hlist_head *iohash(struct cached_dev *dc, uint64_t k)
+{
+       return &dc->io_hash[hash_64(k, RECENT_IO_BITS)];
+}
+
+static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
+{
+       struct cache_set *c = dc->disk.c;
+       unsigned mode = cache_mode(dc, bio);
+       unsigned sectors, congested = bch_get_congested(c);
+       struct task_struct *task = current;
+
+       if (atomic_read(&dc->disk.detaching) ||
+           c->gc_stats.in_use > CUTOFF_CACHE_ADD ||
+           (bio->bi_rw & REQ_DISCARD))
+               goto skip;
+
+       if (mode == CACHE_MODE_NONE ||
+           (mode == CACHE_MODE_WRITEAROUND &&
+            (bio->bi_rw & REQ_WRITE)))
+               goto skip;
+
+       if (bio->bi_sector & (c->sb.block_size - 1) ||
+           bio_sectors(bio) & (c->sb.block_size - 1)) {
+               pr_debug("skipping unaligned io");
+               goto skip;
+       }
+
+       if (!congested && !dc->sequential_cutoff)
+               goto rescale;
+
+       if (!congested &&
+           mode == CACHE_MODE_WRITEBACK &&
+           (bio->bi_rw & REQ_WRITE) &&
+           (bio->bi_rw & REQ_SYNC))
+               goto rescale;
+
+       if (dc->sequential_merge) {
+               struct io *i;
+
+               spin_lock(&dc->io_lock);
+
+               hlist_for_each_entry(i, iohash(dc, bio->bi_sector), hash)
+                       if (i->last == bio->bi_sector &&
+                           time_before(jiffies, i->jiffies))
+                               goto found;
+
+               i = list_first_entry(&dc->io_lru, struct io, lru);
+
+               add_sequential(task);
+               i->sequential = 0;
+found:
+               if (i->sequential + bio->bi_size > i->sequential)
+                       i->sequential   += bio->bi_size;
+
+               i->last                  = bio_end_sector(bio);
+               i->jiffies               = jiffies + msecs_to_jiffies(5000);
+               task->sequential_io      = i->sequential;
+
+               hlist_del(&i->hash);
+               hlist_add_head(&i->hash, iohash(dc, i->last));
+               list_move_tail(&i->lru, &dc->io_lru);
+
+               spin_unlock(&dc->io_lock);
+       } else {
+               task->sequential_io = bio->bi_size;
+
+               add_sequential(task);
+       }
+
+       sectors = max(task->sequential_io,
+                     task->sequential_io_avg) >> 9;
+
+       if (dc->sequential_cutoff &&
+           sectors >= dc->sequential_cutoff >> 9) {
+               trace_bcache_bypass_sequential(bio);
+               goto skip;
+       }
+
+       if (congested && sectors >= congested) {
+               trace_bcache_bypass_congested(bio);
+               goto skip;
+       }
+
+rescale:
+       bch_rescale_priorities(c, bio_sectors(bio));
+       return false;
+skip:
+       bch_mark_sectors_bypassed(c, dc, bio_sectors(bio));
+       return true;
+}
+
 /* Cache lookup */
 
+struct search {
+       /* Stack frame for bio_complete */
+       struct closure          cl;
+
+       struct bcache_device    *d;
+
+       struct bbio             bio;
+       struct bio              *orig_bio;
+       struct bio              *cache_miss;
+
+       unsigned                insert_bio_sectors;
+
+       unsigned                recoverable:1;
+       unsigned                unaligned_bvec:1;
+       unsigned                write:1;
+
+       unsigned long           start_time;
+
+       struct btree_op         op;
+       struct data_insert_op   iop;
+};
+
 static void bch_cache_read_endio(struct bio *bio, int error)
 {
        struct bbio *b = container_of(bio, struct bbio, bio);
         */
 
        if (error)
-               s->error = error;
-       else if (ptr_stale(s->c, &b->key, 0)) {
-               atomic_long_inc(&s->c->cache_read_races);
-               s->error = -EINTR;
+               s->iop.error = error;
+       else if (ptr_stale(s->iop.c, &b->key, 0)) {
+               atomic_long_inc(&s->iop.c->cache_read_races);
+               s->iop.error = -EINTR;
        }
 
-       bch_bbio_endio(s->c, bio, error, "reading from cache");
+       bch_bbio_endio(s->iop.c, bio, error, "reading from cache");
 }
 
 /*
        struct bkey *bio_key;
        unsigned ptr;
 
-       if (bkey_cmp(k, &KEY(s->inode, bio->bi_sector, 0)) <= 0)
+       if (bkey_cmp(k, &KEY(s->iop.inode, bio->bi_sector, 0)) <= 0)
                return MAP_CONTINUE;
 
-       if (KEY_INODE(k) != s->inode ||
+       if (KEY_INODE(k) != s->iop.inode ||
            KEY_START(k) > bio->bi_sector) {
                unsigned bio_sectors = bio_sectors(bio);
-               unsigned sectors = KEY_INODE(k) == s->inode
+               unsigned sectors = KEY_INODE(k) == s->iop.inode
                        ? min_t(uint64_t, INT_MAX,
                                KEY_START(k) - bio->bi_sector)
                        : INT_MAX;
        bio_key = &container_of(n, struct bbio, bio)->key;
        bch_bkey_copy_single_ptr(bio_key, k, ptr);
 
-       bch_cut_front(&KEY(s->inode, n->bi_sector, 0), bio_key);
-       bch_cut_back(&KEY(s->inode, bio_end_sector(n), 0), bio_key);
+       bch_cut_front(&KEY(s->iop.inode, n->bi_sector, 0), bio_key);
+       bch_cut_back(&KEY(s->iop.inode, bio_end_sector(n), 0), bio_key);
 
        n->bi_end_io    = bch_cache_read_endio;
        n->bi_private   = &s->cl;
 
 static void cache_lookup(struct closure *cl)
 {
-       struct search *s = container_of(cl, struct search, btree);
+       struct search *s = container_of(cl, struct search, iop.cl);
        struct bio *bio = &s->bio.bio;
 
-       int ret = bch_btree_map_keys(&s->op, s->c,
-                                    &KEY(s->inode, bio->bi_sector, 0),
+       int ret = bch_btree_map_keys(&s->op, s->iop.c,
+                                    &KEY(s->iop.inode, bio->bi_sector, 0),
                                     cache_lookup_fn, MAP_END_KEY);
        if (ret == -EAGAIN)
                continue_at(cl, cache_lookup, bcache_wq);
 
        if (error) {
                struct search *s = container_of(cl, struct search, cl);
-               s->error = error;
+               s->iop.error = error;
                /* Only cache read errors are recoverable */
                s->recoverable = false;
        }
                part_stat_add(cpu, &s->d->disk->part0, ticks[rw], duration);
                part_stat_unlock();
 
-               trace_bcache_request_end(s, s->orig_bio);
-               bio_endio(s->orig_bio, s->error);
+               trace_bcache_request_end(s->d, s->orig_bio);
+               bio_endio(s->orig_bio, s->iop.error);
                s->orig_bio = NULL;
        }
 }
        struct search *s = container_of(cl, struct search, cl);
        bio_complete(s);
 
-       if (s->cache_bio)
-               bio_put(s->cache_bio);
+       if (s->iop.bio)
+               bio_put(s->iop.bio);
 
        if (s->unaligned_bvec)
                mempool_free(s->bio.bio.bi_io_vec, s->d->unaligned_bvec);
        struct bio_vec *bv;
 
        s = mempool_alloc(d->c->search, GFP_NOIO);
-       memset(s, 0, offsetof(struct search, insert_keys));
+       memset(s, 0, offsetof(struct search, iop.insert_keys));
 
        __closure_init(&s->cl, NULL);
 
-       s->inode                = d->id;
-       s->c                    = d->c;
+       s->iop.inode            = d->id;
+       s->iop.c                = d->c;
        s->d                    = d;
        s->op.lock              = -1;
-       s->task                 = current;
+       s->iop.task             = current;
        s->orig_bio             = bio;
        s->write                = (bio->bi_rw & REQ_WRITE) != 0;
-       s->flush_journal        = (bio->bi_rw & (REQ_FLUSH|REQ_FUA)) != 0;
+       s->iop.flush_journal    = (bio->bi_rw & (REQ_FLUSH|REQ_FUA)) != 0;
        s->recoverable          = 1;
        s->start_time           = jiffies;
        do_bio_hook(s);
        cached_dev_put(dc);
 }
 
-unsigned bch_get_congested(struct cache_set *c)
-{
-       int i;
-       long rand;
-
-       if (!c->congested_read_threshold_us &&
-           !c->congested_write_threshold_us)
-               return 0;
-
-       i = (local_clock_us() - c->congested_last_us) / 1024;
-       if (i < 0)
-               return 0;
-
-       i += atomic_read(&c->congested);
-       if (i >= 0)
-               return 0;
-
-       i += CONGESTED_MAX;
-
-       if (i > 0)
-               i = fract_exp_two(i, 6);
-
-       rand = get_random_int();
-       i -= bitmap_weight(&rand, BITS_PER_LONG);
-
-       return i > 0 ? i : 1;
-}
-
-static void add_sequential(struct task_struct *t)
-{
-       ewma_add(t->sequential_io_avg,
-                t->sequential_io, 8, 0);
-
-       t->sequential_io = 0;
-}
-
-static struct hlist_head *iohash(struct cached_dev *dc, uint64_t k)
-{
-       return &dc->io_hash[hash_64(k, RECENT_IO_BITS)];
-}
-
-static bool check_should_bypass(struct cached_dev *dc, struct search *s)
-{
-       struct cache_set *c = s->c;
-       struct bio *bio = &s->bio.bio;
-       unsigned mode = cache_mode(dc, bio);
-       unsigned sectors, congested = bch_get_congested(c);
-
-       if (atomic_read(&dc->disk.detaching) ||
-           c->gc_stats.in_use > CUTOFF_CACHE_ADD ||
-           (bio->bi_rw & REQ_DISCARD))
-               goto skip;
-
-       if (mode == CACHE_MODE_NONE ||
-           (mode == CACHE_MODE_WRITEAROUND &&
-            (bio->bi_rw & REQ_WRITE)))
-               goto skip;
-
-       if (bio->bi_sector & (c->sb.block_size - 1) ||
-           bio_sectors(bio) & (c->sb.block_size - 1)) {
-               pr_debug("skipping unaligned io");
-               goto skip;
-       }
-
-       if (!congested && !dc->sequential_cutoff)
-               goto rescale;
-
-       if (!congested &&
-           mode == CACHE_MODE_WRITEBACK &&
-           (bio->bi_rw & REQ_WRITE) &&
-           (bio->bi_rw & REQ_SYNC))
-               goto rescale;
-
-       if (dc->sequential_merge) {
-               struct io *i;
-
-               spin_lock(&dc->io_lock);
-
-               hlist_for_each_entry(i, iohash(dc, bio->bi_sector), hash)
-                       if (i->last == bio->bi_sector &&
-                           time_before(jiffies, i->jiffies))
-                               goto found;
-
-               i = list_first_entry(&dc->io_lru, struct io, lru);
-
-               add_sequential(s->task);
-               i->sequential = 0;
-found:
-               if (i->sequential + bio->bi_size > i->sequential)
-                       i->sequential   += bio->bi_size;
-
-               i->last                  = bio_end_sector(bio);
-               i->jiffies               = jiffies + msecs_to_jiffies(5000);
-               s->task->sequential_io   = i->sequential;
-
-               hlist_del(&i->hash);
-               hlist_add_head(&i->hash, iohash(dc, i->last));
-               list_move_tail(&i->lru, &dc->io_lru);
-
-               spin_unlock(&dc->io_lock);
-       } else {
-               s->task->sequential_io = bio->bi_size;
-
-               add_sequential(s->task);
-       }
-
-       sectors = max(s->task->sequential_io,
-                     s->task->sequential_io_avg) >> 9;
-
-       if (dc->sequential_cutoff &&
-           sectors >= dc->sequential_cutoff >> 9) {
-               trace_bcache_bypass_sequential(s->orig_bio);
-               goto skip;
-       }
-
-       if (congested && sectors >= congested) {
-               trace_bcache_bypass_congested(s->orig_bio);
-               goto skip;
-       }
-
-rescale:
-       bch_rescale_priorities(c, bio_sectors(bio));
-       return false;
-skip:
-       bch_mark_sectors_bypassed(s, bio_sectors(bio));
-       return true;
-}
-
 /* Process reads */
 
 static void cached_dev_cache_miss_done(struct closure *cl)
 {
        struct search *s = container_of(cl, struct search, cl);
 
-       if (s->insert_collision)
-               bch_mark_cache_miss_collision(s);
+       if (s->iop.replace_collision)
+               bch_mark_cache_miss_collision(s->iop.c, s->d);
 
-       if (s->cache_bio) {
+       if (s->iop.bio) {
                int i;
                struct bio_vec *bv;
 
-               bio_for_each_segment_all(bv, s->cache_bio, i)
+               bio_for_each_segment_all(bv, s->iop.bio, i)
                        __free_page(bv->bv_page);
        }
 
                /* Retry from the backing device: */
                trace_bcache_read_retry(s->orig_bio);
 
-               s->error = 0;
+               s->iop.error = 0;
                bv = s->bio.bio.bi_io_vec;
                do_bio_hook(s);
                s->bio.bio.bi_io_vec = bv;
         * to the buffers the original bio pointed to:
         */
 
-       if (s->cache_bio) {
-               bio_reset(s->cache_bio);
-               s->cache_bio->bi_sector =
-                       s->cache_miss->bi_sector;
-               s->cache_bio->bi_bdev = s->cache_miss->bi_bdev;
-               s->cache_bio->bi_size = s->cache_bio_sectors << 9;
-               bch_bio_map(s->cache_bio, NULL);
+       if (s->iop.bio) {
+               bio_reset(s->iop.bio);
+               s->iop.bio->bi_sector = s->cache_miss->bi_sector;
+               s->iop.bio->bi_bdev = s->cache_miss->bi_bdev;
+               s->iop.bio->bi_size = s->insert_bio_sectors << 9;
+               bch_bio_map(s->iop.bio, NULL);
 
-               bio_copy_data(s->cache_miss, s->cache_bio);
+               bio_copy_data(s->cache_miss, s->iop.bio);
 
                bio_put(s->cache_miss);
                s->cache_miss = NULL;
        }
 
-       if (verify(dc, &s->bio.bio) && s->recoverable)
-               bch_data_verify(s);
+       if (verify(dc, &s->bio.bio) && s->recoverable && !s->unaligned_bvec)
+               bch_data_verify(dc, s->orig_bio);
 
        bio_complete(s);
 
-       if (s->cache_bio &&
-           !test_bit(CACHE_SET_STOPPING, &s->c->flags)) {
-               BUG_ON(!s->replace);
-               closure_call(&s->btree, bch_data_insert, NULL, cl);
+       if (s->iop.bio &&
+           !test_bit(CACHE_SET_STOPPING, &s->iop.c->flags)) {
+               BUG_ON(!s->iop.replace);
+               closure_call(&s->iop.cl, bch_data_insert, NULL, cl);
        }
 
        continue_at(cl, cached_dev_cache_miss_done, NULL);
        struct search *s = container_of(cl, struct search, cl);
        struct cached_dev *dc = container_of(s->d, struct cached_dev, disk);
 
-       bch_mark_cache_accounting(s, !s->cache_miss, s->bypass);
-       trace_bcache_read(s->orig_bio, !s->cache_miss, s->bypass);
+       bch_mark_cache_accounting(s->iop.c, s->d,
+                                 !s->cache_miss, s->iop.bypass);
+       trace_bcache_read(s->orig_bio, !s->cache_miss, s->iop.bypass);
 
-       if (s->error)
+       if (s->iop.error)
                continue_at_nobarrier(cl, cached_dev_read_error, bcache_wq);
-       else if (s->cache_bio || verify(dc, &s->bio.bio))
+       else if (s->iop.bio || verify(dc, &s->bio.bio))
                continue_at_nobarrier(cl, cached_dev_read_done, bcache_wq);
        else
                continue_at_nobarrier(cl, cached_dev_bio_complete, NULL);
        struct cached_dev *dc = container_of(s->d, struct cached_dev, disk);
        struct bio *miss, *cache_bio;
 
-       if (s->cache_miss || s->bypass) {
+       if (s->cache_miss || s->iop.bypass) {
                miss = bch_bio_split(bio, sectors, GFP_NOIO, s->d->bio_split);
                ret = miss == bio ? MAP_DONE : MAP_CONTINUE;
                goto out_submit;
 
        if (!(bio->bi_rw & REQ_RAHEAD) &&
            !(bio->bi_rw & REQ_META) &&
-           s->c->gc_stats.in_use < CUTOFF_CACHE_READA)
+           s->iop.c->gc_stats.in_use < CUTOFF_CACHE_READA)
                reada = min_t(sector_t, dc->readahead >> 9,
                              bdev_sectors(bio->bi_bdev) - bio_end_sector(bio));
 
-       s->cache_bio_sectors = min(sectors, bio_sectors(bio) + reada);
+       s->insert_bio_sectors = min(sectors, bio_sectors(bio) + reada);
 
-       s->replace_key = KEY(s->inode, bio->bi_sector +
-                            s->cache_bio_sectors, s->cache_bio_sectors);
+       s->iop.replace_key = KEY(s->iop.inode,
+                                bio->bi_sector + s->insert_bio_sectors,
+                                s->insert_bio_sectors);
 
-       ret = bch_btree_insert_check_key(b, &s->op, &s->replace_key);
+       ret = bch_btree_insert_check_key(b, &s->op, &s->iop.replace_key);
        if (ret)
                return ret;
 
-       s->replace = true;
+       s->iop.replace = true;
 
        miss = bch_bio_split(bio, sectors, GFP_NOIO, s->d->bio_split);
 
        ret = miss == bio ? MAP_DONE : -EINTR;
 
        cache_bio = bio_alloc_bioset(GFP_NOWAIT,
-                       DIV_ROUND_UP(s->cache_bio_sectors, PAGE_SECTORS),
+                       DIV_ROUND_UP(s->insert_bio_sectors, PAGE_SECTORS),
                        dc->disk.bio_split);
        if (!cache_bio)
                goto out_submit;
 
        cache_bio->bi_sector    = miss->bi_sector;
        cache_bio->bi_bdev      = miss->bi_bdev;
-       cache_bio->bi_size      = s->cache_bio_sectors << 9;
+       cache_bio->bi_size      = s->insert_bio_sectors << 9;
 
        cache_bio->bi_end_io    = request_endio;
        cache_bio->bi_private   = &s->cl;
        if (bio_alloc_pages(cache_bio, __GFP_NOWARN|GFP_NOIO))
                goto out_put;
 
+       if (reada)
+               bch_mark_cache_readahead(s->iop.c, s->d);
+
        s->cache_miss   = miss;
-       s->cache_bio = cache_bio;
+       s->iop.bio      = cache_bio;
        bio_get(cache_bio);
        closure_bio_submit(cache_bio, &s->cl, s->d);
 
 {
        struct closure *cl = &s->cl;
 
-       closure_call(&s->btree, cache_lookup, NULL, cl);
+       closure_call(&s->iop.cl, cache_lookup, NULL, cl);
        continue_at(cl, cached_dev_read_done_bh, NULL);
 }
 
        struct bkey start = KEY(dc->disk.id, bio->bi_sector, 0);
        struct bkey end = KEY(dc->disk.id, bio_end_sector(bio), 0);
 
-       bch_keybuf_check_overlapping(&s->c->moving_gc_keys, &start, &end);
+       bch_keybuf_check_overlapping(&s->iop.c->moving_gc_keys, &start, &end);
 
        down_read_non_owner(&dc->writeback_lock);
        if (bch_keybuf_check_overlapping(&dc->writeback_keys, &start, &end)) {
                 * We overlap with some dirty data undergoing background
                 * writeback, force this write to writeback
                 */
-               s->bypass       = false;
-               s->writeback    = true;
+               s->iop.bypass = false;
+               s->iop.writeback = true;
        }
 
        /*
         * so we still want to call it.
         */
        if (bio->bi_rw & REQ_DISCARD)
-               s->bypass = true;
+               s->iop.bypass = true;
 
        if (should_writeback(dc, s->orig_bio,
                             cache_mode(dc, bio),
-                            s->bypass)) {
-               s->bypass = false;
-               s->writeback = true;
+                            s->iop.bypass)) {
+               s->iop.bypass = false;
+               s->iop.writeback = true;
        }
 
-       trace_bcache_write(s->orig_bio, s->writeback, s->bypass);
-
-       if (s->bypass) {
-               s->cache_bio = s->orig_bio;
-               bio_get(s->cache_bio);
+       if (s->iop.bypass) {
+               s->iop.bio = s->orig_bio;
+               bio_get(s->iop.bio);
 
                if (!(bio->bi_rw & REQ_DISCARD) ||
                    blk_queue_discard(bdev_get_queue(dc->bdev)))
                        closure_bio_submit(bio, cl, s->d);
-       } else if (s->writeback) {
+       } else if (s->iop.writeback) {
                bch_writeback_add(dc);
-               s->cache_bio = bio;
+               s->iop.bio = bio;
 
                if (bio->bi_rw & REQ_FLUSH) {
                        /* Also need to send a flush to the backing device */
                        closure_bio_submit(flush, cl, s->d);
                }
        } else {
-               s->cache_bio = bio_clone_bioset(bio, GFP_NOIO,
-                                               dc->disk.bio_split);
+               s->iop.bio = bio_clone_bioset(bio, GFP_NOIO,
+                                             dc->disk.bio_split);
 
                closure_bio_submit(bio, cl, s->d);
        }
 
-       closure_call(&s->btree, bch_data_insert, NULL, cl);
+       closure_call(&s->iop.cl, bch_data_insert, NULL, cl);
        continue_at(cl, cached_dev_write_complete, NULL);
 }
 
        struct search *s = container_of(cl, struct search, cl);
        struct bio *bio = &s->bio.bio;
 
-       if (s->flush_journal)
-               bch_journal_meta(s->c, cl);
+       if (s->iop.flush_journal)
+               bch_journal_meta(s->iop.c, cl);
 
        /* If it's a flush, we send the flush to the backing device too */
        closure_bio_submit(bio, cl, s->d);
 
        if (cached_dev_get(dc)) {
                s = search_alloc(bio, d);
-               trace_bcache_request_start(s, bio);
+               trace_bcache_request_start(s->d, bio);
 
                if (!bio->bi_size) {
                        /*
                                              cached_dev_nodata,
                                              bcache_wq);
                } else {
-                       s->bypass = check_should_bypass(dc, s);
+                       s->iop.bypass = check_should_bypass(dc, bio);
 
                        if (rw)
                                cached_dev_write(dc, s);
 {
        struct search *s = container_of(cl, struct search, cl);
 
-       if (s->flush_journal)
-               bch_journal_meta(s->c, cl);
+       if (s->iop.flush_journal)
+               bch_journal_meta(s->iop.c, cl);
 
        continue_at(cl, search_free, NULL);
 }
        cl = &s->cl;
        bio = &s->bio.bio;
 
-       trace_bcache_request_start(s, bio);
+       trace_bcache_request_start(s->d, bio);
 
        if (!bio->bi_size) {
                /*
                                      flash_dev_nodata,
                                      bcache_wq);
        } else if (rw) {
-               bch_keybuf_check_overlapping(&s->c->moving_gc_keys,
+               bch_keybuf_check_overlapping(&s->iop.c->moving_gc_keys,
                                        &KEY(d->id, bio->bi_sector, 0),
                                        &KEY(d->id, bio_end_sector(bio), 0));
 
-               s->bypass       = (bio->bi_rw & REQ_DISCARD) != 0;
-               s->writeback    = true;
-               s->cache_bio    = bio;
+               s->iop.bypass           = (bio->bi_rw & REQ_DISCARD) != 0;
+               s->iop.writeback        = true;
+               s->iop.bio              = bio;
 
-               closure_call(&s->btree, bch_data_insert, NULL, cl);
+               closure_call(&s->iop.cl, bch_data_insert, NULL, cl);
        } else {
-               closure_call(&s->btree, cache_lookup, NULL, cl);
+               closure_call(&s->iop.cl, cache_lookup, NULL, cl);
        }
 
        continue_at(cl, search_free, NULL);