summaryrefslogtreecommitdiff
path: root/drivers/md/persistent-data/dm-btree.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2016-10-10 03:16:18 +0300
committerLinus Torvalds <torvalds@linux-foundation.org>2016-10-10 03:16:18 +0300
commit48915c2cbc77eceec2005afb695ac658fede4e0d (patch)
tree534b680b5203c37741e261b630f9c3584e743c47 /drivers/md/persistent-data/dm-btree.c
parentb9044ac8292fc94bee33f6f08acaed3ac55f0c75 (diff)
parent8ff232c1a819c2e98d85974a3bff0b7b8e2970ed (diff)
downloadlinux-48915c2cbc77eceec2005afb695ac658fede4e0d.tar.xz
Merge tag 'dm-4.9-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
Pull device mapper updates from Mike Snitzer: - various fixes and cleanups for request-based DM core - add support for delaying the requeue of requests; used by DM multipath when all paths have failed and 'queue_if_no_path' is enabled - DM cache improvements to speedup the loading metadata and the writing of the hint array - fix potential for a dm-crypt crash on device teardown - remove dm_bufio_cond_resched() and just using cond_resched() - change DM multipath to return a reservation conflict error immediately; rather than failing the path and retrying (potentially indefinitely) * tag 'dm-4.9-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm: (24 commits) dm mpath: always return reservation conflict without failing over dm bufio: remove dm_bufio_cond_resched() dm crypt: fix crash on exit dm cache metadata: switch to using the new cursor api for loading metadata dm array: introduce cursor api dm btree: introduce cursor api dm cache policy smq: distribute entries to random levels when switching to smq dm cache: speed up writing of the hint array dm array: add dm_array_new() dm mpath: delay the requeue of blk-mq requests while all paths down dm mpath: use dm_mq_kick_requeue_list() dm rq: introduce dm_mq_kick_requeue_list() dm rq: reduce arguments passed to map_request() and dm_requeue_original_request() dm rq: add DM_MAPIO_DELAY_REQUEUE to delay requeue of blk-mq requests dm: convert wait loops to use autoremove_wake_function() dm: use signal_pending_state() in dm_wait_for_completion() dm: rename task state function arguments dm: add two lockdep_assert_held() statements dm rq: simplify dm_old_stop_queue() dm mpath: check if path's request_queue is dying in activate_path() ...
Diffstat (limited to 'drivers/md/persistent-data/dm-btree.c')
-rw-r--r--drivers/md/persistent-data/dm-btree.c162
1 files changed, 162 insertions, 0 deletions
diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c
index 2cc1877804c2..20a40329d84a 100644
--- a/drivers/md/persistent-data/dm-btree.c
+++ b/drivers/md/persistent-data/dm-btree.c
@@ -994,3 +994,165 @@ int dm_btree_walk(struct dm_btree_info *info, dm_block_t root,
return walk_node(info, root, fn, context);
}
EXPORT_SYMBOL_GPL(dm_btree_walk);
+
+/*----------------------------------------------------------------*/
+
+static void prefetch_values(struct dm_btree_cursor *c)
+{
+ unsigned i, nr;
+ __le64 value_le;
+ struct cursor_node *n = c->nodes + c->depth - 1;
+ struct btree_node *bn = dm_block_data(n->b);
+ struct dm_block_manager *bm = dm_tm_get_bm(c->info->tm);
+
+ BUG_ON(c->info->value_type.size != sizeof(value_le));
+
+ nr = le32_to_cpu(bn->header.nr_entries);
+ for (i = 0; i < nr; i++) {
+ memcpy(&value_le, value_ptr(bn, i), sizeof(value_le));
+ dm_bm_prefetch(bm, le64_to_cpu(value_le));
+ }
+}
+
+static bool leaf_node(struct dm_btree_cursor *c)
+{
+ struct cursor_node *n = c->nodes + c->depth - 1;
+ struct btree_node *bn = dm_block_data(n->b);
+
+ return le32_to_cpu(bn->header.flags) & LEAF_NODE;
+}
+
+static int push_node(struct dm_btree_cursor *c, dm_block_t b)
+{
+ int r;
+ struct cursor_node *n = c->nodes + c->depth;
+
+ if (c->depth >= DM_BTREE_CURSOR_MAX_DEPTH - 1) {
+ DMERR("couldn't push cursor node, stack depth too high");
+ return -EINVAL;
+ }
+
+ r = bn_read_lock(c->info, b, &n->b);
+ if (r)
+ return r;
+
+ n->index = 0;
+ c->depth++;
+
+ if (c->prefetch_leaves || !leaf_node(c))
+ prefetch_values(c);
+
+ return 0;
+}
+
+static void pop_node(struct dm_btree_cursor *c)
+{
+ c->depth--;
+ unlock_block(c->info, c->nodes[c->depth].b);
+}
+
+static int inc_or_backtrack(struct dm_btree_cursor *c)
+{
+ struct cursor_node *n;
+ struct btree_node *bn;
+
+ for (;;) {
+ if (!c->depth)
+ return -ENODATA;
+
+ n = c->nodes + c->depth - 1;
+ bn = dm_block_data(n->b);
+
+ n->index++;
+ if (n->index < le32_to_cpu(bn->header.nr_entries))
+ break;
+
+ pop_node(c);
+ }
+
+ return 0;
+}
+
+static int find_leaf(struct dm_btree_cursor *c)
+{
+ int r = 0;
+ struct cursor_node *n;
+ struct btree_node *bn;
+ __le64 value_le;
+
+ for (;;) {
+ n = c->nodes + c->depth - 1;
+ bn = dm_block_data(n->b);
+
+ if (le32_to_cpu(bn->header.flags) & LEAF_NODE)
+ break;
+
+ memcpy(&value_le, value_ptr(bn, n->index), sizeof(value_le));
+ r = push_node(c, le64_to_cpu(value_le));
+ if (r) {
+ DMERR("push_node failed");
+ break;
+ }
+ }
+
+ if (!r && (le32_to_cpu(bn->header.nr_entries) == 0))
+ return -ENODATA;
+
+ return r;
+}
+
+int dm_btree_cursor_begin(struct dm_btree_info *info, dm_block_t root,
+ bool prefetch_leaves, struct dm_btree_cursor *c)
+{
+ int r;
+
+ c->info = info;
+ c->root = root;
+ c->depth = 0;
+ c->prefetch_leaves = prefetch_leaves;
+
+ r = push_node(c, root);
+ if (r)
+ return r;
+
+ return find_leaf(c);
+}
+EXPORT_SYMBOL_GPL(dm_btree_cursor_begin);
+
+void dm_btree_cursor_end(struct dm_btree_cursor *c)
+{
+ while (c->depth)
+ pop_node(c);
+}
+EXPORT_SYMBOL_GPL(dm_btree_cursor_end);
+
+int dm_btree_cursor_next(struct dm_btree_cursor *c)
+{
+ int r = inc_or_backtrack(c);
+ if (!r) {
+ r = find_leaf(c);
+ if (r)
+ DMERR("find_leaf failed");
+ }
+
+ return r;
+}
+EXPORT_SYMBOL_GPL(dm_btree_cursor_next);
+
+int dm_btree_cursor_get_value(struct dm_btree_cursor *c, uint64_t *key, void *value_le)
+{
+ if (c->depth) {
+ struct cursor_node *n = c->nodes + c->depth - 1;
+ struct btree_node *bn = dm_block_data(n->b);
+
+ if (le32_to_cpu(bn->header.flags) & INTERNAL_NODE)
+ return -EINVAL;
+
+ *key = le64_to_cpu(*key_ptr(bn, n->index));
+ memcpy(value_le, value_ptr(bn, n->index), c->info->value_type.size);
+ return 0;
+
+ } else
+ return -ENODATA;
+}
+EXPORT_SYMBOL_GPL(dm_btree_cursor_get_value);