From 2871c69e025e8bc507651d5a9cf81a8a7da9d24b Mon Sep 17 00:00:00 2001 From: Joe Thornber Date: Wed, 21 Oct 2015 18:36:49 +0100 Subject: dm btree remove: fix a bug when rebalancing nodes after removal Commit 4c7e309340ff ("dm btree remove: fix bug in redistribute3") wasn't a complete fix for redistribute3(). The redistribute3 function takes 3 btree nodes and shares out the entries evenly between them. If the three nodes in total contained (MAX_ENTRIES * 3) - 1 entries between them then this was erroneously getting rebalanced as (MAX_ENTRIES - 1) on the left and right, and (MAX_ENTRIES + 1) in the center. Fix this issue by being more careful about calculating the target number of entries for the left and right nodes. Unit tested in userspace using this program: https://github.com/jthornber/redistribute3-test/blob/master/redistribute3_t.c Signed-off-by: Joe Thornber Signed-off-by: Mike Snitzer Cc: stable@vger.kernel.org --- drivers/md/persistent-data/dm-btree-remove.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) (limited to 'drivers/md/persistent-data') diff --git a/drivers/md/persistent-data/dm-btree-remove.c b/drivers/md/persistent-data/dm-btree-remove.c index 421a36c593e3..2e4c4cb79e4d 100644 --- a/drivers/md/persistent-data/dm-btree-remove.c +++ b/drivers/md/persistent-data/dm-btree-remove.c @@ -301,11 +301,16 @@ static void redistribute3(struct dm_btree_info *info, struct btree_node *parent, { int s; uint32_t max_entries = le32_to_cpu(left->header.max_entries); - unsigned target = (nr_left + nr_center + nr_right) / 3; - BUG_ON(target > max_entries); + unsigned total = nr_left + nr_center + nr_right; + unsigned target_right = total / 3; + unsigned remainder = (target_right * 3) != total; + unsigned target_left = target_right + remainder; + + BUG_ON(target_left > max_entries); + BUG_ON(target_right > max_entries); if (nr_left < nr_right) { - s = nr_left - target; + s = nr_left - target_left; if (s < 0 && nr_center < -s) { /* not enough in central node */ @@ -316,10 +321,10 @@ static void redistribute3(struct dm_btree_info *info, struct btree_node *parent, } else shift(left, center, s); - shift(center, right, target - nr_right); + shift(center, right, target_right - nr_right); } else { - s = target - nr_right; + s = target_right - nr_right; if (s > 0 && nr_center < s) { /* not enough in central node */ shift(center, right, nr_center); @@ -329,7 +334,7 @@ static void redistribute3(struct dm_btree_info *info, struct btree_node *parent, } else shift(center, right, s); - shift(left, center, nr_left - target); + shift(left, center, nr_left - target_left); } *key_ptr(parent, c->index) = center->keys[0]; -- cgit v1.2.3 From 4dcb8b57df3593dcb20481d9d6cf79d1dc1534be Mon Sep 17 00:00:00 2001 From: Mike Snitzer Date: Thu, 22 Oct 2015 10:56:40 -0400 Subject: dm btree: fix leak of bufio-backed block in btree_split_beneath error path btree_split_beneath()'s error path had an outstanding FIXME that speaks directly to the potential for _not_ cleaning up a previously allocated bufio-backed block. Fix this by releasing the previously allocated bufio block using unlock_block(). Reported-by: Mikulas Patocka Signed-off-by: Mike Snitzer Acked-by: Joe Thornber Cc: stable@vger.kernel.org --- drivers/md/persistent-data/dm-btree.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/md/persistent-data') diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c index b6cec258cc21..0e09aef43998 100644 --- a/drivers/md/persistent-data/dm-btree.c +++ b/drivers/md/persistent-data/dm-btree.c @@ -523,7 +523,7 @@ static int btree_split_beneath(struct shadow_spine *s, uint64_t key) r = new_block(s->info, &right); if (r < 0) { - /* FIXME: put left */ + unlock_block(s->info, left); return r; } -- cgit v1.2.3 From 4c7da06f5a780bbf44ebd7547789e48536d0a823 Mon Sep 17 00:00:00 2001 From: Mikulas Patocka Date: Thu, 22 Oct 2015 16:46:59 -0400 Subject: dm persistent data: eliminate unnecessary return values dm_bm_unlock and dm_tm_unlock return an integer value but the returned value is always 0. The calling code sometimes checks the return value and sometimes doesn't. Eliminate these unnecessary return values and also the checks for them. Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer --- drivers/md/persistent-data/dm-array.c | 4 +-- drivers/md/persistent-data/dm-block-manager.c | 4 +-- drivers/md/persistent-data/dm-block-manager.h | 2 +- drivers/md/persistent-data/dm-btree-internal.h | 2 +- drivers/md/persistent-data/dm-btree-remove.c | 36 ++++++---------------- drivers/md/persistent-data/dm-btree-spine.c | 20 ++++-------- drivers/md/persistent-data/dm-btree.c | 4 ++- drivers/md/persistent-data/dm-space-map-common.c | 32 +++++++++---------- .../md/persistent-data/dm-transaction-manager.c | 4 +-- .../md/persistent-data/dm-transaction-manager.h | 2 +- 10 files changed, 42 insertions(+), 68 deletions(-) (limited to 'drivers/md/persistent-data') diff --git a/drivers/md/persistent-data/dm-array.c b/drivers/md/persistent-data/dm-array.c index e64b61ad0ef3..431a03067d64 100644 --- a/drivers/md/persistent-data/dm-array.c +++ b/drivers/md/persistent-data/dm-array.c @@ -233,9 +233,9 @@ static int get_ablock(struct dm_array_info *info, dm_block_t b, /* * Unlocks an array block. */ -static int unlock_ablock(struct dm_array_info *info, struct dm_block *block) +static void unlock_ablock(struct dm_array_info *info, struct dm_block *block) { - return dm_tm_unlock(info->btree_info.tm, block); + dm_tm_unlock(info->btree_info.tm, block); } /*----------------------------------------------------------------*/ diff --git a/drivers/md/persistent-data/dm-block-manager.c b/drivers/md/persistent-data/dm-block-manager.c index 88dbe7b97c2c..f2393ba838eb 100644 --- a/drivers/md/persistent-data/dm-block-manager.c +++ b/drivers/md/persistent-data/dm-block-manager.c @@ -578,7 +578,7 @@ int dm_bm_write_lock_zero(struct dm_block_manager *bm, } EXPORT_SYMBOL_GPL(dm_bm_write_lock_zero); -int dm_bm_unlock(struct dm_block *b) +void dm_bm_unlock(struct dm_block *b) { struct buffer_aux *aux; aux = dm_bufio_get_aux_data(to_buffer(b)); @@ -590,8 +590,6 @@ int dm_bm_unlock(struct dm_block *b) bl_up_read(&aux->lock); dm_bufio_release(to_buffer(b)); - - return 0; } EXPORT_SYMBOL_GPL(dm_bm_unlock); diff --git a/drivers/md/persistent-data/dm-block-manager.h b/drivers/md/persistent-data/dm-block-manager.h index 84330f59886d..3627d1b7667a 100644 --- a/drivers/md/persistent-data/dm-block-manager.h +++ b/drivers/md/persistent-data/dm-block-manager.h @@ -94,7 +94,7 @@ int dm_bm_write_lock_zero(struct dm_block_manager *bm, dm_block_t b, struct dm_block_validator *v, struct dm_block **result); -int dm_bm_unlock(struct dm_block *b); +void dm_bm_unlock(struct dm_block *b); /* * It's a common idiom to have a superblock that should be committed last. diff --git a/drivers/md/persistent-data/dm-btree-internal.h b/drivers/md/persistent-data/dm-btree-internal.h index 8731b6ea026b..a240990a7f33 100644 --- a/drivers/md/persistent-data/dm-btree-internal.h +++ b/drivers/md/persistent-data/dm-btree-internal.h @@ -52,7 +52,7 @@ void inc_children(struct dm_transaction_manager *tm, struct btree_node *n, struct dm_btree_value_type *vt); int new_block(struct dm_btree_info *info, struct dm_block **result); -int unlock_block(struct dm_btree_info *info, struct dm_block *b); +void unlock_block(struct dm_btree_info *info, struct dm_block *b); /* * Spines keep track of the rolling locks. There are 2 variants, read-only diff --git a/drivers/md/persistent-data/dm-btree-remove.c b/drivers/md/persistent-data/dm-btree-remove.c index 421a36c593e3..4d7617fc6dbb 100644 --- a/drivers/md/persistent-data/dm-btree-remove.c +++ b/drivers/md/persistent-data/dm-btree-remove.c @@ -165,9 +165,9 @@ static int init_child(struct dm_btree_info *info, struct dm_btree_value_type *vt return 0; } -static int exit_child(struct dm_btree_info *info, struct child *c) +static void exit_child(struct dm_btree_info *info, struct child *c) { - return dm_tm_unlock(info->tm, c->block); + dm_tm_unlock(info->tm, c->block); } static void shift(struct btree_node *left, struct btree_node *right, int count) @@ -249,13 +249,10 @@ static int rebalance2(struct shadow_spine *s, struct dm_btree_info *info, __rebalance2(info, parent, &left, &right); - r = exit_child(info, &left); - if (r) { - exit_child(info, &right); - return r; - } + exit_child(info, &left); + exit_child(info, &right); - return exit_child(info, &right); + return 0; } /* @@ -389,22 +386,9 @@ static int rebalance3(struct shadow_spine *s, struct dm_btree_info *info, __rebalance3(info, parent, &left, ¢er, &right); - r = exit_child(info, &left); - if (r) { - exit_child(info, ¢er); - exit_child(info, &right); - return r; - } - - r = exit_child(info, ¢er); - if (r) { - exit_child(info, &right); - return r; - } - - r = exit_child(info, &right); - if (r) - return r; + exit_child(info, &left); + exit_child(info, ¢er); + exit_child(info, &right); return 0; } @@ -428,9 +412,7 @@ static int rebalance_children(struct shadow_spine *s, memcpy(n, dm_block_data(child), dm_bm_block_size(dm_tm_get_bm(info->tm))); - r = dm_tm_unlock(info->tm, child); - if (r) - return r; + dm_tm_unlock(info->tm, child); dm_tm_dec(info->tm, dm_block_location(child)); return 0; diff --git a/drivers/md/persistent-data/dm-btree-spine.c b/drivers/md/persistent-data/dm-btree-spine.c index 0dee514ba4c5..b27b8091a1ca 100644 --- a/drivers/md/persistent-data/dm-btree-spine.c +++ b/drivers/md/persistent-data/dm-btree-spine.c @@ -117,9 +117,9 @@ int new_block(struct dm_btree_info *info, struct dm_block **result) return dm_tm_new_block(info->tm, &btree_node_validator, result); } -int unlock_block(struct dm_btree_info *info, struct dm_block *b) +void unlock_block(struct dm_btree_info *info, struct dm_block *b) { - return dm_tm_unlock(info->tm, b); + dm_tm_unlock(info->tm, b); } /*----------------------------------------------------------------*/ @@ -137,9 +137,7 @@ int exit_ro_spine(struct ro_spine *s) int r = 0, i; for (i = 0; i < s->count; i++) { - int r2 = unlock_block(s->info, s->nodes[i]); - if (r2 < 0) - r = r2; + unlock_block(s->info, s->nodes[i]); } return r; @@ -150,9 +148,7 @@ int ro_step(struct ro_spine *s, dm_block_t new_child) int r; if (s->count == 2) { - r = unlock_block(s->info, s->nodes[0]); - if (r < 0) - return r; + unlock_block(s->info, s->nodes[0]); s->nodes[0] = s->nodes[1]; s->count--; } @@ -194,9 +190,7 @@ int exit_shadow_spine(struct shadow_spine *s) int r = 0, i; for (i = 0; i < s->count; i++) { - int r2 = unlock_block(s->info, s->nodes[i]); - if (r2 < 0) - r = r2; + unlock_block(s->info, s->nodes[i]); } return r; @@ -208,9 +202,7 @@ int shadow_step(struct shadow_spine *s, dm_block_t b, int r; if (s->count == 2) { - r = unlock_block(s->info, s->nodes[0]); - if (r < 0) - return r; + unlock_block(s->info, s->nodes[0]); s->nodes[0] = s->nodes[1]; s->count--; } diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c index b6cec258cc21..25a62b73cfc0 100644 --- a/drivers/md/persistent-data/dm-btree.c +++ b/drivers/md/persistent-data/dm-btree.c @@ -141,7 +141,9 @@ int dm_btree_empty(struct dm_btree_info *info, dm_block_t *root) n->header.value_size = cpu_to_le32(info->value_type.size); *root = dm_block_location(b); - return unlock_block(info, b); + unlock_block(info, b); + + return 0; } EXPORT_SYMBOL_GPL(dm_btree_empty); diff --git a/drivers/md/persistent-data/dm-space-map-common.c b/drivers/md/persistent-data/dm-space-map-common.c index aacbe70c2c2e..306d2e4502c4 100644 --- a/drivers/md/persistent-data/dm-space-map-common.c +++ b/drivers/md/persistent-data/dm-space-map-common.c @@ -259,9 +259,7 @@ int sm_ll_extend(struct ll_disk *ll, dm_block_t extra_blocks) idx.blocknr = cpu_to_le64(dm_block_location(b)); - r = dm_tm_unlock(ll->tm, b); - if (r < 0) - return r; + dm_tm_unlock(ll->tm, b); idx.nr_free = cpu_to_le32(ll->entries_per_block); idx.none_free_before = 0; @@ -293,7 +291,9 @@ int sm_ll_lookup_bitmap(struct ll_disk *ll, dm_block_t b, uint32_t *result) *result = sm_lookup_bitmap(dm_bitmap_data(blk), b); - return dm_tm_unlock(ll->tm, blk); + dm_tm_unlock(ll->tm, blk); + + return 0; } static int sm_ll_lookup_big_ref_count(struct ll_disk *ll, dm_block_t b, @@ -373,9 +373,7 @@ int sm_ll_find_free_block(struct ll_disk *ll, dm_block_t begin, return r; } - r = dm_tm_unlock(ll->tm, blk); - if (r < 0) - return r; + dm_tm_unlock(ll->tm, blk); *result = i * ll->entries_per_block + (dm_block_t) position; return 0; @@ -429,9 +427,7 @@ static int sm_ll_mutate(struct ll_disk *ll, dm_block_t b, if (ref_count <= 2) { sm_set_bitmap(bm_le, bit, ref_count); - r = dm_tm_unlock(ll->tm, nb); - if (r < 0) - return r; + dm_tm_unlock(ll->tm, nb); if (old > 2) { r = dm_btree_remove(&ll->ref_count_info, @@ -445,9 +441,7 @@ static int sm_ll_mutate(struct ll_disk *ll, dm_block_t b, __le32 le_rc = cpu_to_le32(ref_count); sm_set_bitmap(bm_le, bit, 3); - r = dm_tm_unlock(ll->tm, nb); - if (r < 0) - return r; + dm_tm_unlock(ll->tm, nb); __dm_bless_for_disk(&le_rc); r = dm_btree_insert(&ll->ref_count_info, ll->ref_count_root, @@ -556,7 +550,9 @@ static int metadata_ll_init_index(struct ll_disk *ll) memcpy(dm_block_data(b), &ll->mi_le, sizeof(ll->mi_le)); ll->bitmap_root = dm_block_location(b); - return dm_tm_unlock(ll->tm, b); + dm_tm_unlock(ll->tm, b); + + return 0; } static int metadata_ll_open(struct ll_disk *ll) @@ -570,7 +566,9 @@ static int metadata_ll_open(struct ll_disk *ll) return r; memcpy(&ll->mi_le, dm_block_data(block), sizeof(ll->mi_le)); - return dm_tm_unlock(ll->tm, block); + dm_tm_unlock(ll->tm, block); + + return 0; } static dm_block_t metadata_ll_max_entries(struct ll_disk *ll) @@ -590,7 +588,9 @@ static int metadata_ll_commit(struct ll_disk *ll) memcpy(dm_block_data(b), &ll->mi_le, sizeof(ll->mi_le)); ll->bitmap_root = dm_block_location(b); - return dm_tm_unlock(ll->tm, b); + dm_tm_unlock(ll->tm, b); + + return 0; } int sm_ll_new_metadata(struct ll_disk *ll, struct dm_transaction_manager *tm) diff --git a/drivers/md/persistent-data/dm-transaction-manager.c b/drivers/md/persistent-data/dm-transaction-manager.c index 9cb797d800cf..abe2c5dd0993 100644 --- a/drivers/md/persistent-data/dm-transaction-manager.c +++ b/drivers/md/persistent-data/dm-transaction-manager.c @@ -342,9 +342,9 @@ int dm_tm_read_lock(struct dm_transaction_manager *tm, dm_block_t b, } EXPORT_SYMBOL_GPL(dm_tm_read_lock); -int dm_tm_unlock(struct dm_transaction_manager *tm, struct dm_block *b) +void dm_tm_unlock(struct dm_transaction_manager *tm, struct dm_block *b) { - return dm_bm_unlock(b); + dm_bm_unlock(b); } EXPORT_SYMBOL_GPL(dm_tm_unlock); diff --git a/drivers/md/persistent-data/dm-transaction-manager.h b/drivers/md/persistent-data/dm-transaction-manager.h index 2e0d4d66fb1b..f3a18be68f30 100644 --- a/drivers/md/persistent-data/dm-transaction-manager.h +++ b/drivers/md/persistent-data/dm-transaction-manager.h @@ -94,7 +94,7 @@ int dm_tm_read_lock(struct dm_transaction_manager *tm, dm_block_t b, struct dm_block_validator *v, struct dm_block **result); -int dm_tm_unlock(struct dm_transaction_manager *tm, struct dm_block *b); +void dm_tm_unlock(struct dm_transaction_manager *tm, struct dm_block *b); /* * Functions for altering the reference count of a block directly. -- cgit v1.2.3 From 30ce6e1cc5a0f781d60227e9096c86e188d2c2bd Mon Sep 17 00:00:00 2001 From: Mike Snitzer Date: Mon, 23 Nov 2015 16:24:45 -0500 Subject: dm btree: fix leak of bufio-backed block in btree_split_sibling error path The block allocated at the start of btree_split_sibling() is never released if later insert_at() fails. Fix this by releasing the previously allocated bufio block using unlock_block(). Reported-by: Mikulas Patocka Signed-off-by: Mike Snitzer Cc: stable@vger.kernel.org --- drivers/md/persistent-data/dm-btree.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/md/persistent-data') diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c index c573402033b2..0918a7c5f809 100644 --- a/drivers/md/persistent-data/dm-btree.c +++ b/drivers/md/persistent-data/dm-btree.c @@ -473,8 +473,10 @@ static int btree_split_sibling(struct shadow_spine *s, unsigned parent_index, r = insert_at(sizeof(__le64), pn, parent_index + 1, le64_to_cpu(rn->keys[0]), &location); - if (r) + if (r) { + unlock_block(s->info, right); return r; + } if (key < le64_to_cpu(rn->keys[0])) { unlock_block(s->info, right); -- cgit v1.2.3 From 993ceab91986e2e737ce9a3e23bebc8cce649240 Mon Sep 17 00:00:00 2001 From: Joe Thornber Date: Wed, 2 Dec 2015 12:24:39 +0000 Subject: dm thin metadata: fix bug in dm_thin_remove_range() dm_btree_remove_leaves() only unmaps a contiguous region so we need a loop, in __remove_range(), to handle ranges that contain multiple regions. A new btree function, dm_btree_lookup_next(), is introduced which is more efficiently able to skip over regions of the thin device which aren't mapped. __remove_range() uses dm_btree_lookup_next() for each iteration of __remove_range()'s loop. Also, improve description of dm_btree_remove_leaves(). Fixes: 6550f075 ("dm thin metadata: add dm_thin_remove_range()") Signed-off-by: Joe Thornber Signed-off-by: Mike Snitzer Cc: stable@vger.kernel.org # 4.1+ --- drivers/md/persistent-data/dm-btree.c | 81 +++++++++++++++++++++++++++++++++++ drivers/md/persistent-data/dm-btree.h | 14 ++++-- 2 files changed, 92 insertions(+), 3 deletions(-) (limited to 'drivers/md/persistent-data') diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c index 0918a7c5f809..7e5b7f12958a 100644 --- a/drivers/md/persistent-data/dm-btree.c +++ b/drivers/md/persistent-data/dm-btree.c @@ -63,6 +63,11 @@ int lower_bound(struct btree_node *n, uint64_t key) return bsearch(n, key, 0); } +static int upper_bound(struct btree_node *n, uint64_t key) +{ + return bsearch(n, key, 1); +} + void inc_children(struct dm_transaction_manager *tm, struct btree_node *n, struct dm_btree_value_type *vt) { @@ -392,6 +397,82 @@ int dm_btree_lookup(struct dm_btree_info *info, dm_block_t root, } EXPORT_SYMBOL_GPL(dm_btree_lookup); +static int dm_btree_lookup_next_single(struct dm_btree_info *info, dm_block_t root, + uint64_t key, uint64_t *rkey, void *value_le) +{ + int r, i; + uint32_t flags, nr_entries; + struct dm_block *node; + struct btree_node *n; + + r = bn_read_lock(info, root, &node); + if (r) + return r; + + n = dm_block_data(node); + flags = le32_to_cpu(n->header.flags); + nr_entries = le32_to_cpu(n->header.nr_entries); + + if (flags & INTERNAL_NODE) { + i = lower_bound(n, key); + if (i < 0 || i >= nr_entries) { + r = -ENODATA; + goto out; + } + + r = dm_btree_lookup_next_single(info, value64(n, i), key, rkey, value_le); + if (r == -ENODATA && i < (nr_entries - 1)) { + i++; + r = dm_btree_lookup_next_single(info, value64(n, i), key, rkey, value_le); + } + + } else { + i = upper_bound(n, key); + if (i < 0 || i >= nr_entries) { + r = -ENODATA; + goto out; + } + + *rkey = le64_to_cpu(n->keys[i]); + memcpy(value_le, value_ptr(n, i), info->value_type.size); + } +out: + dm_tm_unlock(info->tm, node); + return r; +} + +int dm_btree_lookup_next(struct dm_btree_info *info, dm_block_t root, + uint64_t *keys, uint64_t *rkey, void *value_le) +{ + unsigned level; + int r = -ENODATA; + __le64 internal_value_le; + struct ro_spine spine; + + init_ro_spine(&spine, info); + for (level = 0; level < info->levels - 1u; level++) { + r = btree_lookup_raw(&spine, root, keys[level], + lower_bound, rkey, + &internal_value_le, sizeof(uint64_t)); + if (r) + goto out; + + if (*rkey != keys[level]) { + r = -ENODATA; + goto out; + } + + root = le64_to_cpu(internal_value_le); + } + + r = dm_btree_lookup_next_single(info, root, keys[level], rkey, value_le); +out: + exit_ro_spine(&spine); + return r; +} + +EXPORT_SYMBOL_GPL(dm_btree_lookup_next); + /* * Splits a node by creating a sibling node and shifting half the nodes * contents across. Assumes there is a parent node, and it has room for diff --git a/drivers/md/persistent-data/dm-btree.h b/drivers/md/persistent-data/dm-btree.h index 11d8cf78621d..c74301fa5a37 100644 --- a/drivers/md/persistent-data/dm-btree.h +++ b/drivers/md/persistent-data/dm-btree.h @@ -109,6 +109,13 @@ int dm_btree_del(struct dm_btree_info *info, dm_block_t root); int dm_btree_lookup(struct dm_btree_info *info, dm_block_t root, uint64_t *keys, void *value_le); +/* + * Tries to find the first key where the bottom level key is >= to that + * given. Useful for skipping empty sections of the btree. + */ +int dm_btree_lookup_next(struct dm_btree_info *info, dm_block_t root, + uint64_t *keys, uint64_t *rkey, void *value_le); + /* * Insertion (or overwrite an existing value). O(ln(n)) */ @@ -135,9 +142,10 @@ int dm_btree_remove(struct dm_btree_info *info, dm_block_t root, uint64_t *keys, dm_block_t *new_root); /* - * Removes values between 'keys' and keys2, where keys2 is keys with the - * final key replaced with 'end_key'. 'end_key' is the one-past-the-end - * value. 'keys' may be altered. + * Removes a _contiguous_ run of values starting from 'keys' and not + * reaching keys2 (where keys2 is keys with the final key replaced with + * 'end_key'). 'end_key' is the one-past-the-end value. 'keys' may be + * altered. */ int dm_btree_remove_leaves(struct dm_btree_info *info, dm_block_t root, uint64_t *keys, uint64_t end_key, -- cgit v1.2.3 From 50dd842ad83b43bed71790efb31cfb2f6c05c9c1 Mon Sep 17 00:00:00 2001 From: Joe Thornber Date: Wed, 9 Dec 2015 16:38:12 +0000 Subject: dm space map metadata: fix ref counting bug when bootstrapping a new space map When applying block operations (BOPs) do not remove them from the uncommitted BOP ring-buffer until after they've been applied -- in case we recurse. Also, perform BOP_INC operation, in dm_sm_metadata_create() and sm_metadata_extend(), in terms of the uncommitted BOP ring-buffer rather than using direct calls to sm_ll_inc(). Signed-off-by: Joe Thornber Signed-off-by: Mike Snitzer Cc: stable@vger.kernel.org --- drivers/md/persistent-data/dm-space-map-metadata.c | 32 +++++++++++++++------- 1 file changed, 22 insertions(+), 10 deletions(-) (limited to 'drivers/md/persistent-data') diff --git a/drivers/md/persistent-data/dm-space-map-metadata.c b/drivers/md/persistent-data/dm-space-map-metadata.c index 53091295fce9..fca6dbcf9a47 100644 --- a/drivers/md/persistent-data/dm-space-map-metadata.c +++ b/drivers/md/persistent-data/dm-space-map-metadata.c @@ -136,7 +136,7 @@ static int brb_push(struct bop_ring_buffer *brb, return 0; } -static int brb_pop(struct bop_ring_buffer *brb, struct block_op *result) +static int brb_peek(struct bop_ring_buffer *brb, struct block_op *result) { struct block_op *bop; @@ -147,6 +147,17 @@ static int brb_pop(struct bop_ring_buffer *brb, struct block_op *result) result->type = bop->type; result->block = bop->block; + return 0; +} + +static int brb_pop(struct bop_ring_buffer *brb) +{ + struct block_op *bop; + + if (brb_empty(brb)) + return -ENODATA; + + bop = brb->bops + brb->begin; brb->begin = brb_next(brb, brb->begin); return 0; @@ -211,7 +222,7 @@ static int apply_bops(struct sm_metadata *smm) while (!brb_empty(&smm->uncommitted)) { struct block_op bop; - r = brb_pop(&smm->uncommitted, &bop); + r = brb_peek(&smm->uncommitted, &bop); if (r) { DMERR("bug in bop ring buffer"); break; @@ -220,6 +231,8 @@ static int apply_bops(struct sm_metadata *smm) r = commit_bop(smm, &bop); if (r) break; + + brb_pop(&smm->uncommitted); } return r; @@ -683,7 +696,6 @@ static struct dm_space_map bootstrap_ops = { static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks) { int r, i; - enum allocation_event ev; struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm); dm_block_t old_len = smm->ll.nr_blocks; @@ -705,11 +717,12 @@ static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks) * allocate any new blocks. */ do { - for (i = old_len; !r && i < smm->begin; i++) { - r = sm_ll_inc(&smm->ll, i, &ev); - if (r) - goto out; - } + for (i = old_len; !r && i < smm->begin; i++) + r = add_bop(smm, BOP_INC, i); + + if (r) + goto out; + old_len = smm->begin; r = apply_bops(smm); @@ -754,7 +767,6 @@ int dm_sm_metadata_create(struct dm_space_map *sm, { int r; dm_block_t i; - enum allocation_event ev; struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm); smm->begin = superblock + 1; @@ -782,7 +794,7 @@ int dm_sm_metadata_create(struct dm_space_map *sm, * allocated blocks that they were built from. */ for (i = superblock; !r && i < smm->begin; i++) - r = sm_ll_inc(&smm->ll, i, &ev); + r = add_bop(smm, BOP_INC, i); if (r) return r; -- cgit v1.2.3 From ed8b45a3679eb49069b094c0711b30833f27c734 Mon Sep 17 00:00:00 2001 From: Joe Thornber Date: Thu, 10 Dec 2015 14:37:53 +0000 Subject: dm btree: fix bufio buffer leaks in dm_btree_del() error path If dm_btree_del()'s call to push_frame() fails, e.g. due to btree_node_validator finding invalid metadata, the dm_btree_del() error path must unlock all frames (which have active dm-bufio buffers) that were pushed onto the del_stack. Otherwise, dm_bufio_client_destroy() will BUG_ON() because dm-bufio buffers have leaked, e.g.: device-mapper: bufio: leaked buffer 3, hold count 1, list 0 Signed-off-by: Joe Thornber Signed-off-by: Mike Snitzer Cc: stable@vger.kernel.org --- drivers/md/persistent-data/dm-btree.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) (limited to 'drivers/md/persistent-data') diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c index 7e5b7f12958a..b1ced58eb5e1 100644 --- a/drivers/md/persistent-data/dm-btree.c +++ b/drivers/md/persistent-data/dm-btree.c @@ -257,6 +257,16 @@ static void pop_frame(struct del_stack *s) dm_tm_unlock(s->tm, f->b); } +static void unlock_all_frames(struct del_stack *s) +{ + struct frame *f; + + while (unprocessed_frames(s)) { + f = s->spine + s->top--; + dm_tm_unlock(s->tm, f->b); + } +} + int dm_btree_del(struct dm_btree_info *info, dm_block_t root) { int r; @@ -313,9 +323,13 @@ int dm_btree_del(struct dm_btree_info *info, dm_block_t root) pop_frame(s); } } - out: + if (r) { + /* cleanup all frames of del_stack */ + unlock_all_frames(s); + } kfree(s); + return r; } EXPORT_SYMBOL_GPL(dm_btree_del); -- cgit v1.2.3