6.1-stable review patch. If anyone has any objections, please let me know.
------------------
From: Hou Tao houtao1@huawei.com
[ Upstream commit 532d6b36b2bfac5514426a97a4df8d103d700d43 ]
When a LPM trie is full, in-place updates of existing elements incorrectly return -ENOSPC.
Fix this by deferring the check of trie->n_entries. For new insertions, n_entries must not exceed max_entries. However, in-place updates are allowed even when the trie is full.
Fixes: b95a5c4db09b ("bpf: add a longest prefix match trie map implementation") Reviewed-by: Toke Høiland-Jørgensen toke@redhat.com Signed-off-by: Hou Tao houtao1@huawei.com Link: https://lore.kernel.org/r/20241206110622.1161752-5-houtao@huaweicloud.com Signed-off-by: Alexei Starovoitov ast@kernel.org Signed-off-by: Sasha Levin sashal@kernel.org --- kernel/bpf/lpm_trie.c | 44 +++++++++++++++++++++---------------------- 1 file changed, 21 insertions(+), 23 deletions(-)
diff --git a/kernel/bpf/lpm_trie.c b/kernel/bpf/lpm_trie.c index 694fbf9891f4f..c5fa2a74de771 100644 --- a/kernel/bpf/lpm_trie.c +++ b/kernel/bpf/lpm_trie.c @@ -302,6 +302,16 @@ static struct lpm_trie_node *lpm_trie_node_alloc(const struct lpm_trie *trie, return node; }
+static int trie_check_add_elem(struct lpm_trie *trie, u64 flags) +{ + if (flags == BPF_EXIST) + return -ENOENT; + if (trie->n_entries == trie->map.max_entries) + return -ENOSPC; + trie->n_entries++; + return 0; +} + /* Called from syscall or from eBPF program */ static int trie_update_elem(struct bpf_map *map, void *_key, void *value, u64 flags) @@ -325,20 +335,12 @@ static int trie_update_elem(struct bpf_map *map, spin_lock_irqsave(&trie->lock, irq_flags);
/* Allocate and fill a new node */ - - if (trie->n_entries == trie->map.max_entries) { - ret = -ENOSPC; - goto out; - } - new_node = lpm_trie_node_alloc(trie, value); if (!new_node) { ret = -ENOMEM; goto out; }
- trie->n_entries++; - new_node->prefixlen = key->prefixlen; RCU_INIT_POINTER(new_node->child[0], NULL); RCU_INIT_POINTER(new_node->child[1], NULL); @@ -368,10 +370,10 @@ static int trie_update_elem(struct bpf_map *map, * simply assign the @new_node to that slot and be done. */ if (!node) { - if (flags == BPF_EXIST) { - ret = -ENOENT; + ret = trie_check_add_elem(trie, flags); + if (ret) goto out; - } + rcu_assign_pointer(*slot, new_node); goto out; } @@ -385,10 +387,10 @@ static int trie_update_elem(struct bpf_map *map, ret = -EEXIST; goto out; } - trie->n_entries--; - } else if (flags == BPF_EXIST) { - ret = -ENOENT; - goto out; + } else { + ret = trie_check_add_elem(trie, flags); + if (ret) + goto out; }
new_node->child[0] = node->child[0]; @@ -400,10 +402,9 @@ static int trie_update_elem(struct bpf_map *map, goto out; }
- if (flags == BPF_EXIST) { - ret = -ENOENT; + ret = trie_check_add_elem(trie, flags); + if (ret) goto out; - }
/* If the new node matches the prefix completely, it must be inserted * as an ancestor. Simply insert it between @node and *@slot. @@ -417,6 +418,7 @@ static int trie_update_elem(struct bpf_map *map,
im_node = lpm_trie_node_alloc(trie, NULL); if (!im_node) { + trie->n_entries--; ret = -ENOMEM; goto out; } @@ -438,12 +440,8 @@ static int trie_update_elem(struct bpf_map *map, rcu_assign_pointer(*slot, im_node);
out: - if (ret) { - if (new_node) - trie->n_entries--; + if (ret) kfree(new_node); - } - spin_unlock_irqrestore(&trie->lock, irq_flags); kfree_rcu(free_node, rcu);