Commit 13ee6ac579574a2a95e982b19920fd2495dce8cd
Committed by
Pablo Neira Ayuso
Pablo Neira Ayuso
1 parent
83723d6071
Exists in
master
and in
4 other branches
netfilter: fix race in conntrack between dump_table and destroy
The netlink interface to dump the connection tracking table has a race when entries are deleted at the same time. A customer reported a crash and the backtrace showed thatctnetlink_dump_table was running while a conntrack entry was being destroyed. (see https://bugzilla.vyatta.com/show_bug.cgi?id=6402). According to RCU documentation, when using hlist_nulls the reader must handle the case of seeing a deleted entry and not proceed further down the linked list. The old code would continue which caused the scan to walk into the free list. This patch uses locking (rather than RCU) for this operation which is guaranteed safe, and no longer requires getting reference while doing dump operation. Signed-off-by: Stephen Hemminger <shemminger@vyatta.com> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Showing 1 changed file with 5 additions and 9 deletions Side-by-side Diff
net/netfilter/nf_conntrack_netlink.c
| ... | ... | @@ -645,25 +645,23 @@ |
| 645 | 645 | struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); |
| 646 | 646 | u_int8_t l3proto = nfmsg->nfgen_family; |
| 647 | 647 | |
| 648 | - rcu_read_lock(); | |
| 648 | + spin_lock_bh(&nf_conntrack_lock); | |
| 649 | 649 | last = (struct nf_conn *)cb->args[1]; |
| 650 | 650 | for (; cb->args[0] < net->ct.htable_size; cb->args[0]++) { |
| 651 | 651 | restart: |
| 652 | - hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[cb->args[0]], | |
| 652 | + hlist_nulls_for_each_entry(h, n, &net->ct.hash[cb->args[0]], | |
| 653 | 653 | hnnode) { |
| 654 | 654 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) |
| 655 | 655 | continue; |
| 656 | 656 | ct = nf_ct_tuplehash_to_ctrack(h); |
| 657 | - if (!atomic_inc_not_zero(&ct->ct_general.use)) | |
| 658 | - continue; | |
| 659 | 657 | /* Dump entries of a given L3 protocol number. |
| 660 | 658 | * If it is not specified, ie. l3proto == 0, |
| 661 | 659 | * then dump everything. */ |
| 662 | 660 | if (l3proto && nf_ct_l3num(ct) != l3proto) |
| 663 | - goto releasect; | |
| 661 | + continue; | |
| 664 | 662 | if (cb->args[1]) { |
| 665 | 663 | if (ct != last) |
| 666 | - goto releasect; | |
| 664 | + continue; | |
| 667 | 665 | cb->args[1] = 0; |
| 668 | 666 | } |
| 669 | 667 | if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid, |
| ... | ... | @@ -681,8 +679,6 @@ |
| 681 | 679 | if (acct) |
| 682 | 680 | memset(acct, 0, sizeof(struct nf_conn_counter[IP_CT_DIR_MAX])); |
| 683 | 681 | } |
| 684 | -releasect: | |
| 685 | - nf_ct_put(ct); | |
| 686 | 682 | } |
| 687 | 683 | if (cb->args[1]) { |
| 688 | 684 | cb->args[1] = 0; |
| ... | ... | @@ -690,7 +686,7 @@ |
| 690 | 686 | } |
| 691 | 687 | } |
| 692 | 688 | out: |
| 693 | - rcu_read_unlock(); | |
| 689 | + spin_unlock_bh(&nf_conntrack_lock); | |
| 694 | 690 | if (last) |
| 695 | 691 | nf_ct_put(last); |
| 696 | 692 |