On 2014-07-29 at 13:41:33 +0200, Thomas Graf <tg...@suug.ch> wrote:
> Heavy Netlink users such as Open vSwitch spend a considerable amount of
> time in netlink_lookup() due to the read-lock on nl_table_lock. Use of
> RCU relieves the lock contention.
>
> Makes use of the new resizable hash table to avoid locking on the
> lookup.
>
> The hash table will grow if entries exceeds 75% of table size up to a
> total table size of 64K. It will automatically shrink if usage falls
> below 50%.
>
> Also splits nl_table_lock into a separate spinlock to protect hash table
> mutations. This avoids a possible deadlock when the hash table growing
> waits on RCU readers to complete via synchronize_rcu() while readers
> holding RCU read lock are waiting on the nl_table_lock() to be released
> to lock the table for broadcasting.
>
> Before:
> 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup
> 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers
> 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker
> 6.23% kpktgend_0 [kernel.kallsyms] [k] memset
> 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup
> 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy
> 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract
> 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2
>
> After:
> 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup
> 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker
> 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers
> 5.11% kpktgend_0 [kernel.kallsyms] [k] memset
> 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract
> 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock
> 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2
> [...]
> 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup
>
> Signed-off-by: Thomas Graf <tg...@suug.ch>
> ---
> net/netlink/af_netlink.c | 285
> ++++++++++++++++++-----------------------------
> net/netlink/af_netlink.h | 18 +--
> net/netlink/diag.c | 11 +-
> 3 files changed, 119 insertions(+), 195 deletions(-)
>
> diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
> index 1b38f7f..7f44468 100644
> --- a/net/netlink/af_netlink.c
> +++ b/net/netlink/af_netlink.c
[...]
> @@ -2996,28 +2939,26 @@ static void *netlink_seq_next(struct seq_file *seq,
> void *v, loff_t *pos)
>
> net = seq_file_net(seq);
> iter = seq->private;
> - s = v;
> - do {
> - s = sk_next(s);
> - } while (s && !nl_table[s->sk_protocol].compare(net, s));
> - if (s)
> - return s;
> + nlk = v;
> +
> + rht_for_each_entry_rcu(nlk, nlk->node.next, node)
> + if (net_eq(sock_net((struct sock *)nlk), net))
> + return nlk;
>
> i = iter->link;
> j = iter->hash_idx + 1;
>
> do {
> - struct nl_portid_hash *hash = &nl_table[i].hash;
> -
> - for (; j <= hash->mask; j++) {
> - s = sk_head(&hash->table[j]);
> + struct rhashtable *ht = &nl_table[i].hash;
> + const struct bucket_table *tbl = rht_dereference(ht->tbl, ht);
>
> - while (s && !nl_table[s->sk_protocol].compare(net, s))
> - s = sk_next(s);
> - if (s) {
> - iter->link = i;
> - iter->hash_idx = j;
> - return s;
> + for (; j <= tbl->size; j++) {
Should the condition j < tbl->size here, since the bucket table only
contains `size' buckets?
> + rht_for_each_entry_rcu(nlk, tbl->buckets[j], node) {
> + if (net_eq(sock_net((struct sock *)nlk), net)) {
> + iter->link = i;
> + iter->hash_idx = j;
> + return nlk;
> + }
> }
> }
[...]
> diff --git a/net/netlink/diag.c b/net/netlink/diag.c
> index 1af29624..7588f34 100644
> --- a/net/netlink/diag.c
> +++ b/net/netlink/diag.c
[...]
> @@ -101,16 +102,20 @@ static int __netlink_diag_dump(struct sk_buff *skb,
> struct netlink_callback *cb,
> int protocol, int s_num)
> {
> struct netlink_table *tbl = &nl_table[protocol];
> - struct nl_portid_hash *hash = &tbl->hash;
> + struct rhashtable *ht = &tbl->hash;
> + const struct bucket_table *htbl = rht_dereference(ht->tbl, ht);
> struct net *net = sock_net(skb->sk);
> struct netlink_diag_req *req;
> + struct netlink_sock *nlsk;
> struct sock *sk;
> int ret = 0, num = 0, i;
>
> req = nlmsg_data(cb->nlh);
>
> - for (i = 0; i <= hash->mask; i++) {
> - sk_for_each(sk, &hash->table[i]) {
> + for (i = 0; i <= htbl->size; i++) {
Same here, condition should be i < htbl->size
> + rht_for_each_entry(nlsk, htbl->buckets[i], ht, node) {
> + sk = (struct sock *)nlsk;
> +
> if (!net_eq(sock_net(sk), net))
> continue;
> if (num < s_num) {
> --
> 1.9.3
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
