ipvs: convert rr scheduler to rcu

The schedule method now needs _rcu list-traversal
primitive for svc->destinations. As the previous entry
could be unlinked, limit the list traversals to 2 when
lookup started from previous entry.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>

diff --git a/net/netfilter/ipvs/ip_vs_rr.c b/net/netfilter/ipvs/ip_vs_rr.c
index c49b388..3942890 100644 (file)
--- a/net/netfilter/ipvs/ip_vs_rr.c
+++ b/net/netfilter/ipvs/ip_vs_rr.c
@@ -35,9 +35,18 @@ static int ip_vs_rr_init_svc(struct ip_vs_service *svc)
 }
 
 
-static int ip_vs_rr_update_svc(struct ip_vs_service *svc)
+static int ip_vs_rr_del_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest)
 {
-       svc->sched_data = &svc->destinations;
+       struct list_head *p;
+
+       write_lock_bh(&svc->sched_lock);
+       p = (struct list_head *) svc->sched_data;
+       /* dest is already unlinked, so p->prev is not valid but
+        * p->next is valid, use it to reach previous entry.
+        */
+       if (p == &dest->n_list)
+               svc->sched_data = p->next->prev;
+       write_unlock_bh(&svc->sched_lock);
        return 0;
 }
 
@@ -48,35 +57,40 @@ static int ip_vs_rr_update_svc(struct ip_vs_service *svc)
 static struct ip_vs_dest *
 ip_vs_rr_schedule(struct ip_vs_service *svc, const struct sk_buff *skb)
 {
-       struct list_head *p, *q;
-       struct ip_vs_dest *dest;
+       struct list_head *p;
+       struct ip_vs_dest *dest, *last;
+       int pass = 0;
 
        IP_VS_DBG(6, "%s(): Scheduling...\n", __func__);
 
        write_lock(&svc->sched_lock);
-       p = (struct list_head *)svc->sched_data;
-       p = p->next;
-       q = p;
+       p = (struct list_head *) svc->sched_data;
+       last = dest = list_entry(p, struct ip_vs_dest, n_list);
+
        do {
-               /* skip list head */
-               if (q == &svc->destinations) {
-                       q = q->next;
-                       continue;
+               list_for_each_entry_continue_rcu(dest,
+                                                &svc->destinations,
+                                                n_list) {
+                       if (!(dest->flags & IP_VS_DEST_F_OVERLOAD) &&
+                           atomic_read(&dest->weight) > 0)
+                               /* HIT */
+                               goto out;
+                       if (dest == last)
+                               goto stop;
                }
+               pass++;
+               /* Previous dest could be unlinked, do not loop forever.
+                * If we stay at head there is no need for 2nd pass.
+                */
+       } while (pass < 2 && p != &svc->destinations);
 
-               dest = list_entry(q, struct ip_vs_dest, n_list);
-               if (!(dest->flags & IP_VS_DEST_F_OVERLOAD) &&
-                   atomic_read(&dest->weight) > 0)
-                       /* HIT */
-                       goto out;
-               q = q->next;
-       } while (q != p);
+stop:
        write_unlock(&svc->sched_lock);
        ip_vs_scheduler_err(svc, "no destination available");
        return NULL;
 
   out:
-       svc->sched_data = q;
+       svc->sched_data = &dest->n_list;
        write_unlock(&svc->sched_lock);
        IP_VS_DBG_BUF(6, "RR: server %s:%u "
                      "activeconns %d refcnt %d weight %d\n",
@@ -94,7 +108,8 @@ static struct ip_vs_scheduler ip_vs_rr_scheduler = {
        .module =               THIS_MODULE,
        .n_list =               LIST_HEAD_INIT(ip_vs_rr_scheduler.n_list),
        .init_service =         ip_vs_rr_init_svc,
-       .update_service =       ip_vs_rr_update_svc,
+       .add_dest =             NULL,
+       .del_dest =             ip_vs_rr_del_dest,
        .schedule =             ip_vs_rr_schedule,
 };