[PATCH 2/3] netfilter: ipvs: release refcnt and improve ip_vs_mh_reassign()

[Inju Song <inju.song@xxxxxxxxxxxxx>]

  1) release refcnt of old dests in lookup entry

  2) handle exception about svc->num_dests == 0
     - skip permutate and populate if dest list is empty

  3) change allocation method
     - in case of allocating permutate array, use vmalloc()
       instead of kalloc()

Signed-off-by: Inju Song <inju.song@xxxxxxxxxxxxx>
---
 net/netfilter/ipvs/ip_vs_mh.c | 58 +++++++++++++++++++++++++++++--------------
 1 file changed, 39 insertions(+), 19 deletions(-)

diff --git a/net/netfilter/ipvs/ip_vs_mh.c b/net/netfilter/ipvs/ip_vs_mh.c
index b3fa8c4..bb20b96 100644
--- a/net/netfilter/ipvs/ip_vs_mh.c
+++ b/net/netfilter/ipvs/ip_vs_mh.c
@@ -5,6 +5,7 @@
 #include <linux/slab.h>
 #include <linux/module.h>
 #include <linux/kernel.h>
+#include <linux/vmalloc.h>
 #include <linux/skbuff.h>
 
 #include <net/ip_vs.h>
@@ -74,6 +75,12 @@ static inline int ip_vs_mh_permutate(struct ip_vs_mh_state 
*s,
        struct list_head *p;
        struct ip_vs_dest *dest;
        unsigned int offset, skip;
+       bool empty;
+
+       p = &svc->destinations;
+       empty = list_empty(p);
+       if (empty)
+               return 0;
 
        /* extending permutation table to 2d arrays */
        for (i = 1; i < svc->num_dests; i++)
@@ -81,7 +88,6 @@ static inline int ip_vs_mh_permutate(struct ip_vs_mh_state *s,
                                    IP_VS_MH_LOOKUP_SIZE;
 
        i = 0;
-       p = &svc->destinations;
        while ((p = p->next) != &svc->destinations) {
                dest = list_entry(p, struct ip_vs_dest, n_list);
                offset = ip_vs_mh_hashkey(svc->af, &dest->addr, dest->port,
@@ -110,6 +116,7 @@ static inline int ip_vs_mh_permutate(struct ip_vs_mh_state 
*s,
        struct list_head *p;
        struct ip_vs_dest *dest;
        unsigned int n, c;
+       bool empty;
 
        ret = 0;
        next = kcalloc(IP_VS_MH_TAB_SIZE, sizeof(unsigned int),
@@ -128,11 +135,17 @@ static inline int ip_vs_mh_permutate(struct 
ip_vs_mh_state *s,
                RCU_INIT_POINTER(entry[i].dest, NULL);
 
        n = 0;
+       p = &svc->destinations;
+       empty = list_empty(p);
+       if (empty)
+               goto out;
+
        pmt = s->permutation;
        while (n < IP_VS_MH_LOOKUP_SIZE) {
-               p = &svc->destinations;
                for (i = 0; i < svc->num_dests; i++) {
-                       p = p->next;
+                       if (p == &svc->destinations)
+                               p = p->next;
+
                        c = pmt[i][next[i]];
 
                        while (entry[c].dest) {
@@ -147,20 +160,29 @@ static inline int ip_vs_mh_permutate(struct 
ip_vs_mh_state *s,
                        n++;
                        if (n == IP_VS_MH_LOOKUP_SIZE)
                                break;
+
+                       p = p->next;
                }
        }
 
+out:
        l = &s->lookup[0];
        for (i = 0; i < IP_VS_MH_LOOKUP_SIZE; i++) {
                dest = rcu_dereference_protected(entry[i].dest, 1);
-               ip_vs_dest_hold(dest);
-               RCU_INIT_POINTER(l->dest, dest);
+               if (dest)
+                       ip_vs_dest_put(dest);
+               if (empty) {
+                       RCU_INIT_POINTER(l->dest, NULL);
+               } else {
+                       ip_vs_dest_hold(dest);
+                       RCU_INIT_POINTER(l->dest, dest);
 
-               IP_VS_DBG_BUF(6, "assigned i: %d dest: %s weight: %d\n",
-                             i, IP_VS_DBG_ADDR(dest->af, &dest->addr),
-                             atomic_read(&dest->weight));
+                       IP_VS_DBG_BUF(6, "assigned i: %d dest: %s weight: %d\n",
+                                     i, IP_VS_DBG_ADDR(dest->af, &dest->addr),
+                                     atomic_read(&dest->weight));
 
-               RCU_INIT_POINTER(entry[i].dest, NULL);
+                       RCU_INIT_POINTER(entry[i].dest, NULL);
+               }
                l++;
        }
 
@@ -250,19 +272,15 @@ static void ip_vs_mh_flush(struct ip_vs_mh_state *s)
 {
        int ret;
 
-       /* flush all the hash entry before assigning mh entry */
-       ip_vs_mh_flush(s);
-
        /* ip_vs_mh_reassign is responsible for assigning
         * and releasing s->permutation table
         */
-       s->permutation = kcalloc(IP_VS_MH_TAB_SIZE, sizeof(unsigned int *),
-                                GFP_KERNEL);
+       s->permutation = vzalloc(IP_VS_MH_TAB_SIZE * sizeof(unsigned int *));
        if (!s->permutation)
                return -ENOMEM;
 
-       s->permutation[0] = kcalloc(IP_VS_MH_TAB_SIZE * IP_VS_MH_LOOKUP_SIZE,
-                                   sizeof(unsigned int), GFP_KERNEL);
+       s->permutation[0] = vzalloc(IP_VS_MH_TAB_SIZE * IP_VS_MH_LOOKUP_SIZE *
+                                   sizeof(unsigned int));
        if (!s->permutation[0]) {
                ret = -ENOMEM;
                goto err_alloc;
@@ -276,12 +294,14 @@ static void ip_vs_mh_flush(struct ip_vs_mh_state *s)
        if (ret < 0)
                goto err_out;
 
-       return 0;
+       IP_VS_DBG_BUF(6, "MH: reassign lookup table of %s:%d\n",
+                     IP_VS_DBG_ADDR(svc->af, &svc->addr),
+                     ntohs(svc->port));
 
 err_out:
-       kfree(s->permutation[0]);
+       vfree(s->permutation[0]);
 err_alloc:
-       kfree(s->permutation);
+       vfree(s->permutation);
        return ret;
 }
 
-- 
1.8.3.1


-- 
Inju Song
NAVER Corporation

--
To unsubscribe from this list: send the line "unsubscribe lvs-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html