ipv6 flowlabel: Convert hash list to RCU.
authorYOSHIFUJI Hideaki / 吉藤英明 <yoshfuji@linux-ipv6.org>
Wed, 30 Jan 2013 09:27:47 +0000 (09:27 +0000)
committerDavid S. Miller <davem@davemloft.net>
Thu, 31 Jan 2013 03:41:13 +0000 (22:41 -0500)
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/ipv6.h
net/ipv6/ip6_flowlabel.c

index dc30b60975efd8eeafe72945f783133b22abc258..1d457161def255dc5a3aa9e842881c4417407710 100644 (file)
@@ -222,6 +222,7 @@ struct ip6_flowlabel {
        struct in6_addr         dst;
        struct ipv6_txoptions   *opt;
        unsigned long           linger;
+       struct rcu_head         rcu;
        u8                      share;
        union {
                struct pid *pid;
index 5d767f1b878016d6458f1e4518da9790fe7bbe2b..da156015d827d6eeeadaae006c61306eefb8aaa3 100644 (file)
 #define FL_HASH(l)     (ntohl(l)&FL_HASH_MASK)
 
 static atomic_t fl_size = ATOMIC_INIT(0);
-static struct ip6_flowlabel *fl_ht[FL_HASH_MASK+1];
+static struct ip6_flowlabel __rcu *fl_ht[FL_HASH_MASK+1];
 
 static void ip6_fl_gc(unsigned long dummy);
 static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc, 0, 0);
 
 /* FL hash table lock: it protects only of GC */
 
-static DEFINE_RWLOCK(ip6_fl_lock);
+static DEFINE_SPINLOCK(ip6_fl_lock);
 
 /* Big socket sock */
 
 static DEFINE_RWLOCK(ip6_sk_fl_lock);
 
+#define for_each_fl_rcu(hash, fl)                              \
+       for (fl = rcu_dereference(fl_ht[(hash)]);               \
+            fl != NULL;                                        \
+            fl = rcu_dereference(fl->next))
+#define for_each_fl_continue_rcu(fl)                           \
+       for (fl = rcu_dereference(fl->next);                    \
+            fl != NULL;                                        \
+            fl = rcu_dereference(fl->next))
 
 static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label)
 {
        struct ip6_flowlabel *fl;
 
-       for (fl=fl_ht[FL_HASH(label)]; fl; fl = fl->next) {
+       for_each_fl_rcu(FL_HASH(label), fl) {
                if (fl->label == label && net_eq(fl->fl_net, net))
                        return fl;
        }
@@ -80,11 +88,11 @@ static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label)
 {
        struct ip6_flowlabel *fl;
 
-       read_lock_bh(&ip6_fl_lock);
+       rcu_read_lock_bh();
        fl = __fl_lookup(net, label);
-       if (fl)
-               atomic_inc(&fl->users);
-       read_unlock_bh(&ip6_fl_lock);
+       if (fl && !atomic_inc_not_zero(&fl->users))
+               fl = NULL;
+       rcu_read_unlock_bh();
        return fl;
 }
 
@@ -96,13 +104,13 @@ static void fl_free(struct ip6_flowlabel *fl)
                        put_pid(fl->owner.pid);
                release_net(fl->fl_net);
                kfree(fl->opt);
+               kfree_rcu(fl, rcu);
        }
-       kfree(fl);
 }
 
 static void fl_release(struct ip6_flowlabel *fl)
 {
-       write_lock_bh(&ip6_fl_lock);
+       spin_lock_bh(&ip6_fl_lock);
 
        fl->lastuse = jiffies;
        if (atomic_dec_and_test(&fl->users)) {
@@ -119,7 +127,7 @@ static void fl_release(struct ip6_flowlabel *fl)
                    time_after(ip6_fl_gc_timer.expires, ttd))
                        mod_timer(&ip6_fl_gc_timer, ttd);
        }
-       write_unlock_bh(&ip6_fl_lock);
+       spin_unlock_bh(&ip6_fl_lock);
 }
 
 static void ip6_fl_gc(unsigned long dummy)
@@ -128,12 +136,13 @@ static void ip6_fl_gc(unsigned long dummy)
        unsigned long now = jiffies;
        unsigned long sched = 0;
 
-       write_lock(&ip6_fl_lock);
+       spin_lock(&ip6_fl_lock);
 
        for (i=0; i<=FL_HASH_MASK; i++) {
                struct ip6_flowlabel *fl, **flp;
                flp = &fl_ht[i];
-               while ((fl=*flp) != NULL) {
+               while ((fl = rcu_dereference_protected(*flp,
+                                                      lockdep_is_held(&ip6_fl_lock))) != NULL) {
                        if (atomic_read(&fl->users) == 0) {
                                unsigned long ttd = fl->lastuse + fl->linger;
                                if (time_after(ttd, fl->expires))
@@ -156,18 +165,19 @@ static void ip6_fl_gc(unsigned long dummy)
        if (sched) {
                mod_timer(&ip6_fl_gc_timer, sched);
        }
-       write_unlock(&ip6_fl_lock);
+       spin_unlock(&ip6_fl_lock);
 }
 
 static void __net_exit ip6_fl_purge(struct net *net)
 {
        int i;
 
-       write_lock(&ip6_fl_lock);
+       spin_lock(&ip6_fl_lock);
        for (i = 0; i <= FL_HASH_MASK; i++) {
                struct ip6_flowlabel *fl, **flp;
                flp = &fl_ht[i];
-               while ((fl = *flp) != NULL) {
+               while ((fl = rcu_dereference_protected(*flp,
+                                                      lockdep_is_held(&ip6_fl_lock))) != NULL) {
                        if (net_eq(fl->fl_net, net) &&
                            atomic_read(&fl->users) == 0) {
                                *flp = fl->next;
@@ -178,7 +188,7 @@ static void __net_exit ip6_fl_purge(struct net *net)
                        flp = &fl->next;
                }
        }
-       write_unlock(&ip6_fl_lock);
+       spin_unlock(&ip6_fl_lock);
 }
 
 static struct ip6_flowlabel *fl_intern(struct net *net,
@@ -188,7 +198,7 @@ static struct ip6_flowlabel *fl_intern(struct net *net,
 
        fl->label = label & IPV6_FLOWLABEL_MASK;
 
-       write_lock_bh(&ip6_fl_lock);
+       spin_lock_bh(&ip6_fl_lock);
        if (label == 0) {
                for (;;) {
                        fl->label = htonl(net_random())&IPV6_FLOWLABEL_MASK;
@@ -210,16 +220,16 @@ static struct ip6_flowlabel *fl_intern(struct net *net,
                lfl = __fl_lookup(net, fl->label);
                if (lfl != NULL) {
                        atomic_inc(&lfl->users);
-                       write_unlock_bh(&ip6_fl_lock);
+                       spin_unlock_bh(&ip6_fl_lock);
                        return lfl;
                }
        }
 
        fl->lastuse = jiffies;
        fl->next = fl_ht[FL_HASH(fl->label)];
-       fl_ht[FL_HASH(fl->label)] = fl;
+       rcu_assign_pointer(fl_ht[FL_HASH(fl->label)], fl);
        atomic_inc(&fl_size);
-       write_unlock_bh(&ip6_fl_lock);
+       spin_unlock_bh(&ip6_fl_lock);
        return NULL;
 }
 
@@ -650,13 +660,13 @@ static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq)
        struct net *net = seq_file_net(seq);
 
        for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) {
-               fl = fl_ht[state->bucket];
-
-               while (fl && !net_eq(fl->fl_net, net))
-                       fl = fl->next;
-               if (fl)
-                       break;
+               for_each_fl_rcu(state->bucket, fl) {
+                       if (net_eq(fl->fl_net, net))
+                               goto out;
+               }
        }
+       fl = NULL;
+out:
        return fl;
 }
 
@@ -665,18 +675,22 @@ static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flo
        struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
        struct net *net = seq_file_net(seq);
 
-       fl = fl->next;
+       for_each_fl_continue_rcu(fl) {
+               if (net_eq(fl->fl_net, net))
+                       goto out;
+       }
+
 try_again:
-       while (fl && !net_eq(fl->fl_net, net))
-               fl = fl->next;
-
-       while (!fl) {
-               if (++state->bucket <= FL_HASH_MASK) {
-                       fl = fl_ht[state->bucket];
-                       goto try_again;
-               } else
-                       break;
+       if (++state->bucket <= FL_HASH_MASK) {
+               for_each_fl_rcu(state->bucket, fl) {
+                       if (net_eq(fl->fl_net, net))
+                               goto out;
+               }
+               goto try_again;
        }
+       fl = NULL;
+
+out:
        return fl;
 }
 
@@ -690,9 +704,9 @@ static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos)
 }
 
 static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos)
-       __acquires(ip6_fl_lock)
+       __acquires(RCU)
 {
-       read_lock_bh(&ip6_fl_lock);
+       rcu_read_lock_bh();
        return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 }
 
@@ -709,9 +723,9 @@ static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 }
 
 static void ip6fl_seq_stop(struct seq_file *seq, void *v)
-       __releases(ip6_fl_lock)
+       __releases(RCU)
 {
-       read_unlock_bh(&ip6_fl_lock);
+       rcu_read_unlock_bh();
 }
 
 static int ip6fl_seq_show(struct seq_file *seq, void *v)