[PATCH] (5/8) Eliminate brlock from netfilter

From: Stephen Hemminger (shemminger@osdl.org)
Date: Tue Mar 11 2003 - 19:14:54 EST


Replace brlock in netfilter code with RCU.

diff -urN -X dontdiff linux-2.5.64/net/core/netfilter.c linux-2.5-nobrlock/net/core/netfilter.c
--- linux-2.5.64/net/core/netfilter.c 2003-03-11 09:08:01.000000000 -0800
+++ linux-2.5-nobrlock/net/core/netfilter.c 2003-03-11 14:39:38.000000000 -0800
@@ -19,7 +19,6 @@
 #include <linux/interrupt.h>
 #include <linux/if.h>
 #include <linux/netdevice.h>
-#include <linux/brlock.h>
 #include <linux/inetdevice.h>
 #include <net/sock.h>
 #include <net/route.h>
@@ -46,6 +45,7 @@
 
 struct list_head nf_hooks[NPROTO][NF_MAX_HOOKS];
 static LIST_HEAD(nf_sockopts);
+static spinlock_t nf_lock = SPIN_LOCK_UNLOCKED;
 
 /*
  * A queue handler may be registered for each protocol. Each is protected by
@@ -56,28 +56,31 @@
         nf_queue_outfn_t outfn;
         void *data;
 } queue_handler[NPROTO];
+static spinlock_t queue_lock = SPIN_LOCK_UNLOCKED;
 
 int nf_register_hook(struct nf_hook_ops *reg)
 {
         struct list_head *i;
 
- br_write_lock_bh(BR_NETPROTO_LOCK);
+ spin_lock_bh(&nf_lock);
         for (i = nf_hooks[reg->pf][reg->hooknum].next;
              i != &nf_hooks[reg->pf][reg->hooknum];
              i = i->next) {
                 if (reg->priority < ((struct nf_hook_ops *)i)->priority)
                         break;
         }
- list_add(&reg->list, i->prev);
- br_write_unlock_bh(BR_NETPROTO_LOCK);
+ list_add_rcu(&reg->list, i->prev);
+ spin_unlock_bh(&nf_lock);
         return 0;
 }
 
 void nf_unregister_hook(struct nf_hook_ops *reg)
 {
- br_write_lock_bh(BR_NETPROTO_LOCK);
+ spin_lock_bh(&nf_lock);
         list_del(&reg->list);
- br_write_unlock_bh(BR_NETPROTO_LOCK);
+ spin_unlock_bh(&nf_lock);
+
+ synchronize_kernel();
 }
 
 /* Do exclusive ranges overlap? */
@@ -344,7 +347,9 @@
                                int (*okfn)(struct sk_buff *),
                                int hook_thresh)
 {
- for (*i = (*i)->next; *i != head; *i = (*i)->next) {
+
+ for (*i = (*i)->next; *i != head; *i = (*i)->next,
+ ({ read_barrier_depends(); 0;})) {
                 struct nf_hook_ops *elem = (struct nf_hook_ops *)*i;
 
                 if (hook_thresh > elem->priority)
@@ -381,15 +386,16 @@
 {
         int ret;
 
- br_write_lock_bh(BR_NETPROTO_LOCK);
+ spin_lock_bh(&queue_lock);
         if (queue_handler[pf].outfn)
                 ret = -EBUSY;
         else {
- queue_handler[pf].outfn = outfn;
                 queue_handler[pf].data = data;
+ wmb();
+ queue_handler[pf].outfn = outfn;
                 ret = 0;
         }
- br_write_unlock_bh(BR_NETPROTO_LOCK);
+ spin_unlock_bh(&queue_lock);
 
         return ret;
 }
@@ -397,10 +403,14 @@
 /* The caller must flush their queue before this */
 int nf_unregister_queue_handler(int pf)
 {
- br_write_lock_bh(BR_NETPROTO_LOCK);
+ spin_lock_bh(&queue_lock);
         queue_handler[pf].outfn = NULL;
+ wmb();
         queue_handler[pf].data = NULL;
- br_write_unlock_bh(BR_NETPROTO_LOCK);
+ spin_unlock_bh(&queue_lock);
+
+ synchronize_kernel();
+
         return 0;
 }
 
@@ -422,9 +432,10 @@
         struct net_device *physoutdev = NULL;
 #endif
 
+ rcu_read_lock();
         if (!queue_handler[pf].outfn) {
                 kfree_skb(skb);
- return;
+ goto out;
         }
 
         info = kmalloc(sizeof(*info), GFP_ATOMIC);
@@ -433,7 +444,7 @@
                         printk(KERN_ERR "OOM queueing packet %p\n",
                                skb);
                 kfree_skb(skb);
- return;
+ goto out;
         }
 
         *info = (struct nf_info) {
@@ -463,8 +474,9 @@
 #endif
                 kfree(info);
                 kfree_skb(skb);
- return;
         }
+ out:
+ rcu_read_unlock();
 }
 
 int nf_hook_slow(int pf, unsigned int hook, struct sk_buff *skb,
@@ -491,7 +503,7 @@
         }
 
         /* We may already have this, but read-locks nest anyway */
- br_read_lock_bh(BR_NETPROTO_LOCK);
+ rcu_read_lock();
 
 #ifdef CONFIG_NETFILTER_DEBUG
         if (skb->nf_debug & (1 << hook)) {
@@ -520,7 +532,7 @@
                 break;
         }
 
- br_read_unlock_bh(BR_NETPROTO_LOCK);
+ rcu_read_unlock();
         return ret;
 }
 
@@ -530,8 +542,7 @@
         struct list_head *elem = &info->elem->list;
         struct list_head *i;
 
- /* We don't have BR_NETPROTO_LOCK here */
- br_read_lock_bh(BR_NETPROTO_LOCK);
+ rcu_read_lock();
         for (i = nf_hooks[info->pf][info->hook].next; i != elem; i = i->next) {
                 if (i == &nf_hooks[info->pf][info->hook]) {
                         /* The module which sent it to userspace is gone. */
@@ -569,7 +580,7 @@
                 kfree_skb(skb);
                 break;
         }
- br_read_unlock_bh(BR_NETPROTO_LOCK);
+ rcu_read_unlock();
 
         /* Release those devices we held, or Alexey will kill me. */
         if (info->indev) dev_put(info->indev);

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Sat Mar 15 2003 - 22:00:28 EST