Commit 923f4902 authored by Patrick McHardy's avatar Patrick McHardy Committed by David S. Miller

[NETFILTER]: nf_conntrack: properly use RCU API for nf_ct_protos/nf_ct_l3protos arrays

Replace preempt_{enable,disable} based RCU by proper use of the
RCU API and add missing rcu_read_lock/rcu_read_unlock calls in
all paths not obviously only used within packet process context
(nfnetlink_conntrack).
Signed-off-by: default avatarPatrick McHardy <kaber@trash.net>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 642d628b
...@@ -106,7 +106,7 @@ __nf_ct_l3proto_find(u_int16_t l3proto) ...@@ -106,7 +106,7 @@ __nf_ct_l3proto_find(u_int16_t l3proto)
{ {
if (unlikely(l3proto >= AF_MAX)) if (unlikely(l3proto >= AF_MAX))
return &nf_conntrack_l3proto_generic; return &nf_conntrack_l3proto_generic;
return nf_ct_l3protos[l3proto]; return rcu_dereference(nf_ct_l3protos[l3proto]);
} }
#endif /*_NF_CONNTRACK_L3PROTO_H*/ #endif /*_NF_CONNTRACK_L3PROTO_H*/
...@@ -170,7 +170,9 @@ icmp_error_message(struct sk_buff *skb, ...@@ -170,7 +170,9 @@ icmp_error_message(struct sk_buff *skb,
return -NF_ACCEPT; return -NF_ACCEPT;
} }
/* rcu_read_lock()ed by nf_hook_slow */
innerproto = __nf_ct_l4proto_find(PF_INET, inside->ip.protocol); innerproto = __nf_ct_l4proto_find(PF_INET, inside->ip.protocol);
dataoff = skb->nh.iph->ihl*4 + sizeof(inside->icmp); dataoff = skb->nh.iph->ihl*4 + sizeof(inside->icmp);
/* Are they talking about one of our connections? */ /* Are they talking about one of our connections? */
if (!nf_ct_get_tuple(skb, dataoff, dataoff + inside->ip.ihl*4, PF_INET, if (!nf_ct_get_tuple(skb, dataoff, dataoff + inside->ip.ihl*4, PF_INET,
......
...@@ -429,6 +429,7 @@ int nf_nat_icmp_reply_translation(struct nf_conn *ct, ...@@ -429,6 +429,7 @@ int nf_nat_icmp_reply_translation(struct nf_conn *ct,
struct icmphdr icmp; struct icmphdr icmp;
struct iphdr ip; struct iphdr ip;
} *inside; } *inside;
struct nf_conntrack_l4proto *l4proto;
struct nf_conntrack_tuple inner, target; struct nf_conntrack_tuple inner, target;
int hdrlen = (*pskb)->nh.iph->ihl * 4; int hdrlen = (*pskb)->nh.iph->ihl * 4;
enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
...@@ -464,16 +465,16 @@ int nf_nat_icmp_reply_translation(struct nf_conn *ct, ...@@ -464,16 +465,16 @@ int nf_nat_icmp_reply_translation(struct nf_conn *ct,
DEBUGP("icmp_reply_translation: translating error %p manp %u dir %s\n", DEBUGP("icmp_reply_translation: translating error %p manp %u dir %s\n",
*pskb, manip, dir == IP_CT_DIR_ORIGINAL ? "ORIG" : "REPLY"); *pskb, manip, dir == IP_CT_DIR_ORIGINAL ? "ORIG" : "REPLY");
/* rcu_read_lock()ed by nf_hook_slow */
l4proto = __nf_ct_l4proto_find(PF_INET, inside->ip.protocol);
if (!nf_ct_get_tuple(*pskb, if (!nf_ct_get_tuple(*pskb,
(*pskb)->nh.iph->ihl*4 + sizeof(struct icmphdr), (*pskb)->nh.iph->ihl*4 + sizeof(struct icmphdr),
(*pskb)->nh.iph->ihl*4 + (*pskb)->nh.iph->ihl*4 +
sizeof(struct icmphdr) + inside->ip.ihl*4, sizeof(struct icmphdr) + inside->ip.ihl*4,
(u_int16_t)AF_INET, (u_int16_t)AF_INET,
inside->ip.protocol, inside->ip.protocol,
&inner, &inner, l3proto, l4proto))
l3proto,
__nf_ct_l4proto_find((u_int16_t)PF_INET,
inside->ip.protocol)))
return 0; return 0;
/* Change inner back to look like incoming packet. We do the /* Change inner back to look like incoming packet. We do the
......
...@@ -182,6 +182,7 @@ icmpv6_error_message(struct sk_buff *skb, ...@@ -182,6 +182,7 @@ icmpv6_error_message(struct sk_buff *skb,
return -NF_ACCEPT; return -NF_ACCEPT;
} }
/* rcu_read_lock()ed by nf_hook_slow */
inproto = __nf_ct_l4proto_find(PF_INET6, inprotonum); inproto = __nf_ct_l4proto_find(PF_INET6, inprotonum);
/* Are they talking about one of our connections? */ /* Are they talking about one of our connections? */
......
...@@ -332,13 +332,16 @@ destroy_conntrack(struct nf_conntrack *nfct) ...@@ -332,13 +332,16 @@ destroy_conntrack(struct nf_conntrack *nfct)
/* To make sure we don't get any weird locking issues here: /* To make sure we don't get any weird locking issues here:
* destroy_conntrack() MUST NOT be called with a write lock * destroy_conntrack() MUST NOT be called with a write lock
* to nf_conntrack_lock!!! -HW */ * to nf_conntrack_lock!!! -HW */
rcu_read_lock();
l3proto = __nf_ct_l3proto_find(ct->tuplehash[IP_CT_DIR_REPLY].tuple.src.l3num); l3proto = __nf_ct_l3proto_find(ct->tuplehash[IP_CT_DIR_REPLY].tuple.src.l3num);
if (l3proto && l3proto->destroy) if (l3proto && l3proto->destroy)
l3proto->destroy(ct); l3proto->destroy(ct);
l4proto = __nf_ct_l4proto_find(ct->tuplehash[IP_CT_DIR_REPLY].tuple.src.l3num, ct->tuplehash[IP_CT_DIR_REPLY].tuple.dst.protonum); l4proto = __nf_ct_l4proto_find(ct->tuplehash[IP_CT_DIR_REPLY].tuple.src.l3num,
ct->tuplehash[IP_CT_DIR_REPLY].tuple.dst.protonum);
if (l4proto && l4proto->destroy) if (l4proto && l4proto->destroy)
l4proto->destroy(ct); l4proto->destroy(ct);
rcu_read_unlock();
if (nf_conntrack_destroyed) if (nf_conntrack_destroyed)
nf_conntrack_destroyed(ct); nf_conntrack_destroyed(ct);
...@@ -647,9 +650,14 @@ struct nf_conn *nf_conntrack_alloc(const struct nf_conntrack_tuple *orig, ...@@ -647,9 +650,14 @@ struct nf_conn *nf_conntrack_alloc(const struct nf_conntrack_tuple *orig,
const struct nf_conntrack_tuple *repl) const struct nf_conntrack_tuple *repl)
{ {
struct nf_conntrack_l3proto *l3proto; struct nf_conntrack_l3proto *l3proto;
struct nf_conn *ct;
rcu_read_lock();
l3proto = __nf_ct_l3proto_find(orig->src.l3num); l3proto = __nf_ct_l3proto_find(orig->src.l3num);
return __nf_conntrack_alloc(orig, repl, l3proto, 0); ct = __nf_conntrack_alloc(orig, repl, l3proto, 0);
rcu_read_unlock();
return ct;
} }
EXPORT_SYMBOL_GPL(nf_conntrack_alloc); EXPORT_SYMBOL_GPL(nf_conntrack_alloc);
...@@ -817,7 +825,9 @@ nf_conntrack_in(int pf, unsigned int hooknum, struct sk_buff **pskb) ...@@ -817,7 +825,9 @@ nf_conntrack_in(int pf, unsigned int hooknum, struct sk_buff **pskb)
return NF_ACCEPT; return NF_ACCEPT;
} }
/* rcu_read_lock()ed by nf_hook_slow */
l3proto = __nf_ct_l3proto_find((u_int16_t)pf); l3proto = __nf_ct_l3proto_find((u_int16_t)pf);
if ((ret = l3proto->prepare(pskb, hooknum, &dataoff, &protonum)) <= 0) { if ((ret = l3proto->prepare(pskb, hooknum, &dataoff, &protonum)) <= 0) {
DEBUGP("not prepared to track yet or error occured\n"); DEBUGP("not prepared to track yet or error occured\n");
return -ret; return -ret;
...@@ -872,10 +882,15 @@ EXPORT_SYMBOL_GPL(nf_conntrack_in); ...@@ -872,10 +882,15 @@ EXPORT_SYMBOL_GPL(nf_conntrack_in);
int nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, int nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse,
const struct nf_conntrack_tuple *orig) const struct nf_conntrack_tuple *orig)
{ {
return nf_ct_invert_tuple(inverse, orig, int ret;
rcu_read_lock();
ret = nf_ct_invert_tuple(inverse, orig,
__nf_ct_l3proto_find(orig->src.l3num), __nf_ct_l3proto_find(orig->src.l3num),
__nf_ct_l4proto_find(orig->src.l3num, __nf_ct_l4proto_find(orig->src.l3num,
orig->dst.protonum)); orig->dst.protonum));
rcu_read_unlock();
return ret;
} }
EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr);
......
...@@ -66,7 +66,7 @@ __nf_ct_l4proto_find(u_int16_t l3proto, u_int8_t l4proto) ...@@ -66,7 +66,7 @@ __nf_ct_l4proto_find(u_int16_t l3proto, u_int8_t l4proto)
if (unlikely(l3proto >= AF_MAX || nf_ct_protos[l3proto] == NULL)) if (unlikely(l3proto >= AF_MAX || nf_ct_protos[l3proto] == NULL))
return &nf_conntrack_l4proto_generic; return &nf_conntrack_l4proto_generic;
return nf_ct_protos[l3proto][l4proto]; return rcu_dereference(nf_ct_protos[l3proto][l4proto]);
} }
EXPORT_SYMBOL_GPL(__nf_ct_l4proto_find); EXPORT_SYMBOL_GPL(__nf_ct_l4proto_find);
...@@ -77,11 +77,11 @@ nf_ct_l4proto_find_get(u_int16_t l3proto, u_int8_t l4proto) ...@@ -77,11 +77,11 @@ nf_ct_l4proto_find_get(u_int16_t l3proto, u_int8_t l4proto)
{ {
struct nf_conntrack_l4proto *p; struct nf_conntrack_l4proto *p;
preempt_disable(); rcu_read_lock();
p = __nf_ct_l4proto_find(l3proto, l4proto); p = __nf_ct_l4proto_find(l3proto, l4proto);
if (!try_module_get(p->me)) if (!try_module_get(p->me))
p = &nf_conntrack_l4proto_generic; p = &nf_conntrack_l4proto_generic;
preempt_enable(); rcu_read_unlock();
return p; return p;
} }
...@@ -98,11 +98,11 @@ nf_ct_l3proto_find_get(u_int16_t l3proto) ...@@ -98,11 +98,11 @@ nf_ct_l3proto_find_get(u_int16_t l3proto)
{ {
struct nf_conntrack_l3proto *p; struct nf_conntrack_l3proto *p;
preempt_disable(); rcu_read_lock();
p = __nf_ct_l3proto_find(l3proto); p = __nf_ct_l3proto_find(l3proto);
if (!try_module_get(p->me)) if (!try_module_get(p->me))
p = &nf_conntrack_l3proto_generic; p = &nf_conntrack_l3proto_generic;
preempt_enable(); rcu_read_unlock();
return p; return p;
} }
...@@ -137,10 +137,8 @@ void nf_ct_l3proto_module_put(unsigned short l3proto) ...@@ -137,10 +137,8 @@ void nf_ct_l3proto_module_put(unsigned short l3proto)
{ {
struct nf_conntrack_l3proto *p; struct nf_conntrack_l3proto *p;
preempt_disable(); /* rcu_read_lock not necessary since the caller holds a reference */
p = __nf_ct_l3proto_find(l3proto); p = __nf_ct_l3proto_find(l3proto);
preempt_enable();
module_put(p->me); module_put(p->me);
} }
EXPORT_SYMBOL_GPL(nf_ct_l3proto_module_put); EXPORT_SYMBOL_GPL(nf_ct_l3proto_module_put);
...@@ -202,7 +200,7 @@ int nf_conntrack_l3proto_register(struct nf_conntrack_l3proto *proto) ...@@ -202,7 +200,7 @@ int nf_conntrack_l3proto_register(struct nf_conntrack_l3proto *proto)
ret = -EBUSY; ret = -EBUSY;
goto out_unlock; goto out_unlock;
} }
nf_ct_l3protos[proto->l3proto] = proto; rcu_assign_pointer(nf_ct_l3protos[proto->l3proto], proto);
write_unlock_bh(&nf_conntrack_lock); write_unlock_bh(&nf_conntrack_lock);
ret = nf_ct_l3proto_register_sysctl(proto); ret = nf_ct_l3proto_register_sysctl(proto);
...@@ -233,14 +231,13 @@ int nf_conntrack_l3proto_unregister(struct nf_conntrack_l3proto *proto) ...@@ -233,14 +231,13 @@ int nf_conntrack_l3proto_unregister(struct nf_conntrack_l3proto *proto)
goto out; goto out;
} }
nf_ct_l3protos[proto->l3proto] = &nf_conntrack_l3proto_generic; rcu_assign_pointer(nf_ct_l3protos[proto->l3proto],
&nf_conntrack_l3proto_generic);
write_unlock_bh(&nf_conntrack_lock); write_unlock_bh(&nf_conntrack_lock);
synchronize_rcu();
nf_ct_l3proto_unregister_sysctl(proto); nf_ct_l3proto_unregister_sysctl(proto);
/* Somebody could be still looking at the proto in bh. */
synchronize_net();
/* Remove all contrack entries for this protocol */ /* Remove all contrack entries for this protocol */
nf_ct_iterate_cleanup(kill_l3proto, proto); nf_ct_iterate_cleanup(kill_l3proto, proto);
...@@ -356,7 +353,7 @@ int nf_conntrack_l4proto_register(struct nf_conntrack_l4proto *l4proto) ...@@ -356,7 +353,7 @@ int nf_conntrack_l4proto_register(struct nf_conntrack_l4proto *l4proto)
goto retry; goto retry;
} }
nf_ct_protos[l4proto->l3proto][l4proto->l4proto] = l4proto; rcu_assign_pointer(nf_ct_protos[l4proto->l3proto][l4proto->l4proto], l4proto);
write_unlock_bh(&nf_conntrack_lock); write_unlock_bh(&nf_conntrack_lock);
ret = nf_ct_l4proto_register_sysctl(l4proto); ret = nf_ct_l4proto_register_sysctl(l4proto);
...@@ -392,15 +389,13 @@ int nf_conntrack_l4proto_unregister(struct nf_conntrack_l4proto *l4proto) ...@@ -392,15 +389,13 @@ int nf_conntrack_l4proto_unregister(struct nf_conntrack_l4proto *l4proto)
ret = -EBUSY; ret = -EBUSY;
goto out; goto out;
} }
nf_ct_protos[l4proto->l3proto][l4proto->l4proto] rcu_assign_pointer(nf_ct_protos[l4proto->l3proto][l4proto->l4proto],
= &nf_conntrack_l4proto_generic; &nf_conntrack_l4proto_generic);
write_unlock_bh(&nf_conntrack_lock); write_unlock_bh(&nf_conntrack_lock);
synchronize_rcu();
nf_ct_l4proto_unregister_sysctl(l4proto); nf_ct_l4proto_unregister_sysctl(l4proto);
/* Somebody could be still looking at the proto in bh. */
synchronize_net();
/* Remove all contrack entries for this protocol */ /* Remove all contrack entries for this protocol */
nf_ct_iterate_cleanup(kill_l4proto, l4proto); nf_ct_iterate_cleanup(kill_l4proto, l4proto);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment