: Netfilter fixes for 2.6.17
Subsystems (3)
cover : Netfilter fixes for 2.6.17
Hi Dave,
following are a couple of netfilter fixes for 2.6.17. The whitespace problem is fixed now, apparently sed doesn't understand \t. I guess I'm lucky that it didn't start stripping trailing t's :)
include/linux/netfilter/x_tables.h | 4 ++++ net/ipv4/netfilter/Kconfig | 2 +- net/ipv6/netfilter/ip6_tables.c | 13 ------------- net/netfilter/nf_conntrack_core.c | 15 ++++----------- net/netfilter/nf_conntrack_l3proto_generic.c | 1 - net/netfilter/x_tables.c | 2 +- net/sched/act_ipt.c | 5 +++++ 7 files changed, 15 insertions(+), 27 deletions(-)
Dmitry Mishin:
[NETFILTER]: x_tables: move table->lock initialization
Patrick McHardy:
[NETFILTER]: Fix compat_xt_counters alignment for non-x86 [NETFILTER]: ip6_tables: remove broken comefrom debugging [NETFILTER]: ipt action: use xt_check_target for basic verification
Thomas Voegtle:
[NETFILTER]: ULOG target is not obsolete
Yasuyuki Kozakai:
[NETFILTER]: nf_conntrack: Fix module refcount dropping too far [NETFILTER]: nf_conntrack: kill unused callback init_conntrack
PATCH 1/7 : nf_conntrack: Fix module refcount dropping too far
[NETFILTER]: nf_conntrack: Fix module refcount dropping too far
If nf_ct_l3proto_find_get() fails to get the refcount of nf_ct_l3proto_generic, nf_ct_l3proto_put() will drop the refcount too far.
This gets rid of '.me = THIS_MODULE' of nf_ct_l3proto_generic so that nf_ct_l3proto_find_get() doesn't try to get refcount of it. It's OK because its symbol is usable until nf_conntrack.ko is unloaded.
This also kills unnecessary NULL pointer check as well. __nf_ct_proto_find() allways returns non-NULL pointer.
---
commit bc79a39fcc0cb09ed73ba0052cb87c2893f4f65f
tree a2702573ee2acdc6fb865921a94042ef8d14f0f5
parent 5a7b46b369419493bab4de67b1526e9f76b22a7f
author Yasuyuki Kozakai <[email protected]> Fri, 21 Apr 2006 01:19:00 +0200 committer Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:19:00 +0200
net/netfilter/nf_conntrack_core.c | 12 ++++-------- net/netfilter/nf_conntrack_l3proto_generic.c | 1 - 2 files changed, 4 insertions(+), 9 deletions(-)
net/netfilter/nf_conntrack_core.c +4 -8
@@ -208,10 +208,8 @@ nf_ct_proto_find_get(u_int16_t l3proto,preempt_disable (); p =__nf_ct_proto_find (l3proto, protocol); -if (p) { -if (!try_module_get (p->me)) - p = &nf_conntrack_generic_protocol; - } +if (!try_module_get (p->me)) + p = &nf_conntrack_generic_protocol;preempt_enable ();return p;
@@ -229,10 +227,8 @@ nf_ct_l3proto_find_get(u_int16_t l3protopreempt_disable (); p =__nf_ct_l3proto_find (l3proto); -if (p) { -if (!try_module_get (p->me)) - p = &nf_conntrack_generic_l3proto; - } +if (!try_module_get (p->me)) + p = &nf_conntrack_generic_l3proto;preempt_enable ();return p;
net/netfilter/nf_conntrack_l3proto_generic.c +0 -1
@@ -94,5 +94,4 @@ struct nf_conntrack_l3proto nf_conntrack .print_conntrack = generic_print_conntrack, .prepare = generic_prepare, .get_features = generic_get_features, - .me = THIS_MODULE, };
PATCH 2/7 : ULOG target is not obsolete
In reply to Patrick McHardy[NETFILTER]: ULOG target is not obsolete
The backend part is obsoleted, but the target itself is still needed.
---
commit 40c6dcb49ac3f091e30fa105c347cc25776e8b72
tree 2e826e3e3c47a7c5ba0fc5d070a05b090bc30a02
parent bc79a39fcc0cb09ed73ba0052cb87c2893f4f65f
author Thomas Voegtle <[email protected]> Fri, 21 Apr 2006 01:19:15 +0200 committer Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:19:15 +0200
net/ipv4/netfilter/Kconfig | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
net/ipv4/netfilter/Kconfig +1 -1
@@ -345,7 +345,7 @@ config IP_NF_TARGET_LOG To compile it as a module, choose M here. If unsure, say N. config IP_NF_TARGET_ULOG - tristate "ULOG target support (OBSOLETE)" + tristate "ULOG target support" depends on IP_NF_IPTABLES ---help---
PATCH 3/7 : Fix compat_xt_counters alignment for non-x86
In reply to Patrick McHardy[NETFILTER]: Fix compat_xt_counters alignment for non-x86
Some (?) non-x86 architectures require 8byte alignment for u_int64_t even when compiled for 32bit, using u_int32_t in compat_xt_counters breaks on these architectures, use u_int64_t for everything but x86.
Reported by Andreas Schwab <[email protected]>.
---
commit 030e806c1db0d3a278585b83475d1a6f9a874788
tree ef069a86e4ac0a1572ea51f7090b0e57df00052f
parent 40c6dcb49ac3f091e30fa105c347cc25776e8b72
author Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:19:29 +0200 committer Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:19:29 +0200
include/linux/netfilter/x_tables.h | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)
include/linux/netfilter/x_tables.h +4 -0
@@ -361,7 +361,11 @@ struct compat_xt_entry_targetstruct compat_xt_counters { +#if defined(CONFIG_X86_64 ) || defined(CONFIG_IA64 )u_int32_t cnt[4 ]; +#else +u_int64_t cnt[2 ]; +#endif };struct compat_xt_counters_info
PATCH 4/7 : nf_conntrack: kill unused callback init_conntrack
In reply to Patrick McHardy[NETFILTER]: nf_conntrack: kill unused callback init_conntrack
---
commit 81d5f60a185910989bc5778e3ea7ee03bda7f36d
tree 03c41d336c6528c11e024a1e42e100281cc0bd3e
parent 030e806c1db0d3a278585b83475d1a6f9a874788
author Yasuyuki Kozakai <[email protected]> Fri, 21 Apr 2006 01:21:30 +0200 committer Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:21:30 +0200
net/netfilter/nf_conntrack_core.c | 3 --- 1 files changed, 0 insertions(+), 3 deletions(-)
net/netfilter/nf_conntrack_core.c +0 -3
@@ -178,9 +178,6 @@ static struct {/* allocated slab cache + modules which uses this slab cache */ int use; -/* Initialization */ -int (*init_conntrack)(struct nf_conn *,u_int32_t ); - } nf_ct_cache[NF_CT_F_NUM ];/* protect members of nf_ct_cache except of "use" */
PATCH 5/7 : ip6_tables: remove broken comefrom debugging
In reply to Patrick McHardy[NETFILTER]: ip6_tables: remove broken comefrom debugging
The introduction of x_tables broke comefrom debugging, remove it from ip6_tables as well (ip_tables already got removed).
---
commit 1d7a13058df694ded6411b12931f3dd42f8193a4
tree c1f64b340b6adc83cec030d6a5d365521cd14003
parent 81d5f60a185910989bc5778e3ea7ee03bda7f36d
author Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:25:26 +0200 committer Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:25:26 +0200
net/ipv6/netfilter/ip6_tables.c | 13 ------------- 1 files changed, 0 insertions(+), 13 deletions(-)
net/ipv6/netfilter/ip6_tables.c +0 -13
@@ -288,19 +288,6 @@ ip6t_do_table(struct sk_buff **pskb, table_base = (void *)private->entries[smp_processor_id ()]; e =get_entry (table_base, private->hook_entry[hook]); -#ifdef CONFIG_NETFILTER_DEBUG -/* Check noone else using our table */ -if (((struct ip6t_entry *)table_base)->comefrom !=0xdead57ac - && ((struct ip6t_entry *)table_base)->comefrom !=0xeeeeeeec ) { -printk ("ASSERT: CPU #%u, %s comefrom(%p) = %X\n" , -smp_processor_id (), - table->name, - &((struct ip6t_entry *)table_base)->comefrom, - ((struct ip6t_entry *)table_base)->comefrom); - } - ((struct ip6t_entry *)table_base)->comefrom =0x57acc001 ; -#endif -/* For return from builtin chain */ back =get_entry (table_base, private->underflow[hook]);
PATCH 6/7 : x_tables: move table->lock initialization
In reply to Patrick McHardy[NETFILTER]: x_tables: move table->lock initialization
xt_table->lock should be initialized before xt_replace_table() call, which uses it. This patch removes strict requirement that table should define lock before registering.
---
commit 81b536eef7989c16a2d59ced319aafaaf9e3ed03
tree f412294f1c7862db8e0b7d457fdfe85036c32c94
parent 1d7a13058df694ded6411b12931f3dd42f8193a4
author Dmitry Mishin <[email protected]> Fri, 21 Apr 2006 01:29:07 +0200 committer Patrick McHardy <[email protected]> Fri, 21 Apr 2006 01:29:07 +0200
net/netfilter/x_tables.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
net/netfilter/x_tables.c +1 -1
@@ -529,6 +529,7 @@ int xt_register_table(struct xt_table *t/* Simplifies replace_table code. */ table->private = bootstrap; +rwlock_init (&table->lock);if (!xt_replace_table (table,0 , newinfo, &ret)) goto unlock;
@@ -538,7 +539,6 @@ int xt_register_table(struct xt_table *t/* save number of initial entries */ private->initial_entries = private->number; -rwlock_init (&table->lock);list_prepend (&xt[table->af].tables, table); ret =0 ;
PATCH 7/7 : ipt action: use xt_check_target for basic verification
In reply to Patrick McHardy[NETFILTER]: ipt action: use xt_check_target for basic verification
The targets don't do the basic verification themselves anymore so the ipt action needs to take care of it.
---
commit 42a2de55185ffa55f70e289e708d492eae2685c5
tree 49d5316c896a1801f2f1abedb73b8e052522c6c9
parent 81b536eef7989c16a2d59ced319aafaaf9e3ed03
author Patrick McHardy <[email protected]> Fri, 21 Apr 2006 03:03:53 +0200 committer Patrick McHardy <[email protected]> Fri, 21 Apr 2006 03:03:53 +0200
net/sched/act_ipt.c | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-)
net/sched/act_ipt.c +5 -0
@@ -69,6 +69,11 @@ ipt_init_target(struct ipt_entry_targetDPRINTK ("ipt_init_target: found %s\n" , target->name); t->u.kernel.target = target; + ret =xt_check_target (target,AF_INET , t->u.target_size -sizeof (*t), + table, hook,0 ,0 ); +if (ret) +return ret; +if (t->u.kernel.target->checkentry && !t->u.kernel.target->checkentry(table, NULL, t->u.kernel.target, t->data,
David S. Miller
maintainerFrom: Patrick McHardy <[email protected]>
Date: Fri, 21 Apr 2006 03:06:48 +0200 (MEST)
Quoted 3 lines
following are a couple of netfilter fixes for 2.6.17. The whitespace problem is fixed now, apparently sed doesn't understand \t. I guess I'm lucky that it didn't start stripping trailing t's :)
:-) All applied, thanks Patrick.