Skip to contentSkip to conversation

netfilter: unregister nf hooks, matches and targets in the reverse order

PATCH In review Changli Gao 3 participants 5 messages Sep 2, 2010 – Oct 4, 2010

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 netfilter: unregister nf hooks, matches and targets in the reverse order

Since we register nf hooks, matches and targets in order, we'd better unregister them in the reverse order.

---
 net/netfilter/core.c     |    6 ++----
 net/netfilter/x_tables.c |   12 ++++--------
 2 files changed, 6 insertions(+), 12 deletions(-)
net/netfilter/core.c +2 -4
index 78b505d..8f014f2 100644
@@ -105,10 +105,8 @@ EXPORT_SYMBOL(nf_register_hooks);
 
 void nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n)
 {
-	unsigned int i;
-
-	for (i = 0; i < n; i++)
-		nf_unregister_hook(&reg[i]);
+	while (n-- > 0)
+		nf_unregister_hook(&reg[n]);
 }
 EXPORT_SYMBOL(nf_unregister_hooks);
 
net/netfilter/x_tables.c +4 -8
index e34622f..8046350 100644
@@ -116,10 +116,8 @@ EXPORT_SYMBOL(xt_register_targets);
 void
 xt_unregister_targets(struct xt_target *target, unsigned int n)
 {
-	unsigned int i;
-
-	for (i = 0; i < n; i++)
-		xt_unregister_target(&target[i]);
+	while (n-- > 0)
+		xt_unregister_target(&target[n]);
 }
 EXPORT_SYMBOL(xt_unregister_targets);
 
@@ -174,10 +172,8 @@ EXPORT_SYMBOL(xt_register_matches);
 void
 xt_unregister_matches(struct xt_match *match, unsigned int n)
 {
-	unsigned int i;
-
-	for (i = 0; i < n; i++)
-		xt_unregister_match(&match[i]);
+	while (n-- > 0)
+		xt_unregister_match(&match[n]);
 }
 EXPORT_SYMBOL(xt_unregister_matches);
 
#1 lore raw

Am 02.09.2010 16:15, schrieb Changli Gao:

Quoted 2 lines

Since we register nf hooks, matches and targets in order, we'd better unregister them in the reverse order.

Why? Is there a specific bug you've noticed?

#2 lore raw
Quoted 6 lines · Patrick McHardy

Am 02.09.2010 16:15, schrieb Changli Gao:

Quoted 2 lines

Since we register nf hooks, matches and targets in order, we'd better unregister them in the reverse order.

Why? Is there a specific bug you've noticed?

No, there isn't any bug. I just think unregistering them in the reverse order is more resonable, like the rollback when failing. And the code patched generates less object:

The original:

00000000000009c9 <xt_unregister_matches>:

     9c9:       55                      push   %rbp
     9ca:       48 89 e5                mov    %rsp,%rbp
     9cd:       41 55                   push   %r13
     9cf:       41 89 f5                mov    %esi,%r13d
     9d2:       41 54                   push   %r12
     9d4:       49 89 fc                mov    %rdi,%r12
     9d7:       53                      push   %rbx
     9d8:       31 db                   xor    %ebx,%ebx
     9da:       48 83 ec 08             sub    $0x8,%rsp
     9de:       eb 0e                   jmp    9ee <xt_unregister_matches+0x25>
     9e0:       4c 89 e7                mov    %r12,%rdi
     9e3:       ff c3                   inc    %ebx
     9e5:       49 83 c4 78             add    $0x78,%r12
     9e9:       e8 00 00 00 00          callq  9ee <xt_unregister_matches+0x25>
     9ee:       44 39 eb                cmp    %r13d,%ebx
     9f1:       72 ed                   jb     9e0 <xt_unregister_matches+0x17>
     9f3:       41 59                   pop    %r9
     9f5:       5b                      pop    %rbx
     9f6:       41 5c                   pop    %r12
     9f8:       41 5d                   pop    %r13
     9fa:       c9                      leaveq
     9fb:       c3                      retq

The patched:

00000000000009c9 <xt_unregister_matches>:

     9c9:       55                      push   %rbp
     9ca:       48 89 e5                mov    %rsp,%rbp
     9cd:       41 54                   push   %r12
     9cf:       41 89 f4                mov    %esi,%r12d
     9d2:       53                      push   %rbx
     9d3:       48 89 fb                mov    %rdi,%rbx
     9d6:       eb 13                   jmp    9eb <xt_unregister_matches+0x22>
     9d8:       41 ff cc                dec    %r12d
     9db:       44 89 e7                mov    %r12d,%edi
     9de:       48 6b ff 78             imul   $0x78,%rdi,%rdi
     9e2:       48 8d 3c 3b             lea    (%rbx,%rdi,1),%rdi
     9e6:       e8 00 00 00 00          callq  9eb <xt_unregister_matches+0x22>
     9eb:       45 85 e4                test   %r12d,%r12d
     9ee:       75 e8                   jne    9d8 <xt_unregister_matches+0xf>
     9f0:       5b                      pop    %rbx
     9f1:       41 5c                   pop    %r12
     9f3:       c9                      leaveq
     9f4:       c3                      retq
Signature · 2 lines
Regards,
Changli Gao([email protected])
#3 lore raw
Quoted 11 lines · Changli Gao
Quoted 6 lines · Patrick McHardy

Am 02.09.2010 16:15, schrieb Changli Gao:

Quoted 2 lines

Since we register nf hooks, matches and targets in order, we'd better unregister them in the reverse order.

Why? Is there a specific bug you've noticed?

No, there isn't any bug. I just think unregistering them in the reverse order is more resonable, like the rollback when failing. And the code patched generates less object:

I support this.

#4 lore raw

Am 02.10.2010 01:57, schrieb Jan Engelhardt:

Quoted 7 lines · Changli Gao
Quoted 6 lines · Patrick McHardy
Quoted 4 lines

Am 02.09.2010 16:15, schrieb Changli Gao:

Quoted 2 lines

Since we register nf hooks, matches and targets in order, we'd better unregister them in the reverse order.

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

Why? Is there a specific bug you've noticed?

No, there isn't any bug. I just think unregistering them in the reverse order is more resonable, like the rollback when failing. And the code patched generates less object:

I support this.

Yeah, it seems reasonable, although I don't like changing things without a concrete reason. Anyways, applied, thanks everyone.