Re: FIX: connlimit NULL pointer kernel panic (was: connlimit patch crashes 2.6.11 kernel)

Re: FIX: connlimit NULL pointer kernel panic (was: connlimit patch crashes 2.6.11 kernel)

[Pablo Neira]

[-- Attachment #1: Type: text/plain, Size: 364 bytes --]

Forte Systems - Iosif Peterfi wrote:
> Here is a diff -urN patch for what has been discussed except the conntrack
> event-api. Works fine on 2.6.1-gentoo-r8, patch-o-matic-ng-20050516. 

Hm, didn't I also tell you that you have to sed 
's/spin_unlock/spin_unlock_bh'?

well, it doesn't matter, attached the correct patch that applies cleanly 
to pom-ng.

--
Pablo

[-- Attachment #2: x --]
[-- Type: text/plain, Size: 1583 bytes --]

Index: linux-2.6.11/net/ipv4/netfilter/ipt_connlimit.c
===================================================================
--- linux-2.6.11/net/ipv4/netfilter/ipt_connlimit.c	(revision 3922)
+++ linux-2.6.11/net/ipv4/netfilter/ipt_connlimit.c	(working copy)
@@ -55,7 +55,7 @@
 	struct ipt_connlimit_conn *conn;
 	struct list_head *hash,*lh;
 
-	spin_lock(&data->lock);
+	spin_lock_bh(&data->lock);
 	tuple = ct->tuplehash[0].tuple;
 	hash = &data->iphash[ipt_iphash(addr & mask)];
 
@@ -64,9 +64,10 @@
 		struct ip_conntrack *found_ct = NULL;
 		conn = list_entry(lh,struct ipt_connlimit_conn,list);
 		found = ip_conntrack_find_get(&conn->tuple,ct);
-		if (0 == memcmp(&conn->tuple,&tuple,sizeof(tuple)) &&
-		    found != NULL && (found_ct = tuplehash_to_ctrack(found)) != NULL &&
-		    found_ct->proto.tcp.state != TCP_CONNTRACK_TIME_WAIT) {
+		 if (found != NULL 
+		     && (found_ct = tuplehash_to_ctrack(found)) != NULL
+		     && 0 == memcmp(&conn->tuple,&tuple,sizeof(tuple)) 
+		     && found_ct->proto.tcp.state != TCP_CONNTRACK_TIME_WAIT) {
 			/* Just to be sure we have it only once in the list.
 			   We should'nt see tuples twice unless someone hooks this
 			   into a table without "-p tcp --syn" */
@@ -111,7 +112,7 @@
 #endif
 		conn = kmalloc(sizeof(*conn),GFP_ATOMIC);
 		if (NULL == conn) {
-			spin_unlock(&data->lock);
+			spin_unlock_bh(&data->lock);
 			return -1;
 		}
 		memset(conn,0,sizeof(*conn));
@@ -120,7 +121,7 @@
 		list_add(&conn->list,hash);
 		matches++;
 	}
-	spin_unlock(&data->lock);
+	spin_unlock_bh(&data->lock);
 	return matches;
 }
 

Re: FIX: connlimit NULL pointer kernel panic

[Patrick McHardy]

Pablo Neira wrote:
> Forte Systems - Iosif Peterfi wrote:
> 
>> Here is a diff -urN patch for what has been discussed except the
>> conntrack
>> event-api. Works fine on 2.6.1-gentoo-r8, patch-o-matic-ng-20050516. 
> 
> 
> Hm, didn't I also tell you that you have to sed
> 's/spin_unlock/spin_unlock_bh'?
> 
> well, it doesn't matter, attached the correct patch that applies cleanly
> to pom-ng.

Thanks Pablo, I've applied the patch. Can you send patches for the 2.4
and the old 2.6 version too?

Regards
Patrick

Re: FIX: connlimit NULL pointer kernel panic

[Pablo Neira]

[-- Attachment #1: Type: text/plain, Size: 558 bytes --]

Patrick McHardy wrote:
> Pablo Neira wrote:
> 
>>Forte Systems - Iosif Peterfi wrote:
>>
>>
>>>Here is a diff -urN patch for what has been discussed except the
>>>conntrack
>>>event-api. Works fine on 2.6.1-gentoo-r8, patch-o-matic-ng-20050516. 
>>
>>
>>Hm, didn't I also tell you that you have to sed
>>'s/spin_unlock/spin_unlock_bh'?
>>
>>well, it doesn't matter, attached the correct patch that applies cleanly
>>to pom-ng.
> 
> 
> Thanks Pablo, I've applied the patch. Can you send patches for the 2.4
> and the old 2.6 version too?

Attached.

--
Pablo

[-- Attachment #2: x --]
[-- Type: text/plain, Size: 2614 bytes --]

Index: linux/net/ipv4/netfilter/ipt_connlimit.c
===================================================================
--- linux/net/ipv4/netfilter/ipt_connlimit.c	(revision 3889)
+++ linux/net/ipv4/netfilter/ipt_connlimit.c	(working copy)
@@ -55,7 +55,7 @@
 	struct ipt_connlimit_conn *conn;
 	struct list_head *hash,*lh;
 
-	spin_lock(&data->lock);
+	spin_lock_bh(&data->lock);
 	tuple = ct->tuplehash[0].tuple;
 	hash = &data->iphash[ipt_iphash(addr & mask)];
 
@@ -63,8 +63,8 @@
 	for (lh = hash->next; lh != hash; lh = lh->next) {
 		conn = list_entry(lh,struct ipt_connlimit_conn,list);
 		found = ip_conntrack_find_get(&conn->tuple,ct);
-		if (0 == memcmp(&conn->tuple,&tuple,sizeof(tuple)) &&
-		    found != NULL &&
+		if (found != NULL &&
+		    0 == memcmp(&conn->tuple,&tuple,sizeof(tuple)) &&
 		    found->ctrack->proto.tcp.state != TCP_CONNTRACK_TIME_WAIT) {
 			/* Just to be sure we have it only once in the list.
 			   We should'nt see tuples twice unless someone hooks this
@@ -117,7 +117,7 @@
 		list_add(&conn->list,hash);
 		matches++;
 	}
-	spin_unlock(&data->lock);
+	spin_unlock_bh(&data->lock);
 	return matches;
 }
 
Index: linux-2.6/net/ipv4/netfilter/ipt_connlimit.c
===================================================================
--- linux-2.6/net/ipv4/netfilter/ipt_connlimit.c	(revision 3889)
+++ linux-2.6/net/ipv4/netfilter/ipt_connlimit.c	(working copy)
@@ -55,7 +55,7 @@
 	struct ipt_connlimit_conn *conn;
 	struct list_head *hash,*lh;
 
-	spin_lock(&data->lock);
+	spin_lock_bh(&data->lock);
 	tuple = ct->tuplehash[0].tuple;
 	hash = &data->iphash[ipt_iphash(addr & mask)];
 
@@ -63,9 +63,9 @@
 	for (lh = hash->next; lh != hash; lh = lh->next) {
 		conn = list_entry(lh,struct ipt_connlimit_conn,list);
 		found = ip_conntrack_find_get(&conn->tuple,ct);
-		if (0 == memcmp(&conn->tuple,&tuple,sizeof(tuple)) &&
-		    found != NULL &&
-		    found->ctrack->proto.tcp.state != TCP_CONNTRACK_TIME_WAIT) {
+		if (found != NULL
+		    && 0 == memcmp(&conn->tuple,&tuple,sizeof(tuple))
+		    && found->proto.tcp.state != TCP_CONNTRACK_TIME_WAIT) {
 			/* Just to be sure we have it only once in the list.
 			   We should'nt see tuples twice unless someone hooks this
 			   into a table without "-p tcp --syn" */
@@ -110,7 +110,7 @@
 #endif
 		conn = kmalloc(sizeof(*conn),GFP_ATOMIC);
 		if (NULL == conn) {
-			spin_unlock(&data->lock);
+			spin_unlock_bh(&data->lock);
 			return -1;
 		}
 		memset(conn,0,sizeof(*conn));
@@ -119,7 +119,7 @@
 		list_add(&conn->list,hash);
 		matches++;
 	}
-	spin_unlock(&data->lock);
+	spin_unlock_bh(&data->lock);
 	return matches;
 }
 

Re: FIX: connlimit NULL pointer kernel panic

[Patrick McHardy]

Pablo Neira wrote:
>> Thanks Pablo, I've applied the patch. Can you send patches for the 2.4
>> and the old 2.6 version too?
> 
> Attached.

Applied, thanks a lot.

Regards
Patrick