Skip to contentSkip to conversation

net/ipv4/netfilter/nf_nat_snmp_basic.c: correct the size argument to kzalloc

PATCH In review [email protected] 2 participants 4 messages Feb 2, 2010 – Feb 3, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net/ipv4/netfilter/nf_nat_snmp_basic.c: correct the size argument to kzalloc

From: Julia Lawall <[email protected]>

obj has type struct snmp_object **, not struct snmp_object *. But indeed it is not even clear why kmalloc is needed. The memory is freed by the end of the function, so the local variable of pointer type should be sufficient.

The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@disable sizeof_type_expr@
type T;
T **x;
@@

x =
<+...sizeof(
- T
+ *x
)...+>
// </smpl>

Cc Patrick McHardy <[email protected]>
Cc "David S. Miller" <[email protected]>

---

 net/ipv4/netfilter/nf_nat_snmp_basic.c |   31 ++++++++---------------
 1 file changed, 11 insertions(+), 20 deletions(-)
net/ipv4/netfilter/nf_nat_snmp_basic.c renamed from net/ipv4/netfilter/nf_nat_snmp_basic.c~net-ipv4-correct-the-size-argument-to-kzalloc +11 -20
diff -puN net/ipv4/netfilter/nf_nat_snmp_basic.c~net-ipv4-correct-the-size-argument-to-kzalloc net/ipv4/netfilter/nf_nat_snmp_basic.c
@@ -1038,7 +1038,7 @@ static int snmp_parse_mangle(unsigned ch
 	unsigned int cls, con, tag, vers, pdutype;
 	struct asn1_ctx ctx;
 	struct asn1_octstr comm;
-	struct snmp_object **obj;
+	struct snmp_object *obj;
 
 	if (debug > 1)
 		hex_dump(msg, len);
@@ -1148,43 +1148,34 @@ static int snmp_parse_mangle(unsigned ch
 	if (cls != ASN1_UNI || con != ASN1_CON || tag != ASN1_SEQ)
 		return 0;
 
-	obj = kmalloc(sizeof(struct snmp_object), GFP_ATOMIC);
-	if (obj == NULL) {
-		if (net_ratelimit())
-			printk(KERN_WARNING "OOM in bsalg(%d)\n", __LINE__);
-		return 0;
-	}
-
 	while (!asn1_eoc_decode(&ctx, eoc)) {
 		unsigned int i;
 
-		if (!snmp_object_decode(&ctx, obj)) {
-			if (*obj) {
-				kfree((*obj)->id);
-				kfree(*obj);
+		if (!snmp_object_decode(&ctx, &obj)) {
+			if (obj) {
+				kfree(obj->id);
+				kfree(obj);
 			}
-			kfree(obj);
 			return 0;
 		}
 
 		if (debug > 1) {
 			printk(KERN_DEBUG "bsalg: object: ");
-			for (i = 0; i < (*obj)->id_len; i++) {
+			for (i = 0; i < obj->id_len; i++) {
 				if (i > 0)
 					printk(".");
-				printk("%lu", (*obj)->id[i]);
+				printk("%lu", obj->id[i]);
 			}
-			printk(": type=%u\n", (*obj)->type);
+			printk(": type=%u\n", obj->type);
 
 		}
 
-		if ((*obj)->type == SNMP_IPADDR)
+		if (obj->type == SNMP_IPADDR)
 			mangle_address(ctx.begin, ctx.pointer - 4 , map, check);
 
-		kfree((*obj)->id);
-		kfree(*obj);
+		kfree(obj->id);
+		kfree(obj);
 	}
-	kfree(obj);
 
 	if (!asn1_eoc_decode(&ctx, eoc))
 		return 0;

_

#1 lore raw
Quoted 6 lines · [email protected] wrote:

From: Julia Lawall <[email protected]>

obj has type struct snmp_object **, not struct snmp_object *. But indeed it is not even clear why kmalloc is needed. The memory is freed by the end of the function, so the local variable of pointer type should be sufficient.

This one is already queued in nf-next-2.6.git.

#2 lore raw
Quoted 9 lines · On Wed, 03 Feb 2010 01:37:20 +0100 Patrick McHardy <[email protected]> wrote:
Quoted 6 lines · [email protected] wrote:

From: Julia Lawall <[email protected]>

obj has type struct snmp_object **, not struct snmp_object *. But indeed it is not even clear why kmalloc is needed. The memory is freed by the end of the function, so the local variable of pointer type should be sufficient.

This one is already queued in nf-next-2.6.git.

Recently? It wasn't in today's linux-next.

#3 lore raw
Quoted 6 lines · Andrew Morton
Quoted 3 lines · On Wed, 03 Feb 2010 01:37:20 +0100 Patrick McHardy <[email protected]> wrote:
Quoted 2 lines · [email protected] wrote:

From: Julia Lawall <[email protected]>

Quoted 8 lines
Quoted 5 lines
Quoted 4 lines

obj has type struct snmp_object **, not struct snmp_object *. But indeed it is not even clear why kmalloc is needed. The memory is freed by the end of the function, so the local variable of pointer type should be sufficient.

This one is already queued in nf-next-2.6.git.

Recently? It wasn't in today's linux-next.

Actually I haven't merged it to Dave yet, but I will do so soon. Its queued and safe however :)