Em Sat, Mar 29, 2008 at 03:06:55PM +0900, Joonwoo Park escreveu:

On Fri, Mar 28, 2008 at 10:12:02AM -0300, Arnaldo Carvalho de Melo wrote:

quoted

quoted

+static int llc_mac_header_len(unsigned short devtype)
+{
+	switch (devtype) {
+	case ARPHRD_ETHER:
+	case ARPHRD_LOOPBACK:
+		return sizeof(struct ethhdr);
+	}
+
+	return __llc_tr_mac_header_len(devtype);
+}

Nah, just drop __llc_tr_mac_header_len altogether and have another case
entry for ARPHDR_IEEE802_TR.

quoted

 /**
  *	llc_alloc_frame - allocates sk_buff for frame
  *	@dev: network device this skb will be sent over
+ *	@type: pdu type to allocate
+ *	@data_size: data size to allocate
  *
  *	Allocates an sk_buff for frame and initializes sk_buff fields.
  *	Returns allocated skb or %NULL when out of memory.
  */
-struct sk_buff *llc_alloc_frame(struct sock *sk, struct net_device *dev)
+struct sk_buff *llc_alloc_frame(struct sock *sk, struct net_device *dev,
+				u8 type, u32 data_size)
 {
-	struct sk_buff *skb = alloc_skb(128, GFP_ATOMIC);
+	int hlen = type == LLC_PDU_TYPE_U ? 3 : 4;
+	struct sk_buff *skb;
+
+	hlen += llc_mac_header_len(dev->type);
+	skb = alloc_skb(hlen + data_size, GFP_ATOMIC);
 
 	if (skb) {
 		skb_reset_mac_header(skb);
-		skb_reserve(skb, 50);
+		skb_reserve(skb, hlen);

Well done, now it is better documented (LLC_PDU_TYPE_U passed) and uses
less memory, no more magic numbers, good.

- Arnaldo

I've respined the patch, thanks Arnaldo.

---
[LLC]: skb allocation size for responses

Allocate the skb for llc responses with the received packet size by
using the size adjustable llc_frame_alloc.
Don't allocate useless extra payload.
Cleanup magic numbers.

So, this fixes oops.
Reported by Jim Westfall:
kernel: skb_over_panic: text:c0541fc7 len:1000 put:997 head:c166ac00 data:c166ac2f tail:0xc166b017 end:0xc166ac80 dev:eth0
kernel: ------------[ cut here ]------------
kernel: kernel BUG at net/core/skbuff.c:95!

Perfect, thanks a lot:

Acked-by: Arnaldo Carvalho de Melo <redacted>

- Arnaldo