Merge branch 'net-permit-skb_segment-on-head_frag-frag_list-skb'

Yonghong Song says:

====================
net: permit skb_segment on head_frag frag_list skb

One of our in-house projects, bpf-based NAT, hits a kernel BUG_ON at
function skb_segment(), line 3667. The bpf program attaches to
clsact ingress, calls bpf_skb_change_proto to change protocol
from ipv4 to ipv6 or from ipv6 to ipv4, and then calls bpf_redirect
to send the changed packet out.
 ...
    3665                 while (pos < offset + len) {
    3666                         if (i >= nfrags) {
    3667                                 BUG_ON(skb_headlen(list_skb));
 ...

The triggering input skb has the following properties:
    list_skb = skb->frag_list;
    skb->nfrags != NULL && skb_headlen(list_skb) != 0
and skb_segment() is not able to handle a frag_list skb
if its headlen (list_skb->len - list_skb->data_len) is not 0.

Patch #1 provides a simple solution to avoid BUG_ON. If
list_skb->head_frag is true, its page-backed frag will
be processed before the list_skb->frags.
Patch #2 provides a test case in test_bpf module which
constructs a skb and calls skb_segment() directly. The test
case is able to trigger the BUG_ON without Patch #1.

The patch has been tested in the following setup:
  ipv6_host <-> nat_server <-> ipv4_host
where nat_server has a bpf program doing ipv4<->ipv6
translation and forwarding through clsact hook
bpf_skb_change_proto.

Changelog:
v5 -> v6:
  . Added back missed BUG_ON(!nfrags) for zero
    skb_headlen(skb) case, plus a couple of
    cosmetic changes, from Alexander.
v4 -> v5:
  . Replace local variable head_frag with
    a static inline function skb_head_frag_to_page_desc
    which gets the head_frag on-demand. This makes
    code more readable and also does not increase
    the stack size, from Alexander.
  . Remove the "if(nfrags)" guard for skb_orphan_frags
    and skb_zerocopy_clone as I found that they can
    handle zero-frag skb (with non-zero skb_headlen(skb))
    properly.
  . Properly release segment list from skb_segment()
    in the test, from Eric.
v3 -> v4:
  . Remove dynamic memory allocation and use rewinding
    for both index and frag to remove one branch in fast path,
    from Alexander.
  . Fix a bunch of issues in test_bpf skb_segment() test,
    including proper way to allocate skb, proper function
    argument for skb_add_rx_frag and not freeint skb, etc.,
    from Eric.
v2 -> v3:
  . Use starting frag index -1 (instead of 0) to
    special process head_frag before other frags in the skb,
    from Alexander Duyck.
v1 -> v2:
  . Removed never-hit BUG_ON, spotted by Linyu Yuan.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

lib/test_bpf.c

@@ -6574,6 +6574,93 @@ static bool exclude_test(int test_id)
 	return test_id < test_range[0] || test_id > test_range[1];
 }
 
+static __init struct sk_buff *build_test_skb(void)
+{
+	u32 headroom = NET_SKB_PAD + NET_IP_ALIGN + ETH_HLEN;
+	struct sk_buff *skb[2];
+	struct page *page[2];
+	int i, data_size = 8;
+
+	for (i = 0; i < 2; i++) {
+		page[i] = alloc_page(GFP_KERNEL);
+		if (!page[i]) {
+			if (i == 0)
+				goto err_page0;
+			else
+				goto err_page1;
+		}
+
+		/* this will set skb[i]->head_frag */
+		skb[i] = dev_alloc_skb(headroom + data_size);
+		if (!skb[i]) {
+			if (i == 0)
+				goto err_skb0;
+			else
+				goto err_skb1;
+		}
+
+		skb_reserve(skb[i], headroom);
+		skb_put(skb[i], data_size);
+		skb[i]->protocol = htons(ETH_P_IP);
+		skb_reset_network_header(skb[i]);
+		skb_set_mac_header(skb[i], -ETH_HLEN);
+
+		skb_add_rx_frag(skb[i], 0, page[i], 0, 64, 64);
+		// skb_headlen(skb[i]): 8, skb[i]->head_frag = 1
+	}
+
+	/* setup shinfo */
+	skb_shinfo(skb[0])->gso_size = 1448;
+	skb_shinfo(skb[0])->gso_type = SKB_GSO_TCPV4;
+	skb_shinfo(skb[0])->gso_type |= SKB_GSO_DODGY;
+	skb_shinfo(skb[0])->gso_segs = 0;
+	skb_shinfo(skb[0])->frag_list = skb[1];
+
+	/* adjust skb[0]'s len */
+	skb[0]->len += skb[1]->len;
+	skb[0]->data_len += skb[1]->data_len;
+	skb[0]->truesize += skb[1]->truesize;
+
+	return skb[0];
+
+err_skb1:
+	__free_page(page[1]);
+err_page1:
+	kfree_skb(skb[0]);
+err_skb0:
+	__free_page(page[0]);
+err_page0:
+	return NULL;
+}
+
+static __init int test_skb_segment(void)
+{
+	netdev_features_t features;
+	struct sk_buff *skb, *segs;
+	int ret = -1;
+
+	features = NETIF_F_SG | NETIF_F_GSO_PARTIAL | NETIF_F_IP_CSUM |
+		   NETIF_F_IPV6_CSUM;
+	features |= NETIF_F_RXCSUM;
+	skb = build_test_skb();
+	if (!skb) {
+		pr_info("%s: failed to build_test_skb", __func__);
+		goto done;
+	}
+
+	segs = skb_segment(skb, features);
+	if (segs) {
+		kfree_skb_list(segs);
+		ret = 0;
+		pr_info("%s: success in skb_segment!", __func__);
+	} else {
+		pr_info("%s: failed in skb_segment!", __func__);
+	}
+	kfree_skb(skb);
+done:
+	return ret;
+}
+
 static __init int test_bpf(void)
 {
 	int i, err_cnt = 0, pass_cnt = 0;
@@ -6632,9 +6719,11 @@ static int __init test_bpf_init(void)
 		return ret;
 
 	ret = test_bpf();
-
 	destroy_bpf_tests();
-	return ret;
+	if (ret)
+		return ret;
+
+	return test_skb_segment();
 }
 
 static void __exit test_bpf_exit(void)

net/core/skbuff.c

@@ -3460,6 +3460,19 @@ void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len)
 }
 EXPORT_SYMBOL_GPL(skb_pull_rcsum);
 
+static inline skb_frag_t skb_head_frag_to_page_desc(struct sk_buff *frag_skb)
+{
+	skb_frag_t head_frag;
+	struct page *page;
+
+	page = virt_to_head_page(frag_skb->head);
+	head_frag.page.p = page;
+	head_frag.page_offset = frag_skb->data -
+		(unsigned char *)page_address(page);
+	head_frag.size = skb_headlen(frag_skb);
+	return head_frag;
+}
+
 /**
  *	skb_segment - Perform protocol segmentation on skb.
  *	@head_skb: buffer to segment
@@ -3664,15 +3677,19 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 
 		while (pos < offset + len) {
 			if (i >= nfrags) {
-				BUG_ON(skb_headlen(list_skb));
-
 				i = 0;
 				nfrags = skb_shinfo(list_skb)->nr_frags;
 				frag = skb_shinfo(list_skb)->frags;
 				frag_skb = list_skb;
+				if (!skb_headlen(list_skb)) {
+					BUG_ON(!nfrags);
+				} else {
+					BUG_ON(!list_skb->head_frag);
 
-				BUG_ON(!nfrags);
-
+					/* to make room for head_frag. */
+					i--;
+					frag--;
+				}
 				if (skb_orphan_frags(frag_skb, GFP_ATOMIC) ||
 				    skb_zerocopy_clone(nskb, frag_skb,
 						       GFP_ATOMIC))
@@ -3689,7 +3706,7 @@ struct sk_buff *skb_segment(struct sk_buff *head_skb,
 				goto err;
 			}
 
-			*nskb_frag = *frag;
+			*nskb_frag = (i < 0) ? skb_head_frag_to_page_desc(frag_skb) : *frag;
 			__skb_frag_ref(nskb_frag);
 			size = skb_frag_size(nskb_frag);