---

yaml
---
r: 256149
b: refs/heads/master
c: a6686f2
h: refs/heads/master
i:
  256147: 08b94f7
v: v3

[refs]

@@ -1,2 +1,2 @@
 ---
-refs/heads/master: 1cdebb423202e255366a321814fc6df079802a0d
+refs/heads/master: a6686f2f382b13f8a7253401a66690c3633b6a74

trunk/include/linux/skbuff.h

@@ -187,6 +187,20 @@ enum {
 
 	/* ensure the originating sk reference is available on driver level */
 	SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
+
+	/* device driver supports TX zero-copy buffers */
+	SKBTX_DEV_ZEROCOPY = 1 << 4,
+};
+
+/*
+ * The callback notifies userspace to release buffers when skb DMA is done in
+ * lower device, the skb last reference should be 0 when calling this.
+ * The desc is used to track userspace buffer index.
+ */
+struct ubuf_info {
+	void (*callback)(void *);
+	void *arg;
+	unsigned long desc;
 };
 
 /* This data is invariant across clones and lives at
@@ -211,6 +225,7 @@ struct skb_shared_info {
 	/* Intermediate layers must ensure that destructor_arg
 	 * remains valid until skb destructor */
 	void *		destructor_arg;
+
 	/* must be last field, see pskb_expand_head() */
 	skb_frag_t	frags[MAX_SKB_FRAGS];
 };
@@ -2265,5 +2280,6 @@ static inline void skb_checksum_none_assert(struct sk_buff *skb)
 }
 
 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
+
 #endif	/* __KERNEL__ */
 #endif	/* _LINUX_SKBUFF_H */

trunk/net/core/skbuff.c

@@ -329,6 +329,18 @@ static void skb_release_data(struct sk_buff *skb)
 				put_page(skb_shinfo(skb)->frags[i].page);
 		}
 
+		/*
+		 * If skb buf is from userspace, we need to notify the caller
+		 * the lower device DMA has done;
+		 */
+		if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
+			struct ubuf_info *uarg;
+
+			uarg = skb_shinfo(skb)->destructor_arg;
+			if (uarg->callback)
+				uarg->callback(uarg);
+		}
+
 		if (skb_has_frag_list(skb))
 			skb_drop_fraglist(skb);
 
@@ -481,6 +493,9 @@ bool skb_recycle_check(struct sk_buff *skb, int skb_size)
 	if (irqs_disabled())
 		return false;
 
+	if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)
+		return false;
+
 	if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE)
 		return false;
 
@@ -596,6 +611,51 @@ struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src)
 }
 EXPORT_SYMBOL_GPL(skb_morph);
 
+/* skb frags copy userspace buffers to kernel */
+static int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask)
+{
+	int i;
+	int num_frags = skb_shinfo(skb)->nr_frags;
+	struct page *page, *head = NULL;
+	struct ubuf_info *uarg = skb_shinfo(skb)->destructor_arg;
+
+	for (i = 0; i < num_frags; i++) {
+		u8 *vaddr;
+		skb_frag_t *f = &skb_shinfo(skb)->frags[i];
+
+		page = alloc_page(GFP_ATOMIC);
+		if (!page) {
+			while (head) {
+				struct page *next = (struct page *)head->private;
+				put_page(head);
+				head = next;
+			}
+			return -ENOMEM;
+		}
+		vaddr = kmap_skb_frag(&skb_shinfo(skb)->frags[i]);
+		memcpy(page_address(page),
+		       vaddr + f->page_offset, f->size);
+		kunmap_skb_frag(vaddr);
+		page->private = (unsigned long)head;
+		head = page;
+	}
+
+	/* skb frags release userspace buffers */
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
+		put_page(skb_shinfo(skb)->frags[i].page);
+
+	uarg->callback(uarg);
+
+	/* skb frags point to kernel buffers */
+	for (i = skb_shinfo(skb)->nr_frags; i > 0; i--) {
+		skb_shinfo(skb)->frags[i - 1].page_offset = 0;
+		skb_shinfo(skb)->frags[i - 1].page = head;
+		head = (struct page *)head->private;
+	}
+	return 0;
+}
+
+
 /**
  *	skb_clone	-	duplicate an sk_buff
  *	@skb: buffer to clone
@@ -614,6 +674,11 @@ struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t gfp_mask)
 {
 	struct sk_buff *n;
 
+	if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
+		if (skb_copy_ubufs(skb, gfp_mask))
+			return NULL;
+	}
+
 	n = skb + 1;
 	if (skb->fclone == SKB_FCLONE_ORIG &&
 	    n->fclone == SKB_FCLONE_UNAVAILABLE) {
@@ -731,6 +796,12 @@ struct sk_buff *pskb_copy(struct sk_buff *skb, gfp_t gfp_mask)
 	if (skb_shinfo(skb)->nr_frags) {
 		int i;
 
+		if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
+			if (skb_copy_ubufs(skb, gfp_mask)) {
+				kfree(n);
+				goto out;
+			}
+		}
 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
 			skb_shinfo(n)->frags[i] = skb_shinfo(skb)->frags[i];
 			get_page(skb_shinfo(n)->frags[i].page);
@@ -788,7 +859,6 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
 		fastpath = true;
 	else {
 		int delta = skb->nohdr ? (1 << SKB_DATAREF_SHIFT) + 1 : 1;
-
 		fastpath = atomic_read(&skb_shinfo(skb)->dataref) == delta;
 	}
 
@@ -819,6 +889,11 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
 	if (fastpath) {
 		kfree(skb->head);
 	} else {
+		/* copy this zero copy skb frags */
+		if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
+			if (skb_copy_ubufs(skb, gfp_mask))
+				goto nofrags;
+		}
 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
 			get_page(skb_shinfo(skb)->frags[i].page);
 
@@ -853,6 +928,8 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
 	atomic_set(&skb_shinfo(skb)->dataref, 1);
 	return 0;
 
+nofrags:
+	kfree(data);
 nodata:
 	return -ENOMEM;
 }
@@ -1354,6 +1431,7 @@ int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len)
 		}
 		start = end;
 	}
+
 	if (!len)
 		return 0;