1 files changed, 82 insertions, 142 deletions

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index b9eb90b39ac7..d67e612bf0ef 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -347,94 +347,18 @@ struct sk_buff *build_skb(void *data, unsigned int frag_size)
 }
 EXPORT_SYMBOL(build_skb);
 
-struct netdev_alloc_cache {
-	struct page_frag	frag;
-	/* we maintain a pagecount bias, so that we dont dirty cache line
-	 * containing page->_count every time we allocate a fragment.
-	 */
-	unsigned int		pagecnt_bias;
-};
-static DEFINE_PER_CPU(struct netdev_alloc_cache, netdev_alloc_cache);
-static DEFINE_PER_CPU(struct netdev_alloc_cache, napi_alloc_cache);
-
-static struct page *__page_frag_refill(struct netdev_alloc_cache *nc,
-				       gfp_t gfp_mask)
-{
-	const unsigned int order = NETDEV_FRAG_PAGE_MAX_ORDER;
-	struct page *page = NULL;
-	gfp_t gfp = gfp_mask;
-
-	if (order) {
-		gfp_mask |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY |
-			    __GFP_NOMEMALLOC;
-		page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
-		nc->frag.size = PAGE_SIZE << (page ? order : 0);
-	}
-
-	if (unlikely(!page))
-		page = alloc_pages_node(NUMA_NO_NODE, gfp, 0);
-
-	nc->frag.page = page;
-
-	return page;
-}
-
-static void *__alloc_page_frag(struct netdev_alloc_cache __percpu *cache,
-			       unsigned int fragsz, gfp_t gfp_mask)
-{
-	struct netdev_alloc_cache *nc = this_cpu_ptr(cache);
-	struct page *page = nc->frag.page;
-	unsigned int size;
-	int offset;
-
-	if (unlikely(!page)) {
-refill:
-		page = __page_frag_refill(nc, gfp_mask);
-		if (!page)
-			return NULL;
-
-		/* if size can vary use frag.size else just use PAGE_SIZE */
-		size = NETDEV_FRAG_PAGE_MAX_ORDER ? nc->frag.size : PAGE_SIZE;
-
-		/* Even if we own the page, we do not use atomic_set().
-		 * This would break get_page_unless_zero() users.
-		 */
-		atomic_add(size - 1, &page->_count);
-
-		/* reset page count bias and offset to start of new frag */
-		nc->pagecnt_bias = size;
-		nc->frag.offset = size;
-	}
-
-	offset = nc->frag.offset - fragsz;
-	if (unlikely(offset < 0)) {
-		if (!atomic_sub_and_test(nc->pagecnt_bias, &page->_count))
-			goto refill;
-
-		/* if size can vary use frag.size else just use PAGE_SIZE */
-		size = NETDEV_FRAG_PAGE_MAX_ORDER ? nc->frag.size : PAGE_SIZE;
-
-		/* OK, page count is 0, we can safely set it */
-		atomic_set(&page->_count, size);
-
-		/* reset page count bias and offset to start of new frag */
-		nc->pagecnt_bias = size;
-		offset = size - fragsz;
-	}
-
-	nc->pagecnt_bias--;
-	nc->frag.offset = offset;
-
-	return page_address(page) + offset;
-}
+static DEFINE_PER_CPU(struct page_frag_cache, netdev_alloc_cache);
+static DEFINE_PER_CPU(struct page_frag_cache, napi_alloc_cache);
 
 static void *__netdev_alloc_frag(unsigned int fragsz, gfp_t gfp_mask)
 {
+	struct page_frag_cache *nc;
 	unsigned long flags;
 	void *data;
 
 	local_irq_save(flags);
-	data = __alloc_page_frag(&netdev_alloc_cache, fragsz, gfp_mask);
+	nc = this_cpu_ptr(&netdev_alloc_cache);
+	data = __alloc_page_frag(nc, fragsz, gfp_mask);
 	local_irq_restore(flags);
 	return data;
 }
@@ -454,7 +378,9 @@ EXPORT_SYMBOL(netdev_alloc_frag);
 
 static void *__napi_alloc_frag(unsigned int fragsz, gfp_t gfp_mask)
 {
-	return __alloc_page_frag(&napi_alloc_cache, fragsz, gfp_mask);
+	struct page_frag_cache *nc = this_cpu_ptr(&napi_alloc_cache);
+
+	return __alloc_page_frag(nc, fragsz, gfp_mask);
 }
 
 void *napi_alloc_frag(unsigned int fragsz)
@@ -464,76 +390,64 @@ void *napi_alloc_frag(unsigned int fragsz)
 EXPORT_SYMBOL(napi_alloc_frag);
 
 /**
- *	__alloc_rx_skb - allocate an skbuff for rx
+ *	__netdev_alloc_skb - allocate an skbuff for rx on a specific device
+ *	@dev: network device to receive on
  *	@length: length to allocate
  *	@gfp_mask: get_free_pages mask, passed to alloc_skb
- *	@flags:	If SKB_ALLOC_RX is set, __GFP_MEMALLOC will be used for
- *		allocations in case we have to fallback to __alloc_skb()
- *		If SKB_ALLOC_NAPI is set, page fragment will be allocated
- *		from napi_cache instead of netdev_cache.
  *
  *	Allocate a new &sk_buff and assign it a usage count of one. The
- *	buffer has unspecified headroom built in. Users should allocate
+ *	buffer has NET_SKB_PAD headroom built in. Users should allocate
  *	the headroom they think they need without accounting for the
  *	built in space. The built in space is used for optimisations.
  *
  *	%NULL is returned if there is no free memory.
  */
-static struct sk_buff *__alloc_rx_skb(unsigned int length, gfp_t gfp_mask,
-				      int flags)
+struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int len,
+				   gfp_t gfp_mask)
 {
-	struct sk_buff *skb = NULL;
-	unsigned int fragsz = SKB_DATA_ALIGN(length) +
-			      SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
+	struct page_frag_cache *nc;
+	unsigned long flags;
+	struct sk_buff *skb;
+	bool pfmemalloc;
+	void *data;
 
-	if (fragsz <= PAGE_SIZE && !(gfp_mask & (__GFP_WAIT | GFP_DMA))) {
-		void *data;
+	len += NET_SKB_PAD;
 
-		if (sk_memalloc_socks())
-			gfp_mask |= __GFP_MEMALLOC;
+	if ((len > SKB_WITH_OVERHEAD(PAGE_SIZE)) ||
+	    (gfp_mask & (__GFP_WAIT | GFP_DMA)))
+		return __alloc_skb(len, gfp_mask, SKB_ALLOC_RX, NUMA_NO_NODE);
 
-		data = (flags & SKB_ALLOC_NAPI) ?
-			__napi_alloc_frag(fragsz, gfp_mask) :
-			__netdev_alloc_frag(fragsz, gfp_mask);
+	len += SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
+	len = SKB_DATA_ALIGN(len);
 
-		if (likely(data)) {
-			skb = build_skb(data, fragsz);
-			if (unlikely(!skb))
-				put_page(virt_to_head_page(data));
-		}
-	} else {
-		skb = __alloc_skb(length, gfp_mask,
-				  SKB_ALLOC_RX, NUMA_NO_NODE);
-	}
-	return skb;
-}
+	if (sk_memalloc_socks())
+		gfp_mask |= __GFP_MEMALLOC;
 
-/**
- *	__netdev_alloc_skb - allocate an skbuff for rx on a specific device
- *	@dev: network device to receive on
- *	@length: length to allocate
- *	@gfp_mask: get_free_pages mask, passed to alloc_skb
- *
- *	Allocate a new &sk_buff and assign it a usage count of one. The
- *	buffer has NET_SKB_PAD headroom built in. Users should allocate
- *	the headroom they think they need without accounting for the
- *	built in space. The built in space is used for optimisations.
- *
- *	%NULL is returned if there is no free memory.
- */
-struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
-				   unsigned int length, gfp_t gfp_mask)
-{
-	struct sk_buff *skb;
+	local_irq_save(flags);
 
-	length += NET_SKB_PAD;
-	skb = __alloc_rx_skb(length, gfp_mask, 0);
+	nc = this_cpu_ptr(&netdev_alloc_cache);
+	data = __alloc_page_frag(nc, len, gfp_mask);
+	pfmemalloc = nc->pfmemalloc;
 
-	if (likely(skb)) {
-		skb_reserve(skb, NET_SKB_PAD);
-		skb->dev = dev;
+	local_irq_restore(flags);
+
+	if (unlikely(!data))
+		return NULL;
+
+	skb = __build_skb(data, len);
+	if (unlikely(!skb)) {
+		skb_free_frag(data);
+		return NULL;
 	}
 
+	/* use OR instead of assignment to avoid clearing of bits in mask */
+	if (pfmemalloc)
+		skb->pfmemalloc = 1;
+	skb->head_frag = 1;
+
+	skb_reserve(skb, NET_SKB_PAD);
+	skb->dev = dev;
+
 	return skb;
 }
 EXPORT_SYMBOL(__netdev_alloc_skb);
@@ -551,19 +465,43 @@ EXPORT_SYMBOL(__netdev_alloc_skb);
  *
  *	%NULL is returned if there is no free memory.
  */
-struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
-				 unsigned int length, gfp_t gfp_mask)
+struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, unsigned int len,
+				 gfp_t gfp_mask)
 {
+	struct page_frag_cache *nc = this_cpu_ptr(&napi_alloc_cache);
 	struct sk_buff *skb;
+	void *data;
+
+	len += NET_SKB_PAD + NET_IP_ALIGN;
 
-	length += NET_SKB_PAD + NET_IP_ALIGN;
-	skb = __alloc_rx_skb(length, gfp_mask, SKB_ALLOC_NAPI);
+	if ((len > SKB_WITH_OVERHEAD(PAGE_SIZE)) ||
+	    (gfp_mask & (__GFP_WAIT | GFP_DMA)))
+		return __alloc_skb(len, gfp_mask, SKB_ALLOC_RX, NUMA_NO_NODE);
 
-	if (likely(skb)) {
-		skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
-		skb->dev = napi->dev;
+	len += SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
+	len = SKB_DATA_ALIGN(len);
+
+	if (sk_memalloc_socks())
+		gfp_mask |= __GFP_MEMALLOC;
+
+	data = __alloc_page_frag(nc, len, gfp_mask);
+	if (unlikely(!data))
+		return NULL;
+
+	skb = __build_skb(data, len);
+	if (unlikely(!skb)) {
+		skb_free_frag(data);
+		return NULL;
 	}
 
+	/* use OR instead of assignment to avoid clearing of bits in mask */
+	if (nc->pfmemalloc)
+		skb->pfmemalloc = 1;
+	skb->head_frag = 1;
+
+	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN);
+	skb->dev = napi->dev;
+
 	return skb;
 }
 EXPORT_SYMBOL(__napi_alloc_skb);
@@ -611,10 +549,12 @@ static void skb_clone_fraglist(struct sk_buff *skb)
 
 static void skb_free_head(struct sk_buff *skb)
 {
+	unsigned char *head = skb->head;
+
 	if (skb->head_frag)
-		put_page(virt_to_head_page(skb->head));
+		skb_free_frag(head);
 	else
-		kfree(skb->head);
+		kfree(head);
 }
 
 static void skb_release_data(struct sk_buff *skb)