xen: netback: handle compound page fragments on transmit.
authorIan Campbell <Ian.Campbell@citrix.com>
Wed, 10 Oct 2012 03:48:42 +0000 (03:48 +0000)
committerDavid S. Miller <davem@davemloft.net>
Thu, 11 Oct 2012 02:50:45 +0000 (22:50 -0400)
An SKB paged fragment can consist of a compound page with order > 0.
However the netchannel protocol deals only in PAGE_SIZE frames.

Handle this in netbk_gop_frag_copy and xen_netbk_count_skb_slots by
iterating over the frames which make up the page.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Konrad Rzeszutek Wilk <konrad@kernel.org>
Cc: Sander Eikelenboom <linux@eikelenboom.it>
Tested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/xen-netback/netback.c

index 4ebfcf3d8a3b26d07c5cd2e742ab59aff3c5ba63..f2d6b78d901d92d3965fe3bd849751f04fb7ccf3 100644 (file)
@@ -335,21 +335,35 @@ unsigned int xen_netbk_count_skb_slots(struct xenvif *vif, struct sk_buff *skb)
 
        for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
                unsigned long size = skb_frag_size(&skb_shinfo(skb)->frags[i]);
+               unsigned long offset = skb_shinfo(skb)->frags[i].page_offset;
                unsigned long bytes;
+
+               offset &= ~PAGE_MASK;
+
                while (size > 0) {
+                       BUG_ON(offset >= PAGE_SIZE);
                        BUG_ON(copy_off > MAX_BUFFER_OFFSET);
 
-                       if (start_new_rx_buffer(copy_off, size, 0)) {
+                       bytes = PAGE_SIZE - offset;
+
+                       if (bytes > size)
+                               bytes = size;
+
+                       if (start_new_rx_buffer(copy_off, bytes, 0)) {
                                count++;
                                copy_off = 0;
                        }
 
-                       bytes = size;
                        if (copy_off + bytes > MAX_BUFFER_OFFSET)
                                bytes = MAX_BUFFER_OFFSET - copy_off;
 
                        copy_off += bytes;
+
+                       offset += bytes;
                        size -= bytes;
+
+                       if (offset == PAGE_SIZE)
+                               offset = 0;
                }
        }
        return count;
@@ -403,14 +417,24 @@ static void netbk_gop_frag_copy(struct xenvif *vif, struct sk_buff *skb,
        unsigned long bytes;
 
        /* Data must not cross a page boundary. */
-       BUG_ON(size + offset > PAGE_SIZE);
+       BUG_ON(size + offset > PAGE_SIZE<<compound_order(page));
 
        meta = npo->meta + npo->meta_prod - 1;
 
+       /* Skip unused frames from start of page */
+       page += offset >> PAGE_SHIFT;
+       offset &= ~PAGE_MASK;
+
        while (size > 0) {
+               BUG_ON(offset >= PAGE_SIZE);
                BUG_ON(npo->copy_off > MAX_BUFFER_OFFSET);
 
-               if (start_new_rx_buffer(npo->copy_off, size, *head)) {
+               bytes = PAGE_SIZE - offset;
+
+               if (bytes > size)
+                       bytes = size;
+
+               if (start_new_rx_buffer(npo->copy_off, bytes, *head)) {
                        /*
                         * Netfront requires there to be some data in the head
                         * buffer.
@@ -420,7 +444,6 @@ static void netbk_gop_frag_copy(struct xenvif *vif, struct sk_buff *skb,
                        meta = get_next_rx_buffer(vif, npo);
                }
 
-               bytes = size;
                if (npo->copy_off + bytes > MAX_BUFFER_OFFSET)
                        bytes = MAX_BUFFER_OFFSET - npo->copy_off;
 
@@ -453,6 +476,13 @@ static void netbk_gop_frag_copy(struct xenvif *vif, struct sk_buff *skb,
                offset += bytes;
                size -= bytes;
 
+               /* Next frame */
+               if (offset == PAGE_SIZE && size) {
+                       BUG_ON(!PageCompound(page));
+                       page++;
+                       offset = 0;
+               }
+
                /* Leave a gap for the GSO descriptor. */
                if (*head && skb_shinfo(skb)->gso_size && !vif->gso_prefix)
                        vif->rx.req_cons++;