dpaa_eth: Use page fragment for egress SGT buffer

Fragmented skbuffs are wrapped in SG frames before being enqueued to
FMan's Tx port. A new buffer needs to be allocated on Tx (and freed on
Tx Confirm) to contain the SGT. Up until now, this buffer was one full
page.
Using the page fragment APIs to get smaller buffers for the SGT.

Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
Change-Id: I1da07141b119ca431e276025d833178cd73a1077
Reviewed-on: http://git.am.freescale.net:8181/2757
Tested-by: Review Code-CDREVIEW <CDREVIEW@freescale.com>
Reviewed-by: Sovaiala Cristian-Constantin-B39531 <Cristian.Sovaiala@freescale.com>
Reviewed-by: Radulescu Ruxandra Ioana-B05472 <ruxandra.radulescu@freescale.com>
Reviewed-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>
Tested-by: Fleming Andrew-AFLEMING <AFLEMING@freescale.com>

1 files changed, 15 insertions, 12 deletions

diff --git a/drivers/net/ethernet/freescale/dpa/dpaa_eth_sg.c b/drivers/net/ethernet/freescale/dpa/dpaa_eth_sg.c
index a624e9a..0104790 100644
--- a/drivers/net/ethernet/freescale/dpa/dpaa_eth_sg.c
+++ b/drivers/net/ethernet/freescale/dpa/dpaa_eth_sg.c
@@ -198,7 +198,8 @@ struct sk_buff *_dpa_cleanup_tx_fd(const struct dpa_priv_s *priv,
 
 	if (fd->format == qm_fd_sg) {
 		/*
-		 * The sgt page is guaranteed to reside in lowmem.
+		 * The sgt buffer has been allocated with netdev_alloc_frag(),
+		 * it's from lowmem.
 		 */
 		sgt = phys_to_virt(addr + dpa_fd_offset(fd));
 #ifdef CONFIG_FSL_DPAA_1588
@@ -234,8 +235,8 @@ struct sk_buff *_dpa_cleanup_tx_fd(const struct dpa_priv_s *priv,
 		 * and we know they're not used by anyone else
 		 */
 
-		/* Free separately the pages that we allocated on Tx */
-		free_page((unsigned long)phys_to_virt(addr));
+		/* Free the page frag that we allocated on Tx */
+		put_page(virt_to_head_page(sgt));
 	}
 #if defined(CONFIG_FSL_DPAA_1588) || defined(CONFIG_FSL_DPAA_TS)
 	else {
@@ -539,18 +540,20 @@ static int __hot skb_to_sg_fd(struct dpa_priv_s *priv,
 	int err;
 
 	struct qm_sg_entry *sgt;
-	unsigned long sgt_page;
+	void *sgt_buf;
 	void *buffer_start;
 	skb_frag_t *frag;
 	int i, j;
 	const enum dma_data_direction dma_dir = DMA_TO_DEVICE;
+	const int nr_frags = skb_shinfo(skb)->nr_frags;
 
 	fd->format = qm_fd_sg;
 
-	/* get a new page to store the SGTable */
-	sgt_page = __get_free_page(GFP_ATOMIC);
-	if (unlikely(!sgt_page)) {
-		dev_err(dpa_bp->dev, "__get_free_page() failed\n");
+	/* get a page frag to store the SGTable */
+	sgt_buf = netdev_alloc_frag(priv->tx_headroom +
+		sizeof(struct qm_sg_entry) * (1 + nr_frags));
+	if (unlikely(!sgt_buf)) {
+		dev_err(dpa_bp->dev, "netdev_alloc_frag() failed\n");
 		return -ENOMEM;
 	}
 
@@ -561,14 +564,14 @@ static int __hot skb_to_sg_fd(struct dpa_priv_s *priv,
 	 * need to write into the skb.
 	 */
 	err = dpa_enable_tx_csum(priv, skb, fd,
-				 (void *)sgt_page + DPA_TX_PRIV_DATA_SIZE);
+				 sgt_buf + DPA_TX_PRIV_DATA_SIZE);
 	if (unlikely(err < 0)) {
 		if (netif_msg_tx_err(priv) && net_ratelimit())
 			netdev_err(net_dev, "HW csum error: %d\n", err);
 		goto csum_failed;
 	}
 
-	sgt = (struct qm_sg_entry *)(sgt_page + priv->tx_headroom);
+	sgt = (struct qm_sg_entry *)(sgt_buf + priv->tx_headroom);
 	sgt[0].bpid = dpa_bp->bpid;
 	sgt[0].offset = 0;
 	sgt[0].length = skb_headlen(skb);
@@ -585,7 +588,7 @@ static int __hot skb_to_sg_fd(struct dpa_priv_s *priv,
 	sgt[0].addr_lo = lower_32_bits(addr);
 
 	/* populate the rest of SGT entries */
-	for (i = 1; i <= skb_shinfo(skb)->nr_frags; i++) {
+	for (i = 1; i <= nr_frags; i++) {
 		frag = &skb_shinfo(skb)->frags[i - 1];
 		sgt[i].bpid = dpa_bp->bpid;
 		sgt[i].offset = 0;
@@ -637,7 +640,7 @@ sg_map_failed:
 			dpa_bp->size, dma_dir);
 sg0_map_failed:
 csum_failed:
-	free_page(sgt_page);
+	put_page(virt_to_head_page(sgt_buf));
 
 	return err;
 }