[PATCH net 0/3] Marvell Armada 375 mvpp2 fixes

[PATCH net 0/3] Marvell Armada 375 mvpp2 fixes

[Marcin Wojtas]

Hi,

During my work on mvneta driver I revised mvpp2, and it occurred that the
initial version of Marvell Armada 375 SoC comprised bugs around
DMA-unmapping in both ingress and egress paths - not all buffers were
umapped in TX path and none(!) in RX. Three patches that I send fix
this situation.

Any feedback would be welcome.

Best regards,
Marcin Wojtas

Marcin Wojtas (3):
  net: mvpp2: fix missing DMA region unmap in egress processing
  net: mvpp2: fix buffers' DMA handling on RX path
  net: mvpp2: fix refilling BM pools in RX path

 drivers/net/ethernet/marvell/mvpp2.c | 52 +++++++++++++++++++++++-------------
 1 file changed, 33 insertions(+), 19 deletions(-)

-- 
1.8.3.1

[PATCH net 1/3] net: mvpp2: fix missing DMA region unmap in egress processing

[Marcin Wojtas]

The Tx descriptor release code currently calls dma_unmap_single() and
dev_kfree_skb_any() if the descriptor is associated with a non-NULL skb.
This condition is true only for the last fragment of the packet.

Since every descriptor's buffer is DMA-mapped it has to be properly
unmapped.

Signed-off-by: Marcin Wojtas <mw@semihalf.com>

Fixes: 3f518509dedc ("ethernet: Add new driver for Marvell Armada 375
network unit")

Cc: <stable@vger.kernel.org> # v3.18+
---
 drivers/net/ethernet/marvell/mvpp2.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index d9884fd..95db519 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -4401,11 +4401,10 @@ static void mvpp2_txq_bufs_free(struct mvpp2_port *port,
 
 		mvpp2_txq_inc_get(txq_pcpu);
 
-		if (!skb)
-			continue;
-
 		dma_unmap_single(port->dev->dev.parent, buf_phys_addr,
 				 skb_headlen(skb), DMA_TO_DEVICE);
+		if (!skb)
+			continue;
 		dev_kfree_skb_any(skb);
 	}
 }
-- 
1.8.3.1

[PATCH net 2/3] net: mvpp2: fix buffers' DMA handling on RX path

[Marcin Wojtas]

Each allocated buffer, whose pointer is put into BM pool is DMA-mapped.
Hence it should be properly unmapped after usage or when removing buffers
from pool.

This commit fixes DMA handling on RX path by adding dma_unmap_single() in
mvpp2_rx() and in mvpp2_bufs_free(). The latter function's argument number
had to be increased for this purpose.

Signed-off-by: Marcin Wojtas <mw@semihalf.com>

Fixes: 3f518509dedc ("ethernet: Add new driver for Marvell Armada 375
network unit")

Cc: <stable@vger.kernel.org> # v3.18+
---
 drivers/net/ethernet/marvell/mvpp2.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 95db519..eaef461 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -3413,16 +3413,23 @@ static void mvpp2_bm_pool_bufsize_set(struct mvpp2 *priv,
 }
 
 /* Free all buffers from the pool */
-static void mvpp2_bm_bufs_free(struct mvpp2 *priv, struct mvpp2_bm_pool *bm_pool)
+static void mvpp2_bm_bufs_free(struct device *dev, struct mvpp2 *priv,
+			       struct mvpp2_bm_pool *bm_pool)
 {
 	int i;
 
 	for (i = 0; i < bm_pool->buf_num; i++) {
+		dma_addr_t buf_phys_addr;
 		u32 vaddr;
 
 		/* Get buffer virtual address (indirect access) */
-		mvpp2_read(priv, MVPP2_BM_PHY_ALLOC_REG(bm_pool->id));
+		buf_phys_addr = mvpp2_read(priv,
+					   MVPP2_BM_PHY_ALLOC_REG(bm_pool->id));
 		vaddr = mvpp2_read(priv, MVPP2_BM_VIRT_ALLOC_REG);
+
+		dma_unmap_single(dev, buf_phys_addr,
+				 bm_pool->buf_size, DMA_FROM_DEVICE);
+
 		if (!vaddr)
 			break;
 		dev_kfree_skb_any((struct sk_buff *)vaddr);
@@ -3439,7 +3446,7 @@ static int mvpp2_bm_pool_destroy(struct platform_device *pdev,
 {
 	u32 val;
 
-	mvpp2_bm_bufs_free(priv, bm_pool);
+	mvpp2_bm_bufs_free(&pdev->dev, priv, bm_pool);
 	if (bm_pool->buf_num) {
 		WARN(1, "cannot free all buffers in pool %d\n", bm_pool->id);
 		return 0;
@@ -3692,7 +3699,8 @@ mvpp2_bm_pool_use(struct mvpp2_port *port, int pool, enum mvpp2_bm_type type,
 				   MVPP2_BM_LONG_BUF_NUM :
 				   MVPP2_BM_SHORT_BUF_NUM;
 		else
-			mvpp2_bm_bufs_free(port->priv, new_pool);
+			mvpp2_bm_bufs_free(port->dev->dev.parent,
+					   port->priv, new_pool);
 
 		new_pool->pkt_size = pkt_size;
 
@@ -3756,7 +3764,7 @@ static int mvpp2_bm_update_mtu(struct net_device *dev, int mtu)
 	int pkt_size = MVPP2_RX_PKT_SIZE(mtu);
 
 	/* Update BM pool with new buffer size */
-	mvpp2_bm_bufs_free(port->priv, port_pool);
+	mvpp2_bm_bufs_free(dev->dev.parent, port->priv, port_pool);
 	if (port_pool->buf_num) {
 		WARN(1, "cannot free all buffers in pool %d\n", port_pool->id);
 		return -EIO;
@@ -5136,6 +5144,9 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 
 		skb = (struct sk_buff *)rx_desc->buf_cookie;
 
+		dma_unmap_single(dev->dev.parent, rx_desc->buf_phys_addr,
+				 bm_pool->buf_size, DMA_FROM_DEVICE);
+
 		rcvd_pkts++;
 		rcvd_bytes += rx_bytes;
 		atomic_inc(&bm_pool->in_use);
-- 
1.8.3.1

[PATCH net 3/3] net: mvpp2: fix refilling BM pools in RX path

[Marcin Wojtas]

In hitherto code in case of RX buffer allocation error during refill,
original buffer is pushed to the network stack, but the amount of
available buffer pointers in BM pool is decreased.

This commit fixes the situation by moving refill call before skb_put(),
and returning original buffer pointer to the pool in case of an error.

Signed-off-by: Marcin Wojtas <mw@semihalf.com>

Fixes: 3f518509dedc ("ethernet: Add new driver for Marvell Armada 375
network unit")

Cc: <stable@vger.kernel.org> # v3.18+
---
 drivers/net/ethernet/marvell/mvpp2.c | 28 ++++++++++++++++------------
 1 file changed, 16 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index eaef461..a4beccf 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -5099,7 +5099,8 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 		    struct mvpp2_rx_queue *rxq)
 {
 	struct net_device *dev = port->dev;
-	int rx_received, rx_filled, i;
+	int rx_received;
+	int rx_done = 0;
 	u32 rcvd_pkts = 0;
 	u32 rcvd_bytes = 0;
 
@@ -5108,17 +5109,18 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 	if (rx_todo > rx_received)
 		rx_todo = rx_received;
 
-	rx_filled = 0;
-	for (i = 0; i < rx_todo; i++) {
+	while (rx_done < rx_todo) {
 		struct mvpp2_rx_desc *rx_desc = mvpp2_rxq_next_desc_get(rxq);
 		struct mvpp2_bm_pool *bm_pool;
 		struct sk_buff *skb;
+		dma_addr_t phys_addr;
 		u32 bm, rx_status;
 		int pool, rx_bytes, err;
 
-		rx_filled++;
+		rx_done++;
 		rx_status = rx_desc->status;
 		rx_bytes = rx_desc->data_size - MVPP2_MH_SIZE;
+		phys_addr = rx_desc->buf_phys_addr;
 
 		bm = mvpp2_bm_cookie_build(rx_desc);
 		pool = mvpp2_bm_cookie_pool_get(bm);
@@ -5135,8 +5137,10 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 		 * comprised by the RX descriptor.
 		 */
 		if (rx_status & MVPP2_RXD_ERR_SUMMARY) {
+		err_drop_frame:
 			dev->stats.rx_errors++;
 			mvpp2_rx_error(port, rx_desc);
+			/* Return the buffer to the pool */
 			mvpp2_pool_refill(port, bm, rx_desc->buf_phys_addr,
 					  rx_desc->buf_cookie);
 			continue;
@@ -5144,7 +5148,13 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 
 		skb = (struct sk_buff *)rx_desc->buf_cookie;
 
-		dma_unmap_single(dev->dev.parent, rx_desc->buf_phys_addr,
+		err = mvpp2_rx_refill(port, bm_pool, bm, 0);
+		if (err) {
+			netdev_err(port->dev, "failed to refill BM pools\n");
+			goto err_drop_frame;
+		}
+
+		dma_unmap_single(dev->dev.parent, phys_addr,
 				 bm_pool->buf_size, DMA_FROM_DEVICE);
 
 		rcvd_pkts++;
@@ -5157,12 +5167,6 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 		mvpp2_rx_csum(port, rx_status, skb);
 
 		napi_gro_receive(&port->napi, skb);
-
-		err = mvpp2_rx_refill(port, bm_pool, bm, 0);
-		if (err) {
-			netdev_err(port->dev, "failed to refill BM pools\n");
-			rx_filled--;
-		}
 	}
 
 	if (rcvd_pkts) {
@@ -5176,7 +5180,7 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 
 	/* Update Rx queue management counters */
 	wmb();
-	mvpp2_rxq_status_update(port, rxq->id, rx_todo, rx_filled);
+	mvpp2_rxq_status_update(port, rxq->id, rx_done, rx_done);
 
 	return rx_todo;
 }
-- 
1.8.3.1

Re: [PATCH net 0/3] Marvell Armada 375 mvpp2 fixes

[David Miller]

From: Marcin Wojtas <mw@semihalf.com>
Date: Thu,  3 Dec 2015 15:20:48 +0100

> During my work on mvneta driver I revised mvpp2, and it occurred that the
> initial version of Marvell Armada 375 SoC comprised bugs around
> DMA-unmapping in both ingress and egress paths - not all buffers were
> umapped in TX path and none(!) in RX. Three patches that I send fix
> this situation.
> 
> Any feedback would be welcome.

Series applied, thank you.