[PATCHv2 net-next 00/11] net: mvpp2: misc improvements and preparation patches

[PATCHv2 net-next 00/11] net: mvpp2: misc improvements and preparation patches

[Thomas Petazzoni]

Hello,

This series contains a number of misc improvements and preparation
patches for an upcoming series that adds support for the new PPv2.2
network controller to the mvpp2 driver.

The most significant improvements are:

 - Switching to using build_skb(), which is necessary for the upcoming
   PPv2.2 support, but anyway a good improvement to the current mvpp2
   driver (supporting PPv2.1).

 - Making the driver build on 64-bit platforms.

Changes since v1:

 - This series is split as a separate series from the larger patch set
   adding support for PPv2.2 in the mvpp2 driver, as requested by
   David Miller.

 - Rebased on top of v4.10-rc1.

Thanks!

Thomas

Thomas Petazzoni (11):
  net: mvpp2: handle too large value handling in
    mvpp2_rx_pkts_coal_set()
  net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()
  net: mvpp2: release reference to txq_cpu[] entry after unmapping
  net: mvpp2: remove unused 'tx_skb' field of 'struct mvpp2_tx_queue'
  net: mvpp2: drop useless fields in mvpp2_bm_pool and related code
  net: mvpp2: simplify mvpp2_bm_bufs_add()
  net: mvpp2: remove unused register definitions
  net: mvpp2: fix indentation of MVPP2_EXT_GLOBAL_CTRL_DEFAULT
  net: mvpp2: simplify MVPP2_PRS_RI_* definitions
  net: mvpp2: switch to build_skb() in the RX path
  net: mvpp2: enable building on 64-bit platforms

 drivers/net/ethernet/marvell/Kconfig |   3 +-
 drivers/net/ethernet/marvell/mvpp2.c | 161 ++++++++++++++++++++---------------
 2 files changed, 93 insertions(+), 71 deletions(-)

-- 
2.7.4

[PATCHv2 net-next 00/11] net: mvpp2: misc improvements and preparation patches

[Thomas Petazzoni]

Hello,

This series contains a number of misc improvements and preparation
patches for an upcoming series that adds support for the new PPv2.2
network controller to the mvpp2 driver.

The most significant improvements are:

 - Switching to using build_skb(), which is necessary for the upcoming
   PPv2.2 support, but anyway a good improvement to the current mvpp2
   driver (supporting PPv2.1).

 - Making the driver build on 64-bit platforms.

Changes since v1:

 - This series is split as a separate series from the larger patch set
   adding support for PPv2.2 in the mvpp2 driver, as requested by
   David Miller.

 - Rebased on top of v4.10-rc1.

Thanks!

Thomas

Thomas Petazzoni (11):
  net: mvpp2: handle too large value handling in
    mvpp2_rx_pkts_coal_set()
  net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()
  net: mvpp2: release reference to txq_cpu[] entry after unmapping
  net: mvpp2: remove unused 'tx_skb' field of 'struct mvpp2_tx_queue'
  net: mvpp2: drop useless fields in mvpp2_bm_pool and related code
  net: mvpp2: simplify mvpp2_bm_bufs_add()
  net: mvpp2: remove unused register definitions
  net: mvpp2: fix indentation of MVPP2_EXT_GLOBAL_CTRL_DEFAULT
  net: mvpp2: simplify MVPP2_PRS_RI_* definitions
  net: mvpp2: switch to build_skb() in the RX path
  net: mvpp2: enable building on 64-bit platforms

 drivers/net/ethernet/marvell/Kconfig |   3 +-
 drivers/net/ethernet/marvell/mvpp2.c | 161 ++++++++++++++++++++---------------
 2 files changed, 93 insertions(+), 71 deletions(-)

-- 
2.7.4

[PATCHv2 net-next 01/11] net: mvpp2: handle too large value handling in mvpp2_rx_pkts_coal_set()

[Thomas Petazzoni]

Currently, mvpp2_rx_pkts_coal_set() does the following to avoid setting
a too large value for the RX coalescing by packet number:

  val = (pkts & MVPP2_OCCUPIED_THRESH_MASK);

This means that if you set a value that is slightly higher the the
maximum number of packets, you in fact get a very low value. It makes a
lot more sense to simply check if the value is too high, and if it's too
high, limit it to the maximum possible value.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 4fe430c..02d91e4 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -4381,11 +4381,11 @@ static void mvpp2_txp_max_tx_size_set(struct mvpp2_port *port)
 static void mvpp2_rx_pkts_coal_set(struct mvpp2_port *port,
 				   struct mvpp2_rx_queue *rxq, u32 pkts)
 {
-	u32 val;
+	if (pkts > MVPP2_OCCUPIED_THRESH_MASK)
+		pkts = MVPP2_OCCUPIED_THRESH_MASK;
 
-	val = (pkts & MVPP2_OCCUPIED_THRESH_MASK);
 	mvpp2_write(port->priv, MVPP2_RXQ_NUM_REG, rxq->id);
-	mvpp2_write(port->priv, MVPP2_RXQ_THRESH_REG, val);
+	mvpp2_write(port->priv, MVPP2_RXQ_THRESH_REG, pkts);
 
 	rxq->pkts_coal = pkts;
 }
-- 
2.7.4

[PATCHv2 net-next 01/11] net: mvpp2: handle too large value handling in mvpp2_rx_pkts_coal_set()

[Thomas Petazzoni]

Currently, mvpp2_rx_pkts_coal_set() does the following to avoid setting
a too large value for the RX coalescing by packet number:

  val = (pkts & MVPP2_OCCUPIED_THRESH_MASK);

This means that if you set a value that is slightly higher the the
maximum number of packets, you in fact get a very low value. It makes a
lot more sense to simply check if the value is too high, and if it's too
high, limit it to the maximum possible value.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 4fe430c..02d91e4 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -4381,11 +4381,11 @@ static void mvpp2_txp_max_tx_size_set(struct mvpp2_port *port)
 static void mvpp2_rx_pkts_coal_set(struct mvpp2_port *port,
 				   struct mvpp2_rx_queue *rxq, u32 pkts)
 {
-	u32 val;
+	if (pkts > MVPP2_OCCUPIED_THRESH_MASK)
+		pkts = MVPP2_OCCUPIED_THRESH_MASK;
 
-	val = (pkts & MVPP2_OCCUPIED_THRESH_MASK);
 	mvpp2_write(port->priv, MVPP2_RXQ_NUM_REG, rxq->id);
-	mvpp2_write(port->priv, MVPP2_RXQ_THRESH_REG, val);
+	mvpp2_write(port->priv, MVPP2_RXQ_THRESH_REG, pkts);
 
 	rxq->pkts_coal = pkts;
 }
-- 
2.7.4

[PATCHv2 net-next 02/11] net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()

[Thomas Petazzoni]

When configuring the MVPP2_ISR_RX_THRESHOLD_REG with the RX coalescing
time threshold, we do not check for the maximum allowed value supported
by the driver, which means we might overflow and use a bogus value. This
commit adds a check for this situation, and if a value higher than what
is supported by the hardware is provided, then we use the maximum value
supported by the hardware.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 02d91e4..a1ba89f 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -154,6 +154,7 @@
 
 /* Interrupt Cause and Mask registers */
 #define MVPP2_ISR_RX_THRESHOLD_REG(rxq)		(0x5200 + 4 * (rxq))
+#define     MVPP2_MAX_ISR_RX_THRESHOLD		0xfffff0
 #define MVPP2_ISR_RXQ_GROUP_REG(rxq)		(0x5400 + 4 * (rxq))
 #define MVPP2_ISR_ENABLE_REG(port)		(0x5420 + 4 * (port))
 #define     MVPP2_ISR_ENABLE_INTERRUPT(mask)	((mask) & 0xffff)
@@ -4397,6 +4398,12 @@ static void mvpp2_rx_time_coal_set(struct mvpp2_port *port,
 	u32 val;
 
 	val = (port->priv->tclk / USEC_PER_SEC) * usec;
+
+	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
+		val = MVPP2_MAX_ISR_RX_THRESHOLD;
+		usec = (val * USEC_PER_SEC) / port->priv->tclk;
+	}
+
 	mvpp2_write(port->priv, MVPP2_ISR_RX_THRESHOLD_REG(rxq->id), val);
 
 	rxq->time_coal = usec;
-- 
2.7.4

[PATCHv2 net-next 02/11] net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()

[Thomas Petazzoni]

When configuring the MVPP2_ISR_RX_THRESHOLD_REG with the RX coalescing
time threshold, we do not check for the maximum allowed value supported
by the driver, which means we might overflow and use a bogus value. This
commit adds a check for this situation, and if a value higher than what
is supported by the hardware is provided, then we use the maximum value
supported by the hardware.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 02d91e4..a1ba89f 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -154,6 +154,7 @@
 
 /* Interrupt Cause and Mask registers */
 #define MVPP2_ISR_RX_THRESHOLD_REG(rxq)		(0x5200 + 4 * (rxq))
+#define     MVPP2_MAX_ISR_RX_THRESHOLD		0xfffff0
 #define MVPP2_ISR_RXQ_GROUP_REG(rxq)		(0x5400 + 4 * (rxq))
 #define MVPP2_ISR_ENABLE_REG(port)		(0x5420 + 4 * (port))
 #define     MVPP2_ISR_ENABLE_INTERRUPT(mask)	((mask) & 0xffff)
@@ -4397,6 +4398,12 @@ static void mvpp2_rx_time_coal_set(struct mvpp2_port *port,
 	u32 val;
 
 	val = (port->priv->tclk / USEC_PER_SEC) * usec;
+
+	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
+		val = MVPP2_MAX_ISR_RX_THRESHOLD;
+		usec = (val * USEC_PER_SEC) / port->priv->tclk;
+	}
+
 	mvpp2_write(port->priv, MVPP2_ISR_RX_THRESHOLD_REG(rxq->id), val);
 
 	rxq->time_coal = usec;
-- 
2.7.4

[PATCHv2 net-next 03/11] net: mvpp2: release reference to txq_cpu[] entry after unmapping

[Thomas Petazzoni]

The mvpp2_txq_bufs_free() function is called upon TX completion to DMA
unmap TX buffers, and free the corresponding SKBs. It gets the
references to the SKB to free and the DMA buffer to unmap from a per-CPU
txq_pcpu data structure.

However, the code currently increments the pointer to the next entry
before doing the DMA unmap and freeing the SKB. It does not cause any
visible problem because for a given SKB the TX completion is guaranteed
to take place on the CPU where the TX was started. However, it is much
more logical to increment the pointer to the next entry once the current
entry has been completely unmapped/released.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index a1ba89f..d098c7b 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -4420,13 +4420,12 @@ static void mvpp2_txq_bufs_free(struct mvpp2_port *port,
 		struct mvpp2_txq_pcpu_buf *tx_buf =
 			txq_pcpu->buffs + txq_pcpu->txq_get_index;
 
-		mvpp2_txq_inc_get(txq_pcpu);
-
 		dma_unmap_single(port->dev->dev.parent, tx_buf->phys,
 				 tx_buf->size, DMA_TO_DEVICE);
-		if (!tx_buf->skb)
-			continue;
-		dev_kfree_skb_any(tx_buf->skb);
+		if (tx_buf->skb)
+			dev_kfree_skb_any(tx_buf->skb);
+
+		mvpp2_txq_inc_get(txq_pcpu);
 	}
 }
 
-- 
2.7.4

[PATCHv2 net-next 03/11] net: mvpp2: release reference to txq_cpu[] entry after unmapping

[Thomas Petazzoni]

The mvpp2_txq_bufs_free() function is called upon TX completion to DMA
unmap TX buffers, and free the corresponding SKBs. It gets the
references to the SKB to free and the DMA buffer to unmap from a per-CPU
txq_pcpu data structure.

However, the code currently increments the pointer to the next entry
before doing the DMA unmap and freeing the SKB. It does not cause any
visible problem because for a given SKB the TX completion is guaranteed
to take place on the CPU where the TX was started. However, it is much
more logical to increment the pointer to the next entry once the current
entry has been completely unmapped/released.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index a1ba89f..d098c7b 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -4420,13 +4420,12 @@ static void mvpp2_txq_bufs_free(struct mvpp2_port *port,
 		struct mvpp2_txq_pcpu_buf *tx_buf =
 			txq_pcpu->buffs + txq_pcpu->txq_get_index;
 
-		mvpp2_txq_inc_get(txq_pcpu);
-
 		dma_unmap_single(port->dev->dev.parent, tx_buf->phys,
 				 tx_buf->size, DMA_TO_DEVICE);
-		if (!tx_buf->skb)
-			continue;
-		dev_kfree_skb_any(tx_buf->skb);
+		if (tx_buf->skb)
+			dev_kfree_skb_any(tx_buf->skb);
+
+		mvpp2_txq_inc_get(txq_pcpu);
 	}
 }
 
-- 
2.7.4

[PATCHv2 net-next 04/11] net: mvpp2: remove unused 'tx_skb' field of 'struct mvpp2_tx_queue'

[Thomas Petazzoni]

This commit remove a field of 'struct mvpp2_tx_queue' that is not used
anywhere.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index d098c7b..6720cdac 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -823,9 +823,6 @@ struct mvpp2_tx_queue {
 	/* Per-CPU control of physical Tx queues */
 	struct mvpp2_txq_pcpu __percpu *pcpu;
 
-	/* Array of transmitted skb */
-	struct sk_buff **tx_skb;
-
 	u32 done_pkts_coal;
 
 	/* Virtual address of thex Tx DMA descriptors array */
-- 
2.7.4

[PATCHv2 net-next 04/11] net: mvpp2: remove unused 'tx_skb' field of 'struct mvpp2_tx_queue'

[Thomas Petazzoni]

This commit remove a field of 'struct mvpp2_tx_queue' that is not used
anywhere.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index d098c7b..6720cdac 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -823,9 +823,6 @@ struct mvpp2_tx_queue {
 	/* Per-CPU control of physical Tx queues */
 	struct mvpp2_txq_pcpu __percpu *pcpu;
 
-	/* Array of transmitted skb */
-	struct sk_buff **tx_skb;
-
 	u32 done_pkts_coal;
 
 	/* Virtual address of thex Tx DMA descriptors array */
-- 
2.7.4

[PATCHv2 net-next 05/11] net: mvpp2: drop useless fields in mvpp2_bm_pool and related code

[Thomas Petazzoni]

This commit drops dead code from the mvpp2 driver. The 'in_use' and
'in_use_thresh' fields of 'struct mvpp2_bm_pool' are
incremented/decremented/initialized in various places. But they are only
used in one place:

       if (is_recycle &&
           (atomic_read(&bm_pool->in_use) < bm_pool->in_use_thresh))
               return 0;

However 'is_recycle', passed as argument to mvpp2_rx_refill() is always
false. So in fact, this code is never reached, and the 'is_recycle'
argument is useless. So let's drop this code.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 18 +++---------------
 1 file changed, 3 insertions(+), 15 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 6720cdac..bfa9f77 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -930,10 +930,6 @@ struct mvpp2_bm_pool {
 
 	/* Ports using BM pool */
 	u32 port_map;
-
-	/* Occupied buffers indicator */
-	atomic_t in_use;
-	int in_use_thresh;
 };
 
 struct mvpp2_buff_hdr {
@@ -3399,7 +3395,6 @@ static int mvpp2_bm_pool_create(struct platform_device *pdev,
 	bm_pool->size = size;
 	bm_pool->pkt_size = 0;
 	bm_pool->buf_num = 0;
-	atomic_set(&bm_pool->in_use, 0);
 
 	return 0;
 }
@@ -3656,7 +3651,6 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 
 	/* Update BM driver with number of buffers added to pool */
 	bm_pool->buf_num += i;
-	bm_pool->in_use_thresh = bm_pool->buf_num / 4;
 
 	netdev_dbg(port->dev,
 		   "%s pool %d: pkt_size=%4d, buf_size=%4d, total_size=%4d\n",
@@ -4997,23 +4991,18 @@ static void mvpp2_rx_csum(struct mvpp2_port *port, u32 status,
 
 /* Reuse skb if possible, or allocate a new skb and add it to BM pool */
 static int mvpp2_rx_refill(struct mvpp2_port *port,
-			   struct mvpp2_bm_pool *bm_pool,
-			   u32 bm, int is_recycle)
+			   struct mvpp2_bm_pool *bm_pool, u32 bm)
 {
 	struct sk_buff *skb;
 	dma_addr_t phys_addr;
 
-	if (is_recycle &&
-	    (atomic_read(&bm_pool->in_use) < bm_pool->in_use_thresh))
-		return 0;
-
 	/* No recycle or too many buffers are in use, so allocate a new skb */
 	skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_ATOMIC);
 	if (!skb)
 		return -ENOMEM;
 
 	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)skb);
-	atomic_dec(&bm_pool->in_use);
+
 	return 0;
 }
 
@@ -5139,7 +5128,7 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 
 		skb = (struct sk_buff *)rx_desc->buf_cookie;
 
-		err = mvpp2_rx_refill(port, bm_pool, bm, 0);
+		err = mvpp2_rx_refill(port, bm_pool, bm);
 		if (err) {
 			netdev_err(port->dev, "failed to refill BM pools\n");
 			goto err_drop_frame;
@@ -5150,7 +5139,6 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 
 		rcvd_pkts++;
 		rcvd_bytes += rx_bytes;
-		atomic_inc(&bm_pool->in_use);
 
 		skb_reserve(skb, MVPP2_MH_SIZE);
 		skb_put(skb, rx_bytes);
-- 
2.7.4

[PATCHv2 net-next 05/11] net: mvpp2: drop useless fields in mvpp2_bm_pool and related code

[Thomas Petazzoni]

This commit drops dead code from the mvpp2 driver. The 'in_use' and
'in_use_thresh' fields of 'struct mvpp2_bm_pool' are
incremented/decremented/initialized in various places. But they are only
used in one place:

       if (is_recycle &&
           (atomic_read(&bm_pool->in_use) < bm_pool->in_use_thresh))
               return 0;

However 'is_recycle', passed as argument to mvpp2_rx_refill() is always
false. So in fact, this code is never reached, and the 'is_recycle'
argument is useless. So let's drop this code.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 18 +++---------------
 1 file changed, 3 insertions(+), 15 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 6720cdac..bfa9f77 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -930,10 +930,6 @@ struct mvpp2_bm_pool {
 
 	/* Ports using BM pool */
 	u32 port_map;
-
-	/* Occupied buffers indicator */
-	atomic_t in_use;
-	int in_use_thresh;
 };
 
 struct mvpp2_buff_hdr {
@@ -3399,7 +3395,6 @@ static int mvpp2_bm_pool_create(struct platform_device *pdev,
 	bm_pool->size = size;
 	bm_pool->pkt_size = 0;
 	bm_pool->buf_num = 0;
-	atomic_set(&bm_pool->in_use, 0);
 
 	return 0;
 }
@@ -3656,7 +3651,6 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 
 	/* Update BM driver with number of buffers added to pool */
 	bm_pool->buf_num += i;
-	bm_pool->in_use_thresh = bm_pool->buf_num / 4;
 
 	netdev_dbg(port->dev,
 		   "%s pool %d: pkt_size=%4d, buf_size=%4d, total_size=%4d\n",
@@ -4997,23 +4991,18 @@ static void mvpp2_rx_csum(struct mvpp2_port *port, u32 status,
 
 /* Reuse skb if possible, or allocate a new skb and add it to BM pool */
 static int mvpp2_rx_refill(struct mvpp2_port *port,
-			   struct mvpp2_bm_pool *bm_pool,
-			   u32 bm, int is_recycle)
+			   struct mvpp2_bm_pool *bm_pool, u32 bm)
 {
 	struct sk_buff *skb;
 	dma_addr_t phys_addr;
 
-	if (is_recycle &&
-	    (atomic_read(&bm_pool->in_use) < bm_pool->in_use_thresh))
-		return 0;
-
 	/* No recycle or too many buffers are in use, so allocate a new skb */
 	skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_ATOMIC);
 	if (!skb)
 		return -ENOMEM;
 
 	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)skb);
-	atomic_dec(&bm_pool->in_use);
+
 	return 0;
 }
 
@@ -5139,7 +5128,7 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 
 		skb = (struct sk_buff *)rx_desc->buf_cookie;
 
-		err = mvpp2_rx_refill(port, bm_pool, bm, 0);
+		err = mvpp2_rx_refill(port, bm_pool, bm);
 		if (err) {
 			netdev_err(port->dev, "failed to refill BM pools\n");
 			goto err_drop_frame;
@@ -5150,7 +5139,6 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 
 		rcvd_pkts++;
 		rcvd_bytes += rx_bytes;
-		atomic_inc(&bm_pool->in_use);
 
 		skb_reserve(skb, MVPP2_MH_SIZE);
 		skb_put(skb, rx_bytes);
-- 
2.7.4

[PATCHv2 net-next 06/11] net: mvpp2: simplify mvpp2_bm_bufs_add()

[Thomas Petazzoni]

The mvpp2_bm_bufs_add() currently creates a fake cookie by calling
mvpp2_bm_cookie_pool_set(), just to be able to call
mvpp2_pool_refill(). But all what mvpp2_pool_refill() does is extract
the pool ID from the cookie, and call mvpp2_bm_pool_put() with this ID.

Instead of doing this convoluted thing, just call mvpp2_bm_pool_put()
directly, since we have the BM pool ID.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index bfa9f77..8174f40 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -3626,7 +3626,6 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 {
 	struct sk_buff *skb;
 	int i, buf_size, total_size;
-	u32 bm;
 	dma_addr_t phys_addr;
 
 	buf_size = MVPP2_RX_BUF_SIZE(bm_pool->pkt_size);
@@ -3640,13 +3639,12 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 		return 0;
 	}
 
-	bm = mvpp2_bm_cookie_pool_set(0, bm_pool->id);
 	for (i = 0; i < buf_num; i++) {
 		skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_KERNEL);
 		if (!skb)
 			break;
 
-		mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)skb);
+		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)skb);
 	}
 
 	/* Update BM driver with number of buffers added to pool */
-- 
2.7.4

[PATCHv2 net-next 06/11] net: mvpp2: simplify mvpp2_bm_bufs_add()

[Thomas Petazzoni]

The mvpp2_bm_bufs_add() currently creates a fake cookie by calling
mvpp2_bm_cookie_pool_set(), just to be able to call
mvpp2_pool_refill(). But all what mvpp2_pool_refill() does is extract
the pool ID from the cookie, and call mvpp2_bm_pool_put() with this ID.

Instead of doing this convoluted thing, just call mvpp2_bm_pool_put()
directly, since we have the BM pool ID.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index bfa9f77..8174f40 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -3626,7 +3626,6 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 {
 	struct sk_buff *skb;
 	int i, buf_size, total_size;
-	u32 bm;
 	dma_addr_t phys_addr;
 
 	buf_size = MVPP2_RX_BUF_SIZE(bm_pool->pkt_size);
@@ -3640,13 +3639,12 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 		return 0;
 	}
 
-	bm = mvpp2_bm_cookie_pool_set(0, bm_pool->id);
 	for (i = 0; i < buf_num; i++) {
 		skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_KERNEL);
 		if (!skb)
 			break;
 
-		mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)skb);
+		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)skb);
 	}
 
 	/* Update BM driver with number of buffers added to pool */
-- 
2.7.4

[PATCHv2 net-next 07/11] net: mvpp2: remove unused register definitions

[Thomas Petazzoni]

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 8174f40..edffcc1 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -253,10 +253,6 @@
 #define MVPP2_SRC_ADDR_HIGH			0x28
 #define MVPP2_PHY_AN_CFG0_REG			0x34
 #define     MVPP2_PHY_AN_STOP_SMI0_MASK		BIT(7)
-#define MVPP2_MIB_COUNTERS_BASE(port)		(0x1000 + ((port) >> 1) * \
-						0x400 + (port) * 0x400)
-#define     MVPP2_MIB_LATE_COLLISION		0x7c
-#define MVPP2_ISR_SUM_MASK_REG			0x220c
 #define MVPP2_MNG_EXTENDED_GLOBAL_CTRL_REG	0x305c
 #define MVPP2_EXT_GLOBAL_CTRL_DEFAULT		0x27
 
-- 
2.7.4

[PATCHv2 net-next 07/11] net: mvpp2: remove unused register definitions

[Thomas Petazzoni]

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 8174f40..edffcc1 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -253,10 +253,6 @@
 #define MVPP2_SRC_ADDR_HIGH			0x28
 #define MVPP2_PHY_AN_CFG0_REG			0x34
 #define     MVPP2_PHY_AN_STOP_SMI0_MASK		BIT(7)
-#define MVPP2_MIB_COUNTERS_BASE(port)		(0x1000 + ((port) >> 1) * \
-						0x400 + (port) * 0x400)
-#define     MVPP2_MIB_LATE_COLLISION		0x7c
-#define MVPP2_ISR_SUM_MASK_REG			0x220c
 #define MVPP2_MNG_EXTENDED_GLOBAL_CTRL_REG	0x305c
 #define MVPP2_EXT_GLOBAL_CTRL_DEFAULT		0x27
 
-- 
2.7.4

[PATCHv2 net-next 08/11] net: mvpp2: fix indentation of MVPP2_EXT_GLOBAL_CTRL_DEFAULT

[Thomas Petazzoni]

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index edffcc1..36c73dc 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -254,7 +254,7 @@
 #define MVPP2_PHY_AN_CFG0_REG			0x34
 #define     MVPP2_PHY_AN_STOP_SMI0_MASK		BIT(7)
 #define MVPP2_MNG_EXTENDED_GLOBAL_CTRL_REG	0x305c
-#define MVPP2_EXT_GLOBAL_CTRL_DEFAULT		0x27
+#define     MVPP2_EXT_GLOBAL_CTRL_DEFAULT	0x27
 
 /* Per-port registers */
 #define MVPP2_GMAC_CTRL_0_REG			0x0
-- 
2.7.4

[PATCHv2 net-next 08/11] net: mvpp2: fix indentation of MVPP2_EXT_GLOBAL_CTRL_DEFAULT

[Thomas Petazzoni]

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index edffcc1..36c73dc 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -254,7 +254,7 @@
 #define MVPP2_PHY_AN_CFG0_REG			0x34
 #define     MVPP2_PHY_AN_STOP_SMI0_MASK		BIT(7)
 #define MVPP2_MNG_EXTENDED_GLOBAL_CTRL_REG	0x305c
-#define MVPP2_EXT_GLOBAL_CTRL_DEFAULT		0x27
+#define     MVPP2_EXT_GLOBAL_CTRL_DEFAULT	0x27
 
 /* Per-port registers */
 #define MVPP2_GMAC_CTRL_0_REG			0x0
-- 
2.7.4

[PATCHv2 net-next 09/11] net: mvpp2: simplify MVPP2_PRS_RI_* definitions

[Thomas Petazzoni]

Some of the MVPP2_PRS_RI_* definitions use the ~(value) syntax, which
doesn't compile nicely on 64-bit. Moreover, those definitions are in
fact unneeded, since they are always used in combination with a bit
mask that ensures only the appropriate bits are modified.

Therefore, such definitions should just be set to 0x0. For example:

 #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
 #define MVPP2_PRS_RI_L2_UCAST                  ~(BIT(9) | BIT(10))
 #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
 #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

becomes

 #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
 #define MVPP2_PRS_RI_L2_UCAST                  0x0
 #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
 #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

Because the values (MVPP2_PRS_RI_L2_UCAST, MVPP2_PRS_RI_L2_MCAST and
MVPP2_PRS_RI_L2_BCAST) are always applied with
MVPP2_PRS_RI_L2_CAST_MASK, and therefore there is no need for
MVPP2_PRS_RI_L2_UCAST to be defined as ~(BIT(9) | BIT(10)).

It fixes the following warnings when building the driver on a 64-bit
platform (which is not possible as of this commit, but will be enabled
in a follow-up commit):

drivers/net/ethernet/marvell/mvpp2.c: In function ‘mvpp2_prs_mac_promisc_set’:
drivers/net/ethernet/marvell/mvpp2.c:524:33: warning: large integer implicitly truncated to unsigned type [-Woverflow]
 #define MVPP2_PRS_RI_L2_UCAST   ~(BIT(9) | BIT(10))
                                 ^
drivers/net/ethernet/marvell/mvpp2.c:1459:33: note: in expansion of macro ‘MVPP2_PRS_RI_L2_UCAST’
   mvpp2_prs_sram_ri_update(&pe, MVPP2_PRS_RI_L2_UCAST,

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 36c73dc..58cf9af 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -511,19 +511,19 @@ enum mvpp2_tag_type {
 #define MVPP2_PRS_RI_MAC_ME_MASK		0x1
 #define MVPP2_PRS_RI_DSA_MASK			0x2
 #define MVPP2_PRS_RI_VLAN_MASK			0xc
-#define MVPP2_PRS_RI_VLAN_NONE			~(BIT(2) | BIT(3))
+#define MVPP2_PRS_RI_VLAN_NONE			0x0
 #define MVPP2_PRS_RI_VLAN_SINGLE		BIT(2)
 #define MVPP2_PRS_RI_VLAN_DOUBLE		BIT(3)
 #define MVPP2_PRS_RI_VLAN_TRIPLE		(BIT(2) | BIT(3))
 #define MVPP2_PRS_RI_CPU_CODE_MASK		0x70
 #define MVPP2_PRS_RI_CPU_CODE_RX_SPEC		BIT(4)
 #define MVPP2_PRS_RI_L2_CAST_MASK		0x600
-#define MVPP2_PRS_RI_L2_UCAST			~(BIT(9) | BIT(10))
+#define MVPP2_PRS_RI_L2_UCAST			0x0
 #define MVPP2_PRS_RI_L2_MCAST			BIT(9)
 #define MVPP2_PRS_RI_L2_BCAST			BIT(10)
 #define MVPP2_PRS_RI_PPPOE_MASK			0x800
 #define MVPP2_PRS_RI_L3_PROTO_MASK		0x7000
-#define MVPP2_PRS_RI_L3_UN			~(BIT(12) | BIT(13) | BIT(14))
+#define MVPP2_PRS_RI_L3_UN			0x0
 #define MVPP2_PRS_RI_L3_IP4			BIT(12)
 #define MVPP2_PRS_RI_L3_IP4_OPT			BIT(13)
 #define MVPP2_PRS_RI_L3_IP4_OTHER		(BIT(12) | BIT(13))
@@ -531,7 +531,7 @@ enum mvpp2_tag_type {
 #define MVPP2_PRS_RI_L3_IP6_EXT			(BIT(12) | BIT(14))
 #define MVPP2_PRS_RI_L3_ARP			(BIT(13) | BIT(14))
 #define MVPP2_PRS_RI_L3_ADDR_MASK		0x18000
-#define MVPP2_PRS_RI_L3_UCAST			~(BIT(15) | BIT(16))
+#define MVPP2_PRS_RI_L3_UCAST			0x0
 #define MVPP2_PRS_RI_L3_MCAST			BIT(15)
 #define MVPP2_PRS_RI_L3_BCAST			(BIT(15) | BIT(16))
 #define MVPP2_PRS_RI_IP_FRAG_MASK		0x20000
-- 
2.7.4

[PATCHv2 net-next 09/11] net: mvpp2: simplify MVPP2_PRS_RI_* definitions

[Thomas Petazzoni]

Some of the MVPP2_PRS_RI_* definitions use the ~(value) syntax, which
doesn't compile nicely on 64-bit. Moreover, those definitions are in
fact unneeded, since they are always used in combination with a bit
mask that ensures only the appropriate bits are modified.

Therefore, such definitions should just be set to 0x0. For example:

 #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
 #define MVPP2_PRS_RI_L2_UCAST                  ~(BIT(9) | BIT(10))
 #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
 #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

becomes

 #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
 #define MVPP2_PRS_RI_L2_UCAST                  0x0
 #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
 #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

Because the values (MVPP2_PRS_RI_L2_UCAST, MVPP2_PRS_RI_L2_MCAST and
MVPP2_PRS_RI_L2_BCAST) are always applied with
MVPP2_PRS_RI_L2_CAST_MASK, and therefore there is no need for
MVPP2_PRS_RI_L2_UCAST to be defined as ~(BIT(9) | BIT(10)).

It fixes the following warnings when building the driver on a 64-bit
platform (which is not possible as of this commit, but will be enabled
in a follow-up commit):

drivers/net/ethernet/marvell/mvpp2.c: In function ?mvpp2_prs_mac_promisc_set?:
drivers/net/ethernet/marvell/mvpp2.c:524:33: warning: large integer implicitly truncated to unsigned type [-Woverflow]
 #define MVPP2_PRS_RI_L2_UCAST   ~(BIT(9) | BIT(10))
                                 ^
drivers/net/ethernet/marvell/mvpp2.c:1459:33: note: in expansion of macro ?MVPP2_PRS_RI_L2_UCAST?
   mvpp2_prs_sram_ri_update(&pe, MVPP2_PRS_RI_L2_UCAST,

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 36c73dc..58cf9af 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -511,19 +511,19 @@ enum mvpp2_tag_type {
 #define MVPP2_PRS_RI_MAC_ME_MASK		0x1
 #define MVPP2_PRS_RI_DSA_MASK			0x2
 #define MVPP2_PRS_RI_VLAN_MASK			0xc
-#define MVPP2_PRS_RI_VLAN_NONE			~(BIT(2) | BIT(3))
+#define MVPP2_PRS_RI_VLAN_NONE			0x0
 #define MVPP2_PRS_RI_VLAN_SINGLE		BIT(2)
 #define MVPP2_PRS_RI_VLAN_DOUBLE		BIT(3)
 #define MVPP2_PRS_RI_VLAN_TRIPLE		(BIT(2) | BIT(3))
 #define MVPP2_PRS_RI_CPU_CODE_MASK		0x70
 #define MVPP2_PRS_RI_CPU_CODE_RX_SPEC		BIT(4)
 #define MVPP2_PRS_RI_L2_CAST_MASK		0x600
-#define MVPP2_PRS_RI_L2_UCAST			~(BIT(9) | BIT(10))
+#define MVPP2_PRS_RI_L2_UCAST			0x0
 #define MVPP2_PRS_RI_L2_MCAST			BIT(9)
 #define MVPP2_PRS_RI_L2_BCAST			BIT(10)
 #define MVPP2_PRS_RI_PPPOE_MASK			0x800
 #define MVPP2_PRS_RI_L3_PROTO_MASK		0x7000
-#define MVPP2_PRS_RI_L3_UN			~(BIT(12) | BIT(13) | BIT(14))
+#define MVPP2_PRS_RI_L3_UN			0x0
 #define MVPP2_PRS_RI_L3_IP4			BIT(12)
 #define MVPP2_PRS_RI_L3_IP4_OPT			BIT(13)
 #define MVPP2_PRS_RI_L3_IP4_OTHER		(BIT(12) | BIT(13))
@@ -531,7 +531,7 @@ enum mvpp2_tag_type {
 #define MVPP2_PRS_RI_L3_IP6_EXT			(BIT(12) | BIT(14))
 #define MVPP2_PRS_RI_L3_ARP			(BIT(13) | BIT(14))
 #define MVPP2_PRS_RI_L3_ADDR_MASK		0x18000
-#define MVPP2_PRS_RI_L3_UCAST			~(BIT(15) | BIT(16))
+#define MVPP2_PRS_RI_L3_UCAST			0x0
 #define MVPP2_PRS_RI_L3_MCAST			BIT(15)
 #define MVPP2_PRS_RI_L3_BCAST			(BIT(15) | BIT(16))
 #define MVPP2_PRS_RI_IP_FRAG_MASK		0x20000
-- 
2.7.4

[PATCHv2 net-next 10/11] net: mvpp2: switch to build_skb() in the RX path

[Thomas Petazzoni]

This commit adapts the mvpp2 RX path to use the build_skb() method. Not
only build_skb() is now the recommended mechanism, but it also
simplifies the addition of support for the PPv2.2 variant.

Indeed, without build_skb(), we have to keep track for each RX
descriptor of the physical address of the packet buffer, and the virtual
address of the SKB. However, in PPv2.2 running on 64 bits platform,
there is not enough space in the descriptor to store the virtual address
of the SKB. So having to take care only of the address of the packet
buffer, and building the SKB upon reception helps in supporting PPv2.2.

The implementation is fairly straightforward:

 - mvpp2_skb_alloc() is renamed to mvpp2_buf_alloc() and no longer
   allocates a SKB. Instead, it allocates a buffer using the new
   mvpp2_frag_alloc() function, with enough space for the data and SKB.

 - The initialization of the RX buffers in mvpp2_bm_bufs_add() as well
   as the refill of the RX buffers in mvpp2_rx_refill() is adjusted
   accordingly.

 - Finally, the mvpp2_rx() is modified to use build_skb().

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 77 +++++++++++++++++++++++++-----------
 1 file changed, 55 insertions(+), 22 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 58cf9af..1e08e8f 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -918,6 +918,7 @@ struct mvpp2_bm_pool {
 	int buf_size;
 	/* Packet size */
 	int pkt_size;
+	int frag_size;
 
 	/* BPPE virtual base address */
 	u32 *virt_addr;
@@ -3354,6 +3355,22 @@ static void mvpp2_cls_oversize_rxq_set(struct mvpp2_port *port)
 	mvpp2_write(port->priv, MVPP2_CLS_SWFWD_PCTRL_REG, val);
 }
 
+static void *mvpp2_frag_alloc(const struct mvpp2_bm_pool *pool)
+{
+	if (likely(pool->frag_size <= PAGE_SIZE))
+		return netdev_alloc_frag(pool->frag_size);
+	else
+		return kmalloc(pool->frag_size, GFP_ATOMIC);
+}
+
+static void mvpp2_frag_free(const struct mvpp2_bm_pool *pool, void *data)
+{
+	if (likely(pool->frag_size <= PAGE_SIZE))
+		skb_free_frag(data);
+	else
+		kfree(data);
+}
+
 /* Buffer Manager configuration routines */
 
 /* Create pool */
@@ -3428,7 +3445,8 @@ static void mvpp2_bm_bufs_free(struct device *dev, struct mvpp2 *priv,
 
 		if (!vaddr)
 			break;
-		dev_kfree_skb_any((struct sk_buff *)vaddr);
+
+		mvpp2_frag_free(bm_pool, (void *)vaddr);
 	}
 
 	/* Update BM driver with number of buffers removed from pool */
@@ -3542,29 +3560,28 @@ static void mvpp2_rxq_short_pool_set(struct mvpp2_port *port,
 	mvpp2_write(port->priv, MVPP2_RXQ_CONFIG_REG(prxq), val);
 }
 
-/* Allocate skb for BM pool */
-static struct sk_buff *mvpp2_skb_alloc(struct mvpp2_port *port,
-				       struct mvpp2_bm_pool *bm_pool,
-				       dma_addr_t *buf_phys_addr,
-				       gfp_t gfp_mask)
+static void *mvpp2_buf_alloc(struct mvpp2_port *port,
+			     struct mvpp2_bm_pool *bm_pool,
+			     dma_addr_t *buf_phys_addr,
+			     gfp_t gfp_mask)
 {
-	struct sk_buff *skb;
 	dma_addr_t phys_addr;
+	void *data;
 
-	skb = __dev_alloc_skb(bm_pool->pkt_size, gfp_mask);
-	if (!skb)
+	data = mvpp2_frag_alloc(bm_pool);
+	if (!data)
 		return NULL;
 
-	phys_addr = dma_map_single(port->dev->dev.parent, skb->head,
+	phys_addr = dma_map_single(port->dev->dev.parent, data,
 				   MVPP2_RX_BUF_SIZE(bm_pool->pkt_size),
 				    DMA_FROM_DEVICE);
 	if (unlikely(dma_mapping_error(port->dev->dev.parent, phys_addr))) {
-		dev_kfree_skb_any(skb);
+		mvpp2_frag_free(bm_pool, data);
 		return NULL;
 	}
 	*buf_phys_addr = phys_addr;
 
-	return skb;
+	return data;
 }
 
 /* Set pool number in a BM cookie */
@@ -3620,9 +3637,9 @@ static void mvpp2_pool_refill(struct mvpp2_port *port, u32 bm,
 static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 			     struct mvpp2_bm_pool *bm_pool, int buf_num)
 {
-	struct sk_buff *skb;
 	int i, buf_size, total_size;
 	dma_addr_t phys_addr;
+	void *buf;
 
 	buf_size = MVPP2_RX_BUF_SIZE(bm_pool->pkt_size);
 	total_size = MVPP2_RX_TOTAL_SIZE(buf_size);
@@ -3636,11 +3653,11 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 	}
 
 	for (i = 0; i < buf_num; i++) {
-		skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_KERNEL);
-		if (!skb)
+		buf = mvpp2_buf_alloc(port, bm_pool, &phys_addr, GFP_KERNEL);
+		if (!buf)
 			break;
 
-		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)skb);
+		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)buf);
 	}
 
 	/* Update BM driver with number of buffers added to pool */
@@ -3696,6 +3713,9 @@ mvpp2_bm_pool_use(struct mvpp2_port *port, int pool, enum mvpp2_bm_type type,
 					   port->priv, new_pool);
 
 		new_pool->pkt_size = pkt_size;
+		new_pool->frag_size =
+			SKB_DATA_ALIGN(MVPP2_RX_BUF_SIZE(pkt_size)) +
+			MVPP2_SKB_SHINFO_SIZE;
 
 		/* Allocate buffers for this pool */
 		num = mvpp2_bm_bufs_add(port, new_pool, pkts_num);
@@ -4987,15 +5007,15 @@ static void mvpp2_rx_csum(struct mvpp2_port *port, u32 status,
 static int mvpp2_rx_refill(struct mvpp2_port *port,
 			   struct mvpp2_bm_pool *bm_pool, u32 bm)
 {
-	struct sk_buff *skb;
 	dma_addr_t phys_addr;
+	void *buf;
 
 	/* No recycle or too many buffers are in use, so allocate a new skb */
-	skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_ATOMIC);
-	if (!skb)
+	buf = mvpp2_buf_alloc(port, bm_pool, &phys_addr, GFP_ATOMIC);
+	if (!buf)
 		return -ENOMEM;
 
-	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)skb);
+	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)buf);
 
 	return 0;
 }
@@ -5087,14 +5107,17 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 		struct mvpp2_rx_desc *rx_desc = mvpp2_rxq_next_desc_get(rxq);
 		struct mvpp2_bm_pool *bm_pool;
 		struct sk_buff *skb;
+		unsigned int frag_size;
 		dma_addr_t phys_addr;
 		u32 bm, rx_status;
 		int pool, rx_bytes, err;
+		void *data;
 
 		rx_done++;
 		rx_status = rx_desc->status;
 		rx_bytes = rx_desc->data_size - MVPP2_MH_SIZE;
 		phys_addr = rx_desc->buf_phys_addr;
+		data = (void *)rx_desc->buf_cookie;
 
 		bm = mvpp2_bm_cookie_build(rx_desc);
 		pool = mvpp2_bm_cookie_pool_get(bm);
@@ -5115,12 +5138,22 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 			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;
 		}
 
-		skb = (struct sk_buff *)rx_desc->buf_cookie;
+		if (bm_pool->frag_size > PAGE_SIZE)
+			frag_size = 0;
+		else
+			frag_size = bm_pool->frag_size;
+
+		skb = build_skb(data, frag_size);
+		if (!skb) {
+			netdev_warn(port->dev, "skb build failed\n");
+			goto err_drop_frame;
+		}
 
 		err = mvpp2_rx_refill(port, bm_pool, bm);
 		if (err) {
@@ -5134,7 +5167,7 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 		rcvd_pkts++;
 		rcvd_bytes += rx_bytes;
 
-		skb_reserve(skb, MVPP2_MH_SIZE);
+		skb_reserve(skb, MVPP2_MH_SIZE + NET_SKB_PAD);
 		skb_put(skb, rx_bytes);
 		skb->protocol = eth_type_trans(skb, dev);
 		mvpp2_rx_csum(port, rx_status, skb);
-- 
2.7.4

[PATCHv2 net-next 10/11] net: mvpp2: switch to build_skb() in the RX path

[Thomas Petazzoni]

This commit adapts the mvpp2 RX path to use the build_skb() method. Not
only build_skb() is now the recommended mechanism, but it also
simplifies the addition of support for the PPv2.2 variant.

Indeed, without build_skb(), we have to keep track for each RX
descriptor of the physical address of the packet buffer, and the virtual
address of the SKB. However, in PPv2.2 running on 64 bits platform,
there is not enough space in the descriptor to store the virtual address
of the SKB. So having to take care only of the address of the packet
buffer, and building the SKB upon reception helps in supporting PPv2.2.

The implementation is fairly straightforward:

 - mvpp2_skb_alloc() is renamed to mvpp2_buf_alloc() and no longer
   allocates a SKB. Instead, it allocates a buffer using the new
   mvpp2_frag_alloc() function, with enough space for the data and SKB.

 - The initialization of the RX buffers in mvpp2_bm_bufs_add() as well
   as the refill of the RX buffers in mvpp2_rx_refill() is adjusted
   accordingly.

 - Finally, the mvpp2_rx() is modified to use build_skb().

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/mvpp2.c | 77 +++++++++++++++++++++++++-----------
 1 file changed, 55 insertions(+), 22 deletions(-)

diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 58cf9af..1e08e8f 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -918,6 +918,7 @@ struct mvpp2_bm_pool {
 	int buf_size;
 	/* Packet size */
 	int pkt_size;
+	int frag_size;
 
 	/* BPPE virtual base address */
 	u32 *virt_addr;
@@ -3354,6 +3355,22 @@ static void mvpp2_cls_oversize_rxq_set(struct mvpp2_port *port)
 	mvpp2_write(port->priv, MVPP2_CLS_SWFWD_PCTRL_REG, val);
 }
 
+static void *mvpp2_frag_alloc(const struct mvpp2_bm_pool *pool)
+{
+	if (likely(pool->frag_size <= PAGE_SIZE))
+		return netdev_alloc_frag(pool->frag_size);
+	else
+		return kmalloc(pool->frag_size, GFP_ATOMIC);
+}
+
+static void mvpp2_frag_free(const struct mvpp2_bm_pool *pool, void *data)
+{
+	if (likely(pool->frag_size <= PAGE_SIZE))
+		skb_free_frag(data);
+	else
+		kfree(data);
+}
+
 /* Buffer Manager configuration routines */
 
 /* Create pool */
@@ -3428,7 +3445,8 @@ static void mvpp2_bm_bufs_free(struct device *dev, struct mvpp2 *priv,
 
 		if (!vaddr)
 			break;
-		dev_kfree_skb_any((struct sk_buff *)vaddr);
+
+		mvpp2_frag_free(bm_pool, (void *)vaddr);
 	}
 
 	/* Update BM driver with number of buffers removed from pool */
@@ -3542,29 +3560,28 @@ static void mvpp2_rxq_short_pool_set(struct mvpp2_port *port,
 	mvpp2_write(port->priv, MVPP2_RXQ_CONFIG_REG(prxq), val);
 }
 
-/* Allocate skb for BM pool */
-static struct sk_buff *mvpp2_skb_alloc(struct mvpp2_port *port,
-				       struct mvpp2_bm_pool *bm_pool,
-				       dma_addr_t *buf_phys_addr,
-				       gfp_t gfp_mask)
+static void *mvpp2_buf_alloc(struct mvpp2_port *port,
+			     struct mvpp2_bm_pool *bm_pool,
+			     dma_addr_t *buf_phys_addr,
+			     gfp_t gfp_mask)
 {
-	struct sk_buff *skb;
 	dma_addr_t phys_addr;
+	void *data;
 
-	skb = __dev_alloc_skb(bm_pool->pkt_size, gfp_mask);
-	if (!skb)
+	data = mvpp2_frag_alloc(bm_pool);
+	if (!data)
 		return NULL;
 
-	phys_addr = dma_map_single(port->dev->dev.parent, skb->head,
+	phys_addr = dma_map_single(port->dev->dev.parent, data,
 				   MVPP2_RX_BUF_SIZE(bm_pool->pkt_size),
 				    DMA_FROM_DEVICE);
 	if (unlikely(dma_mapping_error(port->dev->dev.parent, phys_addr))) {
-		dev_kfree_skb_any(skb);
+		mvpp2_frag_free(bm_pool, data);
 		return NULL;
 	}
 	*buf_phys_addr = phys_addr;
 
-	return skb;
+	return data;
 }
 
 /* Set pool number in a BM cookie */
@@ -3620,9 +3637,9 @@ static void mvpp2_pool_refill(struct mvpp2_port *port, u32 bm,
 static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 			     struct mvpp2_bm_pool *bm_pool, int buf_num)
 {
-	struct sk_buff *skb;
 	int i, buf_size, total_size;
 	dma_addr_t phys_addr;
+	void *buf;
 
 	buf_size = MVPP2_RX_BUF_SIZE(bm_pool->pkt_size);
 	total_size = MVPP2_RX_TOTAL_SIZE(buf_size);
@@ -3636,11 +3653,11 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 	}
 
 	for (i = 0; i < buf_num; i++) {
-		skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_KERNEL);
-		if (!skb)
+		buf = mvpp2_buf_alloc(port, bm_pool, &phys_addr, GFP_KERNEL);
+		if (!buf)
 			break;
 
-		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)skb);
+		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)buf);
 	}
 
 	/* Update BM driver with number of buffers added to pool */
@@ -3696,6 +3713,9 @@ mvpp2_bm_pool_use(struct mvpp2_port *port, int pool, enum mvpp2_bm_type type,
 					   port->priv, new_pool);
 
 		new_pool->pkt_size = pkt_size;
+		new_pool->frag_size =
+			SKB_DATA_ALIGN(MVPP2_RX_BUF_SIZE(pkt_size)) +
+			MVPP2_SKB_SHINFO_SIZE;
 
 		/* Allocate buffers for this pool */
 		num = mvpp2_bm_bufs_add(port, new_pool, pkts_num);
@@ -4987,15 +5007,15 @@ static void mvpp2_rx_csum(struct mvpp2_port *port, u32 status,
 static int mvpp2_rx_refill(struct mvpp2_port *port,
 			   struct mvpp2_bm_pool *bm_pool, u32 bm)
 {
-	struct sk_buff *skb;
 	dma_addr_t phys_addr;
+	void *buf;
 
 	/* No recycle or too many buffers are in use, so allocate a new skb */
-	skb = mvpp2_skb_alloc(port, bm_pool, &phys_addr, GFP_ATOMIC);
-	if (!skb)
+	buf = mvpp2_buf_alloc(port, bm_pool, &phys_addr, GFP_ATOMIC);
+	if (!buf)
 		return -ENOMEM;
 
-	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)skb);
+	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)buf);
 
 	return 0;
 }
@@ -5087,14 +5107,17 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 		struct mvpp2_rx_desc *rx_desc = mvpp2_rxq_next_desc_get(rxq);
 		struct mvpp2_bm_pool *bm_pool;
 		struct sk_buff *skb;
+		unsigned int frag_size;
 		dma_addr_t phys_addr;
 		u32 bm, rx_status;
 		int pool, rx_bytes, err;
+		void *data;
 
 		rx_done++;
 		rx_status = rx_desc->status;
 		rx_bytes = rx_desc->data_size - MVPP2_MH_SIZE;
 		phys_addr = rx_desc->buf_phys_addr;
+		data = (void *)rx_desc->buf_cookie;
 
 		bm = mvpp2_bm_cookie_build(rx_desc);
 		pool = mvpp2_bm_cookie_pool_get(bm);
@@ -5115,12 +5138,22 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 			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;
 		}
 
-		skb = (struct sk_buff *)rx_desc->buf_cookie;
+		if (bm_pool->frag_size > PAGE_SIZE)
+			frag_size = 0;
+		else
+			frag_size = bm_pool->frag_size;
+
+		skb = build_skb(data, frag_size);
+		if (!skb) {
+			netdev_warn(port->dev, "skb build failed\n");
+			goto err_drop_frame;
+		}
 
 		err = mvpp2_rx_refill(port, bm_pool, bm);
 		if (err) {
@@ -5134,7 +5167,7 @@ static int mvpp2_rx(struct mvpp2_port *port, int rx_todo,
 		rcvd_pkts++;
 		rcvd_bytes += rx_bytes;
 
-		skb_reserve(skb, MVPP2_MH_SIZE);
+		skb_reserve(skb, MVPP2_MH_SIZE + NET_SKB_PAD);
 		skb_put(skb, rx_bytes);
 		skb->protocol = eth_type_trans(skb, dev);
 		mvpp2_rx_csum(port, rx_status, skb);
-- 
2.7.4

[PATCHv2 net-next 11/11] net: mvpp2: enable building on 64-bit platforms

[Thomas Petazzoni]

The mvpp2 is going to be extended to support the Marvell Armada 7K/8K
platform, which is ARM64. As a preparation to this work, this commit
enables building the mvpp2 driver on ARM64, by:

 - Adjusting the Kconfig dependency

 - Fixing the types used in the driver so that they are 32/64-bits
   compliant. We use dma_addr_t for DMA addresses, and unsigned long
   for virtual addresses.

It is worth mentioning that after this commit, the driver is for now
still only used on 32-bits platforms.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/Kconfig |  3 +--
 drivers/net/ethernet/marvell/mvpp2.c | 29 +++++++++++++++++------------
 2 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/drivers/net/ethernet/marvell/Kconfig b/drivers/net/ethernet/marvell/Kconfig
index f4b7cf1..d2555e8b 100644
--- a/drivers/net/ethernet/marvell/Kconfig
+++ b/drivers/net/ethernet/marvell/Kconfig
@@ -83,9 +83,8 @@ config MVNETA_BM
 
 config MVPP2
 	tristate "Marvell Armada 375 network interface support"
-	depends on MACH_ARMADA_375 || COMPILE_TEST
+	depends on ARCH_MVEBU || COMPILE_TEST
 	depends on HAS_DMA
-	depends on !64BIT
 	select MVMDIO
 	---help---
 	  This driver supports the network interface units in the
diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 1e08e8f..2268808 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -3388,7 +3388,8 @@ static int mvpp2_bm_pool_create(struct platform_device *pdev,
 	if (!bm_pool->virt_addr)
 		return -ENOMEM;
 
-	if (!IS_ALIGNED((u32)bm_pool->virt_addr, MVPP2_BM_POOL_PTR_ALIGN)) {
+	if (!IS_ALIGNED((unsigned long)bm_pool->virt_addr,
+			MVPP2_BM_POOL_PTR_ALIGN)) {
 		dma_free_coherent(&pdev->dev, size_bytes, bm_pool->virt_addr,
 				  bm_pool->phys_addr);
 		dev_err(&pdev->dev, "BM pool %d is not %d bytes aligned\n",
@@ -3433,7 +3434,7 @@ static void mvpp2_bm_bufs_free(struct device *dev, struct mvpp2 *priv,
 
 	for (i = 0; i < bm_pool->buf_num; i++) {
 		dma_addr_t buf_phys_addr;
-		u32 vaddr;
+		unsigned long vaddr;
 
 		/* Get buffer virtual address (indirect access) */
 		buf_phys_addr = mvpp2_read(priv,
@@ -3596,14 +3597,15 @@ static inline u32 mvpp2_bm_cookie_pool_set(u32 cookie, int pool)
 }
 
 /* Get pool number from a BM cookie */
-static inline int mvpp2_bm_cookie_pool_get(u32 cookie)
+static inline int mvpp2_bm_cookie_pool_get(unsigned long cookie)
 {
 	return (cookie >> MVPP2_BM_COOKIE_POOL_OFFS) & 0xFF;
 }
 
 /* Release buffer to BM */
 static inline void mvpp2_bm_pool_put(struct mvpp2_port *port, int pool,
-				     u32 buf_phys_addr, u32 buf_virt_addr)
+				     dma_addr_t buf_phys_addr,
+				     unsigned long buf_virt_addr)
 {
 	mvpp2_write(port->priv, MVPP2_BM_VIRT_RLS_REG, buf_virt_addr);
 	mvpp2_write(port->priv, MVPP2_BM_PHY_RLS_REG(pool), buf_phys_addr);
@@ -3611,7 +3613,8 @@ static inline void mvpp2_bm_pool_put(struct mvpp2_port *port, int pool,
 
 /* Release multicast buffer */
 static void mvpp2_bm_pool_mc_put(struct mvpp2_port *port, int pool,
-				 u32 buf_phys_addr, u32 buf_virt_addr,
+				 dma_addr_t buf_phys_addr,
+				 unsigned long buf_virt_addr,
 				 int mc_id)
 {
 	u32 val = 0;
@@ -3626,7 +3629,8 @@ static void mvpp2_bm_pool_mc_put(struct mvpp2_port *port, int pool,
 
 /* Refill BM pool */
 static void mvpp2_pool_refill(struct mvpp2_port *port, u32 bm,
-			      u32 phys_addr, u32 cookie)
+			      dma_addr_t phys_addr,
+			      unsigned long cookie)
 {
 	int pool = mvpp2_bm_cookie_pool_get(bm);
 
@@ -3657,7 +3661,8 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 		if (!buf)
 			break;
 
-		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)buf);
+		mvpp2_bm_pool_put(port, bm_pool->id, phys_addr,
+				  (unsigned long)buf);
 	}
 
 	/* Update BM driver with number of buffers added to pool */
@@ -5015,7 +5020,7 @@ static int mvpp2_rx_refill(struct mvpp2_port *port,
 	if (!buf)
 		return -ENOMEM;
 
-	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)buf);
+	mvpp2_pool_refill(port, bm, phys_addr, (unsigned long)buf);
 
 	return 0;
 }
@@ -5057,10 +5062,10 @@ static void mvpp2_buff_hdr_rx(struct mvpp2_port *port,
 	struct mvpp2_buff_hdr *buff_hdr;
 	struct sk_buff *skb;
 	u32 rx_status = rx_desc->status;
-	u32 buff_phys_addr;
-	u32 buff_virt_addr;
-	u32 buff_phys_addr_next;
-	u32 buff_virt_addr_next;
+	dma_addr_t buff_phys_addr;
+	unsigned long buff_virt_addr;
+	dma_addr_t buff_phys_addr_next;
+	unsigned long buff_virt_addr_next;
 	int mc_id;
 	int pool_id;
 
-- 
2.7.4

[PATCHv2 net-next 11/11] net: mvpp2: enable building on 64-bit platforms

[Thomas Petazzoni]

The mvpp2 is going to be extended to support the Marvell Armada 7K/8K
platform, which is ARM64. As a preparation to this work, this commit
enables building the mvpp2 driver on ARM64, by:

 - Adjusting the Kconfig dependency

 - Fixing the types used in the driver so that they are 32/64-bits
   compliant. We use dma_addr_t for DMA addresses, and unsigned long
   for virtual addresses.

It is worth mentioning that after this commit, the driver is for now
still only used on 32-bits platforms.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 drivers/net/ethernet/marvell/Kconfig |  3 +--
 drivers/net/ethernet/marvell/mvpp2.c | 29 +++++++++++++++++------------
 2 files changed, 18 insertions(+), 14 deletions(-)

diff --git a/drivers/net/ethernet/marvell/Kconfig b/drivers/net/ethernet/marvell/Kconfig
index f4b7cf1..d2555e8b 100644
--- a/drivers/net/ethernet/marvell/Kconfig
+++ b/drivers/net/ethernet/marvell/Kconfig
@@ -83,9 +83,8 @@ config MVNETA_BM
 
 config MVPP2
 	tristate "Marvell Armada 375 network interface support"
-	depends on MACH_ARMADA_375 || COMPILE_TEST
+	depends on ARCH_MVEBU || COMPILE_TEST
 	depends on HAS_DMA
-	depends on !64BIT
 	select MVMDIO
 	---help---
 	  This driver supports the network interface units in the
diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
index 1e08e8f..2268808 100644
--- a/drivers/net/ethernet/marvell/mvpp2.c
+++ b/drivers/net/ethernet/marvell/mvpp2.c
@@ -3388,7 +3388,8 @@ static int mvpp2_bm_pool_create(struct platform_device *pdev,
 	if (!bm_pool->virt_addr)
 		return -ENOMEM;
 
-	if (!IS_ALIGNED((u32)bm_pool->virt_addr, MVPP2_BM_POOL_PTR_ALIGN)) {
+	if (!IS_ALIGNED((unsigned long)bm_pool->virt_addr,
+			MVPP2_BM_POOL_PTR_ALIGN)) {
 		dma_free_coherent(&pdev->dev, size_bytes, bm_pool->virt_addr,
 				  bm_pool->phys_addr);
 		dev_err(&pdev->dev, "BM pool %d is not %d bytes aligned\n",
@@ -3433,7 +3434,7 @@ static void mvpp2_bm_bufs_free(struct device *dev, struct mvpp2 *priv,
 
 	for (i = 0; i < bm_pool->buf_num; i++) {
 		dma_addr_t buf_phys_addr;
-		u32 vaddr;
+		unsigned long vaddr;
 
 		/* Get buffer virtual address (indirect access) */
 		buf_phys_addr = mvpp2_read(priv,
@@ -3596,14 +3597,15 @@ static inline u32 mvpp2_bm_cookie_pool_set(u32 cookie, int pool)
 }
 
 /* Get pool number from a BM cookie */
-static inline int mvpp2_bm_cookie_pool_get(u32 cookie)
+static inline int mvpp2_bm_cookie_pool_get(unsigned long cookie)
 {
 	return (cookie >> MVPP2_BM_COOKIE_POOL_OFFS) & 0xFF;
 }
 
 /* Release buffer to BM */
 static inline void mvpp2_bm_pool_put(struct mvpp2_port *port, int pool,
-				     u32 buf_phys_addr, u32 buf_virt_addr)
+				     dma_addr_t buf_phys_addr,
+				     unsigned long buf_virt_addr)
 {
 	mvpp2_write(port->priv, MVPP2_BM_VIRT_RLS_REG, buf_virt_addr);
 	mvpp2_write(port->priv, MVPP2_BM_PHY_RLS_REG(pool), buf_phys_addr);
@@ -3611,7 +3613,8 @@ static inline void mvpp2_bm_pool_put(struct mvpp2_port *port, int pool,
 
 /* Release multicast buffer */
 static void mvpp2_bm_pool_mc_put(struct mvpp2_port *port, int pool,
-				 u32 buf_phys_addr, u32 buf_virt_addr,
+				 dma_addr_t buf_phys_addr,
+				 unsigned long buf_virt_addr,
 				 int mc_id)
 {
 	u32 val = 0;
@@ -3626,7 +3629,8 @@ static void mvpp2_bm_pool_mc_put(struct mvpp2_port *port, int pool,
 
 /* Refill BM pool */
 static void mvpp2_pool_refill(struct mvpp2_port *port, u32 bm,
-			      u32 phys_addr, u32 cookie)
+			      dma_addr_t phys_addr,
+			      unsigned long cookie)
 {
 	int pool = mvpp2_bm_cookie_pool_get(bm);
 
@@ -3657,7 +3661,8 @@ static int mvpp2_bm_bufs_add(struct mvpp2_port *port,
 		if (!buf)
 			break;
 
-		mvpp2_bm_pool_put(port, bm_pool->id, (u32)phys_addr, (u32)buf);
+		mvpp2_bm_pool_put(port, bm_pool->id, phys_addr,
+				  (unsigned long)buf);
 	}
 
 	/* Update BM driver with number of buffers added to pool */
@@ -5015,7 +5020,7 @@ static int mvpp2_rx_refill(struct mvpp2_port *port,
 	if (!buf)
 		return -ENOMEM;
 
-	mvpp2_pool_refill(port, bm, (u32)phys_addr, (u32)buf);
+	mvpp2_pool_refill(port, bm, phys_addr, (unsigned long)buf);
 
 	return 0;
 }
@@ -5057,10 +5062,10 @@ static void mvpp2_buff_hdr_rx(struct mvpp2_port *port,
 	struct mvpp2_buff_hdr *buff_hdr;
 	struct sk_buff *skb;
 	u32 rx_status = rx_desc->status;
-	u32 buff_phys_addr;
-	u32 buff_virt_addr;
-	u32 buff_phys_addr_next;
-	u32 buff_virt_addr_next;
+	dma_addr_t buff_phys_addr;
+	unsigned long buff_virt_addr;
+	dma_addr_t buff_phys_addr_next;
+	unsigned long buff_virt_addr_next;
 	int mc_id;
 	int pool_id;
 
-- 
2.7.4

Re: [PATCHv2 net-next 00/11] net: mvpp2: misc improvements and preparation patches

[David Miller]

From: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Date: Wed, 28 Dec 2016 17:45:56 +0100

> This series contains a number of misc improvements and preparation
> patches for an upcoming series that adds support for the new PPv2.2
> network controller to the mvpp2 driver.
> 
> The most significant improvements are:
> 
>  - Switching to using build_skb(), which is necessary for the upcoming
>    PPv2.2 support, but anyway a good improvement to the current mvpp2
>    driver (supporting PPv2.1).
> 
>  - Making the driver build on 64-bit platforms.
> 
> Changes since v1:
> 
>  - This series is split as a separate series from the larger patch set
>    adding support for PPv2.2 in the mvpp2 driver, as requested by
>    David Miller.
> 
>  - Rebased on top of v4.10-rc1.

You still have warnings to fix for the 64-bit build:

drivers/net/ethernet/marvell/mvpp2.c: In function ‘mvpp2_rx’:
drivers/net/ethernet/marvell/mvpp2.c:5125:10: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
   data = (void *)rx_desc->buf_cookie;
          ^

[PATCHv2 net-next 00/11] net: mvpp2: misc improvements and preparation patches

[David Miller]

From: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Date: Wed, 28 Dec 2016 17:45:56 +0100

> This series contains a number of misc improvements and preparation
> patches for an upcoming series that adds support for the new PPv2.2
> network controller to the mvpp2 driver.
> 
> The most significant improvements are:
> 
>  - Switching to using build_skb(), which is necessary for the upcoming
>    PPv2.2 support, but anyway a good improvement to the current mvpp2
>    driver (supporting PPv2.1).
> 
>  - Making the driver build on 64-bit platforms.
> 
> Changes since v1:
> 
>  - This series is split as a separate series from the larger patch set
>    adding support for PPv2.2 in the mvpp2 driver, as requested by
>    David Miller.
> 
>  - Rebased on top of v4.10-rc1.

You still have warnings to fix for the 64-bit build:

drivers/net/ethernet/marvell/mvpp2.c: In function ?mvpp2_rx?:
drivers/net/ethernet/marvell/mvpp2.c:5125:10: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
   data = (void *)rx_desc->buf_cookie;
          ^

Re: [PATCHv2 net-next 02/11] net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()

[Russell King - ARM Linux]

On Wed, Dec 28, 2016 at 05:45:58PM +0100, Thomas Petazzoni wrote:
> When configuring the MVPP2_ISR_RX_THRESHOLD_REG with the RX coalescing
> time threshold, we do not check for the maximum allowed value supported
> by the driver, which means we might overflow and use a bogus value. This
> commit adds a check for this situation, and if a value higher than what
> is supported by the hardware is provided, then we use the maximum value
> supported by the hardware.
> 
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
> ---
>  drivers/net/ethernet/marvell/mvpp2.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
> index 02d91e4..a1ba89f 100644
> --- a/drivers/net/ethernet/marvell/mvpp2.c
> +++ b/drivers/net/ethernet/marvell/mvpp2.c
> @@ -154,6 +154,7 @@
>  
>  /* Interrupt Cause and Mask registers */
>  #define MVPP2_ISR_RX_THRESHOLD_REG(rxq)		(0x5200 + 4 * (rxq))
> +#define     MVPP2_MAX_ISR_RX_THRESHOLD		0xfffff0
>  #define MVPP2_ISR_RXQ_GROUP_REG(rxq)		(0x5400 + 4 * (rxq))
>  #define MVPP2_ISR_ENABLE_REG(port)		(0x5420 + 4 * (port))
>  #define     MVPP2_ISR_ENABLE_INTERRUPT(mask)	((mask) & 0xffff)
> @@ -4397,6 +4398,12 @@ static void mvpp2_rx_time_coal_set(struct mvpp2_port *port,
>  	u32 val;
>  
>  	val = (port->priv->tclk / USEC_PER_SEC) * usec;
> +
> +	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
> +		val = MVPP2_MAX_ISR_RX_THRESHOLD;
> +		usec = (val * USEC_PER_SEC) / port->priv->tclk;
> +	}
> +

Beware of rounding and overflow errors.  usec and val are u32's.

	MVPP2_MAX_ISR_RX_THRESHOLD = 16777200
	USEC_PER_SEC = 1000000

This equates to 0xF423F0BDC00 for the multiplication, which is a little
larger than 32-bit.  Assuming tclk is 166.666666MHz (as it was on Dove
- I don't know what it would be here) and 64-bit arithmetic, the maximum
value gives 100663us.

Passing that back into the function gives... 16710058, so the second time
around, we end up with a different setting (even though a change wasn't
requested.)

However, 100664 won't trigger your check, neither will values all the way
up to 101067 - the reason being that you're losing so much precision by
dividling the clock by USEC_PER_SEC first.  Only if it's a whole number
of MHz will you get away with that.

So, I'd suggest you switch to using 64 bit math here - it's not a fast
path.  Using bc to evaluate val = port->priv->tclk * usec / USEC_PER_SEC
gives:

	(166666666 * 100663 / 1000000)
	16777166

which is as close as you can come to the limit.

So, I'd suggest (these variants round down, which is acceptable for
this implementation):

static u32 usec_to_cycles(u32 usec, unsigned long clock_rate_hz)
{
	u64 tmp = clock_rate_hz * usec;

	do_div(tmp, USEC_PER_SEC);

	return tmp > 0xffffffff ? 0xffffffff : tmp;
}

static u32 cycles_to_usec(u32 cycles, unsigned long clock_rate_hz)
{
	u64 tmp = cycles * USEC_PER_SEC;

	do_div(tmp, clock_rate_hz);

	return tmp > 0xffffffff ? 0xffffffff : tmp;
}

and:
	u32 val = usec_to_cycles(usec, port->priv->tclk);

	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
		usec = cycles_to_usec(MVPP2_MAX_ISR_RX_THRESHOLD,
				      port->priv->tclk);

		/* re-evaluate to get actual register value for usec */
		val = usec_to_cycles(usec, port->priv->tclk);
	}

>  	mvpp2_write(port->priv, MVPP2_ISR_RX_THRESHOLD_REG(rxq->id), val);
>  
>  	rxq->time_coal = usec;

This function appears to be called from two places:

mvpp2_rxq_init():
        mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);

mvpp2_ethtool_set_coalesce():
                rxq->time_coal = c->rx_coalesce_usecs;
                mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);

It seems rather pointless to pass in rxq->time_coal when you're also
passing in rxq.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

[PATCHv2 net-next 02/11] net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()

[Russell King - ARM Linux]

On Wed, Dec 28, 2016 at 05:45:58PM +0100, Thomas Petazzoni wrote:
> When configuring the MVPP2_ISR_RX_THRESHOLD_REG with the RX coalescing
> time threshold, we do not check for the maximum allowed value supported
> by the driver, which means we might overflow and use a bogus value. This
> commit adds a check for this situation, and if a value higher than what
> is supported by the hardware is provided, then we use the maximum value
> supported by the hardware.
> 
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
> ---
>  drivers/net/ethernet/marvell/mvpp2.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/drivers/net/ethernet/marvell/mvpp2.c b/drivers/net/ethernet/marvell/mvpp2.c
> index 02d91e4..a1ba89f 100644
> --- a/drivers/net/ethernet/marvell/mvpp2.c
> +++ b/drivers/net/ethernet/marvell/mvpp2.c
> @@ -154,6 +154,7 @@
>  
>  /* Interrupt Cause and Mask registers */
>  #define MVPP2_ISR_RX_THRESHOLD_REG(rxq)		(0x5200 + 4 * (rxq))
> +#define     MVPP2_MAX_ISR_RX_THRESHOLD		0xfffff0
>  #define MVPP2_ISR_RXQ_GROUP_REG(rxq)		(0x5400 + 4 * (rxq))
>  #define MVPP2_ISR_ENABLE_REG(port)		(0x5420 + 4 * (port))
>  #define     MVPP2_ISR_ENABLE_INTERRUPT(mask)	((mask) & 0xffff)
> @@ -4397,6 +4398,12 @@ static void mvpp2_rx_time_coal_set(struct mvpp2_port *port,
>  	u32 val;
>  
>  	val = (port->priv->tclk / USEC_PER_SEC) * usec;
> +
> +	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
> +		val = MVPP2_MAX_ISR_RX_THRESHOLD;
> +		usec = (val * USEC_PER_SEC) / port->priv->tclk;
> +	}
> +

Beware of rounding and overflow errors.  usec and val are u32's.

	MVPP2_MAX_ISR_RX_THRESHOLD = 16777200
	USEC_PER_SEC = 1000000

This equates to 0xF423F0BDC00 for the multiplication, which is a little
larger than 32-bit.  Assuming tclk is 166.666666MHz (as it was on Dove
- I don't know what it would be here) and 64-bit arithmetic, the maximum
value gives 100663us.

Passing that back into the function gives... 16710058, so the second time
around, we end up with a different setting (even though a change wasn't
requested.)

However, 100664 won't trigger your check, neither will values all the way
up to 101067 - the reason being that you're losing so much precision by
dividling the clock by USEC_PER_SEC first.  Only if it's a whole number
of MHz will you get away with that.

So, I'd suggest you switch to using 64 bit math here - it's not a fast
path.  Using bc to evaluate val = port->priv->tclk * usec / USEC_PER_SEC
gives:

	(166666666 * 100663 / 1000000)
	16777166

which is as close as you can come to the limit.

So, I'd suggest (these variants round down, which is acceptable for
this implementation):

static u32 usec_to_cycles(u32 usec, unsigned long clock_rate_hz)
{
	u64 tmp = clock_rate_hz * usec;

	do_div(tmp, USEC_PER_SEC);

	return tmp > 0xffffffff ? 0xffffffff : tmp;
}

static u32 cycles_to_usec(u32 cycles, unsigned long clock_rate_hz)
{
	u64 tmp = cycles * USEC_PER_SEC;

	do_div(tmp, clock_rate_hz);

	return tmp > 0xffffffff ? 0xffffffff : tmp;
}

and:
	u32 val = usec_to_cycles(usec, port->priv->tclk);

	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
		usec = cycles_to_usec(MVPP2_MAX_ISR_RX_THRESHOLD,
				      port->priv->tclk);

		/* re-evaluate to get actual register value for usec */
		val = usec_to_cycles(usec, port->priv->tclk);
	}

>  	mvpp2_write(port->priv, MVPP2_ISR_RX_THRESHOLD_REG(rxq->id), val);
>  
>  	rxq->time_coal = usec;

This function appears to be called from two places:

mvpp2_rxq_init():
        mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);

mvpp2_ethtool_set_coalesce():
                rxq->time_coal = c->rx_coalesce_usecs;
                mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);

It seems rather pointless to pass in rxq->time_coal when you're also
passing in rxq.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

Re: [PATCHv2 net-next 09/11] net: mvpp2: simplify MVPP2_PRS_RI_* definitions

[Russell King - ARM Linux]

On Wed, Dec 28, 2016 at 05:46:05PM +0100, Thomas Petazzoni wrote:
> Some of the MVPP2_PRS_RI_* definitions use the ~(value) syntax, which
> doesn't compile nicely on 64-bit. Moreover, those definitions are in
> fact unneeded, since they are always used in combination with a bit
> mask that ensures only the appropriate bits are modified.
> 
> Therefore, such definitions should just be set to 0x0. For example:
> 
>  #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
>  #define MVPP2_PRS_RI_L2_UCAST                  ~(BIT(9) | BIT(10))
>  #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
>  #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)
> 
> becomes
> 
>  #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
>  #define MVPP2_PRS_RI_L2_UCAST                  0x0
>  #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
>  #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

So this is a two-bit field in a register with three defined states - I'm
not sure that using BIT() here is really a good idea.  BIT() is fine for
single-bit controls, but I think it adds an additional level of confusion
for multi-bit controls.

Also, the combination of the mask being defined as hex and the controls
using BIT() is particularly not nice.  I think either use one style or
the other, don't mix them.  So either:

  #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
  #define MVPP2_PRS_RI_L2_UCAST                  0x000
  #define MVPP2_PRS_RI_L2_MCAST                  0x200
  #define MVPP2_PRS_RI_L2_BCAST                  0x400

or:

  #define MVPP2_PRS_RI_L2_CAST_MASK              (BIT(10) | BIT(9))
  #define MVPP2_PRS_RI_L2_UCAST                  0
  #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
  #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

It then becomes obvious that the mask and the settings are changing the
same bits.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

[PATCHv2 net-next 09/11] net: mvpp2: simplify MVPP2_PRS_RI_* definitions

[Russell King - ARM Linux]

On Wed, Dec 28, 2016 at 05:46:05PM +0100, Thomas Petazzoni wrote:
> Some of the MVPP2_PRS_RI_* definitions use the ~(value) syntax, which
> doesn't compile nicely on 64-bit. Moreover, those definitions are in
> fact unneeded, since they are always used in combination with a bit
> mask that ensures only the appropriate bits are modified.
> 
> Therefore, such definitions should just be set to 0x0. For example:
> 
>  #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
>  #define MVPP2_PRS_RI_L2_UCAST                  ~(BIT(9) | BIT(10))
>  #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
>  #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)
> 
> becomes
> 
>  #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
>  #define MVPP2_PRS_RI_L2_UCAST                  0x0
>  #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
>  #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

So this is a two-bit field in a register with three defined states - I'm
not sure that using BIT() here is really a good idea.  BIT() is fine for
single-bit controls, but I think it adds an additional level of confusion
for multi-bit controls.

Also, the combination of the mask being defined as hex and the controls
using BIT() is particularly not nice.  I think either use one style or
the other, don't mix them.  So either:

  #define MVPP2_PRS_RI_L2_CAST_MASK              0x600
  #define MVPP2_PRS_RI_L2_UCAST                  0x000
  #define MVPP2_PRS_RI_L2_MCAST                  0x200
  #define MVPP2_PRS_RI_L2_BCAST                  0x400

or:

  #define MVPP2_PRS_RI_L2_CAST_MASK              (BIT(10) | BIT(9))
  #define MVPP2_PRS_RI_L2_UCAST                  0
  #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
  #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

It then becomes obvious that the mask and the settings are changing the
same bits.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.

Re: [PATCHv2 net-next 02/11] net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()

[Thomas Petazzoni]

Russell,

On Fri, 6 Jan 2017 12:59:24 +0000, Russell King - ARM Linux wrote:

> Beware of rounding and overflow errors.  usec and val are u32's.

[...]

Thanks for all the suggestions, I've taken this into account in my v3
that I've sent a few minutes ago? I've re-used almost exactly your
implementation, with a few changes (see below).

> static u32 usec_to_cycles(u32 usec, unsigned long clock_rate_hz)
> {
> 	u64 tmp = clock_rate_hz * usec;

I had to cast one of the variables to u64 here otherwise the
multiplication was done on 32 bits, and then the result was expanded to
64 bits.

> 	do_div(tmp, USEC_PER_SEC);
> 
> 	return tmp > 0xffffffff ? 0xffffffff : tmp;

I've used U32_MAX here.

> static u32 cycles_to_usec(u32 cycles, unsigned long clock_rate_hz)
> {
> 	u64 tmp = cycles * USEC_PER_SEC;
> 
> 	do_div(tmp, clock_rate_hz);
> 
> 	return tmp > 0xffffffff ? 0xffffffff : tmp;

Same comments for this function.

> and:
> 	u32 val = usec_to_cycles(usec, port->priv->tclk);
> 
> 	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
> 		usec = cycles_to_usec(MVPP2_MAX_ISR_RX_THRESHOLD,
> 				      port->priv->tclk);
> 
> 		/* re-evaluate to get actual register value for usec */
> 		val = usec_to_cycles(usec, port->priv->tclk);
> 	}
> 
> >  	mvpp2_write(port->priv, MVPP2_ISR_RX_THRESHOLD_REG(rxq->id), val);
> >  
> >  	rxq->time_coal = usec;  
> 
> This function appears to be called from two places:
> 
> mvpp2_rxq_init():
>         mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);
> 
> mvpp2_ethtool_set_coalesce():
>                 rxq->time_coal = c->rx_coalesce_usecs;
>                 mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);
> 
> It seems rather pointless to pass in rxq->time_coal when you're also
> passing in rxq.

Indeed, so I've added another patch in the series that does this.

Thanks for the review!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[PATCHv2 net-next 02/11] net: mvpp2: handle too large value in mvpp2_rx_time_coal_set()

[Thomas Petazzoni]

Russell,

On Fri, 6 Jan 2017 12:59:24 +0000, Russell King - ARM Linux wrote:

> Beware of rounding and overflow errors.  usec and val are u32's.

[...]

Thanks for all the suggestions, I've taken this into account in my v3
that I've sent a few minutes ago? I've re-used almost exactly your
implementation, with a few changes (see below).

> static u32 usec_to_cycles(u32 usec, unsigned long clock_rate_hz)
> {
> 	u64 tmp = clock_rate_hz * usec;

I had to cast one of the variables to u64 here otherwise the
multiplication was done on 32 bits, and then the result was expanded to
64 bits.

> 	do_div(tmp, USEC_PER_SEC);
> 
> 	return tmp > 0xffffffff ? 0xffffffff : tmp;

I've used U32_MAX here.

> static u32 cycles_to_usec(u32 cycles, unsigned long clock_rate_hz)
> {
> 	u64 tmp = cycles * USEC_PER_SEC;
> 
> 	do_div(tmp, clock_rate_hz);
> 
> 	return tmp > 0xffffffff ? 0xffffffff : tmp;

Same comments for this function.

> and:
> 	u32 val = usec_to_cycles(usec, port->priv->tclk);
> 
> 	if (val > MVPP2_MAX_ISR_RX_THRESHOLD) {
> 		usec = cycles_to_usec(MVPP2_MAX_ISR_RX_THRESHOLD,
> 				      port->priv->tclk);
> 
> 		/* re-evaluate to get actual register value for usec */
> 		val = usec_to_cycles(usec, port->priv->tclk);
> 	}
> 
> >  	mvpp2_write(port->priv, MVPP2_ISR_RX_THRESHOLD_REG(rxq->id), val);
> >  
> >  	rxq->time_coal = usec;  
> 
> This function appears to be called from two places:
> 
> mvpp2_rxq_init():
>         mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);
> 
> mvpp2_ethtool_set_coalesce():
>                 rxq->time_coal = c->rx_coalesce_usecs;
>                 mvpp2_rx_time_coal_set(port, rxq, rxq->time_coal);
> 
> It seems rather pointless to pass in rxq->time_coal when you're also
> passing in rxq.

Indeed, so I've added another patch in the series that does this.

Thanks for the review!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Re: [PATCHv2 net-next 09/11] net: mvpp2: simplify MVPP2_PRS_RI_* definitions

[Thomas Petazzoni]

Hello,

On Fri, 6 Jan 2017 13:07:27 +0000, Russell King - ARM Linux wrote:

>   #define MVPP2_PRS_RI_L2_CAST_MASK              (BIT(10) | BIT(9))
>   #define MVPP2_PRS_RI_L2_UCAST                  0
>   #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
>   #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

Your comment makes perfect sense, so I've adopted this second option
you suggested. This is in my v3 I sent a few minutes ago.

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[PATCHv2 net-next 09/11] net: mvpp2: simplify MVPP2_PRS_RI_* definitions

[Thomas Petazzoni]

Hello,

On Fri, 6 Jan 2017 13:07:27 +0000, Russell King - ARM Linux wrote:

>   #define MVPP2_PRS_RI_L2_CAST_MASK              (BIT(10) | BIT(9))
>   #define MVPP2_PRS_RI_L2_UCAST                  0
>   #define MVPP2_PRS_RI_L2_MCAST                  BIT(9)
>   #define MVPP2_PRS_RI_L2_BCAST                  BIT(10)

Your comment makes perfect sense, so I've adopted this second option
you suggested. This is in my v3 I sent a few minutes ago.

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Re: [PATCHv2 net-next 00/11] net: mvpp2: misc improvements and preparation patches

[Thomas Petazzoni]

David,

On Thu, 29 Dec 2016 12:03:37 -0500 (EST), David Miller wrote:

> You still have warnings to fix for the 64-bit build:
> 
> drivers/net/ethernet/marvell/mvpp2.c: In function ‘mvpp2_rx’:
> drivers/net/ethernet/marvell/mvpp2.c:5125:10: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
>    data = (void *)rx_desc->buf_cookie;
>           ^

Indeed, sorry about that. The warning did not exist with both series
applied, but did exist with just the first series.

I've fixed that up in my v3 of the first series.

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[PATCHv2 net-next 00/11] net: mvpp2: misc improvements and preparation patches

[Thomas Petazzoni]

David,

On Thu, 29 Dec 2016 12:03:37 -0500 (EST), David Miller wrote:

> You still have warnings to fix for the 64-bit build:
> 
> drivers/net/ethernet/marvell/mvpp2.c: In function ?mvpp2_rx?:
> drivers/net/ethernet/marvell/mvpp2.c:5125:10: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
>    data = (void *)rx_desc->buf_cookie;
>           ^

Indeed, sorry about that. The warning did not exist with both series
applied, but did exist with just the first series.

I've fixed that up in my v3 of the first series.

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com