[PATCH net-next V2 0/8] net: fec: cleanup/fixes

All of lore.kernel.org
 help / color / mirror / Atom feed

* [PATCH net-next V2 0/8] net: fec: cleanup/fixes
@ 2016-02-05 21:52 ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel


V2 is a rebase on top of johannes endian-safe patch and
is only the 1st eight patches.
The testing for this series was done on a nitrogen6x.
The base commit was
commit b45efa30a626e915192a6c548cd8642379cd47cc
    Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Testing showed no change in performance.
Testing used imx_v6_v7_defconfig + CONFIG_MICREL_PHY.
The processor was running at 996Mhz.
The following commands were used to get the transfer rates.

On an x86 ubunto system,
iperf -s -i.5 -u


On a nitrogen6x board, running via SD Card.
I first stopped some background processes

stop cron
stop upstart-file-bridge
stop upstart-socket-bridge
stop upstart-udev-bridge
stop rsyslog
stop dbus
killall dhclient
echo performance >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

taskset 0x2 iperf -c 192.168.0.201 -u -t60 -b500M -r

There is a branch available on github with this series, and the rest of
my fec patches, for those who would like to test it.
https://github.com:boundarydevices/linux-imx6.git branch net-next_master



Troy Kisky (8):
SDCard
 TX/RX
378/405  before any patches
379/407  net: fec: stop the "rcv is not +last, " error messages
378/405  net: fec: fix rx error counts
379/396  net: fec: fix fec_enet_get_free_txdesc_num
379/403  net: fec: add struct bufdesc_prop
390/396  net: fec: add variable reg_desc_active to speed things up
388/396  net: fec: don't disable FEC_ENET_TS_TIMER interrupt
382/403  net: fec: don't transfer ownership until descriptor write is complete
378/403  net: fec: improve error handling

 drivers/net/ethernet/freescale/fec.h      |  38 ++-
 drivers/net/ethernet/freescale/fec_main.c | 396 ++++++++++++++----------------
 2 files changed, 196 insertions(+), 238 deletions(-)

-- 
2.5.0

^ permalink raw reply	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 0/8] net: fec: cleanup/fixes
@ 2016-02-05 21:52 ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky


V2 is a rebase on top of johannes endian-safe patch and
is only the 1st eight patches.
The testing for this series was done on a nitrogen6x.
The base commit was
commit b45efa30a626e915192a6c548cd8642379cd47cc
    Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Testing showed no change in performance.
Testing used imx_v6_v7_defconfig + CONFIG_MICREL_PHY.
The processor was running at 996Mhz.
The following commands were used to get the transfer rates.

On an x86 ubunto system,
iperf -s -i.5 -u


On a nitrogen6x board, running via SD Card.
I first stopped some background processes

stop cron
stop upstart-file-bridge
stop upstart-socket-bridge
stop upstart-udev-bridge
stop rsyslog
stop dbus
killall dhclient
echo performance >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor

taskset 0x2 iperf -c 192.168.0.201 -u -t60 -b500M -r

There is a branch available on github with this series, and the rest of
my fec patches, for those who would like to test it.
https://github.com:boundarydevices/linux-imx6.git branch net-next_master



Troy Kisky (8):
SDCard
 TX/RX
378/405  before any patches
379/407  net: fec: stop the "rcv is not +last, " error messages
378/405  net: fec: fix rx error counts
379/396  net: fec: fix fec_enet_get_free_txdesc_num
379/403  net: fec: add struct bufdesc_prop
390/396  net: fec: add variable reg_desc_active to speed things up
388/396  net: fec: don't disable FEC_ENET_TS_TIMER interrupt
382/403  net: fec: don't transfer ownership until descriptor write is complete
378/403  net: fec: improve error handling

 drivers/net/ethernet/freescale/fec.h      |  38 ++-
 drivers/net/ethernet/freescale/fec_main.c | 396 ++++++++++++++----------------
 2 files changed, 196 insertions(+), 238 deletions(-)

-- 
2.5.0

^ permalink raw reply	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 1/8] net: fec: stop the "rcv is not +last, " error messages
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

Setting the FTRL register will stop the fec from
trying to use multiple receive buffers.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      | 1 +
 drivers/net/ethernet/freescale/fec_main.c | 1 +
 2 files changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index 2106d72..cc9677a 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -64,6 +64,7 @@
 #define FEC_R_FIFO_RSEM		0x194 /* Receive FIFO section empty threshold */
 #define FEC_R_FIFO_RAEM		0x198 /* Receive FIFO almost empty threshold */
 #define FEC_R_FIFO_RAFL		0x19c /* Receive FIFO almost full threshold */
+#define FEC_FTRL		0x1b0 /* Frame truncation receive length*/
 #define FEC_RACC		0x1c4 /* Receive Accelerator function */
 #define FEC_RCMR_1		0x1c8 /* Receive classification match ring 1 */
 #define FEC_RCMR_2		0x1cc /* Receive classification match ring 2 */
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 41c81f6..3e5b24a 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -988,6 +988,7 @@ fec_restart(struct net_device *ndev)
 			val &= ~FEC_RACC_OPTIONS;
 		writel(val, fep->hwp + FEC_RACC);
 	}
+	writel(PKT_MAXBUF_SIZE, fep->hwp + FEC_FTRL);
 #endif
 
 	/*
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 1/8] net: fec: stop the "rcv is not +last, " error messages
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

Setting the FTRL register will stop the fec from
trying to use multiple receive buffers.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      | 1 +
 drivers/net/ethernet/freescale/fec_main.c | 1 +
 2 files changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index 2106d72..cc9677a 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -64,6 +64,7 @@
 #define FEC_R_FIFO_RSEM		0x194 /* Receive FIFO section empty threshold */
 #define FEC_R_FIFO_RAEM		0x198 /* Receive FIFO almost empty threshold */
 #define FEC_R_FIFO_RAFL		0x19c /* Receive FIFO almost full threshold */
+#define FEC_FTRL		0x1b0 /* Frame truncation receive length*/
 #define FEC_RACC		0x1c4 /* Receive Accelerator function */
 #define FEC_RCMR_1		0x1c8 /* Receive classification match ring 1 */
 #define FEC_RCMR_2		0x1cc /* Receive classification match ring 2 */
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 41c81f6..3e5b24a 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -988,6 +988,7 @@ fec_restart(struct net_device *ndev)
 			val &= ~FEC_RACC_OPTIONS;
 		writel(val, fep->hwp + FEC_RACC);
 	}
+	writel(PKT_MAXBUF_SIZE, fep->hwp + FEC_FTRL);
 #endif
 
 	/*
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 2/8] net: fec: fix rx error counts
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

On an overrun, the other flags are not
valid, so don't check them.

Also, don't pass bad frames up the stack.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 36 +++++++++++++------------------
 1 file changed, 15 insertions(+), 21 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 3e5b24a..162fa59 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -1408,37 +1408,31 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
 			break;
 		pkt_received++;
 
-		/* Since we have allocated space to hold a complete frame,
-		 * the last indicator should be set.
-		 */
-		if ((status & BD_ENET_RX_LAST) == 0)
-			netdev_err(ndev, "rcv is not +last\n");
-
 		writel(FEC_ENET_RXF, fep->hwp + FEC_IEVENT);
 
 		/* Check for errors. */
+		status ^= BD_ENET_RX_LAST;
 		if (status & (BD_ENET_RX_LG | BD_ENET_RX_SH | BD_ENET_RX_NO |
-			   BD_ENET_RX_CR | BD_ENET_RX_OV)) {
+			   BD_ENET_RX_CR | BD_ENET_RX_OV | BD_ENET_RX_LAST |
+			   BD_ENET_RX_CL)) {
 			ndev->stats.rx_errors++;
-			if (status & (BD_ENET_RX_LG | BD_ENET_RX_SH)) {
+			if (status & BD_ENET_RX_OV) {
+				/* FIFO overrun */
+				ndev->stats.rx_fifo_errors++;
+				goto rx_processing_done;
+			}
+			if (status & (BD_ENET_RX_LG | BD_ENET_RX_SH
+						| BD_ENET_RX_LAST)) {
 				/* Frame too long or too short. */
 				ndev->stats.rx_length_errors++;
+				if (status & BD_ENET_RX_LAST)
+					netdev_err(ndev, "rcv is not +last\n");
 			}
-			if (status & BD_ENET_RX_NO)	/* Frame alignment */
-				ndev->stats.rx_frame_errors++;
 			if (status & BD_ENET_RX_CR)	/* CRC Error */
 				ndev->stats.rx_crc_errors++;
-			if (status & BD_ENET_RX_OV)	/* FIFO overrun */
-				ndev->stats.rx_fifo_errors++;
-		}
-
-		/* Report late collisions as a frame error.
-		 * On this error, the BD is closed, but we don't know what we
-		 * have in the buffer.  So, just drop this frame on the floor.
-		 */
-		if (status & BD_ENET_RX_CL) {
-			ndev->stats.rx_errors++;
-			ndev->stats.rx_frame_errors++;
+			/* Report late collisions as a frame error. */
+			if (status & (BD_ENET_RX_NO | BD_ENET_RX_CL))
+				ndev->stats.rx_frame_errors++;
 			goto rx_processing_done;
 		}
 
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 2/8] net: fec: fix rx error counts
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

On an overrun, the other flags are not
valid, so don't check them.

Also, don't pass bad frames up the stack.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 36 +++++++++++++------------------
 1 file changed, 15 insertions(+), 21 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 3e5b24a..162fa59 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -1408,37 +1408,31 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
 			break;
 		pkt_received++;
 
-		/* Since we have allocated space to hold a complete frame,
-		 * the last indicator should be set.
-		 */
-		if ((status & BD_ENET_RX_LAST) == 0)
-			netdev_err(ndev, "rcv is not +last\n");
-
 		writel(FEC_ENET_RXF, fep->hwp + FEC_IEVENT);
 
 		/* Check for errors. */
+		status ^= BD_ENET_RX_LAST;
 		if (status & (BD_ENET_RX_LG | BD_ENET_RX_SH | BD_ENET_RX_NO |
-			   BD_ENET_RX_CR | BD_ENET_RX_OV)) {
+			   BD_ENET_RX_CR | BD_ENET_RX_OV | BD_ENET_RX_LAST |
+			   BD_ENET_RX_CL)) {
 			ndev->stats.rx_errors++;
-			if (status & (BD_ENET_RX_LG | BD_ENET_RX_SH)) {
+			if (status & BD_ENET_RX_OV) {
+				/* FIFO overrun */
+				ndev->stats.rx_fifo_errors++;
+				goto rx_processing_done;
+			}
+			if (status & (BD_ENET_RX_LG | BD_ENET_RX_SH
+						| BD_ENET_RX_LAST)) {
 				/* Frame too long or too short. */
 				ndev->stats.rx_length_errors++;
+				if (status & BD_ENET_RX_LAST)
+					netdev_err(ndev, "rcv is not +last\n");
 			}
-			if (status & BD_ENET_RX_NO)	/* Frame alignment */
-				ndev->stats.rx_frame_errors++;
 			if (status & BD_ENET_RX_CR)	/* CRC Error */
 				ndev->stats.rx_crc_errors++;
-			if (status & BD_ENET_RX_OV)	/* FIFO overrun */
-				ndev->stats.rx_fifo_errors++;
-		}
-
-		/* Report late collisions as a frame error.
-		 * On this error, the BD is closed, but we don't know what we
-		 * have in the buffer.  So, just drop this frame on the floor.
-		 */
-		if (status & BD_ENET_RX_CL) {
-			ndev->stats.rx_errors++;
-			ndev->stats.rx_frame_errors++;
+			/* Report late collisions as a frame error. */
+			if (status & (BD_ENET_RX_NO | BD_ENET_RX_CL))
+				ndev->stats.rx_frame_errors++;
 			goto rx_processing_done;
 		}
 
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 3/8] net: fec: fix fec_enet_get_free_txdesc_num
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

When first initialized, cur_tx points to the 1st
entry in the queue, and dirty_tx points to the last.
At this point, fec_enet_get_free_txdesc_num will
return tx_ring_size -2. If tx_ring_size -2 entries
are now queued, then fec_enet_get_free_txdesc_num
should return 0, but it returns tx_ring_size instead.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 162fa59..adbddfd 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -296,7 +296,7 @@ static int fec_enet_get_free_txdesc_num(struct fec_enet_private *fep,
 	entries = ((const char *)txq->dirty_tx -
 			(const char *)txq->cur_tx) / fep->bufdesc_size - 1;
 
-	return entries > 0 ? entries : entries + txq->tx_ring_size;
+	return entries >= 0 ? entries : entries + txq->tx_ring_size;
 }
 
 static void swap_buffer(void *bufaddr, int len)
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 3/8] net: fec: fix fec_enet_get_free_txdesc_num
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

When first initialized, cur_tx points to the 1st
entry in the queue, and dirty_tx points to the last.
At this point, fec_enet_get_free_txdesc_num will
return tx_ring_size -2. If tx_ring_size -2 entries
are now queued, then fec_enet_get_free_txdesc_num
should return 0, but it returns tx_ring_size instead.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 162fa59..adbddfd 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -296,7 +296,7 @@ static int fec_enet_get_free_txdesc_num(struct fec_enet_private *fep,
 	entries = ((const char *)txq->dirty_tx -
 			(const char *)txq->cur_tx) / fep->bufdesc_size - 1;
 
-	return entries > 0 ? entries : entries + txq->tx_ring_size;
+	return entries >= 0 ? entries : entries + txq->tx_ring_size;
 }
 
 static void swap_buffer(void *bufaddr, int len)
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 4/8] net: fec: add struct bufdesc_prop
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

This reduces code and gains speed.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      |  29 ++-
 drivers/net/ethernet/freescale/fec_main.c | 288 ++++++++++++------------------
 2 files changed, 132 insertions(+), 185 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index cc9677a..53ec04f 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -448,33 +448,34 @@ struct bufdesc_ex {
 /* Controller supports RACC register */
 #define FEC_QUIRK_HAS_RACC		(1 << 12)
 
+struct bufdesc_prop {
+	int qid;
+	/* Address of Rx and Tx buffers */
+	struct bufdesc	*base;
+	struct bufdesc	*last;
+	struct bufdesc	*cur;
+	dma_addr_t	dma;
+	unsigned short ring_size;
+	unsigned char dsize;
+	unsigned char dsize_log2;
+};
+
 struct fec_enet_priv_tx_q {
-	int index;
+	struct bufdesc_prop bd;
 	unsigned char *tx_bounce[TX_RING_SIZE];
 	struct  sk_buff *tx_skbuff[TX_RING_SIZE];
 
-	dma_addr_t	bd_dma;
-	struct bufdesc	*tx_bd_base;
-	uint tx_ring_size;
-
 	unsigned short tx_stop_threshold;
 	unsigned short tx_wake_threshold;
 
-	struct bufdesc	*cur_tx;
 	struct bufdesc	*dirty_tx;
 	char *tso_hdrs;
 	dma_addr_t tso_hdrs_dma;
 };
 
 struct fec_enet_priv_rx_q {
-	int index;
+	struct bufdesc_prop bd;
 	struct  sk_buff *rx_skbuff[RX_RING_SIZE];
-
-	dma_addr_t	bd_dma;
-	struct bufdesc	*rx_bd_base;
-	uint rx_ring_size;
-
-	struct bufdesc	*cur_rx;
 };
 
 /* The FEC buffer descriptors track the ring buffers.  The rx_bd_base and
@@ -514,8 +515,6 @@ struct fec_enet_private {
 	unsigned long work_ts;
 	unsigned long work_mdio;
 
-	unsigned short bufdesc_size;
-
 	struct	platform_device *pdev;
 
 	int	dev_id;
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index adbddfd..b039288 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -217,86 +217,38 @@ MODULE_PARM_DESC(macaddr, "FEC Ethernet MAC address");
 
 #define IS_TSO_HEADER(txq, addr) \
 	((addr >= txq->tso_hdrs_dma) && \
-	(addr < txq->tso_hdrs_dma + txq->tx_ring_size * TSO_HEADER_SIZE))
+	(addr < txq->tso_hdrs_dma + txq->bd.ring_size * TSO_HEADER_SIZE))
 
 static int mii_cnt;
 
-static inline
-struct bufdesc *fec_enet_get_nextdesc(struct bufdesc *bdp,
-				      struct fec_enet_private *fep,
-				      int queue_id)
-{
-	struct bufdesc *new_bd = bdp + 1;
-	struct bufdesc_ex *ex_new_bd = (struct bufdesc_ex *)bdp + 1;
-	struct fec_enet_priv_tx_q *txq = fep->tx_queue[queue_id];
-	struct fec_enet_priv_rx_q *rxq = fep->rx_queue[queue_id];
-	struct bufdesc_ex *ex_base;
-	struct bufdesc *base;
-	int ring_size;
-
-	if (bdp >= txq->tx_bd_base) {
-		base = txq->tx_bd_base;
-		ring_size = txq->tx_ring_size;
-		ex_base = (struct bufdesc_ex *)txq->tx_bd_base;
-	} else {
-		base = rxq->rx_bd_base;
-		ring_size = rxq->rx_ring_size;
-		ex_base = (struct bufdesc_ex *)rxq->rx_bd_base;
-	}
-
-	if (fep->bufdesc_ex)
-		return (struct bufdesc *)((ex_new_bd >= (ex_base + ring_size)) ?
-			ex_base : ex_new_bd);
-	else
-		return (new_bd >= (base + ring_size)) ?
-			base : new_bd;
-}
-
-static inline
-struct bufdesc *fec_enet_get_prevdesc(struct bufdesc *bdp,
-				      struct fec_enet_private *fep,
-				      int queue_id)
-{
-	struct bufdesc *new_bd = bdp - 1;
-	struct bufdesc_ex *ex_new_bd = (struct bufdesc_ex *)bdp - 1;
-	struct fec_enet_priv_tx_q *txq = fep->tx_queue[queue_id];
-	struct fec_enet_priv_rx_q *rxq = fep->rx_queue[queue_id];
-	struct bufdesc_ex *ex_base;
-	struct bufdesc *base;
-	int ring_size;
-
-	if (bdp >= txq->tx_bd_base) {
-		base = txq->tx_bd_base;
-		ring_size = txq->tx_ring_size;
-		ex_base = (struct bufdesc_ex *)txq->tx_bd_base;
-	} else {
-		base = rxq->rx_bd_base;
-		ring_size = rxq->rx_ring_size;
-		ex_base = (struct bufdesc_ex *)rxq->rx_bd_base;
-	}
+static struct bufdesc *fec_enet_get_nextdesc(struct bufdesc *bdp,
+					     struct bufdesc_prop *bd)
+{
+	return (bdp >= bd->last) ? bd->base
+			: (struct bufdesc *)(((unsigned)bdp) + bd->dsize);
+}
 
-	if (fep->bufdesc_ex)
-		return (struct bufdesc *)((ex_new_bd < ex_base) ?
-			(ex_new_bd + ring_size) : ex_new_bd);
-	else
-		return (new_bd < base) ? (new_bd + ring_size) : new_bd;
+static struct bufdesc *fec_enet_get_prevdesc(struct bufdesc *bdp,
+					     struct bufdesc_prop *bd)
+{
+	return (bdp <= bd->base) ? bd->last
+			: (struct bufdesc *)(((unsigned)bdp) - bd->dsize);
 }
 
-static int fec_enet_get_bd_index(struct bufdesc *base, struct bufdesc *bdp,
-				struct fec_enet_private *fep)
+static int fec_enet_get_bd_index(struct bufdesc *bdp,
+				 struct bufdesc_prop *bd)
 {
-	return ((const char *)bdp - (const char *)base) / fep->bufdesc_size;
+	return ((const char *)bdp - (const char *)bd->base) >> bd->dsize_log2;
 }
 
-static int fec_enet_get_free_txdesc_num(struct fec_enet_private *fep,
-					struct fec_enet_priv_tx_q *txq)
+static int fec_enet_get_free_txdesc_num(struct fec_enet_priv_tx_q *txq)
 {
 	int entries;
 
-	entries = ((const char *)txq->dirty_tx -
-			(const char *)txq->cur_tx) / fep->bufdesc_size - 1;
+	entries = (((const char *)txq->dirty_tx -
+			(const char *)txq->bd.cur) >> txq->bd.dsize_log2) - 1;
 
-	return entries >= 0 ? entries : entries + txq->tx_ring_size;
+	return entries >= 0 ? entries : entries + txq->bd.ring_size;
 }
 
 static void swap_buffer(void *bufaddr, int len)
@@ -329,20 +281,20 @@ static void fec_dump(struct net_device *ndev)
 	pr_info("Nr     SC     addr       len  SKB\n");
 
 	txq = fep->tx_queue[0];
-	bdp = txq->tx_bd_base;
+	bdp = txq->bd.base;
 
 	do {
 		pr_info("%3u %c%c 0x%04x 0x%08x %4u %p\n",
 			index,
-			bdp == txq->cur_tx ? 'S' : ' ',
+			bdp == txq->bd.cur ? 'S' : ' ',
 			bdp == txq->dirty_tx ? 'H' : ' ',
 			fec16_to_cpu(bdp->cbd_sc),
 			fec32_to_cpu(bdp->cbd_bufaddr),
 			fec16_to_cpu(bdp->cbd_datlen),
 			txq->tx_skbuff[index]);
-		bdp = fec_enet_get_nextdesc(bdp, fep, 0);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		index++;
-	} while (bdp != txq->tx_bd_base);
+	} while (bdp != txq->bd.base);
 }
 
 static inline bool is_ipv4_pkt(struct sk_buff *skb)
@@ -373,7 +325,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 			     struct net_device *ndev)
 {
 	struct fec_enet_private *fep = netdev_priv(ndev);
-	struct bufdesc *bdp = txq->cur_tx;
+	struct bufdesc *bdp = txq->bd.cur;
 	struct bufdesc_ex *ebdp;
 	int nr_frags = skb_shinfo(skb)->nr_frags;
 	unsigned short queue = skb_get_queue_mapping(skb);
@@ -388,7 +340,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 	for (frag = 0; frag < nr_frags; frag++) {
 		this_frag = &skb_shinfo(skb)->frags[frag];
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		ebdp = (struct bufdesc_ex *)bdp;
 
 		status = fec16_to_cpu(bdp->cbd_sc);
@@ -418,7 +370,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 		bufaddr = page_address(this_frag->page.p) + this_frag->page_offset;
 
-		index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &txq->bd);
 		if (((unsigned long) bufaddr) & fep->tx_align ||
 			fep->quirks & FEC_QUIRK_SWAP_FRAME) {
 			memcpy(txq->tx_bounce[index], bufaddr, frag_len);
@@ -444,9 +396,9 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 	return bdp;
 dma_mapping_error:
-	bdp = txq->cur_tx;
+	bdp = txq->bd.cur;
 	for (i = 0; i < frag; i++) {
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		dma_unmap_single(&fep->pdev->dev, fec32_to_cpu(bdp->cbd_bufaddr),
 				 fec16_to_cpu(bdp->cbd_datlen), DMA_TO_DEVICE);
 	}
@@ -468,7 +420,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	unsigned int index;
 	int entries_free;
 
-	entries_free = fec_enet_get_free_txdesc_num(fep, txq);
+	entries_free = fec_enet_get_free_txdesc_num(txq);
 	if (entries_free < MAX_SKB_FRAGS + 1) {
 		dev_kfree_skb_any(skb);
 		if (net_ratelimit())
@@ -483,7 +435,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	}
 
 	/* Fill in a Tx ring entry */
-	bdp = txq->cur_tx;
+	bdp = txq->bd.cur;
 	last_bdp = bdp;
 	status = fec16_to_cpu(bdp->cbd_sc);
 	status &= ~BD_ENET_TX_STATS;
@@ -493,7 +445,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	buflen = skb_headlen(skb);
 
 	queue = skb_get_queue_mapping(skb);
-	index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+	index = fec_enet_get_bd_index(bdp, &txq->bd);
 	if (((unsigned long) bufaddr) & fep->tx_align ||
 		fep->quirks & FEC_QUIRK_SWAP_FRAME) {
 		memcpy(txq->tx_bounce[index], skb->data, buflen);
@@ -544,7 +496,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 		ebdp->cbd_esc = cpu_to_fec32(estatus);
 	}
 
-	index = fec_enet_get_bd_index(txq->tx_bd_base, last_bdp, fep);
+	index = fec_enet_get_bd_index(last_bdp, &txq->bd);
 	/* Save skb pointer */
 	txq->tx_skbuff[index] = skb;
 
@@ -558,15 +510,15 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	bdp->cbd_sc = cpu_to_fec16(status);
 
 	/* If this was the last BD in the ring, start at the beginning again. */
-	bdp = fec_enet_get_nextdesc(last_bdp, fep, queue);
+	bdp = fec_enet_get_nextdesc(last_bdp, &txq->bd);
 
 	skb_tx_timestamp(skb);
 
 	/* Make sure the update to bdp and tx_skbuff are performed before
-	 * cur_tx.
+	 * txq->bd.cur.
 	 */
 	wmb();
-	txq->cur_tx = bdp;
+	txq->bd.cur = bdp;
 
 	/* Trigger transmission start */
 	writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue));
@@ -697,13 +649,13 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 	struct fec_enet_private *fep = netdev_priv(ndev);
 	int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
 	int total_len, data_left;
-	struct bufdesc *bdp = txq->cur_tx;
+	struct bufdesc *bdp = txq->bd.cur;
 	unsigned short queue = skb_get_queue_mapping(skb);
 	struct tso_t tso;
 	unsigned int index = 0;
 	int ret;
 
-	if (tso_count_descs(skb) >= fec_enet_get_free_txdesc_num(fep, txq)) {
+	if (tso_count_descs(skb) >= fec_enet_get_free_txdesc_num(txq)) {
 		dev_kfree_skb_any(skb);
 		if (net_ratelimit())
 			netdev_err(ndev, "NOT enough BD for TSO!\n");
@@ -723,7 +675,7 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 	while (total_len > 0) {
 		char *hdr;
 
-		index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &txq->bd);
 		data_left = min_t(int, skb_shinfo(skb)->gso_size, total_len);
 		total_len -= data_left;
 
@@ -738,9 +690,8 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 			int size;
 
 			size = min_t(int, tso.size, data_left);
-			bdp = fec_enet_get_nextdesc(bdp, fep, queue);
-			index = fec_enet_get_bd_index(txq->tx_bd_base,
-						      bdp, fep);
+			bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
+			index = fec_enet_get_bd_index(bdp, &txq->bd);
 			ret = fec_enet_txq_put_data_tso(txq, skb, ndev,
 							bdp, index,
 							tso.data, size,
@@ -753,14 +704,14 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 			tso_build_data(skb, &tso, size);
 		}
 
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 	}
 
 	/* Save skb pointer */
 	txq->tx_skbuff[index] = skb;
 
 	skb_tx_timestamp(skb);
-	txq->cur_tx = bdp;
+	txq->bd.cur = bdp;
 
 	/* Trigger transmission start */
 	if (!(fep->quirks & FEC_QUIRK_ERR007885) ||
@@ -798,7 +749,7 @@ fec_enet_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 	if (ret)
 		return ret;
 
-	entries_free = fec_enet_get_free_txdesc_num(fep, txq);
+	entries_free = fec_enet_get_free_txdesc_num(txq);
 	if (entries_free <= txq->tx_stop_threshold)
 		netif_tx_stop_queue(nq);
 
@@ -819,32 +770,32 @@ static void fec_enet_bd_init(struct net_device *dev)
 	for (q = 0; q < fep->num_rx_queues; q++) {
 		/* Initialize the receive buffer descriptors. */
 		rxq = fep->rx_queue[q];
-		bdp = rxq->rx_bd_base;
+		bdp = rxq->bd.base;
 
-		for (i = 0; i < rxq->rx_ring_size; i++) {
+		for (i = 0; i < rxq->bd.ring_size; i++) {
 
 			/* Initialize the BD for every fragment in the page. */
 			if (bdp->cbd_bufaddr)
 				bdp->cbd_sc = cpu_to_fec16(BD_ENET_RX_EMPTY);
 			else
 				bdp->cbd_sc = cpu_to_fec16(0);
-			bdp = fec_enet_get_nextdesc(bdp, fep, q);
+			bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 		}
 
 		/* Set the last buffer to wrap */
-		bdp = fec_enet_get_prevdesc(bdp, fep, q);
+		bdp = fec_enet_get_prevdesc(bdp, &rxq->bd);
 		bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 
-		rxq->cur_rx = rxq->rx_bd_base;
+		rxq->bd.cur = rxq->bd.base;
 	}
 
 	for (q = 0; q < fep->num_tx_queues; q++) {
 		/* ...and the same for transmit */
 		txq = fep->tx_queue[q];
-		bdp = txq->tx_bd_base;
-		txq->cur_tx = bdp;
+		bdp = txq->bd.base;
+		txq->bd.cur = bdp;
 
-		for (i = 0; i < txq->tx_ring_size; i++) {
+		for (i = 0; i < txq->bd.ring_size; i++) {
 			/* Initialize the BD for every fragment in the page. */
 			bdp->cbd_sc = cpu_to_fec16(0);
 			if (txq->tx_skbuff[i]) {
@@ -852,11 +803,11 @@ static void fec_enet_bd_init(struct net_device *dev)
 				txq->tx_skbuff[i] = NULL;
 			}
 			bdp->cbd_bufaddr = cpu_to_fec32(0);
-			bdp = fec_enet_get_nextdesc(bdp, fep, q);
+			bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		}
 
 		/* Set the last buffer to wrap */
-		bdp = fec_enet_get_prevdesc(bdp, fep, q);
+		bdp = fec_enet_get_prevdesc(bdp, &txq->bd);
 		bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 		txq->dirty_tx = bdp;
 	}
@@ -880,7 +831,7 @@ static void fec_enet_enable_ring(struct net_device *ndev)
 
 	for (i = 0; i < fep->num_rx_queues; i++) {
 		rxq = fep->rx_queue[i];
-		writel(rxq->bd_dma, fep->hwp + FEC_R_DES_START(i));
+		writel(rxq->bd.dma, fep->hwp + FEC_R_DES_START(i));
 		writel(PKT_MAXBLR_SIZE, fep->hwp + FEC_R_BUFF_SIZE(i));
 
 		/* enable DMA1/2 */
@@ -891,7 +842,7 @@ static void fec_enet_enable_ring(struct net_device *ndev)
 
 	for (i = 0; i < fep->num_tx_queues; i++) {
 		txq = fep->tx_queue[i];
-		writel(txq->bd_dma, fep->hwp + FEC_X_DES_START(i));
+		writel(txq->bd.dma, fep->hwp + FEC_X_DES_START(i));
 
 		/* enable DMA1/2 */
 		if (i)
@@ -909,7 +860,7 @@ static void fec_enet_reset_skb(struct net_device *ndev)
 	for (i = 0; i < fep->num_tx_queues; i++) {
 		txq = fep->tx_queue[i];
 
-		for (j = 0; j < txq->tx_ring_size; j++) {
+		for (j = 0; j < txq->bd.ring_size; j++) {
 			if (txq->tx_skbuff[j]) {
 				dev_kfree_skb_any(txq->tx_skbuff[j]);
 				txq->tx_skbuff[j] = NULL;
@@ -1222,16 +1173,16 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 	bdp = txq->dirty_tx;
 
 	/* get next bdp of dirty_tx */
-	bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+	bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 
-	while (bdp != READ_ONCE(txq->cur_tx)) {
-		/* Order the load of cur_tx and cbd_sc */
+	while (bdp != READ_ONCE(txq->bd.cur)) {
+		/* Order the load of bd.cur and cbd_sc */
 		rmb();
 		status = fec16_to_cpu(READ_ONCE(bdp->cbd_sc));
 		if (status & BD_ENET_TX_READY)
 			break;
 
-		index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &txq->bd);
 
 		skb = txq->tx_skbuff[index];
 		txq->tx_skbuff[index] = NULL;
@@ -1242,7 +1193,7 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 					 DMA_TO_DEVICE);
 		bdp->cbd_bufaddr = cpu_to_fec32(0);
 		if (!skb) {
-			bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+			bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 			continue;
 		}
 
@@ -1291,19 +1242,19 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 		txq->dirty_tx = bdp;
 
 		/* Update pointer to next buffer descriptor to be transmitted */
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 
 		/* Since we have freed up a buffer, the ring is no longer full
 		 */
 		if (netif_queue_stopped(ndev)) {
-			entries_free = fec_enet_get_free_txdesc_num(fep, txq);
+			entries_free = fec_enet_get_free_txdesc_num(txq);
 			if (entries_free >= txq->tx_wake_threshold)
 				netif_tx_wake_queue(nq);
 		}
 	}
 
 	/* ERR006538: Keep the transmitter going */
-	if (bdp != txq->cur_tx &&
+	if (bdp != txq->bd.cur &&
 	    readl(fep->hwp + FEC_X_DES_ACTIVE(queue_id)) == 0)
 		writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue_id));
 }
@@ -1367,7 +1318,7 @@ static bool fec_enet_copybreak(struct net_device *ndev, struct sk_buff **skb,
 	return true;
 }
 
-/* During a receive, the cur_rx points to the current incoming buffer.
+/* During a receive, the bd_rx.cur points to the current incoming buffer.
  * When we update through the ring, if the next incoming buffer has
  * not been given to the system, we just set the empty indicator,
  * effectively tossing the packet.
@@ -1400,7 +1351,7 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
 	/* First, grab all of the stats for the incoming packet.
 	 * These get messed up if we get called due to a busy condition.
 	 */
-	bdp = rxq->cur_rx;
+	bdp = rxq->bd.cur;
 
 	while (!((status = fec16_to_cpu(bdp->cbd_sc)) & BD_ENET_RX_EMPTY)) {
 
@@ -1441,7 +1392,7 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
 		pkt_len = fec16_to_cpu(bdp->cbd_datlen);
 		ndev->stats.rx_bytes += pkt_len;
 
-		index = fec_enet_get_bd_index(rxq->rx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &rxq->bd);
 		skb = rxq->rx_skbuff[index];
 
 		/* The packet length includes FCS, but we don't want to
@@ -1541,7 +1492,7 @@ rx_processing_done:
 		}
 
 		/* Update BD pointer to next entry */
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 
 		/* Doing this here will keep the FEC running while we process
 		 * incoming frames.  On a heavily loaded network, we should be
@@ -1549,7 +1500,7 @@ rx_processing_done:
 		 */
 		writel(0, fep->hwp + FEC_R_DES_ACTIVE(queue_id));
 	}
-	rxq->cur_rx = bdp;
+	rxq->bd.cur = bdp;
 	return pkt_received;
 }
 
@@ -2658,8 +2609,8 @@ static void fec_enet_free_buffers(struct net_device *ndev)
 
 	for (q = 0; q < fep->num_rx_queues; q++) {
 		rxq = fep->rx_queue[q];
-		bdp = rxq->rx_bd_base;
-		for (i = 0; i < rxq->rx_ring_size; i++) {
+		bdp = rxq->bd.base;
+		for (i = 0; i < rxq->bd.ring_size; i++) {
 			skb = rxq->rx_skbuff[i];
 			rxq->rx_skbuff[i] = NULL;
 			if (skb) {
@@ -2669,14 +2620,14 @@ static void fec_enet_free_buffers(struct net_device *ndev)
 						 DMA_FROM_DEVICE);
 				dev_kfree_skb(skb);
 			}
-			bdp = fec_enet_get_nextdesc(bdp, fep, q);
+			bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 		}
 	}
 
 	for (q = 0; q < fep->num_tx_queues; q++) {
 		txq = fep->tx_queue[q];
-		bdp = txq->tx_bd_base;
-		for (i = 0; i < txq->tx_ring_size; i++) {
+		bdp = txq->bd.base;
+		for (i = 0; i < txq->bd.ring_size; i++) {
 			kfree(txq->tx_bounce[i]);
 			txq->tx_bounce[i] = NULL;
 			skb = txq->tx_skbuff[i];
@@ -2696,7 +2647,7 @@ static void fec_enet_free_queue(struct net_device *ndev)
 		if (fep->tx_queue[i] && fep->tx_queue[i]->tso_hdrs) {
 			txq = fep->tx_queue[i];
 			dma_free_coherent(NULL,
-					  txq->tx_ring_size * TSO_HEADER_SIZE,
+					  txq->bd.ring_size * TSO_HEADER_SIZE,
 					  txq->tso_hdrs,
 					  txq->tso_hdrs_dma);
 		}
@@ -2722,15 +2673,15 @@ static int fec_enet_alloc_queue(struct net_device *ndev)
 		}
 
 		fep->tx_queue[i] = txq;
-		txq->tx_ring_size = TX_RING_SIZE;
-		fep->total_tx_ring_size += fep->tx_queue[i]->tx_ring_size;
+		txq->bd.ring_size = TX_RING_SIZE;
+		fep->total_tx_ring_size += fep->tx_queue[i]->bd.ring_size;
 
 		txq->tx_stop_threshold = FEC_MAX_SKB_DESCS;
 		txq->tx_wake_threshold =
-				(txq->tx_ring_size - txq->tx_stop_threshold) / 2;
+			(txq->bd.ring_size - txq->tx_stop_threshold) / 2;
 
 		txq->tso_hdrs = dma_alloc_coherent(NULL,
-					txq->tx_ring_size * TSO_HEADER_SIZE,
+					txq->bd.ring_size * TSO_HEADER_SIZE,
 					&txq->tso_hdrs_dma,
 					GFP_KERNEL);
 		if (!txq->tso_hdrs) {
@@ -2747,8 +2698,8 @@ static int fec_enet_alloc_queue(struct net_device *ndev)
 			goto alloc_failed;
 		}
 
-		fep->rx_queue[i]->rx_ring_size = RX_RING_SIZE;
-		fep->total_rx_ring_size += fep->rx_queue[i]->rx_ring_size;
+		fep->rx_queue[i]->bd.ring_size = RX_RING_SIZE;
+		fep->total_rx_ring_size += fep->rx_queue[i]->bd.ring_size;
 	}
 	return ret;
 
@@ -2767,8 +2718,8 @@ fec_enet_alloc_rxq_buffers(struct net_device *ndev, unsigned int queue)
 	struct fec_enet_priv_rx_q *rxq;
 
 	rxq = fep->rx_queue[queue];
-	bdp = rxq->rx_bd_base;
-	for (i = 0; i < rxq->rx_ring_size; i++) {
+	bdp = rxq->bd.base;
+	for (i = 0; i < rxq->bd.ring_size; i++) {
 		skb = netdev_alloc_skb(ndev, FEC_ENET_RX_FRSIZE);
 		if (!skb)
 			goto err_alloc;
@@ -2786,11 +2737,11 @@ fec_enet_alloc_rxq_buffers(struct net_device *ndev, unsigned int queue)
 			ebdp->cbd_esc = cpu_to_fec32(BD_ENET_RX_INT);
 		}
 
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 	}
 
 	/* Set the last buffer to wrap. */
-	bdp = fec_enet_get_prevdesc(bdp, fep, queue);
+	bdp = fec_enet_get_prevdesc(bdp, &rxq->bd);
 	bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 	return 0;
 
@@ -2808,8 +2759,8 @@ fec_enet_alloc_txq_buffers(struct net_device *ndev, unsigned int queue)
 	struct fec_enet_priv_tx_q *txq;
 
 	txq = fep->tx_queue[queue];
-	bdp = txq->tx_bd_base;
-	for (i = 0; i < txq->tx_ring_size; i++) {
+	bdp = txq->bd.base;
+	for (i = 0; i < txq->bd.ring_size; i++) {
 		txq->tx_bounce[i] = kmalloc(FEC_ENET_TX_FRSIZE, GFP_KERNEL);
 		if (!txq->tx_bounce[i])
 			goto err_alloc;
@@ -2822,11 +2773,11 @@ fec_enet_alloc_txq_buffers(struct net_device *ndev, unsigned int queue)
 			ebdp->cbd_esc = cpu_to_fec32(BD_ENET_TX_INT);
 		}
 
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 	}
 
 	/* Set the last buffer to wrap. */
-	bdp = fec_enet_get_prevdesc(bdp, fep, queue);
+	bdp = fec_enet_get_prevdesc(bdp, &txq->bd);
 	bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 
 	return 0;
@@ -3117,13 +3068,15 @@ static const struct net_device_ops fec_netdev_ops = {
 static int fec_enet_init(struct net_device *ndev)
 {
 	struct fec_enet_private *fep = netdev_priv(ndev);
-	struct fec_enet_priv_tx_q *txq;
-	struct fec_enet_priv_rx_q *rxq;
 	struct bufdesc *cbd_base;
 	dma_addr_t bd_dma;
 	int bd_size;
 	unsigned int i;
+	unsigned dsize = fep->bufdesc_ex ? sizeof(struct bufdesc_ex) :
+			sizeof(struct bufdesc);
+	unsigned dsize_log2 = __fls(dsize);
 
+	WARN_ON(dsize != (1 << dsize_log2));
 #if defined(CONFIG_ARM)
 	fep->rx_align = 0xf;
 	fep->tx_align = 0xf;
@@ -3134,12 +3087,7 @@ static int fec_enet_init(struct net_device *ndev)
 
 	fec_enet_alloc_queue(ndev);
 
-	if (fep->bufdesc_ex)
-		fep->bufdesc_size = sizeof(struct bufdesc_ex);
-	else
-		fep->bufdesc_size = sizeof(struct bufdesc);
-	bd_size = (fep->total_tx_ring_size + fep->total_rx_ring_size) *
-			fep->bufdesc_size;
+	bd_size = (fep->total_tx_ring_size + fep->total_rx_ring_size) * dsize;
 
 	/* Allocate memory for buffer descriptors. */
 	cbd_base = dmam_alloc_coherent(&fep->pdev->dev, bd_size, &bd_dma,
@@ -3157,33 +3105,33 @@ static int fec_enet_init(struct net_device *ndev)
 
 	/* Set receive and transmit descriptor base. */
 	for (i = 0; i < fep->num_rx_queues; i++) {
-		rxq = fep->rx_queue[i];
-		rxq->index = i;
-		rxq->rx_bd_base = (struct bufdesc *)cbd_base;
-		rxq->bd_dma = bd_dma;
-		if (fep->bufdesc_ex) {
-			bd_dma += sizeof(struct bufdesc_ex) * rxq->rx_ring_size;
-			cbd_base = (struct bufdesc *)
-				(((struct bufdesc_ex *)cbd_base) + rxq->rx_ring_size);
-		} else {
-			bd_dma += sizeof(struct bufdesc) * rxq->rx_ring_size;
-			cbd_base += rxq->rx_ring_size;
-		}
+		struct fec_enet_priv_rx_q *rxq = fep->rx_queue[i];
+		unsigned size = dsize * rxq->bd.ring_size;
+
+		rxq->bd.qid = i;
+		rxq->bd.base = cbd_base;
+		rxq->bd.cur = cbd_base;
+		rxq->bd.dma = bd_dma;
+		rxq->bd.dsize = dsize;
+		rxq->bd.dsize_log2 = dsize_log2;
+		bd_dma += size;
+		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
+		rxq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
 	}
 
 	for (i = 0; i < fep->num_tx_queues; i++) {
-		txq = fep->tx_queue[i];
-		txq->index = i;
-		txq->tx_bd_base = (struct bufdesc *)cbd_base;
-		txq->bd_dma = bd_dma;
-		if (fep->bufdesc_ex) {
-			bd_dma += sizeof(struct bufdesc_ex) * txq->tx_ring_size;
-			cbd_base = (struct bufdesc *)
-			 (((struct bufdesc_ex *)cbd_base) + txq->tx_ring_size);
-		} else {
-			bd_dma += sizeof(struct bufdesc) * txq->tx_ring_size;
-			cbd_base += txq->tx_ring_size;
-		}
+		struct fec_enet_priv_tx_q *txq = fep->tx_queue[i];
+		unsigned size = dsize * txq->bd.ring_size;
+
+		txq->bd.qid = i;
+		txq->bd.base = cbd_base;
+		txq->bd.cur = cbd_base;
+		txq->bd.dma = bd_dma;
+		txq->bd.dsize = dsize;
+		txq->bd.dsize_log2 = dsize_log2;
+		bd_dma += size;
+		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
+		txq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
 	}
 
 
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 4/8] net: fec: add struct bufdesc_prop
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

This reduces code and gains speed.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      |  29 ++-
 drivers/net/ethernet/freescale/fec_main.c | 288 ++++++++++++------------------
 2 files changed, 132 insertions(+), 185 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index cc9677a..53ec04f 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -448,33 +448,34 @@ struct bufdesc_ex {
 /* Controller supports RACC register */
 #define FEC_QUIRK_HAS_RACC		(1 << 12)
 
+struct bufdesc_prop {
+	int qid;
+	/* Address of Rx and Tx buffers */
+	struct bufdesc	*base;
+	struct bufdesc	*last;
+	struct bufdesc	*cur;
+	dma_addr_t	dma;
+	unsigned short ring_size;
+	unsigned char dsize;
+	unsigned char dsize_log2;
+};
+
 struct fec_enet_priv_tx_q {
-	int index;
+	struct bufdesc_prop bd;
 	unsigned char *tx_bounce[TX_RING_SIZE];
 	struct  sk_buff *tx_skbuff[TX_RING_SIZE];
 
-	dma_addr_t	bd_dma;
-	struct bufdesc	*tx_bd_base;
-	uint tx_ring_size;
-
 	unsigned short tx_stop_threshold;
 	unsigned short tx_wake_threshold;
 
-	struct bufdesc	*cur_tx;
 	struct bufdesc	*dirty_tx;
 	char *tso_hdrs;
 	dma_addr_t tso_hdrs_dma;
 };
 
 struct fec_enet_priv_rx_q {
-	int index;
+	struct bufdesc_prop bd;
 	struct  sk_buff *rx_skbuff[RX_RING_SIZE];
-
-	dma_addr_t	bd_dma;
-	struct bufdesc	*rx_bd_base;
-	uint rx_ring_size;
-
-	struct bufdesc	*cur_rx;
 };
 
 /* The FEC buffer descriptors track the ring buffers.  The rx_bd_base and
@@ -514,8 +515,6 @@ struct fec_enet_private {
 	unsigned long work_ts;
 	unsigned long work_mdio;
 
-	unsigned short bufdesc_size;
-
 	struct	platform_device *pdev;
 
 	int	dev_id;
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index adbddfd..b039288 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -217,86 +217,38 @@ MODULE_PARM_DESC(macaddr, "FEC Ethernet MAC address");
 
 #define IS_TSO_HEADER(txq, addr) \
 	((addr >= txq->tso_hdrs_dma) && \
-	(addr < txq->tso_hdrs_dma + txq->tx_ring_size * TSO_HEADER_SIZE))
+	(addr < txq->tso_hdrs_dma + txq->bd.ring_size * TSO_HEADER_SIZE))
 
 static int mii_cnt;
 
-static inline
-struct bufdesc *fec_enet_get_nextdesc(struct bufdesc *bdp,
-				      struct fec_enet_private *fep,
-				      int queue_id)
-{
-	struct bufdesc *new_bd = bdp + 1;
-	struct bufdesc_ex *ex_new_bd = (struct bufdesc_ex *)bdp + 1;
-	struct fec_enet_priv_tx_q *txq = fep->tx_queue[queue_id];
-	struct fec_enet_priv_rx_q *rxq = fep->rx_queue[queue_id];
-	struct bufdesc_ex *ex_base;
-	struct bufdesc *base;
-	int ring_size;
-
-	if (bdp >= txq->tx_bd_base) {
-		base = txq->tx_bd_base;
-		ring_size = txq->tx_ring_size;
-		ex_base = (struct bufdesc_ex *)txq->tx_bd_base;
-	} else {
-		base = rxq->rx_bd_base;
-		ring_size = rxq->rx_ring_size;
-		ex_base = (struct bufdesc_ex *)rxq->rx_bd_base;
-	}
-
-	if (fep->bufdesc_ex)
-		return (struct bufdesc *)((ex_new_bd >= (ex_base + ring_size)) ?
-			ex_base : ex_new_bd);
-	else
-		return (new_bd >= (base + ring_size)) ?
-			base : new_bd;
-}
-
-static inline
-struct bufdesc *fec_enet_get_prevdesc(struct bufdesc *bdp,
-				      struct fec_enet_private *fep,
-				      int queue_id)
-{
-	struct bufdesc *new_bd = bdp - 1;
-	struct bufdesc_ex *ex_new_bd = (struct bufdesc_ex *)bdp - 1;
-	struct fec_enet_priv_tx_q *txq = fep->tx_queue[queue_id];
-	struct fec_enet_priv_rx_q *rxq = fep->rx_queue[queue_id];
-	struct bufdesc_ex *ex_base;
-	struct bufdesc *base;
-	int ring_size;
-
-	if (bdp >= txq->tx_bd_base) {
-		base = txq->tx_bd_base;
-		ring_size = txq->tx_ring_size;
-		ex_base = (struct bufdesc_ex *)txq->tx_bd_base;
-	} else {
-		base = rxq->rx_bd_base;
-		ring_size = rxq->rx_ring_size;
-		ex_base = (struct bufdesc_ex *)rxq->rx_bd_base;
-	}
+static struct bufdesc *fec_enet_get_nextdesc(struct bufdesc *bdp,
+					     struct bufdesc_prop *bd)
+{
+	return (bdp >= bd->last) ? bd->base
+			: (struct bufdesc *)(((unsigned)bdp) + bd->dsize);
+}
 
-	if (fep->bufdesc_ex)
-		return (struct bufdesc *)((ex_new_bd < ex_base) ?
-			(ex_new_bd + ring_size) : ex_new_bd);
-	else
-		return (new_bd < base) ? (new_bd + ring_size) : new_bd;
+static struct bufdesc *fec_enet_get_prevdesc(struct bufdesc *bdp,
+					     struct bufdesc_prop *bd)
+{
+	return (bdp <= bd->base) ? bd->last
+			: (struct bufdesc *)(((unsigned)bdp) - bd->dsize);
 }
 
-static int fec_enet_get_bd_index(struct bufdesc *base, struct bufdesc *bdp,
-				struct fec_enet_private *fep)
+static int fec_enet_get_bd_index(struct bufdesc *bdp,
+				 struct bufdesc_prop *bd)
 {
-	return ((const char *)bdp - (const char *)base) / fep->bufdesc_size;
+	return ((const char *)bdp - (const char *)bd->base) >> bd->dsize_log2;
 }
 
-static int fec_enet_get_free_txdesc_num(struct fec_enet_private *fep,
-					struct fec_enet_priv_tx_q *txq)
+static int fec_enet_get_free_txdesc_num(struct fec_enet_priv_tx_q *txq)
 {
 	int entries;
 
-	entries = ((const char *)txq->dirty_tx -
-			(const char *)txq->cur_tx) / fep->bufdesc_size - 1;
+	entries = (((const char *)txq->dirty_tx -
+			(const char *)txq->bd.cur) >> txq->bd.dsize_log2) - 1;
 
-	return entries >= 0 ? entries : entries + txq->tx_ring_size;
+	return entries >= 0 ? entries : entries + txq->bd.ring_size;
 }
 
 static void swap_buffer(void *bufaddr, int len)
@@ -329,20 +281,20 @@ static void fec_dump(struct net_device *ndev)
 	pr_info("Nr     SC     addr       len  SKB\n");
 
 	txq = fep->tx_queue[0];
-	bdp = txq->tx_bd_base;
+	bdp = txq->bd.base;
 
 	do {
 		pr_info("%3u %c%c 0x%04x 0x%08x %4u %p\n",
 			index,
-			bdp == txq->cur_tx ? 'S' : ' ',
+			bdp == txq->bd.cur ? 'S' : ' ',
 			bdp == txq->dirty_tx ? 'H' : ' ',
 			fec16_to_cpu(bdp->cbd_sc),
 			fec32_to_cpu(bdp->cbd_bufaddr),
 			fec16_to_cpu(bdp->cbd_datlen),
 			txq->tx_skbuff[index]);
-		bdp = fec_enet_get_nextdesc(bdp, fep, 0);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		index++;
-	} while (bdp != txq->tx_bd_base);
+	} while (bdp != txq->bd.base);
 }
 
 static inline bool is_ipv4_pkt(struct sk_buff *skb)
@@ -373,7 +325,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 			     struct net_device *ndev)
 {
 	struct fec_enet_private *fep = netdev_priv(ndev);
-	struct bufdesc *bdp = txq->cur_tx;
+	struct bufdesc *bdp = txq->bd.cur;
 	struct bufdesc_ex *ebdp;
 	int nr_frags = skb_shinfo(skb)->nr_frags;
 	unsigned short queue = skb_get_queue_mapping(skb);
@@ -388,7 +340,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 	for (frag = 0; frag < nr_frags; frag++) {
 		this_frag = &skb_shinfo(skb)->frags[frag];
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		ebdp = (struct bufdesc_ex *)bdp;
 
 		status = fec16_to_cpu(bdp->cbd_sc);
@@ -418,7 +370,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 		bufaddr = page_address(this_frag->page.p) + this_frag->page_offset;
 
-		index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &txq->bd);
 		if (((unsigned long) bufaddr) & fep->tx_align ||
 			fep->quirks & FEC_QUIRK_SWAP_FRAME) {
 			memcpy(txq->tx_bounce[index], bufaddr, frag_len);
@@ -444,9 +396,9 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 	return bdp;
 dma_mapping_error:
-	bdp = txq->cur_tx;
+	bdp = txq->bd.cur;
 	for (i = 0; i < frag; i++) {
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		dma_unmap_single(&fep->pdev->dev, fec32_to_cpu(bdp->cbd_bufaddr),
 				 fec16_to_cpu(bdp->cbd_datlen), DMA_TO_DEVICE);
 	}
@@ -468,7 +420,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	unsigned int index;
 	int entries_free;
 
-	entries_free = fec_enet_get_free_txdesc_num(fep, txq);
+	entries_free = fec_enet_get_free_txdesc_num(txq);
 	if (entries_free < MAX_SKB_FRAGS + 1) {
 		dev_kfree_skb_any(skb);
 		if (net_ratelimit())
@@ -483,7 +435,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	}
 
 	/* Fill in a Tx ring entry */
-	bdp = txq->cur_tx;
+	bdp = txq->bd.cur;
 	last_bdp = bdp;
 	status = fec16_to_cpu(bdp->cbd_sc);
 	status &= ~BD_ENET_TX_STATS;
@@ -493,7 +445,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	buflen = skb_headlen(skb);
 
 	queue = skb_get_queue_mapping(skb);
-	index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+	index = fec_enet_get_bd_index(bdp, &txq->bd);
 	if (((unsigned long) bufaddr) & fep->tx_align ||
 		fep->quirks & FEC_QUIRK_SWAP_FRAME) {
 		memcpy(txq->tx_bounce[index], skb->data, buflen);
@@ -544,7 +496,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 		ebdp->cbd_esc = cpu_to_fec32(estatus);
 	}
 
-	index = fec_enet_get_bd_index(txq->tx_bd_base, last_bdp, fep);
+	index = fec_enet_get_bd_index(last_bdp, &txq->bd);
 	/* Save skb pointer */
 	txq->tx_skbuff[index] = skb;
 
@@ -558,15 +510,15 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	bdp->cbd_sc = cpu_to_fec16(status);
 
 	/* If this was the last BD in the ring, start at the beginning again. */
-	bdp = fec_enet_get_nextdesc(last_bdp, fep, queue);
+	bdp = fec_enet_get_nextdesc(last_bdp, &txq->bd);
 
 	skb_tx_timestamp(skb);
 
 	/* Make sure the update to bdp and tx_skbuff are performed before
-	 * cur_tx.
+	 * txq->bd.cur.
 	 */
 	wmb();
-	txq->cur_tx = bdp;
+	txq->bd.cur = bdp;
 
 	/* Trigger transmission start */
 	writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue));
@@ -697,13 +649,13 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 	struct fec_enet_private *fep = netdev_priv(ndev);
 	int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
 	int total_len, data_left;
-	struct bufdesc *bdp = txq->cur_tx;
+	struct bufdesc *bdp = txq->bd.cur;
 	unsigned short queue = skb_get_queue_mapping(skb);
 	struct tso_t tso;
 	unsigned int index = 0;
 	int ret;
 
-	if (tso_count_descs(skb) >= fec_enet_get_free_txdesc_num(fep, txq)) {
+	if (tso_count_descs(skb) >= fec_enet_get_free_txdesc_num(txq)) {
 		dev_kfree_skb_any(skb);
 		if (net_ratelimit())
 			netdev_err(ndev, "NOT enough BD for TSO!\n");
@@ -723,7 +675,7 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 	while (total_len > 0) {
 		char *hdr;
 
-		index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &txq->bd);
 		data_left = min_t(int, skb_shinfo(skb)->gso_size, total_len);
 		total_len -= data_left;
 
@@ -738,9 +690,8 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 			int size;
 
 			size = min_t(int, tso.size, data_left);
-			bdp = fec_enet_get_nextdesc(bdp, fep, queue);
-			index = fec_enet_get_bd_index(txq->tx_bd_base,
-						      bdp, fep);
+			bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
+			index = fec_enet_get_bd_index(bdp, &txq->bd);
 			ret = fec_enet_txq_put_data_tso(txq, skb, ndev,
 							bdp, index,
 							tso.data, size,
@@ -753,14 +704,14 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 			tso_build_data(skb, &tso, size);
 		}
 
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 	}
 
 	/* Save skb pointer */
 	txq->tx_skbuff[index] = skb;
 
 	skb_tx_timestamp(skb);
-	txq->cur_tx = bdp;
+	txq->bd.cur = bdp;
 
 	/* Trigger transmission start */
 	if (!(fep->quirks & FEC_QUIRK_ERR007885) ||
@@ -798,7 +749,7 @@ fec_enet_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 	if (ret)
 		return ret;
 
-	entries_free = fec_enet_get_free_txdesc_num(fep, txq);
+	entries_free = fec_enet_get_free_txdesc_num(txq);
 	if (entries_free <= txq->tx_stop_threshold)
 		netif_tx_stop_queue(nq);
 
@@ -819,32 +770,32 @@ static void fec_enet_bd_init(struct net_device *dev)
 	for (q = 0; q < fep->num_rx_queues; q++) {
 		/* Initialize the receive buffer descriptors. */
 		rxq = fep->rx_queue[q];
-		bdp = rxq->rx_bd_base;
+		bdp = rxq->bd.base;
 
-		for (i = 0; i < rxq->rx_ring_size; i++) {
+		for (i = 0; i < rxq->bd.ring_size; i++) {
 
 			/* Initialize the BD for every fragment in the page. */
 			if (bdp->cbd_bufaddr)
 				bdp->cbd_sc = cpu_to_fec16(BD_ENET_RX_EMPTY);
 			else
 				bdp->cbd_sc = cpu_to_fec16(0);
-			bdp = fec_enet_get_nextdesc(bdp, fep, q);
+			bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 		}
 
 		/* Set the last buffer to wrap */
-		bdp = fec_enet_get_prevdesc(bdp, fep, q);
+		bdp = fec_enet_get_prevdesc(bdp, &rxq->bd);
 		bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 
-		rxq->cur_rx = rxq->rx_bd_base;
+		rxq->bd.cur = rxq->bd.base;
 	}
 
 	for (q = 0; q < fep->num_tx_queues; q++) {
 		/* ...and the same for transmit */
 		txq = fep->tx_queue[q];
-		bdp = txq->tx_bd_base;
-		txq->cur_tx = bdp;
+		bdp = txq->bd.base;
+		txq->bd.cur = bdp;
 
-		for (i = 0; i < txq->tx_ring_size; i++) {
+		for (i = 0; i < txq->bd.ring_size; i++) {
 			/* Initialize the BD for every fragment in the page. */
 			bdp->cbd_sc = cpu_to_fec16(0);
 			if (txq->tx_skbuff[i]) {
@@ -852,11 +803,11 @@ static void fec_enet_bd_init(struct net_device *dev)
 				txq->tx_skbuff[i] = NULL;
 			}
 			bdp->cbd_bufaddr = cpu_to_fec32(0);
-			bdp = fec_enet_get_nextdesc(bdp, fep, q);
+			bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 		}
 
 		/* Set the last buffer to wrap */
-		bdp = fec_enet_get_prevdesc(bdp, fep, q);
+		bdp = fec_enet_get_prevdesc(bdp, &txq->bd);
 		bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 		txq->dirty_tx = bdp;
 	}
@@ -880,7 +831,7 @@ static void fec_enet_enable_ring(struct net_device *ndev)
 
 	for (i = 0; i < fep->num_rx_queues; i++) {
 		rxq = fep->rx_queue[i];
-		writel(rxq->bd_dma, fep->hwp + FEC_R_DES_START(i));
+		writel(rxq->bd.dma, fep->hwp + FEC_R_DES_START(i));
 		writel(PKT_MAXBLR_SIZE, fep->hwp + FEC_R_BUFF_SIZE(i));
 
 		/* enable DMA1/2 */
@@ -891,7 +842,7 @@ static void fec_enet_enable_ring(struct net_device *ndev)
 
 	for (i = 0; i < fep->num_tx_queues; i++) {
 		txq = fep->tx_queue[i];
-		writel(txq->bd_dma, fep->hwp + FEC_X_DES_START(i));
+		writel(txq->bd.dma, fep->hwp + FEC_X_DES_START(i));
 
 		/* enable DMA1/2 */
 		if (i)
@@ -909,7 +860,7 @@ static void fec_enet_reset_skb(struct net_device *ndev)
 	for (i = 0; i < fep->num_tx_queues; i++) {
 		txq = fep->tx_queue[i];
 
-		for (j = 0; j < txq->tx_ring_size; j++) {
+		for (j = 0; j < txq->bd.ring_size; j++) {
 			if (txq->tx_skbuff[j]) {
 				dev_kfree_skb_any(txq->tx_skbuff[j]);
 				txq->tx_skbuff[j] = NULL;
@@ -1222,16 +1173,16 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 	bdp = txq->dirty_tx;
 
 	/* get next bdp of dirty_tx */
-	bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+	bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 
-	while (bdp != READ_ONCE(txq->cur_tx)) {
-		/* Order the load of cur_tx and cbd_sc */
+	while (bdp != READ_ONCE(txq->bd.cur)) {
+		/* Order the load of bd.cur and cbd_sc */
 		rmb();
 		status = fec16_to_cpu(READ_ONCE(bdp->cbd_sc));
 		if (status & BD_ENET_TX_READY)
 			break;
 
-		index = fec_enet_get_bd_index(txq->tx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &txq->bd);
 
 		skb = txq->tx_skbuff[index];
 		txq->tx_skbuff[index] = NULL;
@@ -1242,7 +1193,7 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 					 DMA_TO_DEVICE);
 		bdp->cbd_bufaddr = cpu_to_fec32(0);
 		if (!skb) {
-			bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+			bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 			continue;
 		}
 
@@ -1291,19 +1242,19 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 		txq->dirty_tx = bdp;
 
 		/* Update pointer to next buffer descriptor to be transmitted */
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 
 		/* Since we have freed up a buffer, the ring is no longer full
 		 */
 		if (netif_queue_stopped(ndev)) {
-			entries_free = fec_enet_get_free_txdesc_num(fep, txq);
+			entries_free = fec_enet_get_free_txdesc_num(txq);
 			if (entries_free >= txq->tx_wake_threshold)
 				netif_tx_wake_queue(nq);
 		}
 	}
 
 	/* ERR006538: Keep the transmitter going */
-	if (bdp != txq->cur_tx &&
+	if (bdp != txq->bd.cur &&
 	    readl(fep->hwp + FEC_X_DES_ACTIVE(queue_id)) == 0)
 		writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue_id));
 }
@@ -1367,7 +1318,7 @@ static bool fec_enet_copybreak(struct net_device *ndev, struct sk_buff **skb,
 	return true;
 }
 
-/* During a receive, the cur_rx points to the current incoming buffer.
+/* During a receive, the bd_rx.cur points to the current incoming buffer.
  * When we update through the ring, if the next incoming buffer has
  * not been given to the system, we just set the empty indicator,
  * effectively tossing the packet.
@@ -1400,7 +1351,7 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
 	/* First, grab all of the stats for the incoming packet.
 	 * These get messed up if we get called due to a busy condition.
 	 */
-	bdp = rxq->cur_rx;
+	bdp = rxq->bd.cur;
 
 	while (!((status = fec16_to_cpu(bdp->cbd_sc)) & BD_ENET_RX_EMPTY)) {
 
@@ -1441,7 +1392,7 @@ fec_enet_rx_queue(struct net_device *ndev, int budget, u16 queue_id)
 		pkt_len = fec16_to_cpu(bdp->cbd_datlen);
 		ndev->stats.rx_bytes += pkt_len;
 
-		index = fec_enet_get_bd_index(rxq->rx_bd_base, bdp, fep);
+		index = fec_enet_get_bd_index(bdp, &rxq->bd);
 		skb = rxq->rx_skbuff[index];
 
 		/* The packet length includes FCS, but we don't want to
@@ -1541,7 +1492,7 @@ rx_processing_done:
 		}
 
 		/* Update BD pointer to next entry */
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue_id);
+		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 
 		/* Doing this here will keep the FEC running while we process
 		 * incoming frames.  On a heavily loaded network, we should be
@@ -1549,7 +1500,7 @@ rx_processing_done:
 		 */
 		writel(0, fep->hwp + FEC_R_DES_ACTIVE(queue_id));
 	}
-	rxq->cur_rx = bdp;
+	rxq->bd.cur = bdp;
 	return pkt_received;
 }
 
@@ -2658,8 +2609,8 @@ static void fec_enet_free_buffers(struct net_device *ndev)
 
 	for (q = 0; q < fep->num_rx_queues; q++) {
 		rxq = fep->rx_queue[q];
-		bdp = rxq->rx_bd_base;
-		for (i = 0; i < rxq->rx_ring_size; i++) {
+		bdp = rxq->bd.base;
+		for (i = 0; i < rxq->bd.ring_size; i++) {
 			skb = rxq->rx_skbuff[i];
 			rxq->rx_skbuff[i] = NULL;
 			if (skb) {
@@ -2669,14 +2620,14 @@ static void fec_enet_free_buffers(struct net_device *ndev)
 						 DMA_FROM_DEVICE);
 				dev_kfree_skb(skb);
 			}
-			bdp = fec_enet_get_nextdesc(bdp, fep, q);
+			bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 		}
 	}
 
 	for (q = 0; q < fep->num_tx_queues; q++) {
 		txq = fep->tx_queue[q];
-		bdp = txq->tx_bd_base;
-		for (i = 0; i < txq->tx_ring_size; i++) {
+		bdp = txq->bd.base;
+		for (i = 0; i < txq->bd.ring_size; i++) {
 			kfree(txq->tx_bounce[i]);
 			txq->tx_bounce[i] = NULL;
 			skb = txq->tx_skbuff[i];
@@ -2696,7 +2647,7 @@ static void fec_enet_free_queue(struct net_device *ndev)
 		if (fep->tx_queue[i] && fep->tx_queue[i]->tso_hdrs) {
 			txq = fep->tx_queue[i];
 			dma_free_coherent(NULL,
-					  txq->tx_ring_size * TSO_HEADER_SIZE,
+					  txq->bd.ring_size * TSO_HEADER_SIZE,
 					  txq->tso_hdrs,
 					  txq->tso_hdrs_dma);
 		}
@@ -2722,15 +2673,15 @@ static int fec_enet_alloc_queue(struct net_device *ndev)
 		}
 
 		fep->tx_queue[i] = txq;
-		txq->tx_ring_size = TX_RING_SIZE;
-		fep->total_tx_ring_size += fep->tx_queue[i]->tx_ring_size;
+		txq->bd.ring_size = TX_RING_SIZE;
+		fep->total_tx_ring_size += fep->tx_queue[i]->bd.ring_size;
 
 		txq->tx_stop_threshold = FEC_MAX_SKB_DESCS;
 		txq->tx_wake_threshold =
-				(txq->tx_ring_size - txq->tx_stop_threshold) / 2;
+			(txq->bd.ring_size - txq->tx_stop_threshold) / 2;
 
 		txq->tso_hdrs = dma_alloc_coherent(NULL,
-					txq->tx_ring_size * TSO_HEADER_SIZE,
+					txq->bd.ring_size * TSO_HEADER_SIZE,
 					&txq->tso_hdrs_dma,
 					GFP_KERNEL);
 		if (!txq->tso_hdrs) {
@@ -2747,8 +2698,8 @@ static int fec_enet_alloc_queue(struct net_device *ndev)
 			goto alloc_failed;
 		}
 
-		fep->rx_queue[i]->rx_ring_size = RX_RING_SIZE;
-		fep->total_rx_ring_size += fep->rx_queue[i]->rx_ring_size;
+		fep->rx_queue[i]->bd.ring_size = RX_RING_SIZE;
+		fep->total_rx_ring_size += fep->rx_queue[i]->bd.ring_size;
 	}
 	return ret;
 
@@ -2767,8 +2718,8 @@ fec_enet_alloc_rxq_buffers(struct net_device *ndev, unsigned int queue)
 	struct fec_enet_priv_rx_q *rxq;
 
 	rxq = fep->rx_queue[queue];
-	bdp = rxq->rx_bd_base;
-	for (i = 0; i < rxq->rx_ring_size; i++) {
+	bdp = rxq->bd.base;
+	for (i = 0; i < rxq->bd.ring_size; i++) {
 		skb = netdev_alloc_skb(ndev, FEC_ENET_RX_FRSIZE);
 		if (!skb)
 			goto err_alloc;
@@ -2786,11 +2737,11 @@ fec_enet_alloc_rxq_buffers(struct net_device *ndev, unsigned int queue)
 			ebdp->cbd_esc = cpu_to_fec32(BD_ENET_RX_INT);
 		}
 
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
 	}
 
 	/* Set the last buffer to wrap. */
-	bdp = fec_enet_get_prevdesc(bdp, fep, queue);
+	bdp = fec_enet_get_prevdesc(bdp, &rxq->bd);
 	bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 	return 0;
 
@@ -2808,8 +2759,8 @@ fec_enet_alloc_txq_buffers(struct net_device *ndev, unsigned int queue)
 	struct fec_enet_priv_tx_q *txq;
 
 	txq = fep->tx_queue[queue];
-	bdp = txq->tx_bd_base;
-	for (i = 0; i < txq->tx_ring_size; i++) {
+	bdp = txq->bd.base;
+	for (i = 0; i < txq->bd.ring_size; i++) {
 		txq->tx_bounce[i] = kmalloc(FEC_ENET_TX_FRSIZE, GFP_KERNEL);
 		if (!txq->tx_bounce[i])
 			goto err_alloc;
@@ -2822,11 +2773,11 @@ fec_enet_alloc_txq_buffers(struct net_device *ndev, unsigned int queue)
 			ebdp->cbd_esc = cpu_to_fec32(BD_ENET_TX_INT);
 		}
 
-		bdp = fec_enet_get_nextdesc(bdp, fep, queue);
+		bdp = fec_enet_get_nextdesc(bdp, &txq->bd);
 	}
 
 	/* Set the last buffer to wrap. */
-	bdp = fec_enet_get_prevdesc(bdp, fep, queue);
+	bdp = fec_enet_get_prevdesc(bdp, &txq->bd);
 	bdp->cbd_sc |= cpu_to_fec16(BD_SC_WRAP);
 
 	return 0;
@@ -3117,13 +3068,15 @@ static const struct net_device_ops fec_netdev_ops = {
 static int fec_enet_init(struct net_device *ndev)
 {
 	struct fec_enet_private *fep = netdev_priv(ndev);
-	struct fec_enet_priv_tx_q *txq;
-	struct fec_enet_priv_rx_q *rxq;
 	struct bufdesc *cbd_base;
 	dma_addr_t bd_dma;
 	int bd_size;
 	unsigned int i;
+	unsigned dsize = fep->bufdesc_ex ? sizeof(struct bufdesc_ex) :
+			sizeof(struct bufdesc);
+	unsigned dsize_log2 = __fls(dsize);
 
+	WARN_ON(dsize != (1 << dsize_log2));
 #if defined(CONFIG_ARM)
 	fep->rx_align = 0xf;
 	fep->tx_align = 0xf;
@@ -3134,12 +3087,7 @@ static int fec_enet_init(struct net_device *ndev)
 
 	fec_enet_alloc_queue(ndev);
 
-	if (fep->bufdesc_ex)
-		fep->bufdesc_size = sizeof(struct bufdesc_ex);
-	else
-		fep->bufdesc_size = sizeof(struct bufdesc);
-	bd_size = (fep->total_tx_ring_size + fep->total_rx_ring_size) *
-			fep->bufdesc_size;
+	bd_size = (fep->total_tx_ring_size + fep->total_rx_ring_size) * dsize;
 
 	/* Allocate memory for buffer descriptors. */
 	cbd_base = dmam_alloc_coherent(&fep->pdev->dev, bd_size, &bd_dma,
@@ -3157,33 +3105,33 @@ static int fec_enet_init(struct net_device *ndev)
 
 	/* Set receive and transmit descriptor base. */
 	for (i = 0; i < fep->num_rx_queues; i++) {
-		rxq = fep->rx_queue[i];
-		rxq->index = i;
-		rxq->rx_bd_base = (struct bufdesc *)cbd_base;
-		rxq->bd_dma = bd_dma;
-		if (fep->bufdesc_ex) {
-			bd_dma += sizeof(struct bufdesc_ex) * rxq->rx_ring_size;
-			cbd_base = (struct bufdesc *)
-				(((struct bufdesc_ex *)cbd_base) + rxq->rx_ring_size);
-		} else {
-			bd_dma += sizeof(struct bufdesc) * rxq->rx_ring_size;
-			cbd_base += rxq->rx_ring_size;
-		}
+		struct fec_enet_priv_rx_q *rxq = fep->rx_queue[i];
+		unsigned size = dsize * rxq->bd.ring_size;
+
+		rxq->bd.qid = i;
+		rxq->bd.base = cbd_base;
+		rxq->bd.cur = cbd_base;
+		rxq->bd.dma = bd_dma;
+		rxq->bd.dsize = dsize;
+		rxq->bd.dsize_log2 = dsize_log2;
+		bd_dma += size;
+		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
+		rxq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
 	}
 
 	for (i = 0; i < fep->num_tx_queues; i++) {
-		txq = fep->tx_queue[i];
-		txq->index = i;
-		txq->tx_bd_base = (struct bufdesc *)cbd_base;
-		txq->bd_dma = bd_dma;
-		if (fep->bufdesc_ex) {
-			bd_dma += sizeof(struct bufdesc_ex) * txq->tx_ring_size;
-			cbd_base = (struct bufdesc *)
-			 (((struct bufdesc_ex *)cbd_base) + txq->tx_ring_size);
-		} else {
-			bd_dma += sizeof(struct bufdesc) * txq->tx_ring_size;
-			cbd_base += txq->tx_ring_size;
-		}
+		struct fec_enet_priv_tx_q *txq = fep->tx_queue[i];
+		unsigned size = dsize * txq->bd.ring_size;
+
+		txq->bd.qid = i;
+		txq->bd.base = cbd_base;
+		txq->bd.cur = cbd_base;
+		txq->bd.dma = bd_dma;
+		txq->bd.dsize = dsize;
+		txq->bd.dsize_log2 = dsize_log2;
+		bd_dma += size;
+		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
+		txq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
 	}
 
 
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 5/8] net: fec: add variable reg_desc_active to speed things up
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

There is no need for complex macros every time we need to activate
a queue. Also, no need to call skb_get_queue_mapping when we already
know which queue it is using.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      |  7 +----
 drivers/net/ethernet/freescale/fec_main.c | 44 +++++++++++++++++--------------
 2 files changed, 25 insertions(+), 26 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index 53ec04f..bedd28a 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -310,12 +310,6 @@ struct bufdesc_ex {
 #define FEC_R_BUFF_SIZE(X)	(((X) == 1) ? FEC_R_BUFF_SIZE_1 : \
 				(((X) == 2) ? \
 					FEC_R_BUFF_SIZE_2 : FEC_R_BUFF_SIZE_0))
-#define FEC_R_DES_ACTIVE(X)	(((X) == 1) ? FEC_R_DES_ACTIVE_1 : \
-				(((X) == 2) ? \
-				   FEC_R_DES_ACTIVE_2 : FEC_R_DES_ACTIVE_0))
-#define FEC_X_DES_ACTIVE(X)	(((X) == 1) ? FEC_X_DES_ACTIVE_1 : \
-				(((X) == 2) ? \
-				   FEC_X_DES_ACTIVE_2 : FEC_X_DES_ACTIVE_0))
 
 #define FEC_DMA_CFG(X)		(((X) == 2) ? FEC_DMA_CFG_2 : FEC_DMA_CFG_1)
 
@@ -454,6 +448,7 @@ struct bufdesc_prop {
 	struct bufdesc	*base;
 	struct bufdesc	*last;
 	struct bufdesc	*cur;
+	void __iomem	*reg_desc_active;
 	dma_addr_t	dma;
 	unsigned short ring_size;
 	unsigned char dsize;
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index b039288..712e3bb 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -328,7 +328,6 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 	struct bufdesc *bdp = txq->bd.cur;
 	struct bufdesc_ex *ebdp;
 	int nr_frags = skb_shinfo(skb)->nr_frags;
-	unsigned short queue = skb_get_queue_mapping(skb);
 	int frag, frag_len;
 	unsigned short status;
 	unsigned int estatus = 0;
@@ -361,7 +360,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 		if (fep->bufdesc_ex) {
 			if (fep->quirks & FEC_QUIRK_HAS_AVB)
-				estatus |= FEC_TX_BD_FTYPE(queue);
+				estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 			if (skb->ip_summed == CHECKSUM_PARTIAL)
 				estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
 			ebdp->cbd_bdu = 0;
@@ -415,7 +414,6 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	dma_addr_t addr;
 	unsigned short status;
 	unsigned short buflen;
-	unsigned short queue;
 	unsigned int estatus = 0;
 	unsigned int index;
 	int entries_free;
@@ -444,7 +442,6 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	bufaddr = skb->data;
 	buflen = skb_headlen(skb);
 
-	queue = skb_get_queue_mapping(skb);
 	index = fec_enet_get_bd_index(bdp, &txq->bd);
 	if (((unsigned long) bufaddr) & fep->tx_align ||
 		fep->quirks & FEC_QUIRK_SWAP_FRAME) {
@@ -487,7 +484,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
 
 		if (fep->quirks & FEC_QUIRK_HAS_AVB)
-			estatus |= FEC_TX_BD_FTYPE(queue);
+			estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 
 		if (skb->ip_summed == CHECKSUM_PARTIAL)
 			estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
@@ -521,7 +518,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	txq->bd.cur = bdp;
 
 	/* Trigger transmission start */
-	writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue));
+	writel(0, txq->bd.reg_desc_active);
 
 	return 0;
 }
@@ -534,7 +531,6 @@ fec_enet_txq_put_data_tso(struct fec_enet_priv_tx_q *txq, struct sk_buff *skb,
 {
 	struct fec_enet_private *fep = netdev_priv(ndev);
 	struct bufdesc_ex *ebdp = container_of(bdp, struct bufdesc_ex, desc);
-	unsigned short queue = skb_get_queue_mapping(skb);
 	unsigned short status;
 	unsigned int estatus = 0;
 	dma_addr_t addr;
@@ -566,7 +562,7 @@ fec_enet_txq_put_data_tso(struct fec_enet_priv_tx_q *txq, struct sk_buff *skb,
 
 	if (fep->bufdesc_ex) {
 		if (fep->quirks & FEC_QUIRK_HAS_AVB)
-			estatus |= FEC_TX_BD_FTYPE(queue);
+			estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 		if (skb->ip_summed == CHECKSUM_PARTIAL)
 			estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
 		ebdp->cbd_bdu = 0;
@@ -595,7 +591,6 @@ fec_enet_txq_put_hdr_tso(struct fec_enet_priv_tx_q *txq,
 	struct fec_enet_private *fep = netdev_priv(ndev);
 	int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
 	struct bufdesc_ex *ebdp = container_of(bdp, struct bufdesc_ex, desc);
-	unsigned short queue = skb_get_queue_mapping(skb);
 	void *bufaddr;
 	unsigned long dmabuf;
 	unsigned short status;
@@ -630,7 +625,7 @@ fec_enet_txq_put_hdr_tso(struct fec_enet_priv_tx_q *txq,
 
 	if (fep->bufdesc_ex) {
 		if (fep->quirks & FEC_QUIRK_HAS_AVB)
-			estatus |= FEC_TX_BD_FTYPE(queue);
+			estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 		if (skb->ip_summed == CHECKSUM_PARTIAL)
 			estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
 		ebdp->cbd_bdu = 0;
@@ -650,7 +645,6 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 	int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
 	int total_len, data_left;
 	struct bufdesc *bdp = txq->bd.cur;
-	unsigned short queue = skb_get_queue_mapping(skb);
 	struct tso_t tso;
 	unsigned int index = 0;
 	int ret;
@@ -715,11 +709,11 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 
 	/* Trigger transmission start */
 	if (!(fep->quirks & FEC_QUIRK_ERR007885) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)))
-		writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue));
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active))
+		writel(0, txq->bd.reg_desc_active);
 
 	return 0;
 
@@ -819,7 +813,7 @@ static void fec_enet_active_rxring(struct net_device *ndev)
 	int i;
 
 	for (i = 0; i < fep->num_rx_queues; i++)
-		writel(0, fep->hwp + FEC_R_DES_ACTIVE(i));
+		writel(0, fep->rx_queue[i]->bd.reg_desc_active);
 }
 
 static void fec_enet_enable_ring(struct net_device *ndev)
@@ -1255,8 +1249,8 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 
 	/* ERR006538: Keep the transmitter going */
 	if (bdp != txq->bd.cur &&
-	    readl(fep->hwp + FEC_X_DES_ACTIVE(queue_id)) == 0)
-		writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue_id));
+	    readl(txq->bd.reg_desc_active) == 0)
+		writel(0, txq->bd.reg_desc_active);
 }
 
 static void
@@ -1498,7 +1492,7 @@ rx_processing_done:
 		 * incoming frames.  On a heavily loaded network, we should be
 		 * able to keep up at the expense of system resources.
 		 */
-		writel(0, fep->hwp + FEC_R_DES_ACTIVE(queue_id));
+		writel(0, rxq->bd.reg_desc_active);
 	}
 	rxq->bd.cur = bdp;
 	return pkt_received;
@@ -3061,6 +3055,14 @@ static const struct net_device_ops fec_netdev_ops = {
 	.ndo_set_features	= fec_set_features,
 };
 
+static const unsigned short offset_des_active_rxq[] = {
+	FEC_R_DES_ACTIVE_0, FEC_R_DES_ACTIVE_1, FEC_R_DES_ACTIVE_2
+};
+
+static const unsigned short offset_des_active_txq[] = {
+	FEC_X_DES_ACTIVE_0, FEC_X_DES_ACTIVE_1, FEC_X_DES_ACTIVE_2
+};
+
  /*
   * XXX:  We need to clean up on failure exits here.
   *
@@ -3114,6 +3116,7 @@ static int fec_enet_init(struct net_device *ndev)
 		rxq->bd.dma = bd_dma;
 		rxq->bd.dsize = dsize;
 		rxq->bd.dsize_log2 = dsize_log2;
+		rxq->bd.reg_desc_active = fep->hwp + offset_des_active_rxq[i];
 		bd_dma += size;
 		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
 		rxq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
@@ -3129,6 +3132,7 @@ static int fec_enet_init(struct net_device *ndev)
 		txq->bd.dma = bd_dma;
 		txq->bd.dsize = dsize;
 		txq->bd.dsize_log2 = dsize_log2;
+		txq->bd.reg_desc_active = fep->hwp + offset_des_active_txq[i];
 		bd_dma += size;
 		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
 		txq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 5/8] net: fec: add variable reg_desc_active to speed things up
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

There is no need for complex macros every time we need to activate
a queue. Also, no need to call skb_get_queue_mapping when we already
know which queue it is using.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      |  7 +----
 drivers/net/ethernet/freescale/fec_main.c | 44 +++++++++++++++++--------------
 2 files changed, 25 insertions(+), 26 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index 53ec04f..bedd28a 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -310,12 +310,6 @@ struct bufdesc_ex {
 #define FEC_R_BUFF_SIZE(X)	(((X) == 1) ? FEC_R_BUFF_SIZE_1 : \
 				(((X) == 2) ? \
 					FEC_R_BUFF_SIZE_2 : FEC_R_BUFF_SIZE_0))
-#define FEC_R_DES_ACTIVE(X)	(((X) == 1) ? FEC_R_DES_ACTIVE_1 : \
-				(((X) == 2) ? \
-				   FEC_R_DES_ACTIVE_2 : FEC_R_DES_ACTIVE_0))
-#define FEC_X_DES_ACTIVE(X)	(((X) == 1) ? FEC_X_DES_ACTIVE_1 : \
-				(((X) == 2) ? \
-				   FEC_X_DES_ACTIVE_2 : FEC_X_DES_ACTIVE_0))
 
 #define FEC_DMA_CFG(X)		(((X) == 2) ? FEC_DMA_CFG_2 : FEC_DMA_CFG_1)
 
@@ -454,6 +448,7 @@ struct bufdesc_prop {
 	struct bufdesc	*base;
 	struct bufdesc	*last;
 	struct bufdesc	*cur;
+	void __iomem	*reg_desc_active;
 	dma_addr_t	dma;
 	unsigned short ring_size;
 	unsigned char dsize;
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index b039288..712e3bb 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -328,7 +328,6 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 	struct bufdesc *bdp = txq->bd.cur;
 	struct bufdesc_ex *ebdp;
 	int nr_frags = skb_shinfo(skb)->nr_frags;
-	unsigned short queue = skb_get_queue_mapping(skb);
 	int frag, frag_len;
 	unsigned short status;
 	unsigned int estatus = 0;
@@ -361,7 +360,7 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 		if (fep->bufdesc_ex) {
 			if (fep->quirks & FEC_QUIRK_HAS_AVB)
-				estatus |= FEC_TX_BD_FTYPE(queue);
+				estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 			if (skb->ip_summed == CHECKSUM_PARTIAL)
 				estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
 			ebdp->cbd_bdu = 0;
@@ -415,7 +414,6 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	dma_addr_t addr;
 	unsigned short status;
 	unsigned short buflen;
-	unsigned short queue;
 	unsigned int estatus = 0;
 	unsigned int index;
 	int entries_free;
@@ -444,7 +442,6 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	bufaddr = skb->data;
 	buflen = skb_headlen(skb);
 
-	queue = skb_get_queue_mapping(skb);
 	index = fec_enet_get_bd_index(bdp, &txq->bd);
 	if (((unsigned long) bufaddr) & fep->tx_align ||
 		fep->quirks & FEC_QUIRK_SWAP_FRAME) {
@@ -487,7 +484,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 			skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
 
 		if (fep->quirks & FEC_QUIRK_HAS_AVB)
-			estatus |= FEC_TX_BD_FTYPE(queue);
+			estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 
 		if (skb->ip_summed == CHECKSUM_PARTIAL)
 			estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
@@ -521,7 +518,7 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	txq->bd.cur = bdp;
 
 	/* Trigger transmission start */
-	writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue));
+	writel(0, txq->bd.reg_desc_active);
 
 	return 0;
 }
@@ -534,7 +531,6 @@ fec_enet_txq_put_data_tso(struct fec_enet_priv_tx_q *txq, struct sk_buff *skb,
 {
 	struct fec_enet_private *fep = netdev_priv(ndev);
 	struct bufdesc_ex *ebdp = container_of(bdp, struct bufdesc_ex, desc);
-	unsigned short queue = skb_get_queue_mapping(skb);
 	unsigned short status;
 	unsigned int estatus = 0;
 	dma_addr_t addr;
@@ -566,7 +562,7 @@ fec_enet_txq_put_data_tso(struct fec_enet_priv_tx_q *txq, struct sk_buff *skb,
 
 	if (fep->bufdesc_ex) {
 		if (fep->quirks & FEC_QUIRK_HAS_AVB)
-			estatus |= FEC_TX_BD_FTYPE(queue);
+			estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 		if (skb->ip_summed == CHECKSUM_PARTIAL)
 			estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
 		ebdp->cbd_bdu = 0;
@@ -595,7 +591,6 @@ fec_enet_txq_put_hdr_tso(struct fec_enet_priv_tx_q *txq,
 	struct fec_enet_private *fep = netdev_priv(ndev);
 	int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
 	struct bufdesc_ex *ebdp = container_of(bdp, struct bufdesc_ex, desc);
-	unsigned short queue = skb_get_queue_mapping(skb);
 	void *bufaddr;
 	unsigned long dmabuf;
 	unsigned short status;
@@ -630,7 +625,7 @@ fec_enet_txq_put_hdr_tso(struct fec_enet_priv_tx_q *txq,
 
 	if (fep->bufdesc_ex) {
 		if (fep->quirks & FEC_QUIRK_HAS_AVB)
-			estatus |= FEC_TX_BD_FTYPE(queue);
+			estatus |= FEC_TX_BD_FTYPE(txq->bd.qid);
 		if (skb->ip_summed == CHECKSUM_PARTIAL)
 			estatus |= BD_ENET_TX_PINS | BD_ENET_TX_IINS;
 		ebdp->cbd_bdu = 0;
@@ -650,7 +645,6 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 	int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
 	int total_len, data_left;
 	struct bufdesc *bdp = txq->bd.cur;
-	unsigned short queue = skb_get_queue_mapping(skb);
 	struct tso_t tso;
 	unsigned int index = 0;
 	int ret;
@@ -715,11 +709,11 @@ static int fec_enet_txq_submit_tso(struct fec_enet_priv_tx_q *txq,
 
 	/* Trigger transmission start */
 	if (!(fep->quirks & FEC_QUIRK_ERR007885) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)) ||
-	    !readl(fep->hwp + FEC_X_DES_ACTIVE(queue)))
-		writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue));
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active) ||
+	    !readl(txq->bd.reg_desc_active))
+		writel(0, txq->bd.reg_desc_active);
 
 	return 0;
 
@@ -819,7 +813,7 @@ static void fec_enet_active_rxring(struct net_device *ndev)
 	int i;
 
 	for (i = 0; i < fep->num_rx_queues; i++)
-		writel(0, fep->hwp + FEC_R_DES_ACTIVE(i));
+		writel(0, fep->rx_queue[i]->bd.reg_desc_active);
 }
 
 static void fec_enet_enable_ring(struct net_device *ndev)
@@ -1255,8 +1249,8 @@ fec_enet_tx_queue(struct net_device *ndev, u16 queue_id)
 
 	/* ERR006538: Keep the transmitter going */
 	if (bdp != txq->bd.cur &&
-	    readl(fep->hwp + FEC_X_DES_ACTIVE(queue_id)) == 0)
-		writel(0, fep->hwp + FEC_X_DES_ACTIVE(queue_id));
+	    readl(txq->bd.reg_desc_active) == 0)
+		writel(0, txq->bd.reg_desc_active);
 }
 
 static void
@@ -1498,7 +1492,7 @@ rx_processing_done:
 		 * incoming frames.  On a heavily loaded network, we should be
 		 * able to keep up at the expense of system resources.
 		 */
-		writel(0, fep->hwp + FEC_R_DES_ACTIVE(queue_id));
+		writel(0, rxq->bd.reg_desc_active);
 	}
 	rxq->bd.cur = bdp;
 	return pkt_received;
@@ -3061,6 +3055,14 @@ static const struct net_device_ops fec_netdev_ops = {
 	.ndo_set_features	= fec_set_features,
 };
 
+static const unsigned short offset_des_active_rxq[] = {
+	FEC_R_DES_ACTIVE_0, FEC_R_DES_ACTIVE_1, FEC_R_DES_ACTIVE_2
+};
+
+static const unsigned short offset_des_active_txq[] = {
+	FEC_X_DES_ACTIVE_0, FEC_X_DES_ACTIVE_1, FEC_X_DES_ACTIVE_2
+};
+
  /*
   * XXX:  We need to clean up on failure exits here.
   *
@@ -3114,6 +3116,7 @@ static int fec_enet_init(struct net_device *ndev)
 		rxq->bd.dma = bd_dma;
 		rxq->bd.dsize = dsize;
 		rxq->bd.dsize_log2 = dsize_log2;
+		rxq->bd.reg_desc_active = fep->hwp + offset_des_active_rxq[i];
 		bd_dma += size;
 		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
 		rxq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
@@ -3129,6 +3132,7 @@ static int fec_enet_init(struct net_device *ndev)
 		txq->bd.dma = bd_dma;
 		txq->bd.dsize = dsize;
 		txq->bd.dsize_log2 = dsize_log2;
+		txq->bd.reg_desc_active = fep->hwp + offset_des_active_txq[i];
 		bd_dma += size;
 		cbd_base = (struct bufdesc *)(((void *)cbd_base) + size);
 		txq->bd.last = (struct bufdesc *)(((void *)cbd_base) - dsize);
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 6/8] net: fec: don't disable FEC_ENET_TS_TIMER interrupt
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

Only the interrupt routine processes this condition.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      | 1 +
 drivers/net/ethernet/freescale/fec_main.c | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index bedd28a..195122e 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -375,6 +375,7 @@ struct bufdesc_ex {
 #define FEC_ENET_TS_TIMER       ((uint)0x00008000)
 
 #define FEC_DEFAULT_IMASK (FEC_ENET_TXF | FEC_ENET_RXF | FEC_ENET_MII | FEC_ENET_TS_TIMER)
+#define FEC_NAPI_IMASK	(FEC_ENET_MII | FEC_ENET_TS_TIMER)
 #define FEC_RX_DISABLED_IMASK (FEC_DEFAULT_IMASK & (~FEC_ENET_RXF))
 
 /* ENET interrupt coalescing macro define */
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 712e3bb..ca2708d 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -1553,7 +1553,7 @@ fec_enet_interrupt(int irq, void *dev_id)
 
 		if (napi_schedule_prep(&fep->napi)) {
 			/* Disable the NAPI interrupts */
-			writel(FEC_ENET_MII, fep->hwp + FEC_IMASK);
+			writel(FEC_NAPI_IMASK, fep->hwp + FEC_IMASK);
 			__napi_schedule(&fep->napi);
 		}
 	}
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 6/8] net: fec: don't disable FEC_ENET_TS_TIMER interrupt
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

Only the interrupt routine processes this condition.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec.h      | 1 +
 drivers/net/ethernet/freescale/fec_main.c | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec.h b/drivers/net/ethernet/freescale/fec.h
index bedd28a..195122e 100644
--- a/drivers/net/ethernet/freescale/fec.h
+++ b/drivers/net/ethernet/freescale/fec.h
@@ -375,6 +375,7 @@ struct bufdesc_ex {
 #define FEC_ENET_TS_TIMER       ((uint)0x00008000)
 
 #define FEC_DEFAULT_IMASK (FEC_ENET_TXF | FEC_ENET_RXF | FEC_ENET_MII | FEC_ENET_TS_TIMER)
+#define FEC_NAPI_IMASK	(FEC_ENET_MII | FEC_ENET_TS_TIMER)
 #define FEC_RX_DISABLED_IMASK (FEC_DEFAULT_IMASK & (~FEC_ENET_RXF))
 
 /* ENET interrupt coalescing macro define */
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 712e3bb..ca2708d 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -1553,7 +1553,7 @@ fec_enet_interrupt(int irq, void *dev_id)
 
 		if (napi_schedule_prep(&fep->napi)) {
 			/* Disable the NAPI interrupts */
-			writel(FEC_ENET_MII, fep->hwp + FEC_IMASK);
+			writel(FEC_NAPI_IMASK, fep->hwp + FEC_IMASK);
 			__napi_schedule(&fep->napi);
 		}
 	}
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

If you don't own it, you shouldn't write to it.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index ca2708d..97ca72a 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 		bdp->cbd_bufaddr = cpu_to_fec32(addr);
 		bdp->cbd_datlen = cpu_to_fec16(frag_len);
+		/* Make sure the updates to rest of the descriptor are
+		 * performed before transferring ownership.
+		 */
+		wmb();
 		bdp->cbd_sc = cpu_to_fec16(status);
 	}
 
@@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 
 	bdp->cbd_datlen = cpu_to_fec16(buflen);
 	bdp->cbd_bufaddr = cpu_to_fec32(addr);
+	/* Make sure the updates to rest of the descriptor are performed before
+	 * transferring ownership.
+	 */
+	wmb();
 
 	/* Send it on its way.  Tell FEC it's ready, interrupt when done,
 	 * it's the last BD of the frame, and to put the CRC on the end.
@@ -1475,7 +1483,6 @@ rx_processing_done:
 
 		/* Mark the buffer empty */
 		status |= BD_ENET_RX_EMPTY;
-		bdp->cbd_sc = cpu_to_fec16(status);
 
 		if (fep->bufdesc_ex) {
 			struct bufdesc_ex *ebdp = (struct bufdesc_ex *)bdp;
@@ -1484,6 +1491,11 @@ rx_processing_done:
 			ebdp->cbd_prot = 0;
 			ebdp->cbd_bdu = 0;
 		}
+		/* Make sure the updates to rest of the descriptor are
+		 * performed before transferring ownership.
+		 */
+		wmb();
+		bdp->cbd_sc = cpu_to_fec16(status);
 
 		/* Update BD pointer to next entry */
 		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

If you don't own it, you shouldn't write to it.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index ca2708d..97ca72a 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 
 		bdp->cbd_bufaddr = cpu_to_fec32(addr);
 		bdp->cbd_datlen = cpu_to_fec16(frag_len);
+		/* Make sure the updates to rest of the descriptor are
+		 * performed before transferring ownership.
+		 */
+		wmb();
 		bdp->cbd_sc = cpu_to_fec16(status);
 	}
 
@@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 
 	bdp->cbd_datlen = cpu_to_fec16(buflen);
 	bdp->cbd_bufaddr = cpu_to_fec32(addr);
+	/* Make sure the updates to rest of the descriptor are performed before
+	 * transferring ownership.
+	 */
+	wmb();
 
 	/* Send it on its way.  Tell FEC it's ready, interrupt when done,
 	 * it's the last BD of the frame, and to put the CRC on the end.
@@ -1475,7 +1483,6 @@ rx_processing_done:
 
 		/* Mark the buffer empty */
 		status |= BD_ENET_RX_EMPTY;
-		bdp->cbd_sc = cpu_to_fec16(status);
 
 		if (fep->bufdesc_ex) {
 			struct bufdesc_ex *ebdp = (struct bufdesc_ex *)bdp;
@@ -1484,6 +1491,11 @@ rx_processing_done:
 			ebdp->cbd_prot = 0;
 			ebdp->cbd_bdu = 0;
 		}
+		/* Make sure the updates to rest of the descriptor are
+		 * performed before transferring ownership.
+		 */
+		wmb();
+		bdp->cbd_sc = cpu_to_fec16(status);
 
 		/* Update BD pointer to next entry */
 		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
  2016-02-05 21:52   ` Troy Kisky
@ 2016-02-06  1:03     ` Joshua Clayton
  -1 siblings, 0 replies; 28+ messages in thread
From: Joshua Clayton @ 2016-02-06  1:03 UTC (permalink / raw)
  To: linux-arm-kernel

On Fri,  5 Feb 2016 14:52:49 -0700
Troy Kisky <troy.kisky@boundarydevices.com> wrote:

> If you don't own it, you shouldn't write to it.
> 
> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
> ---
>  drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>  1 file changed, 13 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/freescale/fec_main.c
> b/drivers/net/ethernet/freescale/fec_main.c index ca2708d..97ca72a
> 100644 --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct
> fec_enet_priv_tx_q *txq, 
>  		bdp->cbd_bufaddr = cpu_to_fec32(addr);
>  		bdp->cbd_datlen = cpu_to_fec16(frag_len);
> +		/* Make sure the updates to rest of the descriptor
> are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
>  		bdp->cbd_sc = cpu_to_fec16(status);
You use almost exactly the same code in each place.
I'd prefer to have wmb(); bdp->cbd_sc = cpu_to_fec16(status) wrapped in
a function or function like macro, and put the comment in one place.

e.g.

/* Make sure the updates to rest of the descriptor
 * performed before transferring ownership.
 */
#define fec_enet_xfr_ownership(status) \
	do { \
		wmb();
		bdp->cbd_sc = cpu_to_fec16(status);
	} while (0)

(I don't know if thats the right macro name, I'm just basing it off
your comment)

>  	}
>  
> @@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct
> fec_enet_priv_tx_q *txq, 
>  	bdp->cbd_datlen = cpu_to_fec16(buflen);
>  	bdp->cbd_bufaddr = cpu_to_fec32(addr);
> +	/* Make sure the updates to rest of the descriptor are
> performed before
> +	 * transferring ownership.
> +	 */
> +	wmb();
>  
>  	/* Send it on its way.  Tell FEC it's ready, interrupt when
> done,
>  	 * it's the last BD of the frame, and to put the CRC on the
> end. @@ -1475,7 +1483,6 @@ rx_processing_done:
>  
>  		/* Mark the buffer empty */
>  		status |= BD_ENET_RX_EMPTY;
> -		bdp->cbd_sc = cpu_to_fec16(status);
>  
>  		if (fep->bufdesc_ex) {
>  			struct bufdesc_ex *ebdp = (struct bufdesc_ex
> *)bdp; @@ -1484,6 +1491,11 @@ rx_processing_done:
>  			ebdp->cbd_prot = 0;
>  			ebdp->cbd_bdu = 0;
>  		}
> +		/* Make sure the updates to rest of the descriptor
> are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
> +		bdp->cbd_sc = cpu_to_fec16(status);
>  
>  		/* Update BD pointer to next entry */
>  		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);

^ permalink raw reply	[flat|nested] 28+ messages in thread

* Re: [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
@ 2016-02-06  1:03     ` Joshua Clayton
  0 siblings, 0 replies; 28+ messages in thread
From: Joshua Clayton @ 2016-02-06  1:03 UTC (permalink / raw)
  To: Troy Kisky
  Cc: netdev, davem, B38611, fabio.estevam, andrew, linux, arnd, laci,
	johannes, l.stach, shawnguo, linux-arm-kernel, tremyfr

On Fri,  5 Feb 2016 14:52:49 -0700
Troy Kisky <troy.kisky@boundarydevices.com> wrote:

> If you don't own it, you shouldn't write to it.
> 
> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
> ---
>  drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>  1 file changed, 13 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/freescale/fec_main.c
> b/drivers/net/ethernet/freescale/fec_main.c index ca2708d..97ca72a
> 100644 --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct
> fec_enet_priv_tx_q *txq, 
>  		bdp->cbd_bufaddr = cpu_to_fec32(addr);
>  		bdp->cbd_datlen = cpu_to_fec16(frag_len);
> +		/* Make sure the updates to rest of the descriptor
> are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
>  		bdp->cbd_sc = cpu_to_fec16(status);
You use almost exactly the same code in each place.
I'd prefer to have wmb(); bdp->cbd_sc = cpu_to_fec16(status) wrapped in
a function or function like macro, and put the comment in one place.

e.g.

/* Make sure the updates to rest of the descriptor
 * performed before transferring ownership.
 */
#define fec_enet_xfr_ownership(status) \
	do { \
		wmb();
		bdp->cbd_sc = cpu_to_fec16(status);
	} while (0)

(I don't know if thats the right macro name, I'm just basing it off
your comment)

>  	}
>  
> @@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct
> fec_enet_priv_tx_q *txq, 
>  	bdp->cbd_datlen = cpu_to_fec16(buflen);
>  	bdp->cbd_bufaddr = cpu_to_fec32(addr);
> +	/* Make sure the updates to rest of the descriptor are
> performed before
> +	 * transferring ownership.
> +	 */
> +	wmb();
>  
>  	/* Send it on its way.  Tell FEC it's ready, interrupt when
> done,
>  	 * it's the last BD of the frame, and to put the CRC on the
> end. @@ -1475,7 +1483,6 @@ rx_processing_done:
>  
>  		/* Mark the buffer empty */
>  		status |= BD_ENET_RX_EMPTY;
> -		bdp->cbd_sc = cpu_to_fec16(status);
>  
>  		if (fep->bufdesc_ex) {
>  			struct bufdesc_ex *ebdp = (struct bufdesc_ex
> *)bdp; @@ -1484,6 +1491,11 @@ rx_processing_done:
>  			ebdp->cbd_prot = 0;
>  			ebdp->cbd_bdu = 0;
>  		}
> +		/* Make sure the updates to rest of the descriptor
> are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
> +		bdp->cbd_sc = cpu_to_fec16(status);
>  
>  		/* Update BD pointer to next entry */
>  		bdp = fec_enet_get_nextdesc(bdp, &rxq->bd);

^ permalink raw reply	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
  2016-02-06  1:03     ` Joshua Clayton
@ 2016-02-24 20:38       ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-24 20:38 UTC (permalink / raw)
  To: linux-arm-kernel

On 2/5/2016 6:03 PM, Joshua Clayton wrote:
> On Fri,  5 Feb 2016 14:52:49 -0700
> Troy Kisky <troy.kisky@boundarydevices.com> wrote:
> 
>> If you don't own it, you shouldn't write to it.
>>
>> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
>> ---
>>  drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>>  1 file changed, 13 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ethernet/freescale/fec_main.c
>> b/drivers/net/ethernet/freescale/fec_main.c index ca2708d..97ca72a
>> 100644 --- a/drivers/net/ethernet/freescale/fec_main.c
>> +++ b/drivers/net/ethernet/freescale/fec_main.c
>> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct
>> fec_enet_priv_tx_q *txq, 
>>  		bdp->cbd_bufaddr = cpu_to_fec32(addr);
>>  		bdp->cbd_datlen = cpu_to_fec16(frag_len);
>> +		/* Make sure the updates to rest of the descriptor
>> are
>> +		 * performed before transferring ownership.
>> +		 */
>> +		wmb();
>>  		bdp->cbd_sc = cpu_to_fec16(status);
> You use almost exactly the same code in each place.
> I'd prefer to have wmb(); bdp->cbd_sc = cpu_to_fec16(status) wrapped in
> a function or function like macro, and put the comment in one place.
> 

Thanks for the review. I added a patch to the end of the series to create a common
"trigger_tx" subroutine which should address this.

^ permalink raw reply	[flat|nested] 28+ messages in thread

* Re: [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
@ 2016-02-24 20:38       ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-24 20:38 UTC (permalink / raw)
  To: Joshua Clayton
  Cc: netdev, davem, B38611, fabio.estevam, andrew, linux, arnd, laci,
	johannes, l.stach, shawnguo, linux-arm-kernel, tremyfr

On 2/5/2016 6:03 PM, Joshua Clayton wrote:
> On Fri,  5 Feb 2016 14:52:49 -0700
> Troy Kisky <troy.kisky@boundarydevices.com> wrote:
> 
>> If you don't own it, you shouldn't write to it.
>>
>> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
>> ---
>>  drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>>  1 file changed, 13 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ethernet/freescale/fec_main.c
>> b/drivers/net/ethernet/freescale/fec_main.c index ca2708d..97ca72a
>> 100644 --- a/drivers/net/ethernet/freescale/fec_main.c
>> +++ b/drivers/net/ethernet/freescale/fec_main.c
>> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct
>> fec_enet_priv_tx_q *txq, 
>>  		bdp->cbd_bufaddr = cpu_to_fec32(addr);
>>  		bdp->cbd_datlen = cpu_to_fec16(frag_len);
>> +		/* Make sure the updates to rest of the descriptor
>> are
>> +		 * performed before transferring ownership.
>> +		 */
>> +		wmb();
>>  		bdp->cbd_sc = cpu_to_fec16(status);
> You use almost exactly the same code in each place.
> I'd prefer to have wmb(); bdp->cbd_sc = cpu_to_fec16(status) wrapped in
> a function or function like macro, and put the comment in one place.
> 

Thanks for the review. I added a patch to the end of the series to create a common
"trigger_tx" subroutine which should address this.

^ permalink raw reply	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
  2016-02-05 21:52   ` Troy Kisky
@ 2016-02-06 11:52     ` Sergei Shtylyov
  -1 siblings, 0 replies; 28+ messages in thread
From: Sergei Shtylyov @ 2016-02-06 11:52 UTC (permalink / raw)
  To: linux-arm-kernel

Hello.

On 2/6/2016 12:52 AM, Troy Kisky wrote:

> If you don't own it, you shouldn't write to it.
>
> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
> ---
>   drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>   1 file changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
> index ca2708d..97ca72a 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
>
>   		bdp->cbd_bufaddr = cpu_to_fec32(addr);
>   		bdp->cbd_datlen = cpu_to_fec16(frag_len);
> +		/* Make sure the updates to rest of the descriptor are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
>   		bdp->cbd_sc = cpu_to_fec16(status);
>   	}
>
> @@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
>
>   	bdp->cbd_datlen = cpu_to_fec16(buflen);
>   	bdp->cbd_bufaddr = cpu_to_fec32(addr);
> +	/* Make sure the updates to rest of the descriptor are performed before
> +	 * transferring ownership.
> +	 */
> +	wmb();
>
>   	/* Send it on its way.  Tell FEC it's ready, interrupt when done,
>   	 * it's the last BD of the frame, and to put the CRC on the end.
[...]
> @@ -1484,6 +1491,11 @@ rx_processing_done:
>   			ebdp->cbd_prot = 0;
>   			ebdp->cbd_bdu = 0;
>   		}
> +		/* Make sure the updates to rest of the descriptor are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
> +		bdp->cbd_sc = cpu_to_fec16(status);

    I think you can use "ligter" dma_wmb() in this case.

[...]

MBR, Sergei

^ permalink raw reply	[flat|nested] 28+ messages in thread

* Re: [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
@ 2016-02-06 11:52     ` Sergei Shtylyov
  0 siblings, 0 replies; 28+ messages in thread
From: Sergei Shtylyov @ 2016-02-06 11:52 UTC (permalink / raw)
  To: Troy Kisky, netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd

Hello.

On 2/6/2016 12:52 AM, Troy Kisky wrote:

> If you don't own it, you shouldn't write to it.
>
> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
> ---
>   drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>   1 file changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
> index ca2708d..97ca72a 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
>
>   		bdp->cbd_bufaddr = cpu_to_fec32(addr);
>   		bdp->cbd_datlen = cpu_to_fec16(frag_len);
> +		/* Make sure the updates to rest of the descriptor are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
>   		bdp->cbd_sc = cpu_to_fec16(status);
>   	}
>
> @@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
>
>   	bdp->cbd_datlen = cpu_to_fec16(buflen);
>   	bdp->cbd_bufaddr = cpu_to_fec32(addr);
> +	/* Make sure the updates to rest of the descriptor are performed before
> +	 * transferring ownership.
> +	 */
> +	wmb();
>
>   	/* Send it on its way.  Tell FEC it's ready, interrupt when done,
>   	 * it's the last BD of the frame, and to put the CRC on the end.
[...]
> @@ -1484,6 +1491,11 @@ rx_processing_done:
>   			ebdp->cbd_prot = 0;
>   			ebdp->cbd_bdu = 0;
>   		}
> +		/* Make sure the updates to rest of the descriptor are
> +		 * performed before transferring ownership.
> +		 */
> +		wmb();
> +		bdp->cbd_sc = cpu_to_fec16(status);

    I think you can use "ligter" dma_wmb() in this case.

[...]

MBR, Sergei

^ permalink raw reply	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
  2016-02-06 11:52     ` Sergei Shtylyov
@ 2016-02-24 20:39       ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-24 20:39 UTC (permalink / raw)
  To: linux-arm-kernel

On 2/6/2016 4:52 AM, Sergei Shtylyov wrote:
> Hello.
> 
> On 2/6/2016 12:52 AM, Troy Kisky wrote:
> 
>> If you don't own it, you shouldn't write to it.
>>
>> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
>> ---
>>   drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>>   1 file changed, 13 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
>> index ca2708d..97ca72a 100644
>> --- a/drivers/net/ethernet/freescale/fec_main.c
>> +++ b/drivers/net/ethernet/freescale/fec_main.c
>> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
>>
>>           bdp->cbd_bufaddr = cpu_to_fec32(addr);
>>           bdp->cbd_datlen = cpu_to_fec16(frag_len);
>> +        /* Make sure the updates to rest of the descriptor are
>> +         * performed before transferring ownership.
>> +         */
>> +        wmb();
>>           bdp->cbd_sc = cpu_to_fec16(status);
>>       }
>>
>> @@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
>>
>>       bdp->cbd_datlen = cpu_to_fec16(buflen);
>>       bdp->cbd_bufaddr = cpu_to_fec32(addr);
>> +    /* Make sure the updates to rest of the descriptor are performed before
>> +     * transferring ownership.
>> +     */
>> +    wmb();
>>
>>       /* Send it on its way.  Tell FEC it's ready, interrupt when done,
>>        * it's the last BD of the frame, and to put the CRC on the end.
> [...]
>> @@ -1484,6 +1491,11 @@ rx_processing_done:
>>               ebdp->cbd_prot = 0;
>>               ebdp->cbd_bdu = 0;
>>           }
>> +        /* Make sure the updates to rest of the descriptor are
>> +         * performed before transferring ownership.
>> +         */
>> +        wmb();
>> +        bdp->cbd_sc = cpu_to_fec16(status);
> 
>    I think you can use "ligter" dma_wmb() in this case.
> 


Thanks for the review, I added a patch to the end of the series to address this.

Troy

^ permalink raw reply	[flat|nested] 28+ messages in thread

* Re: [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete
@ 2016-02-24 20:39       ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-24 20:39 UTC (permalink / raw)
  To: Sergei Shtylyov, netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd

On 2/6/2016 4:52 AM, Sergei Shtylyov wrote:
> Hello.
> 
> On 2/6/2016 12:52 AM, Troy Kisky wrote:
> 
>> If you don't own it, you shouldn't write to it.
>>
>> Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
>> ---
>>   drivers/net/ethernet/freescale/fec_main.c | 14 +++++++++++++-
>>   1 file changed, 13 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
>> index ca2708d..97ca72a 100644
>> --- a/drivers/net/ethernet/freescale/fec_main.c
>> +++ b/drivers/net/ethernet/freescale/fec_main.c
>> @@ -390,6 +390,10 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
>>
>>           bdp->cbd_bufaddr = cpu_to_fec32(addr);
>>           bdp->cbd_datlen = cpu_to_fec16(frag_len);
>> +        /* Make sure the updates to rest of the descriptor are
>> +         * performed before transferring ownership.
>> +         */
>> +        wmb();
>>           bdp->cbd_sc = cpu_to_fec16(status);
>>       }
>>
>> @@ -499,6 +503,10 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
>>
>>       bdp->cbd_datlen = cpu_to_fec16(buflen);
>>       bdp->cbd_bufaddr = cpu_to_fec32(addr);
>> +    /* Make sure the updates to rest of the descriptor are performed before
>> +     * transferring ownership.
>> +     */
>> +    wmb();
>>
>>       /* Send it on its way.  Tell FEC it's ready, interrupt when done,
>>        * it's the last BD of the frame, and to put the CRC on the end.
> [...]
>> @@ -1484,6 +1491,11 @@ rx_processing_done:
>>               ebdp->cbd_prot = 0;
>>               ebdp->cbd_bdu = 0;
>>           }
>> +        /* Make sure the updates to rest of the descriptor are
>> +         * performed before transferring ownership.
>> +         */
>> +        wmb();
>> +        bdp->cbd_sc = cpu_to_fec16(status);
> 
>    I think you can use "ligter" dma_wmb() in this case.
> 


Thanks for the review, I added a patch to the end of the series to address this.

Troy

^ permalink raw reply	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 8/8] net: fec: improve error handling
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-05 21:52   ` Troy Kisky
  -1 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: linux-arm-kernel

Unmap initial buffer on error.
Don't free skb until it has been unmapped.
Move cbd_bufaddr assignment closer to the mapping function.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 97ca72a..ef18ca5 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -382,7 +382,6 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 		addr = dma_map_single(&fep->pdev->dev, bufaddr, frag_len,
 				      DMA_TO_DEVICE);
 		if (dma_mapping_error(&fep->pdev->dev, addr)) {
-			dev_kfree_skb_any(skb);
 			if (net_ratelimit())
 				netdev_err(ndev, "Tx DMA memory map failed\n");
 			goto dma_mapping_error;
@@ -467,8 +466,12 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 
 	if (nr_frags) {
 		last_bdp = fec_enet_txq_submit_frag_skb(txq, skb, ndev);
-		if (IS_ERR(last_bdp))
+		if (IS_ERR(last_bdp)) {
+			dma_unmap_single(&fep->pdev->dev, addr,
+					 buflen, DMA_TO_DEVICE);
+			dev_kfree_skb_any(skb);
 			return NETDEV_TX_OK;
+		}
 	} else {
 		status |= (BD_ENET_TX_INTR | BD_ENET_TX_LAST);
 		if (fep->bufdesc_ex) {
@@ -478,6 +481,8 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 				estatus |= BD_ENET_TX_TS;
 		}
 	}
+	bdp->cbd_bufaddr = cpu_to_fec32(addr);
+	bdp->cbd_datlen = cpu_to_fec16(buflen);
 
 	if (fep->bufdesc_ex) {
 
@@ -501,8 +506,6 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	/* Save skb pointer */
 	txq->tx_skbuff[index] = skb;
 
-	bdp->cbd_datlen = cpu_to_fec16(buflen);
-	bdp->cbd_bufaddr = cpu_to_fec32(addr);
 	/* Make sure the updates to rest of the descriptor are performed before
 	 * transferring ownership.
 	 */
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 8/8] net: fec: improve error handling
@ 2016-02-05 21:52   ` Troy Kisky
  0 siblings, 0 replies; 28+ messages in thread
From: Troy Kisky @ 2016-02-05 21:52 UTC (permalink / raw)
  To: netdev, davem, B38611
  Cc: fabio.estevam, l.stach, andrew, tremyfr, linux, linux-arm-kernel,
	laci, shawnguo, johannes, arnd, Troy Kisky

Unmap initial buffer on error.
Don't free skb until it has been unmapped.
Move cbd_bufaddr assignment closer to the mapping function.

Signed-off-by: Troy Kisky <troy.kisky@boundarydevices.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index 97ca72a..ef18ca5 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -382,7 +382,6 @@ fec_enet_txq_submit_frag_skb(struct fec_enet_priv_tx_q *txq,
 		addr = dma_map_single(&fep->pdev->dev, bufaddr, frag_len,
 				      DMA_TO_DEVICE);
 		if (dma_mapping_error(&fep->pdev->dev, addr)) {
-			dev_kfree_skb_any(skb);
 			if (net_ratelimit())
 				netdev_err(ndev, "Tx DMA memory map failed\n");
 			goto dma_mapping_error;
@@ -467,8 +466,12 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 
 	if (nr_frags) {
 		last_bdp = fec_enet_txq_submit_frag_skb(txq, skb, ndev);
-		if (IS_ERR(last_bdp))
+		if (IS_ERR(last_bdp)) {
+			dma_unmap_single(&fep->pdev->dev, addr,
+					 buflen, DMA_TO_DEVICE);
+			dev_kfree_skb_any(skb);
 			return NETDEV_TX_OK;
+		}
 	} else {
 		status |= (BD_ENET_TX_INTR | BD_ENET_TX_LAST);
 		if (fep->bufdesc_ex) {
@@ -478,6 +481,8 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 				estatus |= BD_ENET_TX_TS;
 		}
 	}
+	bdp->cbd_bufaddr = cpu_to_fec32(addr);
+	bdp->cbd_datlen = cpu_to_fec16(buflen);
 
 	if (fep->bufdesc_ex) {
 
@@ -501,8 +506,6 @@ static int fec_enet_txq_submit_skb(struct fec_enet_priv_tx_q *txq,
 	/* Save skb pointer */
 	txq->tx_skbuff[index] = skb;
 
-	bdp->cbd_datlen = cpu_to_fec16(buflen);
-	bdp->cbd_bufaddr = cpu_to_fec32(addr);
 	/* Make sure the updates to rest of the descriptor are performed before
 	 * transferring ownership.
 	 */
-- 
2.5.0

^ permalink raw reply related	[flat|nested] 28+ messages in thread

* [PATCH net-next V2 0/8] net: fec: cleanup/fixes
  2016-02-05 21:52 ` Troy Kisky
@ 2016-02-11 11:15   ` David Miller
  -1 siblings, 0 replies; 28+ messages in thread
From: David Miller @ 2016-02-11 11:15 UTC (permalink / raw)
  To: linux-arm-kernel

From: Troy Kisky <troy.kisky@boundarydevices.com>
Date: Fri,  5 Feb 2016 14:52:42 -0700

> 
> V2 is a rebase on top of johannes endian-safe patch and
> is only the 1st eight patches.
> The testing for this series was done on a nitrogen6x.
> The base commit was
> commit b45efa30a626e915192a6c548cd8642379cd47cc
>     Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
> 
> Testing showed no change in performance.
> Testing used imx_v6_v7_defconfig + CONFIG_MICREL_PHY.
> The processor was running at 996Mhz.
> The following commands were used to get the transfer rates.
> 
> On an x86 ubunto system,
> iperf -s -i.5 -u
> 
> 
> On a nitrogen6x board, running via SD Card.
> I first stopped some background processes
> 
> stop cron
> stop upstart-file-bridge
> stop upstart-socket-bridge
> stop upstart-udev-bridge
> stop rsyslog
> stop dbus
> killall dhclient
> echo performance >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
> 
> taskset 0x2 iperf -c 192.168.0.201 -u -t60 -b500M -r
> 
> There is a branch available on github with this series, and the rest of
> my fec patches, for those who would like to test it.
> https://github.com:boundarydevices/linux-imx6.git branch net-next_master

Series applied, thanks.

^ permalink raw reply	[flat|nested] 28+ messages in thread

* Re: [PATCH net-next V2 0/8] net: fec: cleanup/fixes
@ 2016-02-11 11:15   ` David Miller
  0 siblings, 0 replies; 28+ messages in thread
From: David Miller @ 2016-02-11 11:15 UTC (permalink / raw)
  To: troy.kisky
  Cc: netdev, B38611, fabio.estevam, l.stach, andrew, tremyfr, linux,
	linux-arm-kernel, laci, shawnguo, johannes, arnd

From: Troy Kisky <troy.kisky@boundarydevices.com>
Date: Fri,  5 Feb 2016 14:52:42 -0700

> 
> V2 is a rebase on top of johannes endian-safe patch and
> is only the 1st eight patches.
> The testing for this series was done on a nitrogen6x.
> The base commit was
> commit b45efa30a626e915192a6c548cd8642379cd47cc
>     Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
> 
> Testing showed no change in performance.
> Testing used imx_v6_v7_defconfig + CONFIG_MICREL_PHY.
> The processor was running at 996Mhz.
> The following commands were used to get the transfer rates.
> 
> On an x86 ubunto system,
> iperf -s -i.5 -u
> 
> 
> On a nitrogen6x board, running via SD Card.
> I first stopped some background processes
> 
> stop cron
> stop upstart-file-bridge
> stop upstart-socket-bridge
> stop upstart-udev-bridge
> stop rsyslog
> stop dbus
> killall dhclient
> echo performance >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
> 
> taskset 0x2 iperf -c 192.168.0.201 -u -t60 -b500M -r
> 
> There is a branch available on github with this series, and the rest of
> my fec patches, for those who would like to test it.
> https://github.com:boundarydevices/linux-imx6.git branch net-next_master

Series applied, thanks.

^ permalink raw reply	[flat|nested] 28+ messages in thread

end of thread, other threads:[~2016-02-24 20:39 UTC | newest]

Thread overview: 28+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-02-05 21:52 [PATCH net-next V2 0/8] net: fec: cleanup/fixes Troy Kisky
2016-02-05 21:52 ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 1/8] net: fec: stop the "rcv is not +last, " error messages Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 2/8] net: fec: fix rx error counts Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 3/8] net: fec: fix fec_enet_get_free_txdesc_num Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 4/8] net: fec: add struct bufdesc_prop Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 5/8] net: fec: add variable reg_desc_active to speed things up Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 6/8] net: fec: don't disable FEC_ENET_TS_TIMER interrupt Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 7/8] net: fec: don't transfer ownership until descriptor write is complete Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-06  1:03   ` Joshua Clayton
2016-02-06  1:03     ` Joshua Clayton
2016-02-24 20:38     ` Troy Kisky
2016-02-24 20:38       ` Troy Kisky
2016-02-06 11:52   ` Sergei Shtylyov
2016-02-06 11:52     ` Sergei Shtylyov
2016-02-24 20:39     ` Troy Kisky
2016-02-24 20:39       ` Troy Kisky
2016-02-05 21:52 ` [PATCH net-next V2 8/8] net: fec: improve error handling Troy Kisky
2016-02-05 21:52   ` Troy Kisky
2016-02-11 11:15 ` [PATCH net-next V2 0/8] net: fec: cleanup/fixes David Miller
2016-02-11 11:15   ` David Miller

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.