[PATCH net-next 0/5] net: lan966x: Add support for FDMA

[PATCH net-next 0/5] net: lan966x: Add support for FDMA

[Horatiu Vultur]

Currently when injecting or extracting a frame from CPU, the frame
is given to the HW each word at a time. There is another way to
inject/extract frames from CPU using FDMA(Frame Direct Memory Access).
In this way the entire frame is given to the HW. This improves both
RX and TX bitrate.

Horatiu Vultur (5):
  dt-bindings: net: lan966x: Extend with FDMA interrupt
  net: lan966x: Add registers that are used for FDMA.
  net: lan966x: Expose functions that are needed by FDMA
  net: lan966x: Add FDMA functionality
  net: lan96x: Update FDMA to change MTU.

 .../net/microchip,lan966x-switch.yaml         |   2 +
 .../net/ethernet/microchip/lan966x/Makefile   |   2 +-
 .../ethernet/microchip/lan966x/lan966x_fdma.c | 772 ++++++++++++++++++
 .../ethernet/microchip/lan966x/lan966x_main.c |  44 +-
 .../ethernet/microchip/lan966x/lan966x_main.h | 120 +++
 .../ethernet/microchip/lan966x/lan966x_port.c |   3 +
 .../ethernet/microchip/lan966x/lan966x_regs.h | 106 +++
 7 files changed, 1037 insertions(+), 12 deletions(-)
 create mode 100644 drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c

-- 
2.33.0

[PATCH net-next 1/5] dt-bindings: net: lan966x: Extend with FDMA interrupt

[Horatiu Vultur]

Extend dt-bindings for lan966x with FDMA interrupt. This is generated
when receiving a frame or when a frame was transmitted. The interrupt
needs to be enable for each frame.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
---
 .../devicetree/bindings/net/microchip,lan966x-switch.yaml       | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml b/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
index 13812768b923..14e0bae5965f 100644
--- a/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
+++ b/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
@@ -39,6 +39,7 @@ properties:
       - description: frame dma based extraction
       - description: analyzer interrupt
       - description: ptp interrupt
+      - description: fdma interrupt
 
   interrupt-names:
     minItems: 1
@@ -47,6 +48,7 @@ properties:
       - const: fdma
       - const: ana
       - const: ptp
+      - const: fdma
 
   resets:
     items:
-- 
2.33.0

[PATCH net-next 2/5] net: lan966x: Add registers that are used for FDMA.

[Horatiu Vultur]

Add the registers that are used to configure the FDMA.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
---
 .../ethernet/microchip/lan966x/lan966x_main.c |   1 +
 .../ethernet/microchip/lan966x/lan966x_regs.h | 106 ++++++++++++++++++
 2 files changed, 107 insertions(+)

diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
index e1bcb28039dc..4240db708886 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
@@ -43,6 +43,7 @@ struct lan966x_main_io_resource {
 
 static const struct lan966x_main_io_resource lan966x_main_iomap[] =  {
 	{ TARGET_CPU,                   0xc0000, 0 }, /* 0xe00c0000 */
+	{ TARGET_FDMA,                  0xc0400, 0 }, /* 0xe00c0400 */
 	{ TARGET_ORG,                         0, 1 }, /* 0xe2000000 */
 	{ TARGET_GCB,                    0x4000, 1 }, /* 0xe2004000 */
 	{ TARGET_QS,                     0x8000, 1 }, /* 0xe2008000 */
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_regs.h b/drivers/net/ethernet/microchip/lan966x/lan966x_regs.h
index 0c0b3e173d53..2f59285bef29 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_regs.h
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_regs.h
@@ -17,6 +17,7 @@ enum lan966x_target {
 	TARGET_CHIP_TOP = 5,
 	TARGET_CPU = 6,
 	TARGET_DEV = 13,
+	TARGET_FDMA = 21,
 	TARGET_GCB = 27,
 	TARGET_ORG = 36,
 	TARGET_PTP = 41,
@@ -578,6 +579,111 @@ enum lan966x_target {
 #define DEV_PCS1G_STICKY_LINK_DOWN_STICKY_GET(x)\
 	FIELD_GET(DEV_PCS1G_STICKY_LINK_DOWN_STICKY, x)
 
+/*      FDMA:FDMA:FDMA_CH_ACTIVATE */
+#define FDMA_CH_ACTIVATE          __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 0, 0, 1, 4)
+
+#define FDMA_CH_ACTIVATE_CH_ACTIVATE             GENMASK(7, 0)
+#define FDMA_CH_ACTIVATE_CH_ACTIVATE_SET(x)\
+	FIELD_PREP(FDMA_CH_ACTIVATE_CH_ACTIVATE, x)
+#define FDMA_CH_ACTIVATE_CH_ACTIVATE_GET(x)\
+	FIELD_GET(FDMA_CH_ACTIVATE_CH_ACTIVATE, x)
+
+/*      FDMA:FDMA:FDMA_CH_RELOAD */
+#define FDMA_CH_RELOAD            __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 4, 0, 1, 4)
+
+#define FDMA_CH_RELOAD_CH_RELOAD                 GENMASK(7, 0)
+#define FDMA_CH_RELOAD_CH_RELOAD_SET(x)\
+	FIELD_PREP(FDMA_CH_RELOAD_CH_RELOAD, x)
+#define FDMA_CH_RELOAD_CH_RELOAD_GET(x)\
+	FIELD_GET(FDMA_CH_RELOAD_CH_RELOAD, x)
+
+/*      FDMA:FDMA:FDMA_CH_DISABLE */
+#define FDMA_CH_DISABLE           __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 8, 0, 1, 4)
+
+#define FDMA_CH_DISABLE_CH_DISABLE               GENMASK(7, 0)
+#define FDMA_CH_DISABLE_CH_DISABLE_SET(x)\
+	FIELD_PREP(FDMA_CH_DISABLE_CH_DISABLE, x)
+#define FDMA_CH_DISABLE_CH_DISABLE_GET(x)\
+	FIELD_GET(FDMA_CH_DISABLE_CH_DISABLE, x)
+
+/*      FDMA:FDMA:FDMA_CH_DB_DISCARD */
+#define FDMA_CH_DB_DISCARD        __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 16, 0, 1, 4)
+
+#define FDMA_CH_DB_DISCARD_DB_DISCARD            GENMASK(7, 0)
+#define FDMA_CH_DB_DISCARD_DB_DISCARD_SET(x)\
+	FIELD_PREP(FDMA_CH_DB_DISCARD_DB_DISCARD, x)
+#define FDMA_CH_DB_DISCARD_DB_DISCARD_GET(x)\
+	FIELD_GET(FDMA_CH_DB_DISCARD_DB_DISCARD, x)
+
+/*      FDMA:FDMA:FDMA_DCB_LLP */
+#define FDMA_DCB_LLP(r)           __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 52, r, 8, 4)
+
+/*      FDMA:FDMA:FDMA_DCB_LLP1 */
+#define FDMA_DCB_LLP1(r)          __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 84, r, 8, 4)
+
+/*      FDMA:FDMA:FDMA_CH_ACTIVE */
+#define FDMA_CH_ACTIVE            __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 180, 0, 1, 4)
+
+/*      FDMA:FDMA:FDMA_CH_CFG */
+#define FDMA_CH_CFG(r)            __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 224, r, 8, 4)
+
+#define FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY          BIT(4)
+#define FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY_SET(x)\
+	FIELD_PREP(FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY, x)
+#define FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY_GET(x)\
+	FIELD_GET(FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY, x)
+
+#define FDMA_CH_CFG_CH_INJ_PORT                  BIT(3)
+#define FDMA_CH_CFG_CH_INJ_PORT_SET(x)\
+	FIELD_PREP(FDMA_CH_CFG_CH_INJ_PORT, x)
+#define FDMA_CH_CFG_CH_INJ_PORT_GET(x)\
+	FIELD_GET(FDMA_CH_CFG_CH_INJ_PORT, x)
+
+#define FDMA_CH_CFG_CH_DCB_DB_CNT                GENMASK(2, 1)
+#define FDMA_CH_CFG_CH_DCB_DB_CNT_SET(x)\
+	FIELD_PREP(FDMA_CH_CFG_CH_DCB_DB_CNT, x)
+#define FDMA_CH_CFG_CH_DCB_DB_CNT_GET(x)\
+	FIELD_GET(FDMA_CH_CFG_CH_DCB_DB_CNT, x)
+
+#define FDMA_CH_CFG_CH_MEM                       BIT(0)
+#define FDMA_CH_CFG_CH_MEM_SET(x)\
+	FIELD_PREP(FDMA_CH_CFG_CH_MEM, x)
+#define FDMA_CH_CFG_CH_MEM_GET(x)\
+	FIELD_GET(FDMA_CH_CFG_CH_MEM, x)
+
+/*      FDMA:FDMA:FDMA_PORT_CTRL */
+#define FDMA_PORT_CTRL(r)         __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 376, r, 2, 4)
+
+#define FDMA_PORT_CTRL_INJ_STOP                  BIT(4)
+#define FDMA_PORT_CTRL_INJ_STOP_SET(x)\
+	FIELD_PREP(FDMA_PORT_CTRL_INJ_STOP, x)
+#define FDMA_PORT_CTRL_INJ_STOP_GET(x)\
+	FIELD_GET(FDMA_PORT_CTRL_INJ_STOP, x)
+
+#define FDMA_PORT_CTRL_XTR_STOP                  BIT(2)
+#define FDMA_PORT_CTRL_XTR_STOP_SET(x)\
+	FIELD_PREP(FDMA_PORT_CTRL_XTR_STOP, x)
+#define FDMA_PORT_CTRL_XTR_STOP_GET(x)\
+	FIELD_GET(FDMA_PORT_CTRL_XTR_STOP, x)
+
+/*      FDMA:FDMA:FDMA_INTR_DB */
+#define FDMA_INTR_DB              __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 392, 0, 1, 4)
+
+/*      FDMA:FDMA:FDMA_INTR_DB_ENA */
+#define FDMA_INTR_DB_ENA          __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 396, 0, 1, 4)
+
+#define FDMA_INTR_DB_ENA_INTR_DB_ENA             GENMASK(7, 0)
+#define FDMA_INTR_DB_ENA_INTR_DB_ENA_SET(x)\
+	FIELD_PREP(FDMA_INTR_DB_ENA_INTR_DB_ENA, x)
+#define FDMA_INTR_DB_ENA_INTR_DB_ENA_GET(x)\
+	FIELD_GET(FDMA_INTR_DB_ENA_INTR_DB_ENA, x)
+
+/*      FDMA:FDMA:FDMA_INTR_ERR */
+#define FDMA_INTR_ERR             __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 400, 0, 1, 4)
+
+/*      FDMA:FDMA:FDMA_ERRORS */
+#define FDMA_ERRORS               __REG(TARGET_FDMA, 0, 1, 8, 0, 1, 428, 412, 0, 1, 4)
+
 /*      PTP:PTP_CFG:PTP_DOM_CFG */
 #define PTP_DOM_CFG               __REG(TARGET_PTP, 0, 1, 512, 0, 1, 16, 12, 0, 1, 4)
 
-- 
2.33.0

[PATCH net-next 3/5] net: lan966x: Expose functions that are needed by FDMA

[Horatiu Vultur]

Expose the following functions 'lan966x_hw_offload',
'lan966x_ifh_get_src_port' and 'lan966x_ifh_get_timestamp' in
lan966x_main.h so they can be accessed by FDMA.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
---
 drivers/net/ethernet/microchip/lan966x/lan966x_main.c | 10 +++-------
 drivers/net/ethernet/microchip/lan966x/lan966x_main.h |  8 ++++++++
 2 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
index 4240db708886..2c82f847ae6d 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
@@ -24,9 +24,6 @@
 #define XTR_NOT_READY			0x07000080U
 #define XTR_VALID_BYTES(x)		(4 - (((x) >> 24) & 3))
 
-#define READL_SLEEP_US			10
-#define READL_TIMEOUT_US		100000000
-
 #define IO_RANGES 2
 
 static const struct of_device_id lan966x_match[] = {
@@ -430,8 +427,7 @@ bool lan966x_netdevice_check(const struct net_device *dev)
 	return dev->netdev_ops == &lan966x_port_netdev_ops;
 }
 
-static bool lan966x_hw_offload(struct lan966x *lan966x, u32 port,
-			       struct sk_buff *skb)
+bool lan966x_hw_offload(struct lan966x *lan966x, u32 port, struct sk_buff *skb)
 {
 	u32 val;
 
@@ -512,7 +508,7 @@ static int lan966x_rx_frame_word(struct lan966x *lan966x, u8 grp, u32 *rval)
 	}
 }
 
-static void lan966x_ifh_get_src_port(void *ifh, u64 *src_port)
+void lan966x_ifh_get_src_port(void *ifh, u64 *src_port)
 {
 	packing(ifh, src_port, IFH_POS_SRCPORT + IFH_WID_SRCPORT - 1,
 		IFH_POS_SRCPORT, IFH_LEN * 4, UNPACK, 0);
@@ -524,7 +520,7 @@ static void lan966x_ifh_get_len(void *ifh, u64 *len)
 		IFH_POS_LEN, IFH_LEN * 4, UNPACK, 0);
 }
 
-static void lan966x_ifh_get_timestamp(void *ifh, u64 *timestamp)
+void lan966x_ifh_get_timestamp(void *ifh, u64 *timestamp)
 {
 	packing(ifh, timestamp, IFH_POS_TIMESTAMP + IFH_WID_TIMESTAMP - 1,
 		IFH_POS_TIMESTAMP, IFH_LEN * 4, UNPACK, 0);
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_main.h b/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
index ae282da1da74..b692c612f235 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
@@ -17,6 +17,9 @@
 #define TABLE_UPDATE_SLEEP_US		10
 #define TABLE_UPDATE_TIMEOUT_US		100000
 
+#define READL_SLEEP_US			10
+#define READL_TIMEOUT_US		100000000
+
 #define LAN966X_BUFFER_CELL_SZ		64
 #define LAN966X_BUFFER_MEMORY		(160 * 1024)
 #define LAN966X_BUFFER_MIN_SZ		60
@@ -195,6 +198,11 @@ bool lan966x_netdevice_check(const struct net_device *dev);
 void lan966x_register_notifier_blocks(void);
 void lan966x_unregister_notifier_blocks(void);
 
+bool lan966x_hw_offload(struct lan966x *lan966x, u32 port, struct sk_buff *skb);
+
+void lan966x_ifh_get_src_port(void *ifh, u64 *src_port);
+void lan966x_ifh_get_timestamp(void *ifh, u64 *timestamp);
+
 void lan966x_stats_get(struct net_device *dev,
 		       struct rtnl_link_stats64 *stats);
 int lan966x_stats_init(struct lan966x *lan966x);
-- 
2.33.0

[PATCH net-next 4/5] net: lan966x: Add FDMA functionality

[Horatiu Vultur]

Ethernet frames can be extracted or injected to or from the device's
DDR memory. There is one channel for injection and one channel for
extraction. Each of these channels contain a linked list of DCBs which
contains DB. The DCB for injection contains only 1 DB and the DCB for
extraction contains 3 DBs. Each DB contains a frame. Everytime when a
frame is received or transmitted an interrupt is generated.

It is not possible to use both the FDMA and the manual
injection/extraction of the frames. Therefore the FDMA has priority over
the manual because of better performance values.

FDMA:
[  5]   0.00-10.01  sec   456 MBytes   382 Mbits/sec    0 sender
[  5]   0.00-10.05  sec   315 MBytes   263 Mbits/sec      receiver

Manual:
[  5]   0.00-10.02  sec  94.0 MBytes  78.7 Mbits/sec    0 sender
[  5]   0.00-10.01  sec   118 MBytes  98.6 Mbits/sec      receiver

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
---
 .../net/ethernet/microchip/lan966x/Makefile   |   2 +-
 .../ethernet/microchip/lan966x/lan966x_fdma.c | 677 ++++++++++++++++++
 .../ethernet/microchip/lan966x/lan966x_main.c |  31 +-
 .../ethernet/microchip/lan966x/lan966x_main.h | 111 +++
 .../ethernet/microchip/lan966x/lan966x_port.c |   3 +
 5 files changed, 820 insertions(+), 4 deletions(-)
 create mode 100644 drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c

diff --git a/drivers/net/ethernet/microchip/lan966x/Makefile b/drivers/net/ethernet/microchip/lan966x/Makefile
index a9ffc719aa0e..fd2e0ebb2427 100644
--- a/drivers/net/ethernet/microchip/lan966x/Makefile
+++ b/drivers/net/ethernet/microchip/lan966x/Makefile
@@ -8,4 +8,4 @@ obj-$(CONFIG_LAN966X_SWITCH) += lan966x-switch.o
 lan966x-switch-objs  := lan966x_main.o lan966x_phylink.o lan966x_port.o \
 			lan966x_mac.o lan966x_ethtool.o lan966x_switchdev.o \
 			lan966x_vlan.o lan966x_fdb.o lan966x_mdb.o \
-			lan966x_ptp.o
+			lan966x_ptp.o lan966x_fdma.o
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c b/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
new file mode 100644
index 000000000000..c23e521a1f8b
--- /dev/null
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
@@ -0,0 +1,677 @@
+// SPDX-License-Identifier: GPL-2.0+
+
+#include "lan966x_main.h"
+
+static int lan966x_fdma_channel_active(struct lan966x *lan966x)
+{
+	return lan_rd(lan966x, FDMA_CH_ACTIVE);
+}
+
+static struct sk_buff *lan966x_fdma_rx_alloc_skb(struct lan966x_rx *rx,
+						 struct lan966x_db *db)
+{
+	struct lan966x *lan966x = rx->lan966x;
+	struct sk_buff *skb;
+	dma_addr_t dma_addr;
+	struct page *page;
+	void *buff_addr;
+
+	page = dev_alloc_pages(rx->page_order);
+	if (unlikely(!page))
+		return NULL;
+
+	dma_addr = dma_map_page(lan966x->dev, page, 0,
+				PAGE_SIZE << rx->page_order,
+				DMA_FROM_DEVICE);
+	if (unlikely(dma_mapping_error(lan966x->dev, dma_addr))) {
+		__free_pages(page, rx->page_order);
+		return NULL;
+	}
+
+	buff_addr = page_address(page);
+	skb = build_skb(buff_addr, PAGE_SIZE << rx->page_order);
+
+	if (unlikely(!skb)) {
+		dev_err_ratelimited(lan966x->dev,
+				    "build_skb failed !\n");
+		dma_unmap_single(lan966x->dev, dma_addr,
+				 PAGE_SIZE << rx->page_order,
+				 DMA_FROM_DEVICE);
+		__free_pages(page, rx->page_order);
+		return NULL;
+	}
+
+	db->dataptr = dma_addr;
+	return skb;
+}
+
+static void lan966x_fdma_rx_free_skbs(struct lan966x_rx *rx)
+{
+	struct lan966x *lan966x = rx->lan966x;
+	struct lan966x_rx_dcb *dcb;
+	struct lan966x_db *db;
+	int i, j;
+
+	for (i = 0; i < FDMA_DCB_MAX; ++i) {
+		dcb = &rx->dcbs[i];
+
+		for (j = 0; j < FDMA_RX_DCB_MAX_DBS; ++j) {
+			db = &dcb->db[j];
+			dma_unmap_single(lan966x->dev,
+					 (dma_addr_t)db->dataptr,
+					 PAGE_SIZE << rx->page_order,
+					 DMA_FROM_DEVICE);
+			kfree_skb(rx->skb[i][j]);
+		}
+	}
+}
+
+static void lan966x_fdma_rx_add_dcb(struct lan966x_rx *rx,
+				    struct lan966x_rx_dcb *dcb,
+				    u64 nextptr)
+{
+	struct lan966x_db *db;
+	int i;
+
+	for (i = 0; i < FDMA_RX_DCB_MAX_DBS; ++i) {
+		db = &dcb->db[i];
+		db->status = FDMA_DCB_STATUS_INTR;
+	}
+
+	dcb->nextptr = FDMA_DCB_INVALID_DATA;
+	dcb->info = FDMA_DCB_INFO_DATAL(PAGE_SIZE << rx->page_order);
+
+	rx->last_entry->nextptr = nextptr;
+	rx->last_entry = dcb;
+}
+
+static int lan966x_fdma_rx_alloc(struct lan966x_rx *rx)
+{
+	struct lan966x *lan966x = rx->lan966x;
+	struct lan966x_rx_dcb *dcb;
+	struct lan966x_db *db;
+	struct sk_buff *skb;
+	int i, j;
+	int size;
+
+	/* calculate how many pages are needed to allocate the dcbs */
+	size = sizeof(struct lan966x_rx_dcb) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+
+	rx->dcbs = dma_alloc_coherent(lan966x->dev, size, &rx->dma, GFP_ATOMIC);
+	rx->last_entry = rx->dcbs;
+	rx->db_index = 0;
+	rx->dcb_index = 0;
+
+	/* Now for each dcb allocate the dbs */
+	for (i = 0; i < FDMA_DCB_MAX; ++i) {
+		dcb = &rx->dcbs[i];
+		dcb->info = 0;
+
+		/* For each db allocate a skb and map skb data pointer to the DB
+		 * dataptr. In this way when the frame is received the skb->data
+		 * will contain the frame, so it is not needed any memcpy
+		 */
+		for (j = 0; j < FDMA_RX_DCB_MAX_DBS; ++j) {
+			db = &dcb->db[j];
+			skb = lan966x_fdma_rx_alloc_skb(rx, db);
+			if (!skb)
+				return -ENOMEM;
+
+			db->status = 0;
+			rx->skb[i][j] = skb;
+		}
+
+		lan966x_fdma_rx_add_dcb(rx, dcb, rx->dma + sizeof(*dcb) * i);
+	}
+
+	return 0;
+}
+
+static void lan966x_fdma_rx_free(struct lan966x_rx *rx)
+{
+	struct lan966x *lan966x = rx->lan966x;
+	u32 size;
+
+	/* Now it is possible to do the cleanup of dcb */
+	size = sizeof(struct lan966x_tx_dcb) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+	dma_free_coherent(lan966x->dev, size, rx->dcbs, rx->dma);
+}
+
+static void lan966x_fdma_rx_start(struct lan966x_rx *rx)
+{
+	struct lan966x *lan966x = rx->lan966x;
+	u32 mask;
+
+	/* When activating a channel, first is required to write the first DCB
+	 * address and then to activate it
+	 */
+	lan_wr(((u64)rx->dma) & GENMASK(31, 0), lan966x,
+	       FDMA_DCB_LLP(rx->channel_id));
+	lan_wr(((u64)rx->dma) >> 32, lan966x, FDMA_DCB_LLP1(rx->channel_id));
+
+	lan_wr(FDMA_CH_CFG_CH_DCB_DB_CNT_SET(FDMA_RX_DCB_MAX_DBS) |
+	       FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY_SET(1) |
+	       FDMA_CH_CFG_CH_INJ_PORT_SET(0) |
+	       FDMA_CH_CFG_CH_MEM_SET(1),
+	       lan966x, FDMA_CH_CFG(rx->channel_id));
+
+	/* Start fdma */
+	lan_rmw(FDMA_PORT_CTRL_XTR_STOP_SET(0),
+		FDMA_PORT_CTRL_XTR_STOP,
+		lan966x, FDMA_PORT_CTRL(0));
+
+	/* Enable interrupts */
+	mask = lan_rd(lan966x, FDMA_INTR_DB_ENA);
+	mask = FDMA_INTR_DB_ENA_INTR_DB_ENA_GET(mask);
+	mask |= BIT(rx->channel_id);
+	lan_rmw(FDMA_INTR_DB_ENA_INTR_DB_ENA_SET(mask),
+		FDMA_INTR_DB_ENA_INTR_DB_ENA,
+		lan966x, FDMA_INTR_DB_ENA);
+
+	/* Activate the channel */
+	lan_rmw(FDMA_CH_ACTIVATE_CH_ACTIVATE_SET(BIT(rx->channel_id)),
+		FDMA_CH_ACTIVATE_CH_ACTIVATE,
+		lan966x, FDMA_CH_ACTIVATE);
+}
+
+static void lan966x_fdma_rx_disable(struct lan966x_rx *rx)
+{
+	struct lan966x *lan966x = rx->lan966x;
+	u32 val;
+
+	/* Disable the channel */
+	lan_rmw(FDMA_CH_DISABLE_CH_DISABLE_SET(BIT(rx->channel_id)),
+		FDMA_CH_DISABLE_CH_DISABLE,
+		lan966x, FDMA_CH_DISABLE);
+
+	readx_poll_timeout_atomic(lan966x_fdma_channel_active, lan966x,
+				  val, !(val & BIT(rx->channel_id)),
+				  READL_SLEEP_US, READL_TIMEOUT_US);
+
+	lan_rmw(FDMA_CH_DB_DISCARD_DB_DISCARD_SET(BIT(rx->channel_id)),
+		FDMA_CH_DB_DISCARD_DB_DISCARD,
+		lan966x, FDMA_CH_DB_DISCARD);
+}
+
+static void lan966x_fdma_rx_reload(struct lan966x_rx *rx)
+{
+	struct lan966x *lan966x = rx->lan966x;
+
+	lan_rmw(FDMA_CH_RELOAD_CH_RELOAD_SET(BIT(rx->channel_id)),
+		FDMA_CH_RELOAD_CH_RELOAD,
+		lan966x, FDMA_CH_RELOAD);
+}
+
+static void lan966x_fdma_tx_add_dcb(struct lan966x_tx *tx,
+				    struct lan966x_tx_dcb *dcb)
+{
+	dcb->nextptr = FDMA_DCB_INVALID_DATA;
+	dcb->info = 0;
+}
+
+static int lan966x_fdma_tx_alloc(struct lan966x_tx *tx)
+{
+	struct lan966x *lan966x = tx->lan966x;
+	struct lan966x_tx_dcb *dcb;
+	struct lan966x_db *db;
+	int size;
+	int i, j;
+
+	tx->dcbs_buf = kcalloc(FDMA_DCB_MAX, sizeof(struct lan966x_tx_dcb_buf),
+			       GFP_ATOMIC);
+	if (!tx->dcbs_buf)
+		return -ENOMEM;
+
+	/* calculate how many pages are needed to allocate the dcbs */
+	size = sizeof(struct lan966x_tx_dcb) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+	tx->dcbs = dma_alloc_coherent(lan966x->dev, size, &tx->dma, GFP_ATOMIC);
+
+	/* Now for each dcb allocate the db */
+	for (i = 0; i < FDMA_DCB_MAX; ++i) {
+		dcb = &tx->dcbs[i];
+
+		for (j = 0; j < FDMA_TX_DCB_MAX_DBS; ++j) {
+			db = &dcb->db[j];
+			db->dataptr = 0;
+			db->status = 0;
+		}
+
+		lan966x_fdma_tx_add_dcb(tx, dcb);
+	}
+
+	return 0;
+}
+
+static void lan966x_fdma_tx_free(struct lan966x_tx *tx)
+{
+	struct lan966x *lan966x = tx->lan966x;
+	int size;
+
+	kfree(tx->dcbs_buf);
+
+	size = sizeof(struct lan966x_tx_dcb) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+	dma_free_coherent(lan966x->dev, size, tx->dcbs, tx->dma);
+}
+
+static void lan966x_fdma_tx_activate(struct lan966x_tx *tx)
+{
+	struct lan966x *lan966x = tx->lan966x;
+	u32 mask;
+
+	/* When activating a channel, first is required to write the first DCB
+	 * address and then to activate it
+	 */
+	lan_wr(((u64)tx->dma) & GENMASK(31, 0), lan966x,
+	       FDMA_DCB_LLP(tx->channel_id));
+	lan_wr(((u64)tx->dma) >> 32, lan966x, FDMA_DCB_LLP1(tx->channel_id));
+
+	lan_wr(FDMA_CH_CFG_CH_DCB_DB_CNT_SET(FDMA_TX_DCB_MAX_DBS) |
+	       FDMA_CH_CFG_CH_INTR_DB_EOF_ONLY_SET(1) |
+	       FDMA_CH_CFG_CH_INJ_PORT_SET(0) |
+	       FDMA_CH_CFG_CH_MEM_SET(1),
+	       lan966x, FDMA_CH_CFG(tx->channel_id));
+
+	/* Start fdma */
+	lan_rmw(FDMA_PORT_CTRL_INJ_STOP_SET(0),
+		FDMA_PORT_CTRL_INJ_STOP,
+		lan966x, FDMA_PORT_CTRL(0));
+
+	/* Enable interrupts */
+	mask = lan_rd(lan966x, FDMA_INTR_DB_ENA);
+	mask = FDMA_INTR_DB_ENA_INTR_DB_ENA_GET(mask);
+	mask |= BIT(tx->channel_id);
+	lan_rmw(FDMA_INTR_DB_ENA_INTR_DB_ENA_SET(mask),
+		FDMA_INTR_DB_ENA_INTR_DB_ENA,
+		lan966x, FDMA_INTR_DB_ENA);
+
+	/* Activate the channel */
+	lan_rmw(FDMA_CH_ACTIVATE_CH_ACTIVATE_SET(BIT(tx->channel_id)),
+		FDMA_CH_ACTIVATE_CH_ACTIVATE,
+		lan966x, FDMA_CH_ACTIVATE);
+}
+
+static void lan966x_fdma_tx_disable(struct lan966x_tx *tx)
+{
+	struct lan966x *lan966x = tx->lan966x;
+	u32 val;
+
+	/* Disable the channel */
+	lan_rmw(FDMA_CH_DISABLE_CH_DISABLE_SET(BIT(tx->channel_id)),
+		FDMA_CH_DISABLE_CH_DISABLE,
+		lan966x, FDMA_CH_DISABLE);
+
+	readx_poll_timeout_atomic(lan966x_fdma_channel_active, lan966x,
+				  val, !(val & BIT(tx->channel_id)),
+				  READL_SLEEP_US, READL_TIMEOUT_US);
+
+	lan_rmw(FDMA_CH_DB_DISCARD_DB_DISCARD_SET(BIT(tx->channel_id)),
+		FDMA_CH_DB_DISCARD_DB_DISCARD,
+		lan966x, FDMA_CH_DB_DISCARD);
+
+	tx->activated = false;
+}
+
+static void lan966x_fdma_tx_reload(struct lan966x_tx *tx)
+{
+	struct lan966x *lan966x = tx->lan966x;
+
+	/* Write the registers to reload the channel */
+	lan_rmw(FDMA_CH_RELOAD_CH_RELOAD_SET(BIT(tx->channel_id)),
+		FDMA_CH_RELOAD_CH_RELOAD,
+		lan966x, FDMA_CH_RELOAD);
+}
+
+static void lan966x_fdma_wakeup_netdev(struct lan966x *lan966x)
+{
+	struct lan966x_port *port;
+	int i;
+
+	for (i = 0; i < lan966x->num_phys_ports; ++i) {
+		port = lan966x->ports[i];
+		if (!port)
+			continue;
+
+		if (netif_queue_stopped(port->dev))
+			netif_wake_queue(port->dev);
+	}
+}
+
+static void lan966x_fdma_tx_clear_buf(struct lan966x *lan966x, int weight)
+{
+	struct lan966x_tx *tx = &lan966x->tx;
+	struct lan966x_tx_dcb_buf *dcb_buf;
+	struct lan966x_db *db;
+	unsigned long flags;
+	bool clear = false;
+	int i;
+
+	spin_lock_irqsave(&lan966x->tx_lock, flags);
+	for (i = 0; i < FDMA_DCB_MAX; ++i) {
+		dcb_buf = &tx->dcbs_buf[i];
+
+		if (!dcb_buf->used)
+			continue;
+
+		db = &tx->dcbs[i].db[0];
+		if (!(db->status & FDMA_DCB_STATUS_DONE))
+			continue;
+
+		dcb_buf->used = false;
+		dma_unmap_single(lan966x->dev,
+				 dcb_buf->dma_addr,
+				 dcb_buf->skb->len,
+				 DMA_TO_DEVICE);
+		if (!dcb_buf->ptp)
+			dev_kfree_skb_any(dcb_buf->skb);
+
+		clear = true;
+	}
+	spin_unlock_irqrestore(&lan966x->tx_lock, flags);
+
+	if (clear)
+		lan966x_fdma_wakeup_netdev(lan966x);
+}
+
+static struct sk_buff *lan966x_fdma_rx_get_frame(struct lan966x_rx *rx)
+{
+	struct lan966x *lan966x = rx->lan966x;
+	u64 src_port, timestamp;
+	struct sk_buff *new_skb;
+	struct lan966x_db *db;
+	struct sk_buff *skb;
+
+	/* Check if there is any data */
+	db = &rx->dcbs[rx->dcb_index].db[rx->db_index];
+	if (unlikely(!(db->status & FDMA_DCB_STATUS_DONE)))
+		return NULL;
+
+	/* Get the received frame and unmap it */
+	skb = rx->skb[rx->dcb_index][rx->db_index];
+	dma_unmap_single(lan966x->dev, (dma_addr_t)db->dataptr,
+			 FDMA_DCB_STATUS_BLOCKL(db->status),
+			 DMA_FROM_DEVICE);
+
+	/* Allocate a new skb and map it */
+	new_skb = lan966x_fdma_rx_alloc_skb(rx, db);
+	if (unlikely(!new_skb))
+		return NULL;
+
+	rx->skb[rx->dcb_index][rx->db_index] = new_skb;
+
+	skb_put(skb, FDMA_DCB_STATUS_BLOCKL(db->status));
+
+	lan966x_ifh_get_src_port(skb->data, &src_port);
+	lan966x_ifh_get_timestamp(skb->data, &timestamp);
+
+	WARN_ON(src_port >= lan966x->num_phys_ports);
+
+	skb->dev = lan966x->ports[src_port]->dev;
+	skb_pull(skb, IFH_LEN * sizeof(u32));
+
+	if (likely(!(skb->dev->features & NETIF_F_RXFCS)))
+		skb_trim(skb, skb->len - ETH_FCS_LEN);
+
+	lan966x_ptp_rxtstamp(lan966x, skb, timestamp);
+	skb->protocol = eth_type_trans(skb, skb->dev);
+
+	if (lan966x->bridge_mask & BIT(src_port)) {
+		skb->offload_fwd_mark = 1;
+
+		skb_reset_network_header(skb);
+		if (!lan966x_hw_offload(lan966x, src_port, skb))
+			skb->offload_fwd_mark = 0;
+	}
+
+	skb->dev->stats.rx_bytes += skb->len;
+	skb->dev->stats.rx_packets++;
+
+	return skb;
+}
+
+static int lan966x_fdma_napi_poll(struct napi_struct *napi, int weight)
+{
+	struct lan966x *lan966x = container_of(napi, struct lan966x, napi);
+	struct lan966x_rx *rx = &lan966x->rx;
+	struct list_head rx_list;
+	int counter = 0;
+
+	lan966x_fdma_tx_clear_buf(lan966x, weight);
+
+	INIT_LIST_HEAD(&rx_list);
+
+	while (counter < weight) {
+		struct lan966x_rx_dcb *old_dcb;
+		struct sk_buff *skb;
+		u64 nextptr;
+
+		skb = lan966x_fdma_rx_get_frame(rx);
+		if (!skb)
+			break;
+		list_add_tail(&skb->list, &rx_list);
+
+		rx->db_index++;
+		counter++;
+
+		/* Check if the DCB can be reused */
+		if (rx->db_index != FDMA_RX_DCB_MAX_DBS)
+			continue;
+
+		/* Now the DCB  can be reused, just advance the dcb_index
+		 * pointer and set the nextptr in the DCB
+		 */
+		rx->db_index = 0;
+
+		old_dcb = &rx->dcbs[rx->dcb_index];
+		rx->dcb_index++;
+		rx->dcb_index &= FDMA_DCB_MAX - 1;
+
+		nextptr = rx->dma + ((unsigned long)old_dcb -
+				     (unsigned long)rx->dcbs);
+		lan966x_fdma_rx_add_dcb(rx, old_dcb, nextptr);
+		lan966x_fdma_rx_reload(rx);
+	}
+
+	if (counter < weight) {
+		napi_complete_done(napi, counter);
+		lan_wr(0xff, lan966x, FDMA_INTR_DB_ENA);
+	}
+
+	netif_receive_skb_list(&rx_list);
+
+	return counter;
+}
+
+irqreturn_t lan966x_fdma_irq_handler(int irq, void *args)
+{
+	struct lan966x *lan966x = args;
+	u32 db, err, err_type;
+
+	db = lan_rd(lan966x, FDMA_INTR_DB);
+	err = lan_rd(lan966x, FDMA_INTR_ERR);
+
+	if (db) {
+		lan_wr(0, lan966x, FDMA_INTR_DB_ENA);
+		lan_wr(db, lan966x, FDMA_INTR_DB);
+
+		napi_schedule(&lan966x->napi);
+	}
+
+	if (err) {
+		err_type = lan_rd(lan966x, FDMA_ERRORS);
+
+		WARN(1, "Unexpected error: %d, error_type: %d\n", err, err_type);
+
+		lan_wr(err, lan966x, FDMA_INTR_ERR);
+		lan_wr(err_type, lan966x, FDMA_ERRORS);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static int lan966x_fdma_get_next_dcb(struct lan966x_tx *tx)
+{
+	struct lan966x_tx_dcb_buf *dcb_buf;
+	int i;
+
+	for (i = 0; i < FDMA_DCB_MAX; ++i) {
+		dcb_buf = &tx->dcbs_buf[i];
+		if (!dcb_buf->used && i != tx->last_in_use)
+			return i;
+	}
+
+	return -1;
+}
+
+int lan966x_fdma_xmit(struct sk_buff *skb, __be32 *ifh, struct net_device *dev)
+{
+	struct lan966x_port *port = netdev_priv(dev);
+	struct lan966x *lan966x = port->lan966x;
+	struct lan966x_tx_dcb_buf *next_dcb_buf;
+	struct lan966x_tx_dcb *next_dcb, *dcb;
+	struct lan966x_tx *tx = &lan966x->tx;
+	struct lan966x_db *next_db;
+	dma_addr_t dma_addr;
+	int next_to_use;
+	int err;
+
+	/* Get next index */
+	next_to_use = lan966x_fdma_get_next_dcb(tx);
+	if (next_to_use < 0) {
+		netif_stop_queue(dev);
+		err = NETDEV_TX_BUSY;
+		goto out;
+	}
+
+	if (skb_put_padto(skb, ETH_ZLEN)) {
+		dev->stats.tx_dropped++;
+		err = NETDEV_TX_OK;
+		goto out;
+	}
+
+	/* skb processing */
+	skb_tx_timestamp(skb);
+	if (skb_headroom(skb) < IFH_LEN * sizeof(u32)) {
+		err = pskb_expand_head(skb,
+				       IFH_LEN * sizeof(u32) - skb_headroom(skb),
+				       0, GFP_ATOMIC);
+		if (unlikely(err)) {
+			dev->stats.tx_dropped++;
+			err = NETDEV_TX_OK;
+			goto release;
+		}
+	}
+
+	skb_push(skb, IFH_LEN * sizeof(u32));
+	memcpy(skb->data, ifh, IFH_LEN * sizeof(u32));
+	skb_put(skb, 4);
+
+	dma_addr = dma_map_single(lan966x->dev, skb->data, skb->len,
+				  DMA_TO_DEVICE);
+	if (dma_mapping_error(lan966x->dev, dma_addr)) {
+		dev->stats.tx_dropped++;
+		err = NETDEV_TX_OK;
+		goto release;
+	}
+
+	/* Setup next dcb */
+	next_dcb = &tx->dcbs[next_to_use];
+	next_dcb->nextptr = FDMA_DCB_INVALID_DATA;
+
+	next_db = &next_dcb->db[0];
+	next_db->dataptr = dma_addr;
+	next_db->status = FDMA_DCB_STATUS_SOF |
+			  FDMA_DCB_STATUS_EOF |
+			  FDMA_DCB_STATUS_INTR |
+			  FDMA_DCB_STATUS_BLOCKO(0) |
+			  FDMA_DCB_STATUS_BLOCKL(skb->len);
+
+	/* Fill up the buffer */
+	next_dcb_buf = &tx->dcbs_buf[next_to_use];
+	next_dcb_buf->skb = skb;
+	next_dcb_buf->dma_addr = dma_addr;
+	next_dcb_buf->used = true;
+	next_dcb_buf->ptp = false;
+
+	if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP &&
+	    LAN966X_SKB_CB(skb)->rew_op == IFH_REW_OP_TWO_STEP_PTP)
+		next_dcb_buf->ptp = true;
+
+	if (likely(lan966x->tx.activated)) {
+		/* Connect current dcb to the next db */
+		dcb = &tx->dcbs[tx->last_in_use];
+		dcb->nextptr = tx->dma + (next_to_use *
+					  sizeof(struct lan966x_tx_dcb));
+
+		lan966x_fdma_tx_reload(tx);
+	} else {
+		/* Because it is first time, then just activate */
+		lan966x->tx.activated = true;
+		lan966x_fdma_tx_activate(tx);
+	}
+
+	/* Move to next dcb because this last in use */
+	tx->last_in_use = next_to_use;
+
+	dev->stats.tx_packets++;
+	dev->stats.tx_bytes += skb->len;
+
+	return NETDEV_TX_OK;
+
+out:
+	return err;
+
+release:
+	if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP &&
+	    LAN966X_SKB_CB(skb)->rew_op == IFH_REW_OP_TWO_STEP_PTP)
+		lan966x_ptp_txtstamp_release(port, skb);
+
+	dev_kfree_skb_any(skb);
+	return err;
+}
+
+void lan966x_fdma_netdev_init(struct lan966x *lan966x, struct net_device *dev)
+{
+	if (lan966x->fdma_ndev)
+		return;
+
+	lan966x->fdma_ndev = dev;
+	netif_napi_add(dev, &lan966x->napi, lan966x_fdma_napi_poll,
+		       FDMA_WEIGHT);
+	napi_enable(&lan966x->napi);
+}
+
+void lan966x_fdma_netdev_deinit(struct lan966x *lan966x, struct net_device *dev)
+{
+	if (lan966x->fdma_ndev == dev) {
+		netif_napi_del(&lan966x->napi);
+		lan966x->fdma_ndev = NULL;
+	}
+}
+
+void lan966x_fdma_init(struct lan966x *lan966x)
+{
+	lan966x->rx.lan966x = lan966x;
+	lan966x->rx.channel_id = FDMA_XTR_CHANNEL;
+	lan966x->tx.lan966x = lan966x;
+	lan966x->tx.channel_id = FDMA_INJ_CHANNEL;
+	lan966x->tx.last_in_use = -1;
+
+	lan966x_fdma_rx_alloc(&lan966x->rx);
+	lan966x_fdma_tx_alloc(&lan966x->tx);
+
+	lan966x_fdma_rx_start(&lan966x->rx);
+}
+
+void lan966x_fdma_deinit(struct lan966x *lan966x)
+{
+	lan966x_fdma_rx_disable(&lan966x->rx);
+	lan966x_fdma_tx_disable(&lan966x->tx);
+
+	lan966x_fdma_rx_free_skbs(&lan966x->rx);
+	lan966x_fdma_rx_free(&lan966x->rx);
+	lan966x_fdma_tx_free(&lan966x->tx);
+}
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
index 2c82f847ae6d..6cb9fffc3058 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
@@ -341,7 +341,10 @@ static int lan966x_port_xmit(struct sk_buff *skb, struct net_device *dev)
 	}
 
 	spin_lock(&lan966x->tx_lock);
-	err = lan966x_port_ifh_xmit(skb, ifh, dev);
+	if (port->lan966x->fdma)
+		err = lan966x_fdma_xmit(skb, ifh, dev);
+	else
+		err = lan966x_port_ifh_xmit(skb, ifh, dev);
 	spin_unlock(&lan966x->tx_lock);
 
 	return err;
@@ -640,6 +643,9 @@ static void lan966x_cleanup_ports(struct lan966x *lan966x)
 		if (port->dev)
 			unregister_netdev(port->dev);
 
+		if (lan966x->fdma && lan966x->fdma_ndev == port->dev)
+			lan966x_fdma_netdev_deinit(lan966x, port->dev);
+
 		if (port->phylink) {
 			rtnl_lock();
 			lan966x_port_stop(port->dev);
@@ -659,6 +665,11 @@ static void lan966x_cleanup_ports(struct lan966x *lan966x)
 		disable_irq(lan966x->ana_irq);
 		lan966x->ana_irq = -ENXIO;
 	}
+
+	if (lan966x->fdma_irq) {
+		disable_irq(lan966x->fdma_irq);
+		lan966x->fdma_irq = -ENXIO;
+	}
 }
 
 static int lan966x_probe_port(struct lan966x *lan966x, u32 p,
@@ -787,12 +798,12 @@ static void lan966x_init(struct lan966x *lan966x)
 	/* Do byte-swap and expect status after last data word
 	 * Extraction: Mode: manual extraction) | Byte_swap
 	 */
-	lan_wr(QS_XTR_GRP_CFG_MODE_SET(1) |
+	lan_wr(QS_XTR_GRP_CFG_MODE_SET(lan966x->fdma ? 2 : 1) |
 	       QS_XTR_GRP_CFG_BYTE_SWAP_SET(1),
 	       lan966x, QS_XTR_GRP_CFG(0));
 
 	/* Injection: Mode: manual injection | Byte_swap */
-	lan_wr(QS_INJ_GRP_CFG_MODE_SET(1) |
+	lan_wr(QS_INJ_GRP_CFG_MODE_SET(lan966x->fdma ? 2 : 1) |
 	       QS_INJ_GRP_CFG_BYTE_SWAP_SET(1),
 	       lan966x, QS_INJ_GRP_CFG(0));
 
@@ -1014,6 +1025,17 @@ static int lan966x_probe(struct platform_device *pdev)
 		lan966x->ptp = 1;
 	}
 
+	lan966x->fdma_irq = platform_get_irq_byname(pdev, "fdma");
+	if (lan966x->fdma_irq > 0) {
+		err = devm_request_threaded_irq(&pdev->dev, lan966x->fdma_irq, NULL,
+						lan966x_fdma_irq_handler, IRQF_ONESHOT,
+						"fdma irq", lan966x);
+		if (err)
+			return dev_err_probe(&pdev->dev, err, "Unable to use fdma irq");
+
+		lan966x->fdma = true;
+	}
+
 	/* init switch */
 	lan966x_init(lan966x);
 	lan966x_stats_init(lan966x);
@@ -1043,6 +1065,8 @@ static int lan966x_probe(struct platform_device *pdev)
 		lan966x_port_init(lan966x->ports[p]);
 	}
 
+	lan966x_fdma_init(lan966x);
+
 	lan966x_mdb_init(lan966x);
 	err = lan966x_fdb_init(lan966x);
 	if (err)
@@ -1073,6 +1097,7 @@ static int lan966x_remove(struct platform_device *pdev)
 {
 	struct lan966x *lan966x = platform_get_drvdata(pdev);
 
+	lan966x_fdma_deinit(lan966x);
 	lan966x_cleanup_ports(lan966x);
 
 	cancel_delayed_work_sync(&lan966x->stats_work);
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_main.h b/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
index b692c612f235..bfa7feea2b56 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
@@ -61,6 +61,23 @@
 #define IFH_REW_OP_ONE_STEP_PTP		0x3
 #define IFH_REW_OP_TWO_STEP_PTP		0x4
 
+#define FDMA_RX_DCB_MAX_DBS		3
+#define FDMA_TX_DCB_MAX_DBS		1
+#define FDMA_DCB_INFO_DATAL(x)		((x) & GENMASK(15, 0))
+
+#define FDMA_DCB_STATUS_BLOCKL(x)	((x) & GENMASK(15, 0))
+#define FDMA_DCB_STATUS_SOF		BIT(16)
+#define FDMA_DCB_STATUS_EOF		BIT(17)
+#define FDMA_DCB_STATUS_INTR		BIT(18)
+#define FDMA_DCB_STATUS_DONE		BIT(19)
+#define FDMA_DCB_STATUS_BLOCKO(x)	(((x) << 20) & GENMASK(31, 20))
+#define FDMA_DCB_INVALID_DATA		0x1
+
+#define FDMA_XTR_CHANNEL		6
+#define FDMA_INJ_CHANNEL		0
+#define FDMA_DCB_MAX			512
+#define FDMA_WEIGHT			64
+
 /* MAC table entry types.
  * ENTRYTYPE_NORMAL is subject to aging.
  * ENTRYTYPE_LOCKED is not subject to aging.
@@ -76,6 +93,85 @@ enum macaccess_entry_type {
 
 struct lan966x_port;
 
+struct lan966x_db {
+	u64 dataptr;
+	u64 status;
+};
+
+struct lan966x_rx_dcb {
+	u64 nextptr;
+	u64 info;
+	struct lan966x_db db[FDMA_RX_DCB_MAX_DBS];
+};
+
+struct lan966x_tx_dcb {
+	u64 nextptr;
+	u64 info;
+	struct lan966x_db db[FDMA_TX_DCB_MAX_DBS];
+};
+
+struct lan966x_rx {
+	struct lan966x *lan966x;
+
+	/* Pointer to the array of hardware dcbs. */
+	struct lan966x_rx_dcb *dcbs;
+
+	/* Pointer to the last address in the dcbs. */
+	struct lan966x_rx_dcb *last_entry;
+
+	/* For each DB, there is a skb, and the skb data pointer is mapped in
+	 * the DB. Once a frame is received the skb is given to the upper layers
+	 * and a new skb is added to the db.
+	 */
+	struct sk_buff *skb[FDMA_DCB_MAX][FDMA_RX_DCB_MAX_DBS];
+
+	/* Represents the db_index, it can have a value between 0 and
+	 * FDMA_RX_DCB_MAX_DBS, once it reaches the value of FDMA_RX_DCB_MAX_DBS
+	 * it means that the DCB can be reused.
+	 */
+	int db_index;
+
+	/* Represents the index in the dcbs. It has a value between 0 and
+	 * FDMA_DCB_MAX
+	 */
+	int dcb_index;
+
+	/* Represents the dma address to the dcbs array */
+	dma_addr_t dma;
+
+	/* Represents the page order that is used to allocate the pages for the
+	 * RX buffers. This value is calculated based on max MTU of the devices.
+	 */
+	u8 page_order;
+
+	u8 channel_id;
+};
+
+struct lan966x_tx_dcb_buf {
+	struct sk_buff *skb;
+	dma_addr_t dma_addr;
+	bool used;
+	bool ptp;
+};
+
+struct lan966x_tx {
+	struct lan966x *lan966x;
+
+	/* Pointer to the dcb list */
+	struct lan966x_tx_dcb *dcbs;
+	u16 last_in_use;
+
+	/* Represents the DMA address to the first entry of the dcb entries. */
+	dma_addr_t dma;
+
+	/* Array of dcbs that are given to the HW */
+	struct lan966x_tx_dcb_buf *dcbs_buf;
+
+	u8 channel_id;
+
+	bool activated;
+};
+
 struct lan966x_stat_layout {
 	u32 offset;
 	char name[ETH_GSTRING_LEN];
@@ -137,6 +233,7 @@ struct lan966x {
 	int xtr_irq;
 	int ana_irq;
 	int ptp_irq;
+	int fdma_irq;
 
 	/* worqueue for fdb */
 	struct workqueue_struct *fdb_work;
@@ -153,6 +250,13 @@ struct lan966x {
 	spinlock_t ptp_ts_id_lock; /* lock for ts_id */
 	struct mutex ptp_lock; /* lock for ptp interface state */
 	u16 ptp_skbs;
+
+	/* fdma */
+	bool fdma;
+	struct net_device *fdma_ndev;
+	struct lan966x_rx rx;
+	struct lan966x_tx tx;
+	struct napi_struct napi;
 };
 
 struct lan966x_port_config {
@@ -292,6 +396,13 @@ void lan966x_ptp_txtstamp_release(struct lan966x_port *port,
 				  struct sk_buff *skb);
 irqreturn_t lan966x_ptp_irq_handler(int irq, void *args);
 
+int lan966x_fdma_xmit(struct sk_buff *skb, __be32 *ifh, struct net_device *dev);
+void lan966x_fdma_netdev_init(struct lan966x *lan966x, struct net_device *dev);
+void lan966x_fdma_netdev_deinit(struct lan966x *lan966x, struct net_device *dev);
+void lan966x_fdma_init(struct lan966x *lan966x);
+void lan966x_fdma_deinit(struct lan966x *lan966x);
+irqreturn_t lan966x_fdma_irq_handler(int irq, void *args);
+
 static inline void __iomem *lan_addr(void __iomem *base[],
 				     int id, int tinst, int tcnt,
 				     int gbase, int ginst,
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_port.c b/drivers/net/ethernet/microchip/lan966x/lan966x_port.c
index 237555845a52..f141644e4372 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_port.c
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_port.c
@@ -393,6 +393,9 @@ void lan966x_port_init(struct lan966x_port *port)
 
 	lan966x_port_config_down(port);
 
+	if (lan966x->fdma)
+		lan966x_fdma_netdev_init(lan966x, port->dev);
+
 	if (config->portmode != PHY_INTERFACE_MODE_QSGMII)
 		return;
 
-- 
2.33.0

[PATCH net-next 5/5] net: lan96x: Update FDMA to change MTU.

[Horatiu Vultur]

When changing the MTU, it is required to change also the size of the
DBs. In case those frames will arrive to CPU.

Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
---
 .../ethernet/microchip/lan966x/lan966x_fdma.c | 95 +++++++++++++++++++
 .../ethernet/microchip/lan966x/lan966x_main.c |  2 +-
 .../ethernet/microchip/lan966x/lan966x_main.h |  1 +
 3 files changed, 97 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c b/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
index c23e521a1f8b..a33329cc4834 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
@@ -633,6 +633,101 @@ int lan966x_fdma_xmit(struct sk_buff *skb, __be32 *ifh, struct net_device *dev)
 	return err;
 }
 
+static int lan966x_fdma_get_max_mtu(struct lan966x *lan966x)
+{
+	int max_mtu = 0;
+	int i;
+
+	for (i = 0; i < lan966x->num_phys_ports; ++i) {
+		int mtu;
+
+		if (!lan966x->ports[i])
+			continue;
+
+		mtu = lan966x->ports[i]->dev->mtu;
+		if (mtu > max_mtu)
+			max_mtu = mtu;
+	}
+
+	return max_mtu;
+}
+
+static int lan966x_qsys_sw_status(struct lan966x *lan966x)
+{
+	return lan_rd(lan966x, QSYS_SW_STATUS(CPU_PORT));
+}
+
+static void lan966x_fdma_reload(struct lan966x *lan966x, int new_mtu)
+{
+	void *rx_dcb, *tx_dcb, *tx_dcb_buf;
+	dma_addr_t rx_dma, tx_dma;
+	unsigned long flags;
+	u32 size;
+
+	/* Store these for later to free them */
+	rx_dma = lan966x->rx.dma;
+	tx_dma = lan966x->tx.dma;
+	rx_dcb = lan966x->rx.dcbs;
+	tx_dcb = lan966x->tx.dcbs;
+	tx_dcb_buf = lan966x->tx.dcbs_buf;
+
+	lan966x_fdma_rx_disable(&lan966x->rx);
+	lan966x_fdma_rx_free_skbs(&lan966x->rx);
+	lan966x->rx.page_order = round_up(new_mtu, PAGE_SIZE) / PAGE_SIZE - 1;
+	lan966x_fdma_rx_alloc(&lan966x->rx);
+	lan966x_fdma_rx_start(&lan966x->rx);
+
+	spin_lock_irqsave(&lan966x->tx_lock, flags);
+	lan966x_fdma_tx_disable(&lan966x->tx);
+	lan966x_fdma_tx_alloc(&lan966x->tx);
+	spin_unlock_irqrestore(&lan966x->tx_lock, flags);
+
+	/* Now it is possible to clean */
+	size = sizeof(struct lan966x_tx_dcb) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+	dma_free_coherent(lan966x->dev, size, tx_dcb, tx_dma);
+
+	kfree(tx_dcb_buf);
+
+	size = sizeof(struct lan966x_rx_dcb) * FDMA_DCB_MAX;
+	size = ALIGN(size, PAGE_SIZE);
+	dma_free_coherent(lan966x->dev, size, rx_dcb, rx_dma);
+}
+
+int lan966x_fdma_change_mtu(struct lan966x *lan966x)
+{
+	int max_mtu;
+	u32 val;
+
+	max_mtu = lan966x_fdma_get_max_mtu(lan966x);
+	if (round_up(max_mtu, PAGE_SIZE) / PAGE_SIZE - 1 ==
+	    lan966x->rx.page_order)
+		return 0;
+
+	/* Disable the CPU port */
+	lan_rmw(QSYS_SW_PORT_MODE_PORT_ENA_SET(0),
+		QSYS_SW_PORT_MODE_PORT_ENA,
+		lan966x, QSYS_SW_PORT_MODE(CPU_PORT));
+
+	/* Flush the CPU queues */
+	readx_poll_timeout(lan966x_qsys_sw_status, lan966x,
+			   val, !(QSYS_SW_STATUS_EQ_AVAIL_GET(val)),
+			   READL_SLEEP_US, READL_TIMEOUT_US);
+
+	/* Add a sleep in case there are frames between the queues and the CPU
+	 * port
+	 */
+	usleep_range(1000, 2000);
+
+	lan966x_fdma_reload(lan966x, max_mtu);
+
+	/* Enable back the CPU port */
+	lan_rmw(QSYS_SW_PORT_MODE_PORT_ENA_SET(1),
+		QSYS_SW_PORT_MODE_PORT_ENA,
+		lan966x,  QSYS_SW_PORT_MODE(CPU_PORT));
+	return 0;
+}
+
 void lan966x_fdma_netdev_init(struct lan966x *lan966x, struct net_device *dev)
 {
 	if (lan966x->fdma_ndev)
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
index 6cb9fffc3058..a78fee5471e7 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_main.c
@@ -359,7 +359,7 @@ static int lan966x_port_change_mtu(struct net_device *dev, int new_mtu)
 	       lan966x, DEV_MAC_MAXLEN_CFG(port->chip_port));
 	dev->mtu = new_mtu;
 
-	return 0;
+	return !lan966x->fdma ? 0 : lan966x_fdma_change_mtu(lan966x);
 }
 
 static int lan966x_mc_unsync(struct net_device *dev, const unsigned char *addr)
diff --git a/drivers/net/ethernet/microchip/lan966x/lan966x_main.h b/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
index bfa7feea2b56..fa4016f2b5d4 100644
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_main.h
@@ -397,6 +397,7 @@ void lan966x_ptp_txtstamp_release(struct lan966x_port *port,
 irqreturn_t lan966x_ptp_irq_handler(int irq, void *args);
 
 int lan966x_fdma_xmit(struct sk_buff *skb, __be32 *ifh, struct net_device *dev);
+int lan966x_fdma_change_mtu(struct lan966x *lan966x);
 void lan966x_fdma_netdev_init(struct lan966x *lan966x, struct net_device *dev);
 void lan966x_fdma_netdev_deinit(struct lan966x *lan966x, struct net_device *dev);
 void lan966x_fdma_init(struct lan966x *lan966x);
-- 
2.33.0

Re: [PATCH net-next 1/5] dt-bindings: net: lan966x: Extend with FDMA interrupt

[Michael Walle]

> Extend dt-bindings for lan966x with FDMA interrupt. This is generated
> when receiving a frame or when a frame was transmitted. The interrupt
> needs to be enable for each frame.
> 
> Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
> ---
>  .../devicetree/bindings/net/microchip,lan966x-switch.yaml       | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml b/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
> index 13812768b923..14e0bae5965f 100644
> --- a/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
> +++ b/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
> @@ -39,6 +39,7 @@ properties:
>        - description: frame dma based extraction
>        - description: analyzer interrupt
>        - description: ptp interrupt
> +      - description: fdma interrupt
>  
>    interrupt-names:
>      minItems: 1
> @@ -47,6 +48,7 @@ properties:
>        - const: fdma
>        - const: ana
>        - const: ptp
> +      - const: fdma

This interrupt is already described (three lines above), no?

-michael

Re: [PATCH net-next 1/5] dt-bindings: net: lan966x: Extend with FDMA interrupt

[Horatiu Vultur]

The 03/18/2022 03:14, Michael Walle wrote:
> 
> > Extend dt-bindings for lan966x with FDMA interrupt. This is generated
> > when receiving a frame or when a frame was transmitted. The interrupt
> > needs to be enable for each frame.
> >
> > Signed-off-by: Horatiu Vultur <horatiu.vultur@microchip.com>
> > ---
> >  .../devicetree/bindings/net/microchip,lan966x-switch.yaml       | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml b/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
> > index 13812768b923..14e0bae5965f 100644
> > --- a/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
> > +++ b/Documentation/devicetree/bindings/net/microchip,lan966x-switch.yaml
> > @@ -39,6 +39,7 @@ properties:
> >        - description: frame dma based extraction
> >        - description: analyzer interrupt
> >        - description: ptp interrupt
> > +      - description: fdma interrupt
> >
> >    interrupt-names:
> >      minItems: 1
> > @@ -47,6 +48,7 @@ properties:
> >        - const: fdma
> >        - const: ana
> >        - const: ptp
> > +      - const: fdma
> 
> This interrupt is already described (three lines above), no?

Yes you are right.
So I will drop this patch in the next version.

> 
> -michael

-- 
/Horatiu

Re: [PATCH net-next 0/5] net: lan966x: Add support for FDMA

[Michael Walle]

Hi Horatiu,

> Currently when injecting or extracting a frame from CPU, the frame
> is given to the HW each word at a time. There is another way to
> inject/extract frames from CPU using FDMA(Frame Direct Memory Access).
> In this way the entire frame is given to the HW. This improves both
> RX and TX bitrate.

I wanted to test this. ping and such works fine and I'm also
seeing fdma interrupts. But as soon as I try iperf3 I get a skb_panic
(due to frame size?). Hope that splash below helps.

-michael

[  159.778850] skbuff: skb_over_panic: text:c07960c0 len:106 put:4 head:c8cc6e00 data:c8cc6e9a tail:0xc8cc6f04 end:0xc8cc6f00 dev:eth0
[  159.788067] ------------[ cut here ]------------
[  159.792575] kernel BUG at net/core/skbuff.c:113!
..
[  160.620995] Backtrace:
[  160.623426] [<c0a62350>] (skb_panic) from [<c08876f8>] (skb_put+0x54/0x58)
[  160.630284] [<c08876a4>] (skb_put) from [<c07960c0>] (lan966x_fdma_xmit+0x108/0x504)
[  160.638011]  r5:00000000 r4:00000000
[  160.641566] [<c0795fb8>] (lan966x_fdma_xmit) from [<c078cdf8>] (lan966x_port_xmit+0x1b0/0x450)
[  160.650171]  r10:c1960160 r9:c1960040 r8:00000001 r7:0000001c r6:c196b000 r5:c891a780
[  160.657973]  r4:00000000
[  160.660488] [<c078cc48>] (lan966x_port_xmit) from [<c08a3e24>] (dev_hard_start_xmit+0x114/0x248)
[  160.669267]  r10:c0f016e8 r9:c0f060cc r8:c0e8d350 r7:c196ba00 r6:00000000 r5:c196b000
[  160.677069]  r4:c891a780
[  160.679584] [<c08a3d10>] (dev_hard_start_xmit) from [<c0904728>] (sch_direct_xmit+0x11c/0x31c)
[  160.688188]  r10:2a01000a r9:00000000 r8:00000000 r7:c196b000 r6:c196ba00 r5:c891a780
[  160.695991]  r4:c1e57400
[  160.698507] [<c090460c>] (sch_direct_xmit) from [<c08a4520>] (__dev_queue_xmit+0x508/0xaac)
[  160.706850]  r9:c8cc6f00 r8:0000004a r7:00000001 r6:00000000 r5:c1e57400 r4:c891a780
[  160.714565] [<c08a4018>] (__dev_queue_xmit) from [<c08a4ad8>] (dev_queue_xmit+0x14/0x18)
[  160.722648]  r10:2a01000a r9:00000010 r8:0000004a r7:c1e7038c r6:00000000 r5:c891a780
[  160.730451]  r4:c1e70300
[  160.732966] [<c08a4ac4>] (dev_queue_xmit) from [<c092b0c0>] (ip_finish_output2+0x270/0x5dc)
[  160.741300] [<c092ae50>] (ip_finish_output2) from [<c092ceb8>] (__ip_finish_output+0x9c/0x144)
[  160.749903]  r9:c8cdef00 r8:c10833c0 r7:000005dc r6:c8924000 r5:c891a780 r4:c891a780
[  160.757618] [<c092ce1c>] (__ip_finish_output) from [<c092d01c>] (ip_output+0xbc/0xc4)
[  160.765441]  r10:2a01000a r9:c8cdef00 r8:c8cc6e00 r7:c196b000 r6:c8924000 r5:c10833c0
[  160.773244]  r4:c891a780
[  160.775759] [<c092cf60>] (ip_output) from [<c092ab10>] (ip_local_out+0x60/0x6c)
[  160.783058]  r7:00000000 r6:c8924000 r5:c10833c0 r4:c891a780
[  160.788692] [<c092aab0>] (ip_local_out) from [<c092ad74>] (ip_build_and_send_pkt+0x110/0x1ec)
[  160.797206]  r7:00000000 r6:c8d40000 r5:000000c4 r4:c891a780
[  160.802841] [<c092ac64>] (ip_build_and_send_pkt) from [<c0952bb8>] (tcp_v4_send_synack+0xec/0x138)
[  160.811792]  r10:c8cdef00 r9:c0b9b614 r8:c10833c0 r7:8b01000a r6:2a01000a r5:c8d40000
[  160.819595]  r4:c8924000
[  160.822110] [<c0952acc>] (tcp_v4_send_synack) from [<c0941520>] (tcp_conn_request+0x6cc/0x984)
[  160.830711]  r7:c891ab40 r6:c0952acc r5:c8d40000 r4:c8924000
[  160.836345] [<c0940e54>] (tcp_conn_request) from [<c095299c>] (tcp_v4_conn_request+0x38/0x84)
[  160.844863]  r10:c892902a r9:c892903e r8:c10833c0 r7:00000000 r6:00000000 r5:c891ab40
[  160.852666]  r4:c8d40000
[  160.855181] [<c0952964>] (tcp_v4_conn_request) from [<c09d05a0>] (tcp_v6_conn_request+0x12c/0x138)
[  160.864122] [<c09d0474>] (tcp_v6_conn_request) from [<c0948870>] (tcp_rcv_state_process+0x29c/0xf44)
[  160.873240]  r5:c891ab40 r4:c8d40000
[  160.876794] [<c09485d4>] (tcp_rcv_state_process) from [<c0953374>] (tcp_v4_do_rcv+0x110/0x22c)
[  160.885398]  r10:c892902a r9:c892903e r8:c10833c0 r7:00000000 r6:c8d40000 r5:c891ab40
[  160.893201]  r4:c8d40000
[  160.895716] [<c0953264>] (tcp_v4_do_rcv) from [<c0955158>] (tcp_v4_rcv+0xb08/0xba8)
[  160.903362]  r7:00000000 r6:c8d40000 r5:00000000 r4:c891ab40
[  160.908998] [<c0954650>] (tcp_v4_rcv) from [<c09273f4>] (ip_protocol_deliver_rcu+0x34/0x1d8)
[  160.917428]  r10:c10833c0 r9:c891ab40 r8:00000000 r7:c10833c0 r6:c0f06f64 r5:c891ab40
[  160.925230]  r4:00000006
[  160.927746] [<c09273c0>] (ip_protocol_deliver_rcu) from [<c0927640>] (ip_local_deliver+0xa8/0xf0)
[  160.936609]  r8:c0f01c0c r7:c0f01c0c r6:00000000 r5:c0f01bc4 r4:c891ab40
[  160.943283] [<c0927598>] (ip_local_deliver) from [<c0926794>] (ip_sublist_rcv_finish+0x44/0x58)
[  160.951967]  r5:c0f01bc4 r4:c0f01bc4
[  160.955522] [<c0926750>] (ip_sublist_rcv_finish) from [<c0926d58>] (ip_list_rcv_finish.constprop.0+0x10c/0x134)
[  160.965598]  r7:c0f01c0c r6:c0f01bc4 r5:c0f01c0c r4:c0f01c0c
[  160.971232] [<c0926c4c>] (ip_list_rcv_finish.constprop.0) from [<c09277cc>] (ip_list_rcv+0xec/0x114)
[  160.980358]  r10:c0f01c5c r9:c10833c0 r8:c0f01c0c r7:c0f01c5c r6:c10833c0 r5:c196b000
[  160.988160]  r4:c891ab40
[  160.990676] [<c09276e0>] (ip_list_rcv) from [<c08a6328>] (__netif_receive_skb_list_core+0x90/0x214)
[  160.999714]  r10:c196b000 r9:c196b000 r8:00000000 r7:c0f01c5c r6:c0f07450 r5:c891acc0
[  161.007517]  r4:c09276e0
[  161.010032] [<c08a6298>] (__netif_receive_skb_list_core) from [<c08a6670>] (netif_receive_skb_list_internal+0x1c4/0x2c4)
[  161.020893]  r10:c10833b8 r9:c0f08a00 r8:00000000 r7:c0f01d44 r6:c0f01d44 r5:c0f01d44
[  161.028696]  r4:00000000
[  161.031211] [<c08a64ac>] (netif_receive_skb_list_internal) from [<c08a67a8>] (netif_receive_skb_list+0x38/0xe4)
[  161.041291]  r10:c1963de0 r9:f08527d0 r8:f08527c0 r7:00000003 r6:c1960040 r5:c103d920
[  161.049094]  r4:c0f01d44
[  161.051609] [<c08a6770>] (netif_receive_skb_list) from [<c0795af4>] (lan966x_fdma_napi_poll+0x3ec/0x490)
[  161.061081]  r9:f08527d0 r8:f08527c0 r7:00000003 r6:c1960040 r5:c1962de0 r4:00000040
[  161.068796] [<c0795708>] (lan966x_fdma_napi_poll) from [<c08a6ee4>] (__napi_poll+0x34/0x1dc)
[  161.077226]  r10:ef7e6080 r9:c0f01db4 r8:c0f03d00 r7:c0f01db3 r6:00000040 r5:c1963de0
[  161.085029]  r4:00000001
[  161.087545] [<c08a6eb0>] (__napi_poll) from [<c08a7290>] (net_rx_action+0xec/0x288)
[  161.095190]  r7:2e957000 r6:ffffc936 r5:0000012c r4:c1963de0
[  161.100825] [<c08a71a4>] (net_rx_action) from [<c010143c>] (__do_softirq+0x13c/0x384)
[  161.108647]  r10:c0f08a00 r9:00000008 r8:00000100 r7:c1037d40 r6:00000003 r5:00000004
[  161.116451]  r4:c0f0308c
[  161.118966] [<c0101300>] (__do_softirq) from [<c013bc9c>] (irq_exit+0xac/0xdc)
[  161.126182]  r10:c0f04f78 r9:c0f08a00 r8:00000000 r7:c0f01ed4 r6:00000000 r5:c0f01ea0
[  161.133985]  r4:c0f08a00
[  161.136500] [<c013bbf0>] (irq_exit) from [<c0a62c84>] (generic_handle_arch_irq+0x48/0x4c)
[  161.144663]  r5:c0f01ea0 r4:c0e8d358
[  161.148218] [<c0a62c3c>] (generic_handle_arch_irq) from [<c0100be8>] (__irq_svc+0x88/0xb0)

Re: [PATCH net-next 0/5] net: lan966x: Add support for FDMA

[Horatiu Vultur]

The 03/18/2022 12:07, Michael Walle wrote:
> 
> Hi Horatiu,

Hi Michael,

> 
> > Currently when injecting or extracting a frame from CPU, the frame
> > is given to the HW each word at a time. There is another way to
> > inject/extract frames from CPU using FDMA(Frame Direct Memory Access).
> > In this way the entire frame is given to the HW. This improves both
> > RX and TX bitrate.
> 
> I wanted to test this. ping and such works fine and I'm also
> seeing fdma interrupts. 

Thanks for testing this also on your board.

> But as soon as I try iperf3 I get a skb_panic
> (due to frame size?). Hope that splash below helps.

I have not seen this issue. But it looks like it is a problem that there
is no more space to add the FCS.
Can you tell me how you run iperf3 so I can also try it?

Also I have a small diff that might fix the issue:
---
--- a/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
+++ b/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
@@ -534,6 +534,8 @@ int lan966x_fdma_xmit(struct sk_buff *skb, __be32 *ifh, struct net_device *dev)
        struct lan966x_tx_dcb *next_dcb, *dcb;
        struct lan966x_tx *tx = &lan966x->tx;
        struct lan966x_db *next_db;
+       int needed_headroom;
+       int needed_tailroom;
        dma_addr_t dma_addr;
        int next_to_use;
        int err;
@@ -554,10 +556,11 @@ int lan966x_fdma_xmit(struct sk_buff *skb, __be32 *ifh, struct net_device *dev)

        /* skb processing */
        skb_tx_timestamp(skb);
-       if (skb_headroom(skb) < IFH_LEN * sizeof(u32)) {
-               err = pskb_expand_head(skb,
-                                      IFH_LEN * sizeof(u32) - skb_headroom(skb),
-                                      0, GFP_ATOMIC);
+       needed_headroom = max_t(int, IFH_LEN * sizeof(u32) - skb_headroom(skb), 0);
+       needed_tailroom = max_t(int, ETH_FCS_LEN - skb_tailroom(skb), 0);
+       if (needed_headroom || needed_tailroom) {
+               err = pskb_expand_head(skb, needed_headroom, needed_tailroom,
+                                      GFP_ATOMIC);
                if (unlikely(err)) {
                        dev->stats.tx_dropped++;
                        err = NETDEV_TX_OK;
---

> 
> -michael
> 
> [  159.778850] skbuff: skb_over_panic: text:c07960c0 len:106 put:4 head:c8cc6e00 data:c8cc6e9a tail:0xc8cc6f04 end:0xc8cc6f00 dev:eth0
> [  159.788067] ------------[ cut here ]------------
> [  159.792575] kernel BUG at net/core/skbuff.c:113!
> ..
> [  160.620995] Backtrace:
> [  160.623426] [<c0a62350>] (skb_panic) from [<c08876f8>] (skb_put+0x54/0x58)
> [  160.630284] [<c08876a4>] (skb_put) from [<c07960c0>] (lan966x_fdma_xmit+0x108/0x504)
> [  160.638011]  r5:00000000 r4:00000000
> [  160.641566] [<c0795fb8>] (lan966x_fdma_xmit) from [<c078cdf8>] (lan966x_port_xmit+0x1b0/0x450)
> [  160.650171]  r10:c1960160 r9:c1960040 r8:00000001 r7:0000001c r6:c196b000 r5:c891a780
> [  160.657973]  r4:00000000
> [  160.660488] [<c078cc48>] (lan966x_port_xmit) from [<c08a3e24>] (dev_hard_start_xmit+0x114/0x248)
> [  160.669267]  r10:c0f016e8 r9:c0f060cc r8:c0e8d350 r7:c196ba00 r6:00000000 r5:c196b000
> [  160.677069]  r4:c891a780
> [  160.679584] [<c08a3d10>] (dev_hard_start_xmit) from [<c0904728>] (sch_direct_xmit+0x11c/0x31c)
> [  160.688188]  r10:2a01000a r9:00000000 r8:00000000 r7:c196b000 r6:c196ba00 r5:c891a780
> [  160.695991]  r4:c1e57400
> [  160.698507] [<c090460c>] (sch_direct_xmit) from [<c08a4520>] (__dev_queue_xmit+0x508/0xaac)
> [  160.706850]  r9:c8cc6f00 r8:0000004a r7:00000001 r6:00000000 r5:c1e57400 r4:c891a780
> [  160.714565] [<c08a4018>] (__dev_queue_xmit) from [<c08a4ad8>] (dev_queue_xmit+0x14/0x18)
> [  160.722648]  r10:2a01000a r9:00000010 r8:0000004a r7:c1e7038c r6:00000000 r5:c891a780
> [  160.730451]  r4:c1e70300
> [  160.732966] [<c08a4ac4>] (dev_queue_xmit) from [<c092b0c0>] (ip_finish_output2+0x270/0x5dc)
> [  160.741300] [<c092ae50>] (ip_finish_output2) from [<c092ceb8>] (__ip_finish_output+0x9c/0x144)
> [  160.749903]  r9:c8cdef00 r8:c10833c0 r7:000005dc r6:c8924000 r5:c891a780 r4:c891a780
> [  160.757618] [<c092ce1c>] (__ip_finish_output) from [<c092d01c>] (ip_output+0xbc/0xc4)
> [  160.765441]  r10:2a01000a r9:c8cdef00 r8:c8cc6e00 r7:c196b000 r6:c8924000 r5:c10833c0
> [  160.773244]  r4:c891a780
> [  160.775759] [<c092cf60>] (ip_output) from [<c092ab10>] (ip_local_out+0x60/0x6c)
> [  160.783058]  r7:00000000 r6:c8924000 r5:c10833c0 r4:c891a780
> [  160.788692] [<c092aab0>] (ip_local_out) from [<c092ad74>] (ip_build_and_send_pkt+0x110/0x1ec)
> [  160.797206]  r7:00000000 r6:c8d40000 r5:000000c4 r4:c891a780
> [  160.802841] [<c092ac64>] (ip_build_and_send_pkt) from [<c0952bb8>] (tcp_v4_send_synack+0xec/0x138)
> [  160.811792]  r10:c8cdef00 r9:c0b9b614 r8:c10833c0 r7:8b01000a r6:2a01000a r5:c8d40000
> [  160.819595]  r4:c8924000
> [  160.822110] [<c0952acc>] (tcp_v4_send_synack) from [<c0941520>] (tcp_conn_request+0x6cc/0x984)
> [  160.830711]  r7:c891ab40 r6:c0952acc r5:c8d40000 r4:c8924000
> [  160.836345] [<c0940e54>] (tcp_conn_request) from [<c095299c>] (tcp_v4_conn_request+0x38/0x84)
> [  160.844863]  r10:c892902a r9:c892903e r8:c10833c0 r7:00000000 r6:00000000 r5:c891ab40
> [  160.852666]  r4:c8d40000
> [  160.855181] [<c0952964>] (tcp_v4_conn_request) from [<c09d05a0>] (tcp_v6_conn_request+0x12c/0x138)
> [  160.864122] [<c09d0474>] (tcp_v6_conn_request) from [<c0948870>] (tcp_rcv_state_process+0x29c/0xf44)
> [  160.873240]  r5:c891ab40 r4:c8d40000
> [  160.876794] [<c09485d4>] (tcp_rcv_state_process) from [<c0953374>] (tcp_v4_do_rcv+0x110/0x22c)
> [  160.885398]  r10:c892902a r9:c892903e r8:c10833c0 r7:00000000 r6:c8d40000 r5:c891ab40
> [  160.893201]  r4:c8d40000
> [  160.895716] [<c0953264>] (tcp_v4_do_rcv) from [<c0955158>] (tcp_v4_rcv+0xb08/0xba8)
> [  160.903362]  r7:00000000 r6:c8d40000 r5:00000000 r4:c891ab40
> [  160.908998] [<c0954650>] (tcp_v4_rcv) from [<c09273f4>] (ip_protocol_deliver_rcu+0x34/0x1d8)
> [  160.917428]  r10:c10833c0 r9:c891ab40 r8:00000000 r7:c10833c0 r6:c0f06f64 r5:c891ab40
> [  160.925230]  r4:00000006
> [  160.927746] [<c09273c0>] (ip_protocol_deliver_rcu) from [<c0927640>] (ip_local_deliver+0xa8/0xf0)
> [  160.936609]  r8:c0f01c0c r7:c0f01c0c r6:00000000 r5:c0f01bc4 r4:c891ab40
> [  160.943283] [<c0927598>] (ip_local_deliver) from [<c0926794>] (ip_sublist_rcv_finish+0x44/0x58)
> [  160.951967]  r5:c0f01bc4 r4:c0f01bc4
> [  160.955522] [<c0926750>] (ip_sublist_rcv_finish) from [<c0926d58>] (ip_list_rcv_finish.constprop.0+0x10c/0x134)
> [  160.965598]  r7:c0f01c0c r6:c0f01bc4 r5:c0f01c0c r4:c0f01c0c
> [  160.971232] [<c0926c4c>] (ip_list_rcv_finish.constprop.0) from [<c09277cc>] (ip_list_rcv+0xec/0x114)
> [  160.980358]  r10:c0f01c5c r9:c10833c0 r8:c0f01c0c r7:c0f01c5c r6:c10833c0 r5:c196b000
> [  160.988160]  r4:c891ab40
> [  160.990676] [<c09276e0>] (ip_list_rcv) from [<c08a6328>] (__netif_receive_skb_list_core+0x90/0x214)
> [  160.999714]  r10:c196b000 r9:c196b000 r8:00000000 r7:c0f01c5c r6:c0f07450 r5:c891acc0
> [  161.007517]  r4:c09276e0
> [  161.010032] [<c08a6298>] (__netif_receive_skb_list_core) from [<c08a6670>] (netif_receive_skb_list_internal+0x1c4/0x2c4)
> [  161.020893]  r10:c10833b8 r9:c0f08a00 r8:00000000 r7:c0f01d44 r6:c0f01d44 r5:c0f01d44
> [  161.028696]  r4:00000000
> [  161.031211] [<c08a64ac>] (netif_receive_skb_list_internal) from [<c08a67a8>] (netif_receive_skb_list+0x38/0xe4)
> [  161.041291]  r10:c1963de0 r9:f08527d0 r8:f08527c0 r7:00000003 r6:c1960040 r5:c103d920
> [  161.049094]  r4:c0f01d44
> [  161.051609] [<c08a6770>] (netif_receive_skb_list) from [<c0795af4>] (lan966x_fdma_napi_poll+0x3ec/0x490)
> [  161.061081]  r9:f08527d0 r8:f08527c0 r7:00000003 r6:c1960040 r5:c1962de0 r4:00000040
> [  161.068796] [<c0795708>] (lan966x_fdma_napi_poll) from [<c08a6ee4>] (__napi_poll+0x34/0x1dc)
> [  161.077226]  r10:ef7e6080 r9:c0f01db4 r8:c0f03d00 r7:c0f01db3 r6:00000040 r5:c1963de0
> [  161.085029]  r4:00000001
> [  161.087545] [<c08a6eb0>] (__napi_poll) from [<c08a7290>] (net_rx_action+0xec/0x288)
> [  161.095190]  r7:2e957000 r6:ffffc936 r5:0000012c r4:c1963de0
> [  161.100825] [<c08a71a4>] (net_rx_action) from [<c010143c>] (__do_softirq+0x13c/0x384)
> [  161.108647]  r10:c0f08a00 r9:00000008 r8:00000100 r7:c1037d40 r6:00000003 r5:00000004
> [  161.116451]  r4:c0f0308c
> [  161.118966] [<c0101300>] (__do_softirq) from [<c013bc9c>] (irq_exit+0xac/0xdc)
> [  161.126182]  r10:c0f04f78 r9:c0f08a00 r8:00000000 r7:c0f01ed4 r6:00000000 r5:c0f01ea0
> [  161.133985]  r4:c0f08a00
> [  161.136500] [<c013bbf0>] (irq_exit) from [<c0a62c84>] (generic_handle_arch_irq+0x48/0x4c)
> [  161.144663]  r5:c0f01ea0 r4:c0e8d358
> [  161.148218] [<c0a62c3c>] (generic_handle_arch_irq) from [<c0100be8>] (__irq_svc+0x88/0xb0)

-- 
/Horatiu

Re: [PATCH net-next 0/5] net: lan966x: Add support for FDMA

[Michael Walle]

Am 2022-03-18 13:10, schrieb Horatiu Vultur:
> The 03/18/2022 12:07, Michael Walle wrote:
>> > Currently when injecting or extracting a frame from CPU, the frame
>> > is given to the HW each word at a time. There is another way to
>> > inject/extract frames from CPU using FDMA(Frame Direct Memory Access).
>> > In this way the entire frame is given to the HW. This improves both
>> > RX and TX bitrate.
>> 
>> I wanted to test this. ping and such works fine and I'm also
>> seeing fdma interrupts.
> 
> Thanks for testing this also on your board.
> 
>> But as soon as I try iperf3 I get a skb_panic
>> (due to frame size?). Hope that splash below helps.
> 
> I have not seen this issue. But it looks like it is a problem that 
> there
> is no more space to add the FCS.
> Can you tell me how you run iperf3 so I can also try it?

oh, I forgot to include the commandline.

# on the remote computer
$ iperf3 --version
iperf 3.6 (cJSON 1.5.2)
Linux eddie01 4.19.0-18-686-pae #1 SMP Debian 4.19.208-1 (2021-09-29) 
i686
Optional features available: CPU affinity setting, IPv6 flow label, 
SCTP, TCP congestion algorithm setting, sendfile / zerocopy, socket 
pacing, authentication
$ iperf3 -s

# on the board
$ iperf3 --version
iperf 3.10.1 (cJSON 1.7.13)
Linux buildroot 5.17.0-rc8-next-20220316-00058-gc6cb0628f2a6-dirty #385 
SMP Fri Mar 18 13:34:26 CET 2022 armv7l
Optional features available: CPU affinity setting, IPv6 flow label, TCP 
congestion algorithm setting, sendfile / zerocopy, socket pacing, bind 
to device, support IPv4 don't fragment
$ iperf3 -c eddie01

> Also I have a small diff that might fix the issue:
> ---
> --- a/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
> +++ b/drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c
> @@ -534,6 +534,8 @@ int lan966x_fdma_xmit(struct sk_buff *skb, __be32
> *ifh, struct net_device *dev)
>         struct lan966x_tx_dcb *next_dcb, *dcb;
>         struct lan966x_tx *tx = &lan966x->tx;
>         struct lan966x_db *next_db;
> +       int needed_headroom;
> +       int needed_tailroom;
>         dma_addr_t dma_addr;
>         int next_to_use;
>         int err;
> @@ -554,10 +556,11 @@ int lan966x_fdma_xmit(struct sk_buff *skb,
> __be32 *ifh, struct net_device *dev)
> 
>         /* skb processing */
>         skb_tx_timestamp(skb);

btw. skb_tx_timestamp() should be as close to the handover
of the frame to the hardware as possible, no?

> -       if (skb_headroom(skb) < IFH_LEN * sizeof(u32)) {
> -               err = pskb_expand_head(skb,
> -                                      IFH_LEN * sizeof(u32) -
> skb_headroom(skb),
> -                                      0, GFP_ATOMIC);
> +       needed_headroom = max_t(int, IFH_LEN * sizeof(u32) -
> skb_headroom(skb), 0);
> +       needed_tailroom = max_t(int, ETH_FCS_LEN - skb_tailroom(skb), 
> 0);
> +       if (needed_headroom || needed_tailroom) {
> +               err = pskb_expand_head(skb, needed_headroom, 
> needed_tailroom,
> +                                      GFP_ATOMIC);
>                 if (unlikely(err)) {
>                         dev->stats.tx_dropped++;
>                         err = NETDEV_TX_OK;

Indeed this will fix the issue:

# iperf3 -c eddie01
Connecting to host eddie01, port 5201
[  5] local 10.0.1.143 port 55342 connected to 10.0.1.42 port 5201
[ ID] Interval           Transfer     Bitrate         Retr  Cwnd
[  5]   0.00-1.01   sec  43.8 MBytes   364 Mbits/sec    0    245 KBytes
[  5]   1.01-2.02   sec  43.8 MBytes   364 Mbits/sec    0    246 KBytes
[  5]   2.02-3.03   sec  43.8 MBytes   364 Mbits/sec    0    259 KBytes

# iperf3 -R -c eddie01
Connecting to host eddie01, port 5201
Reverse mode, remote host eddie01 is sending
[  5] local 10.0.1.143 port 55346 connected to 10.0.1.42 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-1.00   sec  28.6 MBytes   240 Mbits/sec
[  5]   1.00-2.00   sec  28.9 MBytes   242 Mbits/sec
[  5]   2.00-3.00   sec  28.7 MBytes   241 Mbits/sec

-michael