net: aquantia: implement data PTP datapath

Here we do alloc/free IRQs for PTP rings.
We also implement processing of PTP packets on TX and RX sides.
Signed-off-by: Egor Pomozov <epomozov@marvell.com>
Co-developed-by: Sergey Samoilenko <sergey.samoilenko@aquantia.com>
Signed-off-by: Sergey Samoilenko <sergey.samoilenko@aquantia.com>
Co-developed-by: Dmitry Bezrukov <dmitry.bezrukov@aquantia.com>
Signed-off-by: Dmitry Bezrukov <dmitry.bezrukov@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/aquantia/atlantic/aq_cfg.h

 /* SPDX-License-Identifier: GPL-2.0-only */
 /*
  * aQuantia Corporation Network Driver
- * Copyright (C) 2014-2017 aQuantia Corporation. All rights reserved
+ * Copyright (C) 2014-2019 aQuantia Corporation. All rights reserved
  */
 
 /* File aq_cfg.h: Definition of configuration parameters and constants. */
@@ -27,7 +27,7 @@
 
 #define AQ_CFG_INTERRUPT_MODERATION_USEC_MAX (0x1FF * 2)
 
-#define AQ_CFG_IRQ_MASK                      0x1FFU
+#define AQ_CFG_IRQ_MASK                      0x3FFU
 
 #define AQ_CFG_VECS_MAX   8U
 #define AQ_CFG_TCS_MAX    8U

drivers/net/ethernet/aquantia/atlantic/aq_hw.h

@@ -244,6 +244,12 @@ struct aq_hw_ops {
 
 	int (*hw_rx_tc_mode_get)(struct aq_hw_s *self, u32 *tc_mode);
 
+	int (*hw_ring_hwts_rx_fill)(struct aq_hw_s *self,
+				    struct aq_ring_s *aq_ring);
+
+	int (*hw_ring_hwts_rx_receive)(struct aq_hw_s *self,
+				       struct aq_ring_s *ring);
+
 	void (*hw_get_ptp_ts)(struct aq_hw_s *self, u64 *stamp);
 
 	int (*hw_adj_clock_freq)(struct aq_hw_s *self, s32 delta);
@@ -252,6 +258,12 @@ struct aq_hw_ops {
 
 	int (*hw_set_sys_clock)(struct aq_hw_s *self, u64 time, u64 ts);
 
+	u16 (*rx_extract_ts)(struct aq_hw_s *self, u8 *p, unsigned int len,
+			     u64 *timestamp);
+
+	int (*extract_hwts)(struct aq_hw_s *self, u8 *p, unsigned int len,
+			    u64 *timestamp);
+
 	int (*hw_set_fc)(struct aq_hw_s *self, u32 fc, u32 tc);
 };
 

drivers/net/ethernet/aquantia/atlantic/aq_main.c

 // SPDX-License-Identifier: GPL-2.0-only
 /*
  * aQuantia Corporation Network Driver
- * Copyright (C) 2014-2017 aQuantia Corporation. All rights reserved
+ * Copyright (C) 2014-2019 aQuantia Corporation. All rights reserved
  */
 
 /* File aq_main.c: Main file for aQuantia Linux driver. */
@@ -10,10 +10,13 @@
 #include "aq_nic.h"
 #include "aq_pci_func.h"
 #include "aq_ethtool.h"
+#include "aq_ptp.h"
 #include "aq_filters.h"
 
 #include <linux/netdevice.h>
 #include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/udp.h>
 
 MODULE_LICENSE("GPL v2");
 MODULE_VERSION(AQ_CFG_DRV_VERSION);
@@ -93,6 +96,24 @@ static int aq_ndev_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 {
 	struct aq_nic_s *aq_nic = netdev_priv(ndev);
 
+	if (unlikely(aq_utils_obj_test(&aq_nic->flags, AQ_NIC_PTP_DPATH_UP))) {
+		/* Hardware adds the Timestamp for PTPv2 802.AS1
+		 * and PTPv2 IPv4 UDP.
+		 * We have to push even general 320 port messages to the ptp
+		 * queue explicitly. This is a limitation of current firmware
+		 * and hardware PTP design of the chip. Otherwise ptp stream
+		 * will fail to sync
+		 */
+		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) ||
+		    unlikely((ip_hdr(skb)->version == 4) &&
+			     (ip_hdr(skb)->protocol == IPPROTO_UDP) &&
+			     ((udp_hdr(skb)->dest == htons(319)) ||
+			      (udp_hdr(skb)->dest == htons(320)))) ||
+		    unlikely(eth_hdr(skb)->h_proto == htons(ETH_P_1588)))
+			return aq_ptp_xmit(aq_nic, skb);
+	}
+
+	skb_tx_timestamp(skb);
 	return aq_nic_xmit(aq_nic, skb);
 }
 

drivers/net/ethernet/aquantia/atlantic/aq_nic.c

@@ -146,8 +146,11 @@ static int aq_nic_update_link_status(struct aq_nic_s *self)
 			self->aq_hw->aq_link_status.mbps);
 		aq_nic_update_interrupt_moderation_settings(self);
 
-		if (self->aq_ptp)
+		if (self->aq_ptp) {
 			aq_ptp_clock_init(self);
+			aq_ptp_tm_offset_set(self,
+					     self->aq_hw->aq_link_status.mbps);
+		}
 
 		/* Driver has to update flow control settings on RX block
 		 * on any link event.
@@ -196,6 +199,8 @@ static void aq_nic_service_task(struct work_struct *work)
 					     service_task);
 	int err;
 
+	aq_ptp_service_task(self);
+
 	if (aq_utils_obj_test(&self->flags, AQ_NIC_FLAGS_IS_NOT_READY))
 		return;
 
@@ -408,6 +413,10 @@ int aq_nic_start(struct aq_nic_s *self)
 				goto err_exit;
 		}
 
+		err = aq_ptp_irq_alloc(self);
+		if (err < 0)
+			goto err_exit;
+
 		if (self->aq_nic_cfg.link_irq_vec) {
 			int irqvec = pci_irq_vector(self->pdev,
 						   self->aq_nic_cfg.link_irq_vec);
@@ -440,8 +449,7 @@ int aq_nic_start(struct aq_nic_s *self)
 	return err;
 }
 
-static unsigned int aq_nic_map_skb(struct aq_nic_s *self,
-				   struct sk_buff *skb,
+unsigned int aq_nic_map_skb(struct aq_nic_s *self, struct sk_buff *skb,
 			    struct aq_ring_s *ring)
 {
 	unsigned int ret = 0U;
@@ -973,6 +981,8 @@ int aq_nic_stop(struct aq_nic_s *self)
 	else
 		aq_pci_func_free_irqs(self);
 
+	aq_ptp_irq_free(self);
+
 	for (i = 0U, aq_vec = self->aq_vec[0];
 		self->aq_vecs > i; ++i, aq_vec = self->aq_vec[i])
 		aq_vec_stop(aq_vec);

drivers/net/ethernet/aquantia/atlantic/aq_nic.h

@@ -54,6 +54,7 @@ struct aq_nic_cfg_s {
 #define AQ_NIC_FLAG_STOPPING    0x00000008U
 #define AQ_NIC_FLAG_RESETTING   0x00000010U
 #define AQ_NIC_FLAG_CLOSING     0x00000020U
+#define AQ_NIC_PTP_DPATH_UP     0x02000000U
 #define AQ_NIC_LINK_DOWN        0x04000000U
 #define AQ_NIC_FLAG_ERR_UNPLUG  0x40000000U
 #define AQ_NIC_FLAG_ERR_HW      0x80000000U
@@ -129,6 +130,8 @@ void aq_nic_cfg_start(struct aq_nic_s *self);
 int aq_nic_ndev_register(struct aq_nic_s *self);
 void aq_nic_ndev_free(struct aq_nic_s *self);
 int aq_nic_start(struct aq_nic_s *self);
+unsigned int aq_nic_map_skb(struct aq_nic_s *self, struct sk_buff *skb,
+			    struct aq_ring_s *ring);
 int aq_nic_xmit(struct aq_nic_s *self, struct sk_buff *skb);
 int aq_nic_get_regs(struct aq_nic_s *self, struct ethtool_regs *regs, void *p);
 int aq_nic_get_regs_count(struct aq_nic_s *self);

drivers/net/ethernet/aquantia/atlantic/aq_pci_func.c

 // SPDX-License-Identifier: GPL-2.0-only
 /*
  * aQuantia Corporation Network Driver
- * Copyright (C) 2014-2017 aQuantia Corporation. All rights reserved
+ * Copyright (C) 2014-2019 aQuantia Corporation. All rights reserved
  */
 
 /* File aq_pci_func.c: Definition of PCI functions. */
@@ -269,6 +269,9 @@ static int aq_pci_probe(struct pci_dev *pdev,
 	numvecs = min((u8)AQ_CFG_VECS_DEF,
 		      aq_nic_get_cfg(self)->aq_hw_caps->msix_irqs);
 	numvecs = min(numvecs, num_online_cpus());
+	/* Request IRQ vector for PTP */
+	numvecs += 1;
+
 	numvecs += AQ_HW_SERVICE_IRQS;
 	/*enable interrupts */
 #if !AQ_CFG_FORCE_LEGACY_INT

drivers/net/ethernet/aquantia/atlantic/aq_ptp.h

@@ -17,6 +17,9 @@ int aq_ptp_init(struct aq_nic_s *aq_nic, unsigned int idx_vec);
 void aq_ptp_unregister(struct aq_nic_s *aq_nic);
 void aq_ptp_free(struct aq_nic_s *aq_nic);
 
+int aq_ptp_irq_alloc(struct aq_nic_s *aq_nic);
+void aq_ptp_irq_free(struct aq_nic_s *aq_nic);
+
 int aq_ptp_ring_alloc(struct aq_nic_s *aq_nic);
 void aq_ptp_ring_free(struct aq_nic_s *aq_nic);
 
@@ -25,6 +28,22 @@ int aq_ptp_ring_start(struct aq_nic_s *aq_nic);
 void aq_ptp_ring_stop(struct aq_nic_s *aq_nic);
 void aq_ptp_ring_deinit(struct aq_nic_s *aq_nic);
 
+void aq_ptp_service_task(struct aq_nic_s *aq_nic);
+
+void aq_ptp_tm_offset_set(struct aq_nic_s *aq_nic, unsigned int mbps);
+
 void aq_ptp_clock_init(struct aq_nic_s *aq_nic);
 
+/* Traffic processing functions */
+int aq_ptp_xmit(struct aq_nic_s *aq_nic, struct sk_buff *skb);
+void aq_ptp_tx_hwtstamp(struct aq_nic_s *aq_nic, u64 timestamp);
+
+/* Return either ring is belong to PTP or not*/
+bool aq_ptp_ring(struct aq_nic_s *aq_nic, struct aq_ring_s *ring);
+
+u16 aq_ptp_extract_ts(struct aq_nic_s *aq_nic, struct sk_buff *skb, u8 *p,
+		      unsigned int len);
+
+struct ptp_clock *aq_ptp_get_ptp_clock(struct aq_ptp_s *aq_ptp);
+
 #endif /* AQ_PTP_H */

drivers/net/ethernet/aquantia/atlantic/aq_ring.c

@@ -10,6 +10,7 @@
 #include "aq_nic.h"
 #include "aq_hw.h"
 #include "aq_hw_utils.h"
+#include "aq_ptp.h"
 
 #include <linux/netdevice.h>
 #include <linux/etherdevice.h>
@@ -314,6 +315,7 @@ int aq_ring_rx_clean(struct aq_ring_s *self,
 		self->sw_head = aq_ring_next_dx(self, self->sw_head),
 		--budget, ++(*work_done)) {
 		struct aq_ring_buff_s *buff = &self->buff_ring[self->sw_head];
+		bool is_ptp_ring = aq_ptp_ring(self->aq_nic, self);
 		struct aq_ring_buff_s *buff_ = NULL;
 		struct sk_buff *skb = NULL;
 		unsigned int next_ = 0U;
@@ -378,6 +380,11 @@ int aq_ring_rx_clean(struct aq_ring_s *self,
 				err = -ENOMEM;
 				goto err_exit;
 			}
+			if (is_ptp_ring)
+				buff->len -=
+					aq_ptp_extract_ts(self->aq_nic, skb,
+						aq_buf_vaddr(&buff->rxdata),
+						buff->len);
 			skb_put(skb, buff->len);
 			page_ref_inc(buff->rxdata.page);
 		} else {
@@ -386,6 +393,11 @@ int aq_ring_rx_clean(struct aq_ring_s *self,
 				err = -ENOMEM;
 				goto err_exit;
 			}
+			if (is_ptp_ring)
+				buff->len -=
+					aq_ptp_extract_ts(self->aq_nic, skb,
+						aq_buf_vaddr(&buff->rxdata),
+						buff->len);
 
 			hdr_len = buff->len;
 			if (hdr_len > AQ_CFG_RX_HDR_SIZE)
@@ -445,8 +457,8 @@ int aq_ring_rx_clean(struct aq_ring_s *self,
 		skb_set_hash(skb, buff->rss_hash,
 			     buff->is_hash_l4 ? PKT_HASH_TYPE_L4 :
 			     PKT_HASH_TYPE_NONE);
-
-		skb_record_rx_queue(skb, self->idx);
+		/* Send all PTP traffic to 0 queue */
+		skb_record_rx_queue(skb, is_ptp_ring ? 0 : self->idx);
 
 		++self->stats.rx.packets;
 		self->stats.rx.bytes += skb->len;
@@ -458,6 +470,21 @@ int aq_ring_rx_clean(struct aq_ring_s *self,
 	return err;
 }
 
+void aq_ring_hwts_rx_clean(struct aq_ring_s *self, struct aq_nic_s *aq_nic)
+{
+	while (self->sw_head != self->hw_head) {
+		u64 ns;
+
+		aq_nic->aq_hw_ops->extract_hwts(aq_nic->aq_hw,
+						self->dx_ring +
+						(self->sw_head * self->dx_size),
+						self->dx_size, &ns);
+		aq_ptp_tx_hwtstamp(aq_nic, ns);
+
+		self->sw_head = aq_ring_next_dx(self, self->sw_head);
+	}
+}
+
 int aq_ring_rx_fill(struct aq_ring_s *self)
 {
 	unsigned int page_order = self->page_order;

drivers/net/ethernet/aquantia/atlantic/aq_ring.h

@@ -177,5 +177,6 @@ int aq_ring_rx_fill(struct aq_ring_s *self);
 struct aq_ring_s *aq_ring_hwts_rx_alloc(struct aq_ring_s *self,
 		struct aq_nic_s *aq_nic, unsigned int idx,
 		unsigned int size, unsigned int dx_size);
+void aq_ring_hwts_rx_clean(struct aq_ring_s *self, struct aq_nic_s *aq_nic);
 
 #endif /* AQ_RING_H */

drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c

@@ -682,6 +682,46 @@ static int hw_atl_b0_hw_ring_rx_fill(struct aq_hw_s *self,
 	return aq_hw_err_from_flags(self);
 }
 
+static int hw_atl_b0_hw_ring_hwts_rx_fill(struct aq_hw_s *self,
+					  struct aq_ring_s *ring)
+{
+	unsigned int i;
+
+	for (i = aq_ring_avail_dx(ring); i--;
+			ring->sw_tail = aq_ring_next_dx(ring, ring->sw_tail)) {
+		struct hw_atl_rxd_s *rxd =
+			(struct hw_atl_rxd_s *)
+			&ring->dx_ring[ring->sw_tail * HW_ATL_B0_RXD_SIZE];
+
+		rxd->buf_addr = ring->dx_ring_pa + ring->size * ring->dx_size;
+		rxd->hdr_addr = 0U;
+	}
+	/* Make sure descriptors are updated before bump tail*/
+	wmb();
+
+	hw_atl_reg_rx_dma_desc_tail_ptr_set(self, ring->sw_tail, ring->idx);
+
+	return aq_hw_err_from_flags(self);
+}
+
+static int hw_atl_b0_hw_ring_hwts_rx_receive(struct aq_hw_s *self,
+					     struct aq_ring_s *ring)
+{
+	while (ring->hw_head != ring->sw_tail) {
+		struct hw_atl_rxd_hwts_wb_s *hwts_wb =
+			(struct hw_atl_rxd_hwts_wb_s *)
+			(ring->dx_ring + (ring->hw_head * HW_ATL_B0_RXD_SIZE));
+
+		/* RxD is not done */
+		if (!(hwts_wb->sec_lw0 & 0x1U))
+			break;
+
+		ring->hw_head = aq_ring_next_dx(ring, ring->hw_head);
+	}
+
+	return aq_hw_err_from_flags(self);
+}
+
 static int hw_atl_b0_hw_ring_tx_head_update(struct aq_hw_s *self,
 					    struct aq_ring_s *ring)
 {
@@ -1133,6 +1173,61 @@ static int hw_atl_b0_adj_clock_freq(struct aq_hw_s *self, s32 ppb)
 	return self->aq_fw_ops->send_fw_request(self, &fwreq, size);
 }
 
+static u16 hw_atl_b0_rx_extract_ts(struct aq_hw_s *self, u8 *p,
+				   unsigned int len, u64 *timestamp)
+{
+	unsigned int offset = 14;
+	struct ethhdr *eth;
+	u64 sec;
+	u8 *ptr;
+	u32 ns;
+
+	if (len <= offset || !timestamp)
+		return 0;
+
+	/* The TIMESTAMP in the end of package has following format:
+	 * (big-endian)
+	 *   struct {
+	 *     uint64_t sec;
+	 *     uint32_t ns;
+	 *     uint16_t stream_id;
+	 *   };
+	 */
+	ptr = p + (len - offset);
+	memcpy(&sec, ptr, sizeof(sec));
+	ptr += sizeof(sec);
+	memcpy(&ns, ptr, sizeof(ns));
+
+	sec = be64_to_cpu(sec) & 0xffffffffffffllu;
+	ns = be32_to_cpu(ns);
+	*timestamp = sec * NSEC_PER_SEC + ns + self->ptp_clk_offset;
+
+	eth = (struct ethhdr *)p;
+
+	return (eth->h_proto == htons(ETH_P_1588)) ? 12 : 14;
+}
+
+static int hw_atl_b0_extract_hwts(struct aq_hw_s *self, u8 *p, unsigned int len,
+				  u64 *timestamp)
+{
+	struct hw_atl_rxd_hwts_wb_s *hwts_wb = (struct hw_atl_rxd_hwts_wb_s *)p;
+	u64 tmp, sec, ns;
+
+	sec = 0;
+	tmp = (hwts_wb->sec_lw0 >> 2) & 0x3ff;
+	sec += tmp;
+	tmp = (u64)((hwts_wb->sec_lw1 >> 16) & 0xffff) << 10;
+	sec += tmp;
+	tmp = (u64)(hwts_wb->sec_hw & 0xfff) << 26;
+	sec += tmp;
+	tmp = (u64)((hwts_wb->sec_hw >> 22) & 0x3ff) << 38;
+	sec += tmp;
+	ns = sec * NSEC_PER_SEC + hwts_wb->ns;
+	if (timestamp)
+		*timestamp = ns + self->ptp_clk_offset;
+	return 0;
+}
+
 static int hw_atl_b0_hw_fl3l4_clear(struct aq_hw_s *self,
 				    struct aq_rx_filter_l3l4 *data)
 {
@@ -1309,11 +1404,16 @@ const struct aq_hw_ops hw_atl_ops_b0 = {
 	.hw_tx_tc_mode_get       = hw_atl_b0_tx_tc_mode_get,
 	.hw_rx_tc_mode_get       = hw_atl_b0_rx_tc_mode_get,
 
+	.hw_ring_hwts_rx_fill        = hw_atl_b0_hw_ring_hwts_rx_fill,
+	.hw_ring_hwts_rx_receive     = hw_atl_b0_hw_ring_hwts_rx_receive,
+
 	.hw_get_ptp_ts           = hw_atl_b0_get_ptp_ts,
 	.hw_adj_sys_clock        = hw_atl_b0_adj_sys_clock,
 	.hw_set_sys_clock        = hw_atl_b0_set_sys_clock,
 	.hw_adj_clock_freq       = hw_atl_b0_adj_clock_freq,
 
+	.rx_extract_ts           = hw_atl_b0_rx_extract_ts,
+	.extract_hwts            = hw_atl_b0_extract_hwts,
 	.hw_set_offload          = hw_atl_b0_hw_offload_set,
 	.hw_set_fc                   = hw_atl_b0_set_fc,
 };