Merge branch 'octeontx2-pf-Add-network-driver-for-physical-function'

Sunil Goutham says:

====================
octeontx2-pf: Add network driver for physical function

OcteonTX2 SOC's resource virtualization unit (RVU) supports
multiple physical and virtual functions. Each of the PF's
functionality is determined by what kind of resources are attached
to it. If NPA and NIX blocks are attached to a PF it can function
as a highly capable network device.

This patch series add a network driver for the PF. Initial set of
patches adds mailbox communication with admin function (RVU AF)
and configuration of queues. Followed by Rx and tx pkts NAPI
handler and then support for HW offloads like RSS, TSO, Rxhash etc.
Ethtool support to extract stats, config RSS, queue sizes, queue
count is also added.

Added documentation to give a high level overview of HW and
different drivers which will be upstreamed and how they interact.

Changes from v5:
   * Fixed otx2_atomic64_add() non ARM64 fallback definition.
     - Suggested by David Miller

Changes from v4:
   * Replaced pci_set_dma_mask and pci_set_consistent_dma_mask
     fn()s with dma_set_mask_and_coherent().
   * Some additonal code cleanup.
   * Fixed receive buffer segmnetation logic in otx2_alloc_rbuf()
   * Removed all unused BIG_ENDIAN structure definitions.
   * Removed unnecessary memory barriers
     - Sugested by Jakub Kicinski
   * Fixed mailbox initalization failure handling
   * Removed unused function parameter in otx2_skb_add_frag()
     - Suggested by Maciej Fijalkowski

Changes from v3:
   * Fixed receive side scaling reinitialization during interface
     DOWN and UP to retain user configured settings, if any.
   * Removed driver version from ethtool.
   * Fixed otx2_set_rss_hash_opts() to return error incase RSS is
     not enabled.
     - Sugested by Jakub Kicinski

Changes from v2:
   * Removed frames, bytes, dropped packet stats from ethtool to avoid
     duplication of same stats in netlink and ethtool.
     - Sugested by Jakub Kicinski
   * Removed number of channels and ringparam upper bound checking
     in ethtool support.
   * Fixed RSS hash option setting to reject unsupported config.
     - Suggested by Michal Kubecek

Changes from v1:
   * Made driver dependent on 64bit, to fix build errors related to
     non availability of writeq/readq APIs for 32bit platforms.
     - Reported by kbuild test robot
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/device_drivers/index.rst

@@ -22,6 +22,7 @@ Contents:
    intel/iavf
    intel/ice
    google/gve
+   marvell/octeontx2
    mellanox/mlx5
    netronome/nfp
    pensando/ionic

Documentation/networking/device_drivers/marvell/octeontx2.rst

+.. SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+
+====================================
+Marvell OcteonTx2 RVU Kernel Drivers
+====================================
+
+Copyright (c) 2020 Marvell International Ltd.
+
+Contents
+========
+
+- `Overview`_
+- `Drivers`_
+- `Basic packet flow`_
+
+Overview
+========
+
+Resource virtualization unit (RVU) on Marvell's OcteonTX2 SOC maps HW
+resources from the network, crypto and other functional blocks into
+PCI-compatible physical and virtual functions. Each functional block
+again has multiple local functions (LFs) for provisioning to PCI devices.
+RVU supports multiple PCIe SRIOV physical functions (PFs) and virtual
+functions (VFs). PF0 is called the administrative / admin function (AF)
+and has privileges to provision RVU functional block's LFs to each of the
+PF/VF.
+
+RVU managed networking functional blocks
+ - Network pool or buffer allocator (NPA)
+ - Network interface controller (NIX)
+ - Network parser CAM (NPC)
+ - Schedule/Synchronize/Order unit (SSO)
+ - Loopback interface (LBK)
+
+RVU managed non-networking functional blocks
+ - Crypto accelerator (CPT)
+ - Scheduled timers unit (TIM)
+ - Schedule/Synchronize/Order unit (SSO)
+   Used for both networking and non networking usecases
+
+Resource provisioning examples
+ - A PF/VF with NIX-LF & NPA-LF resources works as a pure network device
+ - A PF/VF with CPT-LF resource works as a pure crypto offload device.
+
+RVU functional blocks are highly configurable as per software requirements.
+
+Firmware setups following stuff before kernel boots
+ - Enables required number of RVU PFs based on number of physical links.
+ - Number of VFs per PF are either static or configurable at compile time.
+   Based on config, firmware assigns VFs to each of the PFs.
+ - Also assigns MSIX vectors to each of PF and VFs.
+ - These are not changed after kernel boot.
+
+Drivers
+=======
+
+Linux kernel will have multiple drivers registering to different PF and VFs
+of RVU. Wrt networking there will be 3 flavours of drivers.
+
+Admin Function driver
+---------------------
+
+As mentioned above RVU PF0 is called the admin function (AF), this driver
+supports resource provisioning and configuration of functional blocks.
+Doesn't handle any I/O. It sets up few basic stuff but most of the
+funcionality is achieved via configuration requests from PFs and VFs.
+
+PF/VFs communicates with AF via a shared memory region (mailbox). Upon
+receiving requests AF does resource provisioning and other HW configuration.
+AF is always attached to host kernel, but PFs and their VFs may be used by host
+kernel itself, or attached to VMs or to userspace applications like
+DPDK etc. So AF has to handle provisioning/configuration requests sent
+by any device from any domain.
+
+AF driver also interacts with underlying firmware to
+ - Manage physical ethernet links ie CGX LMACs.
+ - Retrieve information like speed, duplex, autoneg etc
+ - Retrieve PHY EEPROM and stats.
+ - Configure FEC, PAM modes
+ - etc
+
+From pure networking side AF driver supports following functionality.
+ - Map a physical link to a RVU PF to which a netdev is registered.
+ - Attach NIX and NPA block LFs to RVU PF/VF which provide buffer pools, RQs, SQs
+   for regular networking functionality.
+ - Flow control (pause frames) enable/disable/config.
+ - HW PTP timestamping related config.
+ - NPC parser profile config, basically how to parse pkt and what info to extract.
+ - NPC extract profile config, what to extract from the pkt to match data in MCAM entries.
+ - Manage NPC MCAM entries, upon request can frame and install requested packet forwarding rules.
+ - Defines receive side scaling (RSS) algorithms.
+ - Defines segmentation offload algorithms (eg TSO)
+ - VLAN stripping, capture and insertion config.
+ - SSO and TIM blocks config which provide packet scheduling support.
+ - Debugfs support, to check current resource provising, current status of
+   NPA pools, NIX RQ, SQ and CQs, various stats etc which helps in debugging issues.
+ - And many more.
+
+Physical Function driver
+------------------------
+
+This RVU PF handles IO, is mapped to a physical ethernet link and this
+driver registers a netdev. This supports SR-IOV. As said above this driver
+communicates with AF with a mailbox. To retrieve information from physical
+links this driver talks to AF and AF gets that info from firmware and responds
+back ie cannot talk to firmware directly.
+
+Supports ethtool for configuring links, RSS, queue count, queue size,
+flow control, ntuple filters, dump PHY EEPROM, config FEC etc.
+
+Virtual Function driver
+-----------------------
+
+There are two types VFs, VFs that share the physical link with their parent
+SR-IOV PF and the VFs which work in pairs using internal HW loopback channels (LBK).
+
+Type1:
+ - These VFs and their parent PF share a physical link and used for outside communication.
+ - VFs cannot communicate with AF directly, they send mbox message to PF and PF
+   forwards that to AF. AF after processing, responds back to PF and PF forwards
+   the reply to VF.
+ - From functionality point of view there is no difference between PF and VF as same type
+   HW resources are attached to both. But user would be able to configure few stuff only
+   from PF as PF is treated as owner/admin of the link.
+
+Type2:
+ - RVU PF0 ie admin function creates these VFs and maps them to loopback block's channels.
+ - A set of two VFs (VF0 & VF1, VF2 & VF3 .. so on) works as a pair ie pkts sent out of
+   VF0 will be received by VF1 and viceversa.
+ - These VFs can be used by applications or virtual machines to communicate between them
+   without sending traffic outside. There is no switch present in HW, hence the support
+   for loopback VFs.
+ - These communicate directly with AF (PF0) via mbox.
+
+Except for the IO channels or links used for packet reception and transmission there is
+no other difference between these VF types. AF driver takes care of IO channel mapping,
+hence same VF driver works for both types of devices.
+
+Basic packet flow
+=================
+
+Ingress
+-------
+
+1. CGX LMAC receives packet.
+2. Forwards the packet to the NIX block.
+3. Then submitted to NPC block for parsing and then MCAM lookup to get the destination RVU device.
+4. NIX LF attached to the destination RVU device allocates a buffer from RQ mapped buffer pool of NPA block LF.
+5. RQ may be selected by RSS or by configuring MCAM rule with a RQ number.
+6. Packet is DMA'ed and driver is notified.
+
+Egress
+------
+
+1. Driver prepares a send descriptor and submits to SQ for transmission.
+2. The SQ is already configured (by AF) to transmit on a specific link/channel.
+3. The SQ descriptor ring is maintained in buffers allocated from SQ mapped pool of NPA block LF.
+4. NIX block transmits the pkt on the designated channel.
+5. NPC MCAM entries can be installed to divert pkt onto a different channel.

MAINTAINERS

@@ -10000,6 +10000,16 @@ M:	Jerin Jacob <jerinj@marvell.com>
 L:	netdev@vger.kernel.org
 S:	Supported
 F:	drivers/net/ethernet/marvell/octeontx2/af/
+F:	Documentation/networking/device_drivers/marvell/octeontx2.rst
+
+MARVELL OCTEONTX2 PHYSICAL FUNCTION DRIVER
+M:	Sunil Goutham <sgoutham@marvell.com>
+M:	Geetha sowjanya <gakula@marvell.com>
+M:	Subbaraya Sundeep <sbhatta@marvell.com>
+M:	hariprasad <hkelam@marvell.com>
+L:	netdev@vger.kernel.org
+S:	Supported
+F:	drivers/net/ethernet/marvell/octeontx2/nic/
 
 MATROX FRAMEBUFFER DRIVER
 L:	linux-fbdev@vger.kernel.org

drivers/net/ethernet/marvell/octeontx2/Kconfig

@@ -25,3 +25,11 @@ config NDC_DIS_DYNAMIC_CACHING
 	  This config option disables caching of dynamic entries such as NIX SQEs
 	  , NPA stack pages etc in NDC. Also locks down NIX SQ/CQ/RQ/RSS and
 	  NPA Aura/Pool contexts.
+
+config OCTEONTX2_PF
+	tristate "Marvell OcteonTX2 NIC Physical Function driver"
+	select OCTEONTX2_MBOX
+	depends on (64BIT && COMPILE_TEST) || ARM64
+	depends on PCI
+	help
+	  This driver supports Marvell's OcteonTX2 NIC physical function.

drivers/net/ethernet/marvell/octeontx2/Makefile

@@ -3,4 +3,6 @@
 # Makefile for Marvell OcteonTX2 device drivers.
 #
 
+obj-$(CONFIG_OCTEONTX2_MBOX) += af/
 obj-$(CONFIG_OCTEONTX2_AF) += af/
+obj-$(CONFIG_OCTEONTX2_PF) += nic/

drivers/net/ethernet/marvell/octeontx2/af/common.h

@@ -143,8 +143,13 @@ enum nix_scheduler {
 	NIX_TXSCH_LVL_CNT = 0x5,
 };
 
-#define TXSCH_TL1_DFLT_RR_QTM      ((1 << 24) - 1)
+#define TXSCH_RR_QTM_MAX		((1 << 24) - 1)
+#define TXSCH_TL1_DFLT_RR_QTM		TXSCH_RR_QTM_MAX
 #define TXSCH_TL1_DFLT_RR_PRIO		(0x1ull)
+#define MAX_SCHED_WEIGHT		0xFF
+#define DFLT_RR_WEIGHT			71
+#define DFLT_RR_QTM	((DFLT_RR_WEIGHT * TXSCH_RR_QTM_MAX) \
+			 / MAX_SCHED_WEIGHT)
 
 /* Min/Max packet sizes, excluding FCS */
 #define	NIC_HW_MIN_FRS			40

drivers/net/ethernet/marvell/octeontx2/af/mbox.h

@@ -210,7 +210,8 @@ M(NIX_SET_RX_CFG,	0x8010, nix_set_rx_cfg, nix_rx_cfg, msg_rsp)	\
 M(NIX_LSO_FORMAT_CFG,	0x8011, nix_lso_format_cfg,			\
 				 nix_lso_format_cfg,			\
 				 nix_lso_format_cfg_rsp)		\
-M(NIX_RXVLAN_ALLOC,	0x8012, nix_rxvlan_alloc, msg_req, msg_rsp)
+M(NIX_RXVLAN_ALLOC,	0x8012, nix_rxvlan_alloc, msg_req, msg_rsp)	\
+M(NIX_GET_MAC_ADDR, 0x8018, nix_get_mac_addr, msg_req, nix_get_mac_addr_rsp) \
 
 /* Messages initiated by AF (range 0xC00 - 0xDFF) */
 #define MBOX_UP_CGX_MESSAGES						\
@@ -618,6 +619,11 @@ struct nix_set_mac_addr {
 	u8 mac_addr[ETH_ALEN]; /* MAC address to be set for this pcifunc */
 };
 
+struct nix_get_mac_addr_rsp {
+	struct mbox_msghdr hdr;
+	u8 mac_addr[ETH_ALEN];
+};
+
 struct nix_mark_format_cfg {
 	struct mbox_msghdr hdr;
 	u8 offset;

drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c

@@ -2546,6 +2546,23 @@ int rvu_mbox_handler_nix_set_mac_addr(struct rvu *rvu,
 	return 0;
 }
 
+int rvu_mbox_handler_nix_get_mac_addr(struct rvu *rvu,
+				      struct msg_req *req,
+				      struct nix_get_mac_addr_rsp *rsp)
+{
+	u16 pcifunc = req->hdr.pcifunc;
+	struct rvu_pfvf *pfvf;
+
+	if (!is_nixlf_attached(rvu, pcifunc))
+		return NIX_AF_ERR_AF_LF_INVALID;
+
+	pfvf = rvu_get_pfvf(rvu, pcifunc);
+
+	ether_addr_copy(rsp->mac_addr, pfvf->mac_addr);
+
+	return 0;
+}
+
 int rvu_mbox_handler_nix_set_rx_mode(struct rvu *rvu, struct nix_rx_mode *req,
 				     struct msg_rsp *rsp)
 {

drivers/net/ethernet/marvell/octeontx2/nic/Makefile

+# SPDX-License-Identifier: GPL-2.0
+#
+# Makefile for Marvell's OcteonTX2 ethernet device drivers
+#
+
+obj-$(CONFIG_OCTEONTX2_PF) += octeontx2_nicpf.o
+
+octeontx2_nicpf-y := otx2_pf.o otx2_common.o otx2_txrx.o otx2_ethtool.o
+
+ccflags-y += -I$(srctree)/drivers/net/ethernet/marvell/octeontx2/af

drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.c

+// SPDX-License-Identifier: GPL-2.0
+/* Marvell OcteonTx2 RVU Ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <net/tso.h>
+
+#include "otx2_reg.h"
+#include "otx2_common.h"
+#include "otx2_struct.h"
+
+static void otx2_nix_rq_op_stats(struct queue_stats *stats,
+				 struct otx2_nic *pfvf, int qidx)
+{
+	u64 incr = (u64)qidx << 32;
+	u64 *ptr;
+
+	ptr = (u64 *)otx2_get_regaddr(pfvf, NIX_LF_RQ_OP_OCTS);
+	stats->bytes = otx2_atomic64_add(incr, ptr);
+
+	ptr = (u64 *)otx2_get_regaddr(pfvf, NIX_LF_RQ_OP_PKTS);
+	stats->pkts = otx2_atomic64_add(incr, ptr);
+}
+
+static void otx2_nix_sq_op_stats(struct queue_stats *stats,
+				 struct otx2_nic *pfvf, int qidx)
+{
+	u64 incr = (u64)qidx << 32;
+	u64 *ptr;
+
+	ptr = (u64 *)otx2_get_regaddr(pfvf, NIX_LF_SQ_OP_OCTS);
+	stats->bytes = otx2_atomic64_add(incr, ptr);
+
+	ptr = (u64 *)otx2_get_regaddr(pfvf, NIX_LF_SQ_OP_PKTS);
+	stats->pkts = otx2_atomic64_add(incr, ptr);
+}
+
+void otx2_update_lmac_stats(struct otx2_nic *pfvf)
+{
+	struct msg_req *req;
+
+	if (!netif_running(pfvf->netdev))
+		return;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	req = otx2_mbox_alloc_msg_cgx_stats(&pfvf->mbox);
+	if (!req) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return;
+	}
+
+	otx2_sync_mbox_msg(&pfvf->mbox);
+	otx2_mbox_unlock(&pfvf->mbox);
+}
+
+int otx2_update_rq_stats(struct otx2_nic *pfvf, int qidx)
+{
+	struct otx2_rcv_queue *rq = &pfvf->qset.rq[qidx];
+
+	if (!pfvf->qset.rq)
+		return 0;
+
+	otx2_nix_rq_op_stats(&rq->stats, pfvf, qidx);
+	return 1;
+}
+
+int otx2_update_sq_stats(struct otx2_nic *pfvf, int qidx)
+{
+	struct otx2_snd_queue *sq = &pfvf->qset.sq[qidx];
+
+	if (!pfvf->qset.sq)
+		return 0;
+
+	otx2_nix_sq_op_stats(&sq->stats, pfvf, qidx);
+	return 1;
+}
+
+void otx2_get_dev_stats(struct otx2_nic *pfvf)
+{
+	struct otx2_dev_stats *dev_stats = &pfvf->hw.dev_stats;
+
+#define OTX2_GET_RX_STATS(reg) \
+	 otx2_read64(pfvf, NIX_LF_RX_STATX(reg))
+#define OTX2_GET_TX_STATS(reg) \
+	 otx2_read64(pfvf, NIX_LF_TX_STATX(reg))
+
+	dev_stats->rx_bytes = OTX2_GET_RX_STATS(RX_OCTS);
+	dev_stats->rx_drops = OTX2_GET_RX_STATS(RX_DROP);
+	dev_stats->rx_bcast_frames = OTX2_GET_RX_STATS(RX_BCAST);
+	dev_stats->rx_mcast_frames = OTX2_GET_RX_STATS(RX_MCAST);
+	dev_stats->rx_ucast_frames = OTX2_GET_RX_STATS(RX_UCAST);
+	dev_stats->rx_frames = dev_stats->rx_bcast_frames +
+			       dev_stats->rx_mcast_frames +
+			       dev_stats->rx_ucast_frames;
+
+	dev_stats->tx_bytes = OTX2_GET_TX_STATS(TX_OCTS);
+	dev_stats->tx_drops = OTX2_GET_TX_STATS(TX_DROP);
+	dev_stats->tx_bcast_frames = OTX2_GET_TX_STATS(TX_BCAST);
+	dev_stats->tx_mcast_frames = OTX2_GET_TX_STATS(TX_MCAST);
+	dev_stats->tx_ucast_frames = OTX2_GET_TX_STATS(TX_UCAST);
+	dev_stats->tx_frames = dev_stats->tx_bcast_frames +
+			       dev_stats->tx_mcast_frames +
+			       dev_stats->tx_ucast_frames;
+}
+
+void otx2_get_stats64(struct net_device *netdev,
+		      struct rtnl_link_stats64 *stats)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	struct otx2_dev_stats *dev_stats;
+
+	otx2_get_dev_stats(pfvf);
+
+	dev_stats = &pfvf->hw.dev_stats;
+	stats->rx_bytes = dev_stats->rx_bytes;
+	stats->rx_packets = dev_stats->rx_frames;
+	stats->rx_dropped = dev_stats->rx_drops;
+	stats->multicast = dev_stats->rx_mcast_frames;
+
+	stats->tx_bytes = dev_stats->tx_bytes;
+	stats->tx_packets = dev_stats->tx_frames;
+	stats->tx_dropped = dev_stats->tx_drops;
+}
+
+/* Sync MAC address with RVU AF */
+static int otx2_hw_set_mac_addr(struct otx2_nic *pfvf, u8 *mac)
+{
+	struct nix_set_mac_addr *req;
+	int err;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	req = otx2_mbox_alloc_msg_nix_set_mac_addr(&pfvf->mbox);
+	if (!req) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+
+	ether_addr_copy(req->mac_addr, mac);
+
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	otx2_mbox_unlock(&pfvf->mbox);
+	return err;
+}
+
+static int otx2_hw_get_mac_addr(struct otx2_nic *pfvf,
+				struct net_device *netdev)
+{
+	struct nix_get_mac_addr_rsp *rsp;
+	struct mbox_msghdr *msghdr;
+	struct msg_req *req;
+	int err;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	req = otx2_mbox_alloc_msg_nix_get_mac_addr(&pfvf->mbox);
+	if (!req) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	if (err) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return err;
+	}
+
+	msghdr = otx2_mbox_get_rsp(&pfvf->mbox.mbox, 0, &req->hdr);
+	if (!msghdr) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+	rsp = (struct nix_get_mac_addr_rsp *)msghdr;
+	ether_addr_copy(netdev->dev_addr, rsp->mac_addr);
+	otx2_mbox_unlock(&pfvf->mbox);
+
+	return 0;
+}
+
+int otx2_set_mac_address(struct net_device *netdev, void *p)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	struct sockaddr *addr = p;
+
+	if (!is_valid_ether_addr(addr->sa_data))
+		return -EADDRNOTAVAIL;
+
+	if (!otx2_hw_set_mac_addr(pfvf, addr->sa_data))
+		memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
+	else
+		return -EPERM;
+
+	return 0;
+}
+
+int otx2_hw_set_mtu(struct otx2_nic *pfvf, int mtu)
+{
+	struct nix_frs_cfg *req;
+	int err;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	req = otx2_mbox_alloc_msg_nix_set_hw_frs(&pfvf->mbox);
+	if (!req) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+
+	/* SMQ config limits maximum pkt size that can be transmitted */
+	req->update_smq = true;
+	pfvf->max_frs = mtu +  OTX2_ETH_HLEN;
+	req->maxlen = pfvf->max_frs;
+
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	otx2_mbox_unlock(&pfvf->mbox);
+	return err;
+}
+
+int otx2_set_flowkey_cfg(struct otx2_nic *pfvf)
+{
+	struct otx2_rss_info *rss = &pfvf->hw.rss_info;
+	struct nix_rss_flowkey_cfg *req;
+	int err;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	req = otx2_mbox_alloc_msg_nix_rss_flowkey_cfg(&pfvf->mbox);
+	if (!req) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+	req->mcam_index = -1; /* Default or reserved index */
+	req->flowkey_cfg = rss->flowkey_cfg;
+	req->group = DEFAULT_RSS_CONTEXT_GROUP;
+
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	otx2_mbox_unlock(&pfvf->mbox);
+	return err;
+}
+
+int otx2_set_rss_table(struct otx2_nic *pfvf)
+{
+	struct otx2_rss_info *rss = &pfvf->hw.rss_info;
+	struct mbox *mbox = &pfvf->mbox;
+	struct nix_aq_enq_req *aq;
+	int idx, err;
+
+	otx2_mbox_lock(mbox);
+	/* Get memory to put this msg */
+	for (idx = 0; idx < rss->rss_size; idx++) {
+		aq = otx2_mbox_alloc_msg_nix_aq_enq(mbox);
+		if (!aq) {
+			/* The shared memory buffer can be full.
+			 * Flush it and retry
+			 */
+			err = otx2_sync_mbox_msg(mbox);
+			if (err) {
+				otx2_mbox_unlock(mbox);
+				return err;
+			}
+			aq = otx2_mbox_alloc_msg_nix_aq_enq(mbox);
+			if (!aq) {
+				otx2_mbox_unlock(mbox);
+				return -ENOMEM;
+			}
+		}
+
+		aq->rss.rq = rss->ind_tbl[idx];
+
+		/* Fill AQ info */
+		aq->qidx = idx;
+		aq->ctype = NIX_AQ_CTYPE_RSS;
+		aq->op = NIX_AQ_INSTOP_INIT;
+	}
+	err = otx2_sync_mbox_msg(mbox);
+	otx2_mbox_unlock(mbox);
+	return err;
+}
+
+void otx2_set_rss_key(struct otx2_nic *pfvf)
+{
+	struct otx2_rss_info *rss = &pfvf->hw.rss_info;
+	u64 *key = (u64 *)&rss->key[4];
+	int idx;
+
+	/* 352bit or 44byte key needs to be configured as below
+	 * NIX_LF_RX_SECRETX0 = key<351:288>
+	 * NIX_LF_RX_SECRETX1 = key<287:224>
+	 * NIX_LF_RX_SECRETX2 = key<223:160>
+	 * NIX_LF_RX_SECRETX3 = key<159:96>
+	 * NIX_LF_RX_SECRETX4 = key<95:32>
+	 * NIX_LF_RX_SECRETX5<63:32> = key<31:0>
+	 */
+	otx2_write64(pfvf, NIX_LF_RX_SECRETX(5),
+		     (u64)(*((u32 *)&rss->key)) << 32);
+	idx = sizeof(rss->key) / sizeof(u64);
+	while (idx > 0) {
+		idx--;
+		otx2_write64(pfvf, NIX_LF_RX_SECRETX(idx), *key++);
+	}
+}
+
+int otx2_rss_init(struct otx2_nic *pfvf)
+{
+	struct otx2_rss_info *rss = &pfvf->hw.rss_info;
+	int idx, ret = 0;
+
+	rss->rss_size = sizeof(rss->ind_tbl);
+
+	/* Init RSS key if it is not setup already */
+	if (!rss->enable)
+		netdev_rss_key_fill(rss->key, sizeof(rss->key));
+	otx2_set_rss_key(pfvf);
+
+	if (!netif_is_rxfh_configured(pfvf->netdev)) {
+		/* Default indirection table */
+		for (idx = 0; idx < rss->rss_size; idx++)
+			rss->ind_tbl[idx] =
+				ethtool_rxfh_indir_default(idx,
+							   pfvf->hw.rx_queues);
+	}
+	ret = otx2_set_rss_table(pfvf);
+	if (ret)
+		return ret;
+
+	/* Flowkey or hash config to be used for generating flow tag */
+	rss->flowkey_cfg = rss->enable ? rss->flowkey_cfg :
+			   NIX_FLOW_KEY_TYPE_IPV4 | NIX_FLOW_KEY_TYPE_IPV6 |
+			   NIX_FLOW_KEY_TYPE_TCP | NIX_FLOW_KEY_TYPE_UDP |
+			   NIX_FLOW_KEY_TYPE_SCTP;
+
+	ret = otx2_set_flowkey_cfg(pfvf);
+	if (ret)
+		return ret;
+
+	rss->enable = true;
+	return 0;
+}
+
+void otx2_config_irq_coalescing(struct otx2_nic *pfvf, int qidx)
+{
+	/* Configure CQE interrupt coalescing parameters
+	 *
+	 * HW triggers an irq when ECOUNT > cq_ecount_wait, hence
+	 * set 1 less than cq_ecount_wait. And cq_time_wait is in
+	 * usecs, convert that to 100ns count.
+	 */
+	otx2_write64(pfvf, NIX_LF_CINTX_WAIT(qidx),
+		     ((u64)(pfvf->hw.cq_time_wait * 10) << 48) |
+		     ((u64)pfvf->hw.cq_qcount_wait << 32) |
+		     (pfvf->hw.cq_ecount_wait - 1));
+}
+
+dma_addr_t otx2_alloc_rbuf(struct otx2_nic *pfvf, struct otx2_pool *pool,
+			   gfp_t gfp)
+{
+	dma_addr_t iova;
+
+	/* Check if request can be accommodated in previous allocated page */
+	if (pool->page && ((pool->page_offset + pool->rbsize) <=
+	    (PAGE_SIZE << pool->rbpage_order))) {
+		pool->pageref++;
+		goto ret;
+	}
+
+	otx2_get_page(pool);
+
+	/* Allocate a new page */
+	pool->page = alloc_pages(gfp | __GFP_COMP | __GFP_NOWARN,
+				 pool->rbpage_order);
+	if (unlikely(!pool->page))
+		return -ENOMEM;
+
+	pool->page_offset = 0;
+ret:
+	iova = (u64)otx2_dma_map_page(pfvf, pool->page, pool->page_offset,
+				      pool->rbsize, DMA_FROM_DEVICE);
+	if (!iova) {
+		if (!pool->page_offset)
+			__free_pages(pool->page, pool->rbpage_order);
+		pool->page = NULL;
+		return -ENOMEM;
+	}
+	pool->page_offset += pool->rbsize;
+	return iova;
+}
+
+void otx2_tx_timeout(struct net_device *netdev, unsigned int txq)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+
+	schedule_work(&pfvf->reset_task);
+}
+
+void otx2_get_mac_from_af(struct net_device *netdev)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	int err;
+
+	err = otx2_hw_get_mac_addr(pfvf, netdev);
+	if (err)
+		dev_warn(pfvf->dev, "Failed to read mac from hardware\n");
+
+	/* If AF doesn't provide a valid MAC, generate a random one */
+	if (!is_valid_ether_addr(netdev->dev_addr))
+		eth_hw_addr_random(netdev);
+}
+
+static int otx2_get_link(struct otx2_nic *pfvf)
+{
+	int link = 0;
+	u16 map;
+
+	/* cgx lmac link */
+	if (pfvf->hw.tx_chan_base >= CGX_CHAN_BASE) {
+		map = pfvf->hw.tx_chan_base & 0x7FF;
+		link = 4 * ((map >> 8) & 0xF) + ((map >> 4) & 0xF);
+	}
+	/* LBK channel */
+	if (pfvf->hw.tx_chan_base < SDP_CHAN_BASE)
+		link = 12;
+
+	return link;
+}
+
+int otx2_txschq_config(struct otx2_nic *pfvf, int lvl)
+{
+	struct otx2_hw *hw = &pfvf->hw;
+	struct nix_txschq_config *req;
+	u64 schq, parent;
+
+	req = otx2_mbox_alloc_msg_nix_txschq_cfg(&pfvf->mbox);
+	if (!req)
+		return -ENOMEM;
+
+	req->lvl = lvl;
+	req->num_regs = 1;
+
+	schq = hw->txschq_list[lvl][0];
+	/* Set topology e.t.c configuration */
+	if (lvl == NIX_TXSCH_LVL_SMQ) {
+		req->reg[0] = NIX_AF_SMQX_CFG(schq);
+		req->regval[0] = ((pfvf->netdev->mtu  + OTX2_ETH_HLEN) << 8) |
+				   OTX2_MIN_MTU;
+
+		req->regval[0] |= (0x20ULL << 51) | (0x80ULL << 39) |
+				  (0x2ULL << 36);
+		req->num_regs++;
+		/* MDQ config */
+		parent =  hw->txschq_list[NIX_TXSCH_LVL_TL4][0];
+		req->reg[1] = NIX_AF_MDQX_PARENT(schq);
+		req->regval[1] = parent << 16;
+		req->num_regs++;
+		/* Set DWRR quantum */
+		req->reg[2] = NIX_AF_MDQX_SCHEDULE(schq);
+		req->regval[2] =  DFLT_RR_QTM;
+	} else if (lvl == NIX_TXSCH_LVL_TL4) {
+		parent =  hw->txschq_list[NIX_TXSCH_LVL_TL3][0];
+		req->reg[0] = NIX_AF_TL4X_PARENT(schq);
+		req->regval[0] = parent << 16;
+		req->num_regs++;
+		req->reg[1] = NIX_AF_TL4X_SCHEDULE(schq);
+		req->regval[1] = DFLT_RR_QTM;
+	} else if (lvl == NIX_TXSCH_LVL_TL3) {
+		parent = hw->txschq_list[NIX_TXSCH_LVL_TL2][0];
+		req->reg[0] = NIX_AF_TL3X_PARENT(schq);
+		req->regval[0] = parent << 16;
+		req->num_regs++;
+		req->reg[1] = NIX_AF_TL3X_SCHEDULE(schq);
+		req->regval[1] = DFLT_RR_QTM;
+	} else if (lvl == NIX_TXSCH_LVL_TL2) {
+		parent =  hw->txschq_list[NIX_TXSCH_LVL_TL1][0];
+		req->reg[0] = NIX_AF_TL2X_PARENT(schq);
+		req->regval[0] = parent << 16;
+
+		req->num_regs++;
+		req->reg[1] = NIX_AF_TL2X_SCHEDULE(schq);
+		req->regval[1] = TXSCH_TL1_DFLT_RR_PRIO << 24 | DFLT_RR_QTM;
+
+		req->num_regs++;
+		req->reg[2] = NIX_AF_TL3_TL2X_LINKX_CFG(schq,
+							otx2_get_link(pfvf));
+		/* Enable this queue and backpressure */
+		req->regval[2] = BIT_ULL(13) | BIT_ULL(12);
+
+	} else if (lvl == NIX_TXSCH_LVL_TL1) {
+		/* Default config for TL1.
+		 * For VF this is always ignored.
+		 */
+
+		/* Set DWRR quantum */
+		req->reg[0] = NIX_AF_TL1X_SCHEDULE(schq);
+		req->regval[0] = TXSCH_TL1_DFLT_RR_QTM;
+
+		req->num_regs++;
+		req->reg[1] = NIX_AF_TL1X_TOPOLOGY(schq);
+		req->regval[1] = (TXSCH_TL1_DFLT_RR_PRIO << 1);
+
+		req->num_regs++;
+		req->reg[2] = NIX_AF_TL1X_CIR(schq);
+		req->regval[2] = 0;
+	}
+
+	return otx2_sync_mbox_msg(&pfvf->mbox);
+}
+
+int otx2_txsch_alloc(struct otx2_nic *pfvf)
+{
+	struct nix_txsch_alloc_req *req;
+	int lvl;
+
+	/* Get memory to put this msg */
+	req = otx2_mbox_alloc_msg_nix_txsch_alloc(&pfvf->mbox);
+	if (!req)
+		return -ENOMEM;
+
+	/* Request one schq per level */
+	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++)
+		req->schq[lvl] = 1;
+
+	return otx2_sync_mbox_msg(&pfvf->mbox);
+}
+
+int otx2_txschq_stop(struct otx2_nic *pfvf)
+{
+	struct nix_txsch_free_req *free_req;
+	int lvl, schq, err;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	/* Free the transmit schedulers */
+	free_req = otx2_mbox_alloc_msg_nix_txsch_free(&pfvf->mbox);
+	if (!free_req) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+
+	free_req->flags = TXSCHQ_FREE_ALL;
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	otx2_mbox_unlock(&pfvf->mbox);
+
+	/* Clear the txschq list */
+	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++) {
+		for (schq = 0; schq < MAX_TXSCHQ_PER_FUNC; schq++)
+			pfvf->hw.txschq_list[lvl][schq] = 0;
+	}
+	return err;
+}
+
+void otx2_sqb_flush(struct otx2_nic *pfvf)
+{
+	int qidx, sqe_tail, sqe_head;
+	u64 incr, *ptr, val;
+
+	ptr = (u64 *)otx2_get_regaddr(pfvf, NIX_LF_SQ_OP_STATUS);
+	for (qidx = 0; qidx < pfvf->hw.tx_queues; qidx++) {
+		incr = (u64)qidx << 32;
+		while (1) {
+			val = otx2_atomic64_add(incr, ptr);
+			sqe_head = (val >> 20) & 0x3F;
+			sqe_tail = (val >> 28) & 0x3F;
+			if (sqe_head == sqe_tail)
+				break;
+			usleep_range(1, 3);
+		}
+	}
+}
+
+/* RED and drop levels of CQ on packet reception.
+ * For CQ level is measure of emptiness ( 0x0 = full, 255 = empty).
+ */
+#define RQ_PASS_LVL_CQ(skid, qsize)	((((skid) + 16) * 256) / (qsize))
+#define RQ_DROP_LVL_CQ(skid, qsize)	(((skid) * 256) / (qsize))
+
+/* RED and drop levels of AURA for packet reception.
+ * For AURA level is measure of fullness (0x0 = empty, 255 = full).
+ * Eg: For RQ length 1K, for pass/drop level 204/230.
+ * RED accepts pkts if free pointers > 102 & <= 205.
+ * Drops pkts if free pointers < 102.
+ */
+#define RQ_PASS_LVL_AURA (255 - ((95 * 256) / 100)) /* RED when 95% is full */
+#define RQ_DROP_LVL_AURA (255 - ((99 * 256) / 100)) /* Drop when 99% is full */
+
+/* Send skid of 2000 packets required for CQ size of 4K CQEs. */
+#define SEND_CQ_SKID	2000
+
+static int otx2_rq_init(struct otx2_nic *pfvf, u16 qidx, u16 lpb_aura)
+{
+	struct otx2_qset *qset = &pfvf->qset;
+	struct nix_aq_enq_req *aq;
+
+	/* Get memory to put this msg */
+	aq = otx2_mbox_alloc_msg_nix_aq_enq(&pfvf->mbox);
+	if (!aq)
+		return -ENOMEM;
+
+	aq->rq.cq = qidx;
+	aq->rq.ena = 1;
+	aq->rq.pb_caching = 1;
+	aq->rq.lpb_aura = lpb_aura; /* Use large packet buffer aura */
+	aq->rq.lpb_sizem1 = (DMA_BUFFER_LEN(pfvf->rbsize) / 8) - 1;
+	aq->rq.xqe_imm_size = 0; /* Copying of packet to CQE not needed */
+	aq->rq.flow_tagw = 32; /* Copy full 32bit flow_tag to CQE header */
+	aq->rq.qint_idx = 0;
+	aq->rq.lpb_drop_ena = 1; /* Enable RED dropping for AURA */
+	aq->rq.xqe_drop_ena = 1; /* Enable RED dropping for CQ/SSO */
+	aq->rq.xqe_pass = RQ_PASS_LVL_CQ(pfvf->hw.rq_skid, qset->rqe_cnt);
+	aq->rq.xqe_drop = RQ_DROP_LVL_CQ(pfvf->hw.rq_skid, qset->rqe_cnt);
+	aq->rq.lpb_aura_pass = RQ_PASS_LVL_AURA;
+	aq->rq.lpb_aura_drop = RQ_DROP_LVL_AURA;
+
+	/* Fill AQ info */
+	aq->qidx = qidx;
+	aq->ctype = NIX_AQ_CTYPE_RQ;
+	aq->op = NIX_AQ_INSTOP_INIT;
+
+	return otx2_sync_mbox_msg(&pfvf->mbox);
+}
+
+static int otx2_sq_init(struct otx2_nic *pfvf, u16 qidx, u16 sqb_aura)
+{
+	struct otx2_qset *qset = &pfvf->qset;
+	struct otx2_snd_queue *sq;
+	struct nix_aq_enq_req *aq;
+	struct otx2_pool *pool;
+	int err;
+
+	pool = &pfvf->qset.pool[sqb_aura];
+	sq = &qset->sq[qidx];
+	sq->sqe_size = NIX_SQESZ_W16 ? 64 : 128;
+	sq->sqe_cnt = qset->sqe_cnt;
+
+	err = qmem_alloc(pfvf->dev, &sq->sqe, 1, sq->sqe_size);
+	if (err)
+		return err;
+
+	err = qmem_alloc(pfvf->dev, &sq->tso_hdrs, qset->sqe_cnt,
+			 TSO_HEADER_SIZE);
+	if (err)
+		return err;
+
+	sq->sqe_base = sq->sqe->base;
+	sq->sg = kcalloc(qset->sqe_cnt, sizeof(struct sg_list), GFP_KERNEL);
+	if (!sq->sg)
+		return -ENOMEM;
+
+	sq->head = 0;
+	sq->sqe_per_sqb = (pfvf->hw.sqb_size / sq->sqe_size) - 1;
+	sq->num_sqbs = (qset->sqe_cnt + sq->sqe_per_sqb) / sq->sqe_per_sqb;
+	/* Set SQE threshold to 10% of total SQEs */
+	sq->sqe_thresh = ((sq->num_sqbs * sq->sqe_per_sqb) * 10) / 100;
+	sq->aura_id = sqb_aura;
+	sq->aura_fc_addr = pool->fc_addr->base;
+	sq->lmt_addr = (__force u64 *)(pfvf->reg_base + LMT_LF_LMTLINEX(qidx));
+	sq->io_addr = (__force u64)otx2_get_regaddr(pfvf, NIX_LF_OP_SENDX(0));
+
+	sq->stats.bytes = 0;
+	sq->stats.pkts = 0;
+
+	/* Get memory to put this msg */
+	aq = otx2_mbox_alloc_msg_nix_aq_enq(&pfvf->mbox);
+	if (!aq)
+		return -ENOMEM;
+
+	aq->sq.cq = pfvf->hw.rx_queues + qidx;
+	aq->sq.max_sqe_size = NIX_MAXSQESZ_W16; /* 128 byte */
+	aq->sq.cq_ena = 1;
+	aq->sq.ena = 1;
+	/* Only one SMQ is allocated, map all SQ's to that SMQ  */
+	aq->sq.smq = pfvf->hw.txschq_list[NIX_TXSCH_LVL_SMQ][0];
+	aq->sq.smq_rr_quantum = DFLT_RR_QTM;
+	aq->sq.default_chan = pfvf->hw.tx_chan_base;
+	aq->sq.sqe_stype = NIX_STYPE_STF; /* Cache SQB */
+	aq->sq.sqb_aura = sqb_aura;
+	aq->sq.sq_int_ena = NIX_SQINT_BITS;
+	aq->sq.qint_idx = 0;
+	/* Due pipelining impact minimum 2000 unused SQ CQE's
+	 * need to maintain to avoid CQ overflow.
+	 */
+	aq->sq.cq_limit = ((SEND_CQ_SKID * 256) / (sq->sqe_cnt));
+
+	/* Fill AQ info */
+	aq->qidx = qidx;
+	aq->ctype = NIX_AQ_CTYPE_SQ;
+	aq->op = NIX_AQ_INSTOP_INIT;
+
+	return otx2_sync_mbox_msg(&pfvf->mbox);
+}
+
+static int otx2_cq_init(struct otx2_nic *pfvf, u16 qidx)
+{
+	struct otx2_qset *qset = &pfvf->qset;
+	struct nix_aq_enq_req *aq;
+	struct otx2_cq_queue *cq;
+	int err, pool_id;
+
+	cq = &qset->cq[qidx];
+	cq->cq_idx = qidx;
+	if (qidx < pfvf->hw.rx_queues) {
+		cq->cq_type = CQ_RX;
+		cq->cint_idx = qidx;
+		cq->cqe_cnt = qset->rqe_cnt;
+	} else {
+		cq->cq_type = CQ_TX;
+		cq->cint_idx = qidx - pfvf->hw.rx_queues;
+		cq->cqe_cnt = qset->sqe_cnt;
+	}
+	cq->cqe_size = pfvf->qset.xqe_size;
+
+	/* Allocate memory for CQEs */
+	err = qmem_alloc(pfvf->dev, &cq->cqe, cq->cqe_cnt, cq->cqe_size);
+	if (err)
+		return err;
+
+	/* Save CQE CPU base for faster reference */
+	cq->cqe_base = cq->cqe->base;
+	/* In case where all RQs auras point to single pool,
+	 * all CQs receive buffer pool also point to same pool.
+	 */
+	pool_id = ((cq->cq_type == CQ_RX) &&
+		   (pfvf->hw.rqpool_cnt != pfvf->hw.rx_queues)) ? 0 : qidx;
+	cq->rbpool = &qset->pool[pool_id];
+	cq->refill_task_sched = false;
+
+	/* Get memory to put this msg */
+	aq = otx2_mbox_alloc_msg_nix_aq_enq(&pfvf->mbox);
+	if (!aq)
+		return -ENOMEM;
+
+	aq->cq.ena = 1;
+	aq->cq.qsize = Q_SIZE(cq->cqe_cnt, 4);
+	aq->cq.caching = 1;
+	aq->cq.base = cq->cqe->iova;
+	aq->cq.cint_idx = cq->cint_idx;
+	aq->cq.cq_err_int_ena = NIX_CQERRINT_BITS;
+	aq->cq.qint_idx = 0;
+	aq->cq.avg_level = 255;
+
+	if (qidx < pfvf->hw.rx_queues) {
+		aq->cq.drop = RQ_DROP_LVL_CQ(pfvf->hw.rq_skid, cq->cqe_cnt);
+		aq->cq.drop_ena = 1;
+	}
+
+	/* Fill AQ info */
+	aq->qidx = qidx;
+	aq->ctype = NIX_AQ_CTYPE_CQ;
+	aq->op = NIX_AQ_INSTOP_INIT;
+
+	return otx2_sync_mbox_msg(&pfvf->mbox);
+}
+
+static void otx2_pool_refill_task(struct work_struct *work)
+{
+	struct otx2_cq_queue *cq;
+	struct otx2_pool *rbpool;
+	struct refill_work *wrk;
+	int qidx, free_ptrs = 0;
+	struct otx2_nic *pfvf;
+	s64 bufptr;
+
+	wrk = container_of(work, struct refill_work, pool_refill_work.work);
+	pfvf = wrk->pf;
+	qidx = wrk - pfvf->refill_wrk;
+	cq = &pfvf->qset.cq[qidx];
+	rbpool = cq->rbpool;
+	free_ptrs = cq->pool_ptrs;
+
+	while (cq->pool_ptrs) {
+		bufptr = otx2_alloc_rbuf(pfvf, rbpool, GFP_KERNEL);
+		if (bufptr <= 0) {
+			/* Schedule a WQ if we fails to free atleast half of the
+			 * pointers else enable napi for this RQ.
+			 */
+			if (!((free_ptrs - cq->pool_ptrs) > free_ptrs / 2)) {
+				struct delayed_work *dwork;
+
+				dwork = &wrk->pool_refill_work;
+				schedule_delayed_work(dwork,
+						      msecs_to_jiffies(100));
+			} else {
+				cq->refill_task_sched = false;
+			}
+			return;
+		}
+		otx2_aura_freeptr(pfvf, qidx, bufptr + OTX2_HEAD_ROOM);
+		cq->pool_ptrs--;
+	}
+	cq->refill_task_sched = false;
+}
+
+int otx2_config_nix_queues(struct otx2_nic *pfvf)
+{
+	int qidx, err;
+
+	/* Initialize RX queues */
+	for (qidx = 0; qidx < pfvf->hw.rx_queues; qidx++) {
+		u16 lpb_aura = otx2_get_pool_idx(pfvf, AURA_NIX_RQ, qidx);
+
+		err = otx2_rq_init(pfvf, qidx, lpb_aura);
+		if (err)
+			return err;
+	}
+
+	/* Initialize TX queues */
+	for (qidx = 0; qidx < pfvf->hw.tx_queues; qidx++) {
+		u16 sqb_aura = otx2_get_pool_idx(pfvf, AURA_NIX_SQ, qidx);
+
+		err = otx2_sq_init(pfvf, qidx, sqb_aura);
+		if (err)
+			return err;
+	}
+
+	/* Initialize completion queues */
+	for (qidx = 0; qidx < pfvf->qset.cq_cnt; qidx++) {
+		err = otx2_cq_init(pfvf, qidx);
+		if (err)
+			return err;
+	}
+
+	/* Initialize work queue for receive buffer refill */
+	pfvf->refill_wrk = devm_kcalloc(pfvf->dev, pfvf->qset.cq_cnt,
+					sizeof(struct refill_work), GFP_KERNEL);
+	if (!pfvf->refill_wrk)
+		return -ENOMEM;
+
+	for (qidx = 0; qidx < pfvf->qset.cq_cnt; qidx++) {
+		pfvf->refill_wrk[qidx].pf = pfvf;
+		INIT_DELAYED_WORK(&pfvf->refill_wrk[qidx].pool_refill_work,
+				  otx2_pool_refill_task);
+	}
+	return 0;
+}
+
+int otx2_config_nix(struct otx2_nic *pfvf)
+{
+	struct nix_lf_alloc_req  *nixlf;
+	struct nix_lf_alloc_rsp *rsp;
+	int err;
+
+	pfvf->qset.xqe_size = NIX_XQESZ_W16 ? 128 : 512;
+
+	/* Get memory to put this msg */
+	nixlf = otx2_mbox_alloc_msg_nix_lf_alloc(&pfvf->mbox);
+	if (!nixlf)
+		return -ENOMEM;
+
+	/* Set RQ/SQ/CQ counts */
+	nixlf->rq_cnt = pfvf->hw.rx_queues;
+	nixlf->sq_cnt = pfvf->hw.tx_queues;
+	nixlf->cq_cnt = pfvf->qset.cq_cnt;
+	nixlf->rss_sz = MAX_RSS_INDIR_TBL_SIZE;
+	nixlf->rss_grps = 1; /* Single RSS indir table supported, for now */
+	nixlf->xqe_sz = NIX_XQESZ_W16;
+	/* We don't know absolute NPA LF idx attached.
+	 * AF will replace 'RVU_DEFAULT_PF_FUNC' with
+	 * NPA LF attached to this RVU PF/VF.
+	 */
+	nixlf->npa_func = RVU_DEFAULT_PF_FUNC;
+	/* Disable alignment pad, enable L2 length check,
+	 * enable L4 TCP/UDP checksum verification.
+	 */
+	nixlf->rx_cfg = BIT_ULL(33) | BIT_ULL(35) | BIT_ULL(37);
+
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	if (err)
+		return err;
+
+	rsp = (struct nix_lf_alloc_rsp *)otx2_mbox_get_rsp(&pfvf->mbox.mbox, 0,
+							   &nixlf->hdr);
+	if (IS_ERR(rsp))
+		return PTR_ERR(rsp);
+
+	if (rsp->qints < 1)
+		return -ENXIO;
+
+	return rsp->hdr.rc;
+}
+
+void otx2_sq_free_sqbs(struct otx2_nic *pfvf)
+{
+	struct otx2_qset *qset = &pfvf->qset;
+	struct otx2_hw *hw = &pfvf->hw;
+	struct otx2_snd_queue *sq;
+	int sqb, qidx;
+	u64 iova, pa;
+
+	for (qidx = 0; qidx < hw->tx_queues; qidx++) {
+		sq = &qset->sq[qidx];
+		if (!sq->sqb_ptrs)
+			continue;
+		for (sqb = 0; sqb < sq->sqb_count; sqb++) {
+			if (!sq->sqb_ptrs[sqb])
+				continue;
+			iova = sq->sqb_ptrs[sqb];
+			pa = otx2_iova_to_phys(pfvf->iommu_domain, iova);
+			dma_unmap_page_attrs(pfvf->dev, iova, hw->sqb_size,
+					     DMA_FROM_DEVICE,
+					     DMA_ATTR_SKIP_CPU_SYNC);
+			put_page(virt_to_page(phys_to_virt(pa)));
+		}
+		sq->sqb_count = 0;
+	}
+}
+
+void otx2_free_aura_ptr(struct otx2_nic *pfvf, int type)
+{
+	int pool_id, pool_start = 0, pool_end = 0, size = 0;
+	u64 iova, pa;
+
+	if (type == AURA_NIX_SQ) {
+		pool_start = otx2_get_pool_idx(pfvf, type, 0);
+		pool_end =  pool_start + pfvf->hw.sqpool_cnt;
+		size = pfvf->hw.sqb_size;
+	}
+	if (type == AURA_NIX_RQ) {
+		pool_start = otx2_get_pool_idx(pfvf, type, 0);
+		pool_end = pfvf->hw.rqpool_cnt;
+		size = pfvf->rbsize;
+	}
+
+	/* Free SQB and RQB pointers from the aura pool */
+	for (pool_id = pool_start; pool_id < pool_end; pool_id++) {
+		iova = otx2_aura_allocptr(pfvf, pool_id);
+		while (iova) {
+			if (type == AURA_NIX_RQ)
+				iova -= OTX2_HEAD_ROOM;
+
+			pa = otx2_iova_to_phys(pfvf->iommu_domain, iova);
+			dma_unmap_page_attrs(pfvf->dev, iova, size,
+					     DMA_FROM_DEVICE,
+					     DMA_ATTR_SKIP_CPU_SYNC);
+			put_page(virt_to_page(phys_to_virt(pa)));
+			iova = otx2_aura_allocptr(pfvf, pool_id);
+		}
+	}
+}
+
+void otx2_aura_pool_free(struct otx2_nic *pfvf)
+{
+	struct otx2_pool *pool;
+	int pool_id;
+
+	if (!pfvf->qset.pool)
+		return;
+
+	for (pool_id = 0; pool_id < pfvf->hw.pool_cnt; pool_id++) {
+		pool = &pfvf->qset.pool[pool_id];
+		qmem_free(pfvf->dev, pool->stack);
+		qmem_free(pfvf->dev, pool->fc_addr);
+	}
+	devm_kfree(pfvf->dev, pfvf->qset.pool);
+}
+
+static int otx2_aura_init(struct otx2_nic *pfvf, int aura_id,
+			  int pool_id, int numptrs)
+{
+	struct npa_aq_enq_req *aq;
+	struct otx2_pool *pool;
+	int err;
+
+	pool = &pfvf->qset.pool[pool_id];
+
+	/* Allocate memory for HW to update Aura count.
+	 * Alloc one cache line, so that it fits all FC_STYPE modes.
+	 */
+	if (!pool->fc_addr) {
+		err = qmem_alloc(pfvf->dev, &pool->fc_addr, 1, OTX2_ALIGN);
+		if (err)
+			return err;
+	}
+
+	/* Initialize this aura's context via AF */
+	aq = otx2_mbox_alloc_msg_npa_aq_enq(&pfvf->mbox);
+	if (!aq) {
+		/* Shared mbox memory buffer is full, flush it and retry */
+		err = otx2_sync_mbox_msg(&pfvf->mbox);
+		if (err)
+			return err;
+		aq = otx2_mbox_alloc_msg_npa_aq_enq(&pfvf->mbox);
+		if (!aq)
+			return -ENOMEM;
+	}
+
+	aq->aura_id = aura_id;
+	/* Will be filled by AF with correct pool context address */
+	aq->aura.pool_addr = pool_id;
+	aq->aura.pool_caching = 1;
+	aq->aura.shift = ilog2(numptrs) - 8;
+	aq->aura.count = numptrs;
+	aq->aura.limit = numptrs;
+	aq->aura.avg_level = 255;
+	aq->aura.ena = 1;
+	aq->aura.fc_ena = 1;
+	aq->aura.fc_addr = pool->fc_addr->iova;
+	aq->aura.fc_hyst_bits = 0; /* Store count on all updates */
+
+	/* Fill AQ info */
+	aq->ctype = NPA_AQ_CTYPE_AURA;
+	aq->op = NPA_AQ_INSTOP_INIT;
+
+	return 0;
+}
+
+static int otx2_pool_init(struct otx2_nic *pfvf, u16 pool_id,
+			  int stack_pages, int numptrs, int buf_size)
+{
+	struct npa_aq_enq_req *aq;
+	struct otx2_pool *pool;
+	int err;
+
+	pool = &pfvf->qset.pool[pool_id];
+	/* Alloc memory for stack which is used to store buffer pointers */
+	err = qmem_alloc(pfvf->dev, &pool->stack,
+			 stack_pages, pfvf->hw.stack_pg_bytes);
+	if (err)
+		return err;
+
+	pool->rbsize = buf_size;
+	pool->rbpage_order = get_order(buf_size);
+
+	/* Initialize this pool's context via AF */
+	aq = otx2_mbox_alloc_msg_npa_aq_enq(&pfvf->mbox);
+	if (!aq) {
+		/* Shared mbox memory buffer is full, flush it and retry */
+		err = otx2_sync_mbox_msg(&pfvf->mbox);
+		if (err) {
+			qmem_free(pfvf->dev, pool->stack);
+			return err;
+		}
+		aq = otx2_mbox_alloc_msg_npa_aq_enq(&pfvf->mbox);
+		if (!aq) {
+			qmem_free(pfvf->dev, pool->stack);
+			return -ENOMEM;
+		}
+	}
+
+	aq->aura_id = pool_id;
+	aq->pool.stack_base = pool->stack->iova;
+	aq->pool.stack_caching = 1;
+	aq->pool.ena = 1;
+	aq->pool.buf_size = buf_size / 128;
+	aq->pool.stack_max_pages = stack_pages;
+	aq->pool.shift = ilog2(numptrs) - 8;
+	aq->pool.ptr_start = 0;
+	aq->pool.ptr_end = ~0ULL;
+
+	/* Fill AQ info */
+	aq->ctype = NPA_AQ_CTYPE_POOL;
+	aq->op = NPA_AQ_INSTOP_INIT;
+
+	return 0;
+}
+
+int otx2_sq_aura_pool_init(struct otx2_nic *pfvf)
+{
+	int qidx, pool_id, stack_pages, num_sqbs;
+	struct otx2_qset *qset = &pfvf->qset;
+	struct otx2_hw *hw = &pfvf->hw;
+	struct otx2_snd_queue *sq;
+	struct otx2_pool *pool;
+	int err, ptr;
+	s64 bufptr;
+
+	/* Calculate number of SQBs needed.
+	 *
+	 * For a 128byte SQE, and 4K size SQB, 31 SQEs will fit in one SQB.
+	 * Last SQE is used for pointing to next SQB.
+	 */
+	num_sqbs = (hw->sqb_size / 128) - 1;
+	num_sqbs = (qset->sqe_cnt + num_sqbs) / num_sqbs;
+
+	/* Get no of stack pages needed */
+	stack_pages =
+		(num_sqbs + hw->stack_pg_ptrs - 1) / hw->stack_pg_ptrs;
+
+	for (qidx = 0; qidx < hw->tx_queues; qidx++) {
+		pool_id = otx2_get_pool_idx(pfvf, AURA_NIX_SQ, qidx);
+		/* Initialize aura context */
+		err = otx2_aura_init(pfvf, pool_id, pool_id, num_sqbs);
+		if (err)
+			goto fail;
+
+		/* Initialize pool context */
+		err = otx2_pool_init(pfvf, pool_id, stack_pages,
+				     num_sqbs, hw->sqb_size);
+		if (err)
+			goto fail;
+	}
+
+	/* Flush accumulated messages */
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	if (err)
+		goto fail;
+
+	/* Allocate pointers and free them to aura/pool */
+	for (qidx = 0; qidx < hw->tx_queues; qidx++) {
+		pool_id = otx2_get_pool_idx(pfvf, AURA_NIX_SQ, qidx);
+		pool = &pfvf->qset.pool[pool_id];
+
+		sq = &qset->sq[qidx];
+		sq->sqb_count = 0;
+		sq->sqb_ptrs = kcalloc(num_sqbs, sizeof(u64 *), GFP_KERNEL);
+		if (!sq->sqb_ptrs)
+			return -ENOMEM;
+
+		for (ptr = 0; ptr < num_sqbs; ptr++) {
+			bufptr = otx2_alloc_rbuf(pfvf, pool, GFP_KERNEL);
+			if (bufptr <= 0)
+				return bufptr;
+			otx2_aura_freeptr(pfvf, pool_id, bufptr);
+			sq->sqb_ptrs[sq->sqb_count++] = (u64)bufptr;
+		}
+		otx2_get_page(pool);
+	}
+
+	return 0;
+fail:
+	otx2_mbox_reset(&pfvf->mbox.mbox, 0);
+	otx2_aura_pool_free(pfvf);
+	return err;
+}
+
+int otx2_rq_aura_pool_init(struct otx2_nic *pfvf)
+{
+	struct otx2_hw *hw = &pfvf->hw;
+	int stack_pages, pool_id, rq;
+	struct otx2_pool *pool;
+	int err, ptr, num_ptrs;
+	s64 bufptr;
+
+	num_ptrs = pfvf->qset.rqe_cnt;
+
+	stack_pages =
+		(num_ptrs + hw->stack_pg_ptrs - 1) / hw->stack_pg_ptrs;
+
+	for (rq = 0; rq < hw->rx_queues; rq++) {
+		pool_id = otx2_get_pool_idx(pfvf, AURA_NIX_RQ, rq);
+		/* Initialize aura context */
+		err = otx2_aura_init(pfvf, pool_id, pool_id, num_ptrs);
+		if (err)
+			goto fail;
+	}
+	for (pool_id = 0; pool_id < hw->rqpool_cnt; pool_id++) {
+		err = otx2_pool_init(pfvf, pool_id, stack_pages,
+				     num_ptrs, pfvf->rbsize);
+		if (err)
+			goto fail;
+	}
+
+	/* Flush accumulated messages */
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	if (err)
+		goto fail;
+
+	/* Allocate pointers and free them to aura/pool */
+	for (pool_id = 0; pool_id < hw->rqpool_cnt; pool_id++) {
+		pool = &pfvf->qset.pool[pool_id];
+		for (ptr = 0; ptr < num_ptrs; ptr++) {
+			bufptr = otx2_alloc_rbuf(pfvf, pool, GFP_KERNEL);
+			if (bufptr <= 0)
+				return bufptr;
+			otx2_aura_freeptr(pfvf, pool_id,
+					  bufptr + OTX2_HEAD_ROOM);
+		}
+		otx2_get_page(pool);
+	}
+
+	return 0;
+fail:
+	otx2_mbox_reset(&pfvf->mbox.mbox, 0);
+	otx2_aura_pool_free(pfvf);
+	return err;
+}
+
+int otx2_config_npa(struct otx2_nic *pfvf)
+{
+	struct otx2_qset *qset = &pfvf->qset;
+	struct npa_lf_alloc_req  *npalf;
+	struct otx2_hw *hw = &pfvf->hw;
+	int aura_cnt;
+
+	/* Pool - Stack of free buffer pointers
+	 * Aura - Alloc/frees pointers from/to pool for NIX DMA.
+	 */
+
+	if (!hw->pool_cnt)
+		return -EINVAL;
+
+	qset->pool = devm_kzalloc(pfvf->dev, sizeof(struct otx2_pool) *
+				  hw->pool_cnt, GFP_KERNEL);
+	if (!qset->pool)
+		return -ENOMEM;
+
+	/* Get memory to put this msg */
+	npalf = otx2_mbox_alloc_msg_npa_lf_alloc(&pfvf->mbox);
+	if (!npalf)
+		return -ENOMEM;
+
+	/* Set aura and pool counts */
+	npalf->nr_pools = hw->pool_cnt;
+	aura_cnt = ilog2(roundup_pow_of_two(hw->pool_cnt));
+	npalf->aura_sz = (aura_cnt >= ilog2(128)) ? (aura_cnt - 6) : 1;
+
+	return otx2_sync_mbox_msg(&pfvf->mbox);
+}
+
+int otx2_detach_resources(struct mbox *mbox)
+{
+	struct rsrc_detach *detach;
+
+	otx2_mbox_lock(mbox);
+	detach = otx2_mbox_alloc_msg_detach_resources(mbox);
+	if (!detach) {
+		otx2_mbox_unlock(mbox);
+		return -ENOMEM;
+	}
+
+	/* detach all */
+	detach->partial = false;
+
+	/* Send detach request to AF */
+	otx2_mbox_msg_send(&mbox->mbox, 0);
+	otx2_mbox_unlock(mbox);
+	return 0;
+}
+
+int otx2_attach_npa_nix(struct otx2_nic *pfvf)
+{
+	struct rsrc_attach *attach;
+	struct msg_req *msix;
+	int err;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	/* Get memory to put this msg */
+	attach = otx2_mbox_alloc_msg_attach_resources(&pfvf->mbox);
+	if (!attach) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+
+	attach->npalf = true;
+	attach->nixlf = true;
+
+	/* Send attach request to AF */
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	if (err) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return err;
+	}
+
+	pfvf->nix_blkaddr = BLKADDR_NIX0;
+
+	/* If the platform has two NIX blocks then LF may be
+	 * allocated from NIX1.
+	 */
+	if (otx2_read64(pfvf, RVU_PF_BLOCK_ADDRX_DISC(BLKADDR_NIX1)) & 0x1FFULL)
+		pfvf->nix_blkaddr = BLKADDR_NIX1;
+
+	/* Get NPA and NIX MSIX vector offsets */
+	msix = otx2_mbox_alloc_msg_msix_offset(&pfvf->mbox);
+	if (!msix) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	if (err) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return err;
+	}
+	otx2_mbox_unlock(&pfvf->mbox);
+
+	if (pfvf->hw.npa_msixoff == MSIX_VECTOR_INVALID ||
+	    pfvf->hw.nix_msixoff == MSIX_VECTOR_INVALID) {
+		dev_err(pfvf->dev,
+			"RVUPF: Invalid MSIX vector offset for NPA/NIX\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+void otx2_ctx_disable(struct mbox *mbox, int type, bool npa)
+{
+	struct hwctx_disable_req *req;
+
+	otx2_mbox_lock(mbox);
+	/* Request AQ to disable this context */
+	if (npa)
+		req = otx2_mbox_alloc_msg_npa_hwctx_disable(mbox);
+	else
+		req = otx2_mbox_alloc_msg_nix_hwctx_disable(mbox);
+
+	if (!req) {
+		otx2_mbox_unlock(mbox);
+		return;
+	}
+
+	req->ctype = type;
+
+	if (otx2_sync_mbox_msg(mbox))
+		dev_err(mbox->pfvf->dev, "%s failed to disable context\n",
+			__func__);
+
+	otx2_mbox_unlock(mbox);
+}
+
+/* Mbox message handlers */
+void mbox_handler_cgx_stats(struct otx2_nic *pfvf,
+			    struct cgx_stats_rsp *rsp)
+{
+	int id;
+
+	for (id = 0; id < CGX_RX_STATS_COUNT; id++)
+		pfvf->hw.cgx_rx_stats[id] = rsp->rx_stats[id];
+	for (id = 0; id < CGX_TX_STATS_COUNT; id++)
+		pfvf->hw.cgx_tx_stats[id] = rsp->tx_stats[id];
+}
+
+void mbox_handler_nix_txsch_alloc(struct otx2_nic *pf,
+				  struct nix_txsch_alloc_rsp *rsp)
+{
+	int lvl, schq;
+
+	/* Setup transmit scheduler list */
+	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++)
+		for (schq = 0; schq < rsp->schq[lvl]; schq++)
+			pf->hw.txschq_list[lvl][schq] =
+				rsp->schq_list[lvl][schq];
+}
+
+void mbox_handler_npa_lf_alloc(struct otx2_nic *pfvf,
+			       struct npa_lf_alloc_rsp *rsp)
+{
+	pfvf->hw.stack_pg_ptrs = rsp->stack_pg_ptrs;
+	pfvf->hw.stack_pg_bytes = rsp->stack_pg_bytes;
+}
+
+void mbox_handler_nix_lf_alloc(struct otx2_nic *pfvf,
+			       struct nix_lf_alloc_rsp *rsp)
+{
+	pfvf->hw.sqb_size = rsp->sqb_size;
+	pfvf->hw.rx_chan_base = rsp->rx_chan_base;
+	pfvf->hw.tx_chan_base = rsp->tx_chan_base;
+	pfvf->hw.lso_tsov4_idx = rsp->lso_tsov4_idx;
+	pfvf->hw.lso_tsov6_idx = rsp->lso_tsov6_idx;
+}
+
+void mbox_handler_msix_offset(struct otx2_nic *pfvf,
+			      struct msix_offset_rsp *rsp)
+{
+	pfvf->hw.npa_msixoff = rsp->npa_msixoff;
+	pfvf->hw.nix_msixoff = rsp->nix_msixoff;
+}
+
+void otx2_free_cints(struct otx2_nic *pfvf, int n)
+{
+	struct otx2_qset *qset = &pfvf->qset;
+	struct otx2_hw *hw = &pfvf->hw;
+	int irq, qidx;
+
+	for (qidx = 0, irq = hw->nix_msixoff + NIX_LF_CINT_VEC_START;
+	     qidx < n;
+	     qidx++, irq++) {
+		int vector = pci_irq_vector(pfvf->pdev, irq);
+
+		irq_set_affinity_hint(vector, NULL);
+		free_cpumask_var(hw->affinity_mask[irq]);
+		free_irq(vector, &qset->napi[qidx]);
+	}
+}
+
+void otx2_set_cints_affinity(struct otx2_nic *pfvf)
+{
+	struct otx2_hw *hw = &pfvf->hw;
+	int vec, cpu, irq, cint;
+
+	vec = hw->nix_msixoff + NIX_LF_CINT_VEC_START;
+	cpu = cpumask_first(cpu_online_mask);
+
+	/* CQ interrupts */
+	for (cint = 0; cint < pfvf->hw.cint_cnt; cint++, vec++) {
+		if (!alloc_cpumask_var(&hw->affinity_mask[vec], GFP_KERNEL))
+			return;
+
+		cpumask_set_cpu(cpu, hw->affinity_mask[vec]);
+
+		irq = pci_irq_vector(pfvf->pdev, vec);
+		irq_set_affinity_hint(irq, hw->affinity_mask[vec]);
+
+		cpu = cpumask_next(cpu, cpu_online_mask);
+		if (unlikely(cpu >= nr_cpu_ids))
+			cpu = 0;
+	}
+}
+
+#define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
+int __weak								\
+otx2_mbox_up_handler_ ## _fn_name(struct otx2_nic *pfvf,		\
+				struct _req_type *req,			\
+				struct _rsp_type *rsp)			\
+{									\
+	/* Nothing to do here */					\
+	return 0;							\
+}									\
+EXPORT_SYMBOL(otx2_mbox_up_handler_ ## _fn_name);
+MBOX_UP_CGX_MESSAGES
+#undef M

drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Marvell OcteonTx2 RVU Ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef OTX2_COMMON_H
+#define OTX2_COMMON_H
+
+#include <linux/pci.h>
+#include <linux/iommu.h>
+
+#include <mbox.h>
+#include "otx2_reg.h"
+#include "otx2_txrx.h"
+
+/* PCI device IDs */
+#define PCI_DEVID_OCTEONTX2_RVU_PF              0xA063
+
+#define PCI_SUBSYS_DEVID_96XX_RVU_PFVF		0xB200
+
+/* PCI BAR nos */
+#define PCI_CFG_REG_BAR_NUM                     2
+#define PCI_MBOX_BAR_NUM                        4
+
+#define NAME_SIZE                               32
+
+enum arua_mapped_qtypes {
+	AURA_NIX_RQ,
+	AURA_NIX_SQ,
+};
+
+/* NIX LF interrupts range*/
+#define NIX_LF_QINT_VEC_START			0x00
+#define NIX_LF_CINT_VEC_START			0x40
+#define NIX_LF_GINT_VEC				0x80
+#define NIX_LF_ERR_VEC				0x81
+#define NIX_LF_POISON_VEC			0x82
+
+/* RSS configuration */
+struct otx2_rss_info {
+	u8 enable;
+	u32 flowkey_cfg;
+	u16 rss_size;
+	u8  ind_tbl[MAX_RSS_INDIR_TBL_SIZE];
+#define RSS_HASH_KEY_SIZE	44   /* 352 bit key */
+	u8  key[RSS_HASH_KEY_SIZE];
+};
+
+/* NIX (or NPC) RX errors */
+enum otx2_errlvl {
+	NPC_ERRLVL_RE,
+	NPC_ERRLVL_LID_LA,
+	NPC_ERRLVL_LID_LB,
+	NPC_ERRLVL_LID_LC,
+	NPC_ERRLVL_LID_LD,
+	NPC_ERRLVL_LID_LE,
+	NPC_ERRLVL_LID_LF,
+	NPC_ERRLVL_LID_LG,
+	NPC_ERRLVL_LID_LH,
+	NPC_ERRLVL_NIX = 0x0F,
+};
+
+enum otx2_errcodes_re {
+	/* NPC_ERRLVL_RE errcodes */
+	ERRCODE_FCS = 0x7,
+	ERRCODE_FCS_RCV = 0x8,
+	ERRCODE_UNDERSIZE = 0x10,
+	ERRCODE_OVERSIZE = 0x11,
+	ERRCODE_OL2_LEN_MISMATCH = 0x12,
+	/* NPC_ERRLVL_NIX errcodes */
+	ERRCODE_OL3_LEN = 0x10,
+	ERRCODE_OL4_LEN = 0x11,
+	ERRCODE_OL4_CSUM = 0x12,
+	ERRCODE_IL3_LEN = 0x20,
+	ERRCODE_IL4_LEN = 0x21,
+	ERRCODE_IL4_CSUM = 0x22,
+};
+
+/* NIX TX stats */
+enum nix_stat_lf_tx {
+	TX_UCAST	= 0x0,
+	TX_BCAST	= 0x1,
+	TX_MCAST	= 0x2,
+	TX_DROP		= 0x3,
+	TX_OCTS		= 0x4,
+	TX_STATS_ENUM_LAST,
+};
+
+/* NIX RX stats */
+enum nix_stat_lf_rx {
+	RX_OCTS		= 0x0,
+	RX_UCAST	= 0x1,
+	RX_BCAST	= 0x2,
+	RX_MCAST	= 0x3,
+	RX_DROP		= 0x4,
+	RX_DROP_OCTS	= 0x5,
+	RX_FCS		= 0x6,
+	RX_ERR		= 0x7,
+	RX_DRP_BCAST	= 0x8,
+	RX_DRP_MCAST	= 0x9,
+	RX_DRP_L3BCAST	= 0xa,
+	RX_DRP_L3MCAST	= 0xb,
+	RX_STATS_ENUM_LAST,
+};
+
+struct otx2_dev_stats {
+	u64 rx_bytes;
+	u64 rx_frames;
+	u64 rx_ucast_frames;
+	u64 rx_bcast_frames;
+	u64 rx_mcast_frames;
+	u64 rx_drops;
+
+	u64 tx_bytes;
+	u64 tx_frames;
+	u64 tx_ucast_frames;
+	u64 tx_bcast_frames;
+	u64 tx_mcast_frames;
+	u64 tx_drops;
+};
+
+/* Driver counted stats */
+struct otx2_drv_stats {
+	atomic_t rx_fcs_errs;
+	atomic_t rx_oversize_errs;
+	atomic_t rx_undersize_errs;
+	atomic_t rx_csum_errs;
+	atomic_t rx_len_errs;
+	atomic_t rx_other_errs;
+};
+
+struct mbox {
+	struct otx2_mbox	mbox;
+	struct work_struct	mbox_wrk;
+	struct otx2_mbox	mbox_up;
+	struct work_struct	mbox_up_wrk;
+	struct otx2_nic		*pfvf;
+	void			*bbuf_base; /* Bounce buffer for mbox memory */
+	struct mutex		lock;	/* serialize mailbox access */
+	int			num_msgs; /* mbox number of messages */
+	int			up_num_msgs; /* mbox_up number of messages */
+};
+
+struct otx2_hw {
+	struct pci_dev		*pdev;
+	struct otx2_rss_info	rss_info;
+	u16                     rx_queues;
+	u16                     tx_queues;
+	u16			max_queues;
+	u16			pool_cnt;
+	u16			rqpool_cnt;
+	u16			sqpool_cnt;
+
+	/* NPA */
+	u32			stack_pg_ptrs;  /* No of ptrs per stack page */
+	u32			stack_pg_bytes; /* Size of stack page */
+	u16			sqb_size;
+
+	/* NIX */
+	u16		txschq_list[NIX_TXSCH_LVL_CNT][MAX_TXSCHQ_PER_FUNC];
+
+	/* HW settings, coalescing etc */
+	u16			rx_chan_base;
+	u16			tx_chan_base;
+	u16			cq_qcount_wait;
+	u16			cq_ecount_wait;
+	u16			rq_skid;
+	u8			cq_time_wait;
+
+	/* For TSO segmentation */
+	u8			lso_tsov4_idx;
+	u8			lso_tsov6_idx;
+	u8			hw_tso;
+
+	/* MSI-X */
+	u8			cint_cnt; /* CQ interrupt count */
+	u16			npa_msixoff; /* Offset of NPA vectors */
+	u16			nix_msixoff; /* Offset of NIX vectors */
+	char			*irq_name;
+	cpumask_var_t           *affinity_mask;
+
+	/* Stats */
+	struct otx2_dev_stats	dev_stats;
+	struct otx2_drv_stats	drv_stats;
+	u64			cgx_rx_stats[CGX_RX_STATS_COUNT];
+	u64			cgx_tx_stats[CGX_TX_STATS_COUNT];
+};
+
+struct refill_work {
+	struct delayed_work pool_refill_work;
+	struct otx2_nic *pf;
+};
+
+struct otx2_nic {
+	void __iomem		*reg_base;
+	struct net_device	*netdev;
+	void			*iommu_domain;
+	u16			max_frs;
+	u16			rbsize; /* Receive buffer size */
+
+#define OTX2_FLAG_INTF_DOWN			BIT_ULL(2)
+	u64			flags;
+
+	struct otx2_qset	qset;
+	struct otx2_hw		hw;
+	struct pci_dev		*pdev;
+	struct device		*dev;
+
+	/* Mbox */
+	struct mbox		mbox;
+	struct workqueue_struct *mbox_wq;
+
+	u16			pcifunc; /* RVU PF_FUNC */
+	struct cgx_link_user_info linfo;
+
+	u64			reset_count;
+	struct work_struct	reset_task;
+	struct refill_work	*refill_wrk;
+
+	/* Ethtool stuff */
+	u32			msg_enable;
+
+	/* Block address of NIX either BLKADDR_NIX0 or BLKADDR_NIX1 */
+	int			nix_blkaddr;
+};
+
+static inline bool is_96xx_A0(struct pci_dev *pdev)
+{
+	return (pdev->revision == 0x00) &&
+		(pdev->subsystem_device == PCI_SUBSYS_DEVID_96XX_RVU_PFVF);
+}
+
+static inline bool is_96xx_B0(struct pci_dev *pdev)
+{
+	return (pdev->revision == 0x01) &&
+		(pdev->subsystem_device == PCI_SUBSYS_DEVID_96XX_RVU_PFVF);
+}
+
+static inline void otx2_setup_dev_hw_settings(struct otx2_nic *pfvf)
+{
+	struct otx2_hw *hw = &pfvf->hw;
+
+	pfvf->hw.cq_time_wait = CQ_TIMER_THRESH_DEFAULT;
+	pfvf->hw.cq_ecount_wait = CQ_CQE_THRESH_DEFAULT;
+	pfvf->hw.cq_qcount_wait = CQ_QCOUNT_DEFAULT;
+
+	hw->hw_tso = true;
+
+	if (is_96xx_A0(pfvf->pdev)) {
+		hw->hw_tso = false;
+
+		/* Time based irq coalescing is not supported */
+		pfvf->hw.cq_qcount_wait = 0x0;
+
+		/* Due to HW issue previous silicons required minimum
+		 * 600 unused CQE to avoid CQ overflow.
+		 */
+		pfvf->hw.rq_skid = 600;
+		pfvf->qset.rqe_cnt = Q_COUNT(Q_SIZE_1K);
+	}
+}
+
+/* Register read/write APIs */
+static inline void __iomem *otx2_get_regaddr(struct otx2_nic *nic, u64 offset)
+{
+	u64 blkaddr;
+
+	switch ((offset >> RVU_FUNC_BLKADDR_SHIFT) & RVU_FUNC_BLKADDR_MASK) {
+	case BLKTYPE_NIX:
+		blkaddr = nic->nix_blkaddr;
+		break;
+	case BLKTYPE_NPA:
+		blkaddr = BLKADDR_NPA;
+		break;
+	default:
+		blkaddr = BLKADDR_RVUM;
+		break;
+	};
+
+	offset &= ~(RVU_FUNC_BLKADDR_MASK << RVU_FUNC_BLKADDR_SHIFT);
+	offset |= (blkaddr << RVU_FUNC_BLKADDR_SHIFT);
+
+	return nic->reg_base + offset;
+}
+
+static inline void otx2_write64(struct otx2_nic *nic, u64 offset, u64 val)
+{
+	void __iomem *addr = otx2_get_regaddr(nic, offset);
+
+	writeq(val, addr);
+}
+
+static inline u64 otx2_read64(struct otx2_nic *nic, u64 offset)
+{
+	void __iomem *addr = otx2_get_regaddr(nic, offset);
+
+	return readq(addr);
+}
+
+/* Mbox bounce buffer APIs */
+static inline int otx2_mbox_bbuf_init(struct mbox *mbox, struct pci_dev *pdev)
+{
+	struct otx2_mbox *otx2_mbox;
+	struct otx2_mbox_dev *mdev;
+
+	mbox->bbuf_base = devm_kmalloc(&pdev->dev, MBOX_SIZE, GFP_KERNEL);
+	if (!mbox->bbuf_base)
+		return -ENOMEM;
+
+	/* Overwrite mbox mbase to point to bounce buffer, so that PF/VF
+	 * prepare all mbox messages in bounce buffer instead of directly
+	 * in hw mbox memory.
+	 */
+	otx2_mbox = &mbox->mbox;
+	mdev = &otx2_mbox->dev[0];
+	mdev->mbase = mbox->bbuf_base;
+
+	otx2_mbox = &mbox->mbox_up;
+	mdev = &otx2_mbox->dev[0];
+	mdev->mbase = mbox->bbuf_base;
+	return 0;
+}
+
+static inline void otx2_sync_mbox_bbuf(struct otx2_mbox *mbox, int devid)
+{
+	u16 msgs_offset = ALIGN(sizeof(struct mbox_hdr), MBOX_MSG_ALIGN);
+	void *hw_mbase = mbox->hwbase + (devid * MBOX_SIZE);
+	struct otx2_mbox_dev *mdev = &mbox->dev[devid];
+	struct mbox_hdr *hdr;
+	u64 msg_size;
+
+	if (mdev->mbase == hw_mbase)
+		return;
+
+	hdr = hw_mbase + mbox->rx_start;
+	msg_size = hdr->msg_size;
+
+	if (msg_size > mbox->rx_size - msgs_offset)
+		msg_size = mbox->rx_size - msgs_offset;
+
+	/* Copy mbox messages from mbox memory to bounce buffer */
+	memcpy(mdev->mbase + mbox->rx_start,
+	       hw_mbase + mbox->rx_start, msg_size + msgs_offset);
+}
+
+static inline void otx2_mbox_lock_init(struct mbox *mbox)
+{
+	mutex_init(&mbox->lock);
+}
+
+static inline void otx2_mbox_lock(struct mbox *mbox)
+{
+	mutex_lock(&mbox->lock);
+}
+
+static inline void otx2_mbox_unlock(struct mbox *mbox)
+{
+	mutex_unlock(&mbox->lock);
+}
+
+/* With the absence of API for 128-bit IO memory access for arm64,
+ * implement required operations at place.
+ */
+#if defined(CONFIG_ARM64)
+static inline void otx2_write128(u64 lo, u64 hi, void __iomem *addr)
+{
+	__asm__ volatile("stp %x[x0], %x[x1], [%x[p1],#0]!"
+			 ::[x0]"r"(lo), [x1]"r"(hi), [p1]"r"(addr));
+}
+
+static inline u64 otx2_atomic64_add(u64 incr, u64 *ptr)
+{
+	u64 result;
+
+	__asm__ volatile(".cpu   generic+lse\n"
+			 "ldadd %x[i], %x[r], [%[b]]"
+			 : [r]"=r"(result), "+m"(*ptr)
+			 : [i]"r"(incr), [b]"r"(ptr)
+			 : "memory");
+	return result;
+}
+
+static inline u64 otx2_lmt_flush(uint64_t addr)
+{
+	u64 result = 0;
+
+	__asm__ volatile(".cpu  generic+lse\n"
+			 "ldeor xzr,%x[rf],[%[rs]]"
+			 : [rf]"=r"(result)
+			 : [rs]"r"(addr));
+	return result;
+}
+
+#else
+#define otx2_write128(lo, hi, addr)
+#define otx2_atomic64_add(incr, ptr)		({ *ptr += incr; })
+#define otx2_lmt_flush(addr)			({ 0; })
+#endif
+
+/* Alloc pointer from pool/aura */
+static inline u64 otx2_aura_allocptr(struct otx2_nic *pfvf, int aura)
+{
+	u64 *ptr = (u64 *)otx2_get_regaddr(pfvf,
+			   NPA_LF_AURA_OP_ALLOCX(0));
+	u64 incr = (u64)aura | BIT_ULL(63);
+
+	return otx2_atomic64_add(incr, ptr);
+}
+
+/* Free pointer to a pool/aura */
+static inline void otx2_aura_freeptr(struct otx2_nic *pfvf,
+				     int aura, s64 buf)
+{
+	otx2_write128((u64)buf, (u64)aura | BIT_ULL(63),
+		      otx2_get_regaddr(pfvf, NPA_LF_AURA_OP_FREE0));
+}
+
+/* Update page ref count */
+static inline void otx2_get_page(struct otx2_pool *pool)
+{
+	if (!pool->page)
+		return;
+
+	if (pool->pageref)
+		page_ref_add(pool->page, pool->pageref);
+	pool->pageref = 0;
+	pool->page = NULL;
+}
+
+static inline int otx2_get_pool_idx(struct otx2_nic *pfvf, int type, int idx)
+{
+	if (type == AURA_NIX_SQ)
+		return pfvf->hw.rqpool_cnt + idx;
+
+	 /* AURA_NIX_RQ */
+	return idx;
+}
+
+/* Mbox APIs */
+static inline int otx2_sync_mbox_msg(struct mbox *mbox)
+{
+	int err;
+
+	if (!otx2_mbox_nonempty(&mbox->mbox, 0))
+		return 0;
+	otx2_mbox_msg_send(&mbox->mbox, 0);
+	err = otx2_mbox_wait_for_rsp(&mbox->mbox, 0);
+	if (err)
+		return err;
+
+	return otx2_mbox_check_rsp_msgs(&mbox->mbox, 0);
+}
+
+static inline int otx2_sync_mbox_up_msg(struct mbox *mbox, int devid)
+{
+	int err;
+
+	if (!otx2_mbox_nonempty(&mbox->mbox_up, devid))
+		return 0;
+	otx2_mbox_msg_send(&mbox->mbox_up, devid);
+	err = otx2_mbox_wait_for_rsp(&mbox->mbox_up, devid);
+	if (err)
+		return err;
+
+	return otx2_mbox_check_rsp_msgs(&mbox->mbox_up, devid);
+}
+
+/* Use this API to send mbox msgs in atomic context
+ * where sleeping is not allowed
+ */
+static inline int otx2_sync_mbox_msg_busy_poll(struct mbox *mbox)
+{
+	int err;
+
+	if (!otx2_mbox_nonempty(&mbox->mbox, 0))
+		return 0;
+	otx2_mbox_msg_send(&mbox->mbox, 0);
+	err = otx2_mbox_busy_poll_for_rsp(&mbox->mbox, 0);
+	if (err)
+		return err;
+
+	return otx2_mbox_check_rsp_msgs(&mbox->mbox, 0);
+}
+
+#define M(_name, _id, _fn_name, _req_type, _rsp_type)                   \
+static struct _req_type __maybe_unused					\
+*otx2_mbox_alloc_msg_ ## _fn_name(struct mbox *mbox)                    \
+{									\
+	struct _req_type *req;						\
+									\
+	req = (struct _req_type *)otx2_mbox_alloc_msg_rsp(		\
+		&mbox->mbox, 0, sizeof(struct _req_type),		\
+		sizeof(struct _rsp_type));				\
+	if (!req)							\
+		return NULL;						\
+	req->hdr.sig = OTX2_MBOX_REQ_SIG;				\
+	req->hdr.id = _id;						\
+	return req;							\
+}
+
+MBOX_MESSAGES
+#undef M
+
+#define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
+int									\
+otx2_mbox_up_handler_ ## _fn_name(struct otx2_nic *pfvf,		\
+				struct _req_type *req,			\
+				struct _rsp_type *rsp);			\
+
+MBOX_UP_CGX_MESSAGES
+#undef M
+
+/* Time to wait before watchdog kicks off */
+#define OTX2_TX_TIMEOUT		(100 * HZ)
+
+#define	RVU_PFVF_PF_SHIFT	10
+#define	RVU_PFVF_PF_MASK	0x3F
+#define	RVU_PFVF_FUNC_SHIFT	0
+#define	RVU_PFVF_FUNC_MASK	0x3FF
+
+static inline int rvu_get_pf(u16 pcifunc)
+{
+	return (pcifunc >> RVU_PFVF_PF_SHIFT) & RVU_PFVF_PF_MASK;
+}
+
+static inline dma_addr_t otx2_dma_map_page(struct otx2_nic *pfvf,
+					   struct page *page,
+					   size_t offset, size_t size,
+					   enum dma_data_direction dir)
+{
+	dma_addr_t iova;
+
+	iova = dma_map_page_attrs(pfvf->dev, page,
+				  offset, size, dir, DMA_ATTR_SKIP_CPU_SYNC);
+	if (unlikely(dma_mapping_error(pfvf->dev, iova)))
+		return (dma_addr_t)NULL;
+	return iova;
+}
+
+static inline void otx2_dma_unmap_page(struct otx2_nic *pfvf,
+				       dma_addr_t addr, size_t size,
+				       enum dma_data_direction dir)
+{
+	dma_unmap_page_attrs(pfvf->dev, addr, size,
+			     dir, DMA_ATTR_SKIP_CPU_SYNC);
+}
+
+/* MSI-X APIs */
+void otx2_free_cints(struct otx2_nic *pfvf, int n);
+void otx2_set_cints_affinity(struct otx2_nic *pfvf);
+int otx2_set_mac_address(struct net_device *netdev, void *p);
+int otx2_hw_set_mtu(struct otx2_nic *pfvf, int mtu);
+void otx2_tx_timeout(struct net_device *netdev, unsigned int txq);
+void otx2_get_mac_from_af(struct net_device *netdev);
+void otx2_config_irq_coalescing(struct otx2_nic *pfvf, int qidx);
+
+/* RVU block related APIs */
+int otx2_attach_npa_nix(struct otx2_nic *pfvf);
+int otx2_detach_resources(struct mbox *mbox);
+int otx2_config_npa(struct otx2_nic *pfvf);
+int otx2_sq_aura_pool_init(struct otx2_nic *pfvf);
+int otx2_rq_aura_pool_init(struct otx2_nic *pfvf);
+void otx2_aura_pool_free(struct otx2_nic *pfvf);
+void otx2_free_aura_ptr(struct otx2_nic *pfvf, int type);
+void otx2_sq_free_sqbs(struct otx2_nic *pfvf);
+int otx2_config_nix(struct otx2_nic *pfvf);
+int otx2_config_nix_queues(struct otx2_nic *pfvf);
+int otx2_txschq_config(struct otx2_nic *pfvf, int lvl);
+int otx2_txsch_alloc(struct otx2_nic *pfvf);
+int otx2_txschq_stop(struct otx2_nic *pfvf);
+void otx2_sqb_flush(struct otx2_nic *pfvf);
+dma_addr_t otx2_alloc_rbuf(struct otx2_nic *pfvf, struct otx2_pool *pool,
+			   gfp_t gfp);
+int otx2_rxtx_enable(struct otx2_nic *pfvf, bool enable);
+void otx2_ctx_disable(struct mbox *mbox, int type, bool npa);
+void otx2_cleanup_rx_cqes(struct otx2_nic *pfvf, struct otx2_cq_queue *cq);
+void otx2_cleanup_tx_cqes(struct otx2_nic *pfvf, struct otx2_cq_queue *cq);
+
+/* RSS configuration APIs*/
+int otx2_rss_init(struct otx2_nic *pfvf);
+int otx2_set_flowkey_cfg(struct otx2_nic *pfvf);
+void otx2_set_rss_key(struct otx2_nic *pfvf);
+int otx2_set_rss_table(struct otx2_nic *pfvf);
+
+/* Mbox handlers */
+void mbox_handler_msix_offset(struct otx2_nic *pfvf,
+			      struct msix_offset_rsp *rsp);
+void mbox_handler_npa_lf_alloc(struct otx2_nic *pfvf,
+			       struct npa_lf_alloc_rsp *rsp);
+void mbox_handler_nix_lf_alloc(struct otx2_nic *pfvf,
+			       struct nix_lf_alloc_rsp *rsp);
+void mbox_handler_nix_txsch_alloc(struct otx2_nic *pf,
+				  struct nix_txsch_alloc_rsp *rsp);
+void mbox_handler_cgx_stats(struct otx2_nic *pfvf,
+			    struct cgx_stats_rsp *rsp);
+
+/* Device stats APIs */
+void otx2_get_dev_stats(struct otx2_nic *pfvf);
+void otx2_get_stats64(struct net_device *netdev,
+		      struct rtnl_link_stats64 *stats);
+void otx2_update_lmac_stats(struct otx2_nic *pfvf);
+int otx2_update_rq_stats(struct otx2_nic *pfvf, int qidx);
+int otx2_update_sq_stats(struct otx2_nic *pfvf, int qidx);
+void otx2_set_ethtool_ops(struct net_device *netdev);
+
+int otx2_open(struct net_device *netdev);
+int otx2_stop(struct net_device *netdev);
+int otx2_set_real_num_queues(struct net_device *netdev,
+			     int tx_queues, int rx_queues);
+#endif /* OTX2_COMMON_H */

drivers/net/ethernet/marvell/octeontx2/nic/otx2_ethtool.c

+// SPDX-License-Identifier: GPL-2.0
+/* Marvell OcteonTx2 RVU Ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/pci.h>
+#include <linux/ethtool.h>
+#include <linux/stddef.h>
+#include <linux/etherdevice.h>
+#include <linux/log2.h>
+
+#include "otx2_common.h"
+
+#define DRV_NAME	"octeontx2-nicpf"
+
+struct otx2_stat {
+	char name[ETH_GSTRING_LEN];
+	unsigned int index;
+};
+
+/* HW device stats */
+#define OTX2_DEV_STAT(stat) { \
+	.name = #stat, \
+	.index = offsetof(struct otx2_dev_stats, stat) / sizeof(u64), \
+}
+
+static const struct otx2_stat otx2_dev_stats[] = {
+	OTX2_DEV_STAT(rx_ucast_frames),
+	OTX2_DEV_STAT(rx_bcast_frames),
+	OTX2_DEV_STAT(rx_mcast_frames),
+
+	OTX2_DEV_STAT(tx_ucast_frames),
+	OTX2_DEV_STAT(tx_bcast_frames),
+	OTX2_DEV_STAT(tx_mcast_frames),
+};
+
+/* Driver level stats */
+#define OTX2_DRV_STAT(stat) { \
+	.name = #stat, \
+	.index = offsetof(struct otx2_drv_stats, stat) / sizeof(atomic_t), \
+}
+
+static const struct otx2_stat otx2_drv_stats[] = {
+	OTX2_DRV_STAT(rx_fcs_errs),
+	OTX2_DRV_STAT(rx_oversize_errs),
+	OTX2_DRV_STAT(rx_undersize_errs),
+	OTX2_DRV_STAT(rx_csum_errs),
+	OTX2_DRV_STAT(rx_len_errs),
+	OTX2_DRV_STAT(rx_other_errs),
+};
+
+static const struct otx2_stat otx2_queue_stats[] = {
+	{ "bytes", 0 },
+	{ "frames", 1 },
+};
+
+static const unsigned int otx2_n_dev_stats = ARRAY_SIZE(otx2_dev_stats);
+static const unsigned int otx2_n_drv_stats = ARRAY_SIZE(otx2_drv_stats);
+static const unsigned int otx2_n_queue_stats = ARRAY_SIZE(otx2_queue_stats);
+
+static void otx2_dev_open(struct net_device *netdev)
+{
+	otx2_open(netdev);
+}
+
+static void otx2_dev_stop(struct net_device *netdev)
+{
+	otx2_stop(netdev);
+}
+
+static void otx2_get_drvinfo(struct net_device *netdev,
+			     struct ethtool_drvinfo *info)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+
+	strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
+	strlcpy(info->bus_info, pci_name(pfvf->pdev), sizeof(info->bus_info));
+}
+
+static void otx2_get_qset_strings(struct otx2_nic *pfvf, u8 **data, int qset)
+{
+	int start_qidx = qset * pfvf->hw.rx_queues;
+	int qidx, stats;
+
+	for (qidx = 0; qidx < pfvf->hw.rx_queues; qidx++) {
+		for (stats = 0; stats < otx2_n_queue_stats; stats++) {
+			sprintf(*data, "rxq%d: %s", qidx + start_qidx,
+				otx2_queue_stats[stats].name);
+			*data += ETH_GSTRING_LEN;
+		}
+	}
+	for (qidx = 0; qidx < pfvf->hw.tx_queues; qidx++) {
+		for (stats = 0; stats < otx2_n_queue_stats; stats++) {
+			sprintf(*data, "txq%d: %s", qidx + start_qidx,
+				otx2_queue_stats[stats].name);
+			*data += ETH_GSTRING_LEN;
+		}
+	}
+}
+
+static void otx2_get_strings(struct net_device *netdev, u32 sset, u8 *data)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	int stats;
+
+	if (sset != ETH_SS_STATS)
+		return;
+
+	for (stats = 0; stats < otx2_n_dev_stats; stats++) {
+		memcpy(data, otx2_dev_stats[stats].name, ETH_GSTRING_LEN);
+		data += ETH_GSTRING_LEN;
+	}
+
+	for (stats = 0; stats < otx2_n_drv_stats; stats++) {
+		memcpy(data, otx2_drv_stats[stats].name, ETH_GSTRING_LEN);
+		data += ETH_GSTRING_LEN;
+	}
+
+	otx2_get_qset_strings(pfvf, &data, 0);
+
+	for (stats = 0; stats < CGX_RX_STATS_COUNT; stats++) {
+		sprintf(data, "cgx_rxstat%d: ", stats);
+		data += ETH_GSTRING_LEN;
+	}
+
+	for (stats = 0; stats < CGX_TX_STATS_COUNT; stats++) {
+		sprintf(data, "cgx_txstat%d: ", stats);
+		data += ETH_GSTRING_LEN;
+	}
+
+	strcpy(data, "reset_count");
+	data += ETH_GSTRING_LEN;
+}
+
+static void otx2_get_qset_stats(struct otx2_nic *pfvf,
+				struct ethtool_stats *stats, u64 **data)
+{
+	int stat, qidx;
+
+	if (!pfvf)
+		return;
+	for (qidx = 0; qidx < pfvf->hw.rx_queues; qidx++) {
+		if (!otx2_update_rq_stats(pfvf, qidx)) {
+			for (stat = 0; stat < otx2_n_queue_stats; stat++)
+				*((*data)++) = 0;
+			continue;
+		}
+		for (stat = 0; stat < otx2_n_queue_stats; stat++)
+			*((*data)++) = ((u64 *)&pfvf->qset.rq[qidx].stats)
+				[otx2_queue_stats[stat].index];
+	}
+
+	for (qidx = 0; qidx < pfvf->hw.tx_queues; qidx++) {
+		if (!otx2_update_sq_stats(pfvf, qidx)) {
+			for (stat = 0; stat < otx2_n_queue_stats; stat++)
+				*((*data)++) = 0;
+			continue;
+		}
+		for (stat = 0; stat < otx2_n_queue_stats; stat++)
+			*((*data)++) = ((u64 *)&pfvf->qset.sq[qidx].stats)
+				[otx2_queue_stats[stat].index];
+	}
+}
+
+/* Get device and per queue statistics */
+static void otx2_get_ethtool_stats(struct net_device *netdev,
+				   struct ethtool_stats *stats, u64 *data)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	int stat;
+
+	otx2_get_dev_stats(pfvf);
+	for (stat = 0; stat < otx2_n_dev_stats; stat++)
+		*(data++) = ((u64 *)&pfvf->hw.dev_stats)
+				[otx2_dev_stats[stat].index];
+
+	for (stat = 0; stat < otx2_n_drv_stats; stat++)
+		*(data++) = atomic_read(&((atomic_t *)&pfvf->hw.drv_stats)
+						[otx2_drv_stats[stat].index]);
+
+	otx2_get_qset_stats(pfvf, stats, &data);
+	otx2_update_lmac_stats(pfvf);
+	for (stat = 0; stat < CGX_RX_STATS_COUNT; stat++)
+		*(data++) = pfvf->hw.cgx_rx_stats[stat];
+	for (stat = 0; stat < CGX_TX_STATS_COUNT; stat++)
+		*(data++) = pfvf->hw.cgx_tx_stats[stat];
+	*(data++) = pfvf->reset_count;
+}
+
+static int otx2_get_sset_count(struct net_device *netdev, int sset)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	int qstats_count;
+
+	if (sset != ETH_SS_STATS)
+		return -EINVAL;
+
+	qstats_count = otx2_n_queue_stats *
+		       (pfvf->hw.rx_queues + pfvf->hw.tx_queues);
+
+	return otx2_n_dev_stats + otx2_n_drv_stats + qstats_count +
+		CGX_RX_STATS_COUNT + CGX_TX_STATS_COUNT + 1;
+}
+
+/* Get no of queues device supports and current queue count */
+static void otx2_get_channels(struct net_device *dev,
+			      struct ethtool_channels *channel)
+{
+	struct otx2_nic *pfvf = netdev_priv(dev);
+
+	channel->max_rx = pfvf->hw.max_queues;
+	channel->max_tx = pfvf->hw.max_queues;
+
+	channel->rx_count = pfvf->hw.rx_queues;
+	channel->tx_count = pfvf->hw.tx_queues;
+}
+
+/* Set no of Tx, Rx queues to be used */
+static int otx2_set_channels(struct net_device *dev,
+			     struct ethtool_channels *channel)
+{
+	struct otx2_nic *pfvf = netdev_priv(dev);
+	bool if_up = netif_running(dev);
+	int err = 0;
+
+	if (!channel->rx_count || !channel->tx_count)
+		return -EINVAL;
+
+	if (if_up)
+		otx2_dev_stop(dev);
+
+	err = otx2_set_real_num_queues(dev, channel->tx_count,
+				       channel->rx_count);
+	if (err)
+		goto fail;
+
+	pfvf->hw.rx_queues = channel->rx_count;
+	pfvf->hw.tx_queues = channel->tx_count;
+	pfvf->qset.cq_cnt = pfvf->hw.tx_queues +  pfvf->hw.rx_queues;
+
+fail:
+	if (if_up)
+		otx2_dev_open(dev);
+
+	netdev_info(dev, "Setting num Tx rings to %d, Rx rings to %d success\n",
+		    pfvf->hw.tx_queues, pfvf->hw.rx_queues);
+
+	return err;
+}
+
+static void otx2_get_ringparam(struct net_device *netdev,
+			       struct ethtool_ringparam *ring)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	struct otx2_qset *qs = &pfvf->qset;
+
+	ring->rx_max_pending = Q_COUNT(Q_SIZE_MAX);
+	ring->rx_pending = qs->rqe_cnt ? qs->rqe_cnt : Q_COUNT(Q_SIZE_256);
+	ring->tx_max_pending = Q_COUNT(Q_SIZE_MAX);
+	ring->tx_pending = qs->sqe_cnt ? qs->sqe_cnt : Q_COUNT(Q_SIZE_4K);
+}
+
+static int otx2_set_ringparam(struct net_device *netdev,
+			      struct ethtool_ringparam *ring)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	bool if_up = netif_running(netdev);
+	struct otx2_qset *qs = &pfvf->qset;
+	u32 rx_count, tx_count;
+
+	if (ring->rx_mini_pending || ring->rx_jumbo_pending)
+		return -EINVAL;
+
+	/* Permitted lengths are 16 64 256 1K 4K 16K 64K 256K 1M  */
+	rx_count = ring->rx_pending;
+	/* On some silicon variants a skid or reserved CQEs are
+	 * needed to avoid CQ overflow.
+	 */
+	if (rx_count < pfvf->hw.rq_skid)
+		rx_count =  pfvf->hw.rq_skid;
+	rx_count = Q_COUNT(Q_SIZE(rx_count, 3));
+
+	/* Due pipelining impact minimum 2000 unused SQ CQE's
+	 * need to be maintained to avoid CQ overflow, hence the
+	 * minimum 4K size.
+	 */
+	tx_count = clamp_t(u32, ring->tx_pending,
+			   Q_COUNT(Q_SIZE_4K), Q_COUNT(Q_SIZE_MAX));
+	tx_count = Q_COUNT(Q_SIZE(tx_count, 3));
+
+	if (tx_count == qs->sqe_cnt && rx_count == qs->rqe_cnt)
+		return 0;
+
+	if (if_up)
+		otx2_dev_stop(netdev);
+
+	/* Assigned to the nearest possible exponent. */
+	qs->sqe_cnt = tx_count;
+	qs->rqe_cnt = rx_count;
+
+	if (if_up)
+		otx2_dev_open(netdev);
+	return 0;
+}
+
+static int otx2_get_coalesce(struct net_device *netdev,
+			     struct ethtool_coalesce *cmd)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	struct otx2_hw *hw = &pfvf->hw;
+
+	cmd->rx_coalesce_usecs = hw->cq_time_wait;
+	cmd->rx_max_coalesced_frames = hw->cq_ecount_wait;
+	cmd->tx_coalesce_usecs = hw->cq_time_wait;
+	cmd->tx_max_coalesced_frames = hw->cq_ecount_wait;
+
+	return 0;
+}
+
+static int otx2_set_coalesce(struct net_device *netdev,
+			     struct ethtool_coalesce *ec)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	struct otx2_hw *hw = &pfvf->hw;
+	int qidx;
+
+	if (ec->use_adaptive_rx_coalesce || ec->use_adaptive_tx_coalesce ||
+	    ec->rx_coalesce_usecs_irq || ec->rx_max_coalesced_frames_irq ||
+	    ec->tx_coalesce_usecs_irq || ec->tx_max_coalesced_frames_irq ||
+	    ec->stats_block_coalesce_usecs || ec->pkt_rate_low ||
+	    ec->rx_coalesce_usecs_low || ec->rx_max_coalesced_frames_low ||
+	    ec->tx_coalesce_usecs_low || ec->tx_max_coalesced_frames_low ||
+	    ec->pkt_rate_high || ec->rx_coalesce_usecs_high ||
+	    ec->rx_max_coalesced_frames_high || ec->tx_coalesce_usecs_high ||
+	    ec->tx_max_coalesced_frames_high || ec->rate_sample_interval)
+		return -EOPNOTSUPP;
+
+	if (!ec->rx_max_coalesced_frames || !ec->tx_max_coalesced_frames)
+		return 0;
+
+	/* 'cq_time_wait' is 8bit and is in multiple of 100ns,
+	 * so clamp the user given value to the range of 1 to 25usec.
+	 */
+	ec->rx_coalesce_usecs = clamp_t(u32, ec->rx_coalesce_usecs,
+					1, CQ_TIMER_THRESH_MAX);
+	ec->tx_coalesce_usecs = clamp_t(u32, ec->tx_coalesce_usecs,
+					1, CQ_TIMER_THRESH_MAX);
+
+	/* Rx and Tx are mapped to same CQ, check which one
+	 * is changed, if both then choose the min.
+	 */
+	if (hw->cq_time_wait == ec->rx_coalesce_usecs)
+		hw->cq_time_wait = ec->tx_coalesce_usecs;
+	else if (hw->cq_time_wait == ec->tx_coalesce_usecs)
+		hw->cq_time_wait = ec->rx_coalesce_usecs;
+	else
+		hw->cq_time_wait = min_t(u8, ec->rx_coalesce_usecs,
+					 ec->tx_coalesce_usecs);
+
+	/* Max ecount_wait supported is 16bit,
+	 * so clamp the user given value to the range of 1 to 64k.
+	 */
+	ec->rx_max_coalesced_frames = clamp_t(u32, ec->rx_max_coalesced_frames,
+					      1, U16_MAX);
+	ec->tx_max_coalesced_frames = clamp_t(u32, ec->tx_max_coalesced_frames,
+					      1, U16_MAX);
+
+	/* Rx and Tx are mapped to same CQ, check which one
+	 * is changed, if both then choose the min.
+	 */
+	if (hw->cq_ecount_wait == ec->rx_max_coalesced_frames)
+		hw->cq_ecount_wait = ec->tx_max_coalesced_frames;
+	else if (hw->cq_ecount_wait == ec->tx_max_coalesced_frames)
+		hw->cq_ecount_wait = ec->rx_max_coalesced_frames;
+	else
+		hw->cq_ecount_wait = min_t(u16, ec->rx_max_coalesced_frames,
+					   ec->tx_max_coalesced_frames);
+
+	if (netif_running(netdev)) {
+		for (qidx = 0; qidx < pfvf->hw.cint_cnt; qidx++)
+			otx2_config_irq_coalescing(pfvf, qidx);
+	}
+
+	return 0;
+}
+
+static int otx2_get_rss_hash_opts(struct otx2_nic *pfvf,
+				  struct ethtool_rxnfc *nfc)
+{
+	struct otx2_rss_info *rss = &pfvf->hw.rss_info;
+
+	if (!(rss->flowkey_cfg &
+	    (NIX_FLOW_KEY_TYPE_IPV4 | NIX_FLOW_KEY_TYPE_IPV6)))
+		return 0;
+
+	/* Mimimum is IPv4 and IPv6, SIP/DIP */
+	nfc->data = RXH_IP_SRC | RXH_IP_DST;
+
+	switch (nfc->flow_type) {
+	case TCP_V4_FLOW:
+	case TCP_V6_FLOW:
+		if (rss->flowkey_cfg & NIX_FLOW_KEY_TYPE_TCP)
+			nfc->data |= RXH_L4_B_0_1 | RXH_L4_B_2_3;
+		break;
+	case UDP_V4_FLOW:
+	case UDP_V6_FLOW:
+		if (rss->flowkey_cfg & NIX_FLOW_KEY_TYPE_UDP)
+			nfc->data |= RXH_L4_B_0_1 | RXH_L4_B_2_3;
+		break;
+	case SCTP_V4_FLOW:
+	case SCTP_V6_FLOW:
+		if (rss->flowkey_cfg & NIX_FLOW_KEY_TYPE_SCTP)
+			nfc->data |= RXH_L4_B_0_1 | RXH_L4_B_2_3;
+		break;
+	case AH_ESP_V4_FLOW:
+	case AH_V4_FLOW:
+	case ESP_V4_FLOW:
+	case IPV4_FLOW:
+	case AH_ESP_V6_FLOW:
+	case AH_V6_FLOW:
+	case ESP_V6_FLOW:
+	case IPV6_FLOW:
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int otx2_set_rss_hash_opts(struct otx2_nic *pfvf,
+				  struct ethtool_rxnfc *nfc)
+{
+	struct otx2_rss_info *rss = &pfvf->hw.rss_info;
+	u32 rxh_l4 = RXH_L4_B_0_1 | RXH_L4_B_2_3;
+	u32 rss_cfg = rss->flowkey_cfg;
+
+	if (!rss->enable) {
+		netdev_err(pfvf->netdev,
+			   "RSS is disabled, cannot change settings\n");
+		return -EIO;
+	}
+
+	/* Mimimum is IPv4 and IPv6, SIP/DIP */
+	if (!(nfc->data & RXH_IP_SRC) || !(nfc->data & RXH_IP_DST))
+		return -EINVAL;
+
+	switch (nfc->flow_type) {
+	case TCP_V4_FLOW:
+	case TCP_V6_FLOW:
+		/* Different config for v4 and v6 is not supported.
+		 * Both of them have to be either 4-tuple or 2-tuple.
+		 */
+		switch (nfc->data & rxh_l4) {
+		case 0:
+			rss_cfg &= ~NIX_FLOW_KEY_TYPE_TCP;
+			break;
+		case (RXH_L4_B_0_1 | RXH_L4_B_2_3):
+			rss_cfg |= NIX_FLOW_KEY_TYPE_TCP;
+			break;
+		default:
+			return -EINVAL;
+		}
+		break;
+	case UDP_V4_FLOW:
+	case UDP_V6_FLOW:
+		switch (nfc->data & rxh_l4) {
+		case 0:
+			rss_cfg &= ~NIX_FLOW_KEY_TYPE_UDP;
+			break;
+		case (RXH_L4_B_0_1 | RXH_L4_B_2_3):
+			rss_cfg |= NIX_FLOW_KEY_TYPE_UDP;
+			break;
+		default:
+			return -EINVAL;
+		}
+		break;
+	case SCTP_V4_FLOW:
+	case SCTP_V6_FLOW:
+		switch (nfc->data & rxh_l4) {
+		case 0:
+			rss_cfg &= ~NIX_FLOW_KEY_TYPE_SCTP;
+			break;
+		case (RXH_L4_B_0_1 | RXH_L4_B_2_3):
+			rss_cfg |= NIX_FLOW_KEY_TYPE_SCTP;
+			break;
+		default:
+			return -EINVAL;
+		}
+		break;
+	case IPV4_FLOW:
+	case IPV6_FLOW:
+		rss_cfg = NIX_FLOW_KEY_TYPE_IPV4 | NIX_FLOW_KEY_TYPE_IPV6;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	rss->flowkey_cfg = rss_cfg;
+	otx2_set_flowkey_cfg(pfvf);
+	return 0;
+}
+
+static int otx2_get_rxnfc(struct net_device *dev,
+			  struct ethtool_rxnfc *nfc, u32 *rules)
+{
+	struct otx2_nic *pfvf = netdev_priv(dev);
+	int ret = -EOPNOTSUPP;
+
+	switch (nfc->cmd) {
+	case ETHTOOL_GRXRINGS:
+		nfc->data = pfvf->hw.rx_queues;
+		ret = 0;
+		break;
+	case ETHTOOL_GRXFH:
+		return otx2_get_rss_hash_opts(pfvf, nfc);
+	default:
+		break;
+	}
+	return ret;
+}
+
+static int otx2_set_rxnfc(struct net_device *dev, struct ethtool_rxnfc *nfc)
+{
+	struct otx2_nic *pfvf = netdev_priv(dev);
+	int ret = -EOPNOTSUPP;
+
+	switch (nfc->cmd) {
+	case ETHTOOL_SRXFH:
+		ret = otx2_set_rss_hash_opts(pfvf, nfc);
+		break;
+	default:
+		break;
+	}
+
+	return ret;
+}
+
+static u32 otx2_get_rxfh_key_size(struct net_device *netdev)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	struct otx2_rss_info *rss;
+
+	rss = &pfvf->hw.rss_info;
+
+	return sizeof(rss->key);
+}
+
+static u32 otx2_get_rxfh_indir_size(struct net_device *dev)
+{
+	struct otx2_nic *pfvf = netdev_priv(dev);
+
+	return pfvf->hw.rss_info.rss_size;
+}
+
+/* Get RSS configuration */
+static int otx2_get_rxfh(struct net_device *dev, u32 *indir,
+			 u8 *hkey, u8 *hfunc)
+{
+	struct otx2_nic *pfvf = netdev_priv(dev);
+	struct otx2_rss_info *rss;
+	int idx;
+
+	rss = &pfvf->hw.rss_info;
+
+	if (indir) {
+		for (idx = 0; idx < rss->rss_size; idx++)
+			indir[idx] = rss->ind_tbl[idx];
+	}
+
+	if (hkey)
+		memcpy(hkey, rss->key, sizeof(rss->key));
+
+	if (hfunc)
+		*hfunc = ETH_RSS_HASH_TOP;
+
+	return 0;
+}
+
+/* Configure RSS table and hash key */
+static int otx2_set_rxfh(struct net_device *dev, const u32 *indir,
+			 const u8 *hkey, const u8 hfunc)
+{
+	struct otx2_nic *pfvf = netdev_priv(dev);
+	struct otx2_rss_info *rss;
+	int idx;
+
+	if (hfunc != ETH_RSS_HASH_NO_CHANGE && hfunc != ETH_RSS_HASH_TOP)
+		return -EOPNOTSUPP;
+
+	rss = &pfvf->hw.rss_info;
+
+	if (!rss->enable) {
+		netdev_err(dev, "RSS is disabled, cannot change settings\n");
+		return -EIO;
+	}
+
+	if (indir) {
+		for (idx = 0; idx < rss->rss_size; idx++)
+			rss->ind_tbl[idx] = indir[idx];
+	}
+
+	if (hkey) {
+		memcpy(rss->key, hkey, sizeof(rss->key));
+		otx2_set_rss_key(pfvf);
+	}
+
+	otx2_set_rss_table(pfvf);
+	return 0;
+}
+
+static u32 otx2_get_msglevel(struct net_device *netdev)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+
+	return pfvf->msg_enable;
+}
+
+static void otx2_set_msglevel(struct net_device *netdev, u32 val)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+
+	pfvf->msg_enable = val;
+}
+
+static u32 otx2_get_link(struct net_device *netdev)
+{
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+
+	return pfvf->linfo.link_up;
+}
+
+static const struct ethtool_ops otx2_ethtool_ops = {
+	.get_link		= otx2_get_link,
+	.get_drvinfo		= otx2_get_drvinfo,
+	.get_strings		= otx2_get_strings,
+	.get_ethtool_stats	= otx2_get_ethtool_stats,
+	.get_sset_count		= otx2_get_sset_count,
+	.set_channels		= otx2_set_channels,
+	.get_channels		= otx2_get_channels,
+	.get_ringparam		= otx2_get_ringparam,
+	.set_ringparam		= otx2_set_ringparam,
+	.get_coalesce		= otx2_get_coalesce,
+	.set_coalesce		= otx2_set_coalesce,
+	.get_rxnfc		= otx2_get_rxnfc,
+	.set_rxnfc              = otx2_set_rxnfc,
+	.get_rxfh_key_size	= otx2_get_rxfh_key_size,
+	.get_rxfh_indir_size	= otx2_get_rxfh_indir_size,
+	.get_rxfh		= otx2_get_rxfh,
+	.set_rxfh		= otx2_set_rxfh,
+	.get_msglevel		= otx2_get_msglevel,
+	.set_msglevel		= otx2_set_msglevel,
+};
+
+void otx2_set_ethtool_ops(struct net_device *netdev)
+{
+	netdev->ethtool_ops = &otx2_ethtool_ops;
+}

drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c

+// SPDX-License-Identifier: GPL-2.0
+/* Marvell OcteonTx2 RVU Physcial Function ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/module.h>
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/etherdevice.h>
+#include <linux/of.h>
+#include <linux/if_vlan.h>
+#include <linux/iommu.h>
+#include <net/ip.h>
+
+#include "otx2_reg.h"
+#include "otx2_common.h"
+#include "otx2_txrx.h"
+#include "otx2_struct.h"
+
+#define DRV_NAME	"octeontx2-nicpf"
+#define DRV_STRING	"Marvell OcteonTX2 NIC Physical Function Driver"
+#define DRV_VERSION	"1.0"
+
+/* Supported devices */
+static const struct pci_device_id otx2_pf_id_table[] = {
+	{ PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, PCI_DEVID_OCTEONTX2_RVU_PF) },
+	{ 0, }  /* end of table */
+};
+
+MODULE_AUTHOR("Marvell International Ltd.");
+MODULE_DESCRIPTION(DRV_STRING);
+MODULE_LICENSE("GPL v2");
+MODULE_VERSION(DRV_VERSION);
+MODULE_DEVICE_TABLE(pci, otx2_pf_id_table);
+
+enum {
+	TYPE_PFAF,
+	TYPE_PFVF,
+};
+
+static int otx2_change_mtu(struct net_device *netdev, int new_mtu)
+{
+	bool if_up = netif_running(netdev);
+	int err = 0;
+
+	if (if_up)
+		otx2_stop(netdev);
+
+	netdev_info(netdev, "Changing MTU from %d to %d\n",
+		    netdev->mtu, new_mtu);
+	netdev->mtu = new_mtu;
+
+	if (if_up)
+		err = otx2_open(netdev);
+
+	return err;
+}
+
+static void otx2_queue_work(struct mbox *mw, struct workqueue_struct *mbox_wq,
+			    int first, int mdevs, u64 intr, int type)
+{
+	struct otx2_mbox_dev *mdev;
+	struct otx2_mbox *mbox;
+	struct mbox_hdr *hdr;
+	int i;
+
+	for (i = first; i < mdevs; i++) {
+		/* start from 0 */
+		if (!(intr & BIT_ULL(i - first)))
+			continue;
+
+		mbox = &mw->mbox;
+		mdev = &mbox->dev[i];
+		if (type == TYPE_PFAF)
+			otx2_sync_mbox_bbuf(mbox, i);
+		hdr = mdev->mbase + mbox->rx_start;
+		/* The hdr->num_msgs is set to zero immediately in the interrupt
+		 * handler to  ensure that it holds a correct value next time
+		 * when the interrupt handler is called.
+		 * pf->mbox.num_msgs holds the data for use in pfaf_mbox_handler
+		 * pf>mbox.up_num_msgs holds the data for use in
+		 * pfaf_mbox_up_handler.
+		 */
+		if (hdr->num_msgs) {
+			mw[i].num_msgs = hdr->num_msgs;
+			hdr->num_msgs = 0;
+			if (type == TYPE_PFAF)
+				memset(mbox->hwbase + mbox->rx_start, 0,
+				       ALIGN(sizeof(struct mbox_hdr),
+					     sizeof(u64)));
+
+			queue_work(mbox_wq, &mw[i].mbox_wrk);
+		}
+
+		mbox = &mw->mbox_up;
+		mdev = &mbox->dev[i];
+		if (type == TYPE_PFAF)
+			otx2_sync_mbox_bbuf(mbox, i);
+		hdr = mdev->mbase + mbox->rx_start;
+		if (hdr->num_msgs) {
+			mw[i].up_num_msgs = hdr->num_msgs;
+			hdr->num_msgs = 0;
+			if (type == TYPE_PFAF)
+				memset(mbox->hwbase + mbox->rx_start, 0,
+				       ALIGN(sizeof(struct mbox_hdr),
+					     sizeof(u64)));
+
+			queue_work(mbox_wq, &mw[i].mbox_up_wrk);
+		}
+	}
+}
+
+static void otx2_process_pfaf_mbox_msg(struct otx2_nic *pf,
+				       struct mbox_msghdr *msg)
+{
+	if (msg->id >= MBOX_MSG_MAX) {
+		dev_err(pf->dev,
+			"Mbox msg with unknown ID 0x%x\n", msg->id);
+		return;
+	}
+
+	if (msg->sig != OTX2_MBOX_RSP_SIG) {
+		dev_err(pf->dev,
+			"Mbox msg with wrong signature %x, ID 0x%x\n",
+			 msg->sig, msg->id);
+		return;
+	}
+
+	switch (msg->id) {
+	case MBOX_MSG_READY:
+		pf->pcifunc = msg->pcifunc;
+		break;
+	case MBOX_MSG_MSIX_OFFSET:
+		mbox_handler_msix_offset(pf, (struct msix_offset_rsp *)msg);
+		break;
+	case MBOX_MSG_NPA_LF_ALLOC:
+		mbox_handler_npa_lf_alloc(pf, (struct npa_lf_alloc_rsp *)msg);
+		break;
+	case MBOX_MSG_NIX_LF_ALLOC:
+		mbox_handler_nix_lf_alloc(pf, (struct nix_lf_alloc_rsp *)msg);
+		break;
+	case MBOX_MSG_NIX_TXSCH_ALLOC:
+		mbox_handler_nix_txsch_alloc(pf,
+					     (struct nix_txsch_alloc_rsp *)msg);
+		break;
+	case MBOX_MSG_CGX_STATS:
+		mbox_handler_cgx_stats(pf, (struct cgx_stats_rsp *)msg);
+		break;
+	default:
+		if (msg->rc)
+			dev_err(pf->dev,
+				"Mbox msg response has err %d, ID 0x%x\n",
+				msg->rc, msg->id);
+		break;
+	}
+}
+
+static void otx2_pfaf_mbox_handler(struct work_struct *work)
+{
+	struct otx2_mbox_dev *mdev;
+	struct mbox_hdr *rsp_hdr;
+	struct mbox_msghdr *msg;
+	struct otx2_mbox *mbox;
+	struct mbox *af_mbox;
+	struct otx2_nic *pf;
+	int offset, id;
+
+	af_mbox = container_of(work, struct mbox, mbox_wrk);
+	mbox = &af_mbox->mbox;
+	mdev = &mbox->dev[0];
+	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
+
+	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
+	pf = af_mbox->pfvf;
+
+	for (id = 0; id < af_mbox->num_msgs; id++) {
+		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
+		otx2_process_pfaf_mbox_msg(pf, msg);
+		offset = mbox->rx_start + msg->next_msgoff;
+		mdev->msgs_acked++;
+	}
+
+	otx2_mbox_reset(mbox, 0);
+}
+
+static void otx2_handle_link_event(struct otx2_nic *pf)
+{
+	struct cgx_link_user_info *linfo = &pf->linfo;
+	struct net_device *netdev = pf->netdev;
+
+	pr_info("%s NIC Link is %s %d Mbps %s duplex\n", netdev->name,
+		linfo->link_up ? "UP" : "DOWN", linfo->speed,
+		linfo->full_duplex ? "Full" : "Half");
+	if (linfo->link_up) {
+		netif_carrier_on(netdev);
+		netif_tx_start_all_queues(netdev);
+	} else {
+		netif_tx_stop_all_queues(netdev);
+		netif_carrier_off(netdev);
+	}
+}
+
+int otx2_mbox_up_handler_cgx_link_event(struct otx2_nic *pf,
+					struct cgx_link_info_msg *msg,
+					struct msg_rsp *rsp)
+{
+	/* Copy the link info sent by AF */
+	pf->linfo = msg->link_info;
+
+	/* interface has not been fully configured yet */
+	if (pf->flags & OTX2_FLAG_INTF_DOWN)
+		return 0;
+
+	otx2_handle_link_event(pf);
+	return 0;
+}
+
+static int otx2_process_mbox_msg_up(struct otx2_nic *pf,
+				    struct mbox_msghdr *req)
+{
+	/* Check if valid, if not reply with a invalid msg */
+	if (req->sig != OTX2_MBOX_REQ_SIG) {
+		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
+		return -ENODEV;
+	}
+
+	switch (req->id) {
+#define M(_name, _id, _fn_name, _req_type, _rsp_type)			\
+	case _id: {							\
+		struct _rsp_type *rsp;					\
+		int err;						\
+									\
+		rsp = (struct _rsp_type *)otx2_mbox_alloc_msg(		\
+			&pf->mbox.mbox_up, 0,				\
+			sizeof(struct _rsp_type));			\
+		if (!rsp)						\
+			return -ENOMEM;					\
+									\
+		rsp->hdr.id = _id;					\
+		rsp->hdr.sig = OTX2_MBOX_RSP_SIG;			\
+		rsp->hdr.pcifunc = 0;					\
+		rsp->hdr.rc = 0;					\
+									\
+		err = otx2_mbox_up_handler_ ## _fn_name(		\
+			pf, (struct _req_type *)req, rsp);		\
+		return err;						\
+	}
+MBOX_UP_CGX_MESSAGES
+#undef M
+		break;
+	default:
+		otx2_reply_invalid_msg(&pf->mbox.mbox_up, 0, 0, req->id);
+		return -ENODEV;
+	}
+	return 0;
+}
+
+static void otx2_pfaf_mbox_up_handler(struct work_struct *work)
+{
+	struct mbox *af_mbox = container_of(work, struct mbox, mbox_up_wrk);
+	struct otx2_mbox *mbox = &af_mbox->mbox_up;
+	struct otx2_mbox_dev *mdev = &mbox->dev[0];
+	struct otx2_nic *pf = af_mbox->pfvf;
+	int offset, id, devid = 0;
+	struct mbox_hdr *rsp_hdr;
+	struct mbox_msghdr *msg;
+
+	rsp_hdr = (struct mbox_hdr *)(mdev->mbase + mbox->rx_start);
+
+	offset = mbox->rx_start + ALIGN(sizeof(*rsp_hdr), MBOX_MSG_ALIGN);
+
+	for (id = 0; id < af_mbox->up_num_msgs; id++) {
+		msg = (struct mbox_msghdr *)(mdev->mbase + offset);
+
+		devid = msg->pcifunc & RVU_PFVF_FUNC_MASK;
+		/* Skip processing VF's messages */
+		if (!devid)
+			otx2_process_mbox_msg_up(pf, msg);
+		offset = mbox->rx_start + msg->next_msgoff;
+	}
+
+	otx2_mbox_msg_send(mbox, 0);
+}
+
+static irqreturn_t otx2_pfaf_mbox_intr_handler(int irq, void *pf_irq)
+{
+	struct otx2_nic *pf = (struct otx2_nic *)pf_irq;
+	struct mbox *mbox;
+
+	/* Clear the IRQ */
+	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
+
+	mbox = &pf->mbox;
+	otx2_queue_work(mbox, pf->mbox_wq, 0, 1, 1, TYPE_PFAF);
+
+	return IRQ_HANDLED;
+}
+
+static void otx2_disable_mbox_intr(struct otx2_nic *pf)
+{
+	int vector = pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX);
+
+	/* Disable AF => PF mailbox IRQ */
+	otx2_write64(pf, RVU_PF_INT_ENA_W1C, BIT_ULL(0));
+	free_irq(vector, pf);
+}
+
+static int otx2_register_mbox_intr(struct otx2_nic *pf, bool probe_af)
+{
+	struct otx2_hw *hw = &pf->hw;
+	struct msg_req *req;
+	char *irq_name;
+	int err;
+
+	/* Register mailbox interrupt handler */
+	irq_name = &hw->irq_name[RVU_PF_INT_VEC_AFPF_MBOX * NAME_SIZE];
+	snprintf(irq_name, NAME_SIZE, "RVUPFAF Mbox");
+	err = request_irq(pci_irq_vector(pf->pdev, RVU_PF_INT_VEC_AFPF_MBOX),
+			  otx2_pfaf_mbox_intr_handler, 0, irq_name, pf);
+	if (err) {
+		dev_err(pf->dev,
+			"RVUPF: IRQ registration failed for PFAF mbox irq\n");
+		return err;
+	}
+
+	/* Enable mailbox interrupt for msgs coming from AF.
+	 * First clear to avoid spurious interrupts, if any.
+	 */
+	otx2_write64(pf, RVU_PF_INT, BIT_ULL(0));
+	otx2_write64(pf, RVU_PF_INT_ENA_W1S, BIT_ULL(0));
+
+	if (!probe_af)
+		return 0;
+
+	/* Check mailbox communication with AF */
+	req = otx2_mbox_alloc_msg_ready(&pf->mbox);
+	if (!req) {
+		otx2_disable_mbox_intr(pf);
+		return -ENOMEM;
+	}
+	err = otx2_sync_mbox_msg(&pf->mbox);
+	if (err) {
+		dev_warn(pf->dev,
+			 "AF not responding to mailbox, deferring probe\n");
+		otx2_disable_mbox_intr(pf);
+		return -EPROBE_DEFER;
+	}
+
+	return 0;
+}
+
+static void otx2_pfaf_mbox_destroy(struct otx2_nic *pf)
+{
+	struct mbox *mbox = &pf->mbox;
+
+	if (pf->mbox_wq) {
+		flush_workqueue(pf->mbox_wq);
+		destroy_workqueue(pf->mbox_wq);
+		pf->mbox_wq = NULL;
+	}
+
+	if (mbox->mbox.hwbase)
+		iounmap((void __iomem *)mbox->mbox.hwbase);
+
+	otx2_mbox_destroy(&mbox->mbox);
+	otx2_mbox_destroy(&mbox->mbox_up);
+}
+
+static int otx2_pfaf_mbox_init(struct otx2_nic *pf)
+{
+	struct mbox *mbox = &pf->mbox;
+	void __iomem *hwbase;
+	int err;
+
+	mbox->pfvf = pf;
+	pf->mbox_wq = alloc_workqueue("otx2_pfaf_mailbox",
+				      WQ_UNBOUND | WQ_HIGHPRI |
+				      WQ_MEM_RECLAIM, 1);
+	if (!pf->mbox_wq)
+		return -ENOMEM;
+
+	/* Mailbox is a reserved memory (in RAM) region shared between
+	 * admin function (i.e AF) and this PF, shouldn't be mapped as
+	 * device memory to allow unaligned accesses.
+	 */
+	hwbase = ioremap_wc(pci_resource_start(pf->pdev, PCI_MBOX_BAR_NUM),
+			    pci_resource_len(pf->pdev, PCI_MBOX_BAR_NUM));
+	if (!hwbase) {
+		dev_err(pf->dev, "Unable to map PFAF mailbox region\n");
+		err = -ENOMEM;
+		goto exit;
+	}
+
+	err = otx2_mbox_init(&mbox->mbox, hwbase, pf->pdev, pf->reg_base,
+			     MBOX_DIR_PFAF, 1);
+	if (err)
+		goto exit;
+
+	err = otx2_mbox_init(&mbox->mbox_up, hwbase, pf->pdev, pf->reg_base,
+			     MBOX_DIR_PFAF_UP, 1);
+	if (err)
+		goto exit;
+
+	err = otx2_mbox_bbuf_init(mbox, pf->pdev);
+	if (err)
+		goto exit;
+
+	INIT_WORK(&mbox->mbox_wrk, otx2_pfaf_mbox_handler);
+	INIT_WORK(&mbox->mbox_up_wrk, otx2_pfaf_mbox_up_handler);
+	otx2_mbox_lock_init(&pf->mbox);
+
+	return 0;
+exit:
+	otx2_pfaf_mbox_destroy(pf);
+	return err;
+}
+
+static int otx2_cgx_config_linkevents(struct otx2_nic *pf, bool enable)
+{
+	struct msg_req *msg;
+	int err;
+
+	otx2_mbox_lock(&pf->mbox);
+	if (enable)
+		msg = otx2_mbox_alloc_msg_cgx_start_linkevents(&pf->mbox);
+	else
+		msg = otx2_mbox_alloc_msg_cgx_stop_linkevents(&pf->mbox);
+
+	if (!msg) {
+		otx2_mbox_unlock(&pf->mbox);
+		return -ENOMEM;
+	}
+
+	err = otx2_sync_mbox_msg(&pf->mbox);
+	otx2_mbox_unlock(&pf->mbox);
+	return err;
+}
+
+static int otx2_cgx_config_loopback(struct otx2_nic *pf, bool enable)
+{
+	struct msg_req *msg;
+	int err;
+
+	otx2_mbox_lock(&pf->mbox);
+	if (enable)
+		msg = otx2_mbox_alloc_msg_cgx_intlbk_enable(&pf->mbox);
+	else
+		msg = otx2_mbox_alloc_msg_cgx_intlbk_disable(&pf->mbox);
+
+	if (!msg) {
+		otx2_mbox_unlock(&pf->mbox);
+		return -ENOMEM;
+	}
+
+	err = otx2_sync_mbox_msg(&pf->mbox);
+	otx2_mbox_unlock(&pf->mbox);
+	return err;
+}
+
+int otx2_set_real_num_queues(struct net_device *netdev,
+			     int tx_queues, int rx_queues)
+{
+	int err;
+
+	err = netif_set_real_num_tx_queues(netdev, tx_queues);
+	if (err) {
+		netdev_err(netdev,
+			   "Failed to set no of Tx queues: %d\n", tx_queues);
+		return err;
+	}
+
+	err = netif_set_real_num_rx_queues(netdev, rx_queues);
+	if (err)
+		netdev_err(netdev,
+			   "Failed to set no of Rx queues: %d\n", rx_queues);
+	return err;
+}
+
+static irqreturn_t otx2_q_intr_handler(int irq, void *data)
+{
+	struct otx2_nic *pf = data;
+	u64 val, *ptr;
+	u64 qidx = 0;
+
+	/* CQ */
+	for (qidx = 0; qidx < pf->qset.cq_cnt; qidx++) {
+		ptr = otx2_get_regaddr(pf, NIX_LF_CQ_OP_INT);
+		val = otx2_atomic64_add((qidx << 44), ptr);
+
+		otx2_write64(pf, NIX_LF_CQ_OP_INT, (qidx << 44) |
+			     (val & NIX_CQERRINT_BITS));
+		if (!(val & (NIX_CQERRINT_BITS | BIT_ULL(42))))
+			continue;
+
+		if (val & BIT_ULL(42)) {
+			netdev_err(pf->netdev, "CQ%lld: error reading NIX_LF_CQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
+				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
+		} else {
+			if (val & BIT_ULL(NIX_CQERRINT_DOOR_ERR))
+				netdev_err(pf->netdev, "CQ%lld: Doorbell error",
+					   qidx);
+			if (val & BIT_ULL(NIX_CQERRINT_CQE_FAULT))
+				netdev_err(pf->netdev, "CQ%lld: Memory fault on CQE write to LLC/DRAM",
+					   qidx);
+		}
+
+		schedule_work(&pf->reset_task);
+	}
+
+	/* SQ */
+	for (qidx = 0; qidx < pf->hw.tx_queues; qidx++) {
+		ptr = otx2_get_regaddr(pf, NIX_LF_SQ_OP_INT);
+		val = otx2_atomic64_add((qidx << 44), ptr);
+		otx2_write64(pf, NIX_LF_SQ_OP_INT, (qidx << 44) |
+			     (val & NIX_SQINT_BITS));
+
+		if (!(val & (NIX_SQINT_BITS | BIT_ULL(42))))
+			continue;
+
+		if (val & BIT_ULL(42)) {
+			netdev_err(pf->netdev, "SQ%lld: error reading NIX_LF_SQ_OP_INT, NIX_LF_ERR_INT 0x%llx\n",
+				   qidx, otx2_read64(pf, NIX_LF_ERR_INT));
+		} else {
+			if (val & BIT_ULL(NIX_SQINT_LMT_ERR)) {
+				netdev_err(pf->netdev, "SQ%lld: LMT store error NIX_LF_SQ_OP_ERR_DBG:0x%llx",
+					   qidx,
+					   otx2_read64(pf,
+						       NIX_LF_SQ_OP_ERR_DBG));
+				otx2_write64(pf, NIX_LF_SQ_OP_ERR_DBG,
+					     BIT_ULL(44));
+			}
+			if (val & BIT_ULL(NIX_SQINT_MNQ_ERR)) {
+				netdev_err(pf->netdev, "SQ%lld: Meta-descriptor enqueue error NIX_LF_MNQ_ERR_DGB:0x%llx\n",
+					   qidx,
+					   otx2_read64(pf, NIX_LF_MNQ_ERR_DBG));
+				otx2_write64(pf, NIX_LF_MNQ_ERR_DBG,
+					     BIT_ULL(44));
+			}
+			if (val & BIT_ULL(NIX_SQINT_SEND_ERR)) {
+				netdev_err(pf->netdev, "SQ%lld: Send error, NIX_LF_SEND_ERR_DBG 0x%llx",
+					   qidx,
+					   otx2_read64(pf,
+						       NIX_LF_SEND_ERR_DBG));
+				otx2_write64(pf, NIX_LF_SEND_ERR_DBG,
+					     BIT_ULL(44));
+			}
+			if (val & BIT_ULL(NIX_SQINT_SQB_ALLOC_FAIL))
+				netdev_err(pf->netdev, "SQ%lld: SQB allocation failed",
+					   qidx);
+		}
+
+		schedule_work(&pf->reset_task);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t otx2_cq_intr_handler(int irq, void *cq_irq)
+{
+	struct otx2_cq_poll *cq_poll = (struct otx2_cq_poll *)cq_irq;
+	struct otx2_nic *pf = (struct otx2_nic *)cq_poll->dev;
+	int qidx = cq_poll->cint_idx;
+
+	/* Disable interrupts.
+	 *
+	 * Completion interrupts behave in a level-triggered interrupt
+	 * fashion, and hence have to be cleared only after it is serviced.
+	 */
+	otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
+
+	/* Schedule NAPI */
+	napi_schedule_irqoff(&cq_poll->napi);
+
+	return IRQ_HANDLED;
+}
+
+static void otx2_disable_napi(struct otx2_nic *pf)
+{
+	struct otx2_qset *qset = &pf->qset;
+	struct otx2_cq_poll *cq_poll;
+	int qidx;
+
+	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
+		cq_poll = &qset->napi[qidx];
+		napi_disable(&cq_poll->napi);
+		netif_napi_del(&cq_poll->napi);
+	}
+}
+
+static void otx2_free_cq_res(struct otx2_nic *pf)
+{
+	struct otx2_qset *qset = &pf->qset;
+	struct otx2_cq_queue *cq;
+	int qidx;
+
+	/* Disable CQs */
+	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_CQ, false);
+	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
+		cq = &qset->cq[qidx];
+		qmem_free(pf->dev, cq->cqe);
+	}
+}
+
+static void otx2_free_sq_res(struct otx2_nic *pf)
+{
+	struct otx2_qset *qset = &pf->qset;
+	struct otx2_snd_queue *sq;
+	int qidx;
+
+	/* Disable SQs */
+	otx2_ctx_disable(&pf->mbox, NIX_AQ_CTYPE_SQ, false);
+	/* Free SQB pointers */
+	otx2_sq_free_sqbs(pf);
+	for (qidx = 0; qidx < pf->hw.tx_queues; qidx++) {
+		sq = &qset->sq[qidx];
+		qmem_free(pf->dev, sq->sqe);
+		qmem_free(pf->dev, sq->tso_hdrs);
+		kfree(sq->sg);
+		kfree(sq->sqb_ptrs);
+	}
+}
+
+static int otx2_init_hw_resources(struct otx2_nic *pf)
+{
+	struct mbox *mbox = &pf->mbox;
+	struct otx2_hw *hw = &pf->hw;
+	struct msg_req *req;
+	int err = 0, lvl;
+
+	/* Set required NPA LF's pool counts
+	 * Auras and Pools are used in a 1:1 mapping,
+	 * so, aura count = pool count.
+	 */
+	hw->rqpool_cnt = hw->rx_queues;
+	hw->sqpool_cnt = hw->tx_queues;
+	hw->pool_cnt = hw->rqpool_cnt + hw->sqpool_cnt;
+
+	/* Get the size of receive buffers to allocate */
+	pf->rbsize = RCV_FRAG_LEN(pf->netdev->mtu + OTX2_ETH_HLEN);
+
+	otx2_mbox_lock(mbox);
+	/* NPA init */
+	err = otx2_config_npa(pf);
+	if (err)
+		goto exit;
+
+	/* NIX init */
+	err = otx2_config_nix(pf);
+	if (err)
+		goto err_free_npa_lf;
+
+	/* Init Auras and pools used by NIX RQ, for free buffer ptrs */
+	err = otx2_rq_aura_pool_init(pf);
+	if (err) {
+		otx2_mbox_unlock(mbox);
+		goto err_free_nix_lf;
+	}
+	/* Init Auras and pools used by NIX SQ, for queueing SQEs */
+	err = otx2_sq_aura_pool_init(pf);
+	if (err) {
+		otx2_mbox_unlock(mbox);
+		goto err_free_rq_ptrs;
+	}
+
+	err = otx2_txsch_alloc(pf);
+	if (err) {
+		otx2_mbox_unlock(mbox);
+		goto err_free_sq_ptrs;
+	}
+
+	err = otx2_config_nix_queues(pf);
+	if (err) {
+		otx2_mbox_unlock(mbox);
+		goto err_free_txsch;
+	}
+	for (lvl = 0; lvl < NIX_TXSCH_LVL_CNT; lvl++) {
+		err = otx2_txschq_config(pf, lvl);
+		if (err) {
+			otx2_mbox_unlock(mbox);
+			goto err_free_nix_queues;
+		}
+	}
+	otx2_mbox_unlock(mbox);
+	return err;
+
+err_free_nix_queues:
+	otx2_free_sq_res(pf);
+	otx2_free_cq_res(pf);
+	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
+err_free_txsch:
+	if (otx2_txschq_stop(pf))
+		dev_err(pf->dev, "%s failed to stop TX schedulers\n", __func__);
+err_free_sq_ptrs:
+	otx2_sq_free_sqbs(pf);
+err_free_rq_ptrs:
+	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
+	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
+	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
+	otx2_aura_pool_free(pf);
+err_free_nix_lf:
+	otx2_mbox_lock(mbox);
+	req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
+	if (req) {
+		if (otx2_sync_mbox_msg(mbox))
+			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
+	}
+err_free_npa_lf:
+	/* Reset NPA LF */
+	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
+	if (req) {
+		if (otx2_sync_mbox_msg(mbox))
+			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
+	}
+exit:
+	otx2_mbox_unlock(mbox);
+	return err;
+}
+
+static void otx2_free_hw_resources(struct otx2_nic *pf)
+{
+	struct otx2_qset *qset = &pf->qset;
+	struct mbox *mbox = &pf->mbox;
+	struct otx2_cq_queue *cq;
+	struct msg_req *req;
+	int qidx, err;
+
+	/* Ensure all SQE are processed */
+	otx2_sqb_flush(pf);
+
+	/* Stop transmission */
+	err = otx2_txschq_stop(pf);
+	if (err)
+		dev_err(pf->dev, "RVUPF: Failed to stop/free TX schedulers\n");
+
+	/* Disable RQs */
+	otx2_ctx_disable(mbox, NIX_AQ_CTYPE_RQ, false);
+
+	/*Dequeue all CQEs */
+	for (qidx = 0; qidx < qset->cq_cnt; qidx++) {
+		cq = &qset->cq[qidx];
+		if (cq->cq_type == CQ_RX)
+			otx2_cleanup_rx_cqes(pf, cq);
+		else
+			otx2_cleanup_tx_cqes(pf, cq);
+	}
+
+	otx2_free_sq_res(pf);
+
+	/* Free RQ buffer pointers*/
+	otx2_free_aura_ptr(pf, AURA_NIX_RQ);
+
+	otx2_free_cq_res(pf);
+
+	otx2_mbox_lock(mbox);
+	/* Reset NIX LF */
+	req = otx2_mbox_alloc_msg_nix_lf_free(mbox);
+	if (req) {
+		if (otx2_sync_mbox_msg(mbox))
+			dev_err(pf->dev, "%s failed to free nixlf\n", __func__);
+	}
+	otx2_mbox_unlock(mbox);
+
+	/* Disable NPA Pool and Aura hw context */
+	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_POOL, true);
+	otx2_ctx_disable(mbox, NPA_AQ_CTYPE_AURA, true);
+	otx2_aura_pool_free(pf);
+
+	otx2_mbox_lock(mbox);
+	/* Reset NPA LF */
+	req = otx2_mbox_alloc_msg_npa_lf_free(mbox);
+	if (req) {
+		if (otx2_sync_mbox_msg(mbox))
+			dev_err(pf->dev, "%s failed to free npalf\n", __func__);
+	}
+	otx2_mbox_unlock(mbox);
+}
+
+int otx2_open(struct net_device *netdev)
+{
+	struct otx2_nic *pf = netdev_priv(netdev);
+	struct otx2_cq_poll *cq_poll = NULL;
+	struct otx2_qset *qset = &pf->qset;
+	int err = 0, qidx, vec;
+	char *irq_name;
+
+	netif_carrier_off(netdev);
+
+	pf->qset.cq_cnt = pf->hw.rx_queues + pf->hw.tx_queues;
+	/* RQ and SQs are mapped to different CQs,
+	 * so find out max CQ IRQs (i.e CINTs) needed.
+	 */
+	pf->hw.cint_cnt = max(pf->hw.rx_queues, pf->hw.tx_queues);
+	qset->napi = kcalloc(pf->hw.cint_cnt, sizeof(*cq_poll), GFP_KERNEL);
+	if (!qset->napi)
+		return -ENOMEM;
+
+	/* CQ size of RQ */
+	qset->rqe_cnt = qset->rqe_cnt ? qset->rqe_cnt : Q_COUNT(Q_SIZE_256);
+	/* CQ size of SQ */
+	qset->sqe_cnt = qset->sqe_cnt ? qset->sqe_cnt : Q_COUNT(Q_SIZE_4K);
+
+	err = -ENOMEM;
+	qset->cq = kcalloc(pf->qset.cq_cnt,
+			   sizeof(struct otx2_cq_queue), GFP_KERNEL);
+	if (!qset->cq)
+		goto err_free_mem;
+
+	qset->sq = kcalloc(pf->hw.tx_queues,
+			   sizeof(struct otx2_snd_queue), GFP_KERNEL);
+	if (!qset->sq)
+		goto err_free_mem;
+
+	qset->rq = kcalloc(pf->hw.rx_queues,
+			   sizeof(struct otx2_rcv_queue), GFP_KERNEL);
+	if (!qset->rq)
+		goto err_free_mem;
+
+	err = otx2_init_hw_resources(pf);
+	if (err)
+		goto err_free_mem;
+
+	/* Register NAPI handler */
+	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
+		cq_poll = &qset->napi[qidx];
+		cq_poll->cint_idx = qidx;
+		/* RQ0 & SQ0 are mapped to CINT0 and so on..
+		 * 'cq_ids[0]' points to RQ's CQ and
+		 * 'cq_ids[1]' points to SQ's CQ and
+		 */
+		cq_poll->cq_ids[CQ_RX] =
+			(qidx <  pf->hw.rx_queues) ? qidx : CINT_INVALID_CQ;
+		cq_poll->cq_ids[CQ_TX] = (qidx < pf->hw.tx_queues) ?
+				      qidx + pf->hw.rx_queues : CINT_INVALID_CQ;
+		cq_poll->dev = (void *)pf;
+		netif_napi_add(netdev, &cq_poll->napi,
+			       otx2_napi_handler, NAPI_POLL_WEIGHT);
+		napi_enable(&cq_poll->napi);
+	}
+
+	/* Set maximum frame size allowed in HW */
+	err = otx2_hw_set_mtu(pf, netdev->mtu);
+	if (err)
+		goto err_disable_napi;
+
+	/* Initialize RSS */
+	err = otx2_rss_init(pf);
+	if (err)
+		goto err_disable_napi;
+
+	/* Register Queue IRQ handlers */
+	vec = pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START;
+	irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
+
+	snprintf(irq_name, NAME_SIZE, "%s-qerr", pf->netdev->name);
+
+	err = request_irq(pci_irq_vector(pf->pdev, vec),
+			  otx2_q_intr_handler, 0, irq_name, pf);
+	if (err) {
+		dev_err(pf->dev,
+			"RVUPF%d: IRQ registration failed for QERR\n",
+			rvu_get_pf(pf->pcifunc));
+		goto err_disable_napi;
+	}
+
+	/* Enable QINT IRQ */
+	otx2_write64(pf, NIX_LF_QINTX_ENA_W1S(0), BIT_ULL(0));
+
+	/* Register CQ IRQ handlers */
+	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
+	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
+		irq_name = &pf->hw.irq_name[vec * NAME_SIZE];
+
+		snprintf(irq_name, NAME_SIZE, "%s-rxtx-%d", pf->netdev->name,
+			 qidx);
+
+		err = request_irq(pci_irq_vector(pf->pdev, vec),
+				  otx2_cq_intr_handler, 0, irq_name,
+				  &qset->napi[qidx]);
+		if (err) {
+			dev_err(pf->dev,
+				"RVUPF%d: IRQ registration failed for CQ%d\n",
+				rvu_get_pf(pf->pcifunc), qidx);
+			goto err_free_cints;
+		}
+		vec++;
+
+		otx2_config_irq_coalescing(pf, qidx);
+
+		/* Enable CQ IRQ */
+		otx2_write64(pf, NIX_LF_CINTX_INT(qidx), BIT_ULL(0));
+		otx2_write64(pf, NIX_LF_CINTX_ENA_W1S(qidx), BIT_ULL(0));
+	}
+
+	otx2_set_cints_affinity(pf);
+
+	pf->flags &= ~OTX2_FLAG_INTF_DOWN;
+	/* 'intf_down' may be checked on any cpu */
+	smp_wmb();
+
+	/* we have already received link status notification */
+	if (pf->linfo.link_up && !(pf->pcifunc & RVU_PFVF_FUNC_MASK))
+		otx2_handle_link_event(pf);
+
+	err = otx2_rxtx_enable(pf, true);
+	if (err)
+		goto err_free_cints;
+
+	return 0;
+
+err_free_cints:
+	otx2_free_cints(pf, qidx);
+	vec = pci_irq_vector(pf->pdev,
+			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
+	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
+	synchronize_irq(vec);
+	free_irq(vec, pf);
+err_disable_napi:
+	otx2_disable_napi(pf);
+	otx2_free_hw_resources(pf);
+err_free_mem:
+	kfree(qset->sq);
+	kfree(qset->cq);
+	kfree(qset->rq);
+	kfree(qset->napi);
+	return err;
+}
+
+int otx2_stop(struct net_device *netdev)
+{
+	struct otx2_nic *pf = netdev_priv(netdev);
+	struct otx2_cq_poll *cq_poll = NULL;
+	struct otx2_qset *qset = &pf->qset;
+	int qidx, vec, wrk;
+
+	netif_carrier_off(netdev);
+	netif_tx_stop_all_queues(netdev);
+
+	pf->flags |= OTX2_FLAG_INTF_DOWN;
+	/* 'intf_down' may be checked on any cpu */
+	smp_wmb();
+
+	/* First stop packet Rx/Tx */
+	otx2_rxtx_enable(pf, false);
+
+	/* Cleanup Queue IRQ */
+	vec = pci_irq_vector(pf->pdev,
+			     pf->hw.nix_msixoff + NIX_LF_QINT_VEC_START);
+	otx2_write64(pf, NIX_LF_QINTX_ENA_W1C(0), BIT_ULL(0));
+	synchronize_irq(vec);
+	free_irq(vec, pf);
+
+	/* Cleanup CQ NAPI and IRQ */
+	vec = pf->hw.nix_msixoff + NIX_LF_CINT_VEC_START;
+	for (qidx = 0; qidx < pf->hw.cint_cnt; qidx++) {
+		/* Disable interrupt */
+		otx2_write64(pf, NIX_LF_CINTX_ENA_W1C(qidx), BIT_ULL(0));
+
+		synchronize_irq(pci_irq_vector(pf->pdev, vec));
+
+		cq_poll = &qset->napi[qidx];
+		napi_synchronize(&cq_poll->napi);
+		vec++;
+	}
+
+	netif_tx_disable(netdev);
+
+	otx2_free_hw_resources(pf);
+	otx2_free_cints(pf, pf->hw.cint_cnt);
+	otx2_disable_napi(pf);
+
+	for (qidx = 0; qidx < netdev->num_tx_queues; qidx++)
+		netdev_tx_reset_queue(netdev_get_tx_queue(netdev, qidx));
+
+	for (wrk = 0; wrk < pf->qset.cq_cnt; wrk++)
+		cancel_delayed_work_sync(&pf->refill_wrk[wrk].pool_refill_work);
+	devm_kfree(pf->dev, pf->refill_wrk);
+
+	kfree(qset->sq);
+	kfree(qset->cq);
+	kfree(qset->rq);
+	kfree(qset->napi);
+	/* Do not clear RQ/SQ ringsize settings */
+	memset((void *)qset + offsetof(struct otx2_qset, sqe_cnt), 0,
+	       sizeof(*qset) - offsetof(struct otx2_qset, sqe_cnt));
+	return 0;
+}
+
+static netdev_tx_t otx2_xmit(struct sk_buff *skb, struct net_device *netdev)
+{
+	struct otx2_nic *pf = netdev_priv(netdev);
+	int qidx = skb_get_queue_mapping(skb);
+	struct otx2_snd_queue *sq;
+	struct netdev_queue *txq;
+
+	/* Check for minimum and maximum packet length */
+	if (skb->len <= ETH_HLEN ||
+	    (!skb_shinfo(skb)->gso_size && skb->len > pf->max_frs)) {
+		dev_kfree_skb(skb);
+		return NETDEV_TX_OK;
+	}
+
+	sq = &pf->qset.sq[qidx];
+	txq = netdev_get_tx_queue(netdev, qidx);
+
+	if (!otx2_sq_append_skb(netdev, sq, skb, qidx)) {
+		netif_tx_stop_queue(txq);
+
+		/* Check again, incase SQBs got freed up */
+		smp_mb();
+		if (((sq->num_sqbs - *sq->aura_fc_addr) * sq->sqe_per_sqb)
+							> sq->sqe_thresh)
+			netif_tx_wake_queue(txq);
+
+		return NETDEV_TX_BUSY;
+	}
+
+	return NETDEV_TX_OK;
+}
+
+static void otx2_set_rx_mode(struct net_device *netdev)
+{
+	struct otx2_nic *pf = netdev_priv(netdev);
+	struct nix_rx_mode *req;
+
+	if (!(netdev->flags & IFF_UP))
+		return;
+
+	otx2_mbox_lock(&pf->mbox);
+	req = otx2_mbox_alloc_msg_nix_set_rx_mode(&pf->mbox);
+	if (!req) {
+		otx2_mbox_unlock(&pf->mbox);
+		return;
+	}
+
+	req->mode = NIX_RX_MODE_UCAST;
+
+	/* We don't support MAC address filtering yet */
+	if (netdev->flags & IFF_PROMISC)
+		req->mode |= NIX_RX_MODE_PROMISC;
+	else if (netdev->flags & (IFF_ALLMULTI | IFF_MULTICAST))
+		req->mode |= NIX_RX_MODE_ALLMULTI;
+
+	otx2_sync_mbox_msg(&pf->mbox);
+	otx2_mbox_unlock(&pf->mbox);
+}
+
+static int otx2_set_features(struct net_device *netdev,
+			     netdev_features_t features)
+{
+	netdev_features_t changed = features ^ netdev->features;
+	struct otx2_nic *pf = netdev_priv(netdev);
+
+	if ((changed & NETIF_F_LOOPBACK) && netif_running(netdev))
+		return otx2_cgx_config_loopback(pf,
+						features & NETIF_F_LOOPBACK);
+	return 0;
+}
+
+static void otx2_reset_task(struct work_struct *work)
+{
+	struct otx2_nic *pf = container_of(work, struct otx2_nic, reset_task);
+
+	if (!netif_running(pf->netdev))
+		return;
+
+	otx2_stop(pf->netdev);
+	pf->reset_count++;
+	otx2_open(pf->netdev);
+	netif_trans_update(pf->netdev);
+}
+
+static const struct net_device_ops otx2_netdev_ops = {
+	.ndo_open		= otx2_open,
+	.ndo_stop		= otx2_stop,
+	.ndo_start_xmit		= otx2_xmit,
+	.ndo_set_mac_address    = otx2_set_mac_address,
+	.ndo_change_mtu		= otx2_change_mtu,
+	.ndo_set_rx_mode	= otx2_set_rx_mode,
+	.ndo_set_features	= otx2_set_features,
+	.ndo_tx_timeout		= otx2_tx_timeout,
+	.ndo_get_stats64	= otx2_get_stats64,
+};
+
+static int otx2_check_pf_usable(struct otx2_nic *nic)
+{
+	u64 rev;
+
+	rev = otx2_read64(nic, RVU_PF_BLOCK_ADDRX_DISC(BLKADDR_RVUM));
+	rev = (rev >> 12) & 0xFF;
+	/* Check if AF has setup revision for RVUM block,
+	 * otherwise this driver probe should be deferred
+	 * until AF driver comes up.
+	 */
+	if (!rev) {
+		dev_warn(nic->dev,
+			 "AF is not initialized, deferring probe\n");
+		return -EPROBE_DEFER;
+	}
+	return 0;
+}
+
+static int otx2_realloc_msix_vectors(struct otx2_nic *pf)
+{
+	struct otx2_hw *hw = &pf->hw;
+	int num_vec, err;
+
+	/* NPA interrupts are inot registered, so alloc only
+	 * upto NIX vector offset.
+	 */
+	num_vec = hw->nix_msixoff;
+	num_vec += NIX_LF_CINT_VEC_START + hw->max_queues;
+
+	otx2_disable_mbox_intr(pf);
+	pci_free_irq_vectors(hw->pdev);
+	pci_free_irq_vectors(hw->pdev);
+	err = pci_alloc_irq_vectors(hw->pdev, num_vec, num_vec, PCI_IRQ_MSIX);
+	if (err < 0) {
+		dev_err(pf->dev, "%s: Failed to realloc %d IRQ vectors\n",
+			__func__, num_vec);
+		return err;
+	}
+
+	return otx2_register_mbox_intr(pf, false);
+}
+
+static int otx2_probe(struct pci_dev *pdev, const struct pci_device_id *id)
+{
+	struct device *dev = &pdev->dev;
+	struct net_device *netdev;
+	struct otx2_nic *pf;
+	struct otx2_hw *hw;
+	int err, qcount;
+	int num_vec;
+
+	err = pcim_enable_device(pdev);
+	if (err) {
+		dev_err(dev, "Failed to enable PCI device\n");
+		return err;
+	}
+
+	err = pci_request_regions(pdev, DRV_NAME);
+	if (err) {
+		dev_err(dev, "PCI request regions failed 0x%x\n", err);
+		return err;
+	}
+
+	err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(48));
+	if (err) {
+		dev_err(dev, "DMA mask config failed, abort\n");
+		goto err_release_regions;
+	}
+
+	pci_set_master(pdev);
+
+	/* Set number of queues */
+	qcount = min_t(int, num_online_cpus(), OTX2_MAX_CQ_CNT);
+
+	netdev = alloc_etherdev_mqs(sizeof(*pf), qcount, qcount);
+	if (!netdev) {
+		err = -ENOMEM;
+		goto err_release_regions;
+	}
+
+	pci_set_drvdata(pdev, netdev);
+	SET_NETDEV_DEV(netdev, &pdev->dev);
+	pf = netdev_priv(netdev);
+	pf->netdev = netdev;
+	pf->pdev = pdev;
+	pf->dev = dev;
+	pf->flags |= OTX2_FLAG_INTF_DOWN;
+
+	hw = &pf->hw;
+	hw->pdev = pdev;
+	hw->rx_queues = qcount;
+	hw->tx_queues = qcount;
+	hw->max_queues = qcount;
+
+	num_vec = pci_msix_vec_count(pdev);
+	hw->irq_name = devm_kmalloc_array(&hw->pdev->dev, num_vec, NAME_SIZE,
+					  GFP_KERNEL);
+	if (!hw->irq_name)
+		goto err_free_netdev;
+
+	hw->affinity_mask = devm_kcalloc(&hw->pdev->dev, num_vec,
+					 sizeof(cpumask_var_t), GFP_KERNEL);
+	if (!hw->affinity_mask)
+		goto err_free_netdev;
+
+	/* Map CSRs */
+	pf->reg_base = pcim_iomap(pdev, PCI_CFG_REG_BAR_NUM, 0);
+	if (!pf->reg_base) {
+		dev_err(dev, "Unable to map physical function CSRs, aborting\n");
+		err = -ENOMEM;
+		goto err_free_netdev;
+	}
+
+	err = otx2_check_pf_usable(pf);
+	if (err)
+		goto err_free_netdev;
+
+	err = pci_alloc_irq_vectors(hw->pdev, RVU_PF_INT_VEC_CNT,
+				    RVU_PF_INT_VEC_CNT, PCI_IRQ_MSIX);
+	if (err < 0) {
+		dev_err(dev, "%s: Failed to alloc %d IRQ vectors\n",
+			__func__, num_vec);
+		goto err_free_netdev;
+	}
+
+	/* Init PF <=> AF mailbox stuff */
+	err = otx2_pfaf_mbox_init(pf);
+	if (err)
+		goto err_free_irq_vectors;
+
+	/* Register mailbox interrupt */
+	err = otx2_register_mbox_intr(pf, true);
+	if (err)
+		goto err_mbox_destroy;
+
+	/* Request AF to attach NPA and NIX LFs to this PF.
+	 * NIX and NPA LFs are needed for this PF to function as a NIC.
+	 */
+	err = otx2_attach_npa_nix(pf);
+	if (err)
+		goto err_disable_mbox_intr;
+
+	err = otx2_realloc_msix_vectors(pf);
+	if (err)
+		goto err_detach_rsrc;
+
+	err = otx2_set_real_num_queues(netdev, hw->tx_queues, hw->rx_queues);
+	if (err)
+		goto err_detach_rsrc;
+
+	otx2_setup_dev_hw_settings(pf);
+
+	/* Assign default mac address */
+	otx2_get_mac_from_af(netdev);
+
+	/* NPA's pool is a stack to which SW frees buffer pointers via Aura.
+	 * HW allocates buffer pointer from stack and uses it for DMA'ing
+	 * ingress packet. In some scenarios HW can free back allocated buffer
+	 * pointers to pool. This makes it impossible for SW to maintain a
+	 * parallel list where physical addresses of buffer pointers (IOVAs)
+	 * given to HW can be saved for later reference.
+	 *
+	 * So the only way to convert Rx packet's buffer address is to use
+	 * IOMMU's iova_to_phys() handler which translates the address by
+	 * walking through the translation tables.
+	 */
+	pf->iommu_domain = iommu_get_domain_for_dev(dev);
+
+	netdev->hw_features = (NETIF_F_RXCSUM | NETIF_F_IP_CSUM |
+			       NETIF_F_IPV6_CSUM | NETIF_F_RXHASH |
+			       NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO6);
+	netdev->features |= netdev->hw_features;
+
+	netdev->hw_features |= NETIF_F_LOOPBACK | NETIF_F_RXALL;
+
+	netdev->gso_max_segs = OTX2_MAX_GSO_SEGS;
+	netdev->watchdog_timeo = OTX2_TX_TIMEOUT;
+
+	netdev->netdev_ops = &otx2_netdev_ops;
+
+	/* MTU range: 64 - 9190 */
+	netdev->min_mtu = OTX2_MIN_MTU;
+	netdev->max_mtu = OTX2_MAX_MTU;
+
+	INIT_WORK(&pf->reset_task, otx2_reset_task);
+
+	err = register_netdev(netdev);
+	if (err) {
+		dev_err(dev, "Failed to register netdevice\n");
+		goto err_detach_rsrc;
+	}
+
+	otx2_set_ethtool_ops(netdev);
+
+	/* Enable link notifications */
+	otx2_cgx_config_linkevents(pf, true);
+
+	return 0;
+
+err_detach_rsrc:
+	otx2_detach_resources(&pf->mbox);
+err_disable_mbox_intr:
+	otx2_disable_mbox_intr(pf);
+err_mbox_destroy:
+	otx2_pfaf_mbox_destroy(pf);
+err_free_irq_vectors:
+	pci_free_irq_vectors(hw->pdev);
+err_free_netdev:
+	pci_set_drvdata(pdev, NULL);
+	free_netdev(netdev);
+err_release_regions:
+	pci_release_regions(pdev);
+	return err;
+}
+
+static void otx2_remove(struct pci_dev *pdev)
+{
+	struct net_device *netdev = pci_get_drvdata(pdev);
+	struct otx2_nic *pf;
+
+	if (!netdev)
+		return;
+
+	pf = netdev_priv(netdev);
+
+	/* Disable link notifications */
+	otx2_cgx_config_linkevents(pf, false);
+
+	unregister_netdev(netdev);
+	otx2_detach_resources(&pf->mbox);
+	otx2_disable_mbox_intr(pf);
+	otx2_pfaf_mbox_destroy(pf);
+	pci_free_irq_vectors(pf->pdev);
+	pci_set_drvdata(pdev, NULL);
+	free_netdev(netdev);
+
+	pci_release_regions(pdev);
+}
+
+static struct pci_driver otx2_pf_driver = {
+	.name = DRV_NAME,
+	.id_table = otx2_pf_id_table,
+	.probe = otx2_probe,
+	.shutdown = otx2_remove,
+	.remove = otx2_remove,
+};
+
+static int __init otx2_rvupf_init_module(void)
+{
+	pr_info("%s: %s\n", DRV_NAME, DRV_STRING);
+
+	return pci_register_driver(&otx2_pf_driver);
+}
+
+static void __exit otx2_rvupf_cleanup_module(void)
+{
+	pci_unregister_driver(&otx2_pf_driver);
+}
+
+module_init(otx2_rvupf_init_module);
+module_exit(otx2_rvupf_cleanup_module);

drivers/net/ethernet/marvell/octeontx2/nic/otx2_reg.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Marvell OcteonTx2 RVU Ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef OTX2_REG_H
+#define OTX2_REG_H
+
+#include <rvu_struct.h>
+
+/* RVU PF registers */
+#define	RVU_PF_VFX_PFVF_MBOX0		    (0x00000)
+#define	RVU_PF_VFX_PFVF_MBOX1		    (0x00008)
+#define RVU_PF_VFX_PFVF_MBOXX(a, b)         (0x0 | (a) << 12 | (b) << 3)
+#define RVU_PF_VF_BAR4_ADDR                 (0x10)
+#define RVU_PF_BLOCK_ADDRX_DISC(a)          (0x200 | (a) << 3)
+#define RVU_PF_VFME_STATUSX(a)              (0x800 | (a) << 3)
+#define RVU_PF_VFTRPENDX(a)                 (0x820 | (a) << 3)
+#define RVU_PF_VFTRPEND_W1SX(a)             (0x840 | (a) << 3)
+#define RVU_PF_VFPF_MBOX_INTX(a)            (0x880 | (a) << 3)
+#define RVU_PF_VFPF_MBOX_INT_W1SX(a)        (0x8A0 | (a) << 3)
+#define RVU_PF_VFPF_MBOX_INT_ENA_W1SX(a)    (0x8C0 | (a) << 3)
+#define RVU_PF_VFPF_MBOX_INT_ENA_W1CX(a)    (0x8E0 | (a) << 3)
+#define RVU_PF_VFFLR_INTX(a)                (0x900 | (a) << 3)
+#define RVU_PF_VFFLR_INT_W1SX(a)            (0x920 | (a) << 3)
+#define RVU_PF_VFFLR_INT_ENA_W1SX(a)        (0x940 | (a) << 3)
+#define RVU_PF_VFFLR_INT_ENA_W1CX(a)        (0x960 | (a) << 3)
+#define RVU_PF_VFME_INTX(a)                 (0x980 | (a) << 3)
+#define RVU_PF_VFME_INT_W1SX(a)             (0x9A0 | (a) << 3)
+#define RVU_PF_VFME_INT_ENA_W1SX(a)         (0x9C0 | (a) << 3)
+#define RVU_PF_VFME_INT_ENA_W1CX(a)         (0x9E0 | (a) << 3)
+#define RVU_PF_PFAF_MBOX0                   (0xC00)
+#define RVU_PF_PFAF_MBOX1                   (0xC08)
+#define RVU_PF_PFAF_MBOXX(a)                (0xC00 | (a) << 3)
+#define RVU_PF_INT                          (0xc20)
+#define RVU_PF_INT_W1S                      (0xc28)
+#define RVU_PF_INT_ENA_W1S                  (0xc30)
+#define RVU_PF_INT_ENA_W1C                  (0xc38)
+#define RVU_PF_MSIX_VECX_ADDR(a)            (0x000 | (a) << 4)
+#define RVU_PF_MSIX_VECX_CTL(a)             (0x008 | (a) << 4)
+#define RVU_PF_MSIX_PBAX(a)                 (0xF0000 | (a) << 3)
+
+#define RVU_FUNC_BLKADDR_SHIFT		20
+#define RVU_FUNC_BLKADDR_MASK		0x1FULL
+
+/* NPA LF registers */
+#define NPA_LFBASE			(BLKTYPE_NPA << RVU_FUNC_BLKADDR_SHIFT)
+#define NPA_LF_AURA_OP_ALLOCX(a)	(NPA_LFBASE | 0x10 | (a) << 3)
+#define NPA_LF_AURA_OP_FREE0            (NPA_LFBASE | 0x20)
+#define NPA_LF_AURA_OP_FREE1            (NPA_LFBASE | 0x28)
+#define NPA_LF_AURA_OP_CNT              (NPA_LFBASE | 0x30)
+#define NPA_LF_AURA_OP_LIMIT            (NPA_LFBASE | 0x50)
+#define NPA_LF_AURA_OP_INT              (NPA_LFBASE | 0x60)
+#define NPA_LF_AURA_OP_THRESH           (NPA_LFBASE | 0x70)
+#define NPA_LF_POOL_OP_PC               (NPA_LFBASE | 0x100)
+#define NPA_LF_POOL_OP_AVAILABLE        (NPA_LFBASE | 0x110)
+#define NPA_LF_POOL_OP_PTR_START0       (NPA_LFBASE | 0x120)
+#define NPA_LF_POOL_OP_PTR_START1       (NPA_LFBASE | 0x128)
+#define NPA_LF_POOL_OP_PTR_END0         (NPA_LFBASE | 0x130)
+#define NPA_LF_POOL_OP_PTR_END1         (NPA_LFBASE | 0x138)
+#define NPA_LF_POOL_OP_INT              (NPA_LFBASE | 0x160)
+#define NPA_LF_POOL_OP_THRESH           (NPA_LFBASE | 0x170)
+#define NPA_LF_ERR_INT                  (NPA_LFBASE | 0x200)
+#define NPA_LF_ERR_INT_W1S              (NPA_LFBASE | 0x208)
+#define NPA_LF_ERR_INT_ENA_W1C          (NPA_LFBASE | 0x210)
+#define NPA_LF_ERR_INT_ENA_W1S          (NPA_LFBASE | 0x218)
+#define NPA_LF_RAS                      (NPA_LFBASE | 0x220)
+#define NPA_LF_RAS_W1S                  (NPA_LFBASE | 0x228)
+#define NPA_LF_RAS_ENA_W1C              (NPA_LFBASE | 0x230)
+#define NPA_LF_RAS_ENA_W1S              (NPA_LFBASE | 0x238)
+#define NPA_LF_QINTX_CNT(a)             (NPA_LFBASE | 0x300 | (a) << 12)
+#define NPA_LF_QINTX_INT(a)             (NPA_LFBASE | 0x310 | (a) << 12)
+#define NPA_LF_QINTX_INT_W1S(a)         (NPA_LFBASE | 0x318 | (a) << 12)
+#define NPA_LF_QINTX_ENA_W1S(a)         (NPA_LFBASE | 0x320 | (a) << 12)
+#define NPA_LF_QINTX_ENA_W1C(a)         (NPA_LFBASE | 0x330 | (a) << 12)
+
+/* NIX LF registers */
+#define	NIX_LFBASE			(BLKTYPE_NIX << RVU_FUNC_BLKADDR_SHIFT)
+#define	NIX_LF_RX_SECRETX(a)		(NIX_LFBASE | 0x0 | (a) << 3)
+#define	NIX_LF_CFG			(NIX_LFBASE | 0x100)
+#define	NIX_LF_GINT			(NIX_LFBASE | 0x200)
+#define	NIX_LF_GINT_W1S			(NIX_LFBASE | 0x208)
+#define	NIX_LF_GINT_ENA_W1C		(NIX_LFBASE | 0x210)
+#define	NIX_LF_GINT_ENA_W1S		(NIX_LFBASE | 0x218)
+#define	NIX_LF_ERR_INT			(NIX_LFBASE | 0x220)
+#define	NIX_LF_ERR_INT_W1S		(NIX_LFBASE | 0x228)
+#define	NIX_LF_ERR_INT_ENA_W1C		(NIX_LFBASE | 0x230)
+#define	NIX_LF_ERR_INT_ENA_W1S		(NIX_LFBASE | 0x238)
+#define	NIX_LF_RAS			(NIX_LFBASE | 0x240)
+#define	NIX_LF_RAS_W1S			(NIX_LFBASE | 0x248)
+#define	NIX_LF_RAS_ENA_W1C		(NIX_LFBASE | 0x250)
+#define	NIX_LF_RAS_ENA_W1S		(NIX_LFBASE | 0x258)
+#define	NIX_LF_SQ_OP_ERR_DBG		(NIX_LFBASE | 0x260)
+#define	NIX_LF_MNQ_ERR_DBG		(NIX_LFBASE | 0x270)
+#define	NIX_LF_SEND_ERR_DBG		(NIX_LFBASE | 0x280)
+#define	NIX_LF_TX_STATX(a)		(NIX_LFBASE | 0x300 | (a) << 3)
+#define	NIX_LF_RX_STATX(a)		(NIX_LFBASE | 0x400 | (a) << 3)
+#define	NIX_LF_OP_SENDX(a)		(NIX_LFBASE | 0x800 | (a) << 3)
+#define	NIX_LF_RQ_OP_INT		(NIX_LFBASE | 0x900)
+#define	NIX_LF_RQ_OP_OCTS		(NIX_LFBASE | 0x910)
+#define	NIX_LF_RQ_OP_PKTS		(NIX_LFBASE | 0x920)
+#define	NIX_LF_OP_IPSEC_DYNO_CN		(NIX_LFBASE | 0x980)
+#define	NIX_LF_SQ_OP_INT		(NIX_LFBASE | 0xa00)
+#define	NIX_LF_SQ_OP_OCTS		(NIX_LFBASE | 0xa10)
+#define	NIX_LF_SQ_OP_PKTS		(NIX_LFBASE | 0xa20)
+#define	NIX_LF_SQ_OP_STATUS		(NIX_LFBASE | 0xa30)
+#define	NIX_LF_CQ_OP_INT		(NIX_LFBASE | 0xb00)
+#define	NIX_LF_CQ_OP_DOOR		(NIX_LFBASE | 0xb30)
+#define	NIX_LF_CQ_OP_STATUS		(NIX_LFBASE | 0xb40)
+#define	NIX_LF_QINTX_CNT(a)		(NIX_LFBASE | 0xC00 | (a) << 12)
+#define	NIX_LF_QINTX_INT(a)		(NIX_LFBASE | 0xC10 | (a) << 12)
+#define	NIX_LF_QINTX_INT_W1S(a)		(NIX_LFBASE | 0xC18 | (a) << 12)
+#define	NIX_LF_QINTX_ENA_W1S(a)		(NIX_LFBASE | 0xC20 | (a) << 12)
+#define	NIX_LF_QINTX_ENA_W1C(a)		(NIX_LFBASE | 0xC30 | (a) << 12)
+#define	NIX_LF_CINTX_CNT(a)		(NIX_LFBASE | 0xD00 | (a) << 12)
+#define	NIX_LF_CINTX_WAIT(a)		(NIX_LFBASE | 0xD10 | (a) << 12)
+#define	NIX_LF_CINTX_INT(a)		(NIX_LFBASE | 0xD20 | (a) << 12)
+#define	NIX_LF_CINTX_INT_W1S(a)		(NIX_LFBASE | 0xD30 | (a) << 12)
+#define	NIX_LF_CINTX_ENA_W1S(a)		(NIX_LFBASE | 0xD40 | (a) << 12)
+#define	NIX_LF_CINTX_ENA_W1C(a)		(NIX_LFBASE | 0xD50 | (a) << 12)
+
+/* NIX AF transmit scheduler registers */
+#define NIX_AF_SMQX_CFG(a)		(0x700 | (a) << 16)
+#define NIX_AF_TL1X_SCHEDULE(a)		(0xC00 | (a) << 16)
+#define NIX_AF_TL1X_CIR(a)		(0xC20 | (a) << 16)
+#define NIX_AF_TL1X_TOPOLOGY(a)		(0xC80 | (a) << 16)
+#define NIX_AF_TL2X_PARENT(a)		(0xE88 | (a) << 16)
+#define NIX_AF_TL2X_SCHEDULE(a)		(0xE00 | (a) << 16)
+#define NIX_AF_TL3X_PARENT(a)		(0x1088 | (a) << 16)
+#define NIX_AF_TL3X_SCHEDULE(a)		(0x1000 | (a) << 16)
+#define NIX_AF_TL4X_PARENT(a)		(0x1288 | (a) << 16)
+#define NIX_AF_TL4X_SCHEDULE(a)		(0x1200 | (a) << 16)
+#define NIX_AF_MDQX_SCHEDULE(a)		(0x1400 | (a) << 16)
+#define NIX_AF_MDQX_PARENT(a)		(0x1480 | (a) << 16)
+#define NIX_AF_TL3_TL2X_LINKX_CFG(a, b)	(0x1700 | (a) << 16 | (b) << 3)
+
+/* LMT LF registers */
+#define LMT_LFBASE			BIT_ULL(RVU_FUNC_BLKADDR_SHIFT)
+#define LMT_LF_LMTLINEX(a)		(LMT_LFBASE | 0x000 | (a) << 12)
+#define LMT_LF_LMTCANCEL		(LMT_LFBASE | 0x400)
+
+#endif /* OTX2_REG_H */

drivers/net/ethernet/marvell/octeontx2/nic/otx2_struct.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Marvell OcteonTx2 RVU Ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef OTX2_STRUCT_H
+#define OTX2_STRUCT_H
+
+/* NIX WQE/CQE size 128 byte or 512 byte */
+enum nix_cqesz_e {
+	NIX_XQESZ_W64 = 0x0,
+	NIX_XQESZ_W16 = 0x1,
+};
+
+enum nix_sqes_e {
+	NIX_SQESZ_W16 = 0x0,
+	NIX_SQESZ_W8 = 0x1,
+};
+
+enum nix_send_ldtype {
+	NIX_SEND_LDTYPE_LDD  = 0x0,
+	NIX_SEND_LDTYPE_LDT  = 0x1,
+	NIX_SEND_LDTYPE_LDWB = 0x2,
+};
+
+/* CSUM offload */
+enum nix_sendl3type {
+	NIX_SENDL3TYPE_NONE = 0x0,
+	NIX_SENDL3TYPE_IP4 = 0x2,
+	NIX_SENDL3TYPE_IP4_CKSUM = 0x3,
+	NIX_SENDL3TYPE_IP6 = 0x4,
+};
+
+enum nix_sendl4type {
+	NIX_SENDL4TYPE_NONE,
+	NIX_SENDL4TYPE_TCP_CKSUM,
+	NIX_SENDL4TYPE_SCTP_CKSUM,
+	NIX_SENDL4TYPE_UDP_CKSUM,
+};
+
+/* NIX wqe/cqe types */
+enum nix_xqe_type {
+	NIX_XQE_TYPE_INVALID   = 0x0,
+	NIX_XQE_TYPE_RX        = 0x1,
+	NIX_XQE_TYPE_RX_IPSECS = 0x2,
+	NIX_XQE_TYPE_RX_IPSECH = 0x3,
+	NIX_XQE_TYPE_RX_IPSECD = 0x4,
+	NIX_XQE_TYPE_SEND      = 0x8,
+};
+
+/* NIX CQE/SQE subdescriptor types */
+enum nix_subdc {
+	NIX_SUBDC_NOP  = 0x0,
+	NIX_SUBDC_EXT  = 0x1,
+	NIX_SUBDC_CRC  = 0x2,
+	NIX_SUBDC_IMM  = 0x3,
+	NIX_SUBDC_SG   = 0x4,
+	NIX_SUBDC_MEM  = 0x5,
+	NIX_SUBDC_JUMP = 0x6,
+	NIX_SUBDC_WORK = 0x7,
+	NIX_SUBDC_SOD  = 0xf,
+};
+
+/* Algorithm for nix_sqe_mem_s header (value of the `alg` field) */
+enum nix_sendmemalg {
+	NIX_SENDMEMALG_E_SET       = 0x0,
+	NIX_SENDMEMALG_E_SETTSTMP  = 0x1,
+	NIX_SENDMEMALG_E_SETRSLT   = 0x2,
+	NIX_SENDMEMALG_E_ADD       = 0x8,
+	NIX_SENDMEMALG_E_SUB       = 0x9,
+	NIX_SENDMEMALG_E_ADDLEN    = 0xa,
+	NIX_SENDMEMALG_E_SUBLEN    = 0xb,
+	NIX_SENDMEMALG_E_ADDMBUF   = 0xc,
+	NIX_SENDMEMALG_E_SUBMBUF   = 0xd,
+	NIX_SENDMEMALG_E_ENUM_LAST = 0xe,
+};
+
+/* NIX CQE header structure */
+struct nix_cqe_hdr_s {
+	u64 flow_tag              : 32;
+	u64 q                     : 20;
+	u64 reserved_52_57        : 6;
+	u64 node                  : 2;
+	u64 cqe_type              : 4;
+};
+
+/* NIX CQE RX parse structure */
+struct nix_rx_parse_s {
+	u64 chan         : 12;
+	u64 desc_sizem1  : 5;
+	u64 rsvd_17      : 1;
+	u64 express      : 1;
+	u64 wqwd         : 1;
+	u64 errlev       : 4;
+	u64 errcode      : 8;
+	u64 latype       : 4;
+	u64 lbtype       : 4;
+	u64 lctype       : 4;
+	u64 ldtype       : 4;
+	u64 letype       : 4;
+	u64 lftype       : 4;
+	u64 lgtype       : 4;
+	u64 lhtype       : 4;
+	u64 pkt_lenm1    : 16; /* W1 */
+	u64 l2m          : 1;
+	u64 l2b          : 1;
+	u64 l3m          : 1;
+	u64 l3b          : 1;
+	u64 vtag0_valid  : 1;
+	u64 vtag0_gone   : 1;
+	u64 vtag1_valid  : 1;
+	u64 vtag1_gone   : 1;
+	u64 pkind        : 6;
+	u64 rsvd_95_94   : 2;
+	u64 vtag0_tci    : 16;
+	u64 vtag1_tci    : 16;
+	u64 laflags      : 8; /* W2 */
+	u64 lbflags      : 8;
+	u64 lcflags      : 8;
+	u64 ldflags      : 8;
+	u64 leflags      : 8;
+	u64 lfflags      : 8;
+	u64 lgflags      : 8;
+	u64 lhflags      : 8;
+	u64 eoh_ptr      : 8; /* W3 */
+	u64 wqe_aura     : 20;
+	u64 pb_aura      : 20;
+	u64 match_id     : 16;
+	u64 laptr        : 8; /* W4 */
+	u64 lbptr        : 8;
+	u64 lcptr        : 8;
+	u64 ldptr        : 8;
+	u64 leptr        : 8;
+	u64 lfptr        : 8;
+	u64 lgptr        : 8;
+	u64 lhptr        : 8;
+	u64 vtag0_ptr    : 8; /* W5 */
+	u64 vtag1_ptr    : 8;
+	u64 flow_key_alg : 5;
+	u64 rsvd_383_341 : 43;
+	u64 rsvd_447_384;     /* W6 */
+};
+
+/* NIX CQE RX scatter/gather subdescriptor structure */
+struct nix_rx_sg_s {
+	u64 seg_size   : 16; /* W0 */
+	u64 seg2_size  : 16;
+	u64 seg3_size  : 16;
+	u64 segs       : 2;
+	u64 rsvd_59_50 : 10;
+	u64 subdc      : 4;
+	u64 seg_addr;
+	u64 seg2_addr;
+	u64 seg3_addr;
+};
+
+struct nix_send_comp_s {
+	u64 status	: 8;
+	u64 sqe_id	: 16;
+	u64 rsvd_24_63	: 40;
+};
+
+struct nix_cqe_rx_s {
+	struct nix_cqe_hdr_s  hdr;
+	struct nix_rx_parse_s parse;
+	struct nix_rx_sg_s sg;
+};
+
+struct nix_cqe_tx_s {
+	struct nix_cqe_hdr_s  hdr;
+	struct nix_send_comp_s comp;
+};
+
+/* NIX SQE header structure */
+struct nix_sqe_hdr_s {
+	u64 total		: 18; /* W0 */
+	u64 reserved_18		: 1;
+	u64 df			: 1;
+	u64 aura		: 20;
+	u64 sizem1		: 3;
+	u64 pnc			: 1;
+	u64 sq			: 20;
+	u64 ol3ptr		: 8; /* W1 */
+	u64 ol4ptr		: 8;
+	u64 il3ptr		: 8;
+	u64 il4ptr		: 8;
+	u64 ol3type		: 4;
+	u64 ol4type		: 4;
+	u64 il3type		: 4;
+	u64 il4type		: 4;
+	u64 sqe_id		: 16;
+
+};
+
+/* NIX send extended header subdescriptor structure */
+struct nix_sqe_ext_s {
+	u64 lso_mps       : 14; /* W0 */
+	u64 lso           : 1;
+	u64 tstmp         : 1;
+	u64 lso_sb        : 8;
+	u64 lso_format    : 5;
+	u64 rsvd_31_29    : 3;
+	u64 shp_chg       : 9;
+	u64 shp_dis       : 1;
+	u64 shp_ra        : 2;
+	u64 markptr       : 8;
+	u64 markform      : 7;
+	u64 mark_en       : 1;
+	u64 subdc         : 4;
+	u64 vlan0_ins_ptr : 8; /* W1 */
+	u64 vlan0_ins_tci : 16;
+	u64 vlan1_ins_ptr : 8;
+	u64 vlan1_ins_tci : 16;
+	u64 vlan0_ins_ena : 1;
+	u64 vlan1_ins_ena : 1;
+	u64 rsvd_127_114  : 14;
+};
+
+struct nix_sqe_sg_s {
+	u64 seg1_size	: 16;
+	u64 seg2_size	: 16;
+	u64 seg3_size	: 16;
+	u64 segs	: 2;
+	u64 rsvd_54_50	: 5;
+	u64 i1		: 1;
+	u64 i2		: 1;
+	u64 i3		: 1;
+	u64 ld_type	: 2;
+	u64 subdc	: 4;
+};
+
+/* NIX send memory subdescriptor structure */
+struct nix_sqe_mem_s {
+	u64 offset        : 16; /* W0 */
+	u64 rsvd_52_16    : 37;
+	u64 wmem          : 1;
+	u64 dsz           : 2;
+	u64 alg           : 4;
+	u64 subdc         : 4;
+	u64 addr; /* W1 */
+};
+
+enum nix_cqerrint_e {
+	NIX_CQERRINT_DOOR_ERR = 0,
+	NIX_CQERRINT_WR_FULL = 1,
+	NIX_CQERRINT_CQE_FAULT = 2,
+};
+
+#define NIX_CQERRINT_BITS (BIT_ULL(NIX_CQERRINT_DOOR_ERR) | \
+			   BIT_ULL(NIX_CQERRINT_CQE_FAULT))
+
+enum nix_rqint_e {
+	NIX_RQINT_DROP = 0,
+	NIX_RQINT_RED = 1,
+};
+
+#define NIX_RQINT_BITS (BIT_ULL(NIX_RQINT_DROP) | BIT_ULL(NIX_RQINT_RED))
+
+enum nix_sqint_e {
+	NIX_SQINT_LMT_ERR = 0,
+	NIX_SQINT_MNQ_ERR = 1,
+	NIX_SQINT_SEND_ERR = 2,
+	NIX_SQINT_SQB_ALLOC_FAIL = 3,
+};
+
+#define NIX_SQINT_BITS (BIT_ULL(NIX_SQINT_LMT_ERR) | \
+			BIT_ULL(NIX_SQINT_MNQ_ERR) | \
+			BIT_ULL(NIX_SQINT_SEND_ERR) | \
+			BIT_ULL(NIX_SQINT_SQB_ALLOC_FAIL))
+
+#endif /* OTX2_STRUCT_H */

drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c

+// SPDX-License-Identifier: GPL-2.0
+/* Marvell OcteonTx2 RVU Ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/etherdevice.h>
+#include <net/ip.h>
+#include <net/tso.h>
+
+#include "otx2_reg.h"
+#include "otx2_common.h"
+#include "otx2_struct.h"
+#include "otx2_txrx.h"
+
+#define CQE_ADDR(CQ, idx) ((CQ)->cqe_base + ((CQ)->cqe_size * (idx)))
+
+static struct nix_cqe_hdr_s *otx2_get_next_cqe(struct otx2_cq_queue *cq)
+{
+	struct nix_cqe_hdr_s *cqe_hdr;
+
+	cqe_hdr = (struct nix_cqe_hdr_s *)CQE_ADDR(cq, cq->cq_head);
+	if (cqe_hdr->cqe_type == NIX_XQE_TYPE_INVALID)
+		return NULL;
+
+	cq->cq_head++;
+	cq->cq_head &= (cq->cqe_cnt - 1);
+
+	return cqe_hdr;
+}
+
+static unsigned int frag_num(unsigned int i)
+{
+#ifdef __BIG_ENDIAN
+	return (i & ~3) + 3 - (i & 3);
+#else
+	return i;
+#endif
+}
+
+static dma_addr_t otx2_dma_map_skb_frag(struct otx2_nic *pfvf,
+					struct sk_buff *skb, int seg, int *len)
+{
+	const skb_frag_t *frag;
+	struct page *page;
+	int offset;
+
+	/* First segment is always skb->data */
+	if (!seg) {
+		page = virt_to_page(skb->data);
+		offset = offset_in_page(skb->data);
+		*len = skb_headlen(skb);
+	} else {
+		frag = &skb_shinfo(skb)->frags[seg - 1];
+		page = skb_frag_page(frag);
+		offset = skb_frag_off(frag);
+		*len = skb_frag_size(frag);
+	}
+	return otx2_dma_map_page(pfvf, page, offset, *len, DMA_TO_DEVICE);
+}
+
+static void otx2_dma_unmap_skb_frags(struct otx2_nic *pfvf, struct sg_list *sg)
+{
+	int seg;
+
+	for (seg = 0; seg < sg->num_segs; seg++) {
+		otx2_dma_unmap_page(pfvf, sg->dma_addr[seg],
+				    sg->size[seg], DMA_TO_DEVICE);
+	}
+	sg->num_segs = 0;
+}
+
+static void otx2_snd_pkt_handler(struct otx2_nic *pfvf,
+				 struct otx2_cq_queue *cq,
+				 struct otx2_snd_queue *sq,
+				 struct nix_cqe_tx_s *cqe,
+				 int budget, int *tx_pkts, int *tx_bytes)
+{
+	struct nix_send_comp_s *snd_comp = &cqe->comp;
+	struct sk_buff *skb = NULL;
+	struct sg_list *sg;
+
+	if (unlikely(snd_comp->status) && netif_msg_tx_err(pfvf))
+		net_err_ratelimited("%s: TX%d: Error in send CQ status:%x\n",
+				    pfvf->netdev->name, cq->cint_idx,
+				    snd_comp->status);
+
+	sg = &sq->sg[snd_comp->sqe_id];
+	skb = (struct sk_buff *)sg->skb;
+	if (unlikely(!skb))
+		return;
+
+	*tx_bytes += skb->len;
+	(*tx_pkts)++;
+	otx2_dma_unmap_skb_frags(pfvf, sg);
+	napi_consume_skb(skb, budget);
+	sg->skb = (u64)NULL;
+}
+
+static void otx2_skb_add_frag(struct otx2_nic *pfvf, struct sk_buff *skb,
+			      u64 iova, int len)
+{
+	struct page *page;
+	void *va;
+
+	va = phys_to_virt(otx2_iova_to_phys(pfvf->iommu_domain, iova));
+	page = virt_to_page(va);
+	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
+			va - page_address(page), len, pfvf->rbsize);
+
+	otx2_dma_unmap_page(pfvf, iova - OTX2_HEAD_ROOM,
+			    pfvf->rbsize, DMA_FROM_DEVICE);
+}
+
+static void otx2_set_rxhash(struct otx2_nic *pfvf,
+			    struct nix_cqe_rx_s *cqe, struct sk_buff *skb)
+{
+	enum pkt_hash_types hash_type = PKT_HASH_TYPE_NONE;
+	struct otx2_rss_info *rss;
+	u32 hash = 0;
+
+	if (!(pfvf->netdev->features & NETIF_F_RXHASH))
+		return;
+
+	rss = &pfvf->hw.rss_info;
+	if (rss->flowkey_cfg) {
+		if (rss->flowkey_cfg &
+		    ~(NIX_FLOW_KEY_TYPE_IPV4 | NIX_FLOW_KEY_TYPE_IPV6))
+			hash_type = PKT_HASH_TYPE_L4;
+		else
+			hash_type = PKT_HASH_TYPE_L3;
+		hash = cqe->hdr.flow_tag;
+	}
+	skb_set_hash(skb, hash, hash_type);
+}
+
+static bool otx2_check_rcv_errors(struct otx2_nic *pfvf,
+				  struct nix_cqe_rx_s *cqe, int qidx)
+{
+	struct otx2_drv_stats *stats = &pfvf->hw.drv_stats;
+	struct nix_rx_parse_s *parse = &cqe->parse;
+
+	if (netif_msg_rx_err(pfvf))
+		netdev_err(pfvf->netdev,
+			   "RQ%d: Error pkt with errlev:0x%x errcode:0x%x\n",
+			   qidx, parse->errlev, parse->errcode);
+
+	if (parse->errlev == NPC_ERRLVL_RE) {
+		switch (parse->errcode) {
+		case ERRCODE_FCS:
+		case ERRCODE_FCS_RCV:
+			atomic_inc(&stats->rx_fcs_errs);
+			break;
+		case ERRCODE_UNDERSIZE:
+			atomic_inc(&stats->rx_undersize_errs);
+			break;
+		case ERRCODE_OVERSIZE:
+			atomic_inc(&stats->rx_oversize_errs);
+			break;
+		case ERRCODE_OL2_LEN_MISMATCH:
+			atomic_inc(&stats->rx_len_errs);
+			break;
+		default:
+			atomic_inc(&stats->rx_other_errs);
+			break;
+		}
+	} else if (parse->errlev == NPC_ERRLVL_NIX) {
+		switch (parse->errcode) {
+		case ERRCODE_OL3_LEN:
+		case ERRCODE_OL4_LEN:
+		case ERRCODE_IL3_LEN:
+		case ERRCODE_IL4_LEN:
+			atomic_inc(&stats->rx_len_errs);
+			break;
+		case ERRCODE_OL4_CSUM:
+		case ERRCODE_IL4_CSUM:
+			atomic_inc(&stats->rx_csum_errs);
+			break;
+		default:
+			atomic_inc(&stats->rx_other_errs);
+			break;
+		}
+	} else {
+		atomic_inc(&stats->rx_other_errs);
+		/* For now ignore all the NPC parser errors and
+		 * pass the packets to stack.
+		 */
+		return false;
+	}
+
+	/* If RXALL is enabled pass on packets to stack. */
+	if (cqe->sg.segs && (pfvf->netdev->features & NETIF_F_RXALL))
+		return false;
+
+	/* Free buffer back to pool */
+	if (cqe->sg.segs)
+		otx2_aura_freeptr(pfvf, qidx, cqe->sg.seg_addr & ~0x07ULL);
+	return true;
+}
+
+static void otx2_rcv_pkt_handler(struct otx2_nic *pfvf,
+				 struct napi_struct *napi,
+				 struct otx2_cq_queue *cq,
+				 struct nix_cqe_rx_s *cqe)
+{
+	struct nix_rx_parse_s *parse = &cqe->parse;
+	struct sk_buff *skb = NULL;
+
+	if (unlikely(parse->errlev || parse->errcode)) {
+		if (otx2_check_rcv_errors(pfvf, cqe, cq->cq_idx))
+			return;
+	}
+
+	skb = napi_get_frags(napi);
+	if (unlikely(!skb))
+		return;
+
+	otx2_skb_add_frag(pfvf, skb, cqe->sg.seg_addr, cqe->sg.seg_size);
+	cq->pool_ptrs++;
+
+	otx2_set_rxhash(pfvf, cqe, skb);
+
+	skb_record_rx_queue(skb, cq->cq_idx);
+	if (pfvf->netdev->features & NETIF_F_RXCSUM)
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
+
+	napi_gro_frags(napi);
+}
+
+static int otx2_rx_napi_handler(struct otx2_nic *pfvf,
+				struct napi_struct *napi,
+				struct otx2_cq_queue *cq, int budget)
+{
+	struct nix_cqe_rx_s *cqe;
+	int processed_cqe = 0;
+	s64 bufptr;
+
+	while (likely(processed_cqe < budget)) {
+		cqe = (struct nix_cqe_rx_s *)CQE_ADDR(cq, cq->cq_head);
+		if (cqe->hdr.cqe_type == NIX_XQE_TYPE_INVALID ||
+		    !cqe->sg.seg_addr) {
+			if (!processed_cqe)
+				return 0;
+			break;
+		}
+		cq->cq_head++;
+		cq->cq_head &= (cq->cqe_cnt - 1);
+
+		otx2_rcv_pkt_handler(pfvf, napi, cq, cqe);
+
+		cqe->hdr.cqe_type = NIX_XQE_TYPE_INVALID;
+		cqe->sg.seg_addr = 0x00;
+		processed_cqe++;
+	}
+
+	/* Free CQEs to HW */
+	otx2_write64(pfvf, NIX_LF_CQ_OP_DOOR,
+		     ((u64)cq->cq_idx << 32) | processed_cqe);
+
+	if (unlikely(!cq->pool_ptrs))
+		return 0;
+
+	/* Refill pool with new buffers */
+	while (cq->pool_ptrs) {
+		bufptr = otx2_alloc_rbuf(pfvf, cq->rbpool, GFP_ATOMIC);
+		if (unlikely(bufptr <= 0)) {
+			struct refill_work *work;
+			struct delayed_work *dwork;
+
+			work = &pfvf->refill_wrk[cq->cq_idx];
+			dwork = &work->pool_refill_work;
+			/* Schedule a task if no other task is running */
+			if (!cq->refill_task_sched) {
+				cq->refill_task_sched = true;
+				schedule_delayed_work(dwork,
+						      msecs_to_jiffies(100));
+			}
+			break;
+		}
+		otx2_aura_freeptr(pfvf, cq->cq_idx, bufptr + OTX2_HEAD_ROOM);
+		cq->pool_ptrs--;
+	}
+
+	return processed_cqe;
+}
+
+static int otx2_tx_napi_handler(struct otx2_nic *pfvf,
+				struct otx2_cq_queue *cq, int budget)
+{
+	int tx_pkts = 0, tx_bytes = 0;
+	struct nix_cqe_tx_s *cqe;
+	int processed_cqe = 0;
+
+	while (likely(processed_cqe < budget)) {
+		cqe = (struct nix_cqe_tx_s *)otx2_get_next_cqe(cq);
+		if (unlikely(!cqe)) {
+			if (!processed_cqe)
+				return 0;
+			break;
+		}
+		otx2_snd_pkt_handler(pfvf, cq, &pfvf->qset.sq[cq->cint_idx],
+				     cqe, budget, &tx_pkts, &tx_bytes);
+
+		cqe->hdr.cqe_type = NIX_XQE_TYPE_INVALID;
+		processed_cqe++;
+	}
+
+	/* Free CQEs to HW */
+	otx2_write64(pfvf, NIX_LF_CQ_OP_DOOR,
+		     ((u64)cq->cq_idx << 32) | processed_cqe);
+
+	if (likely(tx_pkts)) {
+		struct netdev_queue *txq;
+
+		txq = netdev_get_tx_queue(pfvf->netdev, cq->cint_idx);
+		netdev_tx_completed_queue(txq, tx_pkts, tx_bytes);
+		/* Check if queue was stopped earlier due to ring full */
+		smp_mb();
+		if (netif_tx_queue_stopped(txq) &&
+		    netif_carrier_ok(pfvf->netdev))
+			netif_tx_wake_queue(txq);
+	}
+	return 0;
+}
+
+int otx2_napi_handler(struct napi_struct *napi, int budget)
+{
+	struct otx2_cq_poll *cq_poll;
+	int workdone = 0, cq_idx, i;
+	struct otx2_cq_queue *cq;
+	struct otx2_qset *qset;
+	struct otx2_nic *pfvf;
+
+	cq_poll = container_of(napi, struct otx2_cq_poll, napi);
+	pfvf = (struct otx2_nic *)cq_poll->dev;
+	qset = &pfvf->qset;
+
+	for (i = CQS_PER_CINT - 1; i >= 0; i--) {
+		cq_idx = cq_poll->cq_ids[i];
+		if (unlikely(cq_idx == CINT_INVALID_CQ))
+			continue;
+		cq = &qset->cq[cq_idx];
+		if (cq->cq_type == CQ_RX) {
+			/* If the RQ refill WQ task is running, skip napi
+			 * scheduler for this queue.
+			 */
+			if (cq->refill_task_sched)
+				continue;
+			workdone += otx2_rx_napi_handler(pfvf, napi,
+							 cq, budget);
+		} else {
+			workdone += otx2_tx_napi_handler(pfvf, cq, budget);
+		}
+	}
+
+	/* Clear the IRQ */
+	otx2_write64(pfvf, NIX_LF_CINTX_INT(cq_poll->cint_idx), BIT_ULL(0));
+
+	if (workdone < budget && napi_complete_done(napi, workdone)) {
+		/* If interface is going down, don't re-enable IRQ */
+		if (pfvf->flags & OTX2_FLAG_INTF_DOWN)
+			return workdone;
+
+		/* Re-enable interrupts */
+		otx2_write64(pfvf, NIX_LF_CINTX_ENA_W1S(cq_poll->cint_idx),
+			     BIT_ULL(0));
+	}
+	return workdone;
+}
+
+static void otx2_sqe_flush(struct otx2_snd_queue *sq, int size)
+{
+	u64 status;
+
+	/* Packet data stores should finish before SQE is flushed to HW */
+	dma_wmb();
+
+	do {
+		memcpy(sq->lmt_addr, sq->sqe_base, size);
+		status = otx2_lmt_flush(sq->io_addr);
+	} while (status == 0);
+
+	sq->head++;
+	sq->head &= (sq->sqe_cnt - 1);
+}
+
+#define MAX_SEGS_PER_SG	3
+/* Add SQE scatter/gather subdescriptor structure */
+static bool otx2_sqe_add_sg(struct otx2_nic *pfvf, struct otx2_snd_queue *sq,
+			    struct sk_buff *skb, int num_segs, int *offset)
+{
+	struct nix_sqe_sg_s *sg = NULL;
+	u64 dma_addr, *iova = NULL;
+	u16 *sg_lens = NULL;
+	int seg, len;
+
+	sq->sg[sq->head].num_segs = 0;
+
+	for (seg = 0; seg < num_segs; seg++) {
+		if ((seg % MAX_SEGS_PER_SG) == 0) {
+			sg = (struct nix_sqe_sg_s *)(sq->sqe_base + *offset);
+			sg->ld_type = NIX_SEND_LDTYPE_LDD;
+			sg->subdc = NIX_SUBDC_SG;
+			sg->segs = 0;
+			sg_lens = (void *)sg;
+			iova = (void *)sg + sizeof(*sg);
+			/* Next subdc always starts at a 16byte boundary.
+			 * So if sg->segs is whether 2 or 3, offset += 16bytes.
+			 */
+			if ((num_segs - seg) >= (MAX_SEGS_PER_SG - 1))
+				*offset += sizeof(*sg) + (3 * sizeof(u64));
+			else
+				*offset += sizeof(*sg) + sizeof(u64);
+		}
+		dma_addr = otx2_dma_map_skb_frag(pfvf, skb, seg, &len);
+		if (dma_mapping_error(pfvf->dev, dma_addr))
+			return false;
+
+		sg_lens[frag_num(seg % MAX_SEGS_PER_SG)] = len;
+		sg->segs++;
+		*iova++ = dma_addr;
+
+		/* Save DMA mapping info for later unmapping */
+		sq->sg[sq->head].dma_addr[seg] = dma_addr;
+		sq->sg[sq->head].size[seg] = len;
+		sq->sg[sq->head].num_segs++;
+	}
+
+	sq->sg[sq->head].skb = (u64)skb;
+	return true;
+}
+
+/* Add SQE extended header subdescriptor */
+static void otx2_sqe_add_ext(struct otx2_nic *pfvf, struct otx2_snd_queue *sq,
+			     struct sk_buff *skb, int *offset)
+{
+	struct nix_sqe_ext_s *ext;
+
+	ext = (struct nix_sqe_ext_s *)(sq->sqe_base + *offset);
+	ext->subdc = NIX_SUBDC_EXT;
+	if (skb_shinfo(skb)->gso_size) {
+		ext->lso = 1;
+		ext->lso_sb = skb_transport_offset(skb) + tcp_hdrlen(skb);
+		ext->lso_mps = skb_shinfo(skb)->gso_size;
+
+		/* Only TSOv4 and TSOv6 GSO offloads are supported */
+		if (skb_shinfo(skb)->gso_type & SKB_GSO_TCPV4) {
+			ext->lso_format = pfvf->hw.lso_tsov4_idx;
+
+			/* HW adds payload size to 'ip_hdr->tot_len' while
+			 * sending TSO segment, hence set payload length
+			 * in IP header of the packet to just header length.
+			 */
+			ip_hdr(skb)->tot_len =
+				htons(ext->lso_sb - skb_network_offset(skb));
+		} else {
+			ext->lso_format = pfvf->hw.lso_tsov6_idx;
+			ipv6_hdr(skb)->payload_len =
+				htons(ext->lso_sb - skb_network_offset(skb));
+		}
+	}
+	*offset += sizeof(*ext);
+}
+
+/* Add SQE header subdescriptor structure */
+static void otx2_sqe_add_hdr(struct otx2_nic *pfvf, struct otx2_snd_queue *sq,
+			     struct nix_sqe_hdr_s *sqe_hdr,
+			     struct sk_buff *skb, u16 qidx)
+{
+	int proto = 0;
+
+	/* Check if SQE was framed before, if yes then no need to
+	 * set these constants again and again.
+	 */
+	if (!sqe_hdr->total) {
+		/* Don't free Tx buffers to Aura */
+		sqe_hdr->df = 1;
+		sqe_hdr->aura = sq->aura_id;
+		/* Post a CQE Tx after pkt transmission */
+		sqe_hdr->pnc = 1;
+		sqe_hdr->sq = qidx;
+	}
+	sqe_hdr->total = skb->len;
+	/* Set SQE identifier which will be used later for freeing SKB */
+	sqe_hdr->sqe_id = sq->head;
+
+	/* Offload TCP/UDP checksum to HW */
+	if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		sqe_hdr->ol3ptr = skb_network_offset(skb);
+		sqe_hdr->ol4ptr = skb_transport_offset(skb);
+		/* get vlan protocol Ethertype */
+		if (eth_type_vlan(skb->protocol))
+			skb->protocol = vlan_get_protocol(skb);
+
+		if (skb->protocol == htons(ETH_P_IP)) {
+			proto = ip_hdr(skb)->protocol;
+			/* In case of TSO, HW needs this to be explicitly set.
+			 * So set this always, instead of adding a check.
+			 */
+			sqe_hdr->ol3type = NIX_SENDL3TYPE_IP4_CKSUM;
+		} else if (skb->protocol == htons(ETH_P_IPV6)) {
+			proto = ipv6_hdr(skb)->nexthdr;
+		}
+
+		if (proto == IPPROTO_TCP)
+			sqe_hdr->ol4type = NIX_SENDL4TYPE_TCP_CKSUM;
+		else if (proto == IPPROTO_UDP)
+			sqe_hdr->ol4type = NIX_SENDL4TYPE_UDP_CKSUM;
+	}
+}
+
+static int otx2_dma_map_tso_skb(struct otx2_nic *pfvf,
+				struct otx2_snd_queue *sq,
+				struct sk_buff *skb, int sqe, int hdr_len)
+{
+	int num_segs = skb_shinfo(skb)->nr_frags + 1;
+	struct sg_list *sg = &sq->sg[sqe];
+	u64 dma_addr;
+	int seg, len;
+
+	sg->num_segs = 0;
+
+	/* Get payload length at skb->data */
+	len = skb_headlen(skb) - hdr_len;
+
+	for (seg = 0; seg < num_segs; seg++) {
+		/* Skip skb->data, if there is no payload */
+		if (!seg && !len)
+			continue;
+		dma_addr = otx2_dma_map_skb_frag(pfvf, skb, seg, &len);
+		if (dma_mapping_error(pfvf->dev, dma_addr))
+			goto unmap;
+
+		/* Save DMA mapping info for later unmapping */
+		sg->dma_addr[sg->num_segs] = dma_addr;
+		sg->size[sg->num_segs] = len;
+		sg->num_segs++;
+	}
+	return 0;
+unmap:
+	otx2_dma_unmap_skb_frags(pfvf, sg);
+	return -EINVAL;
+}
+
+static u64 otx2_tso_frag_dma_addr(struct otx2_snd_queue *sq,
+				  struct sk_buff *skb, int seg,
+				  u64 seg_addr, int hdr_len, int sqe)
+{
+	struct sg_list *sg = &sq->sg[sqe];
+	const skb_frag_t *frag;
+	int offset;
+
+	if (seg < 0)
+		return sg->dma_addr[0] + (seg_addr - (u64)skb->data);
+
+	frag = &skb_shinfo(skb)->frags[seg];
+	offset = seg_addr - (u64)skb_frag_address(frag);
+	if (skb_headlen(skb) - hdr_len)
+		seg++;
+	return sg->dma_addr[seg] + offset;
+}
+
+static void otx2_sqe_tso_add_sg(struct otx2_snd_queue *sq,
+				struct sg_list *list, int *offset)
+{
+	struct nix_sqe_sg_s *sg = NULL;
+	u16 *sg_lens = NULL;
+	u64 *iova = NULL;
+	int seg;
+
+	/* Add SG descriptors with buffer addresses */
+	for (seg = 0; seg < list->num_segs; seg++) {
+		if ((seg % MAX_SEGS_PER_SG) == 0) {
+			sg = (struct nix_sqe_sg_s *)(sq->sqe_base + *offset);
+			sg->ld_type = NIX_SEND_LDTYPE_LDD;
+			sg->subdc = NIX_SUBDC_SG;
+			sg->segs = 0;
+			sg_lens = (void *)sg;
+			iova = (void *)sg + sizeof(*sg);
+			/* Next subdc always starts at a 16byte boundary.
+			 * So if sg->segs is whether 2 or 3, offset += 16bytes.
+			 */
+			if ((list->num_segs - seg) >= (MAX_SEGS_PER_SG - 1))
+				*offset += sizeof(*sg) + (3 * sizeof(u64));
+			else
+				*offset += sizeof(*sg) + sizeof(u64);
+		}
+		sg_lens[frag_num(seg % MAX_SEGS_PER_SG)] = list->size[seg];
+		*iova++ = list->dma_addr[seg];
+		sg->segs++;
+	}
+}
+
+static void otx2_sq_append_tso(struct otx2_nic *pfvf, struct otx2_snd_queue *sq,
+			       struct sk_buff *skb, u16 qidx)
+{
+	struct netdev_queue *txq = netdev_get_tx_queue(pfvf->netdev, qidx);
+	int hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
+	int tcp_data, seg_len, pkt_len, offset;
+	struct nix_sqe_hdr_s *sqe_hdr;
+	int first_sqe = sq->head;
+	struct sg_list list;
+	struct tso_t tso;
+
+	/* Map SKB's fragments to DMA.
+	 * It's done here to avoid mapping for every TSO segment's packet.
+	 */
+	if (otx2_dma_map_tso_skb(pfvf, sq, skb, first_sqe, hdr_len)) {
+		dev_kfree_skb_any(skb);
+		return;
+	}
+
+	netdev_tx_sent_queue(txq, skb->len);
+
+	tso_start(skb, &tso);
+	tcp_data = skb->len - hdr_len;
+	while (tcp_data > 0) {
+		char *hdr;
+
+		seg_len = min_t(int, skb_shinfo(skb)->gso_size, tcp_data);
+		tcp_data -= seg_len;
+
+		/* Set SQE's SEND_HDR */
+		memset(sq->sqe_base, 0, sq->sqe_size);
+		sqe_hdr = (struct nix_sqe_hdr_s *)(sq->sqe_base);
+		otx2_sqe_add_hdr(pfvf, sq, sqe_hdr, skb, qidx);
+		offset = sizeof(*sqe_hdr);
+
+		/* Add TSO segment's pkt header */
+		hdr = sq->tso_hdrs->base + (sq->head * TSO_HEADER_SIZE);
+		tso_build_hdr(skb, hdr, &tso, seg_len, tcp_data == 0);
+		list.dma_addr[0] =
+			sq->tso_hdrs->iova + (sq->head * TSO_HEADER_SIZE);
+		list.size[0] = hdr_len;
+		list.num_segs = 1;
+
+		/* Add TSO segment's payload data fragments */
+		pkt_len = hdr_len;
+		while (seg_len > 0) {
+			int size;
+
+			size = min_t(int, tso.size, seg_len);
+
+			list.size[list.num_segs] = size;
+			list.dma_addr[list.num_segs] =
+				otx2_tso_frag_dma_addr(sq, skb,
+						       tso.next_frag_idx - 1,
+						       (u64)tso.data, hdr_len,
+						       first_sqe);
+			list.num_segs++;
+			pkt_len += size;
+			seg_len -= size;
+			tso_build_data(skb, &tso, size);
+		}
+		sqe_hdr->total = pkt_len;
+		otx2_sqe_tso_add_sg(sq, &list, &offset);
+
+		/* DMA mappings and skb needs to be freed only after last
+		 * TSO segment is transmitted out. So set 'PNC' only for
+		 * last segment. Also point last segment's sqe_id to first
+		 * segment's SQE index where skb address and DMA mappings
+		 * are saved.
+		 */
+		if (!tcp_data) {
+			sqe_hdr->pnc = 1;
+			sqe_hdr->sqe_id = first_sqe;
+			sq->sg[first_sqe].skb = (u64)skb;
+		} else {
+			sqe_hdr->pnc = 0;
+		}
+
+		sqe_hdr->sizem1 = (offset / 16) - 1;
+
+		/* Flush SQE to HW */
+		otx2_sqe_flush(sq, offset);
+	}
+}
+
+static bool is_hw_tso_supported(struct otx2_nic *pfvf,
+				struct sk_buff *skb)
+{
+	int payload_len, last_seg_size;
+
+	if (!pfvf->hw.hw_tso)
+		return false;
+
+	/* HW has an issue due to which when the payload of the last LSO
+	 * segment is shorter than 16 bytes, some header fields may not
+	 * be correctly modified, hence don't offload such TSO segments.
+	 */
+	if (!is_96xx_B0(pfvf->pdev))
+		return true;
+
+	payload_len = skb->len - (skb_transport_offset(skb) + tcp_hdrlen(skb));
+	last_seg_size = payload_len % skb_shinfo(skb)->gso_size;
+	if (last_seg_size && last_seg_size < 16)
+		return false;
+
+	return true;
+}
+
+static int otx2_get_sqe_count(struct otx2_nic *pfvf, struct sk_buff *skb)
+{
+	if (!skb_shinfo(skb)->gso_size)
+		return 1;
+
+	/* HW TSO */
+	if (is_hw_tso_supported(pfvf, skb))
+		return 1;
+
+	/* SW TSO */
+	return skb_shinfo(skb)->gso_segs;
+}
+
+bool otx2_sq_append_skb(struct net_device *netdev, struct otx2_snd_queue *sq,
+			struct sk_buff *skb, u16 qidx)
+{
+	struct netdev_queue *txq = netdev_get_tx_queue(netdev, qidx);
+	struct otx2_nic *pfvf = netdev_priv(netdev);
+	int offset, num_segs, free_sqe;
+	struct nix_sqe_hdr_s *sqe_hdr;
+
+	/* Check if there is room for new SQE.
+	 * 'Num of SQBs freed to SQ's pool - SQ's Aura count'
+	 * will give free SQE count.
+	 */
+	free_sqe = (sq->num_sqbs - *sq->aura_fc_addr) * sq->sqe_per_sqb;
+
+	if (free_sqe < sq->sqe_thresh ||
+	    free_sqe < otx2_get_sqe_count(pfvf, skb))
+		return false;
+
+	num_segs = skb_shinfo(skb)->nr_frags + 1;
+
+	/* If SKB doesn't fit in a single SQE, linearize it.
+	 * TODO: Consider adding JUMP descriptor instead.
+	 */
+	if (unlikely(num_segs > OTX2_MAX_FRAGS_IN_SQE)) {
+		if (__skb_linearize(skb)) {
+			dev_kfree_skb_any(skb);
+			return true;
+		}
+		num_segs = skb_shinfo(skb)->nr_frags + 1;
+	}
+
+	if (skb_shinfo(skb)->gso_size && !is_hw_tso_supported(pfvf, skb)) {
+		otx2_sq_append_tso(pfvf, sq, skb, qidx);
+		return true;
+	}
+
+	/* Set SQE's SEND_HDR.
+	 * Do not clear the first 64bit as it contains constant info.
+	 */
+	memset(sq->sqe_base + 8, 0, sq->sqe_size - 8);
+	sqe_hdr = (struct nix_sqe_hdr_s *)(sq->sqe_base);
+	otx2_sqe_add_hdr(pfvf, sq, sqe_hdr, skb, qidx);
+	offset = sizeof(*sqe_hdr);
+
+	/* Add extended header if needed */
+	otx2_sqe_add_ext(pfvf, sq, skb, &offset);
+
+	/* Add SG subdesc with data frags */
+	if (!otx2_sqe_add_sg(pfvf, sq, skb, num_segs, &offset)) {
+		otx2_dma_unmap_skb_frags(pfvf, &sq->sg[sq->head]);
+		return false;
+	}
+
+	sqe_hdr->sizem1 = (offset / 16) - 1;
+
+	netdev_tx_sent_queue(txq, skb->len);
+
+	/* Flush SQE to HW */
+	otx2_sqe_flush(sq, offset);
+
+	return true;
+}
+
+void otx2_cleanup_rx_cqes(struct otx2_nic *pfvf, struct otx2_cq_queue *cq)
+{
+	struct nix_cqe_rx_s *cqe;
+	int processed_cqe = 0;
+	u64 iova, pa;
+
+	while ((cqe = (struct nix_cqe_rx_s *)otx2_get_next_cqe(cq))) {
+		if (!cqe->sg.subdc)
+			continue;
+		iova = cqe->sg.seg_addr - OTX2_HEAD_ROOM;
+		pa = otx2_iova_to_phys(pfvf->iommu_domain, iova);
+		otx2_dma_unmap_page(pfvf, iova, pfvf->rbsize, DMA_FROM_DEVICE);
+		put_page(virt_to_page(phys_to_virt(pa)));
+		processed_cqe++;
+	}
+
+	/* Free CQEs to HW */
+	otx2_write64(pfvf, NIX_LF_CQ_OP_DOOR,
+		     ((u64)cq->cq_idx << 32) | processed_cqe);
+}
+
+void otx2_cleanup_tx_cqes(struct otx2_nic *pfvf, struct otx2_cq_queue *cq)
+{
+	struct sk_buff *skb = NULL;
+	struct otx2_snd_queue *sq;
+	struct nix_cqe_tx_s *cqe;
+	int processed_cqe = 0;
+	struct sg_list *sg;
+
+	sq = &pfvf->qset.sq[cq->cint_idx];
+
+	while ((cqe = (struct nix_cqe_tx_s *)otx2_get_next_cqe(cq))) {
+		sg = &sq->sg[cqe->comp.sqe_id];
+		skb = (struct sk_buff *)sg->skb;
+		if (skb) {
+			otx2_dma_unmap_skb_frags(pfvf, sg);
+			dev_kfree_skb_any(skb);
+			sg->skb = (u64)NULL;
+		}
+		processed_cqe++;
+	}
+
+	/* Free CQEs to HW */
+	otx2_write64(pfvf, NIX_LF_CQ_OP_DOOR,
+		     ((u64)cq->cq_idx << 32) | processed_cqe);
+}
+
+int otx2_rxtx_enable(struct otx2_nic *pfvf, bool enable)
+{
+	struct msg_req *msg;
+	int err;
+
+	otx2_mbox_lock(&pfvf->mbox);
+	if (enable)
+		msg = otx2_mbox_alloc_msg_nix_lf_start_rx(&pfvf->mbox);
+	else
+		msg = otx2_mbox_alloc_msg_nix_lf_stop_rx(&pfvf->mbox);
+
+	if (!msg) {
+		otx2_mbox_unlock(&pfvf->mbox);
+		return -ENOMEM;
+	}
+
+	err = otx2_sync_mbox_msg(&pfvf->mbox);
+	otx2_mbox_unlock(&pfvf->mbox);
+	return err;
+}

drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Marvell OcteonTx2 RVU Ethernet driver
+ *
+ * Copyright (C) 2020 Marvell International Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef OTX2_TXRX_H
+#define OTX2_TXRX_H
+
+#include <linux/etherdevice.h>
+#include <linux/iommu.h>
+#include <linux/if_vlan.h>
+
+#define LBK_CHAN_BASE	0x000
+#define SDP_CHAN_BASE	0x700
+#define CGX_CHAN_BASE	0x800
+
+#define OTX2_DATA_ALIGN(X)	ALIGN(X, OTX2_ALIGN)
+#define OTX2_HEAD_ROOM		OTX2_ALIGN
+
+#define	OTX2_ETH_HLEN		(VLAN_ETH_HLEN + VLAN_HLEN)
+#define	OTX2_MIN_MTU		64
+#define	OTX2_MAX_MTU		(9212 - OTX2_ETH_HLEN)
+
+#define OTX2_MAX_GSO_SEGS	255
+#define OTX2_MAX_FRAGS_IN_SQE	9
+
+/* Rx buffer size should be in multiples of 128bytes */
+#define RCV_FRAG_LEN1(x)				\
+		((OTX2_HEAD_ROOM + OTX2_DATA_ALIGN(x)) + \
+		OTX2_DATA_ALIGN(sizeof(struct skb_shared_info)))
+
+/* Prefer 2048 byte buffers for better last level cache
+ * utilization or data distribution across regions.
+ */
+#define RCV_FRAG_LEN(x)	\
+		((RCV_FRAG_LEN1(x) < 2048) ? 2048 : RCV_FRAG_LEN1(x))
+
+#define DMA_BUFFER_LEN(x)		\
+		((x) - OTX2_HEAD_ROOM - \
+		OTX2_DATA_ALIGN(sizeof(struct skb_shared_info)))
+
+/* IRQ triggered when NIX_LF_CINTX_CNT[ECOUNT]
+ * is equal to this value.
+ */
+#define CQ_CQE_THRESH_DEFAULT	10
+
+/* IRQ triggered when NIX_LF_CINTX_CNT[ECOUNT]
+ * is nonzero and this much time elapses after that.
+ */
+#define CQ_TIMER_THRESH_DEFAULT	1  /* 1 usec */
+#define CQ_TIMER_THRESH_MAX     25 /* 25 usec */
+
+/* Min number of CQs (of the ones mapped to this CINT)
+ * with valid CQEs.
+ */
+#define CQ_QCOUNT_DEFAULT	1
+
+struct queue_stats {
+	u64	bytes;
+	u64	pkts;
+};
+
+struct otx2_rcv_queue {
+	struct queue_stats	stats;
+};
+
+struct sg_list {
+	u16	num_segs;
+	u64	skb;
+	u64	size[OTX2_MAX_FRAGS_IN_SQE];
+	u64	dma_addr[OTX2_MAX_FRAGS_IN_SQE];
+};
+
+struct otx2_snd_queue {
+	u8			aura_id;
+	u16			head;
+	u16			sqe_size;
+	u32			sqe_cnt;
+	u16			num_sqbs;
+	u16			sqe_thresh;
+	u8			sqe_per_sqb;
+	u64			 io_addr;
+	u64			*aura_fc_addr;
+	u64			*lmt_addr;
+	void			*sqe_base;
+	struct qmem		*sqe;
+	struct qmem		*tso_hdrs;
+	struct sg_list		*sg;
+	struct queue_stats	stats;
+	u16			sqb_count;
+	u64			*sqb_ptrs;
+} ____cacheline_aligned_in_smp;
+
+enum cq_type {
+	CQ_RX,
+	CQ_TX,
+	CQS_PER_CINT = 2, /* RQ + SQ */
+};
+
+struct otx2_cq_poll {
+	void			*dev;
+#define CINT_INVALID_CQ		255
+	u8			cint_idx;
+	u8			cq_ids[CQS_PER_CINT];
+	struct napi_struct	napi;
+};
+
+struct otx2_pool {
+	struct qmem		*stack;
+	struct qmem		*fc_addr;
+	u8			rbpage_order;
+	u16			rbsize;
+	u32			page_offset;
+	u16			pageref;
+	struct page		*page;
+};
+
+struct otx2_cq_queue {
+	u8			cq_idx;
+	u8			cq_type;
+	u8			cint_idx; /* CQ interrupt id */
+	u8			refill_task_sched;
+	u16			cqe_size;
+	u16			pool_ptrs;
+	u32			cqe_cnt;
+	u32			cq_head;
+	void			*cqe_base;
+	struct qmem		*cqe;
+	struct otx2_pool	*rbpool;
+} ____cacheline_aligned_in_smp;
+
+struct otx2_qset {
+	u32			rqe_cnt;
+	u32			sqe_cnt; /* Keep these two at top */
+#define OTX2_MAX_CQ_CNT		64
+	u16			cq_cnt;
+	u16			xqe_size;
+	struct otx2_pool	*pool;
+	struct otx2_cq_poll	*napi;
+	struct otx2_cq_queue	*cq;
+	struct otx2_snd_queue	*sq;
+	struct otx2_rcv_queue	*rq;
+};
+
+/* Translate IOVA to physical address */
+static inline u64 otx2_iova_to_phys(void *iommu_domain, dma_addr_t dma_addr)
+{
+	/* Translation is installed only when IOMMU is present */
+	if (likely(iommu_domain))
+		return iommu_iova_to_phys(iommu_domain, dma_addr);
+	return dma_addr;
+}
+
+int otx2_napi_handler(struct napi_struct *napi, int budget);
+bool otx2_sq_append_skb(struct net_device *netdev, struct otx2_snd_queue *sq,
+			struct sk_buff *skb, u16 qidx);
+#endif /* OTX2_TXRX_H */