Merge branch 'net-mana-add-pf-and-xdp_redirect-support'

Haiyang Zhang says:

====================
net: mana: Add PF and XDP_REDIRECT support

The patch set adds PF and XDP_REDIRECT support.
====================

Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Paolo Abeni <[e-mail]>

drivers/net/ethernet/microsoft/mana/gdma.h

···
 	struct completion	eq_test_event;
 	u32			test_event_eq_id;
 
+	bool			is_pf;
 	void __iomem		*bar0_va;
 	void __iomem		*shm_base;
 	void __iomem		*db_page_base;
···
 #define GDMA_REG_DB_PAGE_OFFSET	8
 #define GDMA_REG_DB_PAGE_SIZE	0x10
 #define GDMA_REG_SHM_OFFSET	0x18
+
+#define GDMA_PF_REG_DB_PAGE_SIZE	0xD0
+#define GDMA_PF_REG_DB_PAGE_OFF		0xC8
+#define GDMA_PF_REG_SHM_OFF		0x70
+
+#define GDMA_SRIOV_REG_CFG_BASE_OFF	0x108
+
+#define MANA_PF_DEVICE_ID 0x00B9
+#define MANA_VF_DEVICE_ID 0x00BA
 
 struct gdma_posted_wqe_info {
 	u32 wqe_size_in_bu;

drivers/net/ethernet/microsoft/mana/gdma_main.c

···
 	return readq(g->bar0_va + offset);
 }
 
-static void mana_gd_init_registers(struct pci_dev *pdev)
+static void mana_gd_init_pf_regs(struct pci_dev *pdev)
+{
+	struct gdma_context *gc = pci_get_drvdata(pdev);
+	void __iomem *sriov_base_va;
+	u64 sriov_base_off;
+
+	gc->db_page_size = mana_gd_r32(gc, GDMA_PF_REG_DB_PAGE_SIZE) & 0xFFFF;
+	gc->db_page_base = gc->bar0_va +
+				mana_gd_r64(gc, GDMA_PF_REG_DB_PAGE_OFF);
+
+	sriov_base_off = mana_gd_r64(gc, GDMA_SRIOV_REG_CFG_BASE_OFF);
+
+	sriov_base_va = gc->bar0_va + sriov_base_off;
+	gc->shm_base = sriov_base_va +
+			mana_gd_r64(gc, sriov_base_off + GDMA_PF_REG_SHM_OFF);
+}
+
+static void mana_gd_init_vf_regs(struct pci_dev *pdev)
 {
 	struct gdma_context *gc = pci_get_drvdata(pdev);
 
···
 				mana_gd_r64(gc, GDMA_REG_DB_PAGE_OFFSET);
 
 	gc->shm_base = gc->bar0_va + mana_gd_r64(gc, GDMA_REG_SHM_OFFSET);
+}
+
+static void mana_gd_init_registers(struct pci_dev *pdev)
+{
+	struct gdma_context *gc = pci_get_drvdata(pdev);
+
+	if (gc->is_pf)
+		mana_gd_init_pf_regs(pdev);
+	else
+		mana_gd_init_vf_regs(pdev);
 }
 
 static int mana_gd_query_max_resources(struct pci_dev *pdev)
···
 	mana_gd_remove_irqs(pdev);
 }
 
+static bool mana_is_pf(unsigned short dev_id)
+{
+	return dev_id == MANA_PF_DEVICE_ID;
+}
+
 static int mana_gd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 {
 	struct gdma_context *gc;
···
 	if (!bar0_va)
 		goto free_gc;
 
+	gc->is_pf = mana_is_pf(pdev->device);
 	gc->bar0_va = bar0_va;
 	gc->dev = &pdev->dev;
-
 
 	err = mana_gd_setup(pdev);
 	if (err)
···
 #endif
 
 static const struct pci_device_id mana_id_table[] = {
-	{ PCI_DEVICE(PCI_VENDOR_ID_MICROSOFT, 0x00BA) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_MICROSOFT, MANA_PF_DEVICE_ID) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_MICROSOFT, MANA_VF_DEVICE_ID) },
 	{ }
 };
 

drivers/net/ethernet/microsoft/mana/hw_channel.c

···
 			hwc->rxq->msg_buf->gpa_mkey = val;
 			hwc->txq->msg_buf->gpa_mkey = val;
 			break;
+
+		case HWC_INIT_DATA_PF_DEST_RQ_ID:
+			hwc->pf_dest_vrq_id = val;
+			break;
+
+		case HWC_INIT_DATA_PF_DEST_CQ_ID:
+			hwc->pf_dest_vrcq_id = val;
+			break;
 		}
 
 		break;
···
 int mana_hwc_send_request(struct hw_channel_context *hwc, u32 req_len,
 			  const void *req, u32 resp_len, void *resp)
 {
+	struct gdma_context *gc = hwc->gdma_dev->gdma_context;
 	struct hwc_work_request *tx_wr;
 	struct hwc_wq *txq = hwc->txq;
 	struct gdma_req_hdr *req_msg;
 	struct hwc_caller_ctx *ctx;
+	u32 dest_vrcq = 0;
+	u32 dest_vrq = 0;
 	u16 msg_id;
 	int err;
 
···
 
 	tx_wr->msg_size = req_len;
 
-	err = mana_hwc_post_tx_wqe(txq, tx_wr, 0, 0, false);
+	if (gc->is_pf) {
+		dest_vrq = hwc->pf_dest_vrq_id;
+		dest_vrcq = hwc->pf_dest_vrcq_id;
+	}
+
+	err = mana_hwc_post_tx_wqe(txq, tx_wr, dest_vrq, dest_vrcq, false);
 	if (err) {
 		dev_err(hwc->dev, "HWC: Failed to post send WQE: %d\n", err);
 		goto out;

drivers/net/ethernet/microsoft/mana/hw_channel.h

···
 #define HWC_INIT_DATA_MAX_NUM_CQS	7
 #define HWC_INIT_DATA_PDID		8
 #define HWC_INIT_DATA_GPA_MKEY		9
+#define HWC_INIT_DATA_PF_DEST_RQ_ID	10
+#define HWC_INIT_DATA_PF_DEST_CQ_ID	11
 
 /* Structures labeled with "HW DATA" are exchanged with the hardware. All of
  * them are naturally aligned and hence don't need __packed.
···
 
 	struct semaphore sema;
 	struct gdma_resource inflight_msg_res;
+
+	u32 pf_dest_vrq_id;
+	u32 pf_dest_vrcq_id;
 
 	struct hwc_caller_ctx *caller_ctx;
 };

drivers/net/ethernet/microsoft/mana/mana.h

···
 	u64 bytes;
 	u64 xdp_drop;
 	u64 xdp_tx;
+	u64 xdp_redirect;
 	struct u64_stats_sync syncp;
 };
 
 struct mana_stats_tx {
 	u64 packets;
 	u64 bytes;
+	u64 xdp_xmit;
 	struct u64_stats_sync syncp;
 };
 
···
 	struct bpf_prog __rcu *bpf_prog;
 	struct xdp_rxq_info xdp_rxq;
 	struct page *xdp_save_page;
+	bool xdp_flush;
+	int xdp_rc; /* XDP redirect return code */
 
 	/* MUST BE THE LAST MEMBER:
 	 * Each receive buffer has an associated mana_recv_buf_oob.
···
 	unsigned int num_queues;
 
 	mana_handle_t port_handle;
+	mana_handle_t pf_filter_handle;
 
 	u16 port_idx;
 
···
 void mana_remove(struct gdma_dev *gd, bool suspending);
 
 void mana_xdp_tx(struct sk_buff *skb, struct net_device *ndev);
+int mana_xdp_xmit(struct net_device *ndev, int n, struct xdp_frame **frames,
+		  u32 flags);
 u32 mana_run_xdp(struct net_device *ndev, struct mana_rxq *rxq,
 		 struct xdp_buff *xdp, void *buf_va, uint pkt_len);
 struct bpf_prog *mana_xdp_get(struct mana_port_context *apc);
···
 	MANA_FENCE_RQ		= 0x20006,
 	MANA_CONFIG_VPORT_RX	= 0x20007,
 	MANA_QUERY_VPORT_CONFIG	= 0x20008,
+
+	/* Privileged commands for the PF mode */
+	MANA_REGISTER_FILTER	= 0x28000,
+	MANA_DEREGISTER_FILTER	= 0x28001,
+	MANA_REGISTER_HW_PORT	= 0x28003,
+	MANA_DEREGISTER_HW_PORT	= 0x28004,
 };
 
 /* Query Device Configuration */
···
 }; /* HW DATA */
 
 struct mana_cfg_rx_steer_resp {
+	struct gdma_resp_hdr hdr;
+}; /* HW DATA */
+
+/* Register HW vPort */
+struct mana_register_hw_vport_req {
+	struct gdma_req_hdr hdr;
+	u16 attached_gfid;
+	u8 is_pf_default_vport;
+	u8 reserved1;
+	u8 allow_all_ether_types;
+	u8 reserved2;
+	u8 reserved3;
+	u8 reserved4;
+}; /* HW DATA */
+
+struct mana_register_hw_vport_resp {
+	struct gdma_resp_hdr hdr;
+	mana_handle_t hw_vport_handle;
+}; /* HW DATA */
+
+/* Deregister HW vPort */
+struct mana_deregister_hw_vport_req {
+	struct gdma_req_hdr hdr;
+	mana_handle_t hw_vport_handle;
+}; /* HW DATA */
+
+struct mana_deregister_hw_vport_resp {
+	struct gdma_resp_hdr hdr;
+}; /* HW DATA */
+
+/* Register filter */
+struct mana_register_filter_req {
+	struct gdma_req_hdr hdr;
+	mana_handle_t vport;
+	u8 mac_addr[6];
+	u8 reserved1;
+	u8 reserved2;
+	u8 reserved3;
+	u8 reserved4;
+	u16 reserved5;
+	u32 reserved6;
+	u32 reserved7;
+	u32 reserved8;
+}; /* HW DATA */
+
+struct mana_register_filter_resp {
+	struct gdma_resp_hdr hdr;
+	mana_handle_t filter_handle;
+}; /* HW DATA */
+
+/* Deregister filter */
+struct mana_deregister_filter_req {
+	struct gdma_req_hdr hdr;
+	mana_handle_t filter_handle;
+}; /* HW DATA */
+
+struct mana_deregister_filter_resp {
 	struct gdma_resp_hdr hdr;
 }; /* HW DATA */
 

drivers/net/ethernet/microsoft/mana/mana_bpf.c

···
 	ndev->stats.tx_dropped++;
 }
 
+static int mana_xdp_xmit_fm(struct net_device *ndev, struct xdp_frame *frame,
+			    u16 q_idx)
+{
+	struct sk_buff *skb;
+
+	skb = xdp_build_skb_from_frame(frame, ndev);
+	if (unlikely(!skb))
+		return -ENOMEM;
+
+	skb_set_queue_mapping(skb, q_idx);
+
+	mana_xdp_tx(skb, ndev);
+
+	return 0;
+}
+
+int mana_xdp_xmit(struct net_device *ndev, int n, struct xdp_frame **frames,
+		  u32 flags)
+{
+	struct mana_port_context *apc = netdev_priv(ndev);
+	struct mana_stats_tx *tx_stats;
+	int i, count = 0;
+	u16 q_idx;
+
+	if (unlikely(!apc->port_is_up))
+		return 0;
+
+	q_idx = smp_processor_id() % ndev->real_num_tx_queues;
+
+	for (i = 0; i < n; i++) {
+		if (mana_xdp_xmit_fm(ndev, frames[i], q_idx))
+			break;
+
+		count++;
+	}
+
+	tx_stats = &apc->tx_qp[q_idx].txq.stats;
+
+	u64_stats_update_begin(&tx_stats->syncp);
+	tx_stats->xdp_xmit += count;
+	u64_stats_update_end(&tx_stats->syncp);
+
+	return count;
+}
+
 u32 mana_run_xdp(struct net_device *ndev, struct mana_rxq *rxq,
 		 struct xdp_buff *xdp, void *buf_va, uint pkt_len)
 {
+	struct mana_stats_rx *rx_stats;
 	struct bpf_prog *prog;
 	u32 act = XDP_PASS;
 
···
 
 	act = bpf_prog_run_xdp(prog, xdp);
 
+	rx_stats = &rxq->stats;
+
 	switch (act) {
 	case XDP_PASS:
 	case XDP_TX:
 	case XDP_DROP:
 		break;
+
+	case XDP_REDIRECT:
+		rxq->xdp_rc = xdp_do_redirect(ndev, xdp, prog);
+		if (!rxq->xdp_rc) {
+			rxq->xdp_flush = true;
+
+			u64_stats_update_begin(&rx_stats->syncp);
+			rx_stats->packets++;
+			rx_stats->bytes += pkt_len;
+			rx_stats->xdp_redirect++;
+			u64_stats_update_end(&rx_stats->syncp);
+
+			break;
+		}
+
+		fallthrough;
 
 	case XDP_ABORTED:
 		trace_xdp_exception(ndev, prog, act);

drivers/net/ethernet/microsoft/mana/mana_en.c

···
 #include <linux/inetdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/ethtool.h>
+#include <linux/filter.h>
 #include <linux/mm.h>
 
 #include <net/checksum.h>
···
 	.ndo_validate_addr	= eth_validate_addr,
 	.ndo_get_stats64	= mana_get_stats64,
 	.ndo_bpf		= mana_bpf,
+	.ndo_xdp_xmit		= mana_xdp_xmit,
 };
 
 static void mana_cleanup_port_context(struct mana_port_context *apc)
···
 		return -EPROTO;
 
 	return 0;
+}
+
+static int mana_pf_register_hw_vport(struct mana_port_context *apc)
+{
+	struct mana_register_hw_vport_resp resp = {};
+	struct mana_register_hw_vport_req req = {};
+	int err;
+
+	mana_gd_init_req_hdr(&req.hdr, MANA_REGISTER_HW_PORT,
+			     sizeof(req), sizeof(resp));
+	req.attached_gfid = 1;
+	req.is_pf_default_vport = 1;
+	req.allow_all_ether_types = 1;
+
+	err = mana_send_request(apc->ac, &req, sizeof(req), &resp,
+				sizeof(resp));
+	if (err) {
+		netdev_err(apc->ndev, "Failed to register hw vPort: %d\n", err);
+		return err;
+	}
+
+	err = mana_verify_resp_hdr(&resp.hdr, MANA_REGISTER_HW_PORT,
+				   sizeof(resp));
+	if (err || resp.hdr.status) {
+		netdev_err(apc->ndev, "Failed to register hw vPort: %d, 0x%x\n",
+			   err, resp.hdr.status);
+		return err ? err : -EPROTO;
+	}
+
+	apc->port_handle = resp.hw_vport_handle;
+	return 0;
+}
+
+static void mana_pf_deregister_hw_vport(struct mana_port_context *apc)
+{
+	struct mana_deregister_hw_vport_resp resp = {};
+	struct mana_deregister_hw_vport_req req = {};
+	int err;
+
+	mana_gd_init_req_hdr(&req.hdr, MANA_DEREGISTER_HW_PORT,
+			     sizeof(req), sizeof(resp));
+	req.hw_vport_handle = apc->port_handle;
+
+	err = mana_send_request(apc->ac, &req, sizeof(req), &resp,
+				sizeof(resp));
+	if (err) {
+		netdev_err(apc->ndev, "Failed to unregister hw vPort: %d\n",
+			   err);
+		return;
+	}
+
+	err = mana_verify_resp_hdr(&resp.hdr, MANA_DEREGISTER_HW_PORT,
+				   sizeof(resp));
+	if (err || resp.hdr.status)
+		netdev_err(apc->ndev,
+			   "Failed to deregister hw vPort: %d, 0x%x\n",
+			   err, resp.hdr.status);
+}
+
+static int mana_pf_register_filter(struct mana_port_context *apc)
+{
+	struct mana_register_filter_resp resp = {};
+	struct mana_register_filter_req req = {};
+	int err;
+
+	mana_gd_init_req_hdr(&req.hdr, MANA_REGISTER_FILTER,
+			     sizeof(req), sizeof(resp));
+	req.vport = apc->port_handle;
+	memcpy(req.mac_addr, apc->mac_addr, ETH_ALEN);
+
+	err = mana_send_request(apc->ac, &req, sizeof(req), &resp,
+				sizeof(resp));
+	if (err) {
+		netdev_err(apc->ndev, "Failed to register filter: %d\n", err);
+		return err;
+	}
+
+	err = mana_verify_resp_hdr(&resp.hdr, MANA_REGISTER_FILTER,
+				   sizeof(resp));
+	if (err || resp.hdr.status) {
+		netdev_err(apc->ndev, "Failed to register filter: %d, 0x%x\n",
+			   err, resp.hdr.status);
+		return err ? err : -EPROTO;
+	}
+
+	apc->pf_filter_handle = resp.filter_handle;
+	return 0;
+}
+
+static void mana_pf_deregister_filter(struct mana_port_context *apc)
+{
+	struct mana_deregister_filter_resp resp = {};
+	struct mana_deregister_filter_req req = {};
+	int err;
+
+	mana_gd_init_req_hdr(&req.hdr, MANA_DEREGISTER_FILTER,
+			     sizeof(req), sizeof(resp));
+	req.filter_handle = apc->pf_filter_handle;
+
+	err = mana_send_request(apc->ac, &req, sizeof(req), &resp,
+				sizeof(resp));
+	if (err) {
+		netdev_err(apc->ndev, "Failed to unregister filter: %d\n",
+			   err);
+		return;
+	}
+
+	err = mana_verify_resp_hdr(&resp.hdr, MANA_DEREGISTER_FILTER,
+				   sizeof(resp));
+	if (err || resp.hdr.status)
+		netdev_err(apc->ndev,
+			   "Failed to deregister filter: %d, 0x%x\n",
+			   err, resp.hdr.status);
 }
 
 static int mana_query_device_cfg(struct mana_context *ac, u32 proto_major_ver,
···
 
 	act = mana_run_xdp(ndev, rxq, &xdp, buf_va, pkt_len);
 
+	if (act == XDP_REDIRECT && !rxq->xdp_rc)
+		return;
+
 	if (act != XDP_PASS && act != XDP_TX)
 		goto drop_xdp;
 
···
 static void mana_poll_rx_cq(struct mana_cq *cq)
 {
 	struct gdma_comp *comp = cq->gdma_comp_buf;
+	struct mana_rxq *rxq = cq->rxq;
 	int comp_read, i;
 
 	comp_read = mana_gd_poll_cq(cq->gdma_cq, comp, CQE_POLLING_BUFFER);
 	WARN_ON_ONCE(comp_read > CQE_POLLING_BUFFER);
+
+	rxq->xdp_flush = false;
 
 	for (i = 0; i < comp_read; i++) {
 		if (WARN_ON_ONCE(comp[i].is_sq))
···
 		if (WARN_ON_ONCE(comp[i].wq_num != cq->rxq->gdma_id))
 			return;
 
-		mana_process_rx_cqe(cq->rxq, cq, &comp[i]);
+		mana_process_rx_cqe(rxq, cq, &comp[i]);
 	}
+
+	if (rxq->xdp_flush)
+		xdp_do_flush();
 }
 
 static void mana_cq_handler(void *context, struct gdma_queue *gdma_queue)
···
 
 static void mana_destroy_vport(struct mana_port_context *apc)
 {
+	struct gdma_dev *gd = apc->ac->gdma_dev;
 	struct mana_rxq *rxq;
 	u32 rxq_idx;
 
···
 	}
 
 	mana_destroy_txq(apc);
+
+	if (gd->gdma_context->is_pf)
+		mana_pf_deregister_hw_vport(apc);
 }
 
 static int mana_create_vport(struct mana_port_context *apc,
···
 	int err;
 
 	apc->default_rxobj = INVALID_MANA_HANDLE;
+
+	if (gd->gdma_context->is_pf) {
+		err = mana_pf_register_hw_vport(apc);
+		if (err)
+			return err;
+	}
 
 	err = mana_cfg_vport(apc, gd->pdid, gd->doorbell);
 	if (err)
···
 int mana_alloc_queues(struct net_device *ndev)
 {
 	struct mana_port_context *apc = netdev_priv(ndev);
+	struct gdma_dev *gd = apc->ac->gdma_dev;
 	int err;
 
 	err = mana_create_vport(apc, ndev);
···
 	err = mana_config_rss(apc, TRI_STATE_TRUE, true, true);
 	if (err)
 		goto destroy_vport;
+
+	if (gd->gdma_context->is_pf) {
+		err = mana_pf_register_filter(apc);
+		if (err)
+			goto destroy_vport;
+	}
 
 	mana_chn_setxdp(apc, mana_xdp_get(apc));
 
···
 static int mana_dealloc_queues(struct net_device *ndev)
 {
 	struct mana_port_context *apc = netdev_priv(ndev);
+	struct gdma_dev *gd = apc->ac->gdma_dev;
 	struct mana_txq *txq;
 	int i, err;
 
···
 		return -EINVAL;
 
 	mana_chn_setxdp(apc, NULL);
+
+	if (gd->gdma_context->is_pf)
+		mana_pf_deregister_filter(apc);
 
 	/* No packet can be transmitted now since apc->port_is_up is false.
 	 * There is still a tiny chance that mana_poll_tx_cq() can re-enable
···
 	apc->max_queues = gc->max_num_queues;
 	apc->num_queues = gc->max_num_queues;
 	apc->port_handle = INVALID_MANA_HANDLE;
+	apc->pf_filter_handle = INVALID_MANA_HANDLE;
 	apc->port_idx = port_idx;
 
 	ndev->netdev_ops = &mana_devops;

drivers/net/ethernet/microsoft/mana/mana_ethtool.c

···
 	if (stringset != ETH_SS_STATS)
 		return -EINVAL;
 
-	return ARRAY_SIZE(mana_eth_stats) + num_queues * 6;
+	return ARRAY_SIZE(mana_eth_stats) + num_queues * 8;
 }
 
 static void mana_get_strings(struct net_device *ndev, u32 stringset, u8 *data)
···
 		p += ETH_GSTRING_LEN;
 		sprintf(p, "rx_%d_xdp_tx", i);
 		p += ETH_GSTRING_LEN;
+		sprintf(p, "rx_%d_xdp_redirect", i);
+		p += ETH_GSTRING_LEN;
 	}
 
 	for (i = 0; i < num_queues; i++) {
 		sprintf(p, "tx_%d_packets", i);
 		p += ETH_GSTRING_LEN;
 		sprintf(p, "tx_%d_bytes", i);
+		p += ETH_GSTRING_LEN;
+		sprintf(p, "tx_%d_xdp_xmit", i);
 		p += ETH_GSTRING_LEN;
 	}
 }
···
 	struct mana_stats_tx *tx_stats;
 	unsigned int start;
 	u64 packets, bytes;
+	u64 xdp_redirect;
+	u64 xdp_xmit;
 	u64 xdp_drop;
 	u64 xdp_tx;
 	int q, i = 0;
···
 			bytes = rx_stats->bytes;
 			xdp_drop = rx_stats->xdp_drop;
 			xdp_tx = rx_stats->xdp_tx;
+			xdp_redirect = rx_stats->xdp_redirect;
 		} while (u64_stats_fetch_retry_irq(&rx_stats->syncp, start));
 
 		data[i++] = packets;
 		data[i++] = bytes;
 		data[i++] = xdp_drop;
 		data[i++] = xdp_tx;
+		data[i++] = xdp_redirect;
 	}
 
 	for (q = 0; q < num_queues; q++) {
···
 			start = u64_stats_fetch_begin_irq(&tx_stats->syncp);
 			packets = tx_stats->packets;
 			bytes = tx_stats->bytes;
+			xdp_xmit = tx_stats->xdp_xmit;
 		} while (u64_stats_fetch_retry_irq(&tx_stats->syncp, start));
 
 		data[i++] = packets;
 		data[i++] = bytes;
+		data[i++] = xdp_xmit;
 	}
 }