qed: Introduce VFs

This adds the qed VFs for the first time -
The vfs are limited functions, with a very different PCI bar structure
[when compared with PFs] to better impose the related security demands
associated with them.

This patch includes the logic neccesary to allow VFs to successfully probe
[without actually adding the ability to enable iov].
This includes diverging all the flows that would occur as part of the pci
probe of the driver, preventing VF from accessing registers/memories it
can't and instead utilize the VF->PF channel to query the PF for needed
information.

Signed-off-by: Yuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

22 files changed, 1768 insertions, 182 deletions

diff --git a/drivers/net/ethernet/qlogic/qed/Makefile b/drivers/net/ethernet/qlogic/qed/Makefile
index e11a809034ea..a44874562cfd 100644
--- a/drivers/net/ethernet/qlogic/qed/Makefile
+++ b/drivers/net/ethernet/qlogic/qed/Makefile
@@ -3,4 +3,4 @@ obj-$(CONFIG_QED) := qed.o
 qed-y := qed_cxt.o qed_dev.o qed_hw.o qed_init_fw_funcs.o qed_init_ops.o \
 	 qed_int.o qed_main.o qed_mcp.o qed_sp_commands.o qed_spq.o qed_l2.o \
 	 qed_selftest.o
-qed-$(CONFIG_QED_SRIOV) += qed_sriov.o
+qed-$(CONFIG_QED_SRIOV) += qed_sriov.o qed_vf.o
diff --git a/drivers/net/ethernet/qlogic/qed/qed.h b/drivers/net/ethernet/qlogic/qed/qed.h
index 01f9b6c880bd..f9a3576305a1 100644
--- a/drivers/net/ethernet/qlogic/qed/qed.h
+++ b/drivers/net/ethernet/qlogic/qed/qed.h
@@ -311,6 +311,8 @@ struct qed_hwfn {
 	bool				first_on_engine;
 	bool				hw_init_done;
 
+	u8				num_funcs_on_engine;
+
 	/* BAR access */
 	void __iomem			*regview;
 	void __iomem			*doorbells;
@@ -361,6 +363,7 @@ struct qed_hwfn {
 	/* True if the driver requests for the link */
 	bool				b_drv_link_init;
 
+	struct qed_vf_iov		*vf_iov_info;
 	struct qed_pf_iov		*pf_iov_info;
 	struct qed_mcp_info		*mcp_info;
 
@@ -497,6 +500,8 @@ struct qed_dev {
 #define IS_QED_SRIOV(cdev)              (!!(cdev)->p_iov_info)
 
 	unsigned long			tunn_mode;
+
+	bool				b_is_vf;
 	u32				drv_type;
 
 	struct qed_eth_stats		*reset_stats;
diff --git a/drivers/net/ethernet/qlogic/qed/qed_cxt.c b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
index fc767c07a264..ac284c58d8c2 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_cxt.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_cxt.c
@@ -24,11 +24,13 @@
 #include "qed_hw.h"
 #include "qed_init_ops.h"
 #include "qed_reg_addr.h"
+#include "qed_sriov.h"
 
 /* Max number of connection types in HW (DQ/CDU etc.) */
 #define MAX_CONN_TYPES		PROTOCOLID_COMMON
 #define NUM_TASK_TYPES		2
 #define NUM_TASK_PF_SEGMENTS	4
+#define NUM_TASK_VF_SEGMENTS	1
 
 /* QM constants */
 #define QM_PQ_ELEMENT_SIZE	4 /* in bytes */
@@ -63,10 +65,12 @@ union conn_context {
 struct qed_conn_type_cfg {
 	u32 cid_count;
 	u32 cid_start;
+	u32 cids_per_vf;
 };
 
 /* ILT Client configuration, Per connection type (protocol) resources. */
 #define ILT_CLI_PF_BLOCKS	(1 + NUM_TASK_PF_SEGMENTS * 2)
+#define ILT_CLI_VF_BLOCKS       (1 + NUM_TASK_VF_SEGMENTS * 2)
 #define CDUC_BLK		(0)
 
 enum ilt_clients {
@@ -97,6 +101,10 @@ struct qed_ilt_client_cfg {
 	/* ILT client blocks for PF */
 	struct qed_ilt_cli_blk pf_blks[ILT_CLI_PF_BLOCKS];
 	u32 pf_total_lines;
+
+	/* ILT client blocks for VFs */
+	struct qed_ilt_cli_blk vf_blks[ILT_CLI_VF_BLOCKS];
+	u32 vf_total_lines;
 };
 
 /* Per Path -
@@ -123,6 +131,11 @@ struct qed_cxt_mngr {
 	/* computed ILT structure */
 	struct qed_ilt_client_cfg	clients[ILT_CLI_MAX];
 
+	/* total number of VFs for this hwfn -
+	 * ALL VFs are symmetric in terms of HW resources
+	 */
+	u32				vf_count;
+
 	/* Acquired CIDs */
 	struct qed_cid_acquired_map	acquired[MAX_CONN_TYPES];
 
@@ -131,37 +144,60 @@ struct qed_cxt_mngr {
 	u32				pf_start_line;
 };
 
-static u32 qed_cxt_cdu_iids(struct qed_cxt_mngr *p_mngr)
-{
-	u32 type, pf_cids = 0;
+/* counts the iids for the CDU/CDUC ILT client configuration */
+struct qed_cdu_iids {
+	u32 pf_cids;
+	u32 per_vf_cids;
+};
 
-	for (type = 0; type < MAX_CONN_TYPES; type++)
-		pf_cids += p_mngr->conn_cfg[type].cid_count;
+static void qed_cxt_cdu_iids(struct qed_cxt_mngr *p_mngr,
+			     struct qed_cdu_iids *iids)
+{
+	u32 type;
 
-	return pf_cids;
+	for (type = 0; type < MAX_CONN_TYPES; type++) {
+		iids->pf_cids += p_mngr->conn_cfg[type].cid_count;
+		iids->per_vf_cids += p_mngr->conn_cfg[type].cids_per_vf;
+	}
 }
 
 static void qed_cxt_qm_iids(struct qed_hwfn *p_hwfn,
 			    struct qed_qm_iids *iids)
 {
 	struct qed_cxt_mngr *p_mngr = p_hwfn->p_cxt_mngr;
-	int type;
+	u32 vf_cids = 0, type;
 
-	for (type = 0; type < MAX_CONN_TYPES; type++)
+	for (type = 0; type < MAX_CONN_TYPES; type++) {
 		iids->cids += p_mngr->conn_cfg[type].cid_count;
+		vf_cids += p_mngr->conn_cfg[type].cids_per_vf;
+	}
 
-	DP_VERBOSE(p_hwfn, QED_MSG_ILT, "iids: CIDS %08x\n", iids->cids);
+	iids->vf_cids += vf_cids * p_mngr->vf_count;
+	DP_VERBOSE(p_hwfn, QED_MSG_ILT,
+		   "iids: CIDS %08x vf_cids %08x\n",
+		   iids->cids, iids->vf_cids);
 }
 
 /* set the iids count per protocol */
 static void qed_cxt_set_proto_cid_count(struct qed_hwfn *p_hwfn,
 					enum protocol_type type,
-					u32 cid_count)
+					u32 cid_count, u32 vf_cid_cnt)
 {
 	struct qed_cxt_mngr *p_mgr = p_hwfn->p_cxt_mngr;
 	struct qed_conn_type_cfg *p_conn = &p_mgr->conn_cfg[type];
 
 	p_conn->cid_count = roundup(cid_count, DQ_RANGE_ALIGN);
+	p_conn->cids_per_vf = roundup(vf_cid_cnt, DQ_RANGE_ALIGN);
+}
+
+u32 qed_cxt_get_proto_cid_count(struct qed_hwfn		*p_hwfn,
+				enum protocol_type	type,
+				u32			*vf_cid)
+{
+	if (vf_cid)
+		*vf_cid = p_hwfn->p_cxt_mngr->conn_cfg[type].cids_per_vf;
+
+	return p_hwfn->p_cxt_mngr->conn_cfg[type].cid_count;
 }
 
 static void qed_ilt_cli_blk_fill(struct qed_ilt_client_cfg *p_cli,
@@ -210,10 +246,12 @@ int qed_cxt_cfg_ilt_compute(struct qed_hwfn *p_hwfn)
 	struct qed_cxt_mngr *p_mngr = p_hwfn->p_cxt_mngr;
 	struct qed_ilt_client_cfg *p_cli;
 	struct qed_ilt_cli_blk *p_blk;
-	u32 curr_line, total, pf_cids;
+	struct qed_cdu_iids cdu_iids;
 	struct qed_qm_iids qm_iids;
+	u32 curr_line, total, i;
 
 	memset(&qm_iids, 0, sizeof(qm_iids));
+	memset(&cdu_iids, 0, sizeof(cdu_iids));
 
 	p_mngr->pf_start_line = RESC_START(p_hwfn, QED_ILT);
 
@@ -224,14 +262,16 @@ int qed_cxt_cfg_ilt_compute(struct qed_hwfn *p_hwfn)
 	/* CDUC */
 	p_cli = &p_mngr->clients[ILT_CLI_CDUC];
 	curr_line = p_mngr->pf_start_line;
+
+	/* CDUC PF */
 	p_cli->pf_total_lines = 0;
 
 	/* get the counters for the CDUC and QM clients  */
-	pf_cids = qed_cxt_cdu_iids(p_mngr);
+	qed_cxt_cdu_iids(p_mngr, &cdu_iids);
 
 	p_blk = &p_cli->pf_blks[CDUC_BLK];
 
-	total = pf_cids * CONN_CXT_SIZE(p_hwfn);
+	total = cdu_iids.pf_cids * CONN_CXT_SIZE(p_hwfn);
 
 	qed_ilt_cli_blk_fill(p_cli, p_blk, curr_line,
 			     total, CONN_CXT_SIZE(p_hwfn));
@@ -239,17 +279,36 @@ int qed_cxt_cfg_ilt_compute(struct qed_hwfn *p_hwfn)
 	qed_ilt_cli_adv_line(p_hwfn, p_cli, p_blk, &curr_line, ILT_CLI_CDUC);
 	p_cli->pf_total_lines = curr_line - p_blk->start_line;
 
+	/* CDUC VF */
+	p_blk = &p_cli->vf_blks[CDUC_BLK];
+	total = cdu_iids.per_vf_cids * CONN_CXT_SIZE(p_hwfn);
+
+	qed_ilt_cli_blk_fill(p_cli, p_blk, curr_line,
+			     total, CONN_CXT_SIZE(p_hwfn));
+
+	qed_ilt_cli_adv_line(p_hwfn, p_cli, p_blk, &curr_line, ILT_CLI_CDUC);
+	p_cli->vf_total_lines = curr_line - p_blk->start_line;
+
+	for (i = 1; i < p_mngr->vf_count; i++)
+		qed_ilt_cli_adv_line(p_hwfn, p_cli, p_blk, &curr_line,
+				     ILT_CLI_CDUC);
+
 	/* QM */
 	p_cli = &p_mngr->clients[ILT_CLI_QM];
 	p_blk = &p_cli->pf_blks[0];
 
 	qed_cxt_qm_iids(p_hwfn, &qm_iids);
-	total = qed_qm_pf_mem_size(p_hwfn->rel_pf_id, qm_iids.cids, 0, 0,
-				   p_hwfn->qm_info.num_pqs, 0);
-
-	DP_VERBOSE(p_hwfn, QED_MSG_ILT,
-		   "QM ILT Info, (cids=%d, num_pqs=%d, memory_size=%d)\n",
-		   qm_iids.cids, p_hwfn->qm_info.num_pqs, total);
+	total = qed_qm_pf_mem_size(p_hwfn->rel_pf_id, qm_iids.cids,
+				   qm_iids.vf_cids, 0,
+				   p_hwfn->qm_info.num_pqs,
+				   p_hwfn->qm_info.num_vf_pqs);
+
+	DP_VERBOSE(p_hwfn,
+		   QED_MSG_ILT,
+		   "QM ILT Info, (cids=%d, vf_cids=%d, num_pqs=%d, num_vf_pqs=%d, memory_size=%d)\n",
+		   qm_iids.cids,
+		   qm_iids.vf_cids,
+		   p_hwfn->qm_info.num_pqs, p_hwfn->qm_info.num_vf_pqs, total);
 
 	qed_ilt_cli_blk_fill(p_cli, p_blk,
 			     curr_line, total * 0x1000,
@@ -358,7 +417,7 @@ static int qed_ilt_shadow_alloc(struct qed_hwfn *p_hwfn)
 	struct qed_cxt_mngr *p_mngr = p_hwfn->p_cxt_mngr;
 	struct qed_ilt_client_cfg *clients = p_mngr->clients;
 	struct qed_ilt_cli_blk *p_blk;
-	u32 size, i, j;
+	u32 size, i, j, k;
 	int rc;
 
 	size = qed_cxt_ilt_shadow_size(clients);
@@ -383,6 +442,16 @@ static int qed_ilt_shadow_alloc(struct qed_hwfn *p_hwfn)
 			if (rc != 0)
 				goto ilt_shadow_fail;
 		}
+		for (k = 0; k < p_mngr->vf_count; k++) {
+			for (j = 0; j < ILT_CLI_VF_BLOCKS; j++) {
+				u32 lines = clients[i].vf_total_lines * k;
+
+				p_blk = &clients[i].vf_blks[j];
+				rc = qed_ilt_blk_alloc(p_hwfn, p_blk, i, lines);
+				if (rc != 0)
+					goto ilt_shadow_fail;
+			}
+		}
 	}
 
 	return 0;
@@ -467,6 +536,9 @@ int qed_cxt_mngr_alloc(struct qed_hwfn *p_hwfn)
 	for (i = 0; i < ILT_CLI_MAX; i++)
 		p_mngr->clients[i].p_size.val = ILT_DEFAULT_HW_P_SIZE;
 
+	if (p_hwfn->cdev->p_iov_info)
+		p_mngr->vf_count = p_hwfn->cdev->p_iov_info->total_vfs;
+
 	/* Set the cxt mangr pointer priori to further allocations */
 	p_hwfn->p_cxt_mngr = p_mngr;
 
@@ -579,8 +651,10 @@ void qed_qm_init_pf(struct qed_hwfn *p_hwfn)
 	params.max_phys_tcs_per_port = qm_info->max_phys_tcs_per_port;
 	params.is_first_pf = p_hwfn->first_on_engine;
 	params.num_pf_cids = iids.cids;
+	params.num_vf_cids = iids.vf_cids;
 	params.start_pq = qm_info->start_pq;
-	params.num_pf_pqs = qm_info->num_pqs;
+	params.num_pf_pqs = qm_info->num_pqs - qm_info->num_vf_pqs;
+	params.num_vf_pqs = qm_info->num_vf_pqs;
 	params.start_vport = qm_info->start_vport;
 	params.num_vports = qm_info->num_vports;
 	params.pf_wfq = qm_info->pf_wfq;
@@ -610,26 +684,55 @@ static int qed_cm_init_pf(struct qed_hwfn *p_hwfn)
 static void qed_dq_init_pf(struct qed_hwfn *p_hwfn)
 {
 	struct qed_cxt_mngr *p_mngr = p_hwfn->p_cxt_mngr;
-	u32 dq_pf_max_cid = 0;
+	u32 dq_pf_max_cid = 0, dq_vf_max_cid = 0;
 
 	dq_pf_max_cid += (p_mngr->conn_cfg[0].cid_count >> DQ_RANGE_SHIFT);
 	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_0_RT_OFFSET, dq_pf_max_cid);
 
+	dq_vf_max_cid += (p_mngr->conn_cfg[0].cids_per_vf >> DQ_RANGE_SHIFT);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_0_RT_OFFSET, dq_vf_max_cid);
+
 	dq_pf_max_cid += (p_mngr->conn_cfg[1].cid_count >> DQ_RANGE_SHIFT);
 	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_1_RT_OFFSET, dq_pf_max_cid);
 
+	dq_vf_max_cid += (p_mngr->conn_cfg[1].cids_per_vf >> DQ_RANGE_SHIFT);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_1_RT_OFFSET, dq_vf_max_cid);
+
 	dq_pf_max_cid += (p_mngr->conn_cfg[2].cid_count >> DQ_RANGE_SHIFT);
 	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_2_RT_OFFSET, dq_pf_max_cid);
 
+	dq_vf_max_cid += (p_mngr->conn_cfg[2].cids_per_vf >> DQ_RANGE_SHIFT);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_2_RT_OFFSET, dq_vf_max_cid);
+
 	dq_pf_max_cid += (p_mngr->conn_cfg[3].cid_count >> DQ_RANGE_SHIFT);
 	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_3_RT_OFFSET, dq_pf_max_cid);
 
+	dq_vf_max_cid += (p_mngr->conn_cfg[3].cids_per_vf >> DQ_RANGE_SHIFT);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_3_RT_OFFSET, dq_vf_max_cid);
+
 	dq_pf_max_cid += (p_mngr->conn_cfg[4].cid_count >> DQ_RANGE_SHIFT);
 	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_4_RT_OFFSET, dq_pf_max_cid);
 
-	/* 5 - PF */
+	dq_vf_max_cid += (p_mngr->conn_cfg[4].cids_per_vf >> DQ_RANGE_SHIFT);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_4_RT_OFFSET, dq_vf_max_cid);
+
 	dq_pf_max_cid += (p_mngr->conn_cfg[5].cid_count >> DQ_RANGE_SHIFT);
 	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_5_RT_OFFSET, dq_pf_max_cid);
+
+	dq_vf_max_cid += (p_mngr->conn_cfg[5].cids_per_vf >> DQ_RANGE_SHIFT);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_5_RT_OFFSET, dq_vf_max_cid);
+
+	/* Connection types 6 & 7 are not in use, yet they must be configured
+	 * as the highest possible connection. Not configuring them means the
+	 * defaults will be  used, and with a large number of cids a bug may
+	 * occur, if the defaults will be smaller than dq_pf_max_cid /
+	 * dq_vf_max_cid.
+	 */
+	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_6_RT_OFFSET, dq_pf_max_cid);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_6_RT_OFFSET, dq_vf_max_cid);
+
+	STORE_RT_REG(p_hwfn, DORQ_REG_PF_MAX_ICID_7_RT_OFFSET, dq_pf_max_cid);
+	STORE_RT_REG(p_hwfn, DORQ_REG_VF_MAX_ICID_7_RT_OFFSET, dq_vf_max_cid);
 }
 
 static void qed_ilt_bounds_init(struct qed_hwfn *p_hwfn)
@@ -653,6 +756,38 @@ static void qed_ilt_bounds_init(struct qed_hwfn *p_hwfn)
 	}
 }
 
+static void qed_ilt_vf_bounds_init(struct qed_hwfn *p_hwfn)
+{
+	struct qed_ilt_client_cfg *p_cli;
+	u32 blk_factor;
+
+	/* For simplicty  we set the 'block' to be an ILT page */
+	if (p_hwfn->cdev->p_iov_info) {
+		struct qed_hw_sriov_info *p_iov = p_hwfn->cdev->p_iov_info;
+
+		STORE_RT_REG(p_hwfn,
+			     PSWRQ2_REG_VF_BASE_RT_OFFSET,
+			     p_iov->first_vf_in_pf);
+		STORE_RT_REG(p_hwfn,
+			     PSWRQ2_REG_VF_LAST_ILT_RT_OFFSET,
+			     p_iov->first_vf_in_pf + p_iov->total_vfs);
+	}
+
+	p_cli = &p_hwfn->p_cxt_mngr->clients[ILT_CLI_CDUC];
+	blk_factor = ilog2(ILT_PAGE_IN_BYTES(p_cli->p_size.val) >> 10);
+	if (p_cli->active) {
+		STORE_RT_REG(p_hwfn,
+			     PSWRQ2_REG_CDUC_BLOCKS_FACTOR_RT_OFFSET,
+			     blk_factor);
+		STORE_RT_REG(p_hwfn,
+			     PSWRQ2_REG_CDUC_NUMBER_OF_PF_BLOCKS_RT_OFFSET,
+			     p_cli->pf_total_lines);
+		STORE_RT_REG(p_hwfn,
+			     PSWRQ2_REG_CDUC_VF_BLOCKS_RT_OFFSET,
+			     p_cli->vf_total_lines);
+	}
+}
+
 /* ILT (PSWRQ2) PF */
 static void qed_ilt_init_pf(struct qed_hwfn *p_hwfn)
 {
@@ -662,6 +797,7 @@ static void qed_ilt_init_pf(struct qed_hwfn *p_hwfn)
 	u32 line, rt_offst, i;
 
 	qed_ilt_bounds_init(p_hwfn);
+	qed_ilt_vf_bounds_init(p_hwfn);
 
 	p_mngr = p_hwfn->p_cxt_mngr;
 	p_shdw = p_mngr->ilt_shadow;
@@ -839,10 +975,10 @@ int qed_cxt_set_pf_params(struct qed_hwfn *p_hwfn)
 	/* Set the number of required CORE connections */
 	u32 core_cids = 1; /* SPQ */
 
-	qed_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_CORE, core_cids);
+	qed_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_CORE, core_cids, 0);
 
 	qed_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_ETH,
-				    p_params->num_cons);
+				    p_params->num_cons, 1);
 
 	return 0;
 }
diff --git a/drivers/net/ethernet/qlogic/qed/qed_cxt.h b/drivers/net/ethernet/qlogic/qed/qed_cxt.h
index c8e1f5e5c42b..078ff3fd7920 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_cxt.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_cxt.h
@@ -51,6 +51,9 @@ enum qed_cxt_elem_type {
 	QED_ELEM_TASK
 };
 
+u32 qed_cxt_get_proto_cid_count(struct qed_hwfn *p_hwfn,
+				enum protocol_type type, u32 *vf_cid);
+
 /**
  * @brief qed_cxt_set_pf_params - Set the PF params for cxt init
  *
diff --git a/drivers/net/ethernet/qlogic/qed/qed_dev.c b/drivers/net/ethernet/qlogic/qed/qed_dev.c
index 7a359c45360f..362e8db2b374 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_dev.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_dev.c
@@ -41,10 +41,14 @@ enum BAR_ID {
 static u32 qed_hw_bar_size(struct qed_hwfn	*p_hwfn,
 			   enum BAR_ID		bar_id)
 {
-	u32	bar_reg = (bar_id == BAR_ID_0 ?
-			   PGLUE_B_REG_PF_BAR0_SIZE : PGLUE_B_REG_PF_BAR1_SIZE);
-	u32	val = qed_rd(p_hwfn, p_hwfn->p_main_ptt, bar_reg);
+	u32 bar_reg = (bar_id == BAR_ID_0 ?
+		       PGLUE_B_REG_PF_BAR0_SIZE : PGLUE_B_REG_PF_BAR1_SIZE);
+	u32 val;
 
+	if (IS_VF(p_hwfn->cdev))
+		return 1 << 17;
+
+	val = qed_rd(p_hwfn, p_hwfn->p_main_ptt, bar_reg);
 	if (val)
 		return 1 << (val + 15);
 
@@ -114,6 +118,9 @@ void qed_resc_free(struct qed_dev *cdev)
 {
 	int i;
 
+	if (IS_VF(cdev))
+		return;
+
 	kfree(cdev->fw_data);
 	cdev->fw_data = NULL;
 
@@ -144,14 +151,19 @@ void qed_resc_free(struct qed_dev *cdev)
 
 static int qed_init_qm_info(struct qed_hwfn *p_hwfn)
 {
+	u8 num_vports, vf_offset = 0, i, vport_id, num_ports, curr_queue = 0;
 	struct qed_qm_info *qm_info = &p_hwfn->qm_info;
 	struct init_qm_port_params *p_qm_port;
-	u8 num_vports, i, vport_id, num_ports;
 	u16 num_pqs, multi_cos_tcs = 1;
+	u16 num_vfs = 0;
 
+#ifdef CONFIG_QED_SRIOV
+	if (p_hwfn->cdev->p_iov_info)
+		num_vfs = p_hwfn->cdev->p_iov_info->total_vfs;
+#endif
 	memset(qm_info, 0, sizeof(*qm_info));
 
-	num_pqs = multi_cos_tcs + 1; /* The '1' is for pure-LB */
+	num_pqs = multi_cos_tcs + num_vfs + 1;	/* The '1' is for pure-LB */
 	num_vports = (u8)RESC_NUM(p_hwfn, QED_VPORT);
 
 	/* Sanity checking that setup requires legal number of resources */
@@ -187,8 +199,9 @@ static int qed_init_qm_info(struct qed_hwfn *p_hwfn)
 	vport_id = (u8)RESC_START(p_hwfn, QED_VPORT);
 
 	/* First init per-TC PQs */
-	for (i = 0; i < multi_cos_tcs; i++) {
-		struct init_qm_pq_params *params = &qm_info->qm_pq_params[i];
+	for (i = 0; i < multi_cos_tcs; i++, curr_queue++) {
+		struct init_qm_pq_params *params =
+		    &qm_info->qm_pq_params[curr_queue];
 
 		params->vport_id = vport_id;
 		params->tc_id = p_hwfn->hw_info.non_offload_tc;
@@ -196,13 +209,26 @@ static int qed_init_qm_info(struct qed_hwfn *p_hwfn)
 	}
 
 	/* Then init pure-LB PQ */
-	qm_info->pure_lb_pq = i;
-	qm_info->qm_pq_params[i].vport_id = (u8)RESC_START(p_hwfn, QED_VPORT);
-	qm_info->qm_pq_params[i].tc_id = PURE_LB_TC;
-	qm_info->qm_pq_params[i].wrr_group = 1;
-	i++;
+	qm_info->pure_lb_pq = curr_queue;
+	qm_info->qm_pq_params[curr_queue].vport_id =
+	    (u8) RESC_START(p_hwfn, QED_VPORT);
+	qm_info->qm_pq_params[curr_queue].tc_id = PURE_LB_TC;
+	qm_info->qm_pq_params[curr_queue].wrr_group = 1;
+	curr_queue++;
 
 	qm_info->offload_pq = 0;
+	/* Then init per-VF PQs */
+	vf_offset = curr_queue;
+	for (i = 0; i < num_vfs; i++) {
+		/* First vport is used by the PF */
+		qm_info->qm_pq_params[curr_queue].vport_id = vport_id + i + 1;
+		qm_info->qm_pq_params[curr_queue].tc_id =
+		    p_hwfn->hw_info.non_offload_tc;
+		qm_info->qm_pq_params[curr_queue].wrr_group = 1;
+		curr_queue++;
+	}
+
+	qm_info->vf_queues_offset = vf_offset;
 	qm_info->num_pqs = num_pqs;
 	qm_info->num_vports = num_vports;
 
@@ -220,7 +246,8 @@ static int qed_init_qm_info(struct qed_hwfn *p_hwfn)
 
 	qm_info->start_pq = (u16)RESC_START(p_hwfn, QED_PQ);
 
-	qm_info->start_vport = (u8)RESC_START(p_hwfn, QED_VPORT);
+	qm_info->num_vf_pqs = num_vfs;
+	qm_info->start_vport = (u8) RESC_START(p_hwfn, QED_VPORT);
 
 	for (i = 0; i < qm_info->num_vports; i++)
 		qm_info->qm_vport_params[i].vport_wfq = 1;
@@ -244,6 +271,9 @@ int qed_resc_alloc(struct qed_dev *cdev)
 	struct qed_eq *p_eq;
 	int i, rc = 0;
 
+	if (IS_VF(cdev))
+		return rc;
+
 	cdev->fw_data = kzalloc(sizeof(*cdev->fw_data), GFP_KERNEL);
 	if (!cdev->fw_data)
 		return -ENOMEM;
@@ -364,6 +394,9 @@ void qed_resc_setup(struct qed_dev *cdev)
 {
 	int i;
 
+	if (IS_VF(cdev))
+		return;
+
 	for_each_hwfn(cdev, i) {
 		struct qed_hwfn *p_hwfn = &cdev->hwfns[i];
 
@@ -508,7 +541,9 @@ static int qed_hw_init_common(struct qed_hwfn *p_hwfn,
 	struct qed_qm_info *qm_info = &p_hwfn->qm_info;
 	struct qed_qm_common_rt_init_params params;
 	struct qed_dev *cdev = p_hwfn->cdev;
+	u32 concrete_fid;
 	int rc = 0;
+	u8 vf_id;
 
 	qed_init_cau_rt_data(cdev);
 
@@ -558,6 +593,14 @@ static int qed_hw_init_common(struct qed_hwfn *p_hwfn,
 	qed_wr(p_hwfn, p_ptt, 0x20b4,
 	       qed_rd(p_hwfn, p_ptt, 0x20b4) & ~0x10);
 
+	for (vf_id = 0; vf_id < MAX_NUM_VFS_BB; vf_id++) {
+		concrete_fid = qed_vfid_to_concrete(p_hwfn, vf_id);
+		qed_fid_pretend(p_hwfn, p_ptt, (u16) concrete_fid);
+		qed_wr(p_hwfn, p_ptt, CCFC_REG_STRONG_ENABLE_VF, 0x1);
+	}
+	/* pretend to original PF */
+	qed_fid_pretend(p_hwfn, p_ptt, p_hwfn->rel_pf_id);
+
 	return rc;
 }
 
@@ -698,13 +741,20 @@ int qed_hw_init(struct qed_dev *cdev,
 	u32 load_code, param;
 	int rc, mfw_rc, i;
 
-	rc = qed_init_fw_data(cdev, bin_fw_data);
-	if (rc != 0)
-		return rc;
+	if (IS_PF(cdev)) {
+		rc = qed_init_fw_data(cdev, bin_fw_data);
+		if (rc != 0)
+			return rc;
+	}
 
 	for_each_hwfn(cdev, i) {
 		struct qed_hwfn *p_hwfn = &cdev->hwfns[i];
 
+		if (IS_VF(cdev)) {
+			p_hwfn->b_int_enabled = 1;
+			continue;
+		}
+
 		/* Enable DMAE in PXP */
 		rc = qed_change_pci_hwfn(p_hwfn, p_hwfn->p_main_ptt, true);
 
@@ -829,6 +879,11 @@ int qed_hw_stop(struct qed_dev *cdev)
 
 		DP_VERBOSE(p_hwfn, NETIF_MSG_IFDOWN, "Stopping hw/fw\n");
 
+		if (IS_VF(cdev)) {
+			/* To be implemented in a later patch */
+			continue;
+		}
+
 		/* mark the hw as uninitialized... */
 		p_hwfn->hw_init_done = false;
 
@@ -860,15 +915,16 @@ int qed_hw_stop(struct qed_dev *cdev)
 		usleep_range(1000, 2000);
 	}
 
-	/* Disable DMAE in PXP - in CMT, this should only be done for
-	 * first hw-function, and only after all transactions have
-	 * stopped for all active hw-functions.
-	 */
-	t_rc = qed_change_pci_hwfn(&cdev->hwfns[0],
-				   cdev->hwfns[0].p_main_ptt,
-				   false);
-	if (t_rc != 0)
-		rc = t_rc;
+	if (IS_PF(cdev)) {
+		/* Disable DMAE in PXP - in CMT, this should only be done for
+		 * first hw-function, and only after all transactions have
+		 * stopped for all active hw-functions.
+		 */
+		t_rc = qed_change_pci_hwfn(&cdev->hwfns[0],
+					   cdev->hwfns[0].p_main_ptt, false);
+		if (t_rc != 0)
+			rc = t_rc;
+	}
 
 	return rc;
 }
@@ -932,6 +988,11 @@ int qed_hw_reset(struct qed_dev *cdev)
 	for_each_hwfn(cdev, i) {
 		struct qed_hwfn *p_hwfn = &cdev->hwfns[i];
 
+		if (IS_VF(cdev)) {
+			/* Will be implemented in a later patch */
+			continue;
+		}
+
 		DP_VERBOSE(p_hwfn, NETIF_MSG_IFDOWN, "Resetting hw/fw\n");
 
 		/* Check for incorrect states */
@@ -1027,11 +1088,10 @@ static void qed_hw_set_feat(struct qed_hwfn *p_hwfn)
 static void qed_hw_get_resc(struct qed_hwfn *p_hwfn)
 {
 	u32 *resc_start = p_hwfn->hw_info.resc_start;
+	u8 num_funcs = p_hwfn->num_funcs_on_engine;
 	u32 *resc_num = p_hwfn->hw_info.resc_num;
 	struct qed_sb_cnt_info sb_cnt_info;
-	int num_funcs, i;
-
-	num_funcs = MAX_NUM_PFS_BB;
+	int i;
 
 	memset(&sb_cnt_info, 0, sizeof(sb_cnt_info));
 	qed_int_get_num_sbs(p_hwfn, &sb_cnt_info);
@@ -1238,6 +1298,51 @@ static int qed_hw_get_nvm_info(struct qed_hwfn *p_hwfn,
 	return qed_mcp_fill_shmem_func_info(p_hwfn, p_ptt);
 }
 
+static void qed_get_num_funcs(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+	u32 reg_function_hide, tmp, eng_mask;
+	u8 num_funcs;
+
+	num_funcs = MAX_NUM_PFS_BB;
+
+	/* Bit 0 of MISCS_REG_FUNCTION_HIDE indicates whether the bypass values
+	 * in the other bits are selected.
+	 * Bits 1-15 are for functions 1-15, respectively, and their value is
+	 * '0' only for enabled functions (function 0 always exists and
+	 * enabled).
+	 * In case of CMT, only the "even" functions are enabled, and thus the
+	 * number of functions for both hwfns is learnt from the same bits.
+	 */
+	reg_function_hide = qed_rd(p_hwfn, p_ptt, MISCS_REG_FUNCTION_HIDE);
+
+	if (reg_function_hide & 0x1) {
+		if (QED_PATH_ID(p_hwfn) && p_hwfn->cdev->num_hwfns == 1) {
+			num_funcs = 0;
+			eng_mask = 0xaaaa;
+		} else {
+			num_funcs = 1;
+			eng_mask = 0x5554;
+		}
+
+		/* Get the number of the enabled functions on the engine */
+		tmp = (reg_function_hide ^ 0xffffffff) & eng_mask;
+		while (tmp) {
+			if (tmp & 0x1)
+				num_funcs++;
+			tmp >>= 0x1;
+		}
+	}
+
+	p_hwfn->num_funcs_on_engine = num_funcs;
+
+	DP_VERBOSE(p_hwfn,
+		   NETIF_MSG_PROBE,
+		   "PF [rel_id %d, abs_id %d] within the %d enabled functions on the engine\n",
+		   p_hwfn->rel_pf_id,
+		   p_hwfn->abs_pf_id,
+		   p_hwfn->num_funcs_on_engine);
+}
+
 static int
 qed_get_hw_info(struct qed_hwfn *p_hwfn,
 		struct qed_ptt *p_ptt,
@@ -1296,6 +1401,8 @@ qed_get_hw_info(struct qed_hwfn *p_hwfn,
 		p_hwfn->hw_info.personality = protocol;
 	}
 
+	qed_get_num_funcs(p_hwfn, p_ptt);
+
 	qed_hw_get_resc(p_hwfn);
 
 	return rc;
@@ -1361,6 +1468,9 @@ static int qed_hw_prepare_single(struct qed_hwfn *p_hwfn,
 	p_hwfn->regview = p_regview;
 	p_hwfn->doorbells = p_doorbells;
 
+	if (IS_VF(p_hwfn->cdev))
+		return qed_vf_hw_prepare(p_hwfn);
+
 	/* Validate that chip access is feasible */
 	if (REG_RD(p_hwfn, PXP_PF_ME_OPAQUE_ADDR) == 0xffffffff) {
 		DP_ERR(p_hwfn,
@@ -1428,7 +1538,8 @@ int qed_hw_prepare(struct qed_dev *cdev,
 	int rc;
 
 	/* Store the precompiled init data ptrs */
-	qed_init_iro_array(cdev);
+	if (IS_PF(cdev))
+		qed_init_iro_array(cdev);
 
 	/* Initialize the first hwfn - will learn number of hwfns */
 	rc = qed_hw_prepare_single(p_hwfn,
@@ -1460,9 +1571,11 @@ int qed_hw_prepare(struct qed_dev *cdev,
 		 * initiliazed hwfn 0.
 		 */
 		if (rc) {
-			qed_init_free(p_hwfn);
-			qed_mcp_free(p_hwfn);
-			qed_hw_hwfn_free(p_hwfn);
+			if (IS_PF(cdev)) {
+				qed_init_free(p_hwfn);
+				qed_mcp_free(p_hwfn);
+				qed_hw_hwfn_free(p_hwfn);
+			}
 		}
 	}
 
@@ -1476,6 +1589,11 @@ void qed_hw_remove(struct qed_dev *cdev)
 	for_each_hwfn(cdev, i) {
 		struct qed_hwfn *p_hwfn = &cdev->hwfns[i];
 
+		if (IS_VF(cdev)) {
+			/* Will be implemented in a later patch */
+			continue;
+		}
+
 		qed_init_free(p_hwfn);
 		qed_hw_hwfn_free(p_hwfn);
 		qed_mcp_free(p_hwfn);
diff --git a/drivers/net/ethernet/qlogic/qed/qed_hsi.h b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
index 6ba197c107ed..c511106870d0 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_hsi.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_hsi.h
@@ -29,7 +29,7 @@ struct qed_ptt;
 enum common_event_opcode {
 	COMMON_EVENT_PF_START,
 	COMMON_EVENT_PF_STOP,
-	COMMON_EVENT_RESERVED,
+	COMMON_EVENT_VF_START,
 	COMMON_EVENT_RESERVED2,
 	COMMON_EVENT_VF_PF_CHANNEL,
 	COMMON_EVENT_RESERVED4,
@@ -44,7 +44,7 @@ enum common_ramrod_cmd_id {
 	COMMON_RAMROD_UNUSED,
 	COMMON_RAMROD_PF_START /* PF Function Start Ramrod */,
 	COMMON_RAMROD_PF_STOP /* PF Function Stop Ramrod */,
-	COMMON_RAMROD_RESERVED,
+	COMMON_RAMROD_VF_START,
 	COMMON_RAMROD_RESERVED2,
 	COMMON_RAMROD_PF_UPDATE,
 	COMMON_RAMROD_EMPTY,
@@ -573,6 +573,14 @@ union event_ring_element {
 	struct event_ring_next_addr	next_addr;
 };
 
+struct mstorm_non_trigger_vf_zone {
+	struct eth_mstorm_per_queue_stat eth_queue_stat;
+};
+
+struct mstorm_vf_zone {
+	struct mstorm_non_trigger_vf_zone non_trigger;
+};
+
 enum personality_type {
 	BAD_PERSONALITY_TYP,
 	PERSONALITY_RESERVED,
@@ -671,6 +679,16 @@ enum ports_mode {
 	MAX_PORTS_MODE
 };
 
+struct pstorm_non_trigger_vf_zone {
+	struct eth_pstorm_per_queue_stat eth_queue_stat;
+	struct regpair reserved[2];
+};
+
+struct pstorm_vf_zone {
+	struct pstorm_non_trigger_vf_zone non_trigger;
+	struct regpair reserved[7];
+};
+
 /* Ramrod Header of SPQE */
 struct ramrod_header {
 	__le32	cid /* Slowpath Connection CID */;
@@ -700,6 +718,29 @@ struct tstorm_per_port_stat {
 	struct regpair	preroce_irregular_pkt;
 };
 
+struct ustorm_non_trigger_vf_zone {
+	struct eth_ustorm_per_queue_stat eth_queue_stat;
+	struct regpair vf_pf_msg_addr;
+};
+
+struct ustorm_trigger_vf_zone {
+	u8 vf_pf_msg_valid;
+	u8 reserved[7];
+};
+
+struct ustorm_vf_zone {
+	struct ustorm_non_trigger_vf_zone non_trigger;
+	struct ustorm_trigger_vf_zone trigger;
+};
+
+struct vf_start_ramrod_data {
+	u8 vf_id;
+	u8 enable_flr_ack;
+	__le16 opaque_fid;
+	u8 personality;
+	u8 reserved[3];
+};
+
 struct atten_status_block {
 	__le32	atten_bits;
 	__le32	atten_ack;
@@ -1026,7 +1067,7 @@ enum init_phases {
 	PHASE_ENGINE,
 	PHASE_PORT,
 	PHASE_PF,
-	PHASE_RESERVED,
+	PHASE_VF,
 	PHASE_QM_PF,
 	MAX_INIT_PHASES
 };
diff --git a/drivers/net/ethernet/qlogic/qed/qed_hw.c b/drivers/net/ethernet/qlogic/qed/qed_hw.c
index a9be5a422d2d..0ada7fdb91bc 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_hw.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_hw.c
@@ -23,6 +23,7 @@
 #include "qed_hsi.h"
 #include "qed_hw.h"
 #include "qed_reg_addr.h"
+#include "qed_sriov.h"
 
 #define QED_BAR_ACQUIRE_TIMEOUT 1000
 
@@ -236,8 +237,12 @@ static void qed_memcpy_hw(struct qed_hwfn *p_hwfn,
 		quota = min_t(size_t, n - done,
 			      PXP_EXTERNAL_BAR_PF_WINDOW_SINGLE_SIZE);
 
-		qed_ptt_set_win(p_hwfn, p_ptt, hw_addr + done);
-		hw_offset = qed_ptt_get_bar_addr(p_ptt)