Merge branch 'occteontx2-rate-limit-offload'

Subbaraya Sundeep says:

====================
octeontx2: Add ingress ratelimit offload

This patchset adds ingress rate limiting hardware
offload support for CN10K silicons. Police actions
are added for TC matchall and flower filters.
CN10K has ingress rate limiting feature where
a receive queue is mapped to bandwidth profile
and the profile is configured with rate and burst
parameters by software. CN10K hardware supports
three levels of ingress policing or ratelimiting.
Multiple leaf profiles can  point to a single mid
level profile and multiple mid level profile can
point to a single top level one. Only leaf level
profiles are used for configuring rate limiting.

Patch 1 adds the new bandwidth profile contexts
in AF driver similar to other hardware contexts
Patch 2 adds the debugfs changes to dump bandwidth
profile contexts
Patch 3 adds support for police action with TC matchall filter
Patch 4 uses NL_SET_ERR_MSG_MOD for tc code
Patch 5 adds support for police action with TC flower filter
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

14 files changed, 1554 insertions, 35 deletions

diff --git a/drivers/net/ethernet/marvell/octeontx2/af/mbox.h b/drivers/net/ethernet/marvell/octeontx2/af/mbox.h
index aee6a6f31b0d..7d7dfa8d8a3f 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/mbox.h
+++ b/drivers/net/ethernet/marvell/octeontx2/af/mbox.h
@@ -260,7 +260,11 @@ M(NIX_BP_DISABLE,	0x8017, nix_bp_disable, nix_bp_cfg_req, msg_rsp) \
 M(NIX_GET_MAC_ADDR, 0x8018, nix_get_mac_addr, msg_req, nix_get_mac_addr_rsp) \
 M(NIX_CN10K_AQ_ENQ,	0x8019, nix_cn10k_aq_enq, nix_cn10k_aq_enq_req, \
 				nix_cn10k_aq_enq_rsp)			\
-M(NIX_GET_HW_INFO,	0x801a, nix_get_hw_info, msg_req, nix_hw_info)
+M(NIX_GET_HW_INFO,	0x801c, nix_get_hw_info, msg_req, nix_hw_info)	\
+M(NIX_BANDPROF_ALLOC,	0x801d, nix_bandprof_alloc, nix_bandprof_alloc_req, \
+				nix_bandprof_alloc_rsp)			    \
+M(NIX_BANDPROF_FREE,	0x801e, nix_bandprof_free, nix_bandprof_free_req,   \
+				msg_rsp)
 
 /* Messages initiated by AF (range 0xC00 - 0xDFF) */
 #define MBOX_UP_CGX_MESSAGES						\
@@ -615,6 +619,9 @@ enum nix_af_status {
 	NIX_AF_ERR_PTP_CONFIG_FAIL  = -423,
 	NIX_AF_ERR_NPC_KEY_NOT_SUPP = -424,
 	NIX_AF_ERR_INVALID_NIXBLK   = -425,
+	NIX_AF_ERR_INVALID_BANDPROF = -426,
+	NIX_AF_ERR_IPOLICER_NOTSUPP = -427,
+	NIX_AF_ERR_BANDPROF_INVAL_REQ  = -428,
 };
 
 /* For NIX RX vtag action  */
@@ -683,6 +690,7 @@ struct nix_cn10k_aq_enq_req {
 		struct nix_cq_ctx_s cq;
 		struct nix_rsse_s   rss;
 		struct nix_rx_mce_s mce;
+		struct nix_bandprof_s prof;
 	};
 	union {
 		struct nix_cn10k_rq_ctx_s rq_mask;
@@ -690,6 +698,7 @@ struct nix_cn10k_aq_enq_req {
 		struct nix_cq_ctx_s cq_mask;
 		struct nix_rsse_s   rss_mask;
 		struct nix_rx_mce_s mce_mask;
+		struct nix_bandprof_s prof_mask;
 	};
 };
 
@@ -701,6 +710,7 @@ struct nix_cn10k_aq_enq_rsp {
 		struct nix_cq_ctx_s cq;
 		struct nix_rsse_s   rss;
 		struct nix_rx_mce_s mce;
+		struct nix_bandprof_s prof;
 	};
 };
 
@@ -716,6 +726,7 @@ struct nix_aq_enq_req {
 		struct nix_cq_ctx_s cq;
 		struct nix_rsse_s   rss;
 		struct nix_rx_mce_s mce;
+		u64 prof;
 	};
 	union {
 		struct nix_rq_ctx_s rq_mask;
@@ -723,6 +734,7 @@ struct nix_aq_enq_req {
 		struct nix_cq_ctx_s cq_mask;
 		struct nix_rsse_s   rss_mask;
 		struct nix_rx_mce_s mce_mask;
+		u64 prof_mask;
 	};
 };
 
@@ -734,6 +746,7 @@ struct nix_aq_enq_rsp {
 		struct nix_cq_ctx_s cq;
 		struct nix_rsse_s   rss;
 		struct nix_rx_mce_s mce;
+		u64 prof;
 	};
 };
 
@@ -975,6 +988,31 @@ struct nix_hw_info {
 	u16 min_mtu;
 };
 
+struct nix_bandprof_alloc_req {
+	struct mbox_msghdr hdr;
+	/* Count of profiles needed per layer */
+	u16 prof_count[BAND_PROF_NUM_LAYERS];
+};
+
+struct nix_bandprof_alloc_rsp {
+	struct mbox_msghdr hdr;
+	u16 prof_count[BAND_PROF_NUM_LAYERS];
+
+	/* There is no need to allocate morethan 1 bandwidth profile
+	 * per RQ of a PF_FUNC's NIXLF. So limit the maximum
+	 * profiles to 64 per PF_FUNC.
+	 */
+#define MAX_BANDPROF_PER_PFFUNC	64
+	u16 prof_idx[BAND_PROF_NUM_LAYERS][MAX_BANDPROF_PER_PFFUNC];
+};
+
+struct nix_bandprof_free_req {
+	struct mbox_msghdr hdr;
+	u8 free_all;
+	u16 prof_count[BAND_PROF_NUM_LAYERS];
+	u16 prof_idx[BAND_PROF_NUM_LAYERS][MAX_BANDPROF_PER_PFFUNC];
+};
+
 /* NPC mbox message structs */
 
 #define NPC_MCAM_ENTRY_INVALID	0xFFFF
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu.c
index f11a02d6b6ef..0b092949d7ac 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.c
@@ -184,6 +184,14 @@ int rvu_rsrc_free_count(struct rsrc_bmap *rsrc)
 	return (rsrc->max - used);
 }
 
+bool is_rsrc_free(struct rsrc_bmap *rsrc, int id)
+{
+	if (!rsrc->bmap)
+		return false;
+
+	return !test_bit(id, rsrc->bmap);
+}
+
 int rvu_alloc_bitmap(struct rsrc_bmap *rsrc)
 {
 	rsrc->bmap = kcalloc(BITS_TO_LONGS(rsrc->max),
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
index c88dab7747ef..9e5d9ba6f01e 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
@@ -296,6 +296,13 @@ struct nix_txvlan {
 	struct mutex rsrc_lock; /* Serialize resource alloc/free */
 };
 
+struct nix_ipolicer {
+	struct rsrc_bmap band_prof;
+	u16 *pfvf_map;
+	u16 *match_id;
+	u16 *ref_count;
+};
+
 struct nix_hw {
 	int blkaddr;
 	struct rvu *rvu;
@@ -305,6 +312,7 @@ struct nix_hw {
 	struct nix_mark_format mark_format;
 	struct nix_lso lso;
 	struct nix_txvlan txvlan;
+	struct nix_ipolicer *ipolicer;
 };
 
 /* RVU block's capabilities or functionality,
@@ -322,6 +330,7 @@ struct hw_cap {
 	bool	nix_rx_multicast;	 /* Rx packet replication support */
 	bool	per_pf_mbox_regs; /* PF mbox specified in per PF registers ? */
 	bool	programmable_chans; /* Channels programmable ? */
+	bool	ipolicer;
 };
 
 struct rvu_hwinfo {
@@ -587,6 +596,7 @@ static inline bool is_rvu_fwdata_valid(struct rvu *rvu)
 int rvu_alloc_bitmap(struct rsrc_bmap *rsrc);
 int rvu_alloc_rsrc(struct rsrc_bmap *rsrc);
 void rvu_free_rsrc(struct rsrc_bmap *rsrc, int id);
+bool is_rsrc_free(struct rsrc_bmap *rsrc, int id);
 int rvu_rsrc_free_count(struct rsrc_bmap *rsrc);
 int rvu_alloc_rsrc_contig(struct rsrc_bmap *rsrc, int nrsrc);
 bool rvu_rsrc_check_contig(struct rsrc_bmap *rsrc, int nrsrc);
@@ -672,6 +682,12 @@ int rvu_get_next_nix_blkaddr(struct rvu *rvu, int blkaddr);
 void rvu_nix_reset_mac(struct rvu_pfvf *pfvf, int pcifunc);
 int nix_get_struct_ptrs(struct rvu *rvu, u16 pcifunc,
 			struct nix_hw **nix_hw, int *blkaddr);
+int rvu_nix_setup_ratelimit_aggr(struct rvu *rvu, u16 pcifunc,
+				 u16 rq_idx, u16 match_id);
+int nix_aq_context_read(struct rvu *rvu, struct nix_hw *nix_hw,
+			struct nix_cn10k_aq_enq_req *aq_req,
+			struct nix_cn10k_aq_enq_rsp *aq_rsp,
+			u16 pcifunc, u8 ctype, u32 qidx);
 
 /* NPC APIs */
 int rvu_npc_init(struct rvu *rvu);
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c
index 7103f8216ad1..3cc3c6fd1d84 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_debugfs.c
@@ -1632,6 +1632,165 @@ static int rvu_dbg_nix_qsize_display(struct seq_file *filp, void *unused)
 
 RVU_DEBUG_SEQ_FOPS(nix_qsize, nix_qsize_display, nix_qsize_write);
 
+static void print_band_prof_ctx(struct seq_file *m,
+				struct nix_bandprof_s *prof)
+{
+	char *str;
+
+	switch (prof->pc_mode) {
+	case NIX_RX_PC_MODE_VLAN:
+		str = "VLAN";
+		break;
+	case NIX_RX_PC_MODE_DSCP:
+		str = "DSCP";
+		break;
+	case NIX_RX_PC_MODE_GEN:
+		str = "Generic";
+		break;
+	case NIX_RX_PC_MODE_RSVD:
+		str = "Reserved";
+		break;
+	}
+	seq_printf(m, "W0: pc_mode\t\t%s\n", str);
+	str = (prof->icolor == 3) ? "Color blind" :
+		(prof->icolor == 0) ? "Green" :
+		(prof->icolor == 1) ? "Yellow" : "Red";
+	seq_printf(m, "W0: icolor\t\t%s\n", str);
+	seq_printf(m, "W0: tnl_ena\t\t%d\n", prof->tnl_ena);
+	seq_printf(m, "W0: peir_exponent\t%d\n", prof->peir_exponent);
+	seq_printf(m, "W0: pebs_exponent\t%d\n", prof->pebs_exponent);
+	seq_printf(m, "W0: cir_exponent\t%d\n", prof->cir_exponent);
+	seq_printf(m, "W0: cbs_exponent\t%d\n", prof->cbs_exponent);
+	seq_printf(m, "W0: peir_mantissa\t%d\n", prof->peir_mantissa);
+	seq_printf(m, "W0: pebs_mantissa\t%d\n", prof->pebs_mantissa);
+	seq_printf(m, "W0: cir_mantissa\t%d\n", prof->cir_mantissa);
+
+	seq_printf(m, "W1: cbs_mantissa\t%d\n", prof->cbs_mantissa);
+	str = (prof->lmode == 0) ? "byte" : "packet";
+	seq_printf(m, "W1: lmode\t\t%s\n", str);
+	seq_printf(m, "W1: l_select\t\t%d\n", prof->l_sellect);
+	seq_printf(m, "W1: rdiv\t\t%d\n", prof->rdiv);
+	seq_printf(m, "W1: adjust_exponent\t%d\n", prof->adjust_exponent);
+	seq_printf(m, "W1: adjust_mantissa\t%d\n", prof->adjust_mantissa);
+	str = (prof->gc_action == 0) ? "PASS" :
+		(prof->gc_action == 1) ? "DROP" : "RED";
+	seq_printf(m, "W1: gc_action\t\t%s\n", str);
+	str = (prof->yc_action == 0) ? "PASS" :
+		(prof->yc_action == 1) ? "DROP" : "RED";
+	seq_printf(m, "W1: yc_action\t\t%s\n", str);
+	str = (prof->rc_action == 0) ? "PASS" :
+		(prof->rc_action == 1) ? "DROP" : "RED";
+	seq_printf(m, "W1: rc_action\t\t%s\n", str);
+	seq_printf(m, "W1: meter_algo\t\t%d\n", prof->meter_algo);
+	seq_printf(m, "W1: band_prof_id\t%d\n", prof->band_prof_id);
+	seq_printf(m, "W1: hl_en\t\t%d\n", prof->hl_en);
+
+	seq_printf(m, "W2: ts\t\t\t%lld\n", (u64)prof->ts);
+	seq_printf(m, "W3: pe_accum\t\t%d\n", prof->pe_accum);
+	seq_printf(m, "W3: c_accum\t\t%d\n", prof->c_accum);
+	seq_printf(m, "W4: green_pkt_pass\t%lld\n",
+		   (u64)prof->green_pkt_pass);
+	seq_printf(m, "W5: yellow_pkt_pass\t%lld\n",
+		   (u64)prof->yellow_pkt_pass);
+	seq_printf(m, "W6: red_pkt_pass\t%lld\n", (u64)prof->red_pkt_pass);
+	seq_printf(m, "W7: green_octs_pass\t%lld\n",
+		   (u64)prof->green_octs_pass);
+	seq_printf(m, "W8: yellow_octs_pass\t%lld\n",
+		   (u64)prof->yellow_octs_pass);
+	seq_printf(m, "W9: red_octs_pass\t%lld\n", (u64)prof->red_octs_pass);
+	seq_printf(m, "W10: green_pkt_drop\t%lld\n",
+		   (u64)prof->green_pkt_drop);
+	seq_printf(m, "W11: yellow_pkt_drop\t%lld\n",
+		   (u64)prof->yellow_pkt_drop);
+	seq_printf(m, "W12: red_pkt_drop\t%lld\n", (u64)prof->red_pkt_drop);
+	seq_printf(m, "W13: green_octs_drop\t%lld\n",
+		   (u64)prof->green_octs_drop);
+	seq_printf(m, "W14: yellow_octs_drop\t%lld\n",
+		   (u64)prof->yellow_octs_drop);
+	seq_printf(m, "W15: red_octs_drop\t%lld\n", (u64)prof->red_octs_drop);
+	seq_puts(m, "==============================\n");
+}
+
+static int rvu_dbg_nix_band_prof_ctx_display(struct seq_file *m, void *unused)
+{
+	struct nix_hw *nix_hw = m->private;
+	struct nix_cn10k_aq_enq_req aq_req;
+	struct nix_cn10k_aq_enq_rsp aq_rsp;
+	struct rvu *rvu = nix_hw->rvu;
+	struct nix_ipolicer *ipolicer;
+	int layer, prof_idx, idx, rc;
+	u16 pcifunc;
+	char *str;
+
+	for (layer = 0; layer < BAND_PROF_NUM_LAYERS; layer++) {
+		if (layer == BAND_PROF_INVAL_LAYER)
+			continue;
+		str = (layer == BAND_PROF_LEAF_LAYER) ? "Leaf" :
+			(layer == BAND_PROF_MID_LAYER) ? "Mid" : "Top";
+
+		seq_printf(m, "\n%s bandwidth profiles\n", str);
+		seq_puts(m, "=======================\n");
+
+		ipolicer = &nix_hw->ipolicer[layer];
+
+		for (idx = 0; idx < ipolicer->band_prof.max; idx++) {
+			if (is_rsrc_free(&ipolicer->band_prof, idx))
+				continue;
+
+			prof_idx = (idx & 0x3FFF) | (layer << 14);
+			rc = nix_aq_context_read(rvu, nix_hw, &aq_req, &aq_rsp,
+						 0x00, NIX_AQ_CTYPE_BANDPROF,
+						 prof_idx);
+			if (rc) {
+				dev_err(rvu->dev,
+					"%s: Failed to fetch context of %s profile %d, err %d\n",
+					__func__, str, idx, rc);
+				return 0;
+			}
+			seq_printf(m, "\n%s bandwidth profile:: %d\n", str, idx);
+			pcifunc = ipolicer->pfvf_map[idx];
+			if (!(pcifunc & RVU_PFVF_FUNC_MASK))
+				seq_printf(m, "Allocated to :: PF %d\n",
+					   rvu_get_pf(pcifunc));
+			else
+				seq_printf(m, "Allocated to :: PF %d VF %d\n",
+					   rvu_get_pf(pcifunc),
+					   (pcifunc & RVU_PFVF_FUNC_MASK) - 1);
+			print_band_prof_ctx(m, &aq_rsp.prof);
+		}
+	}
+	return 0;
+}
+
+RVU_DEBUG_SEQ_FOPS(nix_band_prof_ctx, nix_band_prof_ctx_display, NULL);
+
+static int rvu_dbg_nix_band_prof_rsrc_display(struct seq_file *m, void *unused)
+{
+	struct nix_hw *nix_hw = m->private;
+	struct nix_ipolicer *ipolicer;
+	int layer;
+	char *str;
+
+	seq_puts(m, "\nBandwidth profile resource free count\n");
+	seq_puts(m, "=====================================\n");
+	for (layer = 0; layer < BAND_PROF_NUM_LAYERS; layer++) {
+		if (layer == BAND_PROF_INVAL_LAYER)
+			continue;
+		str = (layer == BAND_PROF_LEAF_LAYER) ? "Leaf" :
+			(layer == BAND_PROF_MID_LAYER) ? "Mid " : "Top ";
+
+		ipolicer = &nix_hw->ipolicer[layer];
+		seq_printf(m, "%s :: Max: %4d  Free: %4d\n", str,
+			   ipolicer->band_prof.max,
+			   rvu_rsrc_free_count(&ipolicer->band_prof));
+	}
+	seq_puts(m, "=====================================\n");
+
+	return 0;
+}
+
+RVU_DEBUG_SEQ_FOPS(nix_band_prof_rsrc, nix_band_prof_rsrc_display, NULL);
+
 static void rvu_dbg_nix_init(struct rvu *rvu, int blkaddr)
 {
 	struct nix_hw *nix_hw;
@@ -1664,6 +1823,10 @@ static void rvu_dbg_nix_init(struct rvu *rvu, int blkaddr)
 			    &rvu_dbg_nix_ndc_rx_hits_miss_fops);
 	debugfs_create_file("qsize", 0600, rvu->rvu_dbg.nix, rvu,
 			    &rvu_dbg_nix_qsize_fops);
+	debugfs_create_file("ingress_policer_ctx", 0600, rvu->rvu_dbg.nix, nix_hw,
+			    &rvu_dbg_nix_band_prof_ctx_fops);
+	debugfs_create_file("ingress_policer_rsrc", 0600, rvu->rvu_dbg.nix, nix_hw,
+			    &rvu_dbg_nix_band_prof_rsrc_fops);
 }
 
 static void rvu_dbg_npa_init(struct rvu *rvu)
diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
index d8cb665b7d8a..d6f8210652c5 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
@@ -23,6 +23,14 @@ static int rvu_nix_get_bpid(struct rvu *rvu, struct nix_bp_cfg_req *req,
 			    int type, int chan_id);
 static int nix_update_mce_rule(struct rvu *rvu, u16 pcifunc,
 			       int type, bool add);
+static int nix_setup_ipolicers(struct rvu *rvu,
+			       struct nix_hw *nix_hw, int blkaddr);
+static void nix_ipolicer_freemem(struct nix_hw *nix_hw);
+static int nix_verify_bandprof(struct nix_cn10k_aq_enq_req *req,
+			       struct nix_hw *nix_hw, u16 pcifunc);
+static int nix_free_all_bandprof(struct rvu *rvu, u16 pcifunc);
+static void nix_clear_ratelimit_aggr(struct rvu *rvu, struct nix_hw *nix_hw,
+				     u32 leaf_prof);
 
 enum mc_tbl_sz {
 	MC_TBL_SZ_256,
@@ -699,8 +707,11 @@ static int rvu_nix_blk_aq_enq_inst(struct rvu *rvu, struct nix_hw *nix_hw,
 	pfvf = rvu_get_pfvf(rvu, pcifunc);
 	nixlf = rvu_get_lf(rvu, block, pcifunc, 0);
 
-	/* Skip NIXLF check for broadcast MCE entry init */
-	if (!(!rsp && req->ctype == NIX_AQ_CTYPE_MCE)) {
+	/* Skip NIXLF check for broadcast MCE entry and bandwidth profile
+	 * operations done by AF itself.
+	 */
+	if (!((!rsp && req->ctype == NIX_AQ_CTYPE_MCE) ||
+	      (req->ctype == NIX_AQ_CTYPE_BANDPROF && !pcifunc))) {
 		if (!pfvf->nixlf || nixlf < 0)
 			return NIX_AF_ERR_AF_LF_INVALID;
 	}
@@ -740,6 +751,11 @@ static int rvu_nix_blk_aq_enq_inst(struct rvu *rvu, struct nix_hw *nix_hw,
 		if (rsp)
 			rc = NIX_AF_ERR_AQ_ENQUEUE;
 		break;
+	case NIX_AQ_CTYPE_BANDPROF:
+		if (nix_verify_bandprof((struct nix_cn10k_aq_enq_req *)req,
+					nix_hw, pcifunc))
+			rc = NIX_AF_ERR_INVALID_BANDPROF;
+		break;
 	default:
 		rc = NIX_AF_ERR_AQ_ENQUEUE;
 	}
@@ -796,6 +812,9 @@ static int rvu_nix_blk_aq_enq_inst(struct rvu *rvu, struct nix_hw *nix_hw,
 		else if (req->ctype == NIX_AQ_CTYPE_MCE)
 			memcpy(mask, &req->mce_mask,
 			       sizeof(struct nix_rx_mce_s));
+		else if (req->ctype == NIX_AQ_CTYPE_BANDPROF)
+			memcpy(mask, &req->prof_mask,
+			       sizeof(struct nix_bandprof_s));
 		fallthrough;
 	case NIX_AQ_INSTOP_INIT:
 		if (req->ctype == NIX_AQ_CTYPE_RQ)
@@ -808,6 +827,8 @@ static int rvu_nix_blk_aq_enq_inst(struct rvu *rvu, struct nix_hw *nix_hw,
 			memcpy(ctx, &req->rss, sizeof(struct nix_rsse_s));
 		else if (req->ctype == NIX_AQ_CTYPE_MCE)
 			memcpy(ctx, &req->mce, sizeof(struct nix_rx_mce_s));
+		else if (req->ctype == NIX_AQ_CTYPE_BANDPROF)
+			memcpy(ctx, &req->prof, sizeof(struct nix_bandprof_s));
 		break;
 	case NIX_AQ_INSTOP_NOP:
 	case NIX_AQ_INSTOP_READ:
@@ -885,6 +906,9 @@ static int rvu_nix_blk_aq_enq_inst(struct rvu *rvu, struct nix_hw *nix_hw,
 			else if (req->ctype == NIX_AQ_CTYPE_MCE)
 				memcpy(&rsp->mce, ctx,
 				       sizeof(struct nix_rx_mce_s));
+			else if (req->ctype == NIX_AQ_CTYPE_BANDPROF)
+				memcpy(&rsp->prof, ctx,
+				       sizeof(struct nix_bandprof_s));
 		}
 	}
 
@@ -3624,6 +3648,10 @@ static int rvu_nix_block_init(struct rvu *rvu, struct nix_hw *nix_hw)
 		if (err)
 			return err;
 
+		err = nix_setup_ipolicers(rvu, nix_hw, blkaddr);
+		if (err)
+			return err;
+
 		err = nix_af_mark_format_setup(rvu, nix_hw, blkaddr);
 		if (err)
 			return err;
@@ -3772,6 +3800,8 @@ static void rvu_nix_block_freemem(struct rvu *rvu, int blkaddr,
 			kfree(txsch->schq.bmap);
 		}
 
+		nix_ipolicer_freemem(nix_hw);
+
 		vlan = &nix_hw->txvlan;
 		kfree(vlan->rsrc.bmap);
 		mutex_destroy(&vlan->rsrc_lock);
@@ -3879,6 +3909,8 @@ void rvu_nix_lf_teardown(struct rvu *rvu, u16 pcifunc, int blkaddr, int nixlf)
 	}
 
 	nix_ctx_free(rvu, pfvf);
+
+	nix_free_all_bandprof(rvu, pcifunc);
 }
 
 #define NIX_AF_LFX_TX_CFG_PTP_EN	BIT_ULL(32)
@@ -3987,3 +4019,586 @@ void rvu_nix_reset_mac(struct rvu_pfvf *pfvf, int pcifunc)
 	if (from_vf)
 		ether_addr_copy(pfvf->mac_addr, pfvf->default_mac);
 }
+
+/* NIX ingress policers or bandwidth profiles APIs */
+static void nix_config_rx_pkt_policer_precolor(struct rvu *rvu, int blkaddr)
+{
+	struct npc_lt_def_cfg defs, *ltdefs;
+
+	ltdefs = &defs;
+	memcpy(ltdefs, rvu->kpu.lt_def, sizeof(struct npc_lt_def_cfg));
+
+	/* Extract PCP and DEI fields from outer VLAN from byte offset
+	 * 2 from the start of LB_PTR (ie TAG).
+	 * VLAN0 is Outer VLAN and VLAN1 is Inner VLAN. Inner VLAN
+	 * fields are considered when 'Tunnel enable' is set in profile.
+	 */
+	rvu_write64(rvu, blkaddr, NIX_AF_RX_DEF_VLAN0_PCP_DEI,
+		    (2UL << 12) | (ltdefs->ovlan.lid << 8) |
+		    (ltdefs->ovlan.ltype_match << 4) |
+		    ltdefs->ovlan.ltype_mask);
+	rvu_write64(rvu, blkaddr, NIX_AF_RX_DEF_VLAN1_PCP_DEI,
+		    (2UL << 12) | (ltdefs->ivlan.lid << 8) |
+		    (ltdefs->ivlan.ltype_match << 4) |
+		    ltdefs->ivlan.ltype_mask);
+
+	/* DSCP field in outer and tunneled IPv4 packets */
+	rvu_write64(rvu, blkaddr, NIX_AF_RX_DEF_OIP4_DSCP,
+		    (1UL << 12) | (ltdefs->rx_oip4.lid << 8) |
+		    (ltdefs->rx_oip4.ltype_match << 4) |
+		    ltdefs->rx_oip4.ltype_mask);
+	rvu_write64(rvu, blkaddr, NIX_AF_RX_DEF_IIP4_DSCP,
+		    (1UL << 12) | (ltdefs->rx_iip4.lid << 8) |
+		    (ltdefs->rx_iip4.ltype_match << 4) |
+		    ltdefs->rx_iip4.ltype_mask);
+
+	/* DSCP field (traffic class) in outer and tunneled IPv6 packets */
+	rvu_write64(rvu, blkaddr, NIX_AF_RX_DEF_OIP6_DSCP,
+		    (1UL << 11) | (ltdefs->rx_oip6.lid << 8) |
+		    (ltdefs->rx_oip6.ltype_match << 4) |
+		    ltdefs->rx_oip6.ltype_mask);
+	rvu_write64(rvu, blkaddr, NIX_AF_RX_DEF_IIP6_DSCP,
+		    (1UL << 11) | (ltdefs->rx_iip6.lid << 8) |
+		    (ltdefs->rx_iip6.ltype_match << 4) |
+		    ltdefs->rx_iip6.ltype_mask);
+}
+
+static int nix_init_policer_context(struct rvu *rvu, struct nix_hw *nix_hw,
+				    int layer, int prof_idx)
+{
+	struct nix_cn10k_aq_enq_req aq_req;
+	int rc;
+
+	memset(&aq_req, 0, sizeof(struct nix_cn10k_aq_enq_req));
+
+	aq_req.qidx = (prof_idx & 0x3FFF) | (layer << 14);
+	aq_req.ctype = NIX_AQ_CTYPE_BANDPROF;
+	aq_req.op = NIX_AQ_INSTOP_INIT;
+
+	/* Context is all zeros, submit to AQ */
+	rc = rvu_nix_blk_aq_enq_inst(rvu, nix_hw,
+				     (struct nix_aq_enq_req *)&aq_req, NULL);
+	if (rc)
+		dev_err(rvu->dev, "Failed to INIT bandwidth profile layer %d profile %d\n",
+			layer, prof_idx);
+	return rc;
+}
+
+static int nix_setup_ipolicers(struct rvu *rvu,
+			       struct nix_hw *nix_hw, int blkaddr)
+{
+	struct rvu_hwinfo *hw = rvu->hw;
+	struct nix_ipolicer *ipolicer;
+	int err, layer, prof_idx;
+	u64 cfg;
+
+	cfg = rvu_read64(rvu, blkaddr, NIX_AF_CONST);
+	if (!(cfg & BIT_ULL(61))) {
+		hw->cap.ipolicer = false;
+		return 0;
+	}
+
+	hw->cap.ipolicer = true;
+	nix_hw->ipolicer = devm_kcalloc(rvu->dev, BAND_PROF_NUM_LAYERS,
+					sizeof(*ipolicer), GFP_KERNEL);
+	if (!nix_hw->ipolicer)
+		return -ENOMEM;
+
+	cfg = rvu_read64(rvu, blkaddr, NIX_AF_PL_CONST);
+
+	for (layer = 0; layer < BAND_PROF_NUM_LAYERS; layer++) {
+		ipolicer = &nix_hw->ipolicer[layer];
+		switch (layer) {
+		case BAND_PROF_LEAF_LAYER:
+			ipolicer->band_prof.max = cfg & 0XFFFF;
+			break;
+		case BAND_PROF_MID_LAYER:
+			ipolicer->band_prof.max = (cfg >> 16) & 0XFFFF;
+			break;
+		case BAND_PROF_TOP_LAYER:
+			ipolicer->band_prof.max = (cfg >> 32) & 0XFFFF;
+			break;
+		}
+
+		if (!ipolicer->band_prof.max)
+			continue;
+
+		err = rvu_alloc_bitmap(&ipolicer->band_prof);
+		if (err)
+			return err;
+
+		ipolicer->pfvf_map = devm_kcalloc(rvu->dev,
+						  ipolicer->band_prof.max,
+						  sizeof(u16), GFP_KERNEL);
+		if (!ipolicer->pfvf_map)
+			return -ENOMEM;
+
+		ipolicer->match_id = devm_kcalloc(rvu->dev,
+						  ipolicer->band_prof.max,
+						  sizeof(u16), GFP_KERNEL);
+		if (!ipolicer->match_id)
+			return -ENOMEM;
+
+		for (prof_idx = 0;
+		     prof_idx < ipolicer->band_prof.max; prof_idx++) {
+			/* Set AF as current owner for INIT ops to succeed */
+			ipolicer->pfvf_map[prof_idx] = 0x00;
+
+			/* There is no enable bit in the profile context,
+			 * so no context disable. So let's INIT them here
+			 * so that PF/VF later on have to just do WRITE to
+			 * setup policer rates and config.
+			 */
+			err = nix_init_policer_context(rvu, nix_hw,
+						       layer, prof_idx);
+			if (err)
+				return err;
+		}
+
+		/* Allocate memory for maintaining ref_counts for MID level
+		 * profiles, this will be needed for leaf layer profiles'
+		 * aggregation.
+		 */
+		if (layer != BAND_PROF_MID_LAYER)
+			continue;
+
+		ipolicer->ref_count = devm_kcalloc(rvu->dev,
+						   ipolicer->band_prof.max,
+						   sizeof(u16), GFP_KERNEL);
+	}
+
+	/* Set policer timeunit to 2us ie  (19 + 1) * 100 nsec = 2us */
+	rvu_write64(rvu, blkaddr, NIX_AF_PL_TS, 19);
+
+	nix_config_rx_pkt_policer_precolor(rvu, blkaddr);
+
+	return 0;
+}
+
+static void nix_ipolicer_freemem(struct nix_hw *nix_hw)
+{
+	struct nix_ipolicer *ipolicer;
+	int layer;
+
+	for (layer = 0; layer < BAND_PROF_NUM_LAYERS; layer++) {
+		ipolicer = &nix_hw->ipolicer[layer];
+
+		if (!ipolicer->band_prof.max)
+			continue;
+
+		kfree(ipolicer->band_prof.bmap);
+	}
+}
+
+static int nix_verify_bandprof(struct nix_cn10k_aq_enq_req *req,
+			       struct nix_hw *nix_hw, u16 pcifunc)
+{
+	struct nix_ipolicer *ipolicer;
+	int layer, hi_layer, prof_idx;
+
+	/* Bits [15:14] in profile index represent layer */
+	layer = (req->qidx >> 14) & 0x03;
+	prof_idx = req->qidx & 0x3FFF;
+
+	ipolicer = &nix_hw->ipolicer[layer];
+	if (prof_idx >= ipolicer->band_prof.max)
+		return -EINVAL;
+
+	/* Check if the profile is allocated to the requesting PCIFUNC or not
+	 * with the exception of AF. AF is allowed to read and update contexts.
+	 */
+	if (pcifunc && ipolicer->pfvf_map[prof_idx] != pcifunc)
+		return -EINVAL;
+
+	/* If this profile is linked to higher layer profile then check
+	 * if that profile is also allocated to the requesting PCIFUNC
+	 * or not.
+	 */
+	if (!req->prof.hl_en)
+		return 0;
+
+	/* Leaf layer profile can link only to mid layer and
+	 * mid layer to top layer.
+	 */
+	if (layer == BAND_PROF_LEAF_LAYER)
+		hi_layer = BAND_PROF_MID_LAYER;
+	else if (layer == BAND_PROF_MID_LAYER)
+		hi_layer = BAND_PROF_TOP_LAYER;
+	else
+		return -EINVAL;
+
+	ipolicer = &nix_hw->ipolicer[hi_layer];
+	prof_idx = req->prof.band_prof_id;
+	if (prof_idx >= ipolicer->band_prof.max ||
+	    ipolicer->pfvf_map[prof_idx] != pcifunc)
+		return -EINVAL;
+
+	return 0;
+}
+
+int rvu_mbox_handler_nix_bandprof_alloc(struct rvu *rvu,
+					struct nix_bandprof_alloc_req *req,
+					struct nix_bandprof_alloc_rsp *rsp)
+{
+	int blkaddr, layer, prof, idx, err;
+	u16 pcifunc = req->hdr.pcifunc;
+	struct nix_ipolicer *ipolicer;
+	struct nix_hw *nix_hw;
+
+	if (!rvu->hw->cap.ipolicer)
+		return NIX_AF_ERR_IPOLICER_NOTSUPP;
+
+	err = nix_get_struct_ptrs(rvu, pcifunc, &nix_hw, &blkaddr);
+	if (err)
+		return err;
+
+	mutex_lock(&rvu->rsrc_lock);
+	for (layer = 0; layer < BAND_PROF_NUM_LAYERS; layer++) {
+		if (layer == BAND_PROF_INVAL_LAYER)
+			continue;
+		if (!req->prof_count[layer])
+			continue;
+
+		ipolicer = &nix_hw->ipolicer[layer];
+		for (idx = 0; idx < req->prof_count[layer]; idx++) {
+			/* Allocate a max of 'MAX_BANDPROF_PER_PFFUNC' profiles */
+			if (idx == MAX_BANDPROF_PER_PFFUNC)
+				break;
+
+			prof = rvu_alloc_rsrc(&ipolicer->band_prof);
+			if (prof < 0)
+				break;
+			rsp->prof_count[layer]++;
+			rsp->prof_idx[layer][idx] = prof;
+			ipolicer->pfvf_map[prof] = pcifunc;
+		}
+	}
+	mutex_unlock(&rvu->rsrc_lock);
+	return 0;
+}
+
+static int nix_free_all_bandprof(struct rvu *rvu, u16 pcifunc)
+{
+	int blkaddr, layer, prof_idx, err;
+	struct nix_ipolicer *ipolicer;
+	struct nix_hw *nix_hw;
+
+	if (!rvu->hw->cap.ipolicer)
+		return NIX_AF_ERR_IPOLICER_NOTSUPP;
+
+	err = nix_get_struct_ptrs(rvu, pcifunc, &nix_hw, &blkaddr);
+	if (err)
+		return err;
+
+	mutex_lock(&rvu->rsrc_lock);
+	/* Free all the profiles allocated to the PCIFUNC */
+	for (layer = 0; layer < BAND_PROF_NUM_LAYERS; layer++) {
+		if (layer == BAND_PROF_INVAL_LAYER)
+			continue;
+		ipolicer = &nix_hw->ipolicer[layer];
+
+		for (prof_idx = 0; prof_idx < ipolicer->band_prof.max; prof_idx++) {
+			if (ipolicer->pfvf_map[prof_idx] != pcifunc)
+				continue;
+
+			/* Clear ratelimit aggregation, if any */
+			if (layer == BAND_PROF_LEAF_LAYER &&
+			    ipolicer->match_id[prof_idx])
+				nix_clear_ratelimit_aggr(rvu, nix_hw, prof_idx);
+
+			ipolicer->pfvf_map[prof_idx] = 0x00;
+			ipolicer->match_id[prof_idx] = 0;
+			rvu_free_rsrc(&ipolicer->band_prof, prof_idx);
+		}
+	}
+	mutex_unlock(&rvu->rsrc_lock);
+	return 0;
+}
+
+int rvu_mbox_handler_nix_bandprof_free(struct rvu *rvu,
+				       struct nix_bandprof_free_req *req,
+				       struct msg_rsp *rsp)
+{
+	int blkaddr, layer, prof_idx, idx, err;
+	u16 pcifunc = req->hdr.pcifunc;
+	struct nix_ipolicer *ipolicer;
+	struct nix_hw *nix_hw;
+
+	if (req->free_all)
+		return nix_free_all_bandprof(rvu, pcifunc);
+
+	if (!rvu->hw->cap.ipolicer)
+		return NIX_AF_ERR_IPOLICER_NOTSUPP;
+
+	err = nix_get_struct_ptrs(rvu, pcifunc, &nix_hw, &blkaddr);
+	if (err)
+		return err;
+
+	mutex_lock(&rvu->rsrc_lock);
+	/* Free the requested profile indices */
+	for (layer = 0; layer < BAND_PROF_NUM_LAYERS; layer++) {
+		if (layer == BAND_PROF_INVAL_LAYER)
+			continue;
+		if (!req->prof_count[layer])
+			continue;
+
+		ipolicer = &nix_hw->ipolicer[layer];
+		for (idx = 0; idx < req->prof_count[layer]; idx++) {
+			prof_idx = req->prof_idx[layer][idx];
+			if (prof_idx >= ipolicer->band_prof.max ||
+			    ipolicer->pfvf_map[prof_idx] != pcifunc)
+				continue;
+
+			/* Clear ratelimit aggregation, if any */
+			if (layer == BAND_PROF_LEAF_LAYER &&
+			    ipolicer->match_id[prof_idx])
+				nix_clear_ratelimit_aggr(rvu, nix_hw, prof_idx);
+
+			ipolicer->pfvf_map[prof_idx] = 0x00;
+			ipolicer->match_id[prof_idx] = 0;
+			rvu_free_rsrc(&ipolicer->band_prof, prof_idx);
+			if (idx == MAX_BANDPROF_PER_PFFUNC)
+				break;
+		}
+	}
+	mutex_unlock(&rvu->rsrc_lock);
+	return 0;
+}
+
+int nix_aq_context_read(struct rvu *rvu, struct nix_hw *nix_hw,
+			struct nix_cn10k_aq_enq_req *aq_req,
+			struct nix_cn10k_aq_enq_rsp *aq_rsp,
+			u16 pcifunc, u8 ctype, u32 qidx)
+{
+	memset(aq_req, 0, sizeof(struct nix_cn10k_aq_enq_req));
+	aq_req->hdr.pcifunc = pcifunc;
+	aq_req->ctype = ctype;
+	aq_req->op = NIX_AQ_INSTOP_READ;
+	aq_req->qidx = qidx;
+
+	return rvu_nix_blk_aq_enq_inst(rvu, nix_hw,
+				       (struct nix_aq_enq_req *)aq_req,
+				       (struct nix_aq_enq_rsp *)aq_rsp);
+}
+
+static int nix_ipolicer_map_leaf_midprofs(struct rvu *rvu,
+					  struct nix_hw *nix_hw,
+					  struct nix_cn10k_aq_enq_req *aq_req,
+					  struct nix_cn10k_aq_enq_rsp *aq_rsp,
+					  u32 leaf_prof, u16 mid_prof)
+{
+	memset(aq_req, 0, sizeof(struct nix_cn10k_aq_enq_req));
+	aq_req->hdr.pcifunc = 0x00;
+	aq_req->ctype = NIX_AQ_CTYPE_BANDPROF;
+	aq_req->op = NIX_AQ_INSTOP_WRITE;
+	aq_req->qidx = leaf_prof;
+
+	aq_req->prof.band_prof_id = mid_prof;
+	aq_req->prof_mask.band_prof_id = GENMASK(6, 0);
+	aq_req->prof.hl_en = 1;
+	aq_req->prof_mask.hl_en = 1;
+
+	return rvu_nix_blk_aq_enq_inst(rvu, nix_hw,
+				       (struct nix_aq_enq_req *)aq_req,
+				       (struct nix_aq_enq_rsp *)aq_rsp);
+}
+
+int rvu_nix_setup_ratelimit_aggr(struct rvu *rvu, u16 pcifunc,
+				 u16 rq_idx, u16 match_id)
+{
+	int leaf_prof, mid_prof, leaf_match;
+	struct nix_cn10k_aq_enq_req aq_req;
+	struct nix_cn10k_aq_enq_rsp aq_rsp;
+	struct nix_ipolicer *ipolicer;
+	struct nix_hw *nix_hw;
+	int blkaddr, idx, rc;
+
+	if (!rvu->hw->cap.ipolicer)
+		return 0;
+
+	rc = nix_get_struct_ptrs(rvu, pcifunc, &nix_hw, &blkaddr);
+	if (rc)
+		return rc;
+
+	/* Fetch the RQ's context to see if policing is enabled */
+	rc = nix_aq_context_read(rvu, nix_hw, &aq_req, &aq_rsp, pcifunc,
+				 NIX_AQ_CTYPE_RQ, rq_idx);
+	if (rc) {
+		dev_err(rvu->dev,
+			"%s: Failed to fetch RQ%d context of PFFUNC 0x%x\n",
+			__func__, rq_idx, pcifunc);
+		return rc;
+	}
+
+	if (!aq_rsp.rq.policer_ena)
+		return 0;
+