Merge branch 'octeontx2-af-cpt-update'

Srujana Challa says:

====================
octeontx2-af: update CPT block for CN10KB and CN10KA B0

This commit addresses two key updates for the CN10KB and CN10KA B0:

1. The number of FLT interrupt vectors has been reduced to 2 on CN10KB.
The code is updated to reflect this change across the CN10K series.
2. The maximum CPT credits that RX can use are now configurable through
a hardware CSR on CN10KA B0. This patch sets the default value to optimize
peak performance, aligning it with other chip versions.

v2:
- Addressed the review comments.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

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

@@ -1856,8 +1856,9 @@ struct cpt_flt_eng_info_req {
 
 struct cpt_flt_eng_info_rsp {
 	struct mbox_msghdr hdr;
-	u64 flt_eng_map[CPT_10K_AF_INT_VEC_RVU];
-	u64 rcvrd_eng_map[CPT_10K_AF_INT_VEC_RVU];
+#define CPT_AF_MAX_FLT_INT_VECS 3
+	u64 flt_eng_map[CPT_AF_MAX_FLT_INT_VECS];
+	u64 rcvrd_eng_map[CPT_AF_MAX_FLT_INT_VECS];
 	u64 rsvd;
 };
 

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

@@ -400,6 +400,7 @@ struct hw_cap {
 	bool	nix_multiple_dwrr_mtu;   /* Multiple DWRR_MTU to choose from */
 	bool	npc_hash_extract; /* Hash extract enabled ? */
 	bool	npc_exact_match_enabled; /* Exact match supported ? */
+	bool    cpt_rxc;   /* Is CPT-RXC supported */
 };
 
 struct rvu_hwinfo {
@@ -690,6 +691,35 @@ static inline bool is_cnf10ka_a0(struct rvu *rvu)
 	return false;
 }
 
+static inline bool is_cn10ka_a0(struct rvu *rvu)
+{
+	struct pci_dev *pdev = rvu->pdev;
+
+	if (pdev->subsystem_device == PCI_SUBSYS_DEVID_CN10K_A &&
+	    (pdev->revision & 0x0F) == 0x0)
+		return true;
+	return false;
+}
+
+static inline bool is_cn10ka_a1(struct rvu *rvu)
+{
+	struct pci_dev *pdev = rvu->pdev;
+
+	if (pdev->subsystem_device == PCI_SUBSYS_DEVID_CN10K_A &&
+	    (pdev->revision & 0x0F) == 0x1)
+		return true;
+	return false;
+}
+
+static inline bool is_cn10kb(struct rvu *rvu)
+{
+	struct pci_dev *pdev = rvu->pdev;
+
+	if (pdev->subsystem_device == PCI_SUBSYS_DEVID_CN10K_B)
+		return true;
+	return false;
+}
+
 static inline bool is_rvu_npc_hash_extract_en(struct rvu *rvu)
 {
 	u64 npc_const3;

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

@@ -19,6 +19,12 @@
 /* Length of initial context fetch in 128 byte words */
 #define CPT_CTX_ILEN    1ULL
 
+/* Interrupt vector count of CPT RVU and RAS interrupts */
+#define CPT_10K_AF_RVU_RAS_INT_VEC_CNT  2
+
+/* Default CPT_AF_RXC_CFG1:max_rxc_icb_cnt */
+#define CPT_DFLT_MAX_RXC_ICB_CNT  0xC0ULL
+
 #define cpt_get_eng_sts(e_min, e_max, rsp, etype)                   \
 ({                                                                  \
 	u64 free_sts = 0, busy_sts = 0;                             \
@@ -37,6 +43,41 @@
 	(_rsp)->free_sts_##etype = free_sts;                        \
 })
 
+#define MAX_AE  GENMASK_ULL(47, 32)
+#define MAX_IE  GENMASK_ULL(31, 16)
+#define MAX_SE  GENMASK_ULL(15, 0)
+
+static u16 cpt_max_engines_get(struct rvu *rvu)
+{
+	u16 max_ses, max_ies, max_aes;
+	u64 reg;
+
+	reg = rvu_read64(rvu, BLKADDR_CPT0, CPT_AF_CONSTANTS1);
+	max_ses = FIELD_GET(MAX_SE, reg);
+	max_ies = FIELD_GET(MAX_IE, reg);
+	max_aes = FIELD_GET(MAX_AE, reg);
+
+	return max_ses + max_ies + max_aes;
+}
+
+/* Number of flt interrupt vectors are depends on number of engines that the
+ * chip has. Each flt vector represents 64 engines.
+ */
+static int cpt_10k_flt_nvecs_get(struct rvu *rvu, u16 max_engs)
+{
+	int flt_vecs;
+
+	flt_vecs = DIV_ROUND_UP(max_engs, 64);
+
+	if (flt_vecs > CPT_10K_AF_INT_VEC_FLT_MAX) {
+		dev_warn_once(rvu->dev, "flt_vecs:%d exceeds the max vectors:%d\n",
+			      flt_vecs, CPT_10K_AF_INT_VEC_FLT_MAX);
+		flt_vecs = CPT_10K_AF_INT_VEC_FLT_MAX;
+	}
+
+	return flt_vecs;
+}
+
 static irqreturn_t cpt_af_flt_intr_handler(int vec, void *ptr)
 {
 	struct rvu_block *block = ptr;
@@ -150,17 +191,26 @@ static void cpt_10k_unregister_interrupts(struct rvu_block *block, int off)
 {
 	struct rvu *rvu = block->rvu;
 	int blkaddr = block->addr;
-	int i;
+	int i, flt_vecs;
+	u16 max_engs;
+	u8 nr;
+
+	max_engs = cpt_max_engines_get(rvu);
+	flt_vecs = cpt_10k_flt_nvecs_get(rvu, max_engs);
 
 	/* Disable all CPT AF interrupts */
-	rvu_write64(rvu, blkaddr, CPT_AF_FLTX_INT_ENA_W1C(0), ~0ULL);
-	rvu_write64(rvu, blkaddr, CPT_AF_FLTX_INT_ENA_W1C(1), ~0ULL);
-	rvu_write64(rvu, blkaddr, CPT_AF_FLTX_INT_ENA_W1C(2), 0xFFFF);
+	for (i = CPT_10K_AF_INT_VEC_FLT0; i < flt_vecs; i++) {
+		nr = (max_engs > 64) ? 64 : max_engs;
+		max_engs -= nr;
+		rvu_write64(rvu, blkaddr, CPT_AF_FLTX_INT_ENA_W1C(i),
+			    INTR_MASK(nr));
+	}
 
 	rvu_write64(rvu, blkaddr, CPT_AF_RVU_INT_ENA_W1C, 0x1);
 	rvu_write64(rvu, blkaddr, CPT_AF_RAS_INT_ENA_W1C, 0x1);
 
-	for (i = 0; i < CPT_10K_AF_INT_VEC_CNT; i++)
+	/* CPT AF interrupt vectors are flt_int, rvu_int and ras_int. */
+	for (i = 0; i < flt_vecs + CPT_10K_AF_RVU_RAS_INT_VEC_CNT; i++)
 		if (rvu->irq_allocated[off + i]) {
 			free_irq(pci_irq_vector(rvu->pdev, off + i), block);
 			rvu->irq_allocated[off + i] = false;
@@ -206,12 +256,18 @@ void rvu_cpt_unregister_interrupts(struct rvu *rvu)
 
 static int cpt_10k_register_interrupts(struct rvu_block *block, int off)
 {
+	int rvu_intr_vec, ras_intr_vec;
 	struct rvu *rvu = block->rvu;
 	int blkaddr = block->addr;
 	irq_handler_t flt_fn;
-	int i, ret;
+	int i, ret, flt_vecs;
+	u16 max_engs;
+	u8 nr;
+
+	max_engs = cpt_max_engines_get(rvu);
+	flt_vecs = cpt_10k_flt_nvecs_get(rvu, max_engs);
 
-	for (i = CPT_10K_AF_INT_VEC_FLT0; i < CPT_10K_AF_INT_VEC_RVU; i++) {
+	for (i = CPT_10K_AF_INT_VEC_FLT0; i < flt_vecs; i++) {
 		sprintf(&rvu->irq_name[(off + i) * NAME_SIZE], "CPTAF FLT%d", i);
 
 		switch (i) {
@@ -229,20 +285,24 @@ static int cpt_10k_register_interrupts(struct rvu_block *block, int off)
 						    flt_fn, &rvu->irq_name[(off + i) * NAME_SIZE]);
 		if (ret)
 			goto err;
-		if (i == CPT_10K_AF_INT_VEC_FLT2)
-			rvu_write64(rvu, blkaddr, CPT_AF_FLTX_INT_ENA_W1S(i), 0xFFFF);
-		else
-			rvu_write64(rvu, blkaddr, CPT_AF_FLTX_INT_ENA_W1S(i), ~0ULL);
+
+		nr = (max_engs > 64) ? 64 : max_engs;
+		max_engs -= nr;
+		rvu_write64(rvu, blkaddr, CPT_AF_FLTX_INT_ENA_W1S(i),
+			    INTR_MASK(nr));
 	}
 
-	ret = rvu_cpt_do_register_interrupt(block, off + CPT_10K_AF_INT_VEC_RVU,
+	rvu_intr_vec = flt_vecs;
+	ras_intr_vec = rvu_intr_vec + 1;
+
+	ret = rvu_cpt_do_register_interrupt(block, off + rvu_intr_vec,
 					    rvu_cpt_af_rvu_intr_handler,
 					    "CPTAF RVU");
 	if (ret)
 		goto err;
 	rvu_write64(rvu, blkaddr, CPT_AF_RVU_INT_ENA_W1S, 0x1);
 
-	ret = rvu_cpt_do_register_interrupt(block, off + CPT_10K_AF_INT_VEC_RAS,
+	ret = rvu_cpt_do_register_interrupt(block, off + ras_intr_vec,
 					    rvu_cpt_af_ras_intr_handler,
 					    "CPTAF RAS");
 	if (ret)
@@ -680,6 +740,7 @@ static bool validate_and_update_reg_offset(struct rvu *rvu,
 		case CPT_AF_BLK_RST:
 		case CPT_AF_CONSTANTS1:
 		case CPT_AF_CTX_FLUSH_TIMER:
+		case CPT_AF_RXC_CFG1:
 			return true;
 		}
 
@@ -732,6 +793,8 @@ int rvu_mbox_handler_cpt_rd_wr_register(struct rvu *rvu,
 
 static void get_ctx_pc(struct rvu *rvu, struct cpt_sts_rsp *rsp, int blkaddr)
 {
+	struct rvu_hwinfo *hw = rvu->hw;
+
 	if (is_rvu_otx2(rvu))
 		return;
 
@@ -755,14 +818,16 @@ static void get_ctx_pc(struct rvu *rvu, struct cpt_sts_rsp *rsp, int blkaddr)
 	rsp->ctx_err = rvu_read64(rvu, blkaddr, CPT_AF_CTX_ERR);
 	rsp->ctx_enc_id = rvu_read64(rvu, blkaddr, CPT_AF_CTX_ENC_ID);
 	rsp->ctx_flush_timer = rvu_read64(rvu, blkaddr, CPT_AF_CTX_FLUSH_TIMER);
+	rsp->x2p_link_cfg0 = rvu_read64(rvu, blkaddr, CPT_AF_X2PX_LINK_CFG(0));
+	rsp->x2p_link_cfg1 = rvu_read64(rvu, blkaddr, CPT_AF_X2PX_LINK_CFG(1));
 
+	if (!hw->cap.cpt_rxc)
+		return;
 	rsp->rxc_time = rvu_read64(rvu, blkaddr, CPT_AF_RXC_TIME);
 	rsp->rxc_time_cfg = rvu_read64(rvu, blkaddr, CPT_AF_RXC_TIME_CFG);
 	rsp->rxc_active_sts = rvu_read64(rvu, blkaddr, CPT_AF_RXC_ACTIVE_STS);
 	rsp->rxc_zombie_sts = rvu_read64(rvu, blkaddr, CPT_AF_RXC_ZOMBIE_STS);
 	rsp->rxc_dfrg = rvu_read64(rvu, blkaddr, CPT_AF_RXC_DFRG);
-	rsp->x2p_link_cfg0 = rvu_read64(rvu, blkaddr, CPT_AF_X2PX_LINK_CFG(0));
-	rsp->x2p_link_cfg1 = rvu_read64(rvu, blkaddr, CPT_AF_X2PX_LINK_CFG(1));
 }
 
 static void get_eng_sts(struct rvu *rvu, struct cpt_sts_rsp *rsp, int blkaddr)
@@ -921,13 +986,17 @@ int rvu_mbox_handler_cpt_flt_eng_info(struct rvu *rvu, struct cpt_flt_eng_info_r
 	struct rvu_block *block;
 	unsigned long flags;
 	int blkaddr, vec;
+	int flt_vecs;
+	u16 max_engs;
 
 	blkaddr = validate_and_get_cpt_blkaddr(req->blkaddr);
 	if (blkaddr < 0)
 		return blkaddr;
 
 	block = &rvu->hw->block[blkaddr];
-	for (vec = 0; vec < CPT_10K_AF_INT_VEC_RVU; vec++) {
+	max_engs = cpt_max_engines_get(rvu);
+	flt_vecs = cpt_10k_flt_nvecs_get(rvu, max_engs);
+	for (vec = 0; vec < flt_vecs; vec++) {
 		spin_lock_irqsave(&rvu->cpt_intr_lock, flags);
 		rsp->flt_eng_map[vec] = block->cpt_flt_eng_map[vec];
 		rsp->rcvrd_eng_map[vec] = block->cpt_rcvrd_eng_map[vec];
@@ -943,10 +1012,11 @@ int rvu_mbox_handler_cpt_flt_eng_info(struct rvu *rvu, struct cpt_flt_eng_info_r
 static void cpt_rxc_teardown(struct rvu *rvu, int blkaddr)
 {
 	struct cpt_rxc_time_cfg_req req, prev;
+	struct rvu_hwinfo *hw = rvu->hw;
 	int timeout = 2000;
 	u64 reg;
 
-	if (is_rvu_otx2(rvu))
+	if (!hw->cap.cpt_rxc)
 		return;
 
 	/* Set time limit to minimum values, so that rxc entries will be
@@ -1219,10 +1289,30 @@ int rvu_cpt_ctx_flush(struct rvu *rvu, u16 pcifunc)
 	return 0;
 }
 
+#define MAX_RXC_ICB_CNT  GENMASK_ULL(40, 32)
+
 int rvu_cpt_init(struct rvu *rvu)
 {
+	struct rvu_hwinfo *hw = rvu->hw;
+	u64 reg_val;
+
 	/* Retrieve CPT PF number */
 	rvu->cpt_pf_num = get_cpt_pf_num(rvu);
+	if (is_block_implemented(rvu->hw, BLKADDR_CPT0) && !is_rvu_otx2(rvu) &&
+	    !is_cn10kb(rvu))
+		hw->cap.cpt_rxc = true;
+
+	if (hw->cap.cpt_rxc && !is_cn10ka_a0(rvu) && !is_cn10ka_a1(rvu)) {
+		/* Set CPT_AF_RXC_CFG1:max_rxc_icb_cnt to 0xc0 to not effect
+		 * inline inbound peak performance
+		 */
+		reg_val = rvu_read64(rvu, BLKADDR_CPT0, CPT_AF_RXC_CFG1);
+		reg_val &= ~MAX_RXC_ICB_CNT;
+		reg_val |= FIELD_PREP(MAX_RXC_ICB_CNT,
+				      CPT_DFLT_MAX_RXC_ICB_CNT);
+		rvu_write64(rvu, BLKADDR_CPT0, CPT_AF_RXC_CFG1, reg_val);
+	}
+
 	spin_lock_init(&rvu->cpt_intr_lock);
 
 	return 0;

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

@@ -545,6 +545,7 @@
 #define CPT_AF_CTX_PSH_PC               (0x49450ull)
 #define CPT_AF_CTX_PSH_LATENCY_PC       (0x49458ull)
 #define CPT_AF_CTX_CAM_DATA(a)          (0x49800ull | (u64)(a) << 3)
+#define CPT_AF_RXC_CFG1                 (0x50000ull)
 #define CPT_AF_RXC_TIME                 (0x50010ull)
 #define CPT_AF_RXC_TIME_CFG             (0x50018ull)
 #define CPT_AF_RXC_DFRG                 (0x50020ull)

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

@@ -71,13 +71,11 @@ enum cpt_af_int_vec_e {
 	CPT_AF_INT_VEC_CNT	= 0x4,
 };
 
-enum cpt_10k_af_int_vec_e {
+enum cpt_cn10k_flt_int_vec_e {
 	CPT_10K_AF_INT_VEC_FLT0	= 0x0,
 	CPT_10K_AF_INT_VEC_FLT1	= 0x1,
 	CPT_10K_AF_INT_VEC_FLT2	= 0x2,
-	CPT_10K_AF_INT_VEC_RVU	= 0x3,
-	CPT_10K_AF_INT_VEC_RAS	= 0x4,
-	CPT_10K_AF_INT_VEC_CNT	= 0x5,
+	CPT_10K_AF_INT_VEC_FLT_MAX = 0x3,
 };
 
 /* NPA Admin function Interrupt Vector Enumeration */