Merge branch 'support-for-octeontx2-98xx-cpt-block'

Srujana Challa says:

====================
Support for OcteonTX2 98xx CPT block.

OcteonTX2 series of silicons have multiple variants, the
98xx variant has two crypto (CPT) blocks to double the crypto
performance. This patchset adds support for new CPT block(CPT1).
====================

Link: https://lore.kernel.org/r/20210202152709.20450-1-schalla@marvell.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

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

@@ -1073,6 +1073,7 @@ struct cpt_rd_wr_reg_msg {
 	u64 *ret_val;
 	u64 val;
 	u8 is_write;
+	int blkaddr;
 };
 
 struct cpt_lf_alloc_req_msg {
@@ -1080,6 +1081,7 @@ struct cpt_lf_alloc_req_msg {
 	u16 nix_pf_func;
 	u16 sso_pf_func;
 	u16 eng_grpmsk;
+	int blkaddr;
 };
 
 #endif /* MBOX_H */

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

@@ -2150,6 +2150,9 @@ static void rvu_blklf_teardown(struct rvu *rvu, u16 pcifunc, u8 blkaddr)
 			rvu_nix_lf_teardown(rvu, pcifunc, block->addr, lf);
 		else if (block->addr == BLKADDR_NPA)
 			rvu_npa_lf_teardown(rvu, pcifunc, lf);
+		else if ((block->addr == BLKADDR_CPT0) ||
+			 (block->addr == BLKADDR_CPT1))
+			rvu_cpt_lf_teardown(rvu, pcifunc, lf, slot);
 
 		err = rvu_lf_reset(rvu, block, lf);
 		if (err) {

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

@@ -33,6 +33,7 @@
 
 #define NAME_SIZE				32
 #define MAX_NIX_BLKS				2
+#define MAX_CPT_BLKS				2
 
 /* PF_FUNC */
 #define RVU_PFVF_PF_SHIFT	10
@@ -47,6 +48,11 @@ struct dump_ctx {
 	bool	all;
 };
 
+struct cpt_ctx {
+	int blkaddr;
+	struct rvu *rvu;
+};
+
 struct rvu_debugfs {
 	struct dentry *root;
 	struct dentry *cgx_root;
@@ -61,6 +67,7 @@ struct rvu_debugfs {
 	struct dump_ctx nix_cq_ctx;
 	struct dump_ctx nix_rq_ctx;
 	struct dump_ctx nix_sq_ctx;
+	struct cpt_ctx cpt_ctx[MAX_CPT_BLKS];
 	int npa_qsize_id;
 	int nix_qsize_id;
 };
@@ -601,6 +608,8 @@ void npc_enable_mcam_entry(struct rvu *rvu, struct npc_mcam *mcam,
 void npc_read_mcam_entry(struct rvu *rvu, struct npc_mcam *mcam,
 			 int blkaddr, u16 src, struct mcam_entry *entry,
 			 u8 *intf, u8 *ena);
+/* CPT APIs */
+int rvu_cpt_lf_teardown(struct rvu *rvu, u16 pcifunc, int lf, int slot);
 
 #ifdef CONFIG_DEBUG_FS
 void rvu_dbg_init(struct rvu *rvu);

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

@@ -65,13 +65,13 @@ int rvu_mbox_handler_cpt_lf_alloc(struct rvu *rvu,
 	int num_lfs, slot;
 	u64 val;
 
+	blkaddr = req->blkaddr ? req->blkaddr : BLKADDR_CPT0;
+	if (blkaddr != BLKADDR_CPT0 && blkaddr != BLKADDR_CPT1)
+		return -ENODEV;
+
 	if (req->eng_grpmsk == 0x0)
 		return CPT_AF_ERR_GRP_INVALID;
 
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return blkaddr;
-
 	block = &rvu->hw->block[blkaddr];
 	num_lfs = rvu_get_rsrc_mapcount(rvu_get_pfvf(rvu, pcifunc),
 					block->addr);
@@ -114,23 +114,17 @@ int rvu_mbox_handler_cpt_lf_alloc(struct rvu *rvu,
 	return 0;
 }
 
-int rvu_mbox_handler_cpt_lf_free(struct rvu *rvu, struct msg_req *req,
-				 struct msg_rsp *rsp)
+static int cpt_lf_free(struct rvu *rvu, struct msg_req *req, int blkaddr)
 {
 	u16 pcifunc = req->hdr.pcifunc;
+	int num_lfs, cptlf, slot;
 	struct rvu_block *block;
-	int cptlf, blkaddr;
-	int num_lfs, slot;
-
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return blkaddr;
 
 	block = &rvu->hw->block[blkaddr];
 	num_lfs = rvu_get_rsrc_mapcount(rvu_get_pfvf(rvu, pcifunc),
 					block->addr);
 	if (!num_lfs)
-		return CPT_AF_ERR_LF_INVALID;
+		return 0;
 
 	for (slot = 0; slot < num_lfs; slot++) {
 		cptlf = rvu_get_lf(rvu, block, pcifunc, slot);
@@ -146,6 +140,21 @@ int rvu_mbox_handler_cpt_lf_free(struct rvu *rvu, struct msg_req *req,
 	return 0;
 }
 
+int rvu_mbox_handler_cpt_lf_free(struct rvu *rvu, struct msg_req *req,
+				 struct msg_rsp *rsp)
+{
+	int ret;
+
+	ret = cpt_lf_free(rvu, req, BLKADDR_CPT0);
+	if (ret)
+		return ret;
+
+	if (is_block_implemented(rvu->hw, BLKADDR_CPT1))
+		ret = cpt_lf_free(rvu, req, BLKADDR_CPT1);
+
+	return ret;
+}
+
 static bool is_valid_offset(struct rvu *rvu, struct cpt_rd_wr_reg_msg *req)
 {
 	u64 offset = req->reg_offset;
@@ -208,9 +217,9 @@ int rvu_mbox_handler_cpt_rd_wr_register(struct rvu *rvu,
 {
 	int blkaddr;
 
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return blkaddr;
+	blkaddr = req->blkaddr ? req->blkaddr : BLKADDR_CPT0;
+	if (blkaddr != BLKADDR_CPT0 && blkaddr != BLKADDR_CPT1)
+		return -ENODEV;
 
 	/* This message is accepted only if sent from CPT PF/VF */
 	if (!is_cpt_pf(rvu, req->hdr.pcifunc) &&
@@ -231,3 +240,92 @@ int rvu_mbox_handler_cpt_rd_wr_register(struct rvu *rvu,
 
 	return 0;
 }
+
+#define INPROG_INFLIGHT(reg)    ((reg) & 0x1FF)
+#define INPROG_GRB_PARTIAL(reg) ((reg) & BIT_ULL(31))
+#define INPROG_GRB(reg)         (((reg) >> 32) & 0xFF)
+#define INPROG_GWB(reg)         (((reg) >> 40) & 0xFF)
+
+static void cpt_lf_disable_iqueue(struct rvu *rvu, int blkaddr, int slot)
+{
+	int i = 0, hard_lp_ctr = 100000;
+	u64 inprog, grp_ptr;
+	u16 nq_ptr, dq_ptr;
+
+	/* Disable instructions enqueuing */
+	rvu_write64(rvu, blkaddr, CPT_AF_BAR2_ALIASX(slot, CPT_LF_CTL), 0x0);
+
+	/* Disable executions in the LF's queue */
+	inprog = rvu_read64(rvu, blkaddr,
+			    CPT_AF_BAR2_ALIASX(slot, CPT_LF_INPROG));
+	inprog &= ~BIT_ULL(16);
+	rvu_write64(rvu, blkaddr,
+		    CPT_AF_BAR2_ALIASX(slot, CPT_LF_INPROG), inprog);
+
+	/* Wait for CPT queue to become execution-quiescent */
+	do {
+		inprog = rvu_read64(rvu, blkaddr,
+				    CPT_AF_BAR2_ALIASX(slot, CPT_LF_INPROG));
+		if (INPROG_GRB_PARTIAL(inprog)) {
+			i = 0;
+			hard_lp_ctr--;
+		} else {
+			i++;
+		}
+
+		grp_ptr = rvu_read64(rvu, blkaddr,
+				     CPT_AF_BAR2_ALIASX(slot,
+							CPT_LF_Q_GRP_PTR));
+		nq_ptr = (grp_ptr >> 32) & 0x7FFF;
+		dq_ptr = grp_ptr & 0x7FFF;
+
+	} while (hard_lp_ctr && (i < 10) && (nq_ptr != dq_ptr));
+
+	if (hard_lp_ctr == 0)
+		dev_warn(rvu->dev, "CPT FLR hits hard loop counter\n");
+
+	i = 0;
+	hard_lp_ctr = 100000;
+	do {
+		inprog = rvu_read64(rvu, blkaddr,
+				    CPT_AF_BAR2_ALIASX(slot, CPT_LF_INPROG));
+
+		if ((INPROG_INFLIGHT(inprog) == 0) &&
+		    (INPROG_GWB(inprog) < 40) &&
+		    ((INPROG_GRB(inprog) == 0) ||
+		     (INPROG_GRB((inprog)) == 40))) {
+			i++;
+		} else {
+			i = 0;
+			hard_lp_ctr--;
+		}
+	} while (hard_lp_ctr && (i < 10));
+
+	if (hard_lp_ctr == 0)
+		dev_warn(rvu->dev, "CPT FLR hits hard loop counter\n");
+}
+
+int rvu_cpt_lf_teardown(struct rvu *rvu, u16 pcifunc, int lf, int slot)
+{
+	int blkaddr;
+	u64 reg;
+
+	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, pcifunc);
+	if (blkaddr != BLKADDR_CPT0 && blkaddr != BLKADDR_CPT1)
+		return -EINVAL;
+
+	/* Enable BAR2 ALIAS for this pcifunc. */
+	reg = BIT_ULL(16) | pcifunc;
+	rvu_write64(rvu, blkaddr, CPT_AF_BAR2_SEL, reg);
+
+	cpt_lf_disable_iqueue(rvu, blkaddr, slot);
+
+	/* Set group drop to help clear out hardware */
+	reg = rvu_read64(rvu, blkaddr, CPT_AF_BAR2_ALIASX(slot, CPT_LF_INPROG));
+	reg |= BIT_ULL(17);
+	rvu_write64(rvu, blkaddr, CPT_AF_BAR2_ALIASX(slot, CPT_LF_INPROG), reg);
+
+	rvu_write64(rvu, blkaddr, CPT_AF_BAR2_SEL, 0);
+
+	return 0;
+}

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

@@ -1904,20 +1904,16 @@ static void rvu_dbg_npc_init(struct rvu *rvu)
 			    &rvu_dbg_npc_rx_miss_act_fops);
 }
 
-/* CPT debugfs APIs */
 static int cpt_eng_sts_display(struct seq_file *filp, u8 eng_type)
 {
-	struct rvu *rvu = filp->private;
+	struct cpt_ctx *ctx = filp->private;
 	u64 busy_sts = 0, free_sts = 0;
 	u32 e_min = 0, e_max = 0, e, i;
 	u16 max_ses, max_ies, max_aes;
-	int blkaddr;
+	struct rvu *rvu = ctx->rvu;
+	int blkaddr = ctx->blkaddr;
 	u64 reg;
 
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return -ENODEV;
-
 	reg = rvu_read64(rvu, blkaddr, CPT_AF_CONSTANTS1);
 	max_ses = reg & 0xffff;
 	max_ies = (reg >> 16) & 0xffff;
@@ -1977,16 +1973,13 @@ RVU_DEBUG_SEQ_FOPS(cpt_ie_sts, cpt_ie_sts_display, NULL);
 
 static int rvu_dbg_cpt_engines_info_display(struct seq_file *filp, void *unused)
 {
-	struct rvu *rvu = filp->private;
+	struct cpt_ctx *ctx = filp->private;
 	u16 max_ses, max_ies, max_aes;
+	struct rvu *rvu = ctx->rvu;
+	int blkaddr = ctx->blkaddr;
 	u32 e_max, e;
-	int blkaddr;
 	u64 reg;
 
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return -ENODEV;
-
 	reg = rvu_read64(rvu, blkaddr, CPT_AF_CONSTANTS1);
 	max_ses = reg & 0xffff;
 	max_ies = (reg >> 16) & 0xffff;
@@ -2014,17 +2007,15 @@ RVU_DEBUG_SEQ_FOPS(cpt_engines_info, cpt_engines_info_display, NULL);
 
 static int rvu_dbg_cpt_lfs_info_display(struct seq_file *filp, void *unused)
 {
-	struct rvu *rvu = filp->private;
-	struct rvu_hwinfo *hw = rvu->hw;
+	struct cpt_ctx *ctx = filp->private;
+	int blkaddr = ctx->blkaddr;
+	struct rvu *rvu = ctx->rvu;
 	struct rvu_block *block;
-	int blkaddr;
+	struct rvu_hwinfo *hw;
 	u64 reg;
 	u32 lf;
 
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return -ENODEV;
-
+	hw = rvu->hw;
 	block = &hw->block[blkaddr];
 	if (!block->lf.bmap)
 		return -ENODEV;
@@ -2049,13 +2040,10 @@ RVU_DEBUG_SEQ_FOPS(cpt_lfs_info, cpt_lfs_info_display, NULL);
 
 static int rvu_dbg_cpt_err_info_display(struct seq_file *filp, void *unused)
 {
-	struct rvu *rvu = filp->private;
+	struct cpt_ctx *ctx = filp->private;
+	struct rvu *rvu = ctx->rvu;
+	int blkaddr = ctx->blkaddr;
 	u64 reg0, reg1;
-	int blkaddr;
-
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return -ENODEV;
 
 	reg0 = rvu_read64(rvu, blkaddr, CPT_AF_FLTX_INT(0));
 	reg1 = rvu_read64(rvu, blkaddr, CPT_AF_FLTX_INT(1));
@@ -2079,15 +2067,11 @@ RVU_DEBUG_SEQ_FOPS(cpt_err_info, cpt_err_info_display, NULL);
 
 static int rvu_dbg_cpt_pc_display(struct seq_file *filp, void *unused)
 {
-	struct rvu *rvu;
-	int blkaddr;
+	struct cpt_ctx *ctx = filp->private;
+	struct rvu *rvu = ctx->rvu;
+	int blkaddr = ctx->blkaddr;
 	u64 reg;
 
-	rvu = filp->private;
-	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_CPT, 0);
-	if (blkaddr < 0)
-		return -ENODEV;
-
 	reg = rvu_read64(rvu, blkaddr, CPT_AF_INST_REQ_PC);
 	seq_printf(filp, "CPT instruction requests   %llu\n", reg);
 	reg = rvu_read64(rvu, blkaddr, CPT_AF_INST_LATENCY_PC);
@@ -2108,26 +2092,39 @@ static int rvu_dbg_cpt_pc_display(struct seq_file *filp, void *unused)
 
 RVU_DEBUG_SEQ_FOPS(cpt_pc, cpt_pc_display, NULL);
 
-static void rvu_dbg_cpt_init(struct rvu *rvu)
+static void rvu_dbg_cpt_init(struct rvu *rvu, int blkaddr)
 {
-	if (!is_block_implemented(rvu->hw, BLKADDR_CPT0))
+	struct cpt_ctx *ctx;
+
+	if (!is_block_implemented(rvu->hw, blkaddr))
 		return;
 
-	rvu->rvu_dbg.cpt = debugfs_create_dir("cpt", rvu->rvu_dbg.root);
+	if (blkaddr == BLKADDR_CPT0) {
+		rvu->rvu_dbg.cpt = debugfs_create_dir("cpt", rvu->rvu_dbg.root);
+		ctx = &rvu->rvu_dbg.cpt_ctx[0];
+		ctx->blkaddr = BLKADDR_CPT0;
+		ctx->rvu = rvu;
+	} else {
+		rvu->rvu_dbg.cpt = debugfs_create_dir("cpt1",
+						      rvu->rvu_dbg.root);
+		ctx = &rvu->rvu_dbg.cpt_ctx[1];
+		ctx->blkaddr = BLKADDR_CPT1;
+		ctx->rvu = rvu;
+	}
 
-	debugfs_create_file("cpt_pc", 0600, rvu->rvu_dbg.cpt, rvu,
+	debugfs_create_file("cpt_pc", 0600, rvu->rvu_dbg.cpt, ctx,
 			    &rvu_dbg_cpt_pc_fops);
-	debugfs_create_file("cpt_ae_sts", 0600, rvu->rvu_dbg.cpt, rvu,
+	debugfs_create_file("cpt_ae_sts", 0600, rvu->rvu_dbg.cpt, ctx,
 			    &rvu_dbg_cpt_ae_sts_fops);
-	debugfs_create_file("cpt_se_sts", 0600, rvu->rvu_dbg.cpt, rvu,
+	debugfs_create_file("cpt_se_sts", 0600, rvu->rvu_dbg.cpt, ctx,
 			    &rvu_dbg_cpt_se_sts_fops);
-	debugfs_create_file("cpt_ie_sts", 0600, rvu->rvu_dbg.cpt, rvu,
+	debugfs_create_file("cpt_ie_sts", 0600, rvu->rvu_dbg.cpt, ctx,
 			    &rvu_dbg_cpt_ie_sts_fops);
-	debugfs_create_file("cpt_engines_info", 0600, rvu->rvu_dbg.cpt, rvu,
+	debugfs_create_file("cpt_engines_info", 0600, rvu->rvu_dbg.cpt, ctx,
 			    &rvu_dbg_cpt_engines_info_fops);
-	debugfs_create_file("cpt_lfs_info", 0600, rvu->rvu_dbg.cpt, rvu,
+	debugfs_create_file("cpt_lfs_info", 0600, rvu->rvu_dbg.cpt, ctx,
 			    &rvu_dbg_cpt_lfs_info_fops);
-	debugfs_create_file("cpt_err_info", 0600, rvu->rvu_dbg.cpt, rvu,
+	debugfs_create_file("cpt_err_info", 0600, rvu->rvu_dbg.cpt, ctx,
 			    &rvu_dbg_cpt_err_info_fops);
 }
 
@@ -2146,7 +2143,8 @@ void rvu_dbg_init(struct rvu *rvu)
 	rvu_dbg_nix_init(rvu, BLKADDR_NIX1);
 	rvu_dbg_cgx_init(rvu);
 	rvu_dbg_npc_init(rvu);
-	rvu_dbg_cpt_init(rvu);
+	rvu_dbg_cpt_init(rvu, BLKADDR_CPT0);
+	rvu_dbg_cpt_init(rvu, BLKADDR_CPT1);
 }
 
 void rvu_dbg_exit(struct rvu *rvu)

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

@@ -484,9 +484,17 @@
 #define CPT_AF_RAS_INT_ENA_W1S          (0x47030)
 #define CPT_AF_RAS_INT_ENA_W1C          (0x47038)
 
+#define AF_BAR2_ALIASX(a, b)            (0x9100000ull | (a) << 12 | (b))
+#define CPT_AF_BAR2_SEL                 0x9000000
+#define CPT_AF_BAR2_ALIASX(a, b)        AF_BAR2_ALIASX(a, b)
+
 #define CPT_AF_LF_CTL2_SHIFT 3
 #define CPT_AF_LF_SSO_PF_FUNC_SHIFT 32
 
+#define CPT_LF_CTL                      0x10
+#define CPT_LF_INPROG                   0x40
+#define CPT_LF_Q_GRP_PTR                0x120
+
 #define NPC_AF_BLK_RST                  (0x00040)
 
 /* NPC */