[dpdk-dev] [PATCH 06/16] crypto/cpt/base: add sym crypto request prepare for CPT

[Anoob Joseph <anoob.joseph@caviumnetworks.com>]

From: Ankur Dwivedi <ankur.dwivedi@cavium.com>

These functions help in preparing symmetric crypto requests
for the supported cipher/auth/aead. This includes all supported
algos except Kasumi, Snow3G, Zuc, HMAC_ONLY and HASH_ONLY cases.

Signed-off-by: Ankur Dwivedi <ankur.dwivedi@cavium.com>
Signed-off-by: Murthy NSSR <Nidadavolu.Murthy@cavium.com>
Signed-off-by: Nithin Dabilpuram <nithin.dabilpuram@cavium.com>
Signed-off-by: Ragothaman Jayaraman <Ragothaman.Jayaraman@cavium.com>
Signed-off-by: Srisivasubramanian Srinivasan <Srisivasubramanian.Srinivasan@cavium.com>
---
 drivers/crypto/cpt/base/cpt.h     |  129 +++++
 drivers/crypto/cpt/base/cpt_ops.c | 1021 +++++++++++++++++++++++++++++++++++++
 2 files changed, 1150 insertions(+)

diff --git a/drivers/crypto/cpt/base/cpt.h b/drivers/crypto/cpt/base/cpt.h
index 11407ae..54b1cb6 100644
--- a/drivers/crypto/cpt/base/cpt.h
+++ b/drivers/crypto/cpt/base/cpt.h
@@ -54,6 +54,135 @@
 	void *marker;
 } app_data_t;
 
+/*
+ * Parameters for Flexi Crypto
+ * requests
+ */
+#define VALID_AAD_BUF 0x01
+#define VALID_MAC_BUF 0x02
+#define VALID_IV_BUF 0x04
+#define SINGLE_BUF_INPLACE 0x08
+#define SINGLE_BUF_HEADTAILROOM 0x10
+
+#define ENCR_IV_OFFSET(__d_offs) ((__d_offs >> 32) & 0xffff)
+#define ENCR_OFFSET(__d_offs) ((__d_offs >> 16) & 0xffff)
+#define AUTH_OFFSET(__d_offs) (__d_offs & 0xffff)
+#define ENCR_DLEN(__d_lens) (__d_lens >> 32)
+#define AUTH_DLEN(__d_lens) (__d_lens & 0xffffffff)
+
+typedef struct fc_params {
+	/* 0th cache line */
+	union {
+		buf_ptr_t bufs[1];
+		struct {
+			iov_ptr_t *src_iov;
+			iov_ptr_t *dst_iov;
+		};
+	};
+	void *iv_buf;
+	void *auth_iv_buf;
+	buf_ptr_t meta_buf;
+	buf_ptr_t ctx_buf;
+	uint64_t rsvd2;
+
+	/* 1st cache line */
+	buf_ptr_t aad_buf;
+	buf_ptr_t mac_buf;
+
+} fc_params_t;
+
+/*
+ * Parameters for digest
+ * generate requests
+ * Only src_iov, op, ctx_buf, mac_buf, prep_req
+ * meta_buf, auth_data_len are used for digest gen.
+ */
+typedef struct fc_params digest_params_t;
+
+/* Cipher Algorithms */
+typedef mc_cipher_type_t cipher_type_t;
+
+/* Auth Algorithms */
+typedef mc_hash_type_t auth_type_t;
+
+/* Flexi Crypto Operations */
+/*
+ * Encr | Encr + Hmac | HASH-HMAC generation
+ */
+/*
+ * ZUC/SNOW3g enc cipher/cipher+auth/auth-gen operation
+ */
+/*
+ * kasumi enc cipher/cipher+auth/auth-gen operation
+ * F8 iv_buf: 64 bits Bigendian format
+ * COUNT[63-32] || BEARER[31-27] ||
+ * DIRECTION[26] || 0...0[25-0]
+ * F9 mac gen auth_iv_buf: 64 bits BE + 8 bits
+ * COUNT[63-32] || FRESH[31-0]
+ * 0...0[7-1] || DIRECTION[0]
+ */
+void *cpt_fc_enc_hmac_prep(uint32_t flags,
+			   uint64_t d_offs,
+			   uint64_t d_lens,
+			   fc_params_t *params,
+			   void *op,
+			   int *ret);
+/*
+ * Decr | Decr + Hmac
+ */
+/*
+ * ZUC/SNOW3g dec cipher/cipher+auth operation
+ */
+/*
+ * kasumi dec cipher/cipher+auth/ operation
+ * F8 iv_buf: 64 bits Bigendian format
+ * COUNT[63-32] || BEARER[31-27] ||
+ * DIRECTION[26] || 0...0[25-0]
+ */
+void *cpt_fc_dec_hmac_prep(uint32_t flags,
+			   uint64_t d_offs,
+			   uint64_t d_lens,
+			   fc_params_t *params,
+			   void *op,
+			   int *ret);
+
+/* Flexi Crypto Ctrl Operations */
+int32_t cpt_fc_ciph_set_key(cpt_instance_t *inst,
+			    void *ctx,
+			    cipher_type_t type,
+			    uint8_t *key,
+			    uint16_t key_len,
+			    uint8_t *salt);
+
+int32_t cpt_fc_ciph_set_iv(cpt_instance_t *inst,
+			   void *ctx,
+			   uint8_t *iv,
+			   uint16_t iv_len);
+
+int32_t cpt_fc_auth_set_key(cpt_instance_t *inst,
+			    void *ctx,
+			    auth_type_t type,
+			    uint8_t *key,
+			    uint16_t key_len,
+			    uint16_t mac_len);
+
+void
+cpt_fc_salt_update(void *ctx,
+		   uint8_t *salt);
+/*
+ * Get's size of contiguous meta buffer
+ * to be allocated per op
+ */
+int32_t  cpt_fc_get_op_meta_len(void);
+
+/* Get context length for a session */
+int32_t cpt_fc_get_ctx_len(void);
+
+/* Provides meta length required when it is
+ * direct mode i.e single buf inplace
+ */
+int32_t cpt_fc_get_op_sb_meta_len(void);
+
 /* Instance operations */
 
 /* Enqueue an SE/AE request */
diff --git a/drivers/crypto/cpt/base/cpt_ops.c b/drivers/crypto/cpt/base/cpt_ops.c
index e340006..31f8064 100644
--- a/drivers/crypto/cpt/base/cpt_ops.c
+++ b/drivers/crypto/cpt/base/cpt_ops.c
@@ -86,6 +86,14 @@ int32_t cpt_fc_get_ctx_len(void)
 	return sizeof(struct cpt_ctx);
 }
 
+inline void
+cpt_fc_salt_update(void *ctx,
+		   uint8_t *salt)
+{
+	struct cpt_ctx *cpt_ctx = ctx;
+	memcpy(&cpt_ctx->fctx.enc.encr_iv, salt, 4);
+}
+
 int
 cpt_fc_ciph_set_key(cpt_instance_t *instance,
 		    void *ctx, cipher_type_t type, uint8_t *key,
@@ -306,3 +314,1016 @@ int32_t cpt_fc_get_ctx_len(void)
 	*ctrl_flags = htobe64(*ctrl_flags);
 	return 0;
 }
+
+static inline uint32_t
+fill_sg_comp(sg_comp_t *list,
+	     uint32_t i,
+	     phys_addr_t dma_addr,
+	     void *vaddr,
+	     uint32_t size)
+{
+	sg_comp_t *to = &list[i>>2];
+
+	to->u.s.len[i%4] = htobe16(size);
+	to->ptr[i%4] = htobe64(dma_addr);
+	(void) vaddr;
+	i++;
+	return i;
+}
+
+static inline uint32_t
+fill_sg_comp_from_buf(sg_comp_t *list,
+		      uint32_t i,
+		      buf_ptr_t *from)
+{
+	sg_comp_t *to = &list[i>>2];
+
+	to->u.s.len[i%4] = htobe16(from->size);
+	to->ptr[i%4] = htobe64(from->dma_addr);
+	i++;
+	return i;
+}
+
+static inline uint32_t
+fill_sg_comp_from_buf_min(sg_comp_t *list,
+			  uint32_t i,
+			  buf_ptr_t *from,
+			  uint32_t *psize)
+{
+	sg_comp_t *to = &list[i >> 2];
+	uint32_t size = *psize;
+	uint32_t e_len;
+
+	e_len = (size > from->size) ? from->size : size;
+	to->u.s.len[i % 4] = htobe16(e_len);
+	to->ptr[i % 4] = htobe64(from->dma_addr);
+	*psize -= e_len;
+	i++;
+	return i;
+}
+
+/*
+ * This fills the MC expected SGIO list
+ * from IOV given by user.
+ */
+static inline uint32_t
+fill_sg_comp_from_iov(sg_comp_t *list,
+		      uint32_t i,
+		      iov_ptr_t *from, uint32_t from_offset,
+		      uint32_t *psize, buf_ptr_t *extra_buf,
+		      uint32_t extra_offset)
+{
+	int32_t j;
+	uint32_t extra_len = extra_buf ? extra_buf->size : 0;
+	uint32_t size = *psize - extra_len;
+	buf_ptr_t *bufs;
+
+	bufs = from->bufs;
+	for (j = 0; (j < from->buf_cnt) && size; j++) {
+		phys_addr_t e_dma_addr;
+		uint32_t e_len;
+		sg_comp_t *to = &list[i >> 2];
+
+		if (!bufs[j].size)
+			continue;
+
+		if (unlikely(from_offset)) {
+			if (from_offset >= bufs[j].size) {
+				from_offset -= bufs[j].size;
+				continue;
+			}
+			e_dma_addr = bufs[j].dma_addr + from_offset;
+			e_len = (size > (bufs[j].size - from_offset)) ?
+				(bufs[j].size - from_offset) : size;
+			from_offset = 0;
+		} else {
+			e_dma_addr = bufs[j].dma_addr;
+			e_len = (size > bufs[j].size) ?
+				bufs[j].size : size;
+		}
+
+		to->u.s.len[i % 4] = htobe16(e_len);
+		to->ptr[i % 4] = htobe64(e_dma_addr);
+
+		if (extra_len && (e_len >= extra_offset)) {
+			/* Break the data at given offset */
+			uint32_t next_len = e_len - extra_offset;
+			phys_addr_t next_dma = e_dma_addr + extra_offset;
+
+			if (!extra_offset) {
+				i--;
+			} else {
+				e_len = extra_offset;
+				size -= e_len;
+				to->u.s.len[i % 4] = htobe16(e_len);
+			}
+
+			/* Insert extra data ptr */
+			if (extra_len) {
+				i++;
+				to = &list[i >> 2];
+				to->u.s.len[i % 4] = htobe16(extra_buf->size);
+				to->ptr[i % 4] = htobe64(extra_buf->dma_addr);
+
+				/* size already decremented by extra len */
+			}
+
+			/* insert the rest of the data */
+			if (next_len) {
+				i++;
+				to = &list[i >> 2];
+				to->u.s.len[i % 4] = htobe16(next_len);
+				to->ptr[i % 4] = htobe64(next_dma);
+				size -= next_len;
+			}
+			extra_len = 0;
+
+		} else {
+			size -= e_len;
+		}
+		if (extra_offset)
+			extra_offset -= size;
+		i++;
+	}
+
+	*psize = size;
+	return (uint32_t)i;
+}
+
+static inline int  __attribute__((always_inline))
+cpt_enc_hmac_prep(uint32_t flags,
+		  uint64_t d_offs,
+		  uint64_t d_lens,
+		  fc_params_t *fc_params,
+		  void *op,
+		  void **prep_req)
+{
+	uint32_t iv_offset = 0;
+	int32_t inputlen, outputlen, enc_dlen, auth_dlen;
+	struct cpt_ctx *cpt_ctx;
+	uint32_t cipher_type, hash_type;
+	uint32_t mac_len, size;
+	uint8_t iv_len = 16;
+	cpt_request_info_t *req;
+	buf_ptr_t *meta_p, *aad_buf = NULL;
+	uint32_t encr_offset, auth_offset;
+	uint32_t encr_data_len, auth_data_len, aad_len = 0;
+	uint32_t passthrough_len = 0;
+	void *m_vaddr, *offset_vaddr;
+	uint64_t m_dma, offset_dma, ctx_dma;
+	vq_cmd_word0_t vq_cmd_w0;
+	vq_cmd_word3_t vq_cmd_w3;
+	void *c_vaddr;
+	uint64_t c_dma;
+	int32_t m_size;
+	opcode_info_t opcode;
+
+	meta_p = &fc_params->meta_buf;
+#ifdef CPTVF_STRICT_PARAM_CHECK
+	if (!fc_params || !meta_p->vaddr || !meta_p->size)
+		return ERR_BAD_INPUT_ARG;
+#endif
+	m_vaddr = meta_p->vaddr;
+	m_dma = meta_p->dma_addr;
+	m_size = meta_p->size;
+
+	encr_offset = ENCR_OFFSET(d_offs);
+	auth_offset = AUTH_OFFSET(d_offs);
+	encr_data_len = ENCR_DLEN(d_lens);
+	auth_data_len = AUTH_DLEN(d_lens);
+	if (unlikely(flags & VALID_AAD_BUF)) {
+		/*
+		 * We dont support both aad
+		 * and auth data separately
+		 */
+		auth_data_len = 0;
+		auth_offset = 0;
+		aad_len = fc_params->aad_buf.size;
+		aad_buf = &fc_params->aad_buf;
+	}
+	cpt_ctx = fc_params->ctx_buf.vaddr;
+	cipher_type = cpt_ctx->enc_cipher;
+	hash_type = cpt_ctx->hash_type;
+	mac_len = cpt_ctx->mac_len;
+
+	/*
+	 * Save initial space that followed app data for completion code &
+	 * alternate completion code to fall in same cache line as app data
+	 */
+	m_vaddr = (uint8_t *)m_vaddr + COMPLETION_CODE_SIZE;
+	m_dma += COMPLETION_CODE_SIZE;
+	size = (uint8_t *)RTE_PTR_ALIGN((uint8_t *)m_vaddr, 16) -
+		(uint8_t *)m_vaddr;
+
+	c_vaddr = (uint8_t *)m_vaddr + size;
+	c_dma = m_dma + size;
+	size += sizeof(cpt_res_s_t);
+
+	m_vaddr = (uint8_t *)m_vaddr + size;
+	m_dma += size;
+	m_size -= size;
+
+	/* start cpt request info struct at 8 byte boundary */
+	size = (uint8_t *)RTE_PTR_ALIGN(m_vaddr, 8) -
+		(uint8_t *)m_vaddr;
+
+	req = (cpt_request_info_t *)((uint8_t *)m_vaddr + size);
+
+	size += sizeof(cpt_request_info_t);
+	m_vaddr = (uint8_t *)m_vaddr + size;
+	m_dma += size;
+	m_size -= size;
+
+	if (hash_type == GMAC_TYPE)
+		encr_data_len = 0;
+
+	if (unlikely(!(flags & VALID_IV_BUF))) {
+		iv_len = 0;
+		iv_offset = ENCR_IV_OFFSET(d_offs);
+	}
+
+	if (unlikely(flags & VALID_AAD_BUF)) {
+		/*
+		 * When AAD is given, data above encr_offset is pass through
+		 * Since AAD is given as separate pointer and not as offset,
+		 * this is a special case as we need to fragment input data
+		 * into passthrough + encr_data and then insert AAD in between.
+		 */
+		if (hash_type != GMAC_TYPE) {
+			passthrough_len = encr_offset;
+			auth_offset = passthrough_len + iv_len;
+			encr_offset = passthrough_len + aad_len + iv_len;
+			auth_data_len = aad_len + encr_data_len;
+		} else {
+			passthrough_len = 16 + aad_len;
+			auth_offset = passthrough_len + iv_len;
+			auth_data_len = aad_len;
+		}
+	} else {
+		encr_offset += iv_len;
+		auth_offset += iv_len;
+	}
+
+	/* Initialising ctrl and opcode
+	 * fields in cpt request structure
+	 */
+
+	req->se_req = SE_CORE_REQ;
+	/*
+	 * We are using DMA mode but indicate that
+	 * SGIO list is already populated.
+	 */
+	req->dma_mode = CTRL_DMA_MODE_SGIO;
+
+	/* Encryption */
+	opcode.s.major = MAJOR_OP_FC;
+	opcode.s.minor = 0;
+
+	auth_dlen = auth_offset + auth_data_len;
+	enc_dlen = encr_data_len + encr_offset;
+	if (unlikely(encr_data_len & 0xf)) {
+		if ((cipher_type == DES3_CBC) || (cipher_type == DES3_ECB))
+			enc_dlen = ROUNDUP8(encr_data_len) + encr_offset;
+		else if (likely((cipher_type == AES_CBC) ||
+				(cipher_type == AES_ECB)))
+			enc_dlen = ROUNDUP16(encr_data_len) + encr_offset;
+	}
+
+	/* TODO: MC issue */
+	if (unlikely(hash_type == GMAC_TYPE)) {
+		encr_offset = auth_dlen;
+		enc_dlen = 0;
+	}
+
+	if (unlikely(auth_dlen > enc_dlen)) {
+		inputlen = auth_dlen;
+		outputlen = auth_dlen + mac_len;
+	} else {
+		inputlen = enc_dlen;
+		outputlen = enc_dlen + mac_len;
+	}
+
+	/*GP op header */
+	vq_cmd_w0.u64 = 0;
+	vq_cmd_w0.s.param1 = htobe16(encr_data_len);
+	vq_cmd_w0.s.param2 = htobe16(auth_data_len);
+	/*
+	 * In 83XX since we have a limitation of
+	 * IV & Offset control word not part of instruction
+	 * and need to be part of Data Buffer, we check if
+	 * head room is there and then only do the Direct mode processing
+	 */
+	if (likely((flags & SINGLE_BUF_INPLACE) &&
+		   (flags & SINGLE_BUF_HEADTAILROOM))) {
+		void *dm_vaddr = fc_params->bufs[0].vaddr;
+		uint64_t dm_dma_addr = fc_params->bufs[0].dma_addr;
+		/*
+		 * This flag indicates that there is 24 bytes head room and
+		 * 8 bytes tail room available, so that we get to do
+		 * DIRECT MODE with limitation
+		 */
+
+		offset_vaddr = (uint8_t *)dm_vaddr - OFF_CTRL_LEN - iv_len;
+		offset_dma = dm_dma_addr - OFF_CTRL_LEN - iv_len;
+
+		/* DPTR */
+		req->ist.ei1 = offset_dma;
+		/* RPTR should just exclude offset control word */
+		req->ist.ei2 = dm_dma_addr - iv_len;
+		req->alternate_caddr = (uint64_t *)((uint8_t *)dm_vaddr
+						    + outputlen - iv_len);
+		/* *req->alternate_caddr = ((uint64_t)0xdeadbeefdeadbeef) */
+
+		vq_cmd_w0.s.dlen = htobe16(inputlen + OFF_CTRL_LEN);
+
+		vq_cmd_w0.s.opcode = htobe16(opcode.flags);
+
+		if (likely(iv_len)) {
+			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr
+						      + OFF_CTRL_LEN);
+			uint64_t *src = fc_params->iv_buf;
+			dest[0] = src[0];
+			dest[1] = src[1];
+		}
+
+		*(uint64_t *)offset_vaddr =
+			htobe64(((uint64_t)encr_offset << 16) |
+				((uint64_t)iv_offset << 8) |
+				((uint64_t)auth_offset));
+
+	} else {
+		uint32_t i, g_size_bytes, s_size_bytes;
+		uint64_t dptr_dma, rptr_dma;
+		sg_comp_t *gather_comp;
+		sg_comp_t *scatter_comp;
+		uint8_t *in_buffer;
+
+		/* This falls under strict SG mode */
+		offset_vaddr = m_vaddr;
+		offset_dma = m_dma;
+		size = OFF_CTRL_LEN + iv_len;
+
+		m_vaddr = (uint8_t *)m_vaddr + size;
+		m_dma += size;
+		m_size -= size;
+
+		opcode.s.major |= DMA_MODE;
+
+		vq_cmd_w0.s.opcode = htobe16(opcode.flags);
+
+		if (likely(iv_len)) {
+			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr
+						      + OFF_CTRL_LEN);
+			uint64_t *src = fc_params->iv_buf;
+			dest[0] = src[0];
+			dest[1] = src[1];
+		}
+
+		*(uint64_t *)offset_vaddr =
+			htobe64(((uint64_t)encr_offset << 16) |
+				((uint64_t)iv_offset << 8) |
+				((uint64_t)auth_offset));
+
+		/* DPTR has SG list */
+		in_buffer = m_vaddr;
+		dptr_dma = m_dma;
+
+		((uint16_t *)in_buffer)[0] = 0;
+		((uint16_t *)in_buffer)[1] = 0;
+
+		/* TODO Add error check if space will be sufficient */
+		gather_comp = (sg_comp_t *)((uint8_t *)m_vaddr + 8);
+
+		/*
+		 * Input Gather List
+		 */
+
+		i = 0;
+
+		/* Offset control word that includes iv */
+		i = fill_sg_comp(gather_comp, i, offset_dma,
+				 offset_vaddr, OFF_CTRL_LEN + iv_len);
+
+		/* Add input data */
+		size = inputlen - iv_len;
+		if (likely(size)) {
+			uint32_t aad_offset = aad_len ? passthrough_len : 0;
+
+			if (unlikely(flags & SINGLE_BUF_INPLACE)) {
+				i = fill_sg_comp_from_buf_min(gather_comp, i,
+							      fc_params->bufs,
+							      &size);
+			} else {
+
+				i = fill_sg_comp_from_iov(gather_comp, i,
+							  fc_params->src_iov,
+							  0, &size,
+							  aad_buf, aad_offset);
+			}
+
+			if (unlikely(size)) {
+				PMD_TX_LOG(ERR, "Insufficient buffer space,"
+					   " size %d need\n", size);
+				return ERR_BAD_INPUT_ARG;
+			}
+		}
+		((uint16_t *)in_buffer)[2] = htobe16(i);
+		g_size_bytes = ((i + 3) / 4) * sizeof(sg_comp_t);
+
+		/*
+		 * Output Scatter list
+		 */
+		i = 0;
+		scatter_comp =
+			(sg_comp_t *)((uint8_t *)gather_comp + g_size_bytes);
+
+		/* Add IV */
+		if (likely(iv_len)) {
+			i = fill_sg_comp(scatter_comp, i,
+					 offset_dma + OFF_CTRL_LEN,
+					 (uint8_t *)offset_vaddr + OFF_CTRL_LEN,
+					 iv_len);
+		}
+
+		/* output data or output data + digest*/
+		if (unlikely(flags & VALID_MAC_BUF)) {
+			size = outputlen - iv_len - mac_len;
+			if (size) {
+				uint32_t aad_offset =
+					aad_len ? passthrough_len : 0;
+
+				if (unlikely(flags & SINGLE_BUF_INPLACE)) {
+					i = fill_sg_comp_from_buf_min(
+								scatter_comp,
+								i,
+								fc_params->bufs,
+								&size);
+				} else {
+					i = fill_sg_comp_from_iov(scatter_comp,
+								i,
+							     fc_params->dst_iov,
+								0,
+								&size,
+								aad_buf,
+								aad_offset);
+				}
+				if (size)
+					return ERR_BAD_INPUT_ARG;
+			}
+			/* mac_data */
+			if (mac_len) {
+				i = fill_sg_comp_from_buf(scatter_comp, i,
+							  &fc_params->mac_buf);
+			}
+		} else {
+			/* Output including mac */
+			size = outputlen - iv_len;
+			if (likely(size)) {
+				uint32_t aad_offset =
+					aad_len ? passthrough_len : 0;
+
+				if (unlikely(flags & SINGLE_BUF_INPLACE)) {
+					i = fill_sg_comp_from_buf_min(
+								scatter_comp,
+								i,
+								fc_params->bufs,
+								&size);
+				} else {
+					i = fill_sg_comp_from_iov(scatter_comp,
+								i,
+							     fc_params->dst_iov,
+								0,
+								&size,
+								aad_buf,
+								aad_offset);
+				}
+				if (unlikely(size)) {
+					PMD_TX_LOG(ERR, "Insufficient buffer"
+					" space, size %d need\n", size);
+					return ERR_BAD_INPUT_ARG;
+				}
+			}
+		}
+		((uint16_t *)in_buffer)[3] = htobe16(i);
+		s_size_bytes = ((i + 3) / 4) * sizeof(sg_comp_t);
+
+		size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
+
+		/* This is DPTR len incase of SG mode */
+		vq_cmd_w0.s.dlen = htobe16(size);
+
+		m_vaddr = (uint8_t *)m_vaddr + size;
+		m_dma += size;
+		m_size -= size;
+
+		/* cpt alternate completion address saved earlier */
+		req->alternate_caddr = (uint64_t *)((uint8_t *)c_vaddr - 8);
+		*req->alternate_caddr = ~((uint64_t)COMPLETION_CODE_INIT);
+		rptr_dma = c_dma - 8;
+
+		req->ist.ei1 = dptr_dma;
+		req->ist.ei2 = rptr_dma;
+	}
+
+	/* First 16-bit swap then 64-bit swap */
+	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
+	 * to eliminate all the swapping
+	 */
+	vq_cmd_w0.u64 = htobe64(vq_cmd_w0.u64);
+
+	ctx_dma = fc_params->ctx_buf.dma_addr +
+		offsetof(struct cpt_ctx, fctx);
+	/* vq command w3 */
+	vq_cmd_w3.u64 = 0;
+	vq_cmd_w3.s.grp = 0;
+	vq_cmd_w3.s.cptr = ctx_dma;
+
+	/* 16 byte aligned cpt res address */
+	req->completion_addr = (uint64_t *)((uint8_t *)c_vaddr);
+	*req->completion_addr = COMPLETION_CODE_INIT;
+	req->comp_baddr  = c_dma;
+
+	/* Fill microcode part of instruction */
+	req->ist.ei0 = vq_cmd_w0.u64;
+	req->ist.ei3 = vq_cmd_w3.u64;
+
+	req->op  = op;
+
+#ifdef CPTVF_STRICT_PARAM_CHECK
+	if (!(m_size >= 0)) {
+		PMD_TX_LOG(ERR, "!!! Buffer Overflow %d\n",
+			   m_size);
+		abort();
+	}
+#endif
+	*prep_req = req;
+	return 0;
+}
+
+static inline int
+cpt_dec_hmac_prep(uint32_t flags,
+		  uint64_t d_offs,
+		  uint64_t d_lens,
+		  fc_params_t *fc_params,
+		  void *op,
+		  void **prep_req)
+{
+	uint32_t iv_offset = 0, size;
+	int32_t inputlen, outputlen, enc_dlen, auth_dlen;
+	struct cpt_ctx *cpt_ctx;
+	int32_t hash_type, mac_len, m_size;
+	uint8_t iv_len = 16;
+	cpt_request_info_t *req;
+	buf_ptr_t *meta_p, *aad_buf = NULL;
+	uint32_t encr_offset, auth_offset;
+	uint32_t encr_data_len, auth_data_len, aad_len = 0;
+	uint32_t passthrough_len = 0;
+	void *m_vaddr, *offset_vaddr;
+	uint64_t m_dma, offset_dma, ctx_dma;
+	opcode_info_t opcode;
+	vq_cmd_word0_t vq_cmd_w0;
+	vq_cmd_word3_t vq_cmd_w3;
+	void *c_vaddr;
+	uint64_t c_dma;
+
+	meta_p = &fc_params->meta_buf;
+#ifdef CPTVF_STRICT_PARAM_CHECK
+	if (!fc_params || !meta_p->vaddr || !meta_p->size)
+		return ERR_BAD_INPUT_ARG;
+#endif
+	m_vaddr = meta_p->vaddr;
+	m_dma = meta_p->dma_addr;
+	m_size = meta_p->size;
+
+	encr_offset = ENCR_OFFSET(d_offs);
+	auth_offset = AUTH_OFFSET(d_offs);
+	encr_data_len = ENCR_DLEN(d_lens);
+	auth_data_len = AUTH_DLEN(d_lens);
+
+	if (unlikely(flags & VALID_AAD_BUF)) {
+		/*
+		 * We dont support both aad
+		 * and auth data separately
+		 */
+		auth_data_len = 0;
+		auth_offset = 0;
+		aad_len = fc_params->aad_buf.size;
+		aad_buf = &fc_params->aad_buf;
+	}
+
+	cpt_ctx = fc_params->ctx_buf.vaddr;
+	hash_type = cpt_ctx->hash_type;
+	mac_len = cpt_ctx->mac_len;
+
+	if (hash_type == GMAC_TYPE)
+		encr_data_len = 0;
+
+	if (unlikely(!(flags & VALID_IV_BUF))) {
+		iv_len = 0;
+		iv_offset = ENCR_IV_OFFSET(d_offs);
+	}
+
+	if (unlikely(flags & VALID_AAD_BUF)) {
+		/*
+		 * When AAD is given, data above encr_offset is pass through
+		 * Since AAD is given as separate pointer and not as offset,
+		 * this is a special case as we need to fragment input data
+		 * into passthrough + encr_data and then insert AAD in between.
+		 */
+		if (hash_type != GMAC_TYPE) {
+			passthrough_len = encr_offset;
+			auth_offset = passthrough_len + iv_len;
+			encr_offset = passthrough_len + aad_len + iv_len;
+			auth_data_len = aad_len + encr_data_len;
+		} else {
+			passthrough_len = 16 + aad_len;
+			auth_offset = passthrough_len + iv_len;
+			auth_data_len = aad_len;
+		}
+	} else {
+		encr_offset += iv_len;
+		auth_offset += iv_len;
+	}
+
+	/*
+	 * Save initial space that followed app data for completion code &
+	 * alternate completion code to fall in same cache line as app data
+	 */
+	m_vaddr = (uint8_t *)m_vaddr + COMPLETION_CODE_SIZE;
+	m_dma += COMPLETION_CODE_SIZE;
+	size = (uint8_t *)RTE_PTR_ALIGN((uint8_t *)m_vaddr, 16) -
+	       (uint8_t *)m_vaddr;
+	c_vaddr = (uint8_t *)m_vaddr + size;
+	c_dma = m_dma + size;
+	size += sizeof(cpt_res_s_t);
+
+	m_vaddr = (uint8_t *)m_vaddr + size;
+	m_dma += size;
+	m_size -= size;
+
+	/* start cpt request info structure at 8 byte alignment */
+	size = (uint8_t *)RTE_PTR_ALIGN(m_vaddr, 8) -
+		(uint8_t *)m_vaddr;
+
+	req = (cpt_request_info_t *)((uint8_t *)m_vaddr + size);
+
+	size += sizeof(cpt_request_info_t);
+	m_vaddr = (uint8_t *)m_vaddr + size;
+	m_dma += size;
+	m_size -= size;
+
+	/* Initialising ctrl and opcode
+	 * fields in cpt request structure
+	 */
+
+	req->se_req = SE_CORE_REQ;
+	/*
+	 * We are using DMA mode but indicate that
+	 * SGIO list is already populated.
+	 */
+	req->dma_mode = CTRL_DMA_MODE_SGIO;
+
+	/* Decryption */
+	opcode.s.major = MAJOR_OP_FC;
+	opcode.s.minor = 1;
+
+	enc_dlen = encr_offset + encr_data_len;
+	auth_dlen = auth_offset + auth_data_len;
+
+	if (auth_dlen > enc_dlen) {
+		inputlen = auth_dlen + mac_len;
+		outputlen = auth_dlen;
+	} else {
+		inputlen = enc_dlen + mac_len;
+		outputlen = enc_dlen;
+	}
+
+	if (hash_type == GMAC_TYPE)
+		encr_offset = inputlen;
+
+	vq_cmd_w0.u64 = 0;
+	vq_cmd_w0.s.param1 = htobe16(encr_data_len);
+	vq_cmd_w0.s.param2 = htobe16(auth_data_len);
+
+	/*
+	 * In 83XX since we have a limitation of
+	 * IV & Offset control word not part of instruction
+	 * and need to be part of Data Buffer, we check if
+	 * head room is there and then only do the Direct mode processing
+	 */
+	if (likely((flags & SINGLE_BUF_INPLACE) &&
+		   (flags & SINGLE_BUF_HEADTAILROOM))) {
+		void *dm_vaddr = fc_params->bufs[0].vaddr;
+		uint64_t dm_dma_addr = fc_params->bufs[0].dma_addr;
+		/*
+		 * This flag indicates that there is 24 bytes head room and
+		 * 8 bytes tail room available, so that we get to do
+		 * DIRECT MODE with limitation
+		 */
+
+		offset_vaddr = (uint8_t *)dm_vaddr - OFF_CTRL_LEN - iv_len;
+		offset_dma = dm_dma_addr - OFF_CTRL_LEN - iv_len;
+		req->ist.ei1 = offset_dma;
+
+		/* RPTR should just exclude offset control word */
+		req->ist.ei2 = dm_dma_addr - iv_len;
+
+		/* In direct mode,changing the alternate completion code address
+		 * to start of rptr,the assumption is that most auth iv failure
+		 * are reported at first byte only.This will not give the
+		 * correct alternate completion code the auth iv fail is
+		 * reported  after some bytes.
+		 * FIXME
+		 */
+		req->alternate_caddr = (uint64_t *)((uint8_t *)dm_vaddr -
+					iv_len);
+		/* since this is decryption,
+		 * don't touch the content of
+		 * alternate ccode space as it contains
+		 * hmac.
+		 */
+
+		vq_cmd_w0.s.dlen = htobe16(inputlen + OFF_CTRL_LEN);
+
+		vq_cmd_w0.s.opcode = htobe16(opcode.flags);
+
+		if (likely(iv_len)) {
+			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr +
+						      OFF_CTRL_LEN);
+			uint64_t *src = fc_params->iv_buf;
+			dest[0] = src[0];
+			dest[1] = src[1];
+		}
+
+		*(uint64_t *)offset_vaddr =
+			htobe64(((uint64_t)encr_offset << 16) |
+				((uint64_t)iv_offset << 8) |
+				((uint64_t)auth_offset));
+
+	} else {
+		uint64_t dptr_dma, rptr_dma;
+		uint32_t g_size_bytes, s_size_bytes;
+		sg_comp_t *gather_comp;
+		sg_comp_t *scatter_comp;
+		uint8_t *in_buffer;
+		uint8_t i = 0;
+
+		/* This falls under strict SG mode */
+		offset_vaddr = m_vaddr;
+		offset_dma = m_dma;
+		size = OFF_CTRL_LEN + iv_len;
+
+		m_vaddr = (uint8_t *)m_vaddr + size;
+		m_dma += size;
+		m_size -= size;
+
+		opcode.s.major |= DMA_MODE;
+
+		vq_cmd_w0.s.opcode = htobe16(opcode.flags);
+
+		if (likely(iv_len)) {
+			uint64_t *dest = (uint64_t *)((uint8_t *)offset_vaddr +
+						      OFF_CTRL_LEN);
+			uint64_t *src = fc_params->iv_buf;
+			dest[0] = src[0];
+			dest[1] = src[1];
+		}
+
+		*(uint64_t *)offset_vaddr =
+			htobe64(((uint64_t)encr_offset << 16) |
+				((uint64_t)iv_offset << 8) |
+				((uint64_t)auth_offset));
+
+
+		/* DPTR has SG list */
+		in_buffer = m_vaddr;
+		dptr_dma = m_dma;
+
+		((uint16_t *)in_buffer)[0] = 0;
+		((uint16_t *)in_buffer)[1] = 0;
+
+		/* TODO Add error check if space will be sufficient */
+		gather_comp = (sg_comp_t *)((uint8_t *)m_vaddr + 8);
+
+		/*
+		 * Input Gather List
+		 */
+		i = 0;
+
+		/* Offset control word that includes iv */
+		i = fill_sg_comp(gather_comp, i, offset_dma,
+				 offset_vaddr, OFF_CTRL_LEN + iv_len);
+
+		/* Add input data */
+		if (flags & VALID_MAC_BUF) {
+			size = inputlen - iv_len - mac_len;
+			if (size) {
+				/* input data only */
+				if (unlikely(flags & SINGLE_BUF_INPLACE)) {
+					i = fill_sg_comp_from_buf_min(
+								gather_comp, i,
+								fc_params->bufs,
+								&size);
+				} else {
+					uint32_t aad_offset = aad_len ?
+						passthrough_len : 0;
+
+					i = fill_sg_comp_from_iov(gather_comp,
+							   i,
+							   fc_params->src_iov,
+							   0, &size,
+							   aad_buf,
+							   aad_offset);
+				}
+				if (size)
+					return ERR_BAD_INPUT_ARG;
+			}
+
+			/* mac data */
+			if (mac_len) {
+				i = fill_sg_comp_from_buf(gather_comp, i,
+							  &fc_params->mac_buf);
+			}
+		} else {
+			/* input data + mac */
+			size = inputlen - iv_len;
+			if (size) {
+				if (unlikely(flags & SINGLE_BUF_INPLACE)) {
+					i = fill_sg_comp_from_buf_min(
+								gather_comp, i,
+								fc_params->bufs,
+								&size);
+				} else {
+					uint32_t aad_offset = aad_len ?
+						passthrough_len : 0;
+
+					if (!fc_params->src_iov)
+						return ERR_BAD_INPUT_ARG;
+
+				       i = fill_sg_comp_from_iov(gather_comp, i,
+							     fc_params->src_iov,
+							     0, &size,
+							     aad_buf,
+							     aad_offset);
+				}
+
+				if (size)
+					return ERR_BAD_INPUT_ARG;
+			}
+		}
+		((uint16_t *)in_buffer)[2] = htobe16(i);
+		g_size_bytes = ((i + 3) / 4) * sizeof(sg_comp_t);
+
+		/*
+		 * Output Scatter List
+		 */
+
+		i = 0;
+		scatter_comp =
+			(sg_comp_t *)((uint8_t *)gather_comp + g_size_bytes);
+
+		/* Add iv */
+		if (iv_len) {
+			i = fill_sg_comp(scatter_comp, i,
+					 offset_dma + OFF_CTRL_LEN,
+					 (uint8_t *)offset_vaddr + OFF_CTRL_LEN,
+					 iv_len);
+		}
+
+		/* Add output data */
+		size = outputlen - iv_len;
+		if (size) {
+			if (unlikely(flags & SINGLE_BUF_INPLACE)) {
+				/* handle single buffer here */
+				i = fill_sg_comp_from_buf_min(scatter_comp, i,
+							      fc_params->bufs,
+							      &size);
+			} else {
+				uint32_t aad_offset = aad_len ?
+					passthrough_len : 0;
+
+				if (!fc_params->dst_iov)
+					return ERR_BAD_INPUT_ARG;
+
+				i = fill_sg_comp_from_iov(scatter_comp, i,
+							  fc_params->dst_iov, 0,
+							  &size, aad_buf,
+							  aad_offset);
+			}
+
+			if (unlikely(size))
+				return ERR_BAD_INPUT_ARG;
+		}
+
+		((uint16_t *)in_buffer)[3] = htobe16(i);
+		s_size_bytes = ((i + 3) / 4) * sizeof(sg_comp_t);
+
+		size = g_size_bytes + s_size_bytes + SG_LIST_HDR_SIZE;
+
+		/* This is DPTR len incase of SG mode */
+		vq_cmd_w0.s.dlen = htobe16(size);
+
+		m_vaddr = (uint8_t *)m_vaddr + size;
+		m_dma += size;
+		m_size -= size;
+
+		/* cpt alternate completion address saved earlier */
+		req->alternate_caddr = (uint64_t *)((uint8_t *)c_vaddr - 8);
+		*req->alternate_caddr = ~((uint64_t)COMPLETION_CODE_INIT);
+		rptr_dma = c_dma - 8;
+		size += COMPLETION_CODE_SIZE;
+
+		req->ist.ei1 = dptr_dma;
+		req->ist.ei2 = rptr_dma;
+	}
+
+	/* First 16-bit swap then 64-bit swap */
+	/* TODO: HACK: Reverse the vq_cmd and cpt_req bit field definitions
+	 * to eliminate all the swapping
+	 */
+	vq_cmd_w0.u64 = htobe64(vq_cmd_w0.u64);
+
+	ctx_dma = fc_params->ctx_buf.dma_addr +
+		offsetof(struct cpt_ctx, fctx);
+	/* vq command w3 */
+	vq_cmd_w3.u64 = 0;
+	vq_cmd_w3.s.grp = 0;
+	vq_cmd_w3.s.cptr = ctx_dma;
+
+	/* 16 byte aligned cpt res address */
+	req->completion_addr = (uint64_t *)((uint8_t *)c_vaddr);
+	*req->completion_addr = COMPLETION_CODE_INIT;
+	req->comp_baddr  = c_dma;
+
+	/* Fill microcode part of instruction */
+	req->ist.ei0 = vq_cmd_w0.u64;
+	req->ist.ei3 = vq_cmd_w3.u64;
+
+	req->op = op;
+
+#ifdef CPTVF_STRICT_PARAM_CHECK
+	if (!(m_size >= 0)) {
+		PMD_TX_LOG(ERR, "!!! Buffer Overflow %d\n",
+			   m_size);
+		abort();
+	}
+#endif
+	*prep_req = req;
+	return 0;
+}
+
+void *
+cpt_fc_dec_hmac_prep(uint32_t flags,
+		     uint64_t d_offs,
+		     uint64_t d_lens,
+		     fc_params_t *fc_params,
+		     void *op, int *ret_val)
+{
+	struct cpt_ctx *ctx = fc_params->ctx_buf.vaddr;
+	uint8_t fc_type;
+	void *prep_req = NULL;
+	int ret;
+
+	fc_type = ctx->fc_type;
+
+	if (likely(fc_type == FC_GEN)) {
+		ret = cpt_dec_hmac_prep(flags, d_offs, d_lens,
+					fc_params, op, &prep_req);
+	} else {
+		/*
+		 * For AUTH_ONLY case,
+		 * MC only supports digest generation and verification
+		 * should be done in software by memcmp()
+		 */
+
+		ret = ERR_EIO;
+	}
+
+	if (unlikely(!prep_req))
+		*ret_val = ret;
+	return prep_req;
+}
+
+void *__hot
+cpt_fc_enc_hmac_prep(uint32_t flags, uint64_t d_offs, uint64_t d_lens,
+		     fc_params_t *fc_params, void *op, int *ret_val)
+{
+	struct cpt_ctx *ctx = fc_params->ctx_buf.vaddr;
+	uint8_t fc_type;
+	void *prep_req = NULL;
+	int ret;
+
+	fc_type = ctx->fc_type;
+
+	/* Common api for rest of the ops */
+	if (likely(fc_type == FC_GEN)) {
+		ret = cpt_enc_hmac_prep(flags, d_offs, d_lens,
+					fc_params, op, &prep_req);
+	} else {
+		ret = ERR_EIO;
+	}
+
+	if (unlikely(!prep_req))
+		*ret_val = ret;
+	return prep_req;
+}
-- 
1.9.3