From 2c3c1048746a4622d8c89a29670120dc8fab93c4 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 7 Apr 2024 20:49:45 +0200 Subject: Adding upstream version 6.1.76. Signed-off-by: Daniel Baumann --- drivers/crypto/hisilicon/hpre/Makefile | 2 + drivers/crypto/hisilicon/hpre/hpre.h | 107 ++ drivers/crypto/hisilicon/hpre/hpre_crypto.c | 2217 +++++++++++++++++++++++++++ drivers/crypto/hisilicon/hpre/hpre_main.c | 1546 +++++++++++++++++++ 4 files changed, 3872 insertions(+) create mode 100644 drivers/crypto/hisilicon/hpre/Makefile create mode 100644 drivers/crypto/hisilicon/hpre/hpre.h create mode 100644 drivers/crypto/hisilicon/hpre/hpre_crypto.c create mode 100644 drivers/crypto/hisilicon/hpre/hpre_main.c (limited to 'drivers/crypto/hisilicon/hpre') diff --git a/drivers/crypto/hisilicon/hpre/Makefile b/drivers/crypto/hisilicon/hpre/Makefile new file mode 100644 index 000000000..4fd32b789 --- /dev/null +++ b/drivers/crypto/hisilicon/hpre/Makefile @@ -0,0 +1,2 @@ +obj-$(CONFIG_CRYPTO_DEV_HISI_HPRE) += hisi_hpre.o +hisi_hpre-objs = hpre_main.o hpre_crypto.o diff --git a/drivers/crypto/hisilicon/hpre/hpre.h b/drivers/crypto/hisilicon/hpre/hpre.h new file mode 100644 index 000000000..9f0b94c8e --- /dev/null +++ b/drivers/crypto/hisilicon/hpre/hpre.h @@ -0,0 +1,107 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright (c) 2019 HiSilicon Limited. */ +#ifndef __HISI_HPRE_H +#define __HISI_HPRE_H + +#include +#include + +#define HPRE_SQE_SIZE sizeof(struct hpre_sqe) +#define HPRE_PF_DEF_Q_NUM 64 +#define HPRE_PF_DEF_Q_BASE 0 + +/* + * type used in qm sqc DW6. + * 0 - Algorithm which has been supported in V2, like RSA, DH and so on; + * 1 - ECC algorithm in V3. + */ +#define HPRE_V2_ALG_TYPE 0 +#define HPRE_V3_ECC_ALG_TYPE 1 + +enum { + HPRE_CLUSTER0, + HPRE_CLUSTER1, + HPRE_CLUSTER2, + HPRE_CLUSTER3, + HPRE_CLUSTERS_NUM_MAX +}; + +enum hpre_ctrl_dbgfs_file { + HPRE_CLEAR_ENABLE, + HPRE_CLUSTER_CTRL, + HPRE_DEBUG_FILE_NUM, +}; + +enum hpre_dfx_dbgfs_file { + HPRE_SEND_CNT, + HPRE_RECV_CNT, + HPRE_SEND_FAIL_CNT, + HPRE_SEND_BUSY_CNT, + HPRE_OVER_THRHLD_CNT, + HPRE_OVERTIME_THRHLD, + HPRE_INVALID_REQ_CNT, + HPRE_DFX_FILE_NUM +}; + +#define HPRE_DEBUGFS_FILE_NUM (HPRE_DEBUG_FILE_NUM + HPRE_CLUSTERS_NUM_MAX - 1) + +struct hpre_debugfs_file { + int index; + enum hpre_ctrl_dbgfs_file type; + spinlock_t lock; + struct hpre_debug *debug; +}; + +struct hpre_dfx { + atomic64_t value; + enum hpre_dfx_dbgfs_file type; +}; + +/* + * One HPRE controller has one PF and multiple VFs, some global configurations + * which PF has need this structure. + * Just relevant for PF. + */ +struct hpre_debug { + struct hpre_dfx dfx[HPRE_DFX_FILE_NUM]; + struct hpre_debugfs_file files[HPRE_DEBUGFS_FILE_NUM]; +}; + +struct hpre { + struct hisi_qm qm; + struct hpre_debug debug; + unsigned long status; +}; + +enum hpre_alg_type { + HPRE_ALG_NC_NCRT = 0x0, + HPRE_ALG_NC_CRT = 0x1, + HPRE_ALG_KG_STD = 0x2, + HPRE_ALG_KG_CRT = 0x3, + HPRE_ALG_DH_G2 = 0x4, + HPRE_ALG_DH = 0x5, + HPRE_ALG_ECC_MUL = 0xD, + /* shared by x25519 and x448, but x448 is not supported now */ + HPRE_ALG_CURVE25519_MUL = 0x10, +}; + +struct hpre_sqe { + __le32 dw0; + __u8 task_len1; + __u8 task_len2; + __u8 mrttest_num; + __u8 resv1; + __le64 key; + __le64 in; + __le64 out; + __le16 tag; + __le16 resv2; +#define _HPRE_SQE_ALIGN_EXT 7 + __le32 rsvd1[_HPRE_SQE_ALIGN_EXT]; +}; + +struct hisi_qp *hpre_create_qp(u8 type); +int hpre_algs_register(struct hisi_qm *qm); +void hpre_algs_unregister(struct hisi_qm *qm); +bool hpre_check_alg_support(struct hisi_qm *qm, u32 alg); +#endif diff --git a/drivers/crypto/hisilicon/hpre/hpre_crypto.c b/drivers/crypto/hisilicon/hpre/hpre_crypto.c new file mode 100644 index 000000000..ef02dadd6 --- /dev/null +++ b/drivers/crypto/hisilicon/hpre/hpre_crypto.c @@ -0,0 +1,2217 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2019 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "hpre.h" + +struct hpre_ctx; + +#define HPRE_CRYPTO_ALG_PRI 1000 +#define HPRE_ALIGN_SZ 64 +#define HPRE_BITS_2_BYTES_SHIFT 3 +#define HPRE_RSA_512BITS_KSZ 64 +#define HPRE_RSA_1536BITS_KSZ 192 +#define HPRE_CRT_PRMS 5 +#define HPRE_CRT_Q 2 +#define HPRE_CRT_P 3 +#define HPRE_CRT_INV 4 +#define HPRE_DH_G_FLAG 0x02 +#define HPRE_TRY_SEND_TIMES 100 +#define HPRE_INVLD_REQ_ID (-1) + +#define HPRE_SQE_ALG_BITS 5 +#define HPRE_SQE_DONE_SHIFT 30 +#define HPRE_DH_MAX_P_SZ 512 + +#define HPRE_DFX_SEC_TO_US 1000000 +#define HPRE_DFX_US_TO_NS 1000 + +/* due to nist p521 */ +#define HPRE_ECC_MAX_KSZ 66 + +/* size in bytes of the n prime */ +#define HPRE_ECC_NIST_P192_N_SIZE 24 +#define HPRE_ECC_NIST_P256_N_SIZE 32 +#define HPRE_ECC_NIST_P384_N_SIZE 48 + +/* size in bytes */ +#define HPRE_ECC_HW256_KSZ_B 32 +#define HPRE_ECC_HW384_KSZ_B 48 + +/* capability register mask of driver */ +#define HPRE_DRV_RSA_MASK_CAP BIT(0) +#define HPRE_DRV_DH_MASK_CAP BIT(1) +#define HPRE_DRV_ECDH_MASK_CAP BIT(2) +#define HPRE_DRV_X25519_MASK_CAP BIT(5) + +typedef void (*hpre_cb)(struct hpre_ctx *ctx, void *sqe); + +struct hpre_rsa_ctx { + /* low address: e--->n */ + char *pubkey; + dma_addr_t dma_pubkey; + + /* low address: d--->n */ + char *prikey; + dma_addr_t dma_prikey; + + /* low address: dq->dp->q->p->qinv */ + char *crt_prikey; + dma_addr_t dma_crt_prikey; + + struct crypto_akcipher *soft_tfm; +}; + +struct hpre_dh_ctx { + /* + * If base is g we compute the public key + * ya = g^xa mod p; [RFC2631 sec 2.1.1] + * else if base if the counterpart public key we + * compute the shared secret + * ZZ = yb^xa mod p; [RFC2631 sec 2.1.1] + * low address: d--->n, please refer to Hisilicon HPRE UM + */ + char *xa_p; + dma_addr_t dma_xa_p; + + char *g; /* m */ + dma_addr_t dma_g; +}; + +struct hpre_ecdh_ctx { + /* low address: p->a->k->b */ + unsigned char *p; + dma_addr_t dma_p; + + /* low address: x->y */ + unsigned char *g; + dma_addr_t dma_g; +}; + +struct hpre_curve25519_ctx { + /* low address: p->a->k */ + unsigned char *p; + dma_addr_t dma_p; + + /* gx coordinate */ + unsigned char *g; + dma_addr_t dma_g; +}; + +struct hpre_ctx { + struct hisi_qp *qp; + struct device *dev; + struct hpre_asym_request **req_list; + struct hpre *hpre; + spinlock_t req_lock; + unsigned int key_sz; + bool crt_g2_mode; + struct idr req_idr; + union { + struct hpre_rsa_ctx rsa; + struct hpre_dh_ctx dh; + struct hpre_ecdh_ctx ecdh; + struct hpre_curve25519_ctx curve25519; + }; + /* for ecc algorithms */ + unsigned int curve_id; +}; + +struct hpre_asym_request { + char *src; + char *dst; + struct hpre_sqe req; + struct hpre_ctx *ctx; + union { + struct akcipher_request *rsa; + struct kpp_request *dh; + struct kpp_request *ecdh; + struct kpp_request *curve25519; + } areq; + int err; + int req_id; + hpre_cb cb; + struct timespec64 req_time; +}; + +static int hpre_alloc_req_id(struct hpre_ctx *ctx) +{ + unsigned long flags; + int id; + + spin_lock_irqsave(&ctx->req_lock, flags); + id = idr_alloc(&ctx->req_idr, NULL, 0, ctx->qp->sq_depth, GFP_ATOMIC); + spin_unlock_irqrestore(&ctx->req_lock, flags); + + return id; +} + +static void hpre_free_req_id(struct hpre_ctx *ctx, int req_id) +{ + unsigned long flags; + + spin_lock_irqsave(&ctx->req_lock, flags); + idr_remove(&ctx->req_idr, req_id); + spin_unlock_irqrestore(&ctx->req_lock, flags); +} + +static int hpre_add_req_to_ctx(struct hpre_asym_request *hpre_req) +{ + struct hpre_ctx *ctx; + struct hpre_dfx *dfx; + int id; + + ctx = hpre_req->ctx; + id = hpre_alloc_req_id(ctx); + if (unlikely(id < 0)) + return -EINVAL; + + ctx->req_list[id] = hpre_req; + hpre_req->req_id = id; + + dfx = ctx->hpre->debug.dfx; + if (atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value)) + ktime_get_ts64(&hpre_req->req_time); + + return id; +} + +static void hpre_rm_req_from_ctx(struct hpre_asym_request *hpre_req) +{ + struct hpre_ctx *ctx = hpre_req->ctx; + int id = hpre_req->req_id; + + if (hpre_req->req_id >= 0) { + hpre_req->req_id = HPRE_INVLD_REQ_ID; + ctx->req_list[id] = NULL; + hpre_free_req_id(ctx, id); + } +} + +static struct hisi_qp *hpre_get_qp_and_start(u8 type) +{ + struct hisi_qp *qp; + int ret; + + qp = hpre_create_qp(type); + if (!qp) { + pr_err("Can not create hpre qp!\n"); + return ERR_PTR(-ENODEV); + } + + ret = hisi_qm_start_qp(qp, 0); + if (ret < 0) { + hisi_qm_free_qps(&qp, 1); + pci_err(qp->qm->pdev, "Can not start qp!\n"); + return ERR_PTR(-EINVAL); + } + + return qp; +} + +static int hpre_get_data_dma_addr(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len, + int is_src, dma_addr_t *tmp) +{ + struct device *dev = hpre_req->ctx->dev; + enum dma_data_direction dma_dir; + + if (is_src) { + hpre_req->src = NULL; + dma_dir = DMA_TO_DEVICE; + } else { + hpre_req->dst = NULL; + dma_dir = DMA_FROM_DEVICE; + } + *tmp = dma_map_single(dev, sg_virt(data), len, dma_dir); + if (unlikely(dma_mapping_error(dev, *tmp))) { + dev_err(dev, "dma map data err!\n"); + return -ENOMEM; + } + + return 0; +} + +static int hpre_prepare_dma_buf(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len, + int is_src, dma_addr_t *tmp) +{ + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + void *ptr; + int shift; + + shift = ctx->key_sz - len; + if (unlikely(shift < 0)) + return -EINVAL; + + ptr = dma_alloc_coherent(dev, ctx->key_sz, tmp, GFP_ATOMIC); + if (unlikely(!ptr)) + return -ENOMEM; + + if (is_src) { + scatterwalk_map_and_copy(ptr + shift, data, 0, len, 0); + hpre_req->src = ptr; + } else { + hpre_req->dst = ptr; + } + + return 0; +} + +static int hpre_hw_data_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len, + int is_src, int is_dh) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + dma_addr_t tmp = 0; + int ret; + + /* when the data is dh's source, we should format it */ + if ((sg_is_last(data) && len == ctx->key_sz) && + ((is_dh && !is_src) || !is_dh)) + ret = hpre_get_data_dma_addr(hpre_req, data, len, is_src, &tmp); + else + ret = hpre_prepare_dma_buf(hpre_req, data, len, is_src, &tmp); + + if (unlikely(ret)) + return ret; + + if (is_src) + msg->in = cpu_to_le64(tmp); + else + msg->out = cpu_to_le64(tmp); + + return 0; +} + +static void hpre_hw_data_clr_all(struct hpre_ctx *ctx, + struct hpre_asym_request *req, + struct scatterlist *dst, + struct scatterlist *src) +{ + struct device *dev = ctx->dev; + struct hpre_sqe *sqe = &req->req; + dma_addr_t tmp; + + tmp = le64_to_cpu(sqe->in); + if (unlikely(dma_mapping_error(dev, tmp))) + return; + + if (src) { + if (req->src) + dma_free_coherent(dev, ctx->key_sz, req->src, tmp); + else + dma_unmap_single(dev, tmp, ctx->key_sz, DMA_TO_DEVICE); + } + + tmp = le64_to_cpu(sqe->out); + if (unlikely(dma_mapping_error(dev, tmp))) + return; + + if (req->dst) { + if (dst) + scatterwalk_map_and_copy(req->dst, dst, 0, + ctx->key_sz, 1); + dma_free_coherent(dev, ctx->key_sz, req->dst, tmp); + } else { + dma_unmap_single(dev, tmp, ctx->key_sz, DMA_FROM_DEVICE); + } +} + +static int hpre_alg_res_post_hf(struct hpre_ctx *ctx, struct hpre_sqe *sqe, + void **kreq) +{ + struct hpre_asym_request *req; + unsigned int err, done, alg; + int id; + +#define HPRE_NO_HW_ERR 0 +#define HPRE_HW_TASK_DONE 3 +#define HREE_HW_ERR_MASK GENMASK(10, 0) +#define HREE_SQE_DONE_MASK GENMASK(1, 0) +#define HREE_ALG_TYPE_MASK GENMASK(4, 0) + id = (int)le16_to_cpu(sqe->tag); + req = ctx->req_list[id]; + hpre_rm_req_from_ctx(req); + *kreq = req; + + err = (le32_to_cpu(sqe->dw0) >> HPRE_SQE_ALG_BITS) & + HREE_HW_ERR_MASK; + + done = (le32_to_cpu(sqe->dw0) >> HPRE_SQE_DONE_SHIFT) & + HREE_SQE_DONE_MASK; + + if (likely(err == HPRE_NO_HW_ERR && done == HPRE_HW_TASK_DONE)) + return 0; + + alg = le32_to_cpu(sqe->dw0) & HREE_ALG_TYPE_MASK; + dev_err_ratelimited(ctx->dev, "alg[0x%x] error: done[0x%x], etype[0x%x]\n", + alg, done, err); + + return -EINVAL; +} + +static int hpre_ctx_set(struct hpre_ctx *ctx, struct hisi_qp *qp, int qlen) +{ + struct hpre *hpre; + + if (!ctx || !qp || qlen < 0) + return -EINVAL; + + spin_lock_init(&ctx->req_lock); + ctx->qp = qp; + ctx->dev = &qp->qm->pdev->dev; + + hpre = container_of(ctx->qp->qm, struct hpre, qm); + ctx->hpre = hpre; + ctx->req_list = kcalloc(qlen, sizeof(void *), GFP_KERNEL); + if (!ctx->req_list) + return -ENOMEM; + ctx->key_sz = 0; + ctx->crt_g2_mode = false; + idr_init(&ctx->req_idr); + + return 0; +} + +static void hpre_ctx_clear(struct hpre_ctx *ctx, bool is_clear_all) +{ + if (is_clear_all) { + idr_destroy(&ctx->req_idr); + kfree(ctx->req_list); + hisi_qm_free_qps(&ctx->qp, 1); + } + + ctx->crt_g2_mode = false; + ctx->key_sz = 0; +} + +static bool hpre_is_bd_timeout(struct hpre_asym_request *req, + u64 overtime_thrhld) +{ + struct timespec64 reply_time; + u64 time_use_us; + + ktime_get_ts64(&reply_time); + time_use_us = (reply_time.tv_sec - req->req_time.tv_sec) * + HPRE_DFX_SEC_TO_US + + (reply_time.tv_nsec - req->req_time.tv_nsec) / + HPRE_DFX_US_TO_NS; + + if (time_use_us <= overtime_thrhld) + return false; + + return true; +} + +static void hpre_dh_cb(struct hpre_ctx *ctx, void *resp) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req; + struct kpp_request *areq; + u64 overtime_thrhld; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + areq = req->areq.dh; + areq->dst_len = ctx->key_sz; + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + hpre_hw_data_clr_all(ctx, req, areq->dst, areq->src); + kpp_request_complete(areq, ret); + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static void hpre_rsa_cb(struct hpre_ctx *ctx, void *resp) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req; + struct akcipher_request *areq; + u64 overtime_thrhld; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + areq = req->areq.rsa; + areq->dst_len = ctx->key_sz; + hpre_hw_data_clr_all(ctx, req, areq->dst, areq->src); + akcipher_request_complete(areq, ret); + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static void hpre_alg_cb(struct hisi_qp *qp, void *resp) +{ + struct hpre_ctx *ctx = qp->qp_ctx; + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_sqe *sqe = resp; + struct hpre_asym_request *req = ctx->req_list[le16_to_cpu(sqe->tag)]; + + if (unlikely(!req)) { + atomic64_inc(&dfx[HPRE_INVALID_REQ_CNT].value); + return; + } + + req->cb(ctx, resp); +} + +static void hpre_stop_qp_and_put(struct hisi_qp *qp) +{ + hisi_qm_stop_qp(qp); + hisi_qm_free_qps(&qp, 1); +} + +static int hpre_ctx_init(struct hpre_ctx *ctx, u8 type) +{ + struct hisi_qp *qp; + int ret; + + qp = hpre_get_qp_and_start(type); + if (IS_ERR(qp)) + return PTR_ERR(qp); + + qp->qp_ctx = ctx; + qp->req_cb = hpre_alg_cb; + + ret = hpre_ctx_set(ctx, qp, qp->sq_depth); + if (ret) + hpre_stop_qp_and_put(qp); + + return ret; +} + +static int hpre_msg_request_set(struct hpre_ctx *ctx, void *req, bool is_rsa) +{ + struct hpre_asym_request *h_req; + struct hpre_sqe *msg; + int req_id; + void *tmp; + + if (is_rsa) { + struct akcipher_request *akreq = req; + + if (akreq->dst_len < ctx->key_sz) { + akreq->dst_len = ctx->key_sz; + return -EOVERFLOW; + } + + tmp = akcipher_request_ctx(akreq); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_rsa_cb; + h_req->areq.rsa = akreq; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + } else { + struct kpp_request *kreq = req; + + if (kreq->dst_len < ctx->key_sz) { + kreq->dst_len = ctx->key_sz; + return -EOVERFLOW; + } + + tmp = kpp_request_ctx(kreq); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_dh_cb; + h_req->areq.dh = kreq; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + msg->key = cpu_to_le64(ctx->dh.dma_xa_p); + } + + msg->in = cpu_to_le64(DMA_MAPPING_ERROR); + msg->out = cpu_to_le64(DMA_MAPPING_ERROR); + msg->dw0 |= cpu_to_le32(0x1 << HPRE_SQE_DONE_SHIFT); + msg->task_len1 = (ctx->key_sz >> HPRE_BITS_2_BYTES_SHIFT) - 1; + h_req->ctx = ctx; + + req_id = hpre_add_req_to_ctx(h_req); + if (req_id < 0) + return -EBUSY; + + msg->tag = cpu_to_le16((u16)req_id); + + return 0; +} + +static int hpre_send(struct hpre_ctx *ctx, struct hpre_sqe *msg) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + int ctr = 0; + int ret; + + do { + atomic64_inc(&dfx[HPRE_SEND_CNT].value); + ret = hisi_qp_send(ctx->qp, msg); + if (ret != -EBUSY) + break; + atomic64_inc(&dfx[HPRE_SEND_BUSY_CNT].value); + } while (ctr++ < HPRE_TRY_SEND_TIMES); + + if (likely(!ret)) + return ret; + + if (ret != -EBUSY) + atomic64_inc(&dfx[HPRE_SEND_FAIL_CNT].value); + + return ret; +} + +static int hpre_dh_compute_value(struct kpp_request *req) +{ + struct crypto_kpp *tfm = crypto_kpp_reqtfm(req); + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + void *tmp = kpp_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + ret = hpre_msg_request_set(ctx, req, false); + if (unlikely(ret)) + return ret; + + if (req->src) { + ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 1); + if (unlikely(ret)) + goto clear_all; + } else { + msg->in = cpu_to_le64(ctx->dh.dma_g); + } + + ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 1); + if (unlikely(ret)) + goto clear_all; + + if (ctx->crt_g2_mode && !req->src) + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_DH_G2); + else + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_DH); + + /* success */ + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + + return ret; +} + +static int hpre_is_dh_params_length_valid(unsigned int key_sz) +{ +#define _HPRE_DH_GRP1 768 +#define _HPRE_DH_GRP2 1024 +#define _HPRE_DH_GRP5 1536 +#define _HPRE_DH_GRP14 2048 +#define _HPRE_DH_GRP15 3072 +#define _HPRE_DH_GRP16 4096 + switch (key_sz) { + case _HPRE_DH_GRP1: + case _HPRE_DH_GRP2: + case _HPRE_DH_GRP5: + case _HPRE_DH_GRP14: + case _HPRE_DH_GRP15: + case _HPRE_DH_GRP16: + return 0; + default: + return -EINVAL; + } +} + +static int hpre_dh_set_params(struct hpre_ctx *ctx, struct dh *params) +{ + struct device *dev = ctx->dev; + unsigned int sz; + + if (params->p_size > HPRE_DH_MAX_P_SZ) + return -EINVAL; + + if (hpre_is_dh_params_length_valid(params->p_size << + HPRE_BITS_2_BYTES_SHIFT)) + return -EINVAL; + + sz = ctx->key_sz = params->p_size; + ctx->dh.xa_p = dma_alloc_coherent(dev, sz << 1, + &ctx->dh.dma_xa_p, GFP_KERNEL); + if (!ctx->dh.xa_p) + return -ENOMEM; + + memcpy(ctx->dh.xa_p + sz, params->p, sz); + + /* If g equals 2 don't copy it */ + if (params->g_size == 1 && *(char *)params->g == HPRE_DH_G_FLAG) { + ctx->crt_g2_mode = true; + return 0; + } + + ctx->dh.g = dma_alloc_coherent(dev, sz, &ctx->dh.dma_g, GFP_KERNEL); + if (!ctx->dh.g) { + dma_free_coherent(dev, sz << 1, ctx->dh.xa_p, + ctx->dh.dma_xa_p); + ctx->dh.xa_p = NULL; + return -ENOMEM; + } + + memcpy(ctx->dh.g + (sz - params->g_size), params->g, params->g_size); + + return 0; +} + +static void hpre_dh_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all) +{ + struct device *dev = ctx->dev; + unsigned int sz = ctx->key_sz; + + if (is_clear_all) + hisi_qm_stop_qp(ctx->qp); + + if (ctx->dh.g) { + dma_free_coherent(dev, sz, ctx->dh.g, ctx->dh.dma_g); + ctx->dh.g = NULL; + } + + if (ctx->dh.xa_p) { + memzero_explicit(ctx->dh.xa_p, sz); + dma_free_coherent(dev, sz << 1, ctx->dh.xa_p, + ctx->dh.dma_xa_p); + ctx->dh.xa_p = NULL; + } + + hpre_ctx_clear(ctx, is_clear_all); +} + +static int hpre_dh_set_secret(struct crypto_kpp *tfm, const void *buf, + unsigned int len) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct dh params; + int ret; + + if (crypto_dh_decode_key(buf, len, ¶ms) < 0) + return -EINVAL; + + /* Free old secret if any */ + hpre_dh_clear_ctx(ctx, false); + + ret = hpre_dh_set_params(ctx, ¶ms); + if (ret < 0) + goto err_clear_ctx; + + memcpy(ctx->dh.xa_p + (ctx->key_sz - params.key_size), params.key, + params.key_size); + + return 0; + +err_clear_ctx: + hpre_dh_clear_ctx(ctx, false); + return ret; +} + +static unsigned int hpre_dh_max_size(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return ctx->key_sz; +} + +static int hpre_dh_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return hpre_ctx_init(ctx, HPRE_V2_ALG_TYPE); +} + +static void hpre_dh_exit_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + hpre_dh_clear_ctx(ctx, true); +} + +static void hpre_rsa_drop_leading_zeros(const char **ptr, size_t *len) +{ + while (!**ptr && *len) { + (*ptr)++; + (*len)--; + } +} + +static bool hpre_rsa_key_size_is_support(unsigned int len) +{ + unsigned int bits = len << HPRE_BITS_2_BYTES_SHIFT; + +#define _RSA_1024BITS_KEY_WDTH 1024 +#define _RSA_2048BITS_KEY_WDTH 2048 +#define _RSA_3072BITS_KEY_WDTH 3072 +#define _RSA_4096BITS_KEY_WDTH 4096 + + switch (bits) { + case _RSA_1024BITS_KEY_WDTH: + case _RSA_2048BITS_KEY_WDTH: + case _RSA_3072BITS_KEY_WDTH: + case _RSA_4096BITS_KEY_WDTH: + return true; + default: + return false; + } +} + +static int hpre_rsa_enc(struct akcipher_request *req) +{ + struct crypto_akcipher *tfm = crypto_akcipher_reqtfm(req); + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + void *tmp = akcipher_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + /* For 512 and 1536 bits key size, use soft tfm instead */ + if (ctx->key_sz == HPRE_RSA_512BITS_KSZ || + ctx->key_sz == HPRE_RSA_1536BITS_KSZ) { + akcipher_request_set_tfm(req, ctx->rsa.soft_tfm); + ret = crypto_akcipher_encrypt(req); + akcipher_request_set_tfm(req, tfm); + return ret; + } + + if (unlikely(!ctx->rsa.pubkey)) + return -EINVAL; + + ret = hpre_msg_request_set(ctx, req, true); + if (unlikely(ret)) + return ret; + + msg->dw0 |= cpu_to_le32(HPRE_ALG_NC_NCRT); + msg->key = cpu_to_le64(ctx->rsa.dma_pubkey); + + ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0); + if (unlikely(ret)) + goto clear_all; + + ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 0); + if (unlikely(ret)) + goto clear_all; + + /* success */ + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + + return ret; +} + +static int hpre_rsa_dec(struct akcipher_request *req) +{ + struct crypto_akcipher *tfm = crypto_akcipher_reqtfm(req); + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + void *tmp = akcipher_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + /* For 512 and 1536 bits key size, use soft tfm instead */ + if (ctx->key_sz == HPRE_RSA_512BITS_KSZ || + ctx->key_sz == HPRE_RSA_1536BITS_KSZ) { + akcipher_request_set_tfm(req, ctx->rsa.soft_tfm); + ret = crypto_akcipher_decrypt(req); + akcipher_request_set_tfm(req, tfm); + return ret; + } + + if (unlikely(!ctx->rsa.prikey)) + return -EINVAL; + + ret = hpre_msg_request_set(ctx, req, true); + if (unlikely(ret)) + return ret; + + if (ctx->crt_g2_mode) { + msg->key = cpu_to_le64(ctx->rsa.dma_crt_prikey); + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | + HPRE_ALG_NC_CRT); + } else { + msg->key = cpu_to_le64(ctx->rsa.dma_prikey); + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | + HPRE_ALG_NC_NCRT); + } + + ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0); + if (unlikely(ret)) + goto clear_all; + + ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 0); + if (unlikely(ret)) + goto clear_all; + + /* success */ + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + + return ret; +} + +static int hpre_rsa_set_n(struct hpre_ctx *ctx, const char *value, + size_t vlen, bool private) +{ + const char *ptr = value; + + hpre_rsa_drop_leading_zeros(&ptr, &vlen); + + ctx->key_sz = vlen; + + /* if invalid key size provided, we use software tfm */ + if (!hpre_rsa_key_size_is_support(ctx->key_sz)) + return 0; + + ctx->rsa.pubkey = dma_alloc_coherent(ctx->dev, vlen << 1, + &ctx->rsa.dma_pubkey, + GFP_KERNEL); + if (!ctx->rsa.pubkey) + return -ENOMEM; + + if (private) { + ctx->rsa.prikey = dma_alloc_coherent(ctx->dev, vlen << 1, + &ctx->rsa.dma_prikey, + GFP_KERNEL); + if (!ctx->rsa.prikey) { + dma_free_coherent(ctx->dev, vlen << 1, + ctx->rsa.pubkey, + ctx->rsa.dma_pubkey); + ctx->rsa.pubkey = NULL; + return -ENOMEM; + } + memcpy(ctx->rsa.prikey + vlen, ptr, vlen); + } + memcpy(ctx->rsa.pubkey + vlen, ptr, vlen); + + /* Using hardware HPRE to do RSA */ + return 1; +} + +static int hpre_rsa_set_e(struct hpre_ctx *ctx, const char *value, + size_t vlen) +{ + const char *ptr = value; + + hpre_rsa_drop_leading_zeros(&ptr, &vlen); + + if (!ctx->key_sz || !vlen || vlen > ctx->key_sz) + return -EINVAL; + + memcpy(ctx->rsa.pubkey + ctx->key_sz - vlen, ptr, vlen); + + return 0; +} + +static int hpre_rsa_set_d(struct hpre_ctx *ctx, const char *value, + size_t vlen) +{ + const char *ptr = value; + + hpre_rsa_drop_leading_zeros(&ptr, &vlen); + + if (!ctx->key_sz || !vlen || vlen > ctx->key_sz) + return -EINVAL; + + memcpy(ctx->rsa.prikey + ctx->key_sz - vlen, ptr, vlen); + + return 0; +} + +static int hpre_crt_para_get(char *para, size_t para_sz, + const char *raw, size_t raw_sz) +{ + const char *ptr = raw; + size_t len = raw_sz; + + hpre_rsa_drop_leading_zeros(&ptr, &len); + if (!len || len > para_sz) + return -EINVAL; + + memcpy(para + para_sz - len, ptr, len); + + return 0; +} + +static int hpre_rsa_setkey_crt(struct hpre_ctx *ctx, struct rsa_key *rsa_key) +{ + unsigned int hlf_ksz = ctx->key_sz >> 1; + struct device *dev = ctx->dev; + u64 offset; + int ret; + + ctx->rsa.crt_prikey = dma_alloc_coherent(dev, hlf_ksz * HPRE_CRT_PRMS, + &ctx->rsa.dma_crt_prikey, + GFP_KERNEL); + if (!ctx->rsa.crt_prikey) + return -ENOMEM; + + ret = hpre_crt_para_get(ctx->rsa.crt_prikey, hlf_ksz, + rsa_key->dq, rsa_key->dq_sz); + if (ret) + goto free_key; + + offset = hlf_ksz; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->dp, rsa_key->dp_sz); + if (ret) + goto free_key; + + offset = hlf_ksz * HPRE_CRT_Q; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->q, rsa_key->q_sz); + if (ret) + goto free_key; + + offset = hlf_ksz * HPRE_CRT_P; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->p, rsa_key->p_sz); + if (ret) + goto free_key; + + offset = hlf_ksz * HPRE_CRT_INV; + ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz, + rsa_key->qinv, rsa_key->qinv_sz); + if (ret) + goto free_key; + + ctx->crt_g2_mode = true; + + return 0; + +free_key: + offset = hlf_ksz * HPRE_CRT_PRMS; + memzero_explicit(ctx->rsa.crt_prikey, offset); + dma_free_coherent(dev, hlf_ksz * HPRE_CRT_PRMS, ctx->rsa.crt_prikey, + ctx->rsa.dma_crt_prikey); + ctx->rsa.crt_prikey = NULL; + ctx->crt_g2_mode = false; + + return ret; +} + +/* If it is clear all, all the resources of the QP will be cleaned. */ +static void hpre_rsa_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all) +{ + unsigned int half_key_sz = ctx->key_sz >> 1; + struct device *dev = ctx->dev; + + if (is_clear_all) + hisi_qm_stop_qp(ctx->qp); + + if (ctx->rsa.pubkey) { + dma_free_coherent(dev, ctx->key_sz << 1, + ctx->rsa.pubkey, ctx->rsa.dma_pubkey); + ctx->rsa.pubkey = NULL; + } + + if (ctx->rsa.crt_prikey) { + memzero_explicit(ctx->rsa.crt_prikey, + half_key_sz * HPRE_CRT_PRMS); + dma_free_coherent(dev, half_key_sz * HPRE_CRT_PRMS, + ctx->rsa.crt_prikey, ctx->rsa.dma_crt_prikey); + ctx->rsa.crt_prikey = NULL; + } + + if (ctx->rsa.prikey) { + memzero_explicit(ctx->rsa.prikey, ctx->key_sz); + dma_free_coherent(dev, ctx->key_sz << 1, ctx->rsa.prikey, + ctx->rsa.dma_prikey); + ctx->rsa.prikey = NULL; + } + + hpre_ctx_clear(ctx, is_clear_all); +} + +/* + * we should judge if it is CRT or not, + * CRT: return true, N-CRT: return false . + */ +static bool hpre_is_crt_key(struct rsa_key *key) +{ + u16 len = key->p_sz + key->q_sz + key->dp_sz + key->dq_sz + + key->qinv_sz; + +#define LEN_OF_NCRT_PARA 5 + + /* N-CRT less than 5 parameters */ + return len > LEN_OF_NCRT_PARA; +} + +static int hpre_rsa_setkey(struct hpre_ctx *ctx, const void *key, + unsigned int keylen, bool private) +{ + struct rsa_key rsa_key; + int ret; + + hpre_rsa_clear_ctx(ctx, false); + + if (private) + ret = rsa_parse_priv_key(&rsa_key, key, keylen); + else + ret = rsa_parse_pub_key(&rsa_key, key, keylen); + if (ret < 0) + return ret; + + ret = hpre_rsa_set_n(ctx, rsa_key.n, rsa_key.n_sz, private); + if (ret <= 0) + return ret; + + if (private) { + ret = hpre_rsa_set_d(ctx, rsa_key.d, rsa_key.d_sz); + if (ret < 0) + goto free; + + if (hpre_is_crt_key(&rsa_key)) { + ret = hpre_rsa_setkey_crt(ctx, &rsa_key); + if (ret < 0) + goto free; + } + } + + ret = hpre_rsa_set_e(ctx, rsa_key.e, rsa_key.e_sz); + if (ret < 0) + goto free; + + if ((private && !ctx->rsa.prikey) || !ctx->rsa.pubkey) { + ret = -EINVAL; + goto free; + } + + return 0; + +free: + hpre_rsa_clear_ctx(ctx, false); + return ret; +} + +static int hpre_rsa_setpubkey(struct crypto_akcipher *tfm, const void *key, + unsigned int keylen) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + int ret; + + ret = crypto_akcipher_set_pub_key(ctx->rsa.soft_tfm, key, keylen); + if (ret) + return ret; + + return hpre_rsa_setkey(ctx, key, keylen, false); +} + +static int hpre_rsa_setprivkey(struct crypto_akcipher *tfm, const void *key, + unsigned int keylen) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + int ret; + + ret = crypto_akcipher_set_priv_key(ctx->rsa.soft_tfm, key, keylen); + if (ret) + return ret; + + return hpre_rsa_setkey(ctx, key, keylen, true); +} + +static unsigned int hpre_rsa_max_size(struct crypto_akcipher *tfm) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + + /* For 512 and 1536 bits key size, use soft tfm instead */ + if (ctx->key_sz == HPRE_RSA_512BITS_KSZ || + ctx->key_sz == HPRE_RSA_1536BITS_KSZ) + return crypto_akcipher_maxsize(ctx->rsa.soft_tfm); + + return ctx->key_sz; +} + +static int hpre_rsa_init_tfm(struct crypto_akcipher *tfm) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + int ret; + + ctx->rsa.soft_tfm = crypto_alloc_akcipher("rsa-generic", 0, 0); + if (IS_ERR(ctx->rsa.soft_tfm)) { + pr_err("Can not alloc_akcipher!\n"); + return PTR_ERR(ctx->rsa.soft_tfm); + } + + ret = hpre_ctx_init(ctx, HPRE_V2_ALG_TYPE); + if (ret) + crypto_free_akcipher(ctx->rsa.soft_tfm); + + return ret; +} + +static void hpre_rsa_exit_tfm(struct crypto_akcipher *tfm) +{ + struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm); + + hpre_rsa_clear_ctx(ctx, true); + crypto_free_akcipher(ctx->rsa.soft_tfm); +} + +static void hpre_key_to_big_end(u8 *data, int len) +{ + int i, j; + + for (i = 0; i < len / 2; i++) { + j = len - i - 1; + swap(data[j], data[i]); + } +} + +static void hpre_ecc_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all, + bool is_ecdh) +{ + struct device *dev = ctx->dev; + unsigned int sz = ctx->key_sz; + unsigned int shift = sz << 1; + + if (is_clear_all) + hisi_qm_stop_qp(ctx->qp); + + if (is_ecdh && ctx->ecdh.p) { + /* ecdh: p->a->k->b */ + memzero_explicit(ctx->ecdh.p + shift, sz); + dma_free_coherent(dev, sz << 3, ctx->ecdh.p, ctx->ecdh.dma_p); + ctx->ecdh.p = NULL; + } else if (!is_ecdh && ctx->curve25519.p) { + /* curve25519: p->a->k */ + memzero_explicit(ctx->curve25519.p + shift, sz); + dma_free_coherent(dev, sz << 2, ctx->curve25519.p, + ctx->curve25519.dma_p); + ctx->curve25519.p = NULL; + } + + hpre_ctx_clear(ctx, is_clear_all); +} + +/* + * The bits of 192/224/256/384/521 are supported by HPRE, + * and convert the bits like: + * bits<=256, bits=256; 256key_sz << 1; + unsigned int shiftb = ctx->key_sz << 2; + void *p = ctx->ecdh.p + ctx->key_sz - cur_sz; + void *a = ctx->ecdh.p + shifta - cur_sz; + void *b = ctx->ecdh.p + shiftb - cur_sz; + void *x = ctx->ecdh.g + ctx->key_sz - cur_sz; + void *y = ctx->ecdh.g + shifta - cur_sz; + const struct ecc_curve *curve = ecc_get_curve(ctx->curve_id); + char *n; + + if (unlikely(!curve)) + return -EINVAL; + + n = kzalloc(ctx->key_sz, GFP_KERNEL); + if (!n) + return -ENOMEM; + + fill_curve_param(p, curve->p, cur_sz, curve->g.ndigits); + fill_curve_param(a, curve->a, cur_sz, curve->g.ndigits); + fill_curve_param(b, curve->b, cur_sz, curve->g.ndigits); + fill_curve_param(x, curve->g.x, cur_sz, curve->g.ndigits); + fill_curve_param(y, curve->g.y, cur_sz, curve->g.ndigits); + fill_curve_param(n, curve->n, cur_sz, curve->g.ndigits); + + if (params->key_size == cur_sz && memcmp(params->key, n, cur_sz) >= 0) { + kfree(n); + return -EINVAL; + } + + kfree(n); + return 0; +} + +static unsigned int hpre_ecdh_get_curvesz(unsigned short id) +{ + switch (id) { + case ECC_CURVE_NIST_P192: + return HPRE_ECC_NIST_P192_N_SIZE; + case ECC_CURVE_NIST_P256: + return HPRE_ECC_NIST_P256_N_SIZE; + case ECC_CURVE_NIST_P384: + return HPRE_ECC_NIST_P384_N_SIZE; + default: + break; + } + + return 0; +} + +static int hpre_ecdh_set_param(struct hpre_ctx *ctx, struct ecdh *params) +{ + struct device *dev = ctx->dev; + unsigned int sz, shift, curve_sz; + int ret; + + ctx->key_sz = hpre_ecdh_supported_curve(ctx->curve_id); + if (!ctx->key_sz) + return -EINVAL; + + curve_sz = hpre_ecdh_get_curvesz(ctx->curve_id); + if (!curve_sz || params->key_size > curve_sz) + return -EINVAL; + + sz = ctx->key_sz; + + if (!ctx->ecdh.p) { + ctx->ecdh.p = dma_alloc_coherent(dev, sz << 3, &ctx->ecdh.dma_p, + GFP_KERNEL); + if (!ctx->ecdh.p) + return -ENOMEM; + } + + shift = sz << 2; + ctx->ecdh.g = ctx->ecdh.p + shift; + ctx->ecdh.dma_g = ctx->ecdh.dma_p + shift; + + ret = hpre_ecdh_fill_curve(ctx, params, curve_sz); + if (ret) { + dev_err(dev, "failed to fill curve_param, ret = %d!\n", ret); + dma_free_coherent(dev, sz << 3, ctx->ecdh.p, ctx->ecdh.dma_p); + ctx->ecdh.p = NULL; + return ret; + } + + return 0; +} + +static bool hpre_key_is_zero(char *key, unsigned short key_sz) +{ + int i; + + for (i = 0; i < key_sz; i++) + if (key[i]) + return false; + + return true; +} + +static int ecdh_gen_privkey(struct hpre_ctx *ctx, struct ecdh *params) +{ + struct device *dev = ctx->dev; + int ret; + + ret = crypto_get_default_rng(); + if (ret) { + dev_err(dev, "failed to get default rng, ret = %d!\n", ret); + return ret; + } + + ret = crypto_rng_get_bytes(crypto_default_rng, (u8 *)params->key, + params->key_size); + crypto_put_default_rng(); + if (ret) + dev_err(dev, "failed to get rng, ret = %d!\n", ret); + + return ret; +} + +static int hpre_ecdh_set_secret(struct crypto_kpp *tfm, const void *buf, + unsigned int len) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + char key[HPRE_ECC_MAX_KSZ]; + unsigned int sz, sz_shift; + struct ecdh params; + int ret; + + if (crypto_ecdh_decode_key(buf, len, ¶ms) < 0) { + dev_err(dev, "failed to decode ecdh key!\n"); + return -EINVAL; + } + + /* Use stdrng to generate private key */ + if (!params.key || !params.key_size) { + params.key = key; + params.key_size = hpre_ecdh_get_curvesz(ctx->curve_id); + ret = ecdh_gen_privkey(ctx, ¶ms); + if (ret) + return ret; + } + + if (hpre_key_is_zero(params.key, params.key_size)) { + dev_err(dev, "Invalid hpre key!\n"); + return -EINVAL; + } + + hpre_ecc_clear_ctx(ctx, false, true); + + ret = hpre_ecdh_set_param(ctx, ¶ms); + if (ret < 0) { + dev_err(dev, "failed to set hpre param, ret = %d!\n", ret); + return ret; + } + + sz = ctx->key_sz; + sz_shift = (sz << 1) + sz - params.key_size; + memcpy(ctx->ecdh.p + sz_shift, params.key, params.key_size); + + return 0; +} + +static void hpre_ecdh_hw_data_clr_all(struct hpre_ctx *ctx, + struct hpre_asym_request *req, + struct scatterlist *dst, + struct scatterlist *src) +{ + struct device *dev = ctx->dev; + struct hpre_sqe *sqe = &req->req; + dma_addr_t dma; + + dma = le64_to_cpu(sqe->in); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (src && req->src) + dma_free_coherent(dev, ctx->key_sz << 2, req->src, dma); + + dma = le64_to_cpu(sqe->out); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (req->dst) + dma_free_coherent(dev, ctx->key_sz << 1, req->dst, dma); + if (dst) + dma_unmap_single(dev, dma, ctx->key_sz << 1, DMA_FROM_DEVICE); +} + +static void hpre_ecdh_cb(struct hpre_ctx *ctx, void *resp) +{ + unsigned int curve_sz = hpre_ecdh_get_curvesz(ctx->curve_id); + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req = NULL; + struct kpp_request *areq; + u64 overtime_thrhld; + char *p; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + areq = req->areq.ecdh; + areq->dst_len = ctx->key_sz << 1; + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + p = sg_virt(areq->dst); + memmove(p, p + ctx->key_sz - curve_sz, curve_sz); + memmove(p + curve_sz, p + areq->dst_len - curve_sz, curve_sz); + + hpre_ecdh_hw_data_clr_all(ctx, req, areq->dst, areq->src); + kpp_request_complete(areq, ret); + + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static int hpre_ecdh_msg_request_set(struct hpre_ctx *ctx, + struct kpp_request *req) +{ + struct hpre_asym_request *h_req; + struct hpre_sqe *msg; + int req_id; + void *tmp; + + if (req->dst_len < ctx->key_sz << 1) { + req->dst_len = ctx->key_sz << 1; + return -EINVAL; + } + + tmp = kpp_request_ctx(req); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_ecdh_cb; + h_req->areq.ecdh = req; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + msg->in = cpu_to_le64(DMA_MAPPING_ERROR); + msg->out = cpu_to_le64(DMA_MAPPING_ERROR); + msg->key = cpu_to_le64(ctx->ecdh.dma_p); + + msg->dw0 |= cpu_to_le32(0x1U << HPRE_SQE_DONE_SHIFT); + msg->task_len1 = (ctx->key_sz >> HPRE_BITS_2_BYTES_SHIFT) - 1; + h_req->ctx = ctx; + + req_id = hpre_add_req_to_ctx(h_req); + if (req_id < 0) + return -EBUSY; + + msg->tag = cpu_to_le16((u16)req_id); + return 0; +} + +static int hpre_ecdh_src_data_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + unsigned int tmpshift; + dma_addr_t dma = 0; + void *ptr; + int shift; + + /* Src_data include gx and gy. */ + shift = ctx->key_sz - (len >> 1); + if (unlikely(shift < 0)) + return -EINVAL; + + ptr = dma_alloc_coherent(dev, ctx->key_sz << 2, &dma, GFP_KERNEL); + if (unlikely(!ptr)) + return -ENOMEM; + + tmpshift = ctx->key_sz << 1; + scatterwalk_map_and_copy(ptr + tmpshift, data, 0, len, 0); + memcpy(ptr + shift, ptr + tmpshift, len >> 1); + memcpy(ptr + ctx->key_sz + shift, ptr + tmpshift + (len >> 1), len >> 1); + + hpre_req->src = ptr; + msg->in = cpu_to_le64(dma); + return 0; +} + +static int hpre_ecdh_dst_data_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + dma_addr_t dma; + + if (unlikely(!data || !sg_is_last(data) || len != ctx->key_sz << 1)) { + dev_err(dev, "data or data length is illegal!\n"); + return -EINVAL; + } + + hpre_req->dst = NULL; + dma = dma_map_single(dev, sg_virt(data), len, DMA_FROM_DEVICE); + if (unlikely(dma_mapping_error(dev, dma))) { + dev_err(dev, "dma map data err!\n"); + return -ENOMEM; + } + + msg->out = cpu_to_le64(dma); + return 0; +} + +static int hpre_ecdh_compute_value(struct kpp_request *req) +{ + struct crypto_kpp *tfm = crypto_kpp_reqtfm(req); + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + void *tmp = kpp_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + ret = hpre_ecdh_msg_request_set(ctx, req); + if (unlikely(ret)) { + dev_err(dev, "failed to set ecdh request, ret = %d!\n", ret); + return ret; + } + + if (req->src) { + ret = hpre_ecdh_src_data_init(hpre_req, req->src, req->src_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init src data, ret = %d!\n", ret); + goto clear_all; + } + } else { + msg->in = cpu_to_le64(ctx->ecdh.dma_g); + } + + ret = hpre_ecdh_dst_data_init(hpre_req, req->dst, req->dst_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init dst data, ret = %d!\n", ret); + goto clear_all; + } + + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_ECC_MUL); + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_ecdh_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + return ret; +} + +static unsigned int hpre_ecdh_max_size(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + /* max size is the pub_key_size, include x and y */ + return ctx->key_sz << 1; +} + +static int hpre_ecdh_nist_p192_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + ctx->curve_id = ECC_CURVE_NIST_P192; + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static int hpre_ecdh_nist_p256_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + ctx->curve_id = ECC_CURVE_NIST_P256; + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static int hpre_ecdh_nist_p384_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + ctx->curve_id = ECC_CURVE_NIST_P384; + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static void hpre_ecdh_exit_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + hpre_ecc_clear_ctx(ctx, true, true); +} + +static void hpre_curve25519_fill_curve(struct hpre_ctx *ctx, const void *buf, + unsigned int len) +{ + u8 secret[CURVE25519_KEY_SIZE] = { 0 }; + unsigned int sz = ctx->key_sz; + const struct ecc_curve *curve; + unsigned int shift = sz << 1; + void *p; + + /* + * The key from 'buf' is in little-endian, we should preprocess it as + * the description in rfc7748: "k[0] &= 248, k[31] &= 127, k[31] |= 64", + * then convert it to big endian. Only in this way, the result can be + * the same as the software curve-25519 that exists in crypto. + */ + memcpy(secret, buf, len); + curve25519_clamp_secret(secret); + hpre_key_to_big_end(secret, CURVE25519_KEY_SIZE); + + p = ctx->curve25519.p + sz - len; + + curve = ecc_get_curve25519(); + + /* fill curve parameters */ + fill_curve_param(p, curve->p, len, curve->g.ndigits); + fill_curve_param(p + sz, curve->a, len, curve->g.ndigits); + memcpy(p + shift, secret, len); + fill_curve_param(p + shift + sz, curve->g.x, len, curve->g.ndigits); + memzero_explicit(secret, CURVE25519_KEY_SIZE); +} + +static int hpre_curve25519_set_param(struct hpre_ctx *ctx, const void *buf, + unsigned int len) +{ + struct device *dev = ctx->dev; + unsigned int sz = ctx->key_sz; + unsigned int shift = sz << 1; + + /* p->a->k->gx */ + if (!ctx->curve25519.p) { + ctx->curve25519.p = dma_alloc_coherent(dev, sz << 2, + &ctx->curve25519.dma_p, + GFP_KERNEL); + if (!ctx->curve25519.p) + return -ENOMEM; + } + + ctx->curve25519.g = ctx->curve25519.p + shift + sz; + ctx->curve25519.dma_g = ctx->curve25519.dma_p + shift + sz; + + hpre_curve25519_fill_curve(ctx, buf, len); + + return 0; +} + +static int hpre_curve25519_set_secret(struct crypto_kpp *tfm, const void *buf, + unsigned int len) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + int ret = -EINVAL; + + if (len != CURVE25519_KEY_SIZE || + !crypto_memneq(buf, curve25519_null_point, CURVE25519_KEY_SIZE)) { + dev_err(dev, "key is null or key len is not 32bytes!\n"); + return ret; + } + + /* Free old secret if any */ + hpre_ecc_clear_ctx(ctx, false, false); + + ctx->key_sz = CURVE25519_KEY_SIZE; + ret = hpre_curve25519_set_param(ctx, buf, CURVE25519_KEY_SIZE); + if (ret) { + dev_err(dev, "failed to set curve25519 param, ret = %d!\n", ret); + hpre_ecc_clear_ctx(ctx, false, false); + return ret; + } + + return 0; +} + +static void hpre_curve25519_hw_data_clr_all(struct hpre_ctx *ctx, + struct hpre_asym_request *req, + struct scatterlist *dst, + struct scatterlist *src) +{ + struct device *dev = ctx->dev; + struct hpre_sqe *sqe = &req->req; + dma_addr_t dma; + + dma = le64_to_cpu(sqe->in); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (src && req->src) + dma_free_coherent(dev, ctx->key_sz, req->src, dma); + + dma = le64_to_cpu(sqe->out); + if (unlikely(dma_mapping_error(dev, dma))) + return; + + if (req->dst) + dma_free_coherent(dev, ctx->key_sz, req->dst, dma); + if (dst) + dma_unmap_single(dev, dma, ctx->key_sz, DMA_FROM_DEVICE); +} + +static void hpre_curve25519_cb(struct hpre_ctx *ctx, void *resp) +{ + struct hpre_dfx *dfx = ctx->hpre->debug.dfx; + struct hpre_asym_request *req = NULL; + struct kpp_request *areq; + u64 overtime_thrhld; + int ret; + + ret = hpre_alg_res_post_hf(ctx, resp, (void **)&req); + areq = req->areq.curve25519; + areq->dst_len = ctx->key_sz; + + overtime_thrhld = atomic64_read(&dfx[HPRE_OVERTIME_THRHLD].value); + if (overtime_thrhld && hpre_is_bd_timeout(req, overtime_thrhld)) + atomic64_inc(&dfx[HPRE_OVER_THRHLD_CNT].value); + + hpre_key_to_big_end(sg_virt(areq->dst), CURVE25519_KEY_SIZE); + + hpre_curve25519_hw_data_clr_all(ctx, req, areq->dst, areq->src); + kpp_request_complete(areq, ret); + + atomic64_inc(&dfx[HPRE_RECV_CNT].value); +} + +static int hpre_curve25519_msg_request_set(struct hpre_ctx *ctx, + struct kpp_request *req) +{ + struct hpre_asym_request *h_req; + struct hpre_sqe *msg; + int req_id; + void *tmp; + + if (unlikely(req->dst_len < ctx->key_sz)) { + req->dst_len = ctx->key_sz; + return -EINVAL; + } + + tmp = kpp_request_ctx(req); + h_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + h_req->cb = hpre_curve25519_cb; + h_req->areq.curve25519 = req; + msg = &h_req->req; + memset(msg, 0, sizeof(*msg)); + msg->in = cpu_to_le64(DMA_MAPPING_ERROR); + msg->out = cpu_to_le64(DMA_MAPPING_ERROR); + msg->key = cpu_to_le64(ctx->curve25519.dma_p); + + msg->dw0 |= cpu_to_le32(0x1U << HPRE_SQE_DONE_SHIFT); + msg->task_len1 = (ctx->key_sz >> HPRE_BITS_2_BYTES_SHIFT) - 1; + h_req->ctx = ctx; + + req_id = hpre_add_req_to_ctx(h_req); + if (req_id < 0) + return -EBUSY; + + msg->tag = cpu_to_le16((u16)req_id); + return 0; +} + +static void hpre_curve25519_src_modulo_p(u8 *ptr) +{ + int i; + + for (i = 0; i < CURVE25519_KEY_SIZE - 1; i++) + ptr[i] = 0; + + /* The modulus is ptr's last byte minus '0xed'(last byte of p) */ + ptr[i] -= 0xed; +} + +static int hpre_curve25519_src_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + u8 p[CURVE25519_KEY_SIZE] = { 0 }; + const struct ecc_curve *curve; + dma_addr_t dma = 0; + u8 *ptr; + + if (len != CURVE25519_KEY_SIZE) { + dev_err(dev, "sourc_data len is not 32bytes, len = %u!\n", len); + return -EINVAL; + } + + ptr = dma_alloc_coherent(dev, ctx->key_sz, &dma, GFP_KERNEL); + if (unlikely(!ptr)) + return -ENOMEM; + + scatterwalk_map_and_copy(ptr, data, 0, len, 0); + + if (!crypto_memneq(ptr, curve25519_null_point, CURVE25519_KEY_SIZE)) { + dev_err(dev, "gx is null!\n"); + goto err; + } + + /* + * Src_data(gx) is in little-endian order, MSB in the final byte should + * be masked as described in RFC7748, then transform it to big-endian + * form, then hisi_hpre can use the data. + */ + ptr[31] &= 0x7f; + hpre_key_to_big_end(ptr, CURVE25519_KEY_SIZE); + + curve = ecc_get_curve25519(); + + fill_curve_param(p, curve->p, CURVE25519_KEY_SIZE, curve->g.ndigits); + + /* + * When src_data equals (2^255 - 19) ~ (2^255 - 1), it is out of p, + * we get its modulus to p, and then use it. + */ + if (memcmp(ptr, p, ctx->key_sz) == 0) { + dev_err(dev, "gx is p!\n"); + goto err; + } else if (memcmp(ptr, p, ctx->key_sz) > 0) { + hpre_curve25519_src_modulo_p(ptr); + } + + hpre_req->src = ptr; + msg->in = cpu_to_le64(dma); + return 0; + +err: + dma_free_coherent(dev, ctx->key_sz, ptr, dma); + return -EINVAL; +} + +static int hpre_curve25519_dst_init(struct hpre_asym_request *hpre_req, + struct scatterlist *data, unsigned int len) +{ + struct hpre_sqe *msg = &hpre_req->req; + struct hpre_ctx *ctx = hpre_req->ctx; + struct device *dev = ctx->dev; + dma_addr_t dma; + + if (!data || !sg_is_last(data) || len != ctx->key_sz) { + dev_err(dev, "data or data length is illegal!\n"); + return -EINVAL; + } + + hpre_req->dst = NULL; + dma = dma_map_single(dev, sg_virt(data), len, DMA_FROM_DEVICE); + if (unlikely(dma_mapping_error(dev, dma))) { + dev_err(dev, "dma map data err!\n"); + return -ENOMEM; + } + + msg->out = cpu_to_le64(dma); + return 0; +} + +static int hpre_curve25519_compute_value(struct kpp_request *req) +{ + struct crypto_kpp *tfm = crypto_kpp_reqtfm(req); + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + struct device *dev = ctx->dev; + void *tmp = kpp_request_ctx(req); + struct hpre_asym_request *hpre_req = PTR_ALIGN(tmp, HPRE_ALIGN_SZ); + struct hpre_sqe *msg = &hpre_req->req; + int ret; + + ret = hpre_curve25519_msg_request_set(ctx, req); + if (unlikely(ret)) { + dev_err(dev, "failed to set curve25519 request, ret = %d!\n", ret); + return ret; + } + + if (req->src) { + ret = hpre_curve25519_src_init(hpre_req, req->src, req->src_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init src data, ret = %d!\n", + ret); + goto clear_all; + } + } else { + msg->in = cpu_to_le64(ctx->curve25519.dma_g); + } + + ret = hpre_curve25519_dst_init(hpre_req, req->dst, req->dst_len); + if (unlikely(ret)) { + dev_err(dev, "failed to init dst data, ret = %d!\n", ret); + goto clear_all; + } + + msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_CURVE25519_MUL); + ret = hpre_send(ctx, msg); + if (likely(!ret)) + return -EINPROGRESS; + +clear_all: + hpre_rm_req_from_ctx(hpre_req); + hpre_curve25519_hw_data_clr_all(ctx, hpre_req, req->dst, req->src); + return ret; +} + +static unsigned int hpre_curve25519_max_size(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return ctx->key_sz; +} + +static int hpre_curve25519_init_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + return hpre_ctx_init(ctx, HPRE_V3_ECC_ALG_TYPE); +} + +static void hpre_curve25519_exit_tfm(struct crypto_kpp *tfm) +{ + struct hpre_ctx *ctx = kpp_tfm_ctx(tfm); + + hpre_ecc_clear_ctx(ctx, true, false); +} + +static struct akcipher_alg rsa = { + .sign = hpre_rsa_dec, + .verify = hpre_rsa_enc, + .encrypt = hpre_rsa_enc, + .decrypt = hpre_rsa_dec, + .set_pub_key = hpre_rsa_setpubkey, + .set_priv_key = hpre_rsa_setprivkey, + .max_size = hpre_rsa_max_size, + .init = hpre_rsa_init_tfm, + .exit = hpre_rsa_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "rsa", + .cra_driver_name = "hpre-rsa", + .cra_module = THIS_MODULE, + }, +}; + +static struct kpp_alg dh = { + .set_secret = hpre_dh_set_secret, + .generate_public_key = hpre_dh_compute_value, + .compute_shared_secret = hpre_dh_compute_value, + .max_size = hpre_dh_max_size, + .init = hpre_dh_init_tfm, + .exit = hpre_dh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "dh", + .cra_driver_name = "hpre-dh", + .cra_module = THIS_MODULE, + }, +}; + +static struct kpp_alg ecdh_curves[] = { + { + .set_secret = hpre_ecdh_set_secret, + .generate_public_key = hpre_ecdh_compute_value, + .compute_shared_secret = hpre_ecdh_compute_value, + .max_size = hpre_ecdh_max_size, + .init = hpre_ecdh_nist_p192_init_tfm, + .exit = hpre_ecdh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "ecdh-nist-p192", + .cra_driver_name = "hpre-ecdh-nist-p192", + .cra_module = THIS_MODULE, + }, + }, { + .set_secret = hpre_ecdh_set_secret, + .generate_public_key = hpre_ecdh_compute_value, + .compute_shared_secret = hpre_ecdh_compute_value, + .max_size = hpre_ecdh_max_size, + .init = hpre_ecdh_nist_p256_init_tfm, + .exit = hpre_ecdh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "ecdh-nist-p256", + .cra_driver_name = "hpre-ecdh-nist-p256", + .cra_module = THIS_MODULE, + }, + }, { + .set_secret = hpre_ecdh_set_secret, + .generate_public_key = hpre_ecdh_compute_value, + .compute_shared_secret = hpre_ecdh_compute_value, + .max_size = hpre_ecdh_max_size, + .init = hpre_ecdh_nist_p384_init_tfm, + .exit = hpre_ecdh_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "ecdh-nist-p384", + .cra_driver_name = "hpre-ecdh-nist-p384", + .cra_module = THIS_MODULE, + }, + } +}; + +static struct kpp_alg curve25519_alg = { + .set_secret = hpre_curve25519_set_secret, + .generate_public_key = hpre_curve25519_compute_value, + .compute_shared_secret = hpre_curve25519_compute_value, + .max_size = hpre_curve25519_max_size, + .init = hpre_curve25519_init_tfm, + .exit = hpre_curve25519_exit_tfm, + .reqsize = sizeof(struct hpre_asym_request) + HPRE_ALIGN_SZ, + .base = { + .cra_ctxsize = sizeof(struct hpre_ctx), + .cra_priority = HPRE_CRYPTO_ALG_PRI, + .cra_name = "curve25519", + .cra_driver_name = "hpre-curve25519", + .cra_module = THIS_MODULE, + }, +}; + +static int hpre_register_rsa(struct hisi_qm *qm) +{ + int ret; + + if (!hpre_check_alg_support(qm, HPRE_DRV_RSA_MASK_CAP)) + return 0; + + rsa.base.cra_flags = 0; + ret = crypto_register_akcipher(&rsa); + if (ret) + dev_err(&qm->pdev->dev, "failed to register rsa (%d)!\n", ret); + + return ret; +} + +static void hpre_unregister_rsa(struct hisi_qm *qm) +{ + if (!hpre_check_alg_support(qm, HPRE_DRV_RSA_MASK_CAP)) + return; + + crypto_unregister_akcipher(&rsa); +} + +static int hpre_register_dh(struct hisi_qm *qm) +{ + int ret; + + if (!hpre_check_alg_support(qm, HPRE_DRV_DH_MASK_CAP)) + return 0; + + ret = crypto_register_kpp(&dh); + if (ret) + dev_err(&qm->pdev->dev, "failed to register dh (%d)!\n", ret); + + return ret; +} + +static void hpre_unregister_dh(struct hisi_qm *qm) +{ + if (!hpre_check_alg_support(qm, HPRE_DRV_DH_MASK_CAP)) + return; + + crypto_unregister_kpp(&dh); +} + +static int hpre_register_ecdh(struct hisi_qm *qm) +{ + int ret, i; + + if (!hpre_check_alg_support(qm, HPRE_DRV_ECDH_MASK_CAP)) + return 0; + + for (i = 0; i < ARRAY_SIZE(ecdh_curves); i++) { + ret = crypto_register_kpp(&ecdh_curves[i]); + if (ret) { + dev_err(&qm->pdev->dev, "failed to register %s (%d)!\n", + ecdh_curves[i].base.cra_name, ret); + goto unreg_kpp; + } + } + + return 0; + +unreg_kpp: + for (--i; i >= 0; --i) + crypto_unregister_kpp(&ecdh_curves[i]); + + return ret; +} + +static void hpre_unregister_ecdh(struct hisi_qm *qm) +{ + int i; + + if (!hpre_check_alg_support(qm, HPRE_DRV_ECDH_MASK_CAP)) + return; + + for (i = ARRAY_SIZE(ecdh_curves) - 1; i >= 0; --i) + crypto_unregister_kpp(&ecdh_curves[i]); +} + +static int hpre_register_x25519(struct hisi_qm *qm) +{ + int ret; + + if (!hpre_check_alg_support(qm, HPRE_DRV_X25519_MASK_CAP)) + return 0; + + ret = crypto_register_kpp(&curve25519_alg); + if (ret) + dev_err(&qm->pdev->dev, "failed to register x25519 (%d)!\n", ret); + + return ret; +} + +static void hpre_unregister_x25519(struct hisi_qm *qm) +{ + if (!hpre_check_alg_support(qm, HPRE_DRV_X25519_MASK_CAP)) + return; + + crypto_unregister_kpp(&curve25519_alg); +} + +int hpre_algs_register(struct hisi_qm *qm) +{ + int ret; + + ret = hpre_register_rsa(qm); + if (ret) + return ret; + + ret = hpre_register_dh(qm); + if (ret) + goto unreg_rsa; + + ret = hpre_register_ecdh(qm); + if (ret) + goto unreg_dh; + + ret = hpre_register_x25519(qm); + if (ret) + goto unreg_ecdh; + + return ret; + +unreg_ecdh: + hpre_unregister_ecdh(qm); +unreg_dh: + hpre_unregister_dh(qm); +unreg_rsa: + hpre_unregister_rsa(qm); + return ret; +} + +void hpre_algs_unregister(struct hisi_qm *qm) +{ + hpre_unregister_x25519(qm); + hpre_unregister_ecdh(qm); + hpre_unregister_dh(qm); + hpre_unregister_rsa(qm); +} diff --git a/drivers/crypto/hisilicon/hpre/hpre_main.c b/drivers/crypto/hisilicon/hpre/hpre_main.c new file mode 100644 index 000000000..269df4ec1 --- /dev/null +++ b/drivers/crypto/hisilicon/hpre/hpre_main.c @@ -0,0 +1,1546 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2018-2019 HiSilicon Limited. */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "hpre.h" + +#define HPRE_QM_ABNML_INT_MASK 0x100004 +#define HPRE_CTRL_CNT_CLR_CE_BIT BIT(0) +#define HPRE_COMM_CNT_CLR_CE 0x0 +#define HPRE_CTRL_CNT_CLR_CE 0x301000 +#define HPRE_FSM_MAX_CNT 0x301008 +#define HPRE_VFG_AXQOS 0x30100c +#define HPRE_VFG_AXCACHE 0x301010 +#define HPRE_RDCHN_INI_CFG 0x301014 +#define HPRE_AWUSR_FP_CFG 0x301018 +#define HPRE_BD_ENDIAN 0x301020 +#define HPRE_ECC_BYPASS 0x301024 +#define HPRE_RAS_WIDTH_CFG 0x301028 +#define HPRE_POISON_BYPASS 0x30102c +#define HPRE_BD_ARUSR_CFG 0x301030 +#define HPRE_BD_AWUSR_CFG 0x301034 +#define HPRE_TYPES_ENB 0x301038 +#define HPRE_RSA_ENB BIT(0) +#define HPRE_ECC_ENB BIT(1) +#define HPRE_DATA_RUSER_CFG 0x30103c +#define HPRE_DATA_WUSER_CFG 0x301040 +#define HPRE_INT_MASK 0x301400 +#define HPRE_INT_STATUS 0x301800 +#define HPRE_HAC_INT_MSK 0x301400 +#define HPRE_HAC_RAS_CE_ENB 0x301410 +#define HPRE_HAC_RAS_NFE_ENB 0x301414 +#define HPRE_HAC_RAS_FE_ENB 0x301418 +#define HPRE_HAC_INT_SET 0x301500 +#define HPRE_RNG_TIMEOUT_NUM 0x301A34 +#define HPRE_CORE_INT_ENABLE 0 +#define HPRE_CORE_INT_DISABLE GENMASK(21, 0) +#define HPRE_RDCHN_INI_ST 0x301a00 +#define HPRE_CLSTR_BASE 0x302000 +#define HPRE_CORE_EN_OFFSET 0x04 +#define HPRE_CORE_INI_CFG_OFFSET 0x20 +#define HPRE_CORE_INI_STATUS_OFFSET 0x80 +#define HPRE_CORE_HTBT_WARN_OFFSET 0x8c +#define HPRE_CORE_IS_SCHD_OFFSET 0x90 + +#define HPRE_RAS_CE_ENB 0x301410 +#define HPRE_RAS_NFE_ENB 0x301414 +#define HPRE_RAS_FE_ENB 0x301418 +#define HPRE_OOO_SHUTDOWN_SEL 0x301a3c +#define HPRE_HAC_RAS_FE_ENABLE 0 + +#define HPRE_CORE_ENB (HPRE_CLSTR_BASE + HPRE_CORE_EN_OFFSET) +#define HPRE_CORE_INI_CFG (HPRE_CLSTR_BASE + HPRE_CORE_INI_CFG_OFFSET) +#define HPRE_CORE_INI_STATUS (HPRE_CLSTR_BASE + HPRE_CORE_INI_STATUS_OFFSET) +#define HPRE_HAC_ECC1_CNT 0x301a04 +#define HPRE_HAC_ECC2_CNT 0x301a08 +#define HPRE_HAC_SOURCE_INT 0x301600 +#define HPRE_CLSTR_ADDR_INTRVL 0x1000 +#define HPRE_CLUSTER_INQURY 0x100 +#define HPRE_CLSTR_ADDR_INQRY_RSLT 0x104 +#define HPRE_TIMEOUT_ABNML_BIT 6 +#define HPRE_PASID_EN_BIT 9 +#define HPRE_REG_RD_INTVRL_US 10 +#define HPRE_REG_RD_TMOUT_US 1000 +#define HPRE_DBGFS_VAL_MAX_LEN 20 +#define PCI_DEVICE_ID_HUAWEI_HPRE_PF 0xa258 +#define HPRE_QM_USR_CFG_MASK GENMASK(31, 1) +#define HPRE_QM_AXI_CFG_MASK GENMASK(15, 0) +#define HPRE_QM_VFG_AX_MASK GENMASK(7, 0) +#define HPRE_BD_USR_MASK GENMASK(1, 0) +#define HPRE_PREFETCH_CFG 0x301130 +#define HPRE_SVA_PREFTCH_DFX 0x30115C +#define HPRE_PREFETCH_ENABLE (~(BIT(0) | BIT(30))) +#define HPRE_PREFETCH_DISABLE BIT(30) +#define HPRE_SVA_DISABLE_READY (BIT(4) | BIT(8)) + +/* clock gate */ +#define HPRE_CLKGATE_CTL 0x301a10 +#define HPRE_PEH_CFG_AUTO_GATE 0x301a2c +#define HPRE_CLUSTER_DYN_CTL 0x302010 +#define HPRE_CORE_SHB_CFG 0x302088 +#define HPRE_CLKGATE_CTL_EN BIT(0) +#define HPRE_PEH_CFG_AUTO_GATE_EN BIT(0) +#define HPRE_CLUSTER_DYN_CTL_EN BIT(0) +#define HPRE_CORE_GATE_EN (BIT(30) | BIT(31)) + +#define HPRE_AM_OOO_SHUTDOWN_ENB 0x301044 +#define HPRE_AM_OOO_SHUTDOWN_ENABLE BIT(0) +#define HPRE_WR_MSI_PORT BIT(2) + +#define HPRE_CORE_ECC_2BIT_ERR BIT(1) +#define HPRE_OOO_ECC_2BIT_ERR BIT(5) + +#define HPRE_QM_BME_FLR BIT(7) +#define HPRE_QM_PM_FLR BIT(11) +#define HPRE_QM_SRIOV_FLR BIT(12) + +#define HPRE_SHAPER_TYPE_RATE 640 +#define HPRE_VIA_MSI_DSM 1 +#define HPRE_SQE_MASK_OFFSET 8 +#define HPRE_SQE_MASK_LEN 24 + +#define HPRE_DFX_BASE 0x301000 +#define HPRE_DFX_COMMON1 0x301400 +#define HPRE_DFX_COMMON2 0x301A00 +#define HPRE_DFX_CORE 0x302000 +#define HPRE_DFX_BASE_LEN 0x55 +#define HPRE_DFX_COMMON1_LEN 0x41 +#define HPRE_DFX_COMMON2_LEN 0xE +#define HPRE_DFX_CORE_LEN 0x43 + +static const char hpre_name[] = "hisi_hpre"; +static struct dentry *hpre_debugfs_root; +static const struct pci_device_id hpre_dev_ids[] = { + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_HPRE_PF) }, + { PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, PCI_DEVICE_ID_HUAWEI_HPRE_VF) }, + { 0, } +}; + +MODULE_DEVICE_TABLE(pci, hpre_dev_ids); + +struct hpre_hw_error { + u32 int_msk; + const char *msg; +}; + +static const struct qm_dev_alg hpre_dev_algs[] = { + { + .alg_msk = BIT(0), + .alg = "rsa\n" + }, { + .alg_msk = BIT(1), + .alg = "dh\n" + }, { + .alg_msk = BIT(2), + .alg = "ecdh\n" + }, { + .alg_msk = BIT(3), + .alg = "ecdsa\n" + }, { + .alg_msk = BIT(4), + .alg = "sm2\n" + }, { + .alg_msk = BIT(5), + .alg = "x25519\n" + }, { + .alg_msk = BIT(6), + .alg = "x448\n" + }, { + /* sentinel */ + } +}; + +static struct hisi_qm_list hpre_devices = { + .register_to_crypto = hpre_algs_register, + .unregister_from_crypto = hpre_algs_unregister, +}; + +static const char * const hpre_debug_file_name[] = { + [HPRE_CLEAR_ENABLE] = "rdclr_en", + [HPRE_CLUSTER_CTRL] = "cluster_ctrl", +}; + +enum hpre_cap_type { + HPRE_QM_NFE_MASK_CAP, + HPRE_QM_RESET_MASK_CAP, + HPRE_QM_OOO_SHUTDOWN_MASK_CAP, + HPRE_QM_CE_MASK_CAP, + HPRE_NFE_MASK_CAP, + HPRE_RESET_MASK_CAP, + HPRE_OOO_SHUTDOWN_MASK_CAP, + HPRE_CE_MASK_CAP, + HPRE_CLUSTER_NUM_CAP, + HPRE_CORE_TYPE_NUM_CAP, + HPRE_CORE_NUM_CAP, + HPRE_CLUSTER_CORE_NUM_CAP, + HPRE_CORE_ENABLE_BITMAP_CAP, + HPRE_DRV_ALG_BITMAP_CAP, + HPRE_DEV_ALG_BITMAP_CAP, + HPRE_CORE1_ALG_BITMAP_CAP, + HPRE_CORE2_ALG_BITMAP_CAP, + HPRE_CORE3_ALG_BITMAP_CAP, + HPRE_CORE4_ALG_BITMAP_CAP, + HPRE_CORE5_ALG_BITMAP_CAP, + HPRE_CORE6_ALG_BITMAP_CAP, + HPRE_CORE7_ALG_BITMAP_CAP, + HPRE_CORE8_ALG_BITMAP_CAP, + HPRE_CORE9_ALG_BITMAP_CAP, + HPRE_CORE10_ALG_BITMAP_CAP +}; + +static const struct hisi_qm_cap_info hpre_basic_info[] = { + {HPRE_QM_NFE_MASK_CAP, 0x3124, 0, GENMASK(31, 0), 0x0, 0x1C37, 0x7C37}, + {HPRE_QM_RESET_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0xC37, 0x6C37}, + {HPRE_QM_OOO_SHUTDOWN_MASK_CAP, 0x3128, 0, GENMASK(31, 0), 0x0, 0x4, 0x6C37}, + {HPRE_QM_CE_MASK_CAP, 0x312C, 0, GENMASK(31, 0), 0x0, 0x8, 0x8}, + {HPRE_NFE_MASK_CAP, 0x3130, 0, GENMASK(31, 0), 0x0, 0x3FFFFE, 0xFFFFFE}, + {HPRE_RESET_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x3FFFFE, 0xBFFFFE}, + {HPRE_OOO_SHUTDOWN_MASK_CAP, 0x3134, 0, GENMASK(31, 0), 0x0, 0x22, 0xBFFFFE}, + {HPRE_CE_MASK_CAP, 0x3138, 0, GENMASK(31, 0), 0x0, 0x1, 0x1}, + {HPRE_CLUSTER_NUM_CAP, 0x313c, 20, GENMASK(3, 0), 0x0, 0x4, 0x1}, + {HPRE_CORE_TYPE_NUM_CAP, 0x313c, 16, GENMASK(3, 0), 0x0, 0x2, 0x2}, + {HPRE_CORE_NUM_CAP, 0x313c, 8, GENMASK(7, 0), 0x0, 0x8, 0xA}, + {HPRE_CLUSTER_CORE_NUM_CAP, 0x313c, 0, GENMASK(7, 0), 0x0, 0x2, 0xA}, + {HPRE_CORE_ENABLE_BITMAP_CAP, 0x3140, 0, GENMASK(31, 0), 0x0, 0xF, 0x3FF}, + {HPRE_DRV_ALG_BITMAP_CAP, 0x3144, 0, GENMASK(31, 0), 0x0, 0x03, 0x27}, + {HPRE_DEV_ALG_BITMAP_CAP, 0x3148, 0, GENMASK(31, 0), 0x0, 0x03, 0x7F}, + {HPRE_CORE1_ALG_BITMAP_CAP, 0x314c, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE2_ALG_BITMAP_CAP, 0x3150, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE3_ALG_BITMAP_CAP, 0x3154, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE4_ALG_BITMAP_CAP, 0x3158, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE5_ALG_BITMAP_CAP, 0x315c, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE6_ALG_BITMAP_CAP, 0x3160, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE7_ALG_BITMAP_CAP, 0x3164, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE8_ALG_BITMAP_CAP, 0x3168, 0, GENMASK(31, 0), 0x0, 0x7F, 0x7F}, + {HPRE_CORE9_ALG_BITMAP_CAP, 0x316c, 0, GENMASK(31, 0), 0x0, 0x10, 0x10}, + {HPRE_CORE10_ALG_BITMAP_CAP, 0x3170, 0, GENMASK(31, 0), 0x0, 0x10, 0x10} +}; + +enum hpre_pre_store_cap_idx { + HPRE_CLUSTER_NUM_CAP_IDX = 0x0, + HPRE_CORE_ENABLE_BITMAP_CAP_IDX, + HPRE_DRV_ALG_BITMAP_CAP_IDX, + HPRE_DEV_ALG_BITMAP_CAP_IDX, +}; + +static const u32 hpre_pre_store_caps[] = { + HPRE_CLUSTER_NUM_CAP, + HPRE_CORE_ENABLE_BITMAP_CAP, + HPRE_DRV_ALG_BITMAP_CAP, + HPRE_DEV_ALG_BITMAP_CAP, +}; + +static const struct hpre_hw_error hpre_hw_errors[] = { + { + .int_msk = BIT(0), + .msg = "core_ecc_1bit_err_int_set" + }, { + .int_msk = BIT(1), + .msg = "core_ecc_2bit_err_int_set" + }, { + .int_msk = BIT(2), + .msg = "dat_wb_poison_int_set" + }, { + .int_msk = BIT(3), + .msg = "dat_rd_poison_int_set" + }, { + .int_msk = BIT(4), + .msg = "bd_rd_poison_int_set" + }, { + .int_msk = BIT(5), + .msg = "ooo_ecc_2bit_err_int_set" + }, { + .int_msk = BIT(6), + .msg = "cluster1_shb_timeout_int_set" + }, { + .int_msk = BIT(7), + .msg = "cluster2_shb_timeout_int_set" + }, { + .int_msk = BIT(8), + .msg = "cluster3_shb_timeout_int_set" + }, { + .int_msk = BIT(9), + .msg = "cluster4_shb_timeout_int_set" + }, { + .int_msk = GENMASK(15, 10), + .msg = "ooo_rdrsp_err_int_set" + }, { + .int_msk = GENMASK(21, 16), + .msg = "ooo_wrrsp_err_int_set" + }, { + .int_msk = BIT(22), + .msg = "pt_rng_timeout_int_set" + }, { + .int_msk = BIT(23), + .msg = "sva_fsm_timeout_int_set" + }, { + /* sentinel */ + } +}; + +static const u64 hpre_cluster_offsets[] = { + [HPRE_CLUSTER0] = + HPRE_CLSTR_BASE + HPRE_CLUSTER0 * HPRE_CLSTR_ADDR_INTRVL, + [HPRE_CLUSTER1] = + HPRE_CLSTR_BASE + HPRE_CLUSTER1 * HPRE_CLSTR_ADDR_INTRVL, + [HPRE_CLUSTER2] = + HPRE_CLSTR_BASE + HPRE_CLUSTER2 * HPRE_CLSTR_ADDR_INTRVL, + [HPRE_CLUSTER3] = + HPRE_CLSTR_BASE + HPRE_CLUSTER3 * HPRE_CLSTR_ADDR_INTRVL, +}; + +static const struct debugfs_reg32 hpre_cluster_dfx_regs[] = { + {"CORES_EN_STATUS ", HPRE_CORE_EN_OFFSET}, + {"CORES_INI_CFG ", HPRE_CORE_INI_CFG_OFFSET}, + {"CORES_INI_STATUS ", HPRE_CORE_INI_STATUS_OFFSET}, + {"CORES_HTBT_WARN ", HPRE_CORE_HTBT_WARN_OFFSET}, + {"CORES_IS_SCHD ", HPRE_CORE_IS_SCHD_OFFSET}, +}; + +static const struct debugfs_reg32 hpre_com_dfx_regs[] = { + {"READ_CLR_EN ", HPRE_CTRL_CNT_CLR_CE}, + {"AXQOS ", HPRE_VFG_AXQOS}, + {"AWUSR_CFG ", HPRE_AWUSR_FP_CFG}, + {"BD_ENDIAN ", HPRE_BD_ENDIAN}, + {"ECC_CHECK_CTRL ", HPRE_ECC_BYPASS}, + {"RAS_INT_WIDTH ", HPRE_RAS_WIDTH_CFG}, + {"POISON_BYPASS ", HPRE_POISON_BYPASS}, + {"BD_ARUSER ", HPRE_BD_ARUSR_CFG}, + {"BD_AWUSER ", HPRE_BD_AWUSR_CFG}, + {"DATA_ARUSER ", HPRE_DATA_RUSER_CFG}, + {"DATA_AWUSER ", HPRE_DATA_WUSER_CFG}, + {"INT_STATUS ", HPRE_INT_STATUS}, + {"INT_MASK ", HPRE_HAC_INT_MSK}, + {"RAS_CE_ENB ", HPRE_HAC_RAS_CE_ENB}, + {"RAS_NFE_ENB ", HPRE_HAC_RAS_NFE_ENB}, + {"RAS_FE_ENB ", HPRE_HAC_RAS_FE_ENB}, + {"INT_SET ", HPRE_HAC_INT_SET}, + {"RNG_TIMEOUT_NUM ", HPRE_RNG_TIMEOUT_NUM}, +}; + +static const char *hpre_dfx_files[HPRE_DFX_FILE_NUM] = { + "send_cnt", + "recv_cnt", + "send_fail_cnt", + "send_busy_cnt", + "over_thrhld_cnt", + "overtime_thrhld", + "invalid_req_cnt" +}; + +/* define the HPRE's dfx regs region and region length */ +static struct dfx_diff_registers hpre_diff_regs[] = { + { + .reg_offset = HPRE_DFX_BASE, + .reg_len = HPRE_DFX_BASE_LEN, + }, { + .reg_offset = HPRE_DFX_COMMON1, + .reg_len = HPRE_DFX_COMMON1_LEN, + }, { + .reg_offset = HPRE_DFX_COMMON2, + .reg_len = HPRE_DFX_COMMON2_LEN, + }, { + .reg_offset = HPRE_DFX_CORE, + .reg_len = HPRE_DFX_CORE_LEN, + }, +}; + +bool hpre_check_alg_support(struct hisi_qm *qm, u32 alg) +{ + u32 cap_val; + + cap_val = qm->cap_tables.dev_cap_table[HPRE_DRV_ALG_BITMAP_CAP_IDX].cap_val; + if (alg & cap_val) + return true; + + return false; +} + +static int hpre_diff_regs_show(struct seq_file *s, void *unused) +{ + struct hisi_qm *qm = s->private; + + hisi_qm_acc_diff_regs_dump(qm, s, qm->debug.acc_diff_regs, + ARRAY_SIZE(hpre_diff_regs)); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(hpre_diff_regs); + +static int hpre_com_regs_show(struct seq_file *s, void *unused) +{ + hisi_qm_regs_dump(s, s->private); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(hpre_com_regs); + +static int hpre_cluster_regs_show(struct seq_file *s, void *unused) +{ + hisi_qm_regs_dump(s, s->private); + + return 0; +} + +DEFINE_SHOW_ATTRIBUTE(hpre_cluster_regs); + +static const struct kernel_param_ops hpre_uacce_mode_ops = { + .set = uacce_mode_set, + .get = param_get_int, +}; + +/* + * uacce_mode = 0 means hpre only register to crypto, + * uacce_mode = 1 means hpre both register to crypto and uacce. + */ +static u32 uacce_mode = UACCE_MODE_NOUACCE; +module_param_cb(uacce_mode, &hpre_uacce_mode_ops, &uacce_mode, 0444); +MODULE_PARM_DESC(uacce_mode, UACCE_MODE_DESC); + +static bool pf_q_num_flag; +static int pf_q_num_set(const char *val, const struct kernel_param *kp) +{ + pf_q_num_flag = true; + + return q_num_set(val, kp, PCI_DEVICE_ID_HUAWEI_HPRE_PF); +} + +static const struct kernel_param_ops hpre_pf_q_num_ops = { + .set = pf_q_num_set, + .get = param_get_int, +}; + +static u32 pf_q_num = HPRE_PF_DEF_Q_NUM; +module_param_cb(pf_q_num, &hpre_pf_q_num_ops, &pf_q_num, 0444); +MODULE_PARM_DESC(pf_q_num, "Number of queues in PF of CS(2-1024)"); + +static const struct kernel_param_ops vfs_num_ops = { + .set = vfs_num_set, + .get = param_get_int, +}; + +static u32 vfs_num; +module_param_cb(vfs_num, &vfs_num_ops, &vfs_num, 0444); +MODULE_PARM_DESC(vfs_num, "Number of VFs to enable(1-63), 0(default)"); + +struct hisi_qp *hpre_create_qp(u8 type) +{ + int node = cpu_to_node(smp_processor_id()); + struct hisi_qp *qp = NULL; + int ret; + + if (type != HPRE_V2_ALG_TYPE && type != HPRE_V3_ECC_ALG_TYPE) + return NULL; + + /* + * type: 0 - RSA/DH. algorithm supported in V2, + * 1 - ECC algorithm in V3. + */ + ret = hisi_qm_alloc_qps_node(&hpre_devices, 1, type, node, &qp); + if (!ret) + return qp; + + return NULL; +} + +static void hpre_config_pasid(struct hisi_qm *qm) +{ + u32 val1, val2; + + if (qm->ver >= QM_HW_V3) + return; + + val1 = readl_relaxed(qm->io_base + HPRE_DATA_RUSER_CFG); + val2 = readl_relaxed(qm->io_base + HPRE_DATA_WUSER_CFG); + if (qm->use_sva) { + val1 |= BIT(HPRE_PASID_EN_BIT); + val2 |= BIT(HPRE_PASID_EN_BIT); + } else { + val1 &= ~BIT(HPRE_PASID_EN_BIT); + val2 &= ~BIT(HPRE_PASID_EN_BIT); + } + writel_relaxed(val1, qm->io_base + HPRE_DATA_RUSER_CFG); + writel_relaxed(val2, qm->io_base + HPRE_DATA_WUSER_CFG); +} + +static int hpre_cfg_by_dsm(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + union acpi_object *obj; + guid_t guid; + + if (guid_parse("b06b81ab-0134-4a45-9b0c-483447b95fa7", &guid)) { + dev_err(dev, "Hpre GUID failed\n"); + return -EINVAL; + } + + /* Switch over to MSI handling due to non-standard PCI implementation */ + obj = acpi_evaluate_dsm(ACPI_HANDLE(dev), &guid, + 0, HPRE_VIA_MSI_DSM, NULL); + if (!obj) { + dev_err(dev, "ACPI handle failed!\n"); + return -EIO; + } + + ACPI_FREE(obj); + + return 0; +} + +static int hpre_set_cluster(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + unsigned long offset; + u32 cluster_core_mask; + u8 clusters_num; + u32 val = 0; + int ret, i; + + cluster_core_mask = qm->cap_tables.dev_cap_table[HPRE_CORE_ENABLE_BITMAP_CAP_IDX].cap_val; + clusters_num = qm->cap_tables.dev_cap_table[HPRE_CLUSTER_NUM_CAP_IDX].cap_val; + for (i = 0; i < clusters_num; i++) { + offset = i * HPRE_CLSTR_ADDR_INTRVL; + + /* clusters initiating */ + writel(cluster_core_mask, + qm->io_base + offset + HPRE_CORE_ENB); + writel(0x1, qm->io_base + offset + HPRE_CORE_INI_CFG); + ret = readl_relaxed_poll_timeout(qm->io_base + offset + + HPRE_CORE_INI_STATUS, val, + ((val & cluster_core_mask) == + cluster_core_mask), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) { + dev_err(dev, + "cluster %d int st status timeout!\n", i); + return -ETIMEDOUT; + } + } + + return 0; +} + +/* + * For Kunpeng 920, we should disable FLR triggered by hardware (BME/PM/SRIOV). + * Or it may stay in D3 state when we bind and unbind hpre quickly, + * as it does FLR triggered by hardware. + */ +static void disable_flr_of_bme(struct hisi_qm *qm) +{ + u32 val; + + val = readl(qm->io_base + QM_PEH_AXUSER_CFG); + val &= ~(HPRE_QM_BME_FLR | HPRE_QM_SRIOV_FLR); + val |= HPRE_QM_PM_FLR; + writel(val, qm->io_base + QM_PEH_AXUSER_CFG); + writel(PEH_AXUSER_CFG_ENABLE, qm->io_base + QM_PEH_AXUSER_CFG_ENABLE); +} + +static void hpre_open_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + /* Enable prefetch */ + val = readl_relaxed(qm->io_base + HPRE_PREFETCH_CFG); + val &= HPRE_PREFETCH_ENABLE; + writel(val, qm->io_base + HPRE_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + HPRE_PREFETCH_CFG, + val, !(val & HPRE_PREFETCH_DISABLE), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) + pci_err(qm->pdev, "failed to open sva prefetch\n"); +} + +static void hpre_close_sva_prefetch(struct hisi_qm *qm) +{ + u32 val; + int ret; + + if (!test_bit(QM_SUPPORT_SVA_PREFETCH, &qm->caps)) + return; + + val = readl_relaxed(qm->io_base + HPRE_PREFETCH_CFG); + val |= HPRE_PREFETCH_DISABLE; + writel(val, qm->io_base + HPRE_PREFETCH_CFG); + + ret = readl_relaxed_poll_timeout(qm->io_base + HPRE_SVA_PREFTCH_DFX, + val, !(val & HPRE_SVA_DISABLE_READY), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) + pci_err(qm->pdev, "failed to close sva prefetch\n"); +} + +static void hpre_enable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + if (qm->ver < QM_HW_V3) + return; + + val = readl(qm->io_base + HPRE_CLKGATE_CTL); + val |= HPRE_CLKGATE_CTL_EN; + writel(val, qm->io_base + HPRE_CLKGATE_CTL); + + val = readl(qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + val |= HPRE_PEH_CFG_AUTO_GATE_EN; + writel(val, qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + + val = readl(qm->io_base + HPRE_CLUSTER_DYN_CTL); + val |= HPRE_CLUSTER_DYN_CTL_EN; + writel(val, qm->io_base + HPRE_CLUSTER_DYN_CTL); + + val = readl_relaxed(qm->io_base + HPRE_CORE_SHB_CFG); + val |= HPRE_CORE_GATE_EN; + writel(val, qm->io_base + HPRE_CORE_SHB_CFG); +} + +static void hpre_disable_clock_gate(struct hisi_qm *qm) +{ + u32 val; + + if (qm->ver < QM_HW_V3) + return; + + val = readl(qm->io_base + HPRE_CLKGATE_CTL); + val &= ~HPRE_CLKGATE_CTL_EN; + writel(val, qm->io_base + HPRE_CLKGATE_CTL); + + val = readl(qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + val &= ~HPRE_PEH_CFG_AUTO_GATE_EN; + writel(val, qm->io_base + HPRE_PEH_CFG_AUTO_GATE); + + val = readl(qm->io_base + HPRE_CLUSTER_DYN_CTL); + val &= ~HPRE_CLUSTER_DYN_CTL_EN; + writel(val, qm->io_base + HPRE_CLUSTER_DYN_CTL); + + val = readl_relaxed(qm->io_base + HPRE_CORE_SHB_CFG); + val &= ~HPRE_CORE_GATE_EN; + writel(val, qm->io_base + HPRE_CORE_SHB_CFG); +} + +static int hpre_set_user_domain_and_cache(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + u32 val; + int ret; + + /* disabel dynamic clock gate before sram init */ + hpre_disable_clock_gate(qm); + + writel(HPRE_QM_USR_CFG_MASK, qm->io_base + QM_ARUSER_M_CFG_ENABLE); + writel(HPRE_QM_USR_CFG_MASK, qm->io_base + QM_AWUSER_M_CFG_ENABLE); + writel_relaxed(HPRE_QM_AXI_CFG_MASK, qm->io_base + QM_AXI_M_CFG); + + /* HPRE need more time, we close this interrupt */ + val = readl_relaxed(qm->io_base + HPRE_QM_ABNML_INT_MASK); + val |= BIT(HPRE_TIMEOUT_ABNML_BIT); + writel_relaxed(val, qm->io_base + HPRE_QM_ABNML_INT_MASK); + + if (qm->ver >= QM_HW_V3) + writel(HPRE_RSA_ENB | HPRE_ECC_ENB, + qm->io_base + HPRE_TYPES_ENB); + else + writel(HPRE_RSA_ENB, qm->io_base + HPRE_TYPES_ENB); + + writel(HPRE_QM_VFG_AX_MASK, qm->io_base + HPRE_VFG_AXCACHE); + writel(0x0, qm->io_base + HPRE_BD_ENDIAN); + writel(0x0, qm->io_base + HPRE_INT_MASK); + writel(0x0, qm->io_base + HPRE_POISON_BYPASS); + writel(0x0, qm->io_base + HPRE_COMM_CNT_CLR_CE); + writel(0x0, qm->io_base + HPRE_ECC_BYPASS); + + writel(HPRE_BD_USR_MASK, qm->io_base + HPRE_BD_ARUSR_CFG); + writel(HPRE_BD_USR_MASK, qm->io_base + HPRE_BD_AWUSR_CFG); + writel(0x1, qm->io_base + HPRE_RDCHN_INI_CFG); + ret = readl_relaxed_poll_timeout(qm->io_base + HPRE_RDCHN_INI_ST, val, + val & BIT(0), + HPRE_REG_RD_INTVRL_US, + HPRE_REG_RD_TMOUT_US); + if (ret) { + dev_err(dev, "read rd channel timeout fail!\n"); + return -ETIMEDOUT; + } + + ret = hpre_set_cluster(qm); + if (ret) + return -ETIMEDOUT; + + /* This setting is only needed by Kunpeng 920. */ + if (qm->ver == QM_HW_V2) { + ret = hpre_cfg_by_dsm(qm); + if (ret) + return ret; + + disable_flr_of_bme(qm); + } + + /* Config data buffer pasid needed by Kunpeng 920 */ + hpre_config_pasid(qm); + + hpre_enable_clock_gate(qm); + + return ret; +} + +static void hpre_cnt_regs_clear(struct hisi_qm *qm) +{ + unsigned long offset; + u8 clusters_num; + int i; + + /* clear clusterX/cluster_ctrl */ + clusters_num = qm->cap_tables.dev_cap_table[HPRE_CLUSTER_NUM_CAP_IDX].cap_val; + for (i = 0; i < clusters_num; i++) { + offset = HPRE_CLSTR_BASE + i * HPRE_CLSTR_ADDR_INTRVL; + writel(0x0, qm->io_base + offset + HPRE_CLUSTER_INQURY); + } + + /* clear rdclr_en */ + writel(0x0, qm->io_base + HPRE_CTRL_CNT_CLR_CE); + + hisi_qm_debug_regs_clear(qm); +} + +static void hpre_master_ooo_ctrl(struct hisi_qm *qm, bool enable) +{ + u32 val1, val2; + + val1 = readl(qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); + if (enable) { + val1 |= HPRE_AM_OOO_SHUTDOWN_ENABLE; + val2 = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + } else { + val1 &= ~HPRE_AM_OOO_SHUTDOWN_ENABLE; + val2 = 0x0; + } + + if (qm->ver > QM_HW_V2) + writel(val2, qm->io_base + HPRE_OOO_SHUTDOWN_SEL); + + writel(val1, qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); +} + +static void hpre_hw_error_disable(struct hisi_qm *qm) +{ + u32 ce, nfe; + + ce = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_CE_MASK_CAP, qm->cap_ver); + nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_NFE_MASK_CAP, qm->cap_ver); + + /* disable hpre hw error interrupts */ + writel(ce | nfe | HPRE_HAC_RAS_FE_ENABLE, qm->io_base + HPRE_INT_MASK); + /* disable HPRE block master OOO when nfe occurs on Kunpeng930 */ + hpre_master_ooo_ctrl(qm, false); +} + +static void hpre_hw_error_enable(struct hisi_qm *qm) +{ + u32 ce, nfe; + + ce = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_CE_MASK_CAP, qm->cap_ver); + nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_NFE_MASK_CAP, qm->cap_ver); + + /* clear HPRE hw error source if having */ + writel(ce | nfe | HPRE_HAC_RAS_FE_ENABLE, qm->io_base + HPRE_HAC_SOURCE_INT); + + /* configure error type */ + writel(ce, qm->io_base + HPRE_RAS_CE_ENB); + writel(nfe, qm->io_base + HPRE_RAS_NFE_ENB); + writel(HPRE_HAC_RAS_FE_ENABLE, qm->io_base + HPRE_RAS_FE_ENB); + + /* enable HPRE block master OOO when nfe occurs on Kunpeng930 */ + hpre_master_ooo_ctrl(qm, true); + + /* enable hpre hw error interrupts */ + writel(HPRE_CORE_INT_ENABLE, qm->io_base + HPRE_INT_MASK); +} + +static inline struct hisi_qm *hpre_file_to_qm(struct hpre_debugfs_file *file) +{ + struct hpre *hpre = container_of(file->debug, struct hpre, debug); + + return &hpre->qm; +} + +static u32 hpre_clear_enable_read(struct hpre_debugfs_file *file) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + + return readl(qm->io_base + HPRE_CTRL_CNT_CLR_CE) & + HPRE_CTRL_CNT_CLR_CE_BIT; +} + +static int hpre_clear_enable_write(struct hpre_debugfs_file *file, u32 val) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + u32 tmp; + + if (val != 1 && val != 0) + return -EINVAL; + + tmp = (readl(qm->io_base + HPRE_CTRL_CNT_CLR_CE) & + ~HPRE_CTRL_CNT_CLR_CE_BIT) | val; + writel(tmp, qm->io_base + HPRE_CTRL_CNT_CLR_CE); + + return 0; +} + +static u32 hpre_cluster_inqry_read(struct hpre_debugfs_file *file) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + int cluster_index = file->index - HPRE_CLUSTER_CTRL; + unsigned long offset = HPRE_CLSTR_BASE + + cluster_index * HPRE_CLSTR_ADDR_INTRVL; + + return readl(qm->io_base + offset + HPRE_CLSTR_ADDR_INQRY_RSLT); +} + +static void hpre_cluster_inqry_write(struct hpre_debugfs_file *file, u32 val) +{ + struct hisi_qm *qm = hpre_file_to_qm(file); + int cluster_index = file->index - HPRE_CLUSTER_CTRL; + unsigned long offset = HPRE_CLSTR_BASE + cluster_index * + HPRE_CLSTR_ADDR_INTRVL; + + writel(val, qm->io_base + offset + HPRE_CLUSTER_INQURY); +} + +static ssize_t hpre_ctrl_debug_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct hpre_debugfs_file *file = filp->private_data; + struct hisi_qm *qm = hpre_file_to_qm(file); + char tbuf[HPRE_DBGFS_VAL_MAX_LEN]; + u32 val; + int ret; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + switch (file->type) { + case HPRE_CLEAR_ENABLE: + val = hpre_clear_enable_read(file); + break; + case HPRE_CLUSTER_CTRL: + val = hpre_cluster_inqry_read(file); + break; + default: + goto err_input; + } + spin_unlock_irq(&file->lock); + + hisi_qm_put_dfx_access(qm); + ret = snprintf(tbuf, HPRE_DBGFS_VAL_MAX_LEN, "%u\n", val); + return simple_read_from_buffer(buf, count, pos, tbuf, ret); + +err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return -EINVAL; +} + +static ssize_t hpre_ctrl_debug_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct hpre_debugfs_file *file = filp->private_data; + struct hisi_qm *qm = hpre_file_to_qm(file); + char tbuf[HPRE_DBGFS_VAL_MAX_LEN]; + unsigned long val; + int len, ret; + + if (*pos != 0) + return 0; + + if (count >= HPRE_DBGFS_VAL_MAX_LEN) + return -ENOSPC; + + len = simple_write_to_buffer(tbuf, HPRE_DBGFS_VAL_MAX_LEN - 1, + pos, buf, count); + if (len < 0) + return len; + + tbuf[len] = '\0'; + if (kstrtoul(tbuf, 0, &val)) + return -EFAULT; + + ret = hisi_qm_get_dfx_access(qm); + if (ret) + return ret; + + spin_lock_irq(&file->lock); + switch (file->type) { + case HPRE_CLEAR_ENABLE: + ret = hpre_clear_enable_write(file, val); + if (ret) + goto err_input; + break; + case HPRE_CLUSTER_CTRL: + hpre_cluster_inqry_write(file, val); + break; + default: + ret = -EINVAL; + goto err_input; + } + + ret = count; + +err_input: + spin_unlock_irq(&file->lock); + hisi_qm_put_dfx_access(qm); + return ret; +} + +static const struct file_operations hpre_ctrl_debug_fops = { + .owner = THIS_MODULE, + .open = simple_open, + .read = hpre_ctrl_debug_read, + .write = hpre_ctrl_debug_write, +}; + +static int hpre_debugfs_atomic64_get(void *data, u64 *val) +{ + struct hpre_dfx *dfx_item = data; + + *val = atomic64_read(&dfx_item->value); + + return 0; +} + +static int hpre_debugfs_atomic64_set(void *data, u64 val) +{ + struct hpre_dfx *dfx_item = data; + struct hpre_dfx *hpre_dfx = NULL; + + if (dfx_item->type == HPRE_OVERTIME_THRHLD) { + hpre_dfx = dfx_item - HPRE_OVERTIME_THRHLD; + atomic64_set(&hpre_dfx[HPRE_OVER_THRHLD_CNT].value, 0); + } else if (val) { + return -EINVAL; + } + + atomic64_set(&dfx_item->value, val); + + return 0; +} + +DEFINE_DEBUGFS_ATTRIBUTE(hpre_atomic64_ops, hpre_debugfs_atomic64_get, + hpre_debugfs_atomic64_set, "%llu\n"); + +static int hpre_create_debugfs_file(struct hisi_qm *qm, struct dentry *dir, + enum hpre_ctrl_dbgfs_file type, int indx) +{ + struct hpre *hpre = container_of(qm, struct hpre, qm); + struct hpre_debug *dbg = &hpre->debug; + struct dentry *file_dir; + + if (dir) + file_dir = dir; + else + file_dir = qm->debug.debug_root; + + if (type >= HPRE_DEBUG_FILE_NUM) + return -EINVAL; + + spin_lock_init(&dbg->files[indx].lock); + dbg->files[indx].debug = dbg; + dbg->files[indx].type = type; + dbg->files[indx].index = indx; + debugfs_create_file(hpre_debug_file_name[type], 0600, file_dir, + dbg->files + indx, &hpre_ctrl_debug_fops); + + return 0; +} + +static int hpre_pf_comm_regs_debugfs_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + struct debugfs_regset32 *regset; + + regset = devm_kzalloc(dev, sizeof(*regset), GFP_KERNEL); + if (!regset) + return -ENOMEM; + + regset->regs = hpre_com_dfx_regs; + regset->nregs = ARRAY_SIZE(hpre_com_dfx_regs); + regset->base = qm->io_base; + regset->dev = dev; + + debugfs_create_file("regs", 0444, qm->debug.debug_root, + regset, &hpre_com_regs_fops); + + return 0; +} + +static int hpre_cluster_debugfs_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + char buf[HPRE_DBGFS_VAL_MAX_LEN]; + struct debugfs_regset32 *regset; + struct dentry *tmp_d; + u8 clusters_num; + int i, ret; + + clusters_num = qm->cap_tables.dev_cap_table[HPRE_CLUSTER_NUM_CAP_IDX].cap_val; + for (i = 0; i < clusters_num; i++) { + ret = snprintf(buf, HPRE_DBGFS_VAL_MAX_LEN, "cluster%d", i); + if (ret >= HPRE_DBGFS_VAL_MAX_LEN) + return -EINVAL; + tmp_d = debugfs_create_dir(buf, qm->debug.debug_root); + + regset = devm_kzalloc(dev, sizeof(*regset), GFP_KERNEL); + if (!regset) + return -ENOMEM; + + regset->regs = hpre_cluster_dfx_regs; + regset->nregs = ARRAY_SIZE(hpre_cluster_dfx_regs); + regset->base = qm->io_base + hpre_cluster_offsets[i]; + regset->dev = dev; + + debugfs_create_file("regs", 0444, tmp_d, regset, + &hpre_cluster_regs_fops); + ret = hpre_create_debugfs_file(qm, tmp_d, HPRE_CLUSTER_CTRL, + i + HPRE_CLUSTER_CTRL); + if (ret) + return ret; + } + + return 0; +} + +static int hpre_ctrl_debug_init(struct hisi_qm *qm) +{ + int ret; + + ret = hpre_create_debugfs_file(qm, NULL, HPRE_CLEAR_ENABLE, + HPRE_CLEAR_ENABLE); + if (ret) + return ret; + + ret = hpre_pf_comm_regs_debugfs_init(qm); + if (ret) + return ret; + + return hpre_cluster_debugfs_init(qm); +} + +static void hpre_dfx_debug_init(struct hisi_qm *qm) +{ + struct dfx_diff_registers *hpre_regs = qm->debug.acc_diff_regs; + struct hpre *hpre = container_of(qm, struct hpre, qm); + struct hpre_dfx *dfx = hpre->debug.dfx; + struct dentry *parent; + int i; + + parent = debugfs_create_dir("hpre_dfx", qm->debug.debug_root); + for (i = 0; i < HPRE_DFX_FILE_NUM; i++) { + dfx[i].type = i; + debugfs_create_file(hpre_dfx_files[i], 0644, parent, &dfx[i], + &hpre_atomic64_ops); + } + + if (qm->fun_type == QM_HW_PF && hpre_regs) + debugfs_create_file("diff_regs", 0444, parent, + qm, &hpre_diff_regs_fops); +} + +static int hpre_debugfs_init(struct hisi_qm *qm) +{ + struct device *dev = &qm->pdev->dev; + int ret; + + qm->debug.debug_root = debugfs_create_dir(dev_name(dev), + hpre_debugfs_root); + + qm->debug.sqe_mask_offset = HPRE_SQE_MASK_OFFSET; + qm->debug.sqe_mask_len = HPRE_SQE_MASK_LEN; + ret = hisi_qm_regs_debugfs_init(qm, hpre_diff_regs, ARRAY_SIZE(hpre_diff_regs)); + if (ret) { + dev_warn(dev, "Failed to init HPRE diff regs!\n"); + goto debugfs_remove; + } + + hisi_qm_debug_init(qm); + + if (qm->pdev->device == PCI_DEVICE_ID_HUAWEI_HPRE_PF) { + ret = hpre_ctrl_debug_init(qm); + if (ret) + goto failed_to_create; + } + + hpre_dfx_debug_init(qm); + + return 0; + +failed_to_create: + hisi_qm_regs_debugfs_uninit(qm, ARRAY_SIZE(hpre_diff_regs)); +debugfs_remove: + debugfs_remove_recursive(qm->debug.debug_root); + return ret; +} + +static void hpre_debugfs_exit(struct hisi_qm *qm) +{ + hisi_qm_regs_debugfs_uninit(qm, ARRAY_SIZE(hpre_diff_regs)); + + debugfs_remove_recursive(qm->debug.debug_root); +} + +static int hpre_pre_store_cap_reg(struct hisi_qm *qm) +{ + struct hisi_qm_cap_record *hpre_cap; + struct device *dev = &qm->pdev->dev; + size_t i, size; + + size = ARRAY_SIZE(hpre_pre_store_caps); + hpre_cap = devm_kzalloc(dev, sizeof(*hpre_cap) * size, GFP_KERNEL); + if (!hpre_cap) + return -ENOMEM; + + for (i = 0; i < size; i++) { + hpre_cap[i].type = hpre_pre_store_caps[i]; + hpre_cap[i].cap_val = hisi_qm_get_hw_info(qm, hpre_basic_info, + hpre_pre_store_caps[i], qm->cap_ver); + } + + if (hpre_cap[HPRE_CLUSTER_NUM_CAP_IDX].cap_val > HPRE_CLUSTERS_NUM_MAX) { + dev_err(dev, "Device cluster num %u is out of range for driver supports %d!\n", + hpre_cap[HPRE_CLUSTER_NUM_CAP_IDX].cap_val, HPRE_CLUSTERS_NUM_MAX); + return -EINVAL; + } + + qm->cap_tables.dev_cap_table = hpre_cap; + + return 0; +} + +static int hpre_qm_init(struct hisi_qm *qm, struct pci_dev *pdev) +{ + u64 alg_msk; + int ret; + + if (pdev->revision == QM_HW_V1) { + pci_warn(pdev, "HPRE version 1 is not supported!\n"); + return -EINVAL; + } + + qm->mode = uacce_mode; + qm->pdev = pdev; + qm->ver = pdev->revision; + qm->sqe_size = HPRE_SQE_SIZE; + qm->dev_name = hpre_name; + + qm->fun_type = (pdev->device == PCI_DEVICE_ID_HUAWEI_HPRE_PF) ? + QM_HW_PF : QM_HW_VF; + if (qm->fun_type == QM_HW_PF) { + qm->qp_base = HPRE_PF_DEF_Q_BASE; + qm->qp_num = pf_q_num; + qm->debug.curr_qm_qp_num = pf_q_num; + qm->qm_list = &hpre_devices; + if (pf_q_num_flag) + set_bit(QM_MODULE_PARAM, &qm->misc_ctl); + } + + ret = hisi_qm_init(qm); + if (ret) { + pci_err(pdev, "Failed to init hpre qm configures!\n"); + return ret; + } + + /* Fetch and save the value of capability registers */ + ret = hpre_pre_store_cap_reg(qm); + if (ret) { + pci_err(pdev, "Failed to pre-store capability registers!\n"); + hisi_qm_uninit(qm); + return ret; + } + + alg_msk = qm->cap_tables.dev_cap_table[HPRE_DEV_ALG_BITMAP_CAP_IDX].cap_val; + ret = hisi_qm_set_algs(qm, alg_msk, hpre_dev_algs, ARRAY_SIZE(hpre_dev_algs)); + if (ret) { + pci_err(pdev, "Failed to set hpre algs!\n"); + hisi_qm_uninit(qm); + } + + return ret; +} + +static int hpre_show_last_regs_init(struct hisi_qm *qm) +{ + int cluster_dfx_regs_num = ARRAY_SIZE(hpre_cluster_dfx_regs); + int com_dfx_regs_num = ARRAY_SIZE(hpre_com_dfx_regs); + struct qm_debug *debug = &qm->debug; + void __iomem *io_base; + u8 clusters_num; + int i, j, idx; + + clusters_num = qm->cap_tables.dev_cap_table[HPRE_CLUSTER_NUM_CAP_IDX].cap_val; + debug->last_words = kcalloc(cluster_dfx_regs_num * clusters_num + + com_dfx_regs_num, sizeof(unsigned int), GFP_KERNEL); + if (!debug->last_words) + return -ENOMEM; + + for (i = 0; i < com_dfx_regs_num; i++) + debug->last_words[i] = readl_relaxed(qm->io_base + + hpre_com_dfx_regs[i].offset); + + for (i = 0; i < clusters_num; i++) { + io_base = qm->io_base + hpre_cluster_offsets[i]; + for (j = 0; j < cluster_dfx_regs_num; j++) { + idx = com_dfx_regs_num + i * cluster_dfx_regs_num + j; + debug->last_words[idx] = readl_relaxed( + io_base + hpre_cluster_dfx_regs[j].offset); + } + } + + return 0; +} + +static void hpre_show_last_regs_uninit(struct hisi_qm *qm) +{ + struct qm_debug *debug = &qm->debug; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + kfree(debug->last_words); + debug->last_words = NULL; +} + +static void hpre_show_last_dfx_regs(struct hisi_qm *qm) +{ + int cluster_dfx_regs_num = ARRAY_SIZE(hpre_cluster_dfx_regs); + int com_dfx_regs_num = ARRAY_SIZE(hpre_com_dfx_regs); + struct qm_debug *debug = &qm->debug; + struct pci_dev *pdev = qm->pdev; + void __iomem *io_base; + u8 clusters_num; + int i, j, idx; + u32 val; + + if (qm->fun_type == QM_HW_VF || !debug->last_words) + return; + + /* dumps last word of the debugging registers during controller reset */ + for (i = 0; i < com_dfx_regs_num; i++) { + val = readl_relaxed(qm->io_base + hpre_com_dfx_regs[i].offset); + if (debug->last_words[i] != val) + pci_info(pdev, "Common_core:%s \t= 0x%08x => 0x%08x\n", + hpre_com_dfx_regs[i].name, debug->last_words[i], val); + } + + clusters_num = qm->cap_tables.dev_cap_table[HPRE_CLUSTER_NUM_CAP_IDX].cap_val; + for (i = 0; i < clusters_num; i++) { + io_base = qm->io_base + hpre_cluster_offsets[i]; + for (j = 0; j < cluster_dfx_regs_num; j++) { + val = readl_relaxed(io_base + + hpre_cluster_dfx_regs[j].offset); + idx = com_dfx_regs_num + i * cluster_dfx_regs_num + j; + if (debug->last_words[idx] != val) + pci_info(pdev, "cluster-%d:%s \t= 0x%08x => 0x%08x\n", + i, hpre_cluster_dfx_regs[j].name, debug->last_words[idx], val); + } + } +} + +static void hpre_log_hw_error(struct hisi_qm *qm, u32 err_sts) +{ + const struct hpre_hw_error *err = hpre_hw_errors; + struct device *dev = &qm->pdev->dev; + + while (err->msg) { + if (err->int_msk & err_sts) + dev_warn(dev, "%s [error status=0x%x] found\n", + err->msg, err->int_msk); + err++; + } +} + +static u32 hpre_get_hw_err_status(struct hisi_qm *qm) +{ + return readl(qm->io_base + HPRE_INT_STATUS); +} + +static void hpre_clear_hw_err_status(struct hisi_qm *qm, u32 err_sts) +{ + u32 nfe; + + writel(err_sts, qm->io_base + HPRE_HAC_SOURCE_INT); + nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_NFE_MASK_CAP, qm->cap_ver); + writel(nfe, qm->io_base + HPRE_RAS_NFE_ENB); +} + +static void hpre_open_axi_master_ooo(struct hisi_qm *qm) +{ + u32 value; + + value = readl(qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); + writel(value & ~HPRE_AM_OOO_SHUTDOWN_ENABLE, + qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); + writel(value | HPRE_AM_OOO_SHUTDOWN_ENABLE, + qm->io_base + HPRE_AM_OOO_SHUTDOWN_ENB); +} + +static void hpre_err_info_init(struct hisi_qm *qm) +{ + struct hisi_qm_err_info *err_info = &qm->err_info; + + err_info->fe = HPRE_HAC_RAS_FE_ENABLE; + err_info->ce = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_QM_CE_MASK_CAP, qm->cap_ver); + err_info->nfe = hisi_qm_get_hw_info(qm, hpre_basic_info, HPRE_QM_NFE_MASK_CAP, qm->cap_ver); + err_info->ecc_2bits_mask = HPRE_CORE_ECC_2BIT_ERR | HPRE_OOO_ECC_2BIT_ERR; + err_info->dev_shutdown_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->qm_shutdown_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_QM_OOO_SHUTDOWN_MASK_CAP, qm->cap_ver); + err_info->qm_reset_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_QM_RESET_MASK_CAP, qm->cap_ver); + err_info->dev_reset_mask = hisi_qm_get_hw_info(qm, hpre_basic_info, + HPRE_RESET_MASK_CAP, qm->cap_ver); + err_info->msi_wr_port = HPRE_WR_MSI_PORT; + err_info->acpi_rst = "HRST"; +} + +static const struct hisi_qm_err_ini hpre_err_ini = { + .hw_init = hpre_set_user_domain_and_cache, + .hw_err_enable = hpre_hw_error_enable, + .hw_err_disable = hpre_hw_error_disable, + .get_dev_hw_err_status = hpre_get_hw_err_status, + .clear_dev_hw_err_status = hpre_clear_hw_err_status, + .log_dev_hw_err = hpre_log_hw_error, + .open_axi_master_ooo = hpre_open_axi_master_ooo, + .open_sva_prefetch = hpre_open_sva_prefetch, + .close_sva_prefetch = hpre_close_sva_prefetch, + .show_last_dfx_regs = hpre_show_last_dfx_regs, + .err_info_init = hpre_err_info_init, +}; + +static int hpre_pf_probe_init(struct hpre *hpre) +{ + struct hisi_qm *qm = &hpre->qm; + int ret; + + ret = hpre_set_user_domain_and_cache(qm); + if (ret) + return ret; + + hpre_open_sva_prefetch(qm); + + qm->err_ini = &hpre_err_ini; + qm->err_ini->err_info_init(qm); + hisi_qm_dev_err_init(qm); + ret = hpre_show_last_regs_init(qm); + if (ret) + pci_err(qm->pdev, "Failed to init last word regs!\n"); + + return ret; +} + +static int hpre_probe_init(struct hpre *hpre) +{ + u32 type_rate = HPRE_SHAPER_TYPE_RATE; + struct hisi_qm *qm = &hpre->qm; + int ret; + + if (qm->fun_type == QM_HW_PF) { + ret = hpre_pf_probe_init(hpre); + if (ret) + return ret; + /* Enable shaper type 0 */ + if (qm->ver >= QM_HW_V3) { + type_rate |= QM_SHAPER_ENABLE; + qm->type_rate = type_rate; + } + } + + return 0; +} + +static int hpre_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + struct hisi_qm *qm; + struct hpre *hpre; + int ret; + + hpre = devm_kzalloc(&pdev->dev, sizeof(*hpre), GFP_KERNEL); + if (!hpre) + return -ENOMEM; + + qm = &hpre->qm; + ret = hpre_qm_init(qm, pdev); + if (ret) { + pci_err(pdev, "Failed to init HPRE QM (%d)!\n", ret); + return ret; + } + + ret = hpre_probe_init(hpre); + if (ret) { + pci_err(pdev, "Failed to probe (%d)!\n", ret); + goto err_with_qm_init; + } + + ret = hisi_qm_start(qm); + if (ret) + goto err_with_err_init; + + ret = hpre_debugfs_init(qm); + if (ret) + dev_warn(&pdev->dev, "init debugfs fail!\n"); + + ret = hisi_qm_alg_register(qm, &hpre_devices); + if (ret < 0) { + pci_err(pdev, "fail to register algs to crypto!\n"); + goto err_with_qm_start; + } + + if (qm->uacce) { + ret = uacce_register(qm->uacce); + if (ret) { + pci_err(pdev, "failed to register uacce (%d)!\n", ret); + goto err_with_alg_register; + } + } + + if (qm->fun_type == QM_HW_PF && vfs_num) { + ret = hisi_qm_sriov_enable(pdev, vfs_num); + if (ret < 0) + goto err_with_alg_register; + } + + hisi_qm_pm_init(qm); + + return 0; + +err_with_alg_register: + hisi_qm_alg_unregister(qm, &hpre_devices); + +err_with_qm_start: + hpre_debugfs_exit(qm); + hisi_qm_stop(qm, QM_NORMAL); + +err_with_err_init: + hpre_show_last_regs_uninit(qm); + hisi_qm_dev_err_uninit(qm); + +err_with_qm_init: + hisi_qm_uninit(qm); + + return ret; +} + +static void hpre_remove(struct pci_dev *pdev) +{ + struct hisi_qm *qm = pci_get_drvdata(pdev); + + hisi_qm_pm_uninit(qm); + hisi_qm_wait_task_finish(qm, &hpre_devices); + hisi_qm_alg_unregister(qm, &hpre_devices); + if (qm->fun_type == QM_HW_PF && qm->vfs_num) + hisi_qm_sriov_disable(pdev, true); + + hpre_debugfs_exit(qm); + hisi_qm_stop(qm, QM_NORMAL); + + if (qm->fun_type == QM_HW_PF) { + hpre_cnt_regs_clear(qm); + qm->debug.curr_qm_qp_num = 0; + hpre_show_last_regs_uninit(qm); + hisi_qm_dev_err_uninit(qm); + } + + hisi_qm_uninit(qm); +} + +static const struct dev_pm_ops hpre_pm_ops = { + SET_RUNTIME_PM_OPS(hisi_qm_suspend, hisi_qm_resume, NULL) +}; + +static const struct pci_error_handlers hpre_err_handler = { + .error_detected = hisi_qm_dev_err_detected, + .slot_reset = hisi_qm_dev_slot_reset, + .reset_prepare = hisi_qm_reset_prepare, + .reset_done = hisi_qm_reset_done, +}; + +static struct pci_driver hpre_pci_driver = { + .name = hpre_name, + .id_table = hpre_dev_ids, + .probe = hpre_probe, + .remove = hpre_remove, + .sriov_configure = IS_ENABLED(CONFIG_PCI_IOV) ? + hisi_qm_sriov_configure : NULL, + .err_handler = &hpre_err_handler, + .shutdown = hisi_qm_dev_shutdown, + .driver.pm = &hpre_pm_ops, +}; + +struct pci_driver *hisi_hpre_get_pf_driver(void) +{ + return &hpre_pci_driver; +} +EXPORT_SYMBOL_GPL(hisi_hpre_get_pf_driver); + +static void hpre_register_debugfs(void) +{ + if (!debugfs_initialized()) + return; + + hpre_debugfs_root = debugfs_create_dir(hpre_name, NULL); +} + +static void hpre_unregister_debugfs(void) +{ + debugfs_remove_recursive(hpre_debugfs_root); +} + +static int __init hpre_init(void) +{ + int ret; + + hisi_qm_init_list(&hpre_devices); + hpre_register_debugfs(); + + ret = pci_register_driver(&hpre_pci_driver); + if (ret) { + hpre_unregister_debugfs(); + pr_err("hpre: can't register hisi hpre driver.\n"); + } + + return ret; +} + +static void __exit hpre_exit(void) +{ + pci_unregister_driver(&hpre_pci_driver); + hpre_unregister_debugfs(); +} + +module_init(hpre_init); +module_exit(hpre_exit); + +MODULE_LICENSE("GPL v2"); +MODULE_AUTHOR("Zaibo Xu "); +MODULE_AUTHOR("Meng Yu "); +MODULE_DESCRIPTION("Driver for HiSilicon HPRE accelerator"); -- cgit v1.2.3