diff options
Diffstat (limited to 'drivers/infiniband/hw')
23 files changed, 636 insertions, 469 deletions
diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h index 50cb2259bf87..fb8a0c248866 100644 --- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h +++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h @@ -930,8 +930,6 @@ void c4iw_id_table_free(struct c4iw_id_table *alloc); typedef int (*c4iw_handler_func)(struct c4iw_dev *dev, struct sk_buff *skb); -int c4iw_ep_redirect(void *ctx, struct dst_entry *old, struct dst_entry *new, - struct l2t_entry *l2t); void c4iw_put_qpid(struct c4iw_rdev *rdev, u32 qpid, struct c4iw_dev_ucontext *uctx); u32 c4iw_get_resource(struct c4iw_id_table *id_table); diff --git a/drivers/infiniband/hw/efa/efa.h b/drivers/infiniband/hw/efa/efa.h index e2bdec32ae80..926f9ff1f60f 100644 --- a/drivers/infiniband/hw/efa/efa.h +++ b/drivers/infiniband/hw/efa/efa.h @@ -57,6 +57,7 @@ struct efa_dev { u64 db_bar_addr; u64 db_bar_len; + unsigned int num_irq_vectors; int admin_msix_vector_idx; struct efa_irq admin_irq; diff --git a/drivers/infiniband/hw/efa/efa_main.c b/drivers/infiniband/hw/efa/efa_main.c index 7b1910a86216..5fa3603c80d8 100644 --- a/drivers/infiniband/hw/efa/efa_main.c +++ b/drivers/infiniband/hw/efa/efa_main.c @@ -322,7 +322,9 @@ static int efa_create_eqs(struct efa_dev *dev) int err; int i; - neqs = min_t(unsigned int, neqs, num_online_cpus()); + neqs = min_t(unsigned int, neqs, + dev->num_irq_vectors - EFA_COMP_EQS_VEC_BASE); + dev->neqs = neqs; dev->eqs = kcalloc(neqs, sizeof(*dev->eqs), GFP_KERNEL); if (!dev->eqs) @@ -468,34 +470,30 @@ static void efa_disable_msix(struct efa_dev *dev) static int efa_enable_msix(struct efa_dev *dev) { - int msix_vecs, irq_num; + int max_vecs, num_vecs; /* * Reserve the max msix vectors we might need, one vector is reserved * for admin. */ - msix_vecs = min_t(int, pci_msix_vec_count(dev->pdev), - num_online_cpus() + 1); + max_vecs = min_t(int, pci_msix_vec_count(dev->pdev), + num_online_cpus() + 1); dev_dbg(&dev->pdev->dev, "Trying to enable MSI-X, vectors %d\n", - msix_vecs); + max_vecs); dev->admin_msix_vector_idx = EFA_MGMNT_MSIX_VEC_IDX; - irq_num = pci_alloc_irq_vectors(dev->pdev, msix_vecs, - msix_vecs, PCI_IRQ_MSIX); + num_vecs = pci_alloc_irq_vectors(dev->pdev, 1, + max_vecs, PCI_IRQ_MSIX); - if (irq_num < 0) { - dev_err(&dev->pdev->dev, "Failed to enable MSI-X. irq_num %d\n", - irq_num); + if (num_vecs < 0) { + dev_err(&dev->pdev->dev, "Failed to enable MSI-X. error %d\n", + num_vecs); return -ENOSPC; } - if (irq_num != msix_vecs) { - efa_disable_msix(dev); - dev_err(&dev->pdev->dev, - "Allocated %d MSI-X (out of %d requested)\n", - irq_num, msix_vecs); - return -ENOSPC; - } + dev_dbg(&dev->pdev->dev, "Allocated %d MSI-X vectors\n", num_vecs); + + dev->num_irq_vectors = num_vecs; return 0; } diff --git a/drivers/infiniband/hw/hfi1/tid_rdma.c b/drivers/infiniband/hw/hfi1/tid_rdma.c index 18b05ffb415a..c465966a1d9c 100644 --- a/drivers/infiniband/hw/hfi1/tid_rdma.c +++ b/drivers/infiniband/hw/hfi1/tid_rdma.c @@ -315,7 +315,7 @@ int hfi1_kern_exp_rcv_init(struct hfi1_ctxtdata *rcd, int reinit) * This routine returns the receive context associated * with a a qp's qpn. * - * Returns the context. + * Return: the context. */ static struct hfi1_ctxtdata *qp_to_rcd(struct rvt_dev_info *rdi, struct rvt_qp *qp) @@ -710,7 +710,7 @@ void hfi1_tid_rdma_flush_wait(struct rvt_qp *qp) * The exp_lock must be held. * * Return: - * On success: a value postive value between 0 and RXE_NUM_TID_FLOWS - 1 + * On success: a value positive value between 0 and RXE_NUM_TID_FLOWS - 1 * On failure: -EAGAIN */ static int kern_reserve_flow(struct hfi1_ctxtdata *rcd, int last) @@ -1007,7 +1007,7 @@ static u32 tid_flush_pages(struct tid_rdma_pageset *list, * pages are tested two at a time, i, i + 1 for contiguous * pages and i - 1 and i contiguous pages. * - * If any condition is false, any accumlated pages are flushed and + * If any condition is false, any accumulated pages are flushed and * v0,v1 are emitted as separate PAGE_SIZE pagesets * * Otherwise, the current 8k is totaled for a future flush. @@ -1434,7 +1434,7 @@ static void kern_program_rcvarray(struct tid_rdma_flow *flow) * (5) computes a tidarray with formatted TID entries which can be sent * to the sender * (6) Reserves and programs HW flows. - * (7) It also manages queing the QP when TID/flow resources are not + * (7) It also manages queueing the QP when TID/flow resources are not * available. * * @req points to struct tid_rdma_request of which the segments are a part. The @@ -1604,7 +1604,7 @@ void hfi1_kern_exp_rcv_clear_all(struct tid_rdma_request *req) } /** - * hfi1_kern_exp_rcv_free_flows - free priviously allocated flow information + * hfi1_kern_exp_rcv_free_flows - free previously allocated flow information * @req: the tid rdma request to be cleaned */ static void hfi1_kern_exp_rcv_free_flows(struct tid_rdma_request *req) @@ -2055,7 +2055,7 @@ static int tid_rdma_rcv_error(struct hfi1_packet *packet, * req->clear_tail is advanced). However, when an earlier * request is received, this request will not be complete any * more (qp->s_tail_ack_queue is moved back, see below). - * Consequently, we need to update the TID flow info everytime + * Consequently, we need to update the TID flow info every time * a duplicate request is received. */ bth0 = be32_to_cpu(ohdr->bth[0]); @@ -2219,7 +2219,7 @@ void hfi1_rc_rcv_tid_rdma_read_req(struct hfi1_packet *packet) /* * 1. Verify TID RDMA READ REQ as per IB_OPCODE_RC_RDMA_READ * (see hfi1_rc_rcv()) - * 2. Put TID RDMA READ REQ into the response queueu (s_ack_queue) + * 2. Put TID RDMA READ REQ into the response queue (s_ack_queue) * - Setup struct tid_rdma_req with request info * - Initialize struct tid_rdma_flow info; * - Copy TID entries; @@ -2439,7 +2439,7 @@ find_tid_request(struct rvt_qp *qp, u32 psn, enum ib_wr_opcode opcode) void hfi1_rc_rcv_tid_rdma_read_resp(struct hfi1_packet *packet) { - /* HANDLER FOR TID RDMA READ RESPONSE packet (Requestor side */ + /* HANDLER FOR TID RDMA READ RESPONSE packet (Requester side) */ /* * 1. Find matching SWQE @@ -3649,7 +3649,7 @@ void hfi1_rc_rcv_tid_rdma_write_req(struct hfi1_packet *packet) * 1. Verify TID RDMA WRITE REQ as per IB_OPCODE_RC_RDMA_WRITE_FIRST * (see hfi1_rc_rcv()) * - Don't allow 0-length requests. - * 2. Put TID RDMA WRITE REQ into the response queueu (s_ack_queue) + * 2. Put TID RDMA WRITE REQ into the response queue (s_ack_queue) * - Setup struct tid_rdma_req with request info * - Prepare struct tid_rdma_flow array? * 3. Set the qp->s_ack_state as state diagram in design doc. @@ -4026,7 +4026,7 @@ unlock_r_lock: void hfi1_rc_rcv_tid_rdma_write_resp(struct hfi1_packet *packet) { - /* HANDLER FOR TID RDMA WRITE RESPONSE packet (Requestor side */ + /* HANDLER FOR TID RDMA WRITE RESPONSE packet (Requester side) */ /* * 1. Find matching SWQE @@ -5440,8 +5440,9 @@ static bool _hfi1_schedule_tid_send(struct rvt_qp *qp) * the two state machines can step on each other with respect to the * RVT_S_BUSY flag. * Therefore, a modified test is used. - * @return true if the second leg is scheduled; - * false if the second leg is not scheduled. + * + * Return: %true if the second leg is scheduled; + * %false if the second leg is not scheduled. */ bool hfi1_schedule_tid_send(struct rvt_qp *qp) { diff --git a/drivers/infiniband/hw/hns/hns_roce_cmd.h b/drivers/infiniband/hw/hns/hns_roce_cmd.h index 052a3d60905a..11dbbabebdc9 100644 --- a/drivers/infiniband/hw/hns/hns_roce_cmd.h +++ b/drivers/infiniband/hw/hns/hns_roce_cmd.h @@ -108,6 +108,9 @@ enum { HNS_ROCE_CMD_QUERY_CEQC = 0x92, HNS_ROCE_CMD_DESTROY_CEQC = 0x93, + /* SCC CTX commands */ + HNS_ROCE_CMD_QUERY_SCCC = 0xa2, + /* SCC CTX BT commands */ HNS_ROCE_CMD_READ_SCCC_BT0 = 0xa4, HNS_ROCE_CMD_WRITE_SCCC_BT0 = 0xa5, diff --git a/drivers/infiniband/hw/hns/hns_roce_cq.c b/drivers/infiniband/hw/hns/hns_roce_cq.c index 1b6d16af8c12..7250d0643b5c 100644 --- a/drivers/infiniband/hw/hns/hns_roce_cq.c +++ b/drivers/infiniband/hw/hns/hns_roce_cq.c @@ -133,14 +133,12 @@ static int alloc_cqc(struct hns_roce_dev *hr_dev, struct hns_roce_cq *hr_cq) struct hns_roce_cq_table *cq_table = &hr_dev->cq_table; struct ib_device *ibdev = &hr_dev->ib_dev; u64 mtts[MTT_MIN_COUNT] = {}; - dma_addr_t dma_handle; int ret; - ret = hns_roce_mtr_find(hr_dev, &hr_cq->mtr, 0, mtts, ARRAY_SIZE(mtts), - &dma_handle); - if (!ret) { + ret = hns_roce_mtr_find(hr_dev, &hr_cq->mtr, 0, mtts, ARRAY_SIZE(mtts)); + if (ret) { ibdev_err(ibdev, "failed to find CQ mtr, ret = %d.\n", ret); - return -EINVAL; + return ret; } /* Get CQC memory HEM(Hardware Entry Memory) table */ @@ -157,7 +155,8 @@ static int alloc_cqc(struct hns_roce_dev *hr_dev, struct hns_roce_cq *hr_cq) goto err_put; } - ret = hns_roce_create_cqc(hr_dev, hr_cq, mtts, dma_handle); + ret = hns_roce_create_cqc(hr_dev, hr_cq, mtts, + hns_roce_get_mtr_ba(&hr_cq->mtr)); if (ret) goto err_xa; diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index b1fce5ddf631..c3cbd0a494bf 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -179,6 +179,7 @@ enum { #define HNS_ROCE_CMD_SUCCESS 1 +#define HNS_ROCE_MAX_HOP_NUM 3 /* The minimum page size is 4K for hardware */ #define HNS_HW_PAGE_SHIFT 12 #define HNS_HW_PAGE_SIZE (1 << HNS_HW_PAGE_SHIFT) @@ -269,6 +270,11 @@ struct hns_roce_hem_list { dma_addr_t root_ba; /* pointer to the root ba table */ }; +enum mtr_type { + MTR_DEFAULT = 0, + MTR_PBL, +}; + struct hns_roce_buf_attr { struct { size_t size; /* region size */ @@ -277,7 +283,10 @@ struct hns_roce_buf_attr { unsigned int region_count; /* valid region count */ unsigned int page_shift; /* buffer page shift */ unsigned int user_access; /* umem access flag */ + u64 iova; + enum mtr_type type; bool mtt_only; /* only alloc buffer-required MTT memory */ + bool adaptive; /* adaptive for page_shift and hopnum */ }; struct hns_roce_hem_cfg { @@ -585,6 +594,13 @@ struct hns_roce_work { u32 queue_num; }; +enum hns_roce_cong_type { + CONG_TYPE_DCQCN, + CONG_TYPE_LDCP, + CONG_TYPE_HC3, + CONG_TYPE_DIP, +}; + struct hns_roce_qp { struct ib_qp ibqp; struct hns_roce_wq rq; @@ -628,6 +644,7 @@ struct hns_roce_qp { struct list_head sq_node; /* all send qps are on a list */ struct hns_user_mmap_entry *dwqe_mmap_entry; u32 config; + enum hns_roce_cong_type cong_type; }; struct hns_roce_ib_iboe { @@ -699,13 +716,6 @@ struct hns_roce_eq_table { struct hns_roce_eq *eq; }; -enum cong_type { - CONG_TYPE_DCQCN, - CONG_TYPE_LDCP, - CONG_TYPE_HC3, - CONG_TYPE_DIP, -}; - struct hns_roce_caps { u64 fw_ver; u8 num_ports; @@ -835,7 +845,8 @@ struct hns_roce_caps { u16 default_aeq_period; u16 default_aeq_arm_st; u16 default_ceq_arm_st; - enum cong_type cong_type; + u8 cong_cap; + enum hns_roce_cong_type default_cong_type; }; enum hns_roce_device_state { @@ -936,6 +947,7 @@ struct hns_roce_hw { int (*query_qpc)(struct hns_roce_dev *hr_dev, u32 qpn, void *buffer); int (*query_mpt)(struct hns_roce_dev *hr_dev, u32 key, void *buffer); int (*query_srqc)(struct hns_roce_dev *hr_dev, u32 srqn, void *buffer); + int (*query_sccc)(struct hns_roce_dev *hr_dev, u32 qpn, void *buffer); int (*query_hw_counter)(struct hns_roce_dev *hr_dev, u64 *stats, u32 port, int *hw_counters); const struct ib_device_ops *hns_roce_dev_ops; @@ -1152,8 +1164,13 @@ void hns_roce_cmd_use_polling(struct hns_roce_dev *hr_dev); /* hns roce hw need current block and next block addr from mtt */ #define MTT_MIN_COUNT 2 +static inline dma_addr_t hns_roce_get_mtr_ba(struct hns_roce_mtr *mtr) +{ + return mtr->hem_cfg.root_ba; +} + int hns_roce_mtr_find(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, - u32 offset, u64 *mtt_buf, int mtt_max, u64 *base_addr); + u32 offset, u64 *mtt_buf, int mtt_max); int hns_roce_mtr_create(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, struct hns_roce_buf_attr *buf_attr, unsigned int page_shift, struct ib_udata *udata, diff --git a/drivers/infiniband/hw/hns/hns_roce_hem.c b/drivers/infiniband/hw/hns/hns_roce_hem.c index c4ac06a33869..a4b3f19161dc 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hem.c +++ b/drivers/infiniband/hw/hns/hns_roce_hem.c @@ -249,61 +249,34 @@ int hns_roce_calc_hem_mhop(struct hns_roce_dev *hr_dev, } static struct hns_roce_hem *hns_roce_alloc_hem(struct hns_roce_dev *hr_dev, - int npages, unsigned long hem_alloc_size, gfp_t gfp_mask) { - struct hns_roce_hem_chunk *chunk = NULL; struct hns_roce_hem *hem; - struct scatterlist *mem; int order; void *buf; WARN_ON(gfp_mask & __GFP_HIGHMEM); + order = get_order(hem_alloc_size); + if (PAGE_SIZE << order != hem_alloc_size) { + dev_err(hr_dev->dev, "invalid hem_alloc_size: %lu!\n", + hem_alloc_size); + return NULL; + } + hem = kmalloc(sizeof(*hem), gfp_mask & ~(__GFP_HIGHMEM | __GFP_NOWARN)); if (!hem) return NULL; - INIT_LIST_HEAD(&hem->chunk_list); - - order = get_order(hem_alloc_size); - - while (npages > 0) { - if (!chunk) { - chunk = kmalloc(sizeof(*chunk), - gfp_mask & ~(__GFP_HIGHMEM | __GFP_NOWARN)); - if (!chunk) - goto fail; - - sg_init_table(chunk->mem, HNS_ROCE_HEM_CHUNK_LEN); - chunk->npages = 0; - chunk->nsg = 0; - memset(chunk->buf, 0, sizeof(chunk->buf)); - list_add_tail(&chunk->list, &hem->chunk_list); - } + buf = dma_alloc_coherent(hr_dev->dev, hem_alloc_size, + &hem->dma, gfp_mask); + if (!buf) + goto fail; - while (1 << order > npages) - --order; - - /* - * Alloc memory one time. If failed, don't alloc small block - * memory, directly return fail. - */ - mem = &chunk->mem[chunk->npages]; - buf = dma_alloc_coherent(hr_dev->dev, PAGE_SIZE << order, - &sg_dma_address(mem), gfp_mask); - if (!buf) - goto fail; - - chunk->buf[chunk->npages] = buf; - sg_dma_len(mem) = PAGE_SIZE << order; - - ++chunk->npages; - ++chunk->nsg; - npages -= 1 << order; - } + hem->buf = buf; + hem->size = hem_alloc_size; return hem; @@ -314,20 +287,10 @@ fail: void hns_roce_free_hem(struct hns_roce_dev *hr_dev, struct hns_roce_hem *hem) { - struct hns_roce_hem_chunk *chunk, *tmp; - int i; - if (!hem) return; - list_for_each_entry_safe(chunk, tmp, &hem->chunk_list, list) { - for (i = 0; i < chunk->npages; ++i) - dma_free_coherent(hr_dev->dev, - sg_dma_len(&chunk->mem[i]), - chunk->buf[i], - sg_dma_address(&chunk->mem[i])); - kfree(chunk); - } + dma_free_coherent(hr_dev->dev, hem->size, hem->buf, hem->dma); kfree(hem); } @@ -415,7 +378,6 @@ static int alloc_mhop_hem(struct hns_roce_dev *hr_dev, { u32 bt_size = mhop->bt_chunk_size; struct device *dev = hr_dev->dev; - struct hns_roce_hem_iter iter; gfp_t flag; u64 bt_ba; u32 size; @@ -456,16 +418,15 @@ static int alloc_mhop_hem(struct hns_roce_dev *hr_dev, */ size = table->type < HEM_TYPE_MTT ? mhop->buf_chunk_size : bt_size; flag = GFP_KERNEL | __GFP_NOWARN; - table->hem[index->buf] = hns_roce_alloc_hem(hr_dev, size >> PAGE_SHIFT, - size, flag); + table->hem[index->buf] = hns_roce_alloc_hem(hr_dev, size, flag); if (!table->hem[index->buf]) { ret = -ENOMEM; goto err_alloc_hem; } index->inited |= HEM_INDEX_BUF; - hns_roce_hem_first(table->hem[index->buf], &iter); - bt_ba = hns_roce_hem_addr(&iter); + bt_ba = table->hem[index->buf]->dma; + if (table->type < HEM_TYPE_MTT) { if (mhop->hop_num == 2) *(table->bt_l1[index->l1] + mhop->l2_idx) = bt_ba; @@ -586,7 +547,6 @@ int hns_roce_table_get(struct hns_roce_dev *hr_dev, } table->hem[i] = hns_roce_alloc_hem(hr_dev, - table->table_chunk_size >> PAGE_SHIFT, table->table_chunk_size, GFP_KERNEL | __GFP_NOWARN); if (!table->hem[i]) { @@ -725,7 +685,6 @@ void *hns_roce_table_find(struct hns_roce_dev *hr_dev, struct hns_roce_hem_table *table, unsigned long obj, dma_addr_t *dma_handle) { - struct hns_roce_hem_chunk *chunk; struct hns_roce_hem_mhop mhop; struct hns_roce_hem *hem; unsigned long mhop_obj = obj; @@ -734,7 +693,6 @@ void *hns_roce_table_find(struct hns_roce_dev *hr_dev, int offset, dma_offset; void *addr = NULL; u32 hem_idx = 0; - int length; int i, j; mutex_lock(&table->mutex); @@ -767,23 +725,8 @@ void *hns_roce_table_find(struct hns_roce_dev *hr_dev, if (!hem) goto out; - list_for_each_entry(chunk, &hem->chunk_list, list) { - for (i = 0; i < chunk->npages; ++i) { - length = sg_dma_len(&chunk->mem[i]); - if (dma_handle && dma_offset >= 0) { - if (length > (u32)dma_offset) - *dma_handle = sg_dma_address( - &chunk->mem[i]) + dma_offset; - dma_offset -= length; - } - - if (length > (u32)offset) { - addr = chunk->buf[i] + offset; - goto out; - } - offset -= length; - } - } + *dma_handle = hem->dma + dma_offset; + addr = hem->buf + offset; out: mutex_unlock(&table->mutex); diff --git a/drivers/infiniband/hw/hns/hns_roce_hem.h b/drivers/infiniband/hw/hns/hns_roce_hem.h index 7d23d3c51da4..6fb51db9682b 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hem.h +++ b/drivers/infiniband/hw/hns/hns_roce_hem.h @@ -56,10 +56,6 @@ enum { HEM_TYPE_TRRL, }; -#define HNS_ROCE_HEM_CHUNK_LEN \ - ((256 - sizeof(struct list_head) - 2 * sizeof(int)) / \ - (sizeof(struct scatterlist) + sizeof(void *))) - #define check_whether_bt_num_3(type, hop_num) \ (type < HEM_TYPE_MTT && hop_num == 2) @@ -72,25 +68,13 @@ enum { (type >= HEM_TYPE_MTT && hop_num == 1) || \ (type >= HEM_TYPE_MTT && hop_num == HNS_ROCE_HOP_NUM_0)) -struct hns_roce_hem_chunk { - struct list_head list; - int npages; - int nsg; - struct scatterlist mem[HNS_ROCE_HEM_CHUNK_LEN]; - void *buf[HNS_ROCE_HEM_CHUNK_LEN]; -}; - struct hns_roce_hem { - struct list_head chunk_list; + void *buf; + dma_addr_t dma; + unsigned long size; refcount_t refcount; }; -struct hns_roce_hem_iter { - struct hns_roce_hem *hem; - struct hns_roce_hem_chunk *chunk; - int page_idx; -}; - struct hns_roce_hem_mhop { u32 hop_num; u32 buf_chunk_size; @@ -133,38 +117,4 @@ void *hns_roce_hem_list_find_mtt(struct hns_roce_dev *hr_dev, struct hns_roce_hem_list *hem_list, int offset, int *mtt_cnt); -static inline void hns_roce_hem_first(struct hns_roce_hem *hem, - struct hns_roce_hem_iter *iter) -{ - iter->hem = hem; - iter->chunk = list_empty(&hem->chunk_list) ? NULL : - list_entry(hem->chunk_list.next, - struct hns_roce_hem_chunk, list); - iter->page_idx = 0; -} - -static inline int hns_roce_hem_last(struct hns_roce_hem_iter *iter) -{ - return !iter->chunk; -} - -static inline void hns_roce_hem_next(struct hns_roce_hem_iter *iter) -{ - if (++iter->page_idx >= iter->chunk->nsg) { - if (iter->chunk->list.next == &iter->hem->chunk_list) { - iter->chunk = NULL; - return; - } - - iter->chunk = list_entry(iter->chunk->list.next, - struct hns_roce_hem_chunk, list); - iter->page_idx = 0; - } -} - -static inline dma_addr_t hns_roce_hem_addr(struct hns_roce_hem_iter *iter) -{ - return sg_dma_address(&iter->chunk->mem[iter->page_idx]); -} - #endif /* _HNS_ROCE_HEM_H */ diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 8206daea6767..ba7ae792d279 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -2209,11 +2209,12 @@ static int hns_roce_query_caps(struct hns_roce_dev *hr_dev) caps->max_wqes = 1 << le16_to_cpu(resp_c->sq_depth); caps->num_srqs = 1 << hr_reg_read(resp_d, PF_CAPS_D_NUM_SRQS); - caps->cong_type = hr_reg_read(resp_d, PF_CAPS_D_CONG_TYPE); + caps->cong_cap = hr_reg_read(resp_d, PF_CAPS_D_CONG_CAP); caps->max_srq_wrs = 1 << le16_to_cpu(resp_d->srq_depth); caps->ceqe_depth = 1 << hr_reg_read(resp_d, PF_CAPS_D_CEQ_DEPTH); caps->num_comp_vectors = hr_reg_read(resp_d, PF_CAPS_D_NUM_CEQS); caps->aeqe_depth = 1 << hr_reg_read(resp_d, PF_CAPS_D_AEQ_DEPTH); + caps->default_cong_type = hr_reg_read(resp_d, PF_CAPS_D_DEFAULT_ALG); caps->reserved_pds = hr_reg_read(resp_d, PF_CAPS_D_RSV_PDS); caps->num_uars = 1 << hr_reg_read(resp_d, PF_CAPS_D_NUM_UARS); caps->reserved_qps = hr_reg_read(resp_d, PF_CAPS_D_RSV_QPS); @@ -3195,21 +3196,22 @@ static int set_mtpt_pbl(struct hns_roce_dev *hr_dev, u64 pages[HNS_ROCE_V2_MAX_INNER_MTPT_NUM] = { 0 }; struct ib_device *ibdev = &hr_dev->ib_dev; dma_addr_t pbl_ba; - int i, count; + int ret; + int i; - count = hns_roce_mtr_find(hr_dev, &mr->pbl_mtr, 0, pages, - min_t(int, ARRAY_SIZE(pages), mr->npages), - &pbl_ba); - if (count < 1) { - ibdev_err(ibdev, "failed to find PBL mtr, count = %d.\n", - count); - return -ENOBUFS; + ret = hns_roce_mtr_find(hr_dev, &mr->pbl_mtr, 0, pages, + min_t(int, ARRAY_SIZE(pages), mr->npages)); + if (ret) { + ibdev_err(ibdev, "failed to find PBL mtr, ret = %d.\n", ret); + return ret; } /* Aligned to the hardware address access unit */ - for (i = 0; i < count; i++) + for (i = 0; i < ARRAY_SIZE(pages); i++) pages[i] >>= 6; + pbl_ba = hns_roce_get_mtr_ba(&mr->pbl_mtr); + mpt_entry->pbl_size = cpu_to_le32(mr->npages); mpt_entry->pbl_ba_l = cpu_to_le32(pbl_ba >> 3); hr_reg_write(mpt_entry, MPT_PBL_BA_H, upper_32_bits(pbl_ba >> 3)); @@ -3308,18 +3310,12 @@ static int hns_roce_v2_rereg_write_mtpt(struct hns_roce_dev *hr_dev, static int hns_roce_v2_frmr_write_mtpt(struct hns_roce_dev *hr_dev, void *mb_buf, struct hns_roce_mr *mr) { - struct ib_device *ibdev = &hr_dev->ib_dev; + dma_addr_t pbl_ba = hns_roce_get_mtr_ba(&mr->pbl_mtr); struct hns_roce_v2_mpt_entry *mpt_entry; - dma_addr_t pbl_ba = 0; mpt_entry = mb_buf; memset(mpt_entry, 0, sizeof(*mpt_entry)); - if (hns_roce_mtr_find(hr_dev, &mr->pbl_mtr, 0, NULL, 0, &pbl_ba) < 0) { - ibdev_err(ibdev, "failed to find frmr mtr.\n"); - return -ENOBUFS; - } - hr_reg_write(mpt_entry, MPT_ST, V2_MPT_ST_FREE); hr_reg_write(mpt_entry, MPT_PD, mr->pd); @@ -4063,7 +4059,6 @@ static int hns_roce_v2_set_hem(struct hns_roce_dev *hr_dev, struct hns_roce_hem_table *table, int obj, u32 step_idx) { - struct hns_roce_hem_iter iter; struct hns_roce_hem_mhop mhop; struct hns_roce_hem *hem; unsigned long mhop_obj = obj; @@ -4100,12 +4095,8 @@ static int hns_roce_v2_set_hem(struct hns_roce_dev *hr_dev, if (check_whether_last_step(hop_num, step_idx)) { hem = table->hem[hem_idx]; - for (hns_roce_hem_first(hem, &iter); - !hns_roce_hem_last(&iter); hns_roce_hem_next(&iter)) { - bt_ba = hns_roce_hem_addr(&iter); - ret = set_hem_to_hw(hr_dev, obj, bt_ba, table->type, - step_idx); - } + + ret = set_hem_to_hw(hr_dev, obj, hem->dma, table->type, step_idx); } else { if (step_idx == 0) bt_ba = table->bt_l0_dma_addr[i]; @@ -4346,17 +4337,20 @@ static int config_qp_rq_buf(struct hns_roce_dev *hr_dev, { u64 mtts[MTT_MIN_COUNT] = { 0 }; u64 wqe_sge_ba; - int count; + int ret; /* Search qp buf's mtts */ - count = hns_roce_mtr_find(hr_dev, &hr_qp->mtr, hr_qp->rq.offset, mtts, - MTT_MIN_COUNT, &wqe_sge_ba); - if (hr_qp->rq.wqe_cnt && count < 1) { + ret = hns_roce_mtr_find(hr_dev, &hr_qp->mtr, hr_qp->rq.offset, mtts, + MTT_MIN_COUNT); + if (hr_qp->rq.wqe_cnt && ret) { ibdev_err(&hr_dev->ib_dev, - "failed to find RQ WQE, QPN = 0x%lx.\n", hr_qp->qpn); - return -EINVAL; + "failed to find QP(0x%lx) RQ WQE buf, ret = %d.\n", + hr_qp->qpn, ret); + return ret; } + wqe_sge_ba = hns_roce_get_mtr_ba(&hr_qp->mtr); + context->wqe_sge_ba = cpu_to_le32(wqe_sge_ba >> 3); qpc_mask->wqe_sge_ba = 0; @@ -4418,23 +4412,23 @@ static int config_qp_sq_buf(struct hns_roce_dev *hr_dev, struct ib_device *ibdev = &hr_dev->ib_dev; u64 sge_cur_blk = 0; u64 sq_cur_blk = 0; - int count; + int ret; /* search qp buf's mtts */ - count = hns_roce_mtr_find(hr_dev, &hr_qp->mtr, 0, &sq_cur_blk, 1, NULL); - if (count < 1) { - ibdev_err(ibdev, "failed to find QP(0x%lx) SQ buf.\n", - hr_qp->qpn); - return -EINVAL; + ret = hns_roce_mtr_find(hr_dev, &hr_qp->mtr, hr_qp->sq.offset, + &sq_cur_blk, 1); + if (ret) { + ibdev_err(ibdev, "failed to find QP(0x%lx) SQ WQE buf, ret = %d.\n", + hr_qp->qpn, ret); + return ret; } if (hr_qp->sge.sge_cnt > 0) { - count = hns_roce_mtr_find(hr_dev, &hr_qp->mtr, - hr_qp->sge.offset, - &sge_cur_blk, 1, NULL); - if (count < 1) { - ibdev_err(ibdev, "failed to find QP(0x%lx) SGE buf.\n", - hr_qp->qpn); - return -EINVAL; + ret = hns_roce_mtr_find(hr_dev, &hr_qp->mtr, + hr_qp->sge.offset, &sge_cur_blk, 1); + if (ret) { + ibdev_err(ibdev, "failed to find QP(0x%lx) SGE buf, ret = %d.\n", + hr_qp->qpn, ret); + return ret; } } @@ -4744,13 +4738,10 @@ enum { static int check_cong_type(struct ib_qp *ibqp, struct hns_roce_congestion_algorithm *cong_alg) { - struct hns_roce_dev *hr_dev = to_hr_dev(ibqp->device); - - if (ibqp->qp_type == IB_QPT_UD) - hr_dev->caps.cong_type = CONG_TYPE_DCQCN; + struct hns_roce_qp *hr_qp = to_hr_qp(ibqp); /* different congestion types match different configurations */ - switch (hr_dev->caps.cong_type) { + switch (hr_qp->cong_type) { case CONG_TYPE_DCQCN: cong_alg->alg_sel = CONG_DCQCN; cong_alg->alg_sub_sel = UNSUPPORT_CONG_LEVEL; @@ -4776,10 +4767,7 @@ static int check_cong_type(struct ib_qp *ibqp, cong_alg->wnd_mode_sel = WND_LIMIT; break; default: - ibdev_warn(&hr_dev->ib_dev, - "invalid type(%u) for congestion selection.\n", - hr_dev->caps.cong_type); - hr_dev->caps.cong_type = CONG_TYPE_DCQCN; + hr_qp->cong_type = CONG_TYPE_DCQCN; cong_alg->alg_sel = CONG_DCQCN; cong_alg->alg_sub_sel = UNSUPPORT_CONG_LEVEL; cong_alg->dip_vld = DIP_INVALID; @@ -4798,6 +4786,7 @@ static int fill_cong_field(struct ib_qp *ibqp, const struct ib_qp_attr *attr, struct hns_roce_congestion_algorithm cong_field; struct ib_device *ibdev = ibqp->device; struct hns_roce_dev *hr_dev = to_hr_dev(ibdev); + struct hns_roce_qp *hr_qp = to_hr_qp(ibqp); u32 dip_idx = 0; int ret; @@ -4810,7 +4799,7 @@ static int fill_cong_field(struct ib_qp *ibqp, const struct ib_qp_attr *attr, return ret; hr_reg_write(context, QPC_CONG_ALGO_TMPL_ID, hr_dev->cong_algo_tmpl_id + - hr_dev->caps.cong_type * HNS_ROCE_CONG_SIZE); + hr_qp->cong_type * HNS_ROCE_CONG_SIZE); hr_reg_clear(qpc_mask, QPC_CONG_ALGO_TMPL_ID); hr_reg_write(&context->ext, QPCEX_CONG_ALG_SEL, cong_field.alg_sel); hr_reg_clear(&qpc_mask->ext, QPCEX_CONG_ALG_SEL); @@ -5328,6 +5317,30 @@ out: return ret; } +static int hns_roce_v2_query_sccc(struct hns_roce_dev *hr_dev, u32 qpn, + void *buffer) +{ + struct hns_roce_v2_scc_context *context; + struct hns_roce_cmd_mailbox *mailbox; + int ret; + + mailbox = hns_roce_alloc_cmd_mailbox(hr_dev); + if (IS_ERR(mailbox)) + return PTR_ERR(mailbox); + + ret = hns_roce_cmd_mbox(hr_dev, 0, mailbox->dma, HNS_ROCE_CMD_QUERY_SCCC, + qpn); + if (ret) + goto out; + + context = mailbox->buf; + memcpy(buffer, context, sizeof(*context)); + +out: + hns_roce_free_cmd_mailbox(hr_dev, mailbox); + return ret; +} + static u8 get_qp_timeout_attr(struct hns_roce_dev *hr_dev, struct hns_roce_v2_qp_context *context) { @@ -5581,18 +5594,20 @@ static int hns_roce_v2_write_srqc_index_queue(struct hns_roce_srq *srq, struct ib_device *ibdev = srq->ibsrq.device; struct hns_roce_dev *hr_dev = to_hr_dev(ibdev); u64 mtts_idx[MTT_MIN_COUNT] = {}; - dma_addr_t dma_handle_idx = 0; + dma_addr_t dma_handle_idx; int ret; /* Get physical address of idx que buf */ ret = hns_roce_mtr_find(hr_dev, &idx_que->mtr, 0, mtts_idx, - ARRAY_SIZE(mtts_idx), &dma_handle_idx); - if (ret < 1) { + ARRAY_SIZE(mtts_idx)); + if (ret) { ibdev_err(ibdev, "failed to find mtr for SRQ idx, ret = %d.\n", ret); - return -ENOBUFS; + return ret; } + dma_handle_idx = hns_roce_get_mtr_ba(&idx_que->mtr); + hr_reg_write(ctx, SRQC_IDX_HOP_NUM, to_hr_hem_hopnum(hr_dev->caps.idx_hop_num, srq->wqe_cnt)); @@ -5624,20 +5639,22 @@ static int hns_roce_v2_write_srqc(struct hns_roce_srq *srq, void *mb_buf) struct hns_roce_dev *hr_dev = to_hr_dev(ibdev); struct hns_roce_srq_context *ctx = mb_buf; u64 mtts_wqe[MTT_MIN_COUNT] = {}; - dma_addr_t dma_handle_wqe = 0; + dma_addr_t dma_handle_wqe; int ret; memset(ctx, 0, sizeof(*ctx)); /* Get the physical address of srq buf */ ret = hns_roce_mtr_find(hr_dev, &srq->buf_mtr, 0, mtts_wqe, - ARRAY_SIZE(mtts_wqe), &dma_handle_wqe); - if (ret < 1) { + ARRAY_SIZE(mtts_wqe)); + if (ret) { ibdev_err(ibdev, "failed to find mtr for SRQ WQE, ret = %d.\n", ret); - return -ENOBUFS; + return ret; } + dma_handle_wqe = hns_roce_get_mtr_ba(&srq->buf_mtr); + hr_reg_write(ctx, SRQC_SRQ_ST, 1); hr_reg_write_bool(ctx, SRQC_SRQ_TYPE, srq->ibsrq.srq_type == IB_SRQT_XRC); @@ -6353,7 +6370,7 @@ static int config_eqc(struct hns_roce_dev *hr_dev, struct hns_roce_eq *eq, u64 eqe_ba[MTT_MIN_COUNT] = { 0 }; struct hns_roce_eq_context *eqc; u64 bt_ba = 0; - int count; + int ret; eqc = mb_buf; memset(eqc, 0, sizeof(struct hns_roce_eq_context)); @@ -6361,13 +6378,15 @@ static int config_eqc(struct hns_roce_dev *hr_dev, struct hns_roce_eq *eq, init_eq_config(hr_dev, eq); /* if not multi-hop, eqe buffer only use one trunk */ - count = hns_roce_mtr_find(hr_dev, &eq->mtr, 0, eqe_ba, MTT_MIN_COUNT, - &bt_ba); - if (count < 1) { - dev_err(hr_dev->dev, "failed to find EQE mtr\n"); - return -ENOBUFS; + ret = hns_roce_mtr_find(hr_dev, &eq->mtr, 0, eqe_ba, + ARRAY_SIZE(eqe_ba)); + if (ret) { + dev_err(hr_dev->dev, "failed to find EQE mtr, ret = %d\n", ret); + return ret; } + bt_ba = hns_roce_get_mtr_ba(&eq->mtr); + hr_reg_write(eqc, EQC_EQ_ST, HNS_ROCE_V2_EQ_STATE_VALID); hr_reg_write(eqc, EQC_EQE_HOP_NUM, eq->hop_num); hr_reg_write(eqc, EQC_OVER_IGNORE, eq->over_ignore); @@ -6714,6 +6733,7 @@ static const struct hns_roce_hw hns_roce_hw_v2 = { .query_qpc = hns_roce_v2_query_qpc, .query_mpt = hns_roce_v2_query_mpt, .query_srqc = hns_roce_v2_query_srqc, + .query_sccc = hns_roce_v2_query_sccc, .query_hw_counter = hns_roce_hw_v2_query_counter, .hns_roce_dev_ops = &hns_roce_v2_dev_ops, .hns_roce_dev_srq_ops = &hns_roce_v2_dev_srq_ops, diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h index cd97cbee682a..df04bc8ede57 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h @@ -646,6 +646,12 @@ struct hns_roce_v2_qp_context { #define QPCEX_SQ_RQ_NOT_FORBID_EN QPCEX_FIELD_LOC(23, 23) #define QPCEX_STASH QPCEX_FIELD_LOC(82, 82) +#define SCC_CONTEXT_SIZE 16 + +struct hns_roce_v2_scc_context { + __le32 data[SCC_CONTEXT_SIZE]; +}; + #define V2_QP_RWE_S 1 /* rdma write enable */ #define V2_QP_RRE_S 2 /* rdma read enable */ #define V2_QP_ATE_S 3 /* rdma atomic enable */ @@ -1214,12 +1220,13 @@ struct hns_roce_query_pf_caps_d { #define PF_CAPS_D_RQWQE_HOP_NUM PF_CAPS_D_FIELD_LOC(21, 20) #define PF_CAPS_D_EX_SGE_HOP_NUM PF_CAPS_D_FIELD_LOC(23, 22) #define PF_CAPS_D_SQWQE_HOP_NUM PF_CAPS_D_FIELD_LOC(25, 24) -#define PF_CAPS_D_CONG_TYPE PF_CAPS_D_FIELD_LOC(29, 26) +#define PF_CAPS_D_CONG_CAP PF_CAPS_D_FIELD_LOC(29, 26) #define PF_CAPS_D_CEQ_DEPTH PF_CAPS_D_FIELD_LOC(85, 64) #define PF_CAPS_D_NUM_CEQS PF_CAPS_D_FIELD_LOC(95, 86) #define PF_CAPS_D_AEQ_DEPTH PF_CAPS_D_FIELD_LOC(117, 96) #define PF_CAPS_D_AEQ_ARM_ST PF_CAPS_D_FIELD_LOC(119, 118) #define PF_CAPS_D_CEQ_ARM_ST PF_CAPS_D_FIELD_LOC(121, 120) +#define PF_CAPS_D_DEFAULT_ALG PF_CAPS_D_FIELD_LOC(127, 122) #define PF_CAPS_D_RSV_PDS PF_CAPS_D_FIELD_LOC(147, 128) #define PF_CAPS_D_NUM_UARS PF_CAPS_D_FIELD_LOC(155, 148) #define PF_CAPS_D_RSV_QPS PF_CAPS_D_FIELD_LOC(179, 160) diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c index b55fe6911f9f..1dc60c2b2b7a 100644 --- a/drivers/infiniband/hw/hns/hns_roce_main.c +++ b/drivers/infiniband/hw/hns/hns_roce_main.c @@ -394,6 +394,9 @@ static int hns_roce_alloc_ucontext(struct ib_ucontext *uctx, resp.config |= HNS_ROCE_RSP_CQE_INLINE_FLAGS; } + if (hr_dev->pci_dev->revision >= PCI_REVISION_ID_HIP09) + resp.congest_type = hr_dev->caps.cong_cap; + ret = hns_roce_uar_alloc(hr_dev, &context->uar); if (ret) goto error_out; diff --git a/drivers/infiniband/hw/hns/hns_roce_mr.c b/drivers/infiniband/hw/hns/hns_roce_mr.c index 91cd580480fe..9e05b57a2d67 100644 --- a/drivers/infiniband/hw/hns/hns_roce_mr.c +++ b/drivers/infiniband/hw/hns/hns_roce_mr.c @@ -32,6 +32,7 @@ */ #include <linux/vmalloc.h> +#include <linux/count_zeros.h> #include <rdma/ib_umem.h> #include <linux/math.h> #include "hns_roce_device.h" @@ -103,14 +104,21 @@ static int alloc_mr_pbl(struct hns_roce_dev *hr_dev, struct hns_roce_mr *mr, buf_attr.user_access = mr->access; /* fast MR's buffer is alloced before mapping, not at creation */ buf_attr.mtt_only = is_fast; + buf_attr.iova = mr->iova; + /* pagesize and hopnum is fixed for fast MR */ + buf_attr.adaptive = !is_fast; + buf_attr.type = MTR_PBL; err = hns_roce_mtr_create(hr_dev, &mr->pbl_mtr, &buf_attr, hr_dev->caps.pbl_ba_pg_sz + PAGE_SHIFT, udata, start); - if (err) + if (err) { ibdev_err(ibdev, "failed to alloc pbl mtr, ret = %d.\n", err); - else - mr->npages = mr->pbl_mtr.hem_cfg.buf_pg_count; + return err; + } + + mr->npages = mr->pbl_mtr.hem_cfg.buf_pg_count; + mr->pbl_hop_num = buf_attr.region[0].hopnum; return err; } @@ -695,7 +703,7 @@ static int mtr_alloc_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, mtr->umem = NULL; mtr->kmem = hns_roce_buf_alloc(hr_dev, total_size, buf_attr->page_shift, - mtr->hem_cfg.is_direct ? + !mtr_has_mtt(buf_attr) ? HNS_ROCE_BUF_DIRECT : 0); if (IS_ERR(mtr->kmem)) { ibdev_err(ibdev, "failed to alloc kmem, ret = %ld.\n", @@ -707,14 +715,41 @@ static int mtr_alloc_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, return 0; } -static int mtr_map_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, - int page_count, unsigned int page_shift) +static int cal_mtr_pg_cnt(struct hns_roce_mtr *mtr) +{ + struct hns_roce_buf_region *region; + int page_cnt = 0; + int i; + + for (i = 0; i < mtr->hem_cfg.region_count; i++) { + region = &mtr->hem_cfg.region[i]; + page_cnt += region->count; + } + + return page_cnt; +} + +static bool need_split_huge_page(struct hns_roce_mtr *mtr) +{ + /* When HEM buffer uses 0-level addressing, the page size is + * equal to the whole buffer size. If the current MTR has multiple + * regions, we split the buffer into small pages(4k, required by hns + * ROCEE). These pages will be used in multiple regions. + */ + return mtr->hem_cfg.is_direct && mtr->hem_cfg.region_count > 1; +} + +static int mtr_map_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr) { struct ib_device *ibdev = &hr_dev->ib_dev; + int page_count = cal_mtr_pg_cnt(mtr); + unsigned int page_shift; dma_addr_t *pages; int npage; int ret; + page_shift = need_split_huge_page(mtr) ? HNS_HW_PAGE_SHIFT : + mtr->hem_cfg.buf_pg_shift; /* alloc a tmp array to store buffer's dma address */ pages = kvcalloc(page_count, sizeof(dma_addr_t), GFP_KERNEL); if (!pages) @@ -734,7 +769,7 @@ static int mtr_map_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, goto err_alloc_list; } - if (mtr->hem_cfg.is_direct && npage > 1) { + if (need_split_huge_page(mtr) && npage > 1) { ret = mtr_check_direct_pages(pages, npage, page_shift); if (ret) { ibdev_err(ibdev, "failed to check %s page: %d / %d.\n", @@ -809,47 +844,53 @@ int hns_roce_mtr_map(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, return ret; } -int hns_roce_mtr_find(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, - u32 offset, u64 *mtt_buf, int mtt_max, u64 *base_addr) +static int hns_roce_get_direct_addr_mtt(struct hns_roce_hem_cfg *cfg, + u32 start_index, u64 *mtt_buf, + int mtt_cnt) { - struct hns_roce_hem_cfg *cfg = &mtr->hem_cfg; - int mtt_count, left; - u32 start_index; + int mtt_count; int total = 0; - __le64 *mtts; u32 npage; u64 addr; - if (!mtt_buf || mtt_max < 1) - goto done; - - /* no mtt memory in direct mode, so just return the buffer address */ - if (cfg->is_direct) { - start_index = offset >> HNS_HW_PAGE_SHIFT; - for (mtt_count = 0; mtt_count < cfg->region_count && - total < mtt_max; mtt_count++) { - npage = cfg->region[mtt_count].offset; - if (npage < start_index) - continue; + if (mtt_cnt > cfg->region_count) + return -EINVAL; - addr = cfg->root_ba + (npage << HNS_HW_PAGE_SHIFT); - mtt_buf[total] = addr; + for (mtt_count = 0; mtt_count < cfg->region_count && total < mtt_cnt; + mtt_count++) { + npage = cfg->region[mtt_count].offset; + if (npage < start_index) + continue; - total++; - } + addr = cfg->root_ba + (npage << HNS_HW_PAGE_SHIFT); + mtt_buf[total] = addr; - goto done; + total++; } - start_index = offset >> cfg->buf_pg_shift; - left = mtt_max; + if (!total) + return -ENOENT; + + return 0; +} + +static int hns_roce_get_mhop_mtt(struct hns_roce_dev *hr_dev, + struct hns_roce_mtr *mtr, u32 start_index, + u64 *mtt_buf, int mtt_cnt) +{ + int left = mtt_cnt; + int total = 0; + int mtt_count; + __le64 *mtts; + u32 npage; + while (left > 0) { mtt_count = 0; mtts = hns_roce_hem_list_find_mtt(hr_dev, &mtr->hem_list, start_index + total, &mtt_count); if (!mtts || !mtt_count) - goto done; + break; npage = min(mtt_count, left); left -= npage; @@ -857,69 +898,165 @@ int hns_roce_mtr_find(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, mtt_buf[total++] = le64_to_cpu(mtts[mtt_count]); } -done: - if (base_addr) - *base_addr = cfg->root_ba; + if (!total) + return -ENOENT; + + return 0; +} + +int hns_roce_mtr_find(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, + u32 offset, u64 *mtt_buf, int mtt_max) +{ + struct hns_roce_hem_cfg *cfg = &mtr->hem_cfg; + u32 start_index; + int ret; + + if (!mtt_buf || mtt_max < 1) + return -EINVAL; + + /* no mtt memory in direct mode, so just return the buffer address */ + if (cfg->is_direct) { + start_index = offset >> HNS_HW_PAGE_SHIFT; + ret = hns_roce_get_direct_addr_mtt(cfg, start_index, + mtt_buf, mtt_max); + } else { + start_index = offset >> cfg->buf_pg_shift; + ret = hns_roce_get_mhop_mtt(hr_dev, mtr, start_index, + mtt_buf, mtt_max); + } + return ret; +} + +static int get_best_page_shift(struct hns_roce_dev *hr_dev, + struct hns_roce_mtr *mtr, + struct hns_roce_buf_attr *buf_attr) +{ + unsigned int page_sz; + + if (!buf_attr->adaptive || buf_attr->type != MTR_PBL || !mtr->umem) + return 0; + + page_sz = ib_umem_find_best_pgsz(mtr->umem, + hr_dev->caps.page_size_cap, + buf_attr->iova); + if (!page_sz) + return -EINVAL; + + buf_attr->page_shift = order_base_2(page_sz); + return 0; +} + +static int get_best_hop_num(struct hns_roce_dev *hr_dev, + struct hns_roce_mtr *mtr, + struct hns_roce_buf_attr *buf_attr, + unsigned int ba_pg_shift) +{ +#define INVALID_HOPNUM -1 +#define MIN_BA_CNT 1 + size_t buf_pg_sz = 1 << buf_attr->page_shift; + struct ib_device *ibdev = &hr_dev->ib_dev; + size_t ba_pg_sz = 1 << ba_pg_shift; + int hop_num = INVALID_HOPNUM; + size_t unit = MIN_BA_CNT; + size_t ba_cnt; + int j; + + if (!buf_attr->adaptive || buf_attr->type != MTR_PBL) + return 0; + + /* Caculating the number of buf pages, each buf page need a BA */ + if (mtr->umem) + ba_cnt = ib_umem_num_dma_blocks(mtr->umem, buf_pg_sz); + else + ba_cnt = DIV_ROUND_UP(buf_attr->region[0].size, buf_pg_sz); + + for (j = 0; j <= HNS_ROCE_MAX_HOP_NUM; j++) { + if (ba_cnt <= unit) { + hop_num = j; + break; + } + /* Number of BAs can be represented at per hop */ + unit *= ba_pg_sz / BA_BYTE_LEN; + } + + if (hop_num < 0) { + ibdev_err(ibdev, + "failed to calculate a valid hopnum.\n"); + return -EINVAL; + } - return total; + buf_attr->region[0].hopnum = hop_num; + + return 0; +} + +static bool is_buf_attr_valid(struct hns_roce_dev *hr_dev, + struct hns_roce_buf_attr *attr) +{ + struct ib_device *ibdev = &hr_dev->ib_dev; + + if (attr->region_count > ARRAY_SIZE(attr->region) || + attr->region_count < 1 || attr->page_shift < HNS_HW_PAGE_SHIFT) { + ibdev_err(ibdev, + "invalid buf attr, region count %d, page shift %u.\n", + attr->region_count, attr->page_shift); + return false; + } + + return true; } static int mtr_init_buf_cfg(struct hns_roce_dev *hr_dev, - struct hns_roce_buf_attr *attr, - struct hns_roce_hem_cfg *cfg, - unsigned int *buf_page_shift, u64 unalinged_size) + struct hns_roce_mtr *mtr, + struct hns_roce_buf_attr *attr) { + struct hns_roce_hem_cfg *cfg = &mtr->hem_cfg; struct hns_roce_buf_region *r; - u64 first_region_padding; - int page_cnt, region_cnt; - unsigned int page_shift; + size_t buf_pg_sz; size_t buf_size; + int page_cnt, i; + u64 pgoff = 0; + + if (!is_buf_attr_valid(hr_dev, attr)) + return -EINVAL; /* If mtt is disabled, all pages must be within a continuous range */ cfg->is_direct = !mtr_has_mtt(attr); + cfg->region_count = attr->region_count; buf_size = mtr_bufs_size(attr); - if (cfg->is_direct) { - /* When HEM buffer uses 0-level addressing, the page size is - * equal to the whole buffer size, and we split the buffer into - * small pages which is used to check whether the adjacent - * units are in the continuous space and its size is fixed to - * 4K based on hns ROCEE's requirement. - */ - page_shift = HNS_HW_PAGE_SHIFT; - - /* The ROCEE requires the page size to be 4K * 2 ^ N. */ + if (need_split_huge_page(mtr)) { + buf_pg_sz = HNS_HW_PAGE_SIZE; cfg->buf_pg_count = 1; + /* The ROCEE requires the page size to be 4K * 2 ^ N. */ cfg->buf_pg_shift = HNS_HW_PAGE_SHIFT + order_base_2(DIV_ROUND_UP(buf_size, HNS_HW_PAGE_SIZE)); - first_region_padding = 0; } else { - page_shift = attr->page_shift; - cfg->buf_pg_count = DIV_ROUND_UP(buf_size + unalinged_size, - 1 << page_shift); - cfg->buf_pg_shift = page_shift; - first_region_padding = unalinged_size; + buf_pg_sz = 1 << attr->page_shift; + cfg->buf_pg_count = mtr->umem ? + ib_umem_num_dma_blocks(mtr->umem, buf_pg_sz) : + DIV_ROUND_UP(buf_size, buf_pg_sz); + cfg->buf_pg_shift = attr->page_shift; + pgoff = mtr->umem ? mtr->umem->address & ~PAGE_MASK : 0; } /* Convert buffer size to page index and page count for each region and * the buffer's offset needs to be appended to the first region. */ - for (page_cnt = 0, region_cnt = 0; region_cnt < attr->region_count && - region_cnt < ARRAY_SIZE(cfg->region); region_cnt++) { - r = &cfg->region[region_cnt]; + for (page_cnt = 0, i = 0; i < attr->region_count; i++) { + r = &cfg->region[i]; r->offset = page_cnt; - buf_size = hr_hw_page_align(attr->region[region_cnt].size + - first_region_padding); - r->count = DIV_ROUND_UP(buf_size, 1 << page_shift); - first_region_padding = 0; + buf_size = hr_hw_page_align(attr->region[i].size + pgoff); + if (attr->type == MTR_PBL && mtr->umem) + r->count = ib_umem_num_dma_blocks(mtr->umem, buf_pg_sz); + else + r->count = DIV_ROUND_UP(buf_size, buf_pg_sz); + + pgoff = 0; page_cnt += r->count; - r->hopnum = to_hr_hem_hopnum(attr->region[region_cnt].hopnum, - r->count); + r->hopnum = to_hr_hem_hopnum(attr->region[i].hopnum, r->count); } - cfg->region_count = region_cnt; - *buf_page_shift = page_shift; - - return page_cnt; + return 0; } static u64 cal_pages_per_l1ba(unsigned int ba_per_bt, unsigned int hopnum) @@ -1007,50 +1144,58 @@ int hns_roce_mtr_create(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr, unsigned long user_addr) { struct ib_device *ibdev = &hr_dev->ib_dev; - unsigned int buf_page_shift = 0; - int buf_page_cnt; int ret; - buf_page_cnt = mtr_init_buf_cfg(hr_dev, buf_attr, &mtr->hem_cfg, - &buf_page_shift, - udata ? user_addr & ~PAGE_MASK : 0); - if (buf_page_cnt < 1 || buf_page_shift < HNS_HW_PAGE_SHIFT) { - ibdev_err(ibdev, "failed to init mtr cfg, count %d shift %u.\n", - buf_page_cnt, buf_page_shift); - return -EINVAL; - } - - ret = mtr_alloc_mtt(hr_dev, mtr, ba_page_shift); - if (ret) { - ibdev_err(ibdev, "failed to alloc mtr mtt, ret = %d.\n", ret); - return ret; - } - /* The caller has its own buffer list and invokes the hns_roce_mtr_map() * to finish the MTT configuration. */ if (buf_attr->mtt_only) { mtr->umem = NULL; mtr->kmem = NULL; - return 0; + } else { + ret = mtr_alloc_bufs(hr_dev, mtr, buf_attr, udata, user_addr); + if (ret) { + ibdev_err(ibdev, + "failed to alloc mtr bufs, ret = %d.\n", ret); + return ret; + } + + ret = get_best_page_shift(hr_dev, mtr, buf_attr); + if (ret) + goto err_init_buf; + + ret = get_best_hop_num(hr_dev, mtr, buf_attr, ba_page_shift); + if (ret) + goto err_init_buf; } - ret = mtr_alloc_bufs(hr_dev, mtr, buf_attr, udata, user_addr); + ret = mtr_init_buf_cfg(hr_dev, mtr, buf_attr); + if (ret) + goto err_init_buf; + + ret = mtr_alloc_mtt(hr_dev, mtr, ba_page_shift); if (ret) { - ibdev_err(ibdev, "failed to alloc mtr bufs, ret = %d.\n", ret); - goto err_alloc_mtt; + ibdev_err(ibdev, "failed to alloc mtr mtt, ret = %d.\n", ret); + goto err_init_buf; } + if (buf_attr->mtt_only) + return 0; + /* Write buffer's dma address to MTT */ - ret = mtr_map_bufs(hr_dev, mtr, buf_page_cnt, buf_page_shift); - if (ret) + ret = mtr_map_bufs(hr_dev, mtr); + if (ret) { ibdev_err(ibdev, "failed to map mtr bufs, ret = %d.\n", ret); - else - return 0; + goto err_alloc_mtt; + } + + return 0; - mtr_free_bufs(hr_dev, mtr); err_alloc_mtt: mtr_free_mtt(hr_dev, mtr); +err_init_buf: + mtr_free_bufs(hr_dev, mtr); + return ret; } diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c index 31b147210688..f35a66325d9a 100644 --- a/drivers/infiniband/hw/hns/hns_roce_qp.c +++ b/drivers/infiniband/hw/hns/hns_roce_qp.c @@ -1004,6 +1004,60 @@ static void free_kernel_wrid(struct hns_roce_qp *hr_qp) kfree(hr_qp->sq.wrid); } +static void default_congest_type(struct hns_roce_dev *hr_dev, + struct hns_roce_qp *hr_qp) +{ + if (hr_qp->ibqp.qp_type == IB_QPT_UD || + hr_qp->ibqp.qp_type == IB_QPT_GSI) + hr_qp->cong_type = CONG_TYPE_DCQCN; + else + hr_qp->cong_type = hr_dev->caps.default_cong_type; +} + +static int set_congest_type(struct hns_roce_qp *hr_qp, + struct hns_roce_ib_create_qp *ucmd) +{ + struct hns_roce_dev *hr_dev = to_hr_dev(hr_qp->ibqp.device); + + switch (ucmd->cong_type_flags) { + case HNS_ROCE_CREATE_QP_FLAGS_DCQCN: + hr_qp->cong_type = CONG_TYPE_DCQCN; + break; + case HNS_ROCE_CREATE_QP_FLAGS_LDCP: + hr_qp->cong_type = CONG_TYPE_LDCP; + break; + case HNS_ROCE_CREATE_QP_FLAGS_HC3: + hr_qp->cong_type = CONG_TYPE_HC3; + break; + case HNS_ROCE_CREATE_QP_FLAGS_DIP: + hr_qp->cong_type = CONG_TYPE_DIP; + break; + default: + return -EINVAL; + } + + if (!test_bit(hr_qp->cong_type, (unsigned long *)&hr_dev->caps.cong_cap)) + return -EOPNOTSUPP; + + if (hr_qp->ibqp.qp_type == IB_QPT_UD && + hr_qp->cong_type != CONG_TYPE_DCQCN) + return -EOPNOTSUPP; + + return 0; +} + +static int set_congest_param(struct hns_roce_dev *hr_dev, + struct hns_roce_qp *hr_qp, + struct hns_roce_ib_create_qp *ucmd) +{ + if (ucmd->comp_mask & HNS_ROCE_CREATE_QP_MASK_CONGEST_TYPE) + return set_congest_type(hr_qp, ucmd); + + default_congest_type(hr_dev, hr_qp); + + return 0; +} + static int set_qp_param(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp, struct ib_qp_init_attr *init_attr, struct ib_udata *udata, @@ -1043,6 +1097,10 @@ static int set_qp_param(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp, ibdev_err(ibdev, "failed to set user SQ size, ret = %d.\n", ret); + + ret = set_congest_param(hr_dev, hr_qp, ucmd); + if (ret) + return ret; } else { if (hr_dev->pci_dev->revision >= PCI_REVISION_ID_HIP09) hr_qp->config = HNS_ROCE_EXSGE_FLAGS; @@ -1051,6 +1109,8 @@ static int set_qp_param(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp, ibdev_err(ibdev, "failed to set kernel SQ size, ret = %d.\n", ret); + + default_congest_type(hr_dev, hr_qp); } return ret; diff --git a/drivers/infiniband/hw/hns/hns_roce_restrack.c b/drivers/infiniband/hw/hns/hns_roce_restrack.c index f7f3c4cc7426..356d98816949 100644 --- a/drivers/infiniband/hw/hns/hns_roce_restrack.c +++ b/drivers/infiniband/hw/hns/hns_roce_restrack.c @@ -97,16 +97,33 @@ int hns_roce_fill_res_qp_entry_raw(struct sk_buff *msg, struct ib_qp *ib_qp) { struct hns_roce_dev *hr_dev = to_hr_dev(ib_qp->device); struct hns_roce_qp *hr_qp = to_hr_qp(ib_qp); - struct hns_roce_v2_qp_context context; + struct hns_roce_full_qp_ctx { + struct hns_roce_v2_qp_context qpc; + struct hns_roce_v2_scc_context sccc; + } context = {}; int ret; if (!hr_dev->hw->query_qpc) return -EINVAL; - ret = hr_dev->hw->query_qpc(hr_dev, hr_qp->qpn, &context); + ret = hr_dev->hw->query_qpc(hr_dev, hr_qp->qpn, &context.qpc); if (ret) - return -EINVAL; + return ret; + + /* If SCC is disabled or the query fails, the queried SCCC will + * be all 0. + */ + if (!(hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_QP_FLOW_CTRL) || + !hr_dev->hw->query_sccc) + goto out; + + ret = hr_dev->hw->query_sccc(hr_dev, hr_qp->qpn, &context.sccc); + if (ret) + ibdev_warn_ratelimited(&hr_dev->ib_dev, + "failed to query SCCC, ret = %d.\n", + ret); +out: ret = nla_put(msg, RDMA_NLDEV_ATTR_RES_RAW, sizeof(context), &context); return ret; diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c index 0b046c061742..12704efb7b19 100644 --- a/drivers/infiniband/hw/irdma/verbs.c +++ b/drivers/infiniband/hw/irdma/verbs.c @@ -719,7 +719,6 @@ static int irdma_setup_kmode_qp(struct irdma_device *iwdev, info->rq_pa + (ukinfo->rq_depth * IRDMA_QP_WQE_MIN_SIZE); ukinfo->sq_size = ukinfo->sq_depth >> ukinfo->sq_shift; ukinfo->rq_size = ukinfo->rq_depth >> ukinfo->rq_shift; - ukinfo->qp_id = iwqp->ibqp.qp_num; iwqp->max_send_wr = (ukinfo->sq_depth - IRDMA_SQ_RSVD) >> ukinfo->sq_shift; iwqp->max_recv_wr = (ukinfo->rq_depth - IRDMA_RQ_RSVD) >> ukinfo->rq_shift; @@ -944,7 +943,7 @@ static int irdma_create_qp(struct ib_qp *ibqp, iwqp->host_ctx.size = IRDMA_QP_CTX_SIZE; init_info.pd = &iwpd->sc_pd; - init_info.qp_uk_init_info.qp_id = iwqp->ibqp.qp_num; + init_info.qp_uk_init_info.qp_id = qp_num; if (!rdma_protocol_roce(&iwdev->ibdev, 1)) init_info.qp_uk_init_info.first_sq_wq = 1; iwqp->ctx_info.qp_compl_ctx = (uintptr_t)qp; diff --git a/drivers/infiniband/hw/mana/cq.c b/drivers/infiniband/hw/mana/cq.c index 83ebd070535a..4a71e678d09c 100644 --- a/drivers/infiniband/hw/mana/cq.c +++ b/drivers/infiniband/hw/mana/cq.c @@ -16,7 +16,7 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr, int err; mdev = container_of(ibdev, struct mana_ib_dev, ib_dev); - gc = mdev->gdma_dev->gdma_context; + gc = mdev_to_gc(mdev); if (udata->inlen < sizeof(ucmd)) return -EINVAL; @@ -48,7 +48,7 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr, return err; } - err = mana_ib_gd_create_dma_region(mdev, cq->umem, &cq->gdma_region); + err = mana_ib_create_zero_offset_dma_region(mdev, cq->umem, &cq->gdma_region); if (err) { ibdev_dbg(ibdev, "Failed to create dma region for create cq, %d\n", @@ -57,7 +57,7 @@ int mana_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr, } ibdev_dbg(ibdev, - "mana_ib_gd_create_dma_region ret %d gdma_region 0x%llx\n", + "create_dma_region ret %d gdma_region 0x%llx\n", err, cq->gdma_region); /* @@ -81,7 +81,7 @@ int mana_ib_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata) int err; mdev = container_of(ibdev, struct mana_ib_dev, ib_dev); - gc = mdev->gdma_dev->gdma_context; + gc = mdev_to_gc(mdev); err = mana_ib_gd_destroy_dma_region(mdev, cq->gdma_region); if (err) { @@ -100,10 +100,29 @@ int mana_ib_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata) return 0; } -void mana_ib_cq_handler(void *ctx, struct gdma_queue *gdma_cq) +static void mana_ib_cq_handler(void *ctx, struct gdma_queue *gdma_cq) { struct mana_ib_cq *cq = ctx; if (cq->ibcq.comp_handler) cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context); } + +int mana_ib_install_cq_cb(struct mana_ib_dev *mdev, struct mana_ib_cq *cq) +{ + struct gdma_context *gc = mdev_to_gc(mdev); + struct gdma_queue *gdma_cq; + + /* Create CQ table entry */ + WARN_ON(gc->cq_table[cq->id]); + gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL); + if (!gdma_cq) + return -ENOMEM; + + gdma_cq->cq.context = cq; + gdma_cq->type = GDMA_CQ; + gdma_cq->cq.callback = mana_ib_cq_handler; + gdma_cq->id = cq->id; + gc->cq_table[cq->id] = gdma_cq; + return 0; +} diff --git a/drivers/infiniband/hw/mana/main.c b/drivers/infiniband/hw/mana/main.c index faca092456fa..71e33feee61b 100644 --- a/drivers/infiniband/hw/mana/main.c +++ b/drivers/infiniband/hw/mana/main.c @@ -8,13 +8,10 @@ void mana_ib_uncfg_vport(struct mana_ib_dev *dev, struct mana_ib_pd *pd, u32 port) { - struct gdma_dev *gd = &dev->gdma_dev->gdma_context->mana; struct mana_port_context *mpc; struct net_device *ndev; - struct mana_context *mc; - mc = gd->driver_data; - ndev = mc->ports[port]; + ndev = mana_ib_get_netdev(&dev->ib_dev, port); mpc = netdev_priv(ndev); mutex_lock(&pd->vport_mutex); @@ -31,14 +28,11 @@ void mana_ib_uncfg_vport(struct mana_ib_dev *dev, struct mana_ib_pd *pd, int mana_ib_cfg_vport(struct mana_ib_dev *dev, u32 port, struct mana_ib_pd *pd, u32 doorbell_id) { - struct gdma_dev *mdev = &dev->gdma_dev->gdma_context->mana; struct mana_port_context *mpc; - struct mana_context *mc; struct net_device *ndev; int err; - mc = mdev->driver_data; - ndev = mc->ports[port]; + ndev = mana_ib_get_netdev(&dev->ib_dev, port); mpc = netdev_priv(ndev); mutex_lock(&pd->vport_mutex); @@ -79,17 +73,17 @@ int mana_ib_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata) struct gdma_create_pd_req req = {}; enum gdma_pd_flags flags = 0; struct mana_ib_dev *dev; - struct gdma_dev *mdev; + struct gdma_context *gc; int err; dev = container_of(ibdev, struct mana_ib_dev, ib_dev); - mdev = dev->gdma_dev; + gc = mdev_to_gc(dev); mana_gd_init_req_hdr(&req.hdr, GDMA_CREATE_PD, sizeof(req), sizeof(resp)); req.flags = flags; - err = mana_gd_send_request(mdev->gdma_context, sizeof(req), &req, + err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp); if (err || resp.hdr.status) { @@ -119,17 +113,17 @@ int mana_ib_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata) struct gdma_destory_pd_resp resp = {}; struct gdma_destroy_pd_req req = {}; struct mana_ib_dev *dev; - struct gdma_dev *mdev; + struct gdma_context *gc; int err; dev = container_of(ibdev, struct mana_ib_dev, ib_dev); - mdev = dev->gdma_dev; + gc = mdev_to_gc(dev); mana_gd_init_req_hdr(&req.hdr, GDMA_DESTROY_PD, sizeof(req), sizeof(resp)); req.pd_handle = pd->pd_handle; - err = mana_gd_send_request(mdev->gdma_context, sizeof(req), &req, + err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp); if (err || resp.hdr.status) { @@ -206,13 +200,11 @@ int mana_ib_alloc_ucontext(struct ib_ucontext *ibcontext, struct ib_device *ibdev = ibcontext->device; struct mana_ib_dev *mdev; struct gdma_context *gc; - struct gdma_dev *dev; int doorbell_page; int ret; mdev = container_of(ibdev, struct mana_ib_dev, ib_dev); - dev = mdev->gdma_dev; - gc = dev->gdma_context; + gc = mdev_to_gc(mdev); /* Allocate a doorbell page index */ ret = mana_gd_allocate_doorbell_page(gc, &doorbell_page); @@ -238,7 +230,7 @@ void mana_ib_dealloc_ucontext(struct ib_ucontext *ibcontext) int ret; mdev = container_of(ibdev, struct mana_ib_dev, ib_dev); - gc = mdev->gdma_dev->gdma_context; + gc = mdev_to_gc(mdev); ret = mana_gd_destroy_doorbell_page(gc, mana_ucontext->doorbell); if (ret) @@ -309,8 +301,8 @@ mana_ib_gd_add_dma_region(struct mana_ib_dev *dev, struct gdma_context *gc, return 0; } -int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, - mana_handle_t *gdma_region) +static int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, + mana_handle_t *gdma_region, unsigned long page_sz) { struct gdma_dma_region_add_pages_req *add_req = NULL; size_t num_pages_processed = 0, num_pages_to_handle; @@ -322,23 +314,14 @@ int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, size_t max_pgs_create_cmd; struct gdma_context *gc; size_t num_pages_total; - struct gdma_dev *mdev; - unsigned long page_sz; unsigned int tail = 0; u64 *page_addr_list; void *request_buf; int err; - mdev = dev->gdma_dev; - gc = mdev->gdma_context; + gc = mdev_to_gc(dev); hwc = gc->hwc.driver_data; - /* Hardware requires dma region to align to chosen page size */ - page_sz = ib_umem_find_best_pgsz(umem, PAGE_SZ_BM, 0); - if (!page_sz) { - ibdev_dbg(&dev->ib_dev, "failed to find page size.\n"); - return -ENOMEM; - } num_pages_total = ib_umem_num_dma_blocks(umem, page_sz); max_pgs_create_cmd = @@ -358,7 +341,7 @@ int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, sizeof(struct gdma_create_dma_region_resp)); create_req->length = umem->length; - create_req->offset_in_page = umem->address & (page_sz - 1); + create_req->offset_in_page = ib_umem_dma_offset(umem, page_sz); create_req->gdma_page_type = order_base_2(page_sz) - PAGE_SHIFT; create_req->page_count = num_pages_total; @@ -424,12 +407,39 @@ out: return err; } +int mana_ib_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, + mana_handle_t *gdma_region, u64 virt) +{ + unsigned long page_sz; + + page_sz = ib_umem_find_best_pgsz(umem, PAGE_SZ_BM, virt); + if (!page_sz) { + ibdev_dbg(&dev->ib_dev, "Failed to find page size.\n"); + return -EINVAL; + } + + return mana_ib_gd_create_dma_region(dev, umem, gdma_region, page_sz); +} + +int mana_ib_create_zero_offset_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, + mana_handle_t *gdma_region) +{ + unsigned long page_sz; + + /* Hardware requires dma region to align to chosen page size */ + page_sz = ib_umem_find_best_pgoff(umem, PAGE_SZ_BM, 0); + if (!page_sz) { + ibdev_dbg(&dev->ib_dev, "Failed to find page size.\n"); + return -EINVAL; + } + + return mana_ib_gd_create_dma_region(dev, umem, gdma_region, page_sz); +} + int mana_ib_gd_destroy_dma_region(struct mana_ib_dev *dev, u64 gdma_region) { - struct gdma_dev *mdev = dev->gdma_dev; - struct gdma_context *gc; + struct gdma_context *gc = mdev_to_gc(dev); - gc = mdev->gdma_context; ibdev_dbg(&dev->ib_dev, "destroy dma region 0x%llx\n", gdma_region); return mana_gd_destroy_dma_region(gc, gdma_region); @@ -447,7 +457,7 @@ int mana_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vma) int ret; mdev = container_of(ibdev, struct mana_ib_dev, ib_dev); - gc = mdev->gdma_dev->gdma_context; + gc = mdev_to_gc(mdev); if (vma->vm_pgoff != 0) { ibdev_dbg(ibdev, "Unexpected vm_pgoff %lu\n", vma->vm_pgoff); @@ -531,7 +541,7 @@ int mana_ib_gd_query_adapter_caps(struct mana_ib_dev *dev) req.hdr.resp.msg_version = GDMA_MESSAGE_V3; req.hdr.dev_id = dev->gdma_dev->dev_id; - err = mana_gd_send_request(dev->gdma_dev->gdma_context, sizeof(req), + err = mana_gd_send_request(mdev_to_gc(dev), sizeof(req), &req, sizeof(resp), &resp); if (err) { diff --git a/drivers/infiniband/hw/mana/mana_ib.h b/drivers/infiniband/hw/mana/mana_ib.h index 6bdc0f5498d5..f83390eebb7d 100644 --- a/drivers/infiniband/hw/mana/mana_ib.h +++ b/drivers/infiniband/hw/mana/mana_ib.h @@ -142,8 +142,29 @@ struct mana_ib_query_adapter_caps_resp { u32 max_inline_data_size; }; /* HW Data */ -int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, - mana_handle_t *gdma_region); +static inline struct gdma_context *mdev_to_gc(struct mana_ib_dev *mdev) +{ + return mdev->gdma_dev->gdma_context; +} + +static inline struct net_device *mana_ib_get_netdev(struct ib_device *ibdev, u32 port) +{ + struct mana_ib_dev *mdev = container_of(ibdev, struct mana_ib_dev, ib_dev); + struct gdma_context *gc = mdev_to_gc(mdev); + struct mana_context *mc = gc->mana.driver_data; + + if (port < 1 || port > mc->num_ports) + return NULL; + return mc->ports[port - 1]; +} + +int mana_ib_install_cq_cb(struct mana_ib_dev *mdev, struct mana_ib_cq *cq); + +int mana_ib_create_zero_offset_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, + mana_handle_t *gdma_region); + +int mana_ib_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem, + mana_handle_t *gdma_region, u64 virt); int mana_ib_gd_destroy_dma_region(struct mana_ib_dev *dev, mana_handle_t gdma_region); @@ -210,6 +231,4 @@ int mana_ib_query_gid(struct ib_device *ibdev, u32 port, int index, void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext); int mana_ib_gd_query_adapter_caps(struct mana_ib_dev *mdev); - -void mana_ib_cq_handler(void *ctx, struct gdma_queue *gdma_cq); #endif diff --git a/drivers/infiniband/hw/mana/mr.c b/drivers/infiniband/hw/mana/mr.c index 351207c60eb6..b70b13484f09 100644 --- a/drivers/infiniband/hw/mana/mr.c +++ b/drivers/infiniband/hw/mana/mr.c @@ -30,12 +30,9 @@ static int mana_ib_gd_create_mr(struct mana_ib_dev *dev, struct mana_ib_mr *mr, { struct gdma_create_mr_response resp = {}; struct gdma_create_mr_request req = {}; - struct gdma_dev *mdev = dev->gdma_dev; - struct gdma_context *gc; + struct gdma_context *gc = mdev_to_gc(dev); int err; - gc = mdev->gdma_context; - mana_gd_init_req_hdr(&req.hdr, GDMA_CREATE_MR, sizeof(req), sizeof(resp)); req.pd_handle = mr_params->pd_handle; @@ -77,12 +74,9 @@ static int mana_ib_gd_destroy_mr(struct mana_ib_dev *dev, u64 mr_handle) { struct gdma_destroy_mr_response resp = {}; struct gdma_destroy_mr_request req = {}; - struct gdma_dev *mdev = dev->gdma_dev; - struct gdma_context *gc; + struct gdma_context *gc = mdev_to_gc(dev); int err; - gc = mdev->gdma_context; - mana_gd_init_req_hdr(&req.hdr, GDMA_DESTROY_MR, sizeof(req), sizeof(resp)); @@ -133,7 +127,7 @@ struct ib_mr *mana_ib_reg_user_mr(struct ib_pd *ibpd, u64 start, u64 length, goto err_free; } - err = mana_ib_gd_create_dma_region(dev, mr->umem, &dma_region_handle); + err = mana_ib_create_dma_region(dev, mr->umem, &dma_region_handle, iova); if (err) { ibdev_dbg(ibdev, "Failed create dma region for user-mr, %d\n", err); @@ -141,7 +135,7 @@ struct ib_mr *mana_ib_reg_user_mr(struct ib_pd *ibpd, u64 start, u64 length, } ibdev_dbg(ibdev, - "mana_ib_gd_create_dma_region ret %d gdma_region %llx\n", err, + "create_dma_region ret %d gdma_region %llx\n", err, dma_region_handle); mr_params.pd_handle = pd->pd_handle; @@ -164,8 +158,7 @@ struct ib_mr *mana_ib_reg_user_mr(struct ib_pd *ibpd, u64 start, u64 length, return &mr->ibmr; err_dma_region: - mana_gd_destroy_dma_region(dev->gdma_dev->gdma_context, - dma_region_handle); + mana_gd_destroy_dma_region(mdev_to_gc(dev), dma_region_handle); err_umem: ib_umem_release(mr->umem); diff --git a/drivers/infiniband/hw/mana/qp.c b/drivers/infiniband/hw/mana/qp.c index 21ac9fcadf3f..6e7627745c95 100644 --- a/drivers/infiniband/hw/mana/qp.c +++ b/drivers/infiniband/hw/mana/qp.c @@ -17,12 +17,10 @@ static int mana_ib_cfg_vport_steering(struct mana_ib_dev *dev, struct mana_cfg_rx_steer_resp resp = {}; mana_handle_t *req_indir_tab; struct gdma_context *gc; - struct gdma_dev *mdev; u32 req_buf_size; int i, err; - gc = dev->gdma_dev->gdma_context; - mdev = &gc->mana; + gc = mdev_to_gc(dev); req_buf_size = sizeof(*req) + sizeof(mana_handle_t) * MANA_INDIRECT_TABLE_SIZE; @@ -39,7 +37,7 @@ static int mana_ib_cfg_vport_steering(struct mana_ib_dev *dev, req->rx_enable = 1; req->update_default_rxobj = 1; req->default_rxobj = default_rxobj; - req->hdr.dev_id = mdev->dev_id; + req->hdr.dev_id = gc->mana.dev_id; /* If there are more than 1 entries in indirection table, enable RSS */ if (log_ind_tbl_size) @@ -99,20 +97,17 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd, struct mana_ib_qp *qp = container_of(ibqp, struct mana_ib_qp, ibqp); struct mana_ib_dev *mdev = container_of(pd->device, struct mana_ib_dev, ib_dev); + struct gdma_context *gc = mdev_to_gc(mdev); struct ib_rwq_ind_table *ind_tbl = attr->rwq_ind_tbl; struct mana_ib_create_qp_rss_resp resp = {}; struct mana_ib_create_qp_rss ucmd = {}; struct gdma_queue **gdma_cq_allocated; mana_handle_t *mana_ind_table; struct mana_port_context *mpc; - struct gdma_queue *gdma_cq; unsigned int ind_tbl_size; - struct mana_context *mc; struct net_device *ndev; - struct gdma_context *gc; struct mana_ib_cq *cq; struct mana_ib_wq *wq; - struct gdma_dev *gd; struct mana_eq *eq; struct ib_cq *ibcq; struct ib_wq *ibwq; @@ -120,10 +115,6 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd, u32 port; int ret; - gc = mdev->gdma_dev->gdma_context; - gd = &gc->mana; - mc = gd->driver_data; - if (!udata || udata->inlen < sizeof(ucmd)) return -EINVAL; @@ -166,12 +157,12 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd, /* IB ports start with 1, MANA start with 0 */ port = ucmd.port; - if (port < 1 || port > mc->num_ports) { + ndev = mana_ib_get_netdev(pd->device, port); + if (!ndev) { ibdev_dbg(&mdev->ib_dev, "Invalid port %u in creating qp\n", port); return -EINVAL; } - ndev = mc->ports[port - 1]; mpc = netdev_priv(ndev); ibdev_dbg(&mdev->ib_dev, "rx_hash_function %d port %d\n", @@ -209,7 +200,7 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd, cq_spec.gdma_region = cq->gdma_region; cq_spec.queue_size = cq->cqe * COMP_ENTRY_SIZE; cq_spec.modr_ctx_id = 0; - eq = &mc->eqs[cq->comp_vector % gc->max_num_queues]; + eq = &mpc->ac->eqs[cq->comp_vector % gc->max_num_queues]; cq_spec.attached_eq = eq->eq->id; ret = mana_create_wq_obj(mpc, mpc->port_handle, GDMA_RQ, @@ -237,19 +228,11 @@ static int mana_ib_create_qp_rss(struct ib_qp *ibqp, struct ib_pd *pd, mana_ind_table[i] = wq->rx_object; /* Create CQ table entry */ - WARN_ON(gc->cq_table[cq->id]); - gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL); - if (!gdma_cq) { - ret = -ENOMEM; + ret = mana_ib_install_cq_cb(mdev, cq); + if (ret) goto fail; - } - gdma_cq_allocated[i] = gdma_cq; - gdma_cq->cq.context = cq; - gdma_cq->type = GDMA_CQ; - gdma_cq->cq.callback = mana_ib_cq_handler; - gdma_cq->id = cq->id; - gc->cq_table[cq->id] = gdma_cq; + gdma_cq_allocated[i] = gc->cq_table[cq->id]; } resp.num_entries = i; @@ -306,14 +289,13 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, struct mana_ib_ucontext *mana_ucontext = rdma_udata_to_drv_context(udata, struct mana_ib_ucontext, ibucontext); - struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana; + struct gdma_context *gc = mdev_to_gc(mdev); struct mana_ib_create_qp_resp resp = {}; struct mana_ib_create_qp ucmd = {}; struct gdma_queue *gdma_cq = NULL; struct mana_obj_spec wq_spec = {}; struct mana_obj_spec cq_spec = {}; struct mana_port_context *mpc; - struct mana_context *mc; struct net_device *ndev; struct ib_umem *umem; struct mana_eq *eq; @@ -321,8 +303,6 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, u32 port; int err; - mc = gd->driver_data; - if (!mana_ucontext || udata->inlen < sizeof(ucmd)) return -EINVAL; @@ -333,11 +313,6 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, return err; } - /* IB ports start with 1, MANA Ethernet ports start with 0 */ - port = ucmd.port; - if (port < 1 || port > mc->num_ports) - return -EINVAL; - if (attr->cap.max_send_wr > mdev->adapter_caps.max_qp_wr) { ibdev_dbg(&mdev->ib_dev, "Requested max_send_wr %d exceeding limit\n", @@ -352,11 +327,17 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, return -EINVAL; } - ndev = mc->ports[port - 1]; + port = ucmd.port; + ndev = mana_ib_get_netdev(ibpd->device, port); + if (!ndev) { + ibdev_dbg(&mdev->ib_dev, "Invalid port %u in creating qp\n", + port); + return -EINVAL; + } mpc = netdev_priv(ndev); ibdev_dbg(&mdev->ib_dev, "port %u ndev %p mpc %p\n", port, ndev, mpc); - err = mana_ib_cfg_vport(mdev, port - 1, pd, mana_ucontext->doorbell); + err = mana_ib_cfg_vport(mdev, port, pd, mana_ucontext->doorbell); if (err) return -ENODEV; @@ -376,8 +357,8 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, } qp->sq_umem = umem; - err = mana_ib_gd_create_dma_region(mdev, qp->sq_umem, - &qp->sq_gdma_region); + err = mana_ib_create_zero_offset_dma_region(mdev, qp->sq_umem, + &qp->sq_gdma_region); if (err) { ibdev_dbg(&mdev->ib_dev, "Failed to create dma region for create qp-raw, %d\n", @@ -386,7 +367,7 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, } ibdev_dbg(&mdev->ib_dev, - "mana_ib_gd_create_dma_region ret %d gdma_region 0x%llx\n", + "create_dma_region ret %d gdma_region 0x%llx\n", err, qp->sq_gdma_region); /* Create a WQ on the same port handle used by the Ethernet */ @@ -396,8 +377,8 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, cq_spec.gdma_region = send_cq->gdma_region; cq_spec.queue_size = send_cq->cqe * COMP_ENTRY_SIZE; cq_spec.modr_ctx_id = 0; - eq_vec = send_cq->comp_vector % gd->gdma_context->max_num_queues; - eq = &mc->eqs[eq_vec]; + eq_vec = send_cq->comp_vector % gc->max_num_queues; + eq = &mpc->ac->eqs[eq_vec]; cq_spec.attached_eq = eq->eq->id; err = mana_create_wq_obj(mpc, mpc->port_handle, GDMA_SQ, &wq_spec, @@ -417,18 +398,9 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, send_cq->id = cq_spec.queue_index; /* Create CQ table entry */ - WARN_ON(gd->gdma_context->cq_table[send_cq->id]); - gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL); - if (!gdma_cq) { - err = -ENOMEM; + err = mana_ib_install_cq_cb(mdev, send_cq); + if (err) goto err_destroy_wq_obj; - } - - gdma_cq->cq.context = send_cq; - gdma_cq->type = GDMA_CQ; - gdma_cq->cq.callback = mana_ib_cq_handler; - gdma_cq->id = send_cq->id; - gd->gdma_context->cq_table[send_cq->id] = gdma_cq; ibdev_dbg(&mdev->ib_dev, "ret %d qp->tx_object 0x%llx sq id %llu cq id %llu\n", err, @@ -450,7 +422,7 @@ static int mana_ib_create_qp_raw(struct ib_qp *ibqp, struct ib_pd *ibpd, err_release_gdma_cq: kfree(gdma_cq); - gd->gdma_context->cq_table[send_cq->id] = NULL; + gc->cq_table[send_cq->id] = NULL; err_destroy_wq_obj: mana_destroy_wq_obj(mpc, GDMA_SQ, qp->tx_object); @@ -462,7 +434,7 @@ err_release_umem: ib_umem_release(umem); err_free_vport: - mana_ib_uncfg_vport(mdev, pd, port - 1); + mana_ib_uncfg_vport(mdev, pd, port); return err; } @@ -500,16 +472,13 @@ static int mana_ib_destroy_qp_rss(struct mana_ib_qp *qp, { struct mana_ib_dev *mdev = container_of(qp->ibqp.device, struct mana_ib_dev, ib_dev); - struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana; struct mana_port_context *mpc; - struct mana_context *mc; struct net_device *ndev; struct mana_ib_wq *wq; struct ib_wq *ibwq; int i; - mc = gd->driver_data; - ndev = mc->ports[qp->port - 1]; + ndev = mana_ib_get_netdev(qp->ibqp.device, qp->port); mpc = netdev_priv(ndev); for (i = 0; i < (1 << ind_tbl->log_ind_tbl_size); i++) { @@ -527,15 +496,12 @@ static int mana_ib_destroy_qp_raw(struct mana_ib_qp *qp, struct ib_udata *udata) { struct mana_ib_dev *mdev = container_of(qp->ibqp.device, struct mana_ib_dev, ib_dev); - struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana; struct ib_pd *ibpd = qp->ibqp.pd; struct mana_port_context *mpc; - struct mana_context *mc; struct net_device *ndev; struct mana_ib_pd *pd; - mc = gd->driver_data; - ndev = mc->ports[qp->port - 1]; + ndev = mana_ib_get_netdev(qp->ibqp.device, qp->port); mpc = netdev_priv(ndev); pd = container_of(ibpd, struct mana_ib_pd, ibpd); @@ -546,7 +512,7 @@ static int mana_ib_destroy_qp_raw(struct mana_ib_qp *qp, struct ib_udata *udata) ib_umem_release(qp->sq_umem); } - mana_ib_uncfg_vport(mdev, pd, qp->port - 1); + mana_ib_uncfg_vport(mdev, pd, qp->port); return 0; } diff --git a/drivers/infiniband/hw/mana/wq.c b/drivers/infiniband/hw/mana/wq.c index 372d361510e0..7c9c69962573 100644 --- a/drivers/infiniband/hw/mana/wq.c +++ b/drivers/infiniband/hw/mana/wq.c @@ -46,7 +46,7 @@ struct ib_wq *mana_ib_create_wq(struct ib_pd *pd, wq->wq_buf_size = ucmd.wq_buf_size; wq->rx_object = INVALID_MANA_HANDLE; - err = mana_ib_gd_create_dma_region(mdev, wq->umem, &wq->gdma_region); + err = mana_ib_create_zero_offset_dma_region(mdev, wq->umem, &wq->gdma_region); if (err) { ibdev_dbg(&mdev->ib_dev, "Failed to create dma region for create wq, %d\n", @@ -55,7 +55,7 @@ struct ib_wq *mana_ib_create_wq(struct ib_pd *pd, } ibdev_dbg(&mdev->ib_dev, - "mana_ib_gd_create_dma_region ret %d gdma_region 0x%llx\n", + "create_dma_region ret %d gdma_region 0x%llx\n", err, wq->gdma_region); /* WQ ID is returned at wq_create time, doesn't know the value yet */ diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index bbe79b86c717..a8de35c07c9e 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -1377,7 +1377,6 @@ int mlx5_ib_query_port(struct ib_device *ibdev, u32 port, struct ib_port_attr *props); void mlx5_ib_populate_pas(struct ib_umem *umem, size_t page_size, __be64 *pas, u64 access_flags); -void mlx5_ib_copy_pas(u64 *old, u64 *new, int step, int num); int mlx5_ib_get_cqe_size(struct ib_cq *ibcq); int mlx5_mkey_cache_init(struct mlx5_ib_dev *dev); void mlx5_mkey_cache_cleanup(struct mlx5_ib_dev *dev); |