diff options
Diffstat (limited to 'drivers/infiniband/sw/rdmavt')
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/mr.c | 21 | ||||
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/pd.c | 29 | ||||
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/pd.h | 7 | ||||
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/qp.c | 104 | ||||
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/rc.c | 13 | ||||
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/srq.c | 5 | ||||
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/trace_cq.h | 10 | ||||
| -rw-r--r-- | drivers/infiniband/sw/rdmavt/vt.c | 34 |
8 files changed, 110 insertions, 113 deletions
diff --git a/drivers/infiniband/sw/rdmavt/mr.c b/drivers/infiniband/sw/rdmavt/mr.c index 49c9541050d4..728795043496 100644 --- a/drivers/infiniband/sw/rdmavt/mr.c +++ b/drivers/infiniband/sw/rdmavt/mr.c @@ -381,15 +381,14 @@ struct ib_mr *rvt_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, { struct rvt_mr *mr; struct ib_umem *umem; - struct scatterlist *sg; - int n, m, entry; + struct sg_page_iter sg_iter; + int n, m; struct ib_mr *ret; if (length == 0) return ERR_PTR(-EINVAL); - umem = ib_umem_get(pd->uobject->context, start, length, - mr_access_flags, 0); + umem = ib_umem_get(udata, start, length, mr_access_flags, 0); if (IS_ERR(umem)) return (void *)umem; @@ -408,23 +407,21 @@ struct ib_mr *rvt_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, mr->mr.access_flags = mr_access_flags; mr->umem = umem; - mr->mr.page_shift = umem->page_shift; + mr->mr.page_shift = PAGE_SHIFT; m = 0; n = 0; - for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) { + for_each_sg_page (umem->sg_head.sgl, &sg_iter, umem->nmap, 0) { void *vaddr; - vaddr = page_address(sg_page(sg)); + vaddr = page_address(sg_page_iter_page(&sg_iter)); if (!vaddr) { ret = ERR_PTR(-EINVAL); goto bail_inval; } mr->mr.map[m]->segs[n].vaddr = vaddr; - mr->mr.map[m]->segs[n].length = BIT(umem->page_shift); - trace_rvt_mr_user_seg(&mr->mr, m, n, vaddr, - BIT(umem->page_shift)); - n++; - if (n == RVT_SEGSZ) { + mr->mr.map[m]->segs[n].length = PAGE_SIZE; + trace_rvt_mr_user_seg(&mr->mr, m, n, vaddr, PAGE_SIZE); + if (++n == RVT_SEGSZ) { m++; n = 0; } diff --git a/drivers/infiniband/sw/rdmavt/pd.c b/drivers/infiniband/sw/rdmavt/pd.c index 8a89afff3363..6033054b22fa 100644 --- a/drivers/infiniband/sw/rdmavt/pd.c +++ b/drivers/infiniband/sw/rdmavt/pd.c @@ -50,7 +50,7 @@ /** * rvt_alloc_pd - allocate a protection domain - * @ibdev: ib device + * @ibpd: PD * @context: optional user context * @udata: optional user data * @@ -58,19 +58,14 @@ * * Return: 0 on success */ -struct ib_pd *rvt_alloc_pd(struct ib_device *ibdev, - struct ib_ucontext *context, - struct ib_udata *udata) +int rvt_alloc_pd(struct ib_pd *ibpd, struct ib_ucontext *context, + struct ib_udata *udata) { + struct ib_device *ibdev = ibpd->device; struct rvt_dev_info *dev = ib_to_rvt(ibdev); - struct rvt_pd *pd; - struct ib_pd *ret; + struct rvt_pd *pd = ibpd_to_rvtpd(ibpd); + int ret = 0; - pd = kmalloc(sizeof(*pd), GFP_KERNEL); - if (!pd) { - ret = ERR_PTR(-ENOMEM); - goto bail; - } /* * While we could continue allocating protecetion domains, being * constrained only by system resources. The IBTA spec defines that @@ -81,8 +76,7 @@ struct ib_pd *rvt_alloc_pd(struct ib_device *ibdev, spin_lock(&dev->n_pds_lock); if (dev->n_pds_allocated == dev->dparms.props.max_pd) { spin_unlock(&dev->n_pds_lock); - kfree(pd); - ret = ERR_PTR(-ENOMEM); + ret = -ENOMEM; goto bail; } @@ -92,8 +86,6 @@ struct ib_pd *rvt_alloc_pd(struct ib_device *ibdev, /* ib_alloc_pd() will initialize pd->ibpd. */ pd->user = !!udata; - ret = &pd->ibpd; - bail: return ret; } @@ -104,16 +96,11 @@ bail: * * Return: always 0 */ -int rvt_dealloc_pd(struct ib_pd *ibpd) +void rvt_dealloc_pd(struct ib_pd *ibpd) { - struct rvt_pd *pd = ibpd_to_rvtpd(ibpd); struct rvt_dev_info *dev = ib_to_rvt(ibpd->device); spin_lock(&dev->n_pds_lock); dev->n_pds_allocated--; spin_unlock(&dev->n_pds_lock); - - kfree(pd); - - return 0; } diff --git a/drivers/infiniband/sw/rdmavt/pd.h b/drivers/infiniband/sw/rdmavt/pd.h index 1892ca4a9746..7a887e4a45e7 100644 --- a/drivers/infiniband/sw/rdmavt/pd.h +++ b/drivers/infiniband/sw/rdmavt/pd.h @@ -50,9 +50,8 @@ #include <rdma/rdma_vt.h> -struct ib_pd *rvt_alloc_pd(struct ib_device *ibdev, - struct ib_ucontext *context, - struct ib_udata *udata); -int rvt_dealloc_pd(struct ib_pd *ibpd); +int rvt_alloc_pd(struct ib_pd *pd, struct ib_ucontext *context, + struct ib_udata *udata); +void rvt_dealloc_pd(struct ib_pd *ibpd); #endif /* DEF_RDMAVTPD_H */ diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c index c6cc3e4ab71d..a34b9a2a32b6 100644 --- a/drivers/infiniband/sw/rdmavt/qp.c +++ b/drivers/infiniband/sw/rdmavt/qp.c @@ -53,6 +53,7 @@ #include <rdma/ib_verbs.h> #include <rdma/ib_hdrs.h> #include <rdma/opa_addr.h> +#include <rdma/uverbs_ioctl.h> #include "qp.h" #include "vt.h" #include "trace.h" @@ -854,6 +855,7 @@ static void rvt_init_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp, qp->s_mig_state = IB_MIG_MIGRATED; qp->r_head_ack_queue = 0; qp->s_tail_ack_queue = 0; + qp->s_acked_ack_queue = 0; qp->s_num_rd_atomic = 0; if (qp->r_rq.wq) { qp->r_rq.wq->head = 0; @@ -955,6 +957,8 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd, size_t sg_list_sz; struct ib_qp *ret = ERR_PTR(-ENOMEM); struct rvt_dev_info *rdi = ib_to_rvt(ibpd->device); + struct rvt_ucontext *ucontext = rdma_udata_to_drv_context( + udata, struct rvt_ucontext, ibucontext); void *priv = NULL; size_t sqsize; @@ -1128,7 +1132,7 @@ struct ib_qp *rvt_create_qp(struct ib_pd *ibpd, u32 s = sizeof(struct rvt_rwq) + qp->r_rq.size * sz; qp->ip = rvt_create_mmap_info(rdi, s, - ibpd->uobject->context, + &ucontext->ibucontext, qp->r_rq.wq); if (!qp->ip) { ret = ERR_PTR(-ENOMEM); @@ -1642,11 +1646,11 @@ int rvt_destroy_qp(struct ib_qp *ibqp) kref_put(&qp->ip->ref, rvt_release_mmap_info); else vfree(qp->r_rq.wq); - vfree(qp->s_wq); rdi->driver_f.qp_priv_free(rdi, qp); kfree(qp->s_ack_queue); rdma_destroy_ah_attr(&qp->remote_ah_attr); rdma_destroy_ah_attr(&qp->alt_ah_attr); + vfree(qp->s_wq); kfree(qp); return 0; } @@ -2393,11 +2397,12 @@ static inline unsigned long rvt_aeth_to_usec(u32 aeth) } /* - * rvt_add_retry_timer - add/start a retry timer + * rvt_add_retry_timer_ext - add/start a retry timer * @qp - the QP + * @shift - timeout shift to wait for multiple packets * add a retry timer on the QP */ -void rvt_add_retry_timer(struct rvt_qp *qp) +void rvt_add_retry_timer_ext(struct rvt_qp *qp, u8 shift) { struct ib_qp *ibqp = &qp->ibqp; struct rvt_dev_info *rdi = ib_to_rvt(ibqp->device); @@ -2405,11 +2410,11 @@ void rvt_add_retry_timer(struct rvt_qp *qp) lockdep_assert_held(&qp->s_lock); qp->s_flags |= RVT_S_TIMER; /* 4.096 usec. * (1 << qp->timeout) */ - qp->s_timer.expires = jiffies + qp->timeout_jiffies + - rdi->busy_jiffies; + qp->s_timer.expires = jiffies + rdi->busy_jiffies + + (qp->timeout_jiffies << shift); add_timer(&qp->s_timer); } -EXPORT_SYMBOL(rvt_add_retry_timer); +EXPORT_SYMBOL(rvt_add_retry_timer_ext); /** * rvt_add_rnr_timer - add/start an rnr timer @@ -2785,6 +2790,18 @@ again: } EXPORT_SYMBOL(rvt_copy_sge); +static enum ib_wc_status loopback_qp_drop(struct rvt_ibport *rvp, + struct rvt_qp *sqp) +{ + rvp->n_pkt_drops++; + /* + * For RC, the requester would timeout and retry so + * shortcut the timeouts and just signal too many retries. + */ + return sqp->ibqp.qp_type == IB_QPT_RC ? + IB_WC_RETRY_EXC_ERR : IB_WC_SUCCESS; +} + /** * ruc_loopback - handle UC and RC loopback requests * @sqp: the sending QP @@ -2857,17 +2874,14 @@ again: } spin_unlock_irqrestore(&sqp->s_lock, flags); - if (!qp || !(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) || + if (!qp) { + send_status = loopback_qp_drop(rvp, sqp); + goto serr_no_r_lock; + } + spin_lock_irqsave(&qp->r_lock, flags); + if (!(ib_rvt_state_ops[qp->state] & RVT_PROCESS_RECV_OK) || qp->ibqp.qp_type != sqp->ibqp.qp_type) { - rvp->n_pkt_drops++; - /* - * For RC, the requester would timeout and retry so - * shortcut the timeouts and just signal too many retries. - */ - if (sqp->ibqp.qp_type == IB_QPT_RC) - send_status = IB_WC_RETRY_EXC_ERR; - else - send_status = IB_WC_SUCCESS; + send_status = loopback_qp_drop(rvp, sqp); goto serr; } @@ -2893,18 +2907,8 @@ again: goto send_comp; case IB_WR_SEND_WITH_INV: - if (!rvt_invalidate_rkey(qp, wqe->wr.ex.invalidate_rkey)) { - wc.wc_flags = IB_WC_WITH_INVALIDATE; - wc.ex.invalidate_rkey = wqe->wr.ex.invalidate_rkey; - } - goto send; - case IB_WR_SEND_WITH_IMM: - wc.wc_flags = IB_WC_WITH_IMM; - wc.ex.imm_data = wqe->wr.ex.imm_data; - /* FALLTHROUGH */ case IB_WR_SEND: -send: ret = rvt_get_rwqe(qp, false); if (ret < 0) goto op_err; @@ -2912,6 +2916,22 @@ send: goto rnr_nak; if (wqe->length > qp->r_len) goto inv_err; + switch (wqe->wr.opcode) { + case IB_WR_SEND_WITH_INV: + if (!rvt_invalidate_rkey(qp, + wqe->wr.ex.invalidate_rkey)) { + wc.wc_flags = IB_WC_WITH_INVALIDATE; + wc.ex.invalidate_rkey = + wqe->wr.ex.invalidate_rkey; + } + break; + case IB_WR_SEND_WITH_IMM: + wc.wc_flags = IB_WC_WITH_IMM; + wc.ex.imm_data = wqe->wr.ex.imm_data; + break; + default: + break; + } break; case IB_WR_RDMA_WRITE_WITH_IMM: @@ -2988,34 +3008,12 @@ do_write: sge = &sqp->s_sge.sge; while (sqp->s_len) { - u32 len = sqp->s_len; + u32 len = rvt_get_sge_length(sge, sqp->s_len); - if (len > sge->length) - len = sge->length; - if (len > sge->sge_length) - len = sge->sge_length; WARN_ON_ONCE(len == 0); rvt_copy_sge(qp, &qp->r_sge, sge->vaddr, len, release, copy_last); - sge->vaddr += len; - sge->length -= len; - sge->sge_length -= len; - if (sge->sge_length == 0) { - if (!release) - rvt_put_mr(sge->mr); - if (--sqp->s_sge.num_sge) - *sge = *sqp->s_sge.sg_list++; - } else if (sge->length == 0 && sge->mr->lkey) { - if (++sge->n >= RVT_SEGSZ) { - if (++sge->m >= sge->mr->mapsz) - break; - sge->n = 0; - } - sge->vaddr = - sge->mr->map[sge->m]->segs[sge->n].vaddr; - sge->length = - sge->mr->map[sge->m]->segs[sge->n].length; - } + rvt_update_sge(&sqp->s_sge, len, !release); sqp->s_len -= len; } if (release) @@ -3041,6 +3039,7 @@ do_write: wqe->wr.send_flags & IB_SEND_SOLICITED); send_comp: + spin_unlock_irqrestore(&qp->r_lock, flags); spin_lock_irqsave(&sqp->s_lock, flags); rvp->n_loop_pkts++; flush_send: @@ -3067,6 +3066,7 @@ rnr_nak: } if (sqp->s_rnr_retry_cnt < 7) sqp->s_rnr_retry--; + spin_unlock_irqrestore(&qp->r_lock, flags); spin_lock_irqsave(&sqp->s_lock, flags); if (!(ib_rvt_state_ops[sqp->state] & RVT_PROCESS_RECV_OK)) goto clr_busy; @@ -3095,6 +3095,8 @@ err: rvt_rc_error(qp, wc.status); serr: + spin_unlock_irqrestore(&qp->r_lock, flags); +serr_no_r_lock: spin_lock_irqsave(&sqp->s_lock, flags); rvt_send_complete(sqp, wqe, send_status); if (sqp->ibqp.qp_type == IB_QPT_RC) { diff --git a/drivers/infiniband/sw/rdmavt/rc.c b/drivers/infiniband/sw/rdmavt/rc.c index 6131cc558bdb..8d71647820a8 100644 --- a/drivers/infiniband/sw/rdmavt/rc.c +++ b/drivers/infiniband/sw/rdmavt/rc.c @@ -187,3 +187,16 @@ void rvt_get_credit(struct rvt_qp *qp, u32 aeth) } } EXPORT_SYMBOL(rvt_get_credit); + +/* rvt_restart_sge - rewind the sge state for a wqe */ +u32 rvt_restart_sge(struct rvt_sge_state *ss, struct rvt_swqe *wqe, u32 len) +{ + ss->sge = wqe->sg_list[0]; + ss->sg_list = wqe->sg_list + 1; + ss->num_sge = wqe->wr.num_sge; + ss->total_len = wqe->length; + rvt_skip_sge(ss, len, false); + return wqe->length - len; +} +EXPORT_SYMBOL(rvt_restart_sge); + diff --git a/drivers/infiniband/sw/rdmavt/srq.c b/drivers/infiniband/sw/rdmavt/srq.c index 78e06fc456c5..895b3fabd0bf 100644 --- a/drivers/infiniband/sw/rdmavt/srq.c +++ b/drivers/infiniband/sw/rdmavt/srq.c @@ -48,6 +48,7 @@ #include <linux/err.h> #include <linux/slab.h> #include <linux/vmalloc.h> +#include <rdma/uverbs_ioctl.h> #include "srq.h" #include "vt.h" @@ -77,6 +78,8 @@ struct ib_srq *rvt_create_srq(struct ib_pd *ibpd, struct ib_udata *udata) { struct rvt_dev_info *dev = ib_to_rvt(ibpd->device); + struct rvt_ucontext *ucontext = rdma_udata_to_drv_context( + udata, struct rvt_ucontext, ibucontext); struct rvt_srq *srq; u32 sz; struct ib_srq *ret; @@ -119,7 +122,7 @@ struct ib_srq *rvt_create_srq(struct ib_pd *ibpd, u32 s = sizeof(struct rvt_rwq) + srq->rq.size * sz; srq->ip = - rvt_create_mmap_info(dev, s, ibpd->uobject->context, + rvt_create_mmap_info(dev, s, &ucontext->ibucontext, srq->rq.wq); if (!srq->ip) { ret = ERR_PTR(-ENOMEM); diff --git a/drivers/infiniband/sw/rdmavt/trace_cq.h b/drivers/infiniband/sw/rdmavt/trace_cq.h index df8e1adbef9d..e3c416c6f900 100644 --- a/drivers/infiniband/sw/rdmavt/trace_cq.h +++ b/drivers/infiniband/sw/rdmavt/trace_cq.h @@ -105,7 +105,7 @@ DEFINE_EVENT(rvt_cq_template, rvt_create_cq, TP_ARGS(cq, attr)); #define CQ_PRN \ -"[%s] idx %u wr_id %llx status %u opcode %u,%s length %u qpn %x" +"[%s] idx %u wr_id %llx status %u opcode %u,%s length %u qpn %x flags %x imm %x" DECLARE_EVENT_CLASS( rvt_cq_entry_template, @@ -119,6 +119,8 @@ DECLARE_EVENT_CLASS( __field(u32, qpn) __field(u32, length) __field(u32, idx) + __field(u32, flags) + __field(u32, imm) ), TP_fast_assign( RDI_DEV_ASSIGN(cq->rdi) @@ -128,6 +130,8 @@ DECLARE_EVENT_CLASS( __entry->length = wc->byte_len; __entry->qpn = wc->qp->qp_num; __entry->idx = idx; + __entry->flags = wc->wc_flags; + __entry->imm = be32_to_cpu(wc->ex.imm_data); ), TP_printk( CQ_PRN, @@ -137,7 +141,9 @@ DECLARE_EVENT_CLASS( __entry->status, __entry->opcode, show_wc_opcode(__entry->opcode), __entry->length, - __entry->qpn + __entry->qpn, + __entry->flags, + __entry->imm ) ); diff --git a/drivers/infiniband/sw/rdmavt/vt.c b/drivers/infiniband/sw/rdmavt/vt.c index aef3aa3fe667..42c9d35f832d 100644 --- a/drivers/infiniband/sw/rdmavt/vt.c +++ b/drivers/infiniband/sw/rdmavt/vt.c @@ -91,7 +91,7 @@ struct rvt_dev_info *rvt_alloc_device(size_t size, int nports) { struct rvt_dev_info *rdi; - rdi = (struct rvt_dev_info *)ib_alloc_device(size); + rdi = container_of(_ib_alloc_device(size), struct rvt_dev_info, ibdev); if (!rdi) return rdi; @@ -284,10 +284,6 @@ static int rvt_query_gid(struct ib_device *ibdev, u8 port_num, &gid->global.interface_id); } -struct rvt_ucontext { - struct ib_ucontext ibucontext; -}; - static inline struct rvt_ucontext *to_iucontext(struct ib_ucontext *ibucontext) { @@ -296,28 +292,21 @@ static inline struct rvt_ucontext *to_iucontext(struct ib_ucontext /** * rvt_alloc_ucontext - Allocate a user context - * @ibdev: Verbs IB dev + * @uctx: Verbs context * @udata: User data allocated */ -static struct ib_ucontext *rvt_alloc_ucontext(struct ib_device *ibdev, - struct ib_udata *udata) +static int rvt_alloc_ucontext(struct ib_ucontext *uctx, struct ib_udata *udata) { - struct rvt_ucontext *context; - - context = kmalloc(sizeof(*context), GFP_KERNEL); - if (!context) - return ERR_PTR(-ENOMEM); - return &context->ibucontext; + return 0; } /** - *rvt_dealloc_ucontext - Free a user context - *@context - Free this + * rvt_dealloc_ucontext - Free a user context + * @context - Free this */ -static int rvt_dealloc_ucontext(struct ib_ucontext *context) +static void rvt_dealloc_ucontext(struct ib_ucontext *context) { - kfree(to_iucontext(context)); - return 0; + return; } static int rvt_get_port_immutable(struct ib_device *ibdev, u8 port_num, @@ -436,6 +425,8 @@ static const struct ib_device_ops rvt_dev_ops = { .req_notify_cq = rvt_req_notify_cq, .resize_cq = rvt_resize_cq, .unmap_fmr = rvt_unmap_fmr, + INIT_RDMA_OBJ_SIZE(ib_pd, rvt_pd, ibpd), + INIT_RDMA_OBJ_SIZE(ib_ucontext, rvt_ucontext, ibucontext), }; static noinline int check_support(struct rvt_dev_info *rdi, int verb) @@ -446,7 +437,7 @@ static noinline int check_support(struct rvt_dev_info *rdi, int verb) * These functions are not part of verbs specifically but are * required for rdmavt to function. */ - if ((!rdi->driver_f.port_callback) || + if ((!rdi->ibdev.ops.init_port) || (!rdi->driver_f.get_pci_dev)) return -EINVAL; break; @@ -644,8 +635,7 @@ int rvt_register_device(struct rvt_dev_info *rdi, u32 driver_id) rdi->ibdev.driver_id = driver_id; /* We are now good to announce we exist */ - ret = ib_register_device(&rdi->ibdev, dev_name(&rdi->ibdev.dev), - rdi->driver_f.port_callback); + ret = ib_register_device(&rdi->ibdev, dev_name(&rdi->ibdev.dev)); if (ret) { rvt_pr_err(rdi, "Failed to register driver with ib core.\n"); goto bail_wss; |