diff options
Diffstat (limited to 'drivers/infiniband/hw')
22 files changed, 198 insertions, 196 deletions
diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c index e87fc0408470..347dc242fb88 100644 --- a/drivers/infiniband/hw/cxgb4/cm.c +++ b/drivers/infiniband/hw/cxgb4/cm.c @@ -495,7 +495,6 @@ static int _put_ep_safe(struct c4iw_dev *dev, struct sk_buff *skb) ep = *((struct c4iw_ep **)(skb->cb + 2 * sizeof(void *))); release_ep_resources(ep); - kfree_skb(skb); return 0; } @@ -506,7 +505,6 @@ static int _put_pass_ep_safe(struct c4iw_dev *dev, struct sk_buff *skb) ep = *((struct c4iw_ep **)(skb->cb + 2 * sizeof(void *))); c4iw_put_ep(&ep->parent_ep->com); release_ep_resources(ep); - kfree_skb(skb); return 0; } @@ -2424,20 +2422,6 @@ static int accept_cr(struct c4iw_ep *ep, struct sk_buff *skb, enum chip_type adapter_type = ep->com.dev->rdev.lldi.adapter_type; pr_debug("ep %p tid %u\n", ep, ep->hwtid); - - skb_get(skb); - rpl = cplhdr(skb); - if (!is_t4(adapter_type)) { - skb_trim(skb, roundup(sizeof(*rpl5), 16)); - rpl5 = (void *)rpl; - INIT_TP_WR(rpl5, ep->hwtid); - } else { - skb_trim(skb, sizeof(*rpl)); - INIT_TP_WR(rpl, ep->hwtid); - } - OPCODE_TID(rpl) = cpu_to_be32(MK_OPCODE_TID(CPL_PASS_ACCEPT_RPL, - ep->hwtid)); - cxgb_best_mtu(ep->com.dev->rdev.lldi.mtus, ep->mtu, &mtu_idx, enable_tcp_timestamps && req->tcpopt.tstamp, (ep->com.remote_addr.ss_family == AF_INET) ? 0 : 1); @@ -2483,6 +2467,20 @@ static int accept_cr(struct c4iw_ep *ep, struct sk_buff *skb, if (tcph->ece && tcph->cwr) opt2 |= CCTRL_ECN_V(1); } + + skb_get(skb); + rpl = cplhdr(skb); + if (!is_t4(adapter_type)) { + skb_trim(skb, roundup(sizeof(*rpl5), 16)); + rpl5 = (void *)rpl; + INIT_TP_WR(rpl5, ep->hwtid); + } else { + skb_trim(skb, sizeof(*rpl)); + INIT_TP_WR(rpl, ep->hwtid); + } + OPCODE_TID(rpl) = cpu_to_be32(MK_OPCODE_TID(CPL_PASS_ACCEPT_RPL, + ep->hwtid)); + if (CHELSIO_CHIP_VERSION(adapter_type) > CHELSIO_T4) { u32 isn = (prandom_u32() & ~7UL) - 1; opt2 |= T5_OPT_2_VALID_F; diff --git a/drivers/infiniband/hw/cxgb4/device.c b/drivers/infiniband/hw/cxgb4/device.c index a8b9548bd1a2..599340c1f0b8 100644 --- a/drivers/infiniband/hw/cxgb4/device.c +++ b/drivers/infiniband/hw/cxgb4/device.c @@ -242,10 +242,13 @@ static void set_ep_sin6_addrs(struct c4iw_ep *ep, } } -static int dump_qp(struct c4iw_qp *qp, struct c4iw_debugfs_data *qpd) +static int dump_qp(unsigned long id, struct c4iw_qp *qp, + struct c4iw_debugfs_data *qpd) { int space; int cc; + if (id != qp->wq.sq.qid) + return 0; space = qpd->bufsize - qpd->pos - 1; if (space == 0) @@ -350,7 +353,7 @@ static int qp_open(struct inode *inode, struct file *file) xa_lock_irq(&qpd->devp->qps); xa_for_each(&qpd->devp->qps, index, qp) - dump_qp(qp, qpd); + dump_qp(index, qp, qpd); xa_unlock_irq(&qpd->devp->qps); qpd->buf[qpd->pos++] = 0; diff --git a/drivers/infiniband/hw/cxgb4/mem.c b/drivers/infiniband/hw/cxgb4/mem.c index aa772ee0706f..35c284af574d 100644 --- a/drivers/infiniband/hw/cxgb4/mem.c +++ b/drivers/infiniband/hw/cxgb4/mem.c @@ -275,13 +275,17 @@ static int write_tpt_entry(struct c4iw_rdev *rdev, u32 reset_tpt_entry, struct sk_buff *skb, struct c4iw_wr_wait *wr_waitp) { int err; - struct fw_ri_tpte tpt; + struct fw_ri_tpte *tpt; u32 stag_idx; static atomic_t key; if (c4iw_fatal_error(rdev)) return -EIO; + tpt = kmalloc(sizeof(*tpt), GFP_KERNEL); + if (!tpt) + return -ENOMEM; + stag_state = stag_state > 0; stag_idx = (*stag) >> 8; @@ -291,6 +295,7 @@ static int write_tpt_entry(struct c4iw_rdev *rdev, u32 reset_tpt_entry, mutex_lock(&rdev->stats.lock); rdev->stats.stag.fail++; mutex_unlock(&rdev->stats.lock); + kfree(tpt); return -ENOMEM; } mutex_lock(&rdev->stats.lock); @@ -305,28 +310,28 @@ static int write_tpt_entry(struct c4iw_rdev *rdev, u32 reset_tpt_entry, /* write TPT entry */ if (reset_tpt_entry) - memset(&tpt, 0, sizeof(tpt)); + memset(tpt, 0, sizeof(*tpt)); else { - tpt.valid_to_pdid = cpu_to_be32(FW_RI_TPTE_VALID_F | + tpt->valid_to_pdid = cpu_to_be32(FW_RI_TPTE_VALID_F | FW_RI_TPTE_STAGKEY_V((*stag & FW_RI_TPTE_STAGKEY_M)) | FW_RI_TPTE_STAGSTATE_V(stag_state) | FW_RI_TPTE_STAGTYPE_V(type) | FW_RI_TPTE_PDID_V(pdid)); - tpt.locread_to_qpid = cpu_to_be32(FW_RI_TPTE_PERM_V(perm) | + tpt->locread_to_qpid = cpu_to_be32(FW_RI_TPTE_PERM_V(perm) | (bind_enabled ? FW_RI_TPTE_MWBINDEN_F : 0) | FW_RI_TPTE_ADDRTYPE_V((zbva ? FW_RI_ZERO_BASED_TO : FW_RI_VA_BASED_TO))| FW_RI_TPTE_PS_V(page_size)); - tpt.nosnoop_pbladdr = !pbl_size ? 0 : cpu_to_be32( + tpt->nosnoop_pbladdr = !pbl_size ? 0 : cpu_to_be32( FW_RI_TPTE_PBLADDR_V(PBL_OFF(rdev, pbl_addr)>>3)); - tpt.len_lo = cpu_to_be32((u32)(len & 0xffffffffUL)); - tpt.va_hi = cpu_to_be32((u32)(to >> 32)); - tpt.va_lo_fbo = cpu_to_be32((u32)(to & 0xffffffffUL)); - tpt.dca_mwbcnt_pstag = cpu_to_be32(0); - tpt.len_hi = cpu_to_be32((u32)(len >> 32)); + tpt->len_lo = cpu_to_be32((u32)(len & 0xffffffffUL)); + tpt->va_hi = cpu_to_be32((u32)(to >> 32)); + tpt->va_lo_fbo = cpu_to_be32((u32)(to & 0xffffffffUL)); + tpt->dca_mwbcnt_pstag = cpu_to_be32(0); + tpt->len_hi = cpu_to_be32((u32)(len >> 32)); } err = write_adapter_mem(rdev, stag_idx + (rdev->lldi.vr->stag.start >> 5), - sizeof(tpt), &tpt, skb, wr_waitp); + sizeof(*tpt), tpt, skb, wr_waitp); if (reset_tpt_entry) { c4iw_put_resource(&rdev->resource.tpt_table, stag_idx); @@ -334,6 +339,7 @@ static int write_tpt_entry(struct c4iw_rdev *rdev, u32 reset_tpt_entry, rdev->stats.stag.cur -= 32; mutex_unlock(&rdev->stats.lock); } + kfree(tpt); return err; } diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index eb9368be28c1..bbcac539777a 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -2737,15 +2737,11 @@ int c4iw_create_srq(struct ib_srq *ib_srq, struct ib_srq_init_attr *attrs, if (CHELSIO_CHIP_VERSION(rhp->rdev.lldi.adapter_type) > CHELSIO_T6) srq->flags = T4_SRQ_LIMIT_SUPPORT; - ret = xa_insert_irq(&rhp->qps, srq->wq.qid, srq, GFP_KERNEL); - if (ret) - goto err_free_queue; - if (udata) { srq_key_mm = kmalloc(sizeof(*srq_key_mm), GFP_KERNEL); if (!srq_key_mm) { ret = -ENOMEM; - goto err_remove_handle; + goto err_free_queue; } srq_db_key_mm = kmalloc(sizeof(*srq_db_key_mm), GFP_KERNEL); if (!srq_db_key_mm) { @@ -2789,8 +2785,6 @@ err_free_srq_db_key_mm: kfree(srq_db_key_mm); err_free_srq_key_mm: kfree(srq_key_mm); -err_remove_handle: - xa_erase_irq(&rhp->qps, srq->wq.qid); err_free_queue: free_srq_queue(srq, ucontext ? &ucontext->uctx : &rhp->rdev.uctx, srq->wr_waitp); @@ -2813,8 +2807,6 @@ void c4iw_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata) rhp = srq->rhp; pr_debug("%s id %d\n", __func__, srq->wq.qid); - - xa_erase_irq(&rhp->qps, srq->wq.qid); ucontext = rdma_udata_to_drv_context(udata, struct c4iw_ucontext, ibucontext); free_srq_queue(srq, ucontext ? &ucontext->uctx : &rhp->rdev.uctx, diff --git a/drivers/infiniband/hw/hfi1/init.c b/drivers/infiniband/hw/hfi1/init.c index 71cb9525c074..26b792bb1027 100644 --- a/drivers/infiniband/hw/hfi1/init.c +++ b/drivers/infiniband/hw/hfi1/init.c @@ -1489,7 +1489,6 @@ static int __init hfi1_mod_init(void) goto bail_dev; } - hfi1_compute_tid_rdma_flow_wt(); /* * These must be called before the driver is registered with * the PCI subsystem. diff --git a/drivers/infiniband/hw/hfi1/pcie.c b/drivers/infiniband/hw/hfi1/pcie.c index 61aa5504d7c3..61362bd6d3ce 100644 --- a/drivers/infiniband/hw/hfi1/pcie.c +++ b/drivers/infiniband/hw/hfi1/pcie.c @@ -319,7 +319,9 @@ int pcie_speeds(struct hfi1_devdata *dd) /* * bus->max_bus_speed is set from the bridge's linkcap Max Link Speed */ - if (parent && dd->pcidev->bus->max_bus_speed != PCIE_SPEED_8_0GT) { + if (parent && + (dd->pcidev->bus->max_bus_speed == PCIE_SPEED_2_5GT || + dd->pcidev->bus->max_bus_speed == PCIE_SPEED_5_0GT)) { dd_dev_info(dd, "Parent PCIe bridge does not support Gen3\n"); dd->link_gen3_capable = 0; } diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c index 513a8aac9ccd..1a3c647675a7 100644 --- a/drivers/infiniband/hw/hfi1/rc.c +++ b/drivers/infiniband/hw/hfi1/rc.c @@ -2209,15 +2209,15 @@ int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode, if (qp->s_flags & RVT_S_WAIT_RNR) goto bail_stop; rdi = ib_to_rvt(qp->ibqp.device); - if (qp->s_rnr_retry == 0 && - !((rdi->post_parms[wqe->wr.opcode].flags & - RVT_OPERATION_IGN_RNR_CNT) && - qp->s_rnr_retry_cnt == 0)) { - status = IB_WC_RNR_RETRY_EXC_ERR; - goto class_b; + if (!(rdi->post_parms[wqe->wr.opcode].flags & + RVT_OPERATION_IGN_RNR_CNT)) { + if (qp->s_rnr_retry == 0) { + status = IB_WC_RNR_RETRY_EXC_ERR; + goto class_b; + } + if (qp->s_rnr_retry_cnt < 7 && qp->s_rnr_retry_cnt > 0) + qp->s_rnr_retry--; } - if (qp->s_rnr_retry_cnt < 7 && qp->s_rnr_retry_cnt > 0) - qp->s_rnr_retry--; /* * The last valid PSN is the previous PSN. For TID RDMA WRITE diff --git a/drivers/infiniband/hw/hfi1/sdma.c b/drivers/infiniband/hw/hfi1/sdma.c index 2395fd4233a7..c61b6022575e 100644 --- a/drivers/infiniband/hw/hfi1/sdma.c +++ b/drivers/infiniband/hw/hfi1/sdma.c @@ -65,6 +65,7 @@ #define SDMA_DESCQ_CNT 2048 #define SDMA_DESC_INTR 64 #define INVALID_TAIL 0xffff +#define SDMA_PAD max_t(size_t, MAX_16B_PADDING, sizeof(u32)) static uint sdma_descq_cnt = SDMA_DESCQ_CNT; module_param(sdma_descq_cnt, uint, S_IRUGO); @@ -1296,7 +1297,7 @@ void sdma_clean(struct hfi1_devdata *dd, size_t num_engines) struct sdma_engine *sde; if (dd->sdma_pad_dma) { - dma_free_coherent(&dd->pcidev->dev, 4, + dma_free_coherent(&dd->pcidev->dev, SDMA_PAD, (void *)dd->sdma_pad_dma, dd->sdma_pad_phys); dd->sdma_pad_dma = NULL; @@ -1491,7 +1492,7 @@ int sdma_init(struct hfi1_devdata *dd, u8 port) } /* Allocate memory for pad */ - dd->sdma_pad_dma = dma_alloc_coherent(&dd->pcidev->dev, sizeof(u32), + dd->sdma_pad_dma = dma_alloc_coherent(&dd->pcidev->dev, SDMA_PAD, &dd->sdma_pad_phys, GFP_KERNEL); if (!dd->sdma_pad_dma) { dd_dev_err(dd, "failed to allocate SendDMA pad memory\n"); @@ -1526,8 +1527,11 @@ int sdma_init(struct hfi1_devdata *dd, u8 port) } ret = rhashtable_init(tmp_sdma_rht, &sdma_rht_params); - if (ret < 0) + if (ret < 0) { + kfree(tmp_sdma_rht); goto bail; + } + dd->sdma_rht = tmp_sdma_rht; dd_dev_info(dd, "SDMA num_sdma: %u\n", dd->num_sdma); diff --git a/drivers/infiniband/hw/hfi1/tid_rdma.c b/drivers/infiniband/hw/hfi1/tid_rdma.c index b4dcc4d29f84..e53f542b60af 100644 --- a/drivers/infiniband/hw/hfi1/tid_rdma.c +++ b/drivers/infiniband/hw/hfi1/tid_rdma.c @@ -107,8 +107,6 @@ static u32 mask_generation(u32 a) * C - Capcode */ -static u32 tid_rdma_flow_wt; - static void tid_rdma_trigger_resume(struct work_struct *work); static void hfi1_kern_exp_rcv_free_flows(struct tid_rdma_request *req); static int hfi1_kern_exp_rcv_alloc_flows(struct tid_rdma_request *req, @@ -136,6 +134,26 @@ static void update_r_next_psn_fecn(struct hfi1_packet *packet, struct tid_rdma_flow *flow, bool fecn); +static void validate_r_tid_ack(struct hfi1_qp_priv *priv) +{ + if (priv->r_tid_ack == HFI1_QP_WQE_INVALID) + priv->r_tid_ack = priv->r_tid_tail; +} + +static void tid_rdma_schedule_ack(struct rvt_qp *qp) +{ + struct hfi1_qp_priv *priv = qp->priv; + + priv->s_flags |= RVT_S_ACK_PENDING; + hfi1_schedule_tid_send(qp); +} + +static void tid_rdma_trigger_ack(struct rvt_qp *qp) +{ + validate_r_tid_ack(qp->priv); + tid_rdma_schedule_ack(qp); +} + static u64 tid_rdma_opfn_encode(struct tid_rdma_params *p) { return @@ -2736,11 +2754,6 @@ static bool handle_read_kdeth_eflags(struct hfi1_ctxtdata *rcd, diff = cmp_psn(psn, flow->flow_state.r_next_psn); if (diff > 0) { - if (!(qp->r_flags & RVT_R_RDMAR_SEQ)) - restart_tid_rdma_read_req(rcd, - qp, - wqe); - /* Drop the packet.*/ goto s_unlock; } else if (diff < 0) { @@ -3010,10 +3023,7 @@ nak_psn: qpriv->s_nak_state = IB_NAK_PSN_ERROR; /* We are NAK'ing the next expected PSN */ qpriv->s_nak_psn = mask_psn(flow->flow_state.r_next_psn); - qpriv->s_flags |= RVT_S_ACK_PENDING; - if (qpriv->r_tid_ack == HFI1_QP_WQE_INVALID) - qpriv->r_tid_ack = qpriv->r_tid_tail; - hfi1_schedule_tid_send(qp); + tid_rdma_trigger_ack(qp); } goto unlock; } @@ -3376,18 +3386,17 @@ u32 hfi1_build_tid_rdma_write_req(struct rvt_qp *qp, struct rvt_swqe *wqe, return sizeof(ohdr->u.tid_rdma.w_req) / sizeof(u32); } -void hfi1_compute_tid_rdma_flow_wt(void) +static u32 hfi1_compute_tid_rdma_flow_wt(struct rvt_qp *qp) { /* * Heuristic for computing the RNR timeout when waiting on the flow * queue. Rather than a computationaly expensive exact estimate of when * a flow will be available, we assume that if a QP is at position N in * the flow queue it has to wait approximately (N + 1) * (number of - * segments between two sync points), assuming PMTU of 4K. The rationale - * for this is that flows are released and recycled at each sync point. + * segments between two sync points). The rationale for this is that + * flows are released and recycled at each sync point. */ - tid_rdma_flow_wt = MAX_TID_FLOW_PSN * enum_to_mtu(OPA_MTU_4096) / - TID_RDMA_MAX_SEGMENT_SIZE; + return (MAX_TID_FLOW_PSN * qp->pmtu) >> TID_RDMA_SEGMENT_SHIFT; } static u32 position_in_queue(struct hfi1_qp_priv *qpriv, @@ -3510,7 +3519,7 @@ static void hfi1_tid_write_alloc_resources(struct rvt_qp *qp, bool intr_ctx) if (qpriv->flow_state.index >= RXE_NUM_TID_FLOWS) { ret = hfi1_kern_setup_hw_flow(qpriv->rcd, qp); if (ret) { - to_seg = tid_rdma_flow_wt * + to_seg = hfi1_compute_tid_rdma_flow_wt(qp) * position_in_queue(qpriv, &rcd->flow_queue); break; @@ -3531,7 +3540,7 @@ static void hfi1_tid_write_alloc_resources(struct rvt_qp *qp, bool intr_ctx) /* * If overtaking req->acked_tail, send an RNR NAK. Because the * QP is not queued in this case, and the issue can only be - * caused due a delay in scheduling the second leg which we + * caused by a delay in scheduling the second leg which we * cannot estimate, we use a rather arbitrary RNR timeout of * (MAX_FLOWS / 2) segments */ @@ -3539,8 +3548,7 @@ static void hfi1_tid_write_alloc_resources(struct rvt_qp *qp, bool intr_ctx) MAX_FLOWS)) { ret = -EAGAIN; to_seg = MAX_FLOWS >> 1; - qpriv->s_flags |= RVT_S_ACK_PENDING; - hfi1_schedule_tid_send(qp); + tid_rdma_trigger_ack(qp); break; } @@ -4340,8 +4348,7 @@ void hfi1_rc_rcv_tid_rdma_write_data(struct hfi1_packet *packet) trace_hfi1_tid_req_rcv_write_data(qp, 0, e->opcode, e->psn, e->lpsn, req); trace_hfi1_tid_write_rsp_rcv_data(qp); - if (priv->r_tid_ack == HFI1_QP_WQE_INVALID) - priv->r_tid_ack = priv->r_tid_tail; + validate_r_tid_ack(priv); if (opcode == TID_OP(WRITE_DATA_LAST)) { release_rdma_sge_mr(e); @@ -4380,8 +4387,7 @@ void hfi1_rc_rcv_tid_rdma_write_data(struct hfi1_packet *packet) } done: - priv->s_flags |= RVT_S_ACK_PENDING; - hfi1_schedule_tid_send(qp); + tid_rdma_schedule_ack(qp); exit: priv->r_next_psn_kdeth = flow->flow_state.r_next_psn; if (fecn) @@ -4393,10 +4399,7 @@ send_nak: if (!priv->s_nak_state) { priv->s_nak_state = IB_NAK_PSN_ERROR; priv->s_nak_psn = flow->flow_state.r_next_psn; - priv->s_flags |= RVT_S_ACK_PENDING; - if (priv->r_tid_ack == HFI1_QP_WQE_INVALID) - priv->r_tid_ack = priv->r_tid_tail; - hfi1_schedule_tid_send(qp); + tid_rdma_trigger_ack(qp); } goto done; } @@ -4944,8 +4947,7 @@ void hfi1_rc_rcv_tid_rdma_resync(struct hfi1_packet *packet) qpriv->resync = true; /* RESYNC request always gets a TID RDMA ACK. */ qpriv->s_nak_state = 0; - qpriv->s_flags |= RVT_S_ACK_PENDING; - hfi1_schedule_tid_send(qp); + tid_rdma_trigger_ack(qp); bail: if (fecn) qp->s_flags |= RVT_S_ECN; diff --git a/drivers/infiniband/hw/hfi1/tid_rdma.h b/drivers/infiniband/hw/hfi1/tid_rdma.h index 1c536185261e..6e82df2190b7 100644 --- a/drivers/infiniband/hw/hfi1/tid_rdma.h +++ b/drivers/infiniband/hw/hfi1/tid_rdma.h @@ -17,6 +17,7 @@ #define TID_RDMA_MIN_SEGMENT_SIZE BIT(18) /* 256 KiB (for now) */ #define TID_RDMA_MAX_SEGMENT_SIZE BIT(18) /* 256 KiB (for now) */ #define TID_RDMA_MAX_PAGES (BIT(18) >> PAGE_SHIFT) +#define TID_RDMA_SEGMENT_SHIFT 18 /* * Bit definitions for priv->s_flags. @@ -274,8 +275,6 @@ u32 hfi1_build_tid_rdma_write_req(struct rvt_qp *qp, struct rvt_swqe *wqe, struct ib_other_headers *ohdr, u32 *bth1, u32 *bth2, u32 *len); -void hfi1_compute_tid_rdma_flow_wt(void); - void hfi1_rc_rcv_tid_rdma_write_req(struct hfi1_packet *packet); u32 hfi1_build_tid_rdma_write_resp(struct rvt_qp *qp, struct rvt_ack_entry *e, diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c index 7bff0a1e713d..089e201d7550 100644 --- a/drivers/infiniband/hw/hfi1/verbs.c +++ b/drivers/infiniband/hw/hfi1/verbs.c @@ -147,9 +147,6 @@ static int pio_wait(struct rvt_qp *qp, /* Length of buffer to create verbs txreq cache name */ #define TXREQ_NAME_LEN 24 -/* 16B trailing buffer */ -static const u8 trail_buf[MAX_16B_PADDING]; - static uint wss_threshold = 80; module_param(wss_threshold, uint, S_IRUGO); MODULE_PARM_DESC(wss_threshold, "Percentage (1-100) of LLC to use as a threshold for a cacheless copy"); @@ -820,8 +817,8 @@ static int build_verbs_tx_desc( /* add icrc, lt byte, and padding to flit */ if (extra_bytes) - ret = sdma_txadd_kvaddr(sde->dd, &tx->txreq, - (void *)trail_buf, extra_bytes); + ret = sdma_txadd_daddr(sde->dd, &tx->txreq, + sde->dd->sdma_pad_phys, extra_bytes); bail_txadd: return ret; @@ -1089,7 +1086,8 @@ int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps, } /* add icrc, lt byte, and padding to flit */ if (extra_bytes) - seg_pio_copy_mid(pbuf, trail_buf, extra_bytes); + seg_pio_copy_mid(pbuf, ppd->dd->sdma_pad_dma, + extra_bytes); seg_pio_copy_end(pbuf); } diff --git a/drivers/infiniband/hw/hns/hns_roce_hem.h b/drivers/infiniband/hw/hns/hns_roce_hem.h index 86783276fb1f..3bb8f78fb7b0 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hem.h +++ b/drivers/infiniband/hw/hns/hns_roce_hem.h @@ -59,7 +59,7 @@ enum { #define HNS_ROCE_HEM_CHUNK_LEN \ ((256 - sizeof(struct list_head) - 2 * sizeof(int)) / \ - (sizeof(struct scatterlist))) + (sizeof(struct scatterlist) + sizeof(void *))) #define check_whether_bt_num_3(type, hop_num) \ (type < HEM_TYPE_MTT && hop_num == 2) diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c index 7a89d669f8bf..e82567fcdeb7 100644 --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c @@ -5389,9 +5389,9 @@ static void hns_roce_v2_free_eq(struct hns_roce_dev *hr_dev, return; } - if (eq->buf_list) - dma_free_coherent(hr_dev->dev, buf_chk_sz, - eq->buf_list->buf, eq->buf_list->map); + dma_free_coherent(hr_dev->dev, buf_chk_sz, eq->buf_list->buf, + eq->buf_list->map); + kfree(eq->buf_list); } static void hns_roce_config_eqc(struct hns_roce_dev *hr_dev, diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c index 9591457eb768..43ea2c13b212 100644 --- a/drivers/infiniband/hw/hns/hns_roce_srq.c +++ b/drivers/infiniband/hw/hns/hns_roce_srq.c @@ -376,7 +376,7 @@ int hns_roce_create_srq(struct ib_srq *ib_srq, srq->max = roundup_pow_of_two(srq_init_attr->attr.max_wr + 1); srq->max_gs = srq_init_attr->attr.max_sge; - srq_desc_size = max(16, 16 * srq->max_gs); + srq_desc_size = roundup_pow_of_two(max(16, 16 * srq->max_gs)); srq->wqe_shift = ilog2(srq_desc_size); diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c index 8056930bbe2c..cd9ee1664a69 100644 --- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c +++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c @@ -2773,6 +2773,10 @@ int i40iw_register_rdma_device(struct i40iw_device *iwdev) return -ENOMEM; iwibdev = iwdev->iwibdev; rdma_set_device_sysfs_group(&iwibdev->ibdev, &i40iw_attr_group); + ret = ib_device_set_netdev(&iwibdev->ibdev, iwdev->netdev, 1); + if (ret) + goto error; + ret = ib_register_device(&iwibdev->ibdev, "i40iw%d"); if (ret) goto error; diff --git a/drivers/infiniband/hw/mlx5/devx.c b/drivers/infiniband/hw/mlx5/devx.c index 59022b744144..d609f4659afb 100644 --- a/drivers/infiniband/hw/mlx5/devx.c +++ b/drivers/infiniband/hw/mlx5/devx.c @@ -1298,29 +1298,6 @@ static int devx_handle_mkey_create(struct mlx5_ib_dev *dev, return 0; } -static void devx_free_indirect_mkey(struct rcu_head *rcu) -{ - kfree(container_of(rcu, struct devx_obj, devx_mr.rcu)); -} - -/* This function to delete from the radix tree needs to be called before - * destroying the underlying mkey. Otherwise a race might occur in case that - * other thread will get the same mkey before this one will be deleted, - * in that case it will fail via inserting to the tree its own data. - * - * Note: - * An error in the destroy is not expected unless there is some other indirect - * mkey which points to this one. In a kernel cleanup flow it will be just - * destroyed in the iterative destruction call. In a user flow, in case - * the application didn't close in the expected order it's its own problem, - * the mkey won't be part of the tree, in both cases the kernel is safe. - */ -static void devx_cleanup_mkey(struct devx_obj *obj) -{ - xa_erase(&obj->ib_dev->mdev->priv.mkey_table, - mlx5_base_mkey(obj->devx_mr.mmkey.key)); -} - static void devx_cleanup_subscription(struct mlx5_ib_dev *dev, struct devx_event_subscription *sub) { @@ -1362,8 +1339,16 @@ static int devx_obj_cleanup(struct ib_uobject *uobject, int ret; dev = mlx5_udata_to_mdev(&attrs->driver_udata); - if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) - devx_cleanup_mkey(obj); + if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) { + /* + * The pagefault_single_data_segment() does commands against + * the mmkey, we must wait for that to stop before freeing the + * mkey, as another allocation could get the same mkey #. + */ + xa_erase(&obj->ib_dev->mdev->priv.mkey_table, + mlx5_base_mkey(obj->devx_mr.mmkey.key)); + synchronize_srcu(&dev->mr_srcu); + } if (obj->flags & DEVX_OBJ_FLAGS_DCT) ret = mlx5_core_destroy_dct(obj->ib_dev->mdev, &obj->core_dct); @@ -1382,12 +1367,6 @@ static int devx_obj_cleanup(struct ib_uobject *uobject, devx_cleanup_subscription(dev, sub_entry); mutex_unlock(&devx_event_table->event_xa_lock); - if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) { - call_srcu(&dev->mr_srcu, &obj->devx_mr.rcu, - devx_free_indirect_mkey); - return ret; - } - kfree(obj); return ret; } @@ -1491,26 +1470,21 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_DEVX_OBJ_CREATE)( &obj_id); WARN_ON(obj->dinlen > MLX5_MAX_DESTROY_INBOX_SIZE_DW * sizeof(u32)); - if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) { - err = devx_handle_mkey_indirect(obj, dev, cmd_in, cmd_out); - if (err) - goto obj_destroy; - } - err = uverbs_copy_to(attrs, MLX5_IB_ATTR_DEVX_OBJ_CREATE_CMD_OUT, cmd_out, cmd_out_len); if (err) - goto err_copy; + goto obj_destroy; if (opcode == MLX5_CMD_OP_CREATE_GENERAL_OBJECT) obj_type = MLX5_GET(general_obj_in_cmd_hdr, cmd_in, obj_type); - obj->obj_id = get_enc_obj_id(opcode | obj_type << 16, obj_id); + if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) { + err = devx_handle_mkey_indirect(obj, dev, cmd_in, cmd_out); + if (err) + goto obj_destroy; + } return 0; -err_copy: - if (obj->flags & DEVX_OBJ_FLAGS_INDIRECT_MKEY) - devx_cleanup_mkey(obj); obj_destroy: if (obj->flags & DEVX_OBJ_FLAGS_DCT) mlx5_core_destroy_dct(obj->ib_dev->mdev, &obj->core_dct); diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 2ceaef3ea3fb..1a98ee2e01c4 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -606,7 +606,7 @@ struct mlx5_ib_mr { struct mlx5_ib_dev *dev; u32 out[MLX5_ST_SZ_DW(create_mkey_out)]; struct mlx5_core_sig_ctx *sig; - int live; + unsigned int live; void *descs_alloc; int access_flags; /* Needed for rereg MR */ @@ -639,7 +639,6 @@ struct mlx5_ib_mw { struct mlx5_ib_devx_mr { struct mlx5_core_mkey mmkey; int ndescs; - struct rcu_head rcu; }; struct mlx5_ib_umr_context { diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index 1eff031ef048..7019c12005f4 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -84,32 +84,6 @@ static bool use_umr_mtt_update(struct mlx5_ib_mr *mr, u64 start, u64 length) length + (start & (MLX5_ADAPTER_PAGE_SIZE - 1)); } -static void update_odp_mr(struct mlx5_ib_mr *mr) -{ - if (is_odp_mr(mr)) { - /* - * This barrier prevents the compiler from moving the - * setting of umem->odp_data->private to point to our - * MR, before reg_umr finished, to ensure that the MR - * initialization have finished before starting to - * handle invalidations. - */ - smp_wmb(); - to_ib_umem_odp(mr->umem)->private = mr; - /* - * Make sure we will see the new - * umem->odp_data->private value in the invalidation - * routines, before we can get page faults on the - * MR. Page faults can happen once we put the MR in - * the tree, below this line. Without the barrier, - * there can be a fault handling and an invalidation - * before umem->odp_data->private == mr is visible to - * the invalidation handler. - */ - smp_wmb(); - } -} - static void reg_mr_callback(int status, struct mlx5_async_work *context) { struct mlx5_ib_mr *mr = @@ -1346,8 +1320,6 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, mr->umem = umem; set_mr_fields(dev, mr, npages, length, access_flags); - update_odp_mr(mr); - if (use_umr) { int update_xlt_flags = MLX5_IB_UPD_XLT_ENABLE; @@ -1363,10 +1335,12 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, } } - if (IS_ENABLED(CONFIG_INFINIBAND_ON_DEMAND_PAGING)) { - mr->live = 1; + if (is_odp_mr(mr)) { + to_ib_umem_odp(mr->umem)->private = mr; atomic_set(&mr->num_pending_prefetch, 0); } + if (IS_ENABLED(CONFIG_INFINIBAND_ON_DEMAND_PAGING)) + smp_store_release(&mr->live, 1); return &mr->ibmr; error: @@ -1441,6 +1415,9 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start, if (!mr->umem) return -EINVAL; + if (is_odp_mr(mr)) + return -EOPNOTSUPP; + if (flags & IB_MR_REREG_TRANS) { addr = virt_addr; len = length; @@ -1486,8 +1463,6 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start, } mr->allocated_from_cache = 0; - if (IS_ENABLED(CONFIG_INFINIBAND_ON_DEMAND_PAGING)) - mr->live = 1; } else { /* * Send a UMR WQE @@ -1516,7 +1491,6 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start, set_mr_fields(dev, mr, npages, len, access_flags); - update_odp_mr(mr); return 0; err: @@ -1607,15 +1581,16 @@ static void dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr) /* Prevent new page faults and * prefetch requests from succeeding */ - mr->live = 0; + WRITE_ONCE(mr->live, 0); + + /* Wait for all running page-fault handlers to finish. */ + synchronize_srcu(&dev->mr_srcu); /* dequeue pending prefetch requests for the mr */ if (atomic_read(&mr->num_pending_prefetch)) flush_workqueue(system_unbound_wq); WARN_ON(atomic_read(&mr->num_pending_prefetch)); - /* Wait for all running page-fault handlers to finish. */ - synchronize_srcu(&dev->mr_srcu); /* Destroy all page mappings */ if (!umem_odp->is_implicit_odp) mlx5_ib_invalidate_range(umem_odp, @@ -1987,14 +1962,25 @@ free: int mlx5_ib_dealloc_mw(struct ib_mw *mw) { + struct mlx5_ib_dev *dev = to_mdev(mw->device); struct mlx5_ib_mw *mmw = to_mmw(mw); int err; - err = mlx5_core_destroy_mkey((to_mdev(mw->device))->mdev, - &mmw->mmkey); - if (!err) - kfree(mmw); - return err; + if (IS_ENABLED(CONFIG_INFINIBAND_ON_DEMAND_PAGING)) { + xa_erase_irq(&dev->mdev->priv.mkey_table, + mlx5_base_mkey(mmw->mmkey.key)); + /* + * pagefault_single_data_segment() may be accessing mmw under + * SRCU if the user bound an ODP MR to this MW. + */ + synchronize_srcu(&dev->mr_srcu); + } + + err = mlx5_core_destroy_mkey(dev->mdev, &mmw->mmkey); + if (err) + return err; + kfree(mmw); + return 0; } int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask, diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c index 2e9b43061797..3f9478d19376 100644 --- a/drivers/infiniband/hw/mlx5/odp.c +++ b/drivers/infiniband/hw/mlx5/odp.c @@ -178,6 +178,29 @@ void mlx5_odp_populate_klm(struct mlx5_klm *pklm, size_t offset, return; } + /* + * The locking here is pretty subtle. Ideally the implicit children + * list would be protected by the umem_mutex, however that is not + * possible. Instead this uses a weaker update-then-lock pattern: + * + * srcu_read_lock() + * <change children list> + * mutex_lock(umem_mutex) + * mlx5_ib_update_xlt() + * mutex_unlock(umem_mutex) + * destroy lkey + * + * ie any change the children list must be followed by the locked + * update_xlt before destroying. + * + * The umem_mutex provides the acquire/release semantic needed to make + * the children list visible to a racing thread. While SRCU is not + * technically required, using it gives consistent use of the SRCU + * locking around the children list. + */ + lockdep_assert_held(&to_ib_umem_odp(mr->umem)->umem_mutex); + lockdep_assert_held(&mr->dev->mr_srcu); + odp = odp_lookup(offset * MLX5_IMR_MTT_SIZE, nentries * MLX5_IMR_MTT_SIZE, mr); @@ -202,15 +225,22 @@ static void mr_leaf_free_action(struct work_struct *work) struct ib_umem_odp *odp = container_of(work, struct ib_umem_odp, work); int idx = ib_umem_start(odp) >> MLX5_IMR_MTT_SHIFT; struct mlx5_ib_mr *mr = odp->private, *imr = mr->parent; + struct ib_umem_odp *odp_imr = to_ib_umem_odp(imr->umem); + int srcu_key; mr->parent = NULL; synchronize_srcu(&mr->dev->mr_srcu); - ib_umem_odp_release(odp); - if (imr->live) + if (smp_load_acquire(&imr->live)) { + srcu_key = srcu_read_lock(&mr->dev->mr_srcu); + mutex_lock(&odp_imr->umem_mutex); mlx5_ib_update_xlt(imr, idx, 1, 0, MLX5_IB_UPD_XLT_INDIRECT | MLX5_IB_UPD_XLT_ATOMIC); + mutex_unlock(&odp_imr->umem_mutex); + srcu_read_unlock(&mr->dev->mr_srcu, srcu_key); + } + ib_umem_odp_release(odp); mlx5_mr_cache_free(mr->dev, mr); if (atomic_dec_and_test(&imr->num_leaf_free)) @@ -278,7 +308,6 @@ void mlx5_ib_invalidate_range(struct ib_umem_odp *umem_odp, unsigned long start, idx - blk_start_idx + 1, 0, MLX5_IB_UPD_XLT_ZAP | MLX5_IB_UPD_XLT_ATOMIC); - mutex_unlock(&umem_odp->umem_mutex); /* * We are now sure that the device will not access the * memory. We can safely unmap it, and mark it as dirty if @@ -289,10 +318,12 @@ void mlx5_ib_invalidate_range(struct ib_umem_odp *umem_odp, unsigned long start, if (unlikely(!umem_odp->npages && mr->parent && !umem_odp->dying)) { - WRITE_ONCE(umem_odp->dying, 1); + WRITE_ONCE(mr->live, 0); + umem_odp->dying = 1; atomic_inc(&mr->parent->num_leaf_free); schedule_work(&umem_odp->work); } + mutex_unlock(&umem_odp->umem_mutex); } void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev) @@ -429,8 +460,6 @@ static struct mlx5_ib_mr *implicit_mr_alloc(struct ib_pd *pd, mr->ibmr.lkey = mr->mmkey.key; mr->ibmr.rkey = mr->mmkey.key; - mr->live = 1; - mlx5_ib_dbg(dev, "key %x dev %p mr %p\n", mr->mmkey.key, dev->mdev, mr); @@ -484,6 +513,8 @@ next_mr: mtt->parent = mr; INIT_WORK(&odp->work, mr_leaf_free_action); + smp_store_release(&mtt->live, 1); + if (!nentries) start_idx = addr >> MLX5_IMR_MTT_SHIFT; nentries++; @@ -536,6 +567,7 @@ struct mlx5_ib_mr *mlx5_ib_alloc_implicit_mr(struct mlx5_ib_pd *pd, init_waitqueue_head(&imr->q_leaf_free); atomic_set(&imr->num_leaf_free, 0); atomic_set(&imr->num_pending_prefetch, 0); + smp_store_release(&imr->live, 1); return imr; } @@ -555,15 +587,19 @@ void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr) if (mr->parent != imr) continue; + mutex_lock(&umem_odp->umem_mutex); ib_umem_odp_unmap_dma_pages(umem_odp, ib_umem_start(umem_odp), ib_umem_end(umem_odp)); - if (umem_odp->dying) + if (umem_odp->dying) { + mutex_unlock(&umem_odp->umem_mutex); continue; + } - WRITE_ONCE(umem_odp->dying, 1); + umem_odp->dying = 1; atomic_inc(&imr->num_leaf_free); schedule_work(&umem_odp->work); + mutex_unlock(&umem_odp->umem_mutex); } up_read(&per_mm->umem_rwsem); @@ -773,7 +809,7 @@ next_mr: switch (mmkey->type) { case MLX5_MKEY_MR: mr = container_of(mmkey, struct mlx5_ib_mr, mmkey); - if (!mr->live || !mr->ibmr.pd) { + if (!smp_load_acquire(&mr->live) || !mr->ibmr.pd) { mlx5_ib_dbg(dev, "got dead MR\n"); ret = -EFAULT; goto srcu_unlock; @@ -1641,12 +1677,12 @@ static bool num_pending_prefetch_inc(struct ib_pd *pd, mr = container_of(mmkey, struct mlx5_ib_mr, mmkey); - if (mr->ibmr.pd != pd) { + if (!smp_load_acquire(&mr->live)) { ret = false; break; } - if (!mr->live) { + if (mr->ibmr.pd != pd) { ret = false; break; } diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c index 8937d72ddcf6..5fd071c05944 100644 --- a/drivers/infiniband/hw/mlx5/qp.c +++ b/drivers/infiniband/hw/mlx5/qp.c @@ -3249,10 +3249,12 @@ static int modify_raw_packet_qp_sq( } /* Only remove the old rate after new rate was set */ - if ((old_rl.rate && - !mlx5_rl_are_equal(&old_rl, &new_rl)) || - (new_state != MLX5_SQC_STATE_RDY)) + if ((old_rl.rate && !mlx5_rl_are_equal(&old_rl, &new_rl)) || + (new_state != MLX5_SQC_STATE_RDY)) { mlx5_rl_remove_rate(dev, &old_rl); + if (new_state != MLX5_SQC_STATE_RDY) + memset(&new_rl, 0, sizeof(new_rl)); + } ibqp->rl = new_rl; sq->state = new_state; diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c index 5136b835e1ba..dc71b6e16a07 100644 --- a/drivers/infiniband/hw/qedr/main.c +++ b/drivers/infiniband/hw/qedr/main.c @@ -76,7 +76,7 @@ static void qedr_get_dev_fw_str(struct ib_device *ibdev, char *str) struct qedr_dev *qedr = get_qedr_dev(ibdev); u32 fw_ver = (u32)qedr->attr.fw_ver; - snprintf(str, IB_FW_VERSION_NAME_MAX, "%d. %d. %d. %d", + snprintf(str, IB_FW_VERSION_NAME_MAX, "%d.%d.%d.%d", (fw_ver >> 24) & 0xFF, (fw_ver >> 16) & 0xFF, (fw_ver >> 8) & 0xFF, fw_ver & 0xFF); } diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c index 6cac0c88cf39..36cdfbdbd325 100644 --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_srq.c @@ -230,8 +230,6 @@ static void pvrdma_free_srq(struct pvrdma_dev *dev, struct pvrdma_srq *srq) pvrdma_page_dir_cleanup(dev, &srq->pdir); - kfree(srq); - atomic_dec(&dev->num_srqs); } |