diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2015-02-21 20:53:21 (GMT) |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2015-02-21 20:53:21 (GMT) |
commit | b5ccb078c806f4804aaf85bb59faa9b6fedf85a7 (patch) | |
tree | 192402fa8c3951037c4079fceff5cb5d5455bf5f /drivers/infiniband | |
parent | 295324556c427d60b41668ab81a43f604533f456 (diff) | |
parent | 147d1da951cf067badb7d3585002e1eb46719f83 (diff) | |
download | linux-b5ccb078c806f4804aaf85bb59faa9b6fedf85a7.tar.xz |
Merge tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband
Pull InfiniBand/RDMA updates from Roland Dreier:
- Re-enable on-demand paging changes with stable ABI
- Fairly large set of ocrdma HW driver fixes
- Some qib HW driver fixes
- Other miscellaneous changes
* tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (43 commits)
IB/qib: Add blank line after declaration
IB/qib: Fix checkpatch warnings
IB/mlx5: Enable the ODP capability query verb
IB/core: Add on demand paging caps to ib_uverbs_ex_query_device
IB/core: Add support for extended query device caps
RDMA/cxgb4: Don't hang threads forever waiting on WR replies
RDMA/ocrdma: Fix off by one in ocrdma_query_gid()
RDMA/ocrdma: Use unsigned for bit index
RDMA/ocrdma: Help gcc generate better code for ocrdma_srq_toggle_bit
RDMA/ocrdma: Update the ocrdma module version string
RDMA/ocrdma: set vlan present bit for user AH
RDMA/ocrdma: remove reference of ocrdma_dev out of ocrdma_qp structure
RDMA/ocrdma: Add support for interrupt moderation
RDMA/ocrdma: Honor return value of ocrdma_resolve_dmac
RDMA/ocrdma: Allow expansion of the SQ CQEs via buddy CQ expansion of the QP
RDMA/ocrdma: Discontinue support of RDMA-READ-WITH-INVALIDATE
RDMA/ocrdma: Host crash on destroying device resources
RDMA/ocrdma: Report correct state in ibv_query_qp
RDMA/ocrdma: Debugfs enhancments for ocrdma driver
RDMA/ocrdma: Report correct count of interrupt vectors while registering ocrdma device
...
Diffstat (limited to 'drivers/infiniband')
61 files changed, 1188 insertions, 560 deletions
diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c index 56a4b7c..45d67e9 100644 --- a/drivers/infiniband/core/ucma.c +++ b/drivers/infiniband/core/ucma.c @@ -1124,6 +1124,9 @@ static int ucma_set_ib_path(struct ucma_context *ctx, if (!optlen) return -EINVAL; + memset(&sa_path, 0, sizeof(sa_path)); + sa_path.vlan_id = 0xffff; + ib_sa_unpack_path(path_data->path_rec, &sa_path); ret = rdma_set_ib_paths(ctx->cm_id, &sa_path, 1); if (ret) diff --git a/drivers/infiniband/core/umem_odp.c b/drivers/infiniband/core/umem_odp.c index 6095872..8b8cc6f 100644 --- a/drivers/infiniband/core/umem_odp.c +++ b/drivers/infiniband/core/umem_odp.c @@ -294,7 +294,8 @@ int ib_umem_odp_get(struct ib_ucontext *context, struct ib_umem *umem) if (likely(ib_umem_start(umem) != ib_umem_end(umem))) rbt_ib_umem_insert(&umem->odp_data->interval_tree, &context->umem_tree); - if (likely(!atomic_read(&context->notifier_count))) + if (likely(!atomic_read(&context->notifier_count)) || + context->odp_mrs_count == 1) umem->odp_data->mn_counters_active = true; else list_add(&umem->odp_data->no_private_counters, diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h index 643c08a..b716b08 100644 --- a/drivers/infiniband/core/uverbs.h +++ b/drivers/infiniband/core/uverbs.h @@ -258,5 +258,6 @@ IB_UVERBS_DECLARE_CMD(close_xrcd); IB_UVERBS_DECLARE_EX_CMD(create_flow); IB_UVERBS_DECLARE_EX_CMD(destroy_flow); +IB_UVERBS_DECLARE_EX_CMD(query_device); #endif /* UVERBS_H */ diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index b7943ff..a9f0489 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -400,6 +400,52 @@ err: return ret; } +static void copy_query_dev_fields(struct ib_uverbs_file *file, + struct ib_uverbs_query_device_resp *resp, + struct ib_device_attr *attr) +{ + resp->fw_ver = attr->fw_ver; + resp->node_guid = file->device->ib_dev->node_guid; + resp->sys_image_guid = attr->sys_image_guid; + resp->max_mr_size = attr->max_mr_size; + resp->page_size_cap = attr->page_size_cap; + resp->vendor_id = attr->vendor_id; + resp->vendor_part_id = attr->vendor_part_id; + resp->hw_ver = attr->hw_ver; + resp->max_qp = attr->max_qp; + resp->max_qp_wr = attr->max_qp_wr; + resp->device_cap_flags = attr->device_cap_flags; + resp->max_sge = attr->max_sge; + resp->max_sge_rd = attr->max_sge_rd; + resp->max_cq = attr->max_cq; + resp->max_cqe = attr->max_cqe; + resp->max_mr = attr->max_mr; + resp->max_pd = attr->max_pd; + resp->max_qp_rd_atom = attr->max_qp_rd_atom; + resp->max_ee_rd_atom = attr->max_ee_rd_atom; + resp->max_res_rd_atom = attr->max_res_rd_atom; + resp->max_qp_init_rd_atom = attr->max_qp_init_rd_atom; + resp->max_ee_init_rd_atom = attr->max_ee_init_rd_atom; + resp->atomic_cap = attr->atomic_cap; + resp->max_ee = attr->max_ee; + resp->max_rdd = attr->max_rdd; + resp->max_mw = attr->max_mw; + resp->max_raw_ipv6_qp = attr->max_raw_ipv6_qp; + resp->max_raw_ethy_qp = attr->max_raw_ethy_qp; + resp->max_mcast_grp = attr->max_mcast_grp; + resp->max_mcast_qp_attach = attr->max_mcast_qp_attach; + resp->max_total_mcast_qp_attach = attr->max_total_mcast_qp_attach; + resp->max_ah = attr->max_ah; + resp->max_fmr = attr->max_fmr; + resp->max_map_per_fmr = attr->max_map_per_fmr; + resp->max_srq = attr->max_srq; + resp->max_srq_wr = attr->max_srq_wr; + resp->max_srq_sge = attr->max_srq_sge; + resp->max_pkeys = attr->max_pkeys; + resp->local_ca_ack_delay = attr->local_ca_ack_delay; + resp->phys_port_cnt = file->device->ib_dev->phys_port_cnt; +} + ssize_t ib_uverbs_query_device(struct ib_uverbs_file *file, const char __user *buf, int in_len, int out_len) @@ -420,47 +466,7 @@ ssize_t ib_uverbs_query_device(struct ib_uverbs_file *file, return ret; memset(&resp, 0, sizeof resp); - - resp.fw_ver = attr.fw_ver; - resp.node_guid = file->device->ib_dev->node_guid; - resp.sys_image_guid = attr.sys_image_guid; - resp.max_mr_size = attr.max_mr_size; - resp.page_size_cap = attr.page_size_cap; - resp.vendor_id = attr.vendor_id; - resp.vendor_part_id = attr.vendor_part_id; - resp.hw_ver = attr.hw_ver; - resp.max_qp = attr.max_qp; - resp.max_qp_wr = attr.max_qp_wr; - resp.device_cap_flags = attr.device_cap_flags; - resp.max_sge = attr.max_sge; - resp.max_sge_rd = attr.max_sge_rd; - resp.max_cq = attr.max_cq; - resp.max_cqe = attr.max_cqe; - resp.max_mr = attr.max_mr; - resp.max_pd = attr.max_pd; - resp.max_qp_rd_atom = attr.max_qp_rd_atom; - resp.max_ee_rd_atom = attr.max_ee_rd_atom; - resp.max_res_rd_atom = attr.max_res_rd_atom; - resp.max_qp_init_rd_atom = attr.max_qp_init_rd_atom; - resp.max_ee_init_rd_atom = attr.max_ee_init_rd_atom; - resp.atomic_cap = attr.atomic_cap; - resp.max_ee = attr.max_ee; - resp.max_rdd = attr.max_rdd; - resp.max_mw = attr.max_mw; - resp.max_raw_ipv6_qp = attr.max_raw_ipv6_qp; - resp.max_raw_ethy_qp = attr.max_raw_ethy_qp; - resp.max_mcast_grp = attr.max_mcast_grp; - resp.max_mcast_qp_attach = attr.max_mcast_qp_attach; - resp.max_total_mcast_qp_attach = attr.max_total_mcast_qp_attach; - resp.max_ah = attr.max_ah; - resp.max_fmr = attr.max_fmr; - resp.max_map_per_fmr = attr.max_map_per_fmr; - resp.max_srq = attr.max_srq; - resp.max_srq_wr = attr.max_srq_wr; - resp.max_srq_sge = attr.max_srq_sge; - resp.max_pkeys = attr.max_pkeys; - resp.local_ca_ack_delay = attr.local_ca_ack_delay; - resp.phys_port_cnt = file->device->ib_dev->phys_port_cnt; + copy_query_dev_fields(file, &resp, &attr); if (copy_to_user((void __user *) (unsigned long) cmd.response, &resp, sizeof resp)) @@ -2091,20 +2097,21 @@ ssize_t ib_uverbs_modify_qp(struct ib_uverbs_file *file, if (qp->real_qp == qp) { ret = ib_resolve_eth_l2_attrs(qp, attr, &cmd.attr_mask); if (ret) - goto out; + goto release_qp; ret = qp->device->modify_qp(qp, attr, modify_qp_mask(qp->qp_type, cmd.attr_mask), &udata); } else { ret = ib_modify_qp(qp, attr, modify_qp_mask(qp->qp_type, cmd.attr_mask)); } - put_qp_read(qp); - if (ret) - goto out; + goto release_qp; ret = in_len; +release_qp: + put_qp_read(qp); + out: kfree(attr); @@ -3287,3 +3294,64 @@ ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file, return ret ? ret : in_len; } + +int ib_uverbs_ex_query_device(struct ib_uverbs_file *file, + struct ib_udata *ucore, + struct ib_udata *uhw) +{ + struct ib_uverbs_ex_query_device_resp resp; + struct ib_uverbs_ex_query_device cmd; + struct ib_device_attr attr; + struct ib_device *device; + int err; + + device = file->device->ib_dev; + if (ucore->inlen < sizeof(cmd)) + return -EINVAL; + + err = ib_copy_from_udata(&cmd, ucore, sizeof(cmd)); + if (err) + return err; + + if (cmd.comp_mask) + return -EINVAL; + + if (cmd.reserved) + return -EINVAL; + + resp.response_length = offsetof(typeof(resp), odp_caps); + + if (ucore->outlen < resp.response_length) + return -ENOSPC; + + err = device->query_device(device, &attr); + if (err) + return err; + + copy_query_dev_fields(file, &resp.base, &attr); + resp.comp_mask = 0; + + if (ucore->outlen < resp.response_length + sizeof(resp.odp_caps)) + goto end; + +#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING + resp.odp_caps.general_caps = attr.odp_caps.general_caps; + resp.odp_caps.per_transport_caps.rc_odp_caps = + attr.odp_caps.per_transport_caps.rc_odp_caps; + resp.odp_caps.per_transport_caps.uc_odp_caps = + attr.odp_caps.per_transport_caps.uc_odp_caps; + resp.odp_caps.per_transport_caps.ud_odp_caps = + attr.odp_caps.per_transport_caps.ud_odp_caps; + resp.odp_caps.reserved = 0; +#else + memset(&resp.odp_caps, 0, sizeof(resp.odp_caps)); +#endif + resp.response_length += sizeof(resp.odp_caps); + +end: + err = ib_copy_to_udata(ucore, &resp, resp.response_length); + if (err) + return err; + + return 0; +} diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 5db1a8c..259dcc7 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -123,6 +123,7 @@ static int (*uverbs_ex_cmd_table[])(struct ib_uverbs_file *file, struct ib_udata *uhw) = { [IB_USER_VERBS_EX_CMD_CREATE_FLOW] = ib_uverbs_ex_create_flow, [IB_USER_VERBS_EX_CMD_DESTROY_FLOW] = ib_uverbs_ex_destroy_flow, + [IB_USER_VERBS_EX_CMD_QUERY_DEVICE] = ib_uverbs_ex_query_device, }; static void ib_uverbs_add_one(struct ib_device *device); diff --git a/drivers/infiniband/hw/cxgb4/ev.c b/drivers/infiniband/hw/cxgb4/ev.c index 794555d..bdfac2c 100644 --- a/drivers/infiniband/hw/cxgb4/ev.c +++ b/drivers/infiniband/hw/cxgb4/ev.c @@ -225,13 +225,20 @@ int c4iw_ev_handler(struct c4iw_dev *dev, u32 qid) struct c4iw_cq *chp; unsigned long flag; + spin_lock_irqsave(&dev->lock, flag); chp = get_chp(dev, qid); if (chp) { + atomic_inc(&chp->refcnt); + spin_unlock_irqrestore(&dev->lock, flag); t4_clear_cq_armed(&chp->cq); spin_lock_irqsave(&chp->comp_handler_lock, flag); (*chp->ibcq.comp_handler)(&chp->ibcq, chp->ibcq.cq_context); spin_unlock_irqrestore(&chp->comp_handler_lock, flag); - } else + if (atomic_dec_and_test(&chp->refcnt)) + wake_up(&chp->wait); + } else { PDBG("%s unknown cqid 0x%x\n", __func__, qid); + spin_unlock_irqrestore(&dev->lock, flag); + } return 0; } diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h index b5678ac..d87e165 100644 --- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h +++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h @@ -196,7 +196,7 @@ static inline int c4iw_num_stags(struct c4iw_rdev *rdev) return (int)(rdev->lldi.vr->stag.size >> 5); } -#define C4IW_WR_TO (30*HZ) +#define C4IW_WR_TO (60*HZ) struct c4iw_wr_wait { struct completion completion; @@ -220,22 +220,21 @@ static inline int c4iw_wait_for_reply(struct c4iw_rdev *rdev, u32 hwtid, u32 qpid, const char *func) { - unsigned to = C4IW_WR_TO; int ret; - do { - ret = wait_for_completion_timeout(&wr_waitp->completion, to); - if (!ret) { - printk(KERN_ERR MOD "%s - Device %s not responding - " - "tid %u qpid %u\n", func, - pci_name(rdev->lldi.pdev), hwtid, qpid); - if (c4iw_fatal_error(rdev)) { - wr_waitp->ret = -EIO; - break; - } - to = to << 2; - } - } while (!ret); + if (c4iw_fatal_error(rdev)) { + wr_waitp->ret = -EIO; + goto out; + } + + ret = wait_for_completion_timeout(&wr_waitp->completion, C4IW_WR_TO); + if (!ret) { + PDBG("%s - Device %s not responding (disabling device) - tid %u qpid %u\n", + func, pci_name(rdev->lldi.pdev), hwtid, qpid); + rdev->flags |= T4_FATAL_ERROR; + wr_waitp->ret = -EIO; + } +out: if (wr_waitp->ret) PDBG("%s: FW reply %d tid %u qpid %u\n", pci_name(rdev->lldi.pdev), wr_waitp->ret, hwtid, qpid); diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h index 6559af6..e08db70 100644 --- a/drivers/infiniband/hw/ipath/ipath_kernel.h +++ b/drivers/infiniband/hw/ipath/ipath_kernel.h @@ -908,9 +908,6 @@ void ipath_chip_cleanup(struct ipath_devdata *); /* clean up any chip type-specific stuff */ void ipath_chip_done(void); -/* check to see if we have to force ordering for write combining */ -int ipath_unordered_wc(void); - void ipath_disarm_piobufs(struct ipath_devdata *, unsigned first, unsigned cnt); void ipath_cancel_sends(struct ipath_devdata *, int); diff --git a/drivers/infiniband/hw/ipath/ipath_wc_ppc64.c b/drivers/infiniband/hw/ipath/ipath_wc_ppc64.c index 1d7bd82..1a7e20a 100644 --- a/drivers/infiniband/hw/ipath/ipath_wc_ppc64.c +++ b/drivers/infiniband/hw/ipath/ipath_wc_ppc64.c @@ -47,16 +47,3 @@ int ipath_enable_wc(struct ipath_devdata *dd) { return 0; } - -/** - * ipath_unordered_wc - indicate whether write combining is unordered - * - * Because our performance depends on our ability to do write - * combining mmio writes in the most efficient way, we need to - * know if we are on a processor that may reorder stores when - * write combining. - */ -int ipath_unordered_wc(void) -{ - return 1; -} diff --git a/drivers/infiniband/hw/ipath/ipath_wc_x86_64.c b/drivers/infiniband/hw/ipath/ipath_wc_x86_64.c index 3428acb..4ad0b93 100644 --- a/drivers/infiniband/hw/ipath/ipath_wc_x86_64.c +++ b/drivers/infiniband/hw/ipath/ipath_wc_x86_64.c @@ -167,18 +167,3 @@ void ipath_disable_wc(struct ipath_devdata *dd) dd->ipath_wc_cookie = 0; /* even on failure */ } } - -/** - * ipath_unordered_wc - indicate whether write combining is ordered - * - * Because our performance depends on our ability to do write combining mmio - * writes in the most efficient way, we need to know if we are on an Intel - * or AMD x86_64 processor. AMD x86_64 processors flush WC buffers out in - * the order completed, and so no special flushing is required to get - * correct ordering. Intel processors, however, will flush write buffers - * out in "random" orders, and so explicit ordering is needed at times. - */ -int ipath_unordered_wc(void) -{ - return boot_cpu_data.x86_vendor != X86_VENDOR_AMD; -} diff --git a/drivers/infiniband/hw/mlx4/cm.c b/drivers/infiniband/hw/mlx4/cm.c index 56a593e..39a4888 100644 --- a/drivers/infiniband/hw/mlx4/cm.c +++ b/drivers/infiniband/hw/mlx4/cm.c @@ -372,7 +372,7 @@ int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave, *slave = mlx4_ib_find_real_gid(ibdev, port, gid.global.interface_id); if (*slave < 0) { mlx4_ib_warn(ibdev, "failed matching slave_id by gid (0x%llx)\n", - gid.global.interface_id); + be64_to_cpu(gid.global.interface_id)); return -ENOENT; } return 0; diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c index 543ecdd..0176caa 100644 --- a/drivers/infiniband/hw/mlx4/cq.c +++ b/drivers/infiniband/hw/mlx4/cq.c @@ -369,8 +369,7 @@ int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) int err; mutex_lock(&cq->resize_mutex); - - if (entries < 1) { + if (entries < 1 || entries > dev->dev->caps.max_cqes) { err = -EINVAL; goto out; } @@ -381,7 +380,7 @@ int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) goto out; } - if (entries > dev->dev->caps.max_cqes) { + if (entries > dev->dev->caps.max_cqes + 1) { err = -EINVAL; goto out; } @@ -394,7 +393,7 @@ int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata) /* Can't be smaller than the number of outstanding CQEs */ outst_cqe = mlx4_ib_get_outstanding_cqes(cq); if (entries < outst_cqe + 1) { - err = 0; + err = -EINVAL; goto out; } diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index eb8e215..ac6e2b7 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -1269,8 +1269,7 @@ static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) struct mlx4_dev *dev = mdev->dev; struct mlx4_ib_qp *mqp = to_mqp(ibqp); struct mlx4_ib_steering *ib_steering = NULL; - enum mlx4_protocol prot = (gid->raw[1] == 0x0e) ? - MLX4_PROT_IB_IPV4 : MLX4_PROT_IB_IPV6; + enum mlx4_protocol prot = MLX4_PROT_IB_IPV6; struct mlx4_flow_reg_id reg_id; if (mdev->dev->caps.steering_mode == @@ -1284,8 +1283,10 @@ static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) !!(mqp->flags & MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK), prot, ®_id.id); - if (err) + if (err) { + pr_err("multicast attach op failed, err %d\n", err); goto err_malloc; + } reg_id.mirror = 0; if (mlx4_is_bonded(dev)) { @@ -1348,9 +1349,7 @@ static int mlx4_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) struct net_device *ndev; struct mlx4_ib_gid_entry *ge; struct mlx4_flow_reg_id reg_id = {0, 0}; - - enum mlx4_protocol prot = (gid->raw[1] == 0x0e) ? - MLX4_PROT_IB_IPV4 : MLX4_PROT_IB_IPV6; + enum mlx4_protocol prot = MLX4_PROT_IB_IPV6; if (mdev->dev->caps.steering_mode == MLX4_STEERING_MODE_DEVICE_MANAGED) { diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index dfc6ca1..ed2bd67 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -1696,8 +1696,10 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_GSI || qp->mlx4_ib_qp_type == MLX4_IB_QPT_TUN_GSI) { err = handle_eth_ud_smac_index(dev, qp, (u8 *)attr->smac, context); - if (err) - return -EINVAL; + if (err) { + err = -EINVAL; + goto out; + } if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_GSI) dev->qp1_proxy[qp->port - 1] = qp; } diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 03bf812..cc4ac1e 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -997,7 +997,7 @@ static int get_port_caps(struct mlx5_ib_dev *dev) struct ib_device_attr *dprops = NULL; struct ib_port_attr *pprops = NULL; struct mlx5_general_caps *gen; - int err = 0; + int err = -ENOMEM; int port; gen = &dev->mdev->caps.gen; @@ -1331,6 +1331,8 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev) (1ull << IB_USER_VERBS_CMD_DESTROY_SRQ) | (1ull << IB_USER_VERBS_CMD_CREATE_XSRQ) | (1ull << IB_USER_VERBS_CMD_OPEN_QP); + dev->ib_dev.uverbs_ex_cmd_mask = + (1ull << IB_USER_VERBS_EX_CMD_QUERY_DEVICE); dev->ib_dev.query_device = mlx5_ib_query_device; dev->ib_dev.query_port = mlx5_ib_query_port; diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c index 32a28bd..cd9822e 100644 --- a/drivers/infiniband/hw/mlx5/mr.c +++ b/drivers/infiniband/hw/mlx5/mr.c @@ -1012,6 +1012,7 @@ static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, u64 virt_addr, goto err_2; } mr->umem = umem; + mr->dev = dev; mr->live = 1; kvfree(in); diff --git a/drivers/infiniband/hw/ocrdma/ocrdma.h b/drivers/infiniband/hw/ocrdma/ocrdma.h index b43456a..c9780d9 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma.h @@ -40,7 +40,7 @@ #include <be_roce.h> #include "ocrdma_sli.h" -#define OCRDMA_ROCE_DRV_VERSION "10.2.287.0u" +#define OCRDMA_ROCE_DRV_VERSION "10.4.205.0u" #define OCRDMA_ROCE_DRV_DESC "Emulex OneConnect RoCE Driver" #define OCRDMA_NODE_DESC "Emulex OneConnect RoCE HCA" @@ -55,12 +55,19 @@ #define OCRDMA_UVERBS(CMD_NAME) (1ull << IB_USER_VERBS_CMD_##CMD_NAME) #define convert_to_64bit(lo, hi) ((u64)hi << 32 | (u64)lo) +#define EQ_INTR_PER_SEC_THRSH_HI 150000 +#define EQ_INTR_PER_SEC_THRSH_LOW 100000 +#define EQ_AIC_MAX_EQD 20 +#define EQ_AIC_MIN_EQD 0 + +void ocrdma_eqd_set_task(struct work_struct *work); struct ocrdma_dev_attr { u8 fw_ver[32]; u32 vendor_id; u32 device_id; u16 max_pd; + u16 max_dpp_pds; u16 max_cq; u16 max_cqe; u16 max_qp; @@ -116,12 +123,19 @@ struct ocrdma_queue_info { bool created; }; +struct ocrdma_aic_obj { /* Adaptive interrupt coalescing (AIC) info */ + u32 prev_eqd; + u64 eq_intr_cnt; + u64 prev_eq_intr_cnt; +}; + struct ocrdma_eq { struct ocrdma_queue_info q; u32 vector; int cq_cnt; struct ocrdma_dev *dev; char irq_name[32]; + struct ocrdma_aic_obj aic_obj; }; struct ocrdma_mq { @@ -171,6 +185,21 @@ struct ocrdma_stats { struct ocrdma_dev *dev; }; +struct ocrdma_pd_resource_mgr { + u32 pd_norm_start; + u16 pd_norm_count; + u16 pd_norm_thrsh; + u16 max_normal_pd; + u32 pd_dpp_start; + u16 pd_dpp_count; + u16 pd_dpp_thrsh; + u16 max_dpp_pd; + u16 dpp_page_index; + unsigned long *pd_norm_bitmap; + unsigned long *pd_dpp_bitmap; + bool pd_prealloc_valid; +}; + struct stats_mem { struct ocrdma_mqe mqe; void *va; @@ -198,6 +227,7 @@ struct ocrdma_dev { struct ocrdma_eq *eq_tbl; int eq_cnt; + struct delayed_work eqd_work; u16 base_eqid; u16 max_eq; @@ -255,7 +285,12 @@ struct ocrdma_dev { struct ocrdma_stats rx_qp_err_stats; struct ocrdma_stats tx_dbg_stats; struct ocrdma_stats rx_dbg_stats; + struct ocrdma_stats driver_stats; + struct ocrdma_stats reset_stats; struct dentry *dir; + atomic_t async_err_stats[OCRDMA_MAX_ASYNC_ERRORS]; + atomic_t cqe_err_stats[OCRDMA_MAX_CQE_ERR]; + struct ocrdma_pd_resource_mgr *pd_mgr; }; struct ocrdma_cq { @@ -335,7 +370,6 @@ struct ocrdma_srq { struct ocrdma_qp { struct ib_qp ibqp; - struct ocrdma_dev *dev; u8 __iomem *sq_db; struct ocrdma_qp_hwq_info sq; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c index f3cc8c9..d812904 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c @@ -29,19 +29,22 @@ #include <net/netevent.h> #include <rdma/ib_addr.h> +#include <rdma/ib_mad.h> #include "ocrdma.h" #include "ocrdma_verbs.h" #include "ocrdma_ah.h" #include "ocrdma_hw.h" +#include "ocrdma_stats.h" #define OCRDMA_VID_PCP_SHIFT 0xD static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, - struct ib_ah_attr *attr, union ib_gid *sgid, int pdid) + struct ib_ah_attr *attr, union ib_gid *sgid, + int pdid, bool *isvlan) { int status = 0; - u16 vlan_tag; bool vlan_enabled = false; + u16 vlan_tag; struct ocrdma_eth_vlan eth; struct ocrdma_grh grh; int eth_sz; @@ -59,7 +62,7 @@ static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, vlan_tag |= (dev->sl & 0x07) << OCRDMA_VID_PCP_SHIFT; eth.vlan_tag = cpu_to_be16(vlan_tag); eth_sz = sizeof(struct ocrdma_eth_vlan); - vlan_enabled = true; + *isvlan = true; } else { eth.eth_type = cpu_to_be16(OCRDMA_ROCE_ETH_TYPE); eth_sz = sizeof(struct ocrdma_eth_basic); @@ -82,7 +85,7 @@ static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, /* Eth HDR */ memcpy(&ah->av->eth_hdr, ð, eth_sz); memcpy((u8 *)ah->av + eth_sz, &grh, sizeof(struct ocrdma_grh)); - if (vlan_enabled) + if (*isvlan) ah->av->valid |= OCRDMA_AV_VLAN_VALID; ah->av->valid = cpu_to_le32(ah->av->valid); return status; @@ -91,6 +94,7 @@ static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr) { u32 *ahid_addr; + bool isvlan = false; int status; struct ocrdma_ah *ah; struct ocrdma_pd *pd = get_ocrdma_pd(ibpd); @@ -127,15 +131,20 @@ struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr) } } - status = set_av_attr(dev, ah, attr, &sgid, pd->id); + status = set_av_attr(dev, ah, attr, &sgid, pd->id, &isvlan); if (status) goto av_conf_err; /* if pd is for the user process, pass the ah_id to user space */ if ((pd->uctx) && (pd->uctx->ah_tbl.va)) { ahid_addr = pd->uctx->ah_tbl.va + attr->dlid; - *ahid_addr = ah->id; + *ahid_addr = 0; + *ahid_addr |= ah->id & OCRDMA_AH_ID_MASK; + if (isvlan) + *ahid_addr |= (OCRDMA_AH_VLAN_VALID_MASK << + OCRDMA_AH_VLAN_VALID_SHIFT); } + return &ah->ibah; av_conf_err: @@ -191,5 +200,20 @@ int ocrdma_process_mad(struct ib_device *ibdev, struct ib_grh *in_grh, struct ib_mad *in_mad, struct ib_mad *out_mad) { - return IB_MAD_RESULT_SUCCESS; + int status; + struct ocrdma_dev *dev; + + switch (in_mad->mad_hdr.mgmt_class) { + case IB_MGMT_CLASS_PERF_MGMT: + dev = get_ocrdma_dev(ibdev); + if (!ocrdma_pma_counters(dev, out_mad)) + status = IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY; + else + status = IB_MAD_RESULT_SUCCESS; + break; + default: + status = IB_MAD_RESULT_SUCCESS; + break; + } + return status; } diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h index 8ac49e7..726a87c 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h @@ -28,6 +28,12 @@ #ifndef __OCRDMA_AH_H__ #define __OCRDMA_AH_H__ +enum { + OCRDMA_AH_ID_MASK = 0x3FF, + OCRDMA_AH_VLAN_VALID_MASK = 0x01, + OCRDMA_AH_VLAN_VALID_SHIFT = 0x1F +}; + struct ib_ah *ocrdma_create_ah(struct ib_pd *, struct ib_ah_attr *); int ocrdma_destroy_ah(struct ib_ah *); int ocrdma_query_ah(struct ib_ah *, struct ib_ah_attr *); diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index 638bff1..0c9e959 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -734,6 +734,9 @@ static void ocrdma_dispatch_ibevent(struct ocrdma_dev *dev, break; } + if (type < OCRDMA_MAX_ASYNC_ERRORS) + atomic_inc(&dev->async_err_stats[type]); + if (qp_event) { if (qp->ibqp.event_handler) qp->ibqp.event_handler(&ib_evt, qp->ibqp.qp_context); @@ -831,20 +834,20 @@ static int ocrdma_mq_cq_handler(struct ocrdma_dev *dev, u16 cq_id) return 0; } -static void ocrdma_qp_buddy_cq_handler(struct ocrdma_dev *dev, - struct ocrdma_cq *cq) +static struct ocrdma_cq *_ocrdma_qp_buddy_cq_handler(struct ocrdma_dev *dev, + struct ocrdma_cq *cq, bool sq) { - unsigned long flags; struct ocrdma_qp *qp; - bool buddy_cq_found = false; - /* Go through list of QPs in error state which are using this CQ - * and invoke its callback handler to trigger CQE processing for - * error/flushed CQE. It is rare to find more than few entries in - * this list as most consumers stops after getting error CQE. - * List is traversed only once when a matching buddy cq found for a QP. - */ - spin_lock_irqsave(&dev->flush_q_lock, flags); - list_for_each_entry(qp, &cq->sq_head, sq_entry) { + struct list_head *cur; + struct ocrdma_cq *bcq = NULL; + struct list_head *head = sq?(&cq->sq_head):(&cq->rq_head); + + list_for_each(cur, head) { + if (sq) + qp = list_entry(cur, struct ocrdma_qp, sq_entry); + else + qp = list_entry(cur, struct ocrdma_qp, rq_entry); + if (qp->srq) continue; /* if wq and rq share the same cq, than comp_handler @@ -856,19 +859,41 @@ static void ocrdma_qp_buddy_cq_handler(struct ocrdma_dev *dev, * if completion came on rq, sq's cq is buddy cq. */ if (qp->sq_cq == cq) - cq = qp->rq_cq; + bcq = qp->rq_cq; else - cq = qp->sq_cq; - buddy_cq_found = true; - break; + bcq = qp->sq_cq; + return bcq; } + return NULL; +} + +static void ocrdma_qp_buddy_cq_handler(struct ocrdma_dev *dev, + struct ocrdma_cq *cq) +{ + unsigned long flags; + struct ocrdma_cq *bcq = NULL; + + /* Go through list of QPs in error state which are using this CQ + * and invoke its callback handler to trigger CQE processing for + * error/flushed CQE. It is rare to find more than few entries in + * this list as most consumers stops after getting error CQE. + * List is traversed only once when a matching buddy cq found for a QP. + */ + spin_lock_irqsave(&dev->flush_q_lock, flags); + /* Check if buddy CQ is present. + * true - Check for SQ CQ + * false - Check for RQ CQ + */ + bcq = _ocrdma_qp_buddy_cq_handler(dev, cq, true); + if (bcq == NULL) + bcq = _ocrdma_qp_buddy_cq_handler(dev, cq, false); spin_unlock_irqrestore(&dev->flush_q_lock, flags); - if (buddy_cq_found == false) - return; - if (cq->ibcq.comp_handler) { - spin_lock_irqsave(&cq->comp_handler_lock, flags); - (*cq->ibcq.comp_handler) (&cq->ibcq, cq->ibcq.cq_context); - spin_unlock_irqrestore(&cq->comp_handler_lock, flags); + + /* if there is valid buddy cq, look for its completion handler */ + if (bcq && bcq->ibcq.comp_handler) { + spin_lock_irqsave(&bcq->comp_handler_lock, flags); + (*bcq->ibcq.comp_handler) (&bcq->ibcq, bcq->ibcq.cq_context); + spin_unlock_irqrestore(&bcq->comp_handler_lock, flags); } } @@ -935,6 +960,7 @@ static irqreturn_t ocrdma_irq_handler(int irq, void *handle) } while (budget); + eq->aic_obj.eq_intr_cnt++; ocrdma_ring_eq_db(dev, eq->q.id, true, true, 0); return IRQ_HANDLED; } @@ -1050,6 +1076,9 @@ static void ocrdma_get_attr(struct ocrdma_dev *dev, attr->max_pd = (rsp->max_pd_ca_ack_delay & OCRDMA_MBX_QUERY_CFG_MAX_PD_MASK) >> OCRDMA_MBX_QUERY_CFG_MAX_PD_SHIFT; + attr->max_dpp_pds = + (rsp->max_dpp_pds_credits & OCRDMA_MBX_QUERY_CFG_MAX_DPP_PDS_MASK) >> + OCRDMA_MBX_QUERY_CFG_MAX_DPP_PDS_OFFSET; attr->max_qp = (rsp->qp_srq_cq_ird_ord & OCRDMA_MBX_QUERY_CFG_MAX_QP_MASK) >> OCRDMA_MBX_QUERY_CFG_MAX_QP_SHIFT; @@ -1396,6 +1425,122 @@ int ocrdma_mbx_dealloc_pd(struct ocrdma_dev *dev, struct ocrdma_pd *pd) return status; } + +static int ocrdma_mbx_alloc_pd_range(struct ocrdma_dev *dev) +{ + int status = -ENOMEM; + size_t pd_bitmap_size; + struct ocrdma_alloc_pd_range *cmd; + struct ocrdma_alloc_pd_range_rsp *rsp; + + /* Pre allocate the DPP PDs */ + cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_ALLOC_PD_RANGE, sizeof(*cmd)); + if (!cmd) + return -ENOMEM; + cmd->pd_count = dev->attr.max_dpp_pds; + cmd->enable_dpp_rsvd |= OCRDMA_ALLOC_PD_ENABLE_DPP; + status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); + if (status) + goto mbx_err; + rsp = (struct ocrdma_alloc_pd_range_rsp *)cmd; + + if ((rsp->dpp_page_pdid & OCRDMA_ALLOC_PD_RSP_DPP) && rsp->pd_count) { + dev->pd_mgr->dpp_page_index = rsp->dpp_page_pdid >> + OCRDMA_ALLOC_PD_RSP_DPP_PAGE_SHIFT; + dev->pd_mgr->pd_dpp_start = rsp->dpp_page_pdid & + OCRDMA_ALLOC_PD_RNG_RSP_START_PDID_MASK; + dev->pd_mgr->max_dpp_pd = rsp->pd_count; + pd_bitmap_size = BITS_TO_LONGS(rsp->pd_count) * sizeof(long); + dev->pd_mgr->pd_dpp_bitmap = kzalloc(pd_bitmap_size, + GFP_KERNEL); + } + kfree(cmd); + + cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_ALLOC_PD_RANGE, sizeof(*cmd)); + if (!cmd) + return -ENOMEM; + + cmd->pd_count = dev->attr.max_pd - dev->attr.max_dpp_pds; + status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); + if (status) + goto mbx_err; + rsp = (struct ocrdma_alloc_pd_range_rsp *)cmd; + if (rsp->pd_count) { + dev->pd_mgr->pd_norm_start = rsp->dpp_page_pdid & + OCRDMA_ALLOC_PD_RNG_RSP_START_PDID_MASK; + dev->pd_mgr->max_normal_pd = rsp->pd_count; + pd_bitmap_size = BITS_TO_LONGS(rsp->pd_count) * sizeof(long); + dev->pd_mgr->pd_norm_bitmap = kzalloc(pd_bitmap_size, + GFP_KERNEL); + } + + if (dev->pd_mgr->pd_norm_bitmap || dev->pd_mgr->pd_dpp_bitmap) { + /* Enable PD resource manager */ + dev->pd_mgr->pd_prealloc_valid = true; + } else { + return -ENOMEM; + } +mbx_err: + kfree(cmd); + return status; +} + +static void ocrdma_mbx_dealloc_pd_range(struct ocrdma_dev *dev) +{ + struct ocrdma_dealloc_pd_range *cmd; + + /* return normal PDs to firmware */ + cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_DEALLOC_PD_RANGE, sizeof(*cmd)); + if (!cmd) + goto mbx_err; + + if (dev->pd_mgr->max_normal_pd) { + cmd->start_pd_id = dev->pd_mgr->pd_norm_start; + cmd->pd_count = dev->pd_mgr->max_normal_pd; + ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); + } + + if (dev->pd_mgr->max_dpp_pd) { + kfree(cmd); + /* return DPP PDs to firmware */ + cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_DEALLOC_PD_RANGE, + sizeof(*cmd)); + if (!cmd) + goto mbx_err; + + cmd->start_pd_id = dev->pd_mgr->pd_dpp_start; + cmd->pd_count = dev->pd_mgr->max_dpp_pd; + ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); + } +mbx_err: + kfree(cmd); +} + +void ocrdma_alloc_pd_pool(struct ocrdma_dev *dev) +{ + int status; + + dev->pd_mgr = kzalloc(sizeof(struct ocrdma_pd_resource_mgr), + GFP_KERNEL); + if (!dev->pd_mgr) { + pr_err("%s(%d)Memory allocation failure.\n", __func__, dev->id); + return; + } + status = ocrdma_mbx_alloc_pd_range(dev); + if (status) { + pr_err("%s(%d) Unable to initialize PD pool, using default.\n", + __func__, dev->id); + } +} + +static void ocrdma_free_pd_pool(struct ocrdma_dev *dev) +{ + ocrdma_mbx_dealloc_pd_range(dev); + kfree(dev->pd_mgr->pd_norm_bitmap); + kfree(dev->pd_mgr->pd_dpp_bitmap); + kfree(dev->pd_mgr); +} + static int ocrdma_build_q_conf(u32 *num_entries, int entry_size, int *num_pages, int *page_size) { @@ -1896,8 +2041,9 @@ void ocrdma_flush_qp(struct ocrdma_qp *qp) { bool found; unsigned long flags; + struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); - spin_lock_irqsave(&qp->dev->flush_q_lock, flags); + spin_lock_irqsave(&dev->flush_q_lock, flags); found = ocrdma_is_qp_in_sq_flushlist(qp->sq_cq, qp); if (!found) list_add_tail(&qp->sq_entry, &qp->sq_cq->sq_head); @@ -1906,7 +2052,7 @@ void ocrdma_flush_qp(struct ocrdma_qp *qp) if (!found) list_add_tail(&qp->rq_entry, &qp->rq_cq->rq_head); } - spin_unlock_irqrestore(&qp->dev->flush_q_lock, flags); + spin_unlock_irqrestore(&dev->flush_q_lock, flags); } static void ocrdma_init_hwq_ptr(struct ocrdma_qp *qp) @@ -1972,7 +2118,8 @@ static int ocrdma_set_create_qp_sq_cmd(struct ocrdma_create_qp_req *cmd, int status; u32 len, hw_pages, hw_page_size; dma_addr_t pa; - struct ocrdma_dev *dev = qp->dev; + struct ocrdma_pd *pd = qp->pd; + struct ocrdma_dev *dev = get_ocrdma_dev(pd->ibpd.device); struct pci_dev *pdev = dev->nic_info.pdev; u32 max_wqe_allocated; u32 max_sges = attrs->cap.max_send_sge; @@ -2027,7 +2174,8 @@ static int ocrdma_set_create_qp_rq_cmd(struct ocrdma_create_qp_req *cmd, int status; u32 len, hw_pages, hw_page_size; dma_addr_t pa = 0; - struct ocrdma_dev *dev = qp->dev; + struct ocrdma_pd *pd = qp->pd; + struct ocrdma_dev *dev = get_ocrdma_dev(pd->ibpd.device); struct pci_dev *pdev = dev->nic_info.pdev; u32 max_rqe_allocated = attrs->cap.max_recv_wr + 1; @@ -2086,7 +2234,8 @@ static void ocrdma_set_create_qp_dpp_cmd(struct ocrdma_create_qp_req *cmd, static int ocrdma_set_create_qp_ird_cmd(struct ocrdma_create_qp_req *cmd, struct ocrdma_qp *qp) { - struct ocrdma_dev *dev = qp->dev; + struct ocrdma_pd *pd = qp->pd; + struct ocrdma_dev *dev = get_ocrdma_dev(pd->ibpd.device); struct pci_dev *pdev = dev->nic_info.pdev; dma_addr_t pa = 0; int ird_page_size = dev->attr.ird_page_size; @@ -2157,8 +2306,8 @@ int ocrdma_mbx_create_qp(struct ocrdma_qp *qp, struct ib_qp_init_attr *attrs, { int status = -ENOMEM; u32 flags = 0; - struct ocrdma_dev *dev = qp->dev; struct ocrdma_pd *pd = qp->pd; + struct ocrdma_dev *dev = get_ocrdma_dev(pd->ibpd.device); struct pci_dev *pdev = dev->nic_info.pdev; struct ocrdma_cq *cq; struct ocrdma_create_qp_req *cmd; @@ -2281,11 +2430,12 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, union ib_gid sgid, zgid; u32 vlan_id; u8 mac_addr[6]; + struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); if ((ah_attr->ah_flags & IB_AH_GRH) == 0) return -EINVAL; - if (atomic_cmpxchg(&qp->dev->update_sl, 1, 0)) - ocrdma_init_service_level(qp->dev); + if (atomic_cmpxchg(&dev->update_sl, 1, 0)) + ocrdma_init_service_level(dev); cmd->params.tclass_sq_psn |= (ah_attr->grh.traffic_class << OCRDMA_QP_PARAMS_TCLASS_SHIFT); cmd->params.rnt_rc_sl_fl |= @@ -2296,7 +2446,7 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, cmd->flags |= OCRDMA_QP_PARA_FLOW_LBL_VALID; memcpy(&cmd->params.dgid[0], &ah_attr->grh.dgid.raw[0], sizeof(cmd->params.dgid)); - status = ocrdma_query_gid(&qp->dev->ibdev, 1, + status = ocrdma_query_gid(&dev->ibdev, 1, ah_attr->grh.sgid_index, &sgid); if (status) return status; @@ -2307,7 +2457,9 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, qp->sgid_idx = ah_attr->grh.sgid_index; memcpy(&cmd->params.sgid[0], &sgid.raw[0], sizeof(cmd->params.sgid)); - ocrdma_resolve_dmac(qp->dev, ah_attr, &mac_addr[0]); + status = ocrdma_resolve_dmac(dev, ah_attr, &mac_addr[0]); + if (status) + return status; cmd->params.dmac_b0_to_b3 = mac_addr[0] | (mac_addr[1] << 8) | (mac_addr[2] << 16) | (mac_addr[3] << 24); /* convert them to LE format. */ @@ -2320,7 +2472,7 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, vlan_id << OCRDMA_QP_PARAMS_VLAN_SHIFT; cmd->flags |= OCRDMA_QP_PARA_VLAN_EN_VALID; cmd->params.rnt_rc_sl_fl |= - (qp->dev->sl & 0x07) << OCRDMA_QP_PARAMS_SL_SHIFT; + (dev->sl & 0x07) << OCRDMA_QP_PARAMS_SL_SHIFT; } return 0; } @@ -2330,6 +2482,7 @@ static int ocrdma_set_qp_params(struct ocrdma_qp *qp, struct ib_qp_attr *attrs, int attr_mask) { int status = 0; + struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); if (attr_mask & IB_QP_PKEY_INDEX) { cmd->params.path_mtu_pkey_indx |= (attrs->pkey_index & @@ -2347,12 +2500,12 @@ static int ocrdma_set_qp_params(struct ocrdma_qp *qp, return status; } else if (qp->qp_type == IB_QPT_GSI || qp->qp_type == IB_QPT_UD) { /* set the default mac address for UD, GSI QPs */ - cmd->params.dmac_b0_to_b3 = qp->dev->nic_info.mac_addr[0] | - (qp->dev->nic_info.mac_addr[1] << 8) | - (qp->dev->nic_info.mac_addr[2] << 16) | - (qp->dev->nic_info.mac_addr[3] << 24); - cmd->params.vlan_dmac_b4_to_b5 = qp->dev->nic_info.mac_addr[4] | - (qp->dev->nic_info.mac_addr[5] << 8); + cmd->params.dmac_b0_to_b3 = dev->nic_info.mac_addr[0] | + (dev->nic_info.mac_addr[1] << 8) | + (dev->nic_info.mac_addr[2] << 16) | + (dev->nic_info.mac_addr[3] << 24); + cmd->params.vlan_dmac_b4_to_b5 = dev->nic_info.mac_addr[4] | + (dev->nic_info.mac_addr[5] << 8); } if ((attr_mask & IB_QP_EN_SQD_ASYNC_NOTIFY) && attrs->en_sqd_async_notify) { @@ -2409,7 +2562,7 @@ static int ocrdma_set_qp_params(struct ocrdma_qp *qp, cmd->flags |= OCRDMA_QP_PARA_RQPSN_VALID; } if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC) { - if (attrs->max_rd_atomic > qp->dev->attr.max_ord_per_qp) { + if (attrs->max_rd_atomic > dev->attr.max_ord_per_qp) { status = -EINVAL; goto pmtu_err; } @@ -2417,7 +2570,7 @@ static int ocrdma_set_qp_params(struct ocrdma_qp *qp, cmd->flags |= OCRDMA_QP_PARA_MAX_ORD_VALID; } if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC) { - if (attrs->max_dest_rd_atomic > qp->dev->attr.max_ird_per_qp) { + if (attrs->max_dest_rd_atomic > dev->attr.max_ird_per_qp) { status = -EINVAL; goto pmtu_err; } @@ -2870,6 +3023,82 @@ done: return status; } +static int ocrdma_mbx_modify_eqd(struct ocrdma_dev *dev, struct ocrdma_eq *eq, + int num) +{ + int i, status = -ENOMEM; + struct ocrdma_modify_eqd_req *cmd; + + cmd = ocrdma_init_emb_mqe(OCRDMA_CMD_MODIFY_EQ_DELAY, sizeof(*cmd)); + if (!cmd) + return status; + + ocrdma_init_mch(&cmd->cmd.req, OCRDMA_CMD_MODIFY_EQ_DELAY, + OCRDMA_SUBSYS_COMMON, sizeof(*cmd)); + + cmd->cmd.num_eq = num; + for (i = 0; i < num; i++) { + cmd->cmd.set_eqd[i].eq_id = eq[i].q.id; + cmd->cmd.set_eqd[i].phase = 0; + cmd->cmd.set_eqd[i].delay_multiplier = + (eq[i].aic_obj.prev_eqd * 65)/100; + } + status = ocrdma_mbx_cmd(dev, (struct ocrdma_mqe *)cmd); + if (status) + goto mbx_err; +mbx_err: + kfree(cmd); + return status; +} + +static int ocrdma_modify_eqd(struct ocrdma_dev *dev, struct ocrdma_eq *eq, + int num) +{ + int num_eqs, i = 0; + if (num > 8) { + while (num) { + num_eqs = min(num, 8); + ocrdma_mbx_modify_eqd(dev, &eq[i], num_eqs); + i += num_eqs; + num -= num_eqs; + } + } else { + ocrdma_mbx_modify_eqd(dev, eq, num); + } + return 0; +} + +void ocrdma_eqd_set_task(struct work_struct *work) +{ + struct ocrdma_dev *dev = + container_of(work, struct ocrdma_dev, eqd_work.work); + struct ocrdma_eq *eq = 0; + int i, num = 0, status = -EINVAL; + u64 eq_intr; + + for (i = 0; i < dev->eq_cnt; i++) { + eq = &dev->eq_tbl[i]; + if (eq->aic_obj.eq_intr_cnt > eq->aic_obj.prev_eq_intr_cnt) { + eq_intr = eq->aic_obj.eq_intr_cnt - + eq->aic_obj.prev_eq_intr_cnt; + if ((eq_intr > EQ_INTR_PER_SEC_THRSH_HI) && + (eq->aic_obj.prev_eqd == EQ_AIC_MIN_EQD)) { + eq->aic_obj.prev_eqd = EQ_AIC_MAX_EQD; + num++; + } else if ((eq_intr < EQ_INTR_PER_SEC_THRSH_LOW) && + (eq->aic_obj.prev_eqd == EQ_AIC_MAX_EQD)) { + eq->aic_obj.prev_eqd = EQ_AIC_MIN_EQD; + num++; + } + } + eq->aic_obj.prev_eq_intr_cnt = eq->aic_obj.eq_intr_cnt; + } + + if (num) + status = ocrdma_modify_eqd(dev, &dev->eq_tbl[0], num); + schedule_delayed_work(&dev->eqd_work, msecs_to_jiffies(1000)); +} + int ocrdma_init_hw(struct ocrdma_dev *dev) { int status; @@ -2915,6 +3144,7 @@ qpeq_err: void ocrdma_cleanup_hw(struct ocrdma_dev *dev) { + ocrdma_free_pd_pool(dev); ocrdma_mbx_delete_ah_tbl(dev); /* cleanup the eqs */ diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.h b/drivers/infiniband/hw/ocrdma/ocrdma_hw.h index 6eed8f19..e905972 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.h @@ -136,5 +136,7 @@ int ocrdma_get_irq(struct ocrdma_dev *dev, struct ocrdma_eq *eq); int ocrdma_mbx_rdma_stats(struct ocrdma_dev *, bool reset); char *port_speed_string(struct ocrdma_dev *dev); void ocrdma_init_service_level(struct ocrdma_dev *); +void ocrdma_alloc_pd_pool(struct ocrdma_dev *dev); +void ocrdma_free_pd_range(struct ocrdma_dev *dev); #endif /* __OCRDMA_HW_H__ */ diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_main.c b/drivers/infiniband/hw/ocrdma/ocrdma_main.c index b0b2257..7a2b59a 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_main.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_main.c @@ -239,7 +239,7 @@ static int ocrdma_register_device(struct ocrdma_dev *dev) dev->ibdev.node_type = RDMA_NODE_IB_CA; dev->ibdev.phys_port_cnt = 1; - dev->ibdev.num_comp_vectors = 1; + dev->ibdev.num_comp_vectors = dev->eq_cnt; /* mandatory verbs. */ dev->ibdev.query_device = ocrdma_query_device; @@ -329,6 +329,8 @@ static int ocrdma_alloc_resources(struct ocrdma_dev *dev) if (dev->stag_arr == NULL) goto alloc_err; + ocrdma_alloc_pd_pool(dev); + spin_lock_init(&dev->av_tbl.lock); spin_lock_init(&dev->flush_q_lock); return 0; @@ -491,6 +493,9 @@ static struct ocrdma_dev *ocrdma_add(struct be_dev_info *dev_info) spin_unlock(&ocrdma_devlist_lock); /* Init stats */ ocrdma_add_port_stats(dev); + /* Interrupt Moderation */ + INIT_DELAYED_WORK(&dev->eqd_work, ocrdma_eqd_set_task); + schedule_delayed_work(&dev->eqd_work, msecs_to_jiffies(1000)); pr_info("%s %s: %s \"%s\" port %d\n", dev_name(&dev->nic_info.pdev->dev), hca_name(dev), @@ -528,11 +533,12 @@ static void ocrdma_remove(struct ocrdma_dev *dev) /* first unregister with stack to stop all the active traffic * of the registered clients. */ - ocrdma_rem_port_stats(dev); + cancel_delayed_work_sync(&dev->eqd_work); ocrdma_remove_sysfiles(dev); - ib_unregister_device(&dev->ibdev); + ocrdma_rem_port_stats(dev); + spin_lock(&ocrdma_devlist_lock); list_del_rcu(&dev->entry); spin_unlock(&ocrdma_devlist_lock); diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h index 4e03648..243c87c 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h @@ -75,6 +75,8 @@ enum { OCRDMA_CMD_DESTROY_RBQ = 26, OCRDMA_CMD_GET_RDMA_STATS = 27, + OCRDMA_CMD_ALLOC_PD_RANGE = 28, + OCRDMA_CMD_DEALLOC_PD_RANGE = 29, OCRDMA_CMD_MAX }; @@ -87,6 +89,7 @@ enum { OCRDMA_CMD_CREATE_MQ = 21, OCRDMA_CMD_GET_CTRL_ATTRIBUTES = 32, OCRDMA_CMD_GET_FW_VER = 35, + OCRDMA_CMD_MODIFY_EQ_DELAY = 41, OCRDMA_CMD_DELETE_MQ = 53, OCRDMA_CMD_DELETE_CQ = 54, OCRDMA_CMD_DELETE_EQ = 55, @@ -101,7 +104,7 @@ enum { QTYPE_MCCQ = 3 }; -#define OCRDMA_MAX_SGID 8 +#define OCRDMA_MAX_SGID 16 #define OCRDMA_MAX_QP 2048 #define OCRDMA_MAX_CQ 2048 @@ -314,6 +317,29 @@ struct ocrdma_create_eq_rsp { #define OCRDMA_EQ_MINOR_OTHER 0x1 +struct ocrmda_set_eqd { + u32 eq_id; + u32 phase; + u32 delay_multiplier; +}; + +struct ocrdma_modify_eqd_cmd { + struct ocrdma_mbx_hdr req; + u32 num_eq; + struct ocrmda_set_eqd set_eqd[8]; +} __packed; + +struct ocrdma_modify_eqd_req { + struct ocrdma_mqe_hdr hdr; + struct ocrdma_modify_eqd_cmd cmd; +}; + + +struct ocrdma_modify_eq_delay_rsp { + struct ocrdma_mbx_rsp hdr; + u32 rsvd0; +} __packed; + enum { OCRDMA_MCQE_STATUS_SHIFT = 0, OCRDMA_MCQE_STATUS_MASK = 0xFFFF, @@ -441,7 +467,9 @@ enum OCRDMA_ASYNC_EVENT_TYPE { OCRDMA_DEVICE_FATAL_EVENT = 0x08, OCRDMA_SRQCAT_ERROR = 0x0E, OCRDMA_SRQ_LIMIT_EVENT = 0x0F, - OCRDMA_QP_LAST_WQE_EVENT = 0x10 + OCRDMA_QP_LAST_WQE_EVENT = 0x10, + + OCRDMA_MAX_ASYNC_ERRORS }; /* mailbox command request and responses */ @@ -1297,6 +1325,37 @@ struct ocrdma_dealloc_pd_rsp { struct ocrdma_mbx_rsp rsp; }; +struct ocrdma_alloc_pd_range { + struct ocrdma_mqe_hdr hdr; + struct ocrdma_mbx_hdr req; + u32 enable_dpp_rsvd; + u32 pd_count; +}; + +struct ocrdma_alloc_pd_range_rsp { + struct ocrdma_mqe_hdr hdr; + struct ocrdma_mbx_rsp rsp; + u32 dpp_page_pdid; + u32 pd_count; +}; + +enum { + OCRDMA_ALLOC_PD_RNG_RSP_START_PDID_MASK = 0xFFFF, +}; + +struct ocrdma_dealloc_pd_range { + struct ocrdma_mqe_hdr hdr; + struct ocrdma_mbx_hdr req; + u32 start_pd_id; + u32 pd_count; +}; + +struct ocrdma_dealloc_pd_range_rsp { + struct ocrdma_mqe_hdr hdr; + struct ocrdma_mbx_hdr req; + u32 rsvd; +}; + enum { OCRDMA_ADDR_CHECK_ENABLE = 1, OCRDMA_ADDR_CHECK_DISABLE = 0 @@ -1597,7 +1656,9 @@ enum OCRDMA_CQE_STATUS { OCRDMA_CQE_INV_EEC_STATE_ERR, OCRDMA_CQE_FATAL_ERR, OCRDMA_CQE_RESP_TIMEOUT_ERR, - OCRDMA_CQE_GENERAL_ERR + OCRDMA_CQE_GENERAL_ERR, + + OCRDMA_MAX_CQE_ERR }; enum { @@ -1673,6 +1734,7 @@ enum { OCRDMA_FLAG_FENCE_R = 0x8, OCRDMA_FLAG_SOLICIT = 0x10, OCRDMA_FLAG_IMM = 0x20, + OCRDMA_FLAG_AH_VLAN_PR = 0x40, /* Stag flags */ OCRDMA_LKEY_FLAG_LOCAL_WR = 0x1, diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c index 41a9aec..48d7ef5 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c @@ -26,6 +26,7 @@ *******************************************************************/ #include <rdma/ib_addr.h> +#include <rdma/ib_pma.h> #include "ocrdma_stats.h" static struct dentry *ocrdma_dbgfs_dir; @@ -249,6 +250,27 @@ static char *ocrdma_rx_stats(struct ocrdma_dev *dev) return stats; } +static u64 ocrdma_sysfs_rcv_pkts(struct ocrdma_dev *dev) +{ + struct ocrdma_rdma_stats_resp *rdma_stats = + (struct ocrdma_rdma_stats_resp *)dev->stats_mem.va; + struct ocrdma_rx_stats *rx_stats = &rdma_stats->rx_stats; + + return convert_to_64bit(rx_stats->roce_frames_lo, + rx_stats->roce_frames_hi) + (u64)rx_stats->roce_frame_icrc_drops + + (u64)rx_stats->roce_frame_payload_len_drops; +} + +static u64 ocrdma_sysfs_rcv_data(struct ocrdma_dev *dev) +{ + struct ocrdma_rdma_stats_resp *rdma_stats = + (struct ocrdma_rdma_stats_resp *)dev->stats_mem.va; + struct ocrdma_rx_stats *rx_stats = &rdma_stats->rx_stats; + + return (convert_to_64bit(rx_stats->roce_frame_bytes_lo, + rx_stats->roce_frame_bytes_hi))/4; +} + static char *ocrdma_tx_stats(struct ocrdma_dev *dev) { char *stats = dev->stats_mem.debugfs_mem, *pcur; @@ -292,6 +314,37 @@ static char *ocrdma_tx_stats(struct ocrdma_dev *dev) return stats; } +static u64 ocrdma_sysfs_xmit_pkts(struct ocrdma_dev *dev) +{ + struct ocrdma_rdma_stats_resp *rdma_stats = + (struct ocrdma_rdma_stats_resp *)dev->stats_mem.va; + struct ocrdma_tx_stats *tx_stats = &rdma_stats->tx_stats; + + return (convert_to_64bit(tx_stats->send_pkts_lo, + tx_stats->send_pkts_hi) + + convert_to_64bit(tx_stats->write_pkts_lo, tx_stats->write_pkts_hi) + + convert_to_64bit(tx_stats->read_pkts_lo, tx_stats->read_pkts_hi) + + convert_to_64bit(tx_stats->read_rsp_pkts_lo, + tx_stats->read_rsp_pkts_hi) + + convert_to_64bit(tx_stats->ack_pkts_lo, tx_stats->ack_pkts_hi)); +} + +static u64 ocrdma_sysfs_xmit_data(struct ocrdma_dev *dev) +{ + struct ocrdma_rdma_stats_resp *rdma_stats = + (struct ocrdma_rdma_stats_resp *)dev->stats_mem.va; + struct ocrdma_tx_stats *tx_stats = &rdma_stats->tx_stats; + + return (convert_to_64bit(tx_stats->send_bytes_lo, + tx_stats->send_bytes_hi) + + convert_to_64bit(tx_stats->write_bytes_lo, + tx_stats->write_bytes_hi) + + convert_to_64bit(tx_stats->read_req_bytes_lo, + tx_stats->read_req_bytes_hi) + + convert_to_64bit(tx_stats->read_rsp_bytes_lo, + tx_stats->read_rsp_bytes_hi))/4; +} + static char *ocrdma_wqe_stats(struct ocrdma_dev *dev) { char *stats = dev->stats_mem.debugfs_mem, *pcur; @@ -432,10 +485,118 @@ static char *ocrdma_rx_dbg_stats(struct ocrdma_dev *dev) return dev->stats_mem.debugfs_mem; } +static char *ocrdma_driver_dbg_stats(struct ocrdma_dev *dev) +{ + char *stats = dev->stats_mem.debugfs_mem, *pcur; + + + memset(stats, 0, (OCRDMA_MAX_DBGFS_MEM)); + + pcur = stats; + pcur += ocrdma_add_stat(stats, pcur, "async_cq_err", + (u64)(dev->async_err_stats + [OCRDMA_CQ_ERROR].counter)); + pcur += ocrdma_add_stat(stats, pcur, "async_cq_overrun_err", + (u64)dev->async_err_stats + [OCRDMA_CQ_OVERRUN_ERROR].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_cq_qpcat_err", + (u64)dev->async_err_stats + [OCRDMA_CQ_QPCAT_ERROR].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_qp_access_err", + (u64)dev->async_err_stats + [OCRDMA_QP_ACCESS_ERROR].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_qp_commm_est_evt", + (u64)dev->async_err_stats + [OCRDMA_QP_COMM_EST_EVENT].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_sq_drained_evt", + (u64)dev->async_err_stats + [OCRDMA_SQ_DRAINED_EVENT].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_dev_fatal_evt", + (u64)dev->async_err_stats + [OCRDMA_DEVICE_FATAL_EVENT].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_srqcat_err", + (u64)dev->async_err_stats + [OCRDMA_SRQCAT_ERROR].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_srq_limit_evt", + (u64)dev->async_err_stats + [OCRDMA_SRQ_LIMIT_EVENT].counter); + pcur += ocrdma_add_stat(stats, pcur, "async_qp_last_wqe_evt", + (u64)dev->async_err_stats + [OCRDMA_QP_LAST_WQE_EVENT].counter); + + pcur += ocrdma_add_stat(stats, pcur, "cqe_loc_len_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_LOC_LEN_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_loc_qp_op_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_LOC_QP_OP_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_loc_eec_op_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_LOC_EEC_OP_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_loc_prot_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_LOC_PROT_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_wr_flush_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_WR_FLUSH_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_mw_bind_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_MW_BIND_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_bad_resp_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_BAD_RESP_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_loc_access_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_LOC_ACCESS_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_rem_inv_req_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_REM_INV_REQ_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_rem_access_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_REM_ACCESS_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_rem_op_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_REM_OP_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_retry_exc_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_RETRY_EXC_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_rnr_retry_exc_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_RNR_RETRY_EXC_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_loc_rdd_viol_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_LOC_RDD_VIOL_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_rem_inv_rd_req_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_REM_INV_RD_REQ_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_rem_abort_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_REM_ABORT_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_inv_eecn_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_INV_EECN_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_inv_eec_state_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_INV_EEC_STATE_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_fatal_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_FATAL_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_resp_timeout_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_RESP_TIMEOUT_ERR].counter); + pcur += ocrdma_add_stat(stats, pcur, "cqe_general_err", + (u64)dev->cqe_err_stats + [OCRDMA_CQE_GENERAL_ERR].counter); + return stats; +} + static void ocrdma_update_stats(struct ocrdma_dev *dev) { ulong now = jiffies, secs; int status = 0; + struct ocrdma_rdma_stats_resp *rdma_stats = + (struct ocrdma_rdma_stats_resp *)dev->stats_mem.va; + struct ocrdma_rsrc_stats *rsrc_stats = &rdma_stats->act_rsrc_stats; secs = jiffies_to_msecs(now - dev->last_stats_time) / 1000U; if (secs) { @@ -444,10 +605,74 @@ static void ocrdma_update_stats(struct ocrdma_dev *dev) if (status) pr_err("%s: stats mbox failed with status = %d\n", __func__, status); + /* Update PD counters from PD resource manager */ + if (dev->pd_mgr->pd_prealloc_valid) { + rsrc_stats->dpp_pds = dev->pd_mgr->pd_dpp_count; + rsrc_stats->non_dpp_pds = dev->pd_mgr->pd_norm_count; + /* Threshold stata*/ + rsrc_stats = &rdma_stats->th_rsrc_stats; + rsrc_stats->dpp_pds = dev->pd_mgr->pd_dpp_thrsh; + rsrc_stats->non_dpp_pds = dev->pd_mgr->pd_norm_thrsh; + } dev->last_stats_time = jiffies; } } +static ssize_t ocrdma_dbgfs_ops_write(struct file *filp, + const char __user *buffer, + size_t count, loff_t *ppos) +{ + char tmp_str[32]; + long reset; + int status = 0; + struct ocrdma_stats *pstats = filp->private_data; + struct ocrdma_dev *dev = pstats->dev; + + if (count > 32) + goto err; + + if (copy_from_user(tmp_str, buffer, count)) + goto err; + + tmp_str[count-1] = '\0'; + if (kstrtol(tmp_str, 10, &reset)) + goto err; + + switch (pstats->type) { + case OCRDMA_RESET_STATS: + if (reset) { + status = ocrdma_mbx_rdma_stats(dev, true); + if (status) { + pr_err("Failed to reset stats = %d", status); + goto err; + } + } + break; + default: + goto err; + } + + return count; +err: + return -EFAULT; +} + +int ocrdma_pma_counters(struct ocrdma_dev *dev, + struct ib_mad *out_mad) +{ + struct ib_pma_portcounters *pma_cnt; + + memset(out_mad->data, 0, sizeof out_mad->data); + pma_cnt = (void *)(out_mad->data + 40); + ocrdma_update_stats(dev); + + pma_cnt->port_xmit_data = cpu_to_be32(ocrdma_sysfs_xmit_data(dev)); + pma_cnt->port_rcv_data = cpu_to_be32(ocrdma_sysfs_rcv_data(dev)); + pma_cnt->port_xmit_packets = cpu_to_be32(ocrdma_sysfs_xmit_pkts(dev)); + pma_cnt->port_rcv_packets = cpu_to_be32(ocrdma_sysfs_rcv_pkts(dev)); + return 0; +} + static ssize_t ocrdma_dbgfs_ops_read(struct file *filp, char __user *buffer, size_t usr_buf_len, loff_t *ppos) { @@ -492,6 +717,9 @@ static ssize_t ocrdma_dbgfs_ops_read(struct file *filp, char __user *buffer, case OCRDMA_RX_DBG_STATS: data = ocrdma_rx_dbg_stats(dev); break; + case OCRDMA_DRV_STATS: + data = ocrdma_driver_dbg_stats(dev); + break; default: status = -EFAULT; @@ -514,6 +742,7 @@ static const struct file_operations ocrdma_dbg_ops = { .owner = THIS_MODULE, .open = simple_open, .read = ocrdma_dbgfs_ops_read, + .write = ocrdma_dbgfs_ops_write, }; void ocrdma_add_port_stats(struct ocrdma_dev *dev) @@ -582,6 +811,18 @@ void ocrdma_add_port_stats(struct ocrdma_dev *dev) &dev->rx_dbg_stats, &ocrdma_dbg_ops)) goto err; + dev->driver_stats.type = OCRDMA_DRV_STATS; + dev->driver_stats.dev = dev; + if (!debugfs_create_file("driver_dbg_stats", S_IRUSR, dev->dir, + &dev->driver_stats, &ocrdma_dbg_ops)) + goto err; + + dev->reset_stats.type = OCRDMA_RESET_STATS; + dev->reset_stats.dev = dev; + if (!debugfs_create_file("reset_stats", S_IRUSR, dev->dir, + &dev->reset_stats, &ocrdma_dbg_ops)) + goto err; + /* Now create dma_mem for stats mbx command */ if (!ocrdma_alloc_stats_mem(dev)) goto err; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_stats.h b/drivers/infiniband/hw/ocrdma/ocrdma_stats.h index 5f5e20c..091edd6 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_stats.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_stats.h @@ -43,12 +43,16 @@ enum OCRDMA_STATS_TYPE { OCRDMA_RXQP_ERRSTATS, OCRDMA_TXQP_ERRSTATS, OCRDMA_TX_DBG_STATS, - OCRDMA_RX_DBG_STATS + OCRDMA_RX_DBG_STATS, + OCRDMA_DRV_STATS, + OCRDMA_RESET_STATS }; void ocrdma_rem_debugfs(void); void ocrdma_init_debugfs(void); void ocrdma_rem_port_stats(struct ocrdma_dev *dev); void ocrdma_add_port_stats(struct ocrdma_dev *dev); +int ocrdma_pma_counters(struct ocrdma_dev *dev, + struct ib_mad *out_mad); #endif /* __OCRDMA_STATS_H__ */ diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c index fb8d8c4..8771755 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c @@ -53,7 +53,7 @@ int ocrdma_query_gid(struct ib_device *ibdev, u8 port, dev = get_ocrdma_dev(ibdev); memset(sgid, 0, sizeof(*sgid)); - if (index > OCRDMA_MAX_SGID) + if (index >= OCRDMA_MAX_SGID) return -EINVAL; memcpy(sgid, &dev->sgid_tbl[index], sizeof(*sgid)); @@ -253,6 +253,107 @@ static bool ocrdma_search_mmap(struct ocrdma_ucontext *uctx, u64 phy_addr, return found; } + +static u16 _ocrdma_pd_mgr_get_bitmap(struct ocrdma_dev *dev, bool dpp_pool) +{ + u16 pd_bitmap_idx = 0; + const unsigned long *pd_bitmap; + + if (dpp_pool) { + pd_bitmap = dev->pd_mgr->pd_dpp_bitmap; + pd_bitmap_idx = find_first_zero_bit(pd_bitmap, + dev->pd_mgr->max_dpp_pd); + __set_bit(pd_bitmap_idx, dev->pd_mgr->pd_dpp_bitmap); + dev->pd_mgr->pd_dpp_count++; + if (dev->pd_mgr->pd_dpp_count > dev->pd_mgr->pd_dpp_thrsh) + dev->pd_mgr->pd_dpp_thrsh = dev->pd_mgr->pd_dpp_count; + } else { + pd_bitmap = dev->pd_mgr->pd_norm_bitmap; + pd_bitmap_idx = find_first_zero_bit(pd_bitmap, + dev->pd_mgr->max_normal_pd); + __set_bit(pd_bitmap_idx, dev->pd_mgr->pd_norm_bitmap); + dev->pd_mgr->pd_norm_count++; + if (dev->pd_mgr->pd_norm_count > dev->pd_mgr->pd_norm_thrsh) + dev->pd_mgr->pd_norm_thrsh = dev->pd_mgr->pd_norm_count; + } + return pd_bitmap_idx; +} + +static int _ocrdma_pd_mgr_put_bitmap(struct ocrdma_dev *dev, u16 pd_id, + bool dpp_pool) +{ + u16 pd_count; + u16 pd_bit_index; + + pd_count = dpp_pool ? dev->pd_mgr->pd_dpp_count : + dev->pd_mgr->pd_norm_count; + if (pd_count == 0) + return -EINVAL; + + if (dpp_pool) { + pd_bit_index = pd_id - dev->pd_mgr->pd_dpp_start; + if (pd_bit_index >= dev->pd_mgr->max_dpp_pd) { + return -EINVAL; + } else { + __clear_bit(pd_bit_index, dev->pd_mgr->pd_dpp_bitmap); + dev->pd_mgr->pd_dpp_count--; + } + } else { + pd_bit_index = pd_id - dev->pd_mgr->pd_norm_start; + if (pd_bit_index >= dev->pd_mgr->max_normal_pd) { + return -EINVAL; + } else { + __clear_bit(pd_bit_index, dev->pd_mgr->pd_norm_bitmap); + dev->pd_mgr->pd_norm_count--; + } + } + + return 0; +} + +static u8 ocrdma_put_pd_num(struct ocrdma_dev *dev, u16 pd_id, + bool dpp_pool) +{ + int status; + + mutex_lock(&dev->dev_lock); + status = _ocrdma_pd_mgr_put_bitmap(dev, pd_id, dpp_pool); + mutex_unlock(&dev->dev_lock); + return status; +} + +static int ocrdma_get_pd_num(struct ocrdma_dev *dev, struct ocrdma_pd *pd) +{ + u16 pd_idx = 0; + int status = 0; + + mutex_lock(&dev->dev_lock); + if (pd->dpp_enabled) { + /* try allocating DPP PD, if not available then normal PD */ + if (dev->pd_mgr->pd_dpp_count < dev->pd_mgr->max_dpp_pd) { + pd_idx = _ocrdma_pd_mgr_get_bitmap(dev, true); + pd->id = dev->pd_mgr->pd_dpp_start + pd_idx; + pd->dpp_page = dev->pd_mgr->dpp_page_index + pd_idx; + } else if (dev->pd_mgr->pd_norm_count < + dev->pd_mgr->max_normal_pd) { + pd_idx = _ocrdma_pd_mgr_get_bitmap(dev, false); + pd->id = dev->pd_mgr->pd_norm_start + pd_idx; + pd->dpp_enabled = false; + } else { + status = -EINVAL; + } + } else { + if (dev->pd_mgr->pd_norm_count < dev->pd_mgr->max_normal_pd) { + pd_idx = _ocrdma_pd_mgr_get_bitmap(dev, false); + pd->id = dev->pd_mgr->pd_norm_start + pd_idx; + } else { + status = -EINVAL; + } + } + mutex_unlock(&dev->dev_lock); + return status; +} + static struct ocrdma_pd *_ocrdma_alloc_pd(struct ocrdma_dev *dev, struct ocrdma_ucontext *uctx, struct ib_udata *udata) @@ -272,6 +373,11 @@ static struct ocrdma_pd *_ocrdma_alloc_pd(struct ocrdma_dev *dev, dev->attr.wqe_size) : 0; } + if (dev->pd_mgr->pd_prealloc_valid) { + status = ocrdma_get_pd_num(dev, pd); + return (status == 0) ? pd : ERR_PTR(status); + } + retry: status = ocrdma_mbx_alloc_pd(dev, pd); if (status) { @@ -299,7 +405,11 @@ static int _ocrdma_dealloc_pd(struct ocrdma_dev *dev, { int status = 0; - status = ocrdma_mbx_dealloc_pd(dev, pd); + if (dev->pd_mgr->pd_prealloc_valid) + status = ocrdma_put_pd_num(dev, pd->id, pd->dpp_enabled); + else + status = ocrdma_mbx_dealloc_pd(dev, pd); + kfree(pd); return status; } @@ -325,7 +435,6 @@ err: static int ocrdma_dealloc_ucontext_pd(struct ocrdma_ucontext *uctx) { - int status = 0; struct ocrdma_pd *pd = uctx->cntxt_pd; struct ocrdma_dev *dev = get_ocrdma_dev(pd->ibpd.device); @@ -334,8 +443,8 @@ static int ocrdma_dealloc_ucontext_pd(struct ocrdma_ucontext *uctx) __func__, dev->id, pd->id); } uctx->cntxt_pd = NULL; - status = _ocrdma_dealloc_pd(dev, pd); - return status; + (void)_ocrdma_dealloc_pd(dev, pd); + return 0; } static struct ocrdma_pd *ocrdma_get_ucontext_pd(struct ocrdma_ucontext *uctx) @@ -569,7 +678,7 @@ err: if (is_uctx_pd) { ocrdma_release_ucontext_pd(uctx); } else { - status = ocrdma_mbx_dealloc_pd(dev, pd); + status = _ocrdma_dealloc_pd(dev, pd); kfree(pd); } exit: @@ -837,9 +946,8 @@ int ocrdma_dereg_mr(struct ib_mr *ib_mr) { struct ocrdma_mr *mr = get_ocrdma_mr(ib_mr); struct ocrdma_dev *dev = get_ocrdma_dev(ib_mr->device); - int status; - status = ocrdma_mbx_dealloc_lkey(dev, mr->hwmr.fr_mr, mr->hwmr.lkey); + (void) ocrdma_mbx_dealloc_lkey(dev, mr->hwmr.fr_mr, mr->hwmr.lkey); ocrdma_free_mr_pbl_tbl(dev, &mr->hwmr); @@ -850,11 +958,10 @@ int ocrdma_dereg_mr(struct ib_mr *ib_mr) /* Don't stop cleanup, in case FW is unresponsive */ if (dev->mqe_ctx.fw_error_state) { - status = 0; pr_err("%s(%d) fw not responding.\n", __func__, dev->id); } - return status; + return 0; } static int ocrdma_copy_cq_uresp(struct ocrdma_dev *dev, struct ocrdma_cq *cq, @@ -986,7 +1093,6 @@ static void ocrdma_flush_cq(struct ocrdma_cq *cq) int ocrdma_destroy_cq(struct ib_cq *ibcq) { - int status; struct ocrdma_cq *cq = get_ocrdma_cq(ibcq); struct ocrdma_eq *eq = NULL; struct ocrdma_dev *dev = get_ocrdma_dev(ibcq->device); @@ -1003,7 +1109,7 @@ int ocrdma_destroy_cq(struct ib_cq *ibcq) synchronize_irq(irq); ocrdma_flush_cq(cq); - status = ocrdma_mbx_destroy_cq(dev, cq); + (void)ocrdma_mbx_destroy_cq(dev, cq); if (cq->ucontext) { pdid = cq->ucontext->cntxt_pd->id; ocrdma_del_mmap(cq->ucontext, (u64) cq->pa, @@ -1014,7 +1120,7 @@ int ocrdma_destroy_cq(struct ib_cq *ibcq) } kfree(cq); - return status; + return 0; } static int ocrdma_add_qpn_map(struct ocrdma_dev *dev, struct ocrdma_qp *qp) @@ -1113,8 +1219,8 @@ static int ocrdma_copy_qp_uresp(struct ocrdma_qp *qp, int status = 0; u64 usr_db; struct ocrdma_create_qp_uresp uresp; - struct ocrdma_dev *dev = qp->dev; struct ocrdma_pd *pd = qp->pd; + struct ocrdma_dev *dev = get_ocrdma_dev(pd->ibpd.device); memset(&uresp, 0, sizeof(uresp)); usr_db = dev->nic_info.unmapped_db + @@ -1253,7 +1359,6 @@ struct ib_qp *ocrdma_create_qp(struct ib_pd *ibpd, status = -ENOMEM; goto gen_err; } - qp->dev = dev; ocrdma_set_qp_init_params(qp, pd, attrs); if (udata == NULL) qp->cap_flags |= (OCRDMA_QP_MW_BIND | OCRDMA_QP_LKEY0 | @@ -1312,7 +1417,7 @@ int _ocrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, enum ib_qp_state old_qps; qp = get_ocrdma_qp(ibqp); - dev = qp->dev; + dev = get_ocrdma_dev(ibqp->device); if (attr_mask & IB_QP_STATE) status = ocrdma_qp_state_change(qp, attr->qp_state, &old_qps); /* if new and previous states are same hw doesn't need to @@ -1335,7 +1440,7 @@ int ocrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, enum ib_qp_state old_qps, new_qps; qp = get_ocrdma_qp(ibqp); - dev = qp->dev; + dev = get_ocrdma_dev(ibqp->device); /* syncronize with multiple context trying to change, retrive qps */ mutex_lock(&dev->dev_lock); @@ -1402,7 +1507,7 @@ int ocrdma_query_qp(struct ib_qp *ibqp, u32 qp_state; struct ocrdma_qp_params params; struct ocrdma_qp *qp = get_ocrdma_qp(ibqp); - struct ocrdma_dev *dev = qp->dev; + struct ocrdma_dev *dev = get_ocrdma_dev(ibqp->device); memset(¶ms, 0, sizeof(params)); mutex_lock(&dev->dev_lock); @@ -1412,8 +1517,6 @@ int ocrdma_query_qp(struct ib_qp *ibqp, goto mbx_err; if (qp->qp_type == IB_QPT_UD) qp_attr->qkey = params.qkey; - qp_attr->qp_state = get_ibqp_state(IB_QPS_INIT); - qp_attr->cur_qp_state = get_ibqp_state(IB_QPS_INIT); qp_attr->path_mtu = ocrdma_mtu_int_to_enum(params.path_mtu_pkey_indx & OCRDMA_QP_PARAMS_PATH_MTU_MASK) >> @@ -1468,6 +1571,8 @@ int ocrdma_query_qp(struct ib_qp *ibqp, memset(&qp_attr->alt_ah_attr, 0, sizeof(qp_attr->alt_ah_attr)); qp_state = (params.max_sge_recv_flags & OCRDMA_QP_PARAMS_STATE_MASK) >> OCRDMA_QP_PARAMS_STATE_SHIFT; + qp_attr->qp_state = get_ibqp_state(qp_state); + qp_attr->cur_qp_state = qp_attr->qp_state; qp_attr->sq_draining = (qp_state == OCRDMA_QPS_SQ_DRAINING) ? 1 : 0; qp_attr->max_dest_rd_atomic = params.max_ord_ird >> OCRDMA_QP_PARAMS_MAX_ORD_SHIFT; @@ -1475,19 +1580,18 @@ int ocrdma_query_qp(struct ib_qp *ibqp, params.max_ord_ird & OCRDMA_QP_PARAMS_MAX_IRD_MASK; qp_attr->en_sqd_async_notify = (params.max_sge_recv_flags & OCRDMA_QP_PARAMS_FLAGS_SQD_ASYNC) ? 1 : 0; + /* Sync driver QP state with FW */ + ocrdma_qp_state_change(qp, qp_attr->qp_state, NULL); mbx_err: return status; } -static void ocrdma_srq_toggle_bit(struct ocrdma_srq *srq, int idx) +static void ocrdma_srq_toggle_bit(struct ocrdma_srq *srq, unsigned int idx) { - int i = idx / 32; - unsigned int mask = (1 << (idx % 32)); + unsigned int i = idx / 32; + u32 mask = (1U << (idx % 32)); - if (srq->idx_bit_fields[i] & mask) - srq->idx_bit_fields[i] &= ~mask; - else - srq->idx_bit_fields[i] |= mask; + srq->idx_bit_fields[i] ^= mask; } static int ocrdma_hwq_free_cnt(struct ocrdma_qp_hwq_info *q) @@ -1596,7 +1700,7 @@ void ocrdma_del_flush_qp(struct ocrdma_qp *qp) { int found = false; unsigned long flags; - struct ocrdma_dev *dev = qp->dev; + struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); /* sync with any active CQ poll */ spin_lock_irqsave(&dev->flush_q_lock, flags); @@ -1613,7 +1717,6 @@ void ocrdma_del_flush_qp(struct ocrdma_qp *qp) int ocrdma_destroy_qp(struct ib_qp *ibqp) { - int status; struct ocrdma_pd *pd; struct ocrdma_qp *qp; struct ocrdma_dev *dev; @@ -1622,7 +1725,7 @@ int ocrdma_destroy_qp(struct ib_qp *ibqp) unsigned long flags; qp = get_ocrdma_qp(ibqp); - dev = qp->dev; + dev = get_ocrdma_dev(ibqp->device); attrs.qp_state = IB_QPS_ERR; pd = qp->pd; @@ -1635,7 +1738,7 @@ int ocrdma_destroy_qp(struct ib_qp *ibqp) * discarded until the old CQEs are discarded. */ mutex_lock(&dev->dev_lock); - status = ocrdma_mbx_destroy_qp(dev, qp); + (void) ocrdma_mbx_destroy_qp(dev, qp); /* * acquire CQ lock while destroy is in progress, in order to @@ -1670,7 +1773,7 @@ int ocrdma_destroy_qp(struct ib_qp *ibqp) kfree(qp->wqe_wr_id_tbl); kfree(qp->rqe_wr_id_tbl); kfree(qp); - return status; + return 0; } static int ocrdma_copy_srq_uresp(struct ocrdma_dev *dev, struct ocrdma_srq *srq, @@ -1831,6 +1934,8 @@ static void ocrdma_build_ud_hdr(struct ocrdma_qp *qp, else ud_hdr->qkey = wr->wr.ud.remote_qkey; ud_hdr->rsvd_ahid = ah->id; + if (ah->av->valid & OCRDMA_AV_VLAN_VALID) + hdr->cw |= (OCRDMA_FLAG_AH_VLAN_PR << OCRDMA_WQE_FLAGS_SHIFT); } static void ocrdma_build_sges(struct ocrdma_hdr_wqe *hdr, @@ -2007,11 +2112,12 @@ static int ocrdma_build_fr(struct ocrdma_qp *qp, struct ocrdma_hdr_wqe *hdr, u64 fbo; struct ocrdma_ewqe_fr *fast_reg = (struct ocrdma_ewqe_fr *)(hdr + 1); struct ocrdma_mr *mr; + struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); u32 wqe_size = sizeof(*fast_reg) + sizeof(*hdr); wqe_size = roundup(wqe_size, OCRDMA_WQE_ALIGN_BYTES); - if (wr->wr.fast_reg.page_list_len > qp->dev->attr.max_pages_per_frmr) + if (wr->wr.fast_reg.page_list_len > dev->attr.max_pages_per_frmr) return -EINVAL; hdr->cw |= (OCRDMA_FR_MR << OCRDMA_WQE_OPCODE_SHIFT); @@ -2039,7 +2145,7 @@ static int ocrdma_build_fr(struct ocrdma_qp *qp, struct ocrdma_hdr_wqe *hdr, fast_reg->size_sge = get_encoded_page_size(1 << wr->wr.fast_reg.page_shift); mr = (struct ocrdma_mr *) (unsigned long) - qp->dev->stag_arr[(hdr->lkey >> 8) & (OCRDMA_MAX_STAG - 1)]; + dev->stag_arr[(hdr->lkey >> 8) & (OCRDMA_MAX_STAG - 1)]; build_frmr_pbes(wr, mr->hwmr.pbl_table, &mr->hwmr); return 0; } @@ -2112,8 +2218,6 @@ int ocrdma_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, hdr->cw |= (OCRDMA_WRITE << OCRDMA_WQE_OPCODE_SHIFT); status = ocrdma_build_write(qp, hdr, wr); break; - case IB_WR_RDMA_READ_WITH_INV: - hdr->cw |= (OCRDMA_FLAG_INV << OCRDMA_WQE_FLAGS_SHIFT); case IB_WR_RDMA_READ: ocrdma_build_read(qp, hdr, wr); break; @@ -2484,8 +2588,11 @@ static bool ocrdma_poll_err_scqe(struct ocrdma_qp *qp, bool *polled, bool *stop) { bool expand; + struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); int status = (le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_STATUS_MASK) >> OCRDMA_CQE_STATUS_SHIFT; + if (status < OCRDMA_MAX_CQE_ERR) + atomic_inc(&dev->cqe_err_stats[status]); /* when hw sq is empty, but rq is not empty, so we continue * to keep the cqe in order to get the cq event again. @@ -2604,6 +2711,10 @@ static bool ocrdma_poll_err_rcqe(struct ocrdma_qp *qp, struct ocrdma_cqe *cqe, int status) { bool expand; + struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); + + if (status < OCRDMA_MAX_CQE_ERR) + atomic_inc(&dev->cqe_err_stats[status]); /* when hw_rq is empty, but wq is not empty, so continue * to keep the cqe to get the cq event again. diff --git a/drivers/infiniband/hw/qib/qib.h b/drivers/infiniband/hw/qib/qib.h index c00ae09..ffd48bf 100644 --- a/drivers/infiniband/hw/qib/qib.h +++ b/drivers/infiniband/hw/qib/qib.h @@ -1082,12 +1082,6 @@ struct qib_devdata { /* control high-level access to EEPROM */ struct mutex eep_lock; uint64_t traffic_wds; - /* active time is kept in seconds, but logged in hours */ - atomic_t active_time; - /* Below are nominal shadow of EEPROM, new since last EEPROM update */ - uint8_t eep_st_errs[QIB_EEP_LOG_CNT]; - uint8_t eep_st_new_errs[QIB_EEP_LOG_CNT]; - uint16_t eep_hrs; /* * masks for which bits of errs, hwerrs that cause * each of the counters to increment. @@ -1309,8 +1303,7 @@ int qib_twsi_blk_rd(struct qib_devdata *dd, int dev, int addr, void *buffer, int qib_twsi_blk_wr(struct qib_devdata *dd, int dev, int addr, const void *buffer, int len); void qib_get_eeprom_info(struct qib_devdata *); -int qib_update_eeprom_log(struct qib_devdata *dd); -void qib_inc_eeprom_err(struct qib_devdata *dd, u32 eidx, u32 incr); +#define qib_inc_eeprom_err(dd, eidx, incr) void qib_dump_lookup_output_queue(struct qib_devdata *); void qib_force_pio_avail_update(struct qib_devdata *); void qib_clear_symerror_on_linkup(unsigned long opaque); @@ -1467,11 +1460,14 @@ const char *qib_get_unit_name(int unit); * Flush write combining store buffers (if present) and perform a write * barrier. */ +static inline void qib_flush_wc(void) +{ #if defined(CONFIG_X86_64) -#define qib_flush_wc() asm volatile("sfence" : : : "memory") + asm volatile("sfence" : : : "memory"); #else -#define qib_flush_wc() wmb() /* no reorder around wc flush */ + wmb(); /* no reorder around wc flush */ #endif +} /* global module parameter variables */ extern unsigned qib_ibmtu; diff --git a/drivers/infiniband/hw/qib/qib_common.h b/drivers/infiniband/hw/qib/qib_common.h index 5670ace..4fb78ab 100644 --- a/drivers/infiniband/hw/qib/qib_common.h +++ b/drivers/infiniband/hw/qib/qib_common.h @@ -257,7 +257,7 @@ struct qib_base_info { /* shared memory page for send buffer disarm status */ __u64 spi_sendbuf_status; -} __attribute__ ((aligned(8))); +} __aligned(8); /* * This version number is given to the driver by the user code during @@ -361,7 +361,7 @@ struct qib_user_info { */ __u64 spu_base_info; -} __attribute__ ((aligned(8))); +} __aligned(8); /* User commands. */ diff --git a/drivers/infiniband/hw/qib/qib_debugfs.c b/drivers/infiniband/hw/qib/qib_debugfs.c index 6abd3ed..5e75b43 100644 --- a/drivers/infiniband/hw/qib/qib_debugfs.c +++ b/drivers/infiniband/hw/qib/qib_debugfs.c @@ -255,7 +255,6 @@ void qib_dbg_ibdev_init(struct qib_ibdev *ibd) DEBUGFS_FILE_CREATE(opcode_stats); DEBUGFS_FILE_CREATE(ctx_stats); DEBUGFS_FILE_CREATE(qp_stats); - return; } void qib_dbg_ibdev_exit(struct qib_ibdev *ibd) diff --git a/drivers/infiniband/hw/qib/qib_diag.c b/drivers/infiniband/hw/qib/qib_diag.c index 5dfda4c..8c34b23 100644 --- a/drivers/infiniband/hw/qib/qib_diag.c +++ b/drivers/infiniband/hw/qib/qib_diag.c @@ -85,7 +85,7 @@ static struct qib_diag_client *get_client(struct qib_devdata *dd) client_pool = dc->next; else /* None in pool, alloc and init */ - dc = kmalloc(sizeof *dc, GFP_KERNEL); + dc = kmalloc(sizeof(*dc), GFP_KERNEL); if (dc) { dc->next = NULL; @@ -257,6 +257,7 @@ static u32 __iomem *qib_remap_ioaddr32(struct qib_devdata *dd, u32 offset, if (dd->userbase) { /* If user regs mapped, they are after send, so set limit. */ u32 ulim = (dd->cfgctxts * dd->ureg_align) + dd->uregbase; + if (!dd->piovl15base) snd_lim = dd->uregbase; krb32 = (u32 __iomem *)dd->userbase; @@ -280,6 +281,7 @@ static u32 __iomem *qib_remap_ioaddr32(struct qib_devdata *dd, u32 offset, snd_bottom = dd->pio2k_bufbase; if (snd_lim == 0) { u32 tot2k = dd->piobcnt2k * ALIGN(dd->piosize2k, dd->palign); + snd_lim = snd_bottom + tot2k; } /* If 4k buffers exist, account for them by bumping @@ -398,6 +400,7 @@ static int qib_write_umem64(struct qib_devdata *dd, u32 regoffs, /* not very efficient, but it works for now */ while (reg_addr < reg_end) { u64 data; + if (copy_from_user(&data, uaddr, sizeof(data))) { ret = -EFAULT; goto bail; @@ -698,7 +701,7 @@ int qib_register_observer(struct qib_devdata *dd, if (!dd || !op) return -EINVAL; - olp = vmalloc(sizeof *olp); + olp = vmalloc(sizeof(*olp)); if (!olp) { pr_err("vmalloc for observer failed\n"); return -ENOMEM; @@ -796,6 +799,7 @@ static ssize_t qib_diag_read(struct file *fp, char __user *data, op = diag_get_observer(dd, *off); if (op) { u32 offset = *off; + ret = op->hook(dd, op, offset, &data64, 0, use_32); } /* @@ -873,6 +877,7 @@ static ssize_t qib_diag_write(struct file *fp, const char __user *data, if (count == 4 || count == 8) { u64 data64; u32 offset = *off; + ret = copy_from_user(&data64, data, count); if (ret) { ret = -EFAULT; diff --git a/drivers/infiniband/hw/qib/qib_driver.c b/drivers/infiniband/hw/qib/qib_driver.c index 5bee08f..f58fdc3 100644 --- a/drivers/infiniband/hw/qib/qib_driver.c +++ b/drivers/infiniband/hw/qib/qib_driver.c @@ -86,7 +86,7 @@ const char *qib_get_unit_name(int unit) { static char iname[16]; - snprintf(iname, sizeof iname, "infinipath%u", unit); + snprintf(iname, sizeof(iname), "infinipath%u", unit); return iname; } @@ -349,6 +349,7 @@ static u32 qib_rcv_hdrerr(struct qib_ctxtdata *rcd, struct qib_pportdata *ppd, qp_num = be32_to_cpu(ohdr->bth[1]) & QIB_QPN_MASK; if (qp_num != QIB_MULTICAST_QPN) { int ruc_res; + qp = qib_lookup_qpn(ibp, qp_num); if (!qp) goto drop; @@ -461,6 +462,7 @@ u32 qib_kreceive(struct qib_ctxtdata *rcd, u32 *llic, u32 *npkts) rhf_addr = (__le32 *) rcd->rcvhdrq + l + dd->rhf_offset; if (dd->flags & QIB_NODMA_RTAIL) { u32 seq = qib_hdrget_seq(rhf_addr); + if (seq != rcd->seq_cnt) goto bail; hdrqtail = 0; @@ -651,6 +653,7 @@ bail: int qib_set_lid(struct qib_pportdata *ppd, u32 lid, u8 lmc) { struct qib_devdata *dd = ppd->dd; + ppd->lid = lid; ppd->lmc = lmc; diff --git a/drivers/infiniband/hw/qib/qib_eeprom.c b/drivers/infiniband/hw/qib/qib_eeprom.c index 4d5d71a..311ee6c 100644 --- a/drivers/infiniband/hw/qib/qib_eeprom.c +++ b/drivers/infiniband/hw/qib/qib_eeprom.c @@ -153,6 +153,7 @@ void qib_get_eeprom_info(struct qib_devdata *dd) if (t && dd0->nguid > 1 && t <= dd0->nguid) { u8 oguid; + dd->base_guid = dd0->base_guid; bguid = (u8 *) &dd->base_guid; @@ -251,206 +252,25 @@ void qib_get_eeprom_info(struct qib_devdata *dd) * This board has a Serial-prefix, which is stored * elsewhere for backward-compatibility. */ - memcpy(snp, ifp->if_sprefix, sizeof ifp->if_sprefix); - snp[sizeof ifp->if_sprefix] = '\0'; + memcpy(snp, ifp->if_sprefix, sizeof(ifp->if_sprefix)); + snp[sizeof(ifp->if_sprefix)] = '\0'; len = strlen(snp); snp += len; - len = (sizeof dd->serial) - len; - if (len > sizeof ifp->if_serial) - len = sizeof ifp->if_serial; + len = sizeof(dd->serial) - len; + if (len > sizeof(ifp->if_serial)) + len = sizeof(ifp->if_serial); memcpy(snp, ifp->if_serial, len); - } else - memcpy(dd->serial, ifp->if_serial, - sizeof ifp->if_serial); + } else { + memcpy(dd->serial, ifp->if_serial, sizeof(ifp->if_serial)); + } if (!strstr(ifp->if_comment, "Tested successfully")) qib_dev_err(dd, "Board SN %s did not pass functional test: %s\n", dd->serial, ifp->if_comment); - memcpy(&dd->eep_st_errs, &ifp->if_errcntp, QIB_EEP_LOG_CNT); - /* - * Power-on (actually "active") hours are kept as little-endian value - * in EEPROM, but as seconds in a (possibly as small as 24-bit) - * atomic_t while running. - */ - atomic_set(&dd->active_time, 0); - dd->eep_hrs = ifp->if_powerhour[0] | (ifp->if_powerhour[1] << 8); - done: vfree(buf); bail:; } -/** - * qib_update_eeprom_log - copy active-time and error counters to eeprom - * @dd: the qlogic_ib device - * - * Although the time is kept as seconds in the qib_devdata struct, it is - * rounded to hours for re-write, as we have only 16 bits in EEPROM. - * First-cut code reads whole (expected) struct qib_flash, modifies, - * re-writes. Future direction: read/write only what we need, assuming - * that the EEPROM had to have been "good enough" for driver init, and - * if not, we aren't making it worse. - * - */ -int qib_update_eeprom_log(struct qib_devdata *dd) -{ - void *buf; - struct qib_flash *ifp; - int len, hi_water; - uint32_t new_time, new_hrs; - u8 csum; - int ret, idx; - unsigned long flags; - - /* first, check if we actually need to do anything. */ - ret = 0; - for (idx = 0; idx < QIB_EEP_LOG_CNT; ++idx) { - if (dd->eep_st_new_errs[idx]) { - ret = 1; - break; - } - } - new_time = atomic_read(&dd->active_time); - - if (ret == 0 && new_time < 3600) - goto bail; - - /* - * The quick-check above determined that there is something worthy - * of logging, so get current contents and do a more detailed idea. - * read full flash, not just currently used part, since it may have - * been written with a newer definition - */ - len = sizeof(struct qib_flash); - buf = vmalloc(len); - ret = 1; - if (!buf) { - qib_dev_err(dd, - "Couldn't allocate memory to read %u bytes from eeprom for logging\n", - len); - goto bail; - } - - /* Grab semaphore and read current EEPROM. If we get an - * error, let go, but if not, keep it until we finish write. - */ - ret = mutex_lock_interruptible(&dd->eep_lock); - if (ret) { - qib_dev_err(dd, "Unable to acquire EEPROM for logging\n"); - goto free_bail; - } - ret = qib_twsi_blk_rd(dd, dd->twsi_eeprom_dev, 0, buf, len); - if (ret) { - mutex_unlock(&dd->eep_lock); - qib_dev_err(dd, "Unable read EEPROM for logging\n"); - goto free_bail; - } - ifp = (struct qib_flash *)buf; - - csum = flash_csum(ifp, 0); - if (csum != ifp->if_csum) { - mutex_unlock(&dd->eep_lock); - qib_dev_err(dd, "EEPROM cks err (0x%02X, S/B 0x%02X)\n", - csum, ifp->if_csum); - ret = 1; - goto free_bail; - } - hi_water = 0; - spin_lock_irqsave(&dd->eep_st_lock, flags); - for (idx = 0; idx < QIB_EEP_LOG_CNT; ++idx) { - int new_val = dd->eep_st_new_errs[idx]; - if (new_val) { - /* - * If we have seen any errors, add to EEPROM values - * We need to saturate at 0xFF (255) and we also - * would need to adjust the checksum if we were - * trying to minimize EEPROM traffic - * Note that we add to actual current count in EEPROM, - * in case it was altered while we were running. - */ - new_val += ifp->if_errcntp[idx]; - if (new_val > 0xFF) - new_val = 0xFF; - if (ifp->if_errcntp[idx] != new_val) { - ifp->if_errcntp[idx] = new_val; - hi_water = offsetof(struct qib_flash, - if_errcntp) + idx; - } - /* - * update our shadow (used to minimize EEPROM - * traffic), to match what we are about to write. - */ - dd->eep_st_errs[idx] = new_val; - dd->eep_st_new_errs[idx] = 0; - } - } - /* - * Now update active-time. We would like to round to the nearest hour - * but unless atomic_t are sure to be proper signed ints we cannot, - * because we need to account for what we "transfer" to EEPROM and - * if we log an hour at 31 minutes, then we would need to set - * active_time to -29 to accurately count the _next_ hour. - */ - if (new_time >= 3600) { - new_hrs = new_time / 3600; - atomic_sub((new_hrs * 3600), &dd->active_time); - new_hrs += dd->eep_hrs; - if (new_hrs > 0xFFFF) - new_hrs = 0xFFFF; - dd->eep_hrs = new_hrs; - if ((new_hrs & 0xFF) != ifp->if_powerhour[0]) { - ifp->if_powerhour[0] = new_hrs & 0xFF; - hi_water = offsetof(struct qib_flash, if_powerhour); - } - if ((new_hrs >> 8) != ifp->if_powerhour[1]) { - ifp->if_powerhour[1] = new_hrs >> 8; - hi_water = offsetof(struct qib_flash, if_powerhour) + 1; - } - } - /* - * There is a tiny possibility that we could somehow fail to write - * the EEPROM after updating our shadows, but problems from holding - * the spinlock too long are a much bigger issue. - */ - spin_unlock_irqrestore(&dd->eep_st_lock, flags); - if (hi_water) { - /* we made some change to the data, uopdate cksum and write */ - csum = flash_csum(ifp, 1); - ret = eeprom_write_with_enable(dd, 0, buf, hi_water + 1); - } - mutex_unlock(&dd->eep_lock); - if (ret) - qib_dev_err(dd, "Failed updating EEPROM\n"); - -free_bail: - vfree(buf); -bail: - return ret; -} - -/** - * qib_inc_eeprom_err - increment one of the four error counters - * that are logged to EEPROM. - * @dd: the qlogic_ib device - * @eidx: 0..3, the counter to increment - * @incr: how much to add - * - * Each counter is 8-bits, and saturates at 255 (0xFF). They - * are copied to the EEPROM (aka flash) whenever qib_update_eeprom_log() - * is called, but it can only be called in a context that allows sleep. - * This function can be called even at interrupt level. - */ -void qib_inc_eeprom_err(struct qib_devdata *dd, u32 eidx, u32 incr) -{ - uint new_val; - unsigned long flags; - - spin_lock_irqsave(&dd->eep_st_lock, flags); - new_val = dd->eep_st_new_errs[eidx] + incr; - if (new_val > 255) - new_val = 255; - dd->eep_st_new_errs[eidx] = new_val; - spin_unlock_irqrestore(&dd->eep_st_lock, flags); -} diff --git a/drivers/infiniband/hw/qib/qib_file_ops.c b/drivers/infiniband/hw/qib/qib_file_ops.c index b15e34e..41937c6 100644 --- a/drivers/infiniband/hw/qib/qib_file_ops.c +++ b/drivers/infiniband/hw/qib/qib_file_ops.c @@ -351,9 +351,10 @@ static int qib_tid_update(struct qib_ctxtdata *rcd, struct file *fp, * unless perhaps the user has mpin'ed the pages * themselves. */ - qib_devinfo(dd->pcidev, - "Failed to lock addr %p, %u pages: " - "errno %d\n", (void *) vaddr, cnt, -ret); + qib_devinfo( + dd->pcidev, + "Failed to lock addr %p, %u pages: errno %d\n", + (void *) vaddr, cnt, -ret); goto done; } for (i = 0; i < cnt; i++, vaddr += PAGE_SIZE) { @@ -437,7 +438,7 @@ cleanup: goto cleanup; } if (copy_to_user((void __user *) (unsigned long) ti->tidmap, - tidmap, sizeof tidmap)) { + tidmap, sizeof(tidmap))) { ret = -EFAULT; goto cleanup; } @@ -484,7 +485,7 @@ static int qib_tid_free(struct qib_ctxtdata *rcd, unsigned subctxt, } if (copy_from_user(tidmap, (void __user *)(unsigned long)ti->tidmap, - sizeof tidmap)) { + sizeof(tidmap))) { ret = -EFAULT; goto done; } @@ -951,8 +952,8 @@ static int mmap_kvaddr(struct vm_area_struct *vma, u64 pgaddr, /* rcvegrbufs are read-only on the slave */ if (vma->vm_flags & VM_WRITE) { qib_devinfo(dd->pcidev, - "Can't map eager buffers as " - "writable (flags=%lx)\n", vma->vm_flags); + "Can't map eager buffers as writable (flags=%lx)\n", + vma->vm_flags); ret = -EPERM; goto bail; } @@ -1185,6 +1186,7 @@ static void assign_ctxt_affinity(struct file *fp, struct qib_devdata *dd) */ if (weight >= qib_cpulist_count) { int cpu; + cpu = find_first_zero_bit(qib_cpulist, qib_cpulist_count); if (cpu == qib_cpulist_count) @@ -1247,10 +1249,7 @@ static int init_subctxts(struct qib_devdata *dd, if (!qib_compatible_subctxts(uinfo->spu_userversion >> 16, uinfo->spu_userversion & 0xffff)) { qib_devinfo(dd->pcidev, - "Mismatched user version (%d.%d) and driver " - "version (%d.%d) while context sharing. Ensure " - "that driver and library are from the same " - "release.\n", + "Mismatched user version (%d.%d) and driver version (%d.%d) while context sharing. Ensure that driver and library are from the same release.\n", (int) (uinfo->spu_userversion >> 16), (int) (uinfo->spu_userversion & 0xffff), QIB_USER_SWMAJOR, QIB_USER_SWMINOR); @@ -1391,6 +1390,7 @@ static int choose_port_ctxt(struct file *fp, struct qib_devdata *dd, u32 port, } if (!ppd) { u32 pidx = ctxt % dd->num_pports; + if (usable(dd->pport + pidx)) ppd = dd->pport + pidx; else { @@ -1438,10 +1438,12 @@ static int get_a_ctxt(struct file *fp, const struct qib_user_info *uinfo, if (alg == QIB_PORT_ALG_ACROSS) { unsigned inuse = ~0U; + /* find device (with ACTIVE ports) with fewest ctxts in use */ for (ndev = 0; ndev < devmax; ndev++) { struct qib_devdata *dd = qib_lookup(ndev); unsigned cused = 0, cfree = 0, pusable = 0; + if (!dd) continue; if (port && port <= dd->num_pports && @@ -1471,6 +1473,7 @@ static int get_a_ctxt(struct file *fp, const struct qib_user_info *uinfo, } else { for (ndev = 0; ndev < devmax; ndev++) { struct qib_devdata *dd = qib_lookup(ndev); + if (dd) { ret = choose_port_ctxt(fp, dd, port, uinfo); if (!ret) @@ -1556,6 +1559,7 @@ static int find_hca(unsigned int cpu, int *unit) } for (ndev = 0; ndev < devmax; ndev++) { struct qib_devdata *dd = qib_lookup(ndev); + if (dd) { if (pcibus_to_node(dd->pcidev->bus) < 0) { ret = -EINVAL; diff --git a/drivers/infiniband/hw/qib/qib_fs.c b/drivers/infiniband/hw/qib/qib_fs.c index 8185458..55f240a 100644 --- a/drivers/infiniband/hw/qib/qib_fs.c +++ b/drivers/infiniband/hw/qib/qib_fs.c @@ -106,7 +106,7 @@ static ssize_t driver_stats_read(struct file *file, char __user *buf, { qib_stats.sps_ints = qib_sps_ints(); return simple_read_from_buffer(buf, count, ppos, &qib_stats, - sizeof qib_stats); + sizeof(qib_stats)); } /* @@ -133,7 +133,7 @@ static ssize_t driver_names_read(struct file *file, char __user *buf, size_t count, loff_t *ppos) { return simple_read_from_buffer(buf, count, ppos, qib_statnames, - sizeof qib_statnames - 1); /* no null */ + sizeof(qib_statnames) - 1); /* no null */ } static const struct file_operations driver_ops[] = { @@ -379,7 +379,7 @@ static int add_cntr_files(struct super_block *sb, struct qib_devdata *dd) int ret, i; /* create the per-unit directory */ - snprintf(unit, sizeof unit, "%u", dd->unit); + snprintf(unit, sizeof(unit), "%u", dd->unit); ret = create_file(unit, S_IFDIR|S_IRUGO|S_IXUGO, sb->s_root, &dir, &simple_dir_operations, dd); if (ret) { @@ -482,7 +482,7 @@ static int remove_device_files(struct super_block *sb, root = dget(sb->s_root); mutex_lock(&root->d_inode->i_mutex); - snprintf(unit, sizeof unit, "%u", dd->unit); + snprintf(unit, sizeof(unit), "%u", dd->unit); dir = lookup_one_len(unit, root, strlen(unit)); if (IS_ERR(dir)) { @@ -560,6 +560,7 @@ static struct dentry *qibfs_mount(struct file_system_type *fs_type, int flags, const char *dev_name, void *data) { struct dentry *ret; + ret = mount_single(fs_type, flags, data, qibfs_fill_super); if (!IS_ERR(ret)) qib_super = ret->d_sb; diff --git a/drivers/infiniband/hw/qib/qib_iba6120.c b/drivers/infiniband/hw/qib/qib_iba6120.c index d68266a..0d2ba59 100644 --- a/drivers/infiniband/hw/qib/qib_iba6120.c +++ b/drivers/infiniband/hw/qib/qib_iba6120.c @@ -333,6 +333,7 @@ static inline void qib_write_ureg(const struct qib_devdata *dd, enum qib_ureg regno, u64 value, int ctxt) { u64 __iomem *ubase; + if (dd->userbase) ubase = (u64 __iomem *) ((char __iomem *) dd->userbase + @@ -834,14 +835,14 @@ static void qib_handle_6120_hwerrors(struct qib_devdata *dd, char *msg, bits = (u32) ((hwerrs >> QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT) & QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK); - snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf, + snprintf(bitsmsg, sizeof(dd->cspec->bitsmsgbuf), "[PCIe Mem Parity Errs %x] ", bits); strlcat(msg, bitsmsg, msgl); } if (hwerrs & _QIB_PLL_FAIL) { isfatal = 1; - snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf, + snprintf(bitsmsg, sizeof(dd->cspec->bitsmsgbuf), "[PLL failed (%llx), InfiniPath hardware unusable]", (unsigned long long) hwerrs & _QIB_PLL_FAIL); strlcat(msg, bitsmsg, msgl); @@ -1014,7 +1015,7 @@ static void handle_6120_errors(struct qib_devdata *dd, u64 errs) /* do these first, they are most important */ if (errs & ERR_MASK(HardwareErr)) - qib_handle_6120_hwerrors(dd, msg, sizeof dd->cspec->emsgbuf); + qib_handle_6120_hwerrors(dd, msg, sizeof(dd->cspec->emsgbuf)); else for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx) if (errs & dd->eep_st_masks[log_idx].errs_to_log) @@ -1062,7 +1063,7 @@ static void handle_6120_errors(struct qib_devdata *dd, u64 errs) */ mask = ERR_MASK(IBStatusChanged) | ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr) | ERR_MASK(HardwareErr); - qib_decode_6120_err(dd, msg, sizeof dd->cspec->emsgbuf, errs & ~mask); + qib_decode_6120_err(dd, msg, sizeof(dd->cspec->emsgbuf), errs & ~mask); if (errs & E_SUM_PKTERRS) qib_stats.sps_rcverrs++; @@ -1670,6 +1671,7 @@ static irqreturn_t qib_6120intr(int irq, void *data) } if (crcs) { u32 cntr = dd->cspec->lli_counter; + cntr += crcs; if (cntr) { if (cntr > dd->cspec->lli_thresh) { @@ -1722,6 +1724,7 @@ static void qib_setup_6120_interrupt(struct qib_devdata *dd) "irq is 0, BIOS error? Interrupts won't work\n"); else { int ret; + ret = request_irq(dd->cspec->irq, qib_6120intr, 0, QIB_DRV_NAME, dd); if (ret) @@ -2681,8 +2684,6 @@ static void qib_get_6120_faststats(unsigned long opaque) spin_lock_irqsave(&dd->eep_st_lock, flags); traffic_wds -= dd->traffic_wds; dd->traffic_wds += traffic_wds; - if (traffic_wds >= QIB_TRAFFIC_ACTIVE_THRESHOLD) - atomic_add(5, &dd->active_time); /* S/B #define */ spin_unlock_irqrestore(&dd->eep_st_lock, flags); qib_chk_6120_errormask(dd); @@ -2929,6 +2930,7 @@ bail: static int qib_6120_set_loopback(struct qib_pportdata *ppd, const char *what) { int ret = 0; + if (!strncmp(what, "ibc", 3)) { ppd->dd->cspec->ibcctrl |= SYM_MASK(IBCCtrl, Loopback); qib_devinfo(ppd->dd->pcidev, "Enabling IB%u:%u IBC loopback\n", @@ -3170,6 +3172,7 @@ static void get_6120_chip_params(struct qib_devdata *dd) static void set_6120_baseaddrs(struct qib_devdata *dd) { u32 cregbase; + cregbase = qib_read_kreg32(dd, kr_counterregbase); dd->cspec->cregbase = (u64 __iomem *) ((char __iomem *) dd->kregbase + cregbase); diff --git a/drivers/infiniband/hw/qib/qib_iba7220.c b/drivers/infiniband/hw/qib/qib_iba7220.c index 7dec89f..22affda 100644 --- a/drivers/infiniband/hw/qib/qib_iba7220.c +++ b/drivers/infiniband/hw/qib/qib_iba7220.c @@ -902,7 +902,8 @@ static void sdma_7220_errors(struct qib_pportdata *ppd, u64 errs) errs &= QLOGIC_IB_E_SDMAERRS; msg = dd->cspec->sdmamsgbuf; - qib_decode_7220_sdma_errs(ppd, errs, msg, sizeof dd->cspec->sdmamsgbuf); + qib_decode_7220_sdma_errs(ppd, errs, msg, + sizeof(dd->cspec->sdmamsgbuf)); spin_lock_irqsave(&ppd->sdma_lock, flags); if (errs & ERR_MASK(SendBufMisuseErr)) { @@ -1043,6 +1044,7 @@ done: static void reenable_7220_chase(unsigned long opaque) { struct qib_pportdata *ppd = (struct qib_pportdata *)opaque; + ppd->cpspec->chase_timer.expires = 0; qib_set_ib_7220_lstate(ppd, QLOGIC_IB_IBCC_LINKCMD_DOWN, QLOGIC_IB_IBCC_LINKINITCMD_POLL); @@ -1101,7 +1103,7 @@ static void handle_7220_errors(struct qib_devdata *dd, u64 errs) /* do these first, they are most important */ if (errs & ERR_MASK(HardwareErr)) - qib_7220_handle_hwerrors(dd, msg, sizeof dd->cspec->emsgbuf); + qib_7220_handle_hwerrors(dd, msg, sizeof(dd->cspec->emsgbuf)); else for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx) if (errs & dd->eep_st_masks[log_idx].errs_to_log) @@ -1155,7 +1157,7 @@ static void handle_7220_errors(struct qib_devdata *dd, u64 errs) ERR_MASK(RcvEgrFullErr) | ERR_MASK(RcvHdrFullErr) | ERR_MASK(HardwareErr) | ERR_MASK(SDmaDisabledErr); - qib_decode_7220_err(dd, msg, sizeof dd->cspec->emsgbuf, errs & ~mask); + qib_decode_7220_err(dd, msg, sizeof(dd->cspec->emsgbuf), errs & ~mask); if (errs & E_SUM_PKTERRS) qib_stats.sps_rcverrs++; @@ -1380,7 +1382,7 @@ static void qib_7220_handle_hwerrors(struct qib_devdata *dd, char *msg, bits = (u32) ((hwerrs >> QLOGIC_IB_HWE_PCIEMEMPARITYERR_SHIFT) & QLOGIC_IB_HWE_PCIEMEMPARITYERR_MASK); - snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf, + snprintf(bitsmsg, sizeof(dd->cspec->bitsmsgbuf), "[PCIe Mem Parity Errs %x] ", bits); strlcat(msg, bitsmsg, msgl); } @@ -1390,7 +1392,7 @@ static void qib_7220_handle_hwerrors(struct qib_devdata *dd, char *msg, if (hwerrs & _QIB_PLL_FAIL) { isfatal = 1; - snprintf(bitsmsg, sizeof dd->cspec->bitsmsgbuf, + snprintf(bitsmsg, sizeof(dd->cspec->bitsmsgbuf), "[PLL failed (%llx), InfiniPath hardware unusable]", (unsigned long long) hwerrs & _QIB_PLL_FAIL); strlcat(msg, bitsmsg, msgl); @@ -3297,8 +3299,6 @@ static void qib_get_7220_faststats(unsigned long opaque) spin_lock_irqsave(&dd->eep_st_lock, flags); traffic_wds -= dd->traffic_wds; dd->traffic_wds += traffic_wds; - if (traffic_wds >= QIB_TRAFFIC_ACTIVE_THRESHOLD) - atomic_add(5, &dd->active_time); /* S/B #define */ spin_unlock_irqrestore(&dd->eep_st_lock, flags); done: mod_timer(&dd->stats_timer, jiffies + HZ * ACTIVITY_TIMER); diff --git a/drivers/infiniband/hw/qib/qib_iba7322.c b/drivers/infiniband/hw/qib/qib_iba7322.c index a7eb325..ef97b71 100644 --- a/drivers/infiniband/hw/qib/qib_iba7322.c +++ b/drivers/infiniband/hw/qib/qib_iba7322.c @@ -117,7 +117,7 @@ MODULE_PARM_DESC(chase, "Enable state chase handling"); static ushort qib_long_atten = 10; /* 10 dB ~= 5m length */ module_param_named(long_attenuation, qib_long_atten, ushort, S_IRUGO); -MODULE_PARM_DESC(long_attenuation, \ +MODULE_PARM_DESC(long_attenuation, "attenuation cutoff (dB) for long copper cable setup"); static ushort qib_singleport; @@ -153,11 +153,12 @@ static struct kparam_string kp_txselect = { static int setup_txselect(const char *, struct kernel_param *); module_param_call(txselect, setup_txselect, param_get_string, &kp_txselect, S_IWUSR | S_IRUGO); -MODULE_PARM_DESC(txselect, \ +MODULE_PARM_DESC(txselect, "Tx serdes indices (for no QSFP or invalid QSFP data)"); #define BOARD_QME7342 5 #define BOARD_QMH7342 6 +#define BOARD_QMH7360 9 #define IS_QMH(dd) (SYM_FIELD((dd)->revision, Revision, BoardID) == \ BOARD_QMH7342) #define IS_QME(dd) (SYM_FIELD((dd)->revision, Revision, BoardID) == \ @@ -817,6 +818,7 @@ static inline void qib_write_ureg(const struct qib_devdata *dd, enum qib_ureg regno, u64 value, int ctxt) { u64 __iomem *ubase; + if (dd->userbase) ubase = (u64 __iomem *) ((char __iomem *) dd->userbase + @@ -1677,7 +1679,7 @@ static noinline void handle_7322_errors(struct qib_devdata *dd) /* do these first, they are most important */ if (errs & QIB_E_HARDWARE) { *msg = '\0'; - qib_7322_handle_hwerrors(dd, msg, sizeof dd->cspec->emsgbuf); + qib_7322_handle_hwerrors(dd, msg, sizeof(dd->cspec->emsgbuf)); } else for (log_idx = 0; log_idx < QIB_EEP_LOG_CNT; ++log_idx) if (errs & dd->eep_st_masks[log_idx].errs_to_log) @@ -1702,7 +1704,7 @@ static noinline void handle_7322_errors(struct qib_devdata *dd) mask = QIB_E_HARDWARE; *msg = '\0'; - err_decode(msg, sizeof dd->cspec->emsgbuf, errs & ~mask, + err_decode(msg, sizeof(dd->cspec->emsgbuf), errs & ~mask, qib_7322error_msgs); /* @@ -1889,10 +1891,10 @@ static noinline void handle_7322_p_errors(struct qib_pportdata *ppd) *msg = '\0'; if (errs & ~QIB_E_P_BITSEXTANT) { - err_decode(msg, sizeof ppd->cpspec->epmsgbuf, + err_decode(msg, sizeof(ppd->cpspec->epmsgbuf), errs & ~QIB_E_P_BITSEXTANT, qib_7322p_error_msgs); if (!*msg) - snprintf(msg, sizeof ppd->cpspec->epmsgbuf, + snprintf(msg, sizeof(ppd->cpspec->epmsgbuf), "no others"); qib_dev_porterr(dd, ppd->port, "error interrupt with unknown errors 0x%016Lx set (and %s)\n", @@ -1906,7 +1908,7 @@ static noinline void handle_7322_p_errors(struct qib_pportdata *ppd) /* determine cause, then write to clear */ symptom = qib_read_kreg_port(ppd, krp_sendhdrsymptom); qib_write_kreg_port(ppd, krp_sendhdrsymptom, 0); - err_decode(msg, sizeof ppd->cpspec->epmsgbuf, symptom, + err_decode(msg, sizeof(ppd->cpspec->epmsgbuf), symptom, hdrchk_msgs); *msg = '\0'; /* senderrbuf cleared in SPKTERRS below */ @@ -1922,7 +1924,7 @@ static noinline void handle_7322_p_errors(struct qib_pportdata *ppd) * isn't valid. We don't want to confuse people, so * we just don't print them, except at debug */ - err_decode(msg, sizeof ppd->cpspec->epmsgbuf, + err_decode(msg, sizeof(ppd->cpspec->epmsgbuf), (errs & QIB_E_P_LINK_PKTERRS), qib_7322p_error_msgs); *msg = '\0'; @@ -1938,7 +1940,7 @@ static noinline void handle_7322_p_errors(struct qib_pportdata *ppd) * valid. We don't want to confuse people, so we just * don't print them, except at debug */ - err_decode(msg, sizeof ppd->cpspec->epmsgbuf, errs, + err_decode(msg, sizeof(ppd->cpspec->epmsgbuf), errs, qib_7322p_error_msgs); ignore_this_time = errs & QIB_E_P_LINK_PKTERRS; *msg = '\0'; @@ -2031,6 +2033,7 @@ static void qib_7322_set_intr_state(struct qib_devdata *dd, u32 enable) if (dd->cspec->num_msix_entries) { /* and same for MSIx */ u64 val = qib_read_kreg64(dd, kr_intgranted); + if (val) qib_write_kreg(dd, kr_intgranted, val); } @@ -2176,6 +2179,7 @@ static void qib_7322_handle_hwerrors(struct qib_devdata *dd, char *msg, int err; unsigned long flags; struct qib_pportdata *ppd = dd->pport; + for (; pidx < dd->num_pports; ++pidx, ppd++) { err = 0; if (pidx == 0 && (hwerrs & @@ -2801,9 +2805,11 @@ static void qib_irq_notifier_notify(struct irq_affinity_notify *notify, if (n->rcv) { struct qib_ctxtdata *rcd = (struct qib_ctxtdata *)n->arg; + qib_update_rhdrq_dca(rcd, cpu); } else { struct qib_pportdata *ppd = (struct qib_pportdata *)n->arg; + qib_update_sdma_dca(ppd, cpu); } } @@ -2816,9 +2822,11 @@ static void qib_irq_notifier_release(struct kref *ref) if (n->rcv) { struct qib_ctxtdata *rcd = (struct qib_ctxtdata *)n->arg; + dd = rcd->dd; } else { struct qib_pportdata *ppd = (struct qib_pportdata *)n->arg; + dd = ppd->dd; } qib_devinfo(dd->pcidev, @@ -2994,6 +3002,7 @@ static noinline void unknown_7322_gpio_intr(struct qib_devdata *dd) struct qib_pportdata *ppd; struct qib_qsfp_data *qd; u32 mask; + if (!dd->pport[pidx].link_speed_supported) continue; mask = QSFP_GPIO_MOD_PRS_N; @@ -3001,6 +3010,7 @@ static noinline void unknown_7322_gpio_intr(struct qib_devdata *dd) mask <<= (QSFP_GPIO_PORT2_SHIFT * ppd->hw_pidx); if (gpiostatus & dd->cspec->gpio_mask & mask) { u64 pins; + qd = &ppd->cpspec->qsfp_data; gpiostatus &= ~mask; pins = qib_read_kreg64(dd, kr_extstatus); @@ -3442,7 +3452,7 @@ try_intx: } /* Try to get MSIx interrupts */ - memset(redirect, 0, sizeof redirect); + memset(redirect, 0, sizeof(redirect)); mask = ~0ULL; msixnum = 0; local_mask = cpumask_of_pcibus(dd->pcidev->bus); @@ -3617,6 +3627,10 @@ static unsigned qib_7322_boardname(struct qib_devdata *dd) n = "InfiniPath_QME7362"; dd->flags |= QIB_HAS_QSFP; break; + case BOARD_QMH7360: + n = "Intel IB QDR 1P FLR-QSFP Adptr"; + dd->flags |= QIB_HAS_QSFP; + break; case 15: n = "InfiniPath_QLE7342_TEST"; dd->flags |= QIB_HAS_QSFP; @@ -3694,6 +3708,7 @@ static int qib_do_7322_reset(struct qib_devdata *dd) */ for (i = 0; i < msix_entries; i++) { u64 vecaddr, vecdata; + vecaddr = qib_read_kreg64(dd, 2 * i + (QIB_7322_MsixTable_OFFS / sizeof(u64))); vecdata = qib_read_kreg64(dd, 1 + 2 * i + @@ -5178,8 +5193,6 @@ static void qib_get_7322_faststats(unsigned long opaque) spin_lock_irqsave(&ppd->dd->eep_st_lock, flags); traffic_wds -= ppd->dd->traffic_wds; ppd->dd->traffic_wds += traffic_wds; - if (traffic_wds >= QIB_TRAFFIC_ACTIVE_THRESHOLD) - atomic_add(ACTIVITY_TIMER, &ppd->dd->active_time); spin_unlock_irqrestore(&ppd->dd->eep_st_lock, flags); if (ppd->cpspec->qdr_dfe_on && (ppd->link_speed_active & QIB_IB_QDR) && @@ -5357,6 +5370,7 @@ static void qib_autoneg_7322_send(struct qib_pportdata *ppd, int which) static void set_7322_ibspeed_fast(struct qib_pportdata *ppd, u32 speed) { u64 newctrlb; + newctrlb = ppd->cpspec->ibcctrl_b & ~(IBA7322_IBC_SPEED_MASK | IBA7322_IBC_IBTA_1_2_MASK | IBA7322_IBC_MAX_SPEED_MASK); @@ -5843,6 +5857,7 @@ static void get_7322_chip_params(struct qib_devdata *dd) static void qib_7322_set_baseaddrs(struct qib_devdata *dd) { u32 cregbase; + cregbase = qib_read_kreg32(dd, kr_counterregbase); dd->cspec->cregbase = (u64 __iomem *)(cregbase + @@ -6183,6 +6198,7 @@ static int setup_txselect(const char *str, struct kernel_param *kp) struct qib_devdata *dd; unsigned long val; char *n; + if (strlen(str) >= MAX_ATTEN_LEN) { pr_info("txselect_values string too long\n"); return -ENOSPC; @@ -6393,6 +6409,7 @@ static void write_7322_initregs(struct qib_devdata *dd) val = TIDFLOW_ERRBITS; /* these are W1C */ for (i = 0; i < dd->cfgctxts; i++) { int flow; + for (flow = 0; flow < NUM_TIDFLOWS_CTXT; flow++) qib_write_ureg(dd, ur_rcvflowtable+flow, val, i); } @@ -6503,6 +6520,7 @@ static int qib_init_7322_variables(struct qib_devdata *dd) for (pidx = 0; pidx < NUM_IB_PORTS; ++pidx) { struct qib_chippport_specific *cp = ppd->cpspec; + ppd->link_speed_supported = features & PORT_SPD_CAP; features >>= PORT_SPD_CAP_SHIFT; if (!ppd->link_speed_supported) { @@ -6581,8 +6599,7 @@ static int qib_init_7322_variables(struct qib_devdata *dd) ppd->vls_supported = IB_VL_VL0_7; else { qib_devinfo(dd->pcidev, - "Invalid num_vls %u for MTU %d " - ", using 4 VLs\n", + "Invalid num_vls %u for MTU %d , using 4 VLs\n", qib_num_cfg_vls, mtu); ppd->vls_supported = IB_VL_VL0_3; qib_num_cfg_vls = 4; @@ -7890,6 +7907,7 @@ static void serdes_7322_los_enable(struct qib_pportdata *ppd, int enable) static int serdes_7322_init(struct qib_pportdata *ppd) { int ret = 0; + if (ppd->dd->cspec->r1) ret = serdes_7322_init_old(ppd); else @@ -8305,8 +8323,8 @@ static void force_h1(struct qib_pportdata *ppd) static int qib_r_grab(struct qib_devdata *dd) { - u64 val; - val = SJA_EN; + u64 val = SJA_EN; + qib_write_kreg(dd, kr_r_access, val); qib_read_kreg32(dd, kr_scratch); return 0; @@ -8319,6 +8337,7 @@ static int qib_r_wait_for_rdy(struct qib_devdata *dd) { u64 val; int timeout; + for (timeout = 0; timeout < 100 ; ++timeout) { val = qib_read_kreg32(dd, kr_r_access); if (val & R_RDY) @@ -8346,6 +8365,7 @@ static int qib_r_shift(struct qib_devdata *dd, int bisten, } if (inp) { int tdi = inp[pos >> 3] >> (pos & 7); + val |= ((tdi & 1) << R_TDI_LSB); } qib_write_kreg(dd, kr_r_access, val); diff --git a/drivers/infiniband/hw/qib/qib_init.c b/drivers/infiniband/hw/qib/qib_init.c index 729da39c..2ee3695 100644 --- a/drivers/infiniband/hw/qib/qib_init.c +++ b/drivers/infiniband/hw/qib/qib_init.c @@ -140,7 +140,7 @@ int qib_create_ctxts(struct qib_devdata *dd) * Allocate full ctxtcnt array, rather than just cfgctxts, because * cleanup iterates across all possible ctxts. */ - dd->rcd = kzalloc(sizeof(*dd->rcd) * dd->ctxtcnt, GFP_KERNEL); + dd->rcd = kcalloc(dd->ctxtcnt, sizeof(*dd->rcd), GFP_KERNEL); if (!dd->rcd) { qib_dev_err(dd, "Unable to allocate ctxtdata array, failing\n"); @@ -234,6 +234,7 @@ int qib_init_pportdata(struct qib_pportdata *ppd, struct qib_devdata *dd, u8 hw_pidx, u8 port) { int size; + ppd->dd = dd; ppd->hw_pidx = hw_pidx; ppd->port = port; /* IB port number, not index */ @@ -613,6 +614,7 @@ static int qib_create_workqueues(struct qib_devdata *dd) ppd = dd->pport + pidx; if (!ppd->qib_wq) { char wq_name[8]; /* 3 + 2 + 1 + 1 + 1 */ + snprintf(wq_name, sizeof(wq_name), "qib%d_%d", dd->unit, pidx); ppd->qib_wq = @@ -714,6 +716,7 @@ int qib_init(struct qib_devdata *dd, int reinit) for (pidx = 0; pidx < dd->num_pports; ++pidx) { int mtu; + if (lastfail) ret = lastfail; ppd = dd->pport + pidx; @@ -931,7 +934,6 @@ static void qib_shutdown_device(struct qib_devdata *dd) qib_free_pportdata(ppd); } - qib_update_eeprom_log(dd); } /** @@ -1026,8 +1028,7 @@ static void qib_verify_pioperf(struct qib_devdata *dd) addr = vmalloc(cnt); if (!addr) { qib_devinfo(dd->pcidev, - "Couldn't get memory for checking PIO perf," - " skipping\n"); + "Couldn't get memory for checking PIO perf, skipping\n"); goto done; } @@ -1163,6 +1164,7 @@ struct qib_devdata *qib_alloc_devdata(struct pci_dev *pdev, size_t extra) if (!qib_cpulist_count) { u32 count = num_online_cpus(); + qib_cpulist = kzalloc(BITS_TO_LONGS(count) * sizeof(long), GFP_KERNEL); if (qib_cpulist) @@ -1179,7 +1181,7 @@ bail: if (!list_empty(&dd->list)) list_del_init(&dd->list); ib_dealloc_device(&dd->verbs_dev.ibdev); - return ERR_PTR(ret);; + return ERR_PTR(ret); } /* diff --git a/drivers/infiniband/hw/qib/qib_intr.c b/drivers/infiniband/hw/qib/qib_intr.c index f4918f2..086616d 100644 --- a/drivers/infiniband/hw/qib/qib_intr.c +++ b/drivers/infiniband/hw/qib/qib_intr.c @@ -168,7 +168,6 @@ skip_ibchange: ppd->lastibcstat = ibcs; if (ev) signal_ib_event(ppd, ev); - return; } void qib_clear_symerror_on_linkup(unsigned long opaque) diff --git a/drivers/infiniband/hw/qib/qib_keys.c b/drivers/infiniband/hw/qib/qib_keys.c index 3b9afcc..ad843c7 100644 --- a/drivers/infiniband/hw/qib/qib_keys.c +++ b/drivers/infiniband/hw/qib/qib_keys.c @@ -122,10 +122,10 @@ void qib_free_lkey(struct qib_mregion *mr) if (!mr->lkey_published) goto out; if (lkey == 0) - rcu_assign_pointer(dev->dma_mr, NULL); + RCU_INIT_POINTER(dev->dma_mr, NULL); else { r = lkey >> (32 - ib_qib_lkey_table_size); - rcu_assign_pointer(rkt->table[r], NULL); + RCU_INIT_POINTER(rkt->table[r], NULL); } qib_put_mr(mr); mr->lkey_published = 0; diff --git a/drivers/infiniband/hw/qib/qib_mad.c b/drivers/infiniband/hw/qib/qib_mad.c index 636be11..395f404 100644 --- a/drivers/infiniband/hw/qib/qib_mad.c +++ b/drivers/infiniband/hw/qib/qib_mad.c @@ -152,14 +152,14 @@ void qib_bad_pqkey(struct qib_ibport *ibp, __be16 trap_num, u32 key, u32 sl, data.trap_num = trap_num; data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid); data.toggle_count = 0; - memset(&data.details, 0, sizeof data.details); + memset(&data.details, 0, sizeof(data.details)); data.details.ntc_257_258.lid1 = lid1; data.details.ntc_257_258.lid2 = lid2; data.details.ntc_257_258.key = cpu_to_be32(key); data.details.ntc_257_258.sl_qp1 = cpu_to_be32((sl << 28) | qp1); data.details.ntc_257_258.qp2 = cpu_to_be32(qp2); - qib_send_trap(ibp, &data, sizeof data); + qib_send_trap(ibp, &data, sizeof(data)); } /* @@ -176,7 +176,7 @@ static void qib_bad_mkey(struct qib_ibport *ibp, struct ib_smp *smp) data.trap_num = IB_NOTICE_TRAP_BAD_MKEY; data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid); data.toggle_count = 0; - memset(&data.details, 0, sizeof data.details); + memset(&data.details, 0, sizeof(data.details)); data.details.ntc_256.lid = data.issuer_lid; data.details.ntc_256.method = smp->method; data.details.ntc_256.attr_id = smp->attr_id; @@ -198,7 +198,7 @@ static void qib_bad_mkey(struct qib_ibport *ibp, struct ib_smp *smp) hop_cnt); } - qib_send_trap(ibp, &data, sizeof data); + qib_send_trap(ibp, &data, sizeof(data)); } /* @@ -214,11 +214,11 @@ void qib_cap_mask_chg(struct qib_ibport *ibp) data.trap_num = IB_NOTICE_TRAP_CAP_MASK_CHG; data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid); data.toggle_count = 0; - memset(&data.details, 0, sizeof data.details); + memset(&data.details, 0, sizeof(data.details)); data.details.ntc_144.lid = data.issuer_lid; data.details.ntc_144.new_cap_mask = cpu_to_be32(ibp->port_cap_flags); - qib_send_trap(ibp, &data, sizeof data); + qib_send_trap(ibp, &data, sizeof(data)); } /* @@ -234,11 +234,11 @@ void qib_sys_guid_chg(struct qib_ibport *ibp) data.trap_num = IB_NOTICE_TRAP_SYS_GUID_CHG; data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid); data.toggle_count = 0; - memset(&data.details, 0, sizeof data.details); + memset(&data.details, 0, sizeof(data.details)); data.details.ntc_145.lid = data.issuer_lid; data.details.ntc_145.new_sys_guid = ib_qib_sys_image_guid; - qib_send_trap(ibp, &data, sizeof data); + qib_send_trap(ibp, &data, sizeof(data)); } /* @@ -254,12 +254,12 @@ void qib_node_desc_chg(struct qib_ibport *ibp) data.trap_num = IB_NOTICE_TRAP_CAP_MASK_CHG; data.issuer_lid = cpu_to_be16(ppd_from_ibp(ibp)->lid); data.toggle_count = 0; - memset(&data.details, 0, sizeof data.details); + memset(&data.details, 0, sizeof(data.details)); data.details.ntc_144.lid = data.issuer_lid; data.details.ntc_144.local_changes = 1; data.details.ntc_144.change_flags = IB_NOTICE_TRAP_NODE_DESC_CHG; - qib_send_trap(ibp, &data, sizeof data); + qib_send_trap(ibp, &data, sizeof(data)); } static int subn_get_nodedescription(struct ib_smp *smp, diff --git a/drivers/infiniband/hw/qib/qib_mmap.c b/drivers/infiniband/hw/qib/qib_mmap.c index 8b73a11..146cf29 100644 --- a/drivers/infiniband/hw/qib/qib_mmap.c +++ b/drivers/infiniband/hw/qib/qib_mmap.c @@ -134,7 +134,7 @@ struct qib_mmap_info *qib_create_mmap_info(struct qib_ibdev *dev, void *obj) { struct qib_mmap_info *ip; - ip = kmalloc(sizeof *ip, GFP_KERNEL); + ip = kmalloc(sizeof(*ip), GFP_KERNEL); if (!ip) goto bail; diff --git a/drivers/infiniband/hw/qib/qib_mr.c b/drivers/infiniband/hw/qib/qib_mr.c index a77fb4f..c4473db 100644 --- a/drivers/infiniband/hw/qib/qib_mr.c +++ b/drivers/infiniband/hw/qib/qib_mr.c @@ -55,7 +55,7 @@ static int init_qib_mregion(struct qib_mregion *mr, struct ib_pd *pd, m = (count + QIB_SEGSZ - 1) / QIB_SEGSZ; for (; i < m; i++) { - mr->map[i] = kzalloc(sizeof *mr->map[0], GFP_KERNEL); + mr->map[i] = kzalloc(sizeof(*mr->map[0]), GFP_KERNEL); if (!mr->map[i]) goto bail; } @@ -104,7 +104,7 @@ struct ib_mr *qib_get_dma_mr(struct ib_pd *pd, int acc) goto bail; } - mr = kzalloc(sizeof *mr, GFP_KERNEL); + mr = kzalloc(sizeof(*mr), GFP_KERNEL); if (!mr) { ret = ERR_PTR(-ENOMEM); goto bail; @@ -143,7 +143,7 @@ static struct qib_mr *alloc_mr(int count, struct ib_pd *pd) /* Allocate struct plus pointers to first level page tables. */ m = (count + QIB_SEGSZ - 1) / QIB_SEGSZ; - mr = kzalloc(sizeof *mr + m * sizeof mr->mr.map[0], GFP_KERNEL); + mr = kzalloc(sizeof(*mr) + m * sizeof(mr->mr.map[0]), GFP_KERNEL); if (!mr) goto bail; @@ -347,7 +347,7 @@ qib_alloc_fast_reg_page_list(struct ib_device *ibdev, int page_list_len) if (size > PAGE_SIZE) return ERR_PTR(-EINVAL); - pl = kzalloc(sizeof *pl, GFP_KERNEL); + pl = kzalloc(sizeof(*pl), GFP_KERNEL); if (!pl) return ERR_PTR(-ENOMEM); @@ -386,7 +386,7 @@ struct ib_fmr *qib_alloc_fmr(struct ib_pd *pd, int mr_access_flags, /* Allocate struct plus pointers to first level page tables. */ m = (fmr_attr->max_pages + QIB_SEGSZ - 1) / QIB_SEGSZ; - fmr = kzalloc(sizeof *fmr + m * sizeof fmr->mr.map[0], GFP_KERNEL); + fmr = kzalloc(sizeof(*fmr) + m * sizeof(fmr->mr.map[0]), GFP_KERNEL); if (!fmr) goto bail; diff --git a/drivers/infiniband/hw/qib/qib_pcie.c b/drivers/infiniband/hw/qib/qib_pcie.c index 61a0046..4758a38 100644 --- a/drivers/infiniband/hw/qib/qib_pcie.c +++ b/drivers/infiniband/hw/qib/qib_pcie.c @@ -210,7 +210,7 @@ static void qib_msix_setup(struct qib_devdata *dd, int pos, u32 *msixcnt, /* We can't pass qib_msix_entry array to qib_msix_setup * so use a dummy msix_entry array and copy the allocated * irq back to the qib_msix_entry array. */ - msix_entry = kmalloc(nvec * sizeof(*msix_entry), GFP_KERNEL); + msix_entry = kcalloc(nvec, sizeof(*msix_entry), GFP_KERNEL); if (!msix_entry) goto do_intx; @@ -234,8 +234,10 @@ free_msix_entry: kfree(msix_entry); do_intx: - qib_dev_err(dd, "pci_enable_msix_range %d vectors failed: %d, " - "falling back to INTx\n", nvec, ret); + qib_dev_err( + dd, + "pci_enable_msix_range %d vectors failed: %d, falling back to INTx\n", + nvec, ret); *msixcnt = 0; qib_enable_intx(dd->pcidev); } @@ -459,6 +461,7 @@ void qib_pcie_getcmd(struct qib_devdata *dd, u16 *cmd, u8 *iline, u8 *cline) void qib_pcie_reenable(struct qib_devdata *dd, u16 cmd, u8 iline, u8 cline) { int r; + r = pci_write_config_dword(dd->pcidev, PCI_BASE_ADDRESS_0, dd->pcibar0); if (r) @@ -696,6 +699,7 @@ static void qib_pci_resume(struct pci_dev *pdev) { struct qib_devdata *dd = pci_get_drvdata(pdev); + qib_devinfo(pdev, "QIB resume function called\n"); pci_cleanup_aer_uncorrect_error_status(pdev); /* diff --git a/drivers/infiniband/hw/qib/qib_qp.c b/drivers/infiniband/hw/qib/qib_qp.c index 6ddc026..4fa88ba 100644 --- a/drivers/infiniband/hw/qib/qib_qp.c +++ b/drivers/infiniband/hw/qib/qib_qp.c @@ -255,10 +255,10 @@ static void remove_qp(struct qib_ibdev *dev, struct qib_qp *qp) if (rcu_dereference_protected(ibp->qp0, lockdep_is_held(&dev->qpt_lock)) == qp) { - rcu_assign_pointer(ibp->qp0, NULL); + RCU_INIT_POINTER(ibp->qp0, NULL); } else if (rcu_dereference_protected(ibp->qp1, lockdep_is_held(&dev->qpt_lock)) == qp) { - rcu_assign_pointer(ibp->qp1, NULL); + RCU_INIT_POINTER(ibp->qp1, NULL); } else { struct qib_qp *q; struct qib_qp __rcu **qpp; @@ -269,7 +269,7 @@ static void remove_qp(struct qib_ibdev *dev, struct qib_qp *qp) lockdep_is_held(&dev->qpt_lock))) != NULL; qpp = &q->next) if (q == qp) { - rcu_assign_pointer(*qpp, + RCU_INIT_POINTER(*qpp, rcu_dereference_protected(qp->next, lockdep_is_held(&dev->qpt_lock))); removed = 1; @@ -315,7 +315,7 @@ unsigned qib_free_all_qps(struct qib_devdata *dd) for (n = 0; n < dev->qp_table_size; n++) { qp = rcu_dereference_protected(dev->qp_table[n], lockdep_is_held(&dev->qpt_lock)); - rcu_assign_pointer(dev->qp_table[n], NULL); + RCU_INIT_POINTER(dev->qp_table[n], NULL); for (; qp; qp = rcu_dereference_protected(qp->next, lockdep_is_held(&dev->qpt_lock))) diff --git a/drivers/infiniband/hw/qib/qib_qsfp.c b/drivers/infiniband/hw/qib/qib_qsfp.c index fa71b1e..5e27f76 100644 --- a/drivers/infiniband/hw/qib/qib_qsfp.c +++ b/drivers/infiniband/hw/qib/qib_qsfp.c @@ -81,7 +81,7 @@ static int qsfp_read(struct qib_pportdata *ppd, int addr, void *bp, int len) * Module could take up to 2 Msec to respond to MOD_SEL, and there * is no way to tell if it is ready, so we must wait. */ - msleep(2); + msleep(20); /* Make sure TWSI bus is in sane state. */ ret = qib_twsi_reset(dd); @@ -99,6 +99,7 @@ static int qsfp_read(struct qib_pportdata *ppd, int addr, void *bp, int len) while (cnt < len) { unsigned in_page; int wlen = len - cnt; + in_page = addr % QSFP_PAGESIZE; if ((in_page + wlen) > QSFP_PAGESIZE) wlen = QSFP_PAGESIZE - in_page; @@ -139,7 +140,7 @@ deselect: else if (pass) qib_dev_porterr(dd, ppd->port, "QSFP retries: %d\n", pass); - msleep(2); + msleep(20); bail: mutex_unlock(&dd->eep_lock); @@ -189,7 +190,7 @@ static int qib_qsfp_write(struct qib_pportdata *ppd, int addr, void *bp, * Module could take up to 2 Msec to respond to MOD_SEL, * and there is no way to tell if it is ready, so we must wait. */ - msleep(2); + msleep(20); /* Make sure TWSI bus is in sane state. */ ret = qib_twsi_reset(dd); @@ -206,6 +207,7 @@ static int qib_qsfp_write(struct qib_pportdata *ppd, int addr, void *bp, while (cnt < len) { unsigned in_page; int wlen = len - cnt; + in_page = addr % QSFP_PAGESIZE; if ((in_page + wlen) > QSFP_PAGESIZE) wlen = QSFP_PAGESIZE - in_page; @@ -234,7 +236,7 @@ deselect: * going away, and there is no way to tell if it is ready. * so we must wait. */ - msleep(2); + msleep(20); bail: mutex_unlock(&dd->eep_lock); @@ -296,6 +298,7 @@ int qib_refresh_qsfp_cache(struct qib_pportdata *ppd, struct qib_qsfp_cache *cp) * set the page to zero, Even if it already appears to be zero. */ u8 poke = 0; + ret = qib_qsfp_write(ppd, 127, &poke, 1); udelay(50); if (ret != 1) { @@ -480,7 +483,6 @@ void qib_qsfp_init(struct qib_qsfp_data *qd, udelay(20); /* Generous RST dwell */ dd->f_gpio_mod(dd, mask, mask, mask); - return; } void qib_qsfp_deinit(struct qib_qsfp_data *qd) @@ -540,6 +542,7 @@ int qib_qsfp_dump(struct qib_pportdata *ppd, char *buf, int len) while (bidx < QSFP_DEFAULT_HDR_CNT) { int iidx; + ret = qsfp_read(ppd, bidx, bin_buff, QSFP_DUMP_CHUNK); if (ret < 0) goto bail; diff --git a/drivers/infiniband/hw/qib/qib_rc.c b/drivers/infiniband/hw/qib/qib_rc.c index 2f25018..4544d6f 100644 --- a/drivers/infiniband/hw/qib/qib_rc.c +++ b/drivers/infiniband/hw/qib/qib_rc.c @@ -1017,7 +1017,7 @@ void qib_rc_send_complete(struct qib_qp *qp, struct qib_ib_header *hdr) /* Post a send completion queue entry if requested. */ if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) || (wqe->wr.send_flags & IB_SEND_SIGNALED)) { - memset(&wc, 0, sizeof wc); + memset(&wc, 0, sizeof(wc)); wc.wr_id = wqe->wr.wr_id; wc.status = IB_WC_SUCCESS; wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode]; @@ -1073,7 +1073,7 @@ static struct qib_swqe *do_rc_completion(struct qib_qp *qp, /* Post a send completion queue entry if requested. */ if (!(qp->s_flags & QIB_S_SIGNAL_REQ_WR) || (wqe->wr.send_flags & IB_SEND_SIGNALED)) { - memset(&wc, 0, sizeof wc); + memset(&wc, 0, sizeof(wc)); wc.wr_id = wqe->wr.wr_id; wc.status = IB_WC_SUCCESS; wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode]; diff --git a/drivers/infiniband/hw/qib/qib_ruc.c b/drivers/infiniband/hw/qib/qib_ruc.c index 4c07a8b..f42bd0f 100644 --- a/drivers/infiniband/hw/qib/qib_ruc.c +++ b/drivers/infiniband/hw/qib/qib_ruc.c @@ -247,8 +247,8 @@ static __be64 get_sguid(struct qib_ibport *ibp, unsigned index) struct qib_pportdata *ppd = ppd_from_ibp(ibp); return ppd->guid; - } else - return ibp->guids[index - 1]; + } + return ibp->guids[index - 1]; } static int gid_ok(union ib_gid *gid, __be64 gid_prefix, __be64 id) @@ -420,7 +420,7 @@ again: goto serr; } - memset(&wc, 0, sizeof wc); + memset(&wc, 0, sizeof(wc)); send_status = IB_WC_SUCCESS; release = 1; @@ -792,7 +792,7 @@ void qib_send_complete(struct qib_qp *qp, struct qib_swqe *wqe, status != IB_WC_SUCCESS) { struct ib_wc wc; - memset(&wc, 0, sizeof wc); + memset(&wc, 0, sizeof(wc)); wc.wr_id = wqe->wr.wr_id; wc.status = status; wc.opcode = ib_qib_wc_opcode[wqe->wr.opcode]; diff --git a/drivers/infiniband/hw/qib/qib_sd7220.c b/drivers/infiniband/hw/qib/qib_sd7220.c index 911205d..c72775f 100644 --- a/drivers/infiniband/hw/qib/qib_sd7220.c +++ b/drivers/infiniband/hw/qib/qib_sd7220.c @@ -259,6 +259,7 @@ static int qib_ibsd_reset(struct qib_devdata *dd, int assert_rst) * it again during startup. */ u64 val; + rst_val &= ~(1ULL); qib_write_kreg(dd, kr_hwerrmask, dd->cspec->hwerrmask & @@ -590,6 +591,7 @@ static int epb_access(struct qib_devdata *dd, int sdnum, int claim) * Both should be clear */ u64 newval = 0; + qib_write_kreg(dd, acc, newval); /* First read after write is not trustworthy */ pollval = qib_read_kreg32(dd, acc); @@ -601,6 +603,7 @@ static int epb_access(struct qib_devdata *dd, int sdnum, int claim) /* Need to claim */ u64 pollval; u64 newval = EPB_ACC_REQ | oct_sel; + qib_write_kreg(dd, acc, newval); /* First read after write is not trustworthy */ pollval = qib_read_kreg32(dd, acc); @@ -812,6 +815,7 @@ static int qib_sd7220_ram_xfer(struct qib_devdata *dd, int sdnum, u32 loc, if (!sofar) { /* Only set address at start of chunk */ int addrbyte = (addr + sofar) >> 8; + transval = csbit | EPB_MADDRH | addrbyte; tries = epb_trans(dd, trans, transval, &transval); @@ -922,7 +926,7 @@ qib_sd7220_ib_vfy(struct qib_devdata *dd, const struct firmware *fw) * IRQ not set up at this point in init, so we poll. */ #define IB_SERDES_TRIM_DONE (1ULL << 11) -#define TRIM_TMO (30) +#define TRIM_TMO (15) static int qib_sd_trimdone_poll(struct qib_devdata *dd) { @@ -940,7 +944,7 @@ static int qib_sd_trimdone_poll(struct qib_devdata *dd) ret = 1; break; } - msleep(10); + msleep(20); } if (trim_tmo >= TRIM_TMO) { qib_dev_err(dd, "No TRIMDONE in %d tries\n", trim_tmo); @@ -1071,6 +1075,7 @@ static int qib_sd_setvals(struct qib_devdata *dd) dds_reg_map >>= 4; for (midx = 0; midx < DDS_ROWS; ++midx) { u64 __iomem *daddr = taddr + ((midx << 4) + idx); + data = dds_init_vals[midx].reg_vals[idx]; writeq(data, daddr); mmiowb(); diff --git a/drivers/infiniband/hw/qib/qib_sysfs.c b/drivers/infiniband/hw/qib/qib_sysfs.c index 3c8e4e3..81f56cd 100644 --- a/drivers/infiniband/hw/qib/qib_sysfs.c +++ b/drivers/infiniband/hw/qib/qib_sysfs.c @@ -586,8 +586,8 @@ static ssize_t show_serial(struct device *device, container_of(device, struct qib_ibdev, ibdev.dev); struct qib_devdata *dd = dd_from_dev(dev); - buf[sizeof dd->serial] = '\0'; - memcpy(buf, dd->serial, sizeof dd->serial); + buf[sizeof(dd->serial)] = '\0'; + memcpy(buf, dd->serial, sizeof(dd->serial)); strcat(buf, "\n"); return strlen(buf); } @@ -611,28 +611,6 @@ bail: return ret < 0 ? ret : count; } -static ssize_t show_logged_errs(struct device *device, - struct device_attribute *attr, char *buf) -{ - struct qib_ibdev *dev = - container_of(device, struct qib_ibdev, ibdev.dev); - struct qib_devdata *dd = dd_from_dev(dev); - int idx, count; - - /* force consistency with actual EEPROM */ - if (qib_update_eeprom_log(dd) != 0) - return -ENXIO; - - count = 0; - for (idx = 0; idx < QIB_EEP_LOG_CNT; ++idx) { - count += scnprintf(buf + count, PAGE_SIZE - count, "%d%c", - dd->eep_st_errs[idx], - idx == (QIB_EEP_LOG_CNT - 1) ? '\n' : ' '); - } - - return count; -} - /* * Dump tempsense regs. in decimal, to ease shell-scripts. */ @@ -679,7 +657,6 @@ static DEVICE_ATTR(nctxts, S_IRUGO, show_nctxts, NULL); static DEVICE_ATTR(nfreectxts, S_IRUGO, show_nfreectxts, NULL); static DEVICE_ATTR(serial, S_IRUGO, show_serial, NULL); static DEVICE_ATTR(boardversion, S_IRUGO, show_boardversion, NULL); -static DEVICE_ATTR(logged_errors, S_IRUGO, show_logged_errs, NULL); static DEVICE_ATTR(tempsense, S_IRUGO, show_tempsense, NULL); static DEVICE_ATTR(localbus_info, S_IRUGO, show_localbus_info, NULL); static DEVICE_ATTR(chip_reset, S_IWUSR, NULL, store_chip_reset); @@ -693,7 +670,6 @@ static struct device_attribute *qib_attributes[] = { &dev_attr_nfreectxts, &dev_attr_serial, &dev_attr_boardversion, - &dev_attr_logged_errors, &dev_attr_tempsense, &dev_attr_localbus_info, &dev_attr_chip_reset, diff --git a/drivers/infiniband/hw/qib/qib_twsi.c b/drivers/infiniband/hw/qib/qib_twsi.c index 647f7be..f569866 100644 --- a/drivers/infiniband/hw/qib/qib_twsi.c +++ b/drivers/infiniband/hw/qib/qib_twsi.c @@ -105,6 +105,7 @@ static void scl_out(struct qib_devdata *dd, u8 bit) udelay(2); else { int rise_usec; + for (rise_usec = SCL_WAIT_USEC; rise_usec > 0; rise_usec -= 2) { if (mask & dd->f_gpio_mod(dd, 0, 0, 0)) break; @@ -326,6 +327,7 @@ int qib_twsi_reset(struct qib_devdata *dd) static int qib_twsi_wr(struct qib_devdata *dd, int data, int flags) { int ret = 1; + if (flags & QIB_TWSI_START) start_seq(dd); @@ -435,8 +437,7 @@ int qib_twsi_blk_wr(struct qib_devdata *dd, int dev, int addr, int sub_len; const u8 *bp = buffer; int max_wait_time, i; - int ret; - ret = 1; + int ret = 1; while (len > 0) { if (dev == QIB_TWSI_NO_DEV) { diff --git a/drivers/infiniband/hw/qib/qib_tx.c b/drivers/infiniband/hw/qib/qib_tx.c index 31d3561..eface3b 100644 --- a/drivers/infiniband/hw/qib/qib_tx.c +++ b/drivers/infiniband/hw/qib/qib_tx.c @@ -180,6 +180,7 @@ void qib_disarm_piobufs_set(struct qib_devdata *dd, unsigned long *mask, for (i = 0; i < cnt; i++) { int which; + if (!test_bit(i, mask)) continue; /* diff --git a/drivers/infiniband/hw/qib/qib_ud.c b/drivers/infiniband/hw/qib/qib_ud.c index aaf7039..26243b7 100644 --- a/drivers/infiniband/hw/qib/qib_ud.c +++ b/drivers/infiniband/hw/qib/qib_ud.c @@ -127,7 +127,7 @@ static void qib_ud_loopback(struct qib_qp *sqp, struct qib_swqe *swqe) * present on the wire. */ length = swqe->length; - memset(&wc, 0, sizeof wc); + memset(&wc, 0, sizeof(wc)); wc.byte_len = length + sizeof(struct ib_grh); if (swqe->wr.opcode == IB_WR_SEND_WITH_IMM) { diff --git a/drivers/infiniband/hw/qib/qib_user_sdma.c b/drivers/infiniband/hw/qib/qib_user_sdma.c index d2806ca..3e0677c 100644 --- a/drivers/infiniband/hw/qib/qib_user_sdma.c +++ b/drivers/infiniband/hw/qib/qib_user_sdma.c @@ -50,7 +50,7 @@ /* expected size of headers (for dma_pool) */ #define QIB_USER_SDMA_EXP_HEADER_LENGTH 64 /* attempt to drain the queue for 5secs */ -#define QIB_USER_SDMA_DRAIN_TIMEOUT 500 +#define QIB_USER_SDMA_DRAIN_TIMEOUT 250 /* * track how many times a process open this driver. @@ -226,6 +226,7 @@ qib_user_sdma_queue_create(struct device *dev, int unit, int ctxt, int sctxt) sdma_rb_node->refcount++; } else { int ret; + sdma_rb_node = kmalloc(sizeof( struct qib_user_sdma_rb_node), GFP_KERNEL); if (!sdma_rb_node) @@ -936,6 +937,7 @@ static int qib_user_sdma_queue_pkts(const struct qib_devdata *dd, if (tiddma) { char *tidsm = (char *)pkt + pktsize; + cfur = copy_from_user(tidsm, iov[idx].iov_base, tidsmsize); if (cfur) { @@ -1142,7 +1144,7 @@ void qib_user_sdma_queue_drain(struct qib_pportdata *ppd, qib_user_sdma_hwqueue_clean(ppd); qib_user_sdma_queue_clean(ppd, pq); mutex_unlock(&pq->lock); - msleep(10); + msleep(20); } if (pq->num_pending || pq->num_sending) { @@ -1316,8 +1318,6 @@ retry: if (nfree && !list_empty(pktlist)) goto retry; - - return; } /* pq->lock must be held, get packets on the wire... */ diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c index 9bcfbd8..4a35998 100644 --- a/drivers/infiniband/hw/qib/qib_verbs.c +++ b/drivers/infiniband/hw/qib/qib_verbs.c @@ -1342,6 +1342,7 @@ static int qib_verbs_send_pio(struct qib_qp *qp, struct qib_ib_header *ibhdr, done: if (dd->flags & QIB_USE_SPCL_TRIG) { u32 spcl_off = (pbufn >= dd->piobcnt2k) ? 2047 : 1023; + qib_flush_wc(); __raw_writel(0xaebecede, piobuf_orig + spcl_off); } @@ -1744,7 +1745,7 @@ static struct ib_pd *qib_alloc_pd(struct ib_device *ibdev, * we allow allocations of more than we report for this value. */ - pd = kmalloc(sizeof *pd, GFP_KERNEL); + pd = kmalloc(sizeof(*pd), GFP_KERNEL); if (!pd) { ret = ERR_PTR(-ENOMEM); goto bail; @@ -1829,7 +1830,7 @@ static struct ib_ah *qib_create_ah(struct ib_pd *pd, goto bail; } - ah = kmalloc(sizeof *ah, GFP_ATOMIC); + ah = kmalloc(sizeof(*ah), GFP_ATOMIC); if (!ah) { ret = ERR_PTR(-ENOMEM); goto bail; @@ -1862,7 +1863,7 @@ struct ib_ah *qib_create_qp0_ah(struct qib_ibport *ibp, u16 dlid) struct ib_ah *ah = ERR_PTR(-EINVAL); struct qib_qp *qp0; - memset(&attr, 0, sizeof attr); + memset(&attr, 0, sizeof(attr)); attr.dlid = dlid; attr.port_num = ppd_from_ibp(ibp)->port; rcu_read_lock(); @@ -1977,7 +1978,7 @@ static struct ib_ucontext *qib_alloc_ucontext(struct ib_device *ibdev, struct qib_ucontext *context; struct ib_ucontext *ret; - context = kmalloc(sizeof *context, GFP_KERNEL); + context = kmalloc(sizeof(*context), GFP_KERNEL); if (!context) { ret = ERR_PTR(-ENOMEM); goto bail; @@ -2054,7 +2055,9 @@ int qib_register_ib_device(struct qib_devdata *dd) dev->qp_table_size = ib_qib_qp_table_size; get_random_bytes(&dev->qp_rnd, sizeof(dev->qp_rnd)); - dev->qp_table = kmalloc(dev->qp_table_size * sizeof *dev->qp_table, + dev->qp_table = kmalloc_array( + dev->qp_table_size, + sizeof(*dev->qp_table), GFP_KERNEL); if (!dev->qp_table) { ret = -ENOMEM; @@ -2122,7 +2125,7 @@ int qib_register_ib_device(struct qib_devdata *dd) for (i = 0; i < ppd->sdma_descq_cnt; i++) { struct qib_verbs_txreq *tx; - tx = kzalloc(sizeof *tx, GFP_KERNEL); + tx = kzalloc(sizeof(*tx), GFP_KERNEL); if (!tx) { ret = -ENOMEM; goto err_tx; diff --git a/drivers/infiniband/hw/qib/qib_verbs_mcast.c b/drivers/infiniband/hw/qib/qib_verbs_mcast.c index dabb697..f8ea069 100644 --- a/drivers/infiniband/hw/qib/qib_verbs_mcast.c +++ b/drivers/infiniband/hw/qib/qib_verbs_mcast.c @@ -43,7 +43,7 @@ static struct qib_mcast_qp *qib_mcast_qp_alloc(struct qib_qp *qp) { struct qib_mcast_qp *mqp; - mqp = kmalloc(sizeof *mqp, GFP_KERNEL); + mqp = kmalloc(sizeof(*mqp), GFP_KERNEL); if (!mqp) goto bail; @@ -75,7 +75,7 @@ static struct qib_mcast *qib_mcast_alloc(union ib_gid *mgid) { struct qib_mcast *mcast; - mcast = kmalloc(sizeof *mcast, GFP_KERNEL); + mcast = kmalloc(sizeof(*mcast), GFP_KERNEL); if (!mcast) goto bail; diff --git a/drivers/infiniband/hw/qib/qib_wc_x86_64.c b/drivers/infiniband/hw/qib/qib_wc_x86_64.c index 1d7281c..81b225f 100644 --- a/drivers/infiniband/hw/qib/qib_wc_x86_64.c +++ b/drivers/infiniband/hw/qib/qib_wc_x86_64.c @@ -72,6 +72,7 @@ int qib_enable_wc(struct qib_devdata *dd) if (dd->piobcnt2k && dd->piobcnt4k) { /* 2 sizes for chip */ unsigned long pio2kbase, pio4kbase; + pio2kbase = dd->piobufbase & 0xffffffffUL; pio4kbase = (dd->piobufbase >> 32) & 0xffffffffUL; if (pio2kbase < pio4kbase) { @@ -91,7 +92,7 @@ int qib_enable_wc(struct qib_devdata *dd) } for (bits = 0; !(piolen & (1ULL << bits)); bits++) - /* do nothing */ ; + ; /* do nothing */ if (piolen != (1ULL << bits)) { piolen >>= bits; @@ -100,8 +101,8 @@ int qib_enable_wc(struct qib_devdata *dd) piolen = 1ULL << (bits + 1); } if (pioaddr & (piolen - 1)) { - u64 atmp; - atmp = pioaddr & ~(piolen - 1); + u64 atmp = pioaddr & ~(piolen - 1); + if (atmp < addr || (atmp + piolen) > (addr + len)) { qib_dev_err(dd, "No way to align address/size (%llx/%llx), no WC mtrr\n", diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 5ce2681..b47aea1 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -654,7 +654,9 @@ int iser_dma_map_task_data(struct iscsi_iser_task *iser_task, enum dma_data_direction dma_dir); void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task, - struct iser_data_buf *data); + struct iser_data_buf *data, + enum dma_data_direction dir); + int iser_initialize_task_headers(struct iscsi_task *task, struct iser_tx_desc *tx_desc); int iser_alloc_rx_descriptors(struct iser_conn *iser_conn, diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index 3821633..20e859a 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -320,9 +320,6 @@ void iser_free_rx_descriptors(struct iser_conn *iser_conn) struct ib_conn *ib_conn = &iser_conn->ib_conn; struct iser_device *device = ib_conn->device; - if (!iser_conn->rx_descs) - goto free_login_buf; - if (device->iser_free_rdma_reg_res) device->iser_free_rdma_reg_res(ib_conn); @@ -334,7 +331,6 @@ void iser_free_rx_descriptors(struct iser_conn *iser_conn) /* make sure we never redo any unmapping */ iser_conn->rx_descs = NULL; -free_login_buf: iser_free_login_buf(iser_conn); } @@ -714,19 +710,23 @@ void iser_task_rdma_finalize(struct iscsi_iser_task *iser_task) device->iser_unreg_rdma_mem(iser_task, ISER_DIR_IN); if (is_rdma_data_aligned) iser_dma_unmap_task_data(iser_task, - &iser_task->data[ISER_DIR_IN]); + &iser_task->data[ISER_DIR_IN], + DMA_FROM_DEVICE); if (prot_count && is_rdma_prot_aligned) iser_dma_unmap_task_data(iser_task, - &iser_task->prot[ISER_DIR_IN]); + &iser_task->prot[ISER_DIR_IN], + DMA_FROM_DEVICE); } if (iser_task->dir[ISER_DIR_OUT]) { device->iser_unreg_rdma_mem(iser_task, ISER_DIR_OUT); if (is_rdma_data_aligned) iser_dma_unmap_task_data(iser_task, - &iser_task->data[ISER_DIR_OUT]); + &iser_task->data[ISER_DIR_OUT], + DMA_TO_DEVICE); if (prot_count && is_rdma_prot_aligned) iser_dma_unmap_task_data(iser_task, - &iser_task->prot[ISER_DIR_OUT]); + &iser_task->prot[ISER_DIR_OUT], + DMA_TO_DEVICE); } } diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index abce933..341040b 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -332,12 +332,13 @@ int iser_dma_map_task_data(struct iscsi_iser_task *iser_task, } void iser_dma_unmap_task_data(struct iscsi_iser_task *iser_task, - struct iser_data_buf *data) + struct iser_data_buf *data, + enum dma_data_direction dir) { struct ib_device *dev; dev = iser_task->iser_conn->ib_conn.device->ib_device; - ib_dma_unmap_sg(dev, data->buf, data->size, DMA_FROM_DEVICE); + ib_dma_unmap_sg(dev, data->buf, data->size, dir); } static int fall_to_bounce_buf(struct iscsi_iser_task *iser_task, @@ -357,7 +358,9 @@ static int fall_to_bounce_buf(struct iscsi_iser_task *iser_task, iser_data_buf_dump(mem, ibdev); /* unmap the command data before accessing it */ - iser_dma_unmap_task_data(iser_task, mem); + iser_dma_unmap_task_data(iser_task, mem, + (cmd_dir == ISER_DIR_OUT) ? + DMA_TO_DEVICE : DMA_FROM_DEVICE); /* allocate copy buf, if we are writing, copy the */ /* unaligned scatterlist, dma map the copy */ diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index 695a270..4065abe 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -600,16 +600,16 @@ void iser_release_work(struct work_struct *work) /** * iser_free_ib_conn_res - release IB related resources * @iser_conn: iser connection struct - * @destroy_device: indicator if we need to try to release - * the iser device (only iscsi shutdown and DEVICE_REMOVAL - * will use this. + * @destroy: indicator if we need to try to release the + * iser device and memory regoins pool (only iscsi + * shutdown and DEVICE_REMOVAL will use this). * * This routine is called with the iser state mutex held * so the cm_id removal is out of here. It is Safe to * be invoked multiple times. */ static void iser_free_ib_conn_res(struct iser_conn *iser_conn, - bool destroy_device) + bool destroy) { struct ib_conn *ib_conn = &iser_conn->ib_conn; struct iser_device *device = ib_conn->device; @@ -617,17 +617,20 @@ static void iser_free_ib_conn_res(struct iser_conn *iser_conn, iser_info("freeing conn %p cma_id %p qp %p\n", iser_conn, ib_conn->cma_id, ib_conn->qp); - iser_free_rx_descriptors(iser_conn); - if (ib_conn->qp != NULL) { ib_conn->comp->active_qps--; rdma_destroy_qp(ib_conn->cma_id); ib_conn->qp = NULL; } - if (destroy_device && device != NULL) { - iser_device_try_release(device); - ib_conn->device = NULL; + if (destroy) { + if (iser_conn->rx_descs) + iser_free_rx_descriptors(iser_conn); + + if (device != NULL) { + iser_device_try_release(device); + ib_conn->device = NULL; + } } } @@ -643,9 +646,11 @@ void iser_conn_release(struct iser_conn *iser_conn) mutex_unlock(&ig.connlist_mutex); mutex_lock(&iser_conn->state_mutex); + /* In case we endup here without ep_disconnect being invoked. */ if (iser_conn->state != ISER_CONN_DOWN) { iser_warn("iser conn %p state %d, expected state down.\n", iser_conn, iser_conn->state); + iscsi_destroy_endpoint(iser_conn->ep); iser_conn->state = ISER_CONN_DOWN; } /* @@ -840,7 +845,7 @@ static void iser_disconnected_handler(struct rdma_cm_id *cma_id) } static void iser_cleanup_handler(struct rdma_cm_id *cma_id, - bool destroy_device) + bool destroy) { struct iser_conn *iser_conn = (struct iser_conn *)cma_id->context; @@ -850,7 +855,7 @@ static void iser_cleanup_handler(struct rdma_cm_id *cma_id, * and flush errors. */ iser_disconnected_handler(cma_id); - iser_free_ib_conn_res(iser_conn, destroy_device); + iser_free_ib_conn_res(iser_conn, destroy); complete(&iser_conn->ib_completion); }; |