From 609e941a6bcd7ceb1cbb561941c997f6465e8698 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Thu, 1 Sep 2016 06:43:46 -0700 Subject: iw_cxgb4: call dev_put() on l2t allocation failure Reviewed-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Steve Wise Signed-off-by: Sagi Grimberg --- drivers/infiniband/hw/cxgb4/cm.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c index a3a67216bce6..57b5bb56cdbf 100644 --- a/drivers/infiniband/hw/cxgb4/cm.c +++ b/drivers/infiniband/hw/cxgb4/cm.c @@ -2077,8 +2077,10 @@ static int import_ep(struct c4iw_ep *ep, int iptype, __u8 *peer_ip, } ep->l2t = cxgb4_l2t_get(cdev->rdev.lldi.l2t, n, pdev, rt_tos2priority(tos)); - if (!ep->l2t) + if (!ep->l2t) { + dev_put(pdev); goto out; + } ep->mtu = pdev->mtu; ep->tx_chan = cxgb4_port_chan(pdev); ep->smac_idx = cxgb4_tp_smt_idx(adapter_type, -- cgit v1.2.3 From 37eb816c0867b1b0db273d22b530780a0a083980 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Thu, 1 Sep 2016 06:44:52 -0700 Subject: iw_cxgb4: block module unload until all ep resources are released Otherwise an endpoint can be still closing down causing a touch after free crash. Also WARN_ON if ulps have failed to destroy various resources during device removal. Fixes: ad61a4c7a9b7 ("iw_cxgb4: don't block in destroy_qp awaiting the last deref") Reviewed-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Steve Wise Signed-off-by: Sagi Grimberg --- drivers/infiniband/hw/cxgb4/cm.c | 2 ++ drivers/infiniband/hw/cxgb4/device.c | 5 +++++ drivers/infiniband/hw/cxgb4/iw_cxgb4.h | 1 + 3 files changed, 8 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c index 57b5bb56cdbf..5621270664f4 100644 --- a/drivers/infiniband/hw/cxgb4/cm.c +++ b/drivers/infiniband/hw/cxgb4/cm.c @@ -314,6 +314,8 @@ static void remove_ep_tid(struct c4iw_ep *ep) spin_lock_irqsave(&ep->com.dev->lock, flags); _remove_handle(ep->com.dev, &ep->com.dev->hwtid_idr, ep->hwtid, 0); + if (idr_is_empty(&ep->com.dev->hwtid_idr)) + wake_up(&ep->com.dev->wait); spin_unlock_irqrestore(&ep->com.dev->lock, flags); } diff --git a/drivers/infiniband/hw/cxgb4/device.c b/drivers/infiniband/hw/cxgb4/device.c index ae2e8b23d2dd..63561aa9c69e 100644 --- a/drivers/infiniband/hw/cxgb4/device.c +++ b/drivers/infiniband/hw/cxgb4/device.c @@ -872,9 +872,13 @@ static void c4iw_rdev_close(struct c4iw_rdev *rdev) static void c4iw_dealloc(struct uld_ctx *ctx) { c4iw_rdev_close(&ctx->dev->rdev); + WARN_ON_ONCE(!idr_is_empty(&ctx->dev->cqidr)); idr_destroy(&ctx->dev->cqidr); + WARN_ON_ONCE(!idr_is_empty(&ctx->dev->qpidr)); idr_destroy(&ctx->dev->qpidr); + WARN_ON_ONCE(!idr_is_empty(&ctx->dev->mmidr)); idr_destroy(&ctx->dev->mmidr); + wait_event(ctx->dev->wait, idr_is_empty(&ctx->dev->hwtid_idr)); idr_destroy(&ctx->dev->hwtid_idr); idr_destroy(&ctx->dev->stid_idr); idr_destroy(&ctx->dev->atid_idr); @@ -992,6 +996,7 @@ static struct c4iw_dev *c4iw_alloc(const struct cxgb4_lld_info *infop) mutex_init(&devp->rdev.stats.lock); mutex_init(&devp->db_mutex); INIT_LIST_HEAD(&devp->db_fc_list); + init_waitqueue_head(&devp->wait); devp->avail_ird = devp->rdev.lldi.max_ird_adapter; if (c4iw_debugfs_root) { diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h index f6f34a75af27..0e9cd4479bb3 100644 --- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h +++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h @@ -263,6 +263,7 @@ struct c4iw_dev { struct idr stid_idr; struct list_head db_fc_list; u32 avail_ird; + wait_queue_head_t wait; }; static inline struct c4iw_dev *to_c4iw_dev(struct ib_device *ibdev) -- cgit v1.2.3 From e5ac40cd66c2f3cd11bc5edc658f012661b16347 Mon Sep 17 00:00:00 2001 From: Alex Vesker Date: Mon, 12 Sep 2016 19:16:18 +0300 Subject: IB/mlx4: Fix incorrect MC join state bit-masking on SR-IOV Because of an incorrect bit-masking done on the join state bits, when handling a join request we failed to detect a difference between the group join state and the request join state when joining as send only full member (0x8). This caused the MC join request not to be sent. This issue is relevant only when SRIOV is enabled and SM supports send only full member. This fix separates scope bits and join states bits a nibble each. Fixes: b9c5d6a64358 ('IB/mlx4: Add multicast group (MCG) paravirtualization for SR-IOV') Signed-off-by: Alex Vesker Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/mcg.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/mcg.c b/drivers/infiniband/hw/mlx4/mcg.c index 8f7ad07915b0..097bfcc4ee99 100644 --- a/drivers/infiniband/hw/mlx4/mcg.c +++ b/drivers/infiniband/hw/mlx4/mcg.c @@ -489,7 +489,7 @@ static u8 get_leave_state(struct mcast_group *group) if (!group->members[i]) leave_state |= (1 << i); - return leave_state & (group->rec.scope_join_state & 7); + return leave_state & (group->rec.scope_join_state & 0xf); } static int join_group(struct mcast_group *group, int slave, u8 join_mask) @@ -564,8 +564,8 @@ static void mlx4_ib_mcg_timeout_handler(struct work_struct *work) } else mcg_warn_group(group, "DRIVER BUG\n"); } else if (group->state == MCAST_LEAVE_SENT) { - if (group->rec.scope_join_state & 7) - group->rec.scope_join_state &= 0xf8; + if (group->rec.scope_join_state & 0xf) + group->rec.scope_join_state &= 0xf0; group->state = MCAST_IDLE; mutex_unlock(&group->lock); if (release_group(group, 1)) @@ -605,7 +605,7 @@ static int handle_leave_req(struct mcast_group *group, u8 leave_mask, static int handle_join_req(struct mcast_group *group, u8 join_mask, struct mcast_req *req) { - u8 group_join_state = group->rec.scope_join_state & 7; + u8 group_join_state = group->rec.scope_join_state & 0xf; int ref = 0; u16 status; struct ib_sa_mcmember_data *sa_data = (struct ib_sa_mcmember_data *)req->sa_mad.data; @@ -690,8 +690,8 @@ static void mlx4_ib_mcg_work_handler(struct work_struct *work) u8 cur_join_state; resp_join_state = ((struct ib_sa_mcmember_data *) - group->response_sa_mad.data)->scope_join_state & 7; - cur_join_state = group->rec.scope_join_state & 7; + group->response_sa_mad.data)->scope_join_state & 0xf; + cur_join_state = group->rec.scope_join_state & 0xf; if (method == IB_MGMT_METHOD_GET_RESP) { /* successfull join */ @@ -710,7 +710,7 @@ process_requests: req = list_first_entry(&group->pending_list, struct mcast_req, group_list); sa_data = (struct ib_sa_mcmember_data *)req->sa_mad.data; - req_join_state = sa_data->scope_join_state & 0x7; + req_join_state = sa_data->scope_join_state & 0xf; /* For a leave request, we will immediately answer the VF, and * update our internal counters. The actual leave will be sent -- cgit v1.2.3 From baa0be7026e2f7d1d40bfd45909044169e9e3c68 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Mon, 12 Sep 2016 19:16:19 +0300 Subject: IB/mlx4: Fix code indentation in QP1 MAD flow The indentation in the QP1 GRH flow in procedure build_mlx_header is really confusing. Fix it, in preparation for a commit which touches this code. Fixes: 1ffeb2eb8be9 ("IB/mlx4: SR-IOV IB context objects and proxy/tunnel SQP support") Signed-off-by: Jack Morgenstein Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/qp.c | 36 +++++++++++++++++++----------------- 1 file changed, 19 insertions(+), 17 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 768085f59566..e398c04903fa 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -2493,24 +2493,26 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_ud_wr *wr, sqp->ud_header.grh.flow_label = ah->av.ib.sl_tclass_flowlabel & cpu_to_be32(0xfffff); sqp->ud_header.grh.hop_limit = ah->av.ib.hop_limit; - if (is_eth) + if (is_eth) { memcpy(sqp->ud_header.grh.source_gid.raw, sgid.raw, 16); - else { - if (mlx4_is_mfunc(to_mdev(ib_dev)->dev)) { - /* When multi-function is enabled, the ib_core gid - * indexes don't necessarily match the hw ones, so - * we must use our own cache */ - sqp->ud_header.grh.source_gid.global.subnet_prefix = - to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1]. - subnet_prefix; - sqp->ud_header.grh.source_gid.global.interface_id = - to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1]. - guid_cache[ah->av.ib.gid_index]; - } else - ib_get_cached_gid(ib_dev, - be32_to_cpu(ah->av.ib.port_pd) >> 24, - ah->av.ib.gid_index, - &sqp->ud_header.grh.source_gid, NULL); + } else { + if (mlx4_is_mfunc(to_mdev(ib_dev)->dev)) { + /* When multi-function is enabled, the ib_core gid + * indexes don't necessarily match the hw ones, so + * we must use our own cache + */ + sqp->ud_header.grh.source_gid.global.subnet_prefix = + to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1]. + subnet_prefix; + sqp->ud_header.grh.source_gid.global.interface_id = + to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1]. + guid_cache[ah->av.ib.gid_index]; + } else { + ib_get_cached_gid(ib_dev, + be32_to_cpu(ah->av.ib.port_pd) >> 24, + ah->av.ib.gid_index, + &sqp->ud_header.grh.source_gid, NULL); + } } memcpy(sqp->ud_header.grh.destination_gid.raw, ah->av.ib.dgid, 16); -- cgit v1.2.3 From 8ec07bf8a8b57d6c58927a16a0a22c0115cf2855 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Mon, 12 Sep 2016 19:16:20 +0300 Subject: IB/mlx4: Use correct subnet-prefix in QP1 mads under SR-IOV When sending QP1 MAD packets which use a GRH, the source GID (which consists of the 64-bit subnet prefix, and the 64 bit port GUID) must be included in the packet GRH. For SR-IOV, a GID cache is used, since the source GID needs to be the slave's source GID, and not the Hypervisor's GID. This cache also included a subnet_prefix. Unfortunately, the subnet_prefix field in the cache was never initialized (to the default subnet prefix 0xfe80::0). As a result, this field remained all zeroes. Therefore, when SR-IOV was active, all QP1 packets which included a GRH had a source GID subnet prefix of all-zeroes. However, the subnet-prefix should initially be 0xfe80::0 (the default subnet prefix). In addition, if OpenSM modifies a port's subnet prefix, the new subnet prefix must be used in the GRH when sending QP1 packets. To fix this we now initialize the subnet prefix in the SR-IOV GID cache to the default subnet prefix. We update the cached value if/when OpenSM modifies the port's subnet prefix. We take this cached value when sending QP1 packets when SR-IOV is active. Note that the value is stored as an atomic64. This eliminates any need for locking when the subnet prefix is being updated. Note also that we depend on the FW generating the "port management change" event for tracking subnet-prefix changes performed by OpenSM. If running early FW (before 2.9.4630), subnet prefix changes will not be tracked (but the default subnet prefix still will be stored in the cache; therefore users who do not modify the subnet prefix will not have a problem). IF there is a need for such tracking also for early FW, we will add that capability in a subsequent patch. Fixes: 1ffeb2eb8be9 ("IB/mlx4: SR-IOV IB context objects and proxy/tunnel SQP support") Signed-off-by: Jack Morgenstein Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/mad.c | 23 +++++++++++++++++++++++ drivers/infiniband/hw/mlx4/mlx4_ib.h | 2 +- drivers/infiniband/hw/mlx4/qp.c | 5 +++-- 3 files changed, 27 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c index 9c2e53d28f98..0f21c3a25552 100644 --- a/drivers/infiniband/hw/mlx4/mad.c +++ b/drivers/infiniband/hw/mlx4/mad.c @@ -1128,6 +1128,27 @@ void handle_port_mgmt_change_event(struct work_struct *work) /* Generate GUID changed event */ if (changed_attr & MLX4_EQ_PORT_INFO_GID_PFX_CHANGE_MASK) { + if (mlx4_is_master(dev->dev)) { + union ib_gid gid; + int err = 0; + + if (!eqe->event.port_mgmt_change.params.port_info.gid_prefix) + err = __mlx4_ib_query_gid(&dev->ib_dev, port, 0, &gid, 1); + else + gid.global.subnet_prefix = + eqe->event.port_mgmt_change.params.port_info.gid_prefix; + if (err) { + pr_warn("Could not change QP1 subnet prefix for port %d: query_gid error (%d)\n", + port, err); + } else { + pr_debug("Changing QP1 subnet prefix for port %d. old=0x%llx. new=0x%llx\n", + port, + (u64)atomic64_read(&dev->sriov.demux[port - 1].subnet_prefix), + be64_to_cpu(gid.global.subnet_prefix)); + atomic64_set(&dev->sriov.demux[port - 1].subnet_prefix, + be64_to_cpu(gid.global.subnet_prefix)); + } + } mlx4_ib_dispatch_event(dev, port, IB_EVENT_GID_CHANGE); /*if master, notify all slaves*/ if (mlx4_is_master(dev->dev)) @@ -2202,6 +2223,8 @@ int mlx4_ib_init_sriov(struct mlx4_ib_dev *dev) if (err) goto demux_err; dev->sriov.demux[i].guid_cache[0] = gid.global.interface_id; + atomic64_set(&dev->sriov.demux[i].subnet_prefix, + be64_to_cpu(gid.global.subnet_prefix)); err = alloc_pv_object(dev, mlx4_master_func_num(dev->dev), i + 1, &dev->sriov.sqps[i]); if (err) diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index 7c5832ede4bd..686ab48ff644 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -448,7 +448,7 @@ struct mlx4_ib_demux_ctx { struct workqueue_struct *wq; struct workqueue_struct *ud_wq; spinlock_t ud_lock; - __be64 subnet_prefix; + atomic64_t subnet_prefix; __be64 guid_cache[128]; struct mlx4_ib_dev *dev; /* the following lock protects both mcg_table and mcg_mgid0_list */ diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index e398c04903fa..7fb9629bd12b 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -2502,8 +2502,9 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_ud_wr *wr, * we must use our own cache */ sqp->ud_header.grh.source_gid.global.subnet_prefix = - to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1]. - subnet_prefix; + cpu_to_be64(atomic64_read(&(to_mdev(ib_dev)->sriov. + demux[sqp->qp.port - 1]. + subnet_prefix))); sqp->ud_header.grh.source_gid.global.interface_id = to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1]. guid_cache[ah->av.ib.gid_index]; -- cgit v1.2.3 From 69d269d38910e697e461ec5677368f57d2046cbe Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Mon, 12 Sep 2016 19:16:22 +0300 Subject: IB/mlx4: Diagnostic HW counters are not supported in slave mode Modify the mlx4_ib_diag_counters() to avoid the following error in the hypervisor when the slave tries to query the hardware counters in SR-IOV mode. mlx4_core 0000:81:00.0: Unknown command:0x30 accepted from slave:1 Fixes: 3f85f2aaabf7 ("IB/mlx4: Add diagnostic hardware counters") Signed-off-by: Kamal Heib Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx4/main.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 2af44c2de262..87ba9bca4181 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -2202,6 +2202,9 @@ static int mlx4_ib_alloc_diag_counters(struct mlx4_ib_dev *ibdev) bool per_port = !!(ibdev->dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_DIAG_PER_PORT); + if (mlx4_is_slave(ibdev->dev)) + return 0; + for (i = 0; i < MLX4_DIAG_COUNTERS_TYPES; i++) { /* i == 1 means we are building port counters */ if (i && !per_port) -- cgit v1.2.3 From 7fae6655a0c897875bd34501ec092232b526d3e4 Mon Sep 17 00:00:00 2001 From: Noa Osherovich Date: Mon, 12 Sep 2016 19:16:23 +0300 Subject: IB/mlx5: Enable MAD_IFC commands for IB ports only MAD_IFC command is supported only for physical functions (PF) and when physical port is IB. The proposed fix enforces it. Fixes: d603c809ef91 ("IB/mlx5: Fix decision on using MAD_IFC") Reported-by: David Chang Signed-off-by: Noa Osherovich Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 8150ea372c53..0480b643fb9c 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -288,7 +288,9 @@ __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num, static int mlx5_use_mad_ifc(struct mlx5_ib_dev *dev) { - return !MLX5_CAP_GEN(dev->mdev, ib_virt); + if (MLX5_CAP_GEN(dev->mdev, port_type) == MLX5_CAP_PORT_TYPE_IB) + return !MLX5_CAP_GEN(dev->mdev, ib_virt); + return 0; } enum { -- cgit v1.2.3 From ee3da804ad1b1dd4c766199a6e8443542b0aaaef Mon Sep 17 00:00:00 2001 From: Maor Gottlieb Date: Mon, 12 Sep 2016 19:16:24 +0300 Subject: IB/mlx5: Set source mac address in FTE Set the source mac address in the FTE when L2 specification is provided. Fixes: 038d2ef87572 ('IB/mlx5: Add flow steering support') Signed-off-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/hw/mlx5/main.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'drivers/infiniband/hw') diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 0480b643fb9c..e19537cf44ab 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -1430,6 +1430,13 @@ static int parse_flow_attr(u32 *match_c, u32 *match_v, dmac_47_16), ib_spec->eth.val.dst_mac); + ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, outer_headers_c, + smac_47_16), + ib_spec->eth.mask.src_mac); + ether_addr_copy(MLX5_ADDR_OF(fte_match_set_lyr_2_4, outer_headers_v, + smac_47_16), + ib_spec->eth.val.src_mac); + if (ib_spec->eth.mask.vlan_tag) { MLX5_SET(fte_match_set_lyr_2_4, outer_headers_c, vlan_tag, 1); -- cgit v1.2.3