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') 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') 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 5e102b3b4fa944815aead89e63eec2a35069a07b Mon Sep 17 00:00:00 2001 From: Alexey Khoroshilov Date: Fri, 2 Sep 2016 23:46:53 +0300 Subject: IB/rxe: fix GFP_KERNEL in spinlock context There is skb_clone(skb, GFP_KERNEL) in spinlock context in rxe_rcv_mcast_pkt(). Found by Linux Driver Verification project (linuxtesting.org). Signed-off-by: Alexey Khoroshilov Acked-by: Moni Shoua Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_recv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/sw/rxe/rxe_recv.c b/drivers/infiniband/sw/rxe/rxe_recv.c index 3d464c23e08b..144d2f129fcd 100644 --- a/drivers/infiniband/sw/rxe/rxe_recv.c +++ b/drivers/infiniband/sw/rxe/rxe_recv.c @@ -312,7 +312,7 @@ static void rxe_rcv_mcast_pkt(struct rxe_dev *rxe, struct sk_buff *skb) * make a copy of the skb to post to the next qp */ skb_copy = (mce->qp_list.next != &mcg->qp_list) ? - skb_clone(skb, GFP_KERNEL) : NULL; + skb_clone(skb, GFP_ATOMIC) : NULL; pkt->qp = qp; rxe_add_ref(qp); -- cgit v1.2.3 From 344bacca8cd811809fc33a249f2738ab757d327f Mon Sep 17 00:00:00 2001 From: Alex Vesker Date: Mon, 12 Sep 2016 09:55:28 +0300 Subject: IB/ipoib: Don't allow MC joins during light MC flush This fix solves a race between light flush and on the fly joins. Light flush doesn't set the device to down and unset IPOIB_OPER_UP flag, this means that if while flushing we have a MC join in progress and the QP was attached to BC MGID we can have a mismatches when re-attaching a QP to the BC MGID. The light flush would set the broadcast group to NULL causing an on the fly join to rejoin and reattach to the BC MCG as well as adding the BC MGID to the multicast list. The flush process would later on remove the BC MGID and detach it from the QP. On the next flush the BC MGID is present in the multicast list but not found when trying to detach it because of the previous double attach and single detach. [18332.714265] ------------[ cut here ]------------ [18332.717775] WARNING: CPU: 6 PID: 3767 at drivers/infiniband/core/verbs.c:280 ib_dealloc_pd+0xff/0x120 [ib_core] ... [18332.775198] Hardware name: Red Hat KVM, BIOS Bochs 01/01/2011 [18332.779411] 0000000000000000 ffff8800b50dfbb0 ffffffff813fed47 0000000000000000 [18332.784960] 0000000000000000 ffff8800b50dfbf0 ffffffff8109add1 0000011832f58300 [18332.790547] ffff880226a596c0 ffff880032482000 ffff880032482830 ffff880226a59280 [18332.796199] Call Trace: [18332.798015] [] dump_stack+0x63/0x8c [18332.801831] [] __warn+0xd1/0xf0 [18332.805403] [] warn_slowpath_null+0x1d/0x20 [18332.809706] [] ib_dealloc_pd+0xff/0x120 [ib_core] [18332.814384] [] ipoib_transport_dev_cleanup+0xfc/0x1d0 [ib_ipoib] [18332.820031] [] ipoib_ib_dev_cleanup+0x98/0x110 [ib_ipoib] [18332.825220] [] ipoib_dev_cleanup+0x2d8/0x550 [ib_ipoib] [18332.830290] [] ipoib_uninit+0x2f/0x40 [ib_ipoib] [18332.834911] [] rollback_registered_many+0x1aa/0x2c0 [18332.839741] [] rollback_registered+0x31/0x40 [18332.844091] [] unregister_netdevice_queue+0x48/0x80 [18332.848880] [] ipoib_vlan_delete+0x1fb/0x290 [ib_ipoib] [18332.853848] [] delete_child+0x7d/0xf0 [ib_ipoib] [18332.858474] [] dev_attr_store+0x18/0x30 [18332.862510] [] sysfs_kf_write+0x3a/0x50 [18332.866349] [] kernfs_fop_write+0x120/0x170 [18332.870471] [] __vfs_write+0x28/0xe0 [18332.874152] [] ? percpu_down_read+0x1f/0x50 [18332.878274] [] vfs_write+0xa2/0x1a0 [18332.881896] [] SyS_write+0x46/0xa0 [18332.885632] [] do_syscall_64+0x57/0xb0 [18332.889709] [] entry_SYSCALL64_slow_path+0x25/0x25 [18332.894727] ---[ end trace 09ebbe31f831ef17 ]--- Fixes: ee1e2c82c245 ("IPoIB: Refresh paths instead of flushing them on SM change events") Signed-off-by: Alex Vesker Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/ulp/ipoib/ipoib_ib.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index dc6d241b9406..be11d5d5b8c1 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -1161,8 +1161,17 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, } if (level == IPOIB_FLUSH_LIGHT) { + int oper_up; ipoib_mark_paths_invalid(dev); + /* Set IPoIB operation as down to prevent races between: + * the flush flow which leaves MCG and on the fly joins + * which can happen during that time. mcast restart task + * should deal with join requests we missed. + */ + oper_up = test_and_clear_bit(IPOIB_FLAG_OPER_UP, &priv->flags); ipoib_mcast_dev_flush(dev); + if (oper_up) + set_bit(IPOIB_FLAG_OPER_UP, &priv->flags); ipoib_flush_ah(dev); } -- 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') 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') 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') 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') 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') 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') 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 From dfdd6158ca2c8838313470c5ab79243d04da70b2 Mon Sep 17 00:00:00 2001 From: Yonatan Cohen Date: Wed, 7 Sep 2016 14:04:04 +0300 Subject: IB/rxe: Fix kernel panic in udp_setup_tunnel Disable creation of a UDP socket for ipv6 when CONFIG_IPV6 is not enabeld. Since udp_sock_create6() returns 0 when CONFIG_IPV6 is not set [ 46.888632] IP: [] setup_udp_tunnel_sock+0x6/0x4f [ 46.891355] *pdpt = 0000000000000000 *pde = f000ff53f000ff53 [ 46.893918] Oops: 0002 [#1] PREEMPT [ 46.896014] CPU: 0 PID: 1 Comm: swapper Not tainted 4.7.0-rc4-00001-g8700e3e #1 [ 46.900280] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Debian-1.8.2-1 04/01/2014 [ 46.904905] task: cf06c040 ti: cf05e000 task.ti: cf05e000 [ 46.907854] EIP: 0060:[] EFLAGS: 00210246 CPU: 0 [ 46.911137] EIP is at setup_udp_tunnel_sock+0x6/0x4f [ 46.914070] EAX: 00000044 EBX: 00000001 ECX: cf05fef0 EDX: ca8142e0 [ 46.917236] ESI: c2c4505b EDI: cf05fef0 EBP: cf05fed0 ESP: cf05fed0 [ 46.919836] DS: 007b ES: 007b FS: 0000 GS: 00e0 SS: 0068 [ 46.922046] CR0: 80050033 CR2: 000001fc CR3: 02cec000 CR4: 000006b0 [ 46.924550] Stack: [ 46.926014] cf05ff10 c1fd4657 ca8142e0 0000000a 00000000 00000000 0000b712 00000008 [ 46.931274] 00000000 6bb5bd01 c1fd48de 00000000 00000000 cf05ff1c 00000000 00000000 [ 46.936122] cf05ff1c c1fd4bdf 00000000 cf05ff28 c2c4507b ffffffff cf05ff88 c2bf1c74 [ 46.942350] Call Trace: [ 46.944403] [] rxe_setup_udp_tunnel+0x8f/0x99 [ 46.947689] [] ? net_to_rxe+0x4e/0x4e [ 46.950567] [] rxe_net_init+0xe/0xa4 [ 46.953147] [] rxe_module_init+0x20/0x4c [ 46.955448] [] do_one_initcall+0x89/0x113 [ 46.957797] [] ? set_debug_rodata+0xf/0xf [ 46.959966] [] ? kernel_init_freeable+0xbe/0x15b [ 46.962262] [] kernel_init_freeable+0xde/0x15b [ 46.964418] [] kernel_init+0x8/0xd0 [ 46.966618] [] ret_from_kernel_thread+0xe/0x24 [ 46.969592] [] ? rest_init+0x6f/0x6f Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe.c | 25 +++++++++++++++-- drivers/infiniband/sw/rxe/rxe_net.c | 55 +++++++++++++++++-------------------- drivers/infiniband/sw/rxe/rxe_net.h | 5 +++- 3 files changed, 51 insertions(+), 34 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/sw/rxe/rxe.c b/drivers/infiniband/sw/rxe/rxe.c index 55f0e8f0ca79..ddd59270ff6d 100644 --- a/drivers/infiniband/sw/rxe/rxe.c +++ b/drivers/infiniband/sw/rxe/rxe.c @@ -362,15 +362,34 @@ static int __init rxe_module_init(void) return err; } - err = rxe_net_init(); + err = rxe_net_ipv4_init(); if (err) { - pr_err("rxe: unable to init\n"); + pr_err("rxe: unable to init ipv4 tunnel\n"); rxe_cache_exit(); - return err; + goto exit; + } + + err = rxe_net_ipv6_init(); + if (err) { + pr_err("rxe: unable to init ipv6 tunnel\n"); + rxe_cache_exit(); + goto exit; } + + err = register_netdevice_notifier(&rxe_net_notifier); + if (err) { + pr_err("rxe: Failed to rigister netdev notifier\n"); + goto exit; + } + pr_info("rxe: loaded\n"); return 0; + +exit: + rxe_release_udp_tunnel(recv_sockets.sk4); + rxe_release_udp_tunnel(recv_sockets.sk6); + return err; } static void __exit rxe_module_exit(void) diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c index 0b8d2ea8b41d..eedf2f1cafdf 100644 --- a/drivers/infiniband/sw/rxe/rxe_net.c +++ b/drivers/infiniband/sw/rxe/rxe_net.c @@ -275,9 +275,10 @@ static struct socket *rxe_setup_udp_tunnel(struct net *net, __be16 port, return sock; } -static void rxe_release_udp_tunnel(struct socket *sk) +void rxe_release_udp_tunnel(struct socket *sk) { - udp_tunnel_sock_release(sk); + if (sk) + udp_tunnel_sock_release(sk); } static void prepare_udp_hdr(struct sk_buff *skb, __be16 src_port, @@ -658,51 +659,45 @@ out: return NOTIFY_OK; } -static struct notifier_block rxe_net_notifier = { +struct notifier_block rxe_net_notifier = { .notifier_call = rxe_notify, }; -int rxe_net_init(void) +int rxe_net_ipv4_init(void) { - int err; - spin_lock_init(&dev_list_lock); - recv_sockets.sk6 = rxe_setup_udp_tunnel(&init_net, - htons(ROCE_V2_UDP_DPORT), true); - if (IS_ERR(recv_sockets.sk6)) { - recv_sockets.sk6 = NULL; - pr_err("rxe: Failed to create IPv6 UDP tunnel\n"); - return -1; - } - recv_sockets.sk4 = rxe_setup_udp_tunnel(&init_net, - htons(ROCE_V2_UDP_DPORT), false); + htons(ROCE_V2_UDP_DPORT), false); if (IS_ERR(recv_sockets.sk4)) { - rxe_release_udp_tunnel(recv_sockets.sk6); recv_sockets.sk4 = NULL; - recv_sockets.sk6 = NULL; pr_err("rxe: Failed to create IPv4 UDP tunnel\n"); return -1; } - err = register_netdevice_notifier(&rxe_net_notifier); - if (err) { - rxe_release_udp_tunnel(recv_sockets.sk6); - rxe_release_udp_tunnel(recv_sockets.sk4); - pr_err("rxe: Failed to rigister netdev notifier\n"); - } - - return err; + return 0; } -void rxe_net_exit(void) +int rxe_net_ipv6_init(void) { - if (recv_sockets.sk6) - rxe_release_udp_tunnel(recv_sockets.sk6); +#if IS_ENABLED(CONFIG_IPV6) - if (recv_sockets.sk4) - rxe_release_udp_tunnel(recv_sockets.sk4); + spin_lock_init(&dev_list_lock); + recv_sockets.sk6 = rxe_setup_udp_tunnel(&init_net, + htons(ROCE_V2_UDP_DPORT), true); + if (IS_ERR(recv_sockets.sk6)) { + recv_sockets.sk6 = NULL; + pr_err("rxe: Failed to create IPv6 UDP tunnel\n"); + return -1; + } +#endif + return 0; +} + +void rxe_net_exit(void) +{ + rxe_release_udp_tunnel(recv_sockets.sk6); + rxe_release_udp_tunnel(recv_sockets.sk4); unregister_netdevice_notifier(&rxe_net_notifier); } diff --git a/drivers/infiniband/sw/rxe/rxe_net.h b/drivers/infiniband/sw/rxe/rxe_net.h index 7b06f76d16cc..0daf7f09e5b5 100644 --- a/drivers/infiniband/sw/rxe/rxe_net.h +++ b/drivers/infiniband/sw/rxe/rxe_net.h @@ -44,10 +44,13 @@ struct rxe_recv_sockets { }; extern struct rxe_recv_sockets recv_sockets; +extern struct notifier_block rxe_net_notifier; +void rxe_release_udp_tunnel(struct socket *sk); struct rxe_dev *rxe_net_add(struct net_device *ndev); -int rxe_net_init(void); +int rxe_net_ipv4_init(void); +int rxe_net_ipv6_init(void); void rxe_net_exit(void); #endif /* RXE_NET_H */ -- cgit v1.2.3 From 908948877bbbb240f67d7eb27d3a392beb394a07 Mon Sep 17 00:00:00 2001 From: Yonatan Cohen Date: Wed, 7 Sep 2016 14:04:05 +0300 Subject: IB/rxe: Fix duplicate atomic request handling When handling ack for atomic opcodes like "fetch&add" or "cmp&swp", the method send_atomic_ack() saves the ack before sending it, in case it gets lost and never reach the requester. In which case the method duplicate_request() will need to find it using the duplicated request.psn. But send_atomic_ack() used a wrong psn value and thus the above ack was never found. This fix uses the ack.psn to locate the ack in case its needed. This fix also copies the ack packet to the skb's control buffer since duplicate_request() will need it when calling rxe_xmit_packet() Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_resp.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c index ebb03b46e2ad..3e0f0f2baace 100644 --- a/drivers/infiniband/sw/rxe/rxe_resp.c +++ b/drivers/infiniband/sw/rxe/rxe_resp.c @@ -972,11 +972,13 @@ static int send_atomic_ack(struct rxe_qp *qp, struct rxe_pkt_info *pkt, free_rd_atomic_resource(qp, res); rxe_advance_resp_resource(qp); + memcpy(SKB_TO_PKT(skb), &ack_pkt, sizeof(skb->cb)); + res->type = RXE_ATOMIC_MASK; res->atomic.skb = skb; - res->first_psn = qp->resp.psn; - res->last_psn = qp->resp.psn; - res->cur_psn = qp->resp.psn; + res->first_psn = ack_pkt.psn; + res->last_psn = ack_pkt.psn; + res->cur_psn = ack_pkt.psn; rc = rxe_xmit_packet(rxe, qp, &ack_pkt, skb_copy); if (rc) { @@ -1116,8 +1118,7 @@ static enum resp_states duplicate_request(struct rxe_qp *qp, rc = RESPST_CLEANUP; goto out; } - bth_set_psn(SKB_TO_PKT(skb_copy), - qp->resp.psn - 1); + /* Resend the result. */ rc = rxe_xmit_packet(to_rdev(qp->ibqp.device), qp, pkt, skb_copy); -- cgit v1.2.3 From 3050b99850247695cb07a5c15265afcc08bcf400 Mon Sep 17 00:00:00 2001 From: Yonatan Cohen Date: Wed, 7 Sep 2016 14:04:06 +0300 Subject: IB/rxe: Fix race condition between requester and completer rxe_requester() is sending a pkt with rxe_xmit_packet() and then calls rxe_update() to update the wqe and qp's psn values. But sometimes the response is received before the requester had time to update the wqe in which case the completer acts on errornous wqe values. This fix updates the wqe and qp before actually sending the request and rolls back when xmit fails. Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_req.c | 57 ++++++++++++++++++++++++++++--------- 1 file changed, 44 insertions(+), 13 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c index 33b2d9d77021..13a848a518e8 100644 --- a/drivers/infiniband/sw/rxe/rxe_req.c +++ b/drivers/infiniband/sw/rxe/rxe_req.c @@ -511,24 +511,21 @@ static int fill_packet(struct rxe_qp *qp, struct rxe_send_wqe *wqe, } static void update_wqe_state(struct rxe_qp *qp, - struct rxe_send_wqe *wqe, - struct rxe_pkt_info *pkt, - enum wqe_state *prev_state) + struct rxe_send_wqe *wqe, + struct rxe_pkt_info *pkt) { - enum wqe_state prev_state_ = wqe->state; - if (pkt->mask & RXE_END_MASK) { if (qp_type(qp) == IB_QPT_RC) wqe->state = wqe_state_pending; } else { wqe->state = wqe_state_processing; } - - *prev_state = prev_state_; } -static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe, - struct rxe_pkt_info *pkt, int payload) +static void update_wqe_psn(struct rxe_qp *qp, + struct rxe_send_wqe *wqe, + struct rxe_pkt_info *pkt, + int payload) { /* number of packets left to send including current one */ int num_pkt = (wqe->dma.resid + payload + qp->mtu - 1) / qp->mtu; @@ -546,9 +543,34 @@ static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe, qp->req.psn = (wqe->first_psn + num_pkt) & BTH_PSN_MASK; else qp->req.psn = (qp->req.psn + 1) & BTH_PSN_MASK; +} - qp->req.opcode = pkt->opcode; +static void save_state(struct rxe_send_wqe *wqe, + struct rxe_qp *qp, + struct rxe_send_wqe *rollback_wqe, + struct rxe_qp *rollback_qp) +{ + rollback_wqe->state = wqe->state; + rollback_wqe->first_psn = wqe->first_psn; + rollback_wqe->last_psn = wqe->last_psn; + rollback_qp->req.psn = qp->req.psn; +} +static void rollback_state(struct rxe_send_wqe *wqe, + struct rxe_qp *qp, + struct rxe_send_wqe *rollback_wqe, + struct rxe_qp *rollback_qp) +{ + wqe->state = rollback_wqe->state; + wqe->first_psn = rollback_wqe->first_psn; + wqe->last_psn = rollback_wqe->last_psn; + qp->req.psn = rollback_qp->req.psn; +} + +static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe, + struct rxe_pkt_info *pkt, int payload) +{ + qp->req.opcode = pkt->opcode; if (pkt->mask & RXE_END_MASK) qp->req.wqe_index = next_index(qp->sq.queue, qp->req.wqe_index); @@ -571,7 +593,8 @@ int rxe_requester(void *arg) int mtu; int opcode; int ret; - enum wqe_state prev_state; + struct rxe_qp rollback_qp; + struct rxe_send_wqe rollback_wqe; next_wqe: if (unlikely(!qp->valid || qp->req.state == QP_STATE_ERROR)) @@ -688,13 +711,21 @@ next_wqe: goto err; } - update_wqe_state(qp, wqe, &pkt, &prev_state); + /* + * To prevent a race on wqe access between requester and completer, + * wqe members state and psn need to be set before calling + * rxe_xmit_packet(). + * Otherwise, completer might initiate an unjustified retry flow. + */ + save_state(wqe, qp, &rollback_wqe, &rollback_qp); + update_wqe_state(qp, wqe, &pkt); + update_wqe_psn(qp, wqe, &pkt, payload); ret = rxe_xmit_packet(to_rdev(qp->ibqp.device), qp, &pkt, skb); if (ret) { qp->need_req_skb = 1; kfree_skb(skb); - wqe->state = prev_state; + rollback_state(wqe, qp, &rollback_wqe, &rollback_qp); if (ret == -EAGAIN) { rxe_run_task(&qp->req.task, 1); -- cgit v1.2.3 From c1cc72cb6ff95cbe02ed747866576dccab4a5911 Mon Sep 17 00:00:00 2001 From: Yonatan Cohen Date: Wed, 7 Sep 2016 14:04:07 +0300 Subject: IB/rxe: Fix kmem_cache leak Decrement qp reference when handling error path in completer to prevent kmem_cache leak. Fixes: 8700e3e7c485 ("Soft RoCE driver") Signed-off-by: Yonatan Cohen Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rxe/rxe_comp.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/sw/rxe/rxe_comp.c b/drivers/infiniband/sw/rxe/rxe_comp.c index 36f67de44095..1c59ef2c67aa 100644 --- a/drivers/infiniband/sw/rxe/rxe_comp.c +++ b/drivers/infiniband/sw/rxe/rxe_comp.c @@ -689,7 +689,14 @@ int rxe_completer(void *arg) qp->req.need_retry = 1; rxe_run_task(&qp->req.task, 1); } + + if (pkt) { + rxe_drop_ref(pkt->qp); + kfree_skb(skb); + } + goto exit; + } else { wqe->status = IB_WC_RETRY_EXC_ERR; state = COMPST_ERROR; @@ -716,6 +723,12 @@ int rxe_completer(void *arg) case COMPST_ERROR: do_complete(qp, wqe); rxe_qp_error(qp); + + if (pkt) { + rxe_drop_ref(pkt->qp); + kfree_skb(skb); + } + goto exit; } } -- cgit v1.2.3 From e4618d40eb3dc1a6d1f55f7150ea25bb23ab410a Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Fri, 9 Sep 2016 08:15:37 +0100 Subject: IB/rdmavt: Don't vfree a kzalloc'ed memory region The userspace memory region 'mr' is allocated with kzalloc in __rvt_alloc_mr however it is incorrectly being freed with vfree in __rvt_free_mr. Fix this by using kfree to free it. Signed-off-by: Colin Ian King Reviewed-by: Leon Romanovsky Acked-by: Dennis Dalessandro Signed-off-by: Doug Ledford --- drivers/infiniband/sw/rdmavt/mr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/sw/rdmavt/mr.c b/drivers/infiniband/sw/rdmavt/mr.c index 80c4b6b401b8..46b64970058e 100644 --- a/drivers/infiniband/sw/rdmavt/mr.c +++ b/drivers/infiniband/sw/rdmavt/mr.c @@ -294,7 +294,7 @@ static void __rvt_free_mr(struct rvt_mr *mr) { rvt_deinit_mregion(&mr->mr); rvt_free_lkey(&mr->mr); - vfree(mr); + kfree(mr); } /** -- cgit v1.2.3