From aa19a88935f6d93b6b5468dc4b8ca4da9cb2ef29 Mon Sep 17 00:00:00 2001 From: Mark Bloch Date: Thu, 27 Oct 2016 16:36:27 +0300 Subject: [PATCH] IB/cm: Mark stale CM id's whenever the mad agent was unregistered commit 9db0ff53cb9b43ed75bacd42a89c1a0ab048b2b0 upstream. When there is a CM id object that has port assigned to it, it means that the cm-id asked for the specific port that it should go by it, but if that port was removed (hot-unplug event) the cm-id was not updated. In order to fix that the port keeps a list of all the cm-id's that are planning to go by it, whenever the port is removed it marks all of them as invalid. This commit fixes a kernel panic which happens when running traffic between guests and we force reboot a guest mid traffic, it triggers a kernel panic: Call Trace: [] ? panic+0xa7/0x16f [] ? oops_end+0xe4/0x100 [] ? no_context+0xfb/0x260 [] ? del_timer_sync+0x22/0x30 [] ? __bad_area_nosemaphore+0x125/0x1e0 [] ? process_timeout+0x0/0x10 [] ? bad_area_nosemaphore+0x13/0x20 [] ? __do_page_fault+0x31f/0x480 [] ? default_wake_function+0x0/0x20 [] ? free_msg+0x55/0x70 [mlx5_core] [] ? cmd_exec+0x124/0x840 [mlx5_core] [] ? find_busiest_group+0x244/0x9f0 [] ? do_page_fault+0x3e/0xa0 [] ? page_fault+0x25/0x30 [] ? cm_alloc_msg+0x35/0xc0 [ib_cm] [] ? ib_send_cm_dreq+0xb1/0x1e0 [ib_cm] [] ? cm_destroy_id+0x176/0x320 [ib_cm] [] ? ib_destroy_cm_id+0x10/0x20 [ib_cm] [] ? ipoib_cm_free_rx_reap_list+0xa7/0x110 [ib_ipoib] [] ? ipoib_cm_rx_reap+0x0/0x20 [ib_ipoib] [] ? ipoib_cm_rx_reap+0x15/0x20 [ib_ipoib] [] ? worker_thread+0x170/0x2a0 [] ? autoremove_wake_function+0x0/0x40 [] ? worker_thread+0x0/0x2a0 [] ? kthread+0x96/0xa0 [] ? child_rip+0xa/0x20 [] ? kthread+0x0/0xa0 [] ? child_rip+0x0/0x20 Fixes: a977049dacde ("[PATCH] IB: Add the kernel CM implementation") Signed-off-by: Mark Bloch Signed-off-by: Erez Shitrit Reviewed-by: Maor Gottlieb Signed-off-by: Leon Romanovsky Signed-off-by: Doug Ledford Signed-off-by: Willy Tarreau --- drivers/infiniband/core/cm.c | 127 ++++++++++++++++++++++++++++++----- 1 file changed, 111 insertions(+), 16 deletions(-) diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index c410217fbe89..951a4f6a3b11 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -79,6 +79,8 @@ static struct ib_cm { __be32 random_id_operand; struct list_head timewait_list; struct workqueue_struct *wq; + /* Sync on cm change port state */ + spinlock_t state_lock; } cm; /* Counter indexes ordered by attribute ID */ @@ -160,6 +162,8 @@ struct cm_port { struct ib_mad_agent *mad_agent; struct kobject port_obj; u8 port_num; + struct list_head cm_priv_prim_list; + struct list_head cm_priv_altr_list; struct cm_counter_group counter_group[CM_COUNTER_GROUPS]; }; @@ -237,6 +241,12 @@ struct cm_id_private { u8 service_timeout; u8 target_ack_delay; + struct list_head prim_list; + struct list_head altr_list; + /* Indicates that the send port mad is registered and av is set */ + int prim_send_port_not_ready; + int altr_send_port_not_ready; + struct list_head work_list; atomic_t work_count; }; @@ -255,19 +265,46 @@ static int cm_alloc_msg(struct cm_id_private *cm_id_priv, struct ib_mad_agent *mad_agent; struct ib_mad_send_buf *m; struct ib_ah *ah; + struct cm_av *av; + unsigned long flags, flags2; + int ret = 0; + /* don't let the port to be released till the agent is down */ + spin_lock_irqsave(&cm.state_lock, flags2); + spin_lock_irqsave(&cm.lock, flags); + if (!cm_id_priv->prim_send_port_not_ready) + av = &cm_id_priv->av; + else if (!cm_id_priv->altr_send_port_not_ready && + (cm_id_priv->alt_av.port)) + av = &cm_id_priv->alt_av; + else { + pr_info("%s: not valid CM id\n", __func__); + ret = -ENODEV; + spin_unlock_irqrestore(&cm.lock, flags); + goto out; + } + spin_unlock_irqrestore(&cm.lock, flags); + /* Make sure the port haven't released the mad yet */ mad_agent = cm_id_priv->av.port->mad_agent; - ah = ib_create_ah(mad_agent->qp->pd, &cm_id_priv->av.ah_attr); - if (IS_ERR(ah)) - return PTR_ERR(ah); + if (!mad_agent) { + pr_info("%s: not a valid MAD agent\n", __func__); + ret = -ENODEV; + goto out; + } + ah = ib_create_ah(mad_agent->qp->pd, &av->ah_attr); + if (IS_ERR(ah)) { + ret = PTR_ERR(ah); + goto out; + } m = ib_create_send_mad(mad_agent, cm_id_priv->id.remote_cm_qpn, - cm_id_priv->av.pkey_index, + av->pkey_index, 0, IB_MGMT_MAD_HDR, IB_MGMT_MAD_DATA, GFP_ATOMIC); if (IS_ERR(m)) { ib_destroy_ah(ah); - return PTR_ERR(m); + ret = PTR_ERR(m); + goto out; } /* Timeout set by caller if response is expected. */ @@ -277,7 +314,10 @@ static int cm_alloc_msg(struct cm_id_private *cm_id_priv, atomic_inc(&cm_id_priv->refcount); m->context[0] = cm_id_priv; *msg = m; - return 0; + +out: + spin_unlock_irqrestore(&cm.state_lock, flags2); + return ret; } static int cm_alloc_response_msg(struct cm_port *port, @@ -346,7 +386,8 @@ static void cm_init_av_for_response(struct cm_port *port, struct ib_wc *wc, grh, &av->ah_attr); } -static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av) +static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av, + struct cm_id_private *cm_id_priv) { struct cm_device *cm_dev; struct cm_port *port = NULL; @@ -376,7 +417,18 @@ static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av) ib_init_ah_from_path(cm_dev->ib_device, port->port_num, path, &av->ah_attr); av->timeout = path->packet_life_time + 1; - return 0; + + spin_lock_irqsave(&cm.lock, flags); + if (&cm_id_priv->av == av) + list_add_tail(&cm_id_priv->prim_list, &port->cm_priv_prim_list); + else if (&cm_id_priv->alt_av == av) + list_add_tail(&cm_id_priv->altr_list, &port->cm_priv_altr_list); + else + ret = -EINVAL; + + spin_unlock_irqrestore(&cm.lock, flags); + + return ret; } static int cm_alloc_id(struct cm_id_private *cm_id_priv) @@ -716,6 +768,8 @@ struct ib_cm_id *ib_create_cm_id(struct ib_device *device, spin_lock_init(&cm_id_priv->lock); init_completion(&cm_id_priv->comp); INIT_LIST_HEAD(&cm_id_priv->work_list); + INIT_LIST_HEAD(&cm_id_priv->prim_list); + INIT_LIST_HEAD(&cm_id_priv->altr_list); atomic_set(&cm_id_priv->work_count, -1); atomic_set(&cm_id_priv->refcount, 1); return &cm_id_priv->id; @@ -914,6 +968,15 @@ retest: break; } + spin_lock_irq(&cm.lock); + if (!list_empty(&cm_id_priv->altr_list) && + (!cm_id_priv->altr_send_port_not_ready)) + list_del(&cm_id_priv->altr_list); + if (!list_empty(&cm_id_priv->prim_list) && + (!cm_id_priv->prim_send_port_not_ready)) + list_del(&cm_id_priv->prim_list); + spin_unlock_irq(&cm.lock); + cm_free_id(cm_id->local_id); cm_deref_id(cm_id_priv); wait_for_completion(&cm_id_priv->comp); @@ -1137,12 +1200,13 @@ int ib_send_cm_req(struct ib_cm_id *cm_id, goto out; } - ret = cm_init_av_by_path(param->primary_path, &cm_id_priv->av); + ret = cm_init_av_by_path(param->primary_path, &cm_id_priv->av, + cm_id_priv); if (ret) goto error1; if (param->alternate_path) { ret = cm_init_av_by_path(param->alternate_path, - &cm_id_priv->alt_av); + &cm_id_priv->alt_av, cm_id_priv); if (ret) goto error1; } @@ -1562,7 +1626,8 @@ static int cm_req_handler(struct cm_work *work) cm_process_routed_req(req_msg, work->mad_recv_wc->wc); cm_format_paths_from_req(req_msg, &work->path[0], &work->path[1]); - ret = cm_init_av_by_path(&work->path[0], &cm_id_priv->av); + ret = cm_init_av_by_path(&work->path[0], &cm_id_priv->av, + cm_id_priv); if (ret) { ib_get_cached_gid(work->port->cm_dev->ib_device, work->port->port_num, 0, &work->path[0].sgid); @@ -1572,7 +1637,8 @@ static int cm_req_handler(struct cm_work *work) goto rejected; } if (req_msg->alt_local_lid) { - ret = cm_init_av_by_path(&work->path[1], &cm_id_priv->alt_av); + ret = cm_init_av_by_path(&work->path[1], &cm_id_priv->alt_av, + cm_id_priv); if (ret) { ib_send_cm_rej(cm_id, IB_CM_REJ_INVALID_ALT_GID, &work->path[0].sgid, @@ -2627,7 +2693,8 @@ int ib_send_cm_lap(struct ib_cm_id *cm_id, goto out; } - ret = cm_init_av_by_path(alternate_path, &cm_id_priv->alt_av); + ret = cm_init_av_by_path(alternate_path, &cm_id_priv->alt_av, + cm_id_priv); if (ret) goto out; cm_id_priv->alt_av.timeout = @@ -2739,7 +2806,8 @@ static int cm_lap_handler(struct cm_work *work) cm_init_av_for_response(work->port, work->mad_recv_wc->wc, work->mad_recv_wc->recv_buf.grh, &cm_id_priv->av); - cm_init_av_by_path(param->alternate_path, &cm_id_priv->alt_av); + cm_init_av_by_path(param->alternate_path, &cm_id_priv->alt_av, + cm_id_priv); ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); @@ -2931,7 +2999,7 @@ int ib_send_cm_sidr_req(struct ib_cm_id *cm_id, return -EINVAL; cm_id_priv = container_of(cm_id, struct cm_id_private, id); - ret = cm_init_av_by_path(param->path, &cm_id_priv->av); + ret = cm_init_av_by_path(param->path, &cm_id_priv->av, cm_id_priv); if (ret) goto out; @@ -3352,7 +3420,9 @@ out: static int cm_migrate(struct ib_cm_id *cm_id) { struct cm_id_private *cm_id_priv; + struct cm_av tmp_av; unsigned long flags; + int tmp_send_port_not_ready; int ret = 0; cm_id_priv = container_of(cm_id, struct cm_id_private, id); @@ -3361,7 +3431,14 @@ static int cm_migrate(struct ib_cm_id *cm_id) (cm_id->lap_state == IB_CM_LAP_UNINIT || cm_id->lap_state == IB_CM_LAP_IDLE)) { cm_id->lap_state = IB_CM_LAP_IDLE; + /* Swap address vector */ + tmp_av = cm_id_priv->av; cm_id_priv->av = cm_id_priv->alt_av; + cm_id_priv->alt_av = tmp_av; + /* Swap port send ready state */ + tmp_send_port_not_ready = cm_id_priv->prim_send_port_not_ready; + cm_id_priv->prim_send_port_not_ready = cm_id_priv->altr_send_port_not_ready; + cm_id_priv->altr_send_port_not_ready = tmp_send_port_not_ready; } else ret = -EINVAL; spin_unlock_irqrestore(&cm_id_priv->lock, flags); @@ -3767,6 +3844,9 @@ static void cm_add_one(struct ib_device *ib_device) port->cm_dev = cm_dev; port->port_num = i; + INIT_LIST_HEAD(&port->cm_priv_prim_list); + INIT_LIST_HEAD(&port->cm_priv_altr_list); + ret = cm_create_port_fs(port); if (ret) goto error1; @@ -3813,6 +3893,8 @@ static void cm_remove_one(struct ib_device *ib_device) { struct cm_device *cm_dev; struct cm_port *port; + struct cm_id_private *cm_id_priv; + struct ib_mad_agent *cur_mad_agent; struct ib_port_modify port_modify = { .clr_port_cap_mask = IB_PORT_CM_SUP }; @@ -3830,10 +3912,22 @@ static void cm_remove_one(struct ib_device *ib_device) for (i = 1; i <= ib_device->phys_port_cnt; i++) { port = cm_dev->port[i-1]; ib_modify_port(ib_device, port->port_num, 0, &port_modify); - ib_unregister_mad_agent(port->mad_agent); + /* Mark all the cm_id's as not valid */ + spin_lock_irq(&cm.lock); + list_for_each_entry(cm_id_priv, &port->cm_priv_altr_list, altr_list) + cm_id_priv->altr_send_port_not_ready = 1; + list_for_each_entry(cm_id_priv, &port->cm_priv_prim_list, prim_list) + cm_id_priv->prim_send_port_not_ready = 1; + spin_unlock_irq(&cm.lock); flush_workqueue(cm.wq); + spin_lock_irq(&cm.state_lock); + cur_mad_agent = port->mad_agent; + port->mad_agent = NULL; + spin_unlock_irq(&cm.state_lock); + ib_unregister_mad_agent(cur_mad_agent); cm_remove_port_fs(port); } + device_unregister(cm_dev->device); kfree(cm_dev); } @@ -3846,6 +3940,7 @@ static int __init ib_cm_init(void) INIT_LIST_HEAD(&cm.device_list); rwlock_init(&cm.device_lock); spin_lock_init(&cm.lock); + spin_lock_init(&cm.state_lock); cm.listen_service_table = RB_ROOT; cm.listen_service_id = be64_to_cpu(IB_CM_ASSIGN_SERVICE_ID); cm.remote_id_table = RB_ROOT; -- 2.20.1