From a2351c5d86d7acf8eef17fba4ac1fc5b305a37c0 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Wed, 9 Oct 2019 10:07:43 +0200 Subject: net/smc: separate SMCD and SMCR link group lists Currently SMCD and SMCR link groups are maintained in one list. To facilitate abnormal termination handling they are split into a separate list for SMCR link groups and separate lists for SMCD link groups per SMCD device. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 24 +++++++++++++++++------- net/smc/smc_ism.c | 1 + 2 files changed, 18 insertions(+), 7 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 4ca50ddf8d16..46d679542b87 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -198,6 +198,7 @@ static void smc_lgr_free_work(struct work_struct *work) static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) { struct smc_link_group *lgr; + struct list_head *lgr_list; struct smc_link *lnk; u8 rndvec[3]; int rc = 0; @@ -233,6 +234,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) /* SMC-D specific settings */ lgr->peer_gid = ini->ism_gid; lgr->smcd = ini->ism_dev; + lgr_list = &ini->ism_dev->lgr_list; } else { /* SMC-R specific settings */ lgr->role = smc->listen_smc ? SMC_SERV : SMC_CLNT; @@ -245,6 +247,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) lnk->link_id = SMC_SINGLE_LINK; lnk->smcibdev = ini->ib_dev; lnk->ibport = ini->ib_port; + lgr_list = &smc_lgr_list.list; lnk->path_mtu = ini->ib_dev->pattr[ini->ib_port - 1].active_mtu; if (!ini->ib_dev->initialized) @@ -275,7 +278,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) } smc->conn.lgr = lgr; spin_lock_bh(&smc_lgr_list.lock); - list_add(&lgr->list, &smc_lgr_list.list); + list_add(&lgr->list, lgr_list); spin_unlock_bh(&smc_lgr_list.lock); return 0; @@ -512,9 +515,8 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) /* run common cleanup function and build free list */ spin_lock_bh(&smc_lgr_list.lock); - list_for_each_entry_safe(lgr, l, &smc_lgr_list.list, list) { - if (lgr->is_smcd && lgr->smcd == dev && - (!peer_gid || lgr->peer_gid == peer_gid) && + list_for_each_entry_safe(lgr, l, &dev->lgr_list, list) { + if ((!peer_gid || lgr->peer_gid == peer_gid) && (vlan == VLAN_VID_MASK || lgr->vlan_id == vlan)) { __smc_lgr_terminate(lgr); list_move(&lgr->list, &lgr_free_list); @@ -604,10 +606,12 @@ static bool smcd_lgr_match(struct smc_link_group *lgr, int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini) { struct smc_connection *conn = &smc->conn; + struct list_head *lgr_list; struct smc_link_group *lgr; enum smc_lgr_role role; int rc = 0; + lgr_list = ini->is_smcd ? &ini->ism_dev->lgr_list : &smc_lgr_list.list; ini->cln_first_contact = SMC_FIRST_CONTACT; role = smc->listen_smc ? SMC_SERV : SMC_CLNT; if (role == SMC_CLNT && ini->srv_first_contact) @@ -616,7 +620,7 @@ int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini) /* determine if an existing link group can be reused */ spin_lock_bh(&smc_lgr_list.lock); - list_for_each_entry(lgr, &smc_lgr_list.list, list) { + list_for_each_entry(lgr, lgr_list, list) { write_lock_bh(&lgr->conns_lock); if ((ini->is_smcd ? smcd_lgr_match(lgr, ini->ism_dev, ini->ism_gid) : @@ -1029,11 +1033,17 @@ void smc_core_exit(void) { struct smc_link_group *lgr, *lg; LIST_HEAD(lgr_freeing_list); + struct smcd_dev *smcd; spin_lock_bh(&smc_lgr_list.lock); - if (!list_empty(&smc_lgr_list.list)) - list_splice_init(&smc_lgr_list.list, &lgr_freeing_list); + list_splice_init(&smc_lgr_list.list, &lgr_freeing_list); spin_unlock_bh(&smc_lgr_list.lock); + + spin_lock(&smcd_dev_list.lock); + list_for_each_entry(smcd, &smcd_dev_list.list, list) + list_splice_init(&smcd->lgr_list, &lgr_freeing_list); + spin_unlock(&smcd_dev_list.lock); + list_for_each_entry_safe(lgr, lg, &lgr_freeing_list, list) { list_del_init(&lgr->list); if (!lgr->is_smcd) { diff --git a/net/smc/smc_ism.c b/net/smc/smc_ism.c index e89e918b88e0..674eb5ae2320 100644 --- a/net/smc/smc_ism.c +++ b/net/smc/smc_ism.c @@ -287,6 +287,7 @@ struct smcd_dev *smcd_alloc_dev(struct device *parent, const char *name, spin_lock_init(&smcd->lock); INIT_LIST_HEAD(&smcd->vlan); + INIT_LIST_HEAD(&smcd->lgr_list); smcd->event_wq = alloc_ordered_workqueue("ism_evt_wq-%s)", WQ_MEM_RECLAIM, name); if (!smcd->event_wq) { -- cgit v1.2.3 From a0a62ee15a829ebf8aeec55a4f1688230439b3e0 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Wed, 9 Oct 2019 10:07:44 +0200 Subject: net/smc: separate locks for SMCD and SMCR link group lists This patch introduces separate locks for the split SMCD and SMCR link group lists. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 57 ++++++++++++++++++++++++++++++++++++++++-------------- net/smc/smc_ism.c | 1 + 2 files changed, 43 insertions(+), 15 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 46d679542b87..949b1914e11a 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -42,6 +42,19 @@ static struct smc_lgr_list smc_lgr_list = { /* established link groups */ static void smc_buf_free(struct smc_link_group *lgr, bool is_rmb, struct smc_buf_desc *buf_desc); +/* return head of link group list and its lock for a given link group */ +static inline struct list_head *smc_lgr_list_head(struct smc_link_group *lgr, + spinlock_t **lgr_lock) +{ + if (lgr->is_smcd) { + *lgr_lock = &lgr->smcd->lgr_lock; + return &lgr->smcd->lgr_list; + } + + *lgr_lock = &smc_lgr_list.lock; + return &smc_lgr_list.list; +} + static void smc_lgr_schedule_free_work(struct smc_link_group *lgr) { /* client link group creation always follows the server link group @@ -157,19 +170,21 @@ static void smc_lgr_free_work(struct work_struct *work) struct smc_link_group *lgr = container_of(to_delayed_work(work), struct smc_link_group, free_work); + spinlock_t *lgr_lock; bool conns; - spin_lock_bh(&smc_lgr_list.lock); + smc_lgr_list_head(lgr, &lgr_lock); + spin_lock_bh(lgr_lock); read_lock_bh(&lgr->conns_lock); conns = RB_EMPTY_ROOT(&lgr->conns_all); read_unlock_bh(&lgr->conns_lock); if (!conns) { /* number of lgr connections is no longer zero */ - spin_unlock_bh(&smc_lgr_list.lock); + spin_unlock_bh(lgr_lock); return; } if (!list_empty(&lgr->list)) list_del_init(&lgr->list); /* remove from smc_lgr_list */ - spin_unlock_bh(&smc_lgr_list.lock); + spin_unlock_bh(lgr_lock); if (!lgr->is_smcd && !lgr->terminating) { struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK]; @@ -200,6 +215,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) struct smc_link_group *lgr; struct list_head *lgr_list; struct smc_link *lnk; + spinlock_t *lgr_lock; u8 rndvec[3]; int rc = 0; int i; @@ -235,6 +251,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) lgr->peer_gid = ini->ism_gid; lgr->smcd = ini->ism_dev; lgr_list = &ini->ism_dev->lgr_list; + lgr_lock = &lgr->smcd->lgr_lock; } else { /* SMC-R specific settings */ lgr->role = smc->listen_smc ? SMC_SERV : SMC_CLNT; @@ -248,6 +265,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) lnk->smcibdev = ini->ib_dev; lnk->ibport = ini->ib_port; lgr_list = &smc_lgr_list.list; + lgr_lock = &smc_lgr_list.lock; lnk->path_mtu = ini->ib_dev->pattr[ini->ib_port - 1].active_mtu; if (!ini->ib_dev->initialized) @@ -277,9 +295,9 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) goto destroy_qp; } smc->conn.lgr = lgr; - spin_lock_bh(&smc_lgr_list.lock); + spin_lock_bh(lgr_lock); list_add(&lgr->list, lgr_list); - spin_unlock_bh(&smc_lgr_list.lock); + spin_unlock_bh(lgr_lock); return 0; destroy_qp: @@ -442,11 +460,15 @@ static void smc_lgr_free(struct smc_link_group *lgr) void smc_lgr_forget(struct smc_link_group *lgr) { - spin_lock_bh(&smc_lgr_list.lock); + struct list_head *lgr_list; + spinlock_t *lgr_lock; + + lgr_list = smc_lgr_list_head(lgr, &lgr_lock); + spin_lock_bh(lgr_lock); /* do not use this link group for new connections */ - if (!list_empty(&lgr->list)) - list_del_init(&lgr->list); - spin_unlock_bh(&smc_lgr_list.lock); + if (!list_empty(lgr_list)) + list_del_init(lgr_list); + spin_unlock_bh(lgr_lock); } /* terminate linkgroup abnormally */ @@ -487,9 +509,12 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) void smc_lgr_terminate(struct smc_link_group *lgr) { - spin_lock_bh(&smc_lgr_list.lock); + spinlock_t *lgr_lock; + + smc_lgr_list_head(lgr, &lgr_lock); + spin_lock_bh(lgr_lock); __smc_lgr_terminate(lgr); - spin_unlock_bh(&smc_lgr_list.lock); + spin_unlock_bh(lgr_lock); } /* Called when IB port is terminated */ @@ -514,7 +539,7 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) LIST_HEAD(lgr_free_list); /* run common cleanup function and build free list */ - spin_lock_bh(&smc_lgr_list.lock); + spin_lock_bh(&dev->lgr_lock); list_for_each_entry_safe(lgr, l, &dev->lgr_list, list) { if ((!peer_gid || lgr->peer_gid == peer_gid) && (vlan == VLAN_VID_MASK || lgr->vlan_id == vlan)) { @@ -522,7 +547,7 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) list_move(&lgr->list, &lgr_free_list); } } - spin_unlock_bh(&smc_lgr_list.lock); + spin_unlock_bh(&dev->lgr_lock); /* cancel the regular free workers and actually free lgrs */ list_for_each_entry_safe(lgr, l, &lgr_free_list, list) { @@ -609,9 +634,11 @@ int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini) struct list_head *lgr_list; struct smc_link_group *lgr; enum smc_lgr_role role; + spinlock_t *lgr_lock; int rc = 0; lgr_list = ini->is_smcd ? &ini->ism_dev->lgr_list : &smc_lgr_list.list; + lgr_lock = ini->is_smcd ? &ini->ism_dev->lgr_lock : &smc_lgr_list.lock; ini->cln_first_contact = SMC_FIRST_CONTACT; role = smc->listen_smc ? SMC_SERV : SMC_CLNT; if (role == SMC_CLNT && ini->srv_first_contact) @@ -619,7 +646,7 @@ int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini) goto create; /* determine if an existing link group can be reused */ - spin_lock_bh(&smc_lgr_list.lock); + spin_lock_bh(lgr_lock); list_for_each_entry(lgr, lgr_list, list) { write_lock_bh(&lgr->conns_lock); if ((ini->is_smcd ? @@ -640,7 +667,7 @@ int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini) } write_unlock_bh(&lgr->conns_lock); } - spin_unlock_bh(&smc_lgr_list.lock); + spin_unlock_bh(lgr_lock); if (role == SMC_CLNT && !ini->srv_first_contact && ini->cln_first_contact == SMC_FIRST_CONTACT) { diff --git a/net/smc/smc_ism.c b/net/smc/smc_ism.c index 674eb5ae2320..34dc619655e8 100644 --- a/net/smc/smc_ism.c +++ b/net/smc/smc_ism.c @@ -286,6 +286,7 @@ struct smcd_dev *smcd_alloc_dev(struct device *parent, const char *name, smc_pnetid_by_dev_port(parent, 0, smcd->pnetid); spin_lock_init(&smcd->lock); + spin_lock_init(&smcd->lgr_lock); INIT_LIST_HEAD(&smcd->vlan); INIT_LIST_HEAD(&smcd->lgr_list); smcd->event_wq = alloc_ordered_workqueue("ism_evt_wq-%s)", -- cgit v1.2.3 From b3cb53c05f20c5b4026a36a7bbd3010d1f3e0a55 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Wed, 9 Oct 2019 10:07:45 +0200 Subject: net/smc: increase device refcount for added link group SMCD link groups belong to certain ISM-devices and SMCR link group links belong to certain IB-devices. Increase the refcount for these devices, as long as corresponding link groups exist. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 949b1914e11a..a07fbf56c929 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -248,12 +248,14 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) lgr->conns_all = RB_ROOT; if (ini->is_smcd) { /* SMC-D specific settings */ + get_device(&ini->ism_dev->dev); lgr->peer_gid = ini->ism_gid; lgr->smcd = ini->ism_dev; lgr_list = &ini->ism_dev->lgr_list; lgr_lock = &lgr->smcd->lgr_lock; } else { /* SMC-R specific settings */ + get_device(&ini->ib_dev->ibdev->dev); lgr->role = smc->listen_smc ? SMC_SERV : SMC_CLNT; memcpy(lgr->peer_systemid, ini->ib_lcl->id_for_peer, SMC_SYSTEMID_LEN); @@ -451,10 +453,13 @@ static void smc_lgr_free_bufs(struct smc_link_group *lgr) static void smc_lgr_free(struct smc_link_group *lgr) { smc_lgr_free_bufs(lgr); - if (lgr->is_smcd) + if (lgr->is_smcd) { smc_ism_put_vlan(lgr->smcd, lgr->vlan_id); - else + put_device(&lgr->smcd->dev); + } else { smc_link_clear(&lgr->lnk[SMC_SINGLE_LINK]); + put_device(&lgr->lnk[SMC_SINGLE_LINK].smcibdev->ibdev->dev); + } kfree(lgr); } -- cgit v1.2.3 From c3d9494e68c4a5d23227ede822fda9bd68bef8e3 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Wed, 9 Oct 2019 10:07:46 +0200 Subject: net/smc: no new connections on disappearing devices Add a "going_away" indication to ISM devices and IB ports and avoid creation of new connections on such disappearing devices. And do not handle ISM events if ISM device is disappearing. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 23 +++++++++++++++++++++++ net/smc/smc_ib.c | 15 +++++++++++++-- net/smc/smc_ib.h | 1 + net/smc/smc_ism.c | 3 +++ net/smc/smc_pnet.c | 5 ++++- 5 files changed, 44 insertions(+), 3 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index a07fbf56c929..5862784eedd4 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -1060,6 +1060,27 @@ int smc_rmb_rtoken_handling(struct smc_connection *conn, return 0; } +static void smc_core_going_away(void) +{ + struct smc_ib_device *smcibdev; + struct smcd_dev *smcd; + + spin_lock(&smc_ib_devices.lock); + list_for_each_entry(smcibdev, &smc_ib_devices.list, list) { + int i; + + for (i = 0; i < SMC_MAX_PORTS; i++) + set_bit(i, smcibdev->ports_going_away); + } + spin_unlock(&smc_ib_devices.lock); + + spin_lock(&smcd_dev_list.lock); + list_for_each_entry(smcd, &smcd_dev_list.list, list) { + smcd->going_away = 1; + } + spin_unlock(&smcd_dev_list.lock); +} + /* Called (from smc_exit) when module is removed */ void smc_core_exit(void) { @@ -1067,6 +1088,8 @@ void smc_core_exit(void) LIST_HEAD(lgr_freeing_list); struct smcd_dev *smcd; + smc_core_going_away(); + spin_lock_bh(&smc_lgr_list.lock); list_splice_init(&smc_lgr_list.list, &lgr_freeing_list); spin_unlock_bh(&smc_lgr_list.lock); diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c index d14ca4af6f94..af05daeb0538 100644 --- a/net/smc/smc_ib.c +++ b/net/smc/smc_ib.c @@ -242,8 +242,12 @@ static void smc_ib_port_event_work(struct work_struct *work) for_each_set_bit(port_idx, &smcibdev->port_event_mask, SMC_MAX_PORTS) { smc_ib_remember_port_attr(smcibdev, port_idx + 1); clear_bit(port_idx, &smcibdev->port_event_mask); - if (!smc_ib_port_active(smcibdev, port_idx + 1)) + if (!smc_ib_port_active(smcibdev, port_idx + 1)) { + set_bit(port_idx, smcibdev->ports_going_away); smc_port_terminate(smcibdev, port_idx + 1); + } else { + clear_bit(port_idx, smcibdev->ports_going_away); + } } } @@ -259,8 +263,10 @@ static void smc_ib_global_event_handler(struct ib_event_handler *handler, switch (ibevent->event) { case IB_EVENT_DEVICE_FATAL: /* terminate all ports on device */ - for (port_idx = 0; port_idx < SMC_MAX_PORTS; port_idx++) + for (port_idx = 0; port_idx < SMC_MAX_PORTS; port_idx++) { set_bit(port_idx, &smcibdev->port_event_mask); + set_bit(port_idx, smcibdev->ports_going_away); + } schedule_work(&smcibdev->port_event_work); break; case IB_EVENT_PORT_ERR: @@ -269,6 +275,10 @@ static void smc_ib_global_event_handler(struct ib_event_handler *handler, port_idx = ibevent->element.port_num - 1; if (port_idx < SMC_MAX_PORTS) { set_bit(port_idx, &smcibdev->port_event_mask); + if (ibevent->event == IB_EVENT_PORT_ERR) + set_bit(port_idx, smcibdev->ports_going_away); + else if (ibevent->event == IB_EVENT_PORT_ACTIVE) + clear_bit(port_idx, smcibdev->ports_going_away); schedule_work(&smcibdev->port_event_work); } break; @@ -307,6 +317,7 @@ static void smc_ib_qp_event_handler(struct ib_event *ibevent, void *priv) port_idx = ibevent->element.qp->port - 1; if (port_idx < SMC_MAX_PORTS) { set_bit(port_idx, &smcibdev->port_event_mask); + set_bit(port_idx, smcibdev->ports_going_away); schedule_work(&smcibdev->port_event_work); } break; diff --git a/net/smc/smc_ib.h b/net/smc/smc_ib.h index da60ab9e8d70..6a0069db6cae 100644 --- a/net/smc/smc_ib.h +++ b/net/smc/smc_ib.h @@ -47,6 +47,7 @@ struct smc_ib_device { /* ib-device infos for smc */ u8 initialized : 1; /* ib dev CQ, evthdl done */ struct work_struct port_event_work; unsigned long port_event_mask; + DECLARE_BITMAP(ports_going_away, SMC_MAX_PORTS); }; struct smc_buf_desc; diff --git a/net/smc/smc_ism.c b/net/smc/smc_ism.c index 34dc619655e8..ee7340898cb4 100644 --- a/net/smc/smc_ism.c +++ b/net/smc/smc_ism.c @@ -315,6 +315,7 @@ void smcd_unregister_dev(struct smcd_dev *smcd) spin_lock(&smcd_dev_list.lock); list_del(&smcd->list); spin_unlock(&smcd_dev_list.lock); + smcd->going_away = 1; flush_workqueue(smcd->event_wq); destroy_workqueue(smcd->event_wq); smc_smcd_terminate(smcd, 0, VLAN_VID_MASK); @@ -344,6 +345,8 @@ void smcd_handle_event(struct smcd_dev *smcd, struct smcd_event *event) { struct smc_ism_event_work *wrk; + if (smcd->going_away) + return; /* copy event to event work queue, and let it be handled there */ wrk = kmalloc(sizeof(*wrk), GFP_ATOMIC); if (!wrk) diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c index bab2da8cf17a..6b7799b3f5ca 100644 --- a/net/smc/smc_pnet.c +++ b/net/smc/smc_pnet.c @@ -781,6 +781,7 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, dev_put(ndev); if (netdev == ndev && smc_ib_port_active(ibdev, i) && + !test_bit(i - 1, ibdev->ports_going_away) && !smc_ib_determine_gid(ibdev, i, ini->vlan_id, ini->ib_gid, NULL)) { ini->ib_dev = ibdev; @@ -820,6 +821,7 @@ static void smc_pnet_find_roce_by_pnetid(struct net_device *ndev, continue; if (smc_pnet_match(ibdev->pnetid[i - 1], ndev_pnetid) && smc_ib_port_active(ibdev, i) && + !test_bit(i - 1, ibdev->ports_going_away) && !smc_ib_determine_gid(ibdev, i, ini->vlan_id, ini->ib_gid, NULL)) { ini->ib_dev = ibdev; @@ -846,7 +848,8 @@ static void smc_pnet_find_ism_by_pnetid(struct net_device *ndev, spin_lock(&smcd_dev_list.lock); list_for_each_entry(ismdev, &smcd_dev_list.list, list) { - if (smc_pnet_match(ismdev->pnetid, ndev_pnetid)) { + if (smc_pnet_match(ismdev->pnetid, ndev_pnetid) && + !ismdev->going_away) { ini->ism_dev = ismdev; break; } -- cgit v1.2.3 From d18963cf036566690c8bfd8b1d97d69f9a7d130f Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Wed, 9 Oct 2019 10:07:47 +0200 Subject: net/smc: improve close of terminated socket Make sure a terminated SMC socket reaches the CLOSED state. Even if sending of close flags fails, change the socket state to the intended state to avoid dangling sockets not reaching the CLOSED state. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_close.c | 40 +++++++++------------------------------- 1 file changed, 9 insertions(+), 31 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c index fc06720b53c1..1a858e59fc31 100644 --- a/net/smc/smc_close.c +++ b/net/smc/smc_close.c @@ -65,8 +65,8 @@ static void smc_close_stream_wait(struct smc_sock *smc, long timeout) rc = sk_wait_event(sk, &timeout, !smc_tx_prepared_sends(&smc->conn) || - (sk->sk_err == ECONNABORTED) || - (sk->sk_err == ECONNRESET), + sk->sk_err == ECONNABORTED || + sk->sk_err == ECONNRESET, &wait); if (rc) break; @@ -113,9 +113,6 @@ static void smc_close_active_abort(struct smc_sock *smc) { struct sock *sk = &smc->sk; - struct smc_cdc_conn_state_flags *txflags = - &smc->conn.local_tx_ctrl.conn_state_flags; - if (sk->sk_state != SMC_INIT && smc->clcsock && smc->clcsock->sk) { sk->sk_err = ECONNABORTED; if (smc->clcsock && smc->clcsock->sk) { @@ -129,35 +126,26 @@ static void smc_close_active_abort(struct smc_sock *smc) release_sock(sk); cancel_delayed_work_sync(&smc->conn.tx_work); lock_sock(sk); + sk->sk_state = SMC_CLOSED; sock_put(sk); /* passive closing */ break; case SMC_APPCLOSEWAIT1: case SMC_APPCLOSEWAIT2: - if (!smc_cdc_rxed_any_close(&smc->conn)) - sk->sk_state = SMC_PEERABORTWAIT; - else - sk->sk_state = SMC_CLOSED; release_sock(sk); cancel_delayed_work_sync(&smc->conn.tx_work); lock_sock(sk); + sk->sk_state = SMC_CLOSED; break; case SMC_PEERCLOSEWAIT1: case SMC_PEERCLOSEWAIT2: - if (!txflags->peer_conn_closed) { - /* just SHUTDOWN_SEND done */ - sk->sk_state = SMC_PEERABORTWAIT; - } else { - sk->sk_state = SMC_CLOSED; - } + case SMC_PEERFINCLOSEWAIT: + sk->sk_state = SMC_CLOSED; sock_put(sk); /* passive closing */ break; case SMC_PROCESSABORT: case SMC_APPFINCLOSEWAIT: sk->sk_state = SMC_CLOSED; break; - case SMC_PEERFINCLOSEWAIT: - sock_put(sk); /* passive closing */ - break; case SMC_INIT: case SMC_PEERABORTWAIT: case SMC_CLOSED: @@ -215,8 +203,6 @@ again: if (sk->sk_state == SMC_ACTIVE) { /* send close request */ rc = smc_close_final(conn); - if (rc) - break; sk->sk_state = SMC_PEERCLOSEWAIT1; } else { /* peer event has changed the state */ @@ -229,8 +215,6 @@ again: !smc_close_sent_any_close(conn)) { /* just shutdown wr done, send close request */ rc = smc_close_final(conn); - if (rc) - break; } sk->sk_state = SMC_CLOSED; break; @@ -246,8 +230,6 @@ again: goto again; /* confirm close from peer */ rc = smc_close_final(conn); - if (rc) - break; if (smc_cdc_rxed_any_close(conn)) { /* peer has closed the socket already */ sk->sk_state = SMC_CLOSED; @@ -263,8 +245,6 @@ again: !smc_close_sent_any_close(conn)) { /* just shutdown wr done, send close request */ rc = smc_close_final(conn); - if (rc) - break; } /* peer sending PeerConnectionClosed will cause transition */ break; @@ -272,10 +252,12 @@ again: /* peer sending PeerConnectionClosed will cause transition */ break; case SMC_PROCESSABORT: - smc_close_abort(conn); + rc = smc_close_abort(conn); sk->sk_state = SMC_CLOSED; break; case SMC_PEERABORTWAIT: + sk->sk_state = SMC_CLOSED; + break; case SMC_CLOSED: /* nothing to do, add tracing in future patch */ break; @@ -451,8 +433,6 @@ again: goto again; /* send close wr request */ rc = smc_close_wr(conn); - if (rc) - break; sk->sk_state = SMC_PEERCLOSEWAIT1; break; case SMC_APPCLOSEWAIT1: @@ -466,8 +446,6 @@ again: goto again; /* confirm close from peer */ rc = smc_close_wr(conn); - if (rc) - break; sk->sk_state = SMC_APPCLOSEWAIT2; break; case SMC_APPCLOSEWAIT2: -- cgit v1.2.3 From b290098092e4aeaa1712d3326bf5b64d2751c740 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:08 +0200 Subject: net/smc: cancel send and receive for terminated socket The resources for a terminated socket are being cleaned up. This patch makes sure * no more data is received for an actively terminated socket * no more data is sent for an actively or passively terminated socket Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc.h | 1 + net/smc/smc_cdc.c | 4 ++-- net/smc/smc_close.c | 7 +++++-- net/smc/smc_core.c | 1 + net/smc/smc_rx.c | 10 ++++++++-- net/smc/smc_tx.c | 26 +++++++++++++++----------- 6 files changed, 32 insertions(+), 17 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc.h b/net/smc/smc.h index 878313f8d6c1..be11ba41190f 100644 --- a/net/smc/smc.h +++ b/net/smc/smc.h @@ -188,6 +188,7 @@ struct smc_connection { * 0 for SMC-R, 32 for SMC-D */ u64 peer_token; /* SMC-D token of peer */ + u8 killed : 1; /* abnormal termination */ }; struct smc_sock { /* smc sock container */ diff --git a/net/smc/smc_cdc.c b/net/smc/smc_cdc.c index d0b0f4c865b4..7dc07ec2379b 100644 --- a/net/smc/smc_cdc.c +++ b/net/smc/smc_cdc.c @@ -63,7 +63,7 @@ int smc_cdc_get_free_slot(struct smc_connection *conn, rc = smc_wr_tx_get_free_slot(link, smc_cdc_tx_handler, wr_buf, wr_rdma_buf, (struct smc_wr_tx_pend_priv **)pend); - if (!conn->alert_token_local) + if (conn->killed) /* abnormal termination */ rc = -EPIPE; return rc; @@ -328,7 +328,7 @@ static void smcd_cdc_rx_tsklet(unsigned long data) struct smcd_cdc_msg cdc; struct smc_sock *smc; - if (!conn) + if (!conn || conn->killed) return; data_cdc = (struct smcd_cdc_msg *)conn->rmb_desc->cpu_addr; diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c index 1a858e59fc31..1d706c581592 100644 --- a/net/smc/smc_close.c +++ b/net/smc/smc_close.c @@ -66,7 +66,8 @@ static void smc_close_stream_wait(struct smc_sock *smc, long timeout) rc = sk_wait_event(sk, &timeout, !smc_tx_prepared_sends(&smc->conn) || sk->sk_err == ECONNABORTED || - sk->sk_err == ECONNRESET, + sk->sk_err == ECONNRESET || + smc->conn.killed, &wait); if (rc) break; @@ -95,6 +96,8 @@ static int smc_close_final(struct smc_connection *conn) conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1; else conn->local_tx_ctrl.conn_state_flags.peer_conn_closed = 1; + if (conn->killed) + return -EPIPE; return smc_cdc_get_slot_and_msg_send(conn); } @@ -326,7 +329,7 @@ static void smc_close_passive_work(struct work_struct *work) lock_sock(sk); old_state = sk->sk_state; - if (!conn->alert_token_local) { + if (conn->killed) { /* abnormal termination */ smc_close_active_abort(smc); goto wakeup; diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index de9bf035f545..4ee0e33b8c5a 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -500,6 +500,7 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) conn = rb_entry(node, struct smc_connection, alert_node); smc = container_of(conn, struct smc_sock, conn); sock_hold(&smc->sk); /* sock_put in close work */ + conn->killed = 1; conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1; __smc_lgr_unregister_conn(conn); conn->lgr = NULL; diff --git a/net/smc/smc_rx.c b/net/smc/smc_rx.c index 97e8369002d7..39d7b34d06d2 100644 --- a/net/smc/smc_rx.c +++ b/net/smc/smc_rx.c @@ -201,6 +201,8 @@ int smc_rx_wait(struct smc_sock *smc, long *timeo, { DEFINE_WAIT_FUNC(wait, woken_wake_function); struct smc_connection *conn = &smc->conn; + struct smc_cdc_conn_state_flags *cflags = + &conn->local_tx_ctrl.conn_state_flags; struct sock *sk = &smc->sk; int rc; @@ -210,7 +212,9 @@ int smc_rx_wait(struct smc_sock *smc, long *timeo, add_wait_queue(sk_sleep(sk), &wait); rc = sk_wait_event(sk, timeo, sk->sk_err || + cflags->peer_conn_abort || sk->sk_shutdown & RCV_SHUTDOWN || + conn->killed || fcrit(conn), &wait); remove_wait_queue(sk_sleep(sk), &wait); @@ -314,11 +318,13 @@ int smc_rx_recvmsg(struct smc_sock *smc, struct msghdr *msg, if (read_done >= target || (pipe && read_done)) break; + if (conn->killed) + break; + if (smc_rx_recvmsg_data_available(smc)) goto copy; - if (sk->sk_shutdown & RCV_SHUTDOWN || - conn->local_tx_ctrl.conn_state_flags.peer_conn_abort) { + if (sk->sk_shutdown & RCV_SHUTDOWN) { /* smc_cdc_msg_recv_action() could have run after * above smc_rx_recvmsg_data_available() */ diff --git a/net/smc/smc_tx.c b/net/smc/smc_tx.c index 6c8f09c1ce51..824f096ee7de 100644 --- a/net/smc/smc_tx.c +++ b/net/smc/smc_tx.c @@ -86,6 +86,7 @@ static int smc_tx_wait(struct smc_sock *smc, int flags) sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); if (sk->sk_err || (sk->sk_shutdown & SEND_SHUTDOWN) || + conn->killed || conn->local_tx_ctrl.conn_state_flags.peer_done_writing) { rc = -EPIPE; break; @@ -155,7 +156,7 @@ int smc_tx_sendmsg(struct smc_sock *smc, struct msghdr *msg, size_t len) return -ENOTCONN; if (smc->sk.sk_shutdown & SEND_SHUTDOWN || (smc->sk.sk_err == ECONNABORTED) || - conn->local_tx_ctrl.conn_state_flags.peer_conn_abort) + conn->killed) return -EPIPE; if (smc_cdc_rxed_any_close(conn)) return send_done ?: -ECONNRESET; @@ -282,10 +283,8 @@ static int smc_tx_rdma_write(struct smc_connection *conn, int peer_rmbe_offset, peer_rmbe_offset; rdma_wr->rkey = lgr->rtokens[conn->rtoken_idx][SMC_SINGLE_LINK].rkey; rc = ib_post_send(link->roce_qp, &rdma_wr->wr, NULL); - if (rc) { - conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1; + if (rc) smc_lgr_terminate(lgr); - } return rc; } @@ -495,10 +494,11 @@ static int smcr_tx_sndbuf_nonempty(struct smc_connection *conn) if (smc->sk.sk_err == ECONNABORTED) return sock_error(&smc->sk); + if (conn->killed) + return -EPIPE; rc = 0; - if (conn->alert_token_local) /* connection healthy */ - mod_delayed_work(system_wq, &conn->tx_work, - SMC_TX_WORK_DELAY); + mod_delayed_work(system_wq, &conn->tx_work, + SMC_TX_WORK_DELAY); } return rc; } @@ -547,6 +547,9 @@ int smc_tx_sndbuf_nonempty(struct smc_connection *conn) { int rc; + if (conn->killed || + conn->local_rx_ctrl.conn_state_flags.peer_conn_abort) + return -EPIPE; /* connection being aborted */ if (conn->lgr->is_smcd) rc = smcd_tx_sndbuf_nonempty(conn); else @@ -573,9 +576,7 @@ void smc_tx_work(struct work_struct *work) int rc; lock_sock(&smc->sk); - if (smc->sk.sk_err || - !conn->alert_token_local || - conn->local_rx_ctrl.conn_state_flags.peer_conn_abort) + if (smc->sk.sk_err) goto out; rc = smc_tx_sndbuf_nonempty(conn); @@ -608,8 +609,11 @@ void smc_tx_consumer_update(struct smc_connection *conn, bool force) ((to_confirm > conn->rmbe_update_limit) && ((sender_free <= (conn->rmb_desc->len / 2)) || conn->local_rx_ctrl.prod_flags.write_blocked))) { + if (conn->killed || + conn->local_rx_ctrl.conn_state_flags.peer_conn_abort) + return; if ((smc_cdc_get_slot_and_msg_send(conn) < 0) && - conn->alert_token_local) { /* connection healthy */ + !conn->killed) { schedule_delayed_work(&conn->tx_work, SMC_TX_WORK_DELAY); return; -- cgit v1.2.3 From 8caa654451bda40379bff786a63833b2965536e4 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:09 +0200 Subject: net/smc: terminate link group without holding lgr lock When a link group is to be terminated, it is sufficient to hold the lgr lock when unlinking the link group from its list. Move the lock-protected link group unlinking into smc_lgr_terminate(). Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 25 +++++++++++++++++-------- 1 file changed, 17 insertions(+), 8 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 4ee0e33b8c5a..b53ba8f0a833 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -182,8 +182,7 @@ static void smc_lgr_free_work(struct work_struct *work) spin_unlock_bh(lgr_lock); return; } - if (!list_empty(&lgr->list)) - list_del_init(&lgr->list); /* remove from smc_lgr_list */ + list_del_init(&lgr->list); /* remove from smc_lgr_list */ spin_unlock_bh(lgr_lock); if (!lgr->is_smcd && !lgr->terminating) { @@ -479,7 +478,7 @@ void smc_lgr_forget(struct smc_link_group *lgr) spin_unlock_bh(lgr_lock); } -/* terminate linkgroup abnormally */ +/* terminate link group */ static void __smc_lgr_terminate(struct smc_link_group *lgr) { struct smc_connection *conn; @@ -489,8 +488,6 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) if (lgr->terminating) return; /* lgr already terminating */ lgr->terminating = 1; - if (!list_empty(&lgr->list)) /* forget lgr */ - list_del_init(&lgr->list); if (!lgr->is_smcd) smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]); @@ -516,29 +513,41 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) smc_lgr_schedule_free_work(lgr); } +/* unlink and terminate link group */ void smc_lgr_terminate(struct smc_link_group *lgr) { spinlock_t *lgr_lock; smc_lgr_list_head(lgr, &lgr_lock); spin_lock_bh(lgr_lock); - __smc_lgr_terminate(lgr); + if (lgr->terminating) { + spin_unlock_bh(lgr_lock); + return; /* lgr already terminating */ + } + list_del_init(&lgr->list); spin_unlock_bh(lgr_lock); + __smc_lgr_terminate(lgr); } /* Called when IB port is terminated */ void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport) { struct smc_link_group *lgr, *l; + LIST_HEAD(lgr_free_list); spin_lock_bh(&smc_lgr_list.lock); list_for_each_entry_safe(lgr, l, &smc_lgr_list.list, list) { if (!lgr->is_smcd && lgr->lnk[SMC_SINGLE_LINK].smcibdev == smcibdev && lgr->lnk[SMC_SINGLE_LINK].ibport == ibport) - __smc_lgr_terminate(lgr); + list_move(&lgr->list, &lgr_free_list); } spin_unlock_bh(&smc_lgr_list.lock); + + list_for_each_entry_safe(lgr, l, &lgr_free_list, list) { + list_del_init(&lgr->list); + __smc_lgr_terminate(lgr); + } } /* Called when SMC-D device is terminated or peer is lost */ @@ -552,7 +561,6 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) list_for_each_entry_safe(lgr, l, &dev->lgr_list, list) { if ((!peer_gid || lgr->peer_gid == peer_gid) && (vlan == VLAN_VID_MASK || lgr->vlan_id == vlan)) { - __smc_lgr_terminate(lgr); list_move(&lgr->list, &lgr_free_list); } } @@ -561,6 +569,7 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) /* cancel the regular free workers and actually free lgrs */ list_for_each_entry_safe(lgr, l, &lgr_free_list, list) { list_del_init(&lgr->list); + __smc_lgr_terminate(lgr); cancel_delayed_work_sync(&lgr->free_work); if (!peer_gid && vlan == VLAN_VID_MASK) /* dev terminated? */ smc_ism_signal_shutdown(lgr); -- cgit v1.2.3 From 69318b5215f2dc32c345a3d65b98b4b1bf29c007 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:10 +0200 Subject: net/smc: improve abnormal termination locking Locking hierarchy requires that the link group conns_lock can be taken if the socket lock is held, but not vice versa. Nevertheless socket termination during abnormal link group termination should be protected by the socket lock. This patch reduces the time segments the link group conns_lock is held to enable usage of lock_sock in smc_lgr_terminate(). Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index b53ba8f0a833..1f58cd82928c 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -491,23 +491,26 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) if (!lgr->is_smcd) smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]); - write_lock_bh(&lgr->conns_lock); + /* kill remaining link group connections */ + read_lock_bh(&lgr->conns_lock); node = rb_first(&lgr->conns_all); while (node) { + read_unlock_bh(&lgr->conns_lock); conn = rb_entry(node, struct smc_connection, alert_node); smc = container_of(conn, struct smc_sock, conn); + lock_sock(&smc->sk); sock_hold(&smc->sk); /* sock_put in close work */ conn->killed = 1; conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1; - __smc_lgr_unregister_conn(conn); + smc_lgr_unregister_conn(conn); conn->lgr = NULL; - write_unlock_bh(&lgr->conns_lock); if (!schedule_work(&conn->close_work)) sock_put(&smc->sk); - write_lock_bh(&lgr->conns_lock); + release_sock(&smc->sk); + read_lock_bh(&lgr->conns_lock); node = rb_first(&lgr->conns_all); } - write_unlock_bh(&lgr->conns_lock); + read_unlock_bh(&lgr->conns_lock); if (!lgr->is_smcd) wake_up(&lgr->lnk[SMC_SINGLE_LINK].wr_reg_wait); smc_lgr_schedule_free_work(lgr); -- cgit v1.2.3 From 8e316b9e7260cbc61974c2558733dab5de949399 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:11 +0200 Subject: net/smc: improve link group freeing Usually link groups are freed delayed to enable quick connection creation for a follow-on SMC socket. Terminated link groups are freed faster. This patch makes sure, fast schedule of link group freeing is not rescheduled by a delayed schedule. And it makes sure link group freeing is not rescheduled, if the real freeing is already running. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 47 ++++++++++++++++++++++++++++++----------------- net/smc/smc_core.h | 2 ++ 2 files changed, 32 insertions(+), 17 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 1f58cd82928c..e7e9dbcd7d8b 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -61,14 +61,21 @@ static void smc_lgr_schedule_free_work(struct smc_link_group *lgr) * creation. For client use a somewhat higher removal delay time, * otherwise there is a risk of out-of-sync link groups. */ - mod_delayed_work(system_wq, &lgr->free_work, - (!lgr->is_smcd && lgr->role == SMC_CLNT) ? - SMC_LGR_FREE_DELAY_CLNT : SMC_LGR_FREE_DELAY_SERV); + if (!lgr->freeing && !lgr->freefast) { + mod_delayed_work(system_wq, &lgr->free_work, + (!lgr->is_smcd && lgr->role == SMC_CLNT) ? + SMC_LGR_FREE_DELAY_CLNT : + SMC_LGR_FREE_DELAY_SERV); + } } void smc_lgr_schedule_free_work_fast(struct smc_link_group *lgr) { - mod_delayed_work(system_wq, &lgr->free_work, SMC_LGR_FREE_DELAY_FAST); + if (!lgr->freeing && !lgr->freefast) { + lgr->freefast = 1; + mod_delayed_work(system_wq, &lgr->free_work, + SMC_LGR_FREE_DELAY_FAST); + } } /* Register connection's alert token in our lookup structure. @@ -171,10 +178,15 @@ static void smc_lgr_free_work(struct work_struct *work) struct smc_link_group, free_work); spinlock_t *lgr_lock; + struct smc_link *lnk; bool conns; smc_lgr_list_head(lgr, &lgr_lock); spin_lock_bh(lgr_lock); + if (lgr->freeing) { + spin_unlock_bh(lgr_lock); + return; + } read_lock_bh(&lgr->conns_lock); conns = RB_EMPTY_ROOT(&lgr->conns_all); read_unlock_bh(&lgr->conns_lock); @@ -183,29 +195,27 @@ static void smc_lgr_free_work(struct work_struct *work) return; } list_del_init(&lgr->list); /* remove from smc_lgr_list */ - spin_unlock_bh(lgr_lock); + lnk = &lgr->lnk[SMC_SINGLE_LINK]; if (!lgr->is_smcd && !lgr->terminating) { - struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK]; - /* try to send del link msg, on error free lgr immediately */ if (lnk->state == SMC_LNK_ACTIVE && !smc_link_send_delete(lnk)) { /* reschedule in case we never receive a response */ smc_lgr_schedule_free_work(lgr); + spin_unlock_bh(lgr_lock); return; } } + lgr->freeing = 1; /* this instance does the freeing, no new schedule */ + spin_unlock_bh(lgr_lock); + cancel_delayed_work(&lgr->free_work); - if (!delayed_work_pending(&lgr->free_work)) { - struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK]; - - if (!lgr->is_smcd && lnk->state != SMC_LNK_INACTIVE) - smc_llc_link_inactive(lnk); - if (lgr->is_smcd) - smc_ism_signal_shutdown(lgr); - smc_lgr_free(lgr); - } + if (!lgr->is_smcd && lnk->state != SMC_LNK_INACTIVE) + smc_llc_link_inactive(lnk); + if (lgr->is_smcd) + smc_ism_signal_shutdown(lgr); + smc_lgr_free(lgr); } /* create a new SMC link group */ @@ -233,6 +243,9 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) } lgr->is_smcd = ini->is_smcd; lgr->sync_err = 0; + lgr->terminating = 0; + lgr->freefast = 0; + lgr->freeing = 0; lgr->vlan_id = ini->vlan_id; rwlock_init(&lgr->sndbufs_lock); rwlock_init(&lgr->rmbs_lock); @@ -513,7 +526,7 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) read_unlock_bh(&lgr->conns_lock); if (!lgr->is_smcd) wake_up(&lgr->lnk[SMC_SINGLE_LINK].wr_reg_wait); - smc_lgr_schedule_free_work(lgr); + smc_lgr_schedule_free_work_fast(lgr); } /* unlink and terminate link group */ diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index c00ac61dc129..12c2818b293f 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -204,6 +204,8 @@ struct smc_link_group { struct delayed_work free_work; /* delayed freeing of an lgr */ u8 sync_err : 1; /* lgr no longer fits to peer */ u8 terminating : 1;/* lgr is terminating */ + u8 freefast : 1; /* free worker scheduled fast */ + u8 freeing : 1; /* lgr is being freed */ bool is_smcd; /* SMC-R or SMC-D */ union { -- cgit v1.2.3 From 8317976096635110603c3e143bcaf8773f4a3e65 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:12 +0200 Subject: net/smc: tell peers about abnormal link group termination There are lots of link group termination scenarios. Most of them still allow to inform the peer of the terminating sockets about aborting. This patch tries to call smc_close_abort() for terminating sockets. And the internal TCP socket is reset with tcp_abort(). Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_close.c | 9 ++++----- net/smc/smc_close.h | 1 + net/smc/smc_core.c | 2 +- 3 files changed, 6 insertions(+), 6 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c index 1d706c581592..2bbcd45a421e 100644 --- a/net/smc/smc_close.c +++ b/net/smc/smc_close.c @@ -13,6 +13,7 @@ #include #include +#include #include "smc.h" #include "smc_tx.h" @@ -102,7 +103,7 @@ static int smc_close_final(struct smc_connection *conn) return smc_cdc_get_slot_and_msg_send(conn); } -static int smc_close_abort(struct smc_connection *conn) +int smc_close_abort(struct smc_connection *conn) { conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1; @@ -118,10 +119,8 @@ static void smc_close_active_abort(struct smc_sock *smc) if (sk->sk_state != SMC_INIT && smc->clcsock && smc->clcsock->sk) { sk->sk_err = ECONNABORTED; - if (smc->clcsock && smc->clcsock->sk) { - smc->clcsock->sk->sk_err = ECONNABORTED; - smc->clcsock->sk->sk_state_change(smc->clcsock->sk); - } + if (smc->clcsock && smc->clcsock->sk) + tcp_abort(smc->clcsock->sk, ECONNABORTED); } switch (sk->sk_state) { case SMC_ACTIVE: diff --git a/net/smc/smc_close.h b/net/smc/smc_close.h index e0e3b5df25d2..084c4f37aa96 100644 --- a/net/smc/smc_close.h +++ b/net/smc/smc_close.h @@ -24,5 +24,6 @@ int smc_close_active(struct smc_sock *smc); int smc_close_shutdown_write(struct smc_sock *smc); void smc_close_init(struct smc_sock *smc); void smc_clcsock_release(struct smc_sock *smc); +int smc_close_abort(struct smc_connection *conn); #endif /* SMC_CLOSE_H */ diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index e7e9dbcd7d8b..494288f32df6 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -513,8 +513,8 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) smc = container_of(conn, struct smc_sock, conn); lock_sock(&smc->sk); sock_hold(&smc->sk); /* sock_put in close work */ + smc_close_abort(conn); conn->killed = 1; - conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1; smc_lgr_unregister_conn(conn); conn->lgr = NULL; if (!schedule_work(&conn->close_work)) -- cgit v1.2.3 From 2a0674fffb6bc1a7c0f46bb2e0b1bcf1d49c2232 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:13 +0200 Subject: net/smc: improve abnormal termination of link groups If a link group and its connections must be terminated, * wake up socket waiters * do not enable buffer reuse A linkgroup might be terminated while normal connection closing is running. Avoid buffer reuse and its related LLC DELETE RKEY call, if linkgroup termination has started. And use the earliest indication of linkgroup termination possible, namely the removal from the linkgroup list. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 40 +++++++++++++++++++++++++++++----------- 1 file changed, 29 insertions(+), 11 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 494288f32df6..6faaa38412b1 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -154,6 +154,7 @@ static void smc_lgr_unregister_conn(struct smc_connection *conn) __smc_lgr_unregister_conn(conn); } write_unlock_bh(&lgr->conns_lock); + conn->lgr = NULL; } /* Send delete link, either as client to request the initiation @@ -344,7 +345,7 @@ static void smc_buf_unuse(struct smc_connection *conn, conn->sndbuf_desc->used = 0; if (conn->rmb_desc) { if (!conn->rmb_desc->regerr) { - if (!lgr->is_smcd) { + if (!lgr->is_smcd && !list_empty(&lgr->list)) { /* unregister rmb with peer */ smc_llc_do_delete_rkey( &lgr->lnk[SMC_SINGLE_LINK], @@ -375,9 +376,10 @@ void smc_conn_free(struct smc_connection *conn) } else { smc_cdc_tx_dismiss_slots(conn); } - smc_lgr_unregister_conn(conn); - smc_buf_unuse(conn, lgr); /* allow buffer reuse */ - conn->lgr = NULL; + if (!list_empty(&lgr->list)) { + smc_lgr_unregister_conn(conn); + smc_buf_unuse(conn, lgr); /* allow buffer reuse */ + } if (!lgr->conns_num) smc_lgr_schedule_free_work(lgr); @@ -491,6 +493,28 @@ void smc_lgr_forget(struct smc_link_group *lgr) spin_unlock_bh(lgr_lock); } +static void smc_sk_wake_ups(struct smc_sock *smc) +{ + smc->sk.sk_write_space(&smc->sk); + smc->sk.sk_data_ready(&smc->sk); + smc->sk.sk_state_change(&smc->sk); +} + +/* kill a connection */ +static void smc_conn_kill(struct smc_connection *conn) +{ + struct smc_sock *smc = container_of(conn, struct smc_sock, conn); + + smc_close_abort(conn); + conn->killed = 1; + smc_sk_wake_ups(smc); + smc_lgr_unregister_conn(conn); + smc->sk.sk_err = ECONNABORTED; + sock_hold(&smc->sk); /* sock_put in close work */ + if (!schedule_work(&conn->close_work)) + sock_put(&smc->sk); +} + /* terminate link group */ static void __smc_lgr_terminate(struct smc_link_group *lgr) { @@ -512,13 +536,7 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) conn = rb_entry(node, struct smc_connection, alert_node); smc = container_of(conn, struct smc_sock, conn); lock_sock(&smc->sk); - sock_hold(&smc->sk); /* sock_put in close work */ - smc_close_abort(conn); - conn->killed = 1; - smc_lgr_unregister_conn(conn); - conn->lgr = NULL; - if (!schedule_work(&conn->close_work)) - sock_put(&smc->sk); + smc_conn_kill(conn); release_sock(&smc->sk); read_lock_bh(&lgr->conns_lock); node = rb_first(&lgr->conns_all); -- cgit v1.2.3 From f528ba24a8ad61b8a5e55d34cb1da127ce67cf6e Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:14 +0200 Subject: net/smc: introduce link group termination worker Use a worker for link group termination to guarantee process context. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/smc_core.c | 9 +++++++++ net/smc/smc_core.h | 7 +++++++ net/smc/smc_llc.c | 2 +- net/smc/smc_wr.c | 10 +++++----- 4 files changed, 22 insertions(+), 6 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 6faaa38412b1..46d4b944c4c4 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -219,6 +219,14 @@ static void smc_lgr_free_work(struct work_struct *work) smc_lgr_free(lgr); } +static void smc_lgr_terminate_work(struct work_struct *work) +{ + struct smc_link_group *lgr = container_of(work, struct smc_link_group, + terminate_work); + + smc_lgr_terminate(lgr); +} + /* create a new SMC link group */ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) { @@ -258,6 +266,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) smc_lgr_list.num += SMC_LGR_NUM_INCR; memcpy(&lgr->id, (u8 *)&smc_lgr_list.num, SMC_LGR_ID_SIZE); INIT_DELAYED_WORK(&lgr->free_work, smc_lgr_free_work); + INIT_WORK(&lgr->terminate_work, smc_lgr_terminate_work); lgr->conns_all = RB_ROOT; if (ini->is_smcd) { /* SMC-D specific settings */ diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 12c2818b293f..e6fd1ed42064 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -202,6 +202,7 @@ struct smc_link_group { u8 id[SMC_LGR_ID_SIZE]; /* unique lgr id */ struct delayed_work free_work; /* delayed freeing of an lgr */ + struct work_struct terminate_work; /* abnormal lgr termination */ u8 sync_err : 1; /* lgr no longer fits to peer */ u8 terminating : 1;/* lgr is terminating */ u8 freefast : 1; /* free worker scheduled fast */ @@ -282,6 +283,12 @@ static inline struct smc_connection *smc_lgr_find_conn( return res; } +static inline void smc_lgr_terminate_sched(struct smc_link_group *lgr) +{ + if (!lgr->terminating) + schedule_work(&lgr->terminate_work); +} + struct smc_sock; struct smc_clc_msg_accept_confirm; struct smc_clc_msg_local; diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c index 4fd60c522802..e1918ffaf125 100644 --- a/net/smc/smc_llc.c +++ b/net/smc/smc_llc.c @@ -475,7 +475,7 @@ static void smc_llc_rx_delete_link(struct smc_link *link, smc_llc_prep_delete_link(llc, link, SMC_LLC_RESP, true); } smc_llc_send_message(link, llc, sizeof(*llc)); - smc_lgr_schedule_free_work_fast(lgr); + smc_lgr_terminate_sched(lgr); } } diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c index 253aa75dc2b6..50743dc56c86 100644 --- a/net/smc/smc_wr.c +++ b/net/smc/smc_wr.c @@ -101,7 +101,7 @@ static inline void smc_wr_tx_process_cqe(struct ib_wc *wc) clear_bit(i, link->wr_tx_mask); } /* terminate connections of this link group abnormally */ - smc_lgr_terminate(smc_get_lgr(link)); + smc_lgr_terminate_sched(smc_get_lgr(link)); } if (pnd_snd.handler) pnd_snd.handler(&pnd_snd.priv, link, wc->status); @@ -191,7 +191,7 @@ int smc_wr_tx_get_free_slot(struct smc_link *link, SMC_WR_TX_WAIT_FREE_SLOT_TIME); if (!rc) { /* timeout - terminate connections */ - smc_lgr_terminate(smc_get_lgr(link)); + smc_lgr_terminate_sched(smc_get_lgr(link)); return -EPIPE; } if (idx == link->wr_tx_cnt) @@ -247,7 +247,7 @@ int smc_wr_tx_send(struct smc_link *link, struct smc_wr_tx_pend_priv *priv) rc = ib_post_send(link->roce_qp, &link->wr_tx_ibs[pend->idx], NULL); if (rc) { smc_wr_tx_put_slot(link, priv); - smc_lgr_terminate(smc_get_lgr(link)); + smc_lgr_terminate_sched(smc_get_lgr(link)); } return rc; } @@ -272,7 +272,7 @@ int smc_wr_reg_send(struct smc_link *link, struct ib_mr *mr) SMC_WR_REG_MR_WAIT_TIME); if (!rc) { /* timeout - terminate connections */ - smc_lgr_terminate(smc_get_lgr(link)); + smc_lgr_terminate_sched(smc_get_lgr(link)); return -EPIPE; } if (rc == -ERESTARTSYS) @@ -373,7 +373,7 @@ static inline void smc_wr_rx_process_cqes(struct ib_wc wc[], int num) /* terminate connections of this link group * abnormally */ - smc_lgr_terminate(smc_get_lgr(link)); + smc_lgr_terminate_sched(smc_get_lgr(link)); break; default: smc_wr_rx_post(link); /* refill WR RX */ -- cgit v1.2.3 From 81cf4f4707af9704ac1c3dd177c8bd1fcc01da6c Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Mon, 21 Oct 2019 16:13:15 +0200 Subject: net/smc: remove close abort worker With the introduction of the link group termination worker there is no longer a need to postpone smc_close_active_abort() to a worker. To protect socket destruction due to normal and abnormal socket closing, the socket refcount is increased. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: Jakub Kicinski --- net/smc/af_smc.c | 4 ++++ net/smc/smc_close.c | 18 +++++++++++------- net/smc/smc_close.h | 1 + net/smc/smc_core.c | 6 +++--- 4 files changed, 19 insertions(+), 10 deletions(-) (limited to 'net/smc') diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index 5b932583e407..91ea098fabd9 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -167,6 +167,7 @@ static int smc_release(struct socket *sock) if (!sk) goto out; + sock_hold(sk); /* sock_put below */ smc = smc_sk(sk); /* cleanup for a dangling non-blocking connect */ @@ -189,6 +190,7 @@ static int smc_release(struct socket *sock) sock->sk = NULL; release_sock(sk); + sock_put(sk); /* sock_hold above */ sock_put(sk); /* final sock_put */ out: return rc; @@ -970,12 +972,14 @@ void smc_close_non_accepted(struct sock *sk) { struct smc_sock *smc = smc_sk(sk); + sock_hold(sk); /* sock_put below */ lock_sock(sk); if (!sk->sk_lingertime) /* wait for peer closing */ sk->sk_lingertime = SMC_MAX_STREAM_WAIT_TIMEOUT; __smc_release(smc); release_sock(sk); + sock_put(sk); /* sock_hold above */ sock_put(sk); /* final sock_put */ } diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c index 2bbcd45a421e..d34e5adce2eb 100644 --- a/net/smc/smc_close.c +++ b/net/smc/smc_close.c @@ -113,9 +113,10 @@ int smc_close_abort(struct smc_connection *conn) /* terminate smc socket abnormally - active abort * link group is terminated, i.e. RDMA communication no longer possible */ -static void smc_close_active_abort(struct smc_sock *smc) +void smc_close_active_abort(struct smc_sock *smc) { struct sock *sk = &smc->sk; + bool release_clcsock = false; if (sk->sk_state != SMC_INIT && smc->clcsock && smc->clcsock->sk) { sk->sk_err = ECONNABORTED; @@ -137,11 +138,14 @@ static void smc_close_active_abort(struct smc_sock *smc) cancel_delayed_work_sync(&smc->conn.tx_work); lock_sock(sk); sk->sk_state = SMC_CLOSED; + sock_put(sk); /* postponed passive closing */ break; case SMC_PEERCLOSEWAIT1: case SMC_PEERCLOSEWAIT2: case SMC_PEERFINCLOSEWAIT: sk->sk_state = SMC_CLOSED; + smc_conn_free(&smc->conn); + release_clcsock = true; sock_put(sk); /* passive closing */ break; case SMC_PROCESSABORT: @@ -156,6 +160,12 @@ static void smc_close_active_abort(struct smc_sock *smc) sock_set_flag(sk, SOCK_DEAD); sk->sk_state_change(sk); + + if (release_clcsock) { + release_sock(sk); + smc_clcsock_release(smc); + lock_sock(sk); + } } static inline bool smc_close_sent_any_close(struct smc_connection *conn) @@ -328,12 +338,6 @@ static void smc_close_passive_work(struct work_struct *work) lock_sock(sk); old_state = sk->sk_state; - if (conn->killed) { - /* abnormal termination */ - smc_close_active_abort(smc); - goto wakeup; - } - rxflags = &conn->local_rx_ctrl.conn_state_flags; if (rxflags->peer_conn_abort) { /* peer has not received all data */ diff --git a/net/smc/smc_close.h b/net/smc/smc_close.h index 084c4f37aa96..634fea2b7c95 100644 --- a/net/smc/smc_close.h +++ b/net/smc/smc_close.h @@ -25,5 +25,6 @@ int smc_close_shutdown_write(struct smc_sock *smc); void smc_close_init(struct smc_sock *smc); void smc_clcsock_release(struct smc_sock *smc); int smc_close_abort(struct smc_connection *conn); +void smc_close_active_abort(struct smc_sock *smc); #endif /* SMC_CLOSE_H */ diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 46d4b944c4c4..ed02eac636da 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -519,9 +519,7 @@ static void smc_conn_kill(struct smc_connection *conn) smc_sk_wake_ups(smc); smc_lgr_unregister_conn(conn); smc->sk.sk_err = ECONNABORTED; - sock_hold(&smc->sk); /* sock_put in close work */ - if (!schedule_work(&conn->close_work)) - sock_put(&smc->sk); + smc_close_active_abort(smc); } /* terminate link group */ @@ -544,9 +542,11 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) read_unlock_bh(&lgr->conns_lock); conn = rb_entry(node, struct smc_connection, alert_node); smc = container_of(conn, struct smc_sock, conn); + sock_hold(&smc->sk); /* sock_put below */ lock_sock(&smc->sk); smc_conn_kill(conn); release_sock(&smc->sk); + sock_put(&smc->sk); /* sock_hold above */ read_lock_bh(&lgr->conns_lock); node = rb_first(&lgr->conns_all); } -- cgit v1.2.3 From 50c6b20eff8e10cb91f06262d8003a5d9be3dfab Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:40 +0100 Subject: net/smc: fix final cleanup sequence for SMCD devices If peer announces shutdown, use the link group terminate worker for local cleanup of link groups and connections to terminate link group in proper context. Make sure link groups are cleaned up first before destroying the event queue of the SMCD device, because link group cleanup may raise events. Send signal shutdown only if peer has not done it already. Send socket abort or close only, if peer has not already announced shutdown. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_cdc.c | 3 +++ net/smc/smc_core.c | 18 +++++++++++------- net/smc/smc_core.h | 2 ++ net/smc/smc_ism.c | 7 +++++-- 4 files changed, 21 insertions(+), 9 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_cdc.c b/net/smc/smc_cdc.c index 7dc07ec2379b..164f1584861b 100644 --- a/net/smc/smc_cdc.c +++ b/net/smc/smc_cdc.c @@ -131,6 +131,9 @@ int smc_cdc_get_slot_and_msg_send(struct smc_connection *conn) { int rc; + if (!conn->lgr || (conn->lgr->is_smcd && conn->lgr->peer_shutdown)) + return -EPIPE; + if (conn->lgr->is_smcd) { spin_lock_bh(&conn->send_lock); rc = smcd_cdc_msg_send(conn); diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 0d92456729ab..561f069b30de 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -275,6 +275,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) lgr->smcd = ini->ism_dev; lgr_list = &ini->ism_dev->lgr_list; lgr_lock = &lgr->smcd->lgr_lock; + lgr->peer_shutdown = 0; } else { /* SMC-R specific settings */ get_device(&ini->ib_dev->ibdev->dev); @@ -514,11 +515,16 @@ static void smc_conn_kill(struct smc_connection *conn) { struct smc_sock *smc = container_of(conn, struct smc_sock, conn); - smc_close_abort(conn); + if (conn->lgr->is_smcd && conn->lgr->peer_shutdown) + conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1; + else + smc_close_abort(conn); conn->killed = 1; + smc->sk.sk_err = ECONNABORTED; smc_sk_wake_ups(smc); + if (conn->lgr->is_smcd) + tasklet_kill(&conn->rx_tsklet); smc_lgr_unregister_conn(conn); - smc->sk.sk_err = ECONNABORTED; smc_close_active_abort(smc); } @@ -604,6 +610,8 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) list_for_each_entry_safe(lgr, l, &dev->lgr_list, list) { if ((!peer_gid || lgr->peer_gid == peer_gid) && (vlan == VLAN_VID_MASK || lgr->vlan_id == vlan)) { + if (peer_gid) /* peer triggered termination */ + lgr->peer_shutdown = 1; list_move(&lgr->list, &lgr_free_list); } } @@ -612,11 +620,7 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) /* cancel the regular free workers and actually free lgrs */ list_for_each_entry_safe(lgr, l, &lgr_free_list, list) { list_del_init(&lgr->list); - __smc_lgr_terminate(lgr); - cancel_delayed_work_sync(&lgr->free_work); - if (!peer_gid && vlan == VLAN_VID_MASK) /* dev terminated? */ - smc_ism_signal_shutdown(lgr); - smc_lgr_free(lgr); + schedule_work(&lgr->terminate_work); } } diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index e6fd1ed42064..097ceba86caf 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -228,6 +228,8 @@ struct smc_link_group { /* Peer GID (remote) */ struct smcd_dev *smcd; /* ISM device for VLAN reg. */ + u8 peer_shutdown : 1; + /* peer triggered shutdownn */ }; }; }; diff --git a/net/smc/smc_ism.c b/net/smc/smc_ism.c index ee7340898cb4..18946e95a3be 100644 --- a/net/smc/smc_ism.c +++ b/net/smc/smc_ism.c @@ -226,6 +226,9 @@ int smc_ism_signal_shutdown(struct smc_link_group *lgr) int rc; union smcd_sw_event_info ev_info; + if (lgr->peer_shutdown) + return 0; + memcpy(ev_info.uid, lgr->id, SMC_LGR_ID_SIZE); ev_info.vlan_id = lgr->vlan_id; ev_info.code = ISM_EVENT_REQUEST; @@ -313,12 +316,12 @@ EXPORT_SYMBOL_GPL(smcd_register_dev); void smcd_unregister_dev(struct smcd_dev *smcd) { spin_lock(&smcd_dev_list.lock); - list_del(&smcd->list); + list_del_init(&smcd->list); spin_unlock(&smcd_dev_list.lock); smcd->going_away = 1; + smc_smcd_terminate(smcd, 0, VLAN_VID_MASK); flush_workqueue(smcd->event_wq); destroy_workqueue(smcd->event_wq); - smc_smcd_terminate(smcd, 0, VLAN_VID_MASK); device_del(&smcd->dev); } -- cgit v1.2.3 From 42bfba9eaa33dd4af0b50b87508062a41ec26653 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:41 +0100 Subject: net/smc: immediate termination for SMCD link groups SMCD link group termination is called when peer signals its shutdown of its corresponding link group. For regular shutdowns no connections exist anymore. For abnormal shutdowns connections must be killed and their DMBs must be unregistered immediately. That means the SMCR method to delay the link group freeing several seconds does not fit. This patch adds immediate termination of a link group and its SMCD connections and makes sure all SMCD link group related cleanup steps are finished. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_close.c | 25 +++++++++++++++++++------ net/smc/smc_core.c | 46 +++++++++++++++++++++++++++++++++++++++------- net/smc/smc_ism.c | 14 ++++++++++++-- 3 files changed, 70 insertions(+), 15 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c index d34e5adce2eb..d205b2114006 100644 --- a/net/smc/smc_close.c +++ b/net/smc/smc_close.c @@ -110,6 +110,17 @@ int smc_close_abort(struct smc_connection *conn) return smc_cdc_get_slot_and_msg_send(conn); } +static void smc_close_cancel_work(struct smc_sock *smc) +{ + struct sock *sk = &smc->sk; + + release_sock(sk); + cancel_work_sync(&smc->conn.close_work); + cancel_delayed_work_sync(&smc->conn.tx_work); + lock_sock(sk); + sk->sk_state = SMC_CLOSED; +} + /* terminate smc socket abnormally - active abort * link group is terminated, i.e. RDMA communication no longer possible */ @@ -126,23 +137,21 @@ void smc_close_active_abort(struct smc_sock *smc) switch (sk->sk_state) { case SMC_ACTIVE: sk->sk_state = SMC_PEERABORTWAIT; - release_sock(sk); - cancel_delayed_work_sync(&smc->conn.tx_work); - lock_sock(sk); + smc_close_cancel_work(smc); sk->sk_state = SMC_CLOSED; sock_put(sk); /* passive closing */ break; case SMC_APPCLOSEWAIT1: case SMC_APPCLOSEWAIT2: - release_sock(sk); - cancel_delayed_work_sync(&smc->conn.tx_work); - lock_sock(sk); + smc_close_cancel_work(smc); sk->sk_state = SMC_CLOSED; sock_put(sk); /* postponed passive closing */ break; case SMC_PEERCLOSEWAIT1: case SMC_PEERCLOSEWAIT2: case SMC_PEERFINCLOSEWAIT: + sk->sk_state = SMC_PEERABORTWAIT; + smc_close_cancel_work(smc); sk->sk_state = SMC_CLOSED; smc_conn_free(&smc->conn); release_clcsock = true; @@ -150,7 +159,11 @@ void smc_close_active_abort(struct smc_sock *smc) break; case SMC_PROCESSABORT: case SMC_APPFINCLOSEWAIT: + sk->sk_state = SMC_PEERABORTWAIT; + smc_close_cancel_work(smc); sk->sk_state = SMC_CLOSED; + smc_conn_free(&smc->conn); + release_clcsock = true; break; case SMC_INIT: case SMC_PEERABORTWAIT: diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 561f069b30de..9d6da2c7413d 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -214,7 +214,7 @@ static void smc_lgr_free_work(struct work_struct *work) if (!lgr->is_smcd && lnk->state != SMC_LNK_INACTIVE) smc_llc_link_inactive(lnk); - if (lgr->is_smcd) + if (lgr->is_smcd && !lgr->terminating) smc_ism_signal_shutdown(lgr); smc_lgr_free(lgr); } @@ -381,7 +381,8 @@ void smc_conn_free(struct smc_connection *conn) if (!lgr) return; if (lgr->is_smcd) { - smc_ism_unset_conn(conn); + if (!list_empty(&lgr->list)) + smc_ism_unset_conn(conn); tasklet_kill(&conn->rx_tsklet); } else { smc_cdc_tx_dismiss_slots(conn); @@ -481,8 +482,10 @@ static void smc_lgr_free(struct smc_link_group *lgr) { smc_lgr_free_bufs(lgr); if (lgr->is_smcd) { - smc_ism_put_vlan(lgr->smcd, lgr->vlan_id); - put_device(&lgr->smcd->dev); + if (!lgr->terminating) { + smc_ism_put_vlan(lgr->smcd, lgr->vlan_id); + put_device(&lgr->smcd->dev); + } } else { smc_link_clear(&lgr->lnk[SMC_SINGLE_LINK]); put_device(&lgr->lnk[SMC_SINGLE_LINK].smcibdev->ibdev->dev); @@ -503,6 +506,20 @@ void smc_lgr_forget(struct smc_link_group *lgr) spin_unlock_bh(lgr_lock); } +static void smcd_unregister_all_dmbs(struct smc_link_group *lgr) +{ + int i; + + for (i = 0; i < SMC_RMBE_SIZES; i++) { + struct smc_buf_desc *buf_desc; + + list_for_each_entry(buf_desc, &lgr->rmbs[i], list) { + buf_desc->len += sizeof(struct smcd_cdc_msg); + smc_ism_unregister_dmb(lgr->smcd, buf_desc); + } + } +} + static void smc_sk_wake_ups(struct smc_sock *smc) { smc->sk.sk_write_space(&smc->sk); @@ -522,12 +539,28 @@ static void smc_conn_kill(struct smc_connection *conn) conn->killed = 1; smc->sk.sk_err = ECONNABORTED; smc_sk_wake_ups(smc); - if (conn->lgr->is_smcd) + if (conn->lgr->is_smcd) { + smc_ism_unset_conn(conn); tasklet_kill(&conn->rx_tsklet); + } smc_lgr_unregister_conn(conn); smc_close_active_abort(smc); } +static void smc_lgr_cleanup(struct smc_link_group *lgr) +{ + if (lgr->is_smcd) { + smc_ism_signal_shutdown(lgr); + smcd_unregister_all_dmbs(lgr); + smc_ism_put_vlan(lgr->smcd, lgr->vlan_id); + put_device(&lgr->smcd->dev); + } else { + struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK]; + + wake_up(&lnk->wr_reg_wait); + } +} + /* terminate link group */ static void __smc_lgr_terminate(struct smc_link_group *lgr) { @@ -557,8 +590,7 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) node = rb_first(&lgr->conns_all); } read_unlock_bh(&lgr->conns_lock); - if (!lgr->is_smcd) - wake_up(&lgr->lnk[SMC_SINGLE_LINK].wr_reg_wait); + smc_lgr_cleanup(lgr); smc_lgr_schedule_free_work_fast(lgr); } diff --git a/net/smc/smc_ism.c b/net/smc/smc_ism.c index 18946e95a3be..903da947b20d 100644 --- a/net/smc/smc_ism.c +++ b/net/smc/smc_ism.c @@ -146,6 +146,10 @@ out: int smc_ism_unregister_dmb(struct smcd_dev *smcd, struct smc_buf_desc *dmb_desc) { struct smcd_dmb dmb; + int rc = 0; + + if (!dmb_desc->dma_addr) + return rc; memset(&dmb, 0, sizeof(dmb)); dmb.dmb_tok = dmb_desc->token; @@ -153,7 +157,13 @@ int smc_ism_unregister_dmb(struct smcd_dev *smcd, struct smc_buf_desc *dmb_desc) dmb.cpu_addr = dmb_desc->cpu_addr; dmb.dma_addr = dmb_desc->dma_addr; dmb.dmb_len = dmb_desc->len; - return smcd->ops->unregister_dmb(smcd, &dmb); + rc = smcd->ops->unregister_dmb(smcd, &dmb); + if (!rc || rc == ISM_ERROR) { + dmb_desc->cpu_addr = NULL; + dmb_desc->dma_addr = 0; + } + + return rc; } int smc_ism_register_dmb(struct smc_link_group *lgr, int dmb_len, @@ -375,7 +385,7 @@ void smcd_handle_irq(struct smcd_dev *smcd, unsigned int dmbno) spin_lock_irqsave(&smcd->lock, flags); conn = smcd->conn[dmbno]; - if (conn) + if (conn && !conn->killed) tasklet_schedule(&conn->rx_tsklet); spin_unlock_irqrestore(&smcd->lock, flags); } -- cgit v1.2.3 From 5421ec281df9dfda4418c02959e1f76097cabd9a Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:42 +0100 Subject: net/smc: abnormal termination of SMCD link groups A final cleanup due to SMCD device removal means immediate freeing of all link groups belonging to this device in interrupt context. This patch introduces a separate SMCD link group termination routine, which terminates all link groups of an SMCD device. This new routine smcd_terminate_all ()is reused if the smc module is unloaded. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_clc.c | 2 +- net/smc/smc_core.c | 67 +++++++++++++++++++++++++++++++++++++++++------------- net/smc/smc_core.h | 3 ++- net/smc/smc_ism.c | 2 +- net/smc/smc_llc.c | 2 +- net/smc/smc_tx.c | 2 +- 6 files changed, 57 insertions(+), 21 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_clc.c b/net/smc/smc_clc.c index 49bcebff6378..0879f7bed967 100644 --- a/net/smc/smc_clc.c +++ b/net/smc/smc_clc.c @@ -349,7 +349,7 @@ int smc_clc_wait_msg(struct smc_sock *smc, void *buf, int buflen, smc->peer_diagnosis = ntohl(dclc->peer_diagnosis); if (((struct smc_clc_msg_decline *)buf)->hdr.flag) { smc->conn.lgr->sync_err = 1; - smc_lgr_terminate(smc->conn.lgr); + smc_lgr_terminate(smc->conn.lgr, true); } } diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 9d6da2c7413d..d79dd78c1cd8 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -224,7 +224,7 @@ static void smc_lgr_terminate_work(struct work_struct *work) struct smc_link_group *lgr = container_of(work, struct smc_link_group, terminate_work); - smc_lgr_terminate(lgr); + smc_lgr_terminate(lgr, true); } /* create a new SMC link group */ @@ -528,7 +528,7 @@ static void smc_sk_wake_ups(struct smc_sock *smc) } /* kill a connection */ -static void smc_conn_kill(struct smc_connection *conn) +static void smc_conn_kill(struct smc_connection *conn, bool soft) { struct smc_sock *smc = container_of(conn, struct smc_sock, conn); @@ -541,7 +541,10 @@ static void smc_conn_kill(struct smc_connection *conn) smc_sk_wake_ups(smc); if (conn->lgr->is_smcd) { smc_ism_unset_conn(conn); - tasklet_kill(&conn->rx_tsklet); + if (soft) + tasklet_kill(&conn->rx_tsklet); + else + tasklet_unlock_wait(&conn->rx_tsklet); } smc_lgr_unregister_conn(conn); smc_close_active_abort(smc); @@ -562,7 +565,7 @@ static void smc_lgr_cleanup(struct smc_link_group *lgr) } /* terminate link group */ -static void __smc_lgr_terminate(struct smc_link_group *lgr) +static void __smc_lgr_terminate(struct smc_link_group *lgr, bool soft) { struct smc_connection *conn; struct smc_sock *smc; @@ -570,6 +573,8 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) if (lgr->terminating) return; /* lgr already terminating */ + if (!soft) + cancel_delayed_work_sync(&lgr->free_work); lgr->terminating = 1; if (!lgr->is_smcd) smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]); @@ -583,7 +588,7 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) smc = container_of(conn, struct smc_sock, conn); sock_hold(&smc->sk); /* sock_put below */ lock_sock(&smc->sk); - smc_conn_kill(conn); + smc_conn_kill(conn, soft); release_sock(&smc->sk); sock_put(&smc->sk); /* sock_hold above */ read_lock_bh(&lgr->conns_lock); @@ -591,11 +596,17 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr) } read_unlock_bh(&lgr->conns_lock); smc_lgr_cleanup(lgr); - smc_lgr_schedule_free_work_fast(lgr); + if (soft) + smc_lgr_schedule_free_work_fast(lgr); + else + smc_lgr_free(lgr); } -/* unlink and terminate link group */ -void smc_lgr_terminate(struct smc_link_group *lgr) +/* unlink and terminate link group + * @soft: true if link group shutdown can take its time + * false if immediate link group shutdown is required + */ +void smc_lgr_terminate(struct smc_link_group *lgr, bool soft) { spinlock_t *lgr_lock; @@ -605,9 +616,11 @@ void smc_lgr_terminate(struct smc_link_group *lgr) spin_unlock_bh(lgr_lock); return; /* lgr already terminating */ } + if (!soft) + lgr->freeing = 1; list_del_init(&lgr->list); spin_unlock_bh(lgr_lock); - __smc_lgr_terminate(lgr); + __smc_lgr_terminate(lgr, soft); } /* Called when IB port is terminated */ @@ -627,11 +640,11 @@ void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport) list_for_each_entry_safe(lgr, l, &lgr_free_list, list) { list_del_init(&lgr->list); - __smc_lgr_terminate(lgr); + __smc_lgr_terminate(lgr, true); } } -/* Called when SMC-D device is terminated or peer is lost */ +/* Called when peer lgr shutdown (regularly or abnormally) is received */ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) { struct smc_link_group *lgr, *l; @@ -656,6 +669,24 @@ void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan) } } +/* Called when an SMCD device is removed or the smc module is unloaded */ +void smc_smcd_terminate_all(struct smcd_dev *smcd) +{ + struct smc_link_group *lgr, *lg; + LIST_HEAD(lgr_free_list); + + spin_lock_bh(&smcd->lgr_lock); + list_splice_init(&smcd->lgr_list, &lgr_free_list); + list_for_each_entry(lgr, &lgr_free_list, list) + lgr->freeing = 1; + spin_unlock_bh(&smcd->lgr_lock); + + list_for_each_entry_safe(lgr, lg, &lgr_free_list, list) { + list_del_init(&lgr->list); + __smc_lgr_terminate(lgr, false); + } +} + /* Determine vlan of internal TCP socket. * @vlan_id: address to store the determined vlan id into */ @@ -1173,8 +1204,8 @@ static void smc_core_going_away(void) spin_unlock(&smcd_dev_list.lock); } -/* Called (from smc_exit) when module is removed */ -void smc_core_exit(void) +/* Clean up all SMC link groups */ +static void smc_lgrs_shutdown(void) { struct smc_link_group *lgr, *lg; LIST_HEAD(lgr_freeing_list); @@ -1188,7 +1219,7 @@ void smc_core_exit(void) spin_lock(&smcd_dev_list.lock); list_for_each_entry(smcd, &smcd_dev_list.list, list) - list_splice_init(&smcd->lgr_list, &lgr_freeing_list); + smc_smcd_terminate_all(smcd); spin_unlock(&smcd_dev_list.lock); list_for_each_entry_safe(lgr, lg, &lgr_freeing_list, list) { @@ -1202,8 +1233,12 @@ void smc_core_exit(void) smc_llc_link_inactive(lnk); } cancel_delayed_work_sync(&lgr->free_work); - if (lgr->is_smcd) - smc_ism_signal_shutdown(lgr); smc_lgr_free(lgr); /* free link group */ } } + +/* Called (from smc_exit) when module is removed */ +void smc_core_exit(void) +{ + smc_lgrs_shutdown(); +} diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 097ceba86caf..7f34f4d5a514 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -296,10 +296,11 @@ struct smc_clc_msg_accept_confirm; struct smc_clc_msg_local; void smc_lgr_forget(struct smc_link_group *lgr); -void smc_lgr_terminate(struct smc_link_group *lgr); +void smc_lgr_terminate(struct smc_link_group *lgr, bool soft); void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport); void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan); +void smc_smcd_terminate_all(struct smcd_dev *dev); int smc_buf_create(struct smc_sock *smc, bool is_smcd); int smc_uncompress_bufsize(u8 compressed); int smc_rmb_rtoken_handling(struct smc_connection *conn, diff --git a/net/smc/smc_ism.c b/net/smc/smc_ism.c index 903da947b20d..56cdab8be1fa 100644 --- a/net/smc/smc_ism.c +++ b/net/smc/smc_ism.c @@ -329,7 +329,7 @@ void smcd_unregister_dev(struct smcd_dev *smcd) list_del_init(&smcd->list); spin_unlock(&smcd_dev_list.lock); smcd->going_away = 1; - smc_smcd_terminate(smcd, 0, VLAN_VID_MASK); + smc_smcd_terminate_all(smcd); flush_workqueue(smcd->event_wq); destroy_workqueue(smcd->event_wq); diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c index e1918ffaf125..26a18c872455 100644 --- a/net/smc/smc_llc.c +++ b/net/smc/smc_llc.c @@ -614,7 +614,7 @@ static void smc_llc_testlink_work(struct work_struct *work) rc = wait_for_completion_interruptible_timeout(&link->llc_testlink_resp, SMC_LLC_WAIT_TIME); if (rc <= 0) { - smc_lgr_terminate(smc_get_lgr(link)); + smc_lgr_terminate(smc_get_lgr(link), true); return; } next_interval = link->llc_testlink_time; diff --git a/net/smc/smc_tx.c b/net/smc/smc_tx.c index 824f096ee7de..0d42e7716b91 100644 --- a/net/smc/smc_tx.c +++ b/net/smc/smc_tx.c @@ -284,7 +284,7 @@ static int smc_tx_rdma_write(struct smc_connection *conn, int peer_rmbe_offset, rdma_wr->rkey = lgr->rtokens[conn->rtoken_idx][SMC_SINGLE_LINK].rkey; rc = ib_post_send(link->roce_qp, &rdma_wr->wr, NULL); if (rc) - smc_lgr_terminate(lgr); + smc_lgr_terminate(lgr, true); return rc; } -- cgit v1.2.3 From 5edd6b9cb8d7c6c346c93c52a53735591127e879 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:43 +0100 Subject: net/smc: introduce bookkeeping of SMCD link groups If the ism module is unloaded return control from exit routine only, if all link groups are freed. If an IB device is thrown away return control from device removal only, if all link groups belonging to this device are freed. A counters for the total number of SMCD link groups per ISM device is introduced. ism module unloading continues only if the total number of SMCD link groups for all ISM devices is zero. ISM device removal continues only it the total number of SMCD link groups per ISM device has decreased to zero. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_core.c | 6 ++++++ net/smc/smc_ism.c | 1 + 2 files changed, 7 insertions(+) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index d79dd78c1cd8..30854acb846c 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -276,6 +276,7 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) lgr_list = &ini->ism_dev->lgr_list; lgr_lock = &lgr->smcd->lgr_lock; lgr->peer_shutdown = 0; + atomic_inc(&ini->ism_dev->lgr_cnt); } else { /* SMC-R specific settings */ get_device(&ini->ib_dev->ibdev->dev); @@ -486,6 +487,8 @@ static void smc_lgr_free(struct smc_link_group *lgr) smc_ism_put_vlan(lgr->smcd, lgr->vlan_id); put_device(&lgr->smcd->dev); } + if (!atomic_dec_return(&lgr->smcd->lgr_cnt)) + wake_up(&lgr->smcd->lgrs_deleted); } else { smc_link_clear(&lgr->lnk[SMC_SINGLE_LINK]); put_device(&lgr->lnk[SMC_SINGLE_LINK].smcibdev->ibdev->dev); @@ -685,6 +688,9 @@ void smc_smcd_terminate_all(struct smcd_dev *smcd) list_del_init(&lgr->list); __smc_lgr_terminate(lgr, false); } + + if (atomic_read(&smcd->lgr_cnt)) + wait_event(smcd->lgrs_deleted, !atomic_read(&smcd->lgr_cnt)); } /* Determine vlan of internal TCP socket. diff --git a/net/smc/smc_ism.c b/net/smc/smc_ism.c index 56cdab8be1fa..5c4727d5066e 100644 --- a/net/smc/smc_ism.c +++ b/net/smc/smc_ism.c @@ -302,6 +302,7 @@ struct smcd_dev *smcd_alloc_dev(struct device *parent, const char *name, spin_lock_init(&smcd->lgr_lock); INIT_LIST_HEAD(&smcd->vlan); INIT_LIST_HEAD(&smcd->lgr_list); + init_waitqueue_head(&smcd->lgrs_deleted); smcd->event_wq = alloc_ordered_workqueue("ism_evt_wq-%s)", WQ_MEM_RECLAIM, name); if (!smcd->event_wq) { -- cgit v1.2.3 From 15e1b99aadfb2766f9379a23a0fc1d4336c8cd8e Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:44 +0100 Subject: net/smc: no WR buffer wait for terminating link group Avoid waiting for a free work request buffer, if the link group is already terminating. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_llc.c | 3 +++ net/smc/smc_wr.c | 10 ++++++---- net/smc/smc_wr.h | 10 ++++++++++ 3 files changed, 19 insertions(+), 4 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c index 26a18c872455..8d1b076021ed 100644 --- a/net/smc/smc_llc.c +++ b/net/smc/smc_llc.c @@ -656,6 +656,7 @@ void smc_llc_link_active(struct smc_link *link, int testlink_time) void smc_llc_link_deleting(struct smc_link *link) { link->state = SMC_LNK_DELETING; + smc_wr_wakeup_tx_wait(link); } /* called in tasklet context */ @@ -663,6 +664,8 @@ void smc_llc_link_inactive(struct smc_link *link) { link->state = SMC_LNK_INACTIVE; cancel_delayed_work(&link->llc_testlink_wrk); + smc_wr_wakeup_reg_wait(link); + smc_wr_wakeup_tx_wait(link); } /* called in worker context */ diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c index 50743dc56c86..619dd89fbac0 100644 --- a/net/smc/smc_wr.c +++ b/net/smc/smc_wr.c @@ -75,7 +75,7 @@ static inline void smc_wr_tx_process_cqe(struct ib_wc *wc) link->wr_reg_state = FAILED; else link->wr_reg_state = CONFIRMED; - wake_up(&link->wr_reg_wait); + smc_wr_wakeup_reg_wait(link); return; } @@ -171,6 +171,7 @@ int smc_wr_tx_get_free_slot(struct smc_link *link, struct smc_rdma_wr **wr_rdma_buf, struct smc_wr_tx_pend_priv **wr_pend_priv) { + struct smc_link_group *lgr = smc_get_lgr(link); struct smc_wr_tx_pend *wr_pend; u32 idx = link->wr_tx_cnt; struct ib_send_wr *wr_ib; @@ -179,19 +180,20 @@ int smc_wr_tx_get_free_slot(struct smc_link *link, *wr_buf = NULL; *wr_pend_priv = NULL; - if (in_softirq()) { + if (in_softirq() || lgr->terminating) { rc = smc_wr_tx_get_free_slot_index(link, &idx); if (rc) return rc; } else { - rc = wait_event_timeout( + rc = wait_event_interruptible_timeout( link->wr_tx_wait, link->state == SMC_LNK_INACTIVE || + lgr->terminating || (smc_wr_tx_get_free_slot_index(link, &idx) != -EBUSY), SMC_WR_TX_WAIT_FREE_SLOT_TIME); if (!rc) { /* timeout - terminate connections */ - smc_lgr_terminate_sched(smc_get_lgr(link)); + smc_lgr_terminate_sched(lgr); return -EPIPE; } if (idx == link->wr_tx_cnt) diff --git a/net/smc/smc_wr.h b/net/smc/smc_wr.h index 09bf32fd3959..3ac99c898418 100644 --- a/net/smc/smc_wr.h +++ b/net/smc/smc_wr.h @@ -60,6 +60,16 @@ static inline void smc_wr_tx_set_wr_id(atomic_long_t *wr_tx_id, long val) atomic_long_set(wr_tx_id, val); } +static inline void smc_wr_wakeup_tx_wait(struct smc_link *lnk) +{ + wake_up_all(&lnk->wr_tx_wait); +} + +static inline void smc_wr_wakeup_reg_wait(struct smc_link *lnk) +{ + wake_up(&lnk->wr_reg_wait); +} + /* post a new receive work request to fill a completed old work request entry */ static inline int smc_wr_rx_post(struct smc_link *link) { -- cgit v1.2.3 From 2c1d3e50302fe3e0bd6873323877c2ad19db3f49 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:45 +0100 Subject: net/smc: abnormal termination without orderly flag For abnormal termination issue an LLC DELETE_LINK without the orderly flag. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_core.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 30854acb846c..ee44e8244d0c 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -161,10 +161,10 @@ static void smc_lgr_unregister_conn(struct smc_connection *conn) * of the DELETE LINK sequence from server; or as server to * initiate the delete processing. See smc_llc_rx_delete_link(). */ -static int smc_link_send_delete(struct smc_link *lnk) +static int smc_link_send_delete(struct smc_link *lnk, bool orderly) { if (lnk->state == SMC_LNK_ACTIVE && - !smc_llc_send_delete_link(lnk, SMC_LLC_REQ, true)) { + !smc_llc_send_delete_link(lnk, SMC_LLC_REQ, orderly)) { smc_llc_link_deleting(lnk); return 0; } @@ -201,7 +201,7 @@ static void smc_lgr_free_work(struct work_struct *work) if (!lgr->is_smcd && !lgr->terminating) { /* try to send del link msg, on error free lgr immediately */ if (lnk->state == SMC_LNK_ACTIVE && - !smc_link_send_delete(lnk)) { + !smc_link_send_delete(lnk, true)) { /* reschedule in case we never receive a response */ smc_lgr_schedule_free_work(lgr); spin_unlock_bh(lgr_lock); @@ -1233,9 +1233,7 @@ static void smc_lgrs_shutdown(void) if (!lgr->is_smcd) { struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK]; - if (lnk->state == SMC_LNK_ACTIVE) - smc_llc_send_delete_link(lnk, SMC_LLC_REQ, - false); + smc_link_send_delete(&lgr->lnk[SMC_SINGLE_LINK], false); smc_llc_link_inactive(lnk); } cancel_delayed_work_sync(&lgr->free_work); -- cgit v1.2.3 From 6a37ad3da5d64a632d03a8dc272c65e706cc7160 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:46 +0100 Subject: net/smc: wait for tx completions before link freeing Make sure all pending work requests are completed before freeing a link. Dismiss tx pending slots already when terminating a link group to exploit termination shortcut in tx completion queue handler. And kill the completion queue tasklets after destroy of the completion queues, otherwise there is a time window for another tasklet schedule of an already killed tasklet. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_core.c | 2 ++ net/smc/smc_ib.c | 2 +- net/smc/smc_wr.c | 27 +++++++++++++++++++++++++-- 3 files changed, 28 insertions(+), 3 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index ee44e8244d0c..0755bd4b587c 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -548,6 +548,8 @@ static void smc_conn_kill(struct smc_connection *conn, bool soft) tasklet_kill(&conn->rx_tsklet); else tasklet_unlock_wait(&conn->rx_tsklet); + } else { + smc_cdc_tx_dismiss_slots(conn); } smc_lgr_unregister_conn(conn); smc_close_active_abort(smc); diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c index af05daeb0538..c15dcd08dc74 100644 --- a/net/smc/smc_ib.c +++ b/net/smc/smc_ib.c @@ -520,9 +520,9 @@ static void smc_ib_cleanup_per_ibdev(struct smc_ib_device *smcibdev) if (!smcibdev->initialized) return; smcibdev->initialized = 0; - smc_wr_remove_dev(smcibdev); ib_destroy_cq(smcibdev->roce_cq_recv); ib_destroy_cq(smcibdev->roce_cq_send); + smc_wr_remove_dev(smcibdev); } static struct ib_client smc_ib_client; diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c index 619dd89fbac0..337ee52ad3d3 100644 --- a/net/smc/smc_wr.c +++ b/net/smc/smc_wr.c @@ -50,6 +50,26 @@ struct smc_wr_tx_pend { /* control data for a pending send request */ /*------------------------------- completion --------------------------------*/ +/* returns true if at least one tx work request is pending on the given link */ +static inline bool smc_wr_is_tx_pend(struct smc_link *link) +{ + if (find_first_bit(link->wr_tx_mask, link->wr_tx_cnt) != + link->wr_tx_cnt) { + return true; + } + return false; +} + +/* wait till all pending tx work requests on the given link are completed */ +static inline int smc_wr_tx_wait_no_pending_sends(struct smc_link *link) +{ + if (wait_event_timeout(link->wr_tx_wait, !smc_wr_is_tx_pend(link), + SMC_WR_TX_WAIT_PENDING_TIME)) + return 0; + else /* timeout */ + return -EPIPE; +} + static inline int smc_wr_tx_find_pending_index(struct smc_link *link, u64 wr_id) { u32 i; @@ -229,6 +249,7 @@ int smc_wr_tx_put_slot(struct smc_link *link, memset(&link->wr_tx_bufs[idx], 0, sizeof(link->wr_tx_bufs[idx])); test_and_clear_bit(idx, link->wr_tx_mask); + wake_up(&link->wr_tx_wait); return 1; } @@ -512,8 +533,10 @@ void smc_wr_free_link(struct smc_link *lnk) { struct ib_device *ibdev; - memset(lnk->wr_tx_mask, 0, - BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*lnk->wr_tx_mask)); + if (smc_wr_tx_wait_no_pending_sends(lnk)) + memset(lnk->wr_tx_mask, 0, + BITS_TO_LONGS(SMC_WR_BUF_CNT) * + sizeof(*lnk->wr_tx_mask)); if (!lnk->smcibdev) return; -- cgit v1.2.3 From 0b29ec6436138721acf5844e558f7334a0fa61d5 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Thu, 14 Nov 2019 13:02:47 +0100 Subject: net/smc: immediate termination for SMCR link groups If the SMC module is unloaded or an IB device is thrown away, the immediate link group freeing introduced for SMCD is exploited for SMCR as well. That means SMCR-specifics are added to smc_conn_kill(). Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_core.c | 58 ++++++++++++++++++++++++++++++++++++------------------ net/smc/smc_core.h | 3 ++- net/smc/smc_ib.c | 3 ++- net/smc/smc_llc.c | 4 +++- 4 files changed, 46 insertions(+), 22 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 0755bd4b587c..97e9d21c4d1e 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -566,6 +566,10 @@ static void smc_lgr_cleanup(struct smc_link_group *lgr) struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK]; wake_up(&lnk->wr_reg_wait); + if (lnk->state != SMC_LNK_INACTIVE) { + smc_link_send_delete(lnk, false); + smc_llc_link_inactive(lnk); + } } } @@ -638,14 +642,16 @@ void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport) list_for_each_entry_safe(lgr, l, &smc_lgr_list.list, list) { if (!lgr->is_smcd && lgr->lnk[SMC_SINGLE_LINK].smcibdev == smcibdev && - lgr->lnk[SMC_SINGLE_LINK].ibport == ibport) + lgr->lnk[SMC_SINGLE_LINK].ibport == ibport) { list_move(&lgr->list, &lgr_free_list); + lgr->freeing = 1; + } } spin_unlock_bh(&smc_lgr_list.lock); list_for_each_entry_safe(lgr, l, &lgr_free_list, list) { list_del_init(&lgr->list); - __smc_lgr_terminate(lgr, true); + __smc_lgr_terminate(lgr, false); } } @@ -695,6 +701,36 @@ void smc_smcd_terminate_all(struct smcd_dev *smcd) wait_event(smcd->lgrs_deleted, !atomic_read(&smcd->lgr_cnt)); } +/* Called when an SMCR device is removed or the smc module is unloaded. + * If smcibdev is given, all SMCR link groups using this device are terminated. + * If smcibdev is NULL, all SMCR link groups are terminated. + */ +void smc_smcr_terminate_all(struct smc_ib_device *smcibdev) +{ + struct smc_link_group *lgr, *lg; + LIST_HEAD(lgr_free_list); + + spin_lock_bh(&smc_lgr_list.lock); + if (!smcibdev) { + list_splice_init(&smc_lgr_list.list, &lgr_free_list); + list_for_each_entry(lgr, &lgr_free_list, list) + lgr->freeing = 1; + } else { + list_for_each_entry_safe(lgr, lg, &smc_lgr_list.list, list) { + if (lgr->lnk[SMC_SINGLE_LINK].smcibdev == smcibdev) { + list_move(&lgr->list, &lgr_free_list); + lgr->freeing = 1; + } + } + } + spin_unlock_bh(&smc_lgr_list.lock); + + list_for_each_entry_safe(lgr, lg, &lgr_free_list, list) { + list_del_init(&lgr->list); + __smc_lgr_terminate(lgr, false); + } +} + /* Determine vlan of internal TCP socket. * @vlan_id: address to store the determined vlan id into */ @@ -1215,32 +1251,16 @@ static void smc_core_going_away(void) /* Clean up all SMC link groups */ static void smc_lgrs_shutdown(void) { - struct smc_link_group *lgr, *lg; - LIST_HEAD(lgr_freeing_list); struct smcd_dev *smcd; smc_core_going_away(); - spin_lock_bh(&smc_lgr_list.lock); - list_splice_init(&smc_lgr_list.list, &lgr_freeing_list); - spin_unlock_bh(&smc_lgr_list.lock); + smc_smcr_terminate_all(NULL); spin_lock(&smcd_dev_list.lock); list_for_each_entry(smcd, &smcd_dev_list.list, list) smc_smcd_terminate_all(smcd); spin_unlock(&smcd_dev_list.lock); - - list_for_each_entry_safe(lgr, lg, &lgr_freeing_list, list) { - list_del_init(&lgr->list); - if (!lgr->is_smcd) { - struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK]; - - smc_link_send_delete(&lgr->lnk[SMC_SINGLE_LINK], false); - smc_llc_link_inactive(lnk); - } - cancel_delayed_work_sync(&lgr->free_work); - smc_lgr_free(lgr); /* free link group */ - } } /* Called (from smc_exit) when module is removed */ diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 7f34f4d5a514..a428db6cd2e2 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -287,7 +287,7 @@ static inline struct smc_connection *smc_lgr_find_conn( static inline void smc_lgr_terminate_sched(struct smc_link_group *lgr) { - if (!lgr->terminating) + if (!lgr->terminating && !lgr->freeing) schedule_work(&lgr->terminate_work); } @@ -301,6 +301,7 @@ void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport); void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid, unsigned short vlan); void smc_smcd_terminate_all(struct smcd_dev *dev); +void smc_smcr_terminate_all(struct smc_ib_device *smcibdev); int smc_buf_create(struct smc_sock *smc, bool is_smcd); int smc_uncompress_bufsize(u8 compressed); int smc_rmb_rtoken_handling(struct smc_connection *conn, diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c index c15dcd08dc74..0ab122e66328 100644 --- a/net/smc/smc_ib.c +++ b/net/smc/smc_ib.c @@ -565,7 +565,7 @@ static void smc_ib_add_dev(struct ib_device *ibdev) schedule_work(&smcibdev->port_event_work); } -/* callback function for ib_register_client() */ +/* callback function for ib_unregister_client() */ static void smc_ib_remove_dev(struct ib_device *ibdev, void *client_data) { struct smc_ib_device *smcibdev; @@ -575,6 +575,7 @@ static void smc_ib_remove_dev(struct ib_device *ibdev, void *client_data) spin_lock(&smc_ib_devices.lock); list_del_init(&smcibdev->list); /* remove from smc_ib_devices */ spin_unlock(&smc_ib_devices.lock); + smc_smcr_terminate_all(smcibdev); smc_ib_cleanup_per_ibdev(smcibdev); ib_unregister_event_handler(&smcibdev->event_handler); kfree(smcibdev); diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c index 8d1b076021ed..a9f6431dd69a 100644 --- a/net/smc/smc_llc.c +++ b/net/smc/smc_llc.c @@ -698,9 +698,11 @@ int smc_llc_do_confirm_rkey(struct smc_link *link, int smc_llc_do_delete_rkey(struct smc_link *link, struct smc_buf_desc *rmb_desc) { - int rc; + int rc = 0; mutex_lock(&link->llc_delete_rkey_mutex); + if (link->state != SMC_LNK_ACTIVE) + goto out; reinit_completion(&link->llc_delete_rkey); rc = smc_llc_send_delete_rkey(link, rmb_desc); if (rc) -- cgit v1.2.3 From 6dabd405451f35c905dfadb6a06f5c981074fc14 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Sat, 16 Nov 2019 17:47:29 +0100 Subject: net/smc: introduce bookkeeping of SMCR link groups If the smc module is unloaded return control from exit routine only, if all link groups are freed. If an IB device is thrown away return control from device removal only, if all link groups belonging to this device are freed. Counters for the total number of SMCR link groups and for the total number of SMCR links per IB device are introduced. smc module unloading continues only if the total number of SMCR link groups is zero. IB device removal continues only it the total number of SMCR links per IB device has decreased to zero. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/af_smc.c | 18 +++++++++++++----- net/smc/smc_core.c | 25 +++++++++++++++++++++++++ net/smc/smc_core.h | 1 + net/smc/smc_ib.c | 4 +++- net/smc/smc_ib.h | 3 +++ 5 files changed, 45 insertions(+), 6 deletions(-) (limited to 'net/smc') diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index b7d9fd285c71..42b7fb8ab22b 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -2038,22 +2038,28 @@ static int __init smc_init(void) if (rc) goto out_pernet_subsys; + rc = smc_core_init(); + if (rc) { + pr_err("%s: smc_core_init fails with %d\n", __func__, rc); + goto out_pnet; + } + rc = smc_llc_init(); if (rc) { pr_err("%s: smc_llc_init fails with %d\n", __func__, rc); - goto out_pnet; + goto out_core; } rc = smc_cdc_init(); if (rc) { pr_err("%s: smc_cdc_init fails with %d\n", __func__, rc); - goto out_pnet; + goto out_core; } rc = proto_register(&smc_proto, 1); if (rc) { pr_err("%s: proto_register(v4) fails with %d\n", __func__, rc); - goto out_pnet; + goto out_core; } rc = proto_register(&smc_proto6, 1); @@ -2085,6 +2091,8 @@ out_proto6: proto_unregister(&smc_proto6); out_proto: proto_unregister(&smc_proto); +out_core: + smc_core_exit(); out_pnet: smc_pnet_exit(); out_pernet_subsys: @@ -2095,10 +2103,10 @@ out_pernet_subsys: static void __exit smc_exit(void) { - smc_core_exit(); static_branch_disable(&tcp_have_smc); - smc_ib_unregister_client(); sock_unregister(PF_SMC); + smc_core_exit(); + smc_ib_unregister_client(); proto_unregister(&smc_proto6); proto_unregister(&smc_proto); smc_pnet_exit(); diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 97e9d21c4d1e..cf34b9d96595 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -13,6 +13,7 @@ #include #include #include +#include #include #include #include @@ -39,6 +40,9 @@ static struct smc_lgr_list smc_lgr_list = { /* established link groups */ .num = 0, }; +static atomic_t lgr_cnt; /* number of existing link groups */ +static DECLARE_WAIT_QUEUE_HEAD(lgrs_deleted); + static void smc_buf_free(struct smc_link_group *lgr, bool is_rmb, struct smc_buf_desc *buf_desc); @@ -319,6 +323,8 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) rc = smc_wr_create_link(lnk); if (rc) goto destroy_qp; + atomic_inc(&lgr_cnt); + atomic_inc(&ini->ib_dev->lnk_cnt); } smc->conn.lgr = lgr; spin_lock_bh(lgr_lock); @@ -406,6 +412,8 @@ static void smc_link_clear(struct smc_link *lnk) smc_ib_destroy_queue_pair(lnk); smc_ib_dealloc_protection_domain(lnk); smc_wr_free_link_mem(lnk); + if (!atomic_dec_return(&lnk->smcibdev->lnk_cnt)) + wake_up(&lnk->smcibdev->lnks_deleted); } static void smcr_buf_free(struct smc_link_group *lgr, bool is_rmb, @@ -492,6 +500,8 @@ static void smc_lgr_free(struct smc_link_group *lgr) } else { smc_link_clear(&lgr->lnk[SMC_SINGLE_LINK]); put_device(&lgr->lnk[SMC_SINGLE_LINK].smcibdev->ibdev->dev); + if (!atomic_dec_return(&lgr_cnt)) + wake_up(&lgrs_deleted); } kfree(lgr); } @@ -729,6 +739,15 @@ void smc_smcr_terminate_all(struct smc_ib_device *smcibdev) list_del_init(&lgr->list); __smc_lgr_terminate(lgr, false); } + + if (smcibdev) { + if (atomic_read(&smcibdev->lnk_cnt)) + wait_event(smcibdev->lnks_deleted, + !atomic_read(&smcibdev->lnk_cnt)); + } else { + if (atomic_read(&lgr_cnt)) + wait_event(lgrs_deleted, !atomic_read(&lgr_cnt)); + } } /* Determine vlan of internal TCP socket. @@ -1263,6 +1282,12 @@ static void smc_lgrs_shutdown(void) spin_unlock(&smcd_dev_list.lock); } +int __init smc_core_init(void) +{ + atomic_set(&lgr_cnt, 0); + return 0; +} + /* Called (from smc_exit) when module is removed */ void smc_core_exit(void) { diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index a428db6cd2e2..c472e12951d1 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -318,6 +318,7 @@ void smc_conn_free(struct smc_connection *conn); int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini); void smcd_conn_free(struct smc_connection *conn); void smc_lgr_schedule_free_work_fast(struct smc_link_group *lgr); +int smc_core_init(void); void smc_core_exit(void); static inline struct smc_link_group *smc_get_lgr(struct smc_link *link) diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c index 0ab122e66328..548632621f4b 100644 --- a/net/smc/smc_ib.c +++ b/net/smc/smc_ib.c @@ -15,6 +15,7 @@ #include #include #include +#include #include #include @@ -543,7 +544,8 @@ static void smc_ib_add_dev(struct ib_device *ibdev) smcibdev->ibdev = ibdev; INIT_WORK(&smcibdev->port_event_work, smc_ib_port_event_work); - + atomic_set(&smcibdev->lnk_cnt, 0); + init_waitqueue_head(&smcibdev->lnks_deleted); spin_lock(&smc_ib_devices.lock); list_add_tail(&smcibdev->list, &smc_ib_devices.list); spin_unlock(&smc_ib_devices.lock); diff --git a/net/smc/smc_ib.h b/net/smc/smc_ib.h index 6a0069db6cae..255db87547d3 100644 --- a/net/smc/smc_ib.h +++ b/net/smc/smc_ib.h @@ -14,6 +14,7 @@ #include #include +#include #include #include @@ -48,6 +49,8 @@ struct smc_ib_device { /* ib-device infos for smc */ struct work_struct port_event_work; unsigned long port_event_mask; DECLARE_BITMAP(ports_going_away, SMC_MAX_PORTS); + atomic_t lnk_cnt; /* number of links on ibdev */ + wait_queue_head_t lnks_deleted; /* wait 4 removal of all links*/ }; struct smc_buf_desc; -- cgit v1.2.3 From a33a803cfe64309d330540ae4a8df17158bcb6ea Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Sat, 16 Nov 2019 17:47:30 +0100 Subject: net/smc: guarantee removal of link groups in reboot When rebooting it should be guaranteed all link groups are cleaned up and freed. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_core.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) (limited to 'net/smc') diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index cf34b9d96595..bb92c7c6214c 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -14,6 +14,7 @@ #include #include #include +#include #include #include #include @@ -1282,14 +1283,27 @@ static void smc_lgrs_shutdown(void) spin_unlock(&smcd_dev_list.lock); } +static int smc_core_reboot_event(struct notifier_block *this, + unsigned long event, void *ptr) +{ + smc_lgrs_shutdown(); + + return 0; +} + +static struct notifier_block smc_reboot_notifier = { + .notifier_call = smc_core_reboot_event, +}; + int __init smc_core_init(void) { atomic_set(&lgr_cnt, 0); - return 0; + return register_reboot_notifier(&smc_reboot_notifier); } /* Called (from smc_exit) when module is removed */ void smc_core_exit(void) { + unregister_reboot_notifier(&smc_reboot_notifier); smc_lgrs_shutdown(); } -- cgit v1.2.3 From 4ead9c96d528e1b9937382321910a2bf35fc1a86 Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Sat, 16 Nov 2019 17:47:31 +0100 Subject: net/smc: use rcu_barrier() on module unload Add rcu_barrier() to make sure no RCU readers or callbacks are pending when the module is unloaded. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/af_smc.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'net/smc') diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index 42b7fb8ab22b..cde4dc0ed173 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -25,6 +25,7 @@ #include #include #include +#include #include #include @@ -2111,6 +2112,7 @@ static void __exit smc_exit(void) proto_unregister(&smc_proto); smc_pnet_exit(); unregister_pernet_subsys(&smc_net_ops); + rcu_barrier(); } module_init(smc_init); -- cgit v1.2.3 From ab8536ca783db9ef863e0a2246946ebae701df5a Mon Sep 17 00:00:00 2001 From: Ursula Braun Date: Sat, 16 Nov 2019 17:47:32 +0100 Subject: net/smc: remove unused constant Constant SMC_CLOSE_WAIT_LISTEN_CLCSOCK_TIME is defined, but since commit 3d502067599f ("net/smc: simplify wait when closing listen socket") no longer used. Remove it. Signed-off-by: Ursula Braun Signed-off-by: Karsten Graul Signed-off-by: David S. Miller --- net/smc/smc_close.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'net/smc') diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c index d205b2114006..290270c821ca 100644 --- a/net/smc/smc_close.c +++ b/net/smc/smc_close.c @@ -20,8 +20,6 @@ #include "smc_cdc.h" #include "smc_close.h" -#define SMC_CLOSE_WAIT_LISTEN_CLCSOCK_TIME (5 * HZ) - /* release the clcsock that is assigned to the smc_sock */ void smc_clcsock_release(struct smc_sock *smc) { -- cgit v1.2.3