diff options
Diffstat (limited to 'drivers/infiniband/ulp')
19 files changed, 1226 insertions, 1842 deletions
diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 3ede10309754..85be0de3ab26 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -244,6 +244,7 @@ struct ipoib_cm_tx { unsigned tx_tail; unsigned long flags; u32 mtu; + unsigned max_send_sge; }; struct ipoib_cm_rx_buf { @@ -390,6 +391,7 @@ struct ipoib_dev_priv { int hca_caps; struct ipoib_ethtool_st ethtool; struct timer_list poll_timer; + unsigned max_send_sge; }; struct ipoib_ah { @@ -495,7 +497,6 @@ void ipoib_dev_cleanup(struct net_device *dev); void ipoib_mcast_join_task(struct work_struct *work); void ipoib_mcast_carrier_on_task(struct work_struct *work); void ipoib_mcast_send(struct net_device *dev, u8 *daddr, struct sk_buff *skb); -void ipoib_mcast_free(struct ipoib_mcast *mc); void ipoib_mcast_restart_task(struct work_struct *work); int ipoib_mcast_start_thread(struct net_device *dev); @@ -549,8 +550,9 @@ void ipoib_path_iter_read(struct ipoib_path_iter *iter, int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid, int set_qkey); -int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast); -struct ipoib_mcast *__ipoib_mcast_find(struct net_device *dev, void *mgid); +void ipoib_mcast_remove_list(struct list_head *remove_list); +void ipoib_check_and_add_mcast_sendonly(struct ipoib_dev_priv *priv, u8 *mgid, + struct list_head *remove_list); int ipoib_init_qp(struct net_device *dev); int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 3ae9726efb98..c8ed53562c9b 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -70,7 +70,6 @@ static struct ib_qp_attr ipoib_cm_err_attr = { #define IPOIB_CM_RX_DRAIN_WRID 0xffffffff static struct ib_send_wr ipoib_cm_rx_drain_wr = { - .wr_id = IPOIB_CM_RX_DRAIN_WRID, .opcode = IB_WR_SEND, }; @@ -223,6 +222,7 @@ static void ipoib_cm_start_rx_drain(struct ipoib_dev_priv *priv) * error" WC will be immediately generated for each WR we post. */ p = list_entry(priv->cm.rx_flush_list.next, typeof(*p), list); + ipoib_cm_rx_drain_wr.wr_id = IPOIB_CM_RX_DRAIN_WRID; if (ib_post_send(p->qp, &ipoib_cm_rx_drain_wr, &bad_wr)) ipoib_warn(priv, "failed to post drain wr\n"); @@ -710,6 +710,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_ struct ipoib_dev_priv *priv = netdev_priv(dev); struct ipoib_tx_buf *tx_req; int rc; + unsigned usable_sge = tx->max_send_sge - !!skb_headlen(skb); if (unlikely(skb->len > tx->mtu)) { ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n", @@ -719,7 +720,23 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_ ipoib_cm_skb_too_long(dev, skb, tx->mtu - IPOIB_ENCAP_LEN); return; } - + if (skb_shinfo(skb)->nr_frags > usable_sge) { + if (skb_linearize(skb) < 0) { + ipoib_warn(priv, "skb could not be linearized\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + /* Does skb_linearize return ok without reducing nr_frags? */ + if (skb_shinfo(skb)->nr_frags > usable_sge) { + ipoib_warn(priv, "too many frags after skb linearize\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + } ipoib_dbg_data(priv, "sending packet: head 0x%x length %d connection 0x%x\n", tx->tx_head, skb->len, tx->qp->qp_num); @@ -1031,7 +1048,8 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_ struct ib_qp *tx_qp; if (dev->features & NETIF_F_SG) - attr.cap.max_send_sge = MAX_SKB_FRAGS + 1; + attr.cap.max_send_sge = + min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1); tx_qp = ib_create_qp(priv->pd, &attr); if (PTR_ERR(tx_qp) == -EINVAL) { @@ -1040,6 +1058,7 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_ attr.create_flags &= ~IB_QP_CREATE_USE_GFP_NOIO; tx_qp = ib_create_qp(priv->pd, &attr); } + tx->max_send_sge = attr.cap.max_send_sge; return tx_qp; } @@ -1522,8 +1541,7 @@ static void ipoib_cm_create_srq(struct net_device *dev, int max_sge) int ipoib_cm_dev_init(struct net_device *dev) { struct ipoib_dev_priv *priv = netdev_priv(dev); - int i, ret; - struct ib_device_attr attr; + int max_srq_sge, i; INIT_LIST_HEAD(&priv->cm.passive_ids); INIT_LIST_HEAD(&priv->cm.reap_list); @@ -1540,19 +1558,13 @@ int ipoib_cm_dev_init(struct net_device *dev) skb_queue_head_init(&priv->cm.skb_queue); - ret = ib_query_device(priv->ca, &attr); - if (ret) { - printk(KERN_WARNING "ib_query_device() failed with %d\n", ret); - return ret; - } - - ipoib_dbg(priv, "max_srq_sge=%d\n", attr.max_srq_sge); + ipoib_dbg(priv, "max_srq_sge=%d\n", priv->ca->attrs.max_srq_sge); - attr.max_srq_sge = min_t(int, IPOIB_CM_RX_SG, attr.max_srq_sge); - ipoib_cm_create_srq(dev, attr.max_srq_sge); + max_srq_sge = min_t(int, IPOIB_CM_RX_SG, priv->ca->attrs.max_srq_sge); + ipoib_cm_create_srq(dev, max_srq_sge); if (ipoib_cm_has_srq(dev)) { - priv->cm.max_cm_mtu = attr.max_srq_sge * PAGE_SIZE - 0x10; - priv->cm.num_frags = attr.max_srq_sge; + priv->cm.max_cm_mtu = max_srq_sge * PAGE_SIZE - 0x10; + priv->cm.num_frags = max_srq_sge; ipoib_dbg(priv, "max_cm_mtu = 0x%x, num_frags=%d\n", priv->cm.max_cm_mtu, priv->cm.num_frags); } else { diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c index 078cadd6c797..a53fa5fc0dec 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c @@ -40,15 +40,11 @@ static void ipoib_get_drvinfo(struct net_device *netdev, struct ethtool_drvinfo *drvinfo) { struct ipoib_dev_priv *priv = netdev_priv(netdev); - struct ib_device_attr *attr; - - attr = kmalloc(sizeof(*attr), GFP_KERNEL); - if (attr && !ib_query_device(priv->ca, attr)) - snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), - "%d.%d.%d", (int)(attr->fw_ver >> 32), - (int)(attr->fw_ver >> 16) & 0xffff, - (int)attr->fw_ver & 0xffff); - kfree(attr); + + snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), + "%d.%d.%d", (int)(priv->ca->attrs.fw_ver >> 32), + (int)(priv->ca->attrs.fw_ver >> 16) & 0xffff, + (int)priv->ca->attrs.fw_ver & 0xffff); strlcpy(drvinfo->bus_info, dev_name(priv->ca->dma_device), sizeof(drvinfo->bus_info)); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index 5ea0c14070d1..899e6b7fb8a5 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -245,8 +245,6 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc) skb_reset_mac_header(skb); skb_pull(skb, IPOIB_ENCAP_LEN); - skb->truesize = SKB_TRUESIZE(skb->len); - ++dev->stats.rx_packets; dev->stats.rx_bytes += skb->len; @@ -540,6 +538,7 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_tx_buf *tx_req; int hlen, rc; void *phead; + unsigned usable_sge = priv->max_send_sge - !!skb_headlen(skb); if (skb_is_gso(skb)) { hlen = skb_transport_offset(skb) + tcp_hdrlen(skb); @@ -563,6 +562,23 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, phead = NULL; hlen = 0; } + if (skb_shinfo(skb)->nr_frags > usable_sge) { + if (skb_linearize(skb) < 0) { + ipoib_warn(priv, "skb could not be linearized\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + /* Does skb_linearize return ok without reducing nr_frags? */ + if (skb_shinfo(skb)->nr_frags > usable_sge) { + ipoib_warn(priv, "too many frags after skb linearize\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + } ipoib_dbg_data(priv, "sending packet, length=%d address=%p qpn=0x%06x\n", skb->len, address, qpn); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 7d3281866ffc..25509bbd4a05 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -1150,8 +1150,6 @@ static void __ipoib_reap_neigh(struct ipoib_dev_priv *priv) unsigned long flags; int i; LIST_HEAD(remove_list); - struct ipoib_mcast *mcast, *tmcast; - struct net_device *dev = priv->dev; if (test_bit(IPOIB_STOP_NEIGH_GC, &priv->flags)) return; @@ -1179,18 +1177,8 @@ static void __ipoib_reap_neigh(struct ipoib_dev_priv *priv) lockdep_is_held(&priv->lock))) != NULL) { /* was the neigh idle for two GC periods */ if (time_after(neigh_obsolete, neigh->alive)) { - u8 *mgid = neigh->daddr + 4; - /* Is this multicast ? */ - if (*mgid == 0xff) { - mcast = __ipoib_mcast_find(dev, mgid); - - if (mcast && test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { - list_del(&mcast->list); - rb_erase(&mcast->rb_node, &priv->multicast_tree); - list_add_tail(&mcast->list, &remove_list); - } - } + ipoib_check_and_add_mcast_sendonly(priv, neigh->daddr + 4, &remove_list); rcu_assign_pointer(*np, rcu_dereference_protected(neigh->hnext, @@ -1207,10 +1195,7 @@ static void __ipoib_reap_neigh(struct ipoib_dev_priv *priv) out_unlock: spin_unlock_irqrestore(&priv->lock, flags); - list_for_each_entry_safe(mcast, tmcast, &remove_list, list) { - ipoib_mcast_leave(dev, mcast); - ipoib_mcast_free(mcast); - } + ipoib_mcast_remove_list(&remove_list); } static void ipoib_reap_neigh(struct work_struct *work) @@ -1777,26 +1762,7 @@ int ipoib_add_pkey_attr(struct net_device *dev) int ipoib_set_dev_features(struct ipoib_dev_priv *priv, struct ib_device *hca) { - struct ib_device_attr *device_attr; - int result = -ENOMEM; - - device_attr = kmalloc(sizeof *device_attr, GFP_KERNEL); - if (!device_attr) { - printk(KERN_WARNING "%s: allocation of %zu bytes failed\n", - hca->name, sizeof *device_attr); - return result; - } - - result = ib_query_device(hca, device_attr); - if (result) { - printk(KERN_WARNING "%s: ib_query_device failed (ret = %d)\n", - hca->name, result); - kfree(device_attr); - return result; - } - priv->hca_caps = device_attr->device_cap_flags; - - kfree(device_attr); + priv->hca_caps = hca->attrs.device_cap_flags; if (priv->hca_caps & IB_DEVICE_UD_IP_CSUM) { priv->dev->hw_features = NETIF_F_SG | diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index f357ca67a41c..25889311b1e9 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -106,7 +106,7 @@ static void __ipoib_mcast_schedule_join_thread(struct ipoib_dev_priv *priv, queue_delayed_work(priv->wq, &priv->mcast_task, 0); } -void ipoib_mcast_free(struct ipoib_mcast *mcast) +static void ipoib_mcast_free(struct ipoib_mcast *mcast) { struct net_device *dev = mcast->dev; int tx_dropped = 0; @@ -153,7 +153,7 @@ static struct ipoib_mcast *ipoib_mcast_alloc(struct net_device *dev, return mcast; } -struct ipoib_mcast *__ipoib_mcast_find(struct net_device *dev, void *mgid) +static struct ipoib_mcast *__ipoib_mcast_find(struct net_device *dev, void *mgid) { struct ipoib_dev_priv *priv = netdev_priv(dev); struct rb_node *n = priv->multicast_tree.rb_node; @@ -456,7 +456,10 @@ out_locked: return status; } -static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast) +/* + * Caller must hold 'priv->lock' + */ +static int ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast) { struct ipoib_dev_priv *priv = netdev_priv(dev); struct ib_sa_multicast *multicast; @@ -466,6 +469,10 @@ static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast) ib_sa_comp_mask comp_mask; int ret = 0; + if (!priv->broadcast || + !test_bit(IPOIB_FLAG_OPER_UP, &priv->flags)) + return -EINVAL; + ipoib_dbg_mcast(priv, "joining MGID %pI6\n", mcast->mcmember.mgid.raw); rec.mgid = mcast->mcmember.mgid; @@ -525,20 +532,23 @@ static void ipoib_mcast_join(struct net_device *dev, struct ipoib_mcast *mcast) rec.join_state = 4; #endif } + spin_unlock_irq(&priv->lock); multicast = ib_sa_join_multicast(&ipoib_sa_client, priv->ca, priv->port, &rec, comp_mask, GFP_KERNEL, ipoib_mcast_join_complete, mcast); + spin_lock_irq(&priv->lock); if (IS_ERR(multicast)) { ret = PTR_ERR(multicast); ipoib_warn(priv, "ib_sa_join_multicast failed, status %d\n", ret); - spin_lock_irq(&priv->lock); /* Requeue this join task with a backoff delay */ __ipoib_mcast_schedule_join_thread(priv, mcast, 1); clear_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags); spin_unlock_irq(&priv->lock); complete(&mcast->done); + spin_lock_irq(&priv->lock); } + return 0; } void ipoib_mcast_join_task(struct work_struct *work) @@ -620,9 +630,10 @@ void ipoib_mcast_join_task(struct work_struct *work) /* Found the next unjoined group */ init_completion(&mcast->done); set_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags); - spin_unlock_irq(&priv->lock); - ipoib_mcast_join(dev, mcast); - spin_lock_irq(&priv->lock); + if (ipoib_mcast_join(dev, mcast)) { + spin_unlock_irq(&priv->lock); + return; + } } else if (!delay_until || time_before(mcast->delay_until, delay_until)) delay_until = mcast->delay_until; @@ -641,10 +652,9 @@ out: if (mcast) { init_completion(&mcast->done); set_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags); + ipoib_mcast_join(dev, mcast); } spin_unlock_irq(&priv->lock); - if (mcast) - ipoib_mcast_join(dev, mcast); } int ipoib_mcast_start_thread(struct net_device *dev) @@ -677,7 +687,7 @@ int ipoib_mcast_stop_thread(struct net_device *dev) return 0; } -int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast) +static int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast) { struct ipoib_dev_priv *priv = netdev_priv(dev); int ret = 0; @@ -704,6 +714,35 @@ int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast) return 0; } +/* + * Check if the multicast group is sendonly. If so remove it from the maps + * and add to the remove list + */ +void ipoib_check_and_add_mcast_sendonly(struct ipoib_dev_priv *priv, u8 *mgid, + struct list_head *remove_list) +{ + /* Is this multicast ? */ + if (*mgid == 0xff) { + struct ipoib_mcast *mcast = __ipoib_mcast_find(priv->dev, mgid); + + if (mcast && test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) { + list_del(&mcast->list); + rb_erase(&mcast->rb_node, &priv->multicast_tree); + list_add_tail(&mcast->list, remove_list); + } + } +} + +void ipoib_mcast_remove_list(struct list_head *remove_list) +{ + struct ipoib_mcast *mcast, *tmcast; + + list_for_each_entry_safe(mcast, tmcast, remove_list, list) { + ipoib_mcast_leave(mcast->dev, mcast); + ipoib_mcast_free(mcast); + } +} + void ipoib_mcast_send(struct net_device *dev, u8 *daddr, struct sk_buff *skb) { struct ipoib_dev_priv *priv = netdev_priv(dev); @@ -810,10 +849,7 @@ void ipoib_mcast_dev_flush(struct net_device *dev) if (test_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags)) wait_for_completion(&mcast->done); - list_for_each_entry_safe(mcast, tmcast, &remove_list, list) { - ipoib_mcast_leave(dev, mcast); - ipoib_mcast_free(mcast); - } + ipoib_mcast_remove_list(&remove_list); } static int ipoib_mcast_addr_is_valid(const u8 *addr, const u8 *broadcast) @@ -939,10 +975,7 @@ void ipoib_mcast_restart_task(struct work_struct *work) if (test_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags)) wait_for_completion(&mcast->done); - list_for_each_entry_safe(mcast, tmcast, &remove_list, list) { - ipoib_mcast_leave(mcast->dev, mcast); - ipoib_mcast_free(mcast); - } + ipoib_mcast_remove_list(&remove_list); /* * Double check that we are still up diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index d48c5bae7877..b809c373e40e 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -206,7 +206,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca) init_attr.create_flags |= IB_QP_CREATE_NETIF_QP; if (dev->features & NETIF_F_SG) - init_attr.cap.max_send_sge = MAX_SKB_FRAGS + 1; + init_attr.cap.max_send_sge = + min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1); priv->qp = ib_create_qp(priv->pd, &init_attr); if (IS_ERR(priv->qp)) { @@ -233,6 +234,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca) priv->rx_wr.next = NULL; priv->rx_wr.sg_list = priv->rx_sge; + priv->max_send_sge = init_attr.cap.max_send_sge; + return 0; out_free_send_cq: diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 9080161e01af..80b6bedc172f 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -644,7 +644,7 @@ iscsi_iser_session_create(struct iscsi_endpoint *ep, ib_conn = &iser_conn->ib_conn; if (ib_conn->pi_support) { - u32 sig_caps = ib_conn->device->dev_attr.sig_prot_cap; + u32 sig_caps = ib_conn->device->ib_device->attrs.sig_prot_cap; scsi_host_set_prot(shost, iser_dif_prot_caps(sig_caps)); scsi_host_set_guard(shost, SHOST_DIX_GUARD_IP | @@ -656,7 +656,7 @@ iscsi_iser_session_create(struct iscsi_endpoint *ep, * max fastreg page list length. */ shost->sg_tablesize = min_t(unsigned short, shost->sg_tablesize, - ib_conn->device->dev_attr.max_fast_reg_page_list_len); + ib_conn->device->ib_device->attrs.max_fast_reg_page_list_len); shost->max_sectors = min_t(unsigned int, 1024, (shost->sg_tablesize * PAGE_SIZE) >> 9); @@ -969,7 +969,16 @@ static umode_t iser_attr_is_visible(int param_type, int param) static int iscsi_iser_slave_alloc(struct scsi_device *sdev) { - blk_queue_virt_boundary(sdev->request_queue, ~MASK_4K); + struct iscsi_session *session; + struct iser_conn *iser_conn; + struct ib_device *ib_dev; + + session = starget_to_session(scsi_target(sdev))->dd_data; + iser_conn = session->leadconn->dd_data; + ib_dev = iser_conn->ib_conn.device->ib_device; + + if (!(ib_dev->attrs.device_cap_flags & IB_DEVICE_SG_GAPS_REG)) + blk_queue_virt_boundary(sdev->request_queue, ~MASK_4K); return 0; } @@ -1059,7 +1068,8 @@ static int __init iser_init(void) release_wq = alloc_workqueue("release workqueue", 0, 0); if (!release_wq) { iser_err("failed to allocate release workqueue\n"); - return -ENOMEM; + err = -ENOMEM; + goto err_alloc_wq; } iscsi_iser_scsi_transport = iscsi_register_transport( @@ -1067,12 +1077,14 @@ static int __init iser_init(void) if (!iscsi_iser_scsi_transport) { iser_err("iscsi_register_transport failed\n"); err = -EINVAL; - goto register_transport_failure; + goto err_reg; } return 0; -register_transport_failure: +err_reg: + destroy_workqueue(release_wq); +err_alloc_wq: kmem_cache_destroy(ig.desc_cache); return err; diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 8a5998e6a407..0351059783b1 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -48,6 +48,7 @@ #include <scsi/scsi_transport_iscsi.h> #include <scsi/scsi_cmnd.h> #include <scsi/scsi_device.h> +#include <scsi/iser.h> #include <linux/interrupt.h> #include <linux/wait.h> @@ -151,46 +152,10 @@ - ISER_MAX_RX_MISC_PDUS) / \ (1 + ISER_INFLIGHT_DATAOUTS)) -#define ISER_WC_BATCH_COUNT 16 #define ISER_SIGNAL_CMD_COUNT 32 -#define ISER_VER 0x10 -#define ISER_WSV 0x08 -#define ISER_RSV 0x04 - -#define ISER_FASTREG_LI_WRID 0xffffffffffffffffULL -#define ISER_BEACON_WRID 0xfffffffffffffffeULL - -/** - * struct iser_hdr - iSER header - * - * @flags: flags support (zbva, remote_inv) - * @rsvd: reserved - * @write_stag: write rkey - * @write_va: write virtual address - * @reaf_stag: read rkey - * @read_va: read virtual address - */ -struct iser_hdr { - u8 flags; - u8 rsvd[3]; - __be32 write_stag; - __be64 write_va; - __be32 read_stag; - __be64 read_va; -} __attribute__((packed)); - - -#define ISER_ZBVA_NOT_SUPPORTED 0x80 -#define ISER_SEND_W_INV_NOT_SUPPORTED 0x40 - -struct iser_cm_hdr { - u8 flags; - u8 rsvd[3]; -} __packed; - /* Constant PDU lengths calculations */ -#define ISER_HEADERS_LEN (sizeof(struct iser_hdr) + sizeof(struct iscsi_hdr)) +#define ISER_HEADERS_LEN (sizeof(struct iser_ctrl) + sizeof(struct iscsi_hdr)) #define ISER_RECV_DATA_SEG_LEN 128 #define ISER_RX_PAYLOAD_SIZE (ISER_HEADERS_LEN + ISER_RECV_DATA_SEG_LEN) @@ -269,7 +234,7 @@ enum iser_desc_type { #define ISER_MAX_WRS 7 /** - * struct iser_tx_desc - iSER TX descriptor (for send wr_id) + * struct iser_tx_desc - iSER TX descriptor * * @iser_header: iser header * @iscsi_header: iscsi header @@ -287,12 +252,13 @@ enum iser_desc_type { * @sig_attrs: Signature attributes */ struct iser_tx_desc { - struct iser_hdr iser_header; + struct iser_ctrl iser_header; struct iscsi_hdr iscsi_header; enum iser_desc_type type; u64 dma_addr; struct ib_sge tx_sg[2]; int num_sge; + struct ib_cqe cqe; bool mapped; u8 wr_idx; union iser_wr { @@ -306,9 +272,10 @@ struct iser_tx_desc { }; #define ISER_RX_PAD_SIZE (256 - (ISER_RX_PAYLOAD_SIZE + \ - sizeof(u64) + sizeof(struct ib_sge))) + sizeof(u64) + sizeof(struct ib_sge) + \ + sizeof(struct ib_cqe))) /** - * struct iser_rx_desc - iSER RX descriptor (for recv wr_id) + * struct iser_rx_desc - iSER RX descriptor * * @iser_header: iser header * @iscsi_header: iscsi header @@ -318,12 +285,32 @@ struct iser_tx_desc { * @pad: for sense data TODO: Modify to maximum sense length supported */ struct iser_rx_desc { - struct iser_hdr iser_header; + struct iser_ctrl iser_header; struct iscsi_hdr iscsi_header; char data[ISER_RECV_DATA_SEG_LEN]; u64 dma_addr; struct ib_sge rx_sg; + struct ib_cqe cqe; char pad[ISER_RX_PAD_SIZE]; +} __packed; + +/** + * struct iser_login_desc - iSER login descriptor + * + * @req: pointer to login request buffer + * @resp: pointer to login response buffer + * @req_dma: DMA address of login request buffer + * @rsp_dma: DMA address of login response buffer + * @sge: IB sge for login post recv + * @cqe: completion handler + */ +struct iser_login_desc { + void *req; + void *rsp; + u64 req_dma; + u64 rsp_dma; + struct ib_sge sge; + struct ib_cqe cqe; } __attribute__((packed)); struct iser_conn; @@ -333,18 +320,12 @@ struct iscsi_iser_task; /** * struct iser_comp - iSER completion context * - * @device: pointer to device handle * @cq: completion queue - * @wcs: work completion array - * @tasklet: Tasklet handle * @active_qps: Number of active QPs attached * to completion context */ struct iser_comp { - struct iser_device *device; struct ib_cq *cq; - struct ib_wc wcs[ISER_WC_BATCH_COUNT]; - struct tasklet_struct tasklet; int active_qps; }; @@ -380,7 +361,6 @@ struct iser_reg_ops { * * @ib_device: RDMA device * @pd: Protection Domain for this device - * @dev_attr: Device attributes container * @mr: Global DMA memory region * @event_handler: IB events handle routine * @ig_list: entry in devices list @@ -389,18 +369,19 @@ struct iser_reg_ops { * cpus and device max completion vectors * @comps: Dinamically allocated array of completion handlers * @reg_ops: Registration ops + * @remote_inv_sup: Remote invalidate is supported on this device */ struct iser_device { struct ib_device *ib_device; struct ib_pd *pd; - struct ib_device_attr dev_attr; struct ib_mr *mr; struct ib_event_handler event_handler; struct list_head ig_list; int refcount; int comps_used; struct iser_comp *comps; - struct iser_reg_ops *reg_ops; + const struct iser_reg_ops *reg_ops; + bool remote_inv_sup; }; #define ISER_CHECK_GUARD 0xc0 @@ -475,10 +456,8 @@ struct iser_fr_pool { * @rx_wr: receive work request for batch posts * @device: reference to iser device * @comp: iser completion context - * @pi_support: Indicate device T10-PI support - * @beacon: beacon send wr to signal all flush errors were drained - * @flush_comp: completes when all connection completions consumed * @fr_pool: connection fast registration poool + * @pi_support: Indicate device T10-PI support */ struct ib_conn { struct rdma_cm_id *cma_id; @@ -488,10 +467,9 @@ struct ib_conn { struct ib_recv_wr rx_wr[ISER_MIN_POSTED_RX]; struct iser_device *device; struct iser_comp *comp; - bool pi_support; - struct ib_send_wr beacon; - struct completion flush_comp; struct iser_fr_pool fr_pool; + bool pi_support; + struct ib_cqe reg_cqe; }; /** @@ -514,11 +492,7 @@ struct ib_conn { * @up_completion: connection establishment completed * (state is ISER_CONN_UP) * @conn_list: entry in ig conn list - * @login_buf: login data buffer (stores login parameters) - * @login_req_buf: login request buffer - * @login_req_dma: login request buffer dma address - * @login_resp_buf: login response buffer - * @login_resp_dma: login response buffer dma address + * @login_desc: login descriptor * @rx_desc_head: head of rx_descs cyclic buffer * @rx_descs: rx buffers array (cyclic buffer) * @num_rx_descs: number of rx descriptors @@ -541,15 +515,13 @@ struct iser_conn { struct completion ib_completion; struct completion up_completion; struct list_head conn_list; - - char *login_buf; - char *login_req_buf, *login_resp_buf; - u64 login_req_dma, login_resp_dma; + struct iser_login_desc login_desc; unsigned int rx_desc_head; struct iser_rx_desc *rx_descs; u32 num_rx_descs; unsigned short scsi_sg_tablesize; unsigned int scsi_max_sectors; + bool snd_w_inv; }; /** @@ -579,9 +551,8 @@ struct iscsi_iser_task { struct iser_page_vec { u64 *pages; - int length; - int offset; - int data_size; + int npages; + struct ib_mr fake_mr; }; /** @@ -633,12 +604,13 @@ int iser_conn_terminate(struct iser_conn *iser_conn); void iser_release_work(struct work_struct *work); -void iser_rcv_completion(struct iser_rx_desc *desc, - unsigned long dto_xfer_len, - struct ib_conn *ib_conn); - -void iser_snd_completion(struct iser_tx_desc *desc, - struct ib_conn *ib_conn); +void iser_err_comp(struct ib_wc *wc, const char *type); +void iser_login_rsp(struct ib_cq *cq, struct ib_wc *wc); +void iser_task_rsp(struct ib_cq *cq, struct ib_wc *wc); +void iser_cmd_comp(struct ib_cq *cq, struct ib_wc *wc); +void iser_ctrl_comp(struct ib_cq *cq, struct ib_wc *wc); +void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc); +void iser_reg_comp(struct ib_cq *cq, struct ib_wc *wc); void iser_task_rdma_init(struct iscsi_iser_task *task); @@ -651,7 +623,8 @@ void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_task *iser_task, enum iser_data_dir cmd_dir); int iser_reg_rdma_mem(struct iscsi_iser_task *task, - enum iser_data_dir dir); + enum iser_data_dir dir, + bool all_imm); void iser_unreg_rdma_mem(struct iscsi_iser_task *task, enum iser_data_dir dir); @@ -719,4 +692,28 @@ iser_tx_next_wr(struct iser_tx_desc *tx_desc) return cur_wr; } +static inline struct iser_conn * +to_iser_conn(struct ib_conn *ib_conn) +{ + return container_of(ib_conn, struct iser_conn, ib_conn); +} + +static inline struct iser_rx_desc * +iser_rx(struct ib_cqe *cqe) +{ + return container_of(cqe, struct iser_rx_desc, cqe); +} + +static inline struct iser_tx_desc * +iser_tx(struct ib_cqe *cqe) +{ + return container_of(cqe, struct iser_tx_desc, cqe); +} + +static inline struct iser_login_desc * +iser_login(struct ib_cqe *cqe) +{ + return container_of(cqe, struct iser_login_desc, cqe); +} + #endif diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index ffd00c420729..81ae2e30dd12 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -51,7 +51,7 @@ static int iser_prepare_read_cmd(struct iscsi_task *task) struct iscsi_iser_task *iser_task = task->dd_data; struct iser_mem_reg *mem_reg; int err; - struct iser_hdr *hdr = &iser_task->desc.iser_header; + struct iser_ctrl *hdr = &iser_task->desc.iser_header; struct iser_data_buf *buf_in = &iser_task->data[ISER_DIR_IN]; err = iser_dma_map_task_data(iser_task, @@ -72,7 +72,7 @@ static int iser_prepare_read_cmd(struct iscsi_task *task) return err; } - err = iser_reg_rdma_mem(iser_task, ISER_DIR_IN); + err = iser_reg_rdma_mem(iser_task, ISER_DIR_IN, false); if (err) { iser_err("Failed to set up Data-IN RDMA\n"); return err; @@ -104,7 +104,7 @@ iser_prepare_write_cmd(struct iscsi_task *task, struct iscsi_iser_task *iser_task = task->dd_data; struct iser_mem_reg *mem_reg; int err; - struct iser_hdr *hdr = &iser_task->desc.iser_header; + struct iser_ctrl *hdr = &iser_task->desc.iser_header; struct iser_data_buf *buf_out = &iser_task->data[ISER_DIR_OUT]; struct ib_sge *tx_dsg = &iser_task->desc.tx_sg[1]; @@ -126,7 +126,8 @@ iser_prepare_write_cmd(struct iscsi_task *task, return err; } - err = iser_reg_rdma_mem(iser_task, ISER_DIR_OUT); + err = iser_reg_rdma_mem(iser_task, ISER_DIR_OUT, + buf_out->data_len == imm_sz); if (err != 0) { iser_err("Failed to register write cmd RDMA mem\n"); return err; @@ -166,7 +167,7 @@ static void iser_create_send_desc(struct iser_conn *iser_conn, ib_dma_sync_single_for_cpu(device->ib_device, tx_desc->dma_addr, ISER_HEADERS_LEN, DMA_TO_DEVICE); - memset(&tx_desc->iser_header, 0, sizeof(struct iser_hdr)); + memset(&tx_desc->iser_header, 0, sizeof(struct iser_ctrl)); tx_desc->iser_header.flags = ISER_VER; tx_desc->num_sge = 1; } @@ -174,73 +175,63 @@ static void iser_create_send_desc(struct iser_conn *iser_conn, static void iser_free_login_buf(struct iser_conn *iser_conn) { struct iser_device *device = iser_conn->ib_conn.device; + struct iser_login_desc *desc = &iser_conn->login_desc; - if (!iser_conn->login_buf) + if (!desc->req) return; - if (iser_conn->login_req_dma) - ib_dma_unmap_single(device->ib_device, - iser_conn->login_req_dma, - ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_TO_DEVICE); + ib_dma_unmap_single(device->ib_device, desc->req_dma, + ISCSI_DEF_MAX_RECV_SEG_LEN, DMA_TO_DEVICE); - if (iser_conn->login_resp_dma) - ib_dma_unmap_single(device->ib_device, - iser_conn->login_resp_dma, - ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE); + ib_dma_unmap_single(device->ib_device, desc->rsp_dma, + ISER_RX_LOGIN_SIZE, DMA_FROM_DEVICE); - kfree(iser_conn->login_buf); + kfree(desc->req); + kfree(desc->rsp); /* make sure we never redo any unmapping */ - iser_conn->login_req_dma = 0; - iser_conn->login_resp_dma = 0; - iser_conn->login_buf = NULL; + desc->req = NULL; + desc->rsp = NULL; } static int iser_alloc_login_buf(struct iser_conn *iser_conn) { struct iser_device *device = iser_conn->ib_conn.device; - int req_err, resp_err; - - BUG_ON(device == NULL); - - iser_conn->login_buf = kmalloc(ISCSI_DEF_MAX_RECV_SEG_LEN + - ISER_RX_LOGIN_SIZE, GFP_KERNEL); - if (!iser_conn->login_buf) - goto out_err; - - iser_conn->login_req_buf = iser_conn->login_buf; - iser_conn->login_resp_buf = iser_conn->login_buf + - ISCSI_DEF_MAX_RECV_SEG_LEN; - - iser_conn->login_req_dma = ib_dma_map_single(device->ib_device, - iser_conn->login_req_buf, - ISCSI_DEF_MAX_RECV_SEG_LEN, - DMA_TO_DEVICE); - - iser_conn->login_resp_dma = ib_dma_map_single(device->ib_device, - iser_conn->login_resp_buf, - ISER_RX_LOGIN_SIZE, - DMA_FROM_DEVICE); - - req_err = ib_dma_mapping_error(device->ib_device, - iser_conn->login_req_dma); - resp_err = ib_dma_mapping_error(device->ib_device, - iser_conn->login_resp_dma); - - if (req_err || resp_err) { - if (req_err) - iser_conn->login_req_dma = 0; - if (resp_err) - iser_conn->login_resp_dma = 0; - goto free_login_buf; - } + struct iser_login_desc *desc = &iser_conn->login_desc; + + desc->req = kmalloc(ISCSI_DEF_MAX_RECV_SEG_LEN, GFP_KERNEL); + if (!desc->req) + return -ENOMEM; + + desc->req_dma = ib_dma_map_single(device->ib_device, desc->req, + ISCSI_DEF_MAX_RECV_SEG_LEN, + DMA_TO_DEVICE); + if (ib_dma_mapping_error(device->ib_device, + desc->req_dma)) + goto free_req; + + desc->rsp = kmalloc(ISER_RX_LOGIN_SIZE, GFP_KERNEL); + if (!desc->rsp) + goto unmap_req; + + desc->rsp_dma = ib_dma_map_single(device->ib_device, desc->rsp, + ISER_RX_LOGIN_SIZE, + DMA_FROM_DEVICE); + if (ib_dma_mapping_error(device->ib_device, + desc->rsp_dma)) + goto free_rsp; + return 0; -free_login_buf: - iser_free_login_buf(iser_conn); +free_rsp: + kfree(desc->rsp); +unmap_req: + ib_dma_unmap_single(device->ib_device, desc->req_dma, + ISCSI_DEF_MAX_RECV_SEG_LEN, + DMA_TO_DEVICE); +free_req: + kfree(desc->req); -out_err: - iser_err("unable to alloc or map login buf\n"); return -ENOMEM; } @@ -280,11 +271,11 @@ int iser_alloc_rx_descriptors(struct iser_conn *iser_conn, goto rx_desc_dma_map_failed; rx_desc->dma_addr = dma_addr; - + rx_desc->cqe.done = iser_task_rsp; rx_sg = &rx_desc->rx_sg; - rx_sg->addr = rx_desc->dma_addr; + rx_sg->addr = rx_desc->dma_addr; rx_sg->length = ISER_RX_PAYLOAD_SIZE; - rx_sg->lkey = device->pd->local_dma_lkey; + rx_sg->lkey = device->pd->local_dma_lkey; } iser_conn->rx_desc_head = 0; @@ -383,6 +374,7 @@ int iser_send_command(struct iscsi_conn *conn, /* build the tx desc regd header and add it to the tx desc dto */ tx_desc->type = ISCSI_TX_SCSI_COMMAND; + tx_desc->cqe.done = iser_cmd_comp; iser_create_send_desc(iser_conn, tx_desc); if (hdr->flags & ISCSI_FLAG_CMD_READ) { @@ -464,6 +456,7 @@ int iser_send_data_out(struct iscsi_conn *conn, } tx_desc->type = ISCSI_TX_DATAOUT; + tx_desc->cqe.done = iser_dataout_comp; tx_desc->iser_header.flags = ISER_VER; memcpy(&tx_desc->iscsi_header, hdr, sizeof(struct iscsi_hdr)); @@ -513,6 +506,7 @@ int iser_send_control(struct iscsi_conn *conn, /* build the tx desc regd header and add it to the tx desc dto */ mdesc->type = ISCSI_TX_CONTROL; + mdesc->cqe.done = iser_ctrl_comp; iser_create_send_desc(iser_conn, mdesc); device = iser_conn->ib_conn.device; @@ -520,25 +514,25 @@ int iser_send_control(struct iscsi_conn *conn, data_seg_len = ntoh24(task->hdr->dlength); if (data_seg_len > 0) { + struct iser_login_desc *desc = &iser_conn->login_desc; struct ib_sge *tx_dsg = &mdesc->tx_sg[1]; + if (task != conn->login_task) { iser_err("data present on non login task!!!\n"); goto send_control_error; } - ib_dma_sync_single_for_cpu(device->ib_device, - iser_conn->login_req_dma, task->data_count, - DMA_TO_DEVICE); + ib_dma_sync_single_for_cpu(device->ib_device, desc->req_dma, + task->data_count, DMA_TO_DEVICE); - memcpy(iser_conn->login_req_buf, task->data, task->data_count); + memcpy(desc->req, task->data, task->data_count); - ib_dma_sync_single_for_device(device->ib_device, - iser_conn->login_req_dma, task->data_count, - DMA_TO_DEVICE); + ib_dma_sync_single_for_device(device->ib_device, desc->req_dma, + task->data_count, DMA_TO_DEVICE); - tx_dsg->addr = iser_conn->login_req_dma; - tx_dsg->length = task->data_count; - tx_dsg->lkey = device->pd->local_dma_lkey; + tx_dsg->addr = desc->req_dma; + tx_dsg->length = task->data_count; + tx_dsg->lkey = device->pd->local_dma_lkey; mdesc->num_sge = 2; } @@ -562,41 +556,126 @@ send_control_error: return err; } -/** - * iser_rcv_dto_completion - recv DTO completion - */ -void iser_rcv_completion(struct iser_rx_desc *rx_desc, - unsigned long rx_xfer_len, - struct ib_conn *ib_conn) +void iser_login_rsp(struct ib_cq *cq, struct ib_wc *wc) { - struct iser_conn *iser_conn = container_of(ib_conn, struct iser_conn, - ib_conn); + struct ib_conn *ib_conn = wc->qp->qp_context; + struct iser_conn *iser_conn = to_iser_conn(ib_conn); + struct iser_login_desc *desc = iser_login(wc->wr_cqe); struct iscsi_hdr *hdr; - u64 rx_dma; - int rx_buflen, outstanding, count, err; + char *data; + int length; - /* differentiate between login to all other PDUs */ - if ((char *)rx_desc == iser_conn->login_resp_buf) { - rx_dma = iser_conn->login_resp_dma; - rx_buflen = ISER_RX_LOGIN_SIZE; - } else { - rx_dma = rx_desc->dma_addr; - rx_buflen = ISER_RX_PAYLOAD_SIZE; + if (unlikely(wc->status != IB_WC_SUCCESS)) { + iser_err_comp(wc, "login_rsp"); + return; + } + + ib_dma_sync_single_for_cpu(ib_conn->device->ib_device, + desc->rsp_dma, ISER_RX_LOGIN_SIZE, + DMA_FROM_DEVICE); + + hdr = desc->rsp + sizeof(struct iser_ctrl); + data = desc->rsp + ISER_HEADERS_LEN; + length = wc->byte_len - ISER_HEADERS_LEN; + + iser_dbg("op 0x%x itt 0x%x dlen %d\n", hdr->opcode, + hdr->itt, length); + + iscsi_iser_recv(iser_conn->iscsi_conn, hdr, data, length); + + ib_dma_sync_single_for_device(ib_conn->device->ib_device, + desc->rsp_dma, ISER_RX_LOGIN_SIZE, + DMA_FROM_DEVICE); + + ib_conn->post_recv_buf_count--; +} + +static inline void +iser_inv_desc(struct iser_fr_desc *desc, u32 rkey) +{ + if (likely(rkey == desc->rsc.mr->rkey)) + desc->rsc.mr_valid = 0; + else if (likely(rkey == desc->pi_ctx->sig_mr->rkey)) + desc->pi_ctx->sig_mr_valid = 0; +} + +static int +iser_check_remote_inv(struct iser_conn *iser_conn, + struct ib_wc *wc, + struct iscsi_hdr *hdr) +{ + if (wc->wc_flags & IB_WC_WITH_INVALIDATE) { + struct iscsi_task *task; + u32 rkey = wc->ex.invalidate_rkey; + + iser_dbg("conn %p: remote invalidation for rkey %#x\n", + iser_conn, rkey); + + if (unlikely(!iser_conn->snd_w_inv)) { + iser_err("conn %p: unexepected remote invalidation, " + "terminating connection\n", iser_conn); + return -EPROTO; + } + + task = iscsi_itt_to_ctask(iser_conn->iscsi_conn, hdr->itt); + if (likely(task)) { + struct iscsi_iser_task *iser_task = task->dd_data; + struct iser_fr_desc *desc; + + if (iser_task->dir[ISER_DIR_IN]) { + desc = iser_task->rdma_reg[ISER_DIR_IN].mem_h; + iser_inv_desc(desc, rkey); + } + + if (iser_task->dir[ISER_DIR_OUT]) { + desc = iser_task->rdma_reg[ISER_DIR_OUT].mem_h; + iser_inv_desc(desc, rkey); + } + } else { + iser_err("failed to get task for itt=%d\n", hdr->itt); + return -EINVAL; + } + } + + return 0; +} + + +void iser_task_rsp(struct ib_cq *cq, struct ib_wc *wc) +{ + struct ib_conn *ib_conn = wc->qp->qp_context; + struct iser_conn *iser_conn = to_iser_conn(ib_conn); + struct iser_rx_desc *desc = iser_rx(wc->wr_cqe); + struct iscsi_hdr *hdr; + int length; + int outstanding, count, err; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + iser_err_comp(wc, "task_rsp"); + return; } - ib_dma_sync_single_for_cpu(ib_conn->device->ib_device, rx_dma, - rx_buflen, DMA_FROM_DEVICE); + ib_dma_sync_single_for_cpu(ib_conn->device->ib_device, + desc->dma_addr, ISER_RX_PAYLOAD_SIZE, + DMA_FROM_DEVICE); - hdr = &rx_desc->iscsi_header; + hdr = &desc->iscsi_header; + length = wc->byte_len - ISER_HEADERS_LEN; iser_dbg("op 0x%x itt 0x%x dlen %d\n", hdr->opcode, - hdr->itt, (int)(rx_xfer_len - ISER_HEADERS_LEN)); + hdr->itt, length); - iscsi_iser_recv(iser_conn->iscsi_conn, hdr, rx_desc->data, - rx_xfer_len - ISER_HEADERS_LEN); + if (iser_check_remote_inv(iser_conn, wc, hdr)) { + iscsi_conn_failure(iser_conn->iscsi_conn, + ISCSI_ERR_CONN_FAILED); + return; + } - ib_dma_sync_single_for_device(ib_conn->device->ib_device, rx_dma, - rx_buflen, DMA_FROM_DEVICE); + iscsi_iser_recv(iser_conn->iscsi_conn, hdr, desc->data, length); + + ib_dma_sync_single_for_device(ib_conn->device->ib_device, + desc->dma_addr, ISER_RX_PAYLOAD_SIZE, + DMA_FROM_DEVICE); /* decrementing conn->post_recv_buf_count only --after-- freeing the * * task eliminates the need to worry on tasks which are completed in * @@ -604,9 +683,6 @@ void iser_rcv_completion(struct iser_rx_desc *rx_desc, * for the posted rx bufs refcount to become zero handles everything */ ib_conn->post_recv_buf_count--; - if (rx_dma == iser_conn->login_resp_dma) - return; - outstanding = ib_conn->post_recv_buf_count; if (outstanding + iser_conn->min_posted_rx <= iser_conn->qp_max_recv_dtos) { count = min(iser_conn->qp_max_recv_dtos - outstanding, @@ -617,26 +693,40 @@ void iser_rcv_completion(struct iser_rx_desc *rx_desc, } } -void iser_snd_completion(struct iser_tx_desc *tx_desc, - struct ib_conn *ib_conn) +void iser_cmd_comp(struct ib_cq *cq, struct ib_wc *wc) +{ + if (unlikely(wc->status != IB_WC_SUCCESS)) + iser_err_comp(wc, "command"); +} + +void iser_ctrl_comp(struct ib_cq *cq, struct ib_wc *wc) { + struct iser_tx_desc *desc = iser_tx(wc->wr_cqe); struct iscsi_task *task; - struct iser_device *device = ib_conn->device; - if (tx_desc->type == ISCSI_TX_DATAOUT) { - ib_dma_unmap_single(device->ib_device, tx_desc->dma_addr, - ISER_HEADERS_LEN, DMA_TO_DEVICE); - kmem_cache_free(ig.desc_cache, tx_desc); - tx_desc = NULL; + if (unlikely(wc->status != IB_WC_SUCCESS)) { + iser_err_comp(wc, "control"); + return; } - if (tx_desc && tx_desc->type == ISCSI_TX_CONTROL) { - /* this arithmetic is legal by libiscsi dd_data allocation */ - task = (void *) ((long)(void *)tx_desc - - sizeof(struct iscsi_task)); - if (task->hdr->itt == RESERVED_ITT) - iscsi_put_task(task); - } + /* this arithmetic is legal by libiscsi dd_data allocation */ + task = (void *)desc - sizeof(struct iscsi_task); + if (task->hdr->itt == RESERVED_ITT) + iscsi_put_task(task); +} + +void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc) +{ + struct iser_tx_desc *desc = iser_tx(wc->wr_cqe); + struct ib_conn *ib_conn = wc->qp->qp_context; + struct iser_device *device = ib_conn->device; + + if (unlikely(wc->status != IB_WC_SUCCESS)) + iser_err_comp(wc, "dataout"); + + ib_dma_unmap_single(device->ib_device, desc->dma_addr, + ISER_HEADERS_LEN, DMA_TO_DEVICE); + kmem_cache_free(ig.desc_cache, desc); } void iser_task_rdma_init(struct iscsi_iser_task *iser_task) diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index ea765fb9664d..9a391cc5b9b3 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -49,7 +49,7 @@ int iser_fast_reg_mr(struct iscsi_iser_task *iser_task, struct iser_reg_resources *rsc, struct iser_mem_reg *mem_reg); -static struct iser_reg_ops fastreg_ops = { +static const struct iser_reg_ops fastreg_ops = { .alloc_reg_res = iser_alloc_fastreg_pool, .free_reg_res = iser_free_fastreg_pool, .reg_mem = iser_fast_reg_mr, @@ -58,7 +58,7 @@ static struct iser_reg_ops fastreg_ops = { .reg_desc_put = iser_reg_desc_put_fr, }; -static struct iser_reg_ops fmr_ops = { +static const struct iser_reg_ops fmr_ops = { .alloc_reg_res = iser_alloc_fmr_pool, .free_reg_res = iser_free_fmr_pool, .reg_mem = iser_fast_reg_fmr, @@ -67,19 +67,24 @@ static struct iser_reg_ops fmr_ops = { .reg_desc_put = iser_reg_desc_put_fmr, }; +void iser_reg_comp(struct ib_cq *cq, struct ib_wc *wc) +{ + iser_err_comp(wc, "memreg"); +} + int iser_assign_reg_ops(struct iser_device *device) { - struct ib_device_attr *dev_attr = &device->dev_attr; + struct ib_device *ib_dev = device->ib_device; /* Assign function handles - based on FMR support */ - if (device->ib_device->alloc_fmr && device->ib_device->dealloc_fmr && - device->ib_device->map_phys_fmr && device->ib_device->unmap_fmr) { + if (ib_dev->alloc_fmr && ib_dev->dealloc_fmr && + ib_dev->map_phys_fmr && ib_dev->unmap_fmr) { iser_info("FMR supported, using FMR for registration\n"); device->reg_ops = &fmr_ops; - } else - if (dev_attr->device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) { + } else if (ib_dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) { iser_info("FastReg supported, using FastReg for registration\n"); device->reg_ops = &fastreg_ops; + device->remote_inv_sup = iser_always_reg; } else { iser_err("IB device does not support FMRs nor FastRegs, can't register memory\n"); return -1; @@ -131,67 +136,6 @@ iser_reg_desc_put_fmr(struct ib_conn *ib_conn, { } -#define IS_4K_ALIGNED(addr) ((((unsigned long)addr) & ~MASK_4K) == 0) - -/** - * iser_sg_to_page_vec - Translates scatterlist entries to physical addresses - * and returns the length of resulting physical address array (may be less than - * the original due to possible compaction). - * - * we build a "page vec" under the assumption that the SG meets the RDMA - * alignment requirements. Other then the first and last SG elements, all - * the "internal" elements can be compacted into a list whose elements are - * dma addresses of physical pages. The code supports also the weird case - * where --few fragments of the same page-- are present in the SG as - * consecutive elements. Also, it handles one entry SG. - */ - -static int iser_sg_to_page_vec(struct iser_data_buf *data, - struct ib_device *ibdev, u64 *pages, - int *offset, int *data_size) -{ - struct scatterlist *sg, *sgl = data->sg; - u64 start_addr, end_addr, page, chunk_start = 0; - unsigned long total_sz = 0; - unsigned int dma_len; - int i, new_chunk, cur_page, last_ent = data->dma_nents - 1; - - /* compute the offset of first element */ - *offset = (u64) sgl[0].offset & ~MASK_4K; - - new_chunk = 1; - cur_page = 0; - for_each_sg(sgl, sg, data->dma_nents, i) { - start_addr = ib_sg_dma_address(ibdev, sg); - if (new_chunk) - chunk_start = start_addr; - dma_len = ib_sg_dma_len(ibdev, sg); - end_addr = start_addr + dma_len; - total_sz += dma_len; - - /* collect page fragments until aligned or end of SG list */ - if (!IS_4K_ALIGNED(end_addr) && i < last_ent) { - new_chunk = 0; - continue; - } - new_chunk = 1; - - /* address of the first page in the contiguous chunk; - masking relevant for the very first SG entry, - which might be unaligned */ - page = chunk_start & MASK_4K; - do { - pages[cur_page++] = page; - page += SIZE_4K; - } while (page < end_addr); - } - - *data_size = total_sz; - iser_dbg("page_vec->data_size:%d cur_page %d\n", - *data_size, cur_page); - return cur_page; -} - static void iser_data_buf_dump(struct iser_data_buf *data, struct ib_device *ibdev) { @@ -210,10 +154,10 @@ static void iser_dump_page_vec(struct iser_page_vec *page_vec) { int i; - iser_err("page vec length %d data size %d\n", - page_vec->length, page_vec->data_size); - for (i = 0; i < page_vec->length; i++) - iser_err("%d %lx\n",i,(unsigned long)page_vec->pages[i]); + iser_err("page vec npages %d data length %d\n", + page_vec->npages, page_vec->fake_mr.length); + for (i = 0; i < page_vec->npages; i++) + iser_err("vec[%d]: %llx\n", i, page_vec->pages[i]); } int iser_dma_map_task_data(struct iscsi_iser_task *iser_task, @@ -251,7 +195,11 @@ iser_reg_dma(struct iser_device *device, struct iser_data_buf *mem, struct scatterlist *sg = mem->sg; reg->sge.lkey = device->pd->local_dma_lkey; - reg->rkey = device->mr->rkey; + /* + * FIXME: rework the registration code path to differentiate + * rkey/lkey use cases + */ + reg->rkey = device->mr ? device->mr->rkey : 0; reg->sge.addr = ib_sg_dma_address(device->ib_device, &sg[0]); reg->sge.length = ib_sg_dma_len(device->ib_device, &sg[0]); @@ -262,11 +210,16 @@ iser_reg_dma(struct iser_device *device, struct iser_data_buf *mem, return 0; } -/** - * iser_reg_page_vec - Register physical memory - * - * returns: 0 on success, errno code on failure - */ +static int iser_set_page(struct ib_mr *mr, u64 addr) +{ + struct iser_page_vec *page_vec = + container_of(mr, struct iser_page_vec, fake_mr); + + page_vec->pages[page_vec->npages++] = addr; + + return 0; +} + static int iser_fast_reg_fmr(struct iscsi_iser_task *iser_task, struct iser_data_buf *mem, @@ -280,22 +233,19 @@ int iser_fast_reg_fmr(struct iscsi_iser_task *iser_task, struct ib_pool_fmr *fmr; int ret, plen; - plen = iser_sg_to_page_vec(mem, device->ib_device, - page_vec->pages, - &page_vec->offset, - &page_vec->data_size); - page_vec->length = plen; - if (plen * SIZE_4K < page_vec->data_size) { + page_vec->npages = 0; + page_vec->fake_mr.page_size = SIZE_4K; + plen = ib_sg_to_pages(&page_vec->fake_mr, mem->sg, + mem->size, iser_set_page); + if (unlikely(plen < mem->size)) { iser_err("page vec too short to hold this SG\n"); iser_data_buf_dump(mem, device->ib_device); iser_dump_page_vec(page_vec); return -EINVAL; } - fmr = ib_fmr_pool_map_phys(fmr_pool, - page_vec->pages, - page_vec->length, - page_vec->pages[0]); + fmr = ib_fmr_pool_map_phys(fmr_pool, page_vec->pages, + page_vec->npages, page_vec->pages[0]); if (IS_ERR(fmr)) { ret = PTR_ERR(fmr); iser_err("ib_fmr_pool_map_phys failed: %d\n", ret); @@ -304,8 +254,8 @@ int iser_fast_reg_fmr(struct iscsi_iser_task *iser_task, reg->sge.lkey = fmr->fmr->lkey; reg->rkey = fmr->fmr->rkey; - reg->sge.addr = page_vec->pages[0] + page_vec->offset; - reg->sge.length = page_vec->data_size; + reg->sge.addr = page_vec->fake_mr.iova; + reg->sge.length = page_vec->fake_mr.length; reg->mem_h = fmr; iser_dbg("fmr reg: lkey=0x%x, rkey=0x%x, addr=0x%llx," @@ -413,19 +363,16 @@ iser_set_prot_checks(struct scsi_cmnd *sc, u8 *mask) *mask |= ISER_CHECK_GUARD; } -static void -iser_inv_rkey(struct ib_send_wr *inv_wr, struct ib_mr *mr) +static inline void +iser_inv_rkey(struct ib_send_wr *inv_wr, + struct ib_mr *mr, + struct ib_cqe *cqe) { - u32 rkey; - inv_wr->opcode = IB_WR_LOCAL_INV; - inv_wr->wr_id = ISER_FASTREG_LI_WRID; + inv_wr->wr_cqe = cqe; inv_wr->ex.invalidate_rkey = mr->rkey; inv_wr->send_flags = 0; inv_wr->num_sge = 0; - - rkey = ib_inc_rkey(mr->rkey); - ib_update_fast_reg_key(mr, rkey); } static int @@ -437,7 +384,9 @@ iser_reg_sig_mr(struct iscsi_iser_task *iser_task, { struct iser_tx_desc *tx_desc = &iser_task->desc; struct ib_sig_attrs *sig_attrs = &tx_desc->sig_attrs; + struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe; struct ib_sig_handover_wr *wr; + struct ib_mr *mr = pi_ctx->sig_mr; int ret; memset(sig_attrs, 0, sizeof(*sig_attrs)); @@ -447,17 +396,19 @@ iser_reg_sig_mr(struct iscsi_iser_task *iser_task, iser_set_prot_checks(iser_task->sc, &sig_attrs->check_mask); - if (!pi_ctx->sig_mr_valid) - iser_inv_rkey(iser_tx_next_wr(tx_desc), pi_ctx->sig_mr); + if (pi_ctx->sig_mr_valid) + iser_inv_rkey(iser_tx_next_wr(tx_desc), mr, cqe); + + ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey)); wr = sig_handover_wr(iser_tx_next_wr(tx_desc)); wr->wr.opcode = IB_WR_REG_SIG_MR; - wr->wr.wr_id = ISER_FASTREG_LI_WRID; + wr->wr.wr_cqe = cqe; wr->wr.sg_list = &data_reg->sge; wr->wr.num_sge = 1; wr->wr.send_flags = 0; wr->sig_attrs = sig_attrs; - wr->sig_mr = pi_ctx->sig_mr; + wr->sig_mr = mr; if (scsi_prot_sg_count(iser_task->sc)) wr->prot = &prot_reg->sge; else @@ -465,10 +416,10 @@ iser_reg_sig_mr(struct iscsi_iser_task *iser_task, wr->access_flags = IB_ACCESS_LOCAL_WRITE | IB_ACCESS_REMOTE_READ | IB_ACCESS_REMOTE_WRITE; - pi_ctx->sig_mr_valid = 0; + pi_ctx->sig_mr_valid = 1; - sig_reg->sge.lkey = pi_ctx->sig_mr->lkey; - sig_reg->rkey = pi_ctx->sig_mr->rkey; + sig_reg->sge.lkey = mr->lkey; + sig_reg->rkey = mr->rkey; sig_reg->sge.addr = 0; sig_reg->sge.length = scsi_transfer_length(iser_task->sc); @@ -485,12 +436,15 @@ static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task, struct iser_mem_reg *reg) { struct iser_tx_desc *tx_desc = &iser_task->desc; + struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe; struct ib_mr *mr = rsc->mr; struct ib_reg_wr *wr; int n; - if (!rsc->mr_valid) - iser_inv_rkey(iser_tx_next_wr(tx_desc), mr); + if (rsc->mr_valid) + iser_inv_rkey(iser_tx_next_wr(tx_desc), mr, cqe); + + ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey)); n = ib_map_mr_sg(mr, mem->sg, mem->size, SIZE_4K); if (unlikely(n != mem->size)) { @@ -501,7 +455,7 @@ static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task, wr = reg_wr(iser_tx_next_wr(tx_desc)); wr->wr.opcode = IB_WR_REG_MR; - wr->wr.wr_id = ISER_FASTREG_LI_WRID; + wr->wr.wr_cqe = cqe; wr->wr.send_flags = 0; wr->wr.num_sge = 0; wr->mr = mr; @@ -510,7 +464,7 @@ static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task, IB_ACCESS_REMOTE_WRITE | IB_ACCESS_REMOTE_READ; - rsc->mr_valid = 0; + rsc->mr_valid = 1; reg->sge.lkey = mr->lkey; reg->rkey = mr->rkey; @@ -554,7 +508,8 @@ iser_reg_data_sg(struct iscsi_iser_task *task, } int iser_reg_rdma_mem(struct iscsi_iser_task *task, - enum iser_data_dir dir) + enum iser_data_dir dir, + bool all_imm) { struct ib_conn *ib_conn = &task->iser_conn->ib_conn; struct iser_device *device = ib_conn->device; @@ -565,8 +520,8 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task, bool use_dma_key; int err; - use_dma_key = (mem->dma_nents == 1 && !iser_always_reg && - scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL); + use_dma_key = mem->dma_nents == 1 && (all_imm || !iser_always_reg) && + scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL; if (!use_dma_key) { desc = device->reg_ops->reg_desc_get(ib_conn); diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index 42f4da620f2e..1b4945367e4f 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -44,17 +44,6 @@ #define ISER_MAX_CQ_LEN (ISER_MAX_RX_LEN + ISER_MAX_TX_LEN + \ ISCSI_ISER_MAX_CONN) -static int iser_cq_poll_limit = 512; - -static void iser_cq_tasklet_fn(unsigned long data); -static void iser_cq_callback(struct ib_cq *cq, void *cq_context); - -static void iser_cq_event_callback(struct ib_event *cause, void *context) -{ - iser_err("cq event %s (%d)\n", - ib_event_msg(cause->event), cause->event); -} - static void iser_qp_event_callback(struct ib_event *cause, void *context) { iser_err("qp event %s (%d)\n", @@ -78,59 +67,40 @@ static void iser_event_handler(struct ib_event_handler *handler, */ static int iser_create_device_ib_res(struct iser_device *device) { - struct ib_device_attr *dev_attr = &device->dev_attr; + struct ib_device *ib_dev = device->ib_device; int ret, i, max_cqe; - ret = ib_query_device(device->ib_device, dev_attr); - if (ret) { - pr_warn("Query device failed for %s\n", device->ib_device->name); - return ret; - } - ret = iser_assign_reg_ops(device); if (ret) return ret; device->comps_used = min_t(int, num_online_cpus(), - device->ib_device->num_comp_vectors); + ib_dev->num_comp_vectors); device->comps = kcalloc(device->comps_used, sizeof(*device->comps), GFP_KERNEL); if (!device->comps) goto comps_err; - max_cqe = min(ISER_MAX_CQ_LEN, dev_attr->max_cqe); + max_cqe = min(ISER_MAX_CQ_LEN, ib_dev->attrs.max_cqe); iser_info("using %d CQs, device %s supports %d vectors max_cqe %d\n", - device->comps_used, device->ib_device->name, - device->ib_device->num_comp_vectors, max_cqe); + device->comps_used, ib_dev->name, + ib_dev->num_comp_vectors, max_cqe); - device->pd = ib_alloc_pd(device->ib_device); + device->pd = ib_alloc_pd(ib_dev); if (IS_ERR(device->pd)) goto pd_err; for (i = 0; i < device->comps_used; i++) { - struct ib_cq_init_attr cq_attr = {}; struct iser_comp *comp = &device->comps[i]; - comp->device = device; - cq_attr.cqe = max_cqe; - cq_attr.comp_vector = i; - comp->cq = ib_create_cq(device->ib_device, - iser_cq_callback, - iser_cq_event_callback, - (void *)comp, - &cq_attr); + comp->cq = ib_alloc_cq(ib_dev, comp, max_cqe, i, + IB_POLL_SOFTIRQ); if (IS_ERR(comp->cq)) { comp->cq = NULL; goto cq_err; } - - if (ib_req_notify_cq(comp->cq, IB_CQ_NEXT_COMP)) - goto cq_err; - - tasklet_init(&comp->tasklet, iser_cq_tasklet_fn, - (unsigned long)comp); } if (!iser_always_reg) { @@ -140,11 +110,11 @@ static int iser_create_device_ib_res(struct iser_device *device) device->mr = ib_get_dma_mr(device->pd, access); if (IS_ERR(device->mr)) - goto dma_mr_err; + goto cq_err; } - INIT_IB_EVENT_HANDLER(&device->event_handler, device->ib_device, - iser_event_handler); + INIT_IB_EVENT_HANDLER(&device->event_handler, ib_dev, + iser_event_handler); if (ib_register_event_handler(&device->event_handler)) goto handler_err; @@ -153,15 +123,12 @@ static int iser_create_device_ib_res(struct iser_device *device) handler_err: if (device->mr) ib_dereg_mr(device->mr); -dma_mr_err: - for (i = 0; i < device->comps_used; i++) - tasklet_kill(&device->comps[i].tasklet); cq_err: for (i = 0; i < device->comps_used; i++) { struct iser_comp *comp = &device->comps[i]; if (comp->cq) - ib_destroy_cq(comp->cq); + ib_free_cq(comp->cq); } ib_dealloc_pd(device->pd); pd_err: @@ -182,8 +149,7 @@ static void iser_free_device_ib_res(struct iser_device *device) for (i = 0; i < device->comps_used; i++) { struct iser_comp *comp = &device->comps[i]; - tasklet_kill(&comp->tasklet); - ib_destroy_cq(comp->cq); + ib_free_cq(comp->cq); comp->cq = NULL; } @@ -286,20 +252,27 @@ void iser_free_fmr_pool(struct ib_conn *ib_conn) } static int -iser_alloc_reg_res(struct ib_device *ib_device, +iser_alloc_reg_res(struct iser_device *device, struct ib_pd *pd, struct iser_reg_resources *res, unsigned int size) { + struct ib_device *ib_dev = device->ib_device; + enum ib_mr_type mr_type; int ret; - res->mr = ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG, size); + if (ib_dev->attrs.device_cap_flags & IB_DEVICE_SG_GAPS_REG) + mr_type = IB_MR_TYPE_SG_GAPS; + else + mr_type = IB_MR_TYPE_MEM_REG; + + res->mr = ib_alloc_mr(pd, mr_type, size); if (IS_ERR(res->mr)) { ret = PTR_ERR(res->mr); iser_err("Failed to allocate ib_fast_reg_mr err=%d\n", ret); return ret; } - res->mr_valid = 1; + res->mr_valid = 0; return 0; } @@ -311,7 +284,7 @@ iser_free_reg_res(struct iser_reg_resources *rsc) } static int -iser_alloc_pi_ctx(struct ib_device *ib_device, +iser_alloc_pi_ctx(struct iser_device *device, struct ib_pd *pd, struct iser_fr_desc *desc, unsigned int size) @@ -325,7 +298,7 @@ iser_alloc_pi_ctx(struct ib_device *ib_device, pi_ctx = desc->pi_ctx; - ret = iser_alloc_reg_res(ib_device, pd, &pi_ctx->rsc, size); + ret = iser_alloc_reg_res(device, pd, &pi_ctx->rsc, size); if (ret) { iser_err("failed to allocate reg_resources\n"); goto alloc_reg_res_err; @@ -336,7 +309,7 @@ iser_alloc_pi_ctx(struct ib_device *ib_device, ret = PTR_ERR(pi_ctx->sig_mr); goto sig_mr_failure; } - pi_ctx->sig_mr_valid = 1; + pi_ctx->sig_mr_valid = 0; desc->pi_ctx->sig_protected = 0; return 0; @@ -358,7 +331,7 @@ iser_free_pi_ctx(struct iser_pi_context *pi_ctx) } static struct iser_fr_desc * -iser_create_fastreg_desc(struct ib_device *ib_device, +iser_create_fastreg_desc(struct iser_device *device, struct ib_pd *pd, bool pi_enable, unsigned int size) @@ -370,12 +343,12 @@ iser_create_fastreg_desc(struct ib_device *ib_device, if (!desc) return ERR_PTR(-ENOMEM); - ret = iser_alloc_reg_res(ib_device, pd, &desc->rsc, size); + ret = iser_alloc_reg_res(device, pd, &desc->rsc, size); if (ret) goto reg_res_alloc_failure; if (pi_enable) { - ret = iser_alloc_pi_ctx(ib_device, pd, desc, size); + ret = iser_alloc_pi_ctx(device, pd, desc, size); if (ret) goto pi_ctx_alloc_failure; } @@ -408,7 +381,7 @@ int iser_alloc_fastreg_pool(struct ib_conn *ib_conn, spin_lock_init(&fr_pool->lock); fr_pool->size = 0; for (i = 0; i < cmds_max; i++) { - desc = iser_create_fastreg_desc(device->ib_device, device->pd, + desc = iser_create_fastreg_desc(device, device->pd, ib_conn->pi_support, size); if (IS_ERR(desc)) { ret = PTR_ERR(desc); @@ -461,10 +434,9 @@ void iser_free_fastreg_pool(struct ib_conn *ib_conn) */ static int iser_create_ib_conn_res(struct ib_conn *ib_conn) { - struct iser_conn *iser_conn = container_of(ib_conn, struct iser_conn, - ib_conn); + struct iser_conn *iser_conn = to_iser_conn(ib_conn); struct iser_device *device; - struct ib_device_attr *dev_attr; + struct ib_device *ib_dev; struct ib_qp_init_attr init_attr; int ret = -ENOMEM; int index, min_index = 0; @@ -472,7 +444,7 @@ static int iser_create_ib_conn_res(struct ib_conn *ib_conn) BUG_ON(ib_conn->device == NULL); device = ib_conn->device; - dev_attr = &device->dev_attr; + ib_dev = device->ib_device; memset(&init_attr, 0, sizeof init_attr); @@ -503,16 +475,16 @@ static int iser_create_ib_conn_res(struct ib_conn *ib_conn) iser_conn->max_cmds = ISER_GET_MAX_XMIT_CMDS(ISER_QP_SIG_MAX_REQ_DTOS); } else { - if (dev_attr->max_qp_wr > ISER_QP_MAX_REQ_DTOS) { + if (ib_dev->attrs.max_qp_wr > ISER_QP_MAX_REQ_DTOS) { init_attr.cap.max_send_wr = ISER_QP_MAX_REQ_DTOS + 1; iser_conn->max_cmds = ISER_GET_MAX_XMIT_CMDS(ISER_QP_MAX_REQ_DTOS); } else { - init_attr.cap.max_send_wr = dev_attr->max_qp_wr; + init_attr.cap.max_send_wr = ib_dev->attrs.max_qp_wr; iser_conn->max_cmds = - ISER_GET_MAX_XMIT_CMDS(dev_attr->max_qp_wr); + ISER_GET_MAX_XMIT_CMDS(ib_dev->attrs.max_qp_wr); iser_dbg("device %s supports max_send_wr %d\n", - device->ib_device->name, dev_attr->max_qp_wr); + device->ib_device->name, ib_dev->attrs.max_qp_wr); } } @@ -698,7 +670,6 @@ void iser_conn_release(struct iser_conn *iser_conn) int iser_conn_terminate(struct iser_conn *iser_conn) { struct ib_conn *ib_conn = &iser_conn->ib_conn; - struct ib_send_wr *bad_wr; int err = 0; /* terminate the iser conn only if the conn state is UP */ @@ -723,14 +694,8 @@ int iser_conn_terminate(struct iser_conn *iser_conn) iser_err("Failed to disconnect, conn: 0x%p err %d\n", iser_conn, err); - /* post an indication that all flush errors were consumed */ - err = ib_post_send(ib_conn->qp, &ib_conn->beacon, &bad_wr); - if (err) { - iser_err("conn %p failed to post beacon", ib_conn); - return 1; - } - - wait_for_completion(&ib_conn->flush_comp); + /* block until all flush errors are consumed */ + ib_drain_sq(ib_conn->qp); } return 1; @@ -756,7 +721,7 @@ iser_calc_scsi_params(struct iser_conn *iser_conn, sg_tablesize = DIV_ROUND_UP(max_sectors * 512, SIZE_4K); sup_sg_tablesize = min_t(unsigned, ISCSI_ISER_MAX_SG_TABLESIZE, - device->dev_attr.max_fast_reg_page_list_len); + device->ib_device->attrs.max_fast_reg_page_list_len); if (sg_tablesize > sup_sg_tablesize) { sg_tablesize = sup_sg_tablesize; @@ -799,7 +764,7 @@ static void iser_addr_handler(struct rdma_cm_id *cma_id) /* connection T10-PI support */ if (iser_pi_enable) { - if (!(device->dev_attr.device_cap_flags & + if (!(device->ib_device->attrs.device_cap_flags & IB_DEVICE_SIGNATURE_HANDOVER)) { iser_warn("T10-PI requested but not supported on %s, " "continue without T10-PI\n", @@ -841,16 +806,17 @@ static void iser_route_handler(struct rdma_cm_id *cma_id) goto failure; memset(&conn_param, 0, sizeof conn_param); - conn_param.responder_resources = device->dev_attr.max_qp_rd_atom; + conn_param.responder_resources = device->ib_device->attrs.max_qp_rd_atom; conn_param.initiator_depth = 1; conn_param.retry_count = 7; conn_param.rnr_retry_count = 6; memset(&req_hdr, 0, sizeof(req_hdr)); - req_hdr.flags = (ISER_ZBVA_NOT_SUPPORTED | - ISER_SEND_W_INV_NOT_SUPPORTED); - conn_param.private_data = (void *)&req_hdr; - conn_param.private_data_len = sizeof(struct iser_cm_hdr); + req_hdr.flags = ISER_ZBVA_NOT_SUP; + if (!device->remote_inv_sup) + req_hdr.flags |= ISER_SEND_W_INV_NOT_SUP; + conn_param.private_data = (void *)&req_hdr; + conn_param.private_data_len = sizeof(struct iser_cm_hdr); ret = rdma_connect(cma_id, &conn_param); if (ret) { @@ -863,7 +829,8 @@ failure: iser_connect_error(cma_id); } -static void iser_connected_handler(struct rdma_cm_id *cma_id) +static void iser_connected_handler(struct rdma_cm_id *cma_id, + const void *private_data) { struct iser_conn *iser_conn; struct ib_qp_attr attr; @@ -877,6 +844,15 @@ static void iser_connected_handler(struct rdma_cm_id *cma_id) (void)ib_query_qp(cma_id->qp, &attr, ~0, &init_attr); iser_info("remote qpn:%x my qpn:%x\n", attr.dest_qp_num, cma_id->qp->qp_num); + if (private_data) { + u8 flags = *(u8 *)private_data; + + iser_conn->snd_w_inv = !(flags & ISER_SEND_W_INV_NOT_SUP); + } + + iser_info("conn %p: negotiated %s invalidation\n", + iser_conn, iser_conn->snd_w_inv ? "remote" : "local"); + iser_conn->state = ISER_CONN_UP; complete(&iser_conn->up_completion); } @@ -928,7 +904,7 @@ static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *eve iser_route_handler(cma_id); break; case RDMA_CM_EVENT_ESTABLISHED: - iser_connected_handler(cma_id); + iser_connected_handler(cma_id, event->param.conn.private_data); break; case RDMA_CM_EVENT_ADDR_ERROR: case RDMA_CM_EVENT_ROUTE_ERROR: @@ -967,14 +943,17 @@ static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *eve void iser_conn_init(struct iser_conn *iser_conn) { + struct ib_conn *ib_conn = &iser_conn->ib_conn; + iser_conn->state = ISER_CONN_INIT; - iser_conn->ib_conn.post_recv_buf_count = 0; - init_completion(&iser_conn->ib_conn.flush_comp); init_completion(&iser_conn->stop_completion); init_completion(&iser_conn->ib_completion); init_completion(&iser_conn->up_completion); INIT_LIST_HEAD(&iser_conn->conn_list); mutex_init(&iser_conn->state_mutex); + + ib_conn->post_recv_buf_count = 0; + ib_conn->reg_cqe.done = iser_reg_comp; } /** @@ -1000,9 +979,6 @@ int iser_connect(struct iser_conn *iser_conn, iser_conn->state = ISER_CONN_PENDING; - ib_conn->beacon.wr_id = ISER_BEACON_WRID; - ib_conn->beacon.opcode = IB_WR_SEND; - ib_conn->cma_id = rdma_create_id(&init_net, iser_cma_handler, (void *)iser_conn, RDMA_PS_TCP, IB_QPT_RC); @@ -1045,56 +1021,60 @@ connect_failure: int iser_post_recvl(struct iser_conn *iser_conn) { - struct ib_recv_wr rx_wr, *rx_wr_failed; struct ib_conn *ib_conn = &iser_conn->ib_conn; - struct ib_sge sge; + struct iser_login_desc *desc = &iser_conn->login_desc; + struct ib_recv_wr wr, *wr_failed; int ib_ret; - sge.addr = iser_conn->login_resp_dma; - sge.length = ISER_RX_LOGIN_SIZE; - sge.lkey = ib_conn->device->pd->local_dma_lkey; + desc->sge.addr = desc->rsp_dma; + desc->sge.length = ISER_RX_LOGIN_SIZE; + desc->sge.lkey = ib_conn->device->pd->local_dma_lkey; - rx_wr.wr_id = (uintptr_t)iser_conn->login_resp_buf; - rx_wr.sg_list = &sge; - rx_wr.num_sge = 1; - rx_wr.next = NULL; + desc->cqe.done = iser_login_rsp; + wr.wr_cqe = &desc->cqe; + wr.sg_list = &desc->sge; + wr.num_sge = 1; + wr.next = NULL; ib_conn->post_recv_buf_count++; - ib_ret = ib_post_recv(ib_conn->qp, &rx_wr, &rx_wr_failed); + ib_ret = ib_post_recv(ib_conn->qp, &wr, &wr_failed); if (ib_ret) { iser_err("ib_post_recv failed ret=%d\n", ib_ret); ib_conn->post_recv_buf_count--; } + return ib_ret; } int iser_post_recvm(struct iser_conn *iser_conn, int count) { - struct ib_recv_wr *rx_wr, *rx_wr_failed; - int i, ib_ret; struct ib_conn *ib_conn = &iser_conn->ib_conn; unsigned int my_rx_head = iser_conn->rx_desc_head; struct iser_rx_desc *rx_desc; + struct ib_recv_wr *wr, *wr_failed; + int i, ib_ret; - for (rx_wr = ib_conn->rx_wr, i = 0; i < count; i++, rx_wr++) { - rx_desc = &iser_conn->rx_descs[my_rx_head]; - rx_wr->wr_id = (uintptr_t)rx_desc; - rx_wr->sg_list = &rx_desc->rx_sg; - rx_wr->num_sge = 1; - rx_wr->next = rx_wr + 1; + for (wr = ib_conn->rx_wr, i = 0; i < count; i++, wr++) { + rx_desc = &iser_conn->rx_descs[my_rx_head]; + rx_desc->cqe.done = iser_task_rsp; + wr->wr_cqe = &rx_desc->cqe; + wr->sg_list = &rx_desc->rx_sg; + wr->num_sge = 1; + wr->next = wr + 1; my_rx_head = (my_rx_head + 1) & iser_conn->qp_max_recv_dtos_mask; } - rx_wr--; - rx_wr->next = NULL; /* mark end of work requests list */ + wr--; + wr->next = NULL; /* mark end of work requests list */ ib_conn->post_recv_buf_count += count; - ib_ret = ib_post_recv(ib_conn->qp, ib_conn->rx_wr, &rx_wr_failed); + ib_ret = ib_post_recv(ib_conn->qp, ib_conn->rx_wr, &wr_failed); if (ib_ret) { iser_err("ib_post_recv failed ret=%d\n", ib_ret); ib_conn->post_recv_buf_count -= count; } else iser_conn->rx_desc_head = my_rx_head; + return ib_ret; } @@ -1115,7 +1095,7 @@ int iser_post_send(struct ib_conn *ib_conn, struct iser_tx_desc *tx_desc, DMA_TO_DEVICE); wr->next = NULL; - wr->wr_id = (uintptr_t)tx_desc; + wr->wr_cqe = &tx_desc->cqe; wr->sg_list = tx_desc->tx_sg; wr->num_sge = tx_desc->num_sge; wr->opcode = IB_WR_SEND; @@ -1129,149 +1109,6 @@ int iser_post_send(struct ib_conn *ib_conn, struct iser_tx_desc *tx_desc, return ib_ret; } -/** - * is_iser_tx_desc - Indicate if the completion wr_id - * is a TX descriptor or not. - * @iser_conn: iser connection - * @wr_id: completion WR identifier - * - * Since we cannot rely on wc opcode in FLUSH errors - * we must work around it by checking if the wr_id address - * falls in the iser connection rx_descs buffer. If so - * it is an RX descriptor, otherwize it is a TX. - */ -static inline bool -is_iser_tx_desc(struct iser_conn *iser_conn, void *wr_id) -{ - void *start = iser_conn->rx_descs; - int len = iser_conn->num_rx_descs * sizeof(*iser_conn->rx_descs); - - if (wr_id >= start && wr_id < start + len) - return false; - - return true; -} - -/** - * iser_handle_comp_error() - Handle error completion - * @ib_conn: connection RDMA resources - * @wc: work completion - * - * Notes: We may handle a FLUSH error completion and in this case - * we only cleanup in case TX type was DATAOUT. For non-FLUSH - * error completion we should also notify iscsi layer that - * connection is failed (in case we passed bind stage). - */ -static void -iser_handle_comp_error(struct ib_conn *ib_conn, - struct ib_wc *wc) -{ - void *wr_id = (void *)(uintptr_t)wc->wr_id; - struct iser_conn *iser_conn = container_of(ib_conn, struct iser_conn, - ib_conn); - - if (wc->status != IB_WC_WR_FLUSH_ERR) - if (iser_conn->iscsi_conn) - iscsi_conn_failure(iser_conn->iscsi_conn, - ISCSI_ERR_CONN_FAILED); - - if (wc->wr_id == ISER_FASTREG_LI_WRID) - return; - - if (is_iser_tx_desc(iser_conn, wr_id)) { - struct iser_tx_desc *desc = wr_id; - - if (desc->type == ISCSI_TX_DATAOUT) - kmem_cache_free(ig.desc_cache, desc); - } else { - ib_conn->post_recv_buf_count--; - } -} - -/** - * iser_handle_wc - handle a single work completion - * @wc: work completion - * - * Soft-IRQ context, work completion can be either - * SEND or RECV, and can turn out successful or - * with error (or flush error). - */ -static void iser_handle_wc(struct ib_wc *wc) -{ - struct ib_conn *ib_conn; - struct iser_tx_desc *tx_desc; - struct iser_rx_desc *rx_desc; - - ib_conn = wc->qp->qp_context; - if (likely(wc->status == IB_WC_SUCCESS)) { - if (wc->opcode == IB_WC_RECV) { - rx_desc = (struct iser_rx_desc *)(uintptr_t)wc->wr_id; - iser_rcv_completion(rx_desc, wc->byte_len, - ib_conn); - } else - if (wc->opcode == IB_WC_SEND) { - tx_desc = (struct iser_tx_desc *)(uintptr_t)wc->wr_id; - iser_snd_completion(tx_desc, ib_conn); - } else { - iser_err("Unknown wc opcode %d\n", wc->opcode); - } - } else { - if (wc->status != IB_WC_WR_FLUSH_ERR) - iser_err("%s (%d): wr id %llx vend_err %x\n", - ib_wc_status_msg(wc->status), wc->status, - wc->wr_id, wc->vendor_err); - else - iser_dbg("%s (%d): wr id %llx\n", - ib_wc_status_msg(wc->status), wc->status, - wc->wr_id); - - if (wc->wr_id == ISER_BEACON_WRID) - /* all flush errors were consumed */ - complete(&ib_conn->flush_comp); - else - iser_handle_comp_error(ib_conn, wc); - } -} - -/** - * iser_cq_tasklet_fn - iSER completion polling loop - * @data: iSER completion context - * - * Soft-IRQ context, polling connection CQ until - * either CQ was empty or we exausted polling budget - */ -static void iser_cq_tasklet_fn(unsigned long data) -{ - struct iser_comp *comp = (struct iser_comp *)data; - struct ib_cq *cq = comp->cq; - struct ib_wc *const wcs = comp->wcs; - int i, n, completed = 0; - - while ((n = ib_poll_cq(cq, ARRAY_SIZE(comp->wcs), wcs)) > 0) { - for (i = 0; i < n; i++) - iser_handle_wc(&wcs[i]); - - completed += n; - if (completed >= iser_cq_poll_limit) - break; - } - - /* - * It is assumed here that arming CQ only once its empty - * would not cause interrupts to be missed. - */ - ib_req_notify_cq(cq, IB_CQ_NEXT_COMP); - - iser_dbg("got %d completions\n", completed); -} - -static void iser_cq_callback(struct ib_cq *cq, void *cq_context) -{ - struct iser_comp *comp = cq_context; - - tasklet_schedule(&comp->tasklet); -} - u8 iser_check_task_pi_status(struct iscsi_iser_task *iser_task, enum iser_data_dir cmd_dir, sector_t *sector) { @@ -1319,3 +1156,21 @@ err: /* Not alot we can do here, return ambiguous guard error */ return 0x1; } + +void iser_err_comp(struct ib_wc *wc, const char *type) +{ + if (wc->status != IB_WC_WR_FLUSH_ERR) { + struct iser_conn *iser_conn = to_iser_conn(wc->qp->qp_context); + + iser_err("%s failure: %s (%d) vend_err %x\n", type, + ib_wc_status_msg(wc->status), wc->status, + wc->vendor_err); + + if (iser_conn->iscsi_conn) + iscsi_conn_failure(iser_conn->iscsi_conn, + ISCSI_ERR_CONN_FAILED); + } else { + iser_dbg("%s failure: %s (%d)\n", type, + ib_wc_status_msg(wc->status), wc->status); + } +} diff --git a/drivers/infiniband/ulp/isert/ib_isert.c b/drivers/infiniband/ulp/isert/ib_isert.c index 468c5e132563..f121e6129339 100644 --- a/drivers/infiniband/ulp/isert/ib_isert.c +++ b/drivers/infiniband/ulp/isert/ib_isert.c @@ -29,7 +29,6 @@ #include <target/iscsi/iscsi_transport.h> #include <linux/semaphore.h> -#include "isert_proto.h" #include "ib_isert.h" #define ISERT_MAX_CONN 8 @@ -95,22 +94,6 @@ isert_qp_event_callback(struct ib_event *e, void *context) } } -static int -isert_query_device(struct ib_device *ib_dev, struct ib_device_attr *devattr) -{ - int ret; - - ret = ib_query_device(ib_dev, devattr); - if (ret) { - isert_err("ib_query_device() failed: %d\n", ret); - return ret; - } - isert_dbg("devattr->max_sge: %d\n", devattr->max_sge); - isert_dbg("devattr->max_sge_rd: %d\n", devattr->max_sge_rd); - - return 0; -} - static struct isert_comp * isert_comp_get(struct isert_conn *isert_conn) { @@ -157,9 +140,9 @@ isert_create_qp(struct isert_conn *isert_conn, attr.recv_cq = comp->cq; attr.cap.max_send_wr = ISERT_QP_MAX_REQ_DTOS; attr.cap.max_recv_wr = ISERT_QP_MAX_RECV_DTOS + 1; - attr.cap.max_send_sge = device->dev_attr.max_sge; - isert_conn->max_sge = min(device->dev_attr.max_sge, - device->dev_attr.max_sge_rd); + attr.cap.max_send_sge = device->ib_device->attrs.max_sge; + isert_conn->max_sge = min(device->ib_device->attrs.max_sge, + device->ib_device->attrs.max_sge_rd); attr.cap.max_recv_sge = 1; attr.sq_sig_type = IB_SIGNAL_REQ_WR; attr.qp_type = IB_QPT_RC; @@ -287,8 +270,7 @@ isert_free_comps(struct isert_device *device) } static int -isert_alloc_comps(struct isert_device *device, - struct ib_device_attr *attr) +isert_alloc_comps(struct isert_device *device) { int i, max_cqe, ret = 0; @@ -308,7 +290,7 @@ isert_alloc_comps(struct isert_device *device, return -ENOMEM; } - max_cqe = min(ISER_MAX_CQ_LEN, attr->max_cqe); + max_cqe = min(ISER_MAX_CQ_LEN, device->ib_device->attrs.max_cqe); for (i = 0; i < device->comps_used; i++) { struct ib_cq_init_attr cq_attr = {}; @@ -344,17 +326,15 @@ out_cq: static int isert_create_device_ib_res(struct isert_device *device) { - struct ib_device_attr *dev_attr; + struct ib_device *ib_dev = device->ib_device; int ret; - dev_attr = &device->dev_attr; - ret = isert_query_device(device->ib_device, dev_attr); - if (ret) - goto out; + isert_dbg("devattr->max_sge: %d\n", ib_dev->attrs.max_sge); + isert_dbg("devattr->max_sge_rd: %d\n", ib_dev->attrs.max_sge_rd); /* asign function handlers */ - if (dev_attr->device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS && - dev_attr->device_cap_flags & IB_DEVICE_SIGNATURE_HANDOVER) { + if (ib_dev->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS && + ib_dev->attrs.device_cap_flags & IB_DEVICE_SIGNATURE_HANDOVER) { device->use_fastreg = 1; device->reg_rdma_mem = isert_reg_rdma; device->unreg_rdma_mem = isert_unreg_rdma; @@ -364,11 +344,11 @@ isert_create_device_ib_res(struct isert_device *device) device->unreg_rdma_mem = isert_unmap_cmd; } - ret = isert_alloc_comps(device, dev_attr); + ret = isert_alloc_comps(device); if (ret) goto out; - device->pd = ib_alloc_pd(device->ib_device); + device->pd = ib_alloc_pd(ib_dev); if (IS_ERR(device->pd)) { ret = PTR_ERR(device->pd); isert_err("failed to allocate pd, device %p, ret=%d\n", @@ -377,7 +357,7 @@ isert_create_device_ib_res(struct isert_device *device) } /* Check signature cap */ - device->pi_capable = dev_attr->device_cap_flags & + device->pi_capable = ib_dev->attrs.device_cap_flags & IB_DEVICE_SIGNATURE_HANDOVER ? true : false; return 0; @@ -676,6 +656,32 @@ out_login_buf: return ret; } +static void +isert_set_nego_params(struct isert_conn *isert_conn, + struct rdma_conn_param *param) +{ + struct ib_device_attr *attr = &isert_conn->device->ib_device->attrs; + + /* Set max inflight RDMA READ requests */ + isert_conn->initiator_depth = min_t(u8, param->initiator_depth, + attr->max_qp_init_rd_atom); + isert_dbg("Using initiator_depth: %u\n", isert_conn->initiator_depth); + + if (param->private_data) { + u8 flags = *(u8 *)param->private_data; + + /* + * use remote invalidation if the both initiator + * and the HCA support it + */ + isert_conn->snd_w_inv = !(flags & ISER_SEND_W_INV_NOT_SUP) && + (attr->device_cap_flags & + IB_DEVICE_MEM_MGT_EXTENSIONS); + if (isert_conn->snd_w_inv) + isert_info("Using remote invalidation\n"); + } +} + static int isert_connect_request(struct rdma_cm_id *cma_id, struct rdma_cm_event *event) { @@ -714,11 +720,7 @@ isert_connect_request(struct rdma_cm_id *cma_id, struct rdma_cm_event *event) } isert_conn->device = device; - /* Set max inflight RDMA READ requests */ - isert_conn->initiator_depth = min_t(u8, - event->param.conn.initiator_depth, - device->dev_attr.max_qp_init_rd_atom); - isert_dbg("Using initiator_depth: %u\n", isert_conn->initiator_depth); + isert_set_nego_params(isert_conn, &event->param.conn); ret = isert_conn_setup_qp(isert_conn, cma_id); if (ret) @@ -1050,8 +1052,8 @@ isert_create_send_desc(struct isert_conn *isert_conn, ib_dma_sync_single_for_cpu(ib_dev, tx_desc->dma_addr, ISER_HEADERS_LEN, DMA_TO_DEVICE); - memset(&tx_desc->iser_header, 0, sizeof(struct iser_hdr)); - tx_desc->iser_header.flags = ISER_VER; + memset(&tx_desc->iser_header, 0, sizeof(struct iser_ctrl)); + tx_desc->iser_header.flags = ISCSI_CTRL; tx_desc->num_sge = 1; tx_desc->isert_cmd = isert_cmd; @@ -1097,7 +1099,14 @@ isert_init_send_wr(struct isert_conn *isert_conn, struct isert_cmd *isert_cmd, isert_cmd->rdma_wr.iser_ib_op = ISER_IB_SEND; send_wr->wr_id = (uintptr_t)&isert_cmd->tx_desc; - send_wr->opcode = IB_WR_SEND; + + if (isert_conn->snd_w_inv && isert_cmd->inv_rkey) { + send_wr->opcode = IB_WR_SEND_WITH_INV; + send_wr->ex.invalidate_rkey = isert_cmd->inv_rkey; + } else { + send_wr->opcode = IB_WR_SEND; + } + send_wr->sg_list = &tx_desc->tx_sg[0]; send_wr->num_sge = isert_cmd->tx_desc.num_sge; send_wr->send_flags = IB_SEND_SIGNALED; @@ -1486,6 +1495,7 @@ isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc, isert_cmd->read_va = read_va; isert_cmd->write_stag = write_stag; isert_cmd->write_va = write_va; + isert_cmd->inv_rkey = read_stag ? read_stag : write_stag; ret = isert_handle_scsi_cmd(isert_conn, isert_cmd, cmd, rx_desc, (unsigned char *)hdr); @@ -1543,21 +1553,21 @@ isert_rx_opcode(struct isert_conn *isert_conn, struct iser_rx_desc *rx_desc, static void isert_rx_do_work(struct iser_rx_desc *rx_desc, struct isert_conn *isert_conn) { - struct iser_hdr *iser_hdr = &rx_desc->iser_header; + struct iser_ctrl *iser_ctrl = &rx_desc->iser_header; uint64_t read_va = 0, write_va = 0; uint32_t read_stag = 0, write_stag = 0; - switch (iser_hdr->flags & 0xF0) { + switch (iser_ctrl->flags & 0xF0) { case ISCSI_CTRL: - if (iser_hdr->flags & ISER_RSV) { - read_stag = be32_to_cpu(iser_hdr->read_stag); - read_va = be64_to_cpu(iser_hdr->read_va); + if (iser_ctrl->flags & ISER_RSV) { + read_stag = be32_to_cpu(iser_ctrl->read_stag); + read_va = be64_to_cpu(iser_ctrl->read_va); isert_dbg("ISER_RSV: read_stag: 0x%x read_va: 0x%llx\n", read_stag, (unsigned long long)read_va); } - if (iser_hdr->flags & ISER_WSV) { - write_stag = be32_to_cpu(iser_hdr->write_stag); - write_va = be64_to_cpu(iser_hdr->write_va); + if (iser_ctrl->flags & ISER_WSV) { + write_stag = be32_to_cpu(iser_ctrl->write_stag); + write_va = be64_to_cpu(iser_ctrl->write_va); isert_dbg("ISER_WSV: write_stag: 0x%x write_va: 0x%llx\n", write_stag, (unsigned long long)write_va); } @@ -1568,7 +1578,7 @@ isert_rx_do_work(struct iser_rx_desc *rx_desc, struct isert_conn *isert_conn) isert_err("iSER Hello message\n"); break; default: - isert_warn("Unknown iSER hdr flags: 0x%02x\n", iser_hdr->flags); + isert_warn("Unknown iSER hdr flags: 0x%02x\n", iser_ctrl->flags); break; } @@ -3095,12 +3105,20 @@ isert_rdma_accept(struct isert_conn *isert_conn) struct rdma_cm_id *cm_id = isert_conn->cm_id; struct rdma_conn_param cp; int ret; + struct iser_cm_hdr rsp_hdr; memset(&cp, 0, sizeof(struct rdma_conn_param)); cp.initiator_depth = isert_conn->initiator_depth; cp.retry_count = 7; cp.rnr_retry_count = 7; + memset(&rsp_hdr, 0, sizeof(rsp_hdr)); + rsp_hdr.flags = ISERT_ZBVA_NOT_USED; + if (!isert_conn->snd_w_inv) + rsp_hdr.flags = rsp_hdr.flags | ISERT_SEND_W_INV_NOT_USED; + cp.private_data = (void *)&rsp_hdr; + cp.private_data_len = sizeof(rsp_hdr); + ret = rdma_accept(cm_id, &cp); if (ret) { isert_err("rdma_accept() failed with: %d\n", ret); diff --git a/drivers/infiniband/ulp/isert/ib_isert.h b/drivers/infiniband/ulp/isert/ib_isert.h index 3d7fbc47c343..8d50453eef66 100644 --- a/drivers/infiniband/ulp/isert/ib_isert.h +++ b/drivers/infiniband/ulp/isert/ib_isert.h @@ -3,6 +3,8 @@ #include <linux/in6.h> #include <rdma/ib_verbs.h> #include <rdma/rdma_cm.h> +#include <scsi/iser.h> + #define DRV_NAME "isert" #define PFX DRV_NAME ": " @@ -31,6 +33,38 @@ #define isert_err(fmt, arg...) \ pr_err(PFX "%s: " fmt, __func__ , ## arg) +/* Constant PDU lengths calculations */ +#define ISER_HEADERS_LEN (sizeof(struct iser_ctrl) + \ + sizeof(struct iscsi_hdr)) +#define ISER_RECV_DATA_SEG_LEN 8192 +#define ISER_RX_PAYLOAD_SIZE (ISER_HEADERS_LEN + ISER_RECV_DATA_SEG_LEN) +#define ISER_RX_LOGIN_SIZE (ISER_HEADERS_LEN + ISCSI_DEF_MAX_RECV_SEG_LEN) + +/* QP settings */ +/* Maximal bounds on received asynchronous PDUs */ +#define ISERT_MAX_TX_MISC_PDUS 4 /* NOOP_IN(2) , ASYNC_EVENT(2) */ + +#define ISERT_MAX_RX_MISC_PDUS 6 /* + * NOOP_OUT(2), TEXT(1), + * SCSI_TMFUNC(2), LOGOUT(1) + */ + +#define ISCSI_DEF_XMIT_CMDS_MAX 128 /* from libiscsi.h, must be power of 2 */ + +#define ISERT_QP_MAX_RECV_DTOS (ISCSI_DEF_XMIT_CMDS_MAX) + +#define ISERT_MIN_POSTED_RX (ISCSI_DEF_XMIT_CMDS_MAX >> 2) + +#define ISERT_INFLIGHT_DATAOUTS 8 + +#define ISERT_QP_MAX_REQ_DTOS (ISCSI_DEF_XMIT_CMDS_MAX * \ + (1 + ISERT_INFLIGHT_DATAOUTS) + \ + ISERT_MAX_TX_MISC_PDUS + \ + ISERT_MAX_RX_MISC_PDUS) + +#define ISER_RX_PAD_SIZE (ISER_RECV_DATA_SEG_LEN + 4096 - \ + (ISER_RX_PAYLOAD_SIZE + sizeof(u64) + sizeof(struct ib_sge))) + #define ISCSI_ISER_SG_TABLESIZE 256 #define ISER_FASTREG_LI_WRID 0xffffffffffffffffULL #define ISER_BEACON_WRID 0xfffffffffffffffeULL @@ -56,7 +90,7 @@ enum iser_conn_state { }; struct iser_rx_desc { - struct iser_hdr iser_header; + struct iser_ctrl iser_header; struct iscsi_hdr iscsi_header; char data[ISER_RECV_DATA_SEG_LEN]; u64 dma_addr; @@ -65,7 +99,7 @@ struct iser_rx_desc { } __packed; struct iser_tx_desc { - struct iser_hdr iser_header; + struct iser_ctrl iser_header; struct iscsi_hdr iscsi_header; enum isert_desc_type type; u64 dma_addr; @@ -129,6 +163,7 @@ struct isert_cmd { uint32_t write_stag; uint64_t read_va; uint64_t write_va; + uint32_t inv_rkey; u64 pdu_buf_dma; u32 pdu_buf_len; struct isert_conn *conn; @@ -176,6 +211,7 @@ struct isert_conn { struct work_struct release_work; struct ib_recv_wr beacon; bool logout_posted; + bool snd_w_inv; }; #define ISERT_MAX_CQ 64 @@ -207,7 +243,6 @@ struct isert_device { struct isert_comp *comps; int comps_used; struct list_head dev_node; - struct ib_device_attr dev_attr; int (*reg_rdma_mem)(struct iscsi_conn *conn, struct iscsi_cmd *cmd, struct isert_rdma_wr *wr); diff --git a/drivers/infiniband/ulp/isert/isert_proto.h b/drivers/infiniband/ulp/isert/isert_proto.h deleted file mode 100644 index 4dccd313b777..000000000000 --- a/drivers/infiniband/ulp/isert/isert_proto.h +++ /dev/null @@ -1,47 +0,0 @@ -/* From iscsi_iser.h */ - -struct iser_hdr { - u8 flags; - u8 rsvd[3]; - __be32 write_stag; /* write rkey */ - __be64 write_va; - __be32 read_stag; /* read rkey */ - __be64 read_va; -} __packed; - -/*Constant PDU lengths calculations */ -#define ISER_HEADERS_LEN (sizeof(struct iser_hdr) + sizeof(struct iscsi_hdr)) - -#define ISER_RECV_DATA_SEG_LEN 8192 -#define ISER_RX_PAYLOAD_SIZE (ISER_HEADERS_LEN + ISER_RECV_DATA_SEG_LEN) -#define ISER_RX_LOGIN_SIZE (ISER_HEADERS_LEN + ISCSI_DEF_MAX_RECV_SEG_LEN) - -/* QP settings */ -/* Maximal bounds on received asynchronous PDUs */ -#define ISERT_MAX_TX_MISC_PDUS 4 /* NOOP_IN(2) , ASYNC_EVENT(2) */ - -#define ISERT_MAX_RX_MISC_PDUS 6 /* NOOP_OUT(2), TEXT(1), * - * SCSI_TMFUNC(2), LOGOUT(1) */ - -#define ISCSI_DEF_XMIT_CMDS_MAX 128 /* from libiscsi.h, must be power of 2 */ - -#define ISERT_QP_MAX_RECV_DTOS (ISCSI_DEF_XMIT_CMDS_MAX) - -#define ISERT_MIN_POSTED_RX (ISCSI_DEF_XMIT_CMDS_MAX >> 2) - -#define ISERT_INFLIGHT_DATAOUTS 8 - -#define ISERT_QP_MAX_REQ_DTOS (ISCSI_DEF_XMIT_CMDS_MAX * \ - (1 + ISERT_INFLIGHT_DATAOUTS) + \ - ISERT_MAX_TX_MISC_PDUS + \ - ISERT_MAX_RX_MISC_PDUS) - -#define ISER_RX_PAD_SIZE (ISER_RECV_DATA_SEG_LEN + 4096 - \ - (ISER_RX_PAYLOAD_SIZE + sizeof(u64) + sizeof(struct ib_sge))) - -#define ISER_VER 0x10 -#define ISER_WSV 0x08 -#define ISER_RSV 0x04 -#define ISCSI_CTRL 0x10 -#define ISER_HELLO 0x20 -#define ISER_HELLORPLY 0x30 diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 3db9a659719b..b6bf20496021 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -132,8 +132,9 @@ MODULE_PARM_DESC(ch_count, static void srp_add_one(struct ib_device *device); static void srp_remove_one(struct ib_device *device, void *client_data); -static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr); -static void srp_send_completion(struct ib_cq *cq, void *ch_ptr); +static void srp_recv_done(struct ib_cq *cq, struct ib_wc *wc); +static void srp_handle_qp_err(struct ib_cq *cq, struct ib_wc *wc, + const char *opname); static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event); static struct scsi_transport_template *ib_srp_transport_template; @@ -449,33 +450,13 @@ static struct srp_fr_pool *srp_alloc_fr_pool(struct srp_target_port *target) * srp_destroy_qp() - destroy an RDMA queue pair * @ch: SRP RDMA channel. * - * Change a queue pair into the error state and wait until all receive - * completions have been processed before destroying it. This avoids that - * the receive completion handler can access the queue pair while it is + * Drain the qp before destroying it. This avoids that the receive + * completion handler can access the queue pair while it is * being destroyed. */ static void srp_destroy_qp(struct srp_rdma_ch *ch) { - static struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR }; - static struct ib_recv_wr wr = { .wr_id = SRP_LAST_WR_ID }; - struct ib_recv_wr *bad_wr; - int ret; - - /* Destroying a QP and reusing ch->done is only safe if not connected */ - WARN_ON_ONCE(ch->connected); - - ret = ib_modify_qp(ch->qp, &attr, IB_QP_STATE); - WARN_ONCE(ret, "ib_cm_init_qp_attr() returned %d\n", ret); - if (ret) - goto out; - - init_completion(&ch->done); - ret = ib_post_recv(ch->qp, &wr, &bad_wr); - WARN_ONCE(ret, "ib_post_recv() returned %d\n", ret); - if (ret == 0) - wait_for_completion(&ch->done); - -out: + ib_drain_rq(ch->qp); ib_destroy_qp(ch->qp); } @@ -489,34 +470,27 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch) struct ib_fmr_pool *fmr_pool = NULL; struct srp_fr_pool *fr_pool = NULL; const int m = dev->use_fast_reg ? 3 : 1; - struct ib_cq_init_attr cq_attr = {}; int ret; init_attr = kzalloc(sizeof *init_attr, GFP_KERNEL); if (!init_attr) return -ENOMEM; - /* + 1 for SRP_LAST_WR_ID */ - cq_attr.cqe = target->queue_size + 1; - cq_attr.comp_vector = ch->comp_vector; - recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, ch, - &cq_attr); + /* queue_size + 1 for ib_drain_rq() */ + recv_cq = ib_alloc_cq(dev->dev, ch, target->queue_size + 1, + ch->comp_vector, IB_POLL_SOFTIRQ); if (IS_ERR(recv_cq)) { ret = PTR_ERR(recv_cq); goto err; } - cq_attr.cqe = m * target->queue_size; - cq_attr.comp_vector = ch->comp_vector; - send_cq = ib_create_cq(dev->dev, srp_send_completion, NULL, ch, - &cq_attr); + send_cq = ib_alloc_cq(dev->dev, ch, m * target->queue_size, + ch->comp_vector, IB_POLL_DIRECT); if (IS_ERR(send_cq)) { ret = PTR_ERR(send_cq); goto err_recv_cq; } - ib_req_notify_cq(recv_cq, IB_CQ_NEXT_COMP); - init_attr->event_handler = srp_qp_event; init_attr->cap.max_send_wr = m * target->queue_size; init_attr->cap.max_recv_wr = target->queue_size + 1; @@ -558,9 +532,9 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch) if (ch->qp) srp_destroy_qp(ch); if (ch->recv_cq) - ib_destroy_cq(ch->recv_cq); + ib_free_cq(ch->recv_cq); if (ch->send_cq) - ib_destroy_cq(ch->send_cq); + ib_free_cq(ch->send_cq); ch->qp = qp; ch->recv_cq = recv_cq; @@ -580,13 +554,13 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch) return 0; err_qp: - ib_destroy_qp(qp); + srp_destroy_qp(ch); err_send_cq: - ib_destroy_cq(send_cq); + ib_free_cq(send_cq); err_recv_cq: - ib_destroy_cq(recv_cq); + ib_free_cq(recv_cq); err: kfree(init_attr); @@ -622,9 +596,10 @@ static void srp_free_ch_ib(struct srp_target_port *target, if (ch->fmr_pool) ib_destroy_fmr_pool(ch->fmr_pool); } + srp_destroy_qp(ch); - ib_destroy_cq(ch->send_cq); - ib_destroy_cq(ch->recv_cq); + ib_free_cq(ch->send_cq); + ib_free_cq(ch->recv_cq); /* * Avoid that the SCSI error handler tries to use this channel after @@ -1041,18 +1016,25 @@ out: return ret <= 0 ? ret : -ENODEV; } -static int srp_inv_rkey(struct srp_rdma_ch *ch, u32 rkey) +static void srp_inv_rkey_err_done(struct ib_cq *cq, struct ib_wc *wc) +{ + srp_handle_qp_err(cq, wc, "INV RKEY"); +} + +static int srp_inv_rkey(struct srp_request *req, struct srp_rdma_ch *ch, + u32 rkey) { struct ib_send_wr *bad_wr; struct ib_send_wr wr = { .opcode = IB_WR_LOCAL_INV, - .wr_id = LOCAL_INV_WR_ID_MASK, .next = NULL, .num_sge = 0, .send_flags = 0, .ex.invalidate_rkey = rkey, }; + wr.wr_cqe = &req->reg_cqe; + req->reg_cqe.done = srp_inv_rkey_err_done; return ib_post_send(ch->qp, &wr, &bad_wr); } @@ -1074,7 +1056,7 @@ static void srp_unmap_data(struct scsi_cmnd *scmnd, struct srp_fr_desc **pfr; for (i = req->nmdesc, pfr = req->fr_list; i > 0; i--, pfr++) { - res = srp_inv_rkey(ch, (*pfr)->mr->rkey); + res = srp_inv_rkey(req, ch, (*pfr)->mr->rkey); if (res < 0) { shost_printk(KERN_ERR, target->scsi_host, PFX "Queueing INV WR for rkey %#x failed (%d)\n", @@ -1312,7 +1294,13 @@ reset_state: return 0; } +static void srp_reg_mr_err_done(struct ib_cq *cq, struct ib_wc *wc) +{ + srp_handle_qp_err(cq, wc, "FAST REG"); +} + static int srp_map_finish_fr(struct srp_map_state *state, + struct srp_request *req, struct srp_rdma_ch *ch, int sg_nents) { struct srp_target_port *target = ch->target; @@ -1349,9 +1337,11 @@ static int srp_map_finish_fr(struct srp_map_state *state, if (unlikely(n < 0)) return n; + req->reg_cqe.done = srp_reg_mr_err_done; + wr.wr.next = NULL; wr.wr.opcode = IB_WR_REG_MR; - wr.wr.wr_id = FAST_REG_WR_ID_MASK; + wr.wr.wr_cqe = &req->reg_cqe; wr.wr.num_sge = 0; wr.wr.send_flags = 0; wr.mr = desc->mr; @@ -1455,7 +1445,7 @@ static int srp_map_sg_fr(struct srp_map_state *state, struct srp_rdma_ch *ch, while (count) { int i, n; - n = srp_map_finish_fr(state, ch, count); + n = srp_map_finish_fr(state, req, ch, count); if (unlikely(n < 0)) return n; @@ -1524,7 +1514,7 @@ static int srp_map_idb(struct srp_rdma_ch *ch, struct srp_request *req, #ifdef CONFIG_NEED_SG_DMA_LENGTH idb_sg->dma_length = idb_sg->length; /* hack^2 */ #endif - ret = srp_map_finish_fr(&state, ch, 1); + ret = srp_map_finish_fr(&state, req, ch, 1); if (ret < 0) return ret; } else if (dev->use_fmr) { @@ -1719,7 +1709,7 @@ static struct srp_iu *__srp_get_tx_iu(struct srp_rdma_ch *ch, s32 rsv = (iu_type == SRP_IU_TSK_MGMT) ? 0 : SRP_TSK_MGMT_SQ_SIZE; struct srp_iu *iu; - srp_send_completion(ch->send_cq, ch); + ib_process_cq_direct(ch->send_cq, -1); if (list_empty(&ch->free_tx)) return NULL; @@ -1739,6 +1729,19 @@ static struct srp_iu *__srp_get_tx_iu(struct srp_rdma_ch *ch, return iu; } +static void srp_send_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct srp_iu *iu = container_of(wc->wr_cqe, struct srp_iu, cqe); + struct srp_rdma_ch *ch = cq->cq_context; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + srp_handle_qp_err(cq, wc, "SEND"); + return; + } + + list_add(&iu->list, &ch->free_tx); +} + static int srp_post_send(struct srp_rdma_ch *ch, struct srp_iu *iu, int len) { struct srp_target_port *target = ch->target; @@ -1749,8 +1752,10 @@ static int srp_post_send(struct srp_rdma_ch *ch, struct srp_iu *iu, int len) list.length = len; list.lkey = target->lkey; + iu->cqe.done = srp_send_done; + wr.next = NULL; - wr.wr_id = (uintptr_t) iu; + wr.wr_cqe = &iu->cqe; wr.sg_list = &list; wr.num_sge = 1; wr.opcode = IB_WR_SEND; @@ -1769,8 +1774,10 @@ static int srp_post_recv(struct srp_rdma_ch *ch, struct srp_iu *iu) list.length = iu->size; list.lkey = target->lkey; + iu->cqe.done = srp_recv_done; + wr.next = NULL; - wr.wr_id = (uintptr_t) iu; + wr.wr_cqe = &iu->cqe; wr.sg_list = &list; wr.num_sge = 1; @@ -1902,14 +1909,20 @@ static void srp_process_aer_req(struct srp_rdma_ch *ch, "problems processing SRP_AER_REQ\n"); } -static void srp_handle_recv(struct srp_rdma_ch *ch, struct ib_wc *wc) +static void srp_recv_done(struct ib_cq *cq, struct ib_wc *wc) { + struct srp_iu *iu = container_of(wc->wr_cqe, struct srp_iu, cqe); + struct srp_rdma_ch *ch = cq->cq_context; struct srp_target_port *target = ch->target; struct ib_device *dev = target->srp_host->srp_dev->dev; - struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id; int res; u8 opcode; + if (unlikely(wc->status != IB_WC_SUCCESS)) { + srp_handle_qp_err(cq, wc, "RECV"); + return; + } + ib_dma_sync_single_for_cpu(dev, iu->dma, ch->max_ti_iu_len, DMA_FROM_DEVICE); @@ -1972,68 +1985,22 @@ static void srp_tl_err_work(struct work_struct *work) srp_start_tl_fail_timers(target->rport); } -static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status, - bool send_err, struct srp_rdma_ch *ch) +static void srp_handle_qp_err(struct ib_cq *cq, struct ib_wc *wc, + const char *opname) { + struct srp_rdma_ch *ch = cq->cq_context; struct srp_target_port *target = ch->target; - if (wr_id == SRP_LAST_WR_ID) { - complete(&ch->done); - return; - } - if (ch->connected && !target->qp_in_error) { - if (wr_id & LOCAL_INV_WR_ID_MASK) { - shost_printk(KERN_ERR, target->scsi_host, PFX - "LOCAL_INV failed with status %s (%d)\n", - ib_wc_status_msg(wc_status), wc_status); - } else if (wr_id & FAST_REG_WR_ID_MASK) { - shost_printk(KERN_ERR, target->scsi_host, PFX - "FAST_REG_MR failed status %s (%d)\n", - ib_wc_status_msg(wc_status), wc_status); - } else { - shost_printk(KERN_ERR, target->scsi_host, - PFX "failed %s status %s (%d) for iu %p\n", - send_err ? "send" : "receive", - ib_wc_status_msg(wc_status), wc_status, - (void *)(uintptr_t)wr_id); - } + shost_printk(KERN_ERR, target->scsi_host, + PFX "failed %s status %s (%d) for CQE %p\n", + opname, ib_wc_status_msg(wc->status), wc->status, + wc->wr_cqe); queue_work(system_long_wq, &target->tl_err_work); } target->qp_in_error = true; } -static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr) -{ - struct srp_rdma_ch *ch = ch_ptr; - struct ib_wc wc; - - ib_req_notify_cq(cq, IB_CQ_NEXT_COMP); - while (ib_poll_cq(cq, 1, &wc) > 0) { - if (likely(wc.status == IB_WC_SUCCESS)) { - srp_handle_recv(ch, &wc); - } else { - srp_handle_qp_err(wc.wr_id, wc.status, false, ch); - } - } -} - -static void srp_send_completion(struct ib_cq *cq, void *ch_ptr) -{ - struct srp_rdma_ch *ch = ch_ptr; - struct ib_wc wc; - struct srp_iu *iu; - - while (ib_poll_cq(cq, 1, &wc) > 0) { - if (likely(wc.status == IB_WC_SUCCESS)) { - iu = (struct srp_iu *) (uintptr_t) wc.wr_id; - list_add(&iu->list, &ch->free_tx); - } else { - srp_handle_qp_err(wc.wr_id, wc.status, true, ch); - } - } -} - static int srp_queuecommand(struct Scsi_Host *shost, struct scsi_cmnd *scmnd) { struct srp_target_port *target = host_to_target(shost); @@ -3439,27 +3406,17 @@ free_host: static void srp_add_one(struct ib_device *device) { struct srp_device *srp_dev; - struct ib_device_attr *dev_attr; struct srp_host *host; int mr_page_shift, p; u64 max_pages_per_mr; - dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL); - if (!dev_attr) - return; - - if (ib_query_device(device, dev_attr)) { - pr_warn("Query device failed for %s\n", device->name); - goto free_attr; - } - srp_dev = kmalloc(sizeof *srp_dev, GFP_KERNEL); if (!srp_dev) - goto free_attr; + return; srp_dev->has_fmr = (device->alloc_fmr && device->dealloc_fmr && device->map_phys_fmr && device->unmap_fmr); - srp_dev->has_fr = (dev_attr->device_cap_flags & + srp_dev->has_fr = (device->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS); if (!srp_dev->has_fmr && !srp_dev->has_fr) dev_warn(&device->dev, "neither FMR nor FR is supported\n"); @@ -3473,23 +3430,23 @@ static void srp_add_one(struct ib_device *device) * minimum of 4096 bytes. We're unlikely to build large sglists * out of smaller entries. */ - mr_page_shift = max(12, ffs(dev_attr->page_size_cap) - 1); + mr_page_shift = max(12, ffs(device->attrs.page_size_cap) - 1); srp_dev->mr_page_size = 1 << mr_page_shift; srp_dev->mr_page_mask = ~((u64) srp_dev->mr_page_size - 1); - max_pages_per_mr = dev_attr->max_mr_size; + max_pages_per_mr = device->attrs.max_mr_size; do_div(max_pages_per_mr, srp_dev->mr_page_size); srp_dev->max_pages_per_mr = min_t(u64, SRP_MAX_PAGES_PER_MR, max_pages_per_mr); if (srp_dev->use_fast_reg) { srp_dev->max_pages_per_mr = min_t(u32, srp_dev->max_pages_per_mr, - dev_attr->max_fast_reg_page_list_len); + device->attrs.max_fast_reg_page_list_len); } srp_dev->mr_max_size = srp_dev->mr_page_size * srp_dev->max_pages_per_mr; - pr_debug("%s: mr_page_shift = %d, dev_attr->max_mr_size = %#llx, dev_attr->max_fast_reg_page_list_len = %u, max_pages_per_mr = %d, mr_max_size = %#x\n", - device->name, mr_page_shift, dev_attr->max_mr_size, - dev_attr->max_fast_reg_page_list_len, + pr_debug("%s: mr_page_shift = %d, device->max_mr_size = %#llx, device->max_fast_reg_page_list_len = %u, max_pages_per_mr = %d, mr_max_size = %#x\n", + device->name, mr_page_shift, device->attrs.max_mr_size, + device->attrs.max_fast_reg_page_list_len, srp_dev->max_pages_per_mr, srp_dev->mr_max_size); INIT_LIST_HEAD(&srp_dev->dev_list); @@ -3517,17 +3474,13 @@ static void srp_add_one(struct ib_device *device) } ib_set_client_data(device, &srp_client, srp_dev); - - goto free_attr; + return; err_pd: ib_dealloc_pd(srp_dev->pd); free_dev: kfree(srp_dev); - -free_attr: - kfree(dev_attr); } static void srp_remove_one(struct ib_device *device, void *client_data) @@ -3587,8 +3540,6 @@ static int __init srp_init_module(void) { int ret; - BUILD_BUG_ON(FIELD_SIZEOF(struct ib_wc, wr_id) < sizeof(void *)); - if (srp_sg_tablesize) { pr_warn("srp_sg_tablesize is deprecated, please use cmd_sg_entries\n"); if (!cmd_sg_entries) diff --git a/drivers/infiniband/ulp/srp/ib_srp.h b/drivers/infiniband/ulp/srp/ib_srp.h index f6af531f9f32..9e05ce4a04fd 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.h +++ b/drivers/infiniband/ulp/srp/ib_srp.h @@ -66,11 +66,6 @@ enum { SRP_TAG_TSK_MGMT = 1U << 31, SRP_MAX_PAGES_PER_MR = 512, - - LOCAL_INV_WR_ID_MASK = 1, - FAST_REG_WR_ID_MASK = 2, - - SRP_LAST_WR_ID = 0xfffffffcU, }; enum srp_target_state { @@ -128,6 +123,7 @@ struct srp_request { struct srp_direct_buf *indirect_desc; dma_addr_t indirect_dma_addr; short nmdesc; + struct ib_cqe reg_cqe; }; /** @@ -231,6 +227,7 @@ struct srp_iu { void *buf; size_t size; enum dma_data_direction direction; + struct ib_cqe cqe; }; /** diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c index bc5470c43d26..25bdaeef2520 100644 --- a/drivers/infiniband/ulp/srpt/ib_srpt.c +++ b/drivers/infiniband/ulp/srpt/ib_srpt.c @@ -91,74 +91,32 @@ MODULE_PARM_DESC(srpt_service_guid, " instead of using the node_guid of the first HCA."); static struct ib_client srpt_client; -static void srpt_release_channel(struct srpt_rdma_ch *ch); +static void srpt_release_cmd(struct se_cmd *se_cmd); +static void srpt_free_ch(struct kref *kref); static int srpt_queue_status(struct se_cmd *cmd); +static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc); +static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc); +static void srpt_process_wait_list(struct srpt_rdma_ch *ch); -/** - * opposite_dma_dir() - Swap DMA_TO_DEVICE and DMA_FROM_DEVICE. - */ -static inline -enum dma_data_direction opposite_dma_dir(enum dma_data_direction dir) -{ - switch (dir) { - case DMA_TO_DEVICE: return DMA_FROM_DEVICE; - case DMA_FROM_DEVICE: return DMA_TO_DEVICE; - default: return dir; - } -} - -/** - * srpt_sdev_name() - Return the name associated with the HCA. - * - * Examples are ib0, ib1, ... - */ -static inline const char *srpt_sdev_name(struct srpt_device *sdev) -{ - return sdev->device->name; -} - -static enum rdma_ch_state srpt_get_ch_state(struct srpt_rdma_ch *ch) -{ - unsigned long flags; - enum rdma_ch_state state; - - spin_lock_irqsave(&ch->spinlock, flags); - state = ch->state; - spin_unlock_irqrestore(&ch->spinlock, flags); - return state; -} - -static enum rdma_ch_state -srpt_set_ch_state(struct srpt_rdma_ch *ch, enum rdma_ch_state new_state) -{ - unsigned long flags; - enum rdma_ch_state prev; - - spin_lock_irqsave(&ch->spinlock, flags); - prev = ch->state; - ch->state = new_state; - spin_unlock_irqrestore(&ch->spinlock, flags); - return prev; -} - -/** - * srpt_test_and_set_ch_state() - Test and set the channel state. - * - * Returns true if and only if the channel state has been set to the new state. +/* + * The only allowed channel state changes are those that change the channel + * state into a state with a higher numerical value. Hence the new > prev test. */ -static bool -srpt_test_and_set_ch_state(struct srpt_rdma_ch *ch, enum rdma_ch_state old, - enum rdma_ch_state new) +static bool srpt_set_ch_state(struct srpt_rdma_ch *ch, enum rdma_ch_state new) { unsigned long flags; enum rdma_ch_state prev; + bool changed = false; spin_lock_irqsave(&ch->spinlock, flags); prev = ch->state; - if (prev == old) + if (new > prev) { ch->state = new; + changed = true; + } spin_unlock_irqrestore(&ch->spinlock, flags); - return prev == old; + + return changed; } /** @@ -180,7 +138,7 @@ static void srpt_event_handler(struct ib_event_handler *handler, return; pr_debug("ASYNC event= %d on device= %s\n", event->event, - srpt_sdev_name(sdev)); + sdev->device->name); switch (event->event) { case IB_EVENT_PORT_ERR: @@ -218,25 +176,39 @@ static void srpt_srq_event(struct ib_event *event, void *ctx) pr_info("SRQ event %d\n", event->event); } +static const char *get_ch_state_name(enum rdma_ch_state s) +{ + switch (s) { + case CH_CONNECTING: + return "connecting"; + case CH_LIVE: + return "live"; + case CH_DISCONNECTING: + return "disconnecting"; + case CH_DRAINING: + return "draining"; + case CH_DISCONNECTED: + return "disconnected"; + } + return "???"; +} + /** * srpt_qp_event() - QP event callback function. */ static void srpt_qp_event(struct ib_event *event, struct srpt_rdma_ch *ch) { pr_debug("QP event %d on cm_id=%p sess_name=%s state=%d\n", - event->event, ch->cm_id, ch->sess_name, srpt_get_ch_state(ch)); + event->event, ch->cm_id, ch->sess_name, ch->state); switch (event->event) { case IB_EVENT_COMM_EST: ib_cm_notify(ch->cm_id, event->event); break; case IB_EVENT_QP_LAST_WQE_REACHED: - if (srpt_test_and_set_ch_state(ch, CH_DRAINING, - CH_RELEASING)) - srpt_release_channel(ch); - else - pr_debug("%s: state %d - ignored LAST_WQE.\n", - ch->sess_name, srpt_get_ch_state(ch)); + pr_debug("%s-%d, state %s: received Last WQE event.\n", + ch->sess_name, ch->qp->qp_num, + get_ch_state_name(ch->state)); break; default: pr_err("received unrecognized IB QP event %d\n", event->event); @@ -279,7 +251,7 @@ static void srpt_get_class_port_info(struct ib_dm_mad *mad) struct ib_class_port_info *cif; cif = (struct ib_class_port_info *)mad->data; - memset(cif, 0, sizeof *cif); + memset(cif, 0, sizeof(*cif)); cif->base_version = 1; cif->class_version = 1; cif->resp_time_value = 20; @@ -338,13 +310,13 @@ static void srpt_get_ioc(struct srpt_port *sport, u32 slot, return; } - memset(iocp, 0, sizeof *iocp); + memset(iocp, 0, sizeof(*iocp)); strcpy(iocp->id_string, SRPT_ID_STRING); iocp->guid = cpu_to_be64(srpt_service_guid); - iocp->vendor_id = cpu_to_be32(sdev->dev_attr.vendor_id); - iocp->device_id = cpu_to_be32(sdev->dev_attr.vendor_part_id); - iocp->device_version = cpu_to_be16(sdev->dev_attr.hw_ver); - iocp->subsys_vendor_id = cpu_to_be32(sdev->dev_attr.vendor_id); + iocp->vendor_id = cpu_to_be32(sdev->device->attrs.vendor_id); + iocp->device_id = cpu_to_be32(sdev->device->attrs.vendor_part_id); + iocp->device_version = cpu_to_be16(sdev->device->attrs.hw_ver); + iocp->subsys_vendor_id = cpu_to_be32(sdev->device->attrs.vendor_id); iocp->subsys_device_id = 0x0; iocp->io_class = cpu_to_be16(SRP_REV16A_IB_IO_CLASS); iocp->io_subclass = cpu_to_be16(SRP_IO_SUBCLASS); @@ -388,7 +360,7 @@ static void srpt_get_svc_entries(u64 ioc_guid, } svc_entries = (struct ib_dm_svc_entries *)mad->data; - memset(svc_entries, 0, sizeof *svc_entries); + memset(svc_entries, 0, sizeof(*svc_entries)); svc_entries->service_entries[0].id = cpu_to_be64(ioc_guid); snprintf(svc_entries->service_entries[0].name, sizeof(svc_entries->service_entries[0].name), @@ -453,6 +425,7 @@ static void srpt_mad_send_handler(struct ib_mad_agent *mad_agent, * srpt_mad_recv_handler() - MAD reception callback function. */ static void srpt_mad_recv_handler(struct ib_mad_agent *mad_agent, + struct ib_mad_send_buf *send_buf, struct ib_mad_recv_wc *mad_wc) { struct srpt_port *sport = (struct srpt_port *)mad_agent->context; @@ -481,7 +454,7 @@ static void srpt_mad_recv_handler(struct ib_mad_agent *mad_agent, rsp->ah = ah; dm_mad = rsp->mad; - memcpy(dm_mad, mad_wc->recv_buf.mad, sizeof *dm_mad); + memcpy(dm_mad, mad_wc->recv_buf.mad, sizeof(*dm_mad)); dm_mad->mad_hdr.method = IB_MGMT_METHOD_GET_RESP; dm_mad->mad_hdr.status = 0; @@ -529,7 +502,7 @@ static int srpt_refresh_port(struct srpt_port *sport) struct ib_port_attr port_attr; int ret; - memset(&port_modify, 0, sizeof port_modify); + memset(&port_modify, 0, sizeof(port_modify)); port_modify.set_port_cap_mask = IB_PORT_DEVICE_MGMT_SUP; port_modify.clr_port_cap_mask = 0; @@ -550,7 +523,7 @@ static int srpt_refresh_port(struct srpt_port *sport) goto err_query_port; if (!sport->mad_agent) { - memset(®_req, 0, sizeof reg_req); + memset(®_req, 0, sizeof(reg_req)); reg_req.mgmt_class = IB_MGMT_CLASS_DEVICE_MGMT; reg_req.mgmt_class_version = IB_MGMT_BASE_VERSION; set_bit(IB_MGMT_METHOD_GET, reg_req.method_mask); @@ -778,12 +751,12 @@ static int srpt_post_recv(struct srpt_device *sdev, struct ib_recv_wr wr, *bad_wr; BUG_ON(!sdev); - wr.wr_id = encode_wr_id(SRPT_RECV, ioctx->ioctx.index); - list.addr = ioctx->ioctx.dma; list.length = srp_max_req_size; list.lkey = sdev->pd->local_dma_lkey; + ioctx->ioctx.cqe.done = srpt_recv_done; + wr.wr_cqe = &ioctx->ioctx.cqe; wr.next = NULL; wr.sg_list = &list; wr.num_sge = 1; @@ -819,8 +792,9 @@ static int srpt_post_send(struct srpt_rdma_ch *ch, list.length = len; list.lkey = sdev->pd->local_dma_lkey; + ioctx->ioctx.cqe.done = srpt_send_done; wr.next = NULL; - wr.wr_id = encode_wr_id(SRPT_SEND, ioctx->ioctx.index); + wr.wr_cqe = &ioctx->ioctx.cqe; wr.sg_list = &list; wr.num_sge = 1; wr.opcode = IB_WR_SEND; @@ -837,6 +811,39 @@ out: } /** + * srpt_zerolength_write() - Perform a zero-length RDMA write. + * + * A quote from the InfiniBand specification: C9-88: For an HCA responder + * using Reliable Connection service, for each zero-length RDMA READ or WRITE + * request, the R_Key shall not be validated, even if the request includes + * Immediate data. + */ +static int srpt_zerolength_write(struct srpt_rdma_ch *ch) +{ + struct ib_send_wr wr, *bad_wr; + + memset(&wr, 0, sizeof(wr)); + wr.opcode = IB_WR_RDMA_WRITE; + wr.wr_cqe = &ch->zw_cqe; + wr.send_flags = IB_SEND_SIGNALED; + return ib_post_send(ch->qp, &wr, &bad_wr); +} + +static void srpt_zerolength_write_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct srpt_rdma_ch *ch = cq->cq_context; + + if (wc->status == IB_WC_SUCCESS) { + srpt_process_wait_list(ch); + } else { + if (srpt_set_ch_state(ch, CH_DISCONNECTED)) + schedule_work(&ch->release_work); + else + WARN_ONCE("%s-%d\n", ch->sess_name, ch->qp->qp_num); + } +} + +/** * srpt_get_desc_tbl() - Parse the data descriptors of an SRP_CMD request. * @ioctx: Pointer to the I/O context associated with the request. * @srp_cmd: Pointer to the SRP_CMD request data. @@ -899,14 +906,14 @@ static int srpt_get_desc_tbl(struct srpt_send_ioctx *ioctx, db = (struct srp_direct_buf *)(srp_cmd->add_data + add_cdb_offset); - memcpy(ioctx->rbufs, db, sizeof *db); + memcpy(ioctx->rbufs, db, sizeof(*db)); *data_len = be32_to_cpu(db->len); } else if (((srp_cmd->buf_fmt & 0xf) == SRP_DATA_DESC_INDIRECT) || ((srp_cmd->buf_fmt >> 4) == SRP_DATA_DESC_INDIRECT)) { idb = (struct srp_indirect_buf *)(srp_cmd->add_data + add_cdb_offset); - ioctx->n_rbuf = be32_to_cpu(idb->table_desc.len) / sizeof *db; + ioctx->n_rbuf = be32_to_cpu(idb->table_desc.len) / sizeof(*db); if (ioctx->n_rbuf > (srp_cmd->data_out_desc_cnt + srp_cmd->data_in_desc_cnt)) { @@ -925,7 +932,7 @@ static int srpt_get_desc_tbl(struct srpt_send_ioctx *ioctx, ioctx->rbufs = &ioctx->single_rbuf; else { ioctx->rbufs = - kmalloc(ioctx->n_rbuf * sizeof *db, GFP_ATOMIC); + kmalloc(ioctx->n_rbuf * sizeof(*db), GFP_ATOMIC); if (!ioctx->rbufs) { ioctx->n_rbuf = 0; ret = -ENOMEM; @@ -934,7 +941,7 @@ static int srpt_get_desc_tbl(struct srpt_send_ioctx *ioctx, } db = idb->desc_list; - memcpy(ioctx->rbufs, db, ioctx->n_rbuf * sizeof *db); + memcpy(ioctx->rbufs, db, ioctx->n_rbuf * sizeof(*db)); *data_len = be32_to_cpu(idb->len); } out: @@ -952,7 +959,7 @@ static int srpt_init_ch_qp(struct srpt_rdma_ch *ch, struct ib_qp *qp) struct ib_qp_attr *attr; int ret; - attr = kzalloc(sizeof *attr, GFP_KERNEL); + attr = kzalloc(sizeof(*attr), GFP_KERNEL); if (!attr) return -ENOMEM; @@ -1052,13 +1059,13 @@ static void srpt_unmap_sg_to_ib_sge(struct srpt_rdma_ch *ch, BUG_ON(!ch); BUG_ON(!ioctx); - BUG_ON(ioctx->n_rdma && !ioctx->rdma_ius); + BUG_ON(ioctx->n_rdma && !ioctx->rdma_wrs); while (ioctx->n_rdma) - kfree(ioctx->rdma_ius[--ioctx->n_rdma].sge); + kfree(ioctx->rdma_wrs[--ioctx->n_rdma].wr.sg_list); - kfree(ioctx->rdma_ius); - ioctx->rdma_ius = NULL; + kfree(ioctx->rdma_wrs); + ioctx->rdma_wrs = NULL; if (ioctx->mapped_sg_count) { sg = ioctx->sg; @@ -1066,7 +1073,7 @@ static void srpt_unmap_sg_to_ib_sge(struct srpt_rdma_ch *ch, dir = ioctx->cmd.data_direction; BUG_ON(dir == DMA_NONE); ib_dma_unmap_sg(ch->sport->sdev->device, sg, ioctx->sg_cnt, - opposite_dma_dir(dir)); + target_reverse_dma_direction(&ioctx->cmd)); ioctx->mapped_sg_count = 0; } } @@ -1082,7 +1089,7 @@ static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch, struct scatterlist *sg, *sg_orig; int sg_cnt; enum dma_data_direction dir; - struct rdma_iu *riu; + struct ib_rdma_wr *riu; struct srp_direct_buf *db; dma_addr_t dma_addr; struct ib_sge *sge; @@ -1103,29 +1110,30 @@ static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch, ioctx->sg_cnt = sg_cnt = cmd->t_data_nents; count = ib_dma_map_sg(ch->sport->sdev->device, sg, sg_cnt, - opposite_dma_dir(dir)); + target_reverse_dma_direction(cmd)); if (unlikely(!count)) return -EAGAIN; ioctx->mapped_sg_count = count; - if (ioctx->rdma_ius && ioctx->n_rdma_ius) - nrdma = ioctx->n_rdma_ius; + if (ioctx->rdma_wrs && ioctx->n_rdma_wrs) + nrdma = ioctx->n_rdma_wrs; else { nrdma = (count + SRPT_DEF_SG_PER_WQE - 1) / SRPT_DEF_SG_PER_WQE + ioctx->n_rbuf; - ioctx->rdma_ius = kzalloc(nrdma * sizeof *riu, GFP_KERNEL); - if (!ioctx->rdma_ius) + ioctx->rdma_wrs = kcalloc(nrdma, sizeof(*ioctx->rdma_wrs), + GFP_KERNEL); + if (!ioctx->rdma_wrs) goto free_mem; - ioctx->n_rdma_ius = nrdma; + ioctx->n_rdma_wrs = nrdma; } db = ioctx->rbufs; tsize = cmd->data_length; dma_len = ib_sg_dma_len(dev, &sg[0]); - riu = ioctx->rdma_ius; + riu = ioctx->rdma_wrs; /* * For each remote desc - calculate the #ib_sge. @@ -1139,9 +1147,9 @@ static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch, j < count && i < ioctx->n_rbuf && tsize > 0; ++i, ++riu, ++db) { rsize = be32_to_cpu(db->len); raddr = be64_to_cpu(db->va); - riu->raddr = raddr; + riu->remote_addr = raddr; riu->rkey = be32_to_cpu(db->key); - riu->sge_cnt = 0; + riu->wr.num_sge = 0; /* calculate how many sge required for this remote_buf */ while (rsize > 0 && tsize > 0) { @@ -1165,33 +1173,35 @@ static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch, rsize = 0; } - ++riu->sge_cnt; + ++riu->wr.num_sge; - if (rsize > 0 && riu->sge_cnt == SRPT_DEF_SG_PER_WQE) { + if (rsize > 0 && + riu->wr.num_sge == SRPT_DEF_SG_PER_WQE) { ++ioctx->n_rdma; - riu->sge = - kmalloc(riu->sge_cnt * sizeof *riu->sge, - GFP_KERNEL); - if (!riu->sge) + riu->wr.sg_list = kmalloc_array(riu->wr.num_sge, + sizeof(*riu->wr.sg_list), + GFP_KERNEL); + if (!riu->wr.sg_list) goto free_mem; ++riu; - riu->sge_cnt = 0; - riu->raddr = raddr; + riu->wr.num_sge = 0; + riu->remote_addr = raddr; riu->rkey = be32_to_cpu(db->key); } } ++ioctx->n_rdma; - riu->sge = kmalloc(riu->sge_cnt * sizeof *riu->sge, - GFP_KERNEL); - if (!riu->sge) + riu->wr.sg_list = kmalloc_array(riu->wr.num_sge, + sizeof(*riu->wr.sg_list), + GFP_KERNEL); + if (!riu->wr.sg_list) goto free_mem; } db = ioctx->rbufs; tsize = cmd->data_length; - riu = ioctx->rdma_ius; + riu = ioctx->rdma_wrs; sg = sg_orig; dma_len = ib_sg_dma_len(dev, &sg[0]); dma_addr = ib_sg_dma_address(dev, &sg[0]); @@ -1200,7 +1210,7 @@ static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch, for (i = 0, j = 0; j < count && i < ioctx->n_rbuf && tsize > 0; ++i, ++riu, ++db) { rsize = be32_to_cpu(db->len); - sge = riu->sge; + sge = riu->wr.sg_list; k = 0; while (rsize > 0 && tsize > 0) { @@ -1232,9 +1242,9 @@ static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch, } ++k; - if (k == riu->sge_cnt && rsize > 0 && tsize > 0) { + if (k == riu->wr.num_sge && rsize > 0 && tsize > 0) { ++riu; - sge = riu->sge; + sge = riu->wr.sg_list; k = 0; } else if (rsize > 0 && tsize > 0) ++sge; @@ -1277,8 +1287,8 @@ static struct srpt_send_ioctx *srpt_get_send_ioctx(struct srpt_rdma_ch *ch) ioctx->n_rbuf = 0; ioctx->rbufs = NULL; ioctx->n_rdma = 0; - ioctx->n_rdma_ius = 0; - ioctx->rdma_ius = NULL; + ioctx->n_rdma_wrs = 0; + ioctx->rdma_wrs = NULL; ioctx->mapped_sg_count = 0; init_completion(&ioctx->tx_done); ioctx->queue_status_only = false; @@ -1306,10 +1316,7 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) /* * If the command is in a state where the target core is waiting for - * the ib_srpt driver, change the state to the next state. Changing - * the state of the command from SRPT_STATE_NEED_DATA to - * SRPT_STATE_DATA_IN ensures that srpt_xmit_response() will call this - * function a second time. + * the ib_srpt driver, change the state to the next state. */ spin_lock_irqsave(&ioctx->spinlock, flags); @@ -1318,25 +1325,17 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) case SRPT_STATE_NEED_DATA: ioctx->state = SRPT_STATE_DATA_IN; break; - case SRPT_STATE_DATA_IN: case SRPT_STATE_CMD_RSP_SENT: case SRPT_STATE_MGMT_RSP_SENT: ioctx->state = SRPT_STATE_DONE; break; default: + WARN_ONCE(true, "%s: unexpected I/O context state %d\n", + __func__, state); break; } spin_unlock_irqrestore(&ioctx->spinlock, flags); - if (state == SRPT_STATE_DONE) { - struct srpt_rdma_ch *ch = ioctx->ch; - - BUG_ON(ch->sess == NULL); - - target_put_sess_cmd(&ioctx->cmd); - goto out; - } - pr_debug("Aborting cmd with state %d and tag %lld\n", state, ioctx->cmd.tag); @@ -1344,19 +1343,16 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) case SRPT_STATE_NEW: case SRPT_STATE_DATA_IN: case SRPT_STATE_MGMT: + case SRPT_STATE_DONE: /* * Do nothing - defer abort processing until * srpt_queue_response() is invoked. */ - WARN_ON(!transport_check_aborted_status(&ioctx->cmd, false)); break; case SRPT_STATE_NEED_DATA: - /* DMA_TO_DEVICE (write) - RDMA read error. */ - - /* XXX(hch): this is a horrible layering violation.. */ - spin_lock_irqsave(&ioctx->cmd.t_state_lock, flags); - ioctx->cmd.transport_state &= ~CMD_T_ACTIVE; - spin_unlock_irqrestore(&ioctx->cmd.t_state_lock, flags); + pr_debug("tag %#llx: RDMA read error\n", ioctx->cmd.tag); + transport_generic_request_failure(&ioctx->cmd, + TCM_CHECK_CONDITION_ABORT_CMD); break; case SRPT_STATE_CMD_RSP_SENT: /* @@ -1364,134 +1360,63 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) * not been received in time. */ srpt_unmap_sg_to_ib_sge(ioctx->ch, ioctx); - target_put_sess_cmd(&ioctx->cmd); + transport_generic_free_cmd(&ioctx->cmd, 0); break; case SRPT_STATE_MGMT_RSP_SENT: - srpt_set_cmd_state(ioctx, SRPT_STATE_DONE); - target_put_sess_cmd(&ioctx->cmd); + transport_generic_free_cmd(&ioctx->cmd, 0); break; default: WARN(1, "Unexpected command state (%d)", state); break; } -out: return state; } /** - * srpt_handle_send_err_comp() - Process an IB_WC_SEND error completion. - */ -static void srpt_handle_send_err_comp(struct srpt_rdma_ch *ch, u64 wr_id) -{ - struct srpt_send_ioctx *ioctx; - enum srpt_command_state state; - u32 index; - - atomic_inc(&ch->sq_wr_avail); - - index = idx_from_wr_id(wr_id); - ioctx = ch->ioctx_ring[index]; - state = srpt_get_cmd_state(ioctx); - - WARN_ON(state != SRPT_STATE_CMD_RSP_SENT - && state != SRPT_STATE_MGMT_RSP_SENT - && state != SRPT_STATE_NEED_DATA - && state != SRPT_STATE_DONE); - - /* If SRP_RSP sending failed, undo the ch->req_lim change. */ - if (state == SRPT_STATE_CMD_RSP_SENT - || state == SRPT_STATE_MGMT_RSP_SENT) - atomic_dec(&ch->req_lim); - - srpt_abort_cmd(ioctx); -} - -/** - * srpt_handle_send_comp() - Process an IB send completion notification. - */ -static void srpt_handle_send_comp(struct srpt_rdma_ch *ch, - struct srpt_send_ioctx *ioctx) -{ - enum srpt_command_state state; - - atomic_inc(&ch->sq_wr_avail); - - state = srpt_set_cmd_state(ioctx, SRPT_STATE_DONE); - - if (WARN_ON(state != SRPT_STATE_CMD_RSP_SENT - && state != SRPT_STATE_MGMT_RSP_SENT - && state != SRPT_STATE_DONE)) - pr_debug("state = %d\n", state); - - if (state != SRPT_STATE_DONE) { - srpt_unmap_sg_to_ib_sge(ch, ioctx); - transport_generic_free_cmd(&ioctx->cmd, 0); - } else { - pr_err("IB completion has been received too late for" - " wr_id = %u.\n", ioctx->ioctx.index); - } -} - -/** - * srpt_handle_rdma_comp() - Process an IB RDMA completion notification. - * * XXX: what is now target_execute_cmd used to be asynchronous, and unmapping * the data that has been transferred via IB RDMA had to be postponed until the * check_stop_free() callback. None of this is necessary anymore and needs to * be cleaned up. */ -static void srpt_handle_rdma_comp(struct srpt_rdma_ch *ch, - struct srpt_send_ioctx *ioctx, - enum srpt_opcode opcode) +static void srpt_rdma_read_done(struct ib_cq *cq, struct ib_wc *wc) { + struct srpt_rdma_ch *ch = cq->cq_context; + struct srpt_send_ioctx *ioctx = + container_of(wc->wr_cqe, struct srpt_send_ioctx, rdma_cqe); + WARN_ON(ioctx->n_rdma <= 0); atomic_add(ioctx->n_rdma, &ch->sq_wr_avail); - if (opcode == SRPT_RDMA_READ_LAST) { - if (srpt_test_and_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA, - SRPT_STATE_DATA_IN)) - target_execute_cmd(&ioctx->cmd); - else - pr_err("%s[%d]: wrong state = %d\n", __func__, - __LINE__, srpt_get_cmd_state(ioctx)); - } else if (opcode == SRPT_RDMA_ABORT) { - ioctx->rdma_aborted = true; - } else { - WARN(true, "unexpected opcode %d\n", opcode); + if (unlikely(wc->status != IB_WC_SUCCESS)) { + pr_info("RDMA_READ for ioctx 0x%p failed with status %d\n", + ioctx, wc->status); + srpt_abort_cmd(ioctx); + return; } + + if (srpt_test_and_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA, + SRPT_STATE_DATA_IN)) + target_execute_cmd(&ioctx->cmd); + else + pr_err("%s[%d]: wrong state = %d\n", __func__, + __LINE__, srpt_get_cmd_state(ioctx)); } -/** - * srpt_handle_rdma_err_comp() - Process an IB RDMA error completion. - */ -static void srpt_handle_rdma_err_comp(struct srpt_rdma_ch *ch, - struct srpt_send_ioctx *ioctx, - enum srpt_opcode opcode) +static void srpt_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc) { - enum srpt_command_state state; + struct srpt_send_ioctx *ioctx = + container_of(wc->wr_cqe, struct srpt_send_ioctx, rdma_cqe); - state = srpt_get_cmd_state(ioctx); - switch (opcode) { - case SRPT_RDMA_READ_LAST: - if (ioctx->n_rdma <= 0) { - pr_err("Received invalid RDMA read" - " error completion with idx %d\n", - ioctx->ioctx.index); - break; - } - atomic_add(ioctx->n_rdma, &ch->sq_wr_avail); - if (state == SRPT_STATE_NEED_DATA) - srpt_abort_cmd(ioctx); - else - pr_err("%s[%d]: wrong state = %d\n", - __func__, __LINE__, state); - break; - case SRPT_RDMA_WRITE_LAST: - break; - default: - pr_err("%s[%d]: opcode = %u\n", __func__, __LINE__, opcode); - break; + if (unlikely(wc->status != IB_WC_SUCCESS)) { + /* + * Note: if an RDMA write error completion is received that + * means that a SEND also has been posted. Defer further + * processing of the associated command until the send error + * completion has been received. + */ + pr_info("RDMA_WRITE for ioctx 0x%p failed with status %d\n", + ioctx, wc->status); } } @@ -1531,7 +1456,7 @@ static int srpt_build_cmd_rsp(struct srpt_rdma_ch *ch, sense_data_len = ioctx->cmd.scsi_sense_length; WARN_ON(sense_data_len > sizeof(ioctx->sense_data)); - memset(srp_rsp, 0, sizeof *srp_rsp); + memset(srp_rsp, 0, sizeof(*srp_rsp)); srp_rsp->opcode = SRP_RSP; srp_rsp->req_lim_delta = cpu_to_be32(1 + atomic_xchg(&ch->req_lim_delta, 0)); @@ -1581,7 +1506,7 @@ static int srpt_build_tskmgmt_rsp(struct srpt_rdma_ch *ch, srp_rsp = ioctx->ioctx.buf; BUG_ON(!srp_rsp); - memset(srp_rsp, 0, sizeof *srp_rsp); + memset(srp_rsp, 0, sizeof(*srp_rsp)); srp_rsp->opcode = SRP_RSP; srp_rsp->req_lim_delta = @@ -1595,80 +1520,6 @@ static int srpt_build_tskmgmt_rsp(struct srpt_rdma_ch *ch, return resp_len; } -#define NO_SUCH_LUN ((uint64_t)-1LL) - -/* - * SCSI LUN addressing method. See also SAM-2 and the section about - * eight byte LUNs. - */ -enum scsi_lun_addr_method { - SCSI_LUN_ADDR_METHOD_PERIPHERAL = 0, - SCSI_LUN_ADDR_METHOD_FLAT = 1, - SCSI_LUN_ADDR_METHOD_LUN = 2, - SCSI_LUN_ADDR_METHOD_EXTENDED_LUN = 3, -}; - -/* - * srpt_unpack_lun() - Convert from network LUN to linear LUN. - * - * Convert an 2-byte, 4-byte, 6-byte or 8-byte LUN structure in network byte - * order (big endian) to a linear LUN. Supports three LUN addressing methods: - * peripheral, flat and logical unit. See also SAM-2, section 4.9.4 (page 40). - */ -static uint64_t srpt_unpack_lun(const uint8_t *lun, int len) -{ - uint64_t res = NO_SUCH_LUN; - int addressing_method; - - if (unlikely(len < 2)) { - pr_err("Illegal LUN length %d, expected 2 bytes or more\n", - len); - goto out; - } - - switch (len) { - case 8: - if ((*((__be64 *)lun) & - cpu_to_be64(0x0000FFFFFFFFFFFFLL)) != 0) - goto out_err; - break; - case 4: - if (*((__be16 *)&lun[2]) != 0) - goto out_err; - break; - case 6: - if (*((__be32 *)&lun[2]) != 0) - goto out_err; - break; - case 2: - break; - default: - goto out_err; - } - - addressing_method = (*lun) >> 6; /* highest two bits of byte 0 */ - switch (addressing_method) { - case SCSI_LUN_ADDR_METHOD_PERIPHERAL: - case SCSI_LUN_ADDR_METHOD_FLAT: - case SCSI_LUN_ADDR_METHOD_LUN: - res = *(lun + 1) | (((*lun) & 0x3f) << 8); - break; - - case SCSI_LUN_ADDR_METHOD_EXTENDED_LUN: - default: - pr_err("Unimplemented LUN addressing method %u\n", - addressing_method); - break; - } - -out: - return res; - -out_err: - pr_err("Support for multi-level LUNs has not yet been implemented\n"); - goto out; -} - static int srpt_check_stop_free(struct se_cmd *cmd) { struct srpt_send_ioctx *ioctx = container_of(cmd, @@ -1680,16 +1531,14 @@ static int srpt_check_stop_free(struct se_cmd *cmd) /** * srpt_handle_cmd() - Process SRP_CMD. */ -static int srpt_handle_cmd(struct srpt_rdma_ch *ch, - struct srpt_recv_ioctx *recv_ioctx, - struct srpt_send_ioctx *send_ioctx) +static void srpt_handle_cmd(struct srpt_rdma_ch *ch, + struct srpt_recv_ioctx *recv_ioctx, + struct srpt_send_ioctx *send_ioctx) { struct se_cmd *cmd; struct srp_cmd *srp_cmd; - uint64_t unpacked_lun; u64 data_len; enum dma_data_direction dir; - sense_reason_t ret; int rc; BUG_ON(!send_ioctx); @@ -1717,65 +1566,23 @@ static int srpt_handle_cmd(struct srpt_rdma_ch *ch, if (srpt_get_desc_tbl(send_ioctx, srp_cmd, &dir, &data_len)) { pr_err("0x%llx: parsing SRP descriptor table failed.\n", srp_cmd->tag); - ret = TCM_INVALID_CDB_FIELD; - goto send_sense; + goto release_ioctx; } - unpacked_lun = srpt_unpack_lun((uint8_t *)&srp_cmd->lun, - sizeof(srp_cmd->lun)); rc = target_submit_cmd(cmd, ch->sess, srp_cmd->cdb, - &send_ioctx->sense_data[0], unpacked_lun, data_len, - TCM_SIMPLE_TAG, dir, TARGET_SCF_ACK_KREF); + &send_ioctx->sense_data[0], + scsilun_to_int(&srp_cmd->lun), data_len, + TCM_SIMPLE_TAG, dir, TARGET_SCF_ACK_KREF); if (rc != 0) { - ret = TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE; - goto send_sense; + pr_debug("target_submit_cmd() returned %d for tag %#llx\n", rc, + srp_cmd->tag); + goto release_ioctx; } - return 0; - -send_sense: - transport_send_check_condition_and_sense(cmd, ret, 0); - return -1; -} - -/** - * srpt_rx_mgmt_fn_tag() - Process a task management function by tag. - * @ch: RDMA channel of the task management request. - * @fn: Task management function to perform. - * @req_tag: Tag of the SRP task management request. - * @mgmt_ioctx: I/O context of the task management request. - * - * Returns zero if the target core will process the task management - * request asynchronously. - * - * Note: It is assumed that the initiator serializes tag-based task management - * requests. - */ -static int srpt_rx_mgmt_fn_tag(struct srpt_send_ioctx *ioctx, u64 tag) -{ - struct srpt_device *sdev; - struct srpt_rdma_ch *ch; - struct srpt_send_ioctx *target; - int ret, i; + return; - ret = -EINVAL; - ch = ioctx->ch; - BUG_ON(!ch); - BUG_ON(!ch->sport); - sdev = ch->sport->sdev; - BUG_ON(!sdev); - spin_lock_irq(&sdev->spinlock); - for (i = 0; i < ch->rq_size; ++i) { - target = ch->ioctx_ring[i]; - if (target->cmd.se_lun == ioctx->cmd.se_lun && - target->cmd.tag == tag && - srpt_get_cmd_state(target) != SRPT_STATE_DONE) { - ret = 0; - /* now let the target core abort &target->cmd; */ - break; - } - } - spin_unlock_irq(&sdev->spinlock); - return ret; +release_ioctx: + send_ioctx->state = SRPT_STATE_DONE; + srpt_release_cmd(cmd); } static int srp_tmr_to_tcm(int fn) @@ -1811,8 +1618,6 @@ static void srpt_handle_tsk_mgmt(struct srpt_rdma_ch *ch, struct srp_tsk_mgmt *srp_tsk; struct se_cmd *cmd; struct se_session *sess = ch->sess; - uint64_t unpacked_lun; - uint32_t tag = 0; int tcm_tmr; int rc; @@ -1828,26 +1633,10 @@ static void srpt_handle_tsk_mgmt(struct srpt_rdma_ch *ch, srpt_set_cmd_state(send_ioctx, SRPT_STATE_MGMT); send_ioctx->cmd.tag = srp_tsk->tag; tcm_tmr = srp_tmr_to_tcm(srp_tsk->tsk_mgmt_func); - if (tcm_tmr < 0) { - send_ioctx->cmd.se_tmr_req->response = - TMR_TASK_MGMT_FUNCTION_NOT_SUPPORTED; - goto fail; - } - unpacked_lun = srpt_unpack_lun((uint8_t *)&srp_tsk->lun, - sizeof(srp_tsk->lun)); - - if (srp_tsk->tsk_mgmt_func == SRP_TSK_ABORT_TASK) { - rc = srpt_rx_mgmt_fn_tag(send_ioctx, srp_tsk->task_tag); - if (rc < 0) { - send_ioctx->cmd.se_tmr_req->response = - TMR_TASK_DOES_NOT_EXIST; - goto fail; - } - tag = srp_tsk->task_tag; - } - rc = target_submit_tmr(&send_ioctx->cmd, sess, NULL, unpacked_lun, - srp_tsk, tcm_tmr, GFP_KERNEL, tag, - TARGET_SCF_ACK_KREF); + rc = target_submit_tmr(&send_ioctx->cmd, sess, NULL, + scsilun_to_int(&srp_tsk->lun), srp_tsk, tcm_tmr, + GFP_KERNEL, srp_tsk->task_tag, + TARGET_SCF_ACK_KREF); if (rc != 0) { send_ioctx->cmd.se_tmr_req->response = TMR_FUNCTION_REJECTED; goto fail; @@ -1867,7 +1656,6 @@ static void srpt_handle_new_iu(struct srpt_rdma_ch *ch, struct srpt_send_ioctx *send_ioctx) { struct srp_cmd *srp_cmd; - enum rdma_ch_state ch_state; BUG_ON(!ch); BUG_ON(!recv_ioctx); @@ -1876,13 +1664,12 @@ static void srpt_handle_new_iu(struct srpt_rdma_ch *ch, recv_ioctx->ioctx.dma, srp_max_req_size, DMA_FROM_DEVICE); - ch_state = srpt_get_ch_state(ch); - if (unlikely(ch_state == CH_CONNECTING)) { + if (unlikely(ch->state == CH_CONNECTING)) { list_add_tail(&recv_ioctx->wait_list, &ch->cmd_wait_list); goto out; } - if (unlikely(ch_state != CH_LIVE)) + if (unlikely(ch->state != CH_LIVE)) goto out; srp_cmd = recv_ioctx->ioctx.buf; @@ -1926,32 +1713,48 @@ out: return; } -static void srpt_process_rcv_completion(struct ib_cq *cq, - struct srpt_rdma_ch *ch, - struct ib_wc *wc) +static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc) { - struct srpt_device *sdev = ch->sport->sdev; - struct srpt_recv_ioctx *ioctx; - u32 index; + struct srpt_rdma_ch *ch = cq->cq_context; + struct srpt_recv_ioctx *ioctx = + container_of(wc->wr_cqe, struct srpt_recv_ioctx, ioctx.cqe); - index = idx_from_wr_id(wc->wr_id); if (wc->status == IB_WC_SUCCESS) { int req_lim; req_lim = atomic_dec_return(&ch->req_lim); if (unlikely(req_lim < 0)) pr_err("req_lim = %d < 0\n", req_lim); - ioctx = sdev->ioctx_ring[index]; srpt_handle_new_iu(ch, ioctx, NULL); } else { - pr_info("receiving failed for idx %u with status %d\n", - index, wc->status); + pr_info("receiving failed for ioctx %p with status %d\n", + ioctx, wc->status); + } +} + +/* + * This function must be called from the context in which RDMA completions are + * processed because it accesses the wait list without protection against + * access from other threads. + */ +static void srpt_process_wait_list(struct srpt_rdma_ch *ch) +{ + struct srpt_send_ioctx *ioctx; + + while (!list_empty(&ch->cmd_wait_list) && + ch->state >= CH_LIVE && + (ioctx = srpt_get_send_ioctx(ch)) != NULL) { + struct srpt_recv_ioctx *recv_ioctx; + + recv_ioctx = list_first_entry(&ch->cmd_wait_list, + struct srpt_recv_ioctx, + wait_list); + list_del(&recv_ioctx->wait_list); + srpt_handle_new_iu(ch, recv_ioctx, ioctx); } } /** - * srpt_process_send_completion() - Process an IB send completion. - * * Note: Although this has not yet been observed during tests, at least in * theory it is possible that the srpt_get_send_ioctx() call invoked by * srpt_handle_new_iu() fails. This is possible because the req_lim_delta @@ -1964,106 +1767,33 @@ static void srpt_process_rcv_completion(struct ib_cq *cq, * are queued on cmd_wait_list. The code below processes these delayed * requests one at a time. */ -static void srpt_process_send_completion(struct ib_cq *cq, - struct srpt_rdma_ch *ch, - struct ib_wc *wc) +static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc) { - struct srpt_send_ioctx *send_ioctx; - uint32_t index; - enum srpt_opcode opcode; - - index = idx_from_wr_id(wc->wr_id); - opcode = opcode_from_wr_id(wc->wr_id); - send_ioctx = ch->ioctx_ring[index]; - if (wc->status == IB_WC_SUCCESS) { - if (opcode == SRPT_SEND) - srpt_handle_send_comp(ch, send_ioctx); - else { - WARN_ON(opcode != SRPT_RDMA_ABORT && - wc->opcode != IB_WC_RDMA_READ); - srpt_handle_rdma_comp(ch, send_ioctx, opcode); - } - } else { - if (opcode == SRPT_SEND) { - pr_info("sending response for idx %u failed" - " with status %d\n", index, wc->status); - srpt_handle_send_err_comp(ch, wc->wr_id); - } else if (opcode != SRPT_RDMA_MID) { - pr_info("RDMA t %d for idx %u failed with" - " status %d\n", opcode, index, wc->status); - srpt_handle_rdma_err_comp(ch, send_ioctx, opcode); - } - } + struct srpt_rdma_ch *ch = cq->cq_context; + struct srpt_send_ioctx *ioctx = + container_of(wc->wr_cqe, struct srpt_send_ioctx, ioctx.cqe); + enum srpt_command_state state; - while (unlikely(opcode == SRPT_SEND - && !list_empty(&ch->cmd_wait_list) - && srpt_get_ch_state(ch) == CH_LIVE - && (send_ioctx = srpt_get_send_ioctx(ch)) != NULL)) { - struct srpt_recv_ioctx *recv_ioctx; + state = srpt_set_cmd_state(ioctx, SRPT_STATE_DONE); - recv_ioctx = list_first_entry(&ch->cmd_wait_list, - struct srpt_recv_ioctx, - wait_list); - list_del(&recv_ioctx->wait_list); - srpt_handle_new_iu(ch, recv_ioctx, send_ioctx); - } -} + WARN_ON(state != SRPT_STATE_CMD_RSP_SENT && + state != SRPT_STATE_MGMT_RSP_SENT); -static void srpt_process_completion(struct ib_cq *cq, struct srpt_rdma_ch *ch) -{ - struct ib_wc *const wc = ch->wc; - int i, n; + atomic_inc(&ch->sq_wr_avail); - WARN_ON(cq != ch->cq); + if (wc->status != IB_WC_SUCCESS) + pr_info("sending response for ioctx 0x%p failed" + " with status %d\n", ioctx, wc->status); - ib_req_notify_cq(cq, IB_CQ_NEXT_COMP); - while ((n = ib_poll_cq(cq, ARRAY_SIZE(ch->wc), wc)) > 0) { - for (i = 0; i < n; i++) { - if (opcode_from_wr_id(wc[i].wr_id) == SRPT_RECV) - srpt_process_rcv_completion(cq, ch, &wc[i]); - else - srpt_process_send_completion(cq, ch, &wc[i]); - } + if (state != SRPT_STATE_DONE) { + srpt_unmap_sg_to_ib_sge(ch, ioctx); + transport_generic_free_cmd(&ioctx->cmd, 0); + } else { + pr_err("IB completion has been received too late for" + " wr_id = %u.\n", ioctx->ioctx.index); } -} -/** - * srpt_completion() - IB completion queue callback function. - * - * Notes: - * - It is guaranteed that a completion handler will never be invoked - * concurrently on two different CPUs for the same completion queue. See also - * Documentation/infiniband/core_locking.txt and the implementation of - * handle_edge_irq() in kernel/irq/chip.c. - * - When threaded IRQs are enabled, completion handlers are invoked in thread - * context instead of interrupt context. - */ -static void srpt_completion(struct ib_cq *cq, void *ctx) -{ - struct srpt_rdma_ch *ch = ctx; - - wake_up_interruptible(&ch->wait_queue); -} - -static int srpt_compl_thread(void *arg) -{ - struct srpt_rdma_ch *ch; - - /* Hibernation / freezing of the SRPT kernel thread is not supported. */ - current->flags |= PF_NOFREEZE; - - ch = arg; - BUG_ON(!ch); - pr_info("Session %s: kernel thread %s (PID %d) started\n", - ch->sess_name, ch->thread->comm, current->pid); - while (!kthread_should_stop()) { - wait_event_interruptible(ch->wait_queue, - (srpt_process_completion(ch->cq, ch), - kthread_should_stop())); - } - pr_info("Session %s: kernel thread %s (PID %d) stopped\n", - ch->sess_name, ch->thread->comm, current->pid); - return 0; + srpt_process_wait_list(ch); } /** @@ -2075,20 +1805,18 @@ static int srpt_create_ch_ib(struct srpt_rdma_ch *ch) struct srpt_port *sport = ch->sport; struct srpt_device *sdev = sport->sdev; u32 srp_sq_size = sport->port_attrib.srp_sq_size; - struct ib_cq_init_attr cq_attr = {}; int ret; WARN_ON(ch->rq_size < 1); ret = -ENOMEM; - qp_init = kzalloc(sizeof *qp_init, GFP_KERNEL); + qp_init = kzalloc(sizeof(*qp_init), GFP_KERNEL); if (!qp_init) goto out; retry: - cq_attr.cqe = ch->rq_size + srp_sq_size; - ch->cq = ib_create_cq(sdev->device, srpt_completion, NULL, ch, - &cq_attr); + ch->cq = ib_alloc_cq(sdev->device, ch, ch->rq_size + srp_sq_size, + 0 /* XXX: spread CQs */, IB_POLL_WORKQUEUE); if (IS_ERR(ch->cq)) { ret = PTR_ERR(ch->cq); pr_err("failed to create CQ cqe= %d ret= %d\n", @@ -2131,18 +1859,6 @@ retry: if (ret) goto err_destroy_qp; - init_waitqueue_head(&ch->wait_queue); - - pr_debug("creating thread for session %s\n", ch->sess_name); - - ch->thread = kthread_run(srpt_compl_thread, ch, "ib_srpt_compl"); - if (IS_ERR(ch->thread)) { - pr_err("failed to create kernel thread %ld\n", - PTR_ERR(ch->thread)); - ch->thread = NULL; - goto err_destroy_qp; - } - out: kfree(qp_init); return ret; @@ -2150,182 +1866,113 @@ out: err_destroy_qp: ib_destroy_qp(ch->qp); err_destroy_cq: - ib_destroy_cq(ch->cq); + ib_free_cq(ch->cq); goto out; } static void srpt_destroy_ch_ib(struct srpt_rdma_ch *ch) { - if (ch->thread) - kthread_stop(ch->thread); - ib_destroy_qp(ch->qp); - ib_destroy_cq(ch->cq); + ib_free_cq(ch->cq); } /** - * __srpt_close_ch() - Close an RDMA channel by setting the QP error state. + * srpt_close_ch() - Close an RDMA channel. * - * Reset the QP and make sure all resources associated with the channel will - * be deallocated at an appropriate time. + * Make sure all resources associated with the channel will be deallocated at + * an appropriate time. * - * Note: The caller must hold ch->sport->sdev->spinlock. + * Returns true if and only if the channel state has been modified into + * CH_DRAINING. */ -static void __srpt_close_ch(struct srpt_rdma_ch *ch) +static bool srpt_close_ch(struct srpt_rdma_ch *ch) { - enum rdma_ch_state prev_state; - unsigned long flags; + int ret; - spin_lock_irqsave(&ch->spinlock, flags); - prev_state = ch->state; - switch (prev_state) { - case CH_CONNECTING: - case CH_LIVE: - ch->state = CH_DISCONNECTING; - break; - default: - break; + if (!srpt_set_ch_state(ch, CH_DRAINING)) { + pr_debug("%s-%d: already closed\n", ch->sess_name, + ch->qp->qp_num); + return false; } - spin_unlock_irqrestore(&ch->spinlock, flags); - switch (prev_state) { - case CH_CONNECTING: - ib_send_cm_rej(ch->cm_id, IB_CM_REJ_NO_RESOURCES, NULL, 0, - NULL, 0); - /* fall through */ - case CH_LIVE: - if (ib_send_cm_dreq(ch->cm_id, NULL, 0) < 0) - pr_err("sending CM DREQ failed.\n"); - break; - case CH_DISCONNECTING: - break; - case CH_DRAINING: - case CH_RELEASING: - break; - } -} + kref_get(&ch->kref); -/** - * srpt_close_ch() - Close an RDMA channel. - */ -static void srpt_close_ch(struct srpt_rdma_ch *ch) -{ - struct srpt_device *sdev; - - sdev = ch->sport->sdev; - spin_lock_irq(&sdev->spinlock); - __srpt_close_ch(ch); - spin_unlock_irq(&sdev->spinlock); -} - -/** - * srpt_shutdown_session() - Whether or not a session may be shut down. - */ -static int srpt_shutdown_session(struct se_session *se_sess) -{ - struct srpt_rdma_ch *ch = se_sess->fabric_sess_ptr; - unsigned long flags; + ret = srpt_ch_qp_err(ch); + if (ret < 0) + pr_err("%s-%d: changing queue pair into error state failed: %d\n", + ch->sess_name, ch->qp->qp_num, ret); - spin_lock_irqsave(&ch->spinlock, flags); - if (ch->in_shutdown) { - spin_unlock_irqrestore(&ch->spinlock, flags); - return true; + pr_debug("%s-%d: queued zerolength write\n", ch->sess_name, + ch->qp->qp_num); + ret = srpt_zerolength_write(ch); + if (ret < 0) { + pr_err("%s-%d: queuing zero-length write failed: %d\n", + ch->sess_name, ch->qp->qp_num, ret); + if (srpt_set_ch_state(ch, CH_DISCONNECTED)) + schedule_work(&ch->release_work); + else + WARN_ON_ONCE(true); } - ch->in_shutdown = true; - target_sess_cmd_list_set_waiting(se_sess); - spin_unlock_irqrestore(&ch->spinlock, flags); + kref_put(&ch->kref, srpt_free_ch); return true; } -/** - * srpt_drain_channel() - Drain a channel by resetting the IB queue pair. - * @cm_id: Pointer to the CM ID of the channel to be drained. - * - * Note: Must be called from inside srpt_cm_handler to avoid a race between - * accessing sdev->spinlock and the call to kfree(sdev) in srpt_remove_one() - * (the caller of srpt_cm_handler holds the cm_id spinlock; srpt_remove_one() - * waits until all target sessions for the associated IB device have been - * unregistered and target session registration involves a call to - * ib_destroy_cm_id(), which locks the cm_id spinlock and hence waits until - * this function has finished). +/* + * Change the channel state into CH_DISCONNECTING. If a channel has not yet + * reached the connected state, close it. If a channel is in the connected + * state, send a DREQ. If a DREQ has been received, send a DREP. Note: it is + * the responsibility of the caller to ensure that this function is not + * invoked concurrently with the code that accepts a connection. This means + * that this function must either be invoked from inside a CM callback + * function or that it must be invoked with the srpt_port.mutex held. */ -static void srpt_drain_channel(struct ib_cm_id *cm_id) +static int srpt_disconnect_ch(struct srpt_rdma_ch *ch) { - struct srpt_device *sdev; - struct srpt_rdma_ch *ch; int ret; - bool do_reset = false; - WARN_ON_ONCE(irqs_disabled()); + if (!srpt_set_ch_state(ch, CH_DISCONNECTING)) + return -ENOTCONN; - sdev = cm_id->context; - BUG_ON(!sdev); - spin_lock_irq(&sdev->spinlock); - list_for_each_entry(ch, &sdev->rch_list, list) { - if (ch->cm_id == cm_id) { - do_reset = srpt_test_and_set_ch_state(ch, - CH_CONNECTING, CH_DRAINING) || - srpt_test_and_set_ch_state(ch, - CH_LIVE, CH_DRAINING) || - srpt_test_and_set_ch_state(ch, - CH_DISCONNECTING, CH_DRAINING); - break; - } - } - spin_unlock_irq(&sdev->spinlock); + ret = ib_send_cm_dreq(ch->cm_id, NULL, 0); + if (ret < 0) + ret = ib_send_cm_drep(ch->cm_id, NULL, 0); - if (do_reset) { - if (ch->sess) - srpt_shutdown_session(ch->sess); + if (ret < 0 && srpt_close_ch(ch)) + ret = 0; - ret = srpt_ch_qp_err(ch); - if (ret < 0) - pr_err("Setting queue pair in error state" - " failed: %d\n", ret); - } + return ret; } -/** - * srpt_find_channel() - Look up an RDMA channel. - * @cm_id: Pointer to the CM ID of the channel to be looked up. - * - * Return NULL if no matching RDMA channel has been found. - */ -static struct srpt_rdma_ch *srpt_find_channel(struct srpt_device *sdev, - struct ib_cm_id *cm_id) +static void __srpt_close_all_ch(struct srpt_device *sdev) { struct srpt_rdma_ch *ch; - bool found; - WARN_ON_ONCE(irqs_disabled()); - BUG_ON(!sdev); + lockdep_assert_held(&sdev->mutex); - found = false; - spin_lock_irq(&sdev->spinlock); list_for_each_entry(ch, &sdev->rch_list, list) { - if (ch->cm_id == cm_id) { - found = true; - break; - } + if (srpt_disconnect_ch(ch) >= 0) + pr_info("Closing channel %s-%d because target %s has been disabled\n", + ch->sess_name, ch->qp->qp_num, + sdev->device->name); + srpt_close_ch(ch); } - spin_unlock_irq(&sdev->spinlock); - - return found ? ch : NULL; } /** - * srpt_release_channel() - Release channel resources. - * - * Schedules the actual release because: - * - Calling the ib_destroy_cm_id() call from inside an IB CM callback would - * trigger a deadlock. - * - It is not safe to call TCM transport_* functions from interrupt context. + * srpt_shutdown_session() - Whether or not a session may be shut down. */ -static void srpt_release_channel(struct srpt_rdma_ch *ch) +static int srpt_shutdown_session(struct se_session *se_sess) +{ + return 1; +} + +static void srpt_free_ch(struct kref *kref) { - schedule_work(&ch->release_work); + struct srpt_rdma_ch *ch = container_of(kref, struct srpt_rdma_ch, kref); + + kfree(ch); } static void srpt_release_channel_work(struct work_struct *w) @@ -2335,8 +1982,8 @@ static void srpt_release_channel_work(struct work_struct *w) struct se_session *se_sess; ch = container_of(w, struct srpt_rdma_ch, release_work); - pr_debug("ch = %p; ch->sess = %p; release_done = %p\n", ch, ch->sess, - ch->release_done); + pr_debug("%s: %s-%d; release_done = %p\n", __func__, ch->sess_name, + ch->qp->qp_num, ch->release_done); sdev = ch->sport->sdev; BUG_ON(!sdev); @@ -2344,6 +1991,7 @@ static void srpt_release_channel_work(struct work_struct *w) se_sess = ch->sess; BUG_ON(!se_sess); + target_sess_cmd_list_set_waiting(se_sess); target_wait_for_sess_cmds(se_sess); transport_deregister_session_configfs(se_sess); @@ -2358,16 +2006,15 @@ static void srpt_release_channel_work(struct work_struct *w) ch->sport->sdev, ch->rq_size, ch->rsp_size, DMA_TO_DEVICE); - spin_lock_irq(&sdev->spinlock); - list_del(&ch->list); - spin_unlock_irq(&sdev->spinlock); - + mutex_lock(&sdev->mutex); + list_del_init(&ch->list); if (ch->release_done) complete(ch->release_done); + mutex_unlock(&sdev->mutex); wake_up(&sdev->ch_releaseQ); - kfree(ch); + kref_put(&ch->kref, srpt_free_ch); } /** @@ -2413,9 +2060,9 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, be64_to_cpu(*(__be64 *)&sdev->port[param->port - 1].gid.raw[0]), be64_to_cpu(*(__be64 *)&sdev->port[param->port - 1].gid.raw[8])); - rsp = kzalloc(sizeof *rsp, GFP_KERNEL); - rej = kzalloc(sizeof *rej, GFP_KERNEL); - rep_param = kzalloc(sizeof *rep_param, GFP_KERNEL); + rsp = kzalloc(sizeof(*rsp), GFP_KERNEL); + rej = kzalloc(sizeof(*rej), GFP_KERNEL); + rep_param = kzalloc(sizeof(*rep_param), GFP_KERNEL); if (!rsp || !rej || !rep_param) { ret = -ENOMEM; @@ -2444,7 +2091,7 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, if ((req->req_flags & SRP_MTCH_ACTION) == SRP_MULTICHAN_SINGLE) { rsp->rsp_flags = SRP_LOGIN_RSP_MULTICHAN_NO_CHAN; - spin_lock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); list_for_each_entry_safe(ch, tmp_ch, &sdev->rch_list, list) { if (!memcmp(ch->i_port_id, req->initiator_port_id, 16) @@ -2452,26 +2099,16 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, && param->port == ch->sport->port && param->listen_id == ch->sport->sdev->cm_id && ch->cm_id) { - enum rdma_ch_state ch_state; - - ch_state = srpt_get_ch_state(ch); - if (ch_state != CH_CONNECTING - && ch_state != CH_LIVE) + if (srpt_disconnect_ch(ch) < 0) continue; - - /* found an existing channel */ - pr_debug("Found existing channel %s" - " cm_id= %p state= %d\n", - ch->sess_name, ch->cm_id, ch_state); - - __srpt_close_ch(ch); - + pr_info("Relogin - closed existing channel %s\n", + ch->sess_name); rsp->rsp_flags = SRP_LOGIN_RSP_MULTICHAN_TERMINATED; } } - spin_unlock_irq(&sdev->spinlock); + mutex_unlock(&sdev->mutex); } else rsp->rsp_flags = SRP_LOGIN_RSP_MULTICHAN_MAINTAINED; @@ -2487,7 +2124,7 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, goto reject; } - ch = kzalloc(sizeof *ch, GFP_KERNEL); + ch = kzalloc(sizeof(*ch), GFP_KERNEL); if (!ch) { rej->reason = cpu_to_be32( SRP_LOGIN_REJ_INSUFFICIENT_RESOURCES); @@ -2496,11 +2133,14 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, goto reject; } + kref_init(&ch->kref); + ch->zw_cqe.done = srpt_zerolength_write_done; INIT_WORK(&ch->release_work, srpt_release_channel_work); memcpy(ch->i_port_id, req->initiator_port_id, 16); memcpy(ch->t_port_id, req->target_port_id, 16); ch->sport = &sdev->port[param->port - 1]; ch->cm_id = cm_id; + cm_id->context = ch; /* * Avoid QUEUE_FULL conditions by limiting the number of buffers used * for the SRP protocol to the command queue size. @@ -2600,7 +2240,7 @@ try_again: /* create cm reply */ rep_param->qp_num = ch->qp->qp_num; rep_param->private_data = (void *)rsp; - rep_param->private_data_len = sizeof *rsp; + rep_param->private_data_len = sizeof(*rsp); rep_param->rnr_retry_count = 7; rep_param->flow_control = 1; rep_param->failover_accepted = 0; @@ -2615,14 +2255,14 @@ try_again: goto release_channel; } - spin_lock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); list_add_tail(&ch->list, &sdev->rch_list); - spin_unlock_irq(&sdev->spinlock); + mutex_unlock(&sdev->mutex); goto out; release_channel: - srpt_set_ch_state(ch, CH_RELEASING); + srpt_disconnect_ch(ch); transport_deregister_session_configfs(ch->sess); transport_deregister_session(ch->sess); ch->sess = NULL; @@ -2644,7 +2284,7 @@ reject: | SRP_BUF_FORMAT_INDIRECT); ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED, NULL, 0, - (void *)rej, sizeof *rej); + (void *)rej, sizeof(*rej)); out: kfree(rep_param); @@ -2654,10 +2294,23 @@ out: return ret; } -static void srpt_cm_rej_recv(struct ib_cm_id *cm_id) +static void srpt_cm_rej_recv(struct srpt_rdma_ch *ch, + enum ib_cm_rej_reason reason, + const u8 *private_data, + u8 private_data_len) { - pr_info("Received IB REJ for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); + char *priv = NULL; + int i; + + if (private_data_len && (priv = kmalloc(private_data_len * 3 + 1, + GFP_KERNEL))) { + for (i = 0; i < private_data_len; i++) + sprintf(priv + 3 * i, " %02x", private_data[i]); + } + pr_info("Received CM REJ for ch %s-%d; reason %d%s%s.\n", + ch->sess_name, ch->qp->qp_num, reason, private_data_len ? + "; private data" : "", priv ? priv : " (?)"); + kfree(priv); } /** @@ -2666,87 +2319,23 @@ static void srpt_cm_rej_recv(struct ib_cm_id *cm_id) * An IB_CM_RTU_RECEIVED message indicates that the connection is established * and that the recipient may begin transmitting (RTU = ready to use). */ -static void srpt_cm_rtu_recv(struct ib_cm_id *cm_id) +static void srpt_cm_rtu_recv(struct srpt_rdma_ch *ch) { - struct srpt_rdma_ch *ch; int ret; - ch = srpt_find_channel(cm_id->context, cm_id); - BUG_ON(!ch); - - if (srpt_test_and_set_ch_state(ch, CH_CONNECTING, CH_LIVE)) { - struct srpt_recv_ioctx *ioctx, *ioctx_tmp; - + if (srpt_set_ch_state(ch, CH_LIVE)) { ret = srpt_ch_qp_rts(ch, ch->qp); - list_for_each_entry_safe(ioctx, ioctx_tmp, &ch->cmd_wait_list, - wait_list) { - list_del(&ioctx->wait_list); - srpt_handle_new_iu(ch, ioctx, NULL); - } - if (ret) + if (ret == 0) { + /* Trigger wait list processing. */ + ret = srpt_zerolength_write(ch); + WARN_ONCE(ret < 0, "%d\n", ret); + } else { srpt_close_ch(ch); + } } } -static void srpt_cm_timewait_exit(struct ib_cm_id *cm_id) -{ - pr_info("Received IB TimeWait exit for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); -} - -static void srpt_cm_rep_error(struct ib_cm_id *cm_id) -{ - pr_info("Received IB REP error for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); -} - -/** - * srpt_cm_dreq_recv() - Process reception of a DREQ message. - */ -static void srpt_cm_dreq_recv(struct ib_cm_id *cm_id) -{ - struct srpt_rdma_ch *ch; - unsigned long flags; - bool send_drep = false; - - ch = srpt_find_channel(cm_id->context, cm_id); - BUG_ON(!ch); - - pr_debug("cm_id= %p ch->state= %d\n", cm_id, srpt_get_ch_state(ch)); - - spin_lock_irqsave(&ch->spinlock, flags); - switch (ch->state) { - case CH_CONNECTING: - case CH_LIVE: - send_drep = true; - ch->state = CH_DISCONNECTING; - break; - case CH_DISCONNECTING: - case CH_DRAINING: - case CH_RELEASING: - WARN(true, "unexpected channel state %d\n", ch->state); - break; - } - spin_unlock_irqrestore(&ch->spinlock, flags); - - if (send_drep) { - if (ib_send_cm_drep(ch->cm_id, NULL, 0) < 0) - pr_err("Sending IB DREP failed.\n"); - pr_info("Received DREQ and sent DREP for session %s.\n", - ch->sess_name); - } -} - -/** - * srpt_cm_drep_recv() - Process reception of a DREP message. - */ -static void srpt_cm_drep_recv(struct ib_cm_id *cm_id) -{ - pr_info("Received InfiniBand DREP message for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); -} - /** * srpt_cm_handler() - IB connection manager callback function. * @@ -2759,6 +2348,7 @@ static void srpt_cm_drep_recv(struct ib_cm_id *cm_id) */ static int srpt_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) { + struct srpt_rdma_ch *ch = cm_id->context; int ret; ret = 0; @@ -2768,32 +2358,39 @@ static int srpt_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) event->private_data); break; case IB_CM_REJ_RECEIVED: - srpt_cm_rej_recv(cm_id); + srpt_cm_rej_recv(ch, event->param.rej_rcvd.reason, + event->private_data, + IB_CM_REJ_PRIVATE_DATA_SIZE); break; case IB_CM_RTU_RECEIVED: case IB_CM_USER_ESTABLISHED: - srpt_cm_rtu_recv(cm_id); + srpt_cm_rtu_recv(ch); break; case IB_CM_DREQ_RECEIVED: - srpt_cm_dreq_recv(cm_id); + srpt_disconnect_ch(ch); break; case IB_CM_DREP_RECEIVED: - srpt_cm_drep_recv(cm_id); + pr_info("Received CM DREP message for ch %s-%d.\n", + ch->sess_name, ch->qp->qp_num); + srpt_close_ch(ch); break; case IB_CM_TIMEWAIT_EXIT: - srpt_cm_timewait_exit(cm_id); + pr_info("Received CM TimeWait exit for ch %s-%d.\n", + ch->sess_name, ch->qp->qp_num); + srpt_close_ch(ch); break; case IB_CM_REP_ERROR: - srpt_cm_rep_error(cm_id); + pr_info("Received CM REP error for ch %s-%d.\n", ch->sess_name, + ch->qp->qp_num); break; case IB_CM_DREQ_ERROR: - pr_info("Received IB DREQ ERROR event.\n"); + pr_info("Received CM DREQ ERROR event.\n"); break; case IB_CM_MRA_RECEIVED: - pr_info("Received IB MRA event\n"); + pr_info("Received CM MRA event\n"); break; default: - pr_err("received unrecognized IB CM event %d\n", event->event); + pr_err("received unrecognized CM event %d\n", event->event); break; } @@ -2808,12 +2405,8 @@ static int srpt_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) static int srpt_perform_rdmas(struct srpt_rdma_ch *ch, struct srpt_send_ioctx *ioctx) { - struct ib_rdma_wr wr; struct ib_send_wr *bad_wr; - struct rdma_iu *riu; - int i; - int ret; - int sq_wr_avail; + int sq_wr_avail, ret, i; enum dma_data_direction dir; const int n_rdma = ioctx->n_rdma; @@ -2829,59 +2422,32 @@ static int srpt_perform_rdmas(struct srpt_rdma_ch *ch, } } - ioctx->rdma_aborted = false; - ret = 0; - riu = ioctx->rdma_ius; - memset(&wr, 0, sizeof wr); - - for (i = 0; i < n_rdma; ++i, ++riu) { - if (dir == DMA_FROM_DEVICE) { - wr.wr.opcode = IB_WR_RDMA_WRITE; - wr.wr.wr_id = encode_wr_id(i == n_rdma - 1 ? - SRPT_RDMA_WRITE_LAST : - SRPT_RDMA_MID, - ioctx->ioctx.index); - } else { - wr.wr.opcode = IB_WR_RDMA_READ; - wr.wr.wr_id = encode_wr_id(i == n_rdma - 1 ? - SRPT_RDMA_READ_LAST : - SRPT_RDMA_MID, - ioctx->ioctx.index); - } - wr.wr.next = NULL; - wr.remote_addr = riu->raddr; - wr.rkey = riu->rkey; - wr.wr.num_sge = riu->sge_cnt; - wr.wr.sg_list = riu->sge; + for (i = 0; i < n_rdma; i++) { + struct ib_send_wr *wr = &ioctx->rdma_wrs[i].wr; - /* only get completion event for the last rdma write */ - if (i == (n_rdma - 1) && dir == DMA_TO_DEVICE) - wr.wr.send_flags = IB_SEND_SIGNALED; + wr->opcode = (dir == DMA_FROM_DEVICE) ? + IB_WR_RDMA_WRITE : IB_WR_RDMA_READ; - ret = ib_post_send(ch->qp, &wr.wr, &bad_wr); - if (ret) - break; + if (i == n_rdma - 1) { + /* only get completion event for the last rdma read */ + if (dir == DMA_TO_DEVICE) { + wr->send_flags = IB_SEND_SIGNALED; + ioctx->rdma_cqe.done = srpt_rdma_read_done; + } else { + ioctx->rdma_cqe.done = srpt_rdma_write_done; + } + wr->wr_cqe = &ioctx->rdma_cqe; + wr->next = NULL; + } else { + wr->wr_cqe = NULL; + wr->next = &ioctx->rdma_wrs[i + 1].wr; + } } + ret = ib_post_send(ch->qp, &ioctx->rdma_wrs->wr, &bad_wr); if (ret) pr_err("%s[%d]: ib_post_send() returned %d for %d/%d\n", __func__, __LINE__, ret, i, n_rdma); - if (ret && i > 0) { - wr.wr.num_sge = 0; - wr.wr.wr_id = encode_wr_id(SRPT_RDMA_ABORT, ioctx->ioctx.index); - wr.wr.send_flags = IB_SEND_SIGNALED; - while (ch->state == CH_LIVE && - ib_post_send(ch->qp, &wr.wr, &bad_wr) != 0) { - pr_info("Trying to abort failed RDMA transfer [%d]\n", - ioctx->ioctx.index); - msleep(1000); - } - while (ch->state != CH_RELEASING && !ioctx->rdma_aborted) { - pr_info("Waiting until RDMA abort finished [%d]\n", - ioctx->ioctx.index); - msleep(1000); - } - } out: if (unlikely(dir == DMA_TO_DEVICE && ret < 0)) atomic_add(n_rdma, &ch->sq_wr_avail); @@ -2933,41 +2499,14 @@ static int srpt_write_pending_status(struct se_cmd *se_cmd) */ static int srpt_write_pending(struct se_cmd *se_cmd) { - struct srpt_rdma_ch *ch; - struct srpt_send_ioctx *ioctx; + struct srpt_send_ioctx *ioctx = + container_of(se_cmd, struct srpt_send_ioctx, cmd); + struct srpt_rdma_ch *ch = ioctx->ch; enum srpt_command_state new_state; - enum rdma_ch_state ch_state; - int ret; - - ioctx = container_of(se_cmd, struct srpt_send_ioctx, cmd); new_state = srpt_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA); WARN_ON(new_state == SRPT_STATE_DONE); - - ch = ioctx->ch; - BUG_ON(!ch); - - ch_state = srpt_get_ch_state(ch); - switch (ch_state) { - case CH_CONNECTING: - WARN(true, "unexpected channel state %d\n", ch_state); - ret = -EINVAL; - goto out; - case CH_LIVE: - break; - case CH_DISCONNECTING: - case CH_DRAINING: - case CH_RELEASING: - pr_debug("cmd with tag %lld: channel disconnecting\n", - ioctx->cmd.tag); - srpt_set_cmd_state(ioctx, SRPT_STATE_DATA_IN); - ret = -EINVAL; - goto out; - } - ret = srpt_xfer_data(ch, ioctx); - -out: - return ret; + return srpt_xfer_data(ch, ioctx); } static u8 tcm_to_srp_tsk_mgmt_status(const int tcm_mgmt_status) @@ -3098,36 +2637,25 @@ static void srpt_refresh_port_work(struct work_struct *work) srpt_refresh_port(sport); } -static int srpt_ch_list_empty(struct srpt_device *sdev) -{ - int res; - - spin_lock_irq(&sdev->spinlock); - res = list_empty(&sdev->rch_list); - spin_unlock_irq(&sdev->spinlock); - - return res; -} - /** * srpt_release_sdev() - Free the channel resources associated with a target. */ static int srpt_release_sdev(struct srpt_device *sdev) { - struct srpt_rdma_ch *ch, *tmp_ch; - int res; + int i, res; WARN_ON_ONCE(irqs_disabled()); BUG_ON(!sdev); - spin_lock_irq(&sdev->spinlock); - list_for_each_entry_safe(ch, tmp_ch, &sdev->rch_list, list) - __srpt_close_ch(ch); - spin_unlock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); + for (i = 0; i < ARRAY_SIZE(sdev->port); i++) + sdev->port[i].enabled = false; + __srpt_close_all_ch(sdev); + mutex_unlock(&sdev->mutex); res = wait_event_interruptible(sdev->ch_releaseQ, - srpt_ch_list_empty(sdev)); + list_empty_careful(&sdev->rch_list)); if (res) pr_err("%s: interrupted.\n", __func__); @@ -3181,23 +2709,20 @@ static void srpt_add_one(struct ib_device *device) pr_debug("device = %p, device->dma_ops = %p\n", device, device->dma_ops); - sdev = kzalloc(sizeof *sdev, GFP_KERNEL); + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL); if (!sdev) goto err; sdev->device = device; INIT_LIST_HEAD(&sdev->rch_list); init_waitqueue_head(&sdev->ch_releaseQ); - spin_lock_init(&sdev->spinlock); - - if (ib_query_device(device, &sdev->dev_attr)) - goto free_dev; + mutex_init(&sdev->mutex); sdev->pd = ib_alloc_pd(device); if (IS_ERR(sdev->pd)) goto free_dev; - sdev->srq_size = min(srpt_srq_size, sdev->dev_attr.max_srq_wr); + sdev->srq_size = min(srpt_srq_size, sdev->device->attrs.max_srq_wr); srq_attr.event_handler = srpt_srq_event; srq_attr.srq_context = (void *)sdev; @@ -3211,7 +2736,7 @@ static void srpt_add_one(struct ib_device *device) goto err_pd; pr_debug("%s: create SRQ #wr= %d max_allow=%d dev= %s\n", - __func__, sdev->srq_size, sdev->dev_attr.max_srq_wr, + __func__, sdev->srq_size, sdev->device->attrs.max_srq_wr, device->name); if (!srpt_service_guid) @@ -3263,7 +2788,7 @@ static void srpt_add_one(struct ib_device *device) if (srpt_refresh_port(sport)) { pr_err("MAD registration failed for %s-%d.\n", - srpt_sdev_name(sdev), i); + sdev->device->name, i); goto err_ring; } snprintf(sport->port_guid, sizeof(sport->port_guid), @@ -3412,24 +2937,26 @@ static void srpt_release_cmd(struct se_cmd *se_cmd) static void srpt_close_session(struct se_session *se_sess) { DECLARE_COMPLETION_ONSTACK(release_done); - struct srpt_rdma_ch *ch; - struct srpt_device *sdev; - unsigned long res; - - ch = se_sess->fabric_sess_ptr; - WARN_ON(ch->sess != se_sess); + struct srpt_rdma_ch *ch = se_sess->fabric_sess_ptr; + struct srpt_device *sdev = ch->sport->sdev; + bool wait; - pr_debug("ch %p state %d\n", ch, srpt_get_ch_state(ch)); + pr_debug("ch %s-%d state %d\n", ch->sess_name, ch->qp->qp_num, + ch->state); - sdev = ch->sport->sdev; - spin_lock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); BUG_ON(ch->release_done); ch->release_done = &release_done; - __srpt_close_ch(ch); - spin_unlock_irq(&sdev->spinlock); + wait = !list_empty(&ch->list); + srpt_disconnect_ch(ch); + mutex_unlock(&sdev->mutex); - res = wait_for_completion_timeout(&release_done, 60 * HZ); - WARN_ON(res == 0); + if (!wait) + return; + + while (wait_for_completion_timeout(&release_done, 180 * HZ) == 0) + pr_info("%s(%s-%d state %d): still waiting ...\n", __func__, + ch->sess_name, ch->qp->qp_num, ch->state); } /** @@ -3637,6 +3164,8 @@ static ssize_t srpt_tpg_enable_store(struct config_item *item, { struct se_portal_group *se_tpg = to_tpg(item); struct srpt_port *sport = container_of(se_tpg, struct srpt_port, port_tpg_1); + struct srpt_device *sdev = sport->sdev; + struct srpt_rdma_ch *ch; unsigned long tmp; int ret; @@ -3650,11 +3179,24 @@ static ssize_t srpt_tpg_enable_store(struct config_item *item, pr_err("Illegal value for srpt_tpg_store_enable: %lu\n", tmp); return -EINVAL; } - if (tmp == 1) - sport->enabled = true; - else - sport->enabled = false; + if (sport->enabled == tmp) + goto out; + sport->enabled = tmp; + if (sport->enabled) + goto out; + mutex_lock(&sdev->mutex); + list_for_each_entry(ch, &sdev->rch_list, list) { + if (ch->sport == sport) { + pr_debug("%s: ch %p %s-%d\n", __func__, ch, + ch->sess_name, ch->qp->qp_num); + srpt_disconnect_ch(ch); + srpt_close_ch(ch); + } + } + mutex_unlock(&sdev->mutex); + +out: return count; } @@ -3746,7 +3288,6 @@ static struct configfs_attribute *srpt_wwn_attrs[] = { static const struct target_core_fabric_ops srpt_template = { .module = THIS_MODULE, .name = "srpt", - .node_acl_size = sizeof(struct srpt_node_acl), .get_fabric_name = srpt_get_fabric_name, .tpg_get_wwn = srpt_get_fabric_wwn, .tpg_get_tag = srpt_get_tag, diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.h b/drivers/infiniband/ulp/srpt/ib_srpt.h index 5366e0a9fd6d..af9b8b527340 100644 --- a/drivers/infiniband/ulp/srpt/ib_srpt.h +++ b/drivers/infiniband/ulp/srpt/ib_srpt.h @@ -128,36 +128,6 @@ enum { DEFAULT_MAX_RDMA_SIZE = 65536, }; -enum srpt_opcode { - SRPT_RECV, - SRPT_SEND, - SRPT_RDMA_MID, - SRPT_RDMA_ABORT, - SRPT_RDMA_READ_LAST, - SRPT_RDMA_WRITE_LAST, -}; - -static inline u64 encode_wr_id(u8 opcode, u32 idx) -{ - return ((u64)opcode << 32) | idx; -} -static inline enum srpt_opcode opcode_from_wr_id(u64 wr_id) -{ - return wr_id >> 32; -} -static inline u32 idx_from_wr_id(u64 wr_id) -{ - return (u32)wr_id; -} - -struct rdma_iu { - u64 raddr; - u32 rkey; - struct ib_sge *sge; - u32 sge_cnt; - int mem_id; -}; - /** * enum srpt_command_state - SCSI command state managed by SRPT. * @SRPT_STATE_NEW: New command arrived and is being processed. @@ -189,6 +159,7 @@ enum srpt_command_state { * @index: Index of the I/O context in its ioctx_ring array. */ struct srpt_ioctx { + struct ib_cqe cqe; void *buf; dma_addr_t dma; uint32_t index; @@ -215,32 +186,30 @@ struct srpt_recv_ioctx { * @sg: Pointer to sg-list associated with this I/O context. * @sg_cnt: SG-list size. * @mapped_sg_count: ib_dma_map_sg() return value. - * @n_rdma_ius: Number of elements in the rdma_ius array. - * @rdma_ius: Array with information about the RDMA mapping. + * @n_rdma_wrs: Number of elements in the rdma_wrs array. + * @rdma_wrs: Array with information about the RDMA mapping. * @tag: Tag of the received SRP information unit. * @spinlock: Protects 'state'. * @state: I/O context state. - * @rdma_aborted: If initiating a multipart RDMA transfer failed, whether - * the already initiated transfers have finished. * @cmd: Target core command data structure. * @sense_data: SCSI sense data. */ struct srpt_send_ioctx { struct srpt_ioctx ioctx; struct srpt_rdma_ch *ch; - struct rdma_iu *rdma_ius; + struct ib_rdma_wr *rdma_wrs; + struct ib_cqe rdma_cqe; struct srp_direct_buf *rbufs; struct srp_direct_buf single_rbuf; struct scatterlist *sg; struct list_head free_list; spinlock_t spinlock; enum srpt_command_state state; - bool rdma_aborted; struct se_cmd cmd; struct completion tx_done; int sg_cnt; int mapped_sg_count; - u16 n_rdma_ius; + u16 n_rdma_wrs; u8 n_rdma; u8 n_rbuf; bool queue_status_only; @@ -249,27 +218,24 @@ struct srpt_send_ioctx { /** * enum rdma_ch_state - SRP channel state. - * @CH_CONNECTING: QP is in RTR state; waiting for RTU. - * @CH_LIVE: QP is in RTS state. - * @CH_DISCONNECTING: DREQ has been received; waiting for DREP - * or DREQ has been send and waiting for DREP - * or . - * @CH_DRAINING: QP is in ERR state; waiting for last WQE event. - * @CH_RELEASING: Last WQE event has been received; releasing resources. + * @CH_CONNECTING: QP is in RTR state; waiting for RTU. + * @CH_LIVE: QP is in RTS state. + * @CH_DISCONNECTING: DREQ has been sent and waiting for DREP or DREQ has + * been received. + * @CH_DRAINING: DREP has been received or waiting for DREP timed out + * and last work request has been queued. + * @CH_DISCONNECTED: Last completion has been received. */ enum rdma_ch_state { CH_CONNECTING, CH_LIVE, CH_DISCONNECTING, CH_DRAINING, - CH_RELEASING + CH_DISCONNECTED, }; /** * struct srpt_rdma_ch - RDMA channel. - * @wait_queue: Allows the kernel thread to wait for more work. - * @thread: Kernel thread that processes the IB queues associated with - * the channel. * @cm_id: IB CM ID associated with the channel. * @qp: IB queue pair used for communicating over this channel. * @cq: IB completion queue for this channel. @@ -288,7 +254,6 @@ enum rdma_ch_state { * @free_list: Head of list with free send I/O contexts. * @state: channel state. See also enum rdma_ch_state. * @ioctx_ring: Send ring. - * @wc: IB work completion array for srpt_process_completion(). * @list: Node for insertion in the srpt_device.rch_list list. * @cmd_wait_list: List of SCSI commands that arrived before the RTU event. This * list contains struct srpt_ioctx elements and is protected @@ -299,11 +264,11 @@ enum rdma_ch_state { * @release_done: Enables waiting for srpt_release_channel() completion. */ struct srpt_rdma_ch { - wait_queue_head_t wait_queue; - struct task_struct *thread; struct ib_cm_id *cm_id; struct ib_qp *qp; struct ib_cq *cq; + struct ib_cqe zw_cqe; + struct kref kref; int rq_size; u32 rsp_size; atomic_t sq_wr_avail; @@ -317,14 +282,12 @@ struct srpt_rdma_ch { struct list_head free_list; enum rdma_ch_state state; struct srpt_send_ioctx **ioctx_ring; - struct ib_wc wc[16]; struct list_head list; struct list_head cmd_wait_list; struct se_session *sess; u8 sess_name[36]; struct work_struct release_work; struct completion *release_done; - bool in_shutdown; }; /** @@ -377,13 +340,11 @@ struct srpt_port { * @mr: L_Key (local key) with write access to all local memory. * @srq: Per-HCA SRQ (shared receive queue). * @cm_id: Connection identifier. - * @dev_attr: Attributes of the InfiniBand device as obtained during the - * ib_client.add() callback. * @srq_size: SRQ size. * @ioctx_ring: Per-HCA SRQ. * @rch_list: Per-device channel list -- see also srpt_rdma_ch.list. * @ch_releaseQ: Enables waiting for removal from rch_list. - * @spinlock: Protects rch_list and tpg. + * @mutex: Protects rch_list. * @port: Information about the ports owned by this HCA. * @event_handler: Per-HCA asynchronous IB event handler. * @list: Node in srpt_dev_list. @@ -393,23 +354,14 @@ struct srpt_device { struct ib_pd *pd; struct ib_srq *srq; struct ib_cm_id *cm_id; - struct ib_device_attr dev_attr; int srq_size; struct srpt_recv_ioctx **ioctx_ring; struct list_head rch_list; wait_queue_head_t ch_releaseQ; - spinlock_t spinlock; + struct mutex mutex; struct srpt_port port[2]; struct ib_event_handler event_handler; struct list_head list; }; -/** - * struct srpt_node_acl - Per-initiator ACL data (managed via configfs). - * @nacl: Target core node ACL information. - */ -struct srpt_node_acl { - struct se_node_acl nacl; -}; - #endif /* IB_SRPT_H */ |
