From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jack Morgenstein Subject: [PATCH for-next V1 20/29] IB/mlx4: Add CM paravirtualization Date: Tue, 19 Jun 2012 11:21:52 +0300 Message-ID: <1340094121-14858-21-git-send-email-jackm@dev.mellanox.co.il> References: <1340094121-14858-1-git-send-email-jackm@dev.mellanox.co.il> Return-path: In-Reply-To: <1340094121-14858-1-git-send-email-jackm-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org> Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: roland-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, liranl-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org, yevgenyp-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org, Amir Vadai , Jack Morgenstein List-Id: linux-rdma@vger.kernel.org From: Amir Vadai In CM para-virtualization: 1. Incoming requests are steered to the correct vHCA according to the embedded GID 2. Communication IDs on outgoing requests are replaced by a globally unique ID, generated by the PPF, since there is no synchronization of ID generation between guests (and so these IDs are not guaranteed to be globally unique). The guest's comm ID is stored, and is returned to the response MAD when it arrives. Signed-off-by: Amir Vadai Signed-off-by: Jack Morgenstein --- drivers/infiniband/hw/mlx4/Makefile | 2 +- drivers/infiniband/hw/mlx4/cm.c | 437 ++++++++++++++++++++++++++++++++++ drivers/infiniband/hw/mlx4/mad.c | 16 ++- drivers/infiniband/hw/mlx4/mlx4_ib.h | 15 ++ 4 files changed, 468 insertions(+), 2 deletions(-) create mode 100644 drivers/infiniband/hw/mlx4/cm.c diff --git a/drivers/infiniband/hw/mlx4/Makefile b/drivers/infiniband/hw/mlx4/Makefile index 20d627d..bf0aa90 100644 --- a/drivers/infiniband/hw/mlx4/Makefile +++ b/drivers/infiniband/hw/mlx4/Makefile @@ -1,3 +1,3 @@ obj-$(CONFIG_MLX4_INFINIBAND) += mlx4_ib.o -mlx4_ib-y := ah.o cq.o doorbell.o mad.o main.o mr.o qp.o srq.o mcg.o +mlx4_ib-y := ah.o cq.o doorbell.o mad.o main.o mr.o qp.o srq.o mcg.o cm.o diff --git a/drivers/infiniband/hw/mlx4/cm.c b/drivers/infiniband/hw/mlx4/cm.c new file mode 100644 index 0000000..0a52146 --- /dev/null +++ b/drivers/infiniband/hw/mlx4/cm.c @@ -0,0 +1,437 @@ +/* + * Copyright (c) 2012 Mellanox Technologies. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include + +#include +#include +#include +#include + +#include "mlx4_ib.h" + +#define CM_CLEANUP_CACHE_TIMEOUT (5 * HZ) + +struct id_map_entry { + struct rb_node node; + + u32 sl_cm_id; + u32 pv_cm_id; + int slave_id; + int scheduled_delete; + struct mlx4_ib_dev *dev; + + struct list_head list; + struct delayed_work timeout; +}; + +struct cm_generic_msg { + struct ib_mad_hdr hdr; + + __be32 local_comm_id; + __be32 remote_comm_id; +}; + +struct cm_req_msg { + unsigned char unused[0x60]; + union ib_gid primary_path_sgid; +}; + + +static void set_local_comm_id(struct ib_mad *mad, u32 cm_id) +{ + struct cm_generic_msg *msg = (struct cm_generic_msg *)mad; + msg->local_comm_id = cpu_to_be32(cm_id); +} + +static u32 get_local_comm_id(struct ib_mad *mad) +{ + struct cm_generic_msg *msg = (struct cm_generic_msg *)mad; + + return be32_to_cpu(msg->local_comm_id); +} + +static void set_remote_comm_id(struct ib_mad *mad, u32 cm_id) +{ + struct cm_generic_msg *msg = (struct cm_generic_msg *)mad; + msg->remote_comm_id = cpu_to_be32(cm_id); +} + +static u32 get_remote_comm_id(struct ib_mad *mad) +{ + struct cm_generic_msg *msg = (struct cm_generic_msg *)mad; + + return be32_to_cpu(msg->remote_comm_id); +} + +static union ib_gid gid_from_req_msg(struct ib_device *ibdev, struct ib_mad *mad) +{ + struct cm_req_msg *msg = (struct cm_req_msg *)mad; + + return msg->primary_path_sgid; +} + +/* Lock should be taken before called */ +static struct id_map_entry * +id_map_find_by_sl_id(struct ib_device *ibdev, u32 slave_id, u32 sl_cm_id) +{ + struct rb_root *sl_id_map = &to_mdev(ibdev)->sriov.sl_id_map; + struct rb_node *node = sl_id_map->rb_node; + + while (node) { + struct id_map_entry *id_map_entry = + rb_entry(node, struct id_map_entry, node); + + if (id_map_entry->sl_cm_id > sl_cm_id) + node = node->rb_left; + else if (id_map_entry->sl_cm_id < sl_cm_id) + node = node->rb_right; + else if (id_map_entry->slave_id > slave_id) + node = node->rb_left; + else if (id_map_entry->slave_id < slave_id) + node = node->rb_right; + else + return id_map_entry; + } + return NULL; +} + +static void id_map_ent_timeout(struct work_struct *work) +{ + struct delayed_work *delay = to_delayed_work(work); + struct id_map_entry *ent = container_of(delay, struct id_map_entry, timeout); + struct id_map_entry *db_ent, *found_ent; + struct mlx4_ib_dev *dev = ent->dev; + struct mlx4_ib_sriov *sriov = &dev->sriov; + struct rb_root *sl_id_map = &sriov->sl_id_map; + int pv_id = (int) ent->pv_cm_id; + + spin_lock(&sriov->id_map_lock); + db_ent = (struct id_map_entry *)idr_find(&sriov->pv_id_table, pv_id); + if (!db_ent) + goto out; + found_ent = id_map_find_by_sl_id(&dev->ib_dev, ent->slave_id, ent->sl_cm_id); + if (found_ent && found_ent == ent) + rb_erase(&found_ent->node, sl_id_map); + idr_remove(&sriov->pv_id_table, pv_id); + +out: + list_del(&ent->list); + spin_unlock(&sriov->id_map_lock); + kfree(ent); +} + +static void id_map_find_del(struct ib_device *ibdev, int pv_cm_id) +{ + struct mlx4_ib_sriov *sriov = &to_mdev(ibdev)->sriov; + struct rb_root *sl_id_map = &sriov->sl_id_map; + struct id_map_entry *ent, *found_ent; + + spin_lock(&sriov->id_map_lock); + ent = (struct id_map_entry *)idr_find(&sriov->pv_id_table, pv_cm_id); + if (!ent) + goto out; + found_ent = id_map_find_by_sl_id(ibdev, ent->slave_id, ent->sl_cm_id); + if (found_ent && found_ent == ent) + rb_erase(&found_ent->node, sl_id_map); + idr_remove(&sriov->pv_id_table, pv_cm_id); +out: + spin_unlock(&sriov->id_map_lock); +} + +static void sl_id_map_add(struct ib_device *ibdev, struct id_map_entry *new) +{ + struct rb_root *sl_id_map = &to_mdev(ibdev)->sriov.sl_id_map; + struct rb_node **link = &sl_id_map->rb_node, *parent = NULL; + struct id_map_entry *ent; + int slave_id = new->slave_id; + int sl_cm_id = new->sl_cm_id; + + ent = id_map_find_by_sl_id(ibdev, slave_id, sl_cm_id); + if (ent) { + pr_debug("overriding existing sl_id_map entry (cm_id = %x)\n", + sl_cm_id); + + rb_replace_node(&ent->node, &new->node, sl_id_map); + return; + } + + /* Go to the bottom of the tree */ + while (*link) { + parent = *link; + ent = rb_entry(parent, struct id_map_entry, node); + + if (ent->sl_cm_id > sl_cm_id || (ent->sl_cm_id == sl_cm_id && ent->slave_id > slave_id)) + link = &(*link)->rb_left; + else + link = &(*link)->rb_right; + } + + rb_link_node(&new->node, parent, link); + rb_insert_color(&new->node, sl_id_map); +} + +static struct id_map_entry * +id_map_alloc(struct ib_device *ibdev, int slave_id, u32 sl_cm_id) +{ + int ret, id; + static int next_id; + struct id_map_entry *ent; + struct mlx4_ib_sriov *sriov = &to_mdev(ibdev)->sriov; + + ent = kmalloc(sizeof(struct id_map_entry), GFP_KERNEL); + if (!ent) { + mlx4_ib_warn(ibdev, "Couldn't allocate id cache entry - out of memory\n"); + return ERR_PTR(-ENOMEM); + } + + ent->sl_cm_id = sl_cm_id; + ent->slave_id = slave_id; + ent->scheduled_delete = 0; + ent->dev = to_mdev(ibdev); + INIT_DELAYED_WORK(&ent->timeout, id_map_ent_timeout); + + do { + spin_lock(&to_mdev(ibdev)->sriov.id_map_lock); + ret = idr_get_new_above(&sriov->pv_id_table, ent, + next_id, &id); + if (!ret) { + next_id = ((unsigned) id + 1) & MAX_ID_MASK; + ent->pv_cm_id = (u32)id; + sl_id_map_add(ibdev, ent); + } + + spin_unlock(&sriov->id_map_lock); + } while (ret == -EAGAIN && idr_pre_get(&sriov->pv_id_table, GFP_KERNEL)); + /*the function idr_get_new_above can return -ENOSPC, so don't insert in that case.*/ + if (!ret) { + spin_lock(&sriov->id_map_lock); + list_add_tail(&ent->list, &sriov->cm_list); + spin_unlock(&sriov->id_map_lock); + return ent; + } + /*error flow*/ + kfree(ent); + mlx4_ib_warn(ibdev, "No more space in the idr (err:0x%x)\n", ret); + return ERR_PTR(-ENOMEM); +} + +static struct id_map_entry * +id_map_get(struct ib_device *ibdev, int *pv_cm_id, int sl_cm_id, int slave_id) +{ + struct id_map_entry *ent; + struct mlx4_ib_sriov *sriov = &to_mdev(ibdev)->sriov; + + spin_lock(&sriov->id_map_lock); + if (*pv_cm_id == -1) { + ent = id_map_find_by_sl_id(ibdev, sl_cm_id, slave_id); + if (ent) + *pv_cm_id = (int) ent->pv_cm_id; + } else + ent = (struct id_map_entry *)idr_find(&sriov->pv_id_table, *pv_cm_id); + spin_unlock(&sriov->id_map_lock); + + return ent; +} + +static void schedule_delayed(struct ib_device *ibdev, struct id_map_entry *id) +{ + struct mlx4_ib_sriov *sriov = &to_mdev(ibdev)->sriov; + unsigned long flags; + + spin_lock_irqsave(&sriov->going_down_lock, flags); + spin_lock(&sriov->id_map_lock); + /*make sure that there is no schedule inside the scheduled work.*/ + if (!sriov->is_going_down) { + id->scheduled_delete = 1; + schedule_delayed_work(&id->timeout, CM_CLEANUP_CACHE_TIMEOUT); + } + spin_unlock(&sriov->id_map_lock); + spin_unlock_irqrestore(&sriov->going_down_lock, flags); +} + +int mlx4_ib_multiplex_cm_handler(struct ib_device *ibdev, int port, int slave_id, + struct ib_mad *mad) +{ + struct id_map_entry *id; + u32 sl_cm_id; + int pv_cm_id = -1; + + sl_cm_id = get_local_comm_id(mad); + + if (mad->mad_hdr.attr_id == CM_REQ_ATTR_ID || + mad->mad_hdr.attr_id == CM_REP_ATTR_ID) { + id = id_map_alloc(ibdev, slave_id, sl_cm_id); + if (IS_ERR(id)) { + mlx4_ib_warn(ibdev, "%s: id{slave: %d, sl_cm_id: 0x%x} Failed to id_map_alloc\n", + __func__, slave_id, sl_cm_id); + return PTR_ERR(id); + } + } else if (mad->mad_hdr.attr_id == CM_REJ_ATTR_ID) { + return 0; + } else { + id = id_map_get(ibdev, &pv_cm_id, slave_id, sl_cm_id); + } + + if (!id) { + pr_debug("id{slave: %d, sl_cm_id: 0x%x} is NULL!\n", + slave_id, sl_cm_id); + return -EINVAL; + } + + set_local_comm_id(mad, id->pv_cm_id); + + if (mad->mad_hdr.attr_id == CM_DREQ_ATTR_ID) + schedule_delayed(ibdev, id); + else if (mad->mad_hdr.attr_id == CM_DREP_ATTR_ID) + id_map_find_del(ibdev, pv_cm_id); + + return 0; +} + +int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave, + struct ib_mad *mad) +{ + u32 pv_cm_id; + struct id_map_entry *id; + + if (mad->mad_hdr.attr_id == CM_REQ_ATTR_ID) { + union ib_gid gid; + + gid = gid_from_req_msg(ibdev, mad); + *slave = mlx4_ib_find_real_gid(ibdev, port, gid.global.interface_id); + if (*slave < 0) { + mlx4_ib_warn(ibdev, "failed matching slave_id by gid (0x%llx)\n", + gid.global.interface_id); + return -ENOENT; + } + return 0; + } + + pv_cm_id = get_remote_comm_id(mad); + id = id_map_get(ibdev, (int *)&pv_cm_id, -1, -1); + + if (!id) { + pr_debug("Couldn't find an entry for pv_cm_id 0x%x\n", pv_cm_id); + return -ENOENT; + } + + *slave = id->slave_id; + set_remote_comm_id(mad, id->sl_cm_id); + + if (mad->mad_hdr.attr_id == CM_DREQ_ATTR_ID) + schedule_delayed(ibdev, id); + else if (mad->mad_hdr.attr_id == CM_REJ_ATTR_ID || + mad->mad_hdr.attr_id == CM_DREP_ATTR_ID) { + id_map_find_del(ibdev, (int) pv_cm_id); + } + + return 0; +} + +void mlx4_ib_cm_paravirt_init(struct mlx4_ib_dev *dev) +{ + spin_lock_init(&dev->sriov.id_map_lock); + INIT_LIST_HEAD(&dev->sriov.cm_list); + dev->sriov.sl_id_map = RB_ROOT; + idr_init(&dev->sriov.pv_id_table); + idr_pre_get(&dev->sriov.pv_id_table, GFP_KERNEL); +} + +/* slave = -1 ==> all slaves */ +/* TBD -- call paravirt clean for single slave. Need for slave RESET event */ +void mlx4_ib_cm_paravirt_clean(struct mlx4_ib_dev *dev, int slave) +{ + struct mlx4_ib_sriov *sriov = &dev->sriov; + struct rb_root *sl_id_map = &sriov->sl_id_map; + struct list_head lh; + struct rb_node *nd; + int need_flush = 1; + struct id_map_entry *map, *tmp_map; + /* cancel all delayed work queue entries */ + INIT_LIST_HEAD(&lh); + spin_lock(&sriov->id_map_lock); + list_for_each_entry_safe(map, tmp_map, &dev->sriov.cm_list, list) { + if (slave < 0 || slave == map->slave_id) { + if (map->scheduled_delete) + need_flush &= !!cancel_delayed_work(&map->timeout); + } + } + + spin_unlock(&sriov->id_map_lock); + + if (!need_flush) + flush_scheduled_work(); /* make sure all timers were flushed */ + + /* now, remove all leftover entries from databases*/ + spin_lock(&sriov->id_map_lock); + if (slave < 0) { + while (rb_first(sl_id_map)) { + struct id_map_entry *ent = + rb_entry(rb_first(sl_id_map), + struct id_map_entry, node); + + rb_erase(&ent->node, sl_id_map); + idr_remove(&sriov->pv_id_table, (int) ent->pv_cm_id); + } + list_splice_init(&dev->sriov.cm_list, &lh); + } else { + /* first, move nodes belonging to slave to db remove list */ + nd = rb_first(sl_id_map); + while (nd) { + struct id_map_entry *ent = + rb_entry(nd, struct id_map_entry, node); + nd = rb_next(nd); + if (ent->slave_id == slave) + list_move_tail(&ent->list, &lh); + } + /* remove those nodes from databases */ + list_for_each_entry_safe(map, tmp_map, &lh, list) { + rb_erase(&map->node, sl_id_map); + idr_remove(&sriov->pv_id_table, (int) map->pv_cm_id); + } + + /* add remaining nodes from cm_list */ + list_for_each_entry_safe(map, tmp_map, &dev->sriov.cm_list, list) { + if (slave == map->slave_id) + list_move_tail(&map->list, &lh); + } + } + + spin_unlock(&sriov->id_map_lock); + + /* free any map entries left behind due to cancel_delayed_work above */ + list_for_each_entry_safe(map, tmp_map, &lh, list) { + list_del(&map->list); + kfree(map); + } +} diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c index 513bb0d..03eeafa 100644 --- a/drivers/infiniband/hw/mlx4/mad.c +++ b/drivers/infiniband/hw/mlx4/mad.c @@ -541,6 +541,10 @@ static int mlx4_ib_demux_mad(struct ib_device *ibdev, u8 port, (struct ib_sa_mad *) mad)) return 0; break; + case IB_MGMT_CLASS_CM: + if (mlx4_ib_demux_cm_handler(ibdev, port, &slave, mad)) + return 0; + break; case IB_MGMT_CLASS_DEVICE_MGMT: if (mad->mad_hdr.method != IB_MGMT_METHOD_GET_RESP) return 0; @@ -1073,6 +1077,11 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc (struct ib_sa_mad *) &tunnel->mad)) return; break; + case IB_MGMT_CLASS_CM: + if (mlx4_ib_multiplex_cm_handler(ctx->ib_dev, ctx->port, slave, + (struct ib_mad *) &tunnel->mad)) + return; + break; case IB_MGMT_CLASS_DEVICE_MGMT: if (tunnel->mad.mad_hdr.method != IB_MGMT_METHOD_GET && tunnel->mad.mad_hdr.method != IB_MGMT_METHOD_SET) @@ -1787,6 +1796,7 @@ int mlx4_ib_init_sriov(struct mlx4_ib_dev *dev) dev->sriov.is_going_down = 0; spin_lock_init(&dev->sriov.going_down_lock); + mlx4_ib_cm_paravirt_init(dev); mlx4_ib_warn(&dev->ib_dev, "multi-function enabled\n"); @@ -1815,6 +1825,7 @@ demux_err: mlx4_ib_free_demux_ctx(&dev->sriov.demux[i]); --i; } + mlx4_ib_cm_paravirt_clean(dev, -1); return err; } @@ -1830,7 +1841,7 @@ void mlx4_ib_close_sriov(struct mlx4_ib_dev *dev) spin_lock_irqsave(&dev->sriov.going_down_lock, flags); dev->sriov.is_going_down = 1; spin_unlock_irqrestore(&dev->sriov.going_down_lock, flags); - if (mlx4_is_master(dev->dev)) + if (mlx4_is_master(dev->dev)) { for (i = 0; i < dev->num_ports; i++) { flush_workqueue(dev->sriov.demux[i].ud_wq); mlx4_ib_free_sqp_ctx(dev->sriov.sqps[i]); @@ -1838,4 +1849,7 @@ void mlx4_ib_close_sriov(struct mlx4_ib_dev *dev) dev->sriov.sqps[i] = NULL; mlx4_ib_free_demux_ctx(&dev->sriov.demux[i]); } + + mlx4_ib_cm_paravirt_clean(dev, -1); + } } diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h index ab1d7f9..6d297b2 100644 --- a/drivers/infiniband/hw/mlx4/mlx4_ib.h +++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h @@ -347,6 +347,12 @@ struct mlx4_ib_sriov { * it may be called from interrupt context.*/ spinlock_t going_down_lock; int is_going_down; + + /* CM paravirtualization fields */ + struct list_head cm_list; + spinlock_t id_map_lock; + struct rb_root sl_id_map; + struct idr pv_id_table; }; struct mlx4_ib_iboe { @@ -590,4 +596,13 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port, u32 qkey, struct ib_ah_attr *attr, struct ib_mad *mad); __be64 mlx4_ib_get_new_demux_tid(struct mlx4_ib_demux_ctx *ctx); +int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave, + struct ib_mad *mad); + +int mlx4_ib_multiplex_cm_handler(struct ib_device *ibdev, int port, int slave_id, + struct ib_mad *mad); + +void mlx4_ib_cm_paravirt_init(struct mlx4_ib_dev *dev); +void mlx4_ib_cm_paravirt_clean(struct mlx4_ib_dev *dev, int slave_id); + #endif /* MLX4_IB_H */ -- 1.7.1 -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html