diff options
Diffstat (limited to 'drivers/infiniband/core/cm.c')
| -rw-r--r-- | drivers/infiniband/core/cm.c | 1403 |
1 files changed, 1020 insertions, 383 deletions
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index 2514de3480d..c3239170d8b 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2004, 2005 Intel Corporation. All rights reserved. + * Copyright (c) 2004-2007 Intel Corporation. All rights reserved. * Copyright (c) 2004 Topspin Corporation. All rights reserved. * Copyright (c) 2004, 2005 Voltaire Corporation. All rights reserved. * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. @@ -31,17 +31,23 @@ * 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. - * - * $Id: cm.c 2821 2005-07-08 17:07:28Z sean.hefty $ */ + +#include <linux/completion.h> #include <linux/dma-mapping.h> +#include <linux/device.h> +#include <linux/module.h> #include <linux/err.h> #include <linux/idr.h> #include <linux/interrupt.h> -#include <linux/pci.h> +#include <linux/random.h> #include <linux/rbtree.h> #include <linux/spinlock.h> +#include <linux/slab.h> +#include <linux/sysfs.h> #include <linux/workqueue.h> +#include <linux/kdev_t.h> +#include <linux/etherdevice.h> #include <rdma/ib_cache.h> #include <rdma/ib_cm.h> @@ -71,20 +77,99 @@ static struct ib_cm { struct rb_root remote_id_table; struct rb_root remote_sidr_table; struct idr local_id_table; + __be32 random_id_operand; + struct list_head timewait_list; struct workqueue_struct *wq; } cm; +/* Counter indexes ordered by attribute ID */ +enum { + CM_REQ_COUNTER, + CM_MRA_COUNTER, + CM_REJ_COUNTER, + CM_REP_COUNTER, + CM_RTU_COUNTER, + CM_DREQ_COUNTER, + CM_DREP_COUNTER, + CM_SIDR_REQ_COUNTER, + CM_SIDR_REP_COUNTER, + CM_LAP_COUNTER, + CM_APR_COUNTER, + CM_ATTR_COUNT, + CM_ATTR_ID_OFFSET = 0x0010, +}; + +enum { + CM_XMIT, + CM_XMIT_RETRIES, + CM_RECV, + CM_RECV_DUPLICATES, + CM_COUNTER_GROUPS +}; + +static char const counter_group_names[CM_COUNTER_GROUPS] + [sizeof("cm_rx_duplicates")] = { + "cm_tx_msgs", "cm_tx_retries", + "cm_rx_msgs", "cm_rx_duplicates" +}; + +struct cm_counter_group { + struct kobject obj; + atomic_long_t counter[CM_ATTR_COUNT]; +}; + +struct cm_counter_attribute { + struct attribute attr; + int index; +}; + +#define CM_COUNTER_ATTR(_name, _index) \ +struct cm_counter_attribute cm_##_name##_counter_attr = { \ + .attr = { .name = __stringify(_name), .mode = 0444 }, \ + .index = _index \ +} + +static CM_COUNTER_ATTR(req, CM_REQ_COUNTER); +static CM_COUNTER_ATTR(mra, CM_MRA_COUNTER); +static CM_COUNTER_ATTR(rej, CM_REJ_COUNTER); +static CM_COUNTER_ATTR(rep, CM_REP_COUNTER); +static CM_COUNTER_ATTR(rtu, CM_RTU_COUNTER); +static CM_COUNTER_ATTR(dreq, CM_DREQ_COUNTER); +static CM_COUNTER_ATTR(drep, CM_DREP_COUNTER); +static CM_COUNTER_ATTR(sidr_req, CM_SIDR_REQ_COUNTER); +static CM_COUNTER_ATTR(sidr_rep, CM_SIDR_REP_COUNTER); +static CM_COUNTER_ATTR(lap, CM_LAP_COUNTER); +static CM_COUNTER_ATTR(apr, CM_APR_COUNTER); + +static struct attribute *cm_counter_default_attrs[] = { + &cm_req_counter_attr.attr, + &cm_mra_counter_attr.attr, + &cm_rej_counter_attr.attr, + &cm_rep_counter_attr.attr, + &cm_rtu_counter_attr.attr, + &cm_dreq_counter_attr.attr, + &cm_drep_counter_attr.attr, + &cm_sidr_req_counter_attr.attr, + &cm_sidr_rep_counter_attr.attr, + &cm_lap_counter_attr.attr, + &cm_apr_counter_attr.attr, + NULL +}; + struct cm_port { struct cm_device *cm_dev; struct ib_mad_agent *mad_agent; + struct kobject port_obj; u8 port_num; + struct cm_counter_group counter_group[CM_COUNTER_GROUPS]; }; struct cm_device { struct list_head list; - struct ib_device *device; - __be64 ca_guid; - struct cm_port port[0]; + struct ib_device *ib_device; + struct device *device; + u8 ack_delay; + struct cm_port *port[0]; }; struct cm_av { @@ -92,11 +177,13 @@ struct cm_av { union ib_gid dgid; struct ib_ah_attr ah_attr; u16 pkey_index; - u8 packet_life_time; + u8 timeout; + u8 valid; + u8 smac[ETH_ALEN]; }; struct cm_work { - struct work_struct work; + struct delayed_work work; struct list_head list; struct cm_port *port; struct ib_mad_recv_wc *mad_recv_wc; /* Received MADs */ @@ -108,6 +195,7 @@ struct cm_work { struct cm_timewait_info { struct cm_work work; /* Must be first. */ + struct list_head list; struct rb_node remote_qp_node; struct rb_node remote_id_node; __be64 remote_ca_guid; @@ -121,8 +209,8 @@ struct cm_id_private { struct rb_node service_node; struct rb_node sidr_id_node; - spinlock_t lock; - wait_queue_head_t wait; + spinlock_t lock; /* Do not acquire inside cm.lock */ + struct completion comp; atomic_t refcount; struct ib_mad_send_buf *msg; @@ -130,6 +218,7 @@ struct cm_id_private { /* todo: use alternate port on send failure */ struct cm_av av; struct cm_av alt_av; + struct ib_cm_compare_data *compare_data; void *private_data; __be64 tid; @@ -140,26 +229,27 @@ struct cm_id_private { __be32 rq_psn; int timeout_ms; enum ib_mtu path_mtu; + __be16 pkey; u8 private_data_len; u8 max_cm_retries; u8 peer_to_peer; u8 responder_resources; u8 initiator_depth; - u8 local_ack_timeout; u8 retry_count; u8 rnr_retry_count; u8 service_timeout; + u8 target_ack_delay; struct list_head work_list; atomic_t work_count; }; -static void cm_work_handler(void *data); +static void cm_work_handler(struct work_struct *work); static inline void cm_deref_id(struct cm_id_private *cm_id_priv) { if (atomic_dec_and_test(&cm_id_priv->refcount)) - wake_up(&cm_id_priv->wait); + complete(&cm_id_priv->comp); } static int cm_alloc_msg(struct cm_id_private *cm_id_priv, @@ -174,7 +264,7 @@ static int cm_alloc_msg(struct cm_id_private *cm_id_priv, if (IS_ERR(ah)) return PTR_ERR(ah); - m = ib_create_send_mad(mad_agent, cm_id_priv->id.remote_cm_qpn, + m = ib_create_send_mad(mad_agent, cm_id_priv->id.remote_cm_qpn, cm_id_priv->av.pkey_index, 0, IB_MGMT_MAD_HDR, IB_MGMT_MAD_DATA, GFP_ATOMIC); @@ -233,11 +323,10 @@ static void * cm_copy_private_data(const void *private_data, if (!private_data || !private_data_len) return NULL; - data = kmalloc(private_data_len, GFP_KERNEL); + data = kmemdup(private_data, private_data_len, GFP_KERNEL); if (!data) return ERR_PTR(-ENOMEM); - memcpy(data, private_data, private_data_len); return data; } @@ -251,23 +340,13 @@ static void cm_set_private_data(struct cm_id_private *cm_id_priv, cm_id_priv->private_data_len = private_data_len; } -static void cm_set_ah_attr(struct ib_ah_attr *ah_attr, u8 port_num, - u16 dlid, u8 sl, u16 src_path_bits) -{ - memset(ah_attr, 0, sizeof ah_attr); - ah_attr->dlid = dlid; - ah_attr->sl = sl; - ah_attr->src_path_bits = src_path_bits; - ah_attr->port_num = port_num; -} - -static void cm_init_av_for_response(struct cm_port *port, - struct ib_wc *wc, struct cm_av *av) +static void cm_init_av_for_response(struct cm_port *port, struct ib_wc *wc, + struct ib_grh *grh, struct cm_av *av) { av->port = port; av->pkey_index = wc->pkey_index; - cm_set_ah_attr(&av->ah_attr, port->port_num, wc->slid, - wc->sl, wc->dlid_path_bits); + ib_init_ah_from_wc(port->cm_dev->ib_device, port->port_num, wc, + grh, &av->ah_attr); } static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av) @@ -280,9 +359,9 @@ static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av) read_lock_irqsave(&cm.device_lock, flags); list_for_each_entry(cm_dev, &cm.device_list, list) { - if (!ib_find_cached_gid(cm_dev->device, &path->sgid, + if (!ib_find_cached_gid(cm_dev->ib_device, &path->sgid, &p, NULL)) { - port = &cm_dev->port[p-1]; + port = cm_dev->port[p-1]; break; } } @@ -291,48 +370,52 @@ static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av) if (!port) return -EINVAL; - ret = ib_find_cached_pkey(cm_dev->device, port->port_num, + ret = ib_find_cached_pkey(cm_dev->ib_device, port->port_num, be16_to_cpu(path->pkey), &av->pkey_index); if (ret) return ret; av->port = port; - cm_set_ah_attr(&av->ah_attr, av->port->port_num, - be16_to_cpu(path->dlid), path->sl, - be16_to_cpu(path->slid) & 0x7F); - av->packet_life_time = path->packet_life_time; + ib_init_ah_from_path(cm_dev->ib_device, port->port_num, path, + &av->ah_attr); + av->timeout = path->packet_life_time + 1; + memcpy(av->smac, path->smac, sizeof(av->smac)); + + av->valid = 1; return 0; } static int cm_alloc_id(struct cm_id_private *cm_id_priv) { unsigned long flags; - int ret; - static int next_id; + int id; - do { - spin_lock_irqsave(&cm.lock, flags); - ret = idr_get_new_above(&cm.local_id_table, cm_id_priv, next_id++, - (__force int *) &cm_id_priv->id.local_id); - spin_unlock_irqrestore(&cm.lock, flags); - } while( (ret == -EAGAIN) && idr_pre_get(&cm.local_id_table, GFP_KERNEL) ); - return ret; + idr_preload(GFP_KERNEL); + spin_lock_irqsave(&cm.lock, flags); + + id = idr_alloc_cyclic(&cm.local_id_table, cm_id_priv, 0, 0, GFP_NOWAIT); + + spin_unlock_irqrestore(&cm.lock, flags); + idr_preload_end(); + + cm_id_priv->id.local_id = (__force __be32)id ^ cm.random_id_operand; + return id < 0 ? id : 0; } static void cm_free_id(__be32 local_id) { - unsigned long flags; - - spin_lock_irqsave(&cm.lock, flags); - idr_remove(&cm.local_id_table, (__force int) local_id); - spin_unlock_irqrestore(&cm.lock, flags); + spin_lock_irq(&cm.lock); + idr_remove(&cm.local_id_table, + (__force int) (local_id ^ cm.random_id_operand)); + spin_unlock_irq(&cm.lock); } static struct cm_id_private * cm_get_id(__be32 local_id, __be32 remote_id) { struct cm_id_private *cm_id_priv; - cm_id_priv = idr_find(&cm.local_id_table, (__force int) local_id); + cm_id_priv = idr_find(&cm.local_id_table, + (__force int) (local_id ^ cm.random_id_operand)); if (cm_id_priv) { if (cm_id_priv->id.remote_id == remote_id) atomic_inc(&cm_id_priv->refcount); @@ -346,15 +429,74 @@ static struct cm_id_private * cm_get_id(__be32 local_id, __be32 remote_id) static struct cm_id_private * cm_acquire_id(__be32 local_id, __be32 remote_id) { struct cm_id_private *cm_id_priv; - unsigned long flags; - spin_lock_irqsave(&cm.lock, flags); + spin_lock_irq(&cm.lock); cm_id_priv = cm_get_id(local_id, remote_id); - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock_irq(&cm.lock); return cm_id_priv; } +static void cm_mask_copy(u8 *dst, u8 *src, u8 *mask) +{ + int i; + + for (i = 0; i < IB_CM_COMPARE_SIZE / sizeof(unsigned long); i++) + ((unsigned long *) dst)[i] = ((unsigned long *) src)[i] & + ((unsigned long *) mask)[i]; +} + +static int cm_compare_data(struct ib_cm_compare_data *src_data, + struct ib_cm_compare_data *dst_data) +{ + u8 src[IB_CM_COMPARE_SIZE]; + u8 dst[IB_CM_COMPARE_SIZE]; + + if (!src_data || !dst_data) + return 0; + + cm_mask_copy(src, src_data->data, dst_data->mask); + cm_mask_copy(dst, dst_data->data, src_data->mask); + return memcmp(src, dst, IB_CM_COMPARE_SIZE); +} + +static int cm_compare_private_data(u8 *private_data, + struct ib_cm_compare_data *dst_data) +{ + u8 src[IB_CM_COMPARE_SIZE]; + + if (!dst_data) + return 0; + + cm_mask_copy(src, private_data, dst_data->mask); + return memcmp(src, dst_data->data, IB_CM_COMPARE_SIZE); +} + +/* + * Trivial helpers to strip endian annotation and compare; the + * endianness doesn't actually matter since we just need a stable + * order for the RB tree. + */ +static int be32_lt(__be32 a, __be32 b) +{ + return (__force u32) a < (__force u32) b; +} + +static int be32_gt(__be32 a, __be32 b) +{ + return (__force u32) a > (__force u32) b; +} + +static int be64_lt(__be64 a, __be64 b) +{ + return (__force u64) a < (__force u64) b; +} + +static int be64_gt(__be64 a, __be64 b) +{ + return (__force u64) a > (__force u64) b; +} + static struct cm_id_private * cm_insert_listen(struct cm_id_private *cm_id_priv) { struct rb_node **link = &cm.listen_service_table.rb_node; @@ -362,21 +504,29 @@ static struct cm_id_private * cm_insert_listen(struct cm_id_private *cm_id_priv) struct cm_id_private *cur_cm_id_priv; __be64 service_id = cm_id_priv->id.service_id; __be64 service_mask = cm_id_priv->id.service_mask; + int data_cmp; while (*link) { parent = *link; cur_cm_id_priv = rb_entry(parent, struct cm_id_private, service_node); + data_cmp = cm_compare_data(cm_id_priv->compare_data, + cur_cm_id_priv->compare_data); if ((cur_cm_id_priv->id.service_mask & service_id) == (service_mask & cur_cm_id_priv->id.service_id) && - (cm_id_priv->id.device == cur_cm_id_priv->id.device)) + (cm_id_priv->id.device == cur_cm_id_priv->id.device) && + !data_cmp) return cur_cm_id_priv; if (cm_id_priv->id.device < cur_cm_id_priv->id.device) link = &(*link)->rb_left; else if (cm_id_priv->id.device > cur_cm_id_priv->id.device) link = &(*link)->rb_right; - else if (service_id < cur_cm_id_priv->id.service_id) + else if (be64_lt(service_id, cur_cm_id_priv->id.service_id)) + link = &(*link)->rb_left; + else if (be64_gt(service_id, cur_cm_id_priv->id.service_id)) + link = &(*link)->rb_right; + else if (data_cmp < 0) link = &(*link)->rb_left; else link = &(*link)->rb_right; @@ -387,23 +537,31 @@ static struct cm_id_private * cm_insert_listen(struct cm_id_private *cm_id_priv) } static struct cm_id_private * cm_find_listen(struct ib_device *device, - __be64 service_id) + __be64 service_id, + u8 *private_data) { struct rb_node *node = cm.listen_service_table.rb_node; struct cm_id_private *cm_id_priv; + int data_cmp; while (node) { cm_id_priv = rb_entry(node, struct cm_id_private, service_node); + data_cmp = cm_compare_private_data(private_data, + cm_id_priv->compare_data); if ((cm_id_priv->id.service_mask & service_id) == cm_id_priv->id.service_id && - (cm_id_priv->id.device == device)) + (cm_id_priv->id.device == device) && !data_cmp) return cm_id_priv; if (device < cm_id_priv->id.device) node = node->rb_left; else if (device > cm_id_priv->id.device) node = node->rb_right; - else if (service_id < cm_id_priv->id.service_id) + else if (be64_lt(service_id, cm_id_priv->id.service_id)) + node = node->rb_left; + else if (be64_gt(service_id, cm_id_priv->id.service_id)) + node = node->rb_right; + else if (data_cmp < 0) node = node->rb_left; else node = node->rb_right; @@ -424,13 +582,13 @@ static struct cm_timewait_info * cm_insert_remote_id(struct cm_timewait_info parent = *link; cur_timewait_info = rb_entry(parent, struct cm_timewait_info, remote_id_node); - if (remote_id < cur_timewait_info->work.remote_id) + if (be32_lt(remote_id, cur_timewait_info->work.remote_id)) link = &(*link)->rb_left; - else if (remote_id > cur_timewait_info->work.remote_id) + else if (be32_gt(remote_id, cur_timewait_info->work.remote_id)) link = &(*link)->rb_right; - else if (remote_ca_guid < cur_timewait_info->remote_ca_guid) + else if (be64_lt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) link = &(*link)->rb_left; - else if (remote_ca_guid > cur_timewait_info->remote_ca_guid) + else if (be64_gt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) link = &(*link)->rb_right; else return cur_timewait_info; @@ -450,13 +608,13 @@ static struct cm_timewait_info * cm_find_remote_id(__be64 remote_ca_guid, while (node) { timewait_info = rb_entry(node, struct cm_timewait_info, remote_id_node); - if (remote_id < timewait_info->work.remote_id) + if (be32_lt(remote_id, timewait_info->work.remote_id)) node = node->rb_left; - else if (remote_id > timewait_info->work.remote_id) + else if (be32_gt(remote_id, timewait_info->work.remote_id)) node = node->rb_right; - else if (remote_ca_guid < timewait_info->remote_ca_guid) + else if (be64_lt(remote_ca_guid, timewait_info->remote_ca_guid)) node = node->rb_left; - else if (remote_ca_guid > timewait_info->remote_ca_guid) + else if (be64_gt(remote_ca_guid, timewait_info->remote_ca_guid)) node = node->rb_right; else return timewait_info; @@ -477,13 +635,13 @@ static struct cm_timewait_info * cm_insert_remote_qpn(struct cm_timewait_info parent = *link; cur_timewait_info = rb_entry(parent, struct cm_timewait_info, remote_qp_node); - if (remote_qpn < cur_timewait_info->remote_qpn) + if (be32_lt(remote_qpn, cur_timewait_info->remote_qpn)) link = &(*link)->rb_left; - else if (remote_qpn > cur_timewait_info->remote_qpn) + else if (be32_gt(remote_qpn, cur_timewait_info->remote_qpn)) link = &(*link)->rb_right; - else if (remote_ca_guid < cur_timewait_info->remote_ca_guid) + else if (be64_lt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) link = &(*link)->rb_left; - else if (remote_ca_guid > cur_timewait_info->remote_ca_guid) + else if (be64_gt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) link = &(*link)->rb_right; else return cur_timewait_info; @@ -507,9 +665,9 @@ static struct cm_id_private * cm_insert_remote_sidr(struct cm_id_private parent = *link; cur_cm_id_priv = rb_entry(parent, struct cm_id_private, sidr_id_node); - if (remote_id < cur_cm_id_priv->id.remote_id) + if (be32_lt(remote_id, cur_cm_id_priv->id.remote_id)) link = &(*link)->rb_left; - else if (remote_id > cur_cm_id_priv->id.remote_id) + else if (be32_gt(remote_id, cur_cm_id_priv->id.remote_id)) link = &(*link)->rb_right; else { int cmp; @@ -559,7 +717,7 @@ struct ib_cm_id *ib_create_cm_id(struct ib_device *device, goto error; spin_lock_init(&cm_id_priv->lock); - init_waitqueue_head(&cm_id_priv->wait); + init_completion(&cm_id_priv->comp); INIT_LIST_HEAD(&cm_id_priv->work_list); atomic_set(&cm_id_priv->work_count, -1); atomic_set(&cm_id_priv->refcount, 1); @@ -596,15 +754,27 @@ static inline int cm_convert_to_ms(int iba_time) return 1 << max(iba_time - 8, 0); } -static void cm_cleanup_timewait(struct cm_timewait_info *timewait_info) +/* + * calculate: 4.096x2^ack_timeout = 4.096x2^ack_delay + 2x4.096x2^life_time + * Because of how ack_timeout is stored, adding one doubles the timeout. + * To avoid large timeouts, select the max(ack_delay, life_time + 1), and + * increment it (round up) only if the other is within 50%. + */ +static u8 cm_ack_timeout(u8 ca_ack_delay, u8 packet_life_time) { - unsigned long flags; + int ack_timeout = packet_life_time + 1; - if (!timewait_info->inserted_remote_id && - !timewait_info->inserted_remote_qp) - return; + if (ack_timeout >= ca_ack_delay) + ack_timeout += (ca_ack_delay >= (ack_timeout - 1)); + else + ack_timeout = ca_ack_delay + + (ack_timeout >= (ca_ack_delay - 1)); - spin_lock_irqsave(&cm.lock, flags); + return min(31, ack_timeout); +} + +static void cm_cleanup_timewait(struct cm_timewait_info *timewait_info) +{ if (timewait_info->inserted_remote_id) { rb_erase(&timewait_info->remote_id_node, &cm.remote_id_table); timewait_info->inserted_remote_id = 0; @@ -614,7 +784,6 @@ static void cm_cleanup_timewait(struct cm_timewait_info *timewait_info) rb_erase(&timewait_info->remote_qp_node, &cm.remote_qp_table); timewait_info->inserted_remote_qp = 0; } - spin_unlock_irqrestore(&cm.lock, flags); } static struct cm_timewait_info * cm_create_timewait_info(__be32 local_id) @@ -626,8 +795,7 @@ static struct cm_timewait_info * cm_create_timewait_info(__be32 local_id) return ERR_PTR(-ENOMEM); timewait_info->work.local_id = local_id; - INIT_WORK(&timewait_info->work.work, cm_work_handler, - &timewait_info->work); + INIT_DELAYED_WORK(&timewait_info->work.work, cm_work_handler); timewait_info->work.cm_event.event = IB_CM_TIMEWAIT_EXIT; return timewait_info; } @@ -635,6 +803,12 @@ static struct cm_timewait_info * cm_create_timewait_info(__be32 local_id) static void cm_enter_timewait(struct cm_id_private *cm_id_priv) { int wait_time; + unsigned long flags; + + spin_lock_irqsave(&cm.lock, flags); + cm_cleanup_timewait(cm_id_priv->timewait_info); + list_add_tail(&cm_id_priv->timewait_info->list, &cm.timewait_list); + spin_unlock_irqrestore(&cm.lock, flags); /* * The cm_id could be destroyed by the user before we exit timewait. @@ -642,7 +816,7 @@ static void cm_enter_timewait(struct cm_id_private *cm_id_priv) * timewait before notifying the user that we've exited timewait. */ cm_id_priv->id.state = IB_CM_TIMEWAIT; - wait_time = cm_convert_to_ms(cm_id_priv->local_ack_timeout); + wait_time = cm_convert_to_ms(cm_id_priv->av.timeout); queue_delayed_work(cm.wq, &cm_id_priv->timewait_info->work.work, msecs_to_jiffies(wait_time)); cm_id_priv->timewait_info = NULL; @@ -650,114 +824,144 @@ static void cm_enter_timewait(struct cm_id_private *cm_id_priv) static void cm_reset_to_idle(struct cm_id_private *cm_id_priv) { + unsigned long flags; + cm_id_priv->id.state = IB_CM_IDLE; if (cm_id_priv->timewait_info) { + spin_lock_irqsave(&cm.lock, flags); cm_cleanup_timewait(cm_id_priv->timewait_info); + spin_unlock_irqrestore(&cm.lock, flags); kfree(cm_id_priv->timewait_info); cm_id_priv->timewait_info = NULL; } } -void ib_destroy_cm_id(struct ib_cm_id *cm_id) +static void cm_destroy_id(struct ib_cm_id *cm_id, int err) { struct cm_id_private *cm_id_priv; struct cm_work *work; - unsigned long flags; cm_id_priv = container_of(cm_id, struct cm_id_private, id); retest: - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); switch (cm_id->state) { case IB_CM_LISTEN: cm_id->state = IB_CM_IDLE; - spin_unlock_irqrestore(&cm_id_priv->lock, flags); - spin_lock_irqsave(&cm.lock, flags); + spin_unlock_irq(&cm_id_priv->lock); + spin_lock_irq(&cm.lock); rb_erase(&cm_id_priv->service_node, &cm.listen_service_table); - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock_irq(&cm.lock); break; case IB_CM_SIDR_REQ_SENT: cm_id->state = IB_CM_IDLE; ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); break; case IB_CM_SIDR_REQ_RCVD: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); cm_reject_sidr_req(cm_id_priv, IB_SIDR_REJECT); break; case IB_CM_REQ_SENT: ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); ib_send_cm_rej(cm_id, IB_CM_REJ_TIMEOUT, - &cm_id_priv->av.port->cm_dev->ca_guid, - sizeof cm_id_priv->av.port->cm_dev->ca_guid, + &cm_id_priv->id.device->node_guid, + sizeof cm_id_priv->id.device->node_guid, NULL, 0); break; + case IB_CM_REQ_RCVD: + if (err == -ENOMEM) { + /* Do not reject to allow future retries. */ + cm_reset_to_idle(cm_id_priv); + spin_unlock_irq(&cm_id_priv->lock); + } else { + spin_unlock_irq(&cm_id_priv->lock); + ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED, + NULL, 0, NULL, 0); + } + break; case IB_CM_MRA_REQ_RCVD: case IB_CM_REP_SENT: case IB_CM_MRA_REP_RCVD: ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); /* Fall through */ - case IB_CM_REQ_RCVD: case IB_CM_MRA_REQ_SENT: case IB_CM_REP_RCVD: case IB_CM_MRA_REP_SENT: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED, NULL, 0, NULL, 0); break; case IB_CM_ESTABLISHED: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); + if (cm_id_priv->qp_type == IB_QPT_XRC_TGT) + break; ib_send_cm_dreq(cm_id, NULL, 0); goto retest; case IB_CM_DREQ_SENT: ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); cm_enter_timewait(cm_id_priv); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); break; case IB_CM_DREQ_RCVD: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); ib_send_cm_drep(cm_id, NULL, 0); break; default: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); break; } cm_free_id(cm_id->local_id); - atomic_dec(&cm_id_priv->refcount); - wait_event(cm_id_priv->wait, !atomic_read(&cm_id_priv->refcount)); + cm_deref_id(cm_id_priv); + wait_for_completion(&cm_id_priv->comp); while ((work = cm_dequeue_work(cm_id_priv)) != NULL) cm_free_work(work); - if (cm_id_priv->private_data && cm_id_priv->private_data_len) - kfree(cm_id_priv->private_data); + kfree(cm_id_priv->compare_data); + kfree(cm_id_priv->private_data); kfree(cm_id_priv); } + +void ib_destroy_cm_id(struct ib_cm_id *cm_id) +{ + cm_destroy_id(cm_id, 0); +} EXPORT_SYMBOL(ib_destroy_cm_id); -int ib_cm_listen(struct ib_cm_id *cm_id, - __be64 service_id, - __be64 service_mask) +int ib_cm_listen(struct ib_cm_id *cm_id, __be64 service_id, __be64 service_mask, + struct ib_cm_compare_data *compare_data) { struct cm_id_private *cm_id_priv, *cur_cm_id_priv; unsigned long flags; int ret = 0; - service_mask = service_mask ? service_mask : - __constant_cpu_to_be64(~0ULL); + service_mask = service_mask ? service_mask : ~cpu_to_be64(0); service_id &= service_mask; if ((service_id & IB_SERVICE_ID_AGN_MASK) == IB_CM_ASSIGN_SERVICE_ID && (service_id != IB_CM_ASSIGN_SERVICE_ID)) return -EINVAL; cm_id_priv = container_of(cm_id, struct cm_id_private, id); - BUG_ON(cm_id->state != IB_CM_IDLE); + if (cm_id->state != IB_CM_IDLE) + return -EINVAL; + + if (compare_data) { + cm_id_priv->compare_data = kzalloc(sizeof *compare_data, + GFP_KERNEL); + if (!cm_id_priv->compare_data) + return -ENOMEM; + cm_mask_copy(cm_id_priv->compare_data->data, + compare_data->data, compare_data->mask); + memcpy(cm_id_priv->compare_data->mask, compare_data->mask, + IB_CM_COMPARE_SIZE); + } cm_id->state = IB_CM_LISTEN; spin_lock_irqsave(&cm.lock, flags); if (service_id == IB_CM_ASSIGN_SERVICE_ID) { cm_id->service_id = cpu_to_be64(cm.listen_service_id++); - cm_id->service_mask = __constant_cpu_to_be64(~0ULL); + cm_id->service_mask = ~cpu_to_be64(0); } else { cm_id->service_id = service_id; cm_id->service_mask = service_mask; @@ -767,6 +971,8 @@ int ib_cm_listen(struct ib_cm_id *cm_id, if (cur_cm_id_priv) { cm_id->state = IB_CM_IDLE; + kfree(cm_id_priv->compare_data); + cm_id_priv->compare_data = NULL; ret = -EBUSY; } return ret; @@ -799,14 +1005,16 @@ static void cm_format_req(struct cm_req_msg *req_msg, struct cm_id_private *cm_id_priv, struct ib_cm_req_param *param) { + struct ib_sa_path_rec *pri_path = param->primary_path; + struct ib_sa_path_rec *alt_path = param->alternate_path; + cm_format_mad_hdr(&req_msg->hdr, CM_REQ_ATTR_ID, cm_form_tid(cm_id_priv, CM_MSG_SEQUENCE_REQ)); req_msg->local_comm_id = cm_id_priv->id.local_id; req_msg->service_id = param->service_id; - req_msg->local_ca_guid = cm_id_priv->av.port->cm_dev->ca_guid; + req_msg->local_ca_guid = cm_id_priv->id.device->node_guid; cm_req_set_local_qpn(req_msg, cpu_to_be32(param->qp_num)); - cm_req_set_resp_res(req_msg, param->responder_resources); cm_req_set_init_depth(req_msg, param->initiator_depth); cm_req_set_remote_resp_timeout(req_msg, param->remote_cm_response_timeout); @@ -815,40 +1023,57 @@ static void cm_format_req(struct cm_req_msg *req_msg, cm_req_set_starting_psn(req_msg, cpu_to_be32(param->starting_psn)); cm_req_set_local_resp_timeout(req_msg, param->local_cm_response_timeout); - cm_req_set_retry_count(req_msg, param->retry_count); req_msg->pkey = param->primary_path->pkey; cm_req_set_path_mtu(req_msg, param->primary_path->mtu); - cm_req_set_rnr_retry_count(req_msg, param->rnr_retry_count); cm_req_set_max_cm_retries(req_msg, param->max_cm_retries); - cm_req_set_srq(req_msg, param->srq); - - req_msg->primary_local_lid = param->primary_path->slid; - req_msg->primary_remote_lid = param->primary_path->dlid; - req_msg->primary_local_gid = param->primary_path->sgid; - req_msg->primary_remote_gid = param->primary_path->dgid; - cm_req_set_primary_flow_label(req_msg, param->primary_path->flow_label); - cm_req_set_primary_packet_rate(req_msg, param->primary_path->rate); - req_msg->primary_traffic_class = param->primary_path->traffic_class; - req_msg->primary_hop_limit = param->primary_path->hop_limit; - cm_req_set_primary_sl(req_msg, param->primary_path->sl); - cm_req_set_primary_subnet_local(req_msg, 1); /* local only... */ - cm_req_set_primary_local_ack_timeout(req_msg, - min(31, param->primary_path->packet_life_time + 1)); - if (param->alternate_path) { - req_msg->alt_local_lid = param->alternate_path->slid; - req_msg->alt_remote_lid = param->alternate_path->dlid; - req_msg->alt_local_gid = param->alternate_path->sgid; - req_msg->alt_remote_gid = param->alternate_path->dgid; + if (param->qp_type != IB_QPT_XRC_INI) { + cm_req_set_resp_res(req_msg, param->responder_resources); + cm_req_set_retry_count(req_msg, param->retry_count); + cm_req_set_rnr_retry_count(req_msg, param->rnr_retry_count); + cm_req_set_srq(req_msg, param->srq); + } + + if (pri_path->hop_limit <= 1) { + req_msg->primary_local_lid = pri_path->slid; + req_msg->primary_remote_lid = pri_path->dlid; + } else { + /* Work-around until there's a way to obtain remote LID info */ + req_msg->primary_local_lid = IB_LID_PERMISSIVE; + req_msg->primary_remote_lid = IB_LID_PERMISSIVE; + } + req_msg->primary_local_gid = pri_path->sgid; + req_msg->primary_remote_gid = pri_path->dgid; + cm_req_set_primary_flow_label(req_msg, pri_path->flow_label); + cm_req_set_primary_packet_rate(req_msg, pri_path->rate); + req_msg->primary_traffic_class = pri_path->traffic_class; + req_msg->primary_hop_limit = pri_path->hop_limit; + cm_req_set_primary_sl(req_msg, pri_path->sl); + cm_req_set_primary_subnet_local(req_msg, (pri_path->hop_limit <= 1)); + cm_req_set_primary_local_ack_timeout(req_msg, + cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay, + pri_path->packet_life_time)); + + if (alt_path) { + if (alt_path->hop_limit <= 1) { + req_msg->alt_local_lid = alt_path->slid; + req_msg->alt_remote_lid = alt_path->dlid; + } else { + req_msg->alt_local_lid = IB_LID_PERMISSIVE; + req_msg->alt_remote_lid = IB_LID_PERMISSIVE; + } + req_msg->alt_local_gid = alt_path->sgid; + req_msg->alt_remote_gid = alt_path->dgid; cm_req_set_alt_flow_label(req_msg, - param->alternate_path->flow_label); - cm_req_set_alt_packet_rate(req_msg, param->alternate_path->rate); - req_msg->alt_traffic_class = param->alternate_path->traffic_class; - req_msg->alt_hop_limit = param->alternate_path->hop_limit; - cm_req_set_alt_sl(req_msg, param->alternate_path->sl); - cm_req_set_alt_subnet_local(req_msg, 1); /* local only... */ + alt_path->flow_label); + cm_req_set_alt_packet_rate(req_msg, alt_path->rate); + req_msg->alt_traffic_class = alt_path->traffic_class; + req_msg->alt_hop_limit = alt_path->hop_limit; + cm_req_set_alt_sl(req_msg, alt_path->sl); + cm_req_set_alt_subnet_local(req_msg, (alt_path->hop_limit <= 1)); cm_req_set_alt_local_ack_timeout(req_msg, - min(31, param->alternate_path->packet_life_time + 1)); + cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay, + alt_path->packet_life_time)); } if (param->private_data && param->private_data_len) @@ -865,7 +1090,8 @@ static int cm_validate_req_param(struct ib_cm_req_param *param) if (!param->primary_path) return -EINVAL; - if (param->qp_type != IB_QPT_RC && param->qp_type != IB_QPT_UC) + if (param->qp_type != IB_QPT_RC && param->qp_type != IB_QPT_UC && + param->qp_type != IB_QPT_XRC_INI) return -EINVAL; if (param->private_data && @@ -904,8 +1130,10 @@ int ib_send_cm_req(struct ib_cm_id *cm_id, cm_id_priv->timewait_info = cm_create_timewait_info(cm_id_priv-> id.local_id); - if (IS_ERR(cm_id_priv->timewait_info)) + if (IS_ERR(cm_id_priv->timewait_info)) { + ret = PTR_ERR(cm_id_priv->timewait_info); goto out; + } ret = cm_init_av_by_path(param->primary_path, &cm_id_priv->av); if (ret) @@ -917,7 +1145,7 @@ int ib_send_cm_req(struct ib_cm_id *cm_id, goto error1; } cm_id->service_id = param->service_id; - cm_id->service_mask = __constant_cpu_to_be64(~0ULL); + cm_id->service_mask = ~cpu_to_be64(0); cm_id_priv->timeout_ms = cm_convert_to_ms( param->primary_path->packet_life_time) * 2 + cm_convert_to_ms( @@ -927,6 +1155,7 @@ int ib_send_cm_req(struct ib_cm_id *cm_id, cm_id_priv->responder_resources = param->responder_resources; cm_id_priv->retry_count = param->retry_count; cm_id_priv->path_mtu = param->primary_path->mtu; + cm_id_priv->pkey = param->primary_path->pkey; cm_id_priv->qp_type = param->qp_type; ret = cm_alloc_msg(cm_id_priv, &cm_id_priv->msg); @@ -941,8 +1170,6 @@ int ib_send_cm_req(struct ib_cm_id *cm_id, cm_id_priv->local_qpn = cm_req_get_local_qpn(req_msg); cm_id_priv->rq_psn = cm_req_get_starting_psn(req_msg); - cm_id_priv->local_ack_timeout = - cm_req_get_primary_local_ack_timeout(req_msg); spin_lock_irqsave(&cm_id_priv->lock, flags); ret = ib_post_send_mad(cm_id_priv->msg, NULL); @@ -1089,7 +1316,6 @@ static void cm_format_req_event(struct cm_work *work, static void cm_process_work(struct cm_id_private *cm_id_priv, struct cm_work *work) { - unsigned long flags; int ret; /* We will typically only have the current event to report. */ @@ -1097,9 +1323,9 @@ static void cm_process_work(struct cm_id_private *cm_id_priv, cm_free_work(work); while (!ret && !atomic_add_negative(-1, &cm_id_priv->work_count)) { - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); work = cm_dequeue_work(cm_id_priv); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); BUG_ON(!work); ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, &work->cm_event); @@ -1107,7 +1333,7 @@ static void cm_process_work(struct cm_id_private *cm_id_priv, } cm_deref_id(cm_id_priv); if (ret) - ib_destroy_cm_id(&cm_id_priv->id); + cm_destroy_id(&cm_id_priv->id, ret); } static void cm_format_mra(struct cm_mra_msg *mra_msg, @@ -1170,9 +1396,11 @@ static void cm_dup_req_handler(struct cm_work *work, struct cm_id_private *cm_id_priv) { struct ib_mad_send_buf *msg = NULL; - unsigned long flags; int ret; + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_REQ_COUNTER]); + /* Quick state check to discard duplicate REQs. */ if (cm_id_priv->id.state == IB_CM_REQ_RCVD) return; @@ -1181,7 +1409,7 @@ static void cm_dup_req_handler(struct cm_work *work, if (ret) return; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); switch (cm_id_priv->id.state) { case IB_CM_MRA_REQ_SENT: cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, @@ -1196,14 +1424,14 @@ static void cm_dup_req_handler(struct cm_work *work, default: goto unlock; } - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); ret = ib_post_send_mad(msg, NULL); if (ret) goto free; return; -unlock: spin_unlock_irqrestore(&cm_id_priv->lock, flags); +unlock: spin_unlock_irq(&cm_id_priv->lock); free: cm_free_msg(msg); } @@ -1213,49 +1441,81 @@ static struct cm_id_private * cm_match_req(struct cm_work *work, struct cm_id_private *listen_cm_id_priv, *cur_cm_id_priv; struct cm_timewait_info *timewait_info; struct cm_req_msg *req_msg; - unsigned long flags; req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad; - /* Check for duplicate REQ and stale connections. */ - spin_lock_irqsave(&cm.lock, flags); + /* Check for possible duplicate REQ. */ + spin_lock_irq(&cm.lock); timewait_info = cm_insert_remote_id(cm_id_priv->timewait_info); - if (!timewait_info) - timewait_info = cm_insert_remote_qpn(cm_id_priv->timewait_info); - if (timewait_info) { cur_cm_id_priv = cm_get_id(timewait_info->work.local_id, timewait_info->work.remote_id); - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock_irq(&cm.lock); if (cur_cm_id_priv) { cm_dup_req_handler(work, cur_cm_id_priv); cm_deref_id(cur_cm_id_priv); - } else - cm_issue_rej(work->port, work->mad_recv_wc, - IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REQ, - NULL, 0); - goto error; + } + return NULL; + } + + /* Check for stale connections. */ + timewait_info = cm_insert_remote_qpn(cm_id_priv->timewait_info); + if (timewait_info) { + cm_cleanup_timewait(cm_id_priv->timewait_info); + spin_unlock_irq(&cm.lock); + cm_issue_rej(work->port, work->mad_recv_wc, + IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REQ, + NULL, 0); + return NULL; } /* Find matching listen request. */ listen_cm_id_priv = cm_find_listen(cm_id_priv->id.device, - req_msg->service_id); + req_msg->service_id, + req_msg->private_data); if (!listen_cm_id_priv) { - spin_unlock_irqrestore(&cm.lock, flags); + cm_cleanup_timewait(cm_id_priv->timewait_info); + spin_unlock_irq(&cm.lock); cm_issue_rej(work->port, work->mad_recv_wc, IB_CM_REJ_INVALID_SERVICE_ID, CM_MSG_RESPONSE_REQ, NULL, 0); - goto error; + goto out; } atomic_inc(&listen_cm_id_priv->refcount); atomic_inc(&cm_id_priv->refcount); cm_id_priv->id.state = IB_CM_REQ_RCVD; atomic_inc(&cm_id_priv->work_count); - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock_irq(&cm.lock); +out: return listen_cm_id_priv; +} -error: cm_cleanup_timewait(cm_id_priv->timewait_info); - return NULL; +/* + * Work-around for inter-subnet connections. If the LIDs are permissive, + * we need to override the LID/SL data in the REQ with the LID information + * in the work completion. + */ +static void cm_process_routed_req(struct cm_req_msg *req_msg, struct ib_wc *wc) +{ + if (!cm_req_get_primary_subnet_local(req_msg)) { + if (req_msg->primary_local_lid == IB_LID_PERMISSIVE) { + req_msg->primary_local_lid = cpu_to_be16(wc->slid); + cm_req_set_primary_sl(req_msg, wc->sl); + } + + if (req_msg->primary_remote_lid == IB_LID_PERMISSIVE) + req_msg->primary_remote_lid = cpu_to_be16(wc->dlid_path_bits); + } + + if (!cm_req_get_alt_subnet_local(req_msg)) { + if (req_msg->alt_local_lid == IB_LID_PERMISSIVE) { + req_msg->alt_local_lid = cpu_to_be16(wc->slid); + cm_req_set_alt_sl(req_msg, wc->sl); + } + + if (req_msg->alt_remote_lid == IB_LID_PERMISSIVE) + req_msg->alt_remote_lid = cpu_to_be16(wc->dlid_path_bits); + } } static int cm_req_handler(struct cm_work *work) @@ -1267,19 +1527,20 @@ static int cm_req_handler(struct cm_work *work) req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad; - cm_id = ib_create_cm_id(work->port->cm_dev->device, NULL, NULL); + cm_id = ib_create_cm_id(work->port->cm_dev->ib_device, NULL, NULL); if (IS_ERR(cm_id)) return PTR_ERR(cm_id); cm_id_priv = container_of(cm_id, struct cm_id_private, id); cm_id_priv->id.remote_id = req_msg->local_comm_id; cm_init_av_for_response(work->port, work->mad_recv_wc->wc, + work->mad_recv_wc->recv_buf.grh, &cm_id_priv->av); cm_id_priv->timewait_info = cm_create_timewait_info(cm_id_priv-> id.local_id); if (IS_ERR(cm_id_priv->timewait_info)) { ret = PTR_ERR(cm_id_priv->timewait_info); - goto error1; + goto destroy; } cm_id_priv->timewait_info->work.remote_id = req_msg->local_comm_id; cm_id_priv->timewait_info->remote_ca_guid = req_msg->local_ca_guid; @@ -1288,22 +1549,37 @@ static int cm_req_handler(struct cm_work *work) listen_cm_id_priv = cm_match_req(work, cm_id_priv); if (!listen_cm_id_priv) { ret = -EINVAL; - goto error2; + kfree(cm_id_priv->timewait_info); + goto destroy; } cm_id_priv->id.cm_handler = listen_cm_id_priv->id.cm_handler; cm_id_priv->id.context = listen_cm_id_priv->id.context; cm_id_priv->id.service_id = req_msg->service_id; - cm_id_priv->id.service_mask = __constant_cpu_to_be64(~0ULL); + cm_id_priv->id.service_mask = ~cpu_to_be64(0); + cm_process_routed_req(req_msg, work->mad_recv_wc->wc); cm_format_paths_from_req(req_msg, &work->path[0], &work->path[1]); + + memcpy(work->path[0].dmac, cm_id_priv->av.ah_attr.dmac, ETH_ALEN); + work->path[0].vlan_id = cm_id_priv->av.ah_attr.vlan_id; ret = cm_init_av_by_path(&work->path[0], &cm_id_priv->av); - if (ret) - goto error3; + if (ret) { + ib_get_cached_gid(work->port->cm_dev->ib_device, + work->port->port_num, 0, &work->path[0].sgid); + ib_send_cm_rej(cm_id, IB_CM_REJ_INVALID_GID, + &work->path[0].sgid, sizeof work->path[0].sgid, + NULL, 0); + goto rejected; + } if (req_msg->alt_local_lid) { ret = cm_init_av_by_path(&work->path[1], &cm_id_priv->alt_av); - if (ret) - goto error3; + if (ret) { + ib_send_cm_rej(cm_id, IB_CM_REJ_INVALID_ALT_GID, + &work->path[0].sgid, + sizeof work->path[0].sgid, NULL, 0); + goto rejected; + } } cm_id_priv->tid = req_msg->hdr.tid; cm_id_priv->timeout_ms = cm_convert_to_ms( @@ -1313,9 +1589,8 @@ static int cm_req_handler(struct cm_work *work) cm_id_priv->initiator_depth = cm_req_get_resp_res(req_msg); cm_id_priv->responder_resources = cm_req_get_init_depth(req_msg); cm_id_priv->path_mtu = cm_req_get_path_mtu(req_msg); + cm_id_priv->pkey = req_msg->pkey; cm_id_priv->sq_psn = cm_req_get_starting_psn(req_msg); - cm_id_priv->local_ack_timeout = - cm_req_get_primary_local_ack_timeout(req_msg); cm_id_priv->retry_count = cm_req_get_retry_count(req_msg); cm_id_priv->rnr_retry_count = cm_req_get_rnr_retry_count(req_msg); cm_id_priv->qp_type = cm_req_get_qp_type(req_msg); @@ -1325,12 +1600,11 @@ static int cm_req_handler(struct cm_work *work) cm_deref_id(listen_cm_id_priv); return 0; -error3: atomic_dec(&cm_id_priv->refcount); +rejected: + atomic_dec(&cm_id_priv->refcount); cm_deref_id(listen_cm_id_priv); - cm_cleanup_timewait(cm_id_priv->timewait_info); -error2: kfree(cm_id_priv->timewait_info); - cm_id_priv->timewait_info = NULL; -error1: ib_destroy_cm_id(&cm_id_priv->id); +destroy: + ib_destroy_cm_id(cm_id); return ret; } @@ -1341,16 +1615,23 @@ static void cm_format_rep(struct cm_rep_msg *rep_msg, cm_format_mad_hdr(&rep_msg->hdr, CM_REP_ATTR_ID, cm_id_priv->tid); rep_msg->local_comm_id = cm_id_priv->id.local_id; rep_msg->remote_comm_id = cm_id_priv->id.remote_id; - cm_rep_set_local_qpn(rep_msg, cpu_to_be32(param->qp_num)); cm_rep_set_starting_psn(rep_msg, cpu_to_be32(param->starting_psn)); rep_msg->resp_resources = param->responder_resources; - rep_msg->initiator_depth = param->initiator_depth; - cm_rep_set_target_ack_delay(rep_msg, param->target_ack_delay); + cm_rep_set_target_ack_delay(rep_msg, + cm_id_priv->av.port->cm_dev->ack_delay); cm_rep_set_failover(rep_msg, param->failover_accepted); - cm_rep_set_flow_ctrl(rep_msg, param->flow_control); cm_rep_set_rnr_retry_count(rep_msg, param->rnr_retry_count); - cm_rep_set_srq(rep_msg, param->srq); - rep_msg->local_ca_guid = cm_id_priv->av.port->cm_dev->ca_guid; + rep_msg->local_ca_guid = cm_id_priv->id.device->node_guid; + + if (cm_id_priv->qp_type != IB_QPT_XRC_TGT) { + rep_msg->initiator_depth = param->initiator_depth; + cm_rep_set_flow_ctrl(rep_msg, param->flow_control); + cm_rep_set_srq(rep_msg, param->srq); + cm_rep_set_local_qpn(rep_msg, cpu_to_be32(param->qp_num)); + } else { + cm_rep_set_srq(rep_msg, 1); + cm_rep_set_local_eecn(rep_msg, cpu_to_be32(param->qp_num)); + } if (param->private_data && param->private_data_len) memcpy(rep_msg->private_data, param->private_data, @@ -1399,7 +1680,7 @@ int ib_send_cm_rep(struct ib_cm_id *cm_id, cm_id_priv->initiator_depth = param->initiator_depth; cm_id_priv->responder_resources = param->responder_resources; cm_id_priv->rq_psn = cm_rep_get_starting_psn(rep_msg); - cm_id_priv->local_qpn = cm_rep_get_local_qpn(rep_msg); + cm_id_priv->local_qpn = cpu_to_be32(param->qp_num & 0xFFFFFF); out: spin_unlock_irqrestore(&cm_id_priv->lock, flags); return ret; @@ -1470,7 +1751,7 @@ error: spin_unlock_irqrestore(&cm_id_priv->lock, flags); } EXPORT_SYMBOL(ib_send_cm_rtu); -static void cm_format_rep_event(struct cm_work *work) +static void cm_format_rep_event(struct cm_work *work, enum ib_qp_type qp_type) { struct cm_rep_msg *rep_msg; struct ib_cm_rep_event_param *param; @@ -1479,7 +1760,7 @@ static void cm_format_rep_event(struct cm_work *work) param = &work->cm_event.param.rep_rcvd; param->remote_ca_guid = rep_msg->local_ca_guid; param->remote_qkey = be32_to_cpu(rep_msg->local_qkey); - param->remote_qpn = be32_to_cpu(cm_rep_get_local_qpn(rep_msg)); + param->remote_qpn = be32_to_cpu(cm_rep_get_qpn(rep_msg, qp_type)); param->starting_psn = be32_to_cpu(cm_rep_get_starting_psn(rep_msg)); param->responder_resources = rep_msg->initiator_depth; param->initiator_depth = rep_msg->resp_resources; @@ -1496,7 +1777,6 @@ static void cm_dup_rep_handler(struct cm_work *work) struct cm_id_private *cm_id_priv; struct cm_rep_msg *rep_msg; struct ib_mad_send_buf *msg = NULL; - unsigned long flags; int ret; rep_msg = (struct cm_rep_msg *) work->mad_recv_wc->recv_buf.mad; @@ -1505,11 +1785,13 @@ static void cm_dup_rep_handler(struct cm_work *work) if (!cm_id_priv) return; + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_REP_COUNTER]); ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg); if (ret) goto deref; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state == IB_CM_ESTABLISHED) cm_format_rtu((struct cm_rtu_msg *) msg->mad, cm_id_priv, cm_id_priv->private_data, @@ -1521,14 +1803,14 @@ static void cm_dup_rep_handler(struct cm_work *work) cm_id_priv->private_data_len); else goto unlock; - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); ret = ib_post_send_mad(msg, NULL); if (ret) goto free; goto deref; -unlock: spin_unlock_irqrestore(&cm_id_priv->lock, flags); +unlock: spin_unlock_irq(&cm_id_priv->lock); free: cm_free_msg(msg); deref: cm_deref_id(cm_id_priv); } @@ -1537,7 +1819,6 @@ static int cm_rep_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; struct cm_rep_msg *rep_msg; - unsigned long flags; int ret; rep_msg = (struct cm_rep_msg *)work->mad_recv_wc->recv_buf.mad; @@ -1547,47 +1828,60 @@ static int cm_rep_handler(struct cm_work *work) return -EINVAL; } + cm_format_rep_event(work, cm_id_priv->qp_type); + + spin_lock_irq(&cm_id_priv->lock); + switch (cm_id_priv->id.state) { + case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: + break; + default: + spin_unlock_irq(&cm_id_priv->lock); + ret = -EINVAL; + goto error; + } + cm_id_priv->timewait_info->work.remote_id = rep_msg->local_comm_id; cm_id_priv->timewait_info->remote_ca_guid = rep_msg->local_ca_guid; - cm_id_priv->timewait_info->remote_qpn = cm_rep_get_local_qpn(rep_msg); + cm_id_priv->timewait_info->remote_qpn = cm_rep_get_qpn(rep_msg, cm_id_priv->qp_type); - spin_lock_irqsave(&cm.lock, flags); + spin_lock(&cm.lock); /* Check for duplicate REP. */ if (cm_insert_remote_id(cm_id_priv->timewait_info)) { - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock(&cm.lock); + spin_unlock_irq(&cm_id_priv->lock); ret = -EINVAL; goto error; } /* Check for a stale connection. */ if (cm_insert_remote_qpn(cm_id_priv->timewait_info)) { - spin_unlock_irqrestore(&cm.lock, flags); + rb_erase(&cm_id_priv->timewait_info->remote_id_node, + &cm.remote_id_table); + cm_id_priv->timewait_info->inserted_remote_id = 0; + spin_unlock(&cm.lock); + spin_unlock_irq(&cm_id_priv->lock); cm_issue_rej(work->port, work->mad_recv_wc, IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REP, NULL, 0); ret = -EINVAL; goto error; } - spin_unlock_irqrestore(&cm.lock, flags); - - cm_format_rep_event(work); + spin_unlock(&cm.lock); - spin_lock_irqsave(&cm_id_priv->lock, flags); - switch (cm_id_priv->id.state) { - case IB_CM_REQ_SENT: - case IB_CM_MRA_REQ_RCVD: - break; - default: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); - ret = -EINVAL; - goto error; - } cm_id_priv->id.state = IB_CM_REP_RCVD; cm_id_priv->id.remote_id = rep_msg->local_comm_id; - cm_id_priv->remote_qpn = cm_rep_get_local_qpn(rep_msg); + cm_id_priv->remote_qpn = cm_rep_get_qpn(rep_msg, cm_id_priv->qp_type); cm_id_priv->initiator_depth = rep_msg->resp_resources; cm_id_priv->responder_resources = rep_msg->initiator_depth; cm_id_priv->sq_psn = cm_rep_get_starting_psn(rep_msg); cm_id_priv->rnr_retry_count = cm_rep_get_rnr_retry_count(rep_msg); + cm_id_priv->target_ack_delay = cm_rep_get_target_ack_delay(rep_msg); + cm_id_priv->av.timeout = + cm_ack_timeout(cm_id_priv->target_ack_delay, + cm_id_priv->av.timeout - 1); + cm_id_priv->alt_av.timeout = + cm_ack_timeout(cm_id_priv->target_ack_delay, + cm_id_priv->alt_av.timeout - 1); /* todo: handle peer_to_peer */ @@ -1595,7 +1889,7 @@ static int cm_rep_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -1603,7 +1897,7 @@ static int cm_rep_handler(struct cm_work *work) cm_deref_id(cm_id_priv); return 0; -error: cm_cleanup_timewait(cm_id_priv->timewait_info); +error: cm_deref_id(cm_id_priv); return ret; } @@ -1611,17 +1905,16 @@ error: cm_cleanup_timewait(cm_id_priv->timewait_info); static int cm_establish_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; - unsigned long flags; int ret; - /* See comment in ib_cm_establish about lookup. */ + /* See comment in cm_establish about lookup. */ cm_id_priv = cm_acquire_id(work->local_id, work->remote_id); if (!cm_id_priv) return -EINVAL; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state != IB_CM_ESTABLISHED) { - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); goto out; } @@ -1629,7 +1922,7 @@ static int cm_establish_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -1645,7 +1938,6 @@ static int cm_rtu_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; struct cm_rtu_msg *rtu_msg; - unsigned long flags; int ret; rtu_msg = (struct cm_rtu_msg *)work->mad_recv_wc->recv_buf.mad; @@ -1656,10 +1948,12 @@ static int cm_rtu_handler(struct cm_work *work) work->cm_event.private_data = &rtu_msg->private_data; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state != IB_CM_REP_SENT && cm_id_priv->id.state != IB_CM_MRA_REP_RCVD) { - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_RTU_COUNTER]); goto out; } cm_id_priv->id.state = IB_CM_ESTABLISHED; @@ -1668,7 +1962,7 @@ static int cm_rtu_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -1714,6 +2008,10 @@ int ib_send_cm_dreq(struct ib_cm_id *cm_id, goto out; } + if (cm_id->lap_state == IB_CM_LAP_SENT || + cm_id->lap_state == IB_CM_MRA_LAP_RCVD) + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + ret = cm_alloc_msg(cm_id_priv, &msg); if (ret) { cm_enter_timewait(cm_id_priv); @@ -1800,23 +2098,52 @@ out: spin_unlock_irqrestore(&cm_id_priv->lock, flags); } EXPORT_SYMBOL(ib_send_cm_drep); +static int cm_issue_drep(struct cm_port *port, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_mad_send_buf *msg = NULL; + struct cm_dreq_msg *dreq_msg; + struct cm_drep_msg *drep_msg; + int ret; + + ret = cm_alloc_response_msg(port, mad_recv_wc, &msg); + if (ret) + return ret; + + dreq_msg = (struct cm_dreq_msg *) mad_recv_wc->recv_buf.mad; + drep_msg = (struct cm_drep_msg *) msg->mad; + + cm_format_mad_hdr(&drep_msg->hdr, CM_DREP_ATTR_ID, dreq_msg->hdr.tid); + drep_msg->remote_comm_id = dreq_msg->local_comm_id; + drep_msg->local_comm_id = dreq_msg->remote_comm_id; + + ret = ib_post_send_mad(msg, NULL); + if (ret) + cm_free_msg(msg); + + return ret; +} + static int cm_dreq_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; struct cm_dreq_msg *dreq_msg; struct ib_mad_send_buf *msg = NULL; - unsigned long flags; int ret; dreq_msg = (struct cm_dreq_msg *)work->mad_recv_wc->recv_buf.mad; cm_id_priv = cm_acquire_id(dreq_msg->remote_comm_id, dreq_msg->local_comm_id); - if (!cm_id_priv) + if (!cm_id_priv) { + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_DREQ_COUNTER]); + cm_issue_drep(work->port, work->mad_recv_wc); return -EINVAL; + } work->cm_event.private_data = &dreq_msg->private_data; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->local_qpn != cm_dreq_get_remote_qpn(dreq_msg)) goto unlock; @@ -1826,20 +2153,30 @@ static int cm_dreq_handler(struct cm_work *work) ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); break; case IB_CM_ESTABLISHED: + if (cm_id_priv->id.lap_state == IB_CM_LAP_SENT || + cm_id_priv->id.lap_state == IB_CM_MRA_LAP_RCVD) + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + break; case IB_CM_MRA_REP_RCVD: break; case IB_CM_TIMEWAIT: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_DREQ_COUNTER]); if (cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg)) goto unlock; cm_format_drep((struct cm_drep_msg *) msg->mad, cm_id_priv, cm_id_priv->private_data, cm_id_priv->private_data_len); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ib_post_send_mad(msg, NULL)) cm_free_msg(msg); goto deref; + case IB_CM_DREQ_RCVD: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_DREQ_COUNTER]); + goto unlock; default: goto unlock; } @@ -1848,7 +2185,7 @@ static int cm_dreq_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -1856,7 +2193,7 @@ static int cm_dreq_handler(struct cm_work *work) cm_deref_id(cm_id_priv); return 0; -unlock: spin_unlock_irqrestore(&cm_id_priv->lock, flags); +unlock: spin_unlock_irq(&cm_id_priv->lock); deref: cm_deref_id(cm_id_priv); return -EINVAL; } @@ -1865,7 +2202,6 @@ static int cm_drep_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; struct cm_drep_msg *drep_msg; - unsigned long flags; int ret; drep_msg = (struct cm_drep_msg *)work->mad_recv_wc->recv_buf.mad; @@ -1876,10 +2212,10 @@ static int cm_drep_handler(struct cm_work *work) work->cm_event.private_data = &drep_msg->private_data; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state != IB_CM_DREQ_SENT && cm_id_priv->id.state != IB_CM_DREQ_RCVD) { - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); goto out; } cm_enter_timewait(cm_id_priv); @@ -1888,7 +2224,7 @@ static int cm_drep_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -1978,28 +2314,28 @@ static struct cm_id_private * cm_acquire_rejected_id(struct cm_rej_msg *rej_msg) { struct cm_timewait_info *timewait_info; struct cm_id_private *cm_id_priv; - unsigned long flags; __be32 remote_id; remote_id = rej_msg->local_comm_id; if (__be16_to_cpu(rej_msg->reason) == IB_CM_REJ_TIMEOUT) { - spin_lock_irqsave(&cm.lock, flags); + spin_lock_irq(&cm.lock); timewait_info = cm_find_remote_id( *((__be64 *) rej_msg->ari), remote_id); if (!timewait_info) { - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock_irq(&cm.lock); return NULL; } - cm_id_priv = idr_find(&cm.local_id_table, - (__force int) timewait_info->work.local_id); + cm_id_priv = idr_find(&cm.local_id_table, (__force int) + (timewait_info->work.local_id ^ + cm.random_id_operand)); if (cm_id_priv) { if (cm_id_priv->id.remote_id == remote_id) atomic_inc(&cm_id_priv->refcount); else cm_id_priv = NULL; } - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock_irq(&cm.lock); } else if (cm_rej_get_msg_rejected(rej_msg) == CM_MSG_RESPONSE_REQ) cm_id_priv = cm_acquire_id(rej_msg->remote_comm_id, 0); else @@ -2012,7 +2348,6 @@ static int cm_rej_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; struct cm_rej_msg *rej_msg; - unsigned long flags; int ret; rej_msg = (struct cm_rej_msg *)work->mad_recv_wc->recv_buf.mad; @@ -2022,7 +2357,7 @@ static int cm_rej_handler(struct cm_work *work) cm_format_rej_event(work); - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); switch (cm_id_priv->id.state) { case IB_CM_REQ_SENT: case IB_CM_MRA_REQ_RCVD: @@ -2042,11 +2377,20 @@ static int cm_rej_handler(struct cm_work *work) /* fall through */ case IB_CM_REP_RCVD: case IB_CM_MRA_REP_SENT: - case IB_CM_ESTABLISHED: cm_enter_timewait(cm_id_priv); break; + case IB_CM_ESTABLISHED: + if (cm_id_priv->id.lap_state == IB_CM_LAP_UNINIT || + cm_id_priv->id.lap_state == IB_CM_LAP_SENT) { + if (cm_id_priv->id.lap_state == IB_CM_LAP_SENT) + ib_cancel_mad(cm_id_priv->av.port->mad_agent, + cm_id_priv->msg); + cm_enter_timewait(cm_id_priv); + break; + } + /* fall through */ default: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); ret = -EINVAL; goto out; } @@ -2054,7 +2398,7 @@ static int cm_rej_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -2073,6 +2417,9 @@ int ib_send_cm_mra(struct ib_cm_id *cm_id, { struct cm_id_private *cm_id_priv; struct ib_mad_send_buf *msg; + enum ib_cm_state cm_state; + enum ib_cm_lap_state lap_state; + enum cm_msg_response msg_response; void *data; unsigned long flags; int ret; @@ -2089,48 +2436,42 @@ int ib_send_cm_mra(struct ib_cm_id *cm_id, spin_lock_irqsave(&cm_id_priv->lock, flags); switch(cm_id_priv->id.state) { case IB_CM_REQ_RCVD: - ret = cm_alloc_msg(cm_id_priv, &msg); - if (ret) - goto error1; - - cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, - CM_MSG_RESPONSE_REQ, service_timeout, - private_data, private_data_len); - ret = ib_post_send_mad(msg, NULL); - if (ret) - goto error2; - cm_id->state = IB_CM_MRA_REQ_SENT; + cm_state = IB_CM_MRA_REQ_SENT; + lap_state = cm_id->lap_state; + msg_response = CM_MSG_RESPONSE_REQ; break; case IB_CM_REP_RCVD: - ret = cm_alloc_msg(cm_id_priv, &msg); - if (ret) - goto error1; - - cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, - CM_MSG_RESPONSE_REP, service_timeout, - private_data, private_data_len); - ret = ib_post_send_mad(msg, NULL); - if (ret) - goto error2; - cm_id->state = IB_CM_MRA_REP_SENT; + cm_state = IB_CM_MRA_REP_SENT; + lap_state = cm_id->lap_state; + msg_response = CM_MSG_RESPONSE_REP; break; case IB_CM_ESTABLISHED: + if (cm_id->lap_state == IB_CM_LAP_RCVD) { + cm_state = cm_id->state; + lap_state = IB_CM_MRA_LAP_SENT; + msg_response = CM_MSG_RESPONSE_OTHER; + break; + } + default: + ret = -EINVAL; + goto error1; + } + + if (!(service_timeout & IB_CM_MRA_FLAG_DELAY)) { ret = cm_alloc_msg(cm_id_priv, &msg); if (ret) goto error1; cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, - CM_MSG_RESPONSE_OTHER, service_timeout, + msg_response, service_timeout, private_data, private_data_len); ret = ib_post_send_mad(msg, NULL); if (ret) goto error2; - cm_id->lap_state = IB_CM_MRA_LAP_SENT; - break; - default: - ret = -EINVAL; - goto error1; } + + cm_id->state = cm_state; + cm_id->lap_state = lap_state; cm_id_priv->service_timeout = service_timeout; cm_set_private_data(cm_id_priv, data, private_data_len); spin_unlock_irqrestore(&cm_id_priv->lock, flags); @@ -2165,7 +2506,6 @@ static int cm_mra_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; struct cm_mra_msg *mra_msg; - unsigned long flags; int timeout, ret; mra_msg = (struct cm_mra_msg *)work->mad_recv_wc->recv_buf.mad; @@ -2177,9 +2517,9 @@ static int cm_mra_handler(struct cm_work *work) work->cm_event.param.mra_rcvd.service_timeout = cm_mra_get_service_timeout(mra_msg); timeout = cm_convert_to_ms(cm_mra_get_service_timeout(mra_msg)) + - cm_convert_to_ms(cm_id_priv->av.packet_life_time); + cm_convert_to_ms(cm_id_priv->av.timeout); - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); switch (cm_id_priv->id.state) { case IB_CM_REQ_SENT: if (cm_mra_get_msg_mraed(mra_msg) != CM_MSG_RESPONSE_REQ || @@ -2199,10 +2539,20 @@ static int cm_mra_handler(struct cm_work *work) if (cm_mra_get_msg_mraed(mra_msg) != CM_MSG_RESPONSE_OTHER || cm_id_priv->id.lap_state != IB_CM_LAP_SENT || ib_modify_mad(cm_id_priv->av.port->mad_agent, - cm_id_priv->msg, timeout)) + cm_id_priv->msg, timeout)) { + if (cm_id_priv->id.lap_state == IB_CM_MRA_LAP_RCVD) + atomic_long_inc(&work->port-> + counter_group[CM_RECV_DUPLICATES]. + counter[CM_MRA_COUNTER]); goto out; + } cm_id_priv->id.lap_state = IB_CM_MRA_LAP_RCVD; break; + case IB_CM_MRA_REQ_RCVD: + case IB_CM_MRA_REP_RCVD: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_MRA_COUNTER]); + /* fall through */ default: goto out; } @@ -2212,7 +2562,7 @@ static int cm_mra_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -2220,7 +2570,7 @@ static int cm_mra_handler(struct cm_work *work) cm_deref_id(cm_id_priv); return 0; out: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); cm_deref_id(cm_id_priv); return -EINVAL; } @@ -2249,7 +2599,8 @@ static void cm_format_lap(struct cm_lap_msg *lap_msg, cm_lap_set_sl(lap_msg, alternate_path->sl); cm_lap_set_subnet_local(lap_msg, 1); /* local only... */ cm_lap_set_local_ack_timeout(lap_msg, - min(31, alternate_path->packet_life_time + 1)); + cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay, + alternate_path->packet_life_time)); if (private_data && private_data_len) memcpy(lap_msg->private_data, private_data, private_data_len); @@ -2271,11 +2622,19 @@ int ib_send_cm_lap(struct ib_cm_id *cm_id, cm_id_priv = container_of(cm_id, struct cm_id_private, id); spin_lock_irqsave(&cm_id_priv->lock, flags); if (cm_id->state != IB_CM_ESTABLISHED || - cm_id->lap_state != IB_CM_LAP_IDLE) { + (cm_id->lap_state != IB_CM_LAP_UNINIT && + cm_id->lap_state != IB_CM_LAP_IDLE)) { ret = -EINVAL; goto out; } + ret = cm_init_av_by_path(alternate_path, &cm_id_priv->alt_av); + if (ret) + goto out; + cm_id_priv->alt_av.timeout = + cm_ack_timeout(cm_id_priv->target_ack_delay, + cm_id_priv->alt_av.timeout - 1); + ret = cm_alloc_msg(cm_id_priv, &msg); if (ret) goto out; @@ -2300,7 +2659,8 @@ out: spin_unlock_irqrestore(&cm_id_priv->lock, flags); } EXPORT_SYMBOL(ib_send_cm_lap); -static void cm_format_path_from_lap(struct ib_sa_path_rec *path, +static void cm_format_path_from_lap(struct cm_id_private *cm_id_priv, + struct ib_sa_path_rec *path, struct cm_lap_msg *lap_msg) { memset(path, 0, sizeof *path); @@ -2312,10 +2672,10 @@ static void cm_format_path_from_lap(struct ib_sa_path_rec *path, path->hop_limit = lap_msg->alt_hop_limit; path->traffic_class = cm_lap_get_traffic_class(lap_msg); path->reversible = 1; - /* pkey is same as in REQ */ + path->pkey = cm_id_priv->pkey; path->sl = cm_lap_get_sl(lap_msg); path->mtu_selector = IB_SA_EQ; - /* mtu is same as in REQ */ + path->mtu = cm_id_priv->path_mtu; path->rate_selector = IB_SA_EQ; path->rate = cm_lap_get_packet_rate(lap_msg); path->packet_life_time_selector = IB_SA_EQ; @@ -2329,7 +2689,6 @@ static int cm_lap_handler(struct cm_work *work) struct cm_lap_msg *lap_msg; struct ib_cm_lap_event_param *param; struct ib_mad_send_buf *msg = NULL; - unsigned long flags; int ret; /* todo: verify LAP request and send reject APR if invalid. */ @@ -2341,17 +2700,20 @@ static int cm_lap_handler(struct cm_work *work) param = &work->cm_event.param.lap_rcvd; param->alternate_path = &work->path[0]; - cm_format_path_from_lap(param->alternate_path, lap_msg); + cm_format_path_from_lap(cm_id_priv, param->alternate_path, lap_msg); work->cm_event.private_data = &lap_msg->private_data; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state != IB_CM_ESTABLISHED) goto unlock; switch (cm_id_priv->id.lap_state) { + case IB_CM_LAP_UNINIT: case IB_CM_LAP_IDLE: break; case IB_CM_MRA_LAP_SENT: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_LAP_COUNTER]); if (cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg)) goto unlock; @@ -2360,21 +2722,29 @@ static int cm_lap_handler(struct cm_work *work) cm_id_priv->service_timeout, cm_id_priv->private_data, cm_id_priv->private_data_len); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ib_post_send_mad(msg, NULL)) cm_free_msg(msg); goto deref; + case IB_CM_LAP_RCVD: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_LAP_COUNTER]); + goto unlock; default: goto unlock; } cm_id_priv->id.lap_state = IB_CM_LAP_RCVD; cm_id_priv->tid = lap_msg->hdr.tid; + cm_init_av_for_response(work->port, work->mad_recv_wc->wc, + work->mad_recv_wc->recv_buf.grh, + &cm_id_priv->av); + cm_init_av_by_path(param->alternate_path, &cm_id_priv->alt_av); ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -2382,7 +2752,7 @@ static int cm_lap_handler(struct cm_work *work) cm_deref_id(cm_id_priv); return 0; -unlock: spin_unlock_irqrestore(&cm_id_priv->lock, flags); +unlock: spin_unlock_irq(&cm_id_priv->lock); deref: cm_deref_id(cm_id_priv); return -EINVAL; } @@ -2457,7 +2827,6 @@ static int cm_apr_handler(struct cm_work *work) { struct cm_id_private *cm_id_priv; struct cm_apr_msg *apr_msg; - unsigned long flags; int ret; apr_msg = (struct cm_apr_msg *)work->mad_recv_wc->recv_buf.mad; @@ -2471,11 +2840,11 @@ static int cm_apr_handler(struct cm_work *work) work->cm_event.param.apr_rcvd.info_len = apr_msg->info_length; work->cm_event.private_data = &apr_msg->private_data; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state != IB_CM_ESTABLISHED || (cm_id_priv->id.lap_state != IB_CM_LAP_SENT && cm_id_priv->id.lap_state != IB_CM_MRA_LAP_RCVD)) { - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); goto out; } cm_id_priv->id.lap_state = IB_CM_LAP_IDLE; @@ -2485,7 +2854,7 @@ static int cm_apr_handler(struct cm_work *work) ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -2501,28 +2870,29 @@ static int cm_timewait_handler(struct cm_work *work) { struct cm_timewait_info *timewait_info; struct cm_id_private *cm_id_priv; - unsigned long flags; int ret; timewait_info = (struct cm_timewait_info *)work; - cm_cleanup_timewait(timewait_info); + spin_lock_irq(&cm.lock); + list_del(&timewait_info->list); + spin_unlock_irq(&cm.lock); cm_id_priv = cm_acquire_id(timewait_info->work.local_id, timewait_info->work.remote_id); if (!cm_id_priv) return -EINVAL; - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state != IB_CM_TIMEWAIT || cm_id_priv->remote_qpn != timewait_info->remote_qpn) { - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); goto out; } cm_id_priv->id.state = IB_CM_IDLE; ret = atomic_inc_and_test(&cm_id_priv->work_count); if (!ret) list_add_tail(&work->list, &cm_id_priv->work_list); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); if (ret) cm_process_work(cm_id_priv, work); @@ -2541,7 +2911,7 @@ static void cm_format_sidr_req(struct cm_sidr_req_msg *sidr_req_msg, cm_format_mad_hdr(&sidr_req_msg->hdr, CM_SIDR_REQ_ATTR_ID, cm_form_tid(cm_id_priv, CM_MSG_SEQUENCE_SIDR)); sidr_req_msg->request_id = cm_id_priv->id.local_id; - sidr_req_msg->pkey = cpu_to_be16(param->pkey); + sidr_req_msg->pkey = param->path->pkey; sidr_req_msg->service_id = param->service_id; if (param->private_data && param->private_data_len) @@ -2567,7 +2937,7 @@ int ib_send_cm_sidr_req(struct ib_cm_id *cm_id, goto out; cm_id->service_id = param->service_id; - cm_id->service_mask = __constant_cpu_to_be64(~0ULL); + cm_id->service_mask = ~cpu_to_be64(0); cm_id_priv->timeout_ms = param->timeout_ms; cm_id_priv->max_cm_retries = param->max_cm_retries; ret = cm_alloc_msg(cm_id_priv, &msg); @@ -2619,9 +2989,8 @@ static int cm_sidr_req_handler(struct cm_work *work) struct cm_id_private *cm_id_priv, *cur_cm_id_priv; struct cm_sidr_req_msg *sidr_req_msg; struct ib_wc *wc; - unsigned long flags; - cm_id = ib_create_cm_id(work->port->cm_dev->device, NULL, NULL); + cm_id = ib_create_cm_id(work->port->cm_dev->ib_device, NULL, NULL); if (IS_ERR(cm_id)) return PTR_ERR(cm_id); cm_id_priv = container_of(cm_id, struct cm_id_private, id); @@ -2633,33 +3002,37 @@ static int cm_sidr_req_handler(struct cm_work *work) cm_id_priv->av.dgid.global.subnet_prefix = cpu_to_be64(wc->slid); cm_id_priv->av.dgid.global.interface_id = 0; cm_init_av_for_response(work->port, work->mad_recv_wc->wc, + work->mad_recv_wc->recv_buf.grh, &cm_id_priv->av); cm_id_priv->id.remote_id = sidr_req_msg->request_id; - cm_id_priv->id.state = IB_CM_SIDR_REQ_RCVD; cm_id_priv->tid = sidr_req_msg->hdr.tid; atomic_inc(&cm_id_priv->work_count); - spin_lock_irqsave(&cm.lock, flags); + spin_lock_irq(&cm.lock); cur_cm_id_priv = cm_insert_remote_sidr(cm_id_priv); if (cur_cm_id_priv) { - spin_unlock_irqrestore(&cm.lock, flags); + spin_unlock_irq(&cm.lock); + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_SIDR_REQ_COUNTER]); goto out; /* Duplicate message. */ } + cm_id_priv->id.state = IB_CM_SIDR_REQ_RCVD; cur_cm_id_priv = cm_find_listen(cm_id->device, - sidr_req_msg->service_id); + sidr_req_msg->service_id, + sidr_req_msg->private_data); if (!cur_cm_id_priv) { - rb_erase(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table); - spin_unlock_irqrestore(&cm.lock, flags); - /* todo: reply with no match */ + spin_unlock_irq(&cm.lock); + cm_reject_sidr_req(cm_id_priv, IB_SIDR_UNSUPPORTED); goto out; /* No match. */ } atomic_inc(&cur_cm_id_priv->refcount); - spin_unlock_irqrestore(&cm.lock, flags); + atomic_inc(&cm_id_priv->refcount); + spin_unlock_irq(&cm.lock); cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler; cm_id_priv->id.context = cur_cm_id_priv->id.context; cm_id_priv->id.service_id = sidr_req_msg->service_id; - cm_id_priv->id.service_mask = __constant_cpu_to_be64(~0ULL); + cm_id_priv->id.service_mask = ~cpu_to_be64(0); cm_format_sidr_req_event(work, &cur_cm_id_priv->id); cm_process_work(cm_id_priv, work); @@ -2755,7 +3128,6 @@ static int cm_sidr_rep_handler(struct cm_work *work) { struct cm_sidr_rep_msg *sidr_rep_msg; struct cm_id_private *cm_id_priv; - unsigned long flags; sidr_rep_msg = (struct cm_sidr_rep_msg *) work->mad_recv_wc->recv_buf.mad; @@ -2763,14 +3135,14 @@ static int cm_sidr_rep_handler(struct cm_work *work) if (!cm_id_priv) return -EINVAL; /* Unmatched reply. */ - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); if (cm_id_priv->id.state != IB_CM_SIDR_REQ_SENT) { - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); goto out; } cm_id_priv->id.state = IB_CM_IDLE; ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); cm_format_sidr_rep_event(work); cm_process_work(cm_id_priv, work); @@ -2786,14 +3158,13 @@ static void cm_process_send_error(struct ib_mad_send_buf *msg, struct cm_id_private *cm_id_priv; struct ib_cm_event cm_event; enum ib_cm_state state; - unsigned long flags; int ret; memset(&cm_event, 0, sizeof cm_event); cm_id_priv = msg->context[0]; /* Discard old sends or ones without a response. */ - spin_lock_irqsave(&cm_id_priv->lock, flags); + spin_lock_irq(&cm_id_priv->lock); state = (enum ib_cm_state) (unsigned long) msg->context[1]; if (msg != cm_id_priv->msg || state != cm_id_priv->id.state) goto discard; @@ -2820,7 +3191,7 @@ static void cm_process_send_error(struct ib_mad_send_buf *msg, default: goto discard; } - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); cm_event.param.send_status = wc_status; /* No other events can occur on the cm_id at this point. */ @@ -2830,7 +3201,7 @@ static void cm_process_send_error(struct ib_mad_send_buf *msg, ib_destroy_cm_id(&cm_id_priv->id); return; discard: - spin_unlock_irqrestore(&cm_id_priv->lock, flags); + spin_unlock_irq(&cm_id_priv->lock); cm_free_msg(msg); } @@ -2838,6 +3209,27 @@ static void cm_send_handler(struct ib_mad_agent *mad_agent, struct ib_mad_send_wc *mad_send_wc) { struct ib_mad_send_buf *msg = mad_send_wc->send_buf; + struct cm_port *port; + u16 attr_index; + + port = mad_agent->context; + attr_index = be16_to_cpu(((struct ib_mad_hdr *) + msg->mad)->attr_id) - CM_ATTR_ID_OFFSET; + + /* + * If the send was in response to a received message (context[0] is not + * set to a cm_id), and is not a REJ, then it is a send that was + * manually retried. + */ + if (!msg->context[0] && (attr_index != CM_REJ_COUNTER)) + msg->retries = 1; + + atomic_long_add(1 + msg->retries, + &port->counter_group[CM_XMIT].counter[attr_index]); + if (msg->retries) + atomic_long_add(msg->retries, + &port->counter_group[CM_XMIT_RETRIES]. + counter[attr_index]); switch (mad_send_wc->status) { case IB_WC_SUCCESS: @@ -2853,9 +3245,9 @@ static void cm_send_handler(struct ib_mad_agent *mad_agent, } } -static void cm_work_handler(void *data) +static void cm_work_handler(struct work_struct *_work) { - struct cm_work *work = data; + struct cm_work *work = container_of(_work, struct cm_work, work.work); int ret; switch (work->cm_event.event) { @@ -2906,7 +3298,7 @@ static void cm_work_handler(void *data) cm_free_work(work); } -int ib_cm_establish(struct ib_cm_id *cm_id) +static int cm_establish(struct ib_cm_id *cm_id) { struct cm_id_private *cm_id_priv; struct cm_work *work; @@ -2945,22 +3337,61 @@ int ib_cm_establish(struct ib_cm_id *cm_id) * we need to find the cm_id once we're in the context of the * worker thread, rather than holding a reference on it. */ - INIT_WORK(&work->work, cm_work_handler, work); + INIT_DELAYED_WORK(&work->work, cm_work_handler); work->local_id = cm_id->local_id; work->remote_id = cm_id->remote_id; work->mad_recv_wc = NULL; work->cm_event.event = IB_CM_USER_ESTABLISHED; - queue_work(cm.wq, &work->work); + queue_delayed_work(cm.wq, &work->work, 0); out: return ret; } -EXPORT_SYMBOL(ib_cm_establish); + +static int cm_migrate(struct ib_cm_id *cm_id) +{ + struct cm_id_private *cm_id_priv; + unsigned long flags; + int ret = 0; + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id->state == IB_CM_ESTABLISHED && + (cm_id->lap_state == IB_CM_LAP_UNINIT || + cm_id->lap_state == IB_CM_LAP_IDLE)) { + cm_id->lap_state = IB_CM_LAP_IDLE; + cm_id_priv->av = cm_id_priv->alt_av; + } else + ret = -EINVAL; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + return ret; +} + +int ib_cm_notify(struct ib_cm_id *cm_id, enum ib_event_type event) +{ + int ret; + + switch (event) { + case IB_EVENT_COMM_EST: + ret = cm_establish(cm_id); + break; + case IB_EVENT_PATH_MIG: + ret = cm_migrate(cm_id); + break; + default: + ret = -EINVAL; + } + return ret; +} +EXPORT_SYMBOL(ib_cm_notify); static void cm_recv_handler(struct ib_mad_agent *mad_agent, struct ib_mad_recv_wc *mad_recv_wc) { + struct cm_port *port = mad_agent->context; struct cm_work *work; enum ib_cm_event_type event; + u16 attr_id; int paths = 0; switch (mad_recv_wc->recv_buf.mad->mad_hdr.attr_id) { @@ -3005,6 +3436,10 @@ static void cm_recv_handler(struct ib_mad_agent *mad_agent, return; } + attr_id = be16_to_cpu(mad_recv_wc->recv_buf.mad->mad_hdr.attr_id); + atomic_long_inc(&port->counter_group[CM_RECV]. + counter[attr_id - CM_ATTR_ID_OFFSET]); + work = kmalloc(sizeof *work + sizeof(struct ib_sa_path_rec) * paths, GFP_KERNEL); if (!work) { @@ -3012,11 +3447,11 @@ static void cm_recv_handler(struct ib_mad_agent *mad_agent, return; } - INIT_WORK(&work->work, cm_work_handler, work); + INIT_DELAYED_WORK(&work->work, cm_work_handler); work->cm_event.event = event; work->mad_recv_wc = mad_recv_wc; - work->port = (struct cm_port *)mad_agent->context; - queue_work(cm.wq, &work->work); + work->port = port; + queue_delayed_work(cm.wq, &work->work, 0); } static int cm_init_qp_init_attr(struct cm_id_private *cm_id_priv, @@ -3039,10 +3474,10 @@ static int cm_init_qp_init_attr(struct cm_id_private *cm_id_priv, case IB_CM_ESTABLISHED: *qp_attr_mask = IB_QP_STATE | IB_QP_ACCESS_FLAGS | IB_QP_PKEY_INDEX | IB_QP_PORT; - qp_attr->qp_access_flags = IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE; + qp_attr->qp_access_flags = IB_ACCESS_REMOTE_WRITE; if (cm_id_priv->responder_resources) - qp_attr->qp_access_flags |= IB_ACCESS_REMOTE_READ; + qp_attr->qp_access_flags |= IB_ACCESS_REMOTE_READ | + IB_ACCESS_REMOTE_ATOMIC; qp_attr->pkey_index = cm_id_priv->av.pkey_index; qp_attr->port_num = cm_id_priv->av.port->port_num; ret = 0; @@ -3074,10 +3509,37 @@ static int cm_init_qp_rtr_attr(struct cm_id_private *cm_id_priv, *qp_attr_mask = IB_QP_STATE | IB_QP_AV | IB_QP_PATH_MTU | IB_QP_DEST_QPN | IB_QP_RQ_PSN; qp_attr->ah_attr = cm_id_priv->av.ah_attr; + if (!cm_id_priv->av.valid) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return -EINVAL; + } + if (cm_id_priv->av.ah_attr.vlan_id != 0xffff) { + qp_attr->vlan_id = cm_id_priv->av.ah_attr.vlan_id; + *qp_attr_mask |= IB_QP_VID; + } + if (!is_zero_ether_addr(cm_id_priv->av.smac)) { + memcpy(qp_attr->smac, cm_id_priv->av.smac, + sizeof(qp_attr->smac)); + *qp_attr_mask |= IB_QP_SMAC; + } + if (cm_id_priv->alt_av.valid) { + if (cm_id_priv->alt_av.ah_attr.vlan_id != 0xffff) { + qp_attr->alt_vlan_id = + cm_id_priv->alt_av.ah_attr.vlan_id; + *qp_attr_mask |= IB_QP_ALT_VID; + } + if (!is_zero_ether_addr(cm_id_priv->alt_av.smac)) { + memcpy(qp_attr->alt_smac, + cm_id_priv->alt_av.smac, + sizeof(qp_attr->alt_smac)); + *qp_attr_mask |= IB_QP_ALT_SMAC; + } + } qp_attr->path_mtu = cm_id_priv->path_mtu; qp_attr->dest_qp_num = be32_to_cpu(cm_id_priv->remote_qpn); qp_attr->rq_psn = be32_to_cpu(cm_id_priv->rq_psn); - if (cm_id_priv->qp_type == IB_QPT_RC) { + if (cm_id_priv->qp_type == IB_QPT_RC || + cm_id_priv->qp_type == IB_QPT_XRC_TGT) { *qp_attr_mask |= IB_QP_MAX_DEST_RD_ATOMIC | IB_QP_MIN_RNR_TIMER; qp_attr->max_dest_rd_atomic = @@ -3086,6 +3548,9 @@ static int cm_init_qp_rtr_attr(struct cm_id_private *cm_id_priv, } if (cm_id_priv->alt_av.ah_attr.dlid) { *qp_attr_mask |= IB_QP_ALT_PATH; + qp_attr->alt_port_num = cm_id_priv->alt_av.port->port_num; + qp_attr->alt_pkey_index = cm_id_priv->alt_av.pkey_index; + qp_attr->alt_timeout = cm_id_priv->alt_av.timeout; qp_attr->alt_ah_attr = cm_id_priv->alt_av.ah_attr; } ret = 0; @@ -3107,24 +3572,44 @@ static int cm_init_qp_rts_attr(struct cm_id_private *cm_id_priv, spin_lock_irqsave(&cm_id_priv->lock, flags); switch (cm_id_priv->id.state) { + /* Allow transition to RTS before sending REP */ + case IB_CM_REQ_RCVD: + case IB_CM_MRA_REQ_SENT: + case IB_CM_REP_RCVD: case IB_CM_MRA_REP_SENT: case IB_CM_REP_SENT: case IB_CM_MRA_REP_RCVD: case IB_CM_ESTABLISHED: - *qp_attr_mask = IB_QP_STATE | IB_QP_SQ_PSN; - qp_attr->sq_psn = be32_to_cpu(cm_id_priv->sq_psn); - if (cm_id_priv->qp_type == IB_QPT_RC) { - *qp_attr_mask |= IB_QP_TIMEOUT | IB_QP_RETRY_CNT | - IB_QP_RNR_RETRY | - IB_QP_MAX_QP_RD_ATOMIC; - qp_attr->timeout = cm_id_priv->local_ack_timeout; - qp_attr->retry_cnt = cm_id_priv->retry_count; - qp_attr->rnr_retry = cm_id_priv->rnr_retry_count; - qp_attr->max_rd_atomic = cm_id_priv->initiator_depth; - } - if (cm_id_priv->alt_av.ah_attr.dlid) { - *qp_attr_mask |= IB_QP_PATH_MIG_STATE; + if (cm_id_priv->id.lap_state == IB_CM_LAP_UNINIT) { + *qp_attr_mask = IB_QP_STATE | IB_QP_SQ_PSN; + qp_attr->sq_psn = be32_to_cpu(cm_id_priv->sq_psn); + switch (cm_id_priv->qp_type) { + case IB_QPT_RC: + case IB_QPT_XRC_INI: + *qp_attr_mask |= IB_QP_RETRY_CNT | IB_QP_RNR_RETRY | + IB_QP_MAX_QP_RD_ATOMIC; + qp_attr->retry_cnt = cm_id_priv->retry_count; + qp_attr->rnr_retry = cm_id_priv->rnr_retry_count; + qp_attr->max_rd_atomic = cm_id_priv->initiator_depth; + /* fall through */ + case IB_QPT_XRC_TGT: + *qp_attr_mask |= IB_QP_TIMEOUT; + qp_attr->timeout = cm_id_priv->av.timeout; + break; + default: + break; + } + if (cm_id_priv->alt_av.ah_attr.dlid) { + *qp_attr_mask |= IB_QP_PATH_MIG_STATE; + qp_attr->path_mig_state = IB_MIG_REARM; + } + } else { + *qp_attr_mask = IB_QP_ALT_PATH | IB_QP_PATH_MIG_STATE; + qp_attr->alt_port_num = cm_id_priv->alt_av.port->port_num; + qp_attr->alt_pkey_index = cm_id_priv->alt_av.pkey_index; + qp_attr->alt_timeout = cm_id_priv->alt_av.timeout; + qp_attr->alt_ah_attr = cm_id_priv->alt_av.ah_attr; qp_attr->path_mig_state = IB_MIG_REARM; } ret = 0; @@ -3163,7 +3648,106 @@ int ib_cm_init_qp_attr(struct ib_cm_id *cm_id, } EXPORT_SYMBOL(ib_cm_init_qp_attr); -static void cm_add_one(struct ib_device *device) +static void cm_get_ack_delay(struct cm_device *cm_dev) +{ + struct ib_device_attr attr; + + if (ib_query_device(cm_dev->ib_device, &attr)) + cm_dev->ack_delay = 0; /* acks will rely on packet life time */ + else + cm_dev->ack_delay = attr.local_ca_ack_delay; +} + +static ssize_t cm_show_counter(struct kobject *obj, struct attribute *attr, + char *buf) +{ + struct cm_counter_group *group; + struct cm_counter_attribute *cm_attr; + + group = container_of(obj, struct cm_counter_group, obj); + cm_attr = container_of(attr, struct cm_counter_attribute, attr); + + return sprintf(buf, "%ld\n", + atomic_long_read(&group->counter[cm_attr->index])); +} + +static const struct sysfs_ops cm_counter_ops = { + .show = cm_show_counter +}; + +static struct kobj_type cm_counter_obj_type = { + .sysfs_ops = &cm_counter_ops, + .default_attrs = cm_counter_default_attrs +}; + +static void cm_release_port_obj(struct kobject *obj) +{ + struct cm_port *cm_port; + + cm_port = container_of(obj, struct cm_port, port_obj); + kfree(cm_port); +} + +static struct kobj_type cm_port_obj_type = { + .release = cm_release_port_obj +}; + +static char *cm_devnode(struct device *dev, umode_t *mode) +{ + if (mode) + *mode = 0666; + return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); +} + +struct class cm_class = { + .owner = THIS_MODULE, + .name = "infiniband_cm", + .devnode = cm_devnode, +}; +EXPORT_SYMBOL(cm_class); + +static int cm_create_port_fs(struct cm_port *port) +{ + int i, ret; + + ret = kobject_init_and_add(&port->port_obj, &cm_port_obj_type, + &port->cm_dev->device->kobj, + "%d", port->port_num); + if (ret) { + kfree(port); + return ret; + } + + for (i = 0; i < CM_COUNTER_GROUPS; i++) { + ret = kobject_init_and_add(&port->counter_group[i].obj, + &cm_counter_obj_type, + &port->port_obj, + "%s", counter_group_names[i]); + if (ret) + goto error; + } + + return 0; + +error: + while (i--) + kobject_put(&port->counter_group[i].obj); + kobject_put(&port->port_obj); + return ret; + +} + +static void cm_remove_port_fs(struct cm_port *port) +{ + int i; + + for (i = 0; i < CM_COUNTER_GROUPS; i++) + kobject_put(&port->counter_group[i].obj); + + kobject_put(&port->port_obj); +} + +static void cm_add_one(struct ib_device *ib_device) { struct cm_device *cm_dev; struct cm_port *port; @@ -3178,20 +3762,40 @@ static void cm_add_one(struct ib_device *device) int ret; u8 i; - cm_dev = kmalloc(sizeof(*cm_dev) + sizeof(*port) * - device->phys_port_cnt, GFP_KERNEL); + if (rdma_node_get_transport(ib_device->node_type) != RDMA_TRANSPORT_IB) + return; + + cm_dev = kzalloc(sizeof(*cm_dev) + sizeof(*port) * + ib_device->phys_port_cnt, GFP_KERNEL); if (!cm_dev) return; - cm_dev->device = device; - cm_dev->ca_guid = device->node_guid; + cm_dev->ib_device = ib_device; + cm_get_ack_delay(cm_dev); + + cm_dev->device = device_create(&cm_class, &ib_device->dev, + MKDEV(0, 0), NULL, + "%s", ib_device->name); + if (IS_ERR(cm_dev->device)) { + kfree(cm_dev); + return; + } set_bit(IB_MGMT_METHOD_SEND, reg_req.method_mask); - for (i = 1; i <= device->phys_port_cnt; i++) { - port = &cm_dev->port[i-1]; + for (i = 1; i <= ib_device->phys_port_cnt; i++) { + port = kzalloc(sizeof *port, GFP_KERNEL); + if (!port) + goto error1; + + cm_dev->port[i-1] = port; port->cm_dev = cm_dev; port->port_num = i; - port->mad_agent = ib_register_mad_agent(device, i, + + ret = cm_create_port_fs(port); + if (ret) + goto error1; + + port->mad_agent = ib_register_mad_agent(ib_device, i, IB_QPT_GSI, ®_req, 0, @@ -3199,33 +3803,37 @@ static void cm_add_one(struct ib_device *device) cm_recv_handler, port); if (IS_ERR(port->mad_agent)) - goto error1; + goto error2; - ret = ib_modify_port(device, i, 0, &port_modify); + ret = ib_modify_port(ib_device, i, 0, &port_modify); if (ret) - goto error2; + goto error3; } - ib_set_client_data(device, &cm_client, cm_dev); + ib_set_client_data(ib_device, &cm_client, cm_dev); write_lock_irqsave(&cm.device_lock, flags); list_add_tail(&cm_dev->list, &cm.device_list); write_unlock_irqrestore(&cm.device_lock, flags); return; -error2: +error3: ib_unregister_mad_agent(port->mad_agent); +error2: + cm_remove_port_fs(port); error1: port_modify.set_port_cap_mask = 0; port_modify.clr_port_cap_mask = IB_PORT_CM_SUP; while (--i) { - port = &cm_dev->port[i-1]; - ib_modify_port(device, port->port_num, 0, &port_modify); + port = cm_dev->port[i-1]; + ib_modify_port(ib_device, port->port_num, 0, &port_modify); ib_unregister_mad_agent(port->mad_agent); + cm_remove_port_fs(port); } + device_unregister(cm_dev->device); kfree(cm_dev); } -static void cm_remove_one(struct ib_device *device) +static void cm_remove_one(struct ib_device *ib_device) { struct cm_device *cm_dev; struct cm_port *port; @@ -3235,7 +3843,7 @@ static void cm_remove_one(struct ib_device *device) unsigned long flags; int i; - cm_dev = ib_get_client_data(device, &cm_client); + cm_dev = ib_get_client_data(ib_device, &cm_client); if (!cm_dev) return; @@ -3243,11 +3851,14 @@ static void cm_remove_one(struct ib_device *device) list_del(&cm_dev->list); write_unlock_irqrestore(&cm.device_lock, flags); - for (i = 1; i <= device->phys_port_cnt; i++) { - port = &cm_dev->port[i-1]; - ib_modify_port(device, port->port_num, 0, &port_modify); + for (i = 1; i <= ib_device->phys_port_cnt; i++) { + port = cm_dev->port[i-1]; + ib_modify_port(ib_device, port->port_num, 0, &port_modify); ib_unregister_mad_agent(port->mad_agent); + flush_workqueue(cm.wq); + cm_remove_port_fs(port); } + device_unregister(cm_dev->device); kfree(cm_dev); } @@ -3260,32 +3871,58 @@ static int __init ib_cm_init(void) rwlock_init(&cm.device_lock); spin_lock_init(&cm.lock); cm.listen_service_table = RB_ROOT; - cm.listen_service_id = __constant_be64_to_cpu(IB_CM_ASSIGN_SERVICE_ID); + cm.listen_service_id = be64_to_cpu(IB_CM_ASSIGN_SERVICE_ID); cm.remote_id_table = RB_ROOT; cm.remote_qp_table = RB_ROOT; cm.remote_sidr_table = RB_ROOT; idr_init(&cm.local_id_table); - idr_pre_get(&cm.local_id_table, GFP_KERNEL); + get_random_bytes(&cm.random_id_operand, sizeof cm.random_id_operand); + INIT_LIST_HEAD(&cm.timewait_list); + + ret = class_register(&cm_class); + if (ret) { + ret = -ENOMEM; + goto error1; + } cm.wq = create_workqueue("ib_cm"); - if (!cm.wq) - return -ENOMEM; + if (!cm.wq) { + ret = -ENOMEM; + goto error2; + } ret = ib_register_client(&cm_client); if (ret) - goto error; + goto error3; return 0; -error: +error3: destroy_workqueue(cm.wq); +error2: + class_unregister(&cm_class); +error1: + idr_destroy(&cm.local_id_table); return ret; } static void __exit ib_cm_cleanup(void) { - flush_workqueue(cm.wq); - destroy_workqueue(cm.wq); + struct cm_timewait_info *timewait_info, *tmp; + + spin_lock_irq(&cm.lock); + list_for_each_entry(timewait_info, &cm.timewait_list, list) + cancel_delayed_work(&timewait_info->work.work); + spin_unlock_irq(&cm.lock); + ib_unregister_client(&cm_client); + destroy_workqueue(cm.wq); + + list_for_each_entry_safe(timewait_info, tmp, &cm.timewait_list, list) { + list_del(&timewait_info->list); + kfree(timewait_info); + } + + class_unregister(&cm_class); idr_destroy(&cm.local_id_table); } |