diff options
50 files changed, 1285 insertions, 955 deletions
diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c index 53343ffbff7a..cb00d59da456 100644 --- a/drivers/infiniband/core/cache.c +++ b/drivers/infiniband/core/cache.c @@ -1043,8 +1043,8 @@ static void ib_cache_update(struct ib_device *device, ret = ib_query_port(device, port, tprops); if (ret) { - printk(KERN_WARNING "ib_query_port failed (%d) for %s\n", - ret, device->name); + pr_warn("ib_query_port failed (%d) for %s\n", + ret, device->name); goto err; } @@ -1067,8 +1067,8 @@ static void ib_cache_update(struct ib_device *device, for (i = 0; i < pkey_cache->table_len; ++i) { ret = ib_query_pkey(device, port, i, pkey_cache->table + i); if (ret) { - printk(KERN_WARNING "ib_query_pkey failed (%d) for %s (index %d)\n", - ret, device->name, i); + pr_warn("ib_query_pkey failed (%d) for %s (index %d)\n", + ret, device->name, i); goto err; } } @@ -1078,8 +1078,8 @@ static void ib_cache_update(struct ib_device *device, ret = ib_query_gid(device, port, i, gid_cache->table + i, NULL); if (ret) { - printk(KERN_WARNING "ib_query_gid failed (%d) for %s (index %d)\n", - ret, device->name, i); + pr_warn("ib_query_gid failed (%d) for %s (index %d)\n", + ret, device->name, i); goto err; } } @@ -1161,8 +1161,7 @@ int ib_cache_setup_one(struct ib_device *device) GFP_KERNEL); if (!device->cache.pkey_cache || !device->cache.lmc_cache) { - printk(KERN_WARNING "Couldn't allocate cache " - "for %s\n", device->name); + pr_warn("Couldn't allocate cache for %s\n", device->name); return -ENOMEM; } diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 7eace1ffa74d..93ab0ae97208 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -1721,7 +1721,7 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) event.param.conn.private_data_len = IB_CM_REJ_PRIVATE_DATA_SIZE; break; default: - printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d\n", + pr_err("RDMA CMA: unexpected IB CM event: %d\n", ib_event->event); goto out; } @@ -2194,8 +2194,8 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv, ret = rdma_listen(id, id_priv->backlog); if (ret) - printk(KERN_WARNING "RDMA CMA: cma_listen_on_dev, error %d, " - "listening on device %s\n", ret, cma_dev->device->name); + pr_warn("RDMA CMA: cma_listen_on_dev, error %d, listening on device %s\n", + ret, cma_dev->device->name); } static void cma_listen_on_all(struct rdma_id_private *id_priv) @@ -3247,7 +3247,7 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id, event.status = 0; break; default: - printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d\n", + pr_err("RDMA CMA: unexpected IB CM event: %d\n", ib_event->event); goto out; } @@ -4011,8 +4011,8 @@ static int cma_netdev_change(struct net_device *ndev, struct rdma_id_private *id if ((dev_addr->bound_dev_if == ndev->ifindex) && (net_eq(dev_net(ndev), dev_addr->net)) && memcmp(dev_addr->src_dev_addr, ndev->dev_addr, ndev->addr_len)) { - printk(KERN_INFO "RDMA CM addr change for ndev %s used by id %p\n", - ndev->name, &id_priv->id); + pr_info("RDMA CM addr change for ndev %s used by id %p\n", + ndev->name, &id_priv->id); work = kzalloc(sizeof *work, GFP_KERNEL); if (!work) return -ENOMEM; @@ -4295,7 +4295,7 @@ static int __init cma_init(void) goto err; if (ibnl_add_client(RDMA_NL_RDMA_CM, RDMA_NL_RDMA_CM_NUM_OPS, cma_cb_table)) - printk(KERN_WARNING "RDMA CMA: failed to add netlink callback\n"); + pr_warn("RDMA CMA: failed to add netlink callback\n"); cma_configfs_init(); return 0; diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index 00da80e02154..270c7ff6cba7 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -115,8 +115,8 @@ static int ib_device_check_mandatory(struct ib_device *device) for (i = 0; i < ARRAY_SIZE(mandatory_table); ++i) { if (!*(void **) ((void *) device + mandatory_table[i].offset)) { - printk(KERN_WARNING "Device %s is missing mandatory function %s\n", - device->name, mandatory_table[i].name); + pr_warn("Device %s is missing mandatory function %s\n", + device->name, mandatory_table[i].name); return -EINVAL; } } @@ -255,8 +255,8 @@ static int add_client_context(struct ib_device *device, struct ib_client *client context = kmalloc(sizeof *context, GFP_KERNEL); if (!context) { - printk(KERN_WARNING "Couldn't allocate client context for %s/%s\n", - device->name, client->name); + pr_warn("Couldn't allocate client context for %s/%s\n", + device->name, client->name); return -ENOMEM; } @@ -343,28 +343,29 @@ int ib_register_device(struct ib_device *device, ret = read_port_immutable(device); if (ret) { - printk(KERN_WARNING "Couldn't create per port immutable data %s\n", - device->name); + pr_warn("Couldn't create per port immutable data %s\n", + device->name); goto out; } ret = ib_cache_setup_one(device); if (ret) { - printk(KERN_WARNING "Couldn't set up InfiniBand P_Key/GID cache\n"); + pr_warn("Couldn't set up InfiniBand P_Key/GID cache\n"); goto out; } memset(&device->attrs, 0, sizeof(device->attrs)); ret = device->query_device(device, &device->attrs, &uhw); if (ret) { - printk(KERN_WARNING "Couldn't query the device attributes\n"); + pr_warn("Couldn't query the device attributes\n"); + ib_cache_cleanup_one(device); goto out; } ret = ib_device_register_sysfs(device, port_callback); if (ret) { - printk(KERN_WARNING "Couldn't register device %s with driver model\n", - device->name); + pr_warn("Couldn't register device %s with driver model\n", + device->name); ib_cache_cleanup_one(device); goto out; } @@ -565,8 +566,8 @@ void ib_set_client_data(struct ib_device *device, struct ib_client *client, goto out; } - printk(KERN_WARNING "No client context found for %s/%s\n", - device->name, client->name); + pr_warn("No client context found for %s/%s\n", + device->name, client->name); out: spin_unlock_irqrestore(&device->client_data_lock, flags); @@ -959,13 +960,13 @@ static int __init ib_core_init(void) ret = class_register(&ib_class); if (ret) { - printk(KERN_WARNING "Couldn't create InfiniBand device class\n"); + pr_warn("Couldn't create InfiniBand device class\n"); goto err_comp; } ret = ibnl_init(); if (ret) { - printk(KERN_WARNING "Couldn't init IB netlink interface\n"); + pr_warn("Couldn't init IB netlink interface\n"); goto err_sysfs; } diff --git a/drivers/infiniband/core/fmr_pool.c b/drivers/infiniband/core/fmr_pool.c index 6ac3683c144b..cdbb1f1a6d97 100644 --- a/drivers/infiniband/core/fmr_pool.c +++ b/drivers/infiniband/core/fmr_pool.c @@ -150,8 +150,8 @@ static void ib_fmr_batch_release(struct ib_fmr_pool *pool) #ifdef DEBUG if (fmr->ref_count !=0) { - printk(KERN_WARNING PFX "Unmapping FMR 0x%08x with ref count %d\n", - fmr, fmr->ref_count); + pr_warn(PFX "Unmapping FMR 0x%08x with ref count %d\n", + fmr, fmr->ref_count); } #endif } @@ -167,7 +167,7 @@ static void ib_fmr_batch_release(struct ib_fmr_pool *pool) ret = ib_unmap_fmr(&fmr_list); if (ret) - printk(KERN_WARNING PFX "ib_unmap_fmr returned %d\n", ret); + pr_warn(PFX "ib_unmap_fmr returned %d\n", ret); spin_lock_irq(&pool->pool_lock); list_splice(&unmap_list, &pool->free_list); @@ -222,8 +222,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd *pd, device = pd->device; if (!device->alloc_fmr || !device->dealloc_fmr || !device->map_phys_fmr || !device->unmap_fmr) { - printk(KERN_INFO PFX "Device %s does not support FMRs\n", - device->name); + pr_info(PFX "Device %s does not support FMRs\n", device->name); return ERR_PTR(-ENOSYS); } @@ -233,13 +232,10 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd *pd, max_remaps = device->attrs.max_map_per_fmr; pool = kmalloc(sizeof *pool, GFP_KERNEL); - if (!pool) { - printk(KERN_WARNING PFX "couldn't allocate pool struct\n"); + if (!pool) return ERR_PTR(-ENOMEM); - } pool->cache_bucket = NULL; - pool->flush_function = params->flush_function; pool->flush_arg = params->flush_arg; @@ -251,7 +247,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd *pd, kmalloc(IB_FMR_HASH_SIZE * sizeof *pool->cache_bucket, GFP_KERNEL); if (!pool->cache_bucket) { - printk(KERN_WARNING PFX "Failed to allocate cache in pool\n"); + pr_warn(PFX "Failed to allocate cache in pool\n"); ret = -ENOMEM; goto out_free_pool; } @@ -275,7 +271,7 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd *pd, "ib_fmr(%s)", device->name); if (IS_ERR(pool->thread)) { - printk(KERN_WARNING PFX "couldn't start cleanup thread\n"); + pr_warn(PFX "couldn't start cleanup thread\n"); ret = PTR_ERR(pool->thread); goto out_free_pool; } @@ -294,11 +290,8 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd *pd, for (i = 0; i < params->pool_size; ++i) { fmr = kmalloc(bytes_per_fmr, GFP_KERNEL); - if (!fmr) { - printk(KERN_WARNING PFX "failed to allocate fmr " - "struct for FMR %d\n", i); + if (!fmr) goto out_fail; - } fmr->pool = pool; fmr->remap_count = 0; @@ -307,8 +300,8 @@ struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd *pd, fmr->fmr = ib_alloc_fmr(pd, params->access, &fmr_attr); if (IS_ERR(fmr->fmr)) { - printk(KERN_WARNING PFX "fmr_create failed " - "for FMR %d\n", i); + pr_warn(PFX "fmr_create failed for FMR %d\n", + i); kfree(fmr); goto out_fail; } @@ -363,8 +356,8 @@ void ib_destroy_fmr_pool(struct ib_fmr_pool *pool) } if (i < pool->pool_size) - printk(KERN_WARNING PFX "pool still has %d regions registered\n", - pool->pool_size - i); + pr_warn(PFX "pool still has %d regions registered\n", + pool->pool_size - i); kfree(pool->cache_bucket); kfree(pool); @@ -463,7 +456,7 @@ struct ib_pool_fmr *ib_fmr_pool_map_phys(struct ib_fmr_pool *pool_handle, list_add(&fmr->list, &pool->free_list); spin_unlock_irqrestore(&pool->pool_lock, flags); - printk(KERN_WARNING PFX "fmr_map returns %d\n", result); + pr_warn(PFX "fmr_map returns %d\n", result); return ERR_PTR(result); } @@ -517,8 +510,8 @@ int ib_fmr_pool_unmap(struct ib_pool_fmr *fmr) #ifdef DEBUG if (fmr->ref_count < 0) - printk(KERN_WARNING PFX "FMR %p has ref count %d < 0\n", - fmr, fmr->ref_count); + pr_warn(PFX "FMR %p has ref count %d < 0\n", + fmr, fmr->ref_count); #endif spin_unlock_irqrestore(&pool->pool_lock, flags); diff --git a/drivers/infiniband/core/packer.c b/drivers/infiniband/core/packer.c index 1b65986c0be3..19b1ee3279b4 100644 --- a/drivers/infiniband/core/packer.c +++ b/drivers/infiniband/core/packer.c @@ -44,7 +44,7 @@ static u64 value_read(int offset, int size, void *structure) case 4: return be32_to_cpup((__be32 *) (structure + offset)); case 8: return be64_to_cpup((__be64 *) (structure + offset)); default: - printk(KERN_WARNING "Field size %d bits not handled\n", size * 8); + pr_warn("Field size %d bits not handled\n", size * 8); return 0; } } @@ -104,9 +104,8 @@ void ib_pack(const struct ib_field *desc, } else { if (desc[i].offset_bits % 8 || desc[i].size_bits % 8) { - printk(KERN_WARNING "Structure field %s of size %d " - "bits is not byte-aligned\n", - desc[i].field_name, desc[i].size_bits); + pr_warn("Structure field %s of size %d bits is not byte-aligned\n", + desc[i].field_name, desc[i].size_bits); } if (desc[i].struct_size_bytes) @@ -132,7 +131,7 @@ static void value_write(int offset, int size, u64 val, void *structure) case 32: *(__be32 *) (structure + offset) = cpu_to_be32(val); break; case 64: *(__be64 *) (structure + offset) = cpu_to_be64(val); break; default: - printk(KERN_WARNING "Field size %d bits not handled\n", size * 8); + pr_warn("Field size %d bits not handled\n", size * 8); } } @@ -188,9 +187,8 @@ void ib_unpack(const struct ib_field *desc, } else { if (desc[i].offset_bits % 8 || desc[i].size_bits % 8) { - printk(KERN_WARNING "Structure field %s of size %d " - "bits is not byte-aligned\n", - desc[i].field_name, desc[i].size_bits); + pr_warn("Structure field %s of size %d bits is not byte-aligned\n", + desc[i].field_name, desc[i].size_bits); } memcpy(structure + desc[i].struct_offset_bytes, diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index f334090bb612..8e3bf6c8d3c3 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -864,13 +864,12 @@ static void update_sm_ah(struct work_struct *work) struct ib_ah_attr ah_attr; if (ib_query_port(port->agent->device, port->port_num, &port_attr)) { - printk(KERN_WARNING "Couldn't query port\n"); + pr_warn("Couldn't query port\n"); return; } new_ah = kmalloc(sizeof *new_ah, GFP_KERNEL); if (!new_ah) { - printk(KERN_WARNING "Couldn't allocate new SM AH\n"); return; } @@ -880,7 +879,7 @@ static void update_sm_ah(struct work_struct *work) new_ah->pkey_index = 0; if (ib_find_pkey(port->agent->device, port->port_num, IB_DEFAULT_PKEY_FULL, &new_ah->pkey_index)) - printk(KERN_ERR "Couldn't find index for default PKey\n"); + pr_err("Couldn't find index for default PKey\n"); memset(&ah_attr, 0, sizeof ah_attr); ah_attr.dlid = port_attr.sm_lid; @@ -889,7 +888,7 @@ static void update_sm_ah(struct work_struct *work) new_ah->ah = ib_create_ah(port->agent->qp->pd, &ah_attr); if (IS_ERR(new_ah->ah)) { - printk(KERN_WARNING "Couldn't create new SM AH\n"); + pr_warn("Couldn't create new SM AH\n"); kfree(new_ah); return; } @@ -1221,7 +1220,7 @@ static void ib_sa_path_rec_callback(struct ib_sa_query *sa_query, rec.net = NULL; rec.ifindex = 0; rec.gid_type = IB_GID_TYPE_IB; - memset(rec.dmac, 0, ETH_ALEN); + eth_zero_addr(rec.dmac); query->callback(status, &rec, query->context); } else query->callback(status, NULL, query->context); @@ -1800,13 +1799,13 @@ static int __init ib_sa_init(void) ret = ib_register_client(&sa_client); if (ret) { - printk(KERN_ERR "Couldn't register ib_sa client\n"); + pr_err("Couldn't register ib_sa client\n"); goto err1; } ret = mcast_init(); if (ret) { - printk(KERN_ERR "Couldn't initialize multicast handling\n"); + pr_err("Couldn't initialize multicast handling\n"); goto err2; } diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c index 6b4e8a008bc0..4a9aa0433b07 100644 --- a/drivers/infiniband/core/ucm.c +++ b/drivers/infiniband/core/ucm.c @@ -1234,7 +1234,7 @@ static int find_overflow_devnum(void) ret = alloc_chrdev_region(&overflow_maj, 0, IB_UCM_MAX_DEVICES, "infiniband_cm"); if (ret) { - printk(KERN_ERR "ucm: couldn't register dynamic device number\n"); + pr_err("ucm: couldn't register dynamic device number\n"); return ret; } } @@ -1329,19 +1329,19 @@ static int __init ib_ucm_init(void) ret = register_chrdev_region(IB_UCM_BASE_DEV, IB_UCM_MAX_DEVICES, "infiniband_cm"); if (ret) { - printk(KERN_ERR "ucm: couldn't register device number\n"); + pr_err("ucm: couldn't register device number\n"); goto error1; } ret = class_create_file(&cm_class, &class_attr_abi_version.attr); if (ret) { - printk(KERN_ERR "ucm: couldn't create abi_version attribute\n"); + pr_err("ucm: couldn't create abi_version attribute\n"); goto error2; } ret = ib_register_client(&ucm_client); if (ret) { - printk(KERN_ERR "ucm: couldn't register client\n"); + pr_err("ucm: couldn't register client\n"); goto error3; } return 0; diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c index 8b5a934e1133..dd3bcceadfde 100644 --- a/drivers/infiniband/core/ucma.c +++ b/drivers/infiniband/core/ucma.c @@ -314,7 +314,7 @@ static void ucma_removal_event_handler(struct rdma_cm_id *cm_id) } } if (!event_found) - printk(KERN_ERR "ucma_removal_event_handler: warning: connect request event wasn't found\n"); + pr_err("ucma_removal_event_handler: warning: connect request event wasn't found\n"); } static int ucma_event_handler(struct rdma_cm_id *cm_id, @@ -1716,13 +1716,13 @@ static int __init ucma_init(void) ret = device_create_file(ucma_misc.this_device, &dev_attr_abi_version); if (ret) { - printk(KERN_ERR "rdma_ucm: couldn't create abi_version attr\n"); + pr_err("rdma_ucm: couldn't create abi_version attr\n"); goto err1; } ucma_ctl_table_hdr = register_net_sysctl(&init_net, "net/rdma_ucm", ucma_ctl_table); if (!ucma_ctl_table_hdr) { - printk(KERN_ERR "rdma_ucm: couldn't register sysctl paths\n"); + pr_err("rdma_ucm: couldn't register sysctl paths\n"); ret = -ENOMEM; goto err2; } diff --git a/drivers/infiniband/core/ud_header.c b/drivers/infiniband/core/ud_header.c index 2116132568e7..29a45d2f8898 100644 --- a/drivers/infiniband/core/ud_header.c +++ b/drivers/infiniband/core/ud_header.c @@ -479,8 +479,8 @@ int ib_ud_header_unpack(void *buf, buf += IB_LRH_BYTES; if (header->lrh.link_version != 0) { - printk(KERN_WARNING "Invalid LRH.link_version %d\n", - header->lrh.link_version); + pr_warn("Invalid LRH.link_version %d\n", + header->lrh.link_version); return -EINVAL; } @@ -496,20 +496,20 @@ int ib_ud_header_unpack(void *buf, buf += IB_GRH_BYTES; if (header->grh.ip_version != 6) { - printk(KERN_WARNING "Invalid GRH.ip_version %d\n", - header->grh.ip_version); + pr_warn("Invalid GRH.ip_version %d\n", + header->grh.ip_version); return -EINVAL; } if (header->grh.next_header != 0x1b) { - printk(KERN_WARNING "Invalid GRH.next_header 0x%02x\n", - header->grh.next_header); + pr_warn("Invalid GRH.next_header 0x%02x\n", + header->grh.next_header); return -EINVAL; } break; default: - printk(KERN_WARNING "Invalid LRH.link_next_header %d\n", - header->lrh.link_next_header); + pr_warn("Invalid LRH.link_next_header %d\n", + header->lrh.link_next_header); return -EINVAL; } @@ -525,14 +525,13 @@ int ib_ud_header_unpack(void *buf, header->immediate_present = 1; break; default: - printk(KERN_WARNING "Invalid BTH.opcode 0x%02x\n", - header->bth.opcode); + pr_warn("Invalid BTH.opcode 0x%02x\n", header->bth.opcode); return -EINVAL; } if (header->bth.transport_header_version != 0) { - printk(KERN_WARNING "Invalid BTH.transport_header_version %d\n", - header->bth.transport_header_version); + pr_warn("Invalid BTH.transport_header_version %d\n", + header->bth.transport_header_version); return -EINVAL; } diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 2bf751ecad15..3638c787cb7c 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -1976,7 +1976,8 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file, resp_size); INIT_UDATA(&uhw, buf + sizeof(cmd), (unsigned long)cmd.response + resp_size, - in_len - sizeof(cmd), out_len - resp_size); + in_len - sizeof(cmd) - sizeof(struct ib_uverbs_cmd_hdr), + out_len - resp_size); memset(&cmd_ex, 0, sizeof(cmd_ex)); cmd_ex.user_handle = cmd.user_handle; @@ -3091,6 +3092,14 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file, !capable(CAP_NET_ADMIN)) || !capable(CAP_NET_RAW)) return -EPERM; + if (cmd.flow_attr.flags >= IB_FLOW_ATTR_FLAGS_RESERVED) + return -EINVAL; + + if ((cmd.flow_attr.flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP) && + ((cmd.flow_attr.type == IB_FLOW_ATTR_ALL_DEFAULT) || + (cmd.flow_attr.type == IB_FLOW_ATTR_MC_DEFAULT))) + return -EINVAL; + if (cmd.flow_attr.num_of_specs > IB_FLOW_SPEC_SUPPORT_LAYERS) return -EINVAL; @@ -3419,7 +3428,8 @@ ssize_t ib_uverbs_create_srq(struct ib_uverbs_file *file, INIT_UDATA(&udata, buf + sizeof cmd, (unsigned long) cmd.response + sizeof resp, - in_len - sizeof cmd, out_len - sizeof resp); + in_len - sizeof cmd - sizeof(struct ib_uverbs_cmd_hdr), + out_len - sizeof resp); ret = __uverbs_create_xsrq(file, ib_dev, &xcmd, &udata); if (ret) @@ -3445,7 +3455,8 @@ ssize_t ib_uverbs_create_xsrq(struct ib_uverbs_file *file, INIT_UDATA(&udata, buf + sizeof cmd, (unsigned long) cmd.response + sizeof resp, - in_len - sizeof cmd, out_len - sizeof resp); + in_len - sizeof cmd - sizeof(struct ib_uverbs_cmd_hdr), + out_len - sizeof resp); ret = __uverbs_create_xsrq(file, ib_dev, &cmd, &udata); if (ret) diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index 39680aed99dd..28ba2cc81535 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -683,12 +683,28 @@ out: return ev_file; } +static int verify_command_mask(struct ib_device *ib_dev, __u32 command) +{ + u64 mask; + + if (command <= IB_USER_VERBS_CMD_OPEN_QP) + mask = ib_dev->uverbs_cmd_mask; + else + mask = ib_dev->uverbs_ex_cmd_mask; + + if (mask & ((u64)1 << command)) + return 0; + + return -1; +} + static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, size_t count, loff_t *pos) { struct ib_uverbs_file *file = filp->private_data; struct ib_device *ib_dev; struct ib_uverbs_cmd_hdr hdr; + __u32 command; __u32 flags; int srcu_key; ssize_t ret; @@ -707,37 +723,34 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, goto out; } - flags = (hdr.command & - IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT; + if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK | + IB_USER_VERBS_CMD_COMMAND_MASK)) { + ret = -EINVAL; + goto out; + } - if (!flags) { - __u32 command; + command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK; + if (verify_command_mask(ib_dev, command)) { + ret = -EOPNOTSUPP; + goto out; + } - if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK | - IB_USER_VERBS_CMD_COMMAND_MASK)) { - ret = -EINVAL; - goto out; - } + if (!file->ucontext && + command != IB_USER_VERBS_CMD_GET_CONTEXT) { + ret = -EINVAL; + goto out; + } - command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK; + flags = (hdr.command & + IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT; + if (!flags) { if (command >= ARRAY_SIZE(uverbs_cmd_table) || !uverbs_cmd_table[command]) { ret = -EINVAL; goto out; } - if (!file->ucontext && - command != IB_USER_VERBS_CMD_GET_CONTEXT) { - ret = -EINVAL; - goto out; - } - - if (!(ib_dev->uverbs_cmd_mask & (1ull << command))) { - ret = -ENOSYS; - goto out; - } - if (hdr.in_words * 4 != count) { ret = -EINVAL; goto out; @@ -749,21 +762,11 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, hdr.out_words * 4); } else if (flags == IB_USER_VERBS_CMD_FLAG_EXTENDED) { - __u32 command; - struct ib_uverbs_ex_cmd_hdr ex_hdr; struct ib_udata ucore; struct ib_udata uhw; size_t written_count = count; - if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK | - IB_USER_VERBS_CMD_COMMAND_MASK)) { - ret = -EINVAL; - goto out; - } - - command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK; - if (command >= ARRAY_SIZE(uverbs_ex_cmd_table) || !uverbs_ex_cmd_table[command]) { ret = -ENOSYS; @@ -775,11 +778,6 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, goto out; } - if (!(ib_dev->uverbs_ex_cmd_mask & (1ull << command))) { - ret = -ENOSYS; - goto out; - } - if (count < (sizeof(hdr) + sizeof(ex_hdr))) { ret = -EINVAL; goto out; @@ -1058,7 +1056,7 @@ static int find_overflow_devnum(void) ret = alloc_chrdev_region(&overflow_maj, 0, IB_UVERBS_MAX_DEVICES, "infiniband_verbs"); if (ret) { - printk(KERN_ERR "user_verbs: couldn't register dynamic device number\n"); + pr_err("user_verbs: couldn't register dynamic device number\n"); return ret; } } @@ -1279,14 +1277,14 @@ static int __init ib_uverbs_init(void) ret = register_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES, "infiniband_verbs"); if (ret) { - printk(KERN_ERR "user_verbs: couldn't register device number\n"); + pr_err("user_verbs: couldn't register device number\n"); goto out; } uverbs_class = class_create(THIS_MODULE, "infiniband_verbs"); if (IS_ERR(uverbs_class)) { ret = PTR_ERR(uverbs_class); - printk(KERN_ERR "user_verbs: couldn't create class infiniband_verbs\n"); + pr_err("user_verbs: couldn't create class infiniband_verbs\n"); goto out_chrdev; } @@ -1294,13 +1292,13 @@ static int __init ib_uverbs_init(void) ret = class_create_file(uverbs_class, &class_attr_abi_version.attr); if (ret) { - printk(KERN_ERR "user_verbs: couldn't create abi_version attribute\n"); + pr_err("user_verbs: couldn't create abi_version attribute\n"); goto out_class; } ret = ib_register_client(&uverbs_client); if (ret) { - printk(KERN_ERR "user_verbs: couldn't register client\n"); + pr_err("user_verbs: couldn't register client\n"); goto out_class; } diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 16f3fb1b9d75..5cd1e3987f2b 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -1659,3 +1659,167 @@ next_page: return i; } EXPORT_SYMBOL(ib_sg_to_pages); + +struct ib_drain_cqe { + struct ib_cqe cqe; + struct completion done; +}; + +static void ib_drain_qp_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct ib_drain_cqe *cqe = container_of(wc->wr_cqe, struct ib_drain_cqe, + cqe); + + complete(&cqe->done); +} + +/* + * Post a WR and block until its completion is reaped for the SQ. + */ +static void __ib_drain_sq(struct ib_qp *qp) +{ + struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR }; + struct ib_drain_cqe sdrain; + struct ib_send_wr swr = {}, *bad_swr; + int ret; + + if (qp->send_cq->poll_ctx == IB_POLL_DIRECT) { + WARN_ONCE(qp->send_cq->poll_ctx == IB_POLL_DIRECT, + "IB_POLL_DIRECT poll_ctx not supported for drain\n"); + return; + } + + swr.wr_cqe = &sdrain.cqe; + sdrain.cqe.done = ib_drain_qp_done; + init_completion(&sdrain.done); + + ret = ib_modify_qp(qp, &attr, IB_QP_STATE); + if (ret) { + WARN_ONCE(ret, "failed to drain send queue: %d\n", ret); + return; + } + + ret = ib_post_send(qp, &swr, &bad_swr); + if (ret) { + WARN_ONCE(ret, "failed to drain send queue: %d\n", ret); + return; + } + + wait_for_completion(&sdrain.done); +} + +/* + * Post a WR and block until its completion is reaped for the RQ. + */ +static void __ib_drain_rq(struct ib_qp *qp) +{ + struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR }; + struct ib_drain_cqe rdrain; + struct ib_recv_wr rwr = {}, *bad_rwr; + int ret; + + if (qp->recv_cq->poll_ctx == IB_POLL_DIRECT) { + WARN_ONCE(qp->recv_cq->poll_ctx == IB_POLL_DIRECT, + "IB_POLL_DIRECT poll_ctx not supported for drain\n"); + return; + } + + rwr.wr_cqe = &rdrain.cqe; + rdrain.cqe.done = ib_drain_qp_done; + init_completion(&rdrain.done); + + ret = ib_modify_qp(qp, &attr, IB_QP_STATE); + if (ret) { + WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret); + return; + } + + ret = ib_post_recv(qp, &rwr, &bad_rwr); + if (ret) { + WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret); + return; + } + + wait_for_completion(&rdrain.done); +} + +/** + * ib_drain_sq() - Block until all SQ CQEs have been consumed by the + * application. + * @qp: queue pair to drain + * + * If the device has a provider-specific drain function, then + * call that. Otherwise call the generic drain function + * __ib_drain_sq(). + * + * The caller must: + * + * ensure there is room in the CQ and SQ for the drain work request and + * completion. + * + * allocate the CQ using ib_alloc_cq() and the CQ poll context cannot be + * IB_POLL_DIRECT. + * + * ensure that there are no other contexts that are posting WRs concurrently. + * Otherwise the drain is not guaranteed. + */ +void ib_drain_sq(struct ib_qp *qp) +{ + if (qp->device->drain_sq) + qp->device->drain_sq(qp); + else + __ib_drain_sq(qp); +} +EXPORT_SYMBOL(ib_drain_sq); + +/** + * ib_drain_rq() - Block until all RQ CQEs have been consumed by the + * application. + * @qp: queue pair to drain + * + * If the device has a provider-specific drain function, then + * call that. Otherwise call the generic drain function + * __ib_drain_rq(). + * + * The caller must: + * + * ensure there is room in the CQ and RQ for the drain work request and + * completion. + * + * allocate the CQ using ib_alloc_cq() and the CQ poll context cannot be + * IB_POLL_DIRECT. + * + * ensure that there are no other contexts that are posting WRs concurrently. + * Otherwise the drain is not guaranteed. + */ +void ib_drain_rq(struct ib_qp *qp) +{ + if (qp->device->drain_rq) + qp->device->drain_rq(qp); + else + __ib_drain_rq(qp); +} +EXPORT_SYMBOL(ib_drain_rq); + +/** + * ib_drain_qp() - Block until all CQEs have been consumed by the + * application on both the RQ and SQ. + * @qp: queue pair to drain + * + * The caller must: + * + * ensure there is room in the CQ(s), SQ, and RQ for drain work requests + * and completions. + * + * allocate the CQs using ib_alloc_cq() and the CQ poll context cannot be + * IB_POLL_DIRECT. + * + * ensure that there are no other contexts that are posting WRs concurrently. + * Otherwise the drain is not guaranteed. + */ +void ib_drain_qp(struct ib_qp *qp) +{ + ib_drain_sq(qp); + ib_drain_rq(qp); +} +EXPORT_SYMBOL(ib_drain_qp); diff --git a/drivers/infiniband/hw/cxgb4/cq.c b/drivers/infiniband/hw/cxgb4/cq.c index cf21df4a8bf5..b4eeb783573c 100644 --- a/drivers/infiniband/hw/cxgb4/cq.c +++ b/drivers/infiniband/hw/cxgb4/cq.c @@ -815,8 +815,15 @@ static int c4iw_poll_cq_one(struct c4iw_cq *chp, struct ib_wc *wc) } } out: - if (wq) + if (wq) { + if (unlikely(qhp->attr.state != C4IW_QP_STATE_RTS)) { + if (t4_sq_empty(wq)) + complete(&qhp->sq_drained); + if (t4_rq_empty(wq)) + complete(&qhp->rq_drained); + } spin_unlock(&qhp->lock); + } return ret; } diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h index 423a3a91bf41..97c0e5aa5c64 100644 --- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h +++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h @@ -476,6 +476,8 @@ struct c4iw_qp { wait_queue_head_t wait; struct timer_list timer; int sq_sig_all; + struct completion rq_drained; + struct completion sq_drained; }; static inline struct c4iw_qp *to_c4iw_qp(struct ib_qp *ibqp) @@ -1017,6 +1019,8 @@ extern int c4iw_wr_log; extern int db_fc_threshold; extern int db_coalescing_threshold; extern int use_dsgl; +void c4iw_drain_rq(struct ib_qp *qp); +void c4iw_drain_sq(struct ib_qp *qp); #endif diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c index ec04272fbdc2..104662d38d1e 100644 --- a/drivers/infiniband/hw/cxgb4/provider.c +++ b/drivers/infiniband/hw/cxgb4/provider.c @@ -564,6 +564,8 @@ int c4iw_register_device(struct c4iw_dev *dev) dev->ibdev.get_protocol_stats = c4iw_get_mib; dev->ibdev.uverbs_abi_ver = C4IW_UVERBS_ABI_VERSION; dev->ibdev.get_port_immutable = c4iw_port_immutable; + dev->ibdev.drain_sq = c4iw_drain_sq; + dev->ibdev.drain_rq = c4iw_drain_rq; dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL); if (!dev->ibdev.iwcm) diff --git a/drivers/infiniband/hw/cxgb4/qp.c b/drivers/infiniband/hw/cxgb4/qp.c index e99345eb875a..7b1b1e840ef1 100644 --- a/drivers/infiniband/hw/cxgb4/qp.c +++ b/drivers/infiniband/hw/cxgb4/qp.c @@ -1697,6 +1697,8 @@ struct ib_qp *c4iw_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attrs, qhp->attr.max_ird = 0; qhp->sq_sig_all = attrs->sq_sig_type == IB_SIGNAL_ALL_WR; spin_lock_init(&qhp->lock); + init_completion(&qhp->sq_drained); + init_completion(&qhp->rq_drained); mutex_init(&qhp->mutex); init_waitqueue_head(&qhp->wait); atomic_set(&qhp->refcnt, 1); @@ -1888,3 +1890,17 @@ int c4iw_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, init_attr->sq_sig_type = qhp->sq_sig_all ? IB_SIGNAL_ALL_WR : 0; return 0; } + +void c4iw_drain_sq(struct ib_qp *ibqp) +{ + struct c4iw_qp *qp = to_c4iw_qp(ibqp); + + wait_for_completion(&qp->sq_drained); +} + +void c4iw_drain_rq(struct ib_qp *ibqp) +{ + struct c4iw_qp *qp = to_c4iw_qp(ibqp); + + wait_for_completion(&qp->rq_drained); +} diff --git a/drivers/infiniband/hw/mlx4/alias_GUID.c b/drivers/infiniband/hw/mlx4/alias_GUID.c index 21cb41a60fe8..c74ef2620b85 100644 --- a/drivers/infiniband/hw/mlx4/alias_GUID.c +++ b/drivers/infiniband/hw/mlx4/alias_GUID.c @@ -310,7 +310,7 @@ static void aliasguid_query_handler(int status, if (status) { pr_debug("(port: %d) failed: status = %d\n", cb_ctx->port, status); - rec->time_to_run = ktime_get_real_ns() + 1 * NSEC_PER_SEC; + rec->time_to_run = ktime_get_boot_ns() + 1 * NSEC_PER_SEC; goto out; } @@ -416,7 +416,7 @@ next_entry: be64_to_cpu((__force __be64)rec->guid_indexes), be64_to_cpu((__force __be64)applied_guid_indexes), be64_to_cpu((__force __be64)declined_guid_indexes)); - rec->time_to_run = ktime_get_real_ns() + + rec->time_to_run = ktime_get_boot_ns() + resched_delay_sec * NSEC_PER_SEC; } else { rec->status = MLX4_GUID_INFO_STATUS_SET; @@ -708,7 +708,7 @@ static int get_low_record_time_index(struct mlx4_ib_dev *dev, u8 port, } } if (resched_delay_sec) { - u64 curr_time = ktime_get_real_ns(); + u64 curr_time = ktime_get_boot_ns(); *resched_delay_sec = (low_record_time < curr_time) ? 0 : div_u64((low_record_time - curr_time), NSEC_PER_SEC); diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 1c7ab6cabbb8..914bc98e753f 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -1643,6 +1643,56 @@ static int mlx4_ib_tunnel_steer_add(struct ib_qp *qp, struct ib_flow_attr *flow_ return err; } +static int mlx4_ib_add_dont_trap_rule(struct mlx4_dev *dev, + struct ib_flow_attr *flow_attr, + enum mlx4_net_trans_promisc_mode *type) +{ + int err = 0; + + if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_DMFS_UC_MC_SNIFFER) || + (dev->caps.dmfs_high_steer_mode == MLX4_STEERING_DMFS_A0_STATIC) || + (flow_attr->num_of_specs > 1) || (flow_attr->priority != 0)) { + return -EOPNOTSUPP; + } + + if (flow_attr->num_of_specs == 0) { + type[0] = MLX4_FS_MC_SNIFFER; + type[1] = MLX4_FS_UC_SNIFFER; + } else { + union ib_flow_spec *ib_spec; + + ib_spec = (union ib_flow_spec *)(flow_attr + 1); + if (ib_spec->type != IB_FLOW_SPEC_ETH) + return -EINVAL; + + /* if all is zero than MC and UC */ + if (is_zero_ether_addr(ib_spec->eth.mask.dst_mac)) { + type[0] = MLX4_FS_MC_SNIFFER; + type[1] = MLX4_FS_UC_SNIFFER; + } else { + u8 mac[ETH_ALEN] = {ib_spec->eth.mask.dst_mac[0] ^ 0x01, + ib_spec->eth.mask.dst_mac[1], + ib_spec->eth.mask.dst_mac[2], + ib_spec->eth.mask.dst_mac[3], + ib_spec->eth.mask.dst_mac[4], + ib_spec->eth.mask.dst_mac[5]}; + + /* Above xor was only on MC bit, non empty mask is valid + * only if this bit is set and rest are zero. + */ + if (!is_zero_ether_addr(&mac[0])) + return -EINVAL; + + if (is_multicast_ether_addr(ib_spec->eth.val.dst_mac)) + type[0] = MLX4_FS_MC_SNIFFER; + else + type[0] = MLX4_FS_UC_SNIFFER; + } + } + + return err; +} + static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp, struct ib_flow_attr *flow_attr, int domain) @@ -1653,6 +1703,10 @@ static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp, struct mlx4_dev *dev = (to_mdev(qp->device))->dev; int is_bonded = mlx4_is_bonded(dev); + if ((flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP) && + (flow_attr->type != IB_FLOW_ATTR_NORMAL)) + return ERR_PTR(-EOPNOTSUPP); + memset(type, 0, sizeof(type)); mflow = kzalloc(sizeof(*mflow), GFP_KERNEL); @@ -1663,7 +1717,19 @@ static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp, switch (flow_attr->type) { case IB_FLOW_ATTR_NORMAL: - type[0] = MLX4_FS_REGULAR; + /* If dont trap flag (continue match) is set, under specific + * condition traffic be replicated to given qp, + * without stealing it + */ + if (unlikely(flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP)) { + err = mlx4_ib_add_dont_trap_rule(dev, + flow_attr, + type); + if (err) + goto err_free; + } else { + type[0] = MLX4_FS_REGULAR; + } break; case IB_FLOW_ATTR_ALL_DEFAULT: @@ -1675,8 +1741,8 @@ static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp, break; case IB_FLOW_ATTR_SNIFFER: - type[0] = MLX4_FS_UC_SNIFFER; - type[1] = MLX4_FS_MC_SNIFFER; + type[0] = MLX4_FS_MIRROR_RX_PORT; + type[1] = MLX4_FS_MIRROR_SX_PORT; break; default: diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c index 7e89a547bf34..5afbb697e691 100644 --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -1382,11 +1382,20 @@ static int mlx5_ib_destroy_flow(struct ib_flow *flow_id) return 0; } +static int ib_prio_to_core_prio(unsigned int priority, bool dont_trap) +{ + priority *= 2; + if (!dont_trap) + priority++; + return priority; +} + #define MLX5_FS_MAX_TYPES 10 #define MLX5_FS_MAX_ENTRIES 32000UL static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev, struct ib_flow_attr *flow_attr) { + bool dont_trap = flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP; struct mlx5_flow_namespace *ns = NULL; struct mlx5_ib_flow_prio *prio; struct mlx5_flow_table *ft; @@ -1396,10 +1405,12 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev, int err = 0; if (flow_attr->type == IB_FLOW_ATTR_NORMAL) { - if (flow_is_multicast_only(flow_attr)) + if (flow_is_multicast_only(flow_attr) && + !dont_trap) priority = MLX5_IB_FLOW_MCAST_PRIO; else - priority = flow_attr->priority; + priority = ib_prio_to_core_prio(flow_attr->priority, + dont_trap); ns = mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS); num_entries = MLX5_FS_MAX_ENTRIES; @@ -1447,6 +1458,7 @@ static struct mlx5_ib_flow_handler *create_flow_rule(struct mlx5_ib_dev *dev, unsigned int spec_index; u32 *match_c; u32 *match_v; + u32 action; int err = 0; if (!is_valid_attr(flow_attr)) @@ -1472,9 +1484,11 @@ static struct mlx5_ib_flow_handler *create_flow_rule(struct mlx5_ib_dev *dev, /* Outer header support only */ match_criteria_enable = (!outer_header_zero(match_c)) << 0; + action = dst ? MLX5_FLOW_CONTEXT_ACTION_FWD_DEST : + MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO; handler->rule = mlx5_add_flow_rule(ft, match_criteria_enable, match_c, match_v, - MLX5_FLOW_CONTEXT_ACTION_FWD_DEST, + action, MLX5_FS_DEFAULT_FLOW_TAG, dst); @@ -1494,6 +1508,29 @@ free: return err ? ERR_PTR(err) : handler; } +static struct mlx5_ib_flow_handler *create_dont_trap_rule(struct mlx5_ib_dev *dev, + struct mlx5_ib_flow_prio *ft_prio, + struct ib_flow_attr *flow_attr, + struct mlx5_flow_destination *dst) +{ + struct mlx5_ib_flow_handler *handler_dst = NULL; + struct mlx5_ib_flow_handler *handler = NULL; + + handler = create_flow_rule(dev, ft_prio, flow_attr, NULL); + if (!IS_ERR(handler)) { + handler_dst = create_flow_rule(dev, ft_prio, + flow_attr, dst); + if (IS_ERR(handler_dst)) { + mlx5_del_flow_rule(handler->rule); + kfree(handler); + handler = handler_dst; + } else { + list_add(&handler_dst->list, &handler->list); + } + } + + return handler; +} enum { LEFTOVERS_MC, LEFTOVERS_UC, @@ -1571,7 +1608,7 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp, if (domain != IB_FLOW_DOMAIN_USER || flow_attr->port > MLX5_CAP_GEN(dev->mdev, num_ports) || - flow_attr->flags) + (flow_attr->flags & ~IB_FLOW_ATTR_FLAGS_DONT_TRAP)) return ERR_PTR(-EINVAL); dst = kzalloc(sizeof(*dst), GFP_KERNEL); @@ -1590,8 +1627,13 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp, dst->tir_num = to_mqp(qp)->raw_packet_qp.rq.tirn; if (flow_attr->type == IB_FLOW_ATTR_NORMAL) { - handler = create_flow_rule(dev, ft_prio, flow_attr, - dst); + if (flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP) { + handler = create_dont_trap_rule(dev, ft_prio, + flow_attr, dst); + } else { + handler = create_flow_rule(dev, ft_prio, flow_attr, + dst); + } } else if (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT || flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT) { handler = create_leftovers_rule(dev, ft_prio, flow_attr, diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h index 60b89629f091..76b2b42e0535 100644 --- a/drivers/infiniband/hw/mlx5/mlx5_ib.h +++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h @@ -127,7 +127,7 @@ struct mlx5_ib_pd { }; #define MLX5_IB_FLOW_MCAST_PRIO (MLX5_BY_PASS_NUM_PRIOS - 1) -#define MLX5_IB_FLOW_LAST_PRIO (MLX5_IB_FLOW_MCAST_PRIO - 1) +#define MLX5_IB_FLOW_LAST_PRIO (MLX5_BY_PASS_NUM_REGULAR_PRIOS - 1) #if (MLX5_IB_FLOW_LAST_PRIO <= 0) #error "Invalid number of bypass priorities" #endif diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c index 4659256cd95e..3b2ddd64a371 100644 --- a/drivers/infiniband/hw/mlx5/srq.c +++ b/drivers/infiniband/hw/mlx5/srq.c @@ -75,7 +75,8 @@ static void mlx5_ib_srq_event(struct mlx5_core_srq *srq, enum mlx5_event type) static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq, struct mlx5_create_srq_mbox_in **in, - struct ib_udata *udata, int buf_size, int *inlen) + struct ib_udata *udata, int buf_size, int *inlen, + int is_xrc) { struct mlx5_ib_dev *dev = to_mdev(pd->device); struct mlx5_ib_create_srq ucmd = {}; @@ -87,13 +88,8 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq, int ncont; u32 offset; u32 uidx = MLX5_IB_DEFAULT_UIDX; - int drv_data = udata->inlen - sizeof(struct ib_uverbs_cmd_hdr); - if (drv_data < 0) - return -EINVAL; - - ucmdlen = (drv_data < sizeof(ucmd)) ? - drv_data : sizeof(ucmd); + ucmdlen = min(udata->inlen, sizeof(ucmd)); if (ib_copy_from_udata(&ucmd, udata, ucmdlen)) { mlx5_ib_dbg(dev, "failed copy udata\n"); @@ -103,15 +99,17 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq, if (ucmd.reserved0 || ucmd.reserved1) return -EINVAL; - if (drv_data > sizeof(ucmd) && + if (udata->inlen > sizeof(ucmd) && !ib_is_udata_cleared(udata, sizeof(ucmd), - drv_data - sizeof(ucmd))) + udata->inlen - sizeof(ucmd))) return -EINVAL; - err = get_srq_user_index(to_mucontext(pd->uobject->context), - &ucmd, udata->inlen, &uidx); - if (err) - return err; + if (is_xrc) { + err = get_srq_user_index(to_mucontext(pd->uobject->context), + &ucmd, udata->inlen, &uidx); + if (err) + return err; + } srq->wq_sig = !!(ucmd.flags & MLX5_SRQ_FLAG_SIGNATURE); @@ -151,7 +149,8 @@ static int create_srq_user(struct ib_pd *pd, struct mlx5_ib_srq *srq, (*in)->ctx.log_pg_sz = page_shift - MLX5_ADAPTER_PAGE_SHIFT; (*in)->ctx.pgoff_cqn = cpu_to_be32(offset << 26); - if (MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1) { + if ((MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1) && + is_xrc){ xsrqc = MLX5_ADDR_OF(create_xrc_srq_in, *in, xrc_srq_context_entry); MLX5_SET(xrc_srqc, xsrqc, user_index, uidx); @@ -170,7 +169,7 @@ err_umem: static int create_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq, struct mlx5_create_srq_mbox_in **in, int buf_size, - int *inlen) + int *inlen, int is_xrc) { int err; int i; @@ -224,7 +223,8 @@ static int create_srq_kernel(struct mlx5_ib_dev *dev, struct mlx5_ib_srq *srq, (*in)->ctx.log_pg_sz = page_shift - MLX5_ADAPTER_PAGE_SHIFT; - if (MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1) { + if ((MLX5_CAP_GEN(dev->mdev, cqe_version) == MLX5_CQE_VERSION_V1) && + is_xrc){ xsrqc = MLX5_ADDR_OF(create_xrc_srq_in, *in, xrc_srq_context_entry); /* 0xffffff means we ask to work with cqe version 0 */ @@ -302,10 +302,14 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, desc_size, init_attr->attr.max_wr, srq->msrq.max, srq->msrq.max_gs, srq->msrq.max_avail_gather); + is_xrc = (init_attr->srq_type == IB_SRQT_XRC); + if (pd->uobject) - err = create_srq_user(pd, srq, &in, udata, buf_size, &inlen); + err = create_srq_user(pd, srq, &in, udata, buf_size, &inlen, + is_xrc); else - err = create_srq_kernel(dev, srq, &in, buf_size, &inlen); + err = create_srq_kernel(dev, srq, &in, buf_size, &inlen, + is_xrc); if (err) { mlx5_ib_warn(dev, "create srq %s failed, err %d\n", @@ -313,7 +317,6 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd, goto err_srq; } - is_xrc = (init_attr->srq_type == IB_SRQT_XRC); in->ctx.state_log_sz = ilog2(srq->msrq.max); flgs = ((srq->msrq.wqe_shift - 4) | (is_xrc << 5) | (srq->wq_sig << 7)) << 24; xrcdn = 0; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma.h b/drivers/infiniband/hw/ocrdma/ocrdma.h index 12503f15fbd6..45bdfa0e3b2b 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma.h @@ -114,6 +114,7 @@ struct ocrdma_dev_attr { u8 local_ca_ack_delay; u8 ird; u8 num_ird_pages; + u8 udp_encap; }; struct ocrdma_dma_mem { @@ -356,6 +357,7 @@ struct ocrdma_ah { struct ocrdma_av *av; u16 sgid_index; u32 id; + u8 hdr_type; }; struct ocrdma_qp_hwq_info { @@ -598,4 +600,10 @@ static inline u8 ocrdma_get_ae_link_state(u32 ae_state) return ((ae_state & OCRDMA_AE_LSC_LS_MASK) >> OCRDMA_AE_LSC_LS_SHIFT); } +static inline bool ocrdma_is_udp_encap_supported(struct ocrdma_dev *dev) +{ + return (dev->attr.udp_encap & OCRDMA_L3_TYPE_IPV4) || + (dev->attr.udp_encap & OCRDMA_L3_TYPE_IPV6); +} + #endif diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c index 3790771f2baa..797362a297b2 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c @@ -55,18 +55,46 @@ #define OCRDMA_VID_PCP_SHIFT 0xD +static u16 ocrdma_hdr_type_to_proto_num(int devid, u8 hdr_type) +{ + switch (hdr_type) { + case OCRDMA_L3_TYPE_IB_GRH: + return (u16)0x8915; + case OCRDMA_L3_TYPE_IPV4: + return (u16)0x0800; + case OCRDMA_L3_TYPE_IPV6: + return (u16)0x86dd; + default: + pr_err("ocrdma%d: Invalid network header\n", devid); + return 0; + } +} + static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, struct ib_ah_attr *attr, union ib_gid *sgid, int pdid, bool *isvlan, u16 vlan_tag) { - int status = 0; + int status; struct ocrdma_eth_vlan eth; struct ocrdma_grh grh; int eth_sz; + u16 proto_num = 0; + u8 nxthdr = 0x11; + struct iphdr ipv4; + union { + struct sockaddr _sockaddr; + struct sockaddr_in _sockaddr_in; + struct sockaddr_in6 _sockaddr_in6; + } sgid_addr, dgid_addr; memset(ð, 0, sizeof(eth)); memset(&grh, 0, sizeof(grh)); + /* Protocol Number */ + proto_num = ocrdma_hdr_type_to_proto_num(dev->id, ah->hdr_type); + if (!proto_num) + return -EINVAL; + nxthdr = (proto_num == 0x8915) ? 0x1b : 0x11; /* VLAN */ if (!vlan_tag || (vlan_tag > 0xFFF)) vlan_tag = dev->pvid; @@ -78,13 +106,13 @@ static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, dev->id); } eth.eth_type = cpu_to_be16(0x8100); - eth.roce_eth_type = cpu_to_be16(OCRDMA_ROCE_ETH_TYPE); + eth.roce_eth_type = cpu_to_be16(proto_num); vlan_tag |= (dev->sl & 0x07) << OCRDMA_VID_PCP_SHIFT; eth.vlan_tag = cpu_to_be16(vlan_tag); eth_sz = sizeof(struct ocrdma_eth_vlan); *isvlan = true; } else { - eth.eth_type = cpu_to_be16(OCRDMA_ROCE_ETH_TYPE); + eth.eth_type = cpu_to_be16(proto_num); eth_sz = sizeof(struct ocrdma_eth_basic); } /* MAC */ @@ -93,18 +121,33 @@ static inline int set_av_attr(struct ocrdma_dev *dev, struct ocrdma_ah *ah, if (status) return status; ah->sgid_index = attr->grh.sgid_index; - memcpy(&grh.sgid[0], sgid->raw, sizeof(union ib_gid)); - memcpy(&grh.dgid[0], attr->grh.dgid.raw, sizeof(attr->grh.dgid.raw)); - - grh.tclass_flow = cpu_to_be32((6 << 28) | - (attr->grh.traffic_class << 24) | - attr->grh.flow_label); - /* 0x1b is next header value in GRH */ - grh.pdid_hoplimit = cpu_to_be32((pdid << 16) | - (0x1b << 8) | attr->grh.hop_limit); /* Eth HDR */ memcpy(&ah->av->eth_hdr, ð, eth_sz); - memcpy((u8 *)ah->av + eth_sz, &grh, sizeof(struct ocrdma_grh)); + if (ah->hdr_type == RDMA_NETWORK_IPV4) { + *((__be16 *)&ipv4) = htons((4 << 12) | (5 << 8) | + attr->grh.traffic_class); + ipv4.id = cpu_to_be16(pdid); + ipv4.frag_off = htons(IP_DF); + ipv4.tot_len = htons(0); + ipv4.ttl = attr->grh.hop_limit; + ipv4.protocol = nxthdr; + rdma_gid2ip(&sgid_addr._sockaddr, sgid); + ipv4.saddr = sgid_addr._sockaddr_in.sin_addr.s_addr; + rdma_gid2ip(&dgid_addr._sockaddr, &attr->grh.dgid); + ipv4.daddr = dgid_addr._sockaddr_in.sin_addr.s_addr; + memcpy((u8 *)ah->av + eth_sz, &ipv4, sizeof(struct iphdr)); + } else { + memcpy(&grh.sgid[0], sgid->raw, sizeof(union ib_gid)); + grh.tclass_flow = cpu_to_be32((6 << 28) | + (attr->grh.traffic_class << 24) | + attr->grh.flow_label); + memcpy(&grh.dgid[0], attr->grh.dgid.raw, + sizeof(attr->grh.dgid.raw)); + grh.pdid_hoplimit = cpu_to_be32((pdid << 16) | + (nxthdr << 8) | + attr->grh.hop_limit); + memcpy((u8 *)ah->av + eth_sz, &grh, sizeof(struct ocrdma_grh)); + } if (*isvlan) ah->av->valid |= OCRDMA_AV_VLAN_VALID; ah->av->valid = cpu_to_le32(ah->av->valid); @@ -128,6 +171,7 @@ struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr) if (atomic_cmpxchg(&dev->update_sl, 1, 0)) ocrdma_init_service_level(dev); + ah = kzalloc(sizeof(*ah), GFP_ATOMIC); if (!ah) return ERR_PTR(-ENOMEM); @@ -148,6 +192,8 @@ struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr) vlan_tag = vlan_dev_vlan_id(sgid_attr.ndev); dev_put(sgid_attr.ndev); } + /* Get network header type for this GID */ + ah->hdr_type = ib_gid_to_network_type(sgid_attr.gid_type, &sgid); if ((pd->uctx) && (!rdma_is_multicast_addr((struct in6_addr *)attr->grh.dgid.raw)) && @@ -172,6 +218,11 @@ struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr) ahid_addr = pd->uctx->ah_tbl.va + attr->dlid; *ahid_addr = 0; *ahid_addr |= ah->id & OCRDMA_AH_ID_MASK; + if (ocrdma_is_udp_encap_supported(dev)) { + *ahid_addr |= ((u32)ah->hdr_type & + OCRDMA_AH_L3_TYPE_MASK) << + OCRDMA_AH_L3_TYPE_SHIFT; + } if (isvlan) *ahid_addr |= (OCRDMA_AH_VLAN_VALID_MASK << OCRDMA_AH_VLAN_VALID_SHIFT); diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h index 04a30ae67473..3856dd4c7e3d 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.h @@ -46,9 +46,10 @@ enum { OCRDMA_AH_ID_MASK = 0x3FF, OCRDMA_AH_VLAN_VALID_MASK = 0x01, - OCRDMA_AH_VLAN_VALID_SHIFT = 0x1F + OCRDMA_AH_VLAN_VALID_SHIFT = 0x1F, + OCRDMA_AH_L3_TYPE_MASK = 0x03, + OCRDMA_AH_L3_TYPE_SHIFT = 0x1D /* 29 bits */ }; - struct ib_ah *ocrdma_create_ah(struct ib_pd *, struct ib_ah_attr *); int ocrdma_destroy_ah(struct ib_ah *); int ocrdma_query_ah(struct ib_ah *, struct ib_ah_attr *); diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c index 283ca842ff74..16740dcb876b 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c @@ -1113,7 +1113,7 @@ mbx_err: static int ocrdma_nonemb_mbx_cmd(struct ocrdma_dev *dev, struct ocrdma_mqe *mqe, void *payload_va) { - int status = 0; + int status; struct ocrdma_mbx_rsp *rsp = payload_va; if ((mqe->hdr.spcl_sge_cnt_emb & OCRDMA_MQE_HDR_EMB_MASK) >> @@ -1144,6 +1144,9 @@ static void ocrdma_get_attr(struct ocrdma_dev *dev, attr->max_pd = (rsp->max_pd_ca_ack_delay & OCRDMA_MBX_QUERY_CFG_MAX_PD_MASK) >> OCRDMA_MBX_QUERY_CFG_MAX_PD_SHIFT; + attr->udp_encap = (rsp->max_pd_ca_ack_delay & + OCRDMA_MBX_QUERY_CFG_L3_TYPE_MASK) >> + OCRDMA_MBX_QUERY_CFG_L3_TYPE_SHIFT; attr->max_dpp_pds = (rsp->max_dpp_pds_credits & OCRDMA_MBX_QUERY_CFG_MAX_DPP_PDS_MASK) >> OCRDMA_MBX_QUERY_CFG_MAX_DPP_PDS_OFFSET; @@ -2138,7 +2141,6 @@ int ocrdma_qp_state_change(struct ocrdma_qp *qp, enum ib_qp_state new_ib_state, enum ib_qp_state *old_ib_state) { unsigned long flags; - int status = 0; enum ocrdma_qp_state new_state; new_state = get_ocrdma_qp_state(new_ib_state); @@ -2163,7 +2165,7 @@ int ocrdma_qp_state_change(struct ocrdma_qp *qp, enum ib_qp_state new_ib_state, qp->state = new_state; spin_unlock_irqrestore(&qp->q_lock, flags); - return status; + return 0; } static u32 ocrdma_set_create_qp_mbx_access_flags(struct ocrdma_qp *qp) @@ -2501,7 +2503,12 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, union ib_gid sgid, zgid; struct ib_gid_attr sgid_attr; u32 vlan_id = 0xFFFF; - u8 mac_addr[6]; + u8 mac_addr[6], hdr_type; + union { + struct sockaddr _sockaddr; + struct sockaddr_in _sockaddr_in; + struct sockaddr_in6 _sockaddr_in6; + } sgid_addr, dgid_addr; struct ocrdma_dev *dev = get_ocrdma_dev(qp->ibqp.device); if ((ah_attr->ah_flags & IB_AH_GRH) == 0) @@ -2516,6 +2523,8 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, cmd->params.hop_lmt_rq_psn |= (ah_attr->grh.hop_limit << OCRDMA_QP_PARAMS_HOP_LMT_SHIFT); cmd->flags |= OCRDMA_QP_PARA_FLOW_LBL_VALID; + + /* GIDs */ memcpy(&cmd->params.dgid[0], &ah_attr->grh.dgid.raw[0], sizeof(cmd->params.dgid)); @@ -2538,6 +2547,16 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, return status; cmd->params.dmac_b0_to_b3 = mac_addr[0] | (mac_addr[1] << 8) | (mac_addr[2] << 16) | (mac_addr[3] << 24); + + hdr_type = ib_gid_to_network_type(sgid_attr.gid_type, &sgid); + if (hdr_type == RDMA_NETWORK_IPV4) { + rdma_gid2ip(&sgid_addr._sockaddr, &sgid); + rdma_gid2ip(&dgid_addr._sockaddr, &ah_attr->grh.dgid); + memcpy(&cmd->params.dgid[0], + &dgid_addr._sockaddr_in.sin_addr.s_addr, 4); + memcpy(&cmd->params.sgid[0], + &sgid_addr._sockaddr_in.sin_addr.s_addr, 4); + } /* convert them to LE format. */ ocrdma_cpu_to_le32(&cmd->params.dgid[0], sizeof(cmd->params.dgid)); ocrdma_cpu_to_le32(&cmd->params.sgid[0], sizeof(cmd->params.sgid)); @@ -2558,7 +2577,9 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp, cmd->params.rnt_rc_sl_fl |= (dev->sl & 0x07) << OCRDMA_QP_PARAMS_SL_SHIFT; } - + cmd->params.max_sge_recv_flags |= ((hdr_type << + OCRDMA_QP_PARAMS_FLAGS_L3_TYPE_SHIFT) & + OCRDMA_QP_PARAMS_FLAGS_L3_TYPE_MASK); return 0; } @@ -2871,7 +2892,7 @@ int ocrdma_mbx_destroy_srq(struct ocrdma_dev *dev, struct ocrdma_srq *srq) static int ocrdma_mbx_get_dcbx_config(struct ocrdma_dev *dev, u32 ptype, struct ocrdma_dcbx_cfg *dcbxcfg) { - int status = 0; + int status; dma_addr_t pa; struct ocrdma_mqe cmd; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_main.c b/drivers/infiniband/hw/ocrdma/ocrdma_main.c index f38743018cb4..3d75f65ce87e 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_main.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_main.c @@ -89,8 +89,10 @@ static int ocrdma_port_immutable(struct ib_device *ibdev, u8 port_num, struct ib_port_immutable *immutable) { struct ib_port_attr attr; + struct ocrdma_dev *dev; int err; + dev = get_ocrdma_dev(ibdev); err = ocrdma_query_port(ibdev, port_num, &attr); if (err) return err; @@ -98,6 +100,8 @@ static int ocrdma_port_immutable(struct ib_device *ibdev, u8 port_num, immutable->pkey_tbl_len = attr.pkey_tbl_len; immutable->gid_tbl_len = attr.gid_tbl_len; immutable->core_cap_flags = RDMA_CORE_PORT_IBA_ROCE; + if (ocrdma_is_udp_encap_supported(dev)) + immutable->core_cap_flags |= RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP; immutable->max_mad_size = IB_MGMT_MAD_SIZE; return 0; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h index 99dd6fdf06d7..0efc9662c6d8 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_sli.h +++ b/drivers/infiniband/hw/ocrdma/ocrdma_sli.h @@ -140,7 +140,11 @@ enum { OCRDMA_DB_RQ_SHIFT = 24 }; -#define OCRDMA_ROUDP_FLAGS_SHIFT 0x03 +enum { + OCRDMA_L3_TYPE_IB_GRH = 0x00, + OCRDMA_L3_TYPE_IPV4 = 0x01, + OCRDMA_L3_TYPE_IPV6 = 0x02 +}; #define OCRDMA_DB_CQ_RING_ID_MASK 0x3FF /* bits 0 - 9 */ #define OCRDMA_DB_CQ_RING_ID_EXT_MASK 0x0C00 /* bits 10-11 of qid at 12-11 */ @@ -546,7 +550,8 @@ enum { OCRDMA_MBX_QUERY_CFG_CA_ACK_DELAY_SHIFT = 8, OCRDMA_MBX_QUERY_CFG_CA_ACK_DELAY_MASK = 0xFF << OCRDMA_MBX_QUERY_CFG_CA_ACK_DELAY_SHIFT, - + OCRDMA_MBX_QUERY_CFG_L3_TYPE_SHIFT = 3, + OCRDMA_MBX_QUERY_CFG_L3_TYPE_MASK = 0x18, OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_SHIFT = 0, OCRDMA_MBX_QUERY_CFG_MAX_SEND_SGE_MASK = 0xFFFF, OCRDMA_MBX_QUERY_CFG_MAX_WRITE_SGE_SHIFT = 16, @@ -1107,6 +1112,8 @@ enum { OCRDMA_QP_PARAMS_STATE_MASK = BIT(5) | BIT(6) | BIT(7), OCRDMA_QP_PARAMS_FLAGS_SQD_ASYNC = BIT(8), OCRDMA_QP_PARAMS_FLAGS_INB_ATEN = BIT(9), + OCRDMA_QP_PARAMS_FLAGS_L3_TYPE_SHIFT = 11, + OCRDMA_QP_PARAMS_FLAGS_L3_TYPE_MASK = BIT(11) | BIT(12) | BIT(13), OCRDMA_QP_PARAMS_MAX_SGE_RECV_SHIFT = 16, OCRDMA_QP_PARAMS_MAX_SGE_RECV_MASK = 0xFFFF << OCRDMA_QP_PARAMS_MAX_SGE_RECV_SHIFT, @@ -1735,8 +1742,11 @@ enum { /* w1 */ OCRDMA_CQE_UD_XFER_LEN_SHIFT = 16, + OCRDMA_CQE_UD_XFER_LEN_MASK = 0x1FFF, OCRDMA_CQE_PKEY_SHIFT = 0, OCRDMA_CQE_PKEY_MASK = 0xFFFF, + OCRDMA_CQE_UD_L3TYPE_SHIFT = 29, + OCRDMA_CQE_UD_L3TYPE_MASK = 0x07, /* w2 */ OCRDMA_CQE_QPN_SHIFT = 0, @@ -1861,7 +1871,7 @@ struct ocrdma_ewqe_ud_hdr { u32 rsvd_dest_qpn; u32 qkey; u32 rsvd_ahid; - u32 rsvd; + u32 hdr_type; }; /* extended wqe followed by hdr_wqe for Fast Memory register */ diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c index 255f774080a4..8bef09a8c49f 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_stats.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_stats.c @@ -610,7 +610,7 @@ static char *ocrdma_driver_dbg_stats(struct ocrdma_dev *dev) static void ocrdma_update_stats(struct ocrdma_dev *dev) { ulong now = jiffies, secs; - int status = 0; + int status; struct ocrdma_rdma_stats_resp *rdma_stats = (struct ocrdma_rdma_stats_resp *)dev->stats_mem.va; struct ocrdma_rsrc_stats *rsrc_stats = &rdma_stats->act_rsrc_stats; @@ -641,7 +641,7 @@ static ssize_t ocrdma_dbgfs_ops_write(struct file *filp, { char tmp_str[32]; long reset; - int status = 0; + int status; struct ocrdma_stats *pstats = filp->private_data; struct ocrdma_dev *dev = pstats->dev; diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c index 12420e4ecf3d..a8496a18e20d 100644 --- a/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +++ b/drivers/infiniband/hw/ocrdma/ocrdma_verbs.c @@ -419,7 +419,7 @@ static struct ocrdma_pd *_ocrdma_alloc_pd(struct ocrdma_dev *dev, struct ib_udata *udata) { struct ocrdma_pd *pd = NULL; - int status = 0; + int status; pd = kzalloc(sizeof(*pd), GFP_KERNEL); if (!pd) @@ -468,7 +468,7 @@ static inline int is_ucontext_pd(struct ocrdma_ucontext *uctx, static int _ocrdma_dealloc_pd(struct ocrdma_dev *dev, struct ocrdma_pd *pd) { - int status = 0; + int status; if (dev->pd_mgr->pd_prealloc_valid) status = ocrdma_put_pd_num(dev, pd->id, pd->dpp_enabled); @@ -596,7 +596,7 @@ map_err: int ocrdma_dealloc_ucontext(struct ib_ucontext *ibctx) { - int status = 0; + int status; struct ocrdma_mm *mm, *tmp; struct ocrdma_ucontext *uctx = get_ocrdma_ucontext(ibctx); struct ocrdma_dev *dev = get_ocrdma_dev(ibctx->device); @@ -623,7 +623,7 @@ int ocrdma_mmap(struct ib_ucontext *context, struct vm_area_struct *vma) unsigned long vm_page = vma->vm_pgoff << PAGE_SHIFT; u64 unmapped_db = (u64) dev->nic_info.unmapped_db; unsigned long len = (vma->vm_end - vma->vm_start); - int status = 0; + int status; bool found; if (vma->vm_start & (PAGE_SIZE - 1)) @@ -1285,7 +1285,7 @@ static int ocrdma_copy_qp_uresp(struct ocrdma_qp *qp, struct ib_udata *udata, int dpp_offset, int dpp_credit_lmt, int srq) { - int status = 0; + int status; u64 usr_db; struct ocrdma_create_qp_uresp uresp; struct ocrdma_pd *pd = qp->pd; @@ -1494,9 +1494,7 @@ int _ocrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, */ if (status < 0) return status; - status = ocrdma_mbx_modify_qp(dev, qp, attr, attr_mask); - - return status; + return ocrdma_mbx_modify_qp(dev, qp, attr, attr_mask); } int ocrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, @@ -1949,7 +1947,7 @@ int ocrdma_modify_srq(struct ib_srq *ibsrq, enum ib_srq_attr_mask srq_attr_mask, struct ib_udata *udata) { - int status = 0; + int status; struct ocrdma_srq *srq; srq = get_ocrdma_srq(ibsrq); @@ -2005,6 +2003,7 @@ static void ocrdma_build_ud_hdr(struct ocrdma_qp *qp, else ud_hdr->qkey = ud_wr(wr)->remote_qkey; ud_hdr->rsvd_ahid = ah->id; + ud_hdr->hdr_type = ah->hdr_type; if (ah->av->valid & OCRDMA_AV_VLAN_VALID) hdr->cw |= (OCRDMA_FLAG_AH_VLAN_PR << OCRDMA_WQE_FLAGS_SHIFT); } @@ -2717,9 +2716,11 @@ static bool ocrdma_poll_scqe(struct ocrdma_qp *qp, struct ocrdma_cqe *cqe, return expand; } -static int ocrdma_update_ud_rcqe(struct ib_wc *ibwc, struct ocrdma_cqe *cqe) +static int ocrdma_update_ud_rcqe(struct ocrdma_dev *dev, struct ib_wc *ibwc, + struct ocrdma_cqe *cqe) { int status; + u16 hdr_type = 0; status = (le32_to_cpu(cqe->flags_status_srcqpn) & OCRDMA_CQE_UD_STATUS_MASK) >> OCRDMA_CQE_UD_STATUS_SHIFT; @@ -2728,7 +2729,17 @@ static int ocrdma_update_ud_rcqe(struct ib_wc *ibwc, struct ocrdma_cqe *cqe) ibwc->pkey_index = 0; ibwc->wc_flags = IB_WC_GRH; ibwc->byte_len = (le32_to_cpu(cqe->ud.rxlen_pkey) >> - OCRDMA_CQE_UD_XFER_LEN_SHIFT); + OCRDMA_CQE_UD_XFER_LEN_SHIFT) & + OCRDMA_CQE_UD_XFER_LEN_MASK; + + if (ocrdma_is_udp_encap_supported(dev)) { + hdr_type = (le32_to_cpu(cqe->ud.rxlen_pkey) >> + OCRDMA_CQE_UD_L3TYPE_SHIFT) & + OCRDMA_CQE_UD_L3TYPE_MASK; + ibwc->wc_flags |= IB_WC_WITH_NETWORK_HDR_TYPE; + ibwc->network_hdr_type = hdr_type; + } + return status; } @@ -2791,12 +2802,15 @@ static bool ocrdma_poll_err_rcqe(struct ocrdma_qp *qp, struct ocrdma_cqe *cqe, static void ocrdma_poll_success_rcqe(struct ocrdma_qp *qp, struct ocrdma_cqe *cqe, struct ib_wc *ibwc) { + struct ocrdma_dev *dev; + + dev = get_ocrdma_dev(qp->ibqp.device); ibwc->opcode = IB_WC_RECV; ibwc->qp = &qp->ibqp; ibwc->status = IB_WC_SUCCESS; if (qp->qp_type == IB_QPT_UD || qp->qp_type == IB_QPT_GSI) - ocrdma_update_ud_rcqe(ibwc, cqe); + ocrdma_update_ud_rcqe(dev, ibwc, cqe); else ibwc->byte_len = le32_to_cpu(cqe->rq.rxlen); diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index a6f3eab0f350..85be0de3ab26 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -244,6 +244,7 @@ struct ipoib_cm_tx { unsigned tx_tail; unsigned long flags; u32 mtu; + unsigned max_send_sge; }; struct ipoib_cm_rx_buf { @@ -390,6 +391,7 @@ struct ipoib_dev_priv { int hca_caps; struct ipoib_ethtool_st ethtool; struct timer_list poll_timer; + unsigned max_send_sge; }; struct ipoib_ah { diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index 917e46ea3bf6..c8ed53562c9b 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -710,6 +710,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_ struct ipoib_dev_priv *priv = netdev_priv(dev); struct ipoib_tx_buf *tx_req; int rc; + unsigned usable_sge = tx->max_send_sge - !!skb_headlen(skb); if (unlikely(skb->len > tx->mtu)) { ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n", @@ -719,7 +720,23 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_ ipoib_cm_skb_too_long(dev, skb, tx->mtu - IPOIB_ENCAP_LEN); return; } - + if (skb_shinfo(skb)->nr_frags > usable_sge) { + if (skb_linearize(skb) < 0) { + ipoib_warn(priv, "skb could not be linearized\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + /* Does skb_linearize return ok without reducing nr_frags? */ + if (skb_shinfo(skb)->nr_frags > usable_sge) { + ipoib_warn(priv, "too many frags after skb linearize\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + } ipoib_dbg_data(priv, "sending packet: head 0x%x length %d connection 0x%x\n", tx->tx_head, skb->len, tx->qp->qp_num); @@ -1031,7 +1048,8 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_ struct ib_qp *tx_qp; if (dev->features & NETIF_F_SG) - attr.cap.max_send_sge = MAX_SKB_FRAGS + 1; + attr.cap.max_send_sge = + min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1); tx_qp = ib_create_qp(priv->pd, &attr); if (PTR_ERR(tx_qp) == -EINVAL) { @@ -1040,6 +1058,7 @@ static struct ib_qp *ipoib_cm_create_tx_qp(struct net_device *dev, struct ipoib_ attr.create_flags &= ~IB_QP_CREATE_USE_GFP_NOIO; tx_qp = ib_create_qp(priv->pd, &attr); } + tx->max_send_sge = attr.cap.max_send_sge; return tx_qp; } diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index fa9c42ff1fb0..899e6b7fb8a5 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -538,6 +538,7 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_tx_buf *tx_req; int hlen, rc; void *phead; + unsigned usable_sge = priv->max_send_sge - !!skb_headlen(skb); if (skb_is_gso(skb)) { hlen = skb_transport_offset(skb) + tcp_hdrlen(skb); @@ -561,6 +562,23 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, phead = NULL; hlen = 0; } + if (skb_shinfo(skb)->nr_frags > usable_sge) { + if (skb_linearize(skb) < 0) { + ipoib_warn(priv, "skb could not be linearized\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + /* Does skb_linearize return ok without reducing nr_frags? */ + if (skb_shinfo(skb)->nr_frags > usable_sge) { + ipoib_warn(priv, "too many frags after skb linearize\n"); + ++dev->stats.tx_dropped; + ++dev->stats.tx_errors; + dev_kfree_skb_any(skb); + return; + } + } ipoib_dbg_data(priv, "sending packet, length=%d address=%p qpn=0x%06x\n", skb->len, address, qpn); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index d48c5bae7877..b809c373e40e 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -206,7 +206,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca) init_attr.create_flags |= IB_QP_CREATE_NETIF_QP; if (dev->features & NETIF_F_SG) - init_attr.cap.max_send_sge = MAX_SKB_FRAGS + 1; + init_attr.cap.max_send_sge = + min_t(u32, priv->ca->attrs.max_sge, MAX_SKB_FRAGS + 1); priv->qp = ib_create_qp(priv->pd, &init_attr); if (IS_ERR(priv->qp)) { @@ -233,6 +234,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca) priv->rx_wr.next = NULL; priv->rx_wr.sg_list = priv->rx_sge; + priv->max_send_sge = init_attr.cap.max_send_sge; + return 0; out_free_send_cq: diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 95f0a64e076b..0351059783b1 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -458,9 +458,6 @@ struct iser_fr_pool { * @comp: iser completion context * @fr_pool: connection fast registration poool * @pi_support: Indicate device T10-PI support - * @last: last send wr to signal all flush errors were drained - * @last_cqe: cqe handler for last wr - * @last_comp: completes when all connection completions consumed */ struct ib_conn { struct rdma_cm_id *cma_id; @@ -472,10 +469,7 @@ struct ib_conn { struct iser_comp *comp; struct iser_fr_pool fr_pool; bool pi_support; - struct ib_send_wr last; - struct ib_cqe last_cqe; struct ib_cqe reg_cqe; - struct completion last_comp; }; /** @@ -617,7 +611,6 @@ void iser_cmd_comp(struct ib_cq *cq, struct ib_wc *wc); void iser_ctrl_comp(struct ib_cq *cq, struct ib_wc *wc); void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc); void iser_reg_comp(struct ib_cq *cq, struct ib_wc *wc); -void iser_last_comp(struct ib_cq *cq, struct ib_wc *wc); void iser_task_rdma_init(struct iscsi_iser_task *task); diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index ed54b388e7ad..81ae2e30dd12 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -729,13 +729,6 @@ void iser_dataout_comp(struct ib_cq *cq, struct ib_wc *wc) kmem_cache_free(ig.desc_cache, desc); } -void iser_last_comp(struct ib_cq *cq, struct ib_wc *wc) -{ - struct ib_conn *ib_conn = wc->qp->qp_context; - - complete(&ib_conn->last_comp); -} - void iser_task_rdma_init(struct iscsi_iser_task *iser_task) { diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index f21bdcc34d59..1b4945367e4f 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -670,7 +670,6 @@ void iser_conn_release(struct iser_conn *iser_conn) int iser_conn_terminate(struct iser_conn *iser_conn) { struct ib_conn *ib_conn = &iser_conn->ib_conn; - struct ib_send_wr *bad_wr; int err = 0; /* terminate the iser conn only if the conn state is UP */ @@ -695,14 +694,8 @@ int iser_conn_terminate(struct iser_conn *iser_conn) iser_err("Failed to disconnect, conn: 0x%p err %d\n", iser_conn, err); - /* post an indication that all flush errors were consumed */ - err = ib_post_send(ib_conn->qp, &ib_conn->last, &bad_wr); - if (err) { - iser_err("conn %p failed to post last wr", ib_conn); - return 1; - } - - wait_for_completion(&ib_conn->last_comp); + /* block until all flush errors are consumed */ + ib_drain_sq(ib_conn->qp); } return 1; @@ -961,10 +954,6 @@ void iser_conn_init(struct iser_conn *iser_conn) ib_conn->post_recv_buf_count = 0; ib_conn->reg_cqe.done = iser_reg_comp; - ib_conn->last_cqe.done = iser_last_comp; - ib_conn->last.wr_cqe = &ib_conn->last_cqe; - ib_conn->last.opcode = IB_WR_SEND; - init_completion(&ib_conn->last_comp); } /** diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 03022f6420d7..b6bf20496021 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -446,49 +446,17 @@ static struct srp_fr_pool *srp_alloc_fr_pool(struct srp_target_port *target) dev->max_pages_per_mr); } -static void srp_drain_done(struct ib_cq *cq, struct ib_wc *wc) -{ - struct srp_rdma_ch *ch = cq->cq_context; - - complete(&ch->done); -} - -static struct ib_cqe srp_drain_cqe = { - .done = srp_drain_done, -}; - /** * srp_destroy_qp() - destroy an RDMA queue pair * @ch: SRP RDMA channel. * - * Change a queue pair into the error state and wait until all receive - * completions have been processed before destroying it. This avoids that - * the receive completion handler can access the queue pair while it is + * Drain the qp before destroying it. This avoids that the receive + * completion handler can access the queue pair while it is * being destroyed. */ static void srp_destroy_qp(struct srp_rdma_ch *ch) { - static struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR }; - static struct ib_recv_wr wr = { 0 }; - struct ib_recv_wr *bad_wr; - int ret; - - wr.wr_cqe = &srp_drain_cqe; - /* Destroying a QP and reusing ch->done is only safe if not connected */ - WARN_ON_ONCE(ch->connected); - - ret = ib_modify_qp(ch->qp, &attr, IB_QP_STATE); - WARN_ONCE(ret, "ib_cm_init_qp_attr() returned %d\n", ret); - if (ret) - goto out; - - init_completion(&ch->done); - ret = ib_post_recv(ch->qp, &wr, &bad_wr); - WARN_ONCE(ret, "ib_post_recv() returned %d\n", ret); - if (ret == 0) - wait_for_completion(&ch->done); - -out: + ib_drain_rq(ch->qp); ib_destroy_qp(ch->qp); } @@ -508,7 +476,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch) if (!init_attr) return -ENOMEM; - /* queue_size + 1 for ib_drain_qp */ + /* queue_size + 1 for ib_drain_rq() */ recv_cq = ib_alloc_cq(dev->dev, ch, target->queue_size + 1, ch->comp_vector, IB_POLL_SOFTIRQ); if (IS_ERR(recv_cq)) { diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c index 0c37fee363b1..25bdaeef2520 100644 --- a/drivers/infiniband/ulp/srpt/ib_srpt.c +++ b/drivers/infiniband/ulp/srpt/ib_srpt.c @@ -91,76 +91,32 @@ MODULE_PARM_DESC(srpt_service_guid, " instead of using the node_guid of the first HCA."); static struct ib_client srpt_client; -static void srpt_release_channel(struct srpt_rdma_ch *ch); +static void srpt_release_cmd(struct se_cmd *se_cmd); +static void srpt_free_ch(struct kref *kref); static int srpt_queue_status(struct se_cmd *cmd); static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc); static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc); +static void srpt_process_wait_list(struct srpt_rdma_ch *ch); -/** - * opposite_dma_dir() - Swap DMA_TO_DEVICE and DMA_FROM_DEVICE. - */ -static inline -enum dma_data_direction opposite_dma_dir(enum dma_data_direction dir) -{ - switch (dir) { - case DMA_TO_DEVICE: return DMA_FROM_DEVICE; - case DMA_FROM_DEVICE: return DMA_TO_DEVICE; - default: return dir; - } -} - -/** - * srpt_sdev_name() - Return the name associated with the HCA. - * - * Examples are ib0, ib1, ... - */ -static inline const char *srpt_sdev_name(struct srpt_device *sdev) -{ - return sdev->device->name; -} - -static enum rdma_ch_state srpt_get_ch_state(struct srpt_rdma_ch *ch) -{ - unsigned long flags; - enum rdma_ch_state state; - - spin_lock_irqsave(&ch->spinlock, flags); - state = ch->state; - spin_unlock_irqrestore(&ch->spinlock, flags); - return state; -} - -static enum rdma_ch_state -srpt_set_ch_state(struct srpt_rdma_ch *ch, enum rdma_ch_state new_state) -{ - unsigned long flags; - enum rdma_ch_state prev; - - spin_lock_irqsave(&ch->spinlock, flags); - prev = ch->state; - ch->state = new_state; - spin_unlock_irqrestore(&ch->spinlock, flags); - return prev; -} - -/** - * srpt_test_and_set_ch_state() - Test and set the channel state. - * - * Returns true if and only if the channel state has been set to the new state. +/* + * The only allowed channel state changes are those that change the channel + * state into a state with a higher numerical value. Hence the new > prev test. */ -static bool -srpt_test_and_set_ch_state(struct srpt_rdma_ch *ch, enum rdma_ch_state old, - enum rdma_ch_state new) +static bool srpt_set_ch_state(struct srpt_rdma_ch *ch, enum rdma_ch_state new) { unsigned long flags; enum rdma_ch_state prev; + bool changed = false; spin_lock_irqsave(&ch->spinlock, flags); prev = ch->state; - if (prev == old) + if (new > prev) { ch->state = new; + changed = true; + } spin_unlock_irqrestore(&ch->spinlock, flags); - return prev == old; + + return changed; } /** @@ -182,7 +138,7 @@ static void srpt_event_handler(struct ib_event_handler *handler, return; pr_debug("ASYNC event= %d on device= %s\n", event->event, - srpt_sdev_name(sdev)); + sdev->device->name); switch (event->event) { case IB_EVENT_PORT_ERR: @@ -220,25 +176,39 @@ static void srpt_srq_event(struct ib_event *event, void *ctx) pr_info("SRQ event %d\n", event->event); } +static const char *get_ch_state_name(enum rdma_ch_state s) +{ + switch (s) { + case CH_CONNECTING: + return "connecting"; + case CH_LIVE: + return "live"; + case CH_DISCONNECTING: + return "disconnecting"; + case CH_DRAINING: + return "draining"; + case CH_DISCONNECTED: + return "disconnected"; + } + return "???"; +} + /** * srpt_qp_event() - QP event callback function. */ static void srpt_qp_event(struct ib_event *event, struct srpt_rdma_ch *ch) { pr_debug("QP event %d on cm_id=%p sess_name=%s state=%d\n", - event->event, ch->cm_id, ch->sess_name, srpt_get_ch_state(ch)); + event->event, ch->cm_id, ch->sess_name, ch->state); switch (event->event) { case IB_EVENT_COMM_EST: ib_cm_notify(ch->cm_id, event->event); break; case IB_EVENT_QP_LAST_WQE_REACHED: - if (srpt_test_and_set_ch_state(ch, CH_DRAINING, - CH_RELEASING)) - srpt_release_channel(ch); - else - pr_debug("%s: state %d - ignored LAST_WQE.\n", - ch->sess_name, srpt_get_ch_state(ch)); + pr_debug("%s-%d, state %s: received Last WQE event.\n", + ch->sess_name, ch->qp->qp_num, + get_ch_state_name(ch->state)); break; default: pr_err("received unrecognized IB QP event %d\n", event->event); @@ -281,7 +251,7 @@ static void srpt_get_class_port_info(struct ib_dm_mad *mad) struct ib_class_port_info *cif; cif = (struct ib_class_port_info *)mad->data; - memset(cif, 0, sizeof *cif); + memset(cif, 0, sizeof(*cif)); cif->base_version = 1; cif->class_version = 1; cif->resp_time_value = 20; @@ -340,7 +310,7 @@ static void srpt_get_ioc(struct srpt_port *sport, u32 slot, return; } - memset(iocp, 0, sizeof *iocp); + memset(iocp, 0, sizeof(*iocp)); strcpy(iocp->id_string, SRPT_ID_STRING); iocp->guid = cpu_to_be64(srpt_service_guid); iocp->vendor_id = cpu_to_be32(sdev->device->attrs.vendor_id); @@ -390,7 +360,7 @@ static void srpt_get_svc_entries(u64 ioc_guid, } svc_entries = (struct ib_dm_svc_entries *)mad->data; - memset(svc_entries, 0, sizeof *svc_entries); + memset(svc_entries, 0, sizeof(*svc_entries)); svc_entries->service_entries[0].id = cpu_to_be64(ioc_guid); snprintf(svc_entries->service_entries[0].name, sizeof(svc_entries->service_entries[0].name), @@ -484,7 +454,7 @@ static void srpt_mad_recv_handler(struct ib_mad_agent *mad_agent, rsp->ah = ah; dm_mad = rsp->mad; - memcpy(dm_mad, mad_wc->recv_buf.mad, sizeof *dm_mad); + memcpy(dm_mad, mad_wc->recv_buf.mad, sizeof(*dm_mad)); dm_mad->mad_hdr.method = IB_MGMT_METHOD_GET_RESP; dm_mad->mad_hdr.status = 0; @@ -532,7 +502,7 @@ static int srpt_refresh_port(struct srpt_port *sport) struct ib_port_attr port_attr; int ret; - memset(&port_modify, 0, sizeof port_modify); + memset(&port_modify, 0, sizeof(port_modify)); port_modify.set_port_cap_mask = IB_PORT_DEVICE_MGMT_SUP; port_modify.clr_port_cap_mask = 0; @@ -553,7 +523,7 @@ static int srpt_refresh_port(struct srpt_port *sport) goto err_query_port; if (!sport->mad_agent) { - memset(®_req, 0, sizeof reg_req); + memset(®_req, 0, sizeof(reg_req)); reg_req.mgmt_class = IB_MGMT_CLASS_DEVICE_MGMT; reg_req.mgmt_class_version = IB_MGMT_BASE_VERSION; set_bit(IB_MGMT_METHOD_GET, reg_req.method_mask); @@ -841,6 +811,39 @@ out: } /** + * srpt_zerolength_write() - Perform a zero-length RDMA write. + * + * A quote from the InfiniBand specification: C9-88: For an HCA responder + * using Reliable Connection service, for each zero-length RDMA READ or WRITE + * request, the R_Key shall not be validated, even if the request includes + * Immediate data. + */ +static int srpt_zerolength_write(struct srpt_rdma_ch *ch) +{ + struct ib_send_wr wr, *bad_wr; + + memset(&wr, 0, sizeof(wr)); + wr.opcode = IB_WR_RDMA_WRITE; + wr.wr_cqe = &ch->zw_cqe; + wr.send_flags = IB_SEND_SIGNALED; + return ib_post_send(ch->qp, &wr, &bad_wr); +} + +static void srpt_zerolength_write_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct srpt_rdma_ch *ch = cq->cq_context; + + if (wc->status == IB_WC_SUCCESS) { + srpt_process_wait_list(ch); + } else { + if (srpt_set_ch_state(ch, CH_DISCONNECTED)) + schedule_work(&ch->release_work); + else + WARN_ONCE("%s-%d\n", ch->sess_name, ch->qp->qp_num); + } +} + +/** * srpt_get_desc_tbl() - Parse the data descriptors of an SRP_CMD request. * @ioctx: Pointer to the I/O context associated with the request. * @srp_cmd: Pointer to the SRP_CMD request data. @@ -903,14 +906,14 @@ static int srpt_get_desc_tbl(struct srpt_send_ioctx *ioctx, db = (struct srp_direct_buf *)(srp_cmd->add_data + add_cdb_offset); - memcpy(ioctx->rbufs, db, sizeof *db); + memcpy(ioctx->rbufs, db, sizeof(*db)); *data_len = be32_to_cpu(db->len); } else if (((srp_cmd->buf_fmt & 0xf) == SRP_DATA_DESC_INDIRECT) || ((srp_cmd->buf_fmt >> 4) == SRP_DATA_DESC_INDIRECT)) { idb = (struct srp_indirect_buf *)(srp_cmd->add_data + add_cdb_offset); - ioctx->n_rbuf = be32_to_cpu(idb->table_desc.len) / sizeof *db; + ioctx->n_rbuf = be32_to_cpu(idb->table_desc.len) / sizeof(*db); if (ioctx->n_rbuf > (srp_cmd->data_out_desc_cnt + srp_cmd->data_in_desc_cnt)) { @@ -929,7 +932,7 @@ static int srpt_get_desc_tbl(struct srpt_send_ioctx *ioctx, ioctx->rbufs = &ioctx->single_rbuf; else { ioctx->rbufs = - kmalloc(ioctx->n_rbuf * sizeof *db, GFP_ATOMIC); + kmalloc(ioctx->n_rbuf * sizeof(*db), GFP_ATOMIC); if (!ioctx->rbufs) { ioctx->n_rbuf = 0; ret = -ENOMEM; @@ -938,7 +941,7 @@ static int srpt_get_desc_tbl(struct srpt_send_ioctx *ioctx, } db = idb->desc_list; - memcpy(ioctx->rbufs, db, ioctx->n_rbuf * sizeof *db); + memcpy(ioctx->rbufs, db, ioctx->n_rbuf * sizeof(*db)); *data_len = be32_to_cpu(idb->len); } out: @@ -956,7 +959,7 @@ static int srpt_init_ch_qp(struct srpt_rdma_ch *ch, struct ib_qp *qp) struct ib_qp_attr *attr; int ret; - attr = kzalloc(sizeof *attr, GFP_KERNEL); + attr = kzalloc(sizeof(*attr), GFP_KERNEL); if (!attr) return -ENOMEM; @@ -1070,7 +1073,7 @@ static void srpt_unmap_sg_to_ib_sge(struct srpt_rdma_ch *ch, dir = ioctx->cmd.data_direction; BUG_ON(dir == DMA_NONE); ib_dma_unmap_sg(ch->sport->sdev->device, sg, ioctx->sg_cnt, - opposite_dma_dir(dir)); + target_reverse_dma_direction(&ioctx->cmd)); ioctx->mapped_sg_count = 0; } } @@ -1107,7 +1110,7 @@ static int srpt_map_sg_to_ib_sge(struct srpt_rdma_ch *ch, ioctx->sg_cnt = sg_cnt = cmd->t_data_nents; count = ib_dma_map_sg(ch->sport->sdev->device, sg, sg_cnt, - opposite_dma_dir(dir)); + target_reverse_dma_direction(cmd)); if (unlikely(!count)) return -EAGAIN; @@ -1313,10 +1316,7 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) /* * If the command is in a state where the target core is waiting for - * the ib_srpt driver, change the state to the next state. Changing - * the state of the command from SRPT_STATE_NEED_DATA to - * SRPT_STATE_DATA_IN ensures that srpt_xmit_response() will call this - * function a second time. + * the ib_srpt driver, change the state to the next state. */ spin_lock_irqsave(&ioctx->spinlock, flags); @@ -1325,25 +1325,17 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) case SRPT_STATE_NEED_DATA: ioctx->state = SRPT_STATE_DATA_IN; break; - case SRPT_STATE_DATA_IN: case SRPT_STATE_CMD_RSP_SENT: case SRPT_STATE_MGMT_RSP_SENT: ioctx->state = SRPT_STATE_DONE; break; default: + WARN_ONCE(true, "%s: unexpected I/O context state %d\n", + __func__, state); break; } spin_unlock_irqrestore(&ioctx->spinlock, flags); - if (state == SRPT_STATE_DONE) { - struct srpt_rdma_ch *ch = ioctx->ch; - - BUG_ON(ch->sess == NULL); - - target_put_sess_cmd(&ioctx->cmd); - goto out; - } - pr_debug("Aborting cmd with state %d and tag %lld\n", state, ioctx->cmd.tag); @@ -1351,19 +1343,16 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) case SRPT_STATE_NEW: case SRPT_STATE_DATA_IN: case SRPT_STATE_MGMT: + case SRPT_STATE_DONE: /* * Do nothing - defer abort processing until * srpt_queue_response() is invoked. */ - WARN_ON(!transport_check_aborted_status(&ioctx->cmd, false)); break; case SRPT_STATE_NEED_DATA: - /* DMA_TO_DEVICE (write) - RDMA read error. */ - - /* XXX(hch): this is a horrible layering violation.. */ - spin_lock_irqsave(&ioctx->cmd.t_state_lock, flags); - ioctx->cmd.transport_state &= ~CMD_T_ACTIVE; - spin_unlock_irqrestore(&ioctx->cmd.t_state_lock, flags); + pr_debug("tag %#llx: RDMA read error\n", ioctx->cmd.tag); + transport_generic_request_failure(&ioctx->cmd, + TCM_CHECK_CONDITION_ABORT_CMD); break; case SRPT_STATE_CMD_RSP_SENT: /* @@ -1371,18 +1360,16 @@ static int srpt_abort_cmd(struct srpt_send_ioctx *ioctx) * not been received in time. */ srpt_unmap_sg_to_ib_sge(ioctx->ch, ioctx); - target_put_sess_cmd(&ioctx->cmd); + transport_generic_free_cmd(&ioctx->cmd, 0); break; case SRPT_STATE_MGMT_RSP_SENT: - srpt_set_cmd_state(ioctx, SRPT_STATE_DONE); - target_put_sess_cmd(&ioctx->cmd); + transport_generic_free_cmd(&ioctx->cmd, 0); break; default: WARN(1, "Unexpected command state (%d)", state); break; } -out: return state; } @@ -1422,9 +1409,14 @@ static void srpt_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc) container_of(wc->wr_cqe, struct srpt_send_ioctx, rdma_cqe); if (unlikely(wc->status != IB_WC_SUCCESS)) { + /* + * Note: if an RDMA write error completion is received that + * means that a SEND also has been posted. Defer further + * processing of the associated command until the send error + * completion has been received. + */ pr_info("RDMA_WRITE for ioctx 0x%p failed with status %d\n", ioctx, wc->status); - srpt_abort_cmd(ioctx); } } @@ -1464,7 +1456,7 @@ static int srpt_build_cmd_rsp(struct srpt_rdma_ch *ch, sense_data_len = ioctx->cmd.scsi_sense_length; WARN_ON(sense_data_len > sizeof(ioctx->sense_data)); - memset(srp_rsp, 0, sizeof *srp_rsp); + memset(srp_rsp, 0, sizeof(*srp_rsp)); srp_rsp->opcode = SRP_RSP; srp_rsp->req_lim_delta = cpu_to_be32(1 + atomic_xchg(&ch->req_lim_delta, 0)); @@ -1514,7 +1506,7 @@ static int srpt_build_tskmgmt_rsp(struct srpt_rdma_ch *ch, srp_rsp = ioctx->ioctx.buf; BUG_ON(!srp_rsp); - memset(srp_rsp, 0, sizeof *srp_rsp); + memset(srp_rsp, 0, sizeof(*srp_rsp)); srp_rsp->opcode = SRP_RSP; srp_rsp->req_lim_delta = @@ -1528,80 +1520,6 @@ static int srpt_build_tskmgmt_rsp(struct srpt_rdma_ch *ch, return resp_len; } -#define NO_SUCH_LUN ((uint64_t)-1LL) - -/* - * SCSI LUN addressing method. See also SAM-2 and the section about - * eight byte LUNs. - */ -enum scsi_lun_addr_method { - SCSI_LUN_ADDR_METHOD_PERIPHERAL = 0, - SCSI_LUN_ADDR_METHOD_FLAT = 1, - SCSI_LUN_ADDR_METHOD_LUN = 2, - SCSI_LUN_ADDR_METHOD_EXTENDED_LUN = 3, -}; - -/* - * srpt_unpack_lun() - Convert from network LUN to linear LUN. - * - * Convert an 2-byte, 4-byte, 6-byte or 8-byte LUN structure in network byte - * order (big endian) to a linear LUN. Supports three LUN addressing methods: - * peripheral, flat and logical unit. See also SAM-2, section 4.9.4 (page 40). - */ -static uint64_t srpt_unpack_lun(const uint8_t *lun, int len) -{ - uint64_t res = NO_SUCH_LUN; - int addressing_method; - - if (unlikely(len < 2)) { - pr_err("Illegal LUN length %d, expected 2 bytes or more\n", - len); - goto out; - } - - switch (len) { - case 8: - if ((*((__be64 *)lun) & - cpu_to_be64(0x0000FFFFFFFFFFFFLL)) != 0) - goto out_err; - break; - case 4: - if (*((__be16 *)&lun[2]) != 0) - goto out_err; - break; - case 6: - if (*((__be32 *)&lun[2]) != 0) - goto out_err; - break; - case 2: - break; - default: - goto out_err; - } - - addressing_method = (*lun) >> 6; /* highest two bits of byte 0 */ - switch (addressing_method) { - case SCSI_LUN_ADDR_METHOD_PERIPHERAL: - case SCSI_LUN_ADDR_METHOD_FLAT: - case SCSI_LUN_ADDR_METHOD_LUN: - res = *(lun + 1) | (((*lun) & 0x3f) << 8); - break; - - case SCSI_LUN_ADDR_METHOD_EXTENDED_LUN: - default: - pr_err("Unimplemented LUN addressing method %u\n", - addressing_method); - break; - } - -out: - return res; - -out_err: - pr_err("Support for multi-level LUNs has not yet been implemented\n"); - goto out; -} - static int srpt_check_stop_free(struct se_cmd *cmd) { struct srpt_send_ioctx *ioctx = container_of(cmd, @@ -1613,16 +1531,14 @@ static int srpt_check_stop_free(struct se_cmd *cmd) /** * srpt_handle_cmd() - Process SRP_CMD. */ -static int srpt_handle_cmd(struct srpt_rdma_ch *ch, - struct srpt_recv_ioctx *recv_ioctx, - struct srpt_send_ioctx *send_ioctx) +static void srpt_handle_cmd(struct srpt_rdma_ch *ch, + struct srpt_recv_ioctx *recv_ioctx, + struct srpt_send_ioctx *send_ioctx) { struct se_cmd *cmd; struct srp_cmd *srp_cmd; - uint64_t unpacked_lun; u64 data_len; enum dma_data_direction dir; - sense_reason_t ret; int rc; BUG_ON(!send_ioctx); @@ -1650,65 +1566,23 @@ static int srpt_handle_cmd(struct srpt_rdma_ch *ch, if (srpt_get_desc_tbl(send_ioctx, srp_cmd, &dir, &data_len)) { pr_err("0x%llx: parsing SRP descriptor table failed.\n", srp_cmd->tag); - ret = TCM_INVALID_CDB_FIELD; - goto send_sense; + goto release_ioctx; } - unpacked_lun = srpt_unpack_lun((uint8_t *)&srp_cmd->lun, - sizeof(srp_cmd->lun)); rc = target_submit_cmd(cmd, ch->sess, srp_cmd->cdb, - &send_ioctx->sense_data[0], unpacked_lun, data_len, - TCM_SIMPLE_TAG, dir, TARGET_SCF_ACK_KREF); + &send_ioctx->sense_data[0], + scsilun_to_int(&srp_cmd->lun), data_len, + TCM_SIMPLE_TAG, dir, TARGET_SCF_ACK_KREF); if (rc != 0) { - ret = TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE; - goto send_sense; + pr_debug("target_submit_cmd() returned %d for tag %#llx\n", rc, + srp_cmd->tag); + goto release_ioctx; } - return 0; - -send_sense: - transport_send_check_condition_and_sense(cmd, ret, 0); - return -1; -} - -/** - * srpt_rx_mgmt_fn_tag() - Process a task management function by tag. - * @ch: RDMA channel of the task management request. - * @fn: Task management function to perform. - * @req_tag: Tag of the SRP task management request. - * @mgmt_ioctx: I/O context of the task management request. - * - * Returns zero if the target core will process the task management - * request asynchronously. - * - * Note: It is assumed that the initiator serializes tag-based task management - * requests. - */ -static int srpt_rx_mgmt_fn_tag(struct srpt_send_ioctx *ioctx, u64 tag) -{ - struct srpt_device *sdev; - struct srpt_rdma_ch *ch; - struct srpt_send_ioctx *target; - int ret, i; + return; - ret = -EINVAL; - ch = ioctx->ch; - BUG_ON(!ch); - BUG_ON(!ch->sport); - sdev = ch->sport->sdev; - BUG_ON(!sdev); - spin_lock_irq(&sdev->spinlock); - for (i = 0; i < ch->rq_size; ++i) { - target = ch->ioctx_ring[i]; - if (target->cmd.se_lun == ioctx->cmd.se_lun && - target->cmd.tag == tag && - srpt_get_cmd_state(target) != SRPT_STATE_DONE) { - ret = 0; - /* now let the target core abort &target->cmd; */ - break; - } - } - spin_unlock_irq(&sdev->spinlock); - return ret; +release_ioctx: + send_ioctx->state = SRPT_STATE_DONE; + srpt_release_cmd(cmd); } static int srp_tmr_to_tcm(int fn) @@ -1744,8 +1618,6 @@ static void srpt_handle_tsk_mgmt(struct srpt_rdma_ch *ch, struct srp_tsk_mgmt *srp_tsk; struct se_cmd *cmd; struct se_session *sess = ch->sess; - uint64_t unpacked_lun; - uint32_t tag = 0; int tcm_tmr; int rc; @@ -1761,26 +1633,10 @@ static void srpt_handle_tsk_mgmt(struct srpt_rdma_ch *ch, srpt_set_cmd_state(send_ioctx, SRPT_STATE_MGMT); send_ioctx->cmd.tag = srp_tsk->tag; tcm_tmr = srp_tmr_to_tcm(srp_tsk->tsk_mgmt_func); - if (tcm_tmr < 0) { - send_ioctx->cmd.se_tmr_req->response = - TMR_TASK_MGMT_FUNCTION_NOT_SUPPORTED; - goto fail; - } - unpacked_lun = srpt_unpack_lun((uint8_t *)&srp_tsk->lun, - sizeof(srp_tsk->lun)); - - if (srp_tsk->tsk_mgmt_func == SRP_TSK_ABORT_TASK) { - rc = srpt_rx_mgmt_fn_tag(send_ioctx, srp_tsk->task_tag); - if (rc < 0) { - send_ioctx->cmd.se_tmr_req->response = - TMR_TASK_DOES_NOT_EXIST; - goto fail; - } - tag = srp_tsk->task_tag; - } - rc = target_submit_tmr(&send_ioctx->cmd, sess, NULL, unpacked_lun, - srp_tsk, tcm_tmr, GFP_KERNEL, tag, - TARGET_SCF_ACK_KREF); + rc = target_submit_tmr(&send_ioctx->cmd, sess, NULL, + scsilun_to_int(&srp_tsk->lun), srp_tsk, tcm_tmr, + GFP_KERNEL, srp_tsk->task_tag, + TARGET_SCF_ACK_KREF); if (rc != 0) { send_ioctx->cmd.se_tmr_req->response = TMR_FUNCTION_REJECTED; goto fail; @@ -1800,7 +1656,6 @@ static void srpt_handle_new_iu(struct srpt_rdma_ch *ch, struct srpt_send_ioctx *send_ioctx) { struct srp_cmd *srp_cmd; - enum rdma_ch_state ch_state; BUG_ON(!ch); BUG_ON(!recv_ioctx); @@ -1809,13 +1664,12 @@ static void srpt_handle_new_iu(struct srpt_rdma_ch *ch, recv_ioctx->ioctx.dma, srp_max_req_size, DMA_FROM_DEVICE); - ch_state = srpt_get_ch_state(ch); - if (unlikely(ch_state == CH_CONNECTING)) { + if (unlikely(ch->state == CH_CONNECTING)) { list_add_tail(&recv_ioctx->wait_list, &ch->cmd_wait_list); goto out; } - if (unlikely(ch_state != CH_LIVE)) + if (unlikely(ch->state != CH_LIVE)) goto out; srp_cmd = recv_ioctx->ioctx.buf; @@ -1878,6 +1732,28 @@ static void srpt_recv_done(struct ib_cq *cq, struct ib_wc *wc) } } +/* + * This function must be called from the context in which RDMA completions are + * processed because it accesses the wait list without protection against + * access from other threads. + */ +static void srpt_process_wait_list(struct srpt_rdma_ch *ch) +{ + struct srpt_send_ioctx *ioctx; + + while (!list_empty(&ch->cmd_wait_list) && + ch->state >= CH_LIVE && + (ioctx = srpt_get_send_ioctx(ch)) != NULL) { + struct srpt_recv_ioctx *recv_ioctx; + + recv_ioctx = list_first_entry(&ch->cmd_wait_list, + struct srpt_recv_ioctx, + wait_list); + list_del(&recv_ioctx->wait_list); + srpt_handle_new_iu(ch, recv_ioctx, ioctx); + } +} + /** * Note: Although this has not yet been observed during tests, at least in * theory it is possible that the srpt_get_send_ioctx() call invoked by @@ -1905,15 +1781,10 @@ static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc) atomic_inc(&ch->sq_wr_avail); - if (wc->status != IB_WC_SUCCESS) { + if (wc->status != IB_WC_SUCCESS) pr_info("sending response for ioctx 0x%p failed" " with status %d\n", ioctx, wc->status); - atomic_dec(&ch->req_lim); - srpt_abort_cmd(ioctx); - goto out; - } - if (state != SRPT_STATE_DONE) { srpt_unmap_sg_to_ib_sge(ch, ioctx); transport_generic_free_cmd(&ioctx->cmd, 0); @@ -1922,18 +1793,7 @@ static void srpt_send_done(struct ib_cq *cq, struct ib_wc *wc) " wr_id = %u.\n", ioctx->ioctx.index); } -out: - while (!list_empty(&ch->cmd_wait_list) && - srpt_get_ch_state(ch) == CH_LIVE && - (ioctx = srpt_get_send_ioctx(ch)) != NULL) { - struct srpt_recv_ioctx *recv_ioctx; - - recv_ioctx = list_first_entry(&ch->cmd_wait_list, - struct srpt_recv_ioctx, - wait_list); - list_del(&recv_ioctx->wait_list); - srpt_handle_new_iu(ch, recv_ioctx, ioctx); - } + srpt_process_wait_list(ch); } /** @@ -1950,7 +1810,7 @@ static int srpt_create_ch_ib(struct srpt_rdma_ch *ch) WARN_ON(ch->rq_size < 1); ret = -ENOMEM; - qp_init = kzalloc(sizeof *qp_init, GFP_KERNEL); + qp_init = kzalloc(sizeof(*qp_init), GFP_KERNEL); if (!qp_init) goto out; @@ -2017,168 +1877,102 @@ static void srpt_destroy_ch_ib(struct srpt_rdma_ch *ch) } /** - * __srpt_close_ch() - Close an RDMA channel by setting the QP error state. + * srpt_close_ch() - Close an RDMA channel. * - * Reset the QP and make sure all resources associated with the channel will - * be deallocated at an appropriate time. + * Make sure all resources associated with the channel will be deallocated at + * an appropriate time. * - * Note: The caller must hold ch->sport->sdev->spinlock. + * Returns true if and only if the channel state has been modified into + * CH_DRAINING. */ -static void __srpt_close_ch(struct srpt_rdma_ch *ch) +static bool srpt_close_ch(struct srpt_rdma_ch *ch) { - enum rdma_ch_state prev_state; - unsigned long flags; + int ret; - spin_lock_irqsave(&ch->spinlock, flags); - prev_state = ch->state; - switch (prev_state) { - case CH_CONNECTING: - case CH_LIVE: - ch->state = CH_DISCONNECTING; - break; - default: - break; + if (!srpt_set_ch_state(ch, CH_DRAINING)) { + pr_debug("%s-%d: already closed\n", ch->sess_name, + ch->qp->qp_num); + return false; } - spin_unlock_irqrestore(&ch->spinlock, flags); - - switch (prev_state) { - case CH_CONNECTING: - ib_send_cm_rej(ch->cm_id, IB_CM_REJ_NO_RESOURCES, NULL, 0, - NULL, 0); - /* fall through */ - case CH_LIVE: - if (ib_send_cm_dreq(ch->cm_id, NULL, 0) < 0) - pr_err("sending CM DREQ failed.\n"); - break; - case CH_DISCONNECTING: - break; - case CH_DRAINING: - case CH_RELEASING: - break; - } -} - -/** - * srpt_close_ch() - Close an RDMA channel. - */ -static void srpt_close_ch(struct srpt_rdma_ch *ch) -{ - struct srpt_device *sdev; - sdev = ch->sport->sdev; - spin_lock_irq(&sdev->spinlock); - __srpt_close_ch(ch); - spin_unlock_irq(&sdev->spinlock); -} + kref_get(&ch->kref); -/** - * srpt_shutdown_session() - Whether or not a session may be shut down. - */ -static int srpt_shutdown_session(struct se_session *se_sess) -{ - struct srpt_rdma_ch *ch = se_sess->fabric_sess_ptr; - unsigned long flags; + ret = srpt_ch_qp_err(ch); + if (ret < 0) + pr_err("%s-%d: changing queue pair into error state failed: %d\n", + ch->sess_name, ch->qp->qp_num, ret); - spin_lock_irqsave(&ch->spinlock, flags); - if (ch->in_shutdown) { - spin_unlock_irqrestore(&ch->spinlock, flags); - return true; + pr_debug("%s-%d: queued zerolength write\n", ch->sess_name, + ch->qp->qp_num); + ret = srpt_zerolength_write(ch); + if (ret < 0) { + pr_err("%s-%d: queuing zero-length write failed: %d\n", + ch->sess_name, ch->qp->qp_num, ret); + if (srpt_set_ch_state(ch, CH_DISCONNECTED)) + schedule_work(&ch->release_work); + else + WARN_ON_ONCE(true); } - ch->in_shutdown = true; - target_sess_cmd_list_set_waiting(se_sess); - spin_unlock_irqrestore(&ch->spinlock, flags); + kref_put(&ch->kref, srpt_free_ch); return true; } -/** - * srpt_drain_channel() - Drain a channel by resetting the IB queue pair. - * @cm_id: Pointer to the CM ID of the channel to be drained. - * - * Note: Must be called from inside srpt_cm_handler to avoid a race between - * accessing sdev->spinlock and the call to kfree(sdev) in srpt_remove_one() - * (the caller of srpt_cm_handler holds the cm_id spinlock; srpt_remove_one() - * waits until all target sessions for the associated IB device have been - * unregistered and target session registration involves a call to - * ib_destroy_cm_id(), which locks the cm_id spinlock and hence waits until - * this function has finished). +/* + * Change the channel state into CH_DISCONNECTING. If a channel has not yet + * reached the connected state, close it. If a channel is in the connected + * state, send a DREQ. If a DREQ has been received, send a DREP. Note: it is + * the responsibility of the caller to ensure that this function is not + * invoked concurrently with the code that accepts a connection. This means + * that this function must either be invoked from inside a CM callback + * function or that it must be invoked with the srpt_port.mutex held. */ -static void srpt_drain_channel(struct ib_cm_id *cm_id) +static int srpt_disconnect_ch(struct srpt_rdma_ch *ch) { - struct srpt_device *sdev; - struct srpt_rdma_ch *ch; int ret; - bool do_reset = false; - WARN_ON_ONCE(irqs_disabled()); + if (!srpt_set_ch_state(ch, CH_DISCONNECTING)) + return -ENOTCONN; - sdev = cm_id->context; - BUG_ON(!sdev); - spin_lock_irq(&sdev->spinlock); - list_for_each_entry(ch, &sdev->rch_list, list) { - if (ch->cm_id == cm_id) { - do_reset = srpt_test_and_set_ch_state(ch, - CH_CONNECTING, CH_DRAINING) || - srpt_test_and_set_ch_state(ch, - CH_LIVE, CH_DRAINING) || - srpt_test_and_set_ch_state(ch, - CH_DISCONNECTING, CH_DRAINING); - break; - } - } - spin_unlock_irq(&sdev->spinlock); + ret = ib_send_cm_dreq(ch->cm_id, NULL, 0); + if (ret < 0) + ret = ib_send_cm_drep(ch->cm_id, NULL, 0); - if (do_reset) { - if (ch->sess) - srpt_shutdown_session(ch->sess); + if (ret < 0 && srpt_close_ch(ch)) + ret = 0; - ret = srpt_ch_qp_err(ch); - if (ret < 0) - pr_err("Setting queue pair in error state" - " failed: %d\n", ret); - } + return ret; } -/** - * srpt_find_channel() - Look up an RDMA channel. - * @cm_id: Pointer to the CM ID of the channel to be looked up. - * - * Return NULL if no matching RDMA channel has been found. - */ -static struct srpt_rdma_ch *srpt_find_channel(struct srpt_device *sdev, - struct ib_cm_id *cm_id) +static void __srpt_close_all_ch(struct srpt_device *sdev) { struct srpt_rdma_ch *ch; - bool found; - WARN_ON_ONCE(irqs_disabled()); - BUG_ON(!sdev); + lockdep_assert_held(&sdev->mutex); - found = false; - spin_lock_irq(&sdev->spinlock); list_for_each_entry(ch, &sdev->rch_list, list) { - if (ch->cm_id == cm_id) { - found = true; - break; - } + if (srpt_disconnect_ch(ch) >= 0) + pr_info("Closing channel %s-%d because target %s has been disabled\n", + ch->sess_name, ch->qp->qp_num, + sdev->device->name); + srpt_close_ch(ch); } - spin_unlock_irq(&sdev->spinlock); - - return found ? ch : NULL; } /** - * srpt_release_channel() - Release channel resources. - * - * Schedules the actual release because: - * - Calling the ib_destroy_cm_id() call from inside an IB CM callback would - * trigger a deadlock. - * - It is not safe to call TCM transport_* functions from interrupt context. + * srpt_shutdown_session() - Whether or not a session may be shut down. */ -static void srpt_release_channel(struct srpt_rdma_ch *ch) +static int srpt_shutdown_session(struct se_session *se_sess) +{ + return 1; +} + +static void srpt_free_ch(struct kref *kref) { - schedule_work(&ch->release_work); + struct srpt_rdma_ch *ch = container_of(kref, struct srpt_rdma_ch, kref); + + kfree(ch); } static void srpt_release_channel_work(struct work_struct *w) @@ -2188,8 +1982,8 @@ static void srpt_release_channel_work(struct work_struct *w) struct se_session *se_sess; ch = container_of(w, struct srpt_rdma_ch, release_work); - pr_debug("ch = %p; ch->sess = %p; release_done = %p\n", ch, ch->sess, - ch->release_done); + pr_debug("%s: %s-%d; release_done = %p\n", __func__, ch->sess_name, + ch->qp->qp_num, ch->release_done); sdev = ch->sport->sdev; BUG_ON(!sdev); @@ -2197,6 +1991,7 @@ static void srpt_release_channel_work(struct work_struct *w) se_sess = ch->sess; BUG_ON(!se_sess); + target_sess_cmd_list_set_waiting(se_sess); target_wait_for_sess_cmds(se_sess); transport_deregister_session_configfs(se_sess); @@ -2211,16 +2006,15 @@ static void srpt_release_channel_work(struct work_struct *w) ch->sport->sdev, ch->rq_size, ch->rsp_size, DMA_TO_DEVICE); - spin_lock_irq(&sdev->spinlock); - list_del(&ch->list); - spin_unlock_irq(&sdev->spinlock); - + mutex_lock(&sdev->mutex); + list_del_init(&ch->list); if (ch->release_done) complete(ch->release_done); + mutex_unlock(&sdev->mutex); wake_up(&sdev->ch_releaseQ); - kfree(ch); + kref_put(&ch->kref, srpt_free_ch); } /** @@ -2266,9 +2060,9 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, be64_to_cpu(*(__be64 *)&sdev->port[param->port - 1].gid.raw[0]), be64_to_cpu(*(__be64 *)&sdev->port[param->port - 1].gid.raw[8])); - rsp = kzalloc(sizeof *rsp, GFP_KERNEL); - rej = kzalloc(sizeof *rej, GFP_KERNEL); - rep_param = kzalloc(sizeof *rep_param, GFP_KERNEL); + rsp = kzalloc(sizeof(*rsp), GFP_KERNEL); + rej = kzalloc(sizeof(*rej), GFP_KERNEL); + rep_param = kzalloc(sizeof(*rep_param), GFP_KERNEL); if (!rsp || !rej || !rep_param) { ret = -ENOMEM; @@ -2297,7 +2091,7 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, if ((req->req_flags & SRP_MTCH_ACTION) == SRP_MULTICHAN_SINGLE) { rsp->rsp_flags = SRP_LOGIN_RSP_MULTICHAN_NO_CHAN; - spin_lock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); list_for_each_entry_safe(ch, tmp_ch, &sdev->rch_list, list) { if (!memcmp(ch->i_port_id, req->initiator_port_id, 16) @@ -2305,26 +2099,16 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, && param->port == ch->sport->port && param->listen_id == ch->sport->sdev->cm_id && ch->cm_id) { - enum rdma_ch_state ch_state; - - ch_state = srpt_get_ch_state(ch); - if (ch_state != CH_CONNECTING - && ch_state != CH_LIVE) + if (srpt_disconnect_ch(ch) < 0) continue; - - /* found an existing channel */ - pr_debug("Found existing channel %s" - " cm_id= %p state= %d\n", - ch->sess_name, ch->cm_id, ch_state); - - __srpt_close_ch(ch); - + pr_info("Relogin - closed existing channel %s\n", + ch->sess_name); rsp->rsp_flags = SRP_LOGIN_RSP_MULTICHAN_TERMINATED; } } - spin_unlock_irq(&sdev->spinlock); + mutex_unlock(&sdev->mutex); } else rsp->rsp_flags = SRP_LOGIN_RSP_MULTICHAN_MAINTAINED; @@ -2340,7 +2124,7 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, goto reject; } - ch = kzalloc(sizeof *ch, GFP_KERNEL); + ch = kzalloc(sizeof(*ch), GFP_KERNEL); if (!ch) { rej->reason = cpu_to_be32( SRP_LOGIN_REJ_INSUFFICIENT_RESOURCES); @@ -2349,11 +2133,14 @@ static int srpt_cm_req_recv(struct ib_cm_id *cm_id, goto reject; } + kref_init(&ch->kref); + ch->zw_cqe.done = srpt_zerolength_write_done; INIT_WORK(&ch->release_work, srpt_release_channel_work); memcpy(ch->i_port_id, req->initiator_port_id, 16); memcpy(ch->t_port_id, req->target_port_id, 16); ch->sport = &sdev->port[param->port - 1]; ch->cm_id = cm_id; + cm_id->context = ch; /* * Avoid QUEUE_FULL conditions by limiting the number of buffers used * for the SRP protocol to the command queue size. @@ -2453,7 +2240,7 @@ try_again: /* create cm reply */ rep_param->qp_num = ch->qp->qp_num; rep_param->private_data = (void *)rsp; - rep_param->private_data_len = sizeof *rsp; + rep_param->private_data_len = sizeof(*rsp); rep_param->rnr_retry_count = 7; rep_param->flow_control = 1; rep_param->failover_accepted = 0; @@ -2468,14 +2255,14 @@ try_again: goto release_channel; } - spin_lock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); list_add_tail(&ch->list, &sdev->rch_list); - spin_unlock_irq(&sdev->spinlock); + mutex_unlock(&sdev->mutex); goto out; release_channel: - srpt_set_ch_state(ch, CH_RELEASING); + srpt_disconnect_ch(ch); transport_deregister_session_configfs(ch->sess); transport_deregister_session(ch->sess); ch->sess = NULL; @@ -2497,7 +2284,7 @@ reject: | SRP_BUF_FORMAT_INDIRECT); ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED, NULL, 0, - (void *)rej, sizeof *rej); + (void *)rej, sizeof(*rej)); out: kfree(rep_param); @@ -2507,10 +2294,23 @@ out: return ret; } -static void srpt_cm_rej_recv(struct ib_cm_id *cm_id) +static void srpt_cm_rej_recv(struct srpt_rdma_ch *ch, + enum ib_cm_rej_reason reason, + const u8 *private_data, + u8 private_data_len) { - pr_info("Received IB REJ for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); + char *priv = NULL; + int i; + + if (private_data_len && (priv = kmalloc(private_data_len * 3 + 1, + GFP_KERNEL))) { + for (i = 0; i < private_data_len; i++) + sprintf(priv + 3 * i, " %02x", private_data[i]); + } + pr_info("Received CM REJ for ch %s-%d; reason %d%s%s.\n", + ch->sess_name, ch->qp->qp_num, reason, private_data_len ? + "; private data" : "", priv ? priv : " (?)"); + kfree(priv); } /** @@ -2519,87 +2319,23 @@ static void srpt_cm_rej_recv(struct ib_cm_id *cm_id) * An IB_CM_RTU_RECEIVED message indicates that the connection is established * and that the recipient may begin transmitting (RTU = ready to use). */ -static void srpt_cm_rtu_recv(struct ib_cm_id *cm_id) +static void srpt_cm_rtu_recv(struct srpt_rdma_ch *ch) { - struct srpt_rdma_ch *ch; int ret; - ch = srpt_find_channel(cm_id->context, cm_id); - BUG_ON(!ch); - - if (srpt_test_and_set_ch_state(ch, CH_CONNECTING, CH_LIVE)) { - struct srpt_recv_ioctx *ioctx, *ioctx_tmp; - + if (srpt_set_ch_state(ch, CH_LIVE)) { ret = srpt_ch_qp_rts(ch, ch->qp); - list_for_each_entry_safe(ioctx, ioctx_tmp, &ch->cmd_wait_list, - wait_list) { - list_del(&ioctx->wait_list); - srpt_handle_new_iu(ch, ioctx, NULL); - } - if (ret) + if (ret == 0) { + /* Trigger wait list processing. */ + ret = srpt_zerolength_write(ch); + WARN_ONCE(ret < 0, "%d\n", ret); + } else { srpt_close_ch(ch); + } } } -static void srpt_cm_timewait_exit(struct ib_cm_id *cm_id) -{ - pr_info("Received IB TimeWait exit for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); -} - -static void srpt_cm_rep_error(struct ib_cm_id *cm_id) -{ - pr_info("Received IB REP error for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); -} - -/** - * srpt_cm_dreq_recv() - Process reception of a DREQ message. - */ -static void srpt_cm_dreq_recv(struct ib_cm_id *cm_id) -{ - struct srpt_rdma_ch *ch; - unsigned long flags; - bool send_drep = false; - - ch = srpt_find_channel(cm_id->context, cm_id); - BUG_ON(!ch); - - pr_debug("cm_id= %p ch->state= %d\n", cm_id, srpt_get_ch_state(ch)); - - spin_lock_irqsave(&ch->spinlock, flags); - switch (ch->state) { - case CH_CONNECTING: - case CH_LIVE: - send_drep = true; - ch->state = CH_DISCONNECTING; - break; - case CH_DISCONNECTING: - case CH_DRAINING: - case CH_RELEASING: - WARN(true, "unexpected channel state %d\n", ch->state); - break; - } - spin_unlock_irqrestore(&ch->spinlock, flags); - - if (send_drep) { - if (ib_send_cm_drep(ch->cm_id, NULL, 0) < 0) - pr_err("Sending IB DREP failed.\n"); - pr_info("Received DREQ and sent DREP for session %s.\n", - ch->sess_name); - } -} - -/** - * srpt_cm_drep_recv() - Process reception of a DREP message. - */ -static void srpt_cm_drep_recv(struct ib_cm_id *cm_id) -{ - pr_info("Received InfiniBand DREP message for cm_id %p.\n", cm_id); - srpt_drain_channel(cm_id); -} - /** * srpt_cm_handler() - IB connection manager callback function. * @@ -2612,6 +2348,7 @@ static void srpt_cm_drep_recv(struct ib_cm_id *cm_id) */ static int srpt_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) { + struct srpt_rdma_ch *ch = cm_id->context; int ret; ret = 0; @@ -2621,32 +2358,39 @@ static int srpt_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) event->private_data); break; case IB_CM_REJ_RECEIVED: - srpt_cm_rej_recv(cm_id); + srpt_cm_rej_recv(ch, event->param.rej_rcvd.reason, + event->private_data, + IB_CM_REJ_PRIVATE_DATA_SIZE); break; case IB_CM_RTU_RECEIVED: case IB_CM_USER_ESTABLISHED: - srpt_cm_rtu_recv(cm_id); + srpt_cm_rtu_recv(ch); break; case IB_CM_DREQ_RECEIVED: - srpt_cm_dreq_recv(cm_id); + srpt_disconnect_ch(ch); break; case IB_CM_DREP_RECEIVED: - srpt_cm_drep_recv(cm_id); + pr_info("Received CM DREP message for ch %s-%d.\n", + ch->sess_name, ch->qp->qp_num); + srpt_close_ch(ch); break; case IB_CM_TIMEWAIT_EXIT: - srpt_cm_timewait_exit(cm_id); + pr_info("Received CM TimeWait exit for ch %s-%d.\n", + ch->sess_name, ch->qp->qp_num); + srpt_close_ch(ch); break; case IB_CM_REP_ERROR: - srpt_cm_rep_error(cm_id); + pr_info("Received CM REP error for ch %s-%d.\n", ch->sess_name, + ch->qp->qp_num); break; case IB_CM_DREQ_ERROR: - pr_info("Received IB DREQ ERROR event.\n"); + pr_info("Received CM DREQ ERROR event.\n"); break; case IB_CM_MRA_RECEIVED: - pr_info("Received IB MRA event\n"); + pr_info("Received CM MRA event\n"); break; default: - pr_err("received unrecognized IB CM event %d\n", event->event); + pr_err("received unrecognized CM event %d\n", event->event); break; } @@ -2755,41 +2499,14 @@ static int srpt_write_pending_status(struct se_cmd *se_cmd) */ static int srpt_write_pending(struct se_cmd *se_cmd) { - struct srpt_rdma_ch *ch; - struct srpt_send_ioctx *ioctx; + struct srpt_send_ioctx *ioctx = + container_of(se_cmd, struct srpt_send_ioctx, cmd); + struct srpt_rdma_ch *ch = ioctx->ch; enum srpt_command_state new_state; - enum rdma_ch_state ch_state; - int ret; - - ioctx = container_of(se_cmd, struct srpt_send_ioctx, cmd); new_state = srpt_set_cmd_state(ioctx, SRPT_STATE_NEED_DATA); WARN_ON(new_state == SRPT_STATE_DONE); - - ch = ioctx->ch; - BUG_ON(!ch); - - ch_state = srpt_get_ch_state(ch); - switch (ch_state) { - case CH_CONNECTING: - WARN(true, "unexpected channel state %d\n", ch_state); - ret = -EINVAL; - goto out; - case CH_LIVE: - break; - case CH_DISCONNECTING: - case CH_DRAINING: - case CH_RELEASING: - pr_debug("cmd with tag %lld: channel disconnecting\n", - ioctx->cmd.tag); - srpt_set_cmd_state(ioctx, SRPT_STATE_DATA_IN); - ret = -EINVAL; - goto out; - } - ret = srpt_xfer_data(ch, ioctx); - -out: - return ret; + return srpt_xfer_data(ch, ioctx); } static u8 tcm_to_srp_tsk_mgmt_status(const int tcm_mgmt_status) @@ -2920,36 +2637,25 @@ static void srpt_refresh_port_work(struct work_struct *work) srpt_refresh_port(sport); } -static int srpt_ch_list_empty(struct srpt_device *sdev) -{ - int res; - - spin_lock_irq(&sdev->spinlock); - res = list_empty(&sdev->rch_list); - spin_unlock_irq(&sdev->spinlock); - - return res; -} - /** * srpt_release_sdev() - Free the channel resources associated with a target. */ static int srpt_release_sdev(struct srpt_device *sdev) { - struct srpt_rdma_ch *ch, *tmp_ch; - int res; + int i, res; WARN_ON_ONCE(irqs_disabled()); BUG_ON(!sdev); - spin_lock_irq(&sdev->spinlock); - list_for_each_entry_safe(ch, tmp_ch, &sdev->rch_list, list) - __srpt_close_ch(ch); - spin_unlock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); + for (i = 0; i < ARRAY_SIZE(sdev->port); i++) + sdev->port[i].enabled = false; + __srpt_close_all_ch(sdev); + mutex_unlock(&sdev->mutex); res = wait_event_interruptible(sdev->ch_releaseQ, - srpt_ch_list_empty(sdev)); + list_empty_careful(&sdev->rch_list)); if (res) pr_err("%s: interrupted.\n", __func__); @@ -3003,14 +2709,14 @@ static void srpt_add_one(struct ib_device *device) pr_debug("device = %p, device->dma_ops = %p\n", device, device->dma_ops); - sdev = kzalloc(sizeof *sdev, GFP_KERNEL); + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL); if (!sdev) goto err; sdev->device = device; INIT_LIST_HEAD(&sdev->rch_list); init_waitqueue_head(&sdev->ch_releaseQ); - spin_lock_init(&sdev->spinlock); + mutex_init(&sdev->mutex); sdev->pd = ib_alloc_pd(device); if (IS_ERR(sdev->pd)) @@ -3082,7 +2788,7 @@ static void srpt_add_one(struct ib_device *device) if (srpt_refresh_port(sport)) { pr_err("MAD registration failed for %s-%d.\n", - srpt_sdev_name(sdev), i); + sdev->device->name, i); goto err_ring; } snprintf(sport->port_guid, sizeof(sport->port_guid), @@ -3231,24 +2937,26 @@ static void srpt_release_cmd(struct se_cmd *se_cmd) static void srpt_close_session(struct se_session *se_sess) { DECLARE_COMPLETION_ONSTACK(release_done); - struct srpt_rdma_ch *ch; - struct srpt_device *sdev; - unsigned long res; - - ch = se_sess->fabric_sess_ptr; - WARN_ON(ch->sess != se_sess); + struct srpt_rdma_ch *ch = se_sess->fabric_sess_ptr; + struct srpt_device *sdev = ch->sport->sdev; + bool wait; - pr_debug("ch %p state %d\n", ch, srpt_get_ch_state(ch)); + pr_debug("ch %s-%d state %d\n", ch->sess_name, ch->qp->qp_num, + ch->state); - sdev = ch->sport->sdev; - spin_lock_irq(&sdev->spinlock); + mutex_lock(&sdev->mutex); BUG_ON(ch->release_done); ch->release_done = &release_done; - __srpt_close_ch(ch); - spin_unlock_irq(&sdev->spinlock); + wait = !list_empty(&ch->list); + srpt_disconnect_ch(ch); + mutex_unlock(&sdev->mutex); - res = wait_for_completion_timeout(&release_done, 60 * HZ); - WARN_ON(res == 0); + if (!wait) + return; + + while (wait_for_completion_timeout(&release_done, 180 * HZ) == 0) + pr_info("%s(%s-%d state %d): still waiting ...\n", __func__, + ch->sess_name, ch->qp->qp_num, ch->state); } /** @@ -3456,6 +3164,8 @@ static ssize_t srpt_tpg_enable_store(struct config_item *item, { struct se_portal_group *se_tpg = to_tpg(item); struct srpt_port *sport = container_of(se_tpg, struct srpt_port, port_tpg_1); + struct srpt_device *sdev = sport->sdev; + struct srpt_rdma_ch *ch; unsigned long tmp; int ret; @@ -3469,11 +3179,24 @@ static ssize_t srpt_tpg_enable_store(struct config_item *item, pr_err("Illegal value for srpt_tpg_store_enable: %lu\n", tmp); return -EINVAL; } - if (tmp == 1) - sport->enabled = true; - else - sport->enabled = false; + if (sport->enabled == tmp) + goto out; + sport->enabled = tmp; + if (sport->enabled) + goto out; + + mutex_lock(&sdev->mutex); + list_for_each_entry(ch, &sdev->rch_list, list) { + if (ch->sport == sport) { + pr_debug("%s: ch %p %s-%d\n", __func__, ch, + ch->sess_name, ch->qp->qp_num); + srpt_disconnect_ch(ch); + srpt_close_ch(ch); + } + } + mutex_unlock(&sdev->mutex); +out: return count; } @@ -3565,7 +3288,6 @@ static struct configfs_attribute *srpt_wwn_attrs[] = { static const struct target_core_fabric_ops srpt_template = { .module = THIS_MODULE, .name = "srpt", - .node_acl_size = sizeof(struct srpt_node_acl), .get_fabric_name = srpt_get_fabric_name, .tpg_get_wwn = srpt_get_fabric_wwn, .tpg_get_tag = srpt_get_tag, diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.h b/drivers/infiniband/ulp/srpt/ib_srpt.h index 09037f2b0b51..af9b8b527340 100644 --- a/drivers/infiniband/ulp/srpt/ib_srpt.h +++ b/drivers/infiniband/ulp/srpt/ib_srpt.h @@ -218,20 +218,20 @@ struct srpt_send_ioctx { /** * enum rdma_ch_state - SRP channel state. - * @CH_CONNECTING: QP is in RTR state; waiting for RTU. - * @CH_LIVE: QP is in RTS state. - * @CH_DISCONNECTING: DREQ has been received; waiting for DREP - * or DREQ has been send and waiting for DREP - * or . - * @CH_DRAINING: QP is in ERR state; waiting for last WQE event. - * @CH_RELEASING: Last WQE event has been received; releasing resources. + * @CH_CONNECTING: QP is in RTR state; waiting for RTU. + * @CH_LIVE: QP is in RTS state. + * @CH_DISCONNECTING: DREQ has been sent and waiting for DREP or DREQ has + * been received. + * @CH_DRAINING: DREP has been received or waiting for DREP timed out + * and last work request has been queued. + * @CH_DISCONNECTED: Last completion has been received. */ enum rdma_ch_state { CH_CONNECTING, CH_LIVE, CH_DISCONNECTING, CH_DRAINING, - CH_RELEASING + CH_DISCONNECTED, }; /** @@ -267,6 +267,8 @@ struct srpt_rdma_ch { struct ib_cm_id *cm_id; struct ib_qp *qp; struct ib_cq *cq; + struct ib_cqe zw_cqe; + struct kref kref; int rq_size; u32 rsp_size; atomic_t sq_wr_avail; @@ -286,7 +288,6 @@ struct srpt_rdma_ch { u8 sess_name[36]; struct work_struct release_work; struct completion *release_done; - bool in_shutdown; }; /** @@ -343,7 +344,7 @@ struct srpt_port { * @ioctx_ring: Per-HCA SRQ. * @rch_list: Per-device channel list -- see also srpt_rdma_ch.list. * @ch_releaseQ: Enables waiting for removal from rch_list. - * @spinlock: Protects rch_list and tpg. + * @mutex: Protects rch_list. * @port: Information about the ports owned by this HCA. * @event_handler: Per-HCA asynchronous IB event handler. * @list: Node in srpt_dev_list. @@ -357,18 +358,10 @@ struct srpt_device { struct srpt_recv_ioctx **ioctx_ring; struct list_head rch_list; wait_queue_head_t ch_releaseQ; - spinlock_t spinlock; + struct mutex mutex; struct srpt_port port[2]; struct ib_event_handler event_handler; struct list_head list; }; -/** - * struct srpt_node_acl - Per-initiator ACL data (managed via configfs). - * @nacl: Target core node ACL information. - */ -struct srpt_node_acl { - struct se_node_acl nacl; -}; - #endif /* IB_SRPT_H */ diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.c b/drivers/net/ethernet/mellanox/mlx4/fw.c index d66c690a8597..e97094598b2d 100644 --- a/drivers/net/ethernet/mellanox/mlx4/fw.c +++ b/drivers/net/ethernet/mellanox/mlx4/fw.c @@ -157,7 +157,8 @@ static void dump_dev_cap_flags2(struct mlx4_dev *dev, u64 flags) [29] = "802.1ad offload support", [31] = "Modifying loopback source checks using UPDATE_QP support", [32] = "Loopback source checks support", - [33] = "RoCEv2 support" + [33] = "RoCEv2 support", + [34] = "DMFS Sniffer support (UC & MC)" }; int i; @@ -810,6 +811,8 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap) if (field & 0x80) dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_FS_EN; dev_cap->fs_log_max_ucast_qp_range_size = field & 0x1f; + if (field & 0x20) + dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_DMFS_UC_MC_SNIFFER; MLX4_GET(field, outbox, QUERY_DEV_CAP_PORT_BEACON_OFFSET); if (field & 0x80) dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_PORT_BEACON; diff --git a/drivers/net/ethernet/mellanox/mlx4/mcg.c b/drivers/net/ethernet/mellanox/mlx4/mcg.c index 1d4e2e054647..42d8de892bfe 100644 --- a/drivers/net/ethernet/mellanox/mlx4/mcg.c +++ b/drivers/net/ethernet/mellanox/mlx4/mcg.c @@ -752,8 +752,10 @@ static const u8 __promisc_mode[] = { [MLX4_FS_REGULAR] = 0x0, [MLX4_FS_ALL_DEFAULT] = 0x1, [MLX4_FS_MC_DEFAULT] = 0x3, - [MLX4_FS_UC_SNIFFER] = 0x4, - [MLX4_FS_MC_SNIFFER] = 0x5, + [MLX4_FS_MIRROR_RX_PORT] = 0x4, + [MLX4_FS_MIRROR_SX_PORT] = 0x5, + [MLX4_FS_UC_SNIFFER] = 0x6, + [MLX4_FS_MC_SNIFFER] = 0x7, }; int mlx4_map_sw_to_hw_steering_mode(struct mlx4_dev *dev, diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c index 6f68dba8d7ed..bf3446794bd5 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c @@ -77,6 +77,9 @@ #define KERNEL_NUM_PRIOS 1 #define KENREL_MIN_LEVEL 2 +#define ANCHOR_MAX_FT 1 +#define ANCHOR_NUM_PRIOS 1 +#define ANCHOR_MIN_LEVEL (BY_PASS_MIN_LEVEL + 1) struct node_caps { size_t arr_sz; long *caps; @@ -92,7 +95,7 @@ static struct init_tree_node { int max_ft; } root_fs = { .type = FS_TYPE_NAMESPACE, - .ar_size = 3, + .ar_size = 4, .children = (struct init_tree_node[]) { ADD_PRIO(0, BY_PASS_MIN_LEVEL, 0, FS_REQUIRED_CAPS(FS_CAP(flow_table_properties_nic_receive.flow_modify_en), @@ -108,6 +111,8 @@ static struct init_tree_node { FS_CAP(flow_table_properties_nic_receive.identified_miss_table_mode), FS_CAP(flow_table_properties_nic_receive.flow_table_modify)), ADD_NS(ADD_MULTIPLE_PRIO(LEFTOVERS_NUM_PRIOS, LEFTOVERS_MAX_FT))), + ADD_PRIO(0, ANCHOR_MIN_LEVEL, 0, {}, + ADD_NS(ADD_MULTIPLE_PRIO(ANCHOR_NUM_PRIOS, ANCHOR_MAX_FT))), } }; @@ -196,8 +201,10 @@ static void tree_put_node(struct fs_node *node) static int tree_remove_node(struct fs_node *node) { - if (atomic_read(&node->refcount) > 1) - return -EPERM; + if (atomic_read(&node->refcount) > 1) { + atomic_dec(&node->refcount); + return -EEXIST; + } tree_put_node(node); return 0; } @@ -360,6 +367,11 @@ static void del_rule(struct fs_node *node) memcpy(match_value, fte->val, sizeof(fte->val)); fs_get_obj(ft, fg->node.parent); list_del(&rule->node.list); + if (rule->sw_action == MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO) { + mutex_lock(&rule->dest_attr.ft->lock); + list_del(&rule->next_ft); + mutex_unlock(&rule->dest_attr.ft->lock); + } fte->dests_size--; if (fte->dests_size) { err = mlx5_cmd_update_fte(dev, ft, @@ -465,6 +477,8 @@ static struct mlx5_flow_table *alloc_flow_table(int level, int max_fte, ft->node.type = FS_TYPE_FLOW_TABLE; ft->type = table_type; ft->max_fte = max_fte; + INIT_LIST_HEAD(&ft->fwd_rules); + mutex_init(&ft->lock); return ft; } @@ -601,9 +615,63 @@ static int update_root_ft_create(struct mlx5_flow_table *ft, struct fs_prio return err; } +static int mlx5_modify_rule_destination(struct mlx5_flow_rule *rule, + struct mlx5_flow_destination *dest) +{ + struct mlx5_flow_table *ft; + struct mlx5_flow_group *fg; + struct fs_fte *fte; + int err = 0; + + fs_get_obj(fte, rule->node.parent); + if (!(fte->action & MLX5_FLOW_CONTEXT_ACTION_FWD_DEST)) + return -EINVAL; + lock_ref_node(&fte->node); + fs_get_obj(fg, fte->node.parent); + fs_get_obj(ft, fg->node.parent); + + memcpy(&rule->dest_attr, dest, sizeof(*dest)); + err = mlx5_cmd_update_fte(get_dev(&ft->node), + ft, fg->id, fte); + unlock_ref_node(&fte->node); + + return err; +} + +/* Modify/set FWD rules that point on old_next_ft to point on new_next_ft */ +static int connect_fwd_rules(struct mlx5_core_dev *dev, + struct mlx5_flow_table *new_next_ft, + struct mlx5_flow_table *old_next_ft) +{ + struct mlx5_flow_destination dest; + struct mlx5_flow_rule *iter; + int err = 0; + + /* new_next_ft and old_next_ft could be NULL only + * when we create/destroy the anchor flow table. + */ + if (!new_next_ft || !old_next_ft) + return 0; + + dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE; + dest.ft = new_next_ft; + + mutex_lock(&old_next_ft->lock); + list_splice_init(&old_next_ft->fwd_rules, &new_next_ft->fwd_rules); + mutex_unlock(&old_next_ft->lock); + list_for_each_entry(iter, &new_next_ft->fwd_rules, next_ft) { + err = mlx5_modify_rule_destination(iter, &dest); + if (err) + pr_err("mlx5_core: failed to modify rule to point on flow table %d\n", + new_next_ft->id); + } + return 0; +} + static int connect_flow_table(struct mlx5_core_dev *dev, struct mlx5_flow_table *ft, struct fs_prio *prio) { + struct mlx5_flow_table *next_ft; int err = 0; /* Connect_prev_fts and update_root_ft_create are mutually exclusive */ @@ -612,6 +680,11 @@ static int connect_flow_table(struct mlx5_core_dev *dev, struct mlx5_flow_table err = connect_prev_fts(dev, ft, prio); if (err) return err; + + next_ft = find_next_chained_ft(prio); + err = connect_fwd_rules(dev, ft, next_ft); + if (err) + return err; } if (MLX5_CAP_FLOWTABLE(dev, @@ -762,6 +835,7 @@ static struct mlx5_flow_rule *alloc_rule(struct mlx5_flow_destination *dest) if (!rule) return NULL; + INIT_LIST_HEAD(&rule->next_ft); rule->node.type = FS_TYPE_FLOW_DEST; memcpy(&rule->dest_attr, dest, sizeof(*dest)); @@ -782,9 +856,14 @@ static struct mlx5_flow_rule *add_rule_fte(struct fs_fte *fte, return ERR_PTR(-ENOMEM); fs_get_obj(ft, fg->node.parent); - /* Add dest to dests list- added as first element after the head */ + /* Add dest to dests list- we need flow tables to be in the + * end of the list for forward to next prio rules. + */ tree_init_node(&rule->node, 1, del_rule); - list_add_tail(&rule->node.list, &fte->node.children); + if (dest && dest->type != MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE) + list_add(&rule->node.list, &fte->node.children); + else + list_add_tail(&rule->node.list, &fte->node.children); fte->dests_size++; if (fte->dests_size == 1) err = mlx5_cmd_create_fte(get_dev(&ft->node), @@ -903,6 +982,25 @@ out: return fg; } +static struct mlx5_flow_rule *find_flow_rule(struct fs_fte *fte, + struct mlx5_flow_destination *dest) +{ + struct mlx5_flow_rule *rule; + + list_for_each_entry(rule, &fte->node.children, node.list) { + if (rule->dest_attr.type == dest->type) { + if ((dest->type == MLX5_FLOW_DESTINATION_TYPE_VPORT && + dest->vport_num == rule->dest_attr.vport_num) || + (dest->type == MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE && + dest->ft == rule->dest_attr.ft) || + (dest->type == MLX5_FLOW_DESTINATION_TYPE_TIR && + dest->tir_num == rule->dest_attr.tir_num)) + return rule; + } + } + return NULL; +} + static struct mlx5_flow_rule *add_rule_fg(struct mlx5_flow_group *fg, u32 *match_value, u8 action, @@ -919,6 +1017,13 @@ static struct mlx5_flow_rule *add_rule_fg(struct mlx5_flow_group *fg, nested_lock_ref_node(&fte->node, FS_MUTEX_CHILD); if (compare_match_value(&fg->mask, match_value, &fte->val) && action == fte->action && flow_tag == fte->flow_tag) { + rule = find_flow_rule(fte, dest); + if (rule) { + atomic_inc(&rule->node.refcount); + unlock_ref_node(&fte->node); + unlock_ref_node(&fg->node); + return rule; + } rule = add_rule_fte(fte, fg, dest); unlock_ref_node(&fte->node); if (IS_ERR(rule)) @@ -984,14 +1089,14 @@ static struct mlx5_flow_rule *add_rule_to_auto_fg(struct mlx5_flow_table *ft, return rule; } -struct mlx5_flow_rule * -mlx5_add_flow_rule(struct mlx5_flow_table *ft, - u8 match_criteria_enable, - u32 *match_criteria, - u32 *match_value, - u32 action, - u32 flow_tag, - struct mlx5_flow_destination *dest) +static struct mlx5_flow_rule * +_mlx5_add_flow_rule(struct mlx5_flow_table *ft, + u8 match_criteria_enable, + u32 *match_criteria, + u32 *match_value, + u32 action, + u32 flow_tag, + struct mlx5_flow_destination *dest) { struct mlx5_flow_group *g; struct mlx5_flow_rule *rule; @@ -1014,6 +1119,63 @@ unlock: unlock_ref_node(&ft->node); return rule; } + +static bool fwd_next_prio_supported(struct mlx5_flow_table *ft) +{ + return ((ft->type == FS_FT_NIC_RX) && + (MLX5_CAP_FLOWTABLE(get_dev(&ft->node), nic_rx_multi_path_tirs))); +} + +struct mlx5_flow_rule * +mlx5_add_flow_rule(struct mlx5_flow_table *ft, + u8 match_criteria_enable, + u32 *match_criteria, + u32 *match_value, + u32 action, + u32 flow_tag, + struct mlx5_flow_destination *dest) +{ + struct mlx5_flow_root_namespace *root = find_root(&ft->node); + struct mlx5_flow_destination gen_dest; + struct mlx5_flow_table *next_ft = NULL; + struct mlx5_flow_rule *rule = NULL; + u32 sw_action = action; + struct fs_prio *prio; + + fs_get_obj(prio, ft->node.parent); + if (action == MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO) { + if (!fwd_next_prio_supported(ft)) + return ERR_PTR(-EOPNOTSUPP); + if (dest) + return ERR_PTR(-EINVAL); + mutex_lock(&root->chain_lock); + next_ft = find_next_chained_ft(prio); + if (next_ft) { + gen_dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE; + gen_dest.ft = next_ft; + dest = &gen_dest; + action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST; + } else { + mutex_unlock(&root->chain_lock); + return ERR_PTR(-EOPNOTSUPP); + } + } + + rule = _mlx5_add_flow_rule(ft, match_criteria_enable, match_criteria, + match_value, action, flow_tag, dest); + + if (sw_action == MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO) { + if (!IS_ERR_OR_NULL(rule) && + (list_empty(&rule->next_ft))) { + mutex_lock(&next_ft->lock); + list_add(&rule->next_ft, &next_ft->fwd_rules); + mutex_unlock(&next_ft->lock); + rule->sw_action = MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO; + } + mutex_unlock(&root->chain_lock); + } + return rule; +} EXPORT_SYMBOL(mlx5_add_flow_rule); void mlx5_del_flow_rule(struct mlx5_flow_rule *rule) @@ -1077,6 +1239,10 @@ static int disconnect_flow_table(struct mlx5_flow_table *ft) return 0; next_ft = find_next_chained_ft(prio); + err = connect_fwd_rules(dev, next_ft, ft); + if (err) + return err; + err = connect_prev_fts(dev, next_ft, prio); if (err) mlx5_core_warn(dev, "Failed to disconnect flow table %d\n", @@ -1126,6 +1292,7 @@ struct mlx5_flow_namespace *mlx5_get_flow_namespace(struct mlx5_core_dev *dev, case MLX5_FLOW_NAMESPACE_BYPASS: case MLX5_FLOW_NAMESPACE_KERNEL: case MLX5_FLOW_NAMESPACE_LEFTOVERS: + case MLX5_FLOW_NAMESPACE_ANCHOR: prio = type; break; case MLX5_FLOW_NAMESPACE_FDB: @@ -1351,6 +1518,25 @@ static void set_prio_attrs(struct mlx5_flow_root_namespace *root_ns) } } +#define ANCHOR_PRIO 0 +#define ANCHOR_SIZE 1 +static int create_anchor_flow_table(struct mlx5_core_dev + *dev) +{ + struct mlx5_flow_namespace *ns = NULL; + struct mlx5_flow_table *ft; + + ns = mlx5_get_flow_namespace(dev, MLX5_FLOW_NAMESPACE_ANCHOR); + if (!ns) + return -EINVAL; + ft = mlx5_create_flow_table(ns, ANCHOR_PRIO, ANCHOR_SIZE); + if (IS_ERR(ft)) { + mlx5_core_err(dev, "Failed to create last anchor flow table"); + return PTR_ERR(ft); + } + return 0; +} + static int init_root_ns(struct mlx5_core_dev *dev) { @@ -1363,6 +1549,9 @@ static int init_root_ns(struct mlx5_core_dev *dev) set_prio_attrs(dev->priv.root_ns); + if (create_anchor_flow_table(dev)) + goto cleanup; + return 0; cleanup: @@ -1392,6 +1581,15 @@ static void cleanup_single_prio_root_ns(struct mlx5_core_dev *dev, root_ns = NULL; } +static void destroy_flow_tables(struct fs_prio *prio) +{ + struct mlx5_flow_table *iter; + struct mlx5_flow_table *tmp; + + fs_for_each_ft_safe(iter, tmp, prio) + mlx5_destroy_flow_table(iter); +} + static void cleanup_root_ns(struct mlx5_core_dev *dev) { struct mlx5_flow_root_namespace *root_ns = dev->priv.root_ns; @@ -1420,6 +1618,7 @@ static void cleanup_root_ns(struct mlx5_core_dev *dev) list); fs_get_obj(obj_iter_prio2, iter_prio2); + destroy_flow_tables(obj_iter_prio2); if (tree_remove_node(iter_prio2)) { mlx5_core_warn(dev, "Priority %d wasn't destroyed, refcount > 1\n", diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h index 00245fd7e4bc..f37a6248a27b 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h @@ -68,6 +68,11 @@ struct fs_node { struct mlx5_flow_rule { struct fs_node node; struct mlx5_flow_destination dest_attr; + /* next_ft should be accessed under chain_lock and only of + * destination type is FWD_NEXT_fT. + */ + struct list_head next_ft; + u32 sw_action; }; /* Type of children is mlx5_flow_group */ @@ -82,6 +87,10 @@ struct mlx5_flow_table { unsigned int required_groups; unsigned int num_groups; } autogroup; + /* Protect fwd_rules */ + struct mutex lock; + /* FWD rules that point on this flow table */ + struct list_head fwd_rules; }; /* Type of children is mlx5_flow_rule */ @@ -142,6 +151,9 @@ void mlx5_cleanup_fs(struct mlx5_core_dev *dev); #define fs_list_for_each_entry(pos, root) \ list_for_each_entry(pos, root, node.list) +#define fs_list_for_each_entry_safe(pos, tmp, root) \ + list_for_each_entry_safe(pos, tmp, root, node.list) + #define fs_for_each_ns_or_ft_reverse(pos, prio) \ list_for_each_entry_reverse(pos, &(prio)->node.children, list) @@ -157,6 +169,9 @@ void mlx5_cleanup_fs(struct mlx5_core_dev *dev); #define fs_for_each_ft(pos, prio) \ fs_list_for_each_entry(pos, &(prio)->node.children) +#define fs_for_each_ft_safe(pos, tmp, prio) \ + fs_list_for_each_entry_safe(pos, tmp, &(prio)->node.children) + #define fs_for_each_fg(pos, ft) \ fs_list_for_each_entry(pos, &(ft)->node.children) diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h index a0e8cc8dcc67..8541a913f6a3 100644 --- a/include/linux/mlx4/device.h +++ b/include/linux/mlx4/device.h @@ -219,6 +219,7 @@ enum { MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB = 1ULL << 31, MLX4_DEV_CAP_FLAG2_LB_SRC_CHK = 1ULL << 32, MLX4_DEV_CAP_FLAG2_ROCE_V1_V2 = 1ULL << 33, + MLX4_DEV_CAP_FLAG2_DMFS_UC_MC_SNIFFER = 1ULL << 34, }; enum { @@ -1160,6 +1161,8 @@ enum mlx4_net_trans_promisc_mode { MLX4_FS_REGULAR = 1, MLX4_FS_ALL_DEFAULT, MLX4_FS_MC_DEFAULT, + MLX4_FS_MIRROR_RX_PORT, + MLX4_FS_MIRROR_SX_PORT, MLX4_FS_UC_SNIFFER, MLX4_FS_MC_SNIFFER, MLX4_FS_MODE_NUM, /* should be last */ diff --git a/include/linux/mlx5/device.h b/include/linux/mlx5/device.h index 12079fd20413..9566b3b3b2c5 100644 --- a/include/linux/mlx5/device.h +++ b/include/linux/mlx5/device.h @@ -1318,6 +1318,11 @@ static inline u16 mlx5_to_sw_pkey_sz(int pkey_sz) return MLX5_MIN_PKEY_TABLE_SIZE << pkey_sz; } -#define MLX5_BY_PASS_NUM_PRIOS 9 +#define MLX5_BY_PASS_NUM_REGULAR_PRIOS 8 +#define MLX5_BY_PASS_NUM_DONT_TRAP_PRIOS 8 +#define MLX5_BY_PASS_NUM_MULTICAST_PRIOS 1 +#define MLX5_BY_PASS_NUM_PRIOS (MLX5_BY_PASS_NUM_REGULAR_PRIOS +\ + MLX5_BY_PASS_NUM_DONT_TRAP_PRIOS +\ + MLX5_BY_PASS_NUM_MULTICAST_PRIOS) #endif /* MLX5_DEVICE_H */ diff --git a/include/linux/mlx5/fs.h b/include/linux/mlx5/fs.h index 8230caa3fb6e..8dec5508d93d 100644 --- a/include/linux/mlx5/fs.h +++ b/include/linux/mlx5/fs.h @@ -38,6 +38,10 @@ #define MLX5_FS_DEFAULT_FLOW_TAG 0x0 +enum { + MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO = 1 << 16, +}; + #define LEFTOVERS_RULE_NUM 2 static inline void build_leftovers_ft_param(int *priority, int *n_ent, @@ -52,6 +56,7 @@ enum mlx5_flow_namespace_type { MLX5_FLOW_NAMESPACE_BYPASS, MLX5_FLOW_NAMESPACE_KERNEL, MLX5_FLOW_NAMESPACE_LEFTOVERS, + MLX5_FLOW_NAMESPACE_ANCHOR, MLX5_FLOW_NAMESPACE_FDB, }; diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 3044cfa683f1..a3cacab22849 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -458,7 +458,8 @@ struct mlx5_ifc_ads_bits { }; struct mlx5_ifc_flow_table_nic_cap_bits { - u8 reserved_at_0[0x200]; + u8 nic_rx_multi_path_tirs[0x1]; + u8 reserved_at_1[0x1ff]; struct mlx5_ifc_flow_table_prop_layout_bits flow_table_properties_nic_receive; diff --git a/include/rdma/ib_mad.h b/include/rdma/ib_mad.h index 0ff049bd9ad4..37dd534cbeab 100644 --- a/include/rdma/ib_mad.h +++ b/include/rdma/ib_mad.h @@ -424,11 +424,11 @@ typedef void (*ib_mad_send_handler)(struct ib_mad_agent *mad_agent, /** * ib_mad_snoop_handler - Callback handler for snooping sent MADs. * @mad_agent: MAD agent that snooped the MAD. - * @send_wr: Work request information on the sent MAD. + * @send_buf: send MAD data buffer. * @mad_send_wc: Work completion information on the sent MAD. Valid * only for snooping that occurs on a send completion. * - * Clients snooping MADs should not modify data referenced by the @send_wr + * Clients snooping MADs should not modify data referenced by the @send_buf * or @mad_send_wc. */ typedef void (*ib_mad_snoop_handler)(struct ib_mad_agent *mad_agent, diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index bcd5b242e6b1..3a03c1d18afa 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -1493,6 +1493,11 @@ enum ib_flow_domain { IB_FLOW_DOMAIN_NUM /* Must be last */ }; +enum ib_flow_flags { + IB_FLOW_ATTR_FLAGS_DONT_TRAP = 1UL << 1, /* Continue match, no steal */ + IB_FLOW_ATTR_FLAGS_RESERVED = 1UL << 2 /* Must be last */ +}; + struct ib_flow_eth_filter { u8 dst_mac[6]; u8 src_mac[6]; @@ -1853,6 +1858,8 @@ struct ib_device { int (*check_mr_status)(struct ib_mr *mr, u32 check_mask, struct ib_mr_status *mr_status); void (*disassociate_ucontext)(struct ib_ucontext *ibcontext); + void (*drain_rq)(struct ib_qp *qp); + void (*drain_sq)(struct ib_qp *qp); struct ib_dma_mapping_ops *dma_ops; @@ -3101,4 +3108,7 @@ int ib_sg_to_pages(struct ib_mr *mr, int sg_nents, int (*set_page)(struct ib_mr *, u64)); +void ib_drain_rq(struct ib_qp *qp); +void ib_drain_sq(struct ib_qp *qp); +void ib_drain_qp(struct ib_qp *qp); #endif /* IB_VERBS_H */ diff --git a/net/9p/trans_rdma.c b/net/9p/trans_rdma.c index 52b4a2f993f2..1852e383afd6 100644 --- a/net/9p/trans_rdma.c +++ b/net/9p/trans_rdma.c @@ -109,14 +109,13 @@ struct p9_trans_rdma { /** * p9_rdma_context - Keeps track of in-process WR * - * @wc_op: The original WR op for when the CQE completes in error. * @busa: Bus address to unmap when the WR completes * @req: Keeps track of requests (send) * @rc: Keepts track of replies (receive) */ struct p9_rdma_req; struct p9_rdma_context { - enum ib_wc_opcode wc_op; + struct ib_cqe cqe; dma_addr_t busa; union { struct p9_req_t *req; @@ -284,9 +283,12 @@ p9_cm_event_handler(struct rdma_cm_id *id, struct rdma_cm_event *event) } static void -handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma, - struct p9_rdma_context *c, enum ib_wc_status status, u32 byte_len) +recv_done(struct ib_cq *cq, struct ib_wc *wc) { + struct p9_client *client = cq->cq_context; + struct p9_trans_rdma *rdma = client->trans; + struct p9_rdma_context *c = + container_of(wc->wr_cqe, struct p9_rdma_context, cqe); struct p9_req_t *req; int err = 0; int16_t tag; @@ -295,7 +297,7 @@ handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma, ib_dma_unmap_single(rdma->cm_id->device, c->busa, client->msize, DMA_FROM_DEVICE); - if (status != IB_WC_SUCCESS) + if (wc->status != IB_WC_SUCCESS) goto err_out; err = p9_parse_header(c->rc, NULL, NULL, &tag, 1); @@ -316,21 +318,32 @@ handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma, req->rc = c->rc; p9_client_cb(client, req, REQ_STATUS_RCVD); + out: + up(&rdma->rq_sem); + kfree(c); return; err_out: - p9_debug(P9_DEBUG_ERROR, "req %p err %d status %d\n", req, err, status); + p9_debug(P9_DEBUG_ERROR, "req %p err %d status %d\n", + req, err, wc->status); rdma->state = P9_RDMA_FLUSHING; client->status = Disconnected; + goto out; } static void -handle_send(struct p9_client *client, struct p9_trans_rdma *rdma, - struct p9_rdma_context *c, enum ib_wc_status status, u32 byte_len) +send_done(struct ib_cq *cq, struct ib_wc *wc) { + struct p9_client *client = cq->cq_context; + struct p9_trans_rdma *rdma = client->trans; + struct p9_rdma_context *c = + container_of(wc->wr_cqe, struct p9_rdma_context, cqe); + ib_dma_unmap_single(rdma->cm_id->device, c->busa, c->req->tc->size, DMA_TO_DEVICE); + up(&rdma->sq_sem); + kfree(c); } static void qp_event_handler(struct ib_event *event, void *context) @@ -339,42 +352,6 @@ static void qp_event_handler(struct ib_event *event, void *context) event->event, context); } -static void cq_comp_handler(struct ib_cq *cq, void *cq_context) -{ - struct p9_client *client = cq_context; - struct p9_trans_rdma *rdma = client->trans; - int ret; - struct ib_wc wc; - - ib_req_notify_cq(rdma->cq, IB_CQ_NEXT_COMP); - while ((ret = ib_poll_cq(cq, 1, &wc)) > 0) { - struct p9_rdma_context *c = (void *) (unsigned long) wc.wr_id; - - switch (c->wc_op) { - case IB_WC_RECV: - handle_recv(client, rdma, c, wc.status, wc.byte_len); - up(&rdma->rq_sem); - break; - - case IB_WC_SEND: - handle_send(client, rdma, c, wc.status, wc.byte_len); - up(&rdma->sq_sem); - break; - - default: - pr_err("unexpected completion type, c->wc_op=%d, wc.opcode=%d, status=%d\n", - c->wc_op, wc.opcode, wc.status); - break; - } - kfree(c); - } -} - -static void cq_event_handler(struct ib_event *e, void *v) -{ - p9_debug(P9_DEBUG_ERROR, "CQ event %d context %p\n", e->event, v); -} - static void rdma_destroy_trans(struct p9_trans_rdma *rdma) { if (!rdma) @@ -387,7 +364,7 @@ static void rdma_destroy_trans(struct p9_trans_rdma *rdma) ib_dealloc_pd(rdma->pd); if (rdma->cq && !IS_ERR(rdma->cq)) - ib_destroy_cq(rdma->cq); + ib_free_cq(rdma->cq); if (rdma->cm_id && !IS_ERR(rdma->cm_id)) rdma_destroy_id(rdma->cm_id); @@ -408,13 +385,14 @@ post_recv(struct p9_client *client, struct p9_rdma_context *c) if (ib_dma_mapping_error(rdma->cm_id->device, c->busa)) goto error; + c->cqe.done = recv_done; + sge.addr = c->busa; sge.length = client->msize; sge.lkey = rdma->pd->local_dma_lkey; wr.next = NULL; - c->wc_op = IB_WC_RECV; - wr.wr_id = (unsigned long) c; + wr.wr_cqe = &c->cqe; wr.sg_list = &sge; wr.num_sge = 1; return ib_post_recv(rdma->qp, &wr, &bad_wr); @@ -499,13 +477,14 @@ dont_need_post_recv: goto send_error; } + c->cqe.done = send_done; + sge.addr = c->busa; sge.length = c->req->tc->size; sge.lkey = rdma->pd->local_dma_lkey; wr.next = NULL; - c->wc_op = IB_WC_SEND; - wr.wr_id = (unsigned long) c; + wr.wr_cqe = &c->cqe; wr.opcode = IB_WR_SEND; wr.send_flags = IB_SEND_SIGNALED; wr.sg_list = &sge; @@ -642,7 +621,6 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args) struct p9_trans_rdma *rdma; struct rdma_conn_param conn_param; struct ib_qp_init_attr qp_attr; - struct ib_cq_init_attr cq_attr = {}; /* Parse the transport specific mount options */ err = parse_opts(args, &opts); @@ -695,13 +673,11 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args) goto error; /* Create the Completion Queue */ - cq_attr.cqe = opts.sq_depth + opts.rq_depth + 1; - rdma->cq = ib_create_cq(rdma->cm_id->device, cq_comp_handler, - cq_event_handler, client, - &cq_attr); + rdma->cq = ib_alloc_cq(rdma->cm_id->device, client, + opts.sq_depth + opts.rq_depth + 1, + 0, IB_POLL_SOFTIRQ); if (IS_ERR(rdma->cq)) goto error; - ib_req_notify_cq(rdma->cq, IB_CQ_NEXT_COMP); /* Create the Protection Domain */ rdma->pd = ib_alloc_pd(rdma->cm_id->device); |