X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Finfiniband%2Fulp%2Fsrp%2Fib_srp.c;h=39bf057fbc4306c46112d3ac5e9433578a21f4ba;hb=f4fd0b224d60044d2da5ca02f8f2b5150c1d8731;hp=9cbdffa08dc2bca6da821638efc0078b24792374;hpb=093beac189e4295d968f0d38787b46f76cb0eaaa;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 9cbdffa..39bf057 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -62,6 +62,13 @@ MODULE_DESCRIPTION("InfiniBand SCSI RDMA Protocol initiator " "v" DRV_VERSION " (" DRV_RELDATE ")"); MODULE_LICENSE("Dual BSD/GPL"); +static int srp_sg_tablesize = SRP_DEF_SG_TABLESIZE; +static int srp_max_iu_len; + +module_param(srp_sg_tablesize, int, 0444); +MODULE_PARM_DESC(srp_sg_tablesize, + "Max number of gather/scatter entries per I/O (default is 12)"); + static int topspin_workarounds = 1; module_param(topspin_workarounds, int, 0444); @@ -70,6 +77,14 @@ MODULE_PARM_DESC(topspin_workarounds, static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad }; +static int mellanox_workarounds = 1; + +module_param(mellanox_workarounds, int, 0444); +MODULE_PARM_DESC(mellanox_workarounds, + "Enable workarounds for Mellanox SRP target bugs if != 0"); + +static const u8 mellanox_oui[3] = { 0x00, 0x02, 0xc9 }; + static void srp_add_one(struct ib_device *device); static void srp_remove_one(struct ib_device *device); static void srp_completion(struct ib_cq *cq, void *target_ptr); @@ -81,6 +96,8 @@ static struct ib_client srp_client = { .remove = srp_remove_one }; +static struct ib_sa_client srp_sa_client; + static inline struct srp_target_port *host_to_target(struct Scsi_Host *host) { return (struct srp_target_port *) host->hostdata; @@ -105,8 +122,8 @@ static struct srp_iu *srp_alloc_iu(struct srp_host *host, size_t size, if (!iu->buf) goto out_free_iu; - iu->dma = dma_map_single(host->dev->dma_device, iu->buf, size, direction); - if (dma_mapping_error(iu->dma)) + iu->dma = ib_dma_map_single(host->dev->dev, iu->buf, size, direction); + if (ib_dma_mapping_error(host->dev->dev, iu->dma)) goto out_free_buf; iu->size = size; @@ -127,7 +144,7 @@ static void srp_free_iu(struct srp_host *host, struct srp_iu *iu) if (!iu) return; - dma_unmap_single(host->dev->dma_device, iu->dma, iu->size, iu->direction); + ib_dma_unmap_single(host->dev->dev, iu->dma, iu->size, iu->direction); kfree(iu->buf); kfree(iu); } @@ -147,7 +164,7 @@ static int srp_init_qp(struct srp_target_port *target, if (!attr) return -ENOMEM; - ret = ib_find_cached_pkey(target->srp_host->dev, + ret = ib_find_cached_pkey(target->srp_host->dev->dev, target->srp_host->port, be16_to_cpu(target->path.pkey), &attr->pkey_index); @@ -179,8 +196,8 @@ static int srp_create_target_ib(struct srp_target_port *target) if (!init_attr) return -ENOMEM; - target->cq = ib_create_cq(target->srp_host->dev, srp_completion, - NULL, target, SRP_CQ_SIZE); + target->cq = ib_create_cq(target->srp_host->dev->dev, srp_completion, + NULL, target, SRP_CQ_SIZE, 0); if (IS_ERR(target->cq)) { ret = PTR_ERR(target->cq); goto out; @@ -198,7 +215,7 @@ static int srp_create_target_ib(struct srp_target_port *target) init_attr->send_cq = target->cq; init_attr->recv_cq = target->cq; - target->qp = ib_create_qp(target->srp_host->pd, init_attr); + target->qp = ib_create_qp(target->srp_host->dev->pd, init_attr); if (IS_ERR(target->qp)) { ret = PTR_ERR(target->qp); ib_destroy_cq(target->cq); @@ -250,7 +267,8 @@ static int srp_lookup_path(struct srp_target_port *target) init_completion(&target->done); - target->path_query_id = ib_sa_path_rec_get(target->srp_host->dev, + target->path_query_id = ib_sa_path_rec_get(&srp_sa_client, + target->srp_host->dev->dev, target->srp_host->port, &target->path, IB_SA_PATH_REC_DGID | @@ -309,24 +327,47 @@ static int srp_send_req(struct srp_target_port *target) req->priv.opcode = SRP_LOGIN_REQ; req->priv.tag = 0; - req->priv.req_it_iu_len = cpu_to_be32(SRP_MAX_IU_LEN); + req->priv.req_it_iu_len = cpu_to_be32(srp_max_iu_len); req->priv.req_buf_fmt = cpu_to_be16(SRP_BUF_FORMAT_DIRECT | SRP_BUF_FORMAT_INDIRECT); - memcpy(req->priv.initiator_port_id, target->srp_host->initiator_port_id, 16); + /* + * In the published SRP specification (draft rev. 16a), the + * port identifier format is 8 bytes of ID extension followed + * by 8 bytes of GUID. Older drafts put the two halves in the + * opposite order, so that the GUID comes first. + * + * Targets conforming to these obsolete drafts can be + * recognized by the I/O Class they report. + */ + if (target->io_class == SRP_REV10_IB_IO_CLASS) { + memcpy(req->priv.initiator_port_id, + &target->path.sgid.global.interface_id, 8); + memcpy(req->priv.initiator_port_id + 8, + &target->initiator_ext, 8); + memcpy(req->priv.target_port_id, &target->ioc_guid, 8); + memcpy(req->priv.target_port_id + 8, &target->id_ext, 8); + } else { + memcpy(req->priv.initiator_port_id, + &target->initiator_ext, 8); + memcpy(req->priv.initiator_port_id + 8, + &target->path.sgid.global.interface_id, 8); + memcpy(req->priv.target_port_id, &target->id_ext, 8); + memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8); + } + /* * Topspin/Cisco SRP targets will reject our login unless we - * zero out the first 8 bytes of our initiator port ID. The - * second 8 bytes must be our local node GUID, but we always - * use that anyway. + * zero out the first 8 bytes of our initiator port ID and set + * the second 8 bytes to the local node GUID. */ if (topspin_workarounds && !memcmp(&target->ioc_guid, topspin_oui, 3)) { printk(KERN_DEBUG PFX "Topspin/Cisco initiator port ID workaround " "activated for target GUID %016llx\n", (unsigned long long) be64_to_cpu(target->ioc_guid)); memset(req->priv.initiator_port_id, 0, 8); + memcpy(req->priv.initiator_port_id + 8, + &target->srp_host->dev->dev->node_guid, 8); } - memcpy(req->priv.target_port_id, &target->id_ext, 8); - memcpy(req->priv.target_port_id + 8, &target->ioc_guid, 8); status = ib_send_cm_req(target->cm_id, &req->param); @@ -347,9 +388,10 @@ static void srp_disconnect_target(struct srp_target_port *target) wait_for_completion(&target->done); } -static void srp_remove_work(void *target_ptr) +static void srp_remove_work(struct work_struct *work) { - struct srp_target_port *target = target_ptr; + struct srp_target_port *target = + container_of(work, struct srp_target_port, work); spin_lock_irq(target->scsi_host->host_lock); if (target->state != SRP_TARGET_DEAD) { @@ -359,9 +401,9 @@ static void srp_remove_work(void *target_ptr) target->state = SRP_TARGET_REMOVED; spin_unlock_irq(target->scsi_host->host_lock); - mutex_lock(&target->srp_host->target_mutex); + spin_lock(&target->srp_host->target_lock); list_del(&target->list); - mutex_unlock(&target->srp_host->target_mutex); + spin_unlock(&target->srp_host->target_lock); scsi_remove_host(target->scsi_host); ib_destroy_cm_id(target->cm_id); @@ -421,6 +463,11 @@ static void srp_unmap_data(struct scsi_cmnd *scmnd, scmnd->sc_data_direction != DMA_FROM_DEVICE)) return; + if (req->fmr) { + ib_fmr_pool_unmap(req->fmr); + req->fmr = NULL; + } + /* * This handling of non-SG commands can be killed when the * SCSI midlayer no longer generates non-SG commands. @@ -433,18 +480,30 @@ static void srp_unmap_data(struct scsi_cmnd *scmnd, scat = &req->fake_sg; } - dma_unmap_sg(target->srp_host->dev->dma_device, scat, nents, - scmnd->sc_data_direction); + ib_dma_unmap_sg(target->srp_host->dev->dev, scat, nents, + scmnd->sc_data_direction); +} + +static void srp_remove_req(struct srp_target_port *target, struct srp_request *req) +{ + srp_unmap_data(req->scmnd, target, req); + list_move_tail(&req->list, &target->free_reqs); +} + +static void srp_reset_req(struct srp_target_port *target, struct srp_request *req) +{ + req->scmnd->result = DID_RESET << 16; + req->scmnd->scsi_done(req->scmnd); + srp_remove_req(target, req); } static int srp_reconnect_target(struct srp_target_port *target) { struct ib_cm_id *new_cm_id; struct ib_qp_attr qp_attr; - struct srp_request *req; + struct srp_request *req, *tmp; struct ib_wc wc; int ret; - int i; spin_lock_irq(target->scsi_host->host_lock); if (target->state != SRP_TARGET_LIVE) { @@ -459,7 +518,7 @@ static int srp_reconnect_target(struct srp_target_port *target) * Now get a new local CM ID so that we avoid confusing the * target in case things are really fouled up. */ - new_cm_id = ib_create_cm_id(target->srp_host->dev, + new_cm_id = ib_create_cm_id(target->srp_host->dev->dev, srp_cm_handler, target); if (IS_ERR(new_cm_id)) { ret = PTR_ERR(new_cm_id); @@ -480,20 +539,16 @@ static int srp_reconnect_target(struct srp_target_port *target) while (ib_poll_cq(target->cq, 1, &wc) > 0) ; /* nothing */ - list_for_each_entry(req, &target->req_queue, list) { - req->scmnd->result = DID_RESET << 16; - req->scmnd->scsi_done(req->scmnd); - srp_unmap_data(req->scmnd, target, req); - } + spin_lock_irq(target->scsi_host->host_lock); + list_for_each_entry_safe(req, tmp, &target->req_queue, list) + srp_reset_req(target, req); + spin_unlock_irq(target->scsi_host->host_lock); target->rx_head = 0; target->tx_head = 0; target->tx_tail = 0; - INIT_LIST_HEAD(&target->free_reqs); - INIT_LIST_HEAD(&target->req_queue); - for (i = 0; i < SRP_SQ_SIZE; ++i) - list_add_tail(&target->req_ring[i].list, &target->free_reqs); + target->qp_in_error = 0; ret = srp_connect_target(target); if (ret) goto err; @@ -520,7 +575,7 @@ err: spin_lock_irq(target->scsi_host->host_lock); if (target->state == SRP_TARGET_CONNECTING) { target->state = SRP_TARGET_DEAD; - INIT_WORK(&target->work, srp_remove_work, target); + INIT_WORK(&target->work, srp_remove_work); schedule_work(&target->work); } spin_unlock_irq(target->scsi_host->host_lock); @@ -528,14 +583,95 @@ err: return ret; } +static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat, + int sg_cnt, struct srp_request *req, + struct srp_direct_buf *buf) +{ + u64 io_addr = 0; + u64 *dma_pages; + u32 len; + int page_cnt; + int i, j; + int ret; + struct srp_device *dev = target->srp_host->dev; + struct ib_device *ibdev = dev->dev; + + if (!dev->fmr_pool) + return -ENODEV; + + if ((ib_sg_dma_address(ibdev, &scat[0]) & ~dev->fmr_page_mask) && + mellanox_workarounds && !memcmp(&target->ioc_guid, mellanox_oui, 3)) + return -EINVAL; + + len = page_cnt = 0; + for (i = 0; i < sg_cnt; ++i) { + unsigned int dma_len = ib_sg_dma_len(ibdev, &scat[i]); + + if (ib_sg_dma_address(ibdev, &scat[i]) & ~dev->fmr_page_mask) { + if (i > 0) + return -EINVAL; + else + ++page_cnt; + } + if ((ib_sg_dma_address(ibdev, &scat[i]) + dma_len) & + ~dev->fmr_page_mask) { + if (i < sg_cnt - 1) + return -EINVAL; + else + ++page_cnt; + } + + len += dma_len; + } + + page_cnt += len >> dev->fmr_page_shift; + if (page_cnt > SRP_FMR_SIZE) + return -ENOMEM; + + dma_pages = kmalloc(sizeof (u64) * page_cnt, GFP_ATOMIC); + if (!dma_pages) + return -ENOMEM; + + page_cnt = 0; + for (i = 0; i < sg_cnt; ++i) { + unsigned int dma_len = ib_sg_dma_len(ibdev, &scat[i]); + + for (j = 0; j < dma_len; j += dev->fmr_page_size) + dma_pages[page_cnt++] = + (ib_sg_dma_address(ibdev, &scat[i]) & + dev->fmr_page_mask) + j; + } + + req->fmr = ib_fmr_pool_map_phys(dev->fmr_pool, + dma_pages, page_cnt, io_addr); + if (IS_ERR(req->fmr)) { + ret = PTR_ERR(req->fmr); + req->fmr = NULL; + goto out; + } + + buf->va = cpu_to_be64(ib_sg_dma_address(ibdev, &scat[0]) & + ~dev->fmr_page_mask); + buf->key = cpu_to_be32(req->fmr->fmr->rkey); + buf->len = cpu_to_be32(len); + + ret = 0; + +out: + kfree(dma_pages); + + return ret; +} + static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target, struct srp_request *req) { struct scatterlist *scat; struct srp_cmd *cmd = req->cmd->buf; int len, nents, count; - int i; - u8 fmt; + u8 fmt = SRP_DATA_DESC_DIRECT; + struct srp_device *dev; + struct ib_device *ibdev; if (!scmnd->request_buffer || scmnd->sc_data_direction == DMA_NONE) return sizeof (struct srp_cmd); @@ -560,53 +696,66 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target, sg_init_one(scat, scmnd->request_buffer, scmnd->request_bufflen); } - count = dma_map_sg(target->srp_host->dev->dma_device, scat, nents, - scmnd->sc_data_direction); + dev = target->srp_host->dev; + ibdev = dev->dev; - if (count == 1) { - struct srp_direct_buf *buf = (void *) cmd->add_data; + count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction); - fmt = SRP_DATA_DESC_DIRECT; + fmt = SRP_DATA_DESC_DIRECT; + len = sizeof (struct srp_cmd) + sizeof (struct srp_direct_buf); - buf->va = cpu_to_be64(sg_dma_address(scat)); - buf->key = cpu_to_be32(target->srp_host->mr->rkey); - buf->len = cpu_to_be32(sg_dma_len(scat)); + if (count == 1) { + /* + * The midlayer only generated a single gather/scatter + * entry, or DMA mapping coalesced everything to a + * single entry. So a direct descriptor along with + * the DMA MR suffices. + */ + struct srp_direct_buf *buf = (void *) cmd->add_data; - len = sizeof (struct srp_cmd) + - sizeof (struct srp_direct_buf); - } else { + buf->va = cpu_to_be64(ib_sg_dma_address(ibdev, scat)); + buf->key = cpu_to_be32(dev->mr->rkey); + buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat)); + } else if (srp_map_fmr(target, scat, count, req, + (void *) cmd->add_data)) { + /* + * FMR mapping failed, and the scatterlist has more + * than one entry. Generate an indirect memory + * descriptor. + */ struct srp_indirect_buf *buf = (void *) cmd->add_data; u32 datalen = 0; + int i; fmt = SRP_DATA_DESC_INDIRECT; + len = sizeof (struct srp_cmd) + + sizeof (struct srp_indirect_buf) + + count * sizeof (struct srp_direct_buf); + + for (i = 0; i < count; ++i) { + unsigned int dma_len = ib_sg_dma_len(ibdev, &scat[i]); + + buf->desc_list[i].va = + cpu_to_be64(ib_sg_dma_address(ibdev, &scat[i])); + buf->desc_list[i].key = + cpu_to_be32(dev->mr->rkey); + buf->desc_list[i].len = cpu_to_be32(dma_len); + datalen += dma_len; + } if (scmnd->sc_data_direction == DMA_TO_DEVICE) cmd->data_out_desc_cnt = count; else cmd->data_in_desc_cnt = count; - buf->table_desc.va = cpu_to_be64(req->cmd->dma + - sizeof *cmd + - sizeof *buf); + buf->table_desc.va = + cpu_to_be64(req->cmd->dma + sizeof *cmd + sizeof *buf); buf->table_desc.key = - cpu_to_be32(target->srp_host->mr->rkey); + cpu_to_be32(target->srp_host->dev->mr->rkey); buf->table_desc.len = cpu_to_be32(count * sizeof (struct srp_direct_buf)); - for (i = 0; i < count; ++i) { - buf->desc_list[i].va = cpu_to_be64(sg_dma_address(&scat[i])); - buf->desc_list[i].key = - cpu_to_be32(target->srp_host->mr->rkey); - buf->desc_list[i].len = cpu_to_be32(sg_dma_len(&scat[i])); - - datalen += sg_dma_len(&scat[i]); - } - buf->len = cpu_to_be32(datalen); - - len = sizeof (struct srp_cmd) + - sizeof (struct srp_indirect_buf) + - count * sizeof (struct srp_direct_buf); } if (scmnd->sc_data_direction == DMA_TO_DEVICE) @@ -617,12 +766,6 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target, return len; } -static void srp_remove_req(struct srp_target_port *target, struct srp_request *req) -{ - srp_unmap_data(req->scmnd, target, req); - list_move_tail(&req->list, &target->free_reqs); -} - static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp) { struct srp_request *req; @@ -675,22 +818,17 @@ static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp) spin_unlock_irqrestore(target->scsi_host->host_lock, flags); } -static void srp_reconnect_work(void *target_ptr) -{ - struct srp_target_port *target = target_ptr; - - srp_reconnect_target(target); -} - static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc) { + struct ib_device *dev; struct srp_iu *iu; u8 opcode; iu = target->rx_ring[wc->wr_id & ~SRP_OP_RECV]; - dma_sync_single_for_cpu(target->srp_host->dev->dma_device, iu->dma, - target->max_ti_iu_len, DMA_FROM_DEVICE); + dev = target->srp_host->dev->dev; + ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len, + DMA_FROM_DEVICE); opcode = *(u8 *) iu->buf; @@ -726,15 +864,14 @@ static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc) break; } - dma_sync_single_for_device(target->srp_host->dev->dma_device, iu->dma, - target->max_ti_iu_len, DMA_FROM_DEVICE); + ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len, + DMA_FROM_DEVICE); } static void srp_completion(struct ib_cq *cq, void *target_ptr) { struct srp_target_port *target = target_ptr; struct ib_wc wc; - unsigned long flags; ib_req_notify_cq(cq, IB_CQ_NEXT_COMP); while (ib_poll_cq(cq, 1, &wc) > 0) { @@ -742,10 +879,7 @@ static void srp_completion(struct ib_cq *cq, void *target_ptr) printk(KERN_ERR PFX "failed %s status %d\n", wc.wr_id & SRP_OP_RECV ? "receive" : "send", wc.status); - spin_lock_irqsave(target->scsi_host->host_lock, flags); - if (target->state == SRP_TARGET_LIVE) - schedule_work(&target->work); - spin_unlock_irqrestore(target->scsi_host->host_lock, flags); + target->qp_in_error = 1; break; } @@ -770,7 +904,7 @@ static int __srp_post_recv(struct srp_target_port *target) list.addr = iu->dma; list.length = iu->size; - list.lkey = target->srp_host->mr->lkey; + list.lkey = target->srp_host->dev->mr->lkey; wr.next = NULL; wr.sg_list = &list; @@ -805,12 +939,8 @@ static struct srp_iu *__srp_get_tx_iu(struct srp_target_port *target) if (target->tx_head - target->tx_tail >= SRP_SQ_SIZE) return NULL; - if (unlikely(target->req_lim < 1)) { - if (printk_ratelimit()) - printk(KERN_DEBUG PFX "Target has req_lim %d\n", - target->req_lim); - return NULL; - } + if (unlikely(target->req_lim < 1)) + ++target->zero_req_lim; return target->tx_ring[target->tx_head & SRP_SQ_SIZE]; } @@ -828,7 +958,7 @@ static int __srp_post_send(struct srp_target_port *target, list.addr = iu->dma; list.length = len; - list.lkey = target->srp_host->mr->lkey; + list.lkey = target->srp_host->dev->mr->lkey; wr.next = NULL; wr.wr_id = target->tx_head & SRP_SQ_SIZE; @@ -854,6 +984,7 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd, struct srp_request *req; struct srp_iu *iu; struct srp_cmd *cmd; + struct ib_device *dev; int len; if (target->state == SRP_TARGET_CONNECTING) @@ -870,8 +1001,9 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd, if (!iu) goto err; - dma_sync_single_for_cpu(target->srp_host->dev->dma_device, iu->dma, - SRP_MAX_IU_LEN, DMA_TO_DEVICE); + dev = target->srp_host->dev->dev; + ib_dma_sync_single_for_cpu(dev, iu->dma, srp_max_iu_len, + DMA_TO_DEVICE); req = list_entry(target->free_reqs.next, struct srp_request, list); @@ -903,8 +1035,8 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd, goto err_unmap; } - dma_sync_single_for_device(target->srp_host->dev->dma_device, iu->dma, - SRP_MAX_IU_LEN, DMA_TO_DEVICE); + ib_dma_sync_single_for_device(dev, iu->dma, srp_max_iu_len, + DMA_TO_DEVICE); if (__srp_post_send(target, iu, len)) { printk(KERN_ERR PFX "Send failed\n"); @@ -936,7 +1068,7 @@ static int srp_alloc_iu_bufs(struct srp_target_port *target) for (i = 0; i < SRP_SQ_SIZE + 1; ++i) { target->tx_ring[i] = srp_alloc_iu(target->srp_host, - SRP_MAX_IU_LEN, + srp_max_iu_len, GFP_KERNEL, DMA_TO_DEVICE); if (!target->tx_ring[i]) goto err; @@ -1062,9 +1194,11 @@ static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) break; } - target->status = srp_alloc_iu_bufs(target); - if (target->status) - break; + if (!target->rx_ring[0]) { + target->status = srp_alloc_iu_bufs(target); + if (target->status) + break; + } qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL); if (!qp_attr) { @@ -1107,11 +1241,10 @@ static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) srp_cm_rej_handler(cm_id, event, target); break; - case IB_CM_MRA_RECEIVED: - printk(KERN_ERR PFX "MRA received\n"); - break; - - case IB_CM_DREP_RECEIVED: + case IB_CM_DREQ_RECEIVED: + printk(KERN_WARNING PFX "DREQ received - connection closed\n"); + if (ib_send_cm_drep(cm_id, NULL, 0)) + printk(KERN_ERR PFX "Sending CM DREP failed\n"); break; case IB_CM_TIMEWAIT_EXIT: @@ -1121,6 +1254,11 @@ static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) target->status = 0; break; + case IB_CM_MRA_RECEIVED: + case IB_CM_DREQ_ERROR: + case IB_CM_DREP_RECEIVED: + break; + default: printk(KERN_WARNING PFX "Unhandled CM event %d\n", event->event); break; @@ -1201,6 +1339,8 @@ static int srp_abort(struct scsi_cmnd *scmnd) printk(KERN_ERR "SRP abort called\n"); + if (target->qp_in_error) + return FAILED; if (srp_find_req(target, scmnd, &req)) return FAILED; if (srp_send_tsk_mgmt(target, req, SRP_TSK_ABORT_TASK)) @@ -1229,6 +1369,8 @@ static int srp_reset_device(struct scsi_cmnd *scmnd) printk(KERN_ERR "SRP reset_device called\n"); + if (target->qp_in_error) + return FAILED; if (srp_find_req(target, scmnd, &req)) return FAILED; if (srp_send_tsk_mgmt(target, req, SRP_TSK_LUN_RESET)) @@ -1239,11 +1381,8 @@ static int srp_reset_device(struct scsi_cmnd *scmnd) spin_lock_irq(target->scsi_host->host_lock); list_for_each_entry_safe(req, tmp, &target->req_queue, list) - if (req->scmnd->device == scmnd->device) { - req->scmnd->result = DID_RESET << 16; - req->scmnd->scsi_done(req->scmnd); - srp_remove_req(target, req); - } + if (req->scmnd->device == scmnd->device) + srp_reset_req(target, req); spin_unlock_irq(target->scsi_host->host_lock); @@ -1329,11 +1468,59 @@ static ssize_t show_dgid(struct class_device *cdev, char *buf) be16_to_cpu(((__be16 *) target->path.dgid.raw)[7])); } -static CLASS_DEVICE_ATTR(id_ext, S_IRUGO, show_id_ext, NULL); -static CLASS_DEVICE_ATTR(ioc_guid, S_IRUGO, show_ioc_guid, NULL); -static CLASS_DEVICE_ATTR(service_id, S_IRUGO, show_service_id, NULL); -static CLASS_DEVICE_ATTR(pkey, S_IRUGO, show_pkey, NULL); -static CLASS_DEVICE_ATTR(dgid, S_IRUGO, show_dgid, NULL); +static ssize_t show_orig_dgid(struct class_device *cdev, char *buf) +{ + struct srp_target_port *target = host_to_target(class_to_shost(cdev)); + + if (target->state == SRP_TARGET_DEAD || + target->state == SRP_TARGET_REMOVED) + return -ENODEV; + + return sprintf(buf, "%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n", + be16_to_cpu(target->orig_dgid[0]), + be16_to_cpu(target->orig_dgid[1]), + be16_to_cpu(target->orig_dgid[2]), + be16_to_cpu(target->orig_dgid[3]), + be16_to_cpu(target->orig_dgid[4]), + be16_to_cpu(target->orig_dgid[5]), + be16_to_cpu(target->orig_dgid[6]), + be16_to_cpu(target->orig_dgid[7])); +} + +static ssize_t show_zero_req_lim(struct class_device *cdev, char *buf) +{ + struct srp_target_port *target = host_to_target(class_to_shost(cdev)); + + if (target->state == SRP_TARGET_DEAD || + target->state == SRP_TARGET_REMOVED) + return -ENODEV; + + return sprintf(buf, "%d\n", target->zero_req_lim); +} + +static ssize_t show_local_ib_port(struct class_device *cdev, char *buf) +{ + struct srp_target_port *target = host_to_target(class_to_shost(cdev)); + + return sprintf(buf, "%d\n", target->srp_host->port); +} + +static ssize_t show_local_ib_device(struct class_device *cdev, char *buf) +{ + struct srp_target_port *target = host_to_target(class_to_shost(cdev)); + + return sprintf(buf, "%s\n", target->srp_host->dev->dev->name); +} + +static CLASS_DEVICE_ATTR(id_ext, S_IRUGO, show_id_ext, NULL); +static CLASS_DEVICE_ATTR(ioc_guid, S_IRUGO, show_ioc_guid, NULL); +static CLASS_DEVICE_ATTR(service_id, S_IRUGO, show_service_id, NULL); +static CLASS_DEVICE_ATTR(pkey, S_IRUGO, show_pkey, NULL); +static CLASS_DEVICE_ATTR(dgid, S_IRUGO, show_dgid, NULL); +static CLASS_DEVICE_ATTR(orig_dgid, S_IRUGO, show_orig_dgid, NULL); +static CLASS_DEVICE_ATTR(zero_req_lim, S_IRUGO, show_zero_req_lim, NULL); +static CLASS_DEVICE_ATTR(local_ib_port, S_IRUGO, show_local_ib_port, NULL); +static CLASS_DEVICE_ATTR(local_ib_device, S_IRUGO, show_local_ib_device, NULL); static struct class_device_attribute *srp_host_attrs[] = { &class_device_attr_id_ext, @@ -1341,12 +1528,17 @@ static struct class_device_attribute *srp_host_attrs[] = { &class_device_attr_service_id, &class_device_attr_pkey, &class_device_attr_dgid, + &class_device_attr_orig_dgid, + &class_device_attr_zero_req_lim, + &class_device_attr_local_ib_port, + &class_device_attr_local_ib_device, NULL }; static struct scsi_host_template srp_template = { .module = THIS_MODULE, - .name = DRV_NAME, + .name = "InfiniBand SRP initiator", + .proc_name = DRV_NAME, .info = srp_target_info, .queuecommand = srp_queuecommand, .eh_abort_handler = srp_abort, @@ -1354,7 +1546,6 @@ static struct scsi_host_template srp_template = { .eh_host_reset_handler = srp_reset_host, .can_queue = SRP_SQ_SIZE, .this_id = -1, - .sg_tablesize = SRP_MAX_INDIRECT, .cmd_per_lun = SRP_SQ_SIZE, .use_clustering = ENABLE_CLUSTERING, .shost_attrs = srp_host_attrs @@ -1365,18 +1556,17 @@ static int srp_add_target(struct srp_host *host, struct srp_target_port *target) sprintf(target->target_name, "SRP.T10:%016llX", (unsigned long long) be64_to_cpu(target->id_ext)); - if (scsi_add_host(target->scsi_host, host->dev->dma_device)) + if (scsi_add_host(target->scsi_host, host->dev->dev->dma_device)) return -ENODEV; - mutex_lock(&host->target_mutex); + spin_lock(&host->target_lock); list_add_tail(&target->list, &host->target_list); - mutex_unlock(&host->target_mutex); + spin_unlock(&host->target_lock); target->state = SRP_TARGET_LIVE; - /* XXX: are we supposed to have a definition of SCAN_WILD_CARD ?? */ scsi_scan_target(&target->scsi_host->shost_gendev, - 0, target->scsi_id, ~0, 0); + 0, target->scsi_id, SCAN_WILD_CARD, 0); return 0; } @@ -1410,6 +1600,9 @@ enum { SRP_OPT_PKEY = 1 << 3, SRP_OPT_SERVICE_ID = 1 << 4, SRP_OPT_MAX_SECT = 1 << 5, + SRP_OPT_MAX_CMD_PER_LUN = 1 << 6, + SRP_OPT_IO_CLASS = 1 << 7, + SRP_OPT_INITIATOR_EXT = 1 << 8, SRP_OPT_ALL = (SRP_OPT_ID_EXT | SRP_OPT_IOC_GUID | SRP_OPT_DGID | @@ -1418,13 +1611,16 @@ enum { }; static match_table_t srp_opt_tokens = { - { SRP_OPT_ID_EXT, "id_ext=%s" }, - { SRP_OPT_IOC_GUID, "ioc_guid=%s" }, - { SRP_OPT_DGID, "dgid=%s" }, - { SRP_OPT_PKEY, "pkey=%x" }, - { SRP_OPT_SERVICE_ID, "service_id=%s" }, - { SRP_OPT_MAX_SECT, "max_sect=%d" }, - { SRP_OPT_ERR, NULL } + { SRP_OPT_ID_EXT, "id_ext=%s" }, + { SRP_OPT_IOC_GUID, "ioc_guid=%s" }, + { SRP_OPT_DGID, "dgid=%s" }, + { SRP_OPT_PKEY, "pkey=%x" }, + { SRP_OPT_SERVICE_ID, "service_id=%s" }, + { SRP_OPT_MAX_SECT, "max_sect=%d" }, + { SRP_OPT_MAX_CMD_PER_LUN, "max_cmd_per_lun=%d" }, + { SRP_OPT_IO_CLASS, "io_class=%x" }, + { SRP_OPT_INITIATOR_EXT, "initiator_ext=%s" }, + { SRP_OPT_ERR, NULL } }; static int srp_parse_options(const char *buf, struct srp_target_port *target) @@ -1453,18 +1649,30 @@ static int srp_parse_options(const char *buf, struct srp_target_port *target) switch (token) { case SRP_OPT_ID_EXT: p = match_strdup(args); + if (!p) { + ret = -ENOMEM; + goto out; + } target->id_ext = cpu_to_be64(simple_strtoull(p, NULL, 16)); kfree(p); break; case SRP_OPT_IOC_GUID: p = match_strdup(args); + if (!p) { + ret = -ENOMEM; + goto out; + } target->ioc_guid = cpu_to_be64(simple_strtoull(p, NULL, 16)); kfree(p); break; case SRP_OPT_DGID: p = match_strdup(args); + if (!p) { + ret = -ENOMEM; + goto out; + } if (strlen(p) != 32) { printk(KERN_WARNING PFX "bad dest GID parameter '%s'\n", p); kfree(p); @@ -1476,6 +1684,7 @@ static int srp_parse_options(const char *buf, struct srp_target_port *target) target->path.dgid.raw[i] = simple_strtoul(dgid, NULL, 16); } kfree(p); + memcpy(target->orig_dgid, target->path.dgid.raw, 16); break; case SRP_OPT_PKEY: @@ -1488,6 +1697,10 @@ static int srp_parse_options(const char *buf, struct srp_target_port *target) case SRP_OPT_SERVICE_ID: p = match_strdup(args); + if (!p) { + ret = -ENOMEM; + goto out; + } target->service_id = cpu_to_be64(simple_strtoull(p, NULL, 16)); kfree(p); break; @@ -1500,6 +1713,39 @@ static int srp_parse_options(const char *buf, struct srp_target_port *target) target->scsi_host->max_sectors = token; break; + case SRP_OPT_MAX_CMD_PER_LUN: + if (match_int(args, &token)) { + printk(KERN_WARNING PFX "bad max cmd_per_lun parameter '%s'\n", p); + goto out; + } + target->scsi_host->cmd_per_lun = min(token, SRP_SQ_SIZE); + break; + + case SRP_OPT_IO_CLASS: + if (match_hex(args, &token)) { + printk(KERN_WARNING PFX "bad IO class parameter '%s' \n", p); + goto out; + } + if (token != SRP_REV10_IB_IO_CLASS && + token != SRP_REV16A_IB_IO_CLASS) { + printk(KERN_WARNING PFX "unknown IO class parameter value" + " %x specified (use %x or %x).\n", + token, SRP_REV10_IB_IO_CLASS, SRP_REV16A_IB_IO_CLASS); + goto out; + } + target->io_class = token; + break; + + case SRP_OPT_INITIATOR_EXT: + p = match_strdup(args); + if (!p) { + ret = -ENOMEM; + goto out; + } + target->initiator_ext = cpu_to_be64(simple_strtoull(p, NULL, 16)); + kfree(p); + break; + default: printk(KERN_WARNING PFX "unknown parameter or missing value " "'%s' in target creation request\n", p); @@ -1537,16 +1783,15 @@ static ssize_t srp_create_target(struct class_device *class_dev, if (!target_host) return -ENOMEM; - target_host->max_lun = SRP_MAX_LUN; + target_host->max_lun = SRP_MAX_LUN; + target_host->max_cmd_len = sizeof ((struct srp_cmd *) (void *) 0L)->cdb; target = host_to_target(target_host); - memset(target, 0, sizeof *target); + target->io_class = SRP_REV16A_IB_IO_CLASS; target->scsi_host = target_host; target->srp_host = host; - INIT_WORK(&target->work, srp_reconnect_work, target); - INIT_LIST_HEAD(&target->free_reqs); INIT_LIST_HEAD(&target->req_queue); for (i = 0; i < SRP_SQ_SIZE; ++i) { @@ -1558,7 +1803,7 @@ static ssize_t srp_create_target(struct class_device *class_dev, if (ret) goto err; - ib_get_cached_gid(host->dev, host->port, 0, &target->path.sgid); + ib_get_cached_gid(host->dev->dev, host->port, 0, &target->path.sgid); printk(KERN_DEBUG PFX "new target: id_ext %016llx ioc_guid %016llx pkey %04x " "service_id %016llx dgid %04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n", @@ -1579,12 +1824,13 @@ static ssize_t srp_create_target(struct class_device *class_dev, if (ret) goto err; - target->cm_id = ib_create_cm_id(host->dev, srp_cm_handler, target); + target->cm_id = ib_create_cm_id(host->dev->dev, srp_cm_handler, target); if (IS_ERR(target->cm_id)) { ret = PTR_ERR(target->cm_id); goto err_free; } + target->qp_in_error = 0; ret = srp_connect_target(target); if (ret) { printk(KERN_ERR PFX "Connection failed\n"); @@ -1619,7 +1865,7 @@ static ssize_t show_ibdev(struct class_device *class_dev, char *buf) struct srp_host *host = container_of(class_dev, struct srp_host, class_dev); - return sprintf(buf, "%s\n", host->dev->name); + return sprintf(buf, "%s\n", host->dev->dev->name); } static CLASS_DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL); @@ -1634,7 +1880,7 @@ static ssize_t show_port(struct class_device *class_dev, char *buf) static CLASS_DEVICE_ATTR(port, S_IRUGO, show_port, NULL); -static struct srp_host *srp_add_port(struct ib_device *device, u8 port) +static struct srp_host *srp_add_port(struct srp_device *device, u8 port) { struct srp_host *host; @@ -1643,32 +1889,18 @@ static struct srp_host *srp_add_port(struct ib_device *device, u8 port) return NULL; INIT_LIST_HEAD(&host->target_list); - mutex_init(&host->target_mutex); + spin_lock_init(&host->target_lock); init_completion(&host->released); host->dev = device; host->port = port; - host->initiator_port_id[7] = port; - memcpy(host->initiator_port_id + 8, &device->node_guid, 8); - - host->pd = ib_alloc_pd(device); - if (IS_ERR(host->pd)) - goto err_free; - - host->mr = ib_get_dma_mr(host->pd, - IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_READ | - IB_ACCESS_REMOTE_WRITE); - if (IS_ERR(host->mr)) - goto err_pd; - host->class_dev.class = &srp_class; - host->class_dev.dev = device->dma_device; + host->class_dev.dev = device->dev->dma_device; snprintf(host->class_dev.class_id, BUS_ID_SIZE, "srp-%s-%d", - device->name, port); + device->dev->name, port); if (class_device_register(&host->class_dev)) - goto err_mr; + goto free_host; if (class_device_create_file(&host->class_dev, &class_device_attr_add_target)) goto err_class; if (class_device_create_file(&host->class_dev, &class_device_attr_ibdev)) @@ -1681,13 +1913,7 @@ static struct srp_host *srp_add_port(struct ib_device *device, u8 port) err_class: class_device_unregister(&host->class_dev); -err_mr: - ib_dereg_mr(host->mr); - -err_pd: - ib_dealloc_pd(host->pd); - -err_free: +free_host: kfree(host); return NULL; @@ -1695,17 +1921,64 @@ err_free: static void srp_add_one(struct ib_device *device) { - struct list_head *dev_list; + struct srp_device *srp_dev; + struct ib_device_attr *dev_attr; + struct ib_fmr_pool_param fmr_param; struct srp_host *host; int s, e, p; - dev_list = kmalloc(sizeof *dev_list, GFP_KERNEL); - if (!dev_list) + dev_attr = kmalloc(sizeof *dev_attr, GFP_KERNEL); + if (!dev_attr) return; - INIT_LIST_HEAD(dev_list); + if (ib_query_device(device, dev_attr)) { + printk(KERN_WARNING PFX "Query device failed for %s\n", + device->name); + goto free_attr; + } + + srp_dev = kmalloc(sizeof *srp_dev, GFP_KERNEL); + if (!srp_dev) + goto free_attr; + + /* + * Use the smallest page size supported by the HCA, down to a + * minimum of 512 bytes (which is the smallest sector that a + * SCSI command will ever carry). + */ + srp_dev->fmr_page_shift = max(9, ffs(dev_attr->page_size_cap) - 1); + srp_dev->fmr_page_size = 1 << srp_dev->fmr_page_shift; + srp_dev->fmr_page_mask = ~((u64) srp_dev->fmr_page_size - 1); + + INIT_LIST_HEAD(&srp_dev->dev_list); + + srp_dev->dev = device; + srp_dev->pd = ib_alloc_pd(device); + if (IS_ERR(srp_dev->pd)) + goto free_dev; + + srp_dev->mr = ib_get_dma_mr(srp_dev->pd, + IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_READ | + IB_ACCESS_REMOTE_WRITE); + if (IS_ERR(srp_dev->mr)) + goto err_pd; - if (device->node_type == IB_NODE_SWITCH) { + memset(&fmr_param, 0, sizeof fmr_param); + fmr_param.pool_size = SRP_FMR_POOL_SIZE; + fmr_param.dirty_watermark = SRP_FMR_DIRTY_SIZE; + fmr_param.cache = 1; + fmr_param.max_pages_per_fmr = SRP_FMR_SIZE; + fmr_param.page_shift = srp_dev->fmr_page_shift; + fmr_param.access = (IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE | + IB_ACCESS_REMOTE_READ); + + srp_dev->fmr_pool = ib_create_fmr_pool(srp_dev->pd, &fmr_param); + if (IS_ERR(srp_dev->fmr_pool)) + srp_dev->fmr_pool = NULL; + + if (device->node_type == RDMA_NODE_IB_SWITCH) { s = 0; e = 0; } else { @@ -1714,25 +1987,35 @@ static void srp_add_one(struct ib_device *device) } for (p = s; p <= e; ++p) { - host = srp_add_port(device, p); + host = srp_add_port(srp_dev, p); if (host) - list_add_tail(&host->list, dev_list); + list_add_tail(&host->list, &srp_dev->dev_list); } - ib_set_client_data(device, &srp_client, dev_list); + ib_set_client_data(device, &srp_client, srp_dev); + + goto free_attr; + +err_pd: + ib_dealloc_pd(srp_dev->pd); + +free_dev: + kfree(srp_dev); + +free_attr: + kfree(dev_attr); } static void srp_remove_one(struct ib_device *device) { - struct list_head *dev_list; + struct srp_device *srp_dev; struct srp_host *host, *tmp_host; LIST_HEAD(target_list); struct srp_target_port *target, *tmp_target; - unsigned long flags; - dev_list = ib_get_client_data(device, &srp_client); + srp_dev = ib_get_client_data(device, &srp_client); - list_for_each_entry_safe(host, tmp_host, dev_list, list) { + list_for_each_entry_safe(host, tmp_host, &srp_dev->dev_list, list) { class_device_unregister(&host->class_dev); /* * Wait for the sysfs entry to go away, so that no new @@ -1744,15 +2027,13 @@ static void srp_remove_one(struct ib_device *device) * Mark all target ports as removed, so we stop queueing * commands and don't try to reconnect. */ - mutex_lock(&host->target_mutex); - list_for_each_entry_safe(target, tmp_target, - &host->target_list, list) { - spin_lock_irqsave(target->scsi_host->host_lock, flags); - if (target->state != SRP_TARGET_REMOVED) - target->state = SRP_TARGET_REMOVED; - spin_unlock_irqrestore(target->scsi_host->host_lock, flags); + spin_lock(&host->target_lock); + list_for_each_entry(target, &host->target_list, list) { + spin_lock_irq(target->scsi_host->host_lock); + target->state = SRP_TARGET_REMOVED; + spin_unlock_irq(target->scsi_host->host_lock); } - mutex_unlock(&host->target_mutex); + spin_unlock(&host->target_lock); /* * Wait for any reconnection tasks that may have @@ -1770,27 +2051,38 @@ static void srp_remove_one(struct ib_device *device) scsi_host_put(target->scsi_host); } - ib_dereg_mr(host->mr); - ib_dealloc_pd(host->pd); kfree(host); } - kfree(dev_list); + if (srp_dev->fmr_pool) + ib_destroy_fmr_pool(srp_dev->fmr_pool); + ib_dereg_mr(srp_dev->mr); + ib_dealloc_pd(srp_dev->pd); + + kfree(srp_dev); } static int __init srp_init_module(void) { int ret; + srp_template.sg_tablesize = srp_sg_tablesize; + srp_max_iu_len = (sizeof (struct srp_cmd) + + sizeof (struct srp_indirect_buf) + + srp_sg_tablesize * 16); + ret = class_register(&srp_class); if (ret) { printk(KERN_ERR PFX "couldn't register class infiniband_srp\n"); return ret; } + ib_sa_register_client(&srp_sa_client); + ret = ib_register_client(&srp_client); if (ret) { printk(KERN_ERR PFX "couldn't register IB client\n"); + ib_sa_unregister_client(&srp_sa_client); class_unregister(&srp_class); return ret; } @@ -1801,6 +2093,7 @@ static int __init srp_init_module(void) static void __exit srp_cleanup_module(void) { ib_unregister_client(&srp_client); + ib_sa_unregister_client(&srp_sa_client); class_unregister(&srp_class); }