iser_verbs.c

linux内核源码

	}

	return;
failure:
	iser_connect_error(cma_id);
}

static void iser_connected_handler(struct rdma_cm_id *cma_id)
{
	struct iser_conn *ib_conn;

	ib_conn = (struct iser_conn *)cma_id->context;
	ib_conn->state = ISER_CONN_UP;
	wake_up_interruptible(&ib_conn->wait);
}

static void iser_disconnected_handler(struct rdma_cm_id *cma_id)
{
	struct iser_conn *ib_conn;

	ib_conn = (struct iser_conn *)cma_id->context;
	ib_conn->disc_evt_flag = 1;

	/* getting here when the state is UP means that the conn is being *
	 * terminated asynchronously from the iSCSI layer's perspective.  */
	if (iser_conn_state_comp_exch(ib_conn, ISER_CONN_UP,
				      ISER_CONN_TERMINATING))
		iscsi_conn_failure(ib_conn->iser_conn->iscsi_conn,
				   ISCSI_ERR_CONN_FAILED);

	/* Complete the termination process if no posts are pending */
	if ((atomic_read(&ib_conn->post_recv_buf_count) == 0) &&
	    (atomic_read(&ib_conn->post_send_buf_count) == 0)) {
		ib_conn->state = ISER_CONN_DOWN;
		wake_up_interruptible(&ib_conn->wait);
	}
}

static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *event)
{
	int ret = 0;

	iser_err("event %d conn %p id %p\n",event->event,cma_id->context,cma_id);

	switch (event->event) {
	case RDMA_CM_EVENT_ADDR_RESOLVED:
		iser_addr_handler(cma_id);
		break;
	case RDMA_CM_EVENT_ROUTE_RESOLVED:
		iser_route_handler(cma_id);
		break;
	case RDMA_CM_EVENT_ESTABLISHED:
		iser_connected_handler(cma_id);
		break;
	case RDMA_CM_EVENT_ADDR_ERROR:
	case RDMA_CM_EVENT_ROUTE_ERROR:
	case RDMA_CM_EVENT_CONNECT_ERROR:
	case RDMA_CM_EVENT_UNREACHABLE:
	case RDMA_CM_EVENT_REJECTED:
		iser_err("event: %d, error: %d\n", event->event, event->status);
		iser_connect_error(cma_id);
		break;
	case RDMA_CM_EVENT_DISCONNECTED:
		iser_disconnected_handler(cma_id);
		break;
	case RDMA_CM_EVENT_DEVICE_REMOVAL:
		BUG();
		break;
	case RDMA_CM_EVENT_CONNECT_RESPONSE:
		BUG();
		break;
	case RDMA_CM_EVENT_CONNECT_REQUEST:
	default:
		break;
	}
	return ret;
}

int iser_conn_init(struct iser_conn **ibconn)
{
	struct iser_conn *ib_conn;

	ib_conn = kzalloc(sizeof *ib_conn, GFP_KERNEL);
	if (!ib_conn) {
		iser_err("can't alloc memory for struct iser_conn\n");
		return -ENOMEM;
	}
	ib_conn->state = ISER_CONN_INIT;
	init_waitqueue_head(&ib_conn->wait);
	atomic_set(&ib_conn->post_recv_buf_count, 0);
	atomic_set(&ib_conn->post_send_buf_count, 0);
	INIT_LIST_HEAD(&ib_conn->conn_list);
	spin_lock_init(&ib_conn->lock);

	*ibconn = ib_conn;
	return 0;
}

 /**
 * starts the process of connecting to the target
 * sleeps untill the connection is established or rejected
 */
int iser_connect(struct iser_conn   *ib_conn,
		 struct sockaddr_in *src_addr,
		 struct sockaddr_in *dst_addr,
		 int                 non_blocking)
{
	struct sockaddr *src, *dst;
	int err = 0;

	sprintf(ib_conn->name,"%d.%d.%d.%d:%d",
		NIPQUAD(dst_addr->sin_addr.s_addr), dst_addr->sin_port);

	/* the device is known only --after-- address resolution */
	ib_conn->device = NULL;

	iser_err("connecting to: %d.%d.%d.%d, port 0x%x\n",
		 NIPQUAD(dst_addr->sin_addr), dst_addr->sin_port);

	ib_conn->state = ISER_CONN_PENDING;

	ib_conn->cma_id = rdma_create_id(iser_cma_handler,
					     (void *)ib_conn,
					     RDMA_PS_TCP);
	if (IS_ERR(ib_conn->cma_id)) {
		err = PTR_ERR(ib_conn->cma_id);
		iser_err("rdma_create_id failed: %d\n", err);
		goto id_failure;
	}

	src = (struct sockaddr *)src_addr;
	dst = (struct sockaddr *)dst_addr;
	err = rdma_resolve_addr(ib_conn->cma_id, src, dst, 1000);
	if (err) {
		iser_err("rdma_resolve_addr failed: %d\n", err);
		goto addr_failure;
	}

	if (!non_blocking) {
		wait_event_interruptible(ib_conn->wait,
					 (ib_conn->state != ISER_CONN_PENDING));

		if (ib_conn->state != ISER_CONN_UP) {
			err =  -EIO;
			goto connect_failure;
		}
	}

	mutex_lock(&ig.connlist_mutex);
	list_add(&ib_conn->conn_list, &ig.connlist);
	mutex_unlock(&ig.connlist_mutex);
	return 0;

id_failure:
	ib_conn->cma_id = NULL;
addr_failure:
	ib_conn->state = ISER_CONN_DOWN;
connect_failure:
	iser_conn_release(ib_conn);
	return err;
}

/**
 * iser_reg_page_vec - Register physical memory
 *
 * returns: 0 on success, errno code on failure
 */
int iser_reg_page_vec(struct iser_conn     *ib_conn,
		      struct iser_page_vec *page_vec,
		      struct iser_mem_reg  *mem_reg)
{
	struct ib_pool_fmr *mem;
	u64		   io_addr;
	u64		   *page_list;
	int		   status;

	page_list = page_vec->pages;
	io_addr	  = page_list[0];

	mem  = ib_fmr_pool_map_phys(ib_conn->fmr_pool,
				    page_list,
				    page_vec->length,
				    io_addr);

	if (IS_ERR(mem)) {
		status = (int)PTR_ERR(mem);
		iser_err("ib_fmr_pool_map_phys failed: %d\n", status);
		return status;
	}

	mem_reg->lkey  = mem->fmr->lkey;
	mem_reg->rkey  = mem->fmr->rkey;
	mem_reg->len   = page_vec->length * SIZE_4K;
	mem_reg->va    = io_addr;
	mem_reg->is_fmr = 1;
	mem_reg->mem_h = (void *)mem;

	mem_reg->va   += page_vec->offset;
	mem_reg->len   = page_vec->data_size;

	iser_dbg("PHYSICAL Mem.register, [PHYS p_array: 0x%p, sz: %d, "
		 "entry[0]: (0x%08lx,%ld)] -> "
		 "[lkey: 0x%08X mem_h: 0x%p va: 0x%08lX sz: %ld]\n",
		 page_vec, page_vec->length,
		 (unsigned long)page_vec->pages[0],
		 (unsigned long)page_vec->data_size,
		 (unsigned int)mem_reg->lkey, mem_reg->mem_h,
		 (unsigned long)mem_reg->va, (unsigned long)mem_reg->len);
	return 0;
}

/**
 * Unregister (previosuly registered) memory.
 */
void iser_unreg_mem(struct iser_mem_reg *reg)
{
	int ret;

	iser_dbg("PHYSICAL Mem.Unregister mem_h %p\n",reg->mem_h);

	ret = ib_fmr_pool_unmap((struct ib_pool_fmr *)reg->mem_h);
	if (ret)
		iser_err("ib_fmr_pool_unmap failed %d\n", ret);

	reg->mem_h = NULL;
}

/**
 * iser_dto_to_iov - builds IOV from a dto descriptor
 */
static void iser_dto_to_iov(struct iser_dto *dto, struct ib_sge *iov, int iov_len)
{
	int		     i;
	struct ib_sge	     *sge;
	struct iser_regd_buf *regd_buf;

	if (dto->regd_vector_len > iov_len) {
		iser_err("iov size %d too small for posting dto of len %d\n",
			 iov_len, dto->regd_vector_len);
		BUG();
	}

	for (i = 0; i < dto->regd_vector_len; i++) {
		sge	    = &iov[i];
		regd_buf  = dto->regd[i];

		sge->addr   = regd_buf->reg.va;
		sge->length = regd_buf->reg.len;
		sge->lkey   = regd_buf->reg.lkey;

		if (dto->used_sz[i] > 0)  /* Adjust size */
			sge->length = dto->used_sz[i];

		/* offset and length should not exceed the regd buf length */
		if (sge->length + dto->offset[i] > regd_buf->reg.len) {
			iser_err("Used len:%ld + offset:%d, exceed reg.buf.len:"
				 "%ld in dto:0x%p [%d], va:0x%08lX\n",
				 (unsigned long)sge->length, dto->offset[i],
				 (unsigned long)regd_buf->reg.len, dto, i,
				 (unsigned long)sge->addr);
			BUG();
		}

		sge->addr += dto->offset[i]; /* Adjust offset */
	}
}

/**
 * iser_post_recv - Posts a receive buffer.
 *
 * returns 0 on success, -1 on failure
 */
int iser_post_recv(struct iser_desc *rx_desc)
{
	int		  ib_ret, ret_val = 0;
	struct ib_recv_wr recv_wr, *recv_wr_failed;
	struct ib_sge	  iov[2];
	struct iser_conn  *ib_conn;
	struct iser_dto   *recv_dto = &rx_desc->dto;

	/* Retrieve conn */
	ib_conn = recv_dto->ib_conn;

	iser_dto_to_iov(recv_dto, iov, 2);

	recv_wr.next	= NULL;
	recv_wr.sg_list = iov;
	recv_wr.num_sge = recv_dto->regd_vector_len;
	recv_wr.wr_id	= (unsigned long)rx_desc;

	atomic_inc(&ib_conn->post_recv_buf_count);
	ib_ret	= ib_post_recv(ib_conn->qp, &recv_wr, &recv_wr_failed);
	if (ib_ret) {
		iser_err("ib_post_recv failed ret=%d\n", ib_ret);
		atomic_dec(&ib_conn->post_recv_buf_count);
		ret_val = -1;
	}

	return ret_val;
}

/**
 * iser_start_send - Initiate a Send DTO operation
 *
 * returns 0 on success, -1 on failure
 */
int iser_post_send(struct iser_desc *tx_desc)
{
	int		  ib_ret, ret_val = 0;
	struct ib_send_wr send_wr, *send_wr_failed;
	struct ib_sge	  iov[MAX_REGD_BUF_VECTOR_LEN];
	struct iser_conn  *ib_conn;
	struct iser_dto   *dto = &tx_desc->dto;

	ib_conn = dto->ib_conn;

	iser_dto_to_iov(dto, iov, MAX_REGD_BUF_VECTOR_LEN);

	send_wr.next	   = NULL;
	send_wr.wr_id	   = (unsigned long)tx_desc;
	send_wr.sg_list	   = iov;
	send_wr.num_sge	   = dto->regd_vector_len;
	send_wr.opcode	   = IB_WR_SEND;
	send_wr.send_flags = dto->notify_enable ? IB_SEND_SIGNALED : 0;

	atomic_inc(&ib_conn->post_send_buf_count);

	ib_ret = ib_post_send(ib_conn->qp, &send_wr, &send_wr_failed);
	if (ib_ret) {
		iser_err("Failed to start SEND DTO, dto: 0x%p, IOV len: %d\n",
			 dto, dto->regd_vector_len);
		iser_err("ib_post_send failed, ret:%d\n", ib_ret);
		atomic_dec(&ib_conn->post_send_buf_count);
		ret_val = -1;
	}

	return ret_val;
}

static void iser_handle_comp_error(struct iser_desc *desc)
{
	struct iser_dto  *dto     = &desc->dto;
	struct iser_conn *ib_conn = dto->ib_conn;

	iser_dto_buffs_release(dto);

	if (desc->type == ISCSI_RX) {
		kfree(desc->data);
		kmem_cache_free(ig.desc_cache, desc);
		atomic_dec(&ib_conn->post_recv_buf_count);
	} else { /* type is TX control/command/dataout */
		if (desc->type == ISCSI_TX_DATAOUT)
			kmem_cache_free(ig.desc_cache, desc);
		atomic_dec(&ib_conn->post_send_buf_count);
	}

	if (atomic_read(&ib_conn->post_recv_buf_count) == 0 &&
	    atomic_read(&ib_conn->post_send_buf_count) == 0) {
		/* getting here when the state is UP means that the conn is *
		 * being terminated asynchronously from the iSCSI layer's   *
		 * perspective.                                             */
		if (iser_conn_state_comp_exch(ib_conn, ISER_CONN_UP,
		    ISER_CONN_TERMINATING))
			iscsi_conn_failure(ib_conn->iser_conn->iscsi_conn,
					   ISCSI_ERR_CONN_FAILED);

		/* complete the termination process if disconnect event was delivered *
		 * note there are no more non completed posts to the QP               */
		if (ib_conn->disc_evt_flag) {
			ib_conn->state = ISER_CONN_DOWN;
			wake_up_interruptible(&ib_conn->wait);
		}
	}
}

static void iser_cq_tasklet_fn(unsigned long data)
{
	 struct iser_device  *device = (struct iser_device *)data;
	 struct ib_cq	     *cq = device->cq;
	 struct ib_wc	     wc;
	 struct iser_desc    *desc;
	 unsigned long	     xfer_len;

	while (ib_poll_cq(cq, 1, &wc) == 1) {
		desc	 = (struct iser_desc *) (unsigned long) wc.wr_id;
		BUG_ON(desc == NULL);

		if (wc.status == IB_WC_SUCCESS) {
			if (desc->type == ISCSI_RX) {
				xfer_len = (unsigned long)wc.byte_len;
				iser_rcv_completion(desc, xfer_len);
			} else /* type == ISCSI_TX_CONTROL/SCSI_CMD/DOUT */
				iser_snd_completion(desc);
		} else {
			iser_err("comp w. error op %d status %d\n",desc->type,wc.status);
			iser_handle_comp_error(desc);
		}
	}
	/* #warning "it is assumed here that arming CQ only once its empty" *
	 * " would not cause interrupts to be missed"                       */
	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
}

static void iser_cq_callback(struct ib_cq *cq, void *cq_context)
{
	struct iser_device  *device = (struct iser_device *)cq_context;

	tasklet_schedule(&device->cq_tasklet);
}