ipath_verbs.c

来自「linux 内核源代码」· C语言 代码 · 共 1,887 行 · 第 1/4 页

	return 0;
}

/**
 * ipath_get_npkeys - return the size of the PKEY table for port 0
 * @dd: the infinipath device
 */
unsigned ipath_get_npkeys(struct ipath_devdata *dd)
{
	return ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys);
}

/**
 * ipath_get_pkey - return the indexed PKEY from the port 0 PKEY table
 * @dd: the infinipath device
 * @index: the PKEY index
 */
unsigned ipath_get_pkey(struct ipath_devdata *dd, unsigned index)
{
	unsigned ret;

	if (index >= ARRAY_SIZE(dd->ipath_pd[0]->port_pkeys))
		ret = 0;
	else
		ret = dd->ipath_pd[0]->port_pkeys[index];

	return ret;
}

static int ipath_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
			    u16 *pkey)
{
	struct ipath_ibdev *dev = to_idev(ibdev);
	int ret;

	if (index >= ipath_get_npkeys(dev->dd)) {
		ret = -EINVAL;
		goto bail;
	}

	*pkey = ipath_get_pkey(dev->dd, index);
	ret = 0;

bail:
	return ret;
}

/**
 * ipath_alloc_ucontext - allocate a ucontest
 * @ibdev: the infiniband device
 * @udata: not used by the InfiniPath driver
 */

static struct ib_ucontext *ipath_alloc_ucontext(struct ib_device *ibdev,
						struct ib_udata *udata)
{
	struct ipath_ucontext *context;
	struct ib_ucontext *ret;

	context = kmalloc(sizeof *context, GFP_KERNEL);
	if (!context) {
		ret = ERR_PTR(-ENOMEM);
		goto bail;
	}

	ret = &context->ibucontext;

bail:
	return ret;
}

static int ipath_dealloc_ucontext(struct ib_ucontext *context)
{
	kfree(to_iucontext(context));
	return 0;
}

static int ipath_verbs_register_sysfs(struct ib_device *dev);

static void __verbs_timer(unsigned long arg)
{
	struct ipath_devdata *dd = (struct ipath_devdata *) arg;

	/* Handle verbs layer timeouts. */
	ipath_ib_timer(dd->verbs_dev);

	mod_timer(&dd->verbs_timer, jiffies + 1);
}

static int enable_timer(struct ipath_devdata *dd)
{
	/*
	 * Early chips had a design flaw where the chip and kernel idea
	 * of the tail register don't always agree, and therefore we won't
	 * get an interrupt on the next packet received.
	 * If the board supports per packet receive interrupts, use it.
	 * Otherwise, the timer function periodically checks for packets
	 * to cover this case.
	 * Either way, the timer is needed for verbs layer related
	 * processing.
	 */
	if (dd->ipath_flags & IPATH_GPIO_INTR) {
		ipath_write_kreg(dd, dd->ipath_kregs->kr_debugportselect,
				 0x2074076542310ULL);
		/* Enable GPIO bit 2 interrupt */
		dd->ipath_gpio_mask |= (u64) (1 << IPATH_GPIO_PORT0_BIT);
		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask,
				 dd->ipath_gpio_mask);
	}

	init_timer(&dd->verbs_timer);
	dd->verbs_timer.function = __verbs_timer;
	dd->verbs_timer.data = (unsigned long)dd;
	dd->verbs_timer.expires = jiffies + 1;
	add_timer(&dd->verbs_timer);

	return 0;
}

static int disable_timer(struct ipath_devdata *dd)
{
	/* Disable GPIO bit 2 interrupt */
	if (dd->ipath_flags & IPATH_GPIO_INTR) {
                /* Disable GPIO bit 2 interrupt */
		dd->ipath_gpio_mask &= ~((u64) (1 << IPATH_GPIO_PORT0_BIT));
		ipath_write_kreg(dd, dd->ipath_kregs->kr_gpio_mask,
				 dd->ipath_gpio_mask);
		/*
		 * We might want to undo changes to debugportselect,
		 * but how?
		 */
	}

	del_timer_sync(&dd->verbs_timer);

	return 0;
}

/**
 * ipath_register_ib_device - register our device with the infiniband core
 * @dd: the device data structure
 * Return the allocated ipath_ibdev pointer or NULL on error.
 */
int ipath_register_ib_device(struct ipath_devdata *dd)
{
	struct ipath_verbs_counters cntrs;
	struct ipath_ibdev *idev;
	struct ib_device *dev;
	int ret;

	idev = (struct ipath_ibdev *)ib_alloc_device(sizeof *idev);
	if (idev == NULL) {
		ret = -ENOMEM;
		goto bail;
	}

	dev = &idev->ibdev;

	/* Only need to initialize non-zero fields. */
	spin_lock_init(&idev->n_pds_lock);
	spin_lock_init(&idev->n_ahs_lock);
	spin_lock_init(&idev->n_cqs_lock);
	spin_lock_init(&idev->n_qps_lock);
	spin_lock_init(&idev->n_srqs_lock);
	spin_lock_init(&idev->n_mcast_grps_lock);

	spin_lock_init(&idev->qp_table.lock);
	spin_lock_init(&idev->lk_table.lock);
	idev->sm_lid = __constant_be16_to_cpu(IB_LID_PERMISSIVE);
	/* Set the prefix to the default value (see ch. 4.1.1) */
	idev->gid_prefix = __constant_cpu_to_be64(0xfe80000000000000ULL);

	ret = ipath_init_qp_table(idev, ib_ipath_qp_table_size);
	if (ret)
		goto err_qp;

	/*
	 * The top ib_ipath_lkey_table_size bits are used to index the
	 * table.  The lower 8 bits can be owned by the user (copied from
	 * the LKEY).  The remaining bits act as a generation number or tag.
	 */
	idev->lk_table.max = 1 << ib_ipath_lkey_table_size;
	idev->lk_table.table = kzalloc(idev->lk_table.max *
				       sizeof(*idev->lk_table.table),
				       GFP_KERNEL);
	if (idev->lk_table.table == NULL) {
		ret = -ENOMEM;
		goto err_lk;
	}
	INIT_LIST_HEAD(&idev->pending_mmaps);
	spin_lock_init(&idev->pending_lock);
	idev->mmap_offset = PAGE_SIZE;
	spin_lock_init(&idev->mmap_offset_lock);
	INIT_LIST_HEAD(&idev->pending[0]);
	INIT_LIST_HEAD(&idev->pending[1]);
	INIT_LIST_HEAD(&idev->pending[2]);
	INIT_LIST_HEAD(&idev->piowait);
	INIT_LIST_HEAD(&idev->rnrwait);
	idev->pending_index = 0;
	idev->port_cap_flags =
		IB_PORT_SYS_IMAGE_GUID_SUP | IB_PORT_CLIENT_REG_SUP;
	idev->pma_counter_select[0] = IB_PMA_PORT_XMIT_DATA;
	idev->pma_counter_select[1] = IB_PMA_PORT_RCV_DATA;
	idev->pma_counter_select[2] = IB_PMA_PORT_XMIT_PKTS;
	idev->pma_counter_select[3] = IB_PMA_PORT_RCV_PKTS;
	idev->pma_counter_select[4] = IB_PMA_PORT_XMIT_WAIT;
	idev->link_width_enabled = 3;	/* 1x or 4x */

	/* Snapshot current HW counters to "clear" them. */
	ipath_get_counters(dd, &cntrs);
	idev->z_symbol_error_counter = cntrs.symbol_error_counter;
	idev->z_link_error_recovery_counter =
		cntrs.link_error_recovery_counter;
	idev->z_link_downed_counter = cntrs.link_downed_counter;
	idev->z_port_rcv_errors = cntrs.port_rcv_errors;
	idev->z_port_rcv_remphys_errors =
		cntrs.port_rcv_remphys_errors;
	idev->z_port_xmit_discards = cntrs.port_xmit_discards;
	idev->z_port_xmit_data = cntrs.port_xmit_data;
	idev->z_port_rcv_data = cntrs.port_rcv_data;
	idev->z_port_xmit_packets = cntrs.port_xmit_packets;
	idev->z_port_rcv_packets = cntrs.port_rcv_packets;
	idev->z_local_link_integrity_errors =
		cntrs.local_link_integrity_errors;
	idev->z_excessive_buffer_overrun_errors =
		cntrs.excessive_buffer_overrun_errors;

	/*
	 * The system image GUID is supposed to be the same for all
	 * IB HCAs in a single system but since there can be other
	 * device types in the system, we can't be sure this is unique.
	 */
	if (!sys_image_guid)
		sys_image_guid = dd->ipath_guid;
	idev->sys_image_guid = sys_image_guid;
	idev->ib_unit = dd->ipath_unit;
	idev->dd = dd;

	strlcpy(dev->name, "ipath%d", IB_DEVICE_NAME_MAX);
	dev->owner = THIS_MODULE;
	dev->node_guid = dd->ipath_guid;
	dev->uverbs_abi_ver = IPATH_UVERBS_ABI_VERSION;
	dev->uverbs_cmd_mask =
		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE)	|
		(1ull << IB_USER_VERBS_CMD_QUERY_PORT)		|
		(1ull << IB_USER_VERBS_CMD_ALLOC_PD)		|
		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD)		|
		(1ull << IB_USER_VERBS_CMD_CREATE_AH)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_AH)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_AH)		|
		(1ull << IB_USER_VERBS_CMD_REG_MR)		|
		(1ull << IB_USER_VERBS_CMD_DEREG_MR)		|
		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
		(1ull << IB_USER_VERBS_CMD_CREATE_CQ)		|
		(1ull << IB_USER_VERBS_CMD_RESIZE_CQ)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ)		|
		(1ull << IB_USER_VERBS_CMD_POLL_CQ)		|
		(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ)	|
		(1ull << IB_USER_VERBS_CMD_CREATE_QP)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_QP)		|
		(1ull << IB_USER_VERBS_CMD_MODIFY_QP)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_QP)		|
		(1ull << IB_USER_VERBS_CMD_POST_SEND)		|
		(1ull << IB_USER_VERBS_CMD_POST_RECV)		|
		(1ull << IB_USER_VERBS_CMD_ATTACH_MCAST)	|
		(1ull << IB_USER_VERBS_CMD_DETACH_MCAST)	|
		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)		|
		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV);
	dev->node_type = RDMA_NODE_IB_CA;
	dev->phys_port_cnt = 1;
	dev->num_comp_vectors = 1;
	dev->dma_device = &dd->pcidev->dev;
	dev->query_device = ipath_query_device;
	dev->modify_device = ipath_modify_device;
	dev->query_port = ipath_query_port;
	dev->modify_port = ipath_modify_port;
	dev->query_pkey = ipath_query_pkey;
	dev->query_gid = ipath_query_gid;
	dev->alloc_ucontext = ipath_alloc_ucontext;
	dev->dealloc_ucontext = ipath_dealloc_ucontext;
	dev->alloc_pd = ipath_alloc_pd;
	dev->dealloc_pd = ipath_dealloc_pd;
	dev->create_ah = ipath_create_ah;
	dev->destroy_ah = ipath_destroy_ah;
	dev->query_ah = ipath_query_ah;
	dev->create_srq = ipath_create_srq;
	dev->modify_srq = ipath_modify_srq;
	dev->query_srq = ipath_query_srq;
	dev->destroy_srq = ipath_destroy_srq;
	dev->create_qp = ipath_create_qp;
	dev->modify_qp = ipath_modify_qp;
	dev->query_qp = ipath_query_qp;
	dev->destroy_qp = ipath_destroy_qp;
	dev->post_send = ipath_post_send;
	dev->post_recv = ipath_post_receive;
	dev->post_srq_recv = ipath_post_srq_receive;
	dev->create_cq = ipath_create_cq;
	dev->destroy_cq = ipath_destroy_cq;
	dev->resize_cq = ipath_resize_cq;
	dev->poll_cq = ipath_poll_cq;
	dev->req_notify_cq = ipath_req_notify_cq;
	dev->get_dma_mr = ipath_get_dma_mr;
	dev->reg_phys_mr = ipath_reg_phys_mr;
	dev->reg_user_mr = ipath_reg_user_mr;
	dev->dereg_mr = ipath_dereg_mr;
	dev->alloc_fmr = ipath_alloc_fmr;
	dev->map_phys_fmr = ipath_map_phys_fmr;
	dev->unmap_fmr = ipath_unmap_fmr;
	dev->dealloc_fmr = ipath_dealloc_fmr;
	dev->attach_mcast = ipath_multicast_attach;
	dev->detach_mcast = ipath_multicast_detach;
	dev->process_mad = ipath_process_mad;
	dev->mmap = ipath_mmap;
	dev->dma_ops = &ipath_dma_mapping_ops;

	snprintf(dev->node_desc, sizeof(dev->node_desc),
		 IPATH_IDSTR " %s", init_utsname()->nodename);

	ret = ib_register_device(dev);
	if (ret)
		goto err_reg;

	if (ipath_verbs_register_sysfs(dev))
		goto err_class;

	enable_timer(dd);

	goto bail;

err_class:
	ib_unregister_device(dev);
err_reg:
	kfree(idev->lk_table.table);
err_lk:
	kfree(idev->qp_table.table);
err_qp:
	ib_dealloc_device(dev);
	ipath_dev_err(dd, "cannot register verbs: %d!\n", -ret);
	idev = NULL;

bail:
	dd->verbs_dev = idev;
	return ret;
}

void ipath_unregister_ib_device(struct ipath_ibdev *dev)
{
	struct ib_device *ibdev = &dev->ibdev;

	disable_timer(dev->dd);

	ib_unregister_device(ibdev);

	if (!list_empty(&dev->pending[0]) ||
	    !list_empty(&dev->pending[1]) ||
	    !list_empty(&dev->pending[2]))
		ipath_dev_err(dev->dd, "pending list not empty!\n");
	if (!list_empty(&dev->piowait))
		ipath_dev_err(dev->dd, "piowait list not empty!\n");
	if (!list_empty(&dev->rnrwait))
		ipath_dev_err(dev->dd, "rnrwait list not empty!\n");
	if (!ipath_mcast_tree_empty())
		ipath_dev_err(dev->dd, "multicast table memory leak!\n");
	/*
	 * Note that ipath_unregister_ib_device() can be called before all
	 * the QPs are destroyed!
	 */
	ipath_free_all_qps(&dev->qp_table);
	kfree(dev->qp_table.table);
	kfree(dev->lk_table.table);
	ib_dealloc_device(ibdev);
}

static ssize_t show_rev(struct class_device *cdev, char *buf)
{
	struct ipath_ibdev *dev =
		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);

	return sprintf(buf, "%x\n", dev->dd->ipath_pcirev);
}

static ssize_t show_hca(struct class_device *cdev, char *buf)
{
	struct ipath_ibdev *dev =
		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);
	int ret;

	ret = dev->dd->ipath_f_get_boardname(dev->dd, buf, 128);
	if (ret < 0)
		goto bail;
	strcat(buf, "\n");
	ret = strlen(buf);

bail:
	return ret;
}

static ssize_t show_stats(struct class_device *cdev, char *buf)
{
	struct ipath_ibdev *dev =
		container_of(cdev, struct ipath_ibdev, ibdev.class_dev);
	int i;
	int len;

	len = sprintf(buf,
		      "RC resends  %d\n"
		      "RC no QACK  %d\n"
		      "RC ACKs     %d\n"
		      "RC SEQ NAKs %d\n"
		      "RC RDMA seq %d\n"
		      "RC RNR NAKs %d\n"
		      "RC OTH NAKs %d\n"
		      "RC timeouts %d\n"
		      "RC RDMA dup %d\n"
		      "RC stalls   %d\n"
		      "piobuf wait %d\n"
		      "no piobuf   %d\n"
		      "PKT drops   %d\n"
		      "WQE errs    %d\n",
		      dev->n_rc_resends, dev->n_rc_qacks, dev->n_rc_acks,
		      dev->n_seq_naks, dev->n_rdma_seq, dev->n_rnr_naks,
		      dev->n_other_naks, dev->n_timeouts,
		      dev->n_rdma_dup_busy, dev->n_rc_stalls, dev->n_piowait,
		      dev->n_no_piobuf, dev->n_pkt_drops, dev->n_wqe_errs);
	for (i = 0; i < ARRAY_SIZE(dev->opstats); i++) {
		const struct ipath_opcode_stats *si = &dev->opstats[i];

		if (!si->n_packets && !si->n_bytes)
			continue;
		len += sprintf(buf + len, "%02x %llu/%llu\n", i,
			       (unsigned long long) si->n_packets,
			       (unsigned long long) si->n_bytes);
	}
	return len;
}

static CLASS_DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
static CLASS_DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
static CLASS_DEVICE_ATTR(board_id, S_IRUGO, show_hca, NULL);
static CLASS_DEVICE_ATTR(stats, S_IRUGO, show_stats, NULL);

static struct class_device_attribute *ipath_class_attributes[] = {
	&class_device_attr_hw_rev,
	&class_device_attr_hca_type,
	&class_device_attr_board_id,
	&class_device_attr_stats
};

static int ipath_verbs_register_sysfs(struct ib_device *dev)
{
	int i;
	int ret;

	for (i = 0; i < ARRAY_SIZE(ipath_class_attributes); ++i)
		if (class_device_create_file(&dev->class_dev,
					     ipath_class_attributes[i])) {
			ret = 1;
			goto bail;
		}

	ret = 0;

bail:
	return ret;
}