libibverbs/ibdev_nl.c

/*
 * Copyright (c) 2019, Mellanox Technologies. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *      Redistribution and use in source and binary forms, with or
 *      without modification, are permitted provided that the following
 *      conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */
#include <util/rdma_nl.h>

#include <dirent.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/sysmacros.h>

#include <ccan/list.h>
#include <util/util.h>
#include <infiniband/driver.h>

#include "ibverbs.h"

/* Determine the name of the uverbsX class for the sysfs_dev using sysfs. */
static int find_uverbs_sysfs(struct verbs_sysfs_dev *sysfs_dev)
{
	char path[IBV_SYSFS_PATH_MAX];
	struct dirent *dent;
	DIR *class_dir;
	int ret = ENOENT;

	if (!check_snprintf(path, sizeof(path), "%s/device/infiniband_verbs",
			    sysfs_dev->ibdev_path))
		return ENOMEM;

	class_dir = opendir(path);
	if (!class_dir)
		return ENOSYS;

	while ((dent = readdir(class_dir))) {
		int uv_dirfd;
		bool failed;

		if (dent->d_name[0] == '.')
			continue;

		uv_dirfd = openat(dirfd(class_dir), dent->d_name,
				  O_RDONLY | O_DIRECTORY | O_CLOEXEC);
		if (uv_dirfd == -1)
			break;
		failed = setup_sysfs_uverbs(uv_dirfd, dent->d_name, sysfs_dev);
		close(uv_dirfd);
		if (!failed)
			ret = 0;
		break;
	}
	closedir(class_dir);
	return ret;
}

static int find_uverbs_nl_cb(struct nl_msg *msg, void *data)
{
	struct verbs_sysfs_dev *sysfs_dev = data;
	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
	uint64_t cdev64;
	int ret;

	ret = nlmsg_parse(nlmsg_hdr(msg), 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
			  rdmanl_policy);
	if (ret < 0)
		return ret;
	if (!tb[RDMA_NLDEV_ATTR_CHARDEV] || !tb[RDMA_NLDEV_ATTR_CHARDEV_ABI] ||
	    !tb[RDMA_NLDEV_ATTR_CHARDEV_NAME])
		return NLE_PARSE_ERR;

	/*
	 * The global uverbs abi is 6 for the request string 'uverbs'. We
	 * don't expect to ever have to change the ABI version for uverbs
	 * again.
	 */
	abi_ver = 6;

	/*
	 * The top 32 bits of CHARDEV_ABI are reserved for a future use,
	 * current kernels set them to 0
	 */
	sysfs_dev->abi_ver = nla_get_u64(tb[RDMA_NLDEV_ATTR_CHARDEV_ABI]);
	if (tb[RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID])
		sysfs_dev->driver_id =
			nla_get_u32(tb[RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID]);
	else
		sysfs_dev->driver_id = RDMA_DRIVER_UNKNOWN;

	/* Convert from huge_encode_dev to whatever glibc uses */
	cdev64 = nla_get_u64(tb[RDMA_NLDEV_ATTR_CHARDEV]);
	sysfs_dev->sysfs_cdev =
		makedev((cdev64 & 0xfff00) >> 8,
			(cdev64 & 0xff) | ((cdev64 >> 12) & 0xfff00));

	if (!check_snprintf(sysfs_dev->sysfs_name,
			    sizeof(sysfs_dev->sysfs_name), "%s",
			    nla_get_string(tb[RDMA_NLDEV_ATTR_CHARDEV_NAME])))
		return NLE_PARSE_ERR;
	return 0;
}

/* Ask the kernel for the uverbs char device information */
static int find_uverbs_nl(struct nl_sock *nl, struct verbs_sysfs_dev *sysfs_dev)
{
	if (rdmanl_get_chardev(nl, sysfs_dev->ibdev_idx, "uverbs",
				  find_uverbs_nl_cb, sysfs_dev))
		return -1;
	if (!sysfs_dev->sysfs_name[0])
		return -1;
	return 0;
}

static int find_sysfs_devs_nl_cb(struct nl_msg *msg, void *data)
{
	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
	struct list_head *sysfs_list = data;
	struct verbs_sysfs_dev *sysfs_dev;
	int ret;

	ret = nlmsg_parse(nlmsg_hdr(msg), 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
			  rdmanl_policy);
	if (ret < 0)
		return ret;
	if (!tb[RDMA_NLDEV_ATTR_DEV_NAME] ||
	    !tb[RDMA_NLDEV_ATTR_DEV_NODE_TYPE] ||
	    !tb[RDMA_NLDEV_ATTR_DEV_INDEX] ||
	    !tb[RDMA_NLDEV_ATTR_NODE_GUID] ||
	    !tb[RDMA_NLDEV_ATTR_PORT_INDEX])
		return NLE_PARSE_ERR;

	sysfs_dev = calloc(1, sizeof(*sysfs_dev));
	if (!sysfs_dev)
		return NLE_NOMEM;

	sysfs_dev->ibdev_idx = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
	sysfs_dev->num_ports = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]);
	sysfs_dev->node_guid = nla_get_u64(tb[RDMA_NLDEV_ATTR_NODE_GUID]);
	sysfs_dev->flags |= VSYSFS_READ_NODE_GUID;
	if (!check_snprintf(sysfs_dev->ibdev_name,
			    sizeof(sysfs_dev->ibdev_name), "%s",
			    nla_get_string(tb[RDMA_NLDEV_ATTR_DEV_NAME])))
		goto err;
	if (!check_snprintf(
		    sysfs_dev->ibdev_path, sizeof(sysfs_dev->ibdev_path),
		    "%s/class/infiniband/%s", ibv_get_sysfs_path(),
		    sysfs_dev->ibdev_name))
		goto err;
	sysfs_dev->node_type = decode_knode_type(
		nla_get_u8(tb[RDMA_NLDEV_ATTR_DEV_NODE_TYPE]));

	/*
	 * We don't need to check the cdev as netlink only shows us devices in
	 * this namespace
	 */

	list_add(sysfs_list, &sysfs_dev->entry);
	return NL_OK;

err:
	free(sysfs_dev);
	return NLE_PARSE_ERR;
}

/* Fetch the list of IB devices and uverbs from netlink */
int find_sysfs_devs_nl(struct list_head *tmp_sysfs_dev_list)
{
	struct verbs_sysfs_dev *dev, *dev_tmp;
	struct nl_sock *nl;

	nl = rdmanl_socket_alloc();
	if (!nl)
		return EOPNOTSUPP;

	if (rdmanl_get_devices(nl, find_sysfs_devs_nl_cb, tmp_sysfs_dev_list))
		goto err;

	list_for_each_safe (tmp_sysfs_dev_list, dev, dev_tmp, entry) {
		if ((find_uverbs_nl(nl, dev) && find_uverbs_sysfs(dev)) ||
		    try_access_device(dev)) {
			list_del(&dev->entry);
			free(dev);
		}
	}

	nl_socket_free(nl);
	return 0;

err:
	list_for_each_safe (tmp_sysfs_dev_list, dev, dev_tmp, entry) {
		list_del(&dev->entry);
		free(dev);
	}
	nl_socket_free(nl);
	return EINVAL;
}

static int get_copy_on_fork_cb(struct nl_msg *msg, void *data)
{
	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
	int ret;

	ret = nlmsg_parse(nlmsg_hdr(msg), 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
			  rdmanl_policy);
	if (ret < 0)
		return ret;

	/* Older kernels don't support COF and don't report it through nl */
	if (!tb[RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK]) {
		*(uint8_t *)data = 0;
		return NL_OK;
	}

	*(uint8_t *)data = nla_get_u8(tb[RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK]);
	return NL_OK;
}

bool get_copy_on_fork(void)
{
	struct nl_sock *nl;
	uint8_t cof;

	nl = rdmanl_socket_alloc();
	if (!nl)
		return false;

	if (rdmanl_get_copy_on_fork(nl, get_copy_on_fork_cb, &cof))
		cof = false;

	nl_socket_free(nl);
	return cof;
}