freebsd-nq/contrib/ofed/libibumad/umad.c

/*
 * Copyright (c) 2004-2009 Voltaire Inc.  All rights reserved.
 * Copyright (c) 2014 Intel Corporation.  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 <config.h>

#include <sys/poll.h>
#include <unistd.h>
#include <string.h>
#include <stdio.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <dirent.h>
#include <ctype.h>
#include <inttypes.h>
#include <assert.h>

#include <infiniband/umad.h>

#define IB_OPENIB_OUI                 (0x001405)

#include "sysfs.h"

typedef struct ib_user_mad_reg_req {
	uint32_t id;
	uint32_t method_mask[4];
	uint8_t qpn;
	uint8_t mgmt_class;
	uint8_t mgmt_class_version;
	uint8_t oui[3];
	uint8_t rmpp_version;
} ib_user_mad_reg_req_t;

static_assert(sizeof(struct ib_user_mad_reg_req) == IOCPARM_LEN(IB_USER_MAD_REGISTER_AGENT),
    "Invalid structure size");

struct ib_user_mad_reg_req2 {
	uint32_t id;
	uint32_t qpn;
	uint8_t  mgmt_class;
	uint8_t  mgmt_class_version;
	uint16_t res;
	uint32_t flags;
	uint64_t method_mask[2];
	uint32_t oui;
	uint8_t  rmpp_version;
	uint8_t  reserved[3];
};

static_assert(sizeof(struct ib_user_mad_reg_req2) == IOCPARM_LEN(IB_USER_MAD_REGISTER_AGENT2),
    "Invalid structure size");

#define IBWARN(fmt, args...) fprintf(stderr, "ibwarn: [%d] %s: " fmt "\n", getpid(), __func__, ## args)

#define TRACE	if (umaddebug)	IBWARN
#define DEBUG	if (umaddebug)	IBWARN

static int umaddebug = 0;

#define UMAD_DEV_FILE_SZ	256

static const char *def_ca_name = "mthca0";
static int def_ca_port = 1;

static unsigned abi_version;
static unsigned new_user_mad_api;

/*************************************
 * Port
 */
static int find_cached_ca(const char *ca_name, umad_ca_t * ca)
{
	return 0;		/* caching not implemented yet */
}

static int put_ca(umad_ca_t * ca)
{
	return 0;		/* caching not implemented yet */
}

static int release_port(umad_port_t * port)
{
	free(port->pkeys);
	port->pkeys = NULL;
	port->pkeys_size = 0;
	return 0;
}

static int check_for_digit_name(const struct dirent *dent)
{
	const char *p = dent->d_name;
	while (*p && isdigit(*p))
		p++;
	return *p ? 0 : 1;
}

static int get_port(const char *ca_name, const char *dir, int portnum, umad_port_t * port)
{
	char port_dir[256];
	union umad_gid gid;
	struct dirent **namelist = NULL;
	int i, len, num_pkeys = 0;
	uint32_t capmask;

	strncpy(port->ca_name, ca_name, sizeof port->ca_name - 1);
	port->portnum = portnum;
	port->pkeys = NULL;
	port->rate = 0;

	len = snprintf(port_dir, sizeof(port_dir), "%s/%d", dir, portnum);
	if (len < 0 || len > sizeof(port_dir))
		goto clean;

	if (sys_read_uint(port_dir, SYS_PORT_LMC, &port->lmc) < 0)
		goto clean;
	if (sys_read_uint(port_dir, SYS_PORT_SMLID, &port->sm_lid) < 0)
		goto clean;
	if (sys_read_uint(port_dir, SYS_PORT_SMSL, &port->sm_sl) < 0)
		goto clean;
	if (sys_read_uint(port_dir, SYS_PORT_LID, &port->base_lid) < 0)
		goto clean;
	if (sys_read_uint(port_dir, SYS_PORT_STATE, &port->state) < 0)
		goto clean;
	if (sys_read_uint(port_dir, SYS_PORT_PHY_STATE, &port->phys_state) < 0)
		goto clean;
	/*
	 * If width was not set properly this read may fail.
	 * Instead of failing everything, we will just skip the check
	 * and it will be set to 0.
	 */
	sys_read_uint(port_dir, SYS_PORT_RATE, &port->rate);
	if (sys_read_uint(port_dir, SYS_PORT_CAPMASK, &capmask) < 0)
		goto clean;

	if (sys_read_string(port_dir, SYS_PORT_LINK_LAYER,
	    port->link_layer, UMAD_CA_NAME_LEN) < 0)
		/* assume IB by default */
		sprintf(port->link_layer, "IB");

	port->capmask = htobe32(capmask);

	if (sys_read_gid(port_dir, SYS_PORT_GID, &gid) < 0)
		goto clean;

	port->gid_prefix = gid.global.subnet_prefix;
	port->port_guid = gid.global.interface_id;

	snprintf(port_dir + len, sizeof(port_dir) - len, "/pkeys");
	num_pkeys = sys_scandir(port_dir, &namelist, check_for_digit_name, NULL);
	if (num_pkeys <= 0) {
		IBWARN("no pkeys found for %s:%u (at dir %s)...",
		       port->ca_name, port->portnum, port_dir);
		goto clean;
	}
	port->pkeys = calloc(num_pkeys, sizeof(port->pkeys[0]));
	if (!port->pkeys) {
		IBWARN("get_port: calloc failed: %s", strerror(errno));
		goto clean;
	}
	for (i = 0; i < num_pkeys; i++) {
		unsigned idx, val;
		idx = strtoul(namelist[i]->d_name, NULL, 0);
		sys_read_uint(port_dir, namelist[i]->d_name, &val);
		port->pkeys[idx] = val;
		free(namelist[i]);
	}
	port->pkeys_size = num_pkeys;
	free(namelist);
	namelist = NULL;
	port_dir[len] = '\0';

	/* FIXME: handle gids */

	return 0;

clean:
	if (namelist) {
		for (i = 0; i < num_pkeys; i++)
			free(namelist[i]);
		free(namelist);
	}
	if (port->pkeys)
		free(port->pkeys);
	return -EIO;
}

static int release_ca(umad_ca_t * ca)
{
	int i;

	for (i = 0; i <= ca->numports; i++) {
		if (!ca->ports[i])
			continue;
		release_port(ca->ports[i]);
		free(ca->ports[i]);
		ca->ports[i] = NULL;
	}
	return 0;
}

/*
 * if *port > 0, check ca[port] state. Otherwise set *port to
 * the first port that is active, and if such is not found, to
 * the first port that is link up and if none are linkup, then
 * the first port that is not disabled.  Otherwise return -1.
 */
static int resolve_ca_port(const char *ca_name, int *port)
{
	umad_ca_t ca;
	int active = -1, up = -1;
	int i, ret = 0;

	TRACE("checking ca '%s'", ca_name);

	if (umad_get_ca(ca_name, &ca) < 0)
		return -1;

	if (ca.node_type == 2) {
		*port = 0;	/* switch sma port 0 */
		ret = 1;
		goto Exit;
	}

	if (*port > 0) {	/* check only the port the user wants */
		if (*port > ca.numports) {
			ret = -1;
			goto Exit;
		}
		if (!ca.ports[*port]) {
			ret = -1;
			goto Exit;
		}
		if (strcmp(ca.ports[*port]->link_layer, "InfiniBand") &&
		    strcmp(ca.ports[*port]->link_layer, "IB")) {
			ret = -1;
			goto Exit;
		}
		if (ca.ports[*port]->state == 4) {
			ret = 1;
			goto Exit;
		}
		if (ca.ports[*port]->phys_state != 3)
			goto Exit;
		ret = -1;
		goto Exit;
	}

	for (i = 0; i <= ca.numports; i++) {
		DEBUG("checking port %d", i);
		if (!ca.ports[i])
			continue;
		if (strcmp(ca.ports[i]->link_layer, "InfiniBand") &&
		    strcmp(ca.ports[i]->link_layer, "IB"))
			continue;
		if (up < 0 && ca.ports[i]->phys_state == 5)
			up = *port = i;
		if (ca.ports[i]->state == 4) {
			active = *port = i;
			DEBUG("found active port %d", i);
			break;
		}
	}

	if (active == -1 && up == -1) {	/* no active or linkup port found */
		for (i = 0; i <= ca.numports; i++) {
			DEBUG("checking port %d", i);
			if (!ca.ports[i])
				continue;
			if (ca.ports[i]->phys_state != 3) {
				up = *port = i;
				break;
			}
		}
	}

	if (active >= 0) {
		ret = 1;
		goto Exit;
	}
	if (up >= 0) {
		ret = 0;
		goto Exit;
	}
	ret = -1;
Exit:
	release_ca(&ca);
	return ret;
}

static const char *resolve_ca_name(const char *ca_name, int *best_port)
{
	static char names[UMAD_MAX_DEVICES][UMAD_CA_NAME_LEN];
	int phys_found = -1, port_found = 0, port, port_type;
	int caidx, n;

	if (ca_name && (!best_port || *best_port))
		return ca_name;

	if (ca_name) {
		if (resolve_ca_port(ca_name, best_port) < 0)
			return NULL;
		return ca_name;
	}

	/* Get the list of CA names */
	if ((n = umad_get_cas_names((void *)names, UMAD_MAX_DEVICES)) < 0)
		return NULL;

	/* Find the first existing CA with an active port */
	for (caidx = 0; caidx < n; caidx++) {
		TRACE("checking ca '%s'", names[caidx]);

		port = best_port ? *best_port : 0;
		if ((port_type = resolve_ca_port(names[caidx], &port)) < 0)
			continue;

		DEBUG("found ca %s with port %d type %d",
		      names[caidx], port, port_type);

		if (port_type > 0) {
			if (best_port)
				*best_port = port;
			DEBUG("found ca %s with active port %d",
			      names[caidx], port);
			return (char *)(names + caidx);
		}

		if (phys_found == -1) {
			phys_found = caidx;
			port_found = port;
		}
	}

	DEBUG("phys found %d on %s port %d",
	      phys_found, phys_found >= 0 ? names[phys_found] : NULL,
	      port_found);
	if (phys_found >= 0) {
		if (best_port)
			*best_port = port_found;
		return names[phys_found];
	}

	if (best_port)
		*best_port = def_ca_port;
	return def_ca_name;
}

static int get_ca(const char *ca_name, umad_ca_t * ca)
{
	char dir_name[256];
	struct dirent **namelist;
	int r, i, ret;
	int portnum;

	ca->numports = 0;
	memset(ca->ports, 0, sizeof ca->ports);
	strncpy(ca->ca_name, ca_name, sizeof(ca->ca_name) - 1);

	snprintf(dir_name, sizeof(dir_name), "%s/%s", SYS_INFINIBAND,
		 ca->ca_name);

	if ((r = sys_read_uint(dir_name, SYS_NODE_TYPE, &ca->node_type)) < 0)
		return r;
	if (sys_read_string(dir_name, SYS_CA_FW_VERS, ca->fw_ver,
			    sizeof ca->fw_ver) < 0)
		ca->fw_ver[0] = '\0';
	if (sys_read_string(dir_name, SYS_CA_HW_VERS, ca->hw_ver,
			    sizeof ca->hw_ver) < 0)
		ca->hw_ver[0] = '\0';
	if ((r = sys_read_string(dir_name, SYS_CA_TYPE, ca->ca_type,
				 sizeof ca->ca_type)) < 0)
		ca->ca_type[0] = '\0';
	if ((r = sys_read_guid(dir_name, SYS_CA_NODE_GUID, &ca->node_guid)) < 0)
		return r;
	if ((r =
	     sys_read_guid(dir_name, SYS_CA_SYS_GUID, &ca->system_guid)) < 0)
		return r;

	snprintf(dir_name, sizeof(dir_name), "%s/%s/%s",
		 SYS_INFINIBAND, ca->ca_name, SYS_CA_PORTS_DIR);

	if ((r = sys_scandir(dir_name, &namelist, NULL, alphasort)) < 0) {
		ret = errno < 0 ? errno : -EIO;
		goto error;
	}

	ret = 0;
	for (i = 0; i < r; i++) {
		portnum = 0;
		if (!strcmp(".", namelist[i]->d_name) ||
		    !strcmp("..", namelist[i]->d_name))
			continue;
		if (strcmp("0", namelist[i]->d_name) &&
		    ((portnum = atoi(namelist[i]->d_name)) <= 0 ||
		     portnum >= UMAD_CA_MAX_PORTS)) {
			ret = -EIO;
			goto clean;
		}
		if (!(ca->ports[portnum] =
		      calloc(1, sizeof(*ca->ports[portnum])))) {
			ret = -ENOMEM;
			goto clean;
		}
		if (get_port(ca_name, dir_name, portnum, ca->ports[portnum]) <
		    0) {
			free(ca->ports[portnum]);
			ca->ports[portnum] = NULL;
			ret = -EIO;
			goto clean;
		}
		if (ca->numports < portnum)
			ca->numports = portnum;
	}

	for (i = 0; i < r; i++)
		free(namelist[i]);
	free(namelist);

	put_ca(ca);
	return 0;

clean:
	for (i = 0; i < r; i++)
		free(namelist[i]);
	free(namelist);
error:
	release_ca(ca);

	return ret;
}

static int umad_id_to_dev(int umad_id, char *dev, unsigned *port)
{
	char path[256];
	int r;

	snprintf(path, sizeof(path), SYS_INFINIBAND_MAD "/umad%d", umad_id);

	if ((r =
	     sys_read_string(path, SYS_IB_MAD_DEV, dev, UMAD_CA_NAME_LEN)) < 0)
		return r;

	if ((r = sys_read_uint(path, SYS_IB_MAD_PORT, port)) < 0)
		return r;

	return 0;
}

static int dev_to_umad_id(const char *dev, unsigned port)
{
	char umad_dev[UMAD_CA_NAME_LEN];
	unsigned umad_port;
	int id;

	for (id = 0; id < UMAD_MAX_PORTS; id++) {
		if (umad_id_to_dev(id, umad_dev, &umad_port) < 0)
			continue;
		if (strncmp(dev, umad_dev, UMAD_CA_NAME_LEN))
			continue;
		if (port != umad_port)
			continue;

		DEBUG("mapped %s %d to %d", dev, port, id);
		return id;
	}

	return -1;		/* not found */
}

/*******************************
 * Public interface
 */

int umad_init(void)
{
	TRACE("umad_init");
	if (sys_read_uint(IB_UMAD_ABI_DIR, IB_UMAD_ABI_FILE, &abi_version) < 0) {
		IBWARN
		    ("can't read ABI version from %s/%s (%m): is ib_umad module loaded?",
		     IB_UMAD_ABI_DIR, IB_UMAD_ABI_FILE);
		return -1;
	}
	if (abi_version < IB_UMAD_ABI_VERSION) {
		IBWARN
		    ("wrong ABI version: %s/%s is %d but library minimal ABI is %d",
		     IB_UMAD_ABI_DIR, IB_UMAD_ABI_FILE, abi_version,
		     IB_UMAD_ABI_VERSION);
		return -1;
	}
	return 0;
}

int umad_done(void)
{
	TRACE("umad_done");
	/* FIXME - verify that all ports are closed */
	return 0;
}

static unsigned is_ib_type(const char *ca_name)
{
	char dir_name[256];
	unsigned type;

	snprintf(dir_name, sizeof(dir_name), "%s/%s", SYS_INFINIBAND, ca_name);

	if (sys_read_uint(dir_name, SYS_NODE_TYPE, &type) < 0)
		return 0;

	return type >= 1 && type <= 3 ? 1 : 0;
}

int umad_get_cas_names(char cas[][UMAD_CA_NAME_LEN], int max)
{
	struct dirent **namelist;
	int n, i, j = 0;

	TRACE("max %d", max);

	n = sys_scandir(SYS_INFINIBAND, &namelist, NULL, alphasort);
	if (n > 0) {
		for (i = 0; i < n; i++) {
			if (strcmp(namelist[i]->d_name, ".") &&
			    strcmp(namelist[i]->d_name, "..")) {
				if (j < max && is_ib_type(namelist[i]->d_name))
					strncpy(cas[j++], namelist[i]->d_name,
						UMAD_CA_NAME_LEN);
			}
			free(namelist[i]);
		}
		DEBUG("return %d cas", j);
	} else {
		/* Is this still needed ? */
		strncpy((char *)cas, def_ca_name, UMAD_CA_NAME_LEN);
		DEBUG("return 1 ca");
		j = 1;
	}
	if (n >= 0)
		free(namelist);
	return j;
}

int umad_get_ca_portguids(const char *ca_name, __be64 *portguids, int max)
{
	umad_ca_t ca;
	int ports = 0, i;

	TRACE("ca name %s max port guids %d", ca_name, max);
	if (!(ca_name = resolve_ca_name(ca_name, NULL)))
		return -ENODEV;

	if (umad_get_ca(ca_name, &ca) < 0)
		return -1;

	if (portguids) {
		if (ca.numports + 1 > max) {
			release_ca(&ca);
			return -ENOMEM;
		}

		for (i = 0; i <= ca.numports; i++)
			portguids[ports++] = ca.ports[i] ?
				ca.ports[i]->port_guid : htobe64(0);
	}

	release_ca(&ca);
	DEBUG("%s: %d ports", ca_name, ports);

	return ports;
}

int umad_get_issm_path(const char *ca_name, int portnum, char path[], int max)
{
	int umad_id;

	TRACE("ca %s port %d", ca_name, portnum);

	if (!(ca_name = resolve_ca_name(ca_name, &portnum)))
		return -ENODEV;

	if ((umad_id = dev_to_umad_id(ca_name, portnum)) < 0)
		return -EINVAL;

	snprintf(path, max, "%s/issm%u", UMAD_DEV_DIR, umad_id);

	return 0;
}

int umad_open_port(const char *ca_name, int portnum)
{
	char dev_file[UMAD_DEV_FILE_SZ];
	int umad_id, fd;

	TRACE("ca %s port %d", ca_name, portnum);

	if (!(ca_name = resolve_ca_name(ca_name, &portnum)))
		return -ENODEV;

	DEBUG("opening %s port %d", ca_name, portnum);

	if ((umad_id = dev_to_umad_id(ca_name, portnum)) < 0)
		return -EINVAL;

	snprintf(dev_file, sizeof(dev_file), "%s/umad%d",
		 UMAD_DEV_DIR, umad_id);

	if ((fd = open(dev_file, O_RDWR | O_NONBLOCK)) < 0) {
		DEBUG("open %s failed: %s", dev_file, strerror(errno));
		return -EIO;
	}

	if (abi_version > 5 || !ioctl(fd, IB_USER_MAD_ENABLE_PKEY, NULL))
		new_user_mad_api = 1;
	else
		new_user_mad_api = 0;

	DEBUG("opened %s fd %d portid %d", dev_file, fd, umad_id);
	return fd;
}

int umad_get_ca(const char *ca_name, umad_ca_t * ca)
{
	int r;

	TRACE("ca_name %s", ca_name);
	if (!(ca_name = resolve_ca_name(ca_name, NULL)))
		return -ENODEV;

	if (find_cached_ca(ca_name, ca) > 0)
		return 0;

	if ((r = get_ca(ca_name, ca)) < 0)
		return r;

	DEBUG("opened %s", ca_name);
	return 0;
}

int umad_release_ca(umad_ca_t * ca)
{
	int r;

	TRACE("ca_name %s", ca->ca_name);
	if (!ca)
		return -ENODEV;

	if ((r = release_ca(ca)) < 0)
		return r;

	DEBUG("releasing %s", ca->ca_name);
	return 0;
}

int umad_get_port(const char *ca_name, int portnum, umad_port_t * port)
{
	char dir_name[256];

	TRACE("ca_name %s portnum %d", ca_name, portnum);

	if (!(ca_name = resolve_ca_name(ca_name, &portnum)))
		return -ENODEV;

	snprintf(dir_name, sizeof(dir_name), "%s/%s/%s",
		 SYS_INFINIBAND, ca_name, SYS_CA_PORTS_DIR);

	return get_port(ca_name, dir_name, portnum, port);
}

int umad_release_port(umad_port_t * port)
{
	int r;

	TRACE("port %s:%d", port->ca_name, port->portnum);
	if (!port)
		return -ENODEV;

	if ((r = release_port(port)) < 0)
		return r;

	DEBUG("releasing %s:%d", port->ca_name, port->portnum);
	return 0;
}

int umad_close_port(int fd)
{
	close(fd);
	DEBUG("closed fd %d", fd);
	return 0;
}

void *umad_get_mad(void *umad)
{
	return new_user_mad_api ? ((struct ib_user_mad *)umad)->data :
	    (void *)&((struct ib_user_mad *)umad)->addr.pkey_index;
}

size_t umad_size(void)
{
	return new_user_mad_api ? sizeof(struct ib_user_mad) :
	    sizeof(struct ib_user_mad) - 8;
}

int umad_set_grh(void *umad, void *mad_addr)
{
	struct ib_user_mad *mad = umad;
	struct ib_mad_addr *addr = mad_addr;

	if (mad_addr) {
		mad->addr.grh_present = 1;
		mad->addr.ib_gid = addr->ib_gid;
		/* The definition for umad_set_grh requires that the input be
		 * in host order */
		mad->addr.flow_label = htobe32((uint32_t)addr->flow_label);
		mad->addr.hop_limit = addr->hop_limit;
		mad->addr.traffic_class = addr->traffic_class;
	} else
		mad->addr.grh_present = 0;
	return 0;
}

int umad_set_pkey(void *umad, int pkey_index)
{
	struct ib_user_mad *mad = umad;

	if (new_user_mad_api)
		mad->addr.pkey_index = pkey_index;

	return 0;
}

int umad_get_pkey(void *umad)
{
	struct ib_user_mad *mad = umad;

	if (new_user_mad_api)
		return mad->addr.pkey_index;

	return 0;
}

int umad_set_addr(void *umad, int dlid, int dqp, int sl, int qkey)
{
	struct ib_user_mad *mad = umad;

	TRACE("umad %p dlid %u dqp %d sl %d, qkey %x",
	      umad, dlid, dqp, sl, qkey);
	mad->addr.qpn = htobe32(dqp);
	mad->addr.lid = htobe16(dlid);
	mad->addr.qkey = htobe32(qkey);
	mad->addr.sl = sl;

	return 0;
}

int umad_set_addr_net(void *umad, __be16 dlid, __be32 dqp, int sl, __be32 qkey)
{
	struct ib_user_mad *mad = umad;

	TRACE("umad %p dlid %u dqp %d sl %d qkey %x",
	      umad, be16toh(dlid), be32toh(dqp), sl, be32toh(qkey));
	mad->addr.qpn = dqp;
	mad->addr.lid = dlid;
	mad->addr.qkey = qkey;
	mad->addr.sl = sl;

	return 0;
}

int umad_send(int fd, int agentid, void *umad, int length,
	      int timeout_ms, int retries)
{
	struct ib_user_mad *mad = umad;
	int n;

	TRACE("fd %d agentid %d umad %p timeout %u",
	      fd, agentid, umad, timeout_ms);
	errno = 0;

	mad->timeout_ms = timeout_ms;
	mad->retries = retries;
	mad->agent_id = agentid;

	if (umaddebug > 1)
		umad_dump(mad);

	n = write(fd, mad, length + umad_size());
	if (n == length + umad_size())
		return 0;

	DEBUG("write returned %d != sizeof umad %zu + length %d (%m)",
	      n, umad_size(), length);
	if (!errno)
		errno = EIO;
	return -EIO;
}

static int dev_poll(int fd, int timeout_ms)
{
	struct pollfd ufds;
	int n;

	ufds.fd = fd;
	ufds.events = POLLIN;

	if ((n = poll(&ufds, 1, timeout_ms)) == 1)
		return 0;

	if (n == 0)
		return -ETIMEDOUT;

	return -EIO;
}

int umad_recv(int fd, void *umad, int *length, int timeout_ms)
{
	struct ib_user_mad *mad = umad;
	int n;

	errno = 0;
	TRACE("fd %d umad %p timeout %u", fd, umad, timeout_ms);

	if (!umad || !length) {
		errno = EINVAL;
		return -EINVAL;
	}

	if (timeout_ms && (n = dev_poll(fd, timeout_ms)) < 0) {
		if (!errno)
			errno = -n;
		return n;
	}

	n = read(fd, umad, umad_size() + *length);

	VALGRIND_MAKE_MEM_DEFINED(umad, umad_size() + *length);

	if ((n >= 0) && (n <= umad_size() + *length)) {
		DEBUG("mad received by agent %d length %d", mad->agent_id, n);
		if (n > umad_size())
			*length = n - umad_size();
		else
			*length = 0;
		return mad->agent_id;
	}

	if (n == -EWOULDBLOCK) {
		if (!errno)
			errno = EWOULDBLOCK;
		return n;
	}

	DEBUG("read returned %zu > sizeof umad %zu + length %d (%m)",
	      mad->length - umad_size(), umad_size(), *length);

	*length = mad->length - umad_size();
	if (!errno)
		errno = EIO;
	return -errno;
}

int umad_poll(int fd, int timeout_ms)
{
	TRACE("fd %d timeout %u", fd, timeout_ms);
	return dev_poll(fd, timeout_ms);
}

int umad_get_fd(int fd)
{
	TRACE("fd %d", fd);
	return fd;
}

int umad_register_oui(int fd, int mgmt_class, uint8_t rmpp_version,
		      uint8_t oui[3], long method_mask[])
{
	struct ib_user_mad_reg_req req;

	TRACE("fd %d mgmt_class %u rmpp_version %d oui 0x%x%x%x method_mask %p",
	      fd, mgmt_class, (int)rmpp_version, (int)oui[0], (int)oui[1],
	      (int)oui[2], method_mask);

	if (mgmt_class < 0x30 || mgmt_class > 0x4f) {
		DEBUG("mgmt class %d not in vendor range 2", mgmt_class);
		return -EINVAL;
	}

	req.qpn = 1;
	req.mgmt_class = mgmt_class;
	req.mgmt_class_version = 1;
	memcpy(req.oui, oui, sizeof req.oui);
	req.rmpp_version = rmpp_version;

	if (method_mask)
		memcpy(req.method_mask, method_mask, sizeof req.method_mask);
	else
		memset(req.method_mask, 0, sizeof req.method_mask);

	VALGRIND_MAKE_MEM_DEFINED(&req, sizeof req);

	if (!ioctl(fd, IB_USER_MAD_REGISTER_AGENT, (void *)&req)) {
		DEBUG
		    ("fd %d registered to use agent %d qp %d class 0x%x oui %p",
		     fd, req.id, req.qpn, req.mgmt_class, oui);
		return req.id;	/* return agentid */
	}

	DEBUG("fd %d registering qp %d class 0x%x version %d oui %p failed: %m",
	      fd, req.qpn, req.mgmt_class, req.mgmt_class_version, oui);
	return -EPERM;
}

int umad_register(int fd, int mgmt_class, int mgmt_version,
		  uint8_t rmpp_version, long method_mask[])
{
	struct ib_user_mad_reg_req req;
	__be32 oui = htobe32(IB_OPENIB_OUI);
	int qp;

	TRACE
	    ("fd %d mgmt_class %u mgmt_version %u rmpp_version %d method_mask %p",
	     fd, mgmt_class, mgmt_version, rmpp_version, method_mask);

	req.qpn = qp = (mgmt_class == 0x1 || mgmt_class == 0x81) ? 0 : 1;
	req.mgmt_class = mgmt_class;
	req.mgmt_class_version = mgmt_version;
	req.rmpp_version = rmpp_version;

	if (method_mask)
		memcpy(req.method_mask, method_mask, sizeof req.method_mask);
	else
		memset(req.method_mask, 0, sizeof req.method_mask);

	memcpy(&req.oui, (char *)&oui + 1, sizeof req.oui);

	VALGRIND_MAKE_MEM_DEFINED(&req, sizeof req);

	if (!ioctl(fd, IB_USER_MAD_REGISTER_AGENT, (void *)&req)) {
		DEBUG("fd %d registered to use agent %d qp %d", fd, req.id, qp);
		return req.id;	/* return agentid */
	}

	DEBUG("fd %d registering qp %d class 0x%x version %d failed: %m",
	      fd, qp, mgmt_class, mgmt_version);
	return -EPERM;
}

int umad_register2(int port_fd, struct umad_reg_attr *attr, uint32_t *agent_id)
{
	struct ib_user_mad_reg_req2 req;
	int rc;

	if (!attr || !agent_id)
		return EINVAL;

	TRACE("fd %d mgmt_class %u mgmt_class_version %u flags 0x%08x "
	      "method_mask 0x%016" PRIx64 " %016" PRIx64
	      "oui 0x%06x rmpp_version %u ",
	      port_fd, attr->mgmt_class, attr->mgmt_class_version,
	      attr->flags, attr->method_mask[0], attr->method_mask[1],
	      attr->oui, attr->rmpp_version);

	if (attr->mgmt_class >= 0x30 && attr->mgmt_class <= 0x4f &&
	    ((attr->oui & 0x00ffffff) == 0 || (attr->oui & 0xff000000) != 0)) {
		DEBUG("mgmt class %d is in vendor range 2 but oui (0x%08x) is invalid",
		      attr->mgmt_class, attr->oui);
		return EINVAL;
	}

	memset(&req, 0, sizeof(req));

	req.mgmt_class = attr->mgmt_class;
	req.mgmt_class_version = attr->mgmt_class_version;
	req.qpn = (attr->mgmt_class == 0x1 || attr->mgmt_class == 0x81) ? 0 : 1;
	req.flags = attr->flags;
	memcpy(req.method_mask, attr->method_mask, sizeof req.method_mask);
	req.oui = attr->oui;
	req.rmpp_version = attr->rmpp_version;

	VALGRIND_MAKE_MEM_DEFINED(&req, sizeof req);

	if ((rc = ioctl(port_fd, IB_USER_MAD_REGISTER_AGENT2, (void *)&req)) == 0) {
		DEBUG("fd %d registered to use agent %d qp %d class 0x%x oui 0x%06x",
		      port_fd, req.id, req.qpn, req.mgmt_class, attr->oui);
		*agent_id = req.id;
		return 0;
	}

	if (errno == ENOTTY || errno == EINVAL) {

		TRACE("no kernel support for registration flags");
		req.flags = 0;

		if (attr->flags == 0) {
			struct ib_user_mad_reg_req req_v1;

			TRACE("attempting original register ioctl");

			memset(&req_v1, 0, sizeof(req_v1));
			req_v1.mgmt_class = req.mgmt_class;
			req_v1.mgmt_class_version = req.mgmt_class_version;
			req_v1.qpn = req.qpn;
			req_v1.rmpp_version = req.rmpp_version;
			req_v1.oui[0] = (req.oui & 0xff0000) >> 16;
			req_v1.oui[1] = (req.oui & 0x00ff00) >> 8;
			req_v1.oui[2] =  req.oui & 0x0000ff;

			memcpy(req_v1.method_mask, req.method_mask, sizeof req_v1.method_mask);

			if ((rc = ioctl(port_fd, IB_USER_MAD_REGISTER_AGENT,
					(void *)&req_v1)) == 0) {
				DEBUG("fd %d registered to use agent %d qp %d class 0x%x oui 0x%06x",
				      port_fd, req_v1.id, req_v1.qpn, req_v1.mgmt_class, attr->oui);
				*agent_id = req_v1.id;
				return 0;
			}
		}
	}

	rc = errno;
	attr->flags = req.flags;

	DEBUG("fd %d registering qp %d class 0x%x version %d "
	      "oui 0x%06x failed flags returned 0x%x : %m",
	      port_fd, req.qpn, req.mgmt_class, req.mgmt_class_version,
	      attr->oui, req.flags);

	return rc;
}

int umad_unregister(int fd, int agentid)
{
	TRACE("fd %d unregistering agent %d", fd, agentid);
	return ioctl(fd, IB_USER_MAD_UNREGISTER_AGENT, &agentid);
}

int umad_status(void *umad)
{
	struct ib_user_mad *mad = umad;

	return mad->status;
}

ib_mad_addr_t *umad_get_mad_addr(void *umad)
{
	struct ib_user_mad *mad = umad;

	return &mad->addr;
}

int umad_debug(int level)
{
	if (level >= 0)
		umaddebug = level;
	return umaddebug;
}

void umad_addr_dump(ib_mad_addr_t * addr)
{
#define HEX(x)  ((x) < 10 ? '0' + (x) : 'a' + ((x) -10))
	char gid_str[64];
	int i;

	for (i = 0; i < sizeof addr->gid; i++) {
		gid_str[i * 2] = HEX(addr->gid[i] >> 4);
		gid_str[i * 2 + 1] = HEX(addr->gid[i] & 0xf);
	}
	gid_str[i * 2] = 0;
	IBWARN("qpn %d qkey 0x%x lid %u sl %d\n"
	       "grh_present %d gid_index %d hop_limit %d traffic_class %d flow_label 0x%x pkey_index 0x%x\n"
	       "Gid 0x%s",
	       be32toh(addr->qpn), be32toh(addr->qkey), be16toh(addr->lid), addr->sl,
	       addr->grh_present, (int)addr->gid_index, (int)addr->hop_limit,
	       (int)addr->traffic_class, addr->flow_label, addr->pkey_index,
	       gid_str);
}

void umad_dump(void *umad)
{
	struct ib_user_mad *mad = umad;

	IBWARN("agent id %d status %x timeout %d",
	       mad->agent_id, mad->status, mad->timeout_ms);
	umad_addr_dump(&mad->addr);
}