librpma.h

/* SPDX-License-Identifier: BSD-3-Clause */
/* Copyright 2019-2020, Intel Corporation */

/*
 * librpma.h -- definitions of librpma entry points
 *
 * This library provides low-level support for remote access to persistent
 * memory utilizing RDMA-capable NICs.
 */

#ifndef LIBRPMA_H
#define LIBRPMA_H 1

#include <stddef.h>
#include <stdint.h>
#include <stdio.h>
#include <stdbool.h>

#include <infiniband/verbs.h>

/** 7
 * librpma - remote persistent memory access library
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *	cc ... -lrpma
 *
 * DESCRIPTION
 *
 * librpma is a C library to simplify accessing persistent memory (PMem)
 * on remote hosts over Remote Direct Memory Access (RDMA).
 *
 * The librpma library provides two possible schemes of operation:
 * Remote Memory Access and Messaging. Both of them are available over
 * a connection established between two peers. Both of these schemes can make
 * use of PMem as well as DRAM for the sake of building efficient and scalable
 * Remote Persistent Memory Accessing (RPMA) applications.
 *
 * REMOTE MEMORY ACCESS
 *
 * The librpma library implements four basic API calls
 * dedicated for accessing a remote memory:
 * - rpma_read() - initiates transferring data from the remote memory
 *   to the local memory,
 * - rpma_write() - initiates transferring data from the local memory
 *   to the remote memory),
 * - rpma_write_atomic() - works like rpma_write(), but it allows transferring
 *   8 bytes of data (RPMA_ATOMIC_WRITE_ALIGNMENT) and storing them atomically
 *   in the remote memory (see rpma_write_atomic(3) for details
 *   and restrictions), and:
 * - rpma_flush() - initiates finalizing a transfer of data to the remote
 *   memory. Possible types of rpma_flush() operation:
 *   - RPMA_FLUSH_TYPE_PERSISTENT - flush data down to the persistent domain,
 *   - RPMA_FLUSH_TYPE_VISIBILITY - flush data deep enough to make it visible
 *   on the remote node.
 *
 * All the above functions use the attribute flags to set the completion
 * notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generates the completion only on error
 * - RPMA_F_COMPLETION_ALWAYS - generates the completion regardless of a result
 * of the operation.
 *
 * All of these operations are considered as finished
 * when the respective completion is generated.
 *
 * DIRECT WRITE TO PMEM
 *
 * \f[B]Direct Write to PMem\f[R] is a feature of a platform and
 * its configuration which allows an RDMA-capable network interface to write
 * data to platform's PMem in a persistent way. It may be impossible because
 * of e.g. caching mechanisms existing on the data's way. When \f[B]Direct Write
 * to PMem\f[R] is impossible, operating in the way assuming it is possible may
 * corrupt data on PMem, so this is why \f[B]Direct Write to PMem\f[R] is not
 * enabled by default.
 *
 * On the current Intel platforms, the only thing you have to do in order
 * to enable \f[B]Direct Write to PMem\f[R] is turning off
 * Intel Direct Data I/O (DDIO). Sometimes, you can turn off DDIO either
 * globally for the whole platform or for a specific PCIe Root Port.
 * For details, please see the manual of your platform.
 *
 * When you have a platform which allows \f[B]Direct Write to PMem\f[R],
 * you have to declare this is the case in your peer's configuration. The peer's
 * configuration has to be transferred to all the peers which want to execute
 * rpma_flush() with RPMA_FLUSH_TYPE_PERSISTENT against the platform's PMem and
 * applied to the connection object which safeguards access to PMem.
 *
 * - rpma_peer_cfg_set_direct_write_to_pmem() - declare \f[B]Direct Write
 * to PMem\f[R] support
 * - rpma_peer_cfg_get_descriptor() - get the descriptor of the peer
 * configuration
 * - rpma_peer_cfg_from_descriptor() - create a peer configuration from
 * the descriptor
 * - rpma_conn_apply_remote_peer_cfg() - apply remote peer cfg to the connection
 *
 * For details on how to use these APIs please see
 * https://github.com/pmem/rpma/tree/master/examples/05-flush-to-persistent.
 *
 * CLIENT OPERATION
 * A client is the active side of the process of establishing a connection.
 * A role of the peer during the process of establishing connection
 * does not determine direction of the data flow (neither via
 * Remote Memory Access nor via Messaging). After establishing the connection
 * both peers have the same capabilities.
 *
 * The client, in order to establish a connection, has to perform the following
 * steps:
 *
 * - rpma_conn_req_new() - create a new outgoing connection request object
 * - rpma_conn_req_connect() - initiate processing the connection request
 * - rpma_conn_next_event() - wait for the RPMA_CONN_ESTABLISHED event
 *
 * After establishing the connection both peers can perform
 * Remote Memory Access and/or Messaging over the connection.
 *
 * The client, in order to close a connection, has to perform the following
 * steps:
 *
 * - rpma_conn_disconnect() - initiate disconnection
 * - rpma_conn_next_event() - wait for the RPMA_CONN_CLOSED event
 * - rpma_conn_delete() - delete the closed connection
 *
 * SERVER OPERATION
 * A server is the passive side of the process of establishing a connection.
 * Note that after establishing the connection both peers have
 * the same capabilities.
 *
 * The server, in order to establish a connection, has to perform the following
 * steps:
 *
 * - rpma_ep_listen() - create a listening endpoint
 * - rpma_ep_next_conn_req() - obtain an incoming connection request
 * - rpma_conn_req_connect() - initiate connecting the connection request
 * - rpma_conn_next_event() - wait for the RPMA_CONN_ESTABLISHED event
 *
 * After establishing the connection both peers can perform
 * Remote Memory Access and/or Messaging over the connection.
 *
 * The server, in order to close a connection, has to perform the following
 * steps:
 *
 * - rpma_conn_next_event() - wait for the RPMA_CONN_CLOSED event
 * - rpma_conn_disconnect() - disconnect the connection
 * - rpma_conn_delete() - delete the closed connection
 *
 * When no more incoming connections are expected, the server can stop waiting
 * for them:
 *
 * - rpma_ep_shutdown() - stop listening and delete the endpoint
 *
 * MEMORY MANAGEMENT
 *
 * Every piece of memory (either volatile or persistent) must be registered
 * and its usage must be specified in order to be used in Remote Memory Access
 * or Messaging. This can be done using the following memory management
 * librpma functions:
 * - rpma_mr_reg() which registers a memory region and creates a local memory
 * registration object and
 * - rpma_mr_dereg() which deregisters the memory region and deletes
 * the local memory registration object.
 *
 * A description of the registered memory region sometimes has to be
 * transferred via network to the other side of the connection.
 * In order to do that a network-transferable description
 * of the provided memory region (called 'descriptor') has to be created
 * using rpma_mr_get_descriptor(). On the other side of the connection
 * the received descriptor should be decoded using
 * rpma_mr_remote_from_descriptor(). It creates a remote memory region's
 * structure that allows for Remote Memory Access.
 *
 * MESSAGING
 *
 * The librpma messaging API allows transferring messages
 * (buffers of arbitrary data) between the peers.
 * Transferring messages requires preparing buffers
 * (memory regions) on the remote side to receive the sent data.
 * The received data are written to those dedicated buffers
 * and the sender does not have to have a respective remote
 * memory region object to send a message.
 * The memory buffers used for messaging have to be registered
 * using rpma_mr_reg() prior to rpma_send() or rpma_recv() function call.
 *
 * The librpma library implements the following messaging API:
 * - rpma_send() - initiates the send operation which transfers a message
 *   from the local memory to other side of the connection,
 * - rpma_recv() - initiates the receive operation which prepares a buffer
 *   for a message sent from other side of the connection,
 * - rpma_conn_req_recv() works as rpma_recv(), but it may be used
 *   before the connection is established.
 *
 * All of these operations are considered as finished
 * when the respective completion is generated.
 *
 * COMPLETIONS
 *
 * RDMA operations generate complitions that notify a user
 * that the respective operation has been completed.
 *
 * The following operations are available in librpma:
 * - RPMA_OP_READ - RMA read operation
 * - RPMA_OP_WRITE - RMA write operation
 * - RPMA_OP_FLUSH - RMA flush operation
 * - RPMA_OP_SEND - messaging send operation
 * - RPMA_OP_RECV - messaging receive operation
 *
 * All operations generate completion on error. The operations posted
 * with the \f[B]RPMA_F_COMPLETION_ALWAYS\f[R] flag also generate a completion
 * on success. Completion codes are reused from the libibverbs library,
 * where the IBV_WC_SUCCESS status indicates the successful completion
 * of an operation. Completions are collected in the completion queue (CQ)
 * (see the \f[B]QUEUES, PERFORMANCE AND RESOURCE USE\f[R] section
 * for more details on queues).
 *
 * The librpma library implements the following API for handling completions:
 * - rpma_conn_completion_wait() waits for incoming completions. If it
 * succeeds the completions can be collected using rpma_conn_completion_get().
 * - rpma_conn_completion_get() receives the next available completion
 * of an already posted operation.
 *
 * PEER
 *
 * A peer is an abstraction representing an RDMA-capable device.
 * All other RPMA objects have to be created in the context of a peer.
 * A peer allows one to:
 * - establish connections (Client Operation)
 * - register memory regions (Memory Management)
 * - create endpoints for listening for incoming connections (Server Operation)
 *
 * At the beginning, in order to create a peer, a user has to obtain
 * an RDMA device context by the given IPv4/IPv6 address using
 * rpma_utils_get_ibv_context(). Then a new peer object can be created
 * using rpma_peer_new() and deleted using rpma_peer_delete().
 *
 * SYNCHRONOUS AND ASYNCHRONOUS MODES
 * By default, all endpoints and connections operate in the synchronous mode
 * where:
 *
 * - rpma_ep_next_conn_req(),
 * - rpma_conn_completion_wait() and
 * - rpma_conn_get_next_event()
 *
 * are blocking calls. You can make those API calls non-blocking by modifying
 * the respective file descriptors:
 *
 * - rpma_ep_get_fd() - provides a file descriptor for rpma_ep_next_conn_req()
 * - rpma_conn_get_completion_fd() - provides a file descriptor for
 * rpma_conn_completion_wait()
 * - rpma_conn_get_event_fd() - provides a file descriptor for
 * rpma_conn_get_next_event()
 *
 * When you have a file descriptor, you can make it non-blocking using fcntl(2)
 * as follows:
 *
 *	int ret = fcntl(fd, F_GETFL);
 *	fcntl(fd, F_SETFL, flags | O_NONBLOCK);
 *
 * Such change makes the respective API call non-blocking automatically.
 *
 * The provided file descriptors can also be used for scalable I/O handling like
 * epoll(7).
 *
 * Please see the example showing how to make use of RPMA file descriptors:
 * https://github.com/pmem/rpma/tree/master/examples/06-multiple-connections
 *
 * .SH QUEUES, PERFORMANCE AND RESOURCE USE
 *
 * \f[B]Remote Memory Access\f[R] operations, \f[B]Messaging\f[R] operations
 * and their \f[B]Completions\f[R] consume space in queues allocated
 * in an RDMA-capable network interface (RNIC) hardware for each
 * of the connections.
 * You must be aware of the existence of these queues:
 *
 * - completion queue \f[B](CQ)\f[R] where completions of operations are placed,
 * either when a completion was required by a user (RPMA_F_COMPLETION_ALWAYS)
 * or a completion with an error occurred. All \f[B]Remote Memory Access\f[R]
 * operations and \f[B]Messaging\f[R] operations can consume \f[B]CQ\f[R] space.
 * - send queue \f[B](SQ)\f[R] where all \f[B]Remote Memory Access\f[R]
 * operations and rpma_send() operations are placed before they are executed by
 * RNIC.
 * - receive queue \f[B](RQ)\f[R] where rpma_recv() entries are placed before
 * they are consumed by the rpma_send() coming from another side
 * of the connection.
 *
 * You must assume \f[B]SQ\f[R] and \f[B]RQ\f[R] entries occupy the place
 * in their respective queue till:
 *
 * - a respective operation's completion is generated or
 * - a completion of an operation, which was scheduled later, is generated.
 *
 * You must also be aware that RNIC has limited resources so it is impossible
 * to store a very long set of queues for many possibly existing connections.
 * If all of the queues will not fit into RNIC's resources it will start using
 * the platform's memory for this purpose. In this case, the performance will
 * be degraded because of inevitable cache misses.
 *
 * Because the length of queues has so profound impact on the performance of
 * RPMA application you can configure the length of each of the queues
 * separately for each of the connections:
 *
 * - rpma_conn_cfg_set_cq_size() - set length of \f[B]CQ\f[R]
 * - rpma_conn_cfg_set_sq_size() - set length of \f[B]SQ\f[R]
 * - rpma_conn_cfg_set_rq_size() - set length of \f[B]RQ\f[R]
 *
 * When the connection configuration object is ready it has to be used for
 * either rpma_conn_req_new() or rpma_ep_next_conn_req() for the settings
 * to take effect.
 *
 * THREAD SAFETY
 *
 * Most of the core librpma API calls are thread-safe but there are also very
 * important exceptions mainly related to connection's configuration,
 * establishment and tear-down. Here you can find a complete list of
 * NOT thread-safe API calls:
 *
 * - rpma_conn_apply_remote_peer_cfg()
 * - rpma_conn_cfg_get_cq_size()
 * - rpma_conn_cfg_get_rq_size()
 * - rpma_conn_cfg_get_sq_size()
 * - rpma_conn_cfg_get_timeout()
 * - rpma_conn_cfg_set_cq_size()
 * - rpma_conn_cfg_set_rq_size()
 * - rpma_conn_cfg_set_sq_size()
 * - rpma_conn_cfg_set_timeout()
 * - rpma_conn_delete()
 * - rpma_conn_disconnect()
 * - rpma_conn_get_private_data()
 * - rpma_conn_next_event()
 * - rpma_conn_req_connect()
 * - rpma_conn_req_delete()
 * - rpma_conn_req_new()
 * - rpma_ep_listen()
 * - rpma_ep_next_conn_req()
 * - rpma_ep_shutdown()
 * - rpma_peer_cfg_get_descriptor()
 * - rpma_peer_cfg_get_descriptor_size()
 * - rpma_peer_cfg_get_direct_write_to_pmem()
 * - rpma_peer_cfg_set_direct_write_to_pmem()
 * - rpma_utils_get_ibv_context()
 *
 * Other librpma API calls are thread-safe. However, creating RPMA library
 * resources usually involves dynamic memory allocation and destroying
 * resources usually involves a dynamic memory release. The same resource
 * cannot be destroyed more than once, at any thread, and a resource cannot be
 * used after it was destroyed. It is the user's responsibility to follow those
 * rules and not doing so may result in a segmentation fault or undefined
 * behaviour.
 *
 * .SH ON-DEMAND PAGING SUPPORT
 *
 * On-Demand-Paging (ODP) is a technique that simplifies the memory
 * registration process (for example, applications no longer need to pin down
 * the underlying physical pages of the address space and track the validity
 * of the mappings). On-Demand Paging is available if both the hardware
 * and the kernel support it. The detailed description of ODP can be found here:
 * https://community.mellanox.com/s/article/understanding-on-demand-paging--odp-x
 *
 * State of ODP support can be checked using
 * the rpma_utils_ibv_context_is_odp_capable() function
 * that queries the RDMA device context's capabilities
 * and checks if it supports On-Demand Paging.
 *
 * The librpma library uses ODP automatically if it is supported.
 * ODP support is required to register PMem memory region mapped
 * from File System DAX (FSDAX).
 *
 * DEBUGGING AND ERROR HANDLING
 *
 * If a librpma function may fail, it returns a negative error code.
 * Checking if the returned value is non-negative is the only
 * programmatically available way to verify if the API call succeeded.
 * The exact meaning of all error codes is described in the manual
 * of each function.
 *
 * The librpma library implements the logging API which may give additional
 * information in case of an error and during normal operation as well,
 * according to the current logging threshold levels.
 *
 * The function that will handle all generated log messages can be set
 * using rpma_log_set_function(). The logging function can be either
 * the default logging function (built into the library)
 * or a user-defined, thread-safe, function. The default logging function
 * can write messages to syslog(3) and stderr(3).
 * The logging threshold level can be set or got using
 * rpma_log_set_threshold() or rpma_log_get_threshold() respectively.
 *
 * There is an example of the usage of the logging functions:
 * https://github.com/pmem/rpma/tree/master/examples/log
 *
 * EXAMPLES
 *
 * See https://github.com/pmem/rpma/tree/master/examples for examples of using
 * the librpma API.
 *
 * ACKNOWLEDGEMENTS
 *
 * librpma is built on the top of libibverbs and librdmacm APIs.
 *
 * SEE ALSO
 *
 * https://pmem.io/rpma/
 */

#define RPMA_W_WAIT_FOR_COMPLETION	(1)

#define RPMA_E_UNKNOWN		(-100000) /* Unknown error */
#define RPMA_E_NOSUPP		(-100001) /* Not supported */
#define RPMA_E_PROVIDER		(-100002) /* Provider error occurred */
#define RPMA_E_NOMEM		(-100003) /* Out of memory */
#define RPMA_E_INVAL		(-100004) /* Invalid argument */
#define RPMA_E_NO_COMPLETION	(-100005) /* No next completion available */
#define RPMA_E_NO_EVENT		(-100006) /* No next event available */
#define RPMA_E_AGAIN		(-100007) /* Temporary error */

/* picking up an RDMA-capable device */

#define RPMA_DEFAULT_TIMEOUT_MS 1000

/* pick a type of an ibv_context to lookup for */
enum rpma_util_ibv_context_type {
	RPMA_UTIL_IBV_CONTEXT_LOCAL, /* lookup for a local device */
	RPMA_UTIL_IBV_CONTEXT_REMOTE /* lookup for a remote device */
};

/** 3
 * rpma_utils_get_ibv_context - obtain an RDMA device context by IP address
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct ibv_context;
 *	enum rpma_util_ibv_context_type {
 *		RPMA_UTIL_IBV_CONTEXT_LOCAL,
 *		RPMA_UTIL_IBV_CONTEXT_REMOTE
 *	};
 *
 *	int rpma_utils_get_ibv_context(const char *addr,
 *		enum rpma_util_ibv_context_type type,
 *		struct ibv_context **dev_ptr);
 *
 * DESCRIPTION
 * rpma_utils_get_ibv_context() obtains an RDMA device context
 * by the given IPv4/IPv6 address (either local or remote) using
 * the TCP RDMA port space (RDMA_PS_TCP) - reliable, connection-oriented
 * and message-based QP communication.
 * Possible values of the 'type' argument:
 * - RPMA_UTIL_IBV_CONTEXT_LOCAL - lookup for a device
 * based on the given local address
 * - RPMA_UTIL_IBV_CONTEXT_REMOTE - lookup for a device
 * based on the given remote address
 *
 * RETURN VALUE
 * The rpma_utils_get_ibv_context() function returns 0 on success or a negative
 * error code on failure. rpma_utils_get_ibv_context() does not set *dev_ptr
 * value on failure.
 *
 * ERRORS
 * rpma_utils_get_ibv_context() can fail with the following errors:
 *
 * - RPMA_E_INVAL - addr or dev_ptr is NULL or type is unknown
 * - RPMA_E_NOMEM - out of memory
 * - RPMA_E_PROVIDER - rdma_getaddrinfo(), rdma_create_id(), rdma_bind_addr()
 * or rdma_resolve_addr() failed, the exact cause of the error
 * can be read from the log
 *
 * SEE ALSO
 * rpma_peer_new(3), rpma_utils_ibv_context_is_odp_capable(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_utils_get_ibv_context(const char *addr,
		enum rpma_util_ibv_context_type type,
		struct ibv_context **dev_ptr);

/** 3
 * rpma_utils_ibv_context_is_odp_capable - is On-Demand Paging supported
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct ibv_context;
 *	int rpma_utils_ibv_context_is_odp_capable(struct ibv_context *dev,
 *		int *is_odp_capable);
 *
 * DESCRIPTION
 * rpma_utils_ibv_context_is_odp_capable() queries the RDMA device context's
 * capabilities and check if it supports On-Demand Paging.
 *
 * RETURN VALUE
 * The rpma_utils_ibv_context_is_odp_capable() function returns 0 on success or
 * a negative error code on failure. The *is_odp_capable value on failure is
 * undefined.
 *
 * ERRORS
 * rpma_utils_ibv_context_is_odp_capable() can fail with the following errors:
 *
 * - RPMA_E_INVAL - dev or is_odp_capable is NULL
 * - RPMA_E_PROVIDER - ibv_query_device_ex() failed, the exact cause
 * of the error can be read from the log
 *
 * SEE ALSO
 * rpma_utils_get_ibv_context(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_utils_ibv_context_is_odp_capable(struct ibv_context *dev,
		int *is_odp_capable);

/* peer configuration */

struct rpma_peer_cfg;

/** 3
 * rpma_peer_cfg_new - create a new peer configuration object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer_cfg;
 *	int rpma_peer_cfg_new(struct rpma_peer_cfg **pcfg_ptr);
 *
 * DESCRIPTION
 * rpma_peer_cfg_new() creates a new peer configuration object.
 *
 * RETURN VALUE
 * The rpma_peer_cfg_new() function returns 0 on success or a negative
 * error code on failure. rpm_peer_cfg_new() does not set
 * *pcfg_ptr value on failure.
 *
 * ERRORS
 * rpma_peer_cfg_new() can fail with the following errors:
 *
 * - RPMA_E_INVAL - pcfg_ptr is NULL
 * - RPMA_E_NOMEM - out of memory
 *
 * SEE ALSO
 * rpma_conn_apply_remote_peer_cfg(3), rpma_peer_cfg_delete(3),
 * rpma_peer_cfg_from_descriptor(3), rpma_peer_cfg_get_descriptor(3),
 * rpma_peer_cfg_get_descriptor_size(3),
 * rpma_peer_cfg_get_direct_write_to_pmem(3),
 * rpma_peer_cfg_set_direct_write_to_pmem(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_peer_cfg_new(struct rpma_peer_cfg **pcfg_ptr);

/** 3
 * rpma_peer_cfg_delete - delete the peer configuration object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer_cfg;
 *	int rpma_peer_cfg_delete(struct rpma_peer_cfg **pcfg_ptr);
 *
 * DESCRIPTION
 * rpma_peer_cfg_delete() deletes the peer configuration object.
 *
 * RETURN VALUE
 * The rpma_peer_cfg_delete() function returns 0 on success or a negative
 * error code on failure. rpm_peer_cfg_delete() does not set
 * *pcfg_ptr value to NULL on failure.
 *
 * ERRORS
 * rpma_peer_cfg_delete() can fail with the following error:
 *
 * - RPMA_E_INVAL - pcfg_ptr is NULL
 *
 * SEE ALSO
 * rpma_peer_cfg_new(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_peer_cfg_delete(struct rpma_peer_cfg **pcfg_ptr);

/** 3
 * rpma_peer_cfg_set_direct_write_to_pmem - declare direct write to PMEM support
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer_cfg;
 *	int rpma_peer_cfg_set_direct_write_to_pmem(struct rpma_peer_cfg *pcfg,
 *			bool supported);
 *
 * DESCRIPTION
 * rpma_peer_cfg_set_direct_write_to_pmem() declares the support
 * of the direct write to PMEM.
 *
 * RETURN VALUE
 * The rpma_peer_cfg_set_direct_write_to_pmem() function returns 0 on success
 * or a negative error code on failure.
 *
 * ERRORS
 * rpma_peer_cfg_set_direct_write_to_pmem() can fail with the following error:
 *
 * - RPMA_E_INVAL - pcfg is NULL
 *
 * SEE ALSO
 * rpma_conn_apply_remote_peer_cfg(3), rpma_peer_cfg_get_descriptor(3),
 * rpma_peer_cfg_get_direct_write_to_pmem(3), rpma_peer_cfg_new(3), librpma(7)
 * and https://pmem.io/rpma/
 */
int rpma_peer_cfg_set_direct_write_to_pmem(struct rpma_peer_cfg *pcfg,
		bool supported);

/** 3
 * rpma_peer_cfg_get_direct_write_to_pmem - check direct write to PMEM support
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer_cfg;
 *	int rpma_peer_cfg_get_direct_write_to_pmem(
 *			const struct rpma_peer_cfg *pcfg, bool *supported);
 *
 * DESCRIPTION
 * rpma_peer_cfg_get_direct_write_to_pmem() checks the support
 * of the direct write to PMEM.
 *
 * RETURN VALUE
 * The rpma_peer_cfg_get_direct_write_to_pmem() function returns 0 on success
 * or a negative error code on failure.
 *
 * ERRORS
 * rpma_peer_cfg_get_direct_write_to_pmem() can fail with the following error:
 *
 * - RPMA_E_INVAL - pcfg or supported are NULL
 *
 * SEE ALSO
 * rpma_peer_cfg_from_descriptor(3), rpma_peer_cfg_new(3),
 * rpma_peer_cfg_set_direct_write_to_pmem(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_peer_cfg_get_direct_write_to_pmem(const struct rpma_peer_cfg *pcfg,
		bool *supported);

/** 3
 * rpma_peer_cfg_get_descriptor - get the descriptor of the peer configuration
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer_cfg;
 *	int rpma_peer_cfg_get_descriptor(const struct rpma_peer_cfg *pcfg,
 *			void *desc);
 *
 * DESCRIPTION
 * rpma_peer_cfg_get_descriptor() gets the descriptor of the peer configuration.
 *
 * RETURN VALUE
 * The rpma_peer_cfg_get_descriptor() function returns 0 on success or
 * a negative error code on failure.
 *
 * ERRORS
 * rpma_peer_cfg_get_descriptor() can fail with the following error:
 *
 * - RPMA_E_INVAL - pcfg or desc are NULL
 *
 * SEE ALSO
 * rpma_peer_cfg_from_descriptor(3), rpma_peer_cfg_new(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_peer_cfg_get_descriptor(const struct rpma_peer_cfg *pcfg, void *desc);

/** 3
 * rpma_peer_cfg_get_descriptor_size - get size of the peer cfg descriptor
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer_cfg;
 *	int rpma_peer_cfg_get_descriptor_size(const struct rpma_peer_cfg *pcfg,
 *			size_t *desc_size);
 *
 * DESCRIPTION
 * rpma_peer_cfg_get_descriptor_size() gets size of the peer configuration
 * descriptor.
 *
 * RETURN VALUE
 * The rpma_peer_cfg_get_descriptor_size() function returns 0 on success or
 * a negative error code on failure.
 *
 * ERRORS
 * rpma_peer_cfg_get_descriptor_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - pcfg or desc_size is NULL
 *
 * SEE ALSO
 * rpma_peer_cfg_get_descriptor(3), rpma_peer_cfg_new(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int
rpma_peer_cfg_get_descriptor_size(const struct rpma_peer_cfg *pcfg,
		size_t *desc_size);

/** 3
 * rpma_peer_cfg_from_descriptor - create a peer cfg from the descriptor
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer_cfg;
 *	int rpma_peer_cfg_from_descriptor(const void *desc, size_t desc_size,
 *			struct rpma_peer_cfg **pcfg_ptr);
 *
 * DESCRIPTION
 * rpma_peer_cfg_from_descriptor() creates a peer configuration object
 * from the descriptor.
 *
 * RETURN VALUE
 * The rpma_peer_cfg_from_descriptor() function returns 0 on success
 * or a negative error code on failure. rpma_peer_cfg_from_descriptor()
 * does not set *pcfg_ptr value on failure.
 *
 * ERRORS
 * rpma_peer_cfg_from_descriptor() can fail with the following errors:
 *
 * - RPMA_E_INVAL - desc or pcfg_ptr are NULL
 * - RPMA_E_NOMEM - out of memory
 *
 * SEE ALSO
 * rpma_conn_apply_remote_peer_cfg(3), rpma_peer_cfg_get_descriptor(3),
 * rpma_peer_cfg_new(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_peer_cfg_from_descriptor(const void *desc, size_t desc_size,
		struct rpma_peer_cfg **pcfg_ptr);

/* peer */

struct rpma_peer;

/** 3
 * rpma_peer_new - create a peer object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct ibv_context;
 *	struct rpma_peer;
 *	int rpma_peer_new(struct ibv_context *ibv_ctx,
 *			struct rpma_peer **peer_ptr);
 *
 * DESCRIPTION
 * rpma_peer_new() creates a new peer object.
 *
 * RETURN VALUE
 * The rpma_peer_new() function returns 0 on success or a negative error code
 * on failure. rpma_peer_new() does not set *peer_ptr value on failure.
 *
 * ERRORS
 * rpma_peer_new() can fail with the following errors:
 *
 * - RPMA_E_INVAL - ibv_ctx or peer_ptr is NULL
 * - RPMA_E_NOMEM - creating a verbs protection domain failed with ENOMEM.
 * - RPMA_E_PROVIDER - creating a verbs protection domain failed with error
 *   other than ENOMEM.
 * - RPMA_E_UNKNOWN - creating a verbs protection domain failed without error
 *   value.
 * - RPMA_E_NOMEM - out of memory
 *
 * SEE ALSO
 * rpma_conn_req_new(3), rpma_ep_listen(3), rpma_mr_reg(3), rpma_peer_delete(3),
 * rpma_utils_get_ibv_context(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_peer_new(struct ibv_context *ibv_ctx, struct rpma_peer **peer_ptr);

/** 3
 * rpma_peer_delete - delete a peer object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer;
 *	int rpma_peer_delete(struct rpma_peer **peer_ptr);
 *
 * DESCRIPTION
 * rpma_peer_delete() deletes the peer object.
 *
 * RETURN VALUE
 * The rpma_peer_delete() function returns 0 on success or a negative error
 * code on failure. rpm_peer_delete() does not set *peer_ptr value
 * to NULL on failure.
 *
 * RETURN VALUE
 * The rpma_peer_delete() function returns 0 on success or a negative error code
 * on failure. rpma_peer_delete() does not set *peer_ptr to NULL on failure.
 *
 * ERRORS
 * rpma_peer_delete() can fail with the following error:
 *
 * - RPMA_E_PROVIDER - deleting the verbs protection domain failed.
 *
 * SEE ALSO
 * rpma_peer_new(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_peer_delete(struct rpma_peer **peer_ptr);

/* memory-related structures */

struct rpma_mr_local;
struct rpma_mr_remote;

#define RPMA_MR_USAGE_READ_SRC			(1 << 0)
#define RPMA_MR_USAGE_READ_DST			(1 << 1)
#define RPMA_MR_USAGE_WRITE_SRC			(1 << 2)
#define RPMA_MR_USAGE_WRITE_DST			(1 << 3)
#define RPMA_MR_USAGE_FLUSH_TYPE_VISIBILITY	(1 << 4)
#define RPMA_MR_USAGE_FLUSH_TYPE_PERSISTENT	(1 << 5)
#define RPMA_MR_USAGE_SEND			(1 << 6)
#define RPMA_MR_USAGE_RECV			(1 << 7)

/** 3
 * rpma_mr_reg - create a local memory registration object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer;
 *	struct rpma_mr_local;
 *
 *	int rpma_mr_reg(struct rpma_peer *peer, void *ptr, size_t size,
 *		int usage, struct rpma_mr_local **mr_ptr);
 *
 * DESCRIPTION
 * rpma_mr_reg() registers a memory region and creates a local memory
 * registration object. The usage parameter specifies the operations
 * that can be performed on the given memory region which should be expressed
 * as bitwise-inclusive OR of the following:
 * - RPMA_MR_USAGE_READ_SRC - memory used as a source of the read operation
 * - RPMA_MR_USAGE_READ_DST - memory used as a destination of the read operation
 * - RPMA_MR_USAGE_WRITE_SRC - memory used as a source of the write operation
 * - RPMA_MR_USAGE_WRITE_DST - memory used as a destination of
 * the write operation
 * - RPMA_MR_USAGE_FLUSH_TYPE_VISIBILITY - memory with available flush operation
 * - RPMA_MR_USAGE_FLUSH_TYPE_PERSISTENT - memory with available
 * persistent flush operation
 * - RPMA_MR_USAGE_SEND - memory used for send operation
 * - RPMA_MR_USAGE_RECV - memory used for receive operation
 *
 * RETURN VALUE
 * The rpma_mr_reg() function returns 0 on success or a negative error code
 * on failure. rpma_mr_reg() does not set *mr_ptr value on failure.
 *
 * ERRORS
 * rpma_mr_reg() can fail with the following errors:
 *
 * - RPMA_E_INVAL - peer or ptr or mr_ptr is NULL
 * - RPMA_E_INVAL - size equals 0
 * - RPMA_E_NOMEM - out of memory
 * - RPMA_E_PROVIDER - memory registration failed
 *
 * SEE ALSO
 * rpma_conn_req_recv(3), rpma_mr_dereg(3), rpma_mr_get_descriptor(3),
 * rpma_mr_get_descriptor_size(3), rpma_peer_new(3), rpma_read(3), rpma_recv(3),
 * rpma_send(3), rpma_write(3), rpma_write_atomic(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_mr_reg(struct rpma_peer *peer, void *ptr, size_t size,
		int usage, struct rpma_mr_local **mr_ptr);

/** 3
 * rpma_mr_dereg - delete a local memory registration object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_mr_local;
 *	int rpma_mr_dereg(struct rpma_mr_local **mr_ptr);
 *
 * DESCRIPTION
 * rpma_mr_dereg() deregisters a memory region
 * and deletes a local memory registration object.
 *
 * RETURN VALUE
 * The rpma_mr_dereg() function returns 0 on success or a negative error code
 * on failure. rpma_mr_dereg() does not set *mr_ptr value to NULL on failure.
 *
 * ERRORS
 * rpma_mr_dereg() can fail with the following errors:
 *
 * - RPMA_E_INVAL - mr_ptr is NULL
 * - RPMA_E_PROVIDER - memory deregistration failed
 *
 * SEE ALSO
 * rpma_mr_reg(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_mr_dereg(struct rpma_mr_local **mr_ptr);

/** 3
 * rpma_mr_get_descriptor - get a descriptor of a memory region
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_mr_local;
 *	int rpma_mr_get_descriptor(const struct rpma_mr_local *mr, void *desc);
 *
 * DESCRIPTION
 * rpma_mr_get_descriptor() writes a network-transferable description
 * of the provided local memory region (called 'descriptor').
 * Once the descriptor is transferred to the other side it should be decoded
 * by rpma_mr_remote_from_descriptor() to create a remote memory region's
 * structure which allows for Remote Memory Access.
 * Please see librpma(7) for details.
 *
 * RETURN VALUE
 * The rpma_mr_get_descriptor() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_mr_get_descriptor() can fail with the following error:
 *
 * - RPMA_E_INVAL - mr or desc is NULL
 *
 * SEE ALSO
 * rpma_mr_get_descriptor_size(3), rpma_mr_reg(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_mr_get_descriptor(const struct rpma_mr_local *mr, void *desc);

/** 3
 * rpma_mr_remote_from_descriptor - create a memory region from a descriptor
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_mr_remote;
 *	int rpma_mr_remote_from_descriptor(const void *desc,
 *			size_t desc_size, struct rpma_mr_remote **mr_ptr);
 *
 * DESCRIPTION
 * Create a remote memory region's structure based on the provided descriptor
 * with a network-transferable description of the memory region local to
 * the remote peer.
 *
 * RETURN VALUE
 * The rpma_mr_remote_from_descriptor() function returns 0 on success
 * or a negative error code on failure. rpma_mr_remote_from_descriptor()
 * does not set *mr_ptr value on failure.
 *
 * ERRORS
 * rpma_mr_remote_from_descriptor() can fail with the following errors:
 *
 * - RPMA_E_INVAL - desc or mr_ptr is NULL
 * - RPMA_E_INVAL - incompatible descriptor size
 * - RPMA_E_NOSUPP - deserialized information does not represent a valid memory
 * region
 * - RPMA_E_NOMEM - out of memory
 *
 * SEE ALSO
 * rpma_mr_remote_delete(3), rpma_mr_remote_get_flush_type(3),
 * rpma_mr_remote_get_size(3), rpma_flush(3), rpma_read(3), rpma_write(3),
 * rpma_write_atomic(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_mr_remote_from_descriptor(const void *desc,
		size_t desc_size, struct rpma_mr_remote **mr_ptr);

/** 3
 * rpma_mr_get_descriptor_size - get size of the memory region descriptor
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_mr_local;
 *	int rpma_mr_get_descriptor_size(const struct rpma_mr_local *mr,
 *			size_t *desc_size);
 *
 * DESCRIPTION
 * rpma_mr_get_descriptor_size() gets size of the memory region descriptor.
 *
 * RETURN VALUE
 * The rpma_mr_get_descriptor_size() function returns 0 on success
 * or a negative error code on failure. rpma_mr_get_descriptor_size()
 * does not set *desc_size value on failure.
 *
 * ERRORS
 * rpma_mr_get_descriptor_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - mr or desc_size is NULL
 *
 * SEE ALSO
 * rpma_mr_get_descriptor(3), rpma_mr_reg(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_mr_get_descriptor_size(const struct rpma_mr_local *mr,
		size_t *desc_size);

/** 3
 * rpma_mr_remote_get_size - get a remote memory region size
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_mr_remote;
 *	int rpma_mr_remote_get_size(const struct rpma_mr_remote *mr,
 *			size_t *size);
 *
 * DESCRIPTION
 * rpma_mr_remote_get_size() gets the size of the remote memory region.
 *
 * RETURN VALUE
 * The rpma_mr_remote_get_size() function returns 0 on success
 * or a negative error code on failure. rpma_mr_remote_get_size()
 * does not set *size value on failure.
 *
 * ERRORS
 * rpma_mr_remote_get_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - mr or size is NULL
 *
 * SEE ALSO
 * rpma_mr_remote_from_descriptor(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_mr_remote_get_size(const struct rpma_mr_remote *mr, size_t *size);

/** 3
 * rpma_mr_remote_get_flush_type - get a remote memory region's flush types
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_mr_remote;
 *	int rpma_mr_remote_get_flush_type(const struct rpma_mr_remote *mr,
 *			int *flush_type);
 *
 * DESCRIPTION
 * rpma_mr_remote_get_flush_type() gets flush types supported
 * by the remote memory region.
 *
 * RETURN VALUE
 * The rpma_mr_remote_get_flush_type() function returns 0 on success
 * or a negative error code on failure. rpma_mr_remote_get_flush_type()
 * does not set *flush_type value on failure.
 *
 * ERRORS
 * rpma_mr_remote_get_flush_type() can fail with the following error:
 *
 * - RPMA_E_INVAL - mr or flush_type is NULL
 *
 * SEE ALSO
 * rpma_mr_remote_from_descriptor(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_mr_remote_get_flush_type(const struct rpma_mr_remote *mr,
		int *flush_type);

/** 3
 * rpma_mr_remote_delete - delete the remote memory region's structure
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_mr_remote;
 *	int rpma_mr_remote_delete(struct rpma_mr_remote **mr_ptr);
 *
 * DESCRIPTION
 * rpma_mr_remote_delete() deletes the remote memory region's structure.
 *
 * RETURN VALUE
 * The rpma_mr_remote_delete() function returns 0 on success
 * or a negative error code on failure. rpma_mr_remote_delete()
 * does not set *mr_ptr value to NULL on failure.
 *
 * ERRORS
 * rpma_mr_remote_delete() can fail with the following error:
 *
 * - RPMA_E_INVAL - mr_ptr is NULL
 *
 * SEE ALSO
 * rpma_mr_remote_from_descriptor(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_mr_remote_delete(struct rpma_mr_remote **mr_ptr);

/* connection configuration */

struct rpma_conn_cfg;

/** 3
 * rpma_conn_cfg_new - create a new connection configuration object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_new(struct rpma_conn_cfg **cfg_ptr);
 *
 * DESCRIPTION
 * rpma_conn_cfg_new() creates a new connection configuration object.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_new() function returns 0 on success or a negative
 * error code on failure. rpma_conn_cfg_new() does not set
 * *cfg_ptr value on failure.
 *
 * ERRORS
 * rpma_conn_cfg_new() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg_ptr is NULL
 * - RPMA_E_NOMEM - out of memory
 *
 * SEE ALSO
 * rpma_conn_cfg_delete(3), rpma_conn_cfg_get_cq_size(3),
 * rpma_conn_cfg_get_rq_size(3), rpma_conn_cfg_get_sq_size(3),
 * rpma_conn_cfg_get_timeout(3), rpma_conn_cfg_set_cq_size(3),
 * rpma_conn_cfg_set_rq_size(3), rpma_conn_cfg_set_sq_size(3),
 * rpma_conn_cfg_set_timeout(3), rpma_conn_req_new(3), rpma_ep_next_conn_req(3),
 * librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_cfg_new(struct rpma_conn_cfg **cfg_ptr);

/** 3
 * rpma_conn_cfg_delete - delete the connection configuration object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_delete(struct rpma_conn_cfg **cfg_ptr);
 *
 * DESCRIPTION
 * rpma_conn_cfg_delete() deletes the connection configuration object.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_delete() function returns 0 on success or a negative
 * error code on failure. rpma_conn_cfg_delete() sets *cfg_ptr value to NULL
 * on success and on failure.
 *
 * ERRORS
 * rpma_conn_cfg_delete() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg_ptr is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_cfg_delete(struct rpma_conn_cfg **cfg_ptr);

/** 3
 * rpma_conn_cfg_set_timeout - set connection establishment timeout
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_set_timeout(struct rpma_conn_cfg *cfg,
 *			int timeout_ms);
 *
 * DESCRIPTION
 * rpma_conn_cfg_set_timeout() sets the connection establishment timeout.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_set_timeout() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_conn_cfg_set_timeout() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg is NULL or timeout_ms < 0
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_get_timeout(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_set_timeout(struct rpma_conn_cfg *cfg, int timeout_ms);

/** 3
 * rpma_conn_cfg_get_timeout - get connection establishment timeout
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_get_timeout(const struct rpma_conn_cfg *cfg,
 *			int *timeout_ms);
 *
 * DESCRIPTION
 * rpma_conn_cfg_get_timeout() gets the connection establishment timeout.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_get_timeout() function returns 0 on success or a negative
 * error code on failure. rpma_conn_cfg_get_timeout() does not set
 * *timeout_ms value on failure.
 *
 * ERRORS
 * rpma_conn_cfg_get_timeout() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg or timeout_ms is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_set_timeout(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_get_timeout(const struct rpma_conn_cfg *cfg, int *timeout_ms);

/** 3
 * rpma_conn_cfg_set_cq_size - set CQ size for the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_set_cq_size(struct rpma_conn_cfg *cfg,
 *			uint32_t cq_size);
 *
 * DESCRIPTION
 * rpma_conn_cfg_set_cq_size() sets the CQ size for the connection.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_set_cq_size() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_conn_cfg_set_cq_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_get_cq_size(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_set_cq_size(struct rpma_conn_cfg *cfg, uint32_t cq_size);

/** 3
 * rpma_conn_cfg_get_cq_size - get CQ size for the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_get_cq_size(const struct rpma_conn_cfg *cfg,
 *			uint32_t *cq_size);
 *
 * DESCRIPTION
 * rpma_conn_cfg_get_cq_size() gets the CQ size for the connection.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_get_cq_size() function returns 0 on success or a negative
 * error code on failure. rpma_conn_cfg_get_cq_size() does not set
 * *cq_size value on failure.
 *
 * ERRORS
 * rpma_conn_cfg_get_cq_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg or cq_size is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_set_cq_size(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_get_cq_size(const struct rpma_conn_cfg *cfg,
		uint32_t *cq_size);

/** 3
 * rpma_conn_cfg_set_sq_size - set SQ size for the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_set_sq_size(struct rpma_conn_cfg *cfg,
 *			uint32_t sq_size);
 *
 * DESCRIPTION
 * rpma_conn_cfg_set_sq_size() sets the SQ size for the connection.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_set_sq_size() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_conn_cfg_set_sq_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_get_sq_size(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_set_sq_size(struct rpma_conn_cfg *cfg, uint32_t sq_size);

/** 3
 * rpma_conn_cfg_get_sq_size - get SQ size for the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_get_sq_size(const struct rpma_conn_cfg *cfg,
 *			uint32_t *sq_size);
 *
 * DESCRIPTION
 * rpma_conn_cfg_get_sq_size() gets the SQ size for the connection.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_get_sq_size() function returns 0 on success or a negative
 * error code on failure. rpma_conn_cfg_get_sq_size() does not set
 * *sq_size value on failure.
 *
 * ERRORS
 * rpma_conn_cfg_get_sq_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg or sq_size is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_set_sq_size(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_get_sq_size(const struct rpma_conn_cfg *cfg,
		uint32_t *sq_size);

/** 3
 * rpma_conn_cfg_set_rq_size - set RQ size for the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_set_rq_size(struct rpma_conn_cfg *cfg,
 *			uint32_t rq_size);
 *
 * DESCRIPTION
 * rpma_conn_cfg_set_rq_size() sets the RQ size for the connection.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_set_rq_size() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_conn_cfg_set_rq_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_get_rq_size(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_set_rq_size(struct rpma_conn_cfg *cfg, uint32_t rq_size);

/** 3
 * rpma_conn_cfg_get_rq_size - get RQ size for the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_cfg;
 *	int rpma_conn_cfg_get_rq_size(const struct rpma_conn_cfg *cfg,
 *			uint32_t *rq_size);
 *
 * DESCRIPTION
 * rpma_conn_cfg_get_rq_size() gets the RQ size for the connection.
 *
 * RETURN VALUE
 * The rpma_conn_cfg_get_rq_size() function returns 0 on success or a negative
 * error code on failure. rpma_conn_cfg_get_rq_size() does not set
 * *rq_size value on failure.
 *
 * ERRORS
 * rpma_conn_cfg_get_rq_size() can fail with the following error:
 *
 * - RPMA_E_INVAL - cfg or rq_size is NULL
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_cfg_set_rq_size(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_cfg_get_rq_size(const struct rpma_conn_cfg *cfg,
		uint32_t *rq_size);

/* connection */

struct rpma_conn;

/** 3
 * rpma_conn_get_event_fd - get an event file descriptor of the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	int rpma_conn_get_event_fd(const struct rpma_conn *conn, int *fd);
 *
 * DESCRIPTION
 * rpma_conn_get_event_fd() gets an event file descriptor of the connection.
 *
 * RETURN VALUE
 * The rpma_conn_get_event_fd() function returns 0 on success or a negative
 * error code on failure. rpma_conn_get_event_fd() does not set
 * *fd value on failure.
 *
 * ERRORS
 * rpma_conn_get_event_fd() can fail with the following error:
 *
 * - RPMA_E_INVAL - conn or fd is NULL
 *
 * SEE ALSO
 * rpma_conn_next_event(3), rpma_conn_req_connect(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_get_event_fd(const struct rpma_conn *conn, int *fd);

enum rpma_conn_event {
	RPMA_CONN_UNDEFINED = -1,	/* Undefined connection event */
	RPMA_CONN_ESTABLISHED,		/* Connection established */
	RPMA_CONN_CLOSED,			/* Connection closed */
	RPMA_CONN_LOST				/* Connection lost */
};

/** 3
 * rpma_conn_next_event - obtain a connection status
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	enum rpma_conn_event {
 *		RPMA_CONN_UNDEFINED = -1,
 *		RPMA_CONN_ESTABLISHED,
 *		RPMA_CONN_CLOSED,
 *		RPMA_CONN_LOST,
 *	};
 *
 *	int rpma_conn_next_event(struct rpma_conn *conn,
 *			enum rpma_conn_event *event);
 *
 * DESCRIPTION
 * rpma_conn_next_event() obtains the next event from the connection.
 * Types of events:
 * - RPMA_CONN_UNDEFINED - undefined connection event
 * - RPMA_CONN_ESTABLISHED - connection established
 * - RPMA_CONN_CLOSED - connection closed
 * - RPMA_CONN_LOST - connection lost
 *
 * RETURN VALUE
 * The rpma_conn_next_event() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_conn_next_event() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn or event is NULL
 * - RPMA_E_UNKNOWN - unexpected event
 * - RPMA_E_PROVIDER - rdma_get_cm_event() or rdma_ack_cm_event() failed
 * - RPMA_E_NOMEM - out of memory
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_conn_disconnect(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_next_event(struct rpma_conn *conn, enum rpma_conn_event *event);

/** 3
 * rpma_utils_conn_event_2str - convert RPMA_CONN_* enum to a string
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	const char *rpma_utils_conn_event_2str(enum rpma_conn_event conn_event);
 *
 *	enum rpma_conn_event{
 *		RPMA_CONN_UNDEFINED = -1,
 *		RPMA_CONN_ESTABLISHED,
 *		RPMA_CONN_CLOSED,
 *		RPMA_CONN_LOST
 *	};
 *
 * DESCRIPTION
 * rpma_utils_conn_event_2str() converts RPMA_CONN_* enum to the const string
 * representation.
 *
 * RETURN VALUE
 * The rpma_utils_conn_event_2str() function returns the const string
 * representation of RPMA_CONN_* enums.
 *
 * ERRORS
 * rpma_utils_conn_event_2str() can not fail.
 *
 * SEE ALSO
 * rpma_conn_next_event(3), librpma(7) and https://pmem.io/rpma/
 */
const char *rpma_utils_conn_event_2str(enum rpma_conn_event conn_event);

struct rpma_conn_private_data {
	void *ptr;
	uint8_t len;
};

/** 3
 * rpma_conn_get_private_data - get a pointer to the connection's private data
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_conn_private_data;
 *	int rpma_conn_get_private_data(const struct rpma_conn *conn,
 *			struct rpma_conn_private_data *pdata);
 *
 * DESCRIPTION
 * rpma_conn_get_private_data() obtains the pointer to the private data
 * given by the other side of the connection.
 *
 * RETURN VALUE
 * The rpma_conn_get_private_data() function returns 0 on success or a negative
 * error code on failure. rpma_conn_get_private_data() does not set
 * *pdata value on failure.
 *
 * ERRORS
 * rpma_conn_get_private_data() can fail with the following error:
 *
 * - RPMA_E_INVAL - conn or pdata is NULL
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_get_private_data(const struct rpma_conn *conn,
		struct rpma_conn_private_data *pdata);

/** 3
 * rpma_conn_apply_remote_peer_cfg - apply remote peer cfg to the connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_peer_cfg;
 *	int rpma_conn_apply_remote_peer_cfg(struct rpma_conn *conn,
 *			const struct rpma_peer_cfg *pcfg);
 *
 * DESCRIPTION
 * rpma_conn_apply_remote_peer_cfg() applies the remote peer configuration
 * to the connection.
 *
 * RETURN VALUE
 * The rpma_conn_apply_remote_peer_cfg() function returns 0 on success
 * or a negative error code on failure. rpma_conn_apply_remote_peer_cfg()
 * does not set *pcfg value on failure.
 *
 * ERRORS
 * rpma_conn_apply_remote_peer_cfg() can fail with the following error:
 *
 * - RPMA_E_INVAL - conn or pcfg are NULL
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_peer_cfg_new(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_apply_remote_peer_cfg(struct rpma_conn *conn,
		const struct rpma_peer_cfg *pcfg);

/** 3
 * rpma_conn_disconnect - tear the connection down
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	int rpma_conn_disconnect(struct rpma_conn *conn);
 *
 * DESCRIPTION
 * rpma_conn_disconnect() tears the connection down.
 *
 * - It may initiate disconnecting the connection. In this case,
 * the end of disconnecting is signalled by the RPMA_CONN_CLOSED event via
 * rpma_conn_next_event() or
 * - It may be called after receiving the RPMA_CONN_CLOSED event. In this case,
 * the disconnection is done when rpma_conn_disconnect() returns with success.
 *
 * RETURN VALUE
 * The rpma_conn_disconnect() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_conn_disconnect() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn is NULL
 * - RPMA_E_PROVIDER - rdma_disconnect() failed
 *
 * SEE ALSO
 * rpma_conn_delete(3), rpma_conn_next_event(3), rpma_conn_req_connect(3),
 * librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_disconnect(struct rpma_conn *conn);

/** 3
 * rpma_conn_delete - delete already closed connection
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	int rpma_conn_delete(struct rpma_conn **conn_ptr);
 *
 * DESCRIPTION
 * rpma_conn_delete() deletes already closed connection.
 *
 * RETURN VALUE
 * The rpma_conn_delete() function returns 0 on success or a negative error code
 * on failure. rpma_conn_delete() sets *conn_ptr value to NULL on success and on
 * failure.
 *
 * ERRORS
 * rpma_conn_delete() can fail with the following errors:
 * - RPMA_E_INVAL - conn_ptr is NULL
 * - RPMA_E_PROVIDER - ibv_destroy_cq() or rdma_destroy_id() failed
 *
 * SEE ALSO
 * rpma_conn_disconnect(3), rpma_conn_req_connect(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_delete(struct rpma_conn **conn_ptr);

/* incoming / outgoing connection request */

struct rpma_conn_req;

/** 3
 * rpma_conn_req_new - create a new outgoing connection request object
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer;
 *	struct rpma_conn_cfg;
 *	struct rpma_conn_req;
 *	int rpma_conn_req_new(struct rpma_peer *peer, const char *addr,
 *			const char *port, const struct rpma_conn_cfg *cfg,
 *			struct rpma_conn_req **req_ptr);
 *
 * DESCRIPTION
 * rpma_conn_req_new() creates a new outgoing connection request object
 * using reliable, connection-oriented
 * and message-based (RDMA_PS_TCP) QP communication.
 *
 * RETURN VALUE
 * The rpma_conn_req_new() function returns 0 on success or a negative
 * error code on failure. rpma_conn_req_new() does not set
 * *req_ptr value on failure.
 *
 * ERRORS
 * rpma_conn_req_new() can fail with the following errors:
 *
 * - RPMA_E_INVAL - peer, addr, port or req_ptr is NULL
 * - RPMA_E_NOMEM - out of memory
 * - RPMA_E_PROVIDER - rdma_create_id(3), rdma_resolve_addr(3),
 *   rdma_resolve_route(3) or ibv_create_cq(3) failed
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_req_connect(3), rpma_conn_req_delete(3),
 * rpma_conn_req_recv(3), rpma_ep_next_conn_req(3), rpma_peer_new(3), librpma(7)
 * and https://pmem.io/rpma/
 */
int rpma_conn_req_new(struct rpma_peer *peer, const char *addr,
		const char *port, const struct rpma_conn_cfg *cfg,
		struct rpma_conn_req **req_ptr);

/** 3
 * rpma_conn_req_delete - delete the connection requests
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_req;
 *	int rpma_conn_req_delete(struct rpma_conn_req **req_ptr);
 *
 * DESCRIPTION
 * rpma_conn_req_delete() deletes the connection requests both
 * incoming and outgoing.
 *
 * RETURN VALUE
 * The rpma_conn_req_delete() function returns 0 on success or a negative
 * error code on failure. rpma_conn_req_delete() sets *req_ptr value to NULL
 * on success and on failure.
 *
 * ERRORS
 * rpma_conn_req_delete() can fail with the following errors:
 *
 * - RPMA_E_INVAL - req_ptr is NULL
 * - RPMA_E_PROVIDER
 *	- rdma_destroy_qp(3) or ibv_destroy_cq(3) failed
 *	- rdma_reject(3) or rdma_ack_cm_event(3) failed (passive side only)
 *	- rdma_destroy_id(3) failed (active side only)
 *
 * SEE ALSO
 * rpma_conn_req_new(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_req_delete(struct rpma_conn_req **req_ptr);

/** 3
 * rpma_conn_req_connect - initiate processing the connection request
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_req;
 *	struct rpma_conn_private_data;
 *	struct rpma_conn;
 *	int rpma_conn_req_connect(struct rpma_conn_req **req_ptr,
 *			const struct rpma_conn_private_data *pdata,
 *			struct rpma_conn **conn_ptr);
 *
 * DESCRIPTION
 * rpma_conn_req_connect() initiates processing the connection requests both
 * incoming and outgoing. The end of processing is signalled by
 * the RPMA_CONN_ESTABLISHED event via rpma_conn_next_event().
 *
 * RETURN VALUE
 * The rpma_conn_req_connect() function returns 0 on success or a negative error
 * code on failure. On success, the newly created connection object is stored in
 * *conn_ptr whereas *req_ptr is consumed and set to NULL. On failure,
 * rpma_conn_req_connect() does not set *conn_ptr whereas *req_ptr is consumed
 * and set to NULL.
 *
 * ERRORS
 * rpma_conn_req_connect() can fail with the following errors:
 *
 * - RPMA_E_INVAL - req_ptr, *req_ptr or conn_ptr is NULL
 * - RPMA_E_INVAL - pdata is not NULL whereas pdata->len == 0
 * - RPMA_E_NOMEM - out of memory
 * - RPMA_E_PROVIDER - initiating a connection request failed (active side only)
 * - RPMA_E_PROVIDER - accepting the connection request failed
 * (passive side only)
 * - RPMA_E_PROVIDER - freeing a communication event failed (passive side only)
 *
 * SEE ALSO
 * rpma_conn_apply_remote_peer_cfg(3), rpma_conn_delete(3),
 * rpma_conn_disconnect(3), rpma_conn_get_completion_fd(3),
 * rpma_conn_get_event_fd(3), rpma_conn_get_private_data(3),
 * rpma_conn_completion_get(3), rpma_conn_next_event(3),
 * rpma_conn_completion_wait(3), rpma_conn_req_new(3),
 * rpma_ep_next_conn_req(3), rpma_flush(3), rpma_read(3), rpma_recv(3),
 * rpma_send(3), rpma_write(3), rpma_write_atomic(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_req_connect(struct rpma_conn_req **req_ptr,
		const struct rpma_conn_private_data *pdata,
		struct rpma_conn **conn_ptr);

/** 3
 * rpma_conn_req_recv - initiate the receive operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn_req;
 *	struct rpma_mr_local;
 *	int rpma_conn_req_recv(struct rpma_conn_req *req,
 *			struct rpma_mr_local *dst, size_t offset,
 *			size_t len, const void *op_context);
 *
 * DESCRIPTION
 * rpma_conn_req_recv() initiates the receive operation. It prepares a buffer
 * for a message sent from other side of the connection.
 * Please see rpma_send(3). This is a variant of rpma_recv(3) which may be used
 * before the connection is established.
 *
 * RETURN VALUE
 * The rpma_conn_req_recv() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_conn_req_recv() can fail with the following errors:
 *
 * - RPMA_E_INVAL - req or src or op_context is NULL
 * - RPMA_E_PROVIDER - ibv_post_recv(3) failed
 *
 * SEE ALSO
 * rpma_conn_req_new(3), rpma_mr_reg(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_req_recv(struct rpma_conn_req *req,
		struct rpma_mr_local *dst, size_t offset,
		size_t len, const void *op_context);

/* server-side setup */

struct rpma_ep;

/** 3
 * rpma_ep_listen - create a listening endpoint
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_peer;
 *	struct rpma_ep;
 *	int rpma_ep_listen(struct rpma_peer *peer, const char *addr,
 *			const char *port, struct rpma_ep **ep_ptr);
 *
 * DESCRIPTION
 * rpma_ep_listen() creates an endpoint and initiates listening for incoming
 * connections using reliable, connection-oriented and message-based
 * (RDMA_PS_TCP) QP communication.
 *
 * RETURN VALUE
 * The rpma_ep_listen() function returns 0 on success or a negative
 * error code on failure. rpma_ep_listen() does not set
 * *ep_ptr value on failure.
 *
 * ERRORS
 * rpma_ep_listen() can fail with the following errors:
 *
 * - RPMA_E_INVAL - peer, addr, port or ep_ptr is NULL
 * - RPMA_E_PROVIDER - rdma_create_event_channel(3), rdma_create_id(3),
 *   rdma_getaddrinfo(3), rdma_listen(3) failed
 * - RPMA_E_NOMEM - out of memory
 *
 * SEE ALSO
 * rpma_ep_get_fd(3), rpma_ep_next_conn_req(3), rpma_ep_shutdown(3),
 * rpma_peer_new(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_ep_listen(struct rpma_peer *peer, const char *addr,
		const char *port, struct rpma_ep **ep_ptr);

/** 3
 * rpma_ep_shutdown - stop listening and delete the endpoint
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_ep;
 *	int rpma_ep_shutdown(struct rpma_ep **ep_ptr);
 *
 * DESCRIPTION
 * rpma_ep_shutdown() stops listening for incoming connections
 * and deletes the endpoint. The connections established using the endpoint
 * will still exist after deleting the endpoint.
 *
 * RETURN VALUE
 * The rpma_ep_shutdown() function returns 0 on success or a negative
 * error code on failure. rpma_ep_shutdown() does not set
 * *ep_ptr value to NULL on failure.
 *
 * ERRORS
 * rpma_ep_shutdown() can fail with the following errors:
 *
 * - RPMA_E_INVAL - ep_ptr is NULL
 * - RPMA_E_PROVIDER - rdma_destroy_id(3) failed
 *
 * SEE ALSO
 * rpma_ep_listen(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_ep_shutdown(struct rpma_ep **ep_ptr);

/** 3
 * rpma_ep_get_fd - get a file descriptor of the endpoint
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_ep;
 *	int rpma_ep_get_fd(const struct rpma_ep *ep, int *fd);
 *
 * DESCRIPTION
 * rpma_ep_get_fd() gets the file descriptor of the endpoint.
 *
 * RETURN VALUE
 * The rpma_ep_get_fd() function returns 0 on success or a negative
 * error code on failure. rpma_ep_get_fd() does not set
 * *fd value on failure.
 *
 * ERRORS
 * rpma_ep_get_fd() can fail with the following error:
 *
 * - RPMA_E_INVAL - ep or fd is NULL
 *
 * SEE ALSO
 * rpma_ep_listen(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_ep_get_fd(const struct rpma_ep *ep, int *fd);

/** 3
 * rpma_ep_next_conn_req - obtain an incoming connection request
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_ep;
 *	struct rpma_conn_cfg;
 *	struct rpma_conn_req;
 *	int rpma_ep_next_conn_req(struct rpma_ep *ep,
 *			const struct rpma_conn_cfg *cfg,
 *			struct rpma_conn_req **req_ptr);
 *
 * DESCRIPTION
 * rpma_ep_next_conn_req() obtains the next connection request
 * from the endpoint.
 *
 * RETURN VALUE
 * The rpma_ep_next_conn_req() function returns 0 on success or a negative
 * error code on failure. rpma_ep_next_conn_req() does not set
 * *req_ptr value on failure.
 *
 * ERRORS
 * rpma_ep_next_conn_req() can fail with the following errors:
 *
 * - RPMA_E_INVAL - ep or req_ptr is NULL
 * - RPMA_E_INVAL - obtained an event different than a connection request
 * - RPMA_E_PROVIDER - rdma_get_cm_event(3) failed
 * - RPMA_E_NOMEM - out of memory
 * - RPMA_E_NO_EVENT - no next connection request available
 *
 * SEE ALSO
 * rpma_conn_cfg_new(3), rpma_conn_req_delete(3), rpma_conn_req_connect(3),
 * rpma_ep_listen(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_ep_next_conn_req(struct rpma_ep *ep,
		const struct rpma_conn_cfg *cfg,
		struct rpma_conn_req **req_ptr);

/* remote memory access functions */

/* generate operation completion on error */
#define RPMA_F_COMPLETION_ON_ERROR	(1 << 0)
/* generate operation completion regardless of its result */
#define RPMA_F_COMPLETION_ALWAYS	(1 << 1 | RPMA_F_COMPLETION_ON_ERROR)

/** 3
 * rpma_read - initiate the read operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_mr_local;
 *	struct rpma_mr_remote;
 *	int rpma_read(struct rpma_conn *conn,
 *			struct rpma_mr_local *dst, size_t dst_offset,
 *			const struct rpma_mr_remote *src,  size_t src_offset,
 *			size_t len, int flags, const void *op_context);
 *
 * DESCRIPTION
 * rpma_read() initiates transferring data from the remote memory
 * to the local memory.
 * The attribute flags set the completion notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generate the completion on error
 * - RPMA_F_COMPLETION_ALWAYS - generate the completion regardless of result of
 * the operation
 *
 * RETURN VALUE
 * The rpma_read() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_read() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn, dst or src is NULL
 * - RPMA_E_INVAL - flags are not set
 * - RPMA_E_PROVIDER - ibv_post_send(3) failed
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_mr_reg(3), rpma_mr_remote_from_descriptor(3),
 * librpma(7) and https://pmem.io/rpma/
 */
int rpma_read(struct rpma_conn *conn,
		struct rpma_mr_local *dst, size_t dst_offset,
		const struct rpma_mr_remote *src,  size_t src_offset,
		size_t len, int flags, const void *op_context);

/** 3
 * rpma_write - initiate the write operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_mr_local;
 *	struct rpma_mr_remote;
 *	int rpma_write(struct rpma_conn *conn,
 *			struct rpma_mr_remote *dst, size_t dst_offset,
 *			const struct rpma_mr_local *src,  size_t src_offset,
 *			size_t len, int flags, const void *op_context);
 *
 * DESCRIPTION
 * rpma_write() initiates transferring data from the local memory
 * to the remote memory.
 * The attribute flags set the completion notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generate the completion on error
 * - RPMA_F_COMPLETION_ALWAYS - generate the completion regardless of result of
 * the operation
 *
 * RETURN VALUE
 * The rpma_write() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_write() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn, dst or src is NULL
 * - RPMA_E_INVAL - flags are not set
 * - RPMA_E_PROVIDER - ibv_post_send(3) failed
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_mr_reg(3),
 * rpma_mr_remote_from_descriptor(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_write(struct rpma_conn *conn,
		struct rpma_mr_remote *dst, size_t dst_offset,
		const struct rpma_mr_local *src,  size_t src_offset,
		size_t len, int flags, const void *op_context);

#define RPMA_ATOMIC_WRITE_ALIGNMENT 8

/** 3
 * rpma_write_atomic - initiate the atomic write operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_mr_local;
 *	struct rpma_mr_remote;
 *	int rpma_write_atomic(struct rpma_conn *conn,
 *			struct rpma_mr_remote *dst, size_t dst_offset,
 *			const struct rpma_mr_local *src,  size_t src_offset,
 *			int flags, const void *op_context);
 *
 * DESCRIPTION
 * rpma_write_atomic() initiates the atomic write operation (transferring
 * data from the local memory to the remote memory). The atomic write operation
 * allows transferring 8 bytes of data (RPMA_ATOMIC_WRITE_ALIGNMENT) and storing
 * them atomically in the remote memory.
 * The attribute flags set the completion notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generate the completion on error
 * - RPMA_F_COMPLETION_ALWAYS - generate the completion regardless of result of
 * the operation
 *
 * RETURN VALUE
 * The rpma_write_atomic() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_write_atomic() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn, dst or src is NULL
 * - RPMA_E_INVAL - dst_offset is not aligned to 8 bytes
 * - RPMA_E_INVAL - flags are not set
 * - RPMA_E_PROVIDER - ibv_post_send(3) failed
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_mr_reg(3),
 * rpma_mr_remote_from_descriptor(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_write_atomic(struct rpma_conn *conn,
		struct rpma_mr_remote *dst, size_t dst_offset,
		const struct rpma_mr_local *src,  size_t src_offset,
		int flags, const void *op_context);

/*
 * possible types of rpma_flush() operation
 */
enum rpma_flush_type {
	/* flush data down to the persistent domain */
	RPMA_FLUSH_TYPE_PERSISTENT,
	/* flush data deep enough to make it visible on the remote node */
	RPMA_FLUSH_TYPE_VISIBILITY,
};

/** 3
 * rpma_flush - initiate the flush operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_mr_remote;
 *	enum rpma_flush_type {
 *		RPMA_FLUSH_TYPE_PERSISTENT,
 *		RPMA_FLUSH_TYPE_VISIBILITY,
 *	};
 *
 *	int rpma_flush(struct rpma_conn *conn,
 *			struct rpma_mr_remote *dst, size_t dst_offset,
 *			size_t len, enum rpma_flush_type type, int flags,
 *			const void *op_context);
 *
 * DESCRIPTION
 * rpma_flush() initiates the flush operation (finalizing a transfer of data
 * to the remote memory).
 * Possible types of rpma_flush() operation:
 * - RPMA_FLUSH_TYPE_PERSISTENT - flush data down to the persistent domain
 * - RPMA_FLUSH_TYPE_VISIBILITY - flush data deep enough to make it visible
 * on the remote node
 *
 * The attribute flags set the completion notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generate the completion on error
 * - RPMA_F_COMPLETION_ALWAYS - generate the completion regardless of result of
 * the operation
 *
 * RETURN VALUE
 * The rpma_flush() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_flush() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn or dst is NULL
 * - RPMA_E_INVAL - unknown type value
 * - RPMA_E_INVAL - flags are not set
 * - RPMA_E_PROVIDER - ibv_post_send(3) failed
 * - RPMA_E_NOSUPP - type is RPMA_FLUSH_TYPE_PERSISTENT and
 * the direct write to pmem is not supported
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_mr_remote_from_descriptor(3), librpma(7)
 * and https://pmem.io/rpma/
 */
int rpma_flush(struct rpma_conn *conn,
		struct rpma_mr_remote *dst, size_t dst_offset, size_t len,
		enum rpma_flush_type type, int flags, const void *op_context);

/** 3
 * rpma_send - initiate the send operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_mr_local;
 *	int rpma_send(struct rpma_conn *conn,
 *			const struct rpma_mr_local *src, size_t offset,
 *			size_t len, int flags, const void *op_context);
 *
 * DESCRIPTION
 * rpma_send() initiates the send operation which transfers a message from
 * the local memory to other side of the connection.
 * The attribute flags set the completion notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generate the completion on error
 * - RPMA_F_COMPLETION_ALWAYS - generate the completion regardless of result of
 * the operation
 *
 * RETURN VALUE
 * The rpma_send() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_send() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn or src is NULL
 * - RPMA_E_INVAL - flags are not set
 * - RPMA_E_PROVIDER - ibv_post_send(3) failed
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_mr_reg(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_send(struct rpma_conn *conn,
		const struct rpma_mr_local *src, size_t offset, size_t len,
		int flags, const void *op_context);

/** 3
 * rpma_send_with_imm - initiate the send operation with immediate data
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_mr_local;
 *	int rpma_send_with_imm(struct rpma_conn *conn,
 *			const struct rpma_mr_local *src, size_t offset,
 *			size_t len, int flags, const void *op_context,
 *			uint32_t imm_data);
 *
 * DESCRIPTION
 * rpma_send_with_imm() initiates the send operation with immediate data which
 * transfers a message from the local memory to other side of the connection.
 * The attribute flags set the completion notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generate the completion on error
 * - RPMA_F_COMPLETION_ALWAYS - generate the completion regardless of result of
 * the operation
 *
 * RETURN VALUE
 * The rpma_send_with_imm() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_send_with_imm() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn or src is NULL
 * - RPMA_E_INVAL - flags are not set
 * - RPMA_E_PROVIDER - ibv_post_send(3) failed
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_mr_reg(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_send_with_imm(struct rpma_conn *conn,
		const struct rpma_mr_local *src, size_t offset, size_t len,
		int flags, const void *op_context, uint32_t imm_data);

/** 3
 * rpma_recv - initiate the receive operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_mr_local;
 *	int rpma_recv(struct rpma_conn *conn,
 *			struct rpma_mr_local *dst, size_t offset,
 *			size_t len, const void *op_context);
 *
 * DESCRIPTION
 * rpma_recv() initiates the receive operation which prepares a buffer for
 * a message sent from other side of the connection. Please see rpma_send(3).
 *
 * All buffers prepared via rpma_recv(3) form an unordered set. When a message
 * arrives it is placed in one of the buffers awaitaning and a completion for
 * the receive operation is generated.
 *
 * A buffer for an incoming message have to be prepared beforehand.
 *
 * The order of buffers in the set does not affect the order of completions of
 * receive operations get via rpma_conn_completion_get(3).
 *
 * The attribute flags set the completion notification indicator:
 * - RPMA_F_COMPLETION_ON_ERROR - generate the completion on error
 * - RPMA_F_COMPLETION_ALWAYS - generate the completion regardless of result of
 * the operation
 *
 * NOTE
 * In the RDMA standard, receive requests form an ordered queue.
 * The RPMA does NOT inherit this guarantee.
 *
 * RETURN VALUE
 * The rpma_recv() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_recv() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn or src is NULL
 * - RPMA_E_PROVIDER - ibv_post_recv(3) failed
 *
 * SEE ALSO
 * rpma_conn_req_connect(3), rpma_mr_reg(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_recv(struct rpma_conn *conn,
		struct rpma_mr_local *dst, size_t offset, size_t len,
		const void *op_context);

/* completion handling */

/** 3
 * rpma_conn_get_completion_fd - get the completion file descriptor
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	int rpma_conn_get_completion_fd(const struct rpma_conn *conn, int *fd);
 *
 * DESCRIPTION
 * rpma_conn_get_completion_fd() gets the completion file descriptor
 * of the connection.
 *
 * RETURN VALUE
 * The rpma_conn_get_completion_fd() function returns 0 on success
 * or a negative error code on failure. rpma_conn_get_completion_fd()
 * does not set *fd value on failure.
 *
 * ERRORS
 * rpma_conn_get_completion_fd() can fail with the following error:
 *
 * - RPMA_E_INVAL - conn or fd is NULL
 *
 * SEE ALSO
 * rpma_conn_completion_get(3), rpma_conn_completion_wait(3),
 * rpma_conn_req_connect(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_get_completion_fd(const struct rpma_conn *conn, int *fd);

enum rpma_op {
	RPMA_OP_READ,
	RPMA_OP_WRITE,
	RPMA_OP_FLUSH,
	RPMA_OP_SEND,
	RPMA_OP_RECV,
};

struct rpma_completion {
	void *op_context;
	enum rpma_op op;
	uint32_t byte_len;
	enum ibv_wc_status op_status;
	unsigned flags;
	union {
		uint32_t imm_data;
		uint32_t invalidate_rkey;
	};
};

/** 3
 * rpma_conn_completion_wait - wait for completions
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	int rpma_conn_completion_wait(struct rpma_conn *conn);
 *
 * DESCRIPTION
 * rpma_conn_completion_wait() waits for incoming completions. If it
 * succeeds the completions can be collected using rpma_conn_completion_get().
 *
 * RETURN VALUE
 * The rpma_conn_completion_wait() function returns 0 on success
 * or a negative error code on failure.
 *
 * ERRORS
 * rpma_conn_completion_wait() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn is NULL
 * - RPMA_E_PROVIDER - ibv_req_notify_cq(3) failed with a provider error
 * - RPMA_E_NO_COMPLETION - no completions available
 *
 * SEE ALSO
 * rpma_conn_get_completion_fd(3), rpma_conn_completion_get(3),
 * rpma_conn_req_connect(3), librpma(7) and https://pmem.io/rpma/
 */
int rpma_conn_completion_wait(struct rpma_conn *conn);

/** 3
 * rpma_conn_completion_get - receive a completion of an operation
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	struct rpma_conn;
 *	struct rpma_completion;
 *	enum rpma_op {
 *		RPMA_OP_READ,
 *		RPMA_OP_WRITE,
 *		RPMA_OP_FLUSH,
 *		RPMA_OP_SEND,
 *		RPMA_OP_RECV,
 *	};
 *
 *	int rpma_conn_completion_get(struct rpma_conn *conn,
 *			struct rpma_completion *cmpl);
 *
 * DESCRIPTION
 * rpma_conn_completion_get() receives the next available completion
 * of an already posted operation. All operations generate completion on
 * error. The operations posted with the **RPMA_F_COMPLETION_ALWAYS** flag
 * also generate a completion on success.
 * The following operations are available:
 * - RPMA_OP_READ - RMA read operation
 * - RPMA_OP_WRITE - RMA write operation
 * - RPMA_OP_FLUSH - RMA flush operation
 * - RPMA_OP_SEND - messaging send operation
 * - RPMA_OP_RECV - messaging receive operation
 *
 * RETURN VALUE
 * The rpma_conn_completion_get() function returns 0 on success
 * or a negative error code on failure.
 *
 * ERRORS
 * rpma_conn_completion_get() can fail with the following errors:
 *
 * - RPMA_E_INVAL - conn or cmpl is NULL
 * - RPMA_E_NO_COMPLETION - no completions available
 * - RPMA_E_PROVIDER - ibv_poll_cq(3) failed with a provider error
 * - RPMA_E_UNKNOWN - ibv_poll_cq(3) failed but no provider error is available
 * - RPMA_E_NOSUPP - not supported opcode
 *
 * SEE ALSO
 * rpma_conn_get_completion_fd(3), rpma_conn_completion_wait(3),
 * rpma_conn_req_connect(3), rpma_flush(3), rpma_read(3), rpma_recv(3),
 * rpma_send(3), rpma_write(3), rpma_write_atomic(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_conn_completion_get(struct rpma_conn *conn,
		struct rpma_completion *cmpl);

/* error handling */

/** 3
 * rpma_err_2str - convert RPMA error code to a string
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	const char *rpma_err_2str(int ret);
 *
 * DESCRIPTION
 * rpma_err_2str() returns the const string representation of RPMA error codes.
 *
 * ERRORS
 * rpma_err_2str() can not fail.
 *
 * SEE ALSO
 * librpma(7) and https://pmem.io/rpma/
 */
const char *rpma_err_2str(int ret);

/* librpma logging mechanism control */

/*
 * Available log levels in librpma. Log levels (except RPMA_LOG_DISABLED)
 * are used in logging API calls to indicate logging message severity.
 * Log levels are also used to define thresholds for logging.
 */
enum rpma_log_level {
	/* all messages will be suppressed */
	RPMA_LOG_DISABLED = -1,
	/* an error that causes the library to stop working immediately */
	RPMA_LOG_LEVEL_FATAL,
	/* an error that causes the library to stop working properly */
	RPMA_LOG_LEVEL_ERROR,
	/* an errors that could be handled in the upper level */
	RPMA_LOG_LEVEL_WARNING,
	/*
	 * non-massive info mainly related to public API function completions
	 * e.g. connection established
	 */
	RPMA_LOG_LEVEL_NOTICE,
	/* massive info e.g. every write operation indication */
	RPMA_LOG_LEVEL_INFO,
	/* debug info e.g. write operation dump */
	RPMA_LOG_LEVEL_DEBUG,
};

enum rpma_log_threshold {
	/*
	 * the main threshold level - the logging messages above this level
	 * won't trigger the logging functions
	 */
	RPMA_LOG_THRESHOLD,
	/*
	 * the auxiliary threshold level - may or may not be used by
	 * the logging function
	 */
	RPMA_LOG_THRESHOLD_AUX,
	RPMA_LOG_THRESHOLD_MAX
};

/** 3
 * rpma_log_set_threshold - set the logging threshold level
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	int rpma_log_set_threshold(enum rpma_log_threshold threshold,
 *			enum rpma_log_level level);
 *
 *	enum rpma_log_level {
 *		RPMA_LOG_DISABLED,
 *		RPMA_LOG_LEVEL_FATAL,
 *		RPMA_LOG_LEVEL_ERROR,
 *		RPMA_LOG_LEVEL_WARNING,
 *		RPMA_LOG_LEVEL_NOTICE,
 *		RPMA_LOG_LEVEL_INFO,
 *		RPMA_LOG_LEVEL_DEBUG,
 *	};
 *
 *	enum rpma_log_threshold {
 *		RPMA_LOG_THRESHOLD,
 *		RPMA_LOG_THRESHOLD_AUX,
 *		RPMA_LOG_THRESHOLD_MAX
 *	};
 *
 * DESCRIPTION
 * rpma_log_set_threshold() sets the logging threshold level.
 *
 * Available thresholds are:
 * - RPMA_LOG_THRESHOLD - the main threshold used to filter out
 * undesired logging messages. Messages on a higher level than the primary
 * threshold level are ignored. RPMA_LOG_DISABLED shall be used to suppress
 * logging. The default value is RPMA_LOG_WARNING.
 * - RPMA_LOG_THRESHOLD_AUX - the auxiliary threshold intended for use
 * inside the logging function (please see rpma_log_get_threshold(3)).
 * The logging function may or may not take this threshold into consideration.
 * The default value is RPMA_LOG_DISABLED.
 *
 * Available threshold levels are defined by enum rpma_log_level:
 * - RPMA_LOG_DISABLED - all messages will be suppressed
 * - RPMA_LOG_LEVEL_FATAL - an error that causes the library to stop working
 * immediately
 * - RPMA_LOG_LEVEL_ERROR - an error that causes the library to stop working
 * properly
 * - RPMA_LOG_LEVEL_WARNING - an error that could be handled in the upper level
 * - RPMA_LOG_LEVEL_NOTICE - non-massive info mainly related to public
 * API function completions e.g. connection established
 * - RPMA_LOG_LEVEL_INFO - massive info e.g. every write operation indication
 * - RPMA_LOG_LEVEL_DEBUG - debug info e.g. write operation dump
 *
 * THE DEFAULT LOGGING FUNCTION
 * The default logging function writes messages to syslog(3) and to stderr(3),
 * where syslog(3) is the primary destination
 * (RPMA_LOG_THRESHOLD applies) whereas stderr(3) is the secondary
 * destination (RPMA_LOG_THRESHOLD_AUX applies).
 *
 * RETURN VALUE
 * rpma_log_syslog_set_threshold() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_log_set_threshold() can fail with the following errors:
 * - RPMA_E_INVAL - threshold is not RPMA_LOG_THRESHOLD nor
 * RPMA_LOG_THRESHOLD_AUX
 * - RPMA_E_INVAL - level is not a value defined by enum rpma_log_level type
 * - RPMA_E_AGAIN - a temporary error occurred, the retry may fix the problem
 *
 * SEE ALSO
 * rpma_log_get_threshold(3), rpma_log_set_function(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_log_set_threshold(enum rpma_log_threshold threshold,
				enum rpma_log_level level);

/** 3
 * rpma_log_get_threshold - get the logging threshold level
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	int rpma_log_get_threshold(enum rpma_log_threshold threshold,
 *			enum rpma_log_level *level);
 *
 * DESCRIPTION
 * rpma_log_get_threshold() gets the current level of the threshold.
 * See rpma_log_set_threshold(3) for available thresholds and levels.
 *
 * RETURN VALUE
 * rpma_log_get_threshold() function returns 0 on success or a negative
 * error code on failure.
 *
 * ERRORS
 * rpma_log_get_threshold() can fail with the following errors:
 * - RPMA_E_INVAL - threshold is not RPMA_LOG_THRESHOLD nor
 * RPMA_LOG_THRESHOLD_AUX
 * - RPMA_E_INVAL - *level is NULL
 *
 * SEE ALSO
 * rpma_log_set_function(3), rpma_log_set_threshold(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_log_get_threshold(enum rpma_log_threshold threshold,
				enum rpma_log_level *level);

/*
 * the type used for defining logging functions
 */
typedef void rpma_log_function(
	/* the log level of the message */
	enum rpma_log_level level,
	/* name of the source file where the message coming from */
	const char *file_name,
	/* the source file line where the message coming from */
	const int line_no,
	/* the function name where the message coming from */
	const char *function_name,
	/* printf(3)-like format string of the message */
	const char *message_format,
	/* additional arguments of the message format string */
	...);

#define RPMA_LOG_USE_DEFAULT_FUNCTION (NULL)

/** 3
 * rpma_log_set_function - set the logging function
 *
 * SYNOPSIS
 *
 *	#include <librpma.h>
 *
 *	typedef void rpma_log_function(
 *		enum rpma_log_level level,
 *		const char *file_name,
 *		const int line_no,
 *		const char *function_name,
 *		const char *message_format,
 *		...);
 *
 *	int rpma_log_set_function(rpma_log_function *log_function);
 *
 * DESCRIPTION
 * rpma_log_set_function() allows choosing the function which will get all
 * the generated logging messages. The log_function can be either
 * RPMA_LOG_USE_DEFAULT_FUNCTION which will use the default logging function
 * (built into the library) or a pointer to a user-defined function.
 *
 * Parameters of a user-defined log function are as follow:
 * - level - the log level of the message
 * - file_name - name of the source file where the message coming from.
 * It could be set to NULL and in such case neither line_no nor function_name
 * are provided.
 * - line_no - the source file line where the message coming from
 * - function_name - the function name where the message coming from
 * - message_format - printf(3)-like format string of the message
 * - "..." - additional arguments of the message format string
 *
 * THE DEFAULT LOGGING FUNCTION
 * The initial value of the logging function is RPMA_LOG_USE_DEFAULT_FUNCTION.
 * This function writes messages to syslog(3) (the primary destination) and to
 * stderr(3) (the secondary destination).
 *
 * RETURN VALUE
 * rpma_log_set_function() function returns 0 on success or error code
 * on failure.
 *
 * ERRORS
 * - RPMA_E_AGAIN - a temporary error occurred, the retry may fix the problem
 *
 * NOTE
 * The logging messages on the levels above the RPMA_LOG_THRESHOLD
 * level won't trigger the logging function.
 *
 * The user defined function must be thread-safe.
 *
 * SEE ALSO
 * rpma_log_get_threshold(3), rpma_log_set_threshold(3), librpma(7) and
 * https://pmem.io/rpma/
 */
int rpma_log_set_function(rpma_log_function *log_function);

#endif /* LIBRPMA_H */