path: root/libdm/libdevmapper.h

/*
 * Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
 * Copyright (C) 2004-2017 Red Hat, Inc. All rights reserved.
 * Copyright (C) 2006 Rackable Systems All rights reserved.
 *
 * This file is part of the device-mapper userspace tools.
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
 * of the GNU Lesser General Public License v.2.1.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 */

#ifndef LIB_DEVICE_MAPPER_H
#define LIB_DEVICE_MAPPER_H

#include <inttypes.h>
#include <stdarg.h>
#include <sys/types.h>
#include <sys/stat.h>

#ifdef __linux__
#  include <linux/types.h>
#endif

#include <limits.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#include <stddef.h> /* offsetof */

#ifndef __GNUC__
# define __typeof__ typeof
#endif

/* Macros to make string defines */
#define DM_TO_STRING_EXP(A) #A
#define DM_TO_STRING(A) DM_TO_STRING_EXP(A)

#define DM_ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))

#ifdef __cplusplus
extern "C" {
#endif

/*****************************************************************
 * The first section of this file provides direct access to the
 * individual device-mapper ioctls.  Since it is quite laborious to
 * build the ioctl arguments for the device-mapper, people are
 * encouraged to use this library.
 ****************************************************************/

/*
 * The library user may wish to register their own
 * logging function.  By default errors go to stderr.
 * Use dm_log_with_errno_init(NULL) to restore the default log fn.
 * Error messages may have a non-zero errno.
 * Debug messages may have a non-zero class.
 * Aborts on internal error when env DM_ABORT_ON_INTERNAL_ERRORS is 1
 */

typedef void (*dm_log_with_errno_fn) (int level, const char *file, int line,
				      int dm_errno_or_class, const char *f, ...)
    __attribute__ ((format(printf, 5, 6)));

void dm_log_with_errno_init(dm_log_with_errno_fn fn);
void dm_log_init_verbose(int level);

/*
 * Original version of this function.
 * dm_errno is set to 0.
 *
 * Deprecated: Use the _with_errno_ versions above instead.
 */
typedef void (*dm_log_fn) (int level, const char *file, int line,
			   const char *f, ...)
    __attribute__ ((format(printf, 4, 5)));

void dm_log_init(dm_log_fn fn);
/*
 * For backward-compatibility, indicate that dm_log_init() was used
 * to set a non-default value of dm_log().
 */
int dm_log_is_non_default(void);

/*
 * Number of devices currently in suspended state (via the library).
 */
int dm_get_suspended_counter(void);

enum {
	DM_DEVICE_CREATE,
	DM_DEVICE_RELOAD,
	DM_DEVICE_REMOVE,
	DM_DEVICE_REMOVE_ALL,

	DM_DEVICE_SUSPEND,
	DM_DEVICE_RESUME,

	DM_DEVICE_INFO,
	DM_DEVICE_DEPS,
	DM_DEVICE_RENAME,

	DM_DEVICE_VERSION,

	DM_DEVICE_STATUS,
	DM_DEVICE_TABLE,
	DM_DEVICE_WAITEVENT,

	DM_DEVICE_LIST,

	DM_DEVICE_CLEAR,

	DM_DEVICE_MKNODES,

	DM_DEVICE_LIST_VERSIONS,
	
	DM_DEVICE_TARGET_MSG,

	DM_DEVICE_SET_GEOMETRY,

	DM_DEVICE_ARM_POLL,

	DM_DEVICE_GET_TARGET_VERSION
};

/*
 * You will need to build a struct dm_task for
 * each ioctl command you want to execute.
 */

struct dm_pool;
struct dm_task;
struct dm_timestamp;

struct dm_task *dm_task_create(int type);
void dm_task_destroy(struct dm_task *dmt);

int dm_task_set_name(struct dm_task *dmt, const char *name);
int dm_task_set_uuid(struct dm_task *dmt, const char *uuid);

/*
 * Retrieve attributes after an info.
 */
struct dm_info {
	int exists;
	int suspended;
	int live_table;
	int inactive_table;
	int32_t open_count;
	uint32_t event_nr;
	uint32_t major;
	uint32_t minor;		/* minor device number */
	int read_only;		/* 0:read-write; 1:read-only */

	int32_t target_count;

	int deferred_remove;
	int internal_suspend;
};

struct dm_deps {
	uint32_t count;
	uint32_t filler;
	uint64_t device[];
};

struct dm_names {
	uint64_t dev;
	uint32_t next;		/* Offset to next struct from start of this struct */
	char name[];
};

struct dm_versions {
	uint32_t next;		/* Offset to next struct from start of this struct */
	uint32_t version[3];

	char name[];
};

int dm_get_library_version(char *version, size_t size);
int dm_task_get_driver_version(struct dm_task *dmt, char *version, size_t size);
int dm_task_get_info(struct dm_task *dmt, struct dm_info *dmi);

/*
 * This function returns dm device's UUID based on the value
 * of the mangling mode set during preceding dm_task_run call:
 *   - unmangled UUID for DM_STRING_MANGLING_{AUTO, HEX},
 *   - UUID without any changes for DM_STRING_MANGLING_NONE.
 *
 * To get mangled or unmangled form of the UUID directly, use
 * dm_task_get_uuid_mangled or dm_task_get_uuid_unmangled function.
 */
const char *dm_task_get_uuid(const struct dm_task *dmt);

struct dm_deps *dm_task_get_deps(struct dm_task *dmt);
struct dm_versions *dm_task_get_versions(struct dm_task *dmt);
const char *dm_task_get_message_response(struct dm_task *dmt);

/*
 * These functions return device-mapper names based on the value
 * of the mangling mode set during preceding dm_task_run call:
 *   - unmangled name for DM_STRING_MANGLING_{AUTO, HEX},
 *   - name without any changes for DM_STRING_MANGLING_NONE.
 *
 * To get mangled or unmangled form of the name directly, use
 * dm_task_get_name_mangled or dm_task_get_name_unmangled function.
 */
const char *dm_task_get_name(const struct dm_task *dmt);
struct dm_names *dm_task_get_names(struct dm_task *dmt);

int dm_task_set_ro(struct dm_task *dmt);
int dm_task_set_newname(struct dm_task *dmt, const char *newname);
int dm_task_set_newuuid(struct dm_task *dmt, const char *newuuid);
int dm_task_set_minor(struct dm_task *dmt, int minor);
int dm_task_set_major(struct dm_task *dmt, int major);
int dm_task_set_major_minor(struct dm_task *dmt, int major, int minor, int allow_default_major_fallback);
int dm_task_set_uid(struct dm_task *dmt, uid_t uid);
int dm_task_set_gid(struct dm_task *dmt, gid_t gid);
int dm_task_set_mode(struct dm_task *dmt, mode_t mode);
/* See also description for DM_UDEV_DISABLE_LIBRARY_FALLBACK flag! */
int dm_task_set_cookie(struct dm_task *dmt, uint32_t *cookie, uint16_t flags);
int dm_task_set_event_nr(struct dm_task *dmt, uint32_t event_nr);
int dm_task_set_geometry(struct dm_task *dmt, const char *cylinders, const char *heads, const char *sectors, const char *start);
int dm_task_set_message(struct dm_task *dmt, const char *message);
int dm_task_set_sector(struct dm_task *dmt, uint64_t sector);
int dm_task_no_flush(struct dm_task *dmt);
int dm_task_no_open_count(struct dm_task *dmt);
int dm_task_skip_lockfs(struct dm_task *dmt);
int dm_task_query_inactive_table(struct dm_task *dmt);
int dm_task_suppress_identical_reload(struct dm_task *dmt);
int dm_task_secure_data(struct dm_task *dmt);
int dm_task_retry_remove(struct dm_task *dmt);
int dm_task_deferred_remove(struct dm_task *dmt);
int dm_task_ima_measurement(struct dm_task *dmt);

/*
 * Record timestamp immediately after the ioctl returns.
 */
int dm_task_set_record_timestamp(struct dm_task *dmt);
struct dm_timestamp *dm_task_get_ioctl_timestamp(struct dm_task *dmt);

/*
 * Enable checks for common mistakes such as issuing ioctls in an unsafe order.
 */
int dm_task_enable_checks(struct dm_task *dmt);

typedef enum dm_add_node_e {
	DM_ADD_NODE_ON_RESUME, /* add /dev/mapper node with dmsetup resume */
	DM_ADD_NODE_ON_CREATE  /* add /dev/mapper node with dmsetup create */
} dm_add_node_t;
int dm_task_set_add_node(struct dm_task *dmt, dm_add_node_t add_node);

/*
 * Control read_ahead.
 */
#define DM_READ_AHEAD_AUTO UINT32_MAX	/* Use kernel default readahead */
#define DM_READ_AHEAD_NONE 0		/* Disable readahead */

#define DM_READ_AHEAD_MINIMUM_FLAG	0x1	/* Value supplied is minimum */

/*
 * Read ahead is set with DM_DEVICE_CREATE with a table or DM_DEVICE_RESUME.
 */
int dm_task_set_read_ahead(struct dm_task *dmt, uint32_t read_ahead,
			   uint32_t read_ahead_flags);
uint32_t dm_task_get_read_ahead(const struct dm_task *dmt,
				uint32_t *read_ahead);

/*
 * Use these to prepare for a create or reload.
 */
int dm_task_add_target(struct dm_task *dmt,
		       uint64_t start,
		       uint64_t size, const char *ttype, const char *params);

/*
 * Format major/minor numbers correctly for input to driver.
 */
#define DM_FORMAT_DEV_BUFSIZE	13	/* Minimum bufsize to handle worst case. */
int dm_format_dev(char *buf, int bufsize, uint32_t dev_major, uint32_t dev_minor);

/* Use this to retrive target information returned from a STATUS call */
void *dm_get_next_target(struct dm_task *dmt,
			 void *next, uint64_t *start, uint64_t *length,
			 char **target_type, char **params);

/*
 * Following dm_get_status_* functions will allocate approriate status structure
 * from passed mempool together with the necessary character arrays.
 * Destroying the mempool will release all asociated allocation.
 */

/* Parse params from STATUS call for mirror target */
typedef enum dm_status_mirror_health_e {
	DM_STATUS_MIRROR_ALIVE	      = 'A',/* No failures */
	DM_STATUS_MIRROR_FLUSH_FAILED = 'F',/* Mirror out-of-sync */
	DM_STATUS_MIRROR_WRITE_FAILED = 'D',/* Mirror out-of-sync */
	DM_STATUS_MIRROR_SYNC_FAILED  = 'S',/* Mirror out-of-sync */
	DM_STATUS_MIRROR_READ_FAILED  = 'R',/* Mirror data unaffected */
	DM_STATUS_MIRROR_UNCLASSIFIED = 'U' /* Bug */
} dm_status_mirror_health_t;

struct dm_status_mirror {
	uint64_t total_regions;
	uint64_t insync_regions;
	uint32_t dev_count;             /* # of devs[] elements (<= 8) */
	struct dm_dev_leg_health_s {
		dm_status_mirror_health_t health;
		uint32_t major;
		uint32_t minor;
	} *devs;                        /* array with individual legs */
	const char *log_type;           /* core, disk,.... */
	uint32_t log_count;		/* # of logs[] elements */
	struct dm_dev_log_health_s {
		dm_status_mirror_health_t health;
		uint32_t major;
		uint32_t minor;
	} *logs;			/* array with individual logs */
};

int dm_get_status_mirror(struct dm_pool *mem, const char *params,
			 struct dm_status_mirror **status);

/* Parse params from STATUS call for raid target */
struct dm_status_raid {
	uint64_t reserved;
	uint64_t total_regions;		/* sectors */
	uint64_t insync_regions;	/* sectors */
	uint64_t mismatch_count;
	uint32_t dev_count;
	char *raid_type;
	/* A - alive,  a - alive not in-sync,  D - dead/failed */
	char *dev_health;
	/* idle, frozen, resync, recover, check, repair */
	char *sync_action;
	uint64_t data_offset; /* RAID out-of-place reshaping */
};

int dm_get_status_raid(struct dm_pool *mem, const char *params,
		       struct dm_status_raid **status);

/* Parse params from STATUS call for cache target */
struct dm_status_cache {
	uint64_t version;  /* zero for now */

	uint32_t metadata_block_size;   /* in 512B sectors */
	uint32_t block_size;            /* AKA 'chunk_size' */

	uint64_t metadata_used_blocks;
	uint64_t metadata_total_blocks;

	uint64_t used_blocks;
	uint64_t dirty_blocks;
	uint64_t total_blocks;

	uint64_t read_hits;
	uint64_t read_misses;
	uint64_t write_hits;
	uint64_t write_misses;

	uint64_t demotions;
	uint64_t promotions;

	uint64_t feature_flags;		/* DM_CACHE_FEATURE_? */

	int core_argc;
	char **core_argv;

	char *policy_name;
	int policy_argc;
	char **policy_argv;

	unsigned error : 1;		/* detected error (switches to fail soon) */
	unsigned fail : 1;		/* all I/O fails */
	unsigned needs_check : 1;	/* metadata needs check */
	unsigned read_only : 1;		/* metadata may not be changed */
	uint32_t reserved : 28;
};

int dm_get_status_cache(struct dm_pool *mem, const char *params,
			struct dm_status_cache **status);

/*
 * Parse params from STATUS call for snapshot target
 *
 * Snapshot target's format:
 * <= 1.7.0: <used_sectors>/<total_sectors>
 * >= 1.8.0: <used_sectors>/<total_sectors> <metadata_sectors>
 */
struct dm_status_snapshot {
	uint64_t used_sectors;          /* in 512b units */
	uint64_t total_sectors;
	uint64_t metadata_sectors;
	unsigned has_metadata_sectors : 1; /* set when metadata_sectors is present */
	unsigned invalid : 1;		/* set when snapshot is invalidated */
	unsigned merge_failed : 1;	/* set when snapshot merge failed */
	unsigned overflow : 1;		/* set when snapshot overflows */
};

int dm_get_status_snapshot(struct dm_pool *mem, const char *params,
			   struct dm_status_snapshot **status);

/* Parse params from STATUS call for thin_pool target */
typedef enum dm_thin_discards_e {
	DM_THIN_DISCARDS_IGNORE,
	DM_THIN_DISCARDS_NO_PASSDOWN,
	DM_THIN_DISCARDS_PASSDOWN
} dm_thin_discards_t;

struct dm_status_thin_pool {
	uint64_t transaction_id;
	uint64_t used_metadata_blocks;
	uint64_t total_metadata_blocks;
	uint64_t used_data_blocks;
	uint64_t total_data_blocks;
	uint64_t held_metadata_root;
	uint32_t read_only;		/* metadata may not be changed */
	dm_thin_discards_t discards;
	uint32_t fail : 1;		/* all I/O fails */
	uint32_t error_if_no_space : 1;	/* otherwise queue_if_no_space */
	uint32_t out_of_data_space : 1;	/* metadata may be changed, but data may not be allocated (no rw) */
	uint32_t needs_check : 1;	/* metadata needs check */
	uint32_t error : 1;		/* detected error (switches to fail soon) */
	uint32_t reserved : 27;
};

int dm_get_status_thin_pool(struct dm_pool *mem, const char *params,
			    struct dm_status_thin_pool **status);

/* Parse params from STATUS call for thin target */
struct dm_status_thin {
	uint64_t mapped_sectors;
	uint64_t highest_mapped_sector;
	uint32_t fail : 1;              /* Thin volume fails I/O */
	uint32_t reserved : 31;
};

int dm_get_status_thin(struct dm_pool *mem, const char *params,
		       struct dm_status_thin **status);

/*
 * device-mapper statistics support
 */

/*
 * Statistics handle.
 *
 * Operations on dm_stats objects include managing statistics regions
 * and obtaining and manipulating current counter values from the
 * kernel. Methods are provided to return baisc count values and to
 * derive time-based metrics when a suitable interval estimate is
 * provided.
 *
 * Internally the dm_stats handle contains a pointer to a table of one
 * or more dm_stats_region objects representing the regions registered
 * with the dm_stats_create_region() method. These in turn point to a
 * table of one or more dm_stats_counters objects containing the
 * counter sets for each defined area within the region:
 *
 * dm_stats->dm_stats_region[nr_regions]->dm_stats_counters[nr_areas]
 *
 * This structure is private to the library and may change in future
 * versions: all users should make use of the public interface and treat
 * the dm_stats type as an opaque handle.
 *
 * Regions and counter sets are stored in order of increasing region_id.
 * Depending on region specifications and the sequence of create and
 * delete operations this may not correspond to increasing sector
 * number: users of the library should not assume that this is the case
 * unless region creation is deliberately managed to ensure this (by
 * always creating regions in strict order of ascending sector address).
 *
 * Regions may also overlap so the same sector range may be included in
 * more than one region or area: applications should be prepared to deal
 * with this or manage regions such that it does not occur.
 */
struct dm_stats;

/*
 * Histogram handle.
 *
 * A histogram object represents the latency histogram values and bin
 * boundaries of the histogram associated with a particular area.
 *
 * Operations on the handle allow the number of bins, bin boundaries,
 * counts and relative proportions to be obtained as well as the
 * conversion of a histogram or its bounds to a compact string
 * representation.
 */
struct dm_histogram;

/*
 * Allocate a dm_stats handle to use for subsequent device-mapper
 * statistics operations. A program_id may be specified and will be
 * used by default for subsequent operations on this handle.
 *
 * If program_id is NULL or the empty string a program_id will be
 * automatically set to the value contained in /proc/self/comm.
 */
struct dm_stats *dm_stats_create(const char *program_id);

/*
 * Bind a dm_stats handle to the specified device major and minor
 * values. Any previous binding is cleared and any preexisting counter
 * data contained in the handle is released.
 */
int dm_stats_bind_devno(struct dm_stats *dms, int major, int minor);

/*
 * Bind a dm_stats handle to the specified device name.
 * Any previous binding is cleared and any preexisting counter
 * data contained in the handle is released.
 */
int dm_stats_bind_name(struct dm_stats *dms, const char *name);

/*
 * Bind a dm_stats handle to the specified device UUID.
 * Any previous binding is cleared and any preexisting counter
 * data contained in the handle is released.
 */
int dm_stats_bind_uuid(struct dm_stats *dms, const char *uuid);

/*
 * Bind a dm_stats handle to the device backing the file referenced
 * by the specified file descriptor.
 *
 * File descriptor fd must reference a regular file, open for reading,
 * in a local file system, backed by a device-mapper device, that
 * supports the FIEMAP ioctl, and that returns data describing the
 * physical location of extents.
 */
int dm_stats_bind_from_fd(struct dm_stats *dms, int fd);
/*
 * Test whether the running kernel supports the precise_timestamps
 * feature. Presence of this feature also implies histogram support.
 * The library will check this call internally and fails any attempt
 * to use nanosecond counters or histograms on kernels that fail to
 * meet this check.
 */
int dm_message_supports_precise_timestamps(void);

/*
 * Precise timetamps and histogram support.
 * 
 * Test for the presence of precise_timestamps and histogram support.
 */
int dm_stats_driver_supports_precise(void);
int dm_stats_driver_supports_histogram(void);

/*
 * Returns 1 if the specified region has the precise_timestamps feature
 * enabled (i.e. produces nanosecond-precision counter values) or 0 for
 * a region using the default milisecond precision.
 */
int dm_stats_get_region_precise_timestamps(const struct dm_stats *dms,
					   uint64_t region_id);

/*
 * Returns 1 if the region at the current cursor location has the
 * precise_timestamps feature enabled (i.e. produces
 * nanosecond-precision counter values) or 0 for a region using the
 * default milisecond precision.
 */
int dm_stats_get_current_region_precise_timestamps(const struct dm_stats *dms);

#define DM_STATS_ALL_PROGRAMS ""
/*
 * Parse the response from a @stats_list message. dm_stats_list will
 * allocate the necessary dm_stats and dm_stats region structures from
 * the embedded dm_pool. No counter data will be obtained (the counters
 * members of dm_stats_region objects are set to NULL).
 *
 * A program_id may optionally be supplied; if the argument is non-NULL
 * only regions with a matching program_id value will be considered. If
 * the argument is NULL then the default program_id associated with the
 * dm_stats handle will be used. Passing the special value
 * DM_STATS_ALL_PROGRAMS will cause all regions to be queried
 * regardless of region program_id.
 */
int dm_stats_list(struct dm_stats *dms, const char *program_id);

#define DM_STATS_REGIONS_ALL UINT64_MAX
/*
 * Populate a dm_stats object with statistics for one or more regions of
 * the specified device.
 *
 * A program_id may optionally be supplied; if the argument is non-NULL
 * only regions with a matching program_id value will be considered. If
 * the argument is NULL then the default program_id associated with the
 * dm_stats handle will be used. Passing the special value
 * DM_STATS_ALL_PROGRAMS will cause all regions to be queried
 * regardless of region program_id.
 *
 * Passing the special value DM_STATS_REGIONS_ALL as the region_id
 * argument will attempt to retrieve all regions selected by the
 * program_id argument.
 *
 * If region_id is used to request a single region_id to be populated
 * the program_id is ignored.
 */
int dm_stats_populate(struct dm_stats *dms, const char *program_id,
		      uint64_t region_id);

/*
 * Create a new statistics region on the device bound to dms.
 *
 * start and len specify the region start and length in 512b sectors.
 * Passing zero for both start and len will create a region spanning
 * the entire device.
 *
 * Step determines how to subdivide the region into discrete counter
 * sets: a positive value specifies the size of areas into which the
 * region should be split while a negative value will split the region
 * into a number of areas equal to the absolute value of step:
 *
 * - a region with one area spanning the entire device:
 *
 *   dm_stats_create_region(dms, 0, 0, -1, p, a);
 *
 * - a region with areas of 1MiB:
 *
 *   dm_stats_create_region(dms, 0, 0, 1 << 11, p, a);
 *
 * - one 1MiB region starting at 1024 sectors with two areas:
 *
 *   dm_stats_create_region(dms, 1024, 1 << 11, -2, p, a);
 *
 * If precise is non-zero attempt to create a region with nanosecond
 * precision counters using the kernel precise_timestamps feature.
 *
 * precise - A flag to request nanosecond precision counters
 * to be used for this region.
 *
 * histogram_bounds - specify the boundaries of a latency histogram to
 * be tracked for the region. The values are expressed as an array of
 * uint64_t terminated with a zero. Values must be in order of ascending
 * magnitude and specify the upper bounds of successive histogram bins
 * in nanoseconds (with an implicit lower bound of zero on the first bin
 * and an implicit upper bound of infinity on the final bin). For
 * example:
 *
 *   uint64_t bounds_ary[] = { 1000, 2000, 3000, 0 };
 *
 * Specifies a histogram with four bins: 0-1000ns, 1000-2000ns,
 * 2000-3000ns and >3000ns.
 *
 * The smallest latency value that can be tracked for a region not using
 * precise_timestamps is 1ms: attempting to create a region with
 * histogram boundaries < 1ms will cause the precise_timestamps feature
 * to be enabled for that region automatically if it was not requested
 * explicitly.
 *
 * program_id is an optional string argument that identifies the
 * program creating the region. If program_id is NULL or the empty
 * string the default program_id stored in the handle will be used.
 *
 * user_data is an optional string argument that is added to the
 * content of the aux_data field stored with the statistics region by
 * the kernel.
 *
 * The library may also use this space internally, for example, to
 * store a group descriptor or other metadata: in this case the
 * library will strip any internal data fields from the value before
 * it is returned via a call to dm_stats_get_region_aux_data().
 *
 * The user data stored is not accessed by the library or kernel and
 * may be used to store an arbitrary data word (embedded whitespace is
 * not permitted).
 *
 * An application using both the library and direct access to the
 * @stats_list device-mapper message may see the internal values stored
 * in this field by the library. In such cases any string up to and
 * including the first '#' in the field must be treated as an opaque
 * value and preserved across any external modification of aux_data.
 *
 * The region_id of the newly-created region is returned in *region_id
 * if it is non-NULL.
 */
int dm_stats_create_region(struct dm_stats *dms, uint64_t *region_id,
			   uint64_t start, uint64_t len, int64_t step,
			   int precise, struct dm_histogram *bounds,
			   const char *program_id, const char *user_data);

/*
 * Delete the specified statistics region. This will also mark the
 * region as not-present and discard any existing statistics data.
 */
int dm_stats_delete_region(struct dm_stats *dms, uint64_t region_id);

/*
 * Clear the specified statistics region. This requests the kernel to
 * zero all counter values (except in-flight I/O). Note that this
 * operation is not atomic with respect to reads of the counters; any IO
 * events occurring between the last print operation and the clear will
 * be lost. This can be avoided by using the atomic print-and-clear
 * function of the dm_stats_print_region() call or by using the higher
 * level dm_stats_populate() interface.
 */
int dm_stats_clear_region(struct dm_stats *dms, uint64_t region_id);

/*
 * Print the current counter values for the specified statistics region
 * and return them as a string. The memory for the string buffer will
 * be allocated from the dm_stats handle's private pool and should be
 * returned by calling dm_stats_buffer_destroy() when no longer
 * required. The pointer will become invalid following any call that
 * clears or reinitializes the handle (destroy, list, populate, bind).
 *
 * This allows applications that wish to access the raw message response
 * to obtain it via a dm_stats handle; no parsing of the textual counter
 * data is carried out by this function.
 *
 * Most users are recommended to use the dm_stats_populate() call
 * instead since this will automatically parse the statistics data into
 * numeric form accessible via the dm_stats_get_*() counter access
 * methods.
 *
 * A subset of the data lines may be requested by setting the
 * start_line and num_lines parameters. If both are zero all data
 * lines are returned.
 *
 * If the clear parameter is non-zero the operation will also
 * atomically reset all counter values to zero (except in-flight IO).
 */
char *dm_stats_print_region(struct dm_stats *dms, uint64_t region_id,
			    unsigned start_line, unsigned num_lines,
			    unsigned clear);

/*
 * Destroy a statistics response buffer obtained from a call to
 * dm_stats_print_region().
 */
void dm_stats_buffer_destroy(struct dm_stats *dms, char *buffer);

/*
 * Determine the number of regions contained in a dm_stats handle
 * following a dm_stats_list() or dm_stats_populate() call.
 *
 * The value returned is the number of registered regions visible with the
 * progam_id value used for the list or populate operation and may not be
 * equal to the highest present region_id (either due to program_id
 * filtering or gaps in the sequence of region_id values).
 *
 * Always returns zero on an empty handle.
 */
uint64_t dm_stats_get_nr_regions(const struct dm_stats *dms);

/*
 * Determine the number of groups contained in a dm_stats handle
 * following a dm_stats_list() or dm_stats_populate() call.
 *
 * The value returned is the number of registered groups visible with the
 * progam_id value used for the list or populate operation and may not be
 * equal to the highest present group_id (either due to program_id
 * filtering or gaps in the sequence of group_id values).
 *
 * Always returns zero on an empty handle.
 */
uint64_t dm_stats_get_nr_groups(const struct dm_stats *dms);

/*
 * Test whether region_id is present in this dm_stats handle.
 */
int dm_stats_region_present(const struct dm_stats *dms, uint64_t region_id);

/*
 * Returns the number of areas (counter sets) contained in the specified
 * region_id of the supplied dm_stats handle.
 */
uint64_t dm_stats_get_region_nr_areas(const struct dm_stats *dms,
				      uint64_t region_id);

/*
 * Returns the total number of areas (counter sets) in all regions of the
 * given dm_stats object.
 */
uint64_t dm_stats_get_nr_areas(const struct dm_stats *dms);

/*
 * Test whether group_id is present in this dm_stats handle.
 */
int dm_stats_group_present(const struct dm_stats *dms, uint64_t group_id);

/*
 * Return the number of bins in the histogram configuration for the
 * specified region or zero if no histogram specification is configured.
 * Valid following a dm_stats_list() or dm_stats_populate() operation.
 */
int dm_stats_get_region_nr_histogram_bins(const struct dm_stats *dms,
					  uint64_t region_id);

/*
 * Parse a histogram string with optional unit suffixes into a
 * dm_histogram bounds description.
 *
 * A histogram string is a string of numbers "n1,n2,n3,..." that
 * represent the boundaries of a histogram. The first and final bins
 * have implicit lower and upper bounds of zero and infinity
 * respectively and boundary values must occur in order of ascending
 * magnitude.  Unless a unit suffix is given all values are specified in
 * nanoseconds.
 *
 * For example, if bounds_str="300,600,900", the region will be created
 * with a histogram containing four bins. Each report will include four
 * numbers a:b:c:d. a is the number of requests that took between 0 and
 * 300ns to complete, b is the number of requests that took 300-600ns to
 * complete, c is the number of requests that took 600-900ns to complete
 * and d is the number of requests that took more than 900ns to
 * complete.
 *
 * An optional unit suffix of 's', 'ms', 'us', or 'ns' may be used to
 * specify units of seconds, miliseconds, microseconds, or nanoseconds:
 *
 *   bounds_str="1ns,1us,1ms,1s"
 *   bounds_str="500us,1ms,1500us,2ms"
 *   bounds_str="200ms,400ms,600ms,800ms,1s"
 *
 * The smallest valid unit of time for a histogram specification depends
 * on whether the region uses precise timestamps: for a region with the
 * default milisecond precision the smallest possible histogram boundary
 * magnitude is one milisecond: attempting to use a histogram with a
 * boundary less than one milisecond when creating a region will cause
 * the region to be created with the precise_timestamps feature enabled.
 *
 * On sucess a pointer to the struct dm_histogram representing the
 * bounds values is returned, or NULL in the case of error. The returned
 * pointer should be freed using dm_free() when no longer required.
 */
struct dm_histogram *dm_histogram_bounds_from_string(const char *bounds_str);

/*
 * Parse a zero terminated array of uint64_t into a dm_histogram bounds
 * description.
 *
 * Each value in the array specifies the upper bound of a bin in the
 * latency histogram in nanoseconds. Values must appear in ascending
 * order of magnitude.
 *
 * The smallest valid unit of time for a histogram specification depends
 * on whether the region uses precise timestamps: for a region with the
 * default milisecond precision the smallest possible histogram boundary
 * magnitude is one milisecond: attempting to use a histogram with a
 * boundary less than one milisecond when creating a region will cause
 * the region to be created with the precise_timestamps feature enabled.
 */
struct dm_histogram *dm_histogram_bounds_from_uint64(const uint64_t *bounds);

/*
 * Destroy the histogram bounds array obtained from a call to
 * dm_histogram_bounds_from_string().
 */
void dm_histogram_bounds_destroy(struct dm_histogram *bounds);

/*
 * Destroy a dm_stats object and all associated regions, counter
 * sets and histograms.
 */
void dm_stats_destroy(struct dm_stats *dms);

/*
 * Counter sampling interval
 */

/*
 * Set the sampling interval for counter data to the specified value in
 * either nanoseconds or milliseconds.
 *
 * The interval is used to calculate time-based metrics from the basic
 * counter data: an interval must be set before calling any of the
 * metric methods.
 *
 * For best accuracy the duration should be measured and updated at the
 * end of each interval.
 *
 * All values are stored internally with nanosecond precision and are
 * converted to or from ms when the millisecond interfaces are used.
 */
void dm_stats_set_sampling_interval_ns(struct dm_stats *dms,
				       uint64_t interval_ns);

void dm_stats_set_sampling_interval_ms(struct dm_stats *dms,
				       uint64_t interval_ms);

/*
 * Retrieve the configured sampling interval in either nanoseconds or
 * milliseconds.
 */
uint64_t dm_stats_get_sampling_interval_ns(const struct dm_stats *dms);
uint64_t dm_stats_get_sampling_interval_ms(const struct dm_stats *dms);

/*
 * Override program_id. This may be used to change the default
 * program_id value for an existing handle. If the allow_empty argument
 * is non-zero a NULL or empty program_id is permitted.
 *
 * Use with caution! Most users of the library should set a valid,
 * non-NULL program_id for every statistics region created. Failing to
 * do so may result in confusing state when multiple programs are
 * creating and managing statistics regions.
 *
 * All users of the library are encouraged to choose an unambiguous,
 * unique program_id: this could be based on PID (for programs that
 * create, report, and delete regions in a single process), session id,
 * executable name, or some other distinguishing string.
 *
 * Use of the empty string as a program_id does not simplify use of the
 * library or the command line tools and use of this value is strongly
 * discouraged.
 */
int dm_stats_set_program_id(struct dm_stats *dms, int allow_empty,
			    const char *program_id);

/*
 * Region properties: size, length & area_len.
 *
 * Region start and length are returned in units of 512b as specified
 * at region creation time. The area_len value gives the size of areas
 * into which the region has been subdivided. For regions with a single
 * area spanning the range this value is equal to the region length.
 *
 * For regions created with a specified number of areas the value
 * represents the size of the areas into which the kernel divided the
 * region excluding any rounding of the last area size. The number of
 * areas may be obtained using the dm_stats_nr_areas_region() call.
 *
 * All values are returned in units of 512b sectors.
 */
int dm_stats_get_region_start(const struct dm_stats *dms, uint64_t *start,
			      uint64_t region_id);

int dm_stats_get_region_len(const struct dm_stats *dms, uint64_t *len,
			    uint64_t region_id);

int dm_stats_get_region_area_len(const struct dm_stats *dms,
				 uint64_t *len, uint64_t region_id);

/*
 * Area properties: start, offset and length.
 *
 * The area length is always equal to the area length of the region
 * that contains it and is obtained from dm_stats_get_region_area_len().
 *
 * The start of an area is a function of the area_id and the containing
 * region's start and area length: it gives the absolute offset into the
 * containing device of the beginning of the area.
 *
 * The offset expresses the area's relative offset into the current
 * region. I.e. the area start minus the start offset of the containing
 * region.
 *
 * All values are returned in units of 512b sectors.
 */
int dm_stats_get_area_start(const struct dm_stats *dms, uint64_t *start,
			    uint64_t region_id, uint64_t area_id);

int dm_stats_get_area_offset(const struct dm_stats *dms, uint64_t *offset,
			     uint64_t region_id, uint64_t area_id);

/*
 * Retrieve program_id and user aux_data for a specific region.
 *
 * Only valid following a call to dm_stats_list().
 */

/*
 * Retrieve program_id for the specified region.
 *
 * The returned pointer does not need to be freed separately from the
 * dm_stats handle but will become invalid after a dm_stats_destroy(),
 * dm_stats_list(), dm_stats_populate(), or dm_stats_bind*() of the
 * handle from which it was obtained.
 */
const char *dm_stats_get_region_program_id(const struct dm_stats *dms,
					   uint64_t region_id);

/*
 * Retrieve user aux_data set for the specified region. This function
 * will return any stored user aux_data as a string in the memory
 * pointed to by the aux_data argument.
 *
 * Any library internal aux_data fields, such as DMS_GROUP descriptors,
 * are stripped before the value is returned.
 *
 * The returned pointer does not need to be freed separately from the
 * dm_stats handle but will become invalid after a dm_stats_destroy(),
 * dm_stats_list(), dm_stats_populate(), or dm_stats_bind*() of the
 * handle from which it was obtained.
 */
const char *dm_stats_get_region_aux_data(const struct dm_stats *dms,
					 uint64_t region_id);

typedef enum dm_stats_obj_type_e {
	DM_STATS_OBJECT_TYPE_NONE,
	DM_STATS_OBJECT_TYPE_AREA,
	DM_STATS_OBJECT_TYPE_REGION,
	DM_STATS_OBJECT_TYPE_GROUP
} dm_stats_obj_type_t;

/*
 * Statistics cursor
 *
 * A dm_stats handle maintains an optional cursor into the statistics
 * tables that it stores. Iterators are provided to visit each region,
 * area, or group in a handle and accessor methods are provided to
 * obtain properties and values for the object at the current cursor
 * position.
 *
 * Using the cursor simplifies walking all regions or groups when
 * the tables are sparse (i.e. contains some present and some
 * non-present region_id or group_id values either due to program_id
 * filtering or the ordering of region and group creation and deletion).
 *
 * Simple macros are provided to visit each area, region, or group,
 * contained in a handle and applications are encouraged to use these
 * where possible.
 */

/*
 * Walk flags are used to initialise a dm_stats handle's cursor control
 * and to select region or group aggregation when calling a metric or
 * counter property method with immediate group, region, and area ID
 * values.
 *
 * Walk flags are stored in the uppermost word of a uint64_t so that
 * a region_id or group_id may be encoded in the lower bits. This
 * allows an aggregate region_id or group_id to be specified when
 * retrieving counter or metric values.
 *
 * Flags may be ORred together when used to initialise a dm_stats_walk:
 * the resulting walk will visit instance of each type specified by
 * the flag combination.
 */
#define DM_STATS_WALK_AREA   0x1000000000000ULL
#define DM_STATS_WALK_REGION 0x2000000000000ULL
#define DM_STATS_WALK_GROUP  0x4000000000000ULL

#define DM_STATS_WALK_ALL    0x7000000000000ULL
#define DM_STATS_WALK_DEFAULT (DM_STATS_WALK_AREA | DM_STATS_WALK_REGION)

/*
 * Skip regions from a DM_STATS_WALK_REGION that contain only a single
 * area: in this case the region's aggregate values are identical to
 * the values of the single contained area. Setting this flag will
 * suppress these duplicate entries during a dm_stats_walk_* with the
 * DM_STATS_WALK_REGION flag set.
 */
#define DM_STATS_WALK_SKIP_SINGLE_AREA   0x8000000000000ULL

/*
 * Initialise the cursor control of a dm_stats handle for the specified
 * walk type(s). Including a walk flag in the flags argument will cause
 * any subsequent walk to visit that type of object (until the next
 * call to dm_stats_walk_init()).
 */
int dm_stats_walk_init(struct dm_stats *dms, uint64_t flags);

/*
 * Set the cursor of a dm_stats handle to address the first present
 * group, region, or area of the currently configured walk. It is
 * valid to attempt to walk a NULL stats handle or a handle containing
 * no present regions; in this case any call to dm_stats_walk_next()
 * becomes a no-op and all calls to dm_stats_walk_end() return true.
 */
void dm_stats_walk_start(struct dm_stats *dms);

/*
 * Advance the statistics cursor to the next area, or to the next
 * present region if at the end of the current region. If the end of
 * the region, area, or group tables is reached a subsequent call to
 * dm_stats_walk_end() will return 1 and dm_stats_object_type() called
 * on the location will return DM_STATS_OBJECT_TYPE_NONE,
 */
void dm_stats_walk_next(struct dm_stats *dms);

/*
 * Force the statistics cursor to advance to the next region. This will
 * stop any in-progress area walk (by clearing DM_STATS_WALK_AREA) and
 * advance the cursor to the next present region, the first present
 * group (if DM_STATS_GROUP_WALK is set), or to the end. In this case a
 * subsequent call to dm_stats_walk_end() will return 1 and a call to
 * dm_stats_object_type() for the location will return
 * DM_STATS_OBJECT_TYPE_NONE.
 */
void dm_stats_walk_next_region(struct dm_stats *dms);

/*
 * Test whether the end of a statistics walk has been reached.
 */
int dm_stats_walk_end(struct dm_stats *dms);

/*
 * Return the type of object at the location specified by region_id
 * and area_id. If either region_id or area_id uses one of the special
 * values DM_STATS_REGION_CURRENT or DM_STATS_AREA_CURRENT the
 * corresponding region or area identifier will be taken from the
 * current cursor location. If the cursor location or the value encoded
 * by region_id and area_id indicates an aggregate region or group,
 * this will be reflected in the value returned.
 */
dm_stats_obj_type_t dm_stats_object_type(const struct dm_stats *dms,
					 uint64_t region_id,
					 uint64_t area_id);

/*
 * Return the type of object at the current stats cursor location.
 */
dm_stats_obj_type_t dm_stats_current_object_type(const struct dm_stats *dms);

/*
 * Stats iterators
 *
 * C 'for' and 'do'/'while' style iterators for dm_stats data.
 *
 * It is not safe to call any function that modifies the region table
 * within the loop body (i.e. dm_stats_list(), dm_stats_populate(),
 * dm_stats_init(), or dm_stats_destroy()).
 *
 * All counter and property (dm_stats_get_*) access methods, as well as
 * dm_stats_populate_region() can be safely called from loops.
 *
 */

/*
 * Iterate over the regions table visiting each region.
 *
 * If the region table is empty or unpopulated the loop body will not be
 * executed.
 */
#define dm_stats_foreach_region(dms)				\
for (dm_stats_walk_init((dms), DM_STATS_WALK_REGION),		\
     dm_stats_walk_start((dms));				\
     !dm_stats_walk_end((dms)); dm_stats_walk_next_region((dms)))

/*
 * Iterate over the regions table visiting each area.
 *
 * If the region table is empty or unpopulated the loop body will not
 * be executed.
 */
#define dm_stats_foreach_area(dms)				\
for (dm_stats_walk_init((dms), DM_STATS_WALK_AREA),		\
     dm_stats_walk_start((dms));				\
     !dm_stats_walk_end((dms)); dm_stats_walk_next((dms)))

/*
 * Iterate over the regions table visiting each group. Metric and
 * counter methods will return values for the group.
 *
 * If the group table is empty or unpopulated the loop body will not
 * be executed.
 */
#define dm_stats_foreach_group(dms)				\
for (dm_stats_walk_init((dms), DM_STATS_WALK_GROUP),		\
     dm_stats_walk_start(dms);					\
     !dm_stats_walk_end(dms);					\
     dm_stats_walk_next(dms))

/*
 * Start a walk iterating over the regions contained in dm_stats handle
 * 'dms'.
 *
 * The body of the loop should call dm_stats_walk_next() or
 * dm_stats_walk_next_region() to advance to the next element.
 *
 * The loop body is executed at least once even if the stats handle is
 * empty.
 */
#define dm_stats_walk_do(dms)					\
do {								\
	dm_stats_walk_start((dms));				\
	do

/*
 * Start a 'while' style loop or end a 'do..while' loop iterating over the
 * regions contained in dm_stats handle 'dms'.
 */
#define dm_stats_walk_while(dms)				\
	while(!dm_stats_walk_end((dms)));			\
} while (0)

/*
 * Cursor relative property methods
 *
 * Calls with the prefix dm_stats_get_current_* operate relative to the
 * current cursor location, returning properties for the current region
 * or area of the supplied dm_stats handle.
 *
 */

/*
 * Returns the number of areas (counter sets) contained in the current
 * region of the supplied dm_stats handle.
 */
uint64_t dm_stats_get_current_nr_areas(const struct dm_stats *dms);

/*
 * Retrieve the current values of the stats cursor.
 */
uint64_t dm_stats_get_current_region(const struct dm_stats *dms);
uint64_t dm_stats_get_current_area(const struct dm_stats *dms);

/*
 * Current region properties: size, length & area_len.
 *
 * See the comments for the equivalent dm_stats_get_* versions for a
 * complete description of these methods.
 *
 * All values are returned in units of 512b sectors.
 */
int dm_stats_get_current_region_start(const struct dm_stats *dms,
				      uint64_t *start);

int dm_stats_get_current_region_len(const struct dm_stats *dms,
				    uint64_t *len);

int dm_stats_get_current_region_area_len(const struct dm_stats *dms,
					 uint64_t *area_len);

/*
 * Current area properties: start and length.
 *
 * See the comments for the equivalent dm_stats_get_* versions for a
 * complete description of these methods.
 *
 * All values are returned in units of 512b sectors.
 */
int dm_stats_get_current_area_start(const struct dm_stats *dms,
				    uint64_t *start);

int dm_stats_get_current_area_offset(const struct dm_stats *dms,
				     uint64_t *offset);

int dm_stats_get_current_area_len(const struct dm_stats *dms,
				       uint64_t *start);

/*
 * Return a pointer to the program_id string for region at the current
 * cursor location.
 */
const char *dm_stats_get_current_region_program_id(const struct dm_stats *dms);

/*
 * Return a pointer to the user aux_data string for the region at the
 * current cursor location.
 */
const char *dm_stats_get_current_region_aux_data(const struct dm_stats *dms);

/*
 * Statistics groups and data aggregation.
 */

/*
 * Create a new group in stats handle dms from the group descriptor
 * passed in group. The group descriptor is a string containing a list
 * of region_id values that will be included in the group. The first
 * region_id found will be the group leader. Ranges of identifiers may
 * be expressed as "M-N", where M and N are the start and end region_id
 * values for the range.
 */
int dm_stats_create_group(struct dm_stats *dms, const char *group,
			  const char *alias, uint64_t *group_id);

/*
 * Remove the specified group_id. If the remove argument is zero the
 * group will be removed but the regions that it contained will remain.
 * If remove is non-zero then all regions that belong to the group will
 * also be removed.
 */
int dm_stats_delete_group(struct dm_stats *dms, uint64_t group_id, int remove);

/*
 * Set an alias for this group or region. The alias will be returned
 * instead of the normal dm-stats name for this region or group.
 */
int dm_stats_set_alias(struct dm_stats *dms, uint64_t group_id,
		       const char *alias);

/*
 * Returns a pointer to the currently configured alias for id, or the
 * name of the dm device the handle is bound to if no alias has been
 * set. The pointer will be freed automatically when a new alias is set
 * or when the stats handle is cleared.
 */
const char *dm_stats_get_alias(const struct dm_stats *dms, uint64_t id);

#define DM_STATS_GROUP_NONE UINT64_MAX
/*
 * Return the group_id that the specified region_id belongs to, or the
 * special value DM_STATS_GROUP_NONE if the region does not belong
 * to any group.
 */
uint64_t dm_stats_get_group_id(const struct dm_stats *dms, uint64_t region_id);

/*
 * Store a pointer to a string describing the regions that are members
 * of the group specified by group_id in the memory pointed to by buf.
 * The string is in the same format as the 'group' argument to
 * dm_stats_create_group().
 *
 * The pointer does not need to be freed explicitly by the caller: it
 * will become invalid following a subsequent dm_stats_list(),
 * dm_stats_populate() or dm_stats_destroy() of the corresponding
 * dm_stats handle.
 */
int dm_stats_get_group_descriptor(const struct dm_stats *dms,
				  uint64_t group_id, char **buf);

/*
 * Create regions that correspond to the extents of a file in the
 * filesystem and optionally place them into a group.
 *
 * File descriptor fd must reference a regular file, open for reading,
 * in a local file system that supports the FIEMAP ioctl, and that
 * returns data describing the physical location of extents.
 *
 * The file descriptor can be closed by the caller following the call
 * to dm_stats_create_regions_from_fd().
 *
 * Unless nogroup is non-zero the regions will be placed into a group
 * and the group alias set to the value supplied (if alias is NULL no
 * group alias will be assigned).
 *
 * On success the function returns a pointer to an array of uint64_t
 * containing the IDs of the newly created regions. The region_id
 * array is terminated by the value DM_STATS_REGION_NOT_PRESENT and
 * should be freed using dm_free() when no longer required.
 *
 * On error NULL is returned.
 *
 * Following a call to dm_stats_create_regions_from_fd() the handle
 * is guaranteed to be in a listed state, and to contain any region
 * and group identifiers created by the operation.
 *
 * The group_id for the new group is equal to the region_id value in
 * the first array element.
 */
uint64_t *dm_stats_create_regions_from_fd(struct dm_stats *dms, int fd,
					  int group, int precise,
					  struct dm_histogram *bounds,
					  const char *alias);
/*
 * Update a group of regions that correspond to the extents of a file
 * in the filesystem, adding and removing regions to account for
 * allocation changes in the underlying file.
 *
 * File descriptor fd must reference a regular file, open for reading,
 * in a local file system that supports the FIEMAP ioctl, and that
 * returns data describing the physical location of extents.
 *
 * The file descriptor can be closed by the caller following the call
 * to dm_stats_update_regions_from_fd().
 *
 * On success the function returns a pointer to an array of uint64_t
 * containing the IDs of the updated regions (including any existing
 * regions that were not modified by the call).
 *
 * The region_id array is terminated by the special value
 * DM_STATS_REGION_NOT_PRESENT and should be freed using dm_free()
 * when no longer required.
 *
 * On error NULL is returned.
 *
 * Following a call to dm_stats_update_regions_from_fd() the handle
 * is guaranteed to be in a listed state, and to contain any region
 * and group identifiers created by the operation.
 *
 * This function cannot be used with file mapped regions that are
 * not members of a group: either group the regions, or remove them
 * and re-map them with dm_stats_create_regions_from_fd().
 */
uint64_t *dm_stats_update_regions_from_fd(struct dm_stats *dms, int fd,
					  uint64_t group_id);


/*
 * The file map monitoring daemon can monitor files in two distinct
 * ways: the mode affects the behaviour of the daemon when a file
 * under monitoring is renamed or unlinked, and the conditions which
 * cause the daemon to terminate.
 *
 * In both modes, the daemon will always shut down when the group
 * being monitored is deleted.
 *
 * Follow inode:
 * The daemon follows the inode of the file, as it was at the time the
 * daemon started. The file descriptor referencing the file is kept
 * open at all times, and the daemon will exit when it detects that
 * the file has been unlinked and it is the last holder of a reference
 * to the file.
 *
 * This mode is useful if the file is expected to be renamed, or moved
 * within the file system, while it is being monitored.
 *
 * Follow path:
 * The daemon follows the path that was given on the daemon command
 * line. The file descriptor referencing the file is re-opened on each
 * iteration of the daemon, and the daemon will exit if no file exists
 * at this location (a tolerance is allowed so that a brief delay
 * between unlink() and creat() is permitted).
 *
 * This mode is useful if the file is updated by unlinking the original
 * and placing a new file at the same path.
 */

typedef enum dm_filemapd_mode_e {
	DM_FILEMAPD_FOLLOW_INODE,
	DM_FILEMAPD_FOLLOW_PATH,
	DM_FILEMAPD_FOLLOW_NONE
} dm_filemapd_mode_t;

/*
 * Parse a string representation of a dmfilemapd mode.
 *
 * Returns a valid dm_filemapd_mode_t value on success, or
 * DM_FILEMAPD_FOLLOW_NONE on error.
 */
dm_filemapd_mode_t dm_filemapd_mode_from_string(const char *mode_str);

/*
 * Start the dmfilemapd filemap monitoring daemon for the specified
 * file descriptor, group, and file system path. The daemon will
 * monitor the file for allocation changes, and when a change is
 * detected, call dm_stats_update_regions_from_fd() to update the
 * mapped regions for the file.
 *
 * The path provided to dm_stats_start_filemapd() must be an absolute
 * path, and should reflect the path of 'fd' at the time that it was
 * opened.
 *
 * The mode parameter controls the behaviour of the daemon when the
 * file being monitored is unlinked or moved: see the comments for
 * dm_filemapd_mode_t for a full description and possible values.
 *
 * The daemon can be stopped at any time by sending SIGTERM to the
 * daemon pid.
 */
int dm_stats_start_filemapd(int fd, uint64_t group_id, const char *path,
			    dm_filemapd_mode_t mode, unsigned foreground,
			    unsigned verbose);

/*
 * Call this to actually run the ioctl.
 */
int dm_task_run(struct dm_task *dmt);

/*
 * The errno from the last device-mapper ioctl performed by dm_task_run.
 */
int dm_task_get_errno(struct dm_task *dmt);

/*
 * Call this to make or remove the device nodes associated with previously
 * issued commands.
 */
void dm_task_update_nodes(void);

/*
 * Mangling support
 *
 * Character whitelist: 0-9, A-Z, a-z, #+-.:=@_
 * HEX mangling format: \xNN, NN being the hex value of the character.
 * (whitelist and format supported by udev)
*/
typedef enum dm_string_mangling_e {
	DM_STRING_MANGLING_NONE, /* do not mangle at all */
	DM_STRING_MANGLING_AUTO, /* mangle only if not already mangled with hex, error when mixed */
	DM_STRING_MANGLING_HEX	 /* always mangle with hex encoding, no matter what the input is */
} dm_string_mangling_t;

/*
 * Set/get mangling mode used for device-mapper names and uuids.
 */
int dm_set_name_mangling_mode(dm_string_mangling_t name_mangling);
dm_string_mangling_t dm_get_name_mangling_mode(void);

/*
 * Get mangled/unmangled form of the device-mapper name or uuid
 * irrespective of the global setting (set by dm_set_name_mangling_mode).
 * The name or uuid returned needs to be freed after use by calling dm_free!
 */
char *dm_task_get_name_mangled(const struct dm_task *dmt);
char *dm_task_get_name_unmangled(const struct dm_task *dmt);
char *dm_task_get_uuid_mangled(const struct dm_task *dmt);
char *dm_task_get_uuid_unmangled(const struct dm_task *dmt);

/*
 * Configure the device-mapper directory
 */
int dm_set_dev_dir(const char *dir);
const char *dm_dir(void);

/*
 * Configure sysfs directory, /sys by default
 */
int dm_set_sysfs_dir(const char *dir);
const char *dm_sysfs_dir(void);

/*
 * Configure default UUID prefix string.
 * Conventionally this is a short capitalised prefix indicating the subsystem
 * that is managing the devices, e.g. "LVM-" or "MPATH-".
 * To support stacks of devices from different subsystems, recursive functions
 * stop recursing if they reach a device with a different prefix.
 */
int dm_set_uuid_prefix(const char *uuid_prefix);
const char *dm_uuid_prefix(void);

/*
 * Determine whether a major number belongs to device-mapper or not.
 */
int dm_is_dm_major(uint32_t major);

/*
 * Get associated device name for given major and minor number by reading
 * the sysfs content. If this is a dm device, get associated dm name, the one
 * that appears in /dev/mapper. DM names could be resolved this way only if
 * kernel used >= 2.6.29, kernel name is found otherwise (e.g. dm-0).
 * If prefer_kernel_name is set, the kernel name is always preferred over
 * device-mapper name for dm devices no matter what the kernel version is.
 * For non-dm devices, we always get associated kernel name, e.g sda, md0 etc.
 * Returns 0 on error or if sysfs is not used (or configured incorrectly),
 * otherwise returns 1 and the supplied buffer holds the device name.
 */
int dm_device_get_name(uint32_t major, uint32_t minor,
		       int prefer_kernel_name,
		       char *buf, size_t buf_size);

/*
 * Determine whether a device has any holders (devices
 * using this device). If sysfs is not used (or configured
 * incorrectly), returns 0.
 */
int dm_device_has_holders(uint32_t major, uint32_t minor);

/*
 * Determine whether a device contains mounted filesystem.
 * If sysfs is not used (or configured incorrectly), returns 0.
 */
int dm_device_has_mounted_fs(uint32_t major, uint32_t minor);


/*
 * Callback is invoked for individal mountinfo lines,
 * minor, major and mount target are parsed and unmangled.
 */
typedef int (*dm_mountinfo_line_callback_fn) (char *line, unsigned maj, unsigned min,
					      char *target, void *cb_data);

/*
 * Read all lines from /proc/self/mountinfo,
 * for each line calls read_fn callback.
 */
int dm_mountinfo_read(dm_mountinfo_line_callback_fn read_fn, void *cb_data);

/*
 * Initialise library
 */
void dm_lib_init(void) __attribute__((constructor));

/*
 * Release library resources
 */
void dm_lib_release(void);
void dm_lib_exit(void) __attribute__((destructor));

/* An optimisation for clients making repeated calls involving dm ioctls */
void dm_hold_control_dev(int hold_open);

/*
 * Use NULL for all devices.
 */
int dm_mknodes(const char *name);
int dm_driver_version(char *version, size_t size);

/******************************************************
 * Functions to build and manipulate trees of devices *
 ******************************************************/
struct dm_tree;
struct dm_tree_node;

/*
 * Initialise an empty dependency tree.
 *
 * The tree consists of a root node together with one node for each mapped
 * device which has child nodes for each device referenced in its table.
 *
 * Every node in the tree has one or more children and one or more parents.
 *
 * The root node is the parent/child of every node that doesn't have other
 * parents/children.
 */
struct dm_tree *dm_tree_create(void);
void dm_tree_free(struct dm_tree *tree);

/*
 * List of suffixes to be ignored when matching uuids against existing devices.
 */
void dm_tree_set_optional_uuid_suffixes(struct dm_tree *dtree, const char **optional_uuid_suffixes);

/*
 * Add nodes to the tree for a given device and all the devices it uses.
 */
int dm_tree_add_dev(struct dm_tree *tree, uint32_t major, uint32_t minor);
int dm_tree_add_dev_with_udev_flags(struct dm_tree *tree, uint32_t major,
				    uint32_t minor, uint16_t udev_flags);

/*
 * Add a new node to the tree if it doesn't already exist.
 */
struct dm_tree_node *dm_tree_add_new_dev(struct dm_tree *tree,
					 const char *name,
					 const char *uuid,
					 uint32_t major, uint32_t minor,
					 int read_only,
					 int clear_inactive,
					 void *context);
struct dm_tree_node *dm_tree_add_new_dev_with_udev_flags(struct dm_tree *tree,
							 const char *name,
							 const char *uuid,
							 uint32_t major,
							 uint32_t minor,
							 int read_only,
							 int clear_inactive,
							 void *context,
							 uint16_t udev_flags);

/*
 * Search for a node in the tree.
 * Set major and minor to 0 or uuid to NULL to get the root node.
 */
struct dm_tree_node *dm_tree_find_node(struct dm_tree *tree,
				       uint32_t major,
				       uint32_t minor);
struct dm_tree_node *dm_tree_find_node_by_uuid(struct dm_tree *tree,
					       const char *uuid);

/*
 * Use this to walk through all children of a given node.
 * Set handle to NULL in first call.
 * Returns NULL after the last child.
 * Set inverted to use inverted tree.
 */
struct dm_tree_node *dm_tree_next_child(void **handle,
					const struct dm_tree_node *parent,
					uint32_t inverted);

/*
 * Get properties of a node.
 */
const char *dm_tree_node_get_name(const struct dm_tree_node *node);
const char *dm_tree_node_get_uuid(const struct dm_tree_node *node);
const struct dm_info *dm_tree_node_get_info(const struct dm_tree_node *node);
void *dm_tree_node_get_context(const struct dm_tree_node *node);
/*
 * Returns  0 when node size and its children is unchanged.
 * Returns  1 when node or any of its children has increased size.
 * Rerurns -1 when node or any of its children has reduced size.
 */
int dm_tree_node_size_changed(const struct dm_tree_node *dnode);

/*
 * Returns the number of children of the given node (excluding the root node).
 * Set inverted for the number of parents.
 */
int dm_tree_node_num_children(const struct dm_tree_node *node, uint32_t inverted);

/*
 * Deactivate a device plus all dependencies.
 * Ignores devices that don't have a uuid starting with uuid_prefix.
 */
int dm_tree_deactivate_children(struct dm_tree_node *dnode,
				const char *uuid_prefix,
				size_t uuid_prefix_len);
/*
 * Preload/create a device plus all dependencies.
 * Ignores devices that don't have a uuid starting with uuid_prefix.
 */
int dm_tree_preload_children(struct dm_tree_node *dnode,
			     const char *uuid_prefix,
			     size_t uuid_prefix_len);

/*
 * Resume a device plus all dependencies.
 * Ignores devices that don't have a uuid starting with uuid_prefix.
 */
int dm_tree_activate_children(struct dm_tree_node *dnode,
			      const char *uuid_prefix,
			      size_t uuid_prefix_len);

/*
 * Suspend a device plus all dependencies.
 * Ignores devices that don't have a uuid starting with uuid_prefix.
 */
int dm_tree_suspend_children(struct dm_tree_node *dnode,
			     const char *uuid_prefix,
			     size_t uuid_prefix_len);

/*
 * Skip the filesystem sync when suspending.
 * Does nothing with other functions.
 * Use this when no snapshots are involved.
 */
void dm_tree_skip_lockfs(struct dm_tree_node *dnode);

/*
 * Set the 'noflush' flag when suspending devices.
 * If the kernel supports it, instead of erroring outstanding I/O that
 * cannot be completed, the I/O is queued and resubmitted when the
 * device is resumed.  This affects multipath devices when all paths
 * have failed and queue_if_no_path is set, and mirror devices when
 * block_on_error is set and the mirror log has failed.
 */
void dm_tree_use_no_flush_suspend(struct dm_tree_node *dnode);

/*
 * Retry removal of each device if not successful.
 */
void dm_tree_retry_remove(struct dm_tree_node *dnode);

/*
 * Is the uuid prefix present in the tree?
 * Only returns 0 if every node was checked successfully.
 * Returns 1 if the tree walk has to be aborted.
 */
int dm_tree_children_use_uuid(struct dm_tree_node *dnode,
			      const char *uuid_prefix,
			      size_t uuid_prefix_len);

/*
 * Construct tables for new nodes before activating them.
 */
int dm_tree_node_add_snapshot_origin_target(struct dm_tree_node *dnode,
					    uint64_t size,
					    const char *origin_uuid);
int dm_tree_node_add_snapshot_target(struct dm_tree_node *node,
				     uint64_t size,
				     const char *origin_uuid,
				     const char *cow_uuid,
				     int persistent,
				     uint32_t chunk_size);
int dm_tree_node_add_snapshot_merge_target(struct dm_tree_node *node,
					   uint64_t size,
					   const char *origin_uuid,
					   const char *cow_uuid,
					   const char *merge_uuid,
					   uint32_t chunk_size);
int dm_tree_node_add_error_target(struct dm_tree_node *node,
				  uint64_t size);
int dm_tree_node_add_zero_target(struct dm_tree_node *node,
				 uint64_t size);
int dm_tree_node_add_linear_target(struct dm_tree_node *node,
				   uint64_t size);
int dm_tree_node_add_striped_target(struct dm_tree_node *node,
				    uint64_t size,
				    uint32_t stripe_size);

#define DM_CRYPT_IV_DEFAULT	UINT64_C(-1)	/* iv_offset == seg offset */
/*
 * Function accepts one string in cipher specification
 * (chainmode and iv should be NULL because included in cipher string)
 *   or
 * separate arguments which will be joined to "cipher-chainmode-iv"
 */
int dm_tree_node_add_crypt_target(struct dm_tree_node *node,
				  uint64_t size,
				  const char *cipher,
				  const char *chainmode,
				  const char *iv,
				  uint64_t iv_offset,
				  const char *key);
int dm_tree_node_add_mirror_target(struct dm_tree_node *node,
				   uint64_t size);

/* Mirror log flags */
#define DM_NOSYNC		0x00000001	/* Known already in sync */
#define DM_FORCESYNC		0x00000002	/* Force resync */
#define DM_BLOCK_ON_ERROR	0x00000004	/* On error, suspend I/O */
#define DM_CORELOG		0x00000008	/* In-memory log */

int dm_tree_node_add_mirror_target_log(struct dm_tree_node *node,
				       uint32_t region_size,
				       unsigned clustered,
				       const char *log_uuid,
				       unsigned area_count,
				       uint32_t flags);

int dm_tree_node_add_raid_target(struct dm_tree_node *node,
				 uint64_t size,
				 const char *raid_type,
				 uint32_t region_size,
				 uint32_t stripe_size,
				 uint64_t rebuilds,
				 uint64_t flags);

/*
 * Defines below are based on kernel's dm-cache.c defines
 * DM_CACHE_MIN_DATA_BLOCK_SIZE (32 * 1024 >> SECTOR_SHIFT)
 * DM_CACHE_MAX_DATA_BLOCK_SIZE (1024 * 1024 * 1024 >> SECTOR_SHIFT)
 */
#define DM_CACHE_MIN_DATA_BLOCK_SIZE (UINT32_C(64))
#define DM_CACHE_MAX_DATA_BLOCK_SIZE (UINT32_C(2097152))
/*
 * Max supported size for cache pool metadata device.
 * Limitation is hardcoded into the kernel and bigger device sizes
 * are not accepted.
 *
 * Limit defined in drivers/md/dm-cache-metadata.h
 */
#define DM_CACHE_METADATA_MAX_SECTORS DM_THIN_METADATA_MAX_SECTORS

/*
 * Define number of elements in rebuild and writemostly arrays
 * 'of struct dm_tree_node_raid_params'.
 */

struct dm_tree_node_raid_params {
	const char *raid_type;

	uint32_t stripes;
	uint32_t mirrors;
	uint32_t region_size;
	uint32_t stripe_size;

	/*
	 * 'rebuilds' and 'writemostly' are bitfields that signify
	 * which devices in the array are to be rebuilt or marked
	 * writemostly.  The kernel supports up to 253 legs.
	 * We limit ourselves by choosing a lower value
	 * for DEFAULT_RAID{1}_MAX_IMAGES in defaults.h.
	 */
	uint64_t rebuilds;
	uint64_t writemostly;
	uint32_t writebehind;	    /* I/Os (kernel default COUNTER_MAX / 2) */
	uint32_t sync_daemon_sleep; /* ms (kernel default = 5sec) */
	uint32_t max_recovery_rate; /* kB/sec/disk */
	uint32_t min_recovery_rate; /* kB/sec/disk */
	uint32_t stripe_cache;      /* sectors */

	uint64_t flags;             /* [no]sync */
	uint32_t reserved2;
};

/*
 * Version 2 of above node raid params struct to keeep API compatibility.
 *
 * Extended for more than 64 legs (max 253 in the MD kernel runtime!),
 * delta_disks for disk add/remove reshaping,
 * data_offset for out-of-place reshaping
 * and data_copies for odd number of raid10 legs.
 */
#define	RAID_BITMAP_SIZE 4 /* 4 * 64 bit elements in rebuilds/writemostly arrays */
struct dm_tree_node_raid_params_v2 {
	const char *raid_type;

	uint32_t stripes;
	uint32_t mirrors;
	uint32_t region_size;
	uint32_t stripe_size;

	int delta_disks; /* +/- number of disks to add/remove (reshaping) */
	int data_offset; /* data offset to set (out-of-place reshaping) */

	/*
	 * 'rebuilds' and 'writemostly' are bitfields that signify
	 * which devices in the array are to be rebuilt or marked
	 * writemostly.  The kernel supports up to 253 legs.
	 * We limit ourselvs by choosing a lower value
	 * for DEFAULT_RAID_MAX_IMAGES.
	 */
	uint64_t rebuilds[RAID_BITMAP_SIZE];
	uint64_t writemostly[RAID_BITMAP_SIZE];
	uint32_t writebehind;	    /* I/Os (kernel default COUNTER_MAX / 2) */
	uint32_t data_copies;	    /* RAID # of data copies */
	uint32_t sync_daemon_sleep; /* ms (kernel default = 5sec) */
	uint32_t max_recovery_rate; /* kB/sec/disk */
	uint32_t min_recovery_rate; /* kB/sec/disk */
	uint32_t stripe_cache;      /* sectors */

	uint64_t flags;             /* [no]sync */
};

int dm_tree_node_add_raid_target_with_params(struct dm_tree_node *node,
					     uint64_t size,
					     const struct dm_tree_node_raid_params *p);

/* Version 2 API function taking dm_tree_node_raid_params_v2 for aforementioned extensions. */
int dm_tree_node_add_raid_target_with_params_v2(struct dm_tree_node *node,
						uint64_t size,
						const struct dm_tree_node_raid_params_v2 *p);

/* Cache feature_flags */
#define DM_CACHE_FEATURE_WRITEBACK    0x00000001
#define DM_CACHE_FEATURE_WRITETHROUGH 0x00000002
#define DM_CACHE_FEATURE_PASSTHROUGH  0x00000004
#define DM_CACHE_FEATURE_METADATA2    0x00000008 /* cache v1.10 */
#define DM_CACHE_FEATURE_NO_DISCARD_PASSDOWN 0x00000010

struct dm_config_node;
/*
 * Use for passing cache policy and all its args e.g.:
 *
 * policy_settings {
 *    migration_threshold=2048
 *    sequention_threashold=100
 *    ...
 * }
 *
 * For policy without any parameters use NULL.
 */
int dm_tree_node_add_cache_target(struct dm_tree_node *node,
				  uint64_t size,
				  uint64_t feature_flags, /* DM_CACHE_FEATURE_* */
				  const char *metadata_uuid,
				  const char *data_uuid,
				  const char *origin_uuid,
				  const char *policy_name,
				  const struct dm_config_node *policy_settings,
				  uint32_t data_block_size);

/*
 * FIXME Add individual cache policy pairs  <key> = value, like:
 * int dm_tree_node_add_cache_policy_arg(struct dm_tree_node *dnode,
 *				      const char *key, uint64_t value);
 */

/*
 * Replicator operation mode
 * Note: API for Replicator is not yet stable
 */
typedef enum dm_replicator_mode_e {
	DM_REPLICATOR_SYNC,			/* Synchronous replication */
	DM_REPLICATOR_ASYNC_WARN,		/* Warn if async replicator is slow */
	DM_REPLICATOR_ASYNC_STALL,		/* Stall replicator if not fast enough */
	DM_REPLICATOR_ASYNC_DROP,		/* Drop sites out of sync */
	DM_REPLICATOR_ASYNC_FAIL,		/* Fail replicator if slow */
	NUM_DM_REPLICATOR_MODES
} dm_replicator_mode_t;

int dm_tree_node_add_replicator_target(struct dm_tree_node *node,
				       uint64_t size,
				       const char *rlog_uuid,
				       const char *rlog_type,
				       unsigned rsite_index,
				       dm_replicator_mode_t mode,
				       uint32_t async_timeout,
				       uint64_t fall_behind_data,
				       uint32_t fall_behind_ios);

int dm_tree_node_add_replicator_dev_target(struct dm_tree_node *node,
					   uint64_t size,
					   const char *replicator_uuid,	/* Replicator control device */
					   uint64_t rdevice_index,
					   const char *rdev_uuid,	/* Rimage device name/uuid */
					   unsigned rsite_index,
					   const char *slog_uuid,
					   uint32_t slog_flags,		/* Mirror log flags */
					   uint32_t slog_region_size);
/* End of Replicator API */

/*
 * FIXME: Defines bellow are based on kernel's dm-thin.c defines
 * DATA_DEV_BLOCK_SIZE_MIN_SECTORS (64 * 1024 >> SECTOR_SHIFT)
 * DATA_DEV_BLOCK_SIZE_MAX_SECTORS (1024 * 1024 * 1024 >> SECTOR_SHIFT)
 */
#define DM_THIN_MIN_DATA_BLOCK_SIZE (UINT32_C(128))
#define DM_THIN_MAX_DATA_BLOCK_SIZE (UINT32_C(2097152))
/*
 * Max supported size for thin pool metadata device (17045913600 bytes)
 * drivers/md/dm-thin-metadata.h THIN_METADATA_MAX_SECTORS
 * But here DM_THIN_MAX_METADATA_SIZE got defined incorrectly
 * Correct size is (UINT64_C(255) * ((1 << 14) - 64) * (4096 / (1 << 9)))
 */
#define DM_THIN_MAX_METADATA_SIZE   (UINT64_C(255) * (1 << 14) * (4096 / (1 << 9)) - 256 * 1024)

int dm_tree_node_add_thin_pool_target(struct dm_tree_node *node,
				      uint64_t size,
				      uint64_t transaction_id,
				      const char *metadata_uuid,
				      const char *pool_uuid,
				      uint32_t data_block_size,
				      uint64_t low_water_mark,
				      unsigned skip_block_zeroing);

int dm_tree_node_add_thin_pool_target_v1(struct dm_tree_node *node,
					 uint64_t size,
					 uint64_t transaction_id,
					 const char *metadata_uuid,
					 const char *pool_uuid,
					 uint32_t data_block_size,
					 uint64_t low_water_mark,
					 unsigned skip_block_zeroing,
					 unsigned crop_metadata);

/* Supported messages for thin provision target */
typedef enum dm_thin_message_e {
	DM_THIN_MESSAGE_CREATE_SNAP,		/* device_id, origin_id */
	DM_THIN_MESSAGE_CREATE_THIN,		/* device_id */
	DM_THIN_MESSAGE_DELETE,			/* device_id */
	DM_THIN_MESSAGE_SET_TRANSACTION_ID,	/* current_id, new_id */
	DM_THIN_MESSAGE_RESERVE_METADATA_SNAP,	/* target version >= 1.1 */
	DM_THIN_MESSAGE_RELEASE_METADATA_SNAP,	/* target version >= 1.1 */
} dm_thin_message_t;

int dm_tree_node_add_thin_pool_message(struct dm_tree_node *node,
				       dm_thin_message_t type,
				       uint64_t id1, uint64_t id2);

/*
 * Set thin pool discard features
 *   ignore      - Disable support for discards
 *   no_passdown - Don't pass discards down to underlying data device,
 *                 just remove the mapping
 * Feature is available since version 1.1 of the thin target.
 */
int dm_tree_node_set_thin_pool_discard(struct dm_tree_node *node,
				       unsigned ignore,
				       unsigned no_passdown);
/*
 * Set error if no space, instead of queueing for thin pool.
 */
int dm_tree_node_set_thin_pool_error_if_no_space(struct dm_tree_node *node,
						 unsigned error_if_no_space);
/* Start thin pool with metadata in read-only mode */
int dm_tree_node_set_thin_pool_read_only(struct dm_tree_node *node,
					 unsigned read_only);
/*
 * FIXME: Defines bellow are based on kernel's dm-thin.c defines
 * MAX_DEV_ID ((1 << 24) - 1)
 */
#define DM_THIN_MAX_DEVICE_ID (UINT32_C((1 << 24) - 1))
int dm_tree_node_add_thin_target(struct dm_tree_node *node,
				 uint64_t size,
				 const char *pool_uuid,
				 uint32_t device_id);

int dm_tree_node_set_thin_external_origin(struct dm_tree_node *node,
					  const char *external_uuid);

void dm_tree_node_set_udev_flags(struct dm_tree_node *node, uint16_t udev_flags);

void dm_tree_node_set_presuspend_node(struct dm_tree_node *node,
				      struct dm_tree_node *presuspend_node);

int dm_tree_node_add_target_area(struct dm_tree_node *node,
				    const char *dev_name,
				    const char *dlid,
				    uint64_t offset);

/*
 * Only for temporarily-missing raid devices where changes are tracked.
 */
int dm_tree_node_add_null_area(struct dm_tree_node *node, uint64_t offset);

/*
 * Set readahead (in sectors) after loading the node.
 */
void dm_tree_node_set_read_ahead(struct dm_tree_node *dnode,
				 uint32_t read_ahead,
				 uint32_t read_ahead_flags);

/*
 * Set node callback hook before de/activation.
 * Callback is called before 'activation' of node for activation tree,
 * or 'deactivation' of node for deactivation tree.
 */
typedef enum dm_node_callback_e {
	DM_NODE_CALLBACK_PRELOADED,   /* Node has preload deps */
	DM_NODE_CALLBACK_DEACTIVATED, /* Node is deactivated */
} dm_node_callback_t;
typedef int (*dm_node_callback_fn) (struct dm_tree_node *node,
				    dm_node_callback_t type, void *cb_data);
void dm_tree_node_set_callback(struct dm_tree_node *node,
			       dm_node_callback_fn cb, void *cb_data);

void dm_tree_set_cookie(struct dm_tree_node *node, uint32_t cookie);
uint32_t dm_tree_get_cookie(struct dm_tree_node *node);

/*****************************************************************************
 * Library functions
 *****************************************************************************/

/*******************
 * Memory management
 *******************/

/*
 * Never use these functions directly - use the macros following instead.
 */
void *dm_malloc_wrapper(size_t s, const char *file, int line)
	__attribute__((__malloc__)) __attribute__((__warn_unused_result__));
void *dm_malloc_aligned_wrapper(size_t s, size_t a, const char *file, int line)
	__attribute__((__malloc__)) __attribute__((__warn_unused_result__));
void *dm_zalloc_wrapper(size_t s, const char *file, int line)
	__attribute__((__malloc__)) __attribute__((__warn_unused_result__));
void *dm_realloc_wrapper(void *p, unsigned int s, const char *file, int line)
	__attribute__((__warn_unused_result__));
void dm_free_wrapper(void *ptr);
char *dm_strdup_wrapper(const char *s, const char *file, int line)
	__attribute__((__warn_unused_result__));
int dm_dump_memory_wrapper(void);
void dm_bounds_check_wrapper(void);

#define dm_malloc(s) dm_malloc_wrapper((s), __FILE__, __LINE__)
#define dm_malloc_aligned(s, a) dm_malloc_aligned_wrapper((s), (a),  __FILE__, __LINE__)
#define dm_zalloc(s) dm_zalloc_wrapper((s), __FILE__, __LINE__)
#define dm_strdup(s) dm_strdup_wrapper((s), __FILE__, __LINE__)
#define dm_free(p) dm_free_wrapper(p)
#define dm_realloc(p, s) dm_realloc_wrapper((p), (s), __FILE__, __LINE__)
#define dm_dump_memory() dm_dump_memory_wrapper()
#define dm_bounds_check() dm_bounds_check_wrapper()

/*
 * The pool allocator is useful when you are going to allocate
 * lots of memory, use the memory for a bit, and then free the
 * memory in one go.  A surprising amount of code has this usage
 * profile.
 *
 * You should think of the pool as an infinite, contiguous chunk
 * of memory.  The front of this chunk of memory contains
 * allocated objects, the second half is free.  dm_pool_alloc grabs
 * the next 'size' bytes from the free half, in effect moving it
 * into the allocated half.  This operation is very efficient.
 *
 * dm_pool_free frees the allocated object *and* all objects
 * allocated after it.  It is important to note this semantic
 * difference from malloc/free.  This is also extremely
 * efficient, since a single dm_pool_free can dispose of a large
 * complex object.
 *
 * dm_pool_destroy frees all allocated memory.
 *
 * eg, If you are building a binary tree in your program, and
 * know that you are only ever going to insert into your tree,
 * and not delete (eg, maintaining a symbol table for a
 * compiler).  You can create yourself a pool, allocate the nodes
 * from it, and when the tree becomes redundant call dm_pool_destroy
 * (no nasty iterating through the tree to free nodes).
 *
 * eg, On the other hand if you wanted to repeatedly insert and
 * remove objects into the tree, you would be better off
 * allocating the nodes from a free list; you cannot free a
 * single arbitrary node with pool.
 */

struct dm_pool;

/* constructor and destructor */
struct dm_pool *dm_pool_create(const char *name, size_t chunk_hint)
	__attribute__((__warn_unused_result__));
void dm_pool_destroy(struct dm_pool *p);

/* simple allocation/free routines */
void *dm_pool_alloc(struct dm_pool *p, size_t s)
	__attribute__((__warn_unused_result__));
void *dm_pool_alloc_aligned(struct dm_pool *p, size_t s, unsigned alignment)
	__attribute__((__warn_unused_result__));
void dm_pool_empty(struct dm_pool *p);
void dm_pool_free(struct dm_pool *p, void *ptr);

/*
 * To aid debugging, a pool can be locked. Any modifications made
 * to the content of the pool while it is locked can be detected.
 * Default compilation is using a crc checksum to notice modifications.
 * The pool locking is using the mprotect with the compilation flag
 * DEBUG_ENFORCE_POOL_LOCKING to enforce the memory protection.
 */
/* query pool lock status */
int dm_pool_locked(struct dm_pool *p);
/* mark pool as locked */
int dm_pool_lock(struct dm_pool *p, int crc)
	__attribute__((__warn_unused_result__));
/* mark pool as unlocked */
int dm_pool_unlock(struct dm_pool *p, int crc)
	__attribute__((__warn_unused_result__));

/*
 * Object building routines:
 *
 * These allow you to 'grow' an object, useful for
 * building strings, or filling in dynamic
 * arrays.
 *
 * It's probably best explained with an example:
 *
 * char *build_string(struct dm_pool *mem)
 * {
 *      int i;
 *      char buffer[16];
 *
 *      if (!dm_pool_begin_object(mem, 128))
 *              return NULL;
 *
 *      for (i = 0; i < 50; i++) {
 *              snprintf(buffer, sizeof(buffer), "%d, ", i);
 *              if (!dm_pool_grow_object(mem, buffer, 0))
 *                      goto bad;
 *      }
 *
 *	// add null
 *      if (!dm_pool_grow_object(mem, "\0", 1))
 *              goto bad;
 *
 *      return dm_pool_end_object(mem);
 *
 * bad:
 *
 *      dm_pool_abandon_object(mem);
 *      return NULL;
 *}
 *
 * So start an object by calling dm_pool_begin_object
 * with a guess at the final object size - if in
 * doubt make the guess too small.
 *
 * Then append chunks of data to your object with
 * dm_pool_grow_object.  Finally get your object with
 * a call to dm_pool_end_object.
 *
 * Setting delta to 0 means it will use strlen(extra).
 */
int dm_pool_begin_object(struct dm_pool *p, size_t hint);
int dm_pool_grow_object(struct dm_pool *p, const void *extra, size_t delta);
void *dm_pool_end_object(struct dm_pool *p);
void dm_pool_abandon_object(struct dm_pool *p);

/* utilities */
char *dm_pool_strdup(struct dm_pool *p, const char *str)
	__attribute__((__warn_unused_result__));
char *dm_pool_strndup(struct dm_pool *p, const char *str, size_t n)
	__attribute__((__warn_unused_result__));
void *dm_pool_zalloc(struct dm_pool *p, size_t s)
	__attribute__((__warn_unused_result__));

/******************
 * bitset functions
 ******************/

typedef uint32_t *dm_bitset_t;

dm_bitset_t dm_bitset_create(struct dm_pool *mem, unsigned num_bits);
void dm_bitset_destroy(dm_bitset_t bs);

int dm_bitset_equal(dm_bitset_t in1, dm_bitset_t in2);

void dm_bit_and(dm_bitset_t out, dm_bitset_t in1, dm_bitset_t in2);
void dm_bit_union(dm_bitset_t out, dm_bitset_t in1, dm_bitset_t in2);
int dm_bit_get_first(dm_bitset_t bs);
int dm_bit_get_next(dm_bitset_t bs, int last_bit);
int dm_bit_get_last(dm_bitset_t bs);
int dm_bit_get_prev(dm_bitset_t bs, int last_bit);

#define DM_BITS_PER_INT ((unsigned)sizeof(int) * CHAR_BIT)

#define dm_bit(bs, i) \
   ((bs)[((i) / DM_BITS_PER_INT) + 1] & (0x1 << ((i) & (DM_BITS_PER_INT - 1))))

#define dm_bit_set(bs, i) \
   ((bs)[((i) / DM_BITS_PER_INT) + 1] |= (0x1 << ((i) & (DM_BITS_PER_INT - 1))))

#define dm_bit_clear(bs, i) \
   ((bs)[((i) / DM_BITS_PER_INT) + 1] &= ~(0x1 << ((i) & (DM_BITS_PER_INT - 1))))

#define dm_bit_set_all(bs) \
   memset((bs) + 1, -1, ((*(bs) / DM_BITS_PER_INT) + 1) * sizeof(int))

#define dm_bit_clear_all(bs) \
   memset((bs) + 1, 0, ((*(bs) / DM_BITS_PER_INT) + 1) * sizeof(int))

#define dm_bit_copy(bs1, bs2) \
   memcpy((bs1) + 1, (bs2) + 1, ((*(bs2) / DM_BITS_PER_INT) + 1) * sizeof(int))

/*
 * Parse a string representation of a bitset into a dm_bitset_t. The
 * notation used is identical to the kernel bitmap parser (cpuset etc.)
 * and supports both lists ("1,2,3") and ranges ("1-2,5-8"). If the mem
 * parameter is NULL memory for the bitset will be allocated using
 * dm_malloc(). Otherwise the bitset will be allocated using the supplied
 * dm_pool.
 */
dm_bitset_t dm_bitset_parse_list(const char *str, struct dm_pool *mem,
				 size_t min_num_bits);

/* Returns number of set bits */
static inline unsigned hweight32(uint32_t i)
{
	unsigned r = (i & 0x55555555) + ((i >> 1) & 0x55555555);

	r =    (r & 0x33333333) + ((r >>  2) & 0x33333333);
	r =    (r & 0x0F0F0F0F) + ((r >>  4) & 0x0F0F0F0F);
	r =    (r & 0x00FF00FF) + ((r >>  8) & 0x00FF00FF);
	return (r & 0x0000FFFF) + ((r >> 16) & 0x0000FFFF);
}

/****************
 * hash functions
 ****************/

struct dm_hash_table;
struct dm_hash_node;

typedef void (*dm_hash_iterate_fn) (void *data);

struct dm_hash_table *dm_hash_create(unsigned size_hint)
	__attribute__((__warn_unused_result__));
void dm_hash_destroy(struct dm_hash_table *t);
void dm_hash_wipe(struct dm_hash_table *t);

void *dm_hash_lookup(struct dm_hash_table *t, const char *key);
int dm_hash_insert(struct dm_hash_table *t, const char *key, void *data);
void dm_hash_remove(struct dm_hash_table *t, const char *key);

void *dm_hash_lookup_binary(struct dm_hash_table *t, const void *key, uint32_t len);
int dm_hash_insert_binary(struct dm_hash_table *t, const void *key, uint32_t len,
			  void *data);
void dm_hash_remove_binary(struct dm_hash_table *t, const void *key, uint32_t len);

unsigned dm_hash_get_num_entries(struct dm_hash_table *t);
void dm_hash_iter(struct dm_hash_table *t, dm_hash_iterate_fn f);

char *dm_hash_get_key(struct dm_hash_table *t, struct dm_hash_node *n);
void *dm_hash_get_data(struct dm_hash_table *t, struct dm_hash_node *n);
struct dm_hash_node *dm_hash_get_first(struct dm_hash_table *t);
struct dm_hash_node *dm_hash_get_next(struct dm_hash_table *t, struct dm_hash_node *n);

/*
 * dm_hash_insert() replaces the value of an existing
 * entry with a matching key if one exists.  Otherwise
 * it adds a new entry.
 *
 * dm_hash_insert_with_val() inserts a new entry if
 * another entry with the same key already exists.
 * val_len is the size of the data being inserted.
 *
 * If two entries with the same key exist,
 * (added using dm_hash_insert_allow_multiple), then:
 * . dm_hash_lookup() returns the first one it finds, and
 *   dm_hash_lookup_with_val() returns the one with a matching
 *   val_len/val.
 * . dm_hash_remove() removes the first one it finds, and
 *   dm_hash_remove_with_val() removes the one with a matching
 *   val_len/val.
 *
 * If a single entry with a given key exists, and it has
 * zero val_len, then:
 * . dm_hash_lookup() returns it
 * . dm_hash_lookup_with_val(val_len=0) returns it
 * . dm_hash_remove() removes it
 * . dm_hash_remove_with_val(val_len=0) removes it
 *
 * dm_hash_lookup_with_count() is a single call that will
 * both lookup a key's value and check if there is more
 * than one entry with the given key.
 *
 * (It is not meant to retrieve all the entries with the
 * given key.  In the common case where a single entry exists
 * for the key, it is useful to have a single call that will
 * both look up the value and indicate if multiple values
 * exist for the key.)
 *
 * dm_hash_lookup_with_count:
 * . If no entries exist, the function returns NULL, and
 *   the count is set to 0.
 * . If only one entry exists, the value of that entry is
 *   returned and count is set to 1.
 * . If N entries exists, the value of the first entry is
 *   returned and count is set to N.
 */

void *dm_hash_lookup_with_val(struct dm_hash_table *t, const char *key,
                              const void *val, uint32_t val_len);
void dm_hash_remove_with_val(struct dm_hash_table *t, const char *key,
                             const void *val, uint32_t val_len);
int dm_hash_insert_allow_multiple(struct dm_hash_table *t, const char *key,
                                  const void *val, uint32_t val_len);
void *dm_hash_lookup_with_count(struct dm_hash_table *t, const char *key, int *count);


#define dm_hash_iterate(v, h) \
	for (v = dm_hash_get_first((h)); v; \
	     v = dm_hash_get_next((h), v))

/****************
 * list functions
 ****************/

/*
 * A list consists of a list head plus elements.
 * Each element has 'next' and 'previous' pointers.
 * The list head's pointers point to the first and the last element.
 */

struct dm_list {
	struct dm_list *n, *p;
};

/*
 * String list.
 */
struct dm_str_list {
	struct dm_list list;
	const char *str;
};

/*
 * Initialise a list before use.
 * The list head's next and previous pointers point back to itself.
 */
#define DM_LIST_HEAD_INIT(name)	 { &(name), &(name) }
#define DM_LIST_INIT(name)	struct dm_list name = DM_LIST_HEAD_INIT(name)
void dm_list_init(struct dm_list *head);

/*
 * Insert an element before 'head'.
 * If 'head' is the list head, this adds an element to the end of the list.
 */
void dm_list_add(struct dm_list *head, struct dm_list *elem);

/*
 * Insert an element after 'head'.
 * If 'head' is the list head, this adds an element to the front of the list.
 */
void dm_list_add_h(struct dm_list *head, struct dm_list *elem);

/*
 * Delete an element from its list.
 * Note that this doesn't change the element itself - it may still be safe
 * to follow its pointers.
 */
void dm_list_del(struct dm_list *elem);

/*
 * Remove an element from existing list and insert before 'head'.
 */
void dm_list_move(struct dm_list *head, struct dm_list *elem);

/*
 * Join 'head1' to the end of 'head'.
 */
void dm_list_splice(struct dm_list *head, struct dm_list *head1);

/*
 * Is the list empty?
 */
int dm_list_empty(const struct dm_list *head);

/*
 * Is this the first element of the list?
 */
int dm_list_start(const struct dm_list *head, const struct dm_list *elem);

/*
 * Is this the last element of the list?
 */
int dm_list_end(const struct dm_list *head, const struct dm_list *elem);

/*
 * Return first element of the list or NULL if empty
 */
struct dm_list *dm_list_first(const struct dm_list *head);

/*
 * Return last element of the list or NULL if empty
 */
struct dm_list *dm_list_last(const struct dm_list *head);

/*
 * Return the previous element of the list, or NULL if we've reached the start.
 */
struct dm_list *dm_list_prev(const struct dm_list *head, const struct dm_list *elem);

/*
 * Return the next element of the list, or NULL if we've reached the end.
 */
struct dm_list *dm_list_next(const struct dm_list *head, const struct dm_list *elem);

/*
 * Given the address v of an instance of 'struct dm_list' called 'head'
 * contained in a structure of type t, return the containing structure.
 */
#define dm_list_struct_base(v, t, head) \
    ((t *)((char *)(v) - offsetof(t, head)))

/*
 * Given the address v of an instance of 'struct dm_list list' contained in
 * a structure of type t, return the containing structure.
 */
#define dm_list_item(v, t) dm_list_struct_base((v), t, list)

/*
 * Given the address v of one known element e in a known structure of type t,
 * return another element f.
 */
#define dm_struct_field(v, t, e, f) \
    (((t *)((uintptr_t)(v) - offsetof(t, e))->f)

/*
 * Given the address v of a known element e in a known structure of type t,
 * return the list head 'list'
 */
#define dm_list_head(v, t, e) dm_struct_field(v, t, e, list)

/*
 * Set v to each element of a list in turn.
 */
#define dm_list_iterate(v, head) \
	for (v = (head)->n; v != head; v = v->n)

/*
 * Set v to each element in a list in turn, starting from the element
 * in front of 'start'.
 * You can use this to 'unwind' a list_iterate and back out actions on
 * already-processed elements.
 * If 'start' is 'head' it walks the list backwards.
 */
#define dm_list_uniterate(v, head, start) \
	for (v = (start)->p; v != head; v = v->p)

/*
 * A safe way to walk a list and delete and free some elements along
 * the way.
 * t must be defined as a temporary variable of the same type as v.
 */
#define dm_list_iterate_safe(v, t, head) \
	for (v = (head)->n, t = v->n; v != head; v = t, t = v->n)

/*
 * Walk a list, setting 'v' in turn to the containing structure of each item.
 * The containing structure should be the same type as 'v'.
 * The 'struct dm_list' variable within the containing structure is 'field'.
 */
#define dm_list_iterate_items_gen(v, head, field) \
	for (v = dm_list_struct_base((head)->n, __typeof__(*v), field); \
	     &v->field != (head); \
	     v = dm_list_struct_base(v->field.n, __typeof__(*v), field))

/*
 * Walk a list, setting 'v' in turn to the containing structure of each item.
 * The containing structure should be the same type as 'v'.
 * The list should be 'struct dm_list list' within the containing structure.
 */
#define dm_list_iterate_items(v, head) dm_list_iterate_items_gen(v, (head), list)

/*
 * Walk a list, setting 'v' in turn to the containing structure of each item.
 * The containing structure should be the same type as 'v'.
 * The 'struct dm_list' variable within the containing structure is 'field'.
 * t must be defined as a temporary variable of the same type as v.
 */
#define dm_list_iterate_items_gen_safe(v, t, head, field) \
	for (v = dm_list_struct_base((head)->n, __typeof__(*v), field), \
	     t = dm_list_struct_base(v->field.n, __typeof__(*v), field); \
	     &v->field != (head); \
	     v = t, t = dm_list_struct_base(v->field.n, __typeof__(*v), field))
/*
 * Walk a list, setting 'v' in turn to the containing structure of each item.
 * The containing structure should be the same type as 'v'.
 * The list should be 'struct dm_list list' within the containing structure.
 * t must be defined as a temporary variable of the same type as v.
 */
#define dm_list_iterate_items_safe(v, t, head) \
	dm_list_iterate_items_gen_safe(v, t, (head), list)

/*
 * Walk a list backwards, setting 'v' in turn to the containing structure
 * of each item.
 * The containing structure should be the same type as 'v'.
 * The 'struct dm_list' variable within the containing structure is 'field'.
 */
#define dm_list_iterate_back_items_gen(v, head, field) \
	for (v = dm_list_struct_base((head)->p, __typeof__(*v), field); \
	     &v->field != (head); \
	     v = dm_list_struct_base(v->field.p, __typeof__(*v), field))

/*
 * Walk a list backwards, setting 'v' in turn to the containing structure
 * of each item.
 * The containing structure should be the same type as 'v'.
 * The list should be 'struct dm_list list' within the containing structure.
 */
#define dm_list_iterate_back_items(v, head) dm_list_iterate_back_items_gen(v, (head), list)

/*
 * Return the number of elements in a list by walking it.
 */
unsigned int dm_list_size(const struct dm_list *head);

/*********
 * selinux
 *********/

/*
 * Obtain SELinux security context assigned for the path and set this
 * context for creating a new file system object. This security context
 * is global and it is used until reset to default policy behaviour
 * by calling 'dm_prepare_selinux_context(NULL, 0)'.
 */
int dm_prepare_selinux_context(const char *path, mode_t mode);
/*
 * Set SELinux context for existing file system object.
 */
int dm_set_selinux_context(const char *path, mode_t mode);

/*********************
 * string manipulation
 *********************/

/*
 * Break up the name of a mapped device into its constituent
 * Volume Group, Logical Volume and Layer (if present).
 * If mem is supplied, the result is allocated from the mempool.
 * Otherwise the strings are changed in situ.
 */
int dm_split_lvm_name(struct dm_pool *mem, const char *dmname,
		      char **vgname, char **lvname, char **layer);

/*
 * Destructively split buffer into NULL-separated words in argv.
 * Returns number of words.
 */
int dm_split_words(char *buffer, unsigned max,
		   unsigned ignore_comments, /* Not implemented */
		   char **argv);

/*
 * Returns -1 if buffer too small
 */
int dm_snprintf(char *buf, size_t bufsize, const char *format, ...)
    __attribute__ ((format(printf, 3, 4)));

/*
 * Returns pointer to the last component of the path.
 */
const char *dm_basename(const char *path);

/*
 * Returns number of occurrences of 'c' in 'str' of length 'size'.
 */
unsigned dm_count_chars(const char *str, size_t len, const int c);

/*
 * Length of string after escaping double quotes and backslashes.
 */
size_t dm_escaped_len(const char *str);

/*
 * <vg>-<lv>-<layer> or if !layer just <vg>-<lv>.
 */
char *dm_build_dm_name(struct dm_pool *mem, const char *vgname,
		       const char *lvname, const char *layer);
char *dm_build_dm_uuid(struct dm_pool *mem, const char *prefix, const char *lvid, const char *layer);

/*
 * Copies a string, quoting double quotes with backslashes.
 */
char *dm_escape_double_quotes(char *out, const char *src);

/*
 * Undo quoting in situ.
 */
void dm_unescape_double_quotes(char *src);

/*
 * Unescape colons and "at" signs in situ and save the substrings
 * starting at the position of the first unescaped colon and the
 * first unescaped "at" sign. This is normally used to unescape
 * device names used as PVs.
 */
void dm_unescape_colons_and_at_signs(char *src,
				     char **substr_first_unquoted_colon,
				     char **substr_first_unquoted_at_sign);

/*
 * Replacement for strncpy() function.
 *
 * Copies no more than n bytes from string pointed by src to the buffer
 * pointed by dest and ensure string is finished with '\0'.
 * Returns 0 if the whole string does not fit.
 */
int dm_strncpy(char *dest, const char *src, size_t n);

/*
 * Recognize unit specifier in the 'units' arg and return a factor
 * representing that unit. If the 'units' contains a prefix with digits,
 * the 'units' is considered to be a custom unit.
 *
 * Also, set 'unit_type' output arg to the character that represents
 * the unit specified. The 'unit_type' character equals to the unit
 * character itself recognized in the 'units' arg for canonical units.
 * Otherwise, the 'unit_type' character is set to 'U' for custom unit.
 *
 * An example for k/K canonical units and 8k/8K custom units:
 *
 *   units  unit_type  return value (factor)
 *   k      k          1024
 *   K      K          1000
 *   8k     U          1024*8
 *   8K     U          1000*8
 *   etc...
 *
 * Recognized units:
 *
 *   h/H - human readable (returns 1 for both)
 *   b/B - byte (returns 1 for both)
 *   s/S - sector (returns 512 for both)
 *   k/K - kilo (returns 1024/1000 respectively)
 *   m/M - mega (returns 1024^2/1000^2 respectively)
 *   g/G - giga (returns 1024^3/1000^3 respectively)
 *   t/T - tera (returns 1024^4/1000^4 respectively)
 *   p/P - peta (returns 1024^5/1000^5 respectively)
 *   e/E - exa (returns 1024^6/1000^6 respectively)
 *
 * Only one units character is allowed in the 'units' arg
 * if strict mode is enabled by 'strict' arg.
 *
 * The 'endptr' output arg, if not NULL, saves the pointer
 * in the 'units' string which follows the unit specifier
 * recognized (IOW the position where the parsing of the
 * unit specifier stopped).
 *
 * Returns the unit factor or 0 if no unit is recognized.
 */
uint64_t dm_units_to_factor(const char *units, char *unit_type,
			    int strict, const char **endptr);

/*
 * Type of unit specifier used by dm_size_to_string().
 */
typedef enum dm_size_suffix_e {
	DM_SIZE_LONG = 0,	/* Megabyte */
	DM_SIZE_SHORT = 1,	/* MB or MiB */
	DM_SIZE_UNIT = 2	/* M or m */
} dm_size_suffix_t;

/*
 * Convert a size (in 512-byte sectors) into a printable string using units of unit_type.
 * An upper-case unit_type indicates output units based on powers of 1000 are
 * required; a lower-case unit_type indicates powers of 1024.
 * For correct operation, unit_factor must be one of:
 * 	0 - the correct value will be calculated internally;
 *   or the output from dm_units_to_factor() corresponding to unit_type;
 *   or 'u' or 'U', an arbitrary number of bytes to use as the power base.
 * Set include_suffix to 1 to include a suffix of suffix_type.
 * Set use_si_units to 0 for suffixes that don't distinguish between 1000 and 1024.
 * Set use_si_units to 1 for a suffix that does distinguish.
 */
const char *dm_size_to_string(struct dm_pool *mem, uint64_t size,
			      char unit_type, int use_si_units,
			      uint64_t unit_factor, int include_suffix,
			      dm_size_suffix_t suffix_type);

/**************************
 * file/stream manipulation
 **************************/

/*
 * Create a directory (with parent directories if necessary).
 * Returns 1 on success, 0 on failure.
 */
int dm_create_dir(const char *dir);

int dm_is_empty_dir(const char *dir);

/*
 * Close a stream, with nicer error checking than fclose's.
 * Derived from gnulib's close-stream.c.
 *
 * Close "stream".  Return 0 if successful, and EOF (setting errno)
 * otherwise.  Upon failure, set errno to 0 if the error number
 * cannot be determined.  Useful mainly for writable streams.
 */
int dm_fclose(FILE *stream);

/*
 * Returns size of a buffer which is allocated with dm_malloc.
 * Pointer to the buffer is stored in *buf.
 * Returns -1 on failure leaving buf undefined.
 */
int dm_asprintf(char **buf, const char *format, ...)
    __attribute__ ((format(printf, 2, 3)));
int dm_vasprintf(char **buf, const char *format, va_list ap)
    __attribute__ ((format(printf, 2, 0)));

/*
 * create lockfile (pidfile) - create and lock a lock file
 * @lockfile: location of lock file
 *
 * Returns: 1 on success, 0 otherwise, errno is handled internally
 */
int dm_create_lockfile(const char* lockfile);

/*
 * Query whether a daemon is running based on its lockfile
 *
 * Returns: 1 if running, 0 if not
 */
int dm_daemon_is_running(const char* lockfile);

/*********************
 * regular expressions
 *********************/
struct dm_regex;

/*
 * Initialise an array of num patterns for matching.
 * Uses memory from mem.
 */
struct dm_regex *dm_regex_create(struct dm_pool *mem, const char * const *patterns,
				 unsigned num_patterns);

/*
 * Match string s against the patterns.
 * Returns the index of the highest pattern in the array that matches,
 * or -1 if none match.
 */
int dm_regex_match(struct dm_regex *regex, const char *s);

/*
 * This is useful for regression testing only.  The idea is if two
 * fingerprints are different, then the two dfas are certainly not
 * isomorphic.  If two fingerprints _are_ the same then it's very likely
 * that the dfas are isomorphic.
 *
 * This function must be called before any matching is done.
 */
uint32_t dm_regex_fingerprint(struct dm_regex *regex);

/******************
 * percent handling
 ******************/
/*
 * A fixed-point representation of percent values. One percent equals to
 * DM_PERCENT_1 as defined below. Values that are not multiples of DM_PERCENT_1
 * represent fractions, with precision of 1/1000000 of a percent. See
 * dm_percent_to_float for a conversion to a floating-point representation.
 *
 * You should always use dm_make_percent when building dm_percent_t values. The
 * implementation of dm_make_percent is biased towards the middle: it ensures that
 * the result is DM_PERCENT_0 or DM_PERCENT_100 if and only if this is the actual
 * value -- it never rounds any intermediate value (> 0 or < 100) to either 0
 * or 100.
*/
#define DM_PERCENT_CHAR '%'

typedef enum dm_percent_range_e {
	DM_PERCENT_0 = 0,
	DM_PERCENT_1 = 1000000,
	DM_PERCENT_100 = 100 * DM_PERCENT_1,
	DM_PERCENT_INVALID = -1,
	DM_PERCENT_FAILED = -2
} dm_percent_range_t;

typedef int32_t dm_percent_t;

float dm_percent_to_float(dm_percent_t percent);
/*
 * Return adjusted/rounded float for better percent value printing.
 * Function ensures for given precision of digits:
 * 100.0% returns only when the value is DM_PERCENT_100
 *        for close smaller values rounds to nearest smaller value
 * 0.0% returns only for value DM_PERCENT_0
 *        for close bigger values rounds to nearest bigger value
 * In all other cases returns same value as dm_percent_to_float()
 */
float dm_percent_to_round_float(dm_percent_t percent, unsigned digits);
dm_percent_t dm_make_percent(uint64_t numerator, uint64_t denominator);

/********************
 * timestamp handling
 ********************/

/*
 * Create a dm_timestamp object to use with dm_timestamp_get.
 */
struct dm_timestamp *dm_timestamp_alloc(void);

/*
 * Update dm_timestamp object to represent the current time.
 */
int dm_timestamp_get(struct dm_timestamp *ts);

/*
 * Copy a timestamp from ts_old to ts_new.
 */
void dm_timestamp_copy(struct dm_timestamp *ts_new, struct dm_timestamp *ts_old);

/*
 * Compare two timestamps.
 *
 * Return: -1 if ts1 is less than ts2
 *  	    0 if ts1 is equal to ts2
 *          1 if ts1 is greater than ts2
 */
int dm_timestamp_compare(struct dm_timestamp *ts1, struct dm_timestamp *ts2);

/*
 * Return the absolute difference in nanoseconds between
 * the dm_timestamp objects ts1 and ts2.
 *
 * Callers that need to know whether ts1 is before, equal to, or after ts2
 * in addition to the magnitude should use dm_timestamp_compare.
 */
uint64_t dm_timestamp_delta(struct dm_timestamp *ts1, struct dm_timestamp *ts2);

/*
 * Destroy a dm_timestamp object.
 */
void dm_timestamp_destroy(struct dm_timestamp *ts);

/*********************
 * reporting functions
 *********************/

struct dm_report_object_type {
	uint32_t id;			/* Powers of 2 */
	const char *desc;
	const char *prefix;		/* field id string prefix (optional) */
	/* FIXME: convert to proper usage of const pointers here */
	void *(*data_fn)(void *object);	/* callback from report_object() */
};

struct dm_report_field;

/*
 * dm_report_field_type flags
 */
#define DM_REPORT_FIELD_MASK				0x00000FFF
#define DM_REPORT_FIELD_ALIGN_MASK			0x0000000F
#define DM_REPORT_FIELD_ALIGN_LEFT			0x00000001
#define DM_REPORT_FIELD_ALIGN_RIGHT			0x00000002
#define DM_REPORT_FIELD_TYPE_MASK			0x00000FF0
#define DM_REPORT_FIELD_TYPE_NONE			0x00000000
#define DM_REPORT_FIELD_TYPE_STRING			0x00000010
#define DM_REPORT_FIELD_TYPE_NUMBER			0x00000020
#define DM_REPORT_FIELD_TYPE_SIZE			0x00000040
#define DM_REPORT_FIELD_TYPE_PERCENT			0x00000080
#define DM_REPORT_FIELD_TYPE_STRING_LIST		0x00000100
#define DM_REPORT_FIELD_TYPE_TIME			0x00000200

/* For use with reserved values only! */
#define DM_REPORT_FIELD_RESERVED_VALUE_MASK		0x0000000F
#define DM_REPORT_FIELD_RESERVED_VALUE_NAMED		0x00000001 /* only named value, less strict form of reservation */
#define DM_REPORT_FIELD_RESERVED_VALUE_RANGE		0x00000002 /* value is range - low and high value defined */
#define DM_REPORT_FIELD_RESERVED_VALUE_DYNAMIC_VALUE	0x00000004 /* value is computed in runtime */
#define DM_REPORT_FIELD_RESERVED_VALUE_FUZZY_NAMES	0x00000008 /* value names are recognized in runtime */

#define DM_REPORT_FIELD_TYPE_ID_LEN 32
#define DM_REPORT_FIELD_TYPE_HEADING_LEN 32

struct dm_report;
struct dm_report_field_type {
	uint32_t type;		/* object type id */
	uint32_t flags;		/* DM_REPORT_FIELD_* */
	uint32_t offset;	/* byte offset in the object */
	int32_t width;		/* default width */
	/* string used to specify the field */
	const char id[DM_REPORT_FIELD_TYPE_ID_LEN];
	/* string printed in header */
	const char heading[DM_REPORT_FIELD_TYPE_HEADING_LEN];
	int (*report_fn)(struct dm_report *rh, struct dm_pool *mem,
			 struct dm_report_field *field, const void *data,
			 void *private_data);
	const char *desc;	/* description of the field */
};

/*
 * Per-field reserved value.
 */
struct dm_report_field_reserved_value {
	/* field_num is the position of the field in 'fields'
	   array passed to dm_report_init_with_selection */
	uint32_t field_num;
	/* the value is of the same type as the field
	   identified by field_num */
	const void *value;
};

/*
 * Reserved value is a 'value' that is used directly if any of the 'names' is hit
 * or in case of fuzzy names, if such fuzzy name matches.
 *
 * If type is any of DM_REPORT_FIELD_TYPE_*, the reserved value is recognized
 * for all fields of that type.
 *
 * If type is DM_REPORT_FIELD_TYPE_NONE, the reserved value is recognized
 * for the exact field specified - hence the type of the value is automatically
 * the same as the type of the field itself.
 *
 * The array of reserved values is used to initialize reporting with
 * selection enabled (see also dm_report_init_with_selection function).
 */
struct dm_report_reserved_value {
	const uint32_t type;		/* DM_REPORT_FIELD_RESERVED_VALUE_* and DM_REPORT_FIELD_TYPE_*  */
	const void *value;		/* reserved value:
						uint64_t for DM_REPORT_FIELD_TYPE_NUMBER
						uint64_t for DM_REPORT_FIELD_TYPE_SIZE (number of 512-byte sectors)
						uint64_t for DM_REPORT_FIELD_TYPE_PERCENT
						const char* for DM_REPORT_FIELD_TYPE_STRING
						struct dm_report_field_reserved_value for DM_REPORT_FIELD_TYPE_NONE
						dm_report_reserved_handler* if DM_REPORT_FIELD_RESERVED_VALUE_{DYNAMIC_VALUE,FUZZY_NAMES} is used */
	const char **names;		/* null-terminated array of static names for this reserved value */
	const char *description;	/* description of the reserved value */
};

/*
 * Available actions for dm_report_reserved_value_handler.
 */
typedef enum dm_report_reserved_action_e {
	DM_REPORT_RESERVED_PARSE_FUZZY_NAME,
	DM_REPORT_RESERVED_GET_DYNAMIC_VALUE,
} dm_report_reserved_action_t;

/*
 * Generic reserved value handler to process reserved value names and/or values.
 *
 * Actions and their input/output:
 *
 * 	DM_REPORT_RESERVED_PARSE_FUZZY_NAME
 *		data_in:  const char *fuzzy_name
 *		data_out: const char *canonical_name, NULL if fuzzy_name not recognized
 *
 * 	DM_REPORT_RESERVED_GET_DYNAMIC_VALUE
 * 		data_in:  const char *canonical_name
 * 		data_out: void *value, NULL if canonical_name not recognized
 *
 * All actions return:
 *
 *	-1 if action not implemented
 * 	0 on error
 * 	1 on success
 */
typedef int (*dm_report_reserved_handler) (struct dm_report *rh,
					   struct dm_pool *mem,
					   uint32_t field_num,
					   dm_report_reserved_action_t action,
					   const void *data_in,
					   const void **data_out);

/*
 * The dm_report_value_cache_{set,get} are helper functions to store and retrieve
 * various values used during reporting (dm_report_field_type.report_fn) and/or
 * selection processing (dm_report_reserved_handler instances) to avoid
 * recalculation of these values or to share values among calls.
 */
int dm_report_value_cache_set(struct dm_report *rh, const char *name, const void *data);
const void *dm_report_value_cache_get(struct dm_report *rh, const char *name);
/*
 * dm_report_init output_flags
 */
#define DM_REPORT_OUTPUT_MASK			0x000000FF
#define DM_REPORT_OUTPUT_ALIGNED		0x00000001
#define DM_REPORT_OUTPUT_BUFFERED		0x00000002
#define DM_REPORT_OUTPUT_HEADINGS		0x00000004
#define DM_REPORT_OUTPUT_FIELD_NAME_PREFIX	0x00000008
#define DM_REPORT_OUTPUT_FIELD_UNQUOTED		0x00000010
#define DM_REPORT_OUTPUT_COLUMNS_AS_ROWS	0x00000020
#define DM_REPORT_OUTPUT_MULTIPLE_TIMES		0x00000040

struct dm_report *dm_report_init(uint32_t *report_types,
				 const struct dm_report_object_type *types,
				 const struct dm_report_field_type *fields,
				 const char *output_fields,
				 const char *output_separator,
				 uint32_t output_flags,
				 const char *sort_keys,
				 void *private_data);
struct dm_report *dm_report_init_with_selection(uint32_t *report_types,
						const struct dm_report_object_type *types,
						const struct dm_report_field_type *fields,
						const char *output_fields,
						const char *output_separator,
						uint32_t output_flags,
						const char *sort_keys,
						const char *selection,
						const struct dm_report_reserved_value reserved_values[],
						void *private_data);
/*
 * Report an object, pass it through the selection criteria if they
 * are present and display the result on output if it passes the criteria.
 */
int dm_report_object(struct dm_report *rh, void *object);
/*
 * The same as dm_report_object, but display the result on output only if
 * 'do_output' arg is set. Also, save the result of selection in 'selected'
 * arg if it's not NULL (either 1 if the object passes, otherwise 0).
 */
int dm_report_object_is_selected(struct dm_report *rh, void *object, int do_output, int *selected);

/*
 * Compact report output so that if field value is empty for all rows in
 * the report, drop the field from output completely (including headers).
 * Compact output is applicable only if report is buffered, otherwise
 * this function has no effect.
 */
int dm_report_compact_fields(struct dm_report *rh);

/*
 * The same as dm_report_compact_fields, but for selected fields only.
 * The "fields" arg is comma separated list of field names (the same format
 * as used for "output_fields" arg in dm_report_init fn).
 */
int dm_report_compact_given_fields(struct dm_report *rh, const char *fields);

/*
 * Returns 1 if there is no data waiting to be output.
 */
int dm_report_is_empty(struct dm_report *rh);

/*
 * Destroy report content without doing output.
 */
void dm_report_destroy_rows(struct dm_report *rh);

int dm_report_output(struct dm_report *rh);

/*
 * Output the report headings for a columns-based report, even if they
 * have already been shown. Useful for repeating reports that wish to
 * issue a periodic reminder of the column headings.
 */
int dm_report_column_headings(struct dm_report *rh);

void dm_report_free(struct dm_report *rh);

/*
 * Prefix added to each field name with DM_REPORT_OUTPUT_FIELD_NAME_PREFIX
 */
int dm_report_set_output_field_name_prefix(struct dm_report *rh,
					   const char *report_prefix);

int dm_report_set_selection(struct dm_report *rh, const char *selection);

/*
 * Report functions are provided for simple data types.
 * They take care of allocating copies of the data.
 */
int dm_report_field_string(struct dm_report *rh, struct dm_report_field *field,
			   const char *const *data);
int dm_report_field_string_list(struct dm_report *rh, struct dm_report_field *field,
				const struct dm_list *data, const char *delimiter);
int dm_report_field_string_list_unsorted(struct dm_report *rh, struct dm_report_field *field,
					 const struct dm_list *data, const char *delimiter);
int dm_report_field_int32(struct dm_report *rh, struct dm_report_field *field,
			  const int32_t *data);
int dm_report_field_uint32(struct dm_report *rh, struct dm_report_field *field,
			   const uint32_t *data);
int dm_report_field_int(struct dm_report *rh, struct dm_report_field *field,
			const int *data);
int dm_report_field_uint64(struct dm_report *rh, struct dm_report_field *field,
			   const uint64_t *data);
int dm_report_field_percent(struct dm_report *rh, struct dm_report_field *field,
			    const dm_percent_t *data);

/*
 * For custom fields, allocate the data in 'mem' and use
 * dm_report_field_set_value().
 * 'sortvalue' may be NULL if it matches 'value'
 */
void dm_report_field_set_value(struct dm_report_field *field, const void *value,
			       const void *sortvalue);

/*
 * Report group support.
 */
struct dm_report_group;

typedef enum dm_report_group_type_e {
	DM_REPORT_GROUP_SINGLE,
	DM_REPORT_GROUP_BASIC,
	DM_REPORT_GROUP_JSON,
	DM_REPORT_GROUP_JSON_STD
} dm_report_group_type_t;

struct dm_report_group *dm_report_group_create(dm_report_group_type_t type, void *data);
int dm_report_group_push(struct dm_report_group *group, struct dm_report *report, void *data);
int dm_report_group_pop(struct dm_report_group *group);
int dm_report_group_output_and_pop_all(struct dm_report_group *group);
int dm_report_group_destroy(struct dm_report_group *group);

/*
 * Stats counter access methods
 *
 * Each method returns the corresponding stats counter value from the
 * supplied dm_stats handle for the specified region_id and area_id.
 * If either region_id or area_id uses one of the special values
 * DM_STATS_REGION_CURRENT or DM_STATS_AREA_CURRENT then the region
 * or area is selected according to the current state of the dm_stats
 * handle's embedded cursor.
 *
 * Two methods are provided to access counter values: a named function
 * for each available counter field and a single function that accepts
 * an enum value specifying the required field. New code is encouraged
 * to use the enum based interface as calls to the named functions are
 * implemented using the enum method internally.
 *
 * See the kernel documentation for complete descriptions of each
 * counter field:
 *
 * Documentation/device-mapper/statistics.txt
 * Documentation/iostats.txt
 *
 * reads: the number of reads completed
 * reads_merged: the number of reads merged
 * read_sectors: the number of sectors read
 * read_nsecs: the number of nanoseconds spent reading
 * writes: the number of writes completed
 * writes_merged: the number of writes merged
 * write_sectors: the number of sectors written
 * write_nsecs: the number of nanoseconds spent writing
 * io_in_progress: the number of I/Os currently in progress
 * io_nsecs: the number of nanoseconds spent doing I/Os
 * weighted_io_nsecs: the weighted number of nanoseconds spent doing I/Os
 * total_read_nsecs: the total time spent reading in nanoseconds
 * total_write_nsecs: the total time spent writing in nanoseconds
 */

#define DM_STATS_REGION_CURRENT UINT64_MAX
#define DM_STATS_AREA_CURRENT UINT64_MAX

typedef enum dm_stats_counter_e {
	DM_STATS_READS_COUNT,
	DM_STATS_READS_MERGED_COUNT,
	DM_STATS_READ_SECTORS_COUNT,
	DM_STATS_READ_NSECS,
	DM_STATS_WRITES_COUNT,
	DM_STATS_WRITES_MERGED_COUNT,
	DM_STATS_WRITE_SECTORS_COUNT,
	DM_STATS_WRITE_NSECS,
	DM_STATS_IO_IN_PROGRESS_COUNT,
	DM_STATS_IO_NSECS,
	DM_STATS_WEIGHTED_IO_NSECS,
	DM_STATS_TOTAL_READ_NSECS,
	DM_STATS_TOTAL_WRITE_NSECS,
	DM_STATS_NR_COUNTERS
} dm_stats_counter_t;

uint64_t dm_stats_get_counter(const struct dm_stats *dms,
			      dm_stats_counter_t counter,
			      uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_reads(const struct dm_stats *dms,
			    uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_reads_merged(const struct dm_stats *dms,
				   uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_read_sectors(const struct dm_stats *dms,
				   uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_read_nsecs(const struct dm_stats *dms,
				 uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_writes(const struct dm_stats *dms,
			     uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_writes_merged(const struct dm_stats *dms,
				    uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_write_sectors(const struct dm_stats *dms,
				    uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_write_nsecs(const struct dm_stats *dms,
				  uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_io_in_progress(const struct dm_stats *dms,
				     uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_io_nsecs(const struct dm_stats *dms,
			       uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_weighted_io_nsecs(const struct dm_stats *dms,
					uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_total_read_nsecs(const struct dm_stats *dms,
				       uint64_t region_id, uint64_t area_id);

uint64_t dm_stats_get_total_write_nsecs(const struct dm_stats *dms,
					uint64_t region_id, uint64_t area_id);

/*
 * Derived statistics access methods
 *
 * Each method returns the corresponding value calculated from the
 * counters stored in the supplied dm_stats handle for the specified
 * region_id and area_id. If either region_id or area_id uses one of the
 * special values DM_STATS_REGION_CURRENT or DM_STATS_AREA_CURRENT then
 * the region or area is selected according to the current state of the
 * dm_stats handle's embedded cursor.
 *
 * The set of metrics is based on the fields provided by the Linux
 * iostats program.
 *
 * rd_merges_per_sec: the number of reads merged per second
 * wr_merges_per_sec: the number of writes merged per second
 * reads_per_sec: the number of reads completed per second
 * writes_per_sec: the number of writes completed per second
 * read_sectors_per_sec: the number of sectors read per second
 * write_sectors_per_sec: the number of sectors written per second
 * average_request_size: the average size of requests submitted
 * service_time: the average service time (in ns) for requests issued
 * average_queue_size: the average queue length
 * average_wait_time: the average time for requests to be served (in ns)
 * average_rd_wait_time: the average read wait time
 * average_wr_wait_time: the average write wait time
 */

typedef enum dm_stats_metric_e {
	DM_STATS_RD_MERGES_PER_SEC,
	DM_STATS_WR_MERGES_PER_SEC,
	DM_STATS_READS_PER_SEC,
	DM_STATS_WRITES_PER_SEC,
	DM_STATS_READ_SECTORS_PER_SEC,
	DM_STATS_WRITE_SECTORS_PER_SEC,
	DM_STATS_AVERAGE_REQUEST_SIZE,
	DM_STATS_AVERAGE_QUEUE_SIZE,
	DM_STATS_AVERAGE_WAIT_TIME,
	DM_STATS_AVERAGE_RD_WAIT_TIME,
	DM_STATS_AVERAGE_WR_WAIT_TIME,
	DM_STATS_SERVICE_TIME,
	DM_STATS_THROUGHPUT,
	DM_STATS_UTILIZATION,
	DM_STATS_NR_METRICS
} dm_stats_metric_t;

int dm_stats_get_metric(const struct dm_stats *dms, int metric,
			uint64_t region_id, uint64_t area_id, double *value);

int dm_stats_get_rd_merges_per_sec(const struct dm_stats *dms, double *rrqm,
				   uint64_t region_id, uint64_t area_id);

int dm_stats_get_wr_merges_per_sec(const struct dm_stats *dms, double *rrqm,
				   uint64_t region_id, uint64_t area_id);

int dm_stats_get_reads_per_sec(const struct dm_stats *dms, double *rd_s,
			       uint64_t region_id, uint64_t area_id);

int dm_stats_get_writes_per_sec(const struct dm_stats *dms, double *wr_s,
				uint64_t region_id, uint64_t area_id);

int dm_stats_get_read_sectors_per_sec(const struct dm_stats *dms,
				      double *rsec_s, uint64_t region_id,
				      uint64_t area_id);

int dm_stats_get_write_sectors_per_sec(const struct dm_stats *dms,
				       double *wr_s, uint64_t region_id,
				       uint64_t area_id);

int dm_stats_get_average_request_size(const struct dm_stats *dms,
				      double *arqsz, uint64_t region_id,
				      uint64_t area_id);

int dm_stats_get_service_time(const struct dm_stats *dms, double *svctm,
			      uint64_t region_id, uint64_t area_id);

int dm_stats_get_average_queue_size(const struct dm_stats *dms, double *qusz,
				    uint64_t region_id, uint64_t area_id);

int dm_stats_get_average_wait_time(const struct dm_stats *dms, double *await,
				   uint64_t region_id, uint64_t area_id);

int dm_stats_get_average_rd_wait_time(const struct dm_stats *dms,
				      double *await, uint64_t region_id,
				      uint64_t area_id);

int dm_stats_get_average_wr_wait_time(const struct dm_stats *dms,
				      double *await, uint64_t region_id,
				      uint64_t area_id);

int dm_stats_get_throughput(const struct dm_stats *dms, double *tput,
			    uint64_t region_id, uint64_t area_id);

int dm_stats_get_utilization(const struct dm_stats *dms, dm_percent_t *util,
			     uint64_t region_id, uint64_t area_id);

/*
 * Statistics histogram access methods.
 *
 * Methods to access latency histograms for regions that have them
 * enabled. Each histogram contains a configurable number of bins
 * spanning a user defined latency interval.
 *
 * The bin count, upper and lower bin bounds, and bin values are
 * made available via the following area methods.
 *
 * Methods to obtain a simple string representation of the histogram
 * and its bounds are also provided.
 */

/*
 * Retrieve a pointer to the histogram associated with the specified
 * area. If the area does not have a histogram configured this function
 * returns NULL.
 *
 * The pointer does not need to be freed explicitly by the caller: it
 * will become invalid following a subsequent dm_stats_list(),
 * dm_stats_populate() or dm_stats_destroy() of the corresponding
 * dm_stats handle.
 *
 * If region_id or area_id is one of the special values
 * DM_STATS_REGION_CURRENT or DM_STATS_AREA_CURRENT the current cursor
 * value is used to select the region or area.
 */
struct dm_histogram *dm_stats_get_histogram(const struct dm_stats *dms,
					    uint64_t region_id,
					    uint64_t area_id);

/*
 * Return the number of bins in the specified histogram handle.
 */
int dm_histogram_get_nr_bins(const struct dm_histogram *dmh);

/*
 * Get the lower bound of the specified bin of the histogram for the
 * area specified by region_id and area_id. The value is returned in
 * nanoseconds.
 */
uint64_t dm_histogram_get_bin_lower(const struct dm_histogram *dmh, int bin);

/*
 * Get the upper bound of the specified bin of the histogram for the
 * area specified by region_id and area_id. The value is returned in
 * nanoseconds.
 */
uint64_t dm_histogram_get_bin_upper(const struct dm_histogram *dmh, int bin);

/*
 * Get the width of the specified bin of the histogram for the area
 * specified by region_id and area_id. The width is equal to the bin
 * upper bound minus the lower bound and yields the range of latency
 * values covered by this bin. The value is returned in nanoseconds.
 */
uint64_t dm_histogram_get_bin_width(const struct dm_histogram *dmh, int bin);

/*
 * Get the value of the specified bin of the histogram for the area
 * specified by region_id and area_id.
 */
uint64_t dm_histogram_get_bin_count(const struct dm_histogram *dmh, int bin);

/*
 * Get the percentage (relative frequency) of the specified bin of the
 * histogram for the area specified by region_id and area_id.
 */
dm_percent_t dm_histogram_get_bin_percent(const struct dm_histogram *dmh,
					  int bin);

/*
 * Return the total observations (sum of bin counts) for the histogram
 * of the area specified by region_id and area_id.
 */
uint64_t dm_histogram_get_sum(const struct dm_histogram *dmh);

/*
 * Histogram formatting flags.
 */
#define DM_HISTOGRAM_SUFFIX  0x1
#define DM_HISTOGRAM_VALUES  0x2
#define DM_HISTOGRAM_PERCENT 0X4
#define DM_HISTOGRAM_BOUNDS_LOWER 0x10
#define DM_HISTOGRAM_BOUNDS_UPPER 0x20
#define DM_HISTOGRAM_BOUNDS_RANGE 0x30

/*
 * Return a string representation of the supplied histogram's values and
 * bin boundaries.
 *
 * The bin argument selects the bin to format. If this argument is less
 * than zero all bins will be included in the resulting string.
 *
 * width specifies a minimum width for the field in characters; if it is
 * zero the width will be determined automatically based on the options
 * selected for formatting. A value less than zero disables field width
 * control: bin boundaries and values will be output with a minimum
 * amount of whitespace.
 *
 * flags is a collection of flag arguments that control the string format:
 *
 * DM_HISTOGRAM_VALUES  - Include bin values in the string.
 * DM_HISTOGRAM_SUFFIX  - Include time unit suffixes when printing bounds.
 * DM_HISTOGRAM_PERCENT - Format bin values as a percentage.
 *
 * DM_HISTOGRAM_BOUNDS_LOWER - Include the lower bound of each bin.
 * DM_HISTOGRAM_BOUNDS_UPPER - Include the upper bound of each bin.
 * DM_HISTOGRAM_BOUNDS_RANGE - Show the span of each bin as "lo-up".
 *
 * The returned pointer does not need to be freed explicitly by the
 * caller: it will become invalid following a subsequent
 * dm_stats_list(), dm_stats_populate() or dm_stats_destroy() of the
 * corresponding dm_stats handle.
 */
const char *dm_histogram_to_string(const struct dm_histogram *dmh, int bin,
				   int width, int flags);

/*************************
 * config file parse/print
 *************************/
typedef enum dm_config_value_type_e {
	DM_CFG_INT,
	DM_CFG_FLOAT,
	DM_CFG_STRING,
	DM_CFG_EMPTY_ARRAY
} dm_config_value_type_t;

struct dm_config_value {
	dm_config_value_type_t type;

	union dm_config_value_u {
		int64_t i;
		float f;
		double d;       	/* Unused. */
		const char *str;
	} v;

	struct dm_config_value *next;	/* For arrays */
	uint32_t format_flags;
};

struct dm_config_node {
	const char *key;
	struct dm_config_node *parent, *sib, *child;
	struct dm_config_value *v;
	int id;
};

struct dm_config_tree {
	struct dm_config_node *root;
	struct dm_config_tree *cascade;
	struct dm_pool *mem;
	void *custom;
};

struct dm_config_tree *dm_config_create(void);
struct dm_config_tree *dm_config_from_string(const char *config_settings);
int dm_config_parse(struct dm_config_tree *cft, const char *start, const char *end);
int dm_config_parse_without_dup_node_check(struct dm_config_tree *cft, const char *start, const char *end);

void *dm_config_get_custom(struct dm_config_tree *cft);
void dm_config_set_custom(struct dm_config_tree *cft, void *custom);

/*
 * When searching, first_cft is checked before second_cft.
 */
struct dm_config_tree *dm_config_insert_cascaded_tree(struct dm_config_tree *first_cft, struct dm_config_tree *second_cft);

/*
 * If there's a cascaded dm_config_tree, remove the top layer
 * and return the layer below.  Otherwise return NULL.
 */
struct dm_config_tree *dm_config_remove_cascaded_tree(struct dm_config_tree *cft);

/*
 * Create a new, uncascaded config tree equivalent to the input cascade.
 */
struct dm_config_tree *dm_config_flatten(struct dm_config_tree *cft);

void dm_config_destroy(struct dm_config_tree *cft);

/* Simple output line by line. */
typedef int (*dm_putline_fn)(const char *line, void *baton);
/* More advaced output with config node reference. */
typedef int (*dm_config_node_out_fn)(const struct dm_config_node *cn, const char *line, void *baton);

/*
 * Specification for advanced config node output.
 */
struct dm_config_node_out_spec {
	dm_config_node_out_fn prefix_fn; /* called before processing config node lines */
	dm_config_node_out_fn line_fn; /* called for each config node line */
	dm_config_node_out_fn suffix_fn; /* called after processing config node lines */
};

/* Write the node and any subsequent siblings it has. */
int dm_config_write_node(const struct dm_config_node *cn, dm_putline_fn putline, void *baton);
int dm_config_write_node_out(const struct dm_config_node *cn, const struct dm_config_node_out_spec *out_spec, void *baton);

/* Write given node only without subsequent siblings. */
int dm_config_write_one_node(const struct dm_config_node *cn, dm_putline_fn putline, void *baton);
int dm_config_write_one_node_out(const struct dm_config_node *cn, const struct dm_config_node_out_spec *out_spec, void *baton);

struct dm_config_node *dm_config_find_node(const struct dm_config_node *cn, const char *path);
int dm_config_has_node(const struct dm_config_node *cn, const char *path);
int dm_config_remove_node(struct dm_config_node *parent, struct dm_config_node *remove);
const char *dm_config_find_str(const struct dm_config_node *cn, const char *path, const char *fail);
const char *dm_config_find_str_allow_empty(const struct dm_config_node *cn, const char *path, const char *fail);
int dm_config_find_int(const struct dm_config_node *cn, const char *path, int fail);
int64_t dm_config_find_int64(const struct dm_config_node *cn, const char *path, int64_t fail);
float dm_config_find_float(const struct dm_config_node *cn, const char *path, float fail);

const struct dm_config_node *dm_config_tree_find_node(const struct dm_config_tree *cft, const char *path);
const char *dm_config_tree_find_str(const struct dm_config_tree *cft, const char *path, const char *fail);
const char *dm_config_tree_find_str_allow_empty(const struct dm_config_tree *cft, const char *path, const char *fail);
int dm_config_tree_find_int(const struct dm_config_tree *cft, const char *path, int fail);
int64_t dm_config_tree_find_int64(const struct dm_config_tree *cft, const char *path, int64_t fail);
float dm_config_tree_find_float(const struct dm_config_tree *cft, const char *path, float fail);
int dm_config_tree_find_bool(const struct dm_config_tree *cft, const char *path, int fail);

/*
 * Understands (0, ~0), (y, n), (yes, no), (on,
 * off), (true, false).
 */
int dm_config_find_bool(const struct dm_config_node *cn, const char *path, int fail);
int dm_config_value_is_bool(const struct dm_config_value *v);

int dm_config_get_uint32(const struct dm_config_node *cn, const char *path, uint32_t *result);
int dm_config_get_uint64(const struct dm_config_node *cn, const char *path, uint64_t *result);
int dm_config_get_str(const struct dm_config_node *cn, const char *path, const char **result);
int dm_config_get_list(const struct dm_config_node *cn, const char *path, const struct dm_config_value **result);
int dm_config_get_section(const struct dm_config_node *cn, const char *path, const struct dm_config_node **result);

unsigned dm_config_maybe_section(const char *str, unsigned len);

const char *dm_config_parent_name(const struct dm_config_node *n);

struct dm_config_node *dm_config_clone_node_with_mem(struct dm_pool *mem, const struct dm_config_node *node, int siblings);
struct dm_config_node *dm_config_create_node(struct dm_config_tree *cft, const char *key);
struct dm_config_value *dm_config_create_value(struct dm_config_tree *cft);
struct dm_config_node *dm_config_clone_node(struct dm_config_tree *cft, const struct dm_config_node *cn, int siblings);

/*
 * Common formatting flags applicable to all config node types (lower 16 bits).
 */
#define DM_CONFIG_VALUE_FMT_COMMON_ARRAY             0x00000001 /* value is array */
#define DM_CONFIG_VALUE_FMT_COMMON_EXTRA_SPACES      0x00000002 /* add spaces in "key = value" pairs in constrast to "key=value" for better readability */

/*
 * Type-related config node formatting flags (higher 16 bits).
 */
/* int-related formatting flags */
#define DM_CONFIG_VALUE_FMT_INT_OCTAL                0x00010000 /* print number in octal form */

/* string-related formatting flags */
#define DM_CONFIG_VALUE_FMT_STRING_NO_QUOTES         0x00010000 /* do not print quotes around string value */

void dm_config_value_set_format_flags(struct dm_config_value *cv, uint32_t format_flags);
uint32_t dm_config_value_get_format_flags(struct dm_config_value *cv);

struct dm_pool *dm_config_memory(struct dm_config_tree *cft);

/* Udev device directory. */
#define DM_UDEV_DEV_DIR "/dev/"

/* Cookie prefixes.
 *
 * The cookie value consists of a prefix (16 bits) and a base (16 bits).
 * We can use the prefix to store the flags. These flags are sent to
 * kernel within given dm task. When returned back to userspace in
 * DM_COOKIE udev environment variable, we can control several aspects
 * of udev rules we use by decoding the cookie prefix. When doing the
 * notification, we replace the cookie prefix with DM_COOKIE_MAGIC,
 * so we notify the right semaphore.
 *
 * It is still possible to use cookies for passing the flags to udev
 * rules even when udev_sync is disabled. The base part of the cookie
 * will be zero (there's no notification semaphore) and prefix will be
 * set then. However, having udev_sync enabled is highly recommended.
 */
#define DM_COOKIE_MAGIC 0x0D4D
#define DM_UDEV_FLAGS_MASK 0xFFFF0000
#define DM_UDEV_FLAGS_SHIFT 16

/*
 * DM_UDEV_DISABLE_DM_RULES_FLAG is set in case we need to disable
 * basic device-mapper udev rules that create symlinks in /dev/<DM_DIR>
 * directory. However, we can't reliably prevent creating default
 * nodes by udev (commonly /dev/dm-X, where X is a number).
 */
#define DM_UDEV_DISABLE_DM_RULES_FLAG 0x0001
/*
 * DM_UDEV_DISABLE_SUBSYTEM_RULES_FLAG is set in case we need to disable
 * subsystem udev rules, but still we need the general DM udev rules to
 * be applied (to create the nodes and symlinks under /dev and /dev/disk).
 */
#define DM_UDEV_DISABLE_SUBSYSTEM_RULES_FLAG 0x0002
/*
 * DM_UDEV_DISABLE_DISK_RULES_FLAG is set in case we need to disable
 * general DM rules that set symlinks in /dev/disk directory.
 */
#define DM_UDEV_DISABLE_DISK_RULES_FLAG 0x0004
/*
 * DM_UDEV_DISABLE_OTHER_RULES_FLAG is set in case we need to disable
 * all the other rules that are not general device-mapper nor subsystem
 * related (the rules belong to other software or packages). All foreign
 * rules should check this flag directly and they should ignore further
 * rule processing for such event.
 */
#define DM_UDEV_DISABLE_OTHER_RULES_FLAG 0x0008
/*
 * DM_UDEV_LOW_PRIORITY_FLAG is set in case we need to instruct the
 * udev rules to give low priority to the device that is currently
 * processed. For example, this provides a way to select which symlinks
 * could be overwritten by high priority ones if their names are equal.
 * Common situation is a name based on FS UUID while using origin and
 * snapshot devices.
 */
#define DM_UDEV_LOW_PRIORITY_FLAG 0x0010
/*
 * DM_UDEV_DISABLE_LIBRARY_FALLBACK is set in case we need to disable
 * libdevmapper's node management. We will rely on udev completely
 * and there will be no fallback action provided by libdevmapper if
 * udev does something improperly. Using the library fallback code has
 * a consequence that you need to take into account: any device node
 * or symlink created without udev is not recorded in udev database
 * which other applications may read to get complete list of devices.
 * For this reason, use of DM_UDEV_DISABLE_LIBRARY_FALLBACK is
 * recommended on systems where udev is used. Keep library fallback
 * enabled just for exceptional cases where you need to debug udev-related
 * problems. If you hit such problems, please contact us through upstream
 * LVM2 development mailing list (see also README file). This flag is
 * currently not set by default in libdevmapper so you need to set it
 * explicitly if you're sure that udev is behaving correctly on your
 * setups.
 */
#define DM_UDEV_DISABLE_LIBRARY_FALLBACK 0x0020
/*
 * DM_UDEV_PRIMARY_SOURCE_FLAG is automatically appended by
 * libdevmapper for all ioctls generating udev uevents. Once used in
 * udev rules, we know if this is a real "primary sourced" event or not.
 * We need to distinguish real events originated in libdevmapper from
 * any spurious events to gather all missing information (e.g. events
 * generated as a result of "udevadm trigger" command or as a result
 * of the "watch" udev rule).
 */
#define DM_UDEV_PRIMARY_SOURCE_FLAG 0x0040

/*
 * Udev flags reserved for use by any device-mapper subsystem.
 */
#define DM_SUBSYSTEM_UDEV_FLAG0 0x0100
#define DM_SUBSYSTEM_UDEV_FLAG1 0x0200
#define DM_SUBSYSTEM_UDEV_FLAG2 0x0400
#define DM_SUBSYSTEM_UDEV_FLAG3 0x0800
#define DM_SUBSYSTEM_UDEV_FLAG4 0x1000
#define DM_SUBSYSTEM_UDEV_FLAG5 0x2000
#define DM_SUBSYSTEM_UDEV_FLAG6 0x4000
#define DM_SUBSYSTEM_UDEV_FLAG7 0x8000

int dm_cookie_supported(void);

/*
 * Udev synchronization functions.
 */
void dm_udev_set_sync_support(int sync_with_udev);
int dm_udev_get_sync_support(void);
void dm_udev_set_checking(int checking);
int dm_udev_get_checking(void);

/*
 * Default value to get new auto generated cookie created
 */
#define DM_COOKIE_AUTO_CREATE 0
int dm_udev_create_cookie(uint32_t *cookie);
int dm_udev_complete(uint32_t cookie);
int dm_udev_wait(uint32_t cookie);

/*
 * dm_dev_wait_immediate 
 * If *ready is 1 on return, the wait is complete.
 * If *ready is 0 on return, the wait is incomplete and either
 * this function or dm_udev_wait() must be called again.
 * Returns 0 on error, when neither function should be called again.
 */
int dm_udev_wait_immediate(uint32_t cookie, int *ready);

#define DM_DEV_DIR_UMASK 0022
#define DM_CONTROL_NODE_UMASK 0177

#ifdef __cplusplus
}
#endif
#endif				/* LIB_DEVICE_MAPPER_H */