switch to go vendoring

vendor/github.com/ceph/go-ceph/LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2014 Noah Watkins
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

vendor/github.com/ceph/go-ceph/cephfs/admin/bytecount.go

@@ -0,0 +1,66 @@
+package admin
+
+import (
+	"encoding/json"
+	"fmt"
+)
+
+// ByteCount represents the size of a volume in bytes.
+type ByteCount uint64
+
+// SI byte size constants. keep these private for now.
+const (
+	kibiByte ByteCount = 1024
+	mebiByte           = 1024 * kibiByte
+	gibiByte           = 1024 * mebiByte
+	tebiByte           = 1024 * gibiByte
+)
+
+// resizeValue returns a size value as a string, as needed by the subvolume
+// resize command json.
+func (bc ByteCount) resizeValue() string {
+	return uint64String(uint64(bc))
+}
+
+// QuotaSize interface values can be used to change the size of a volume.
+type QuotaSize interface {
+	resizeValue() string
+}
+
+// specialSize is a custom non-numeric quota size value.
+type specialSize string
+
+// resizeValue for a specialSize returns the original string value.
+func (s specialSize) resizeValue() string {
+	return string(s)
+}
+
+// Infinite is a special QuotaSize value that can be used to clear size limits
+// on a subvolume.
+const Infinite = specialSize("infinite")
+
+// quotaSizePlaceholder types are helpful to extract QuotaSize typed values
+// from JSON responses.
+type quotaSizePlaceholder struct {
+	Value QuotaSize
+}
+
+func (p *quotaSizePlaceholder) UnmarshalJSON(b []byte) error {
+	var val interface{}
+	if err := json.Unmarshal(b, &val); err != nil {
+		return err
+	}
+	switch v := val.(type) {
+	case string:
+		if v == string(Infinite) {
+			p.Value = Infinite
+		} else {
+			return fmt.Errorf("quota size: invalid string value: %q", v)
+		}
+	case float64:
+		p.Value = ByteCount(v)
+	default:
+		return fmt.Errorf("quota size: invalid type, string or number required: %v (%T)", val, val)
+	}
+	return nil
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/clone.go

@@ -0,0 +1,147 @@
+package admin
+
+import (
+	"strings"
+)
+
+const notProtectedSuffix = "is not protected"
+
+// NotProtectedError error values will be returned by CloneSubVolumeSnapshot in
+// the case that the source snapshot needs to be protected but is not.  The
+// requirement for a snapshot to be protected prior to cloning varies by Ceph
+// version.
+type NotProtectedError struct {
+	response
+}
+
+// CloneOptions are used to specify optional values to be used when creating a
+// new subvolume clone.
+type CloneOptions struct {
+	TargetGroup string
+	PoolLayout  string
+}
+
+// CloneSubVolumeSnapshot clones the specified snapshot from the subvolume.
+// The group, subvolume, and snapshot parameters specify the source for the
+// clone, and only the source.  Additional properties of the clone, such as the
+// subvolume group that the clone will be created in and the pool layout may be
+// specified using the clone options parameter.
+//
+// Similar To:
+//  ceph fs subvolume snapshot clone <volume> --group_name=<group> <subvolume> <snapshot> <name> [...]
+func (fsa *FSAdmin) CloneSubVolumeSnapshot(volume, group, subvolume, snapshot, name string, o *CloneOptions) error {
+	m := map[string]string{
+		"prefix":          "fs subvolume snapshot clone",
+		"vol_name":        volume,
+		"sub_name":        subvolume,
+		"snap_name":       snapshot,
+		"target_sub_name": name,
+		"format":          "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	if o != nil && o.TargetGroup != NoGroup {
+		m["target_group_name"] = group
+	}
+	if o != nil && o.PoolLayout != "" {
+		m["pool_layout"] = o.PoolLayout
+	}
+	return checkCloneResponse(fsa.marshalMgrCommand(m))
+}
+
+func checkCloneResponse(res response) error {
+	if strings.HasSuffix(res.Status(), notProtectedSuffix) {
+		return NotProtectedError{response: res}
+	}
+	return res.NoData().End()
+}
+
+// CloneState is used to define constant values used to determine the state of
+// a clone.
+type CloneState string
+
+const (
+	// ClonePending is the state of a pending clone.
+	ClonePending = CloneState("pending")
+	// CloneInProgress is the state of a clone in progress.
+	CloneInProgress = CloneState("in-progress")
+	// CloneComplete is the state of a complete clone.
+	CloneComplete = CloneState("complete")
+	// CloneFailed is the state of a failed clone.
+	CloneFailed = CloneState("failed")
+)
+
+// CloneSource contains values indicating the source of a clone.
+type CloneSource struct {
+	Volume    string `json:"volume"`
+	Group     string `json:"group"`
+	SubVolume string `json:"subvolume"`
+	Snapshot  string `json:"snapshot"`
+}
+
+// CloneStatus reports on the status of a subvolume clone.
+type CloneStatus struct {
+	State  CloneState  `json:"state"`
+	Source CloneSource `json:"source"`
+
+	// failure can be obtained through .GetFailure()
+	failure *CloneFailure
+}
+
+// CloneFailure reports details of a failure after a subvolume clone failed.
+type CloneFailure struct {
+	Errno  string `json:"errno"`
+	ErrStr string `json:"errstr"`
+}
+
+type cloneStatusWrapper struct {
+	Status  CloneStatus  `json:"status"`
+	Failure CloneFailure `json:"failure"`
+}
+
+func parseCloneStatus(res response) (*CloneStatus, error) {
+	var status cloneStatusWrapper
+	if err := res.NoStatus().Unmarshal(&status).End(); err != nil {
+		return nil, err
+	}
+	if status.Failure.Errno != "" || status.Failure.ErrStr != "" {
+		status.Status.failure = &status.Failure
+	}
+	return &status.Status, nil
+}
+
+// CloneStatus returns data reporting the status of a subvolume clone.
+//
+// Similar To:
+//  ceph fs clone status <volume> --group_name=<group> <clone>
+func (fsa *FSAdmin) CloneStatus(volume, group, clone string) (*CloneStatus, error) {
+	m := map[string]string{
+		"prefix":     "fs clone status",
+		"vol_name":   volume,
+		"clone_name": clone,
+		"format":     "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return parseCloneStatus(fsa.marshalMgrCommand(m))
+}
+
+// CancelClone stops the background processes that populate a clone.
+// CancelClone does not delete the clone.
+//
+// Similar To:
+//  ceph fs clone cancel <volume> --group_name=<group> <clone>
+func (fsa *FSAdmin) CancelClone(volume, group, clone string) error {
+	m := map[string]string{
+		"prefix":     "fs clone cancel",
+		"vol_name":   volume,
+		"clone_name": clone,
+		"format":     "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return fsa.marshalMgrCommand(m).NoData().End()
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/clone_nautilus.go

@@ -0,0 +1,10 @@
+package admin
+
+// GetFailure returns details about the CloneStatus when in CloneFailed state.
+//
+// Similar To:
+//  Reading the .failure object from the JSON returned by "ceph fs subvolume
+//  snapshot clone"
+func (cs *CloneStatus) GetFailure() *CloneFailure {
+	return cs.failure
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/doc.go

@@ -0,0 +1,11 @@
+/*
+Package admin is a convenience layer to support the administration of
+CephFS volumes, subvolumes, etc.
+
+Unlike the cephfs package this API does not map to APIs provided by
+ceph libraries themselves. This API is not yet stable and is subject
+to change.
+
+This package only supports ceph "nautilus" and "octopus" at this time.
+*/
+package admin

vendor/github.com/ceph/go-ceph/cephfs/admin/flags.go

@@ -0,0 +1,58 @@
+package admin
+
+// For APIs that accept extra sets of "boolean" flags we may end up wanting
+// multiple different sets of supported flags. Example: most rm functions
+// accept a force flag, but only subvolume delete has retain snapshots.
+// To make this somewhat uniform in the admin package we define a utility
+// interface and helper function to merge flags with naming options.
+
+type flagSet interface {
+	flags() map[string]bool
+}
+
+type commonRmFlags struct {
+	force bool
+}
+
+func (f commonRmFlags) flags() map[string]bool {
+	o := make(map[string]bool)
+	if f.force {
+		o["force"] = true
+	}
+	return o
+}
+
+// SubVolRmFlags does not embed other types to simplify and keep the
+// interface with the type flat and simple. At the cost of some code
+// duplication we get a nicer UX for those using the library.
+
+// SubVolRmFlags may be used to specify behavior modifying flags when
+// removing sub volumes.
+type SubVolRmFlags struct {
+	Force           bool
+	RetainSnapshots bool
+}
+
+func (f SubVolRmFlags) flags() map[string]bool {
+	o := make(map[string]bool)
+	if f.Force {
+		o["force"] = true
+	}
+	if f.RetainSnapshots {
+		o["retain_snapshots"] = true
+	}
+	return o
+}
+
+// mergeFlags combines a set of key-value settings with any type implementing
+// the flagSet interface.
+func mergeFlags(m map[string]string, f flagSet) map[string]interface{} {
+	o := make(map[string]interface{})
+	for k, v := range m {
+		o[k] = v
+	}
+	for k, v := range f.flags() {
+		o[k] = v
+	}
+	return o
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/fsadmin.go

@@ -0,0 +1,136 @@
+package admin
+
+import (
+	"strconv"
+
+	ccom "github.com/ceph/go-ceph/common/commands"
+	"github.com/ceph/go-ceph/internal/commands"
+	"github.com/ceph/go-ceph/rados"
+)
+
+// RadosCommander provides an interface to execute JSON-formatted commands that
+// allow the cephfs administrative functions to interact with the Ceph cluster.
+type RadosCommander = ccom.RadosCommander
+
+// FSAdmin is used to administrate CephFS within a ceph cluster.
+type FSAdmin struct {
+	conn RadosCommander
+}
+
+// New creates an FSAdmin automatically based on the default ceph
+// configuration file. If more customization is needed, create a
+// *rados.Conn as you see fit and use NewFromConn to use that
+// connection with these administrative functions.
+func New() (*FSAdmin, error) {
+	conn, err := rados.NewConn()
+	if err != nil {
+		return nil, err
+	}
+	err = conn.ReadDefaultConfigFile()
+	if err != nil {
+		return nil, err
+	}
+	err = conn.Connect()
+	if err != nil {
+		return nil, err
+	}
+	return NewFromConn(conn), nil
+}
+
+// NewFromConn creates an FSAdmin management object from a preexisting
+// rados connection. The existing connection can be rados.Conn or any
+// type implementing the RadosCommander interface. This may be useful
+// if the calling layer needs to inject additional logging, error handling,
+// fault injection, etc.
+func NewFromConn(conn RadosCommander) *FSAdmin {
+	return &FSAdmin{conn}
+}
+
+func (fsa *FSAdmin) validate() error {
+	if fsa.conn == nil {
+		return rados.ErrNotConnected
+	}
+	return nil
+}
+
+// rawMgrCommand takes a byte buffer and sends it to the MGR as a command.
+// The buffer is expected to contain preformatted JSON.
+func (fsa *FSAdmin) rawMgrCommand(buf []byte) response {
+	return commands.RawMgrCommand(fsa.conn, buf)
+}
+
+// marshalMgrCommand takes an generic interface{} value, converts it to JSON and
+// sends the json to the MGR as a command.
+func (fsa *FSAdmin) marshalMgrCommand(v interface{}) response {
+	return commands.MarshalMgrCommand(fsa.conn, v)
+}
+
+// rawMonCommand takes a byte buffer and sends it to the MON as a command.
+// The buffer is expected to contain preformatted JSON.
+func (fsa *FSAdmin) rawMonCommand(buf []byte) response {
+	return commands.RawMonCommand(fsa.conn, buf)
+}
+
+// marshalMonCommand takes an generic interface{} value, converts it to JSON and
+// sends the json to the MGR as a command.
+func (fsa *FSAdmin) marshalMonCommand(v interface{}) response {
+	return commands.MarshalMonCommand(fsa.conn, v)
+}
+
+type listNamedResult struct {
+	Name string `json:"name"`
+}
+
+func parseListNames(res response) ([]string, error) {
+	var r []listNamedResult
+	if err := res.NoStatus().Unmarshal(&r).End(); err != nil {
+		return nil, err
+	}
+	vl := make([]string, len(r))
+	for i := range r {
+		vl[i] = r[i].Name
+	}
+	return vl, nil
+}
+
+func parseListKeyValues(res response) (map[string]string, error) {
+	var x map[string]string
+	if err := res.NoStatus().Unmarshal(&x).End(); err != nil {
+		return nil, err
+	}
+
+	return x, nil
+}
+
+// parsePathResponse returns a cleaned up path from requests that get a path
+// unless an error is encountered, then an error is returned.
+func parsePathResponse(res response) (string, error) {
+	if res2 := res.NoStatus(); !res2.Ok() {
+		return "", res.End()
+	}
+	b := res.Body()
+	// if there's a trailing newline in the buffer strip it.
+	// ceph assumes a CLI wants the output of the buffer and there's
+	// no format=json mode available currently.
+	for len(b) >= 1 && b[len(b)-1] == '\n' {
+		b = b[:len(b)-1]
+	}
+	return string(b), nil
+}
+
+// modeString converts a unix-style mode value to a string-ified version in an
+// octal representation (e.g. "777", "700", etc). This format is expected by
+// some of the ceph JSON command inputs.
+func modeString(m int, force bool) string {
+	if force || m != 0 {
+		return strconv.FormatInt(int64(m), 8)
+	}
+	return ""
+}
+
+// uint64String converts a uint64 to a string. Some of the ceph json commands
+// can take a string or "int" (as a string). This is a common function for
+// doing that conversion.
+func uint64String(v uint64) string {
+	return strconv.FormatUint(uint64(v), 10)
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/metadata.go

@@ -0,0 +1,102 @@
+//go:build !(nautilus || octopus) && ceph_preview && ceph_pre_quincy
+// +build !nautilus,!octopus,ceph_preview,ceph_pre_quincy
+
+package admin
+
+// GetMetadata gets custom metadata on the subvolume in a volume belonging to
+// an optional subvolume group based on provided key name.
+//
+// Similar To:
+//  ceph fs subvolume metadata get <vol_name> <sub_name> <key_name> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) GetMetadata(volume, group, subvolume, key string) (string, error) {
+	m := map[string]string{
+		"prefix":   "fs subvolume metadata get",
+		"format":   "json",
+		"vol_name": volume,
+		"sub_name": subvolume,
+		"key_name": key,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return parsePathResponse(fsa.marshalMgrCommand(m))
+}
+
+// SetMetadata sets custom metadata on the subvolume in a volume belonging to
+// an optional subvolume group as a key-value pair.
+//
+// Similar To:
+//  ceph fs subvolume metadata set <vol_name> <sub_name> <key_name> <value> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) SetMetadata(volume, group, subvolume, key, value string) error {
+	m := map[string]string{
+		"prefix":   "fs subvolume metadata set",
+		"format":   "json",
+		"vol_name": volume,
+		"sub_name": subvolume,
+		"key_name": key,
+		"value":    value,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return fsa.marshalMgrCommand(m).NoData().End()
+}
+
+// RemoveMetadata removes custom metadata set on the subvolume in a volume
+// belonging to an optional subvolume group using the metadata key.
+//
+// Similar To:
+//  ceph fs subvolume metadata rm <vol_name> <sub_name> <key_name> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) RemoveMetadata(volume, group, subvolume, key string) error {
+	return fsa.rmSubVolumeMetadata(volume, group, subvolume, key, commonRmFlags{})
+}
+
+// ForceRemoveMetadata attempt to forcefully remove custom metadata set on
+// the subvolume in a volume belonging to an optional subvolume group using
+// the metadata key.
+//
+// Similar To:
+//  ceph fs subvolume metadata rm <vol_name> <sub_name> <key_name> [--group_name <subvol_group_name>] --force
+func (fsa *FSAdmin) ForceRemoveMetadata(volume, group, subvolume, key string) error {
+	return fsa.rmSubVolumeMetadata(volume, group, subvolume, key, commonRmFlags{force: true})
+}
+
+func (fsa *FSAdmin) rmSubVolumeMetadata(volume, group, subvolume, key string, o commonRmFlags) error {
+	m := map[string]string{
+		"prefix":   "fs subvolume metadata rm",
+		"format":   "json",
+		"vol_name": volume,
+		"sub_name": subvolume,
+		"key_name": key,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return fsa.marshalMgrCommand(mergeFlags(m, o)).NoData().End()
+}
+
+// ListMetadata lists custom metadata (key-value pairs) set on the subvolume
+// in a volume belonging to an optional subvolume group.
+//
+// Similar To:
+//  ceph fs subvolume metadata ls <vol_name> <sub_name> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) ListMetadata(volume, group, subvolume string) (map[string]string, error) {
+	m := map[string]string{
+		"prefix":   "fs subvolume metadata ls",
+		"format":   "json",
+		"vol_name": volume,
+		"sub_name": subvolume,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return parseListKeyValues(fsa.marshalMgrCommand(m))
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/mgrmodule.go

@@ -0,0 +1,25 @@
+package admin
+
+import (
+	"github.com/ceph/go-ceph/common/admin/manager"
+)
+
+const mirroring = "mirroring"
+
+// EnableMirroringModule will enable the mirroring module for cephfs.
+//
+// Similar To:
+//  ceph mgr module enable mirroring [--force]
+func (fsa *FSAdmin) EnableMirroringModule(force bool) error {
+	mgradmin := manager.NewFromConn(fsa.conn)
+	return mgradmin.EnableModule(mirroring, force)
+}
+
+// DisableMirroringModule will disable the mirroring module for cephfs.
+//
+// Similar To:
+//  ceph mgr module disable mirroring
+func (fsa *FSAdmin) DisableMirroringModule() error {
+	mgradmin := manager.NewFromConn(fsa.conn)
+	return mgradmin.DisableModule(mirroring)
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/mirror.go

@@ -0,0 +1,234 @@
+package admin
+
+import (
+	ccom "github.com/ceph/go-ceph/common/commands"
+	"github.com/ceph/go-ceph/internal/commands"
+)
+
+// SnapshotMirrorAdmin helps administer the snapshot mirroring features of
+// cephfs. Snapshot mirroring is only available in ceph pacific and later.
+type SnapshotMirrorAdmin struct {
+	conn ccom.MgrCommander
+}
+
+// SnapshotMirror returns a new SnapshotMirrorAdmin to be used for the
+// administration of snapshot mirroring features.
+func (fsa *FSAdmin) SnapshotMirror() *SnapshotMirrorAdmin {
+	return &SnapshotMirrorAdmin{conn: fsa.conn}
+}
+
+// Enable snapshot mirroring for the given file system.
+//
+// Similar To:
+//  ceph fs snapshot mirror enable <fs_name>
+func (sma *SnapshotMirrorAdmin) Enable(fsname string) error {
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror enable",
+		"fs_name": fsname,
+		"format":  "json",
+	}
+	return commands.MarshalMgrCommand(sma.conn, m).NoStatus().EmptyBody().End()
+}
+
+// Disable snapshot mirroring for the given file system.
+//
+// Similar To:
+//  ceph fs snapshot mirror disable <fs_name>
+func (sma *SnapshotMirrorAdmin) Disable(fsname string) error {
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror disable",
+		"fs_name": fsname,
+		"format":  "json",
+	}
+	return commands.MarshalMgrCommand(sma.conn, m).NoStatus().EmptyBody().End()
+}
+
+// Add a path in the file system to be mirrored.
+//
+// Similar To:
+//  ceph fs snapshot mirror add <fs_name> <path>
+func (sma *SnapshotMirrorAdmin) Add(fsname, path string) error {
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror add",
+		"fs_name": fsname,
+		"path":    path,
+		"format":  "json",
+	}
+	return commands.MarshalMgrCommand(sma.conn, m).NoStatus().EmptyBody().End()
+}
+
+// Remove a path in the file system from mirroring.
+//
+// Similar To:
+//  ceph fs snapshot mirror remove <fs_name> <path>
+func (sma *SnapshotMirrorAdmin) Remove(fsname, path string) error {
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror remove",
+		"fs_name": fsname,
+		"path":    path,
+		"format":  "json",
+	}
+	return commands.MarshalMgrCommand(sma.conn, m).NoStatus().EmptyBody().End()
+}
+
+type bootstrapTokenResponse struct {
+	Token string `json:"token"`
+}
+
+// CreatePeerBootstrapToken returns a token that can be used to create
+// a peering association between this site an another site.
+//
+// Similar To:
+//  ceph fs snapshot mirror peer_bootstrap create <fs_name> <client_entity> <site-name>
+func (sma *SnapshotMirrorAdmin) CreatePeerBootstrapToken(
+	fsname, client, site string) (string, error) {
+	m := map[string]string{
+		"prefix":      "fs snapshot mirror peer_bootstrap create",
+		"fs_name":     fsname,
+		"client_name": client,
+		"format":      "json",
+	}
+	if site != "" {
+		m["site_name"] = site
+	}
+	var bt bootstrapTokenResponse
+	err := commands.MarshalMgrCommand(sma.conn, m).NoStatus().Unmarshal(&bt).End()
+	return bt.Token, err
+}
+
+// ImportPeerBoostrapToken creates an association between another site, one
+// that has provided a token, with the current site.
+//
+// Similar To:
+//  ceph fs snapshot mirror peer_bootstrap import <fs_name> <token>
+func (sma *SnapshotMirrorAdmin) ImportPeerBoostrapToken(fsname, token string) error {
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror peer_bootstrap import",
+		"fs_name": fsname,
+		"token":   token,
+		"format":  "json",
+	}
+	return commands.MarshalMgrCommand(sma.conn, m).NoStatus().EmptyBody().End()
+}
+
+// DaemonID represents the ID of a cephfs mirroring daemon.
+type DaemonID uint
+
+// FileSystemID represents the ID of a cephfs file system.
+type FileSystemID uint
+
+// PeerUUID represents the UUID of a cephfs mirroring peer.
+type PeerUUID string
+
+// DaemonStatusPeer contains fields detailing a remote peer.
+type DaemonStatusPeer struct {
+	ClientName  string `json:"client_name"`
+	ClusterName string `json:"cluster_name"`
+	FSName      string `json:"fs_name"`
+}
+
+// DaemonStatusPeerStats contains fields detailing the a remote peer's stats.
+type DaemonStatusPeerStats struct {
+	FailureCount  uint64 `json:"failure_count"`
+	RecoveryCount uint64 `json:"recovery_count"`
+}
+
+// DaemonStatusPeerInfo contains fields representing information about a remote peer.
+type DaemonStatusPeerInfo struct {
+	UUID   PeerUUID              `json:"uuid"`
+	Remote DaemonStatusPeer      `json:"remote"`
+	Stats  DaemonStatusPeerStats `json:"stats"`
+}
+
+// DaemonStatusFileSystemInfo represents information about a mirrored file system.
+type DaemonStatusFileSystemInfo struct {
+	FileSystemID   FileSystemID           `json:"filesystem_id"`
+	Name           string                 `json:"name"`
+	DirectoryCount int64                  `json:"directory_count"`
+	Peers          []DaemonStatusPeerInfo `json:"peers"`
+}
+
+// DaemonStatusInfo maps file system IDs to information about that file system.
+type DaemonStatusInfo struct {
+	DaemonID    DaemonID                     `json:"daemon_id"`
+	FileSystems []DaemonStatusFileSystemInfo `json:"filesystems"`
+}
+
+// DaemonStatusResults maps mirroring daemon IDs to information about that
+// mirroring daemon.
+type DaemonStatusResults []DaemonStatusInfo
+
+func parseDaemonStatus(res response) (DaemonStatusResults, error) {
+	var dsr DaemonStatusResults
+	if err := res.NoStatus().Unmarshal(&dsr).End(); err != nil {
+		return nil, err
+	}
+	return dsr, nil
+}
+
+// DaemonStatus returns information on the status of cephfs mirroring daemons
+// associated with the given file system.
+//
+// Similar To:
+//  ceph fs snapshot mirror daemon status <fs_name>
+func (sma *SnapshotMirrorAdmin) DaemonStatus(fsname string) (
+	DaemonStatusResults, error) {
+	// ---
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror daemon status",
+		"fs_name": fsname,
+		"format":  "json",
+	}
+	return parseDaemonStatus(commands.MarshalMgrCommand(sma.conn, m))
+}
+
+// PeerInfo includes information about a cephfs mirroring peer.
+type PeerInfo struct {
+	ClientName string `json:"client_name"`
+	SiteName   string `json:"site_name"`
+	FSName     string `json:"fs_name"`
+	MonHost    string `json:"mon_host"`
+}
+
+// PeerListResults maps a peer's UUID to information about that peer.
+type PeerListResults map[PeerUUID]PeerInfo
+
+func parsePeerList(res response) (PeerListResults, error) {
+	var plr PeerListResults
+	if err := res.NoStatus().Unmarshal(&plr).End(); err != nil {
+		return nil, err
+	}
+	return plr, nil
+}
+
+// PeerList returns information about peers associated with the given file system.
+//
+// Similar To:
+//  ceph fs snapshot mirror peer_list <fs_name>
+func (sma *SnapshotMirrorAdmin) PeerList(fsname string) (
+	PeerListResults, error) {
+	// ---
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror peer_list",
+		"fs_name": fsname,
+		"format":  "json",
+	}
+	return parsePeerList(commands.MarshalMgrCommand(sma.conn, m))
+}
+
+/*
+ DirMap - figure out what last_shuffled is supposed to mean and, if it is a time
+          like it seems to be, how best to represent in Go.
+
+ DirMap TODO
+ceph fs snapshot mirror dirmap
+func (sma *SnapshotMirrorAdmin) DirMap(fsname, path string) error {
+	m := map[string]string{
+		"prefix":  "fs snapshot mirror dirmap",
+		"fs_name": fsname,
+		"path":    path,
+		"format":  "json",
+	}
+	return commands.MarshalMgrCommand(sma.conn, m).NoStatus().EmptyBody().End()
+}
+*/

vendor/github.com/ceph/go-ceph/cephfs/admin/response.go

@@ -0,0 +1,22 @@
+package admin
+
+import (
+	"github.com/ceph/go-ceph/internal/commands"
+)
+
+var (
+	// ErrStatusNotEmpty is an alias for commands.ErrStatusNotEmpty
+	ErrStatusNotEmpty = commands.ErrStatusNotEmpty
+	// ErrBodyNotEmpty is an alias for commands.ErrBodyNotEmpty
+	ErrBodyNotEmpty = commands.ErrBodyNotEmpty
+)
+
+type response = commands.Response
+
+// NotImplementedError is an alias for commands.NotImplementedError.
+type NotImplementedError = commands.NotImplementedError
+
+// newResponse returns a response.
+func newResponse(b []byte, s string, e error) response {
+	return commands.NewResponse(b, s, e)
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/snapshot_metadata.go

@@ -0,0 +1,107 @@
+//go:build !(nautilus || octopus) && ceph_preview && ceph_pre_quincy
+// +build !nautilus,!octopus,ceph_preview,ceph_pre_quincy
+
+package admin
+
+// GetSnapshotMetadata gets custom metadata on the subvolume snapshot in a
+// volume belonging to an optional subvolume group based on provided key name.
+//
+// Similar To:
+//  ceph fs subvolume snapshot metadata get <vol_name> <sub_name> <snap_name> <key_name> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) GetSnapshotMetadata(volume, group, subvolume, snapname, key string) (string, error) {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot metadata get",
+		"format":    "json",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": snapname,
+		"key_name":  key,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return parsePathResponse(fsa.marshalMgrCommand(m))
+}
+
+// SetSnapshotMetadata sets custom metadata on the subvolume snapshot in a
+// volume belonging to an optional subvolume group as a key-value pair.
+//
+// Similar To:
+//  ceph fs subvolume snapshot metadata set <vol_name> <sub_name> <snap_name> <key_name> <value> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) SetSnapshotMetadata(volume, group, subvolume, snapname, key, value string) error {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot metadata set",
+		"format":    "json",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": snapname,
+		"key_name":  key,
+		"value":     value,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return fsa.marshalMgrCommand(m).NoData().End()
+}
+
+// RemoveSnapshotMetadata removes custom metadata set on the subvolume
+// snapshot in a volume belonging to an optional subvolume group using the
+// metadata key.
+//
+// Similar To:
+//  ceph fs subvolume snapshot metadata rm <vol_name> <sub_name> <snap_name> <key_name> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) RemoveSnapshotMetadata(volume, group, subvolume, snapname, key string) error {
+	return fsa.rmSubVolumeSnapShotMetadata(volume, group, subvolume, snapname, key, commonRmFlags{})
+}
+
+// ForceRemoveSnapshotMetadata attempt to forcefully remove custom metadata
+// set on the subvolume snapshot in a volume belonging to an optional
+// subvolume group using the metadata key.
+//
+// Similar To:
+//  ceph fs subvolume snapshot metadata rm <vol_name> <sub_name> <snap_name> <key_name> [--group_name <subvol_group_name>] --force
+func (fsa *FSAdmin) ForceRemoveSnapshotMetadata(volume, group, subvolume, snapname, key string) error {
+	return fsa.rmSubVolumeSnapShotMetadata(volume, group, subvolume, snapname, key, commonRmFlags{force: true})
+}
+
+func (fsa *FSAdmin) rmSubVolumeSnapShotMetadata(volume, group, subvolume, snapname, key string, o commonRmFlags) error {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot metadata rm",
+		"format":    "json",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": snapname,
+		"key_name":  key,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return fsa.marshalMgrCommand(mergeFlags(m, o)).NoData().End()
+}
+
+// ListSnapshotMetadata lists custom metadata (key-value pairs) set on the subvolume
+// snapshot in a volume belonging to an optional subvolume group.
+//
+// Similar To:
+//  ceph fs subvolume snapshot metadata ls <vol_name> <sub_name> <snap_name> [--group_name <subvol_group_name>]
+func (fsa *FSAdmin) ListSnapshotMetadata(volume, group, subvolume, snapname string) (map[string]string, error) {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot metadata ls",
+		"format":    "json",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": snapname,
+	}
+
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+
+	return parseListKeyValues(fsa.marshalMgrCommand(m))
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/subvolume.go

@@ -0,0 +1,410 @@
+package admin
+
+// this is the internal type used to create JSON for ceph.
+// See SubVolumeOptions for the type that users of the library
+// interact with.
+// note that the ceph json takes mode as a string.
+type subVolumeFields struct {
+	Prefix            string    `json:"prefix"`
+	Format            string    `json:"format"`
+	VolName           string    `json:"vol_name"`
+	GroupName         string    `json:"group_name,omitempty"`
+	SubName           string    `json:"sub_name"`
+	Size              ByteCount `json:"size,omitempty"`
+	Uid               int       `json:"uid,omitempty"`
+	Gid               int       `json:"gid,omitempty"`
+	Mode              string    `json:"mode,omitempty"`
+	PoolLayout        string    `json:"pool_layout,omitempty"`
+	NamespaceIsolated bool      `json:"namespace_isolated"`
+}
+
+// SubVolumeOptions are used to specify optional, non-identifying, values
+// to be used when creating a new subvolume.
+type SubVolumeOptions struct {
+	Size              ByteCount
+	Uid               int
+	Gid               int
+	Mode              int
+	PoolLayout        string
+	NamespaceIsolated bool
+}
+
+func (s *SubVolumeOptions) toFields(v, g, n string) *subVolumeFields {
+	return &subVolumeFields{
+		Prefix:            "fs subvolume create",
+		Format:            "json",
+		VolName:           v,
+		GroupName:         g,
+		SubName:           n,
+		Size:              s.Size,
+		Uid:               s.Uid,
+		Gid:               s.Gid,
+		Mode:              modeString(s.Mode, false),
+		PoolLayout:        s.PoolLayout,
+		NamespaceIsolated: s.NamespaceIsolated,
+	}
+}
+
+// NoGroup should be used when an optional subvolume group name is not
+// specified.
+const NoGroup = ""
+
+// CreateSubVolume sends a request to create a CephFS subvolume in a volume,
+// belonging to an optional subvolume group.
+//
+// Similar To:
+//  ceph fs subvolume create <volume> --group-name=<group> <name> ...
+func (fsa *FSAdmin) CreateSubVolume(volume, group, name string, o *SubVolumeOptions) error {
+	if o == nil {
+		o = &SubVolumeOptions{}
+	}
+	f := o.toFields(volume, group, name)
+	return fsa.marshalMgrCommand(f).NoData().End()
+}
+
+// ListSubVolumes returns a list of subvolumes belonging to the volume and
+// optional subvolume group.
+//
+// Similar To:
+//  ceph fs subvolume ls <volume> --group-name=<group>
+func (fsa *FSAdmin) ListSubVolumes(volume, group string) ([]string, error) {
+	m := map[string]string{
+		"prefix":   "fs subvolume ls",
+		"vol_name": volume,
+		"format":   "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return parseListNames(fsa.marshalMgrCommand(m))
+}
+
+// RemoveSubVolume will delete a CephFS subvolume in a volume and optional
+// subvolume group.
+//
+// Similar To:
+//  ceph fs subvolume rm <volume> --group-name=<group> <name>
+func (fsa *FSAdmin) RemoveSubVolume(volume, group, name string) error {
+	return fsa.RemoveSubVolumeWithFlags(volume, group, name, SubVolRmFlags{})
+}
+
+// ForceRemoveSubVolume will delete a CephFS subvolume in a volume and optional
+// subvolume group.
+//
+// Similar To:
+//  ceph fs subvolume rm <volume> --group-name=<group> <name> --force
+func (fsa *FSAdmin) ForceRemoveSubVolume(volume, group, name string) error {
+	return fsa.RemoveSubVolumeWithFlags(volume, group, name, SubVolRmFlags{Force: true})
+}
+
+// RemoveSubVolumeWithFlags will delete a CephFS subvolume in a volume and
+// optional subvolume group. This function accepts a SubVolRmFlags type that
+// can be used to specify flags that modify the operations behavior.
+// Equivalent to RemoveSubVolume with no flags set.
+// Equivalent to ForceRemoveSubVolume if only the "Force" flag is set.
+//
+// Similar To:
+//  ceph fs subvolume rm <volume> --group-name=<group> <name> [...flags...]
+func (fsa *FSAdmin) RemoveSubVolumeWithFlags(volume, group, name string, o SubVolRmFlags) error {
+	m := map[string]string{
+		"prefix":   "fs subvolume rm",
+		"vol_name": volume,
+		"sub_name": name,
+		"format":   "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return fsa.marshalMgrCommand(mergeFlags(m, o)).NoData().End()
+}
+
+type subVolumeResizeFields struct {
+	Prefix    string `json:"prefix"`
+	Format    string `json:"format"`
+	VolName   string `json:"vol_name"`
+	GroupName string `json:"group_name,omitempty"`
+	SubName   string `json:"sub_name"`
+	NewSize   string `json:"new_size"`
+	NoShrink  bool   `json:"no_shrink"`
+}
+
+// SubVolumeResizeResult reports the size values returned by the
+// ResizeSubVolume function, as reported by Ceph.
+type SubVolumeResizeResult struct {
+	BytesUsed    ByteCount `json:"bytes_used"`
+	BytesQuota   ByteCount `json:"bytes_quota"`
+	BytesPercent string    `json:"bytes_pcent"`
+}
+
+// ResizeSubVolume will resize a CephFS subvolume. The newSize value may be a
+// ByteCount or the special Infinite constant. Setting noShrink to true will
+// prevent reducing the size of the volume below the current used size.
+//
+// Similar To:
+//  ceph fs subvolume resize <volume> --group-name=<group> <name> ...
+func (fsa *FSAdmin) ResizeSubVolume(
+	volume, group, name string,
+	newSize QuotaSize, noShrink bool) (*SubVolumeResizeResult, error) {
+
+	f := &subVolumeResizeFields{
+		Prefix:    "fs subvolume resize",
+		Format:    "json",
+		VolName:   volume,
+		GroupName: group,
+		SubName:   name,
+		NewSize:   newSize.resizeValue(),
+		NoShrink:  noShrink,
+	}
+	var result []*SubVolumeResizeResult
+	res := fsa.marshalMgrCommand(f)
+	if err := res.NoStatus().Unmarshal(&result).End(); err != nil {
+		return nil, err
+	}
+	return result[0], nil
+}
+
+// SubVolumePath returns the path to the subvolume from the root of the file system.
+//
+// Similar To:
+//  ceph fs subvolume getpath <volume> --group-name=<group> <name>
+func (fsa *FSAdmin) SubVolumePath(volume, group, name string) (string, error) {
+	m := map[string]string{
+		"prefix":   "fs subvolume getpath",
+		"vol_name": volume,
+		"sub_name": name,
+		// ceph doesn't respond in json for this cmd (even if you ask)
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return parsePathResponse(fsa.marshalMgrCommand(m))
+}
+
+// Feature is used to define constant values for optional features on
+// subvolumes.
+type Feature string
+
+const (
+	// SnapshotCloneFeature indicates a subvolume supports cloning.
+	SnapshotCloneFeature = Feature("snapshot-clone")
+	// SnapshotAutoprotectFeature indicates a subvolume does not require
+	// manually protecting a subvolume before cloning.
+	SnapshotAutoprotectFeature = Feature("snapshot-autoprotect")
+	// SnapshotRetentionFeature indicates a subvolume supports retaining
+	// snapshots on subvolume removal.
+	SnapshotRetentionFeature = Feature("snapshot-retention")
+)
+
+// SubVolumeState is used to define constant value for the state of
+// a subvolume.
+type SubVolumeState string
+
+const (
+	// StateUnset indicates a subvolume without any state.
+	StateUnset = SubVolumeState("")
+	// StateInit indicates that the subvolume is in initializing state.
+	StateInit = SubVolumeState("init")
+	// StatePending indicates that the subvolume is in pending state.
+	StatePending = SubVolumeState("pending")
+	// StateInProgress indicates that the subvolume is in in-progress state.
+	StateInProgress = SubVolumeState("in-progress")
+	// StateFailed indicates that the subvolume is in failed state.
+	StateFailed = SubVolumeState("failed")
+	// StateComplete indicates that the subvolume is in complete state.
+	StateComplete = SubVolumeState("complete")
+	// StateCanceled indicates that the subvolume is in canceled state.
+	StateCanceled = SubVolumeState("canceled")
+	// StateSnapRetained indicates that the subvolume is in
+	// snapshot-retained state.
+	StateSnapRetained = SubVolumeState("snapshot-retained")
+)
+
+// SubVolumeInfo reports various informational values about a subvolume.
+type SubVolumeInfo struct {
+	Type          string         `json:"type"`
+	Path          string         `json:"path"`
+	State         SubVolumeState `json:"state"`
+	Uid           int            `json:"uid"`
+	Gid           int            `json:"gid"`
+	Mode          int            `json:"mode"`
+	BytesPercent  string         `json:"bytes_pcent"`
+	BytesUsed     ByteCount      `json:"bytes_used"`
+	BytesQuota    QuotaSize      `json:"-"`
+	DataPool      string         `json:"data_pool"`
+	PoolNamespace string         `json:"pool_namespace"`
+	Atime         TimeStamp      `json:"atime"`
+	Mtime         TimeStamp      `json:"mtime"`
+	Ctime         TimeStamp      `json:"ctime"`
+	CreatedAt     TimeStamp      `json:"created_at"`
+	Features      []Feature      `json:"features"`
+}
+
+type subVolumeInfoWrapper struct {
+	SubVolumeInfo
+	VBytesQuota *quotaSizePlaceholder `json:"bytes_quota"`
+}
+
+func parseSubVolumeInfo(res response) (*SubVolumeInfo, error) {
+	var info subVolumeInfoWrapper
+	if err := res.NoStatus().Unmarshal(&info).End(); err != nil {
+		return nil, err
+	}
+	if info.VBytesQuota != nil {
+		info.BytesQuota = info.VBytesQuota.Value
+	}
+	return &info.SubVolumeInfo, nil
+}
+
+// SubVolumeInfo returns information about the specified subvolume.
+//
+// Similar To:
+//  ceph fs subvolume info <volume> --group-name=<group> <name>
+func (fsa *FSAdmin) SubVolumeInfo(volume, group, name string) (*SubVolumeInfo, error) {
+	m := map[string]string{
+		"prefix":   "fs subvolume info",
+		"vol_name": volume,
+		"sub_name": name,
+		"format":   "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return parseSubVolumeInfo(fsa.marshalMgrCommand(m))
+}
+
+// CreateSubVolumeSnapshot creates a new snapshot from the source subvolume.
+//
+// Similar To:
+//  ceph fs subvolume snapshot create <volume> --group-name=<group> <source> <name>
+func (fsa *FSAdmin) CreateSubVolumeSnapshot(volume, group, source, name string) error {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot create",
+		"vol_name":  volume,
+		"sub_name":  source,
+		"snap_name": name,
+		"format":    "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return fsa.marshalMgrCommand(m).NoData().End()
+}
+
+// RemoveSubVolumeSnapshot removes the specified snapshot from the subvolume.
+//
+// Similar To:
+//  ceph fs subvolume snapshot rm <volume> --group-name=<group> <subvolume> <name>
+func (fsa *FSAdmin) RemoveSubVolumeSnapshot(volume, group, subvolume, name string) error {
+	return fsa.rmSubVolumeSnapshot(volume, group, subvolume, name, commonRmFlags{})
+}
+
+// ForceRemoveSubVolumeSnapshot removes the specified snapshot from the subvolume.
+//
+// Similar To:
+//  ceph fs subvolume snapshot rm <volume> --group-name=<group> <subvolume> <name> --force
+func (fsa *FSAdmin) ForceRemoveSubVolumeSnapshot(volume, group, subvolume, name string) error {
+	return fsa.rmSubVolumeSnapshot(volume, group, subvolume, name, commonRmFlags{force: true})
+}
+
+func (fsa *FSAdmin) rmSubVolumeSnapshot(volume, group, subvolume, name string, o commonRmFlags) error {
+
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot rm",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": name,
+		"format":    "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return fsa.marshalMgrCommand(mergeFlags(m, o)).NoData().End()
+}
+
+// ListSubVolumeSnapshots returns a listing of snapshots for a given subvolume.
+//
+// Similar To:
+//  ceph fs subvolume snapshot ls <volume> --group-name=<group> <name>
+func (fsa *FSAdmin) ListSubVolumeSnapshots(volume, group, name string) ([]string, error) {
+	m := map[string]string{
+		"prefix":   "fs subvolume snapshot ls",
+		"vol_name": volume,
+		"sub_name": name,
+		"format":   "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return parseListNames(fsa.marshalMgrCommand(m))
+}
+
+// SubVolumeSnapshotInfo reports various informational values about a subvolume.
+type SubVolumeSnapshotInfo struct {
+	CreatedAt        TimeStamp `json:"created_at"`
+	DataPool         string    `json:"data_pool"`
+	HasPendingClones string    `json:"has_pending_clones"`
+	Protected        string    `json:"protected"`
+	Size             ByteCount `json:"size"`
+}
+
+func parseSubVolumeSnapshotInfo(res response) (*SubVolumeSnapshotInfo, error) {
+	var info SubVolumeSnapshotInfo
+	if err := res.NoStatus().Unmarshal(&info).End(); err != nil {
+		return nil, err
+	}
+	return &info, nil
+}
+
+// SubVolumeSnapshotInfo returns information about the specified subvolume snapshot.
+//
+// Similar To:
+//  ceph fs subvolume snapshot info <volume> --group-name=<group> <subvolume> <name>
+func (fsa *FSAdmin) SubVolumeSnapshotInfo(volume, group, subvolume, name string) (*SubVolumeSnapshotInfo, error) {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot info",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": name,
+		"format":    "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return parseSubVolumeSnapshotInfo(fsa.marshalMgrCommand(m))
+}
+
+// ProtectSubVolumeSnapshot protects the specified snapshot.
+//
+// Similar To:
+//  ceph fs subvolume snapshot protect <volume> --group-name=<group> <subvolume> <name>
+func (fsa *FSAdmin) ProtectSubVolumeSnapshot(volume, group, subvolume, name string) error {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot protect",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": name,
+		"format":    "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return fsa.marshalMgrCommand(m).FilterDeprecated().NoData().End()
+}
+
+// UnprotectSubVolumeSnapshot removes protection from the specified snapshot.
+//
+// Similar To:
+//  ceph fs subvolume snapshot unprotect <volume> --group-name=<group> <subvolume> <name>
+func (fsa *FSAdmin) UnprotectSubVolumeSnapshot(volume, group, subvolume, name string) error {
+	m := map[string]string{
+		"prefix":    "fs subvolume snapshot unprotect",
+		"vol_name":  volume,
+		"sub_name":  subvolume,
+		"snap_name": name,
+		"format":    "json",
+	}
+	if group != NoGroup {
+		m["group_name"] = group
+	}
+	return fsa.marshalMgrCommand(m).FilterDeprecated().NoData().End()
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/subvolumegroup.go

@@ -0,0 +1,103 @@
+package admin
+
+// this is the internal type used to create JSON for ceph.
+// See SubVolumeGroupOptions for the type that users of the library
+// interact with.
+// note that the ceph json takes mode as a string.
+type subVolumeGroupFields struct {
+	Prefix     string `json:"prefix"`
+	Format     string `json:"format"`
+	VolName    string `json:"vol_name"`
+	GroupName  string `json:"group_name"`
+	Uid        int    `json:"uid,omitempty"`
+	Gid        int    `json:"gid,omitempty"`
+	Mode       string `json:"mode,omitempty"`
+	PoolLayout string `json:"pool_layout,omitempty"`
+}
+
+// SubVolumeGroupOptions are used to specify optional, non-identifying, values
+// to be used when creating a new subvolume group.
+type SubVolumeGroupOptions struct {
+	Uid        int
+	Gid        int
+	Mode       int
+	PoolLayout string
+}
+
+func (s *SubVolumeGroupOptions) toFields(v, g string) *subVolumeGroupFields {
+	return &subVolumeGroupFields{
+		Prefix:     "fs subvolumegroup create",
+		Format:     "json",
+		VolName:    v,
+		GroupName:  g,
+		Uid:        s.Uid,
+		Gid:        s.Gid,
+		Mode:       modeString(s.Mode, false),
+		PoolLayout: s.PoolLayout,
+	}
+}
+
+// CreateSubVolumeGroup sends a request to create a subvolume group in a volume.
+//
+// Similar To:
+//  ceph fs subvolumegroup create <volume> <group_name>  ...
+func (fsa *FSAdmin) CreateSubVolumeGroup(volume, name string, o *SubVolumeGroupOptions) error {
+	if o == nil {
+		o = &SubVolumeGroupOptions{}
+	}
+	res := fsa.marshalMgrCommand(o.toFields(volume, name))
+	return res.NoData().End()
+}
+
+// ListSubVolumeGroups returns a list of subvolume groups belonging to the
+// specified volume.
+//
+// Similar To:
+//  ceph fs subvolumegroup ls cephfs <volume>
+func (fsa *FSAdmin) ListSubVolumeGroups(volume string) ([]string, error) {
+	res := fsa.marshalMgrCommand(map[string]string{
+		"prefix":   "fs subvolumegroup ls",
+		"vol_name": volume,
+		"format":   "json",
+	})
+	return parseListNames(res)
+}
+
+// RemoveSubVolumeGroup will delete a subvolume group in a volume.
+// Similar To:
+//  ceph fs subvolumegroup rm <volume> <group_name>
+func (fsa *FSAdmin) RemoveSubVolumeGroup(volume, name string) error {
+	return fsa.rmSubVolumeGroup(volume, name, commonRmFlags{})
+}
+
+// ForceRemoveSubVolumeGroup will delete a subvolume group in a volume.
+// Similar To:
+//  ceph fs subvolumegroup rm <volume> <group_name> --force
+func (fsa *FSAdmin) ForceRemoveSubVolumeGroup(volume, name string) error {
+	return fsa.rmSubVolumeGroup(volume, name, commonRmFlags{force: true})
+}
+
+func (fsa *FSAdmin) rmSubVolumeGroup(volume, name string, o commonRmFlags) error {
+	res := fsa.marshalMgrCommand(mergeFlags(map[string]string{
+		"prefix":     "fs subvolumegroup rm",
+		"vol_name":   volume,
+		"group_name": name,
+		"format":     "json",
+	}, o))
+	return res.NoData().End()
+}
+
+// SubVolumeGroupPath returns the path to the subvolume from the root of the
+// file system.
+//
+// Similar To:
+//  ceph fs subvolumegroup getpath <volume> <group_name>
+func (fsa *FSAdmin) SubVolumeGroupPath(volume, name string) (string, error) {
+	m := map[string]string{
+		"prefix":     "fs subvolumegroup getpath",
+		"vol_name":   volume,
+		"group_name": name,
+		// ceph doesn't respond in json for this cmd (even if you ask)
+	}
+	return parsePathResponse(fsa.marshalMgrCommand(m))
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/timestamp.go

@@ -0,0 +1,37 @@
+package admin
+
+import (
+	"encoding/json"
+	"time"
+)
+
+// golang's date parsing approach is rather bizarre
+var cephTSLayout = "2006-01-02 15:04:05"
+
+// TimeStamp abstracts some of the details about date+time stamps
+// returned by ceph via JSON.
+type TimeStamp struct {
+	time.Time
+}
+
+// String returns a string representing the date+time as presented
+// by ceph.
+func (ts TimeStamp) String() string {
+	return ts.Format(cephTSLayout)
+}
+
+// UnmarshalJSON implements the json Unmarshaler interface.
+func (ts *TimeStamp) UnmarshalJSON(b []byte) error {
+	var raw string
+	if err := json.Unmarshal(b, &raw); err != nil {
+		return err
+	}
+	// AFAICT, ceph always returns the time in UTC so Parse, as opposed to
+	// ParseInLocation, is appropriate here.
+	t, err := time.Parse(cephTSLayout, raw)
+	if err != nil {
+		return err
+	}
+	*ts = TimeStamp{t}
+	return nil
+}

vendor/github.com/ceph/go-ceph/cephfs/admin/volume.go

@@ -0,0 +1,183 @@
+package admin
+
+import (
+	"bytes"
+	"encoding/json"
+)
+
+var (
+	listVolumesCmd = []byte(`{"prefix":"fs volume ls"}`)
+	dumpVolumesCmd = []byte(`{"prefix":"fs dump","format":"json"}`)
+	listFsCmd      = []byte(`{"prefix":"fs ls","format":"json"}`)
+)
+
+// ListVolumes return a list of volumes in this Ceph cluster.
+//
+// Similar To:
+//  ceph fs volume ls
+func (fsa *FSAdmin) ListVolumes() ([]string, error) {
+	res := fsa.rawMgrCommand(listVolumesCmd)
+	return parseListNames(res)
+}
+
+// FSPoolInfo contains the name of a file system as well as the metadata and
+// data pools. Pool information is available by ID or by name.
+type FSPoolInfo struct {
+	Name           string   `json:"name"`
+	MetadataPool   string   `json:"metadata_pool"`
+	MetadataPoolID int      `json:"metadata_pool_id"`
+	DataPools      []string `json:"data_pools"`
+	DataPoolIDs    []int    `json:"data_pool_ids"`
+}
+
+// ListFileSystems lists file systems along with the pools occupied by those
+// file systems.
+//
+// Similar To:
+//  ceph fs ls
+func (fsa *FSAdmin) ListFileSystems() ([]FSPoolInfo, error) {
+	res := fsa.rawMonCommand(listFsCmd)
+	return parseFsList(res)
+}
+
+func parseFsList(res response) ([]FSPoolInfo, error) {
+	var listing []FSPoolInfo
+	if err := res.NoStatus().Unmarshal(&listing).End(); err != nil {
+		return nil, err
+	}
+	return listing, nil
+}
+
+// VolumeIdent contains a pair of file system identifying values: the volume
+// name and the volume ID.
+type VolumeIdent struct {
+	Name string
+	ID   int64
+}
+
+type cephFileSystem struct {
+	ID     int64 `json:"id"`
+	MDSMap struct {
+		FSName string `json:"fs_name"`
+	} `json:"mdsmap"`
+}
+
+type fsDump struct {
+	FileSystems []cephFileSystem `json:"filesystems"`
+}
+
+const (
+	dumpOkPrefix = "dumped fsmap epoch"
+	dumpOkLen    = len(dumpOkPrefix)
+
+	invalidTextualResponse = "this ceph version returns a non-parsable volume status response"
+)
+
+func parseDumpToIdents(res response) ([]VolumeIdent, error) {
+	if !res.Ok() {
+		return nil, res.End()
+	}
+	var dump fsDump
+	if err := res.FilterPrefix(dumpOkPrefix).NoStatus().Unmarshal(&dump).End(); err != nil {
+		return nil, err
+	}
+	// copy the dump json into the simpler enumeration list
+	idents := make([]VolumeIdent, len(dump.FileSystems))
+	for i := range dump.FileSystems {
+		idents[i].ID = dump.FileSystems[i].ID
+		idents[i].Name = dump.FileSystems[i].MDSMap.FSName
+	}
+	return idents, nil
+}
+
+// EnumerateVolumes returns a list of volume-name volume-id pairs.
+func (fsa *FSAdmin) EnumerateVolumes() ([]VolumeIdent, error) {
+	// We base our enumeration on the ceph fs dump json.  This may not be the
+	// only way to do it, but it's the only one I know of currently. Because of
+	// this and to keep our initial implementation simple, we expose our own
+	// simplified type only, rather do a partial implementation of dump.
+	return parseDumpToIdents(fsa.rawMonCommand(dumpVolumesCmd))
+}
+
+// VolumePool reports on the pool status for a CephFS volume.
+type VolumePool struct {
+	ID        int    `json:"id"`
+	Name      string `json:"name"`
+	Type      string `json:"type"`
+	Available uint64 `json:"avail"`
+	Used      uint64 `json:"used"`
+}
+
+// VolumeStatus reports various properties of a CephFS volume.
+// TODO: Fill in.
+type VolumeStatus struct {
+	MDSVersion string       `json:"mds_version"`
+	Pools      []VolumePool `json:"pools"`
+}
+
+type mdsVersionField struct {
+	Version string
+	Items   []struct {
+		Version string `json:"version"`
+	}
+}
+
+func (m *mdsVersionField) UnmarshalJSON(data []byte) (err error) {
+	if err = json.Unmarshal(data, &m.Version); err == nil {
+		return
+	}
+	return json.Unmarshal(data, &m.Items)
+}
+
+// volumeStatusResponse deals with the changing output of the mgr
+// api json
+type volumeStatusResponse struct {
+	Pools      []VolumePool    `json:"pools"`
+	MDSVersion mdsVersionField `json:"mds_version"`
+}
+
+func (v *volumeStatusResponse) volumeStatus() *VolumeStatus {
+	vstatus := &VolumeStatus{}
+	vstatus.Pools = v.Pools
+	if v.MDSVersion.Version != "" {
+		vstatus.MDSVersion = v.MDSVersion.Version
+	} else if len(v.MDSVersion.Items) > 0 {
+		vstatus.MDSVersion = v.MDSVersion.Items[0].Version
+	}
+	return vstatus
+}
+
+func parseVolumeStatus(res response) (*volumeStatusResponse, error) {
+	var vs volumeStatusResponse
+	res = res.NoStatus()
+	if !res.Ok() {
+		return nil, res.End()
+	}
+	res = res.Unmarshal(&vs)
+	if !res.Ok() {
+		if bytes.HasPrefix(res.Body(), []byte("ceph")) {
+			return nil, NotImplementedError{
+				Response: newResponse(res.Body(), invalidTextualResponse, res.Unwrap()),
+			}
+		}
+		return nil, res.End()
+	}
+	return &vs, nil
+}
+
+// VolumeStatus returns a VolumeStatus object for the given volume name.
+//
+// Similar To:
+//  ceph fs status cephfs <name>
+func (fsa *FSAdmin) VolumeStatus(name string) (*VolumeStatus, error) {
+	res := fsa.marshalMgrCommand(map[string]string{
+		"fs":     name,
+		"prefix": "fs status",
+		"format": "json",
+	})
+	v, err := parseVolumeStatus(res)
+	if err != nil {
+		return nil, err
+	}
+	return v.volumeStatus(), nil
+}

vendor/github.com/ceph/go-ceph/cephfs/cephfs.go

@@ -0,0 +1,236 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <stdlib.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/retry"
+	"github.com/ceph/go-ceph/rados"
+)
+
+// MountInfo exports ceph's ceph_mount_info from libcephfs.cc
+type MountInfo struct {
+	mount *C.struct_ceph_mount_info
+}
+
+func createMount(id *C.char) (*MountInfo, error) {
+	mount := &MountInfo{}
+	ret := C.ceph_create(&mount.mount, id)
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+	return mount, nil
+}
+
+// validate checks whether mount.mount is ready to use or not.
+func (mount *MountInfo) validate() error {
+	if mount.mount == nil {
+		return ErrNotConnected
+	}
+	return nil
+}
+
+// Version returns the major, minor, and patch level of the libcephfs library.
+func Version() (int, int, int) {
+	var cMajor, cMinor, cPatch C.int
+	C.ceph_version(&cMajor, &cMinor, &cPatch)
+	return int(cMajor), int(cMinor), int(cPatch)
+}
+
+// CreateMount creates a mount handle for interacting with Ceph.
+func CreateMount() (*MountInfo, error) {
+	return createMount(nil)
+}
+
+// CreateMountWithId creates a mount handle for interacting with Ceph.
+// The caller can specify a unique id that will identify this client.
+func CreateMountWithId(id string) (*MountInfo, error) {
+	cid := C.CString(id)
+	defer C.free(unsafe.Pointer(cid))
+	return createMount(cid)
+}
+
+// CreateFromRados creates a mount handle using an existing rados cluster
+// connection.
+//
+// Implements:
+//  int ceph_create_from_rados(struct ceph_mount_info **cmount, rados_t cluster);
+func CreateFromRados(conn *rados.Conn) (*MountInfo, error) {
+	mount := &MountInfo{}
+	ret := C.ceph_create_from_rados(&mount.mount, C.rados_t(conn.Cluster()))
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+	return mount, nil
+}
+
+// ReadDefaultConfigFile loads the ceph configuration from the default config file.
+//
+// Implements:
+//  int ceph_conf_read_file(struct ceph_mount_info *cmount, const char *path_list);
+func (mount *MountInfo) ReadDefaultConfigFile() error {
+	ret := C.ceph_conf_read_file(mount.mount, nil)
+	return getError(ret)
+}
+
+// ReadConfigFile loads the ceph configuration from the specified config file.
+//
+// Implements:
+//  int ceph_conf_read_file(struct ceph_mount_info *cmount, const char *path_list);
+func (mount *MountInfo) ReadConfigFile(path string) error {
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	ret := C.ceph_conf_read_file(mount.mount, cPath)
+	return getError(ret)
+}
+
+// ParseConfigArgv configures the mount using a unix style command line
+// argument vector.
+//
+// Implements:
+//  int ceph_conf_parse_argv(struct ceph_mount_info *cmount, int argc, const char **argv);
+func (mount *MountInfo) ParseConfigArgv(argv []string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	if len(argv) == 0 {
+		return ErrEmptyArgument
+	}
+	cargv := make([]*C.char, len(argv))
+	for i := range argv {
+		cargv[i] = C.CString(argv[i])
+		defer C.free(unsafe.Pointer(cargv[i]))
+	}
+
+	ret := C.ceph_conf_parse_argv(mount.mount, C.int(len(cargv)), &cargv[0])
+	return getError(ret)
+}
+
+// ParseDefaultConfigEnv configures the mount from the default Ceph
+// environment variable CEPH_ARGS.
+//
+// Implements:
+//  int ceph_conf_parse_env(struct ceph_mount_info *cmount, const char *var);
+func (mount *MountInfo) ParseDefaultConfigEnv() error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	ret := C.ceph_conf_parse_env(mount.mount, nil)
+	return getError(ret)
+}
+
+// SetConfigOption sets the value of the configuration option identified by
+// the given name.
+//
+// Implements:
+//  int ceph_conf_set(struct ceph_mount_info *cmount, const char *option, const char *value);
+func (mount *MountInfo) SetConfigOption(option, value string) error {
+	cOption := C.CString(option)
+	defer C.free(unsafe.Pointer(cOption))
+	cValue := C.CString(value)
+	defer C.free(unsafe.Pointer(cValue))
+	return getError(C.ceph_conf_set(mount.mount, cOption, cValue))
+}
+
+// GetConfigOption returns the value of the Ceph configuration option
+// identified by the given name.
+//
+// Implements:
+//  int ceph_conf_get(struct ceph_mount_info *cmount, const char *option, char *buf, size_t len);
+func (mount *MountInfo) GetConfigOption(option string) (string, error) {
+	cOption := C.CString(option)
+	defer C.free(unsafe.Pointer(cOption))
+
+	var (
+		err error
+		buf []byte
+	)
+	// range from 4k to 256KiB
+	retry.WithSizes(4096, 1<<18, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret := C.ceph_conf_get(
+			mount.mount,
+			cOption,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			C.size_t(len(buf)))
+		err = getError(ret)
+		return retry.DoubleSize.If(err == errNameTooLong)
+	})
+	if err != nil {
+		return "", err
+	}
+	value := C.GoString((*C.char)(unsafe.Pointer(&buf[0])))
+	return value, nil
+}
+
+// Init the file system client without actually mounting the file system.
+//
+// Implements:
+//  int ceph_init(struct ceph_mount_info *cmount);
+func (mount *MountInfo) Init() error {
+	return getError(C.ceph_init(mount.mount))
+}
+
+// Mount the file system, establishing a connection capable of I/O.
+//
+// Implements:
+//  int ceph_mount(struct ceph_mount_info *cmount, const char *root);
+func (mount *MountInfo) Mount() error {
+	ret := C.ceph_mount(mount.mount, nil)
+	return getError(ret)
+}
+
+// MountWithRoot mounts the file system using the path provided for the root of
+// the mount. This establishes a connection capable of I/O.
+//
+// Implements:
+//  int ceph_mount(struct ceph_mount_info *cmount, const char *root);
+func (mount *MountInfo) MountWithRoot(root string) error {
+	croot := C.CString(root)
+	defer C.free(unsafe.Pointer(croot))
+	return getError(C.ceph_mount(mount.mount, croot))
+}
+
+// Unmount the file system.
+//
+// Implements:
+//  int ceph_unmount(struct ceph_mount_info *cmount);
+func (mount *MountInfo) Unmount() error {
+	ret := C.ceph_unmount(mount.mount)
+	return getError(ret)
+}
+
+// Release destroys the mount handle.
+//
+// Implements:
+//  int ceph_release(struct ceph_mount_info *cmount);
+func (mount *MountInfo) Release() error {
+	if mount.mount == nil {
+		return nil
+	}
+	ret := C.ceph_release(mount.mount)
+	if err := getError(ret); err != nil {
+		return err
+	}
+	mount.mount = nil
+	return nil
+}
+
+// SyncFs synchronizes all filesystem data to persistent media.
+func (mount *MountInfo) SyncFs() error {
+	ret := C.ceph_sync_fs(mount.mount)
+	return getError(ret)
+}
+
+// IsMounted checks mount status.
+func (mount *MountInfo) IsMounted() bool {
+	ret := C.ceph_is_mounted(mount.mount)
+	return ret == 1
+}

vendor/github.com/ceph/go-ceph/cephfs/command.go

@@ -0,0 +1,63 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <stdlib.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+)
+
+func cephBufferFree(p unsafe.Pointer) {
+	C.ceph_buffer_free((*C.char)(p))
+}
+
+// MdsCommand sends commands to the specified MDS.
+func (mount *MountInfo) MdsCommand(mdsSpec string, args [][]byte) ([]byte, string, error) {
+	return mount.mdsCommand(mdsSpec, args, nil)
+}
+
+// MdsCommandWithInputBuffer sends commands to the specified MDS, with an input
+// buffer.
+func (mount *MountInfo) MdsCommandWithInputBuffer(mdsSpec string, args [][]byte, inputBuffer []byte) ([]byte, string, error) {
+	return mount.mdsCommand(mdsSpec, args, inputBuffer)
+}
+
+// mdsCommand supports sending formatted commands to MDS.
+//
+// Implements:
+//  int ceph_mds_command(struct ceph_mount_info *cmount,
+//      const char *mds_spec,
+//      const char **cmd,
+//      size_t cmdlen,
+//      const char *inbuf, size_t inbuflen,
+//      char **outbuf, size_t *outbuflen,
+//      char **outs, size_t *outslen);
+func (mount *MountInfo) mdsCommand(mdsSpec string, args [][]byte, inputBuffer []byte) (buffer []byte, info string, err error) {
+	spec := C.CString(mdsSpec)
+	defer C.free(unsafe.Pointer(spec))
+	ci := cutil.NewCommandInput(args, inputBuffer)
+	defer ci.Free()
+	co := cutil.NewCommandOutput().SetFreeFunc(cephBufferFree)
+	defer co.Free()
+
+	ret := C.ceph_mds_command(
+		mount.mount, // cephfs mount ref
+		spec,        // mds spec
+		(**C.char)(ci.Cmd()),
+		C.size_t(ci.CmdLen()),
+		(*C.char)(ci.InBuf()),
+		C.size_t(ci.InBufLen()),
+		(**C.char)(co.OutBuf()),
+		(*C.size_t)(co.OutBufLen()),
+		(**C.char)(co.Outs()),
+		(*C.size_t)(co.OutsLen()))
+	buf, status := co.GoValues()
+	return buf, status, getError(ret)
+}

vendor/github.com/ceph/go-ceph/cephfs/conn_nautilus.go

@@ -0,0 +1,28 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+// Some general connectivity and mounting functions are new in
+// Ceph Nautilus.
+
+// GetFsCid returns the cluster ID for a mounted ceph file system.
+// If the object does not refer to a mounted file system, an error
+// will be returned.
+//
+// Note:
+//  Only supported in Ceph Nautilus and newer.
+//
+// Implements:
+//  int64_t ceph_get_fs_cid(struct ceph_mount_info *cmount);
+func (mount *MountInfo) GetFsCid() (int64, error) {
+	ret := C.ceph_get_fs_cid(mount.mount)
+	if ret < 0 {
+		return 0, getError(C.int(ret))
+	}
+	return int64(ret), nil
+}

vendor/github.com/ceph/go-ceph/cephfs/directory.go

@@ -0,0 +1,231 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <stdlib.h>
+#include <dirent.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+// Directory represents an open directory handle.
+type Directory struct {
+	mount *MountInfo
+	dir   *C.struct_ceph_dir_result
+}
+
+// OpenDir returns a new Directory handle open for I/O.
+//
+// Implements:
+//  int ceph_opendir(struct ceph_mount_info *cmount, const char *name, struct ceph_dir_result **dirpp);
+func (mount *MountInfo) OpenDir(path string) (*Directory, error) {
+	var dir *C.struct_ceph_dir_result
+
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_opendir(mount.mount, cPath, &dir)
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+
+	return &Directory{
+		mount: mount,
+		dir:   dir,
+	}, nil
+}
+
+// Close the open directory handle.
+//
+// Implements:
+//  int ceph_closedir(struct ceph_mount_info *cmount, struct ceph_dir_result *dirp);
+func (dir *Directory) Close() error {
+	return getError(C.ceph_closedir(dir.mount.mount, dir.dir))
+}
+
+// Inode represents an inode number in the file system.
+type Inode uint64
+
+// DType values are used to determine, when possible, the file type
+// of a directory entry.
+type DType uint8
+
+const (
+	// DTypeBlk indicates a directory entry is a block device.
+	DTypeBlk = DType(C.DT_BLK)
+	// DTypeChr indicates a directory entry is a character device.
+	DTypeChr = DType(C.DT_CHR)
+	// DTypeDir indicates a directory entry is a directory.
+	DTypeDir = DType(C.DT_DIR)
+	// DTypeFIFO indicates a directory entry is a named pipe (FIFO).
+	DTypeFIFO = DType(C.DT_FIFO)
+	// DTypeLnk indicates a directory entry is a symbolic link.
+	DTypeLnk = DType(C.DT_LNK)
+	// DTypeReg indicates a directory entry is a regular file.
+	DTypeReg = DType(C.DT_REG)
+	// DTypeSock indicates a directory entry is a UNIX domain socket.
+	DTypeSock = DType(C.DT_SOCK)
+	// DTypeUnknown indicates that the file type could not be determined.
+	DTypeUnknown = DType(C.DT_UNKNOWN)
+)
+
+// DirEntry represents an entry within a directory.
+type DirEntry struct {
+	inode Inode
+	name  string
+	dtype DType
+}
+
+// Name returns the directory entry's name.
+func (d *DirEntry) Name() string {
+	return d.name
+}
+
+// Inode returns the directory entry's inode number.
+func (d *DirEntry) Inode() Inode {
+	return d.inode
+}
+
+// DType returns the Directory-entry's Type, indicating if it
+// is a regular file, directory, etc.
+// DType may be unknown and thus require an additional call
+// (stat for example) if Unknown.
+func (d *DirEntry) DType() DType {
+	return d.dtype
+}
+
+// DirEntryPlus is a DirEntry plus additional data (stat) for an entry
+// within a directory.
+type DirEntryPlus struct {
+	DirEntry
+	// statx: the converted statx returned by ceph_readdirplus_r
+	statx *CephStatx
+}
+
+// Statx returns cached stat metadata for the directory entry.
+// This call does not incur an actual file system stat.
+func (d *DirEntryPlus) Statx() *CephStatx {
+	return d.statx
+}
+
+// toDirEntry converts a c struct dirent to our go wrapper.
+func toDirEntry(de *C.struct_dirent) *DirEntry {
+	return &DirEntry{
+		inode: Inode(de.d_ino),
+		name:  C.GoString(&de.d_name[0]),
+		dtype: DType(de.d_type),
+	}
+}
+
+// toDirEntryPlus converts c structs set by ceph_readdirplus_r to our go
+// wrapper.
+func toDirEntryPlus(de *C.struct_dirent, s C.struct_ceph_statx) *DirEntryPlus {
+	return &DirEntryPlus{
+		DirEntry: *toDirEntry(de),
+		statx:    cStructToCephStatx(s),
+	}
+}
+
+// ReadDir reads a single directory entry from the open Directory.
+// A nil DirEntry pointer will be returned when the Directory stream has been
+// exhausted.
+//
+// Implements:
+//  int ceph_readdir_r(struct ceph_mount_info *cmount, struct ceph_dir_result *dirp, struct dirent *de);
+func (dir *Directory) ReadDir() (*DirEntry, error) {
+	var de C.struct_dirent
+	ret := C.ceph_readdir_r(dir.mount.mount, dir.dir, &de)
+	if ret < 0 {
+		return nil, getError(ret)
+	}
+	if ret == 0 {
+		return nil, nil // End-of-stream
+	}
+	return toDirEntry(&de), nil
+}
+
+// ReadDirPlus reads a single directory entry and stat information from the
+// open Directory.
+// A nil DirEntryPlus pointer will be returned when the Directory stream has
+// been exhausted.
+// See Statx for a description of the wants and flags parameters.
+//
+// Implements:
+//  int ceph_readdirplus_r(struct ceph_mount_info *cmount, struct ceph_dir_result *dirp, struct dirent *de,
+//                         struct ceph_statx *stx, unsigned want, unsigned flags, struct Inode **out);
+func (dir *Directory) ReadDirPlus(
+	want StatxMask, flags AtFlags) (*DirEntryPlus, error) {
+
+	var (
+		de C.struct_dirent
+		s  C.struct_ceph_statx
+	)
+	ret := C.ceph_readdirplus_r(
+		dir.mount.mount,
+		dir.dir,
+		&de,
+		&s,
+		C.uint(want),
+		C.uint(flags),
+		nil, // unused, internal Inode type not needed for high level api
+	)
+	if ret < 0 {
+		return nil, getError(ret)
+	}
+	if ret == 0 {
+		return nil, nil // End-of-stream
+	}
+	return toDirEntryPlus(&de, s), nil
+}
+
+// RewindDir sets the directory stream to the beginning of the directory.
+//
+// Implements:
+//  void ceph_rewinddir(struct ceph_mount_info *cmount, struct ceph_dir_result *dirp);
+func (dir *Directory) RewindDir() {
+	C.ceph_rewinddir(dir.mount.mount, dir.dir)
+}
+
+// dirEntries provides a convenient wrapper around slices of DirEntry items.
+// For example, use the Names() call to easily get only the names from a
+// DirEntry slice.
+type dirEntries []*DirEntry
+
+// list returns all the contents of a directory as a dirEntries slice.
+//
+// list is implemented using ReadDir. If any of the calls to ReadDir returns
+// an error List will return an error. However, all previous entries
+// collected will still be returned. Callers of this function may want to check
+// the entries return value even when an error is returned.
+// List rewinds the handle every time it is called to get a full
+// listing of directory contents.
+func (dir *Directory) list() (dirEntries, error) {
+	var (
+		err     error
+		entry   *DirEntry
+		entries = make(dirEntries, 0)
+	)
+	dir.RewindDir()
+	for {
+		entry, err = dir.ReadDir()
+		if err != nil || entry == nil {
+			break
+		}
+		entries = append(entries, entry)
+	}
+	return entries, err
+}
+
+// names returns a slice of only the name fields from dir entries.
+func (entries dirEntries) names() []string {
+	names := make([]string, len(entries))
+	for i, v := range entries {
+		names[i] = v.Name()
+	}
+	return names
+}

vendor/github.com/ceph/go-ceph/cephfs/doc.go

@@ -0,0 +1,4 @@
+/*
+Package cephfs contains a set of wrappers around Ceph's libcephfs API.
+*/
+package cephfs

vendor/github.com/ceph/go-ceph/cephfs/errors.go

@@ -0,0 +1,66 @@
+package cephfs
+
+/*
+#include <errno.h>
+*/
+import "C"
+
+import (
+	"errors"
+
+	"github.com/ceph/go-ceph/internal/errutil"
+)
+
+// cephFSError represents an error condition returned from the CephFS APIs.
+type cephFSError int
+
+// Error returns the error string for the cephFSError type.
+func (e cephFSError) Error() string {
+	return errutil.FormatErrorCode("cephfs", int(e))
+}
+
+func (e cephFSError) ErrorCode() int {
+	return int(e)
+}
+
+func getError(e C.int) error {
+	if e == 0 {
+		return nil
+	}
+	return cephFSError(e)
+}
+
+// getErrorIfNegative converts a ceph return code to error if negative.
+// This is useful for functions that return a usable positive value on
+// success but a negative error number on error.
+func getErrorIfNegative(ret C.int) error {
+	if ret >= 0 {
+		return nil
+	}
+	return getError(ret)
+}
+
+// Public go errors:
+
+var (
+	// ErrEmptyArgument may be returned if a function argument is passed
+	// a zero-length slice or map.
+	ErrEmptyArgument = errors.New("Argument must contain at least one item")
+)
+
+// Public CephFSErrors:
+
+const (
+	// ErrNotConnected may be returned when client is not connected
+	// to a cluster.
+	ErrNotConnected = cephFSError(-C.ENOTCONN)
+)
+
+// Private errors:
+
+const (
+	errInvalid     = cephFSError(-C.EINVAL)
+	errNameTooLong = cephFSError(-C.ENAMETOOLONG)
+	errNoEntry     = cephFSError(-C.ENOENT)
+	errRange       = cephFSError(-C.ERANGE)
+)

vendor/github.com/ceph/go-ceph/cephfs/file.go

@@ -0,0 +1,416 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#define _GNU_SOURCE
+#include <stdlib.h>
+#include <fcntl.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"io"
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+)
+
+const (
+	// SeekSet is used with Seek to set the absolute position in the file.
+	SeekSet = int(C.SEEK_SET)
+	// SeekCur is used with Seek to position the file relative to the current
+	// position.
+	SeekCur = int(C.SEEK_CUR)
+	// SeekEnd is used with Seek to position the file relative to the end.
+	SeekEnd = int(C.SEEK_END)
+)
+
+// SyncChoice is used to control how metadata and/or data is sync'ed to
+// the file system.
+type SyncChoice int
+
+const (
+	// SyncAll will synchronize both data and metadata.
+	SyncAll = SyncChoice(0)
+	// SyncDataOnly will synchronize only data.
+	SyncDataOnly = SyncChoice(1)
+)
+
+// File represents an open file descriptor in cephfs.
+type File struct {
+	mount *MountInfo
+	fd    C.int
+}
+
+// Open a file at the given path. The flags are the same os flags as
+// a local open call. Mode is the same mode bits as a local open call.
+//
+// Implements:
+//  int ceph_open(struct ceph_mount_info *cmount, const char *path, int flags, mode_t mode);
+func (mount *MountInfo) Open(path string, flags int, mode uint32) (*File, error) {
+	if mount.mount == nil {
+		return nil, ErrNotConnected
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	ret := C.ceph_open(mount.mount, cPath, C.int(flags), C.mode_t(mode))
+	if ret < 0 {
+		return nil, getError(ret)
+	}
+	return &File{mount: mount, fd: ret}, nil
+}
+
+func (f *File) validate() error {
+	if f.mount == nil {
+		return ErrNotConnected
+	}
+	return nil
+}
+
+// Close the file.
+//
+// Implements:
+//  int ceph_close(struct ceph_mount_info *cmount, int fd);
+func (f *File) Close() error {
+	if f.fd == -1 {
+		// already closed
+		return nil
+	}
+	if err := f.validate(); err != nil {
+		return err
+	}
+	if err := getError(C.ceph_close(f.mount.mount, f.fd)); err != nil {
+		return err
+	}
+	f.fd = -1
+	return nil
+}
+
+// read directly wraps the ceph_read call. Because read is such a common
+// operation we deviate from the ceph naming and expose Read and ReadAt
+// wrappers for external callers of the library.
+//
+// Implements:
+//  int ceph_read(struct ceph_mount_info *cmount, int fd, char *buf, int64_t size, int64_t offset);
+func (f *File) read(buf []byte, offset int64) (int, error) {
+	if err := f.validate(); err != nil {
+		return 0, err
+	}
+	bufptr := (*C.char)(unsafe.Pointer(&buf[0]))
+	ret := C.ceph_read(
+		f.mount.mount, f.fd, bufptr, C.int64_t(len(buf)), C.int64_t(offset))
+	switch {
+	case ret < 0:
+		return 0, getError(ret)
+	case ret == 0:
+		return 0, io.EOF
+	}
+	return int(ret), nil
+}
+
+// Read data from file. Up to len(buf) bytes will be read from the file.
+// The number of bytes read will be returned.
+// When nothing is left to read from the file, Read returns, 0, io.EOF.
+func (f *File) Read(buf []byte) (int, error) {
+	// to-consider: should we mimic Go's behavior of returning an
+	// io.ErrShortWrite error if write length < buf size?
+	return f.read(buf, -1)
+}
+
+// ReadAt will read data from the file starting at the given offset.
+// Up to len(buf) bytes will be read from the file.
+// The number of bytes read will be returned.
+// When nothing is left to read from the file, ReadAt returns, 0, io.EOF.
+func (f *File) ReadAt(buf []byte, offset int64) (int, error) {
+	if offset < 0 {
+		return 0, errInvalid
+	}
+	return f.read(buf, offset)
+}
+
+// Preadv will read data from the file, starting at the given offset,
+// into the byte-slice data buffers sequentially.
+// The number of bytes read will be returned.
+// When nothing is left to read from the file the return values will be:
+// 0, io.EOF.
+//
+// Implements:
+//  int ceph_preadv(struct ceph_mount_info *cmount, int fd, const struct iovec *iov, int iovcnt,
+//                  int64_t offset);
+func (f *File) Preadv(data [][]byte, offset int64) (int, error) {
+	if err := f.validate(); err != nil {
+		return 0, err
+	}
+	iov := cutil.ByteSlicesToIovec(data)
+	defer iov.Free()
+
+	ret := C.ceph_preadv(
+		f.mount.mount,
+		f.fd,
+		(*C.struct_iovec)(iov.Pointer()),
+		C.int(iov.Len()),
+		C.int64_t(offset))
+	switch {
+	case ret < 0:
+		return 0, getError(ret)
+	case ret == 0:
+		return 0, io.EOF
+	}
+	iov.Sync()
+	return int(ret), nil
+}
+
+// write directly wraps the ceph_write call. Because write is such a common
+// operation we deviate from the ceph naming and expose Write and WriteAt
+// wrappers for external callers of the library.
+//
+// Implements:
+//  int ceph_write(struct ceph_mount_info *cmount, int fd, const char *buf,
+//                 int64_t size, int64_t offset);
+func (f *File) write(buf []byte, offset int64) (int, error) {
+	if err := f.validate(); err != nil {
+		return 0, err
+	}
+	bufptr := (*C.char)(unsafe.Pointer(&buf[0]))
+	ret := C.ceph_write(
+		f.mount.mount, f.fd, bufptr, C.int64_t(len(buf)), C.int64_t(offset))
+	if ret < 0 {
+		return 0, getError(ret)
+	}
+	return int(ret), nil
+}
+
+// Write data from buf to the file.
+// The number of bytes written is returned.
+func (f *File) Write(buf []byte) (int, error) {
+	return f.write(buf, -1)
+}
+
+// WriteAt writes data from buf to the file at the specified offset.
+// The number of bytes written is returned.
+func (f *File) WriteAt(buf []byte, offset int64) (int, error) {
+	if offset < 0 {
+		return 0, errInvalid
+	}
+	return f.write(buf, offset)
+}
+
+// Pwritev writes data from the slice of byte-slice buffers to the file at the
+// specified offset.
+// The number of bytes written is returned.
+//
+// Implements:
+//  int ceph_pwritev(struct ceph_mount_info *cmount, int fd, const struct iovec *iov, int iovcnt,
+//                   int64_t offset);
+func (f *File) Pwritev(data [][]byte, offset int64) (int, error) {
+	if err := f.validate(); err != nil {
+		return 0, err
+	}
+	iov := cutil.ByteSlicesToIovec(data)
+	defer iov.Free()
+
+	ret := C.ceph_pwritev(
+		f.mount.mount,
+		f.fd,
+		(*C.struct_iovec)(iov.Pointer()),
+		C.int(iov.Len()),
+		C.int64_t(offset))
+	if ret < 0 {
+		return 0, getError(ret)
+	}
+	return int(ret), nil
+}
+
+// Seek will reposition the file stream based on the given offset.
+//
+// Implements:
+//  int64_t ceph_lseek(struct ceph_mount_info *cmount, int fd, int64_t offset, int whence);
+func (f *File) Seek(offset int64, whence int) (int64, error) {
+	if err := f.validate(); err != nil {
+		return 0, err
+	}
+	// validate the seek whence value in case the caller skews
+	// from the seek values we technically support from C as documented.
+	// TODO: need to support seek-(hole|data) in mimic and later.
+	switch whence {
+	case SeekSet, SeekCur, SeekEnd:
+	default:
+		return 0, errInvalid
+	}
+
+	ret := C.ceph_lseek(f.mount.mount, f.fd, C.int64_t(offset), C.int(whence))
+	if ret < 0 {
+		return 0, getError(C.int(ret))
+	}
+	return int64(ret), nil
+}
+
+// Fchmod changes the mode bits (permissions) of a file.
+//
+// Implements:
+//  int ceph_fchmod(struct ceph_mount_info *cmount, int fd, mode_t mode);
+func (f *File) Fchmod(mode uint32) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+
+	ret := C.ceph_fchmod(f.mount.mount, f.fd, C.mode_t(mode))
+	return getError(ret)
+}
+
+// Fchown changes the ownership of a file.
+//
+// Implements:
+//  int ceph_fchown(struct ceph_mount_info *cmount, int fd, int uid, int gid);
+func (f *File) Fchown(user uint32, group uint32) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+
+	ret := C.ceph_fchown(f.mount.mount, f.fd, C.int(user), C.int(group))
+	return getError(ret)
+}
+
+// Fstatx returns information about an open file.
+//
+// Implements:
+//  int ceph_fstatx(struct ceph_mount_info *cmount, int fd, struct ceph_statx *stx,
+//                  unsigned int want, unsigned int flags);
+func (f *File) Fstatx(want StatxMask, flags AtFlags) (*CephStatx, error) {
+	if err := f.validate(); err != nil {
+		return nil, err
+	}
+
+	var stx C.struct_ceph_statx
+	ret := C.ceph_fstatx(
+		f.mount.mount,
+		f.fd,
+		&stx,
+		C.uint(want),
+		C.uint(flags),
+	)
+	if err := getError(ret); err != nil {
+		return nil, err
+	}
+	return cStructToCephStatx(stx), nil
+}
+
+// FallocFlags represent flags which determine the operation to be
+// performed on the given range.
+// CephFS supports only following two flags.
+type FallocFlags int
+
+const (
+	// FallocNoFlag means default option.
+	FallocNoFlag = FallocFlags(0)
+	// FallocFlKeepSize specifies that the file size will not be changed.
+	FallocFlKeepSize = FallocFlags(C.FALLOC_FL_KEEP_SIZE)
+	// FallocFlPunchHole specifies that the operation is to deallocate
+	// space and zero the byte range.
+	FallocFlPunchHole = FallocFlags(C.FALLOC_FL_PUNCH_HOLE)
+)
+
+// Fallocate preallocates or releases disk space for the file for the
+// given byte range, the flags determine the operation to be performed
+// on the given range.
+//
+// Implements:
+//	int ceph_fallocate(struct ceph_mount_info *cmount, int fd, int mode,
+//								  int64_t offset, int64_t length);
+func (f *File) Fallocate(mode FallocFlags, offset, length int64) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+	ret := C.ceph_fallocate(f.mount.mount, f.fd, C.int(mode), C.int64_t(offset), C.int64_t(length))
+	return getError(ret)
+}
+
+// LockOp determines operations/type of locks which can be applied on a file.
+type LockOp int
+
+const (
+	// LockSH places a shared lock.
+	// More than one process may hold a shared lock for a given file at a given time.
+	LockSH = LockOp(C.LOCK_SH)
+	// LockEX places an exclusive lock.
+	// Only one process may hold an exclusive lock for a given file at a given time.
+	LockEX = LockOp(C.LOCK_EX)
+	// LockUN removes an existing lock held by this process.
+	LockUN = LockOp(C.LOCK_UN)
+	// LockNB can be ORed with any of the above to make a nonblocking call.
+	LockNB = LockOp(C.LOCK_NB)
+)
+
+// Flock applies or removes an advisory lock on an open file.
+// Param owner is the user-supplied identifier for the owner of the
+// lock, must be an arbitrary integer.
+//
+// Implements:
+//  int ceph_flock(struct ceph_mount_info *cmount, int fd, int operation, uint64_t owner);
+func (f *File) Flock(operation LockOp, owner uint64) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+
+	// validate the operation values before passing it on.
+	switch operation &^ LockNB {
+	case LockSH, LockEX, LockUN:
+	default:
+		return errInvalid
+	}
+
+	ret := C.ceph_flock(f.mount.mount, f.fd, C.int(operation), C.uint64_t(owner))
+	return getError(ret)
+}
+
+// Fsync ensures the file content that may be cached is committed to stable
+// storage.
+// Pass SyncAll to have this call behave like standard fsync and synchronize
+// all data and metadata.
+// Pass SyncDataOnly to have this call behave more like fdatasync (on linux).
+//
+// Implements:
+//  int ceph_fsync(struct ceph_mount_info *cmount, int fd, int syncdataonly);
+func (f *File) Fsync(sync SyncChoice) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+
+	ret := C.ceph_fsync(
+		f.mount.mount,
+		f.fd,
+		C.int(sync),
+	)
+	return getError(ret)
+}
+
+// Sync ensures the file content that may be cached is committed to stable
+// storage.
+// Sync behaves like Go's os package File.Sync function.
+func (f *File) Sync() error {
+	return f.Fsync(SyncAll)
+}
+
+// Truncate sets the size of the open file.
+// NOTE: In some versions of ceph a bug exists where calling ftruncate on a
+// file open for read-only is permitted. The go-ceph wrapper does no additional
+// checking and will inherit the issue on affected versions of ceph.  Please
+// refer to the following issue for details:
+// https://tracker.ceph.com/issues/48202
+//
+// Implements:
+//  int ceph_ftruncate(struct ceph_mount_info *cmount, int fd, int64_t size);
+func (f *File) Truncate(size int64) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+
+	ret := C.ceph_ftruncate(
+		f.mount.mount,
+		f.fd,
+		C.int64_t(size),
+	)
+	return getError(ret)
+}

vendor/github.com/ceph/go-ceph/cephfs/file_xattr.go

@@ -0,0 +1,159 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#define _GNU_SOURCE
+#include <stdlib.h>
+#include <sys/types.h>
+#include <sys/xattr.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+	"github.com/ceph/go-ceph/internal/retry"
+)
+
+// XattrFlags are used to control the behavior of set-xattr calls.
+type XattrFlags int
+
+const (
+	// XattrDefault specifies that set-xattr calls use the default behavior of
+	// creating or updating an xattr.
+	XattrDefault = XattrFlags(0)
+	// XattrCreate specifies that set-xattr calls only set new xattrs.
+	XattrCreate = XattrFlags(C.XATTR_CREATE)
+	// XattrReplace specifies that set-xattr calls only replace existing xattr
+	// values.
+	XattrReplace = XattrFlags(C.XATTR_REPLACE)
+)
+
+// SetXattr sets an extended attribute on the open file.
+//
+// NOTE: Attempting to set an xattr value with an empty value may cause the
+// xattr to be unset on some older versions of ceph.
+// Please refer to https://tracker.ceph.com/issues/46084
+//
+// Implements:
+//  int ceph_fsetxattr(struct ceph_mount_info *cmount, int fd, const char *name,
+//                     const void *value, size_t size, int flags);
+func (f *File) SetXattr(name string, value []byte, flags XattrFlags) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+	if name == "" {
+		return errInvalid
+	}
+	var vptr unsafe.Pointer
+	if len(value) > 0 {
+		vptr = unsafe.Pointer(&value[0])
+	}
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.ceph_fsetxattr(
+		f.mount.mount,
+		f.fd,
+		cName,
+		vptr,
+		C.size_t(len(value)),
+		C.int(flags))
+	return getError(ret)
+}
+
+// GetXattr gets an extended attribute from the open file.
+//
+// Implements:
+//  int ceph_fgetxattr(struct ceph_mount_info *cmount, int fd, const char *name,
+//                     void *value, size_t size);
+func (f *File) GetXattr(name string) ([]byte, error) {
+	if err := f.validate(); err != nil {
+		return nil, err
+	}
+	if name == "" {
+		return nil, errInvalid
+	}
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	var (
+		ret C.int
+		err error
+		buf []byte
+	)
+	// range from 1k to 64KiB
+	retry.WithSizes(1024, 1<<16, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.ceph_fgetxattr(
+			f.mount.mount,
+			f.fd,
+			cName,
+			unsafe.Pointer(&buf[0]),
+			C.size_t(size))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return nil, err
+	}
+	return buf[:ret], nil
+}
+
+// ListXattr returns a slice containing strings for the name of each xattr set
+// on the file.
+//
+// Implements:
+//  int ceph_flistxattr(struct ceph_mount_info *cmount, int fd, char *list, size_t size);
+func (f *File) ListXattr() ([]string, error) {
+	if err := f.validate(); err != nil {
+		return nil, err
+	}
+
+	var (
+		ret C.int
+		err error
+		buf []byte
+	)
+	// range from 1k to 64KiB
+	retry.WithSizes(1024, 1<<16, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.ceph_flistxattr(
+			f.mount.mount,
+			f.fd,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			C.size_t(size))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	names := cutil.SplitSparseBuffer(buf[:ret])
+	return names, nil
+}
+
+// RemoveXattr removes the named xattr from the open file.
+//
+// Implements:
+//  int ceph_fremovexattr(struct ceph_mount_info *cmount, int fd, const char *name);
+func (f *File) RemoveXattr(name string) error {
+	if err := f.validate(); err != nil {
+		return err
+	}
+	if name == "" {
+		return errInvalid
+	}
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.ceph_fremovexattr(
+		f.mount.mount,
+		f.fd,
+		cName)
+	return getError(ret)
+}

vendor/github.com/ceph/go-ceph/cephfs/mount_perms_mimic.go

@@ -0,0 +1,21 @@
+//
+// ceph_mount_perms_set available in mimic & later
+
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+// SetMountPerms applies the given UserPerm to the mount object, which it will
+// then use to define the connection's ownership credentials.
+// This function must be called after Init but before Mount.
+//
+// Implements:
+//  int ceph_mount_perms_set(struct ceph_mount_info *cmount, UserPerm *perm);
+func (mount *MountInfo) SetMountPerms(perm *UserPerm) error {
+	return getError(C.ceph_mount_perms_set(mount.mount, perm.userPerm))
+}

vendor/github.com/ceph/go-ceph/cephfs/path.go

@@ -0,0 +1,194 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <stdlib.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+// CurrentDir gets the current working directory.
+func (mount *MountInfo) CurrentDir() string {
+	if err := mount.validate(); err != nil {
+		return ""
+	}
+	cDir := C.ceph_getcwd(mount.mount)
+	return C.GoString(cDir)
+}
+
+// ChangeDir changes the current working directory.
+func (mount *MountInfo) ChangeDir(path string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_chdir(mount.mount, cPath)
+	return getError(ret)
+}
+
+// MakeDir creates a directory.
+func (mount *MountInfo) MakeDir(path string, mode uint32) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_mkdir(mount.mount, cPath, C.mode_t(mode))
+	return getError(ret)
+}
+
+// RemoveDir removes a directory.
+func (mount *MountInfo) RemoveDir(path string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_rmdir(mount.mount, cPath)
+	return getError(ret)
+}
+
+// Unlink removes a file.
+//
+// Implements:
+//  int ceph_unlink(struct ceph_mount_info *cmount, const char *path);
+func (mount *MountInfo) Unlink(path string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_unlink(mount.mount, cPath)
+	return getError(ret)
+}
+
+// Link creates a new link to an existing file.
+//
+// Implements:
+//  int ceph_link (struct ceph_mount_info *cmount, const char *existing, const char *newname);
+func (mount *MountInfo) Link(oldname, newname string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cOldname := C.CString(oldname)
+	defer C.free(unsafe.Pointer(cOldname))
+
+	cNewname := C.CString(newname)
+	defer C.free(unsafe.Pointer(cNewname))
+
+	ret := C.ceph_link(mount.mount, cOldname, cNewname)
+	return getError(ret)
+}
+
+// Symlink creates a symbolic link to an existing path.
+//
+// Implements:
+//  int ceph_symlink(struct ceph_mount_info *cmount, const char *existing, const char *newname);
+func (mount *MountInfo) Symlink(existing, newname string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cExisting := C.CString(existing)
+	defer C.free(unsafe.Pointer(cExisting))
+
+	cNewname := C.CString(newname)
+	defer C.free(unsafe.Pointer(cNewname))
+
+	ret := C.ceph_symlink(mount.mount, cExisting, cNewname)
+	return getError(ret)
+}
+
+// Readlink returns the value of a symbolic link.
+//
+// Implements:
+//  int ceph_readlink(struct ceph_mount_info *cmount, const char *path, char *buf, int64_t size);
+func (mount *MountInfo) Readlink(path string) (string, error) {
+	if err := mount.validate(); err != nil {
+		return "", err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	buf := make([]byte, 4096)
+	ret := C.ceph_readlink(mount.mount,
+		cPath,
+		(*C.char)(unsafe.Pointer(&buf[0])),
+		C.int64_t(len(buf)))
+	if ret < 0 {
+		return "", getError(ret)
+	}
+
+	return string(buf[:ret]), nil
+}
+
+// Statx returns information about a file/directory.
+//
+// Implements:
+//  int ceph_statx(struct ceph_mount_info *cmount, const char *path, struct ceph_statx *stx,
+//                 unsigned int want, unsigned int flags);
+func (mount *MountInfo) Statx(path string, want StatxMask, flags AtFlags) (*CephStatx, error) {
+	if err := mount.validate(); err != nil {
+		return nil, err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	var stx C.struct_ceph_statx
+	ret := C.ceph_statx(
+		mount.mount,
+		cPath,
+		&stx,
+		C.uint(want),
+		C.uint(flags),
+	)
+	if err := getError(ret); err != nil {
+		return nil, err
+	}
+	return cStructToCephStatx(stx), nil
+}
+
+// Rename a file or directory.
+//
+// Implements:
+//  int ceph_rename(struct ceph_mount_info *cmount, const char *from, const char *to);
+func (mount *MountInfo) Rename(from, to string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cFrom := C.CString(from)
+	defer C.free(unsafe.Pointer(cFrom))
+	cTo := C.CString(to)
+	defer C.free(unsafe.Pointer(cTo))
+
+	ret := C.ceph_rename(mount.mount, cFrom, cTo)
+	return getError(ret)
+}
+
+// Truncate sets the size of the specified file.
+//
+// Implements:
+//  int ceph_truncate(struct ceph_mount_info *cmount, const char *path, int64_t size);
+func (mount *MountInfo) Truncate(path string, size int64) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_truncate(
+		mount.mount,
+		cPath,
+		C.int64_t(size),
+	)
+	return getError(ret)
+}

vendor/github.com/ceph/go-ceph/cephfs/path_xattr.go

@@ -0,0 +1,283 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#define _GNU_SOURCE
+#include <stdlib.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+	"github.com/ceph/go-ceph/internal/retry"
+)
+
+// SetXattr sets an extended attribute on the file at the supplied path.
+//
+// NOTE: Attempting to set an xattr value with an empty value may cause
+// the xattr to be unset. Please refer to https://tracker.ceph.com/issues/46084
+//
+// Implements:
+//  int ceph_setxattr(struct ceph_mount_info *cmount, const char *path, const char *name,
+//                    const void *value, size_t size, int flags);
+func (mount *MountInfo) SetXattr(path, name string, value []byte, flags XattrFlags) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	if name == "" {
+		return errInvalid
+	}
+	var vptr unsafe.Pointer
+	if len(value) > 0 {
+		vptr = unsafe.Pointer(&value[0])
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.ceph_setxattr(
+		mount.mount,
+		cPath,
+		cName,
+		vptr,
+		C.size_t(len(value)),
+		C.int(flags))
+	return getError(ret)
+}
+
+// GetXattr gets an extended attribute from the file at the supplied path.
+//
+// Implements:
+//  int ceph_getxattr(struct ceph_mount_info *cmount, const char *path, const char *name,
+//                    void *value, size_t size);
+func (mount *MountInfo) GetXattr(path, name string) ([]byte, error) {
+	if err := mount.validate(); err != nil {
+		return nil, err
+	}
+	if name == "" {
+		return nil, errInvalid
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	var (
+		ret C.int
+		err error
+		buf []byte
+	)
+	// range from 1k to 64KiB
+	retry.WithSizes(1024, 1<<16, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.ceph_getxattr(
+			mount.mount,
+			cPath,
+			cName,
+			unsafe.Pointer(&buf[0]),
+			C.size_t(size))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return nil, err
+	}
+	return buf[:ret], nil
+}
+
+// ListXattr returns a slice containing strings for the name of each xattr set
+// on the file at the supplied path.
+//
+// Implements:
+//  int ceph_listxattr(struct ceph_mount_info *cmount, const char *path, char *list, size_t size);
+func (mount *MountInfo) ListXattr(path string) ([]string, error) {
+	if err := mount.validate(); err != nil {
+		return nil, err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	var (
+		ret C.int
+		err error
+		buf []byte
+	)
+	// range from 1k to 64KiB
+	retry.WithSizes(1024, 1<<16, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.ceph_listxattr(
+			mount.mount,
+			cPath,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			C.size_t(size))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	names := cutil.SplitSparseBuffer(buf[:ret])
+	return names, nil
+}
+
+// RemoveXattr removes the named xattr from the open file.
+//
+// Implements:
+//  int ceph_removexattr(struct ceph_mount_info *cmount, const char *path, const char *name);
+func (mount *MountInfo) RemoveXattr(path, name string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	if name == "" {
+		return errInvalid
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.ceph_removexattr(
+		mount.mount,
+		cPath,
+		cName)
+	return getError(ret)
+}
+
+// LsetXattr sets an extended attribute on the file at the supplied path.
+//
+// NOTE: Attempting to set an xattr value with an empty value may cause
+// the xattr to be unset. Please refer to https://tracker.ceph.com/issues/46084
+//
+// Implements:
+//  int ceph_lsetxattr(struct ceph_mount_info *cmount, const char *path, const char *name,
+//                    const void *value, size_t size, int flags);
+func (mount *MountInfo) LsetXattr(path, name string, value []byte, flags XattrFlags) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	if name == "" {
+		return errInvalid
+	}
+	var vptr unsafe.Pointer
+	if len(value) > 0 {
+		vptr = unsafe.Pointer(&value[0])
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.ceph_lsetxattr(
+		mount.mount,
+		cPath,
+		cName,
+		vptr,
+		C.size_t(len(value)),
+		C.int(flags))
+	return getError(ret)
+}
+
+// LgetXattr gets an extended attribute from the file at the supplied path.
+//
+// Implements:
+//  int ceph_lgetxattr(struct ceph_mount_info *cmount, const char *path, const char *name,
+//                    void *value, size_t size);
+func (mount *MountInfo) LgetXattr(path, name string) ([]byte, error) {
+	if err := mount.validate(); err != nil {
+		return nil, err
+	}
+	if name == "" {
+		return nil, errInvalid
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	var (
+		ret C.int
+		err error
+		buf []byte
+	)
+	// range from 1k to 64KiB
+	retry.WithSizes(1024, 1<<16, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.ceph_lgetxattr(
+			mount.mount,
+			cPath,
+			cName,
+			unsafe.Pointer(&buf[0]),
+			C.size_t(size))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return nil, err
+	}
+	return buf[:ret], nil
+}
+
+// LlistXattr returns a slice containing strings for the name of each xattr set
+// on the file at the supplied path.
+//
+// Implements:
+//  int ceph_llistxattr(struct ceph_mount_info *cmount, const char *path, char *list, size_t size);
+func (mount *MountInfo) LlistXattr(path string) ([]string, error) {
+	if err := mount.validate(); err != nil {
+		return nil, err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	var (
+		ret C.int
+		err error
+		buf []byte
+	)
+	// range from 1k to 64KiB
+	retry.WithSizes(1024, 1<<16, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.ceph_llistxattr(
+			mount.mount,
+			cPath,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			C.size_t(size))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	names := cutil.SplitSparseBuffer(buf[:ret])
+	return names, nil
+}
+
+// LremoveXattr removes the named xattr from the open file.
+//
+// Implements:
+//  int ceph_lremovexattr(struct ceph_mount_info *cmount, const char *path, const char *name);
+func (mount *MountInfo) LremoveXattr(path, name string) error {
+	if err := mount.validate(); err != nil {
+		return err
+	}
+	if name == "" {
+		return errInvalid
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.ceph_lremovexattr(
+		mount.mount,
+		cPath,
+		cName)
+	return getError(ret)
+}

vendor/github.com/ceph/go-ceph/cephfs/permissions.go

@@ -0,0 +1,40 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <stdlib.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+// Chmod changes the mode bits (permissions) of a file/directory.
+func (mount *MountInfo) Chmod(path string, mode uint32) error {
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_chmod(mount.mount, cPath, C.mode_t(mode))
+	return getError(ret)
+}
+
+// Chown changes the ownership of a file/directory.
+func (mount *MountInfo) Chown(path string, user uint32, group uint32) error {
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_chown(mount.mount, cPath, C.int(user), C.int(group))
+	return getError(ret)
+}
+
+// Lchown changes the ownership of a file/directory/etc without following symbolic links
+func (mount *MountInfo) Lchown(path string, user uint32, group uint32) error {
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	ret := C.ceph_lchown(mount.mount, cPath, C.int(user), C.int(group))
+	return getError(ret)
+}

vendor/github.com/ceph/go-ceph/cephfs/statfs.go

@@ -0,0 +1,74 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <stdlib.h>
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+// CephStatVFS instances are returned from the StatFS call. It reports
+// file-system wide statistics.
+type CephStatVFS struct {
+	// Bsize reports the file system's block size.
+	Bsize int64
+	// Fragment reports the file system's fragment size.
+	Frsize int64
+	// Blocks reports the number of blocks in the file system.
+	Blocks uint64
+	// Bfree reports the number of free blocks.
+	Bfree uint64
+	// Bavail reports the number of free blocks for unprivileged users.
+	Bavail uint64
+	// Files reports the number of inodes in the file system.
+	Files uint64
+	// Ffree reports the number of free indoes.
+	Ffree uint64
+	// Favail reports the number of free indoes for unprivileged users.
+	Favail uint64
+	// Fsid reports the file system ID number.
+	Fsid int64
+	// Flag reports the file system mount flags.
+	Flag int64
+	// Namemax reports the maximum file name length.
+	Namemax int64
+}
+
+// StatFS returns file system wide statistics.
+// NOTE: Many of the statistics fields reported by ceph are not filled in with
+// useful values.
+//
+// Implements:
+//  int ceph_statfs(struct ceph_mount_info *cmount, const char *path, struct statvfs *stbuf);
+func (mount *MountInfo) StatFS(path string) (*CephStatVFS, error) {
+	if err := mount.validate(); err != nil {
+		return nil, err
+	}
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+
+	var statvfs C.struct_statvfs
+	ret := C.ceph_statfs(mount.mount, cPath, &statvfs)
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+	csfs := &CephStatVFS{
+		Bsize:   int64(statvfs.f_bsize),
+		Frsize:  int64(statvfs.f_frsize),
+		Blocks:  uint64(statvfs.f_blocks),
+		Bfree:   uint64(statvfs.f_bfree),
+		Bavail:  uint64(statvfs.f_bavail),
+		Files:   uint64(statvfs.f_files),
+		Ffree:   uint64(statvfs.f_ffree),
+		Favail:  uint64(statvfs.f_favail),
+		Fsid:    int64(statvfs.f_fsid),
+		Flag:    int64(statvfs.f_flag),
+		Namemax: int64(statvfs.f_namemax),
+	}
+	return csfs, nil
+}

vendor/github.com/ceph/go-ceph/cephfs/statx.go

@@ -0,0 +1,156 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <cephfs/libcephfs.h>
+#ifndef AT_STATX_DONT_SYNC
+// for versions earlier than Pacific
+#define AT_STATX_DONT_SYNC AT_NO_ATTR_SYNC
+#endif
+*/
+import "C"
+
+import (
+	ts "github.com/ceph/go-ceph/internal/timespec"
+)
+
+// Timespec is a public type for the internal C 'struct timespec'
+type Timespec ts.Timespec
+
+// StatxMask values contain bit-flags indicating what data should be
+// populated by a statx-type call.
+type StatxMask uint32
+
+const (
+	// StatxMode requests the mode value be filled in.
+	StatxMode = StatxMask(C.CEPH_STATX_MODE)
+	// StatxNlink requests the nlink value be filled in.
+	StatxNlink = StatxMask(C.CEPH_STATX_NLINK)
+	// StatxUid requests the uid value be filled in.
+	StatxUid = StatxMask(C.CEPH_STATX_UID)
+	// StatxRdev requests the rdev value be filled in.
+	StatxRdev = StatxMask(C.CEPH_STATX_RDEV)
+	// StatxAtime requests the access-time value be filled in.
+	StatxAtime = StatxMask(C.CEPH_STATX_ATIME)
+	// StatxMtime requests the modified-time value be filled in.
+	StatxMtime = StatxMask(C.CEPH_STATX_MTIME)
+	// StatxIno requests the inode be filled in.
+	StatxIno = StatxMask(C.CEPH_STATX_INO)
+	// StatxSize requests the size value be filled in.
+	StatxSize = StatxMask(C.CEPH_STATX_SIZE)
+	// StatxBlocks requests the blocks value be filled in.
+	StatxBlocks = StatxMask(C.CEPH_STATX_BLOCKS)
+	// StatxBasicStats requests all the fields that are part of a
+	// traditional stat call.
+	StatxBasicStats = StatxMask(C.CEPH_STATX_BASIC_STATS)
+	// StatxBtime requests the birth-time value be filled in.
+	StatxBtime = StatxMask(C.CEPH_STATX_BTIME)
+	// StatxVersion requests the version value be filled in.
+	StatxVersion = StatxMask(C.CEPH_STATX_VERSION)
+	// StatxAllStats requests all known stat values be filled in.
+	StatxAllStats = StatxMask(C.CEPH_STATX_ALL_STATS)
+)
+
+// AtFlags represent flags to be passed to calls that control how files
+// are used or referenced. For example, not following symlinks.
+type AtFlags uint
+
+const (
+	// AtStatxDontSync requests that the stat call only fetch locally-cached
+	// values if possible, avoiding round trips to a back-end server.
+	AtStatxDontSync = AtFlags(C.AT_STATX_DONT_SYNC)
+	// AtNoAttrSync requests that the stat call only fetch locally-cached
+	// values if possible, avoiding round trips to a back-end server.
+	//
+	// Deprecated: replaced by AtStatxDontSync
+	AtNoAttrSync = AtStatxDontSync
+	// AtSymlinkNofollow indicates the call should not follow symlinks
+	// but operate on the symlink itself.
+	AtSymlinkNofollow = AtFlags(C.AT_SYMLINK_NOFOLLOW)
+)
+
+// NOTE: CephStatx fields are meant to be settable by the callers.
+// This is the primary reason we use public fields and not accessors
+// for the CephStatx type.
+
+// CephStatx instances are returned by extended stat (statx) calls.
+// Note that CephStatx results are similar to but not identical
+// to (Linux) system statx results.
+type CephStatx struct {
+	// Mask is a bitmask indicating what fields have been set.
+	Mask StatxMask
+	// Blksize represents the file system's block size.
+	Blksize uint32
+	// Nlink is the number of links for the file.
+	Nlink uint32
+	// Uid (user id) value for the file.
+	Uid uint32
+	// Gid (group id) value for the file.
+	Gid uint32
+	// Mode is the file's type and mode value.
+	Mode uint16
+	// Inode value for the file.
+	Inode Inode
+	// Size of the file in bytes.
+	Size uint64
+	// Blocks indicates the number of blocks allocated to the file.
+	Blocks uint64
+	// Dev describes the device containing this file system.
+	Dev uint64
+	// Rdev describes the device of this file, if the file is a device.
+	Rdev uint64
+	// Atime is the access time of this file.
+	Atime Timespec
+	// Ctime is the status change time of this file.
+	Ctime Timespec
+	// Mtime is the modification time of this file.
+	Mtime Timespec
+	// Btime is the creation (birth) time of this file.
+	Btime Timespec
+	// Version value for the file.
+	Version uint64
+}
+
+func cStructToCephStatx(s C.struct_ceph_statx) *CephStatx {
+	return &CephStatx{
+		Mask:    StatxMask(s.stx_mask),
+		Blksize: uint32(s.stx_blksize),
+		Nlink:   uint32(s.stx_nlink),
+		Uid:     uint32(s.stx_uid),
+		Gid:     uint32(s.stx_gid),
+		Mode:    uint16(s.stx_mode),
+		Inode:   Inode(s.stx_ino),
+		Size:    uint64(s.stx_size),
+		Blocks:  uint64(s.stx_blocks),
+		Dev:     uint64(s.stx_dev),
+		Rdev:    uint64(s.stx_rdev),
+		Atime:   Timespec(ts.CStructToTimespec(ts.CTimespecPtr(&s.stx_atime))),
+		Ctime:   Timespec(ts.CStructToTimespec(ts.CTimespecPtr(&s.stx_ctime))),
+		Mtime:   Timespec(ts.CStructToTimespec(ts.CTimespecPtr(&s.stx_mtime))),
+		Btime:   Timespec(ts.CStructToTimespec(ts.CTimespecPtr(&s.stx_btime))),
+		Version: uint64(s.stx_version),
+	}
+}
+
+/* TODO:
+   - enable later when we can test round -trips
+   - add time fields
+
+func (c *CephStatx) toCStruct() C.struct_ceph_statx {
+	var s C.struct_ceph_statx
+	s.stx_mask = C.uint32_t(c.Mask)
+	s.stx_blksize = C.uint32_t(c.Blksize)
+	s.stx_nlink = C.uint32_t(c.Nlink)
+	s.stx_uid = C.uint32_t(c.Uid)
+	s.stx_gid = C.uint32_t(c.Gid)
+	s.stx_mode = C.uint16_t(c.Mode)
+	s.stx_ino = C.uint64_t(c.Inode)
+	s.stx_size = C.uint64_t(c.Size)
+	s.stx_blocks = C.uint64_t(c.Blocks)
+	s.stx_dev = C.uint64_t(c.Dev)
+	s.stx_rdev = C.uint64_t(c.Rdev)
+	s.stx_version = C.uint64_t(c.Version)
+	return s
+}
+*/

vendor/github.com/ceph/go-ceph/cephfs/userperm.go

@@ -0,0 +1,77 @@
+package cephfs
+
+/*
+#cgo LDFLAGS: -lcephfs
+#cgo CPPFLAGS: -D_FILE_OFFSET_BITS=64
+#include <cephfs/libcephfs.h>
+*/
+import "C"
+
+import (
+	"runtime"
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/log"
+)
+
+// UserPerm types may be used to get or change the credentials used by the
+// connection or some operations.
+type UserPerm struct {
+	userPerm *C.UserPerm
+
+	// cache create-time params
+	managed bool // if set, the userPerm was created by go-ceph
+	uid     C.uid_t
+	gid     C.gid_t
+	gidList []C.gid_t
+}
+
+// NewUserPerm creates a UserPerm pointer and the underlying ceph resources.
+//
+// Implements:
+//  UserPerm *ceph_userperm_new(uid_t uid, gid_t gid, int ngids, gid_t *gidlist);
+func NewUserPerm(uid, gid int, gidlist []int) *UserPerm {
+	// the C code does not copy the content of the gid list so we keep the
+	// inputs stashed in the go type. For completeness we stash everything.
+	p := &UserPerm{
+		managed: true,
+		uid:     C.uid_t(uid),
+		gid:     C.gid_t(gid),
+		gidList: make([]C.gid_t, len(gidlist)),
+	}
+	var cgids *C.gid_t
+	if len(p.gidList) > 0 {
+		for i, gid := range gidlist {
+			p.gidList[i] = C.gid_t(gid)
+		}
+		cgids = (*C.gid_t)(unsafe.Pointer(&p.gidList[0]))
+	}
+	p.userPerm = C.ceph_userperm_new(
+		p.uid, p.gid, C.int(len(p.gidList)), cgids)
+	// if the go object is unreachable, we would like to free the c-memory
+	// since this has no other resources than memory associated with it.
+	// This is only valid for UserPerm objects created by new, and thus have
+	// the managed var set.
+	runtime.SetFinalizer(p, destroyUserPerm)
+	return p
+}
+
+// Destroy will explicitly free ceph resources associated with the UserPerm.
+//
+// Implements:
+//  void ceph_userperm_destroy(UserPerm *perm);
+func (p *UserPerm) Destroy() {
+	if p.userPerm == nil || !p.managed {
+		return
+	}
+	C.ceph_userperm_destroy(p.userPerm)
+	p.userPerm = nil
+	p.gidList = nil
+}
+
+func destroyUserPerm(p *UserPerm) {
+	if p.userPerm != nil && p.managed {
+		log.Warnf("unreachable UserPerm object has not been destroyed. Cleaning up.")
+	}
+	p.Destroy()
+}

vendor/github.com/ceph/go-ceph/common/admin/manager/admin.go

@@ -0,0 +1,17 @@
+package manager
+
+import (
+	ccom "github.com/ceph/go-ceph/common/commands"
+)
+
+// MgrAdmin is used to administrate ceph's manager (mgr).
+type MgrAdmin struct {
+	conn ccom.RadosCommander
+}
+
+// NewFromConn creates an new management object from a preexisting
+// rados connection. The existing connection can be rados.Conn or any
+// type implementing the RadosCommander interface.
+func NewFromConn(conn ccom.RadosCommander) *MgrAdmin {
+	return &MgrAdmin{conn}
+}

vendor/github.com/ceph/go-ceph/common/admin/manager/doc.go

@@ -0,0 +1,5 @@
+/*
+Package manager from common/admin contains a set of APIs used to interact
+with and administer the Ceph manager (mgr).
+*/
+package manager

vendor/github.com/ceph/go-ceph/common/admin/manager/module.go

@@ -0,0 +1,75 @@
+package manager
+
+import (
+	"github.com/ceph/go-ceph/internal/commands"
+)
+
+// EnableModule will enable the specified manager module.
+//
+// Similar To:
+//  ceph mgr module enable <module> [--force]
+func (fsa *MgrAdmin) EnableModule(module string, force bool) error {
+	m := map[string]string{
+		"prefix": "mgr module enable",
+		"module": module,
+		"format": "json",
+	}
+	if force {
+		m["force"] = "--force"
+	}
+	// Why is this _only_ part of the mon command json? You'd think a mgr
+	// command would be available as a MgrCommand but I couldn't figure it out.
+	return commands.MarshalMonCommand(fsa.conn, m).NoData().End()
+}
+
+// DisableModule will disable the specified manager module.
+//
+// Similar To:
+//  ceph mgr module disable <module>
+func (fsa *MgrAdmin) DisableModule(module string) error {
+	m := map[string]string{
+		"prefix": "mgr module disable",
+		"module": module,
+		"format": "json",
+	}
+	return commands.MarshalMonCommand(fsa.conn, m).NoData().End()
+}
+
+// DisabledModule describes a disabled Ceph mgr module.
+// The Ceph JSON structure contains a complex module_options
+// substructure that go-ceph does not currently implement.
+type DisabledModule struct {
+	Name        string `json:"name"`
+	CanRun      bool   `json:"can_run"`
+	ErrorString string `json:"error_string"`
+}
+
+// ModuleInfo contains fields that report the status of modules within the
+// ceph mgr.
+type ModuleInfo struct {
+	// EnabledModules lists the names of the enabled modules.
+	EnabledModules []string `json:"enabled_modules"`
+	// AlwaysOnModules lists the names of the always-on modules.
+	AlwaysOnModules []string `json:"always_on_modules"`
+	// DisabledModules lists structures describing modules that are
+	// not currently enabled.
+	DisabledModules []DisabledModule `json:"disabled_modules"`
+}
+
+func parseModuleInfo(res commands.Response) (*ModuleInfo, error) {
+	m := &ModuleInfo{}
+	if err := res.NoStatus().Unmarshal(m).End(); err != nil {
+		return nil, err
+	}
+	return m, nil
+}
+
+// ListModules returns a module info struct reporting the lists of
+// enabled, disabled, and always-on modules in the Ceph mgr.
+func (fsa *MgrAdmin) ListModules() (*ModuleInfo, error) {
+	m := map[string]string{
+		"prefix": "mgr module ls",
+		"format": "json",
+	}
+	return parseModuleInfo(commands.MarshalMonCommand(fsa.conn, m))
+}

vendor/github.com/ceph/go-ceph/common/commands/doc.go

@@ -0,0 +1,7 @@
+/*
+Package commands provides types and utility functions that are used for
+interfacing with the JSON based command infrastructure in Ceph.
+
+The *rados.Conn type implements many of the interfaces found in this package.
+*/
+package commands

vendor/github.com/ceph/go-ceph/common/commands/interfaces.go

@@ -0,0 +1,20 @@
+package commands
+
+// MgrCommander in an interface for the API needed to execute JSON formatted
+// commands on the ceph mgr.
+type MgrCommander interface {
+	MgrCommand(buf [][]byte) ([]byte, string, error)
+}
+
+// MonCommander is an interface for the API needed to execute JSON formatted
+// commands on the ceph mon(s).
+type MonCommander interface {
+	MonCommand(buf []byte) ([]byte, string, error)
+}
+
+// RadosCommander provides an interface for APIs needed to execute JSON
+// formatted commands on the Ceph cluster.
+type RadosCommander interface {
+	MgrCommander
+	MonCommander
+}

vendor/github.com/ceph/go-ceph/internal/commands/commands.go

@@ -0,0 +1,53 @@
+package commands
+
+import (
+	"encoding/json"
+
+	ccom "github.com/ceph/go-ceph/common/commands"
+	"github.com/ceph/go-ceph/rados"
+)
+
+func validate(m interface{}) error {
+	if m == nil {
+		return rados.ErrNotConnected
+	}
+	return nil
+}
+
+// RawMgrCommand takes a byte buffer and sends it to the MGR as a command.
+// The buffer is expected to contain preformatted JSON.
+func RawMgrCommand(m ccom.MgrCommander, buf []byte) Response {
+	if err := validate(m); err != nil {
+		return Response{err: err}
+	}
+	return NewResponse(m.MgrCommand([][]byte{buf}))
+}
+
+// MarshalMgrCommand takes an generic interface{} value, converts it to JSON
+// and sends the json to the MGR as a command.
+func MarshalMgrCommand(m ccom.MgrCommander, v interface{}) Response {
+	b, err := json.Marshal(v)
+	if err != nil {
+		return Response{err: err}
+	}
+	return RawMgrCommand(m, b)
+}
+
+// RawMonCommand takes a byte buffer and sends it to the MON as a command.
+// The buffer is expected to contain preformatted JSON.
+func RawMonCommand(m ccom.MonCommander, buf []byte) Response {
+	if err := validate(m); err != nil {
+		return Response{err: err}
+	}
+	return NewResponse(m.MonCommand(buf))
+}
+
+// MarshalMonCommand takes an generic interface{} value, converts it to JSON
+// and sends the json to the MGR as a command.
+func MarshalMonCommand(m ccom.MonCommander, v interface{}) Response {
+	b, err := json.Marshal(v)
+	if err != nil {
+		return Response{err: err}
+	}
+	return RawMonCommand(m, b)
+}

vendor/github.com/ceph/go-ceph/internal/commands/response.go

@@ -0,0 +1,193 @@
+package commands
+
+import (
+	"bytes"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"strings"
+)
+
+var (
+	// ErrStatusNotEmpty may be returned if a call should not have a status
+	// string set but one is.
+	ErrStatusNotEmpty = errors.New("response status not empty")
+	// ErrBodyNotEmpty may be returned if a call should have an empty body but
+	// a body value is present.
+	ErrBodyNotEmpty = errors.New("response body not empty")
+)
+
+const (
+	deprecatedSuffix = "call is deprecated and will be removed in a future release"
+	missingPrefix    = "No handler found"
+	einval           = -22
+)
+
+type cephError interface {
+	ErrorCode() int
+}
+
+// NotImplementedError error values will be returned in the case that an API
+// call is not available in the version of Ceph that is running in the target
+// cluster.
+type NotImplementedError struct {
+	Response
+}
+
+// Error implements the error interface.
+func (e NotImplementedError) Error() string {
+	return fmt.Sprintf("API call not implemented server-side: %s", e.status)
+}
+
+// Response encapsulates the data returned by ceph and supports easy processing
+// pipelines.
+type Response struct {
+	body   []byte
+	status string
+	err    error
+}
+
+// Ok returns true if the response contains no error.
+func (r Response) Ok() bool {
+	return r.err == nil
+}
+
+// Error implements the error interface.
+func (r Response) Error() string {
+	if r.status == "" {
+		return r.err.Error()
+	}
+	return fmt.Sprintf("%s: %q", r.err, r.status)
+}
+
+// Unwrap returns the error this response contains.
+func (r Response) Unwrap() error {
+	return r.err
+}
+
+// Status returns the status string value.
+func (r Response) Status() string {
+	return r.status
+}
+
+// Body returns the response body as a raw byte-slice.
+func (r Response) Body() []byte {
+	return r.body
+}
+
+// End returns an error if the response contains an error or nil, indicating
+// that response is no longer needed for processing.
+func (r Response) End() error {
+	if !r.Ok() {
+		if ce, ok := r.err.(cephError); ok {
+			if ce.ErrorCode() == einval && strings.HasPrefix(r.status, missingPrefix) {
+				return NotImplementedError{Response: r}
+			}
+		}
+		return r
+	}
+	return nil
+}
+
+// NoStatus asserts that the input response has no status value.
+func (r Response) NoStatus() Response {
+	if !r.Ok() {
+		return r
+	}
+	if r.status != "" {
+		return Response{r.body, r.status, ErrStatusNotEmpty}
+	}
+	return r
+}
+
+// NoBody asserts that the input response has no body value.
+func (r Response) NoBody() Response {
+	if !r.Ok() {
+		return r
+	}
+	if len(r.body) != 0 {
+		return Response{r.body, r.status, ErrBodyNotEmpty}
+	}
+	return r
+}
+
+// EmptyBody is similar to NoBody but also accepts an empty JSON object.
+func (r Response) EmptyBody() Response {
+	if !r.Ok() {
+		return r
+	}
+	if len(r.body) != 0 {
+		d := map[string]interface{}{}
+		if err := json.Unmarshal(r.body, &d); err != nil {
+			return Response{r.body, r.status, err}
+		}
+		if len(d) != 0 {
+			return Response{r.body, r.status, ErrBodyNotEmpty}
+		}
+	}
+	return r
+}
+
+// NoData asserts that the input response has no status or body values.
+func (r Response) NoData() Response {
+	return r.NoStatus().NoBody()
+}
+
+// FilterPrefix sets the status value to an empty string if the status
+// value contains the given prefix string.
+func (r Response) FilterPrefix(p string) Response {
+	if !r.Ok() {
+		return r
+	}
+	if strings.HasPrefix(r.status, p) {
+		return Response{r.body, "", r.err}
+	}
+	return r
+}
+
+// FilterSuffix sets the status value to an empty string if the status
+// value contains the given suffix string.
+func (r Response) FilterSuffix(s string) Response {
+	if !r.Ok() {
+		return r
+	}
+	if strings.HasSuffix(r.status, s) {
+		return Response{r.body, "", r.err}
+	}
+	return r
+}
+
+// FilterBodyPrefix sets the body value equivalent to an empty string if the
+// body value contains the given prefix string.
+func (r Response) FilterBodyPrefix(p string) Response {
+	if !r.Ok() {
+		return r
+	}
+	if bytes.HasPrefix(r.body, []byte(p)) {
+		return Response{[]byte(""), r.status, r.err}
+	}
+	return r
+}
+
+// FilterDeprecated removes deprecation warnings from the response status.
+// Use it when checking the response from calls that may be deprecated in ceph
+// if you want those calls to continue working if the warning is present.
+func (r Response) FilterDeprecated() Response {
+	return r.FilterSuffix(deprecatedSuffix)
+}
+
+// Unmarshal data from the response body into v.
+func (r Response) Unmarshal(v interface{}) Response {
+	if !r.Ok() {
+		return r
+	}
+	if err := json.Unmarshal(r.body, v); err != nil {
+		return Response{body: r.body, err: err}
+	}
+	return r
+}
+
+// NewResponse returns a response.
+func NewResponse(b []byte, s string, e error) Response {
+	return Response{b, s, e}
+}

vendor/github.com/ceph/go-ceph/internal/commands/trace.go

@@ -0,0 +1,53 @@
+package commands
+
+import (
+	"fmt"
+
+	ccom "github.com/ceph/go-ceph/common/commands"
+)
+
+// NewTraceCommander is a RadosCommander that wraps a given RadosCommander
+// and when commands are executes prints debug level "traces" to the
+// standard output.
+func NewTraceCommander(c ccom.RadosCommander) ccom.RadosCommander {
+	return &tracingCommander{c}
+}
+
+// tracingCommander serves two purposes: first, it allows one to trace the
+// input and output json when running the tests. It can help with actually
+// debugging the tests. Second, it demonstrates the rationale for using an
+// interface in FSAdmin. You can layer any sort of debugging, error injection,
+// or whatnot between the FSAdmin layer and the RADOS layer.
+type tracingCommander struct {
+	conn ccom.RadosCommander
+}
+
+func (t *tracingCommander) MgrCommand(buf [][]byte) ([]byte, string, error) {
+	fmt.Println("(MGR Command)")
+	for i := range buf {
+		fmt.Println("IN:", string(buf[i]))
+	}
+	r, s, err := t.conn.MgrCommand(buf)
+	fmt.Println("OUT(result):", string(r))
+	if s != "" {
+		fmt.Println("OUT(status):", s)
+	}
+	if err != nil {
+		fmt.Println("OUT(error):", err.Error())
+	}
+	return r, s, err
+}
+
+func (t *tracingCommander) MonCommand(buf []byte) ([]byte, string, error) {
+	fmt.Println("(MON Command)")
+	fmt.Println("IN:", string(buf))
+	r, s, err := t.conn.MonCommand(buf)
+	fmt.Println("OUT(result):", string(r))
+	if s != "" {
+		fmt.Println("OUT(status):", s)
+	}
+	if err != nil {
+		fmt.Println("OUT(error):", err.Error())
+	}
+	return r, s, err
+}

vendor/github.com/ceph/go-ceph/internal/cutil/aliases.go

@@ -0,0 +1,66 @@
+package cutil
+
+/*
+#include <stdlib.h>
+#include <string.h>
+typedef void* voidptr;
+*/
+import "C"
+
+import (
+	"math"
+	"unsafe"
+)
+
+const (
+	// MaxIdx is the maximum index on 32 bit systems
+	MaxIdx = math.MaxInt32 // 2GB, max int32 value, should be safe
+
+	// PtrSize is the size of a pointer
+	PtrSize = C.sizeof_voidptr
+
+	// SizeTSize is the size of C.size_t
+	SizeTSize = C.sizeof_size_t
+)
+
+// Compile-time assertion ensuring that Go's `int` is at least as large as C's.
+const _ = unsafe.Sizeof(int(0)) - C.sizeof_int
+
+// SizeT wraps size_t from C.
+type SizeT C.size_t
+
+// This section contains a bunch of types that are basically just
+// unsafe.Pointer but have specific types to help "self document" what the
+// underlying pointer is really meant to represent.
+
+// CPtr is an unsafe.Pointer to C allocated memory
+type CPtr unsafe.Pointer
+
+// CharPtrPtr is an unsafe pointer wrapping C's `char**`.
+type CharPtrPtr unsafe.Pointer
+
+// CharPtr is an unsafe pointer wrapping C's `char*`.
+type CharPtr unsafe.Pointer
+
+// SizeTPtr is an unsafe pointer wrapping C's `size_t*`.
+type SizeTPtr unsafe.Pointer
+
+// FreeFunc is a wrapper around calls to, or act like, C's free function.
+type FreeFunc func(unsafe.Pointer)
+
+// Malloc is C.malloc
+func Malloc(s SizeT) CPtr { return CPtr(C.malloc(C.size_t(s))) }
+
+// Free is C.free
+func Free(p CPtr) { C.free(unsafe.Pointer(p)) }
+
+// CString is C.CString
+func CString(s string) CharPtr { return CharPtr((C.CString(s))) }
+
+// CBytes is C.CBytes
+func CBytes(b []byte) CPtr { return CPtr(C.CBytes(b)) }
+
+// Memcpy is C.memcpy
+func Memcpy(dst, src CPtr, n SizeT) {
+	C.memcpy(unsafe.Pointer(dst), unsafe.Pointer(src), C.size_t(n))
+}

vendor/github.com/ceph/go-ceph/internal/cutil/buffergroup.go

@@ -0,0 +1,89 @@
+package cutil
+
+// #include <stdlib.h>
+import "C"
+
+import (
+	"unsafe"
+)
+
+// BufferGroup is a helper structure that holds Go-allocated slices of
+// C-allocated strings and their respective lengths. Useful for C functions
+// that consume byte buffers with explicit length instead of null-terminated
+// strings. When used as input arguments in C functions, caller must make sure
+// the C code will not hold any pointers to either of the struct's attributes
+// after that C function returns.
+type BufferGroup struct {
+	// C-allocated buffers.
+	Buffers []CharPtr
+	// Lengths of C buffers, where Lengths[i] = length(Buffers[i]).
+	Lengths []SizeT
+}
+
+// TODO: should BufferGroup implementation change and the slices would contain
+//       nested Go pointers, they must be pinned with PtrGuard.
+
+// NewBufferGroupStrings returns new BufferGroup constructed from strings.
+func NewBufferGroupStrings(strs []string) *BufferGroup {
+	s := &BufferGroup{
+		Buffers: make([]CharPtr, len(strs)),
+		Lengths: make([]SizeT, len(strs)),
+	}
+
+	for i, str := range strs {
+		bs := []byte(str)
+		s.Buffers[i] = CharPtr(C.CBytes(bs))
+		s.Lengths[i] = SizeT(len(bs))
+	}
+
+	return s
+}
+
+// NewBufferGroupBytes returns new BufferGroup constructed
+// from slice of byte slices.
+func NewBufferGroupBytes(bss [][]byte) *BufferGroup {
+	s := &BufferGroup{
+		Buffers: make([]CharPtr, len(bss)),
+		Lengths: make([]SizeT, len(bss)),
+	}
+
+	for i, bs := range bss {
+		s.Buffers[i] = CharPtr(C.CBytes(bs))
+		s.Lengths[i] = SizeT(len(bs))
+	}
+
+	return s
+}
+
+// Free free()s the C-allocated memory.
+func (s *BufferGroup) Free() {
+	for _, ptr := range s.Buffers {
+		C.free(unsafe.Pointer(ptr))
+	}
+
+	s.Buffers = nil
+	s.Lengths = nil
+}
+
+// BuffersPtr returns a pointer to the beginning of the Buffers slice.
+func (s *BufferGroup) BuffersPtr() CharPtrPtr {
+	if len(s.Buffers) == 0 {
+		return nil
+	}
+
+	return CharPtrPtr(&s.Buffers[0])
+}
+
+// LengthsPtr returns a pointer to the beginning of the Lengths slice.
+func (s *BufferGroup) LengthsPtr() SizeTPtr {
+	if len(s.Lengths) == 0 {
+		return nil
+	}
+
+	return SizeTPtr(&s.Lengths[0])
+}
+
+func testBufferGroupGet(s *BufferGroup, index int) (str string, length int) {
+	bs := C.GoBytes(unsafe.Pointer(s.Buffers[index]), C.int(s.Lengths[index]))
+	return string(bs), int(s.Lengths[index])
+}

vendor/github.com/ceph/go-ceph/internal/cutil/command_input.go

@@ -0,0 +1,62 @@
+package cutil
+
+/*
+#include <stdlib.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+// CommandInput can be used to manage the input to ceph's *_command functions.
+type CommandInput struct {
+	cmd   []*C.char
+	inbuf []byte
+}
+
+// NewCommandInput creates C-level pointers from go byte buffers suitable
+// for passing off to ceph's *_command functions.
+func NewCommandInput(cmd [][]byte, inputBuffer []byte) *CommandInput {
+	ci := &CommandInput{
+		cmd:   make([]*C.char, len(cmd)),
+		inbuf: inputBuffer,
+	}
+	for i := range cmd {
+		ci.cmd[i] = C.CString(string(cmd[i]))
+	}
+	return ci
+}
+
+// Free any C memory managed by this CommandInput.
+func (ci *CommandInput) Free() {
+	for i := range ci.cmd {
+		C.free(unsafe.Pointer(ci.cmd[i]))
+	}
+	ci.cmd = nil
+}
+
+// Cmd returns an unsafe wrapper around an array of C-strings.
+func (ci *CommandInput) Cmd() CharPtrPtr {
+	ptr := &ci.cmd[0]
+	return CharPtrPtr(ptr)
+}
+
+// CmdLen returns the length of the array of C-strings returned by
+// Cmd.
+func (ci *CommandInput) CmdLen() SizeT {
+	return SizeT(len(ci.cmd))
+}
+
+// InBuf returns an unsafe wrapper to a C char*.
+func (ci *CommandInput) InBuf() CharPtr {
+	if len(ci.inbuf) == 0 {
+		return nil
+	}
+	return CharPtr(&ci.inbuf[0])
+}
+
+// InBufLen returns the length of the buffer returned by InBuf.
+func (ci *CommandInput) InBufLen() SizeT {
+	return SizeT(len(ci.inbuf))
+}

vendor/github.com/ceph/go-ceph/internal/cutil/command_output.go

@@ -0,0 +1,100 @@
+package cutil
+
+/*
+#include <stdlib.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+// CommandOutput can be used to manage the outputs of ceph's *_command
+// functions.
+type CommandOutput struct {
+	free      FreeFunc
+	outBuf    *C.char
+	outBufLen C.size_t
+	outs      *C.char
+	outsLen   C.size_t
+}
+
+// NewCommandOutput returns an empty CommandOutput. The pointers that
+// a CommandOutput provides can be used to get the results of ceph's
+// *_command functions.
+func NewCommandOutput() *CommandOutput {
+	return &CommandOutput{
+		free: free,
+	}
+}
+
+// SetFreeFunc sets the function used to free memory held by CommandOutput.
+// Not all uses of CommandOutput expect to use the basic C.free function
+// and either require or prefer the use of a custom deallocation function.
+// Use SetFreeFunc to change the free function and return the modified
+// CommandOutput object.
+func (co *CommandOutput) SetFreeFunc(f FreeFunc) *CommandOutput {
+	co.free = f
+	return co
+}
+
+// Free any C memory tracked by this object.
+func (co *CommandOutput) Free() {
+	if co.outBuf != nil {
+		co.free(unsafe.Pointer(co.outBuf))
+	}
+	if co.outs != nil {
+		co.free(unsafe.Pointer(co.outs))
+	}
+}
+
+// OutBuf returns an unsafe wrapper around a pointer to a `char*`.
+func (co *CommandOutput) OutBuf() CharPtrPtr {
+	return CharPtrPtr(&co.outBuf)
+}
+
+// OutBufLen returns an unsafe wrapper around a pointer to a size_t.
+func (co *CommandOutput) OutBufLen() SizeTPtr {
+	return SizeTPtr(&co.outBufLen)
+}
+
+// Outs returns an unsafe wrapper around a pointer to a `char*`.
+func (co *CommandOutput) Outs() CharPtrPtr {
+	return CharPtrPtr(&co.outs)
+}
+
+// OutsLen returns an unsafe wrapper around a pointer to a size_t.
+func (co *CommandOutput) OutsLen() SizeTPtr {
+	return SizeTPtr(&co.outsLen)
+}
+
+// GoValues returns native go values converted from the internal C-language
+// values tracked by this object.
+func (co *CommandOutput) GoValues() (buf []byte, status string) {
+	if co.outBufLen > 0 {
+		buf = C.GoBytes(unsafe.Pointer(co.outBuf), C.int(co.outBufLen))
+	}
+	if co.outsLen > 0 {
+		status = C.GoStringN(co.outs, C.int(co.outsLen))
+	}
+	return buf, status
+}
+
+// testSetString is only used by the unit tests for this file.
+// It is located here due to the restriction on having import "C" in
+// go test files. :-(
+// It mimics a C function that takes a pointer to a
+// string and length and allocates memory and sets the pointers
+// to the new string and its length.
+func testSetString(strp CharPtrPtr, lenp SizeTPtr, s string) {
+	sp := (**C.char)(strp)
+	lp := (*C.size_t)(lenp)
+	*sp = C.CString(s)
+	*lp = C.size_t(len(s))
+}
+
+// free wraps C.free.
+// Required for unit tests that may not use cgo directly.
+func free(p unsafe.Pointer) {
+	C.free(p)
+}

vendor/github.com/ceph/go-ceph/internal/cutil/cslice.go

@@ -0,0 +1,76 @@
+package cutil
+
+// The following code needs some explanation:
+// This creates slices on top of the C memory buffers allocated before in
+// order to safely and comfortably use them as arrays. First the void pointer
+// is cast to a pointer to an array of the type that will be stored in the
+// array. Because the size of an array is a constant, but the real array size
+// is dynamic, we just use the biggest possible index value MaxIdx, to make
+// sure it's always big enough. (Nothing is allocated by casting, so the size
+// can be arbitrarily big.) So, if the array should store items of myType, the
+// cast would be (*[MaxIdx]myItem)(myCMemPtr).
+// From that array pointer a slice is created with the [start:end:capacity]
+// syntax. The capacity must be set explicitly here, because by default it
+// would be set to the size of the original array, which is MaxIdx, which
+// doesn't reflect reality in this case. This results in definitions like:
+// cSlice := (*[MaxIdx]myItem)(myCMemPtr)[:numOfItems:numOfItems]
+
+////////// CPtr //////////
+
+// CPtrCSlice is a C allocated slice of C pointers.
+type CPtrCSlice []CPtr
+
+// NewCPtrCSlice returns a CPtrSlice.
+// Similar to CString it must be freed with slice.Free()
+func NewCPtrCSlice(size int) CPtrCSlice {
+	if size == 0 {
+		return nil
+	}
+	cMem := Malloc(SizeT(size) * PtrSize)
+	cSlice := (*[MaxIdx]CPtr)(cMem)[:size:size]
+	return cSlice
+}
+
+// Ptr returns a pointer to CPtrSlice
+func (v *CPtrCSlice) Ptr() CPtr {
+	if len(*v) == 0 {
+		return nil
+	}
+	return CPtr(&(*v)[0])
+}
+
+// Free frees a CPtrSlice
+func (v *CPtrCSlice) Free() {
+	Free(v.Ptr())
+	*v = nil
+}
+
+////////// SizeT //////////
+
+// SizeTCSlice is a C allocated slice of C.size_t.
+type SizeTCSlice []SizeT
+
+// NewSizeTCSlice returns a SizeTCSlice.
+// Similar to CString it must be freed with slice.Free()
+func NewSizeTCSlice(size int) SizeTCSlice {
+	if size == 0 {
+		return nil
+	}
+	cMem := Malloc(SizeT(size) * SizeTSize)
+	cSlice := (*[MaxIdx]SizeT)(cMem)[:size:size]
+	return cSlice
+}
+
+// Ptr returns a pointer to SizeTCSlice
+func (v *SizeTCSlice) Ptr() CPtr {
+	if len(*v) == 0 {
+		return nil
+	}
+	return CPtr(&(*v)[0])
+}
+
+// Free frees a SizeTCSlice
+func (v *SizeTCSlice) Free() {
+	Free(v.Ptr())
+	*v = nil
+}

vendor/github.com/ceph/go-ceph/internal/cutil/iovec.go

@@ -0,0 +1,60 @@
+package cutil
+
+/*
+#include <stdlib.h>
+#include <sys/uio.h>
+*/
+import "C"
+import (
+	"unsafe"
+)
+
+// Iovec is a slice of iovec structs. Might have allocated C memory, so it must
+// be freed with the Free() method.
+type Iovec struct {
+	iovec []C.struct_iovec
+	sbs   []*SyncBuffer
+}
+
+const iovecSize = C.sizeof_struct_iovec
+
+// ByteSlicesToIovec creates an Iovec and links it to Go buffers in data.
+func ByteSlicesToIovec(data [][]byte) (v Iovec) {
+	n := len(data)
+	iovecMem := C.malloc(iovecSize * C.size_t(n))
+	v.iovec = (*[MaxIdx]C.struct_iovec)(iovecMem)[:n:n]
+	for i, b := range data {
+		sb := NewSyncBuffer(CPtr(&v.iovec[i].iov_base), b)
+		v.sbs = append(v.sbs, sb)
+		v.iovec[i].iov_len = C.size_t(len(b))
+	}
+	return
+}
+
+// Sync makes sure the slices contain the same as the C buffers
+func (v *Iovec) Sync() {
+	for _, sb := range v.sbs {
+		sb.Sync()
+	}
+}
+
+// Pointer returns a pointer to the iovec
+func (v *Iovec) Pointer() unsafe.Pointer {
+	return unsafe.Pointer(&v.iovec[0])
+}
+
+// Len returns a pointer to the iovec
+func (v *Iovec) Len() int {
+	return len(v.iovec)
+}
+
+// Free the C memory in the Iovec.
+func (v *Iovec) Free() {
+	for _, sb := range v.sbs {
+		sb.Release()
+	}
+	if len(v.iovec) != 0 {
+		C.free(unsafe.Pointer(&v.iovec[0]))
+	}
+	v.iovec = nil
+}

vendor/github.com/ceph/go-ceph/internal/cutil/ptrguard.go

@@ -0,0 +1,90 @@
+package cutil
+
+import (
+	"sync"
+	"unsafe"
+)
+
+// PtrGuard respresents a guarded Go pointer (pointing to memory allocated by Go
+// runtime) stored in C memory (allocated by C)
+type PtrGuard struct {
+	// These mutexes will be used as binary semaphores for signalling events from
+	// one thread to another, which - in contrast to other languages like C++ - is
+	// possible in Go, that is a Mutex can be locked in one thread and unlocked in
+	// another.
+	stored, release sync.Mutex
+	released        bool
+}
+
+// WARNING: using binary semaphores (mutexes) for signalling like this is quite
+// a delicate task in order to avoid deadlocks or panics. Whenever changing the
+// code logic, please review at least three times that there is no unexpected
+// state possible. Usually the natural choice would be to use channels instead,
+// but these can not easily passed to C code because of the pointer-to-pointer
+// cgo rule, and would require the use of a Go object registry.
+
+// NewPtrGuard writes the goPtr (pointing to Go memory) into C memory at the
+// position cPtr, and returns a PtrGuard object.
+func NewPtrGuard(cPtr CPtr, goPtr unsafe.Pointer) *PtrGuard {
+	var v PtrGuard
+	// Since the mutexes are used for signalling, they have to be initialized to
+	// locked state, so that following lock attempts will block.
+	v.release.Lock()
+	v.stored.Lock()
+	// Start a background go routine that lives until Release is called. This
+	// calls a special function that makes sure the garbage collector doesn't touch
+	// goPtr, stores it into C memory at position cPtr and then waits until it
+	// reveices the "release" signal, after which it nulls out the C memory at
+	// cPtr and then exits.
+	go func() {
+		storeUntilRelease(&v, (*CPtr)(cPtr), uintptr(goPtr))
+	}()
+	// Wait for the "stored" signal from the go routine when the Go pointer has
+	// been stored to the C memory. <--(1)
+	v.stored.Lock()
+	return &v
+}
+
+// Release removes the guarded Go pointer from the C memory by overwriting it
+// with NULL.
+func (v *PtrGuard) Release() {
+	if !v.released {
+		v.released = true
+		v.release.Unlock() // Send the "release" signal to the go routine. -->(2)
+		v.stored.Lock()    // Wait for the second "stored" signal when the C memory
+		//                    has been nulled out. <--(3)
+
+	}
+}
+
+// The uintptrPtr() helper function below assumes that uintptr has the same size
+// as a pointer, although in theory it could be larger.  Therefore we use this
+// constant expression to assert size equality as a safeguard at compile time.
+// How it works: if sizes are different, either the inner or outer expression is
+// negative, which always fails with "constant ... overflows uintptr", because
+// unsafe.Sizeof() is a uintptr typed constant.
+const _ = -(unsafe.Sizeof(uintptr(0)) - PtrSize) // size assert
+func uintptrPtr(p *CPtr) *uintptr {
+	return (*uintptr)(unsafe.Pointer(p))
+}
+
+//go:uintptrescapes
+
+// From https://golang.org/src/cmd/compile/internal/gc/lex.go:
+// For the next function declared in the file any uintptr arguments may be
+// pointer values converted to uintptr. This directive ensures that the
+// referenced allocated object, if any, is retained and not moved until the call
+// completes, even though from the types alone it would appear that the object
+// is no longer needed during the call. The conversion to uintptr must appear in
+// the argument list.
+// Also see https://golang.org/cmd/compile/#hdr-Compiler_Directives
+
+func storeUntilRelease(v *PtrGuard, cPtr *CPtr, goPtr uintptr) {
+	uip := uintptrPtr(cPtr)
+	*uip = goPtr      // store Go pointer in C memory at c_ptr
+	v.stored.Unlock() // send "stored" signal to main thread -->(1)
+	v.release.Lock()  // wait for "release" signal from main thread when
+	//                   Release() has been called. <--(2)
+	*uip = 0          // reset C memory to NULL
+	v.stored.Unlock() // send second "stored" signal to main thread -->(3)
+}

vendor/github.com/ceph/go-ceph/internal/cutil/splitbuf.go

@@ -0,0 +1,49 @@
+package cutil
+
+import "C"
+
+import (
+	"bytes"
+)
+
+// SplitBuffer splits a byte-slice buffer, typically returned from C code,
+// into a slice of strings.
+// The contents of the buffer are assumed to be null-byte separated.
+// If the buffer contains a sequence of null-bytes it will assume that the
+// "space" between the bytes are meant to be empty strings.
+func SplitBuffer(b []byte) []string {
+	return splitBufStrings(b, true)
+}
+
+// SplitSparseBuffer splits a byte-slice buffer, typically returned from C code,
+// into a slice of strings.
+// The contents of the buffer are assumed to be null-byte separated.
+// This function assumes that buffer to be "sparse" such that only non-null-byte
+// strings will be returned, and no "empty" strings exist if null-bytes
+// are found adjacent to each other.
+func SplitSparseBuffer(b []byte) []string {
+	return splitBufStrings(b, false)
+}
+
+// If keepEmpty is true, empty substrings will be returned, by default they are
+// excluded from the results.
+// This is almost certainly a suboptimal implementation, especially for
+// keepEmpty=true case. Optimizing the functions is a job for another day.
+func splitBufStrings(b []byte, keepEmpty bool) []string {
+	values := make([]string, 0)
+	// the final null byte should be the terminating null in C
+	// we never want to preserve the empty string after it
+	if len(b) > 0 && b[len(b)-1] == 0 {
+		b = b[:len(b)-1]
+	}
+	if len(b) == 0 {
+		return values
+	}
+	for _, s := range bytes.Split(b, []byte{0}) {
+		if !keepEmpty && len(s) == 0 {
+			continue
+		}
+		values = append(values, string(s))
+	}
+	return values
+}

vendor/github.com/ceph/go-ceph/internal/cutil/sync_buffer.go

@@ -0,0 +1,29 @@
+//go:build !no_ptrguard
+// +build !no_ptrguard
+
+package cutil
+
+import (
+	"unsafe"
+)
+
+// SyncBuffer is a C buffer connected to a data slice
+type SyncBuffer struct {
+	pg *PtrGuard
+}
+
+// NewSyncBuffer creates a C buffer from a data slice and stores it at CPtr
+func NewSyncBuffer(cPtr CPtr, data []byte) *SyncBuffer {
+	var v SyncBuffer
+	v.pg = NewPtrGuard(cPtr, unsafe.Pointer(&data[0]))
+	return &v
+}
+
+// Release releases the C buffer and nulls its stored pointer
+func (v *SyncBuffer) Release() {
+	v.pg.Release()
+}
+
+// Sync asserts that changes in the C buffer are available in the data
+// slice
+func (*SyncBuffer) Sync() {}

vendor/github.com/ceph/go-ceph/internal/cutil/sync_buffer_memcpy.go

@@ -0,0 +1,38 @@
+//go:build no_ptrguard
+// +build no_ptrguard
+
+package cutil
+
+// SyncBuffer is a C buffer connected to a data slice
+type SyncBuffer struct {
+	data []byte
+	cPtr *CPtr
+}
+
+// NewSyncBuffer creates a C buffer from a data slice and stores it at CPtr
+func NewSyncBuffer(cPtr CPtr, data []byte) *SyncBuffer {
+	var v SyncBuffer
+	v.data = data
+	v.cPtr = (*CPtr)(cPtr)
+	*v.cPtr = CBytes(data)
+	return &v
+}
+
+// Release releases the C buffer and nulls its stored pointer
+func (v *SyncBuffer) Release() {
+	if v.cPtr != nil {
+		Free(*v.cPtr)
+		*v.cPtr = nil
+		v.cPtr = nil
+	}
+	v.data = nil
+}
+
+// Sync asserts that changes in the C buffer are available in the data
+// slice
+func (v *SyncBuffer) Sync() {
+	if v.cPtr == nil {
+		return
+	}
+	Memcpy(CPtr(&v.data[0]), CPtr(*v.cPtr), SizeT(len(v.data)))
+}

vendor/github.com/ceph/go-ceph/internal/errutil/strerror.go

@@ -0,0 +1,52 @@
+/*
+Package errutil provides common functions for dealing with error conditions for
+all ceph api wrappers.
+*/
+package errutil
+
+/* force XSI-complaint strerror_r() */
+
+// #define _POSIX_C_SOURCE 200112L
+// #undef _GNU_SOURCE
+// #include <stdlib.h>
+// #include <errno.h>
+// #include <string.h>
+import "C"
+
+import (
+	"fmt"
+	"unsafe"
+)
+
+// FormatErrno returns the absolute value of the errno as well as a string
+// describing the errno. The string will be empty is the errno is not known.
+func FormatErrno(errno int) (int, string) {
+	buf := make([]byte, 1024)
+	// strerror expects errno >= 0
+	if errno < 0 {
+		errno = -errno
+	}
+
+	ret := C.strerror_r(
+		C.int(errno),
+		(*C.char)(unsafe.Pointer(&buf[0])),
+		C.size_t(len(buf)))
+	if ret != 0 {
+		return errno, ""
+	}
+
+	return errno, C.GoString((*C.char)(unsafe.Pointer(&buf[0])))
+}
+
+// FormatErrorCode returns a string that describes the supplied error source
+// and error code as a string. Suitable to use in Error() methods.  If the
+// error code maps to an errno the string will contain a description of the
+// error. Otherwise the string will only indicate the source and value if the
+// value does not map to a known errno.
+func FormatErrorCode(source string, errValue int) string {
+	_, s := FormatErrno(errValue)
+	if s == "" {
+		return fmt.Sprintf("%s: ret=%d", source, errValue)
+	}
+	return fmt.Sprintf("%s: ret=%d, %s", source, errValue, s)
+}

vendor/github.com/ceph/go-ceph/internal/log/log.go

@@ -0,0 +1,14 @@
+// Package log is the internal package for go-ceph logging. This package is only
+// used from go-ceph code, not from consumers of go-ceph. go-ceph code uses the
+// functions in this package to log information that can't be returned as
+// errors. The functions default to no-ops and can be set with the external log
+// package common/log by the go-ceph consumers.
+package log
+
+func noop(string, ...interface{}) {}
+
+// These variables are set by the common log package.
+var (
+	Warnf  = noop
+	Debugf = noop
+)

vendor/github.com/ceph/go-ceph/internal/retry/sizer.go

@@ -0,0 +1,64 @@
+package retry
+
+// Hint is a type for retry hints
+type Hint interface {
+	If(bool) Hint
+	size() int
+}
+
+type hintInt int
+
+func (hint hintInt) size() int {
+	return int(hint)
+}
+
+// If is a convenience function, that returns a given hint only if a certain
+// condition is met (for example a test for a "buffer too small" error).
+// Otherwise it returns a nil which stops the retries.
+func (hint hintInt) If(cond bool) Hint {
+	if cond {
+		return hint
+	}
+	return nil
+}
+
+// DoubleSize is a hint to retry with double the size
+const DoubleSize = hintInt(0)
+
+// Size returns a hint for a specific size
+func Size(s int) Hint {
+	return hintInt(s)
+}
+
+// SizeFunc is used to implement 'resize loops' that hides the complexity of the
+// sizing away from most of the application. It's a function that takes a size
+// argument and returns nil, if no retry is necessary, or a hint indicating the
+// size for the next retry. If errors or other results are required from the
+// function, the function can write them to function closures of the surrounding
+// scope. See tests for examples.
+type SizeFunc func(size int) (hint Hint)
+
+// WithSizes repeatingly calls a SizeFunc with increasing sizes until either it
+// returns nil, or the max size has been reached. If the returned hint is
+// DoubleSize or indicating a size not greater than the current size, the size
+// is doubled. If the hint or next size is greater than the max size, the max
+// size is used for a last retry.
+func WithSizes(size int, max int, f SizeFunc) {
+	if size > max {
+		return
+	}
+	for {
+		hint := f(size)
+		if hint == nil || size == max {
+			break
+		}
+		if hint.size() > size {
+			size = hint.size()
+		} else {
+			size *= 2
+		}
+		if size > max {
+			size = max
+		}
+	}
+}

vendor/github.com/ceph/go-ceph/internal/timespec/timespec.go

@@ -0,0 +1,39 @@
+package timespec
+
+/*
+#include <time.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+
+	"golang.org/x/sys/unix"
+)
+
+// Timespec behaves similarly to C's struct timespec.
+// Timespec is used to retain fidelity to the C based file systems
+// apis that could be lossy with the use of Go time types.
+type Timespec unix.Timespec
+
+// CTimespecPtr is an unsafe pointer wrapping C's `struct timespec`.
+type CTimespecPtr unsafe.Pointer
+
+// CStructToTimespec creates a new Timespec for the C 'struct timespec'.
+func CStructToTimespec(cts CTimespecPtr) Timespec {
+	t := (*C.struct_timespec)(cts)
+
+	return Timespec{
+		Sec:  int64(t.tv_sec),
+		Nsec: int64(t.tv_nsec),
+	}
+}
+
+// CopyToCStruct copies the time values from a Timespec to a previously
+// allocated C `struct timespec`. Due to restrictions on Cgo the C pointer
+// must be passed via the CTimespecPtr wrapper.
+func CopyToCStruct(ts Timespec, cts CTimespecPtr) {
+	t := (*C.struct_timespec)(cts)
+	t.tv_sec = C.time_t(ts.Sec)
+	t.tv_nsec = C.long(ts.Nsec)
+}

vendor/github.com/ceph/go-ceph/rados/alloc_hint_flags.go

@@ -0,0 +1,37 @@
+//go:build ceph_preview
+// +build ceph_preview
+
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+//
+import "C"
+
+// AllocHintFlags control the behavior of read and write operations.
+type AllocHintFlags uint32
+
+const (
+	// AllocHintNoHint indicates no predefined behavior
+	AllocHintNoHint = AllocHintFlags(0)
+	// AllocHintSequentialWrite TODO
+	AllocHintSequentialWrite = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_SEQUENTIAL_WRITE)
+	// AllocHintRandomWrite TODO
+	AllocHintRandomWrite = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_RANDOM_WRITE)
+	// AllocHintSequentialRead TODO
+	AllocHintSequentialRead = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_SEQUENTIAL_READ)
+	// AllocHintRandomRead TODO
+	AllocHintRandomRead = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_RANDOM_READ)
+	// AllocHintAppendOnly TODO
+	AllocHintAppendOnly = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_APPEND_ONLY)
+	// AllocHintImmutable TODO
+	AllocHintImmutable = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_IMMUTABLE)
+	// AllocHintShortlived TODO
+	AllocHintShortlived = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_SHORTLIVED)
+	// AllocHintLonglived TODO
+	AllocHintLonglived = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_LONGLIVED)
+	// AllocHintCompressible TODO
+	AllocHintCompressible = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_COMPRESSIBLE)
+	// AllocHintIncompressible TODO
+	AllocHintIncompressible = AllocHintFlags(C.LIBRADOS_ALLOC_HINT_FLAG_INCOMPRESSIBLE)
+)

vendor/github.com/ceph/go-ceph/rados/command.go

@@ -0,0 +1,193 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <stdlib.h>
+// #include <rados/librados.h>
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+)
+
+func radosBufferFree(p unsafe.Pointer) {
+	C.rados_buffer_free((*C.char)(p))
+}
+
+// MonCommand sends a command to one of the monitors
+func (c *Conn) MonCommand(args []byte) ([]byte, string, error) {
+	return c.MonCommandWithInputBuffer(args, nil)
+}
+
+// MonCommandWithInputBuffer sends a command to one of the monitors, with an input buffer
+func (c *Conn) MonCommandWithInputBuffer(args, inputBuffer []byte) ([]byte, string, error) {
+	ci := cutil.NewCommandInput([][]byte{args}, inputBuffer)
+	defer ci.Free()
+	co := cutil.NewCommandOutput().SetFreeFunc(radosBufferFree)
+	defer co.Free()
+
+	ret := C.rados_mon_command(
+		c.cluster,
+		(**C.char)(ci.Cmd()),
+		C.size_t(ci.CmdLen()),
+		(*C.char)(ci.InBuf()),
+		C.size_t(ci.InBufLen()),
+		(**C.char)(co.OutBuf()),
+		(*C.size_t)(co.OutBufLen()),
+		(**C.char)(co.Outs()),
+		(*C.size_t)(co.OutsLen()))
+	buf, status := co.GoValues()
+	return buf, status, getError(ret)
+}
+
+// PGCommand sends a command to one of the PGs
+//
+// Implements:
+//  int rados_pg_command(rados_t cluster, const char *pgstr,
+//                       const char **cmd, size_t cmdlen,
+//                       const char *inbuf, size_t inbuflen,
+//                       char **outbuf, size_t *outbuflen,
+//                       char **outs, size_t *outslen);
+func (c *Conn) PGCommand(pgid []byte, args [][]byte) ([]byte, string, error) {
+	return c.PGCommandWithInputBuffer(pgid, args, nil)
+}
+
+// PGCommandWithInputBuffer sends a command to one of the PGs, with an input buffer
+//
+// Implements:
+//  int rados_pg_command(rados_t cluster, const char *pgstr,
+//                       const char **cmd, size_t cmdlen,
+//                       const char *inbuf, size_t inbuflen,
+//                       char **outbuf, size_t *outbuflen,
+//                       char **outs, size_t *outslen);
+func (c *Conn) PGCommandWithInputBuffer(pgid []byte, args [][]byte, inputBuffer []byte) ([]byte, string, error) {
+	name := C.CString(string(pgid))
+	defer C.free(unsafe.Pointer(name))
+	ci := cutil.NewCommandInput(args, inputBuffer)
+	defer ci.Free()
+	co := cutil.NewCommandOutput().SetFreeFunc(radosBufferFree)
+	defer co.Free()
+
+	ret := C.rados_pg_command(
+		c.cluster,
+		name,
+		(**C.char)(ci.Cmd()),
+		C.size_t(ci.CmdLen()),
+		(*C.char)(ci.InBuf()),
+		C.size_t(ci.InBufLen()),
+		(**C.char)(co.OutBuf()),
+		(*C.size_t)(co.OutBufLen()),
+		(**C.char)(co.Outs()),
+		(*C.size_t)(co.OutsLen()))
+	buf, status := co.GoValues()
+	return buf, status, getError(ret)
+}
+
+// MgrCommand sends a command to a ceph-mgr.
+func (c *Conn) MgrCommand(args [][]byte) ([]byte, string, error) {
+	return c.MgrCommandWithInputBuffer(args, nil)
+}
+
+// MgrCommandWithInputBuffer sends a command, with an input buffer, to a ceph-mgr.
+//
+// Implements:
+//  int rados_mgr_command(rados_t cluster, const char **cmd,
+//                         size_t cmdlen, const char *inbuf,
+//                         size_t inbuflen, char **outbuf,
+//                         size_t *outbuflen, char **outs,
+//                          size_t *outslen);
+func (c *Conn) MgrCommandWithInputBuffer(args [][]byte, inputBuffer []byte) ([]byte, string, error) {
+	ci := cutil.NewCommandInput(args, inputBuffer)
+	defer ci.Free()
+	co := cutil.NewCommandOutput().SetFreeFunc(radosBufferFree)
+	defer co.Free()
+
+	ret := C.rados_mgr_command(
+		c.cluster,
+		(**C.char)(ci.Cmd()),
+		C.size_t(ci.CmdLen()),
+		(*C.char)(ci.InBuf()),
+		C.size_t(ci.InBufLen()),
+		(**C.char)(co.OutBuf()),
+		(*C.size_t)(co.OutBufLen()),
+		(**C.char)(co.Outs()),
+		(*C.size_t)(co.OutsLen()))
+	buf, status := co.GoValues()
+	return buf, status, getError(ret)
+}
+
+// OsdCommand sends a command to the specified ceph OSD.
+func (c *Conn) OsdCommand(osd int, args [][]byte) ([]byte, string, error) {
+	return c.OsdCommandWithInputBuffer(osd, args, nil)
+}
+
+// OsdCommandWithInputBuffer sends a command, with an input buffer, to the
+// specified ceph OSD.
+//
+// Implements:
+//  int rados_osd_command(rados_t cluster, int osdid,
+//                                       const char **cmd, size_t cmdlen,
+//                                       const char *inbuf, size_t inbuflen,
+//                                       char **outbuf, size_t *outbuflen,
+//                                       char **outs, size_t *outslen);
+func (c *Conn) OsdCommandWithInputBuffer(
+	osd int, args [][]byte, inputBuffer []byte) ([]byte, string, error) {
+
+	ci := cutil.NewCommandInput(args, inputBuffer)
+	defer ci.Free()
+	co := cutil.NewCommandOutput().SetFreeFunc(radosBufferFree)
+	defer co.Free()
+
+	ret := C.rados_osd_command(
+		c.cluster,
+		C.int(osd),
+		(**C.char)(ci.Cmd()),
+		C.size_t(ci.CmdLen()),
+		(*C.char)(ci.InBuf()),
+		C.size_t(ci.InBufLen()),
+		(**C.char)(co.OutBuf()),
+		(*C.size_t)(co.OutBufLen()),
+		(**C.char)(co.Outs()),
+		(*C.size_t)(co.OutsLen()))
+	buf, status := co.GoValues()
+	return buf, status, getError(ret)
+}
+
+// MonCommandTarget sends a command to a specified monitor.
+func (c *Conn) MonCommandTarget(name string, args [][]byte) ([]byte, string, error) {
+	return c.MonCommandTargetWithInputBuffer(name, args, nil)
+}
+
+// MonCommandTargetWithInputBuffer sends a command, with an input buffer, to a specified monitor.
+//
+// Implements:
+//  int rados_mon_command_target(rados_t cluster, const char *name,
+//                               const char **cmd, size_t cmdlen,
+//                               const char *inbuf, size_t inbuflen,
+//                               char **outbuf, size_t *outbuflen,
+//                               char **outs, size_t *outslen);
+func (c *Conn) MonCommandTargetWithInputBuffer(
+	name string, args [][]byte, inputBuffer []byte) ([]byte, string, error) {
+
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+	ci := cutil.NewCommandInput(args, inputBuffer)
+	defer ci.Free()
+	co := cutil.NewCommandOutput().SetFreeFunc(radosBufferFree)
+	defer co.Free()
+
+	ret := C.rados_mon_command_target(
+		c.cluster,
+		cName,
+		(**C.char)(ci.Cmd()),
+		C.size_t(ci.CmdLen()),
+		(*C.char)(ci.InBuf()),
+		C.size_t(ci.InBufLen()),
+		(**C.char)(co.OutBuf()),
+		(*C.size_t)(co.OutBufLen()),
+		(**C.char)(co.Outs()),
+		(*C.size_t)(co.OutsLen()))
+	buf, status := co.GoValues()
+	return buf, status, getError(ret)
+}

vendor/github.com/ceph/go-ceph/rados/conn.go

@@ -0,0 +1,311 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <stdlib.h>
+// #include <rados/librados.h>
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+	"github.com/ceph/go-ceph/internal/retry"
+)
+
+var argvPlaceholder = "placeholder"
+
+//revive:disable:var-naming old-yet-exported public api
+
+// ClusterStat represents Ceph cluster statistics.
+type ClusterStat struct {
+	Kb          uint64
+	Kb_used     uint64
+	Kb_avail    uint64
+	Num_objects uint64
+}
+
+//revive:enable:var-naming
+
+// Conn is a connection handle to a Ceph cluster.
+type Conn struct {
+	cluster   C.rados_t
+	connected bool
+}
+
+// ClusterRef represents a fundamental RADOS cluster connection.
+type ClusterRef C.rados_t
+
+// Cluster returns the underlying RADOS cluster reference for this Conn.
+func (c *Conn) Cluster() ClusterRef {
+	return ClusterRef(c.cluster)
+}
+
+// PingMonitor sends a ping to a monitor and returns the reply.
+func (c *Conn) PingMonitor(id string) (string, error) {
+	cid := C.CString(id)
+	defer C.free(unsafe.Pointer(cid))
+
+	var strlen C.size_t
+	var strout *C.char
+
+	ret := C.rados_ping_monitor(c.cluster, cid, &strout, &strlen)
+	defer C.rados_buffer_free(strout)
+
+	if ret == 0 {
+		reply := C.GoStringN(strout, (C.int)(strlen))
+		return reply, nil
+	}
+	return "", getError(ret)
+}
+
+// Connect establishes a connection to a RADOS cluster. It returns an error,
+// if any.
+func (c *Conn) Connect() error {
+	ret := C.rados_connect(c.cluster)
+	if ret != 0 {
+		return getError(ret)
+	}
+	c.connected = true
+	return nil
+}
+
+// Shutdown disconnects from the cluster.
+func (c *Conn) Shutdown() {
+	if err := c.ensureConnected(); err != nil {
+		return
+	}
+	freeConn(c)
+}
+
+// ReadConfigFile configures the connection using a Ceph configuration file.
+func (c *Conn) ReadConfigFile(path string) error {
+	cPath := C.CString(path)
+	defer C.free(unsafe.Pointer(cPath))
+	ret := C.rados_conf_read_file(c.cluster, cPath)
+	return getError(ret)
+}
+
+// ReadDefaultConfigFile configures the connection using a Ceph configuration
+// file located at default locations.
+func (c *Conn) ReadDefaultConfigFile() error {
+	ret := C.rados_conf_read_file(c.cluster, nil)
+	return getError(ret)
+}
+
+// OpenIOContext creates and returns a new IOContext for the given pool.
+//
+// Implements:
+//  int rados_ioctx_create(rados_t cluster, const char *pool_name,
+//                         rados_ioctx_t *ioctx);
+func (c *Conn) OpenIOContext(pool string) (*IOContext, error) {
+	cPool := C.CString(pool)
+	defer C.free(unsafe.Pointer(cPool))
+	ioctx := &IOContext{conn: c}
+	ret := C.rados_ioctx_create(c.cluster, cPool, &ioctx.ioctx)
+	if ret == 0 {
+		return ioctx, nil
+	}
+	return nil, getError(ret)
+}
+
+// ListPools returns the names of all existing pools.
+func (c *Conn) ListPools() (names []string, err error) {
+	buf := make([]byte, 4096)
+	for {
+		ret := C.rados_pool_list(c.cluster,
+			(*C.char)(unsafe.Pointer(&buf[0])), C.size_t(len(buf)))
+		if ret < 0 {
+			return nil, getError(ret)
+		}
+
+		if int(ret) > len(buf) {
+			buf = make([]byte, ret)
+			continue
+		}
+
+		names = cutil.SplitSparseBuffer(buf[:ret])
+		return names, nil
+	}
+}
+
+// SetConfigOption sets the value of the configuration option identified by
+// the given name.
+func (c *Conn) SetConfigOption(option, value string) error {
+	cOpt, cVal := C.CString(option), C.CString(value)
+	defer C.free(unsafe.Pointer(cOpt))
+	defer C.free(unsafe.Pointer(cVal))
+	ret := C.rados_conf_set(c.cluster, cOpt, cVal)
+	return getError(ret)
+}
+
+// GetConfigOption returns the value of the Ceph configuration option
+// identified by the given name.
+func (c *Conn) GetConfigOption(name string) (value string, err error) {
+	cOption := C.CString(name)
+	defer C.free(unsafe.Pointer(cOption))
+
+	var buf []byte
+	// range from 4k to 256KiB
+	retry.WithSizes(4096, 1<<18, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret := C.rados_conf_get(
+			c.cluster,
+			cOption,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			C.size_t(len(buf)))
+		err = getError(ret)
+		return retry.DoubleSize.If(err == errNameTooLong)
+	})
+	if err != nil {
+		return "", err
+	}
+	value = C.GoString((*C.char)(unsafe.Pointer(&buf[0])))
+	return value, nil
+}
+
+// WaitForLatestOSDMap blocks the caller until the latest OSD map has been
+// retrieved.
+func (c *Conn) WaitForLatestOSDMap() error {
+	ret := C.rados_wait_for_latest_osdmap(c.cluster)
+	return getError(ret)
+}
+
+func (c *Conn) ensureConnected() error {
+	if c.connected {
+		return nil
+	}
+	return ErrNotConnected
+}
+
+// GetClusterStats returns statistics about the cluster associated with the
+// connection.
+func (c *Conn) GetClusterStats() (stat ClusterStat, err error) {
+	if err := c.ensureConnected(); err != nil {
+		return ClusterStat{}, err
+	}
+	cStat := C.struct_rados_cluster_stat_t{}
+	ret := C.rados_cluster_stat(c.cluster, &cStat)
+	if ret < 0 {
+		return ClusterStat{}, getError(ret)
+	}
+	return ClusterStat{
+		Kb:          uint64(cStat.kb),
+		Kb_used:     uint64(cStat.kb_used),
+		Kb_avail:    uint64(cStat.kb_avail),
+		Num_objects: uint64(cStat.num_objects),
+	}, nil
+}
+
+// ParseConfigArgv configures the connection using a unix style command line
+// argument vector.
+//
+// Implements:
+//  int rados_conf_parse_argv(rados_t cluster, int argc,
+//                            const char **argv);
+func (c *Conn) ParseConfigArgv(argv []string) error {
+	if c.cluster == nil {
+		return ErrNotConnected
+	}
+	if len(argv) == 0 {
+		return ErrEmptyArgument
+	}
+	cargv := make([]*C.char, len(argv))
+	for i := range argv {
+		cargv[i] = C.CString(argv[i])
+		defer C.free(unsafe.Pointer(cargv[i]))
+	}
+
+	ret := C.rados_conf_parse_argv(c.cluster, C.int(len(cargv)), &cargv[0])
+	return getError(ret)
+}
+
+// ParseCmdLineArgs configures the connection from command line arguments.
+//
+// This function passes a placeholder value to Ceph as argv[0], see
+// ParseConfigArgv for a version of this function that allows the caller to
+// specify argv[0].
+func (c *Conn) ParseCmdLineArgs(args []string) error {
+	argv := make([]string, len(args)+1)
+	// Ceph expects a proper argv array as the actual contents with the
+	// first element containing the executable name
+	argv[0] = argvPlaceholder
+	for i := range args {
+		argv[i+1] = args[i]
+	}
+	return c.ParseConfigArgv(argv)
+}
+
+// ParseDefaultConfigEnv configures the connection from the default Ceph
+// environment variable CEPH_ARGS.
+func (c *Conn) ParseDefaultConfigEnv() error {
+	ret := C.rados_conf_parse_env(c.cluster, nil)
+	return getError(ret)
+}
+
+// GetFSID returns the fsid of the cluster as a hexadecimal string. The fsid
+// is a unique identifier of an entire Ceph cluster.
+func (c *Conn) GetFSID() (fsid string, err error) {
+	buf := make([]byte, 37)
+	ret := C.rados_cluster_fsid(c.cluster,
+		(*C.char)(unsafe.Pointer(&buf[0])), C.size_t(len(buf)))
+	// FIXME: the success case isn't documented correctly in librados.h
+	if ret == 36 {
+		fsid = C.GoString((*C.char)(unsafe.Pointer(&buf[0])))
+		return fsid, nil
+	}
+	return "", getError(ret)
+}
+
+// GetInstanceID returns a globally unique identifier for the cluster
+// connection instance.
+func (c *Conn) GetInstanceID() uint64 {
+	// FIXME: are there any error cases for this?
+	return uint64(C.rados_get_instance_id(c.cluster))
+}
+
+// MakePool creates a new pool with default settings.
+func (c *Conn) MakePool(name string) error {
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+	ret := C.rados_pool_create(c.cluster, cName)
+	return getError(ret)
+}
+
+// DeletePool deletes a pool and all the data inside the pool.
+func (c *Conn) DeletePool(name string) error {
+	if err := c.ensureConnected(); err != nil {
+		return err
+	}
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+	ret := C.rados_pool_delete(c.cluster, cName)
+	return getError(ret)
+}
+
+// GetPoolByName returns the ID of the pool with a given name.
+func (c *Conn) GetPoolByName(name string) (int64, error) {
+	if err := c.ensureConnected(); err != nil {
+		return 0, err
+	}
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+	ret := int64(C.rados_pool_lookup(c.cluster, cName))
+	if ret < 0 {
+		return 0, radosError(ret)
+	}
+	return ret, nil
+}
+
+// GetPoolByID returns the name of a pool by a given ID.
+func (c *Conn) GetPoolByID(id int64) (string, error) {
+	buf := make([]byte, 4096)
+	if err := c.ensureConnected(); err != nil {
+		return "", err
+	}
+	cid := C.int64_t(id)
+	ret := int(C.rados_pool_reverse_lookup(c.cluster, cid, (*C.char)(unsafe.Pointer(&buf[0])), C.size_t(len(buf))))
+	if ret < 0 {
+		return "", radosError(ret)
+	}
+	return C.GoString((*C.char)(unsafe.Pointer(&buf[0]))), nil
+}

vendor/github.com/ceph/go-ceph/rados/doc.go

@@ -0,0 +1,4 @@
+/*
+Package rados contains a set of wrappers around Ceph's librados API.
+*/
+package rados

vendor/github.com/ceph/go-ceph/rados/errors.go

@@ -0,0 +1,86 @@
+package rados
+
+/*
+#include <errno.h>
+*/
+import "C"
+
+import (
+	"errors"
+
+	"github.com/ceph/go-ceph/internal/errutil"
+)
+
+// radosError represents an error condition returned from the Ceph RADOS APIs.
+type radosError int
+
+// Error returns the error string for the radosError type.
+func (e radosError) Error() string {
+	return errutil.FormatErrorCode("rados", int(e))
+}
+
+func (e radosError) ErrorCode() int {
+	return int(e)
+}
+
+func getError(e C.int) error {
+	if e == 0 {
+		return nil
+	}
+	return radosError(e)
+}
+
+// getErrorIfNegative converts a ceph return code to error if negative.
+// This is useful for functions that return a usable positive value on
+// success but a negative error number on error.
+func getErrorIfNegative(ret C.int) error {
+	if ret >= 0 {
+		return nil
+	}
+	return getError(ret)
+}
+
+// Public go errors:
+
+var (
+	// ErrNotConnected is returned when functions are called
+	// without a RADOS connection.
+	ErrNotConnected = errors.New("RADOS not connected")
+	// ErrEmptyArgument may be returned if a function argument is passed
+	// a zero-length slice or map.
+	ErrEmptyArgument = errors.New("Argument must contain at least one item")
+	// ErrInvalidIOContext may be returned if an api call requires an IOContext
+	// but IOContext is not ready for use.
+	ErrInvalidIOContext = errors.New("IOContext is not ready for use")
+	// ErrOperationIncomplete is returned from write op or read op steps for
+	// which the operation has not been performed yet.
+	ErrOperationIncomplete = errors.New("Operation has not been performed yet")
+)
+
+// Public radosErrors:
+
+const (
+	// ErrNotFound indicates a missing resource.
+	ErrNotFound = radosError(-C.ENOENT)
+	// ErrPermissionDenied indicates a permissions issue.
+	ErrPermissionDenied = radosError(-C.EPERM)
+	// ErrObjectExists indicates that an exclusive object creation failed.
+	ErrObjectExists = radosError(-C.EEXIST)
+
+	// RadosErrorNotFound indicates a missing resource.
+	//
+	// Deprecated: use ErrNotFound instead
+	RadosErrorNotFound = ErrNotFound
+	// RadosErrorPermissionDenied indicates a permissions issue.
+	//
+	// Deprecated: use ErrPermissionDenied instead
+	RadosErrorPermissionDenied = ErrPermissionDenied
+)
+
+// Private errors:
+
+const (
+	errNameTooLong = radosError(-C.ENAMETOOLONG)
+
+	errRange = radosError(-C.ERANGE)
+)

vendor/github.com/ceph/go-ceph/rados/ioctx.go

@@ -0,0 +1,717 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <errno.h>
+// #include <stdlib.h>
+// #include <rados/librados.h>
+//
+// char* nextChunk(char **idx) {
+// 	char *copy;
+// 	copy = strdup(*idx);
+// 	*idx += strlen(*idx) + 1;
+// 	return copy;
+// }
+//
+// #if __APPLE__
+// #define ceph_time_t __darwin_time_t
+// #define ceph_suseconds_t __darwin_suseconds_t
+// #elif __GLIBC__
+// #define ceph_time_t __time_t
+// #define ceph_suseconds_t __suseconds_t
+// #else
+// #define ceph_time_t time_t
+// #define ceph_suseconds_t suseconds_t
+// #endif
+import "C"
+
+import (
+	"syscall"
+	"time"
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/retry"
+)
+
+// CreateOption is passed to IOContext.Create() and should be one of
+// CreateExclusive or CreateIdempotent.
+type CreateOption int
+
+const (
+	// CreateExclusive if used with IOContext.Create() and the object
+	// already exists, the function will return an error.
+	CreateExclusive = C.LIBRADOS_CREATE_EXCLUSIVE
+	// CreateIdempotent if used with IOContext.Create() and the object
+	// already exists, the function will not return an error.
+	CreateIdempotent = C.LIBRADOS_CREATE_IDEMPOTENT
+
+	defaultListObjectsResultSize = 1000
+	// listEndSentinel is the value returned by rados_list_object_list_is_end
+	// when a cursor has reached the end of a pool
+	listEndSentinel = 1
+)
+
+//revive:disable:var-naming old-yet-exported public api
+
+// PoolStat represents Ceph pool statistics.
+type PoolStat struct {
+	// space used in bytes
+	Num_bytes uint64
+	// space used in KB
+	Num_kb uint64
+	// number of objects in the pool
+	Num_objects uint64
+	// number of clones of objects
+	Num_object_clones uint64
+	// num_objects * num_replicas
+	Num_object_copies              uint64
+	Num_objects_missing_on_primary uint64
+	// number of objects found on no OSDs
+	Num_objects_unfound uint64
+	// number of objects replicated fewer times than they should be
+	// (but found on at least one OSD)
+	Num_objects_degraded uint64
+	Num_rd               uint64
+	Num_rd_kb            uint64
+	Num_wr               uint64
+	Num_wr_kb            uint64
+}
+
+//revive:enable:var-naming
+
+// ObjectStat represents an object stat information
+type ObjectStat struct {
+	// current length in bytes
+	Size uint64
+	// last modification time
+	ModTime time.Time
+}
+
+// LockInfo represents information on a current Ceph lock
+type LockInfo struct {
+	NumLockers int
+	Exclusive  bool
+	Tag        string
+	Clients    []string
+	Cookies    []string
+	Addrs      []string
+}
+
+// IOContext represents a context for performing I/O within a pool.
+type IOContext struct {
+	ioctx C.rados_ioctx_t
+
+	// Hold a reference back to the connection that the ioctx depends on so
+	// that Go's GC doesn't trigger the Conn's finalizer before this
+	// IOContext is destroyed.
+	conn *Conn
+}
+
+// validate returns an error if the ioctx is not ready to be used
+// with ceph C calls.
+func (ioctx *IOContext) validate() error {
+	if ioctx.ioctx == nil {
+		return ErrInvalidIOContext
+	}
+	return nil
+}
+
+// Pointer returns a pointer reference to an internal structure.
+// This function should NOT be used outside of go-ceph itself.
+func (ioctx *IOContext) Pointer() unsafe.Pointer {
+	return unsafe.Pointer(ioctx.ioctx)
+}
+
+// SetNamespace sets the namespace for objects within this IO context (pool).
+// Setting namespace to a empty or zero length string sets the pool to the default namespace.
+//
+// Implements:
+//  void rados_ioctx_set_namespace(rados_ioctx_t io,
+//                                 const char *nspace);
+func (ioctx *IOContext) SetNamespace(namespace string) {
+	var cns *C.char
+	if len(namespace) > 0 {
+		cns = C.CString(namespace)
+		defer C.free(unsafe.Pointer(cns))
+	}
+	C.rados_ioctx_set_namespace(ioctx.ioctx, cns)
+}
+
+// Create a new object with key oid.
+//
+// Implements:
+//  void rados_write_op_create(rados_write_op_t write_op, int exclusive,
+//                             const char* category)
+func (ioctx *IOContext) Create(oid string, exclusive CreateOption) error {
+	op := CreateWriteOp()
+	defer op.Release()
+	op.Create(exclusive)
+	return op.operateCompat(ioctx, oid)
+}
+
+// Write writes len(data) bytes to the object with key oid starting at byte
+// offset offset. It returns an error, if any.
+func (ioctx *IOContext) Write(oid string, data []byte, offset uint64) error {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	dataPointer := unsafe.Pointer(nil)
+	if len(data) > 0 {
+		dataPointer = unsafe.Pointer(&data[0])
+	}
+
+	ret := C.rados_write(ioctx.ioctx, coid,
+		(*C.char)(dataPointer),
+		(C.size_t)(len(data)),
+		(C.uint64_t)(offset))
+
+	return getError(ret)
+}
+
+// WriteFull writes len(data) bytes to the object with key oid.
+// The object is filled with the provided data. If the object exists,
+// it is atomically truncated and then written. It returns an error, if any.
+func (ioctx *IOContext) WriteFull(oid string, data []byte) error {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	ret := C.rados_write_full(ioctx.ioctx, coid,
+		(*C.char)(unsafe.Pointer(&data[0])),
+		(C.size_t)(len(data)))
+	return getError(ret)
+}
+
+// Append appends len(data) bytes to the object with key oid.
+// The object is appended with the provided data. If the object exists,
+// it is atomically appended to. It returns an error, if any.
+func (ioctx *IOContext) Append(oid string, data []byte) error {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	ret := C.rados_append(ioctx.ioctx, coid,
+		(*C.char)(unsafe.Pointer(&data[0])),
+		(C.size_t)(len(data)))
+	return getError(ret)
+}
+
+// Read reads up to len(data) bytes from the object with key oid starting at byte
+// offset offset. It returns the number of bytes read and an error, if any.
+func (ioctx *IOContext) Read(oid string, data []byte, offset uint64) (int, error) {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	var buf *C.char
+	if len(data) > 0 {
+		buf = (*C.char)(unsafe.Pointer(&data[0]))
+	}
+
+	ret := C.rados_read(
+		ioctx.ioctx,
+		coid,
+		buf,
+		(C.size_t)(len(data)),
+		(C.uint64_t)(offset))
+
+	if ret >= 0 {
+		return int(ret), nil
+	}
+	return 0, getError(ret)
+}
+
+// Delete deletes the object with key oid. It returns an error, if any.
+func (ioctx *IOContext) Delete(oid string) error {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	return getError(C.rados_remove(ioctx.ioctx, coid))
+}
+
+// Truncate resizes the object with key oid to size size. If the operation
+// enlarges the object, the new area is logically filled with zeroes. If the
+// operation shrinks the object, the excess data is removed. It returns an
+// error, if any.
+func (ioctx *IOContext) Truncate(oid string, size uint64) error {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	return getError(C.rados_trunc(ioctx.ioctx, coid, (C.uint64_t)(size)))
+}
+
+// Destroy informs librados that the I/O context is no longer in use.
+// Resources associated with the context may not be freed immediately, and the
+// context should not be used again after calling this method.
+func (ioctx *IOContext) Destroy() {
+	C.rados_ioctx_destroy(ioctx.ioctx)
+}
+
+// GetPoolStats returns a set of statistics about the pool associated with this I/O
+// context.
+//
+// Implements:
+//  int rados_ioctx_pool_stat(rados_ioctx_t io,
+//                            struct rados_pool_stat_t *stats);
+func (ioctx *IOContext) GetPoolStats() (stat PoolStat, err error) {
+	cStat := C.struct_rados_pool_stat_t{}
+	ret := C.rados_ioctx_pool_stat(ioctx.ioctx, &cStat)
+	if ret < 0 {
+		return PoolStat{}, getError(ret)
+	}
+	return PoolStat{
+		Num_bytes:                      uint64(cStat.num_bytes),
+		Num_kb:                         uint64(cStat.num_kb),
+		Num_objects:                    uint64(cStat.num_objects),
+		Num_object_clones:              uint64(cStat.num_object_clones),
+		Num_object_copies:              uint64(cStat.num_object_copies),
+		Num_objects_missing_on_primary: uint64(cStat.num_objects_missing_on_primary),
+		Num_objects_unfound:            uint64(cStat.num_objects_unfound),
+		Num_objects_degraded:           uint64(cStat.num_objects_degraded),
+		Num_rd:                         uint64(cStat.num_rd),
+		Num_rd_kb:                      uint64(cStat.num_rd_kb),
+		Num_wr:                         uint64(cStat.num_wr),
+		Num_wr_kb:                      uint64(cStat.num_wr_kb),
+	}, nil
+}
+
+// GetPoolID returns the pool ID associated with the I/O context.
+//
+// Implements:
+//  int64_t rados_ioctx_get_id(rados_ioctx_t io)
+func (ioctx *IOContext) GetPoolID() int64 {
+	ret := C.rados_ioctx_get_id(ioctx.ioctx)
+	return int64(ret)
+}
+
+// GetPoolName returns the name of the pool associated with the I/O context.
+func (ioctx *IOContext) GetPoolName() (name string, err error) {
+	var (
+		buf []byte
+		ret C.int
+	)
+	retry.WithSizes(128, 8192, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.rados_ioctx_get_pool_name(
+			ioctx.ioctx,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			C.unsigned(len(buf)))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return "", err
+	}
+	name = C.GoStringN((*C.char)(unsafe.Pointer(&buf[0])), ret)
+	return name, nil
+}
+
+// ObjectListFunc is the type of the function called for each object visited
+// by ListObjects.
+type ObjectListFunc func(oid string)
+
+// ListObjects lists all of the objects in the pool associated with the I/O
+// context, and called the provided listFn function for each object, passing
+// to the function the name of the object. Call SetNamespace with
+// RadosAllNamespaces before calling this function to return objects from all
+// namespaces
+func (ioctx *IOContext) ListObjects(listFn ObjectListFunc) error {
+	pageResults := C.size_t(defaultListObjectsResultSize)
+	var filterLen C.size_t
+	results := make([]C.rados_object_list_item, pageResults)
+
+	next := C.rados_object_list_begin(ioctx.ioctx)
+	if next == nil {
+		return ErrNotFound
+	}
+	defer C.rados_object_list_cursor_free(ioctx.ioctx, next)
+	finish := C.rados_object_list_end(ioctx.ioctx)
+	if finish == nil {
+		return ErrNotFound
+	}
+	defer C.rados_object_list_cursor_free(ioctx.ioctx, finish)
+
+	for {
+		ret := C.rados_object_list(ioctx.ioctx, next, finish, pageResults, nil, filterLen, (*C.rados_object_list_item)(unsafe.Pointer(&results[0])), &next)
+		if ret < 0 {
+			return getError(ret)
+		}
+
+		numEntries := int(ret)
+		for i := 0; i < numEntries; i++ {
+			item := results[i]
+			listFn(C.GoStringN(item.oid, (C.int)(item.oid_length)))
+		}
+
+		if C.rados_object_list_is_end(ioctx.ioctx, next) == listEndSentinel {
+			return nil
+		}
+	}
+}
+
+// Stat returns the size of the object and its last modification time
+func (ioctx *IOContext) Stat(object string) (stat ObjectStat, err error) {
+	var cPsize C.uint64_t
+	var cPmtime C.time_t
+	cObject := C.CString(object)
+	defer C.free(unsafe.Pointer(cObject))
+
+	ret := C.rados_stat(
+		ioctx.ioctx,
+		cObject,
+		&cPsize,
+		&cPmtime)
+
+	if ret < 0 {
+		return ObjectStat{}, getError(ret)
+	}
+	return ObjectStat{
+		Size:    uint64(cPsize),
+		ModTime: time.Unix(int64(cPmtime), 0),
+	}, nil
+}
+
+// GetXattr gets an xattr with key `name`, it returns the length of
+// the key read or an error if not successful
+func (ioctx *IOContext) GetXattr(object string, name string, data []byte) (int, error) {
+	cObject := C.CString(object)
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cObject))
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.rados_getxattr(
+		ioctx.ioctx,
+		cObject,
+		cName,
+		(*C.char)(unsafe.Pointer(&data[0])),
+		(C.size_t)(len(data)))
+
+	if ret >= 0 {
+		return int(ret), nil
+	}
+	return 0, getError(ret)
+}
+
+// SetXattr sets an xattr for an object with key `name` with value as `data`
+func (ioctx *IOContext) SetXattr(object string, name string, data []byte) error {
+	cObject := C.CString(object)
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cObject))
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.rados_setxattr(
+		ioctx.ioctx,
+		cObject,
+		cName,
+		(*C.char)(unsafe.Pointer(&data[0])),
+		(C.size_t)(len(data)))
+
+	return getError(ret)
+}
+
+// ListXattrs lists all the xattrs for an object. The xattrs are returned as a
+// mapping of string keys and byte-slice values.
+func (ioctx *IOContext) ListXattrs(oid string) (map[string][]byte, error) {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	var it C.rados_xattrs_iter_t
+
+	ret := C.rados_getxattrs(ioctx.ioctx, coid, &it)
+	if ret < 0 {
+		return nil, getError(ret)
+	}
+	defer func() { C.rados_getxattrs_end(it) }()
+	m := make(map[string][]byte)
+	for {
+		var cName, cVal *C.char
+		var cLen C.size_t
+		defer C.free(unsafe.Pointer(cName))
+		defer C.free(unsafe.Pointer(cVal))
+
+		ret := C.rados_getxattrs_next(it, &cName, &cVal, &cLen)
+		if ret < 0 {
+			return nil, getError(ret)
+		}
+		// rados api returns a null name,val & 0-length upon
+		// end of iteration
+		if cName == nil {
+			return m, nil // stop iteration
+		}
+		m[C.GoString(cName)] = C.GoBytes(unsafe.Pointer(cVal), (C.int)(cLen))
+	}
+}
+
+// RmXattr removes an xattr with key `name` from object `oid`
+func (ioctx *IOContext) RmXattr(oid string, name string) error {
+	coid := C.CString(oid)
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(coid))
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.rados_rmxattr(
+		ioctx.ioctx,
+		coid,
+		cName)
+
+	return getError(ret)
+}
+
+// LockExclusive takes an exclusive lock on an object.
+func (ioctx *IOContext) LockExclusive(oid, name, cookie, desc string, duration time.Duration, flags *byte) (int, error) {
+	coid := C.CString(oid)
+	cName := C.CString(name)
+	cCookie := C.CString(cookie)
+	cDesc := C.CString(desc)
+
+	var cDuration C.struct_timeval
+	if duration != 0 {
+		tv := syscall.NsecToTimeval(duration.Nanoseconds())
+		cDuration = C.struct_timeval{tv_sec: C.ceph_time_t(tv.Sec), tv_usec: C.ceph_suseconds_t(tv.Usec)}
+	}
+
+	var cFlags C.uint8_t
+	if flags != nil {
+		cFlags = C.uint8_t(*flags)
+	}
+
+	defer C.free(unsafe.Pointer(coid))
+	defer C.free(unsafe.Pointer(cName))
+	defer C.free(unsafe.Pointer(cCookie))
+	defer C.free(unsafe.Pointer(cDesc))
+
+	ret := C.rados_lock_exclusive(
+		ioctx.ioctx,
+		coid,
+		cName,
+		cCookie,
+		cDesc,
+		&cDuration,
+		cFlags)
+
+	// 0 on success, negative error code on failure
+	// -EBUSY if the lock is already held by another (client, cookie) pair
+	// -EEXIST if the lock is already held by the same (client, cookie) pair
+
+	switch ret {
+	case 0:
+		return int(ret), nil
+	case -C.EBUSY:
+		return int(ret), nil
+	case -C.EEXIST:
+		return int(ret), nil
+	default:
+		return int(ret), getError(ret)
+	}
+}
+
+// LockShared takes a shared lock on an object.
+func (ioctx *IOContext) LockShared(oid, name, cookie, tag, desc string, duration time.Duration, flags *byte) (int, error) {
+	coid := C.CString(oid)
+	cName := C.CString(name)
+	cCookie := C.CString(cookie)
+	cTag := C.CString(tag)
+	cDesc := C.CString(desc)
+
+	var cDuration C.struct_timeval
+	if duration != 0 {
+		tv := syscall.NsecToTimeval(duration.Nanoseconds())
+		cDuration = C.struct_timeval{tv_sec: C.ceph_time_t(tv.Sec), tv_usec: C.ceph_suseconds_t(tv.Usec)}
+	}
+
+	var cFlags C.uint8_t
+	if flags != nil {
+		cFlags = C.uint8_t(*flags)
+	}
+
+	defer C.free(unsafe.Pointer(coid))
+	defer C.free(unsafe.Pointer(cName))
+	defer C.free(unsafe.Pointer(cCookie))
+	defer C.free(unsafe.Pointer(cTag))
+	defer C.free(unsafe.Pointer(cDesc))
+
+	ret := C.rados_lock_shared(
+		ioctx.ioctx,
+		coid,
+		cName,
+		cCookie,
+		cTag,
+		cDesc,
+		&cDuration,
+		cFlags)
+
+	// 0 on success, negative error code on failure
+	// -EBUSY if the lock is already held by another (client, cookie) pair
+	// -EEXIST if the lock is already held by the same (client, cookie) pair
+
+	switch ret {
+	case 0:
+		return int(ret), nil
+	case -C.EBUSY:
+		return int(ret), nil
+	case -C.EEXIST:
+		return int(ret), nil
+	default:
+		return int(ret), getError(ret)
+	}
+}
+
+// Unlock releases a shared or exclusive lock on an object.
+func (ioctx *IOContext) Unlock(oid, name, cookie string) (int, error) {
+	coid := C.CString(oid)
+	cName := C.CString(name)
+	cCookie := C.CString(cookie)
+
+	defer C.free(unsafe.Pointer(coid))
+	defer C.free(unsafe.Pointer(cName))
+	defer C.free(unsafe.Pointer(cCookie))
+
+	// 0 on success, negative error code on failure
+	// -ENOENT if the lock is not held by the specified (client, cookie) pair
+
+	ret := C.rados_unlock(
+		ioctx.ioctx,
+		coid,
+		cName,
+		cCookie)
+
+	switch ret {
+	case 0:
+		return int(ret), nil
+	case -C.ENOENT:
+		return int(ret), nil
+	default:
+		return int(ret), getError(ret)
+	}
+}
+
+// ListLockers lists clients that have locked the named object lock and
+// information about the lock.
+// The number of bytes required in each buffer is put in the corresponding size
+// out parameter.  If any of the provided buffers are too short, -ERANGE is
+// returned after these sizes are filled in.
+func (ioctx *IOContext) ListLockers(oid, name string) (*LockInfo, error) {
+	coid := C.CString(oid)
+	cName := C.CString(name)
+
+	cTag := (*C.char)(C.malloc(C.size_t(1024)))
+	cClients := (*C.char)(C.malloc(C.size_t(1024)))
+	cCookies := (*C.char)(C.malloc(C.size_t(1024)))
+	cAddrs := (*C.char)(C.malloc(C.size_t(1024)))
+
+	var cExclusive C.int
+	cTagLen := C.size_t(1024)
+	cClientsLen := C.size_t(1024)
+	cCookiesLen := C.size_t(1024)
+	cAddrsLen := C.size_t(1024)
+
+	defer C.free(unsafe.Pointer(coid))
+	defer C.free(unsafe.Pointer(cName))
+	defer C.free(unsafe.Pointer(cTag))
+	defer C.free(unsafe.Pointer(cClients))
+	defer C.free(unsafe.Pointer(cCookies))
+	defer C.free(unsafe.Pointer(cAddrs))
+
+	ret := C.rados_list_lockers(
+		ioctx.ioctx,
+		coid,
+		cName,
+		&cExclusive,
+		cTag,
+		&cTagLen,
+		cClients,
+		&cClientsLen,
+		cCookies,
+		&cCookiesLen,
+		cAddrs,
+		&cAddrsLen)
+
+	splitCString := func(items *C.char, itemsLen C.size_t) []string {
+		currLen := 0
+		clients := []string{}
+		for currLen < int(itemsLen) {
+			client := C.GoString(C.nextChunk(&items))
+			clients = append(clients, client)
+			currLen += len(client) + 1
+		}
+		return clients
+	}
+
+	if ret < 0 {
+		return nil, radosError(ret)
+	}
+	return &LockInfo{int(ret), cExclusive == 1, C.GoString(cTag), splitCString(cClients, cClientsLen), splitCString(cCookies, cCookiesLen), splitCString(cAddrs, cAddrsLen)}, nil
+}
+
+// BreakLock releases a shared or exclusive lock on an object, which was taken by the specified client.
+func (ioctx *IOContext) BreakLock(oid, name, client, cookie string) (int, error) {
+	coid := C.CString(oid)
+	cName := C.CString(name)
+	cClient := C.CString(client)
+	cCookie := C.CString(cookie)
+
+	defer C.free(unsafe.Pointer(coid))
+	defer C.free(unsafe.Pointer(cName))
+	defer C.free(unsafe.Pointer(cClient))
+	defer C.free(unsafe.Pointer(cCookie))
+
+	// 0 on success, negative error code on failure
+	// -ENOENT if the lock is not held by the specified (client, cookie) pair
+	// -EINVAL if the client cannot be parsed
+
+	ret := C.rados_break_lock(
+		ioctx.ioctx,
+		coid,
+		cName,
+		cClient,
+		cCookie)
+
+	switch ret {
+	case 0:
+		return int(ret), nil
+	case -C.ENOENT:
+		return int(ret), nil
+	case -C.EINVAL: // -EINVAL
+		return int(ret), nil
+	default:
+		return int(ret), getError(ret)
+	}
+}
+
+// GetLastVersion will return the version number of the last object read or
+// written to.
+//
+// Implements:
+//  uint64_t rados_get_last_version(rados_ioctx_t io);
+func (ioctx *IOContext) GetLastVersion() (uint64, error) {
+	if err := ioctx.validate(); err != nil {
+		return 0, err
+	}
+	v := C.rados_get_last_version(ioctx.ioctx)
+	return uint64(v), nil
+}
+
+// GetNamespace gets the namespace used for objects within this IO context.
+//
+// Implements:
+//  int rados_ioctx_get_namespace(rados_ioctx_t io, char *buf,
+//                                unsigned maxlen);
+func (ioctx *IOContext) GetNamespace() (string, error) {
+	if err := ioctx.validate(); err != nil {
+		return "", err
+	}
+	var (
+		err error
+		buf []byte
+		ret C.int
+	)
+	retry.WithSizes(128, 8192, func(size int) retry.Hint {
+		buf = make([]byte, size)
+		ret = C.rados_ioctx_get_namespace(
+			ioctx.ioctx,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			C.unsigned(len(buf)))
+		err = getErrorIfNegative(ret)
+		return retry.DoubleSize.If(err == errRange)
+	})
+	if err != nil {
+		return "", err
+	}
+	return string(buf[:ret]), nil
+}

vendor/github.com/ceph/go-ceph/rados/ioctx_nautilus.go

@@ -0,0 +1,35 @@
+//go:build nautilus
+// +build nautilus
+
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+//
+import "C"
+
+// SetPoolFullTry makes sure to send requests to the cluster despite
+// the cluster or pool being marked full; ops will either succeed(e.g., delete)
+// or return EDQUOT or ENOSPC.
+//
+// Implements:
+//  void rados_set_osdmap_full_try(rados_ioctx_t io);
+func (ioctx *IOContext) SetPoolFullTry() error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+	C.rados_set_osdmap_full_try(ioctx.ioctx)
+	return nil
+}
+
+// UnsetPoolFullTry unsets the flag set by SetPoolFullTry()
+//
+// Implements:
+//  void rados_unset_osdmap_full_try(rados_ioctx_t io);
+func (ioctx *IOContext) UnsetPoolFullTry() error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+	C.rados_unset_osdmap_full_try(ioctx.ioctx)
+	return nil
+}

vendor/github.com/ceph/go-ceph/rados/ioctx_octopus.go

@@ -0,0 +1,38 @@
+//go:build !nautilus
+// +build !nautilus
+
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+//
+import "C"
+
+// Ceph octopus deprecates rados_set_osdmap_full_try() and implements rados_set_pool_full_try()
+// Ceph octopus deprecates rados_unset_osdmap_full_try() and implements rados_unset_pool_full_try()
+
+// SetPoolFullTry makes sure to send requests to the cluster despite
+// the cluster or pool being marked full; ops will either succeed(e.g., delete)
+// or return EDQUOT or ENOSPC.
+//
+// Implements:
+//  void rados_set_pool_full_try(rados_ioctx_t io);
+func (ioctx *IOContext) SetPoolFullTry() error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+	C.rados_set_pool_full_try(ioctx.ioctx)
+	return nil
+}
+
+// UnsetPoolFullTry unsets the flag set by SetPoolFullTry()
+//
+// Implements:
+//  void rados_unset_pool_full_try(rados_ioctx_t io);
+func (ioctx *IOContext) UnsetPoolFullTry() error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+	C.rados_unset_pool_full_try(ioctx.ioctx)
+	return nil
+}

vendor/github.com/ceph/go-ceph/rados/ioctx_pool_alignment.go

@@ -0,0 +1,27 @@
+//go:build ceph_preview
+// +build ceph_preview
+
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+// Alignment returns the required stripe size in bytes for pools supporting/requiring it, or an error if unsuccessful.
+// For an EC pool, a buffer size multiple of its stripe size is required to call Append. To know if the pool requires
+// alignment or not, use RequiresAlignment.
+//
+// Implements:
+//  int rados_ioctx_pool_required_alignment2(rados_ioctx_t io, uint64_t *alignment)
+func (ioctx *IOContext) Alignment() (uint64, error) {
+	var alignSizeBytes C.uint64_t
+	ret := C.rados_ioctx_pool_required_alignment2(
+		ioctx.ioctx,
+		&alignSizeBytes)
+	if ret != 0 {
+		return 0, getError(ret)
+	}
+	return uint64(alignSizeBytes), nil
+}

vendor/github.com/ceph/go-ceph/rados/ioctx_pool_requires_alignment.go

@@ -0,0 +1,27 @@
+//go:build ceph_preview
+// +build ceph_preview
+
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+// RequiresAlignment returns true if the pool supports/requires alignment or an error if not successful.
+// For an EC pool, a buffer size multiple of its stripe size is required to call Append. See
+// Alignment to know how to get the stripe size for pools requiring it.
+//
+// Implements:
+//  int rados_ioctx_pool_requires_alignment2(rados_ioctx_t io, int *req)
+func (ioctx *IOContext) RequiresAlignment() (bool, error) {
+	var alignRequired C.int
+	ret := C.rados_ioctx_pool_requires_alignment2(
+		ioctx.ioctx,
+		&alignRequired)
+	if ret != 0 {
+		return false, getError(ret)
+	}
+	return (alignRequired != 0), nil
+}

vendor/github.com/ceph/go-ceph/rados/ioctx_set_alloc_hint.go

@@ -0,0 +1,38 @@
+//go:build ceph_preview
+// +build ceph_preview
+
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+import (
+	"unsafe"
+)
+
+// SetAllocationHint sets allocation hint for an object. This is an advisory
+// operation, it will always succeed (as if it was submitted with a
+// LIBRADOS_OP_FLAG_FAILOK flag set) and is not guaranteed to do anything on
+// the backend.
+//
+// Implements:
+//  int rados_set_alloc_hint2(rados_ioctx_t io,
+//                            const char *o,
+//                            uint64_t expected_object_size,
+//                            uint64_t expected_write_size,
+//                            uint32_t flags);
+func (ioctx *IOContext) SetAllocationHint(oid string, expectedObjectSize uint64, expectedWriteSize uint64, flags AllocHintFlags) error {
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+
+	return getError(C.rados_set_alloc_hint2(
+		ioctx.ioctx,
+		coid,
+		(C.uint64_t)(expectedObjectSize),
+		(C.uint64_t)(expectedWriteSize),
+		(C.uint32_t)(flags),
+	))
+}

vendor/github.com/ceph/go-ceph/rados/object_iter.go

@@ -0,0 +1,92 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+//
+import "C"
+
+// Iter supports iterating over objects in the ioctx.
+type Iter struct {
+	ctx       C.rados_list_ctx_t
+	err       error
+	entry     string
+	namespace string
+}
+
+// IterToken supports reporting on and seeking to different positions.
+type IterToken uint32
+
+// Iter returns a Iterator object that can be used to list the object names in the current pool
+func (ioctx *IOContext) Iter() (*Iter, error) {
+	iter := Iter{}
+	if cerr := C.rados_nobjects_list_open(ioctx.ioctx, &iter.ctx); cerr < 0 {
+		return nil, getError(cerr)
+	}
+	return &iter, nil
+}
+
+// Token returns a token marking the current position of the iterator. To be used in combination with Iter.Seek()
+func (iter *Iter) Token() IterToken {
+	return IterToken(C.rados_nobjects_list_get_pg_hash_position(iter.ctx))
+}
+
+// Seek moves the iterator to the position indicated by the token.
+func (iter *Iter) Seek(token IterToken) {
+	C.rados_nobjects_list_seek(iter.ctx, C.uint32_t(token))
+}
+
+// Next retrieves the next object name in the pool/namespace iterator.
+// Upon a successful invocation (return value of true), the Value method should
+// be used to obtain the name of the retrieved object name. When the iterator is
+// exhausted, Next returns false. The Err method should used to verify whether the
+// end of the iterator was reached, or the iterator received an error.
+//
+// Example:
+//	iter := pool.Iter()
+//	defer iter.Close()
+//	for iter.Next() {
+//		fmt.Printf("%v\n", iter.Value())
+//	}
+//	return iter.Err()
+//
+func (iter *Iter) Next() bool {
+	var cEntry *C.char
+	var cNamespace *C.char
+	if cerr := C.rados_nobjects_list_next(iter.ctx, &cEntry, nil, &cNamespace); cerr < 0 {
+		iter.err = getError(cerr)
+		return false
+	}
+	iter.entry = C.GoString(cEntry)
+	iter.namespace = C.GoString(cNamespace)
+	return true
+}
+
+// Value returns the current value of the iterator (object name), after a successful call to Next.
+func (iter *Iter) Value() string {
+	if iter.err != nil {
+		return ""
+	}
+	return iter.entry
+}
+
+// Namespace returns the namespace associated with the current value of the iterator (object name), after a successful call to Next.
+func (iter *Iter) Namespace() string {
+	if iter.err != nil {
+		return ""
+	}
+	return iter.namespace
+}
+
+// Err checks whether the iterator has encountered an error.
+func (iter *Iter) Err() error {
+	if iter.err == ErrNotFound {
+		return nil
+	}
+	return iter.err
+}
+
+// Close the iterator cursor on the server. Be aware that iterators are not closed automatically
+// at the end of iteration.
+func (iter *Iter) Close() {
+	C.rados_nobjects_list_close(iter.ctx)
+}

vendor/github.com/ceph/go-ceph/rados/omap.go

@@ -0,0 +1,205 @@
+package rados
+
+/*
+#cgo LDFLAGS: -lrados
+#include <stdlib.h>
+#include <rados/librados.h>
+*/
+import "C"
+
+import (
+	"runtime"
+	"unsafe"
+)
+
+// OmapKeyValue items are returned by the GetOmapStep's Next call.
+type OmapKeyValue struct {
+	Key   string
+	Value []byte
+}
+
+// GetOmapStep values are used to get the results of an GetOmapValues call
+// on a WriteOp. Until the Operate method of the WriteOp is called the Next
+// call will return an error. After Operate is called, the Next call will
+// return valid results.
+//
+// The life cycle of the GetOmapStep is bound to the ReadOp, if the ReadOp
+// Release method is called the public methods of the step must no longer be
+// used and may return errors.
+type GetOmapStep struct {
+	// C returned data:
+	iter C.rados_omap_iter_t
+	more *C.uchar
+	rval *C.int
+
+	// internal state:
+
+	// canIterate is only set after the operation is performed and is
+	// intended to prevent premature fetching of data
+	canIterate bool
+}
+
+func newGetOmapStep() *GetOmapStep {
+	gos := &GetOmapStep{
+		more: (*C.uchar)(C.malloc(C.sizeof_uchar)),
+		rval: (*C.int)(C.malloc(C.sizeof_int)),
+	}
+	runtime.SetFinalizer(gos, opStepFinalizer)
+	return gos
+}
+
+func (gos *GetOmapStep) free() {
+	gos.canIterate = false
+	if gos.iter != nil {
+		C.rados_omap_get_end(gos.iter)
+	}
+	gos.iter = nil
+	C.free(unsafe.Pointer(gos.more))
+	gos.more = nil
+	C.free(unsafe.Pointer(gos.rval))
+	gos.rval = nil
+}
+
+func (gos *GetOmapStep) update() error {
+	err := getError(*gos.rval)
+	gos.canIterate = (err == nil)
+	return err
+}
+
+// Next returns the next key value pair or nil if iteration is exhausted.
+func (gos *GetOmapStep) Next() (*OmapKeyValue, error) {
+	if !gos.canIterate {
+		return nil, ErrOperationIncomplete
+	}
+	var (
+		cKey    *C.char
+		cVal    *C.char
+		cKeyLen C.size_t
+		cValLen C.size_t
+	)
+	ret := C.rados_omap_get_next2(gos.iter, &cKey, &cVal, &cKeyLen, &cValLen)
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+	if cKey == nil {
+		return nil, nil
+	}
+	return &OmapKeyValue{
+		Key:   string(C.GoBytes(unsafe.Pointer(cKey), C.int(cKeyLen))),
+		Value: C.GoBytes(unsafe.Pointer(cVal), C.int(cValLen)),
+	}, nil
+}
+
+// More returns true if there are more matching keys available.
+func (gos *GetOmapStep) More() bool {
+	// tad bit hacky, but go can't automatically convert from
+	// unsigned char to bool
+	return *gos.more != 0
+}
+
+// SetOmap appends the map `pairs` to the omap `oid`
+func (ioctx *IOContext) SetOmap(oid string, pairs map[string][]byte) error {
+	op := CreateWriteOp()
+	defer op.Release()
+	op.SetOmap(pairs)
+	return op.operateCompat(ioctx, oid)
+}
+
+// OmapListFunc is the type of the function called for each omap key
+// visited by ListOmapValues
+type OmapListFunc func(key string, value []byte)
+
+// ListOmapValues iterates over the keys and values in an omap by way of
+// a callback function.
+//
+// `startAfter`: iterate only on the keys after this specified one
+// `filterPrefix`: iterate only on the keys beginning with this prefix
+// `maxReturn`: iterate no more than `maxReturn` key/value pairs
+// `listFn`: the function called at each iteration
+func (ioctx *IOContext) ListOmapValues(oid string, startAfter string, filterPrefix string, maxReturn int64, listFn OmapListFunc) error {
+
+	op := CreateReadOp()
+	defer op.Release()
+	gos := op.GetOmapValues(startAfter, filterPrefix, uint64(maxReturn))
+	err := op.operateCompat(ioctx, oid)
+	if err != nil {
+		return err
+	}
+
+	for {
+		kv, err := gos.Next()
+		if err != nil {
+			return err
+		}
+		if kv == nil {
+			break
+		}
+		listFn(kv.Key, kv.Value)
+	}
+	return nil
+}
+
+// GetOmapValues fetches a set of keys and their values from an omap and returns then as a map
+// `startAfter`: retrieve only the keys after this specified one
+// `filterPrefix`: retrieve only the keys beginning with this prefix
+// `maxReturn`: retrieve no more than `maxReturn` key/value pairs
+func (ioctx *IOContext) GetOmapValues(oid string, startAfter string, filterPrefix string, maxReturn int64) (map[string][]byte, error) {
+	omap := map[string][]byte{}
+
+	err := ioctx.ListOmapValues(
+		oid, startAfter, filterPrefix, maxReturn,
+		func(key string, value []byte) {
+			omap[key] = value
+		},
+	)
+
+	return omap, err
+}
+
+// GetAllOmapValues fetches all the keys and their values from an omap and returns then as a map
+// `startAfter`: retrieve only the keys after this specified one
+// `filterPrefix`: retrieve only the keys beginning with this prefix
+// `iteratorSize`: internal number of keys to fetch during a read operation
+func (ioctx *IOContext) GetAllOmapValues(oid string, startAfter string, filterPrefix string, iteratorSize int64) (map[string][]byte, error) {
+	omap := map[string][]byte{}
+	omapSize := 0
+
+	for {
+		err := ioctx.ListOmapValues(
+			oid, startAfter, filterPrefix, iteratorSize,
+			func(key string, value []byte) {
+				omap[key] = value
+				startAfter = key
+			},
+		)
+
+		if err != nil {
+			return omap, err
+		}
+
+		// End of omap
+		if len(omap) == omapSize {
+			break
+		}
+
+		omapSize = len(omap)
+	}
+
+	return omap, nil
+}
+
+// RmOmapKeys removes the specified `keys` from the omap `oid`
+func (ioctx *IOContext) RmOmapKeys(oid string, keys []string) error {
+	op := CreateWriteOp()
+	defer op.Release()
+	op.RmOmapKeys(keys)
+	return op.operateCompat(ioctx, oid)
+}
+
+// CleanOmap clears the omap `oid`
+func (ioctx *IOContext) CleanOmap(oid string) error {
+	op := CreateWriteOp()
+	defer op.Release()
+	op.CleanOmap()
+	return op.operateCompat(ioctx, oid)
+}

vendor/github.com/ceph/go-ceph/rados/operation.go

@@ -0,0 +1,154 @@
+package rados
+
+// #include <stdlib.h>
+import "C"
+
+import (
+	"fmt"
+	"strings"
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/log"
+)
+
+// The file operation.go exists to support both read op and write op types that
+// have some pretty common behaviors between them. In C/C++ its assumed that
+// the buffer types and other pointers will not be freed between passing them
+// to the action setup calls (things like rados_write_op_write or
+// rados_read_op_omap_get_vals2) and the call to Operate(...).  Since there's
+// nothing stopping one from sleeping for hours between these calls, or passing
+// the op to other functions and calling Operate there, we want a mechanism
+// that's (fairly) simple to understand and won't run afoul of Go's garbage
+// collection.  That's one reason the operation type tracks the steps (the
+// parts that track complex inputs and outputs) so that as long as the op
+// exists it will have a reference to the step, which will have references
+// to the C language types.
+
+type opKind string
+
+const (
+	readOp  opKind = "read"
+	writeOp opKind = "write"
+)
+
+// OperationError is an error type that may be returned by an Operate call.
+// It captures the error from the operate call itself and any errors from
+// steps that can return an error.
+type OperationError struct {
+	kind       opKind
+	OpError    error
+	StepErrors map[int]error
+}
+
+func (e OperationError) Error() string {
+	subErrors := []string{}
+	if e.OpError != nil {
+		subErrors = append(subErrors,
+			fmt.Sprintf("op=%s", e.OpError))
+	}
+	for idx, es := range e.StepErrors {
+		subErrors = append(subErrors,
+			fmt.Sprintf("Step#%d=%s", idx, es))
+	}
+	return fmt.Sprintf(
+		"%s operation error: %s",
+		e.kind,
+		strings.Join(subErrors, ", "))
+}
+
+// opStep provides an interface for types that are tied to the management of
+// data being input or output from write ops and read ops. The steps are
+// meant to simplify the internals of the ops themselves and be exportable when
+// appropriate. If a step is not being exported it should not be returned
+// from an ops action function. If the step is exported it should be
+// returned from an ops action function.
+//
+// Not all types implementing opStep are expected to need all the functions
+// in the interface. However, for the sake of simplicity on the op side, we use
+// the same interface for all cases and expect those implementing opStep
+// just embed the without* types that provide no-op implementation of
+// functions that make up this interface.
+type opStep interface {
+	// update the state of the step after the call to Operate.
+	// It can be used to convert values from C and cache them and/or
+	// communicate a failure of the action associated with the step.  The
+	// update call will only be made once. Implementations are not required to
+	// handle this call being made more than once.
+	update() error
+	// free will be called to free any resources, especially C memory, that
+	// the step is managing. The behavior of free should be idempotent and
+	// handle being called more than once.
+	free()
+}
+
+// operation represents some of the shared underlying mechanisms for
+// both read and write op types.
+type operation struct {
+	steps []opStep
+}
+
+// free will call the free method of all the steps this operation
+// contains.
+func (o *operation) free() {
+	for i := range o.steps {
+		o.steps[i].free()
+	}
+}
+
+// update the operation and the steps it contains. The top-level result
+// of the rados call is passed in as ret and used to construct errors.
+// The update call of each step is used to update the contents of each
+// step and gather any errors from those steps.
+func (o *operation) update(kind opKind, ret C.int) error {
+	stepErrors := map[int]error{}
+	for i := range o.steps {
+		if err := o.steps[i].update(); err != nil {
+			stepErrors[i] = err
+		}
+	}
+	if ret == 0 && len(stepErrors) == 0 {
+		return nil
+	}
+	return OperationError{
+		kind:       kind,
+		OpError:    getError(ret),
+		StepErrors: stepErrors,
+	}
+}
+
+func opStepFinalizer(s opStep) {
+	if s != nil {
+		log.Warnf("unreachable opStep object found. Cleaning up.")
+		s.free()
+	}
+}
+
+// withoutUpdate can be embedded in a struct to help indicate
+// the type implements the opStep interface but has a no-op
+// update function.
+type withoutUpdate struct{}
+
+func (*withoutUpdate) update() error { return nil }
+
+// withoutFree can be embedded in a struct to help indicate
+// the type implements the opStep interface but has a no-op
+// free function.
+type withoutFree struct{}
+
+func (*withoutFree) free() {}
+
+// withRefs is a embeddable type to help track and free C memory.
+type withRefs struct {
+	refs []unsafe.Pointer
+}
+
+func (w *withRefs) free() {
+	for i := range w.refs {
+		C.free(w.refs[i])
+	}
+	w.refs = nil
+}
+
+func (w *withRefs) add(ptr unsafe.Pointer) {
+	w.refs = append(w.refs, ptr)
+}

vendor/github.com/ceph/go-ceph/rados/operation_flags.go

@@ -0,0 +1,37 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <errno.h>
+// #include <stdlib.h>
+// #include <rados/librados.h>
+//
+import "C"
+
+// OperationFlags control the behavior of read and write operations.
+type OperationFlags int
+
+const (
+	// OperationNoFlag indicates no special behavior is requested.
+	OperationNoFlag = OperationFlags(C.LIBRADOS_OPERATION_NOFLAG)
+	// OperationBalanceReads TODO
+	OperationBalanceReads = OperationFlags(C.LIBRADOS_OPERATION_BALANCE_READS)
+	// OperationLocalizeReads TODO
+	OperationLocalizeReads = OperationFlags(C.LIBRADOS_OPERATION_LOCALIZE_READS)
+	// OperationOrderReadsWrites TODO
+	OperationOrderReadsWrites = OperationFlags(C.LIBRADOS_OPERATION_ORDER_READS_WRITES)
+	// OperationIgnoreCache TODO
+	OperationIgnoreCache = OperationFlags(C.LIBRADOS_OPERATION_IGNORE_CACHE)
+	// OperationSkipRWLocks TODO
+	OperationSkipRWLocks = OperationFlags(C.LIBRADOS_OPERATION_SKIPRWLOCKS)
+	// OperationIgnoreOverlay TODO
+	OperationIgnoreOverlay = OperationFlags(C.LIBRADOS_OPERATION_IGNORE_OVERLAY)
+	// OperationFullTry send request to a full cluster or pool, ops such as delete
+	// can succeed while other ops will return out-of-space errors.
+	OperationFullTry = OperationFlags(C.LIBRADOS_OPERATION_FULL_TRY)
+	// OperationFullForce TODO
+	OperationFullForce = OperationFlags(C.LIBRADOS_OPERATION_FULL_FORCE)
+	// OperationIgnoreRedirect TODO
+	OperationIgnoreRedirect = OperationFlags(C.LIBRADOS_OPERATION_IGNORE_REDIRECT)
+	// OperationOrderSnap TODO
+	OperationOrderSnap = OperationFlags(C.LIBRADOS_OPERATION_ORDERSNAP)
+)

vendor/github.com/ceph/go-ceph/rados/rados.go

@@ -0,0 +1,130 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <errno.h>
+// #include <stdlib.h>
+// #include <rados/librados.h>
+import "C"
+
+import (
+	"runtime"
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/log"
+)
+
+const (
+	// AllNamespaces is used to reset a selected namespace to all
+	// namespaces. See the IOContext SetNamespace function.
+	AllNamespaces = C.LIBRADOS_ALL_NSPACES
+
+	// FIXME: for backwards compatibility
+
+	// RadosAllNamespaces is used to reset a selected namespace to all
+	// namespaces. See the IOContext SetNamespace function.
+	//
+	// Deprecated: use AllNamespaces instead
+	RadosAllNamespaces = AllNamespaces
+)
+
+// OpFlags are flags that can be set on a per-op basis.
+type OpFlags uint
+
+const (
+	// OpFlagNone can be use to not set any flags.
+	OpFlagNone = OpFlags(0)
+	// OpFlagExcl marks an op to fail a create operation if the object
+	// already exists.
+	OpFlagExcl = OpFlags(C.LIBRADOS_OP_FLAG_EXCL)
+	// OpFlagFailOk allows the transaction to succeed even if the flagged
+	// op fails.
+	OpFlagFailOk = OpFlags(C.LIBRADOS_OP_FLAG_FAILOK)
+	// OpFlagFAdviseRandom indicates read/write op random.
+	OpFlagFAdviseRandom = OpFlags(C.LIBRADOS_OP_FLAG_FADVISE_RANDOM)
+	// OpFlagFAdviseSequential indicates read/write op sequential.
+	OpFlagFAdviseSequential = OpFlags(C.LIBRADOS_OP_FLAG_FADVISE_SEQUENTIAL)
+	// OpFlagFAdviseWillNeed indicates read/write data will be accessed in
+	// the near future (by someone).
+	OpFlagFAdviseWillNeed = OpFlags(C.LIBRADOS_OP_FLAG_FADVISE_WILLNEED)
+	// OpFlagFAdviseDontNeed indicates read/write data will not accessed in
+	// the near future (by anyone).
+	OpFlagFAdviseDontNeed = OpFlags(C.LIBRADOS_OP_FLAG_FADVISE_DONTNEED)
+	// OpFlagFAdviseNoCache indicates read/write data will not accessed
+	// again (by *this* client).
+	OpFlagFAdviseNoCache = OpFlags(C.LIBRADOS_OP_FLAG_FADVISE_NOCACHE)
+)
+
+// Version returns the major, minor, and patch components of the version of
+// the RADOS library linked against.
+func Version() (int, int, int) {
+	var cMajor, cMinor, cPatch C.int
+	C.rados_version(&cMajor, &cMinor, &cPatch)
+	return int(cMajor), int(cMinor), int(cPatch)
+}
+
+func makeConn() *Conn {
+	return &Conn{connected: false}
+}
+
+func newConn(user *C.char) (*Conn, error) {
+	conn := makeConn()
+	ret := C.rados_create(&conn.cluster, user)
+
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+
+	runtime.SetFinalizer(conn, freeConn)
+	return conn, nil
+}
+
+// NewConn creates a new connection object. It returns the connection and an
+// error, if any.
+func NewConn() (*Conn, error) {
+	return newConn(nil)
+}
+
+// NewConnWithUser creates a new connection object with a custom username.
+// It returns the connection and an error, if any.
+func NewConnWithUser(user string) (*Conn, error) {
+	cUser := C.CString(user)
+	defer C.free(unsafe.Pointer(cUser))
+	return newConn(cUser)
+}
+
+// NewConnWithClusterAndUser creates a new connection object for a specific cluster and username.
+// It returns the connection and an error, if any.
+func NewConnWithClusterAndUser(clusterName string, userName string) (*Conn, error) {
+	cClusterName := C.CString(clusterName)
+	defer C.free(unsafe.Pointer(cClusterName))
+
+	cName := C.CString(userName)
+	defer C.free(unsafe.Pointer(cName))
+
+	conn := makeConn()
+	ret := C.rados_create2(&conn.cluster, cClusterName, cName, 0)
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+
+	runtime.SetFinalizer(conn, freeConn)
+	return conn, nil
+}
+
+// freeConn releases resources that are allocated while configuring the
+// connection to the cluster. rados_shutdown() should only be needed after a
+// successful call to rados_connect(), however if the connection has been
+// configured with non-default parameters, some of the parameters may be
+// allocated before connecting. rados_shutdown() will free the allocated
+// resources, even if there has not been a connection yet.
+//
+// This function is setup as a destructor/finalizer when rados_create() is
+// called.
+func freeConn(conn *Conn) {
+	if conn.cluster != nil {
+		log.Warnf("unreachable Conn object has not been shut down. Cleaning up.")
+		C.rados_shutdown(conn.cluster)
+		// prevent calling rados_shutdown() more than once
+		conn.cluster = nil
+	}
+}

vendor/github.com/ceph/go-ceph/rados/rados_nautilus.go

@@ -0,0 +1,13 @@
+//go:build !mimic
+// +build !mimic
+
+package rados
+
+// #include <rados/librados.h>
+import "C"
+
+const (
+	// OpFlagFAdviseFUA optionally support FUA (force unit access) on write
+	// requests.
+	OpFlagFAdviseFUA = OpFlags(C.LIBRADOS_OP_FLAG_FADVISE_FUA)
+)

vendor/github.com/ceph/go-ceph/rados/rados_read_op_assert_version.go

@@ -0,0 +1,18 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+// AssertVersion ensures that the object exists and that its internal version
+// number is equal to "ver" before reading. "ver" should be a version number
+// previously obtained with IOContext.GetLastVersion().
+//
+// Implements:
+//  void rados_read_op_assert_version(rados_read_op_t read_op,
+//                                    uint64_t ver)
+func (r *ReadOp) AssertVersion(ver uint64) {
+	C.rados_read_op_assert_version(r.op, C.uint64_t(ver))
+}

vendor/github.com/ceph/go-ceph/rados/rados_set_locator.go

@@ -0,0 +1,27 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+import (
+	"unsafe"
+)
+
+// SetLocator sets the key for mapping objects to pgs within an io context.
+// Until a different locator key is set, all objects in this io context will be placed in the same pg.
+// To reset the locator, an empty string must be set.
+//
+// Implements:
+//  void rados_ioctx_locator_set_key(rados_ioctx_t io, const char *key);
+func (ioctx *IOContext) SetLocator(locator string) {
+	if locator == "" {
+		C.rados_ioctx_locator_set_key(ioctx.ioctx, nil)
+	} else {
+		var cLoc *C.char = C.CString(locator)
+		defer C.free(unsafe.Pointer(cLoc))
+		C.rados_ioctx_locator_set_key(ioctx.ioctx, cLoc)
+	}
+}

vendor/github.com/ceph/go-ceph/rados/rados_write_op_assert_version.go

@@ -0,0 +1,18 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+// AssertVersion ensures that the object exists and that its internal version
+// number is equal to "ver" before writing. "ver" should be a version number
+// previously obtained with IOContext.GetLastVersion().
+//
+// Implements:
+//  void rados_read_op_assert_version(rados_read_op_t read_op,
+//                                    uint64_t ver)
+func (w *WriteOp) AssertVersion(ver uint64) {
+	C.rados_write_op_assert_version(w.op, C.uint64_t(ver))
+}

vendor/github.com/ceph/go-ceph/rados/rados_write_op_remove.go

@@ -0,0 +1,15 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+// Remove object.
+//
+// Implements:
+//  void rados_write_op_remove(rados_write_op_t write_op)
+func (w *WriteOp) Remove() {
+	C.rados_write_op_remove(w.op)
+}

vendor/github.com/ceph/go-ceph/rados/rados_write_op_setxattr.go

@@ -0,0 +1,30 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+import (
+	"unsafe"
+)
+
+// SetXattr sets an xattr.
+//
+// Implements:
+//  void rados_write_op_setxattr(rados_write_op_t write_op,
+//                               const char * name,
+//                               const char * value,
+//                               size_t value_len)
+func (w *WriteOp) SetXattr(name string, value []byte) {
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	C.rados_write_op_setxattr(
+		w.op,
+		cName,
+		(*C.char)(unsafe.Pointer(&value[0])),
+		C.size_t(len(value)),
+	)
+}

vendor/github.com/ceph/go-ceph/rados/read_op.go

@@ -0,0 +1,90 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <errno.h>
+// #include <stdlib.h>
+// #include <rados/librados.h>
+//
+import "C"
+
+import (
+	"unsafe"
+)
+
+// ReadOp manages a set of discrete object read actions that will be performed
+// together atomically.
+type ReadOp struct {
+	operation
+	op C.rados_read_op_t
+}
+
+// CreateReadOp returns a newly constructed read operation.
+func CreateReadOp() *ReadOp {
+	return &ReadOp{
+		op: C.rados_create_read_op(),
+	}
+}
+
+// Release the resources associated with this read operation.
+func (r *ReadOp) Release() {
+	C.rados_release_read_op(r.op)
+	r.op = nil
+	r.free()
+}
+
+// Operate will perform the operation(s).
+func (r *ReadOp) Operate(ioctx *IOContext, oid string, flags OperationFlags) error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+
+	cOid := C.CString(oid)
+	defer C.free(unsafe.Pointer(cOid))
+
+	ret := C.rados_read_op_operate(r.op, ioctx.ioctx, cOid, C.int(flags))
+	return r.update(readOp, ret)
+}
+
+func (r *ReadOp) operateCompat(ioctx *IOContext, oid string) error {
+	switch err := r.Operate(ioctx, oid, OperationNoFlag).(type) {
+	case nil:
+		return nil
+	case OperationError:
+		return err.OpError
+	default:
+		return err
+	}
+}
+
+// AssertExists assures the object targeted by the read op exists.
+//
+// Implements:
+//  void rados_read_op_assert_exists(rados_read_op_t read_op);
+func (r *ReadOp) AssertExists() {
+	C.rados_read_op_assert_exists(r.op)
+}
+
+// GetOmapValues is used to iterate over a set, or sub-set, of omap keys
+// as part of a read operation. An GetOmapStep is returned from this
+// function. The GetOmapStep may be used to iterate over the key-value
+// pairs after the Operate call has been performed.
+func (r *ReadOp) GetOmapValues(startAfter, filterPrefix string, maxReturn uint64) *GetOmapStep {
+	gos := newGetOmapStep()
+	r.steps = append(r.steps, gos)
+
+	cStartAfter := C.CString(startAfter)
+	cFilterPrefix := C.CString(filterPrefix)
+	defer C.free(unsafe.Pointer(cStartAfter))
+	defer C.free(unsafe.Pointer(cFilterPrefix))
+
+	C.rados_read_op_omap_get_vals2(
+		r.op,
+		cStartAfter,
+		cFilterPrefix,
+		C.uint64_t(maxReturn),
+		&gos.iter,
+		gos.more,
+		gos.rval,
+	)
+	return gos
+}

vendor/github.com/ceph/go-ceph/rados/read_op_omap_get_vals_by_keys.go

@@ -0,0 +1,112 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+)
+
+// ReadOpOmapGetValsByKeysStep holds the result of the
+// GetOmapValuesByKeys read operation.
+// Result is valid only after Operate() was called.
+type ReadOpOmapGetValsByKeysStep struct {
+	// C arguments
+
+	iter  C.rados_omap_iter_t
+	prval *C.int
+
+	// Internal state
+
+	// canIterate is only set after the operation is performed and is
+	// intended to prevent premature fetching of data.
+	canIterate bool
+}
+
+func newReadOpOmapGetValsByKeysStep() *ReadOpOmapGetValsByKeysStep {
+	s := &ReadOpOmapGetValsByKeysStep{
+		prval: (*C.int)(C.malloc(C.sizeof_int)),
+	}
+
+	return s
+}
+
+func (s *ReadOpOmapGetValsByKeysStep) free() {
+	s.canIterate = false
+	C.rados_omap_get_end(s.iter)
+
+	C.free(unsafe.Pointer(s.prval))
+	s.prval = nil
+}
+
+func (s *ReadOpOmapGetValsByKeysStep) update() error {
+	err := getError(*s.prval)
+	s.canIterate = (err == nil)
+
+	return err
+}
+
+// Next gets the next omap key/value pair referenced by
+// ReadOpOmapGetValsByKeysStep's internal iterator.
+// If there are no more elements to retrieve, (nil, nil) is returned.
+// May be called only after Operate() finished.
+func (s *ReadOpOmapGetValsByKeysStep) Next() (*OmapKeyValue, error) {
+	if !s.canIterate {
+		return nil, ErrOperationIncomplete
+	}
+
+	var (
+		cKey    *C.char
+		cVal    *C.char
+		cKeyLen C.size_t
+		cValLen C.size_t
+	)
+
+	ret := C.rados_omap_get_next2(s.iter, &cKey, &cVal, &cKeyLen, &cValLen)
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+
+	if cKey == nil {
+		// Iterator has reached the end of the list.
+		return nil, nil
+	}
+
+	return &OmapKeyValue{
+		Key:   string(C.GoBytes(unsafe.Pointer(cKey), C.int(cKeyLen))),
+		Value: C.GoBytes(unsafe.Pointer(cVal), C.int(cValLen)),
+	}, nil
+}
+
+// GetOmapValuesByKeys starts iterating over specific key/value pairs.
+//
+// Implements:
+//  void rados_read_op_omap_get_vals_by_keys2(rados_read_op_t read_op,
+//                                            char const * const * keys,
+//                                            size_t num_keys,
+//                                            const size_t * key_lens,
+//                                            rados_omap_iter_t * iter,
+//                                            int * prval)
+func (r *ReadOp) GetOmapValuesByKeys(keys []string) *ReadOpOmapGetValsByKeysStep {
+	s := newReadOpOmapGetValsByKeysStep()
+	r.steps = append(r.steps, s)
+
+	cKeys := cutil.NewBufferGroupStrings(keys)
+	defer cKeys.Free()
+
+	C.rados_read_op_omap_get_vals_by_keys2(
+		r.op,
+		(**C.char)(cKeys.BuffersPtr()),
+		C.size_t(len(keys)),
+		(*C.size_t)(cKeys.LengthsPtr()),
+		&s.iter,
+		s.prval,
+	)
+
+	return s
+}

vendor/github.com/ceph/go-ceph/rados/read_op_read.go

@@ -0,0 +1,71 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+import (
+	"unsafe"
+)
+
+// ReadOpReadStep holds the result of the Read read operation.
+// Result is valid only after Operate() was called.
+type ReadOpReadStep struct {
+	// C returned data:
+	bytesRead *C.size_t
+	prval     *C.int
+
+	BytesRead int64 // Bytes read by this action.
+	Result    int   // Result of this action.
+}
+
+func (s *ReadOpReadStep) update() error {
+	s.BytesRead = (int64)(*s.bytesRead)
+	s.Result = (int)(*s.prval)
+
+	return nil
+}
+
+func (s *ReadOpReadStep) free() {
+	C.free(unsafe.Pointer(s.bytesRead))
+	C.free(unsafe.Pointer(s.prval))
+
+	s.bytesRead = nil
+	s.prval = nil
+}
+
+func newReadOpReadStep() *ReadOpReadStep {
+	return &ReadOpReadStep{
+		bytesRead: (*C.size_t)(C.malloc(C.sizeof_size_t)),
+		prval:     (*C.int)(C.malloc(C.sizeof_int)),
+	}
+}
+
+// Read bytes from offset into buffer.
+// len(buffer) is the maximum number of bytes read from the object.
+// buffer[:ReadOpReadStep.BytesRead] then contains object data.
+//
+// Implements:
+//  void rados_read_op_read(rados_read_op_t read_op,
+//                          uint64_t offset,
+//                          size_t len,
+//                          char * buffer,
+//                          size_t * bytes_read,
+//                          int * prval)
+func (r *ReadOp) Read(offset uint64, buffer []byte) *ReadOpReadStep {
+	oe := newReadStep(buffer, offset)
+	readStep := newReadOpReadStep()
+	r.steps = append(r.steps, oe, readStep)
+	C.rados_read_op_read(
+		r.op,
+		oe.cOffset,
+		oe.cReadLen,
+		oe.cBuffer,
+		readStep.bytesRead,
+		readStep.prval,
+	)
+
+	return readStep
+}

vendor/github.com/ceph/go-ceph/rados/read_step.go

@@ -0,0 +1,31 @@
+package rados
+
+// #include <stdint.h>
+import "C"
+
+import (
+	"unsafe"
+)
+
+type readStep struct {
+	withoutUpdate
+	withoutFree
+	// the c pointer utilizes the Go byteslice data and no free is needed
+
+	// inputs:
+	b []byte
+
+	// arguments:
+	cBuffer  *C.char
+	cReadLen C.size_t
+	cOffset  C.uint64_t
+}
+
+func newReadStep(b []byte, offset uint64) *readStep {
+	return &readStep{
+		b:        b,
+		cBuffer:  (*C.char)(unsafe.Pointer(&b[0])), // TODO: must be pinned
+		cReadLen: C.size_t(len(b)),
+		cOffset:  C.uint64_t(offset),
+	}
+}

vendor/github.com/ceph/go-ceph/rados/snapshot.go

@@ -0,0 +1,189 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <stdlib.h>
+// #include <rados/librados.h>
+import "C"
+
+import (
+	"time"
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/retry"
+)
+
+// CreateSnap creates a pool-wide snapshot.
+//
+// Implements:
+// int rados_ioctx_snap_create(rados_ioctx_t io, const char *snapname)
+func (ioctx *IOContext) CreateSnap(snapName string) error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+
+	cSnapName := C.CString(snapName)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ret := C.rados_ioctx_snap_create(ioctx.ioctx, cSnapName)
+	return getError(ret)
+}
+
+// RemoveSnap deletes the pool snapshot.
+//
+// Implements:
+//  int rados_ioctx_snap_remove(rados_ioctx_t io, const char *snapname)
+func (ioctx *IOContext) RemoveSnap(snapName string) error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+
+	cSnapName := C.CString(snapName)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ret := C.rados_ioctx_snap_remove(ioctx.ioctx, cSnapName)
+	return getError(ret)
+}
+
+// SnapID represents the ID of a rados snapshot.
+type SnapID C.rados_snap_t
+
+// LookupSnap returns the ID of a pool snapshot.
+//
+// Implements:
+//  int rados_ioctx_snap_lookup(rados_ioctx_t io, const char *name, rados_snap_t *id)
+func (ioctx *IOContext) LookupSnap(snapName string) (SnapID, error) {
+	var snapID SnapID
+
+	if err := ioctx.validate(); err != nil {
+		return snapID, err
+	}
+
+	cSnapName := C.CString(snapName)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ret := C.rados_ioctx_snap_lookup(
+		ioctx.ioctx,
+		cSnapName,
+		(*C.rados_snap_t)(&snapID))
+	return snapID, getError(ret)
+}
+
+// GetSnapName returns the name of a pool snapshot with the given snapshot ID.
+//
+// Implements:
+//  int rados_ioctx_snap_get_name(rados_ioctx_t io, rados_snap_t id, char *name, int maxlen)
+func (ioctx *IOContext) GetSnapName(snapID SnapID) (string, error) {
+	if err := ioctx.validate(); err != nil {
+		return "", err
+	}
+
+	var (
+		buf []byte
+		err error
+	)
+	// range from 1k to 64KiB
+	retry.WithSizes(1024, 1<<16, func(len int) retry.Hint {
+		cLen := C.int(len)
+		buf = make([]byte, cLen)
+		ret := C.rados_ioctx_snap_get_name(
+			ioctx.ioctx,
+			(C.rados_snap_t)(snapID),
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			cLen)
+		err = getError(ret)
+		return retry.Size(int(cLen)).If(err == errRange)
+	})
+
+	if err != nil {
+		return "", err
+	}
+	return C.GoString((*C.char)(unsafe.Pointer(&buf[0]))), nil
+}
+
+// GetSnapStamp returns the time of the pool snapshot creation.
+//
+// Implements:
+//  int rados_ioctx_snap_get_stamp(rados_ioctx_t io, rados_snap_t id, time_t *t)
+func (ioctx *IOContext) GetSnapStamp(snapID SnapID) (time.Time, error) {
+	var cTime C.time_t
+
+	if err := ioctx.validate(); err != nil {
+		return time.Unix(int64(cTime), 0), err
+	}
+
+	ret := C.rados_ioctx_snap_get_stamp(
+		ioctx.ioctx,
+		(C.rados_snap_t)(snapID),
+		&cTime)
+	return time.Unix(int64(cTime), 0), getError(ret)
+}
+
+// ListSnaps returns a slice containing the SnapIDs of existing pool snapshots.
+//
+// Implements:
+//  int rados_ioctx_snap_list(rados_ioctx_t io, rados_snap_t *snaps, int maxlen)
+func (ioctx *IOContext) ListSnaps() ([]SnapID, error) {
+	if err := ioctx.validate(); err != nil {
+		return nil, err
+	}
+
+	var (
+		snapList []SnapID
+		cLen     C.int
+		err      error
+		ret      C.int
+	)
+	retry.WithSizes(100, 1000, func(maxlen int) retry.Hint {
+		cLen = C.int(maxlen)
+		snapList = make([]SnapID, cLen)
+		ret = C.rados_ioctx_snap_list(
+			ioctx.ioctx,
+			(*C.rados_snap_t)(unsafe.Pointer(&snapList[0])),
+			cLen)
+		err = getErrorIfNegative(ret)
+		return retry.Size(int(cLen)).If(err == errRange)
+	})
+
+	if err != nil {
+		return nil, err
+	}
+	return snapList[:ret], nil
+}
+
+// RollbackSnap rollbacks the object with key oID to the pool snapshot.
+// The contents of the object will be the same as when the snapshot was taken.
+//
+// Implements:
+//  int rados_ioctx_snap_rollback(rados_ioctx_t io, const char *oid, const char *snapname);
+func (ioctx *IOContext) RollbackSnap(oid, snapName string) error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+
+	coid := C.CString(oid)
+	defer C.free(unsafe.Pointer(coid))
+	cSnapName := C.CString(snapName)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ret := C.rados_ioctx_snap_rollback(ioctx.ioctx, coid, cSnapName)
+	return getError(ret)
+}
+
+// SnapHead is the representation of LIBRADOS_SNAP_HEAD from librados.
+// SnapHead can be used to reset the IOContext to stop reading from a snapshot.
+const SnapHead = SnapID(C.LIBRADOS_SNAP_HEAD)
+
+// SetReadSnap sets the snapshot from which reads are performed.
+// Subsequent reads will return data as it was at the time of that snapshot.
+// Pass SnapHead for no snapshot (i.e. normal operation).
+//
+// Implements:
+//  void rados_ioctx_snap_set_read(rados_ioctx_t io, rados_snap_t snap);
+func (ioctx *IOContext) SetReadSnap(snapID SnapID) error {
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+
+	C.rados_ioctx_snap_set_read(ioctx.ioctx, (C.rados_snap_t)(snapID))
+	return nil
+}

vendor/github.com/ceph/go-ceph/rados/watcher.go

@@ -0,0 +1,367 @@
+package rados
+
+/*
+#cgo LDFLAGS: -lrados
+#include <stdlib.h>
+#include <rados/librados.h>
+extern void watchNotifyCb(void*, uint64_t, uint64_t, uint64_t, void*, size_t);
+extern void watchErrorCb(void*, uint64_t, int);
+*/
+import "C"
+
+import (
+	"encoding/binary"
+	"fmt"
+	"math"
+	"sync"
+	"time"
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/log"
+)
+
+type (
+	// WatcherID is the unique id of a Watcher.
+	WatcherID uint64
+	// NotifyID is the unique id of a NotifyEvent.
+	NotifyID uint64
+	// NotifierID is the unique id of a notifying client.
+	NotifierID uint64
+)
+
+// NotifyEvent is received by a watcher for each notification.
+type NotifyEvent struct {
+	ID         NotifyID
+	WatcherID  WatcherID
+	NotifierID NotifierID
+	Data       []byte
+}
+
+// NotifyAck represents an acknowleged notification.
+type NotifyAck struct {
+	WatcherID  WatcherID
+	NotifierID NotifierID
+	Response   []byte
+}
+
+// NotifyTimeout represents an unacknowleged notification.
+type NotifyTimeout struct {
+	WatcherID  WatcherID
+	NotifierID NotifierID
+}
+
+// Watcher receives all notifications for certain object.
+type Watcher struct {
+	id     WatcherID
+	oid    string
+	ioctx  *IOContext
+	events chan NotifyEvent
+	errors chan error
+	done   chan struct{}
+}
+
+var (
+	watchers    = map[WatcherID]*Watcher{}
+	watchersMtx sync.RWMutex
+)
+
+// Watch creates a Watcher for the specified object.
+//
+// A Watcher receives all notifications that are sent to the object on which it
+// has been created. It exposes two read-only channels: Events() receives all
+// the NotifyEvents and Errors() receives all occuring errors. A typical code
+// creating a Watcher could look like this:
+//
+//  watcher, err := ioctx.Watch(oid)
+//  go func() { // event handler
+//    for ne := range watcher.Events() {
+//      ...
+//      ne.Ack([]byte("response data..."))
+//      ...
+//    }
+//  }()
+//  go func() { // error handler
+//    for err := range watcher.Errors() {
+//      ... handle err ...
+//    }
+//  }()
+//
+// CAUTION: the Watcher references the IOContext in which it has been created.
+// Therefore all watchers must be deleted with the Delete() method before the
+// IOContext is being destroyed.
+//
+// Implements:
+//  int rados_watch2(rados_ioctx_t io, const char* o, uint64_t* cookie,
+//    rados_watchcb2_t watchcb, rados_watcherrcb_t watcherrcb, void* arg)
+func (ioctx *IOContext) Watch(obj string) (*Watcher, error) {
+	return ioctx.WatchWithTimeout(obj, 0)
+}
+
+// WatchWithTimeout creates a watcher on an object. Same as Watcher(), but
+// different timeout than the default can be specified.
+//
+// Implements:
+//  int rados_watch3(rados_ioctx_t io, const char *o, uint64_t *cookie,
+// 	  rados_watchcb2_t watchcb, rados_watcherrcb_t watcherrcb, uint32_t timeout,
+// 	  void *arg);
+func (ioctx *IOContext) WatchWithTimeout(oid string, timeout time.Duration) (*Watcher, error) {
+	cObj := C.CString(oid)
+	defer C.free(unsafe.Pointer(cObj))
+	var id C.uint64_t
+	watchersMtx.Lock()
+	defer watchersMtx.Unlock()
+	ret := C.rados_watch3(
+		ioctx.ioctx,
+		cObj,
+		&id,
+		(C.rados_watchcb2_t)(C.watchNotifyCb),
+		(C.rados_watcherrcb_t)(C.watchErrorCb),
+		C.uint32_t(timeout.Milliseconds()/1000),
+		nil,
+	)
+	if err := getError(ret); err != nil {
+		return nil, err
+	}
+	evCh := make(chan NotifyEvent)
+	errCh := make(chan error)
+	w := &Watcher{
+		id:     WatcherID(id),
+		ioctx:  ioctx,
+		oid:    oid,
+		events: evCh,
+		errors: errCh,
+		done:   make(chan struct{}),
+	}
+	watchers[WatcherID(id)] = w
+	return w, nil
+}
+
+// ID returns the WatcherId of the Watcher
+func (w *Watcher) ID() WatcherID {
+	return w.id
+}
+
+// Events returns a read-only channel, that receives all notifications that are
+// sent to the object of the Watcher.
+func (w *Watcher) Events() <-chan NotifyEvent {
+	return w.events
+}
+
+// Errors returns a read-only channel, that receives all errors for the Watcher.
+func (w *Watcher) Errors() <-chan error {
+	return w.errors
+}
+
+// Check on the status of a Watcher.
+//
+// Returns the time since it was last confirmed. If there is an error, the
+// Watcher is no longer valid, and should be destroyed with the Delete() method.
+//
+// Implements:
+//  int rados_watch_check(rados_ioctx_t io, uint64_t cookie)
+func (w *Watcher) Check() (time.Duration, error) {
+	ret := C.rados_watch_check(w.ioctx.ioctx, C.uint64_t(w.id))
+	if ret < 0 {
+		return 0, getError(ret)
+	}
+	return time.Millisecond * time.Duration(ret), nil
+}
+
+// Delete the watcher. This closes both the event and error channel.
+//
+// Implements:
+//  int rados_unwatch2(rados_ioctx_t io, uint64_t cookie)
+func (w *Watcher) Delete() error {
+	watchersMtx.Lock()
+	_, ok := watchers[w.id]
+	if ok {
+		delete(watchers, w.id)
+	}
+	watchersMtx.Unlock()
+	if !ok {
+		return nil
+	}
+	ret := C.rados_unwatch2(w.ioctx.ioctx, C.uint64_t(w.id))
+	if ret != 0 {
+		return getError(ret)
+	}
+	close(w.done) // unblock blocked callbacks
+	close(w.events)
+	close(w.errors)
+	return nil
+}
+
+// Notify sends a notification with the provided data to all Watchers of the
+// specified object.
+//
+// CAUTION: even if the error is not nil. the returned slices
+// might still contain data.
+func (ioctx *IOContext) Notify(obj string, data []byte) ([]NotifyAck, []NotifyTimeout, error) {
+	return ioctx.NotifyWithTimeout(obj, data, 0)
+}
+
+// NotifyWithTimeout is like Notify() but with a different timeout than the
+// default.
+//
+// Implements:
+//  int rados_notify2(rados_ioctx_t io, const char* o, const char* buf, int buf_len,
+//    uint64_t timeout_ms, char** reply_buffer, size_t* reply_buffer_len)
+func (ioctx *IOContext) NotifyWithTimeout(obj string, data []byte, timeout time.Duration) ([]NotifyAck,
+	[]NotifyTimeout, error) {
+	cObj := C.CString(obj)
+	defer C.free(unsafe.Pointer(cObj))
+	var cResponse *C.char
+	defer C.rados_buffer_free(cResponse)
+	var responseLen C.size_t
+	var dataPtr *C.char
+	if len(data) > 0 {
+		dataPtr = (*C.char)(unsafe.Pointer(&data[0]))
+	}
+	ret := C.rados_notify2(
+		ioctx.ioctx,
+		cObj,
+		dataPtr,
+		C.int(len(data)),
+		C.uint64_t(timeout.Milliseconds()),
+		&cResponse,
+		&responseLen,
+	)
+	// cResponse has been set even if an error is returned, so we decode it anyway
+	acks, timeouts := decodeNotifyResponse(cResponse, responseLen)
+	return acks, timeouts, getError(ret)
+}
+
+// Ack sends an acknowledgement with the specified response data to the notfier
+// of the NotifyEvent. If a notify is not ack'ed, the originating Notify() call
+// blocks and eventiually times out.
+//
+// Implements:
+//  int rados_notify_ack(rados_ioctx_t io, const char *o, uint64_t notify_id,
+//    uint64_t cookie, const char *buf, int buf_len)
+func (ne *NotifyEvent) Ack(response []byte) error {
+	watchersMtx.RLock()
+	w, ok := watchers[ne.WatcherID]
+	watchersMtx.RUnlock()
+	if !ok {
+		return fmt.Errorf("can't ack on deleted watcher %v", ne.WatcherID)
+	}
+	cOID := C.CString(w.oid)
+	defer C.free(unsafe.Pointer(cOID))
+	var respPtr *C.char
+	if len(response) > 0 {
+		respPtr = (*C.char)(unsafe.Pointer(&response[0]))
+	}
+	ret := C.rados_notify_ack(
+		w.ioctx.ioctx,
+		cOID,
+		C.uint64_t(ne.ID),
+		C.uint64_t(ne.WatcherID),
+		respPtr,
+		C.int(len(response)),
+	)
+	return getError(ret)
+}
+
+// WatcherFlush flushes all pending notifications of the cluster.
+//
+// Implements:
+//  int rados_watch_flush(rados_t cluster)
+func (c *Conn) WatcherFlush() error {
+	if !c.connected {
+		return ErrNotConnected
+	}
+	ret := C.rados_watch_flush(c.cluster)
+	return getError(ret)
+}
+
+// decoder for this notify response format:
+//    le32 num_acks
+//    {
+//      le64 gid     global id for the client (for client.1234 that's 1234)
+//      le64 cookie  cookie for the client
+//      le32 buflen  length of reply message buffer
+//      u8 buflen  payload
+//    } num_acks
+//    le32 num_timeouts
+//    {
+//      le64 gid     global id for the client
+//      le64 cookie  cookie for the client
+//    } num_timeouts
+//
+// NOTE: starting with pacific this is implemented as a C function and this can
+// be replaced later
+func decodeNotifyResponse(response *C.char, len C.size_t) ([]NotifyAck, []NotifyTimeout) {
+	if len == 0 || response == nil {
+		return nil, nil
+	}
+	b := (*[math.MaxInt32]byte)(unsafe.Pointer(response))[:len:len]
+	pos := 0
+
+	num := binary.LittleEndian.Uint32(b[pos:])
+	pos += 4
+	acks := make([]NotifyAck, num)
+	for i := range acks {
+		acks[i].NotifierID = NotifierID(binary.LittleEndian.Uint64(b[pos:]))
+		pos += 8
+		acks[i].WatcherID = WatcherID(binary.LittleEndian.Uint64(b[pos:]))
+		pos += 8
+		dataLen := binary.LittleEndian.Uint32(b[pos:])
+		pos += 4
+		if dataLen > 0 {
+			acks[i].Response = C.GoBytes(unsafe.Pointer(&b[pos]), C.int(dataLen))
+			pos += int(dataLen)
+		}
+	}
+
+	num = binary.LittleEndian.Uint32(b[pos:])
+	pos += 4
+	timeouts := make([]NotifyTimeout, num)
+	for i := range timeouts {
+		timeouts[i].NotifierID = NotifierID(binary.LittleEndian.Uint64(b[pos:]))
+		pos += 8
+		timeouts[i].WatcherID = WatcherID(binary.LittleEndian.Uint64(b[pos:]))
+		pos += 8
+	}
+	return acks, timeouts
+}
+
+//export watchNotifyCb
+func watchNotifyCb(_ unsafe.Pointer, notifyID C.uint64_t, id C.uint64_t,
+	notifierID C.uint64_t, cData unsafe.Pointer, dataLen C.size_t) {
+	ev := NotifyEvent{
+		ID:         NotifyID(notifyID),
+		WatcherID:  WatcherID(id),
+		NotifierID: NotifierID(notifierID),
+	}
+	if dataLen > 0 {
+		ev.Data = C.GoBytes(cData, C.int(dataLen))
+	}
+	watchersMtx.RLock()
+	w, ok := watchers[WatcherID(id)]
+	watchersMtx.RUnlock()
+	if !ok {
+		// usually this should not happen, but who knows
+		log.Warnf("received notification for unknown watcher ID: %#v", ev)
+		return
+	}
+	select {
+	case <-w.done: // unblock when deleted
+	case w.events <- ev:
+	}
+}
+
+//export watchErrorCb
+func watchErrorCb(_ unsafe.Pointer, id C.uint64_t, err C.int) {
+	watchersMtx.RLock()
+	w, ok := watchers[WatcherID(id)]
+	watchersMtx.RUnlock()
+	if !ok {
+		// usually this should not happen, but who knows
+		log.Warnf("received error for unknown watcher ID: id=%d err=%#v", id, err)
+		return
+	}
+	select {
+	case <-w.done: // unblock when deleted
+	case w.errors <- getError(err):
+	}
+}

vendor/github.com/ceph/go-ceph/rados/write_op.go

@@ -0,0 +1,195 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <errno.h>
+// #include <stdlib.h>
+// #include <rados/librados.h>
+//
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+	ts "github.com/ceph/go-ceph/internal/timespec"
+)
+
+// Timespec is a public type for the internal C 'struct timespec'
+type Timespec ts.Timespec
+
+// WriteOp manages a set of discrete actions that will be performed together
+// atomically.
+type WriteOp struct {
+	operation
+	op C.rados_write_op_t
+}
+
+// CreateWriteOp returns a newly constructed write operation.
+func CreateWriteOp() *WriteOp {
+	return &WriteOp{
+		op: C.rados_create_write_op(),
+	}
+}
+
+// Release the resources associated with this write operation.
+func (w *WriteOp) Release() {
+	C.rados_release_write_op(w.op)
+	w.op = nil
+	w.free()
+}
+
+func (w WriteOp) operate2(
+	ioctx *IOContext, oid string, mtime *Timespec, flags OperationFlags) error {
+
+	if err := ioctx.validate(); err != nil {
+		return err
+	}
+
+	cOid := C.CString(oid)
+	defer C.free(unsafe.Pointer(cOid))
+	var cMtime *C.struct_timespec
+	if mtime != nil {
+		cMtime = &C.struct_timespec{}
+		ts.CopyToCStruct(
+			ts.Timespec(*mtime),
+			ts.CTimespecPtr(cMtime))
+	}
+
+	ret := C.rados_write_op_operate2(
+		w.op, ioctx.ioctx, cOid, cMtime, C.int(flags))
+	return w.update(writeOp, ret)
+}
+
+// Operate will perform the operation(s).
+func (w *WriteOp) Operate(ioctx *IOContext, oid string, flags OperationFlags) error {
+	return w.operate2(ioctx, oid, nil, flags)
+}
+
+// OperateWithMtime will perform the operation while setting the modification
+// time stamp to the supplied value.
+func (w *WriteOp) OperateWithMtime(
+	ioctx *IOContext, oid string, mtime Timespec, flags OperationFlags) error {
+
+	return w.operate2(ioctx, oid, &mtime, flags)
+}
+
+func (w *WriteOp) operateCompat(ioctx *IOContext, oid string) error {
+	switch err := w.Operate(ioctx, oid, OperationNoFlag).(type) {
+	case nil:
+		return nil
+	case OperationError:
+		return err.OpError
+	default:
+		return err
+	}
+}
+
+// Create a rados object.
+func (w *WriteOp) Create(exclusive CreateOption) {
+	// category, the 3rd param, is deprecated and has no effect so we do not
+	// implement it in go-ceph
+	C.rados_write_op_create(w.op, C.int(exclusive), nil)
+}
+
+// SetOmap appends the map `pairs` to the omap `oid`.
+func (w *WriteOp) SetOmap(pairs map[string][]byte) {
+	keys := make([]string, len(pairs))
+	values := make([][]byte, len(pairs))
+	idx := 0
+	for k, v := range pairs {
+		keys[idx] = k
+		values[idx] = v
+		idx++
+	}
+
+	cKeys := cutil.NewBufferGroupStrings(keys)
+	cValues := cutil.NewBufferGroupBytes(values)
+	defer cKeys.Free()
+	defer cValues.Free()
+
+	C.rados_write_op_omap_set2(
+		w.op,
+		(**C.char)(cKeys.BuffersPtr()),
+		(**C.char)(cValues.BuffersPtr()),
+		(*C.size_t)(cKeys.LengthsPtr()),
+		(*C.size_t)(cValues.LengthsPtr()),
+		(C.size_t)(len(pairs)))
+}
+
+// RmOmapKeys removes the specified `keys` from the omap `oid`.
+func (w *WriteOp) RmOmapKeys(keys []string) {
+	cKeys := cutil.NewBufferGroupStrings(keys)
+	defer cKeys.Free()
+
+	C.rados_write_op_omap_rm_keys2(
+		w.op,
+		(**C.char)(cKeys.BuffersPtr()),
+		(*C.size_t)(cKeys.LengthsPtr()),
+		(C.size_t)(len(keys)))
+}
+
+// CleanOmap clears the omap `oid`.
+func (w *WriteOp) CleanOmap() {
+	C.rados_write_op_omap_clear(w.op)
+}
+
+// AssertExists assures the object targeted by the write op exists.
+//
+// Implements:
+//  void rados_write_op_assert_exists(rados_write_op_t write_op);
+func (w *WriteOp) AssertExists() {
+	C.rados_write_op_assert_exists(w.op)
+}
+
+// Write a given byte slice at the supplied offset.
+//
+// Implements:
+//  void rados_write_op_write(rados_write_op_t write_op,
+//                                       const char *buffer,
+//                                       size_t len,
+//                                       uint64_t offset);
+func (w *WriteOp) Write(b []byte, offset uint64) {
+	oe := newWriteStep(b, 0, offset)
+	w.steps = append(w.steps, oe)
+	C.rados_write_op_write(
+		w.op,
+		oe.cBuffer,
+		oe.cDataLen,
+		oe.cOffset)
+}
+
+// WriteFull writes a given byte slice as the whole object,
+// atomically replacing it.
+//
+// Implements:
+//  void rados_write_op_write_full(rados_write_op_t write_op,
+//                                 const char *buffer,
+//                                 size_t len);
+func (w *WriteOp) WriteFull(b []byte) {
+	oe := newWriteStep(b, 0, 0)
+	w.steps = append(w.steps, oe)
+	C.rados_write_op_write_full(
+		w.op,
+		oe.cBuffer,
+		oe.cDataLen)
+}
+
+// WriteSame write a given byte slice to the object multiple times, until
+// writeLen is satisfied.
+//
+// Implements:
+//  void rados_write_op_writesame(rados_write_op_t write_op,
+//                                const char *buffer,
+//                                size_t data_len,
+//                                size_t write_len,
+//                                uint64_t offset);
+func (w *WriteOp) WriteSame(b []byte, writeLen, offset uint64) {
+	oe := newWriteStep(b, writeLen, offset)
+	w.steps = append(w.steps, oe)
+	C.rados_write_op_writesame(
+		w.op,
+		oe.cBuffer,
+		oe.cDataLen,
+		oe.cWriteLen,
+		oe.cOffset)
+}

vendor/github.com/ceph/go-ceph/rados/write_op_cmpext.go

@@ -0,0 +1,59 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+import (
+	"unsafe"
+)
+
+// WriteOpCmpExtStep holds result of the CmpExt write operation.
+// Result is valid only after Operate() was called.
+type WriteOpCmpExtStep struct {
+	// C returned data:
+	prval *C.int
+
+	// Result of the CmpExt write operation.
+	Result int
+}
+
+func (s *WriteOpCmpExtStep) update() error {
+	s.Result = int(*s.prval)
+	return nil
+}
+
+func (s *WriteOpCmpExtStep) free() {
+	C.free(unsafe.Pointer(s.prval))
+	s.prval = nil
+}
+
+func newWriteOpCmpExtStep() *WriteOpCmpExtStep {
+	return &WriteOpCmpExtStep{
+		prval: (*C.int)(C.malloc(C.sizeof_int)),
+	}
+}
+
+// CmpExt ensures that given object range (extent) satisfies comparison.
+//
+// Implements:
+//  void rados_write_op_cmpext(rados_write_op_t write_op,
+//                             const char * cmp_buf,
+//                             size_t cmp_len,
+//                             uint64_t off,
+//                             int * prval);
+func (w *WriteOp) CmpExt(b []byte, offset uint64) *WriteOpCmpExtStep {
+	oe := newWriteStep(b, 0, offset)
+	cmpExtStep := newWriteOpCmpExtStep()
+	w.steps = append(w.steps, oe, cmpExtStep)
+	C.rados_write_op_cmpext(
+		w.op,
+		oe.cBuffer,
+		oe.cDataLen,
+		oe.cOffset,
+		cmpExtStep.prval)
+
+	return cmpExtStep
+}

vendor/github.com/ceph/go-ceph/rados/write_op_set_alloc_hint.go

@@ -0,0 +1,28 @@
+//go:build ceph_preview
+// +build ceph_preview
+
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <rados/librados.h>
+// #include <stdlib.h>
+//
+import "C"
+
+// SetAllocationHint sets allocation hint for an object. This is an advisory
+// operation, it will always succeed (as if it was submitted with a
+// LIBRADOS_OP_FLAG_FAILOK flag set) and is not guaranteed to do anything on
+// the backend.
+//
+// Implements:
+//  void rados_write_op_set_alloc_hint2(rados_write_op_t write_op,
+//                                      uint64_t expected_object_size,
+//                                      uint64_t expected_write_size,
+//                                      uint32_t flags);
+func (w *WriteOp) SetAllocationHint(expectedObjectSize uint64, expectedWriteSize uint64, flags AllocHintFlags) {
+	C.rados_write_op_set_alloc_hint2(
+		w.op,
+		C.uint64_t(expectedObjectSize),
+		C.uint64_t(expectedWriteSize),
+		C.uint32_t(flags))
+}

vendor/github.com/ceph/go-ceph/rados/write_step.go

@@ -0,0 +1,33 @@
+package rados
+
+// #include <stdint.h>
+import "C"
+
+import (
+	"unsafe"
+)
+
+type writeStep struct {
+	withoutUpdate
+	withoutFree
+	// the c pointer utilizes the Go byteslice data and no free is needed
+
+	// inputs:
+	b []byte
+
+	// arguments:
+	cBuffer   *C.char
+	cDataLen  C.size_t
+	cWriteLen C.size_t
+	cOffset   C.uint64_t
+}
+
+func newWriteStep(b []byte, writeLen, offset uint64) *writeStep {
+	return &writeStep{
+		b:         b,
+		cBuffer:   (*C.char)(unsafe.Pointer(&b[0])), // TODO: must be pinned
+		cDataLen:  C.size_t(len(b)),
+		cWriteLen: C.size_t(writeLen),
+		cOffset:   C.uint64_t(offset),
+	}
+}