rebase: update go-ceph to v0.8.0

Updating go-ceph to v0.8.0.

Signed-off-by: Mudit Agarwal <muagarwa@redhat.com>

go.mod

@@ -3,7 +3,7 @@ module github.com/ceph/ceph-csi
 go 1.13
 
 require (
-	github.com/ceph/go-ceph v0.7.0
+	github.com/ceph/go-ceph v0.8.0
 	github.com/container-storage-interface/spec v1.3.0
 	github.com/go-logr/logr v0.2.1 // indirect
 	github.com/golang/protobuf v1.4.3

go.sum

@@ -203,8 +203,8 @@ github.com/cenkalti/backoff v2.2.1+incompatible/go.mod h1:90ReRw6GdpyfrHakVjL/QH
 github.com/census-instrumentation/opencensus-proto v0.2.1/go.mod h1:f6KPmirojxKA12rnyqOA5BBL4O983OfeGPqjHWSTneU=
 github.com/centrify/cloud-golang-sdk v0.0.0-20190214225812-119110094d0f h1:gJzxrodnNd/CtPXjO3WYiakyNzHg3rtAi7rO74ejHYU=
 github.com/centrify/cloud-golang-sdk v0.0.0-20190214225812-119110094d0f/go.mod h1:C0rtzmGXgN78pYR0tGJFhtHgkbAs0lIbHwkB81VxDQE=
-github.com/ceph/go-ceph v0.7.0 h1:+4oWZCuvh9B7/kZVJVw/DSuk9Qby38KWY1pMQ5gYGyY=
+github.com/ceph/go-ceph v0.8.0 h1:d+VP0eyconBl9RrvKVUq7S0npyK969ErLkCt5pg2fp0=
-github.com/ceph/go-ceph v0.7.0/go.mod h1:wd+keAOqrcsN//20VQnHBGtnBnY0KHl0PA024Ng8HfQ=
+github.com/ceph/go-ceph v0.8.0/go.mod h1:wd+keAOqrcsN//20VQnHBGtnBnY0KHl0PA024Ng8HfQ=
 github.com/cespare/prettybench v0.0.0-20150116022406-03b8cfe5406c/go.mod h1:Xe6ZsFhtM8HrDku0pxJ3/Lr51rwykrzgFwpmTzleatY=
 github.com/cespare/xxhash v1.1.0 h1:a6HrQnmkObjyL+Gs60czilIUGqrzKutQD6XZog3p+ko=
 github.com/cespare/xxhash v1.1.0/go.mod h1:XrSqR1VqqWfGrhpAt58auRo0WTKS1nRRg3ghfAqPWnc=

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

@@ -41,7 +41,7 @@ func (f SubVolRmFlags) flags() map[string]bool {
 		o["force"] = true
 	}
 	if f.RetainSnapshots {
-		o["retain-snapshots"] = true
+		o["retain_snapshots"] = true
 	}
 	return o
 }

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

@@ -197,23 +197,48 @@ const (
 	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"`
+	Type          string         `json:"type"`
-	Path          string    `json:"path"`
+	Path          string         `json:"path"`
-	Uid           int       `json:"uid"`
+	State         SubVolumeState `json:"state"`
-	Gid           int       `json:"gid"`
+	Uid           int            `json:"uid"`
-	Mode          int       `json:"mode"`
+	Gid           int            `json:"gid"`
-	BytesPercent  string    `json:"bytes_pcent"`
+	Mode          int            `json:"mode"`
-	BytesUsed     ByteCount `json:"bytes_used"`
+	BytesPercent  string         `json:"bytes_pcent"`
-	BytesQuota    QuotaSize `json:"-"`
+	BytesUsed     ByteCount      `json:"bytes_used"`
-	DataPool      string    `json:"data_pool"`
+	BytesQuota    QuotaSize      `json:"-"`
-	PoolNamespace string    `json:"pool_namespace"`
+	DataPool      string         `json:"data_pool"`
-	Atime         TimeStamp `json:"atime"`
+	PoolNamespace string         `json:"pool_namespace"`
-	Mtime         TimeStamp `json:"mtime"`
+	Atime         TimeStamp      `json:"atime"`
-	Ctime         TimeStamp `json:"ctime"`
+	Mtime         TimeStamp      `json:"mtime"`
-	CreatedAt     TimeStamp `json:"created_at"`
+	Ctime         TimeStamp      `json:"ctime"`
-	Features      []Feature `json:"features"`
+	CreatedAt     TimeStamp      `json:"created_at"`
+	Features      []Feature      `json:"features"`
 }
 
 type subVolumeInfoWrapper struct {

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

@@ -0,0 +1,62 @@
+package cutil
+
+/*
+#include <stdlib.h>
+#include <string.h>
+typedef void* voidptr;
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+const (
+	// MaxIdx is the maximum index on 32 bit systems
+	MaxIdx = 1<<31 - 1 // 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
+)
+
+// 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/iovec.go

@@ -5,74 +5,56 @@ package cutil
 #include <sys/uio.h>
 */
 import "C"
-
 import (
 	"unsafe"
 )
 
-var iovecSize uintptr
+// Iovec is a slice of iovec structs. Might have allocated C memory, so it must
-
+// be freed with the Free() method.
-// StructIovecPtr is an unsafe pointer wrapping C's `*struct iovec`.
-type StructIovecPtr unsafe.Pointer
-
-// Iovec helps manage struct iovec arrays needed by some C functions.
 type Iovec struct {
-	// cvec represents an array of struct iovec C memory
+	iovec []C.struct_iovec
-	cvec unsafe.Pointer
+	sbs   []*SyncBuffer
-	// length of the array (in elements)
-	length int
 }
 
-// NewIovec creates an Iovec, and underlying C memory, of the specified size.
+const iovecSize = C.sizeof_struct_iovec
-func NewIovec(l int) *Iovec {
+
-	r := &Iovec{
+// ByteSlicesToIovec creates an Iovec and links it to Go buffers in data.
-		cvec:   C.malloc(C.size_t(l) * C.size_t(iovecSize)),
+func ByteSlicesToIovec(data [][]byte) (v Iovec) {
-		length: l,
+	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 r
+	return
 }
 
-// ByteSlicesToIovec takes a slice of byte slices and returns a new iovec that
+// Sync makes sure the slices contain the same as the C buffers
-// maps the slice data to struct iovec entries.
+func (v *Iovec) Sync() {
-func ByteSlicesToIovec(data [][]byte) *Iovec {
+	for _, sb := range v.sbs {
-	iov := NewIovec(len(data))
+		sb.Sync()
-	for i := range data {
-		iov.Set(i, data[i])
 	}
-	return iov
 }
 
-// Pointer returns a StructIovecPtr that represents the C memory of the
+// Pointer returns a pointer to the iovec
-// underlying array.
+func (v *Iovec) Pointer() unsafe.Pointer {
-func (v *Iovec) Pointer() StructIovecPtr {
+	return unsafe.Pointer(&v.iovec[0])
-	return StructIovecPtr(unsafe.Pointer(v.cvec))
 }
 
-// Len returns the number of entries in the Iovec.
+// Len returns a pointer to the iovec
 func (v *Iovec) Len() int {
-	return v.length
+	return len(v.iovec)
 }
 
 // Free the C memory in the Iovec.
 func (v *Iovec) Free() {
-	if v.cvec != nil {
+	for _, sb := range v.sbs {
-		C.free(v.cvec)
+		sb.Release()
-		v.cvec = nil
-		v.length = 0
 	}
-}
+	if len(v.iovec) != 0 {
-
+		C.free(unsafe.Pointer(&v.iovec[0]))
-// Set will map the memory of the given byte slice to the iovec at the
+	}
-// specified position.
+	v.iovec = nil
-func (v *Iovec) Set(i int, buf []byte) {
-	offset := uintptr(i) * iovecSize
-	iov := (*C.struct_iovec)(unsafe.Pointer(
-		uintptr(unsafe.Pointer(v.cvec)) + offset))
-	iov.iov_base = unsafe.Pointer(&buf[0])
-	iov.iov_len = C.size_t(len(buf))
-}
-
-func init() {
-	var iovec C.struct_iovec
-	iovecSize = unsafe.Sizeof(iovec)
 }

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

@@ -0,0 +1,79 @@
+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)
+
+	}
+}
+
+//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 := (*uintptr)(unsafe.Pointer(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/sync_buffer.go

@@ -0,0 +1,28 @@
+// +build 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 (v *SyncBuffer) Sync() {}

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

@@ -0,0 +1,37 @@
+// +build !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/cutil/type_aliases.go

@@ -1,28 +0,0 @@
-package cutil
-
-import "C"
-
-import (
-	"unsafe"
-)
-
-// Basic types from C that we can make "public" without too much fuss.
-
-// 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.
-
-// 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)

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

@@ -28,3 +28,12 @@ func CStructToTimespec(cts CTimespecPtr) Timespec {
 		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/errors.go

@@ -52,6 +52,9 @@ var (
 	// 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:

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

@@ -128,15 +128,10 @@ func (ioctx *IOContext) SetNamespace(namespace string) {
 //  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 {
-	c_oid := C.CString(oid)
+	op := CreateWriteOp()
-	defer C.free(unsafe.Pointer(c_oid))
+	defer op.Release()
-
+	op.Create(exclusive)
-	op := C.rados_create_write_op()
+	return op.operateCompat(ioctx, oid)
-	C.rados_write_op_create(op, C.int(exclusive), nil)
-	ret := C.rados_write_op_operate(op, ioctx.ioctx, c_oid, nil, 0)
-	C.rados_release_write_op(op)
-
-	return getError(ret)
 }
 
 // Write writes len(data) bytes to the object with key oid starting at byte

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

@@ -1,69 +1,230 @@
 package rados
 
-// #cgo LDFLAGS: -lrados
+/*
-// #include <stdlib.h>
+#cgo LDFLAGS: -lrados
-// #include <rados/librados.h>
+#include <stdlib.h>
-//
+#include <rados/librados.h>
+
+typedef void* voidptr;
+
+*/
 import "C"
 
 import (
+	"runtime"
 	"unsafe"
 )
 
-// SetOmap appends the map `pairs` to the omap `oid`
+const (
-func (ioctx *IOContext) SetOmap(oid string, pairs map[string][]byte) error {
+	ptrSize   = C.sizeof_voidptr
-	c_oid := C.CString(oid)
+	sizeTSize = C.sizeof_size_t
-	defer C.free(unsafe.Pointer(c_oid))
+)
 
-	var s C.size_t
+// setOmapStep is a write op step. It holds C memory used in the operation.
-	var c *C.char
+type setOmapStep struct {
-	ptrSize := unsafe.Sizeof(c)
+	withRefs
+	withoutUpdate
 
-	c_keys := C.malloc(C.size_t(len(pairs)) * C.size_t(ptrSize))
+	// C arguments
-	c_values := C.malloc(C.size_t(len(pairs)) * C.size_t(ptrSize))
+	cKeys    **C.char
-	c_lengths := C.malloc(C.size_t(len(pairs)) * C.size_t(unsafe.Sizeof(s)))
+	cValues  **C.char
+	cLengths *C.size_t
+	cNum     C.size_t
+}
 
-	defer C.free(unsafe.Pointer(c_keys))
+func newSetOmapStep(pairs map[string][]byte) *setOmapStep {
-	defer C.free(unsafe.Pointer(c_values))
-	defer C.free(unsafe.Pointer(c_lengths))
 
-	i := 0
+	maplen := C.size_t(len(pairs))
+	cKeys := C.malloc(maplen * ptrSize)
+	cValues := C.malloc(maplen * ptrSize)
+	cLengths := C.malloc(maplen * sizeTSize)
+
+	sos := &setOmapStep{
+		cKeys:    (**C.char)(cKeys),
+		cValues:  (**C.char)(cValues),
+		cLengths: (*C.size_t)(cLengths),
+		cNum:     C.size_t(len(pairs)),
+	}
+	sos.add(cKeys)
+	sos.add(cValues)
+	sos.add(cLengths)
+
+	var i uintptr
 	for key, value := range pairs {
 		// key
-		c_key_ptr := (**C.char)(unsafe.Pointer(uintptr(c_keys) + uintptr(i)*ptrSize))
+		ck := C.CString(key)
-		*c_key_ptr = C.CString(key)
+		sos.add(unsafe.Pointer(ck))
-		defer C.free(unsafe.Pointer(*c_key_ptr))
+		ckp := (**C.char)(unsafe.Pointer(uintptr(cKeys) + i*ptrSize))
+		*ckp = ck
 
 		// value and its length
-		c_value_ptr := (**C.char)(unsafe.Pointer(uintptr(c_values) + uintptr(i)*ptrSize))
+		cvp := (**C.char)(unsafe.Pointer(uintptr(cValues) + i*ptrSize))
-
+		vlen := C.size_t(len(value))
-		var c_length C.size_t
+		if vlen > 0 {
-		if len(value) > 0 {
+			cv := C.CBytes(value)
-			*c_value_ptr = (*C.char)(unsafe.Pointer(&value[0]))
+			sos.add(cv)
-			c_length = C.size_t(len(value))
+			*cvp = (*C.char)(cv)
 		} else {
-			*c_value_ptr = nil
+			*cvp = nil
-			c_length = C.size_t(0)
 		}
 
-		c_length_ptr := (*C.size_t)(unsafe.Pointer(uintptr(c_lengths) + uintptr(i)*ptrSize))
+		clp := (*C.size_t)(unsafe.Pointer(uintptr(cLengths) + i*ptrSize))
-		*c_length_ptr = c_length
+		*clp = vlen
 
 		i++
 	}
 
-	op := C.rados_create_write_op()
+	runtime.SetFinalizer(sos, opStepFinalizer)
-	C.rados_write_op_omap_set(
+	return sos
-		op,
+}
-		(**C.char)(c_keys),
-		(**C.char)(c_values),
-		(*C.size_t)(c_lengths),
-		C.size_t(len(pairs)))
 
-	ret := C.rados_write_op_operate(op, ioctx.ioctx, c_oid, nil, 0)
+func (sos *setOmapStep) free() {
-	C.rados_release_write_op(op)
+	sos.cKeys = nil
+	sos.cValues = nil
+	sos.cLengths = nil
+	sos.withRefs.free()
+}
 
-	return getError(ret)
+// 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 {
+	// inputs:
+	startAfter   string
+	filterPrefix string
+	maxReturn    uint64
+
+	// arguments:
+	cStartAfter   *C.char
+	cFilterPrefix *C.char
+
+	// 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(startAfter, filterPrefix string, maxReturn uint64) *GetOmapStep {
+	gos := &GetOmapStep{
+		startAfter:    startAfter,
+		filterPrefix:  filterPrefix,
+		maxReturn:     maxReturn,
+		cStartAfter:   C.CString(startAfter),
+		cFilterPrefix: C.CString(filterPrefix),
+	}
+	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
+	gos.more = 0
+	gos.rval = 0
+	C.free(unsafe.Pointer(gos.cStartAfter))
+	gos.cStartAfter = nil
+	C.free(unsafe.Pointer(gos.cFilterPrefix))
+	gos.cFilterPrefix = 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
+		cLen C.size_t
+	)
+	ret := C.rados_omap_get_next(gos.iter, &cKey, &cVal, &cLen)
+	if ret != 0 {
+		return nil, getError(ret)
+	}
+	if cKey == nil {
+		return nil, nil
+	}
+	return &OmapKeyValue{
+		Key:   C.GoString(cKey),
+		Value: C.GoBytes(unsafe.Pointer(cVal), C.int(cLen)),
+	}, 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
+}
+
+// removeOmapKeysStep is a write operation step used to track state, especially
+// C memory, across the setup and use of a WriteOp.
+type removeOmapKeysStep struct {
+	withRefs
+	withoutUpdate
+
+	// arguments:
+	cKeys **C.char
+	cNum  C.size_t
+}
+
+func newRemoveOmapKeysStep(keys []string) *removeOmapKeysStep {
+	cKeys := C.malloc(C.size_t(len(keys)) * ptrSize)
+	roks := &removeOmapKeysStep{
+		cKeys: (**C.char)(cKeys),
+		cNum:  C.size_t(len(keys)),
+	}
+	roks.add(cKeys)
+
+	i := 0
+	for _, key := range keys {
+		ckp := (**C.char)(unsafe.Pointer(uintptr(cKeys) + uintptr(i)*ptrSize))
+		*ckp = C.CString(key)
+		roks.add(unsafe.Pointer(*ckp))
+		i++
+	}
+
+	runtime.SetFinalizer(roks, opStepFinalizer)
+	return roks
+}
+
+func (roks *removeOmapKeysStep) free() {
+	roks.cKeys = nil
+	roks.withRefs.free()
+}
+
+// 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
@@ -78,58 +239,25 @@ type OmapListFunc func(key string, value []byte)
 // `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 {
-	c_oid := C.CString(oid)
-	c_start_after := C.CString(startAfter)
-	c_filter_prefix := C.CString(filterPrefix)
-	c_max_return := C.uint64_t(maxReturn)
 
-	defer C.free(unsafe.Pointer(c_oid))
+	op := CreateReadOp()
-	defer C.free(unsafe.Pointer(c_start_after))
+	defer op.Release()
-	defer C.free(unsafe.Pointer(c_filter_prefix))
+	gos := op.GetOmapValues(startAfter, filterPrefix, uint64(maxReturn))
-
+	err := op.operateCompat(ioctx, oid)
-	op := C.rados_create_read_op()
+	if err != nil {
-
+		return err
-	var c_iter C.rados_omap_iter_t
-	var c_prval C.int
-	C.rados_read_op_omap_get_vals2(
-		op,
-		c_start_after,
-		c_filter_prefix,
-		c_max_return,
-		&c_iter,
-		nil,
-		&c_prval,
-	)
-
-	ret := C.rados_read_op_operate(op, ioctx.ioctx, c_oid, 0)
-
-	if int(ret) != 0 {
-		return getError(ret)
-	} else if int(c_prval) != 0 {
-		return getError(c_prval)
 	}
 
 	for {
-		var c_key *C.char
+		kv, err := gos.Next()
-		var c_val *C.char
+		if err != nil {
-		var c_len C.size_t
+			return err
-
-		ret = C.rados_omap_get_next(c_iter, &c_key, &c_val, &c_len)
-
-		if int(ret) != 0 {
-			return getError(ret)
 		}
-
+		if kv == nil {
-		if c_key == nil {
 			break
 		}
-
+		listFn(kv.Key, kv.Value)
-		listFn(C.GoString(c_key), C.GoBytes(unsafe.Pointer(c_val), C.int(c_len)))
 	}
-
-	C.rados_omap_get_end(c_iter)
-	C.rados_release_read_op(op)
-
 	return nil
 }
 
@@ -184,45 +312,16 @@ func (ioctx *IOContext) GetAllOmapValues(oid string, startAfter string, filterPr
 
 // RmOmapKeys removes the specified `keys` from the omap `oid`
 func (ioctx *IOContext) RmOmapKeys(oid string, keys []string) error {
-	c_oid := C.CString(oid)
+	op := CreateWriteOp()
-	defer C.free(unsafe.Pointer(c_oid))
+	defer op.Release()
-
+	op.RmOmapKeys(keys)
-	var c *C.char
+	return op.operateCompat(ioctx, oid)
-	ptrSize := unsafe.Sizeof(c)
-
-	c_keys := C.malloc(C.size_t(len(keys)) * C.size_t(ptrSize))
-	defer C.free(unsafe.Pointer(c_keys))
-
-	i := 0
-	for _, key := range keys {
-		c_key_ptr := (**C.char)(unsafe.Pointer(uintptr(c_keys) + uintptr(i)*ptrSize))
-		*c_key_ptr = C.CString(key)
-		defer C.free(unsafe.Pointer(*c_key_ptr))
-		i++
-	}
-
-	op := C.rados_create_write_op()
-	C.rados_write_op_omap_rm_keys(
-		op,
-		(**C.char)(c_keys),
-		C.size_t(len(keys)))
-
-	ret := C.rados_write_op_operate(op, ioctx.ioctx, c_oid, nil, 0)
-	C.rados_release_write_op(op)
-
-	return getError(ret)
 }
 
 // CleanOmap clears the omap `oid`
 func (ioctx *IOContext) CleanOmap(oid string) error {
-	c_oid := C.CString(oid)
+	op := CreateWriteOp()
-	defer C.free(unsafe.Pointer(c_oid))
+	defer op.Release()
-
+	op.CleanOmap()
-	op := C.rados_create_write_op()
+	return op.operateCompat(ioctx, oid)
-	C.rados_write_op_omap_clear(op)
-
-	ret := C.rados_write_op_operate(op, ioctx.ioctx, c_oid, nil, 0)
-	C.rados_release_write_op(op)
-
-	return getError(ret)
 }

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

@@ -0,0 +1,151 @@
+package rados
+
+// #include <stdlib.h>
+import "C"
+
+import (
+	"fmt"
+	"strings"
+	"unsafe"
+)
+
+// 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 {
+		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/read_op.go

@@ -0,0 +1,84 @@
+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(startAfter, filterPrefix, maxReturn)
+	r.steps = append(r.steps, gos)
+	C.rados_read_op_omap_get_vals2(
+		r.op,
+		gos.cStartAfter,
+		gos.cFilterPrefix,
+		C.uint64_t(gos.maxReturn),
+		&gos.iter,
+		&gos.more,
+		&gos.rval,
+	)
+	return gos
+}

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

@@ -0,0 +1,179 @@
+package rados
+
+// #cgo LDFLAGS: -lrados
+// #include <errno.h>
+// #include <stdlib.h>
+// #include <rados/librados.h>
+//
+import "C"
+
+import (
+	"unsafe"
+
+	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) {
+	sos := newSetOmapStep(pairs)
+	w.steps = append(w.steps, sos)
+	C.rados_write_op_omap_set(
+		w.op,
+		sos.cKeys,
+		sos.cValues,
+		sos.cLengths,
+		sos.cNum)
+}
+
+// RmOmapKeys removes the specified `keys` from the omap `oid`.
+func (w *WriteOp) RmOmapKeys(keys []string) {
+	roks := newRemoveOmapKeysStep(keys)
+	w.steps = append(w.steps, roks)
+	C.rados_write_op_omap_rm_keys(
+		w.op,
+		roks.cKeys,
+		roks.cNum)
+}
+
+// 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_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])),
+		cDataLen:  C.size_t(len(b)),
+		cWriteLen: C.size_t(writeLen),
+		cOffset:   C.uint64_t(offset),
+	}
+}

vendor/github.com/ceph/go-ceph/rbd/group.go

@@ -0,0 +1,258 @@
+package rbd
+
+/*
+#cgo LDFLAGS: -lrbd
+#include <stdlib.h>
+#include <rbd/librbd.h>
+*/
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/cutil"
+	"github.com/ceph/go-ceph/internal/retry"
+	"github.com/ceph/go-ceph/rados"
+)
+
+// GroupCreate is used to create an image group.
+//
+// Implements:
+//  int rbd_group_create(rados_ioctx_t p, const char *name);
+func GroupCreate(ioctx *rados.IOContext, name string) error {
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.rbd_group_create(cephIoctx(ioctx), cName)
+	return getError(ret)
+}
+
+// GroupRemove is used to remove an image group.
+//
+// Implements:
+//  int rbd_group_remove(rados_ioctx_t p, const char *name);
+func GroupRemove(ioctx *rados.IOContext, name string) error {
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	ret := C.rbd_group_remove(cephIoctx(ioctx), cName)
+	return getError(ret)
+}
+
+// GroupRename will rename an existing image group.
+//
+// Implements:
+//  int rbd_group_rename(rados_ioctx_t p, const char *src_name,
+//                       const char *dest_name);
+func GroupRename(ioctx *rados.IOContext, src, dest string) error {
+	cSrc := C.CString(src)
+	defer C.free(unsafe.Pointer(cSrc))
+	cDest := C.CString(dest)
+	defer C.free(unsafe.Pointer(cDest))
+
+	ret := C.rbd_group_rename(cephIoctx(ioctx), cSrc, cDest)
+	return getError(ret)
+}
+
+// GroupList returns a slice of image group names.
+//
+// Implements:
+//  int rbd_group_list(rados_ioctx_t p, char *names, size_t *size);
+func GroupList(ioctx *rados.IOContext) ([]string, error) {
+	var (
+		buf []byte
+		err error
+		ret C.int
+	)
+	retry.WithSizes(1024, 262144, func(size int) retry.Hint {
+		cSize := C.size_t(size)
+		buf = make([]byte, cSize)
+		ret = C.rbd_group_list(
+			cephIoctx(ioctx),
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			&cSize)
+		err = getErrorIfNegative(ret)
+		return retry.Size(int(cSize)).If(err == errRange)
+	})
+
+	if err != nil {
+		return nil, err
+	}
+
+	// cSize is not set to the expected size when it is sufficiently large
+	// but ret will be set to the size in a non-error condition.
+	groups := cutil.SplitBuffer(buf[:ret])
+	return groups, nil
+}
+
+// GroupImageAdd will add the specified image to the named group.
+// An io context must be supplied for both the group and image.
+//
+// Implements:
+//  int rbd_group_image_add(rados_ioctx_t group_p,
+//                          const char *group_name,
+//                          rados_ioctx_t image_p,
+//                          const char *image_name);
+func GroupImageAdd(groupIoctx *rados.IOContext, groupName string,
+	imageIoctx *rados.IOContext, imageName string) error {
+
+	cGroupName := C.CString(groupName)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cImageName := C.CString(imageName)
+	defer C.free(unsafe.Pointer(cImageName))
+
+	ret := C.rbd_group_image_add(
+		cephIoctx(groupIoctx),
+		cGroupName,
+		cephIoctx(imageIoctx),
+		cImageName)
+	return getError(ret)
+}
+
+// GroupImageRemove will remove the specified image from the named group.
+// An io context must be supplied for both the group and image.
+//
+// Implements:
+//  int rbd_group_image_remove(rados_ioctx_t group_p,
+//                             const char *group_name,
+//                             rados_ioctx_t image_p,
+//                             const char *image_name);
+func GroupImageRemove(groupIoctx *rados.IOContext, groupName string,
+	imageIoctx *rados.IOContext, imageName string) error {
+
+	cGroupName := C.CString(groupName)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cImageName := C.CString(imageName)
+	defer C.free(unsafe.Pointer(cImageName))
+
+	ret := C.rbd_group_image_remove(
+		cephIoctx(groupIoctx),
+		cGroupName,
+		cephIoctx(imageIoctx),
+		cImageName)
+	return getError(ret)
+}
+
+// GroupImageRemoveByID will remove the specified image from the named group.
+// An io context must be supplied for both the group and image.
+//
+// Implements:
+//  CEPH_RBD_API int rbd_group_image_remove_by_id(rados_ioctx_t group_p,
+//                                               const char *group_name,
+//                                               rados_ioctx_t image_p,
+//                                               const char *image_id);
+func GroupImageRemoveByID(groupIoctx *rados.IOContext, groupName string,
+	imageIoctx *rados.IOContext, imageID string) error {
+
+	cGroupName := C.CString(groupName)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cid := C.CString(imageID)
+	defer C.free(unsafe.Pointer(cid))
+
+	ret := C.rbd_group_image_remove_by_id(
+		cephIoctx(groupIoctx),
+		cGroupName,
+		cephIoctx(imageIoctx),
+		cid)
+	return getError(ret)
+}
+
+// GroupImageState indicates an image's state in a group.
+type GroupImageState int
+
+const (
+	// GroupImageStateAttached is equivalent to RBD_GROUP_IMAGE_STATE_ATTACHED
+	GroupImageStateAttached = GroupImageState(C.RBD_GROUP_IMAGE_STATE_ATTACHED)
+	// GroupImageStateIncomplete is equivalent to RBD_GROUP_IMAGE_STATE_INCOMPLETE
+	GroupImageStateIncomplete = GroupImageState(C.RBD_GROUP_IMAGE_STATE_INCOMPLETE)
+)
+
+// GroupImageInfo reports on images within a group.
+type GroupImageInfo struct {
+	Name   string
+	PoolID int64
+	State  GroupImageState
+}
+
+// GroupImageList returns a slice of GroupImageInfo types based on the
+// images that are part of the named group.
+//
+// Implements:
+//  int rbd_group_image_list(rados_ioctx_t group_p,
+//                           const char *group_name,
+//                           rbd_group_image_info_t *images,
+//                           size_t group_image_info_size,
+//                           size_t *num_entries);
+func GroupImageList(ioctx *rados.IOContext, name string) ([]GroupImageInfo, error) {
+	cName := C.CString(name)
+	defer C.free(unsafe.Pointer(cName))
+
+	var (
+		cImages []C.rbd_group_image_info_t
+		cSize   C.size_t
+		err     error
+	)
+	retry.WithSizes(1024, 262144, func(size int) retry.Hint {
+		cSize = C.size_t(size)
+		cImages = make([]C.rbd_group_image_info_t, cSize)
+		ret := C.rbd_group_image_list(
+			cephIoctx(ioctx),
+			cName,
+			(*C.rbd_group_image_info_t)(unsafe.Pointer(&cImages[0])),
+			C.sizeof_rbd_group_image_info_t,
+			&cSize)
+		err = getErrorIfNegative(ret)
+		return retry.Size(int(cSize)).If(err == errRange)
+	})
+
+	if err != nil {
+		return nil, err
+	}
+
+	images := make([]GroupImageInfo, cSize)
+	for i := range images {
+		images[i].Name = C.GoString(cImages[i].name)
+		images[i].PoolID = int64(cImages[i].pool)
+		images[i].State = GroupImageState(cImages[i].state)
+	}
+
+	// free C memory allocated by C.rbd_group_image_list call
+	ret := C.rbd_group_image_list_cleanup(
+		(*C.rbd_group_image_info_t)(unsafe.Pointer(&cImages[0])),
+		C.sizeof_rbd_group_image_info_t,
+		cSize)
+	return images, getError(ret)
+}
+
+// GroupInfo contains the name and pool id of a RBD group.
+type GroupInfo struct {
+	Name   string
+	PoolID int64
+}
+
+// GetGroup returns group info for the group this image is part of.
+//
+// Implements:
+//  int rbd_get_group(rbd_image_t image, rbd_group_info_t *group_info,
+//                    size_t group_info_size);
+func (image *Image) GetGroup() (GroupInfo, error) {
+	if err := image.validate(imageIsOpen); err != nil {
+		return GroupInfo{}, err
+	}
+
+	var cgi C.rbd_group_info_t
+	ret := C.rbd_get_group(
+		image.image,
+		&cgi,
+		C.sizeof_rbd_group_info_t)
+	if err := getErrorIfNegative(ret); err != nil {
+		return GroupInfo{}, err
+	}
+
+	gi := GroupInfo{
+		Name:   C.GoString(cgi.name),
+		PoolID: int64(cgi.pool),
+	}
+	ret = C.rbd_group_info_cleanup(&cgi, C.sizeof_rbd_group_info_t)
+	return gi, getError(ret)
+}

vendor/github.com/ceph/go-ceph/rbd/group_snap.go

@@ -0,0 +1,223 @@
+package rbd
+
+/*
+#cgo LDFLAGS: -lrbd
+#include <errno.h>
+#include <stdlib.h>
+#include <rbd/librbd.h>
+
+extern int snapRollbackCallback(uint64_t, uint64_t, uintptr_t);
+
+// inline wrapper to cast uintptr_t to void*
+static inline int wrap_rbd_group_snap_rollback_with_progress(
+		rados_ioctx_t group_p, const char *group_name,
+		const char *snap_name, uintptr_t arg) {
+	return rbd_group_snap_rollback_with_progress(
+		group_p, group_name, snap_name, (librbd_progress_fn_t)snapRollbackCallback, (void*)arg);
+};
+*/
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/callbacks"
+	"github.com/ceph/go-ceph/internal/retry"
+	"github.com/ceph/go-ceph/rados"
+)
+
+// GroupSnapCreate will create a group snapshot.
+//
+// Implements:
+//  int rbd_group_snap_create(rados_ioctx_t group_p,
+//                            const char *group_name,
+//                            const char *snap_name);
+func GroupSnapCreate(ioctx *rados.IOContext, group, snap string) error {
+	cGroupName := C.CString(group)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cSnapName := C.CString(snap)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ret := C.rbd_group_snap_create(cephIoctx(ioctx), cGroupName, cSnapName)
+	return getError(ret)
+}
+
+// GroupSnapRemove removes an existing group snapshot.
+//
+// Implements:
+//  int rbd_group_snap_remove(rados_ioctx_t group_p,
+//                            const char *group_name,
+//                            const char *snap_name);
+func GroupSnapRemove(ioctx *rados.IOContext, group, snap string) error {
+	cGroupName := C.CString(group)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cSnapName := C.CString(snap)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ret := C.rbd_group_snap_remove(cephIoctx(ioctx), cGroupName, cSnapName)
+	return getError(ret)
+}
+
+// GroupSnapRename will rename an existing group snapshot.
+//
+// Implements:
+//  int rbd_group_snap_rename(rados_ioctx_t group_p,
+//                            const char *group_name,
+//                            const char *old_snap_name,
+//                            const char *new_snap_name);
+func GroupSnapRename(ioctx *rados.IOContext, group, src, dest string) error {
+	cGroupName := C.CString(group)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cOldSnapName := C.CString(src)
+	defer C.free(unsafe.Pointer(cOldSnapName))
+	cNewSnapName := C.CString(dest)
+	defer C.free(unsafe.Pointer(cNewSnapName))
+
+	ret := C.rbd_group_snap_rename(
+		cephIoctx(ioctx), cGroupName, cOldSnapName, cNewSnapName)
+	return getError(ret)
+}
+
+// GroupSnapState represents the state of a group snapshot in GroupSnapInfo.
+type GroupSnapState int
+
+const (
+	// GroupSnapStateIncomplete is equivalent to RBD_GROUP_SNAP_STATE_INCOMPLETE.
+	GroupSnapStateIncomplete = GroupSnapState(C.RBD_GROUP_SNAP_STATE_INCOMPLETE)
+	// GroupSnapStateComplete is equivalent to RBD_GROUP_SNAP_STATE_COMPLETE.
+	GroupSnapStateComplete = GroupSnapState(C.RBD_GROUP_SNAP_STATE_COMPLETE)
+)
+
+// GroupSnapInfo values are returned by GroupSnapList, representing the
+// snapshots that are part of an rbd group.
+type GroupSnapInfo struct {
+	Name  string
+	State GroupSnapState
+}
+
+// GroupSnapList returns a slice of snapshots in a group.
+//
+// Implements:
+//  int rbd_group_snap_list(rados_ioctx_t group_p,
+//                          const char *group_name,
+//                          rbd_group_snap_info_t *snaps,
+//                          size_t group_snap_info_size,
+//                          size_t *num_entries);
+func GroupSnapList(ioctx *rados.IOContext, group string) ([]GroupSnapInfo, error) {
+	cGroupName := C.CString(group)
+	defer C.free(unsafe.Pointer(cGroupName))
+
+	var (
+		cSnaps []C.rbd_group_snap_info_t
+		cSize  C.size_t
+		err    error
+	)
+	retry.WithSizes(1024, 262144, func(size int) retry.Hint {
+		cSize = C.size_t(size)
+		cSnaps = make([]C.rbd_group_snap_info_t, cSize)
+		ret := C.rbd_group_snap_list(
+			cephIoctx(ioctx),
+			cGroupName,
+			(*C.rbd_group_snap_info_t)(unsafe.Pointer(&cSnaps[0])),
+			C.sizeof_rbd_group_snap_info_t,
+			&cSize)
+		err = getErrorIfNegative(ret)
+		return retry.Size(int(cSize)).If(err == errRange)
+	})
+
+	if err != nil {
+		return nil, err
+	}
+
+	snaps := make([]GroupSnapInfo, cSize)
+	for i := range snaps {
+		snaps[i].Name = C.GoString(cSnaps[i].name)
+		snaps[i].State = GroupSnapState(cSnaps[i].state)
+	}
+
+	// free C memory allocated by C.rbd_group_snap_list call
+	ret := C.rbd_group_snap_list_cleanup(
+		(*C.rbd_group_snap_info_t)(unsafe.Pointer(&cSnaps[0])),
+		C.sizeof_rbd_group_snap_info_t,
+		cSize)
+	return snaps, getError(ret)
+}
+
+// GroupSnapRollback will roll back the images in the group to that of the
+// given snapshot.
+//
+// Implements:
+//  int rbd_group_snap_rollback(rados_ioctx_t group_p,
+//                              const char *group_name,
+//                              const char *snap_name);
+func GroupSnapRollback(ioctx *rados.IOContext, group, snap string) error {
+	cGroupName := C.CString(group)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cSnapName := C.CString(snap)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ret := C.rbd_group_snap_rollback(cephIoctx(ioctx), cGroupName, cSnapName)
+	return getError(ret)
+}
+
+// GroupSnapRollbackCallback defines the function signature needed for the
+// GroupSnapRollbackWithProgress callback.
+//
+// This callback will be called by GroupSnapRollbackWithProgress when it
+// wishes to report progress rolling back a group snapshot.
+type GroupSnapRollbackCallback func(uint64, uint64, interface{}) int
+
+var groupSnapRollbackCallbacks = callbacks.New()
+
+// GroupSnapRollbackWithProgress will roll back the images in the group
+// to that of given snapshot. The given progress callback will be called
+// to report on the progress of the snapshot rollback.
+//
+// Implements:
+//  int rbd_group_snap_rollback_with_progress(rados_ioctx_t group_p,
+//                                            const char *group_name,
+//                                            const char *snap_name,
+//                                            librbd_progress_fn_t cb,
+//                                            void *cbdata);
+func GroupSnapRollbackWithProgress(
+	ioctx *rados.IOContext, group, snap string,
+	cb GroupSnapRollbackCallback, data interface{}) error {
+	// the provided callback must be a real function
+	if cb == nil {
+		return rbdError(C.EINVAL)
+	}
+
+	cGroupName := C.CString(group)
+	defer C.free(unsafe.Pointer(cGroupName))
+	cSnapName := C.CString(snap)
+	defer C.free(unsafe.Pointer(cSnapName))
+
+	ctx := gsnapRollbackCallbackCtx{
+		callback: cb,
+		data:     data,
+	}
+	cbIndex := groupSnapRollbackCallbacks.Add(ctx)
+	defer diffIterateCallbacks.Remove(cbIndex)
+
+	ret := C.wrap_rbd_group_snap_rollback_with_progress(
+		cephIoctx(ioctx),
+		cGroupName,
+		cSnapName,
+		C.uintptr_t(cbIndex))
+
+	return getError(ret)
+}
+
+type gsnapRollbackCallbackCtx struct {
+	callback GroupSnapRollbackCallback
+	data     interface{}
+}
+
+//export snapRollbackCallback
+func snapRollbackCallback(
+	offset, total C.uint64_t, index uintptr) C.int {
+
+	v := groupSnapRollbackCallbacks.Lookup(index)
+	ctx := v.(gsnapRollbackCallbackCtx)
+	return C.int(ctx.callback(uint64(offset), uint64(total), ctx.data))
+}

vendor/github.com/ceph/go-ceph/rbd/mirror.go

@@ -0,0 +1,235 @@
+// +build !nautilus
+
+// Initially, we're only providing mirroring related functions for octopus as
+// that version of ceph deprecated a number of the functions in nautilus. If
+// you need mirroring on an earlier supported version of ceph please file an
+// issue in our tracker.
+
+package rbd
+
+// #cgo LDFLAGS: -lrbd
+// #include <stdlib.h>
+// #include <rbd/librbd.h>
+import "C"
+
+import (
+	"unsafe"
+
+	"github.com/ceph/go-ceph/internal/retry"
+	"github.com/ceph/go-ceph/rados"
+)
+
+// MirrorMode is used to indicate an approach used for RBD mirroring.
+type MirrorMode int64
+
+const (
+	// MirrorModeDisabled disables mirroring.
+	MirrorModeDisabled = MirrorMode(C.RBD_MIRROR_MODE_DISABLED)
+	// MirrorModeImage enables mirroring on a per-image basis.
+	MirrorModeImage = MirrorMode(C.RBD_MIRROR_MODE_IMAGE)
+	// MirrorModePool enables mirroring on all journaled images.
+	MirrorModePool = MirrorMode(C.RBD_MIRROR_MODE_POOL)
+)
+
+// ImageMirrorMode is used to indicate the mirroring approach for an RBD image.
+type ImageMirrorMode int64
+
+const (
+	// ImageMirrorModeJournal uses journaling to propagate RBD images between
+	// ceph clusters.
+	ImageMirrorModeJournal = ImageMirrorMode(C.RBD_MIRROR_IMAGE_MODE_JOURNAL)
+	// ImageMirrorModeSnapshot uses snapshots to propagate RBD images between
+	// ceph clusters.
+	ImageMirrorModeSnapshot = ImageMirrorMode(C.RBD_MIRROR_IMAGE_MODE_SNAPSHOT)
+)
+
+// SetMirrorMode is used to enable or disable pool level mirroring with either
+// an automatic or per-image behavior.
+//
+// Implements:
+//  int rbd_mirror_mode_set(rados_ioctx_t io_ctx,
+//                          rbd_mirror_mode_t mirror_mode);
+func SetMirrorMode(ioctx *rados.IOContext, mode MirrorMode) error {
+	ret := C.rbd_mirror_mode_set(
+		cephIoctx(ioctx),
+		C.rbd_mirror_mode_t(mode))
+	return getError(ret)
+}
+
+// GetMirrorMode is used to fetch the current mirroring mode for a pool.
+//
+// Implements:
+//  int rbd_mirror_mode_get(rados_ioctx_t io_ctx,
+//                          rbd_mirror_mode_t *mirror_mode);
+func GetMirrorMode(ioctx *rados.IOContext) (MirrorMode, error) {
+	var mode C.rbd_mirror_mode_t
+
+	ret := C.rbd_mirror_mode_get(
+		cephIoctx(ioctx),
+		&mode)
+	if err := getError(ret); err != nil {
+		return MirrorModeDisabled, err
+	}
+	return MirrorMode(mode), nil
+}
+
+// MirrorEnable will enable mirroring for an image using the specified mode.
+//
+// Implements:
+//  int rbd_mirror_image_enable2(rbd_image_t image,
+//                               rbd_mirror_image_mode_t mode);
+func (image *Image) MirrorEnable(mode ImageMirrorMode) error {
+	if err := image.validate(imageIsOpen); err != nil {
+		return err
+	}
+	ret := C.rbd_mirror_image_enable2(image.image, C.rbd_mirror_image_mode_t(mode))
+	return getError(ret)
+}
+
+// MirrorDisable will disable mirroring for the image.
+//
+// Implements:
+//  int rbd_mirror_image_disable(rbd_image_t image, bool force);
+func (image *Image) MirrorDisable(force bool) error {
+	if err := image.validate(imageIsOpen); err != nil {
+		return err
+	}
+	ret := C.rbd_mirror_image_disable(image.image, C.bool(force))
+	return getError(ret)
+}
+
+// MirrorPromote will promote the image to primary status.
+//
+// Implements:
+//  int rbd_mirror_image_promote(rbd_image_t image, bool force);
+func (image *Image) MirrorPromote(force bool) error {
+	if err := image.validate(imageIsOpen); err != nil {
+		return err
+	}
+	ret := C.rbd_mirror_image_promote(image.image, C.bool(force))
+	return getError(ret)
+}
+
+// MirrorDemote will demote the image to secondary status.
+//
+// Implements:
+//  int rbd_mirror_image_demote(rbd_image_t image);
+func (image *Image) MirrorDemote() error {
+	if err := image.validate(imageIsOpen); err != nil {
+		return err
+	}
+	ret := C.rbd_mirror_image_demote(image.image)
+	return getError(ret)
+}
+
+// MirrorResync is used to manually resolve split-brain status by triggering
+// resynchronization.
+//
+// Implements:
+//  int rbd_mirror_image_resync(rbd_image_t image);
+func (image *Image) MirrorResync() error {
+	if err := image.validate(imageIsOpen); err != nil {
+		return err
+	}
+	ret := C.rbd_mirror_image_resync(image.image)
+	return getError(ret)
+}
+
+// MirrorInstanceID returns a string naming the instance id for the image.
+//
+// Implements:
+//  int rbd_mirror_image_get_instance_id(rbd_image_t image,
+//                                       char *instance_id,
+//                                       size_t *id_max_length);
+func (image *Image) MirrorInstanceID() (string, error) {
+	if err := image.validate(imageIsOpen); err != nil {
+		return "", err
+	}
+	var (
+		err   error
+		buf   []byte
+		cSize C.size_t
+	)
+	retry.WithSizes(1024, 1<<16, func(size int) retry.Hint {
+		cSize = C.size_t(size)
+		buf = make([]byte, cSize)
+		ret := C.rbd_mirror_image_get_instance_id(
+			image.image,
+			(*C.char)(unsafe.Pointer(&buf[0])),
+			&cSize)
+		err = getErrorIfNegative(ret)
+		return retry.Size(int(cSize)).If(err == errRange)
+	})
+	if err != nil {
+		return "", err
+	}
+	return string(buf[:cSize]), nil
+}
+
+// MirrorImageState represents the mirroring state of a RBD image.
+type MirrorImageState C.rbd_mirror_image_state_t
+
+const (
+	// MirrorImageDisabling is the representation of
+	// RBD_MIRROR_IMAGE_DISABLING from librbd.
+	MirrorImageDisabling = MirrorImageState(C.RBD_MIRROR_IMAGE_DISABLING)
+	// MirrorImageEnabled is the representation of
+	// RBD_MIRROR_IMAGE_ENABLED from librbd.
+	MirrorImageEnabled = MirrorImageState(C.RBD_MIRROR_IMAGE_ENABLED)
+	// MirrorImageDisabled is the representation of
+	// RBD_MIRROR_IMAGE_DISABLED from librbd.
+	MirrorImageDisabled = MirrorImageState(C.RBD_MIRROR_IMAGE_DISABLED)
+)
+
+// MirrorImageInfo represents the mirroring status information of a RBD image.
+type MirrorImageInfo struct {
+	GlobalID string
+	State    MirrorImageState
+	Primary  bool
+}
+
+// GetMirrorImageInfo fetches the mirroring status information of a RBD image.
+//
+// Implements:
+//  int rbd_mirror_image_get_info(rbd_image_t image,
+//                                rbd_mirror_image_info_t *mirror_image_info,
+//                                size_t info_size)
+func (image *Image) GetMirrorImageInfo() (*MirrorImageInfo, error) {
+	if err := image.validate(imageIsOpen); err != nil {
+		return nil, err
+	}
+
+	var cInfo C.rbd_mirror_image_info_t
+
+	ret := C.rbd_mirror_image_get_info(
+		image.image,
+		&cInfo,
+		C.sizeof_rbd_mirror_image_info_t)
+	if ret < 0 {
+		return nil, getError(ret)
+	}
+
+	mii := MirrorImageInfo{
+		GlobalID: C.GoString(cInfo.global_id),
+		State:    MirrorImageState(cInfo.state),
+		Primary:  bool(cInfo.primary),
+	}
+
+	// free C memory allocated by C.rbd_mirror_image_get_info call
+	C.rbd_mirror_image_get_info_cleanup(&cInfo)
+	return &mii, nil
+}
+
+// GetImageMirrorMode fetches the mirroring approach for an RBD image.
+//
+// Implements:
+//  int rbd_mirror_image_get_mode(rbd_image_t image, rbd_mirror_image_mode_t *mode);
+func (image *Image) GetImageMirrorMode() (ImageMirrorMode, error) {
+	var mode C.rbd_mirror_image_mode_t
+	if err := image.validate(imageIsOpen); err != nil {
+		return ImageMirrorMode(mode), err
+	}
+
+	ret := C.rbd_mirror_image_get_mode(image.image, &mode)
+	return ImageMirrorMode(mode), getError(ret)
+}

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

@@ -95,7 +95,11 @@ type TrashInfo struct {
 
 // cephIoctx returns a ceph rados_ioctx_t given a go-ceph rados IOContext.
 func cephIoctx(radosIoctx *rados.IOContext) C.rados_ioctx_t {
-	return C.rados_ioctx_t(radosIoctx.Pointer())
+	p := radosIoctx.Pointer()
+	if p == nil {
+		panic("invalid IOContext pointer")
+	}
+	return C.rados_ioctx_t(p)
 }
 
 // test if a bit is set in the given value

vendor/modules.txt

@@ -26,7 +26,7 @@ github.com/aws/aws-sdk-go/service/ec2
 github.com/beorn7/perks/quantile
 # github.com/blang/semver v3.5.1+incompatible
 github.com/blang/semver
-# github.com/ceph/go-ceph v0.7.0
+# github.com/ceph/go-ceph v0.8.0
 github.com/ceph/go-ceph/cephfs/admin
 github.com/ceph/go-ceph/internal/callbacks
 github.com/ceph/go-ceph/internal/cutil