go-ceph/cephfs/file.go

335 lines
9.3 KiB
Go

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"
)
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)
}
// 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)
}
// 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 and 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)
}