@@ -0,0 +1,2510 @@
1+// Copyright 2009 The Go Authors. All rights reserved.
2+// Use of this source code is governed by a BSD-style
3+// license that can be found in the LICENSE file.
4+
5+// Linux system calls.
6+// This file is compiled as ordinary Go code,
7+// but it is also input to mksyscall,
8+// which parses the //sys lines and generates system call stubs.
9+// Note that sometimes we use a lowercase //sys name and
10+// wrap it in our own nicer implementation.
11+
12+package unix
13+
14+import (
15+ "encoding/binary"
16+ "runtime"
17+ "syscall"
18+ "unsafe"
19+)
20+
21+/*
22+ * Wrapped
23+ */
24+
25+func Access(path string, mode uint32) (err error) {
26+ return Faccessat(AT_FDCWD, path, mode, 0)
27+}
28+
29+func Chmod(path string, mode uint32) (err error) {
30+ return Fchmodat(AT_FDCWD, path, mode, 0)
31+}
32+
33+func Chown(path string, uid int, gid int) (err error) {
34+ return Fchownat(AT_FDCWD, path, uid, gid, 0)
35+}
36+
37+func Creat(path string, mode uint32) (fd int, err error) {
38+ return Open(path, O_CREAT|O_WRONLY|O_TRUNC, mode)
39+}
40+
41+//sys FanotifyInit(flags uint, event_f_flags uint) (fd int, err error)
42+//sys fanotifyMark(fd int, flags uint, mask uint64, dirFd int, pathname *byte) (err error)
43+
44+func FanotifyMark(fd int, flags uint, mask uint64, dirFd int, pathname string) (err error) {
45+ if pathname == "" {
46+ return fanotifyMark(fd, flags, mask, dirFd, nil)
47+ }
48+ p, err := BytePtrFromString(pathname)
49+ if err != nil {
50+ return err
51+ }
52+ return fanotifyMark(fd, flags, mask, dirFd, p)
53+}
54+
55+//sys fchmodat(dirfd int, path string, mode uint32) (err error)
56+
57+func Fchmodat(dirfd int, path string, mode uint32, flags int) (err error) {
58+ // Linux fchmodat doesn't support the flags parameter. Mimick glibc's behavior
59+ // and check the flags. Otherwise the mode would be applied to the symlink
60+ // destination which is not what the user expects.
61+ if flags&^AT_SYMLINK_NOFOLLOW != 0 {
62+ return EINVAL
63+ } else if flags&AT_SYMLINK_NOFOLLOW != 0 {
64+ return EOPNOTSUPP
65+ }
66+ return fchmodat(dirfd, path, mode)
67+}
68+
69+//sys ioctl(fd int, req uint, arg uintptr) (err error)
70+
71+// ioctl itself should not be exposed directly, but additional get/set
72+// functions for specific types are permissible.
73+
74+// IoctlRetInt performs an ioctl operation specified by req on a device
75+// associated with opened file descriptor fd, and returns a non-negative
76+// integer that is returned by the ioctl syscall.
77+func IoctlRetInt(fd int, req uint) (int, error) {
78+ ret, _, err := Syscall(SYS_IOCTL, uintptr(fd), uintptr(req), 0)
79+ if err != 0 {
80+ return 0, err
81+ }
82+ return int(ret), nil
83+}
84+
85+func IoctlSetRTCTime(fd int, value *RTCTime) error {
86+ err := ioctl(fd, RTC_SET_TIME, uintptr(unsafe.Pointer(value)))
87+ runtime.KeepAlive(value)
88+ return err
89+}
90+
91+func IoctlSetRTCWkAlrm(fd int, value *RTCWkAlrm) error {
92+ err := ioctl(fd, RTC_WKALM_SET, uintptr(unsafe.Pointer(value)))
93+ runtime.KeepAlive(value)
94+ return err
95+}
96+
97+func IoctlGetUint32(fd int, req uint) (uint32, error) {
98+ var value uint32
99+ err := ioctl(fd, req, uintptr(unsafe.Pointer(&value)))
100+ return value, err
101+}
102+
103+func IoctlGetRTCTime(fd int) (*RTCTime, error) {
104+ var value RTCTime
105+ err := ioctl(fd, RTC_RD_TIME, uintptr(unsafe.Pointer(&value)))
106+ return &value, err
107+}
108+
109+type ifreqEthtool struct {
110+ name [IFNAMSIZ]byte
111+ data unsafe.Pointer
112+}
113+
114+// IoctlGetEthtoolDrvinfo fetches ethtool driver information for the network
115+// device specified by ifname.
116+func IoctlGetEthtoolDrvinfo(fd int, ifname string) (*EthtoolDrvinfo, error) {
117+ // Leave room for terminating NULL byte.
118+ if len(ifname) >= IFNAMSIZ {
119+ return nil, EINVAL
120+ }
121+
122+ value := EthtoolDrvinfo{
123+ Cmd: ETHTOOL_GDRVINFO,
124+ }
125+ ifreq := ifreqEthtool{
126+ data: unsafe.Pointer(&value),
127+ }
128+ copy(ifreq.name[:], ifname)
129+ err := ioctl(fd, SIOCETHTOOL, uintptr(unsafe.Pointer(&ifreq)))
130+ runtime.KeepAlive(ifreq)
131+ return &value, err
132+}
133+
134+// IoctlGetWatchdogInfo fetches information about a watchdog device from the
135+// Linux watchdog API. For more information, see:
136+// https://www.kernel.org/doc/html/latest/watchdog/watchdog-api.html.
137+func IoctlGetWatchdogInfo(fd int) (*WatchdogInfo, error) {
138+ var value WatchdogInfo
139+ err := ioctl(fd, WDIOC_GETSUPPORT, uintptr(unsafe.Pointer(&value)))
140+ return &value, err
141+}
142+
143+func IoctlGetRTCWkAlrm(fd int) (*RTCWkAlrm, error) {
144+ var value RTCWkAlrm
145+ err := ioctl(fd, RTC_WKALM_RD, uintptr(unsafe.Pointer(&value)))
146+ return &value, err
147+}
148+
149+// IoctlFileCloneRange performs an FICLONERANGE ioctl operation to clone the
150+// range of data conveyed in value to the file associated with the file
151+// descriptor destFd. See the ioctl_ficlonerange(2) man page for details.
152+func IoctlFileCloneRange(destFd int, value *FileCloneRange) error {
153+ err := ioctl(destFd, FICLONERANGE, uintptr(unsafe.Pointer(value)))
154+ runtime.KeepAlive(value)
155+ return err
156+}
157+
158+// IoctlFileClone performs an FICLONE ioctl operation to clone the entire file
159+// associated with the file description srcFd to the file associated with the
160+// file descriptor destFd. See the ioctl_ficlone(2) man page for details.
161+func IoctlFileClone(destFd, srcFd int) error {
162+ return ioctl(destFd, FICLONE, uintptr(srcFd))
163+}
164+
165+type FileDedupeRange struct {
166+ Src_offset uint64
167+ Src_length uint64
168+ Reserved1 uint16
169+ Reserved2 uint32
170+ Info []FileDedupeRangeInfo
171+}
172+
173+type FileDedupeRangeInfo struct {
174+ Dest_fd int64
175+ Dest_offset uint64
176+ Bytes_deduped uint64
177+ Status int32
178+ Reserved uint32
179+}
180+
181+// IoctlFileDedupeRange performs an FIDEDUPERANGE ioctl operation to share the
182+// range of data conveyed in value from the file associated with the file
183+// descriptor srcFd to the value.Info destinations. See the
184+// ioctl_fideduperange(2) man page for details.
185+func IoctlFileDedupeRange(srcFd int, value *FileDedupeRange) error {
186+ buf := make([]byte, SizeofRawFileDedupeRange+
187+ len(value.Info)*SizeofRawFileDedupeRangeInfo)
188+ rawrange := (*RawFileDedupeRange)(unsafe.Pointer(&buf[0]))
189+ rawrange.Src_offset = value.Src_offset
190+ rawrange.Src_length = value.Src_length
191+ rawrange.Dest_count = uint16(len(value.Info))
192+ rawrange.Reserved1 = value.Reserved1
193+ rawrange.Reserved2 = value.Reserved2
194+
195+ for i := range value.Info {
196+ rawinfo := (*RawFileDedupeRangeInfo)(unsafe.Pointer(
197+ uintptr(unsafe.Pointer(&buf[0])) + uintptr(SizeofRawFileDedupeRange) +
198+ uintptr(i*SizeofRawFileDedupeRangeInfo)))
199+ rawinfo.Dest_fd = value.Info[i].Dest_fd
200+ rawinfo.Dest_offset = value.Info[i].Dest_offset
201+ rawinfo.Bytes_deduped = value.Info[i].Bytes_deduped
202+ rawinfo.Status = value.Info[i].Status
203+ rawinfo.Reserved = value.Info[i].Reserved
204+ }
205+
206+ err := ioctl(srcFd, FIDEDUPERANGE, uintptr(unsafe.Pointer(&buf[0])))
207+
208+ // Output
209+ for i := range value.Info {
210+ rawinfo := (*RawFileDedupeRangeInfo)(unsafe.Pointer(
211+ uintptr(unsafe.Pointer(&buf[0])) + uintptr(SizeofRawFileDedupeRange) +
212+ uintptr(i*SizeofRawFileDedupeRangeInfo)))
213+ value.Info[i].Dest_fd = rawinfo.Dest_fd
214+ value.Info[i].Dest_offset = rawinfo.Dest_offset
215+ value.Info[i].Bytes_deduped = rawinfo.Bytes_deduped
216+ value.Info[i].Status = rawinfo.Status
217+ value.Info[i].Reserved = rawinfo.Reserved
218+ }
219+
220+ return err
221+}
222+
223+// IoctlWatchdogKeepalive issues a keepalive ioctl to a watchdog device. For
224+// more information, see:
225+// https://www.kernel.org/doc/html/latest/watchdog/watchdog-api.html.
226+func IoctlWatchdogKeepalive(fd int) error {
227+ return ioctl(fd, WDIOC_KEEPALIVE, 0)
228+}
229+
230+func IoctlHIDGetDesc(fd int, value *HIDRawReportDescriptor) error {
231+ err := ioctl(fd, HIDIOCGRDESC, uintptr(unsafe.Pointer(value)))
232+ runtime.KeepAlive(value)
233+ return err
234+}
235+
236+func IoctlHIDGetRawInfo(fd int) (*HIDRawDevInfo, error) {
237+ var value HIDRawDevInfo
238+ err := ioctl(fd, HIDIOCGRAWINFO, uintptr(unsafe.Pointer(&value)))
239+ return &value, err
240+}
241+
242+func IoctlHIDGetRawName(fd int) (string, error) {
243+ var value [_HIDIOCGRAWNAME_LEN]byte
244+ err := ioctl(fd, _HIDIOCGRAWNAME, uintptr(unsafe.Pointer(&value[0])))
245+ return ByteSliceToString(value[:]), err
246+}
247+
248+func IoctlHIDGetRawPhys(fd int) (string, error) {
249+ var value [_HIDIOCGRAWPHYS_LEN]byte
250+ err := ioctl(fd, _HIDIOCGRAWPHYS, uintptr(unsafe.Pointer(&value[0])))
251+ return ByteSliceToString(value[:]), err
252+}
253+
254+func IoctlHIDGetRawUniq(fd int) (string, error) {
255+ var value [_HIDIOCGRAWUNIQ_LEN]byte
256+ err := ioctl(fd, _HIDIOCGRAWUNIQ, uintptr(unsafe.Pointer(&value[0])))
257+ return ByteSliceToString(value[:]), err
258+}
259+
260+//sys Linkat(olddirfd int, oldpath string, newdirfd int, newpath string, flags int) (err error)
261+
262+func Link(oldpath string, newpath string) (err error) {
263+ return Linkat(AT_FDCWD, oldpath, AT_FDCWD, newpath, 0)
264+}
265+
266+func Mkdir(path string, mode uint32) (err error) {
267+ return Mkdirat(AT_FDCWD, path, mode)
268+}
269+
270+func Mknod(path string, mode uint32, dev int) (err error) {
271+ return Mknodat(AT_FDCWD, path, mode, dev)
272+}
273+
274+func Open(path string, mode int, perm uint32) (fd int, err error) {
275+ return openat(AT_FDCWD, path, mode|O_LARGEFILE, perm)
276+}
277+
278+//sys openat(dirfd int, path string, flags int, mode uint32) (fd int, err error)
279+
280+func Openat(dirfd int, path string, flags int, mode uint32) (fd int, err error) {
281+ return openat(dirfd, path, flags|O_LARGEFILE, mode)
282+}
283+
284+//sys openat2(dirfd int, path string, open_how *OpenHow, size int) (fd int, err error)
285+
286+func Openat2(dirfd int, path string, how *OpenHow) (fd int, err error) {
287+ return openat2(dirfd, path, how, SizeofOpenHow)
288+}
289+
290+//sys ppoll(fds *PollFd, nfds int, timeout *Timespec, sigmask *Sigset_t) (n int, err error)
291+
292+func Ppoll(fds []PollFd, timeout *Timespec, sigmask *Sigset_t) (n int, err error) {
293+ if len(fds) == 0 {
294+ return ppoll(nil, 0, timeout, sigmask)
295+ }
296+ return ppoll(&fds[0], len(fds), timeout, sigmask)
297+}
298+
299+//sys Readlinkat(dirfd int, path string, buf []byte) (n int, err error)
300+
301+func Readlink(path string, buf []byte) (n int, err error) {
302+ return Readlinkat(AT_FDCWD, path, buf)
303+}
304+
305+func Rename(oldpath string, newpath string) (err error) {
306+ return Renameat(AT_FDCWD, oldpath, AT_FDCWD, newpath)
307+}
308+
309+func Rmdir(path string) error {
310+ return Unlinkat(AT_FDCWD, path, AT_REMOVEDIR)
311+}
312+
313+//sys Symlinkat(oldpath string, newdirfd int, newpath string) (err error)
314+
315+func Symlink(oldpath string, newpath string) (err error) {
316+ return Symlinkat(oldpath, AT_FDCWD, newpath)
317+}
318+
319+func Unlink(path string) error {
320+ return Unlinkat(AT_FDCWD, path, 0)
321+}
322+
323+//sys Unlinkat(dirfd int, path string, flags int) (err error)
324+
325+func Utimes(path string, tv []Timeval) error {
326+ if tv == nil {
327+ err := utimensat(AT_FDCWD, path, nil, 0)
328+ if err != ENOSYS {
329+ return err
330+ }
331+ return utimes(path, nil)
332+ }
333+ if len(tv) != 2 {
334+ return EINVAL
335+ }
336+ var ts [2]Timespec
337+ ts[0] = NsecToTimespec(TimevalToNsec(tv[0]))
338+ ts[1] = NsecToTimespec(TimevalToNsec(tv[1]))
339+ err := utimensat(AT_FDCWD, path, (*[2]Timespec)(unsafe.Pointer(&ts[0])), 0)
340+ if err != ENOSYS {
341+ return err
342+ }
343+ return utimes(path, (*[2]Timeval)(unsafe.Pointer(&tv[0])))
344+}
345+
346+//sys utimensat(dirfd int, path string, times *[2]Timespec, flags int) (err error)
347+
348+func UtimesNano(path string, ts []Timespec) error {
349+ if ts == nil {
350+ err := utimensat(AT_FDCWD, path, nil, 0)
351+ if err != ENOSYS {
352+ return err
353+ }
354+ return utimes(path, nil)
355+ }
356+ if len(ts) != 2 {
357+ return EINVAL
358+ }
359+ err := utimensat(AT_FDCWD, path, (*[2]Timespec)(unsafe.Pointer(&ts[0])), 0)
360+ if err != ENOSYS {
361+ return err
362+ }
363+ // If the utimensat syscall isn't available (utimensat was added to Linux
364+ // in 2.6.22, Released, 8 July 2007) then fall back to utimes
365+ var tv [2]Timeval
366+ for i := 0; i < 2; i++ {
367+ tv[i] = NsecToTimeval(TimespecToNsec(ts[i]))
368+ }
369+ return utimes(path, (*[2]Timeval)(unsafe.Pointer(&tv[0])))
370+}
371+
372+func UtimesNanoAt(dirfd int, path string, ts []Timespec, flags int) error {
373+ if ts == nil {
374+ return utimensat(dirfd, path, nil, flags)
375+ }
376+ if len(ts) != 2 {
377+ return EINVAL
378+ }
379+ return utimensat(dirfd, path, (*[2]Timespec)(unsafe.Pointer(&ts[0])), flags)
380+}
381+
382+func Futimesat(dirfd int, path string, tv []Timeval) error {
383+ if tv == nil {
384+ return futimesat(dirfd, path, nil)
385+ }
386+ if len(tv) != 2 {
387+ return EINVAL
388+ }
389+ return futimesat(dirfd, path, (*[2]Timeval)(unsafe.Pointer(&tv[0])))
390+}
391+
392+func Futimes(fd int, tv []Timeval) (err error) {
393+ // Believe it or not, this is the best we can do on Linux
394+ // (and is what glibc does).
395+ return Utimes("/proc/self/fd/"+itoa(fd), tv)
396+}
397+
398+const ImplementsGetwd = true
399+
400+//sys Getcwd(buf []byte) (n int, err error)
401+
402+func Getwd() (wd string, err error) {
403+ var buf [PathMax]byte
404+ n, err := Getcwd(buf[0:])
405+ if err != nil {
406+ return "", err
407+ }
408+ // Getcwd returns the number of bytes written to buf, including the NUL.
409+ if n < 1 || n > len(buf) || buf[n-1] != 0 {
410+ return "", EINVAL
411+ }
412+ return string(buf[0 : n-1]), nil
413+}
414+
415+func Getgroups() (gids []int, err error) {
416+ n, err := getgroups(0, nil)
417+ if err != nil {
418+ return nil, err
419+ }
420+ if n == 0 {
421+ return nil, nil
422+ }
423+
424+ // Sanity check group count. Max is 1<<16 on Linux.
425+ if n < 0 || n > 1<<20 {
426+ return nil, EINVAL
427+ }
428+
429+ a := make([]_Gid_t, n)
430+ n, err = getgroups(n, &a[0])
431+ if err != nil {
432+ return nil, err
433+ }
434+ gids = make([]int, n)
435+ for i, v := range a[0:n] {
436+ gids[i] = int(v)
437+ }
438+ return
439+}
440+
441+func Setgroups(gids []int) (err error) {
442+ if len(gids) == 0 {
443+ return setgroups(0, nil)
444+ }
445+
446+ a := make([]_Gid_t, len(gids))
447+ for i, v := range gids {
448+ a[i] = _Gid_t(v)
449+ }
450+ return setgroups(len(a), &a[0])
451+}
452+
453+type WaitStatus uint32
454+
455+// Wait status is 7 bits at bottom, either 0 (exited),
456+// 0x7F (stopped), or a signal number that caused an exit.
457+// The 0x80 bit is whether there was a core dump.
458+// An extra number (exit code, signal causing a stop)
459+// is in the high bits. At least that's the idea.
460+// There are various irregularities. For example, the
461+// "continued" status is 0xFFFF, distinguishing itself
462+// from stopped via the core dump bit.
463+
464+const (
465+ mask = 0x7F
466+ core = 0x80
467+ exited = 0x00
468+ stopped = 0x7F
469+ shift = 8
470+)
471+
472+func (w WaitStatus) Exited() bool { return w&mask == exited }
473+
474+func (w WaitStatus) Signaled() bool { return w&mask != stopped && w&mask != exited }
475+
476+func (w WaitStatus) Stopped() bool { return w&0xFF == stopped }
477+
478+func (w WaitStatus) Continued() bool { return w == 0xFFFF }
479+
480+func (w WaitStatus) CoreDump() bool { return w.Signaled() && w&core != 0 }
481+
482+func (w WaitStatus) ExitStatus() int {
483+ if !w.Exited() {
484+ return -1
485+ }
486+ return int(w>>shift) & 0xFF
487+}
488+
489+func (w WaitStatus) Signal() syscall.Signal {
490+ if !w.Signaled() {
491+ return -1
492+ }
493+ return syscall.Signal(w & mask)
494+}
495+
496+func (w WaitStatus) StopSignal() syscall.Signal {
497+ if !w.Stopped() {
498+ return -1
499+ }
500+ return syscall.Signal(w>>shift) & 0xFF
501+}
502+
503+func (w WaitStatus) TrapCause() int {
504+ if w.StopSignal() != SIGTRAP {
505+ return -1
506+ }
507+ return int(w>>shift) >> 8
508+}
509+
510+//sys wait4(pid int, wstatus *_C_int, options int, rusage *Rusage) (wpid int, err error)
511+
512+func Wait4(pid int, wstatus *WaitStatus, options int, rusage *Rusage) (wpid int, err error) {
513+ var status _C_int
514+ wpid, err = wait4(pid, &status, options, rusage)
515+ if wstatus != nil {
516+ *wstatus = WaitStatus(status)
517+ }
518+ return
519+}
520+
521+func Mkfifo(path string, mode uint32) error {
522+ return Mknod(path, mode|S_IFIFO, 0)
523+}
524+
525+func Mkfifoat(dirfd int, path string, mode uint32) error {
526+ return Mknodat(dirfd, path, mode|S_IFIFO, 0)
527+}
528+
529+func (sa *SockaddrInet4) sockaddr() (unsafe.Pointer, _Socklen, error) {
530+ if sa.Port < 0 || sa.Port > 0xFFFF {
531+ return nil, 0, EINVAL
532+ }
533+ sa.raw.Family = AF_INET
534+ p := (*[2]byte)(unsafe.Pointer(&sa.raw.Port))
535+ p[0] = byte(sa.Port >> 8)
536+ p[1] = byte(sa.Port)
537+ for i := 0; i < len(sa.Addr); i++ {
538+ sa.raw.Addr[i] = sa.Addr[i]
539+ }
540+ return unsafe.Pointer(&sa.raw), SizeofSockaddrInet4, nil
541+}
542+
543+func (sa *SockaddrInet6) sockaddr() (unsafe.Pointer, _Socklen, error) {
544+ if sa.Port < 0 || sa.Port > 0xFFFF {
545+ return nil, 0, EINVAL
546+ }
547+ sa.raw.Family = AF_INET6
548+ p := (*[2]byte)(unsafe.Pointer(&sa.raw.Port))
549+ p[0] = byte(sa.Port >> 8)
550+ p[1] = byte(sa.Port)
551+ sa.raw.Scope_id = sa.ZoneId
552+ for i := 0; i < len(sa.Addr); i++ {
553+ sa.raw.Addr[i] = sa.Addr[i]
554+ }
555+ return unsafe.Pointer(&sa.raw), SizeofSockaddrInet6, nil
556+}
557+
558+func (sa *SockaddrUnix) sockaddr() (unsafe.Pointer, _Socklen, error) {
559+ name := sa.Name
560+ n := len(name)
561+ if n >= len(sa.raw.Path) {
562+ return nil, 0, EINVAL
563+ }
564+ sa.raw.Family = AF_UNIX
565+ for i := 0; i < n; i++ {
566+ sa.raw.Path[i] = int8(name[i])
567+ }
568+ // length is family (uint16), name, NUL.
569+ sl := _Socklen(2)
570+ if n > 0 {
571+ sl += _Socklen(n) + 1
572+ }
573+ if sa.raw.Path[0] == '@' {
574+ sa.raw.Path[0] = 0
575+ // Don't count trailing NUL for abstract address.
576+ sl--
577+ }
578+
579+ return unsafe.Pointer(&sa.raw), sl, nil
580+}
581+
582+// SockaddrLinklayer implements the Sockaddr interface for AF_PACKET type sockets.
583+type SockaddrLinklayer struct {
584+ Protocol uint16
585+ Ifindex int
586+ Hatype uint16
587+ Pkttype uint8
588+ Halen uint8
589+ Addr [8]byte
590+ raw RawSockaddrLinklayer
591+}
592+
593+func (sa *SockaddrLinklayer) sockaddr() (unsafe.Pointer, _Socklen, error) {
594+ if sa.Ifindex < 0 || sa.Ifindex > 0x7fffffff {
595+ return nil, 0, EINVAL
596+ }
597+ sa.raw.Family = AF_PACKET
598+ sa.raw.Protocol = sa.Protocol
599+ sa.raw.Ifindex = int32(sa.Ifindex)
600+ sa.raw.Hatype = sa.Hatype
601+ sa.raw.Pkttype = sa.Pkttype
602+ sa.raw.Halen = sa.Halen
603+ for i := 0; i < len(sa.Addr); i++ {
604+ sa.raw.Addr[i] = sa.Addr[i]
605+ }
606+ return unsafe.Pointer(&sa.raw), SizeofSockaddrLinklayer, nil
607+}
608+
609+// SockaddrNetlink implements the Sockaddr interface for AF_NETLINK type sockets.
610+type SockaddrNetlink struct {
611+ Family uint16
612+ Pad uint16
613+ Pid uint32
614+ Groups uint32
615+ raw RawSockaddrNetlink
616+}
617+
618+func (sa *SockaddrNetlink) sockaddr() (unsafe.Pointer, _Socklen, error) {
619+ sa.raw.Family = AF_NETLINK
620+ sa.raw.Pad = sa.Pad
621+ sa.raw.Pid = sa.Pid
622+ sa.raw.Groups = sa.Groups
623+ return unsafe.Pointer(&sa.raw), SizeofSockaddrNetlink, nil
624+}
625+
626+// SockaddrHCI implements the Sockaddr interface for AF_BLUETOOTH type sockets
627+// using the HCI protocol.
628+type SockaddrHCI struct {
629+ Dev uint16
630+ Channel uint16
631+ raw RawSockaddrHCI
632+}
633+
634+func (sa *SockaddrHCI) sockaddr() (unsafe.Pointer, _Socklen, error) {
635+ sa.raw.Family = AF_BLUETOOTH
636+ sa.raw.Dev = sa.Dev
637+ sa.raw.Channel = sa.Channel
638+ return unsafe.Pointer(&sa.raw), SizeofSockaddrHCI, nil
639+}
640+
641+// SockaddrL2 implements the Sockaddr interface for AF_BLUETOOTH type sockets
642+// using the L2CAP protocol.
643+type SockaddrL2 struct {
644+ PSM uint16
645+ CID uint16
646+ Addr [6]uint8
647+ AddrType uint8
648+ raw RawSockaddrL2
649+}
650+
651+func (sa *SockaddrL2) sockaddr() (unsafe.Pointer, _Socklen, error) {
652+ sa.raw.Family = AF_BLUETOOTH
653+ psm := (*[2]byte)(unsafe.Pointer(&sa.raw.Psm))
654+ psm[0] = byte(sa.PSM)
655+ psm[1] = byte(sa.PSM >> 8)
656+ for i := 0; i < len(sa.Addr); i++ {
657+ sa.raw.Bdaddr[i] = sa.Addr[len(sa.Addr)-1-i]
658+ }
659+ cid := (*[2]byte)(unsafe.Pointer(&sa.raw.Cid))
660+ cid[0] = byte(sa.CID)
661+ cid[1] = byte(sa.CID >> 8)
662+ sa.raw.Bdaddr_type = sa.AddrType
663+ return unsafe.Pointer(&sa.raw), SizeofSockaddrL2, nil
664+}
665+
666+// SockaddrRFCOMM implements the Sockaddr interface for AF_BLUETOOTH type sockets
667+// using the RFCOMM protocol.
668+//
669+// Server example:
670+//
671+// fd, _ := Socket(AF_BLUETOOTH, SOCK_STREAM, BTPROTO_RFCOMM)
672+// _ = unix.Bind(fd, &unix.SockaddrRFCOMM{
673+// Channel: 1,
674+// Addr: [6]uint8{0, 0, 0, 0, 0, 0}, // BDADDR_ANY or 00:00:00:00:00:00
675+// })
676+// _ = Listen(fd, 1)
677+// nfd, sa, _ := Accept(fd)
678+// fmt.Printf("conn addr=%v fd=%d", sa.(*unix.SockaddrRFCOMM).Addr, nfd)
679+// Read(nfd, buf)
680+//
681+// Client example:
682+//
683+// fd, _ := Socket(AF_BLUETOOTH, SOCK_STREAM, BTPROTO_RFCOMM)
684+// _ = Connect(fd, &SockaddrRFCOMM{
685+// Channel: 1,
686+// Addr: [6]byte{0x11, 0x22, 0x33, 0xaa, 0xbb, 0xcc}, // CC:BB:AA:33:22:11
687+// })
688+// Write(fd, []byte(`hello`))
689+type SockaddrRFCOMM struct {
690+ // Addr represents a bluetooth address, byte ordering is little-endian.
691+ Addr [6]uint8
692+
693+ // Channel is a designated bluetooth channel, only 1-30 are available for use.
694+ // Since Linux 2.6.7 and further zero value is the first available channel.
695+ Channel uint8
696+
697+ raw RawSockaddrRFCOMM
698+}
699+
700+func (sa *SockaddrRFCOMM) sockaddr() (unsafe.Pointer, _Socklen, error) {
701+ sa.raw.Family = AF_BLUETOOTH
702+ sa.raw.Channel = sa.Channel
703+ sa.raw.Bdaddr = sa.Addr
704+ return unsafe.Pointer(&sa.raw), SizeofSockaddrRFCOMM, nil
705+}
706+
707+// SockaddrCAN implements the Sockaddr interface for AF_CAN type sockets.
708+// The RxID and TxID fields are used for transport protocol addressing in
709+// (CAN_TP16, CAN_TP20, CAN_MCNET, and CAN_ISOTP), they can be left with
710+// zero values for CAN_RAW and CAN_BCM sockets as they have no meaning.
711+//
712+// The SockaddrCAN struct must be bound to the socket file descriptor
713+// using Bind before the CAN socket can be used.
714+//
715+// // Read one raw CAN frame
716+// fd, _ := Socket(AF_CAN, SOCK_RAW, CAN_RAW)
717+// addr := &SockaddrCAN{Ifindex: index}
718+// Bind(fd, addr)
719+// frame := make([]byte, 16)
720+// Read(fd, frame)
721+//
722+// The full SocketCAN documentation can be found in the linux kernel
723+// archives at: https://www.kernel.org/doc/Documentation/networking/can.txt
724+type SockaddrCAN struct {
725+ Ifindex int
726+ RxID uint32
727+ TxID uint32
728+ raw RawSockaddrCAN
729+}
730+
731+func (sa *SockaddrCAN) sockaddr() (unsafe.Pointer, _Socklen, error) {
732+ if sa.Ifindex < 0 || sa.Ifindex > 0x7fffffff {
733+ return nil, 0, EINVAL
734+ }
735+ sa.raw.Family = AF_CAN
736+ sa.raw.Ifindex = int32(sa.Ifindex)
737+ rx := (*[4]byte)(unsafe.Pointer(&sa.RxID))
738+ for i := 0; i < 4; i++ {
739+ sa.raw.Addr[i] = rx[i]
740+ }
741+ tx := (*[4]byte)(unsafe.Pointer(&sa.TxID))
742+ for i := 0; i < 4; i++ {
743+ sa.raw.Addr[i+4] = tx[i]
744+ }
745+ return unsafe.Pointer(&sa.raw), SizeofSockaddrCAN, nil
746+}
747+
748+// SockaddrCANJ1939 implements the Sockaddr interface for AF_CAN using J1939
749+// protocol (https://en.wikipedia.org/wiki/SAE_J1939). For more information
750+// on the purposes of the fields, check the official linux kernel documentation
751+// available here: https://www.kernel.org/doc/Documentation/networking/j1939.rst
752+type SockaddrCANJ1939 struct {
753+ Ifindex int
754+ Name uint64
755+ PGN uint32
756+ Addr uint8
757+ raw RawSockaddrCAN
758+}
759+
760+func (sa *SockaddrCANJ1939) sockaddr() (unsafe.Pointer, _Socklen, error) {
761+ if sa.Ifindex < 0 || sa.Ifindex > 0x7fffffff {
762+ return nil, 0, EINVAL
763+ }
764+ sa.raw.Family = AF_CAN
765+ sa.raw.Ifindex = int32(sa.Ifindex)
766+ n := (*[8]byte)(unsafe.Pointer(&sa.Name))
767+ for i := 0; i < 8; i++ {
768+ sa.raw.Addr[i] = n[i]
769+ }
770+ p := (*[4]byte)(unsafe.Pointer(&sa.PGN))
771+ for i := 0; i < 4; i++ {
772+ sa.raw.Addr[i+8] = p[i]
773+ }
774+ sa.raw.Addr[12] = sa.Addr
775+ return unsafe.Pointer(&sa.raw), SizeofSockaddrCAN, nil
776+}
777+
778+// SockaddrALG implements the Sockaddr interface for AF_ALG type sockets.
779+// SockaddrALG enables userspace access to the Linux kernel's cryptography
780+// subsystem. The Type and Name fields specify which type of hash or cipher
781+// should be used with a given socket.
782+//
783+// To create a file descriptor that provides access to a hash or cipher, both
784+// Bind and Accept must be used. Once the setup process is complete, input
785+// data can be written to the socket, processed by the kernel, and then read
786+// back as hash output or ciphertext.
787+//
788+// Here is an example of using an AF_ALG socket with SHA1 hashing.
789+// The initial socket setup process is as follows:
790+//
791+// // Open a socket to perform SHA1 hashing.
792+// fd, _ := unix.Socket(unix.AF_ALG, unix.SOCK_SEQPACKET, 0)
793+// addr := &unix.SockaddrALG{Type: "hash", Name: "sha1"}
794+// unix.Bind(fd, addr)
795+// // Note: unix.Accept does not work at this time; must invoke accept()
796+// // manually using unix.Syscall.
797+// hashfd, _, _ := unix.Syscall(unix.SYS_ACCEPT, uintptr(fd), 0, 0)
798+//
799+// Once a file descriptor has been returned from Accept, it may be used to
800+// perform SHA1 hashing. The descriptor is not safe for concurrent use, but
801+// may be re-used repeatedly with subsequent Write and Read operations.
802+//
803+// When hashing a small byte slice or string, a single Write and Read may
804+// be used:
805+//
806+// // Assume hashfd is already configured using the setup process.
807+// hash := os.NewFile(hashfd, "sha1")
808+// // Hash an input string and read the results. Each Write discards
809+// // previous hash state. Read always reads the current state.
810+// b := make([]byte, 20)
811+// for i := 0; i < 2; i++ {
812+// io.WriteString(hash, "Hello, world.")
813+// hash.Read(b)
814+// fmt.Println(hex.EncodeToString(b))
815+// }
816+// // Output:
817+// // 2ae01472317d1935a84797ec1983ae243fc6aa28
818+// // 2ae01472317d1935a84797ec1983ae243fc6aa28
819+//
820+// For hashing larger byte slices, or byte streams such as those read from
821+// a file or socket, use Sendto with MSG_MORE to instruct the kernel to update
822+// the hash digest instead of creating a new one for a given chunk and finalizing it.
823+//
824+// // Assume hashfd and addr are already configured using the setup process.
825+// hash := os.NewFile(hashfd, "sha1")
826+// // Hash the contents of a file.
827+// f, _ := os.Open("/tmp/linux-4.10-rc7.tar.xz")
828+// b := make([]byte, 4096)
829+// for {
830+// n, err := f.Read(b)
831+// if err == io.EOF {
832+// break
833+// }
834+// unix.Sendto(hashfd, b[:n], unix.MSG_MORE, addr)
835+// }
836+// hash.Read(b)
837+// fmt.Println(hex.EncodeToString(b))
838+// // Output: 85cdcad0c06eef66f805ecce353bec9accbeecc5
839+//
840+// For more information, see: http://www.chronox.de/crypto-API/crypto/userspace-if.html.
841+type SockaddrALG struct {
842+ Type string
843+ Name string
844+ Feature uint32
845+ Mask uint32
846+ raw RawSockaddrALG
847+}
848+
849+func (sa *SockaddrALG) sockaddr() (unsafe.Pointer, _Socklen, error) {
850+ // Leave room for NUL byte terminator.
851+ if len(sa.Type) > 13 {
852+ return nil, 0, EINVAL
853+ }
854+ if len(sa.Name) > 63 {
855+ return nil, 0, EINVAL
856+ }
857+
858+ sa.raw.Family = AF_ALG
859+ sa.raw.Feat = sa.Feature
860+ sa.raw.Mask = sa.Mask
861+
862+ typ, err := ByteSliceFromString(sa.Type)
863+ if err != nil {
864+ return nil, 0, err
865+ }
866+ name, err := ByteSliceFromString(sa.Name)
867+ if err != nil {
868+ return nil, 0, err
869+ }
870+
871+ copy(sa.raw.Type[:], typ)
872+ copy(sa.raw.Name[:], name)
873+
874+ return unsafe.Pointer(&sa.raw), SizeofSockaddrALG, nil
875+}
876+
877+// SockaddrVM implements the Sockaddr interface for AF_VSOCK type sockets.
878+// SockaddrVM provides access to Linux VM sockets: a mechanism that enables
879+// bidirectional communication between a hypervisor and its guest virtual
880+// machines.
881+type SockaddrVM struct {
882+ // CID and Port specify a context ID and port address for a VM socket.
883+ // Guests have a unique CID, and hosts may have a well-known CID of:
884+ // - VMADDR_CID_HYPERVISOR: refers to the hypervisor process.
885+ // - VMADDR_CID_LOCAL: refers to local communication (loopback).
886+ // - VMADDR_CID_HOST: refers to other processes on the host.
887+ CID uint32
888+ Port uint32
889+ Flags uint8
890+ raw RawSockaddrVM
891+}
892+
893+func (sa *SockaddrVM) sockaddr() (unsafe.Pointer, _Socklen, error) {
894+ sa.raw.Family = AF_VSOCK
895+ sa.raw.Port = sa.Port
896+ sa.raw.Cid = sa.CID
897+ sa.raw.Flags = sa.Flags
898+
899+ return unsafe.Pointer(&sa.raw), SizeofSockaddrVM, nil
900+}
901+
902+type SockaddrXDP struct {
903+ Flags uint16
904+ Ifindex uint32
905+ QueueID uint32
906+ SharedUmemFD uint32
907+ raw RawSockaddrXDP
908+}
909+
910+func (sa *SockaddrXDP) sockaddr() (unsafe.Pointer, _Socklen, error) {
911+ sa.raw.Family = AF_XDP
912+ sa.raw.Flags = sa.Flags
913+ sa.raw.Ifindex = sa.Ifindex
914+ sa.raw.Queue_id = sa.QueueID
915+ sa.raw.Shared_umem_fd = sa.SharedUmemFD
916+
917+ return unsafe.Pointer(&sa.raw), SizeofSockaddrXDP, nil
918+}
919+
920+// This constant mirrors the #define of PX_PROTO_OE in
921+// linux/if_pppox.h. We're defining this by hand here instead of
922+// autogenerating through mkerrors.sh because including
923+// linux/if_pppox.h causes some declaration conflicts with other
924+// includes (linux/if_pppox.h includes linux/in.h, which conflicts
925+// with netinet/in.h). Given that we only need a single zero constant
926+// out of that file, it's cleaner to just define it by hand here.
927+const px_proto_oe = 0
928+
929+type SockaddrPPPoE struct {
930+ SID uint16
931+ Remote []byte
932+ Dev string
933+ raw RawSockaddrPPPoX
934+}
935+
936+func (sa *SockaddrPPPoE) sockaddr() (unsafe.Pointer, _Socklen, error) {
937+ if len(sa.Remote) != 6 {
938+ return nil, 0, EINVAL
939+ }
940+ if len(sa.Dev) > IFNAMSIZ-1 {
941+ return nil, 0, EINVAL
942+ }
943+
944+ *(*uint16)(unsafe.Pointer(&sa.raw[0])) = AF_PPPOX
945+ // This next field is in host-endian byte order. We can't use the
946+ // same unsafe pointer cast as above, because this value is not
947+ // 32-bit aligned and some architectures don't allow unaligned
948+ // access.
949+ //
950+ // However, the value of px_proto_oe is 0, so we can use
951+ // encoding/binary helpers to write the bytes without worrying
952+ // about the ordering.
953+ binary.BigEndian.PutUint32(sa.raw[2:6], px_proto_oe)
954+ // This field is deliberately big-endian, unlike the previous
955+ // one. The kernel expects SID to be in network byte order.
956+ binary.BigEndian.PutUint16(sa.raw[6:8], sa.SID)
957+ copy(sa.raw[8:14], sa.Remote)
958+ for i := 14; i < 14+IFNAMSIZ; i++ {
959+ sa.raw[i] = 0
960+ }
961+ copy(sa.raw[14:], sa.Dev)
962+ return unsafe.Pointer(&sa.raw), SizeofSockaddrPPPoX, nil
963+}
964+
965+// SockaddrTIPC implements the Sockaddr interface for AF_TIPC type sockets.
966+// For more information on TIPC, see: http://tipc.sourceforge.net/.
967+type SockaddrTIPC struct {
968+ // Scope is the publication scopes when binding service/service range.
969+ // Should be set to TIPC_CLUSTER_SCOPE or TIPC_NODE_SCOPE.
970+ Scope int
971+
972+ // Addr is the type of address used to manipulate a socket. Addr must be
973+ // one of:
974+ // - *TIPCSocketAddr: "id" variant in the C addr union
975+ // - *TIPCServiceRange: "nameseq" variant in the C addr union
976+ // - *TIPCServiceName: "name" variant in the C addr union
977+ //
978+ // If nil, EINVAL will be returned when the structure is used.
979+ Addr TIPCAddr
980+
981+ raw RawSockaddrTIPC
982+}
983+
984+// TIPCAddr is implemented by types that can be used as an address for
985+// SockaddrTIPC. It is only implemented by *TIPCSocketAddr, *TIPCServiceRange,
986+// and *TIPCServiceName.
987+type TIPCAddr interface {
988+ tipcAddrtype() uint8
989+ tipcAddr() [12]byte
990+}
991+
992+func (sa *TIPCSocketAddr) tipcAddr() [12]byte {
993+ var out [12]byte
994+ copy(out[:], (*(*[unsafe.Sizeof(TIPCSocketAddr{})]byte)(unsafe.Pointer(sa)))[:])
995+ return out
996+}
997+
998+func (sa *TIPCSocketAddr) tipcAddrtype() uint8 { return TIPC_SOCKET_ADDR }
999+
1000+func (sa *TIPCServiceRange) tipcAddr() [12]byte {
1001+ var out [12]byte
1002+ copy(out[:], (*(*[unsafe.Sizeof(TIPCServiceRange{})]byte)(unsafe.Pointer(sa)))[:])
1003+ return out
1004+}
1005+
1006+func (sa *TIPCServiceRange) tipcAddrtype() uint8 { return TIPC_SERVICE_RANGE }
1007+
1008+func (sa *TIPCServiceName) tipcAddr() [12]byte {
1009+ var out [12]byte
1010+ copy(out[:], (*(*[unsafe.Sizeof(TIPCServiceName{})]byte)(unsafe.Pointer(sa)))[:])
1011+ return out
1012+}
1013+
1014+func (sa *TIPCServiceName) tipcAddrtype() uint8 { return TIPC_SERVICE_ADDR }
1015+
1016+func (sa *SockaddrTIPC) sockaddr() (unsafe.Pointer, _Socklen, error) {
1017+ if sa.Addr == nil {
1018+ return nil, 0, EINVAL
1019+ }
1020+
1021+ sa.raw.Family = AF_TIPC
1022+ sa.raw.Scope = int8(sa.Scope)
1023+ sa.raw.Addrtype = sa.Addr.tipcAddrtype()
1024+ sa.raw.Addr = sa.Addr.tipcAddr()
1025+
1026+ return unsafe.Pointer(&sa.raw), SizeofSockaddrTIPC, nil
1027+}
1028+
1029+// SockaddrL2TPIP implements the Sockaddr interface for IPPROTO_L2TP/AF_INET sockets.
1030+type SockaddrL2TPIP struct {
1031+ Addr [4]byte
1032+ ConnId uint32
1033+ raw RawSockaddrL2TPIP
1034+}
1035+
1036+func (sa *SockaddrL2TPIP) sockaddr() (unsafe.Pointer, _Socklen, error) {
1037+ sa.raw.Family = AF_INET
1038+ sa.raw.Conn_id = sa.ConnId
1039+ for i := 0; i < len(sa.Addr); i++ {
1040+ sa.raw.Addr[i] = sa.Addr[i]
1041+ }
1042+ return unsafe.Pointer(&sa.raw), SizeofSockaddrL2TPIP, nil
1043+}
1044+
1045+// SockaddrL2TPIP6 implements the Sockaddr interface for IPPROTO_L2TP/AF_INET6 sockets.
1046+type SockaddrL2TPIP6 struct {
1047+ Addr [16]byte
1048+ ZoneId uint32
1049+ ConnId uint32
1050+ raw RawSockaddrL2TPIP6
1051+}
1052+
1053+func (sa *SockaddrL2TPIP6) sockaddr() (unsafe.Pointer, _Socklen, error) {
1054+ sa.raw.Family = AF_INET6
1055+ sa.raw.Conn_id = sa.ConnId
1056+ sa.raw.Scope_id = sa.ZoneId
1057+ for i := 0; i < len(sa.Addr); i++ {
1058+ sa.raw.Addr[i] = sa.Addr[i]
1059+ }
1060+ return unsafe.Pointer(&sa.raw), SizeofSockaddrL2TPIP6, nil
1061+}
1062+
1063+// SockaddrIUCV implements the Sockaddr interface for AF_IUCV sockets.
1064+type SockaddrIUCV struct {
1065+ UserID string
1066+ Name string
1067+ raw RawSockaddrIUCV
1068+}
1069+
1070+func (sa *SockaddrIUCV) sockaddr() (unsafe.Pointer, _Socklen, error) {
1071+ sa.raw.Family = AF_IUCV
1072+ // These are EBCDIC encoded by the kernel, but we still need to pad them
1073+ // with blanks. Initializing with blanks allows the caller to feed in either
1074+ // a padded or an unpadded string.
1075+ for i := 0; i < 8; i++ {
1076+ sa.raw.Nodeid[i] = ' '
1077+ sa.raw.User_id[i] = ' '
1078+ sa.raw.Name[i] = ' '
1079+ }
1080+ if len(sa.UserID) > 8 || len(sa.Name) > 8 {
1081+ return nil, 0, EINVAL
1082+ }
1083+ for i, b := range []byte(sa.UserID[:]) {
1084+ sa.raw.User_id[i] = int8(b)
1085+ }
1086+ for i, b := range []byte(sa.Name[:]) {
1087+ sa.raw.Name[i] = int8(b)
1088+ }
1089+ return unsafe.Pointer(&sa.raw), SizeofSockaddrIUCV, nil
1090+}
1091+
1092+var socketProtocol = func(fd int) (int, error) {
1093+ return GetsockoptInt(fd, SOL_SOCKET, SO_PROTOCOL)
1094+}
1095+
1096+func anyToSockaddr(fd int, rsa *RawSockaddrAny) (Sockaddr, error) {
1097+ switch rsa.Addr.Family {
1098+ case AF_NETLINK:
1099+ pp := (*RawSockaddrNetlink)(unsafe.Pointer(rsa))
1100+ sa := new(SockaddrNetlink)
1101+ sa.Family = pp.Family
1102+ sa.Pad = pp.Pad
1103+ sa.Pid = pp.Pid
1104+ sa.Groups = pp.Groups
1105+ return sa, nil
1106+
1107+ case AF_PACKET:
1108+ pp := (*RawSockaddrLinklayer)(unsafe.Pointer(rsa))
1109+ sa := new(SockaddrLinklayer)
1110+ sa.Protocol = pp.Protocol
1111+ sa.Ifindex = int(pp.Ifindex)
1112+ sa.Hatype = pp.Hatype
1113+ sa.Pkttype = pp.Pkttype
1114+ sa.Halen = pp.Halen
1115+ for i := 0; i < len(sa.Addr); i++ {
1116+ sa.Addr[i] = pp.Addr[i]
1117+ }
1118+ return sa, nil
1119+
1120+ case AF_UNIX:
1121+ pp := (*RawSockaddrUnix)(unsafe.Pointer(rsa))
1122+ sa := new(SockaddrUnix)
1123+ if pp.Path[0] == 0 {
1124+ // "Abstract" Unix domain socket.
1125+ // Rewrite leading NUL as @ for textual display.
1126+ // (This is the standard convention.)
1127+ // Not friendly to overwrite in place,
1128+ // but the callers below don't care.
1129+ pp.Path[0] = '@'
1130+ }
1131+
1132+ // Assume path ends at NUL.
1133+ // This is not technically the Linux semantics for
1134+ // abstract Unix domain sockets--they are supposed
1135+ // to be uninterpreted fixed-size binary blobs--but
1136+ // everyone uses this convention.
1137+ n := 0
1138+ for n < len(pp.Path) && pp.Path[n] != 0 {
1139+ n++
1140+ }
1141+ bytes := (*[len(pp.Path)]byte)(unsafe.Pointer(&pp.Path[0]))[0:n]
1142+ sa.Name = string(bytes)
1143+ return sa, nil
1144+
1145+ case AF_INET:
1146+ proto, err := socketProtocol(fd)
1147+ if err != nil {
1148+ return nil, err
1149+ }
1150+
1151+ switch proto {
1152+ case IPPROTO_L2TP:
1153+ pp := (*RawSockaddrL2TPIP)(unsafe.Pointer(rsa))
1154+ sa := new(SockaddrL2TPIP)
1155+ sa.ConnId = pp.Conn_id
1156+ for i := 0; i < len(sa.Addr); i++ {
1157+ sa.Addr[i] = pp.Addr[i]
1158+ }
1159+ return sa, nil
1160+ default:
1161+ pp := (*RawSockaddrInet4)(unsafe.Pointer(rsa))
1162+ sa := new(SockaddrInet4)
1163+ p := (*[2]byte)(unsafe.Pointer(&pp.Port))
1164+ sa.Port = int(p[0])<<8 + int(p[1])
1165+ for i := 0; i < len(sa.Addr); i++ {
1166+ sa.Addr[i] = pp.Addr[i]
1167+ }
1168+ return sa, nil
1169+ }
1170+
1171+ case AF_INET6:
1172+ proto, err := socketProtocol(fd)
1173+ if err != nil {
1174+ return nil, err
1175+ }
1176+
1177+ switch proto {
1178+ case IPPROTO_L2TP:
1179+ pp := (*RawSockaddrL2TPIP6)(unsafe.Pointer(rsa))
1180+ sa := new(SockaddrL2TPIP6)
1181+ sa.ConnId = pp.Conn_id
1182+ sa.ZoneId = pp.Scope_id
1183+ for i := 0; i < len(sa.Addr); i++ {
1184+ sa.Addr[i] = pp.Addr[i]
1185+ }
1186+ return sa, nil
1187+ default:
1188+ pp := (*RawSockaddrInet6)(unsafe.Pointer(rsa))
1189+ sa := new(SockaddrInet6)
1190+ p := (*[2]byte)(unsafe.Pointer(&pp.Port))
1191+ sa.Port = int(p[0])<<8 + int(p[1])
1192+ sa.ZoneId = pp.Scope_id
1193+ for i := 0; i < len(sa.Addr); i++ {
1194+ sa.Addr[i] = pp.Addr[i]
1195+ }
1196+ return sa, nil
1197+ }
1198+
1199+ case AF_VSOCK:
1200+ pp := (*RawSockaddrVM)(unsafe.Pointer(rsa))
1201+ sa := &SockaddrVM{
1202+ CID: pp.Cid,
1203+ Port: pp.Port,
1204+ Flags: pp.Flags,
1205+ }
1206+ return sa, nil
1207+ case AF_BLUETOOTH:
1208+ proto, err := socketProtocol(fd)
1209+ if err != nil {
1210+ return nil, err
1211+ }
1212+ // only BTPROTO_L2CAP and BTPROTO_RFCOMM can accept connections
1213+ switch proto {
1214+ case BTPROTO_L2CAP:
1215+ pp := (*RawSockaddrL2)(unsafe.Pointer(rsa))
1216+ sa := &SockaddrL2{
1217+ PSM: pp.Psm,
1218+ CID: pp.Cid,
1219+ Addr: pp.Bdaddr,
1220+ AddrType: pp.Bdaddr_type,
1221+ }
1222+ return sa, nil
1223+ case BTPROTO_RFCOMM:
1224+ pp := (*RawSockaddrRFCOMM)(unsafe.Pointer(rsa))
1225+ sa := &SockaddrRFCOMM{
1226+ Channel: pp.Channel,
1227+ Addr: pp.Bdaddr,
1228+ }
1229+ return sa, nil
1230+ }
1231+ case AF_XDP:
1232+ pp := (*RawSockaddrXDP)(unsafe.Pointer(rsa))
1233+ sa := &SockaddrXDP{
1234+ Flags: pp.Flags,
1235+ Ifindex: pp.Ifindex,
1236+ QueueID: pp.Queue_id,
1237+ SharedUmemFD: pp.Shared_umem_fd,
1238+ }
1239+ return sa, nil
1240+ case AF_PPPOX:
1241+ pp := (*RawSockaddrPPPoX)(unsafe.Pointer(rsa))
1242+ if binary.BigEndian.Uint32(pp[2:6]) != px_proto_oe {
1243+ return nil, EINVAL
1244+ }
1245+ sa := &SockaddrPPPoE{
1246+ SID: binary.BigEndian.Uint16(pp[6:8]),
1247+ Remote: pp[8:14],
1248+ }
1249+ for i := 14; i < 14+IFNAMSIZ; i++ {
1250+ if pp[i] == 0 {
1251+ sa.Dev = string(pp[14:i])
1252+ break
1253+ }
1254+ }
1255+ return sa, nil
1256+ case AF_TIPC:
1257+ pp := (*RawSockaddrTIPC)(unsafe.Pointer(rsa))
1258+
1259+ sa := &SockaddrTIPC{
1260+ Scope: int(pp.Scope),
1261+ }
1262+
1263+ // Determine which union variant is present in pp.Addr by checking
1264+ // pp.Addrtype.
1265+ switch pp.Addrtype {
1266+ case TIPC_SERVICE_RANGE:
1267+ sa.Addr = (*TIPCServiceRange)(unsafe.Pointer(&pp.Addr))
1268+ case TIPC_SERVICE_ADDR:
1269+ sa.Addr = (*TIPCServiceName)(unsafe.Pointer(&pp.Addr))
1270+ case TIPC_SOCKET_ADDR:
1271+ sa.Addr = (*TIPCSocketAddr)(unsafe.Pointer(&pp.Addr))
1272+ default:
1273+ return nil, EINVAL
1274+ }
1275+
1276+ return sa, nil
1277+ case AF_IUCV:
1278+ pp := (*RawSockaddrIUCV)(unsafe.Pointer(rsa))
1279+
1280+ var user [8]byte
1281+ var name [8]byte
1282+
1283+ for i := 0; i < 8; i++ {
1284+ user[i] = byte(pp.User_id[i])
1285+ name[i] = byte(pp.Name[i])
1286+ }
1287+
1288+ sa := &SockaddrIUCV{
1289+ UserID: string(user[:]),
1290+ Name: string(name[:]),
1291+ }
1292+ return sa, nil
1293+
1294+ case AF_CAN:
1295+ proto, err := socketProtocol(fd)
1296+ if err != nil {
1297+ return nil, err
1298+ }
1299+
1300+ pp := (*RawSockaddrCAN)(unsafe.Pointer(rsa))
1301+
1302+ switch proto {
1303+ case CAN_J1939:
1304+ sa := &SockaddrCANJ1939{
1305+ Ifindex: int(pp.Ifindex),
1306+ }
1307+ name := (*[8]byte)(unsafe.Pointer(&sa.Name))
1308+ for i := 0; i < 8; i++ {
1309+ name[i] = pp.Addr[i]
1310+ }
1311+ pgn := (*[4]byte)(unsafe.Pointer(&sa.PGN))
1312+ for i := 0; i < 4; i++ {
1313+ pgn[i] = pp.Addr[i+8]
1314+ }
1315+ addr := (*[1]byte)(unsafe.Pointer(&sa.Addr))
1316+ addr[0] = pp.Addr[12]
1317+ return sa, nil
1318+ default:
1319+ sa := &SockaddrCAN{
1320+ Ifindex: int(pp.Ifindex),
1321+ }
1322+ rx := (*[4]byte)(unsafe.Pointer(&sa.RxID))
1323+ for i := 0; i < 4; i++ {
1324+ rx[i] = pp.Addr[i]
1325+ }
1326+ tx := (*[4]byte)(unsafe.Pointer(&sa.TxID))
1327+ for i := 0; i < 4; i++ {
1328+ tx[i] = pp.Addr[i+4]
1329+ }
1330+ return sa, nil
1331+ }
1332+ }
1333+ return nil, EAFNOSUPPORT
1334+}
1335+
1336+func Accept(fd int) (nfd int, sa Sockaddr, err error) {
1337+ var rsa RawSockaddrAny
1338+ var len _Socklen = SizeofSockaddrAny
1339+ nfd, err = accept(fd, &rsa, &len)
1340+ if err != nil {
1341+ return
1342+ }
1343+ sa, err = anyToSockaddr(fd, &rsa)
1344+ if err != nil {
1345+ Close(nfd)
1346+ nfd = 0
1347+ }
1348+ return
1349+}
1350+
1351+func Accept4(fd int, flags int) (nfd int, sa Sockaddr, err error) {
1352+ var rsa RawSockaddrAny
1353+ var len _Socklen = SizeofSockaddrAny
1354+ nfd, err = accept4(fd, &rsa, &len, flags)
1355+ if err != nil {
1356+ return
1357+ }
1358+ if len > SizeofSockaddrAny {
1359+ panic("RawSockaddrAny too small")
1360+ }
1361+ sa, err = anyToSockaddr(fd, &rsa)
1362+ if err != nil {
1363+ Close(nfd)
1364+ nfd = 0
1365+ }
1366+ return
1367+}
1368+
1369+func Getsockname(fd int) (sa Sockaddr, err error) {
1370+ var rsa RawSockaddrAny
1371+ var len _Socklen = SizeofSockaddrAny
1372+ if err = getsockname(fd, &rsa, &len); err != nil {
1373+ return
1374+ }
1375+ return anyToSockaddr(fd, &rsa)
1376+}
1377+
1378+func GetsockoptIPMreqn(fd, level, opt int) (*IPMreqn, error) {
1379+ var value IPMreqn
1380+ vallen := _Socklen(SizeofIPMreqn)
1381+ err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1382+ return &value, err
1383+}
1384+
1385+func GetsockoptUcred(fd, level, opt int) (*Ucred, error) {
1386+ var value Ucred
1387+ vallen := _Socklen(SizeofUcred)
1388+ err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1389+ return &value, err
1390+}
1391+
1392+func GetsockoptTCPInfo(fd, level, opt int) (*TCPInfo, error) {
1393+ var value TCPInfo
1394+ vallen := _Socklen(SizeofTCPInfo)
1395+ err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1396+ return &value, err
1397+}
1398+
1399+// GetsockoptString returns the string value of the socket option opt for the
1400+// socket associated with fd at the given socket level.
1401+func GetsockoptString(fd, level, opt int) (string, error) {
1402+ buf := make([]byte, 256)
1403+ vallen := _Socklen(len(buf))
1404+ err := getsockopt(fd, level, opt, unsafe.Pointer(&buf[0]), &vallen)
1405+ if err != nil {
1406+ if err == ERANGE {
1407+ buf = make([]byte, vallen)
1408+ err = getsockopt(fd, level, opt, unsafe.Pointer(&buf[0]), &vallen)
1409+ }
1410+ if err != nil {
1411+ return "", err
1412+ }
1413+ }
1414+ return string(buf[:vallen-1]), nil
1415+}
1416+
1417+func GetsockoptTpacketStats(fd, level, opt int) (*TpacketStats, error) {
1418+ var value TpacketStats
1419+ vallen := _Socklen(SizeofTpacketStats)
1420+ err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1421+ return &value, err
1422+}
1423+
1424+func GetsockoptTpacketStatsV3(fd, level, opt int) (*TpacketStatsV3, error) {
1425+ var value TpacketStatsV3
1426+ vallen := _Socklen(SizeofTpacketStatsV3)
1427+ err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1428+ return &value, err
1429+}
1430+
1431+func SetsockoptIPMreqn(fd, level, opt int, mreq *IPMreqn) (err error) {
1432+ return setsockopt(fd, level, opt, unsafe.Pointer(mreq), unsafe.Sizeof(*mreq))
1433+}
1434+
1435+func SetsockoptPacketMreq(fd, level, opt int, mreq *PacketMreq) error {
1436+ return setsockopt(fd, level, opt, unsafe.Pointer(mreq), unsafe.Sizeof(*mreq))
1437+}
1438+
1439+// SetsockoptSockFprog attaches a classic BPF or an extended BPF program to a
1440+// socket to filter incoming packets. See 'man 7 socket' for usage information.
1441+func SetsockoptSockFprog(fd, level, opt int, fprog *SockFprog) error {
1442+ return setsockopt(fd, level, opt, unsafe.Pointer(fprog), unsafe.Sizeof(*fprog))
1443+}
1444+
1445+func SetsockoptCanRawFilter(fd, level, opt int, filter []CanFilter) error {
1446+ var p unsafe.Pointer
1447+ if len(filter) > 0 {
1448+ p = unsafe.Pointer(&filter[0])
1449+ }
1450+ return setsockopt(fd, level, opt, p, uintptr(len(filter)*SizeofCanFilter))
1451+}
1452+
1453+func SetsockoptTpacketReq(fd, level, opt int, tp *TpacketReq) error {
1454+ return setsockopt(fd, level, opt, unsafe.Pointer(tp), unsafe.Sizeof(*tp))
1455+}
1456+
1457+func SetsockoptTpacketReq3(fd, level, opt int, tp *TpacketReq3) error {
1458+ return setsockopt(fd, level, opt, unsafe.Pointer(tp), unsafe.Sizeof(*tp))
1459+}
1460+
1461+// Keyctl Commands (http://man7.org/linux/man-pages/man2/keyctl.2.html)
1462+
1463+// KeyctlInt calls keyctl commands in which each argument is an int.
1464+// These commands are KEYCTL_REVOKE, KEYCTL_CHOWN, KEYCTL_CLEAR, KEYCTL_LINK,
1465+// KEYCTL_UNLINK, KEYCTL_NEGATE, KEYCTL_SET_REQKEY_KEYRING, KEYCTL_SET_TIMEOUT,
1466+// KEYCTL_ASSUME_AUTHORITY, KEYCTL_SESSION_TO_PARENT, KEYCTL_REJECT,
1467+// KEYCTL_INVALIDATE, and KEYCTL_GET_PERSISTENT.
1468+//sys KeyctlInt(cmd int, arg2 int, arg3 int, arg4 int, arg5 int) (ret int, err error) = SYS_KEYCTL
1469+
1470+// KeyctlBuffer calls keyctl commands in which the third and fourth
1471+// arguments are a buffer and its length, respectively.
1472+// These commands are KEYCTL_UPDATE, KEYCTL_READ, and KEYCTL_INSTANTIATE.
1473+//sys KeyctlBuffer(cmd int, arg2 int, buf []byte, arg5 int) (ret int, err error) = SYS_KEYCTL
1474+
1475+// KeyctlString calls keyctl commands which return a string.
1476+// These commands are KEYCTL_DESCRIBE and KEYCTL_GET_SECURITY.
1477+func KeyctlString(cmd int, id int) (string, error) {
1478+ // We must loop as the string data may change in between the syscalls.
1479+ // We could allocate a large buffer here to reduce the chance that the
1480+ // syscall needs to be called twice; however, this is unnecessary as
1481+ // the performance loss is negligible.
1482+ var buffer []byte
1483+ for {
1484+ // Try to fill the buffer with data
1485+ length, err := KeyctlBuffer(cmd, id, buffer, 0)
1486+ if err != nil {
1487+ return "", err
1488+ }
1489+
1490+ // Check if the data was written
1491+ if length <= len(buffer) {
1492+ // Exclude the null terminator
1493+ return string(buffer[:length-1]), nil
1494+ }
1495+
1496+ // Make a bigger buffer if needed
1497+ buffer = make([]byte, length)
1498+ }
1499+}
1500+
1501+// Keyctl commands with special signatures.
1502+
1503+// KeyctlGetKeyringID implements the KEYCTL_GET_KEYRING_ID command.
1504+// See the full documentation at:
1505+// http://man7.org/linux/man-pages/man3/keyctl_get_keyring_ID.3.html
1506+func KeyctlGetKeyringID(id int, create bool) (ringid int, err error) {
1507+ createInt := 0
1508+ if create {
1509+ createInt = 1
1510+ }
1511+ return KeyctlInt(KEYCTL_GET_KEYRING_ID, id, createInt, 0, 0)
1512+}
1513+
1514+// KeyctlSetperm implements the KEYCTL_SETPERM command. The perm value is the
1515+// key handle permission mask as described in the "keyctl setperm" section of
1516+// http://man7.org/linux/man-pages/man1/keyctl.1.html.
1517+// See the full documentation at:
1518+// http://man7.org/linux/man-pages/man3/keyctl_setperm.3.html
1519+func KeyctlSetperm(id int, perm uint32) error {
1520+ _, err := KeyctlInt(KEYCTL_SETPERM, id, int(perm), 0, 0)
1521+ return err
1522+}
1523+
1524+//sys keyctlJoin(cmd int, arg2 string) (ret int, err error) = SYS_KEYCTL
1525+
1526+// KeyctlJoinSessionKeyring implements the KEYCTL_JOIN_SESSION_KEYRING command.
1527+// See the full documentation at:
1528+// http://man7.org/linux/man-pages/man3/keyctl_join_session_keyring.3.html
1529+func KeyctlJoinSessionKeyring(name string) (ringid int, err error) {
1530+ return keyctlJoin(KEYCTL_JOIN_SESSION_KEYRING, name)
1531+}
1532+
1533+//sys keyctlSearch(cmd int, arg2 int, arg3 string, arg4 string, arg5 int) (ret int, err error) = SYS_KEYCTL
1534+
1535+// KeyctlSearch implements the KEYCTL_SEARCH command.
1536+// See the full documentation at:
1537+// http://man7.org/linux/man-pages/man3/keyctl_search.3.html
1538+func KeyctlSearch(ringid int, keyType, description string, destRingid int) (id int, err error) {
1539+ return keyctlSearch(KEYCTL_SEARCH, ringid, keyType, description, destRingid)
1540+}
1541+
1542+//sys keyctlIOV(cmd int, arg2 int, payload []Iovec, arg5 int) (err error) = SYS_KEYCTL
1543+
1544+// KeyctlInstantiateIOV implements the KEYCTL_INSTANTIATE_IOV command. This
1545+// command is similar to KEYCTL_INSTANTIATE, except that the payload is a slice
1546+// of Iovec (each of which represents a buffer) instead of a single buffer.
1547+// See the full documentation at:
1548+// http://man7.org/linux/man-pages/man3/keyctl_instantiate_iov.3.html
1549+func KeyctlInstantiateIOV(id int, payload []Iovec, ringid int) error {
1550+ return keyctlIOV(KEYCTL_INSTANTIATE_IOV, id, payload, ringid)
1551+}
1552+
1553+//sys keyctlDH(cmd int, arg2 *KeyctlDHParams, buf []byte) (ret int, err error) = SYS_KEYCTL
1554+
1555+// KeyctlDHCompute implements the KEYCTL_DH_COMPUTE command. This command
1556+// computes a Diffie-Hellman shared secret based on the provide params. The
1557+// secret is written to the provided buffer and the returned size is the number
1558+// of bytes written (returning an error if there is insufficient space in the
1559+// buffer). If a nil buffer is passed in, this function returns the minimum
1560+// buffer length needed to store the appropriate data. Note that this differs
1561+// from KEYCTL_READ's behavior which always returns the requested payload size.
1562+// See the full documentation at:
1563+// http://man7.org/linux/man-pages/man3/keyctl_dh_compute.3.html
1564+func KeyctlDHCompute(params *KeyctlDHParams, buffer []byte) (size int, err error) {
1565+ return keyctlDH(KEYCTL_DH_COMPUTE, params, buffer)
1566+}
1567+
1568+// KeyctlRestrictKeyring implements the KEYCTL_RESTRICT_KEYRING command. This
1569+// command limits the set of keys that can be linked to the keyring, regardless
1570+// of keyring permissions. The command requires the "setattr" permission.
1571+//
1572+// When called with an empty keyType the command locks the keyring, preventing
1573+// any further keys from being linked to the keyring.
1574+//
1575+// The "asymmetric" keyType defines restrictions requiring key payloads to be
1576+// DER encoded X.509 certificates signed by keys in another keyring. Restrictions
1577+// for "asymmetric" include "builtin_trusted", "builtin_and_secondary_trusted",
1578+// "key_or_keyring:<key>", and "key_or_keyring:<key>:chain".
1579+//
1580+// As of Linux 4.12, only the "asymmetric" keyType defines type-specific
1581+// restrictions.
1582+//
1583+// See the full documentation at:
1584+// http://man7.org/linux/man-pages/man3/keyctl_restrict_keyring.3.html
1585+// http://man7.org/linux/man-pages/man2/keyctl.2.html
1586+func KeyctlRestrictKeyring(ringid int, keyType string, restriction string) error {
1587+ if keyType == "" {
1588+ return keyctlRestrictKeyring(KEYCTL_RESTRICT_KEYRING, ringid)
1589+ }
1590+ return keyctlRestrictKeyringByType(KEYCTL_RESTRICT_KEYRING, ringid, keyType, restriction)
1591+}
1592+
1593+//sys keyctlRestrictKeyringByType(cmd int, arg2 int, keyType string, restriction string) (err error) = SYS_KEYCTL
1594+//sys keyctlRestrictKeyring(cmd int, arg2 int) (err error) = SYS_KEYCTL
1595+
1596+func Recvmsg(fd int, p, oob []byte, flags int) (n, oobn int, recvflags int, from Sockaddr, err error) {
1597+ var msg Msghdr
1598+ var rsa RawSockaddrAny
1599+ msg.Name = (*byte)(unsafe.Pointer(&rsa))
1600+ msg.Namelen = uint32(SizeofSockaddrAny)
1601+ var iov Iovec
1602+ if len(p) > 0 {
1603+ iov.Base = &p[0]
1604+ iov.SetLen(len(p))
1605+ }
1606+ var dummy byte
1607+ if len(oob) > 0 {
1608+ if len(p) == 0 {
1609+ var sockType int
1610+ sockType, err = GetsockoptInt(fd, SOL_SOCKET, SO_TYPE)
1611+ if err != nil {
1612+ return
1613+ }
1614+ // receive at least one normal byte
1615+ if sockType != SOCK_DGRAM {
1616+ iov.Base = &dummy
1617+ iov.SetLen(1)
1618+ }
1619+ }
1620+ msg.Control = &oob[0]
1621+ msg.SetControllen(len(oob))
1622+ }
1623+ msg.Iov = &iov
1624+ msg.Iovlen = 1
1625+ if n, err = recvmsg(fd, &msg, flags); err != nil {
1626+ return
1627+ }
1628+ oobn = int(msg.Controllen)
1629+ recvflags = int(msg.Flags)
1630+ // source address is only specified if the socket is unconnected
1631+ if rsa.Addr.Family != AF_UNSPEC {
1632+ from, err = anyToSockaddr(fd, &rsa)
1633+ }
1634+ return
1635+}
1636+
1637+func Sendmsg(fd int, p, oob []byte, to Sockaddr, flags int) (err error) {
1638+ _, err = SendmsgN(fd, p, oob, to, flags)
1639+ return
1640+}
1641+
1642+func SendmsgN(fd int, p, oob []byte, to Sockaddr, flags int) (n int, err error) {
1643+ var ptr unsafe.Pointer
1644+ var salen _Socklen
1645+ if to != nil {
1646+ var err error
1647+ ptr, salen, err = to.sockaddr()
1648+ if err != nil {
1649+ return 0, err
1650+ }
1651+ }
1652+ var msg Msghdr
1653+ msg.Name = (*byte)(ptr)
1654+ msg.Namelen = uint32(salen)
1655+ var iov Iovec
1656+ if len(p) > 0 {
1657+ iov.Base = &p[0]
1658+ iov.SetLen(len(p))
1659+ }
1660+ var dummy byte
1661+ if len(oob) > 0 {
1662+ if len(p) == 0 {
1663+ var sockType int
1664+ sockType, err = GetsockoptInt(fd, SOL_SOCKET, SO_TYPE)
1665+ if err != nil {
1666+ return 0, err
1667+ }
1668+ // send at least one normal byte
1669+ if sockType != SOCK_DGRAM {
1670+ iov.Base = &dummy
1671+ iov.SetLen(1)
1672+ }
1673+ }
1674+ msg.Control = &oob[0]
1675+ msg.SetControllen(len(oob))
1676+ }
1677+ msg.Iov = &iov
1678+ msg.Iovlen = 1
1679+ if n, err = sendmsg(fd, &msg, flags); err != nil {
1680+ return 0, err
1681+ }
1682+ if len(oob) > 0 && len(p) == 0 {
1683+ n = 0
1684+ }
1685+ return n, nil
1686+}
1687+
1688+// BindToDevice binds the socket associated with fd to device.
1689+func BindToDevice(fd int, device string) (err error) {
1690+ return SetsockoptString(fd, SOL_SOCKET, SO_BINDTODEVICE, device)
1691+}
1692+
1693+//sys ptrace(request int, pid int, addr uintptr, data uintptr) (err error)
1694+
1695+func ptracePeek(req int, pid int, addr uintptr, out []byte) (count int, err error) {
1696+ // The peek requests are machine-size oriented, so we wrap it
1697+ // to retrieve arbitrary-length data.
1698+
1699+ // The ptrace syscall differs from glibc's ptrace.
1700+ // Peeks returns the word in *data, not as the return value.
1701+
1702+ var buf [SizeofPtr]byte
1703+
1704+ // Leading edge. PEEKTEXT/PEEKDATA don't require aligned
1705+ // access (PEEKUSER warns that it might), but if we don't
1706+ // align our reads, we might straddle an unmapped page
1707+ // boundary and not get the bytes leading up to the page
1708+ // boundary.
1709+ n := 0
1710+ if addr%SizeofPtr != 0 {
1711+ err = ptrace(req, pid, addr-addr%SizeofPtr, uintptr(unsafe.Pointer(&buf[0])))
1712+ if err != nil {
1713+ return 0, err
1714+ }
1715+ n += copy(out, buf[addr%SizeofPtr:])
1716+ out = out[n:]
1717+ }
1718+
1719+ // Remainder.
1720+ for len(out) > 0 {
1721+ // We use an internal buffer to guarantee alignment.
1722+ // It's not documented if this is necessary, but we're paranoid.
1723+ err = ptrace(req, pid, addr+uintptr(n), uintptr(unsafe.Pointer(&buf[0])))
1724+ if err != nil {
1725+ return n, err
1726+ }
1727+ copied := copy(out, buf[0:])
1728+ n += copied
1729+ out = out[copied:]
1730+ }
1731+
1732+ return n, nil
1733+}
1734+
1735+func PtracePeekText(pid int, addr uintptr, out []byte) (count int, err error) {
1736+ return ptracePeek(PTRACE_PEEKTEXT, pid, addr, out)
1737+}
1738+
1739+func PtracePeekData(pid int, addr uintptr, out []byte) (count int, err error) {
1740+ return ptracePeek(PTRACE_PEEKDATA, pid, addr, out)
1741+}
1742+
1743+func PtracePeekUser(pid int, addr uintptr, out []byte) (count int, err error) {
1744+ return ptracePeek(PTRACE_PEEKUSR, pid, addr, out)
1745+}
1746+
1747+func ptracePoke(pokeReq int, peekReq int, pid int, addr uintptr, data []byte) (count int, err error) {
1748+ // As for ptracePeek, we need to align our accesses to deal
1749+ // with the possibility of straddling an invalid page.
1750+
1751+ // Leading edge.
1752+ n := 0
1753+ if addr%SizeofPtr != 0 {
1754+ var buf [SizeofPtr]byte
1755+ err = ptrace(peekReq, pid, addr-addr%SizeofPtr, uintptr(unsafe.Pointer(&buf[0])))
1756+ if err != nil {
1757+ return 0, err
1758+ }
1759+ n += copy(buf[addr%SizeofPtr:], data)
1760+ word := *((*uintptr)(unsafe.Pointer(&buf[0])))
1761+ err = ptrace(pokeReq, pid, addr-addr%SizeofPtr, word)
1762+ if err != nil {
1763+ return 0, err
1764+ }
1765+ data = data[n:]
1766+ }
1767+
1768+ // Interior.
1769+ for len(data) > SizeofPtr {
1770+ word := *((*uintptr)(unsafe.Pointer(&data[0])))
1771+ err = ptrace(pokeReq, pid, addr+uintptr(n), word)
1772+ if err != nil {
1773+ return n, err
1774+ }
1775+ n += SizeofPtr
1776+ data = data[SizeofPtr:]
1777+ }
1778+
1779+ // Trailing edge.
1780+ if len(data) > 0 {
1781+ var buf [SizeofPtr]byte
1782+ err = ptrace(peekReq, pid, addr+uintptr(n), uintptr(unsafe.Pointer(&buf[0])))
1783+ if err != nil {
1784+ return n, err
1785+ }
1786+ copy(buf[0:], data)
1787+ word := *((*uintptr)(unsafe.Pointer(&buf[0])))
1788+ err = ptrace(pokeReq, pid, addr+uintptr(n), word)
1789+ if err != nil {
1790+ return n, err
1791+ }
1792+ n += len(data)
1793+ }
1794+
1795+ return n, nil
1796+}
1797+
1798+func PtracePokeText(pid int, addr uintptr, data []byte) (count int, err error) {
1799+ return ptracePoke(PTRACE_POKETEXT, PTRACE_PEEKTEXT, pid, addr, data)
1800+}
1801+
1802+func PtracePokeData(pid int, addr uintptr, data []byte) (count int, err error) {
1803+ return ptracePoke(PTRACE_POKEDATA, PTRACE_PEEKDATA, pid, addr, data)
1804+}
1805+
1806+func PtracePokeUser(pid int, addr uintptr, data []byte) (count int, err error) {
1807+ return ptracePoke(PTRACE_POKEUSR, PTRACE_PEEKUSR, pid, addr, data)
1808+}
1809+
1810+func PtraceGetRegs(pid int, regsout *PtraceRegs) (err error) {
1811+ return ptrace(PTRACE_GETREGS, pid, 0, uintptr(unsafe.Pointer(regsout)))
1812+}
1813+
1814+func PtraceSetRegs(pid int, regs *PtraceRegs) (err error) {
1815+ return ptrace(PTRACE_SETREGS, pid, 0, uintptr(unsafe.Pointer(regs)))
1816+}
1817+
1818+func PtraceSetOptions(pid int, options int) (err error) {
1819+ return ptrace(PTRACE_SETOPTIONS, pid, 0, uintptr(options))
1820+}
1821+
1822+func PtraceGetEventMsg(pid int) (msg uint, err error) {
1823+ var data _C_long
1824+ err = ptrace(PTRACE_GETEVENTMSG, pid, 0, uintptr(unsafe.Pointer(&data)))
1825+ msg = uint(data)
1826+ return
1827+}
1828+
1829+func PtraceCont(pid int, signal int) (err error) {
1830+ return ptrace(PTRACE_CONT, pid, 0, uintptr(signal))
1831+}
1832+
1833+func PtraceSyscall(pid int, signal int) (err error) {
1834+ return ptrace(PTRACE_SYSCALL, pid, 0, uintptr(signal))
1835+}
1836+
1837+func PtraceSingleStep(pid int) (err error) { return ptrace(PTRACE_SINGLESTEP, pid, 0, 0) }
1838+
1839+func PtraceInterrupt(pid int) (err error) { return ptrace(PTRACE_INTERRUPT, pid, 0, 0) }
1840+
1841+func PtraceAttach(pid int) (err error) { return ptrace(PTRACE_ATTACH, pid, 0, 0) }
1842+
1843+func PtraceSeize(pid int) (err error) { return ptrace(PTRACE_SEIZE, pid, 0, 0) }
1844+
1845+func PtraceDetach(pid int) (err error) { return ptrace(PTRACE_DETACH, pid, 0, 0) }
1846+
1847+//sys reboot(magic1 uint, magic2 uint, cmd int, arg string) (err error)
1848+
1849+func Reboot(cmd int) (err error) {
1850+ return reboot(LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, cmd, "")
1851+}
1852+
1853+func direntIno(buf []byte) (uint64, bool) {
1854+ return readInt(buf, unsafe.Offsetof(Dirent{}.Ino), unsafe.Sizeof(Dirent{}.Ino))
1855+}
1856+
1857+func direntReclen(buf []byte) (uint64, bool) {
1858+ return readInt(buf, unsafe.Offsetof(Dirent{}.Reclen), unsafe.Sizeof(Dirent{}.Reclen))
1859+}
1860+
1861+func direntNamlen(buf []byte) (uint64, bool) {
1862+ reclen, ok := direntReclen(buf)
1863+ if !ok {
1864+ return 0, false
1865+ }
1866+ return reclen - uint64(unsafe.Offsetof(Dirent{}.Name)), true
1867+}
1868+
1869+//sys mount(source string, target string, fstype string, flags uintptr, data *byte) (err error)
1870+
1871+func Mount(source string, target string, fstype string, flags uintptr, data string) (err error) {
1872+ // Certain file systems get rather angry and EINVAL if you give
1873+ // them an empty string of data, rather than NULL.
1874+ if data == "" {
1875+ return mount(source, target, fstype, flags, nil)
1876+ }
1877+ datap, err := BytePtrFromString(data)
1878+ if err != nil {
1879+ return err
1880+ }
1881+ return mount(source, target, fstype, flags, datap)
1882+}
1883+
1884+func Sendfile(outfd int, infd int, offset *int64, count int) (written int, err error) {
1885+ if raceenabled {
1886+ raceReleaseMerge(unsafe.Pointer(&ioSync))
1887+ }
1888+ return sendfile(outfd, infd, offset, count)
1889+}
1890+
1891+// Sendto
1892+// Recvfrom
1893+// Socketpair
1894+
1895+/*
1896+ * Direct access
1897+ */
1898+//sys Acct(path string) (err error)
1899+//sys AddKey(keyType string, description string, payload []byte, ringid int) (id int, err error)
1900+//sys Adjtimex(buf *Timex) (state int, err error)
1901+//sysnb Capget(hdr *CapUserHeader, data *CapUserData) (err error)
1902+//sysnb Capset(hdr *CapUserHeader, data *CapUserData) (err error)
1903+//sys Chdir(path string) (err error)
1904+//sys Chroot(path string) (err error)
1905+//sys ClockGetres(clockid int32, res *Timespec) (err error)
1906+//sys ClockGettime(clockid int32, time *Timespec) (err error)
1907+//sys ClockNanosleep(clockid int32, flags int, request *Timespec, remain *Timespec) (err error)
1908+//sys Close(fd int) (err error)
1909+//sys CloseRange(first uint, last uint, flags uint) (err error)
1910+//sys CopyFileRange(rfd int, roff *int64, wfd int, woff *int64, len int, flags int) (n int, err error)
1911+//sys DeleteModule(name string, flags int) (err error)
1912+//sys Dup(oldfd int) (fd int, err error)
1913+
1914+func Dup2(oldfd, newfd int) error {
1915+ // Android O and newer blocks dup2; riscv and arm64 don't implement dup2.
1916+ if runtime.GOOS == "android" || runtime.GOARCH == "riscv64" || runtime.GOARCH == "arm64" {
1917+ return Dup3(oldfd, newfd, 0)
1918+ }
1919+ return dup2(oldfd, newfd)
1920+}
1921+
1922+//sys Dup3(oldfd int, newfd int, flags int) (err error)
1923+//sysnb EpollCreate1(flag int) (fd int, err error)
1924+//sysnb EpollCtl(epfd int, op int, fd int, event *EpollEvent) (err error)
1925+//sys Eventfd(initval uint, flags int) (fd int, err error) = SYS_EVENTFD2
1926+//sys Exit(code int) = SYS_EXIT_GROUP
1927+//sys Fallocate(fd int, mode uint32, off int64, len int64) (err error)
1928+//sys Fchdir(fd int) (err error)
1929+//sys Fchmod(fd int, mode uint32) (err error)
1930+//sys Fchownat(dirfd int, path string, uid int, gid int, flags int) (err error)
1931+//sys Fdatasync(fd int) (err error)
1932+//sys Fgetxattr(fd int, attr string, dest []byte) (sz int, err error)
1933+//sys FinitModule(fd int, params string, flags int) (err error)
1934+//sys Flistxattr(fd int, dest []byte) (sz int, err error)
1935+//sys Flock(fd int, how int) (err error)
1936+//sys Fremovexattr(fd int, attr string) (err error)
1937+//sys Fsetxattr(fd int, attr string, dest []byte, flags int) (err error)
1938+//sys Fsync(fd int) (err error)
1939+//sys Getdents(fd int, buf []byte) (n int, err error) = SYS_GETDENTS64
1940+//sysnb Getpgid(pid int) (pgid int, err error)
1941+
1942+func Getpgrp() (pid int) {
1943+ pid, _ = Getpgid(0)
1944+ return
1945+}
1946+
1947+//sysnb Getpid() (pid int)
1948+//sysnb Getppid() (ppid int)
1949+//sys Getpriority(which int, who int) (prio int, err error)
1950+//sys Getrandom(buf []byte, flags int) (n int, err error)
1951+//sysnb Getrusage(who int, rusage *Rusage) (err error)
1952+//sysnb Getsid(pid int) (sid int, err error)
1953+//sysnb Gettid() (tid int)
1954+//sys Getxattr(path string, attr string, dest []byte) (sz int, err error)
1955+//sys InitModule(moduleImage []byte, params string) (err error)
1956+//sys InotifyAddWatch(fd int, pathname string, mask uint32) (watchdesc int, err error)
1957+//sysnb InotifyInit1(flags int) (fd int, err error)
1958+//sysnb InotifyRmWatch(fd int, watchdesc uint32) (success int, err error)
1959+//sysnb Kill(pid int, sig syscall.Signal) (err error)
1960+//sys Klogctl(typ int, buf []byte) (n int, err error) = SYS_SYSLOG
1961+//sys Lgetxattr(path string, attr string, dest []byte) (sz int, err error)
1962+//sys Listxattr(path string, dest []byte) (sz int, err error)
1963+//sys Llistxattr(path string, dest []byte) (sz int, err error)
1964+//sys Lremovexattr(path string, attr string) (err error)
1965+//sys Lsetxattr(path string, attr string, data []byte, flags int) (err error)
1966+//sys MemfdCreate(name string, flags int) (fd int, err error)
1967+//sys Mkdirat(dirfd int, path string, mode uint32) (err error)
1968+//sys Mknodat(dirfd int, path string, mode uint32, dev int) (err error)
1969+//sys Nanosleep(time *Timespec, leftover *Timespec) (err error)
1970+//sys PerfEventOpen(attr *PerfEventAttr, pid int, cpu int, groupFd int, flags int) (fd int, err error)
1971+//sys PivotRoot(newroot string, putold string) (err error) = SYS_PIVOT_ROOT
1972+//sysnb prlimit(pid int, resource int, newlimit *Rlimit, old *Rlimit) (err error) = SYS_PRLIMIT64
1973+//sys Prctl(option int, arg2 uintptr, arg3 uintptr, arg4 uintptr, arg5 uintptr) (err error)
1974+//sys Pselect(nfd int, r *FdSet, w *FdSet, e *FdSet, timeout *Timespec, sigmask *Sigset_t) (n int, err error) = SYS_PSELECT6
1975+//sys read(fd int, p []byte) (n int, err error)
1976+//sys Removexattr(path string, attr string) (err error)
1977+//sys Renameat2(olddirfd int, oldpath string, newdirfd int, newpath string, flags uint) (err error)
1978+//sys RequestKey(keyType string, description string, callback string, destRingid int) (id int, err error)
1979+//sys Setdomainname(p []byte) (err error)
1980+//sys Sethostname(p []byte) (err error)
1981+//sysnb Setpgid(pid int, pgid int) (err error)
1982+//sysnb Setsid() (pid int, err error)
1983+//sysnb Settimeofday(tv *Timeval) (err error)
1984+//sys Setns(fd int, nstype int) (err error)
1985+
1986+// PrctlRetInt performs a prctl operation specified by option and further
1987+// optional arguments arg2 through arg5 depending on option. It returns a
1988+// non-negative integer that is returned by the prctl syscall.
1989+func PrctlRetInt(option int, arg2 uintptr, arg3 uintptr, arg4 uintptr, arg5 uintptr) (int, error) {
1990+ ret, _, err := Syscall6(SYS_PRCTL, uintptr(option), uintptr(arg2), uintptr(arg3), uintptr(arg4), uintptr(arg5), 0)
1991+ if err != 0 {
1992+ return 0, err
1993+ }
1994+ return int(ret), nil
1995+}
1996+
1997+// issue 1435.
1998+// On linux Setuid and Setgid only affects the current thread, not the process.
1999+// This does not match what most callers expect so we must return an error
2000+// here rather than letting the caller think that the call succeeded.
2001+
2002+func Setuid(uid int) (err error) {
2003+ return EOPNOTSUPP
2004+}
2005+
2006+func Setgid(uid int) (err error) {
2007+ return EOPNOTSUPP
2008+}
2009+
2010+// SetfsgidRetGid sets fsgid for current thread and returns previous fsgid set.
2011+// setfsgid(2) will return a non-nil error only if its caller lacks CAP_SETUID capability.
2012+// If the call fails due to other reasons, current fsgid will be returned.
2013+func SetfsgidRetGid(gid int) (int, error) {
2014+ return setfsgid(gid)
2015+}
2016+
2017+// SetfsuidRetUid sets fsuid for current thread and returns previous fsuid set.
2018+// setfsgid(2) will return a non-nil error only if its caller lacks CAP_SETUID capability
2019+// If the call fails due to other reasons, current fsuid will be returned.
2020+func SetfsuidRetUid(uid int) (int, error) {
2021+ return setfsuid(uid)
2022+}
2023+
2024+func Setfsgid(gid int) error {
2025+ _, err := setfsgid(gid)
2026+ return err
2027+}
2028+
2029+func Setfsuid(uid int) error {
2030+ _, err := setfsuid(uid)
2031+ return err
2032+}
2033+
2034+func Signalfd(fd int, sigmask *Sigset_t, flags int) (newfd int, err error) {
2035+ return signalfd(fd, sigmask, _C__NSIG/8, flags)
2036+}
2037+
2038+//sys Setpriority(which int, who int, prio int) (err error)
2039+//sys Setxattr(path string, attr string, data []byte, flags int) (err error)
2040+//sys signalfd(fd int, sigmask *Sigset_t, maskSize uintptr, flags int) (newfd int, err error) = SYS_SIGNALFD4
2041+//sys Statx(dirfd int, path string, flags int, mask int, stat *Statx_t) (err error)
2042+//sys Sync()
2043+//sys Syncfs(fd int) (err error)
2044+//sysnb Sysinfo(info *Sysinfo_t) (err error)
2045+//sys Tee(rfd int, wfd int, len int, flags int) (n int64, err error)
2046+//sysnb TimerfdCreate(clockid int, flags int) (fd int, err error)
2047+//sysnb TimerfdGettime(fd int, currValue *ItimerSpec) (err error)
2048+//sysnb TimerfdSettime(fd int, flags int, newValue *ItimerSpec, oldValue *ItimerSpec) (err error)
2049+//sysnb Tgkill(tgid int, tid int, sig syscall.Signal) (err error)
2050+//sysnb Times(tms *Tms) (ticks uintptr, err error)
2051+//sysnb Umask(mask int) (oldmask int)
2052+//sysnb Uname(buf *Utsname) (err error)
2053+//sys Unmount(target string, flags int) (err error) = SYS_UMOUNT2
2054+//sys Unshare(flags int) (err error)
2055+//sys write(fd int, p []byte) (n int, err error)
2056+//sys exitThread(code int) (err error) = SYS_EXIT
2057+//sys readlen(fd int, p *byte, np int) (n int, err error) = SYS_READ
2058+//sys writelen(fd int, p *byte, np int) (n int, err error) = SYS_WRITE
2059+//sys readv(fd int, iovs []Iovec) (n int, err error) = SYS_READV
2060+//sys writev(fd int, iovs []Iovec) (n int, err error) = SYS_WRITEV
2061+//sys preadv(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr) (n int, err error) = SYS_PREADV
2062+//sys pwritev(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr) (n int, err error) = SYS_PWRITEV
2063+//sys preadv2(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr, flags int) (n int, err error) = SYS_PREADV2
2064+//sys pwritev2(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr, flags int) (n int, err error) = SYS_PWRITEV2
2065+
2066+func bytes2iovec(bs [][]byte) []Iovec {
2067+ iovecs := make([]Iovec, len(bs))
2068+ for i, b := range bs {
2069+ iovecs[i].SetLen(len(b))
2070+ if len(b) > 0 {
2071+ iovecs[i].Base = &b[0]
2072+ } else {
2073+ iovecs[i].Base = (*byte)(unsafe.Pointer(&_zero))
2074+ }
2075+ }
2076+ return iovecs
2077+}
2078+
2079+// offs2lohi splits offs into its lower and upper unsigned long. On 64-bit
2080+// systems, hi will always be 0. On 32-bit systems, offs will be split in half.
2081+// preadv/pwritev chose this calling convention so they don't need to add a
2082+// padding-register for alignment on ARM.
2083+func offs2lohi(offs int64) (lo, hi uintptr) {
2084+ return uintptr(offs), uintptr(uint64(offs) >> SizeofLong)
2085+}
2086+
2087+func Readv(fd int, iovs [][]byte) (n int, err error) {
2088+ iovecs := bytes2iovec(iovs)
2089+ n, err = readv(fd, iovecs)
2090+ readvRacedetect(iovecs, n, err)
2091+ return n, err
2092+}
2093+
2094+func Preadv(fd int, iovs [][]byte, offset int64) (n int, err error) {
2095+ iovecs := bytes2iovec(iovs)
2096+ lo, hi := offs2lohi(offset)
2097+ n, err = preadv(fd, iovecs, lo, hi)
2098+ readvRacedetect(iovecs, n, err)
2099+ return n, err
2100+}
2101+
2102+func Preadv2(fd int, iovs [][]byte, offset int64, flags int) (n int, err error) {
2103+ iovecs := bytes2iovec(iovs)
2104+ lo, hi := offs2lohi(offset)
2105+ n, err = preadv2(fd, iovecs, lo, hi, flags)
2106+ readvRacedetect(iovecs, n, err)
2107+ return n, err
2108+}
2109+
2110+func readvRacedetect(iovecs []Iovec, n int, err error) {
2111+ if !raceenabled {
2112+ return
2113+ }
2114+ for i := 0; n > 0 && i < len(iovecs); i++ {
2115+ m := int(iovecs[i].Len)
2116+ if m > n {
2117+ m = n
2118+ }
2119+ n -= m
2120+ if m > 0 {
2121+ raceWriteRange(unsafe.Pointer(iovecs[i].Base), m)
2122+ }
2123+ }
2124+ if err == nil {
2125+ raceAcquire(unsafe.Pointer(&ioSync))
2126+ }
2127+}
2128+
2129+func Writev(fd int, iovs [][]byte) (n int, err error) {
2130+ iovecs := bytes2iovec(iovs)
2131+ if raceenabled {
2132+ raceReleaseMerge(unsafe.Pointer(&ioSync))
2133+ }
2134+ n, err = writev(fd, iovecs)
2135+ writevRacedetect(iovecs, n)
2136+ return n, err
2137+}
2138+
2139+func Pwritev(fd int, iovs [][]byte, offset int64) (n int, err error) {
2140+ iovecs := bytes2iovec(iovs)
2141+ if raceenabled {
2142+ raceReleaseMerge(unsafe.Pointer(&ioSync))
2143+ }
2144+ lo, hi := offs2lohi(offset)
2145+ n, err = pwritev(fd, iovecs, lo, hi)
2146+ writevRacedetect(iovecs, n)
2147+ return n, err
2148+}
2149+
2150+func Pwritev2(fd int, iovs [][]byte, offset int64, flags int) (n int, err error) {
2151+ iovecs := bytes2iovec(iovs)
2152+ if raceenabled {
2153+ raceReleaseMerge(unsafe.Pointer(&ioSync))
2154+ }
2155+ lo, hi := offs2lohi(offset)
2156+ n, err = pwritev2(fd, iovecs, lo, hi, flags)
2157+ writevRacedetect(iovecs, n)
2158+ return n, err
2159+}
2160+
2161+func writevRacedetect(iovecs []Iovec, n int) {
2162+ if !raceenabled {
2163+ return
2164+ }
2165+ for i := 0; n > 0 && i < len(iovecs); i++ {
2166+ m := int(iovecs[i].Len)
2167+ if m > n {
2168+ m = n
2169+ }
2170+ n -= m
2171+ if m > 0 {
2172+ raceReadRange(unsafe.Pointer(iovecs[i].Base), m)
2173+ }
2174+ }
2175+}
2176+
2177+// mmap varies by architecture; see syscall_linux_*.go.
2178+//sys munmap(addr uintptr, length uintptr) (err error)
2179+
2180+var mapper = &mmapper{
2181+ active: make(map[*byte][]byte),
2182+ mmap: mmap,
2183+ munmap: munmap,
2184+}
2185+
2186+func Mmap(fd int, offset int64, length int, prot int, flags int) (data []byte, err error) {
2187+ return mapper.Mmap(fd, offset, length, prot, flags)
2188+}
2189+
2190+func Munmap(b []byte) (err error) {
2191+ return mapper.Munmap(b)
2192+}
2193+
2194+//sys Madvise(b []byte, advice int) (err error)
2195+//sys Mprotect(b []byte, prot int) (err error)
2196+//sys Mlock(b []byte) (err error)
2197+//sys Mlockall(flags int) (err error)
2198+//sys Msync(b []byte, flags int) (err error)
2199+//sys Munlock(b []byte) (err error)
2200+//sys Munlockall() (err error)
2201+
2202+// Vmsplice splices user pages from a slice of Iovecs into a pipe specified by fd,
2203+// using the specified flags.
2204+func Vmsplice(fd int, iovs []Iovec, flags int) (int, error) {
2205+ var p unsafe.Pointer
2206+ if len(iovs) > 0 {
2207+ p = unsafe.Pointer(&iovs[0])
2208+ }
2209+
2210+ n, _, errno := Syscall6(SYS_VMSPLICE, uintptr(fd), uintptr(p), uintptr(len(iovs)), uintptr(flags), 0, 0)
2211+ if errno != 0 {
2212+ return 0, syscall.Errno(errno)
2213+ }
2214+
2215+ return int(n), nil
2216+}
2217+
2218+func isGroupMember(gid int) bool {
2219+ groups, err := Getgroups()
2220+ if err != nil {
2221+ return false
2222+ }
2223+
2224+ for _, g := range groups {
2225+ if g == gid {
2226+ return true
2227+ }
2228+ }
2229+ return false
2230+}
2231+
2232+//sys faccessat(dirfd int, path string, mode uint32) (err error)
2233+//sys Faccessat2(dirfd int, path string, mode uint32, flags int) (err error)
2234+
2235+func Faccessat(dirfd int, path string, mode uint32, flags int) (err error) {
2236+ if flags == 0 {
2237+ return faccessat(dirfd, path, mode)
2238+ }
2239+
2240+ if err := Faccessat2(dirfd, path, mode, flags); err != ENOSYS && err != EPERM {
2241+ return err
2242+ }
2243+
2244+ // The Linux kernel faccessat system call does not take any flags.
2245+ // The glibc faccessat implements the flags itself; see
2246+ // https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/faccessat.c;hb=HEAD
2247+ // Because people naturally expect syscall.Faccessat to act
2248+ // like C faccessat, we do the same.
2249+
2250+ if flags & ^(AT_SYMLINK_NOFOLLOW|AT_EACCESS) != 0 {
2251+ return EINVAL
2252+ }
2253+
2254+ var st Stat_t
2255+ if err := Fstatat(dirfd, path, &st, flags&AT_SYMLINK_NOFOLLOW); err != nil {
2256+ return err
2257+ }
2258+
2259+ mode &= 7
2260+ if mode == 0 {
2261+ return nil
2262+ }
2263+
2264+ var uid int
2265+ if flags&AT_EACCESS != 0 {
2266+ uid = Geteuid()
2267+ } else {
2268+ uid = Getuid()
2269+ }
2270+
2271+ if uid == 0 {
2272+ if mode&1 == 0 {
2273+ // Root can read and write any file.
2274+ return nil
2275+ }
2276+ if st.Mode&0111 != 0 {
2277+ // Root can execute any file that anybody can execute.
2278+ return nil
2279+ }
2280+ return EACCES
2281+ }
2282+
2283+ var fmode uint32
2284+ if uint32(uid) == st.Uid {
2285+ fmode = (st.Mode >> 6) & 7
2286+ } else {
2287+ var gid int
2288+ if flags&AT_EACCESS != 0 {
2289+ gid = Getegid()
2290+ } else {
2291+ gid = Getgid()
2292+ }
2293+
2294+ if uint32(gid) == st.Gid || isGroupMember(gid) {
2295+ fmode = (st.Mode >> 3) & 7
2296+ } else {
2297+ fmode = st.Mode & 7
2298+ }
2299+ }
2300+
2301+ if fmode&mode == mode {
2302+ return nil
2303+ }
2304+
2305+ return EACCES
2306+}
2307+
2308+//sys nameToHandleAt(dirFD int, pathname string, fh *fileHandle, mountID *_C_int, flags int) (err error) = SYS_NAME_TO_HANDLE_AT
2309+//sys openByHandleAt(mountFD int, fh *fileHandle, flags int) (fd int, err error) = SYS_OPEN_BY_HANDLE_AT
2310+
2311+// fileHandle is the argument to nameToHandleAt and openByHandleAt. We
2312+// originally tried to generate it via unix/linux/types.go with "type
2313+// fileHandle C.struct_file_handle" but that generated empty structs
2314+// for mips64 and mips64le. Instead, hard code it for now (it's the
2315+// same everywhere else) until the mips64 generator issue is fixed.
2316+type fileHandle struct {
2317+ Bytes uint32
2318+ Type int32
2319+}
2320+
2321+// FileHandle represents the C struct file_handle used by
2322+// name_to_handle_at (see NameToHandleAt) and open_by_handle_at (see
2323+// OpenByHandleAt).
2324+type FileHandle struct {
2325+ *fileHandle
2326+}
2327+
2328+// NewFileHandle constructs a FileHandle.
2329+func NewFileHandle(handleType int32, handle []byte) FileHandle {
2330+ const hdrSize = unsafe.Sizeof(fileHandle{})
2331+ buf := make([]byte, hdrSize+uintptr(len(handle)))
2332+ copy(buf[hdrSize:], handle)
2333+ fh := (*fileHandle)(unsafe.Pointer(&buf[0]))
2334+ fh.Type = handleType
2335+ fh.Bytes = uint32(len(handle))
2336+ return FileHandle{fh}
2337+}
2338+
2339+func (fh *FileHandle) Size() int { return int(fh.fileHandle.Bytes) }
2340+func (fh *FileHandle) Type() int32 { return fh.fileHandle.Type }
2341+func (fh *FileHandle) Bytes() []byte {
2342+ n := fh.Size()
2343+ if n == 0 {
2344+ return nil
2345+ }
2346+ return (*[1 << 30]byte)(unsafe.Pointer(uintptr(unsafe.Pointer(&fh.fileHandle.Type)) + 4))[:n:n]
2347+}
2348+
2349+// NameToHandleAt wraps the name_to_handle_at system call; it obtains
2350+// a handle for a path name.
2351+func NameToHandleAt(dirfd int, path string, flags int) (handle FileHandle, mountID int, err error) {
2352+ var mid _C_int
2353+ // Try first with a small buffer, assuming the handle will
2354+ // only be 32 bytes.
2355+ size := uint32(32 + unsafe.Sizeof(fileHandle{}))
2356+ didResize := false
2357+ for {
2358+ buf := make([]byte, size)
2359+ fh := (*fileHandle)(unsafe.Pointer(&buf[0]))
2360+ fh.Bytes = size - uint32(unsafe.Sizeof(fileHandle{}))
2361+ err = nameToHandleAt(dirfd, path, fh, &mid, flags)
2362+ if err == EOVERFLOW {
2363+ if didResize {
2364+ // We shouldn't need to resize more than once
2365+ return
2366+ }
2367+ didResize = true
2368+ size = fh.Bytes + uint32(unsafe.Sizeof(fileHandle{}))
2369+ continue
2370+ }
2371+ if err != nil {
2372+ return
2373+ }
2374+ return FileHandle{fh}, int(mid), nil
2375+ }
2376+}
2377+
2378+// OpenByHandleAt wraps the open_by_handle_at system call; it opens a
2379+// file via a handle as previously returned by NameToHandleAt.
2380+func OpenByHandleAt(mountFD int, handle FileHandle, flags int) (fd int, err error) {
2381+ return openByHandleAt(mountFD, handle.fileHandle, flags)
2382+}
2383+
2384+// Klogset wraps the sys_syslog system call; it sets console_loglevel to
2385+// the value specified by arg and passes a dummy pointer to bufp.
2386+func Klogset(typ int, arg int) (err error) {
2387+ var p unsafe.Pointer
2388+ _, _, errno := Syscall(SYS_SYSLOG, uintptr(typ), uintptr(p), uintptr(arg))
2389+ if errno != 0 {
2390+ return errnoErr(errno)
2391+ }
2392+ return nil
2393+}
2394+
2395+// RemoteIovec is Iovec with the pointer replaced with an integer.
2396+// It is used for ProcessVMReadv and ProcessVMWritev, where the pointer
2397+// refers to a location in a different process' address space, which
2398+// would confuse the Go garbage collector.
2399+type RemoteIovec struct {
2400+ Base uintptr
2401+ Len int
2402+}
2403+
2404+//sys ProcessVMReadv(pid int, localIov []Iovec, remoteIov []RemoteIovec, flags uint) (n int, err error) = SYS_PROCESS_VM_READV
2405+//sys ProcessVMWritev(pid int, localIov []Iovec, remoteIov []RemoteIovec, flags uint) (n int, err error) = SYS_PROCESS_VM_WRITEV
2406+
2407+/*
2408+ * Unimplemented
2409+ */
2410+// AfsSyscall
2411+// Alarm
2412+// ArchPrctl
2413+// Brk
2414+// ClockNanosleep
2415+// ClockSettime
2416+// Clone
2417+// EpollCtlOld
2418+// EpollPwait
2419+// EpollWaitOld
2420+// Execve
2421+// Fork
2422+// Futex
2423+// GetKernelSyms
2424+// GetMempolicy
2425+// GetRobustList
2426+// GetThreadArea
2427+// Getitimer
2428+// Getpmsg
2429+// IoCancel
2430+// IoDestroy
2431+// IoGetevents
2432+// IoSetup
2433+// IoSubmit
2434+// IoprioGet
2435+// IoprioSet
2436+// KexecLoad
2437+// LookupDcookie
2438+// Mbind
2439+// MigratePages
2440+// Mincore
2441+// ModifyLdt
2442+// Mount
2443+// MovePages
2444+// MqGetsetattr
2445+// MqNotify
2446+// MqOpen
2447+// MqTimedreceive
2448+// MqTimedsend
2449+// MqUnlink
2450+// Mremap
2451+// Msgctl
2452+// Msgget
2453+// Msgrcv
2454+// Msgsnd
2455+// Nfsservctl
2456+// Personality
2457+// Pselect6
2458+// Ptrace
2459+// Putpmsg
2460+// Quotactl
2461+// Readahead
2462+// Readv
2463+// RemapFilePages
2464+// RestartSyscall
2465+// RtSigaction
2466+// RtSigpending
2467+// RtSigprocmask
2468+// RtSigqueueinfo
2469+// RtSigreturn
2470+// RtSigsuspend
2471+// RtSigtimedwait
2472+// SchedGetPriorityMax
2473+// SchedGetPriorityMin
2474+// SchedGetparam
2475+// SchedGetscheduler
2476+// SchedRrGetInterval
2477+// SchedSetparam
2478+// SchedYield
2479+// Security
2480+// Semctl
2481+// Semget
2482+// Semop
2483+// Semtimedop
2484+// SetMempolicy
2485+// SetRobustList
2486+// SetThreadArea
2487+// SetTidAddress
2488+// Shmat
2489+// Shmctl
2490+// Shmdt
2491+// Shmget
2492+// Sigaltstack
2493+// Swapoff
2494+// Swapon
2495+// Sysfs
2496+// TimerCreate
2497+// TimerDelete
2498+// TimerGetoverrun
2499+// TimerGettime
2500+// TimerSettime
2501+// Tkill (obsolete)
2502+// Tuxcall
2503+// Umount2
2504+// Uselib
2505+// Utimensat
2506+// Vfork
2507+// Vhangup
2508+// Vserver
2509+// Waitid
2510+// _Sysctl