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
12package unix
13
14import (
15 "encoding/binary"
16 "strconv"
17 "syscall"
18 "time"
19 "unsafe"
20)
21
22/*
23 * Wrapped
24 */
25
26func Access(path string, mode uint32) (err error) {
27 return Faccessat(AT_FDCWD, path, mode, 0)
28}
29
30func Chmod(path string, mode uint32) (err error) {
31 return Fchmodat(AT_FDCWD, path, mode, 0)
32}
33
34func Chown(path string, uid int, gid int) (err error) {
35 return Fchownat(AT_FDCWD, path, uid, gid, 0)
36}
37
38func Creat(path string, mode uint32) (fd int, err error) {
39 return Open(path, O_CREAT|O_WRONLY|O_TRUNC, mode)
40}
41
42func EpollCreate(size int) (fd int, err error) {
43 if size <= 0 {
44 return -1, EINVAL
45 }
46 return EpollCreate1(0)
47}
48
49//sys FanotifyInit(flags uint, event_f_flags uint) (fd int, err error)
50//sys fanotifyMark(fd int, flags uint, mask uint64, dirFd int, pathname *byte) (err error)
51
52func FanotifyMark(fd int, flags uint, mask uint64, dirFd int, pathname string) (err error) {
53 if pathname == "" {
54 return fanotifyMark(fd, flags, mask, dirFd, nil)
55 }
56 p, err := BytePtrFromString(pathname)
57 if err != nil {
58 return err
59 }
60 return fanotifyMark(fd, flags, mask, dirFd, p)
61}
62
63//sys fchmodat(dirfd int, path string, mode uint32) (err error)
64//sys fchmodat2(dirfd int, path string, mode uint32, flags int) (err error)
65
66func Fchmodat(dirfd int, path string, mode uint32, flags int) error {
67 // Linux fchmodat doesn't support the flags parameter, but fchmodat2 does.
68 // Try fchmodat2 if flags are specified.
69 if flags != 0 {
70 err := fchmodat2(dirfd, path, mode, flags)
71 if err == ENOSYS {
72 // fchmodat2 isn't available. If the flags are known to be valid,
73 // return EOPNOTSUPP to indicate that fchmodat doesn't support them.
74 if flags&^(AT_SYMLINK_NOFOLLOW|AT_EMPTY_PATH) != 0 {
75 return EINVAL
76 } else if flags&(AT_SYMLINK_NOFOLLOW|AT_EMPTY_PATH) != 0 {
77 return EOPNOTSUPP
78 }
79 }
80 return err
81 }
82 return fchmodat(dirfd, path, mode)
83}
84
85func InotifyInit() (fd int, err error) {
86 return InotifyInit1(0)
87}
88
89//sys ioctl(fd int, req uint, arg uintptr) (err error) = SYS_IOCTL
90//sys ioctlPtr(fd int, req uint, arg unsafe.Pointer) (err error) = SYS_IOCTL
91
92// ioctl itself should not be exposed directly, but additional get/set functions
93// for specific types are permissible. These are defined in ioctl.go and
94// ioctl_linux.go.
95//
96// The third argument to ioctl is often a pointer but sometimes an integer.
97// Callers should use ioctlPtr when the third argument is a pointer and ioctl
98// when the third argument is an integer.
99//
100// TODO: some existing code incorrectly uses ioctl when it should use ioctlPtr.
101
102//sys Linkat(olddirfd int, oldpath string, newdirfd int, newpath string, flags int) (err error)
103
104func Link(oldpath string, newpath string) (err error) {
105 return Linkat(AT_FDCWD, oldpath, AT_FDCWD, newpath, 0)
106}
107
108func Mkdir(path string, mode uint32) (err error) {
109 return Mkdirat(AT_FDCWD, path, mode)
110}
111
112func Mknod(path string, mode uint32, dev int) (err error) {
113 return Mknodat(AT_FDCWD, path, mode, dev)
114}
115
116func Open(path string, mode int, perm uint32) (fd int, err error) {
117 return openat(AT_FDCWD, path, mode|O_LARGEFILE, perm)
118}
119
120//sys openat(dirfd int, path string, flags int, mode uint32) (fd int, err error)
121
122func Openat(dirfd int, path string, flags int, mode uint32) (fd int, err error) {
123 return openat(dirfd, path, flags|O_LARGEFILE, mode)
124}
125
126//sys openat2(dirfd int, path string, open_how *OpenHow, size int) (fd int, err error)
127
128func Openat2(dirfd int, path string, how *OpenHow) (fd int, err error) {
129 return openat2(dirfd, path, how, SizeofOpenHow)
130}
131
132func Pipe(p []int) error {
133 return Pipe2(p, 0)
134}
135
136//sysnb pipe2(p *[2]_C_int, flags int) (err error)
137
138func Pipe2(p []int, flags int) error {
139 if len(p) != 2 {
140 return EINVAL
141 }
142 var pp [2]_C_int
143 err := pipe2(&pp, flags)
144 if err == nil {
145 p[0] = int(pp[0])
146 p[1] = int(pp[1])
147 }
148 return err
149}
150
151//sys ppoll(fds *PollFd, nfds int, timeout *Timespec, sigmask *Sigset_t) (n int, err error)
152
153func Ppoll(fds []PollFd, timeout *Timespec, sigmask *Sigset_t) (n int, err error) {
154 if len(fds) == 0 {
155 return ppoll(nil, 0, timeout, sigmask)
156 }
157 return ppoll(&fds[0], len(fds), timeout, sigmask)
158}
159
160func Poll(fds []PollFd, timeout int) (n int, err error) {
161 var ts *Timespec
162 if timeout >= 0 {
163 ts = new(Timespec)
164 *ts = NsecToTimespec(int64(timeout) * 1e6)
165 }
166 return Ppoll(fds, ts, nil)
167}
168
169//sys Readlinkat(dirfd int, path string, buf []byte) (n int, err error)
170
171func Readlink(path string, buf []byte) (n int, err error) {
172 return Readlinkat(AT_FDCWD, path, buf)
173}
174
175func Rename(oldpath string, newpath string) (err error) {
176 return Renameat(AT_FDCWD, oldpath, AT_FDCWD, newpath)
177}
178
179func Rmdir(path string) error {
180 return Unlinkat(AT_FDCWD, path, AT_REMOVEDIR)
181}
182
183//sys Symlinkat(oldpath string, newdirfd int, newpath string) (err error)
184
185func Symlink(oldpath string, newpath string) (err error) {
186 return Symlinkat(oldpath, AT_FDCWD, newpath)
187}
188
189func Unlink(path string) error {
190 return Unlinkat(AT_FDCWD, path, 0)
191}
192
193//sys Unlinkat(dirfd int, path string, flags int) (err error)
194
195func Utimes(path string, tv []Timeval) error {
196 if tv == nil {
197 err := utimensat(AT_FDCWD, path, nil, 0)
198 if err != ENOSYS {
199 return err
200 }
201 return utimes(path, nil)
202 }
203 if len(tv) != 2 {
204 return EINVAL
205 }
206 var ts [2]Timespec
207 ts[0] = NsecToTimespec(TimevalToNsec(tv[0]))
208 ts[1] = NsecToTimespec(TimevalToNsec(tv[1]))
209 err := utimensat(AT_FDCWD, path, (*[2]Timespec)(unsafe.Pointer(&ts[0])), 0)
210 if err != ENOSYS {
211 return err
212 }
213 return utimes(path, (*[2]Timeval)(unsafe.Pointer(&tv[0])))
214}
215
216//sys utimensat(dirfd int, path string, times *[2]Timespec, flags int) (err error)
217
218func UtimesNano(path string, ts []Timespec) error {
219 return UtimesNanoAt(AT_FDCWD, path, ts, 0)
220}
221
222func UtimesNanoAt(dirfd int, path string, ts []Timespec, flags int) error {
223 if ts == nil {
224 return utimensat(dirfd, path, nil, flags)
225 }
226 if len(ts) != 2 {
227 return EINVAL
228 }
229 return utimensat(dirfd, path, (*[2]Timespec)(unsafe.Pointer(&ts[0])), flags)
230}
231
232func Futimesat(dirfd int, path string, tv []Timeval) error {
233 if tv == nil {
234 return futimesat(dirfd, path, nil)
235 }
236 if len(tv) != 2 {
237 return EINVAL
238 }
239 return futimesat(dirfd, path, (*[2]Timeval)(unsafe.Pointer(&tv[0])))
240}
241
242func Futimes(fd int, tv []Timeval) (err error) {
243 // Believe it or not, this is the best we can do on Linux
244 // (and is what glibc does).
245 return Utimes("/proc/self/fd/"+strconv.Itoa(fd), tv)
246}
247
248const ImplementsGetwd = true
249
250//sys Getcwd(buf []byte) (n int, err error)
251
252func Getwd() (wd string, err error) {
253 var buf [PathMax]byte
254 n, err := Getcwd(buf[0:])
255 if err != nil {
256 return "", err
257 }
258 // Getcwd returns the number of bytes written to buf, including the NUL.
259 if n < 1 || n > len(buf) || buf[n-1] != 0 {
260 return "", EINVAL
261 }
262 // In some cases, Linux can return a path that starts with the
263 // "(unreachable)" prefix, which can potentially be a valid relative
264 // path. To work around that, return ENOENT if path is not absolute.
265 if buf[0] != '/' {
266 return "", ENOENT
267 }
268
269 return string(buf[0 : n-1]), nil
270}
271
272func Getgroups() (gids []int, err error) {
273 n, err := getgroups(0, nil)
274 if err != nil {
275 return nil, err
276 }
277 if n == 0 {
278 return nil, nil
279 }
280
281 // Sanity check group count. Max is 1<<16 on Linux.
282 if n < 0 || n > 1<<20 {
283 return nil, EINVAL
284 }
285
286 a := make([]_Gid_t, n)
287 n, err = getgroups(n, &a[0])
288 if err != nil {
289 return nil, err
290 }
291 gids = make([]int, n)
292 for i, v := range a[0:n] {
293 gids[i] = int(v)
294 }
295 return
296}
297
298func Setgroups(gids []int) (err error) {
299 if len(gids) == 0 {
300 return setgroups(0, nil)
301 }
302
303 a := make([]_Gid_t, len(gids))
304 for i, v := range gids {
305 a[i] = _Gid_t(v)
306 }
307 return setgroups(len(a), &a[0])
308}
309
310type WaitStatus uint32
311
312// Wait status is 7 bits at bottom, either 0 (exited),
313// 0x7F (stopped), or a signal number that caused an exit.
314// The 0x80 bit is whether there was a core dump.
315// An extra number (exit code, signal causing a stop)
316// is in the high bits. At least that's the idea.
317// There are various irregularities. For example, the
318// "continued" status is 0xFFFF, distinguishing itself
319// from stopped via the core dump bit.
320
321const (
322 mask = 0x7F
323 core = 0x80
324 exited = 0x00
325 stopped = 0x7F
326 shift = 8
327)
328
329func (w WaitStatus) Exited() bool { return w&mask == exited }
330
331func (w WaitStatus) Signaled() bool { return w&mask != stopped && w&mask != exited }
332
333func (w WaitStatus) Stopped() bool { return w&0xFF == stopped }
334
335func (w WaitStatus) Continued() bool { return w == 0xFFFF }
336
337func (w WaitStatus) CoreDump() bool { return w.Signaled() && w&core != 0 }
338
339func (w WaitStatus) ExitStatus() int {
340 if !w.Exited() {
341 return -1
342 }
343 return int(w>>shift) & 0xFF
344}
345
346func (w WaitStatus) Signal() syscall.Signal {
347 if !w.Signaled() {
348 return -1
349 }
350 return syscall.Signal(w & mask)
351}
352
353func (w WaitStatus) StopSignal() syscall.Signal {
354 if !w.Stopped() {
355 return -1
356 }
357 return syscall.Signal(w>>shift) & 0xFF
358}
359
360func (w WaitStatus) TrapCause() int {
361 if w.StopSignal() != SIGTRAP {
362 return -1
363 }
364 return int(w>>shift) >> 8
365}
366
367//sys wait4(pid int, wstatus *_C_int, options int, rusage *Rusage) (wpid int, err error)
368
369func Wait4(pid int, wstatus *WaitStatus, options int, rusage *Rusage) (wpid int, err error) {
370 var status _C_int
371 wpid, err = wait4(pid, &status, options, rusage)
372 if wstatus != nil {
373 *wstatus = WaitStatus(status)
374 }
375 return
376}
377
378//sys Waitid(idType int, id int, info *Siginfo, options int, rusage *Rusage) (err error)
379
380func Mkfifo(path string, mode uint32) error {
381 return Mknod(path, mode|S_IFIFO, 0)
382}
383
384func Mkfifoat(dirfd int, path string, mode uint32) error {
385 return Mknodat(dirfd, path, mode|S_IFIFO, 0)
386}
387
388func (sa *SockaddrInet4) sockaddr() (unsafe.Pointer, _Socklen, error) {
389 if sa.Port < 0 || sa.Port > 0xFFFF {
390 return nil, 0, EINVAL
391 }
392 sa.raw.Family = AF_INET
393 p := (*[2]byte)(unsafe.Pointer(&sa.raw.Port))
394 p[0] = byte(sa.Port >> 8)
395 p[1] = byte(sa.Port)
396 sa.raw.Addr = sa.Addr
397 return unsafe.Pointer(&sa.raw), SizeofSockaddrInet4, nil
398}
399
400func (sa *SockaddrInet6) sockaddr() (unsafe.Pointer, _Socklen, error) {
401 if sa.Port < 0 || sa.Port > 0xFFFF {
402 return nil, 0, EINVAL
403 }
404 sa.raw.Family = AF_INET6
405 p := (*[2]byte)(unsafe.Pointer(&sa.raw.Port))
406 p[0] = byte(sa.Port >> 8)
407 p[1] = byte(sa.Port)
408 sa.raw.Scope_id = sa.ZoneId
409 sa.raw.Addr = sa.Addr
410 return unsafe.Pointer(&sa.raw), SizeofSockaddrInet6, nil
411}
412
413func (sa *SockaddrUnix) sockaddr() (unsafe.Pointer, _Socklen, error) {
414 name := sa.Name
415 n := len(name)
416 if n >= len(sa.raw.Path) {
417 return nil, 0, EINVAL
418 }
419 sa.raw.Family = AF_UNIX
420 for i := 0; i < n; i++ {
421 sa.raw.Path[i] = int8(name[i])
422 }
423 // length is family (uint16), name, NUL.
424 sl := _Socklen(2)
425 if n > 0 {
426 sl += _Socklen(n) + 1
427 }
428 if sa.raw.Path[0] == '@' || (sa.raw.Path[0] == 0 && sl > 3) {
429 // Check sl > 3 so we don't change unnamed socket behavior.
430 sa.raw.Path[0] = 0
431 // Don't count trailing NUL for abstract address.
432 sl--
433 }
434
435 return unsafe.Pointer(&sa.raw), sl, nil
436}
437
438// SockaddrLinklayer implements the Sockaddr interface for AF_PACKET type sockets.
439type SockaddrLinklayer struct {
440 Protocol uint16
441 Ifindex int
442 Hatype uint16
443 Pkttype uint8
444 Halen uint8
445 Addr [8]byte
446 raw RawSockaddrLinklayer
447}
448
449func (sa *SockaddrLinklayer) sockaddr() (unsafe.Pointer, _Socklen, error) {
450 if sa.Ifindex < 0 || sa.Ifindex > 0x7fffffff {
451 return nil, 0, EINVAL
452 }
453 sa.raw.Family = AF_PACKET
454 sa.raw.Protocol = sa.Protocol
455 sa.raw.Ifindex = int32(sa.Ifindex)
456 sa.raw.Hatype = sa.Hatype
457 sa.raw.Pkttype = sa.Pkttype
458 sa.raw.Halen = sa.Halen
459 sa.raw.Addr = sa.Addr
460 return unsafe.Pointer(&sa.raw), SizeofSockaddrLinklayer, nil
461}
462
463// SockaddrNetlink implements the Sockaddr interface for AF_NETLINK type sockets.
464type SockaddrNetlink struct {
465 Family uint16
466 Pad uint16
467 Pid uint32
468 Groups uint32
469 raw RawSockaddrNetlink
470}
471
472func (sa *SockaddrNetlink) sockaddr() (unsafe.Pointer, _Socklen, error) {
473 sa.raw.Family = AF_NETLINK
474 sa.raw.Pad = sa.Pad
475 sa.raw.Pid = sa.Pid
476 sa.raw.Groups = sa.Groups
477 return unsafe.Pointer(&sa.raw), SizeofSockaddrNetlink, nil
478}
479
480// SockaddrHCI implements the Sockaddr interface for AF_BLUETOOTH type sockets
481// using the HCI protocol.
482type SockaddrHCI struct {
483 Dev uint16
484 Channel uint16
485 raw RawSockaddrHCI
486}
487
488func (sa *SockaddrHCI) sockaddr() (unsafe.Pointer, _Socklen, error) {
489 sa.raw.Family = AF_BLUETOOTH
490 sa.raw.Dev = sa.Dev
491 sa.raw.Channel = sa.Channel
492 return unsafe.Pointer(&sa.raw), SizeofSockaddrHCI, nil
493}
494
495// SockaddrL2 implements the Sockaddr interface for AF_BLUETOOTH type sockets
496// using the L2CAP protocol.
497type SockaddrL2 struct {
498 PSM uint16
499 CID uint16
500 Addr [6]uint8
501 AddrType uint8
502 raw RawSockaddrL2
503}
504
505func (sa *SockaddrL2) sockaddr() (unsafe.Pointer, _Socklen, error) {
506 sa.raw.Family = AF_BLUETOOTH
507 psm := (*[2]byte)(unsafe.Pointer(&sa.raw.Psm))
508 psm[0] = byte(sa.PSM)
509 psm[1] = byte(sa.PSM >> 8)
510 for i := 0; i < len(sa.Addr); i++ {
511 sa.raw.Bdaddr[i] = sa.Addr[len(sa.Addr)-1-i]
512 }
513 cid := (*[2]byte)(unsafe.Pointer(&sa.raw.Cid))
514 cid[0] = byte(sa.CID)
515 cid[1] = byte(sa.CID >> 8)
516 sa.raw.Bdaddr_type = sa.AddrType
517 return unsafe.Pointer(&sa.raw), SizeofSockaddrL2, nil
518}
519
520// SockaddrRFCOMM implements the Sockaddr interface for AF_BLUETOOTH type sockets
521// using the RFCOMM protocol.
522//
523// Server example:
524//
525// fd, _ := Socket(AF_BLUETOOTH, SOCK_STREAM, BTPROTO_RFCOMM)
526// _ = unix.Bind(fd, &unix.SockaddrRFCOMM{
527// Channel: 1,
528// Addr: [6]uint8{0, 0, 0, 0, 0, 0}, // BDADDR_ANY or 00:00:00:00:00:00
529// })
530// _ = Listen(fd, 1)
531// nfd, sa, _ := Accept(fd)
532// fmt.Printf("conn addr=%v fd=%d", sa.(*unix.SockaddrRFCOMM).Addr, nfd)
533// Read(nfd, buf)
534//
535// Client example:
536//
537// fd, _ := Socket(AF_BLUETOOTH, SOCK_STREAM, BTPROTO_RFCOMM)
538// _ = Connect(fd, &SockaddrRFCOMM{
539// Channel: 1,
540// Addr: [6]byte{0x11, 0x22, 0x33, 0xaa, 0xbb, 0xcc}, // CC:BB:AA:33:22:11
541// })
542// Write(fd, []byte(`hello`))
543type SockaddrRFCOMM struct {
544 // Addr represents a bluetooth address, byte ordering is little-endian.
545 Addr [6]uint8
546
547 // Channel is a designated bluetooth channel, only 1-30 are available for use.
548 // Since Linux 2.6.7 and further zero value is the first available channel.
549 Channel uint8
550
551 raw RawSockaddrRFCOMM
552}
553
554func (sa *SockaddrRFCOMM) sockaddr() (unsafe.Pointer, _Socklen, error) {
555 sa.raw.Family = AF_BLUETOOTH
556 sa.raw.Channel = sa.Channel
557 sa.raw.Bdaddr = sa.Addr
558 return unsafe.Pointer(&sa.raw), SizeofSockaddrRFCOMM, nil
559}
560
561// SockaddrCAN implements the Sockaddr interface for AF_CAN type sockets.
562// The RxID and TxID fields are used for transport protocol addressing in
563// (CAN_TP16, CAN_TP20, CAN_MCNET, and CAN_ISOTP), they can be left with
564// zero values for CAN_RAW and CAN_BCM sockets as they have no meaning.
565//
566// The SockaddrCAN struct must be bound to the socket file descriptor
567// using Bind before the CAN socket can be used.
568//
569// // Read one raw CAN frame
570// fd, _ := Socket(AF_CAN, SOCK_RAW, CAN_RAW)
571// addr := &SockaddrCAN{Ifindex: index}
572// Bind(fd, addr)
573// frame := make([]byte, 16)
574// Read(fd, frame)
575//
576// The full SocketCAN documentation can be found in the linux kernel
577// archives at: https://www.kernel.org/doc/Documentation/networking/can.txt
578type SockaddrCAN struct {
579 Ifindex int
580 RxID uint32
581 TxID uint32
582 raw RawSockaddrCAN
583}
584
585func (sa *SockaddrCAN) sockaddr() (unsafe.Pointer, _Socklen, error) {
586 if sa.Ifindex < 0 || sa.Ifindex > 0x7fffffff {
587 return nil, 0, EINVAL
588 }
589 sa.raw.Family = AF_CAN
590 sa.raw.Ifindex = int32(sa.Ifindex)
591 rx := (*[4]byte)(unsafe.Pointer(&sa.RxID))
592 for i := 0; i < 4; i++ {
593 sa.raw.Addr[i] = rx[i]
594 }
595 tx := (*[4]byte)(unsafe.Pointer(&sa.TxID))
596 for i := 0; i < 4; i++ {
597 sa.raw.Addr[i+4] = tx[i]
598 }
599 return unsafe.Pointer(&sa.raw), SizeofSockaddrCAN, nil
600}
601
602// SockaddrCANJ1939 implements the Sockaddr interface for AF_CAN using J1939
603// protocol (https://en.wikipedia.org/wiki/SAE_J1939). For more information
604// on the purposes of the fields, check the official linux kernel documentation
605// available here: https://www.kernel.org/doc/Documentation/networking/j1939.rst
606type SockaddrCANJ1939 struct {
607 Ifindex int
608 Name uint64
609 PGN uint32
610 Addr uint8
611 raw RawSockaddrCAN
612}
613
614func (sa *SockaddrCANJ1939) sockaddr() (unsafe.Pointer, _Socklen, error) {
615 if sa.Ifindex < 0 || sa.Ifindex > 0x7fffffff {
616 return nil, 0, EINVAL
617 }
618 sa.raw.Family = AF_CAN
619 sa.raw.Ifindex = int32(sa.Ifindex)
620 n := (*[8]byte)(unsafe.Pointer(&sa.Name))
621 for i := 0; i < 8; i++ {
622 sa.raw.Addr[i] = n[i]
623 }
624 p := (*[4]byte)(unsafe.Pointer(&sa.PGN))
625 for i := 0; i < 4; i++ {
626 sa.raw.Addr[i+8] = p[i]
627 }
628 sa.raw.Addr[12] = sa.Addr
629 return unsafe.Pointer(&sa.raw), SizeofSockaddrCAN, nil
630}
631
632// SockaddrALG implements the Sockaddr interface for AF_ALG type sockets.
633// SockaddrALG enables userspace access to the Linux kernel's cryptography
634// subsystem. The Type and Name fields specify which type of hash or cipher
635// should be used with a given socket.
636//
637// To create a file descriptor that provides access to a hash or cipher, both
638// Bind and Accept must be used. Once the setup process is complete, input
639// data can be written to the socket, processed by the kernel, and then read
640// back as hash output or ciphertext.
641//
642// Here is an example of using an AF_ALG socket with SHA1 hashing.
643// The initial socket setup process is as follows:
644//
645// // Open a socket to perform SHA1 hashing.
646// fd, _ := unix.Socket(unix.AF_ALG, unix.SOCK_SEQPACKET, 0)
647// addr := &unix.SockaddrALG{Type: "hash", Name: "sha1"}
648// unix.Bind(fd, addr)
649// // Note: unix.Accept does not work at this time; must invoke accept()
650// // manually using unix.Syscall.
651// hashfd, _, _ := unix.Syscall(unix.SYS_ACCEPT, uintptr(fd), 0, 0)
652//
653// Once a file descriptor has been returned from Accept, it may be used to
654// perform SHA1 hashing. The descriptor is not safe for concurrent use, but
655// may be re-used repeatedly with subsequent Write and Read operations.
656//
657// When hashing a small byte slice or string, a single Write and Read may
658// be used:
659//
660// // Assume hashfd is already configured using the setup process.
661// hash := os.NewFile(hashfd, "sha1")
662// // Hash an input string and read the results. Each Write discards
663// // previous hash state. Read always reads the current state.
664// b := make([]byte, 20)
665// for i := 0; i < 2; i++ {
666// io.WriteString(hash, "Hello, world.")
667// hash.Read(b)
668// fmt.Println(hex.EncodeToString(b))
669// }
670// // Output:
671// // 2ae01472317d1935a84797ec1983ae243fc6aa28
672// // 2ae01472317d1935a84797ec1983ae243fc6aa28
673//
674// For hashing larger byte slices, or byte streams such as those read from
675// a file or socket, use Sendto with MSG_MORE to instruct the kernel to update
676// the hash digest instead of creating a new one for a given chunk and finalizing it.
677//
678// // Assume hashfd and addr are already configured using the setup process.
679// hash := os.NewFile(hashfd, "sha1")
680// // Hash the contents of a file.
681// f, _ := os.Open("/tmp/linux-4.10-rc7.tar.xz")
682// b := make([]byte, 4096)
683// for {
684// n, err := f.Read(b)
685// if err == io.EOF {
686// break
687// }
688// unix.Sendto(hashfd, b[:n], unix.MSG_MORE, addr)
689// }
690// hash.Read(b)
691// fmt.Println(hex.EncodeToString(b))
692// // Output: 85cdcad0c06eef66f805ecce353bec9accbeecc5
693//
694// For more information, see: http://www.chronox.de/crypto-API/crypto/userspace-if.html.
695type SockaddrALG struct {
696 Type string
697 Name string
698 Feature uint32
699 Mask uint32
700 raw RawSockaddrALG
701}
702
703func (sa *SockaddrALG) sockaddr() (unsafe.Pointer, _Socklen, error) {
704 // Leave room for NUL byte terminator.
705 if len(sa.Type) > len(sa.raw.Type)-1 {
706 return nil, 0, EINVAL
707 }
708 if len(sa.Name) > len(sa.raw.Name)-1 {
709 return nil, 0, EINVAL
710 }
711
712 sa.raw.Family = AF_ALG
713 sa.raw.Feat = sa.Feature
714 sa.raw.Mask = sa.Mask
715
716 copy(sa.raw.Type[:], sa.Type)
717 copy(sa.raw.Name[:], sa.Name)
718
719 return unsafe.Pointer(&sa.raw), SizeofSockaddrALG, nil
720}
721
722// SockaddrVM implements the Sockaddr interface for AF_VSOCK type sockets.
723// SockaddrVM provides access to Linux VM sockets: a mechanism that enables
724// bidirectional communication between a hypervisor and its guest virtual
725// machines.
726type SockaddrVM struct {
727 // CID and Port specify a context ID and port address for a VM socket.
728 // Guests have a unique CID, and hosts may have a well-known CID of:
729 // - VMADDR_CID_HYPERVISOR: refers to the hypervisor process.
730 // - VMADDR_CID_LOCAL: refers to local communication (loopback).
731 // - VMADDR_CID_HOST: refers to other processes on the host.
732 CID uint32
733 Port uint32
734 Flags uint8
735 raw RawSockaddrVM
736}
737
738func (sa *SockaddrVM) sockaddr() (unsafe.Pointer, _Socklen, error) {
739 sa.raw.Family = AF_VSOCK
740 sa.raw.Port = sa.Port
741 sa.raw.Cid = sa.CID
742 sa.raw.Flags = sa.Flags
743
744 return unsafe.Pointer(&sa.raw), SizeofSockaddrVM, nil
745}
746
747type SockaddrXDP struct {
748 Flags uint16
749 Ifindex uint32
750 QueueID uint32
751 SharedUmemFD uint32
752 raw RawSockaddrXDP
753}
754
755func (sa *SockaddrXDP) sockaddr() (unsafe.Pointer, _Socklen, error) {
756 sa.raw.Family = AF_XDP
757 sa.raw.Flags = sa.Flags
758 sa.raw.Ifindex = sa.Ifindex
759 sa.raw.Queue_id = sa.QueueID
760 sa.raw.Shared_umem_fd = sa.SharedUmemFD
761
762 return unsafe.Pointer(&sa.raw), SizeofSockaddrXDP, nil
763}
764
765// This constant mirrors the #define of PX_PROTO_OE in
766// linux/if_pppox.h. We're defining this by hand here instead of
767// autogenerating through mkerrors.sh because including
768// linux/if_pppox.h causes some declaration conflicts with other
769// includes (linux/if_pppox.h includes linux/in.h, which conflicts
770// with netinet/in.h). Given that we only need a single zero constant
771// out of that file, it's cleaner to just define it by hand here.
772const px_proto_oe = 0
773
774type SockaddrPPPoE struct {
775 SID uint16
776 Remote []byte
777 Dev string
778 raw RawSockaddrPPPoX
779}
780
781func (sa *SockaddrPPPoE) sockaddr() (unsafe.Pointer, _Socklen, error) {
782 if len(sa.Remote) != 6 {
783 return nil, 0, EINVAL
784 }
785 if len(sa.Dev) > IFNAMSIZ-1 {
786 return nil, 0, EINVAL
787 }
788
789 *(*uint16)(unsafe.Pointer(&sa.raw[0])) = AF_PPPOX
790 // This next field is in host-endian byte order. We can't use the
791 // same unsafe pointer cast as above, because this value is not
792 // 32-bit aligned and some architectures don't allow unaligned
793 // access.
794 //
795 // However, the value of px_proto_oe is 0, so we can use
796 // encoding/binary helpers to write the bytes without worrying
797 // about the ordering.
798 binary.BigEndian.PutUint32(sa.raw[2:6], px_proto_oe)
799 // This field is deliberately big-endian, unlike the previous
800 // one. The kernel expects SID to be in network byte order.
801 binary.BigEndian.PutUint16(sa.raw[6:8], sa.SID)
802 copy(sa.raw[8:14], sa.Remote)
803 for i := 14; i < 14+IFNAMSIZ; i++ {
804 sa.raw[i] = 0
805 }
806 copy(sa.raw[14:], sa.Dev)
807 return unsafe.Pointer(&sa.raw), SizeofSockaddrPPPoX, nil
808}
809
810// SockaddrTIPC implements the Sockaddr interface for AF_TIPC type sockets.
811// For more information on TIPC, see: http://tipc.sourceforge.net/.
812type SockaddrTIPC struct {
813 // Scope is the publication scopes when binding service/service range.
814 // Should be set to TIPC_CLUSTER_SCOPE or TIPC_NODE_SCOPE.
815 Scope int
816
817 // Addr is the type of address used to manipulate a socket. Addr must be
818 // one of:
819 // - *TIPCSocketAddr: "id" variant in the C addr union
820 // - *TIPCServiceRange: "nameseq" variant in the C addr union
821 // - *TIPCServiceName: "name" variant in the C addr union
822 //
823 // If nil, EINVAL will be returned when the structure is used.
824 Addr TIPCAddr
825
826 raw RawSockaddrTIPC
827}
828
829// TIPCAddr is implemented by types that can be used as an address for
830// SockaddrTIPC. It is only implemented by *TIPCSocketAddr, *TIPCServiceRange,
831// and *TIPCServiceName.
832type TIPCAddr interface {
833 tipcAddrtype() uint8
834 tipcAddr() [12]byte
835}
836
837func (sa *TIPCSocketAddr) tipcAddr() [12]byte {
838 var out [12]byte
839 copy(out[:], (*(*[unsafe.Sizeof(TIPCSocketAddr{})]byte)(unsafe.Pointer(sa)))[:])
840 return out
841}
842
843func (sa *TIPCSocketAddr) tipcAddrtype() uint8 { return TIPC_SOCKET_ADDR }
844
845func (sa *TIPCServiceRange) tipcAddr() [12]byte {
846 var out [12]byte
847 copy(out[:], (*(*[unsafe.Sizeof(TIPCServiceRange{})]byte)(unsafe.Pointer(sa)))[:])
848 return out
849}
850
851func (sa *TIPCServiceRange) tipcAddrtype() uint8 { return TIPC_SERVICE_RANGE }
852
853func (sa *TIPCServiceName) tipcAddr() [12]byte {
854 var out [12]byte
855 copy(out[:], (*(*[unsafe.Sizeof(TIPCServiceName{})]byte)(unsafe.Pointer(sa)))[:])
856 return out
857}
858
859func (sa *TIPCServiceName) tipcAddrtype() uint8 { return TIPC_SERVICE_ADDR }
860
861func (sa *SockaddrTIPC) sockaddr() (unsafe.Pointer, _Socklen, error) {
862 if sa.Addr == nil {
863 return nil, 0, EINVAL
864 }
865 sa.raw.Family = AF_TIPC
866 sa.raw.Scope = int8(sa.Scope)
867 sa.raw.Addrtype = sa.Addr.tipcAddrtype()
868 sa.raw.Addr = sa.Addr.tipcAddr()
869 return unsafe.Pointer(&sa.raw), SizeofSockaddrTIPC, nil
870}
871
872// SockaddrL2TPIP implements the Sockaddr interface for IPPROTO_L2TP/AF_INET sockets.
873type SockaddrL2TPIP struct {
874 Addr [4]byte
875 ConnId uint32
876 raw RawSockaddrL2TPIP
877}
878
879func (sa *SockaddrL2TPIP) sockaddr() (unsafe.Pointer, _Socklen, error) {
880 sa.raw.Family = AF_INET
881 sa.raw.Conn_id = sa.ConnId
882 sa.raw.Addr = sa.Addr
883 return unsafe.Pointer(&sa.raw), SizeofSockaddrL2TPIP, nil
884}
885
886// SockaddrL2TPIP6 implements the Sockaddr interface for IPPROTO_L2TP/AF_INET6 sockets.
887type SockaddrL2TPIP6 struct {
888 Addr [16]byte
889 ZoneId uint32
890 ConnId uint32
891 raw RawSockaddrL2TPIP6
892}
893
894func (sa *SockaddrL2TPIP6) sockaddr() (unsafe.Pointer, _Socklen, error) {
895 sa.raw.Family = AF_INET6
896 sa.raw.Conn_id = sa.ConnId
897 sa.raw.Scope_id = sa.ZoneId
898 sa.raw.Addr = sa.Addr
899 return unsafe.Pointer(&sa.raw), SizeofSockaddrL2TPIP6, nil
900}
901
902// SockaddrIUCV implements the Sockaddr interface for AF_IUCV sockets.
903type SockaddrIUCV struct {
904 UserID string
905 Name string
906 raw RawSockaddrIUCV
907}
908
909func (sa *SockaddrIUCV) sockaddr() (unsafe.Pointer, _Socklen, error) {
910 sa.raw.Family = AF_IUCV
911 // These are EBCDIC encoded by the kernel, but we still need to pad them
912 // with blanks. Initializing with blanks allows the caller to feed in either
913 // a padded or an unpadded string.
914 for i := 0; i < 8; i++ {
915 sa.raw.Nodeid[i] = ' '
916 sa.raw.User_id[i] = ' '
917 sa.raw.Name[i] = ' '
918 }
919 if len(sa.UserID) > 8 || len(sa.Name) > 8 {
920 return nil, 0, EINVAL
921 }
922 for i, b := range []byte(sa.UserID[:]) {
923 sa.raw.User_id[i] = int8(b)
924 }
925 for i, b := range []byte(sa.Name[:]) {
926 sa.raw.Name[i] = int8(b)
927 }
928 return unsafe.Pointer(&sa.raw), SizeofSockaddrIUCV, nil
929}
930
931type SockaddrNFC struct {
932 DeviceIdx uint32
933 TargetIdx uint32
934 NFCProtocol uint32
935 raw RawSockaddrNFC
936}
937
938func (sa *SockaddrNFC) sockaddr() (unsafe.Pointer, _Socklen, error) {
939 sa.raw.Sa_family = AF_NFC
940 sa.raw.Dev_idx = sa.DeviceIdx
941 sa.raw.Target_idx = sa.TargetIdx
942 sa.raw.Nfc_protocol = sa.NFCProtocol
943 return unsafe.Pointer(&sa.raw), SizeofSockaddrNFC, nil
944}
945
946type SockaddrNFCLLCP struct {
947 DeviceIdx uint32
948 TargetIdx uint32
949 NFCProtocol uint32
950 DestinationSAP uint8
951 SourceSAP uint8
952 ServiceName string
953 raw RawSockaddrNFCLLCP
954}
955
956func (sa *SockaddrNFCLLCP) sockaddr() (unsafe.Pointer, _Socklen, error) {
957 sa.raw.Sa_family = AF_NFC
958 sa.raw.Dev_idx = sa.DeviceIdx
959 sa.raw.Target_idx = sa.TargetIdx
960 sa.raw.Nfc_protocol = sa.NFCProtocol
961 sa.raw.Dsap = sa.DestinationSAP
962 sa.raw.Ssap = sa.SourceSAP
963 if len(sa.ServiceName) > len(sa.raw.Service_name) {
964 return nil, 0, EINVAL
965 }
966 copy(sa.raw.Service_name[:], sa.ServiceName)
967 sa.raw.SetServiceNameLen(len(sa.ServiceName))
968 return unsafe.Pointer(&sa.raw), SizeofSockaddrNFCLLCP, nil
969}
970
971var socketProtocol = func(fd int) (int, error) {
972 return GetsockoptInt(fd, SOL_SOCKET, SO_PROTOCOL)
973}
974
975func anyToSockaddr(fd int, rsa *RawSockaddrAny) (Sockaddr, error) {
976 switch rsa.Addr.Family {
977 case AF_NETLINK:
978 pp := (*RawSockaddrNetlink)(unsafe.Pointer(rsa))
979 sa := new(SockaddrNetlink)
980 sa.Family = pp.Family
981 sa.Pad = pp.Pad
982 sa.Pid = pp.Pid
983 sa.Groups = pp.Groups
984 return sa, nil
985
986 case AF_PACKET:
987 pp := (*RawSockaddrLinklayer)(unsafe.Pointer(rsa))
988 sa := new(SockaddrLinklayer)
989 sa.Protocol = pp.Protocol
990 sa.Ifindex = int(pp.Ifindex)
991 sa.Hatype = pp.Hatype
992 sa.Pkttype = pp.Pkttype
993 sa.Halen = pp.Halen
994 sa.Addr = pp.Addr
995 return sa, nil
996
997 case AF_UNIX:
998 pp := (*RawSockaddrUnix)(unsafe.Pointer(rsa))
999 sa := new(SockaddrUnix)
1000 if pp.Path[0] == 0 {
1001 // "Abstract" Unix domain socket.
1002 // Rewrite leading NUL as @ for textual display.
1003 // (This is the standard convention.)
1004 // Not friendly to overwrite in place,
1005 // but the callers below don't care.
1006 pp.Path[0] = '@'
1007 }
1008
1009 // Assume path ends at NUL.
1010 // This is not technically the Linux semantics for
1011 // abstract Unix domain sockets--they are supposed
1012 // to be uninterpreted fixed-size binary blobs--but
1013 // everyone uses this convention.
1014 n := 0
1015 for n < len(pp.Path) && pp.Path[n] != 0 {
1016 n++
1017 }
1018 sa.Name = string(unsafe.Slice((*byte)(unsafe.Pointer(&pp.Path[0])), n))
1019 return sa, nil
1020
1021 case AF_INET:
1022 proto, err := socketProtocol(fd)
1023 if err != nil {
1024 return nil, err
1025 }
1026
1027 switch proto {
1028 case IPPROTO_L2TP:
1029 pp := (*RawSockaddrL2TPIP)(unsafe.Pointer(rsa))
1030 sa := new(SockaddrL2TPIP)
1031 sa.ConnId = pp.Conn_id
1032 sa.Addr = pp.Addr
1033 return sa, nil
1034 default:
1035 pp := (*RawSockaddrInet4)(unsafe.Pointer(rsa))
1036 sa := new(SockaddrInet4)
1037 p := (*[2]byte)(unsafe.Pointer(&pp.Port))
1038 sa.Port = int(p[0])<<8 + int(p[1])
1039 sa.Addr = pp.Addr
1040 return sa, nil
1041 }
1042
1043 case AF_INET6:
1044 proto, err := socketProtocol(fd)
1045 if err != nil {
1046 return nil, err
1047 }
1048
1049 switch proto {
1050 case IPPROTO_L2TP:
1051 pp := (*RawSockaddrL2TPIP6)(unsafe.Pointer(rsa))
1052 sa := new(SockaddrL2TPIP6)
1053 sa.ConnId = pp.Conn_id
1054 sa.ZoneId = pp.Scope_id
1055 sa.Addr = pp.Addr
1056 return sa, nil
1057 default:
1058 pp := (*RawSockaddrInet6)(unsafe.Pointer(rsa))
1059 sa := new(SockaddrInet6)
1060 p := (*[2]byte)(unsafe.Pointer(&pp.Port))
1061 sa.Port = int(p[0])<<8 + int(p[1])
1062 sa.ZoneId = pp.Scope_id
1063 sa.Addr = pp.Addr
1064 return sa, nil
1065 }
1066
1067 case AF_VSOCK:
1068 pp := (*RawSockaddrVM)(unsafe.Pointer(rsa))
1069 sa := &SockaddrVM{
1070 CID: pp.Cid,
1071 Port: pp.Port,
1072 Flags: pp.Flags,
1073 }
1074 return sa, nil
1075 case AF_BLUETOOTH:
1076 proto, err := socketProtocol(fd)
1077 if err != nil {
1078 return nil, err
1079 }
1080 // only BTPROTO_L2CAP and BTPROTO_RFCOMM can accept connections
1081 switch proto {
1082 case BTPROTO_L2CAP:
1083 pp := (*RawSockaddrL2)(unsafe.Pointer(rsa))
1084 sa := &SockaddrL2{
1085 PSM: pp.Psm,
1086 CID: pp.Cid,
1087 Addr: pp.Bdaddr,
1088 AddrType: pp.Bdaddr_type,
1089 }
1090 return sa, nil
1091 case BTPROTO_RFCOMM:
1092 pp := (*RawSockaddrRFCOMM)(unsafe.Pointer(rsa))
1093 sa := &SockaddrRFCOMM{
1094 Channel: pp.Channel,
1095 Addr: pp.Bdaddr,
1096 }
1097 return sa, nil
1098 }
1099 case AF_XDP:
1100 pp := (*RawSockaddrXDP)(unsafe.Pointer(rsa))
1101 sa := &SockaddrXDP{
1102 Flags: pp.Flags,
1103 Ifindex: pp.Ifindex,
1104 QueueID: pp.Queue_id,
1105 SharedUmemFD: pp.Shared_umem_fd,
1106 }
1107 return sa, nil
1108 case AF_PPPOX:
1109 pp := (*RawSockaddrPPPoX)(unsafe.Pointer(rsa))
1110 if binary.BigEndian.Uint32(pp[2:6]) != px_proto_oe {
1111 return nil, EINVAL
1112 }
1113 sa := &SockaddrPPPoE{
1114 SID: binary.BigEndian.Uint16(pp[6:8]),
1115 Remote: pp[8:14],
1116 }
1117 for i := 14; i < 14+IFNAMSIZ; i++ {
1118 if pp[i] == 0 {
1119 sa.Dev = string(pp[14:i])
1120 break
1121 }
1122 }
1123 return sa, nil
1124 case AF_TIPC:
1125 pp := (*RawSockaddrTIPC)(unsafe.Pointer(rsa))
1126
1127 sa := &SockaddrTIPC{
1128 Scope: int(pp.Scope),
1129 }
1130
1131 // Determine which union variant is present in pp.Addr by checking
1132 // pp.Addrtype.
1133 switch pp.Addrtype {
1134 case TIPC_SERVICE_RANGE:
1135 sa.Addr = (*TIPCServiceRange)(unsafe.Pointer(&pp.Addr))
1136 case TIPC_SERVICE_ADDR:
1137 sa.Addr = (*TIPCServiceName)(unsafe.Pointer(&pp.Addr))
1138 case TIPC_SOCKET_ADDR:
1139 sa.Addr = (*TIPCSocketAddr)(unsafe.Pointer(&pp.Addr))
1140 default:
1141 return nil, EINVAL
1142 }
1143
1144 return sa, nil
1145 case AF_IUCV:
1146 pp := (*RawSockaddrIUCV)(unsafe.Pointer(rsa))
1147
1148 var user [8]byte
1149 var name [8]byte
1150
1151 for i := 0; i < 8; i++ {
1152 user[i] = byte(pp.User_id[i])
1153 name[i] = byte(pp.Name[i])
1154 }
1155
1156 sa := &SockaddrIUCV{
1157 UserID: string(user[:]),
1158 Name: string(name[:]),
1159 }
1160 return sa, nil
1161
1162 case AF_CAN:
1163 proto, err := socketProtocol(fd)
1164 if err != nil {
1165 return nil, err
1166 }
1167
1168 pp := (*RawSockaddrCAN)(unsafe.Pointer(rsa))
1169
1170 switch proto {
1171 case CAN_J1939:
1172 sa := &SockaddrCANJ1939{
1173 Ifindex: int(pp.Ifindex),
1174 }
1175 name := (*[8]byte)(unsafe.Pointer(&sa.Name))
1176 for i := 0; i < 8; i++ {
1177 name[i] = pp.Addr[i]
1178 }
1179 pgn := (*[4]byte)(unsafe.Pointer(&sa.PGN))
1180 for i := 0; i < 4; i++ {
1181 pgn[i] = pp.Addr[i+8]
1182 }
1183 addr := (*[1]byte)(unsafe.Pointer(&sa.Addr))
1184 addr[0] = pp.Addr[12]
1185 return sa, nil
1186 default:
1187 sa := &SockaddrCAN{
1188 Ifindex: int(pp.Ifindex),
1189 }
1190 rx := (*[4]byte)(unsafe.Pointer(&sa.RxID))
1191 for i := 0; i < 4; i++ {
1192 rx[i] = pp.Addr[i]
1193 }
1194 tx := (*[4]byte)(unsafe.Pointer(&sa.TxID))
1195 for i := 0; i < 4; i++ {
1196 tx[i] = pp.Addr[i+4]
1197 }
1198 return sa, nil
1199 }
1200 case AF_NFC:
1201 proto, err := socketProtocol(fd)
1202 if err != nil {
1203 return nil, err
1204 }
1205 switch proto {
1206 case NFC_SOCKPROTO_RAW:
1207 pp := (*RawSockaddrNFC)(unsafe.Pointer(rsa))
1208 sa := &SockaddrNFC{
1209 DeviceIdx: pp.Dev_idx,
1210 TargetIdx: pp.Target_idx,
1211 NFCProtocol: pp.Nfc_protocol,
1212 }
1213 return sa, nil
1214 case NFC_SOCKPROTO_LLCP:
1215 pp := (*RawSockaddrNFCLLCP)(unsafe.Pointer(rsa))
1216 if uint64(pp.Service_name_len) > uint64(len(pp.Service_name)) {
1217 return nil, EINVAL
1218 }
1219 sa := &SockaddrNFCLLCP{
1220 DeviceIdx: pp.Dev_idx,
1221 TargetIdx: pp.Target_idx,
1222 NFCProtocol: pp.Nfc_protocol,
1223 DestinationSAP: pp.Dsap,
1224 SourceSAP: pp.Ssap,
1225 ServiceName: string(pp.Service_name[:pp.Service_name_len]),
1226 }
1227 return sa, nil
1228 default:
1229 return nil, EINVAL
1230 }
1231 }
1232 return nil, EAFNOSUPPORT
1233}
1234
1235func Accept(fd int) (nfd int, sa Sockaddr, err error) {
1236 var rsa RawSockaddrAny
1237 var len _Socklen = SizeofSockaddrAny
1238 nfd, err = accept4(fd, &rsa, &len, 0)
1239 if err != nil {
1240 return
1241 }
1242 sa, err = anyToSockaddr(fd, &rsa)
1243 if err != nil {
1244 Close(nfd)
1245 nfd = 0
1246 }
1247 return
1248}
1249
1250func Accept4(fd int, flags int) (nfd int, sa Sockaddr, err error) {
1251 var rsa RawSockaddrAny
1252 var len _Socklen = SizeofSockaddrAny
1253 nfd, err = accept4(fd, &rsa, &len, flags)
1254 if err != nil {
1255 return
1256 }
1257 if len > SizeofSockaddrAny {
1258 panic("RawSockaddrAny too small")
1259 }
1260 sa, err = anyToSockaddr(fd, &rsa)
1261 if err != nil {
1262 Close(nfd)
1263 nfd = 0
1264 }
1265 return
1266}
1267
1268func Getsockname(fd int) (sa Sockaddr, err error) {
1269 var rsa RawSockaddrAny
1270 var len _Socklen = SizeofSockaddrAny
1271 if err = getsockname(fd, &rsa, &len); err != nil {
1272 return
1273 }
1274 return anyToSockaddr(fd, &rsa)
1275}
1276
1277func GetsockoptIPMreqn(fd, level, opt int) (*IPMreqn, error) {
1278 var value IPMreqn
1279 vallen := _Socklen(SizeofIPMreqn)
1280 err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1281 return &value, err
1282}
1283
1284func GetsockoptUcred(fd, level, opt int) (*Ucred, error) {
1285 var value Ucred
1286 vallen := _Socklen(SizeofUcred)
1287 err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1288 return &value, err
1289}
1290
1291func GetsockoptTCPInfo(fd, level, opt int) (*TCPInfo, error) {
1292 var value TCPInfo
1293 vallen := _Socklen(SizeofTCPInfo)
1294 err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1295 return &value, err
1296}
1297
1298// GetsockoptTCPCCVegasInfo returns algorithm specific congestion control information for a socket using the "vegas"
1299// algorithm.
1300//
1301// The socket's congestion control algorighm can be retrieved via [GetsockoptString] with the [TCP_CONGESTION] option:
1302//
1303// algo, err := unix.GetsockoptString(fd, unix.IPPROTO_TCP, unix.TCP_CONGESTION)
1304func GetsockoptTCPCCVegasInfo(fd, level, opt int) (*TCPVegasInfo, error) {
1305 var value [SizeofTCPCCInfo / 4]uint32 // ensure proper alignment
1306 vallen := _Socklen(SizeofTCPCCInfo)
1307 err := getsockopt(fd, level, opt, unsafe.Pointer(&value[0]), &vallen)
1308 out := (*TCPVegasInfo)(unsafe.Pointer(&value[0]))
1309 return out, err
1310}
1311
1312// GetsockoptTCPCCDCTCPInfo returns algorithm specific congestion control information for a socket using the "dctp"
1313// algorithm.
1314//
1315// The socket's congestion control algorighm can be retrieved via [GetsockoptString] with the [TCP_CONGESTION] option:
1316//
1317// algo, err := unix.GetsockoptString(fd, unix.IPPROTO_TCP, unix.TCP_CONGESTION)
1318func GetsockoptTCPCCDCTCPInfo(fd, level, opt int) (*TCPDCTCPInfo, error) {
1319 var value [SizeofTCPCCInfo / 4]uint32 // ensure proper alignment
1320 vallen := _Socklen(SizeofTCPCCInfo)
1321 err := getsockopt(fd, level, opt, unsafe.Pointer(&value[0]), &vallen)
1322 out := (*TCPDCTCPInfo)(unsafe.Pointer(&value[0]))
1323 return out, err
1324}
1325
1326// GetsockoptTCPCCBBRInfo returns algorithm specific congestion control information for a socket using the "bbr"
1327// algorithm.
1328//
1329// The socket's congestion control algorighm can be retrieved via [GetsockoptString] with the [TCP_CONGESTION] option:
1330//
1331// algo, err := unix.GetsockoptString(fd, unix.IPPROTO_TCP, unix.TCP_CONGESTION)
1332func GetsockoptTCPCCBBRInfo(fd, level, opt int) (*TCPBBRInfo, error) {
1333 var value [SizeofTCPCCInfo / 4]uint32 // ensure proper alignment
1334 vallen := _Socklen(SizeofTCPCCInfo)
1335 err := getsockopt(fd, level, opt, unsafe.Pointer(&value[0]), &vallen)
1336 out := (*TCPBBRInfo)(unsafe.Pointer(&value[0]))
1337 return out, err
1338}
1339
1340// GetsockoptString returns the string value of the socket option opt for the
1341// socket associated with fd at the given socket level.
1342func GetsockoptString(fd, level, opt int) (string, error) {
1343 buf := make([]byte, 256)
1344 vallen := _Socklen(len(buf))
1345 err := getsockopt(fd, level, opt, unsafe.Pointer(&buf[0]), &vallen)
1346 if err != nil {
1347 if err == ERANGE {
1348 buf = make([]byte, vallen)
1349 err = getsockopt(fd, level, opt, unsafe.Pointer(&buf[0]), &vallen)
1350 }
1351 if err != nil {
1352 return "", err
1353 }
1354 }
1355 return ByteSliceToString(buf[:vallen]), nil
1356}
1357
1358func GetsockoptTpacketStats(fd, level, opt int) (*TpacketStats, error) {
1359 var value TpacketStats
1360 vallen := _Socklen(SizeofTpacketStats)
1361 err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1362 return &value, err
1363}
1364
1365func GetsockoptTpacketStatsV3(fd, level, opt int) (*TpacketStatsV3, error) {
1366 var value TpacketStatsV3
1367 vallen := _Socklen(SizeofTpacketStatsV3)
1368 err := getsockopt(fd, level, opt, unsafe.Pointer(&value), &vallen)
1369 return &value, err
1370}
1371
1372func SetsockoptIPMreqn(fd, level, opt int, mreq *IPMreqn) (err error) {
1373 return setsockopt(fd, level, opt, unsafe.Pointer(mreq), unsafe.Sizeof(*mreq))
1374}
1375
1376func SetsockoptPacketMreq(fd, level, opt int, mreq *PacketMreq) error {
1377 return setsockopt(fd, level, opt, unsafe.Pointer(mreq), unsafe.Sizeof(*mreq))
1378}
1379
1380// SetsockoptSockFprog attaches a classic BPF or an extended BPF program to a
1381// socket to filter incoming packets. See 'man 7 socket' for usage information.
1382func SetsockoptSockFprog(fd, level, opt int, fprog *SockFprog) error {
1383 return setsockopt(fd, level, opt, unsafe.Pointer(fprog), unsafe.Sizeof(*fprog))
1384}
1385
1386func SetsockoptCanRawFilter(fd, level, opt int, filter []CanFilter) error {
1387 var p unsafe.Pointer
1388 if len(filter) > 0 {
1389 p = unsafe.Pointer(&filter[0])
1390 }
1391 return setsockopt(fd, level, opt, p, uintptr(len(filter)*SizeofCanFilter))
1392}
1393
1394func SetsockoptTpacketReq(fd, level, opt int, tp *TpacketReq) error {
1395 return setsockopt(fd, level, opt, unsafe.Pointer(tp), unsafe.Sizeof(*tp))
1396}
1397
1398func SetsockoptTpacketReq3(fd, level, opt int, tp *TpacketReq3) error {
1399 return setsockopt(fd, level, opt, unsafe.Pointer(tp), unsafe.Sizeof(*tp))
1400}
1401
1402func SetsockoptTCPRepairOpt(fd, level, opt int, o []TCPRepairOpt) (err error) {
1403 if len(o) == 0 {
1404 return EINVAL
1405 }
1406 return setsockopt(fd, level, opt, unsafe.Pointer(&o[0]), uintptr(SizeofTCPRepairOpt*len(o)))
1407}
1408
1409func SetsockoptTCPMD5Sig(fd, level, opt int, s *TCPMD5Sig) error {
1410 return setsockopt(fd, level, opt, unsafe.Pointer(s), unsafe.Sizeof(*s))
1411}
1412
1413// Keyctl Commands (http://man7.org/linux/man-pages/man2/keyctl.2.html)
1414
1415// KeyctlInt calls keyctl commands in which each argument is an int.
1416// These commands are KEYCTL_REVOKE, KEYCTL_CHOWN, KEYCTL_CLEAR, KEYCTL_LINK,
1417// KEYCTL_UNLINK, KEYCTL_NEGATE, KEYCTL_SET_REQKEY_KEYRING, KEYCTL_SET_TIMEOUT,
1418// KEYCTL_ASSUME_AUTHORITY, KEYCTL_SESSION_TO_PARENT, KEYCTL_REJECT,
1419// KEYCTL_INVALIDATE, and KEYCTL_GET_PERSISTENT.
1420//sys KeyctlInt(cmd int, arg2 int, arg3 int, arg4 int, arg5 int) (ret int, err error) = SYS_KEYCTL
1421
1422// KeyctlBuffer calls keyctl commands in which the third and fourth
1423// arguments are a buffer and its length, respectively.
1424// These commands are KEYCTL_UPDATE, KEYCTL_READ, and KEYCTL_INSTANTIATE.
1425//sys KeyctlBuffer(cmd int, arg2 int, buf []byte, arg5 int) (ret int, err error) = SYS_KEYCTL
1426
1427// KeyctlString calls keyctl commands which return a string.
1428// These commands are KEYCTL_DESCRIBE and KEYCTL_GET_SECURITY.
1429func KeyctlString(cmd int, id int) (string, error) {
1430 // We must loop as the string data may change in between the syscalls.
1431 // We could allocate a large buffer here to reduce the chance that the
1432 // syscall needs to be called twice; however, this is unnecessary as
1433 // the performance loss is negligible.
1434 var buffer []byte
1435 for {
1436 // Try to fill the buffer with data
1437 length, err := KeyctlBuffer(cmd, id, buffer, 0)
1438 if err != nil {
1439 return "", err
1440 }
1441
1442 // Check if the data was written
1443 if length <= len(buffer) {
1444 // Exclude the null terminator
1445 return string(buffer[:length-1]), nil
1446 }
1447
1448 // Make a bigger buffer if needed
1449 buffer = make([]byte, length)
1450 }
1451}
1452
1453// Keyctl commands with special signatures.
1454
1455// KeyctlGetKeyringID implements the KEYCTL_GET_KEYRING_ID command.
1456// See the full documentation at:
1457// http://man7.org/linux/man-pages/man3/keyctl_get_keyring_ID.3.html
1458func KeyctlGetKeyringID(id int, create bool) (ringid int, err error) {
1459 createInt := 0
1460 if create {
1461 createInt = 1
1462 }
1463 return KeyctlInt(KEYCTL_GET_KEYRING_ID, id, createInt, 0, 0)
1464}
1465
1466// KeyctlSetperm implements the KEYCTL_SETPERM command. The perm value is the
1467// key handle permission mask as described in the "keyctl setperm" section of
1468// http://man7.org/linux/man-pages/man1/keyctl.1.html.
1469// See the full documentation at:
1470// http://man7.org/linux/man-pages/man3/keyctl_setperm.3.html
1471func KeyctlSetperm(id int, perm uint32) error {
1472 _, err := KeyctlInt(KEYCTL_SETPERM, id, int(perm), 0, 0)
1473 return err
1474}
1475
1476//sys keyctlJoin(cmd int, arg2 string) (ret int, err error) = SYS_KEYCTL
1477
1478// KeyctlJoinSessionKeyring implements the KEYCTL_JOIN_SESSION_KEYRING command.
1479// See the full documentation at:
1480// http://man7.org/linux/man-pages/man3/keyctl_join_session_keyring.3.html
1481func KeyctlJoinSessionKeyring(name string) (ringid int, err error) {
1482 return keyctlJoin(KEYCTL_JOIN_SESSION_KEYRING, name)
1483}
1484
1485//sys keyctlSearch(cmd int, arg2 int, arg3 string, arg4 string, arg5 int) (ret int, err error) = SYS_KEYCTL
1486
1487// KeyctlSearch implements the KEYCTL_SEARCH command.
1488// See the full documentation at:
1489// http://man7.org/linux/man-pages/man3/keyctl_search.3.html
1490func KeyctlSearch(ringid int, keyType, description string, destRingid int) (id int, err error) {
1491 return keyctlSearch(KEYCTL_SEARCH, ringid, keyType, description, destRingid)
1492}
1493
1494//sys keyctlIOV(cmd int, arg2 int, payload []Iovec, arg5 int) (err error) = SYS_KEYCTL
1495
1496// KeyctlInstantiateIOV implements the KEYCTL_INSTANTIATE_IOV command. This
1497// command is similar to KEYCTL_INSTANTIATE, except that the payload is a slice
1498// of Iovec (each of which represents a buffer) instead of a single buffer.
1499// See the full documentation at:
1500// http://man7.org/linux/man-pages/man3/keyctl_instantiate_iov.3.html
1501func KeyctlInstantiateIOV(id int, payload []Iovec, ringid int) error {
1502 return keyctlIOV(KEYCTL_INSTANTIATE_IOV, id, payload, ringid)
1503}
1504
1505//sys keyctlDH(cmd int, arg2 *KeyctlDHParams, buf []byte) (ret int, err error) = SYS_KEYCTL
1506
1507// KeyctlDHCompute implements the KEYCTL_DH_COMPUTE command. This command
1508// computes a Diffie-Hellman shared secret based on the provide params. The
1509// secret is written to the provided buffer and the returned size is the number
1510// of bytes written (returning an error if there is insufficient space in the
1511// buffer). If a nil buffer is passed in, this function returns the minimum
1512// buffer length needed to store the appropriate data. Note that this differs
1513// from KEYCTL_READ's behavior which always returns the requested payload size.
1514// See the full documentation at:
1515// http://man7.org/linux/man-pages/man3/keyctl_dh_compute.3.html
1516func KeyctlDHCompute(params *KeyctlDHParams, buffer []byte) (size int, err error) {
1517 return keyctlDH(KEYCTL_DH_COMPUTE, params, buffer)
1518}
1519
1520// KeyctlRestrictKeyring implements the KEYCTL_RESTRICT_KEYRING command. This
1521// command limits the set of keys that can be linked to the keyring, regardless
1522// of keyring permissions. The command requires the "setattr" permission.
1523//
1524// When called with an empty keyType the command locks the keyring, preventing
1525// any further keys from being linked to the keyring.
1526//
1527// The "asymmetric" keyType defines restrictions requiring key payloads to be
1528// DER encoded X.509 certificates signed by keys in another keyring. Restrictions
1529// for "asymmetric" include "builtin_trusted", "builtin_and_secondary_trusted",
1530// "key_or_keyring:<key>", and "key_or_keyring:<key>:chain".
1531//
1532// As of Linux 4.12, only the "asymmetric" keyType defines type-specific
1533// restrictions.
1534//
1535// See the full documentation at:
1536// http://man7.org/linux/man-pages/man3/keyctl_restrict_keyring.3.html
1537// http://man7.org/linux/man-pages/man2/keyctl.2.html
1538func KeyctlRestrictKeyring(ringid int, keyType string, restriction string) error {
1539 if keyType == "" {
1540 return keyctlRestrictKeyring(KEYCTL_RESTRICT_KEYRING, ringid)
1541 }
1542 return keyctlRestrictKeyringByType(KEYCTL_RESTRICT_KEYRING, ringid, keyType, restriction)
1543}
1544
1545//sys keyctlRestrictKeyringByType(cmd int, arg2 int, keyType string, restriction string) (err error) = SYS_KEYCTL
1546//sys keyctlRestrictKeyring(cmd int, arg2 int) (err error) = SYS_KEYCTL
1547
1548func recvmsgRaw(fd int, iov []Iovec, oob []byte, flags int, rsa *RawSockaddrAny) (n, oobn int, recvflags int, err error) {
1549 var msg Msghdr
1550 msg.Name = (*byte)(unsafe.Pointer(rsa))
1551 msg.Namelen = uint32(SizeofSockaddrAny)
1552 var dummy byte
1553 if len(oob) > 0 {
1554 if emptyIovecs(iov) {
1555 var sockType int
1556 sockType, err = GetsockoptInt(fd, SOL_SOCKET, SO_TYPE)
1557 if err != nil {
1558 return
1559 }
1560 // receive at least one normal byte
1561 if sockType != SOCK_DGRAM {
1562 var iova [1]Iovec
1563 iova[0].Base = &dummy
1564 iova[0].SetLen(1)
1565 iov = iova[:]
1566 }
1567 }
1568 msg.Control = &oob[0]
1569 msg.SetControllen(len(oob))
1570 }
1571 if len(iov) > 0 {
1572 msg.Iov = &iov[0]
1573 msg.SetIovlen(len(iov))
1574 }
1575 if n, err = recvmsg(fd, &msg, flags); err != nil {
1576 return
1577 }
1578 oobn = int(msg.Controllen)
1579 recvflags = int(msg.Flags)
1580 return
1581}
1582
1583func sendmsgN(fd int, iov []Iovec, oob []byte, ptr unsafe.Pointer, salen _Socklen, flags int) (n int, err error) {
1584 var msg Msghdr
1585 msg.Name = (*byte)(ptr)
1586 msg.Namelen = uint32(salen)
1587 var dummy byte
1588 var empty bool
1589 if len(oob) > 0 {
1590 empty = emptyIovecs(iov)
1591 if empty {
1592 var sockType int
1593 sockType, err = GetsockoptInt(fd, SOL_SOCKET, SO_TYPE)
1594 if err != nil {
1595 return 0, err
1596 }
1597 // send at least one normal byte
1598 if sockType != SOCK_DGRAM {
1599 var iova [1]Iovec
1600 iova[0].Base = &dummy
1601 iova[0].SetLen(1)
1602 iov = iova[:]
1603 }
1604 }
1605 msg.Control = &oob[0]
1606 msg.SetControllen(len(oob))
1607 }
1608 if len(iov) > 0 {
1609 msg.Iov = &iov[0]
1610 msg.SetIovlen(len(iov))
1611 }
1612 if n, err = sendmsg(fd, &msg, flags); err != nil {
1613 return 0, err
1614 }
1615 if len(oob) > 0 && empty {
1616 n = 0
1617 }
1618 return n, nil
1619}
1620
1621// BindToDevice binds the socket associated with fd to device.
1622func BindToDevice(fd int, device string) (err error) {
1623 return SetsockoptString(fd, SOL_SOCKET, SO_BINDTODEVICE, device)
1624}
1625
1626//sys ptrace(request int, pid int, addr uintptr, data uintptr) (err error)
1627//sys ptracePtr(request int, pid int, addr uintptr, data unsafe.Pointer) (err error) = SYS_PTRACE
1628
1629func ptracePeek(req int, pid int, addr uintptr, out []byte) (count int, err error) {
1630 // The peek requests are machine-size oriented, so we wrap it
1631 // to retrieve arbitrary-length data.
1632
1633 // The ptrace syscall differs from glibc's ptrace.
1634 // Peeks returns the word in *data, not as the return value.
1635
1636 var buf [SizeofPtr]byte
1637
1638 // Leading edge. PEEKTEXT/PEEKDATA don't require aligned
1639 // access (PEEKUSER warns that it might), but if we don't
1640 // align our reads, we might straddle an unmapped page
1641 // boundary and not get the bytes leading up to the page
1642 // boundary.
1643 n := 0
1644 if addr%SizeofPtr != 0 {
1645 err = ptracePtr(req, pid, addr-addr%SizeofPtr, unsafe.Pointer(&buf[0]))
1646 if err != nil {
1647 return 0, err
1648 }
1649 n += copy(out, buf[addr%SizeofPtr:])
1650 out = out[n:]
1651 }
1652
1653 // Remainder.
1654 for len(out) > 0 {
1655 // We use an internal buffer to guarantee alignment.
1656 // It's not documented if this is necessary, but we're paranoid.
1657 err = ptracePtr(req, pid, addr+uintptr(n), unsafe.Pointer(&buf[0]))
1658 if err != nil {
1659 return n, err
1660 }
1661 copied := copy(out, buf[0:])
1662 n += copied
1663 out = out[copied:]
1664 }
1665
1666 return n, nil
1667}
1668
1669func PtracePeekText(pid int, addr uintptr, out []byte) (count int, err error) {
1670 return ptracePeek(PTRACE_PEEKTEXT, pid, addr, out)
1671}
1672
1673func PtracePeekData(pid int, addr uintptr, out []byte) (count int, err error) {
1674 return ptracePeek(PTRACE_PEEKDATA, pid, addr, out)
1675}
1676
1677func PtracePeekUser(pid int, addr uintptr, out []byte) (count int, err error) {
1678 return ptracePeek(PTRACE_PEEKUSR, pid, addr, out)
1679}
1680
1681func ptracePoke(pokeReq int, peekReq int, pid int, addr uintptr, data []byte) (count int, err error) {
1682 // As for ptracePeek, we need to align our accesses to deal
1683 // with the possibility of straddling an invalid page.
1684
1685 // Leading edge.
1686 n := 0
1687 if addr%SizeofPtr != 0 {
1688 var buf [SizeofPtr]byte
1689 err = ptracePtr(peekReq, pid, addr-addr%SizeofPtr, unsafe.Pointer(&buf[0]))
1690 if err != nil {
1691 return 0, err
1692 }
1693 n += copy(buf[addr%SizeofPtr:], data)
1694 word := *((*uintptr)(unsafe.Pointer(&buf[0])))
1695 err = ptrace(pokeReq, pid, addr-addr%SizeofPtr, word)
1696 if err != nil {
1697 return 0, err
1698 }
1699 data = data[n:]
1700 }
1701
1702 // Interior.
1703 for len(data) > SizeofPtr {
1704 word := *((*uintptr)(unsafe.Pointer(&data[0])))
1705 err = ptrace(pokeReq, pid, addr+uintptr(n), word)
1706 if err != nil {
1707 return n, err
1708 }
1709 n += SizeofPtr
1710 data = data[SizeofPtr:]
1711 }
1712
1713 // Trailing edge.
1714 if len(data) > 0 {
1715 var buf [SizeofPtr]byte
1716 err = ptracePtr(peekReq, pid, addr+uintptr(n), unsafe.Pointer(&buf[0]))
1717 if err != nil {
1718 return n, err
1719 }
1720 copy(buf[0:], data)
1721 word := *((*uintptr)(unsafe.Pointer(&buf[0])))
1722 err = ptrace(pokeReq, pid, addr+uintptr(n), word)
1723 if err != nil {
1724 return n, err
1725 }
1726 n += len(data)
1727 }
1728
1729 return n, nil
1730}
1731
1732func PtracePokeText(pid int, addr uintptr, data []byte) (count int, err error) {
1733 return ptracePoke(PTRACE_POKETEXT, PTRACE_PEEKTEXT, pid, addr, data)
1734}
1735
1736func PtracePokeData(pid int, addr uintptr, data []byte) (count int, err error) {
1737 return ptracePoke(PTRACE_POKEDATA, PTRACE_PEEKDATA, pid, addr, data)
1738}
1739
1740func PtracePokeUser(pid int, addr uintptr, data []byte) (count int, err error) {
1741 return ptracePoke(PTRACE_POKEUSR, PTRACE_PEEKUSR, pid, addr, data)
1742}
1743
1744// elfNT_PRSTATUS is a copy of the debug/elf.NT_PRSTATUS constant so
1745// x/sys/unix doesn't need to depend on debug/elf and thus
1746// compress/zlib, debug/dwarf, and other packages.
1747const elfNT_PRSTATUS = 1
1748
1749func PtraceGetRegs(pid int, regsout *PtraceRegs) (err error) {
1750 var iov Iovec
1751 iov.Base = (*byte)(unsafe.Pointer(regsout))
1752 iov.SetLen(int(unsafe.Sizeof(*regsout)))
1753 return ptracePtr(PTRACE_GETREGSET, pid, uintptr(elfNT_PRSTATUS), unsafe.Pointer(&iov))
1754}
1755
1756func PtraceSetRegs(pid int, regs *PtraceRegs) (err error) {
1757 var iov Iovec
1758 iov.Base = (*byte)(unsafe.Pointer(regs))
1759 iov.SetLen(int(unsafe.Sizeof(*regs)))
1760 return ptracePtr(PTRACE_SETREGSET, pid, uintptr(elfNT_PRSTATUS), unsafe.Pointer(&iov))
1761}
1762
1763func PtraceSetOptions(pid int, options int) (err error) {
1764 return ptrace(PTRACE_SETOPTIONS, pid, 0, uintptr(options))
1765}
1766
1767func PtraceGetEventMsg(pid int) (msg uint, err error) {
1768 var data _C_long
1769 err = ptracePtr(PTRACE_GETEVENTMSG, pid, 0, unsafe.Pointer(&data))
1770 msg = uint(data)
1771 return
1772}
1773
1774func PtraceCont(pid int, signal int) (err error) {
1775 return ptrace(PTRACE_CONT, pid, 0, uintptr(signal))
1776}
1777
1778func PtraceSyscall(pid int, signal int) (err error) {
1779 return ptrace(PTRACE_SYSCALL, pid, 0, uintptr(signal))
1780}
1781
1782func PtraceSingleStep(pid int) (err error) { return ptrace(PTRACE_SINGLESTEP, pid, 0, 0) }
1783
1784func PtraceInterrupt(pid int) (err error) { return ptrace(PTRACE_INTERRUPT, pid, 0, 0) }
1785
1786func PtraceAttach(pid int) (err error) { return ptrace(PTRACE_ATTACH, pid, 0, 0) }
1787
1788func PtraceSeize(pid int) (err error) { return ptrace(PTRACE_SEIZE, pid, 0, 0) }
1789
1790func PtraceDetach(pid int) (err error) { return ptrace(PTRACE_DETACH, pid, 0, 0) }
1791
1792//sys reboot(magic1 uint, magic2 uint, cmd int, arg string) (err error)
1793
1794func Reboot(cmd int) (err error) {
1795 return reboot(LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, cmd, "")
1796}
1797
1798func direntIno(buf []byte) (uint64, bool) {
1799 return readInt(buf, unsafe.Offsetof(Dirent{}.Ino), unsafe.Sizeof(Dirent{}.Ino))
1800}
1801
1802func direntReclen(buf []byte) (uint64, bool) {
1803 return readInt(buf, unsafe.Offsetof(Dirent{}.Reclen), unsafe.Sizeof(Dirent{}.Reclen))
1804}
1805
1806func direntNamlen(buf []byte) (uint64, bool) {
1807 reclen, ok := direntReclen(buf)
1808 if !ok {
1809 return 0, false
1810 }
1811 return reclen - uint64(unsafe.Offsetof(Dirent{}.Name)), true
1812}
1813
1814//sys mount(source string, target string, fstype string, flags uintptr, data *byte) (err error)
1815
1816func Mount(source string, target string, fstype string, flags uintptr, data string) (err error) {
1817 // Certain file systems get rather angry and EINVAL if you give
1818 // them an empty string of data, rather than NULL.
1819 if data == "" {
1820 return mount(source, target, fstype, flags, nil)
1821 }
1822 datap, err := BytePtrFromString(data)
1823 if err != nil {
1824 return err
1825 }
1826 return mount(source, target, fstype, flags, datap)
1827}
1828
1829//sys mountSetattr(dirfd int, pathname string, flags uint, attr *MountAttr, size uintptr) (err error) = SYS_MOUNT_SETATTR
1830
1831// MountSetattr is a wrapper for mount_setattr(2).
1832// https://man7.org/linux/man-pages/man2/mount_setattr.2.html
1833//
1834// Requires kernel >= 5.12.
1835func MountSetattr(dirfd int, pathname string, flags uint, attr *MountAttr) error {
1836 return mountSetattr(dirfd, pathname, flags, attr, unsafe.Sizeof(*attr))
1837}
1838
1839func Sendfile(outfd int, infd int, offset *int64, count int) (written int, err error) {
1840 if raceenabled {
1841 raceReleaseMerge(unsafe.Pointer(&ioSync))
1842 }
1843 return sendfile(outfd, infd, offset, count)
1844}
1845
1846// Sendto
1847// Recvfrom
1848// Socketpair
1849
1850/*
1851 * Direct access
1852 */
1853//sys Acct(path string) (err error)
1854//sys AddKey(keyType string, description string, payload []byte, ringid int) (id int, err error)
1855//sys Adjtimex(buf *Timex) (state int, err error)
1856//sysnb Capget(hdr *CapUserHeader, data *CapUserData) (err error)
1857//sysnb Capset(hdr *CapUserHeader, data *CapUserData) (err error)
1858//sys Chdir(path string) (err error)
1859//sys Chroot(path string) (err error)
1860//sys ClockAdjtime(clockid int32, buf *Timex) (state int, err error)
1861//sys ClockGetres(clockid int32, res *Timespec) (err error)
1862//sys ClockGettime(clockid int32, time *Timespec) (err error)
1863//sys ClockSettime(clockid int32, time *Timespec) (err error)
1864//sys ClockNanosleep(clockid int32, flags int, request *Timespec, remain *Timespec) (err error)
1865//sys Close(fd int) (err error)
1866//sys CloseRange(first uint, last uint, flags uint) (err error)
1867//sys CopyFileRange(rfd int, roff *int64, wfd int, woff *int64, len int, flags int) (n int, err error)
1868//sys DeleteModule(name string, flags int) (err error)
1869//sys Dup(oldfd int) (fd int, err error)
1870
1871func Dup2(oldfd, newfd int) error {
1872 return Dup3(oldfd, newfd, 0)
1873}
1874
1875//sys Dup3(oldfd int, newfd int, flags int) (err error)
1876//sysnb EpollCreate1(flag int) (fd int, err error)
1877//sysnb EpollCtl(epfd int, op int, fd int, event *EpollEvent) (err error)
1878//sys Eventfd(initval uint, flags int) (fd int, err error) = SYS_EVENTFD2
1879//sys Exit(code int) = SYS_EXIT_GROUP
1880//sys Fallocate(fd int, mode uint32, off int64, len int64) (err error)
1881//sys Fchdir(fd int) (err error)
1882//sys Fchmod(fd int, mode uint32) (err error)
1883//sys Fchownat(dirfd int, path string, uid int, gid int, flags int) (err error)
1884//sys Fdatasync(fd int) (err error)
1885//sys Fgetxattr(fd int, attr string, dest []byte) (sz int, err error)
1886//sys FinitModule(fd int, params string, flags int) (err error)
1887//sys Flistxattr(fd int, dest []byte) (sz int, err error)
1888//sys Flock(fd int, how int) (err error)
1889//sys Fremovexattr(fd int, attr string) (err error)
1890//sys Fsetxattr(fd int, attr string, dest []byte, flags int) (err error)
1891//sys Fsync(fd int) (err error)
1892//sys Fsmount(fd int, flags int, mountAttrs int) (fsfd int, err error)
1893//sys Fsopen(fsName string, flags int) (fd int, err error)
1894//sys Fspick(dirfd int, pathName string, flags int) (fd int, err error)
1895
1896//sys fsconfig(fd int, cmd uint, key *byte, value *byte, aux int) (err error)
1897
1898func fsconfigCommon(fd int, cmd uint, key string, value *byte, aux int) (err error) {
1899 var keyp *byte
1900 if keyp, err = BytePtrFromString(key); err != nil {
1901 return
1902 }
1903 return fsconfig(fd, cmd, keyp, value, aux)
1904}
1905
1906// FsconfigSetFlag is equivalent to fsconfig(2) called
1907// with cmd == FSCONFIG_SET_FLAG.
1908//
1909// fd is the filesystem context to act upon.
1910// key the parameter key to set.
1911func FsconfigSetFlag(fd int, key string) (err error) {
1912 return fsconfigCommon(fd, FSCONFIG_SET_FLAG, key, nil, 0)
1913}
1914
1915// FsconfigSetString is equivalent to fsconfig(2) called
1916// with cmd == FSCONFIG_SET_STRING.
1917//
1918// fd is the filesystem context to act upon.
1919// key the parameter key to set.
1920// value is the parameter value to set.
1921func FsconfigSetString(fd int, key string, value string) (err error) {
1922 var valuep *byte
1923 if valuep, err = BytePtrFromString(value); err != nil {
1924 return
1925 }
1926 return fsconfigCommon(fd, FSCONFIG_SET_STRING, key, valuep, 0)
1927}
1928
1929// FsconfigSetBinary is equivalent to fsconfig(2) called
1930// with cmd == FSCONFIG_SET_BINARY.
1931//
1932// fd is the filesystem context to act upon.
1933// key the parameter key to set.
1934// value is the parameter value to set.
1935func FsconfigSetBinary(fd int, key string, value []byte) (err error) {
1936 if len(value) == 0 {
1937 return EINVAL
1938 }
1939 return fsconfigCommon(fd, FSCONFIG_SET_BINARY, key, &value[0], len(value))
1940}
1941
1942// FsconfigSetPath is equivalent to fsconfig(2) called
1943// with cmd == FSCONFIG_SET_PATH.
1944//
1945// fd is the filesystem context to act upon.
1946// key the parameter key to set.
1947// path is a non-empty path for specified key.
1948// atfd is a file descriptor at which to start lookup from or AT_FDCWD.
1949func FsconfigSetPath(fd int, key string, path string, atfd int) (err error) {
1950 var valuep *byte
1951 if valuep, err = BytePtrFromString(path); err != nil {
1952 return
1953 }
1954 return fsconfigCommon(fd, FSCONFIG_SET_PATH, key, valuep, atfd)
1955}
1956
1957// FsconfigSetPathEmpty is equivalent to fsconfig(2) called
1958// with cmd == FSCONFIG_SET_PATH_EMPTY. The same as
1959// FconfigSetPath but with AT_PATH_EMPTY implied.
1960func FsconfigSetPathEmpty(fd int, key string, path string, atfd int) (err error) {
1961 var valuep *byte
1962 if valuep, err = BytePtrFromString(path); err != nil {
1963 return
1964 }
1965 return fsconfigCommon(fd, FSCONFIG_SET_PATH_EMPTY, key, valuep, atfd)
1966}
1967
1968// FsconfigSetFd is equivalent to fsconfig(2) called
1969// with cmd == FSCONFIG_SET_FD.
1970//
1971// fd is the filesystem context to act upon.
1972// key the parameter key to set.
1973// value is a file descriptor to be assigned to specified key.
1974func FsconfigSetFd(fd int, key string, value int) (err error) {
1975 return fsconfigCommon(fd, FSCONFIG_SET_FD, key, nil, value)
1976}
1977
1978// FsconfigCreate is equivalent to fsconfig(2) called
1979// with cmd == FSCONFIG_CMD_CREATE.
1980//
1981// fd is the filesystem context to act upon.
1982func FsconfigCreate(fd int) (err error) {
1983 return fsconfig(fd, FSCONFIG_CMD_CREATE, nil, nil, 0)
1984}
1985
1986// FsconfigReconfigure is equivalent to fsconfig(2) called
1987// with cmd == FSCONFIG_CMD_RECONFIGURE.
1988//
1989// fd is the filesystem context to act upon.
1990func FsconfigReconfigure(fd int) (err error) {
1991 return fsconfig(fd, FSCONFIG_CMD_RECONFIGURE, nil, nil, 0)
1992}
1993
1994//sys Getdents(fd int, buf []byte) (n int, err error) = SYS_GETDENTS64
1995//sysnb Getpgid(pid int) (pgid int, err error)
1996
1997func Getpgrp() (pid int) {
1998 pid, _ = Getpgid(0)
1999 return
2000}
2001
2002//sysnb Getpid() (pid int)
2003//sysnb Getppid() (ppid int)
2004//sys Getpriority(which int, who int) (prio int, err error)
2005
2006func Getrandom(buf []byte, flags int) (n int, err error) {
2007 vdsoRet, supported := vgetrandom(buf, uint32(flags))
2008 if supported {
2009 if vdsoRet < 0 {
2010 return 0, errnoErr(syscall.Errno(-vdsoRet))
2011 }
2012 return vdsoRet, nil
2013 }
2014 var p *byte
2015 if len(buf) > 0 {
2016 p = &buf[0]
2017 }
2018 r, _, e := Syscall(SYS_GETRANDOM, uintptr(unsafe.Pointer(p)), uintptr(len(buf)), uintptr(flags))
2019 if e != 0 {
2020 return 0, errnoErr(e)
2021 }
2022 return int(r), nil
2023}
2024
2025//sysnb Getrusage(who int, rusage *Rusage) (err error)
2026//sysnb Getsid(pid int) (sid int, err error)
2027//sysnb Gettid() (tid int)
2028//sys Getxattr(path string, attr string, dest []byte) (sz int, err error)
2029//sys InitModule(moduleImage []byte, params string) (err error)
2030//sys InotifyAddWatch(fd int, pathname string, mask uint32) (watchdesc int, err error)
2031//sysnb InotifyInit1(flags int) (fd int, err error)
2032//sysnb InotifyRmWatch(fd int, watchdesc uint32) (success int, err error)
2033//sysnb Kill(pid int, sig syscall.Signal) (err error)
2034//sys Klogctl(typ int, buf []byte) (n int, err error) = SYS_SYSLOG
2035//sys Lgetxattr(path string, attr string, dest []byte) (sz int, err error)
2036//sys Listxattr(path string, dest []byte) (sz int, err error)
2037//sys Llistxattr(path string, dest []byte) (sz int, err error)
2038//sys Lremovexattr(path string, attr string) (err error)
2039//sys Lsetxattr(path string, attr string, data []byte, flags int) (err error)
2040//sys MemfdCreate(name string, flags int) (fd int, err error)
2041//sys Mkdirat(dirfd int, path string, mode uint32) (err error)
2042//sys Mknodat(dirfd int, path string, mode uint32, dev int) (err error)
2043//sys MoveMount(fromDirfd int, fromPathName string, toDirfd int, toPathName string, flags int) (err error)
2044//sys Nanosleep(time *Timespec, leftover *Timespec) (err error)
2045//sys OpenTree(dfd int, fileName string, flags uint) (r int, err error)
2046//sys PerfEventOpen(attr *PerfEventAttr, pid int, cpu int, groupFd int, flags int) (fd int, err error)
2047//sys PivotRoot(newroot string, putold string) (err error) = SYS_PIVOT_ROOT
2048//sys Prctl(option int, arg2 uintptr, arg3 uintptr, arg4 uintptr, arg5 uintptr) (err error)
2049//sys pselect6(nfd int, r *FdSet, w *FdSet, e *FdSet, timeout *Timespec, sigmask *sigset_argpack) (n int, err error)
2050//sys read(fd int, p []byte) (n int, err error)
2051//sys Removexattr(path string, attr string) (err error)
2052//sys Renameat2(olddirfd int, oldpath string, newdirfd int, newpath string, flags uint) (err error)
2053//sys RequestKey(keyType string, description string, callback string, destRingid int) (id int, err error)
2054//sys Setdomainname(p []byte) (err error)
2055//sys Sethostname(p []byte) (err error)
2056//sysnb Setpgid(pid int, pgid int) (err error)
2057//sysnb Setsid() (pid int, err error)
2058//sysnb Settimeofday(tv *Timeval) (err error)
2059//sys Setns(fd int, nstype int) (err error)
2060
2061//go:linkname syscall_prlimit syscall.prlimit
2062func syscall_prlimit(pid, resource int, newlimit, old *syscall.Rlimit) error
2063
2064func Prlimit(pid, resource int, newlimit, old *Rlimit) error {
2065 // Just call the syscall version, because as of Go 1.21
2066 // it will affect starting a new process.
2067 return syscall_prlimit(pid, resource, (*syscall.Rlimit)(newlimit), (*syscall.Rlimit)(old))
2068}
2069
2070// PrctlRetInt performs a prctl operation specified by option and further
2071// optional arguments arg2 through arg5 depending on option. It returns a
2072// non-negative integer that is returned by the prctl syscall.
2073func PrctlRetInt(option int, arg2 uintptr, arg3 uintptr, arg4 uintptr, arg5 uintptr) (int, error) {
2074 ret, _, err := Syscall6(SYS_PRCTL, uintptr(option), uintptr(arg2), uintptr(arg3), uintptr(arg4), uintptr(arg5), 0)
2075 if err != 0 {
2076 return 0, err
2077 }
2078 return int(ret), nil
2079}
2080
2081func Setuid(uid int) (err error) {
2082 return syscall.Setuid(uid)
2083}
2084
2085func Setgid(gid int) (err error) {
2086 return syscall.Setgid(gid)
2087}
2088
2089func Setreuid(ruid, euid int) (err error) {
2090 return syscall.Setreuid(ruid, euid)
2091}
2092
2093func Setregid(rgid, egid int) (err error) {
2094 return syscall.Setregid(rgid, egid)
2095}
2096
2097func Setresuid(ruid, euid, suid int) (err error) {
2098 return syscall.Setresuid(ruid, euid, suid)
2099}
2100
2101func Setresgid(rgid, egid, sgid int) (err error) {
2102 return syscall.Setresgid(rgid, egid, sgid)
2103}
2104
2105// SetfsgidRetGid sets fsgid for current thread and returns previous fsgid set.
2106// setfsgid(2) will return a non-nil error only if its caller lacks CAP_SETUID capability.
2107// If the call fails due to other reasons, current fsgid will be returned.
2108func SetfsgidRetGid(gid int) (int, error) {
2109 return setfsgid(gid)
2110}
2111
2112// SetfsuidRetUid sets fsuid for current thread and returns previous fsuid set.
2113// setfsgid(2) will return a non-nil error only if its caller lacks CAP_SETUID capability
2114// If the call fails due to other reasons, current fsuid will be returned.
2115func SetfsuidRetUid(uid int) (int, error) {
2116 return setfsuid(uid)
2117}
2118
2119func Setfsgid(gid int) error {
2120 _, err := setfsgid(gid)
2121 return err
2122}
2123
2124func Setfsuid(uid int) error {
2125 _, err := setfsuid(uid)
2126 return err
2127}
2128
2129func Signalfd(fd int, sigmask *Sigset_t, flags int) (newfd int, err error) {
2130 return signalfd(fd, sigmask, _C__NSIG/8, flags)
2131}
2132
2133//sys Setpriority(which int, who int, prio int) (err error)
2134//sys Setxattr(path string, attr string, data []byte, flags int) (err error)
2135//sys signalfd(fd int, sigmask *Sigset_t, maskSize uintptr, flags int) (newfd int, err error) = SYS_SIGNALFD4
2136//sys Statx(dirfd int, path string, flags int, mask int, stat *Statx_t) (err error)
2137//sys Sync()
2138//sys Syncfs(fd int) (err error)
2139//sysnb Sysinfo(info *Sysinfo_t) (err error)
2140//sys Tee(rfd int, wfd int, len int, flags int) (n int64, err error)
2141//sysnb TimerfdCreate(clockid int, flags int) (fd int, err error)
2142//sysnb TimerfdGettime(fd int, currValue *ItimerSpec) (err error)
2143//sysnb TimerfdSettime(fd int, flags int, newValue *ItimerSpec, oldValue *ItimerSpec) (err error)
2144//sysnb Tgkill(tgid int, tid int, sig syscall.Signal) (err error)
2145//sysnb Times(tms *Tms) (ticks uintptr, err error)
2146//sysnb Umask(mask int) (oldmask int)
2147//sysnb Uname(buf *Utsname) (err error)
2148//sys Unmount(target string, flags int) (err error) = SYS_UMOUNT2
2149//sys Unshare(flags int) (err error)
2150//sys write(fd int, p []byte) (n int, err error)
2151//sys exitThread(code int) (err error) = SYS_EXIT
2152//sys readv(fd int, iovs []Iovec) (n int, err error) = SYS_READV
2153//sys writev(fd int, iovs []Iovec) (n int, err error) = SYS_WRITEV
2154//sys preadv(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr) (n int, err error) = SYS_PREADV
2155//sys pwritev(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr) (n int, err error) = SYS_PWRITEV
2156//sys preadv2(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr, flags int) (n int, err error) = SYS_PREADV2
2157//sys pwritev2(fd int, iovs []Iovec, offs_l uintptr, offs_h uintptr, flags int) (n int, err error) = SYS_PWRITEV2
2158
2159// minIovec is the size of the small initial allocation used by
2160// Readv, Writev, etc.
2161//
2162// This small allocation gets stack allocated, which lets the
2163// common use case of len(iovs) <= minIovs avoid more expensive
2164// heap allocations.
2165const minIovec = 8
2166
2167// appendBytes converts bs to Iovecs and appends them to vecs.
2168func appendBytes(vecs []Iovec, bs [][]byte) []Iovec {
2169 for _, b := range bs {
2170 var v Iovec
2171 v.SetLen(len(b))
2172 if len(b) > 0 {
2173 v.Base = &b[0]
2174 } else {
2175 v.Base = (*byte)(unsafe.Pointer(&_zero))
2176 }
2177 vecs = append(vecs, v)
2178 }
2179 return vecs
2180}
2181
2182// offs2lohi splits offs into its low and high order bits.
2183func offs2lohi(offs int64) (lo, hi uintptr) {
2184 const longBits = SizeofLong * 8
2185 return uintptr(offs), uintptr(uint64(offs) >> (longBits - 1) >> 1) // two shifts to avoid false positive in vet
2186}
2187
2188func Readv(fd int, iovs [][]byte) (n int, err error) {
2189 iovecs := make([]Iovec, 0, minIovec)
2190 iovecs = appendBytes(iovecs, iovs)
2191 n, err = readv(fd, iovecs)
2192 readvRacedetect(iovecs, n, err)
2193 return n, err
2194}
2195
2196func Preadv(fd int, iovs [][]byte, offset int64) (n int, err error) {
2197 iovecs := make([]Iovec, 0, minIovec)
2198 iovecs = appendBytes(iovecs, iovs)
2199 lo, hi := offs2lohi(offset)
2200 n, err = preadv(fd, iovecs, lo, hi)
2201 readvRacedetect(iovecs, n, err)
2202 return n, err
2203}
2204
2205func Preadv2(fd int, iovs [][]byte, offset int64, flags int) (n int, err error) {
2206 iovecs := make([]Iovec, 0, minIovec)
2207 iovecs = appendBytes(iovecs, iovs)
2208 lo, hi := offs2lohi(offset)
2209 n, err = preadv2(fd, iovecs, lo, hi, flags)
2210 readvRacedetect(iovecs, n, err)
2211 return n, err
2212}
2213
2214func readvRacedetect(iovecs []Iovec, n int, err error) {
2215 if !raceenabled {
2216 return
2217 }
2218 for i := 0; n > 0 && i < len(iovecs); i++ {
2219 m := int(iovecs[i].Len)
2220 if m > n {
2221 m = n
2222 }
2223 n -= m
2224 if m > 0 {
2225 raceWriteRange(unsafe.Pointer(iovecs[i].Base), m)
2226 }
2227 }
2228 if err == nil {
2229 raceAcquire(unsafe.Pointer(&ioSync))
2230 }
2231}
2232
2233func Writev(fd int, iovs [][]byte) (n int, err error) {
2234 iovecs := make([]Iovec, 0, minIovec)
2235 iovecs = appendBytes(iovecs, iovs)
2236 if raceenabled {
2237 raceReleaseMerge(unsafe.Pointer(&ioSync))
2238 }
2239 n, err = writev(fd, iovecs)
2240 writevRacedetect(iovecs, n)
2241 return n, err
2242}
2243
2244func Pwritev(fd int, iovs [][]byte, offset int64) (n int, err error) {
2245 iovecs := make([]Iovec, 0, minIovec)
2246 iovecs = appendBytes(iovecs, iovs)
2247 if raceenabled {
2248 raceReleaseMerge(unsafe.Pointer(&ioSync))
2249 }
2250 lo, hi := offs2lohi(offset)
2251 n, err = pwritev(fd, iovecs, lo, hi)
2252 writevRacedetect(iovecs, n)
2253 return n, err
2254}
2255
2256func Pwritev2(fd int, iovs [][]byte, offset int64, flags int) (n int, err error) {
2257 iovecs := make([]Iovec, 0, minIovec)
2258 iovecs = appendBytes(iovecs, iovs)
2259 if raceenabled {
2260 raceReleaseMerge(unsafe.Pointer(&ioSync))
2261 }
2262 lo, hi := offs2lohi(offset)
2263 n, err = pwritev2(fd, iovecs, lo, hi, flags)
2264 writevRacedetect(iovecs, n)
2265 return n, err
2266}
2267
2268func writevRacedetect(iovecs []Iovec, n int) {
2269 if !raceenabled {
2270 return
2271 }
2272 for i := 0; n > 0 && i < len(iovecs); i++ {
2273 m := int(iovecs[i].Len)
2274 if m > n {
2275 m = n
2276 }
2277 n -= m
2278 if m > 0 {
2279 raceReadRange(unsafe.Pointer(iovecs[i].Base), m)
2280 }
2281 }
2282}
2283
2284// mmap varies by architecture; see syscall_linux_*.go.
2285//sys munmap(addr uintptr, length uintptr) (err error)
2286//sys mremap(oldaddr uintptr, oldlength uintptr, newlength uintptr, flags int, newaddr uintptr) (xaddr uintptr, err error)
2287//sys Madvise(b []byte, advice int) (err error)
2288//sys Mprotect(b []byte, prot int) (err error)
2289//sys Mlock(b []byte) (err error)
2290//sys Mlockall(flags int) (err error)
2291//sys Msync(b []byte, flags int) (err error)
2292//sys Munlock(b []byte) (err error)
2293//sys Munlockall() (err error)
2294
2295const (
2296 mremapFixed = MREMAP_FIXED
2297 mremapDontunmap = MREMAP_DONTUNMAP
2298 mremapMaymove = MREMAP_MAYMOVE
2299)
2300
2301// Vmsplice splices user pages from a slice of Iovecs into a pipe specified by fd,
2302// using the specified flags.
2303func Vmsplice(fd int, iovs []Iovec, flags int) (int, error) {
2304 var p unsafe.Pointer
2305 if len(iovs) > 0 {
2306 p = unsafe.Pointer(&iovs[0])
2307 }
2308
2309 n, _, errno := Syscall6(SYS_VMSPLICE, uintptr(fd), uintptr(p), uintptr(len(iovs)), uintptr(flags), 0, 0)
2310 if errno != 0 {
2311 return 0, syscall.Errno(errno)
2312 }
2313
2314 return int(n), nil
2315}
2316
2317func isGroupMember(gid int) bool {
2318 groups, err := Getgroups()
2319 if err != nil {
2320 return false
2321 }
2322
2323 for _, g := range groups {
2324 if g == gid {
2325 return true
2326 }
2327 }
2328 return false
2329}
2330
2331func isCapDacOverrideSet() bool {
2332 hdr := CapUserHeader{Version: LINUX_CAPABILITY_VERSION_3}
2333 data := [2]CapUserData{}
2334 err := Capget(&hdr, &data[0])
2335
2336 return err == nil && data[0].Effective&(1<<CAP_DAC_OVERRIDE) != 0
2337}
2338
2339//sys faccessat(dirfd int, path string, mode uint32) (err error)
2340//sys Faccessat2(dirfd int, path string, mode uint32, flags int) (err error)
2341
2342func Faccessat(dirfd int, path string, mode uint32, flags int) (err error) {
2343 if flags == 0 {
2344 return faccessat(dirfd, path, mode)
2345 }
2346
2347 if err := Faccessat2(dirfd, path, mode, flags); err != ENOSYS && err != EPERM {
2348 return err
2349 }
2350
2351 // The Linux kernel faccessat system call does not take any flags.
2352 // The glibc faccessat implements the flags itself; see
2353 // https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/faccessat.c;hb=HEAD
2354 // Because people naturally expect syscall.Faccessat to act
2355 // like C faccessat, we do the same.
2356
2357 if flags & ^(AT_SYMLINK_NOFOLLOW|AT_EACCESS) != 0 {
2358 return EINVAL
2359 }
2360
2361 var st Stat_t
2362 if err := Fstatat(dirfd, path, &st, flags&AT_SYMLINK_NOFOLLOW); err != nil {
2363 return err
2364 }
2365
2366 mode &= 7
2367 if mode == 0 {
2368 return nil
2369 }
2370
2371 var uid int
2372 if flags&AT_EACCESS != 0 {
2373 uid = Geteuid()
2374 if uid != 0 && isCapDacOverrideSet() {
2375 // If CAP_DAC_OVERRIDE is set, file access check is
2376 // done by the kernel in the same way as for root
2377 // (see generic_permission() in the Linux sources).
2378 uid = 0
2379 }
2380 } else {
2381 uid = Getuid()
2382 }
2383
2384 if uid == 0 {
2385 if mode&1 == 0 {
2386 // Root can read and write any file.
2387 return nil
2388 }
2389 if st.Mode&0111 != 0 {
2390 // Root can execute any file that anybody can execute.
2391 return nil
2392 }
2393 return EACCES
2394 }
2395
2396 var fmode uint32
2397 if uint32(uid) == st.Uid {
2398 fmode = (st.Mode >> 6) & 7
2399 } else {
2400 var gid int
2401 if flags&AT_EACCESS != 0 {
2402 gid = Getegid()
2403 } else {
2404 gid = Getgid()
2405 }
2406
2407 if uint32(gid) == st.Gid || isGroupMember(int(st.Gid)) {
2408 fmode = (st.Mode >> 3) & 7
2409 } else {
2410 fmode = st.Mode & 7
2411 }
2412 }
2413
2414 if fmode&mode == mode {
2415 return nil
2416 }
2417
2418 return EACCES
2419}
2420
2421//sys nameToHandleAt(dirFD int, pathname string, fh *fileHandle, mountID *_C_int, flags int) (err error) = SYS_NAME_TO_HANDLE_AT
2422//sys openByHandleAt(mountFD int, fh *fileHandle, flags int) (fd int, err error) = SYS_OPEN_BY_HANDLE_AT
2423
2424// fileHandle is the argument to nameToHandleAt and openByHandleAt. We
2425// originally tried to generate it via unix/linux/types.go with "type
2426// fileHandle C.struct_file_handle" but that generated empty structs
2427// for mips64 and mips64le. Instead, hard code it for now (it's the
2428// same everywhere else) until the mips64 generator issue is fixed.
2429type fileHandle struct {
2430 Bytes uint32
2431 Type int32
2432}
2433
2434// FileHandle represents the C struct file_handle used by
2435// name_to_handle_at (see NameToHandleAt) and open_by_handle_at (see
2436// OpenByHandleAt).
2437type FileHandle struct {
2438 *fileHandle
2439}
2440
2441// NewFileHandle constructs a FileHandle.
2442func NewFileHandle(handleType int32, handle []byte) FileHandle {
2443 const hdrSize = unsafe.Sizeof(fileHandle{})
2444 buf := make([]byte, hdrSize+uintptr(len(handle)))
2445 copy(buf[hdrSize:], handle)
2446 fh := (*fileHandle)(unsafe.Pointer(&buf[0]))
2447 fh.Type = handleType
2448 fh.Bytes = uint32(len(handle))
2449 return FileHandle{fh}
2450}
2451
2452func (fh *FileHandle) Size() int { return int(fh.fileHandle.Bytes) }
2453func (fh *FileHandle) Type() int32 { return fh.fileHandle.Type }
2454func (fh *FileHandle) Bytes() []byte {
2455 n := fh.Size()
2456 if n == 0 {
2457 return nil
2458 }
2459 return unsafe.Slice((*byte)(unsafe.Pointer(uintptr(unsafe.Pointer(&fh.fileHandle.Type))+4)), n)
2460}
2461
2462// NameToHandleAt wraps the name_to_handle_at system call; it obtains
2463// a handle for a path name.
2464func NameToHandleAt(dirfd int, path string, flags int) (handle FileHandle, mountID int, err error) {
2465 var mid _C_int
2466 // Try first with a small buffer, assuming the handle will
2467 // only be 32 bytes.
2468 size := uint32(32 + unsafe.Sizeof(fileHandle{}))
2469 didResize := false
2470 for {
2471 buf := make([]byte, size)
2472 fh := (*fileHandle)(unsafe.Pointer(&buf[0]))
2473 fh.Bytes = size - uint32(unsafe.Sizeof(fileHandle{}))
2474 err = nameToHandleAt(dirfd, path, fh, &mid, flags)
2475 if err == EOVERFLOW {
2476 if didResize {
2477 // We shouldn't need to resize more than once
2478 return
2479 }
2480 didResize = true
2481 size = fh.Bytes + uint32(unsafe.Sizeof(fileHandle{}))
2482 continue
2483 }
2484 if err != nil {
2485 return
2486 }
2487 return FileHandle{fh}, int(mid), nil
2488 }
2489}
2490
2491// OpenByHandleAt wraps the open_by_handle_at system call; it opens a
2492// file via a handle as previously returned by NameToHandleAt.
2493func OpenByHandleAt(mountFD int, handle FileHandle, flags int) (fd int, err error) {
2494 return openByHandleAt(mountFD, handle.fileHandle, flags)
2495}
2496
2497// Klogset wraps the sys_syslog system call; it sets console_loglevel to
2498// the value specified by arg and passes a dummy pointer to bufp.
2499func Klogset(typ int, arg int) (err error) {
2500 var p unsafe.Pointer
2501 _, _, errno := Syscall(SYS_SYSLOG, uintptr(typ), uintptr(p), uintptr(arg))
2502 if errno != 0 {
2503 return errnoErr(errno)
2504 }
2505 return nil
2506}
2507
2508// RemoteIovec is Iovec with the pointer replaced with an integer.
2509// It is used for ProcessVMReadv and ProcessVMWritev, where the pointer
2510// refers to a location in a different process' address space, which
2511// would confuse the Go garbage collector.
2512type RemoteIovec struct {
2513 Base uintptr
2514 Len int
2515}
2516
2517//sys ProcessVMReadv(pid int, localIov []Iovec, remoteIov []RemoteIovec, flags uint) (n int, err error) = SYS_PROCESS_VM_READV
2518//sys ProcessVMWritev(pid int, localIov []Iovec, remoteIov []RemoteIovec, flags uint) (n int, err error) = SYS_PROCESS_VM_WRITEV
2519
2520//sys PidfdOpen(pid int, flags int) (fd int, err error) = SYS_PIDFD_OPEN
2521//sys PidfdGetfd(pidfd int, targetfd int, flags int) (fd int, err error) = SYS_PIDFD_GETFD
2522//sys PidfdSendSignal(pidfd int, sig Signal, info *Siginfo, flags int) (err error) = SYS_PIDFD_SEND_SIGNAL
2523
2524//sys shmat(id int, addr uintptr, flag int) (ret uintptr, err error)
2525//sys shmctl(id int, cmd int, buf *SysvShmDesc) (result int, err error)
2526//sys shmdt(addr uintptr) (err error)
2527//sys shmget(key int, size int, flag int) (id int, err error)
2528
2529//sys getitimer(which int, currValue *Itimerval) (err error)
2530//sys setitimer(which int, newValue *Itimerval, oldValue *Itimerval) (err error)
2531
2532// MakeItimerval creates an Itimerval from interval and value durations.
2533func MakeItimerval(interval, value time.Duration) Itimerval {
2534 return Itimerval{
2535 Interval: NsecToTimeval(interval.Nanoseconds()),
2536 Value: NsecToTimeval(value.Nanoseconds()),
2537 }
2538}
2539
2540// A value which may be passed to the which parameter for Getitimer and
2541// Setitimer.
2542type ItimerWhich int
2543
2544// Possible which values for Getitimer and Setitimer.
2545const (
2546 ItimerReal ItimerWhich = ITIMER_REAL
2547 ItimerVirtual ItimerWhich = ITIMER_VIRTUAL
2548 ItimerProf ItimerWhich = ITIMER_PROF
2549)
2550
2551// Getitimer wraps getitimer(2) to return the current value of the timer
2552// specified by which.
2553func Getitimer(which ItimerWhich) (Itimerval, error) {
2554 var it Itimerval
2555 if err := getitimer(int(which), &it); err != nil {
2556 return Itimerval{}, err
2557 }
2558
2559 return it, nil
2560}
2561
2562// Setitimer wraps setitimer(2) to arm or disarm the timer specified by which.
2563// It returns the previous value of the timer.
2564//
2565// If the Itimerval argument is the zero value, the timer will be disarmed.
2566func Setitimer(which ItimerWhich, it Itimerval) (Itimerval, error) {
2567 var prev Itimerval
2568 if err := setitimer(int(which), &it, &prev); err != nil {
2569 return Itimerval{}, err
2570 }
2571
2572 return prev, nil
2573}
2574
2575//sysnb rtSigprocmask(how int, set *Sigset_t, oldset *Sigset_t, sigsetsize uintptr) (err error) = SYS_RT_SIGPROCMASK
2576
2577func PthreadSigmask(how int, set, oldset *Sigset_t) error {
2578 if oldset != nil {
2579 // Explicitly clear in case Sigset_t is larger than _C__NSIG.
2580 *oldset = Sigset_t{}
2581 }
2582 return rtSigprocmask(how, set, oldset, _C__NSIG/8)
2583}
2584
2585//sysnb getresuid(ruid *_C_int, euid *_C_int, suid *_C_int)
2586//sysnb getresgid(rgid *_C_int, egid *_C_int, sgid *_C_int)
2587
2588func Getresuid() (ruid, euid, suid int) {
2589 var r, e, s _C_int
2590 getresuid(&r, &e, &s)
2591 return int(r), int(e), int(s)
2592}
2593
2594func Getresgid() (rgid, egid, sgid int) {
2595 var r, e, s _C_int
2596 getresgid(&r, &e, &s)
2597 return int(r), int(e), int(s)
2598}
2599
2600// Pselect is a wrapper around the Linux pselect6 system call.
2601// This version does not modify the timeout argument.
2602func Pselect(nfd int, r *FdSet, w *FdSet, e *FdSet, timeout *Timespec, sigmask *Sigset_t) (n int, err error) {
2603 // Per https://man7.org/linux/man-pages/man2/select.2.html#NOTES,
2604 // The Linux pselect6() system call modifies its timeout argument.
2605 // [Not modifying the argument] is the behavior required by POSIX.1-2001.
2606 var mutableTimeout *Timespec
2607 if timeout != nil {
2608 mutableTimeout = new(Timespec)
2609 *mutableTimeout = *timeout
2610 }
2611
2612 // The final argument of the pselect6() system call is not a
2613 // sigset_t * pointer, but is instead a structure
2614 var kernelMask *sigset_argpack
2615 if sigmask != nil {
2616 wordBits := 32 << (^uintptr(0) >> 63) // see math.intSize
2617
2618 // A sigset stores one bit per signal,
2619 // offset by 1 (because signal 0 does not exist).
2620 // So the number of words needed is ⌈__C_NSIG - 1 / wordBits⌉.
2621 sigsetWords := (_C__NSIG - 1 + wordBits - 1) / (wordBits)
2622
2623 sigsetBytes := uintptr(sigsetWords * (wordBits / 8))
2624 kernelMask = &sigset_argpack{
2625 ss: sigmask,
2626 ssLen: sigsetBytes,
2627 }
2628 }
2629
2630 return pselect6(nfd, r, w, e, mutableTimeout, kernelMask)
2631}
2632
2633//sys schedSetattr(pid int, attr *SchedAttr, flags uint) (err error)
2634//sys schedGetattr(pid int, attr *SchedAttr, size uint, flags uint) (err error)
2635
2636// SchedSetAttr is a wrapper for sched_setattr(2) syscall.
2637// https://man7.org/linux/man-pages/man2/sched_setattr.2.html
2638func SchedSetAttr(pid int, attr *SchedAttr, flags uint) error {
2639 if attr == nil {
2640 return EINVAL
2641 }
2642 attr.Size = SizeofSchedAttr
2643 return schedSetattr(pid, attr, flags)
2644}
2645
2646// SchedGetAttr is a wrapper for sched_getattr(2) syscall.
2647// https://man7.org/linux/man-pages/man2/sched_getattr.2.html
2648func SchedGetAttr(pid int, flags uint) (*SchedAttr, error) {
2649 attr := &SchedAttr{}
2650 if err := schedGetattr(pid, attr, SizeofSchedAttr, flags); err != nil {
2651 return nil, err
2652 }
2653 return attr, nil
2654}
2655
2656//sys Cachestat(fd uint, crange *CachestatRange, cstat *Cachestat_t, flags uint) (err error)
2657//sys Mseal(b []byte, flags uint) (err error)