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