1 /* $NetBSD: linux_misc.c,v 1.267 2024/10/01 16:41:29 riastradh Exp $ */
2
3 /*-
4 * Copyright (c) 1995, 1998, 1999, 2008 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Frank van der Linden and Eric Haszlakiewicz; by Jason R. Thorpe
9 * of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30 * POSSIBILITY OF SUCH DAMAGE.
31 */
32
33 /*
34 * Linux compatibility module. Try to deal with various Linux system calls.
35 */
36
37 /*
38 * These functions have been moved to multiarch to allow
39 * selection of which machines include them to be
40 * determined by the individual files.linux_<arch> files.
41 *
42 * Function in multiarch:
43 * linux_sys_break : linux_break.c
44 * linux_sys_alarm : linux_misc_notalpha.c
45 * linux_sys_getresgid : linux_misc_notalpha.c
46 * linux_sys_nice : linux_misc_notalpha.c
47 * linux_sys_readdir : linux_misc_notalpha.c
48 * linux_sys_setresgid : linux_misc_notalpha.c
49 * linux_sys_time : linux_misc_notalpha.c
50 * linux_sys_utime : linux_misc_notalpha.c
51 * linux_sys_waitpid : linux_misc_notalpha.c
52 * linux_sys_old_mmap : linux_oldmmap.c
53 * linux_sys_oldolduname : linux_oldolduname.c
54 * linux_sys_oldselect : linux_oldselect.c
55 * linux_sys_olduname : linux_olduname.c
56 * linux_sys_pipe : linux_pipe.c
57 */
58
59 #include <sys/cdefs.h>
60 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.267 2024/10/01 16:41:29 riastradh Exp $");
61
62 #include <sys/param.h>
63 #include <sys/systm.h>
64 #include <sys/namei.h>
65 #include <sys/proc.h>
66 #include <sys/dirent.h>
67 #include <sys/epoll.h>
68 #include <sys/eventfd.h>
69 #include <sys/file.h>
70 #include <sys/stat.h>
71 #include <sys/filedesc.h>
72 #include <sys/ioctl.h>
73 #include <sys/kernel.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/mman.h>
77 #include <sys/mount.h>
78 #include <sys/poll.h>
79 #include <sys/prot.h>
80 #include <sys/reboot.h>
81 #include <sys/resource.h>
82 #include <sys/resourcevar.h>
83 #include <sys/select.h>
84 #include <sys/signal.h>
85 #include <sys/signalvar.h>
86 #include <sys/socket.h>
87 #include <sys/time.h>
88 #include <sys/times.h>
89 #include <sys/vnode.h>
90 #include <sys/uio.h>
91 #include <sys/wait.h>
92 #include <sys/utsname.h>
93 #include <sys/unistd.h>
94 #include <sys/vfs_syscalls.h>
95 #include <sys/swap.h> /* for SWAP_ON */
96 #include <sys/sysctl.h> /* for KERN_DOMAINNAME */
97 #include <sys/kauth.h>
98 #include <sys/futex.h>
99
100 #include <sys/ptrace.h>
101 #include <machine/ptrace.h>
102
103 #include <sys/syscall.h>
104 #include <sys/syscallargs.h>
105
106 #include <compat/sys/resource.h>
107
108 #include <compat/linux/common/linux_machdep.h>
109 #include <compat/linux/common/linux_types.h>
110 #include <compat/linux/common/linux_signal.h>
111 #include <compat/linux/common/linux_ipc.h>
112 #include <compat/linux/common/linux_sem.h>
113
114 #include <compat/linux/common/linux_fcntl.h>
115 #include <compat/linux/common/linux_mmap.h>
116 #include <compat/linux/common/linux_dirent.h>
117 #include <compat/linux/common/linux_util.h>
118 #include <compat/linux/common/linux_misc.h>
119 #include <compat/linux/common/linux_statfs.h>
120 #include <compat/linux/common/linux_limit.h>
121 #include <compat/linux/common/linux_ptrace.h>
122 #include <compat/linux/common/linux_reboot.h>
123 #include <compat/linux/common/linux_emuldata.h>
124 #include <compat/linux/common/linux_sched.h>
125
126 #include <compat/linux/linux_syscallargs.h>
127
128 const int linux_ptrace_request_map[] = {
129 LINUX_PTRACE_TRACEME, PT_TRACE_ME,
130 LINUX_PTRACE_PEEKTEXT, PT_READ_I,
131 LINUX_PTRACE_PEEKDATA, PT_READ_D,
132 LINUX_PTRACE_POKETEXT, PT_WRITE_I,
133 LINUX_PTRACE_POKEDATA, PT_WRITE_D,
134 LINUX_PTRACE_CONT, PT_CONTINUE,
135 LINUX_PTRACE_KILL, PT_KILL,
136 LINUX_PTRACE_ATTACH, PT_ATTACH,
137 LINUX_PTRACE_DETACH, PT_DETACH,
138 # ifdef PT_STEP
139 LINUX_PTRACE_SINGLESTEP, PT_STEP,
140 # endif
141 LINUX_PTRACE_SYSCALL, PT_SYSCALL,
142 -1
143 };
144
145 const struct linux_mnttypes linux_fstypes[] = {
146 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC },
147 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC },
148 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC },
149 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC },
150 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC },
151 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC },
152 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC },
153 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC },
154 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC },
155 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC },
156 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC },
157 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC },
158 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC },
159 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC },
160 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC },
161 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC },
162 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC },
163 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC },
164 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC },
165 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC },
166 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC },
167 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC },
168 { MOUNT_TMPFS, LINUX_TMPFS_SUPER_MAGIC }
169 };
170 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
171
172 #ifdef DEBUG_LINUX
173 #define DPRINTF(a, ...) uprintf(a, __VA_ARGS__)
174 #else
175 #define DPRINTF(a, ...)
176 #endif
177
178 /* Local linux_misc.c functions: */
179 static void linux_to_bsd_mmap_args(struct sys_mmap_args *,
180 const struct linux_sys_mmap_args *);
181 static int linux_mmap(struct lwp *, const struct linux_sys_mmap_args *,
182 register_t *, off_t);
183 static int linux_to_native_wait_options(int);
184
185 /*
186 * The information on a terminated (or stopped) process needs
187 * to be converted in order for Linux binaries to get a valid signal
188 * number out of it.
189 */
190 int
bsd_to_linux_wstat(int st)191 bsd_to_linux_wstat(int st)
192 {
193
194 int sig;
195
196 if (WIFSIGNALED(st)) {
197 sig = WTERMSIG(st);
198 if (sig >= 0 && sig < NSIG)
199 st= (st & ~0177) | native_to_linux_signo[sig];
200 } else if (WIFSTOPPED(st)) {
201 sig = WSTOPSIG(st);
202 if (sig >= 0 && sig < NSIG)
203 st = (st & ~0xff00) |
204 (native_to_linux_signo[sig] << 8);
205 }
206 return st;
207 }
208
209 /*
210 * wait4(2). Passed on to the NetBSD call, surrounded by code to
211 * reserve some space for a NetBSD-style wait status, and converting
212 * it to what Linux wants.
213 */
214 int
linux_sys_wait4(struct lwp * l,const struct linux_sys_wait4_args * uap,register_t * retval)215 linux_sys_wait4(struct lwp *l, const struct linux_sys_wait4_args *uap, register_t *retval)
216 {
217 /* {
218 syscallarg(int) pid;
219 syscallarg(int *) status;
220 syscallarg(int) options;
221 syscallarg(struct rusage50 *) rusage;
222 } */
223 int error, status, options, linux_options, pid = SCARG(uap, pid);
224 struct rusage50 ru50;
225 struct rusage ru;
226 proc_t *p;
227
228 linux_options = SCARG(uap, options);
229 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
230 return (EINVAL);
231
232 options = linux_to_native_wait_options(linux_options);
233 # ifdef DIAGNOSTIC
234 if (linux_options & LINUX_WNOTHREAD)
235 printf("WARNING: %s: linux process %d.%d called "
236 "waitpid with __WNOTHREAD set!\n",
237 __FILE__, l->l_proc->p_pid, l->l_lid);
238
239 # endif
240
241 error = do_sys_wait(&pid, &status, options,
242 SCARG(uap, rusage) != NULL ? &ru : NULL);
243
244 retval[0] = pid;
245 if (pid == 0)
246 return error;
247
248 p = curproc;
249 mutex_enter(p->p_lock);
250 sigdelset(&p->p_sigpend.sp_set, SIGCHLD); /* XXXAD ksiginfo leak */
251 mutex_exit(p->p_lock);
252
253 if (SCARG(uap, rusage) != NULL) {
254 rusage_to_rusage50(&ru, &ru50);
255 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
256 }
257
258 if (error == 0 && SCARG(uap, status) != NULL) {
259 status = bsd_to_linux_wstat(status);
260 error = copyout(&status, SCARG(uap, status), sizeof status);
261 }
262
263 return error;
264 }
265
266 /*
267 * waitid(2). Converting arguments to the NetBSD equivalent and
268 * calling it.
269 */
270 int
linux_sys_waitid(struct lwp * l,const struct linux_sys_waitid_args * uap,register_t * retval)271 linux_sys_waitid(struct lwp *l, const struct linux_sys_waitid_args *uap, register_t *retval)
272 {
273 /* {
274 syscallarg(int) idtype;
275 syscallarg(id_t) id;
276 syscallarg(linux_siginfo_t *) infop;
277 syscallarg(int) options;
278 syscallarg(struct rusage50 *) rusage;
279 } */
280 int error, linux_options, options, linux_idtype, status;
281 pid_t pid;
282 idtype_t idtype;
283 id_t id;
284 siginfo_t info;
285 linux_siginfo_t linux_info;
286 struct wrusage wru;
287 struct rusage50 ru50;
288
289 linux_idtype = SCARG(uap, idtype);
290 switch (linux_idtype) {
291 case LINUX_P_ALL:
292 idtype = P_ALL;
293 break;
294 case LINUX_P_PID:
295 idtype = P_PID;
296 break;
297 case LINUX_P_PGID:
298 idtype = P_PGID;
299 break;
300 case LINUX_P_PIDFD:
301 return EOPNOTSUPP;
302 default:
303 return EINVAL;
304 }
305
306 linux_options = SCARG(uap, options);
307 if (linux_options & ~(LINUX_WAITID_KNOWNFLAGS))
308 return EINVAL;
309
310 options = linux_to_native_wait_options(linux_options);
311 id = SCARG(uap, id);
312
313 error = do_sys_waitid(idtype, id, &pid, &status, options, &wru, &info);
314 if (pid == 0 && options & WNOHANG) {
315 info.si_signo = 0;
316 info.si_pid = 0;
317 }
318
319 if (error == 0 && SCARG(uap, infop) != NULL) {
320 /* POSIX says that this NULL check is a bug, but Linux does this. */
321 native_to_linux_siginfo(&linux_info, &info._info);
322 error = copyout(&linux_info, SCARG(uap, infop), sizeof(linux_info));
323 }
324
325 if (error == 0 && SCARG(uap, rusage) != NULL) {
326 rusage_to_rusage50(&wru.wru_children, &ru50);
327 error = copyout(&ru50, SCARG(uap, rusage), sizeof(ru50));
328 }
329
330 return error;
331 }
332
333 /*
334 * Convert the options argument for wait4(2) and waitid(2) from what
335 * Linux wants to what NetBSD wants.
336 */
337 static int
linux_to_native_wait_options(int linux_options)338 linux_to_native_wait_options(int linux_options)
339 {
340 int options = 0;
341
342 if (linux_options & LINUX_WNOHANG)
343 options |= WNOHANG;
344 if (linux_options & LINUX_WUNTRACED)
345 options |= WUNTRACED;
346 if (linux_options & LINUX_WEXITED)
347 options |= WEXITED;
348 if (linux_options & LINUX_WCONTINUED)
349 options |= WCONTINUED;
350 if (linux_options & LINUX_WNOWAIT)
351 options |= WNOWAIT;
352 if (linux_options & LINUX_WALL)
353 options |= WALLSIG;
354 if (linux_options & LINUX_WCLONE)
355 options |= WALTSIG;
356
357 return options;
358 }
359
360 /*
361 * Linux brk(2). Like native, but always return the new break value.
362 */
363 int
linux_sys_brk(struct lwp * l,const struct linux_sys_brk_args * uap,register_t * retval)364 linux_sys_brk(struct lwp *l, const struct linux_sys_brk_args *uap, register_t *retval)
365 {
366 /* {
367 syscallarg(char *) nsize;
368 } */
369 struct proc *p = l->l_proc;
370 struct vmspace *vm = p->p_vmspace;
371 struct sys_obreak_args oba;
372
373 SCARG(&oba, nsize) = SCARG(uap, nsize);
374
375 (void) sys_obreak(l, &oba, retval);
376 retval[0] = (register_t)((char *)vm->vm_daddr + ptoa(vm->vm_dsize));
377 return 0;
378 }
379
380 /*
381 * Implement the fs stat functions. Straightforward.
382 */
383 int
linux_sys_statfs(struct lwp * l,const struct linux_sys_statfs_args * uap,register_t * retval)384 linux_sys_statfs(struct lwp *l, const struct linux_sys_statfs_args *uap, register_t *retval)
385 {
386 /* {
387 syscallarg(const char *) path;
388 syscallarg(struct linux_statfs *) sp;
389 } */
390 struct statvfs *sb;
391 struct linux_statfs ltmp;
392 int error;
393
394 sb = STATVFSBUF_GET();
395 error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb);
396 if (error == 0) {
397 bsd_to_linux_statfs(sb, <mp);
398 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp);
399 }
400 STATVFSBUF_PUT(sb);
401
402 return error;
403 }
404
405 int
linux_sys_fstatfs(struct lwp * l,const struct linux_sys_fstatfs_args * uap,register_t * retval)406 linux_sys_fstatfs(struct lwp *l, const struct linux_sys_fstatfs_args *uap, register_t *retval)
407 {
408 /* {
409 syscallarg(int) fd;
410 syscallarg(struct linux_statfs *) sp;
411 } */
412 struct statvfs *sb;
413 struct linux_statfs ltmp;
414 int error;
415
416 sb = STATVFSBUF_GET();
417 error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb);
418 if (error == 0) {
419 bsd_to_linux_statfs(sb, <mp);
420 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp);
421 }
422 STATVFSBUF_PUT(sb);
423
424 return error;
425 }
426
427 /*
428 * uname(). Just copy the info from the various strings stored in the
429 * kernel, and put it in the Linux utsname structure. That structure
430 * is almost the same as the NetBSD one, only it has fields 65 characters
431 * long, and an extra domainname field.
432 */
433 int
linux_sys_uname(struct lwp * l,const struct linux_sys_uname_args * uap,register_t * retval)434 linux_sys_uname(struct lwp *l, const struct linux_sys_uname_args *uap, register_t *retval)
435 {
436 /* {
437 syscallarg(struct linux_utsname *) up;
438 } */
439 struct linux_utsname luts;
440
441 memset(&luts, 0, sizeof(luts));
442 strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
443 strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
444 strlcpy(luts.l_release, linux_release, sizeof(luts.l_release));
445 strlcpy(luts.l_version, linux_version, sizeof(luts.l_version));
446 strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
447 strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
448
449 return copyout(&luts, SCARG(uap, up), sizeof(luts));
450 }
451
452 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
453 /* Used indirectly on: arm, i386, m68k */
454
455 /*
456 * New type Linux mmap call.
457 * Only called directly on machines with >= 6 free regs.
458 */
459 int
linux_sys_mmap(struct lwp * l,const struct linux_sys_mmap_args * uap,register_t * retval)460 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval)
461 {
462 /* {
463 syscallarg(unsigned long) addr;
464 syscallarg(size_t) len;
465 syscallarg(int) prot;
466 syscallarg(int) flags;
467 syscallarg(int) fd;
468 syscallarg(linux_off_t) offset;
469 } */
470
471 if (SCARG(uap, offset) & PAGE_MASK)
472 return EINVAL;
473
474 return linux_mmap(l, uap, retval, SCARG(uap, offset));
475 }
476
477 /*
478 * Guts of most architectures' mmap64() implementations. This shares
479 * its list of arguments with linux_sys_mmap().
480 *
481 * The difference in linux_sys_mmap2() is that "offset" is actually
482 * (offset / pagesize), not an absolute byte count. This translation
483 * to pagesize offsets is done inside glibc between the mmap64() call
484 * point, and the actual syscall.
485 */
486 int
linux_sys_mmap2(struct lwp * l,const struct linux_sys_mmap2_args * uap,register_t * retval)487 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval)
488 {
489 /* {
490 syscallarg(unsigned long) addr;
491 syscallarg(size_t) len;
492 syscallarg(int) prot;
493 syscallarg(int) flags;
494 syscallarg(int) fd;
495 syscallarg(linux_off_t) offset;
496 } */
497
498 return linux_mmap(l, uap, retval,
499 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
500 }
501
502 /*
503 * Massage arguments and call system mmap(2).
504 */
505 static int
linux_mmap(struct lwp * l,const struct linux_sys_mmap_args * uap,register_t * retval,off_t offset)506 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset)
507 {
508 struct sys_mmap_args cma;
509 int error;
510 size_t mmoff=0;
511
512 linux_to_bsd_mmap_args(&cma, uap);
513 SCARG(&cma, pos) = offset;
514
515 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
516 /*
517 * Request for stack-like memory segment. On linux, this
518 * works by mmap()ping (small) segment, which is automatically
519 * extended when page fault happens below the currently
520 * allocated area. We emulate this by allocating (typically
521 * bigger) segment sized at current stack size limit, and
522 * offsetting the requested and returned address accordingly.
523 * Since physical pages are only allocated on-demand, this
524 * is effectively identical.
525 */
526 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
527
528 if (SCARG(&cma, len) < ssl) {
529 /* Compute the address offset */
530 mmoff = round_page(ssl) - SCARG(uap, len);
531
532 if (SCARG(&cma, addr))
533 SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff;
534
535 SCARG(&cma, len) = (size_t) ssl;
536 }
537 }
538
539 error = sys_mmap(l, &cma, retval);
540 if (error)
541 return (error);
542
543 /* Shift the returned address for stack-like segment if necessary */
544 retval[0] += mmoff;
545
546 return (0);
547 }
548
549 static void
linux_to_bsd_mmap_args(struct sys_mmap_args * cma,const struct linux_sys_mmap_args * uap)550 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap)
551 {
552 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
553
554 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
555 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
556 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
557 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
558 flags |= cvtto_bsd_mask(fl, LINUX_MAP_LOCKED, MAP_WIRED);
559 /* XXX XAX ERH: Any other flags here? There are more defined... */
560
561 SCARG(cma, addr) = (void *)SCARG(uap, addr);
562 SCARG(cma, len) = SCARG(uap, len);
563 SCARG(cma, prot) = SCARG(uap, prot);
564 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
565 SCARG(cma, prot) |= VM_PROT_READ;
566 SCARG(cma, flags) = flags;
567 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
568 SCARG(cma, PAD) = 0;
569 }
570
571 #define LINUX_MREMAP_MAYMOVE 1
572 #define LINUX_MREMAP_FIXED 2
573
574 int
linux_sys_mremap(struct lwp * l,const struct linux_sys_mremap_args * uap,register_t * retval)575 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval)
576 {
577 /* {
578 syscallarg(void *) old_address;
579 syscallarg(size_t) old_size;
580 syscallarg(size_t) new_size;
581 syscallarg(u_long) flags;
582 } */
583
584 struct proc *p;
585 struct vm_map *map;
586 vaddr_t oldva;
587 vaddr_t newva;
588 size_t oldsize;
589 size_t newsize;
590 int flags;
591 int uvmflags;
592 int error;
593
594 flags = SCARG(uap, flags);
595 oldva = (vaddr_t)SCARG(uap, old_address);
596 oldsize = round_page(SCARG(uap, old_size));
597 newsize = round_page(SCARG(uap, new_size));
598 if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
599 error = EINVAL;
600 goto done;
601 }
602 if ((flags & LINUX_MREMAP_FIXED) != 0) {
603 if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
604 error = EINVAL;
605 goto done;
606 }
607 #if 0 /* notyet */
608 newva = SCARG(uap, new_address);
609 uvmflags = MAP_FIXED;
610 #else /* notyet */
611 error = EOPNOTSUPP;
612 goto done;
613 #endif /* notyet */
614 } else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
615 uvmflags = 0;
616 } else {
617 newva = oldva;
618 uvmflags = MAP_FIXED;
619 }
620 p = l->l_proc;
621 map = &p->p_vmspace->vm_map;
622 error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
623 uvmflags);
624
625 done:
626 *retval = (error != 0) ? 0 : (register_t)newva;
627 return error;
628 }
629
630 #ifdef USRSTACK
631 int
linux_sys_mprotect(struct lwp * l,const struct linux_sys_mprotect_args * uap,register_t * retval)632 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval)
633 {
634 /* {
635 syscallarg(const void *) start;
636 syscallarg(unsigned long) len;
637 syscallarg(int) prot;
638 } */
639 struct vm_map_entry *entry;
640 struct vm_map *map;
641 struct proc *p;
642 vaddr_t end, start, len, stacklim;
643 int prot, grows;
644
645 start = (vaddr_t)SCARG(uap, start);
646 len = round_page(SCARG(uap, len));
647 prot = SCARG(uap, prot);
648 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
649 prot &= ~grows;
650 end = start + len;
651
652 if (start & PAGE_MASK)
653 return EINVAL;
654 if (end < start)
655 return EINVAL;
656 if (end == start)
657 return 0;
658
659 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
660 return EINVAL;
661 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
662 return EINVAL;
663
664 p = l->l_proc;
665 map = &p->p_vmspace->vm_map;
666 vm_map_lock(map);
667 # ifdef notdef
668 VM_MAP_RANGE_CHECK(map, start, end);
669 # endif
670 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
671 vm_map_unlock(map);
672 return ENOMEM;
673 }
674
675 /*
676 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
677 */
678
679 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
680 if (grows & LINUX_PROT_GROWSDOWN) {
681 if (USRSTACK - stacklim <= start && start < USRSTACK) {
682 start = USRSTACK - stacklim;
683 } else {
684 start = entry->start;
685 }
686 } else if (grows & LINUX_PROT_GROWSUP) {
687 if (USRSTACK <= end && end < USRSTACK + stacklim) {
688 end = USRSTACK + stacklim;
689 } else {
690 end = entry->end;
691 }
692 }
693 vm_map_unlock(map);
694 return uvm_map_protect_user(l, start, end, prot);
695 }
696 #endif /* USRSTACK */
697
698 /*
699 * This code is partly stolen from src/lib/libc/compat-43/times.c
700 */
701
702 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz))
703
704 int
linux_sys_times(struct lwp * l,const struct linux_sys_times_args * uap,register_t * retval)705 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval)
706 {
707 /* {
708 syscallarg(struct times *) tms;
709 } */
710 struct proc *p = l->l_proc;
711 struct timeval t;
712 int error;
713
714 if (SCARG(uap, tms)) {
715 struct linux_tms ltms;
716 struct rusage ru;
717
718 memset(<ms, 0, sizeof(ltms));
719
720 mutex_enter(p->p_lock);
721 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
722 ltms.ltms_utime = CONVTCK(ru.ru_utime);
723 ltms.ltms_stime = CONVTCK(ru.ru_stime);
724 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
725 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
726 mutex_exit(p->p_lock);
727
728 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms)))
729 return error;
730 }
731
732 getmicrouptime(&t);
733
734 retval[0] = ((linux_clock_t)(CONVTCK(t)));
735 return 0;
736 }
737
738 #undef CONVTCK
739
740 #if !defined(__aarch64__)
741 /*
742 * Linux 'readdir' call. This code is mostly taken from the
743 * SunOS getdents call (see compat/sunos/sunos_misc.c), though
744 * an attempt has been made to keep it a little cleaner (failing
745 * miserably, because of the cruft needed if count 1 is passed).
746 *
747 * The d_off field should contain the offset of the next valid entry,
748 * but in Linux it has the offset of the entry itself. We emulate
749 * that bug here.
750 *
751 * Read in BSD-style entries, convert them, and copy them out.
752 *
753 * Note that this doesn't handle union-mounted filesystems.
754 */
755 int
linux_sys_getdents(struct lwp * l,const struct linux_sys_getdents_args * uap,register_t * retval)756 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval)
757 {
758 /* {
759 syscallarg(int) fd;
760 syscallarg(struct linux_dirent *) dent;
761 syscallarg(unsigned int) count;
762 } */
763 struct dirent *bdp;
764 struct vnode *vp;
765 char *inp, *tbuf; /* BSD-format */
766 int len, reclen; /* BSD-format */
767 char *outp; /* Linux-format */
768 int resid, linux_reclen = 0; /* Linux-format */
769 struct file *fp;
770 struct uio auio;
771 struct iovec aiov;
772 struct linux_dirent idb;
773 off_t off; /* true file offset */
774 int buflen, error, eofflag, nbytes, oldcall;
775 struct vattr va;
776 off_t *cookiebuf = NULL, *cookie;
777 int ncookies;
778
779 /* fd_getvnode() will use the descriptor for us */
780 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
781 return (error);
782
783 if ((fp->f_flag & FREAD) == 0) {
784 error = EBADF;
785 goto out1;
786 }
787
788 vp = (struct vnode *)fp->f_data;
789 if (vp->v_type != VDIR) {
790 error = ENOTDIR;
791 goto out1;
792 }
793
794 vn_lock(vp, LK_SHARED | LK_RETRY);
795 error = VOP_GETATTR(vp, &va, l->l_cred);
796 VOP_UNLOCK(vp);
797 if (error)
798 goto out1;
799
800 nbytes = SCARG(uap, count);
801 if (nbytes == 1) { /* emulating old, broken behaviour */
802 nbytes = sizeof (idb);
803 buflen = uimax(va.va_blocksize, nbytes);
804 oldcall = 1;
805 } else {
806 buflen = uimin(MAXBSIZE, nbytes);
807 if (buflen < va.va_blocksize)
808 buflen = va.va_blocksize;
809 oldcall = 0;
810 }
811 tbuf = malloc(buflen, M_TEMP, M_WAITOK);
812
813 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
814 off = fp->f_offset;
815 again:
816 aiov.iov_base = tbuf;
817 aiov.iov_len = buflen;
818 auio.uio_iov = &aiov;
819 auio.uio_iovcnt = 1;
820 auio.uio_rw = UIO_READ;
821 auio.uio_resid = buflen;
822 auio.uio_offset = off;
823 UIO_SETUP_SYSSPACE(&auio);
824 /*
825 * First we read into the malloc'ed buffer, then
826 * we massage it into user space, one record at a time.
827 */
828 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
829 &ncookies);
830 if (error)
831 goto out;
832
833 inp = tbuf;
834 outp = (void *)SCARG(uap, dent);
835 resid = nbytes;
836 if ((len = buflen - auio.uio_resid) == 0)
837 goto eof;
838
839 for (cookie = cookiebuf; len > 0; len -= reclen) {
840 bdp = (struct dirent *)inp;
841 reclen = bdp->d_reclen;
842 if (reclen & 3) {
843 error = EIO;
844 goto out;
845 }
846 if (bdp->d_fileno == 0) {
847 inp += reclen; /* it is a hole; squish it out */
848 if (cookie)
849 off = *cookie++;
850 else
851 off += reclen;
852 continue;
853 }
854 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
855 if (reclen > len || resid < linux_reclen) {
856 /* entry too big for buffer, so just stop */
857 outp++;
858 break;
859 }
860 /*
861 * Massage in place to make a Linux-shaped dirent (otherwise
862 * we have to worry about touching user memory outside of
863 * the copyout() call).
864 */
865 memset(&idb, 0, sizeof(idb));
866 idb.d_ino = bdp->d_fileno;
867 /*
868 * The old readdir() call misuses the offset and reclen fields.
869 */
870 if (oldcall) {
871 idb.d_off = (linux_off_t)linux_reclen;
872 idb.d_reclen = (u_short)bdp->d_namlen;
873 } else {
874 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
875 compat_offseterr(vp, "linux_getdents");
876 error = EINVAL;
877 goto out;
878 }
879 idb.d_off = (linux_off_t)off;
880 idb.d_reclen = (u_short)linux_reclen;
881 /* Linux puts d_type at the end of each record */
882 *((char *)&idb + idb.d_reclen - 1) = bdp->d_type;
883 }
884 memcpy(idb.d_name, bdp->d_name,
885 MIN(sizeof(idb.d_name), bdp->d_namlen + 1));
886 if ((error = copyout((void *)&idb, outp, linux_reclen)))
887 goto out;
888 /* advance past this real entry */
889 inp += reclen;
890 if (cookie)
891 off = *cookie++; /* each entry points to itself */
892 else
893 off += reclen;
894 /* advance output past Linux-shaped entry */
895 outp += linux_reclen;
896 resid -= linux_reclen;
897 if (oldcall)
898 break;
899 }
900
901 /* if we squished out the whole block, try again */
902 if (outp == (void *)SCARG(uap, dent)) {
903 if (cookiebuf)
904 free(cookiebuf, M_TEMP);
905 cookiebuf = NULL;
906 goto again;
907 }
908 fp->f_offset = off; /* update the vnode offset */
909
910 if (oldcall)
911 nbytes = resid + linux_reclen;
912
913 eof:
914 *retval = nbytes - resid;
915 out:
916 VOP_UNLOCK(vp);
917 if (cookiebuf)
918 free(cookiebuf, M_TEMP);
919 free(tbuf, M_TEMP);
920 out1:
921 fd_putfile(SCARG(uap, fd));
922 return error;
923 }
924 #endif
925
926 #if !defined(__aarch64__)
927 /*
928 * Even when just using registers to pass arguments to syscalls you can
929 * have 5 of them on the i386. So this newer version of select() does
930 * this.
931 */
932 int
linux_sys_select(struct lwp * l,const struct linux_sys_select_args * uap,register_t * retval)933 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval)
934 {
935 /* {
936 syscallarg(int) nfds;
937 syscallarg(fd_set *) readfds;
938 syscallarg(fd_set *) writefds;
939 syscallarg(fd_set *) exceptfds;
940 syscallarg(struct timeval50 *) timeout;
941 } */
942
943 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
944 SCARG(uap, writefds), SCARG(uap, exceptfds),
945 (struct linux_timeval *)SCARG(uap, timeout));
946 }
947
948 /*
949 * Common code for the old and new versions of select(). A couple of
950 * things are important:
951 * 1) return the amount of time left in the 'timeout' parameter
952 * 2) select never returns ERESTART on Linux, always return EINTR
953 */
954 int
linux_select1(struct lwp * l,register_t * retval,int nfds,fd_set * readfds,fd_set * writefds,fd_set * exceptfds,struct linux_timeval * timeout)955 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds,
956 fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout)
957 {
958 struct timespec ts0, ts1, uts, *ts = NULL;
959 struct linux_timeval ltv;
960 int error;
961
962 /*
963 * Store current time for computation of the amount of
964 * time left.
965 */
966 if (timeout) {
967 if ((error = copyin(timeout, <v, sizeof(ltv))))
968 return error;
969 uts.tv_sec = ltv.tv_sec;
970 uts.tv_nsec = (long)((unsigned long)ltv.tv_usec * 1000);
971 if (itimespecfix(&uts)) {
972 /*
973 * The timeval was invalid. Convert it to something
974 * valid that will act as it does under Linux.
975 */
976 uts.tv_sec += uts.tv_nsec / 1000000000;
977 uts.tv_nsec %= 1000000000;
978 if (uts.tv_nsec < 0) {
979 uts.tv_sec -= 1;
980 uts.tv_nsec += 1000000000;
981 }
982 if (uts.tv_sec < 0)
983 timespecclear(&uts);
984 }
985 ts = &uts;
986 nanotime(&ts0);
987 }
988
989 error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL);
990
991 if (error) {
992 /*
993 * See fs/select.c in the Linux kernel. Without this,
994 * Maelstrom doesn't work.
995 */
996 if (error == ERESTART)
997 error = EINTR;
998 return error;
999 }
1000
1001 if (timeout) {
1002 if (*retval) {
1003 /*
1004 * Compute how much time was left of the timeout,
1005 * by subtracting the current time and the time
1006 * before we started the call, and subtracting
1007 * that result from the user-supplied value.
1008 */
1009 nanotime(&ts1);
1010 timespecsub(&ts1, &ts0, &ts1);
1011 timespecsub(&uts, &ts1, &uts);
1012 if (uts.tv_sec < 0)
1013 timespecclear(&uts);
1014 } else
1015 timespecclear(&uts);
1016 ltv.tv_sec = uts.tv_sec;
1017 ltv.tv_usec = uts.tv_nsec / 1000;
1018 if ((error = copyout(<v, timeout, sizeof(ltv))))
1019 return error;
1020 }
1021
1022 return 0;
1023 }
1024 #endif
1025
1026 /*
1027 * Derived from FreeBSD's sys/compat/linux/linux_misc.c:linux_pselect6()
1028 * which was contributed by Dmitry Chagin
1029 * https://svnweb.freebsd.org/base?view=revision&revision=283403
1030 */
1031 int
linux_sys_pselect6(struct lwp * l,const struct linux_sys_pselect6_args * uap,register_t * retval)1032 linux_sys_pselect6(struct lwp *l,
1033 const struct linux_sys_pselect6_args *uap, register_t *retval)
1034 {
1035 /* {
1036 syscallarg(int) nfds;
1037 syscallarg(fd_set *) readfds;
1038 syscallarg(fd_set *) writefds;
1039 syscallarg(fd_set *) exceptfds;
1040 syscallarg(struct timespec *) timeout;
1041 syscallarg(linux_sized_sigset_t *) ss;
1042 } */
1043 struct timespec uts, ts0, ts1, *tsp;
1044 linux_sized_sigset_t lsss;
1045 struct linux_timespec lts;
1046 linux_sigset_t lss;
1047 sigset_t *ssp;
1048 sigset_t ss;
1049 int error;
1050
1051 ssp = NULL;
1052 if (SCARG(uap, ss) != NULL) {
1053 if ((error = copyin(SCARG(uap, ss), &lsss, sizeof(lsss))) != 0)
1054 return (error);
1055 if (lsss.ss_len != sizeof(lss))
1056 return (EINVAL);
1057 if (lsss.ss != NULL) {
1058 if ((error = copyin(lsss.ss, &lss, sizeof(lss))) != 0)
1059 return (error);
1060 linux_to_native_sigset(&ss, &lss);
1061 ssp = &ss;
1062 }
1063 }
1064
1065 if (SCARG(uap, timeout) != NULL) {
1066 error = copyin(SCARG(uap, timeout), <s, sizeof(lts));
1067 if (error != 0)
1068 return (error);
1069 linux_to_native_timespec(&uts, <s);
1070
1071 if (itimespecfix(&uts))
1072 return (EINVAL);
1073
1074 nanotime(&ts0);
1075 tsp = &uts;
1076 } else {
1077 tsp = NULL;
1078 }
1079
1080 error = selcommon(retval, SCARG(uap, nfds), SCARG(uap, readfds),
1081 SCARG(uap, writefds), SCARG(uap, exceptfds), tsp, ssp);
1082
1083 if (error == 0 && tsp != NULL) {
1084 if (retval != 0) {
1085 /*
1086 * Compute how much time was left of the timeout,
1087 * by subtracting the current time and the time
1088 * before we started the call, and subtracting
1089 * that result from the user-supplied value.
1090 */
1091 nanotime(&ts1);
1092 timespecsub(&ts1, &ts0, &ts1);
1093 timespecsub(&uts, &ts1, &uts);
1094 if (uts.tv_sec < 0)
1095 timespecclear(&uts);
1096 } else {
1097 timespecclear(&uts);
1098 }
1099
1100 native_to_linux_timespec(<s, &uts);
1101 error = copyout(<s, SCARG(uap, timeout), sizeof(lts));
1102 }
1103
1104 return (error);
1105 }
1106
1107 int
linux_sys_ppoll(struct lwp * l,const struct linux_sys_ppoll_args * uap,register_t * retval)1108 linux_sys_ppoll(struct lwp *l,
1109 const struct linux_sys_ppoll_args *uap, register_t *retval)
1110 {
1111 /* {
1112 syscallarg(struct pollfd *) fds;
1113 syscallarg(u_int) nfds;
1114 syscallarg(struct linux_timespec *) timeout;
1115 syscallarg(linux_sigset_t *) sigset;
1116 } */
1117 struct linux_timespec lts0, *lts;
1118 struct timespec ts0, *ts = NULL;
1119 linux_sigset_t lsigmask0, *lsigmask;
1120 sigset_t sigmask0, *sigmask = NULL;
1121 int error;
1122
1123 lts = SCARG(uap, timeout);
1124 if (lts) {
1125 if ((error = copyin(lts, <s0, sizeof(lts0))) != 0)
1126 return error;
1127 linux_to_native_timespec(&ts0, <s0);
1128 ts = &ts0;
1129 }
1130
1131 lsigmask = SCARG(uap, sigset);
1132 if (lsigmask) {
1133 if ((error = copyin(lsigmask, &lsigmask0, sizeof(lsigmask0))))
1134 return error;
1135 linux_to_native_sigset(&sigmask0, &lsigmask0);
1136 sigmask = &sigmask0;
1137 }
1138
1139 return pollcommon(retval, SCARG(uap, fds), SCARG(uap, nfds),
1140 ts, sigmask);
1141 }
1142
1143 /*
1144 * Set the 'personality' (emulation mode) for the current process. Only
1145 * accept the Linux personality here (0). This call is needed because
1146 * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1147 * ELF binaries run in Linux mode, not SVR4 mode.
1148 */
1149 int
linux_sys_personality(struct lwp * l,const struct linux_sys_personality_args * uap,register_t * retval)1150 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval)
1151 {
1152 /* {
1153 syscallarg(unsigned long) per;
1154 } */
1155 struct linux_emuldata *led;
1156 int per;
1157
1158 per = SCARG(uap, per);
1159 led = l->l_emuldata;
1160 if (per == LINUX_PER_QUERY) {
1161 retval[0] = led->led_personality;
1162 return 0;
1163 }
1164
1165 switch (per & LINUX_PER_MASK) {
1166 case LINUX_PER_LINUX:
1167 case LINUX_PER_LINUX32:
1168 led->led_personality = per;
1169 break;
1170
1171 default:
1172 return EINVAL;
1173 }
1174
1175 retval[0] = per;
1176 return 0;
1177 }
1178
1179 /*
1180 * We have nonexistent fsuid equal to uid.
1181 * If modification is requested, refuse.
1182 */
1183 int
linux_sys_setfsuid(struct lwp * l,const struct linux_sys_setfsuid_args * uap,register_t * retval)1184 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval)
1185 {
1186 /* {
1187 syscallarg(uid_t) uid;
1188 } */
1189 uid_t uid;
1190
1191 uid = SCARG(uap, uid);
1192 if (kauth_cred_getuid(l->l_cred) != uid)
1193 return sys_nosys(l, uap, retval);
1194
1195 *retval = uid;
1196 return 0;
1197 }
1198
1199 int
linux_sys_setfsgid(struct lwp * l,const struct linux_sys_setfsgid_args * uap,register_t * retval)1200 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval)
1201 {
1202 /* {
1203 syscallarg(gid_t) gid;
1204 } */
1205 gid_t gid;
1206
1207 gid = SCARG(uap, gid);
1208 if (kauth_cred_getgid(l->l_cred) != gid)
1209 return sys_nosys(l, uap, retval);
1210
1211 *retval = gid;
1212 return 0;
1213 }
1214
1215 int
linux_sys_setresuid(struct lwp * l,const struct linux_sys_setresuid_args * uap,register_t * retval)1216 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval)
1217 {
1218 /* {
1219 syscallarg(uid_t) ruid;
1220 syscallarg(uid_t) euid;
1221 syscallarg(uid_t) suid;
1222 } */
1223
1224 /*
1225 * Note: These checks are a little different than the NetBSD
1226 * setreuid(2) call performs. This precisely follows the
1227 * behavior of the Linux kernel.
1228 */
1229
1230 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1231 SCARG(uap, suid),
1232 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1233 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1234 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1235 }
1236
1237 int
linux_sys_getresuid(struct lwp * l,const struct linux_sys_getresuid_args * uap,register_t * retval)1238 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval)
1239 {
1240 /* {
1241 syscallarg(uid_t *) ruid;
1242 syscallarg(uid_t *) euid;
1243 syscallarg(uid_t *) suid;
1244 } */
1245 kauth_cred_t pc = l->l_cred;
1246 int error;
1247 uid_t uid;
1248
1249 /*
1250 * Linux copies these values out to userspace like so:
1251 *
1252 * 1. Copy out ruid.
1253 * 2. If that succeeds, copy out euid.
1254 * 3. If both of those succeed, copy out suid.
1255 */
1256 uid = kauth_cred_getuid(pc);
1257 if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1258 return (error);
1259
1260 uid = kauth_cred_geteuid(pc);
1261 if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1262 return (error);
1263
1264 uid = kauth_cred_getsvuid(pc);
1265
1266 return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1267 }
1268
1269 int
linux_sys_ptrace(struct lwp * l,const struct linux_sys_ptrace_args * uap,register_t * retval)1270 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval)
1271 {
1272 /* {
1273 i386, m68k, powerpc: T=int
1274 alpha, amd64: T=long
1275 syscallarg(T) request;
1276 syscallarg(T) pid;
1277 syscallarg(T) addr;
1278 syscallarg(T) data;
1279 } */
1280 const int *ptr;
1281 int request;
1282 int error;
1283
1284 ptr = linux_ptrace_request_map;
1285 request = SCARG(uap, request);
1286 while (*ptr != -1)
1287 if (*ptr++ == request) {
1288 struct sys_ptrace_args pta;
1289
1290 SCARG(&pta, req) = *ptr;
1291 SCARG(&pta, pid) = SCARG(uap, pid);
1292 SCARG(&pta, addr) = (void *)SCARG(uap, addr);
1293 SCARG(&pta, data) = SCARG(uap, data);
1294
1295 /*
1296 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1297 * to continue where the process left off previously.
1298 * The same thing is achieved by addr == (void *) 1
1299 * on NetBSD, so rewrite 'addr' appropriately.
1300 */
1301 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1302 SCARG(&pta, addr) = (void *) 1;
1303
1304 error = sysent[SYS_ptrace].sy_call(l, &pta, retval);
1305 if (error)
1306 return error;
1307 switch (request) {
1308 case LINUX_PTRACE_PEEKTEXT:
1309 case LINUX_PTRACE_PEEKDATA:
1310 error = copyout (retval,
1311 (void *)SCARG(uap, data),
1312 sizeof *retval);
1313 *retval = SCARG(uap, data);
1314 break;
1315 default:
1316 break;
1317 }
1318 return error;
1319 }
1320 else
1321 ptr++;
1322
1323 return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1324 }
1325
1326 int
linux_sys_reboot(struct lwp * l,const struct linux_sys_reboot_args * uap,register_t * retval)1327 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval)
1328 {
1329 /* {
1330 syscallarg(int) magic1;
1331 syscallarg(int) magic2;
1332 syscallarg(int) cmd;
1333 syscallarg(void *) arg;
1334 } */
1335 struct sys_reboot_args /* {
1336 syscallarg(int) opt;
1337 syscallarg(char *) bootstr;
1338 } */ sra;
1339 int error;
1340
1341 if ((error = kauth_authorize_system(l->l_cred,
1342 KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0)
1343 return(error);
1344
1345 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1346 return(EINVAL);
1347 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1348 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1349 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1350 return(EINVAL);
1351
1352 switch ((unsigned long)SCARG(uap, cmd)) {
1353 case LINUX_REBOOT_CMD_RESTART:
1354 SCARG(&sra, opt) = RB_AUTOBOOT;
1355 break;
1356 case LINUX_REBOOT_CMD_HALT:
1357 SCARG(&sra, opt) = RB_HALT;
1358 break;
1359 case LINUX_REBOOT_CMD_POWER_OFF:
1360 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1361 break;
1362 case LINUX_REBOOT_CMD_RESTART2:
1363 /* Reboot with an argument. */
1364 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1365 SCARG(&sra, bootstr) = SCARG(uap, arg);
1366 break;
1367 case LINUX_REBOOT_CMD_CAD_ON:
1368 return(EINVAL); /* We don't implement ctrl-alt-delete */
1369 case LINUX_REBOOT_CMD_CAD_OFF:
1370 return(0);
1371 default:
1372 return(EINVAL);
1373 }
1374
1375 return(sys_reboot(l, &sra, retval));
1376 }
1377
1378 /*
1379 * Copy of compat_12_sys_swapon().
1380 */
1381 int
linux_sys_swapon(struct lwp * l,const struct linux_sys_swapon_args * uap,register_t * retval)1382 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval)
1383 {
1384 /* {
1385 syscallarg(const char *) name;
1386 } */
1387 struct sys_swapctl_args ua;
1388
1389 SCARG(&ua, cmd) = SWAP_ON;
1390 SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1391 SCARG(&ua, misc) = 0; /* priority */
1392 return (sys_swapctl(l, &ua, retval));
1393 }
1394
1395 /*
1396 * Stop swapping to the file or block device specified by path.
1397 */
1398 int
linux_sys_swapoff(struct lwp * l,const struct linux_sys_swapoff_args * uap,register_t * retval)1399 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval)
1400 {
1401 /* {
1402 syscallarg(const char *) path;
1403 } */
1404 struct sys_swapctl_args ua;
1405
1406 SCARG(&ua, cmd) = SWAP_OFF;
1407 SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1408 return (sys_swapctl(l, &ua, retval));
1409 }
1410
1411 /*
1412 * Copy of compat_09_sys_setdomainname()
1413 */
1414 /* ARGSUSED */
1415 int
linux_sys_setdomainname(struct lwp * l,const struct linux_sys_setdomainname_args * uap,register_t * retval)1416 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval)
1417 {
1418 /* {
1419 syscallarg(char *) domainname;
1420 syscallarg(int) len;
1421 } */
1422 int name[2];
1423
1424 name[0] = CTL_KERN;
1425 name[1] = KERN_DOMAINNAME;
1426 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1427 SCARG(uap, len), l));
1428 }
1429
1430 /*
1431 * sysinfo()
1432 */
1433 /* ARGSUSED */
1434 int
linux_sys_sysinfo(struct lwp * l,const struct linux_sys_sysinfo_args * uap,register_t * retval)1435 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval)
1436 {
1437 /* {
1438 syscallarg(struct linux_sysinfo *) arg;
1439 } */
1440 struct linux_sysinfo si;
1441 struct loadavg *la;
1442 int64_t filepg;
1443
1444 memset(&si, 0, sizeof(si));
1445 si.uptime = time_uptime;
1446 la = &averunnable;
1447 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1448 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1449 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1450 si.totalram = ctob((u_long)physmem);
1451 /* uvm_availmem() may sync the counters. */
1452 si.freeram = (u_long)uvm_availmem(true) * uvmexp.pagesize;
1453 filepg = cpu_count_get(CPU_COUNT_FILECLEAN) +
1454 cpu_count_get(CPU_COUNT_FILEDIRTY) +
1455 cpu_count_get(CPU_COUNT_FILEUNKNOWN) -
1456 cpu_count_get(CPU_COUNT_EXECPAGES);
1457 si.sharedram = 0; /* XXX */
1458 si.bufferram = (u_long)(filepg * uvmexp.pagesize);
1459 si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1460 si.freeswap =
1461 (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1462 si.procs = atomic_load_relaxed(&nprocs);
1463
1464 /* The following are only present in newer Linux kernels. */
1465 si.totalbig = 0;
1466 si.freebig = 0;
1467 si.mem_unit = 1;
1468
1469 return (copyout(&si, SCARG(uap, arg), sizeof si));
1470 }
1471
1472 int
linux_sys_getrlimit(struct lwp * l,const struct linux_sys_getrlimit_args * uap,register_t * retval)1473 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval)
1474 {
1475 /* {
1476 syscallarg(int) which;
1477 # ifdef LINUX_LARGEFILE64
1478 syscallarg(struct rlimit *) rlp;
1479 # else
1480 syscallarg(struct orlimit *) rlp;
1481 # endif
1482 } */
1483 # ifdef LINUX_LARGEFILE64
1484 struct rlimit orl;
1485 # else
1486 struct orlimit orl;
1487 # endif
1488 int which;
1489
1490 which = linux_to_bsd_limit(SCARG(uap, which));
1491 if (which < 0)
1492 return -which;
1493
1494 memset(&orl, 0, sizeof(orl));
1495 bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]);
1496
1497 return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1498 }
1499
1500 int
linux_sys_setrlimit(struct lwp * l,const struct linux_sys_setrlimit_args * uap,register_t * retval)1501 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval)
1502 {
1503 /* {
1504 syscallarg(int) which;
1505 # ifdef LINUX_LARGEFILE64
1506 syscallarg(struct rlimit *) rlp;
1507 # else
1508 syscallarg(struct orlimit *) rlp;
1509 # endif
1510 } */
1511 struct rlimit rl;
1512 # ifdef LINUX_LARGEFILE64
1513 struct rlimit orl;
1514 # else
1515 struct orlimit orl;
1516 # endif
1517 int error;
1518 int which;
1519
1520 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1521 return error;
1522
1523 which = linux_to_bsd_limit(SCARG(uap, which));
1524 if (which < 0)
1525 return -which;
1526
1527 linux_to_bsd_rlimit(&rl, &orl);
1528 return dosetrlimit(l, l->l_proc, which, &rl);
1529 }
1530
1531 # if !defined(__aarch64__) && !defined(__mips__) && !defined(__amd64__)
1532 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1533 int
linux_sys_ugetrlimit(struct lwp * l,const struct linux_sys_ugetrlimit_args * uap,register_t * retval)1534 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval)
1535 {
1536 return linux_sys_getrlimit(l, (const void *)uap, retval);
1537 }
1538 # endif
1539
1540 int
linux_sys_prlimit64(struct lwp * l,const struct linux_sys_prlimit64_args * uap,register_t * retval)1541 linux_sys_prlimit64(struct lwp *l, const struct linux_sys_prlimit64_args *uap, register_t *retval)
1542 {
1543 /* {
1544 syscallarg(pid_t) pid;
1545 syscallarg(int) witch;
1546 syscallarg(struct rlimit *) new_rlp;
1547 syscallarg(struct rlimit *) old_rlp;
1548 }; */
1549 struct rlimit rl, nrl, orl;
1550 struct rlimit *p;
1551 int which;
1552 int error;
1553
1554 /* XXX: Cannot operate any process other than its own */
1555 if (SCARG(uap, pid) != 0)
1556 return EPERM;
1557
1558 which = linux_to_bsd_limit(SCARG(uap, which));
1559 if (which < 0)
1560 return -which;
1561
1562 p = SCARG(uap, old_rlp);
1563 if (p != NULL) {
1564 memset(&orl, 0, sizeof(orl));
1565 bsd_to_linux_rlimit64(&orl, &l->l_proc->p_rlimit[which]);
1566 if ((error = copyout(&orl, p, sizeof(orl))) != 0)
1567 return error;
1568 }
1569
1570 p = SCARG(uap, new_rlp);
1571 if (p != NULL) {
1572 if ((error = copyin(p, &nrl, sizeof(nrl))) != 0)
1573 return error;
1574
1575 linux_to_bsd_rlimit(&rl, &nrl);
1576 return dosetrlimit(l, l->l_proc, which, &rl);
1577 }
1578
1579 return 0;
1580 }
1581
1582 /*
1583 * This gets called for unsupported syscalls. The difference to sys_nosys()
1584 * is that process does not get SIGSYS, the call just returns with ENOSYS.
1585 * This is the way Linux does it and glibc depends on this behaviour.
1586 */
1587 int
linux_sys_nosys(struct lwp * l,const void * v,register_t * retval)1588 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval)
1589 {
1590 return (ENOSYS);
1591 }
1592
1593 int
linux_sys_getpriority(struct lwp * l,const struct linux_sys_getpriority_args * uap,register_t * retval)1594 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval)
1595 {
1596 /* {
1597 syscallarg(int) which;
1598 syscallarg(int) who;
1599 } */
1600 struct sys_getpriority_args bsa;
1601 int error;
1602
1603 SCARG(&bsa, which) = SCARG(uap, which);
1604 SCARG(&bsa, who) = SCARG(uap, who);
1605
1606 if ((error = sys_getpriority(l, &bsa, retval)))
1607 return error;
1608
1609 *retval = NZERO - *retval;
1610
1611 return 0;
1612 }
1613
1614 int
linux_do_sys_utimensat(struct lwp * l,int fd,const char * path,struct timespec * tsp,int flags,register_t * retval)1615 linux_do_sys_utimensat(struct lwp *l, int fd, const char *path, struct timespec *tsp, int flags, register_t *retval)
1616 {
1617 int follow, error;
1618
1619 follow = (flags & LINUX_AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
1620
1621 if (path == NULL && fd != AT_FDCWD) {
1622 file_t *fp;
1623
1624 /* fd_getvnode() will use the descriptor for us */
1625 if ((error = fd_getvnode(fd, &fp)) != 0)
1626 return error;
1627 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0,
1628 tsp, UIO_SYSSPACE);
1629 fd_putfile(fd);
1630 return error;
1631 }
1632
1633 return do_sys_utimensat(l, fd, NULL, path, follow, tsp, UIO_SYSSPACE);
1634 }
1635
1636 int
linux_sys_utimensat(struct lwp * l,const struct linux_sys_utimensat_args * uap,register_t * retval)1637 linux_sys_utimensat(struct lwp *l, const struct linux_sys_utimensat_args *uap,
1638 register_t *retval)
1639 {
1640 /* {
1641 syscallarg(int) fd;
1642 syscallarg(const char *) path;
1643 syscallarg(const struct linux_timespec *) times;
1644 syscallarg(int) flag;
1645 } */
1646 int error;
1647 struct linux_timespec lts[2];
1648 struct timespec *tsp = NULL, ts[2];
1649
1650 if (SCARG(uap, times)) {
1651 error = copyin(SCARG(uap, times), <s, sizeof(lts));
1652 if (error != 0)
1653 return error;
1654 linux_to_native_timespec(&ts[0], <s[0]);
1655 linux_to_native_timespec(&ts[1], <s[1]);
1656 tsp = ts;
1657 }
1658
1659 return linux_do_sys_utimensat(l, SCARG(uap, fd), SCARG(uap, path),
1660 tsp, SCARG(uap, flag), retval);
1661 }
1662
1663 int
linux_sys_futex(struct lwp * l,const struct linux_sys_futex_args * uap,register_t * retval)1664 linux_sys_futex(struct lwp *l, const struct linux_sys_futex_args *uap,
1665 register_t *retval)
1666 {
1667 /* {
1668 syscallarg(int *) uaddr;
1669 syscallarg(int) op;
1670 syscallarg(int) val;
1671 syscallarg(const struct linux_timespec *) timeout;
1672 syscallarg(int *) uaddr2;
1673 syscallarg(int) val3;
1674 } */
1675 struct linux_timespec lts;
1676 struct timespec ts, *tsp = NULL;
1677 int val2 = 0;
1678 int error;
1679
1680 /*
1681 * Linux overlays the "timeout" field and the "val2" field.
1682 * "timeout" is only valid for FUTEX_WAIT and FUTEX_WAIT_BITSET
1683 * on Linux.
1684 */
1685 const int op = (SCARG(uap, op) & FUTEX_CMD_MASK);
1686 if ((op == FUTEX_WAIT || op == FUTEX_WAIT_BITSET) &&
1687 SCARG(uap, timeout) != NULL) {
1688 if ((error = copyin(SCARG(uap, timeout),
1689 <s, sizeof(lts))) != 0) {
1690 return error;
1691 }
1692 linux_to_native_timespec(&ts, <s);
1693 tsp = &ts;
1694 } else {
1695 val2 = (int)(uintptr_t)SCARG(uap, timeout);
1696 }
1697
1698 return linux_do_futex(SCARG(uap, uaddr), SCARG(uap, op),
1699 SCARG(uap, val), tsp, SCARG(uap, uaddr2), val2,
1700 SCARG(uap, val3), retval);
1701 }
1702
1703 int
linux_do_futex(int * uaddr,int op,int val,struct timespec * timeout,int * uaddr2,int val2,int val3,register_t * retval)1704 linux_do_futex(int *uaddr, int op, int val, struct timespec *timeout,
1705 int *uaddr2, int val2, int val3, register_t *retval)
1706 {
1707 /*
1708 * Always clear FUTEX_PRIVATE_FLAG for Linux processes.
1709 * NetBSD-native futexes exist in different namespace
1710 * depending on FUTEX_PRIVATE_FLAG. This appears not
1711 * to be the case in Linux, and some futex users will
1712 * mix private and non-private ops on the same futex
1713 * object.
1714 */
1715 return do_futex(uaddr, op & ~FUTEX_PRIVATE_FLAG,
1716 val, timeout, uaddr2, val2, val3, retval);
1717 }
1718
1719 #define LINUX_EFD_SEMAPHORE 0x0001
1720 #define LINUX_EFD_CLOEXEC LINUX_O_CLOEXEC
1721 #define LINUX_EFD_NONBLOCK LINUX_O_NONBLOCK
1722
1723 static int
linux_do_eventfd2(struct lwp * l,unsigned int initval,int flags,register_t * retval)1724 linux_do_eventfd2(struct lwp *l, unsigned int initval, int flags,
1725 register_t *retval)
1726 {
1727 int nflags = 0;
1728
1729 if (flags & ~(LINUX_EFD_SEMAPHORE | LINUX_EFD_CLOEXEC |
1730 LINUX_EFD_NONBLOCK)) {
1731 return EINVAL;
1732 }
1733 if (flags & LINUX_EFD_SEMAPHORE) {
1734 nflags |= EFD_SEMAPHORE;
1735 }
1736 if (flags & LINUX_EFD_CLOEXEC) {
1737 nflags |= EFD_CLOEXEC;
1738 }
1739 if (flags & LINUX_EFD_NONBLOCK) {
1740 nflags |= EFD_NONBLOCK;
1741 }
1742
1743 return do_eventfd(l, initval, nflags, retval);
1744 }
1745
1746 int
linux_sys_eventfd(struct lwp * l,const struct linux_sys_eventfd_args * uap,register_t * retval)1747 linux_sys_eventfd(struct lwp *l, const struct linux_sys_eventfd_args *uap,
1748 register_t *retval)
1749 {
1750 /* {
1751 syscallarg(unsigned int) initval;
1752 } */
1753
1754 return linux_do_eventfd2(l, SCARG(uap, initval), 0, retval);
1755 }
1756
1757 int
linux_sys_eventfd2(struct lwp * l,const struct linux_sys_eventfd2_args * uap,register_t * retval)1758 linux_sys_eventfd2(struct lwp *l, const struct linux_sys_eventfd2_args *uap,
1759 register_t *retval)
1760 {
1761 /* {
1762 syscallarg(unsigned int) initval;
1763 syscallarg(int) flags;
1764 } */
1765
1766 return linux_do_eventfd2(l, SCARG(uap, initval), SCARG(uap, flags),
1767 retval);
1768 }
1769
1770 #ifndef __aarch64__
1771 /*
1772 * epoll_create(2). Check size and call sys_epoll_create1.
1773 */
1774 int
linux_sys_epoll_create(struct lwp * l,const struct linux_sys_epoll_create_args * uap,register_t * retval)1775 linux_sys_epoll_create(struct lwp *l,
1776 const struct linux_sys_epoll_create_args *uap, register_t *retval)
1777 {
1778 /* {
1779 syscallarg(int) size;
1780 } */
1781 struct sys_epoll_create1_args ca;
1782
1783 /*
1784 * SCARG(uap, size) is unused. Linux just tests it and then
1785 * forgets it as well.
1786 */
1787 if (SCARG(uap, size) <= 0)
1788 return EINVAL;
1789
1790 SCARG(&ca, flags) = 0;
1791 return sys_epoll_create1(l, &ca, retval);
1792 }
1793 #endif /* !__aarch64__ */
1794
1795 /*
1796 * epoll_create1(2). Translate the flags and call sys_epoll_create1.
1797 */
1798 int
linux_sys_epoll_create1(struct lwp * l,const struct linux_sys_epoll_create1_args * uap,register_t * retval)1799 linux_sys_epoll_create1(struct lwp *l,
1800 const struct linux_sys_epoll_create1_args *uap, register_t *retval)
1801 {
1802 /* {
1803 syscallarg(int) flags;
1804 } */
1805 struct sys_epoll_create1_args ca;
1806
1807 if ((SCARG(uap, flags) & ~(LINUX_O_CLOEXEC)) != 0)
1808 return EINVAL;
1809
1810 SCARG(&ca, flags) = 0;
1811 if ((SCARG(uap, flags) & LINUX_O_CLOEXEC) != 0)
1812 SCARG(&ca, flags) |= EPOLL_CLOEXEC;
1813
1814 return sys_epoll_create1(l, &ca, retval);
1815 }
1816
1817 /*
1818 * epoll_ctl(2). Copyin event and translate it if necessary and then
1819 * call epoll_ctl_common().
1820 */
1821 int
linux_sys_epoll_ctl(struct lwp * l,const struct linux_sys_epoll_ctl_args * uap,register_t * retval)1822 linux_sys_epoll_ctl(struct lwp *l, const struct linux_sys_epoll_ctl_args *uap,
1823 register_t *retval)
1824 {
1825 /* {
1826 syscallarg(int) epfd;
1827 syscallarg(int) op;
1828 syscallarg(int) fd;
1829 syscallarg(struct linux_epoll_event *) event;
1830 } */
1831 struct linux_epoll_event lee;
1832 struct epoll_event ee;
1833 struct epoll_event *eep;
1834 int error;
1835
1836 if (SCARG(uap, op) != EPOLL_CTL_DEL) {
1837 error = copyin(SCARG(uap, event), &lee, sizeof(lee));
1838 if (error != 0)
1839 return error;
1840
1841 /*
1842 * On some architectures, struct linux_epoll_event and
1843 * struct epoll_event are packed differently... but otherwise
1844 * the contents are the same.
1845 */
1846 ee.events = lee.events;
1847 ee.data = lee.data;
1848
1849 eep = ⅇ
1850 } else
1851 eep = NULL;
1852
1853 return epoll_ctl_common(l, retval, SCARG(uap, epfd), SCARG(uap, op),
1854 SCARG(uap, fd), eep);
1855 }
1856
1857 #ifndef __aarch64__
1858 /*
1859 * epoll_wait(2). Call sys_epoll_pwait().
1860 */
1861 int
linux_sys_epoll_wait(struct lwp * l,const struct linux_sys_epoll_wait_args * uap,register_t * retval)1862 linux_sys_epoll_wait(struct lwp *l,
1863 const struct linux_sys_epoll_wait_args *uap, register_t *retval)
1864 {
1865 /* {
1866 syscallarg(int) epfd;
1867 syscallarg(struct linux_epoll_event *) events;
1868 syscallarg(int) maxevents;
1869 syscallarg(int) timeout;
1870 } */
1871 struct linux_sys_epoll_pwait_args ea;
1872
1873 SCARG(&ea, epfd) = SCARG(uap, epfd);
1874 SCARG(&ea, events) = SCARG(uap, events);
1875 SCARG(&ea, maxevents) = SCARG(uap, maxevents);
1876 SCARG(&ea, timeout) = SCARG(uap, timeout);
1877 SCARG(&ea, sigmask) = NULL;
1878
1879 return linux_sys_epoll_pwait(l, &ea, retval);
1880 }
1881 #endif /* !__aarch64__ */
1882
1883 /*
1884 * Main body of epoll_pwait2(2). Translate timeout and sigmask and
1885 * call epoll_wait_common.
1886 */
1887 static int
linux_epoll_pwait2_common(struct lwp * l,register_t * retval,int epfd,struct linux_epoll_event * events,int maxevents,struct linux_timespec * timeout,const linux_sigset_t * sigmask)1888 linux_epoll_pwait2_common(struct lwp *l, register_t *retval, int epfd,
1889 struct linux_epoll_event *events, int maxevents,
1890 struct linux_timespec *timeout, const linux_sigset_t *sigmask)
1891 {
1892 struct timespec ts, *tsp;
1893 linux_sigset_t lss;
1894 sigset_t ss, *ssp;
1895 struct epoll_event *eep;
1896 struct linux_epoll_event *leep;
1897 int i, error;
1898
1899 if (maxevents <= 0 || maxevents > EPOLL_MAX_EVENTS)
1900 return EINVAL;
1901
1902 if (timeout != NULL) {
1903 linux_to_native_timespec(&ts, timeout);
1904 tsp = &ts;
1905 } else
1906 tsp = NULL;
1907
1908 if (sigmask != NULL) {
1909 error = copyin(sigmask, &lss, sizeof(lss));
1910 if (error != 0)
1911 return error;
1912
1913 linux_to_native_sigset(&ss, &lss);
1914 ssp = &ss;
1915 } else
1916 ssp = NULL;
1917
1918 eep = kmem_alloc(maxevents * sizeof(*eep), KM_SLEEP);
1919
1920 error = epoll_wait_common(l, retval, epfd, eep, maxevents, tsp,
1921 ssp);
1922 if (error == 0 && *retval > 0) {
1923 leep = kmem_alloc((*retval) * sizeof(*leep), KM_SLEEP);
1924
1925 /* Translate the events (because of packing). */
1926 for (i = 0; i < *retval; i++) {
1927 leep[i].events = eep[i].events;
1928 leep[i].data = eep[i].data;
1929 }
1930
1931 error = copyout(leep, events, (*retval) * sizeof(*leep));
1932 kmem_free(leep, (*retval) * sizeof(*leep));
1933 }
1934
1935 kmem_free(eep, maxevents * sizeof(*eep));
1936 return error;
1937 }
1938
1939 /*
1940 * epoll_pwait(2). Translate timeout and call sys_epoll_pwait2.
1941 */
1942 int
linux_sys_epoll_pwait(struct lwp * l,const struct linux_sys_epoll_pwait_args * uap,register_t * retval)1943 linux_sys_epoll_pwait(struct lwp *l,
1944 const struct linux_sys_epoll_pwait_args *uap, register_t *retval)
1945 {
1946 /* {
1947 syscallarg(int) epfd;
1948 syscallarg(struct linux_epoll_event *) events;
1949 syscallarg(int) maxevents;
1950 syscallarg(int) timeout;
1951 syscallarg(linux_sigset_t *) sigmask;
1952 } */
1953 struct linux_timespec lts, *ltsp;
1954 const int timeout = SCARG(uap, timeout);
1955
1956 if (timeout >= 0) {
1957 /* Convert from milliseconds to timespec. */
1958 lts.tv_sec = timeout / 1000;
1959 lts.tv_nsec = (timeout % 1000) * 1000000;
1960
1961 ltsp = <s;
1962 } else
1963 ltsp = NULL;
1964
1965 return linux_epoll_pwait2_common(l, retval, SCARG(uap, epfd),
1966 SCARG(uap, events), SCARG(uap, maxevents), ltsp,
1967 SCARG(uap, sigmask));
1968 }
1969
1970
1971 /*
1972 * epoll_pwait2(2). Copyin timeout and call linux_epoll_pwait2_common().
1973 */
1974 int
linux_sys_epoll_pwait2(struct lwp * l,const struct linux_sys_epoll_pwait2_args * uap,register_t * retval)1975 linux_sys_epoll_pwait2(struct lwp *l,
1976 const struct linux_sys_epoll_pwait2_args *uap, register_t *retval)
1977 {
1978 /* {
1979 syscallarg(int) epfd;
1980 syscallarg(struct linux_epoll_event *) events;
1981 syscallarg(int) maxevents;
1982 syscallarg(struct linux_timespec *) timeout;
1983 syscallarg(linux_sigset_t *) sigmask;
1984 } */
1985 struct linux_timespec lts, *ltsp;
1986 int error;
1987
1988 if (SCARG(uap, timeout) != NULL) {
1989 error = copyin(SCARG(uap, timeout), <s, sizeof(lts));
1990 if (error != 0)
1991 return error;
1992
1993 ltsp = <s;
1994 } else
1995 ltsp = NULL;
1996
1997 return linux_epoll_pwait2_common(l, retval, SCARG(uap, epfd),
1998 SCARG(uap, events), SCARG(uap, maxevents), ltsp,
1999 SCARG(uap, sigmask));
2000 }
2001
2002 #define LINUX_MFD_CLOEXEC 0x0001U
2003 #define LINUX_MFD_ALLOW_SEALING 0x0002U
2004 #define LINUX_MFD_HUGETLB 0x0004U
2005 #define LINUX_MFD_NOEXEC_SEAL 0x0008U
2006 #define LINUX_MFD_EXEC 0x0010U
2007 #define LINUX_MFD_HUGE_FLAGS (0x3f << 26)
2008
2009 #define LINUX_MFD_ALL_FLAGS (LINUX_MFD_CLOEXEC|LINUX_MFD_ALLOW_SEALING \
2010 |LINUX_MFD_HUGETLB|LINUX_MFD_NOEXEC_SEAL \
2011 |LINUX_MFD_EXEC|LINUX_MFD_HUGE_FLAGS)
2012 #define LINUX_MFD_KNOWN_FLAGS (LINUX_MFD_CLOEXEC|LINUX_MFD_ALLOW_SEALING)
2013
2014 #define LINUX_MFD_NAME_MAX 249
2015
2016 /*
2017 * memfd_create(2). Do some error checking and then call NetBSD's
2018 * version.
2019 */
2020 int
linux_sys_memfd_create(struct lwp * l,const struct linux_sys_memfd_create_args * uap,register_t * retval)2021 linux_sys_memfd_create(struct lwp *l,
2022 const struct linux_sys_memfd_create_args *uap, register_t *retval)
2023 {
2024 /* {
2025 syscallarg(const char *) name;
2026 syscallarg(unsigned int) flags;
2027 } */
2028 int error;
2029 char *pbuf;
2030 struct sys_memfd_create_args muap;
2031 const unsigned int lflags = SCARG(uap, flags);
2032
2033 KASSERT(LINUX_MFD_NAME_MAX < NAME_MAX); /* sanity check */
2034
2035 if (lflags & ~LINUX_MFD_ALL_FLAGS)
2036 return EINVAL;
2037 if ((lflags & LINUX_MFD_HUGE_FLAGS) != 0 &&
2038 (lflags & LINUX_MFD_HUGETLB) == 0)
2039 return EINVAL;
2040 if ((lflags & LINUX_MFD_HUGETLB) && (lflags & LINUX_MFD_ALLOW_SEALING))
2041 return EINVAL;
2042
2043 /* Linux has a stricter limit for name size */
2044 pbuf = PNBUF_GET();
2045 error = copyinstr(SCARG(uap, name), pbuf, LINUX_MFD_NAME_MAX+1, NULL);
2046 PNBUF_PUT(pbuf);
2047 pbuf = NULL;
2048 if (error != 0) {
2049 if (error == ENAMETOOLONG)
2050 error = EINVAL;
2051 return error;
2052 }
2053
2054 if (lflags & ~LINUX_MFD_KNOWN_FLAGS) {
2055 DPRINTF("%s: ignored flags %#x\n", __func__,
2056 lflags & ~LINUX_MFD_KNOWN_FLAGS);
2057 }
2058
2059 SCARG(&muap, name) = SCARG(uap, name);
2060 SCARG(&muap, flags) = lflags & LINUX_MFD_KNOWN_FLAGS;
2061
2062 return sys_memfd_create(l, &muap, retval);
2063 }
2064
2065 #define LINUX_CLOSE_RANGE_UNSHARE 0x02U
2066 #define LINUX_CLOSE_RANGE_CLOEXEC 0x04U
2067
2068 /*
2069 * close_range(2).
2070 */
2071 int
linux_sys_close_range(struct lwp * l,const struct linux_sys_close_range_args * uap,register_t * retval)2072 linux_sys_close_range(struct lwp *l,
2073 const struct linux_sys_close_range_args *uap, register_t *retval)
2074 {
2075 /* {
2076 syscallarg(unsigned int) first;
2077 syscallarg(unsigned int) last;
2078 syscallarg(unsigned int) flags;
2079 } */
2080 unsigned int fd, last;
2081 file_t *fp;
2082 filedesc_t *fdp;
2083 const unsigned int flags = SCARG(uap, flags);
2084
2085 if (flags & ~(LINUX_CLOSE_RANGE_CLOEXEC|LINUX_CLOSE_RANGE_UNSHARE))
2086 return EINVAL;
2087 if (SCARG(uap, first) > SCARG(uap, last))
2088 return EINVAL;
2089
2090 if (flags & LINUX_CLOSE_RANGE_UNSHARE) {
2091 fdp = fd_copy();
2092 fd_free();
2093 l->l_proc->p_fd = fdp;
2094 l->l_fd = fdp;
2095 }
2096
2097 last = MIN(SCARG(uap, last), l->l_proc->p_fd->fd_lastfile);
2098 for (fd = SCARG(uap, first); fd <= last; fd++) {
2099 fp = fd_getfile(fd);
2100 if (fp == NULL)
2101 continue;
2102
2103 if (flags & LINUX_CLOSE_RANGE_CLOEXEC) {
2104 fd_set_exclose(l, fd, true);
2105 fd_putfile(fd);
2106 } else
2107 fd_close(fd);
2108 }
2109
2110 return 0;
2111 }
2112
2113 /*
2114 * readahead(2). Call posix_fadvise with POSIX_FADV_WILLNEED with some extra
2115 * error checking.
2116 */
2117 int
linux_sys_readahead(struct lwp * l,const struct linux_sys_readahead_args * uap,register_t * retval)2118 linux_sys_readahead(struct lwp *l, const struct linux_sys_readahead_args *uap,
2119 register_t *retval)
2120 {
2121 /* {
2122 syscallarg(int) fd;
2123 syscallarg(off_t) offset;
2124 syscallarg(size_t) count;
2125 } */
2126 file_t *fp;
2127 int error = 0;
2128 const int fd = SCARG(uap, fd);
2129
2130 fp = fd_getfile(fd);
2131 if (fp == NULL)
2132 return EBADF;
2133 if ((fp->f_flag & FREAD) == 0)
2134 error = EBADF;
2135 else if (fp->f_type != DTYPE_VNODE || fp->f_vnode->v_type != VREG)
2136 error = EINVAL;
2137 fd_putfile(fd);
2138 if (error != 0)
2139 return error;
2140
2141 return do_posix_fadvise(fd, SCARG(uap, offset), SCARG(uap, count),
2142 POSIX_FADV_WILLNEED);
2143 }
2144
2145 int
linux_sys_getcpu(lwp_t * l,const struct linux_sys_getcpu_args * uap,register_t * retval)2146 linux_sys_getcpu(lwp_t *l, const struct linux_sys_getcpu_args *uap,
2147 register_t *retval)
2148 {
2149 /* {
2150 syscallarg(unsigned int *) cpu;
2151 syscallarg(unsigned int *) node;
2152 syscallarg(struct linux_getcpu_cache *) tcache;
2153 }*/
2154 int error;
2155
2156 if (SCARG(uap, cpu)) {
2157 u_int cpu_id = l->l_cpu->ci_data.cpu_index;
2158 error = copyout(&cpu_id, SCARG(uap, cpu), sizeof(cpu_id));
2159 if (error)
2160 return error;
2161
2162 }
2163
2164 // TO-DO: Test on a NUMA machine if the node_id returned is correct
2165 if (SCARG(uap, node)) {
2166 u_int node_id = l->l_cpu->ci_data.cpu_numa_id;
2167 error = copyout(&node_id, SCARG(uap, node), sizeof(node_id));
2168 if (error)
2169 return error;
2170 }
2171
2172 return 0;
2173 }
2174