1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 2002 Doug Rabson
5 * Copyright (c) 1994-1995 Søren Schmidt
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer
13 * in this position and unchanged.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. The name of the author may not be used to endorse or promote products
18 * derived from this software without specific prior written permission
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
21 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34
35 #include "opt_compat.h"
36
37 #include <sys/param.h>
38 #include <sys/blist.h>
39 #include <sys/fcntl.h>
40 #if defined(__i386__)
41 #include <sys/imgact_aout.h>
42 #endif
43 #include <sys/jail.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mman.h>
49 #include <sys/mount.h>
50 #include <sys/msgbuf.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/procctl.h>
56 #include <sys/reboot.h>
57 #include <sys/racct.h>
58 #include <sys/random.h>
59 #include <sys/resourcevar.h>
60 #include <sys/sched.h>
61 #include <sys/sdt.h>
62 #include <sys/signalvar.h>
63 #include <sys/stat.h>
64 #include <sys/syscallsubr.h>
65 #include <sys/sysctl.h>
66 #include <sys/sysproto.h>
67 #include <sys/systm.h>
68 #include <sys/time.h>
69 #include <sys/vmmeter.h>
70 #include <sys/vnode.h>
71 #include <sys/wait.h>
72 #include <sys/cpuset.h>
73 #include <sys/uio.h>
74
75 #include <security/mac/mac_framework.h>
76
77 #include <vm/vm.h>
78 #include <vm/pmap.h>
79 #include <vm/vm_kern.h>
80 #include <vm/vm_map.h>
81 #include <vm/vm_extern.h>
82 #include <vm/swap_pager.h>
83
84 #ifdef COMPAT_LINUX32
85 #include <machine/../linux32/linux.h>
86 #include <machine/../linux32/linux32_proto.h>
87 #else
88 #include <machine/../linux/linux.h>
89 #include <machine/../linux/linux_proto.h>
90 #endif
91
92 #include <compat/linux/linux_dtrace.h>
93 #include <compat/linux/linux_file.h>
94 #include <compat/linux/linux_mib.h>
95 #include <compat/linux/linux_signal.h>
96 #include <compat/linux/linux_timer.h>
97 #include <compat/linux/linux_util.h>
98 #include <compat/linux/linux_sysproto.h>
99 #include <compat/linux/linux_emul.h>
100 #include <compat/linux/linux_misc.h>
101
102 int stclohz; /* Statistics clock frequency */
103
104 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
105 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
106 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
107 RLIMIT_MEMLOCK, RLIMIT_AS
108 };
109
110 struct l_sysinfo {
111 l_long uptime; /* Seconds since boot */
112 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
113 #define LINUX_SYSINFO_LOADS_SCALE 65536
114 l_ulong totalram; /* Total usable main memory size */
115 l_ulong freeram; /* Available memory size */
116 l_ulong sharedram; /* Amount of shared memory */
117 l_ulong bufferram; /* Memory used by buffers */
118 l_ulong totalswap; /* Total swap space size */
119 l_ulong freeswap; /* swap space still available */
120 l_ushort procs; /* Number of current processes */
121 l_ushort pads;
122 l_ulong totalhigh;
123 l_ulong freehigh;
124 l_uint mem_unit;
125 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
126 };
127
128 struct l_pselect6arg {
129 l_uintptr_t ss;
130 l_size_t ss_len;
131 };
132
133 static int linux_utimensat_nsec_valid(l_long);
134
135
136 int
137 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
138 {
139 struct l_sysinfo sysinfo;
140 int i, j;
141 struct timespec ts;
142
143 bzero(&sysinfo, sizeof(sysinfo));
144 getnanouptime(&ts);
145 if (ts.tv_nsec != 0)
146 ts.tv_sec++;
147 sysinfo.uptime = ts.tv_sec;
148
149 /* Use the information from the mib to get our load averages */
150 for (i = 0; i < 3; i++)
151 sysinfo.loads[i] = averunnable.ldavg[i] *
152 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
153
154 sysinfo.totalram = physmem * PAGE_SIZE;
155 sysinfo.freeram = (u_long)vm_free_count() * PAGE_SIZE;
156
157 /*
158 * sharedram counts pages allocated to named, swap-backed objects such
159 * as shared memory segments and tmpfs files. There is no cheap way to
160 * compute this, so just leave the field unpopulated. Linux itself only
161 * started setting this field in the 3.x timeframe.
162 */
163 sysinfo.sharedram = 0;
164 sysinfo.bufferram = 0;
165
166 swap_pager_status(&i, &j);
167 sysinfo.totalswap = i * PAGE_SIZE;
168 sysinfo.freeswap = (i - j) * PAGE_SIZE;
169
170 sysinfo.procs = nprocs;
171
172 /*
173 * Platforms supported by the emulation layer do not have a notion of
174 * high memory.
175 */
176 sysinfo.totalhigh = 0;
177 sysinfo.freehigh = 0;
178
179 sysinfo.mem_unit = 1;
180
181 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
182 }
183
184 #ifdef LINUX_LEGACY_SYSCALLS
185 int
186 linux_alarm(struct thread *td, struct linux_alarm_args *args)
187 {
188 struct itimerval it, old_it;
189 u_int secs;
190 int error;
191
192 secs = args->secs;
193 /*
194 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
195 * to match kern_setitimer()'s limit to avoid error from it.
196 *
197 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
198 * platforms.
199 */
200 if (secs > INT32_MAX / 2)
201 secs = INT32_MAX / 2;
202
203 it.it_value.tv_sec = secs;
204 it.it_value.tv_usec = 0;
205 timevalclear(&it.it_interval);
206 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
207 KASSERT(error == 0, ("kern_setitimer returns %d", error));
208
209 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
210 old_it.it_value.tv_usec >= 500000)
211 old_it.it_value.tv_sec++;
212 td->td_retval[0] = old_it.it_value.tv_sec;
213 return (0);
214 }
215 #endif
216
217 int
218 linux_brk(struct thread *td, struct linux_brk_args *args)
219 {
220 struct vmspace *vm = td->td_proc->p_vmspace;
221 uintptr_t new, old;
222
223 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
224 new = (uintptr_t)args->dsend;
225 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
226 td->td_retval[0] = (register_t)new;
227 else
228 td->td_retval[0] = (register_t)old;
229
230 return (0);
231 }
232
233 #if defined(__i386__)
234 /* XXX: what about amd64/linux32? */
235
236 int
237 linux_uselib(struct thread *td, struct linux_uselib_args *args)
238 {
239 struct nameidata ni;
240 struct vnode *vp;
241 struct exec *a_out;
242 vm_map_t map;
243 vm_map_entry_t entry;
244 struct vattr attr;
245 vm_offset_t vmaddr;
246 unsigned long file_offset;
247 unsigned long bss_size;
248 char *library;
249 ssize_t aresid;
250 int error;
251 bool locked, opened, textset;
252
253 LCONVPATHEXIST(td, args->library, &library);
254
255 a_out = NULL;
256 vp = NULL;
257 locked = false;
258 textset = false;
259 opened = false;
260
261 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
262 UIO_SYSSPACE, library, td);
263 error = namei(&ni);
264 LFREEPATH(library);
265 if (error)
266 goto cleanup;
267
268 vp = ni.ni_vp;
269 NDFREE(&ni, NDF_ONLY_PNBUF);
270
271 /*
272 * From here on down, we have a locked vnode that must be unlocked.
273 * XXX: The code below largely duplicates exec_check_permissions().
274 */
275 locked = true;
276
277 /* Executable? */
278 error = VOP_GETATTR(vp, &attr, td->td_ucred);
279 if (error)
280 goto cleanup;
281
282 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
283 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
284 /* EACCESS is what exec(2) returns. */
285 error = ENOEXEC;
286 goto cleanup;
287 }
288
289 /* Sensible size? */
290 if (attr.va_size == 0) {
291 error = ENOEXEC;
292 goto cleanup;
293 }
294
295 /* Can we access it? */
296 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
297 if (error)
298 goto cleanup;
299
300 /*
301 * XXX: This should use vn_open() so that it is properly authorized,
302 * and to reduce code redundancy all over the place here.
303 * XXX: Not really, it duplicates far more of exec_check_permissions()
304 * than vn_open().
305 */
306 #ifdef MAC
307 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
308 if (error)
309 goto cleanup;
310 #endif
311 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
312 if (error)
313 goto cleanup;
314 opened = true;
315
316 /* Pull in executable header into exec_map */
317 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
318 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
319 if (error)
320 goto cleanup;
321
322 /* Is it a Linux binary ? */
323 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
324 error = ENOEXEC;
325 goto cleanup;
326 }
327
328 /*
329 * While we are here, we should REALLY do some more checks
330 */
331
332 /* Set file/virtual offset based on a.out variant. */
333 switch ((int)(a_out->a_magic & 0xffff)) {
334 case 0413: /* ZMAGIC */
335 file_offset = 1024;
336 break;
337 case 0314: /* QMAGIC */
338 file_offset = 0;
339 break;
340 default:
341 error = ENOEXEC;
342 goto cleanup;
343 }
344
345 bss_size = round_page(a_out->a_bss);
346
347 /* Check various fields in header for validity/bounds. */
348 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
349 error = ENOEXEC;
350 goto cleanup;
351 }
352
353 /* text + data can't exceed file size */
354 if (a_out->a_data + a_out->a_text > attr.va_size) {
355 error = EFAULT;
356 goto cleanup;
357 }
358
359 /*
360 * text/data/bss must not exceed limits
361 * XXX - this is not complete. it should check current usage PLUS
362 * the resources needed by this library.
363 */
364 PROC_LOCK(td->td_proc);
365 if (a_out->a_text > maxtsiz ||
366 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
367 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
368 bss_size) != 0) {
369 PROC_UNLOCK(td->td_proc);
370 error = ENOMEM;
371 goto cleanup;
372 }
373 PROC_UNLOCK(td->td_proc);
374
375 /*
376 * Prevent more writers.
377 */
378 error = VOP_SET_TEXT(vp);
379 if (error != 0)
380 goto cleanup;
381 textset = true;
382
383 /*
384 * Lock no longer needed
385 */
386 locked = false;
387 VOP_UNLOCK(vp, 0);
388
389 /*
390 * Check if file_offset page aligned. Currently we cannot handle
391 * misalinged file offsets, and so we read in the entire image
392 * (what a waste).
393 */
394 if (file_offset & PAGE_MASK) {
395 /* Map text+data read/write/execute */
396
397 /* a_entry is the load address and is page aligned */
398 vmaddr = trunc_page(a_out->a_entry);
399
400 /* get anon user mapping, read+write+execute */
401 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
402 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
403 VM_PROT_ALL, VM_PROT_ALL, 0);
404 if (error)
405 goto cleanup;
406
407 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
408 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
409 td->td_ucred, NOCRED, &aresid, td);
410 if (error != 0)
411 goto cleanup;
412 if (aresid != 0) {
413 error = ENOEXEC;
414 goto cleanup;
415 }
416 } else {
417 /*
418 * for QMAGIC, a_entry is 20 bytes beyond the load address
419 * to skip the executable header
420 */
421 vmaddr = trunc_page(a_out->a_entry);
422
423 /*
424 * Map it all into the process's space as a single
425 * copy-on-write "data" segment.
426 */
427 map = &td->td_proc->p_vmspace->vm_map;
428 error = vm_mmap(map, &vmaddr,
429 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
430 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
431 if (error)
432 goto cleanup;
433 vm_map_lock(map);
434 if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
435 vm_map_unlock(map);
436 error = EDOOFUS;
437 goto cleanup;
438 }
439 entry->eflags |= MAP_ENTRY_VN_EXEC;
440 vm_map_unlock(map);
441 textset = false;
442 }
443
444 if (bss_size != 0) {
445 /* Calculate BSS start address */
446 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
447 a_out->a_data;
448
449 /* allocate some 'anon' space */
450 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
451 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
452 VM_PROT_ALL, 0);
453 if (error)
454 goto cleanup;
455 }
456
457 cleanup:
458 if (opened) {
459 if (locked)
460 VOP_UNLOCK(vp, 0);
461 locked = false;
462 VOP_CLOSE(vp, FREAD, td->td_ucred, td);
463 }
464 if (textset) {
465 if (!locked) {
466 locked = true;
467 VOP_LOCK(vp, LK_SHARED | LK_RETRY);
468 }
469 VOP_UNSET_TEXT_CHECKED(vp);
470 }
471 if (locked)
472 VOP_UNLOCK(vp, 0);
473
474 /* Release the temporary mapping. */
475 if (a_out)
476 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
477
478 return (error);
479 }
480
481 #endif /* __i386__ */
482
483 #ifdef LINUX_LEGACY_SYSCALLS
484 int
485 linux_select(struct thread *td, struct linux_select_args *args)
486 {
487 l_timeval ltv;
488 struct timeval tv0, tv1, utv, *tvp;
489 int error;
490
491 /*
492 * Store current time for computation of the amount of
493 * time left.
494 */
495 if (args->timeout) {
496 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
497 goto select_out;
498 utv.tv_sec = ltv.tv_sec;
499 utv.tv_usec = ltv.tv_usec;
500
501 if (itimerfix(&utv)) {
502 /*
503 * The timeval was invalid. Convert it to something
504 * valid that will act as it does under Linux.
505 */
506 utv.tv_sec += utv.tv_usec / 1000000;
507 utv.tv_usec %= 1000000;
508 if (utv.tv_usec < 0) {
509 utv.tv_sec -= 1;
510 utv.tv_usec += 1000000;
511 }
512 if (utv.tv_sec < 0)
513 timevalclear(&utv);
514 }
515 microtime(&tv0);
516 tvp = &utv;
517 } else
518 tvp = NULL;
519
520 error = kern_select(td, args->nfds, args->readfds, args->writefds,
521 args->exceptfds, tvp, LINUX_NFDBITS);
522 if (error)
523 goto select_out;
524
525 if (args->timeout) {
526 if (td->td_retval[0]) {
527 /*
528 * Compute how much time was left of the timeout,
529 * by subtracting the current time and the time
530 * before we started the call, and subtracting
531 * that result from the user-supplied value.
532 */
533 microtime(&tv1);
534 timevalsub(&tv1, &tv0);
535 timevalsub(&utv, &tv1);
536 if (utv.tv_sec < 0)
537 timevalclear(&utv);
538 } else
539 timevalclear(&utv);
540 ltv.tv_sec = utv.tv_sec;
541 ltv.tv_usec = utv.tv_usec;
542 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
543 goto select_out;
544 }
545
546 select_out:
547 return (error);
548 }
549 #endif
550
551 int
552 linux_mremap(struct thread *td, struct linux_mremap_args *args)
553 {
554 uintptr_t addr;
555 size_t len;
556 int error = 0;
557
558 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
559 td->td_retval[0] = 0;
560 return (EINVAL);
561 }
562
563 /*
564 * Check for the page alignment.
565 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
566 */
567 if (args->addr & PAGE_MASK) {
568 td->td_retval[0] = 0;
569 return (EINVAL);
570 }
571
572 args->new_len = round_page(args->new_len);
573 args->old_len = round_page(args->old_len);
574
575 if (args->new_len > args->old_len) {
576 td->td_retval[0] = 0;
577 return (ENOMEM);
578 }
579
580 if (args->new_len < args->old_len) {
581 addr = args->addr + args->new_len;
582 len = args->old_len - args->new_len;
583 error = kern_munmap(td, addr, len);
584 }
585
586 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
587 return (error);
588 }
589
590 #define LINUX_MS_ASYNC 0x0001
591 #define LINUX_MS_INVALIDATE 0x0002
592 #define LINUX_MS_SYNC 0x0004
593
594 int
595 linux_msync(struct thread *td, struct linux_msync_args *args)
596 {
597
598 return (kern_msync(td, args->addr, args->len,
599 args->fl & ~LINUX_MS_SYNC));
600 }
601
602 #ifdef LINUX_LEGACY_SYSCALLS
603 int
604 linux_time(struct thread *td, struct linux_time_args *args)
605 {
606 struct timeval tv;
607 l_time_t tm;
608 int error;
609
610 microtime(&tv);
611 tm = tv.tv_sec;
612 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
613 return (error);
614 td->td_retval[0] = tm;
615 return (0);
616 }
617 #endif
618
619 struct l_times_argv {
620 l_clock_t tms_utime;
621 l_clock_t tms_stime;
622 l_clock_t tms_cutime;
623 l_clock_t tms_cstime;
624 };
625
626
627 /*
628 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
629 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
630 * auxiliary vector entry.
631 */
632 #define CLK_TCK 100
633
634 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
635 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
636
637 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
638 CONVNTCK(r) : CONVOTCK(r))
639
640 int
641 linux_times(struct thread *td, struct linux_times_args *args)
642 {
643 struct timeval tv, utime, stime, cutime, cstime;
644 struct l_times_argv tms;
645 struct proc *p;
646 int error;
647
648 if (args->buf != NULL) {
649 p = td->td_proc;
650 PROC_LOCK(p);
651 PROC_STATLOCK(p);
652 calcru(p, &utime, &stime);
653 PROC_STATUNLOCK(p);
654 calccru(p, &cutime, &cstime);
655 PROC_UNLOCK(p);
656
657 tms.tms_utime = CONVTCK(utime);
658 tms.tms_stime = CONVTCK(stime);
659
660 tms.tms_cutime = CONVTCK(cutime);
661 tms.tms_cstime = CONVTCK(cstime);
662
663 if ((error = copyout(&tms, args->buf, sizeof(tms))))
664 return (error);
665 }
666
667 microuptime(&tv);
668 td->td_retval[0] = (int)CONVTCK(tv);
669 return (0);
670 }
671
672 int
673 linux_newuname(struct thread *td, struct linux_newuname_args *args)
674 {
675 struct l_new_utsname utsname;
676 char osname[LINUX_MAX_UTSNAME];
677 char osrelease[LINUX_MAX_UTSNAME];
678 char *p;
679
680 linux_get_osname(td, osname);
681 linux_get_osrelease(td, osrelease);
682
683 bzero(&utsname, sizeof(utsname));
684 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
685 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
686 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
687 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
688 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
689 for (p = utsname.version; *p != '\0'; ++p)
690 if (*p == '\n') {
691 *p = '\0';
692 break;
693 }
694 #if defined(__amd64__)
695 /*
696 * On amd64, Linux uname(2) needs to return "x86_64"
697 * for both 64-bit and 32-bit applications. On 32-bit,
698 * the string returned by getauxval(AT_PLATFORM) needs
699 * to remain "i686", though.
700 */
701 strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
702 #else
703 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
704 #endif
705
706 return (copyout(&utsname, args->buf, sizeof(utsname)));
707 }
708
709 struct l_utimbuf {
710 l_time_t l_actime;
711 l_time_t l_modtime;
712 };
713
714 #ifdef LINUX_LEGACY_SYSCALLS
715 int
716 linux_utime(struct thread *td, struct linux_utime_args *args)
717 {
718 struct timeval tv[2], *tvp;
719 struct l_utimbuf lut;
720 char *fname;
721 int error;
722
723 LCONVPATHEXIST(td, args->fname, &fname);
724
725 if (args->times) {
726 if ((error = copyin(args->times, &lut, sizeof lut))) {
727 LFREEPATH(fname);
728 return (error);
729 }
730 tv[0].tv_sec = lut.l_actime;
731 tv[0].tv_usec = 0;
732 tv[1].tv_sec = lut.l_modtime;
733 tv[1].tv_usec = 0;
734 tvp = tv;
735 } else
736 tvp = NULL;
737
738 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
739 UIO_SYSSPACE);
740 LFREEPATH(fname);
741 return (error);
742 }
743 #endif
744
745 #ifdef LINUX_LEGACY_SYSCALLS
746 int
747 linux_utimes(struct thread *td, struct linux_utimes_args *args)
748 {
749 l_timeval ltv[2];
750 struct timeval tv[2], *tvp = NULL;
751 char *fname;
752 int error;
753
754 LCONVPATHEXIST(td, args->fname, &fname);
755
756 if (args->tptr != NULL) {
757 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
758 LFREEPATH(fname);
759 return (error);
760 }
761 tv[0].tv_sec = ltv[0].tv_sec;
762 tv[0].tv_usec = ltv[0].tv_usec;
763 tv[1].tv_sec = ltv[1].tv_sec;
764 tv[1].tv_usec = ltv[1].tv_usec;
765 tvp = tv;
766 }
767
768 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
769 tvp, UIO_SYSSPACE);
770 LFREEPATH(fname);
771 return (error);
772 }
773 #endif
774
775 static int
776 linux_utimensat_nsec_valid(l_long nsec)
777 {
778
779 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
780 return (0);
781 if (nsec >= 0 && nsec <= 999999999)
782 return (0);
783 return (1);
784 }
785
786 int
787 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
788 {
789 struct l_timespec l_times[2];
790 struct timespec times[2], *timesp = NULL;
791 char *path = NULL;
792 int error, dfd, flags = 0;
793
794 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
795
796 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
797 return (EINVAL);
798
799 if (args->times != NULL) {
800 error = copyin(args->times, l_times, sizeof(l_times));
801 if (error != 0)
802 return (error);
803
804 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
805 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
806 return (EINVAL);
807
808 times[0].tv_sec = l_times[0].tv_sec;
809 switch (l_times[0].tv_nsec)
810 {
811 case LINUX_UTIME_OMIT:
812 times[0].tv_nsec = UTIME_OMIT;
813 break;
814 case LINUX_UTIME_NOW:
815 times[0].tv_nsec = UTIME_NOW;
816 break;
817 default:
818 times[0].tv_nsec = l_times[0].tv_nsec;
819 }
820
821 times[1].tv_sec = l_times[1].tv_sec;
822 switch (l_times[1].tv_nsec)
823 {
824 case LINUX_UTIME_OMIT:
825 times[1].tv_nsec = UTIME_OMIT;
826 break;
827 case LINUX_UTIME_NOW:
828 times[1].tv_nsec = UTIME_NOW;
829 break;
830 default:
831 times[1].tv_nsec = l_times[1].tv_nsec;
832 break;
833 }
834 timesp = times;
835
836 /* This breaks POSIX, but is what the Linux kernel does
837 * _on purpose_ (documented in the man page for utimensat(2)),
838 * so we must follow that behaviour. */
839 if (times[0].tv_nsec == UTIME_OMIT &&
840 times[1].tv_nsec == UTIME_OMIT)
841 return (0);
842 }
843
844 if (args->pathname != NULL)
845 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
846 else if (args->flags != 0)
847 return (EINVAL);
848
849 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
850 flags |= AT_SYMLINK_NOFOLLOW;
851
852 if (path == NULL)
853 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
854 else {
855 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
856 UIO_SYSSPACE, flags);
857 LFREEPATH(path);
858 }
859
860 return (error);
861 }
862
863 #ifdef LINUX_LEGACY_SYSCALLS
864 int
865 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
866 {
867 l_timeval ltv[2];
868 struct timeval tv[2], *tvp = NULL;
869 char *fname;
870 int error, dfd;
871
872 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
873 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
874
875 if (args->utimes != NULL) {
876 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
877 LFREEPATH(fname);
878 return (error);
879 }
880 tv[0].tv_sec = ltv[0].tv_sec;
881 tv[0].tv_usec = ltv[0].tv_usec;
882 tv[1].tv_sec = ltv[1].tv_sec;
883 tv[1].tv_usec = ltv[1].tv_usec;
884 tvp = tv;
885 }
886
887 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
888 LFREEPATH(fname);
889 return (error);
890 }
891 #endif
892
893 static int
894 linux_common_wait(struct thread *td, int pid, int *statusp,
895 int options, struct __wrusage *wrup)
896 {
897 siginfo_t siginfo;
898 idtype_t idtype;
899 id_t id;
900 int error, status, tmpstat;
901
902 if (pid == WAIT_ANY) {
903 idtype = P_ALL;
904 id = 0;
905 } else if (pid < 0) {
906 idtype = P_PGID;
907 id = (id_t)-pid;
908 } else {
909 idtype = P_PID;
910 id = (id_t)pid;
911 }
912
913 /*
914 * For backward compatibility we implicitly add flags WEXITED
915 * and WTRAPPED here.
916 */
917 options |= WEXITED | WTRAPPED;
918 error = kern_wait6(td, idtype, id, &status, options, wrup, &siginfo);
919 if (error)
920 return (error);
921
922 if (statusp) {
923 tmpstat = status & 0xffff;
924 if (WIFSIGNALED(tmpstat)) {
925 tmpstat = (tmpstat & 0xffffff80) |
926 bsd_to_linux_signal(WTERMSIG(tmpstat));
927 } else if (WIFSTOPPED(tmpstat)) {
928 tmpstat = (tmpstat & 0xffff00ff) |
929 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
930 #if defined(__amd64__) && !defined(COMPAT_LINUX32)
931 if (WSTOPSIG(status) == SIGTRAP) {
932 tmpstat = linux_ptrace_status(td,
933 siginfo.si_pid, tmpstat);
934 }
935 #endif
936 } else if (WIFCONTINUED(tmpstat)) {
937 tmpstat = 0xffff;
938 }
939 error = copyout(&tmpstat, statusp, sizeof(int));
940 }
941
942 return (error);
943 }
944
945 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
946 int
947 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
948 {
949 struct linux_wait4_args wait4_args;
950
951 wait4_args.pid = args->pid;
952 wait4_args.status = args->status;
953 wait4_args.options = args->options;
954 wait4_args.rusage = NULL;
955
956 return (linux_wait4(td, &wait4_args));
957 }
958 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
959
960 int
961 linux_wait4(struct thread *td, struct linux_wait4_args *args)
962 {
963 int error, options;
964 struct __wrusage wru, *wrup;
965
966 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
967 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
968 return (EINVAL);
969
970 options = WEXITED;
971 linux_to_bsd_waitopts(args->options, &options);
972
973 if (args->rusage != NULL)
974 wrup = &wru;
975 else
976 wrup = NULL;
977 error = linux_common_wait(td, args->pid, args->status, options, wrup);
978 if (error != 0)
979 return (error);
980 if (args->rusage != NULL)
981 error = linux_copyout_rusage(&wru.wru_self, args->rusage);
982 return (error);
983 }
984
985 int
986 linux_waitid(struct thread *td, struct linux_waitid_args *args)
987 {
988 int status, options, sig;
989 struct __wrusage wru;
990 siginfo_t siginfo;
991 l_siginfo_t lsi;
992 idtype_t idtype;
993 struct proc *p;
994 int error;
995
996 options = 0;
997 linux_to_bsd_waitopts(args->options, &options);
998
999 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1000 return (EINVAL);
1001 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1002 return (EINVAL);
1003
1004 switch (args->idtype) {
1005 case LINUX_P_ALL:
1006 idtype = P_ALL;
1007 break;
1008 case LINUX_P_PID:
1009 if (args->id <= 0)
1010 return (EINVAL);
1011 idtype = P_PID;
1012 break;
1013 case LINUX_P_PGID:
1014 if (args->id <= 0)
1015 return (EINVAL);
1016 idtype = P_PGID;
1017 break;
1018 default:
1019 return (EINVAL);
1020 }
1021
1022 error = kern_wait6(td, idtype, args->id, &status, options,
1023 &wru, &siginfo);
1024 if (error != 0)
1025 return (error);
1026 if (args->rusage != NULL) {
1027 error = linux_copyout_rusage(&wru.wru_children,
1028 args->rusage);
1029 if (error != 0)
1030 return (error);
1031 }
1032 if (args->info != NULL) {
1033 p = td->td_proc;
1034 bzero(&lsi, sizeof(lsi));
1035 if (td->td_retval[0] != 0) {
1036 sig = bsd_to_linux_signal(siginfo.si_signo);
1037 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1038 }
1039 error = copyout(&lsi, args->info, sizeof(lsi));
1040 }
1041 td->td_retval[0] = 0;
1042
1043 return (error);
1044 }
1045
1046 #ifdef LINUX_LEGACY_SYSCALLS
1047 int
1048 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1049 {
1050 char *path;
1051 int error;
1052
1053 LCONVPATHCREAT(td, args->path, &path);
1054
1055 switch (args->mode & S_IFMT) {
1056 case S_IFIFO:
1057 case S_IFSOCK:
1058 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1059 args->mode);
1060 break;
1061
1062 case S_IFCHR:
1063 case S_IFBLK:
1064 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1065 args->mode, args->dev);
1066 break;
1067
1068 case S_IFDIR:
1069 error = EPERM;
1070 break;
1071
1072 case 0:
1073 args->mode |= S_IFREG;
1074 /* FALLTHROUGH */
1075 case S_IFREG:
1076 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1077 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1078 if (error == 0)
1079 kern_close(td, td->td_retval[0]);
1080 break;
1081
1082 default:
1083 error = EINVAL;
1084 break;
1085 }
1086 LFREEPATH(path);
1087 return (error);
1088 }
1089 #endif
1090
1091 int
1092 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1093 {
1094 char *path;
1095 int error, dfd;
1096
1097 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1098 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1099
1100 switch (args->mode & S_IFMT) {
1101 case S_IFIFO:
1102 case S_IFSOCK:
1103 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1104 break;
1105
1106 case S_IFCHR:
1107 case S_IFBLK:
1108 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1109 args->dev);
1110 break;
1111
1112 case S_IFDIR:
1113 error = EPERM;
1114 break;
1115
1116 case 0:
1117 args->mode |= S_IFREG;
1118 /* FALLTHROUGH */
1119 case S_IFREG:
1120 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1121 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1122 if (error == 0)
1123 kern_close(td, td->td_retval[0]);
1124 break;
1125
1126 default:
1127 error = EINVAL;
1128 break;
1129 }
1130 LFREEPATH(path);
1131 return (error);
1132 }
1133
1134 /*
1135 * UGH! This is just about the dumbest idea I've ever heard!!
1136 */
1137 int
1138 linux_personality(struct thread *td, struct linux_personality_args *args)
1139 {
1140 struct linux_pemuldata *pem;
1141 struct proc *p = td->td_proc;
1142 uint32_t old;
1143
1144 PROC_LOCK(p);
1145 pem = pem_find(p);
1146 old = pem->persona;
1147 if (args->per != 0xffffffff)
1148 pem->persona = args->per;
1149 PROC_UNLOCK(p);
1150
1151 td->td_retval[0] = old;
1152 return (0);
1153 }
1154
1155 struct l_itimerval {
1156 l_timeval it_interval;
1157 l_timeval it_value;
1158 };
1159
1160 #define B2L_ITIMERVAL(bip, lip) \
1161 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1162 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1163 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1164 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1165
1166 int
1167 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1168 {
1169 int error;
1170 struct l_itimerval ls;
1171 struct itimerval aitv, oitv;
1172
1173 if (uap->itv == NULL) {
1174 uap->itv = uap->oitv;
1175 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1176 }
1177
1178 error = copyin(uap->itv, &ls, sizeof(ls));
1179 if (error != 0)
1180 return (error);
1181 B2L_ITIMERVAL(&aitv, &ls);
1182 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1183 if (error != 0 || uap->oitv == NULL)
1184 return (error);
1185 B2L_ITIMERVAL(&ls, &oitv);
1186
1187 return (copyout(&ls, uap->oitv, sizeof(ls)));
1188 }
1189
1190 int
1191 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1192 {
1193 int error;
1194 struct l_itimerval ls;
1195 struct itimerval aitv;
1196
1197 error = kern_getitimer(td, uap->which, &aitv);
1198 if (error != 0)
1199 return (error);
1200 B2L_ITIMERVAL(&ls, &aitv);
1201 return (copyout(&ls, uap->itv, sizeof(ls)));
1202 }
1203
1204 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1205 int
1206 linux_nice(struct thread *td, struct linux_nice_args *args)
1207 {
1208 struct setpriority_args bsd_args;
1209
1210 bsd_args.which = PRIO_PROCESS;
1211 bsd_args.who = 0; /* current process */
1212 bsd_args.prio = args->inc;
1213 return (sys_setpriority(td, &bsd_args));
1214 }
1215 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1216
1217 int
1218 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1219 {
1220 struct ucred *newcred, *oldcred;
1221 l_gid_t *linux_gidset;
1222 gid_t *bsd_gidset;
1223 int ngrp, error;
1224 struct proc *p;
1225
1226 ngrp = args->gidsetsize;
1227 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1228 return (EINVAL);
1229 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1230 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1231 if (error)
1232 goto out;
1233 newcred = crget();
1234 crextend(newcred, ngrp + 1);
1235 p = td->td_proc;
1236 PROC_LOCK(p);
1237 oldcred = p->p_ucred;
1238 crcopy(newcred, oldcred);
1239
1240 /*
1241 * cr_groups[0] holds egid. Setting the whole set from
1242 * the supplied set will cause egid to be changed too.
1243 * Keep cr_groups[0] unchanged to prevent that.
1244 */
1245
1246 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1247 PROC_UNLOCK(p);
1248 crfree(newcred);
1249 goto out;
1250 }
1251
1252 if (ngrp > 0) {
1253 newcred->cr_ngroups = ngrp + 1;
1254
1255 bsd_gidset = newcred->cr_groups;
1256 ngrp--;
1257 while (ngrp >= 0) {
1258 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1259 ngrp--;
1260 }
1261 } else
1262 newcred->cr_ngroups = 1;
1263
1264 setsugid(p);
1265 proc_set_cred(p, newcred);
1266 PROC_UNLOCK(p);
1267 crfree(oldcred);
1268 error = 0;
1269 out:
1270 free(linux_gidset, M_LINUX);
1271 return (error);
1272 }
1273
1274 int
1275 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1276 {
1277 struct ucred *cred;
1278 l_gid_t *linux_gidset;
1279 gid_t *bsd_gidset;
1280 int bsd_gidsetsz, ngrp, error;
1281
1282 cred = td->td_ucred;
1283 bsd_gidset = cred->cr_groups;
1284 bsd_gidsetsz = cred->cr_ngroups - 1;
1285
1286 /*
1287 * cr_groups[0] holds egid. Returning the whole set
1288 * here will cause a duplicate. Exclude cr_groups[0]
1289 * to prevent that.
1290 */
1291
1292 if ((ngrp = args->gidsetsize) == 0) {
1293 td->td_retval[0] = bsd_gidsetsz;
1294 return (0);
1295 }
1296
1297 if (ngrp < bsd_gidsetsz)
1298 return (EINVAL);
1299
1300 ngrp = 0;
1301 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1302 M_LINUX, M_WAITOK);
1303 while (ngrp < bsd_gidsetsz) {
1304 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1305 ngrp++;
1306 }
1307
1308 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1309 free(linux_gidset, M_LINUX);
1310 if (error)
1311 return (error);
1312
1313 td->td_retval[0] = ngrp;
1314 return (0);
1315 }
1316
1317 int
1318 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1319 {
1320 struct rlimit bsd_rlim;
1321 struct l_rlimit rlim;
1322 u_int which;
1323 int error;
1324
1325 if (args->resource >= LINUX_RLIM_NLIMITS)
1326 return (EINVAL);
1327
1328 which = linux_to_bsd_resource[args->resource];
1329 if (which == -1)
1330 return (EINVAL);
1331
1332 error = copyin(args->rlim, &rlim, sizeof(rlim));
1333 if (error)
1334 return (error);
1335
1336 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1337 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1338 return (kern_setrlimit(td, which, &bsd_rlim));
1339 }
1340
1341 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1342 int
1343 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1344 {
1345 struct l_rlimit rlim;
1346 struct rlimit bsd_rlim;
1347 u_int which;
1348
1349 if (args->resource >= LINUX_RLIM_NLIMITS)
1350 return (EINVAL);
1351
1352 which = linux_to_bsd_resource[args->resource];
1353 if (which == -1)
1354 return (EINVAL);
1355
1356 lim_rlimit(td, which, &bsd_rlim);
1357
1358 #ifdef COMPAT_LINUX32
1359 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1360 if (rlim.rlim_cur == UINT_MAX)
1361 rlim.rlim_cur = INT_MAX;
1362 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1363 if (rlim.rlim_max == UINT_MAX)
1364 rlim.rlim_max = INT_MAX;
1365 #else
1366 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1367 if (rlim.rlim_cur == ULONG_MAX)
1368 rlim.rlim_cur = LONG_MAX;
1369 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1370 if (rlim.rlim_max == ULONG_MAX)
1371 rlim.rlim_max = LONG_MAX;
1372 #endif
1373 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1374 }
1375 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1376
1377 int
1378 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1379 {
1380 struct l_rlimit rlim;
1381 struct rlimit bsd_rlim;
1382 u_int which;
1383
1384 if (args->resource >= LINUX_RLIM_NLIMITS)
1385 return (EINVAL);
1386
1387 which = linux_to_bsd_resource[args->resource];
1388 if (which == -1)
1389 return (EINVAL);
1390
1391 lim_rlimit(td, which, &bsd_rlim);
1392
1393 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1394 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1395 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1396 }
1397
1398 int
1399 linux_sched_setscheduler(struct thread *td,
1400 struct linux_sched_setscheduler_args *args)
1401 {
1402 struct sched_param sched_param;
1403 struct thread *tdt;
1404 int error, policy;
1405
1406 switch (args->policy) {
1407 case LINUX_SCHED_OTHER:
1408 policy = SCHED_OTHER;
1409 break;
1410 case LINUX_SCHED_FIFO:
1411 policy = SCHED_FIFO;
1412 break;
1413 case LINUX_SCHED_RR:
1414 policy = SCHED_RR;
1415 break;
1416 default:
1417 return (EINVAL);
1418 }
1419
1420 error = copyin(args->param, &sched_param, sizeof(sched_param));
1421 if (error)
1422 return (error);
1423
1424 if (linux_map_sched_prio) {
1425 switch (policy) {
1426 case SCHED_OTHER:
1427 if (sched_param.sched_priority != 0)
1428 return (EINVAL);
1429
1430 sched_param.sched_priority =
1431 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1432 break;
1433 case SCHED_FIFO:
1434 case SCHED_RR:
1435 if (sched_param.sched_priority < 1 ||
1436 sched_param.sched_priority >= LINUX_MAX_RT_PRIO)
1437 return (EINVAL);
1438
1439 /*
1440 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1441 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1442 */
1443 sched_param.sched_priority =
1444 (sched_param.sched_priority - 1) *
1445 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1446 (LINUX_MAX_RT_PRIO - 1);
1447 break;
1448 }
1449 }
1450
1451 tdt = linux_tdfind(td, args->pid, -1);
1452 if (tdt == NULL)
1453 return (ESRCH);
1454
1455 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1456 PROC_UNLOCK(tdt->td_proc);
1457 return (error);
1458 }
1459
1460 int
1461 linux_sched_getscheduler(struct thread *td,
1462 struct linux_sched_getscheduler_args *args)
1463 {
1464 struct thread *tdt;
1465 int error, policy;
1466
1467 tdt = linux_tdfind(td, args->pid, -1);
1468 if (tdt == NULL)
1469 return (ESRCH);
1470
1471 error = kern_sched_getscheduler(td, tdt, &policy);
1472 PROC_UNLOCK(tdt->td_proc);
1473
1474 switch (policy) {
1475 case SCHED_OTHER:
1476 td->td_retval[0] = LINUX_SCHED_OTHER;
1477 break;
1478 case SCHED_FIFO:
1479 td->td_retval[0] = LINUX_SCHED_FIFO;
1480 break;
1481 case SCHED_RR:
1482 td->td_retval[0] = LINUX_SCHED_RR;
1483 break;
1484 }
1485 return (error);
1486 }
1487
1488 int
1489 linux_sched_get_priority_max(struct thread *td,
1490 struct linux_sched_get_priority_max_args *args)
1491 {
1492 struct sched_get_priority_max_args bsd;
1493
1494 if (linux_map_sched_prio) {
1495 switch (args->policy) {
1496 case LINUX_SCHED_OTHER:
1497 td->td_retval[0] = 0;
1498 return (0);
1499 case LINUX_SCHED_FIFO:
1500 case LINUX_SCHED_RR:
1501 td->td_retval[0] = LINUX_MAX_RT_PRIO - 1;
1502 return (0);
1503 default:
1504 return (EINVAL);
1505 }
1506 }
1507
1508 switch (args->policy) {
1509 case LINUX_SCHED_OTHER:
1510 bsd.policy = SCHED_OTHER;
1511 break;
1512 case LINUX_SCHED_FIFO:
1513 bsd.policy = SCHED_FIFO;
1514 break;
1515 case LINUX_SCHED_RR:
1516 bsd.policy = SCHED_RR;
1517 break;
1518 default:
1519 return (EINVAL);
1520 }
1521 return (sys_sched_get_priority_max(td, &bsd));
1522 }
1523
1524 int
1525 linux_sched_get_priority_min(struct thread *td,
1526 struct linux_sched_get_priority_min_args *args)
1527 {
1528 struct sched_get_priority_min_args bsd;
1529
1530 if (linux_map_sched_prio) {
1531 switch (args->policy) {
1532 case LINUX_SCHED_OTHER:
1533 td->td_retval[0] = 0;
1534 return (0);
1535 case LINUX_SCHED_FIFO:
1536 case LINUX_SCHED_RR:
1537 td->td_retval[0] = 1;
1538 return (0);
1539 default:
1540 return (EINVAL);
1541 }
1542 }
1543
1544 switch (args->policy) {
1545 case LINUX_SCHED_OTHER:
1546 bsd.policy = SCHED_OTHER;
1547 break;
1548 case LINUX_SCHED_FIFO:
1549 bsd.policy = SCHED_FIFO;
1550 break;
1551 case LINUX_SCHED_RR:
1552 bsd.policy = SCHED_RR;
1553 break;
1554 default:
1555 return (EINVAL);
1556 }
1557 return (sys_sched_get_priority_min(td, &bsd));
1558 }
1559
1560 #define REBOOT_CAD_ON 0x89abcdef
1561 #define REBOOT_CAD_OFF 0
1562 #define REBOOT_HALT 0xcdef0123
1563 #define REBOOT_RESTART 0x01234567
1564 #define REBOOT_RESTART2 0xA1B2C3D4
1565 #define REBOOT_POWEROFF 0x4321FEDC
1566 #define REBOOT_MAGIC1 0xfee1dead
1567 #define REBOOT_MAGIC2 0x28121969
1568 #define REBOOT_MAGIC2A 0x05121996
1569 #define REBOOT_MAGIC2B 0x16041998
1570
1571 int
1572 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1573 {
1574 struct reboot_args bsd_args;
1575
1576 if (args->magic1 != REBOOT_MAGIC1)
1577 return (EINVAL);
1578
1579 switch (args->magic2) {
1580 case REBOOT_MAGIC2:
1581 case REBOOT_MAGIC2A:
1582 case REBOOT_MAGIC2B:
1583 break;
1584 default:
1585 return (EINVAL);
1586 }
1587
1588 switch (args->cmd) {
1589 case REBOOT_CAD_ON:
1590 case REBOOT_CAD_OFF:
1591 return (priv_check(td, PRIV_REBOOT));
1592 case REBOOT_HALT:
1593 bsd_args.opt = RB_HALT;
1594 break;
1595 case REBOOT_RESTART:
1596 case REBOOT_RESTART2:
1597 bsd_args.opt = 0;
1598 break;
1599 case REBOOT_POWEROFF:
1600 bsd_args.opt = RB_POWEROFF;
1601 break;
1602 default:
1603 return (EINVAL);
1604 }
1605 return (sys_reboot(td, &bsd_args));
1606 }
1607
1608
1609 int
1610 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1611 {
1612
1613 td->td_retval[0] = td->td_proc->p_pid;
1614
1615 return (0);
1616 }
1617
1618 int
1619 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1620 {
1621 struct linux_emuldata *em;
1622
1623 em = em_find(td);
1624 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1625
1626 td->td_retval[0] = em->em_tid;
1627
1628 return (0);
1629 }
1630
1631
1632 int
1633 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1634 {
1635
1636 td->td_retval[0] = kern_getppid(td);
1637 return (0);
1638 }
1639
1640 int
1641 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1642 {
1643
1644 td->td_retval[0] = td->td_ucred->cr_rgid;
1645 return (0);
1646 }
1647
1648 int
1649 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1650 {
1651
1652 td->td_retval[0] = td->td_ucred->cr_ruid;
1653 return (0);
1654 }
1655
1656
1657 int
1658 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1659 {
1660 struct getsid_args bsd;
1661
1662 bsd.pid = args->pid;
1663 return (sys_getsid(td, &bsd));
1664 }
1665
1666 int
1667 linux_nosys(struct thread *td, struct nosys_args *ignore)
1668 {
1669
1670 return (ENOSYS);
1671 }
1672
1673 int
1674 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1675 {
1676 struct getpriority_args bsd_args;
1677 int error;
1678
1679 bsd_args.which = args->which;
1680 bsd_args.who = args->who;
1681 error = sys_getpriority(td, &bsd_args);
1682 td->td_retval[0] = 20 - td->td_retval[0];
1683 return (error);
1684 }
1685
1686 int
1687 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1688 {
1689 int name[2];
1690
1691 name[0] = CTL_KERN;
1692 name[1] = KERN_HOSTNAME;
1693 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1694 args->len, 0, 0));
1695 }
1696
1697 int
1698 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1699 {
1700 int name[2];
1701
1702 name[0] = CTL_KERN;
1703 name[1] = KERN_NISDOMAINNAME;
1704 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1705 args->len, 0, 0));
1706 }
1707
1708 int
1709 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1710 {
1711
1712 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1713 args->error_code);
1714
1715 /*
1716 * XXX: we should send a signal to the parent if
1717 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1718 * as it doesnt occur often.
1719 */
1720 exit1(td, args->error_code, 0);
1721 /* NOTREACHED */
1722 }
1723
1724 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1725 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1726 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1727
1728 struct l_user_cap_header {
1729 l_int version;
1730 l_int pid;
1731 };
1732
1733 struct l_user_cap_data {
1734 l_int effective;
1735 l_int permitted;
1736 l_int inheritable;
1737 };
1738
1739 int
1740 linux_capget(struct thread *td, struct linux_capget_args *uap)
1741 {
1742 struct l_user_cap_header luch;
1743 struct l_user_cap_data lucd[2];
1744 int error, u32s;
1745
1746 if (uap->hdrp == NULL)
1747 return (EFAULT);
1748
1749 error = copyin(uap->hdrp, &luch, sizeof(luch));
1750 if (error != 0)
1751 return (error);
1752
1753 switch (luch.version) {
1754 case _LINUX_CAPABILITY_VERSION_1:
1755 u32s = 1;
1756 break;
1757 case _LINUX_CAPABILITY_VERSION_2:
1758 case _LINUX_CAPABILITY_VERSION_3:
1759 u32s = 2;
1760 break;
1761 default:
1762 luch.version = _LINUX_CAPABILITY_VERSION_1;
1763 error = copyout(&luch, uap->hdrp, sizeof(luch));
1764 if (error)
1765 return (error);
1766 return (EINVAL);
1767 }
1768
1769 if (luch.pid)
1770 return (EPERM);
1771
1772 if (uap->datap) {
1773 /*
1774 * The current implementation doesn't support setting
1775 * a capability (it's essentially a stub) so indicate
1776 * that no capabilities are currently set or available
1777 * to request.
1778 */
1779 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1780 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1781 }
1782
1783 return (error);
1784 }
1785
1786 int
1787 linux_capset(struct thread *td, struct linux_capset_args *uap)
1788 {
1789 struct l_user_cap_header luch;
1790 struct l_user_cap_data lucd[2];
1791 int error, i, u32s;
1792
1793 if (uap->hdrp == NULL || uap->datap == NULL)
1794 return (EFAULT);
1795
1796 error = copyin(uap->hdrp, &luch, sizeof(luch));
1797 if (error != 0)
1798 return (error);
1799
1800 switch (luch.version) {
1801 case _LINUX_CAPABILITY_VERSION_1:
1802 u32s = 1;
1803 break;
1804 case _LINUX_CAPABILITY_VERSION_2:
1805 case _LINUX_CAPABILITY_VERSION_3:
1806 u32s = 2;
1807 break;
1808 default:
1809 luch.version = _LINUX_CAPABILITY_VERSION_1;
1810 error = copyout(&luch, uap->hdrp, sizeof(luch));
1811 if (error)
1812 return (error);
1813 return (EINVAL);
1814 }
1815
1816 if (luch.pid)
1817 return (EPERM);
1818
1819 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1820 if (error != 0)
1821 return (error);
1822
1823 /* We currently don't support setting any capabilities. */
1824 for (i = 0; i < u32s; i++) {
1825 if (lucd[i].effective || lucd[i].permitted ||
1826 lucd[i].inheritable) {
1827 linux_msg(td,
1828 "capset[%d] effective=0x%x, permitted=0x%x, "
1829 "inheritable=0x%x is not implemented", i,
1830 (int)lucd[i].effective, (int)lucd[i].permitted,
1831 (int)lucd[i].inheritable);
1832 return (EPERM);
1833 }
1834 }
1835
1836 return (0);
1837 }
1838
1839 int
1840 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1841 {
1842 int error = 0, max_size;
1843 struct proc *p = td->td_proc;
1844 char comm[LINUX_MAX_COMM_LEN];
1845 int pdeath_signal;
1846
1847 switch (args->option) {
1848 case LINUX_PR_SET_PDEATHSIG:
1849 if (!LINUX_SIG_VALID(args->arg2))
1850 return (EINVAL);
1851 pdeath_signal = linux_to_bsd_signal(args->arg2);
1852 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
1853 &pdeath_signal));
1854 case LINUX_PR_GET_PDEATHSIG:
1855 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
1856 &pdeath_signal);
1857 if (error != 0)
1858 return (error);
1859 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
1860 return (copyout(&pdeath_signal,
1861 (void *)(register_t)args->arg2,
1862 sizeof(pdeath_signal)));
1863 break;
1864 case LINUX_PR_GET_KEEPCAPS:
1865 /*
1866 * Indicate that we always clear the effective and
1867 * permitted capability sets when the user id becomes
1868 * non-zero (actually the capability sets are simply
1869 * always zero in the current implementation).
1870 */
1871 td->td_retval[0] = 0;
1872 break;
1873 case LINUX_PR_SET_KEEPCAPS:
1874 /*
1875 * Ignore requests to keep the effective and permitted
1876 * capability sets when the user id becomes non-zero.
1877 */
1878 break;
1879 case LINUX_PR_SET_NAME:
1880 /*
1881 * To be on the safe side we need to make sure to not
1882 * overflow the size a Linux program expects. We already
1883 * do this here in the copyin, so that we don't need to
1884 * check on copyout.
1885 */
1886 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1887 error = copyinstr((void *)(register_t)args->arg2, comm,
1888 max_size, NULL);
1889
1890 /* Linux silently truncates the name if it is too long. */
1891 if (error == ENAMETOOLONG) {
1892 /*
1893 * XXX: copyinstr() isn't documented to populate the
1894 * array completely, so do a copyin() to be on the
1895 * safe side. This should be changed in case
1896 * copyinstr() is changed to guarantee this.
1897 */
1898 error = copyin((void *)(register_t)args->arg2, comm,
1899 max_size - 1);
1900 comm[max_size - 1] = '\0';
1901 }
1902 if (error)
1903 return (error);
1904
1905 PROC_LOCK(p);
1906 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1907 PROC_UNLOCK(p);
1908 break;
1909 case LINUX_PR_GET_NAME:
1910 PROC_LOCK(p);
1911 strlcpy(comm, p->p_comm, sizeof(comm));
1912 PROC_UNLOCK(p);
1913 error = copyout(comm, (void *)(register_t)args->arg2,
1914 strlen(comm) + 1);
1915 break;
1916 default:
1917 error = EINVAL;
1918 break;
1919 }
1920
1921 return (error);
1922 }
1923
1924 int
1925 linux_sched_setparam(struct thread *td,
1926 struct linux_sched_setparam_args *uap)
1927 {
1928 struct sched_param sched_param;
1929 struct thread *tdt;
1930 int error, policy;
1931
1932 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1933 if (error)
1934 return (error);
1935
1936 tdt = linux_tdfind(td, uap->pid, -1);
1937 if (tdt == NULL)
1938 return (ESRCH);
1939
1940 if (linux_map_sched_prio) {
1941 error = kern_sched_getscheduler(td, tdt, &policy);
1942 if (error)
1943 goto out;
1944
1945 switch (policy) {
1946 case SCHED_OTHER:
1947 if (sched_param.sched_priority != 0) {
1948 error = EINVAL;
1949 goto out;
1950 }
1951 sched_param.sched_priority =
1952 PRI_MAX_TIMESHARE - PRI_MIN_TIMESHARE;
1953 break;
1954 case SCHED_FIFO:
1955 case SCHED_RR:
1956 if (sched_param.sched_priority < 1 ||
1957 sched_param.sched_priority >= LINUX_MAX_RT_PRIO) {
1958 error = EINVAL;
1959 goto out;
1960 }
1961 /*
1962 * Map [1, LINUX_MAX_RT_PRIO - 1] to
1963 * [0, RTP_PRIO_MAX - RTP_PRIO_MIN] (rounding down).
1964 */
1965 sched_param.sched_priority =
1966 (sched_param.sched_priority - 1) *
1967 (RTP_PRIO_MAX - RTP_PRIO_MIN + 1) /
1968 (LINUX_MAX_RT_PRIO - 1);
1969 break;
1970 }
1971 }
1972
1973 error = kern_sched_setparam(td, tdt, &sched_param);
1974 out: PROC_UNLOCK(tdt->td_proc);
1975 return (error);
1976 }
1977
1978 int
1979 linux_sched_getparam(struct thread *td,
1980 struct linux_sched_getparam_args *uap)
1981 {
1982 struct sched_param sched_param;
1983 struct thread *tdt;
1984 int error, policy;
1985
1986 tdt = linux_tdfind(td, uap->pid, -1);
1987 if (tdt == NULL)
1988 return (ESRCH);
1989
1990 error = kern_sched_getparam(td, tdt, &sched_param);
1991 if (error) {
1992 PROC_UNLOCK(tdt->td_proc);
1993 return (error);
1994 }
1995
1996 if (linux_map_sched_prio) {
1997 error = kern_sched_getscheduler(td, tdt, &policy);
1998 PROC_UNLOCK(tdt->td_proc);
1999 if (error)
2000 return (error);
2001
2002 switch (policy) {
2003 case SCHED_OTHER:
2004 sched_param.sched_priority = 0;
2005 break;
2006 case SCHED_FIFO:
2007 case SCHED_RR:
2008 /*
2009 * Map [0, RTP_PRIO_MAX - RTP_PRIO_MIN] to
2010 * [1, LINUX_MAX_RT_PRIO - 1] (rounding up).
2011 */
2012 sched_param.sched_priority =
2013 (sched_param.sched_priority *
2014 (LINUX_MAX_RT_PRIO - 1) +
2015 (RTP_PRIO_MAX - RTP_PRIO_MIN - 1)) /
2016 (RTP_PRIO_MAX - RTP_PRIO_MIN) + 1;
2017 break;
2018 }
2019 } else
2020 PROC_UNLOCK(tdt->td_proc);
2021
2022 error = copyout(&sched_param, uap->param, sizeof(sched_param));
2023 return (error);
2024 }
2025
2026 static const struct cpuset_copy_cb copy_set = {
2027 .cpuset_copyin = copyin,
2028 .cpuset_copyout = copyout
2029 };
2030
2031 /*
2032 * Get affinity of a process.
2033 */
2034 int
2035 linux_sched_getaffinity(struct thread *td,
2036 struct linux_sched_getaffinity_args *args)
2037 {
2038 int error;
2039 struct thread *tdt;
2040
2041 if (args->len < sizeof(cpuset_t))
2042 return (EINVAL);
2043
2044 tdt = linux_tdfind(td, args->pid, -1);
2045 if (tdt == NULL)
2046 return (ESRCH);
2047
2048 PROC_UNLOCK(tdt->td_proc);
2049
2050 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2051 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr,
2052 ©_set);
2053 if (error == 0)
2054 td->td_retval[0] = sizeof(cpuset_t);
2055
2056 return (error);
2057 }
2058
2059 /*
2060 * Set affinity of a process.
2061 */
2062 int
2063 linux_sched_setaffinity(struct thread *td,
2064 struct linux_sched_setaffinity_args *args)
2065 {
2066 struct thread *tdt;
2067
2068 if (args->len < sizeof(cpuset_t))
2069 return (EINVAL);
2070
2071 tdt = linux_tdfind(td, args->pid, -1);
2072 if (tdt == NULL)
2073 return (ESRCH);
2074
2075 PROC_UNLOCK(tdt->td_proc);
2076
2077 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2078 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr,
2079 ©_set));
2080 }
2081
2082 struct linux_rlimit64 {
2083 uint64_t rlim_cur;
2084 uint64_t rlim_max;
2085 };
2086
2087 int
2088 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2089 {
2090 struct rlimit rlim, nrlim;
2091 struct linux_rlimit64 lrlim;
2092 struct proc *p;
2093 u_int which;
2094 int flags;
2095 int error;
2096
2097 if (args->resource >= LINUX_RLIM_NLIMITS)
2098 return (EINVAL);
2099
2100 which = linux_to_bsd_resource[args->resource];
2101 if (which == -1)
2102 return (EINVAL);
2103
2104 if (args->new != NULL) {
2105 /*
2106 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2107 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2108 * as INFINITY so we do not need a conversion even.
2109 */
2110 error = copyin(args->new, &nrlim, sizeof(nrlim));
2111 if (error != 0)
2112 return (error);
2113 }
2114
2115 flags = PGET_HOLD | PGET_NOTWEXIT;
2116 if (args->new != NULL)
2117 flags |= PGET_CANDEBUG;
2118 else
2119 flags |= PGET_CANSEE;
2120 if (args->pid == 0) {
2121 p = td->td_proc;
2122 PHOLD(p);
2123 } else {
2124 error = pget(args->pid, flags, &p);
2125 if (error != 0)
2126 return (error);
2127 }
2128 if (args->old != NULL) {
2129 PROC_LOCK(p);
2130 lim_rlimit_proc(p, which, &rlim);
2131 PROC_UNLOCK(p);
2132 if (rlim.rlim_cur == RLIM_INFINITY)
2133 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2134 else
2135 lrlim.rlim_cur = rlim.rlim_cur;
2136 if (rlim.rlim_max == RLIM_INFINITY)
2137 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2138 else
2139 lrlim.rlim_max = rlim.rlim_max;
2140 error = copyout(&lrlim, args->old, sizeof(lrlim));
2141 if (error != 0)
2142 goto out;
2143 }
2144
2145 if (args->new != NULL)
2146 error = kern_proc_setrlimit(td, p, which, &nrlim);
2147
2148 out:
2149 PRELE(p);
2150 return (error);
2151 }
2152
2153 int
2154 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2155 {
2156 struct timeval utv, tv0, tv1, *tvp;
2157 struct l_pselect6arg lpse6;
2158 struct l_timespec lts;
2159 struct timespec uts;
2160 l_sigset_t l_ss;
2161 sigset_t *ssp;
2162 sigset_t ss;
2163 int error;
2164
2165 ssp = NULL;
2166 if (args->sig != NULL) {
2167 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2168 if (error != 0)
2169 return (error);
2170 if (lpse6.ss_len != sizeof(l_ss))
2171 return (EINVAL);
2172 if (lpse6.ss != 0) {
2173 error = copyin(PTRIN(lpse6.ss), &l_ss,
2174 sizeof(l_ss));
2175 if (error != 0)
2176 return (error);
2177 linux_to_bsd_sigset(&l_ss, &ss);
2178 ssp = &ss;
2179 }
2180 }
2181
2182 /*
2183 * Currently glibc changes nanosecond number to microsecond.
2184 * This mean losing precision but for now it is hardly seen.
2185 */
2186 if (args->tsp != NULL) {
2187 error = copyin(args->tsp, <s, sizeof(lts));
2188 if (error != 0)
2189 return (error);
2190 error = linux_to_native_timespec(&uts, <s);
2191 if (error != 0)
2192 return (error);
2193
2194 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2195 if (itimerfix(&utv))
2196 return (EINVAL);
2197
2198 microtime(&tv0);
2199 tvp = &utv;
2200 } else
2201 tvp = NULL;
2202
2203 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2204 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2205
2206 if (error == 0 && args->tsp != NULL) {
2207 if (td->td_retval[0] != 0) {
2208 /*
2209 * Compute how much time was left of the timeout,
2210 * by subtracting the current time and the time
2211 * before we started the call, and subtracting
2212 * that result from the user-supplied value.
2213 */
2214
2215 microtime(&tv1);
2216 timevalsub(&tv1, &tv0);
2217 timevalsub(&utv, &tv1);
2218 if (utv.tv_sec < 0)
2219 timevalclear(&utv);
2220 } else
2221 timevalclear(&utv);
2222
2223 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2224
2225 error = native_to_linux_timespec(<s, &uts);
2226 if (error == 0)
2227 error = copyout(<s, args->tsp, sizeof(lts));
2228 }
2229
2230 return (error);
2231 }
2232
2233 int
2234 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2235 {
2236 struct timespec ts0, ts1;
2237 struct l_timespec lts;
2238 struct timespec uts, *tsp;
2239 l_sigset_t l_ss;
2240 sigset_t *ssp;
2241 sigset_t ss;
2242 int error;
2243
2244 if (args->sset != NULL) {
2245 if (args->ssize != sizeof(l_ss))
2246 return (EINVAL);
2247 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2248 if (error)
2249 return (error);
2250 linux_to_bsd_sigset(&l_ss, &ss);
2251 ssp = &ss;
2252 } else
2253 ssp = NULL;
2254 if (args->tsp != NULL) {
2255 error = copyin(args->tsp, <s, sizeof(lts));
2256 if (error)
2257 return (error);
2258 error = linux_to_native_timespec(&uts, <s);
2259 if (error != 0)
2260 return (error);
2261
2262 nanotime(&ts0);
2263 tsp = &uts;
2264 } else
2265 tsp = NULL;
2266
2267 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2268
2269 if (error == 0 && args->tsp != NULL) {
2270 if (td->td_retval[0]) {
2271 nanotime(&ts1);
2272 timespecsub(&ts1, &ts0, &ts1);
2273 timespecsub(&uts, &ts1, &uts);
2274 if (uts.tv_sec < 0)
2275 timespecclear(&uts);
2276 } else
2277 timespecclear(&uts);
2278
2279 error = native_to_linux_timespec(<s, &uts);
2280 if (error == 0)
2281 error = copyout(<s, args->tsp, sizeof(lts));
2282 }
2283
2284 return (error);
2285 }
2286
2287 int
2288 linux_sched_rr_get_interval(struct thread *td,
2289 struct linux_sched_rr_get_interval_args *uap)
2290 {
2291 struct timespec ts;
2292 struct l_timespec lts;
2293 struct thread *tdt;
2294 int error;
2295
2296 /*
2297 * According to man in case the invalid pid specified
2298 * EINVAL should be returned.
2299 */
2300 if (uap->pid < 0)
2301 return (EINVAL);
2302
2303 tdt = linux_tdfind(td, uap->pid, -1);
2304 if (tdt == NULL)
2305 return (ESRCH);
2306
2307 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2308 PROC_UNLOCK(tdt->td_proc);
2309 if (error != 0)
2310 return (error);
2311 error = native_to_linux_timespec(<s, &ts);
2312 if (error != 0)
2313 return (error);
2314 return (copyout(<s, uap->interval, sizeof(lts)));
2315 }
2316
2317 /*
2318 * In case when the Linux thread is the initial thread in
2319 * the thread group thread id is equal to the process id.
2320 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2321 */
2322 struct thread *
2323 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2324 {
2325 struct linux_emuldata *em;
2326 struct thread *tdt;
2327 struct proc *p;
2328
2329 tdt = NULL;
2330 if (tid == 0 || tid == td->td_tid) {
2331 tdt = td;
2332 PROC_LOCK(tdt->td_proc);
2333 } else if (tid > PID_MAX)
2334 tdt = tdfind(tid, pid);
2335 else {
2336 /*
2337 * Initial thread where the tid equal to the pid.
2338 */
2339 p = pfind(tid);
2340 if (p != NULL) {
2341 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2342 /*
2343 * p is not a Linuxulator process.
2344 */
2345 PROC_UNLOCK(p);
2346 return (NULL);
2347 }
2348 FOREACH_THREAD_IN_PROC(p, tdt) {
2349 em = em_find(tdt);
2350 if (tid == em->em_tid)
2351 return (tdt);
2352 }
2353 PROC_UNLOCK(p);
2354 }
2355 return (NULL);
2356 }
2357
2358 return (tdt);
2359 }
2360
2361 void
2362 linux_to_bsd_waitopts(int options, int *bsdopts)
2363 {
2364
2365 if (options & LINUX_WNOHANG)
2366 *bsdopts |= WNOHANG;
2367 if (options & LINUX_WUNTRACED)
2368 *bsdopts |= WUNTRACED;
2369 if (options & LINUX_WEXITED)
2370 *bsdopts |= WEXITED;
2371 if (options & LINUX_WCONTINUED)
2372 *bsdopts |= WCONTINUED;
2373 if (options & LINUX_WNOWAIT)
2374 *bsdopts |= WNOWAIT;
2375
2376 if (options & __WCLONE)
2377 *bsdopts |= WLINUXCLONE;
2378 }
2379
2380 int
2381 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2382 {
2383 struct uio uio;
2384 struct iovec iov;
2385 int error;
2386
2387 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2388 return (EINVAL);
2389 if (args->count > INT_MAX)
2390 args->count = INT_MAX;
2391
2392 iov.iov_base = args->buf;
2393 iov.iov_len = args->count;
2394
2395 uio.uio_iov = &iov;
2396 uio.uio_iovcnt = 1;
2397 uio.uio_resid = iov.iov_len;
2398 uio.uio_segflg = UIO_USERSPACE;
2399 uio.uio_rw = UIO_READ;
2400 uio.uio_td = td;
2401
2402 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2403 if (error == 0)
2404 td->td_retval[0] = args->count - uio.uio_resid;
2405 return (error);
2406 }
2407
2408 int
2409 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2410 {
2411
2412 /* Needs to be page-aligned */
2413 if (args->start & PAGE_MASK)
2414 return (EINVAL);
2415 return (kern_mincore(td, args->start, args->len, args->vec));
2416 }
2417
2418 #define SYSLOG_TAG "<6>"
2419
2420 int
2421 linux_syslog(struct thread *td, struct linux_syslog_args *args)
2422 {
2423 char buf[128], *src, *dst;
2424 u_int seq;
2425 int buflen, error;
2426
2427 if (args->type != LINUX_SYSLOG_ACTION_READ_ALL) {
2428 linux_msg(td, "syslog unsupported type 0x%x", args->type);
2429 return (EINVAL);
2430 }
2431
2432 if (args->len < 6) {
2433 td->td_retval[0] = 0;
2434 return (0);
2435 }
2436
2437 error = priv_check(td, PRIV_MSGBUF);
2438 if (error)
2439 return (error);
2440
2441 mtx_lock(&msgbuf_lock);
2442 msgbuf_peekbytes(msgbufp, NULL, 0, &seq);
2443 mtx_unlock(&msgbuf_lock);
2444
2445 dst = args->buf;
2446 error = copyout(&SYSLOG_TAG, dst, sizeof(SYSLOG_TAG));
2447 /* The -1 is to skip the trailing '\0'. */
2448 dst += sizeof(SYSLOG_TAG) - 1;
2449
2450 while (error == 0) {
2451 mtx_lock(&msgbuf_lock);
2452 buflen = msgbuf_peekbytes(msgbufp, buf, sizeof(buf), &seq);
2453 mtx_unlock(&msgbuf_lock);
2454
2455 if (buflen == 0)
2456 break;
2457
2458 for (src = buf; src < buf + buflen && error == 0; src++) {
2459 if (*src == '\0')
2460 continue;
2461
2462 if (dst >= args->buf + args->len)
2463 goto out;
2464
2465 error = copyout(src, dst, 1);
2466 dst++;
2467
2468 if (*src == '\n' && *(src + 1) != '<' &&
2469 dst + sizeof(SYSLOG_TAG) < args->buf + args->len) {
2470 error = copyout(&SYSLOG_TAG,
2471 dst, sizeof(SYSLOG_TAG));
2472 dst += sizeof(SYSLOG_TAG) - 1;
2473 }
2474 }
2475 }
2476 out:
2477 td->td_retval[0] = dst - args->buf;
2478 return (error);
2479 }
2480
2481 int
2482 linux_getcpu(struct thread *td, struct linux_getcpu_args *args)
2483 {
2484 int cpu, error, node;
2485
2486 cpu = td->td_oncpu; /* Make sure it doesn't change during copyout(9) */
2487 error = 0;
2488 node = cpuid_to_pcpu[cpu]->pc_domain;
2489
2490 if (args->cpu != NULL)
2491 error = copyout(&cpu, args->cpu, sizeof(l_int));
2492 if (args->node != NULL)
2493 error = copyout(&node, args->node, sizeof(l_int));
2494 return (error);
2495 }
Cache object: c46062af77db4d6acb41adf8cefa9b41
|