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/mutex.h>
51 #include <sys/namei.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/procctl.h>
55 #include <sys/reboot.h>
56 #include <sys/racct.h>
57 #include <sys/random.h>
58 #include <sys/resourcevar.h>
59 #include <sys/sched.h>
60 #include <sys/sdt.h>
61 #include <sys/signalvar.h>
62 #include <sys/stat.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
65 #include <sys/sysproto.h>
66 #include <sys/systm.h>
67 #include <sys/time.h>
68 #include <sys/vmmeter.h>
69 #include <sys/vnode.h>
70 #include <sys/wait.h>
71 #include <sys/cpuset.h>
72 #include <sys/uio.h>
73
74 #include <security/mac/mac_framework.h>
75
76 #include <vm/vm.h>
77 #include <vm/pmap.h>
78 #include <vm/vm_kern.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_extern.h>
81 #include <vm/vm_object.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 /**
103 * Special DTrace provider for the linuxulator.
104 *
105 * In this file we define the provider for the entire linuxulator. All
106 * modules (= files of the linuxulator) use it.
107 *
108 * We define a different name depending on the emulated bitsize, see
109 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
110 * native bitsize = linuxulator
111 * amd64, 32bit emulation = linuxulator32
112 */
113 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
114
115 int stclohz; /* Statistics clock frequency */
116
117 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
118 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
119 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
120 RLIMIT_MEMLOCK, RLIMIT_AS
121 };
122
123 struct l_sysinfo {
124 l_long uptime; /* Seconds since boot */
125 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
126 #define LINUX_SYSINFO_LOADS_SCALE 65536
127 l_ulong totalram; /* Total usable main memory size */
128 l_ulong freeram; /* Available memory size */
129 l_ulong sharedram; /* Amount of shared memory */
130 l_ulong bufferram; /* Memory used by buffers */
131 l_ulong totalswap; /* Total swap space size */
132 l_ulong freeswap; /* swap space still available */
133 l_ushort procs; /* Number of current processes */
134 l_ushort pads;
135 l_ulong totalbig;
136 l_ulong freebig;
137 l_uint mem_unit;
138 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
139 };
140
141 struct l_pselect6arg {
142 l_uintptr_t ss;
143 l_size_t ss_len;
144 };
145
146 static int linux_utimensat_nsec_valid(l_long);
147
148
149 int
linux_sysinfo(struct thread * td,struct linux_sysinfo_args * args)150 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
151 {
152 struct l_sysinfo sysinfo;
153 vm_object_t object;
154 int i, j;
155 struct timespec ts;
156
157 bzero(&sysinfo, sizeof(sysinfo));
158 getnanouptime(&ts);
159 if (ts.tv_nsec != 0)
160 ts.tv_sec++;
161 sysinfo.uptime = ts.tv_sec;
162
163 /* Use the information from the mib to get our load averages */
164 for (i = 0; i < 3; i++)
165 sysinfo.loads[i] = averunnable.ldavg[i] *
166 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
167
168 sysinfo.totalram = physmem * PAGE_SIZE;
169 sysinfo.freeram = sysinfo.totalram - vm_wire_count() * PAGE_SIZE;
170
171 sysinfo.sharedram = 0;
172 mtx_lock(&vm_object_list_mtx);
173 TAILQ_FOREACH(object, &vm_object_list, object_list)
174 if (object->shadow_count > 1)
175 sysinfo.sharedram += object->resident_page_count;
176 mtx_unlock(&vm_object_list_mtx);
177
178 sysinfo.sharedram *= PAGE_SIZE;
179 sysinfo.bufferram = 0;
180
181 swap_pager_status(&i, &j);
182 sysinfo.totalswap = i * PAGE_SIZE;
183 sysinfo.freeswap = (i - j) * PAGE_SIZE;
184
185 sysinfo.procs = nprocs;
186
187 /* The following are only present in newer Linux kernels. */
188 sysinfo.totalbig = 0;
189 sysinfo.freebig = 0;
190 sysinfo.mem_unit = 1;
191
192 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
193 }
194
195 #ifdef LINUX_LEGACY_SYSCALLS
196 int
linux_alarm(struct thread * td,struct linux_alarm_args * args)197 linux_alarm(struct thread *td, struct linux_alarm_args *args)
198 {
199 struct itimerval it, old_it;
200 u_int secs;
201 int error;
202
203 #ifdef DEBUG
204 if (ldebug(alarm))
205 printf(ARGS(alarm, "%u"), args->secs);
206 #endif
207 secs = args->secs;
208 /*
209 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
210 * to match kern_setitimer()'s limit to avoid error from it.
211 *
212 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
213 * platforms.
214 */
215 if (secs > INT32_MAX / 2)
216 secs = INT32_MAX / 2;
217
218 it.it_value.tv_sec = secs;
219 it.it_value.tv_usec = 0;
220 timevalclear(&it.it_interval);
221 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
222 KASSERT(error == 0, ("kern_setitimer returns %d", error));
223
224 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
225 old_it.it_value.tv_usec >= 500000)
226 old_it.it_value.tv_sec++;
227 td->td_retval[0] = old_it.it_value.tv_sec;
228 return (0);
229 }
230 #endif
231
232 int
linux_brk(struct thread * td,struct linux_brk_args * args)233 linux_brk(struct thread *td, struct linux_brk_args *args)
234 {
235 struct vmspace *vm = td->td_proc->p_vmspace;
236 uintptr_t new, old;
237
238 #ifdef DEBUG
239 if (ldebug(brk))
240 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
241 #endif
242 old = (uintptr_t)vm->vm_daddr + ctob(vm->vm_dsize);
243 new = (uintptr_t)args->dsend;
244 if ((caddr_t)new > vm->vm_daddr && !kern_break(td, &new))
245 td->td_retval[0] = (register_t)new;
246 else
247 td->td_retval[0] = (register_t)old;
248
249 return (0);
250 }
251
252 #if defined(__i386__)
253 /* XXX: what about amd64/linux32? */
254
255 int
linux_uselib(struct thread * td,struct linux_uselib_args * args)256 linux_uselib(struct thread *td, struct linux_uselib_args *args)
257 {
258 struct nameidata ni;
259 struct vnode *vp;
260 struct exec *a_out;
261 vm_map_t map;
262 vm_map_entry_t entry;
263 struct vattr attr;
264 vm_offset_t vmaddr;
265 unsigned long file_offset;
266 unsigned long bss_size;
267 char *library;
268 ssize_t aresid;
269 int error;
270 bool locked, opened, textset;
271
272 LCONVPATHEXIST(td, args->library, &library);
273
274 #ifdef DEBUG
275 if (ldebug(uselib))
276 printf(ARGS(uselib, "%s"), library);
277 #endif
278
279 a_out = NULL;
280 vp = NULL;
281 locked = false;
282 textset = false;
283 opened = false;
284
285 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
286 UIO_SYSSPACE, library, td);
287 error = namei(&ni);
288 LFREEPATH(library);
289 if (error)
290 goto cleanup;
291
292 vp = ni.ni_vp;
293 NDFREE(&ni, NDF_ONLY_PNBUF);
294
295 /*
296 * From here on down, we have a locked vnode that must be unlocked.
297 * XXX: The code below largely duplicates exec_check_permissions().
298 */
299 locked = true;
300
301 /* Executable? */
302 error = VOP_GETATTR(vp, &attr, td->td_ucred);
303 if (error)
304 goto cleanup;
305
306 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
307 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
308 /* EACCESS is what exec(2) returns. */
309 error = ENOEXEC;
310 goto cleanup;
311 }
312
313 /* Sensible size? */
314 if (attr.va_size == 0) {
315 error = ENOEXEC;
316 goto cleanup;
317 }
318
319 /* Can we access it? */
320 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
321 if (error)
322 goto cleanup;
323
324 /*
325 * XXX: This should use vn_open() so that it is properly authorized,
326 * and to reduce code redundancy all over the place here.
327 * XXX: Not really, it duplicates far more of exec_check_permissions()
328 * than vn_open().
329 */
330 #ifdef MAC
331 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
332 if (error)
333 goto cleanup;
334 #endif
335 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
336 if (error)
337 goto cleanup;
338 opened = true;
339
340 /* Pull in executable header into exec_map */
341 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
342 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
343 if (error)
344 goto cleanup;
345
346 /* Is it a Linux binary ? */
347 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
348 error = ENOEXEC;
349 goto cleanup;
350 }
351
352 /*
353 * While we are here, we should REALLY do some more checks
354 */
355
356 /* Set file/virtual offset based on a.out variant. */
357 switch ((int)(a_out->a_magic & 0xffff)) {
358 case 0413: /* ZMAGIC */
359 file_offset = 1024;
360 break;
361 case 0314: /* QMAGIC */
362 file_offset = 0;
363 break;
364 default:
365 error = ENOEXEC;
366 goto cleanup;
367 }
368
369 bss_size = round_page(a_out->a_bss);
370
371 /* Check various fields in header for validity/bounds. */
372 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
373 error = ENOEXEC;
374 goto cleanup;
375 }
376
377 /* text + data can't exceed file size */
378 if (a_out->a_data + a_out->a_text > attr.va_size) {
379 error = EFAULT;
380 goto cleanup;
381 }
382
383 /*
384 * text/data/bss must not exceed limits
385 * XXX - this is not complete. it should check current usage PLUS
386 * the resources needed by this library.
387 */
388 PROC_LOCK(td->td_proc);
389 if (a_out->a_text > maxtsiz ||
390 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
391 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
392 bss_size) != 0) {
393 PROC_UNLOCK(td->td_proc);
394 error = ENOMEM;
395 goto cleanup;
396 }
397 PROC_UNLOCK(td->td_proc);
398
399 /*
400 * Prevent more writers.
401 */
402 error = VOP_SET_TEXT(vp);
403 if (error != 0)
404 goto cleanup;
405 textset = true;
406
407 /*
408 * Lock no longer needed
409 */
410 locked = false;
411 VOP_UNLOCK(vp, 0);
412
413 /*
414 * Check if file_offset page aligned. Currently we cannot handle
415 * misalinged file offsets, and so we read in the entire image
416 * (what a waste).
417 */
418 if (file_offset & PAGE_MASK) {
419 #ifdef DEBUG
420 printf("uselib: Non page aligned binary %lu\n", file_offset);
421 #endif
422 /* Map text+data read/write/execute */
423
424 /* a_entry is the load address and is page aligned */
425 vmaddr = trunc_page(a_out->a_entry);
426
427 /* get anon user mapping, read+write+execute */
428 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
429 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
430 VM_PROT_ALL, VM_PROT_ALL, 0);
431 if (error)
432 goto cleanup;
433
434 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
435 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
436 td->td_ucred, NOCRED, &aresid, td);
437 if (error != 0)
438 goto cleanup;
439 if (aresid != 0) {
440 error = ENOEXEC;
441 goto cleanup;
442 }
443 } else {
444 #ifdef DEBUG
445 printf("uselib: Page aligned binary %lu\n", file_offset);
446 #endif
447 /*
448 * for QMAGIC, a_entry is 20 bytes beyond the load address
449 * to skip the executable header
450 */
451 vmaddr = trunc_page(a_out->a_entry);
452
453 /*
454 * Map it all into the process's space as a single
455 * copy-on-write "data" segment.
456 */
457 map = &td->td_proc->p_vmspace->vm_map;
458 error = vm_mmap(map, &vmaddr,
459 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
460 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
461 if (error)
462 goto cleanup;
463 vm_map_lock(map);
464 if (!vm_map_lookup_entry(map, vmaddr, &entry)) {
465 vm_map_unlock(map);
466 error = EDOOFUS;
467 goto cleanup;
468 }
469 entry->eflags |= MAP_ENTRY_VN_EXEC;
470 vm_map_unlock(map);
471 textset = false;
472 }
473 #ifdef DEBUG
474 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
475 ((long *)vmaddr)[1]);
476 #endif
477 if (bss_size != 0) {
478 /* Calculate BSS start address */
479 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
480 a_out->a_data;
481
482 /* allocate some 'anon' space */
483 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
484 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
485 VM_PROT_ALL, 0);
486 if (error)
487 goto cleanup;
488 }
489
490 cleanup:
491 if (opened) {
492 if (locked)
493 VOP_UNLOCK(vp, 0);
494 locked = false;
495 VOP_CLOSE(vp, FREAD, td->td_ucred, td);
496 }
497 if (textset) {
498 if (!locked) {
499 locked = true;
500 VOP_LOCK(vp, LK_SHARED | LK_RETRY);
501 }
502 VOP_UNSET_TEXT_CHECKED(vp);
503 }
504 if (locked)
505 VOP_UNLOCK(vp, 0);
506
507 /* Release the temporary mapping. */
508 if (a_out)
509 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
510
511 return (error);
512 }
513
514 #endif /* __i386__ */
515
516 #ifdef LINUX_LEGACY_SYSCALLS
517 int
linux_select(struct thread * td,struct linux_select_args * args)518 linux_select(struct thread *td, struct linux_select_args *args)
519 {
520 l_timeval ltv;
521 struct timeval tv0, tv1, utv, *tvp;
522 int error;
523
524 #ifdef DEBUG
525 if (ldebug(select))
526 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
527 (void *)args->readfds, (void *)args->writefds,
528 (void *)args->exceptfds, (void *)args->timeout);
529 #endif
530
531 /*
532 * Store current time for computation of the amount of
533 * time left.
534 */
535 if (args->timeout) {
536 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
537 goto select_out;
538 utv.tv_sec = ltv.tv_sec;
539 utv.tv_usec = ltv.tv_usec;
540 #ifdef DEBUG
541 if (ldebug(select))
542 printf(LMSG("incoming timeout (%jd/%ld)"),
543 (intmax_t)utv.tv_sec, utv.tv_usec);
544 #endif
545
546 if (itimerfix(&utv)) {
547 /*
548 * The timeval was invalid. Convert it to something
549 * valid that will act as it does under Linux.
550 */
551 utv.tv_sec += utv.tv_usec / 1000000;
552 utv.tv_usec %= 1000000;
553 if (utv.tv_usec < 0) {
554 utv.tv_sec -= 1;
555 utv.tv_usec += 1000000;
556 }
557 if (utv.tv_sec < 0)
558 timevalclear(&utv);
559 }
560 microtime(&tv0);
561 tvp = &utv;
562 } else
563 tvp = NULL;
564
565 error = kern_select(td, args->nfds, args->readfds, args->writefds,
566 args->exceptfds, tvp, LINUX_NFDBITS);
567
568 #ifdef DEBUG
569 if (ldebug(select))
570 printf(LMSG("real select returns %d"), error);
571 #endif
572 if (error)
573 goto select_out;
574
575 if (args->timeout) {
576 if (td->td_retval[0]) {
577 /*
578 * Compute how much time was left of the timeout,
579 * by subtracting the current time and the time
580 * before we started the call, and subtracting
581 * that result from the user-supplied value.
582 */
583 microtime(&tv1);
584 timevalsub(&tv1, &tv0);
585 timevalsub(&utv, &tv1);
586 if (utv.tv_sec < 0)
587 timevalclear(&utv);
588 } else
589 timevalclear(&utv);
590 #ifdef DEBUG
591 if (ldebug(select))
592 printf(LMSG("outgoing timeout (%jd/%ld)"),
593 (intmax_t)utv.tv_sec, utv.tv_usec);
594 #endif
595 ltv.tv_sec = utv.tv_sec;
596 ltv.tv_usec = utv.tv_usec;
597 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
598 goto select_out;
599 }
600
601 select_out:
602 #ifdef DEBUG
603 if (ldebug(select))
604 printf(LMSG("select_out -> %d"), error);
605 #endif
606 return (error);
607 }
608 #endif
609
610 int
linux_mremap(struct thread * td,struct linux_mremap_args * args)611 linux_mremap(struct thread *td, struct linux_mremap_args *args)
612 {
613 uintptr_t addr;
614 size_t len;
615 int error = 0;
616
617 #ifdef DEBUG
618 if (ldebug(mremap))
619 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
620 (void *)(uintptr_t)args->addr,
621 (unsigned long)args->old_len,
622 (unsigned long)args->new_len,
623 (unsigned long)args->flags);
624 #endif
625
626 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
627 td->td_retval[0] = 0;
628 return (EINVAL);
629 }
630
631 /*
632 * Check for the page alignment.
633 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
634 */
635 if (args->addr & PAGE_MASK) {
636 td->td_retval[0] = 0;
637 return (EINVAL);
638 }
639
640 args->new_len = round_page(args->new_len);
641 args->old_len = round_page(args->old_len);
642
643 if (args->new_len > args->old_len) {
644 td->td_retval[0] = 0;
645 return (ENOMEM);
646 }
647
648 if (args->new_len < args->old_len) {
649 addr = args->addr + args->new_len;
650 len = args->old_len - args->new_len;
651 error = kern_munmap(td, addr, len);
652 }
653
654 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
655 return (error);
656 }
657
658 #define LINUX_MS_ASYNC 0x0001
659 #define LINUX_MS_INVALIDATE 0x0002
660 #define LINUX_MS_SYNC 0x0004
661
662 int
linux_msync(struct thread * td,struct linux_msync_args * args)663 linux_msync(struct thread *td, struct linux_msync_args *args)
664 {
665
666 return (kern_msync(td, args->addr, args->len,
667 args->fl & ~LINUX_MS_SYNC));
668 }
669
670 #ifdef LINUX_LEGACY_SYSCALLS
671 int
linux_time(struct thread * td,struct linux_time_args * args)672 linux_time(struct thread *td, struct linux_time_args *args)
673 {
674 struct timeval tv;
675 l_time_t tm;
676 int error;
677
678 #ifdef DEBUG
679 if (ldebug(time))
680 printf(ARGS(time, "*"));
681 #endif
682
683 microtime(&tv);
684 tm = tv.tv_sec;
685 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
686 return (error);
687 td->td_retval[0] = tm;
688 return (0);
689 }
690 #endif
691
692 struct l_times_argv {
693 l_clock_t tms_utime;
694 l_clock_t tms_stime;
695 l_clock_t tms_cutime;
696 l_clock_t tms_cstime;
697 };
698
699
700 /*
701 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
702 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
703 * auxiliary vector entry.
704 */
705 #define CLK_TCK 100
706
707 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
708 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
709
710 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
711 CONVNTCK(r) : CONVOTCK(r))
712
713 int
linux_times(struct thread * td,struct linux_times_args * args)714 linux_times(struct thread *td, struct linux_times_args *args)
715 {
716 struct timeval tv, utime, stime, cutime, cstime;
717 struct l_times_argv tms;
718 struct proc *p;
719 int error;
720
721 #ifdef DEBUG
722 if (ldebug(times))
723 printf(ARGS(times, "*"));
724 #endif
725
726 if (args->buf != NULL) {
727 p = td->td_proc;
728 PROC_LOCK(p);
729 PROC_STATLOCK(p);
730 calcru(p, &utime, &stime);
731 PROC_STATUNLOCK(p);
732 calccru(p, &cutime, &cstime);
733 PROC_UNLOCK(p);
734
735 tms.tms_utime = CONVTCK(utime);
736 tms.tms_stime = CONVTCK(stime);
737
738 tms.tms_cutime = CONVTCK(cutime);
739 tms.tms_cstime = CONVTCK(cstime);
740
741 if ((error = copyout(&tms, args->buf, sizeof(tms))))
742 return (error);
743 }
744
745 microuptime(&tv);
746 td->td_retval[0] = (int)CONVTCK(tv);
747 return (0);
748 }
749
750 int
linux_newuname(struct thread * td,struct linux_newuname_args * args)751 linux_newuname(struct thread *td, struct linux_newuname_args *args)
752 {
753 struct l_new_utsname utsname;
754 char osname[LINUX_MAX_UTSNAME];
755 char osrelease[LINUX_MAX_UTSNAME];
756 char *p;
757
758 #ifdef DEBUG
759 if (ldebug(newuname))
760 printf(ARGS(newuname, "*"));
761 #endif
762
763 linux_get_osname(td, osname);
764 linux_get_osrelease(td, osrelease);
765
766 bzero(&utsname, sizeof(utsname));
767 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
768 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
769 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
770 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
771 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
772 for (p = utsname.version; *p != '\0'; ++p)
773 if (*p == '\n') {
774 *p = '\0';
775 break;
776 }
777 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
778
779 return (copyout(&utsname, args->buf, sizeof(utsname)));
780 }
781
782 struct l_utimbuf {
783 l_time_t l_actime;
784 l_time_t l_modtime;
785 };
786
787 #ifdef LINUX_LEGACY_SYSCALLS
788 int
linux_utime(struct thread * td,struct linux_utime_args * args)789 linux_utime(struct thread *td, struct linux_utime_args *args)
790 {
791 struct timeval tv[2], *tvp;
792 struct l_utimbuf lut;
793 char *fname;
794 int error;
795
796 LCONVPATHEXIST(td, args->fname, &fname);
797
798 #ifdef DEBUG
799 if (ldebug(utime))
800 printf(ARGS(utime, "%s, *"), fname);
801 #endif
802
803 if (args->times) {
804 if ((error = copyin(args->times, &lut, sizeof lut))) {
805 LFREEPATH(fname);
806 return (error);
807 }
808 tv[0].tv_sec = lut.l_actime;
809 tv[0].tv_usec = 0;
810 tv[1].tv_sec = lut.l_modtime;
811 tv[1].tv_usec = 0;
812 tvp = tv;
813 } else
814 tvp = NULL;
815
816 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
817 UIO_SYSSPACE);
818 LFREEPATH(fname);
819 return (error);
820 }
821 #endif
822
823 #ifdef LINUX_LEGACY_SYSCALLS
824 int
linux_utimes(struct thread * td,struct linux_utimes_args * args)825 linux_utimes(struct thread *td, struct linux_utimes_args *args)
826 {
827 l_timeval ltv[2];
828 struct timeval tv[2], *tvp = NULL;
829 char *fname;
830 int error;
831
832 LCONVPATHEXIST(td, args->fname, &fname);
833
834 #ifdef DEBUG
835 if (ldebug(utimes))
836 printf(ARGS(utimes, "%s, *"), fname);
837 #endif
838
839 if (args->tptr != NULL) {
840 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
841 LFREEPATH(fname);
842 return (error);
843 }
844 tv[0].tv_sec = ltv[0].tv_sec;
845 tv[0].tv_usec = ltv[0].tv_usec;
846 tv[1].tv_sec = ltv[1].tv_sec;
847 tv[1].tv_usec = ltv[1].tv_usec;
848 tvp = tv;
849 }
850
851 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
852 tvp, UIO_SYSSPACE);
853 LFREEPATH(fname);
854 return (error);
855 }
856 #endif
857
858 static int
linux_utimensat_nsec_valid(l_long nsec)859 linux_utimensat_nsec_valid(l_long nsec)
860 {
861
862 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
863 return (0);
864 if (nsec >= 0 && nsec <= 999999999)
865 return (0);
866 return (1);
867 }
868
869 int
linux_utimensat(struct thread * td,struct linux_utimensat_args * args)870 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
871 {
872 struct l_timespec l_times[2];
873 struct timespec times[2], *timesp = NULL;
874 char *path = NULL;
875 int error, dfd, flags = 0;
876
877 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
878
879 #ifdef DEBUG
880 if (ldebug(utimensat))
881 printf(ARGS(utimensat, "%d, *"), dfd);
882 #endif
883
884 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
885 return (EINVAL);
886
887 if (args->times != NULL) {
888 error = copyin(args->times, l_times, sizeof(l_times));
889 if (error != 0)
890 return (error);
891
892 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
893 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
894 return (EINVAL);
895
896 times[0].tv_sec = l_times[0].tv_sec;
897 switch (l_times[0].tv_nsec)
898 {
899 case LINUX_UTIME_OMIT:
900 times[0].tv_nsec = UTIME_OMIT;
901 break;
902 case LINUX_UTIME_NOW:
903 times[0].tv_nsec = UTIME_NOW;
904 break;
905 default:
906 times[0].tv_nsec = l_times[0].tv_nsec;
907 }
908
909 times[1].tv_sec = l_times[1].tv_sec;
910 switch (l_times[1].tv_nsec)
911 {
912 case LINUX_UTIME_OMIT:
913 times[1].tv_nsec = UTIME_OMIT;
914 break;
915 case LINUX_UTIME_NOW:
916 times[1].tv_nsec = UTIME_NOW;
917 break;
918 default:
919 times[1].tv_nsec = l_times[1].tv_nsec;
920 break;
921 }
922 timesp = times;
923
924 /* This breaks POSIX, but is what the Linux kernel does
925 * _on purpose_ (documented in the man page for utimensat(2)),
926 * so we must follow that behaviour. */
927 if (times[0].tv_nsec == UTIME_OMIT &&
928 times[1].tv_nsec == UTIME_OMIT)
929 return (0);
930 }
931
932 if (args->pathname != NULL)
933 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
934 else if (args->flags != 0)
935 return (EINVAL);
936
937 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
938 flags |= AT_SYMLINK_NOFOLLOW;
939
940 if (path == NULL)
941 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
942 else {
943 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
944 UIO_SYSSPACE, flags);
945 LFREEPATH(path);
946 }
947
948 return (error);
949 }
950
951 #ifdef LINUX_LEGACY_SYSCALLS
952 int
linux_futimesat(struct thread * td,struct linux_futimesat_args * args)953 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
954 {
955 l_timeval ltv[2];
956 struct timeval tv[2], *tvp = NULL;
957 char *fname;
958 int error, dfd;
959
960 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
961 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
962
963 #ifdef DEBUG
964 if (ldebug(futimesat))
965 printf(ARGS(futimesat, "%s, *"), fname);
966 #endif
967
968 if (args->utimes != NULL) {
969 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
970 LFREEPATH(fname);
971 return (error);
972 }
973 tv[0].tv_sec = ltv[0].tv_sec;
974 tv[0].tv_usec = ltv[0].tv_usec;
975 tv[1].tv_sec = ltv[1].tv_sec;
976 tv[1].tv_usec = ltv[1].tv_usec;
977 tvp = tv;
978 }
979
980 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
981 LFREEPATH(fname);
982 return (error);
983 }
984 #endif
985
986 int
linux_common_wait(struct thread * td,int pid,int * status,int options,struct rusage * ru)987 linux_common_wait(struct thread *td, int pid, int *status,
988 int options, struct rusage *ru)
989 {
990 int error, tmpstat;
991
992 error = kern_wait(td, pid, &tmpstat, options, ru);
993 if (error)
994 return (error);
995
996 if (status) {
997 tmpstat &= 0xffff;
998 if (WIFSIGNALED(tmpstat))
999 tmpstat = (tmpstat & 0xffffff80) |
1000 bsd_to_linux_signal(WTERMSIG(tmpstat));
1001 else if (WIFSTOPPED(tmpstat))
1002 tmpstat = (tmpstat & 0xffff00ff) |
1003 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
1004 else if (WIFCONTINUED(tmpstat))
1005 tmpstat = 0xffff;
1006 error = copyout(&tmpstat, status, sizeof(int));
1007 }
1008
1009 return (error);
1010 }
1011
1012 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1013 int
linux_waitpid(struct thread * td,struct linux_waitpid_args * args)1014 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
1015 {
1016 struct linux_wait4_args wait4_args;
1017
1018 #ifdef DEBUG
1019 if (ldebug(waitpid))
1020 printf(ARGS(waitpid, "%d, %p, %d"),
1021 args->pid, (void *)args->status, args->options);
1022 #endif
1023
1024 wait4_args.pid = args->pid;
1025 wait4_args.status = args->status;
1026 wait4_args.options = args->options;
1027 wait4_args.rusage = NULL;
1028
1029 return (linux_wait4(td, &wait4_args));
1030 }
1031 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1032
1033 int
linux_wait4(struct thread * td,struct linux_wait4_args * args)1034 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1035 {
1036 int error, options;
1037 struct rusage ru, *rup;
1038
1039 #ifdef DEBUG
1040 if (ldebug(wait4))
1041 printf(ARGS(wait4, "%d, %p, %d, %p"),
1042 args->pid, (void *)args->status, args->options,
1043 (void *)args->rusage);
1044 #endif
1045 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1046 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1047 return (EINVAL);
1048
1049 options = WEXITED;
1050 linux_to_bsd_waitopts(args->options, &options);
1051
1052 if (args->rusage != NULL)
1053 rup = &ru;
1054 else
1055 rup = NULL;
1056 error = linux_common_wait(td, args->pid, args->status, options, rup);
1057 if (error != 0)
1058 return (error);
1059 if (args->rusage != NULL)
1060 error = linux_copyout_rusage(&ru, args->rusage);
1061 return (error);
1062 }
1063
1064 int
linux_waitid(struct thread * td,struct linux_waitid_args * args)1065 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1066 {
1067 int status, options, sig;
1068 struct __wrusage wru;
1069 siginfo_t siginfo;
1070 l_siginfo_t lsi;
1071 idtype_t idtype;
1072 struct proc *p;
1073 int error;
1074
1075 options = 0;
1076 linux_to_bsd_waitopts(args->options, &options);
1077
1078 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1079 return (EINVAL);
1080 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1081 return (EINVAL);
1082
1083 switch (args->idtype) {
1084 case LINUX_P_ALL:
1085 idtype = P_ALL;
1086 break;
1087 case LINUX_P_PID:
1088 if (args->id <= 0)
1089 return (EINVAL);
1090 idtype = P_PID;
1091 break;
1092 case LINUX_P_PGID:
1093 if (args->id <= 0)
1094 return (EINVAL);
1095 idtype = P_PGID;
1096 break;
1097 default:
1098 return (EINVAL);
1099 }
1100
1101 error = kern_wait6(td, idtype, args->id, &status, options,
1102 &wru, &siginfo);
1103 if (error != 0)
1104 return (error);
1105 if (args->rusage != NULL) {
1106 error = linux_copyout_rusage(&wru.wru_children,
1107 args->rusage);
1108 if (error != 0)
1109 return (error);
1110 }
1111 if (args->info != NULL) {
1112 p = td->td_proc;
1113 bzero(&lsi, sizeof(lsi));
1114 if (td->td_retval[0] != 0) {
1115 sig = bsd_to_linux_signal(siginfo.si_signo);
1116 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1117 }
1118 error = copyout(&lsi, args->info, sizeof(lsi));
1119 }
1120 td->td_retval[0] = 0;
1121
1122 return (error);
1123 }
1124
1125 #ifdef LINUX_LEGACY_SYSCALLS
1126 int
linux_mknod(struct thread * td,struct linux_mknod_args * args)1127 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1128 {
1129 char *path;
1130 int error;
1131
1132 LCONVPATHCREAT(td, args->path, &path);
1133
1134 #ifdef DEBUG
1135 if (ldebug(mknod))
1136 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1137 (uintmax_t)args->dev);
1138 #endif
1139
1140 switch (args->mode & S_IFMT) {
1141 case S_IFIFO:
1142 case S_IFSOCK:
1143 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1144 args->mode);
1145 break;
1146
1147 case S_IFCHR:
1148 case S_IFBLK:
1149 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1150 args->mode, args->dev);
1151 break;
1152
1153 case S_IFDIR:
1154 error = EPERM;
1155 break;
1156
1157 case 0:
1158 args->mode |= S_IFREG;
1159 /* FALLTHROUGH */
1160 case S_IFREG:
1161 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1162 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1163 if (error == 0)
1164 kern_close(td, td->td_retval[0]);
1165 break;
1166
1167 default:
1168 error = EINVAL;
1169 break;
1170 }
1171 LFREEPATH(path);
1172 return (error);
1173 }
1174 #endif
1175
1176 int
linux_mknodat(struct thread * td,struct linux_mknodat_args * args)1177 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1178 {
1179 char *path;
1180 int error, dfd;
1181
1182 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1183 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1184
1185 #ifdef DEBUG
1186 if (ldebug(mknodat))
1187 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1188 #endif
1189
1190 switch (args->mode & S_IFMT) {
1191 case S_IFIFO:
1192 case S_IFSOCK:
1193 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1194 break;
1195
1196 case S_IFCHR:
1197 case S_IFBLK:
1198 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1199 args->dev);
1200 break;
1201
1202 case S_IFDIR:
1203 error = EPERM;
1204 break;
1205
1206 case 0:
1207 args->mode |= S_IFREG;
1208 /* FALLTHROUGH */
1209 case S_IFREG:
1210 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1211 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1212 if (error == 0)
1213 kern_close(td, td->td_retval[0]);
1214 break;
1215
1216 default:
1217 error = EINVAL;
1218 break;
1219 }
1220 LFREEPATH(path);
1221 return (error);
1222 }
1223
1224 /*
1225 * UGH! This is just about the dumbest idea I've ever heard!!
1226 */
1227 int
linux_personality(struct thread * td,struct linux_personality_args * args)1228 linux_personality(struct thread *td, struct linux_personality_args *args)
1229 {
1230 struct linux_pemuldata *pem;
1231 struct proc *p = td->td_proc;
1232 uint32_t old;
1233
1234 #ifdef DEBUG
1235 if (ldebug(personality))
1236 printf(ARGS(personality, "%u"), args->per);
1237 #endif
1238
1239 PROC_LOCK(p);
1240 pem = pem_find(p);
1241 old = pem->persona;
1242 if (args->per != 0xffffffff)
1243 pem->persona = args->per;
1244 PROC_UNLOCK(p);
1245
1246 td->td_retval[0] = old;
1247 return (0);
1248 }
1249
1250 struct l_itimerval {
1251 l_timeval it_interval;
1252 l_timeval it_value;
1253 };
1254
1255 #define B2L_ITIMERVAL(bip, lip) \
1256 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1257 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1258 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1259 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1260
1261 int
linux_setitimer(struct thread * td,struct linux_setitimer_args * uap)1262 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1263 {
1264 int error;
1265 struct l_itimerval ls;
1266 struct itimerval aitv, oitv;
1267
1268 #ifdef DEBUG
1269 if (ldebug(setitimer))
1270 printf(ARGS(setitimer, "%p, %p"),
1271 (void *)uap->itv, (void *)uap->oitv);
1272 #endif
1273
1274 if (uap->itv == NULL) {
1275 uap->itv = uap->oitv;
1276 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1277 }
1278
1279 error = copyin(uap->itv, &ls, sizeof(ls));
1280 if (error != 0)
1281 return (error);
1282 B2L_ITIMERVAL(&aitv, &ls);
1283 #ifdef DEBUG
1284 if (ldebug(setitimer)) {
1285 printf("setitimer: value: sec: %jd, usec: %ld\n",
1286 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1287 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1288 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1289 }
1290 #endif
1291 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1292 if (error != 0 || uap->oitv == NULL)
1293 return (error);
1294 B2L_ITIMERVAL(&ls, &oitv);
1295
1296 return (copyout(&ls, uap->oitv, sizeof(ls)));
1297 }
1298
1299 int
linux_getitimer(struct thread * td,struct linux_getitimer_args * uap)1300 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1301 {
1302 int error;
1303 struct l_itimerval ls;
1304 struct itimerval aitv;
1305
1306 #ifdef DEBUG
1307 if (ldebug(getitimer))
1308 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1309 #endif
1310 error = kern_getitimer(td, uap->which, &aitv);
1311 if (error != 0)
1312 return (error);
1313 B2L_ITIMERVAL(&ls, &aitv);
1314 return (copyout(&ls, uap->itv, sizeof(ls)));
1315 }
1316
1317 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1318 int
linux_nice(struct thread * td,struct linux_nice_args * args)1319 linux_nice(struct thread *td, struct linux_nice_args *args)
1320 {
1321 struct setpriority_args bsd_args;
1322
1323 bsd_args.which = PRIO_PROCESS;
1324 bsd_args.who = 0; /* current process */
1325 bsd_args.prio = args->inc;
1326 return (sys_setpriority(td, &bsd_args));
1327 }
1328 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1329
1330 int
linux_setgroups(struct thread * td,struct linux_setgroups_args * args)1331 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1332 {
1333 struct ucred *newcred, *oldcred;
1334 l_gid_t *linux_gidset;
1335 gid_t *bsd_gidset;
1336 int ngrp, error;
1337 struct proc *p;
1338
1339 ngrp = args->gidsetsize;
1340 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1341 return (EINVAL);
1342 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1343 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1344 if (error)
1345 goto out;
1346 newcred = crget();
1347 crextend(newcred, ngrp + 1);
1348 p = td->td_proc;
1349 PROC_LOCK(p);
1350 oldcred = p->p_ucred;
1351 crcopy(newcred, oldcred);
1352
1353 /*
1354 * cr_groups[0] holds egid. Setting the whole set from
1355 * the supplied set will cause egid to be changed too.
1356 * Keep cr_groups[0] unchanged to prevent that.
1357 */
1358
1359 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1360 PROC_UNLOCK(p);
1361 crfree(newcred);
1362 goto out;
1363 }
1364
1365 if (ngrp > 0) {
1366 newcred->cr_ngroups = ngrp + 1;
1367
1368 bsd_gidset = newcred->cr_groups;
1369 ngrp--;
1370 while (ngrp >= 0) {
1371 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1372 ngrp--;
1373 }
1374 } else
1375 newcred->cr_ngroups = 1;
1376
1377 setsugid(p);
1378 proc_set_cred(p, newcred);
1379 PROC_UNLOCK(p);
1380 crfree(oldcred);
1381 error = 0;
1382 out:
1383 free(linux_gidset, M_LINUX);
1384 return (error);
1385 }
1386
1387 int
linux_getgroups(struct thread * td,struct linux_getgroups_args * args)1388 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1389 {
1390 struct ucred *cred;
1391 l_gid_t *linux_gidset;
1392 gid_t *bsd_gidset;
1393 int bsd_gidsetsz, ngrp, error;
1394
1395 cred = td->td_ucred;
1396 bsd_gidset = cred->cr_groups;
1397 bsd_gidsetsz = cred->cr_ngroups - 1;
1398
1399 /*
1400 * cr_groups[0] holds egid. Returning the whole set
1401 * here will cause a duplicate. Exclude cr_groups[0]
1402 * to prevent that.
1403 */
1404
1405 if ((ngrp = args->gidsetsize) == 0) {
1406 td->td_retval[0] = bsd_gidsetsz;
1407 return (0);
1408 }
1409
1410 if (ngrp < bsd_gidsetsz)
1411 return (EINVAL);
1412
1413 ngrp = 0;
1414 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1415 M_LINUX, M_WAITOK);
1416 while (ngrp < bsd_gidsetsz) {
1417 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1418 ngrp++;
1419 }
1420
1421 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1422 free(linux_gidset, M_LINUX);
1423 if (error)
1424 return (error);
1425
1426 td->td_retval[0] = ngrp;
1427 return (0);
1428 }
1429
1430 int
linux_setrlimit(struct thread * td,struct linux_setrlimit_args * args)1431 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1432 {
1433 struct rlimit bsd_rlim;
1434 struct l_rlimit rlim;
1435 u_int which;
1436 int error;
1437
1438 #ifdef DEBUG
1439 if (ldebug(setrlimit))
1440 printf(ARGS(setrlimit, "%d, %p"),
1441 args->resource, (void *)args->rlim);
1442 #endif
1443
1444 if (args->resource >= LINUX_RLIM_NLIMITS)
1445 return (EINVAL);
1446
1447 which = linux_to_bsd_resource[args->resource];
1448 if (which == -1)
1449 return (EINVAL);
1450
1451 error = copyin(args->rlim, &rlim, sizeof(rlim));
1452 if (error)
1453 return (error);
1454
1455 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1456 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1457 return (kern_setrlimit(td, which, &bsd_rlim));
1458 }
1459
1460 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1461 int
linux_old_getrlimit(struct thread * td,struct linux_old_getrlimit_args * args)1462 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1463 {
1464 struct l_rlimit rlim;
1465 struct rlimit bsd_rlim;
1466 u_int which;
1467
1468 #ifdef DEBUG
1469 if (ldebug(old_getrlimit))
1470 printf(ARGS(old_getrlimit, "%d, %p"),
1471 args->resource, (void *)args->rlim);
1472 #endif
1473
1474 if (args->resource >= LINUX_RLIM_NLIMITS)
1475 return (EINVAL);
1476
1477 which = linux_to_bsd_resource[args->resource];
1478 if (which == -1)
1479 return (EINVAL);
1480
1481 lim_rlimit(td, which, &bsd_rlim);
1482
1483 #ifdef COMPAT_LINUX32
1484 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1485 if (rlim.rlim_cur == UINT_MAX)
1486 rlim.rlim_cur = INT_MAX;
1487 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1488 if (rlim.rlim_max == UINT_MAX)
1489 rlim.rlim_max = INT_MAX;
1490 #else
1491 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1492 if (rlim.rlim_cur == ULONG_MAX)
1493 rlim.rlim_cur = LONG_MAX;
1494 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1495 if (rlim.rlim_max == ULONG_MAX)
1496 rlim.rlim_max = LONG_MAX;
1497 #endif
1498 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1499 }
1500 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1501
1502 int
linux_getrlimit(struct thread * td,struct linux_getrlimit_args * args)1503 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1504 {
1505 struct l_rlimit rlim;
1506 struct rlimit bsd_rlim;
1507 u_int which;
1508
1509 #ifdef DEBUG
1510 if (ldebug(getrlimit))
1511 printf(ARGS(getrlimit, "%d, %p"),
1512 args->resource, (void *)args->rlim);
1513 #endif
1514
1515 if (args->resource >= LINUX_RLIM_NLIMITS)
1516 return (EINVAL);
1517
1518 which = linux_to_bsd_resource[args->resource];
1519 if (which == -1)
1520 return (EINVAL);
1521
1522 lim_rlimit(td, which, &bsd_rlim);
1523
1524 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1525 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1526 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1527 }
1528
1529 int
linux_sched_setscheduler(struct thread * td,struct linux_sched_setscheduler_args * args)1530 linux_sched_setscheduler(struct thread *td,
1531 struct linux_sched_setscheduler_args *args)
1532 {
1533 struct sched_param sched_param;
1534 struct thread *tdt;
1535 int error, policy;
1536
1537 #ifdef DEBUG
1538 if (ldebug(sched_setscheduler))
1539 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1540 args->pid, args->policy, (const void *)args->param);
1541 #endif
1542
1543 switch (args->policy) {
1544 case LINUX_SCHED_OTHER:
1545 policy = SCHED_OTHER;
1546 break;
1547 case LINUX_SCHED_FIFO:
1548 policy = SCHED_FIFO;
1549 break;
1550 case LINUX_SCHED_RR:
1551 policy = SCHED_RR;
1552 break;
1553 default:
1554 return (EINVAL);
1555 }
1556
1557 error = copyin(args->param, &sched_param, sizeof(sched_param));
1558 if (error)
1559 return (error);
1560
1561 tdt = linux_tdfind(td, args->pid, -1);
1562 if (tdt == NULL)
1563 return (ESRCH);
1564
1565 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1566 PROC_UNLOCK(tdt->td_proc);
1567 return (error);
1568 }
1569
1570 int
linux_sched_getscheduler(struct thread * td,struct linux_sched_getscheduler_args * args)1571 linux_sched_getscheduler(struct thread *td,
1572 struct linux_sched_getscheduler_args *args)
1573 {
1574 struct thread *tdt;
1575 int error, policy;
1576
1577 #ifdef DEBUG
1578 if (ldebug(sched_getscheduler))
1579 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1580 #endif
1581
1582 tdt = linux_tdfind(td, args->pid, -1);
1583 if (tdt == NULL)
1584 return (ESRCH);
1585
1586 error = kern_sched_getscheduler(td, tdt, &policy);
1587 PROC_UNLOCK(tdt->td_proc);
1588
1589 switch (policy) {
1590 case SCHED_OTHER:
1591 td->td_retval[0] = LINUX_SCHED_OTHER;
1592 break;
1593 case SCHED_FIFO:
1594 td->td_retval[0] = LINUX_SCHED_FIFO;
1595 break;
1596 case SCHED_RR:
1597 td->td_retval[0] = LINUX_SCHED_RR;
1598 break;
1599 }
1600 return (error);
1601 }
1602
1603 int
linux_sched_get_priority_max(struct thread * td,struct linux_sched_get_priority_max_args * args)1604 linux_sched_get_priority_max(struct thread *td,
1605 struct linux_sched_get_priority_max_args *args)
1606 {
1607 struct sched_get_priority_max_args bsd;
1608
1609 #ifdef DEBUG
1610 if (ldebug(sched_get_priority_max))
1611 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1612 #endif
1613
1614 switch (args->policy) {
1615 case LINUX_SCHED_OTHER:
1616 bsd.policy = SCHED_OTHER;
1617 break;
1618 case LINUX_SCHED_FIFO:
1619 bsd.policy = SCHED_FIFO;
1620 break;
1621 case LINUX_SCHED_RR:
1622 bsd.policy = SCHED_RR;
1623 break;
1624 default:
1625 return (EINVAL);
1626 }
1627 return (sys_sched_get_priority_max(td, &bsd));
1628 }
1629
1630 int
linux_sched_get_priority_min(struct thread * td,struct linux_sched_get_priority_min_args * args)1631 linux_sched_get_priority_min(struct thread *td,
1632 struct linux_sched_get_priority_min_args *args)
1633 {
1634 struct sched_get_priority_min_args bsd;
1635
1636 #ifdef DEBUG
1637 if (ldebug(sched_get_priority_min))
1638 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1639 #endif
1640
1641 switch (args->policy) {
1642 case LINUX_SCHED_OTHER:
1643 bsd.policy = SCHED_OTHER;
1644 break;
1645 case LINUX_SCHED_FIFO:
1646 bsd.policy = SCHED_FIFO;
1647 break;
1648 case LINUX_SCHED_RR:
1649 bsd.policy = SCHED_RR;
1650 break;
1651 default:
1652 return (EINVAL);
1653 }
1654 return (sys_sched_get_priority_min(td, &bsd));
1655 }
1656
1657 #define REBOOT_CAD_ON 0x89abcdef
1658 #define REBOOT_CAD_OFF 0
1659 #define REBOOT_HALT 0xcdef0123
1660 #define REBOOT_RESTART 0x01234567
1661 #define REBOOT_RESTART2 0xA1B2C3D4
1662 #define REBOOT_POWEROFF 0x4321FEDC
1663 #define REBOOT_MAGIC1 0xfee1dead
1664 #define REBOOT_MAGIC2 0x28121969
1665 #define REBOOT_MAGIC2A 0x05121996
1666 #define REBOOT_MAGIC2B 0x16041998
1667
1668 int
linux_reboot(struct thread * td,struct linux_reboot_args * args)1669 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1670 {
1671 struct reboot_args bsd_args;
1672
1673 #ifdef DEBUG
1674 if (ldebug(reboot))
1675 printf(ARGS(reboot, "0x%x"), args->cmd);
1676 #endif
1677
1678 if (args->magic1 != REBOOT_MAGIC1)
1679 return (EINVAL);
1680
1681 switch (args->magic2) {
1682 case REBOOT_MAGIC2:
1683 case REBOOT_MAGIC2A:
1684 case REBOOT_MAGIC2B:
1685 break;
1686 default:
1687 return (EINVAL);
1688 }
1689
1690 switch (args->cmd) {
1691 case REBOOT_CAD_ON:
1692 case REBOOT_CAD_OFF:
1693 return (priv_check(td, PRIV_REBOOT));
1694 case REBOOT_HALT:
1695 bsd_args.opt = RB_HALT;
1696 break;
1697 case REBOOT_RESTART:
1698 case REBOOT_RESTART2:
1699 bsd_args.opt = 0;
1700 break;
1701 case REBOOT_POWEROFF:
1702 bsd_args.opt = RB_POWEROFF;
1703 break;
1704 default:
1705 return (EINVAL);
1706 }
1707 return (sys_reboot(td, &bsd_args));
1708 }
1709
1710
1711 int
linux_getpid(struct thread * td,struct linux_getpid_args * args)1712 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1713 {
1714
1715 #ifdef DEBUG
1716 if (ldebug(getpid))
1717 printf(ARGS(getpid, ""));
1718 #endif
1719 td->td_retval[0] = td->td_proc->p_pid;
1720
1721 return (0);
1722 }
1723
1724 int
linux_gettid(struct thread * td,struct linux_gettid_args * args)1725 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1726 {
1727 struct linux_emuldata *em;
1728
1729 #ifdef DEBUG
1730 if (ldebug(gettid))
1731 printf(ARGS(gettid, ""));
1732 #endif
1733
1734 em = em_find(td);
1735 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1736
1737 td->td_retval[0] = em->em_tid;
1738
1739 return (0);
1740 }
1741
1742
1743 int
linux_getppid(struct thread * td,struct linux_getppid_args * args)1744 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1745 {
1746
1747 #ifdef DEBUG
1748 if (ldebug(getppid))
1749 printf(ARGS(getppid, ""));
1750 #endif
1751
1752 td->td_retval[0] = kern_getppid(td);
1753 return (0);
1754 }
1755
1756 int
linux_getgid(struct thread * td,struct linux_getgid_args * args)1757 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1758 {
1759
1760 #ifdef DEBUG
1761 if (ldebug(getgid))
1762 printf(ARGS(getgid, ""));
1763 #endif
1764
1765 td->td_retval[0] = td->td_ucred->cr_rgid;
1766 return (0);
1767 }
1768
1769 int
linux_getuid(struct thread * td,struct linux_getuid_args * args)1770 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1771 {
1772
1773 #ifdef DEBUG
1774 if (ldebug(getuid))
1775 printf(ARGS(getuid, ""));
1776 #endif
1777
1778 td->td_retval[0] = td->td_ucred->cr_ruid;
1779 return (0);
1780 }
1781
1782
1783 int
linux_getsid(struct thread * td,struct linux_getsid_args * args)1784 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1785 {
1786 struct getsid_args bsd;
1787
1788 #ifdef DEBUG
1789 if (ldebug(getsid))
1790 printf(ARGS(getsid, "%i"), args->pid);
1791 #endif
1792
1793 bsd.pid = args->pid;
1794 return (sys_getsid(td, &bsd));
1795 }
1796
1797 int
linux_nosys(struct thread * td,struct nosys_args * ignore)1798 linux_nosys(struct thread *td, struct nosys_args *ignore)
1799 {
1800
1801 return (ENOSYS);
1802 }
1803
1804 int
linux_getpriority(struct thread * td,struct linux_getpriority_args * args)1805 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1806 {
1807 struct getpriority_args bsd_args;
1808 int error;
1809
1810 #ifdef DEBUG
1811 if (ldebug(getpriority))
1812 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1813 #endif
1814
1815 bsd_args.which = args->which;
1816 bsd_args.who = args->who;
1817 error = sys_getpriority(td, &bsd_args);
1818 td->td_retval[0] = 20 - td->td_retval[0];
1819 return (error);
1820 }
1821
1822 int
linux_sethostname(struct thread * td,struct linux_sethostname_args * args)1823 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1824 {
1825 int name[2];
1826
1827 #ifdef DEBUG
1828 if (ldebug(sethostname))
1829 printf(ARGS(sethostname, "*, %i"), args->len);
1830 #endif
1831
1832 name[0] = CTL_KERN;
1833 name[1] = KERN_HOSTNAME;
1834 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1835 args->len, 0, 0));
1836 }
1837
1838 int
linux_setdomainname(struct thread * td,struct linux_setdomainname_args * args)1839 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1840 {
1841 int name[2];
1842
1843 #ifdef DEBUG
1844 if (ldebug(setdomainname))
1845 printf(ARGS(setdomainname, "*, %i"), args->len);
1846 #endif
1847
1848 name[0] = CTL_KERN;
1849 name[1] = KERN_NISDOMAINNAME;
1850 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1851 args->len, 0, 0));
1852 }
1853
1854 int
linux_exit_group(struct thread * td,struct linux_exit_group_args * args)1855 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1856 {
1857
1858 #ifdef DEBUG
1859 if (ldebug(exit_group))
1860 printf(ARGS(exit_group, "%i"), args->error_code);
1861 #endif
1862
1863 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1864 args->error_code);
1865
1866 /*
1867 * XXX: we should send a signal to the parent if
1868 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1869 * as it doesnt occur often.
1870 */
1871 exit1(td, args->error_code, 0);
1872 /* NOTREACHED */
1873 }
1874
1875 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1876 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1877 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1878
1879 struct l_user_cap_header {
1880 l_int version;
1881 l_int pid;
1882 };
1883
1884 struct l_user_cap_data {
1885 l_int effective;
1886 l_int permitted;
1887 l_int inheritable;
1888 };
1889
1890 int
linux_capget(struct thread * td,struct linux_capget_args * uap)1891 linux_capget(struct thread *td, struct linux_capget_args *uap)
1892 {
1893 struct l_user_cap_header luch;
1894 struct l_user_cap_data lucd[2];
1895 int error, u32s;
1896
1897 if (uap->hdrp == NULL)
1898 return (EFAULT);
1899
1900 error = copyin(uap->hdrp, &luch, sizeof(luch));
1901 if (error != 0)
1902 return (error);
1903
1904 switch (luch.version) {
1905 case _LINUX_CAPABILITY_VERSION_1:
1906 u32s = 1;
1907 break;
1908 case _LINUX_CAPABILITY_VERSION_2:
1909 case _LINUX_CAPABILITY_VERSION_3:
1910 u32s = 2;
1911 break;
1912 default:
1913 #ifdef DEBUG
1914 if (ldebug(capget))
1915 printf(LMSG("invalid capget capability version 0x%x"),
1916 luch.version);
1917 #endif
1918 luch.version = _LINUX_CAPABILITY_VERSION_1;
1919 error = copyout(&luch, uap->hdrp, sizeof(luch));
1920 if (error)
1921 return (error);
1922 return (EINVAL);
1923 }
1924
1925 if (luch.pid)
1926 return (EPERM);
1927
1928 if (uap->datap) {
1929 /*
1930 * The current implementation doesn't support setting
1931 * a capability (it's essentially a stub) so indicate
1932 * that no capabilities are currently set or available
1933 * to request.
1934 */
1935 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1936 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1937 }
1938
1939 return (error);
1940 }
1941
1942 int
linux_capset(struct thread * td,struct linux_capset_args * uap)1943 linux_capset(struct thread *td, struct linux_capset_args *uap)
1944 {
1945 struct l_user_cap_header luch;
1946 struct l_user_cap_data lucd[2];
1947 int error, i, u32s;
1948
1949 if (uap->hdrp == NULL || uap->datap == NULL)
1950 return (EFAULT);
1951
1952 error = copyin(uap->hdrp, &luch, sizeof(luch));
1953 if (error != 0)
1954 return (error);
1955
1956 switch (luch.version) {
1957 case _LINUX_CAPABILITY_VERSION_1:
1958 u32s = 1;
1959 break;
1960 case _LINUX_CAPABILITY_VERSION_2:
1961 case _LINUX_CAPABILITY_VERSION_3:
1962 u32s = 2;
1963 break;
1964 default:
1965 #ifdef DEBUG
1966 if (ldebug(capset))
1967 printf(LMSG("invalid capset capability version 0x%x"),
1968 luch.version);
1969 #endif
1970 luch.version = _LINUX_CAPABILITY_VERSION_1;
1971 error = copyout(&luch, uap->hdrp, sizeof(luch));
1972 if (error)
1973 return (error);
1974 return (EINVAL);
1975 }
1976
1977 if (luch.pid)
1978 return (EPERM);
1979
1980 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1981 if (error != 0)
1982 return (error);
1983
1984 /* We currently don't support setting any capabilities. */
1985 for (i = 0; i < u32s; i++) {
1986 if (lucd[i].effective || lucd[i].permitted ||
1987 lucd[i].inheritable) {
1988 linux_msg(td,
1989 "capset[%d] effective=0x%x, permitted=0x%x, "
1990 "inheritable=0x%x is not implemented", i,
1991 (int)lucd[i].effective, (int)lucd[i].permitted,
1992 (int)lucd[i].inheritable);
1993 return (EPERM);
1994 }
1995 }
1996
1997 return (0);
1998 }
1999
2000 int
linux_prctl(struct thread * td,struct linux_prctl_args * args)2001 linux_prctl(struct thread *td, struct linux_prctl_args *args)
2002 {
2003 int error = 0, max_size;
2004 struct proc *p = td->td_proc;
2005 char comm[LINUX_MAX_COMM_LEN];
2006 int pdeath_signal;
2007
2008 #ifdef DEBUG
2009 if (ldebug(prctl))
2010 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
2011 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
2012 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
2013 #endif
2014
2015 switch (args->option) {
2016 case LINUX_PR_SET_PDEATHSIG:
2017 if (!LINUX_SIG_VALID(args->arg2))
2018 return (EINVAL);
2019 pdeath_signal = linux_to_bsd_signal(args->arg2);
2020 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
2021 &pdeath_signal));
2022 case LINUX_PR_GET_PDEATHSIG:
2023 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
2024 &pdeath_signal);
2025 if (error != 0)
2026 return (error);
2027 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
2028 return (copyout(&pdeath_signal,
2029 (void *)(register_t)args->arg2,
2030 sizeof(pdeath_signal)));
2031 break;
2032 case LINUX_PR_GET_KEEPCAPS:
2033 /*
2034 * Indicate that we always clear the effective and
2035 * permitted capability sets when the user id becomes
2036 * non-zero (actually the capability sets are simply
2037 * always zero in the current implementation).
2038 */
2039 td->td_retval[0] = 0;
2040 break;
2041 case LINUX_PR_SET_KEEPCAPS:
2042 /*
2043 * Ignore requests to keep the effective and permitted
2044 * capability sets when the user id becomes non-zero.
2045 */
2046 break;
2047 case LINUX_PR_SET_NAME:
2048 /*
2049 * To be on the safe side we need to make sure to not
2050 * overflow the size a Linux program expects. We already
2051 * do this here in the copyin, so that we don't need to
2052 * check on copyout.
2053 */
2054 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2055 error = copyinstr((void *)(register_t)args->arg2, comm,
2056 max_size, NULL);
2057
2058 /* Linux silently truncates the name if it is too long. */
2059 if (error == ENAMETOOLONG) {
2060 /*
2061 * XXX: copyinstr() isn't documented to populate the
2062 * array completely, so do a copyin() to be on the
2063 * safe side. This should be changed in case
2064 * copyinstr() is changed to guarantee this.
2065 */
2066 error = copyin((void *)(register_t)args->arg2, comm,
2067 max_size - 1);
2068 comm[max_size - 1] = '\0';
2069 }
2070 if (error)
2071 return (error);
2072
2073 PROC_LOCK(p);
2074 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2075 PROC_UNLOCK(p);
2076 break;
2077 case LINUX_PR_GET_NAME:
2078 PROC_LOCK(p);
2079 strlcpy(comm, p->p_comm, sizeof(comm));
2080 PROC_UNLOCK(p);
2081 error = copyout(comm, (void *)(register_t)args->arg2,
2082 strlen(comm) + 1);
2083 break;
2084 default:
2085 error = EINVAL;
2086 break;
2087 }
2088
2089 return (error);
2090 }
2091
2092 int
linux_sched_setparam(struct thread * td,struct linux_sched_setparam_args * uap)2093 linux_sched_setparam(struct thread *td,
2094 struct linux_sched_setparam_args *uap)
2095 {
2096 struct sched_param sched_param;
2097 struct thread *tdt;
2098 int error;
2099
2100 #ifdef DEBUG
2101 if (ldebug(sched_setparam))
2102 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2103 #endif
2104
2105 error = copyin(uap->param, &sched_param, sizeof(sched_param));
2106 if (error)
2107 return (error);
2108
2109 tdt = linux_tdfind(td, uap->pid, -1);
2110 if (tdt == NULL)
2111 return (ESRCH);
2112
2113 error = kern_sched_setparam(td, tdt, &sched_param);
2114 PROC_UNLOCK(tdt->td_proc);
2115 return (error);
2116 }
2117
2118 int
linux_sched_getparam(struct thread * td,struct linux_sched_getparam_args * uap)2119 linux_sched_getparam(struct thread *td,
2120 struct linux_sched_getparam_args *uap)
2121 {
2122 struct sched_param sched_param;
2123 struct thread *tdt;
2124 int error;
2125
2126 #ifdef DEBUG
2127 if (ldebug(sched_getparam))
2128 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2129 #endif
2130
2131 tdt = linux_tdfind(td, uap->pid, -1);
2132 if (tdt == NULL)
2133 return (ESRCH);
2134
2135 error = kern_sched_getparam(td, tdt, &sched_param);
2136 PROC_UNLOCK(tdt->td_proc);
2137 if (error == 0)
2138 error = copyout(&sched_param, uap->param,
2139 sizeof(sched_param));
2140 return (error);
2141 }
2142
2143 /*
2144 * Get affinity of a process.
2145 */
2146 int
linux_sched_getaffinity(struct thread * td,struct linux_sched_getaffinity_args * args)2147 linux_sched_getaffinity(struct thread *td,
2148 struct linux_sched_getaffinity_args *args)
2149 {
2150 int error;
2151 struct thread *tdt;
2152
2153 #ifdef DEBUG
2154 if (ldebug(sched_getaffinity))
2155 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2156 args->len);
2157 #endif
2158 if (args->len < sizeof(cpuset_t))
2159 return (EINVAL);
2160
2161 tdt = linux_tdfind(td, args->pid, -1);
2162 if (tdt == NULL)
2163 return (ESRCH);
2164
2165 PROC_UNLOCK(tdt->td_proc);
2166
2167 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2168 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2169 if (error == 0)
2170 td->td_retval[0] = sizeof(cpuset_t);
2171
2172 return (error);
2173 }
2174
2175 /*
2176 * Set affinity of a process.
2177 */
2178 int
linux_sched_setaffinity(struct thread * td,struct linux_sched_setaffinity_args * args)2179 linux_sched_setaffinity(struct thread *td,
2180 struct linux_sched_setaffinity_args *args)
2181 {
2182 struct thread *tdt;
2183
2184 #ifdef DEBUG
2185 if (ldebug(sched_setaffinity))
2186 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2187 args->len);
2188 #endif
2189 if (args->len < sizeof(cpuset_t))
2190 return (EINVAL);
2191
2192 tdt = linux_tdfind(td, args->pid, -1);
2193 if (tdt == NULL)
2194 return (ESRCH);
2195
2196 PROC_UNLOCK(tdt->td_proc);
2197
2198 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2199 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2200 }
2201
2202 struct linux_rlimit64 {
2203 uint64_t rlim_cur;
2204 uint64_t rlim_max;
2205 };
2206
2207 int
linux_prlimit64(struct thread * td,struct linux_prlimit64_args * args)2208 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2209 {
2210 struct rlimit rlim, nrlim;
2211 struct linux_rlimit64 lrlim;
2212 struct proc *p;
2213 u_int which;
2214 int flags;
2215 int error;
2216
2217 #ifdef DEBUG
2218 if (ldebug(prlimit64))
2219 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2220 args->resource, (void *)args->new, (void *)args->old);
2221 #endif
2222
2223 if (args->resource >= LINUX_RLIM_NLIMITS)
2224 return (EINVAL);
2225
2226 which = linux_to_bsd_resource[args->resource];
2227 if (which == -1)
2228 return (EINVAL);
2229
2230 if (args->new != NULL) {
2231 /*
2232 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2233 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2234 * as INFINITY so we do not need a conversion even.
2235 */
2236 error = copyin(args->new, &nrlim, sizeof(nrlim));
2237 if (error != 0)
2238 return (error);
2239 }
2240
2241 flags = PGET_HOLD | PGET_NOTWEXIT;
2242 if (args->new != NULL)
2243 flags |= PGET_CANDEBUG;
2244 else
2245 flags |= PGET_CANSEE;
2246 error = pget(args->pid, flags, &p);
2247 if (error != 0)
2248 return (error);
2249
2250 if (args->old != NULL) {
2251 PROC_LOCK(p);
2252 lim_rlimit_proc(p, which, &rlim);
2253 PROC_UNLOCK(p);
2254 if (rlim.rlim_cur == RLIM_INFINITY)
2255 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2256 else
2257 lrlim.rlim_cur = rlim.rlim_cur;
2258 if (rlim.rlim_max == RLIM_INFINITY)
2259 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2260 else
2261 lrlim.rlim_max = rlim.rlim_max;
2262 error = copyout(&lrlim, args->old, sizeof(lrlim));
2263 if (error != 0)
2264 goto out;
2265 }
2266
2267 if (args->new != NULL)
2268 error = kern_proc_setrlimit(td, p, which, &nrlim);
2269
2270 out:
2271 PRELE(p);
2272 return (error);
2273 }
2274
2275 int
linux_pselect6(struct thread * td,struct linux_pselect6_args * args)2276 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2277 {
2278 struct timeval utv, tv0, tv1, *tvp;
2279 struct l_pselect6arg lpse6;
2280 struct l_timespec lts;
2281 struct timespec uts;
2282 l_sigset_t l_ss;
2283 sigset_t *ssp;
2284 sigset_t ss;
2285 int error;
2286
2287 ssp = NULL;
2288 if (args->sig != NULL) {
2289 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2290 if (error != 0)
2291 return (error);
2292 if (lpse6.ss_len != sizeof(l_ss))
2293 return (EINVAL);
2294 if (lpse6.ss != 0) {
2295 error = copyin(PTRIN(lpse6.ss), &l_ss,
2296 sizeof(l_ss));
2297 if (error != 0)
2298 return (error);
2299 linux_to_bsd_sigset(&l_ss, &ss);
2300 ssp = &ss;
2301 }
2302 }
2303
2304 /*
2305 * Currently glibc changes nanosecond number to microsecond.
2306 * This mean losing precision but for now it is hardly seen.
2307 */
2308 if (args->tsp != NULL) {
2309 error = copyin(args->tsp, <s, sizeof(lts));
2310 if (error != 0)
2311 return (error);
2312 error = linux_to_native_timespec(&uts, <s);
2313 if (error != 0)
2314 return (error);
2315
2316 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2317 if (itimerfix(&utv))
2318 return (EINVAL);
2319
2320 microtime(&tv0);
2321 tvp = &utv;
2322 } else
2323 tvp = NULL;
2324
2325 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2326 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2327
2328 if (error == 0 && args->tsp != NULL) {
2329 if (td->td_retval[0] != 0) {
2330 /*
2331 * Compute how much time was left of the timeout,
2332 * by subtracting the current time and the time
2333 * before we started the call, and subtracting
2334 * that result from the user-supplied value.
2335 */
2336
2337 microtime(&tv1);
2338 timevalsub(&tv1, &tv0);
2339 timevalsub(&utv, &tv1);
2340 if (utv.tv_sec < 0)
2341 timevalclear(&utv);
2342 } else
2343 timevalclear(&utv);
2344
2345 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2346
2347 error = native_to_linux_timespec(<s, &uts);
2348 if (error == 0)
2349 error = copyout(<s, args->tsp, sizeof(lts));
2350 }
2351
2352 return (error);
2353 }
2354
2355 int
linux_ppoll(struct thread * td,struct linux_ppoll_args * args)2356 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2357 {
2358 struct timespec ts0, ts1;
2359 struct l_timespec lts;
2360 struct timespec uts, *tsp;
2361 l_sigset_t l_ss;
2362 sigset_t *ssp;
2363 sigset_t ss;
2364 int error;
2365
2366 if (args->sset != NULL) {
2367 if (args->ssize != sizeof(l_ss))
2368 return (EINVAL);
2369 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2370 if (error)
2371 return (error);
2372 linux_to_bsd_sigset(&l_ss, &ss);
2373 ssp = &ss;
2374 } else
2375 ssp = NULL;
2376 if (args->tsp != NULL) {
2377 error = copyin(args->tsp, <s, sizeof(lts));
2378 if (error)
2379 return (error);
2380 error = linux_to_native_timespec(&uts, <s);
2381 if (error != 0)
2382 return (error);
2383
2384 nanotime(&ts0);
2385 tsp = &uts;
2386 } else
2387 tsp = NULL;
2388
2389 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2390
2391 if (error == 0 && args->tsp != NULL) {
2392 if (td->td_retval[0]) {
2393 nanotime(&ts1);
2394 timespecsub(&ts1, &ts0, &ts1);
2395 timespecsub(&uts, &ts1, &uts);
2396 if (uts.tv_sec < 0)
2397 timespecclear(&uts);
2398 } else
2399 timespecclear(&uts);
2400
2401 error = native_to_linux_timespec(<s, &uts);
2402 if (error == 0)
2403 error = copyout(<s, args->tsp, sizeof(lts));
2404 }
2405
2406 return (error);
2407 }
2408
2409 #if defined(DEBUG) || defined(KTR)
2410 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2411
2412 #ifdef COMPAT_LINUX32
2413 #define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL
2414 #else
2415 #define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL
2416 #endif
2417
2418 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2419
2420 static int
linux_debug(int syscall,int toggle,int global)2421 linux_debug(int syscall, int toggle, int global)
2422 {
2423
2424 if (global) {
2425 char c = toggle ? 0 : 0xff;
2426
2427 memset(linux_debug_map, c, sizeof(linux_debug_map));
2428 return (0);
2429 }
2430 if (syscall < 0 || syscall >= L_MAXSYSCALL)
2431 return (EINVAL);
2432 if (toggle)
2433 clrbit(linux_debug_map, syscall);
2434 else
2435 setbit(linux_debug_map, syscall);
2436 return (0);
2437 }
2438 #undef L_MAXSYSCALL
2439
2440 /*
2441 * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2442 *
2443 * E.g.: sysctl linux.debug=21.0
2444 *
2445 * As a special case, syscall "all" will apply to all syscalls globally.
2446 */
2447 #define LINUX_MAX_DEBUGSTR 16
2448 int
linux_sysctl_debug(SYSCTL_HANDLER_ARGS)2449 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2450 {
2451 char value[LINUX_MAX_DEBUGSTR], *p;
2452 int error, sysc, toggle;
2453 int global = 0;
2454
2455 value[0] = '\0';
2456 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2457 if (error || req->newptr == NULL)
2458 return (error);
2459 for (p = value; *p != '\0' && *p != '.'; p++);
2460 if (*p == '\0')
2461 return (EINVAL);
2462 *p++ = '\0';
2463 sysc = strtol(value, NULL, 0);
2464 toggle = strtol(p, NULL, 0);
2465 if (strcmp(value, "all") == 0)
2466 global = 1;
2467 error = linux_debug(sysc, toggle, global);
2468 return (error);
2469 }
2470
2471 #endif /* DEBUG || KTR */
2472
2473 int
linux_sched_rr_get_interval(struct thread * td,struct linux_sched_rr_get_interval_args * uap)2474 linux_sched_rr_get_interval(struct thread *td,
2475 struct linux_sched_rr_get_interval_args *uap)
2476 {
2477 struct timespec ts;
2478 struct l_timespec lts;
2479 struct thread *tdt;
2480 int error;
2481
2482 /*
2483 * According to man in case the invalid pid specified
2484 * EINVAL should be returned.
2485 */
2486 if (uap->pid < 0)
2487 return (EINVAL);
2488
2489 tdt = linux_tdfind(td, uap->pid, -1);
2490 if (tdt == NULL)
2491 return (ESRCH);
2492
2493 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2494 PROC_UNLOCK(tdt->td_proc);
2495 if (error != 0)
2496 return (error);
2497 error = native_to_linux_timespec(<s, &ts);
2498 if (error != 0)
2499 return (error);
2500 return (copyout(<s, uap->interval, sizeof(lts)));
2501 }
2502
2503 /*
2504 * In case when the Linux thread is the initial thread in
2505 * the thread group thread id is equal to the process id.
2506 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2507 */
2508 struct thread *
linux_tdfind(struct thread * td,lwpid_t tid,pid_t pid)2509 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2510 {
2511 struct linux_emuldata *em;
2512 struct thread *tdt;
2513 struct proc *p;
2514
2515 tdt = NULL;
2516 if (tid == 0 || tid == td->td_tid) {
2517 tdt = td;
2518 PROC_LOCK(tdt->td_proc);
2519 } else if (tid > PID_MAX)
2520 tdt = tdfind(tid, pid);
2521 else {
2522 /*
2523 * Initial thread where the tid equal to the pid.
2524 */
2525 p = pfind(tid);
2526 if (p != NULL) {
2527 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2528 /*
2529 * p is not a Linuxulator process.
2530 */
2531 PROC_UNLOCK(p);
2532 return (NULL);
2533 }
2534 FOREACH_THREAD_IN_PROC(p, tdt) {
2535 em = em_find(tdt);
2536 if (tid == em->em_tid)
2537 return (tdt);
2538 }
2539 PROC_UNLOCK(p);
2540 }
2541 return (NULL);
2542 }
2543
2544 return (tdt);
2545 }
2546
2547 void
linux_to_bsd_waitopts(int options,int * bsdopts)2548 linux_to_bsd_waitopts(int options, int *bsdopts)
2549 {
2550
2551 if (options & LINUX_WNOHANG)
2552 *bsdopts |= WNOHANG;
2553 if (options & LINUX_WUNTRACED)
2554 *bsdopts |= WUNTRACED;
2555 if (options & LINUX_WEXITED)
2556 *bsdopts |= WEXITED;
2557 if (options & LINUX_WCONTINUED)
2558 *bsdopts |= WCONTINUED;
2559 if (options & LINUX_WNOWAIT)
2560 *bsdopts |= WNOWAIT;
2561
2562 if (options & __WCLONE)
2563 *bsdopts |= WLINUXCLONE;
2564 }
2565
2566 int
linux_getrandom(struct thread * td,struct linux_getrandom_args * args)2567 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2568 {
2569 struct uio uio;
2570 struct iovec iov;
2571 int error;
2572
2573 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2574 return (EINVAL);
2575 if (args->count > INT_MAX)
2576 args->count = INT_MAX;
2577
2578 iov.iov_base = args->buf;
2579 iov.iov_len = args->count;
2580
2581 uio.uio_iov = &iov;
2582 uio.uio_iovcnt = 1;
2583 uio.uio_resid = iov.iov_len;
2584 uio.uio_segflg = UIO_USERSPACE;
2585 uio.uio_rw = UIO_READ;
2586 uio.uio_td = td;
2587
2588 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2589 if (error == 0)
2590 td->td_retval[0] = args->count - uio.uio_resid;
2591 return (error);
2592 }
2593
2594 int
linux_mincore(struct thread * td,struct linux_mincore_args * args)2595 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2596 {
2597
2598 /* Needs to be page-aligned */
2599 if (args->start & PAGE_MASK)
2600 return (EINVAL);
2601 return (kern_mincore(td, args->start, args->len, args->vec));
2602 }
2603