1 /*-
2 * Copyright (c) 2002 Doug Rabson
3 * Copyright (c) 1994-1995 Søren Schmidt
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer
11 * in this position and unchanged.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. The name of the author may not be used to endorse or promote products
16 * derived from this software without specific prior written permission
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 */
29
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32
33 #include "opt_compat.h"
34
35 #include <sys/param.h>
36 #include <sys/blist.h>
37 #include <sys/fcntl.h>
38 #if defined(__i386__)
39 #include <sys/imgact_aout.h>
40 #endif
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mman.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/procctl.h>
53 #include <sys/reboot.h>
54 #include <sys/racct.h>
55 #include <sys/random.h>
56 #include <sys/resourcevar.h>
57 #include <sys/sched.h>
58 #include <sys/sdt.h>
59 #include <sys/signalvar.h>
60 #include <sys/stat.h>
61 #include <sys/syscallsubr.h>
62 #include <sys/sysctl.h>
63 #include <sys/sysproto.h>
64 #include <sys/systm.h>
65 #include <sys/time.h>
66 #include <sys/vmmeter.h>
67 #include <sys/vnode.h>
68 #include <sys/wait.h>
69 #include <sys/cpuset.h>
70 #include <sys/uio.h>
71
72 #include <security/mac/mac_framework.h>
73
74 #include <vm/vm.h>
75 #include <vm/pmap.h>
76 #include <vm/vm_kern.h>
77 #include <vm/vm_map.h>
78 #include <vm/vm_extern.h>
79 #include <vm/vm_object.h>
80 #include <vm/swap_pager.h>
81
82 #ifdef COMPAT_LINUX32
83 #include <machine/../linux32/linux.h>
84 #include <machine/../linux32/linux32_proto.h>
85 #else
86 #include <machine/../linux/linux.h>
87 #include <machine/../linux/linux_proto.h>
88 #endif
89
90 #include <compat/linux/linux_dtrace.h>
91 #include <compat/linux/linux_file.h>
92 #include <compat/linux/linux_mib.h>
93 #include <compat/linux/linux_signal.h>
94 #include <compat/linux/linux_timer.h>
95 #include <compat/linux/linux_util.h>
96 #include <compat/linux/linux_sysproto.h>
97 #include <compat/linux/linux_emul.h>
98 #include <compat/linux/linux_misc.h>
99
100 /**
101 * Special DTrace provider for the linuxulator.
102 *
103 * In this file we define the provider for the entire linuxulator. All
104 * modules (= files of the linuxulator) use it.
105 *
106 * We define a different name depending on the emulated bitsize, see
107 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
108 * native bitsize = linuxulator
109 * amd64, 32bit emulation = linuxulator32
110 */
111 LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
112
113 int stclohz; /* Statistics clock frequency */
114
115 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
116 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
117 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
118 RLIMIT_MEMLOCK, RLIMIT_AS
119 };
120
121 struct l_sysinfo {
122 l_long uptime; /* Seconds since boot */
123 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
124 #define LINUX_SYSINFO_LOADS_SCALE 65536
125 l_ulong totalram; /* Total usable main memory size */
126 l_ulong freeram; /* Available memory size */
127 l_ulong sharedram; /* Amount of shared memory */
128 l_ulong bufferram; /* Memory used by buffers */
129 l_ulong totalswap; /* Total swap space size */
130 l_ulong freeswap; /* swap space still available */
131 l_ushort procs; /* Number of current processes */
132 l_ushort pads;
133 l_ulong totalbig;
134 l_ulong freebig;
135 l_uint mem_unit;
136 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
137 };
138
139 struct l_pselect6arg {
140 l_uintptr_t ss;
141 l_size_t ss_len;
142 };
143
144 static int linux_utimensat_nsec_valid(l_long);
145
146
147 int
linux_sysinfo(struct thread * td,struct linux_sysinfo_args * args)148 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
149 {
150 struct l_sysinfo sysinfo;
151 vm_object_t object;
152 int i, j;
153 struct timespec ts;
154
155 bzero(&sysinfo, sizeof(sysinfo));
156 getnanouptime(&ts);
157 if (ts.tv_nsec != 0)
158 ts.tv_sec++;
159 sysinfo.uptime = ts.tv_sec;
160
161 /* Use the information from the mib to get our load averages */
162 for (i = 0; i < 3; i++)
163 sysinfo.loads[i] = averunnable.ldavg[i] *
164 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
165
166 sysinfo.totalram = physmem * PAGE_SIZE;
167 sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
168
169 sysinfo.sharedram = 0;
170 mtx_lock(&vm_object_list_mtx);
171 TAILQ_FOREACH(object, &vm_object_list, object_list)
172 if (object->shadow_count > 1)
173 sysinfo.sharedram += object->resident_page_count;
174 mtx_unlock(&vm_object_list_mtx);
175
176 sysinfo.sharedram *= PAGE_SIZE;
177 sysinfo.bufferram = 0;
178
179 swap_pager_status(&i, &j);
180 sysinfo.totalswap = i * PAGE_SIZE;
181 sysinfo.freeswap = (i - j) * PAGE_SIZE;
182
183 sysinfo.procs = nprocs;
184
185 /* The following are only present in newer Linux kernels. */
186 sysinfo.totalbig = 0;
187 sysinfo.freebig = 0;
188 sysinfo.mem_unit = 1;
189
190 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
191 }
192
193 #ifdef LINUX_LEGACY_SYSCALLS
194 int
linux_alarm(struct thread * td,struct linux_alarm_args * args)195 linux_alarm(struct thread *td, struct linux_alarm_args *args)
196 {
197 struct itimerval it, old_it;
198 u_int secs;
199 int error;
200
201 #ifdef DEBUG
202 if (ldebug(alarm))
203 printf(ARGS(alarm, "%u"), args->secs);
204 #endif
205 secs = args->secs;
206 /*
207 * Linux alarm() is always successful. Limit secs to INT32_MAX / 2
208 * to match kern_setitimer()'s limit to avoid error from it.
209 *
210 * XXX. Linux limit secs to INT_MAX on 32 and does not limit on 64-bit
211 * platforms.
212 */
213 if (secs > INT32_MAX / 2)
214 secs = INT32_MAX / 2;
215
216 it.it_value.tv_sec = secs;
217 it.it_value.tv_usec = 0;
218 timevalclear(&it.it_interval);
219 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
220 KASSERT(error == 0, ("kern_setitimer returns %d", error));
221
222 if ((old_it.it_value.tv_sec == 0 && old_it.it_value.tv_usec > 0) ||
223 old_it.it_value.tv_usec >= 500000)
224 old_it.it_value.tv_sec++;
225 td->td_retval[0] = old_it.it_value.tv_sec;
226 return (0);
227 }
228 #endif
229
230 int
linux_brk(struct thread * td,struct linux_brk_args * args)231 linux_brk(struct thread *td, struct linux_brk_args *args)
232 {
233 struct vmspace *vm = td->td_proc->p_vmspace;
234 vm_offset_t new, old;
235 struct obreak_args /* {
236 char * nsize;
237 } */ tmp;
238
239 #ifdef DEBUG
240 if (ldebug(brk))
241 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
242 #endif
243 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
244 new = (vm_offset_t)args->dsend;
245 tmp.nsize = (char *)new;
246 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
247 td->td_retval[0] = (long)new;
248 else
249 td->td_retval[0] = (long)old;
250
251 return (0);
252 }
253
254 #if defined(__i386__)
255 /* XXX: what about amd64/linux32? */
256
257 int
linux_uselib(struct thread * td,struct linux_uselib_args * args)258 linux_uselib(struct thread *td, struct linux_uselib_args *args)
259 {
260 struct nameidata ni;
261 struct vnode *vp;
262 struct exec *a_out;
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, locked, writecount;
270
271 LCONVPATHEXIST(td, args->library, &library);
272
273 #ifdef DEBUG
274 if (ldebug(uselib))
275 printf(ARGS(uselib, "%s"), library);
276 #endif
277
278 a_out = NULL;
279 locked = 0;
280 vp = NULL;
281
282 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
283 UIO_SYSSPACE, library, td);
284 error = namei(&ni);
285 LFREEPATH(library);
286 if (error)
287 goto cleanup;
288
289 vp = ni.ni_vp;
290 NDFREE(&ni, NDF_ONLY_PNBUF);
291
292 /*
293 * From here on down, we have a locked vnode that must be unlocked.
294 * XXX: The code below largely duplicates exec_check_permissions().
295 */
296 locked = 1;
297
298 /* Writable? */
299 error = VOP_GET_WRITECOUNT(vp, &writecount);
300 if (error != 0)
301 goto cleanup;
302 if (writecount != 0) {
303 error = ETXTBSY;
304 goto cleanup;
305 }
306
307 /* Executable? */
308 error = VOP_GETATTR(vp, &attr, td->td_ucred);
309 if (error)
310 goto cleanup;
311
312 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
313 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
314 /* EACCESS is what exec(2) returns. */
315 error = ENOEXEC;
316 goto cleanup;
317 }
318
319 /* Sensible size? */
320 if (attr.va_size == 0) {
321 error = ENOEXEC;
322 goto cleanup;
323 }
324
325 /* Can we access it? */
326 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
327 if (error)
328 goto cleanup;
329
330 /*
331 * XXX: This should use vn_open() so that it is properly authorized,
332 * and to reduce code redundancy all over the place here.
333 * XXX: Not really, it duplicates far more of exec_check_permissions()
334 * than vn_open().
335 */
336 #ifdef MAC
337 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
338 if (error)
339 goto cleanup;
340 #endif
341 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
342 if (error)
343 goto cleanup;
344
345 /* Pull in executable header into exec_map */
346 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
347 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
348 if (error)
349 goto cleanup;
350
351 /* Is it a Linux binary ? */
352 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
353 error = ENOEXEC;
354 goto cleanup;
355 }
356
357 /*
358 * While we are here, we should REALLY do some more checks
359 */
360
361 /* Set file/virtual offset based on a.out variant. */
362 switch ((int)(a_out->a_magic & 0xffff)) {
363 case 0413: /* ZMAGIC */
364 file_offset = 1024;
365 break;
366 case 0314: /* QMAGIC */
367 file_offset = 0;
368 break;
369 default:
370 error = ENOEXEC;
371 goto cleanup;
372 }
373
374 bss_size = round_page(a_out->a_bss);
375
376 /* Check various fields in header for validity/bounds. */
377 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
378 error = ENOEXEC;
379 goto cleanup;
380 }
381
382 /* text + data can't exceed file size */
383 if (a_out->a_data + a_out->a_text > attr.va_size) {
384 error = EFAULT;
385 goto cleanup;
386 }
387
388 /*
389 * text/data/bss must not exceed limits
390 * XXX - this is not complete. it should check current usage PLUS
391 * the resources needed by this library.
392 */
393 PROC_LOCK(td->td_proc);
394 if (a_out->a_text > maxtsiz ||
395 a_out->a_data + bss_size > lim_cur_proc(td->td_proc, RLIMIT_DATA) ||
396 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
397 bss_size) != 0) {
398 PROC_UNLOCK(td->td_proc);
399 error = ENOMEM;
400 goto cleanup;
401 }
402 PROC_UNLOCK(td->td_proc);
403
404 /*
405 * Prevent more writers.
406 * XXX: Note that if any of the VM operations fail below we don't
407 * clear this flag.
408 */
409 VOP_SET_TEXT(vp);
410
411 /*
412 * Lock no longer needed
413 */
414 locked = 0;
415 VOP_UNLOCK(vp, 0);
416
417 /*
418 * Check if file_offset page aligned. Currently we cannot handle
419 * misalinged file offsets, and so we read in the entire image
420 * (what a waste).
421 */
422 if (file_offset & PAGE_MASK) {
423 #ifdef DEBUG
424 printf("uselib: Non page aligned binary %lu\n", file_offset);
425 #endif
426 /* Map text+data read/write/execute */
427
428 /* a_entry is the load address and is page aligned */
429 vmaddr = trunc_page(a_out->a_entry);
430
431 /* get anon user mapping, read+write+execute */
432 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
433 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
434 VM_PROT_ALL, VM_PROT_ALL, 0);
435 if (error)
436 goto cleanup;
437
438 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
439 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
440 td->td_ucred, NOCRED, &aresid, td);
441 if (error != 0)
442 goto cleanup;
443 if (aresid != 0) {
444 error = ENOEXEC;
445 goto cleanup;
446 }
447 } else {
448 #ifdef DEBUG
449 printf("uselib: Page aligned binary %lu\n", file_offset);
450 #endif
451 /*
452 * for QMAGIC, a_entry is 20 bytes beyond the load address
453 * to skip the executable header
454 */
455 vmaddr = trunc_page(a_out->a_entry);
456
457 /*
458 * Map it all into the process's space as a single
459 * copy-on-write "data" segment.
460 */
461 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
462 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
463 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
464 if (error)
465 goto cleanup;
466 }
467 #ifdef DEBUG
468 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
469 ((long *)vmaddr)[1]);
470 #endif
471 if (bss_size != 0) {
472 /* Calculate BSS start address */
473 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
474 a_out->a_data;
475
476 /* allocate some 'anon' space */
477 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
478 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
479 VM_PROT_ALL, 0);
480 if (error)
481 goto cleanup;
482 }
483
484 cleanup:
485 /* Unlock vnode if needed */
486 if (locked)
487 VOP_UNLOCK(vp, 0);
488
489 /* Release the temporary mapping. */
490 if (a_out)
491 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
492
493 return (error);
494 }
495
496 #endif /* __i386__ */
497
498 #ifdef LINUX_LEGACY_SYSCALLS
499 int
linux_select(struct thread * td,struct linux_select_args * args)500 linux_select(struct thread *td, struct linux_select_args *args)
501 {
502 l_timeval ltv;
503 struct timeval tv0, tv1, utv, *tvp;
504 int error;
505
506 #ifdef DEBUG
507 if (ldebug(select))
508 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
509 (void *)args->readfds, (void *)args->writefds,
510 (void *)args->exceptfds, (void *)args->timeout);
511 #endif
512
513 /*
514 * Store current time for computation of the amount of
515 * time left.
516 */
517 if (args->timeout) {
518 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
519 goto select_out;
520 utv.tv_sec = ltv.tv_sec;
521 utv.tv_usec = ltv.tv_usec;
522 #ifdef DEBUG
523 if (ldebug(select))
524 printf(LMSG("incoming timeout (%jd/%ld)"),
525 (intmax_t)utv.tv_sec, utv.tv_usec);
526 #endif
527
528 if (itimerfix(&utv)) {
529 /*
530 * The timeval was invalid. Convert it to something
531 * valid that will act as it does under Linux.
532 */
533 utv.tv_sec += utv.tv_usec / 1000000;
534 utv.tv_usec %= 1000000;
535 if (utv.tv_usec < 0) {
536 utv.tv_sec -= 1;
537 utv.tv_usec += 1000000;
538 }
539 if (utv.tv_sec < 0)
540 timevalclear(&utv);
541 }
542 microtime(&tv0);
543 tvp = &utv;
544 } else
545 tvp = NULL;
546
547 error = kern_select(td, args->nfds, args->readfds, args->writefds,
548 args->exceptfds, tvp, LINUX_NFDBITS);
549
550 #ifdef DEBUG
551 if (ldebug(select))
552 printf(LMSG("real select returns %d"), error);
553 #endif
554 if (error)
555 goto select_out;
556
557 if (args->timeout) {
558 if (td->td_retval[0]) {
559 /*
560 * Compute how much time was left of the timeout,
561 * by subtracting the current time and the time
562 * before we started the call, and subtracting
563 * that result from the user-supplied value.
564 */
565 microtime(&tv1);
566 timevalsub(&tv1, &tv0);
567 timevalsub(&utv, &tv1);
568 if (utv.tv_sec < 0)
569 timevalclear(&utv);
570 } else
571 timevalclear(&utv);
572 #ifdef DEBUG
573 if (ldebug(select))
574 printf(LMSG("outgoing timeout (%jd/%ld)"),
575 (intmax_t)utv.tv_sec, utv.tv_usec);
576 #endif
577 ltv.tv_sec = utv.tv_sec;
578 ltv.tv_usec = utv.tv_usec;
579 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
580 goto select_out;
581 }
582
583 select_out:
584 #ifdef DEBUG
585 if (ldebug(select))
586 printf(LMSG("select_out -> %d"), error);
587 #endif
588 return (error);
589 }
590 #endif
591
592 int
linux_mremap(struct thread * td,struct linux_mremap_args * args)593 linux_mremap(struct thread *td, struct linux_mremap_args *args)
594 {
595 uintptr_t addr;
596 size_t len;
597 int error = 0;
598
599 #ifdef DEBUG
600 if (ldebug(mremap))
601 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
602 (void *)(uintptr_t)args->addr,
603 (unsigned long)args->old_len,
604 (unsigned long)args->new_len,
605 (unsigned long)args->flags);
606 #endif
607
608 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
609 td->td_retval[0] = 0;
610 return (EINVAL);
611 }
612
613 /*
614 * Check for the page alignment.
615 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
616 */
617 if (args->addr & PAGE_MASK) {
618 td->td_retval[0] = 0;
619 return (EINVAL);
620 }
621
622 args->new_len = round_page(args->new_len);
623 args->old_len = round_page(args->old_len);
624
625 if (args->new_len > args->old_len) {
626 td->td_retval[0] = 0;
627 return (ENOMEM);
628 }
629
630 if (args->new_len < args->old_len) {
631 addr = args->addr + args->new_len;
632 len = args->old_len - args->new_len;
633 error = kern_munmap(td, addr, len);
634 }
635
636 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
637 return (error);
638 }
639
640 #define LINUX_MS_ASYNC 0x0001
641 #define LINUX_MS_INVALIDATE 0x0002
642 #define LINUX_MS_SYNC 0x0004
643
644 int
linux_msync(struct thread * td,struct linux_msync_args * args)645 linux_msync(struct thread *td, struct linux_msync_args *args)
646 {
647
648 return (kern_msync(td, args->addr, args->len,
649 args->fl & ~LINUX_MS_SYNC));
650 }
651
652 #ifdef LINUX_LEGACY_SYSCALLS
653 int
linux_time(struct thread * td,struct linux_time_args * args)654 linux_time(struct thread *td, struct linux_time_args *args)
655 {
656 struct timeval tv;
657 l_time_t tm;
658 int error;
659
660 #ifdef DEBUG
661 if (ldebug(time))
662 printf(ARGS(time, "*"));
663 #endif
664
665 microtime(&tv);
666 tm = tv.tv_sec;
667 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
668 return (error);
669 td->td_retval[0] = tm;
670 return (0);
671 }
672 #endif
673
674 struct l_times_argv {
675 l_clock_t tms_utime;
676 l_clock_t tms_stime;
677 l_clock_t tms_cutime;
678 l_clock_t tms_cstime;
679 };
680
681
682 /*
683 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
684 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
685 * auxiliary vector entry.
686 */
687 #define CLK_TCK 100
688
689 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
690 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
691
692 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
693 CONVNTCK(r) : CONVOTCK(r))
694
695 int
linux_times(struct thread * td,struct linux_times_args * args)696 linux_times(struct thread *td, struct linux_times_args *args)
697 {
698 struct timeval tv, utime, stime, cutime, cstime;
699 struct l_times_argv tms;
700 struct proc *p;
701 int error;
702
703 #ifdef DEBUG
704 if (ldebug(times))
705 printf(ARGS(times, "*"));
706 #endif
707
708 if (args->buf != NULL) {
709 p = td->td_proc;
710 PROC_LOCK(p);
711 PROC_STATLOCK(p);
712 calcru(p, &utime, &stime);
713 PROC_STATUNLOCK(p);
714 calccru(p, &cutime, &cstime);
715 PROC_UNLOCK(p);
716
717 tms.tms_utime = CONVTCK(utime);
718 tms.tms_stime = CONVTCK(stime);
719
720 tms.tms_cutime = CONVTCK(cutime);
721 tms.tms_cstime = CONVTCK(cstime);
722
723 if ((error = copyout(&tms, args->buf, sizeof(tms))))
724 return (error);
725 }
726
727 microuptime(&tv);
728 td->td_retval[0] = (int)CONVTCK(tv);
729 return (0);
730 }
731
732 int
linux_newuname(struct thread * td,struct linux_newuname_args * args)733 linux_newuname(struct thread *td, struct linux_newuname_args *args)
734 {
735 struct l_new_utsname utsname;
736 char osname[LINUX_MAX_UTSNAME];
737 char osrelease[LINUX_MAX_UTSNAME];
738 char *p;
739
740 #ifdef DEBUG
741 if (ldebug(newuname))
742 printf(ARGS(newuname, "*"));
743 #endif
744
745 linux_get_osname(td, osname);
746 linux_get_osrelease(td, osrelease);
747
748 bzero(&utsname, sizeof(utsname));
749 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
750 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
751 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
752 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
753 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
754 for (p = utsname.version; *p != '\0'; ++p)
755 if (*p == '\n') {
756 *p = '\0';
757 break;
758 }
759 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
760
761 return (copyout(&utsname, args->buf, sizeof(utsname)));
762 }
763
764 struct l_utimbuf {
765 l_time_t l_actime;
766 l_time_t l_modtime;
767 };
768
769 #ifdef LINUX_LEGACY_SYSCALLS
770 int
linux_utime(struct thread * td,struct linux_utime_args * args)771 linux_utime(struct thread *td, struct linux_utime_args *args)
772 {
773 struct timeval tv[2], *tvp;
774 struct l_utimbuf lut;
775 char *fname;
776 int error;
777
778 LCONVPATHEXIST(td, args->fname, &fname);
779
780 #ifdef DEBUG
781 if (ldebug(utime))
782 printf(ARGS(utime, "%s, *"), fname);
783 #endif
784
785 if (args->times) {
786 if ((error = copyin(args->times, &lut, sizeof lut))) {
787 LFREEPATH(fname);
788 return (error);
789 }
790 tv[0].tv_sec = lut.l_actime;
791 tv[0].tv_usec = 0;
792 tv[1].tv_sec = lut.l_modtime;
793 tv[1].tv_usec = 0;
794 tvp = tv;
795 } else
796 tvp = NULL;
797
798 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
799 UIO_SYSSPACE);
800 LFREEPATH(fname);
801 return (error);
802 }
803 #endif
804
805 #ifdef LINUX_LEGACY_SYSCALLS
806 int
linux_utimes(struct thread * td,struct linux_utimes_args * args)807 linux_utimes(struct thread *td, struct linux_utimes_args *args)
808 {
809 l_timeval ltv[2];
810 struct timeval tv[2], *tvp = NULL;
811 char *fname;
812 int error;
813
814 LCONVPATHEXIST(td, args->fname, &fname);
815
816 #ifdef DEBUG
817 if (ldebug(utimes))
818 printf(ARGS(utimes, "%s, *"), fname);
819 #endif
820
821 if (args->tptr != NULL) {
822 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
823 LFREEPATH(fname);
824 return (error);
825 }
826 tv[0].tv_sec = ltv[0].tv_sec;
827 tv[0].tv_usec = ltv[0].tv_usec;
828 tv[1].tv_sec = ltv[1].tv_sec;
829 tv[1].tv_usec = ltv[1].tv_usec;
830 tvp = tv;
831 }
832
833 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
834 tvp, UIO_SYSSPACE);
835 LFREEPATH(fname);
836 return (error);
837 }
838 #endif
839
840 static int
linux_utimensat_nsec_valid(l_long nsec)841 linux_utimensat_nsec_valid(l_long nsec)
842 {
843
844 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
845 return (0);
846 if (nsec >= 0 && nsec <= 999999999)
847 return (0);
848 return (1);
849 }
850
851 int
linux_utimensat(struct thread * td,struct linux_utimensat_args * args)852 linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
853 {
854 struct l_timespec l_times[2];
855 struct timespec times[2], *timesp = NULL;
856 char *path = NULL;
857 int error, dfd, flags = 0;
858
859 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
860
861 #ifdef DEBUG
862 if (ldebug(utimensat))
863 printf(ARGS(utimensat, "%d, *"), dfd);
864 #endif
865
866 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
867 return (EINVAL);
868
869 if (args->times != NULL) {
870 error = copyin(args->times, l_times, sizeof(l_times));
871 if (error != 0)
872 return (error);
873
874 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
875 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
876 return (EINVAL);
877
878 times[0].tv_sec = l_times[0].tv_sec;
879 switch (l_times[0].tv_nsec)
880 {
881 case LINUX_UTIME_OMIT:
882 times[0].tv_nsec = UTIME_OMIT;
883 break;
884 case LINUX_UTIME_NOW:
885 times[0].tv_nsec = UTIME_NOW;
886 break;
887 default:
888 times[0].tv_nsec = l_times[0].tv_nsec;
889 }
890
891 times[1].tv_sec = l_times[1].tv_sec;
892 switch (l_times[1].tv_nsec)
893 {
894 case LINUX_UTIME_OMIT:
895 times[1].tv_nsec = UTIME_OMIT;
896 break;
897 case LINUX_UTIME_NOW:
898 times[1].tv_nsec = UTIME_NOW;
899 break;
900 default:
901 times[1].tv_nsec = l_times[1].tv_nsec;
902 break;
903 }
904 timesp = times;
905
906 /* This breaks POSIX, but is what the Linux kernel does
907 * _on purpose_ (documented in the man page for utimensat(2)),
908 * so we must follow that behaviour. */
909 if (times[0].tv_nsec == UTIME_OMIT &&
910 times[1].tv_nsec == UTIME_OMIT)
911 return (0);
912 }
913
914 if (args->pathname != NULL)
915 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
916 else if (args->flags != 0)
917 return (EINVAL);
918
919 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
920 flags |= AT_SYMLINK_NOFOLLOW;
921
922 if (path == NULL)
923 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
924 else {
925 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
926 UIO_SYSSPACE, flags);
927 LFREEPATH(path);
928 }
929
930 return (error);
931 }
932
933 #ifdef LINUX_LEGACY_SYSCALLS
934 int
linux_futimesat(struct thread * td,struct linux_futimesat_args * args)935 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
936 {
937 l_timeval ltv[2];
938 struct timeval tv[2], *tvp = NULL;
939 char *fname;
940 int error, dfd;
941
942 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
943 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
944
945 #ifdef DEBUG
946 if (ldebug(futimesat))
947 printf(ARGS(futimesat, "%s, *"), fname);
948 #endif
949
950 if (args->utimes != NULL) {
951 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
952 LFREEPATH(fname);
953 return (error);
954 }
955 tv[0].tv_sec = ltv[0].tv_sec;
956 tv[0].tv_usec = ltv[0].tv_usec;
957 tv[1].tv_sec = ltv[1].tv_sec;
958 tv[1].tv_usec = ltv[1].tv_usec;
959 tvp = tv;
960 }
961
962 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
963 LFREEPATH(fname);
964 return (error);
965 }
966 #endif
967
968 int
linux_common_wait(struct thread * td,int pid,int * status,int options,struct rusage * ru)969 linux_common_wait(struct thread *td, int pid, int *status,
970 int options, struct rusage *ru)
971 {
972 int error, tmpstat;
973
974 error = kern_wait(td, pid, &tmpstat, options, ru);
975 if (error)
976 return (error);
977
978 if (status) {
979 tmpstat &= 0xffff;
980 if (WIFSIGNALED(tmpstat))
981 tmpstat = (tmpstat & 0xffffff80) |
982 bsd_to_linux_signal(WTERMSIG(tmpstat));
983 else if (WIFSTOPPED(tmpstat))
984 tmpstat = (tmpstat & 0xffff00ff) |
985 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
986 else if (WIFCONTINUED(tmpstat))
987 tmpstat = 0xffff;
988 error = copyout(&tmpstat, status, sizeof(int));
989 }
990
991 return (error);
992 }
993
994 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
995 int
linux_waitpid(struct thread * td,struct linux_waitpid_args * args)996 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
997 {
998 struct linux_wait4_args wait4_args;
999
1000 #ifdef DEBUG
1001 if (ldebug(waitpid))
1002 printf(ARGS(waitpid, "%d, %p, %d"),
1003 args->pid, (void *)args->status, args->options);
1004 #endif
1005
1006 wait4_args.pid = args->pid;
1007 wait4_args.status = args->status;
1008 wait4_args.options = args->options;
1009 wait4_args.rusage = NULL;
1010
1011 return (linux_wait4(td, &wait4_args));
1012 }
1013 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1014
1015 int
linux_wait4(struct thread * td,struct linux_wait4_args * args)1016 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1017 {
1018 int error, options;
1019 struct rusage ru, *rup;
1020
1021 #ifdef DEBUG
1022 if (ldebug(wait4))
1023 printf(ARGS(wait4, "%d, %p, %d, %p"),
1024 args->pid, (void *)args->status, args->options,
1025 (void *)args->rusage);
1026 #endif
1027 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1028 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1029 return (EINVAL);
1030
1031 options = WEXITED;
1032 linux_to_bsd_waitopts(args->options, &options);
1033
1034 if (args->rusage != NULL)
1035 rup = &ru;
1036 else
1037 rup = NULL;
1038 error = linux_common_wait(td, args->pid, args->status, options, rup);
1039 if (error != 0)
1040 return (error);
1041 if (args->rusage != NULL)
1042 error = linux_copyout_rusage(&ru, args->rusage);
1043 return (error);
1044 }
1045
1046 int
linux_waitid(struct thread * td,struct linux_waitid_args * args)1047 linux_waitid(struct thread *td, struct linux_waitid_args *args)
1048 {
1049 int status, options, sig;
1050 struct __wrusage wru;
1051 siginfo_t siginfo;
1052 l_siginfo_t lsi;
1053 idtype_t idtype;
1054 struct proc *p;
1055 int error;
1056
1057 options = 0;
1058 linux_to_bsd_waitopts(args->options, &options);
1059
1060 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1061 return (EINVAL);
1062 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1063 return (EINVAL);
1064
1065 switch (args->idtype) {
1066 case LINUX_P_ALL:
1067 idtype = P_ALL;
1068 break;
1069 case LINUX_P_PID:
1070 if (args->id <= 0)
1071 return (EINVAL);
1072 idtype = P_PID;
1073 break;
1074 case LINUX_P_PGID:
1075 if (args->id <= 0)
1076 return (EINVAL);
1077 idtype = P_PGID;
1078 break;
1079 default:
1080 return (EINVAL);
1081 }
1082
1083 error = kern_wait6(td, idtype, args->id, &status, options,
1084 &wru, &siginfo);
1085 if (error != 0)
1086 return (error);
1087 if (args->rusage != NULL) {
1088 error = linux_copyout_rusage(&wru.wru_children,
1089 args->rusage);
1090 if (error != 0)
1091 return (error);
1092 }
1093 if (args->info != NULL) {
1094 p = td->td_proc;
1095 bzero(&lsi, sizeof(lsi));
1096 if (td->td_retval[0] != 0) {
1097 sig = bsd_to_linux_signal(siginfo.si_signo);
1098 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1099 }
1100 error = copyout(&lsi, args->info, sizeof(lsi));
1101 }
1102 td->td_retval[0] = 0;
1103
1104 return (error);
1105 }
1106
1107 #ifdef LINUX_LEGACY_SYSCALLS
1108 int
linux_mknod(struct thread * td,struct linux_mknod_args * args)1109 linux_mknod(struct thread *td, struct linux_mknod_args *args)
1110 {
1111 char *path;
1112 int error;
1113
1114 LCONVPATHCREAT(td, args->path, &path);
1115
1116 #ifdef DEBUG
1117 if (ldebug(mknod))
1118 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1119 (uintmax_t)args->dev);
1120 #endif
1121
1122 switch (args->mode & S_IFMT) {
1123 case S_IFIFO:
1124 case S_IFSOCK:
1125 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1126 args->mode);
1127 break;
1128
1129 case S_IFCHR:
1130 case S_IFBLK:
1131 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1132 args->mode, args->dev);
1133 break;
1134
1135 case S_IFDIR:
1136 error = EPERM;
1137 break;
1138
1139 case 0:
1140 args->mode |= S_IFREG;
1141 /* FALLTHROUGH */
1142 case S_IFREG:
1143 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1144 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1145 if (error == 0)
1146 kern_close(td, td->td_retval[0]);
1147 break;
1148
1149 default:
1150 error = EINVAL;
1151 break;
1152 }
1153 LFREEPATH(path);
1154 return (error);
1155 }
1156 #endif
1157
1158 int
linux_mknodat(struct thread * td,struct linux_mknodat_args * args)1159 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1160 {
1161 char *path;
1162 int error, dfd;
1163
1164 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1165 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1166
1167 #ifdef DEBUG
1168 if (ldebug(mknodat))
1169 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1170 #endif
1171
1172 switch (args->mode & S_IFMT) {
1173 case S_IFIFO:
1174 case S_IFSOCK:
1175 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1176 break;
1177
1178 case S_IFCHR:
1179 case S_IFBLK:
1180 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1181 args->dev);
1182 break;
1183
1184 case S_IFDIR:
1185 error = EPERM;
1186 break;
1187
1188 case 0:
1189 args->mode |= S_IFREG;
1190 /* FALLTHROUGH */
1191 case S_IFREG:
1192 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1193 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1194 if (error == 0)
1195 kern_close(td, td->td_retval[0]);
1196 break;
1197
1198 default:
1199 error = EINVAL;
1200 break;
1201 }
1202 LFREEPATH(path);
1203 return (error);
1204 }
1205
1206 /*
1207 * UGH! This is just about the dumbest idea I've ever heard!!
1208 */
1209 int
linux_personality(struct thread * td,struct linux_personality_args * args)1210 linux_personality(struct thread *td, struct linux_personality_args *args)
1211 {
1212 struct linux_pemuldata *pem;
1213 struct proc *p = td->td_proc;
1214 uint32_t old;
1215
1216 #ifdef DEBUG
1217 if (ldebug(personality))
1218 printf(ARGS(personality, "%u"), args->per);
1219 #endif
1220
1221 PROC_LOCK(p);
1222 pem = pem_find(p);
1223 old = pem->persona;
1224 if (args->per != 0xffffffff)
1225 pem->persona = args->per;
1226 PROC_UNLOCK(p);
1227
1228 td->td_retval[0] = old;
1229 return (0);
1230 }
1231
1232 struct l_itimerval {
1233 l_timeval it_interval;
1234 l_timeval it_value;
1235 };
1236
1237 #define B2L_ITIMERVAL(bip, lip) \
1238 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1239 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1240 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1241 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1242
1243 int
linux_setitimer(struct thread * td,struct linux_setitimer_args * uap)1244 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1245 {
1246 int error;
1247 struct l_itimerval ls;
1248 struct itimerval aitv, oitv;
1249
1250 #ifdef DEBUG
1251 if (ldebug(setitimer))
1252 printf(ARGS(setitimer, "%p, %p"),
1253 (void *)uap->itv, (void *)uap->oitv);
1254 #endif
1255
1256 if (uap->itv == NULL) {
1257 uap->itv = uap->oitv;
1258 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1259 }
1260
1261 error = copyin(uap->itv, &ls, sizeof(ls));
1262 if (error != 0)
1263 return (error);
1264 B2L_ITIMERVAL(&aitv, &ls);
1265 #ifdef DEBUG
1266 if (ldebug(setitimer)) {
1267 printf("setitimer: value: sec: %jd, usec: %ld\n",
1268 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1269 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1270 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1271 }
1272 #endif
1273 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1274 if (error != 0 || uap->oitv == NULL)
1275 return (error);
1276 B2L_ITIMERVAL(&ls, &oitv);
1277
1278 return (copyout(&ls, uap->oitv, sizeof(ls)));
1279 }
1280
1281 int
linux_getitimer(struct thread * td,struct linux_getitimer_args * uap)1282 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1283 {
1284 int error;
1285 struct l_itimerval ls;
1286 struct itimerval aitv;
1287
1288 #ifdef DEBUG
1289 if (ldebug(getitimer))
1290 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1291 #endif
1292 error = kern_getitimer(td, uap->which, &aitv);
1293 if (error != 0)
1294 return (error);
1295 B2L_ITIMERVAL(&ls, &aitv);
1296 return (copyout(&ls, uap->itv, sizeof(ls)));
1297 }
1298
1299 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1300 int
linux_nice(struct thread * td,struct linux_nice_args * args)1301 linux_nice(struct thread *td, struct linux_nice_args *args)
1302 {
1303 struct setpriority_args bsd_args;
1304
1305 bsd_args.which = PRIO_PROCESS;
1306 bsd_args.who = 0; /* current process */
1307 bsd_args.prio = args->inc;
1308 return (sys_setpriority(td, &bsd_args));
1309 }
1310 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1311
1312 int
linux_setgroups(struct thread * td,struct linux_setgroups_args * args)1313 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1314 {
1315 struct ucred *newcred, *oldcred;
1316 l_gid_t *linux_gidset;
1317 gid_t *bsd_gidset;
1318 int ngrp, error;
1319 struct proc *p;
1320
1321 ngrp = args->gidsetsize;
1322 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1323 return (EINVAL);
1324 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1325 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1326 if (error)
1327 goto out;
1328 newcred = crget();
1329 crextend(newcred, ngrp + 1);
1330 p = td->td_proc;
1331 PROC_LOCK(p);
1332 oldcred = p->p_ucred;
1333 crcopy(newcred, oldcred);
1334
1335 /*
1336 * cr_groups[0] holds egid. Setting the whole set from
1337 * the supplied set will cause egid to be changed too.
1338 * Keep cr_groups[0] unchanged to prevent that.
1339 */
1340
1341 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1342 PROC_UNLOCK(p);
1343 crfree(newcred);
1344 goto out;
1345 }
1346
1347 if (ngrp > 0) {
1348 newcred->cr_ngroups = ngrp + 1;
1349
1350 bsd_gidset = newcred->cr_groups;
1351 ngrp--;
1352 while (ngrp >= 0) {
1353 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1354 ngrp--;
1355 }
1356 } else
1357 newcred->cr_ngroups = 1;
1358
1359 setsugid(p);
1360 proc_set_cred(p, newcred);
1361 PROC_UNLOCK(p);
1362 crfree(oldcred);
1363 error = 0;
1364 out:
1365 free(linux_gidset, M_LINUX);
1366 return (error);
1367 }
1368
1369 int
linux_getgroups(struct thread * td,struct linux_getgroups_args * args)1370 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1371 {
1372 struct ucred *cred;
1373 l_gid_t *linux_gidset;
1374 gid_t *bsd_gidset;
1375 int bsd_gidsetsz, ngrp, error;
1376
1377 cred = td->td_ucred;
1378 bsd_gidset = cred->cr_groups;
1379 bsd_gidsetsz = cred->cr_ngroups - 1;
1380
1381 /*
1382 * cr_groups[0] holds egid. Returning the whole set
1383 * here will cause a duplicate. Exclude cr_groups[0]
1384 * to prevent that.
1385 */
1386
1387 if ((ngrp = args->gidsetsize) == 0) {
1388 td->td_retval[0] = bsd_gidsetsz;
1389 return (0);
1390 }
1391
1392 if (ngrp < bsd_gidsetsz)
1393 return (EINVAL);
1394
1395 ngrp = 0;
1396 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1397 M_LINUX, M_WAITOK);
1398 while (ngrp < bsd_gidsetsz) {
1399 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1400 ngrp++;
1401 }
1402
1403 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1404 free(linux_gidset, M_LINUX);
1405 if (error)
1406 return (error);
1407
1408 td->td_retval[0] = ngrp;
1409 return (0);
1410 }
1411
1412 int
linux_setrlimit(struct thread * td,struct linux_setrlimit_args * args)1413 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1414 {
1415 struct rlimit bsd_rlim;
1416 struct l_rlimit rlim;
1417 u_int which;
1418 int error;
1419
1420 #ifdef DEBUG
1421 if (ldebug(setrlimit))
1422 printf(ARGS(setrlimit, "%d, %p"),
1423 args->resource, (void *)args->rlim);
1424 #endif
1425
1426 if (args->resource >= LINUX_RLIM_NLIMITS)
1427 return (EINVAL);
1428
1429 which = linux_to_bsd_resource[args->resource];
1430 if (which == -1)
1431 return (EINVAL);
1432
1433 error = copyin(args->rlim, &rlim, sizeof(rlim));
1434 if (error)
1435 return (error);
1436
1437 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1438 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1439 return (kern_setrlimit(td, which, &bsd_rlim));
1440 }
1441
1442 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1443 int
linux_old_getrlimit(struct thread * td,struct linux_old_getrlimit_args * args)1444 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1445 {
1446 struct l_rlimit rlim;
1447 struct rlimit bsd_rlim;
1448 u_int which;
1449
1450 #ifdef DEBUG
1451 if (ldebug(old_getrlimit))
1452 printf(ARGS(old_getrlimit, "%d, %p"),
1453 args->resource, (void *)args->rlim);
1454 #endif
1455
1456 if (args->resource >= LINUX_RLIM_NLIMITS)
1457 return (EINVAL);
1458
1459 which = linux_to_bsd_resource[args->resource];
1460 if (which == -1)
1461 return (EINVAL);
1462
1463 lim_rlimit(td, which, &bsd_rlim);
1464
1465 #ifdef COMPAT_LINUX32
1466 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1467 if (rlim.rlim_cur == UINT_MAX)
1468 rlim.rlim_cur = INT_MAX;
1469 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1470 if (rlim.rlim_max == UINT_MAX)
1471 rlim.rlim_max = INT_MAX;
1472 #else
1473 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1474 if (rlim.rlim_cur == ULONG_MAX)
1475 rlim.rlim_cur = LONG_MAX;
1476 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1477 if (rlim.rlim_max == ULONG_MAX)
1478 rlim.rlim_max = LONG_MAX;
1479 #endif
1480 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1481 }
1482 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1483
1484 int
linux_getrlimit(struct thread * td,struct linux_getrlimit_args * args)1485 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1486 {
1487 struct l_rlimit rlim;
1488 struct rlimit bsd_rlim;
1489 u_int which;
1490
1491 #ifdef DEBUG
1492 if (ldebug(getrlimit))
1493 printf(ARGS(getrlimit, "%d, %p"),
1494 args->resource, (void *)args->rlim);
1495 #endif
1496
1497 if (args->resource >= LINUX_RLIM_NLIMITS)
1498 return (EINVAL);
1499
1500 which = linux_to_bsd_resource[args->resource];
1501 if (which == -1)
1502 return (EINVAL);
1503
1504 lim_rlimit(td, which, &bsd_rlim);
1505
1506 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1507 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1508 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1509 }
1510
1511 int
linux_sched_setscheduler(struct thread * td,struct linux_sched_setscheduler_args * args)1512 linux_sched_setscheduler(struct thread *td,
1513 struct linux_sched_setscheduler_args *args)
1514 {
1515 struct sched_param sched_param;
1516 struct thread *tdt;
1517 int error, policy;
1518
1519 #ifdef DEBUG
1520 if (ldebug(sched_setscheduler))
1521 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1522 args->pid, args->policy, (const void *)args->param);
1523 #endif
1524
1525 switch (args->policy) {
1526 case LINUX_SCHED_OTHER:
1527 policy = SCHED_OTHER;
1528 break;
1529 case LINUX_SCHED_FIFO:
1530 policy = SCHED_FIFO;
1531 break;
1532 case LINUX_SCHED_RR:
1533 policy = SCHED_RR;
1534 break;
1535 default:
1536 return (EINVAL);
1537 }
1538
1539 error = copyin(args->param, &sched_param, sizeof(sched_param));
1540 if (error)
1541 return (error);
1542
1543 tdt = linux_tdfind(td, args->pid, -1);
1544 if (tdt == NULL)
1545 return (ESRCH);
1546
1547 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1548 PROC_UNLOCK(tdt->td_proc);
1549 return (error);
1550 }
1551
1552 int
linux_sched_getscheduler(struct thread * td,struct linux_sched_getscheduler_args * args)1553 linux_sched_getscheduler(struct thread *td,
1554 struct linux_sched_getscheduler_args *args)
1555 {
1556 struct thread *tdt;
1557 int error, policy;
1558
1559 #ifdef DEBUG
1560 if (ldebug(sched_getscheduler))
1561 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1562 #endif
1563
1564 tdt = linux_tdfind(td, args->pid, -1);
1565 if (tdt == NULL)
1566 return (ESRCH);
1567
1568 error = kern_sched_getscheduler(td, tdt, &policy);
1569 PROC_UNLOCK(tdt->td_proc);
1570
1571 switch (policy) {
1572 case SCHED_OTHER:
1573 td->td_retval[0] = LINUX_SCHED_OTHER;
1574 break;
1575 case SCHED_FIFO:
1576 td->td_retval[0] = LINUX_SCHED_FIFO;
1577 break;
1578 case SCHED_RR:
1579 td->td_retval[0] = LINUX_SCHED_RR;
1580 break;
1581 }
1582 return (error);
1583 }
1584
1585 int
linux_sched_get_priority_max(struct thread * td,struct linux_sched_get_priority_max_args * args)1586 linux_sched_get_priority_max(struct thread *td,
1587 struct linux_sched_get_priority_max_args *args)
1588 {
1589 struct sched_get_priority_max_args bsd;
1590
1591 #ifdef DEBUG
1592 if (ldebug(sched_get_priority_max))
1593 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1594 #endif
1595
1596 switch (args->policy) {
1597 case LINUX_SCHED_OTHER:
1598 bsd.policy = SCHED_OTHER;
1599 break;
1600 case LINUX_SCHED_FIFO:
1601 bsd.policy = SCHED_FIFO;
1602 break;
1603 case LINUX_SCHED_RR:
1604 bsd.policy = SCHED_RR;
1605 break;
1606 default:
1607 return (EINVAL);
1608 }
1609 return (sys_sched_get_priority_max(td, &bsd));
1610 }
1611
1612 int
linux_sched_get_priority_min(struct thread * td,struct linux_sched_get_priority_min_args * args)1613 linux_sched_get_priority_min(struct thread *td,
1614 struct linux_sched_get_priority_min_args *args)
1615 {
1616 struct sched_get_priority_min_args bsd;
1617
1618 #ifdef DEBUG
1619 if (ldebug(sched_get_priority_min))
1620 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1621 #endif
1622
1623 switch (args->policy) {
1624 case LINUX_SCHED_OTHER:
1625 bsd.policy = SCHED_OTHER;
1626 break;
1627 case LINUX_SCHED_FIFO:
1628 bsd.policy = SCHED_FIFO;
1629 break;
1630 case LINUX_SCHED_RR:
1631 bsd.policy = SCHED_RR;
1632 break;
1633 default:
1634 return (EINVAL);
1635 }
1636 return (sys_sched_get_priority_min(td, &bsd));
1637 }
1638
1639 #define REBOOT_CAD_ON 0x89abcdef
1640 #define REBOOT_CAD_OFF 0
1641 #define REBOOT_HALT 0xcdef0123
1642 #define REBOOT_RESTART 0x01234567
1643 #define REBOOT_RESTART2 0xA1B2C3D4
1644 #define REBOOT_POWEROFF 0x4321FEDC
1645 #define REBOOT_MAGIC1 0xfee1dead
1646 #define REBOOT_MAGIC2 0x28121969
1647 #define REBOOT_MAGIC2A 0x05121996
1648 #define REBOOT_MAGIC2B 0x16041998
1649
1650 int
linux_reboot(struct thread * td,struct linux_reboot_args * args)1651 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1652 {
1653 struct reboot_args bsd_args;
1654
1655 #ifdef DEBUG
1656 if (ldebug(reboot))
1657 printf(ARGS(reboot, "0x%x"), args->cmd);
1658 #endif
1659
1660 if (args->magic1 != REBOOT_MAGIC1)
1661 return (EINVAL);
1662
1663 switch (args->magic2) {
1664 case REBOOT_MAGIC2:
1665 case REBOOT_MAGIC2A:
1666 case REBOOT_MAGIC2B:
1667 break;
1668 default:
1669 return (EINVAL);
1670 }
1671
1672 switch (args->cmd) {
1673 case REBOOT_CAD_ON:
1674 case REBOOT_CAD_OFF:
1675 return (priv_check(td, PRIV_REBOOT));
1676 case REBOOT_HALT:
1677 bsd_args.opt = RB_HALT;
1678 break;
1679 case REBOOT_RESTART:
1680 case REBOOT_RESTART2:
1681 bsd_args.opt = 0;
1682 break;
1683 case REBOOT_POWEROFF:
1684 bsd_args.opt = RB_POWEROFF;
1685 break;
1686 default:
1687 return (EINVAL);
1688 }
1689 return (sys_reboot(td, &bsd_args));
1690 }
1691
1692
1693 /*
1694 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1695 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1696 * are assumed to be preserved. The following lightweight syscalls fixes
1697 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1698 *
1699 * linux_getpid() - MP SAFE
1700 * linux_getgid() - MP SAFE
1701 * linux_getuid() - MP SAFE
1702 */
1703
1704 int
linux_getpid(struct thread * td,struct linux_getpid_args * args)1705 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1706 {
1707
1708 #ifdef DEBUG
1709 if (ldebug(getpid))
1710 printf(ARGS(getpid, ""));
1711 #endif
1712 td->td_retval[0] = td->td_proc->p_pid;
1713
1714 return (0);
1715 }
1716
1717 int
linux_gettid(struct thread * td,struct linux_gettid_args * args)1718 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1719 {
1720 struct linux_emuldata *em;
1721
1722 #ifdef DEBUG
1723 if (ldebug(gettid))
1724 printf(ARGS(gettid, ""));
1725 #endif
1726
1727 em = em_find(td);
1728 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1729
1730 td->td_retval[0] = em->em_tid;
1731
1732 return (0);
1733 }
1734
1735
1736 int
linux_getppid(struct thread * td,struct linux_getppid_args * args)1737 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1738 {
1739
1740 #ifdef DEBUG
1741 if (ldebug(getppid))
1742 printf(ARGS(getppid, ""));
1743 #endif
1744
1745 td->td_retval[0] = kern_getppid(td);
1746 return (0);
1747 }
1748
1749 int
linux_getgid(struct thread * td,struct linux_getgid_args * args)1750 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1751 {
1752
1753 #ifdef DEBUG
1754 if (ldebug(getgid))
1755 printf(ARGS(getgid, ""));
1756 #endif
1757
1758 td->td_retval[0] = td->td_ucred->cr_rgid;
1759 return (0);
1760 }
1761
1762 int
linux_getuid(struct thread * td,struct linux_getuid_args * args)1763 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1764 {
1765
1766 #ifdef DEBUG
1767 if (ldebug(getuid))
1768 printf(ARGS(getuid, ""));
1769 #endif
1770
1771 td->td_retval[0] = td->td_ucred->cr_ruid;
1772 return (0);
1773 }
1774
1775
1776 int
linux_getsid(struct thread * td,struct linux_getsid_args * args)1777 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1778 {
1779 struct getsid_args bsd;
1780
1781 #ifdef DEBUG
1782 if (ldebug(getsid))
1783 printf(ARGS(getsid, "%i"), args->pid);
1784 #endif
1785
1786 bsd.pid = args->pid;
1787 return (sys_getsid(td, &bsd));
1788 }
1789
1790 int
linux_nosys(struct thread * td,struct nosys_args * ignore)1791 linux_nosys(struct thread *td, struct nosys_args *ignore)
1792 {
1793
1794 return (ENOSYS);
1795 }
1796
1797 int
linux_getpriority(struct thread * td,struct linux_getpriority_args * args)1798 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1799 {
1800 struct getpriority_args bsd_args;
1801 int error;
1802
1803 #ifdef DEBUG
1804 if (ldebug(getpriority))
1805 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1806 #endif
1807
1808 bsd_args.which = args->which;
1809 bsd_args.who = args->who;
1810 error = sys_getpriority(td, &bsd_args);
1811 td->td_retval[0] = 20 - td->td_retval[0];
1812 return (error);
1813 }
1814
1815 int
linux_sethostname(struct thread * td,struct linux_sethostname_args * args)1816 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1817 {
1818 int name[2];
1819
1820 #ifdef DEBUG
1821 if (ldebug(sethostname))
1822 printf(ARGS(sethostname, "*, %i"), args->len);
1823 #endif
1824
1825 name[0] = CTL_KERN;
1826 name[1] = KERN_HOSTNAME;
1827 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1828 args->len, 0, 0));
1829 }
1830
1831 int
linux_setdomainname(struct thread * td,struct linux_setdomainname_args * args)1832 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1833 {
1834 int name[2];
1835
1836 #ifdef DEBUG
1837 if (ldebug(setdomainname))
1838 printf(ARGS(setdomainname, "*, %i"), args->len);
1839 #endif
1840
1841 name[0] = CTL_KERN;
1842 name[1] = KERN_NISDOMAINNAME;
1843 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1844 args->len, 0, 0));
1845 }
1846
1847 int
linux_exit_group(struct thread * td,struct linux_exit_group_args * args)1848 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1849 {
1850
1851 #ifdef DEBUG
1852 if (ldebug(exit_group))
1853 printf(ARGS(exit_group, "%i"), args->error_code);
1854 #endif
1855
1856 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1857 args->error_code);
1858
1859 /*
1860 * XXX: we should send a signal to the parent if
1861 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1862 * as it doesnt occur often.
1863 */
1864 exit1(td, args->error_code, 0);
1865 /* NOTREACHED */
1866 }
1867
1868 #define _LINUX_CAPABILITY_VERSION_1 0x19980330
1869 #define _LINUX_CAPABILITY_VERSION_2 0x20071026
1870 #define _LINUX_CAPABILITY_VERSION_3 0x20080522
1871
1872 struct l_user_cap_header {
1873 l_int version;
1874 l_int pid;
1875 };
1876
1877 struct l_user_cap_data {
1878 l_int effective;
1879 l_int permitted;
1880 l_int inheritable;
1881 };
1882
1883 int
linux_capget(struct thread * td,struct linux_capget_args * uap)1884 linux_capget(struct thread *td, struct linux_capget_args *uap)
1885 {
1886 struct l_user_cap_header luch;
1887 struct l_user_cap_data lucd[2];
1888 int error, u32s;
1889
1890 if (uap->hdrp == NULL)
1891 return (EFAULT);
1892
1893 error = copyin(uap->hdrp, &luch, sizeof(luch));
1894 if (error != 0)
1895 return (error);
1896
1897 switch (luch.version) {
1898 case _LINUX_CAPABILITY_VERSION_1:
1899 u32s = 1;
1900 break;
1901 case _LINUX_CAPABILITY_VERSION_2:
1902 case _LINUX_CAPABILITY_VERSION_3:
1903 u32s = 2;
1904 break;
1905 default:
1906 #ifdef DEBUG
1907 if (ldebug(capget))
1908 printf(LMSG("invalid capget capability version 0x%x"),
1909 luch.version);
1910 #endif
1911 luch.version = _LINUX_CAPABILITY_VERSION_1;
1912 error = copyout(&luch, uap->hdrp, sizeof(luch));
1913 if (error)
1914 return (error);
1915 return (EINVAL);
1916 }
1917
1918 if (luch.pid)
1919 return (EPERM);
1920
1921 if (uap->datap) {
1922 /*
1923 * The current implementation doesn't support setting
1924 * a capability (it's essentially a stub) so indicate
1925 * that no capabilities are currently set or available
1926 * to request.
1927 */
1928 memset(&lucd, 0, u32s * sizeof(lucd[0]));
1929 error = copyout(&lucd, uap->datap, u32s * sizeof(lucd[0]));
1930 }
1931
1932 return (error);
1933 }
1934
1935 int
linux_capset(struct thread * td,struct linux_capset_args * uap)1936 linux_capset(struct thread *td, struct linux_capset_args *uap)
1937 {
1938 struct l_user_cap_header luch;
1939 struct l_user_cap_data lucd[2];
1940 int error, i, u32s;
1941
1942 if (uap->hdrp == NULL || uap->datap == NULL)
1943 return (EFAULT);
1944
1945 error = copyin(uap->hdrp, &luch, sizeof(luch));
1946 if (error != 0)
1947 return (error);
1948
1949 switch (luch.version) {
1950 case _LINUX_CAPABILITY_VERSION_1:
1951 u32s = 1;
1952 break;
1953 case _LINUX_CAPABILITY_VERSION_2:
1954 case _LINUX_CAPABILITY_VERSION_3:
1955 u32s = 2;
1956 break;
1957 default:
1958 #ifdef DEBUG
1959 if (ldebug(capset))
1960 printf(LMSG("invalid capset capability version 0x%x"),
1961 luch.version);
1962 #endif
1963 luch.version = _LINUX_CAPABILITY_VERSION_1;
1964 error = copyout(&luch, uap->hdrp, sizeof(luch));
1965 if (error)
1966 return (error);
1967 return (EINVAL);
1968 }
1969
1970 if (luch.pid)
1971 return (EPERM);
1972
1973 error = copyin(uap->datap, &lucd, u32s * sizeof(lucd[0]));
1974 if (error != 0)
1975 return (error);
1976
1977 /* We currently don't support setting any capabilities. */
1978 for (i = 0; i < u32s; i++) {
1979 if (lucd[i].effective || lucd[i].permitted ||
1980 lucd[i].inheritable) {
1981 linux_msg(td,
1982 "capset[%d] effective=0x%x, permitted=0x%x, "
1983 "inheritable=0x%x is not implemented", i,
1984 (int)lucd[i].effective, (int)lucd[i].permitted,
1985 (int)lucd[i].inheritable);
1986 return (EPERM);
1987 }
1988 }
1989
1990 return (0);
1991 }
1992
1993 int
linux_prctl(struct thread * td,struct linux_prctl_args * args)1994 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1995 {
1996 int error = 0, max_size;
1997 struct proc *p = td->td_proc;
1998 char comm[LINUX_MAX_COMM_LEN];
1999 int pdeath_signal;
2000
2001 #ifdef DEBUG
2002 if (ldebug(prctl))
2003 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
2004 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
2005 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
2006 #endif
2007
2008 switch (args->option) {
2009 case LINUX_PR_SET_PDEATHSIG:
2010 if (!LINUX_SIG_VALID(args->arg2))
2011 return (EINVAL);
2012 pdeath_signal = linux_to_bsd_signal(args->arg2);
2013 return (kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_CTL,
2014 &pdeath_signal));
2015 case LINUX_PR_GET_PDEATHSIG:
2016 error = kern_procctl(td, P_PID, 0, PROC_PDEATHSIG_STATUS,
2017 &pdeath_signal);
2018 if (error != 0)
2019 return (error);
2020 pdeath_signal = bsd_to_linux_signal(pdeath_signal);
2021 return (copyout(&pdeath_signal,
2022 (void *)(register_t)args->arg2,
2023 sizeof(pdeath_signal)));
2024 break;
2025 case LINUX_PR_GET_KEEPCAPS:
2026 /*
2027 * Indicate that we always clear the effective and
2028 * permitted capability sets when the user id becomes
2029 * non-zero (actually the capability sets are simply
2030 * always zero in the current implementation).
2031 */
2032 td->td_retval[0] = 0;
2033 break;
2034 case LINUX_PR_SET_KEEPCAPS:
2035 /*
2036 * Ignore requests to keep the effective and permitted
2037 * capability sets when the user id becomes non-zero.
2038 */
2039 break;
2040 case LINUX_PR_SET_NAME:
2041 /*
2042 * To be on the safe side we need to make sure to not
2043 * overflow the size a Linux program expects. We already
2044 * do this here in the copyin, so that we don't need to
2045 * check on copyout.
2046 */
2047 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2048 error = copyinstr((void *)(register_t)args->arg2, comm,
2049 max_size, NULL);
2050
2051 /* Linux silently truncates the name if it is too long. */
2052 if (error == ENAMETOOLONG) {
2053 /*
2054 * XXX: copyinstr() isn't documented to populate the
2055 * array completely, so do a copyin() to be on the
2056 * safe side. This should be changed in case
2057 * copyinstr() is changed to guarantee this.
2058 */
2059 error = copyin((void *)(register_t)args->arg2, comm,
2060 max_size - 1);
2061 comm[max_size - 1] = '\0';
2062 }
2063 if (error)
2064 return (error);
2065
2066 PROC_LOCK(p);
2067 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2068 PROC_UNLOCK(p);
2069 break;
2070 case LINUX_PR_GET_NAME:
2071 PROC_LOCK(p);
2072 strlcpy(comm, p->p_comm, sizeof(comm));
2073 PROC_UNLOCK(p);
2074 error = copyout(comm, (void *)(register_t)args->arg2,
2075 strlen(comm) + 1);
2076 break;
2077 default:
2078 error = EINVAL;
2079 break;
2080 }
2081
2082 return (error);
2083 }
2084
2085 int
linux_sched_setparam(struct thread * td,struct linux_sched_setparam_args * uap)2086 linux_sched_setparam(struct thread *td,
2087 struct linux_sched_setparam_args *uap)
2088 {
2089 struct sched_param sched_param;
2090 struct thread *tdt;
2091 int error;
2092
2093 #ifdef DEBUG
2094 if (ldebug(sched_setparam))
2095 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2096 #endif
2097
2098 error = copyin(uap->param, &sched_param, sizeof(sched_param));
2099 if (error)
2100 return (error);
2101
2102 tdt = linux_tdfind(td, uap->pid, -1);
2103 if (tdt == NULL)
2104 return (ESRCH);
2105
2106 error = kern_sched_setparam(td, tdt, &sched_param);
2107 PROC_UNLOCK(tdt->td_proc);
2108 return (error);
2109 }
2110
2111 int
linux_sched_getparam(struct thread * td,struct linux_sched_getparam_args * uap)2112 linux_sched_getparam(struct thread *td,
2113 struct linux_sched_getparam_args *uap)
2114 {
2115 struct sched_param sched_param;
2116 struct thread *tdt;
2117 int error;
2118
2119 #ifdef DEBUG
2120 if (ldebug(sched_getparam))
2121 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2122 #endif
2123
2124 tdt = linux_tdfind(td, uap->pid, -1);
2125 if (tdt == NULL)
2126 return (ESRCH);
2127
2128 error = kern_sched_getparam(td, tdt, &sched_param);
2129 PROC_UNLOCK(tdt->td_proc);
2130 if (error == 0)
2131 error = copyout(&sched_param, uap->param,
2132 sizeof(sched_param));
2133 return (error);
2134 }
2135
2136 /*
2137 * Get affinity of a process.
2138 */
2139 int
linux_sched_getaffinity(struct thread * td,struct linux_sched_getaffinity_args * args)2140 linux_sched_getaffinity(struct thread *td,
2141 struct linux_sched_getaffinity_args *args)
2142 {
2143 int error;
2144 struct thread *tdt;
2145
2146 #ifdef DEBUG
2147 if (ldebug(sched_getaffinity))
2148 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2149 args->len);
2150 #endif
2151 if (args->len < sizeof(cpuset_t))
2152 return (EINVAL);
2153
2154 tdt = linux_tdfind(td, args->pid, -1);
2155 if (tdt == NULL)
2156 return (ESRCH);
2157
2158 PROC_UNLOCK(tdt->td_proc);
2159
2160 error = kern_cpuset_getaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2161 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *)args->user_mask_ptr);
2162 if (error == 0)
2163 td->td_retval[0] = sizeof(cpuset_t);
2164
2165 return (error);
2166 }
2167
2168 /*
2169 * Set affinity of a process.
2170 */
2171 int
linux_sched_setaffinity(struct thread * td,struct linux_sched_setaffinity_args * args)2172 linux_sched_setaffinity(struct thread *td,
2173 struct linux_sched_setaffinity_args *args)
2174 {
2175 struct thread *tdt;
2176
2177 #ifdef DEBUG
2178 if (ldebug(sched_setaffinity))
2179 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2180 args->len);
2181 #endif
2182 if (args->len < sizeof(cpuset_t))
2183 return (EINVAL);
2184
2185 tdt = linux_tdfind(td, args->pid, -1);
2186 if (tdt == NULL)
2187 return (ESRCH);
2188
2189 PROC_UNLOCK(tdt->td_proc);
2190
2191 return (kern_cpuset_setaffinity(td, CPU_LEVEL_WHICH, CPU_WHICH_TID,
2192 tdt->td_tid, sizeof(cpuset_t), (cpuset_t *) args->user_mask_ptr));
2193 }
2194
2195 struct linux_rlimit64 {
2196 uint64_t rlim_cur;
2197 uint64_t rlim_max;
2198 };
2199
2200 int
linux_prlimit64(struct thread * td,struct linux_prlimit64_args * args)2201 linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2202 {
2203 struct rlimit rlim, nrlim;
2204 struct linux_rlimit64 lrlim;
2205 struct proc *p;
2206 u_int which;
2207 int flags;
2208 int error;
2209
2210 #ifdef DEBUG
2211 if (ldebug(prlimit64))
2212 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2213 args->resource, (void *)args->new, (void *)args->old);
2214 #endif
2215
2216 if (args->resource >= LINUX_RLIM_NLIMITS)
2217 return (EINVAL);
2218
2219 which = linux_to_bsd_resource[args->resource];
2220 if (which == -1)
2221 return (EINVAL);
2222
2223 if (args->new != NULL) {
2224 /*
2225 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2226 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2227 * as INFINITY so we do not need a conversion even.
2228 */
2229 error = copyin(args->new, &nrlim, sizeof(nrlim));
2230 if (error != 0)
2231 return (error);
2232 }
2233
2234 flags = PGET_HOLD | PGET_NOTWEXIT;
2235 if (args->new != NULL)
2236 flags |= PGET_CANDEBUG;
2237 else
2238 flags |= PGET_CANSEE;
2239 error = pget(args->pid, flags, &p);
2240 if (error != 0)
2241 return (error);
2242
2243 if (args->old != NULL) {
2244 PROC_LOCK(p);
2245 lim_rlimit_proc(p, which, &rlim);
2246 PROC_UNLOCK(p);
2247 if (rlim.rlim_cur == RLIM_INFINITY)
2248 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2249 else
2250 lrlim.rlim_cur = rlim.rlim_cur;
2251 if (rlim.rlim_max == RLIM_INFINITY)
2252 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2253 else
2254 lrlim.rlim_max = rlim.rlim_max;
2255 error = copyout(&lrlim, args->old, sizeof(lrlim));
2256 if (error != 0)
2257 goto out;
2258 }
2259
2260 if (args->new != NULL)
2261 error = kern_proc_setrlimit(td, p, which, &nrlim);
2262
2263 out:
2264 PRELE(p);
2265 return (error);
2266 }
2267
2268 int
linux_pselect6(struct thread * td,struct linux_pselect6_args * args)2269 linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2270 {
2271 struct timeval utv, tv0, tv1, *tvp;
2272 struct l_pselect6arg lpse6;
2273 struct l_timespec lts;
2274 struct timespec uts;
2275 l_sigset_t l_ss;
2276 sigset_t *ssp;
2277 sigset_t ss;
2278 int error;
2279
2280 ssp = NULL;
2281 if (args->sig != NULL) {
2282 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2283 if (error != 0)
2284 return (error);
2285 if (lpse6.ss_len != sizeof(l_ss))
2286 return (EINVAL);
2287 if (lpse6.ss != 0) {
2288 error = copyin(PTRIN(lpse6.ss), &l_ss,
2289 sizeof(l_ss));
2290 if (error != 0)
2291 return (error);
2292 linux_to_bsd_sigset(&l_ss, &ss);
2293 ssp = &ss;
2294 }
2295 }
2296
2297 /*
2298 * Currently glibc changes nanosecond number to microsecond.
2299 * This mean losing precision but for now it is hardly seen.
2300 */
2301 if (args->tsp != NULL) {
2302 error = copyin(args->tsp, <s, sizeof(lts));
2303 if (error != 0)
2304 return (error);
2305 error = linux_to_native_timespec(&uts, <s);
2306 if (error != 0)
2307 return (error);
2308
2309 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2310 if (itimerfix(&utv))
2311 return (EINVAL);
2312
2313 microtime(&tv0);
2314 tvp = &utv;
2315 } else
2316 tvp = NULL;
2317
2318 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2319 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2320
2321 if (error == 0 && args->tsp != NULL) {
2322 if (td->td_retval[0] != 0) {
2323 /*
2324 * Compute how much time was left of the timeout,
2325 * by subtracting the current time and the time
2326 * before we started the call, and subtracting
2327 * that result from the user-supplied value.
2328 */
2329
2330 microtime(&tv1);
2331 timevalsub(&tv1, &tv0);
2332 timevalsub(&utv, &tv1);
2333 if (utv.tv_sec < 0)
2334 timevalclear(&utv);
2335 } else
2336 timevalclear(&utv);
2337
2338 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2339
2340 error = native_to_linux_timespec(<s, &uts);
2341 if (error == 0)
2342 error = copyout(<s, args->tsp, sizeof(lts));
2343 }
2344
2345 return (error);
2346 }
2347
2348 int
linux_ppoll(struct thread * td,struct linux_ppoll_args * args)2349 linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2350 {
2351 struct timespec ts0, ts1;
2352 struct l_timespec lts;
2353 struct timespec uts, *tsp;
2354 l_sigset_t l_ss;
2355 sigset_t *ssp;
2356 sigset_t ss;
2357 int error;
2358
2359 if (args->sset != NULL) {
2360 if (args->ssize != sizeof(l_ss))
2361 return (EINVAL);
2362 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2363 if (error)
2364 return (error);
2365 linux_to_bsd_sigset(&l_ss, &ss);
2366 ssp = &ss;
2367 } else
2368 ssp = NULL;
2369 if (args->tsp != NULL) {
2370 error = copyin(args->tsp, <s, sizeof(lts));
2371 if (error)
2372 return (error);
2373 error = linux_to_native_timespec(&uts, <s);
2374 if (error != 0)
2375 return (error);
2376
2377 nanotime(&ts0);
2378 tsp = &uts;
2379 } else
2380 tsp = NULL;
2381
2382 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2383
2384 if (error == 0 && args->tsp != NULL) {
2385 if (td->td_retval[0]) {
2386 nanotime(&ts1);
2387 timespecsub(&ts1, &ts0);
2388 timespecsub(&uts, &ts1);
2389 if (uts.tv_sec < 0)
2390 timespecclear(&uts);
2391 } else
2392 timespecclear(&uts);
2393
2394 error = native_to_linux_timespec(<s, &uts);
2395 if (error == 0)
2396 error = copyout(<s, args->tsp, sizeof(lts));
2397 }
2398
2399 return (error);
2400 }
2401
2402 #if defined(DEBUG) || defined(KTR)
2403 /* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2404
2405 #ifdef COMPAT_LINUX32
2406 #define L_MAXSYSCALL LINUX32_SYS_MAXSYSCALL
2407 #else
2408 #define L_MAXSYSCALL LINUX_SYS_MAXSYSCALL
2409 #endif
2410
2411 u_char linux_debug_map[howmany(L_MAXSYSCALL, sizeof(u_char))];
2412
2413 static int
linux_debug(int syscall,int toggle,int global)2414 linux_debug(int syscall, int toggle, int global)
2415 {
2416
2417 if (global) {
2418 char c = toggle ? 0 : 0xff;
2419
2420 memset(linux_debug_map, c, sizeof(linux_debug_map));
2421 return (0);
2422 }
2423 if (syscall < 0 || syscall >= L_MAXSYSCALL)
2424 return (EINVAL);
2425 if (toggle)
2426 clrbit(linux_debug_map, syscall);
2427 else
2428 setbit(linux_debug_map, syscall);
2429 return (0);
2430 }
2431 #undef L_MAXSYSCALL
2432
2433 /*
2434 * Usage: sysctl linux.debug=<syscall_nr>.<0/1>
2435 *
2436 * E.g.: sysctl linux.debug=21.0
2437 *
2438 * As a special case, syscall "all" will apply to all syscalls globally.
2439 */
2440 #define LINUX_MAX_DEBUGSTR 16
2441 int
linux_sysctl_debug(SYSCTL_HANDLER_ARGS)2442 linux_sysctl_debug(SYSCTL_HANDLER_ARGS)
2443 {
2444 char value[LINUX_MAX_DEBUGSTR], *p;
2445 int error, sysc, toggle;
2446 int global = 0;
2447
2448 value[0] = '\0';
2449 error = sysctl_handle_string(oidp, value, LINUX_MAX_DEBUGSTR, req);
2450 if (error || req->newptr == NULL)
2451 return (error);
2452 for (p = value; *p != '\0' && *p != '.'; p++);
2453 if (*p == '\0')
2454 return (EINVAL);
2455 *p++ = '\0';
2456 sysc = strtol(value, NULL, 0);
2457 toggle = strtol(p, NULL, 0);
2458 if (strcmp(value, "all") == 0)
2459 global = 1;
2460 error = linux_debug(sysc, toggle, global);
2461 return (error);
2462 }
2463
2464 #endif /* DEBUG || KTR */
2465
2466 int
linux_sched_rr_get_interval(struct thread * td,struct linux_sched_rr_get_interval_args * uap)2467 linux_sched_rr_get_interval(struct thread *td,
2468 struct linux_sched_rr_get_interval_args *uap)
2469 {
2470 struct timespec ts;
2471 struct l_timespec lts;
2472 struct thread *tdt;
2473 int error;
2474
2475 /*
2476 * According to man in case the invalid pid specified
2477 * EINVAL should be returned.
2478 */
2479 if (uap->pid < 0)
2480 return (EINVAL);
2481
2482 tdt = linux_tdfind(td, uap->pid, -1);
2483 if (tdt == NULL)
2484 return (ESRCH);
2485
2486 error = kern_sched_rr_get_interval_td(td, tdt, &ts);
2487 PROC_UNLOCK(tdt->td_proc);
2488 if (error != 0)
2489 return (error);
2490 error = native_to_linux_timespec(<s, &ts);
2491 if (error != 0)
2492 return (error);
2493 return (copyout(<s, uap->interval, sizeof(lts)));
2494 }
2495
2496 /*
2497 * In case when the Linux thread is the initial thread in
2498 * the thread group thread id is equal to the process id.
2499 * Glibc depends on this magic (assert in pthread_getattr_np.c).
2500 */
2501 struct thread *
linux_tdfind(struct thread * td,lwpid_t tid,pid_t pid)2502 linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid)
2503 {
2504 struct linux_emuldata *em;
2505 struct thread *tdt;
2506 struct proc *p;
2507
2508 tdt = NULL;
2509 if (tid == 0 || tid == td->td_tid) {
2510 tdt = td;
2511 PROC_LOCK(tdt->td_proc);
2512 } else if (tid > PID_MAX)
2513 tdt = tdfind(tid, pid);
2514 else {
2515 /*
2516 * Initial thread where the tid equal to the pid.
2517 */
2518 p = pfind(tid);
2519 if (p != NULL) {
2520 if (SV_PROC_ABI(p) != SV_ABI_LINUX) {
2521 /*
2522 * p is not a Linuxulator process.
2523 */
2524 PROC_UNLOCK(p);
2525 return (NULL);
2526 }
2527 FOREACH_THREAD_IN_PROC(p, tdt) {
2528 em = em_find(tdt);
2529 if (tid == em->em_tid)
2530 return (tdt);
2531 }
2532 PROC_UNLOCK(p);
2533 }
2534 return (NULL);
2535 }
2536
2537 return (tdt);
2538 }
2539
2540 void
linux_to_bsd_waitopts(int options,int * bsdopts)2541 linux_to_bsd_waitopts(int options, int *bsdopts)
2542 {
2543
2544 if (options & LINUX_WNOHANG)
2545 *bsdopts |= WNOHANG;
2546 if (options & LINUX_WUNTRACED)
2547 *bsdopts |= WUNTRACED;
2548 if (options & LINUX_WEXITED)
2549 *bsdopts |= WEXITED;
2550 if (options & LINUX_WCONTINUED)
2551 *bsdopts |= WCONTINUED;
2552 if (options & LINUX_WNOWAIT)
2553 *bsdopts |= WNOWAIT;
2554
2555 if (options & __WCLONE)
2556 *bsdopts |= WLINUXCLONE;
2557 }
2558
2559 int
linux_getrandom(struct thread * td,struct linux_getrandom_args * args)2560 linux_getrandom(struct thread *td, struct linux_getrandom_args *args)
2561 {
2562 struct uio uio;
2563 struct iovec iov;
2564 int error;
2565
2566 if (args->flags & ~(LINUX_GRND_NONBLOCK|LINUX_GRND_RANDOM))
2567 return (EINVAL);
2568 if (args->count > INT_MAX)
2569 args->count = INT_MAX;
2570
2571 iov.iov_base = args->buf;
2572 iov.iov_len = args->count;
2573
2574 uio.uio_iov = &iov;
2575 uio.uio_iovcnt = 1;
2576 uio.uio_resid = iov.iov_len;
2577 uio.uio_segflg = UIO_USERSPACE;
2578 uio.uio_rw = UIO_READ;
2579 uio.uio_td = td;
2580
2581 error = read_random_uio(&uio, args->flags & LINUX_GRND_NONBLOCK);
2582 if (error == 0)
2583 td->td_retval[0] = args->count - uio.uio_resid;
2584 return (error);
2585 }
2586
2587 int
linux_mincore(struct thread * td,struct linux_mincore_args * args)2588 linux_mincore(struct thread *td, struct linux_mincore_args *args)
2589 {
2590
2591 /* Needs to be page-aligned */
2592 if (args->start & PAGE_MASK)
2593 return (EINVAL);
2594 return (kern_mincore(td, args->start, args->len, args->vec));
2595 }
2596