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: releng/9.2/sys/compat/linux/linux_misc.c 246290 2013-02-03 18:14:37Z dchagin $");
32
33 #include "opt_compat.h"
34 #include "opt_kdtrace.h"
35
36 #include <sys/param.h>
37 #include <sys/blist.h>
38 #include <sys/fcntl.h>
39 #if defined(__i386__)
40 #include <sys/imgact_aout.h>
41 #endif
42 #include <sys/jail.h>
43 #include <sys/kernel.h>
44 #include <sys/limits.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mman.h>
48 #include <sys/mount.h>
49 #include <sys/mutex.h>
50 #include <sys/namei.h>
51 #include <sys/priv.h>
52 #include <sys/proc.h>
53 #include <sys/reboot.h>
54 #include <sys/racct.h>
55 #include <sys/resourcevar.h>
56 #include <sys/sched.h>
57 #include <sys/sdt.h>
58 #include <sys/signalvar.h>
59 #include <sys/stat.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysproto.h>
63 #include <sys/systm.h>
64 #include <sys/time.h>
65 #include <sys/vmmeter.h>
66 #include <sys/vnode.h>
67 #include <sys/wait.h>
68 #include <sys/cpuset.h>
69
70 #include <security/mac/mac_framework.h>
71
72 #include <vm/vm.h>
73 #include <vm/pmap.h>
74 #include <vm/vm_kern.h>
75 #include <vm/vm_map.h>
76 #include <vm/vm_extern.h>
77 #include <vm/vm_object.h>
78 #include <vm/swap_pager.h>
79
80 #ifdef COMPAT_LINUX32
81 #include <machine/../linux32/linux.h>
82 #include <machine/../linux32/linux32_proto.h>
83 #else
84 #include <machine/../linux/linux.h>
85 #include <machine/../linux/linux_proto.h>
86 #endif
87
88 #include <compat/linux/linux_dtrace.h>
89 #include <compat/linux/linux_file.h>
90 #include <compat/linux/linux_mib.h>
91 #include <compat/linux/linux_signal.h>
92 #include <compat/linux/linux_util.h>
93 #include <compat/linux/linux_sysproto.h>
94 #include <compat/linux/linux_emul.h>
95 #include <compat/linux/linux_misc.h>
96
97 /* DTrace init */
98 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
99
100 /* Linuxulator-global DTrace probes */
101 LIN_SDT_PROBE_DECLARE(locks, emul_lock, locked);
102 LIN_SDT_PROBE_DECLARE(locks, emul_lock, unlock);
103 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, locked);
104 LIN_SDT_PROBE_DECLARE(locks, emul_shared_rlock, unlock);
105 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, locked);
106 LIN_SDT_PROBE_DECLARE(locks, emul_shared_wlock, unlock);
107
108 int stclohz; /* Statistics clock frequency */
109
110 static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
111 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
112 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
113 RLIMIT_MEMLOCK, RLIMIT_AS
114 };
115
116 struct l_sysinfo {
117 l_long uptime; /* Seconds since boot */
118 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
119 #define LINUX_SYSINFO_LOADS_SCALE 65536
120 l_ulong totalram; /* Total usable main memory size */
121 l_ulong freeram; /* Available memory size */
122 l_ulong sharedram; /* Amount of shared memory */
123 l_ulong bufferram; /* Memory used by buffers */
124 l_ulong totalswap; /* Total swap space size */
125 l_ulong freeswap; /* swap space still available */
126 l_ushort procs; /* Number of current processes */
127 l_ushort pads;
128 l_ulong totalbig;
129 l_ulong freebig;
130 l_uint mem_unit;
131 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
132 };
133 int
134 linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
135 {
136 struct l_sysinfo sysinfo;
137 vm_object_t object;
138 int i, j;
139 struct timespec ts;
140
141 getnanouptime(&ts);
142 if (ts.tv_nsec != 0)
143 ts.tv_sec++;
144 sysinfo.uptime = ts.tv_sec;
145
146 /* Use the information from the mib to get our load averages */
147 for (i = 0; i < 3; i++)
148 sysinfo.loads[i] = averunnable.ldavg[i] *
149 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
150
151 sysinfo.totalram = physmem * PAGE_SIZE;
152 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
153
154 sysinfo.sharedram = 0;
155 mtx_lock(&vm_object_list_mtx);
156 TAILQ_FOREACH(object, &vm_object_list, object_list)
157 if (object->shadow_count > 1)
158 sysinfo.sharedram += object->resident_page_count;
159 mtx_unlock(&vm_object_list_mtx);
160
161 sysinfo.sharedram *= PAGE_SIZE;
162 sysinfo.bufferram = 0;
163
164 swap_pager_status(&i, &j);
165 sysinfo.totalswap = i * PAGE_SIZE;
166 sysinfo.freeswap = (i - j) * PAGE_SIZE;
167
168 sysinfo.procs = nprocs;
169
170 /* The following are only present in newer Linux kernels. */
171 sysinfo.totalbig = 0;
172 sysinfo.freebig = 0;
173 sysinfo.mem_unit = 1;
174
175 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
176 }
177
178 int
179 linux_alarm(struct thread *td, struct linux_alarm_args *args)
180 {
181 struct itimerval it, old_it;
182 u_int secs;
183 int error;
184
185 #ifdef DEBUG
186 if (ldebug(alarm))
187 printf(ARGS(alarm, "%u"), args->secs);
188 #endif
189
190 secs = args->secs;
191
192 if (secs > INT_MAX)
193 secs = INT_MAX;
194
195 it.it_value.tv_sec = (long) secs;
196 it.it_value.tv_usec = 0;
197 it.it_interval.tv_sec = 0;
198 it.it_interval.tv_usec = 0;
199 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
200 if (error)
201 return (error);
202 if (timevalisset(&old_it.it_value)) {
203 if (old_it.it_value.tv_usec != 0)
204 old_it.it_value.tv_sec++;
205 td->td_retval[0] = old_it.it_value.tv_sec;
206 }
207 return (0);
208 }
209
210 int
211 linux_brk(struct thread *td, struct linux_brk_args *args)
212 {
213 struct vmspace *vm = td->td_proc->p_vmspace;
214 vm_offset_t new, old;
215 struct obreak_args /* {
216 char * nsize;
217 } */ tmp;
218
219 #ifdef DEBUG
220 if (ldebug(brk))
221 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
222 #endif
223 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
224 new = (vm_offset_t)args->dsend;
225 tmp.nsize = (char *)new;
226 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
227 td->td_retval[0] = (long)new;
228 else
229 td->td_retval[0] = (long)old;
230
231 return (0);
232 }
233
234 #if defined(__i386__)
235 /* XXX: what about amd64/linux32? */
236
237 int
238 linux_uselib(struct thread *td, struct linux_uselib_args *args)
239 {
240 struct nameidata ni;
241 struct vnode *vp;
242 struct exec *a_out;
243 struct vattr attr;
244 vm_offset_t vmaddr;
245 unsigned long file_offset;
246 unsigned long bss_size;
247 char *library;
248 ssize_t aresid;
249 int error, locked, vfslocked, writecount;
250
251 LCONVPATHEXIST(td, args->library, &library);
252
253 #ifdef DEBUG
254 if (ldebug(uselib))
255 printf(ARGS(uselib, "%s"), library);
256 #endif
257
258 a_out = NULL;
259 vfslocked = 0;
260 locked = 0;
261 vp = NULL;
262
263 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
264 UIO_SYSSPACE, library, td);
265 error = namei(&ni);
266 LFREEPATH(library);
267 if (error)
268 goto cleanup;
269
270 vp = ni.ni_vp;
271 vfslocked = NDHASGIANT(&ni);
272 NDFREE(&ni, NDF_ONLY_PNBUF);
273
274 /*
275 * From here on down, we have a locked vnode that must be unlocked.
276 * XXX: The code below largely duplicates exec_check_permissions().
277 */
278 locked = 1;
279
280 /* Writable? */
281 error = VOP_GET_WRITECOUNT(vp, &writecount);
282 if (error != 0)
283 goto cleanup;
284 if (writecount != 0) {
285 error = ETXTBSY;
286 goto cleanup;
287 }
288
289 /* Executable? */
290 error = VOP_GETATTR(vp, &attr, td->td_ucred);
291 if (error)
292 goto cleanup;
293
294 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
295 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
296 /* EACCESS is what exec(2) returns. */
297 error = ENOEXEC;
298 goto cleanup;
299 }
300
301 /* Sensible size? */
302 if (attr.va_size == 0) {
303 error = ENOEXEC;
304 goto cleanup;
305 }
306
307 /* Can we access it? */
308 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
309 if (error)
310 goto cleanup;
311
312 /*
313 * XXX: This should use vn_open() so that it is properly authorized,
314 * and to reduce code redundancy all over the place here.
315 * XXX: Not really, it duplicates far more of exec_check_permissions()
316 * than vn_open().
317 */
318 #ifdef MAC
319 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
320 if (error)
321 goto cleanup;
322 #endif
323 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
324 if (error)
325 goto cleanup;
326
327 /* Pull in executable header into exec_map */
328 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
329 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
330 if (error)
331 goto cleanup;
332
333 /* Is it a Linux binary ? */
334 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
335 error = ENOEXEC;
336 goto cleanup;
337 }
338
339 /*
340 * While we are here, we should REALLY do some more checks
341 */
342
343 /* Set file/virtual offset based on a.out variant. */
344 switch ((int)(a_out->a_magic & 0xffff)) {
345 case 0413: /* ZMAGIC */
346 file_offset = 1024;
347 break;
348 case 0314: /* QMAGIC */
349 file_offset = 0;
350 break;
351 default:
352 error = ENOEXEC;
353 goto cleanup;
354 }
355
356 bss_size = round_page(a_out->a_bss);
357
358 /* Check various fields in header for validity/bounds. */
359 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
360 error = ENOEXEC;
361 goto cleanup;
362 }
363
364 /* text + data can't exceed file size */
365 if (a_out->a_data + a_out->a_text > attr.va_size) {
366 error = EFAULT;
367 goto cleanup;
368 }
369
370 /*
371 * text/data/bss must not exceed limits
372 * XXX - this is not complete. it should check current usage PLUS
373 * the resources needed by this library.
374 */
375 PROC_LOCK(td->td_proc);
376 if (a_out->a_text > maxtsiz ||
377 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
378 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
379 bss_size) != 0) {
380 PROC_UNLOCK(td->td_proc);
381 error = ENOMEM;
382 goto cleanup;
383 }
384 PROC_UNLOCK(td->td_proc);
385
386 /*
387 * Prevent more writers.
388 * XXX: Note that if any of the VM operations fail below we don't
389 * clear this flag.
390 */
391 VOP_SET_TEXT(vp);
392
393 /*
394 * Lock no longer needed
395 */
396 locked = 0;
397 VOP_UNLOCK(vp, 0);
398 VFS_UNLOCK_GIANT(vfslocked);
399
400 /*
401 * Check if file_offset page aligned. Currently we cannot handle
402 * misalinged file offsets, and so we read in the entire image
403 * (what a waste).
404 */
405 if (file_offset & PAGE_MASK) {
406 #ifdef DEBUG
407 printf("uselib: Non page aligned binary %lu\n", file_offset);
408 #endif
409 /* Map text+data read/write/execute */
410
411 /* a_entry is the load address and is page aligned */
412 vmaddr = trunc_page(a_out->a_entry);
413
414 /* get anon user mapping, read+write+execute */
415 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
416 &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
417 VM_PROT_ALL, 0);
418 if (error)
419 goto cleanup;
420
421 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
422 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
423 td->td_ucred, NOCRED, &aresid, td);
424 if (error != 0)
425 goto cleanup;
426 if (aresid != 0) {
427 error = ENOEXEC;
428 goto cleanup;
429 }
430 } else {
431 #ifdef DEBUG
432 printf("uselib: Page aligned binary %lu\n", file_offset);
433 #endif
434 /*
435 * for QMAGIC, a_entry is 20 bytes beyond the load address
436 * to skip the executable header
437 */
438 vmaddr = trunc_page(a_out->a_entry);
439
440 /*
441 * Map it all into the process's space as a single
442 * copy-on-write "data" segment.
443 */
444 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
445 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
446 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
447 if (error)
448 goto cleanup;
449 }
450 #ifdef DEBUG
451 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
452 ((long *)vmaddr)[1]);
453 #endif
454 if (bss_size != 0) {
455 /* Calculate BSS start address */
456 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
457 a_out->a_data;
458
459 /* allocate some 'anon' space */
460 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
461 &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
462 if (error)
463 goto cleanup;
464 }
465
466 cleanup:
467 /* Unlock vnode if needed */
468 if (locked) {
469 VOP_UNLOCK(vp, 0);
470 VFS_UNLOCK_GIANT(vfslocked);
471 }
472
473 /* Release the temporary mapping. */
474 if (a_out)
475 kmem_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
476
477 return (error);
478 }
479
480 #endif /* __i386__ */
481
482 int
483 linux_select(struct thread *td, struct linux_select_args *args)
484 {
485 l_timeval ltv;
486 struct timeval tv0, tv1, utv, *tvp;
487 int error;
488
489 #ifdef DEBUG
490 if (ldebug(select))
491 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
492 (void *)args->readfds, (void *)args->writefds,
493 (void *)args->exceptfds, (void *)args->timeout);
494 #endif
495
496 /*
497 * Store current time for computation of the amount of
498 * time left.
499 */
500 if (args->timeout) {
501 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
502 goto select_out;
503 utv.tv_sec = ltv.tv_sec;
504 utv.tv_usec = ltv.tv_usec;
505 #ifdef DEBUG
506 if (ldebug(select))
507 printf(LMSG("incoming timeout (%jd/%ld)"),
508 (intmax_t)utv.tv_sec, utv.tv_usec);
509 #endif
510
511 if (itimerfix(&utv)) {
512 /*
513 * The timeval was invalid. Convert it to something
514 * valid that will act as it does under Linux.
515 */
516 utv.tv_sec += utv.tv_usec / 1000000;
517 utv.tv_usec %= 1000000;
518 if (utv.tv_usec < 0) {
519 utv.tv_sec -= 1;
520 utv.tv_usec += 1000000;
521 }
522 if (utv.tv_sec < 0)
523 timevalclear(&utv);
524 }
525 microtime(&tv0);
526 tvp = &utv;
527 } else
528 tvp = NULL;
529
530 error = kern_select(td, args->nfds, args->readfds, args->writefds,
531 args->exceptfds, tvp, sizeof(l_int) * 8);
532
533 #ifdef DEBUG
534 if (ldebug(select))
535 printf(LMSG("real select returns %d"), error);
536 #endif
537 if (error)
538 goto select_out;
539
540 if (args->timeout) {
541 if (td->td_retval[0]) {
542 /*
543 * Compute how much time was left of the timeout,
544 * by subtracting the current time and the time
545 * before we started the call, and subtracting
546 * that result from the user-supplied value.
547 */
548 microtime(&tv1);
549 timevalsub(&tv1, &tv0);
550 timevalsub(&utv, &tv1);
551 if (utv.tv_sec < 0)
552 timevalclear(&utv);
553 } else
554 timevalclear(&utv);
555 #ifdef DEBUG
556 if (ldebug(select))
557 printf(LMSG("outgoing timeout (%jd/%ld)"),
558 (intmax_t)utv.tv_sec, utv.tv_usec);
559 #endif
560 ltv.tv_sec = utv.tv_sec;
561 ltv.tv_usec = utv.tv_usec;
562 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
563 goto select_out;
564 }
565
566 select_out:
567 #ifdef DEBUG
568 if (ldebug(select))
569 printf(LMSG("select_out -> %d"), error);
570 #endif
571 return (error);
572 }
573
574 int
575 linux_mremap(struct thread *td, struct linux_mremap_args *args)
576 {
577 struct munmap_args /* {
578 void *addr;
579 size_t len;
580 } */ bsd_args;
581 int error = 0;
582
583 #ifdef DEBUG
584 if (ldebug(mremap))
585 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
586 (void *)(uintptr_t)args->addr,
587 (unsigned long)args->old_len,
588 (unsigned long)args->new_len,
589 (unsigned long)args->flags);
590 #endif
591
592 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
593 td->td_retval[0] = 0;
594 return (EINVAL);
595 }
596
597 /*
598 * Check for the page alignment.
599 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
600 */
601 if (args->addr & PAGE_MASK) {
602 td->td_retval[0] = 0;
603 return (EINVAL);
604 }
605
606 args->new_len = round_page(args->new_len);
607 args->old_len = round_page(args->old_len);
608
609 if (args->new_len > args->old_len) {
610 td->td_retval[0] = 0;
611 return (ENOMEM);
612 }
613
614 if (args->new_len < args->old_len) {
615 bsd_args.addr =
616 (caddr_t)((uintptr_t)args->addr + args->new_len);
617 bsd_args.len = args->old_len - args->new_len;
618 error = sys_munmap(td, &bsd_args);
619 }
620
621 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
622 return (error);
623 }
624
625 #define LINUX_MS_ASYNC 0x0001
626 #define LINUX_MS_INVALIDATE 0x0002
627 #define LINUX_MS_SYNC 0x0004
628
629 int
630 linux_msync(struct thread *td, struct linux_msync_args *args)
631 {
632 struct msync_args bsd_args;
633
634 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
635 bsd_args.len = (uintptr_t)args->len;
636 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
637
638 return (sys_msync(td, &bsd_args));
639 }
640
641 int
642 linux_time(struct thread *td, struct linux_time_args *args)
643 {
644 struct timeval tv;
645 l_time_t tm;
646 int error;
647
648 #ifdef DEBUG
649 if (ldebug(time))
650 printf(ARGS(time, "*"));
651 #endif
652
653 microtime(&tv);
654 tm = tv.tv_sec;
655 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
656 return (error);
657 td->td_retval[0] = tm;
658 return (0);
659 }
660
661 struct l_times_argv {
662 l_clock_t tms_utime;
663 l_clock_t tms_stime;
664 l_clock_t tms_cutime;
665 l_clock_t tms_cstime;
666 };
667
668
669 /*
670 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
671 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
672 * auxiliary vector entry.
673 */
674 #define CLK_TCK 100
675
676 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
677 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
678
679 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
680 CONVNTCK(r) : CONVOTCK(r))
681
682 int
683 linux_times(struct thread *td, struct linux_times_args *args)
684 {
685 struct timeval tv, utime, stime, cutime, cstime;
686 struct l_times_argv tms;
687 struct proc *p;
688 int error;
689
690 #ifdef DEBUG
691 if (ldebug(times))
692 printf(ARGS(times, "*"));
693 #endif
694
695 if (args->buf != NULL) {
696 p = td->td_proc;
697 PROC_LOCK(p);
698 PROC_SLOCK(p);
699 calcru(p, &utime, &stime);
700 PROC_SUNLOCK(p);
701 calccru(p, &cutime, &cstime);
702 PROC_UNLOCK(p);
703
704 tms.tms_utime = CONVTCK(utime);
705 tms.tms_stime = CONVTCK(stime);
706
707 tms.tms_cutime = CONVTCK(cutime);
708 tms.tms_cstime = CONVTCK(cstime);
709
710 if ((error = copyout(&tms, args->buf, sizeof(tms))))
711 return (error);
712 }
713
714 microuptime(&tv);
715 td->td_retval[0] = (int)CONVTCK(tv);
716 return (0);
717 }
718
719 int
720 linux_newuname(struct thread *td, struct linux_newuname_args *args)
721 {
722 struct l_new_utsname utsname;
723 char osname[LINUX_MAX_UTSNAME];
724 char osrelease[LINUX_MAX_UTSNAME];
725 char *p;
726
727 #ifdef DEBUG
728 if (ldebug(newuname))
729 printf(ARGS(newuname, "*"));
730 #endif
731
732 linux_get_osname(td, osname);
733 linux_get_osrelease(td, osrelease);
734
735 bzero(&utsname, sizeof(utsname));
736 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
737 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
738 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
739 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
740 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
741 for (p = utsname.version; *p != '\0'; ++p)
742 if (*p == '\n') {
743 *p = '\0';
744 break;
745 }
746 strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
747
748 return (copyout(&utsname, args->buf, sizeof(utsname)));
749 }
750
751 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
752 struct l_utimbuf {
753 l_time_t l_actime;
754 l_time_t l_modtime;
755 };
756
757 int
758 linux_utime(struct thread *td, struct linux_utime_args *args)
759 {
760 struct timeval tv[2], *tvp;
761 struct l_utimbuf lut;
762 char *fname;
763 int error;
764
765 LCONVPATHEXIST(td, args->fname, &fname);
766
767 #ifdef DEBUG
768 if (ldebug(utime))
769 printf(ARGS(utime, "%s, *"), fname);
770 #endif
771
772 if (args->times) {
773 if ((error = copyin(args->times, &lut, sizeof lut))) {
774 LFREEPATH(fname);
775 return (error);
776 }
777 tv[0].tv_sec = lut.l_actime;
778 tv[0].tv_usec = 0;
779 tv[1].tv_sec = lut.l_modtime;
780 tv[1].tv_usec = 0;
781 tvp = tv;
782 } else
783 tvp = NULL;
784
785 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
786 LFREEPATH(fname);
787 return (error);
788 }
789
790 int
791 linux_utimes(struct thread *td, struct linux_utimes_args *args)
792 {
793 l_timeval ltv[2];
794 struct timeval tv[2], *tvp = NULL;
795 char *fname;
796 int error;
797
798 LCONVPATHEXIST(td, args->fname, &fname);
799
800 #ifdef DEBUG
801 if (ldebug(utimes))
802 printf(ARGS(utimes, "%s, *"), fname);
803 #endif
804
805 if (args->tptr != NULL) {
806 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
807 LFREEPATH(fname);
808 return (error);
809 }
810 tv[0].tv_sec = ltv[0].tv_sec;
811 tv[0].tv_usec = ltv[0].tv_usec;
812 tv[1].tv_sec = ltv[1].tv_sec;
813 tv[1].tv_usec = ltv[1].tv_usec;
814 tvp = tv;
815 }
816
817 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
818 LFREEPATH(fname);
819 return (error);
820 }
821
822 int
823 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
824 {
825 l_timeval ltv[2];
826 struct timeval tv[2], *tvp = NULL;
827 char *fname;
828 int error, dfd;
829
830 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
831 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
832
833 #ifdef DEBUG
834 if (ldebug(futimesat))
835 printf(ARGS(futimesat, "%s, *"), fname);
836 #endif
837
838 if (args->utimes != NULL) {
839 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
840 LFREEPATH(fname);
841 return (error);
842 }
843 tv[0].tv_sec = ltv[0].tv_sec;
844 tv[0].tv_usec = ltv[0].tv_usec;
845 tv[1].tv_sec = ltv[1].tv_sec;
846 tv[1].tv_usec = ltv[1].tv_usec;
847 tvp = tv;
848 }
849
850 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
851 LFREEPATH(fname);
852 return (error);
853 }
854 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
855
856 int
857 linux_common_wait(struct thread *td, int pid, int *status,
858 int options, struct rusage *ru)
859 {
860 int error, tmpstat;
861
862 error = kern_wait(td, pid, &tmpstat, options, ru);
863 if (error)
864 return (error);
865
866 if (status) {
867 tmpstat &= 0xffff;
868 if (WIFSIGNALED(tmpstat))
869 tmpstat = (tmpstat & 0xffffff80) |
870 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
871 else if (WIFSTOPPED(tmpstat))
872 tmpstat = (tmpstat & 0xffff00ff) |
873 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
874 error = copyout(&tmpstat, status, sizeof(int));
875 }
876
877 return (error);
878 }
879
880 int
881 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
882 {
883 int options;
884
885 #ifdef DEBUG
886 if (ldebug(waitpid))
887 printf(ARGS(waitpid, "%d, %p, %d"),
888 args->pid, (void *)args->status, args->options);
889 #endif
890 /*
891 * this is necessary because the test in kern_wait doesn't work
892 * because we mess with the options here
893 */
894 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
895 return (EINVAL);
896
897 options = (args->options & (WNOHANG | WUNTRACED));
898 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
899 if (args->options & __WCLONE)
900 options |= WLINUXCLONE;
901
902 return (linux_common_wait(td, args->pid, args->status, options, NULL));
903 }
904
905
906 int
907 linux_mknod(struct thread *td, struct linux_mknod_args *args)
908 {
909 char *path;
910 int error;
911
912 LCONVPATHCREAT(td, args->path, &path);
913
914 #ifdef DEBUG
915 if (ldebug(mknod))
916 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
917 #endif
918
919 switch (args->mode & S_IFMT) {
920 case S_IFIFO:
921 case S_IFSOCK:
922 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
923 break;
924
925 case S_IFCHR:
926 case S_IFBLK:
927 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
928 args->dev);
929 break;
930
931 case S_IFDIR:
932 error = EPERM;
933 break;
934
935 case 0:
936 args->mode |= S_IFREG;
937 /* FALLTHROUGH */
938 case S_IFREG:
939 error = kern_open(td, path, UIO_SYSSPACE,
940 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
941 if (error == 0)
942 kern_close(td, td->td_retval[0]);
943 break;
944
945 default:
946 error = EINVAL;
947 break;
948 }
949 LFREEPATH(path);
950 return (error);
951 }
952
953 int
954 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
955 {
956 char *path;
957 int error, dfd;
958
959 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
960 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
961
962 #ifdef DEBUG
963 if (ldebug(mknodat))
964 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
965 #endif
966
967 switch (args->mode & S_IFMT) {
968 case S_IFIFO:
969 case S_IFSOCK:
970 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
971 break;
972
973 case S_IFCHR:
974 case S_IFBLK:
975 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
976 args->dev);
977 break;
978
979 case S_IFDIR:
980 error = EPERM;
981 break;
982
983 case 0:
984 args->mode |= S_IFREG;
985 /* FALLTHROUGH */
986 case S_IFREG:
987 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
988 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
989 if (error == 0)
990 kern_close(td, td->td_retval[0]);
991 break;
992
993 default:
994 error = EINVAL;
995 break;
996 }
997 LFREEPATH(path);
998 return (error);
999 }
1000
1001 /*
1002 * UGH! This is just about the dumbest idea I've ever heard!!
1003 */
1004 int
1005 linux_personality(struct thread *td, struct linux_personality_args *args)
1006 {
1007 #ifdef DEBUG
1008 if (ldebug(personality))
1009 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1010 #endif
1011 if (args->per != 0)
1012 return (EINVAL);
1013
1014 /* Yes Jim, it's still a Linux... */
1015 td->td_retval[0] = 0;
1016 return (0);
1017 }
1018
1019 struct l_itimerval {
1020 l_timeval it_interval;
1021 l_timeval it_value;
1022 };
1023
1024 #define B2L_ITIMERVAL(bip, lip) \
1025 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1026 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1027 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1028 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1029
1030 int
1031 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1032 {
1033 int error;
1034 struct l_itimerval ls;
1035 struct itimerval aitv, oitv;
1036
1037 #ifdef DEBUG
1038 if (ldebug(setitimer))
1039 printf(ARGS(setitimer, "%p, %p"),
1040 (void *)uap->itv, (void *)uap->oitv);
1041 #endif
1042
1043 if (uap->itv == NULL) {
1044 uap->itv = uap->oitv;
1045 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1046 }
1047
1048 error = copyin(uap->itv, &ls, sizeof(ls));
1049 if (error != 0)
1050 return (error);
1051 B2L_ITIMERVAL(&aitv, &ls);
1052 #ifdef DEBUG
1053 if (ldebug(setitimer)) {
1054 printf("setitimer: value: sec: %jd, usec: %ld\n",
1055 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1056 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1057 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1058 }
1059 #endif
1060 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1061 if (error != 0 || uap->oitv == NULL)
1062 return (error);
1063 B2L_ITIMERVAL(&ls, &oitv);
1064
1065 return (copyout(&ls, uap->oitv, sizeof(ls)));
1066 }
1067
1068 int
1069 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1070 {
1071 int error;
1072 struct l_itimerval ls;
1073 struct itimerval aitv;
1074
1075 #ifdef DEBUG
1076 if (ldebug(getitimer))
1077 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1078 #endif
1079 error = kern_getitimer(td, uap->which, &aitv);
1080 if (error != 0)
1081 return (error);
1082 B2L_ITIMERVAL(&ls, &aitv);
1083 return (copyout(&ls, uap->itv, sizeof(ls)));
1084 }
1085
1086 int
1087 linux_nice(struct thread *td, struct linux_nice_args *args)
1088 {
1089 struct setpriority_args bsd_args;
1090
1091 bsd_args.which = PRIO_PROCESS;
1092 bsd_args.who = 0; /* current process */
1093 bsd_args.prio = args->inc;
1094 return (sys_setpriority(td, &bsd_args));
1095 }
1096
1097 int
1098 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1099 {
1100 struct ucred *newcred, *oldcred;
1101 l_gid_t *linux_gidset;
1102 gid_t *bsd_gidset;
1103 int ngrp, error;
1104 struct proc *p;
1105
1106 ngrp = args->gidsetsize;
1107 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1108 return (EINVAL);
1109 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1110 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1111 if (error)
1112 goto out;
1113 newcred = crget();
1114 p = td->td_proc;
1115 PROC_LOCK(p);
1116 oldcred = crcopysafe(p, newcred);
1117
1118 /*
1119 * cr_groups[0] holds egid. Setting the whole set from
1120 * the supplied set will cause egid to be changed too.
1121 * Keep cr_groups[0] unchanged to prevent that.
1122 */
1123
1124 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1125 PROC_UNLOCK(p);
1126 crfree(newcred);
1127 goto out;
1128 }
1129
1130 if (ngrp > 0) {
1131 newcred->cr_ngroups = ngrp + 1;
1132
1133 bsd_gidset = newcred->cr_groups;
1134 ngrp--;
1135 while (ngrp >= 0) {
1136 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1137 ngrp--;
1138 }
1139 } else
1140 newcred->cr_ngroups = 1;
1141
1142 setsugid(p);
1143 p->p_ucred = newcred;
1144 PROC_UNLOCK(p);
1145 crfree(oldcred);
1146 error = 0;
1147 out:
1148 free(linux_gidset, M_TEMP);
1149 return (error);
1150 }
1151
1152 int
1153 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1154 {
1155 struct ucred *cred;
1156 l_gid_t *linux_gidset;
1157 gid_t *bsd_gidset;
1158 int bsd_gidsetsz, ngrp, error;
1159
1160 cred = td->td_ucred;
1161 bsd_gidset = cred->cr_groups;
1162 bsd_gidsetsz = cred->cr_ngroups - 1;
1163
1164 /*
1165 * cr_groups[0] holds egid. Returning the whole set
1166 * here will cause a duplicate. Exclude cr_groups[0]
1167 * to prevent that.
1168 */
1169
1170 if ((ngrp = args->gidsetsize) == 0) {
1171 td->td_retval[0] = bsd_gidsetsz;
1172 return (0);
1173 }
1174
1175 if (ngrp < bsd_gidsetsz)
1176 return (EINVAL);
1177
1178 ngrp = 0;
1179 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1180 M_TEMP, M_WAITOK);
1181 while (ngrp < bsd_gidsetsz) {
1182 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1183 ngrp++;
1184 }
1185
1186 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1187 free(linux_gidset, M_TEMP);
1188 if (error)
1189 return (error);
1190
1191 td->td_retval[0] = ngrp;
1192 return (0);
1193 }
1194
1195 int
1196 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1197 {
1198 struct rlimit bsd_rlim;
1199 struct l_rlimit rlim;
1200 u_int which;
1201 int error;
1202
1203 #ifdef DEBUG
1204 if (ldebug(setrlimit))
1205 printf(ARGS(setrlimit, "%d, %p"),
1206 args->resource, (void *)args->rlim);
1207 #endif
1208
1209 if (args->resource >= LINUX_RLIM_NLIMITS)
1210 return (EINVAL);
1211
1212 which = linux_to_bsd_resource[args->resource];
1213 if (which == -1)
1214 return (EINVAL);
1215
1216 error = copyin(args->rlim, &rlim, sizeof(rlim));
1217 if (error)
1218 return (error);
1219
1220 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1221 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1222 return (kern_setrlimit(td, which, &bsd_rlim));
1223 }
1224
1225 int
1226 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1227 {
1228 struct l_rlimit rlim;
1229 struct proc *p = td->td_proc;
1230 struct rlimit bsd_rlim;
1231 u_int which;
1232
1233 #ifdef DEBUG
1234 if (ldebug(old_getrlimit))
1235 printf(ARGS(old_getrlimit, "%d, %p"),
1236 args->resource, (void *)args->rlim);
1237 #endif
1238
1239 if (args->resource >= LINUX_RLIM_NLIMITS)
1240 return (EINVAL);
1241
1242 which = linux_to_bsd_resource[args->resource];
1243 if (which == -1)
1244 return (EINVAL);
1245
1246 PROC_LOCK(p);
1247 lim_rlimit(p, which, &bsd_rlim);
1248 PROC_UNLOCK(p);
1249
1250 #ifdef COMPAT_LINUX32
1251 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1252 if (rlim.rlim_cur == UINT_MAX)
1253 rlim.rlim_cur = INT_MAX;
1254 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1255 if (rlim.rlim_max == UINT_MAX)
1256 rlim.rlim_max = INT_MAX;
1257 #else
1258 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1259 if (rlim.rlim_cur == ULONG_MAX)
1260 rlim.rlim_cur = LONG_MAX;
1261 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1262 if (rlim.rlim_max == ULONG_MAX)
1263 rlim.rlim_max = LONG_MAX;
1264 #endif
1265 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1266 }
1267
1268 int
1269 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1270 {
1271 struct l_rlimit rlim;
1272 struct proc *p = td->td_proc;
1273 struct rlimit bsd_rlim;
1274 u_int which;
1275
1276 #ifdef DEBUG
1277 if (ldebug(getrlimit))
1278 printf(ARGS(getrlimit, "%d, %p"),
1279 args->resource, (void *)args->rlim);
1280 #endif
1281
1282 if (args->resource >= LINUX_RLIM_NLIMITS)
1283 return (EINVAL);
1284
1285 which = linux_to_bsd_resource[args->resource];
1286 if (which == -1)
1287 return (EINVAL);
1288
1289 PROC_LOCK(p);
1290 lim_rlimit(p, which, &bsd_rlim);
1291 PROC_UNLOCK(p);
1292
1293 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1294 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1295 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1296 }
1297
1298 int
1299 linux_sched_setscheduler(struct thread *td,
1300 struct linux_sched_setscheduler_args *args)
1301 {
1302 struct sched_setscheduler_args bsd;
1303
1304 #ifdef DEBUG
1305 if (ldebug(sched_setscheduler))
1306 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1307 args->pid, args->policy, (const void *)args->param);
1308 #endif
1309
1310 switch (args->policy) {
1311 case LINUX_SCHED_OTHER:
1312 bsd.policy = SCHED_OTHER;
1313 break;
1314 case LINUX_SCHED_FIFO:
1315 bsd.policy = SCHED_FIFO;
1316 break;
1317 case LINUX_SCHED_RR:
1318 bsd.policy = SCHED_RR;
1319 break;
1320 default:
1321 return (EINVAL);
1322 }
1323
1324 bsd.pid = args->pid;
1325 bsd.param = (struct sched_param *)args->param;
1326 return (sys_sched_setscheduler(td, &bsd));
1327 }
1328
1329 int
1330 linux_sched_getscheduler(struct thread *td,
1331 struct linux_sched_getscheduler_args *args)
1332 {
1333 struct sched_getscheduler_args bsd;
1334 int error;
1335
1336 #ifdef DEBUG
1337 if (ldebug(sched_getscheduler))
1338 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1339 #endif
1340
1341 bsd.pid = args->pid;
1342 error = sys_sched_getscheduler(td, &bsd);
1343
1344 switch (td->td_retval[0]) {
1345 case SCHED_OTHER:
1346 td->td_retval[0] = LINUX_SCHED_OTHER;
1347 break;
1348 case SCHED_FIFO:
1349 td->td_retval[0] = LINUX_SCHED_FIFO;
1350 break;
1351 case SCHED_RR:
1352 td->td_retval[0] = LINUX_SCHED_RR;
1353 break;
1354 }
1355
1356 return (error);
1357 }
1358
1359 int
1360 linux_sched_get_priority_max(struct thread *td,
1361 struct linux_sched_get_priority_max_args *args)
1362 {
1363 struct sched_get_priority_max_args bsd;
1364
1365 #ifdef DEBUG
1366 if (ldebug(sched_get_priority_max))
1367 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1368 #endif
1369
1370 switch (args->policy) {
1371 case LINUX_SCHED_OTHER:
1372 bsd.policy = SCHED_OTHER;
1373 break;
1374 case LINUX_SCHED_FIFO:
1375 bsd.policy = SCHED_FIFO;
1376 break;
1377 case LINUX_SCHED_RR:
1378 bsd.policy = SCHED_RR;
1379 break;
1380 default:
1381 return (EINVAL);
1382 }
1383 return (sys_sched_get_priority_max(td, &bsd));
1384 }
1385
1386 int
1387 linux_sched_get_priority_min(struct thread *td,
1388 struct linux_sched_get_priority_min_args *args)
1389 {
1390 struct sched_get_priority_min_args bsd;
1391
1392 #ifdef DEBUG
1393 if (ldebug(sched_get_priority_min))
1394 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1395 #endif
1396
1397 switch (args->policy) {
1398 case LINUX_SCHED_OTHER:
1399 bsd.policy = SCHED_OTHER;
1400 break;
1401 case LINUX_SCHED_FIFO:
1402 bsd.policy = SCHED_FIFO;
1403 break;
1404 case LINUX_SCHED_RR:
1405 bsd.policy = SCHED_RR;
1406 break;
1407 default:
1408 return (EINVAL);
1409 }
1410 return (sys_sched_get_priority_min(td, &bsd));
1411 }
1412
1413 #define REBOOT_CAD_ON 0x89abcdef
1414 #define REBOOT_CAD_OFF 0
1415 #define REBOOT_HALT 0xcdef0123
1416 #define REBOOT_RESTART 0x01234567
1417 #define REBOOT_RESTART2 0xA1B2C3D4
1418 #define REBOOT_POWEROFF 0x4321FEDC
1419 #define REBOOT_MAGIC1 0xfee1dead
1420 #define REBOOT_MAGIC2 0x28121969
1421 #define REBOOT_MAGIC2A 0x05121996
1422 #define REBOOT_MAGIC2B 0x16041998
1423
1424 int
1425 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1426 {
1427 struct reboot_args bsd_args;
1428
1429 #ifdef DEBUG
1430 if (ldebug(reboot))
1431 printf(ARGS(reboot, "0x%x"), args->cmd);
1432 #endif
1433
1434 if (args->magic1 != REBOOT_MAGIC1)
1435 return (EINVAL);
1436
1437 switch (args->magic2) {
1438 case REBOOT_MAGIC2:
1439 case REBOOT_MAGIC2A:
1440 case REBOOT_MAGIC2B:
1441 break;
1442 default:
1443 return (EINVAL);
1444 }
1445
1446 switch (args->cmd) {
1447 case REBOOT_CAD_ON:
1448 case REBOOT_CAD_OFF:
1449 return (priv_check(td, PRIV_REBOOT));
1450 case REBOOT_HALT:
1451 bsd_args.opt = RB_HALT;
1452 break;
1453 case REBOOT_RESTART:
1454 case REBOOT_RESTART2:
1455 bsd_args.opt = 0;
1456 break;
1457 case REBOOT_POWEROFF:
1458 bsd_args.opt = RB_POWEROFF;
1459 break;
1460 default:
1461 return (EINVAL);
1462 }
1463 return (sys_reboot(td, &bsd_args));
1464 }
1465
1466
1467 /*
1468 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1469 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1470 * are assumed to be preserved. The following lightweight syscalls fixes
1471 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1472 *
1473 * linux_getpid() - MP SAFE
1474 * linux_getgid() - MP SAFE
1475 * linux_getuid() - MP SAFE
1476 */
1477
1478 int
1479 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1480 {
1481 struct linux_emuldata *em;
1482
1483 #ifdef DEBUG
1484 if (ldebug(getpid))
1485 printf(ARGS(getpid, ""));
1486 #endif
1487
1488 if (linux_use26(td)) {
1489 em = em_find(td->td_proc, EMUL_DONTLOCK);
1490 KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1491 td->td_retval[0] = em->shared->group_pid;
1492 } else {
1493 td->td_retval[0] = td->td_proc->p_pid;
1494 }
1495
1496 return (0);
1497 }
1498
1499 int
1500 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1501 {
1502
1503 #ifdef DEBUG
1504 if (ldebug(gettid))
1505 printf(ARGS(gettid, ""));
1506 #endif
1507
1508 td->td_retval[0] = td->td_proc->p_pid;
1509 return (0);
1510 }
1511
1512
1513 int
1514 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1515 {
1516 struct linux_emuldata *em;
1517 struct proc *p, *pp;
1518
1519 #ifdef DEBUG
1520 if (ldebug(getppid))
1521 printf(ARGS(getppid, ""));
1522 #endif
1523
1524 if (!linux_use26(td)) {
1525 PROC_LOCK(td->td_proc);
1526 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1527 PROC_UNLOCK(td->td_proc);
1528 return (0);
1529 }
1530
1531 em = em_find(td->td_proc, EMUL_DONTLOCK);
1532
1533 KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1534
1535 /* find the group leader */
1536 p = pfind(em->shared->group_pid);
1537
1538 if (p == NULL) {
1539 #ifdef DEBUG
1540 printf(LMSG("parent process not found.\n"));
1541 #endif
1542 return (0);
1543 }
1544
1545 pp = p->p_pptr; /* switch to parent */
1546 PROC_LOCK(pp);
1547 PROC_UNLOCK(p);
1548
1549 /* if its also linux process */
1550 if (pp->p_sysent == &elf_linux_sysvec) {
1551 em = em_find(pp, EMUL_DONTLOCK);
1552 KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1553
1554 td->td_retval[0] = em->shared->group_pid;
1555 } else
1556 td->td_retval[0] = pp->p_pid;
1557
1558 PROC_UNLOCK(pp);
1559
1560 return (0);
1561 }
1562
1563 int
1564 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1565 {
1566
1567 #ifdef DEBUG
1568 if (ldebug(getgid))
1569 printf(ARGS(getgid, ""));
1570 #endif
1571
1572 td->td_retval[0] = td->td_ucred->cr_rgid;
1573 return (0);
1574 }
1575
1576 int
1577 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1578 {
1579
1580 #ifdef DEBUG
1581 if (ldebug(getuid))
1582 printf(ARGS(getuid, ""));
1583 #endif
1584
1585 td->td_retval[0] = td->td_ucred->cr_ruid;
1586 return (0);
1587 }
1588
1589
1590 int
1591 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1592 {
1593 struct getsid_args bsd;
1594
1595 #ifdef DEBUG
1596 if (ldebug(getsid))
1597 printf(ARGS(getsid, "%i"), args->pid);
1598 #endif
1599
1600 bsd.pid = args->pid;
1601 return (sys_getsid(td, &bsd));
1602 }
1603
1604 int
1605 linux_nosys(struct thread *td, struct nosys_args *ignore)
1606 {
1607
1608 return (ENOSYS);
1609 }
1610
1611 int
1612 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1613 {
1614 struct getpriority_args bsd_args;
1615 int error;
1616
1617 #ifdef DEBUG
1618 if (ldebug(getpriority))
1619 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1620 #endif
1621
1622 bsd_args.which = args->which;
1623 bsd_args.who = args->who;
1624 error = sys_getpriority(td, &bsd_args);
1625 td->td_retval[0] = 20 - td->td_retval[0];
1626 return (error);
1627 }
1628
1629 int
1630 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1631 {
1632 int name[2];
1633
1634 #ifdef DEBUG
1635 if (ldebug(sethostname))
1636 printf(ARGS(sethostname, "*, %i"), args->len);
1637 #endif
1638
1639 name[0] = CTL_KERN;
1640 name[1] = KERN_HOSTNAME;
1641 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1642 args->len, 0, 0));
1643 }
1644
1645 int
1646 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1647 {
1648 int name[2];
1649
1650 #ifdef DEBUG
1651 if (ldebug(setdomainname))
1652 printf(ARGS(setdomainname, "*, %i"), args->len);
1653 #endif
1654
1655 name[0] = CTL_KERN;
1656 name[1] = KERN_NISDOMAINNAME;
1657 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1658 args->len, 0, 0));
1659 }
1660
1661 int
1662 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1663 {
1664 struct linux_emuldata *em;
1665
1666 #ifdef DEBUG
1667 if (ldebug(exit_group))
1668 printf(ARGS(exit_group, "%i"), args->error_code);
1669 #endif
1670
1671 em = em_find(td->td_proc, EMUL_DONTLOCK);
1672 if (em->shared->refs > 1) {
1673 EMUL_SHARED_WLOCK(&emul_shared_lock);
1674 em->shared->flags |= EMUL_SHARED_HASXSTAT;
1675 em->shared->xstat = W_EXITCODE(args->error_code, 0);
1676 EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1677 if (linux_use26(td))
1678 linux_kill_threads(td, SIGKILL);
1679 }
1680
1681 /*
1682 * XXX: we should send a signal to the parent if
1683 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1684 * as it doesnt occur often.
1685 */
1686 exit1(td, W_EXITCODE(args->error_code, 0));
1687
1688 return (0);
1689 }
1690
1691 #define _LINUX_CAPABILITY_VERSION 0x19980330
1692
1693 struct l_user_cap_header {
1694 l_int version;
1695 l_int pid;
1696 };
1697
1698 struct l_user_cap_data {
1699 l_int effective;
1700 l_int permitted;
1701 l_int inheritable;
1702 };
1703
1704 int
1705 linux_capget(struct thread *td, struct linux_capget_args *args)
1706 {
1707 struct l_user_cap_header luch;
1708 struct l_user_cap_data lucd;
1709 int error;
1710
1711 if (args->hdrp == NULL)
1712 return (EFAULT);
1713
1714 error = copyin(args->hdrp, &luch, sizeof(luch));
1715 if (error != 0)
1716 return (error);
1717
1718 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1719 luch.version = _LINUX_CAPABILITY_VERSION;
1720 error = copyout(&luch, args->hdrp, sizeof(luch));
1721 if (error)
1722 return (error);
1723 return (EINVAL);
1724 }
1725
1726 if (luch.pid)
1727 return (EPERM);
1728
1729 if (args->datap) {
1730 /*
1731 * The current implementation doesn't support setting
1732 * a capability (it's essentially a stub) so indicate
1733 * that no capabilities are currently set or available
1734 * to request.
1735 */
1736 bzero (&lucd, sizeof(lucd));
1737 error = copyout(&lucd, args->datap, sizeof(lucd));
1738 }
1739
1740 return (error);
1741 }
1742
1743 int
1744 linux_capset(struct thread *td, struct linux_capset_args *args)
1745 {
1746 struct l_user_cap_header luch;
1747 struct l_user_cap_data lucd;
1748 int error;
1749
1750 if (args->hdrp == NULL || args->datap == NULL)
1751 return (EFAULT);
1752
1753 error = copyin(args->hdrp, &luch, sizeof(luch));
1754 if (error != 0)
1755 return (error);
1756
1757 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1758 luch.version = _LINUX_CAPABILITY_VERSION;
1759 error = copyout(&luch, args->hdrp, sizeof(luch));
1760 if (error)
1761 return (error);
1762 return (EINVAL);
1763 }
1764
1765 if (luch.pid)
1766 return (EPERM);
1767
1768 error = copyin(args->datap, &lucd, sizeof(lucd));
1769 if (error != 0)
1770 return (error);
1771
1772 /* We currently don't support setting any capabilities. */
1773 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1774 linux_msg(td,
1775 "capset effective=0x%x, permitted=0x%x, "
1776 "inheritable=0x%x is not implemented",
1777 (int)lucd.effective, (int)lucd.permitted,
1778 (int)lucd.inheritable);
1779 return (EPERM);
1780 }
1781
1782 return (0);
1783 }
1784
1785 int
1786 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1787 {
1788 int error = 0, max_size;
1789 struct proc *p = td->td_proc;
1790 char comm[LINUX_MAX_COMM_LEN];
1791 struct linux_emuldata *em;
1792 int pdeath_signal;
1793
1794 #ifdef DEBUG
1795 if (ldebug(prctl))
1796 printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1797 args->arg2, args->arg3, args->arg4, args->arg5);
1798 #endif
1799
1800 switch (args->option) {
1801 case LINUX_PR_SET_PDEATHSIG:
1802 if (!LINUX_SIG_VALID(args->arg2))
1803 return (EINVAL);
1804 em = em_find(p, EMUL_DOLOCK);
1805 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1806 em->pdeath_signal = args->arg2;
1807 EMUL_UNLOCK(&emul_lock);
1808 break;
1809 case LINUX_PR_GET_PDEATHSIG:
1810 em = em_find(p, EMUL_DOLOCK);
1811 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1812 pdeath_signal = em->pdeath_signal;
1813 EMUL_UNLOCK(&emul_lock);
1814 error = copyout(&pdeath_signal,
1815 (void *)(register_t)args->arg2,
1816 sizeof(pdeath_signal));
1817 break;
1818 case LINUX_PR_GET_KEEPCAPS:
1819 /*
1820 * Indicate that we always clear the effective and
1821 * permitted capability sets when the user id becomes
1822 * non-zero (actually the capability sets are simply
1823 * always zero in the current implementation).
1824 */
1825 td->td_retval[0] = 0;
1826 break;
1827 case LINUX_PR_SET_KEEPCAPS:
1828 /*
1829 * Ignore requests to keep the effective and permitted
1830 * capability sets when the user id becomes non-zero.
1831 */
1832 break;
1833 case LINUX_PR_SET_NAME:
1834 /*
1835 * To be on the safe side we need to make sure to not
1836 * overflow the size a linux program expects. We already
1837 * do this here in the copyin, so that we don't need to
1838 * check on copyout.
1839 */
1840 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1841 error = copyinstr((void *)(register_t)args->arg2, comm,
1842 max_size, NULL);
1843
1844 /* Linux silently truncates the name if it is too long. */
1845 if (error == ENAMETOOLONG) {
1846 /*
1847 * XXX: copyinstr() isn't documented to populate the
1848 * array completely, so do a copyin() to be on the
1849 * safe side. This should be changed in case
1850 * copyinstr() is changed to guarantee this.
1851 */
1852 error = copyin((void *)(register_t)args->arg2, comm,
1853 max_size - 1);
1854 comm[max_size - 1] = '\0';
1855 }
1856 if (error)
1857 return (error);
1858
1859 PROC_LOCK(p);
1860 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1861 PROC_UNLOCK(p);
1862 break;
1863 case LINUX_PR_GET_NAME:
1864 PROC_LOCK(p);
1865 strlcpy(comm, p->p_comm, sizeof(comm));
1866 PROC_UNLOCK(p);
1867 error = copyout(comm, (void *)(register_t)args->arg2,
1868 strlen(comm) + 1);
1869 break;
1870 default:
1871 error = EINVAL;
1872 break;
1873 }
1874
1875 return (error);
1876 }
1877
1878 /*
1879 * Get affinity of a process.
1880 */
1881 int
1882 linux_sched_getaffinity(struct thread *td,
1883 struct linux_sched_getaffinity_args *args)
1884 {
1885 int error;
1886 struct cpuset_getaffinity_args cga;
1887
1888 #ifdef DEBUG
1889 if (ldebug(sched_getaffinity))
1890 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1891 args->len);
1892 #endif
1893 if (args->len < sizeof(cpuset_t))
1894 return (EINVAL);
1895
1896 cga.level = CPU_LEVEL_WHICH;
1897 cga.which = CPU_WHICH_PID;
1898 cga.id = args->pid;
1899 cga.cpusetsize = sizeof(cpuset_t);
1900 cga.mask = (cpuset_t *) args->user_mask_ptr;
1901
1902 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1903 td->td_retval[0] = sizeof(cpuset_t);
1904
1905 return (error);
1906 }
1907
1908 /*
1909 * Set affinity of a process.
1910 */
1911 int
1912 linux_sched_setaffinity(struct thread *td,
1913 struct linux_sched_setaffinity_args *args)
1914 {
1915 struct cpuset_setaffinity_args csa;
1916
1917 #ifdef DEBUG
1918 if (ldebug(sched_setaffinity))
1919 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1920 args->len);
1921 #endif
1922 if (args->len < sizeof(cpuset_t))
1923 return (EINVAL);
1924
1925 csa.level = CPU_LEVEL_WHICH;
1926 csa.which = CPU_WHICH_PID;
1927 csa.id = args->pid;
1928 csa.cpusetsize = sizeof(cpuset_t);
1929 csa.mask = (cpuset_t *) args->user_mask_ptr;
1930
1931 return (sys_cpuset_setaffinity(td, &csa));
1932 }
Cache object: 7d94884dae828a45dd9e1dcd06b06bc9
|