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 #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 bzero(&sysinfo, sizeof(sysinfo));
142 getnanouptime(&ts);
143 if (ts.tv_nsec != 0)
144 ts.tv_sec++;
145 sysinfo.uptime = ts.tv_sec;
146
147 /* Use the information from the mib to get our load averages */
148 for (i = 0; i < 3; i++)
149 sysinfo.loads[i] = averunnable.ldavg[i] *
150 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
151
152 sysinfo.totalram = physmem * PAGE_SIZE;
153 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
154
155 sysinfo.sharedram = 0;
156 mtx_lock(&vm_object_list_mtx);
157 TAILQ_FOREACH(object, &vm_object_list, object_list)
158 if (object->shadow_count > 1)
159 sysinfo.sharedram += object->resident_page_count;
160 mtx_unlock(&vm_object_list_mtx);
161
162 sysinfo.sharedram *= PAGE_SIZE;
163 sysinfo.bufferram = 0;
164
165 swap_pager_status(&i, &j);
166 sysinfo.totalswap = i * PAGE_SIZE;
167 sysinfo.freeswap = (i - j) * PAGE_SIZE;
168
169 sysinfo.procs = nprocs;
170
171 /* The following are only present in newer Linux kernels. */
172 sysinfo.totalbig = 0;
173 sysinfo.freebig = 0;
174 sysinfo.mem_unit = 1;
175
176 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
177 }
178
179 int
180 linux_alarm(struct thread *td, struct linux_alarm_args *args)
181 {
182 struct itimerval it, old_it;
183 u_int secs;
184 int error;
185
186 #ifdef DEBUG
187 if (ldebug(alarm))
188 printf(ARGS(alarm, "%u"), args->secs);
189 #endif
190
191 secs = args->secs;
192
193 if (secs > INT_MAX)
194 secs = INT_MAX;
195
196 it.it_value.tv_sec = (long) secs;
197 it.it_value.tv_usec = 0;
198 it.it_interval.tv_sec = 0;
199 it.it_interval.tv_usec = 0;
200 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
201 if (error)
202 return (error);
203 if (timevalisset(&old_it.it_value)) {
204 if (old_it.it_value.tv_usec != 0)
205 old_it.it_value.tv_sec++;
206 td->td_retval[0] = old_it.it_value.tv_sec;
207 }
208 return (0);
209 }
210
211 int
212 linux_brk(struct thread *td, struct linux_brk_args *args)
213 {
214 struct vmspace *vm = td->td_proc->p_vmspace;
215 vm_offset_t new, old;
216 struct obreak_args /* {
217 char * nsize;
218 } */ tmp;
219
220 #ifdef DEBUG
221 if (ldebug(brk))
222 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
223 #endif
224 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
225 new = (vm_offset_t)args->dsend;
226 tmp.nsize = (char *)new;
227 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
228 td->td_retval[0] = (long)new;
229 else
230 td->td_retval[0] = (long)old;
231
232 return (0);
233 }
234
235 #if defined(__i386__)
236 /* XXX: what about amd64/linux32? */
237
238 int
239 linux_uselib(struct thread *td, struct linux_uselib_args *args)
240 {
241 struct nameidata ni;
242 struct vnode *vp;
243 struct exec *a_out;
244 struct vattr attr;
245 vm_offset_t vmaddr;
246 unsigned long file_offset;
247 unsigned long bss_size;
248 char *library;
249 ssize_t aresid;
250 int error, locked, vfslocked, writecount;
251
252 LCONVPATHEXIST(td, args->library, &library);
253
254 #ifdef DEBUG
255 if (ldebug(uselib))
256 printf(ARGS(uselib, "%s"), library);
257 #endif
258
259 a_out = NULL;
260 vfslocked = 0;
261 locked = 0;
262 vp = NULL;
263
264 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
265 UIO_SYSSPACE, library, td);
266 error = namei(&ni);
267 LFREEPATH(library);
268 if (error)
269 goto cleanup;
270
271 vp = ni.ni_vp;
272 vfslocked = NDHASGIANT(&ni);
273 NDFREE(&ni, NDF_ONLY_PNBUF);
274
275 /*
276 * From here on down, we have a locked vnode that must be unlocked.
277 * XXX: The code below largely duplicates exec_check_permissions().
278 */
279 locked = 1;
280
281 /* Writable? */
282 error = VOP_GET_WRITECOUNT(vp, &writecount);
283 if (error != 0)
284 goto cleanup;
285 if (writecount != 0) {
286 error = ETXTBSY;
287 goto cleanup;
288 }
289
290 /* Executable? */
291 error = VOP_GETATTR(vp, &attr, td->td_ucred);
292 if (error)
293 goto cleanup;
294
295 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
296 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
297 /* EACCESS is what exec(2) returns. */
298 error = ENOEXEC;
299 goto cleanup;
300 }
301
302 /* Sensible size? */
303 if (attr.va_size == 0) {
304 error = ENOEXEC;
305 goto cleanup;
306 }
307
308 /* Can we access it? */
309 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
310 if (error)
311 goto cleanup;
312
313 /*
314 * XXX: This should use vn_open() so that it is properly authorized,
315 * and to reduce code redundancy all over the place here.
316 * XXX: Not really, it duplicates far more of exec_check_permissions()
317 * than vn_open().
318 */
319 #ifdef MAC
320 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
321 if (error)
322 goto cleanup;
323 #endif
324 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
325 if (error)
326 goto cleanup;
327
328 /* Pull in executable header into exec_map */
329 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
330 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
331 if (error)
332 goto cleanup;
333
334 /* Is it a Linux binary ? */
335 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
336 error = ENOEXEC;
337 goto cleanup;
338 }
339
340 /*
341 * While we are here, we should REALLY do some more checks
342 */
343
344 /* Set file/virtual offset based on a.out variant. */
345 switch ((int)(a_out->a_magic & 0xffff)) {
346 case 0413: /* ZMAGIC */
347 file_offset = 1024;
348 break;
349 case 0314: /* QMAGIC */
350 file_offset = 0;
351 break;
352 default:
353 error = ENOEXEC;
354 goto cleanup;
355 }
356
357 bss_size = round_page(a_out->a_bss);
358
359 /* Check various fields in header for validity/bounds. */
360 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
361 error = ENOEXEC;
362 goto cleanup;
363 }
364
365 /* text + data can't exceed file size */
366 if (a_out->a_data + a_out->a_text > attr.va_size) {
367 error = EFAULT;
368 goto cleanup;
369 }
370
371 /*
372 * text/data/bss must not exceed limits
373 * XXX - this is not complete. it should check current usage PLUS
374 * the resources needed by this library.
375 */
376 PROC_LOCK(td->td_proc);
377 if (a_out->a_text > maxtsiz ||
378 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
379 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
380 bss_size) != 0) {
381 PROC_UNLOCK(td->td_proc);
382 error = ENOMEM;
383 goto cleanup;
384 }
385 PROC_UNLOCK(td->td_proc);
386
387 /*
388 * Prevent more writers.
389 * XXX: Note that if any of the VM operations fail below we don't
390 * clear this flag.
391 */
392 VOP_SET_TEXT(vp);
393
394 /*
395 * Lock no longer needed
396 */
397 locked = 0;
398 VOP_UNLOCK(vp, 0);
399 VFS_UNLOCK_GIANT(vfslocked);
400
401 /*
402 * Check if file_offset page aligned. Currently we cannot handle
403 * misalinged file offsets, and so we read in the entire image
404 * (what a waste).
405 */
406 if (file_offset & PAGE_MASK) {
407 #ifdef DEBUG
408 printf("uselib: Non page aligned binary %lu\n", file_offset);
409 #endif
410 /* Map text+data read/write/execute */
411
412 /* a_entry is the load address and is page aligned */
413 vmaddr = trunc_page(a_out->a_entry);
414
415 /* get anon user mapping, read+write+execute */
416 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
417 &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
418 VM_PROT_ALL, 0);
419 if (error)
420 goto cleanup;
421
422 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
423 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
424 td->td_ucred, NOCRED, &aresid, td);
425 if (error != 0)
426 goto cleanup;
427 if (aresid != 0) {
428 error = ENOEXEC;
429 goto cleanup;
430 }
431 } else {
432 #ifdef DEBUG
433 printf("uselib: Page aligned binary %lu\n", file_offset);
434 #endif
435 /*
436 * for QMAGIC, a_entry is 20 bytes beyond the load address
437 * to skip the executable header
438 */
439 vmaddr = trunc_page(a_out->a_entry);
440
441 /*
442 * Map it all into the process's space as a single
443 * copy-on-write "data" segment.
444 */
445 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
446 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
447 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
448 if (error)
449 goto cleanup;
450 }
451 #ifdef DEBUG
452 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
453 ((long *)vmaddr)[1]);
454 #endif
455 if (bss_size != 0) {
456 /* Calculate BSS start address */
457 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
458 a_out->a_data;
459
460 /* allocate some 'anon' space */
461 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
462 &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
463 if (error)
464 goto cleanup;
465 }
466
467 cleanup:
468 /* Unlock vnode if needed */
469 if (locked) {
470 VOP_UNLOCK(vp, 0);
471 VFS_UNLOCK_GIANT(vfslocked);
472 }
473
474 /* Release the temporary mapping. */
475 if (a_out)
476 kmem_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
477
478 return (error);
479 }
480
481 #endif /* __i386__ */
482
483 int
484 linux_select(struct thread *td, struct linux_select_args *args)
485 {
486 l_timeval ltv;
487 struct timeval tv0, tv1, utv, *tvp;
488 int error;
489
490 #ifdef DEBUG
491 if (ldebug(select))
492 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
493 (void *)args->readfds, (void *)args->writefds,
494 (void *)args->exceptfds, (void *)args->timeout);
495 #endif
496
497 /*
498 * Store current time for computation of the amount of
499 * time left.
500 */
501 if (args->timeout) {
502 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
503 goto select_out;
504 utv.tv_sec = ltv.tv_sec;
505 utv.tv_usec = ltv.tv_usec;
506 #ifdef DEBUG
507 if (ldebug(select))
508 printf(LMSG("incoming timeout (%jd/%ld)"),
509 (intmax_t)utv.tv_sec, utv.tv_usec);
510 #endif
511
512 if (itimerfix(&utv)) {
513 /*
514 * The timeval was invalid. Convert it to something
515 * valid that will act as it does under Linux.
516 */
517 utv.tv_sec += utv.tv_usec / 1000000;
518 utv.tv_usec %= 1000000;
519 if (utv.tv_usec < 0) {
520 utv.tv_sec -= 1;
521 utv.tv_usec += 1000000;
522 }
523 if (utv.tv_sec < 0)
524 timevalclear(&utv);
525 }
526 microtime(&tv0);
527 tvp = &utv;
528 } else
529 tvp = NULL;
530
531 error = kern_select(td, args->nfds, args->readfds, args->writefds,
532 args->exceptfds, tvp, sizeof(l_int) * 8);
533
534 #ifdef DEBUG
535 if (ldebug(select))
536 printf(LMSG("real select returns %d"), error);
537 #endif
538 if (error)
539 goto select_out;
540
541 if (args->timeout) {
542 if (td->td_retval[0]) {
543 /*
544 * Compute how much time was left of the timeout,
545 * by subtracting the current time and the time
546 * before we started the call, and subtracting
547 * that result from the user-supplied value.
548 */
549 microtime(&tv1);
550 timevalsub(&tv1, &tv0);
551 timevalsub(&utv, &tv1);
552 if (utv.tv_sec < 0)
553 timevalclear(&utv);
554 } else
555 timevalclear(&utv);
556 #ifdef DEBUG
557 if (ldebug(select))
558 printf(LMSG("outgoing timeout (%jd/%ld)"),
559 (intmax_t)utv.tv_sec, utv.tv_usec);
560 #endif
561 ltv.tv_sec = utv.tv_sec;
562 ltv.tv_usec = utv.tv_usec;
563 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
564 goto select_out;
565 }
566
567 select_out:
568 #ifdef DEBUG
569 if (ldebug(select))
570 printf(LMSG("select_out -> %d"), error);
571 #endif
572 return (error);
573 }
574
575 int
576 linux_mremap(struct thread *td, struct linux_mremap_args *args)
577 {
578 struct munmap_args /* {
579 void *addr;
580 size_t len;
581 } */ bsd_args;
582 int error = 0;
583
584 #ifdef DEBUG
585 if (ldebug(mremap))
586 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
587 (void *)(uintptr_t)args->addr,
588 (unsigned long)args->old_len,
589 (unsigned long)args->new_len,
590 (unsigned long)args->flags);
591 #endif
592
593 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
594 td->td_retval[0] = 0;
595 return (EINVAL);
596 }
597
598 /*
599 * Check for the page alignment.
600 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
601 */
602 if (args->addr & PAGE_MASK) {
603 td->td_retval[0] = 0;
604 return (EINVAL);
605 }
606
607 args->new_len = round_page(args->new_len);
608 args->old_len = round_page(args->old_len);
609
610 if (args->new_len > args->old_len) {
611 td->td_retval[0] = 0;
612 return (ENOMEM);
613 }
614
615 if (args->new_len < args->old_len) {
616 bsd_args.addr =
617 (caddr_t)((uintptr_t)args->addr + args->new_len);
618 bsd_args.len = args->old_len - args->new_len;
619 error = sys_munmap(td, &bsd_args);
620 }
621
622 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
623 return (error);
624 }
625
626 #define LINUX_MS_ASYNC 0x0001
627 #define LINUX_MS_INVALIDATE 0x0002
628 #define LINUX_MS_SYNC 0x0004
629
630 int
631 linux_msync(struct thread *td, struct linux_msync_args *args)
632 {
633 struct msync_args bsd_args;
634
635 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
636 bsd_args.len = (uintptr_t)args->len;
637 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
638
639 return (sys_msync(td, &bsd_args));
640 }
641
642 int
643 linux_time(struct thread *td, struct linux_time_args *args)
644 {
645 struct timeval tv;
646 l_time_t tm;
647 int error;
648
649 #ifdef DEBUG
650 if (ldebug(time))
651 printf(ARGS(time, "*"));
652 #endif
653
654 microtime(&tv);
655 tm = tv.tv_sec;
656 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
657 return (error);
658 td->td_retval[0] = tm;
659 return (0);
660 }
661
662 struct l_times_argv {
663 l_clock_t tms_utime;
664 l_clock_t tms_stime;
665 l_clock_t tms_cutime;
666 l_clock_t tms_cstime;
667 };
668
669
670 /*
671 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
672 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
673 * auxiliary vector entry.
674 */
675 #define CLK_TCK 100
676
677 #define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
678 #define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
679
680 #define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
681 CONVNTCK(r) : CONVOTCK(r))
682
683 int
684 linux_times(struct thread *td, struct linux_times_args *args)
685 {
686 struct timeval tv, utime, stime, cutime, cstime;
687 struct l_times_argv tms;
688 struct proc *p;
689 int error;
690
691 #ifdef DEBUG
692 if (ldebug(times))
693 printf(ARGS(times, "*"));
694 #endif
695
696 if (args->buf != NULL) {
697 p = td->td_proc;
698 PROC_LOCK(p);
699 PROC_SLOCK(p);
700 calcru(p, &utime, &stime);
701 PROC_SUNLOCK(p);
702 calccru(p, &cutime, &cstime);
703 PROC_UNLOCK(p);
704
705 tms.tms_utime = CONVTCK(utime);
706 tms.tms_stime = CONVTCK(stime);
707
708 tms.tms_cutime = CONVTCK(cutime);
709 tms.tms_cstime = CONVTCK(cstime);
710
711 if ((error = copyout(&tms, args->buf, sizeof(tms))))
712 return (error);
713 }
714
715 microuptime(&tv);
716 td->td_retval[0] = (int)CONVTCK(tv);
717 return (0);
718 }
719
720 int
721 linux_newuname(struct thread *td, struct linux_newuname_args *args)
722 {
723 struct l_new_utsname utsname;
724 char osname[LINUX_MAX_UTSNAME];
725 char osrelease[LINUX_MAX_UTSNAME];
726 char *p;
727
728 #ifdef DEBUG
729 if (ldebug(newuname))
730 printf(ARGS(newuname, "*"));
731 #endif
732
733 linux_get_osname(td, osname);
734 linux_get_osrelease(td, osrelease);
735
736 bzero(&utsname, sizeof(utsname));
737 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
738 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
739 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
740 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
741 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
742 for (p = utsname.version; *p != '\0'; ++p)
743 if (*p == '\n') {
744 *p = '\0';
745 break;
746 }
747 strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME);
748
749 return (copyout(&utsname, args->buf, sizeof(utsname)));
750 }
751
752 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
753 struct l_utimbuf {
754 l_time_t l_actime;
755 l_time_t l_modtime;
756 };
757
758 int
759 linux_utime(struct thread *td, struct linux_utime_args *args)
760 {
761 struct timeval tv[2], *tvp;
762 struct l_utimbuf lut;
763 char *fname;
764 int error;
765
766 LCONVPATHEXIST(td, args->fname, &fname);
767
768 #ifdef DEBUG
769 if (ldebug(utime))
770 printf(ARGS(utime, "%s, *"), fname);
771 #endif
772
773 if (args->times) {
774 if ((error = copyin(args->times, &lut, sizeof lut))) {
775 LFREEPATH(fname);
776 return (error);
777 }
778 tv[0].tv_sec = lut.l_actime;
779 tv[0].tv_usec = 0;
780 tv[1].tv_sec = lut.l_modtime;
781 tv[1].tv_usec = 0;
782 tvp = tv;
783 } else
784 tvp = NULL;
785
786 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
787 LFREEPATH(fname);
788 return (error);
789 }
790
791 int
792 linux_utimes(struct thread *td, struct linux_utimes_args *args)
793 {
794 l_timeval ltv[2];
795 struct timeval tv[2], *tvp = NULL;
796 char *fname;
797 int error;
798
799 LCONVPATHEXIST(td, args->fname, &fname);
800
801 #ifdef DEBUG
802 if (ldebug(utimes))
803 printf(ARGS(utimes, "%s, *"), fname);
804 #endif
805
806 if (args->tptr != NULL) {
807 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
808 LFREEPATH(fname);
809 return (error);
810 }
811 tv[0].tv_sec = ltv[0].tv_sec;
812 tv[0].tv_usec = ltv[0].tv_usec;
813 tv[1].tv_sec = ltv[1].tv_sec;
814 tv[1].tv_usec = ltv[1].tv_usec;
815 tvp = tv;
816 }
817
818 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
819 LFREEPATH(fname);
820 return (error);
821 }
822
823 int
824 linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
825 {
826 l_timeval ltv[2];
827 struct timeval tv[2], *tvp = NULL;
828 char *fname;
829 int error, dfd;
830
831 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
832 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
833
834 #ifdef DEBUG
835 if (ldebug(futimesat))
836 printf(ARGS(futimesat, "%s, *"), fname);
837 #endif
838
839 if (args->utimes != NULL) {
840 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
841 LFREEPATH(fname);
842 return (error);
843 }
844 tv[0].tv_sec = ltv[0].tv_sec;
845 tv[0].tv_usec = ltv[0].tv_usec;
846 tv[1].tv_sec = ltv[1].tv_sec;
847 tv[1].tv_usec = ltv[1].tv_usec;
848 tvp = tv;
849 }
850
851 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
852 LFREEPATH(fname);
853 return (error);
854 }
855 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
856
857 int
858 linux_common_wait(struct thread *td, int pid, int *status,
859 int options, struct rusage *ru)
860 {
861 int error, tmpstat;
862
863 error = kern_wait(td, pid, &tmpstat, options, ru);
864 if (error)
865 return (error);
866
867 if (status) {
868 tmpstat &= 0xffff;
869 if (WIFSIGNALED(tmpstat))
870 tmpstat = (tmpstat & 0xffffff80) |
871 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
872 else if (WIFSTOPPED(tmpstat))
873 tmpstat = (tmpstat & 0xffff00ff) |
874 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
875 error = copyout(&tmpstat, status, sizeof(int));
876 }
877
878 return (error);
879 }
880
881 int
882 linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
883 {
884 int options;
885
886 #ifdef DEBUG
887 if (ldebug(waitpid))
888 printf(ARGS(waitpid, "%d, %p, %d"),
889 args->pid, (void *)args->status, args->options);
890 #endif
891 /*
892 * this is necessary because the test in kern_wait doesn't work
893 * because we mess with the options here
894 */
895 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
896 return (EINVAL);
897
898 options = (args->options & (WNOHANG | WUNTRACED));
899 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
900 if (args->options & __WCLONE)
901 options |= WLINUXCLONE;
902
903 return (linux_common_wait(td, args->pid, args->status, options, NULL));
904 }
905
906
907 int
908 linux_mknod(struct thread *td, struct linux_mknod_args *args)
909 {
910 char *path;
911 int error;
912
913 LCONVPATHCREAT(td, args->path, &path);
914
915 #ifdef DEBUG
916 if (ldebug(mknod))
917 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
918 #endif
919
920 switch (args->mode & S_IFMT) {
921 case S_IFIFO:
922 case S_IFSOCK:
923 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
924 break;
925
926 case S_IFCHR:
927 case S_IFBLK:
928 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
929 args->dev);
930 break;
931
932 case S_IFDIR:
933 error = EPERM;
934 break;
935
936 case 0:
937 args->mode |= S_IFREG;
938 /* FALLTHROUGH */
939 case S_IFREG:
940 error = kern_open(td, path, UIO_SYSSPACE,
941 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
942 if (error == 0)
943 kern_close(td, td->td_retval[0]);
944 break;
945
946 default:
947 error = EINVAL;
948 break;
949 }
950 LFREEPATH(path);
951 return (error);
952 }
953
954 int
955 linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
956 {
957 char *path;
958 int error, dfd;
959
960 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
961 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
962
963 #ifdef DEBUG
964 if (ldebug(mknodat))
965 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
966 #endif
967
968 switch (args->mode & S_IFMT) {
969 case S_IFIFO:
970 case S_IFSOCK:
971 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
972 break;
973
974 case S_IFCHR:
975 case S_IFBLK:
976 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
977 args->dev);
978 break;
979
980 case S_IFDIR:
981 error = EPERM;
982 break;
983
984 case 0:
985 args->mode |= S_IFREG;
986 /* FALLTHROUGH */
987 case S_IFREG:
988 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
989 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
990 if (error == 0)
991 kern_close(td, td->td_retval[0]);
992 break;
993
994 default:
995 error = EINVAL;
996 break;
997 }
998 LFREEPATH(path);
999 return (error);
1000 }
1001
1002 /*
1003 * UGH! This is just about the dumbest idea I've ever heard!!
1004 */
1005 int
1006 linux_personality(struct thread *td, struct linux_personality_args *args)
1007 {
1008 #ifdef DEBUG
1009 if (ldebug(personality))
1010 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1011 #endif
1012 if (args->per != 0)
1013 return (EINVAL);
1014
1015 /* Yes Jim, it's still a Linux... */
1016 td->td_retval[0] = 0;
1017 return (0);
1018 }
1019
1020 struct l_itimerval {
1021 l_timeval it_interval;
1022 l_timeval it_value;
1023 };
1024
1025 #define B2L_ITIMERVAL(bip, lip) \
1026 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1027 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1028 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1029 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1030
1031 int
1032 linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1033 {
1034 int error;
1035 struct l_itimerval ls;
1036 struct itimerval aitv, oitv;
1037
1038 #ifdef DEBUG
1039 if (ldebug(setitimer))
1040 printf(ARGS(setitimer, "%p, %p"),
1041 (void *)uap->itv, (void *)uap->oitv);
1042 #endif
1043
1044 if (uap->itv == NULL) {
1045 uap->itv = uap->oitv;
1046 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1047 }
1048
1049 error = copyin(uap->itv, &ls, sizeof(ls));
1050 if (error != 0)
1051 return (error);
1052 B2L_ITIMERVAL(&aitv, &ls);
1053 #ifdef DEBUG
1054 if (ldebug(setitimer)) {
1055 printf("setitimer: value: sec: %jd, usec: %ld\n",
1056 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1057 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1058 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1059 }
1060 #endif
1061 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1062 if (error != 0 || uap->oitv == NULL)
1063 return (error);
1064 B2L_ITIMERVAL(&ls, &oitv);
1065
1066 return (copyout(&ls, uap->oitv, sizeof(ls)));
1067 }
1068
1069 int
1070 linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1071 {
1072 int error;
1073 struct l_itimerval ls;
1074 struct itimerval aitv;
1075
1076 #ifdef DEBUG
1077 if (ldebug(getitimer))
1078 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1079 #endif
1080 error = kern_getitimer(td, uap->which, &aitv);
1081 if (error != 0)
1082 return (error);
1083 B2L_ITIMERVAL(&ls, &aitv);
1084 return (copyout(&ls, uap->itv, sizeof(ls)));
1085 }
1086
1087 int
1088 linux_nice(struct thread *td, struct linux_nice_args *args)
1089 {
1090 struct setpriority_args bsd_args;
1091
1092 bsd_args.which = PRIO_PROCESS;
1093 bsd_args.who = 0; /* current process */
1094 bsd_args.prio = args->inc;
1095 return (sys_setpriority(td, &bsd_args));
1096 }
1097
1098 int
1099 linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1100 {
1101 struct ucred *newcred, *oldcred;
1102 l_gid_t *linux_gidset;
1103 gid_t *bsd_gidset;
1104 int ngrp, error;
1105 struct proc *p;
1106
1107 ngrp = args->gidsetsize;
1108 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1109 return (EINVAL);
1110 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1111 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1112 if (error)
1113 goto out;
1114 newcred = crget();
1115 crextend(newcred, ngrp + 1);
1116 p = td->td_proc;
1117 PROC_LOCK(p);
1118 oldcred = p->p_ucred;
1119 crcopy(newcred, oldcred);
1120
1121 /*
1122 * cr_groups[0] holds egid. Setting the whole set from
1123 * the supplied set will cause egid to be changed too.
1124 * Keep cr_groups[0] unchanged to prevent that.
1125 */
1126
1127 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1128 PROC_UNLOCK(p);
1129 crfree(newcred);
1130 goto out;
1131 }
1132
1133 if (ngrp > 0) {
1134 newcred->cr_ngroups = ngrp + 1;
1135
1136 bsd_gidset = newcred->cr_groups;
1137 ngrp--;
1138 while (ngrp >= 0) {
1139 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1140 ngrp--;
1141 }
1142 } else
1143 newcred->cr_ngroups = 1;
1144
1145 setsugid(p);
1146 p->p_ucred = newcred;
1147 PROC_UNLOCK(p);
1148 crfree(oldcred);
1149 error = 0;
1150 out:
1151 free(linux_gidset, M_TEMP);
1152 return (error);
1153 }
1154
1155 int
1156 linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1157 {
1158 struct ucred *cred;
1159 l_gid_t *linux_gidset;
1160 gid_t *bsd_gidset;
1161 int bsd_gidsetsz, ngrp, error;
1162
1163 cred = td->td_ucred;
1164 bsd_gidset = cred->cr_groups;
1165 bsd_gidsetsz = cred->cr_ngroups - 1;
1166
1167 /*
1168 * cr_groups[0] holds egid. Returning the whole set
1169 * here will cause a duplicate. Exclude cr_groups[0]
1170 * to prevent that.
1171 */
1172
1173 if ((ngrp = args->gidsetsize) == 0) {
1174 td->td_retval[0] = bsd_gidsetsz;
1175 return (0);
1176 }
1177
1178 if (ngrp < bsd_gidsetsz)
1179 return (EINVAL);
1180
1181 ngrp = 0;
1182 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1183 M_TEMP, M_WAITOK);
1184 while (ngrp < bsd_gidsetsz) {
1185 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1186 ngrp++;
1187 }
1188
1189 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1190 free(linux_gidset, M_TEMP);
1191 if (error)
1192 return (error);
1193
1194 td->td_retval[0] = ngrp;
1195 return (0);
1196 }
1197
1198 int
1199 linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1200 {
1201 struct rlimit bsd_rlim;
1202 struct l_rlimit rlim;
1203 u_int which;
1204 int error;
1205
1206 #ifdef DEBUG
1207 if (ldebug(setrlimit))
1208 printf(ARGS(setrlimit, "%d, %p"),
1209 args->resource, (void *)args->rlim);
1210 #endif
1211
1212 if (args->resource >= LINUX_RLIM_NLIMITS)
1213 return (EINVAL);
1214
1215 which = linux_to_bsd_resource[args->resource];
1216 if (which == -1)
1217 return (EINVAL);
1218
1219 error = copyin(args->rlim, &rlim, sizeof(rlim));
1220 if (error)
1221 return (error);
1222
1223 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1224 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1225 return (kern_setrlimit(td, which, &bsd_rlim));
1226 }
1227
1228 int
1229 linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1230 {
1231 struct l_rlimit rlim;
1232 struct proc *p = td->td_proc;
1233 struct rlimit bsd_rlim;
1234 u_int which;
1235
1236 #ifdef DEBUG
1237 if (ldebug(old_getrlimit))
1238 printf(ARGS(old_getrlimit, "%d, %p"),
1239 args->resource, (void *)args->rlim);
1240 #endif
1241
1242 if (args->resource >= LINUX_RLIM_NLIMITS)
1243 return (EINVAL);
1244
1245 which = linux_to_bsd_resource[args->resource];
1246 if (which == -1)
1247 return (EINVAL);
1248
1249 PROC_LOCK(p);
1250 lim_rlimit(p, which, &bsd_rlim);
1251 PROC_UNLOCK(p);
1252
1253 #ifdef COMPAT_LINUX32
1254 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1255 if (rlim.rlim_cur == UINT_MAX)
1256 rlim.rlim_cur = INT_MAX;
1257 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1258 if (rlim.rlim_max == UINT_MAX)
1259 rlim.rlim_max = INT_MAX;
1260 #else
1261 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1262 if (rlim.rlim_cur == ULONG_MAX)
1263 rlim.rlim_cur = LONG_MAX;
1264 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1265 if (rlim.rlim_max == ULONG_MAX)
1266 rlim.rlim_max = LONG_MAX;
1267 #endif
1268 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1269 }
1270
1271 int
1272 linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1273 {
1274 struct l_rlimit rlim;
1275 struct proc *p = td->td_proc;
1276 struct rlimit bsd_rlim;
1277 u_int which;
1278
1279 #ifdef DEBUG
1280 if (ldebug(getrlimit))
1281 printf(ARGS(getrlimit, "%d, %p"),
1282 args->resource, (void *)args->rlim);
1283 #endif
1284
1285 if (args->resource >= LINUX_RLIM_NLIMITS)
1286 return (EINVAL);
1287
1288 which = linux_to_bsd_resource[args->resource];
1289 if (which == -1)
1290 return (EINVAL);
1291
1292 PROC_LOCK(p);
1293 lim_rlimit(p, which, &bsd_rlim);
1294 PROC_UNLOCK(p);
1295
1296 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1297 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1298 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1299 }
1300
1301 int
1302 linux_sched_setscheduler(struct thread *td,
1303 struct linux_sched_setscheduler_args *args)
1304 {
1305 struct sched_setscheduler_args bsd;
1306
1307 #ifdef DEBUG
1308 if (ldebug(sched_setscheduler))
1309 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1310 args->pid, args->policy, (const void *)args->param);
1311 #endif
1312
1313 switch (args->policy) {
1314 case LINUX_SCHED_OTHER:
1315 bsd.policy = SCHED_OTHER;
1316 break;
1317 case LINUX_SCHED_FIFO:
1318 bsd.policy = SCHED_FIFO;
1319 break;
1320 case LINUX_SCHED_RR:
1321 bsd.policy = SCHED_RR;
1322 break;
1323 default:
1324 return (EINVAL);
1325 }
1326
1327 bsd.pid = args->pid;
1328 bsd.param = (struct sched_param *)args->param;
1329 return (sys_sched_setscheduler(td, &bsd));
1330 }
1331
1332 int
1333 linux_sched_getscheduler(struct thread *td,
1334 struct linux_sched_getscheduler_args *args)
1335 {
1336 struct sched_getscheduler_args bsd;
1337 int error;
1338
1339 #ifdef DEBUG
1340 if (ldebug(sched_getscheduler))
1341 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1342 #endif
1343
1344 bsd.pid = args->pid;
1345 error = sys_sched_getscheduler(td, &bsd);
1346
1347 switch (td->td_retval[0]) {
1348 case SCHED_OTHER:
1349 td->td_retval[0] = LINUX_SCHED_OTHER;
1350 break;
1351 case SCHED_FIFO:
1352 td->td_retval[0] = LINUX_SCHED_FIFO;
1353 break;
1354 case SCHED_RR:
1355 td->td_retval[0] = LINUX_SCHED_RR;
1356 break;
1357 }
1358
1359 return (error);
1360 }
1361
1362 int
1363 linux_sched_get_priority_max(struct thread *td,
1364 struct linux_sched_get_priority_max_args *args)
1365 {
1366 struct sched_get_priority_max_args bsd;
1367
1368 #ifdef DEBUG
1369 if (ldebug(sched_get_priority_max))
1370 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1371 #endif
1372
1373 switch (args->policy) {
1374 case LINUX_SCHED_OTHER:
1375 bsd.policy = SCHED_OTHER;
1376 break;
1377 case LINUX_SCHED_FIFO:
1378 bsd.policy = SCHED_FIFO;
1379 break;
1380 case LINUX_SCHED_RR:
1381 bsd.policy = SCHED_RR;
1382 break;
1383 default:
1384 return (EINVAL);
1385 }
1386 return (sys_sched_get_priority_max(td, &bsd));
1387 }
1388
1389 int
1390 linux_sched_get_priority_min(struct thread *td,
1391 struct linux_sched_get_priority_min_args *args)
1392 {
1393 struct sched_get_priority_min_args bsd;
1394
1395 #ifdef DEBUG
1396 if (ldebug(sched_get_priority_min))
1397 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1398 #endif
1399
1400 switch (args->policy) {
1401 case LINUX_SCHED_OTHER:
1402 bsd.policy = SCHED_OTHER;
1403 break;
1404 case LINUX_SCHED_FIFO:
1405 bsd.policy = SCHED_FIFO;
1406 break;
1407 case LINUX_SCHED_RR:
1408 bsd.policy = SCHED_RR;
1409 break;
1410 default:
1411 return (EINVAL);
1412 }
1413 return (sys_sched_get_priority_min(td, &bsd));
1414 }
1415
1416 #define REBOOT_CAD_ON 0x89abcdef
1417 #define REBOOT_CAD_OFF 0
1418 #define REBOOT_HALT 0xcdef0123
1419 #define REBOOT_RESTART 0x01234567
1420 #define REBOOT_RESTART2 0xA1B2C3D4
1421 #define REBOOT_POWEROFF 0x4321FEDC
1422 #define REBOOT_MAGIC1 0xfee1dead
1423 #define REBOOT_MAGIC2 0x28121969
1424 #define REBOOT_MAGIC2A 0x05121996
1425 #define REBOOT_MAGIC2B 0x16041998
1426
1427 int
1428 linux_reboot(struct thread *td, struct linux_reboot_args *args)
1429 {
1430 struct reboot_args bsd_args;
1431
1432 #ifdef DEBUG
1433 if (ldebug(reboot))
1434 printf(ARGS(reboot, "0x%x"), args->cmd);
1435 #endif
1436
1437 if (args->magic1 != REBOOT_MAGIC1)
1438 return (EINVAL);
1439
1440 switch (args->magic2) {
1441 case REBOOT_MAGIC2:
1442 case REBOOT_MAGIC2A:
1443 case REBOOT_MAGIC2B:
1444 break;
1445 default:
1446 return (EINVAL);
1447 }
1448
1449 switch (args->cmd) {
1450 case REBOOT_CAD_ON:
1451 case REBOOT_CAD_OFF:
1452 return (priv_check(td, PRIV_REBOOT));
1453 case REBOOT_HALT:
1454 bsd_args.opt = RB_HALT;
1455 break;
1456 case REBOOT_RESTART:
1457 case REBOOT_RESTART2:
1458 bsd_args.opt = 0;
1459 break;
1460 case REBOOT_POWEROFF:
1461 bsd_args.opt = RB_POWEROFF;
1462 break;
1463 default:
1464 return (EINVAL);
1465 }
1466 return (sys_reboot(td, &bsd_args));
1467 }
1468
1469
1470 /*
1471 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1472 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1473 * are assumed to be preserved. The following lightweight syscalls fixes
1474 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1475 *
1476 * linux_getpid() - MP SAFE
1477 * linux_getgid() - MP SAFE
1478 * linux_getuid() - MP SAFE
1479 */
1480
1481 int
1482 linux_getpid(struct thread *td, struct linux_getpid_args *args)
1483 {
1484 struct linux_emuldata *em;
1485
1486 #ifdef DEBUG
1487 if (ldebug(getpid))
1488 printf(ARGS(getpid, ""));
1489 #endif
1490
1491 if (linux_use26(td)) {
1492 em = em_find(td->td_proc, EMUL_DONTLOCK);
1493 KASSERT(em != NULL, ("getpid: emuldata not found.\n"));
1494 td->td_retval[0] = em->shared->group_pid;
1495 } else {
1496 td->td_retval[0] = td->td_proc->p_pid;
1497 }
1498
1499 return (0);
1500 }
1501
1502 int
1503 linux_gettid(struct thread *td, struct linux_gettid_args *args)
1504 {
1505
1506 #ifdef DEBUG
1507 if (ldebug(gettid))
1508 printf(ARGS(gettid, ""));
1509 #endif
1510
1511 td->td_retval[0] = td->td_proc->p_pid;
1512 return (0);
1513 }
1514
1515
1516 int
1517 linux_getppid(struct thread *td, struct linux_getppid_args *args)
1518 {
1519 struct linux_emuldata *em;
1520 struct proc *p, *pp;
1521
1522 #ifdef DEBUG
1523 if (ldebug(getppid))
1524 printf(ARGS(getppid, ""));
1525 #endif
1526
1527 if (!linux_use26(td)) {
1528 PROC_LOCK(td->td_proc);
1529 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1530 PROC_UNLOCK(td->td_proc);
1531 return (0);
1532 }
1533
1534 em = em_find(td->td_proc, EMUL_DONTLOCK);
1535
1536 KASSERT(em != NULL, ("getppid: process emuldata not found.\n"));
1537
1538 /* find the group leader */
1539 p = pfind(em->shared->group_pid);
1540
1541 if (p == NULL) {
1542 #ifdef DEBUG
1543 printf(LMSG("parent process not found.\n"));
1544 #endif
1545 return (0);
1546 }
1547
1548 pp = p->p_pptr; /* switch to parent */
1549 PROC_LOCK(pp);
1550 PROC_UNLOCK(p);
1551
1552 /* if its also linux process */
1553 if (pp->p_sysent == &elf_linux_sysvec) {
1554 em = em_find(pp, EMUL_DONTLOCK);
1555 KASSERT(em != NULL, ("getppid: parent emuldata not found.\n"));
1556
1557 td->td_retval[0] = em->shared->group_pid;
1558 } else
1559 td->td_retval[0] = pp->p_pid;
1560
1561 PROC_UNLOCK(pp);
1562
1563 return (0);
1564 }
1565
1566 int
1567 linux_getgid(struct thread *td, struct linux_getgid_args *args)
1568 {
1569
1570 #ifdef DEBUG
1571 if (ldebug(getgid))
1572 printf(ARGS(getgid, ""));
1573 #endif
1574
1575 td->td_retval[0] = td->td_ucred->cr_rgid;
1576 return (0);
1577 }
1578
1579 int
1580 linux_getuid(struct thread *td, struct linux_getuid_args *args)
1581 {
1582
1583 #ifdef DEBUG
1584 if (ldebug(getuid))
1585 printf(ARGS(getuid, ""));
1586 #endif
1587
1588 td->td_retval[0] = td->td_ucred->cr_ruid;
1589 return (0);
1590 }
1591
1592
1593 int
1594 linux_getsid(struct thread *td, struct linux_getsid_args *args)
1595 {
1596 struct getsid_args bsd;
1597
1598 #ifdef DEBUG
1599 if (ldebug(getsid))
1600 printf(ARGS(getsid, "%i"), args->pid);
1601 #endif
1602
1603 bsd.pid = args->pid;
1604 return (sys_getsid(td, &bsd));
1605 }
1606
1607 int
1608 linux_nosys(struct thread *td, struct nosys_args *ignore)
1609 {
1610
1611 return (ENOSYS);
1612 }
1613
1614 int
1615 linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1616 {
1617 struct getpriority_args bsd_args;
1618 int error;
1619
1620 #ifdef DEBUG
1621 if (ldebug(getpriority))
1622 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1623 #endif
1624
1625 bsd_args.which = args->which;
1626 bsd_args.who = args->who;
1627 error = sys_getpriority(td, &bsd_args);
1628 td->td_retval[0] = 20 - td->td_retval[0];
1629 return (error);
1630 }
1631
1632 int
1633 linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1634 {
1635 int name[2];
1636
1637 #ifdef DEBUG
1638 if (ldebug(sethostname))
1639 printf(ARGS(sethostname, "*, %i"), args->len);
1640 #endif
1641
1642 name[0] = CTL_KERN;
1643 name[1] = KERN_HOSTNAME;
1644 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1645 args->len, 0, 0));
1646 }
1647
1648 int
1649 linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1650 {
1651 int name[2];
1652
1653 #ifdef DEBUG
1654 if (ldebug(setdomainname))
1655 printf(ARGS(setdomainname, "*, %i"), args->len);
1656 #endif
1657
1658 name[0] = CTL_KERN;
1659 name[1] = KERN_NISDOMAINNAME;
1660 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1661 args->len, 0, 0));
1662 }
1663
1664 int
1665 linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1666 {
1667 struct linux_emuldata *em;
1668
1669 #ifdef DEBUG
1670 if (ldebug(exit_group))
1671 printf(ARGS(exit_group, "%i"), args->error_code);
1672 #endif
1673
1674 em = em_find(td->td_proc, EMUL_DONTLOCK);
1675 if (em->shared->refs > 1) {
1676 EMUL_SHARED_WLOCK(&emul_shared_lock);
1677 em->shared->flags |= EMUL_SHARED_HASXSTAT;
1678 em->shared->xstat = W_EXITCODE(args->error_code, 0);
1679 EMUL_SHARED_WUNLOCK(&emul_shared_lock);
1680 if (linux_use26(td))
1681 linux_kill_threads(td, SIGKILL);
1682 }
1683
1684 /*
1685 * XXX: we should send a signal to the parent if
1686 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1687 * as it doesnt occur often.
1688 */
1689 exit1(td, W_EXITCODE(args->error_code, 0));
1690
1691 return (0);
1692 }
1693
1694 #define _LINUX_CAPABILITY_VERSION 0x19980330
1695
1696 struct l_user_cap_header {
1697 l_int version;
1698 l_int pid;
1699 };
1700
1701 struct l_user_cap_data {
1702 l_int effective;
1703 l_int permitted;
1704 l_int inheritable;
1705 };
1706
1707 int
1708 linux_capget(struct thread *td, struct linux_capget_args *args)
1709 {
1710 struct l_user_cap_header luch;
1711 struct l_user_cap_data lucd;
1712 int error;
1713
1714 if (args->hdrp == NULL)
1715 return (EFAULT);
1716
1717 error = copyin(args->hdrp, &luch, sizeof(luch));
1718 if (error != 0)
1719 return (error);
1720
1721 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1722 luch.version = _LINUX_CAPABILITY_VERSION;
1723 error = copyout(&luch, args->hdrp, sizeof(luch));
1724 if (error)
1725 return (error);
1726 return (EINVAL);
1727 }
1728
1729 if (luch.pid)
1730 return (EPERM);
1731
1732 if (args->datap) {
1733 /*
1734 * The current implementation doesn't support setting
1735 * a capability (it's essentially a stub) so indicate
1736 * that no capabilities are currently set or available
1737 * to request.
1738 */
1739 bzero (&lucd, sizeof(lucd));
1740 error = copyout(&lucd, args->datap, sizeof(lucd));
1741 }
1742
1743 return (error);
1744 }
1745
1746 int
1747 linux_capset(struct thread *td, struct linux_capset_args *args)
1748 {
1749 struct l_user_cap_header luch;
1750 struct l_user_cap_data lucd;
1751 int error;
1752
1753 if (args->hdrp == NULL || args->datap == NULL)
1754 return (EFAULT);
1755
1756 error = copyin(args->hdrp, &luch, sizeof(luch));
1757 if (error != 0)
1758 return (error);
1759
1760 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1761 luch.version = _LINUX_CAPABILITY_VERSION;
1762 error = copyout(&luch, args->hdrp, sizeof(luch));
1763 if (error)
1764 return (error);
1765 return (EINVAL);
1766 }
1767
1768 if (luch.pid)
1769 return (EPERM);
1770
1771 error = copyin(args->datap, &lucd, sizeof(lucd));
1772 if (error != 0)
1773 return (error);
1774
1775 /* We currently don't support setting any capabilities. */
1776 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1777 linux_msg(td,
1778 "capset effective=0x%x, permitted=0x%x, "
1779 "inheritable=0x%x is not implemented",
1780 (int)lucd.effective, (int)lucd.permitted,
1781 (int)lucd.inheritable);
1782 return (EPERM);
1783 }
1784
1785 return (0);
1786 }
1787
1788 int
1789 linux_prctl(struct thread *td, struct linux_prctl_args *args)
1790 {
1791 int error = 0, max_size;
1792 struct proc *p = td->td_proc;
1793 char comm[LINUX_MAX_COMM_LEN];
1794 struct linux_emuldata *em;
1795 int pdeath_signal;
1796
1797 #ifdef DEBUG
1798 if (ldebug(prctl))
1799 printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option,
1800 args->arg2, args->arg3, args->arg4, args->arg5);
1801 #endif
1802
1803 switch (args->option) {
1804 case LINUX_PR_SET_PDEATHSIG:
1805 if (!LINUX_SIG_VALID(args->arg2))
1806 return (EINVAL);
1807 em = em_find(p, EMUL_DOLOCK);
1808 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1809 em->pdeath_signal = args->arg2;
1810 EMUL_UNLOCK(&emul_lock);
1811 break;
1812 case LINUX_PR_GET_PDEATHSIG:
1813 em = em_find(p, EMUL_DOLOCK);
1814 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1815 pdeath_signal = em->pdeath_signal;
1816 EMUL_UNLOCK(&emul_lock);
1817 error = copyout(&pdeath_signal,
1818 (void *)(register_t)args->arg2,
1819 sizeof(pdeath_signal));
1820 break;
1821 case LINUX_PR_GET_KEEPCAPS:
1822 /*
1823 * Indicate that we always clear the effective and
1824 * permitted capability sets when the user id becomes
1825 * non-zero (actually the capability sets are simply
1826 * always zero in the current implementation).
1827 */
1828 td->td_retval[0] = 0;
1829 break;
1830 case LINUX_PR_SET_KEEPCAPS:
1831 /*
1832 * Ignore requests to keep the effective and permitted
1833 * capability sets when the user id becomes non-zero.
1834 */
1835 break;
1836 case LINUX_PR_SET_NAME:
1837 /*
1838 * To be on the safe side we need to make sure to not
1839 * overflow the size a linux program expects. We already
1840 * do this here in the copyin, so that we don't need to
1841 * check on copyout.
1842 */
1843 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1844 error = copyinstr((void *)(register_t)args->arg2, comm,
1845 max_size, NULL);
1846
1847 /* Linux silently truncates the name if it is too long. */
1848 if (error == ENAMETOOLONG) {
1849 /*
1850 * XXX: copyinstr() isn't documented to populate the
1851 * array completely, so do a copyin() to be on the
1852 * safe side. This should be changed in case
1853 * copyinstr() is changed to guarantee this.
1854 */
1855 error = copyin((void *)(register_t)args->arg2, comm,
1856 max_size - 1);
1857 comm[max_size - 1] = '\0';
1858 }
1859 if (error)
1860 return (error);
1861
1862 PROC_LOCK(p);
1863 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1864 PROC_UNLOCK(p);
1865 break;
1866 case LINUX_PR_GET_NAME:
1867 PROC_LOCK(p);
1868 strlcpy(comm, p->p_comm, sizeof(comm));
1869 PROC_UNLOCK(p);
1870 error = copyout(comm, (void *)(register_t)args->arg2,
1871 strlen(comm) + 1);
1872 break;
1873 default:
1874 error = EINVAL;
1875 break;
1876 }
1877
1878 return (error);
1879 }
1880
1881 /*
1882 * Get affinity of a process.
1883 */
1884 int
1885 linux_sched_getaffinity(struct thread *td,
1886 struct linux_sched_getaffinity_args *args)
1887 {
1888 int error;
1889 struct cpuset_getaffinity_args cga;
1890
1891 #ifdef DEBUG
1892 if (ldebug(sched_getaffinity))
1893 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1894 args->len);
1895 #endif
1896 if (args->len < sizeof(cpuset_t))
1897 return (EINVAL);
1898
1899 cga.level = CPU_LEVEL_WHICH;
1900 cga.which = CPU_WHICH_PID;
1901 cga.id = args->pid;
1902 cga.cpusetsize = sizeof(cpuset_t);
1903 cga.mask = (cpuset_t *) args->user_mask_ptr;
1904
1905 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
1906 td->td_retval[0] = sizeof(cpuset_t);
1907
1908 return (error);
1909 }
1910
1911 /*
1912 * Set affinity of a process.
1913 */
1914 int
1915 linux_sched_setaffinity(struct thread *td,
1916 struct linux_sched_setaffinity_args *args)
1917 {
1918 struct cpuset_setaffinity_args csa;
1919
1920 #ifdef DEBUG
1921 if (ldebug(sched_setaffinity))
1922 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
1923 args->len);
1924 #endif
1925 if (args->len < sizeof(cpuset_t))
1926 return (EINVAL);
1927
1928 csa.level = CPU_LEVEL_WHICH;
1929 csa.which = CPU_WHICH_PID;
1930 csa.id = args->pid;
1931 csa.cpusetsize = sizeof(cpuset_t);
1932 csa.mask = (cpuset_t *) args->user_mask_ptr;
1933
1934 return (sys_cpuset_setaffinity(td, &csa));
1935 }
Cache object: d5db8541bc39cda4af94b06d027cbba9
|