32
33#include "opt_compat.h"
34
35#include <sys/param.h>
36#include <sys/blist.h>
37#include <sys/fcntl.h>
38#if defined(__i386__)
39#include <sys/imgact_aout.h>
40#endif
41#include <sys/jail.h>
42#include <sys/kernel.h>
43#include <sys/limits.h>
44#include <sys/lock.h>
45#include <sys/malloc.h>
46#include <sys/mman.h>
47#include <sys/mount.h>
48#include <sys/mutex.h>
49#include <sys/namei.h>
50#include <sys/priv.h>
51#include <sys/proc.h>
52#include <sys/reboot.h>
53#include <sys/racct.h>
54#include <sys/resourcevar.h>
55#include <sys/sched.h>
56#include <sys/signalvar.h>
57#include <sys/stat.h>
58#include <sys/syscallsubr.h>
59#include <sys/sysctl.h>
60#include <sys/sysproto.h>
61#include <sys/systm.h>
62#include <sys/time.h>
63#include <sys/vmmeter.h>
64#include <sys/vnode.h>
65#include <sys/wait.h>
66#include <sys/cpuset.h>
67
68#include <security/mac/mac_framework.h>
69
70#include <vm/vm.h>
71#include <vm/pmap.h>
72#include <vm/vm_kern.h>
73#include <vm/vm_map.h>
74#include <vm/vm_extern.h>
75#include <vm/vm_object.h>
76#include <vm/swap_pager.h>
77
78#ifdef COMPAT_LINUX32
79#include <machine/../linux32/linux.h>
80#include <machine/../linux32/linux32_proto.h>
81#else
82#include <machine/../linux/linux.h>
83#include <machine/../linux/linux_proto.h>
84#endif
85
86#include <compat/linux/linux_file.h>
87#include <compat/linux/linux_mib.h>
88#include <compat/linux/linux_signal.h>
89#include <compat/linux/linux_util.h>
90#include <compat/linux/linux_sysproto.h>
91#include <compat/linux/linux_emul.h>
92#include <compat/linux/linux_misc.h>
93
94int stclohz; /* Statistics clock frequency */
95
96static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
97 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
98 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
99 RLIMIT_MEMLOCK, RLIMIT_AS
100};
101
102struct l_sysinfo {
103 l_long uptime; /* Seconds since boot */
104 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
105#define LINUX_SYSINFO_LOADS_SCALE 65536
106 l_ulong totalram; /* Total usable main memory size */
107 l_ulong freeram; /* Available memory size */
108 l_ulong sharedram; /* Amount of shared memory */
109 l_ulong bufferram; /* Memory used by buffers */
110 l_ulong totalswap; /* Total swap space size */
111 l_ulong freeswap; /* swap space still available */
112 l_ushort procs; /* Number of current processes */
113 l_ushort pads;
114 l_ulong totalbig;
115 l_ulong freebig;
116 l_uint mem_unit;
117 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
118};
119
120struct l_pselect6arg {
121 l_uintptr_t ss;
122 l_size_t ss_len;
123};
124
125int
126linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
127{
128 struct l_sysinfo sysinfo;
129 vm_object_t object;
130 int i, j;
131 struct timespec ts;
132
133 getnanouptime(&ts);
134 if (ts.tv_nsec != 0)
135 ts.tv_sec++;
136 sysinfo.uptime = ts.tv_sec;
137
138 /* Use the information from the mib to get our load averages */
139 for (i = 0; i < 3; i++)
140 sysinfo.loads[i] = averunnable.ldavg[i] *
141 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
142
143 sysinfo.totalram = physmem * PAGE_SIZE;
144 sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
145
146 sysinfo.sharedram = 0;
147 mtx_lock(&vm_object_list_mtx);
148 TAILQ_FOREACH(object, &vm_object_list, object_list)
149 if (object->shadow_count > 1)
150 sysinfo.sharedram += object->resident_page_count;
151 mtx_unlock(&vm_object_list_mtx);
152
153 sysinfo.sharedram *= PAGE_SIZE;
154 sysinfo.bufferram = 0;
155
156 swap_pager_status(&i, &j);
157 sysinfo.totalswap = i * PAGE_SIZE;
158 sysinfo.freeswap = (i - j) * PAGE_SIZE;
159
160 sysinfo.procs = nprocs;
161
162 /* The following are only present in newer Linux kernels. */
163 sysinfo.totalbig = 0;
164 sysinfo.freebig = 0;
165 sysinfo.mem_unit = 1;
166
167 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
168}
169
170int
171linux_alarm(struct thread *td, struct linux_alarm_args *args)
172{
173 struct itimerval it, old_it;
174 u_int secs;
175 int error;
176
177#ifdef DEBUG
178 if (ldebug(alarm))
179 printf(ARGS(alarm, "%u"), args->secs);
180#endif
181
182 secs = args->secs;
183
184 if (secs > INT_MAX)
185 secs = INT_MAX;
186
187 it.it_value.tv_sec = (long) secs;
188 it.it_value.tv_usec = 0;
189 it.it_interval.tv_sec = 0;
190 it.it_interval.tv_usec = 0;
191 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
192 if (error)
193 return (error);
194 if (timevalisset(&old_it.it_value)) {
195 if (old_it.it_value.tv_usec != 0)
196 old_it.it_value.tv_sec++;
197 td->td_retval[0] = old_it.it_value.tv_sec;
198 }
199 return (0);
200}
201
202int
203linux_brk(struct thread *td, struct linux_brk_args *args)
204{
205 struct vmspace *vm = td->td_proc->p_vmspace;
206 vm_offset_t new, old;
207 struct obreak_args /* {
208 char * nsize;
209 } */ tmp;
210
211#ifdef DEBUG
212 if (ldebug(brk))
213 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
214#endif
215 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
216 new = (vm_offset_t)args->dsend;
217 tmp.nsize = (char *)new;
218 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
219 td->td_retval[0] = (long)new;
220 else
221 td->td_retval[0] = (long)old;
222
223 return (0);
224}
225
226#if defined(__i386__)
227/* XXX: what about amd64/linux32? */
228
229int
230linux_uselib(struct thread *td, struct linux_uselib_args *args)
231{
232 struct nameidata ni;
233 struct vnode *vp;
234 struct exec *a_out;
235 struct vattr attr;
236 vm_offset_t vmaddr;
237 unsigned long file_offset;
238 unsigned long bss_size;
239 char *library;
240 ssize_t aresid;
241 int error, locked, writecount;
242
243 LCONVPATHEXIST(td, args->library, &library);
244
245#ifdef DEBUG
246 if (ldebug(uselib))
247 printf(ARGS(uselib, "%s"), library);
248#endif
249
250 a_out = NULL;
251 locked = 0;
252 vp = NULL;
253
254 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
255 UIO_SYSSPACE, library, td);
256 error = namei(&ni);
257 LFREEPATH(library);
258 if (error)
259 goto cleanup;
260
261 vp = ni.ni_vp;
262 NDFREE(&ni, NDF_ONLY_PNBUF);
263
264 /*
265 * From here on down, we have a locked vnode that must be unlocked.
266 * XXX: The code below largely duplicates exec_check_permissions().
267 */
268 locked = 1;
269
270 /* Writable? */
271 error = VOP_GET_WRITECOUNT(vp, &writecount);
272 if (error != 0)
273 goto cleanup;
274 if (writecount != 0) {
275 error = ETXTBSY;
276 goto cleanup;
277 }
278
279 /* Executable? */
280 error = VOP_GETATTR(vp, &attr, td->td_ucred);
281 if (error)
282 goto cleanup;
283
284 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
285 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
286 /* EACCESS is what exec(2) returns. */
287 error = ENOEXEC;
288 goto cleanup;
289 }
290
291 /* Sensible size? */
292 if (attr.va_size == 0) {
293 error = ENOEXEC;
294 goto cleanup;
295 }
296
297 /* Can we access it? */
298 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
299 if (error)
300 goto cleanup;
301
302 /*
303 * XXX: This should use vn_open() so that it is properly authorized,
304 * and to reduce code redundancy all over the place here.
305 * XXX: Not really, it duplicates far more of exec_check_permissions()
306 * than vn_open().
307 */
308#ifdef MAC
309 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
310 if (error)
311 goto cleanup;
312#endif
313 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
314 if (error)
315 goto cleanup;
316
317 /* Pull in executable header into exec_map */
318 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
319 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
320 if (error)
321 goto cleanup;
322
323 /* Is it a Linux binary ? */
324 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
325 error = ENOEXEC;
326 goto cleanup;
327 }
328
329 /*
330 * While we are here, we should REALLY do some more checks
331 */
332
333 /* Set file/virtual offset based on a.out variant. */
334 switch ((int)(a_out->a_magic & 0xffff)) {
335 case 0413: /* ZMAGIC */
336 file_offset = 1024;
337 break;
338 case 0314: /* QMAGIC */
339 file_offset = 0;
340 break;
341 default:
342 error = ENOEXEC;
343 goto cleanup;
344 }
345
346 bss_size = round_page(a_out->a_bss);
347
348 /* Check various fields in header for validity/bounds. */
349 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
350 error = ENOEXEC;
351 goto cleanup;
352 }
353
354 /* text + data can't exceed file size */
355 if (a_out->a_data + a_out->a_text > attr.va_size) {
356 error = EFAULT;
357 goto cleanup;
358 }
359
360 /*
361 * text/data/bss must not exceed limits
362 * XXX - this is not complete. it should check current usage PLUS
363 * the resources needed by this library.
364 */
365 PROC_LOCK(td->td_proc);
366 if (a_out->a_text > maxtsiz ||
367 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
368 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
369 bss_size) != 0) {
370 PROC_UNLOCK(td->td_proc);
371 error = ENOMEM;
372 goto cleanup;
373 }
374 PROC_UNLOCK(td->td_proc);
375
376 /*
377 * Prevent more writers.
378 * XXX: Note that if any of the VM operations fail below we don't
379 * clear this flag.
380 */
381 VOP_SET_TEXT(vp);
382
383 /*
384 * Lock no longer needed
385 */
386 locked = 0;
387 VOP_UNLOCK(vp, 0);
388
389 /*
390 * Check if file_offset page aligned. Currently we cannot handle
391 * misalinged file offsets, and so we read in the entire image
392 * (what a waste).
393 */
394 if (file_offset & PAGE_MASK) {
395#ifdef DEBUG
396 printf("uselib: Non page aligned binary %lu\n", file_offset);
397#endif
398 /* Map text+data read/write/execute */
399
400 /* a_entry is the load address and is page aligned */
401 vmaddr = trunc_page(a_out->a_entry);
402
403 /* get anon user mapping, read+write+execute */
404 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
405 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
406 VM_PROT_ALL, VM_PROT_ALL, 0);
407 if (error)
408 goto cleanup;
409
410 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
411 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
412 td->td_ucred, NOCRED, &aresid, td);
413 if (error != 0)
414 goto cleanup;
415 if (aresid != 0) {
416 error = ENOEXEC;
417 goto cleanup;
418 }
419 } else {
420#ifdef DEBUG
421 printf("uselib: Page aligned binary %lu\n", file_offset);
422#endif
423 /*
424 * for QMAGIC, a_entry is 20 bytes beyond the load address
425 * to skip the executable header
426 */
427 vmaddr = trunc_page(a_out->a_entry);
428
429 /*
430 * Map it all into the process's space as a single
431 * copy-on-write "data" segment.
432 */
433 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
434 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
435 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
436 if (error)
437 goto cleanup;
438 }
439#ifdef DEBUG
440 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
441 ((long *)vmaddr)[1]);
442#endif
443 if (bss_size != 0) {
444 /* Calculate BSS start address */
445 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
446 a_out->a_data;
447
448 /* allocate some 'anon' space */
449 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
450 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
451 VM_PROT_ALL, 0);
452 if (error)
453 goto cleanup;
454 }
455
456cleanup:
457 /* Unlock vnode if needed */
458 if (locked)
459 VOP_UNLOCK(vp, 0);
460
461 /* Release the temporary mapping. */
462 if (a_out)
463 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
464
465 return (error);
466}
467
468#endif /* __i386__ */
469
470int
471linux_select(struct thread *td, struct linux_select_args *args)
472{
473 l_timeval ltv;
474 struct timeval tv0, tv1, utv, *tvp;
475 int error;
476
477#ifdef DEBUG
478 if (ldebug(select))
479 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
480 (void *)args->readfds, (void *)args->writefds,
481 (void *)args->exceptfds, (void *)args->timeout);
482#endif
483
484 /*
485 * Store current time for computation of the amount of
486 * time left.
487 */
488 if (args->timeout) {
489 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
490 goto select_out;
491 utv.tv_sec = ltv.tv_sec;
492 utv.tv_usec = ltv.tv_usec;
493#ifdef DEBUG
494 if (ldebug(select))
495 printf(LMSG("incoming timeout (%jd/%ld)"),
496 (intmax_t)utv.tv_sec, utv.tv_usec);
497#endif
498
499 if (itimerfix(&utv)) {
500 /*
501 * The timeval was invalid. Convert it to something
502 * valid that will act as it does under Linux.
503 */
504 utv.tv_sec += utv.tv_usec / 1000000;
505 utv.tv_usec %= 1000000;
506 if (utv.tv_usec < 0) {
507 utv.tv_sec -= 1;
508 utv.tv_usec += 1000000;
509 }
510 if (utv.tv_sec < 0)
511 timevalclear(&utv);
512 }
513 microtime(&tv0);
514 tvp = &utv;
515 } else
516 tvp = NULL;
517
518 error = kern_select(td, args->nfds, args->readfds, args->writefds,
519 args->exceptfds, tvp, sizeof(l_int) * 8);
520
521#ifdef DEBUG
522 if (ldebug(select))
523 printf(LMSG("real select returns %d"), error);
524#endif
525 if (error)
526 goto select_out;
527
528 if (args->timeout) {
529 if (td->td_retval[0]) {
530 /*
531 * Compute how much time was left of the timeout,
532 * by subtracting the current time and the time
533 * before we started the call, and subtracting
534 * that result from the user-supplied value.
535 */
536 microtime(&tv1);
537 timevalsub(&tv1, &tv0);
538 timevalsub(&utv, &tv1);
539 if (utv.tv_sec < 0)
540 timevalclear(&utv);
541 } else
542 timevalclear(&utv);
543#ifdef DEBUG
544 if (ldebug(select))
545 printf(LMSG("outgoing timeout (%jd/%ld)"),
546 (intmax_t)utv.tv_sec, utv.tv_usec);
547#endif
548 ltv.tv_sec = utv.tv_sec;
549 ltv.tv_usec = utv.tv_usec;
550 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
551 goto select_out;
552 }
553
554select_out:
555#ifdef DEBUG
556 if (ldebug(select))
557 printf(LMSG("select_out -> %d"), error);
558#endif
559 return (error);
560}
561
562int
563linux_mremap(struct thread *td, struct linux_mremap_args *args)
564{
565 struct munmap_args /* {
566 void *addr;
567 size_t len;
568 } */ bsd_args;
569 int error = 0;
570
571#ifdef DEBUG
572 if (ldebug(mremap))
573 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
574 (void *)(uintptr_t)args->addr,
575 (unsigned long)args->old_len,
576 (unsigned long)args->new_len,
577 (unsigned long)args->flags);
578#endif
579
580 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
581 td->td_retval[0] = 0;
582 return (EINVAL);
583 }
584
585 /*
586 * Check for the page alignment.
587 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
588 */
589 if (args->addr & PAGE_MASK) {
590 td->td_retval[0] = 0;
591 return (EINVAL);
592 }
593
594 args->new_len = round_page(args->new_len);
595 args->old_len = round_page(args->old_len);
596
597 if (args->new_len > args->old_len) {
598 td->td_retval[0] = 0;
599 return (ENOMEM);
600 }
601
602 if (args->new_len < args->old_len) {
603 bsd_args.addr =
604 (caddr_t)((uintptr_t)args->addr + args->new_len);
605 bsd_args.len = args->old_len - args->new_len;
606 error = sys_munmap(td, &bsd_args);
607 }
608
609 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
610 return (error);
611}
612
613#define LINUX_MS_ASYNC 0x0001
614#define LINUX_MS_INVALIDATE 0x0002
615#define LINUX_MS_SYNC 0x0004
616
617int
618linux_msync(struct thread *td, struct linux_msync_args *args)
619{
620 struct msync_args bsd_args;
621
622 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
623 bsd_args.len = (uintptr_t)args->len;
624 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
625
626 return (sys_msync(td, &bsd_args));
627}
628
629int
630linux_time(struct thread *td, struct linux_time_args *args)
631{
632 struct timeval tv;
633 l_time_t tm;
634 int error;
635
636#ifdef DEBUG
637 if (ldebug(time))
638 printf(ARGS(time, "*"));
639#endif
640
641 microtime(&tv);
642 tm = tv.tv_sec;
643 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
644 return (error);
645 td->td_retval[0] = tm;
646 return (0);
647}
648
649struct l_times_argv {
650 l_clock_t tms_utime;
651 l_clock_t tms_stime;
652 l_clock_t tms_cutime;
653 l_clock_t tms_cstime;
654};
655
656
657/*
658 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
659 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
660 * auxiliary vector entry.
661 */
662#define CLK_TCK 100
663
664#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
665#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
666
667#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
668 CONVNTCK(r) : CONVOTCK(r))
669
670int
671linux_times(struct thread *td, struct linux_times_args *args)
672{
673 struct timeval tv, utime, stime, cutime, cstime;
674 struct l_times_argv tms;
675 struct proc *p;
676 int error;
677
678#ifdef DEBUG
679 if (ldebug(times))
680 printf(ARGS(times, "*"));
681#endif
682
683 if (args->buf != NULL) {
684 p = td->td_proc;
685 PROC_LOCK(p);
686 PROC_STATLOCK(p);
687 calcru(p, &utime, &stime);
688 PROC_STATUNLOCK(p);
689 calccru(p, &cutime, &cstime);
690 PROC_UNLOCK(p);
691
692 tms.tms_utime = CONVTCK(utime);
693 tms.tms_stime = CONVTCK(stime);
694
695 tms.tms_cutime = CONVTCK(cutime);
696 tms.tms_cstime = CONVTCK(cstime);
697
698 if ((error = copyout(&tms, args->buf, sizeof(tms))))
699 return (error);
700 }
701
702 microuptime(&tv);
703 td->td_retval[0] = (int)CONVTCK(tv);
704 return (0);
705}
706
707int
708linux_newuname(struct thread *td, struct linux_newuname_args *args)
709{
710 struct l_new_utsname utsname;
711 char osname[LINUX_MAX_UTSNAME];
712 char osrelease[LINUX_MAX_UTSNAME];
713 char *p;
714
715#ifdef DEBUG
716 if (ldebug(newuname))
717 printf(ARGS(newuname, "*"));
718#endif
719
720 linux_get_osname(td, osname);
721 linux_get_osrelease(td, osrelease);
722
723 bzero(&utsname, sizeof(utsname));
724 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
725 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
726 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
727 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
728 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
729 for (p = utsname.version; *p != '\0'; ++p)
730 if (*p == '\n') {
731 *p = '\0';
732 break;
733 }
734 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
735
736 return (copyout(&utsname, args->buf, sizeof(utsname)));
737}
738
739struct l_utimbuf {
740 l_time_t l_actime;
741 l_time_t l_modtime;
742};
743
744int
745linux_utime(struct thread *td, struct linux_utime_args *args)
746{
747 struct timeval tv[2], *tvp;
748 struct l_utimbuf lut;
749 char *fname;
750 int error;
751
752 LCONVPATHEXIST(td, args->fname, &fname);
753
754#ifdef DEBUG
755 if (ldebug(utime))
756 printf(ARGS(utime, "%s, *"), fname);
757#endif
758
759 if (args->times) {
760 if ((error = copyin(args->times, &lut, sizeof lut))) {
761 LFREEPATH(fname);
762 return (error);
763 }
764 tv[0].tv_sec = lut.l_actime;
765 tv[0].tv_usec = 0;
766 tv[1].tv_sec = lut.l_modtime;
767 tv[1].tv_usec = 0;
768 tvp = tv;
769 } else
770 tvp = NULL;
771
772 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
773 UIO_SYSSPACE);
774 LFREEPATH(fname);
775 return (error);
776}
777
778int
779linux_utimes(struct thread *td, struct linux_utimes_args *args)
780{
781 l_timeval ltv[2];
782 struct timeval tv[2], *tvp = NULL;
783 char *fname;
784 int error;
785
786 LCONVPATHEXIST(td, args->fname, &fname);
787
788#ifdef DEBUG
789 if (ldebug(utimes))
790 printf(ARGS(utimes, "%s, *"), fname);
791#endif
792
793 if (args->tptr != NULL) {
794 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
795 LFREEPATH(fname);
796 return (error);
797 }
798 tv[0].tv_sec = ltv[0].tv_sec;
799 tv[0].tv_usec = ltv[0].tv_usec;
800 tv[1].tv_sec = ltv[1].tv_sec;
801 tv[1].tv_usec = ltv[1].tv_usec;
802 tvp = tv;
803 }
804
805 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
806 tvp, UIO_SYSSPACE);
807 LFREEPATH(fname);
808 return (error);
809}
810
811int
812linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
813{
814 l_timeval ltv[2];
815 struct timeval tv[2], *tvp = NULL;
816 char *fname;
817 int error, dfd;
818
819 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
820 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
821
822#ifdef DEBUG
823 if (ldebug(futimesat))
824 printf(ARGS(futimesat, "%s, *"), fname);
825#endif
826
827 if (args->utimes != NULL) {
828 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
829 LFREEPATH(fname);
830 return (error);
831 }
832 tv[0].tv_sec = ltv[0].tv_sec;
833 tv[0].tv_usec = ltv[0].tv_usec;
834 tv[1].tv_sec = ltv[1].tv_sec;
835 tv[1].tv_usec = ltv[1].tv_usec;
836 tvp = tv;
837 }
838
839 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
840 LFREEPATH(fname);
841 return (error);
842}
843
844int
845linux_common_wait(struct thread *td, int pid, int *status,
846 int options, struct rusage *ru)
847{
848 int error, tmpstat;
849
850 error = kern_wait(td, pid, &tmpstat, options, ru);
851 if (error)
852 return (error);
853
854 if (status) {
855 tmpstat &= 0xffff;
856 if (WIFSIGNALED(tmpstat))
857 tmpstat = (tmpstat & 0xffffff80) |
858 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
859 else if (WIFSTOPPED(tmpstat))
860 tmpstat = (tmpstat & 0xffff00ff) |
861 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
862 error = copyout(&tmpstat, status, sizeof(int));
863 }
864
865 return (error);
866}
867
868#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
869int
870linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
871{
872 int options;
873
874#ifdef DEBUG
875 if (ldebug(waitpid))
876 printf(ARGS(waitpid, "%d, %p, %d"),
877 args->pid, (void *)args->status, args->options);
878#endif
879 /*
880 * this is necessary because the test in kern_wait doesn't work
881 * because we mess with the options here
882 */
883 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
884 return (EINVAL);
885
886 options = (args->options & (WNOHANG | WUNTRACED));
887 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
888 if (args->options & __WCLONE)
889 options |= WLINUXCLONE;
890
891 return (linux_common_wait(td, args->pid, args->status, options, NULL));
892}
893#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
894
895int
896linux_wait4(struct thread *td, struct linux_wait4_args *args)
897{
898 int error, options;
899 struct rusage ru, *rup;
900
901#ifdef DEBUG
902 if (ldebug(wait4))
903 printf(ARGS(wait4, "%d, %p, %d, %p"),
904 args->pid, (void *)args->status, args->options,
905 (void *)args->rusage);
906#endif
907
908 options = (args->options & (WNOHANG | WUNTRACED));
909 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
910 if (args->options & __WCLONE)
911 options |= WLINUXCLONE;
912
913 if (args->rusage != NULL)
914 rup = &ru;
915 else
916 rup = NULL;
917 error = linux_common_wait(td, args->pid, args->status, options, rup);
918 if (error != 0)
919 return (error);
920 if (args->rusage != NULL)
921 error = linux_copyout_rusage(&ru, args->rusage);
922 return (error);
923}
924
925int
926linux_waitid(struct thread *td, struct linux_waitid_args *args)
927{
928 int status, options, sig;
929 struct __wrusage wru;
930 siginfo_t siginfo;
931 l_siginfo_t lsi;
932 idtype_t idtype;
933 struct proc *p;
934 int error;
935
936 options = 0;
937 linux_to_bsd_waitopts(args->options, &options);
938
939 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
940 return (EINVAL);
941 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
942 return (EINVAL);
943
944 switch (args->idtype) {
945 case LINUX_P_ALL:
946 idtype = P_ALL;
947 break;
948 case LINUX_P_PID:
949 if (args->id <= 0)
950 return (EINVAL);
951 idtype = P_PID;
952 break;
953 case LINUX_P_PGID:
954 if (args->id <= 0)
955 return (EINVAL);
956 idtype = P_PGID;
957 break;
958 default:
959 return (EINVAL);
960 }
961
962 error = kern_wait6(td, idtype, args->id, &status, options,
963 &wru, &siginfo);
964 if (error != 0)
965 return (error);
966 if (args->rusage != NULL) {
967 error = linux_copyout_rusage(&wru.wru_children,
968 args->rusage);
969 if (error != 0)
970 return (error);
971 }
972 if (args->info != NULL) {
973 p = td->td_proc;
974 if (td->td_retval[0] == 0)
975 bzero(&lsi, sizeof(lsi));
976 else {
977 sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo);
978 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
979 }
980 error = copyout(&lsi, args->info, sizeof(lsi));
981 }
982 td->td_retval[0] = 0;
983
984 return (error);
985}
986
987int
988linux_mknod(struct thread *td, struct linux_mknod_args *args)
989{
990 char *path;
991 int error;
992
993 LCONVPATHCREAT(td, args->path, &path);
994
995#ifdef DEBUG
996 if (ldebug(mknod))
997 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
998 (uintmax_t)args->dev);
999#endif
1000
1001 switch (args->mode & S_IFMT) {
1002 case S_IFIFO:
1003 case S_IFSOCK:
1004 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1005 args->mode);
1006 break;
1007
1008 case S_IFCHR:
1009 case S_IFBLK:
1010 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1011 args->mode, args->dev);
1012 break;
1013
1014 case S_IFDIR:
1015 error = EPERM;
1016 break;
1017
1018 case 0:
1019 args->mode |= S_IFREG;
1020 /* FALLTHROUGH */
1021 case S_IFREG:
1022 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1023 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1024 if (error == 0)
1025 kern_close(td, td->td_retval[0]);
1026 break;
1027
1028 default:
1029 error = EINVAL;
1030 break;
1031 }
1032 LFREEPATH(path);
1033 return (error);
1034}
1035
1036int
1037linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1038{
1039 char *path;
1040 int error, dfd;
1041
1042 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1043 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1044
1045#ifdef DEBUG
1046 if (ldebug(mknodat))
1047 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1048#endif
1049
1050 switch (args->mode & S_IFMT) {
1051 case S_IFIFO:
1052 case S_IFSOCK:
1053 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1054 break;
1055
1056 case S_IFCHR:
1057 case S_IFBLK:
1058 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1059 args->dev);
1060 break;
1061
1062 case S_IFDIR:
1063 error = EPERM;
1064 break;
1065
1066 case 0:
1067 args->mode |= S_IFREG;
1068 /* FALLTHROUGH */
1069 case S_IFREG:
1070 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1071 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1072 if (error == 0)
1073 kern_close(td, td->td_retval[0]);
1074 break;
1075
1076 default:
1077 error = EINVAL;
1078 break;
1079 }
1080 LFREEPATH(path);
1081 return (error);
1082}
1083
1084/*
1085 * UGH! This is just about the dumbest idea I've ever heard!!
1086 */
1087int
1088linux_personality(struct thread *td, struct linux_personality_args *args)
1089{
1090#ifdef DEBUG
1091 if (ldebug(personality))
1092 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1093#endif
1094 if (args->per != 0)
1095 return (EINVAL);
1096
1097 /* Yes Jim, it's still a Linux... */
1098 td->td_retval[0] = 0;
1099 return (0);
1100}
1101
1102struct l_itimerval {
1103 l_timeval it_interval;
1104 l_timeval it_value;
1105};
1106
1107#define B2L_ITIMERVAL(bip, lip) \
1108 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1109 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1110 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1111 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1112
1113int
1114linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1115{
1116 int error;
1117 struct l_itimerval ls;
1118 struct itimerval aitv, oitv;
1119
1120#ifdef DEBUG
1121 if (ldebug(setitimer))
1122 printf(ARGS(setitimer, "%p, %p"),
1123 (void *)uap->itv, (void *)uap->oitv);
1124#endif
1125
1126 if (uap->itv == NULL) {
1127 uap->itv = uap->oitv;
1128 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1129 }
1130
1131 error = copyin(uap->itv, &ls, sizeof(ls));
1132 if (error != 0)
1133 return (error);
1134 B2L_ITIMERVAL(&aitv, &ls);
1135#ifdef DEBUG
1136 if (ldebug(setitimer)) {
1137 printf("setitimer: value: sec: %jd, usec: %ld\n",
1138 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1139 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1140 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1141 }
1142#endif
1143 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1144 if (error != 0 || uap->oitv == NULL)
1145 return (error);
1146 B2L_ITIMERVAL(&ls, &oitv);
1147
1148 return (copyout(&ls, uap->oitv, sizeof(ls)));
1149}
1150
1151int
1152linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1153{
1154 int error;
1155 struct l_itimerval ls;
1156 struct itimerval aitv;
1157
1158#ifdef DEBUG
1159 if (ldebug(getitimer))
1160 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1161#endif
1162 error = kern_getitimer(td, uap->which, &aitv);
1163 if (error != 0)
1164 return (error);
1165 B2L_ITIMERVAL(&ls, &aitv);
1166 return (copyout(&ls, uap->itv, sizeof(ls)));
1167}
1168
1169#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1170int
1171linux_nice(struct thread *td, struct linux_nice_args *args)
1172{
1173 struct setpriority_args bsd_args;
1174
1175 bsd_args.which = PRIO_PROCESS;
1176 bsd_args.who = 0; /* current process */
1177 bsd_args.prio = args->inc;
1178 return (sys_setpriority(td, &bsd_args));
1179}
1180#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1181
1182int
1183linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1184{
1185 struct ucred *newcred, *oldcred;
1186 l_gid_t *linux_gidset;
1187 gid_t *bsd_gidset;
1188 int ngrp, error;
1189 struct proc *p;
1190
1191 ngrp = args->gidsetsize;
1192 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1193 return (EINVAL);
1194 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1195 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1196 if (error)
1197 goto out;
1198 newcred = crget();
1199 p = td->td_proc;
1200 PROC_LOCK(p);
1201 oldcred = crcopysafe(p, newcred);
1202
1203 /*
1204 * cr_groups[0] holds egid. Setting the whole set from
1205 * the supplied set will cause egid to be changed too.
1206 * Keep cr_groups[0] unchanged to prevent that.
1207 */
1208
1209 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1210 PROC_UNLOCK(p);
1211 crfree(newcred);
1212 goto out;
1213 }
1214
1215 if (ngrp > 0) {
1216 newcred->cr_ngroups = ngrp + 1;
1217
1218 bsd_gidset = newcred->cr_groups;
1219 ngrp--;
1220 while (ngrp >= 0) {
1221 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1222 ngrp--;
1223 }
1224 } else
1225 newcred->cr_ngroups = 1;
1226
1227 setsugid(p);
1228 proc_set_cred(p, newcred);
1229 PROC_UNLOCK(p);
1230 crfree(oldcred);
1231 error = 0;
1232out:
1233 free(linux_gidset, M_TEMP);
1234 return (error);
1235}
1236
1237int
1238linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1239{
1240 struct ucred *cred;
1241 l_gid_t *linux_gidset;
1242 gid_t *bsd_gidset;
1243 int bsd_gidsetsz, ngrp, error;
1244
1245 cred = td->td_ucred;
1246 bsd_gidset = cred->cr_groups;
1247 bsd_gidsetsz = cred->cr_ngroups - 1;
1248
1249 /*
1250 * cr_groups[0] holds egid. Returning the whole set
1251 * here will cause a duplicate. Exclude cr_groups[0]
1252 * to prevent that.
1253 */
1254
1255 if ((ngrp = args->gidsetsize) == 0) {
1256 td->td_retval[0] = bsd_gidsetsz;
1257 return (0);
1258 }
1259
1260 if (ngrp < bsd_gidsetsz)
1261 return (EINVAL);
1262
1263 ngrp = 0;
1264 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1265 M_TEMP, M_WAITOK);
1266 while (ngrp < bsd_gidsetsz) {
1267 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1268 ngrp++;
1269 }
1270
1271 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1272 free(linux_gidset, M_TEMP);
1273 if (error)
1274 return (error);
1275
1276 td->td_retval[0] = ngrp;
1277 return (0);
1278}
1279
1280int
1281linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1282{
1283 struct rlimit bsd_rlim;
1284 struct l_rlimit rlim;
1285 u_int which;
1286 int error;
1287
1288#ifdef DEBUG
1289 if (ldebug(setrlimit))
1290 printf(ARGS(setrlimit, "%d, %p"),
1291 args->resource, (void *)args->rlim);
1292#endif
1293
1294 if (args->resource >= LINUX_RLIM_NLIMITS)
1295 return (EINVAL);
1296
1297 which = linux_to_bsd_resource[args->resource];
1298 if (which == -1)
1299 return (EINVAL);
1300
1301 error = copyin(args->rlim, &rlim, sizeof(rlim));
1302 if (error)
1303 return (error);
1304
1305 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1306 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1307 return (kern_setrlimit(td, which, &bsd_rlim));
1308}
1309
1310#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1311int
1312linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1313{
1314 struct l_rlimit rlim;
1315 struct proc *p = td->td_proc;
1316 struct rlimit bsd_rlim;
1317 u_int which;
1318
1319#ifdef DEBUG
1320 if (ldebug(old_getrlimit))
1321 printf(ARGS(old_getrlimit, "%d, %p"),
1322 args->resource, (void *)args->rlim);
1323#endif
1324
1325 if (args->resource >= LINUX_RLIM_NLIMITS)
1326 return (EINVAL);
1327
1328 which = linux_to_bsd_resource[args->resource];
1329 if (which == -1)
1330 return (EINVAL);
1331
1332 PROC_LOCK(p);
1333 lim_rlimit(p, which, &bsd_rlim);
1334 PROC_UNLOCK(p);
1335
1336#ifdef COMPAT_LINUX32
1337 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1338 if (rlim.rlim_cur == UINT_MAX)
1339 rlim.rlim_cur = INT_MAX;
1340 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1341 if (rlim.rlim_max == UINT_MAX)
1342 rlim.rlim_max = INT_MAX;
1343#else
1344 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1345 if (rlim.rlim_cur == ULONG_MAX)
1346 rlim.rlim_cur = LONG_MAX;
1347 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1348 if (rlim.rlim_max == ULONG_MAX)
1349 rlim.rlim_max = LONG_MAX;
1350#endif
1351 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1352}
1353#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1354
1355int
1356linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1357{
1358 struct l_rlimit rlim;
1359 struct proc *p = td->td_proc;
1360 struct rlimit bsd_rlim;
1361 u_int which;
1362
1363#ifdef DEBUG
1364 if (ldebug(getrlimit))
1365 printf(ARGS(getrlimit, "%d, %p"),
1366 args->resource, (void *)args->rlim);
1367#endif
1368
1369 if (args->resource >= LINUX_RLIM_NLIMITS)
1370 return (EINVAL);
1371
1372 which = linux_to_bsd_resource[args->resource];
1373 if (which == -1)
1374 return (EINVAL);
1375
1376 PROC_LOCK(p);
1377 lim_rlimit(p, which, &bsd_rlim);
1378 PROC_UNLOCK(p);
1379
1380 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1381 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1382 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1383}
1384
1385int
1386linux_sched_setscheduler(struct thread *td,
1387 struct linux_sched_setscheduler_args *args)
1388{
1389 struct sched_param sched_param;
1390 struct thread *tdt;
1391 int error, policy;
1392
1393#ifdef DEBUG
1394 if (ldebug(sched_setscheduler))
1395 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1396 args->pid, args->policy, (const void *)args->param);
1397#endif
1398
1399 switch (args->policy) {
1400 case LINUX_SCHED_OTHER:
1401 policy = SCHED_OTHER;
1402 break;
1403 case LINUX_SCHED_FIFO:
1404 policy = SCHED_FIFO;
1405 break;
1406 case LINUX_SCHED_RR:
1407 policy = SCHED_RR;
1408 break;
1409 default:
1410 return (EINVAL);
1411 }
1412
1413 error = copyin(args->param, &sched_param, sizeof(sched_param));
1414 if (error)
1415 return (error);
1416
1417 tdt = linux_tdfind(td, args->pid, -1);
1418 if (tdt == NULL)
1419 return (ESRCH);
1420
1421 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1422 PROC_UNLOCK(tdt->td_proc);
1423 return (error);
1424}
1425
1426int
1427linux_sched_getscheduler(struct thread *td,
1428 struct linux_sched_getscheduler_args *args)
1429{
1430 struct thread *tdt;
1431 int error, policy;
1432
1433#ifdef DEBUG
1434 if (ldebug(sched_getscheduler))
1435 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1436#endif
1437
1438 tdt = linux_tdfind(td, args->pid, -1);
1439 if (tdt == NULL)
1440 return (ESRCH);
1441
1442 error = kern_sched_getscheduler(td, tdt, &policy);
1443 PROC_UNLOCK(tdt->td_proc);
1444
1445 switch (policy) {
1446 case SCHED_OTHER:
1447 td->td_retval[0] = LINUX_SCHED_OTHER;
1448 break;
1449 case SCHED_FIFO:
1450 td->td_retval[0] = LINUX_SCHED_FIFO;
1451 break;
1452 case SCHED_RR:
1453 td->td_retval[0] = LINUX_SCHED_RR;
1454 break;
1455 }
1456 return (error);
1457}
1458
1459int
1460linux_sched_get_priority_max(struct thread *td,
1461 struct linux_sched_get_priority_max_args *args)
1462{
1463 struct sched_get_priority_max_args bsd;
1464
1465#ifdef DEBUG
1466 if (ldebug(sched_get_priority_max))
1467 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1468#endif
1469
1470 switch (args->policy) {
1471 case LINUX_SCHED_OTHER:
1472 bsd.policy = SCHED_OTHER;
1473 break;
1474 case LINUX_SCHED_FIFO:
1475 bsd.policy = SCHED_FIFO;
1476 break;
1477 case LINUX_SCHED_RR:
1478 bsd.policy = SCHED_RR;
1479 break;
1480 default:
1481 return (EINVAL);
1482 }
1483 return (sys_sched_get_priority_max(td, &bsd));
1484}
1485
1486int
1487linux_sched_get_priority_min(struct thread *td,
1488 struct linux_sched_get_priority_min_args *args)
1489{
1490 struct sched_get_priority_min_args bsd;
1491
1492#ifdef DEBUG
1493 if (ldebug(sched_get_priority_min))
1494 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1495#endif
1496
1497 switch (args->policy) {
1498 case LINUX_SCHED_OTHER:
1499 bsd.policy = SCHED_OTHER;
1500 break;
1501 case LINUX_SCHED_FIFO:
1502 bsd.policy = SCHED_FIFO;
1503 break;
1504 case LINUX_SCHED_RR:
1505 bsd.policy = SCHED_RR;
1506 break;
1507 default:
1508 return (EINVAL);
1509 }
1510 return (sys_sched_get_priority_min(td, &bsd));
1511}
1512
1513#define REBOOT_CAD_ON 0x89abcdef
1514#define REBOOT_CAD_OFF 0
1515#define REBOOT_HALT 0xcdef0123
1516#define REBOOT_RESTART 0x01234567
1517#define REBOOT_RESTART2 0xA1B2C3D4
1518#define REBOOT_POWEROFF 0x4321FEDC
1519#define REBOOT_MAGIC1 0xfee1dead
1520#define REBOOT_MAGIC2 0x28121969
1521#define REBOOT_MAGIC2A 0x05121996
1522#define REBOOT_MAGIC2B 0x16041998
1523
1524int
1525linux_reboot(struct thread *td, struct linux_reboot_args *args)
1526{
1527 struct reboot_args bsd_args;
1528
1529#ifdef DEBUG
1530 if (ldebug(reboot))
1531 printf(ARGS(reboot, "0x%x"), args->cmd);
1532#endif
1533
1534 if (args->magic1 != REBOOT_MAGIC1)
1535 return (EINVAL);
1536
1537 switch (args->magic2) {
1538 case REBOOT_MAGIC2:
1539 case REBOOT_MAGIC2A:
1540 case REBOOT_MAGIC2B:
1541 break;
1542 default:
1543 return (EINVAL);
1544 }
1545
1546 switch (args->cmd) {
1547 case REBOOT_CAD_ON:
1548 case REBOOT_CAD_OFF:
1549 return (priv_check(td, PRIV_REBOOT));
1550 case REBOOT_HALT:
1551 bsd_args.opt = RB_HALT;
1552 break;
1553 case REBOOT_RESTART:
1554 case REBOOT_RESTART2:
1555 bsd_args.opt = 0;
1556 break;
1557 case REBOOT_POWEROFF:
1558 bsd_args.opt = RB_POWEROFF;
1559 break;
1560 default:
1561 return (EINVAL);
1562 }
1563 return (sys_reboot(td, &bsd_args));
1564}
1565
1566
1567/*
1568 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1569 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1570 * are assumed to be preserved. The following lightweight syscalls fixes
1571 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1572 *
1573 * linux_getpid() - MP SAFE
1574 * linux_getgid() - MP SAFE
1575 * linux_getuid() - MP SAFE
1576 */
1577
1578int
1579linux_getpid(struct thread *td, struct linux_getpid_args *args)
1580{
1581
1582#ifdef DEBUG
1583 if (ldebug(getpid))
1584 printf(ARGS(getpid, ""));
1585#endif
1586 td->td_retval[0] = td->td_proc->p_pid;
1587
1588 return (0);
1589}
1590
1591int
1592linux_gettid(struct thread *td, struct linux_gettid_args *args)
1593{
1594 struct linux_emuldata *em;
1595
1596#ifdef DEBUG
1597 if (ldebug(gettid))
1598 printf(ARGS(gettid, ""));
1599#endif
1600
1601 em = em_find(td);
1602 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1603
1604 td->td_retval[0] = em->em_tid;
1605
1606 return (0);
1607}
1608
1609
1610int
1611linux_getppid(struct thread *td, struct linux_getppid_args *args)
1612{
1613
1614#ifdef DEBUG
1615 if (ldebug(getppid))
1616 printf(ARGS(getppid, ""));
1617#endif
1618
1619 PROC_LOCK(td->td_proc);
1620 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1621 PROC_UNLOCK(td->td_proc);
1622 return (0);
1623}
1624
1625int
1626linux_getgid(struct thread *td, struct linux_getgid_args *args)
1627{
1628
1629#ifdef DEBUG
1630 if (ldebug(getgid))
1631 printf(ARGS(getgid, ""));
1632#endif
1633
1634 td->td_retval[0] = td->td_ucred->cr_rgid;
1635 return (0);
1636}
1637
1638int
1639linux_getuid(struct thread *td, struct linux_getuid_args *args)
1640{
1641
1642#ifdef DEBUG
1643 if (ldebug(getuid))
1644 printf(ARGS(getuid, ""));
1645#endif
1646
1647 td->td_retval[0] = td->td_ucred->cr_ruid;
1648 return (0);
1649}
1650
1651
1652int
1653linux_getsid(struct thread *td, struct linux_getsid_args *args)
1654{
1655 struct getsid_args bsd;
1656
1657#ifdef DEBUG
1658 if (ldebug(getsid))
1659 printf(ARGS(getsid, "%i"), args->pid);
1660#endif
1661
1662 bsd.pid = args->pid;
1663 return (sys_getsid(td, &bsd));
1664}
1665
1666int
1667linux_nosys(struct thread *td, struct nosys_args *ignore)
1668{
1669
1670 return (ENOSYS);
1671}
1672
1673int
1674linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1675{
1676 struct getpriority_args bsd_args;
1677 int error;
1678
1679#ifdef DEBUG
1680 if (ldebug(getpriority))
1681 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1682#endif
1683
1684 bsd_args.which = args->which;
1685 bsd_args.who = args->who;
1686 error = sys_getpriority(td, &bsd_args);
1687 td->td_retval[0] = 20 - td->td_retval[0];
1688 return (error);
1689}
1690
1691int
1692linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1693{
1694 int name[2];
1695
1696#ifdef DEBUG
1697 if (ldebug(sethostname))
1698 printf(ARGS(sethostname, "*, %i"), args->len);
1699#endif
1700
1701 name[0] = CTL_KERN;
1702 name[1] = KERN_HOSTNAME;
1703 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1704 args->len, 0, 0));
1705}
1706
1707int
1708linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1709{
1710 int name[2];
1711
1712#ifdef DEBUG
1713 if (ldebug(setdomainname))
1714 printf(ARGS(setdomainname, "*, %i"), args->len);
1715#endif
1716
1717 name[0] = CTL_KERN;
1718 name[1] = KERN_NISDOMAINNAME;
1719 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1720 args->len, 0, 0));
1721}
1722
1723int
1724linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1725{
1726
1727#ifdef DEBUG
1728 if (ldebug(exit_group))
1729 printf(ARGS(exit_group, "%i"), args->error_code);
1730#endif
1731
1732 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1733 args->error_code);
1734
1735 /*
1736 * XXX: we should send a signal to the parent if
1737 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1738 * as it doesnt occur often.
1739 */
1740 exit1(td, W_EXITCODE(args->error_code, 0));
1741 /* NOTREACHED */
1742}
1743
1744#define _LINUX_CAPABILITY_VERSION 0x19980330
1745
1746struct l_user_cap_header {
1747 l_int version;
1748 l_int pid;
1749};
1750
1751struct l_user_cap_data {
1752 l_int effective;
1753 l_int permitted;
1754 l_int inheritable;
1755};
1756
1757int
1758linux_capget(struct thread *td, struct linux_capget_args *args)
1759{
1760 struct l_user_cap_header luch;
1761 struct l_user_cap_data lucd;
1762 int error;
1763
1764 if (args->hdrp == NULL)
1765 return (EFAULT);
1766
1767 error = copyin(args->hdrp, &luch, sizeof(luch));
1768 if (error != 0)
1769 return (error);
1770
1771 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1772 luch.version = _LINUX_CAPABILITY_VERSION;
1773 error = copyout(&luch, args->hdrp, sizeof(luch));
1774 if (error)
1775 return (error);
1776 return (EINVAL);
1777 }
1778
1779 if (luch.pid)
1780 return (EPERM);
1781
1782 if (args->datap) {
1783 /*
1784 * The current implementation doesn't support setting
1785 * a capability (it's essentially a stub) so indicate
1786 * that no capabilities are currently set or available
1787 * to request.
1788 */
1789 bzero (&lucd, sizeof(lucd));
1790 error = copyout(&lucd, args->datap, sizeof(lucd));
1791 }
1792
1793 return (error);
1794}
1795
1796int
1797linux_capset(struct thread *td, struct linux_capset_args *args)
1798{
1799 struct l_user_cap_header luch;
1800 struct l_user_cap_data lucd;
1801 int error;
1802
1803 if (args->hdrp == NULL || args->datap == NULL)
1804 return (EFAULT);
1805
1806 error = copyin(args->hdrp, &luch, sizeof(luch));
1807 if (error != 0)
1808 return (error);
1809
1810 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1811 luch.version = _LINUX_CAPABILITY_VERSION;
1812 error = copyout(&luch, args->hdrp, sizeof(luch));
1813 if (error)
1814 return (error);
1815 return (EINVAL);
1816 }
1817
1818 if (luch.pid)
1819 return (EPERM);
1820
1821 error = copyin(args->datap, &lucd, sizeof(lucd));
1822 if (error != 0)
1823 return (error);
1824
1825 /* We currently don't support setting any capabilities. */
1826 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1827 linux_msg(td,
1828 "capset effective=0x%x, permitted=0x%x, "
1829 "inheritable=0x%x is not implemented",
1830 (int)lucd.effective, (int)lucd.permitted,
1831 (int)lucd.inheritable);
1832 return (EPERM);
1833 }
1834
1835 return (0);
1836}
1837
1838int
1839linux_prctl(struct thread *td, struct linux_prctl_args *args)
1840{
1841 int error = 0, max_size;
1842 struct proc *p = td->td_proc;
1843 char comm[LINUX_MAX_COMM_LEN];
1844 struct linux_emuldata *em;
1845 int pdeath_signal;
1846
1847#ifdef DEBUG
1848 if (ldebug(prctl))
1849 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1850 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1851 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1852#endif
1853
1854 switch (args->option) {
1855 case LINUX_PR_SET_PDEATHSIG:
1856 if (!LINUX_SIG_VALID(args->arg2))
1857 return (EINVAL);
1858 em = em_find(td);
1859 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1860 em->pdeath_signal = args->arg2;
1861 break;
1862 case LINUX_PR_GET_PDEATHSIG:
1863 em = em_find(td);
1864 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1865 pdeath_signal = em->pdeath_signal;
1866 error = copyout(&pdeath_signal,
1867 (void *)(register_t)args->arg2,
1868 sizeof(pdeath_signal));
1869 break;
1870 case LINUX_PR_GET_KEEPCAPS:
1871 /*
1872 * Indicate that we always clear the effective and
1873 * permitted capability sets when the user id becomes
1874 * non-zero (actually the capability sets are simply
1875 * always zero in the current implementation).
1876 */
1877 td->td_retval[0] = 0;
1878 break;
1879 case LINUX_PR_SET_KEEPCAPS:
1880 /*
1881 * Ignore requests to keep the effective and permitted
1882 * capability sets when the user id becomes non-zero.
1883 */
1884 break;
1885 case LINUX_PR_SET_NAME:
1886 /*
1887 * To be on the safe side we need to make sure to not
1888 * overflow the size a linux program expects. We already
1889 * do this here in the copyin, so that we don't need to
1890 * check on copyout.
1891 */
1892 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1893 error = copyinstr((void *)(register_t)args->arg2, comm,
1894 max_size, NULL);
1895
1896 /* Linux silently truncates the name if it is too long. */
1897 if (error == ENAMETOOLONG) {
1898 /*
1899 * XXX: copyinstr() isn't documented to populate the
1900 * array completely, so do a copyin() to be on the
1901 * safe side. This should be changed in case
1902 * copyinstr() is changed to guarantee this.
1903 */
1904 error = copyin((void *)(register_t)args->arg2, comm,
1905 max_size - 1);
1906 comm[max_size - 1] = '\0';
1907 }
1908 if (error)
1909 return (error);
1910
1911 PROC_LOCK(p);
1912 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1913 PROC_UNLOCK(p);
1914 break;
1915 case LINUX_PR_GET_NAME:
1916 PROC_LOCK(p);
1917 strlcpy(comm, p->p_comm, sizeof(comm));
1918 PROC_UNLOCK(p);
1919 error = copyout(comm, (void *)(register_t)args->arg2,
1920 strlen(comm) + 1);
1921 break;
1922 default:
1923 error = EINVAL;
1924 break;
1925 }
1926
1927 return (error);
1928}
1929
1930int
1931linux_sched_setparam(struct thread *td,
1932 struct linux_sched_setparam_args *uap)
1933{
1934 struct sched_param sched_param;
1935 struct thread *tdt;
1936 int error;
1937
1938#ifdef DEBUG
1939 if (ldebug(sched_setparam))
1940 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
1941#endif
1942
1943 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1944 if (error)
1945 return (error);
1946
1947 tdt = linux_tdfind(td, uap->pid, -1);
1948 if (tdt == NULL)
1949 return (ESRCH);
1950
1951 error = kern_sched_setparam(td, tdt, &sched_param);
1952 PROC_UNLOCK(tdt->td_proc);
1953 return (error);
1954}
1955
1956int
1957linux_sched_getparam(struct thread *td,
1958 struct linux_sched_getparam_args *uap)
1959{
1960 struct sched_param sched_param;
1961 struct thread *tdt;
1962 int error;
1963
1964#ifdef DEBUG
1965 if (ldebug(sched_getparam))
1966 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
1967#endif
1968
1969 tdt = linux_tdfind(td, uap->pid, -1);
1970 if (tdt == NULL)
1971 return (ESRCH);
1972
1973 error = kern_sched_getparam(td, tdt, &sched_param);
1974 PROC_UNLOCK(tdt->td_proc);
1975 if (error == 0)
1976 error = copyout(&sched_param, uap->param,
1977 sizeof(sched_param));
1978 return (error);
1979}
1980
1981/*
1982 * Get affinity of a process.
1983 */
1984int
1985linux_sched_getaffinity(struct thread *td,
1986 struct linux_sched_getaffinity_args *args)
1987{
1988 int error;
1989 struct thread *tdt;
1990 struct cpuset_getaffinity_args cga;
1991
1992#ifdef DEBUG
1993 if (ldebug(sched_getaffinity))
1994 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1995 args->len);
1996#endif
1997 if (args->len < sizeof(cpuset_t))
1998 return (EINVAL);
1999
2000 tdt = linux_tdfind(td, args->pid, -1);
2001 if (tdt == NULL)
2002 return (ESRCH);
2003
2004 PROC_UNLOCK(tdt->td_proc);
2005 cga.level = CPU_LEVEL_WHICH;
2006 cga.which = CPU_WHICH_TID;
2007 cga.id = tdt->td_tid;
2008 cga.cpusetsize = sizeof(cpuset_t);
2009 cga.mask = (cpuset_t *) args->user_mask_ptr;
2010
2011 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2012 td->td_retval[0] = sizeof(cpuset_t);
2013
2014 return (error);
2015}
2016
2017/*
2018 * Set affinity of a process.
2019 */
2020int
2021linux_sched_setaffinity(struct thread *td,
2022 struct linux_sched_setaffinity_args *args)
2023{
2024 struct cpuset_setaffinity_args csa;
2025 struct thread *tdt;
2026
2027#ifdef DEBUG
2028 if (ldebug(sched_setaffinity))
2029 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2030 args->len);
2031#endif
2032 if (args->len < sizeof(cpuset_t))
2033 return (EINVAL);
2034
2035 tdt = linux_tdfind(td, args->pid, -1);
2036 if (tdt == NULL)
2037 return (ESRCH);
2038
2039 PROC_UNLOCK(tdt->td_proc);
2040 csa.level = CPU_LEVEL_WHICH;
2041 csa.which = CPU_WHICH_TID;
2042 csa.id = tdt->td_tid;
2043 csa.cpusetsize = sizeof(cpuset_t);
2044 csa.mask = (cpuset_t *) args->user_mask_ptr;
2045
2046 return (sys_cpuset_setaffinity(td, &csa));
2047}
2048
2049struct linux_rlimit64 {
2050 uint64_t rlim_cur;
2051 uint64_t rlim_max;
2052};
2053
2054int
2055linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2056{
2057 struct rlimit rlim, nrlim;
2058 struct linux_rlimit64 lrlim;
2059 struct proc *p;
2060 u_int which;
2061 int flags;
2062 int error;
2063
2064#ifdef DEBUG
2065 if (ldebug(prlimit64))
2066 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2067 args->resource, (void *)args->new, (void *)args->old);
2068#endif
2069
2070 if (args->resource >= LINUX_RLIM_NLIMITS)
2071 return (EINVAL);
2072
2073 which = linux_to_bsd_resource[args->resource];
2074 if (which == -1)
2075 return (EINVAL);
2076
2077 if (args->new != NULL) {
2078 /*
2079 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2080 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2081 * as INFINITY so we do not need a conversion even.
2082 */
2083 error = copyin(args->new, &nrlim, sizeof(nrlim));
2084 if (error != 0)
2085 return (error);
2086 }
2087
2088 flags = PGET_HOLD | PGET_NOTWEXIT;
2089 if (args->new != NULL)
2090 flags |= PGET_CANDEBUG;
2091 else
2092 flags |= PGET_CANSEE;
2093 error = pget(args->pid, flags, &p);
2094 if (error != 0)
2095 return (error);
2096
2097 if (args->old != NULL) {
2098 PROC_LOCK(p);
2099 lim_rlimit(p, which, &rlim);
2100 PROC_UNLOCK(p);
2101 if (rlim.rlim_cur == RLIM_INFINITY)
2102 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2103 else
2104 lrlim.rlim_cur = rlim.rlim_cur;
2105 if (rlim.rlim_max == RLIM_INFINITY)
2106 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2107 else
2108 lrlim.rlim_max = rlim.rlim_max;
2109 error = copyout(&lrlim, args->old, sizeof(lrlim));
2110 if (error != 0)
2111 goto out;
2112 }
2113
2114 if (args->new != NULL)
2115 error = kern_proc_setrlimit(td, p, which, &nrlim);
2116
2117 out:
2118 PRELE(p);
2119 return (error);
2120}
2121
2122int
2123linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2124{
2125 struct timeval utv, tv0, tv1, *tvp;
2126 struct l_pselect6arg lpse6;
2127 struct l_timespec lts;
2128 struct timespec uts;
2129 l_sigset_t l_ss;
2130 sigset_t *ssp;
2131 sigset_t ss;
2132 int error;
2133
2134 ssp = NULL;
2135 if (args->sig != NULL) {
2136 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2137 if (error != 0)
2138 return (error);
2139 if (lpse6.ss_len != sizeof(l_ss))
2140 return (EINVAL);
2141 if (lpse6.ss != 0) {
2142 error = copyin(PTRIN(lpse6.ss), &l_ss,
2143 sizeof(l_ss));
2144 if (error != 0)
2145 return (error);
2146 linux_to_bsd_sigset(&l_ss, &ss);
2147 ssp = &ss;
2148 }
2149 }
2150
2151 /*
2152 * Currently glibc changes nanosecond number to microsecond.
2153 * This mean losing precision but for now it is hardly seen.
2154 */
2155 if (args->tsp != NULL) {
2156 error = copyin(args->tsp, <s, sizeof(lts));
2157 if (error != 0)
2158 return (error);
2159 uts.tv_sec = lts.tv_sec;
2160 uts.tv_nsec = lts.tv_nsec;
2161
2162 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2163 if (itimerfix(&utv))
2164 return (EINVAL);
2165
2166 microtime(&tv0);
2167 tvp = &utv;
2168 } else
2169 tvp = NULL;
2170
2171 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2172 args->exceptfds, tvp, ssp, sizeof(l_int) * 8);
2173
2174 if (error == 0 && args->tsp != NULL) {
2175 if (td->td_retval[0] != 0) {
2176 /*
2177 * Compute how much time was left of the timeout,
2178 * by subtracting the current time and the time
2179 * before we started the call, and subtracting
2180 * that result from the user-supplied value.
2181 */
2182
2183 microtime(&tv1);
2184 timevalsub(&tv1, &tv0);
2185 timevalsub(&utv, &tv1);
2186 if (utv.tv_sec < 0)
2187 timevalclear(&utv);
2188 } else
2189 timevalclear(&utv);
2190
2191 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2192 lts.tv_sec = uts.tv_sec;
2193 lts.tv_nsec = uts.tv_nsec;
2194 error = copyout(<s, args->tsp, sizeof(lts));
2195 }
2196
2197 return (error);
2198}
2199
| 32
33#include "opt_compat.h"
34
35#include <sys/param.h>
36#include <sys/blist.h>
37#include <sys/fcntl.h>
38#if defined(__i386__)
39#include <sys/imgact_aout.h>
40#endif
41#include <sys/jail.h>
42#include <sys/kernel.h>
43#include <sys/limits.h>
44#include <sys/lock.h>
45#include <sys/malloc.h>
46#include <sys/mman.h>
47#include <sys/mount.h>
48#include <sys/mutex.h>
49#include <sys/namei.h>
50#include <sys/priv.h>
51#include <sys/proc.h>
52#include <sys/reboot.h>
53#include <sys/racct.h>
54#include <sys/resourcevar.h>
55#include <sys/sched.h>
56#include <sys/signalvar.h>
57#include <sys/stat.h>
58#include <sys/syscallsubr.h>
59#include <sys/sysctl.h>
60#include <sys/sysproto.h>
61#include <sys/systm.h>
62#include <sys/time.h>
63#include <sys/vmmeter.h>
64#include <sys/vnode.h>
65#include <sys/wait.h>
66#include <sys/cpuset.h>
67
68#include <security/mac/mac_framework.h>
69
70#include <vm/vm.h>
71#include <vm/pmap.h>
72#include <vm/vm_kern.h>
73#include <vm/vm_map.h>
74#include <vm/vm_extern.h>
75#include <vm/vm_object.h>
76#include <vm/swap_pager.h>
77
78#ifdef COMPAT_LINUX32
79#include <machine/../linux32/linux.h>
80#include <machine/../linux32/linux32_proto.h>
81#else
82#include <machine/../linux/linux.h>
83#include <machine/../linux/linux_proto.h>
84#endif
85
86#include <compat/linux/linux_file.h>
87#include <compat/linux/linux_mib.h>
88#include <compat/linux/linux_signal.h>
89#include <compat/linux/linux_util.h>
90#include <compat/linux/linux_sysproto.h>
91#include <compat/linux/linux_emul.h>
92#include <compat/linux/linux_misc.h>
93
94int stclohz; /* Statistics clock frequency */
95
96static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
97 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
98 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
99 RLIMIT_MEMLOCK, RLIMIT_AS
100};
101
102struct l_sysinfo {
103 l_long uptime; /* Seconds since boot */
104 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
105#define LINUX_SYSINFO_LOADS_SCALE 65536
106 l_ulong totalram; /* Total usable main memory size */
107 l_ulong freeram; /* Available memory size */
108 l_ulong sharedram; /* Amount of shared memory */
109 l_ulong bufferram; /* Memory used by buffers */
110 l_ulong totalswap; /* Total swap space size */
111 l_ulong freeswap; /* swap space still available */
112 l_ushort procs; /* Number of current processes */
113 l_ushort pads;
114 l_ulong totalbig;
115 l_ulong freebig;
116 l_uint mem_unit;
117 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
118};
119
120struct l_pselect6arg {
121 l_uintptr_t ss;
122 l_size_t ss_len;
123};
124
125int
126linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
127{
128 struct l_sysinfo sysinfo;
129 vm_object_t object;
130 int i, j;
131 struct timespec ts;
132
133 getnanouptime(&ts);
134 if (ts.tv_nsec != 0)
135 ts.tv_sec++;
136 sysinfo.uptime = ts.tv_sec;
137
138 /* Use the information from the mib to get our load averages */
139 for (i = 0; i < 3; i++)
140 sysinfo.loads[i] = averunnable.ldavg[i] *
141 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
142
143 sysinfo.totalram = physmem * PAGE_SIZE;
144 sysinfo.freeram = sysinfo.totalram - vm_cnt.v_wire_count * PAGE_SIZE;
145
146 sysinfo.sharedram = 0;
147 mtx_lock(&vm_object_list_mtx);
148 TAILQ_FOREACH(object, &vm_object_list, object_list)
149 if (object->shadow_count > 1)
150 sysinfo.sharedram += object->resident_page_count;
151 mtx_unlock(&vm_object_list_mtx);
152
153 sysinfo.sharedram *= PAGE_SIZE;
154 sysinfo.bufferram = 0;
155
156 swap_pager_status(&i, &j);
157 sysinfo.totalswap = i * PAGE_SIZE;
158 sysinfo.freeswap = (i - j) * PAGE_SIZE;
159
160 sysinfo.procs = nprocs;
161
162 /* The following are only present in newer Linux kernels. */
163 sysinfo.totalbig = 0;
164 sysinfo.freebig = 0;
165 sysinfo.mem_unit = 1;
166
167 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
168}
169
170int
171linux_alarm(struct thread *td, struct linux_alarm_args *args)
172{
173 struct itimerval it, old_it;
174 u_int secs;
175 int error;
176
177#ifdef DEBUG
178 if (ldebug(alarm))
179 printf(ARGS(alarm, "%u"), args->secs);
180#endif
181
182 secs = args->secs;
183
184 if (secs > INT_MAX)
185 secs = INT_MAX;
186
187 it.it_value.tv_sec = (long) secs;
188 it.it_value.tv_usec = 0;
189 it.it_interval.tv_sec = 0;
190 it.it_interval.tv_usec = 0;
191 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
192 if (error)
193 return (error);
194 if (timevalisset(&old_it.it_value)) {
195 if (old_it.it_value.tv_usec != 0)
196 old_it.it_value.tv_sec++;
197 td->td_retval[0] = old_it.it_value.tv_sec;
198 }
199 return (0);
200}
201
202int
203linux_brk(struct thread *td, struct linux_brk_args *args)
204{
205 struct vmspace *vm = td->td_proc->p_vmspace;
206 vm_offset_t new, old;
207 struct obreak_args /* {
208 char * nsize;
209 } */ tmp;
210
211#ifdef DEBUG
212 if (ldebug(brk))
213 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
214#endif
215 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
216 new = (vm_offset_t)args->dsend;
217 tmp.nsize = (char *)new;
218 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
219 td->td_retval[0] = (long)new;
220 else
221 td->td_retval[0] = (long)old;
222
223 return (0);
224}
225
226#if defined(__i386__)
227/* XXX: what about amd64/linux32? */
228
229int
230linux_uselib(struct thread *td, struct linux_uselib_args *args)
231{
232 struct nameidata ni;
233 struct vnode *vp;
234 struct exec *a_out;
235 struct vattr attr;
236 vm_offset_t vmaddr;
237 unsigned long file_offset;
238 unsigned long bss_size;
239 char *library;
240 ssize_t aresid;
241 int error, locked, writecount;
242
243 LCONVPATHEXIST(td, args->library, &library);
244
245#ifdef DEBUG
246 if (ldebug(uselib))
247 printf(ARGS(uselib, "%s"), library);
248#endif
249
250 a_out = NULL;
251 locked = 0;
252 vp = NULL;
253
254 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
255 UIO_SYSSPACE, library, td);
256 error = namei(&ni);
257 LFREEPATH(library);
258 if (error)
259 goto cleanup;
260
261 vp = ni.ni_vp;
262 NDFREE(&ni, NDF_ONLY_PNBUF);
263
264 /*
265 * From here on down, we have a locked vnode that must be unlocked.
266 * XXX: The code below largely duplicates exec_check_permissions().
267 */
268 locked = 1;
269
270 /* Writable? */
271 error = VOP_GET_WRITECOUNT(vp, &writecount);
272 if (error != 0)
273 goto cleanup;
274 if (writecount != 0) {
275 error = ETXTBSY;
276 goto cleanup;
277 }
278
279 /* Executable? */
280 error = VOP_GETATTR(vp, &attr, td->td_ucred);
281 if (error)
282 goto cleanup;
283
284 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
285 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
286 /* EACCESS is what exec(2) returns. */
287 error = ENOEXEC;
288 goto cleanup;
289 }
290
291 /* Sensible size? */
292 if (attr.va_size == 0) {
293 error = ENOEXEC;
294 goto cleanup;
295 }
296
297 /* Can we access it? */
298 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
299 if (error)
300 goto cleanup;
301
302 /*
303 * XXX: This should use vn_open() so that it is properly authorized,
304 * and to reduce code redundancy all over the place here.
305 * XXX: Not really, it duplicates far more of exec_check_permissions()
306 * than vn_open().
307 */
308#ifdef MAC
309 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
310 if (error)
311 goto cleanup;
312#endif
313 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
314 if (error)
315 goto cleanup;
316
317 /* Pull in executable header into exec_map */
318 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
319 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
320 if (error)
321 goto cleanup;
322
323 /* Is it a Linux binary ? */
324 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
325 error = ENOEXEC;
326 goto cleanup;
327 }
328
329 /*
330 * While we are here, we should REALLY do some more checks
331 */
332
333 /* Set file/virtual offset based on a.out variant. */
334 switch ((int)(a_out->a_magic & 0xffff)) {
335 case 0413: /* ZMAGIC */
336 file_offset = 1024;
337 break;
338 case 0314: /* QMAGIC */
339 file_offset = 0;
340 break;
341 default:
342 error = ENOEXEC;
343 goto cleanup;
344 }
345
346 bss_size = round_page(a_out->a_bss);
347
348 /* Check various fields in header for validity/bounds. */
349 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
350 error = ENOEXEC;
351 goto cleanup;
352 }
353
354 /* text + data can't exceed file size */
355 if (a_out->a_data + a_out->a_text > attr.va_size) {
356 error = EFAULT;
357 goto cleanup;
358 }
359
360 /*
361 * text/data/bss must not exceed limits
362 * XXX - this is not complete. it should check current usage PLUS
363 * the resources needed by this library.
364 */
365 PROC_LOCK(td->td_proc);
366 if (a_out->a_text > maxtsiz ||
367 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
368 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
369 bss_size) != 0) {
370 PROC_UNLOCK(td->td_proc);
371 error = ENOMEM;
372 goto cleanup;
373 }
374 PROC_UNLOCK(td->td_proc);
375
376 /*
377 * Prevent more writers.
378 * XXX: Note that if any of the VM operations fail below we don't
379 * clear this flag.
380 */
381 VOP_SET_TEXT(vp);
382
383 /*
384 * Lock no longer needed
385 */
386 locked = 0;
387 VOP_UNLOCK(vp, 0);
388
389 /*
390 * Check if file_offset page aligned. Currently we cannot handle
391 * misalinged file offsets, and so we read in the entire image
392 * (what a waste).
393 */
394 if (file_offset & PAGE_MASK) {
395#ifdef DEBUG
396 printf("uselib: Non page aligned binary %lu\n", file_offset);
397#endif
398 /* Map text+data read/write/execute */
399
400 /* a_entry is the load address and is page aligned */
401 vmaddr = trunc_page(a_out->a_entry);
402
403 /* get anon user mapping, read+write+execute */
404 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
405 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
406 VM_PROT_ALL, VM_PROT_ALL, 0);
407 if (error)
408 goto cleanup;
409
410 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
411 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
412 td->td_ucred, NOCRED, &aresid, td);
413 if (error != 0)
414 goto cleanup;
415 if (aresid != 0) {
416 error = ENOEXEC;
417 goto cleanup;
418 }
419 } else {
420#ifdef DEBUG
421 printf("uselib: Page aligned binary %lu\n", file_offset);
422#endif
423 /*
424 * for QMAGIC, a_entry is 20 bytes beyond the load address
425 * to skip the executable header
426 */
427 vmaddr = trunc_page(a_out->a_entry);
428
429 /*
430 * Map it all into the process's space as a single
431 * copy-on-write "data" segment.
432 */
433 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
434 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
435 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
436 if (error)
437 goto cleanup;
438 }
439#ifdef DEBUG
440 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
441 ((long *)vmaddr)[1]);
442#endif
443 if (bss_size != 0) {
444 /* Calculate BSS start address */
445 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
446 a_out->a_data;
447
448 /* allocate some 'anon' space */
449 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
450 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
451 VM_PROT_ALL, 0);
452 if (error)
453 goto cleanup;
454 }
455
456cleanup:
457 /* Unlock vnode if needed */
458 if (locked)
459 VOP_UNLOCK(vp, 0);
460
461 /* Release the temporary mapping. */
462 if (a_out)
463 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
464
465 return (error);
466}
467
468#endif /* __i386__ */
469
470int
471linux_select(struct thread *td, struct linux_select_args *args)
472{
473 l_timeval ltv;
474 struct timeval tv0, tv1, utv, *tvp;
475 int error;
476
477#ifdef DEBUG
478 if (ldebug(select))
479 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
480 (void *)args->readfds, (void *)args->writefds,
481 (void *)args->exceptfds, (void *)args->timeout);
482#endif
483
484 /*
485 * Store current time for computation of the amount of
486 * time left.
487 */
488 if (args->timeout) {
489 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
490 goto select_out;
491 utv.tv_sec = ltv.tv_sec;
492 utv.tv_usec = ltv.tv_usec;
493#ifdef DEBUG
494 if (ldebug(select))
495 printf(LMSG("incoming timeout (%jd/%ld)"),
496 (intmax_t)utv.tv_sec, utv.tv_usec);
497#endif
498
499 if (itimerfix(&utv)) {
500 /*
501 * The timeval was invalid. Convert it to something
502 * valid that will act as it does under Linux.
503 */
504 utv.tv_sec += utv.tv_usec / 1000000;
505 utv.tv_usec %= 1000000;
506 if (utv.tv_usec < 0) {
507 utv.tv_sec -= 1;
508 utv.tv_usec += 1000000;
509 }
510 if (utv.tv_sec < 0)
511 timevalclear(&utv);
512 }
513 microtime(&tv0);
514 tvp = &utv;
515 } else
516 tvp = NULL;
517
518 error = kern_select(td, args->nfds, args->readfds, args->writefds,
519 args->exceptfds, tvp, sizeof(l_int) * 8);
520
521#ifdef DEBUG
522 if (ldebug(select))
523 printf(LMSG("real select returns %d"), error);
524#endif
525 if (error)
526 goto select_out;
527
528 if (args->timeout) {
529 if (td->td_retval[0]) {
530 /*
531 * Compute how much time was left of the timeout,
532 * by subtracting the current time and the time
533 * before we started the call, and subtracting
534 * that result from the user-supplied value.
535 */
536 microtime(&tv1);
537 timevalsub(&tv1, &tv0);
538 timevalsub(&utv, &tv1);
539 if (utv.tv_sec < 0)
540 timevalclear(&utv);
541 } else
542 timevalclear(&utv);
543#ifdef DEBUG
544 if (ldebug(select))
545 printf(LMSG("outgoing timeout (%jd/%ld)"),
546 (intmax_t)utv.tv_sec, utv.tv_usec);
547#endif
548 ltv.tv_sec = utv.tv_sec;
549 ltv.tv_usec = utv.tv_usec;
550 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
551 goto select_out;
552 }
553
554select_out:
555#ifdef DEBUG
556 if (ldebug(select))
557 printf(LMSG("select_out -> %d"), error);
558#endif
559 return (error);
560}
561
562int
563linux_mremap(struct thread *td, struct linux_mremap_args *args)
564{
565 struct munmap_args /* {
566 void *addr;
567 size_t len;
568 } */ bsd_args;
569 int error = 0;
570
571#ifdef DEBUG
572 if (ldebug(mremap))
573 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
574 (void *)(uintptr_t)args->addr,
575 (unsigned long)args->old_len,
576 (unsigned long)args->new_len,
577 (unsigned long)args->flags);
578#endif
579
580 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
581 td->td_retval[0] = 0;
582 return (EINVAL);
583 }
584
585 /*
586 * Check for the page alignment.
587 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
588 */
589 if (args->addr & PAGE_MASK) {
590 td->td_retval[0] = 0;
591 return (EINVAL);
592 }
593
594 args->new_len = round_page(args->new_len);
595 args->old_len = round_page(args->old_len);
596
597 if (args->new_len > args->old_len) {
598 td->td_retval[0] = 0;
599 return (ENOMEM);
600 }
601
602 if (args->new_len < args->old_len) {
603 bsd_args.addr =
604 (caddr_t)((uintptr_t)args->addr + args->new_len);
605 bsd_args.len = args->old_len - args->new_len;
606 error = sys_munmap(td, &bsd_args);
607 }
608
609 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
610 return (error);
611}
612
613#define LINUX_MS_ASYNC 0x0001
614#define LINUX_MS_INVALIDATE 0x0002
615#define LINUX_MS_SYNC 0x0004
616
617int
618linux_msync(struct thread *td, struct linux_msync_args *args)
619{
620 struct msync_args bsd_args;
621
622 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
623 bsd_args.len = (uintptr_t)args->len;
624 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
625
626 return (sys_msync(td, &bsd_args));
627}
628
629int
630linux_time(struct thread *td, struct linux_time_args *args)
631{
632 struct timeval tv;
633 l_time_t tm;
634 int error;
635
636#ifdef DEBUG
637 if (ldebug(time))
638 printf(ARGS(time, "*"));
639#endif
640
641 microtime(&tv);
642 tm = tv.tv_sec;
643 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
644 return (error);
645 td->td_retval[0] = tm;
646 return (0);
647}
648
649struct l_times_argv {
650 l_clock_t tms_utime;
651 l_clock_t tms_stime;
652 l_clock_t tms_cutime;
653 l_clock_t tms_cstime;
654};
655
656
657/*
658 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
659 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
660 * auxiliary vector entry.
661 */
662#define CLK_TCK 100
663
664#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
665#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
666
667#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
668 CONVNTCK(r) : CONVOTCK(r))
669
670int
671linux_times(struct thread *td, struct linux_times_args *args)
672{
673 struct timeval tv, utime, stime, cutime, cstime;
674 struct l_times_argv tms;
675 struct proc *p;
676 int error;
677
678#ifdef DEBUG
679 if (ldebug(times))
680 printf(ARGS(times, "*"));
681#endif
682
683 if (args->buf != NULL) {
684 p = td->td_proc;
685 PROC_LOCK(p);
686 PROC_STATLOCK(p);
687 calcru(p, &utime, &stime);
688 PROC_STATUNLOCK(p);
689 calccru(p, &cutime, &cstime);
690 PROC_UNLOCK(p);
691
692 tms.tms_utime = CONVTCK(utime);
693 tms.tms_stime = CONVTCK(stime);
694
695 tms.tms_cutime = CONVTCK(cutime);
696 tms.tms_cstime = CONVTCK(cstime);
697
698 if ((error = copyout(&tms, args->buf, sizeof(tms))))
699 return (error);
700 }
701
702 microuptime(&tv);
703 td->td_retval[0] = (int)CONVTCK(tv);
704 return (0);
705}
706
707int
708linux_newuname(struct thread *td, struct linux_newuname_args *args)
709{
710 struct l_new_utsname utsname;
711 char osname[LINUX_MAX_UTSNAME];
712 char osrelease[LINUX_MAX_UTSNAME];
713 char *p;
714
715#ifdef DEBUG
716 if (ldebug(newuname))
717 printf(ARGS(newuname, "*"));
718#endif
719
720 linux_get_osname(td, osname);
721 linux_get_osrelease(td, osrelease);
722
723 bzero(&utsname, sizeof(utsname));
724 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
725 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
726 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
727 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
728 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
729 for (p = utsname.version; *p != '\0'; ++p)
730 if (*p == '\n') {
731 *p = '\0';
732 break;
733 }
734 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
735
736 return (copyout(&utsname, args->buf, sizeof(utsname)));
737}
738
739struct l_utimbuf {
740 l_time_t l_actime;
741 l_time_t l_modtime;
742};
743
744int
745linux_utime(struct thread *td, struct linux_utime_args *args)
746{
747 struct timeval tv[2], *tvp;
748 struct l_utimbuf lut;
749 char *fname;
750 int error;
751
752 LCONVPATHEXIST(td, args->fname, &fname);
753
754#ifdef DEBUG
755 if (ldebug(utime))
756 printf(ARGS(utime, "%s, *"), fname);
757#endif
758
759 if (args->times) {
760 if ((error = copyin(args->times, &lut, sizeof lut))) {
761 LFREEPATH(fname);
762 return (error);
763 }
764 tv[0].tv_sec = lut.l_actime;
765 tv[0].tv_usec = 0;
766 tv[1].tv_sec = lut.l_modtime;
767 tv[1].tv_usec = 0;
768 tvp = tv;
769 } else
770 tvp = NULL;
771
772 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE, tvp,
773 UIO_SYSSPACE);
774 LFREEPATH(fname);
775 return (error);
776}
777
778int
779linux_utimes(struct thread *td, struct linux_utimes_args *args)
780{
781 l_timeval ltv[2];
782 struct timeval tv[2], *tvp = NULL;
783 char *fname;
784 int error;
785
786 LCONVPATHEXIST(td, args->fname, &fname);
787
788#ifdef DEBUG
789 if (ldebug(utimes))
790 printf(ARGS(utimes, "%s, *"), fname);
791#endif
792
793 if (args->tptr != NULL) {
794 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
795 LFREEPATH(fname);
796 return (error);
797 }
798 tv[0].tv_sec = ltv[0].tv_sec;
799 tv[0].tv_usec = ltv[0].tv_usec;
800 tv[1].tv_sec = ltv[1].tv_sec;
801 tv[1].tv_usec = ltv[1].tv_usec;
802 tvp = tv;
803 }
804
805 error = kern_utimesat(td, AT_FDCWD, fname, UIO_SYSSPACE,
806 tvp, UIO_SYSSPACE);
807 LFREEPATH(fname);
808 return (error);
809}
810
811int
812linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
813{
814 l_timeval ltv[2];
815 struct timeval tv[2], *tvp = NULL;
816 char *fname;
817 int error, dfd;
818
819 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
820 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
821
822#ifdef DEBUG
823 if (ldebug(futimesat))
824 printf(ARGS(futimesat, "%s, *"), fname);
825#endif
826
827 if (args->utimes != NULL) {
828 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
829 LFREEPATH(fname);
830 return (error);
831 }
832 tv[0].tv_sec = ltv[0].tv_sec;
833 tv[0].tv_usec = ltv[0].tv_usec;
834 tv[1].tv_sec = ltv[1].tv_sec;
835 tv[1].tv_usec = ltv[1].tv_usec;
836 tvp = tv;
837 }
838
839 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
840 LFREEPATH(fname);
841 return (error);
842}
843
844int
845linux_common_wait(struct thread *td, int pid, int *status,
846 int options, struct rusage *ru)
847{
848 int error, tmpstat;
849
850 error = kern_wait(td, pid, &tmpstat, options, ru);
851 if (error)
852 return (error);
853
854 if (status) {
855 tmpstat &= 0xffff;
856 if (WIFSIGNALED(tmpstat))
857 tmpstat = (tmpstat & 0xffffff80) |
858 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
859 else if (WIFSTOPPED(tmpstat))
860 tmpstat = (tmpstat & 0xffff00ff) |
861 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
862 error = copyout(&tmpstat, status, sizeof(int));
863 }
864
865 return (error);
866}
867
868#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
869int
870linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
871{
872 int options;
873
874#ifdef DEBUG
875 if (ldebug(waitpid))
876 printf(ARGS(waitpid, "%d, %p, %d"),
877 args->pid, (void *)args->status, args->options);
878#endif
879 /*
880 * this is necessary because the test in kern_wait doesn't work
881 * because we mess with the options here
882 */
883 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
884 return (EINVAL);
885
886 options = (args->options & (WNOHANG | WUNTRACED));
887 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
888 if (args->options & __WCLONE)
889 options |= WLINUXCLONE;
890
891 return (linux_common_wait(td, args->pid, args->status, options, NULL));
892}
893#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
894
895int
896linux_wait4(struct thread *td, struct linux_wait4_args *args)
897{
898 int error, options;
899 struct rusage ru, *rup;
900
901#ifdef DEBUG
902 if (ldebug(wait4))
903 printf(ARGS(wait4, "%d, %p, %d, %p"),
904 args->pid, (void *)args->status, args->options,
905 (void *)args->rusage);
906#endif
907
908 options = (args->options & (WNOHANG | WUNTRACED));
909 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
910 if (args->options & __WCLONE)
911 options |= WLINUXCLONE;
912
913 if (args->rusage != NULL)
914 rup = &ru;
915 else
916 rup = NULL;
917 error = linux_common_wait(td, args->pid, args->status, options, rup);
918 if (error != 0)
919 return (error);
920 if (args->rusage != NULL)
921 error = linux_copyout_rusage(&ru, args->rusage);
922 return (error);
923}
924
925int
926linux_waitid(struct thread *td, struct linux_waitid_args *args)
927{
928 int status, options, sig;
929 struct __wrusage wru;
930 siginfo_t siginfo;
931 l_siginfo_t lsi;
932 idtype_t idtype;
933 struct proc *p;
934 int error;
935
936 options = 0;
937 linux_to_bsd_waitopts(args->options, &options);
938
939 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
940 return (EINVAL);
941 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
942 return (EINVAL);
943
944 switch (args->idtype) {
945 case LINUX_P_ALL:
946 idtype = P_ALL;
947 break;
948 case LINUX_P_PID:
949 if (args->id <= 0)
950 return (EINVAL);
951 idtype = P_PID;
952 break;
953 case LINUX_P_PGID:
954 if (args->id <= 0)
955 return (EINVAL);
956 idtype = P_PGID;
957 break;
958 default:
959 return (EINVAL);
960 }
961
962 error = kern_wait6(td, idtype, args->id, &status, options,
963 &wru, &siginfo);
964 if (error != 0)
965 return (error);
966 if (args->rusage != NULL) {
967 error = linux_copyout_rusage(&wru.wru_children,
968 args->rusage);
969 if (error != 0)
970 return (error);
971 }
972 if (args->info != NULL) {
973 p = td->td_proc;
974 if (td->td_retval[0] == 0)
975 bzero(&lsi, sizeof(lsi));
976 else {
977 sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo);
978 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
979 }
980 error = copyout(&lsi, args->info, sizeof(lsi));
981 }
982 td->td_retval[0] = 0;
983
984 return (error);
985}
986
987int
988linux_mknod(struct thread *td, struct linux_mknod_args *args)
989{
990 char *path;
991 int error;
992
993 LCONVPATHCREAT(td, args->path, &path);
994
995#ifdef DEBUG
996 if (ldebug(mknod))
997 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
998 (uintmax_t)args->dev);
999#endif
1000
1001 switch (args->mode & S_IFMT) {
1002 case S_IFIFO:
1003 case S_IFSOCK:
1004 error = kern_mkfifoat(td, AT_FDCWD, path, UIO_SYSSPACE,
1005 args->mode);
1006 break;
1007
1008 case S_IFCHR:
1009 case S_IFBLK:
1010 error = kern_mknodat(td, AT_FDCWD, path, UIO_SYSSPACE,
1011 args->mode, args->dev);
1012 break;
1013
1014 case S_IFDIR:
1015 error = EPERM;
1016 break;
1017
1018 case 0:
1019 args->mode |= S_IFREG;
1020 /* FALLTHROUGH */
1021 case S_IFREG:
1022 error = kern_openat(td, AT_FDCWD, path, UIO_SYSSPACE,
1023 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1024 if (error == 0)
1025 kern_close(td, td->td_retval[0]);
1026 break;
1027
1028 default:
1029 error = EINVAL;
1030 break;
1031 }
1032 LFREEPATH(path);
1033 return (error);
1034}
1035
1036int
1037linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1038{
1039 char *path;
1040 int error, dfd;
1041
1042 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1043 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1044
1045#ifdef DEBUG
1046 if (ldebug(mknodat))
1047 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1048#endif
1049
1050 switch (args->mode & S_IFMT) {
1051 case S_IFIFO:
1052 case S_IFSOCK:
1053 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1054 break;
1055
1056 case S_IFCHR:
1057 case S_IFBLK:
1058 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1059 args->dev);
1060 break;
1061
1062 case S_IFDIR:
1063 error = EPERM;
1064 break;
1065
1066 case 0:
1067 args->mode |= S_IFREG;
1068 /* FALLTHROUGH */
1069 case S_IFREG:
1070 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1071 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1072 if (error == 0)
1073 kern_close(td, td->td_retval[0]);
1074 break;
1075
1076 default:
1077 error = EINVAL;
1078 break;
1079 }
1080 LFREEPATH(path);
1081 return (error);
1082}
1083
1084/*
1085 * UGH! This is just about the dumbest idea I've ever heard!!
1086 */
1087int
1088linux_personality(struct thread *td, struct linux_personality_args *args)
1089{
1090#ifdef DEBUG
1091 if (ldebug(personality))
1092 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1093#endif
1094 if (args->per != 0)
1095 return (EINVAL);
1096
1097 /* Yes Jim, it's still a Linux... */
1098 td->td_retval[0] = 0;
1099 return (0);
1100}
1101
1102struct l_itimerval {
1103 l_timeval it_interval;
1104 l_timeval it_value;
1105};
1106
1107#define B2L_ITIMERVAL(bip, lip) \
1108 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1109 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1110 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1111 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1112
1113int
1114linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1115{
1116 int error;
1117 struct l_itimerval ls;
1118 struct itimerval aitv, oitv;
1119
1120#ifdef DEBUG
1121 if (ldebug(setitimer))
1122 printf(ARGS(setitimer, "%p, %p"),
1123 (void *)uap->itv, (void *)uap->oitv);
1124#endif
1125
1126 if (uap->itv == NULL) {
1127 uap->itv = uap->oitv;
1128 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1129 }
1130
1131 error = copyin(uap->itv, &ls, sizeof(ls));
1132 if (error != 0)
1133 return (error);
1134 B2L_ITIMERVAL(&aitv, &ls);
1135#ifdef DEBUG
1136 if (ldebug(setitimer)) {
1137 printf("setitimer: value: sec: %jd, usec: %ld\n",
1138 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1139 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1140 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1141 }
1142#endif
1143 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1144 if (error != 0 || uap->oitv == NULL)
1145 return (error);
1146 B2L_ITIMERVAL(&ls, &oitv);
1147
1148 return (copyout(&ls, uap->oitv, sizeof(ls)));
1149}
1150
1151int
1152linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1153{
1154 int error;
1155 struct l_itimerval ls;
1156 struct itimerval aitv;
1157
1158#ifdef DEBUG
1159 if (ldebug(getitimer))
1160 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1161#endif
1162 error = kern_getitimer(td, uap->which, &aitv);
1163 if (error != 0)
1164 return (error);
1165 B2L_ITIMERVAL(&ls, &aitv);
1166 return (copyout(&ls, uap->itv, sizeof(ls)));
1167}
1168
1169#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1170int
1171linux_nice(struct thread *td, struct linux_nice_args *args)
1172{
1173 struct setpriority_args bsd_args;
1174
1175 bsd_args.which = PRIO_PROCESS;
1176 bsd_args.who = 0; /* current process */
1177 bsd_args.prio = args->inc;
1178 return (sys_setpriority(td, &bsd_args));
1179}
1180#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1181
1182int
1183linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1184{
1185 struct ucred *newcred, *oldcred;
1186 l_gid_t *linux_gidset;
1187 gid_t *bsd_gidset;
1188 int ngrp, error;
1189 struct proc *p;
1190
1191 ngrp = args->gidsetsize;
1192 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1193 return (EINVAL);
1194 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK);
1195 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1196 if (error)
1197 goto out;
1198 newcred = crget();
1199 p = td->td_proc;
1200 PROC_LOCK(p);
1201 oldcred = crcopysafe(p, newcred);
1202
1203 /*
1204 * cr_groups[0] holds egid. Setting the whole set from
1205 * the supplied set will cause egid to be changed too.
1206 * Keep cr_groups[0] unchanged to prevent that.
1207 */
1208
1209 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1210 PROC_UNLOCK(p);
1211 crfree(newcred);
1212 goto out;
1213 }
1214
1215 if (ngrp > 0) {
1216 newcred->cr_ngroups = ngrp + 1;
1217
1218 bsd_gidset = newcred->cr_groups;
1219 ngrp--;
1220 while (ngrp >= 0) {
1221 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1222 ngrp--;
1223 }
1224 } else
1225 newcred->cr_ngroups = 1;
1226
1227 setsugid(p);
1228 proc_set_cred(p, newcred);
1229 PROC_UNLOCK(p);
1230 crfree(oldcred);
1231 error = 0;
1232out:
1233 free(linux_gidset, M_TEMP);
1234 return (error);
1235}
1236
1237int
1238linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1239{
1240 struct ucred *cred;
1241 l_gid_t *linux_gidset;
1242 gid_t *bsd_gidset;
1243 int bsd_gidsetsz, ngrp, error;
1244
1245 cred = td->td_ucred;
1246 bsd_gidset = cred->cr_groups;
1247 bsd_gidsetsz = cred->cr_ngroups - 1;
1248
1249 /*
1250 * cr_groups[0] holds egid. Returning the whole set
1251 * here will cause a duplicate. Exclude cr_groups[0]
1252 * to prevent that.
1253 */
1254
1255 if ((ngrp = args->gidsetsize) == 0) {
1256 td->td_retval[0] = bsd_gidsetsz;
1257 return (0);
1258 }
1259
1260 if (ngrp < bsd_gidsetsz)
1261 return (EINVAL);
1262
1263 ngrp = 0;
1264 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1265 M_TEMP, M_WAITOK);
1266 while (ngrp < bsd_gidsetsz) {
1267 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1268 ngrp++;
1269 }
1270
1271 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1272 free(linux_gidset, M_TEMP);
1273 if (error)
1274 return (error);
1275
1276 td->td_retval[0] = ngrp;
1277 return (0);
1278}
1279
1280int
1281linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1282{
1283 struct rlimit bsd_rlim;
1284 struct l_rlimit rlim;
1285 u_int which;
1286 int error;
1287
1288#ifdef DEBUG
1289 if (ldebug(setrlimit))
1290 printf(ARGS(setrlimit, "%d, %p"),
1291 args->resource, (void *)args->rlim);
1292#endif
1293
1294 if (args->resource >= LINUX_RLIM_NLIMITS)
1295 return (EINVAL);
1296
1297 which = linux_to_bsd_resource[args->resource];
1298 if (which == -1)
1299 return (EINVAL);
1300
1301 error = copyin(args->rlim, &rlim, sizeof(rlim));
1302 if (error)
1303 return (error);
1304
1305 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1306 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1307 return (kern_setrlimit(td, which, &bsd_rlim));
1308}
1309
1310#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1311int
1312linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1313{
1314 struct l_rlimit rlim;
1315 struct proc *p = td->td_proc;
1316 struct rlimit bsd_rlim;
1317 u_int which;
1318
1319#ifdef DEBUG
1320 if (ldebug(old_getrlimit))
1321 printf(ARGS(old_getrlimit, "%d, %p"),
1322 args->resource, (void *)args->rlim);
1323#endif
1324
1325 if (args->resource >= LINUX_RLIM_NLIMITS)
1326 return (EINVAL);
1327
1328 which = linux_to_bsd_resource[args->resource];
1329 if (which == -1)
1330 return (EINVAL);
1331
1332 PROC_LOCK(p);
1333 lim_rlimit(p, which, &bsd_rlim);
1334 PROC_UNLOCK(p);
1335
1336#ifdef COMPAT_LINUX32
1337 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1338 if (rlim.rlim_cur == UINT_MAX)
1339 rlim.rlim_cur = INT_MAX;
1340 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1341 if (rlim.rlim_max == UINT_MAX)
1342 rlim.rlim_max = INT_MAX;
1343#else
1344 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1345 if (rlim.rlim_cur == ULONG_MAX)
1346 rlim.rlim_cur = LONG_MAX;
1347 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1348 if (rlim.rlim_max == ULONG_MAX)
1349 rlim.rlim_max = LONG_MAX;
1350#endif
1351 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1352}
1353#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1354
1355int
1356linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1357{
1358 struct l_rlimit rlim;
1359 struct proc *p = td->td_proc;
1360 struct rlimit bsd_rlim;
1361 u_int which;
1362
1363#ifdef DEBUG
1364 if (ldebug(getrlimit))
1365 printf(ARGS(getrlimit, "%d, %p"),
1366 args->resource, (void *)args->rlim);
1367#endif
1368
1369 if (args->resource >= LINUX_RLIM_NLIMITS)
1370 return (EINVAL);
1371
1372 which = linux_to_bsd_resource[args->resource];
1373 if (which == -1)
1374 return (EINVAL);
1375
1376 PROC_LOCK(p);
1377 lim_rlimit(p, which, &bsd_rlim);
1378 PROC_UNLOCK(p);
1379
1380 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1381 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1382 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1383}
1384
1385int
1386linux_sched_setscheduler(struct thread *td,
1387 struct linux_sched_setscheduler_args *args)
1388{
1389 struct sched_param sched_param;
1390 struct thread *tdt;
1391 int error, policy;
1392
1393#ifdef DEBUG
1394 if (ldebug(sched_setscheduler))
1395 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1396 args->pid, args->policy, (const void *)args->param);
1397#endif
1398
1399 switch (args->policy) {
1400 case LINUX_SCHED_OTHER:
1401 policy = SCHED_OTHER;
1402 break;
1403 case LINUX_SCHED_FIFO:
1404 policy = SCHED_FIFO;
1405 break;
1406 case LINUX_SCHED_RR:
1407 policy = SCHED_RR;
1408 break;
1409 default:
1410 return (EINVAL);
1411 }
1412
1413 error = copyin(args->param, &sched_param, sizeof(sched_param));
1414 if (error)
1415 return (error);
1416
1417 tdt = linux_tdfind(td, args->pid, -1);
1418 if (tdt == NULL)
1419 return (ESRCH);
1420
1421 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1422 PROC_UNLOCK(tdt->td_proc);
1423 return (error);
1424}
1425
1426int
1427linux_sched_getscheduler(struct thread *td,
1428 struct linux_sched_getscheduler_args *args)
1429{
1430 struct thread *tdt;
1431 int error, policy;
1432
1433#ifdef DEBUG
1434 if (ldebug(sched_getscheduler))
1435 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1436#endif
1437
1438 tdt = linux_tdfind(td, args->pid, -1);
1439 if (tdt == NULL)
1440 return (ESRCH);
1441
1442 error = kern_sched_getscheduler(td, tdt, &policy);
1443 PROC_UNLOCK(tdt->td_proc);
1444
1445 switch (policy) {
1446 case SCHED_OTHER:
1447 td->td_retval[0] = LINUX_SCHED_OTHER;
1448 break;
1449 case SCHED_FIFO:
1450 td->td_retval[0] = LINUX_SCHED_FIFO;
1451 break;
1452 case SCHED_RR:
1453 td->td_retval[0] = LINUX_SCHED_RR;
1454 break;
1455 }
1456 return (error);
1457}
1458
1459int
1460linux_sched_get_priority_max(struct thread *td,
1461 struct linux_sched_get_priority_max_args *args)
1462{
1463 struct sched_get_priority_max_args bsd;
1464
1465#ifdef DEBUG
1466 if (ldebug(sched_get_priority_max))
1467 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1468#endif
1469
1470 switch (args->policy) {
1471 case LINUX_SCHED_OTHER:
1472 bsd.policy = SCHED_OTHER;
1473 break;
1474 case LINUX_SCHED_FIFO:
1475 bsd.policy = SCHED_FIFO;
1476 break;
1477 case LINUX_SCHED_RR:
1478 bsd.policy = SCHED_RR;
1479 break;
1480 default:
1481 return (EINVAL);
1482 }
1483 return (sys_sched_get_priority_max(td, &bsd));
1484}
1485
1486int
1487linux_sched_get_priority_min(struct thread *td,
1488 struct linux_sched_get_priority_min_args *args)
1489{
1490 struct sched_get_priority_min_args bsd;
1491
1492#ifdef DEBUG
1493 if (ldebug(sched_get_priority_min))
1494 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1495#endif
1496
1497 switch (args->policy) {
1498 case LINUX_SCHED_OTHER:
1499 bsd.policy = SCHED_OTHER;
1500 break;
1501 case LINUX_SCHED_FIFO:
1502 bsd.policy = SCHED_FIFO;
1503 break;
1504 case LINUX_SCHED_RR:
1505 bsd.policy = SCHED_RR;
1506 break;
1507 default:
1508 return (EINVAL);
1509 }
1510 return (sys_sched_get_priority_min(td, &bsd));
1511}
1512
1513#define REBOOT_CAD_ON 0x89abcdef
1514#define REBOOT_CAD_OFF 0
1515#define REBOOT_HALT 0xcdef0123
1516#define REBOOT_RESTART 0x01234567
1517#define REBOOT_RESTART2 0xA1B2C3D4
1518#define REBOOT_POWEROFF 0x4321FEDC
1519#define REBOOT_MAGIC1 0xfee1dead
1520#define REBOOT_MAGIC2 0x28121969
1521#define REBOOT_MAGIC2A 0x05121996
1522#define REBOOT_MAGIC2B 0x16041998
1523
1524int
1525linux_reboot(struct thread *td, struct linux_reboot_args *args)
1526{
1527 struct reboot_args bsd_args;
1528
1529#ifdef DEBUG
1530 if (ldebug(reboot))
1531 printf(ARGS(reboot, "0x%x"), args->cmd);
1532#endif
1533
1534 if (args->magic1 != REBOOT_MAGIC1)
1535 return (EINVAL);
1536
1537 switch (args->magic2) {
1538 case REBOOT_MAGIC2:
1539 case REBOOT_MAGIC2A:
1540 case REBOOT_MAGIC2B:
1541 break;
1542 default:
1543 return (EINVAL);
1544 }
1545
1546 switch (args->cmd) {
1547 case REBOOT_CAD_ON:
1548 case REBOOT_CAD_OFF:
1549 return (priv_check(td, PRIV_REBOOT));
1550 case REBOOT_HALT:
1551 bsd_args.opt = RB_HALT;
1552 break;
1553 case REBOOT_RESTART:
1554 case REBOOT_RESTART2:
1555 bsd_args.opt = 0;
1556 break;
1557 case REBOOT_POWEROFF:
1558 bsd_args.opt = RB_POWEROFF;
1559 break;
1560 default:
1561 return (EINVAL);
1562 }
1563 return (sys_reboot(td, &bsd_args));
1564}
1565
1566
1567/*
1568 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1569 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1570 * are assumed to be preserved. The following lightweight syscalls fixes
1571 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1572 *
1573 * linux_getpid() - MP SAFE
1574 * linux_getgid() - MP SAFE
1575 * linux_getuid() - MP SAFE
1576 */
1577
1578int
1579linux_getpid(struct thread *td, struct linux_getpid_args *args)
1580{
1581
1582#ifdef DEBUG
1583 if (ldebug(getpid))
1584 printf(ARGS(getpid, ""));
1585#endif
1586 td->td_retval[0] = td->td_proc->p_pid;
1587
1588 return (0);
1589}
1590
1591int
1592linux_gettid(struct thread *td, struct linux_gettid_args *args)
1593{
1594 struct linux_emuldata *em;
1595
1596#ifdef DEBUG
1597 if (ldebug(gettid))
1598 printf(ARGS(gettid, ""));
1599#endif
1600
1601 em = em_find(td);
1602 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1603
1604 td->td_retval[0] = em->em_tid;
1605
1606 return (0);
1607}
1608
1609
1610int
1611linux_getppid(struct thread *td, struct linux_getppid_args *args)
1612{
1613
1614#ifdef DEBUG
1615 if (ldebug(getppid))
1616 printf(ARGS(getppid, ""));
1617#endif
1618
1619 PROC_LOCK(td->td_proc);
1620 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1621 PROC_UNLOCK(td->td_proc);
1622 return (0);
1623}
1624
1625int
1626linux_getgid(struct thread *td, struct linux_getgid_args *args)
1627{
1628
1629#ifdef DEBUG
1630 if (ldebug(getgid))
1631 printf(ARGS(getgid, ""));
1632#endif
1633
1634 td->td_retval[0] = td->td_ucred->cr_rgid;
1635 return (0);
1636}
1637
1638int
1639linux_getuid(struct thread *td, struct linux_getuid_args *args)
1640{
1641
1642#ifdef DEBUG
1643 if (ldebug(getuid))
1644 printf(ARGS(getuid, ""));
1645#endif
1646
1647 td->td_retval[0] = td->td_ucred->cr_ruid;
1648 return (0);
1649}
1650
1651
1652int
1653linux_getsid(struct thread *td, struct linux_getsid_args *args)
1654{
1655 struct getsid_args bsd;
1656
1657#ifdef DEBUG
1658 if (ldebug(getsid))
1659 printf(ARGS(getsid, "%i"), args->pid);
1660#endif
1661
1662 bsd.pid = args->pid;
1663 return (sys_getsid(td, &bsd));
1664}
1665
1666int
1667linux_nosys(struct thread *td, struct nosys_args *ignore)
1668{
1669
1670 return (ENOSYS);
1671}
1672
1673int
1674linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1675{
1676 struct getpriority_args bsd_args;
1677 int error;
1678
1679#ifdef DEBUG
1680 if (ldebug(getpriority))
1681 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1682#endif
1683
1684 bsd_args.which = args->which;
1685 bsd_args.who = args->who;
1686 error = sys_getpriority(td, &bsd_args);
1687 td->td_retval[0] = 20 - td->td_retval[0];
1688 return (error);
1689}
1690
1691int
1692linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1693{
1694 int name[2];
1695
1696#ifdef DEBUG
1697 if (ldebug(sethostname))
1698 printf(ARGS(sethostname, "*, %i"), args->len);
1699#endif
1700
1701 name[0] = CTL_KERN;
1702 name[1] = KERN_HOSTNAME;
1703 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1704 args->len, 0, 0));
1705}
1706
1707int
1708linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1709{
1710 int name[2];
1711
1712#ifdef DEBUG
1713 if (ldebug(setdomainname))
1714 printf(ARGS(setdomainname, "*, %i"), args->len);
1715#endif
1716
1717 name[0] = CTL_KERN;
1718 name[1] = KERN_NISDOMAINNAME;
1719 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1720 args->len, 0, 0));
1721}
1722
1723int
1724linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1725{
1726
1727#ifdef DEBUG
1728 if (ldebug(exit_group))
1729 printf(ARGS(exit_group, "%i"), args->error_code);
1730#endif
1731
1732 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1733 args->error_code);
1734
1735 /*
1736 * XXX: we should send a signal to the parent if
1737 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1738 * as it doesnt occur often.
1739 */
1740 exit1(td, W_EXITCODE(args->error_code, 0));
1741 /* NOTREACHED */
1742}
1743
1744#define _LINUX_CAPABILITY_VERSION 0x19980330
1745
1746struct l_user_cap_header {
1747 l_int version;
1748 l_int pid;
1749};
1750
1751struct l_user_cap_data {
1752 l_int effective;
1753 l_int permitted;
1754 l_int inheritable;
1755};
1756
1757int
1758linux_capget(struct thread *td, struct linux_capget_args *args)
1759{
1760 struct l_user_cap_header luch;
1761 struct l_user_cap_data lucd;
1762 int error;
1763
1764 if (args->hdrp == NULL)
1765 return (EFAULT);
1766
1767 error = copyin(args->hdrp, &luch, sizeof(luch));
1768 if (error != 0)
1769 return (error);
1770
1771 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1772 luch.version = _LINUX_CAPABILITY_VERSION;
1773 error = copyout(&luch, args->hdrp, sizeof(luch));
1774 if (error)
1775 return (error);
1776 return (EINVAL);
1777 }
1778
1779 if (luch.pid)
1780 return (EPERM);
1781
1782 if (args->datap) {
1783 /*
1784 * The current implementation doesn't support setting
1785 * a capability (it's essentially a stub) so indicate
1786 * that no capabilities are currently set or available
1787 * to request.
1788 */
1789 bzero (&lucd, sizeof(lucd));
1790 error = copyout(&lucd, args->datap, sizeof(lucd));
1791 }
1792
1793 return (error);
1794}
1795
1796int
1797linux_capset(struct thread *td, struct linux_capset_args *args)
1798{
1799 struct l_user_cap_header luch;
1800 struct l_user_cap_data lucd;
1801 int error;
1802
1803 if (args->hdrp == NULL || args->datap == NULL)
1804 return (EFAULT);
1805
1806 error = copyin(args->hdrp, &luch, sizeof(luch));
1807 if (error != 0)
1808 return (error);
1809
1810 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1811 luch.version = _LINUX_CAPABILITY_VERSION;
1812 error = copyout(&luch, args->hdrp, sizeof(luch));
1813 if (error)
1814 return (error);
1815 return (EINVAL);
1816 }
1817
1818 if (luch.pid)
1819 return (EPERM);
1820
1821 error = copyin(args->datap, &lucd, sizeof(lucd));
1822 if (error != 0)
1823 return (error);
1824
1825 /* We currently don't support setting any capabilities. */
1826 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1827 linux_msg(td,
1828 "capset effective=0x%x, permitted=0x%x, "
1829 "inheritable=0x%x is not implemented",
1830 (int)lucd.effective, (int)lucd.permitted,
1831 (int)lucd.inheritable);
1832 return (EPERM);
1833 }
1834
1835 return (0);
1836}
1837
1838int
1839linux_prctl(struct thread *td, struct linux_prctl_args *args)
1840{
1841 int error = 0, max_size;
1842 struct proc *p = td->td_proc;
1843 char comm[LINUX_MAX_COMM_LEN];
1844 struct linux_emuldata *em;
1845 int pdeath_signal;
1846
1847#ifdef DEBUG
1848 if (ldebug(prctl))
1849 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1850 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1851 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1852#endif
1853
1854 switch (args->option) {
1855 case LINUX_PR_SET_PDEATHSIG:
1856 if (!LINUX_SIG_VALID(args->arg2))
1857 return (EINVAL);
1858 em = em_find(td);
1859 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1860 em->pdeath_signal = args->arg2;
1861 break;
1862 case LINUX_PR_GET_PDEATHSIG:
1863 em = em_find(td);
1864 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1865 pdeath_signal = em->pdeath_signal;
1866 error = copyout(&pdeath_signal,
1867 (void *)(register_t)args->arg2,
1868 sizeof(pdeath_signal));
1869 break;
1870 case LINUX_PR_GET_KEEPCAPS:
1871 /*
1872 * Indicate that we always clear the effective and
1873 * permitted capability sets when the user id becomes
1874 * non-zero (actually the capability sets are simply
1875 * always zero in the current implementation).
1876 */
1877 td->td_retval[0] = 0;
1878 break;
1879 case LINUX_PR_SET_KEEPCAPS:
1880 /*
1881 * Ignore requests to keep the effective and permitted
1882 * capability sets when the user id becomes non-zero.
1883 */
1884 break;
1885 case LINUX_PR_SET_NAME:
1886 /*
1887 * To be on the safe side we need to make sure to not
1888 * overflow the size a linux program expects. We already
1889 * do this here in the copyin, so that we don't need to
1890 * check on copyout.
1891 */
1892 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
1893 error = copyinstr((void *)(register_t)args->arg2, comm,
1894 max_size, NULL);
1895
1896 /* Linux silently truncates the name if it is too long. */
1897 if (error == ENAMETOOLONG) {
1898 /*
1899 * XXX: copyinstr() isn't documented to populate the
1900 * array completely, so do a copyin() to be on the
1901 * safe side. This should be changed in case
1902 * copyinstr() is changed to guarantee this.
1903 */
1904 error = copyin((void *)(register_t)args->arg2, comm,
1905 max_size - 1);
1906 comm[max_size - 1] = '\0';
1907 }
1908 if (error)
1909 return (error);
1910
1911 PROC_LOCK(p);
1912 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
1913 PROC_UNLOCK(p);
1914 break;
1915 case LINUX_PR_GET_NAME:
1916 PROC_LOCK(p);
1917 strlcpy(comm, p->p_comm, sizeof(comm));
1918 PROC_UNLOCK(p);
1919 error = copyout(comm, (void *)(register_t)args->arg2,
1920 strlen(comm) + 1);
1921 break;
1922 default:
1923 error = EINVAL;
1924 break;
1925 }
1926
1927 return (error);
1928}
1929
1930int
1931linux_sched_setparam(struct thread *td,
1932 struct linux_sched_setparam_args *uap)
1933{
1934 struct sched_param sched_param;
1935 struct thread *tdt;
1936 int error;
1937
1938#ifdef DEBUG
1939 if (ldebug(sched_setparam))
1940 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
1941#endif
1942
1943 error = copyin(uap->param, &sched_param, sizeof(sched_param));
1944 if (error)
1945 return (error);
1946
1947 tdt = linux_tdfind(td, uap->pid, -1);
1948 if (tdt == NULL)
1949 return (ESRCH);
1950
1951 error = kern_sched_setparam(td, tdt, &sched_param);
1952 PROC_UNLOCK(tdt->td_proc);
1953 return (error);
1954}
1955
1956int
1957linux_sched_getparam(struct thread *td,
1958 struct linux_sched_getparam_args *uap)
1959{
1960 struct sched_param sched_param;
1961 struct thread *tdt;
1962 int error;
1963
1964#ifdef DEBUG
1965 if (ldebug(sched_getparam))
1966 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
1967#endif
1968
1969 tdt = linux_tdfind(td, uap->pid, -1);
1970 if (tdt == NULL)
1971 return (ESRCH);
1972
1973 error = kern_sched_getparam(td, tdt, &sched_param);
1974 PROC_UNLOCK(tdt->td_proc);
1975 if (error == 0)
1976 error = copyout(&sched_param, uap->param,
1977 sizeof(sched_param));
1978 return (error);
1979}
1980
1981/*
1982 * Get affinity of a process.
1983 */
1984int
1985linux_sched_getaffinity(struct thread *td,
1986 struct linux_sched_getaffinity_args *args)
1987{
1988 int error;
1989 struct thread *tdt;
1990 struct cpuset_getaffinity_args cga;
1991
1992#ifdef DEBUG
1993 if (ldebug(sched_getaffinity))
1994 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
1995 args->len);
1996#endif
1997 if (args->len < sizeof(cpuset_t))
1998 return (EINVAL);
1999
2000 tdt = linux_tdfind(td, args->pid, -1);
2001 if (tdt == NULL)
2002 return (ESRCH);
2003
2004 PROC_UNLOCK(tdt->td_proc);
2005 cga.level = CPU_LEVEL_WHICH;
2006 cga.which = CPU_WHICH_TID;
2007 cga.id = tdt->td_tid;
2008 cga.cpusetsize = sizeof(cpuset_t);
2009 cga.mask = (cpuset_t *) args->user_mask_ptr;
2010
2011 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2012 td->td_retval[0] = sizeof(cpuset_t);
2013
2014 return (error);
2015}
2016
2017/*
2018 * Set affinity of a process.
2019 */
2020int
2021linux_sched_setaffinity(struct thread *td,
2022 struct linux_sched_setaffinity_args *args)
2023{
2024 struct cpuset_setaffinity_args csa;
2025 struct thread *tdt;
2026
2027#ifdef DEBUG
2028 if (ldebug(sched_setaffinity))
2029 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2030 args->len);
2031#endif
2032 if (args->len < sizeof(cpuset_t))
2033 return (EINVAL);
2034
2035 tdt = linux_tdfind(td, args->pid, -1);
2036 if (tdt == NULL)
2037 return (ESRCH);
2038
2039 PROC_UNLOCK(tdt->td_proc);
2040 csa.level = CPU_LEVEL_WHICH;
2041 csa.which = CPU_WHICH_TID;
2042 csa.id = tdt->td_tid;
2043 csa.cpusetsize = sizeof(cpuset_t);
2044 csa.mask = (cpuset_t *) args->user_mask_ptr;
2045
2046 return (sys_cpuset_setaffinity(td, &csa));
2047}
2048
2049struct linux_rlimit64 {
2050 uint64_t rlim_cur;
2051 uint64_t rlim_max;
2052};
2053
2054int
2055linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2056{
2057 struct rlimit rlim, nrlim;
2058 struct linux_rlimit64 lrlim;
2059 struct proc *p;
2060 u_int which;
2061 int flags;
2062 int error;
2063
2064#ifdef DEBUG
2065 if (ldebug(prlimit64))
2066 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2067 args->resource, (void *)args->new, (void *)args->old);
2068#endif
2069
2070 if (args->resource >= LINUX_RLIM_NLIMITS)
2071 return (EINVAL);
2072
2073 which = linux_to_bsd_resource[args->resource];
2074 if (which == -1)
2075 return (EINVAL);
2076
2077 if (args->new != NULL) {
2078 /*
2079 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2080 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2081 * as INFINITY so we do not need a conversion even.
2082 */
2083 error = copyin(args->new, &nrlim, sizeof(nrlim));
2084 if (error != 0)
2085 return (error);
2086 }
2087
2088 flags = PGET_HOLD | PGET_NOTWEXIT;
2089 if (args->new != NULL)
2090 flags |= PGET_CANDEBUG;
2091 else
2092 flags |= PGET_CANSEE;
2093 error = pget(args->pid, flags, &p);
2094 if (error != 0)
2095 return (error);
2096
2097 if (args->old != NULL) {
2098 PROC_LOCK(p);
2099 lim_rlimit(p, which, &rlim);
2100 PROC_UNLOCK(p);
2101 if (rlim.rlim_cur == RLIM_INFINITY)
2102 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2103 else
2104 lrlim.rlim_cur = rlim.rlim_cur;
2105 if (rlim.rlim_max == RLIM_INFINITY)
2106 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2107 else
2108 lrlim.rlim_max = rlim.rlim_max;
2109 error = copyout(&lrlim, args->old, sizeof(lrlim));
2110 if (error != 0)
2111 goto out;
2112 }
2113
2114 if (args->new != NULL)
2115 error = kern_proc_setrlimit(td, p, which, &nrlim);
2116
2117 out:
2118 PRELE(p);
2119 return (error);
2120}
2121
2122int
2123linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2124{
2125 struct timeval utv, tv0, tv1, *tvp;
2126 struct l_pselect6arg lpse6;
2127 struct l_timespec lts;
2128 struct timespec uts;
2129 l_sigset_t l_ss;
2130 sigset_t *ssp;
2131 sigset_t ss;
2132 int error;
2133
2134 ssp = NULL;
2135 if (args->sig != NULL) {
2136 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2137 if (error != 0)
2138 return (error);
2139 if (lpse6.ss_len != sizeof(l_ss))
2140 return (EINVAL);
2141 if (lpse6.ss != 0) {
2142 error = copyin(PTRIN(lpse6.ss), &l_ss,
2143 sizeof(l_ss));
2144 if (error != 0)
2145 return (error);
2146 linux_to_bsd_sigset(&l_ss, &ss);
2147 ssp = &ss;
2148 }
2149 }
2150
2151 /*
2152 * Currently glibc changes nanosecond number to microsecond.
2153 * This mean losing precision but for now it is hardly seen.
2154 */
2155 if (args->tsp != NULL) {
2156 error = copyin(args->tsp, <s, sizeof(lts));
2157 if (error != 0)
2158 return (error);
2159 uts.tv_sec = lts.tv_sec;
2160 uts.tv_nsec = lts.tv_nsec;
2161
2162 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2163 if (itimerfix(&utv))
2164 return (EINVAL);
2165
2166 microtime(&tv0);
2167 tvp = &utv;
2168 } else
2169 tvp = NULL;
2170
2171 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2172 args->exceptfds, tvp, ssp, sizeof(l_int) * 8);
2173
2174 if (error == 0 && args->tsp != NULL) {
2175 if (td->td_retval[0] != 0) {
2176 /*
2177 * Compute how much time was left of the timeout,
2178 * by subtracting the current time and the time
2179 * before we started the call, and subtracting
2180 * that result from the user-supplied value.
2181 */
2182
2183 microtime(&tv1);
2184 timevalsub(&tv1, &tv0);
2185 timevalsub(&utv, &tv1);
2186 if (utv.tv_sec < 0)
2187 timevalclear(&utv);
2188 } else
2189 timevalclear(&utv);
2190
2191 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2192 lts.tv_sec = uts.tv_sec;
2193 lts.tv_nsec = uts.tv_nsec;
2194 error = copyout(<s, args->tsp, sizeof(lts));
2195 }
2196
2197 return (error);
2198}
2199
|