32
33#include "opt_compat.h"
34#include "opt_mac.h"
35
36#include <sys/param.h>
37#include <sys/blist.h>
38#include <sys/fcntl.h>
39#if defined(__i386__)
40#include <sys/imgact_aout.h>
41#endif
42#include <sys/jail.h>
43#include <sys/kernel.h>
44#include <sys/limits.h>
45#include <sys/lock.h>
46#include <sys/malloc.h>
47#include <sys/mman.h>
48#include <sys/mount.h>
49#include <sys/mutex.h>
50#include <sys/namei.h>
51#include <sys/priv.h>
52#include <sys/proc.h>
53#include <sys/reboot.h>
54#include <sys/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
67#include <security/mac/mac_framework.h>
68
69#include <vm/vm.h>
70#include <vm/pmap.h>
71#include <vm/vm_kern.h>
72#include <vm/vm_map.h>
73#include <vm/vm_extern.h>
74#include <vm/vm_object.h>
75#include <vm/swap_pager.h>
76
77#include <compat/linux/linux_sysproto.h>
78#include <compat/linux/linux_emul.h>
79#include <compat/linux/linux_misc.h>
80
81#ifdef COMPAT_LINUX32
82#include <machine/../linux32/linux.h>
83#include <machine/../linux32/linux32_proto.h>
84#else
85#include <machine/../linux/linux.h>
86#include <machine/../linux/linux_proto.h>
87#endif
88
89#include <compat/linux/linux_mib.h>
90#include <compat/linux/linux_signal.h>
91#include <compat/linux/linux_util.h>
92
93#ifdef __i386__
94#include <machine/cputypes.h>
95#endif
96
97#define BSD_TO_LINUX_SIGNAL(sig) \
98 (((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig)
99
100static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
101 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
102 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
103 RLIMIT_MEMLOCK, -1
104};
105
106struct l_sysinfo {
107 l_long uptime; /* Seconds since boot */
108 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
109#define LINUX_SYSINFO_LOADS_SCALE 65536
110 l_ulong totalram; /* Total usable main memory size */
111 l_ulong freeram; /* Available memory size */
112 l_ulong sharedram; /* Amount of shared memory */
113 l_ulong bufferram; /* Memory used by buffers */
114 l_ulong totalswap; /* Total swap space size */
115 l_ulong freeswap; /* swap space still available */
116 l_ushort procs; /* Number of current processes */
117 l_ushort pads;
118 l_ulong totalbig;
119 l_ulong freebig;
120 l_uint mem_unit;
121 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
122};
123int
124linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
125{
126 struct l_sysinfo sysinfo;
127 vm_object_t object;
128 int i, j;
129 struct timespec ts;
130
131 getnanouptime(&ts);
132 if (ts.tv_nsec != 0)
133 ts.tv_sec++;
134 sysinfo.uptime = ts.tv_sec;
135
136 /* Use the information from the mib to get our load averages */
137 for (i = 0; i < 3; i++)
138 sysinfo.loads[i] = averunnable.ldavg[i] *
139 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
140
141 sysinfo.totalram = physmem * PAGE_SIZE;
142 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
143
144 sysinfo.sharedram = 0;
145 mtx_lock(&vm_object_list_mtx);
146 TAILQ_FOREACH(object, &vm_object_list, object_list)
147 if (object->shadow_count > 1)
148 sysinfo.sharedram += object->resident_page_count;
149 mtx_unlock(&vm_object_list_mtx);
150
151 sysinfo.sharedram *= PAGE_SIZE;
152 sysinfo.bufferram = 0;
153
154 swap_pager_status(&i, &j);
155 sysinfo.totalswap = i * PAGE_SIZE;
156 sysinfo.freeswap = (i - j) * PAGE_SIZE;
157
158 sysinfo.procs = nprocs;
159
160 /* The following are only present in newer Linux kernels. */
161 sysinfo.totalbig = 0;
162 sysinfo.freebig = 0;
163 sysinfo.mem_unit = 1;
164
165 return copyout(&sysinfo, args->info, sizeof(sysinfo));
166}
167
168int
169linux_alarm(struct thread *td, struct linux_alarm_args *args)
170{
171 struct itimerval it, old_it;
172 int error;
173
174#ifdef DEBUG
175 if (ldebug(alarm))
176 printf(ARGS(alarm, "%u"), args->secs);
177#endif
178
179 if (args->secs > 100000000)
180 return (EINVAL);
181
182 it.it_value.tv_sec = (long)args->secs;
183 it.it_value.tv_usec = 0;
184 it.it_interval.tv_sec = 0;
185 it.it_interval.tv_usec = 0;
186 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
187 if (error)
188 return (error);
189 if (timevalisset(&old_it.it_value)) {
190 if (old_it.it_value.tv_usec != 0)
191 old_it.it_value.tv_sec++;
192 td->td_retval[0] = old_it.it_value.tv_sec;
193 }
194 return (0);
195}
196
197int
198linux_brk(struct thread *td, struct linux_brk_args *args)
199{
200 struct vmspace *vm = td->td_proc->p_vmspace;
201 vm_offset_t new, old;
202 struct obreak_args /* {
203 char * nsize;
204 } */ tmp;
205
206#ifdef DEBUG
207 if (ldebug(brk))
208 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
209#endif
210 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
211 new = (vm_offset_t)args->dsend;
212 tmp.nsize = (char *)new;
213 if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
214 td->td_retval[0] = (long)new;
215 else
216 td->td_retval[0] = (long)old;
217
218 return 0;
219}
220
221#if defined(__i386__)
222/* XXX: what about amd64/linux32? */
223
224int
225linux_uselib(struct thread *td, struct linux_uselib_args *args)
226{
227 struct nameidata ni;
228 struct vnode *vp;
229 struct exec *a_out;
230 struct vattr attr;
231 vm_offset_t vmaddr;
232 unsigned long file_offset;
233 vm_offset_t buffer;
234 unsigned long bss_size;
235 char *library;
236 int error;
237 int locked, vfslocked;
238
239 LCONVPATHEXIST(td, args->library, &library);
240
241#ifdef DEBUG
242 if (ldebug(uselib))
243 printf(ARGS(uselib, "%s"), library);
244#endif
245
246 a_out = NULL;
247 vfslocked = 0;
248 locked = 0;
249 vp = NULL;
250
251 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
252 UIO_SYSSPACE, library, td);
253 error = namei(&ni);
254 LFREEPATH(library);
255 if (error)
256 goto cleanup;
257
258 vp = ni.ni_vp;
259 vfslocked = NDHASGIANT(&ni);
260 NDFREE(&ni, NDF_ONLY_PNBUF);
261
262 /*
263 * From here on down, we have a locked vnode that must be unlocked.
264 * XXX: The code below largely duplicates exec_check_permissions().
265 */
266 locked = 1;
267
268 /* Writable? */
269 if (vp->v_writecount) {
270 error = ETXTBSY;
271 goto cleanup;
272 }
273
274 /* Executable? */
275 error = VOP_GETATTR(vp, &attr, td->td_ucred, td);
276 if (error)
277 goto cleanup;
278
279 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
280 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
281 /* EACCESS is what exec(2) returns. */
282 error = ENOEXEC;
283 goto cleanup;
284 }
285
286 /* Sensible size? */
287 if (attr.va_size == 0) {
288 error = ENOEXEC;
289 goto cleanup;
290 }
291
292 /* Can we access it? */
293 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
294 if (error)
295 goto cleanup;
296
297 /*
298 * XXX: This should use vn_open() so that it is properly authorized,
299 * and to reduce code redundancy all over the place here.
300 * XXX: Not really, it duplicates far more of exec_check_permissions()
301 * than vn_open().
302 */
303#ifdef MAC
304 error = mac_check_vnode_open(td->td_ucred, vp, FREAD);
305 if (error)
306 goto cleanup;
307#endif
308 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, -1);
309 if (error)
310 goto cleanup;
311
312 /* Pull in executable header into kernel_map */
313 error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
314 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
315 if (error)
316 goto cleanup;
317
318 /* Is it a Linux binary ? */
319 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
320 error = ENOEXEC;
321 goto cleanup;
322 }
323
324 /*
325 * While we are here, we should REALLY do some more checks
326 */
327
328 /* Set file/virtual offset based on a.out variant. */
329 switch ((int)(a_out->a_magic & 0xffff)) {
330 case 0413: /* ZMAGIC */
331 file_offset = 1024;
332 break;
333 case 0314: /* QMAGIC */
334 file_offset = 0;
335 break;
336 default:
337 error = ENOEXEC;
338 goto cleanup;
339 }
340
341 bss_size = round_page(a_out->a_bss);
342
343 /* Check various fields in header for validity/bounds. */
344 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
345 error = ENOEXEC;
346 goto cleanup;
347 }
348
349 /* text + data can't exceed file size */
350 if (a_out->a_data + a_out->a_text > attr.va_size) {
351 error = EFAULT;
352 goto cleanup;
353 }
354
355 /*
356 * text/data/bss must not exceed limits
357 * XXX - this is not complete. it should check current usage PLUS
358 * the resources needed by this library.
359 */
360 PROC_LOCK(td->td_proc);
361 if (a_out->a_text > maxtsiz ||
362 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
363 PROC_UNLOCK(td->td_proc);
364 error = ENOMEM;
365 goto cleanup;
366 }
367 PROC_UNLOCK(td->td_proc);
368
369 /*
370 * Prevent more writers.
371 * XXX: Note that if any of the VM operations fail below we don't
372 * clear this flag.
373 */
374 vp->v_vflag |= VV_TEXT;
375
376 /*
377 * Lock no longer needed
378 */
379 locked = 0;
380 VOP_UNLOCK(vp, 0, td);
381 VFS_UNLOCK_GIANT(vfslocked);
382
383 /*
384 * Check if file_offset page aligned. Currently we cannot handle
385 * misalinged file offsets, and so we read in the entire image
386 * (what a waste).
387 */
388 if (file_offset & PAGE_MASK) {
389#ifdef DEBUG
390 printf("uselib: Non page aligned binary %lu\n", file_offset);
391#endif
392 /* Map text+data read/write/execute */
393
394 /* a_entry is the load address and is page aligned */
395 vmaddr = trunc_page(a_out->a_entry);
396
397 /* get anon user mapping, read+write+execute */
398 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
399 &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
400 VM_PROT_ALL, 0);
401 if (error)
402 goto cleanup;
403
404 /* map file into kernel_map */
405 error = vm_mmap(kernel_map, &buffer,
406 round_page(a_out->a_text + a_out->a_data + file_offset),
407 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
408 trunc_page(file_offset));
409 if (error)
410 goto cleanup;
411
412 /* copy from kernel VM space to user space */
413 error = copyout(PTRIN(buffer + file_offset),
414 (void *)vmaddr, a_out->a_text + a_out->a_data);
415
416 /* release temporary kernel space */
417 vm_map_remove(kernel_map, buffer, buffer +
418 round_page(a_out->a_text + a_out->a_data + file_offset));
419
420 if (error)
421 goto cleanup;
422 } else {
423#ifdef DEBUG
424 printf("uselib: Page aligned binary %lu\n", file_offset);
425#endif
426 /*
427 * for QMAGIC, a_entry is 20 bytes beyond the load address
428 * to skip the executable header
429 */
430 vmaddr = trunc_page(a_out->a_entry);
431
432 /*
433 * Map it all into the process's space as a single
434 * copy-on-write "data" segment.
435 */
436 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
437 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
438 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
439 if (error)
440 goto cleanup;
441 }
442#ifdef DEBUG
443 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
444 ((long *)vmaddr)[1]);
445#endif
446 if (bss_size != 0) {
447 /* Calculate BSS start address */
448 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
449 a_out->a_data;
450
451 /* allocate some 'anon' space */
452 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
453 &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
454 if (error)
455 goto cleanup;
456 }
457
458cleanup:
459 /* Unlock vnode if needed */
460 if (locked) {
461 VOP_UNLOCK(vp, 0, td);
462 VFS_UNLOCK_GIANT(vfslocked);
463 }
464
465 /* Release the kernel mapping. */
466 if (a_out)
467 vm_map_remove(kernel_map, (vm_offset_t)a_out,
468 (vm_offset_t)a_out + PAGE_SIZE);
469
470 return error;
471}
472
473#endif /* __i386__ */
474
475int
476linux_select(struct thread *td, struct linux_select_args *args)
477{
478 l_timeval ltv;
479 struct timeval tv0, tv1, utv, *tvp;
480 int error;
481
482#ifdef DEBUG
483 if (ldebug(select))
484 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
485 (void *)args->readfds, (void *)args->writefds,
486 (void *)args->exceptfds, (void *)args->timeout);
487#endif
488
489 /*
490 * Store current time for computation of the amount of
491 * time left.
492 */
493 if (args->timeout) {
494 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
495 goto select_out;
496 utv.tv_sec = ltv.tv_sec;
497 utv.tv_usec = ltv.tv_usec;
498#ifdef DEBUG
499 if (ldebug(select))
500 printf(LMSG("incoming timeout (%jd/%ld)"),
501 (intmax_t)utv.tv_sec, utv.tv_usec);
502#endif
503
504 if (itimerfix(&utv)) {
505 /*
506 * The timeval was invalid. Convert it to something
507 * valid that will act as it does under Linux.
508 */
509 utv.tv_sec += utv.tv_usec / 1000000;
510 utv.tv_usec %= 1000000;
511 if (utv.tv_usec < 0) {
512 utv.tv_sec -= 1;
513 utv.tv_usec += 1000000;
514 }
515 if (utv.tv_sec < 0)
516 timevalclear(&utv);
517 }
518 microtime(&tv0);
519 tvp = &utv;
520 } else
521 tvp = NULL;
522
523 error = kern_select(td, args->nfds, args->readfds, args->writefds,
524 args->exceptfds, tvp);
525
526#ifdef DEBUG
527 if (ldebug(select))
528 printf(LMSG("real select returns %d"), error);
529#endif
530 if (error) {
531 /*
532 * See fs/select.c in the Linux kernel. Without this,
533 * Maelstrom doesn't work.
534 */
535 if (error == ERESTART)
536 error = EINTR;
537 goto select_out;
538 }
539
540 if (args->timeout) {
541 if (td->td_retval[0]) {
542 /*
543 * Compute how much time was left of the timeout,
544 * by subtracting the current time and the time
545 * before we started the call, and subtracting
546 * that result from the user-supplied value.
547 */
548 microtime(&tv1);
549 timevalsub(&tv1, &tv0);
550 timevalsub(&utv, &tv1);
551 if (utv.tv_sec < 0)
552 timevalclear(&utv);
553 } else
554 timevalclear(&utv);
555#ifdef DEBUG
556 if (ldebug(select))
557 printf(LMSG("outgoing timeout (%jd/%ld)"),
558 (intmax_t)utv.tv_sec, utv.tv_usec);
559#endif
560 ltv.tv_sec = utv.tv_sec;
561 ltv.tv_usec = utv.tv_usec;
562 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
563 goto select_out;
564 }
565
566select_out:
567#ifdef DEBUG
568 if (ldebug(select))
569 printf(LMSG("select_out -> %d"), error);
570#endif
571 return error;
572}
573
574int
575linux_mremap(struct thread *td, struct linux_mremap_args *args)
576{
577 struct munmap_args /* {
578 void *addr;
579 size_t len;
580 } */ bsd_args;
581 int error = 0;
582
583#ifdef DEBUG
584 if (ldebug(mremap))
585 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
586 (void *)(uintptr_t)args->addr,
587 (unsigned long)args->old_len,
588 (unsigned long)args->new_len,
589 (unsigned long)args->flags);
590#endif
591 args->new_len = round_page(args->new_len);
592 args->old_len = round_page(args->old_len);
593
594 if (args->new_len > args->old_len) {
595 td->td_retval[0] = 0;
596 return ENOMEM;
597 }
598
599 if (args->new_len < args->old_len) {
600 bsd_args.addr =
601 (caddr_t)((uintptr_t)args->addr + args->new_len);
602 bsd_args.len = args->old_len - args->new_len;
603 error = munmap(td, &bsd_args);
604 }
605
606 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
607 return error;
608}
609
610#define LINUX_MS_ASYNC 0x0001
611#define LINUX_MS_INVALIDATE 0x0002
612#define LINUX_MS_SYNC 0x0004
613
614int
615linux_msync(struct thread *td, struct linux_msync_args *args)
616{
617 struct msync_args bsd_args;
618
619 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
620 bsd_args.len = (uintptr_t)args->len;
621 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
622
623 return msync(td, &bsd_args);
624}
625
626int
627linux_time(struct thread *td, struct linux_time_args *args)
628{
629 struct timeval tv;
630 l_time_t tm;
631 int error;
632
633#ifdef DEBUG
634 if (ldebug(time))
635 printf(ARGS(time, "*"));
636#endif
637
638 microtime(&tv);
639 tm = tv.tv_sec;
640 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
641 return error;
642 td->td_retval[0] = tm;
643 return 0;
644}
645
646struct l_times_argv {
647 l_long tms_utime;
648 l_long tms_stime;
649 l_long tms_cutime;
650 l_long tms_cstime;
651};
652
653#define CLK_TCK 100 /* Linux uses 100 */
654
655#define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
656
657int
658linux_times(struct thread *td, struct linux_times_args *args)
659{
660 struct timeval tv, utime, stime, cutime, cstime;
661 struct l_times_argv tms;
662 struct proc *p;
663 int error;
664
665#ifdef DEBUG
666 if (ldebug(times))
667 printf(ARGS(times, "*"));
668#endif
669
670 if (args->buf != NULL) {
671 p = td->td_proc;
672 PROC_LOCK(p);
673 calcru(p, &utime, &stime);
674 calccru(p, &cutime, &cstime);
675 PROC_UNLOCK(p);
676
677 tms.tms_utime = CONVTCK(utime);
678 tms.tms_stime = CONVTCK(stime);
679
680 tms.tms_cutime = CONVTCK(cutime);
681 tms.tms_cstime = CONVTCK(cstime);
682
683 if ((error = copyout(&tms, args->buf, sizeof(tms))))
684 return error;
685 }
686
687 microuptime(&tv);
688 td->td_retval[0] = (int)CONVTCK(tv);
689 return 0;
690}
691
692int
693linux_newuname(struct thread *td, struct linux_newuname_args *args)
694{
695 struct l_new_utsname utsname;
696 char osname[LINUX_MAX_UTSNAME];
697 char osrelease[LINUX_MAX_UTSNAME];
698 char *p;
699
700#ifdef DEBUG
701 if (ldebug(newuname))
702 printf(ARGS(newuname, "*"));
703#endif
704
705 linux_get_osname(td, osname);
706 linux_get_osrelease(td, osrelease);
707
708 bzero(&utsname, sizeof(utsname));
709 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
710 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
711 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
712 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
713 for (p = utsname.version; *p != '\0'; ++p)
714 if (*p == '\n') {
715 *p = '\0';
716 break;
717 }
718#ifdef __i386__
719 {
720 const char *class;
721
722 switch (cpu_class) {
723 case CPUCLASS_686:
724 class = "i686";
725 break;
726 case CPUCLASS_586:
727 class = "i586";
728 break;
729 case CPUCLASS_486:
730 class = "i486";
731 break;
732 default:
733 class = "i386";
734 }
735 strlcpy(utsname.machine, class, LINUX_MAX_UTSNAME);
736 }
737#elif defined(__amd64__) /* XXX: Linux can change 'personality'. */
738#ifdef COMPAT_LINUX32
739 strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
740#else
741 strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
742#endif /* COMPAT_LINUX32 */
743#else /* something other than i386 or amd64 - assume we and Linux agree */
744 strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME);
745#endif /* __i386__ */
746 strlcpy(utsname.domainname, domainname, LINUX_MAX_UTSNAME);
747
748 return (copyout(&utsname, args->buf, sizeof(utsname)));
749}
750
751#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
752struct l_utimbuf {
753 l_time_t l_actime;
754 l_time_t l_modtime;
755};
756
757int
758linux_utime(struct thread *td, struct linux_utime_args *args)
759{
760 struct timeval tv[2], *tvp;
761 struct l_utimbuf lut;
762 char *fname;
763 int error;
764
765 LCONVPATHEXIST(td, args->fname, &fname);
766
767#ifdef DEBUG
768 if (ldebug(utime))
769 printf(ARGS(utime, "%s, *"), fname);
770#endif
771
772 if (args->times) {
773 if ((error = copyin(args->times, &lut, sizeof lut))) {
774 LFREEPATH(fname);
775 return error;
776 }
777 tv[0].tv_sec = lut.l_actime;
778 tv[0].tv_usec = 0;
779 tv[1].tv_sec = lut.l_modtime;
780 tv[1].tv_usec = 0;
781 tvp = tv;
782 } else
783 tvp = NULL;
784
785 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
786 LFREEPATH(fname);
787 return (error);
788}
789
790int
791linux_utimes(struct thread *td, struct linux_utimes_args *args)
792{
793 l_timeval ltv[2];
794 struct timeval tv[2], *tvp = NULL;
795 char *fname;
796 int error;
797
798 LCONVPATHEXIST(td, args->fname, &fname);
799
800#ifdef DEBUG
801 if (ldebug(utimes))
802 printf(ARGS(utimes, "%s, *"), fname);
803#endif
804
805 if (args->tptr != NULL) {
806 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
807 LFREEPATH(fname);
808 return (error);
809 }
810 tv[0].tv_sec = ltv[0].tv_sec;
811 tv[0].tv_usec = ltv[0].tv_usec;
812 tv[1].tv_sec = ltv[1].tv_sec;
813 tv[1].tv_usec = ltv[1].tv_usec;
814 tvp = tv;
815 }
816
817 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
818 LFREEPATH(fname);
819 return (error);
820}
821#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
822
823#define __WCLONE 0x80000000
824
825int
826linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
827{
828 int error, options, tmpstat;
829
830#ifdef DEBUG
831 if (ldebug(waitpid))
832 printf(ARGS(waitpid, "%d, %p, %d"),
833 args->pid, (void *)args->status, args->options);
834#endif
835 /*
836 * this is necessary because the test in kern_wait doesnt work
837 * because we mess with the options here
838 */
839 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
840 return (EINVAL);
841
842 options = (args->options & (WNOHANG | WUNTRACED));
843 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
844 if (args->options & __WCLONE)
845 options |= WLINUXCLONE;
846
847 error = kern_wait(td, args->pid, &tmpstat, options, NULL);
848 if (error)
849 return error;
850
851 if (args->status) {
852 tmpstat &= 0xffff;
853 if (WIFSIGNALED(tmpstat))
854 tmpstat = (tmpstat & 0xffffff80) |
855 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
856 else if (WIFSTOPPED(tmpstat))
857 tmpstat = (tmpstat & 0xffff00ff) |
858 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
859 return copyout(&tmpstat, args->status, sizeof(int));
860 }
861
862 return 0;
863}
864
865int
866linux_wait4(struct thread *td, struct linux_wait4_args *args)
867{
868 int error, options, tmpstat;
869 struct rusage ru, *rup;
870 struct proc *p;
871
872#ifdef DEBUG
873 if (ldebug(wait4))
874 printf(ARGS(wait4, "%d, %p, %d, %p"),
875 args->pid, (void *)args->status, args->options,
876 (void *)args->rusage);
877#endif
878
879 options = (args->options & (WNOHANG | WUNTRACED));
880 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
881 if (args->options & __WCLONE)
882 options |= WLINUXCLONE;
883
884 if (args->rusage != NULL)
885 rup = &ru;
886 else
887 rup = NULL;
888 error = kern_wait(td, args->pid, &tmpstat, options, rup);
889 if (error)
890 return error;
891
892 p = td->td_proc;
893 PROC_LOCK(p);
894 sigqueue_delete(&p->p_sigqueue, SIGCHLD);
895 PROC_UNLOCK(p);
896
897 if (args->status) {
898 tmpstat &= 0xffff;
899 if (WIFSIGNALED(tmpstat))
900 tmpstat = (tmpstat & 0xffffff80) |
901 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
902 else if (WIFSTOPPED(tmpstat))
903 tmpstat = (tmpstat & 0xffff00ff) |
904 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
905 error = copyout(&tmpstat, args->status, sizeof(int));
906 }
907 if (args->rusage != NULL && error == 0)
908 error = copyout(&ru, args->rusage, sizeof(ru));
909
910 return (error);
911}
912
913int
914linux_mknod(struct thread *td, struct linux_mknod_args *args)
915{
916 char *path;
917 int error;
918
919 LCONVPATHCREAT(td, args->path, &path);
920
921#ifdef DEBUG
922 if (ldebug(mknod))
923 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
924#endif
925
926 switch (args->mode & S_IFMT) {
927 case S_IFIFO:
928 case S_IFSOCK:
929 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
930 break;
931
932 case S_IFCHR:
933 case S_IFBLK:
934 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
935 args->dev);
936 break;
937
938 case S_IFDIR:
939 error = EPERM;
940 break;
941
942 case 0:
943 args->mode |= S_IFREG;
944 /* FALLTHROUGH */
945 case S_IFREG:
946 error = kern_open(td, path, UIO_SYSSPACE,
947 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
948 break;
949
950 default:
951 error = EINVAL;
952 break;
953 }
954 LFREEPATH(path);
955 return (error);
956}
957
958/*
959 * UGH! This is just about the dumbest idea I've ever heard!!
960 */
961int
962linux_personality(struct thread *td, struct linux_personality_args *args)
963{
964#ifdef DEBUG
965 if (ldebug(personality))
966 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
967#endif
968 if (args->per != 0)
969 return EINVAL;
970
971 /* Yes Jim, it's still a Linux... */
972 td->td_retval[0] = 0;
973 return 0;
974}
975
976struct l_itimerval {
977 l_timeval it_interval;
978 l_timeval it_value;
979};
980
981#define B2L_ITIMERVAL(bip, lip) \
982 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
983 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
984 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
985 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
986
987int
988linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
989{
990 int error;
991 struct l_itimerval ls;
992 struct itimerval aitv, oitv;
993
994#ifdef DEBUG
995 if (ldebug(setitimer))
996 printf(ARGS(setitimer, "%p, %p"),
997 (void *)uap->itv, (void *)uap->oitv);
998#endif
999
1000 if (uap->itv == NULL) {
1001 uap->itv = uap->oitv;
1002 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1003 }
1004
1005 error = copyin(uap->itv, &ls, sizeof(ls));
1006 if (error != 0)
1007 return (error);
1008 B2L_ITIMERVAL(&aitv, &ls);
1009#ifdef DEBUG
1010 if (ldebug(setitimer)) {
1011 printf("setitimer: value: sec: %jd, usec: %ld\n",
1012 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1013 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1014 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1015 }
1016#endif
1017 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1018 if (error != 0 || uap->oitv == NULL)
1019 return (error);
1020 B2L_ITIMERVAL(&ls, &oitv);
1021
1022 return (copyout(&ls, uap->oitv, sizeof(ls)));
1023}
1024
1025int
1026linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1027{
1028 int error;
1029 struct l_itimerval ls;
1030 struct itimerval aitv;
1031
1032#ifdef DEBUG
1033 if (ldebug(getitimer))
1034 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1035#endif
1036 error = kern_getitimer(td, uap->which, &aitv);
1037 if (error != 0)
1038 return (error);
1039 B2L_ITIMERVAL(&ls, &aitv);
1040 return (copyout(&ls, uap->itv, sizeof(ls)));
1041}
1042
1043int
1044linux_nice(struct thread *td, struct linux_nice_args *args)
1045{
1046 struct setpriority_args bsd_args;
1047
1048 bsd_args.which = PRIO_PROCESS;
1049 bsd_args.who = 0; /* current process */
1050 bsd_args.prio = args->inc;
1051 return setpriority(td, &bsd_args);
1052}
1053
1054int
1055linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1056{
1057 struct ucred *newcred, *oldcred;
1058 l_gid_t linux_gidset[NGROUPS];
1059 gid_t *bsd_gidset;
1060 int ngrp, error;
1061 struct proc *p;
1062
1063 ngrp = args->gidsetsize;
1064 if (ngrp < 0 || ngrp >= NGROUPS)
1065 return (EINVAL);
1066 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1067 if (error)
1068 return (error);
1069 newcred = crget();
1070 p = td->td_proc;
1071 PROC_LOCK(p);
1072 oldcred = p->p_ucred;
1073
1074 /*
1075 * cr_groups[0] holds egid. Setting the whole set from
1076 * the supplied set will cause egid to be changed too.
1077 * Keep cr_groups[0] unchanged to prevent that.
1078 */
1079
1080 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS,
1081 SUSER_ALLOWJAIL)) != 0) {
1082 PROC_UNLOCK(p);
1083 crfree(newcred);
1084 return (error);
1085 }
1086
1087 crcopy(newcred, oldcred);
1088 if (ngrp > 0) {
1089 newcred->cr_ngroups = ngrp + 1;
1090
1091 bsd_gidset = newcred->cr_groups;
1092 ngrp--;
1093 while (ngrp >= 0) {
1094 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1095 ngrp--;
1096 }
1097 } else
1098 newcred->cr_ngroups = 1;
1099
1100 setsugid(p);
1101 p->p_ucred = newcred;
1102 PROC_UNLOCK(p);
1103 crfree(oldcred);
1104 return (0);
1105}
1106
1107int
1108linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1109{
1110 struct ucred *cred;
1111 l_gid_t linux_gidset[NGROUPS];
1112 gid_t *bsd_gidset;
1113 int bsd_gidsetsz, ngrp, error;
1114
1115 cred = td->td_ucred;
1116 bsd_gidset = cred->cr_groups;
1117 bsd_gidsetsz = cred->cr_ngroups - 1;
1118
1119 /*
1120 * cr_groups[0] holds egid. Returning the whole set
1121 * here will cause a duplicate. Exclude cr_groups[0]
1122 * to prevent that.
1123 */
1124
1125 if ((ngrp = args->gidsetsize) == 0) {
1126 td->td_retval[0] = bsd_gidsetsz;
1127 return (0);
1128 }
1129
1130 if (ngrp < bsd_gidsetsz)
1131 return (EINVAL);
1132
1133 ngrp = 0;
1134 while (ngrp < bsd_gidsetsz) {
1135 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1136 ngrp++;
1137 }
1138
1139 if ((error = copyout(linux_gidset, args->grouplist,
1140 ngrp * sizeof(l_gid_t))))
1141 return (error);
1142
1143 td->td_retval[0] = ngrp;
1144 return (0);
1145}
1146
1147int
1148linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1149{
1150 struct rlimit bsd_rlim;
1151 struct l_rlimit rlim;
1152 u_int which;
1153 int error;
1154
1155#ifdef DEBUG
1156 if (ldebug(setrlimit))
1157 printf(ARGS(setrlimit, "%d, %p"),
1158 args->resource, (void *)args->rlim);
1159#endif
1160
1161 if (args->resource >= LINUX_RLIM_NLIMITS)
1162 return (EINVAL);
1163
1164 which = linux_to_bsd_resource[args->resource];
1165 if (which == -1)
1166 return (EINVAL);
1167
1168 error = copyin(args->rlim, &rlim, sizeof(rlim));
1169 if (error)
1170 return (error);
1171
1172 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1173 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1174 return (kern_setrlimit(td, which, &bsd_rlim));
1175}
1176
1177int
1178linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1179{
1180 struct l_rlimit rlim;
1181 struct proc *p = td->td_proc;
1182 struct rlimit bsd_rlim;
1183 u_int which;
1184
1185#ifdef DEBUG
1186 if (ldebug(old_getrlimit))
1187 printf(ARGS(old_getrlimit, "%d, %p"),
1188 args->resource, (void *)args->rlim);
1189#endif
1190
1191 if (args->resource >= LINUX_RLIM_NLIMITS)
1192 return (EINVAL);
1193
1194 which = linux_to_bsd_resource[args->resource];
1195 if (which == -1)
1196 return (EINVAL);
1197
1198 PROC_LOCK(p);
1199 lim_rlimit(p, which, &bsd_rlim);
1200 PROC_UNLOCK(p);
1201
1202#ifdef COMPAT_LINUX32
1203 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1204 if (rlim.rlim_cur == UINT_MAX)
1205 rlim.rlim_cur = INT_MAX;
1206 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1207 if (rlim.rlim_max == UINT_MAX)
1208 rlim.rlim_max = INT_MAX;
1209#else
1210 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1211 if (rlim.rlim_cur == ULONG_MAX)
1212 rlim.rlim_cur = LONG_MAX;
1213 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1214 if (rlim.rlim_max == ULONG_MAX)
1215 rlim.rlim_max = LONG_MAX;
1216#endif
1217 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1218}
1219
1220int
1221linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1222{
1223 struct l_rlimit rlim;
1224 struct proc *p = td->td_proc;
1225 struct rlimit bsd_rlim;
1226 u_int which;
1227
1228#ifdef DEBUG
1229 if (ldebug(getrlimit))
1230 printf(ARGS(getrlimit, "%d, %p"),
1231 args->resource, (void *)args->rlim);
1232#endif
1233
1234 if (args->resource >= LINUX_RLIM_NLIMITS)
1235 return (EINVAL);
1236
1237 which = linux_to_bsd_resource[args->resource];
1238 if (which == -1)
1239 return (EINVAL);
1240
1241 PROC_LOCK(p);
1242 lim_rlimit(p, which, &bsd_rlim);
1243 PROC_UNLOCK(p);
1244
1245 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1246 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1247 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1248}
1249
1250int
1251linux_sched_setscheduler(struct thread *td,
1252 struct linux_sched_setscheduler_args *args)
1253{
1254 struct sched_setscheduler_args bsd;
1255
1256#ifdef DEBUG
1257 if (ldebug(sched_setscheduler))
1258 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1259 args->pid, args->policy, (const void *)args->param);
1260#endif
1261
1262 switch (args->policy) {
1263 case LINUX_SCHED_OTHER:
1264 bsd.policy = SCHED_OTHER;
1265 break;
1266 case LINUX_SCHED_FIFO:
1267 bsd.policy = SCHED_FIFO;
1268 break;
1269 case LINUX_SCHED_RR:
1270 bsd.policy = SCHED_RR;
1271 break;
1272 default:
1273 return EINVAL;
1274 }
1275
1276 bsd.pid = args->pid;
1277 bsd.param = (struct sched_param *)args->param;
1278 return sched_setscheduler(td, &bsd);
1279}
1280
1281int
1282linux_sched_getscheduler(struct thread *td,
1283 struct linux_sched_getscheduler_args *args)
1284{
1285 struct sched_getscheduler_args bsd;
1286 int error;
1287
1288#ifdef DEBUG
1289 if (ldebug(sched_getscheduler))
1290 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1291#endif
1292
1293 bsd.pid = args->pid;
1294 error = sched_getscheduler(td, &bsd);
1295
1296 switch (td->td_retval[0]) {
1297 case SCHED_OTHER:
1298 td->td_retval[0] = LINUX_SCHED_OTHER;
1299 break;
1300 case SCHED_FIFO:
1301 td->td_retval[0] = LINUX_SCHED_FIFO;
1302 break;
1303 case SCHED_RR:
1304 td->td_retval[0] = LINUX_SCHED_RR;
1305 break;
1306 }
1307
1308 return error;
1309}
1310
1311int
1312linux_sched_get_priority_max(struct thread *td,
1313 struct linux_sched_get_priority_max_args *args)
1314{
1315 struct sched_get_priority_max_args bsd;
1316
1317#ifdef DEBUG
1318 if (ldebug(sched_get_priority_max))
1319 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1320#endif
1321
1322 switch (args->policy) {
1323 case LINUX_SCHED_OTHER:
1324 bsd.policy = SCHED_OTHER;
1325 break;
1326 case LINUX_SCHED_FIFO:
1327 bsd.policy = SCHED_FIFO;
1328 break;
1329 case LINUX_SCHED_RR:
1330 bsd.policy = SCHED_RR;
1331 break;
1332 default:
1333 return EINVAL;
1334 }
1335 return sched_get_priority_max(td, &bsd);
1336}
1337
1338int
1339linux_sched_get_priority_min(struct thread *td,
1340 struct linux_sched_get_priority_min_args *args)
1341{
1342 struct sched_get_priority_min_args bsd;
1343
1344#ifdef DEBUG
1345 if (ldebug(sched_get_priority_min))
1346 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1347#endif
1348
1349 switch (args->policy) {
1350 case LINUX_SCHED_OTHER:
1351 bsd.policy = SCHED_OTHER;
1352 break;
1353 case LINUX_SCHED_FIFO:
1354 bsd.policy = SCHED_FIFO;
1355 break;
1356 case LINUX_SCHED_RR:
1357 bsd.policy = SCHED_RR;
1358 break;
1359 default:
1360 return EINVAL;
1361 }
1362 return sched_get_priority_min(td, &bsd);
1363}
1364
1365#define REBOOT_CAD_ON 0x89abcdef
1366#define REBOOT_CAD_OFF 0
1367#define REBOOT_HALT 0xcdef0123
1368#define REBOOT_RESTART 0x01234567
1369#define REBOOT_RESTART2 0xA1B2C3D4
1370#define REBOOT_POWEROFF 0x4321FEDC
1371#define REBOOT_MAGIC1 0xfee1dead
1372#define REBOOT_MAGIC2 0x28121969
1373#define REBOOT_MAGIC2A 0x05121996
1374#define REBOOT_MAGIC2B 0x16041998
1375
1376int
1377linux_reboot(struct thread *td, struct linux_reboot_args *args)
1378{
1379 struct reboot_args bsd_args;
1380
1381#ifdef DEBUG
1382 if (ldebug(reboot))
1383 printf(ARGS(reboot, "0x%x"), args->cmd);
1384#endif
1385
1386 if (args->magic1 != REBOOT_MAGIC1)
1387 return EINVAL;
1388
1389 switch (args->magic2) {
1390 case REBOOT_MAGIC2:
1391 case REBOOT_MAGIC2A:
1392 case REBOOT_MAGIC2B:
1393 break;
1394 default:
1395 return EINVAL;
1396 }
1397
1398 switch (args->cmd) {
1399 case REBOOT_CAD_ON:
1400 case REBOOT_CAD_OFF:
1401 return (priv_check(td, PRIV_REBOOT));
1402 case REBOOT_HALT:
1403 bsd_args.opt = RB_HALT;
1404 break;
1405 case REBOOT_RESTART:
1406 case REBOOT_RESTART2:
1407 bsd_args.opt = 0;
1408 break;
1409 case REBOOT_POWEROFF:
1410 bsd_args.opt = RB_POWEROFF;
1411 break;
1412 default:
1413 return EINVAL;
1414 }
1415 return reboot(td, &bsd_args);
1416}
1417
1418
1419/*
1420 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1421 * td->td_retval[1] when COMPAT_43 is defined. This
1422 * globbers registers that are assumed to be preserved. The following
1423 * lightweight syscalls fixes this. See also linux_getgid16() and
1424 * linux_getuid16() in linux_uid16.c.
1425 *
1426 * linux_getpid() - MP SAFE
1427 * linux_getgid() - MP SAFE
1428 * linux_getuid() - MP SAFE
1429 */
1430
1431int
1432linux_getpid(struct thread *td, struct linux_getpid_args *args)
1433{
1434 struct linux_emuldata *em;
1435
1436#ifdef DEBUG
1437 if (ldebug(getpid))
1438 printf(ARGS(getpid, ""));
1439#endif
1440
1441 if (linux_use26(td)) {
| 32
33#include "opt_compat.h"
34#include "opt_mac.h"
35
36#include <sys/param.h>
37#include <sys/blist.h>
38#include <sys/fcntl.h>
39#if defined(__i386__)
40#include <sys/imgact_aout.h>
41#endif
42#include <sys/jail.h>
43#include <sys/kernel.h>
44#include <sys/limits.h>
45#include <sys/lock.h>
46#include <sys/malloc.h>
47#include <sys/mman.h>
48#include <sys/mount.h>
49#include <sys/mutex.h>
50#include <sys/namei.h>
51#include <sys/priv.h>
52#include <sys/proc.h>
53#include <sys/reboot.h>
54#include <sys/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
67#include <security/mac/mac_framework.h>
68
69#include <vm/vm.h>
70#include <vm/pmap.h>
71#include <vm/vm_kern.h>
72#include <vm/vm_map.h>
73#include <vm/vm_extern.h>
74#include <vm/vm_object.h>
75#include <vm/swap_pager.h>
76
77#include <compat/linux/linux_sysproto.h>
78#include <compat/linux/linux_emul.h>
79#include <compat/linux/linux_misc.h>
80
81#ifdef COMPAT_LINUX32
82#include <machine/../linux32/linux.h>
83#include <machine/../linux32/linux32_proto.h>
84#else
85#include <machine/../linux/linux.h>
86#include <machine/../linux/linux_proto.h>
87#endif
88
89#include <compat/linux/linux_mib.h>
90#include <compat/linux/linux_signal.h>
91#include <compat/linux/linux_util.h>
92
93#ifdef __i386__
94#include <machine/cputypes.h>
95#endif
96
97#define BSD_TO_LINUX_SIGNAL(sig) \
98 (((sig) <= LINUX_SIGTBLSZ) ? bsd_to_linux_signal[_SIG_IDX(sig)] : sig)
99
100static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
101 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
102 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
103 RLIMIT_MEMLOCK, -1
104};
105
106struct l_sysinfo {
107 l_long uptime; /* Seconds since boot */
108 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
109#define LINUX_SYSINFO_LOADS_SCALE 65536
110 l_ulong totalram; /* Total usable main memory size */
111 l_ulong freeram; /* Available memory size */
112 l_ulong sharedram; /* Amount of shared memory */
113 l_ulong bufferram; /* Memory used by buffers */
114 l_ulong totalswap; /* Total swap space size */
115 l_ulong freeswap; /* swap space still available */
116 l_ushort procs; /* Number of current processes */
117 l_ushort pads;
118 l_ulong totalbig;
119 l_ulong freebig;
120 l_uint mem_unit;
121 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
122};
123int
124linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
125{
126 struct l_sysinfo sysinfo;
127 vm_object_t object;
128 int i, j;
129 struct timespec ts;
130
131 getnanouptime(&ts);
132 if (ts.tv_nsec != 0)
133 ts.tv_sec++;
134 sysinfo.uptime = ts.tv_sec;
135
136 /* Use the information from the mib to get our load averages */
137 for (i = 0; i < 3; i++)
138 sysinfo.loads[i] = averunnable.ldavg[i] *
139 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
140
141 sysinfo.totalram = physmem * PAGE_SIZE;
142 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
143
144 sysinfo.sharedram = 0;
145 mtx_lock(&vm_object_list_mtx);
146 TAILQ_FOREACH(object, &vm_object_list, object_list)
147 if (object->shadow_count > 1)
148 sysinfo.sharedram += object->resident_page_count;
149 mtx_unlock(&vm_object_list_mtx);
150
151 sysinfo.sharedram *= PAGE_SIZE;
152 sysinfo.bufferram = 0;
153
154 swap_pager_status(&i, &j);
155 sysinfo.totalswap = i * PAGE_SIZE;
156 sysinfo.freeswap = (i - j) * PAGE_SIZE;
157
158 sysinfo.procs = nprocs;
159
160 /* The following are only present in newer Linux kernels. */
161 sysinfo.totalbig = 0;
162 sysinfo.freebig = 0;
163 sysinfo.mem_unit = 1;
164
165 return copyout(&sysinfo, args->info, sizeof(sysinfo));
166}
167
168int
169linux_alarm(struct thread *td, struct linux_alarm_args *args)
170{
171 struct itimerval it, old_it;
172 int error;
173
174#ifdef DEBUG
175 if (ldebug(alarm))
176 printf(ARGS(alarm, "%u"), args->secs);
177#endif
178
179 if (args->secs > 100000000)
180 return (EINVAL);
181
182 it.it_value.tv_sec = (long)args->secs;
183 it.it_value.tv_usec = 0;
184 it.it_interval.tv_sec = 0;
185 it.it_interval.tv_usec = 0;
186 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
187 if (error)
188 return (error);
189 if (timevalisset(&old_it.it_value)) {
190 if (old_it.it_value.tv_usec != 0)
191 old_it.it_value.tv_sec++;
192 td->td_retval[0] = old_it.it_value.tv_sec;
193 }
194 return (0);
195}
196
197int
198linux_brk(struct thread *td, struct linux_brk_args *args)
199{
200 struct vmspace *vm = td->td_proc->p_vmspace;
201 vm_offset_t new, old;
202 struct obreak_args /* {
203 char * nsize;
204 } */ tmp;
205
206#ifdef DEBUG
207 if (ldebug(brk))
208 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
209#endif
210 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
211 new = (vm_offset_t)args->dsend;
212 tmp.nsize = (char *)new;
213 if (((caddr_t)new > vm->vm_daddr) && !obreak(td, &tmp))
214 td->td_retval[0] = (long)new;
215 else
216 td->td_retval[0] = (long)old;
217
218 return 0;
219}
220
221#if defined(__i386__)
222/* XXX: what about amd64/linux32? */
223
224int
225linux_uselib(struct thread *td, struct linux_uselib_args *args)
226{
227 struct nameidata ni;
228 struct vnode *vp;
229 struct exec *a_out;
230 struct vattr attr;
231 vm_offset_t vmaddr;
232 unsigned long file_offset;
233 vm_offset_t buffer;
234 unsigned long bss_size;
235 char *library;
236 int error;
237 int locked, vfslocked;
238
239 LCONVPATHEXIST(td, args->library, &library);
240
241#ifdef DEBUG
242 if (ldebug(uselib))
243 printf(ARGS(uselib, "%s"), library);
244#endif
245
246 a_out = NULL;
247 vfslocked = 0;
248 locked = 0;
249 vp = NULL;
250
251 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | MPSAFE | AUDITVNODE1,
252 UIO_SYSSPACE, library, td);
253 error = namei(&ni);
254 LFREEPATH(library);
255 if (error)
256 goto cleanup;
257
258 vp = ni.ni_vp;
259 vfslocked = NDHASGIANT(&ni);
260 NDFREE(&ni, NDF_ONLY_PNBUF);
261
262 /*
263 * From here on down, we have a locked vnode that must be unlocked.
264 * XXX: The code below largely duplicates exec_check_permissions().
265 */
266 locked = 1;
267
268 /* Writable? */
269 if (vp->v_writecount) {
270 error = ETXTBSY;
271 goto cleanup;
272 }
273
274 /* Executable? */
275 error = VOP_GETATTR(vp, &attr, td->td_ucred, td);
276 if (error)
277 goto cleanup;
278
279 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
280 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
281 /* EACCESS is what exec(2) returns. */
282 error = ENOEXEC;
283 goto cleanup;
284 }
285
286 /* Sensible size? */
287 if (attr.va_size == 0) {
288 error = ENOEXEC;
289 goto cleanup;
290 }
291
292 /* Can we access it? */
293 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
294 if (error)
295 goto cleanup;
296
297 /*
298 * XXX: This should use vn_open() so that it is properly authorized,
299 * and to reduce code redundancy all over the place here.
300 * XXX: Not really, it duplicates far more of exec_check_permissions()
301 * than vn_open().
302 */
303#ifdef MAC
304 error = mac_check_vnode_open(td->td_ucred, vp, FREAD);
305 if (error)
306 goto cleanup;
307#endif
308 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, -1);
309 if (error)
310 goto cleanup;
311
312 /* Pull in executable header into kernel_map */
313 error = vm_mmap(kernel_map, (vm_offset_t *)&a_out, PAGE_SIZE,
314 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
315 if (error)
316 goto cleanup;
317
318 /* Is it a Linux binary ? */
319 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
320 error = ENOEXEC;
321 goto cleanup;
322 }
323
324 /*
325 * While we are here, we should REALLY do some more checks
326 */
327
328 /* Set file/virtual offset based on a.out variant. */
329 switch ((int)(a_out->a_magic & 0xffff)) {
330 case 0413: /* ZMAGIC */
331 file_offset = 1024;
332 break;
333 case 0314: /* QMAGIC */
334 file_offset = 0;
335 break;
336 default:
337 error = ENOEXEC;
338 goto cleanup;
339 }
340
341 bss_size = round_page(a_out->a_bss);
342
343 /* Check various fields in header for validity/bounds. */
344 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
345 error = ENOEXEC;
346 goto cleanup;
347 }
348
349 /* text + data can't exceed file size */
350 if (a_out->a_data + a_out->a_text > attr.va_size) {
351 error = EFAULT;
352 goto cleanup;
353 }
354
355 /*
356 * text/data/bss must not exceed limits
357 * XXX - this is not complete. it should check current usage PLUS
358 * the resources needed by this library.
359 */
360 PROC_LOCK(td->td_proc);
361 if (a_out->a_text > maxtsiz ||
362 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA)) {
363 PROC_UNLOCK(td->td_proc);
364 error = ENOMEM;
365 goto cleanup;
366 }
367 PROC_UNLOCK(td->td_proc);
368
369 /*
370 * Prevent more writers.
371 * XXX: Note that if any of the VM operations fail below we don't
372 * clear this flag.
373 */
374 vp->v_vflag |= VV_TEXT;
375
376 /*
377 * Lock no longer needed
378 */
379 locked = 0;
380 VOP_UNLOCK(vp, 0, td);
381 VFS_UNLOCK_GIANT(vfslocked);
382
383 /*
384 * Check if file_offset page aligned. Currently we cannot handle
385 * misalinged file offsets, and so we read in the entire image
386 * (what a waste).
387 */
388 if (file_offset & PAGE_MASK) {
389#ifdef DEBUG
390 printf("uselib: Non page aligned binary %lu\n", file_offset);
391#endif
392 /* Map text+data read/write/execute */
393
394 /* a_entry is the load address and is page aligned */
395 vmaddr = trunc_page(a_out->a_entry);
396
397 /* get anon user mapping, read+write+execute */
398 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
399 &vmaddr, a_out->a_text + a_out->a_data, FALSE, VM_PROT_ALL,
400 VM_PROT_ALL, 0);
401 if (error)
402 goto cleanup;
403
404 /* map file into kernel_map */
405 error = vm_mmap(kernel_map, &buffer,
406 round_page(a_out->a_text + a_out->a_data + file_offset),
407 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp,
408 trunc_page(file_offset));
409 if (error)
410 goto cleanup;
411
412 /* copy from kernel VM space to user space */
413 error = copyout(PTRIN(buffer + file_offset),
414 (void *)vmaddr, a_out->a_text + a_out->a_data);
415
416 /* release temporary kernel space */
417 vm_map_remove(kernel_map, buffer, buffer +
418 round_page(a_out->a_text + a_out->a_data + file_offset));
419
420 if (error)
421 goto cleanup;
422 } else {
423#ifdef DEBUG
424 printf("uselib: Page aligned binary %lu\n", file_offset);
425#endif
426 /*
427 * for QMAGIC, a_entry is 20 bytes beyond the load address
428 * to skip the executable header
429 */
430 vmaddr = trunc_page(a_out->a_entry);
431
432 /*
433 * Map it all into the process's space as a single
434 * copy-on-write "data" segment.
435 */
436 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
437 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
438 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
439 if (error)
440 goto cleanup;
441 }
442#ifdef DEBUG
443 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
444 ((long *)vmaddr)[1]);
445#endif
446 if (bss_size != 0) {
447 /* Calculate BSS start address */
448 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
449 a_out->a_data;
450
451 /* allocate some 'anon' space */
452 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
453 &vmaddr, bss_size, FALSE, VM_PROT_ALL, VM_PROT_ALL, 0);
454 if (error)
455 goto cleanup;
456 }
457
458cleanup:
459 /* Unlock vnode if needed */
460 if (locked) {
461 VOP_UNLOCK(vp, 0, td);
462 VFS_UNLOCK_GIANT(vfslocked);
463 }
464
465 /* Release the kernel mapping. */
466 if (a_out)
467 vm_map_remove(kernel_map, (vm_offset_t)a_out,
468 (vm_offset_t)a_out + PAGE_SIZE);
469
470 return error;
471}
472
473#endif /* __i386__ */
474
475int
476linux_select(struct thread *td, struct linux_select_args *args)
477{
478 l_timeval ltv;
479 struct timeval tv0, tv1, utv, *tvp;
480 int error;
481
482#ifdef DEBUG
483 if (ldebug(select))
484 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
485 (void *)args->readfds, (void *)args->writefds,
486 (void *)args->exceptfds, (void *)args->timeout);
487#endif
488
489 /*
490 * Store current time for computation of the amount of
491 * time left.
492 */
493 if (args->timeout) {
494 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
495 goto select_out;
496 utv.tv_sec = ltv.tv_sec;
497 utv.tv_usec = ltv.tv_usec;
498#ifdef DEBUG
499 if (ldebug(select))
500 printf(LMSG("incoming timeout (%jd/%ld)"),
501 (intmax_t)utv.tv_sec, utv.tv_usec);
502#endif
503
504 if (itimerfix(&utv)) {
505 /*
506 * The timeval was invalid. Convert it to something
507 * valid that will act as it does under Linux.
508 */
509 utv.tv_sec += utv.tv_usec / 1000000;
510 utv.tv_usec %= 1000000;
511 if (utv.tv_usec < 0) {
512 utv.tv_sec -= 1;
513 utv.tv_usec += 1000000;
514 }
515 if (utv.tv_sec < 0)
516 timevalclear(&utv);
517 }
518 microtime(&tv0);
519 tvp = &utv;
520 } else
521 tvp = NULL;
522
523 error = kern_select(td, args->nfds, args->readfds, args->writefds,
524 args->exceptfds, tvp);
525
526#ifdef DEBUG
527 if (ldebug(select))
528 printf(LMSG("real select returns %d"), error);
529#endif
530 if (error) {
531 /*
532 * See fs/select.c in the Linux kernel. Without this,
533 * Maelstrom doesn't work.
534 */
535 if (error == ERESTART)
536 error = EINTR;
537 goto select_out;
538 }
539
540 if (args->timeout) {
541 if (td->td_retval[0]) {
542 /*
543 * Compute how much time was left of the timeout,
544 * by subtracting the current time and the time
545 * before we started the call, and subtracting
546 * that result from the user-supplied value.
547 */
548 microtime(&tv1);
549 timevalsub(&tv1, &tv0);
550 timevalsub(&utv, &tv1);
551 if (utv.tv_sec < 0)
552 timevalclear(&utv);
553 } else
554 timevalclear(&utv);
555#ifdef DEBUG
556 if (ldebug(select))
557 printf(LMSG("outgoing timeout (%jd/%ld)"),
558 (intmax_t)utv.tv_sec, utv.tv_usec);
559#endif
560 ltv.tv_sec = utv.tv_sec;
561 ltv.tv_usec = utv.tv_usec;
562 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
563 goto select_out;
564 }
565
566select_out:
567#ifdef DEBUG
568 if (ldebug(select))
569 printf(LMSG("select_out -> %d"), error);
570#endif
571 return error;
572}
573
574int
575linux_mremap(struct thread *td, struct linux_mremap_args *args)
576{
577 struct munmap_args /* {
578 void *addr;
579 size_t len;
580 } */ bsd_args;
581 int error = 0;
582
583#ifdef DEBUG
584 if (ldebug(mremap))
585 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
586 (void *)(uintptr_t)args->addr,
587 (unsigned long)args->old_len,
588 (unsigned long)args->new_len,
589 (unsigned long)args->flags);
590#endif
591 args->new_len = round_page(args->new_len);
592 args->old_len = round_page(args->old_len);
593
594 if (args->new_len > args->old_len) {
595 td->td_retval[0] = 0;
596 return ENOMEM;
597 }
598
599 if (args->new_len < args->old_len) {
600 bsd_args.addr =
601 (caddr_t)((uintptr_t)args->addr + args->new_len);
602 bsd_args.len = args->old_len - args->new_len;
603 error = munmap(td, &bsd_args);
604 }
605
606 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
607 return error;
608}
609
610#define LINUX_MS_ASYNC 0x0001
611#define LINUX_MS_INVALIDATE 0x0002
612#define LINUX_MS_SYNC 0x0004
613
614int
615linux_msync(struct thread *td, struct linux_msync_args *args)
616{
617 struct msync_args bsd_args;
618
619 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
620 bsd_args.len = (uintptr_t)args->len;
621 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
622
623 return msync(td, &bsd_args);
624}
625
626int
627linux_time(struct thread *td, struct linux_time_args *args)
628{
629 struct timeval tv;
630 l_time_t tm;
631 int error;
632
633#ifdef DEBUG
634 if (ldebug(time))
635 printf(ARGS(time, "*"));
636#endif
637
638 microtime(&tv);
639 tm = tv.tv_sec;
640 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
641 return error;
642 td->td_retval[0] = tm;
643 return 0;
644}
645
646struct l_times_argv {
647 l_long tms_utime;
648 l_long tms_stime;
649 l_long tms_cutime;
650 l_long tms_cstime;
651};
652
653#define CLK_TCK 100 /* Linux uses 100 */
654
655#define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
656
657int
658linux_times(struct thread *td, struct linux_times_args *args)
659{
660 struct timeval tv, utime, stime, cutime, cstime;
661 struct l_times_argv tms;
662 struct proc *p;
663 int error;
664
665#ifdef DEBUG
666 if (ldebug(times))
667 printf(ARGS(times, "*"));
668#endif
669
670 if (args->buf != NULL) {
671 p = td->td_proc;
672 PROC_LOCK(p);
673 calcru(p, &utime, &stime);
674 calccru(p, &cutime, &cstime);
675 PROC_UNLOCK(p);
676
677 tms.tms_utime = CONVTCK(utime);
678 tms.tms_stime = CONVTCK(stime);
679
680 tms.tms_cutime = CONVTCK(cutime);
681 tms.tms_cstime = CONVTCK(cstime);
682
683 if ((error = copyout(&tms, args->buf, sizeof(tms))))
684 return error;
685 }
686
687 microuptime(&tv);
688 td->td_retval[0] = (int)CONVTCK(tv);
689 return 0;
690}
691
692int
693linux_newuname(struct thread *td, struct linux_newuname_args *args)
694{
695 struct l_new_utsname utsname;
696 char osname[LINUX_MAX_UTSNAME];
697 char osrelease[LINUX_MAX_UTSNAME];
698 char *p;
699
700#ifdef DEBUG
701 if (ldebug(newuname))
702 printf(ARGS(newuname, "*"));
703#endif
704
705 linux_get_osname(td, osname);
706 linux_get_osrelease(td, osrelease);
707
708 bzero(&utsname, sizeof(utsname));
709 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
710 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
711 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
712 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
713 for (p = utsname.version; *p != '\0'; ++p)
714 if (*p == '\n') {
715 *p = '\0';
716 break;
717 }
718#ifdef __i386__
719 {
720 const char *class;
721
722 switch (cpu_class) {
723 case CPUCLASS_686:
724 class = "i686";
725 break;
726 case CPUCLASS_586:
727 class = "i586";
728 break;
729 case CPUCLASS_486:
730 class = "i486";
731 break;
732 default:
733 class = "i386";
734 }
735 strlcpy(utsname.machine, class, LINUX_MAX_UTSNAME);
736 }
737#elif defined(__amd64__) /* XXX: Linux can change 'personality'. */
738#ifdef COMPAT_LINUX32
739 strlcpy(utsname.machine, "i686", LINUX_MAX_UTSNAME);
740#else
741 strlcpy(utsname.machine, "x86_64", LINUX_MAX_UTSNAME);
742#endif /* COMPAT_LINUX32 */
743#else /* something other than i386 or amd64 - assume we and Linux agree */
744 strlcpy(utsname.machine, machine, LINUX_MAX_UTSNAME);
745#endif /* __i386__ */
746 strlcpy(utsname.domainname, domainname, LINUX_MAX_UTSNAME);
747
748 return (copyout(&utsname, args->buf, sizeof(utsname)));
749}
750
751#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
752struct l_utimbuf {
753 l_time_t l_actime;
754 l_time_t l_modtime;
755};
756
757int
758linux_utime(struct thread *td, struct linux_utime_args *args)
759{
760 struct timeval tv[2], *tvp;
761 struct l_utimbuf lut;
762 char *fname;
763 int error;
764
765 LCONVPATHEXIST(td, args->fname, &fname);
766
767#ifdef DEBUG
768 if (ldebug(utime))
769 printf(ARGS(utime, "%s, *"), fname);
770#endif
771
772 if (args->times) {
773 if ((error = copyin(args->times, &lut, sizeof lut))) {
774 LFREEPATH(fname);
775 return error;
776 }
777 tv[0].tv_sec = lut.l_actime;
778 tv[0].tv_usec = 0;
779 tv[1].tv_sec = lut.l_modtime;
780 tv[1].tv_usec = 0;
781 tvp = tv;
782 } else
783 tvp = NULL;
784
785 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
786 LFREEPATH(fname);
787 return (error);
788}
789
790int
791linux_utimes(struct thread *td, struct linux_utimes_args *args)
792{
793 l_timeval ltv[2];
794 struct timeval tv[2], *tvp = NULL;
795 char *fname;
796 int error;
797
798 LCONVPATHEXIST(td, args->fname, &fname);
799
800#ifdef DEBUG
801 if (ldebug(utimes))
802 printf(ARGS(utimes, "%s, *"), fname);
803#endif
804
805 if (args->tptr != NULL) {
806 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
807 LFREEPATH(fname);
808 return (error);
809 }
810 tv[0].tv_sec = ltv[0].tv_sec;
811 tv[0].tv_usec = ltv[0].tv_usec;
812 tv[1].tv_sec = ltv[1].tv_sec;
813 tv[1].tv_usec = ltv[1].tv_usec;
814 tvp = tv;
815 }
816
817 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
818 LFREEPATH(fname);
819 return (error);
820}
821#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
822
823#define __WCLONE 0x80000000
824
825int
826linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
827{
828 int error, options, tmpstat;
829
830#ifdef DEBUG
831 if (ldebug(waitpid))
832 printf(ARGS(waitpid, "%d, %p, %d"),
833 args->pid, (void *)args->status, args->options);
834#endif
835 /*
836 * this is necessary because the test in kern_wait doesnt work
837 * because we mess with the options here
838 */
839 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE))
840 return (EINVAL);
841
842 options = (args->options & (WNOHANG | WUNTRACED));
843 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
844 if (args->options & __WCLONE)
845 options |= WLINUXCLONE;
846
847 error = kern_wait(td, args->pid, &tmpstat, options, NULL);
848 if (error)
849 return error;
850
851 if (args->status) {
852 tmpstat &= 0xffff;
853 if (WIFSIGNALED(tmpstat))
854 tmpstat = (tmpstat & 0xffffff80) |
855 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
856 else if (WIFSTOPPED(tmpstat))
857 tmpstat = (tmpstat & 0xffff00ff) |
858 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
859 return copyout(&tmpstat, args->status, sizeof(int));
860 }
861
862 return 0;
863}
864
865int
866linux_wait4(struct thread *td, struct linux_wait4_args *args)
867{
868 int error, options, tmpstat;
869 struct rusage ru, *rup;
870 struct proc *p;
871
872#ifdef DEBUG
873 if (ldebug(wait4))
874 printf(ARGS(wait4, "%d, %p, %d, %p"),
875 args->pid, (void *)args->status, args->options,
876 (void *)args->rusage);
877#endif
878
879 options = (args->options & (WNOHANG | WUNTRACED));
880 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
881 if (args->options & __WCLONE)
882 options |= WLINUXCLONE;
883
884 if (args->rusage != NULL)
885 rup = &ru;
886 else
887 rup = NULL;
888 error = kern_wait(td, args->pid, &tmpstat, options, rup);
889 if (error)
890 return error;
891
892 p = td->td_proc;
893 PROC_LOCK(p);
894 sigqueue_delete(&p->p_sigqueue, SIGCHLD);
895 PROC_UNLOCK(p);
896
897 if (args->status) {
898 tmpstat &= 0xffff;
899 if (WIFSIGNALED(tmpstat))
900 tmpstat = (tmpstat & 0xffffff80) |
901 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat));
902 else if (WIFSTOPPED(tmpstat))
903 tmpstat = (tmpstat & 0xffff00ff) |
904 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8);
905 error = copyout(&tmpstat, args->status, sizeof(int));
906 }
907 if (args->rusage != NULL && error == 0)
908 error = copyout(&ru, args->rusage, sizeof(ru));
909
910 return (error);
911}
912
913int
914linux_mknod(struct thread *td, struct linux_mknod_args *args)
915{
916 char *path;
917 int error;
918
919 LCONVPATHCREAT(td, args->path, &path);
920
921#ifdef DEBUG
922 if (ldebug(mknod))
923 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev);
924#endif
925
926 switch (args->mode & S_IFMT) {
927 case S_IFIFO:
928 case S_IFSOCK:
929 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
930 break;
931
932 case S_IFCHR:
933 case S_IFBLK:
934 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
935 args->dev);
936 break;
937
938 case S_IFDIR:
939 error = EPERM;
940 break;
941
942 case 0:
943 args->mode |= S_IFREG;
944 /* FALLTHROUGH */
945 case S_IFREG:
946 error = kern_open(td, path, UIO_SYSSPACE,
947 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
948 break;
949
950 default:
951 error = EINVAL;
952 break;
953 }
954 LFREEPATH(path);
955 return (error);
956}
957
958/*
959 * UGH! This is just about the dumbest idea I've ever heard!!
960 */
961int
962linux_personality(struct thread *td, struct linux_personality_args *args)
963{
964#ifdef DEBUG
965 if (ldebug(personality))
966 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
967#endif
968 if (args->per != 0)
969 return EINVAL;
970
971 /* Yes Jim, it's still a Linux... */
972 td->td_retval[0] = 0;
973 return 0;
974}
975
976struct l_itimerval {
977 l_timeval it_interval;
978 l_timeval it_value;
979};
980
981#define B2L_ITIMERVAL(bip, lip) \
982 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
983 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
984 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
985 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
986
987int
988linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
989{
990 int error;
991 struct l_itimerval ls;
992 struct itimerval aitv, oitv;
993
994#ifdef DEBUG
995 if (ldebug(setitimer))
996 printf(ARGS(setitimer, "%p, %p"),
997 (void *)uap->itv, (void *)uap->oitv);
998#endif
999
1000 if (uap->itv == NULL) {
1001 uap->itv = uap->oitv;
1002 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1003 }
1004
1005 error = copyin(uap->itv, &ls, sizeof(ls));
1006 if (error != 0)
1007 return (error);
1008 B2L_ITIMERVAL(&aitv, &ls);
1009#ifdef DEBUG
1010 if (ldebug(setitimer)) {
1011 printf("setitimer: value: sec: %jd, usec: %ld\n",
1012 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1013 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1014 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1015 }
1016#endif
1017 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1018 if (error != 0 || uap->oitv == NULL)
1019 return (error);
1020 B2L_ITIMERVAL(&ls, &oitv);
1021
1022 return (copyout(&ls, uap->oitv, sizeof(ls)));
1023}
1024
1025int
1026linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1027{
1028 int error;
1029 struct l_itimerval ls;
1030 struct itimerval aitv;
1031
1032#ifdef DEBUG
1033 if (ldebug(getitimer))
1034 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1035#endif
1036 error = kern_getitimer(td, uap->which, &aitv);
1037 if (error != 0)
1038 return (error);
1039 B2L_ITIMERVAL(&ls, &aitv);
1040 return (copyout(&ls, uap->itv, sizeof(ls)));
1041}
1042
1043int
1044linux_nice(struct thread *td, struct linux_nice_args *args)
1045{
1046 struct setpriority_args bsd_args;
1047
1048 bsd_args.which = PRIO_PROCESS;
1049 bsd_args.who = 0; /* current process */
1050 bsd_args.prio = args->inc;
1051 return setpriority(td, &bsd_args);
1052}
1053
1054int
1055linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1056{
1057 struct ucred *newcred, *oldcred;
1058 l_gid_t linux_gidset[NGROUPS];
1059 gid_t *bsd_gidset;
1060 int ngrp, error;
1061 struct proc *p;
1062
1063 ngrp = args->gidsetsize;
1064 if (ngrp < 0 || ngrp >= NGROUPS)
1065 return (EINVAL);
1066 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1067 if (error)
1068 return (error);
1069 newcred = crget();
1070 p = td->td_proc;
1071 PROC_LOCK(p);
1072 oldcred = p->p_ucred;
1073
1074 /*
1075 * cr_groups[0] holds egid. Setting the whole set from
1076 * the supplied set will cause egid to be changed too.
1077 * Keep cr_groups[0] unchanged to prevent that.
1078 */
1079
1080 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS,
1081 SUSER_ALLOWJAIL)) != 0) {
1082 PROC_UNLOCK(p);
1083 crfree(newcred);
1084 return (error);
1085 }
1086
1087 crcopy(newcred, oldcred);
1088 if (ngrp > 0) {
1089 newcred->cr_ngroups = ngrp + 1;
1090
1091 bsd_gidset = newcred->cr_groups;
1092 ngrp--;
1093 while (ngrp >= 0) {
1094 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1095 ngrp--;
1096 }
1097 } else
1098 newcred->cr_ngroups = 1;
1099
1100 setsugid(p);
1101 p->p_ucred = newcred;
1102 PROC_UNLOCK(p);
1103 crfree(oldcred);
1104 return (0);
1105}
1106
1107int
1108linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1109{
1110 struct ucred *cred;
1111 l_gid_t linux_gidset[NGROUPS];
1112 gid_t *bsd_gidset;
1113 int bsd_gidsetsz, ngrp, error;
1114
1115 cred = td->td_ucred;
1116 bsd_gidset = cred->cr_groups;
1117 bsd_gidsetsz = cred->cr_ngroups - 1;
1118
1119 /*
1120 * cr_groups[0] holds egid. Returning the whole set
1121 * here will cause a duplicate. Exclude cr_groups[0]
1122 * to prevent that.
1123 */
1124
1125 if ((ngrp = args->gidsetsize) == 0) {
1126 td->td_retval[0] = bsd_gidsetsz;
1127 return (0);
1128 }
1129
1130 if (ngrp < bsd_gidsetsz)
1131 return (EINVAL);
1132
1133 ngrp = 0;
1134 while (ngrp < bsd_gidsetsz) {
1135 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1136 ngrp++;
1137 }
1138
1139 if ((error = copyout(linux_gidset, args->grouplist,
1140 ngrp * sizeof(l_gid_t))))
1141 return (error);
1142
1143 td->td_retval[0] = ngrp;
1144 return (0);
1145}
1146
1147int
1148linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1149{
1150 struct rlimit bsd_rlim;
1151 struct l_rlimit rlim;
1152 u_int which;
1153 int error;
1154
1155#ifdef DEBUG
1156 if (ldebug(setrlimit))
1157 printf(ARGS(setrlimit, "%d, %p"),
1158 args->resource, (void *)args->rlim);
1159#endif
1160
1161 if (args->resource >= LINUX_RLIM_NLIMITS)
1162 return (EINVAL);
1163
1164 which = linux_to_bsd_resource[args->resource];
1165 if (which == -1)
1166 return (EINVAL);
1167
1168 error = copyin(args->rlim, &rlim, sizeof(rlim));
1169 if (error)
1170 return (error);
1171
1172 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1173 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1174 return (kern_setrlimit(td, which, &bsd_rlim));
1175}
1176
1177int
1178linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1179{
1180 struct l_rlimit rlim;
1181 struct proc *p = td->td_proc;
1182 struct rlimit bsd_rlim;
1183 u_int which;
1184
1185#ifdef DEBUG
1186 if (ldebug(old_getrlimit))
1187 printf(ARGS(old_getrlimit, "%d, %p"),
1188 args->resource, (void *)args->rlim);
1189#endif
1190
1191 if (args->resource >= LINUX_RLIM_NLIMITS)
1192 return (EINVAL);
1193
1194 which = linux_to_bsd_resource[args->resource];
1195 if (which == -1)
1196 return (EINVAL);
1197
1198 PROC_LOCK(p);
1199 lim_rlimit(p, which, &bsd_rlim);
1200 PROC_UNLOCK(p);
1201
1202#ifdef COMPAT_LINUX32
1203 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1204 if (rlim.rlim_cur == UINT_MAX)
1205 rlim.rlim_cur = INT_MAX;
1206 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1207 if (rlim.rlim_max == UINT_MAX)
1208 rlim.rlim_max = INT_MAX;
1209#else
1210 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1211 if (rlim.rlim_cur == ULONG_MAX)
1212 rlim.rlim_cur = LONG_MAX;
1213 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1214 if (rlim.rlim_max == ULONG_MAX)
1215 rlim.rlim_max = LONG_MAX;
1216#endif
1217 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1218}
1219
1220int
1221linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1222{
1223 struct l_rlimit rlim;
1224 struct proc *p = td->td_proc;
1225 struct rlimit bsd_rlim;
1226 u_int which;
1227
1228#ifdef DEBUG
1229 if (ldebug(getrlimit))
1230 printf(ARGS(getrlimit, "%d, %p"),
1231 args->resource, (void *)args->rlim);
1232#endif
1233
1234 if (args->resource >= LINUX_RLIM_NLIMITS)
1235 return (EINVAL);
1236
1237 which = linux_to_bsd_resource[args->resource];
1238 if (which == -1)
1239 return (EINVAL);
1240
1241 PROC_LOCK(p);
1242 lim_rlimit(p, which, &bsd_rlim);
1243 PROC_UNLOCK(p);
1244
1245 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1246 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1247 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1248}
1249
1250int
1251linux_sched_setscheduler(struct thread *td,
1252 struct linux_sched_setscheduler_args *args)
1253{
1254 struct sched_setscheduler_args bsd;
1255
1256#ifdef DEBUG
1257 if (ldebug(sched_setscheduler))
1258 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1259 args->pid, args->policy, (const void *)args->param);
1260#endif
1261
1262 switch (args->policy) {
1263 case LINUX_SCHED_OTHER:
1264 bsd.policy = SCHED_OTHER;
1265 break;
1266 case LINUX_SCHED_FIFO:
1267 bsd.policy = SCHED_FIFO;
1268 break;
1269 case LINUX_SCHED_RR:
1270 bsd.policy = SCHED_RR;
1271 break;
1272 default:
1273 return EINVAL;
1274 }
1275
1276 bsd.pid = args->pid;
1277 bsd.param = (struct sched_param *)args->param;
1278 return sched_setscheduler(td, &bsd);
1279}
1280
1281int
1282linux_sched_getscheduler(struct thread *td,
1283 struct linux_sched_getscheduler_args *args)
1284{
1285 struct sched_getscheduler_args bsd;
1286 int error;
1287
1288#ifdef DEBUG
1289 if (ldebug(sched_getscheduler))
1290 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1291#endif
1292
1293 bsd.pid = args->pid;
1294 error = sched_getscheduler(td, &bsd);
1295
1296 switch (td->td_retval[0]) {
1297 case SCHED_OTHER:
1298 td->td_retval[0] = LINUX_SCHED_OTHER;
1299 break;
1300 case SCHED_FIFO:
1301 td->td_retval[0] = LINUX_SCHED_FIFO;
1302 break;
1303 case SCHED_RR:
1304 td->td_retval[0] = LINUX_SCHED_RR;
1305 break;
1306 }
1307
1308 return error;
1309}
1310
1311int
1312linux_sched_get_priority_max(struct thread *td,
1313 struct linux_sched_get_priority_max_args *args)
1314{
1315 struct sched_get_priority_max_args bsd;
1316
1317#ifdef DEBUG
1318 if (ldebug(sched_get_priority_max))
1319 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1320#endif
1321
1322 switch (args->policy) {
1323 case LINUX_SCHED_OTHER:
1324 bsd.policy = SCHED_OTHER;
1325 break;
1326 case LINUX_SCHED_FIFO:
1327 bsd.policy = SCHED_FIFO;
1328 break;
1329 case LINUX_SCHED_RR:
1330 bsd.policy = SCHED_RR;
1331 break;
1332 default:
1333 return EINVAL;
1334 }
1335 return sched_get_priority_max(td, &bsd);
1336}
1337
1338int
1339linux_sched_get_priority_min(struct thread *td,
1340 struct linux_sched_get_priority_min_args *args)
1341{
1342 struct sched_get_priority_min_args bsd;
1343
1344#ifdef DEBUG
1345 if (ldebug(sched_get_priority_min))
1346 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1347#endif
1348
1349 switch (args->policy) {
1350 case LINUX_SCHED_OTHER:
1351 bsd.policy = SCHED_OTHER;
1352 break;
1353 case LINUX_SCHED_FIFO:
1354 bsd.policy = SCHED_FIFO;
1355 break;
1356 case LINUX_SCHED_RR:
1357 bsd.policy = SCHED_RR;
1358 break;
1359 default:
1360 return EINVAL;
1361 }
1362 return sched_get_priority_min(td, &bsd);
1363}
1364
1365#define REBOOT_CAD_ON 0x89abcdef
1366#define REBOOT_CAD_OFF 0
1367#define REBOOT_HALT 0xcdef0123
1368#define REBOOT_RESTART 0x01234567
1369#define REBOOT_RESTART2 0xA1B2C3D4
1370#define REBOOT_POWEROFF 0x4321FEDC
1371#define REBOOT_MAGIC1 0xfee1dead
1372#define REBOOT_MAGIC2 0x28121969
1373#define REBOOT_MAGIC2A 0x05121996
1374#define REBOOT_MAGIC2B 0x16041998
1375
1376int
1377linux_reboot(struct thread *td, struct linux_reboot_args *args)
1378{
1379 struct reboot_args bsd_args;
1380
1381#ifdef DEBUG
1382 if (ldebug(reboot))
1383 printf(ARGS(reboot, "0x%x"), args->cmd);
1384#endif
1385
1386 if (args->magic1 != REBOOT_MAGIC1)
1387 return EINVAL;
1388
1389 switch (args->magic2) {
1390 case REBOOT_MAGIC2:
1391 case REBOOT_MAGIC2A:
1392 case REBOOT_MAGIC2B:
1393 break;
1394 default:
1395 return EINVAL;
1396 }
1397
1398 switch (args->cmd) {
1399 case REBOOT_CAD_ON:
1400 case REBOOT_CAD_OFF:
1401 return (priv_check(td, PRIV_REBOOT));
1402 case REBOOT_HALT:
1403 bsd_args.opt = RB_HALT;
1404 break;
1405 case REBOOT_RESTART:
1406 case REBOOT_RESTART2:
1407 bsd_args.opt = 0;
1408 break;
1409 case REBOOT_POWEROFF:
1410 bsd_args.opt = RB_POWEROFF;
1411 break;
1412 default:
1413 return EINVAL;
1414 }
1415 return reboot(td, &bsd_args);
1416}
1417
1418
1419/*
1420 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1421 * td->td_retval[1] when COMPAT_43 is defined. This
1422 * globbers registers that are assumed to be preserved. The following
1423 * lightweight syscalls fixes this. See also linux_getgid16() and
1424 * linux_getuid16() in linux_uid16.c.
1425 *
1426 * linux_getpid() - MP SAFE
1427 * linux_getgid() - MP SAFE
1428 * linux_getuid() - MP SAFE
1429 */
1430
1431int
1432linux_getpid(struct thread *td, struct linux_getpid_args *args)
1433{
1434 struct linux_emuldata *em;
1435
1436#ifdef DEBUG
1437 if (ldebug(getpid))
1438 printf(ARGS(getpid, ""));
1439#endif
1440
1441 if (linux_use26(td)) {
|