32
33#include "opt_compat.h"
34#include "opt_kdtrace.h"
35
36#include <sys/param.h>
37#include <sys/blist.h>
38#include <sys/fcntl.h>
39#if defined(__i386__)
40#include <sys/imgact_aout.h>
41#endif
42#include <sys/jail.h>
43#include <sys/kernel.h>
44#include <sys/limits.h>
45#include <sys/lock.h>
46#include <sys/malloc.h>
47#include <sys/mman.h>
48#include <sys/mount.h>
49#include <sys/mutex.h>
50#include <sys/namei.h>
51#include <sys/priv.h>
52#include <sys/proc.h>
53#include <sys/reboot.h>
54#include <sys/racct.h>
55#include <sys/resourcevar.h>
56#include <sys/sched.h>
57#include <sys/sdt.h>
58#include <sys/signalvar.h>
59#include <sys/stat.h>
60#include <sys/syscallsubr.h>
61#include <sys/sysctl.h>
62#include <sys/sysproto.h>
63#include <sys/systm.h>
64#include <sys/time.h>
65#include <sys/vmmeter.h>
66#include <sys/vnode.h>
67#include <sys/wait.h>
68#include <sys/cpuset.h>
69
70#include <security/mac/mac_framework.h>
71
72#include <vm/vm.h>
73#include <vm/pmap.h>
74#include <vm/vm_kern.h>
75#include <vm/vm_map.h>
76#include <vm/vm_extern.h>
77#include <vm/vm_object.h>
78#include <vm/swap_pager.h>
79
80#ifdef COMPAT_LINUX32
81#include <machine/../linux32/linux.h>
82#include <machine/../linux32/linux32_proto.h>
83#else
84#include <machine/../linux/linux.h>
85#include <machine/../linux/linux_proto.h>
86#endif
87
88#include <compat/linux/linux_dtrace.h>
89#include <compat/linux/linux_file.h>
90#include <compat/linux/linux_mib.h>
91#include <compat/linux/linux_signal.h>
92#include <compat/linux/linux_timer.h>
93#include <compat/linux/linux_util.h>
94#include <compat/linux/linux_sysproto.h>
95#include <compat/linux/linux_emul.h>
96#include <compat/linux/linux_misc.h>
97
98/**
99 * Special DTrace provider for the linuxulator.
100 *
101 * In this file we define the provider for the entire linuxulator. All
102 * modules (= files of the linuxulator) use it.
103 *
104 * We define a different name depending on the emulated bitsize, see
105 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
106 * native bitsize = linuxulator
107 * amd64, 32bit emulation = linuxulator32
108 */
109LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
110
111int stclohz; /* Statistics clock frequency */
112
113static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
114 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
115 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
116 RLIMIT_MEMLOCK, RLIMIT_AS
117};
118
119struct l_sysinfo {
120 l_long uptime; /* Seconds since boot */
121 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
122#define LINUX_SYSINFO_LOADS_SCALE 65536
123 l_ulong totalram; /* Total usable main memory size */
124 l_ulong freeram; /* Available memory size */
125 l_ulong sharedram; /* Amount of shared memory */
126 l_ulong bufferram; /* Memory used by buffers */
127 l_ulong totalswap; /* Total swap space size */
128 l_ulong freeswap; /* swap space still available */
129 l_ushort procs; /* Number of current processes */
130 l_ushort pads;
131 l_ulong totalbig;
132 l_ulong freebig;
133 l_uint mem_unit;
134 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
135};
136
137struct l_pselect6arg {
138 l_uintptr_t ss;
139 l_size_t ss_len;
140};
141
142static int linux_utimensat_nsec_valid(l_long);
143
144
145int
146linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
147{
148 struct l_sysinfo sysinfo;
149 vm_object_t object;
150 int i, j;
151 struct timespec ts;
152
153 getnanouptime(&ts);
154 if (ts.tv_nsec != 0)
155 ts.tv_sec++;
156 sysinfo.uptime = ts.tv_sec;
157
158 /* Use the information from the mib to get our load averages */
159 for (i = 0; i < 3; i++)
160 sysinfo.loads[i] = averunnable.ldavg[i] *
161 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
162
163 sysinfo.totalram = physmem * PAGE_SIZE;
164 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
165
166 sysinfo.sharedram = 0;
167 mtx_lock(&vm_object_list_mtx);
168 TAILQ_FOREACH(object, &vm_object_list, object_list)
169 if (object->shadow_count > 1)
170 sysinfo.sharedram += object->resident_page_count;
171 mtx_unlock(&vm_object_list_mtx);
172
173 sysinfo.sharedram *= PAGE_SIZE;
174 sysinfo.bufferram = 0;
175
176 swap_pager_status(&i, &j);
177 sysinfo.totalswap = i * PAGE_SIZE;
178 sysinfo.freeswap = (i - j) * PAGE_SIZE;
179
180 sysinfo.procs = nprocs;
181
182 /* The following are only present in newer Linux kernels. */
183 sysinfo.totalbig = 0;
184 sysinfo.freebig = 0;
185 sysinfo.mem_unit = 1;
186
187 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
188}
189
190int
191linux_alarm(struct thread *td, struct linux_alarm_args *args)
192{
193 struct itimerval it, old_it;
194 u_int secs;
195 int error;
196
197#ifdef DEBUG
198 if (ldebug(alarm))
199 printf(ARGS(alarm, "%u"), args->secs);
200#endif
201
202 secs = args->secs;
203
204 if (secs > INT_MAX)
205 secs = INT_MAX;
206
207 it.it_value.tv_sec = (long) secs;
208 it.it_value.tv_usec = 0;
209 it.it_interval.tv_sec = 0;
210 it.it_interval.tv_usec = 0;
211 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
212 if (error)
213 return (error);
214 if (timevalisset(&old_it.it_value)) {
215 if (old_it.it_value.tv_usec != 0)
216 old_it.it_value.tv_sec++;
217 td->td_retval[0] = old_it.it_value.tv_sec;
218 }
219 return (0);
220}
221
222int
223linux_brk(struct thread *td, struct linux_brk_args *args)
224{
225 struct vmspace *vm = td->td_proc->p_vmspace;
226 vm_offset_t new, old;
227 struct obreak_args /* {
228 char * nsize;
229 } */ tmp;
230
231#ifdef DEBUG
232 if (ldebug(brk))
233 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
234#endif
235 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
236 new = (vm_offset_t)args->dsend;
237 tmp.nsize = (char *)new;
238 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
239 td->td_retval[0] = (long)new;
240 else
241 td->td_retval[0] = (long)old;
242
243 return (0);
244}
245
246#if defined(__i386__)
247/* XXX: what about amd64/linux32? */
248
249int
250linux_uselib(struct thread *td, struct linux_uselib_args *args)
251{
252 struct nameidata ni;
253 struct vnode *vp;
254 struct exec *a_out;
255 struct vattr attr;
256 vm_offset_t vmaddr;
257 unsigned long file_offset;
258 unsigned long bss_size;
259 char *library;
260 ssize_t aresid;
261 int error, locked, writecount;
262
263 LCONVPATHEXIST(td, args->library, &library);
264
265#ifdef DEBUG
266 if (ldebug(uselib))
267 printf(ARGS(uselib, "%s"), library);
268#endif
269
270 a_out = NULL;
271 locked = 0;
272 vp = NULL;
273
274 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
275 UIO_SYSSPACE, library, td);
276 error = namei(&ni);
277 LFREEPATH(library);
278 if (error)
279 goto cleanup;
280
281 vp = ni.ni_vp;
282 NDFREE(&ni, NDF_ONLY_PNBUF);
283
284 /*
285 * From here on down, we have a locked vnode that must be unlocked.
286 * XXX: The code below largely duplicates exec_check_permissions().
287 */
288 locked = 1;
289
290 /* Writable? */
291 error = VOP_GET_WRITECOUNT(vp, &writecount);
292 if (error != 0)
293 goto cleanup;
294 if (writecount != 0) {
295 error = ETXTBSY;
296 goto cleanup;
297 }
298
299 /* Executable? */
300 error = VOP_GETATTR(vp, &attr, td->td_ucred);
301 if (error)
302 goto cleanup;
303
304 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
305 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
306 /* EACCESS is what exec(2) returns. */
307 error = ENOEXEC;
308 goto cleanup;
309 }
310
311 /* Sensible size? */
312 if (attr.va_size == 0) {
313 error = ENOEXEC;
314 goto cleanup;
315 }
316
317 /* Can we access it? */
318 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
319 if (error)
320 goto cleanup;
321
322 /*
323 * XXX: This should use vn_open() so that it is properly authorized,
324 * and to reduce code redundancy all over the place here.
325 * XXX: Not really, it duplicates far more of exec_check_permissions()
326 * than vn_open().
327 */
328#ifdef MAC
329 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
330 if (error)
331 goto cleanup;
332#endif
333 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
334 if (error)
335 goto cleanup;
336
337 /* Pull in executable header into exec_map */
338 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
339 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
340 if (error)
341 goto cleanup;
342
343 /* Is it a Linux binary ? */
344 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
345 error = ENOEXEC;
346 goto cleanup;
347 }
348
349 /*
350 * While we are here, we should REALLY do some more checks
351 */
352
353 /* Set file/virtual offset based on a.out variant. */
354 switch ((int)(a_out->a_magic & 0xffff)) {
355 case 0413: /* ZMAGIC */
356 file_offset = 1024;
357 break;
358 case 0314: /* QMAGIC */
359 file_offset = 0;
360 break;
361 default:
362 error = ENOEXEC;
363 goto cleanup;
364 }
365
366 bss_size = round_page(a_out->a_bss);
367
368 /* Check various fields in header for validity/bounds. */
369 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
370 error = ENOEXEC;
371 goto cleanup;
372 }
373
374 /* text + data can't exceed file size */
375 if (a_out->a_data + a_out->a_text > attr.va_size) {
376 error = EFAULT;
377 goto cleanup;
378 }
379
380 /*
381 * text/data/bss must not exceed limits
382 * XXX - this is not complete. it should check current usage PLUS
383 * the resources needed by this library.
384 */
385 PROC_LOCK(td->td_proc);
386 if (a_out->a_text > maxtsiz ||
387 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
388 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
389 bss_size) != 0) {
390 PROC_UNLOCK(td->td_proc);
391 error = ENOMEM;
392 goto cleanup;
393 }
394 PROC_UNLOCK(td->td_proc);
395
396 /*
397 * Prevent more writers.
398 * XXX: Note that if any of the VM operations fail below we don't
399 * clear this flag.
400 */
401 VOP_SET_TEXT(vp);
402
403 /*
404 * Lock no longer needed
405 */
406 locked = 0;
407 VOP_UNLOCK(vp, 0);
408
409 /*
410 * Check if file_offset page aligned. Currently we cannot handle
411 * misalinged file offsets, and so we read in the entire image
412 * (what a waste).
413 */
414 if (file_offset & PAGE_MASK) {
415#ifdef DEBUG
416 printf("uselib: Non page aligned binary %lu\n", file_offset);
417#endif
418 /* Map text+data read/write/execute */
419
420 /* a_entry is the load address and is page aligned */
421 vmaddr = trunc_page(a_out->a_entry);
422
423 /* get anon user mapping, read+write+execute */
424 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
425 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
426 VM_PROT_ALL, VM_PROT_ALL, 0);
427 if (error)
428 goto cleanup;
429
430 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
431 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
432 td->td_ucred, NOCRED, &aresid, td);
433 if (error != 0)
434 goto cleanup;
435 if (aresid != 0) {
436 error = ENOEXEC;
437 goto cleanup;
438 }
439 } else {
440#ifdef DEBUG
441 printf("uselib: Page aligned binary %lu\n", file_offset);
442#endif
443 /*
444 * for QMAGIC, a_entry is 20 bytes beyond the load address
445 * to skip the executable header
446 */
447 vmaddr = trunc_page(a_out->a_entry);
448
449 /*
450 * Map it all into the process's space as a single
451 * copy-on-write "data" segment.
452 */
453 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
454 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
455 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
456 if (error)
457 goto cleanup;
458 }
459#ifdef DEBUG
460 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
461 ((long *)vmaddr)[1]);
462#endif
463 if (bss_size != 0) {
464 /* Calculate BSS start address */
465 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
466 a_out->a_data;
467
468 /* allocate some 'anon' space */
469 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
470 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
471 VM_PROT_ALL, 0);
472 if (error)
473 goto cleanup;
474 }
475
476cleanup:
477 /* Unlock vnode if needed */
478 if (locked)
479 VOP_UNLOCK(vp, 0);
480
481 /* Release the temporary mapping. */
482 if (a_out)
483 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
484
485 return (error);
486}
487
488#endif /* __i386__ */
489
490int
491linux_select(struct thread *td, struct linux_select_args *args)
492{
493 l_timeval ltv;
494 struct timeval tv0, tv1, utv, *tvp;
495 int error;
496
497#ifdef DEBUG
498 if (ldebug(select))
499 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
500 (void *)args->readfds, (void *)args->writefds,
501 (void *)args->exceptfds, (void *)args->timeout);
502#endif
503
504 /*
505 * Store current time for computation of the amount of
506 * time left.
507 */
508 if (args->timeout) {
509 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
510 goto select_out;
511 utv.tv_sec = ltv.tv_sec;
512 utv.tv_usec = ltv.tv_usec;
513#ifdef DEBUG
514 if (ldebug(select))
515 printf(LMSG("incoming timeout (%jd/%ld)"),
516 (intmax_t)utv.tv_sec, utv.tv_usec);
517#endif
518
519 if (itimerfix(&utv)) {
520 /*
521 * The timeval was invalid. Convert it to something
522 * valid that will act as it does under Linux.
523 */
524 utv.tv_sec += utv.tv_usec / 1000000;
525 utv.tv_usec %= 1000000;
526 if (utv.tv_usec < 0) {
527 utv.tv_sec -= 1;
528 utv.tv_usec += 1000000;
529 }
530 if (utv.tv_sec < 0)
531 timevalclear(&utv);
532 }
533 microtime(&tv0);
534 tvp = &utv;
535 } else
536 tvp = NULL;
537
538 error = kern_select(td, args->nfds, args->readfds, args->writefds,
539 args->exceptfds, tvp, LINUX_NFDBITS);
540
541#ifdef DEBUG
542 if (ldebug(select))
543 printf(LMSG("real select returns %d"), error);
544#endif
545 if (error)
546 goto select_out;
547
548 if (args->timeout) {
549 if (td->td_retval[0]) {
550 /*
551 * Compute how much time was left of the timeout,
552 * by subtracting the current time and the time
553 * before we started the call, and subtracting
554 * that result from the user-supplied value.
555 */
556 microtime(&tv1);
557 timevalsub(&tv1, &tv0);
558 timevalsub(&utv, &tv1);
559 if (utv.tv_sec < 0)
560 timevalclear(&utv);
561 } else
562 timevalclear(&utv);
563#ifdef DEBUG
564 if (ldebug(select))
565 printf(LMSG("outgoing timeout (%jd/%ld)"),
566 (intmax_t)utv.tv_sec, utv.tv_usec);
567#endif
568 ltv.tv_sec = utv.tv_sec;
569 ltv.tv_usec = utv.tv_usec;
570 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
571 goto select_out;
572 }
573
574select_out:
575#ifdef DEBUG
576 if (ldebug(select))
577 printf(LMSG("select_out -> %d"), error);
578#endif
579 return (error);
580}
581
582int
583linux_mremap(struct thread *td, struct linux_mremap_args *args)
584{
585 struct munmap_args /* {
586 void *addr;
587 size_t len;
588 } */ bsd_args;
589 int error = 0;
590
591#ifdef DEBUG
592 if (ldebug(mremap))
593 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
594 (void *)(uintptr_t)args->addr,
595 (unsigned long)args->old_len,
596 (unsigned long)args->new_len,
597 (unsigned long)args->flags);
598#endif
599
600 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
601 td->td_retval[0] = 0;
602 return (EINVAL);
603 }
604
605 /*
606 * Check for the page alignment.
607 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
608 */
609 if (args->addr & PAGE_MASK) {
610 td->td_retval[0] = 0;
611 return (EINVAL);
612 }
613
614 args->new_len = round_page(args->new_len);
615 args->old_len = round_page(args->old_len);
616
617 if (args->new_len > args->old_len) {
618 td->td_retval[0] = 0;
619 return (ENOMEM);
620 }
621
622 if (args->new_len < args->old_len) {
623 bsd_args.addr =
624 (caddr_t)((uintptr_t)args->addr + args->new_len);
625 bsd_args.len = args->old_len - args->new_len;
626 error = sys_munmap(td, &bsd_args);
627 }
628
629 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
630 return (error);
631}
632
633#define LINUX_MS_ASYNC 0x0001
634#define LINUX_MS_INVALIDATE 0x0002
635#define LINUX_MS_SYNC 0x0004
636
637int
638linux_msync(struct thread *td, struct linux_msync_args *args)
639{
640 struct msync_args bsd_args;
641
642 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
643 bsd_args.len = (uintptr_t)args->len;
644 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
645
646 return (sys_msync(td, &bsd_args));
647}
648
649int
650linux_time(struct thread *td, struct linux_time_args *args)
651{
652 struct timeval tv;
653 l_time_t tm;
654 int error;
655
656#ifdef DEBUG
657 if (ldebug(time))
658 printf(ARGS(time, "*"));
659#endif
660
661 microtime(&tv);
662 tm = tv.tv_sec;
663 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
664 return (error);
665 td->td_retval[0] = tm;
666 return (0);
667}
668
669struct l_times_argv {
670 l_clock_t tms_utime;
671 l_clock_t tms_stime;
672 l_clock_t tms_cutime;
673 l_clock_t tms_cstime;
674};
675
676
677/*
678 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
679 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
680 * auxiliary vector entry.
681 */
682#define CLK_TCK 100
683
684#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
685#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
686
687#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
688 CONVNTCK(r) : CONVOTCK(r))
689
690int
691linux_times(struct thread *td, struct linux_times_args *args)
692{
693 struct timeval tv, utime, stime, cutime, cstime;
694 struct l_times_argv tms;
695 struct proc *p;
696 int error;
697
698#ifdef DEBUG
699 if (ldebug(times))
700 printf(ARGS(times, "*"));
701#endif
702
703 if (args->buf != NULL) {
704 p = td->td_proc;
705 PROC_LOCK(p);
706 PROC_STATLOCK(p);
707 calcru(p, &utime, &stime);
708 PROC_STATUNLOCK(p);
709 calccru(p, &cutime, &cstime);
710 PROC_UNLOCK(p);
711
712 tms.tms_utime = CONVTCK(utime);
713 tms.tms_stime = CONVTCK(stime);
714
715 tms.tms_cutime = CONVTCK(cutime);
716 tms.tms_cstime = CONVTCK(cstime);
717
718 if ((error = copyout(&tms, args->buf, sizeof(tms))))
719 return (error);
720 }
721
722 microuptime(&tv);
723 td->td_retval[0] = (int)CONVTCK(tv);
724 return (0);
725}
726
727int
728linux_newuname(struct thread *td, struct linux_newuname_args *args)
729{
730 struct l_new_utsname utsname;
731 char osname[LINUX_MAX_UTSNAME];
732 char osrelease[LINUX_MAX_UTSNAME];
733 char *p;
734
735#ifdef DEBUG
736 if (ldebug(newuname))
737 printf(ARGS(newuname, "*"));
738#endif
739
740 linux_get_osname(td, osname);
741 linux_get_osrelease(td, osrelease);
742
743 bzero(&utsname, sizeof(utsname));
744 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
745 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
746 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
747 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
748 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
749 for (p = utsname.version; *p != '\0'; ++p)
750 if (*p == '\n') {
751 *p = '\0';
752 break;
753 }
754 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
755
756 return (copyout(&utsname, args->buf, sizeof(utsname)));
757}
758
759struct l_utimbuf {
760 l_time_t l_actime;
761 l_time_t l_modtime;
762};
763
764int
765linux_utime(struct thread *td, struct linux_utime_args *args)
766{
767 struct timeval tv[2], *tvp;
768 struct l_utimbuf lut;
769 char *fname;
770 int error;
771
772 LCONVPATHEXIST(td, args->fname, &fname);
773
774#ifdef DEBUG
775 if (ldebug(utime))
776 printf(ARGS(utime, "%s, *"), fname);
777#endif
778
779 if (args->times) {
780 if ((error = copyin(args->times, &lut, sizeof lut))) {
781 LFREEPATH(fname);
782 return (error);
783 }
784 tv[0].tv_sec = lut.l_actime;
785 tv[0].tv_usec = 0;
786 tv[1].tv_sec = lut.l_modtime;
787 tv[1].tv_usec = 0;
788 tvp = tv;
789 } else
790 tvp = NULL;
791
792 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
793 LFREEPATH(fname);
794 return (error);
795}
796
797int
798linux_utimes(struct thread *td, struct linux_utimes_args *args)
799{
800 l_timeval ltv[2];
801 struct timeval tv[2], *tvp = NULL;
802 char *fname;
803 int error;
804
805 LCONVPATHEXIST(td, args->fname, &fname);
806
807#ifdef DEBUG
808 if (ldebug(utimes))
809 printf(ARGS(utimes, "%s, *"), fname);
810#endif
811
812 if (args->tptr != NULL) {
813 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
814 LFREEPATH(fname);
815 return (error);
816 }
817 tv[0].tv_sec = ltv[0].tv_sec;
818 tv[0].tv_usec = ltv[0].tv_usec;
819 tv[1].tv_sec = ltv[1].tv_sec;
820 tv[1].tv_usec = ltv[1].tv_usec;
821 tvp = tv;
822 }
823
824 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
825 LFREEPATH(fname);
826 return (error);
827}
828
829static int
830linux_utimensat_nsec_valid(l_long nsec)
831{
832
833 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
834 return (0);
835 if (nsec >= 0 && nsec <= 999999999)
836 return (0);
837 return (1);
838}
839
840int
841linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
842{
843 struct l_timespec l_times[2];
844 struct timespec times[2], *timesp = NULL;
845 char *path = NULL;
846 int error, dfd, flags = 0;
847
848 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
849
850#ifdef DEBUG
851 if (ldebug(utimensat))
852 printf(ARGS(utimensat, "%d, *"), dfd);
853#endif
854
855 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
856 return (EINVAL);
857
858 if (args->times != NULL) {
859 error = copyin(args->times, l_times, sizeof(l_times));
860 if (error != 0)
861 return (error);
862
863 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
864 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
865 return (EINVAL);
866
867 times[0].tv_sec = l_times[0].tv_sec;
868 switch (l_times[0].tv_nsec)
869 {
870 case LINUX_UTIME_OMIT:
871 times[0].tv_nsec = UTIME_OMIT;
872 break;
873 case LINUX_UTIME_NOW:
874 times[0].tv_nsec = UTIME_NOW;
875 break;
876 default:
877 times[0].tv_nsec = l_times[0].tv_nsec;
878 }
879
880 times[1].tv_sec = l_times[1].tv_sec;
881 switch (l_times[1].tv_nsec)
882 {
883 case LINUX_UTIME_OMIT:
884 times[1].tv_nsec = UTIME_OMIT;
885 break;
886 case LINUX_UTIME_NOW:
887 times[1].tv_nsec = UTIME_NOW;
888 break;
889 default:
890 times[1].tv_nsec = l_times[1].tv_nsec;
891 break;
892 }
893 timesp = times;
894 }
895
896 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT)
897 /* This breaks POSIX, but is what the Linux kernel does
898 * _on purpose_ (documented in the man page for utimensat(2)),
899 * so we must follow that behaviour. */
900 return (0);
901
902 if (args->pathname != NULL)
903 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
904 else if (args->flags != 0)
905 return (EINVAL);
906
907 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
908 flags |= AT_SYMLINK_NOFOLLOW;
909
910 if (path == NULL)
911 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
912 else {
913 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
914 UIO_SYSSPACE, flags);
915 LFREEPATH(path);
916 }
917
918 return (error);
919}
920
921int
922linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
923{
924 l_timeval ltv[2];
925 struct timeval tv[2], *tvp = NULL;
926 char *fname;
927 int error, dfd;
928
929 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
930 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
931
932#ifdef DEBUG
933 if (ldebug(futimesat))
934 printf(ARGS(futimesat, "%s, *"), fname);
935#endif
936
937 if (args->utimes != NULL) {
938 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
939 LFREEPATH(fname);
940 return (error);
941 }
942 tv[0].tv_sec = ltv[0].tv_sec;
943 tv[0].tv_usec = ltv[0].tv_usec;
944 tv[1].tv_sec = ltv[1].tv_sec;
945 tv[1].tv_usec = ltv[1].tv_usec;
946 tvp = tv;
947 }
948
949 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
950 LFREEPATH(fname);
951 return (error);
952}
953
954int
955linux_common_wait(struct thread *td, int pid, int *status,
956 int options, struct rusage *ru)
957{
958 int error, tmpstat;
959
960 error = kern_wait(td, pid, &tmpstat, options, ru);
961 if (error)
962 return (error);
963
964 if (status) {
965 tmpstat &= 0xffff;
966 if (WIFSIGNALED(tmpstat))
967 tmpstat = (tmpstat & 0xffffff80) |
968 bsd_to_linux_signal(WTERMSIG(tmpstat));
969 else if (WIFSTOPPED(tmpstat))
970 tmpstat = (tmpstat & 0xffff00ff) |
971 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
972 else if (WIFCONTINUED(tmpstat))
973 tmpstat = 0xffff;
974 error = copyout(&tmpstat, status, sizeof(int));
975 }
976
977 return (error);
978}
979
980#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
981int
982linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
983{
984 struct linux_wait4_args wait4_args;
985
986#ifdef DEBUG
987 if (ldebug(waitpid))
988 printf(ARGS(waitpid, "%d, %p, %d"),
989 args->pid, (void *)args->status, args->options);
990#endif
991
992 wait4_args.pid = args->pid;
993 wait4_args.status = args->status;
994 wait4_args.options = args->options;
995 wait4_args.rusage = NULL;
996
997 return (linux_wait4(td, &wait4_args));
998}
999#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1000
1001int
1002linux_wait4(struct thread *td, struct linux_wait4_args *args)
1003{
1004 int error, options;
1005 struct rusage ru, *rup;
1006
1007#ifdef DEBUG
1008 if (ldebug(wait4))
1009 printf(ARGS(wait4, "%d, %p, %d, %p"),
1010 args->pid, (void *)args->status, args->options,
1011 (void *)args->rusage);
1012#endif
1013 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1014 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1015 return (EINVAL);
1016
1017 options = WEXITED;
1018 linux_to_bsd_waitopts(args->options, &options);
1019
1020 if (args->rusage != NULL)
1021 rup = &ru;
1022 else
1023 rup = NULL;
1024 error = linux_common_wait(td, args->pid, args->status, options, rup);
1025 if (error != 0)
1026 return (error);
1027 if (args->rusage != NULL)
1028 error = linux_copyout_rusage(&ru, args->rusage);
1029 return (error);
1030}
1031
1032int
1033linux_waitid(struct thread *td, struct linux_waitid_args *args)
1034{
1035 int status, options, sig;
1036 struct __wrusage wru;
1037 siginfo_t siginfo;
1038 l_siginfo_t lsi;
1039 idtype_t idtype;
1040 struct proc *p;
1041 int error;
1042
1043 options = 0;
1044 linux_to_bsd_waitopts(args->options, &options);
1045
1046 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1047 return (EINVAL);
1048 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1049 return (EINVAL);
1050
1051 switch (args->idtype) {
1052 case LINUX_P_ALL:
1053 idtype = P_ALL;
1054 break;
1055 case LINUX_P_PID:
1056 if (args->id <= 0)
1057 return (EINVAL);
1058 idtype = P_PID;
1059 break;
1060 case LINUX_P_PGID:
1061 if (args->id <= 0)
1062 return (EINVAL);
1063 idtype = P_PGID;
1064 break;
1065 default:
1066 return (EINVAL);
1067 }
1068
1069 error = kern_wait6(td, idtype, args->id, &status, options,
1070 &wru, &siginfo);
1071 if (error != 0)
1072 return (error);
1073 if (args->rusage != NULL) {
1074 error = linux_copyout_rusage(&wru.wru_children,
1075 args->rusage);
1076 if (error != 0)
1077 return (error);
1078 }
1079 if (args->info != NULL) {
1080 p = td->td_proc;
1081 if (td->td_retval[0] == 0)
1082 bzero(&lsi, sizeof(lsi));
1083 else {
1084 sig = bsd_to_linux_signal(siginfo.si_signo);
1085 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1086 }
1087 error = copyout(&lsi, args->info, sizeof(lsi));
1088 }
1089 td->td_retval[0] = 0;
1090
1091 return (error);
1092}
1093
1094int
1095linux_mknod(struct thread *td, struct linux_mknod_args *args)
1096{
1097 char *path;
1098 int error;
1099
1100 LCONVPATHCREAT(td, args->path, &path);
1101
1102#ifdef DEBUG
1103 if (ldebug(mknod))
1104 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1105 (uintmax_t)args->dev);
1106#endif
1107
1108 switch (args->mode & S_IFMT) {
1109 case S_IFIFO:
1110 case S_IFSOCK:
1111 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
1112 break;
1113
1114 case S_IFCHR:
1115 case S_IFBLK:
1116 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
1117 args->dev);
1118 break;
1119
1120 case S_IFDIR:
1121 error = EPERM;
1122 break;
1123
1124 case 0:
1125 args->mode |= S_IFREG;
1126 /* FALLTHROUGH */
1127 case S_IFREG:
1128 error = kern_open(td, path, UIO_SYSSPACE,
1129 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1130 if (error == 0)
1131 kern_close(td, td->td_retval[0]);
1132 break;
1133
1134 default:
1135 error = EINVAL;
1136 break;
1137 }
1138 LFREEPATH(path);
1139 return (error);
1140}
1141
1142int
1143linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1144{
1145 char *path;
1146 int error, dfd;
1147
1148 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1149 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1150
1151#ifdef DEBUG
1152 if (ldebug(mknodat))
1153 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1154#endif
1155
1156 switch (args->mode & S_IFMT) {
1157 case S_IFIFO:
1158 case S_IFSOCK:
1159 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1160 break;
1161
1162 case S_IFCHR:
1163 case S_IFBLK:
1164 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1165 args->dev);
1166 break;
1167
1168 case S_IFDIR:
1169 error = EPERM;
1170 break;
1171
1172 case 0:
1173 args->mode |= S_IFREG;
1174 /* FALLTHROUGH */
1175 case S_IFREG:
1176 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1177 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1178 if (error == 0)
1179 kern_close(td, td->td_retval[0]);
1180 break;
1181
1182 default:
1183 error = EINVAL;
1184 break;
1185 }
1186 LFREEPATH(path);
1187 return (error);
1188}
1189
1190/*
1191 * UGH! This is just about the dumbest idea I've ever heard!!
1192 */
1193int
1194linux_personality(struct thread *td, struct linux_personality_args *args)
1195{
1196#ifdef DEBUG
1197 if (ldebug(personality))
1198 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1199#endif
1200 if (args->per != 0)
1201 return (EINVAL);
1202
1203 /* Yes Jim, it's still a Linux... */
1204 td->td_retval[0] = 0;
1205 return (0);
1206}
1207
1208struct l_itimerval {
1209 l_timeval it_interval;
1210 l_timeval it_value;
1211};
1212
1213#define B2L_ITIMERVAL(bip, lip) \
1214 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1215 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1216 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1217 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1218
1219int
1220linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1221{
1222 int error;
1223 struct l_itimerval ls;
1224 struct itimerval aitv, oitv;
1225
1226#ifdef DEBUG
1227 if (ldebug(setitimer))
1228 printf(ARGS(setitimer, "%p, %p"),
1229 (void *)uap->itv, (void *)uap->oitv);
1230#endif
1231
1232 if (uap->itv == NULL) {
1233 uap->itv = uap->oitv;
1234 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1235 }
1236
1237 error = copyin(uap->itv, &ls, sizeof(ls));
1238 if (error != 0)
1239 return (error);
1240 B2L_ITIMERVAL(&aitv, &ls);
1241#ifdef DEBUG
1242 if (ldebug(setitimer)) {
1243 printf("setitimer: value: sec: %jd, usec: %ld\n",
1244 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1245 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1246 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1247 }
1248#endif
1249 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1250 if (error != 0 || uap->oitv == NULL)
1251 return (error);
1252 B2L_ITIMERVAL(&ls, &oitv);
1253
1254 return (copyout(&ls, uap->oitv, sizeof(ls)));
1255}
1256
1257int
1258linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1259{
1260 int error;
1261 struct l_itimerval ls;
1262 struct itimerval aitv;
1263
1264#ifdef DEBUG
1265 if (ldebug(getitimer))
1266 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1267#endif
1268 error = kern_getitimer(td, uap->which, &aitv);
1269 if (error != 0)
1270 return (error);
1271 B2L_ITIMERVAL(&ls, &aitv);
1272 return (copyout(&ls, uap->itv, sizeof(ls)));
1273}
1274
1275#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1276int
1277linux_nice(struct thread *td, struct linux_nice_args *args)
1278{
1279 struct setpriority_args bsd_args;
1280
1281 bsd_args.which = PRIO_PROCESS;
1282 bsd_args.who = 0; /* current process */
1283 bsd_args.prio = args->inc;
1284 return (sys_setpriority(td, &bsd_args));
1285}
1286#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1287
1288int
1289linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1290{
1291 struct ucred *newcred, *oldcred;
1292 l_gid_t *linux_gidset;
1293 gid_t *bsd_gidset;
1294 int ngrp, error;
1295 struct proc *p;
1296
1297 ngrp = args->gidsetsize;
1298 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1299 return (EINVAL);
1300 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1301 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1302 if (error)
1303 goto out;
1304 newcred = crget();
1305 crextend(newcred, ngrp + 1);
1306 p = td->td_proc;
1307 PROC_LOCK(p);
1308 oldcred = p->p_ucred;
1309 crcopy(newcred, oldcred);
1310
1311 /*
1312 * cr_groups[0] holds egid. Setting the whole set from
1313 * the supplied set will cause egid to be changed too.
1314 * Keep cr_groups[0] unchanged to prevent that.
1315 */
1316
1317 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1318 PROC_UNLOCK(p);
1319 crfree(newcred);
1320 goto out;
1321 }
1322
1323 if (ngrp > 0) {
1324 newcred->cr_ngroups = ngrp + 1;
1325
1326 bsd_gidset = newcred->cr_groups;
1327 ngrp--;
1328 while (ngrp >= 0) {
1329 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1330 ngrp--;
1331 }
1332 } else
1333 newcred->cr_ngroups = 1;
1334
1335 setsugid(p);
1336 p->p_ucred = newcred;
1337 PROC_UNLOCK(p);
1338 crfree(oldcred);
1339 error = 0;
1340out:
1341 free(linux_gidset, M_LINUX);
1342 return (error);
1343}
1344
1345int
1346linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1347{
1348 struct ucred *cred;
1349 l_gid_t *linux_gidset;
1350 gid_t *bsd_gidset;
1351 int bsd_gidsetsz, ngrp, error;
1352
1353 cred = td->td_ucred;
1354 bsd_gidset = cred->cr_groups;
1355 bsd_gidsetsz = cred->cr_ngroups - 1;
1356
1357 /*
1358 * cr_groups[0] holds egid. Returning the whole set
1359 * here will cause a duplicate. Exclude cr_groups[0]
1360 * to prevent that.
1361 */
1362
1363 if ((ngrp = args->gidsetsize) == 0) {
1364 td->td_retval[0] = bsd_gidsetsz;
1365 return (0);
1366 }
1367
1368 if (ngrp < bsd_gidsetsz)
1369 return (EINVAL);
1370
1371 ngrp = 0;
1372 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1373 M_LINUX, M_WAITOK);
1374 while (ngrp < bsd_gidsetsz) {
1375 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1376 ngrp++;
1377 }
1378
1379 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1380 free(linux_gidset, M_LINUX);
1381 if (error)
1382 return (error);
1383
1384 td->td_retval[0] = ngrp;
1385 return (0);
1386}
1387
1388int
1389linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1390{
1391 struct rlimit bsd_rlim;
1392 struct l_rlimit rlim;
1393 u_int which;
1394 int error;
1395
1396#ifdef DEBUG
1397 if (ldebug(setrlimit))
1398 printf(ARGS(setrlimit, "%d, %p"),
1399 args->resource, (void *)args->rlim);
1400#endif
1401
1402 if (args->resource >= LINUX_RLIM_NLIMITS)
1403 return (EINVAL);
1404
1405 which = linux_to_bsd_resource[args->resource];
1406 if (which == -1)
1407 return (EINVAL);
1408
1409 error = copyin(args->rlim, &rlim, sizeof(rlim));
1410 if (error)
1411 return (error);
1412
1413 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1414 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1415 return (kern_setrlimit(td, which, &bsd_rlim));
1416}
1417
1418#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1419int
1420linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1421{
1422 struct l_rlimit rlim;
1423 struct proc *p = td->td_proc;
1424 struct rlimit bsd_rlim;
1425 u_int which;
1426
1427#ifdef DEBUG
1428 if (ldebug(old_getrlimit))
1429 printf(ARGS(old_getrlimit, "%d, %p"),
1430 args->resource, (void *)args->rlim);
1431#endif
1432
1433 if (args->resource >= LINUX_RLIM_NLIMITS)
1434 return (EINVAL);
1435
1436 which = linux_to_bsd_resource[args->resource];
1437 if (which == -1)
1438 return (EINVAL);
1439
1440 PROC_LOCK(p);
1441 lim_rlimit(p, which, &bsd_rlim);
1442 PROC_UNLOCK(p);
1443
1444#ifdef COMPAT_LINUX32
1445 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1446 if (rlim.rlim_cur == UINT_MAX)
1447 rlim.rlim_cur = INT_MAX;
1448 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1449 if (rlim.rlim_max == UINT_MAX)
1450 rlim.rlim_max = INT_MAX;
1451#else
1452 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1453 if (rlim.rlim_cur == ULONG_MAX)
1454 rlim.rlim_cur = LONG_MAX;
1455 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1456 if (rlim.rlim_max == ULONG_MAX)
1457 rlim.rlim_max = LONG_MAX;
1458#endif
1459 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1460}
1461#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1462
1463int
1464linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1465{
1466 struct l_rlimit rlim;
1467 struct proc *p = td->td_proc;
1468 struct rlimit bsd_rlim;
1469 u_int which;
1470
1471#ifdef DEBUG
1472 if (ldebug(getrlimit))
1473 printf(ARGS(getrlimit, "%d, %p"),
1474 args->resource, (void *)args->rlim);
1475#endif
1476
1477 if (args->resource >= LINUX_RLIM_NLIMITS)
1478 return (EINVAL);
1479
1480 which = linux_to_bsd_resource[args->resource];
1481 if (which == -1)
1482 return (EINVAL);
1483
1484 PROC_LOCK(p);
1485 lim_rlimit(p, which, &bsd_rlim);
1486 PROC_UNLOCK(p);
1487
1488 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1489 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1490 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1491}
1492
1493int
1494linux_sched_setscheduler(struct thread *td,
1495 struct linux_sched_setscheduler_args *args)
1496{
1497 struct sched_param sched_param;
1498 struct thread *tdt;
1499 int error, policy;
1500
1501#ifdef DEBUG
1502 if (ldebug(sched_setscheduler))
1503 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1504 args->pid, args->policy, (const void *)args->param);
1505#endif
1506
1507 switch (args->policy) {
1508 case LINUX_SCHED_OTHER:
1509 policy = SCHED_OTHER;
1510 break;
1511 case LINUX_SCHED_FIFO:
1512 policy = SCHED_FIFO;
1513 break;
1514 case LINUX_SCHED_RR:
1515 policy = SCHED_RR;
1516 break;
1517 default:
1518 return (EINVAL);
1519 }
1520
1521 error = copyin(args->param, &sched_param, sizeof(sched_param));
1522 if (error)
1523 return (error);
1524
1525 tdt = linux_tdfind(td, args->pid, -1);
1526 if (tdt == NULL)
1527 return (ESRCH);
1528
1529 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1530 PROC_UNLOCK(tdt->td_proc);
1531 return (error);
1532}
1533
1534int
1535linux_sched_getscheduler(struct thread *td,
1536 struct linux_sched_getscheduler_args *args)
1537{
1538 struct thread *tdt;
1539 int error, policy;
1540
1541#ifdef DEBUG
1542 if (ldebug(sched_getscheduler))
1543 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1544#endif
1545
1546 tdt = linux_tdfind(td, args->pid, -1);
1547 if (tdt == NULL)
1548 return (ESRCH);
1549
1550 error = kern_sched_getscheduler(td, tdt, &policy);
1551 PROC_UNLOCK(tdt->td_proc);
1552
1553 switch (policy) {
1554 case SCHED_OTHER:
1555 td->td_retval[0] = LINUX_SCHED_OTHER;
1556 break;
1557 case SCHED_FIFO:
1558 td->td_retval[0] = LINUX_SCHED_FIFO;
1559 break;
1560 case SCHED_RR:
1561 td->td_retval[0] = LINUX_SCHED_RR;
1562 break;
1563 }
1564 return (error);
1565}
1566
1567int
1568linux_sched_get_priority_max(struct thread *td,
1569 struct linux_sched_get_priority_max_args *args)
1570{
1571 struct sched_get_priority_max_args bsd;
1572
1573#ifdef DEBUG
1574 if (ldebug(sched_get_priority_max))
1575 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1576#endif
1577
1578 switch (args->policy) {
1579 case LINUX_SCHED_OTHER:
1580 bsd.policy = SCHED_OTHER;
1581 break;
1582 case LINUX_SCHED_FIFO:
1583 bsd.policy = SCHED_FIFO;
1584 break;
1585 case LINUX_SCHED_RR:
1586 bsd.policy = SCHED_RR;
1587 break;
1588 default:
1589 return (EINVAL);
1590 }
1591 return (sys_sched_get_priority_max(td, &bsd));
1592}
1593
1594int
1595linux_sched_get_priority_min(struct thread *td,
1596 struct linux_sched_get_priority_min_args *args)
1597{
1598 struct sched_get_priority_min_args bsd;
1599
1600#ifdef DEBUG
1601 if (ldebug(sched_get_priority_min))
1602 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1603#endif
1604
1605 switch (args->policy) {
1606 case LINUX_SCHED_OTHER:
1607 bsd.policy = SCHED_OTHER;
1608 break;
1609 case LINUX_SCHED_FIFO:
1610 bsd.policy = SCHED_FIFO;
1611 break;
1612 case LINUX_SCHED_RR:
1613 bsd.policy = SCHED_RR;
1614 break;
1615 default:
1616 return (EINVAL);
1617 }
1618 return (sys_sched_get_priority_min(td, &bsd));
1619}
1620
1621#define REBOOT_CAD_ON 0x89abcdef
1622#define REBOOT_CAD_OFF 0
1623#define REBOOT_HALT 0xcdef0123
1624#define REBOOT_RESTART 0x01234567
1625#define REBOOT_RESTART2 0xA1B2C3D4
1626#define REBOOT_POWEROFF 0x4321FEDC
1627#define REBOOT_MAGIC1 0xfee1dead
1628#define REBOOT_MAGIC2 0x28121969
1629#define REBOOT_MAGIC2A 0x05121996
1630#define REBOOT_MAGIC2B 0x16041998
1631
1632int
1633linux_reboot(struct thread *td, struct linux_reboot_args *args)
1634{
1635 struct reboot_args bsd_args;
1636
1637#ifdef DEBUG
1638 if (ldebug(reboot))
1639 printf(ARGS(reboot, "0x%x"), args->cmd);
1640#endif
1641
1642 if (args->magic1 != REBOOT_MAGIC1)
1643 return (EINVAL);
1644
1645 switch (args->magic2) {
1646 case REBOOT_MAGIC2:
1647 case REBOOT_MAGIC2A:
1648 case REBOOT_MAGIC2B:
1649 break;
1650 default:
1651 return (EINVAL);
1652 }
1653
1654 switch (args->cmd) {
1655 case REBOOT_CAD_ON:
1656 case REBOOT_CAD_OFF:
1657 return (priv_check(td, PRIV_REBOOT));
1658 case REBOOT_HALT:
1659 bsd_args.opt = RB_HALT;
1660 break;
1661 case REBOOT_RESTART:
1662 case REBOOT_RESTART2:
1663 bsd_args.opt = 0;
1664 break;
1665 case REBOOT_POWEROFF:
1666 bsd_args.opt = RB_POWEROFF;
1667 break;
1668 default:
1669 return (EINVAL);
1670 }
1671 return (sys_reboot(td, &bsd_args));
1672}
1673
1674
1675/*
1676 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1677 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1678 * are assumed to be preserved. The following lightweight syscalls fixes
1679 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1680 *
1681 * linux_getpid() - MP SAFE
1682 * linux_getgid() - MP SAFE
1683 * linux_getuid() - MP SAFE
1684 */
1685
1686int
1687linux_getpid(struct thread *td, struct linux_getpid_args *args)
1688{
1689
1690#ifdef DEBUG
1691 if (ldebug(getpid))
1692 printf(ARGS(getpid, ""));
1693#endif
1694 td->td_retval[0] = td->td_proc->p_pid;
1695
1696 return (0);
1697}
1698
1699int
1700linux_gettid(struct thread *td, struct linux_gettid_args *args)
1701{
1702 struct linux_emuldata *em;
1703
1704#ifdef DEBUG
1705 if (ldebug(gettid))
1706 printf(ARGS(gettid, ""));
1707#endif
1708
1709 em = em_find(td);
1710 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1711
1712 td->td_retval[0] = em->em_tid;
1713
1714 return (0);
1715}
1716
1717
1718int
1719linux_getppid(struct thread *td, struct linux_getppid_args *args)
1720{
1721
1722#ifdef DEBUG
1723 if (ldebug(getppid))
1724 printf(ARGS(getppid, ""));
1725#endif
1726
1727 PROC_LOCK(td->td_proc);
1728 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1729 PROC_UNLOCK(td->td_proc);
1730 return (0);
1731}
1732
1733int
1734linux_getgid(struct thread *td, struct linux_getgid_args *args)
1735{
1736
1737#ifdef DEBUG
1738 if (ldebug(getgid))
1739 printf(ARGS(getgid, ""));
1740#endif
1741
1742 td->td_retval[0] = td->td_ucred->cr_rgid;
1743 return (0);
1744}
1745
1746int
1747linux_getuid(struct thread *td, struct linux_getuid_args *args)
1748{
1749
1750#ifdef DEBUG
1751 if (ldebug(getuid))
1752 printf(ARGS(getuid, ""));
1753#endif
1754
1755 td->td_retval[0] = td->td_ucred->cr_ruid;
1756 return (0);
1757}
1758
1759
1760int
1761linux_getsid(struct thread *td, struct linux_getsid_args *args)
1762{
1763 struct getsid_args bsd;
1764
1765#ifdef DEBUG
1766 if (ldebug(getsid))
1767 printf(ARGS(getsid, "%i"), args->pid);
1768#endif
1769
1770 bsd.pid = args->pid;
1771 return (sys_getsid(td, &bsd));
1772}
1773
1774int
1775linux_nosys(struct thread *td, struct nosys_args *ignore)
1776{
1777
1778 return (ENOSYS);
1779}
1780
1781int
1782linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1783{
1784 struct getpriority_args bsd_args;
1785 int error;
1786
1787#ifdef DEBUG
1788 if (ldebug(getpriority))
1789 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1790#endif
1791
1792 bsd_args.which = args->which;
1793 bsd_args.who = args->who;
1794 error = sys_getpriority(td, &bsd_args);
1795 td->td_retval[0] = 20 - td->td_retval[0];
1796 return (error);
1797}
1798
1799int
1800linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1801{
1802 int name[2];
1803
1804#ifdef DEBUG
1805 if (ldebug(sethostname))
1806 printf(ARGS(sethostname, "*, %i"), args->len);
1807#endif
1808
1809 name[0] = CTL_KERN;
1810 name[1] = KERN_HOSTNAME;
1811 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1812 args->len, 0, 0));
1813}
1814
1815int
1816linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1817{
1818 int name[2];
1819
1820#ifdef DEBUG
1821 if (ldebug(setdomainname))
1822 printf(ARGS(setdomainname, "*, %i"), args->len);
1823#endif
1824
1825 name[0] = CTL_KERN;
1826 name[1] = KERN_NISDOMAINNAME;
1827 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1828 args->len, 0, 0));
1829}
1830
1831int
1832linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1833{
1834
1835#ifdef DEBUG
1836 if (ldebug(exit_group))
1837 printf(ARGS(exit_group, "%i"), args->error_code);
1838#endif
1839
1840 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1841 args->error_code);
1842
1843 /*
1844 * XXX: we should send a signal to the parent if
1845 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1846 * as it doesnt occur often.
1847 */
1848 exit1(td, W_EXITCODE(args->error_code, 0));
1849 /* NOTREACHED */
1850}
1851
1852#define _LINUX_CAPABILITY_VERSION 0x19980330
1853
1854struct l_user_cap_header {
1855 l_int version;
1856 l_int pid;
1857};
1858
1859struct l_user_cap_data {
1860 l_int effective;
1861 l_int permitted;
1862 l_int inheritable;
1863};
1864
1865int
1866linux_capget(struct thread *td, struct linux_capget_args *args)
1867{
1868 struct l_user_cap_header luch;
1869 struct l_user_cap_data lucd;
1870 int error;
1871
1872 if (args->hdrp == NULL)
1873 return (EFAULT);
1874
1875 error = copyin(args->hdrp, &luch, sizeof(luch));
1876 if (error != 0)
1877 return (error);
1878
1879 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1880 luch.version = _LINUX_CAPABILITY_VERSION;
1881 error = copyout(&luch, args->hdrp, sizeof(luch));
1882 if (error)
1883 return (error);
1884 return (EINVAL);
1885 }
1886
1887 if (luch.pid)
1888 return (EPERM);
1889
1890 if (args->datap) {
1891 /*
1892 * The current implementation doesn't support setting
1893 * a capability (it's essentially a stub) so indicate
1894 * that no capabilities are currently set or available
1895 * to request.
1896 */
1897 bzero (&lucd, sizeof(lucd));
1898 error = copyout(&lucd, args->datap, sizeof(lucd));
1899 }
1900
1901 return (error);
1902}
1903
1904int
1905linux_capset(struct thread *td, struct linux_capset_args *args)
1906{
1907 struct l_user_cap_header luch;
1908 struct l_user_cap_data lucd;
1909 int error;
1910
1911 if (args->hdrp == NULL || args->datap == NULL)
1912 return (EFAULT);
1913
1914 error = copyin(args->hdrp, &luch, sizeof(luch));
1915 if (error != 0)
1916 return (error);
1917
1918 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1919 luch.version = _LINUX_CAPABILITY_VERSION;
1920 error = copyout(&luch, args->hdrp, sizeof(luch));
1921 if (error)
1922 return (error);
1923 return (EINVAL);
1924 }
1925
1926 if (luch.pid)
1927 return (EPERM);
1928
1929 error = copyin(args->datap, &lucd, sizeof(lucd));
1930 if (error != 0)
1931 return (error);
1932
1933 /* We currently don't support setting any capabilities. */
1934 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1935 linux_msg(td,
1936 "capset effective=0x%x, permitted=0x%x, "
1937 "inheritable=0x%x is not implemented",
1938 (int)lucd.effective, (int)lucd.permitted,
1939 (int)lucd.inheritable);
1940 return (EPERM);
1941 }
1942
1943 return (0);
1944}
1945
1946int
1947linux_prctl(struct thread *td, struct linux_prctl_args *args)
1948{
1949 int error = 0, max_size;
1950 struct proc *p = td->td_proc;
1951 char comm[LINUX_MAX_COMM_LEN];
1952 struct linux_emuldata *em;
1953 int pdeath_signal;
1954
1955#ifdef DEBUG
1956 if (ldebug(prctl))
1957 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1958 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1959 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1960#endif
1961
1962 switch (args->option) {
1963 case LINUX_PR_SET_PDEATHSIG:
1964 if (!LINUX_SIG_VALID(args->arg2))
1965 return (EINVAL);
1966 em = em_find(td);
1967 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1968 em->pdeath_signal = args->arg2;
1969 break;
1970 case LINUX_PR_GET_PDEATHSIG:
1971 em = em_find(td);
1972 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1973 pdeath_signal = em->pdeath_signal;
1974 error = copyout(&pdeath_signal,
1975 (void *)(register_t)args->arg2,
1976 sizeof(pdeath_signal));
1977 break;
1978 case LINUX_PR_GET_KEEPCAPS:
1979 /*
1980 * Indicate that we always clear the effective and
1981 * permitted capability sets when the user id becomes
1982 * non-zero (actually the capability sets are simply
1983 * always zero in the current implementation).
1984 */
1985 td->td_retval[0] = 0;
1986 break;
1987 case LINUX_PR_SET_KEEPCAPS:
1988 /*
1989 * Ignore requests to keep the effective and permitted
1990 * capability sets when the user id becomes non-zero.
1991 */
1992 break;
1993 case LINUX_PR_SET_NAME:
1994 /*
1995 * To be on the safe side we need to make sure to not
1996 * overflow the size a linux program expects. We already
1997 * do this here in the copyin, so that we don't need to
1998 * check on copyout.
1999 */
2000 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2001 error = copyinstr((void *)(register_t)args->arg2, comm,
2002 max_size, NULL);
2003
2004 /* Linux silently truncates the name if it is too long. */
2005 if (error == ENAMETOOLONG) {
2006 /*
2007 * XXX: copyinstr() isn't documented to populate the
2008 * array completely, so do a copyin() to be on the
2009 * safe side. This should be changed in case
2010 * copyinstr() is changed to guarantee this.
2011 */
2012 error = copyin((void *)(register_t)args->arg2, comm,
2013 max_size - 1);
2014 comm[max_size - 1] = '\0';
2015 }
2016 if (error)
2017 return (error);
2018
2019 PROC_LOCK(p);
2020 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2021 PROC_UNLOCK(p);
2022 break;
2023 case LINUX_PR_GET_NAME:
2024 PROC_LOCK(p);
2025 strlcpy(comm, p->p_comm, sizeof(comm));
2026 PROC_UNLOCK(p);
2027 error = copyout(comm, (void *)(register_t)args->arg2,
2028 strlen(comm) + 1);
2029 break;
2030 default:
2031 error = EINVAL;
2032 break;
2033 }
2034
2035 return (error);
2036}
2037
2038int
2039linux_sched_setparam(struct thread *td,
2040 struct linux_sched_setparam_args *uap)
2041{
2042 struct sched_param sched_param;
2043 struct thread *tdt;
2044 int error;
2045
2046#ifdef DEBUG
2047 if (ldebug(sched_setparam))
2048 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2049#endif
2050
2051 error = copyin(uap->param, &sched_param, sizeof(sched_param));
2052 if (error)
2053 return (error);
2054
2055 tdt = linux_tdfind(td, uap->pid, -1);
2056 if (tdt == NULL)
2057 return (ESRCH);
2058
2059 error = kern_sched_setparam(td, tdt, &sched_param);
2060 PROC_UNLOCK(tdt->td_proc);
2061 return (error);
2062}
2063
2064int
2065linux_sched_getparam(struct thread *td,
2066 struct linux_sched_getparam_args *uap)
2067{
2068 struct sched_param sched_param;
2069 struct thread *tdt;
2070 int error;
2071
2072#ifdef DEBUG
2073 if (ldebug(sched_getparam))
2074 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2075#endif
2076
2077 tdt = linux_tdfind(td, uap->pid, -1);
2078 if (tdt == NULL)
2079 return (ESRCH);
2080
2081 error = kern_sched_getparam(td, tdt, &sched_param);
2082 PROC_UNLOCK(tdt->td_proc);
2083 if (error == 0)
2084 error = copyout(&sched_param, uap->param,
2085 sizeof(sched_param));
2086 return (error);
2087}
2088
2089/*
2090 * Get affinity of a process.
2091 */
2092int
2093linux_sched_getaffinity(struct thread *td,
2094 struct linux_sched_getaffinity_args *args)
2095{
2096 int error;
2097 struct thread *tdt;
2098 struct cpuset_getaffinity_args cga;
2099
2100#ifdef DEBUG
2101 if (ldebug(sched_getaffinity))
2102 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2103 args->len);
2104#endif
2105 if (args->len < sizeof(cpuset_t))
2106 return (EINVAL);
2107
2108 tdt = linux_tdfind(td, args->pid, -1);
2109 if (tdt == NULL)
2110 return (ESRCH);
2111
2112 PROC_UNLOCK(tdt->td_proc);
2113 cga.level = CPU_LEVEL_WHICH;
2114 cga.which = CPU_WHICH_TID;
2115 cga.id = tdt->td_tid;
2116 cga.cpusetsize = sizeof(cpuset_t);
2117 cga.mask = (cpuset_t *) args->user_mask_ptr;
2118
2119 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2120 td->td_retval[0] = sizeof(cpuset_t);
2121
2122 return (error);
2123}
2124
2125/*
2126 * Set affinity of a process.
2127 */
2128int
2129linux_sched_setaffinity(struct thread *td,
2130 struct linux_sched_setaffinity_args *args)
2131{
2132 struct cpuset_setaffinity_args csa;
2133 struct thread *tdt;
2134
2135#ifdef DEBUG
2136 if (ldebug(sched_setaffinity))
2137 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2138 args->len);
2139#endif
2140 if (args->len < sizeof(cpuset_t))
2141 return (EINVAL);
2142
2143 tdt = linux_tdfind(td, args->pid, -1);
2144 if (tdt == NULL)
2145 return (ESRCH);
2146
2147 PROC_UNLOCK(tdt->td_proc);
2148 csa.level = CPU_LEVEL_WHICH;
2149 csa.which = CPU_WHICH_TID;
2150 csa.id = tdt->td_tid;
2151 csa.cpusetsize = sizeof(cpuset_t);
2152 csa.mask = (cpuset_t *) args->user_mask_ptr;
2153
2154 return (sys_cpuset_setaffinity(td, &csa));
2155}
2156
2157struct linux_rlimit64 {
2158 uint64_t rlim_cur;
2159 uint64_t rlim_max;
2160};
2161
2162int
2163linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2164{
2165 struct rlimit rlim, nrlim;
2166 struct linux_rlimit64 lrlim;
2167 struct proc *p;
2168 u_int which;
2169 int flags;
2170 int error;
2171
2172#ifdef DEBUG
2173 if (ldebug(prlimit64))
2174 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2175 args->resource, (void *)args->new, (void *)args->old);
2176#endif
2177
2178 if (args->resource >= LINUX_RLIM_NLIMITS)
2179 return (EINVAL);
2180
2181 which = linux_to_bsd_resource[args->resource];
2182 if (which == -1)
2183 return (EINVAL);
2184
2185 if (args->new != NULL) {
2186 /*
2187 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2188 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2189 * as INFINITY so we do not need a conversion even.
2190 */
2191 error = copyin(args->new, &nrlim, sizeof(nrlim));
2192 if (error != 0)
2193 return (error);
2194 }
2195
2196 flags = PGET_HOLD | PGET_NOTWEXIT;
2197 if (args->new != NULL)
2198 flags |= PGET_CANDEBUG;
2199 else
2200 flags |= PGET_CANSEE;
2201 error = pget(args->pid, flags, &p);
2202 if (error != 0)
2203 return (error);
2204
2205 if (args->old != NULL) {
2206 PROC_LOCK(p);
2207 lim_rlimit(p, which, &rlim);
2208 PROC_UNLOCK(p);
2209 if (rlim.rlim_cur == RLIM_INFINITY)
2210 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2211 else
2212 lrlim.rlim_cur = rlim.rlim_cur;
2213 if (rlim.rlim_max == RLIM_INFINITY)
2214 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2215 else
2216 lrlim.rlim_max = rlim.rlim_max;
2217 error = copyout(&lrlim, args->old, sizeof(lrlim));
2218 if (error != 0)
2219 goto out;
2220 }
2221
2222 if (args->new != NULL)
2223 error = kern_proc_setrlimit(td, p, which, &nrlim);
2224
2225 out:
2226 PRELE(p);
2227 return (error);
2228}
2229
2230int
2231linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2232{
2233 struct timeval utv, tv0, tv1, *tvp;
2234 struct l_pselect6arg lpse6;
2235 struct l_timespec lts;
2236 struct timespec uts;
2237 l_sigset_t l_ss;
2238 sigset_t *ssp;
2239 sigset_t ss;
2240 int error;
2241
2242 ssp = NULL;
2243 if (args->sig != NULL) {
2244 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2245 if (error != 0)
2246 return (error);
2247 if (lpse6.ss_len != sizeof(l_ss))
2248 return (EINVAL);
2249 if (lpse6.ss != 0) {
2250 error = copyin(PTRIN(lpse6.ss), &l_ss,
2251 sizeof(l_ss));
2252 if (error != 0)
2253 return (error);
2254 linux_to_bsd_sigset(&l_ss, &ss);
2255 ssp = &ss;
2256 }
2257 }
2258
2259 /*
2260 * Currently glibc changes nanosecond number to microsecond.
2261 * This mean losing precision but for now it is hardly seen.
2262 */
2263 if (args->tsp != NULL) {
2264 error = copyin(args->tsp, <s, sizeof(lts));
2265 if (error != 0)
2266 return (error);
2267 error = linux_to_native_timespec(&uts, <s);
2268 if (error != 0)
2269 return (error);
2270
2271 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2272 if (itimerfix(&utv))
2273 return (EINVAL);
2274
2275 microtime(&tv0);
2276 tvp = &utv;
2277 } else
2278 tvp = NULL;
2279
2280 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2281 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2282
2283 if (error == 0 && args->tsp != NULL) {
2284 if (td->td_retval[0] != 0) {
2285 /*
2286 * Compute how much time was left of the timeout,
2287 * by subtracting the current time and the time
2288 * before we started the call, and subtracting
2289 * that result from the user-supplied value.
2290 */
2291
2292 microtime(&tv1);
2293 timevalsub(&tv1, &tv0);
2294 timevalsub(&utv, &tv1);
2295 if (utv.tv_sec < 0)
2296 timevalclear(&utv);
2297 } else
2298 timevalclear(&utv);
2299
2300 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2301
2302 native_to_linux_timespec(<s, &uts);
2303 error = copyout(<s, args->tsp, sizeof(lts));
2304 }
2305
2306 return (error);
2307}
2308
2309int
2310linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2311{
2312 struct timespec ts0, ts1;
2313 struct l_timespec lts;
2314 struct timespec uts, *tsp;
2315 l_sigset_t l_ss;
2316 sigset_t *ssp;
2317 sigset_t ss;
2318 int error;
2319
2320 if (args->sset != NULL) {
2321 if (args->ssize != sizeof(l_ss))
2322 return (EINVAL);
2323 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2324 if (error)
2325 return (error);
2326 linux_to_bsd_sigset(&l_ss, &ss);
2327 ssp = &ss;
2328 } else
2329 ssp = NULL;
2330 if (args->tsp != NULL) {
2331 error = copyin(args->tsp, <s, sizeof(lts));
2332 if (error)
2333 return (error);
2334 error = linux_to_native_timespec(&uts, <s);
2335 if (error != 0)
2336 return (error);
2337
2338 nanotime(&ts0);
2339 tsp = &uts;
2340 } else
2341 tsp = NULL;
2342
2343 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2344
2345 if (error == 0 && args->tsp != NULL) {
2346 if (td->td_retval[0]) {
2347 nanotime(&ts1);
2348 timespecsub(&ts1, &ts0);
2349 timespecsub(&uts, &ts1);
2350 if (uts.tv_sec < 0)
2351 timespecclear(&uts);
2352 } else
2353 timespecclear(&uts);
2354
2355 native_to_linux_timespec(<s, &uts);
2356 error = copyout(<s, args->tsp, sizeof(lts));
2357 }
2358
2359 return (error);
2360}
2361
2362#if defined(DEBUG) || defined(KTR)
2363/* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2364
| 32
33#include "opt_compat.h"
34#include "opt_kdtrace.h"
35
36#include <sys/param.h>
37#include <sys/blist.h>
38#include <sys/fcntl.h>
39#if defined(__i386__)
40#include <sys/imgact_aout.h>
41#endif
42#include <sys/jail.h>
43#include <sys/kernel.h>
44#include <sys/limits.h>
45#include <sys/lock.h>
46#include <sys/malloc.h>
47#include <sys/mman.h>
48#include <sys/mount.h>
49#include <sys/mutex.h>
50#include <sys/namei.h>
51#include <sys/priv.h>
52#include <sys/proc.h>
53#include <sys/reboot.h>
54#include <sys/racct.h>
55#include <sys/resourcevar.h>
56#include <sys/sched.h>
57#include <sys/sdt.h>
58#include <sys/signalvar.h>
59#include <sys/stat.h>
60#include <sys/syscallsubr.h>
61#include <sys/sysctl.h>
62#include <sys/sysproto.h>
63#include <sys/systm.h>
64#include <sys/time.h>
65#include <sys/vmmeter.h>
66#include <sys/vnode.h>
67#include <sys/wait.h>
68#include <sys/cpuset.h>
69
70#include <security/mac/mac_framework.h>
71
72#include <vm/vm.h>
73#include <vm/pmap.h>
74#include <vm/vm_kern.h>
75#include <vm/vm_map.h>
76#include <vm/vm_extern.h>
77#include <vm/vm_object.h>
78#include <vm/swap_pager.h>
79
80#ifdef COMPAT_LINUX32
81#include <machine/../linux32/linux.h>
82#include <machine/../linux32/linux32_proto.h>
83#else
84#include <machine/../linux/linux.h>
85#include <machine/../linux/linux_proto.h>
86#endif
87
88#include <compat/linux/linux_dtrace.h>
89#include <compat/linux/linux_file.h>
90#include <compat/linux/linux_mib.h>
91#include <compat/linux/linux_signal.h>
92#include <compat/linux/linux_timer.h>
93#include <compat/linux/linux_util.h>
94#include <compat/linux/linux_sysproto.h>
95#include <compat/linux/linux_emul.h>
96#include <compat/linux/linux_misc.h>
97
98/**
99 * Special DTrace provider for the linuxulator.
100 *
101 * In this file we define the provider for the entire linuxulator. All
102 * modules (= files of the linuxulator) use it.
103 *
104 * We define a different name depending on the emulated bitsize, see
105 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
106 * native bitsize = linuxulator
107 * amd64, 32bit emulation = linuxulator32
108 */
109LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
110
111int stclohz; /* Statistics clock frequency */
112
113static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = {
114 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK,
115 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE,
116 RLIMIT_MEMLOCK, RLIMIT_AS
117};
118
119struct l_sysinfo {
120 l_long uptime; /* Seconds since boot */
121 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */
122#define LINUX_SYSINFO_LOADS_SCALE 65536
123 l_ulong totalram; /* Total usable main memory size */
124 l_ulong freeram; /* Available memory size */
125 l_ulong sharedram; /* Amount of shared memory */
126 l_ulong bufferram; /* Memory used by buffers */
127 l_ulong totalswap; /* Total swap space size */
128 l_ulong freeswap; /* swap space still available */
129 l_ushort procs; /* Number of current processes */
130 l_ushort pads;
131 l_ulong totalbig;
132 l_ulong freebig;
133 l_uint mem_unit;
134 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */
135};
136
137struct l_pselect6arg {
138 l_uintptr_t ss;
139 l_size_t ss_len;
140};
141
142static int linux_utimensat_nsec_valid(l_long);
143
144
145int
146linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args)
147{
148 struct l_sysinfo sysinfo;
149 vm_object_t object;
150 int i, j;
151 struct timespec ts;
152
153 getnanouptime(&ts);
154 if (ts.tv_nsec != 0)
155 ts.tv_sec++;
156 sysinfo.uptime = ts.tv_sec;
157
158 /* Use the information from the mib to get our load averages */
159 for (i = 0; i < 3; i++)
160 sysinfo.loads[i] = averunnable.ldavg[i] *
161 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale;
162
163 sysinfo.totalram = physmem * PAGE_SIZE;
164 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE;
165
166 sysinfo.sharedram = 0;
167 mtx_lock(&vm_object_list_mtx);
168 TAILQ_FOREACH(object, &vm_object_list, object_list)
169 if (object->shadow_count > 1)
170 sysinfo.sharedram += object->resident_page_count;
171 mtx_unlock(&vm_object_list_mtx);
172
173 sysinfo.sharedram *= PAGE_SIZE;
174 sysinfo.bufferram = 0;
175
176 swap_pager_status(&i, &j);
177 sysinfo.totalswap = i * PAGE_SIZE;
178 sysinfo.freeswap = (i - j) * PAGE_SIZE;
179
180 sysinfo.procs = nprocs;
181
182 /* The following are only present in newer Linux kernels. */
183 sysinfo.totalbig = 0;
184 sysinfo.freebig = 0;
185 sysinfo.mem_unit = 1;
186
187 return (copyout(&sysinfo, args->info, sizeof(sysinfo)));
188}
189
190int
191linux_alarm(struct thread *td, struct linux_alarm_args *args)
192{
193 struct itimerval it, old_it;
194 u_int secs;
195 int error;
196
197#ifdef DEBUG
198 if (ldebug(alarm))
199 printf(ARGS(alarm, "%u"), args->secs);
200#endif
201
202 secs = args->secs;
203
204 if (secs > INT_MAX)
205 secs = INT_MAX;
206
207 it.it_value.tv_sec = (long) secs;
208 it.it_value.tv_usec = 0;
209 it.it_interval.tv_sec = 0;
210 it.it_interval.tv_usec = 0;
211 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it);
212 if (error)
213 return (error);
214 if (timevalisset(&old_it.it_value)) {
215 if (old_it.it_value.tv_usec != 0)
216 old_it.it_value.tv_sec++;
217 td->td_retval[0] = old_it.it_value.tv_sec;
218 }
219 return (0);
220}
221
222int
223linux_brk(struct thread *td, struct linux_brk_args *args)
224{
225 struct vmspace *vm = td->td_proc->p_vmspace;
226 vm_offset_t new, old;
227 struct obreak_args /* {
228 char * nsize;
229 } */ tmp;
230
231#ifdef DEBUG
232 if (ldebug(brk))
233 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend);
234#endif
235 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize);
236 new = (vm_offset_t)args->dsend;
237 tmp.nsize = (char *)new;
238 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp))
239 td->td_retval[0] = (long)new;
240 else
241 td->td_retval[0] = (long)old;
242
243 return (0);
244}
245
246#if defined(__i386__)
247/* XXX: what about amd64/linux32? */
248
249int
250linux_uselib(struct thread *td, struct linux_uselib_args *args)
251{
252 struct nameidata ni;
253 struct vnode *vp;
254 struct exec *a_out;
255 struct vattr attr;
256 vm_offset_t vmaddr;
257 unsigned long file_offset;
258 unsigned long bss_size;
259 char *library;
260 ssize_t aresid;
261 int error, locked, writecount;
262
263 LCONVPATHEXIST(td, args->library, &library);
264
265#ifdef DEBUG
266 if (ldebug(uselib))
267 printf(ARGS(uselib, "%s"), library);
268#endif
269
270 a_out = NULL;
271 locked = 0;
272 vp = NULL;
273
274 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1,
275 UIO_SYSSPACE, library, td);
276 error = namei(&ni);
277 LFREEPATH(library);
278 if (error)
279 goto cleanup;
280
281 vp = ni.ni_vp;
282 NDFREE(&ni, NDF_ONLY_PNBUF);
283
284 /*
285 * From here on down, we have a locked vnode that must be unlocked.
286 * XXX: The code below largely duplicates exec_check_permissions().
287 */
288 locked = 1;
289
290 /* Writable? */
291 error = VOP_GET_WRITECOUNT(vp, &writecount);
292 if (error != 0)
293 goto cleanup;
294 if (writecount != 0) {
295 error = ETXTBSY;
296 goto cleanup;
297 }
298
299 /* Executable? */
300 error = VOP_GETATTR(vp, &attr, td->td_ucred);
301 if (error)
302 goto cleanup;
303
304 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
305 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) {
306 /* EACCESS is what exec(2) returns. */
307 error = ENOEXEC;
308 goto cleanup;
309 }
310
311 /* Sensible size? */
312 if (attr.va_size == 0) {
313 error = ENOEXEC;
314 goto cleanup;
315 }
316
317 /* Can we access it? */
318 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
319 if (error)
320 goto cleanup;
321
322 /*
323 * XXX: This should use vn_open() so that it is properly authorized,
324 * and to reduce code redundancy all over the place here.
325 * XXX: Not really, it duplicates far more of exec_check_permissions()
326 * than vn_open().
327 */
328#ifdef MAC
329 error = mac_vnode_check_open(td->td_ucred, vp, VREAD);
330 if (error)
331 goto cleanup;
332#endif
333 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
334 if (error)
335 goto cleanup;
336
337 /* Pull in executable header into exec_map */
338 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE,
339 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0);
340 if (error)
341 goto cleanup;
342
343 /* Is it a Linux binary ? */
344 if (((a_out->a_magic >> 16) & 0xff) != 0x64) {
345 error = ENOEXEC;
346 goto cleanup;
347 }
348
349 /*
350 * While we are here, we should REALLY do some more checks
351 */
352
353 /* Set file/virtual offset based on a.out variant. */
354 switch ((int)(a_out->a_magic & 0xffff)) {
355 case 0413: /* ZMAGIC */
356 file_offset = 1024;
357 break;
358 case 0314: /* QMAGIC */
359 file_offset = 0;
360 break;
361 default:
362 error = ENOEXEC;
363 goto cleanup;
364 }
365
366 bss_size = round_page(a_out->a_bss);
367
368 /* Check various fields in header for validity/bounds. */
369 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) {
370 error = ENOEXEC;
371 goto cleanup;
372 }
373
374 /* text + data can't exceed file size */
375 if (a_out->a_data + a_out->a_text > attr.va_size) {
376 error = EFAULT;
377 goto cleanup;
378 }
379
380 /*
381 * text/data/bss must not exceed limits
382 * XXX - this is not complete. it should check current usage PLUS
383 * the resources needed by this library.
384 */
385 PROC_LOCK(td->td_proc);
386 if (a_out->a_text > maxtsiz ||
387 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) ||
388 racct_set(td->td_proc, RACCT_DATA, a_out->a_data +
389 bss_size) != 0) {
390 PROC_UNLOCK(td->td_proc);
391 error = ENOMEM;
392 goto cleanup;
393 }
394 PROC_UNLOCK(td->td_proc);
395
396 /*
397 * Prevent more writers.
398 * XXX: Note that if any of the VM operations fail below we don't
399 * clear this flag.
400 */
401 VOP_SET_TEXT(vp);
402
403 /*
404 * Lock no longer needed
405 */
406 locked = 0;
407 VOP_UNLOCK(vp, 0);
408
409 /*
410 * Check if file_offset page aligned. Currently we cannot handle
411 * misalinged file offsets, and so we read in the entire image
412 * (what a waste).
413 */
414 if (file_offset & PAGE_MASK) {
415#ifdef DEBUG
416 printf("uselib: Non page aligned binary %lu\n", file_offset);
417#endif
418 /* Map text+data read/write/execute */
419
420 /* a_entry is the load address and is page aligned */
421 vmaddr = trunc_page(a_out->a_entry);
422
423 /* get anon user mapping, read+write+execute */
424 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
425 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE,
426 VM_PROT_ALL, VM_PROT_ALL, 0);
427 if (error)
428 goto cleanup;
429
430 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset,
431 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0,
432 td->td_ucred, NOCRED, &aresid, td);
433 if (error != 0)
434 goto cleanup;
435 if (aresid != 0) {
436 error = ENOEXEC;
437 goto cleanup;
438 }
439 } else {
440#ifdef DEBUG
441 printf("uselib: Page aligned binary %lu\n", file_offset);
442#endif
443 /*
444 * for QMAGIC, a_entry is 20 bytes beyond the load address
445 * to skip the executable header
446 */
447 vmaddr = trunc_page(a_out->a_entry);
448
449 /*
450 * Map it all into the process's space as a single
451 * copy-on-write "data" segment.
452 */
453 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr,
454 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL,
455 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset);
456 if (error)
457 goto cleanup;
458 }
459#ifdef DEBUG
460 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0],
461 ((long *)vmaddr)[1]);
462#endif
463 if (bss_size != 0) {
464 /* Calculate BSS start address */
465 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text +
466 a_out->a_data;
467
468 /* allocate some 'anon' space */
469 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0,
470 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL,
471 VM_PROT_ALL, 0);
472 if (error)
473 goto cleanup;
474 }
475
476cleanup:
477 /* Unlock vnode if needed */
478 if (locked)
479 VOP_UNLOCK(vp, 0);
480
481 /* Release the temporary mapping. */
482 if (a_out)
483 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE);
484
485 return (error);
486}
487
488#endif /* __i386__ */
489
490int
491linux_select(struct thread *td, struct linux_select_args *args)
492{
493 l_timeval ltv;
494 struct timeval tv0, tv1, utv, *tvp;
495 int error;
496
497#ifdef DEBUG
498 if (ldebug(select))
499 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds,
500 (void *)args->readfds, (void *)args->writefds,
501 (void *)args->exceptfds, (void *)args->timeout);
502#endif
503
504 /*
505 * Store current time for computation of the amount of
506 * time left.
507 */
508 if (args->timeout) {
509 if ((error = copyin(args->timeout, <v, sizeof(ltv))))
510 goto select_out;
511 utv.tv_sec = ltv.tv_sec;
512 utv.tv_usec = ltv.tv_usec;
513#ifdef DEBUG
514 if (ldebug(select))
515 printf(LMSG("incoming timeout (%jd/%ld)"),
516 (intmax_t)utv.tv_sec, utv.tv_usec);
517#endif
518
519 if (itimerfix(&utv)) {
520 /*
521 * The timeval was invalid. Convert it to something
522 * valid that will act as it does under Linux.
523 */
524 utv.tv_sec += utv.tv_usec / 1000000;
525 utv.tv_usec %= 1000000;
526 if (utv.tv_usec < 0) {
527 utv.tv_sec -= 1;
528 utv.tv_usec += 1000000;
529 }
530 if (utv.tv_sec < 0)
531 timevalclear(&utv);
532 }
533 microtime(&tv0);
534 tvp = &utv;
535 } else
536 tvp = NULL;
537
538 error = kern_select(td, args->nfds, args->readfds, args->writefds,
539 args->exceptfds, tvp, LINUX_NFDBITS);
540
541#ifdef DEBUG
542 if (ldebug(select))
543 printf(LMSG("real select returns %d"), error);
544#endif
545 if (error)
546 goto select_out;
547
548 if (args->timeout) {
549 if (td->td_retval[0]) {
550 /*
551 * Compute how much time was left of the timeout,
552 * by subtracting the current time and the time
553 * before we started the call, and subtracting
554 * that result from the user-supplied value.
555 */
556 microtime(&tv1);
557 timevalsub(&tv1, &tv0);
558 timevalsub(&utv, &tv1);
559 if (utv.tv_sec < 0)
560 timevalclear(&utv);
561 } else
562 timevalclear(&utv);
563#ifdef DEBUG
564 if (ldebug(select))
565 printf(LMSG("outgoing timeout (%jd/%ld)"),
566 (intmax_t)utv.tv_sec, utv.tv_usec);
567#endif
568 ltv.tv_sec = utv.tv_sec;
569 ltv.tv_usec = utv.tv_usec;
570 if ((error = copyout(<v, args->timeout, sizeof(ltv))))
571 goto select_out;
572 }
573
574select_out:
575#ifdef DEBUG
576 if (ldebug(select))
577 printf(LMSG("select_out -> %d"), error);
578#endif
579 return (error);
580}
581
582int
583linux_mremap(struct thread *td, struct linux_mremap_args *args)
584{
585 struct munmap_args /* {
586 void *addr;
587 size_t len;
588 } */ bsd_args;
589 int error = 0;
590
591#ifdef DEBUG
592 if (ldebug(mremap))
593 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"),
594 (void *)(uintptr_t)args->addr,
595 (unsigned long)args->old_len,
596 (unsigned long)args->new_len,
597 (unsigned long)args->flags);
598#endif
599
600 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) {
601 td->td_retval[0] = 0;
602 return (EINVAL);
603 }
604
605 /*
606 * Check for the page alignment.
607 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK.
608 */
609 if (args->addr & PAGE_MASK) {
610 td->td_retval[0] = 0;
611 return (EINVAL);
612 }
613
614 args->new_len = round_page(args->new_len);
615 args->old_len = round_page(args->old_len);
616
617 if (args->new_len > args->old_len) {
618 td->td_retval[0] = 0;
619 return (ENOMEM);
620 }
621
622 if (args->new_len < args->old_len) {
623 bsd_args.addr =
624 (caddr_t)((uintptr_t)args->addr + args->new_len);
625 bsd_args.len = args->old_len - args->new_len;
626 error = sys_munmap(td, &bsd_args);
627 }
628
629 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr;
630 return (error);
631}
632
633#define LINUX_MS_ASYNC 0x0001
634#define LINUX_MS_INVALIDATE 0x0002
635#define LINUX_MS_SYNC 0x0004
636
637int
638linux_msync(struct thread *td, struct linux_msync_args *args)
639{
640 struct msync_args bsd_args;
641
642 bsd_args.addr = (caddr_t)(uintptr_t)args->addr;
643 bsd_args.len = (uintptr_t)args->len;
644 bsd_args.flags = args->fl & ~LINUX_MS_SYNC;
645
646 return (sys_msync(td, &bsd_args));
647}
648
649int
650linux_time(struct thread *td, struct linux_time_args *args)
651{
652 struct timeval tv;
653 l_time_t tm;
654 int error;
655
656#ifdef DEBUG
657 if (ldebug(time))
658 printf(ARGS(time, "*"));
659#endif
660
661 microtime(&tv);
662 tm = tv.tv_sec;
663 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm))))
664 return (error);
665 td->td_retval[0] = tm;
666 return (0);
667}
668
669struct l_times_argv {
670 l_clock_t tms_utime;
671 l_clock_t tms_stime;
672 l_clock_t tms_cutime;
673 l_clock_t tms_cstime;
674};
675
676
677/*
678 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
679 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK
680 * auxiliary vector entry.
681 */
682#define CLK_TCK 100
683
684#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK))
685#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz))
686
687#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \
688 CONVNTCK(r) : CONVOTCK(r))
689
690int
691linux_times(struct thread *td, struct linux_times_args *args)
692{
693 struct timeval tv, utime, stime, cutime, cstime;
694 struct l_times_argv tms;
695 struct proc *p;
696 int error;
697
698#ifdef DEBUG
699 if (ldebug(times))
700 printf(ARGS(times, "*"));
701#endif
702
703 if (args->buf != NULL) {
704 p = td->td_proc;
705 PROC_LOCK(p);
706 PROC_STATLOCK(p);
707 calcru(p, &utime, &stime);
708 PROC_STATUNLOCK(p);
709 calccru(p, &cutime, &cstime);
710 PROC_UNLOCK(p);
711
712 tms.tms_utime = CONVTCK(utime);
713 tms.tms_stime = CONVTCK(stime);
714
715 tms.tms_cutime = CONVTCK(cutime);
716 tms.tms_cstime = CONVTCK(cstime);
717
718 if ((error = copyout(&tms, args->buf, sizeof(tms))))
719 return (error);
720 }
721
722 microuptime(&tv);
723 td->td_retval[0] = (int)CONVTCK(tv);
724 return (0);
725}
726
727int
728linux_newuname(struct thread *td, struct linux_newuname_args *args)
729{
730 struct l_new_utsname utsname;
731 char osname[LINUX_MAX_UTSNAME];
732 char osrelease[LINUX_MAX_UTSNAME];
733 char *p;
734
735#ifdef DEBUG
736 if (ldebug(newuname))
737 printf(ARGS(newuname, "*"));
738#endif
739
740 linux_get_osname(td, osname);
741 linux_get_osrelease(td, osrelease);
742
743 bzero(&utsname, sizeof(utsname));
744 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME);
745 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME);
746 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME);
747 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME);
748 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME);
749 for (p = utsname.version; *p != '\0'; ++p)
750 if (*p == '\n') {
751 *p = '\0';
752 break;
753 }
754 strlcpy(utsname.machine, linux_kplatform, LINUX_MAX_UTSNAME);
755
756 return (copyout(&utsname, args->buf, sizeof(utsname)));
757}
758
759struct l_utimbuf {
760 l_time_t l_actime;
761 l_time_t l_modtime;
762};
763
764int
765linux_utime(struct thread *td, struct linux_utime_args *args)
766{
767 struct timeval tv[2], *tvp;
768 struct l_utimbuf lut;
769 char *fname;
770 int error;
771
772 LCONVPATHEXIST(td, args->fname, &fname);
773
774#ifdef DEBUG
775 if (ldebug(utime))
776 printf(ARGS(utime, "%s, *"), fname);
777#endif
778
779 if (args->times) {
780 if ((error = copyin(args->times, &lut, sizeof lut))) {
781 LFREEPATH(fname);
782 return (error);
783 }
784 tv[0].tv_sec = lut.l_actime;
785 tv[0].tv_usec = 0;
786 tv[1].tv_sec = lut.l_modtime;
787 tv[1].tv_usec = 0;
788 tvp = tv;
789 } else
790 tvp = NULL;
791
792 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
793 LFREEPATH(fname);
794 return (error);
795}
796
797int
798linux_utimes(struct thread *td, struct linux_utimes_args *args)
799{
800 l_timeval ltv[2];
801 struct timeval tv[2], *tvp = NULL;
802 char *fname;
803 int error;
804
805 LCONVPATHEXIST(td, args->fname, &fname);
806
807#ifdef DEBUG
808 if (ldebug(utimes))
809 printf(ARGS(utimes, "%s, *"), fname);
810#endif
811
812 if (args->tptr != NULL) {
813 if ((error = copyin(args->tptr, ltv, sizeof ltv))) {
814 LFREEPATH(fname);
815 return (error);
816 }
817 tv[0].tv_sec = ltv[0].tv_sec;
818 tv[0].tv_usec = ltv[0].tv_usec;
819 tv[1].tv_sec = ltv[1].tv_sec;
820 tv[1].tv_usec = ltv[1].tv_usec;
821 tvp = tv;
822 }
823
824 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
825 LFREEPATH(fname);
826 return (error);
827}
828
829static int
830linux_utimensat_nsec_valid(l_long nsec)
831{
832
833 if (nsec == LINUX_UTIME_OMIT || nsec == LINUX_UTIME_NOW)
834 return (0);
835 if (nsec >= 0 && nsec <= 999999999)
836 return (0);
837 return (1);
838}
839
840int
841linux_utimensat(struct thread *td, struct linux_utimensat_args *args)
842{
843 struct l_timespec l_times[2];
844 struct timespec times[2], *timesp = NULL;
845 char *path = NULL;
846 int error, dfd, flags = 0;
847
848 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
849
850#ifdef DEBUG
851 if (ldebug(utimensat))
852 printf(ARGS(utimensat, "%d, *"), dfd);
853#endif
854
855 if (args->flags & ~LINUX_AT_SYMLINK_NOFOLLOW)
856 return (EINVAL);
857
858 if (args->times != NULL) {
859 error = copyin(args->times, l_times, sizeof(l_times));
860 if (error != 0)
861 return (error);
862
863 if (linux_utimensat_nsec_valid(l_times[0].tv_nsec) != 0 ||
864 linux_utimensat_nsec_valid(l_times[1].tv_nsec) != 0)
865 return (EINVAL);
866
867 times[0].tv_sec = l_times[0].tv_sec;
868 switch (l_times[0].tv_nsec)
869 {
870 case LINUX_UTIME_OMIT:
871 times[0].tv_nsec = UTIME_OMIT;
872 break;
873 case LINUX_UTIME_NOW:
874 times[0].tv_nsec = UTIME_NOW;
875 break;
876 default:
877 times[0].tv_nsec = l_times[0].tv_nsec;
878 }
879
880 times[1].tv_sec = l_times[1].tv_sec;
881 switch (l_times[1].tv_nsec)
882 {
883 case LINUX_UTIME_OMIT:
884 times[1].tv_nsec = UTIME_OMIT;
885 break;
886 case LINUX_UTIME_NOW:
887 times[1].tv_nsec = UTIME_NOW;
888 break;
889 default:
890 times[1].tv_nsec = l_times[1].tv_nsec;
891 break;
892 }
893 timesp = times;
894 }
895
896 if (times[0].tv_nsec == UTIME_OMIT && times[1].tv_nsec == UTIME_OMIT)
897 /* This breaks POSIX, but is what the Linux kernel does
898 * _on purpose_ (documented in the man page for utimensat(2)),
899 * so we must follow that behaviour. */
900 return (0);
901
902 if (args->pathname != NULL)
903 LCONVPATHEXIST_AT(td, args->pathname, &path, dfd);
904 else if (args->flags != 0)
905 return (EINVAL);
906
907 if (args->flags & LINUX_AT_SYMLINK_NOFOLLOW)
908 flags |= AT_SYMLINK_NOFOLLOW;
909
910 if (path == NULL)
911 error = kern_futimens(td, dfd, timesp, UIO_SYSSPACE);
912 else {
913 error = kern_utimensat(td, dfd, path, UIO_SYSSPACE, timesp,
914 UIO_SYSSPACE, flags);
915 LFREEPATH(path);
916 }
917
918 return (error);
919}
920
921int
922linux_futimesat(struct thread *td, struct linux_futimesat_args *args)
923{
924 l_timeval ltv[2];
925 struct timeval tv[2], *tvp = NULL;
926 char *fname;
927 int error, dfd;
928
929 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
930 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd);
931
932#ifdef DEBUG
933 if (ldebug(futimesat))
934 printf(ARGS(futimesat, "%s, *"), fname);
935#endif
936
937 if (args->utimes != NULL) {
938 if ((error = copyin(args->utimes, ltv, sizeof ltv))) {
939 LFREEPATH(fname);
940 return (error);
941 }
942 tv[0].tv_sec = ltv[0].tv_sec;
943 tv[0].tv_usec = ltv[0].tv_usec;
944 tv[1].tv_sec = ltv[1].tv_sec;
945 tv[1].tv_usec = ltv[1].tv_usec;
946 tvp = tv;
947 }
948
949 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE);
950 LFREEPATH(fname);
951 return (error);
952}
953
954int
955linux_common_wait(struct thread *td, int pid, int *status,
956 int options, struct rusage *ru)
957{
958 int error, tmpstat;
959
960 error = kern_wait(td, pid, &tmpstat, options, ru);
961 if (error)
962 return (error);
963
964 if (status) {
965 tmpstat &= 0xffff;
966 if (WIFSIGNALED(tmpstat))
967 tmpstat = (tmpstat & 0xffffff80) |
968 bsd_to_linux_signal(WTERMSIG(tmpstat));
969 else if (WIFSTOPPED(tmpstat))
970 tmpstat = (tmpstat & 0xffff00ff) |
971 (bsd_to_linux_signal(WSTOPSIG(tmpstat)) << 8);
972 else if (WIFCONTINUED(tmpstat))
973 tmpstat = 0xffff;
974 error = copyout(&tmpstat, status, sizeof(int));
975 }
976
977 return (error);
978}
979
980#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
981int
982linux_waitpid(struct thread *td, struct linux_waitpid_args *args)
983{
984 struct linux_wait4_args wait4_args;
985
986#ifdef DEBUG
987 if (ldebug(waitpid))
988 printf(ARGS(waitpid, "%d, %p, %d"),
989 args->pid, (void *)args->status, args->options);
990#endif
991
992 wait4_args.pid = args->pid;
993 wait4_args.status = args->status;
994 wait4_args.options = args->options;
995 wait4_args.rusage = NULL;
996
997 return (linux_wait4(td, &wait4_args));
998}
999#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1000
1001int
1002linux_wait4(struct thread *td, struct linux_wait4_args *args)
1003{
1004 int error, options;
1005 struct rusage ru, *rup;
1006
1007#ifdef DEBUG
1008 if (ldebug(wait4))
1009 printf(ARGS(wait4, "%d, %p, %d, %p"),
1010 args->pid, (void *)args->status, args->options,
1011 (void *)args->rusage);
1012#endif
1013 if (args->options & ~(LINUX_WUNTRACED | LINUX_WNOHANG |
1014 LINUX_WCONTINUED | __WCLONE | __WNOTHREAD | __WALL))
1015 return (EINVAL);
1016
1017 options = WEXITED;
1018 linux_to_bsd_waitopts(args->options, &options);
1019
1020 if (args->rusage != NULL)
1021 rup = &ru;
1022 else
1023 rup = NULL;
1024 error = linux_common_wait(td, args->pid, args->status, options, rup);
1025 if (error != 0)
1026 return (error);
1027 if (args->rusage != NULL)
1028 error = linux_copyout_rusage(&ru, args->rusage);
1029 return (error);
1030}
1031
1032int
1033linux_waitid(struct thread *td, struct linux_waitid_args *args)
1034{
1035 int status, options, sig;
1036 struct __wrusage wru;
1037 siginfo_t siginfo;
1038 l_siginfo_t lsi;
1039 idtype_t idtype;
1040 struct proc *p;
1041 int error;
1042
1043 options = 0;
1044 linux_to_bsd_waitopts(args->options, &options);
1045
1046 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED))
1047 return (EINVAL);
1048 if (!(options & (WEXITED | WUNTRACED | WCONTINUED)))
1049 return (EINVAL);
1050
1051 switch (args->idtype) {
1052 case LINUX_P_ALL:
1053 idtype = P_ALL;
1054 break;
1055 case LINUX_P_PID:
1056 if (args->id <= 0)
1057 return (EINVAL);
1058 idtype = P_PID;
1059 break;
1060 case LINUX_P_PGID:
1061 if (args->id <= 0)
1062 return (EINVAL);
1063 idtype = P_PGID;
1064 break;
1065 default:
1066 return (EINVAL);
1067 }
1068
1069 error = kern_wait6(td, idtype, args->id, &status, options,
1070 &wru, &siginfo);
1071 if (error != 0)
1072 return (error);
1073 if (args->rusage != NULL) {
1074 error = linux_copyout_rusage(&wru.wru_children,
1075 args->rusage);
1076 if (error != 0)
1077 return (error);
1078 }
1079 if (args->info != NULL) {
1080 p = td->td_proc;
1081 if (td->td_retval[0] == 0)
1082 bzero(&lsi, sizeof(lsi));
1083 else {
1084 sig = bsd_to_linux_signal(siginfo.si_signo);
1085 siginfo_to_lsiginfo(&siginfo, &lsi, sig);
1086 }
1087 error = copyout(&lsi, args->info, sizeof(lsi));
1088 }
1089 td->td_retval[0] = 0;
1090
1091 return (error);
1092}
1093
1094int
1095linux_mknod(struct thread *td, struct linux_mknod_args *args)
1096{
1097 char *path;
1098 int error;
1099
1100 LCONVPATHCREAT(td, args->path, &path);
1101
1102#ifdef DEBUG
1103 if (ldebug(mknod))
1104 printf(ARGS(mknod, "%s, %d, %ju"), path, args->mode,
1105 (uintmax_t)args->dev);
1106#endif
1107
1108 switch (args->mode & S_IFMT) {
1109 case S_IFIFO:
1110 case S_IFSOCK:
1111 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode);
1112 break;
1113
1114 case S_IFCHR:
1115 case S_IFBLK:
1116 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode,
1117 args->dev);
1118 break;
1119
1120 case S_IFDIR:
1121 error = EPERM;
1122 break;
1123
1124 case 0:
1125 args->mode |= S_IFREG;
1126 /* FALLTHROUGH */
1127 case S_IFREG:
1128 error = kern_open(td, path, UIO_SYSSPACE,
1129 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1130 if (error == 0)
1131 kern_close(td, td->td_retval[0]);
1132 break;
1133
1134 default:
1135 error = EINVAL;
1136 break;
1137 }
1138 LFREEPATH(path);
1139 return (error);
1140}
1141
1142int
1143linux_mknodat(struct thread *td, struct linux_mknodat_args *args)
1144{
1145 char *path;
1146 int error, dfd;
1147
1148 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1149 LCONVPATHCREAT_AT(td, args->filename, &path, dfd);
1150
1151#ifdef DEBUG
1152 if (ldebug(mknodat))
1153 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev);
1154#endif
1155
1156 switch (args->mode & S_IFMT) {
1157 case S_IFIFO:
1158 case S_IFSOCK:
1159 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode);
1160 break;
1161
1162 case S_IFCHR:
1163 case S_IFBLK:
1164 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode,
1165 args->dev);
1166 break;
1167
1168 case S_IFDIR:
1169 error = EPERM;
1170 break;
1171
1172 case 0:
1173 args->mode |= S_IFREG;
1174 /* FALLTHROUGH */
1175 case S_IFREG:
1176 error = kern_openat(td, dfd, path, UIO_SYSSPACE,
1177 O_WRONLY | O_CREAT | O_TRUNC, args->mode);
1178 if (error == 0)
1179 kern_close(td, td->td_retval[0]);
1180 break;
1181
1182 default:
1183 error = EINVAL;
1184 break;
1185 }
1186 LFREEPATH(path);
1187 return (error);
1188}
1189
1190/*
1191 * UGH! This is just about the dumbest idea I've ever heard!!
1192 */
1193int
1194linux_personality(struct thread *td, struct linux_personality_args *args)
1195{
1196#ifdef DEBUG
1197 if (ldebug(personality))
1198 printf(ARGS(personality, "%lu"), (unsigned long)args->per);
1199#endif
1200 if (args->per != 0)
1201 return (EINVAL);
1202
1203 /* Yes Jim, it's still a Linux... */
1204 td->td_retval[0] = 0;
1205 return (0);
1206}
1207
1208struct l_itimerval {
1209 l_timeval it_interval;
1210 l_timeval it_value;
1211};
1212
1213#define B2L_ITIMERVAL(bip, lip) \
1214 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \
1215 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \
1216 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \
1217 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec;
1218
1219int
1220linux_setitimer(struct thread *td, struct linux_setitimer_args *uap)
1221{
1222 int error;
1223 struct l_itimerval ls;
1224 struct itimerval aitv, oitv;
1225
1226#ifdef DEBUG
1227 if (ldebug(setitimer))
1228 printf(ARGS(setitimer, "%p, %p"),
1229 (void *)uap->itv, (void *)uap->oitv);
1230#endif
1231
1232 if (uap->itv == NULL) {
1233 uap->itv = uap->oitv;
1234 return (linux_getitimer(td, (struct linux_getitimer_args *)uap));
1235 }
1236
1237 error = copyin(uap->itv, &ls, sizeof(ls));
1238 if (error != 0)
1239 return (error);
1240 B2L_ITIMERVAL(&aitv, &ls);
1241#ifdef DEBUG
1242 if (ldebug(setitimer)) {
1243 printf("setitimer: value: sec: %jd, usec: %ld\n",
1244 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec);
1245 printf("setitimer: interval: sec: %jd, usec: %ld\n",
1246 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec);
1247 }
1248#endif
1249 error = kern_setitimer(td, uap->which, &aitv, &oitv);
1250 if (error != 0 || uap->oitv == NULL)
1251 return (error);
1252 B2L_ITIMERVAL(&ls, &oitv);
1253
1254 return (copyout(&ls, uap->oitv, sizeof(ls)));
1255}
1256
1257int
1258linux_getitimer(struct thread *td, struct linux_getitimer_args *uap)
1259{
1260 int error;
1261 struct l_itimerval ls;
1262 struct itimerval aitv;
1263
1264#ifdef DEBUG
1265 if (ldebug(getitimer))
1266 printf(ARGS(getitimer, "%p"), (void *)uap->itv);
1267#endif
1268 error = kern_getitimer(td, uap->which, &aitv);
1269 if (error != 0)
1270 return (error);
1271 B2L_ITIMERVAL(&ls, &aitv);
1272 return (copyout(&ls, uap->itv, sizeof(ls)));
1273}
1274
1275#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1276int
1277linux_nice(struct thread *td, struct linux_nice_args *args)
1278{
1279 struct setpriority_args bsd_args;
1280
1281 bsd_args.which = PRIO_PROCESS;
1282 bsd_args.who = 0; /* current process */
1283 bsd_args.prio = args->inc;
1284 return (sys_setpriority(td, &bsd_args));
1285}
1286#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1287
1288int
1289linux_setgroups(struct thread *td, struct linux_setgroups_args *args)
1290{
1291 struct ucred *newcred, *oldcred;
1292 l_gid_t *linux_gidset;
1293 gid_t *bsd_gidset;
1294 int ngrp, error;
1295 struct proc *p;
1296
1297 ngrp = args->gidsetsize;
1298 if (ngrp < 0 || ngrp >= ngroups_max + 1)
1299 return (EINVAL);
1300 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_LINUX, M_WAITOK);
1301 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t));
1302 if (error)
1303 goto out;
1304 newcred = crget();
1305 crextend(newcred, ngrp + 1);
1306 p = td->td_proc;
1307 PROC_LOCK(p);
1308 oldcred = p->p_ucred;
1309 crcopy(newcred, oldcred);
1310
1311 /*
1312 * cr_groups[0] holds egid. Setting the whole set from
1313 * the supplied set will cause egid to be changed too.
1314 * Keep cr_groups[0] unchanged to prevent that.
1315 */
1316
1317 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) {
1318 PROC_UNLOCK(p);
1319 crfree(newcred);
1320 goto out;
1321 }
1322
1323 if (ngrp > 0) {
1324 newcred->cr_ngroups = ngrp + 1;
1325
1326 bsd_gidset = newcred->cr_groups;
1327 ngrp--;
1328 while (ngrp >= 0) {
1329 bsd_gidset[ngrp + 1] = linux_gidset[ngrp];
1330 ngrp--;
1331 }
1332 } else
1333 newcred->cr_ngroups = 1;
1334
1335 setsugid(p);
1336 p->p_ucred = newcred;
1337 PROC_UNLOCK(p);
1338 crfree(oldcred);
1339 error = 0;
1340out:
1341 free(linux_gidset, M_LINUX);
1342 return (error);
1343}
1344
1345int
1346linux_getgroups(struct thread *td, struct linux_getgroups_args *args)
1347{
1348 struct ucred *cred;
1349 l_gid_t *linux_gidset;
1350 gid_t *bsd_gidset;
1351 int bsd_gidsetsz, ngrp, error;
1352
1353 cred = td->td_ucred;
1354 bsd_gidset = cred->cr_groups;
1355 bsd_gidsetsz = cred->cr_ngroups - 1;
1356
1357 /*
1358 * cr_groups[0] holds egid. Returning the whole set
1359 * here will cause a duplicate. Exclude cr_groups[0]
1360 * to prevent that.
1361 */
1362
1363 if ((ngrp = args->gidsetsize) == 0) {
1364 td->td_retval[0] = bsd_gidsetsz;
1365 return (0);
1366 }
1367
1368 if (ngrp < bsd_gidsetsz)
1369 return (EINVAL);
1370
1371 ngrp = 0;
1372 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset),
1373 M_LINUX, M_WAITOK);
1374 while (ngrp < bsd_gidsetsz) {
1375 linux_gidset[ngrp] = bsd_gidset[ngrp + 1];
1376 ngrp++;
1377 }
1378
1379 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t));
1380 free(linux_gidset, M_LINUX);
1381 if (error)
1382 return (error);
1383
1384 td->td_retval[0] = ngrp;
1385 return (0);
1386}
1387
1388int
1389linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args)
1390{
1391 struct rlimit bsd_rlim;
1392 struct l_rlimit rlim;
1393 u_int which;
1394 int error;
1395
1396#ifdef DEBUG
1397 if (ldebug(setrlimit))
1398 printf(ARGS(setrlimit, "%d, %p"),
1399 args->resource, (void *)args->rlim);
1400#endif
1401
1402 if (args->resource >= LINUX_RLIM_NLIMITS)
1403 return (EINVAL);
1404
1405 which = linux_to_bsd_resource[args->resource];
1406 if (which == -1)
1407 return (EINVAL);
1408
1409 error = copyin(args->rlim, &rlim, sizeof(rlim));
1410 if (error)
1411 return (error);
1412
1413 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur;
1414 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max;
1415 return (kern_setrlimit(td, which, &bsd_rlim));
1416}
1417
1418#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1419int
1420linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args)
1421{
1422 struct l_rlimit rlim;
1423 struct proc *p = td->td_proc;
1424 struct rlimit bsd_rlim;
1425 u_int which;
1426
1427#ifdef DEBUG
1428 if (ldebug(old_getrlimit))
1429 printf(ARGS(old_getrlimit, "%d, %p"),
1430 args->resource, (void *)args->rlim);
1431#endif
1432
1433 if (args->resource >= LINUX_RLIM_NLIMITS)
1434 return (EINVAL);
1435
1436 which = linux_to_bsd_resource[args->resource];
1437 if (which == -1)
1438 return (EINVAL);
1439
1440 PROC_LOCK(p);
1441 lim_rlimit(p, which, &bsd_rlim);
1442 PROC_UNLOCK(p);
1443
1444#ifdef COMPAT_LINUX32
1445 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur;
1446 if (rlim.rlim_cur == UINT_MAX)
1447 rlim.rlim_cur = INT_MAX;
1448 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max;
1449 if (rlim.rlim_max == UINT_MAX)
1450 rlim.rlim_max = INT_MAX;
1451#else
1452 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur;
1453 if (rlim.rlim_cur == ULONG_MAX)
1454 rlim.rlim_cur = LONG_MAX;
1455 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max;
1456 if (rlim.rlim_max == ULONG_MAX)
1457 rlim.rlim_max = LONG_MAX;
1458#endif
1459 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1460}
1461#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1462
1463int
1464linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args)
1465{
1466 struct l_rlimit rlim;
1467 struct proc *p = td->td_proc;
1468 struct rlimit bsd_rlim;
1469 u_int which;
1470
1471#ifdef DEBUG
1472 if (ldebug(getrlimit))
1473 printf(ARGS(getrlimit, "%d, %p"),
1474 args->resource, (void *)args->rlim);
1475#endif
1476
1477 if (args->resource >= LINUX_RLIM_NLIMITS)
1478 return (EINVAL);
1479
1480 which = linux_to_bsd_resource[args->resource];
1481 if (which == -1)
1482 return (EINVAL);
1483
1484 PROC_LOCK(p);
1485 lim_rlimit(p, which, &bsd_rlim);
1486 PROC_UNLOCK(p);
1487
1488 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur;
1489 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max;
1490 return (copyout(&rlim, args->rlim, sizeof(rlim)));
1491}
1492
1493int
1494linux_sched_setscheduler(struct thread *td,
1495 struct linux_sched_setscheduler_args *args)
1496{
1497 struct sched_param sched_param;
1498 struct thread *tdt;
1499 int error, policy;
1500
1501#ifdef DEBUG
1502 if (ldebug(sched_setscheduler))
1503 printf(ARGS(sched_setscheduler, "%d, %d, %p"),
1504 args->pid, args->policy, (const void *)args->param);
1505#endif
1506
1507 switch (args->policy) {
1508 case LINUX_SCHED_OTHER:
1509 policy = SCHED_OTHER;
1510 break;
1511 case LINUX_SCHED_FIFO:
1512 policy = SCHED_FIFO;
1513 break;
1514 case LINUX_SCHED_RR:
1515 policy = SCHED_RR;
1516 break;
1517 default:
1518 return (EINVAL);
1519 }
1520
1521 error = copyin(args->param, &sched_param, sizeof(sched_param));
1522 if (error)
1523 return (error);
1524
1525 tdt = linux_tdfind(td, args->pid, -1);
1526 if (tdt == NULL)
1527 return (ESRCH);
1528
1529 error = kern_sched_setscheduler(td, tdt, policy, &sched_param);
1530 PROC_UNLOCK(tdt->td_proc);
1531 return (error);
1532}
1533
1534int
1535linux_sched_getscheduler(struct thread *td,
1536 struct linux_sched_getscheduler_args *args)
1537{
1538 struct thread *tdt;
1539 int error, policy;
1540
1541#ifdef DEBUG
1542 if (ldebug(sched_getscheduler))
1543 printf(ARGS(sched_getscheduler, "%d"), args->pid);
1544#endif
1545
1546 tdt = linux_tdfind(td, args->pid, -1);
1547 if (tdt == NULL)
1548 return (ESRCH);
1549
1550 error = kern_sched_getscheduler(td, tdt, &policy);
1551 PROC_UNLOCK(tdt->td_proc);
1552
1553 switch (policy) {
1554 case SCHED_OTHER:
1555 td->td_retval[0] = LINUX_SCHED_OTHER;
1556 break;
1557 case SCHED_FIFO:
1558 td->td_retval[0] = LINUX_SCHED_FIFO;
1559 break;
1560 case SCHED_RR:
1561 td->td_retval[0] = LINUX_SCHED_RR;
1562 break;
1563 }
1564 return (error);
1565}
1566
1567int
1568linux_sched_get_priority_max(struct thread *td,
1569 struct linux_sched_get_priority_max_args *args)
1570{
1571 struct sched_get_priority_max_args bsd;
1572
1573#ifdef DEBUG
1574 if (ldebug(sched_get_priority_max))
1575 printf(ARGS(sched_get_priority_max, "%d"), args->policy);
1576#endif
1577
1578 switch (args->policy) {
1579 case LINUX_SCHED_OTHER:
1580 bsd.policy = SCHED_OTHER;
1581 break;
1582 case LINUX_SCHED_FIFO:
1583 bsd.policy = SCHED_FIFO;
1584 break;
1585 case LINUX_SCHED_RR:
1586 bsd.policy = SCHED_RR;
1587 break;
1588 default:
1589 return (EINVAL);
1590 }
1591 return (sys_sched_get_priority_max(td, &bsd));
1592}
1593
1594int
1595linux_sched_get_priority_min(struct thread *td,
1596 struct linux_sched_get_priority_min_args *args)
1597{
1598 struct sched_get_priority_min_args bsd;
1599
1600#ifdef DEBUG
1601 if (ldebug(sched_get_priority_min))
1602 printf(ARGS(sched_get_priority_min, "%d"), args->policy);
1603#endif
1604
1605 switch (args->policy) {
1606 case LINUX_SCHED_OTHER:
1607 bsd.policy = SCHED_OTHER;
1608 break;
1609 case LINUX_SCHED_FIFO:
1610 bsd.policy = SCHED_FIFO;
1611 break;
1612 case LINUX_SCHED_RR:
1613 bsd.policy = SCHED_RR;
1614 break;
1615 default:
1616 return (EINVAL);
1617 }
1618 return (sys_sched_get_priority_min(td, &bsd));
1619}
1620
1621#define REBOOT_CAD_ON 0x89abcdef
1622#define REBOOT_CAD_OFF 0
1623#define REBOOT_HALT 0xcdef0123
1624#define REBOOT_RESTART 0x01234567
1625#define REBOOT_RESTART2 0xA1B2C3D4
1626#define REBOOT_POWEROFF 0x4321FEDC
1627#define REBOOT_MAGIC1 0xfee1dead
1628#define REBOOT_MAGIC2 0x28121969
1629#define REBOOT_MAGIC2A 0x05121996
1630#define REBOOT_MAGIC2B 0x16041998
1631
1632int
1633linux_reboot(struct thread *td, struct linux_reboot_args *args)
1634{
1635 struct reboot_args bsd_args;
1636
1637#ifdef DEBUG
1638 if (ldebug(reboot))
1639 printf(ARGS(reboot, "0x%x"), args->cmd);
1640#endif
1641
1642 if (args->magic1 != REBOOT_MAGIC1)
1643 return (EINVAL);
1644
1645 switch (args->magic2) {
1646 case REBOOT_MAGIC2:
1647 case REBOOT_MAGIC2A:
1648 case REBOOT_MAGIC2B:
1649 break;
1650 default:
1651 return (EINVAL);
1652 }
1653
1654 switch (args->cmd) {
1655 case REBOOT_CAD_ON:
1656 case REBOOT_CAD_OFF:
1657 return (priv_check(td, PRIV_REBOOT));
1658 case REBOOT_HALT:
1659 bsd_args.opt = RB_HALT;
1660 break;
1661 case REBOOT_RESTART:
1662 case REBOOT_RESTART2:
1663 bsd_args.opt = 0;
1664 break;
1665 case REBOOT_POWEROFF:
1666 bsd_args.opt = RB_POWEROFF;
1667 break;
1668 default:
1669 return (EINVAL);
1670 }
1671 return (sys_reboot(td, &bsd_args));
1672}
1673
1674
1675/*
1676 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify
1677 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that
1678 * are assumed to be preserved. The following lightweight syscalls fixes
1679 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c
1680 *
1681 * linux_getpid() - MP SAFE
1682 * linux_getgid() - MP SAFE
1683 * linux_getuid() - MP SAFE
1684 */
1685
1686int
1687linux_getpid(struct thread *td, struct linux_getpid_args *args)
1688{
1689
1690#ifdef DEBUG
1691 if (ldebug(getpid))
1692 printf(ARGS(getpid, ""));
1693#endif
1694 td->td_retval[0] = td->td_proc->p_pid;
1695
1696 return (0);
1697}
1698
1699int
1700linux_gettid(struct thread *td, struct linux_gettid_args *args)
1701{
1702 struct linux_emuldata *em;
1703
1704#ifdef DEBUG
1705 if (ldebug(gettid))
1706 printf(ARGS(gettid, ""));
1707#endif
1708
1709 em = em_find(td);
1710 KASSERT(em != NULL, ("gettid: emuldata not found.\n"));
1711
1712 td->td_retval[0] = em->em_tid;
1713
1714 return (0);
1715}
1716
1717
1718int
1719linux_getppid(struct thread *td, struct linux_getppid_args *args)
1720{
1721
1722#ifdef DEBUG
1723 if (ldebug(getppid))
1724 printf(ARGS(getppid, ""));
1725#endif
1726
1727 PROC_LOCK(td->td_proc);
1728 td->td_retval[0] = td->td_proc->p_pptr->p_pid;
1729 PROC_UNLOCK(td->td_proc);
1730 return (0);
1731}
1732
1733int
1734linux_getgid(struct thread *td, struct linux_getgid_args *args)
1735{
1736
1737#ifdef DEBUG
1738 if (ldebug(getgid))
1739 printf(ARGS(getgid, ""));
1740#endif
1741
1742 td->td_retval[0] = td->td_ucred->cr_rgid;
1743 return (0);
1744}
1745
1746int
1747linux_getuid(struct thread *td, struct linux_getuid_args *args)
1748{
1749
1750#ifdef DEBUG
1751 if (ldebug(getuid))
1752 printf(ARGS(getuid, ""));
1753#endif
1754
1755 td->td_retval[0] = td->td_ucred->cr_ruid;
1756 return (0);
1757}
1758
1759
1760int
1761linux_getsid(struct thread *td, struct linux_getsid_args *args)
1762{
1763 struct getsid_args bsd;
1764
1765#ifdef DEBUG
1766 if (ldebug(getsid))
1767 printf(ARGS(getsid, "%i"), args->pid);
1768#endif
1769
1770 bsd.pid = args->pid;
1771 return (sys_getsid(td, &bsd));
1772}
1773
1774int
1775linux_nosys(struct thread *td, struct nosys_args *ignore)
1776{
1777
1778 return (ENOSYS);
1779}
1780
1781int
1782linux_getpriority(struct thread *td, struct linux_getpriority_args *args)
1783{
1784 struct getpriority_args bsd_args;
1785 int error;
1786
1787#ifdef DEBUG
1788 if (ldebug(getpriority))
1789 printf(ARGS(getpriority, "%i, %i"), args->which, args->who);
1790#endif
1791
1792 bsd_args.which = args->which;
1793 bsd_args.who = args->who;
1794 error = sys_getpriority(td, &bsd_args);
1795 td->td_retval[0] = 20 - td->td_retval[0];
1796 return (error);
1797}
1798
1799int
1800linux_sethostname(struct thread *td, struct linux_sethostname_args *args)
1801{
1802 int name[2];
1803
1804#ifdef DEBUG
1805 if (ldebug(sethostname))
1806 printf(ARGS(sethostname, "*, %i"), args->len);
1807#endif
1808
1809 name[0] = CTL_KERN;
1810 name[1] = KERN_HOSTNAME;
1811 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname,
1812 args->len, 0, 0));
1813}
1814
1815int
1816linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args)
1817{
1818 int name[2];
1819
1820#ifdef DEBUG
1821 if (ldebug(setdomainname))
1822 printf(ARGS(setdomainname, "*, %i"), args->len);
1823#endif
1824
1825 name[0] = CTL_KERN;
1826 name[1] = KERN_NISDOMAINNAME;
1827 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name,
1828 args->len, 0, 0));
1829}
1830
1831int
1832linux_exit_group(struct thread *td, struct linux_exit_group_args *args)
1833{
1834
1835#ifdef DEBUG
1836 if (ldebug(exit_group))
1837 printf(ARGS(exit_group, "%i"), args->error_code);
1838#endif
1839
1840 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid,
1841 args->error_code);
1842
1843 /*
1844 * XXX: we should send a signal to the parent if
1845 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?)
1846 * as it doesnt occur often.
1847 */
1848 exit1(td, W_EXITCODE(args->error_code, 0));
1849 /* NOTREACHED */
1850}
1851
1852#define _LINUX_CAPABILITY_VERSION 0x19980330
1853
1854struct l_user_cap_header {
1855 l_int version;
1856 l_int pid;
1857};
1858
1859struct l_user_cap_data {
1860 l_int effective;
1861 l_int permitted;
1862 l_int inheritable;
1863};
1864
1865int
1866linux_capget(struct thread *td, struct linux_capget_args *args)
1867{
1868 struct l_user_cap_header luch;
1869 struct l_user_cap_data lucd;
1870 int error;
1871
1872 if (args->hdrp == NULL)
1873 return (EFAULT);
1874
1875 error = copyin(args->hdrp, &luch, sizeof(luch));
1876 if (error != 0)
1877 return (error);
1878
1879 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1880 luch.version = _LINUX_CAPABILITY_VERSION;
1881 error = copyout(&luch, args->hdrp, sizeof(luch));
1882 if (error)
1883 return (error);
1884 return (EINVAL);
1885 }
1886
1887 if (luch.pid)
1888 return (EPERM);
1889
1890 if (args->datap) {
1891 /*
1892 * The current implementation doesn't support setting
1893 * a capability (it's essentially a stub) so indicate
1894 * that no capabilities are currently set or available
1895 * to request.
1896 */
1897 bzero (&lucd, sizeof(lucd));
1898 error = copyout(&lucd, args->datap, sizeof(lucd));
1899 }
1900
1901 return (error);
1902}
1903
1904int
1905linux_capset(struct thread *td, struct linux_capset_args *args)
1906{
1907 struct l_user_cap_header luch;
1908 struct l_user_cap_data lucd;
1909 int error;
1910
1911 if (args->hdrp == NULL || args->datap == NULL)
1912 return (EFAULT);
1913
1914 error = copyin(args->hdrp, &luch, sizeof(luch));
1915 if (error != 0)
1916 return (error);
1917
1918 if (luch.version != _LINUX_CAPABILITY_VERSION) {
1919 luch.version = _LINUX_CAPABILITY_VERSION;
1920 error = copyout(&luch, args->hdrp, sizeof(luch));
1921 if (error)
1922 return (error);
1923 return (EINVAL);
1924 }
1925
1926 if (luch.pid)
1927 return (EPERM);
1928
1929 error = copyin(args->datap, &lucd, sizeof(lucd));
1930 if (error != 0)
1931 return (error);
1932
1933 /* We currently don't support setting any capabilities. */
1934 if (lucd.effective || lucd.permitted || lucd.inheritable) {
1935 linux_msg(td,
1936 "capset effective=0x%x, permitted=0x%x, "
1937 "inheritable=0x%x is not implemented",
1938 (int)lucd.effective, (int)lucd.permitted,
1939 (int)lucd.inheritable);
1940 return (EPERM);
1941 }
1942
1943 return (0);
1944}
1945
1946int
1947linux_prctl(struct thread *td, struct linux_prctl_args *args)
1948{
1949 int error = 0, max_size;
1950 struct proc *p = td->td_proc;
1951 char comm[LINUX_MAX_COMM_LEN];
1952 struct linux_emuldata *em;
1953 int pdeath_signal;
1954
1955#ifdef DEBUG
1956 if (ldebug(prctl))
1957 printf(ARGS(prctl, "%d, %ju, %ju, %ju, %ju"), args->option,
1958 (uintmax_t)args->arg2, (uintmax_t)args->arg3,
1959 (uintmax_t)args->arg4, (uintmax_t)args->arg5);
1960#endif
1961
1962 switch (args->option) {
1963 case LINUX_PR_SET_PDEATHSIG:
1964 if (!LINUX_SIG_VALID(args->arg2))
1965 return (EINVAL);
1966 em = em_find(td);
1967 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1968 em->pdeath_signal = args->arg2;
1969 break;
1970 case LINUX_PR_GET_PDEATHSIG:
1971 em = em_find(td);
1972 KASSERT(em != NULL, ("prctl: emuldata not found.\n"));
1973 pdeath_signal = em->pdeath_signal;
1974 error = copyout(&pdeath_signal,
1975 (void *)(register_t)args->arg2,
1976 sizeof(pdeath_signal));
1977 break;
1978 case LINUX_PR_GET_KEEPCAPS:
1979 /*
1980 * Indicate that we always clear the effective and
1981 * permitted capability sets when the user id becomes
1982 * non-zero (actually the capability sets are simply
1983 * always zero in the current implementation).
1984 */
1985 td->td_retval[0] = 0;
1986 break;
1987 case LINUX_PR_SET_KEEPCAPS:
1988 /*
1989 * Ignore requests to keep the effective and permitted
1990 * capability sets when the user id becomes non-zero.
1991 */
1992 break;
1993 case LINUX_PR_SET_NAME:
1994 /*
1995 * To be on the safe side we need to make sure to not
1996 * overflow the size a linux program expects. We already
1997 * do this here in the copyin, so that we don't need to
1998 * check on copyout.
1999 */
2000 max_size = MIN(sizeof(comm), sizeof(p->p_comm));
2001 error = copyinstr((void *)(register_t)args->arg2, comm,
2002 max_size, NULL);
2003
2004 /* Linux silently truncates the name if it is too long. */
2005 if (error == ENAMETOOLONG) {
2006 /*
2007 * XXX: copyinstr() isn't documented to populate the
2008 * array completely, so do a copyin() to be on the
2009 * safe side. This should be changed in case
2010 * copyinstr() is changed to guarantee this.
2011 */
2012 error = copyin((void *)(register_t)args->arg2, comm,
2013 max_size - 1);
2014 comm[max_size - 1] = '\0';
2015 }
2016 if (error)
2017 return (error);
2018
2019 PROC_LOCK(p);
2020 strlcpy(p->p_comm, comm, sizeof(p->p_comm));
2021 PROC_UNLOCK(p);
2022 break;
2023 case LINUX_PR_GET_NAME:
2024 PROC_LOCK(p);
2025 strlcpy(comm, p->p_comm, sizeof(comm));
2026 PROC_UNLOCK(p);
2027 error = copyout(comm, (void *)(register_t)args->arg2,
2028 strlen(comm) + 1);
2029 break;
2030 default:
2031 error = EINVAL;
2032 break;
2033 }
2034
2035 return (error);
2036}
2037
2038int
2039linux_sched_setparam(struct thread *td,
2040 struct linux_sched_setparam_args *uap)
2041{
2042 struct sched_param sched_param;
2043 struct thread *tdt;
2044 int error;
2045
2046#ifdef DEBUG
2047 if (ldebug(sched_setparam))
2048 printf(ARGS(sched_setparam, "%d, *"), uap->pid);
2049#endif
2050
2051 error = copyin(uap->param, &sched_param, sizeof(sched_param));
2052 if (error)
2053 return (error);
2054
2055 tdt = linux_tdfind(td, uap->pid, -1);
2056 if (tdt == NULL)
2057 return (ESRCH);
2058
2059 error = kern_sched_setparam(td, tdt, &sched_param);
2060 PROC_UNLOCK(tdt->td_proc);
2061 return (error);
2062}
2063
2064int
2065linux_sched_getparam(struct thread *td,
2066 struct linux_sched_getparam_args *uap)
2067{
2068 struct sched_param sched_param;
2069 struct thread *tdt;
2070 int error;
2071
2072#ifdef DEBUG
2073 if (ldebug(sched_getparam))
2074 printf(ARGS(sched_getparam, "%d, *"), uap->pid);
2075#endif
2076
2077 tdt = linux_tdfind(td, uap->pid, -1);
2078 if (tdt == NULL)
2079 return (ESRCH);
2080
2081 error = kern_sched_getparam(td, tdt, &sched_param);
2082 PROC_UNLOCK(tdt->td_proc);
2083 if (error == 0)
2084 error = copyout(&sched_param, uap->param,
2085 sizeof(sched_param));
2086 return (error);
2087}
2088
2089/*
2090 * Get affinity of a process.
2091 */
2092int
2093linux_sched_getaffinity(struct thread *td,
2094 struct linux_sched_getaffinity_args *args)
2095{
2096 int error;
2097 struct thread *tdt;
2098 struct cpuset_getaffinity_args cga;
2099
2100#ifdef DEBUG
2101 if (ldebug(sched_getaffinity))
2102 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid,
2103 args->len);
2104#endif
2105 if (args->len < sizeof(cpuset_t))
2106 return (EINVAL);
2107
2108 tdt = linux_tdfind(td, args->pid, -1);
2109 if (tdt == NULL)
2110 return (ESRCH);
2111
2112 PROC_UNLOCK(tdt->td_proc);
2113 cga.level = CPU_LEVEL_WHICH;
2114 cga.which = CPU_WHICH_TID;
2115 cga.id = tdt->td_tid;
2116 cga.cpusetsize = sizeof(cpuset_t);
2117 cga.mask = (cpuset_t *) args->user_mask_ptr;
2118
2119 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0)
2120 td->td_retval[0] = sizeof(cpuset_t);
2121
2122 return (error);
2123}
2124
2125/*
2126 * Set affinity of a process.
2127 */
2128int
2129linux_sched_setaffinity(struct thread *td,
2130 struct linux_sched_setaffinity_args *args)
2131{
2132 struct cpuset_setaffinity_args csa;
2133 struct thread *tdt;
2134
2135#ifdef DEBUG
2136 if (ldebug(sched_setaffinity))
2137 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid,
2138 args->len);
2139#endif
2140 if (args->len < sizeof(cpuset_t))
2141 return (EINVAL);
2142
2143 tdt = linux_tdfind(td, args->pid, -1);
2144 if (tdt == NULL)
2145 return (ESRCH);
2146
2147 PROC_UNLOCK(tdt->td_proc);
2148 csa.level = CPU_LEVEL_WHICH;
2149 csa.which = CPU_WHICH_TID;
2150 csa.id = tdt->td_tid;
2151 csa.cpusetsize = sizeof(cpuset_t);
2152 csa.mask = (cpuset_t *) args->user_mask_ptr;
2153
2154 return (sys_cpuset_setaffinity(td, &csa));
2155}
2156
2157struct linux_rlimit64 {
2158 uint64_t rlim_cur;
2159 uint64_t rlim_max;
2160};
2161
2162int
2163linux_prlimit64(struct thread *td, struct linux_prlimit64_args *args)
2164{
2165 struct rlimit rlim, nrlim;
2166 struct linux_rlimit64 lrlim;
2167 struct proc *p;
2168 u_int which;
2169 int flags;
2170 int error;
2171
2172#ifdef DEBUG
2173 if (ldebug(prlimit64))
2174 printf(ARGS(prlimit64, "%d, %d, %p, %p"), args->pid,
2175 args->resource, (void *)args->new, (void *)args->old);
2176#endif
2177
2178 if (args->resource >= LINUX_RLIM_NLIMITS)
2179 return (EINVAL);
2180
2181 which = linux_to_bsd_resource[args->resource];
2182 if (which == -1)
2183 return (EINVAL);
2184
2185 if (args->new != NULL) {
2186 /*
2187 * Note. Unlike FreeBSD where rlim is signed 64-bit Linux
2188 * rlim is unsigned 64-bit. FreeBSD treats negative limits
2189 * as INFINITY so we do not need a conversion even.
2190 */
2191 error = copyin(args->new, &nrlim, sizeof(nrlim));
2192 if (error != 0)
2193 return (error);
2194 }
2195
2196 flags = PGET_HOLD | PGET_NOTWEXIT;
2197 if (args->new != NULL)
2198 flags |= PGET_CANDEBUG;
2199 else
2200 flags |= PGET_CANSEE;
2201 error = pget(args->pid, flags, &p);
2202 if (error != 0)
2203 return (error);
2204
2205 if (args->old != NULL) {
2206 PROC_LOCK(p);
2207 lim_rlimit(p, which, &rlim);
2208 PROC_UNLOCK(p);
2209 if (rlim.rlim_cur == RLIM_INFINITY)
2210 lrlim.rlim_cur = LINUX_RLIM_INFINITY;
2211 else
2212 lrlim.rlim_cur = rlim.rlim_cur;
2213 if (rlim.rlim_max == RLIM_INFINITY)
2214 lrlim.rlim_max = LINUX_RLIM_INFINITY;
2215 else
2216 lrlim.rlim_max = rlim.rlim_max;
2217 error = copyout(&lrlim, args->old, sizeof(lrlim));
2218 if (error != 0)
2219 goto out;
2220 }
2221
2222 if (args->new != NULL)
2223 error = kern_proc_setrlimit(td, p, which, &nrlim);
2224
2225 out:
2226 PRELE(p);
2227 return (error);
2228}
2229
2230int
2231linux_pselect6(struct thread *td, struct linux_pselect6_args *args)
2232{
2233 struct timeval utv, tv0, tv1, *tvp;
2234 struct l_pselect6arg lpse6;
2235 struct l_timespec lts;
2236 struct timespec uts;
2237 l_sigset_t l_ss;
2238 sigset_t *ssp;
2239 sigset_t ss;
2240 int error;
2241
2242 ssp = NULL;
2243 if (args->sig != NULL) {
2244 error = copyin(args->sig, &lpse6, sizeof(lpse6));
2245 if (error != 0)
2246 return (error);
2247 if (lpse6.ss_len != sizeof(l_ss))
2248 return (EINVAL);
2249 if (lpse6.ss != 0) {
2250 error = copyin(PTRIN(lpse6.ss), &l_ss,
2251 sizeof(l_ss));
2252 if (error != 0)
2253 return (error);
2254 linux_to_bsd_sigset(&l_ss, &ss);
2255 ssp = &ss;
2256 }
2257 }
2258
2259 /*
2260 * Currently glibc changes nanosecond number to microsecond.
2261 * This mean losing precision but for now it is hardly seen.
2262 */
2263 if (args->tsp != NULL) {
2264 error = copyin(args->tsp, <s, sizeof(lts));
2265 if (error != 0)
2266 return (error);
2267 error = linux_to_native_timespec(&uts, <s);
2268 if (error != 0)
2269 return (error);
2270
2271 TIMESPEC_TO_TIMEVAL(&utv, &uts);
2272 if (itimerfix(&utv))
2273 return (EINVAL);
2274
2275 microtime(&tv0);
2276 tvp = &utv;
2277 } else
2278 tvp = NULL;
2279
2280 error = kern_pselect(td, args->nfds, args->readfds, args->writefds,
2281 args->exceptfds, tvp, ssp, LINUX_NFDBITS);
2282
2283 if (error == 0 && args->tsp != NULL) {
2284 if (td->td_retval[0] != 0) {
2285 /*
2286 * Compute how much time was left of the timeout,
2287 * by subtracting the current time and the time
2288 * before we started the call, and subtracting
2289 * that result from the user-supplied value.
2290 */
2291
2292 microtime(&tv1);
2293 timevalsub(&tv1, &tv0);
2294 timevalsub(&utv, &tv1);
2295 if (utv.tv_sec < 0)
2296 timevalclear(&utv);
2297 } else
2298 timevalclear(&utv);
2299
2300 TIMEVAL_TO_TIMESPEC(&utv, &uts);
2301
2302 native_to_linux_timespec(<s, &uts);
2303 error = copyout(<s, args->tsp, sizeof(lts));
2304 }
2305
2306 return (error);
2307}
2308
2309int
2310linux_ppoll(struct thread *td, struct linux_ppoll_args *args)
2311{
2312 struct timespec ts0, ts1;
2313 struct l_timespec lts;
2314 struct timespec uts, *tsp;
2315 l_sigset_t l_ss;
2316 sigset_t *ssp;
2317 sigset_t ss;
2318 int error;
2319
2320 if (args->sset != NULL) {
2321 if (args->ssize != sizeof(l_ss))
2322 return (EINVAL);
2323 error = copyin(args->sset, &l_ss, sizeof(l_ss));
2324 if (error)
2325 return (error);
2326 linux_to_bsd_sigset(&l_ss, &ss);
2327 ssp = &ss;
2328 } else
2329 ssp = NULL;
2330 if (args->tsp != NULL) {
2331 error = copyin(args->tsp, <s, sizeof(lts));
2332 if (error)
2333 return (error);
2334 error = linux_to_native_timespec(&uts, <s);
2335 if (error != 0)
2336 return (error);
2337
2338 nanotime(&ts0);
2339 tsp = &uts;
2340 } else
2341 tsp = NULL;
2342
2343 error = kern_poll(td, args->fds, args->nfds, tsp, ssp);
2344
2345 if (error == 0 && args->tsp != NULL) {
2346 if (td->td_retval[0]) {
2347 nanotime(&ts1);
2348 timespecsub(&ts1, &ts0);
2349 timespecsub(&uts, &ts1);
2350 if (uts.tv_sec < 0)
2351 timespecclear(&uts);
2352 } else
2353 timespecclear(&uts);
2354
2355 native_to_linux_timespec(<s, &uts);
2356 error = copyout(<s, args->tsp, sizeof(lts));
2357 }
2358
2359 return (error);
2360}
2361
2362#if defined(DEBUG) || defined(KTR)
2363/* XXX: can be removed when every ldebug(...) and KTR stuff are removed. */
2364
|