1 /*-
2 * Copyright (c) 2004 Tim J. Robbins
3 * Copyright (c) 2002 Doug Rabson
4 * Copyright (c) 2000 Marcel Moolenaar
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer
12 * in this position and unchanged.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. The name of the author may not be used to endorse or promote products
17 * derived from this software without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 */
30
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD: releng/10.1/sys/amd64/linux32/linux32_machdep.c 255219 2013-09-05 00:09:56Z pjd $");
33
34 #include <sys/param.h>
35 #include <sys/kernel.h>
36 #include <sys/systm.h>
37 #include <sys/capability.h>
38 #include <sys/file.h>
39 #include <sys/fcntl.h>
40 #include <sys/clock.h>
41 #include <sys/imgact.h>
42 #include <sys/limits.h>
43 #include <sys/lock.h>
44 #include <sys/malloc.h>
45 #include <sys/mman.h>
46 #include <sys/mutex.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/resource.h>
50 #include <sys/resourcevar.h>
51 #include <sys/sched.h>
52 #include <sys/syscallsubr.h>
53 #include <sys/sysproto.h>
54 #include <sys/unistd.h>
55 #include <sys/wait.h>
56
57 #include <machine/frame.h>
58 #include <machine/pcb.h>
59 #include <machine/psl.h>
60 #include <machine/segments.h>
61 #include <machine/specialreg.h>
62
63 #include <vm/vm.h>
64 #include <vm/pmap.h>
65 #include <vm/vm_map.h>
66
67 #include <compat/freebsd32/freebsd32_util.h>
68 #include <amd64/linux32/linux.h>
69 #include <amd64/linux32/linux32_proto.h>
70 #include <compat/linux/linux_ipc.h>
71 #include <compat/linux/linux_misc.h>
72 #include <compat/linux/linux_signal.h>
73 #include <compat/linux/linux_util.h>
74 #include <compat/linux/linux_emul.h>
75
76 struct l_old_select_argv {
77 l_int nfds;
78 l_uintptr_t readfds;
79 l_uintptr_t writefds;
80 l_uintptr_t exceptfds;
81 l_uintptr_t timeout;
82 } __packed;
83
84 int
85 linux_to_bsd_sigaltstack(int lsa)
86 {
87 int bsa = 0;
88
89 if (lsa & LINUX_SS_DISABLE)
90 bsa |= SS_DISABLE;
91 if (lsa & LINUX_SS_ONSTACK)
92 bsa |= SS_ONSTACK;
93 return (bsa);
94 }
95
96 static int linux_mmap_common(struct thread *td, l_uintptr_t addr,
97 l_size_t len, l_int prot, l_int flags, l_int fd,
98 l_loff_t pos);
99
100 int
101 bsd_to_linux_sigaltstack(int bsa)
102 {
103 int lsa = 0;
104
105 if (bsa & SS_DISABLE)
106 lsa |= LINUX_SS_DISABLE;
107 if (bsa & SS_ONSTACK)
108 lsa |= LINUX_SS_ONSTACK;
109 return (lsa);
110 }
111
112 static void
113 bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru)
114 {
115
116 lru->ru_utime.tv_sec = ru->ru_utime.tv_sec;
117 lru->ru_utime.tv_usec = ru->ru_utime.tv_usec;
118 lru->ru_stime.tv_sec = ru->ru_stime.tv_sec;
119 lru->ru_stime.tv_usec = ru->ru_stime.tv_usec;
120 lru->ru_maxrss = ru->ru_maxrss;
121 lru->ru_ixrss = ru->ru_ixrss;
122 lru->ru_idrss = ru->ru_idrss;
123 lru->ru_isrss = ru->ru_isrss;
124 lru->ru_minflt = ru->ru_minflt;
125 lru->ru_majflt = ru->ru_majflt;
126 lru->ru_nswap = ru->ru_nswap;
127 lru->ru_inblock = ru->ru_inblock;
128 lru->ru_oublock = ru->ru_oublock;
129 lru->ru_msgsnd = ru->ru_msgsnd;
130 lru->ru_msgrcv = ru->ru_msgrcv;
131 lru->ru_nsignals = ru->ru_nsignals;
132 lru->ru_nvcsw = ru->ru_nvcsw;
133 lru->ru_nivcsw = ru->ru_nivcsw;
134 }
135
136 int
137 linux_execve(struct thread *td, struct linux_execve_args *args)
138 {
139 struct image_args eargs;
140 char *path;
141 int error;
142
143 LCONVPATHEXIST(td, args->path, &path);
144
145 #ifdef DEBUG
146 if (ldebug(execve))
147 printf(ARGS(execve, "%s"), path);
148 #endif
149
150 error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE,
151 args->argp, args->envp);
152 free(path, M_TEMP);
153 if (error == 0)
154 error = kern_execve(td, &eargs, NULL);
155 if (error == 0)
156 /* Linux process can execute FreeBSD one, do not attempt
157 * to create emuldata for such process using
158 * linux_proc_init, this leads to a panic on KASSERT
159 * because such process has p->p_emuldata == NULL.
160 */
161 if (SV_PROC_ABI(td->td_proc) == SV_ABI_LINUX)
162 error = linux_proc_init(td, 0, 0);
163 return (error);
164 }
165
166 CTASSERT(sizeof(struct l_iovec32) == 8);
167
168 static int
169 linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop)
170 {
171 struct l_iovec32 iov32;
172 struct iovec *iov;
173 struct uio *uio;
174 uint32_t iovlen;
175 int error, i;
176
177 *uiop = NULL;
178 if (iovcnt > UIO_MAXIOV)
179 return (EINVAL);
180 iovlen = iovcnt * sizeof(struct iovec);
181 uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK);
182 iov = (struct iovec *)(uio + 1);
183 for (i = 0; i < iovcnt; i++) {
184 error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32));
185 if (error) {
186 free(uio, M_IOV);
187 return (error);
188 }
189 iov[i].iov_base = PTRIN(iov32.iov_base);
190 iov[i].iov_len = iov32.iov_len;
191 }
192 uio->uio_iov = iov;
193 uio->uio_iovcnt = iovcnt;
194 uio->uio_segflg = UIO_USERSPACE;
195 uio->uio_offset = -1;
196 uio->uio_resid = 0;
197 for (i = 0; i < iovcnt; i++) {
198 if (iov->iov_len > INT_MAX - uio->uio_resid) {
199 free(uio, M_IOV);
200 return (EINVAL);
201 }
202 uio->uio_resid += iov->iov_len;
203 iov++;
204 }
205 *uiop = uio;
206 return (0);
207 }
208
209 int
210 linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp,
211 int error)
212 {
213 struct l_iovec32 iov32;
214 struct iovec *iov;
215 uint32_t iovlen;
216 int i;
217
218 *iovp = NULL;
219 if (iovcnt > UIO_MAXIOV)
220 return (error);
221 iovlen = iovcnt * sizeof(struct iovec);
222 iov = malloc(iovlen, M_IOV, M_WAITOK);
223 for (i = 0; i < iovcnt; i++) {
224 error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32));
225 if (error) {
226 free(iov, M_IOV);
227 return (error);
228 }
229 iov[i].iov_base = PTRIN(iov32.iov_base);
230 iov[i].iov_len = iov32.iov_len;
231 }
232 *iovp = iov;
233 return(0);
234
235 }
236
237 int
238 linux_readv(struct thread *td, struct linux_readv_args *uap)
239 {
240 struct uio *auio;
241 int error;
242
243 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
244 if (error)
245 return (error);
246 error = kern_readv(td, uap->fd, auio);
247 free(auio, M_IOV);
248 return (error);
249 }
250
251 int
252 linux_writev(struct thread *td, struct linux_writev_args *uap)
253 {
254 struct uio *auio;
255 int error;
256
257 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
258 if (error)
259 return (error);
260 error = kern_writev(td, uap->fd, auio);
261 free(auio, M_IOV);
262 return (error);
263 }
264
265 struct l_ipc_kludge {
266 l_uintptr_t msgp;
267 l_long msgtyp;
268 } __packed;
269
270 int
271 linux_ipc(struct thread *td, struct linux_ipc_args *args)
272 {
273
274 switch (args->what & 0xFFFF) {
275 case LINUX_SEMOP: {
276 struct linux_semop_args a;
277
278 a.semid = args->arg1;
279 a.tsops = args->ptr;
280 a.nsops = args->arg2;
281 return (linux_semop(td, &a));
282 }
283 case LINUX_SEMGET: {
284 struct linux_semget_args a;
285
286 a.key = args->arg1;
287 a.nsems = args->arg2;
288 a.semflg = args->arg3;
289 return (linux_semget(td, &a));
290 }
291 case LINUX_SEMCTL: {
292 struct linux_semctl_args a;
293 int error;
294
295 a.semid = args->arg1;
296 a.semnum = args->arg2;
297 a.cmd = args->arg3;
298 error = copyin(args->ptr, &a.arg, sizeof(a.arg));
299 if (error)
300 return (error);
301 return (linux_semctl(td, &a));
302 }
303 case LINUX_MSGSND: {
304 struct linux_msgsnd_args a;
305
306 a.msqid = args->arg1;
307 a.msgp = args->ptr;
308 a.msgsz = args->arg2;
309 a.msgflg = args->arg3;
310 return (linux_msgsnd(td, &a));
311 }
312 case LINUX_MSGRCV: {
313 struct linux_msgrcv_args a;
314
315 a.msqid = args->arg1;
316 a.msgsz = args->arg2;
317 a.msgflg = args->arg3;
318 if ((args->what >> 16) == 0) {
319 struct l_ipc_kludge tmp;
320 int error;
321
322 if (args->ptr == 0)
323 return (EINVAL);
324 error = copyin(args->ptr, &tmp, sizeof(tmp));
325 if (error)
326 return (error);
327 a.msgp = PTRIN(tmp.msgp);
328 a.msgtyp = tmp.msgtyp;
329 } else {
330 a.msgp = args->ptr;
331 a.msgtyp = args->arg5;
332 }
333 return (linux_msgrcv(td, &a));
334 }
335 case LINUX_MSGGET: {
336 struct linux_msgget_args a;
337
338 a.key = args->arg1;
339 a.msgflg = args->arg2;
340 return (linux_msgget(td, &a));
341 }
342 case LINUX_MSGCTL: {
343 struct linux_msgctl_args a;
344
345 a.msqid = args->arg1;
346 a.cmd = args->arg2;
347 a.buf = args->ptr;
348 return (linux_msgctl(td, &a));
349 }
350 case LINUX_SHMAT: {
351 struct linux_shmat_args a;
352
353 a.shmid = args->arg1;
354 a.shmaddr = args->ptr;
355 a.shmflg = args->arg2;
356 a.raddr = PTRIN((l_uint)args->arg3);
357 return (linux_shmat(td, &a));
358 }
359 case LINUX_SHMDT: {
360 struct linux_shmdt_args a;
361
362 a.shmaddr = args->ptr;
363 return (linux_shmdt(td, &a));
364 }
365 case LINUX_SHMGET: {
366 struct linux_shmget_args a;
367
368 a.key = args->arg1;
369 a.size = args->arg2;
370 a.shmflg = args->arg3;
371 return (linux_shmget(td, &a));
372 }
373 case LINUX_SHMCTL: {
374 struct linux_shmctl_args a;
375
376 a.shmid = args->arg1;
377 a.cmd = args->arg2;
378 a.buf = args->ptr;
379 return (linux_shmctl(td, &a));
380 }
381 default:
382 break;
383 }
384
385 return (EINVAL);
386 }
387
388 int
389 linux_old_select(struct thread *td, struct linux_old_select_args *args)
390 {
391 struct l_old_select_argv linux_args;
392 struct linux_select_args newsel;
393 int error;
394
395 #ifdef DEBUG
396 if (ldebug(old_select))
397 printf(ARGS(old_select, "%p"), args->ptr);
398 #endif
399
400 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
401 if (error)
402 return (error);
403
404 newsel.nfds = linux_args.nfds;
405 newsel.readfds = PTRIN(linux_args.readfds);
406 newsel.writefds = PTRIN(linux_args.writefds);
407 newsel.exceptfds = PTRIN(linux_args.exceptfds);
408 newsel.timeout = PTRIN(linux_args.timeout);
409 return (linux_select(td, &newsel));
410 }
411
412 int
413 linux_set_cloned_tls(struct thread *td, void *desc)
414 {
415 struct user_segment_descriptor sd;
416 struct l_user_desc info;
417 struct pcb *pcb;
418 int error;
419 int a[2];
420
421 error = copyin(desc, &info, sizeof(struct l_user_desc));
422 if (error) {
423 printf(LMSG("copyin failed!"));
424 } else {
425 /* We might copy out the entry_number as GUGS32_SEL. */
426 info.entry_number = GUGS32_SEL;
427 error = copyout(&info, desc, sizeof(struct l_user_desc));
428 if (error)
429 printf(LMSG("copyout failed!"));
430
431 a[0] = LINUX_LDT_entry_a(&info);
432 a[1] = LINUX_LDT_entry_b(&info);
433
434 memcpy(&sd, &a, sizeof(a));
435 #ifdef DEBUG
436 if (ldebug(clone))
437 printf("Segment created in clone with "
438 "CLONE_SETTLS: lobase: %x, hibase: %x, "
439 "lolimit: %x, hilimit: %x, type: %i, "
440 "dpl: %i, p: %i, xx: %i, long: %i, "
441 "def32: %i, gran: %i\n", sd.sd_lobase,
442 sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit,
443 sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx,
444 sd.sd_long, sd.sd_def32, sd.sd_gran);
445 #endif
446 pcb = td->td_pcb;
447 pcb->pcb_gsbase = (register_t)info.base_addr;
448 /* XXXKIB pcb->pcb_gs32sd = sd; */
449 td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL);
450 set_pcb_flags(pcb, PCB_32BIT);
451 }
452
453 return (error);
454 }
455
456 int
457 linux_set_upcall_kse(struct thread *td, register_t stack)
458 {
459
460 td->td_frame->tf_rsp = stack;
461
462 return (0);
463 }
464
465 #define STACK_SIZE (2 * 1024 * 1024)
466 #define GUARD_SIZE (4 * PAGE_SIZE)
467
468 int
469 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
470 {
471
472 #ifdef DEBUG
473 if (ldebug(mmap2))
474 printf(ARGS(mmap2, "0x%08x, %d, %d, 0x%08x, %d, %d"),
475 args->addr, args->len, args->prot,
476 args->flags, args->fd, args->pgoff);
477 #endif
478
479 return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot,
480 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff *
481 PAGE_SIZE));
482 }
483
484 int
485 linux_mmap(struct thread *td, struct linux_mmap_args *args)
486 {
487 int error;
488 struct l_mmap_argv linux_args;
489
490 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
491 if (error)
492 return (error);
493
494 #ifdef DEBUG
495 if (ldebug(mmap))
496 printf(ARGS(mmap, "0x%08x, %d, %d, 0x%08x, %d, %d"),
497 linux_args.addr, linux_args.len, linux_args.prot,
498 linux_args.flags, linux_args.fd, linux_args.pgoff);
499 #endif
500
501 return (linux_mmap_common(td, linux_args.addr, linux_args.len,
502 linux_args.prot, linux_args.flags, linux_args.fd,
503 (uint32_t)linux_args.pgoff));
504 }
505
506 static int
507 linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot,
508 l_int flags, l_int fd, l_loff_t pos)
509 {
510 struct proc *p = td->td_proc;
511 struct mmap_args /* {
512 caddr_t addr;
513 size_t len;
514 int prot;
515 int flags;
516 int fd;
517 long pad;
518 off_t pos;
519 } */ bsd_args;
520 int error;
521 struct file *fp;
522 cap_rights_t rights;
523
524 error = 0;
525 bsd_args.flags = 0;
526 fp = NULL;
527
528 /*
529 * Linux mmap(2):
530 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE
531 */
532 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE)))
533 return (EINVAL);
534
535 if (flags & LINUX_MAP_SHARED)
536 bsd_args.flags |= MAP_SHARED;
537 if (flags & LINUX_MAP_PRIVATE)
538 bsd_args.flags |= MAP_PRIVATE;
539 if (flags & LINUX_MAP_FIXED)
540 bsd_args.flags |= MAP_FIXED;
541 if (flags & LINUX_MAP_ANON) {
542 /* Enforce pos to be on page boundary, then ignore. */
543 if ((pos & PAGE_MASK) != 0)
544 return (EINVAL);
545 pos = 0;
546 bsd_args.flags |= MAP_ANON;
547 } else
548 bsd_args.flags |= MAP_NOSYNC;
549 if (flags & LINUX_MAP_GROWSDOWN)
550 bsd_args.flags |= MAP_STACK;
551
552 /*
553 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC
554 * on Linux/i386. We do this to ensure maximum compatibility.
555 * Linux/ia64 does the same in i386 emulation mode.
556 */
557 bsd_args.prot = prot;
558 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
559 bsd_args.prot |= PROT_READ | PROT_EXEC;
560
561 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */
562 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd;
563 if (bsd_args.fd != -1) {
564 /*
565 * Linux follows Solaris mmap(2) description:
566 * The file descriptor fildes is opened with
567 * read permission, regardless of the
568 * protection options specified.
569 */
570
571 error = fget(td, bsd_args.fd,
572 cap_rights_init(&rights, CAP_MMAP), &fp);
573 if (error != 0)
574 return (error);
575 if (fp->f_type != DTYPE_VNODE) {
576 fdrop(fp, td);
577 return (EINVAL);
578 }
579
580 /* Linux mmap() just fails for O_WRONLY files */
581 if (!(fp->f_flag & FREAD)) {
582 fdrop(fp, td);
583 return (EACCES);
584 }
585
586 fdrop(fp, td);
587 }
588
589 if (flags & LINUX_MAP_GROWSDOWN) {
590 /*
591 * The Linux MAP_GROWSDOWN option does not limit auto
592 * growth of the region. Linux mmap with this option
593 * takes as addr the inital BOS, and as len, the initial
594 * region size. It can then grow down from addr without
595 * limit. However, Linux threads has an implicit internal
596 * limit to stack size of STACK_SIZE. Its just not
597 * enforced explicitly in Linux. But, here we impose
598 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
599 * region, since we can do this with our mmap.
600 *
601 * Our mmap with MAP_STACK takes addr as the maximum
602 * downsize limit on BOS, and as len the max size of
603 * the region. It then maps the top SGROWSIZ bytes,
604 * and auto grows the region down, up to the limit
605 * in addr.
606 *
607 * If we don't use the MAP_STACK option, the effect
608 * of this code is to allocate a stack region of a
609 * fixed size of (STACK_SIZE - GUARD_SIZE).
610 */
611
612 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) {
613 /*
614 * Some Linux apps will attempt to mmap
615 * thread stacks near the top of their
616 * address space. If their TOS is greater
617 * than vm_maxsaddr, vm_map_growstack()
618 * will confuse the thread stack with the
619 * process stack and deliver a SEGV if they
620 * attempt to grow the thread stack past their
621 * current stacksize rlimit. To avoid this,
622 * adjust vm_maxsaddr upwards to reflect
623 * the current stacksize rlimit rather
624 * than the maximum possible stacksize.
625 * It would be better to adjust the
626 * mmap'ed region, but some apps do not check
627 * mmap's return value.
628 */
629 PROC_LOCK(p);
630 p->p_vmspace->vm_maxsaddr = (char *)LINUX32_USRSTACK -
631 lim_cur(p, RLIMIT_STACK);
632 PROC_UNLOCK(p);
633 }
634
635 /*
636 * This gives us our maximum stack size and a new BOS.
637 * If we're using VM_STACK, then mmap will just map
638 * the top SGROWSIZ bytes, and let the stack grow down
639 * to the limit at BOS. If we're not using VM_STACK
640 * we map the full stack, since we don't have a way
641 * to autogrow it.
642 */
643 if (len > STACK_SIZE - GUARD_SIZE) {
644 bsd_args.addr = (caddr_t)PTRIN(addr);
645 bsd_args.len = len;
646 } else {
647 bsd_args.addr = (caddr_t)PTRIN(addr) -
648 (STACK_SIZE - GUARD_SIZE - len);
649 bsd_args.len = STACK_SIZE - GUARD_SIZE;
650 }
651 } else {
652 bsd_args.addr = (caddr_t)PTRIN(addr);
653 bsd_args.len = len;
654 }
655 bsd_args.pos = pos;
656
657 #ifdef DEBUG
658 if (ldebug(mmap))
659 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n",
660 __func__,
661 (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot,
662 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos);
663 #endif
664 error = sys_mmap(td, &bsd_args);
665 #ifdef DEBUG
666 if (ldebug(mmap))
667 printf("-> %s() return: 0x%x (0x%08x)\n",
668 __func__, error, (u_int)td->td_retval[0]);
669 #endif
670 return (error);
671 }
672
673 int
674 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
675 {
676 struct mprotect_args bsd_args;
677
678 bsd_args.addr = uap->addr;
679 bsd_args.len = uap->len;
680 bsd_args.prot = uap->prot;
681 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
682 bsd_args.prot |= PROT_READ | PROT_EXEC;
683 return (sys_mprotect(td, &bsd_args));
684 }
685
686 int
687 linux_iopl(struct thread *td, struct linux_iopl_args *args)
688 {
689 int error;
690
691 if (args->level < 0 || args->level > 3)
692 return (EINVAL);
693 if ((error = priv_check(td, PRIV_IO)) != 0)
694 return (error);
695 if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
696 return (error);
697 td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
698 (args->level * (PSL_IOPL / 3));
699
700 return (0);
701 }
702
703 int
704 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
705 {
706 l_osigaction_t osa;
707 l_sigaction_t act, oact;
708 int error;
709
710 #ifdef DEBUG
711 if (ldebug(sigaction))
712 printf(ARGS(sigaction, "%d, %p, %p"),
713 args->sig, (void *)args->nsa, (void *)args->osa);
714 #endif
715
716 if (args->nsa != NULL) {
717 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
718 if (error)
719 return (error);
720 act.lsa_handler = osa.lsa_handler;
721 act.lsa_flags = osa.lsa_flags;
722 act.lsa_restorer = osa.lsa_restorer;
723 LINUX_SIGEMPTYSET(act.lsa_mask);
724 act.lsa_mask.__bits[0] = osa.lsa_mask;
725 }
726
727 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
728 args->osa ? &oact : NULL);
729
730 if (args->osa != NULL && !error) {
731 osa.lsa_handler = oact.lsa_handler;
732 osa.lsa_flags = oact.lsa_flags;
733 osa.lsa_restorer = oact.lsa_restorer;
734 osa.lsa_mask = oact.lsa_mask.__bits[0];
735 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
736 }
737
738 return (error);
739 }
740
741 /*
742 * Linux has two extra args, restart and oldmask. We don't use these,
743 * but it seems that "restart" is actually a context pointer that
744 * enables the signal to happen with a different register set.
745 */
746 int
747 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
748 {
749 sigset_t sigmask;
750 l_sigset_t mask;
751
752 #ifdef DEBUG
753 if (ldebug(sigsuspend))
754 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
755 #endif
756
757 LINUX_SIGEMPTYSET(mask);
758 mask.__bits[0] = args->mask;
759 linux_to_bsd_sigset(&mask, &sigmask);
760 return (kern_sigsuspend(td, sigmask));
761 }
762
763 int
764 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
765 {
766 l_sigset_t lmask;
767 sigset_t sigmask;
768 int error;
769
770 #ifdef DEBUG
771 if (ldebug(rt_sigsuspend))
772 printf(ARGS(rt_sigsuspend, "%p, %d"),
773 (void *)uap->newset, uap->sigsetsize);
774 #endif
775
776 if (uap->sigsetsize != sizeof(l_sigset_t))
777 return (EINVAL);
778
779 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
780 if (error)
781 return (error);
782
783 linux_to_bsd_sigset(&lmask, &sigmask);
784 return (kern_sigsuspend(td, sigmask));
785 }
786
787 int
788 linux_pause(struct thread *td, struct linux_pause_args *args)
789 {
790 struct proc *p = td->td_proc;
791 sigset_t sigmask;
792
793 #ifdef DEBUG
794 if (ldebug(pause))
795 printf(ARGS(pause, ""));
796 #endif
797
798 PROC_LOCK(p);
799 sigmask = td->td_sigmask;
800 PROC_UNLOCK(p);
801 return (kern_sigsuspend(td, sigmask));
802 }
803
804 int
805 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
806 {
807 stack_t ss, oss;
808 l_stack_t lss;
809 int error;
810
811 #ifdef DEBUG
812 if (ldebug(sigaltstack))
813 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
814 #endif
815
816 if (uap->uss != NULL) {
817 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
818 if (error)
819 return (error);
820
821 ss.ss_sp = PTRIN(lss.ss_sp);
822 ss.ss_size = lss.ss_size;
823 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
824 }
825 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
826 (uap->uoss != NULL) ? &oss : NULL);
827 if (!error && uap->uoss != NULL) {
828 lss.ss_sp = PTROUT(oss.ss_sp);
829 lss.ss_size = oss.ss_size;
830 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
831 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
832 }
833
834 return (error);
835 }
836
837 int
838 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
839 {
840 struct ftruncate_args sa;
841
842 #ifdef DEBUG
843 if (ldebug(ftruncate64))
844 printf(ARGS(ftruncate64, "%u, %jd"), args->fd,
845 (intmax_t)args->length);
846 #endif
847
848 sa.fd = args->fd;
849 sa.length = args->length;
850 return sys_ftruncate(td, &sa);
851 }
852
853 int
854 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
855 {
856 struct timeval atv;
857 l_timeval atv32;
858 struct timezone rtz;
859 int error = 0;
860
861 if (uap->tp) {
862 microtime(&atv);
863 atv32.tv_sec = atv.tv_sec;
864 atv32.tv_usec = atv.tv_usec;
865 error = copyout(&atv32, uap->tp, sizeof(atv32));
866 }
867 if (error == 0 && uap->tzp != NULL) {
868 rtz.tz_minuteswest = tz_minuteswest;
869 rtz.tz_dsttime = tz_dsttime;
870 error = copyout(&rtz, uap->tzp, sizeof(rtz));
871 }
872 return (error);
873 }
874
875 int
876 linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap)
877 {
878 l_timeval atv32;
879 struct timeval atv, *tvp;
880 struct timezone atz, *tzp;
881 int error;
882
883 if (uap->tp) {
884 error = copyin(uap->tp, &atv32, sizeof(atv32));
885 if (error)
886 return (error);
887 atv.tv_sec = atv32.tv_sec;
888 atv.tv_usec = atv32.tv_usec;
889 tvp = &atv;
890 } else
891 tvp = NULL;
892 if (uap->tzp) {
893 error = copyin(uap->tzp, &atz, sizeof(atz));
894 if (error)
895 return (error);
896 tzp = &atz;
897 } else
898 tzp = NULL;
899 return (kern_settimeofday(td, tvp, tzp));
900 }
901
902 int
903 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
904 {
905 struct l_rusage s32;
906 struct rusage s;
907 int error;
908
909 error = kern_getrusage(td, uap->who, &s);
910 if (error != 0)
911 return (error);
912 if (uap->rusage != NULL) {
913 bsd_to_linux_rusage(&s, &s32);
914 error = copyout(&s32, uap->rusage, sizeof(s32));
915 }
916 return (error);
917 }
918
919 int
920 linux_sched_rr_get_interval(struct thread *td,
921 struct linux_sched_rr_get_interval_args *uap)
922 {
923 struct timespec ts;
924 struct l_timespec ts32;
925 int error;
926
927 error = kern_sched_rr_get_interval(td, uap->pid, &ts);
928 if (error != 0)
929 return (error);
930 ts32.tv_sec = ts.tv_sec;
931 ts32.tv_nsec = ts.tv_nsec;
932 return (copyout(&ts32, uap->interval, sizeof(ts32)));
933 }
934
935 int
936 linux_set_thread_area(struct thread *td,
937 struct linux_set_thread_area_args *args)
938 {
939 struct l_user_desc info;
940 struct user_segment_descriptor sd;
941 struct pcb *pcb;
942 int a[2];
943 int error;
944
945 error = copyin(args->desc, &info, sizeof(struct l_user_desc));
946 if (error)
947 return (error);
948
949 #ifdef DEBUG
950 if (ldebug(set_thread_area))
951 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, "
952 "%i, %i, %i"), info.entry_number, info.base_addr,
953 info.limit, info.seg_32bit, info.contents,
954 info.read_exec_only, info.limit_in_pages,
955 info.seg_not_present, info.useable);
956 #endif
957
958 /*
959 * Semantics of Linux version: every thread in the system has array
960 * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown.
961 * This syscall loads one of the selected TLS decriptors with a value
962 * and also loads GDT descriptors 6, 7 and 8 with the content of
963 * the per-thread descriptors.
964 *
965 * Semantics of FreeBSD version: I think we can ignore that Linux has
966 * three per-thread descriptors and use just the first one.
967 * The tls_array[] is used only in [gs]et_thread_area() syscalls and
968 * for loading the GDT descriptors. We use just one GDT descriptor
969 * for TLS, so we will load just one.
970 *
971 * XXX: This doesn't work when a user space process tries to use more
972 * than one TLS segment. Comment in the Linux source says wine might
973 * do this.
974 */
975
976 /*
977 * GLIBC reads current %gs and call set_thread_area() with it.
978 * We should let GUDATA_SEL and GUGS32_SEL proceed as well because
979 * we use these segments.
980 */
981 switch (info.entry_number) {
982 case GUGS32_SEL:
983 case GUDATA_SEL:
984 case 6:
985 case -1:
986 info.entry_number = GUGS32_SEL;
987 break;
988 default:
989 return (EINVAL);
990 }
991
992 /*
993 * We have to copy out the GDT entry we use.
994 *
995 * XXX: What if a user space program does not check the return value
996 * and tries to use 6, 7 or 8?
997 */
998 error = copyout(&info, args->desc, sizeof(struct l_user_desc));
999 if (error)
1000 return (error);
1001
1002 if (LINUX_LDT_empty(&info)) {
1003 a[0] = 0;
1004 a[1] = 0;
1005 } else {
1006 a[0] = LINUX_LDT_entry_a(&info);
1007 a[1] = LINUX_LDT_entry_b(&info);
1008 }
1009
1010 memcpy(&sd, &a, sizeof(a));
1011 #ifdef DEBUG
1012 if (ldebug(set_thread_area))
1013 printf("Segment created in set_thread_area: "
1014 "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, "
1015 "type: %i, dpl: %i, p: %i, xx: %i, long: %i, "
1016 "def32: %i, gran: %i\n",
1017 sd.sd_lobase,
1018 sd.sd_hibase,
1019 sd.sd_lolimit,
1020 sd.sd_hilimit,
1021 sd.sd_type,
1022 sd.sd_dpl,
1023 sd.sd_p,
1024 sd.sd_xx,
1025 sd.sd_long,
1026 sd.sd_def32,
1027 sd.sd_gran);
1028 #endif
1029
1030 pcb = td->td_pcb;
1031 pcb->pcb_gsbase = (register_t)info.base_addr;
1032 set_pcb_flags(pcb, PCB_32BIT);
1033 update_gdt_gsbase(td, info.base_addr);
1034
1035 return (0);
1036 }
1037
1038 int
1039 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1040 {
1041 int error, options;
1042 struct rusage ru, *rup;
1043 struct l_rusage lru;
1044
1045 #ifdef DEBUG
1046 if (ldebug(wait4))
1047 printf(ARGS(wait4, "%d, %p, %d, %p"),
1048 args->pid, (void *)args->status, args->options,
1049 (void *)args->rusage);
1050 #endif
1051
1052 options = (args->options & (WNOHANG | WUNTRACED));
1053 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
1054 if (args->options & __WCLONE)
1055 options |= WLINUXCLONE;
1056
1057 if (args->rusage != NULL)
1058 rup = &ru;
1059 else
1060 rup = NULL;
1061 error = linux_common_wait(td, args->pid, args->status, options, rup);
1062 if (error)
1063 return (error);
1064 if (args->rusage != NULL) {
1065 bsd_to_linux_rusage(rup, &lru);
1066 error = copyout(&lru, args->rusage, sizeof(lru));
1067 }
1068
1069 return (error);
1070 }
Cache object: 7426da3df2d8956a84d53730fc72ac41
|