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/9.2/sys/amd64/linux32/linux32_machdep.c 250785 2013-05-18 19:06:15Z dchagin $");
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
523 error = 0;
524 bsd_args.flags = 0;
525 fp = NULL;
526
527 /*
528 * Linux mmap(2):
529 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE
530 */
531 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE)))
532 return (EINVAL);
533
534 if (flags & LINUX_MAP_SHARED)
535 bsd_args.flags |= MAP_SHARED;
536 if (flags & LINUX_MAP_PRIVATE)
537 bsd_args.flags |= MAP_PRIVATE;
538 if (flags & LINUX_MAP_FIXED)
539 bsd_args.flags |= MAP_FIXED;
540 if (flags & LINUX_MAP_ANON) {
541 /* Enforce pos to be on page boundary, then ignore. */
542 if ((pos & PAGE_MASK) != 0)
543 return (EINVAL);
544 pos = 0;
545 bsd_args.flags |= MAP_ANON;
546 } else
547 bsd_args.flags |= MAP_NOSYNC;
548 if (flags & LINUX_MAP_GROWSDOWN)
549 bsd_args.flags |= MAP_STACK;
550
551 /*
552 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC
553 * on Linux/i386. We do this to ensure maximum compatibility.
554 * Linux/ia64 does the same in i386 emulation mode.
555 */
556 bsd_args.prot = prot;
557 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
558 bsd_args.prot |= PROT_READ | PROT_EXEC;
559
560 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */
561 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd;
562 if (bsd_args.fd != -1) {
563 /*
564 * Linux follows Solaris mmap(2) description:
565 * The file descriptor fildes is opened with
566 * read permission, regardless of the
567 * protection options specified.
568 */
569
570 if ((error = fget(td, bsd_args.fd, CAP_MMAP, &fp)) != 0)
571 return (error);
572 if (fp->f_type != DTYPE_VNODE) {
573 fdrop(fp, td);
574 return (EINVAL);
575 }
576
577 /* Linux mmap() just fails for O_WRONLY files */
578 if (!(fp->f_flag & FREAD)) {
579 fdrop(fp, td);
580 return (EACCES);
581 }
582
583 fdrop(fp, td);
584 }
585
586 if (flags & LINUX_MAP_GROWSDOWN) {
587 /*
588 * The Linux MAP_GROWSDOWN option does not limit auto
589 * growth of the region. Linux mmap with this option
590 * takes as addr the inital BOS, and as len, the initial
591 * region size. It can then grow down from addr without
592 * limit. However, Linux threads has an implicit internal
593 * limit to stack size of STACK_SIZE. Its just not
594 * enforced explicitly in Linux. But, here we impose
595 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack
596 * region, since we can do this with our mmap.
597 *
598 * Our mmap with MAP_STACK takes addr as the maximum
599 * downsize limit on BOS, and as len the max size of
600 * the region. It then maps the top SGROWSIZ bytes,
601 * and auto grows the region down, up to the limit
602 * in addr.
603 *
604 * If we don't use the MAP_STACK option, the effect
605 * of this code is to allocate a stack region of a
606 * fixed size of (STACK_SIZE - GUARD_SIZE).
607 */
608
609 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) {
610 /*
611 * Some Linux apps will attempt to mmap
612 * thread stacks near the top of their
613 * address space. If their TOS is greater
614 * than vm_maxsaddr, vm_map_growstack()
615 * will confuse the thread stack with the
616 * process stack and deliver a SEGV if they
617 * attempt to grow the thread stack past their
618 * current stacksize rlimit. To avoid this,
619 * adjust vm_maxsaddr upwards to reflect
620 * the current stacksize rlimit rather
621 * than the maximum possible stacksize.
622 * It would be better to adjust the
623 * mmap'ed region, but some apps do not check
624 * mmap's return value.
625 */
626 PROC_LOCK(p);
627 p->p_vmspace->vm_maxsaddr = (char *)LINUX32_USRSTACK -
628 lim_cur(p, RLIMIT_STACK);
629 PROC_UNLOCK(p);
630 }
631
632 /*
633 * This gives us our maximum stack size and a new BOS.
634 * If we're using VM_STACK, then mmap will just map
635 * the top SGROWSIZ bytes, and let the stack grow down
636 * to the limit at BOS. If we're not using VM_STACK
637 * we map the full stack, since we don't have a way
638 * to autogrow it.
639 */
640 if (len > STACK_SIZE - GUARD_SIZE) {
641 bsd_args.addr = (caddr_t)PTRIN(addr);
642 bsd_args.len = len;
643 } else {
644 bsd_args.addr = (caddr_t)PTRIN(addr) -
645 (STACK_SIZE - GUARD_SIZE - len);
646 bsd_args.len = STACK_SIZE - GUARD_SIZE;
647 }
648 } else {
649 bsd_args.addr = (caddr_t)PTRIN(addr);
650 bsd_args.len = len;
651 }
652 bsd_args.pos = pos;
653
654 #ifdef DEBUG
655 if (ldebug(mmap))
656 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n",
657 __func__,
658 (void *)bsd_args.addr, (int)bsd_args.len, bsd_args.prot,
659 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos);
660 #endif
661 error = sys_mmap(td, &bsd_args);
662 #ifdef DEBUG
663 if (ldebug(mmap))
664 printf("-> %s() return: 0x%x (0x%08x)\n",
665 __func__, error, (u_int)td->td_retval[0]);
666 #endif
667 return (error);
668 }
669
670 int
671 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
672 {
673 struct mprotect_args bsd_args;
674
675 bsd_args.addr = uap->addr;
676 bsd_args.len = uap->len;
677 bsd_args.prot = uap->prot;
678 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
679 bsd_args.prot |= PROT_READ | PROT_EXEC;
680 return (sys_mprotect(td, &bsd_args));
681 }
682
683 int
684 linux_iopl(struct thread *td, struct linux_iopl_args *args)
685 {
686 int error;
687
688 if (args->level < 0 || args->level > 3)
689 return (EINVAL);
690 if ((error = priv_check(td, PRIV_IO)) != 0)
691 return (error);
692 if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
693 return (error);
694 td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
695 (args->level * (PSL_IOPL / 3));
696
697 return (0);
698 }
699
700 int
701 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
702 {
703 l_osigaction_t osa;
704 l_sigaction_t act, oact;
705 int error;
706
707 #ifdef DEBUG
708 if (ldebug(sigaction))
709 printf(ARGS(sigaction, "%d, %p, %p"),
710 args->sig, (void *)args->nsa, (void *)args->osa);
711 #endif
712
713 if (args->nsa != NULL) {
714 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
715 if (error)
716 return (error);
717 act.lsa_handler = osa.lsa_handler;
718 act.lsa_flags = osa.lsa_flags;
719 act.lsa_restorer = osa.lsa_restorer;
720 LINUX_SIGEMPTYSET(act.lsa_mask);
721 act.lsa_mask.__bits[0] = osa.lsa_mask;
722 }
723
724 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
725 args->osa ? &oact : NULL);
726
727 if (args->osa != NULL && !error) {
728 osa.lsa_handler = oact.lsa_handler;
729 osa.lsa_flags = oact.lsa_flags;
730 osa.lsa_restorer = oact.lsa_restorer;
731 osa.lsa_mask = oact.lsa_mask.__bits[0];
732 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
733 }
734
735 return (error);
736 }
737
738 /*
739 * Linux has two extra args, restart and oldmask. We don't use these,
740 * but it seems that "restart" is actually a context pointer that
741 * enables the signal to happen with a different register set.
742 */
743 int
744 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
745 {
746 sigset_t sigmask;
747 l_sigset_t mask;
748
749 #ifdef DEBUG
750 if (ldebug(sigsuspend))
751 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
752 #endif
753
754 LINUX_SIGEMPTYSET(mask);
755 mask.__bits[0] = args->mask;
756 linux_to_bsd_sigset(&mask, &sigmask);
757 return (kern_sigsuspend(td, sigmask));
758 }
759
760 int
761 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
762 {
763 l_sigset_t lmask;
764 sigset_t sigmask;
765 int error;
766
767 #ifdef DEBUG
768 if (ldebug(rt_sigsuspend))
769 printf(ARGS(rt_sigsuspend, "%p, %d"),
770 (void *)uap->newset, uap->sigsetsize);
771 #endif
772
773 if (uap->sigsetsize != sizeof(l_sigset_t))
774 return (EINVAL);
775
776 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
777 if (error)
778 return (error);
779
780 linux_to_bsd_sigset(&lmask, &sigmask);
781 return (kern_sigsuspend(td, sigmask));
782 }
783
784 int
785 linux_pause(struct thread *td, struct linux_pause_args *args)
786 {
787 struct proc *p = td->td_proc;
788 sigset_t sigmask;
789
790 #ifdef DEBUG
791 if (ldebug(pause))
792 printf(ARGS(pause, ""));
793 #endif
794
795 PROC_LOCK(p);
796 sigmask = td->td_sigmask;
797 PROC_UNLOCK(p);
798 return (kern_sigsuspend(td, sigmask));
799 }
800
801 int
802 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
803 {
804 stack_t ss, oss;
805 l_stack_t lss;
806 int error;
807
808 #ifdef DEBUG
809 if (ldebug(sigaltstack))
810 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
811 #endif
812
813 if (uap->uss != NULL) {
814 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
815 if (error)
816 return (error);
817
818 ss.ss_sp = PTRIN(lss.ss_sp);
819 ss.ss_size = lss.ss_size;
820 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
821 }
822 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
823 (uap->uoss != NULL) ? &oss : NULL);
824 if (!error && uap->uoss != NULL) {
825 lss.ss_sp = PTROUT(oss.ss_sp);
826 lss.ss_size = oss.ss_size;
827 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
828 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
829 }
830
831 return (error);
832 }
833
834 int
835 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
836 {
837 struct ftruncate_args sa;
838
839 #ifdef DEBUG
840 if (ldebug(ftruncate64))
841 printf(ARGS(ftruncate64, "%u, %jd"), args->fd,
842 (intmax_t)args->length);
843 #endif
844
845 sa.fd = args->fd;
846 sa.length = args->length;
847 return sys_ftruncate(td, &sa);
848 }
849
850 int
851 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
852 {
853 struct timeval atv;
854 l_timeval atv32;
855 struct timezone rtz;
856 int error = 0;
857
858 if (uap->tp) {
859 microtime(&atv);
860 atv32.tv_sec = atv.tv_sec;
861 atv32.tv_usec = atv.tv_usec;
862 error = copyout(&atv32, uap->tp, sizeof(atv32));
863 }
864 if (error == 0 && uap->tzp != NULL) {
865 rtz.tz_minuteswest = tz_minuteswest;
866 rtz.tz_dsttime = tz_dsttime;
867 error = copyout(&rtz, uap->tzp, sizeof(rtz));
868 }
869 return (error);
870 }
871
872 int
873 linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap)
874 {
875 l_timeval atv32;
876 struct timeval atv, *tvp;
877 struct timezone atz, *tzp;
878 int error;
879
880 if (uap->tp) {
881 error = copyin(uap->tp, &atv32, sizeof(atv32));
882 if (error)
883 return (error);
884 atv.tv_sec = atv32.tv_sec;
885 atv.tv_usec = atv32.tv_usec;
886 tvp = &atv;
887 } else
888 tvp = NULL;
889 if (uap->tzp) {
890 error = copyin(uap->tzp, &atz, sizeof(atz));
891 if (error)
892 return (error);
893 tzp = &atz;
894 } else
895 tzp = NULL;
896 return (kern_settimeofday(td, tvp, tzp));
897 }
898
899 int
900 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
901 {
902 struct l_rusage s32;
903 struct rusage s;
904 int error;
905
906 error = kern_getrusage(td, uap->who, &s);
907 if (error != 0)
908 return (error);
909 if (uap->rusage != NULL) {
910 bsd_to_linux_rusage(&s, &s32);
911 error = copyout(&s32, uap->rusage, sizeof(s32));
912 }
913 return (error);
914 }
915
916 int
917 linux_sched_rr_get_interval(struct thread *td,
918 struct linux_sched_rr_get_interval_args *uap)
919 {
920 struct timespec ts;
921 struct l_timespec ts32;
922 int error;
923
924 error = kern_sched_rr_get_interval(td, uap->pid, &ts);
925 if (error != 0)
926 return (error);
927 ts32.tv_sec = ts.tv_sec;
928 ts32.tv_nsec = ts.tv_nsec;
929 return (copyout(&ts32, uap->interval, sizeof(ts32)));
930 }
931
932 int
933 linux_set_thread_area(struct thread *td,
934 struct linux_set_thread_area_args *args)
935 {
936 struct l_user_desc info;
937 struct user_segment_descriptor sd;
938 struct pcb *pcb;
939 int a[2];
940 int error;
941
942 error = copyin(args->desc, &info, sizeof(struct l_user_desc));
943 if (error)
944 return (error);
945
946 #ifdef DEBUG
947 if (ldebug(set_thread_area))
948 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, "
949 "%i, %i, %i"), info.entry_number, info.base_addr,
950 info.limit, info.seg_32bit, info.contents,
951 info.read_exec_only, info.limit_in_pages,
952 info.seg_not_present, info.useable);
953 #endif
954
955 /*
956 * Semantics of Linux version: every thread in the system has array
957 * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown.
958 * This syscall loads one of the selected TLS decriptors with a value
959 * and also loads GDT descriptors 6, 7 and 8 with the content of
960 * the per-thread descriptors.
961 *
962 * Semantics of FreeBSD version: I think we can ignore that Linux has
963 * three per-thread descriptors and use just the first one.
964 * The tls_array[] is used only in [gs]et_thread_area() syscalls and
965 * for loading the GDT descriptors. We use just one GDT descriptor
966 * for TLS, so we will load just one.
967 *
968 * XXX: This doesn't work when a user space process tries to use more
969 * than one TLS segment. Comment in the Linux source says wine might
970 * do this.
971 */
972
973 /*
974 * GLIBC reads current %gs and call set_thread_area() with it.
975 * We should let GUDATA_SEL and GUGS32_SEL proceed as well because
976 * we use these segments.
977 */
978 switch (info.entry_number) {
979 case GUGS32_SEL:
980 case GUDATA_SEL:
981 case 6:
982 case -1:
983 info.entry_number = GUGS32_SEL;
984 break;
985 default:
986 return (EINVAL);
987 }
988
989 /*
990 * We have to copy out the GDT entry we use.
991 *
992 * XXX: What if a user space program does not check the return value
993 * and tries to use 6, 7 or 8?
994 */
995 error = copyout(&info, args->desc, sizeof(struct l_user_desc));
996 if (error)
997 return (error);
998
999 if (LINUX_LDT_empty(&info)) {
1000 a[0] = 0;
1001 a[1] = 0;
1002 } else {
1003 a[0] = LINUX_LDT_entry_a(&info);
1004 a[1] = LINUX_LDT_entry_b(&info);
1005 }
1006
1007 memcpy(&sd, &a, sizeof(a));
1008 #ifdef DEBUG
1009 if (ldebug(set_thread_area))
1010 printf("Segment created in set_thread_area: "
1011 "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, "
1012 "type: %i, dpl: %i, p: %i, xx: %i, long: %i, "
1013 "def32: %i, gran: %i\n",
1014 sd.sd_lobase,
1015 sd.sd_hibase,
1016 sd.sd_lolimit,
1017 sd.sd_hilimit,
1018 sd.sd_type,
1019 sd.sd_dpl,
1020 sd.sd_p,
1021 sd.sd_xx,
1022 sd.sd_long,
1023 sd.sd_def32,
1024 sd.sd_gran);
1025 #endif
1026
1027 pcb = td->td_pcb;
1028 pcb->pcb_gsbase = (register_t)info.base_addr;
1029 set_pcb_flags(pcb, PCB_32BIT);
1030 update_gdt_gsbase(td, info.base_addr);
1031
1032 return (0);
1033 }
1034
1035 int
1036 linux_wait4(struct thread *td, struct linux_wait4_args *args)
1037 {
1038 int error, options;
1039 struct rusage ru, *rup;
1040 struct l_rusage lru;
1041
1042 #ifdef DEBUG
1043 if (ldebug(wait4))
1044 printf(ARGS(wait4, "%d, %p, %d, %p"),
1045 args->pid, (void *)args->status, args->options,
1046 (void *)args->rusage);
1047 #endif
1048
1049 options = (args->options & (WNOHANG | WUNTRACED));
1050 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */
1051 if (args->options & __WCLONE)
1052 options |= WLINUXCLONE;
1053
1054 if (args->rusage != NULL)
1055 rup = &ru;
1056 else
1057 rup = NULL;
1058 error = linux_common_wait(td, args->pid, args->status, options, rup);
1059 if (error)
1060 return (error);
1061 if (args->rusage != NULL) {
1062 bsd_to_linux_rusage(rup, &lru);
1063 error = copyout(&lru, args->rusage, sizeof(lru));
1064 }
1065
1066 return (error);
1067 }
Cache object: def328b9c8082d2d33a5302ee9a5a38c
|