linux_machdep.c revision 293575
1/*- 2 * Copyright (c) 2000 Marcel Moolenaar 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29#include <sys/cdefs.h> 30__FBSDID("$FreeBSD: stable/10/sys/i386/linux/linux_machdep.c 293575 2016-01-09 17:29:08Z dchagin $"); 31 32#include <sys/param.h> 33#include <sys/systm.h> 34#include <sys/capsicum.h> 35#include <sys/file.h> 36#include <sys/fcntl.h> 37#include <sys/imgact.h> 38#include <sys/lock.h> 39#include <sys/malloc.h> 40#include <sys/mman.h> 41#include <sys/mutex.h> 42#include <sys/sx.h> 43#include <sys/priv.h> 44#include <sys/proc.h> 45#include <sys/queue.h> 46#include <sys/resource.h> 47#include <sys/resourcevar.h> 48#include <sys/signalvar.h> 49#include <sys/syscallsubr.h> 50#include <sys/sysproto.h> 51#include <sys/unistd.h> 52#include <sys/wait.h> 53#include <sys/sched.h> 54 55#include <machine/frame.h> 56#include <machine/psl.h> 57#include <machine/segments.h> 58#include <machine/sysarch.h> 59 60#include <vm/vm.h> 61#include <vm/pmap.h> 62#include <vm/vm_map.h> 63 64#include <i386/linux/linux.h> 65#include <i386/linux/linux_proto.h> 66#include <compat/linux/linux_ipc.h> 67#include <compat/linux/linux_misc.h> 68#include <compat/linux/linux_signal.h> 69#include <compat/linux/linux_util.h> 70#include <compat/linux/linux_emul.h> 71 72#include <i386/include/pcb.h> /* needed for pcb definition in linux_set_thread_area */ 73 74#include "opt_posix.h" 75 76extern struct sysentvec elf32_freebsd_sysvec; /* defined in i386/i386/elf_machdep.c */ 77 78struct l_descriptor { 79 l_uint entry_number; 80 l_ulong base_addr; 81 l_uint limit; 82 l_uint seg_32bit:1; 83 l_uint contents:2; 84 l_uint read_exec_only:1; 85 l_uint limit_in_pages:1; 86 l_uint seg_not_present:1; 87 l_uint useable:1; 88}; 89 90struct l_old_select_argv { 91 l_int nfds; 92 l_fd_set *readfds; 93 l_fd_set *writefds; 94 l_fd_set *exceptfds; 95 struct l_timeval *timeout; 96}; 97 98static int linux_mmap_common(struct thread *td, l_uintptr_t addr, 99 l_size_t len, l_int prot, l_int flags, l_int fd, 100 l_loff_t pos); 101 102 103int 104linux_execve(struct thread *td, struct linux_execve_args *args) 105{ 106 struct image_args eargs; 107 struct vmspace *oldvmspace; 108 char *newpath; 109 int error; 110 111 LCONVPATHEXIST(td, args->path, &newpath); 112 113#ifdef DEBUG 114 if (ldebug(execve)) 115 printf(ARGS(execve, "%s"), newpath); 116#endif 117 118 error = pre_execve(td, &oldvmspace); 119 if (error != 0) { 120 free(newpath, M_TEMP); 121 return (error); 122 } 123 error = exec_copyin_args(&eargs, newpath, UIO_SYSSPACE, 124 args->argp, args->envp); 125 free(newpath, M_TEMP); 126 if (error == 0) 127 error = kern_execve(td, &eargs, NULL); 128 if (error == 0) 129 error = linux_common_execve(td, &eargs); 130 post_execve(td, error, oldvmspace); 131 return (error); 132} 133 134struct l_ipc_kludge { 135 struct l_msgbuf *msgp; 136 l_long msgtyp; 137}; 138 139int 140linux_ipc(struct thread *td, struct linux_ipc_args *args) 141{ 142 143 switch (args->what & 0xFFFF) { 144 case LINUX_SEMOP: { 145 struct linux_semop_args a; 146 147 a.semid = args->arg1; 148 a.tsops = args->ptr; 149 a.nsops = args->arg2; 150 return (linux_semop(td, &a)); 151 } 152 case LINUX_SEMGET: { 153 struct linux_semget_args a; 154 155 a.key = args->arg1; 156 a.nsems = args->arg2; 157 a.semflg = args->arg3; 158 return (linux_semget(td, &a)); 159 } 160 case LINUX_SEMCTL: { 161 struct linux_semctl_args a; 162 int error; 163 164 a.semid = args->arg1; 165 a.semnum = args->arg2; 166 a.cmd = args->arg3; 167 error = copyin(args->ptr, &a.arg, sizeof(a.arg)); 168 if (error) 169 return (error); 170 return (linux_semctl(td, &a)); 171 } 172 case LINUX_MSGSND: { 173 struct linux_msgsnd_args a; 174 175 a.msqid = args->arg1; 176 a.msgp = args->ptr; 177 a.msgsz = args->arg2; 178 a.msgflg = args->arg3; 179 return (linux_msgsnd(td, &a)); 180 } 181 case LINUX_MSGRCV: { 182 struct linux_msgrcv_args a; 183 184 a.msqid = args->arg1; 185 a.msgsz = args->arg2; 186 a.msgflg = args->arg3; 187 if ((args->what >> 16) == 0) { 188 struct l_ipc_kludge tmp; 189 int error; 190 191 if (args->ptr == NULL) 192 return (EINVAL); 193 error = copyin(args->ptr, &tmp, sizeof(tmp)); 194 if (error) 195 return (error); 196 a.msgp = tmp.msgp; 197 a.msgtyp = tmp.msgtyp; 198 } else { 199 a.msgp = args->ptr; 200 a.msgtyp = args->arg5; 201 } 202 return (linux_msgrcv(td, &a)); 203 } 204 case LINUX_MSGGET: { 205 struct linux_msgget_args a; 206 207 a.key = args->arg1; 208 a.msgflg = args->arg2; 209 return (linux_msgget(td, &a)); 210 } 211 case LINUX_MSGCTL: { 212 struct linux_msgctl_args a; 213 214 a.msqid = args->arg1; 215 a.cmd = args->arg2; 216 a.buf = args->ptr; 217 return (linux_msgctl(td, &a)); 218 } 219 case LINUX_SHMAT: { 220 struct linux_shmat_args a; 221 222 a.shmid = args->arg1; 223 a.shmaddr = args->ptr; 224 a.shmflg = args->arg2; 225 a.raddr = (l_ulong *)args->arg3; 226 return (linux_shmat(td, &a)); 227 } 228 case LINUX_SHMDT: { 229 struct linux_shmdt_args a; 230 231 a.shmaddr = args->ptr; 232 return (linux_shmdt(td, &a)); 233 } 234 case LINUX_SHMGET: { 235 struct linux_shmget_args a; 236 237 a.key = args->arg1; 238 a.size = args->arg2; 239 a.shmflg = args->arg3; 240 return (linux_shmget(td, &a)); 241 } 242 case LINUX_SHMCTL: { 243 struct linux_shmctl_args a; 244 245 a.shmid = args->arg1; 246 a.cmd = args->arg2; 247 a.buf = args->ptr; 248 return (linux_shmctl(td, &a)); 249 } 250 default: 251 break; 252 } 253 254 return (EINVAL); 255} 256 257int 258linux_old_select(struct thread *td, struct linux_old_select_args *args) 259{ 260 struct l_old_select_argv linux_args; 261 struct linux_select_args newsel; 262 int error; 263 264#ifdef DEBUG 265 if (ldebug(old_select)) 266 printf(ARGS(old_select, "%p"), args->ptr); 267#endif 268 269 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 270 if (error) 271 return (error); 272 273 newsel.nfds = linux_args.nfds; 274 newsel.readfds = linux_args.readfds; 275 newsel.writefds = linux_args.writefds; 276 newsel.exceptfds = linux_args.exceptfds; 277 newsel.timeout = linux_args.timeout; 278 return (linux_select(td, &newsel)); 279} 280 281int 282linux_set_cloned_tls(struct thread *td, void *desc) 283{ 284 struct segment_descriptor sd; 285 struct l_user_desc info; 286 int idx, error; 287 int a[2]; 288 289 error = copyin(desc, &info, sizeof(struct l_user_desc)); 290 if (error) { 291 printf(LMSG("copyin failed!")); 292 } else { 293 idx = info.entry_number; 294 295 /* 296 * looks like we're getting the idx we returned 297 * in the set_thread_area() syscall 298 */ 299 if (idx != 6 && idx != 3) { 300 printf(LMSG("resetting idx!")); 301 idx = 3; 302 } 303 304 /* this doesnt happen in practice */ 305 if (idx == 6) { 306 /* we might copy out the entry_number as 3 */ 307 info.entry_number = 3; 308 error = copyout(&info, desc, sizeof(struct l_user_desc)); 309 if (error) 310 printf(LMSG("copyout failed!")); 311 } 312 313 a[0] = LINUX_LDT_entry_a(&info); 314 a[1] = LINUX_LDT_entry_b(&info); 315 316 memcpy(&sd, &a, sizeof(a)); 317#ifdef DEBUG 318 if (ldebug(clone)) 319 printf("Segment created in clone with " 320 "CLONE_SETTLS: lobase: %x, hibase: %x, " 321 "lolimit: %x, hilimit: %x, type: %i, " 322 "dpl: %i, p: %i, xx: %i, def32: %i, " 323 "gran: %i\n", sd.sd_lobase, sd.sd_hibase, 324 sd.sd_lolimit, sd.sd_hilimit, sd.sd_type, 325 sd.sd_dpl, sd.sd_p, sd.sd_xx, 326 sd.sd_def32, sd.sd_gran); 327#endif 328 329 /* set %gs */ 330 td->td_pcb->pcb_gsd = sd; 331 td->td_pcb->pcb_gs = GSEL(GUGS_SEL, SEL_UPL); 332 } 333 334 return (error); 335} 336 337int 338linux_set_upcall_kse(struct thread *td, register_t stack) 339{ 340 341 if (stack) 342 td->td_frame->tf_esp = stack; 343 344 /* 345 * The newly created Linux thread returns 346 * to the user space by the same path that a parent do. 347 */ 348 td->td_frame->tf_eax = 0; 349 return (0); 350} 351 352#define STACK_SIZE (2 * 1024 * 1024) 353#define GUARD_SIZE (4 * PAGE_SIZE) 354 355int 356linux_mmap2(struct thread *td, struct linux_mmap2_args *args) 357{ 358 359#ifdef DEBUG 360 if (ldebug(mmap2)) 361 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"), 362 (void *)args->addr, args->len, args->prot, 363 args->flags, args->fd, args->pgoff); 364#endif 365 366 return (linux_mmap_common(td, args->addr, args->len, args->prot, 367 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff * 368 PAGE_SIZE)); 369} 370 371int 372linux_mmap(struct thread *td, struct linux_mmap_args *args) 373{ 374 int error; 375 struct l_mmap_argv linux_args; 376 377 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 378 if (error) 379 return (error); 380 381#ifdef DEBUG 382 if (ldebug(mmap)) 383 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"), 384 (void *)linux_args.addr, linux_args.len, linux_args.prot, 385 linux_args.flags, linux_args.fd, linux_args.pgoff); 386#endif 387 388 return (linux_mmap_common(td, linux_args.addr, linux_args.len, 389 linux_args.prot, linux_args.flags, linux_args.fd, 390 (uint32_t)linux_args.pgoff)); 391} 392 393static int 394linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot, 395 l_int flags, l_int fd, l_loff_t pos) 396{ 397 struct proc *p = td->td_proc; 398 struct mmap_args /* { 399 caddr_t addr; 400 size_t len; 401 int prot; 402 int flags; 403 int fd; 404 long pad; 405 off_t pos; 406 } */ bsd_args; 407 int error; 408 struct file *fp; 409 cap_rights_t rights; 410 411 error = 0; 412 bsd_args.flags = 0; 413 fp = NULL; 414 415 /* 416 * Linux mmap(2): 417 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE 418 */ 419 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) 420 return (EINVAL); 421 422 if (flags & LINUX_MAP_SHARED) 423 bsd_args.flags |= MAP_SHARED; 424 if (flags & LINUX_MAP_PRIVATE) 425 bsd_args.flags |= MAP_PRIVATE; 426 if (flags & LINUX_MAP_FIXED) 427 bsd_args.flags |= MAP_FIXED; 428 if (flags & LINUX_MAP_ANON) { 429 /* Enforce pos to be on page boundary, then ignore. */ 430 if ((pos & PAGE_MASK) != 0) 431 return (EINVAL); 432 pos = 0; 433 bsd_args.flags |= MAP_ANON; 434 } else 435 bsd_args.flags |= MAP_NOSYNC; 436 if (flags & LINUX_MAP_GROWSDOWN) 437 bsd_args.flags |= MAP_STACK; 438 439 /* 440 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC 441 * on Linux/i386. We do this to ensure maximum compatibility. 442 * Linux/ia64 does the same in i386 emulation mode. 443 */ 444 bsd_args.prot = prot; 445 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 446 bsd_args.prot |= PROT_READ | PROT_EXEC; 447 448 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ 449 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd; 450 if (bsd_args.fd != -1) { 451 /* 452 * Linux follows Solaris mmap(2) description: 453 * The file descriptor fildes is opened with 454 * read permission, regardless of the 455 * protection options specified. 456 * 457 * Checking just CAP_MMAP is fine here, since the real work 458 * is done in the FreeBSD mmap(). 459 */ 460 461 error = fget(td, bsd_args.fd, 462 cap_rights_init(&rights, CAP_MMAP), &fp); 463 if (error != 0) 464 return (error); 465 if (fp->f_type != DTYPE_VNODE) { 466 fdrop(fp, td); 467 return (EINVAL); 468 } 469 470 /* Linux mmap() just fails for O_WRONLY files */ 471 if (!(fp->f_flag & FREAD)) { 472 fdrop(fp, td); 473 return (EACCES); 474 } 475 476 fdrop(fp, td); 477 } 478 479 if (flags & LINUX_MAP_GROWSDOWN) { 480 /* 481 * The Linux MAP_GROWSDOWN option does not limit auto 482 * growth of the region. Linux mmap with this option 483 * takes as addr the inital BOS, and as len, the initial 484 * region size. It can then grow down from addr without 485 * limit. However, linux threads has an implicit internal 486 * limit to stack size of STACK_SIZE. Its just not 487 * enforced explicitly in linux. But, here we impose 488 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 489 * region, since we can do this with our mmap. 490 * 491 * Our mmap with MAP_STACK takes addr as the maximum 492 * downsize limit on BOS, and as len the max size of 493 * the region. It them maps the top SGROWSIZ bytes, 494 * and auto grows the region down, up to the limit 495 * in addr. 496 * 497 * If we don't use the MAP_STACK option, the effect 498 * of this code is to allocate a stack region of a 499 * fixed size of (STACK_SIZE - GUARD_SIZE). 500 */ 501 502 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) { 503 /* 504 * Some linux apps will attempt to mmap 505 * thread stacks near the top of their 506 * address space. If their TOS is greater 507 * than vm_maxsaddr, vm_map_growstack() 508 * will confuse the thread stack with the 509 * process stack and deliver a SEGV if they 510 * attempt to grow the thread stack past their 511 * current stacksize rlimit. To avoid this, 512 * adjust vm_maxsaddr upwards to reflect 513 * the current stacksize rlimit rather 514 * than the maximum possible stacksize. 515 * It would be better to adjust the 516 * mmap'ed region, but some apps do not check 517 * mmap's return value. 518 */ 519 PROC_LOCK(p); 520 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 521 lim_cur(p, RLIMIT_STACK); 522 PROC_UNLOCK(p); 523 } 524 525 /* 526 * This gives us our maximum stack size and a new BOS. 527 * If we're using VM_STACK, then mmap will just map 528 * the top SGROWSIZ bytes, and let the stack grow down 529 * to the limit at BOS. If we're not using VM_STACK 530 * we map the full stack, since we don't have a way 531 * to autogrow it. 532 */ 533 if (len > STACK_SIZE - GUARD_SIZE) { 534 bsd_args.addr = (caddr_t)PTRIN(addr); 535 bsd_args.len = len; 536 } else { 537 bsd_args.addr = (caddr_t)PTRIN(addr) - 538 (STACK_SIZE - GUARD_SIZE - len); 539 bsd_args.len = STACK_SIZE - GUARD_SIZE; 540 } 541 } else { 542 bsd_args.addr = (caddr_t)PTRIN(addr); 543 bsd_args.len = len; 544 } 545 bsd_args.pos = pos; 546 547#ifdef DEBUG 548 if (ldebug(mmap)) 549 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", 550 __func__, 551 (void *)bsd_args.addr, bsd_args.len, bsd_args.prot, 552 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 553#endif 554 error = sys_mmap(td, &bsd_args); 555#ifdef DEBUG 556 if (ldebug(mmap)) 557 printf("-> %s() return: 0x%x (0x%08x)\n", 558 __func__, error, (u_int)td->td_retval[0]); 559#endif 560 return (error); 561} 562 563int 564linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) 565{ 566 struct mprotect_args bsd_args; 567 568 bsd_args.addr = uap->addr; 569 bsd_args.len = uap->len; 570 bsd_args.prot = uap->prot; 571 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 572 bsd_args.prot |= PROT_READ | PROT_EXEC; 573 return (sys_mprotect(td, &bsd_args)); 574} 575 576int 577linux_ioperm(struct thread *td, struct linux_ioperm_args *args) 578{ 579 int error; 580 struct i386_ioperm_args iia; 581 582 iia.start = args->start; 583 iia.length = args->length; 584 iia.enable = args->enable; 585 error = i386_set_ioperm(td, &iia); 586 return (error); 587} 588 589int 590linux_iopl(struct thread *td, struct linux_iopl_args *args) 591{ 592 int error; 593 594 if (args->level < 0 || args->level > 3) 595 return (EINVAL); 596 if ((error = priv_check(td, PRIV_IO)) != 0) 597 return (error); 598 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 599 return (error); 600 td->td_frame->tf_eflags = (td->td_frame->tf_eflags & ~PSL_IOPL) | 601 (args->level * (PSL_IOPL / 3)); 602 return (0); 603} 604 605int 606linux_modify_ldt(struct thread *td, struct linux_modify_ldt_args *uap) 607{ 608 int error; 609 struct i386_ldt_args ldt; 610 struct l_descriptor ld; 611 union descriptor desc; 612 int size, written; 613 614 switch (uap->func) { 615 case 0x00: /* read_ldt */ 616 ldt.start = 0; 617 ldt.descs = uap->ptr; 618 ldt.num = uap->bytecount / sizeof(union descriptor); 619 error = i386_get_ldt(td, &ldt); 620 td->td_retval[0] *= sizeof(union descriptor); 621 break; 622 case 0x02: /* read_default_ldt = 0 */ 623 size = 5*sizeof(struct l_desc_struct); 624 if (size > uap->bytecount) 625 size = uap->bytecount; 626 for (written = error = 0; written < size && error == 0; written++) 627 error = subyte((char *)uap->ptr + written, 0); 628 td->td_retval[0] = written; 629 break; 630 case 0x01: /* write_ldt */ 631 case 0x11: /* write_ldt */ 632 if (uap->bytecount != sizeof(ld)) 633 return (EINVAL); 634 635 error = copyin(uap->ptr, &ld, sizeof(ld)); 636 if (error) 637 return (error); 638 639 ldt.start = ld.entry_number; 640 ldt.descs = &desc; 641 ldt.num = 1; 642 desc.sd.sd_lolimit = (ld.limit & 0x0000ffff); 643 desc.sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 644 desc.sd.sd_lobase = (ld.base_addr & 0x00ffffff); 645 desc.sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 646 desc.sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 647 (ld.contents << 2); 648 desc.sd.sd_dpl = 3; 649 desc.sd.sd_p = (ld.seg_not_present ^ 1); 650 desc.sd.sd_xx = 0; 651 desc.sd.sd_def32 = ld.seg_32bit; 652 desc.sd.sd_gran = ld.limit_in_pages; 653 error = i386_set_ldt(td, &ldt, &desc); 654 break; 655 default: 656 error = ENOSYS; 657 break; 658 } 659 660 if (error == EOPNOTSUPP) { 661 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 662 error = ENOSYS; 663 } 664 665 return (error); 666} 667 668int 669linux_sigaction(struct thread *td, struct linux_sigaction_args *args) 670{ 671 l_osigaction_t osa; 672 l_sigaction_t act, oact; 673 int error; 674 675#ifdef DEBUG 676 if (ldebug(sigaction)) 677 printf(ARGS(sigaction, "%d, %p, %p"), 678 args->sig, (void *)args->nsa, (void *)args->osa); 679#endif 680 681 if (args->nsa != NULL) { 682 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); 683 if (error) 684 return (error); 685 act.lsa_handler = osa.lsa_handler; 686 act.lsa_flags = osa.lsa_flags; 687 act.lsa_restorer = osa.lsa_restorer; 688 LINUX_SIGEMPTYSET(act.lsa_mask); 689 act.lsa_mask.__mask = osa.lsa_mask; 690 } 691 692 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, 693 args->osa ? &oact : NULL); 694 695 if (args->osa != NULL && !error) { 696 osa.lsa_handler = oact.lsa_handler; 697 osa.lsa_flags = oact.lsa_flags; 698 osa.lsa_restorer = oact.lsa_restorer; 699 osa.lsa_mask = oact.lsa_mask.__mask; 700 error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); 701 } 702 703 return (error); 704} 705 706/* 707 * Linux has two extra args, restart and oldmask. We dont use these, 708 * but it seems that "restart" is actually a context pointer that 709 * enables the signal to happen with a different register set. 710 */ 711int 712linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) 713{ 714 sigset_t sigmask; 715 l_sigset_t mask; 716 717#ifdef DEBUG 718 if (ldebug(sigsuspend)) 719 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 720#endif 721 722 LINUX_SIGEMPTYSET(mask); 723 mask.__mask = args->mask; 724 linux_to_bsd_sigset(&mask, &sigmask); 725 return (kern_sigsuspend(td, sigmask)); 726} 727 728int 729linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) 730{ 731 l_sigset_t lmask; 732 sigset_t sigmask; 733 int error; 734 735#ifdef DEBUG 736 if (ldebug(rt_sigsuspend)) 737 printf(ARGS(rt_sigsuspend, "%p, %d"), 738 (void *)uap->newset, uap->sigsetsize); 739#endif 740 741 if (uap->sigsetsize != sizeof(l_sigset_t)) 742 return (EINVAL); 743 744 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t)); 745 if (error) 746 return (error); 747 748 linux_to_bsd_sigset(&lmask, &sigmask); 749 return (kern_sigsuspend(td, sigmask)); 750} 751 752int 753linux_pause(struct thread *td, struct linux_pause_args *args) 754{ 755 struct proc *p = td->td_proc; 756 sigset_t sigmask; 757 758#ifdef DEBUG 759 if (ldebug(pause)) 760 printf(ARGS(pause, "")); 761#endif 762 763 PROC_LOCK(p); 764 sigmask = td->td_sigmask; 765 PROC_UNLOCK(p); 766 return (kern_sigsuspend(td, sigmask)); 767} 768 769int 770linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) 771{ 772 stack_t ss, oss; 773 l_stack_t lss; 774 int error; 775 776#ifdef DEBUG 777 if (ldebug(sigaltstack)) 778 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 779#endif 780 781 if (uap->uss != NULL) { 782 error = copyin(uap->uss, &lss, sizeof(l_stack_t)); 783 if (error) 784 return (error); 785 786 ss.ss_sp = lss.ss_sp; 787 ss.ss_size = lss.ss_size; 788 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 789 } 790 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL, 791 (uap->uoss != NULL) ? &oss : NULL); 792 if (!error && uap->uoss != NULL) { 793 lss.ss_sp = oss.ss_sp; 794 lss.ss_size = oss.ss_size; 795 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags); 796 error = copyout(&lss, uap->uoss, sizeof(l_stack_t)); 797 } 798 799 return (error); 800} 801 802int 803linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 804{ 805 struct ftruncate_args sa; 806 807#ifdef DEBUG 808 if (ldebug(ftruncate64)) 809 printf(ARGS(ftruncate64, "%u, %jd"), args->fd, 810 (intmax_t)args->length); 811#endif 812 813 sa.fd = args->fd; 814 sa.length = args->length; 815 return sys_ftruncate(td, &sa); 816} 817 818int 819linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args) 820{ 821 struct l_user_desc info; 822 int error; 823 int idx; 824 int a[2]; 825 struct segment_descriptor sd; 826 827 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 828 if (error) 829 return (error); 830 831#ifdef DEBUG 832 if (ldebug(set_thread_area)) 833 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"), 834 info.entry_number, 835 info.base_addr, 836 info.limit, 837 info.seg_32bit, 838 info.contents, 839 info.read_exec_only, 840 info.limit_in_pages, 841 info.seg_not_present, 842 info.useable); 843#endif 844 845 idx = info.entry_number; 846 /* 847 * Semantics of linux version: every thread in the system has array of 848 * 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This 849 * syscall loads one of the selected tls decriptors with a value and 850 * also loads GDT descriptors 6, 7 and 8 with the content of the 851 * per-thread descriptors. 852 * 853 * Semantics of fbsd version: I think we can ignore that linux has 3 854 * per-thread descriptors and use just the 1st one. The tls_array[] 855 * is used only in set/get-thread_area() syscalls and for loading the 856 * GDT descriptors. In fbsd we use just one GDT descriptor for TLS so 857 * we will load just one. 858 * 859 * XXX: this doesn't work when a user space process tries to use more 860 * than 1 TLS segment. Comment in the linux sources says wine might do 861 * this. 862 */ 863 864 /* 865 * we support just GLIBC TLS now 866 * we should let 3 proceed as well because we use this segment so 867 * if code does two subsequent calls it should succeed 868 */ 869 if (idx != 6 && idx != -1 && idx != 3) 870 return (EINVAL); 871 872 /* 873 * we have to copy out the GDT entry we use 874 * FreeBSD uses GDT entry #3 for storing %gs so load that 875 * 876 * XXX: what if a user space program doesn't check this value and tries 877 * to use 6, 7 or 8? 878 */ 879 idx = info.entry_number = 3; 880 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 881 if (error) 882 return (error); 883 884 if (LINUX_LDT_empty(&info)) { 885 a[0] = 0; 886 a[1] = 0; 887 } else { 888 a[0] = LINUX_LDT_entry_a(&info); 889 a[1] = LINUX_LDT_entry_b(&info); 890 } 891 892 memcpy(&sd, &a, sizeof(a)); 893#ifdef DEBUG 894 if (ldebug(set_thread_area)) 895 printf("Segment created in set_thread_area: lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, type: %i, dpl: %i, p: %i, xx: %i, def32: %i, gran: %i\n", sd.sd_lobase, 896 sd.sd_hibase, 897 sd.sd_lolimit, 898 sd.sd_hilimit, 899 sd.sd_type, 900 sd.sd_dpl, 901 sd.sd_p, 902 sd.sd_xx, 903 sd.sd_def32, 904 sd.sd_gran); 905#endif 906 907 /* this is taken from i386 version of cpu_set_user_tls() */ 908 critical_enter(); 909 /* set %gs */ 910 td->td_pcb->pcb_gsd = sd; 911 PCPU_GET(fsgs_gdt)[1] = sd; 912 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 913 critical_exit(); 914 915 return (0); 916} 917 918int 919linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args) 920{ 921 922 struct l_user_desc info; 923 int error; 924 int idx; 925 struct l_desc_struct desc; 926 struct segment_descriptor sd; 927 928#ifdef DEBUG 929 if (ldebug(get_thread_area)) 930 printf(ARGS(get_thread_area, "%p"), args->desc); 931#endif 932 933 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 934 if (error) 935 return (error); 936 937 idx = info.entry_number; 938 /* XXX: I am not sure if we want 3 to be allowed too. */ 939 if (idx != 6 && idx != 3) 940 return (EINVAL); 941 942 idx = 3; 943 944 memset(&info, 0, sizeof(info)); 945 946 sd = PCPU_GET(fsgs_gdt)[1]; 947 948 memcpy(&desc, &sd, sizeof(desc)); 949 950 info.entry_number = idx; 951 info.base_addr = LINUX_GET_BASE(&desc); 952 info.limit = LINUX_GET_LIMIT(&desc); 953 info.seg_32bit = LINUX_GET_32BIT(&desc); 954 info.contents = LINUX_GET_CONTENTS(&desc); 955 info.read_exec_only = !LINUX_GET_WRITABLE(&desc); 956 info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc); 957 info.seg_not_present = !LINUX_GET_PRESENT(&desc); 958 info.useable = LINUX_GET_USEABLE(&desc); 959 960 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 961 if (error) 962 return (EFAULT); 963 964 return (0); 965} 966 967/* XXX: this wont work with module - convert it */ 968int 969linux_mq_open(struct thread *td, struct linux_mq_open_args *args) 970{ 971#ifdef P1003_1B_MQUEUE 972 return sys_kmq_open(td, (struct kmq_open_args *) args); 973#else 974 return (ENOSYS); 975#endif 976} 977 978int 979linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args) 980{ 981#ifdef P1003_1B_MQUEUE 982 return sys_kmq_unlink(td, (struct kmq_unlink_args *) args); 983#else 984 return (ENOSYS); 985#endif 986} 987 988int 989linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args) 990{ 991#ifdef P1003_1B_MQUEUE 992 return sys_kmq_timedsend(td, (struct kmq_timedsend_args *) args); 993#else 994 return (ENOSYS); 995#endif 996} 997 998int 999linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args) 1000{ 1001#ifdef P1003_1B_MQUEUE 1002 return sys_kmq_timedreceive(td, (struct kmq_timedreceive_args *) args); 1003#else 1004 return (ENOSYS); 1005#endif 1006} 1007 1008int 1009linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args) 1010{ 1011#ifdef P1003_1B_MQUEUE 1012 return sys_kmq_notify(td, (struct kmq_notify_args *) args); 1013#else 1014 return (ENOSYS); 1015#endif 1016} 1017 1018int 1019linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args) 1020{ 1021#ifdef P1003_1B_MQUEUE 1022 return sys_kmq_setattr(td, (struct kmq_setattr_args *) args); 1023#else 1024 return (ENOSYS); 1025#endif 1026} 1027