linux_machdep.c revision 293500
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 293500 2016-01-09 15:23:54Z 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 102int 103linux_to_bsd_sigaltstack(int lsa) 104{ 105 int bsa = 0; 106 107 if (lsa & LINUX_SS_DISABLE) 108 bsa |= SS_DISABLE; 109 if (lsa & LINUX_SS_ONSTACK) 110 bsa |= SS_ONSTACK; 111 return (bsa); 112} 113 114int 115bsd_to_linux_sigaltstack(int bsa) 116{ 117 int lsa = 0; 118 119 if (bsa & SS_DISABLE) 120 lsa |= LINUX_SS_DISABLE; 121 if (bsa & SS_ONSTACK) 122 lsa |= LINUX_SS_ONSTACK; 123 return (lsa); 124} 125 126int 127linux_execve(struct thread *td, struct linux_execve_args *args) 128{ 129 struct image_args eargs; 130 struct vmspace *oldvmspace; 131 char *newpath; 132 int error; 133 134 LCONVPATHEXIST(td, args->path, &newpath); 135 136#ifdef DEBUG 137 if (ldebug(execve)) 138 printf(ARGS(execve, "%s"), newpath); 139#endif 140 141 error = pre_execve(td, &oldvmspace); 142 if (error != 0) { 143 free(newpath, M_TEMP); 144 return (error); 145 } 146 error = exec_copyin_args(&eargs, newpath, UIO_SYSSPACE, 147 args->argp, args->envp); 148 free(newpath, M_TEMP); 149 if (error == 0) 150 error = kern_execve(td, &eargs, NULL); 151 if (error == 0) 152 error = linux_common_execve(td, &eargs); 153 post_execve(td, error, oldvmspace); 154 return (error); 155} 156 157struct l_ipc_kludge { 158 struct l_msgbuf *msgp; 159 l_long msgtyp; 160}; 161 162int 163linux_ipc(struct thread *td, struct linux_ipc_args *args) 164{ 165 166 switch (args->what & 0xFFFF) { 167 case LINUX_SEMOP: { 168 struct linux_semop_args a; 169 170 a.semid = args->arg1; 171 a.tsops = args->ptr; 172 a.nsops = args->arg2; 173 return (linux_semop(td, &a)); 174 } 175 case LINUX_SEMGET: { 176 struct linux_semget_args a; 177 178 a.key = args->arg1; 179 a.nsems = args->arg2; 180 a.semflg = args->arg3; 181 return (linux_semget(td, &a)); 182 } 183 case LINUX_SEMCTL: { 184 struct linux_semctl_args a; 185 int error; 186 187 a.semid = args->arg1; 188 a.semnum = args->arg2; 189 a.cmd = args->arg3; 190 error = copyin(args->ptr, &a.arg, sizeof(a.arg)); 191 if (error) 192 return (error); 193 return (linux_semctl(td, &a)); 194 } 195 case LINUX_MSGSND: { 196 struct linux_msgsnd_args a; 197 198 a.msqid = args->arg1; 199 a.msgp = args->ptr; 200 a.msgsz = args->arg2; 201 a.msgflg = args->arg3; 202 return (linux_msgsnd(td, &a)); 203 } 204 case LINUX_MSGRCV: { 205 struct linux_msgrcv_args a; 206 207 a.msqid = args->arg1; 208 a.msgsz = args->arg2; 209 a.msgflg = args->arg3; 210 if ((args->what >> 16) == 0) { 211 struct l_ipc_kludge tmp; 212 int error; 213 214 if (args->ptr == NULL) 215 return (EINVAL); 216 error = copyin(args->ptr, &tmp, sizeof(tmp)); 217 if (error) 218 return (error); 219 a.msgp = tmp.msgp; 220 a.msgtyp = tmp.msgtyp; 221 } else { 222 a.msgp = args->ptr; 223 a.msgtyp = args->arg5; 224 } 225 return (linux_msgrcv(td, &a)); 226 } 227 case LINUX_MSGGET: { 228 struct linux_msgget_args a; 229 230 a.key = args->arg1; 231 a.msgflg = args->arg2; 232 return (linux_msgget(td, &a)); 233 } 234 case LINUX_MSGCTL: { 235 struct linux_msgctl_args a; 236 237 a.msqid = args->arg1; 238 a.cmd = args->arg2; 239 a.buf = args->ptr; 240 return (linux_msgctl(td, &a)); 241 } 242 case LINUX_SHMAT: { 243 struct linux_shmat_args a; 244 245 a.shmid = args->arg1; 246 a.shmaddr = args->ptr; 247 a.shmflg = args->arg2; 248 a.raddr = (l_ulong *)args->arg3; 249 return (linux_shmat(td, &a)); 250 } 251 case LINUX_SHMDT: { 252 struct linux_shmdt_args a; 253 254 a.shmaddr = args->ptr; 255 return (linux_shmdt(td, &a)); 256 } 257 case LINUX_SHMGET: { 258 struct linux_shmget_args a; 259 260 a.key = args->arg1; 261 a.size = args->arg2; 262 a.shmflg = args->arg3; 263 return (linux_shmget(td, &a)); 264 } 265 case LINUX_SHMCTL: { 266 struct linux_shmctl_args a; 267 268 a.shmid = args->arg1; 269 a.cmd = args->arg2; 270 a.buf = args->ptr; 271 return (linux_shmctl(td, &a)); 272 } 273 default: 274 break; 275 } 276 277 return (EINVAL); 278} 279 280int 281linux_old_select(struct thread *td, struct linux_old_select_args *args) 282{ 283 struct l_old_select_argv linux_args; 284 struct linux_select_args newsel; 285 int error; 286 287#ifdef DEBUG 288 if (ldebug(old_select)) 289 printf(ARGS(old_select, "%p"), args->ptr); 290#endif 291 292 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 293 if (error) 294 return (error); 295 296 newsel.nfds = linux_args.nfds; 297 newsel.readfds = linux_args.readfds; 298 newsel.writefds = linux_args.writefds; 299 newsel.exceptfds = linux_args.exceptfds; 300 newsel.timeout = linux_args.timeout; 301 return (linux_select(td, &newsel)); 302} 303 304int 305linux_set_cloned_tls(struct thread *td, void *desc) 306{ 307 struct segment_descriptor sd; 308 struct l_user_desc info; 309 int idx, error; 310 int a[2]; 311 312 error = copyin(desc, &info, sizeof(struct l_user_desc)); 313 if (error) { 314 printf(LMSG("copyin failed!")); 315 } else { 316 idx = info.entry_number; 317 318 /* 319 * looks like we're getting the idx we returned 320 * in the set_thread_area() syscall 321 */ 322 if (idx != 6 && idx != 3) { 323 printf(LMSG("resetting idx!")); 324 idx = 3; 325 } 326 327 /* this doesnt happen in practice */ 328 if (idx == 6) { 329 /* we might copy out the entry_number as 3 */ 330 info.entry_number = 3; 331 error = copyout(&info, desc, sizeof(struct l_user_desc)); 332 if (error) 333 printf(LMSG("copyout failed!")); 334 } 335 336 a[0] = LINUX_LDT_entry_a(&info); 337 a[1] = LINUX_LDT_entry_b(&info); 338 339 memcpy(&sd, &a, sizeof(a)); 340#ifdef DEBUG 341 if (ldebug(clone)) 342 printf("Segment created in clone with " 343 "CLONE_SETTLS: lobase: %x, hibase: %x, " 344 "lolimit: %x, hilimit: %x, type: %i, " 345 "dpl: %i, p: %i, xx: %i, def32: %i, " 346 "gran: %i\n", sd.sd_lobase, sd.sd_hibase, 347 sd.sd_lolimit, sd.sd_hilimit, sd.sd_type, 348 sd.sd_dpl, sd.sd_p, sd.sd_xx, 349 sd.sd_def32, sd.sd_gran); 350#endif 351 352 /* set %gs */ 353 td->td_pcb->pcb_gsd = sd; 354 td->td_pcb->pcb_gs = GSEL(GUGS_SEL, SEL_UPL); 355 } 356 357 return (error); 358} 359 360int 361linux_set_upcall_kse(struct thread *td, register_t stack) 362{ 363 364 if (stack) 365 td->td_frame->tf_esp = stack; 366 367 /* 368 * The newly created Linux thread returns 369 * to the user space by the same path that a parent do. 370 */ 371 td->td_frame->tf_eax = 0; 372 return (0); 373} 374 375#define STACK_SIZE (2 * 1024 * 1024) 376#define GUARD_SIZE (4 * PAGE_SIZE) 377 378int 379linux_mmap2(struct thread *td, struct linux_mmap2_args *args) 380{ 381 382#ifdef DEBUG 383 if (ldebug(mmap2)) 384 printf(ARGS(mmap2, "%p, %d, %d, 0x%08x, %d, %d"), 385 (void *)args->addr, args->len, args->prot, 386 args->flags, args->fd, args->pgoff); 387#endif 388 389 return (linux_mmap_common(td, args->addr, args->len, args->prot, 390 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff * 391 PAGE_SIZE)); 392} 393 394int 395linux_mmap(struct thread *td, struct linux_mmap_args *args) 396{ 397 int error; 398 struct l_mmap_argv linux_args; 399 400 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 401 if (error) 402 return (error); 403 404#ifdef DEBUG 405 if (ldebug(mmap)) 406 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"), 407 (void *)linux_args.addr, linux_args.len, linux_args.prot, 408 linux_args.flags, linux_args.fd, linux_args.pgoff); 409#endif 410 411 return (linux_mmap_common(td, linux_args.addr, linux_args.len, 412 linux_args.prot, linux_args.flags, linux_args.fd, 413 (uint32_t)linux_args.pgoff)); 414} 415 416static int 417linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot, 418 l_int flags, l_int fd, l_loff_t pos) 419{ 420 struct proc *p = td->td_proc; 421 struct mmap_args /* { 422 caddr_t addr; 423 size_t len; 424 int prot; 425 int flags; 426 int fd; 427 long pad; 428 off_t pos; 429 } */ bsd_args; 430 int error; 431 struct file *fp; 432 cap_rights_t rights; 433 434 error = 0; 435 bsd_args.flags = 0; 436 fp = NULL; 437 438 /* 439 * Linux mmap(2): 440 * You must specify exactly one of MAP_SHARED and MAP_PRIVATE 441 */ 442 if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) 443 return (EINVAL); 444 445 if (flags & LINUX_MAP_SHARED) 446 bsd_args.flags |= MAP_SHARED; 447 if (flags & LINUX_MAP_PRIVATE) 448 bsd_args.flags |= MAP_PRIVATE; 449 if (flags & LINUX_MAP_FIXED) 450 bsd_args.flags |= MAP_FIXED; 451 if (flags & LINUX_MAP_ANON) { 452 /* Enforce pos to be on page boundary, then ignore. */ 453 if ((pos & PAGE_MASK) != 0) 454 return (EINVAL); 455 pos = 0; 456 bsd_args.flags |= MAP_ANON; 457 } else 458 bsd_args.flags |= MAP_NOSYNC; 459 if (flags & LINUX_MAP_GROWSDOWN) 460 bsd_args.flags |= MAP_STACK; 461 462 /* 463 * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC 464 * on Linux/i386. We do this to ensure maximum compatibility. 465 * Linux/ia64 does the same in i386 emulation mode. 466 */ 467 bsd_args.prot = prot; 468 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 469 bsd_args.prot |= PROT_READ | PROT_EXEC; 470 471 /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ 472 bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd; 473 if (bsd_args.fd != -1) { 474 /* 475 * Linux follows Solaris mmap(2) description: 476 * The file descriptor fildes is opened with 477 * read permission, regardless of the 478 * protection options specified. 479 * 480 * Checking just CAP_MMAP is fine here, since the real work 481 * is done in the FreeBSD mmap(). 482 */ 483 484 error = fget(td, bsd_args.fd, 485 cap_rights_init(&rights, CAP_MMAP), &fp); 486 if (error != 0) 487 return (error); 488 if (fp->f_type != DTYPE_VNODE) { 489 fdrop(fp, td); 490 return (EINVAL); 491 } 492 493 /* Linux mmap() just fails for O_WRONLY files */ 494 if (!(fp->f_flag & FREAD)) { 495 fdrop(fp, td); 496 return (EACCES); 497 } 498 499 fdrop(fp, td); 500 } 501 502 if (flags & LINUX_MAP_GROWSDOWN) { 503 /* 504 * The Linux MAP_GROWSDOWN option does not limit auto 505 * growth of the region. Linux mmap with this option 506 * takes as addr the inital BOS, and as len, the initial 507 * region size. It can then grow down from addr without 508 * limit. However, linux threads has an implicit internal 509 * limit to stack size of STACK_SIZE. Its just not 510 * enforced explicitly in linux. But, here we impose 511 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 512 * region, since we can do this with our mmap. 513 * 514 * Our mmap with MAP_STACK takes addr as the maximum 515 * downsize limit on BOS, and as len the max size of 516 * the region. It them maps the top SGROWSIZ bytes, 517 * and auto grows the region down, up to the limit 518 * in addr. 519 * 520 * If we don't use the MAP_STACK option, the effect 521 * of this code is to allocate a stack region of a 522 * fixed size of (STACK_SIZE - GUARD_SIZE). 523 */ 524 525 if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) { 526 /* 527 * Some linux apps will attempt to mmap 528 * thread stacks near the top of their 529 * address space. If their TOS is greater 530 * than vm_maxsaddr, vm_map_growstack() 531 * will confuse the thread stack with the 532 * process stack and deliver a SEGV if they 533 * attempt to grow the thread stack past their 534 * current stacksize rlimit. To avoid this, 535 * adjust vm_maxsaddr upwards to reflect 536 * the current stacksize rlimit rather 537 * than the maximum possible stacksize. 538 * It would be better to adjust the 539 * mmap'ed region, but some apps do not check 540 * mmap's return value. 541 */ 542 PROC_LOCK(p); 543 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 544 lim_cur(p, RLIMIT_STACK); 545 PROC_UNLOCK(p); 546 } 547 548 /* 549 * This gives us our maximum stack size and a new BOS. 550 * If we're using VM_STACK, then mmap will just map 551 * the top SGROWSIZ bytes, and let the stack grow down 552 * to the limit at BOS. If we're not using VM_STACK 553 * we map the full stack, since we don't have a way 554 * to autogrow it. 555 */ 556 if (len > STACK_SIZE - GUARD_SIZE) { 557 bsd_args.addr = (caddr_t)PTRIN(addr); 558 bsd_args.len = len; 559 } else { 560 bsd_args.addr = (caddr_t)PTRIN(addr) - 561 (STACK_SIZE - GUARD_SIZE - len); 562 bsd_args.len = STACK_SIZE - GUARD_SIZE; 563 } 564 } else { 565 bsd_args.addr = (caddr_t)PTRIN(addr); 566 bsd_args.len = len; 567 } 568 bsd_args.pos = pos; 569 570#ifdef DEBUG 571 if (ldebug(mmap)) 572 printf("-> %s(%p, %d, %d, 0x%08x, %d, 0x%x)\n", 573 __func__, 574 (void *)bsd_args.addr, bsd_args.len, bsd_args.prot, 575 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 576#endif 577 error = sys_mmap(td, &bsd_args); 578#ifdef DEBUG 579 if (ldebug(mmap)) 580 printf("-> %s() return: 0x%x (0x%08x)\n", 581 __func__, error, (u_int)td->td_retval[0]); 582#endif 583 return (error); 584} 585 586int 587linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) 588{ 589 struct mprotect_args bsd_args; 590 591 bsd_args.addr = uap->addr; 592 bsd_args.len = uap->len; 593 bsd_args.prot = uap->prot; 594 if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) 595 bsd_args.prot |= PROT_READ | PROT_EXEC; 596 return (sys_mprotect(td, &bsd_args)); 597} 598 599int 600linux_ioperm(struct thread *td, struct linux_ioperm_args *args) 601{ 602 int error; 603 struct i386_ioperm_args iia; 604 605 iia.start = args->start; 606 iia.length = args->length; 607 iia.enable = args->enable; 608 error = i386_set_ioperm(td, &iia); 609 return (error); 610} 611 612int 613linux_iopl(struct thread *td, struct linux_iopl_args *args) 614{ 615 int error; 616 617 if (args->level < 0 || args->level > 3) 618 return (EINVAL); 619 if ((error = priv_check(td, PRIV_IO)) != 0) 620 return (error); 621 if ((error = securelevel_gt(td->td_ucred, 0)) != 0) 622 return (error); 623 td->td_frame->tf_eflags = (td->td_frame->tf_eflags & ~PSL_IOPL) | 624 (args->level * (PSL_IOPL / 3)); 625 return (0); 626} 627 628int 629linux_modify_ldt(struct thread *td, struct linux_modify_ldt_args *uap) 630{ 631 int error; 632 struct i386_ldt_args ldt; 633 struct l_descriptor ld; 634 union descriptor desc; 635 int size, written; 636 637 switch (uap->func) { 638 case 0x00: /* read_ldt */ 639 ldt.start = 0; 640 ldt.descs = uap->ptr; 641 ldt.num = uap->bytecount / sizeof(union descriptor); 642 error = i386_get_ldt(td, &ldt); 643 td->td_retval[0] *= sizeof(union descriptor); 644 break; 645 case 0x02: /* read_default_ldt = 0 */ 646 size = 5*sizeof(struct l_desc_struct); 647 if (size > uap->bytecount) 648 size = uap->bytecount; 649 for (written = error = 0; written < size && error == 0; written++) 650 error = subyte((char *)uap->ptr + written, 0); 651 td->td_retval[0] = written; 652 break; 653 case 0x01: /* write_ldt */ 654 case 0x11: /* write_ldt */ 655 if (uap->bytecount != sizeof(ld)) 656 return (EINVAL); 657 658 error = copyin(uap->ptr, &ld, sizeof(ld)); 659 if (error) 660 return (error); 661 662 ldt.start = ld.entry_number; 663 ldt.descs = &desc; 664 ldt.num = 1; 665 desc.sd.sd_lolimit = (ld.limit & 0x0000ffff); 666 desc.sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 667 desc.sd.sd_lobase = (ld.base_addr & 0x00ffffff); 668 desc.sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 669 desc.sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 670 (ld.contents << 2); 671 desc.sd.sd_dpl = 3; 672 desc.sd.sd_p = (ld.seg_not_present ^ 1); 673 desc.sd.sd_xx = 0; 674 desc.sd.sd_def32 = ld.seg_32bit; 675 desc.sd.sd_gran = ld.limit_in_pages; 676 error = i386_set_ldt(td, &ldt, &desc); 677 break; 678 default: 679 error = ENOSYS; 680 break; 681 } 682 683 if (error == EOPNOTSUPP) { 684 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 685 error = ENOSYS; 686 } 687 688 return (error); 689} 690 691int 692linux_sigaction(struct thread *td, struct linux_sigaction_args *args) 693{ 694 l_osigaction_t osa; 695 l_sigaction_t act, oact; 696 int error; 697 698#ifdef DEBUG 699 if (ldebug(sigaction)) 700 printf(ARGS(sigaction, "%d, %p, %p"), 701 args->sig, (void *)args->nsa, (void *)args->osa); 702#endif 703 704 if (args->nsa != NULL) { 705 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t)); 706 if (error) 707 return (error); 708 act.lsa_handler = osa.lsa_handler; 709 act.lsa_flags = osa.lsa_flags; 710 act.lsa_restorer = osa.lsa_restorer; 711 LINUX_SIGEMPTYSET(act.lsa_mask); 712 act.lsa_mask.__bits[0] = osa.lsa_mask; 713 } 714 715 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL, 716 args->osa ? &oact : NULL); 717 718 if (args->osa != NULL && !error) { 719 osa.lsa_handler = oact.lsa_handler; 720 osa.lsa_flags = oact.lsa_flags; 721 osa.lsa_restorer = oact.lsa_restorer; 722 osa.lsa_mask = oact.lsa_mask.__bits[0]; 723 error = copyout(&osa, args->osa, sizeof(l_osigaction_t)); 724 } 725 726 return (error); 727} 728 729/* 730 * Linux has two extra args, restart and oldmask. We dont use these, 731 * but it seems that "restart" is actually a context pointer that 732 * enables the signal to happen with a different register set. 733 */ 734int 735linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args) 736{ 737 sigset_t sigmask; 738 l_sigset_t mask; 739 740#ifdef DEBUG 741 if (ldebug(sigsuspend)) 742 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 743#endif 744 745 LINUX_SIGEMPTYSET(mask); 746 mask.__bits[0] = args->mask; 747 linux_to_bsd_sigset(&mask, &sigmask); 748 return (kern_sigsuspend(td, sigmask)); 749} 750 751int 752linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap) 753{ 754 l_sigset_t lmask; 755 sigset_t sigmask; 756 int error; 757 758#ifdef DEBUG 759 if (ldebug(rt_sigsuspend)) 760 printf(ARGS(rt_sigsuspend, "%p, %d"), 761 (void *)uap->newset, uap->sigsetsize); 762#endif 763 764 if (uap->sigsetsize != sizeof(l_sigset_t)) 765 return (EINVAL); 766 767 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t)); 768 if (error) 769 return (error); 770 771 linux_to_bsd_sigset(&lmask, &sigmask); 772 return (kern_sigsuspend(td, sigmask)); 773} 774 775int 776linux_pause(struct thread *td, struct linux_pause_args *args) 777{ 778 struct proc *p = td->td_proc; 779 sigset_t sigmask; 780 781#ifdef DEBUG 782 if (ldebug(pause)) 783 printf(ARGS(pause, "")); 784#endif 785 786 PROC_LOCK(p); 787 sigmask = td->td_sigmask; 788 PROC_UNLOCK(p); 789 return (kern_sigsuspend(td, sigmask)); 790} 791 792int 793linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap) 794{ 795 stack_t ss, oss; 796 l_stack_t lss; 797 int error; 798 799#ifdef DEBUG 800 if (ldebug(sigaltstack)) 801 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 802#endif 803 804 if (uap->uss != NULL) { 805 error = copyin(uap->uss, &lss, sizeof(l_stack_t)); 806 if (error) 807 return (error); 808 809 ss.ss_sp = lss.ss_sp; 810 ss.ss_size = lss.ss_size; 811 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 812 } 813 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL, 814 (uap->uoss != NULL) ? &oss : NULL); 815 if (!error && uap->uoss != NULL) { 816 lss.ss_sp = oss.ss_sp; 817 lss.ss_size = oss.ss_size; 818 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags); 819 error = copyout(&lss, uap->uoss, sizeof(l_stack_t)); 820 } 821 822 return (error); 823} 824 825int 826linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args) 827{ 828 struct ftruncate_args sa; 829 830#ifdef DEBUG 831 if (ldebug(ftruncate64)) 832 printf(ARGS(ftruncate64, "%u, %jd"), args->fd, 833 (intmax_t)args->length); 834#endif 835 836 sa.fd = args->fd; 837 sa.length = args->length; 838 return sys_ftruncate(td, &sa); 839} 840 841int 842linux_set_thread_area(struct thread *td, struct linux_set_thread_area_args *args) 843{ 844 struct l_user_desc info; 845 int error; 846 int idx; 847 int a[2]; 848 struct segment_descriptor sd; 849 850 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 851 if (error) 852 return (error); 853 854#ifdef DEBUG 855 if (ldebug(set_thread_area)) 856 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, %i, %i, %i\n"), 857 info.entry_number, 858 info.base_addr, 859 info.limit, 860 info.seg_32bit, 861 info.contents, 862 info.read_exec_only, 863 info.limit_in_pages, 864 info.seg_not_present, 865 info.useable); 866#endif 867 868 idx = info.entry_number; 869 /* 870 * Semantics of linux version: every thread in the system has array of 871 * 3 tls descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown. This 872 * syscall loads one of the selected tls decriptors with a value and 873 * also loads GDT descriptors 6, 7 and 8 with the content of the 874 * per-thread descriptors. 875 * 876 * Semantics of fbsd version: I think we can ignore that linux has 3 877 * per-thread descriptors and use just the 1st one. The tls_array[] 878 * is used only in set/get-thread_area() syscalls and for loading the 879 * GDT descriptors. In fbsd we use just one GDT descriptor for TLS so 880 * we will load just one. 881 * 882 * XXX: this doesn't work when a user space process tries to use more 883 * than 1 TLS segment. Comment in the linux sources says wine might do 884 * this. 885 */ 886 887 /* 888 * we support just GLIBC TLS now 889 * we should let 3 proceed as well because we use this segment so 890 * if code does two subsequent calls it should succeed 891 */ 892 if (idx != 6 && idx != -1 && idx != 3) 893 return (EINVAL); 894 895 /* 896 * we have to copy out the GDT entry we use 897 * FreeBSD uses GDT entry #3 for storing %gs so load that 898 * 899 * XXX: what if a user space program doesn't check this value and tries 900 * to use 6, 7 or 8? 901 */ 902 idx = info.entry_number = 3; 903 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 904 if (error) 905 return (error); 906 907 if (LINUX_LDT_empty(&info)) { 908 a[0] = 0; 909 a[1] = 0; 910 } else { 911 a[0] = LINUX_LDT_entry_a(&info); 912 a[1] = LINUX_LDT_entry_b(&info); 913 } 914 915 memcpy(&sd, &a, sizeof(a)); 916#ifdef DEBUG 917 if (ldebug(set_thread_area)) 918 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, 919 sd.sd_hibase, 920 sd.sd_lolimit, 921 sd.sd_hilimit, 922 sd.sd_type, 923 sd.sd_dpl, 924 sd.sd_p, 925 sd.sd_xx, 926 sd.sd_def32, 927 sd.sd_gran); 928#endif 929 930 /* this is taken from i386 version of cpu_set_user_tls() */ 931 critical_enter(); 932 /* set %gs */ 933 td->td_pcb->pcb_gsd = sd; 934 PCPU_GET(fsgs_gdt)[1] = sd; 935 load_gs(GSEL(GUGS_SEL, SEL_UPL)); 936 critical_exit(); 937 938 return (0); 939} 940 941int 942linux_get_thread_area(struct thread *td, struct linux_get_thread_area_args *args) 943{ 944 945 struct l_user_desc info; 946 int error; 947 int idx; 948 struct l_desc_struct desc; 949 struct segment_descriptor sd; 950 951#ifdef DEBUG 952 if (ldebug(get_thread_area)) 953 printf(ARGS(get_thread_area, "%p"), args->desc); 954#endif 955 956 error = copyin(args->desc, &info, sizeof(struct l_user_desc)); 957 if (error) 958 return (error); 959 960 idx = info.entry_number; 961 /* XXX: I am not sure if we want 3 to be allowed too. */ 962 if (idx != 6 && idx != 3) 963 return (EINVAL); 964 965 idx = 3; 966 967 memset(&info, 0, sizeof(info)); 968 969 sd = PCPU_GET(fsgs_gdt)[1]; 970 971 memcpy(&desc, &sd, sizeof(desc)); 972 973 info.entry_number = idx; 974 info.base_addr = LINUX_GET_BASE(&desc); 975 info.limit = LINUX_GET_LIMIT(&desc); 976 info.seg_32bit = LINUX_GET_32BIT(&desc); 977 info.contents = LINUX_GET_CONTENTS(&desc); 978 info.read_exec_only = !LINUX_GET_WRITABLE(&desc); 979 info.limit_in_pages = LINUX_GET_LIMIT_PAGES(&desc); 980 info.seg_not_present = !LINUX_GET_PRESENT(&desc); 981 info.useable = LINUX_GET_USEABLE(&desc); 982 983 error = copyout(&info, args->desc, sizeof(struct l_user_desc)); 984 if (error) 985 return (EFAULT); 986 987 return (0); 988} 989 990/* XXX: this wont work with module - convert it */ 991int 992linux_mq_open(struct thread *td, struct linux_mq_open_args *args) 993{ 994#ifdef P1003_1B_MQUEUE 995 return sys_kmq_open(td, (struct kmq_open_args *) args); 996#else 997 return (ENOSYS); 998#endif 999} 1000 1001int 1002linux_mq_unlink(struct thread *td, struct linux_mq_unlink_args *args) 1003{ 1004#ifdef P1003_1B_MQUEUE 1005 return sys_kmq_unlink(td, (struct kmq_unlink_args *) args); 1006#else 1007 return (ENOSYS); 1008#endif 1009} 1010 1011int 1012linux_mq_timedsend(struct thread *td, struct linux_mq_timedsend_args *args) 1013{ 1014#ifdef P1003_1B_MQUEUE 1015 return sys_kmq_timedsend(td, (struct kmq_timedsend_args *) args); 1016#else 1017 return (ENOSYS); 1018#endif 1019} 1020 1021int 1022linux_mq_timedreceive(struct thread *td, struct linux_mq_timedreceive_args *args) 1023{ 1024#ifdef P1003_1B_MQUEUE 1025 return sys_kmq_timedreceive(td, (struct kmq_timedreceive_args *) args); 1026#else 1027 return (ENOSYS); 1028#endif 1029} 1030 1031int 1032linux_mq_notify(struct thread *td, struct linux_mq_notify_args *args) 1033{ 1034#ifdef P1003_1B_MQUEUE 1035 return sys_kmq_notify(td, (struct kmq_notify_args *) args); 1036#else 1037 return (ENOSYS); 1038#endif 1039} 1040 1041int 1042linux_mq_getsetattr(struct thread *td, struct linux_mq_getsetattr_args *args) 1043{ 1044#ifdef P1003_1B_MQUEUE 1045 return sys_kmq_setattr(td, (struct kmq_setattr_args *) args); 1046#else 1047 return (ENOSYS); 1048#endif 1049} 1050