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 * $FreeBSD$ 29 */ 30 31 #include <sys/param.h> 32 #include <sys/mman.h> 33 #include <sys/proc.h> 34 #include <sys/sysproto.h> 35 #include <sys/systm.h> 36 #include <sys/unistd.h> 37 #include <sys/resource.h> 38 #include <sys/resourcevar.h> 39 40 #include <machine/frame.h> 41 #include <machine/psl.h> 42 #include <machine/segments.h> 43 #include <machine/sysarch.h> 44 45 #include <vm/vm.h> 46 #include <sys/lock.h> 47 #include <vm/pmap.h> 48 #include <vm/vm_map.h> 49 50 #include <i386/linux/linux.h> 51 #include <i386/linux/linux_proto.h> 52 #include <compat/linux/linux_ipc.h> 53 #include <compat/linux/linux_signal.h> 54 #include <compat/linux/linux_util.h> 55 56 struct linux_descriptor { 57 unsigned int entry_number; 58 unsigned long base_addr; 59 unsigned int limit; 60 unsigned int seg_32bit:1; 61 unsigned int contents:2; 62 unsigned int read_exec_only:1; 63 unsigned int limit_in_pages:1; 64 unsigned int seg_not_present:1; 65 unsigned int useable:1; 66 }; 67 68 struct linux_select_argv { 69 int nfds; 70 fd_set *readfds; 71 fd_set *writefds; 72 fd_set *exceptfds; 73 struct timeval *timeout; 74 }; 75 76 int 77 linux_to_bsd_sigaltstack(int lsa) 78 { 79 int bsa = 0; 80 81 if (lsa & LINUX_SS_DISABLE) 82 bsa |= SS_DISABLE; 83 if (lsa & LINUX_SS_ONSTACK) 84 bsa |= SS_ONSTACK; 85 if (lsa == LINUX_SS_ONSTACK_BC) 86 bsa = SS_ONSTACK; 87 return (bsa); 88 } 89 90 int 91 bsd_to_linux_sigaltstack(int bsa) 92 { 93 int lsa = 0; 94 95 if (bsa & SS_DISABLE) 96 lsa |= LINUX_SS_DISABLE; 97 if (bsa & SS_ONSTACK) 98 lsa |= LINUX_SS_ONSTACK; 99 return (lsa); 100 } 101 102 int 103 linux_execve(struct proc *p, struct linux_execve_args *args) 104 { 105 struct execve_args bsd; 106 caddr_t sg; 107 108 sg = stackgap_init(); 109 CHECKALTEXIST(p, &sg, args->path); 110 111 #ifdef DEBUG 112 printf("Linux-emul(%d): execve(%s)\n", 113 p->p_pid, args->path); 114 #endif 115 116 bsd.fname = args->path; 117 bsd.argv = args->argp; 118 bsd.envv = args->envp; 119 return (execve(p, &bsd)); 120 } 121 122 int 123 linux_ipc(struct proc *p, struct linux_ipc_args *args) 124 { 125 switch (args->what) { 126 case LINUX_SEMOP: 127 return (linux_semop(p, args)); 128 case LINUX_SEMGET: 129 return (linux_semget(p, args)); 130 case LINUX_SEMCTL: 131 return (linux_semctl(p, args)); 132 case LINUX_MSGSND: 133 return (linux_msgsnd(p, args)); 134 case LINUX_MSGRCV: 135 return (linux_msgrcv(p, args)); 136 case LINUX_MSGGET: 137 return (linux_msgget(p, args)); 138 case LINUX_MSGCTL: 139 return (linux_msgctl(p, args)); 140 case LINUX_SHMAT: 141 return (linux_shmat(p, args)); 142 case LINUX_SHMDT: 143 return (linux_shmdt(p, args)); 144 case LINUX_SHMGET: 145 return (linux_shmget(p, args)); 146 case LINUX_SHMCTL: 147 return (linux_shmctl(p, args)); 148 } 149 150 uprintf("LINUX: 'ipc' typ=%d not implemented\n", args->what); 151 return (ENOSYS); 152 } 153 154 int 155 linux_select(struct proc *p, struct linux_select_args *args) 156 { 157 struct linux_select_argv linux_args; 158 struct linux_newselect_args newsel; 159 int error; 160 161 #ifdef SELECT_DEBUG 162 printf("Linux-emul(%ld): select(%x)\n", (long)p->p_pid, args->ptr); 163 #endif 164 165 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 166 if (error) 167 return (error); 168 169 newsel.nfds = linux_args.nfds; 170 newsel.readfds = linux_args.readfds; 171 newsel.writefds = linux_args.writefds; 172 newsel.exceptfds = linux_args.exceptfds; 173 newsel.timeout = linux_args.timeout; 174 return (linux_newselect(p, &newsel)); 175 } 176 177 int 178 linux_fork(struct proc *p, struct linux_fork_args *args) 179 { 180 int error; 181 182 #ifdef DEBUG 183 printf("Linux-emul(%ld): fork()\n", (long)p->p_pid); 184 #endif 185 186 if ((error = fork(p, (struct fork_args *)args)) != 0) 187 return (error); 188 189 if (p->p_retval[1] == 1) 190 p->p_retval[0] = 0; 191 return (0); 192 } 193 194 int 195 linux_vfork(struct proc *p, struct linux_vfork_args *args) 196 { 197 int error; 198 199 #ifdef DEBUG 200 printf("Linux-emul(%ld): vfork()\n", (long)p->p_pid); 201 #endif 202 203 if ((error = vfork(p, (struct vfork_args *)args)) != 0) 204 return (error); 205 /* Are we the child? */ 206 if (p->p_retval[1] == 1) 207 p->p_retval[0] = 0; 208 return (0); 209 } 210 211 #define CLONE_VM 0x100 212 #define CLONE_FS 0x200 213 #define CLONE_FILES 0x400 214 #define CLONE_SIGHAND 0x800 215 #define CLONE_PID 0x1000 216 217 int 218 linux_clone(struct proc *p, struct linux_clone_args *args) 219 { 220 int error, ff = RFPROC; 221 struct proc *p2; 222 int exit_signal; 223 vm_offset_t start; 224 struct rfork_args rf_args; 225 226 #ifdef DEBUG 227 if (args->flags & CLONE_PID) 228 printf("linux_clone(%ld): CLONE_PID not yet supported\n", 229 (long)p->p_pid); 230 printf("linux_clone(%ld): invoked with flags %x and stack %x\n", 231 (long)p->p_pid, (unsigned int)args->flags, 232 (unsigned int)args->stack); 233 #endif 234 235 if (!args->stack) 236 return (EINVAL); 237 238 exit_signal = args->flags & 0x000000ff; 239 if (exit_signal >= LINUX_NSIG) 240 return (EINVAL); 241 242 if (exit_signal <= LINUX_SIGTBLSZ) 243 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 244 245 /* RFTHREAD probably not necessary here, but it shouldn't hurt */ 246 ff |= RFTHREAD; 247 248 if (args->flags & CLONE_VM) 249 ff |= RFMEM; 250 if (args->flags & CLONE_SIGHAND) 251 ff |= RFSIGSHARE; 252 if (!(args->flags & CLONE_FILES)) 253 ff |= RFFDG; 254 255 error = 0; 256 start = 0; 257 258 rf_args.flags = ff; 259 if ((error = rfork(p, &rf_args)) != 0) 260 return (error); 261 262 p2 = pfind(p->p_retval[0]); 263 if (p2 == 0) 264 return (ESRCH); 265 266 p2->p_sigparent = exit_signal; 267 p2->p_md.md_regs->tf_esp = (unsigned int)args->stack; 268 269 #ifdef DEBUG 270 printf ("linux_clone(%ld): successful rfork to %ld\n", (long)p->p_pid, 271 (long)p2->p_pid); 272 #endif 273 274 return (0); 275 } 276 277 /* XXX move */ 278 struct linux_mmap_argv { 279 linux_caddr_t addr; 280 int len; 281 int prot; 282 int flags; 283 int fd; 284 int pos; 285 }; 286 287 #define STACK_SIZE (2 * 1024 * 1024) 288 #define GUARD_SIZE (4 * PAGE_SIZE) 289 290 int 291 linux_mmap(struct proc *p, struct linux_mmap_args *args) 292 { 293 struct mmap_args /* { 294 caddr_t addr; 295 size_t len; 296 int prot; 297 int flags; 298 int fd; 299 long pad; 300 off_t pos; 301 } */ bsd_args; 302 int error; 303 struct linux_mmap_argv linux_args; 304 305 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 306 if (error) 307 return (error); 308 309 #ifdef DEBUG 310 printf("Linux-emul(%ld): mmap(%p, %d, %d, 0x%08x, %d, %d)", 311 (long)p->p_pid, (void *)linux_args.addr, linux_args.len, 312 linux_args.prot, linux_args.flags, linux_args.fd, linux_args.pos); 313 #endif 314 315 bsd_args.flags = 0; 316 if (linux_args.flags & LINUX_MAP_SHARED) 317 bsd_args.flags |= MAP_SHARED; 318 if (linux_args.flags & LINUX_MAP_PRIVATE) 319 bsd_args.flags |= MAP_PRIVATE; 320 if (linux_args.flags & LINUX_MAP_FIXED) 321 bsd_args.flags |= MAP_FIXED; 322 if (linux_args.flags & LINUX_MAP_ANON) 323 bsd_args.flags |= MAP_ANON; 324 if (linux_args.flags & LINUX_MAP_GROWSDOWN) { 325 bsd_args.flags |= MAP_STACK; 326 327 /* The linux MAP_GROWSDOWN option does not limit auto 328 * growth of the region. Linux mmap with this option 329 * takes as addr the inital BOS, and as len, the initial 330 * region size. It can then grow down from addr without 331 * limit. However, linux threads has an implicit internal 332 * limit to stack size of STACK_SIZE. Its just not 333 * enforced explicitly in linux. But, here we impose 334 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 335 * region, since we can do this with our mmap. 336 * 337 * Our mmap with MAP_STACK takes addr as the maximum 338 * downsize limit on BOS, and as len the max size of 339 * the region. It them maps the top SGROWSIZ bytes, 340 * and autgrows the region down, up to the limit 341 * in addr. 342 * 343 * If we don't use the MAP_STACK option, the effect 344 * of this code is to allocate a stack region of a 345 * fixed size of (STACK_SIZE - GUARD_SIZE). 346 */ 347 348 /* This gives us TOS */ 349 bsd_args.addr = linux_args.addr + linux_args.len; 350 351 if (bsd_args.addr > p->p_vmspace->vm_maxsaddr) { 352 /* Some linux apps will attempt to mmap 353 * thread stacks near the top of their 354 * address space. If their TOS is greater 355 * than vm_maxsaddr, vm_map_growstack() 356 * will confuse the thread stack with the 357 * process stack and deliver a SEGV if they 358 * attempt to grow the thread stack past their 359 * current stacksize rlimit. To avoid this, 360 * adjust vm_maxsaddr upwards to reflect 361 * the current stacksize rlimit rather 362 * than the maximum possible stacksize. 363 * It would be better to adjust the 364 * mmap'ed region, but some apps do not check 365 * mmap's return value. 366 */ 367 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 368 p->p_rlimit[RLIMIT_STACK].rlim_cur; 369 } 370 371 /* This gives us our maximum stack size */ 372 if (linux_args.len > STACK_SIZE - GUARD_SIZE) 373 bsd_args.len = linux_args.len; 374 else 375 bsd_args.len = STACK_SIZE - GUARD_SIZE; 376 377 /* This gives us a new BOS. If we're using VM_STACK, then 378 * mmap will just map the top SGROWSIZ bytes, and let 379 * the stack grow down to the limit at BOS. If we're 380 * not using VM_STACK we map the full stack, since we 381 * don't have a way to autogrow it. 382 */ 383 bsd_args.addr -= bsd_args.len; 384 } else { 385 bsd_args.addr = linux_args.addr; 386 bsd_args.len = linux_args.len; 387 } 388 389 bsd_args.prot = linux_args.prot | PROT_READ; /* always required */ 390 if (linux_args.flags & LINUX_MAP_ANON) 391 bsd_args.fd = -1; 392 else 393 bsd_args.fd = linux_args.fd; 394 bsd_args.pos = linux_args.pos; 395 bsd_args.pad = 0; 396 397 #ifdef DEBUG 398 printf("-> (%p, %d, %d, 0x%08x, %d, %d)\n", (void *)bsd_args.addr, 399 bsd_args.len, bsd_args.prot, bsd_args.flags, bsd_args.fd, 400 (int)bsd_args.pos); 401 #endif 402 403 return (mmap(p, &bsd_args)); 404 } 405 406 int 407 linux_pipe(struct proc *p, struct linux_pipe_args *args) 408 { 409 int error; 410 int reg_edx; 411 412 #ifdef DEBUG 413 printf("Linux-emul(%ld): pipe(*)\n", (long)p->p_pid); 414 #endif 415 416 reg_edx = p->p_retval[1]; 417 error = pipe(p, 0); 418 if (error) { 419 p->p_retval[1] = reg_edx; 420 return (error); 421 } 422 423 error = copyout(p->p_retval, args->pipefds, 2*sizeof(int)); 424 if (error) { 425 p->p_retval[1] = reg_edx; 426 return (error); 427 } 428 429 p->p_retval[1] = reg_edx; 430 p->p_retval[0] = 0; 431 return (0); 432 } 433 434 int 435 linux_ioperm(struct proc *p, struct linux_ioperm_args *args) 436 { 437 struct sysarch_args sa; 438 struct i386_ioperm_args *iia; 439 caddr_t sg; 440 441 sg = stackgap_init(); 442 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args)); 443 iia->start = args->start; 444 iia->length = args->length; 445 iia->enable = args->enable; 446 sa.op = I386_SET_IOPERM; 447 sa.parms = (char *)iia; 448 return (sysarch(p, &sa)); 449 } 450 451 int 452 linux_iopl(struct proc *p, struct linux_iopl_args *args) 453 { 454 int error; 455 456 if (args->level < 0 || args->level > 3) 457 return (EINVAL); 458 if ((error = suser(p)) != 0) 459 return (error); 460 if (securelevel > 0) 461 return (EPERM); 462 p->p_md.md_regs->tf_eflags = (p->p_md.md_regs->tf_eflags & ~PSL_IOPL) | 463 (args->level * (PSL_IOPL / 3)); 464 return (0); 465 } 466 467 int 468 linux_modify_ldt(p, uap) 469 struct proc *p; 470 struct linux_modify_ldt_args *uap; 471 { 472 int error; 473 caddr_t sg; 474 struct sysarch_args args; 475 struct i386_ldt_args *ldt; 476 struct linux_descriptor ld; 477 union descriptor *desc; 478 479 sg = stackgap_init(); 480 481 if (uap->ptr == NULL) 482 return (EINVAL); 483 484 switch (uap->func) { 485 case 0x00: /* read_ldt */ 486 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 487 ldt->start = 0; 488 ldt->descs = uap->ptr; 489 ldt->num = uap->bytecount / sizeof(union descriptor); 490 args.op = I386_GET_LDT; 491 args.parms = (char*)ldt; 492 error = sysarch(p, &args); 493 p->p_retval[0] *= sizeof(union descriptor); 494 break; 495 case 0x01: /* write_ldt */ 496 case 0x11: /* write_ldt */ 497 if (uap->bytecount != sizeof(ld)) 498 return (EINVAL); 499 500 error = copyin(uap->ptr, &ld, sizeof(ld)); 501 if (error) 502 return (error); 503 504 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 505 desc = stackgap_alloc(&sg, sizeof(*desc)); 506 ldt->start = ld.entry_number; 507 ldt->descs = desc; 508 ldt->num = 1; 509 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff); 510 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 511 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff); 512 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 513 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 514 (ld.contents << 2); 515 desc->sd.sd_dpl = 3; 516 desc->sd.sd_p = (ld.seg_not_present ^ 1); 517 desc->sd.sd_xx = 0; 518 desc->sd.sd_def32 = ld.seg_32bit; 519 desc->sd.sd_gran = ld.limit_in_pages; 520 args.op = I386_SET_LDT; 521 args.parms = (char*)ldt; 522 error = sysarch(p, &args); 523 break; 524 default: 525 error = EINVAL; 526 break; 527 } 528 529 if (error == EOPNOTSUPP) { 530 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 531 error = ENOSYS; 532 } 533 534 return (error); 535 } 536 537 int 538 linux_sigaction(struct proc *p, struct linux_sigaction_args *args) 539 { 540 linux_osigaction_t osa; 541 linux_sigaction_t act, oact; 542 int error; 543 544 #ifdef DEBUG 545 printf("Linux-emul(%ld): sigaction(%d, %p, %p)\n", (long)p->p_pid, 546 args->sig, (void *)args->nsa, (void *)args->osa); 547 #endif 548 549 if (args->nsa != NULL) { 550 error = copyin(args->nsa, &osa, sizeof(linux_osigaction_t)); 551 if (error) 552 return (error); 553 act.lsa_handler = osa.lsa_handler; 554 act.lsa_flags = osa.lsa_flags; 555 act.lsa_restorer = osa.lsa_restorer; 556 LINUX_SIGEMPTYSET(act.lsa_mask); 557 act.lsa_mask.__bits[0] = osa.lsa_mask; 558 } 559 560 error = linux_do_sigaction(p, args->sig, args->nsa ? &act : NULL, 561 args->osa ? &oact : NULL); 562 563 if (args->osa != NULL && !error) { 564 osa.lsa_handler = oact.lsa_handler; 565 osa.lsa_flags = oact.lsa_flags; 566 osa.lsa_restorer = oact.lsa_restorer; 567 osa.lsa_mask = oact.lsa_mask.__bits[0]; 568 error = copyout(&osa, args->osa, sizeof(linux_osigaction_t)); 569 } 570 571 return (error); 572 } 573 574 /* 575 * Linux has two extra args, restart and oldmask. We dont use these, 576 * but it seems that "restart" is actually a context pointer that 577 * enables the signal to happen with a different register set. 578 */ 579 int 580 linux_sigsuspend(struct proc *p, struct linux_sigsuspend_args *args) 581 { 582 struct sigsuspend_args bsd; 583 sigset_t *sigmask; 584 linux_sigset_t mask; 585 caddr_t sg = stackgap_init(); 586 587 #ifdef DEBUG 588 printf("Linux-emul(%ld): sigsuspend(%08lx)\n", 589 (long)p->p_pid, (unsigned long)args->mask); 590 #endif 591 592 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 593 LINUX_SIGEMPTYSET(mask); 594 mask.__bits[0] = args->mask; 595 linux_to_bsd_sigset(&mask, sigmask); 596 bsd.sigmask = sigmask; 597 return (sigsuspend(p, &bsd)); 598 } 599 600 int 601 linux_rt_sigsuspend(p, uap) 602 struct proc *p; 603 struct linux_rt_sigsuspend_args *uap; 604 { 605 linux_sigset_t lmask; 606 sigset_t *bmask; 607 struct sigsuspend_args bsd; 608 caddr_t sg = stackgap_init(); 609 int error; 610 611 #ifdef DEBUG 612 printf("Linux-emul(%ld): rt_sigsuspend(%p, %d)\n", (long)p->p_pid, 613 (void *)uap->newset, uap->sigsetsize); 614 #endif 615 616 if (uap->sigsetsize != sizeof(linux_sigset_t)) 617 return (EINVAL); 618 619 error = copyin(uap->newset, &lmask, sizeof(linux_sigset_t)); 620 if (error) 621 return (error); 622 623 bmask = stackgap_alloc(&sg, sizeof(sigset_t)); 624 linux_to_bsd_sigset(&lmask, bmask); 625 bsd.sigmask = bmask; 626 return (sigsuspend(p, &bsd)); 627 } 628 629 int 630 linux_pause(struct proc *p, struct linux_pause_args *args) 631 { 632 struct sigsuspend_args bsd; 633 sigset_t *sigmask; 634 caddr_t sg = stackgap_init(); 635 636 #ifdef DEBUG 637 printf("Linux-emul(%d): pause()\n", p->p_pid); 638 #endif 639 640 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 641 *sigmask = p->p_sigmask; 642 bsd.sigmask = sigmask; 643 return (sigsuspend(p, &bsd)); 644 } 645 646 int 647 linux_sigaltstack(p, uap) 648 struct proc *p; 649 struct linux_sigaltstack_args *uap; 650 { 651 struct sigaltstack_args bsd; 652 stack_t *ss, *oss; 653 linux_stack_t lss; 654 int error; 655 caddr_t sg = stackgap_init(); 656 657 #ifdef DEBUG 658 printf("Linux-emul(%ld): sigaltstack(%p, %p)\n", 659 (long)p->p_pid, uap->uss, uap->uoss); 660 #endif 661 662 if (uap->uss == NULL) { 663 ss = NULL; 664 } else { 665 error = copyin(uap->uss, &lss, sizeof(linux_stack_t)); 666 if (error) 667 return (error); 668 669 ss = stackgap_alloc(&sg, sizeof(stack_t)); 670 ss->ss_sp = lss.ss_sp; 671 ss->ss_size = lss.ss_size; 672 ss->ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 673 } 674 oss = (uap->uoss != NULL) 675 ? stackgap_alloc(&sg, sizeof(stack_t)) 676 : NULL; 677 678 bsd.ss = ss; 679 bsd.oss = oss; 680 error = sigaltstack(p, &bsd); 681 682 if (!error && oss != NULL) { 683 lss.ss_sp = oss->ss_sp; 684 lss.ss_size = oss->ss_size; 685 lss.ss_flags = bsd_to_linux_sigaltstack(oss->ss_flags); 686 error = copyout(&lss, uap->uoss, sizeof(linux_stack_t)); 687 } 688 689 return (error); 690 } 691