linux_machdep.c revision 72543
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: head/sys/i386/linux/linux_machdep.c 72543 2001-02-16 16:40:43Z jlemon $ 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 56struct 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 68struct linux_select_argv { 69 int nfds; 70 fd_set *readfds; 71 fd_set *writefds; 72 fd_set *exceptfds; 73 struct timeval *timeout; 74}; 75 76int 77linux_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 return (bsa); 86} 87 88int 89bsd_to_linux_sigaltstack(int bsa) 90{ 91 int lsa = 0; 92 93 if (bsa & SS_DISABLE) 94 lsa |= LINUX_SS_DISABLE; 95 if (bsa & SS_ONSTACK) 96 lsa |= LINUX_SS_ONSTACK; 97 return (lsa); 98} 99 100int 101linux_execve(struct proc *p, struct linux_execve_args *args) 102{ 103 struct execve_args bsd; 104 caddr_t sg; 105 106 sg = stackgap_init(); 107 CHECKALTEXIST(p, &sg, args->path); 108 109#ifdef DEBUG 110 if (ldebug(execve)) 111 printf(ARGS(execve, "%s"), args->path); 112#endif 113 114 bsd.fname = args->path; 115 bsd.argv = args->argp; 116 bsd.envv = args->envp; 117 return (execve(p, &bsd)); 118} 119 120int 121linux_ipc(struct proc *p, struct linux_ipc_args *args) 122{ 123 switch (args->what) { 124 case LINUX_SEMOP: 125 return (linux_semop(p, args)); 126 case LINUX_SEMGET: 127 return (linux_semget(p, args)); 128 case LINUX_SEMCTL: 129 return (linux_semctl(p, args)); 130 case LINUX_MSGSND: 131 return (linux_msgsnd(p, args)); 132 case LINUX_MSGRCV: 133 return (linux_msgrcv(p, args)); 134 case LINUX_MSGGET: 135 return (linux_msgget(p, args)); 136 case LINUX_MSGCTL: 137 return (linux_msgctl(p, args)); 138 case LINUX_SHMAT: 139 return (linux_shmat(p, args)); 140 case LINUX_SHMDT: 141 return (linux_shmdt(p, args)); 142 case LINUX_SHMGET: 143 return (linux_shmget(p, args)); 144 case LINUX_SHMCTL: 145 return (linux_shmctl(p, args)); 146 } 147 148 uprintf("LINUX: 'ipc' typ=%d not implemented\n", args->what); 149 return (ENOSYS); 150} 151 152int 153linux_select(struct proc *p, struct linux_select_args *args) 154{ 155 struct linux_select_argv linux_args; 156 struct linux_newselect_args newsel; 157 int error; 158 159#ifdef SELECT_DEBUG 160 if (ldebug(select)) 161 printf(ARGS(select, "%x"), args->ptr); 162#endif 163 164 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 165 if (error) 166 return (error); 167 168 newsel.nfds = linux_args.nfds; 169 newsel.readfds = linux_args.readfds; 170 newsel.writefds = linux_args.writefds; 171 newsel.exceptfds = linux_args.exceptfds; 172 newsel.timeout = linux_args.timeout; 173 return (linux_newselect(p, &newsel)); 174} 175 176int 177linux_fork(struct proc *p, struct linux_fork_args *args) 178{ 179 int error; 180 181#ifdef DEBUG 182 if (ldebug(fork)) 183 printf(ARGS(fork, "")); 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 194int 195linux_vfork(struct proc *p, struct linux_vfork_args *args) 196{ 197 int error; 198 199#ifdef DEBUG 200 if (ldebug(vfork)) 201 printf(ARGS(vfork, "")); 202#endif 203 204 if ((error = vfork(p, (struct vfork_args *)args)) != 0) 205 return (error); 206 /* Are we the child? */ 207 if (p->p_retval[1] == 1) 208 p->p_retval[0] = 0; 209 return (0); 210} 211 212#define CLONE_VM 0x100 213#define CLONE_FS 0x200 214#define CLONE_FILES 0x400 215#define CLONE_SIGHAND 0x800 216#define CLONE_PID 0x1000 217 218int 219linux_clone(struct proc *p, struct linux_clone_args *args) 220{ 221 int error, ff = RFPROC; 222 struct proc *p2; 223 int exit_signal; 224 vm_offset_t start; 225 struct rfork_args rf_args; 226 227#ifdef DEBUG 228 if (ldebug(clone)) { 229 printf(ARGS(clone, "flags %x, stack %x"), 230 (unsigned int)args->flags, (unsigned int)args->stack); 231 if (args->flags & CLONE_PID) 232 printf(LMSG("CLONE_PID not yet supported")); 233 } 234#endif 235 236 if (!args->stack) 237 return (EINVAL); 238 239 exit_signal = args->flags & 0x000000ff; 240 if (exit_signal >= LINUX_NSIG) 241 return (EINVAL); 242 243 if (exit_signal <= LINUX_SIGTBLSZ) 244 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 245 246 /* RFTHREAD probably not necessary here, but it shouldn't hurt */ 247 ff |= RFTHREAD; 248 249 if (args->flags & CLONE_VM) 250 ff |= RFMEM; 251 if (args->flags & CLONE_SIGHAND) 252 ff |= RFSIGSHARE; 253 if (!(args->flags & CLONE_FILES)) 254 ff |= RFFDG; 255 256 error = 0; 257 start = 0; 258 259 rf_args.flags = ff; 260 if ((error = rfork(p, &rf_args)) != 0) 261 return (error); 262 263 p2 = pfind(p->p_retval[0]); 264 if (p2 == NULL) 265 return (ESRCH); 266 267 PROC_LOCK(p2); 268 p2->p_sigparent = exit_signal; 269 PROC_UNLOCK(p2); 270 p2->p_md.md_regs->tf_esp = (unsigned int)args->stack; 271 272#ifdef DEBUG 273 if (ldebug(clone)) 274 printf(LMSG("clone: successful rfork to %ld"), 275 (long)p2->p_pid); 276#endif 277 278 return (0); 279} 280 281/* XXX move */ 282struct linux_mmap_argv { 283 linux_caddr_t addr; 284 int len; 285 int prot; 286 int flags; 287 int fd; 288 int pos; 289}; 290 291#define STACK_SIZE (2 * 1024 * 1024) 292#define GUARD_SIZE (4 * PAGE_SIZE) 293 294int 295linux_mmap(struct proc *p, struct linux_mmap_args *args) 296{ 297 struct mmap_args /* { 298 caddr_t addr; 299 size_t len; 300 int prot; 301 int flags; 302 int fd; 303 long pad; 304 off_t pos; 305 } */ bsd_args; 306 int error; 307 struct linux_mmap_argv linux_args; 308 309 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 310 if (error) 311 return (error); 312 313#ifdef DEBUG 314 if (ldebug(mmap)) 315 printf(ARGS(mmap, "%p, %d, %d, 0x%08x, %d, %d"), 316 (void *)linux_args.addr, linux_args.len, linux_args.prot, 317 linux_args.flags, linux_args.fd, linux_args.pos); 318#endif 319 320 bsd_args.flags = 0; 321 if (linux_args.flags & LINUX_MAP_SHARED) 322 bsd_args.flags |= MAP_SHARED; 323 if (linux_args.flags & LINUX_MAP_PRIVATE) 324 bsd_args.flags |= MAP_PRIVATE; 325 if (linux_args.flags & LINUX_MAP_FIXED) 326 bsd_args.flags |= MAP_FIXED; 327 if (linux_args.flags & LINUX_MAP_ANON) 328 bsd_args.flags |= MAP_ANON; 329 if (linux_args.flags & LINUX_MAP_GROWSDOWN) { 330 bsd_args.flags |= MAP_STACK; 331 332 /* The linux MAP_GROWSDOWN option does not limit auto 333 * growth of the region. Linux mmap with this option 334 * takes as addr the inital BOS, and as len, the initial 335 * region size. It can then grow down from addr without 336 * limit. However, linux threads has an implicit internal 337 * limit to stack size of STACK_SIZE. Its just not 338 * enforced explicitly in linux. But, here we impose 339 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 340 * region, since we can do this with our mmap. 341 * 342 * Our mmap with MAP_STACK takes addr as the maximum 343 * downsize limit on BOS, and as len the max size of 344 * the region. It them maps the top SGROWSIZ bytes, 345 * and autgrows the region down, up to the limit 346 * in addr. 347 * 348 * If we don't use the MAP_STACK option, the effect 349 * of this code is to allocate a stack region of a 350 * fixed size of (STACK_SIZE - GUARD_SIZE). 351 */ 352 353 /* This gives us TOS */ 354 bsd_args.addr = linux_args.addr + linux_args.len; 355 356 if (bsd_args.addr > p->p_vmspace->vm_maxsaddr) { 357 /* Some linux apps will attempt to mmap 358 * thread stacks near the top of their 359 * address space. If their TOS is greater 360 * than vm_maxsaddr, vm_map_growstack() 361 * will confuse the thread stack with the 362 * process stack and deliver a SEGV if they 363 * attempt to grow the thread stack past their 364 * current stacksize rlimit. To avoid this, 365 * adjust vm_maxsaddr upwards to reflect 366 * the current stacksize rlimit rather 367 * than the maximum possible stacksize. 368 * It would be better to adjust the 369 * mmap'ed region, but some apps do not check 370 * mmap's return value. 371 */ 372 mtx_assert(&Giant, MA_OWNED); 373 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 374 p->p_rlimit[RLIMIT_STACK].rlim_cur; 375 } 376 377 /* This gives us our maximum stack size */ 378 if (linux_args.len > STACK_SIZE - GUARD_SIZE) 379 bsd_args.len = linux_args.len; 380 else 381 bsd_args.len = STACK_SIZE - GUARD_SIZE; 382 383 /* This gives us a new BOS. If we're using VM_STACK, then 384 * mmap will just map the top SGROWSIZ bytes, and let 385 * the stack grow down to the limit at BOS. If we're 386 * not using VM_STACK we map the full stack, since we 387 * don't have a way to autogrow it. 388 */ 389 bsd_args.addr -= bsd_args.len; 390 } else { 391 bsd_args.addr = linux_args.addr; 392 bsd_args.len = linux_args.len; 393 } 394 395 bsd_args.prot = linux_args.prot | PROT_READ; /* always required */ 396 if (linux_args.flags & LINUX_MAP_ANON) 397 bsd_args.fd = -1; 398 else 399 bsd_args.fd = linux_args.fd; 400 bsd_args.pos = linux_args.pos; 401 bsd_args.pad = 0; 402 403#ifdef DEBUG 404 if (ldebug(mmap)) 405 printf("-> (%p, %d, %d, 0x%08x, %d, %d)\n", 406 (void *)bsd_args.addr, bsd_args.len, bsd_args.prot, 407 bsd_args.flags, bsd_args.fd, (int)bsd_args.pos); 408#endif 409 410 return (mmap(p, &bsd_args)); 411} 412 413int 414linux_pipe(struct proc *p, struct linux_pipe_args *args) 415{ 416 int error; 417 int reg_edx; 418 419#ifdef DEBUG 420 if (ldebug(pipe)) 421 printf(ARGS(pipe, "*")); 422#endif 423 424 reg_edx = p->p_retval[1]; 425 error = pipe(p, 0); 426 if (error) { 427 p->p_retval[1] = reg_edx; 428 return (error); 429 } 430 431 error = copyout(p->p_retval, args->pipefds, 2*sizeof(int)); 432 if (error) { 433 p->p_retval[1] = reg_edx; 434 return (error); 435 } 436 437 p->p_retval[1] = reg_edx; 438 p->p_retval[0] = 0; 439 return (0); 440} 441 442int 443linux_ioperm(struct proc *p, struct linux_ioperm_args *args) 444{ 445 struct sysarch_args sa; 446 struct i386_ioperm_args *iia; 447 caddr_t sg; 448 449 sg = stackgap_init(); 450 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args)); 451 iia->start = args->start; 452 iia->length = args->length; 453 iia->enable = args->enable; 454 sa.op = I386_SET_IOPERM; 455 sa.parms = (char *)iia; 456 return (sysarch(p, &sa)); 457} 458 459int 460linux_iopl(struct proc *p, struct linux_iopl_args *args) 461{ 462 int error; 463 464 if (args->level < 0 || args->level > 3) 465 return (EINVAL); 466 if ((error = suser(p)) != 0) 467 return (error); 468 if (securelevel > 0) 469 return (EPERM); 470 p->p_md.md_regs->tf_eflags = (p->p_md.md_regs->tf_eflags & ~PSL_IOPL) | 471 (args->level * (PSL_IOPL / 3)); 472 return (0); 473} 474 475int 476linux_modify_ldt(p, uap) 477 struct proc *p; 478 struct linux_modify_ldt_args *uap; 479{ 480 int error; 481 caddr_t sg; 482 struct sysarch_args args; 483 struct i386_ldt_args *ldt; 484 struct linux_descriptor ld; 485 union descriptor *desc; 486 487 sg = stackgap_init(); 488 489 if (uap->ptr == NULL) 490 return (EINVAL); 491 492 switch (uap->func) { 493 case 0x00: /* read_ldt */ 494 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 495 ldt->start = 0; 496 ldt->descs = uap->ptr; 497 ldt->num = uap->bytecount / sizeof(union descriptor); 498 args.op = I386_GET_LDT; 499 args.parms = (char*)ldt; 500 error = sysarch(p, &args); 501 p->p_retval[0] *= sizeof(union descriptor); 502 break; 503 case 0x01: /* write_ldt */ 504 case 0x11: /* write_ldt */ 505 if (uap->bytecount != sizeof(ld)) 506 return (EINVAL); 507 508 error = copyin(uap->ptr, &ld, sizeof(ld)); 509 if (error) 510 return (error); 511 512 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 513 desc = stackgap_alloc(&sg, sizeof(*desc)); 514 ldt->start = ld.entry_number; 515 ldt->descs = desc; 516 ldt->num = 1; 517 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff); 518 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 519 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff); 520 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 521 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 522 (ld.contents << 2); 523 desc->sd.sd_dpl = 3; 524 desc->sd.sd_p = (ld.seg_not_present ^ 1); 525 desc->sd.sd_xx = 0; 526 desc->sd.sd_def32 = ld.seg_32bit; 527 desc->sd.sd_gran = ld.limit_in_pages; 528 args.op = I386_SET_LDT; 529 args.parms = (char*)ldt; 530 error = sysarch(p, &args); 531 break; 532 default: 533 error = EINVAL; 534 break; 535 } 536 537 if (error == EOPNOTSUPP) { 538 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 539 error = ENOSYS; 540 } 541 542 return (error); 543} 544 545int 546linux_sigaction(struct proc *p, struct linux_sigaction_args *args) 547{ 548 linux_osigaction_t osa; 549 linux_sigaction_t act, oact; 550 int error; 551 552#ifdef DEBUG 553 if (ldebug(sigaction)) 554 printf(ARGS(sigaction, "%d, %p, %p"), 555 args->sig, (void *)args->nsa, (void *)args->osa); 556#endif 557 558 if (args->nsa != NULL) { 559 error = copyin(args->nsa, &osa, sizeof(linux_osigaction_t)); 560 if (error) 561 return (error); 562 act.lsa_handler = osa.lsa_handler; 563 act.lsa_flags = osa.lsa_flags; 564 act.lsa_restorer = osa.lsa_restorer; 565 LINUX_SIGEMPTYSET(act.lsa_mask); 566 act.lsa_mask.__bits[0] = osa.lsa_mask; 567 } 568 569 error = linux_do_sigaction(p, args->sig, args->nsa ? &act : NULL, 570 args->osa ? &oact : NULL); 571 572 if (args->osa != NULL && !error) { 573 osa.lsa_handler = oact.lsa_handler; 574 osa.lsa_flags = oact.lsa_flags; 575 osa.lsa_restorer = oact.lsa_restorer; 576 osa.lsa_mask = oact.lsa_mask.__bits[0]; 577 error = copyout(&osa, args->osa, sizeof(linux_osigaction_t)); 578 } 579 580 return (error); 581} 582 583/* 584 * Linux has two extra args, restart and oldmask. We dont use these, 585 * but it seems that "restart" is actually a context pointer that 586 * enables the signal to happen with a different register set. 587 */ 588int 589linux_sigsuspend(struct proc *p, struct linux_sigsuspend_args *args) 590{ 591 struct sigsuspend_args bsd; 592 sigset_t *sigmask; 593 linux_sigset_t mask; 594 caddr_t sg = stackgap_init(); 595 596#ifdef DEBUG 597 if (ldebug(sigsuspend)) 598 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask); 599#endif 600 601 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 602 LINUX_SIGEMPTYSET(mask); 603 mask.__bits[0] = args->mask; 604 linux_to_bsd_sigset(&mask, sigmask); 605 bsd.sigmask = sigmask; 606 return (sigsuspend(p, &bsd)); 607} 608 609int 610linux_rt_sigsuspend(p, uap) 611 struct proc *p; 612 struct linux_rt_sigsuspend_args *uap; 613{ 614 linux_sigset_t lmask; 615 sigset_t *bmask; 616 struct sigsuspend_args bsd; 617 caddr_t sg = stackgap_init(); 618 int error; 619 620#ifdef DEBUG 621 if (ldebug(rt_sigsuspend)) 622 printf(ARGS(rt_sigsuspend, "%p, %d"), 623 (void *)uap->newset, uap->sigsetsize); 624#endif 625 626 if (uap->sigsetsize != sizeof(linux_sigset_t)) 627 return (EINVAL); 628 629 error = copyin(uap->newset, &lmask, sizeof(linux_sigset_t)); 630 if (error) 631 return (error); 632 633 bmask = stackgap_alloc(&sg, sizeof(sigset_t)); 634 linux_to_bsd_sigset(&lmask, bmask); 635 bsd.sigmask = bmask; 636 return (sigsuspend(p, &bsd)); 637} 638 639int 640linux_pause(struct proc *p, struct linux_pause_args *args) 641{ 642 struct sigsuspend_args bsd; 643 sigset_t *sigmask; 644 caddr_t sg = stackgap_init(); 645 646#ifdef DEBUG 647 if (ldebug(pause)) 648 printf(ARGS(pause, "")); 649#endif 650 651 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 652 PROC_LOCK(p); 653 *sigmask = p->p_sigmask; 654 PROC_UNLOCK(p); 655 bsd.sigmask = sigmask; 656 return (sigsuspend(p, &bsd)); 657} 658 659int 660linux_sigaltstack(p, uap) 661 struct proc *p; 662 struct linux_sigaltstack_args *uap; 663{ 664 struct sigaltstack_args bsd; 665 stack_t *ss, *oss; 666 linux_stack_t lss; 667 int error; 668 caddr_t sg = stackgap_init(); 669 670#ifdef DEBUG 671 if (ldebug(sigaltstack)) 672 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss); 673#endif 674 675 if (uap->uss == NULL) { 676 ss = NULL; 677 } else { 678 error = copyin(uap->uss, &lss, sizeof(linux_stack_t)); 679 if (error) 680 return (error); 681 682 ss = stackgap_alloc(&sg, sizeof(stack_t)); 683 ss->ss_sp = lss.ss_sp; 684 ss->ss_size = lss.ss_size; 685 ss->ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 686 } 687 oss = (uap->uoss != NULL) 688 ? stackgap_alloc(&sg, sizeof(stack_t)) 689 : NULL; 690 691 bsd.ss = ss; 692 bsd.oss = oss; 693 error = sigaltstack(p, &bsd); 694 695 if (!error && oss != NULL) { 696 lss.ss_sp = oss->ss_sp; 697 lss.ss_size = oss->ss_size; 698 lss.ss_flags = bsd_to_linux_sigaltstack(oss->ss_flags); 699 error = copyout(&lss, uap->uoss, sizeof(linux_stack_t)); 700 } 701 702 return (error); 703} 704