linux_machdep.c revision 71494
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 71494 2001-01-24 00:23:30Z jhb $ 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 printf("Linux-emul(%d): execve(%s)\n", 111 p->p_pid, 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 printf("Linux-emul(%ld): select(%x)\n", (long)p->p_pid, args->ptr); 161#endif 162 163 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 164 if (error) 165 return (error); 166 167 newsel.nfds = linux_args.nfds; 168 newsel.readfds = linux_args.readfds; 169 newsel.writefds = linux_args.writefds; 170 newsel.exceptfds = linux_args.exceptfds; 171 newsel.timeout = linux_args.timeout; 172 return (linux_newselect(p, &newsel)); 173} 174 175int 176linux_fork(struct proc *p, struct linux_fork_args *args) 177{ 178 int error; 179 180#ifdef DEBUG 181 printf("Linux-emul(%ld): fork()\n", (long)p->p_pid); 182#endif 183 184 if ((error = fork(p, (struct fork_args *)args)) != 0) 185 return (error); 186 187 if (p->p_retval[1] == 1) 188 p->p_retval[0] = 0; 189 return (0); 190} 191 192int 193linux_vfork(struct proc *p, struct linux_vfork_args *args) 194{ 195 int error; 196 197#ifdef DEBUG 198 printf("Linux-emul(%ld): vfork()\n", (long)p->p_pid); 199#endif 200 201 if ((error = vfork(p, (struct vfork_args *)args)) != 0) 202 return (error); 203 /* Are we the child? */ 204 if (p->p_retval[1] == 1) 205 p->p_retval[0] = 0; 206 return (0); 207} 208 209#define CLONE_VM 0x100 210#define CLONE_FS 0x200 211#define CLONE_FILES 0x400 212#define CLONE_SIGHAND 0x800 213#define CLONE_PID 0x1000 214 215int 216linux_clone(struct proc *p, struct linux_clone_args *args) 217{ 218 int error, ff = RFPROC; 219 struct proc *p2; 220 int exit_signal; 221 vm_offset_t start; 222 struct rfork_args rf_args; 223 224#ifdef DEBUG 225 if (args->flags & CLONE_PID) 226 printf("linux_clone(%ld): CLONE_PID not yet supported\n", 227 (long)p->p_pid); 228 printf("linux_clone(%ld): invoked with flags %x and stack %x\n", 229 (long)p->p_pid, (unsigned int)args->flags, 230 (unsigned int)args->stack); 231#endif 232 233 if (!args->stack) 234 return (EINVAL); 235 236 exit_signal = args->flags & 0x000000ff; 237 if (exit_signal >= LINUX_NSIG) 238 return (EINVAL); 239 240 if (exit_signal <= LINUX_SIGTBLSZ) 241 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 242 243 /* RFTHREAD probably not necessary here, but it shouldn't hurt */ 244 ff |= RFTHREAD; 245 246 if (args->flags & CLONE_VM) 247 ff |= RFMEM; 248 if (args->flags & CLONE_SIGHAND) 249 ff |= RFSIGSHARE; 250 if (!(args->flags & CLONE_FILES)) 251 ff |= RFFDG; 252 253 error = 0; 254 start = 0; 255 256 rf_args.flags = ff; 257 if ((error = rfork(p, &rf_args)) != 0) 258 return (error); 259 260 p2 = pfind(p->p_retval[0]); 261 if (p2 == NULL) 262 return (ESRCH); 263 264 PROC_LOCK(p2); 265 p2->p_sigparent = exit_signal; 266 PROC_UNLOCK(p2); 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 */ 278struct 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 290int 291linux_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 mtx_assert(&Giant, MA_OWNED); 368 p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - 369 p->p_rlimit[RLIMIT_STACK].rlim_cur; 370 } 371 372 /* This gives us our maximum stack size */ 373 if (linux_args.len > STACK_SIZE - GUARD_SIZE) 374 bsd_args.len = linux_args.len; 375 else 376 bsd_args.len = STACK_SIZE - GUARD_SIZE; 377 378 /* This gives us a new BOS. If we're using VM_STACK, then 379 * mmap will just map the top SGROWSIZ bytes, and let 380 * the stack grow down to the limit at BOS. If we're 381 * not using VM_STACK we map the full stack, since we 382 * don't have a way to autogrow it. 383 */ 384 bsd_args.addr -= bsd_args.len; 385 } else { 386 bsd_args.addr = linux_args.addr; 387 bsd_args.len = linux_args.len; 388 } 389 390 bsd_args.prot = linux_args.prot | PROT_READ; /* always required */ 391 if (linux_args.flags & LINUX_MAP_ANON) 392 bsd_args.fd = -1; 393 else 394 bsd_args.fd = linux_args.fd; 395 bsd_args.pos = linux_args.pos; 396 bsd_args.pad = 0; 397 398#ifdef DEBUG 399 printf("-> (%p, %d, %d, 0x%08x, %d, %d)\n", (void *)bsd_args.addr, 400 bsd_args.len, bsd_args.prot, bsd_args.flags, bsd_args.fd, 401 (int)bsd_args.pos); 402#endif 403 404 return (mmap(p, &bsd_args)); 405} 406 407int 408linux_pipe(struct proc *p, struct linux_pipe_args *args) 409{ 410 int error; 411 int reg_edx; 412 413#ifdef DEBUG 414 printf("Linux-emul(%ld): pipe(*)\n", (long)p->p_pid); 415#endif 416 417 reg_edx = p->p_retval[1]; 418 error = pipe(p, 0); 419 if (error) { 420 p->p_retval[1] = reg_edx; 421 return (error); 422 } 423 424 error = copyout(p->p_retval, args->pipefds, 2*sizeof(int)); 425 if (error) { 426 p->p_retval[1] = reg_edx; 427 return (error); 428 } 429 430 p->p_retval[1] = reg_edx; 431 p->p_retval[0] = 0; 432 return (0); 433} 434 435int 436linux_ioperm(struct proc *p, struct linux_ioperm_args *args) 437{ 438 struct sysarch_args sa; 439 struct i386_ioperm_args *iia; 440 caddr_t sg; 441 442 sg = stackgap_init(); 443 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args)); 444 iia->start = args->start; 445 iia->length = args->length; 446 iia->enable = args->enable; 447 sa.op = I386_SET_IOPERM; 448 sa.parms = (char *)iia; 449 return (sysarch(p, &sa)); 450} 451 452int 453linux_iopl(struct proc *p, struct linux_iopl_args *args) 454{ 455 int error; 456 457 if (args->level < 0 || args->level > 3) 458 return (EINVAL); 459 if ((error = suser(p)) != 0) 460 return (error); 461 if (securelevel > 0) 462 return (EPERM); 463 p->p_md.md_regs->tf_eflags = (p->p_md.md_regs->tf_eflags & ~PSL_IOPL) | 464 (args->level * (PSL_IOPL / 3)); 465 return (0); 466} 467 468int 469linux_modify_ldt(p, uap) 470 struct proc *p; 471 struct linux_modify_ldt_args *uap; 472{ 473 int error; 474 caddr_t sg; 475 struct sysarch_args args; 476 struct i386_ldt_args *ldt; 477 struct linux_descriptor ld; 478 union descriptor *desc; 479 480 sg = stackgap_init(); 481 482 if (uap->ptr == NULL) 483 return (EINVAL); 484 485 switch (uap->func) { 486 case 0x00: /* read_ldt */ 487 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 488 ldt->start = 0; 489 ldt->descs = uap->ptr; 490 ldt->num = uap->bytecount / sizeof(union descriptor); 491 args.op = I386_GET_LDT; 492 args.parms = (char*)ldt; 493 error = sysarch(p, &args); 494 p->p_retval[0] *= sizeof(union descriptor); 495 break; 496 case 0x01: /* write_ldt */ 497 case 0x11: /* write_ldt */ 498 if (uap->bytecount != sizeof(ld)) 499 return (EINVAL); 500 501 error = copyin(uap->ptr, &ld, sizeof(ld)); 502 if (error) 503 return (error); 504 505 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 506 desc = stackgap_alloc(&sg, sizeof(*desc)); 507 ldt->start = ld.entry_number; 508 ldt->descs = desc; 509 ldt->num = 1; 510 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff); 511 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 512 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff); 513 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 514 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 515 (ld.contents << 2); 516 desc->sd.sd_dpl = 3; 517 desc->sd.sd_p = (ld.seg_not_present ^ 1); 518 desc->sd.sd_xx = 0; 519 desc->sd.sd_def32 = ld.seg_32bit; 520 desc->sd.sd_gran = ld.limit_in_pages; 521 args.op = I386_SET_LDT; 522 args.parms = (char*)ldt; 523 error = sysarch(p, &args); 524 break; 525 default: 526 error = EINVAL; 527 break; 528 } 529 530 if (error == EOPNOTSUPP) { 531 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 532 error = ENOSYS; 533 } 534 535 return (error); 536} 537 538int 539linux_sigaction(struct proc *p, struct linux_sigaction_args *args) 540{ 541 linux_osigaction_t osa; 542 linux_sigaction_t act, oact; 543 int error; 544 545#ifdef DEBUG 546 printf("Linux-emul(%ld): sigaction(%d, %p, %p)\n", (long)p->p_pid, 547 args->sig, (void *)args->nsa, (void *)args->osa); 548#endif 549 550 if (args->nsa != NULL) { 551 error = copyin(args->nsa, &osa, sizeof(linux_osigaction_t)); 552 if (error) 553 return (error); 554 act.lsa_handler = osa.lsa_handler; 555 act.lsa_flags = osa.lsa_flags; 556 act.lsa_restorer = osa.lsa_restorer; 557 LINUX_SIGEMPTYSET(act.lsa_mask); 558 act.lsa_mask.__bits[0] = osa.lsa_mask; 559 } 560 561 error = linux_do_sigaction(p, args->sig, args->nsa ? &act : NULL, 562 args->osa ? &oact : NULL); 563 564 if (args->osa != NULL && !error) { 565 osa.lsa_handler = oact.lsa_handler; 566 osa.lsa_flags = oact.lsa_flags; 567 osa.lsa_restorer = oact.lsa_restorer; 568 osa.lsa_mask = oact.lsa_mask.__bits[0]; 569 error = copyout(&osa, args->osa, sizeof(linux_osigaction_t)); 570 } 571 572 return (error); 573} 574 575/* 576 * Linux has two extra args, restart and oldmask. We dont use these, 577 * but it seems that "restart" is actually a context pointer that 578 * enables the signal to happen with a different register set. 579 */ 580int 581linux_sigsuspend(struct proc *p, struct linux_sigsuspend_args *args) 582{ 583 struct sigsuspend_args bsd; 584 sigset_t *sigmask; 585 linux_sigset_t mask; 586 caddr_t sg = stackgap_init(); 587 588#ifdef DEBUG 589 printf("Linux-emul(%ld): sigsuspend(%08lx)\n", 590 (long)p->p_pid, (unsigned long)args->mask); 591#endif 592 593 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 594 LINUX_SIGEMPTYSET(mask); 595 mask.__bits[0] = args->mask; 596 linux_to_bsd_sigset(&mask, sigmask); 597 bsd.sigmask = sigmask; 598 return (sigsuspend(p, &bsd)); 599} 600 601int 602linux_rt_sigsuspend(p, uap) 603 struct proc *p; 604 struct linux_rt_sigsuspend_args *uap; 605{ 606 linux_sigset_t lmask; 607 sigset_t *bmask; 608 struct sigsuspend_args bsd; 609 caddr_t sg = stackgap_init(); 610 int error; 611 612#ifdef DEBUG 613 printf("Linux-emul(%ld): rt_sigsuspend(%p, %d)\n", (long)p->p_pid, 614 (void *)uap->newset, uap->sigsetsize); 615#endif 616 617 if (uap->sigsetsize != sizeof(linux_sigset_t)) 618 return (EINVAL); 619 620 error = copyin(uap->newset, &lmask, sizeof(linux_sigset_t)); 621 if (error) 622 return (error); 623 624 bmask = stackgap_alloc(&sg, sizeof(sigset_t)); 625 linux_to_bsd_sigset(&lmask, bmask); 626 bsd.sigmask = bmask; 627 return (sigsuspend(p, &bsd)); 628} 629 630int 631linux_pause(struct proc *p, struct linux_pause_args *args) 632{ 633 struct sigsuspend_args bsd; 634 sigset_t *sigmask; 635 caddr_t sg = stackgap_init(); 636 637#ifdef DEBUG 638 printf("Linux-emul(%d): pause()\n", p->p_pid); 639#endif 640 641 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 642 PROC_LOCK(p); 643 *sigmask = p->p_sigmask; 644 PROC_UNLOCK(p); 645 bsd.sigmask = sigmask; 646 return (sigsuspend(p, &bsd)); 647} 648 649int 650linux_sigaltstack(p, uap) 651 struct proc *p; 652 struct linux_sigaltstack_args *uap; 653{ 654 struct sigaltstack_args bsd; 655 stack_t *ss, *oss; 656 linux_stack_t lss; 657 int error; 658 caddr_t sg = stackgap_init(); 659 660#ifdef DEBUG 661 printf("Linux-emul(%ld): sigaltstack(%p, %p)\n", 662 (long)p->p_pid, uap->uss, uap->uoss); 663#endif 664 665 if (uap->uss == NULL) { 666 ss = NULL; 667 } else { 668 error = copyin(uap->uss, &lss, sizeof(linux_stack_t)); 669 if (error) 670 return (error); 671 672 ss = stackgap_alloc(&sg, sizeof(stack_t)); 673 ss->ss_sp = lss.ss_sp; 674 ss->ss_size = lss.ss_size; 675 ss->ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 676 } 677 oss = (uap->uoss != NULL) 678 ? stackgap_alloc(&sg, sizeof(stack_t)) 679 : NULL; 680 681 bsd.ss = ss; 682 bsd.oss = oss; 683 error = sigaltstack(p, &bsd); 684 685 if (!error && oss != NULL) { 686 lss.ss_sp = oss->ss_sp; 687 lss.ss_size = oss->ss_size; 688 lss.ss_flags = bsd_to_linux_sigaltstack(oss->ss_flags); 689 error = copyout(&lss, uap->uoss, sizeof(linux_stack_t)); 690 } 691 692 return (error); 693} 694