linux_machdep.c revision 67051
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 67051 2000-10-13 01:57:43Z gallatin $ 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 38#include <machine/frame.h> 39#include <machine/psl.h> 40#include <machine/segments.h> 41#include <machine/sysarch.h> 42 43#include <i386/linux/linux.h> 44#include <i386/linux/linux_proto.h> 45#include <compat/linux/linux_ipc.h> 46#include <compat/linux/linux_signal.h> 47#include <compat/linux/linux_util.h> 48 49struct linux_descriptor { 50 unsigned int entry_number; 51 unsigned long base_addr; 52 unsigned int limit; 53 unsigned int seg_32bit:1; 54 unsigned int contents:2; 55 unsigned int read_exec_only:1; 56 unsigned int limit_in_pages:1; 57 unsigned int seg_not_present:1; 58 unsigned int useable:1; 59}; 60 61struct linux_select_argv { 62 int nfds; 63 fd_set *readfds; 64 fd_set *writefds; 65 fd_set *exceptfds; 66 struct timeval *timeout; 67}; 68 69int 70linux_to_bsd_sigaltstack(int lsa) 71{ 72 int bsa = 0; 73 74 if (lsa & LINUX_SS_DISABLE) 75 bsa |= SS_DISABLE; 76 if (lsa & LINUX_SS_ONSTACK) 77 bsa |= SS_ONSTACK; 78 if (lsa == LINUX_SS_ONSTACK_BC) 79 bsa = SS_ONSTACK; 80 return (bsa); 81} 82 83int 84bsd_to_linux_sigaltstack(int bsa) 85{ 86 int lsa = 0; 87 88 if (bsa & SS_DISABLE) 89 lsa |= LINUX_SS_DISABLE; 90 if (bsa & SS_ONSTACK) 91 lsa |= LINUX_SS_ONSTACK; 92 return (lsa); 93} 94 95int 96linux_execve(struct proc *p, struct linux_execve_args *args) 97{ 98 struct execve_args bsd; 99 caddr_t sg; 100 101 sg = stackgap_init(); 102 CHECKALTEXIST(p, &sg, args->path); 103 104#ifdef DEBUG 105 printf("Linux-emul(%d): execve(%s)\n", 106 p->p_pid, args->path); 107#endif 108 109 bsd.fname = args->path; 110 bsd.argv = args->argp; 111 bsd.envv = args->envp; 112 return (execve(p, &bsd)); 113} 114 115int 116linux_ipc(struct proc *p, struct linux_ipc_args *args) 117{ 118 switch (args->what) { 119 case LINUX_SEMOP: 120 return (linux_semop(p, args)); 121 case LINUX_SEMGET: 122 return (linux_semget(p, args)); 123 case LINUX_SEMCTL: 124 return (linux_semctl(p, args)); 125 case LINUX_MSGSND: 126 return (linux_msgsnd(p, args)); 127 case LINUX_MSGRCV: 128 return (linux_msgrcv(p, args)); 129 case LINUX_MSGGET: 130 return (linux_msgget(p, args)); 131 case LINUX_MSGCTL: 132 return (linux_msgctl(p, args)); 133 case LINUX_SHMAT: 134 return (linux_shmat(p, args)); 135 case LINUX_SHMDT: 136 return (linux_shmdt(p, args)); 137 case LINUX_SHMGET: 138 return (linux_shmget(p, args)); 139 case LINUX_SHMCTL: 140 return (linux_shmctl(p, args)); 141 } 142 143 uprintf("LINUX: 'ipc' typ=%d not implemented\n", args->what); 144 return (ENOSYS); 145} 146 147int 148linux_select(struct proc *p, struct linux_select_args *args) 149{ 150 struct linux_select_argv linux_args; 151 struct linux_newselect_args newsel; 152 int error; 153 154#ifdef SELECT_DEBUG 155 printf("Linux-emul(%ld): select(%x)\n", (long)p->p_pid, args->ptr); 156#endif 157 158 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 159 if (error) 160 return (error); 161 162 newsel.nfds = linux_args.nfds; 163 newsel.readfds = linux_args.readfds; 164 newsel.writefds = linux_args.writefds; 165 newsel.exceptfds = linux_args.exceptfds; 166 newsel.timeout = linux_args.timeout; 167 return (linux_newselect(p, &newsel)); 168} 169 170int 171linux_fork(struct proc *p, struct linux_fork_args *args) 172{ 173 int error; 174 175#ifdef DEBUG 176 printf("Linux-emul(%ld): fork()\n", (long)p->p_pid); 177#endif 178 179 if ((error = fork(p, (struct fork_args *)args)) != 0) 180 return (error); 181 182 if (p->p_retval[1] == 1) 183 p->p_retval[0] = 0; 184 return (0); 185} 186 187int 188linux_vfork(struct proc *p, struct linux_vfork_args *args) 189{ 190 int error; 191 192#ifdef DEBUG 193 printf("Linux-emul(%ld): vfork()\n", (long)p->p_pid); 194#endif 195 196 if ((error = vfork(p, (struct vfork_args *)args)) != 0) 197 return (error); 198 /* Are we the child? */ 199 if (p->p_retval[1] == 1) 200 p->p_retval[0] = 0; 201 return (0); 202} 203 204#define CLONE_VM 0x100 205#define CLONE_FS 0x200 206#define CLONE_FILES 0x400 207#define CLONE_SIGHAND 0x800 208#define CLONE_PID 0x1000 209 210int 211linux_clone(struct proc *p, struct linux_clone_args *args) 212{ 213 int error, ff = RFPROC; 214 struct proc *p2; 215 int exit_signal; 216 vm_offset_t start; 217 struct rfork_args rf_args; 218 219#ifdef DEBUG 220 if (args->flags & CLONE_PID) 221 printf("linux_clone(%ld): CLONE_PID not yet supported\n", 222 (long)p->p_pid); 223 printf("linux_clone(%ld): invoked with flags %x and stack %x\n", 224 (long)p->p_pid, (unsigned int)args->flags, 225 (unsigned int)args->stack); 226#endif 227 228 if (!args->stack) 229 return (EINVAL); 230 231 exit_signal = args->flags & 0x000000ff; 232 if (exit_signal >= LINUX_NSIG) 233 return (EINVAL); 234 235 if (exit_signal <= LINUX_SIGTBLSZ) 236 exit_signal = linux_to_bsd_signal[_SIG_IDX(exit_signal)]; 237 238 /* RFTHREAD probably not necessary here, but it shouldn't hurt */ 239 ff |= RFTHREAD; 240 241 if (args->flags & CLONE_VM) 242 ff |= RFMEM; 243 if (args->flags & CLONE_SIGHAND) 244 ff |= RFSIGSHARE; 245 if (!(args->flags & CLONE_FILES)) 246 ff |= RFFDG; 247 248 error = 0; 249 start = 0; 250 251 rf_args.flags = ff; 252 if ((error = rfork(p, &rf_args)) != 0) 253 return (error); 254 255 p2 = pfind(p->p_retval[0]); 256 if (p2 == 0) 257 return (ESRCH); 258 259 p2->p_sigparent = exit_signal; 260 p2->p_md.md_regs->tf_esp = (unsigned int)args->stack; 261 262#ifdef DEBUG 263 printf ("linux_clone(%ld): successful rfork to %ld\n", (long)p->p_pid, 264 (long)p2->p_pid); 265#endif 266 267 return (0); 268} 269 270/* XXX move */ 271struct linux_mmap_argv { 272 linux_caddr_t addr; 273 int len; 274 int prot; 275 int flags; 276 int fd; 277 int pos; 278}; 279 280#define STACK_SIZE (2 * 1024 * 1024) 281#define GUARD_SIZE (4 * PAGE_SIZE) 282 283int 284linux_mmap(struct proc *p, struct linux_mmap_args *args) 285{ 286 struct mmap_args /* { 287 caddr_t addr; 288 size_t len; 289 int prot; 290 int flags; 291 int fd; 292 long pad; 293 off_t pos; 294 } */ bsd_args; 295 int error; 296 struct linux_mmap_argv linux_args; 297 298 error = copyin(args->ptr, &linux_args, sizeof(linux_args)); 299 if (error) 300 return (error); 301 302#ifdef DEBUG 303 printf("Linux-emul(%ld): mmap(%p, %d, %d, 0x%08x, %d, %d)", 304 (long)p->p_pid, (void *)linux_args.addr, linux_args.len, 305 linux_args.prot, linux_args.flags, linux_args.fd, linux_args.pos); 306#endif 307 308 bsd_args.flags = 0; 309 if (linux_args.flags & LINUX_MAP_SHARED) 310 bsd_args.flags |= MAP_SHARED; 311 if (linux_args.flags & LINUX_MAP_PRIVATE) 312 bsd_args.flags |= MAP_PRIVATE; 313 if (linux_args.flags & LINUX_MAP_FIXED) 314 bsd_args.flags |= MAP_FIXED; 315 if (linux_args.flags & LINUX_MAP_ANON) 316 bsd_args.flags |= MAP_ANON; 317 if (linux_args.flags & LINUX_MAP_GROWSDOWN) { 318 bsd_args.flags |= MAP_STACK; 319 320 /* The linux MAP_GROWSDOWN option does not limit auto 321 * growth of the region. Linux mmap with this option 322 * takes as addr the inital BOS, and as len, the initial 323 * region size. It can then grow down from addr without 324 * limit. However, linux threads has an implicit internal 325 * limit to stack size of STACK_SIZE. Its just not 326 * enforced explicitly in linux. But, here we impose 327 * a limit of (STACK_SIZE - GUARD_SIZE) on the stack 328 * region, since we can do this with our mmap. 329 * 330 * Our mmap with MAP_STACK takes addr as the maximum 331 * downsize limit on BOS, and as len the max size of 332 * the region. It them maps the top SGROWSIZ bytes, 333 * and autgrows the region down, up to the limit 334 * in addr. 335 * 336 * If we don't use the MAP_STACK option, the effect 337 * of this code is to allocate a stack region of a 338 * fixed size of (STACK_SIZE - GUARD_SIZE). 339 */ 340 341 /* This gives us TOS */ 342 bsd_args.addr = linux_args.addr + linux_args.len; 343 344 /* This gives us our maximum stack size */ 345 if (linux_args.len > STACK_SIZE - GUARD_SIZE) 346 bsd_args.len = linux_args.len; 347 else 348 bsd_args.len = STACK_SIZE - GUARD_SIZE; 349 350 /* This gives us a new BOS. If we're using VM_STACK, then 351 * mmap will just map the top SGROWSIZ bytes, and let 352 * the stack grow down to the limit at BOS. If we're 353 * not using VM_STACK we map the full stack, since we 354 * don't have a way to autogrow it. 355 */ 356 bsd_args.addr -= bsd_args.len; 357 } else { 358 bsd_args.addr = linux_args.addr; 359 bsd_args.len = linux_args.len; 360 } 361 362 bsd_args.prot = linux_args.prot | PROT_READ; /* always required */ 363 if (linux_args.flags & LINUX_MAP_ANON) 364 bsd_args.fd = -1; 365 else 366 bsd_args.fd = linux_args.fd; 367 bsd_args.pos = linux_args.pos; 368 bsd_args.pad = 0; 369 370#ifdef DEBUG 371 printf("-> (%p, %d, %d, 0x%08x, %d, %d)\n", (void *)bsd_args.addr, 372 bsd_args.len, bsd_args.prot, bsd_args.flags, bsd_args.fd, 373 (int)bsd_args.pos); 374#endif 375 376 return (mmap(p, &bsd_args)); 377} 378 379int 380linux_pipe(struct proc *p, struct linux_pipe_args *args) 381{ 382 int error; 383 int reg_edx; 384 385#ifdef DEBUG 386 printf("Linux-emul(%ld): pipe(*)\n", (long)p->p_pid); 387#endif 388 389 reg_edx = p->p_retval[1]; 390 error = pipe(p, 0); 391 if (error) { 392 p->p_retval[1] = reg_edx; 393 return (error); 394 } 395 396 error = copyout(p->p_retval, args->pipefds, 2*sizeof(int)); 397 if (error) { 398 p->p_retval[1] = reg_edx; 399 return (error); 400 } 401 402 p->p_retval[1] = reg_edx; 403 p->p_retval[0] = 0; 404 return (0); 405} 406 407int 408linux_ioperm(struct proc *p, struct linux_ioperm_args *args) 409{ 410 struct sysarch_args sa; 411 struct i386_ioperm_args *iia; 412 caddr_t sg; 413 414 sg = stackgap_init(); 415 iia = stackgap_alloc(&sg, sizeof(struct i386_ioperm_args)); 416 iia->start = args->start; 417 iia->length = args->length; 418 iia->enable = args->enable; 419 sa.op = I386_SET_IOPERM; 420 sa.parms = (char *)iia; 421 return (sysarch(p, &sa)); 422} 423 424int 425linux_iopl(struct proc *p, struct linux_iopl_args *args) 426{ 427 int error; 428 429 if (args->level < 0 || args->level > 3) 430 return (EINVAL); 431 if ((error = suser(p)) != 0) 432 return (error); 433 if (securelevel > 0) 434 return (EPERM); 435 p->p_md.md_regs->tf_eflags = (p->p_md.md_regs->tf_eflags & ~PSL_IOPL) | 436 (args->level * (PSL_IOPL / 3)); 437 return (0); 438} 439 440int 441linux_modify_ldt(p, uap) 442 struct proc *p; 443 struct linux_modify_ldt_args *uap; 444{ 445 int error; 446 caddr_t sg; 447 struct sysarch_args args; 448 struct i386_ldt_args *ldt; 449 struct linux_descriptor ld; 450 union descriptor *desc; 451 452 sg = stackgap_init(); 453 454 if (uap->ptr == NULL) 455 return (EINVAL); 456 457 switch (uap->func) { 458 case 0x00: /* read_ldt */ 459 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 460 ldt->start = 0; 461 ldt->descs = uap->ptr; 462 ldt->num = uap->bytecount / sizeof(union descriptor); 463 args.op = I386_GET_LDT; 464 args.parms = (char*)ldt; 465 error = sysarch(p, &args); 466 p->p_retval[0] *= sizeof(union descriptor); 467 break; 468 case 0x01: /* write_ldt */ 469 case 0x11: /* write_ldt */ 470 if (uap->bytecount != sizeof(ld)) 471 return (EINVAL); 472 473 error = copyin(uap->ptr, &ld, sizeof(ld)); 474 if (error) 475 return (error); 476 477 ldt = stackgap_alloc(&sg, sizeof(*ldt)); 478 desc = stackgap_alloc(&sg, sizeof(*desc)); 479 ldt->start = ld.entry_number; 480 ldt->descs = desc; 481 ldt->num = 1; 482 desc->sd.sd_lolimit = (ld.limit & 0x0000ffff); 483 desc->sd.sd_hilimit = (ld.limit & 0x000f0000) >> 16; 484 desc->sd.sd_lobase = (ld.base_addr & 0x00ffffff); 485 desc->sd.sd_hibase = (ld.base_addr & 0xff000000) >> 24; 486 desc->sd.sd_type = SDT_MEMRO | ((ld.read_exec_only ^ 1) << 1) | 487 (ld.contents << 2); 488 desc->sd.sd_dpl = 3; 489 desc->sd.sd_p = (ld.seg_not_present ^ 1); 490 desc->sd.sd_xx = 0; 491 desc->sd.sd_def32 = ld.seg_32bit; 492 desc->sd.sd_gran = ld.limit_in_pages; 493 args.op = I386_SET_LDT; 494 args.parms = (char*)ldt; 495 error = sysarch(p, &args); 496 break; 497 default: 498 error = EINVAL; 499 break; 500 } 501 502 if (error == EOPNOTSUPP) { 503 printf("linux: modify_ldt needs kernel option USER_LDT\n"); 504 error = ENOSYS; 505 } 506 507 return (error); 508} 509 510int 511linux_sigaction(struct proc *p, struct linux_sigaction_args *args) 512{ 513 linux_osigaction_t osa; 514 linux_sigaction_t act, oact; 515 int error; 516 517#ifdef DEBUG 518 printf("Linux-emul(%ld): sigaction(%d, %p, %p)\n", (long)p->p_pid, 519 args->sig, (void *)args->nsa, (void *)args->osa); 520#endif 521 522 if (args->nsa != NULL) { 523 error = copyin(args->nsa, &osa, sizeof(linux_osigaction_t)); 524 if (error) 525 return (error); 526 act.lsa_handler = osa.lsa_handler; 527 act.lsa_flags = osa.lsa_flags; 528 act.lsa_restorer = osa.lsa_restorer; 529 LINUX_SIGEMPTYSET(act.lsa_mask); 530 act.lsa_mask.__bits[0] = osa.lsa_mask; 531 } 532 533 error = linux_do_sigaction(p, args->sig, args->nsa ? &act : NULL, 534 args->osa ? &oact : NULL); 535 536 if (args->osa != NULL && !error) { 537 osa.lsa_handler = oact.lsa_handler; 538 osa.lsa_flags = oact.lsa_flags; 539 osa.lsa_restorer = oact.lsa_restorer; 540 osa.lsa_mask = oact.lsa_mask.__bits[0]; 541 error = copyout(&osa, args->osa, sizeof(linux_osigaction_t)); 542 } 543 544 return (error); 545} 546 547/* 548 * Linux has two extra args, restart and oldmask. We dont use these, 549 * but it seems that "restart" is actually a context pointer that 550 * enables the signal to happen with a different register set. 551 */ 552int 553linux_sigsuspend(struct proc *p, struct linux_sigsuspend_args *args) 554{ 555 struct sigsuspend_args bsd; 556 sigset_t *sigmask; 557 linux_sigset_t mask; 558 caddr_t sg = stackgap_init(); 559 560#ifdef DEBUG 561 printf("Linux-emul(%ld): sigsuspend(%08lx)\n", 562 (long)p->p_pid, (unsigned long)args->mask); 563#endif 564 565 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 566 LINUX_SIGEMPTYSET(mask); 567 mask.__bits[0] = args->mask; 568 linux_to_bsd_sigset(&mask, sigmask); 569 bsd.sigmask = sigmask; 570 return (sigsuspend(p, &bsd)); 571} 572 573int 574linux_rt_sigsuspend(p, uap) 575 struct proc *p; 576 struct linux_rt_sigsuspend_args *uap; 577{ 578 linux_sigset_t lmask; 579 sigset_t *bmask; 580 struct sigsuspend_args bsd; 581 caddr_t sg = stackgap_init(); 582 int error; 583 584#ifdef DEBUG 585 printf("Linux-emul(%ld): rt_sigsuspend(%p, %d)\n", (long)p->p_pid, 586 (void *)uap->newset, uap->sigsetsize); 587#endif 588 589 if (uap->sigsetsize != sizeof(linux_sigset_t)) 590 return (EINVAL); 591 592 error = copyin(uap->newset, &lmask, sizeof(linux_sigset_t)); 593 if (error) 594 return (error); 595 596 bmask = stackgap_alloc(&sg, sizeof(sigset_t)); 597 linux_to_bsd_sigset(&lmask, bmask); 598 bsd.sigmask = bmask; 599 return (sigsuspend(p, &bsd)); 600} 601 602int 603linux_pause(struct proc *p, struct linux_pause_args *args) 604{ 605 struct sigsuspend_args bsd; 606 sigset_t *sigmask; 607 caddr_t sg = stackgap_init(); 608 609#ifdef DEBUG 610 printf("Linux-emul(%d): pause()\n", p->p_pid); 611#endif 612 613 sigmask = stackgap_alloc(&sg, sizeof(sigset_t)); 614 *sigmask = p->p_sigmask; 615 bsd.sigmask = sigmask; 616 return (sigsuspend(p, &bsd)); 617} 618 619int 620linux_sigaltstack(p, uap) 621 struct proc *p; 622 struct linux_sigaltstack_args *uap; 623{ 624 struct sigaltstack_args bsd; 625 stack_t *ss, *oss; 626 linux_stack_t lss; 627 int error; 628 caddr_t sg = stackgap_init(); 629 630#ifdef DEBUG 631 printf("Linux-emul(%ld): sigaltstack(%p, %p)\n", 632 (long)p->p_pid, uap->uss, uap->uoss); 633#endif 634 635 if (uap->uss == NULL) { 636 ss = NULL; 637 } else { 638 error = copyin(uap->uss, &lss, sizeof(linux_stack_t)); 639 if (error) 640 return (error); 641 642 ss = stackgap_alloc(&sg, sizeof(stack_t)); 643 ss->ss_sp = lss.ss_sp; 644 ss->ss_size = (lss.ss_size >= LINUX_MINSIGSTKSZ && 645 lss.ss_size < MINSIGSTKSZ) ? MINSIGSTKSZ : lss.ss_size; 646 ss->ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags); 647 } 648 oss = (uap->uoss != NULL) 649 ? stackgap_alloc(&sg, sizeof(stack_t)) 650 : NULL; 651 652 bsd.ss = ss; 653 bsd.oss = oss; 654 error = sigaltstack(p, &bsd); 655 656 if (!error && oss != NULL) { 657 lss.ss_sp = oss->ss_sp; 658 lss.ss_size = oss->ss_size; 659 lss.ss_flags = bsd_to_linux_sigaltstack(oss->ss_flags); 660 error = copyout(&lss, uap->uoss, sizeof(linux_stack_t)); 661 } 662 663 return (error); 664} 665