linux_machdep.c revision 1.37
1 /* $NetBSD: linux_machdep.c,v 1.37 1998/01/24 12:42:24 mycroft Exp $ */ 2 3 /* 4 * Copyright (c) 1995 Frank van der Linden 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed for the NetBSD Project 18 * by Frank van der Linden 19 * 4. The name of the author may not be used to endorse or promote products 20 * derived from this software without specific prior written permission 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 23 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 24 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 25 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 27 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 31 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 #include "opt_vm86.h" 35 #include "opt_user_ldt.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/signalvar.h> 40 #include <sys/kernel.h> 41 #include <sys/map.h> 42 #include <sys/proc.h> 43 #include <sys/user.h> 44 #include <sys/buf.h> 45 #include <sys/reboot.h> 46 #include <sys/conf.h> 47 #include <sys/exec.h> 48 #include <sys/file.h> 49 #include <sys/callout.h> 50 #include <sys/malloc.h> 51 #include <sys/mbuf.h> 52 #include <sys/msgbuf.h> 53 #include <sys/mount.h> 54 #include <sys/vnode.h> 55 #include <sys/device.h> 56 #include <sys/syscallargs.h> 57 #include <sys/filedesc.h> 58 59 #include <compat/linux/linux_types.h> 60 #include <compat/linux/linux_signal.h> 61 #include <compat/linux/linux_syscallargs.h> 62 #include <compat/linux/linux_util.h> 63 #include <compat/linux/linux_ioctl.h> 64 65 #include <machine/cpu.h> 66 #include <machine/cpufunc.h> 67 #include <machine/psl.h> 68 #include <machine/reg.h> 69 #include <machine/segments.h> 70 #include <machine/specialreg.h> 71 #include <machine/sysarch.h> 72 #include <machine/vm86.h> 73 #include <machine/vmparam.h> 74 #include <machine/linux_machdep.h> 75 76 /* 77 * To see whether pcvt is configured (for virtual console ioctl calls). 78 */ 79 #ifndef NVT 80 #include "vt.h" 81 #endif 82 #if NVT > 0 83 #include <arch/i386/isa/pcvt/pcvt_ioctl.h> 84 #endif 85 86 #ifdef USER_LDT 87 #include <machine/cpu.h> 88 int linux_read_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *, 89 register_t *)); 90 int linux_write_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *, 91 register_t *)); 92 #endif 93 94 /* 95 * Deal with some i386-specific things in the Linux emulation code. 96 */ 97 98 void 99 linux_setregs(p, epp, stack) 100 struct proc *p; 101 struct exec_package *epp; 102 u_long stack; 103 { 104 105 setregs(p, epp, stack); 106 } 107 108 /* 109 * Send an interrupt to process. 110 * 111 * Stack is set up to allow sigcode stored 112 * in u. to call routine, followed by kcall 113 * to sigreturn routine below. After sigreturn 114 * resets the signal mask, the stack, and the 115 * frame pointer, it returns to the user 116 * specified pc, psl. 117 */ 118 119 void 120 linux_sendsig(catcher, sig, mask, code) 121 sig_t catcher; 122 int sig, mask; 123 u_long code; 124 { 125 register struct proc *p = curproc; 126 register struct trapframe *tf; 127 struct linux_sigframe *fp, frame; 128 struct sigacts *psp = p->p_sigacts; 129 int oonstack; 130 extern char linux_sigcode[], linux_esigcode[]; 131 132 tf = p->p_md.md_regs; 133 oonstack = psp->ps_sigstk.ss_flags & SS_ONSTACK; 134 135 /* 136 * Allocate space for the signal handler context. 137 */ 138 if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack && 139 (psp->ps_sigonstack & sigmask(sig))) { 140 fp = (struct linux_sigframe *)(psp->ps_sigstk.ss_sp + 141 psp->ps_sigstk.ss_size - sizeof(struct linux_sigframe)); 142 psp->ps_sigstk.ss_flags |= SS_ONSTACK; 143 } else { 144 fp = (struct linux_sigframe *)tf->tf_esp - 1; 145 } 146 147 frame.sf_handler = catcher; 148 frame.sf_sig = bsd_to_linux_sig[sig]; 149 150 /* 151 * Build the signal context to be used by sigreturn. 152 */ 153 frame.sf_sc.sc_mask = mask; 154 #ifdef VM86 155 if (tf->tf_eflags & PSL_VM) { 156 frame.sf_sc.sc_gs = tf->tf_vm86_gs; 157 frame.sf_sc.sc_fs = tf->tf_vm86_fs; 158 frame.sf_sc.sc_es = tf->tf_vm86_es; 159 frame.sf_sc.sc_ds = tf->tf_vm86_ds; 160 frame.sf_sc.sc_eflags = get_vflags(p); 161 } else 162 #endif 163 { 164 __asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs)); 165 __asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs)); 166 frame.sf_sc.sc_es = tf->tf_es; 167 frame.sf_sc.sc_ds = tf->tf_ds; 168 frame.sf_sc.sc_eflags = tf->tf_eflags; 169 } 170 frame.sf_sc.sc_edi = tf->tf_edi; 171 frame.sf_sc.sc_esi = tf->tf_esi; 172 frame.sf_sc.sc_ebp = tf->tf_ebp; 173 frame.sf_sc.sc_ebx = tf->tf_ebx; 174 frame.sf_sc.sc_edx = tf->tf_edx; 175 frame.sf_sc.sc_ecx = tf->tf_ecx; 176 frame.sf_sc.sc_eax = tf->tf_eax; 177 frame.sf_sc.sc_eip = tf->tf_eip; 178 frame.sf_sc.sc_cs = tf->tf_cs; 179 frame.sf_sc.sc_esp_at_signal = tf->tf_esp; 180 frame.sf_sc.sc_ss = tf->tf_ss; 181 frame.sf_sc.sc_err = tf->tf_err; 182 frame.sf_sc.sc_trapno = tf->tf_trapno; 183 184 if (copyout(&frame, fp, sizeof(frame)) != 0) { 185 /* 186 * Process has trashed its stack; give it an illegal 187 * instruction to halt it in its tracks. 188 */ 189 sigexit(p, SIGILL); 190 /* NOTREACHED */ 191 } 192 193 /* 194 * Build context to run handler in. 195 */ 196 tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL); 197 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 198 tf->tf_eip = (int)(((char *)PS_STRINGS) - 199 (linux_esigcode - linux_sigcode)); 200 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 201 tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC); 202 tf->tf_esp = (int)fp; 203 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 204 } 205 206 /* 207 * System call to cleanup state after a signal 208 * has been taken. Reset signal mask and 209 * stack state from context left by sendsig (above). 210 * Return to previous pc and psl as specified by 211 * context left by sendsig. Check carefully to 212 * make sure that the user has not modified the 213 * psl to gain improper privileges or to cause 214 * a machine fault. 215 */ 216 int 217 linux_sys_sigreturn(p, v, retval) 218 struct proc *p; 219 void *v; 220 register_t *retval; 221 { 222 struct linux_sys_sigreturn_args /* { 223 syscallarg(struct linux_sigcontext *) scp; 224 } */ *uap = v; 225 struct linux_sigcontext *scp, context; 226 register struct trapframe *tf; 227 228 tf = p->p_md.md_regs; 229 230 /* 231 * The trampoline code hands us the context. 232 * It is unsafe to keep track of it ourselves, in the event that a 233 * program jumps out of a signal handler. 234 */ 235 scp = SCARG(uap, scp); 236 if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0) 237 return (EFAULT); 238 239 /* 240 * Restore signal context. 241 */ 242 #ifdef VM86 243 if (context.sc_eflags & PSL_VM) { 244 tf->tf_vm86_gs = context.sc_gs; 245 tf->tf_vm86_fs = context.sc_fs; 246 tf->tf_vm86_es = context.sc_es; 247 tf->tf_vm86_ds = context.sc_ds; 248 set_vflags(p, context.sc_eflags); 249 } else 250 #endif 251 { 252 /* 253 * Check for security violations. If we're returning to 254 * protected mode, the CPU will validate the segment registers 255 * automatically and generate a trap on violations. We handle 256 * the trap, rather than doing all of the checking here. 257 */ 258 if (((context.sc_eflags ^ tf->tf_eflags) & PSL_USERSTATIC) != 0 || 259 !USERMODE(context.sc_cs, context.sc_eflags)) 260 return (EINVAL); 261 262 /* %fs and %gs were restored by the trampoline. */ 263 tf->tf_es = context.sc_es; 264 tf->tf_ds = context.sc_ds; 265 tf->tf_eflags = context.sc_eflags; 266 } 267 tf->tf_edi = context.sc_edi; 268 tf->tf_esi = context.sc_esi; 269 tf->tf_ebp = context.sc_ebp; 270 tf->tf_ebx = context.sc_ebx; 271 tf->tf_edx = context.sc_edx; 272 tf->tf_ecx = context.sc_ecx; 273 tf->tf_eax = context.sc_eax; 274 tf->tf_eip = context.sc_eip; 275 tf->tf_cs = context.sc_cs; 276 tf->tf_esp = context.sc_esp_at_signal; 277 tf->tf_ss = context.sc_ss; 278 279 p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK; 280 p->p_sigmask = context.sc_mask & ~sigcantmask; 281 282 return (EJUSTRETURN); 283 } 284 285 #ifdef USER_LDT 286 287 int 288 linux_read_ldt(p, uap, retval) 289 struct proc *p; 290 struct linux_sys_modify_ldt_args /* { 291 syscallarg(int) func; 292 syscallarg(void *) ptr; 293 syscallarg(size_t) bytecount; 294 } */ *uap; 295 register_t *retval; 296 { 297 struct i386_get_ldt_args gl; 298 int error; 299 caddr_t sg; 300 char *parms; 301 302 sg = stackgap_init(p->p_emul); 303 304 gl.start = 0; 305 gl.desc = SCARG(uap, ptr); 306 gl.num = SCARG(uap, bytecount) / sizeof(union descriptor); 307 308 parms = stackgap_alloc(&sg, sizeof(gl)); 309 310 if ((error = copyout(&gl, parms, sizeof(gl))) != 0) 311 return (error); 312 313 if ((error = i386_get_ldt(p, parms, retval)) != 0) 314 return (error); 315 316 *retval *= sizeof(union descriptor); 317 return (0); 318 } 319 320 struct linux_ldt_info { 321 u_int entry_number; 322 u_long base_addr; 323 u_int limit; 324 u_int seg_32bit:1; 325 u_int contents:2; 326 u_int read_exec_only:1; 327 u_int limit_in_pages:1; 328 u_int seg_not_present:1; 329 }; 330 331 int 332 linux_write_ldt(p, uap, retval) 333 struct proc *p; 334 struct linux_sys_modify_ldt_args /* { 335 syscallarg(int) func; 336 syscallarg(void *) ptr; 337 syscallarg(size_t) bytecount; 338 } */ *uap; 339 register_t *retval; 340 { 341 struct linux_ldt_info ldt_info; 342 struct segment_descriptor sd; 343 struct i386_set_ldt_args sl; 344 int error; 345 caddr_t sg; 346 char *parms; 347 348 if (SCARG(uap, bytecount) != sizeof(ldt_info)) 349 return (EINVAL); 350 if ((error = copyin(SCARG(uap, ptr), &ldt_info, sizeof(ldt_info))) != 0) 351 return error; 352 if (ldt_info.contents == 3) 353 return (EINVAL); 354 355 sg = stackgap_init(p->p_emul); 356 357 sd.sd_lobase = ldt_info.base_addr & 0xffffff; 358 sd.sd_hibase = (ldt_info.base_addr >> 24) & 0xff; 359 sd.sd_lolimit = ldt_info.limit & 0xffff; 360 sd.sd_hilimit = (ldt_info.limit >> 16) & 0xf; 361 sd.sd_type = 362 16 | (ldt_info.contents << 2) | (!ldt_info.read_exec_only << 1); 363 sd.sd_dpl = SEL_UPL; 364 sd.sd_p = !ldt_info.seg_not_present; 365 sd.sd_def32 = ldt_info.seg_32bit; 366 sd.sd_gran = ldt_info.limit_in_pages; 367 368 sl.start = ldt_info.entry_number; 369 sl.desc = stackgap_alloc(&sg, sizeof(sd)); 370 sl.num = 1; 371 372 #if 0 373 printf("linux_write_ldt: idx=%d, base=%x, limit=%x\n", 374 ldt_info.entry_number, ldt_info.base_addr, ldt_info.limit); 375 #endif 376 377 parms = stackgap_alloc(&sg, sizeof(sl)); 378 379 if ((error = copyout(&sd, sl.desc, sizeof(sd))) != 0) 380 return (error); 381 if ((error = copyout(&sl, parms, sizeof(sl))) != 0) 382 return (error); 383 384 if ((error = i386_set_ldt(p, parms, retval)) != 0) 385 return (error); 386 387 *retval = 0; 388 return (0); 389 } 390 391 #endif /* USER_LDT */ 392 393 int 394 linux_sys_modify_ldt(p, v, retval) 395 struct proc *p; 396 void *v; 397 register_t *retval; 398 { 399 struct linux_sys_modify_ldt_args /* { 400 syscallarg(int) func; 401 syscallarg(void *) ptr; 402 syscallarg(size_t) bytecount; 403 } */ *uap = v; 404 405 switch (SCARG(uap, func)) { 406 #ifdef USER_LDT 407 case 0: 408 return (linux_read_ldt(p, uap, retval)); 409 410 case 1: 411 return (linux_write_ldt(p, uap, retval)); 412 #endif /* USER_LDT */ 413 414 default: 415 return (ENOSYS); 416 } 417 } 418 419 /* 420 * XXX Pathetic hack to make svgalib work. This will fake the major 421 * device number of an opened VT so that svgalib likes it. grmbl. 422 * Should probably do it 'wrong the right way' and use a mapping 423 * array for all major device numbers, and map linux_mknod too. 424 */ 425 dev_t 426 linux_fakedev(dev) 427 dev_t dev; 428 { 429 430 if (major(dev) == NETBSD_CONS_MAJOR) 431 return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1)); 432 return dev; 433 } 434 435 /* 436 * We come here in a last attempt to satisfy a Linux ioctl() call 437 */ 438 int 439 linux_machdepioctl(p, v, retval) 440 struct proc *p; 441 void *v; 442 register_t *retval; 443 { 444 struct linux_sys_ioctl_args /* { 445 syscallarg(int) fd; 446 syscallarg(u_long) com; 447 syscallarg(caddr_t) data; 448 } */ *uap = v; 449 struct sys_ioctl_args bia; 450 u_long com; 451 #if NVT > 0 452 int error; 453 struct vt_mode lvt; 454 caddr_t bvtp, sg; 455 #endif 456 457 SCARG(&bia, fd) = SCARG(uap, fd); 458 SCARG(&bia, data) = SCARG(uap, data); 459 com = SCARG(uap, com); 460 461 switch (com) { 462 #if NVT > 0 463 case LINUX_KDGKBMODE: 464 com = KDGKBMODE; 465 break; 466 case LINUX_KDSKBMODE: 467 com = KDSKBMODE; 468 if ((unsigned)SCARG(uap, data) == LINUX_K_MEDIUMRAW) 469 SCARG(&bia, data) = (caddr_t)K_RAW; 470 break; 471 case LINUX_KDMKTONE: 472 com = KDMKTONE; 473 break; 474 case LINUX_KDSETMODE: 475 com = KDSETMODE; 476 break; 477 case LINUX_KDENABIO: 478 com = KDENABIO; 479 break; 480 case LINUX_KDDISABIO: 481 com = KDDISABIO; 482 break; 483 case LINUX_KDGETLED: 484 com = KDGETLED; 485 break; 486 case LINUX_KDSETLED: 487 com = KDSETLED; 488 break; 489 case LINUX_VT_OPENQRY: 490 com = VT_OPENQRY; 491 break; 492 case LINUX_VT_GETMODE: 493 SCARG(&bia, com) = VT_GETMODE; 494 if ((error = sys_ioctl(p, &bia, retval))) 495 return error; 496 if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt, 497 sizeof (struct vt_mode)))) 498 return error; 499 lvt.relsig = bsd_to_linux_sig[lvt.relsig]; 500 lvt.acqsig = bsd_to_linux_sig[lvt.acqsig]; 501 lvt.frsig = bsd_to_linux_sig[lvt.frsig]; 502 return copyout((caddr_t)&lvt, SCARG(uap, data), 503 sizeof (struct vt_mode)); 504 case LINUX_VT_SETMODE: 505 com = VT_SETMODE; 506 if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt, 507 sizeof (struct vt_mode)))) 508 return error; 509 lvt.relsig = linux_to_bsd_sig[lvt.relsig]; 510 lvt.acqsig = linux_to_bsd_sig[lvt.acqsig]; 511 lvt.frsig = linux_to_bsd_sig[lvt.frsig]; 512 sg = stackgap_init(p->p_emul); 513 bvtp = stackgap_alloc(&sg, sizeof (struct vt_mode)); 514 if ((error = copyout(&lvt, bvtp, sizeof (struct vt_mode)))) 515 return error; 516 SCARG(&bia, data) = bvtp; 517 break; 518 case LINUX_VT_RELDISP: 519 com = VT_RELDISP; 520 break; 521 case LINUX_VT_ACTIVATE: 522 com = VT_ACTIVATE; 523 break; 524 case LINUX_VT_WAITACTIVE: 525 com = VT_WAITACTIVE; 526 break; 527 #endif 528 default: 529 printf("linux_machdepioctl: invalid ioctl %08lx\n", com); 530 return EINVAL; 531 } 532 SCARG(&bia, com) = com; 533 return sys_ioctl(p, &bia, retval); 534 } 535 536 /* 537 * Set I/O permissions for a process. Just set the maximum level 538 * right away (ignoring the argument), otherwise we would have 539 * to rely on I/O permission maps, which are not implemented. 540 */ 541 int 542 linux_sys_iopl(p, v, retval) 543 struct proc *p; 544 void *v; 545 register_t *retval; 546 { 547 #if 0 548 struct linux_sys_iopl_args /* { 549 syscallarg(int) level; 550 } */ *uap = v; 551 #endif 552 struct trapframe *fp = p->p_md.md_regs; 553 554 if (suser(p->p_ucred, &p->p_acflag) != 0) 555 return EPERM; 556 fp->tf_eflags |= PSL_IOPL; 557 *retval = 0; 558 return 0; 559 } 560 561 /* 562 * See above. If a root process tries to set access to an I/O port, 563 * just let it have the whole range. 564 */ 565 int 566 linux_sys_ioperm(p, v, retval) 567 struct proc *p; 568 void *v; 569 register_t *retval; 570 { 571 struct linux_sys_ioperm_args /* { 572 syscallarg(unsigned int) lo; 573 syscallarg(unsigned int) hi; 574 syscallarg(int) val; 575 } */ *uap = v; 576 struct trapframe *fp = p->p_md.md_regs; 577 578 if (suser(p->p_ucred, &p->p_acflag) != 0) 579 return EPERM; 580 if (SCARG(uap, val)) 581 fp->tf_eflags |= PSL_IOPL; 582 *retval = 0; 583 return 0; 584 } 585
Indexes created Mon Sep 29 18:09:42 GMT 2025