linux_machdep.c revision 1.33.12.1
1 /* $NetBSD: linux_machdep.c,v 1.33.12.1 1997/09/08 23:35:26 thorpej 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 <sys/param.h> 35 #include <sys/systm.h> 36 #include <sys/signalvar.h> 37 #include <sys/kernel.h> 38 #include <sys/map.h> 39 #include <sys/proc.h> 40 #include <sys/user.h> 41 #include <sys/buf.h> 42 #include <sys/reboot.h> 43 #include <sys/conf.h> 44 #include <sys/exec.h> 45 #include <sys/file.h> 46 #include <sys/callout.h> 47 #include <sys/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/msgbuf.h> 50 #include <sys/mount.h> 51 #include <sys/vnode.h> 52 #include <sys/device.h> 53 #include <sys/sysctl.h> 54 #include <sys/syscallargs.h> 55 #include <sys/filedesc.h> 56 57 #include <compat/linux/linux_types.h> 58 #include <compat/linux/linux_signal.h> 59 #include <compat/linux/linux_syscallargs.h> 60 #include <compat/linux/linux_util.h> 61 #include <compat/linux/linux_ioctl.h> 62 63 #include <machine/cpu.h> 64 #include <machine/cpufunc.h> 65 #include <machine/psl.h> 66 #include <machine/reg.h> 67 #include <machine/segments.h> 68 #include <machine/specialreg.h> 69 #include <machine/sysarch.h> 70 #include <machine/vm86.h> 71 #include <machine/linux_machdep.h> 72 73 /* 74 * To see whether pcvt is configured (for virtual console ioctl calls). 75 */ 76 #ifndef NVT 77 #include "vt.h" 78 #endif 79 #if NVT > 0 80 #include <arch/i386/isa/pcvt/pcvt_ioctl.h> 81 #endif 82 83 #ifdef USER_LDT 84 #include <machine/cpu.h> 85 int linux_read_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *, 86 register_t *)); 87 int linux_write_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *, 88 register_t *)); 89 #endif 90 91 /* 92 * Deal with some i386-specific things in the Linux emulation code. 93 * This means just signals for now, will include stuff like 94 * I/O map permissions and V86 mode sometime. 95 */ 96 97 /* 98 * Send an interrupt to process. 99 * 100 * Stack is set up to allow sigcode stored 101 * in u. to call routine, followed by kcall 102 * to sigreturn routine below. After sigreturn 103 * resets the signal mask, the stack, and the 104 * frame pointer, it returns to the user 105 * specified pc, psl. 106 */ 107 108 void 109 linux_sendsig(catcher, sig, mask, code) 110 sig_t catcher; 111 int sig, mask; 112 u_long code; 113 { 114 register struct proc *p = curproc; 115 register struct trapframe *tf; 116 struct linux_sigframe *fp, frame; 117 struct sigacts *psp = p->p_sigacts; 118 struct sigaction *sa = &psp->ps_sigact[sig]; 119 int oonstack; 120 extern char linux_sigcode[], linux_esigcode[]; 121 122 tf = p->p_md.md_regs; 123 oonstack = p->p_sigstk.ss_flags & SS_ONSTACK; 124 125 /* 126 * Allocate space for the signal handler context. 127 */ 128 if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack && 129 (sa->sa_flags & SA_ONSTACK)) { 130 fp = (struct linux_sigframe *)(p->p_sigstk.ss_sp + 131 p->p_sigstk.ss_size - sizeof(struct linux_sigframe)); 132 p->p_sigstk.ss_flags |= SS_ONSTACK; 133 } else { 134 fp = (struct linux_sigframe *)tf->tf_esp - 1; 135 } 136 137 frame.sf_handler = catcher; 138 frame.sf_sig = bsd_to_linux_sig[sig]; 139 140 /* 141 * Build the signal context to be used by sigreturn. 142 */ 143 frame.sf_sc.sc_mask = mask; 144 #ifdef VM86 145 if (tf->tf_eflags & PSL_VM) { 146 frame.sf_sc.sc_gs = tf->tf_vm86_gs; 147 frame.sf_sc.sc_fs = tf->tf_vm86_fs; 148 frame.sf_sc.sc_es = tf->tf_vm86_es; 149 frame.sf_sc.sc_ds = tf->tf_vm86_ds; 150 frame.sf_sc.sc_eflags = get_vflags(p); 151 } else 152 #endif 153 { 154 __asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs)); 155 __asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs)); 156 frame.sf_sc.sc_es = tf->tf_es; 157 frame.sf_sc.sc_ds = tf->tf_ds; 158 frame.sf_sc.sc_eflags = tf->tf_eflags; 159 } 160 frame.sf_sc.sc_edi = tf->tf_edi; 161 frame.sf_sc.sc_esi = tf->tf_esi; 162 frame.sf_sc.sc_ebp = tf->tf_ebp; 163 frame.sf_sc.sc_ebx = tf->tf_ebx; 164 frame.sf_sc.sc_edx = tf->tf_edx; 165 frame.sf_sc.sc_ecx = tf->tf_ecx; 166 frame.sf_sc.sc_eax = tf->tf_eax; 167 frame.sf_sc.sc_eip = tf->tf_eip; 168 frame.sf_sc.sc_cs = tf->tf_cs; 169 frame.sf_sc.sc_esp_at_signal = tf->tf_esp; 170 frame.sf_sc.sc_ss = tf->tf_ss; 171 frame.sf_sc.sc_err = tf->tf_err; 172 frame.sf_sc.sc_trapno = tf->tf_trapno; 173 174 if (copyout(&frame, fp, sizeof(frame)) != 0) { 175 /* 176 * Process has trashed its stack; give it an illegal 177 * instruction to halt it in its tracks. 178 */ 179 sigexit(p, SIGILL); 180 /* NOTREACHED */ 181 } 182 183 /* 184 * Build context to run handler in. 185 */ 186 tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL); 187 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 188 tf->tf_eip = (int)(((char *)PS_STRINGS) - 189 (linux_esigcode - linux_sigcode)); 190 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 191 tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC); 192 tf->tf_esp = (int)fp; 193 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 194 } 195 196 /* 197 * System call to cleanup state after a signal 198 * has been taken. Reset signal mask and 199 * stack state from context left by sendsig (above). 200 * Return to previous pc and psl as specified by 201 * context left by sendsig. Check carefully to 202 * make sure that the user has not modified the 203 * psl to gain improper privileges or to cause 204 * a machine fault. 205 */ 206 int 207 linux_sys_sigreturn(p, v, retval) 208 struct proc *p; 209 void *v; 210 register_t *retval; 211 { 212 struct linux_sys_sigreturn_args /* { 213 syscallarg(struct linux_sigcontext *) scp; 214 } */ *uap = v; 215 struct linux_sigcontext *scp, context; 216 register struct trapframe *tf; 217 218 tf = p->p_md.md_regs; 219 220 /* 221 * The trampoline code hands us the context. 222 * It is unsafe to keep track of it ourselves, in the event that a 223 * program jumps out of a signal handler. 224 */ 225 scp = SCARG(uap, scp); 226 if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0) 227 return (EFAULT); 228 229 /* 230 * Restore signal context. 231 */ 232 #ifdef VM86 233 if (context.sc_eflags & PSL_VM) { 234 tf->tf_vm86_gs = context.sc_gs; 235 tf->tf_vm86_fs = context.sc_fs; 236 tf->tf_vm86_es = context.sc_es; 237 tf->tf_vm86_ds = context.sc_ds; 238 set_vflags(p, context.sc_eflags); 239 } else 240 #endif 241 { 242 /* 243 * Check for security violations. If we're returning to 244 * protected mode, the CPU will validate the segment registers 245 * automatically and generate a trap on violations. We handle 246 * the trap, rather than doing all of the checking here. 247 */ 248 if (((context.sc_eflags ^ tf->tf_eflags) & PSL_USERSTATIC) != 0 || 249 !USERMODE(context.sc_cs, context.sc_eflags)) 250 return (EINVAL); 251 252 /* %fs and %gs were restored by the trampoline. */ 253 tf->tf_es = context.sc_es; 254 tf->tf_ds = context.sc_ds; 255 tf->tf_eflags = context.sc_eflags; 256 } 257 tf->tf_edi = context.sc_edi; 258 tf->tf_esi = context.sc_esi; 259 tf->tf_ebp = context.sc_ebp; 260 tf->tf_ebx = context.sc_ebx; 261 tf->tf_edx = context.sc_edx; 262 tf->tf_ecx = context.sc_ecx; 263 tf->tf_eax = context.sc_eax; 264 tf->tf_eip = context.sc_eip; 265 tf->tf_cs = context.sc_cs; 266 tf->tf_esp = context.sc_esp_at_signal; 267 tf->tf_ss = context.sc_ss; 268 269 p->p_sigstk.ss_flags &= ~SS_ONSTACK; 270 p->p_sigmask = context.sc_mask & ~sigcantmask; 271 272 return (EJUSTRETURN); 273 } 274 275 #ifdef USER_LDT 276 277 int 278 linux_read_ldt(p, uap, retval) 279 struct proc *p; 280 struct linux_sys_modify_ldt_args /* { 281 syscallarg(int) func; 282 syscallarg(void *) ptr; 283 syscallarg(size_t) bytecount; 284 } */ *uap; 285 register_t *retval; 286 { 287 struct i386_get_ldt_args gl; 288 int error; 289 caddr_t sg; 290 char *parms; 291 292 sg = stackgap_init(p->p_emul); 293 294 gl.start = 0; 295 gl.desc = SCARG(uap, ptr); 296 gl.num = SCARG(uap, bytecount) / sizeof(union descriptor); 297 298 parms = stackgap_alloc(&sg, sizeof(gl)); 299 300 if ((error = copyout(&gl, parms, sizeof(gl))) != 0) 301 return (error); 302 303 if ((error = i386_get_ldt(p, parms, retval)) != 0) 304 return (error); 305 306 *retval *= sizeof(union descriptor); 307 return (0); 308 } 309 310 struct linux_ldt_info { 311 u_int entry_number; 312 u_long base_addr; 313 u_int limit; 314 u_int seg_32bit:1; 315 u_int contents:2; 316 u_int read_exec_only:1; 317 u_int limit_in_pages:1; 318 u_int seg_not_present:1; 319 }; 320 321 int 322 linux_write_ldt(p, uap, retval) 323 struct proc *p; 324 struct linux_sys_modify_ldt_args /* { 325 syscallarg(int) func; 326 syscallarg(void *) ptr; 327 syscallarg(size_t) bytecount; 328 } */ *uap; 329 register_t *retval; 330 { 331 struct linux_ldt_info ldt_info; 332 struct segment_descriptor sd; 333 struct i386_set_ldt_args sl; 334 int error; 335 caddr_t sg; 336 char *parms; 337 338 if (SCARG(uap, bytecount) != sizeof(ldt_info)) 339 return (EINVAL); 340 if ((error = copyin(SCARG(uap, ptr), &ldt_info, sizeof(ldt_info))) != 0) 341 return error; 342 if (ldt_info.contents == 3) 343 return (EINVAL); 344 345 sg = stackgap_init(p->p_emul); 346 347 sd.sd_lobase = ldt_info.base_addr & 0xffffff; 348 sd.sd_hibase = (ldt_info.base_addr >> 24) & 0xff; 349 sd.sd_lolimit = ldt_info.limit & 0xffff; 350 sd.sd_hilimit = (ldt_info.limit >> 16) & 0xf; 351 sd.sd_type = 352 16 | (ldt_info.contents << 2) | (!ldt_info.read_exec_only << 1); 353 sd.sd_dpl = SEL_UPL; 354 sd.sd_p = !ldt_info.seg_not_present; 355 sd.sd_def32 = ldt_info.seg_32bit; 356 sd.sd_gran = ldt_info.limit_in_pages; 357 358 sl.start = ldt_info.entry_number; 359 sl.desc = stackgap_alloc(&sg, sizeof(sd)); 360 sl.num = 1; 361 362 #if 0 363 printf("linux_write_ldt: idx=%d, base=%x, limit=%x\n", 364 ldt_info.entry_number, ldt_info.base_addr, ldt_info.limit); 365 #endif 366 367 parms = stackgap_alloc(&sg, sizeof(sl)); 368 369 if ((error = copyout(&sd, sl.desc, sizeof(sd))) != 0) 370 return (error); 371 if ((error = copyout(&sl, parms, sizeof(sl))) != 0) 372 return (error); 373 374 if ((error = i386_set_ldt(p, parms, retval)) != 0) 375 return (error); 376 377 *retval = 0; 378 return (0); 379 } 380 381 #endif /* USER_LDT */ 382 383 int 384 linux_sys_modify_ldt(p, v, retval) 385 struct proc *p; 386 void *v; 387 register_t *retval; 388 { 389 struct linux_sys_modify_ldt_args /* { 390 syscallarg(int) func; 391 syscallarg(void *) ptr; 392 syscallarg(size_t) bytecount; 393 } */ *uap = v; 394 395 switch (SCARG(uap, func)) { 396 #ifdef USER_LDT 397 case 0: 398 return (linux_read_ldt(p, uap, retval)); 399 400 case 1: 401 return (linux_write_ldt(p, uap, retval)); 402 #endif /* USER_LDT */ 403 404 default: 405 return (ENOSYS); 406 } 407 } 408 409 /* 410 * XXX Pathetic hack to make svgalib work. This will fake the major 411 * device number of an opened VT so that svgalib likes it. grmbl. 412 * Should probably do it 'wrong the right way' and use a mapping 413 * array for all major device numbers, and map linux_mknod too. 414 */ 415 dev_t 416 linux_fakedev(dev) 417 dev_t dev; 418 { 419 420 if (major(dev) == NETBSD_CONS_MAJOR) 421 return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1)); 422 return dev; 423 } 424 425 /* 426 * We come here in a last attempt to satisfy a Linux ioctl() call 427 */ 428 int 429 linux_machdepioctl(p, v, retval) 430 struct proc *p; 431 void *v; 432 register_t *retval; 433 { 434 struct linux_sys_ioctl_args /* { 435 syscallarg(int) fd; 436 syscallarg(u_long) com; 437 syscallarg(caddr_t) data; 438 } */ *uap = v; 439 struct sys_ioctl_args bia; 440 u_long com; 441 #if NVT > 0 442 int error; 443 struct vt_mode lvt; 444 caddr_t bvtp, sg; 445 #endif 446 447 SCARG(&bia, fd) = SCARG(uap, fd); 448 SCARG(&bia, data) = SCARG(uap, data); 449 com = SCARG(uap, com); 450 451 switch (com) { 452 #if NVT > 0 453 case LINUX_KDGKBMODE: 454 com = KDGKBMODE; 455 break; 456 case LINUX_KDSKBMODE: 457 com = KDSKBMODE; 458 if ((unsigned)SCARG(uap, data) == LINUX_K_MEDIUMRAW) 459 SCARG(&bia, data) = (caddr_t)K_RAW; 460 break; 461 case LINUX_KDMKTONE: 462 com = KDMKTONE; 463 break; 464 case LINUX_KDSETMODE: 465 com = KDSETMODE; 466 break; 467 case LINUX_KDENABIO: 468 com = KDENABIO; 469 break; 470 case LINUX_KDDISABIO: 471 com = KDDISABIO; 472 break; 473 case LINUX_KDGETLED: 474 com = KDGETLED; 475 break; 476 case LINUX_KDSETLED: 477 com = KDSETLED; 478 break; 479 case LINUX_VT_OPENQRY: 480 com = VT_OPENQRY; 481 break; 482 case LINUX_VT_GETMODE: 483 SCARG(&bia, com) = VT_GETMODE; 484 if ((error = sys_ioctl(p, &bia, retval))) 485 return error; 486 if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt, 487 sizeof (struct vt_mode)))) 488 return error; 489 lvt.relsig = bsd_to_linux_sig[lvt.relsig]; 490 lvt.acqsig = bsd_to_linux_sig[lvt.acqsig]; 491 lvt.frsig = bsd_to_linux_sig[lvt.frsig]; 492 return copyout((caddr_t)&lvt, SCARG(uap, data), 493 sizeof (struct vt_mode)); 494 case LINUX_VT_SETMODE: 495 com = VT_SETMODE; 496 if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt, 497 sizeof (struct vt_mode)))) 498 return error; 499 lvt.relsig = linux_to_bsd_sig[lvt.relsig]; 500 lvt.acqsig = linux_to_bsd_sig[lvt.acqsig]; 501 lvt.frsig = linux_to_bsd_sig[lvt.frsig]; 502 sg = stackgap_init(p->p_emul); 503 bvtp = stackgap_alloc(&sg, sizeof (struct vt_mode)); 504 if ((error = copyout(&lvt, bvtp, sizeof (struct vt_mode)))) 505 return error; 506 SCARG(&bia, data) = bvtp; 507 break; 508 case LINUX_VT_RELDISP: 509 com = VT_RELDISP; 510 break; 511 case LINUX_VT_ACTIVATE: 512 com = VT_ACTIVATE; 513 break; 514 case LINUX_VT_WAITACTIVE: 515 com = VT_WAITACTIVE; 516 break; 517 #endif 518 default: 519 printf("linux_machdepioctl: invalid ioctl %08lx\n", com); 520 return EINVAL; 521 } 522 SCARG(&bia, com) = com; 523 return sys_ioctl(p, &bia, retval); 524 } 525 526 /* 527 * Set I/O permissions for a process. Just set the maximum level 528 * right away (ignoring the argument), otherwise we would have 529 * to rely on I/O permission maps, which are not implemented. 530 */ 531 int 532 linux_sys_iopl(p, v, retval) 533 struct proc *p; 534 void *v; 535 register_t *retval; 536 { 537 #if 0 538 struct linux_sys_iopl_args /* { 539 syscallarg(int) level; 540 } */ *uap = v; 541 #endif 542 struct trapframe *fp = p->p_md.md_regs; 543 544 if (suser(p->p_ucred, &p->p_acflag) != 0) 545 return EPERM; 546 fp->tf_eflags |= PSL_IOPL; 547 *retval = 0; 548 return 0; 549 } 550 551 /* 552 * See above. If a root process tries to set access to an I/O port, 553 * just let it have the whole range. 554 */ 555 int 556 linux_sys_ioperm(p, v, retval) 557 struct proc *p; 558 void *v; 559 register_t *retval; 560 { 561 struct linux_sys_ioperm_args /* { 562 syscallarg(unsigned int) lo; 563 syscallarg(unsigned int) hi; 564 syscallarg(int) val; 565 } */ *uap = v; 566 struct trapframe *fp = p->p_md.md_regs; 567 568 if (suser(p->p_ucred, &p->p_acflag) != 0) 569 return EPERM; 570 if (SCARG(uap, val)) 571 fp->tf_eflags |= PSL_IOPL; 572 *retval = 0; 573 return 0; 574 } 575
Indexes created Wed Oct 01 19:09:53 GMT 2025