linux_machdep.c revision 1.31
1 /* $NetBSD: linux_machdep.c,v 1.31 1996/10/12 02:20:37 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 #include "vt.h" 77 #if NVT > 0 78 #include <arch/i386/isa/pcvt/pcvt_ioctl.h> 79 #endif 80 81 #ifdef USER_LDT 82 #include <machine/cpu.h> 83 int linux_read_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *, 84 register_t *)); 85 int linux_write_ldt __P((struct proc *, struct linux_sys_modify_ldt_args *, 86 register_t *)); 87 #endif 88 89 /* 90 * Deal with some i386-specific things in the Linux emulation code. 91 * This means just signals for now, will include stuff like 92 * I/O map permissions and V86 mode sometime. 93 */ 94 95 /* 96 * Send an interrupt to process. 97 * 98 * Stack is set up to allow sigcode stored 99 * in u. to call routine, followed by kcall 100 * to sigreturn routine below. After sigreturn 101 * resets the signal mask, the stack, and the 102 * frame pointer, it returns to the user 103 * specified pc, psl. 104 */ 105 106 void 107 linux_sendsig(catcher, sig, mask, code) 108 sig_t catcher; 109 int sig, mask; 110 u_long code; 111 { 112 register struct proc *p = curproc; 113 register struct trapframe *tf; 114 struct linux_sigframe *fp, frame; 115 struct sigacts *psp = p->p_sigacts; 116 int oonstack; 117 extern char linux_sigcode[], linux_esigcode[]; 118 119 tf = p->p_md.md_regs; 120 oonstack = psp->ps_sigstk.ss_flags & SS_ONSTACK; 121 122 /* 123 * Allocate space for the signal handler context. 124 */ 125 if ((psp->ps_flags & SAS_ALTSTACK) && !oonstack && 126 (psp->ps_sigonstack & sigmask(sig))) { 127 fp = (struct linux_sigframe *)(psp->ps_sigstk.ss_sp + 128 psp->ps_sigstk.ss_size - sizeof(struct linux_sigframe)); 129 psp->ps_sigstk.ss_flags |= SS_ONSTACK; 130 } else { 131 fp = (struct linux_sigframe *)tf->tf_esp - 1; 132 } 133 134 frame.sf_handler = catcher; 135 frame.sf_sig = bsd_to_linux_sig[sig]; 136 137 /* 138 * Build the signal context to be used by sigreturn. 139 */ 140 frame.sf_sc.sc_mask = mask; 141 #ifdef VM86 142 if (tf->tf_eflags & PSL_VM) { 143 frame.sf_sc.sc_gs = tf->tf_vm86_gs; 144 frame.sf_sc.sc_fs = tf->tf_vm86_fs; 145 frame.sf_sc.sc_es = tf->tf_vm86_es; 146 frame.sf_sc.sc_ds = tf->tf_vm86_ds; 147 frame.sf_sc.sc_eflags = get_vflags(p); 148 } else 149 #endif 150 { 151 __asm("movl %%gs,%w0" : "=r" (frame.sf_sc.sc_gs)); 152 __asm("movl %%fs,%w0" : "=r" (frame.sf_sc.sc_fs)); 153 frame.sf_sc.sc_es = tf->tf_es; 154 frame.sf_sc.sc_ds = tf->tf_ds; 155 frame.sf_sc.sc_eflags = tf->tf_eflags; 156 } 157 frame.sf_sc.sc_edi = tf->tf_edi; 158 frame.sf_sc.sc_esi = tf->tf_esi; 159 frame.sf_sc.sc_ebp = tf->tf_ebp; 160 frame.sf_sc.sc_ebx = tf->tf_ebx; 161 frame.sf_sc.sc_edx = tf->tf_edx; 162 frame.sf_sc.sc_ecx = tf->tf_ecx; 163 frame.sf_sc.sc_eax = tf->tf_eax; 164 frame.sf_sc.sc_eip = tf->tf_eip; 165 frame.sf_sc.sc_cs = tf->tf_cs; 166 frame.sf_sc.sc_esp_at_signal = tf->tf_esp; 167 frame.sf_sc.sc_ss = tf->tf_ss; 168 frame.sf_sc.sc_err = tf->tf_err; 169 frame.sf_sc.sc_trapno = tf->tf_trapno; 170 171 if (copyout(&frame, fp, sizeof(frame)) != 0) { 172 /* 173 * Process has trashed its stack; give it an illegal 174 * instruction to halt it in its tracks. 175 */ 176 sigexit(p, SIGILL); 177 /* NOTREACHED */ 178 } 179 180 /* 181 * Build context to run handler in. 182 */ 183 tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL); 184 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 185 tf->tf_eip = (int)(((char *)PS_STRINGS) - 186 (linux_esigcode - linux_sigcode)); 187 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 188 tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC); 189 tf->tf_esp = (int)fp; 190 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 191 } 192 193 /* 194 * System call to cleanup state after a signal 195 * has been taken. Reset signal mask and 196 * stack state from context left by sendsig (above). 197 * Return to previous pc and psl as specified by 198 * context left by sendsig. Check carefully to 199 * make sure that the user has not modified the 200 * psl to gain improper privileges or to cause 201 * a machine fault. 202 */ 203 int 204 linux_sys_sigreturn(p, v, retval) 205 struct proc *p; 206 void *v; 207 register_t *retval; 208 { 209 struct linux_sys_sigreturn_args /* { 210 syscallarg(struct linux_sigcontext *) scp; 211 } */ *uap = v; 212 struct linux_sigcontext *scp, context; 213 register struct trapframe *tf; 214 215 tf = p->p_md.md_regs; 216 217 /* 218 * The trampoline code hands us the context. 219 * It is unsafe to keep track of it ourselves, in the event that a 220 * program jumps out of a signal handler. 221 */ 222 scp = SCARG(uap, scp); 223 if (copyin((caddr_t)scp, &context, sizeof(*scp)) != 0) 224 return (EFAULT); 225 226 /* 227 * Restore signal context. 228 */ 229 #ifdef VM86 230 if (context.sc_eflags & PSL_VM) { 231 tf->tf_vm86_gs = context.sc_gs; 232 tf->tf_vm86_fs = context.sc_fs; 233 tf->tf_vm86_es = context.sc_es; 234 tf->tf_vm86_ds = context.sc_ds; 235 set_vflags(p, context.sc_eflags); 236 } else 237 #endif 238 { 239 /* 240 * Check for security violations. If we're returning to 241 * protected mode, the CPU will validate the segment registers 242 * automatically and generate a trap on violations. We handle 243 * the trap, rather than doing all of the checking here. 244 */ 245 if (((context.sc_eflags ^ tf->tf_eflags) & PSL_USERSTATIC) != 0 || 246 !USERMODE(context.sc_cs, context.sc_eflags)) 247 return (EINVAL); 248 249 /* %fs and %gs were restored by the trampoline. */ 250 tf->tf_es = context.sc_es; 251 tf->tf_ds = context.sc_ds; 252 tf->tf_eflags = context.sc_eflags; 253 } 254 tf->tf_edi = context.sc_edi; 255 tf->tf_esi = context.sc_esi; 256 tf->tf_ebp = context.sc_ebp; 257 tf->tf_ebx = context.sc_ebx; 258 tf->tf_edx = context.sc_edx; 259 tf->tf_ecx = context.sc_ecx; 260 tf->tf_eax = context.sc_eax; 261 tf->tf_eip = context.sc_eip; 262 tf->tf_cs = context.sc_cs; 263 tf->tf_esp = context.sc_esp_at_signal; 264 tf->tf_ss = context.sc_ss; 265 266 p->p_sigacts->ps_sigstk.ss_flags &= ~SS_ONSTACK; 267 p->p_sigmask = context.sc_mask & ~sigcantmask; 268 269 return (EJUSTRETURN); 270 } 271 272 #ifdef USER_LDT 273 274 int 275 linux_read_ldt(p, uap, retval) 276 struct proc *p; 277 struct linux_sys_modify_ldt_args /* { 278 syscallarg(int) func; 279 syscallarg(void *) ptr; 280 syscallarg(size_t) bytecount; 281 } */ *uap; 282 register_t *retval; 283 { 284 struct i386_get_ldt_args gl; 285 int error; 286 caddr_t sg; 287 char *parms; 288 289 sg = stackgap_init(p->p_emul); 290 291 gl.start = 0; 292 gl.desc = SCARG(uap, ptr); 293 gl.num = SCARG(uap, bytecount) / sizeof(union descriptor); 294 295 parms = stackgap_alloc(&sg, sizeof(gl)); 296 297 if ((error = copyout(&gl, parms, sizeof(gl))) != 0) 298 return (error); 299 300 if ((error = i386_get_ldt(p, parms, retval)) != 0) 301 return (error); 302 303 *retval *= sizeof(union descriptor); 304 return (0); 305 } 306 307 struct linux_ldt_info { 308 u_int entry_number; 309 u_long base_addr; 310 u_int limit; 311 u_int seg_32bit:1; 312 u_int contents:2; 313 u_int read_exec_only:1; 314 u_int limit_in_pages:1; 315 u_int seg_not_present:1; 316 }; 317 318 int 319 linux_write_ldt(p, uap, retval) 320 struct proc *p; 321 struct linux_sys_modify_ldt_args /* { 322 syscallarg(int) func; 323 syscallarg(void *) ptr; 324 syscallarg(size_t) bytecount; 325 } */ *uap; 326 register_t *retval; 327 { 328 struct linux_ldt_info ldt_info; 329 struct segment_descriptor sd; 330 struct i386_set_ldt_args sl; 331 int error; 332 caddr_t sg; 333 char *parms; 334 335 if (SCARG(uap, bytecount) != sizeof(ldt_info)) 336 return (EINVAL); 337 if ((error = copyin(SCARG(uap, ptr), &ldt_info, sizeof(ldt_info))) != 0) 338 return error; 339 if (ldt_info.contents == 3) 340 return (EINVAL); 341 342 sg = stackgap_init(p->p_emul); 343 344 sd.sd_lobase = ldt_info.base_addr & 0xffffff; 345 sd.sd_hibase = (ldt_info.base_addr >> 24) & 0xff; 346 sd.sd_lolimit = ldt_info.limit & 0xffff; 347 sd.sd_hilimit = (ldt_info.limit >> 16) & 0xf; 348 sd.sd_type = 349 16 | (ldt_info.contents << 2) | (!ldt_info.read_exec_only << 1); 350 sd.sd_dpl = SEL_UPL; 351 sd.sd_p = !ldt_info.seg_not_present; 352 sd.sd_def32 = ldt_info.seg_32bit; 353 sd.sd_gran = ldt_info.limit_in_pages; 354 355 sl.start = ldt_info.entry_number; 356 sl.desc = stackgap_alloc(&sg, sizeof(sd)); 357 sl.num = 1; 358 359 #if 0 360 kprintf("linux_write_ldt: idx=%d, base=%x, limit=%x\n", 361 ldt_info.entry_number, ldt_info.base_addr, ldt_info.limit); 362 #endif 363 364 parms = stackgap_alloc(&sg, sizeof(sl)); 365 366 if ((error = copyout(&sd, sl.desc, sizeof(sd))) != 0) 367 return (error); 368 if ((error = copyout(&sl, parms, sizeof(sl))) != 0) 369 return (error); 370 371 if ((error = i386_set_ldt(p, parms, retval)) != 0) 372 return (error); 373 374 *retval = 0; 375 return (0); 376 } 377 378 #endif /* USER_LDT */ 379 380 int 381 linux_sys_modify_ldt(p, v, retval) 382 struct proc *p; 383 void *v; 384 register_t *retval; 385 { 386 struct linux_sys_modify_ldt_args /* { 387 syscallarg(int) func; 388 syscallarg(void *) ptr; 389 syscallarg(size_t) bytecount; 390 } */ *uap = v; 391 392 switch (SCARG(uap, func)) { 393 #ifdef USER_LDT 394 case 0: 395 return (linux_read_ldt(p, uap, retval)); 396 397 case 1: 398 return (linux_write_ldt(p, uap, retval)); 399 #endif /* USER_LDT */ 400 401 default: 402 return (ENOSYS); 403 } 404 } 405 406 /* 407 * XXX Pathetic hack to make svgalib work. This will fake the major 408 * device number of an opened VT so that svgalib likes it. grmbl. 409 * Should probably do it 'wrong the right way' and use a mapping 410 * array for all major device numbers, and map linux_mknod too. 411 */ 412 dev_t 413 linux_fakedev(dev) 414 dev_t dev; 415 { 416 417 if (major(dev) == NETBSD_CONS_MAJOR) 418 return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1)); 419 return dev; 420 } 421 422 /* 423 * We come here in a last attempt to satisfy a Linux ioctl() call 424 */ 425 int 426 linux_machdepioctl(p, v, retval) 427 struct proc *p; 428 void *v; 429 register_t *retval; 430 { 431 struct linux_sys_ioctl_args /* { 432 syscallarg(int) fd; 433 syscallarg(u_long) com; 434 syscallarg(caddr_t) data; 435 } */ *uap = v; 436 struct sys_ioctl_args bia; 437 u_long com; 438 #if NVT > 0 439 int error; 440 struct vt_mode lvt; 441 caddr_t bvtp, sg; 442 #endif 443 444 SCARG(&bia, fd) = SCARG(uap, fd); 445 SCARG(&bia, data) = SCARG(uap, data); 446 com = SCARG(uap, com); 447 448 switch (com) { 449 #if NVT > 0 450 case LINUX_KDGKBMODE: 451 com = KDGKBMODE; 452 break; 453 case LINUX_KDSKBMODE: 454 com = KDSKBMODE; 455 if ((unsigned)SCARG(uap, data) == LINUX_K_MEDIUMRAW) 456 SCARG(&bia, data) = (caddr_t)K_RAW; 457 break; 458 case LINUX_KDMKTONE: 459 com = KDMKTONE; 460 break; 461 case LINUX_KDSETMODE: 462 com = KDSETMODE; 463 break; 464 case LINUX_KDENABIO: 465 com = KDENABIO; 466 break; 467 case LINUX_KDDISABIO: 468 com = KDDISABIO; 469 break; 470 case LINUX_KDGETLED: 471 com = KDGETLED; 472 break; 473 case LINUX_KDSETLED: 474 com = KDSETLED; 475 break; 476 case LINUX_VT_OPENQRY: 477 com = VT_OPENQRY; 478 break; 479 case LINUX_VT_GETMODE: 480 SCARG(&bia, com) = VT_GETMODE; 481 if ((error = sys_ioctl(p, &bia, retval))) 482 return error; 483 if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt, 484 sizeof (struct vt_mode)))) 485 return error; 486 lvt.relsig = bsd_to_linux_sig[lvt.relsig]; 487 lvt.acqsig = bsd_to_linux_sig[lvt.acqsig]; 488 lvt.frsig = bsd_to_linux_sig[lvt.frsig]; 489 return copyout((caddr_t)&lvt, SCARG(uap, data), 490 sizeof (struct vt_mode)); 491 case LINUX_VT_SETMODE: 492 com = VT_SETMODE; 493 if ((error = copyin(SCARG(uap, data), (caddr_t)&lvt, 494 sizeof (struct vt_mode)))) 495 return error; 496 lvt.relsig = linux_to_bsd_sig[lvt.relsig]; 497 lvt.acqsig = linux_to_bsd_sig[lvt.acqsig]; 498 lvt.frsig = linux_to_bsd_sig[lvt.frsig]; 499 sg = stackgap_init(p->p_emul); 500 bvtp = stackgap_alloc(&sg, sizeof (struct vt_mode)); 501 if ((error = copyout(&lvt, bvtp, sizeof (struct vt_mode)))) 502 return error; 503 SCARG(&bia, data) = bvtp; 504 break; 505 case LINUX_VT_RELDISP: 506 com = VT_RELDISP; 507 break; 508 case LINUX_VT_ACTIVATE: 509 com = VT_ACTIVATE; 510 break; 511 case LINUX_VT_WAITACTIVE: 512 com = VT_WAITACTIVE; 513 break; 514 #endif 515 default: 516 kprintf("linux_machdepioctl: invalid ioctl %08lx\n", com); 517 return EINVAL; 518 } 519 SCARG(&bia, com) = com; 520 return sys_ioctl(p, &bia, retval); 521 } 522 523 /* 524 * Set I/O permissions for a process. Just set the maximum level 525 * right away (ignoring the argument), otherwise we would have 526 * to rely on I/O permission maps, which are not implemented. 527 */ 528 int 529 linux_sys_iopl(p, v, retval) 530 struct proc *p; 531 void *v; 532 register_t *retval; 533 { 534 #if 0 535 struct linux_sys_iopl_args /* { 536 syscallarg(int) level; 537 } */ *uap = v; 538 #endif 539 struct trapframe *fp = p->p_md.md_regs; 540 541 if (suser(p->p_ucred, &p->p_acflag) != 0) 542 return EPERM; 543 fp->tf_eflags |= PSL_IOPL; 544 *retval = 0; 545 return 0; 546 } 547 548 /* 549 * See above. If a root process tries to set access to an I/O port, 550 * just let it have the whole range. 551 */ 552 int 553 linux_sys_ioperm(p, v, retval) 554 struct proc *p; 555 void *v; 556 register_t *retval; 557 { 558 struct linux_sys_ioperm_args /* { 559 syscallarg(unsigned int) lo; 560 syscallarg(unsigned int) hi; 561 syscallarg(int) val; 562 } */ *uap = v; 563 struct trapframe *fp = p->p_md.md_regs; 564 565 if (suser(p->p_ucred, &p->p_acflag) != 0) 566 return EPERM; 567 if (SCARG(uap, val)) 568 fp->tf_eflags |= PSL_IOPL; 569 *retval = 0; 570 return 0; 571 } 572
Indexes created Sun Sep 28 16:09:52 GMT 2025