1 /* $NetBSD: linux_machdep.c,v 1.8 2001/04/11 19:41:03 manu Exp $ */ 2 3 /*- 4 * Copyright (c) 1995, 2000, 2001 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Frank van der Linden and Emmanuel Dreyfus. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/signalvar.h> 42 #include <sys/kernel.h> 43 #include <sys/map.h> 44 #include <sys/proc.h> 45 #include <sys/user.h> 46 #include <sys/buf.h> 47 #include <sys/reboot.h> 48 #include <sys/conf.h> 49 #include <sys/exec.h> 50 #include <sys/file.h> 51 #include <sys/callout.h> 52 #include <sys/malloc.h> 53 #include <sys/mbuf.h> 54 #include <sys/msgbuf.h> 55 #include <sys/mount.h> 56 #include <sys/vnode.h> 57 #include <sys/device.h> 58 #include <sys/syscallargs.h> 59 #include <sys/filedesc.h> 60 #include <sys/exec_elf.h> 61 #include <sys/disklabel.h> 62 #include <sys/ioctl.h> 63 #include <miscfs/specfs/specdev.h> 64 65 #include <compat/linux/common/linux_types.h> 66 #include <compat/linux/common/linux_signal.h> 67 #include <compat/linux/common/linux_util.h> 68 #include <compat/linux/common/linux_ioctl.h> 69 #include <compat/linux/common/linux_hdio.h> 70 #include <compat/linux/common/linux_exec.h> 71 #include <compat/linux/common/linux_machdep.h> 72 73 #include <compat/linux/linux_syscallargs.h> 74 75 #include <machine/cpu.h> 76 #include <machine/psl.h> 77 #include <machine/reg.h> 78 #include <machine/vmparam.h> 79 80 /* 81 * To see whether wscons is configured (for virtual console ioctl calls). 82 */ 83 #if defined(_KERNEL) && !defined(_LKM) 84 #include "wsdisplay.h" 85 #endif 86 #if (NWSDISPLAY > 0) 87 #include <dev/wscons/wsconsio.h> 88 #include <dev/wscons/wsdisplay_usl_io.h> 89 #if defined(_KERNEL) && !defined(_LKM) 90 #endif 91 #endif 92 93 /* 94 * Set set up registers on exec. 95 * XXX not used at the moment since in sys/kern/exec_conf, LINUX_COMPAT 96 * entry uses NetBSD's native setregs instead of linux_setregs 97 */ 98 void 99 linux_setregs(p, pack, stack) 100 struct proc *p; 101 struct exec_package *pack; 102 u_long stack; 103 { 104 setregs(p, pack, stack); 105 } 106 107 /* 108 * Send an interrupt to process. 109 * 110 * Adapted from arch/powerpc/powerpc/sig_machdep.c:sendsig and 111 * compat/linux/arch/i386/linux_machdep.c:linux_sendsig 112 * 113 * XXX Does not work well yet with RT signals 114 * 115 */ 116 117 void 118 linux_sendsig(catcher, sig, mask, code) /* XXX Check me */ 119 sig_t catcher; 120 int sig; 121 sigset_t *mask; 122 u_long code; 123 { 124 struct proc *p = curproc; 125 struct trapframe *tf; 126 struct linux_sigregs frame; 127 struct linux_pt_regs linux_regs; 128 struct linux_sigcontext sc; 129 register_t fp; 130 int onstack; 131 int i; 132 133 tf = trapframe(p); 134 135 /* 136 * Do we need to jump onto the signal stack? 137 */ 138 onstack = 139 (p->p_sigctx.ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 140 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 141 142 /* 143 * Signal stack is broken (see at the end of linux_sigreturn), so we do 144 * not use it yet. XXX fix this. 145 */ 146 onstack=0; 147 148 /* 149 * Allocate space for the signal handler context. 150 */ 151 if (onstack) { 152 fp = (register_t) 153 ((caddr_t)p->p_sigctx.ps_sigstk.ss_sp + 154 p->p_sigctx.ps_sigstk.ss_size); 155 } else { 156 fp = tf->fixreg[1]; 157 } 158 #ifdef DEBUG_LINUX 159 printf("fp at start of linux_sendsig = %x\n", fp); 160 #endif 161 fp -= sizeof(struct linux_sigregs); 162 fp &= ~0xf; 163 164 /* 165 * Prepare a sigcontext for later. 166 */ 167 memset(&sc, 0, sizeof sc); 168 sc.lsignal = (int)native_to_linux_sig[sig]; 169 sc.lhandler = (unsigned long)catcher; 170 native_to_linux_old_extra_sigset(mask, &sc.lmask, &sc._unused[3]); 171 sc.lregs = (struct linux_pt_regs*)fp; 172 173 /* 174 * Setup the signal stack frame as Linux does it in 175 * arch/ppc/kernel/signal.c:setup_frame() 176 * 177 * Save register context. 178 */ 179 for (i = 0; i < 32; i++) 180 linux_regs.lgpr[i] = tf->fixreg[i]; 181 linux_regs.lnip = tf->srr0; 182 linux_regs.lmsr = tf->srr1; 183 linux_regs.lorig_gpr3 = tf->fixreg[3]; /* XXX Is that right? */ 184 linux_regs.lctr = tf->ctr; 185 linux_regs.llink = tf->lr; 186 linux_regs.lxer = tf->xer; 187 linux_regs.lccr = tf->cr; 188 linux_regs.lmq = 0; /* Unused, 601 only */ 189 linux_regs.ltrap = tf->exc; 190 linux_regs.ldar = tf->dar; 191 linux_regs.ldsisr = tf->dsisr; 192 linux_regs.lresult = 0; 193 194 memset(&frame, 0, sizeof(frame)); 195 memcpy(&frame.lgp_regs, &linux_regs, sizeof(linux_regs)); 196 197 if (curproc == fpuproc) 198 save_fpu(curproc); 199 memcpy(&frame.lfp_regs, curpcb->pcb_fpu.fpr, sizeof(frame.lfp_regs)); 200 201 /* 202 * Copy Linux's signal trampoline on the user stack It should not 203 * be used, but Linux binaries might expect it to be there. 204 */ 205 frame.ltramp[0] = 0x38997777; /* li r0, 0x7777 */ 206 frame.ltramp[1] = 0x44000002; /* sc */ 207 208 /* 209 * Move it to the user stack 210 * There is a little trick here, about the LINUX_ABIGAP: the 211 * linux_sigreg structure has a 56 int gap to support rs6000/xcoff 212 * binaries. But the Linux kernel seems to do without it, and it 213 * just skip it when building the stack frame. Hence the LINUX_ABIGAP. 214 */ 215 if (copyout(&frame, (caddr_t)fp, sizeof (frame) - LINUX_ABIGAP) != 0) { 216 /* 217 * Process has trashed its stack; give it an illegal 218 * instructoin to halt it in its tracks. 219 */ 220 sigexit(p, SIGILL); 221 /* NOTREACHED */ 222 } 223 224 /* 225 * Add a sigcontext on the stack 226 */ 227 fp -= sizeof(struct linux_sigcontext); 228 if (copyout(&sc, (caddr_t)fp, sizeof (struct linux_sigcontext)) != 0) { 229 /* 230 * Process has trashed its stack; give it an illegal 231 * instructoin to halt it in its tracks. 232 */ 233 sigexit(p, SIGILL); 234 /* NOTREACHED */ 235 } 236 237 /* 238 * Set the registers according to how the Linux process expects them. 239 * "Mind the gap" Linux expects a gap here. 240 */ 241 tf->fixreg[1] = fp - LINUX__SIGNAL_FRAMESIZE; 242 tf->lr = (int)catcher; 243 tf->fixreg[3] = (int)native_to_linux_sig[sig]; 244 tf->fixreg[4] = fp; 245 tf->srr0 = (int)p->p_sigctx.ps_sigcode; 246 247 #ifdef DEBUG_LINUX 248 printf("fp at end of linux_sendsig = %x\n", fp); 249 #endif 250 /* 251 * Remember that we're now on the signal stack. 252 */ 253 if (onstack) 254 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 255 #ifdef DEBUG_LINUX 256 printf("linux_sendsig: exitting. fp=0x%lx\n",(long)fp); 257 #endif 258 } 259 260 /* 261 * System call to cleanup state after a signal 262 * has been taken. Reset signal mask and 263 * stack state from context left by sendsig (above). 264 * Return to previous pc and psl as specified by 265 * context left by sendsig. Check carefully to 266 * make sure that the user has not modified the 267 * psl to gain improper privileges or to cause 268 * a machine fault. 269 * 270 * XXX not tested 271 */ 272 int 273 linux_sys_rt_sigreturn(p, v, retval) 274 struct proc *p; 275 void *v; 276 register_t *retval; 277 { 278 struct linux_sys_rt_sigreturn_args /* { 279 syscallarg(struct linux_rt_sigframe *) sfp; 280 } */ *uap = v; 281 struct linux_rt_sigframe *scp, sigframe; 282 struct linux_sigregs sregs; 283 struct linux_pt_regs *lregs; 284 struct trapframe *tf; 285 sigset_t mask; 286 int i; 287 288 /* 289 * The trampoline code hands us the context. 290 * It is unsafe to keep track of it ourselves, in the event that a 291 * program jumps out of a signal handler. 292 */ 293 scp = SCARG(uap, sfp); 294 295 /* 296 * Get the context from user stack 297 */ 298 if (copyin((caddr_t)scp, &sigframe, sizeof(*scp))) 299 return (EFAULT); 300 301 /* 302 * Make sure, fpu is sync'ed 303 */ 304 if (curproc == fpuproc) 305 save_fpu(curproc); 306 307 /* 308 * Restore register context. 309 */ 310 if (copyin((caddr_t)sigframe.luc.luc_context.lregs, 311 &sregs, sizeof(sregs))) 312 return (EFAULT); 313 lregs = (struct linux_pt_regs *)&sregs.lgp_regs; 314 315 tf = trapframe(p); 316 #ifdef DEBUG_LINUX 317 printf("linux_sys_sigreturn: trapframe=0x%lx scp=0x%lx\n", 318 (unsigned long)tf, (unsigned long)scp); 319 #endif 320 321 if ((lregs->lmsr & PSL_USERSTATIC) != (tf->srr1 & PSL_USERSTATIC)) 322 return (EINVAL); 323 324 for (i = 0; i < 32; i++) 325 tf->fixreg[i] = lregs->lgpr[i]; 326 tf->lr = lregs->llink; 327 tf->cr = lregs->lccr; 328 tf->xer = lregs->lxer; 329 tf->ctr = lregs->lctr; 330 tf->srr0 = lregs->lnip; 331 tf->srr1 = lregs->lmsr; 332 333 memcpy(curpcb->pcb_fpu.fpr, (caddr_t)&sregs.lfp_regs, 334 sizeof(curpcb->pcb_fpu.fpr)); 335 336 /* 337 * Restore signal stack. 338 * 339 * XXX cannot find the onstack information in Linux sig context. 340 * Is signal stack really supported on Linux? 341 * 342 * It seems to be supported in libc6... 343 */ 344 /* if (sc.sc_onstack & SS_ONSTACK) 345 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 346 else */ 347 p->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK; 348 349 /* 350 * Grab the signal mask 351 */ 352 linux_to_native_sigset(&sigframe.luc.luc_sigmask, &mask); 353 (void) sigprocmask1(p, SIG_SETMASK, &mask, 0); 354 355 return (EJUSTRETURN); 356 } 357 358 359 /* 360 * The following needs code review for potential security issues 361 */ 362 int 363 linux_sys_sigreturn(p, v, retval) 364 struct proc *p; 365 void *v; 366 register_t *retval; 367 { 368 struct linux_sys_sigreturn_args /* { 369 syscallarg(struct linux_sigcontext *) scp; 370 } */ *uap = v; 371 struct linux_sigcontext *scp, context; 372 struct linux_sigregs sregs; 373 struct linux_pt_regs *lregs; 374 struct trapframe *tf; 375 sigset_t mask; 376 int i; 377 378 /* 379 * The trampoline code hands us the context. 380 * It is unsafe to keep track of it ourselves, in the event that a 381 * program jumps out of a signal handler. 382 */ 383 scp = SCARG(uap, scp); 384 385 /* 386 * Get the context from user stack 387 */ 388 if (copyin(scp, &context, sizeof(*scp))) 389 return (EFAULT); 390 391 /* 392 * Make sure, fpu is in sync 393 */ 394 if (curproc == fpuproc) 395 save_fpu(curproc); 396 397 /* 398 * Restore register context. 399 */ 400 if (copyin((caddr_t)context.lregs, &sregs, sizeof(sregs))) 401 return (EFAULT); 402 lregs = (struct linux_pt_regs *)&sregs.lgp_regs; 403 404 tf = trapframe(p); 405 #ifdef DEBUG_LINUX 406 printf("linux_sys_sigreturn: trapframe=0x%lx scp=0x%lx\n", 407 (unsigned long)tf, (unsigned long)scp); 408 #endif 409 410 if ((lregs->lmsr & PSL_USERSTATIC) != (tf->srr1 & PSL_USERSTATIC)) 411 return (EINVAL); 412 413 for (i = 0; i < 32; i++) 414 tf->fixreg[i] = lregs->lgpr[i]; 415 tf->lr = lregs->llink; 416 tf->cr = lregs->lccr; 417 tf->xer = lregs->lxer; 418 tf->ctr = lregs->lctr; 419 tf->srr0 = lregs->lnip; 420 tf->srr1 = lregs->lmsr; 421 422 memcpy(curpcb->pcb_fpu.fpr, (caddr_t)&sregs.lfp_regs, 423 sizeof(curpcb->pcb_fpu.fpr)); 424 425 /* 426 * Restore signal stack. 427 * 428 * XXX cannot find the onstack information in Linux sig context. 429 * Is signal stack really supported on Linux? 430 */ 431 #if 0 432 if (sc.sc_onstack & SS_ONSTACK) 433 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 434 else 435 #endif 436 p->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK; 437 438 /* Restore signal mask. */ 439 linux_old_extra_to_native_sigset(&context.lmask, 440 &context._unused[3], 441 &mask); 442 (void) sigprocmask1(p, SIG_SETMASK, &mask, 0); 443 444 return (EJUSTRETURN); 445 } 446 447 448 int 449 linux_sys_modify_ldt(p, v, retval) 450 struct proc *p; 451 void *v; 452 register_t *retval; 453 { 454 /* 455 * This syscall is not implemented in Linux/PowerPC: we should not 456 * be here 457 */ 458 #ifdef DEBUG_LINUX 459 printf("linux_sys_modify_ldt: should not be here.\n"); 460 #endif 461 return 0; 462 } 463 464 /* 465 * major device numbers remapping 466 */ 467 dev_t 468 linux_fakedev(dev) 469 dev_t dev; 470 { 471 /* XXX write me */ 472 return dev; 473 } 474 475 /* 476 * We come here in a last attempt to satisfy a Linux ioctl() call 477 */ 478 int 479 linux_machdepioctl(p, v, retval) 480 struct proc *p; 481 void *v; 482 register_t *retval; 483 { 484 struct linux_sys_ioctl_args /* { 485 syscallarg(int) fd; 486 syscallarg(u_long) com; 487 syscallarg(caddr_t) data; 488 } */ *uap = v; 489 struct sys_ioctl_args bia; 490 u_long com; 491 492 SCARG(&bia, fd) = SCARG(uap, fd); 493 SCARG(&bia, data) = SCARG(uap, data); 494 com = SCARG(uap, com); 495 496 switch (com) { 497 default: 498 printf("linux_machdepioctl: invalid ioctl %08lx\n", com); 499 return EINVAL; 500 } 501 SCARG(&bia, com) = com; 502 return sys_ioctl(p, &bia, retval); 503 } 504 /* 505 * Set I/O permissions for a process. Just set the maximum level 506 * right away (ignoring the argument), otherwise we would have 507 * to rely on I/O permission maps, which are not implemented. 508 */ 509 int 510 linux_sys_iopl(p, v, retval) 511 struct proc *p; 512 void *v; 513 register_t *retval; 514 { 515 /* 516 * This syscall is not implemented in Linux/PowerPC: we should not be here 517 */ 518 #ifdef DEBUG_LINUX 519 printf("linux_sys_iopl: should not be here.\n"); 520 #endif 521 return 0; 522 } 523 524 /* 525 * See above. If a root process tries to set access to an I/O port, 526 * just let it have the whole range. 527 */ 528 int 529 linux_sys_ioperm(p, v, retval) 530 struct proc *p; 531 void *v; 532 register_t *retval; 533 { 534 /* 535 * This syscall is not implemented in Linux/PowerPC: we should not be here 536 */ 537 #ifdef DEBUG_LINUX 538 printf("linux_sys_ioperm: should not be here.\n"); 539 #endif 540 return 0; 541 } 542 543 /* 544 * wrapper linux_sys_new_uname() -> linux_sys_uname() 545 */ 546 int 547 linux_sys_new_uname(p, v, retval) 548 struct proc *p; 549 void *v; 550 register_t *retval; 551 { 552 return linux_sys_uname(p, v, retval); 553 } 554 555 /* 556 * wrapper linux_sys_new_select() -> linux_sys_select() 557 */ 558 int 559 linux_sys_new_select(p, v, retval) 560 struct proc *p; 561 void *v; 562 register_t *retval; 563 { 564 return linux_sys_select(p, v, retval); 565 } 566