1 /* $NetBSD: linux_machdep.c,v 1.34 2006/05/15 09:17:05 yamt Exp $ */ 2 3 /*- 4 * Copyright (c) 1998 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Eric Haszlakiewicz. 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 * Based on sys/arch/i386/i386/linux_machdep.c: 39 * linux_machdep.c,v 1.42 1998/09/11 12:50:06 mycroft Exp 40 * written by Frank van der Linden 41 * 42 */ 43 44 #include <sys/cdefs.h> 45 __KERNEL_RCSID(0, "$NetBSD: linux_machdep.c,v 1.34 2006/05/15 09:17:05 yamt Exp $"); 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/signalvar.h> 50 #include <sys/kernel.h> 51 #include <sys/proc.h> 52 #include <sys/user.h> 53 #include <sys/buf.h> 54 #include <sys/reboot.h> 55 #include <sys/conf.h> 56 #include <sys/exec.h> 57 #include <sys/file.h> 58 #include <sys/callout.h> 59 #include <sys/malloc.h> 60 #include <sys/mbuf.h> 61 #include <sys/msgbuf.h> 62 #include <sys/mount.h> 63 #include <sys/vnode.h> 64 #include <sys/device.h> 65 #include <sys/sa.h> 66 #include <sys/syscallargs.h> 67 #include <sys/filedesc.h> 68 #include <sys/exec_elf.h> 69 #include <sys/ioctl.h> 70 #include <sys/kauth.h> 71 72 #include <uvm/uvm_extern.h> 73 74 #include <compat/linux/common/linux_types.h> 75 #include <compat/linux/common/linux_signal.h> 76 #include <compat/linux/common/linux_siginfo.h> 77 #include <compat/linux/common/linux_util.h> 78 #include <compat/linux/common/linux_ioctl.h> 79 #include <compat/linux/common/linux_exec.h> 80 #include <compat/linux/common/linux_machdep.h> 81 #include <compat/linux/common/linux_emuldata.h> 82 83 #include <compat/linux/linux_syscallargs.h> 84 85 #include <machine/alpha.h> 86 #include <machine/reg.h> 87 88 #if defined(_KERNEL_OPT) 89 #include "wsdisplay.h" 90 #endif 91 #if (NWSDISPLAY >0) 92 #include <dev/wscons/wsdisplay_usl_io.h> 93 #endif 94 #ifdef DEBUG 95 #include <machine/sigdebug.h> 96 #endif 97 98 /* 99 * Deal with some alpha-specific things in the Linux emulation code. 100 */ 101 102 void 103 linux_setregs(l, epp, stack) 104 struct lwp *l; 105 struct exec_package *epp; 106 u_long stack; 107 { 108 #ifdef DEBUG 109 struct trapframe *tfp = l->l_md.md_tf; 110 #endif 111 112 setregs(l, epp, stack); 113 #ifdef DEBUG 114 /* 115 * Linux has registers set to zero on entry; for DEBUG kernels 116 * the alpha setregs() fills registers with 0xbabefacedeadbeef. 117 */ 118 memset(tfp->tf_regs, 0, FRAME_SIZE * sizeof tfp->tf_regs[0]); 119 #endif 120 } 121 122 void 123 setup_linux_rt_sigframe(struct trapframe *tf, int sig, const sigset_t *mask) 124 { 125 struct lwp *l = curlwp; 126 struct proc *p = l->l_proc; 127 struct linux_rt_sigframe *sfp, sigframe; 128 int onstack; 129 int fsize, rndfsize; 130 extern char linux_rt_sigcode[], linux_rt_esigcode[]; 131 132 /* Do we need to jump onto the signal stack? */ 133 onstack = (p->p_sigctx.ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 134 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 135 136 /* Allocate space for the signal handler context. */ 137 fsize = sizeof(struct linux_rt_sigframe); 138 rndfsize = ((fsize + 15) / 16) * 16; 139 140 if (onstack) 141 sfp = (struct linux_rt_sigframe *) 142 ((caddr_t)p->p_sigctx.ps_sigstk.ss_sp + 143 p->p_sigctx.ps_sigstk.ss_size); 144 else 145 sfp = (struct linux_rt_sigframe *)(alpha_pal_rdusp()); 146 sfp = (struct linux_rt_sigframe *)((caddr_t)sfp - rndfsize); 147 148 #ifdef DEBUG 149 if ((sigdebug & SDB_KSTACK) && (p->p_pid == sigpid)) 150 printf("linux_sendsig(%d): sig %d ssp %p usp %p\n", p->p_pid, 151 sig, &onstack, sfp); 152 #endif /* DEBUG */ 153 154 /* 155 * Build the signal context to be used by sigreturn. 156 */ 157 memset(&sigframe.uc, 0, sizeof(struct linux_ucontext)); 158 sigframe.uc.uc_mcontext.sc_onstack = onstack; 159 160 /* Setup potentially partial signal mask in sc_mask. */ 161 /* But get all of it in uc_sigmask */ 162 native_to_linux_old_sigset(&sigframe.uc.uc_mcontext.sc_mask, mask); 163 native_to_linux_sigset(&sigframe.uc.uc_sigmask, mask); 164 165 sigframe.uc.uc_mcontext.sc_pc = tf->tf_regs[FRAME_PC]; 166 sigframe.uc.uc_mcontext.sc_ps = ALPHA_PSL_USERMODE; 167 frametoreg(tf, (struct reg *)sigframe.uc.uc_mcontext.sc_regs); 168 sigframe.uc.uc_mcontext.sc_regs[R_SP] = alpha_pal_rdusp(); 169 170 alpha_enable_fp(l, 1); 171 sigframe.uc.uc_mcontext.sc_fpcr = alpha_read_fpcr(); 172 sigframe.uc.uc_mcontext.sc_fp_control = alpha_read_fp_c(l); 173 alpha_pal_wrfen(0); 174 175 sigframe.uc.uc_mcontext.sc_traparg_a0 = tf->tf_regs[FRAME_A0]; 176 sigframe.uc.uc_mcontext.sc_traparg_a1 = tf->tf_regs[FRAME_A1]; 177 sigframe.uc.uc_mcontext.sc_traparg_a2 = tf->tf_regs[FRAME_A2]; 178 179 /* 180 * XXX XAX Create bogus siginfo data. This can't really 181 * XXX be fixed until NetBSD has realtime signals. 182 * XXX Or we do the emuldata thing. 183 * XXX -erh 184 */ 185 memset(&sigframe.info, 0, sizeof(struct linux_siginfo)); 186 sigframe.info.lsi_signo = sig; 187 sigframe.info.lsi_code = LINUX_SI_USER; 188 sigframe.info.lsi_pid = p->p_pid; 189 sigframe.info.lsi_uid = kauth_cred_geteuid(p->p_cred); /* Use real uid here? */ 190 191 if (copyout((caddr_t)&sigframe, (caddr_t)sfp, fsize) != 0) { 192 #ifdef DEBUG 193 if ((sigdebug & SDB_KSTACK) && p->p_pid == sigpid) 194 printf("sendsig(%d): copyout failed on sig %d\n", 195 p->p_pid, sig); 196 #endif 197 /* 198 * Process has trashed its stack; give it an illegal 199 * instruction to halt it in its tracks. 200 */ 201 sigexit(l, SIGILL); 202 /* NOTREACHED */ 203 } 204 205 /* Pass pointers to siginfo and ucontext in the regs */ 206 tf->tf_regs[FRAME_A1] = (unsigned long)&sfp->info; 207 tf->tf_regs[FRAME_A2] = (unsigned long)&sfp->uc; 208 209 /* Address of trampoline code. End up at this PC after mi_switch */ 210 tf->tf_regs[FRAME_PC] = 211 (u_int64_t)(p->p_psstr - (linux_rt_esigcode - linux_rt_sigcode)); 212 213 /* Adjust the stack */ 214 alpha_pal_wrusp((unsigned long)sfp); 215 216 /* Remember that we're now on the signal stack. */ 217 if (onstack) 218 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 219 } 220 221 void setup_linux_sigframe(tf, sig, mask) 222 struct trapframe *tf; 223 int sig; 224 const sigset_t *mask; 225 { 226 struct lwp *l = curlwp; 227 struct proc *p = l->l_proc; 228 struct linux_sigframe *sfp, sigframe; 229 int onstack; 230 int fsize, rndfsize; 231 extern char linux_sigcode[], linux_esigcode[]; 232 233 /* Do we need to jump onto the signal stack? */ 234 onstack = (p->p_sigctx.ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 235 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 236 237 /* Allocate space for the signal handler context. */ 238 fsize = sizeof(struct linux_sigframe); 239 rndfsize = ((fsize + 15) / 16) * 16; 240 241 if (onstack) 242 sfp = (struct linux_sigframe *) 243 ((caddr_t)p->p_sigctx.ps_sigstk.ss_sp + 244 p->p_sigctx.ps_sigstk.ss_size); 245 else 246 sfp = (struct linux_sigframe *)(alpha_pal_rdusp()); 247 sfp = (struct linux_sigframe *)((caddr_t)sfp - rndfsize); 248 249 #ifdef DEBUG 250 if ((sigdebug & SDB_KSTACK) && (p->p_pid == sigpid)) 251 printf("linux_sendsig(%d): sig %d ssp %p usp %p\n", p->p_pid, 252 sig, &onstack, sfp); 253 #endif /* DEBUG */ 254 255 /* 256 * Build the signal context to be used by sigreturn. 257 */ 258 memset(&sigframe.sf_sc, 0, sizeof(struct linux_ucontext)); 259 sigframe.sf_sc.sc_onstack = onstack; 260 native_to_linux_old_sigset(&sigframe.sf_sc.sc_mask, mask); 261 sigframe.sf_sc.sc_pc = tf->tf_regs[FRAME_PC]; 262 sigframe.sf_sc.sc_ps = ALPHA_PSL_USERMODE; 263 frametoreg(tf, (struct reg *)sigframe.sf_sc.sc_regs); 264 sigframe.sf_sc.sc_regs[R_SP] = alpha_pal_rdusp(); 265 266 if (l == fpcurlwp) { 267 alpha_pal_wrfen(1); 268 savefpstate(&l->l_addr->u_pcb.pcb_fp); 269 alpha_pal_wrfen(0); 270 sigframe.sf_sc.sc_fpcr = l->l_addr->u_pcb.pcb_fp.fpr_cr; 271 fpcurlwp = NULL; 272 } 273 /* XXX ownedfp ? etc...? */ 274 275 sigframe.sf_sc.sc_traparg_a0 = tf->tf_regs[FRAME_A0]; 276 sigframe.sf_sc.sc_traparg_a1 = tf->tf_regs[FRAME_A1]; 277 sigframe.sf_sc.sc_traparg_a2 = tf->tf_regs[FRAME_A2]; 278 279 if (copyout((caddr_t)&sigframe, (caddr_t)sfp, fsize) != 0) { 280 #ifdef DEBUG 281 if ((sigdebug & SDB_KSTACK) && p->p_pid == sigpid) 282 printf("sendsig(%d): copyout failed on sig %d\n", 283 p->p_pid, sig); 284 #endif 285 /* 286 * Process has trashed its stack; give it an illegal 287 * instruction to halt it in its tracks. 288 */ 289 sigexit(l, SIGILL); 290 /* NOTREACHED */ 291 } 292 293 /* Pass pointers to sigcontext in the regs */ 294 tf->tf_regs[FRAME_A1] = 0; 295 tf->tf_regs[FRAME_A2] = (unsigned long)&sfp->sf_sc; 296 297 /* Address of trampoline code. End up at this PC after mi_switch */ 298 tf->tf_regs[FRAME_PC] = 299 (u_int64_t)(p->p_psstr - (linux_esigcode - linux_sigcode)); 300 301 /* Adjust the stack */ 302 alpha_pal_wrusp((unsigned long)sfp); 303 304 /* Remember that we're now on the signal stack. */ 305 if (onstack) 306 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 307 } 308 309 /* 310 * Send an interrupt to process. 311 * 312 * Stack is set up to allow sigcode stored 313 * in u. to call routine, followed by kcall 314 * to sigreturn routine below. After sigreturn 315 * resets the signal mask, the stack, and the 316 * frame pointer, it returns to the user 317 * specified pc, psl. 318 */ 319 void 320 linux_sendsig(const ksiginfo_t *ksi, const sigset_t *mask) 321 { 322 struct lwp *l = curlwp; 323 struct proc *p = l->l_proc; 324 struct trapframe *tf = l->l_md.md_tf; 325 const int sig = ksi->ksi_signo; 326 sig_t catcher = SIGACTION(p, sig).sa_handler; 327 #ifdef notyet 328 struct linux_emuldata *edp; 329 330 /* Setup the signal frame (and part of the trapframe) */ 331 /*OLD: if (p->p_sigacts->ps_siginfo & sigmask(sig))*/ 332 /* XXX XAX this is broken now. need someplace to store what 333 XXX XAX kind of signal handler a signal has.*/ 334 #if 0 335 edp = (struct linux_emuldata *)p->p_emuldata; 336 #else 337 edp = 0; 338 #endif 339 if (edp && sigismember(&edp->ps_siginfo, sig)) 340 setup_linux_rt_sigframe(tf, sig, mask); 341 else 342 #endif /* notyet */ 343 setup_linux_sigframe(tf, sig, mask); 344 345 /* Signal handler for trampoline code */ 346 tf->tf_regs[FRAME_T12] = (u_int64_t)catcher; 347 tf->tf_regs[FRAME_A0] = native_to_linux_signo[sig]; 348 349 /* 350 * Linux has a custom restorer option. To support it we would 351 * need to store an array of restorers and a sigcode block 352 * which knew to use it. Doesn't seem worth the trouble. 353 * -erh 354 */ 355 356 #ifdef DEBUG 357 if (sigdebug & SDB_FOLLOW) 358 printf("sendsig(%d): pc %lx, catcher %lx\n", l->l_proc->p_pid, 359 tf->tf_regs[FRAME_PC], tf->tf_regs[FRAME_A3]); 360 if ((sigdebug & SDB_KSTACK) && l->l_proc->p_pid == sigpid) 361 printf("sendsig(%d): sig %d returns\n", l->l_proc->p_pid, sig); 362 #endif 363 } 364 365 /* 366 * System call to cleanup state after a signal 367 * has been taken. Reset signal mask and 368 * stack state from context left by sendsig (above). 369 * Return to previous pc as specified by context 370 * left by sendsig. 371 * Linux real-time signals use a different sigframe, 372 * but the sigcontext is the same. 373 */ 374 375 int 376 linux_restore_sigcontext(struct lwp *l, struct linux_sigcontext context, 377 sigset_t *mask) 378 { 379 380 struct proc *p = l->l_proc; 381 /* 382 * Linux doesn't (yet) have alternate signal stacks. 383 * However, the OSF/1 sigcontext which they use has 384 * an onstack member. This could be needed in the future. 385 */ 386 if (context.sc_onstack & LINUX_SA_ONSTACK) 387 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 388 else 389 p->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK; 390 391 /* Reset the signal mask */ 392 (void) sigprocmask1(p, SIG_SETMASK, mask, 0); 393 394 /* 395 * Check for security violations. 396 * Linux doesn't allow any changes to the PSL. 397 */ 398 if (context.sc_ps != ALPHA_PSL_USERMODE) 399 return(EINVAL); 400 401 l->l_md.md_tf->tf_regs[FRAME_PC] = context.sc_pc; 402 l->l_md.md_tf->tf_regs[FRAME_PS] = context.sc_ps; 403 404 regtoframe((struct reg *)context.sc_regs, l->l_md.md_tf); 405 alpha_pal_wrusp(context.sc_regs[R_SP]); 406 407 if (l == fpcurlwp) 408 fpcurlwp = NULL; 409 410 /* Restore fp regs and fpr_cr */ 411 bcopy((struct fpreg *)context.sc_fpregs, &l->l_addr->u_pcb.pcb_fp, 412 sizeof(struct fpreg)); 413 /* XXX sc_ownedfp ? */ 414 /* XXX sc_fp_control ? */ 415 416 #ifdef DEBUG 417 if (sigdebug & SDB_FOLLOW) 418 printf("linux_rt_sigreturn(%d): returns\n", p->p_pid); 419 #endif 420 return (EJUSTRETURN); 421 } 422 423 int 424 linux_sys_rt_sigreturn(l, v, retval) 425 struct lwp *l; 426 void *v; 427 register_t *retval; 428 { 429 struct linux_sys_rt_sigreturn_args /* { 430 syscallarg(struct linux_rt_sigframe *) sfp; 431 } */ *uap = v; 432 struct linux_rt_sigframe *sfp, sigframe; 433 sigset_t mask; 434 435 /* 436 * The trampoline code hands us the context. 437 * It is unsafe to keep track of it ourselves, in the event that a 438 * program jumps out of a signal handler. 439 */ 440 441 sfp = SCARG(uap, sfp); 442 443 if (ALIGN(sfp) != (u_int64_t)sfp) 444 return(EINVAL); 445 446 /* 447 * Fetch the frame structure. 448 */ 449 if (copyin((caddr_t)sfp, &sigframe, 450 sizeof(struct linux_rt_sigframe)) != 0) 451 return (EFAULT); 452 453 /* Grab the signal mask */ 454 linux_to_native_sigset(&mask, &sigframe.uc.uc_sigmask); 455 456 return(linux_restore_sigcontext(l, sigframe.uc.uc_mcontext, &mask)); 457 } 458 459 460 int 461 linux_sys_sigreturn(l, v, retval) 462 struct lwp *l; 463 void *v; 464 register_t *retval; 465 { 466 struct linux_sys_sigreturn_args /* { 467 syscallarg(struct linux_sigframe *) sfp; 468 } */ *uap = v; 469 struct linux_sigframe *sfp, frame; 470 sigset_t mask; 471 472 /* 473 * The trampoline code hands us the context. 474 * It is unsafe to keep track of it ourselves, in the event that a 475 * program jumps out of a signal handler. 476 */ 477 478 sfp = SCARG(uap, sfp); 479 if (ALIGN(sfp) != (u_int64_t)sfp) 480 return(EINVAL); 481 482 /* 483 * Fetch the frame structure. 484 */ 485 if (copyin((caddr_t)sfp, &frame, sizeof(struct linux_sigframe)) != 0) 486 return(EFAULT); 487 488 /* Grab the signal mask. */ 489 /* XXX use frame.extramask */ 490 linux_old_to_native_sigset(&mask, frame.sf_sc.sc_mask); 491 492 return(linux_restore_sigcontext(l, frame.sf_sc, &mask)); 493 } 494 495 /* 496 * We come here in a last attempt to satisfy a Linux ioctl() call 497 */ 498 /* XXX XAX update this, add maps, etc... */ 499 int 500 linux_machdepioctl(l, v, retval) 501 struct lwp *l; 502 void *v; 503 register_t *retval; 504 { 505 struct linux_sys_ioctl_args /* { 506 syscallarg(int) fd; 507 syscallarg(u_long) com; 508 syscallarg(caddr_t) data; 509 } */ *uap = v; 510 struct sys_ioctl_args bia; 511 u_long com; 512 513 SCARG(&bia, fd) = SCARG(uap, fd); 514 SCARG(&bia, data) = SCARG(uap, data); 515 com = SCARG(uap, com); 516 517 switch (com) { 518 default: 519 printf("linux_machdepioctl: invalid ioctl %08lx\n", com); 520 return EINVAL; 521 } 522 SCARG(&bia, com) = com; 523 return sys_ioctl(l, &bia, retval); 524 } 525 526 /* XXX XAX fix this */ 527 dev_t 528 linux_fakedev(dev, raw) 529 dev_t dev; 530 int raw; 531 { 532 return dev; 533 } 534 535 int 536 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg) 537 { 538 return 0; 539 } 540