1 /* $NetBSD: linux_machdep.c,v 1.48 2014/11/09 17:48:08 maxv 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 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __KERNEL_RCSID(0, "$NetBSD: linux_machdep.c,v 1.48 2014/11/09 17:48:08 maxv Exp $"); 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/signalvar.h> 38 #include <sys/kernel.h> 39 #include <sys/proc.h> 40 #include <sys/buf.h> 41 #include <sys/reboot.h> 42 #include <sys/conf.h> 43 #include <sys/exec.h> 44 #include <sys/file.h> 45 #include <sys/callout.h> 46 #include <sys/mbuf.h> 47 #include <sys/msgbuf.h> 48 #include <sys/mount.h> 49 #include <sys/vnode.h> 50 #include <sys/device.h> 51 #include <sys/syscallargs.h> 52 #include <sys/filedesc.h> 53 #include <sys/exec_elf.h> 54 #include <sys/disklabel.h> 55 #include <sys/ioctl.h> 56 #include <miscfs/specfs/specdev.h> 57 58 #include <compat/linux/common/linux_types.h> 59 #include <compat/linux/common/linux_signal.h> 60 #include <compat/linux/common/linux_util.h> 61 #include <compat/linux/common/linux_ioctl.h> 62 #include <compat/linux/common/linux_hdio.h> 63 #include <compat/linux/common/linux_exec.h> 64 #include <compat/linux/common/linux_machdep.h> 65 66 #include <compat/linux/linux_syscallargs.h> 67 68 #include <sys/cpu.h> 69 #include <machine/fpu.h> 70 #include <machine/psl.h> 71 #include <machine/pcb.h> 72 #include <machine/vmparam.h> 73 74 /* 75 * To see whether wscons is configured (for virtual console ioctl calls). 76 */ 77 #if defined(_KERNEL_OPT) 78 #include "wsdisplay.h" 79 #endif 80 #if (NWSDISPLAY > 0) 81 #include <dev/wscons/wsconsio.h> 82 #include <dev/wscons/wsdisplay_usl_io.h> 83 #endif 84 85 /* 86 * Set set up registers on exec. 87 */ 88 void 89 linux_setregs(struct lwp *l, struct exec_package *pack, vaddr_t stack) 90 { 91 setregs(l, pack, stack); 92 } 93 94 /* 95 * Send an interrupt to process. 96 * 97 * Adapted from arch/powerpc/powerpc/sig_machdep.c:sendsig and 98 * compat/linux/arch/i386/linux_machdep.c:linux_sendsig 99 * 100 * XXX Does not work well yet with RT signals 101 * 102 */ 103 104 void 105 linux_sendsig(const ksiginfo_t *ksi, const sigset_t *mask) 106 { 107 const int sig = ksi->ksi_signo; 108 struct lwp *l = curlwp; 109 struct proc *p = l->l_proc; 110 struct trapframe *tf; 111 sig_t catcher = SIGACTION(p, sig).sa_handler; 112 struct linux_sigregs frame; 113 struct linux_pt_regs linux_regs; 114 struct linux_sigcontext sc; 115 register_t fp; 116 int onstack, error; 117 int i; 118 119 tf = trapframe(l); 120 121 /* 122 * Do we need to jump onto the signal stack? 123 */ 124 onstack = 125 (l->l_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 126 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 127 128 /* 129 * Signal stack is broken (see at the end of linux_sigreturn), so we do 130 * not use it yet. XXX fix this. 131 */ 132 onstack=0; 133 134 /* 135 * Allocate space for the signal handler context. 136 */ 137 if (onstack) { 138 fp = (register_t) 139 ((char *)l->l_sigstk.ss_sp + 140 l->l_sigstk.ss_size); 141 } else { 142 fp = tf->tf_fixreg[1]; 143 } 144 #ifdef DEBUG_LINUX 145 printf("fp at start of linux_sendsig = %x\n", fp); 146 #endif 147 fp -= sizeof(struct linux_sigregs); 148 fp &= ~0xf; 149 150 /* 151 * Prepare a sigcontext for later. 152 */ 153 memset(&sc, 0, sizeof sc); 154 sc.lsignal = (int)native_to_linux_signo[sig]; 155 sc.lhandler = (unsigned long)catcher; 156 native_to_linux_old_extra_sigset(&sc.lmask, &sc._unused[3], mask); 157 sc.lregs = (struct linux_pt_regs*)fp; 158 159 /* 160 * Setup the signal stack frame as Linux does it in 161 * arch/ppc/kernel/signal.c:setup_frame() 162 * 163 * Save register context. 164 */ 165 for (i = 0; i < 32; i++) 166 linux_regs.lgpr[i] = tf->tf_fixreg[i]; 167 linux_regs.lnip = tf->tf_srr0; 168 linux_regs.lmsr = tf->tf_srr1 & PSL_USERSRR1; 169 linux_regs.lorig_gpr3 = tf->tf_fixreg[3]; /* XXX Is that right? */ 170 linux_regs.lctr = tf->tf_ctr; 171 linux_regs.llink = tf->tf_lr; 172 linux_regs.lxer = tf->tf_xer; 173 linux_regs.lccr = tf->tf_cr; 174 linux_regs.lmq = 0; /* Unused, 601 only */ 175 linux_regs.ltrap = tf->tf_exc; 176 linux_regs.ldar = tf->tf_dar; 177 linux_regs.ldsisr = tf->tf_dsisr; 178 linux_regs.lresult = 0; 179 180 memset(&frame, 0, sizeof(frame)); 181 memcpy(&frame.lgp_regs, &linux_regs, sizeof(linux_regs)); 182 183 #ifdef PPC_HAVE_FPU 184 fpu_save(); 185 #endif 186 memcpy(&frame.lfp_regs, curpcb->pcb_fpu.fpreg, sizeof(frame.lfp_regs)); 187 188 /* 189 * Copy Linux's signal trampoline on the user stack It should not 190 * be used, but Linux binaries might expect it to be there. 191 */ 192 frame.ltramp[0] = 0x38997777; /* li r0, 0x7777 */ 193 frame.ltramp[1] = 0x44000002; /* sc */ 194 195 /* 196 * Move it to the user stack 197 * There is a little trick here, about the LINUX_ABIGAP: the 198 * linux_sigreg structure has a 56 int gap to support rs6000/xcoff 199 * binaries. But the Linux kernel seems to do without it, and it 200 * just skip it when building the stack frame. Hence the LINUX_ABIGAP. 201 */ 202 sendsig_reset(l, sig); 203 mutex_exit(p->p_lock); 204 error = copyout(&frame, (void *)fp, sizeof (frame) - LINUX_ABIGAP); 205 206 if (error != 0) { 207 /* 208 * Process has trashed its stack; give it an illegal 209 * instruction to halt it in its tracks. 210 */ 211 mutex_enter(p->p_lock); 212 sigexit(l, SIGILL); 213 /* NOTREACHED */ 214 } 215 216 /* 217 * Add a sigcontext on the stack 218 */ 219 fp -= sizeof(struct linux_sigcontext); 220 error = copyout(&sc, (void *)fp, sizeof (struct linux_sigcontext)); 221 mutex_enter(p->p_lock); 222 223 if (error != 0) { 224 /* 225 * Process has trashed its stack; give it an illegal 226 * instruction to halt it in its tracks. 227 */ 228 sigexit(l, SIGILL); 229 /* NOTREACHED */ 230 } 231 232 /* 233 * Set the registers according to how the Linux process expects them. 234 * "Mind the gap" Linux expects a gap here. 235 */ 236 tf->tf_fixreg[1] = fp - LINUX__SIGNAL_FRAMESIZE; 237 tf->tf_lr = (int)catcher; 238 tf->tf_fixreg[3] = (int)native_to_linux_signo[sig]; 239 tf->tf_fixreg[4] = fp; 240 tf->tf_srr0 = (int)p->p_sigctx.ps_sigcode; 241 242 #ifdef DEBUG_LINUX 243 printf("fp at end of linux_sendsig = %x\n", fp); 244 #endif 245 /* 246 * Remember that we're now on the signal stack. 247 */ 248 if (onstack) 249 l->l_sigstk.ss_flags |= SS_ONSTACK; 250 #ifdef DEBUG_LINUX 251 printf("linux_sendsig: exiting. fp=0x%lx\n",(long)fp); 252 #endif 253 } 254 255 /* 256 * System call to cleanup state after a signal 257 * has been taken. Reset signal mask and 258 * stack state from context left by sendsig (above). 259 * Return to previous pc and psl as specified by 260 * context left by sendsig. Check carefully to 261 * make sure that the user has not modified the 262 * psl to gain improper privileges or to cause 263 * a machine fault. 264 * 265 * XXX not tested 266 */ 267 int 268 linux_sys_rt_sigreturn(struct lwp *l, const struct linux_sys_rt_sigreturn_args *uap, register_t *retval) 269 { 270 /* { 271 syscallarg(struct linux_rt_sigframe *) sfp; 272 } */ 273 struct proc *p = l->l_proc; 274 struct linux_rt_sigframe *scp, sigframe; 275 struct linux_sigregs sregs; 276 struct linux_pt_regs *lregs; 277 struct trapframe *tf; 278 sigset_t mask; 279 int i; 280 281 /* 282 * The trampoline code hands us the context. 283 * It is unsafe to keep track of it ourselves, in the event that a 284 * program jumps out of a signal handler. 285 */ 286 scp = SCARG(uap, sfp); 287 288 /* 289 * Get the context from user stack 290 */ 291 if (copyin((void *)scp, &sigframe, sizeof(*scp))) 292 return (EFAULT); 293 294 /* 295 * Restore register context. 296 */ 297 if (copyin((void *)sigframe.luc.luc_context.lregs, 298 &sregs, sizeof(sregs))) 299 return (EFAULT); 300 lregs = (struct linux_pt_regs *)&sregs.lgp_regs; 301 302 tf = trapframe(l); 303 #ifdef DEBUG_LINUX 304 printf("linux_sys_rt_sigreturn: trapframe=0x%lx scp=0x%lx\n", 305 (unsigned long)tf, (unsigned long)scp); 306 #endif 307 308 if (!PSL_USEROK_P(lregs->lmsr)) 309 return (EINVAL); 310 311 for (i = 0; i < 32; i++) 312 tf->tf_fixreg[i] = lregs->lgpr[i]; 313 tf->tf_lr = lregs->llink; 314 tf->tf_cr = lregs->lccr; 315 tf->tf_xer = lregs->lxer; 316 tf->tf_ctr = lregs->lctr; 317 tf->tf_srr0 = lregs->lnip; 318 tf->tf_srr1 = lregs->lmsr; 319 320 /* 321 * Make sure the fpu state is discarded 322 */ 323 #ifdef PPC_HAVE_FPU 324 fpu_discard(); 325 #endif 326 327 memcpy(curpcb->pcb_fpu.fpreg, (void *)&sregs.lfp_regs, 328 sizeof(curpcb->pcb_fpu.fpreg)); 329 330 fpu_mark_used(curlwp); 331 332 mutex_enter(p->p_lock); 333 334 /* 335 * Restore signal stack. 336 * 337 * XXX cannot find the onstack information in Linux sig context. 338 * Is signal stack really supported on Linux? 339 * 340 * It seems to be supported in libc6... 341 */ 342 /* if (sc.sc_onstack & SS_ONSTACK) 343 l->l_sigstk.ss_flags |= SS_ONSTACK; 344 else */ 345 l->l_sigstk.ss_flags &= ~SS_ONSTACK; 346 347 /* 348 * Grab the signal mask 349 */ 350 linux_to_native_sigset(&mask, &sigframe.luc.luc_sigmask); 351 (void) sigprocmask1(l, SIG_SETMASK, &mask, 0); 352 353 mutex_exit(p->p_lock); 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(struct lwp *l, const struct linux_sys_sigreturn_args *uap, register_t *retval) 364 { 365 /* { 366 syscallarg(struct linux_sigcontext *) scp; 367 } */ 368 struct proc *p = l->l_proc; 369 struct linux_sigcontext *scp, context; 370 struct linux_sigregs sregs; 371 struct linux_pt_regs *lregs; 372 struct trapframe *tf; 373 sigset_t mask; 374 int i; 375 376 /* 377 * The trampoline code hands us the context. 378 * It is unsafe to keep track of it ourselves, in the event that a 379 * program jumps out of a signal handler. 380 */ 381 scp = SCARG(uap, scp); 382 383 /* 384 * Get the context from user stack 385 */ 386 if (copyin(scp, &context, sizeof(*scp))) 387 return (EFAULT); 388 389 /* 390 * Restore register context. 391 */ 392 if (copyin((void *)context.lregs, &sregs, sizeof(sregs))) 393 return (EFAULT); 394 lregs = (struct linux_pt_regs *)&sregs.lgp_regs; 395 396 tf = trapframe(l); 397 #ifdef DEBUG_LINUX 398 printf("linux_sys_sigreturn: trapframe=0x%lx scp=0x%lx\n", 399 (unsigned long)tf, (unsigned long)scp); 400 #endif 401 402 if (!PSL_USEROK_P(lregs->lmsr)) 403 return (EINVAL); 404 405 for (i = 0; i < 32; i++) 406 tf->tf_fixreg[i] = lregs->lgpr[i]; 407 tf->tf_lr = lregs->llink; 408 tf->tf_cr = lregs->lccr; 409 tf->tf_xer = lregs->lxer; 410 tf->tf_ctr = lregs->lctr; 411 tf->tf_srr0 = lregs->lnip; 412 tf->tf_srr1 = lregs->lmsr; 413 414 /* 415 * Make sure the fpu state is discarded 416 */ 417 #ifdef PPC_HAVE_FPU 418 fpu_discard(); 419 #endif 420 421 memcpy(curpcb->pcb_fpu.fpreg, (void *)&sregs.lfp_regs, 422 sizeof(curpcb->pcb_fpu.fpreg)); 423 424 fpu_mark_used(curlwp); 425 426 mutex_enter(p->p_lock); 427 428 /* 429 * Restore signal stack. 430 * 431 * XXX cannot find the onstack information in Linux sig context. 432 * Is signal stack really supported on Linux? 433 */ 434 #if 0 435 if (sc.sc_onstack & SS_ONSTACK) 436 l->l_sigstk.ss_flags |= SS_ONSTACK; 437 else 438 #endif 439 l->l_sigstk.ss_flags &= ~SS_ONSTACK; 440 441 /* Restore signal mask. */ 442 linux_old_extra_to_native_sigset(&mask, &context.lmask, 443 &context._unused[3]); 444 (void) sigprocmask1(l, SIG_SETMASK, &mask, 0); 445 446 mutex_exit(p->p_lock); 447 448 return (EJUSTRETURN); 449 } 450 451 /* 452 * major device numbers remapping 453 */ 454 dev_t 455 linux_fakedev(dev_t dev, int raw) 456 { 457 /* XXX write me */ 458 return dev; 459 } 460 461 /* 462 * We come here in a last attempt to satisfy a Linux ioctl() call 463 */ 464 int 465 linux_machdepioctl(struct lwp *l, const struct linux_sys_ioctl_args *uap, register_t *retval) 466 { 467 /* { 468 syscallarg(int) fd; 469 syscallarg(u_long) com; 470 syscallarg(void *) data; 471 } */ 472 struct sys_ioctl_args bia; 473 u_long com; 474 475 SCARG(&bia, fd) = SCARG(uap, fd); 476 SCARG(&bia, data) = SCARG(uap, data); 477 com = SCARG(uap, com); 478 479 switch (com) { 480 default: 481 printf("linux_machdepioctl: invalid ioctl %08lx\n", com); 482 return EINVAL; 483 } 484 SCARG(&bia, com) = com; 485 /* XXX NJWLWP */ 486 return sys_ioctl(curlwp, &bia, retval); 487 } 488 489 /* 490 * wrapper linux_sys_new_uname() -> linux_sys_uname() 491 */ 492 int 493 linux_sys_new_uname(struct lwp *l, const struct linux_sys_new_uname_args *uap, register_t *retval) 494 { 495 return linux_sys_uname(l, (const void *)uap, retval); 496 } 497 498 /* 499 * wrapper linux_sys_new_select() -> linux_sys_select() 500 */ 501 int 502 linux_sys_new_select(struct lwp *l, const struct linux_sys_new_select_args *uap, register_t *retval) 503 { 504 return linux_sys_select(l, (const void *)uap, retval); 505 } 506 507 int 508 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg) 509 { 510 return 0; 511 } 512