1 /* $NetBSD: linux32_machdep.c,v 1.23 2009/11/23 00:46:07 rmind Exp $ */ 2 3 /*- 4 * Copyright (c) 2006 Emmanuel Dreyfus, all rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. All advertising materials mentioning features or use of this software 15 * must display the following acknowledgement: 16 * This product includes software developed by Emmanuel Dreyfus 17 * 4. The name of the author may not be used to endorse or promote 18 * products derived from this software without specific prior written 19 * permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE THE AUTHOR AND CONTRIBUTORS ``AS IS'' 22 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, 23 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 24 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS 25 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 */ 33 #include <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: linux32_machdep.c,v 1.23 2009/11/23 00:46:07 rmind Exp $"); 35 36 #include <sys/param.h> 37 #include <sys/systm.h> 38 #include <sys/signalvar.h> 39 #include <sys/kernel.h> 40 #include <sys/proc.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/syscallargs.h> 54 #include <sys/filedesc.h> 55 #include <sys/exec_elf.h> 56 #include <sys/disklabel.h> 57 #include <sys/ioctl.h> 58 #include <miscfs/specfs/specdev.h> 59 60 #include <machine/netbsd32_machdep.h> 61 62 #include <compat/netbsd32/netbsd32.h> 63 #include <compat/netbsd32/netbsd32_syscallargs.h> 64 65 #include <compat/linux/common/linux_signal.h> 66 #include <compat/linux/common/linux_errno.h> 67 68 #include <compat/linux32/common/linux32_types.h> 69 #include <compat/linux32/common/linux32_errno.h> 70 #include <compat/linux32/common/linux32_machdep.h> 71 #include <compat/linux32/common/linux32_signal.h> 72 #include <compat/linux32/common/linux32_exec.h> 73 #include <compat/linux32/linux32_syscallargs.h> 74 75 #include <sys/cpu.h> 76 #include <machine/cpufunc.h> 77 #include <machine/psl.h> 78 #include <machine/reg.h> 79 #include <machine/segments.h> 80 #include <machine/specialreg.h> 81 #include <machine/sysarch.h> 82 #include <machine/vmparam.h> 83 84 extern char linux32_sigcode[1]; 85 extern char linux32_rt_sigcode[1]; 86 extern char linux32_esigcode[1]; 87 88 extern void (osyscall_return)(void); 89 90 static void linux32_save_ucontext(struct lwp *, struct trapframe *, 91 const sigset_t *, struct sigaltstack *, struct linux32_ucontext *); 92 static void linux32_save_sigcontext(struct lwp *, struct trapframe *, 93 const sigset_t *, struct linux32_sigcontext *); 94 static void linux32_rt_sendsig(const ksiginfo_t *, const sigset_t *); 95 static void linux32_old_sendsig(const ksiginfo_t *, const sigset_t *); 96 static int linux32_restore_sigcontext(struct lwp *, 97 struct linux32_sigcontext *, register_t *); 98 99 void 100 linux32_sendsig(const ksiginfo_t *ksi, const sigset_t *mask) 101 { 102 if (SIGACTION(curproc, ksi->ksi_signo).sa_flags & SA_SIGINFO) 103 linux32_rt_sendsig(ksi, mask); 104 else 105 linux32_old_sendsig(ksi, mask); 106 return; 107 } 108 109 void 110 linux32_old_sendsig(const ksiginfo_t *ksi, const sigset_t *mask) 111 { 112 struct lwp *l = curlwp; 113 struct proc *p = l->l_proc; 114 struct trapframe *tf; 115 struct linux32_sigframe *fp, frame; 116 int onstack, error; 117 int sig = ksi->ksi_signo; 118 sig_t catcher = SIGACTION(p, sig).sa_handler; 119 struct sigaltstack *sas = &l->l_sigstk; 120 121 tf = l->l_md.md_regs; 122 /* Do we need to jump onto the signal stack? */ 123 onstack = (sas->ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 124 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 125 126 127 /* Allocate space for the signal handler context. */ 128 if (onstack) 129 fp = (struct linux32_sigframe *)((char *)sas->ss_sp + 130 sas->ss_size); 131 else 132 fp = (struct linux32_sigframe *)tf->tf_rsp; 133 fp--; 134 135 /* Build stack frame for signal trampoline. */ 136 NETBSD32PTR32(frame.sf_handler, catcher); 137 frame.sf_sig = native_to_linux32_signo[sig]; 138 139 linux32_save_sigcontext(l, tf, mask, &frame.sf_sc); 140 141 sendsig_reset(l, sig); 142 mutex_exit(p->p_lock); 143 error = copyout(&frame, fp, sizeof(frame)); 144 mutex_enter(p->p_lock); 145 146 if (error != 0) { 147 /* 148 * Process has trashed its stack; give it an illegal 149 * instruction to halt it in its tracks. 150 */ 151 sigexit(l, SIGILL); 152 /* NOTREACHED */ 153 } 154 155 /* 156 * Build context to run handler in. 157 */ 158 tf->tf_gs = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 159 tf->tf_fs = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 160 tf->tf_es = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 161 tf->tf_ds = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 162 tf->tf_rip = ((long)p->p_sigctx.ps_sigcode) & 0xffffffff; 163 tf->tf_cs = GSEL(GUCODE32_SEL, SEL_UPL) & 0xffffffff; 164 tf->tf_rflags &= ~PSL_CLEARSIG & 0xffffffff; 165 tf->tf_rsp = (long)fp & 0xffffffff; 166 tf->tf_ss = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 167 168 /* Remember that we're now on the signal stack. */ 169 if (onstack) 170 sas->ss_flags |= SS_ONSTACK; 171 172 return; 173 } 174 175 void 176 linux32_rt_sendsig(const ksiginfo_t *ksi, const sigset_t *mask) 177 { 178 struct lwp *l = curlwp; 179 struct proc *p = l->l_proc; 180 struct trapframe *tf; 181 struct linux32_rt_sigframe *fp, frame; 182 int onstack, error; 183 linux32_siginfo_t *lsi; 184 int sig = ksi->ksi_signo; 185 sig_t catcher = SIGACTION(p, sig).sa_handler; 186 struct sigaltstack *sas = &l->l_sigstk; 187 188 tf = l->l_md.md_regs; 189 /* Do we need to jump onto the signal stack? */ 190 onstack = (sas->ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 191 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 192 193 194 /* Allocate space for the signal handler context. */ 195 if (onstack) 196 fp = (struct linux32_rt_sigframe *)((char *)sas->ss_sp + 197 sas->ss_size); 198 else 199 fp = (struct linux32_rt_sigframe *)tf->tf_rsp; 200 fp--; 201 202 /* Build stack frame for signal trampoline. */ 203 NETBSD32PTR32(frame.sf_handler, catcher); 204 frame.sf_sig = native_to_linux32_signo[sig]; 205 NETBSD32PTR32(frame.sf_sip, &fp->sf_si); 206 NETBSD32PTR32(frame.sf_ucp, &fp->sf_uc); 207 208 lsi = &frame.sf_si; 209 (void)memset(lsi, 0, sizeof(frame.sf_si)); 210 lsi->lsi_errno = native_to_linux32_errno[ksi->ksi_errno]; 211 lsi->lsi_code = native_to_linux_si_code(ksi->ksi_code); 212 lsi->lsi_signo = frame.sf_sig; 213 switch (lsi->lsi_signo) { 214 case LINUX32_SIGILL: 215 case LINUX32_SIGFPE: 216 case LINUX32_SIGSEGV: 217 case LINUX32_SIGBUS: 218 case LINUX32_SIGTRAP: 219 NETBSD32PTR32(lsi->lsi_addr, ksi->ksi_addr); 220 break; 221 case LINUX32_SIGCHLD: 222 lsi->lsi_uid = ksi->ksi_uid; 223 lsi->lsi_pid = ksi->ksi_pid; 224 lsi->lsi_utime = ksi->ksi_utime; 225 lsi->lsi_stime = ksi->ksi_stime; 226 lsi->lsi_status = native_to_linux_si_status(ksi->ksi_code, 227 ksi->ksi_status); 228 break; 229 case LINUX32_SIGIO: 230 lsi->lsi_band = ksi->ksi_band; 231 lsi->lsi_fd = ksi->ksi_fd; 232 break; 233 default: 234 lsi->lsi_uid = ksi->ksi_uid; 235 lsi->lsi_pid = ksi->ksi_pid; 236 if (lsi->lsi_signo == LINUX32_SIGALRM || 237 lsi->lsi_signo >= LINUX32_SIGRTMIN) 238 NETBSD32PTR32(lsi->lsi_value.sival_ptr, 239 ksi->ksi_value.sival_ptr); 240 break; 241 } 242 243 /* Save register context. */ 244 linux32_save_ucontext(l, tf, mask, sas, &frame.sf_uc); 245 sendsig_reset(l, sig); 246 mutex_exit(p->p_lock); 247 error = copyout(&frame, fp, sizeof(frame)); 248 mutex_enter(p->p_lock); 249 250 if (error != 0) { 251 /* 252 * Process has trashed its stack; give it an illegal 253 * instruction to halt it in its tracks. 254 */ 255 sigexit(l, SIGILL); 256 /* NOTREACHED */ 257 } 258 259 /* 260 * Build context to run handler in. 261 */ 262 tf->tf_gs = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 263 tf->tf_fs = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 264 tf->tf_es = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 265 tf->tf_ds = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 266 tf->tf_rip = (((long)p->p_sigctx.ps_sigcode) + 267 (linux32_rt_sigcode - linux32_sigcode)) & 0xffffffff; 268 tf->tf_cs = GSEL(GUCODE32_SEL, SEL_UPL) & 0xffffffff; 269 tf->tf_rflags &= ~PSL_CLEARSIG & 0xffffffff; 270 tf->tf_rsp = (long)fp & 0xffffffff; 271 tf->tf_ss = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 272 273 /* Remember that we're now on the signal stack. */ 274 if (onstack) 275 sas->ss_flags |= SS_ONSTACK; 276 277 return; 278 } 279 280 void 281 linux32_setregs(struct lwp *l, struct exec_package *pack, u_long stack) 282 { 283 struct pcb *pcb = lwp_getpcb(l); 284 struct trapframe *tf; 285 struct proc *p = l->l_proc; 286 void **retaddr; 287 288 /* If we were using the FPU, forget about it. */ 289 if (pcb->pcb_fpcpu != NULL) 290 fpusave_lwp(l, 0); 291 292 #if defined(USER_LDT) && 0 293 pmap_ldt_cleanup(p); 294 #endif 295 296 netbsd32_adjust_limits(p); 297 298 l->l_md.md_flags &= ~MDP_USEDFPU; 299 pcb->pcb_flags = 0; 300 pcb->pcb_savefpu.fp_fxsave.fx_fcw = __Linux_NPXCW__; 301 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr = __INITIAL_MXCSR__; 302 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr_mask = __INITIAL_MXCSR_MASK__; 303 pcb->pcb_fs = 0; 304 pcb->pcb_gs = 0; 305 306 307 p->p_flag |= PK_32; 308 309 tf = l->l_md.md_regs; 310 tf->tf_rax = 0; 311 tf->tf_rbx = (u_int64_t)p->p_psstr & 0xffffffff; 312 tf->tf_rcx = pack->ep_entry & 0xffffffff; 313 tf->tf_rdx = 0; 314 tf->tf_rsi = 0; 315 tf->tf_rdi = 0; 316 tf->tf_rbp = 0; 317 tf->tf_rsp = stack & 0xffffffff; 318 tf->tf_r8 = 0; 319 tf->tf_r9 = 0; 320 tf->tf_r10 = 0; 321 tf->tf_r11 = 0; 322 tf->tf_r12 = 0; 323 tf->tf_r13 = 0; 324 tf->tf_r14 = 0; 325 tf->tf_r15 = 0; 326 tf->tf_rip = pack->ep_entry & 0xffffffff; 327 tf->tf_rflags = PSL_USERSET; 328 tf->tf_cs = GSEL(GUCODE32_SEL, SEL_UPL) & 0xffffffff; 329 tf->tf_ss = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 330 tf->tf_ds = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 331 tf->tf_es = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 332 tf->tf_fs = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 333 tf->tf_gs = GSEL(GUDATA32_SEL, SEL_UPL) & 0xffffffff; 334 335 /* XXX frob return address to return via old iret method, not sysret */ 336 retaddr = (void **)tf - 1; 337 *retaddr = (void *)osyscall_return; 338 return; 339 } 340 341 static void 342 linux32_save_ucontext(struct lwp *l, struct trapframe *tf, 343 const sigset_t *mask, struct sigaltstack *sas, struct linux32_ucontext *uc) 344 { 345 346 uc->uc_flags = 0; 347 NETBSD32PTR32(uc->uc_link, NULL); 348 native_to_linux32_sigaltstack(&uc->uc_stack, sas); 349 linux32_save_sigcontext(l, tf, mask, &uc->uc_mcontext); 350 native_to_linux32_sigset(&uc->uc_sigmask, mask); 351 (void)memset(&uc->uc_fpregs_mem, 0, sizeof(uc->uc_fpregs_mem)); 352 } 353 354 static void 355 linux32_save_sigcontext(struct lwp *l, struct trapframe *tf, 356 const sigset_t *mask, struct linux32_sigcontext *sc) 357 { 358 struct pcb *pcb = lwp_getpcb(l); 359 360 /* Save register context. */ 361 sc->sc_gs = tf->tf_gs; 362 sc->sc_fs = tf->tf_fs; 363 sc->sc_es = tf->tf_es; 364 sc->sc_ds = tf->tf_ds; 365 sc->sc_eflags = tf->tf_rflags; 366 sc->sc_edi = tf->tf_rdi; 367 sc->sc_esi = tf->tf_rsi; 368 sc->sc_esp = tf->tf_rsp; 369 sc->sc_ebp = tf->tf_rbp; 370 sc->sc_ebx = tf->tf_rbx; 371 sc->sc_edx = tf->tf_rdx; 372 sc->sc_ecx = tf->tf_rcx; 373 sc->sc_eax = tf->tf_rax; 374 sc->sc_eip = tf->tf_rip; 375 sc->sc_cs = tf->tf_cs; 376 sc->sc_esp_at_signal = tf->tf_rsp; 377 sc->sc_ss = tf->tf_ss; 378 sc->sc_err = tf->tf_err; 379 sc->sc_trapno = tf->tf_trapno; 380 sc->sc_cr2 = pcb->pcb_cr2; 381 NETBSD32PTR32(sc->sc_387, NULL); 382 383 /* Save signal stack. */ 384 /* Linux doesn't save the onstack flag in sigframe */ 385 386 /* Save signal mask. */ 387 native_to_linux32_old_sigset(&sc->sc_mask, mask); 388 } 389 390 int 391 linux32_sys_sigreturn(struct lwp *l, 392 const struct linux32_sys_sigreturn_args *uap, register_t *retval) 393 { 394 /* { 395 syscallarg(linux32_sigcontextp_t) scp; 396 } */ 397 struct linux32_sigcontext ctx; 398 int error; 399 400 if ((error = copyin(SCARG_P32(uap, scp), &ctx, sizeof(ctx))) != 0) 401 return error; 402 403 return linux32_restore_sigcontext(l, &ctx, retval); 404 } 405 406 int 407 linux32_sys_rt_sigreturn(struct lwp *l, 408 const struct linux32_sys_rt_sigreturn_args *uap, register_t *retval) 409 { 410 /* { 411 syscallarg(linux32_ucontextp_t) ucp; 412 } */ 413 struct linux32_ucontext ctx; 414 int error; 415 416 if ((error = copyin(SCARG_P32(uap, ucp), &ctx, sizeof(ctx))) != 0) 417 return error; 418 419 return linux32_restore_sigcontext(l, &ctx.uc_mcontext, retval); 420 } 421 422 static int 423 linux32_restore_sigcontext(struct lwp *l, struct linux32_sigcontext *scp, 424 register_t *retval) 425 { 426 struct trapframe *tf; 427 struct proc *p = l->l_proc; 428 struct sigaltstack *sas = &l->l_sigstk; 429 sigset_t mask; 430 ssize_t ss_gap; 431 432 /* Restore register context. */ 433 tf = l->l_md.md_regs; 434 435 /* 436 * Check for security violations. If we're returning to 437 * protected mode, the CPU will validate the segment registers 438 * automatically and generate a trap on violations. We handle 439 * the trap, rather than doing all of the checking here. 440 */ 441 if (((scp->sc_eflags ^ tf->tf_rflags) & PSL_USERSTATIC) != 0 || 442 !USERMODE(scp->sc_cs, scp->sc_eflags)) 443 return EINVAL; 444 445 if (scp->sc_fs != 0 && !VALID_USER_DSEL32(scp->sc_fs)) 446 return EINVAL; 447 448 if (scp->sc_gs != 0 && !VALID_USER_DSEL32(scp->sc_gs)) 449 return EINVAL; 450 451 if (scp->sc_es != 0 && !VALID_USER_DSEL32(scp->sc_es)) 452 return EINVAL; 453 454 if (!VALID_USER_DSEL32(scp->sc_ds) || 455 !VALID_USER_DSEL32(scp->sc_ss)) 456 return EINVAL; 457 458 if (scp->sc_eip >= VM_MAXUSER_ADDRESS32) 459 return EINVAL; 460 461 tf->tf_gs = (register_t)scp->sc_gs & 0xffffffff; 462 tf->tf_fs = (register_t)scp->sc_fs & 0xffffffff; 463 tf->tf_es = (register_t)scp->sc_es & 0xffffffff; 464 tf->tf_ds = (register_t)scp->sc_ds & 0xffffffff; 465 tf->tf_rflags &= ~PSL_USER; 466 tf->tf_rflags |= ((register_t)scp->sc_eflags & PSL_USER); 467 tf->tf_rdi = (register_t)scp->sc_edi & 0xffffffff; 468 tf->tf_rsi = (register_t)scp->sc_esi & 0xffffffff; 469 tf->tf_rbp = (register_t)scp->sc_ebp & 0xffffffff; 470 tf->tf_rbx = (register_t)scp->sc_ebx & 0xffffffff; 471 tf->tf_rdx = (register_t)scp->sc_edx & 0xffffffff; 472 tf->tf_rcx = (register_t)scp->sc_ecx & 0xffffffff; 473 tf->tf_rax = (register_t)scp->sc_eax & 0xffffffff; 474 tf->tf_rip = (register_t)scp->sc_eip & 0xffffffff; 475 tf->tf_cs = (register_t)scp->sc_cs & 0xffffffff; 476 tf->tf_rsp = (register_t)scp->sc_esp_at_signal & 0xffffffff; 477 tf->tf_ss = (register_t)scp->sc_ss & 0xffffffff; 478 479 mutex_enter(p->p_lock); 480 481 /* Restore signal stack. */ 482 ss_gap = (ssize_t) 483 ((char *)NETBSD32IPTR64(scp->sc_esp_at_signal) 484 - (char *)sas->ss_sp); 485 if (ss_gap >= 0 && ss_gap < sas->ss_size) 486 sas->ss_flags |= SS_ONSTACK; 487 else 488 sas->ss_flags &= ~SS_ONSTACK; 489 490 /* Restore signal mask. */ 491 linux32_old_to_native_sigset(&mask, &scp->sc_mask); 492 (void) sigprocmask1(l, SIG_SETMASK, &mask, 0); 493 494 mutex_exit(p->p_lock); 495 496 #ifdef DEBUG_LINUX 497 printf("linux32_sigreturn: rip = 0x%lx, rsp = 0x%lx, flags = 0x%lx\n", 498 tf->tf_rip, tf->tf_rsp, tf->tf_rflags); 499 #endif 500 return EJUSTRETURN; 501 } 502