1 /* $NetBSD: linux_machdep.c,v 1.25 2007/10/19 18:52:10 njoly Exp $ */ 2 3 /*- 4 * Copyright (c) 2005 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 34 #include <sys/cdefs.h> 35 36 __KERNEL_RCSID(0, "$NetBSD: linux_machdep.c,v 1.25 2007/10/19 18:52:10 njoly Exp $"); 37 38 #include <sys/param.h> 39 #include <sys/types.h> 40 #include <sys/systm.h> 41 #include <sys/signal.h> 42 #include <sys/exec.h> 43 #include <sys/proc.h> 44 #include <sys/ptrace.h> /* for process_read_fpregs() */ 45 #include <sys/user.h> 46 #include <sys/wait.h> 47 #include <sys/ucontext.h> 48 #include <sys/conf.h> 49 50 #include <machine/reg.h> 51 #include <machine/pcb.h> 52 #include <machine/fpu.h> 53 #include <machine/mcontext.h> 54 #include <machine/specialreg.h> 55 #include <machine/vmparam.h> 56 57 /* 58 * To see whether wscons is configured (for virtual console ioctl calls). 59 */ 60 #if defined(_KERNEL_OPT) 61 #include "wsdisplay.h" 62 #endif 63 #if (NWSDISPLAY > 0) 64 #include <dev/wscons/wsconsio.h> 65 #include <dev/wscons/wsdisplay_usl_io.h> 66 #endif 67 68 69 #include <compat/linux/common/linux_signal.h> 70 #include <compat/linux/common/linux_errno.h> 71 #include <compat/linux/common/linux_exec.h> 72 #include <compat/linux/common/linux_ioctl.h> 73 #include <compat/linux/common/linux_prctl.h> 74 #include <compat/linux/common/linux_machdep.h> 75 #include <compat/linux/common/linux_ipc.h> 76 #include <compat/linux/common/linux_sem.h> 77 #include <compat/linux/linux_syscall.h> 78 #include <compat/linux/linux_syscallargs.h> 79 80 static void linux_buildcontext(struct lwp *, void *, void *); 81 82 void 83 linux_setregs(l, epp, stack) 84 struct lwp *l; 85 struct exec_package *epp; 86 u_long stack; 87 { 88 struct pcb *pcb = &l->l_addr->u_pcb; 89 struct trapframe *tf; 90 91 /* If we were using the FPU, forget about it. */ 92 if (l->l_addr->u_pcb.pcb_fpcpu != NULL) 93 fpusave_lwp(l, 0); 94 95 l->l_md.md_flags &= ~MDP_USEDFPU; 96 pcb->pcb_flags = 0; 97 pcb->pcb_savefpu.fp_fxsave.fx_fcw = __NetBSD_NPXCW__; 98 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr = __INITIAL_MXCSR__; 99 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr_mask = __INITIAL_MXCSR_MASK__; 100 pcb->pcb_fs = 0; 101 pcb->pcb_gs = 0; 102 103 l->l_proc->p_flag &= ~PK_32; 104 105 tf = l->l_md.md_regs; 106 tf->tf_rax = 0; 107 tf->tf_rbx = 0; 108 tf->tf_rcx = epp->ep_entry; 109 tf->tf_rdx = 0; 110 tf->tf_rsi = 0; 111 tf->tf_rdi = 0; 112 tf->tf_rbp = 0; 113 tf->tf_rsp = stack; 114 tf->tf_r8 = 0; 115 tf->tf_r9 = 0; 116 tf->tf_r10 = 0; 117 tf->tf_r11 = 0; 118 tf->tf_r12 = 0; 119 tf->tf_r13 = 0; 120 tf->tf_r14 = 0; 121 tf->tf_r15 = 0; 122 tf->tf_rip = epp->ep_entry; 123 tf->tf_rflags = PSL_USERSET; 124 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 125 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 126 tf->tf_ds = 0; 127 tf->tf_es = 0; 128 tf->tf_fs = 0; 129 tf->tf_gs = 0; 130 131 return; 132 } 133 134 void 135 linux_sendsig(ksi, mask) 136 const ksiginfo_t *ksi; 137 const sigset_t *mask; 138 { 139 struct lwp *l = curlwp; 140 struct proc *p = l->l_proc; 141 struct sigacts *ps = p->p_sigacts; 142 int onstack, error; 143 int sig = ksi->ksi_signo; 144 struct linux_rt_sigframe *sfp, sigframe; 145 struct linux__fpstate *fpsp, fpstate; 146 struct fpreg fpregs; 147 struct trapframe *tf = l->l_md.md_regs; 148 sig_t catcher = SIGACTION(p, sig).sa_handler; 149 linux_sigset_t lmask; 150 char *sp; 151 152 /* Do we need to jump onto the signal stack? */ 153 onstack = 154 (l->l_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 155 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 156 157 /* Allocate space for the signal handler context. */ 158 if (onstack) 159 sp = ((char *)l->l_sigstk.ss_sp + 160 l->l_sigstk.ss_size); 161 else 162 sp = (char *)tf->tf_rsp - 128; 163 164 /* 165 * Save FPU state, if any 166 */ 167 if (l->l_md.md_flags & MDP_USEDFPU) { 168 sp = (char *) 169 (((long)sp - sizeof(struct linux__fpstate)) & ~0xfUL); 170 fpsp = (struct linux__fpstate *)sp; 171 } else 172 fpsp = NULL; 173 174 /* 175 * Populate the rt_sigframe 176 */ 177 sp = (char *) 178 ((((long)sp - sizeof(struct linux_rt_sigframe)) & ~0xfUL) - 8); 179 sfp = (struct linux_rt_sigframe *)sp; 180 181 bzero(&sigframe, sizeof(sigframe)); 182 if (ps->sa_sigdesc[sig].sd_vers != 0) 183 sigframe.pretcode = 184 (char *)(u_long)ps->sa_sigdesc[sig].sd_tramp; 185 else 186 sigframe.pretcode = NULL; 187 188 /* 189 * The user context 190 */ 191 sigframe.uc.luc_flags = 0; 192 sigframe.uc.luc_link = NULL; 193 194 /* This is used regardless of SA_ONSTACK in Linux */ 195 sigframe.uc.luc_stack.ss_sp = l->l_sigstk.ss_sp; 196 sigframe.uc.luc_stack.ss_size = l->l_sigstk.ss_size; 197 sigframe.uc.luc_stack.ss_flags = 0; 198 if (l->l_sigstk.ss_flags & SS_ONSTACK) 199 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_ONSTACK; 200 if (l->l_sigstk.ss_flags & SS_DISABLE) 201 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_DISABLE; 202 203 sigframe.uc.luc_mcontext.r8 = tf->tf_r8; 204 sigframe.uc.luc_mcontext.r9 = tf->tf_r9; 205 sigframe.uc.luc_mcontext.r10 = tf->tf_r10; 206 sigframe.uc.luc_mcontext.r11 = tf->tf_r11; 207 sigframe.uc.luc_mcontext.r12 = tf->tf_r12; 208 sigframe.uc.luc_mcontext.r13 = tf->tf_r13; 209 sigframe.uc.luc_mcontext.r14 = tf->tf_r14; 210 sigframe.uc.luc_mcontext.r15 = tf->tf_r15; 211 sigframe.uc.luc_mcontext.rdi = tf->tf_rdi; 212 sigframe.uc.luc_mcontext.rsi = tf->tf_rsi; 213 sigframe.uc.luc_mcontext.rbp = tf->tf_rbp; 214 sigframe.uc.luc_mcontext.rbx = tf->tf_rbx; 215 sigframe.uc.luc_mcontext.rdx = tf->tf_rdx; 216 sigframe.uc.luc_mcontext.rax = tf->tf_rax; 217 sigframe.uc.luc_mcontext.rcx = tf->tf_rcx; 218 sigframe.uc.luc_mcontext.rsp = tf->tf_rsp; 219 sigframe.uc.luc_mcontext.rip = tf->tf_rip; 220 sigframe.uc.luc_mcontext.eflags = tf->tf_rflags; 221 sigframe.uc.luc_mcontext.cs = tf->tf_cs; 222 sigframe.uc.luc_mcontext.gs = tf->tf_gs; 223 sigframe.uc.luc_mcontext.fs = tf->tf_fs; 224 sigframe.uc.luc_mcontext.err = tf->tf_err; 225 sigframe.uc.luc_mcontext.trapno = tf->tf_trapno; 226 native_to_linux_sigset(&lmask, mask); 227 sigframe.uc.luc_mcontext.oldmask = lmask.sig[0]; 228 sigframe.uc.luc_mcontext.cr2 = (long)l->l_addr->u_pcb.pcb_onfault; 229 sigframe.uc.luc_mcontext.fpstate = fpsp; 230 native_to_linux_sigset(&sigframe.uc.luc_sigmask, mask); 231 232 /* 233 * the siginfo structure 234 */ 235 sigframe.info.lsi_signo = native_to_linux_signo[sig]; 236 sigframe.info.lsi_errno = native_to_linux_errno[ksi->ksi_errno]; 237 sigframe.info.lsi_code = ksi->ksi_code; 238 239 /* XXX This is a rought conversion, taken from i386 code */ 240 switch (sigframe.info.lsi_signo) { 241 case LINUX_SIGILL: 242 case LINUX_SIGFPE: 243 case LINUX_SIGSEGV: 244 case LINUX_SIGBUS: 245 case LINUX_SIGTRAP: 246 sigframe.info._sifields._sigfault._addr = ksi->ksi_addr; 247 break; 248 case LINUX_SIGCHLD: 249 sigframe.info._sifields._sigchld._pid = ksi->ksi_pid; 250 sigframe.info._sifields._sigchld._uid = ksi->ksi_uid; 251 sigframe.info._sifields._sigchld._utime = ksi->ksi_utime; 252 sigframe.info._sifields._sigchld._stime = ksi->ksi_stime; 253 254 if (WCOREDUMP(ksi->ksi_status)) { 255 sigframe.info.lsi_code = LINUX_CLD_DUMPED; 256 sigframe.info._sifields._sigchld._status = 257 _WSTATUS(ksi->ksi_status); 258 } else if (_WSTATUS(ksi->ksi_status)) { 259 sigframe.info.lsi_code = LINUX_CLD_KILLED; 260 sigframe.info._sifields._sigchld._status = 261 _WSTATUS(ksi->ksi_status); 262 } else { 263 sigframe.info.lsi_code = LINUX_CLD_EXITED; 264 sigframe.info._sifields._sigchld._status = 265 ((ksi->ksi_status & 0xff00U) >> 8); 266 } 267 break; 268 case LINUX_SIGIO: 269 sigframe.info._sifields._sigpoll._band = ksi->ksi_band; 270 sigframe.info._sifields._sigpoll._fd = ksi->ksi_fd; 271 break; 272 default: 273 sigframe.info._sifields._sigchld._pid = ksi->ksi_pid; 274 sigframe.info._sifields._sigchld._uid = ksi->ksi_uid; 275 if ((sigframe.info.lsi_signo == LINUX_SIGALRM) || 276 (sigframe.info.lsi_signo >= LINUX_SIGRTMIN)) 277 sigframe.info._sifields._timer._sigval.sival_ptr = 278 ksi->ksi_value.sival_ptr; 279 break; 280 } 281 282 sendsig_reset(l, sig); 283 mutex_exit(&p->p_smutex); 284 error = 0; 285 286 /* 287 * Save FPU state, if any 288 */ 289 if (fpsp != NULL) { 290 (void)process_read_fpregs(l, &fpregs); 291 bzero(&fpstate, sizeof(fpstate)); 292 fpstate.cwd = fpregs.fp_fcw; 293 fpstate.swd = fpregs.fp_fsw; 294 fpstate.twd = fpregs.fp_ftw; 295 fpstate.fop = fpregs.fp_fop; 296 fpstate.rip = fpregs.fp_rip; 297 fpstate.rdp = fpregs.fp_rdp; 298 fpstate.mxcsr = fpregs.fp_mxcsr; 299 fpstate.mxcsr_mask = fpregs.fp_mxcsr_mask; 300 memcpy(&fpstate.st_space, &fpregs.fp_st, 301 sizeof(fpstate.st_space)); 302 memcpy(&fpstate.xmm_space, &fpregs.fp_xmm, 303 sizeof(fpstate.xmm_space)); 304 error = copyout(&fpstate, fpsp, sizeof(fpstate)); 305 } 306 307 if (error == 0) 308 error = copyout(&sigframe, sp, sizeof(sigframe)); 309 310 mutex_enter(&p->p_smutex); 311 312 if (error != 0) { 313 sigexit(l, SIGILL); 314 return; 315 } 316 317 linux_buildcontext(l, catcher, sp); 318 tf->tf_rdi = sigframe.info.lsi_signo; 319 tf->tf_rax = 0; 320 tf->tf_rsi = (long)&sfp->info; 321 tf->tf_rdx = (long)&sfp->uc; 322 323 /* 324 * Remember we use signal stack 325 */ 326 if (onstack) 327 l->l_sigstk.ss_flags |= SS_ONSTACK; 328 return; 329 } 330 331 int 332 linux_sys_modify_ldt(l, v, retval) 333 struct lwp *l; 334 void *v; 335 register_t *retval; 336 { 337 printf("linux_sys_modify_ldt\n"); 338 return 0; 339 } 340 341 int 342 linux_sys_iopl(l, v, retval) 343 struct lwp *l; 344 void *v; 345 register_t *retval; 346 { 347 return 0; 348 } 349 350 int 351 linux_sys_ioperm(l, v, retval) 352 struct lwp *l; 353 void *v; 354 register_t *retval; 355 { 356 return 0; 357 } 358 359 dev_t 360 linux_fakedev(dev, raw) 361 dev_t dev; 362 int raw; 363 { 364 365 extern const struct cdevsw ptc_cdevsw, pts_cdevsw; 366 const struct cdevsw *cd = cdevsw_lookup(dev); 367 368 if (raw) { 369 #if (NWSDISPLAY > 0) 370 extern const struct cdevsw wsdisplay_cdevsw; 371 if (cd == &wsdisplay_cdevsw) 372 return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1)); 373 #endif 374 } 375 376 if (cd == &ptc_cdevsw) 377 return makedev(LINUX_PTC_MAJOR, minor(dev)); 378 if (cd == &pts_cdevsw) 379 return makedev(LINUX_PTS_MAJOR, minor(dev)); 380 381 return ((minor(dev) & 0xff) | ((major(dev) & 0xfff) << 8) 382 | (((unsigned long long int) (minor(dev) & ~0xff)) << 12) 383 | (((unsigned long long int) (major(dev) & ~0xfff)) << 32)); 384 } 385 386 int 387 linux_machdepioctl(l, v, retval) 388 struct lwp *l; 389 void *v; 390 register_t *retval; 391 { 392 return 0; 393 } 394 395 int 396 linux_sys_rt_sigreturn(l, v, retval) 397 struct lwp *l; 398 void *v; 399 register_t *retval; 400 { 401 struct linux_ucontext *luctx; 402 struct trapframe *tf = l->l_md.md_regs; 403 struct linux_sigcontext *lsigctx; 404 struct linux__fpstate fpstate; 405 struct linux_rt_sigframe frame, *fp; 406 ucontext_t uctx; 407 mcontext_t *mctx; 408 struct fxsave64 *fxarea; 409 int error; 410 411 fp = (struct linux_rt_sigframe *)(tf->tf_rsp - 8); 412 if ((error = copyin(fp, &frame, sizeof(frame))) != 0) { 413 mutex_enter(&l->l_proc->p_smutex); 414 sigexit(l, SIGILL); 415 return error; 416 } 417 luctx = &frame.uc; 418 lsigctx = &luctx->luc_mcontext; 419 420 bzero(&uctx, sizeof(uctx)); 421 mctx = (mcontext_t *)&uctx.uc_mcontext; 422 fxarea = (struct fxsave64 *)&mctx->__fpregs; 423 424 /* 425 * Set the flags. Linux always have CPU, stack and signal state, 426 * FPU is optional. uc_flags is not used to tell what we have. 427 */ 428 uctx.uc_flags = (_UC_SIGMASK|_UC_CPU|_UC_STACK|_UC_CLRSTACK); 429 if (lsigctx->fpstate != NULL) 430 uctx.uc_flags |= _UC_FPU; 431 uctx.uc_link = NULL; 432 433 /* 434 * Signal set 435 */ 436 linux_to_native_sigset(&uctx.uc_sigmask, &luctx->luc_sigmask); 437 438 /* 439 * CPU state 440 */ 441 mctx->__gregs[_REG_R8] = lsigctx->r8; 442 mctx->__gregs[_REG_R9] = lsigctx->r9; 443 mctx->__gregs[_REG_R10] = lsigctx->r10; 444 mctx->__gregs[_REG_R11] = lsigctx->r11; 445 mctx->__gregs[_REG_R12] = lsigctx->r12; 446 mctx->__gregs[_REG_R13] = lsigctx->r13; 447 mctx->__gregs[_REG_R14] = lsigctx->r14; 448 mctx->__gregs[_REG_R15] = lsigctx->r15; 449 mctx->__gregs[_REG_RDI] = lsigctx->rdi; 450 mctx->__gregs[_REG_RSI] = lsigctx->rsi; 451 mctx->__gregs[_REG_RBP] = lsigctx->rbp; 452 mctx->__gregs[_REG_RBX] = lsigctx->rbx; 453 mctx->__gregs[_REG_RAX] = lsigctx->rax; 454 mctx->__gregs[_REG_RDX] = lsigctx->rdx; 455 mctx->__gregs[_REG_RCX] = lsigctx->rcx; 456 mctx->__gregs[_REG_RIP] = lsigctx->rip; 457 mctx->__gregs[_REG_RFL] = lsigctx->eflags; 458 mctx->__gregs[_REG_CS] = lsigctx->cs; 459 mctx->__gregs[_REG_GS] = lsigctx->gs; 460 mctx->__gregs[_REG_FS] = lsigctx->fs; 461 mctx->__gregs[_REG_ERR] = lsigctx->err; 462 mctx->__gregs[_REG_TRAPNO] = lsigctx->trapno; 463 mctx->__gregs[_REG_ES] = tf->tf_es; 464 mctx->__gregs[_REG_DS] = tf->tf_ds; 465 mctx->__gregs[_REG_URSP] = lsigctx->rsp; /* XXX */ 466 mctx->__gregs[_REG_SS] = tf->tf_ss; 467 468 /* 469 * FPU state 470 */ 471 if (lsigctx->fpstate != NULL) { 472 error = copyin(lsigctx->fpstate, &fpstate, sizeof(fpstate)); 473 if (error != 0) { 474 mutex_enter(&l->l_proc->p_smutex); 475 sigexit(l, SIGILL); 476 return error; 477 } 478 479 fxarea->fx_fcw = fpstate.cwd; 480 fxarea->fx_fsw = fpstate.swd; 481 fxarea->fx_ftw = fpstate.twd; 482 fxarea->fx_fop = fpstate.fop; 483 fxarea->fx_rip = fpstate.rip; 484 fxarea->fx_rdp = fpstate.rdp; 485 fxarea->fx_mxcsr = fpstate.mxcsr; 486 fxarea->fx_mxcsr_mask = fpstate.mxcsr_mask; 487 memcpy(&fxarea->fx_st, &fpstate.st_space, 488 sizeof(fxarea->fx_st)); 489 memcpy(&fxarea->fx_xmm, &fpstate.xmm_space, 490 sizeof(fxarea->fx_xmm)); 491 } 492 493 /* 494 * And the stack 495 */ 496 uctx.uc_stack.ss_flags = 0; 497 if (luctx->luc_stack.ss_flags & LINUX_SS_ONSTACK); 498 uctx.uc_stack.ss_flags = SS_ONSTACK; 499 500 if (luctx->luc_stack.ss_flags & LINUX_SS_DISABLE); 501 uctx.uc_stack.ss_flags = SS_DISABLE; 502 503 uctx.uc_stack.ss_sp = luctx->luc_stack.ss_sp; 504 uctx.uc_stack.ss_size = luctx->luc_stack.ss_size; 505 506 /* 507 * And let setucontext deal with that. 508 */ 509 mutex_enter(&l->l_proc->p_smutex); 510 error = setucontext(l, &uctx); 511 mutex_exit(&l->l_proc->p_smutex); 512 if (error) 513 return error; 514 515 return EJUSTRETURN; 516 } 517 518 int 519 linux_sys_arch_prctl(l, v, retval) 520 struct lwp *l; 521 void *v; 522 register_t *retval; 523 { 524 struct linux_sys_arch_prctl_args /* { 525 syscallarg(int) code; 526 syscallarg(unsigned long) addr; 527 } */ *uap = v; 528 struct pcb *pcb = &l->l_addr->u_pcb; 529 struct trapframe *tf = l->l_md.md_regs; 530 int error; 531 uint64_t taddr; 532 533 switch(SCARG(uap, code)) { 534 case LINUX_ARCH_SET_GS: 535 taddr = SCARG(uap, addr); 536 if (taddr >= VM_MAXUSER_ADDRESS) 537 return EINVAL; 538 pcb->pcb_gs = taddr; 539 pcb->pcb_flags |= PCB_GS64; 540 if (l == curlwp) 541 wrmsr(MSR_KERNELGSBASE, taddr); 542 break; 543 544 case LINUX_ARCH_GET_GS: 545 if (pcb->pcb_flags & PCB_GS64) 546 taddr = pcb->pcb_gs; 547 else { 548 error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr); 549 if (error != 0) 550 return error; 551 } 552 error = copyout(&taddr, (char *)SCARG(uap, addr), 8); 553 if (error != 0) 554 return error; 555 break; 556 557 case LINUX_ARCH_SET_FS: 558 taddr = SCARG(uap, addr); 559 if (taddr >= VM_MAXUSER_ADDRESS) 560 return EINVAL; 561 pcb->pcb_fs = taddr; 562 pcb->pcb_flags |= PCB_FS64; 563 if (l == curlwp) 564 wrmsr(MSR_FSBASE, taddr); 565 break; 566 567 case LINUX_ARCH_GET_FS: 568 if (pcb->pcb_flags & PCB_FS64) 569 taddr = pcb->pcb_fs; 570 else { 571 error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr); 572 if (error != 0) 573 return error; 574 } 575 error = copyout(&taddr, (char *)SCARG(uap, addr), 8); 576 if (error != 0) 577 return error; 578 break; 579 580 default: 581 #ifdef DEBUG_LINUX 582 printf("linux_sys_arch_prctl: unexpected code %d\n", 583 SCARG(uap, code)); 584 #endif 585 return EINVAL; 586 } 587 588 return 0; 589 } 590 591 const int linux_vsyscall_to_syscall[] = { 592 LINUX_SYS_gettimeofday, 593 LINUX_SYS_time, 594 LINUX_SYS_nosys, 595 LINUX_SYS_nosys, 596 }; 597 598 int 599 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg) 600 { 601 struct trapframe *tf = arg; 602 uint64_t retaddr; 603 int vsyscallnr; 604 605 /* 606 * Check for a vsyscall. %rip must be the fault address, 607 * and the address must be in the Linux vsyscall area. 608 * Also, vsyscalls are only done at 1024-byte boundaries. 609 */ 610 611 if (__predict_true(trapaddr < LINUX_VSYSCALL_START)) 612 return 0; 613 614 if (trapaddr != tf->tf_rip) 615 return 0; 616 617 if ((tf->tf_rip & (LINUX_VSYSCALL_SIZE - 1)) != 0) 618 return 0; 619 620 vsyscallnr = (tf->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SIZE; 621 622 if (vsyscallnr > LINUX_VSYSCALL_MAXNR) 623 return 0; 624 625 /* 626 * Get the return address from the top of the stack, 627 * and fix up the return address. 628 * This assumes the faulting instruction was callq *reg, 629 * which is the only way that vsyscalls are ever entered. 630 */ 631 if (copyin((void *)tf->tf_rsp, &retaddr, sizeof retaddr) != 0) 632 return 0; 633 tf->tf_rip = retaddr; 634 tf->tf_rax = linux_vsyscall_to_syscall[vsyscallnr]; 635 tf->tf_rsp += 8; /* "pop" the return address */ 636 637 #if 0 638 printf("usertrap: rip %p rsp %p retaddr %p vsys %d sys %d\n", 639 (void *)tf->tf_rip, (void *)tf->tf_rsp, (void *)retaddr, 640 vsyscallnr, (int)tf->tf_rax); 641 #endif 642 643 (*l->l_proc->p_md.md_syscall)(tf); 644 645 return 1; 646 } 647 648 static void 649 linux_buildcontext(struct lwp *l, void *catcher, void *f) 650 { 651 struct trapframe *tf = l->l_md.md_regs; 652 653 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 654 tf->tf_rip = (u_int64_t)catcher; 655 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 656 tf->tf_rflags &= ~(PSL_T|PSL_VM|PSL_AC); 657 tf->tf_rsp = (u_int64_t)f; 658 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 659 } 660 661 void * 662 linux_get_newtls(l) 663 struct lwp *l; 664 { 665 struct trapframe *tf = l->l_md.md_regs; 666 667 return (void *)tf->tf_r8; 668 } 669 670 int 671 linux_set_newtls(l, tls) 672 struct lwp *l; 673 void *tls; 674 { 675 struct linux_sys_arch_prctl_args cup; 676 register_t retval; 677 678 SCARG(&cup, code) = LINUX_ARCH_SET_FS; 679 SCARG(&cup, addr) = (unsigned long)tls; 680 681 return linux_sys_arch_prctl(l, &cup, &retval); 682 } 683