linux_machdep.c revision 1.15.2.1.2.1
1 /* $NetBSD: linux_machdep.c,v 1.15.2.1.2.1 2007/06/04 01:54:19 wrstuden 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 1.15.2.1.2.1 2007/06/04 01:54:19 wrstuden 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/linux_syscall.h> 76 #include <compat/linux/linux_syscallargs.h> 77 78 static void linux_buildcontext(struct lwp *, void *, void *); 79 80 void 81 linux_setregs(l, epp, stack) 82 struct lwp *l; 83 struct exec_package *epp; 84 u_long stack; 85 { 86 struct pcb *pcb = &l->l_addr->u_pcb; 87 struct trapframe *tf; 88 89 /* If we were using the FPU, forget about it. */ 90 if (l->l_addr->u_pcb.pcb_fpcpu != NULL) 91 fpusave_lwp(l, 0); 92 93 l->l_md.md_flags &= ~MDP_USEDFPU; 94 pcb->pcb_flags = 0; 95 pcb->pcb_savefpu.fp_fxsave.fx_fcw = __NetBSD_NPXCW__; 96 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr = __INITIAL_MXCSR__; 97 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr_mask = __INITIAL_MXCSR_MASK__; 98 pcb->pcb_fs = 0; 99 pcb->pcb_gs = 0; 100 101 l->l_proc->p_flag &= ~P_32; 102 103 tf = l->l_md.md_regs; 104 tf->tf_rax = 0; 105 tf->tf_rbx = 0; 106 tf->tf_rcx = epp->ep_entry; 107 tf->tf_rdx = 0; 108 tf->tf_rsi = 0; 109 tf->tf_rdi = 0; 110 tf->tf_rbp = 0; 111 tf->tf_rsp = stack; 112 tf->tf_r8 = 0; 113 tf->tf_r9 = 0; 114 tf->tf_r10 = 0; 115 tf->tf_r11 = 0; 116 tf->tf_r12 = 0; 117 tf->tf_r13 = 0; 118 tf->tf_r14 = 0; 119 tf->tf_r15 = 0; 120 tf->tf_rip = epp->ep_entry; 121 tf->tf_rflags = PSL_USERSET; 122 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 123 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 124 tf->tf_ds = 0; 125 tf->tf_es = 0; 126 tf->tf_fs = 0; 127 tf->tf_gs = 0; 128 129 return; 130 } 131 132 void 133 linux_sendsig(ksi, mask) 134 const ksiginfo_t *ksi; 135 const sigset_t *mask; 136 { 137 struct lwp *l = curlwp; 138 struct proc *p = l->l_proc; 139 struct sigacts *ps = p->p_sigacts; 140 int onstack; 141 int sig = ksi->ksi_signo; 142 struct linux_rt_sigframe *sfp, sigframe; 143 struct linux__fpstate *fpsp, fpstate; 144 struct fpreg fpregs; 145 struct trapframe *tf = l->l_md.md_regs; 146 sig_t catcher = SIGACTION(p, sig).sa_handler; 147 linux_sigset_t lmask; 148 char *sp; 149 int error; 150 151 /* Do we need to jump onto the signal stack? */ 152 onstack = 153 (p->p_sigctx.ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 154 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 155 156 /* Allocate space for the signal handler context. */ 157 if (onstack) 158 sp = ((caddr_t)p->p_sigctx.ps_sigstk.ss_sp + 159 p->p_sigctx.ps_sigstk.ss_size); 160 else 161 sp = (caddr_t)tf->tf_rsp - 128; 162 163 164 /* 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; 171 172 (void)process_read_fpregs(l, &fpregs); 173 bzero(&fpstate, sizeof(fpstate)); 174 175 fpstate.cwd = fpregs.fp_fcw; 176 fpstate.swd = fpregs.fp_fsw; 177 fpstate.twd = fpregs.fp_ftw; 178 fpstate.fop = fpregs.fp_fop; 179 fpstate.rip = fpregs.fp_rip; 180 fpstate.rdp = fpregs.fp_rdp; 181 fpstate.mxcsr = fpregs.fp_mxcsr; 182 fpstate.mxcsr_mask = fpregs.fp_mxcsr_mask; 183 memcpy(&fpstate.st_space, &fpregs.fp_st, 184 sizeof(fpstate.st_space)); 185 memcpy(&fpstate.xmm_space, &fpregs.fp_xmm, 186 sizeof(fpstate.xmm_space)); 187 188 if ((error = copyout(&fpstate, fpsp, sizeof(fpstate))) != 0) { 189 sigexit(l, SIGILL); 190 return; 191 } 192 } else { 193 fpsp = NULL; 194 } 195 196 /* 197 * Populate the rt_sigframe 198 */ 199 sp = (char *) 200 ((((long)sp - sizeof(struct linux_rt_sigframe)) & ~0xfUL) - 8); 201 sfp = (struct linux_rt_sigframe *)sp; 202 203 bzero(&sigframe, sizeof(sigframe)); 204 if (ps->sa_sigdesc[sig].sd_vers != 0) 205 sigframe.pretcode = 206 (char *)(u_long)ps->sa_sigdesc[sig].sd_tramp; 207 else 208 sigframe.pretcode = NULL; 209 210 /* 211 * The user context 212 */ 213 sigframe.uc.luc_flags = 0; 214 sigframe.uc.luc_link = NULL; 215 216 /* This is used regardless of SA_ONSTACK in Linux */ 217 sigframe.uc.luc_stack.ss_sp = p->p_sigctx.ps_sigstk.ss_sp; 218 sigframe.uc.luc_stack.ss_size = p->p_sigctx.ps_sigstk.ss_size; 219 sigframe.uc.luc_stack.ss_flags = 0; 220 if (p->p_sigctx.ps_sigstk.ss_flags & SS_ONSTACK) 221 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_ONSTACK; 222 if (p->p_sigctx.ps_sigstk.ss_flags & SS_DISABLE) 223 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_DISABLE; 224 225 sigframe.uc.luc_mcontext.r8 = tf->tf_r8; 226 sigframe.uc.luc_mcontext.r9 = tf->tf_r9; 227 sigframe.uc.luc_mcontext.r10 = tf->tf_r10; 228 sigframe.uc.luc_mcontext.r11 = tf->tf_r11; 229 sigframe.uc.luc_mcontext.r12 = tf->tf_r12; 230 sigframe.uc.luc_mcontext.r13 = tf->tf_r13; 231 sigframe.uc.luc_mcontext.r14 = tf->tf_r14; 232 sigframe.uc.luc_mcontext.r15 = tf->tf_r15; 233 sigframe.uc.luc_mcontext.rdi = tf->tf_rdi; 234 sigframe.uc.luc_mcontext.rsi = tf->tf_rsi; 235 sigframe.uc.luc_mcontext.rbp = tf->tf_rbp; 236 sigframe.uc.luc_mcontext.rbx = tf->tf_rbx; 237 sigframe.uc.luc_mcontext.rdx = tf->tf_rdx; 238 sigframe.uc.luc_mcontext.rax = tf->tf_rax; 239 sigframe.uc.luc_mcontext.rcx = tf->tf_rcx; 240 sigframe.uc.luc_mcontext.rsp = tf->tf_rsp; 241 sigframe.uc.luc_mcontext.rip = tf->tf_rip; 242 sigframe.uc.luc_mcontext.eflags = tf->tf_rflags; 243 sigframe.uc.luc_mcontext.cs = tf->tf_cs; 244 sigframe.uc.luc_mcontext.gs = tf->tf_gs; 245 sigframe.uc.luc_mcontext.fs = tf->tf_fs; 246 sigframe.uc.luc_mcontext.err = tf->tf_err; 247 sigframe.uc.luc_mcontext.trapno = tf->tf_trapno; 248 native_to_linux_sigset(&lmask, mask); 249 sigframe.uc.luc_mcontext.oldmask = lmask.sig[0]; 250 sigframe.uc.luc_mcontext.cr2 = (long)l->l_addr->u_pcb.pcb_onfault; 251 sigframe.uc.luc_mcontext.fpstate = fpsp; 252 native_to_linux_sigset(&sigframe.uc.luc_sigmask, mask); 253 254 /* 255 * the siginfo structure 256 */ 257 sigframe.info.lsi_signo = native_to_linux_signo[sig]; 258 sigframe.info.lsi_errno = native_to_linux_errno[ksi->ksi_errno]; 259 sigframe.info.lsi_code = ksi->ksi_code; 260 261 /* XXX This is a rought conversion, taken from i386 code */ 262 switch (sigframe.info.lsi_signo) { 263 case LINUX_SIGILL: 264 case LINUX_SIGFPE: 265 case LINUX_SIGSEGV: 266 case LINUX_SIGBUS: 267 case LINUX_SIGTRAP: 268 sigframe.info._sifields._sigfault._addr = ksi->ksi_addr; 269 break; 270 case LINUX_SIGCHLD: 271 sigframe.info._sifields._sigchld._pid = ksi->ksi_pid; 272 sigframe.info._sifields._sigchld._uid = ksi->ksi_uid; 273 sigframe.info._sifields._sigchld._utime = ksi->ksi_utime; 274 sigframe.info._sifields._sigchld._stime = ksi->ksi_stime; 275 276 if (WCOREDUMP(ksi->ksi_status)) { 277 sigframe.info.lsi_code = LINUX_CLD_DUMPED; 278 sigframe.info._sifields._sigchld._status = 279 _WSTATUS(ksi->ksi_status); 280 } else if (_WSTATUS(ksi->ksi_status)) { 281 sigframe.info.lsi_code = LINUX_CLD_KILLED; 282 sigframe.info._sifields._sigchld._status = 283 _WSTATUS(ksi->ksi_status); 284 } else { 285 sigframe.info.lsi_code = LINUX_CLD_EXITED; 286 sigframe.info._sifields._sigchld._status = 287 ((ksi->ksi_status & 0xff00U) >> 8); 288 } 289 break; 290 case LINUX_SIGIO: 291 sigframe.info._sifields._sigpoll._band = ksi->ksi_band; 292 sigframe.info._sifields._sigpoll._fd = ksi->ksi_fd; 293 break; 294 default: 295 sigframe.info._sifields._sigchld._pid = ksi->ksi_pid; 296 sigframe.info._sifields._sigchld._uid = ksi->ksi_uid; 297 if ((sigframe.info.lsi_signo == LINUX_SIGALRM) || 298 (sigframe.info.lsi_signo >= LINUX_SIGRTMIN)) 299 sigframe.info._sifields._timer._sigval.sival_ptr = 300 ksi->ksi_sigval.sival_ptr; 301 break; 302 } 303 304 if ((error = copyout(&sigframe, sp, sizeof(sigframe))) != 0) { 305 sigexit(l, SIGILL); 306 return; 307 } 308 309 linux_buildcontext(l, catcher, sp); 310 tf->tf_rdi = sigframe.info.lsi_signo; 311 tf->tf_rax = 0; 312 tf->tf_rsi = (long)&sfp->info; 313 tf->tf_rdx = (long)&sfp->uc; 314 315 /* 316 * Remember we use signal stack 317 */ 318 if (onstack) 319 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 320 return; 321 } 322 323 int 324 linux_sys_modify_ldt(l, v, retval) 325 struct lwp *l; 326 void *v; 327 register_t *retval; 328 { 329 printf("linux_sys_modify_ldt\n"); 330 return 0; 331 } 332 333 int 334 linux_sys_iopl(l, v, retval) 335 struct lwp *l; 336 void *v; 337 register_t *retval; 338 { 339 return 0; 340 } 341 342 int 343 linux_sys_ioperm(l, v, retval) 344 struct lwp *l; 345 void *v; 346 register_t *retval; 347 { 348 return 0; 349 } 350 351 dev_t 352 linux_fakedev(dev, raw) 353 dev_t dev; 354 int raw; 355 { 356 357 extern const struct cdevsw ptc_cdevsw, pts_cdevsw; 358 const struct cdevsw *cd = cdevsw_lookup(dev); 359 360 if (raw) { 361 #if (NWSDISPLAY > 0) 362 extern const struct cdevsw wsdisplay_cdevsw; 363 if (cd == &wsdisplay_cdevsw) 364 return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1)); 365 #endif 366 } 367 368 if (cd == &ptc_cdevsw) 369 return makedev(LINUX_PTC_MAJOR, minor(dev)); 370 if (cd == &pts_cdevsw) 371 return makedev(LINUX_PTS_MAJOR, minor(dev)); 372 373 return ((minor(dev) & 0xff) | ((major(dev) & 0xfff) << 8) 374 | (((unsigned long long int) (minor(dev) & ~0xff)) << 12) 375 | (((unsigned long long int) (major(dev) & ~0xfff)) << 32)); 376 } 377 378 int 379 linux_machdepioctl(l, v, retval) 380 struct lwp *l; 381 void *v; 382 register_t *retval; 383 { 384 return 0; 385 } 386 387 int 388 linux_sys_rt_sigreturn(l, v, retval) 389 struct lwp *l; 390 void *v; 391 register_t *retval; 392 { 393 struct linux_ucontext *luctx; 394 struct trapframe *tf = l->l_md.md_regs; 395 struct linux_sigcontext *lsigctx; 396 struct linux__fpstate fpstate; 397 struct linux_rt_sigframe frame, *fp; 398 ucontext_t uctx; 399 mcontext_t *mctx; 400 struct fxsave64 *fxsave; 401 int error; 402 403 fp = (struct linux_rt_sigframe *)(tf->tf_rsp - 8); 404 if ((error = copyin(fp, &frame, sizeof(frame))) != 0) { 405 sigexit(l, SIGILL); 406 return error; 407 } 408 luctx = &frame.uc; 409 lsigctx = &luctx->luc_mcontext; 410 411 bzero(&uctx, sizeof(uctx)); 412 mctx = (mcontext_t *)&uctx.uc_mcontext; 413 fxsave = (struct fxsave64 *)&mctx->__fpregs; 414 415 /* 416 * Set the flags. Linux always have CPU, stack and signal state, 417 * FPU is optional. uc_flags is not used to tell what we have. 418 */ 419 uctx.uc_flags = (_UC_SIGMASK|_UC_CPU|_UC_STACK|_UC_CLRSTACK); 420 if (lsigctx->fpstate != NULL) 421 uctx.uc_flags |= _UC_FPU; 422 uctx.uc_link = NULL; 423 424 /* 425 * Signal set 426 */ 427 linux_to_native_sigset(&uctx.uc_sigmask, &luctx->luc_sigmask); 428 429 /* 430 * CPU state 431 */ 432 mctx->__gregs[_REG_R8] = lsigctx->r8; 433 mctx->__gregs[_REG_R9] = lsigctx->r9; 434 mctx->__gregs[_REG_R10] = lsigctx->r10; 435 mctx->__gregs[_REG_R11] = lsigctx->r11; 436 mctx->__gregs[_REG_R12] = lsigctx->r12; 437 mctx->__gregs[_REG_R13] = lsigctx->r13; 438 mctx->__gregs[_REG_R14] = lsigctx->r14; 439 mctx->__gregs[_REG_R15] = lsigctx->r15; 440 mctx->__gregs[_REG_RDI] = lsigctx->rdi; 441 mctx->__gregs[_REG_RSI] = lsigctx->rsi; 442 mctx->__gregs[_REG_RBP] = lsigctx->rbp; 443 mctx->__gregs[_REG_RBX] = lsigctx->rbx; 444 mctx->__gregs[_REG_RAX] = lsigctx->rax; 445 mctx->__gregs[_REG_RDX] = lsigctx->rdx; 446 mctx->__gregs[_REG_RCX] = lsigctx->rcx; 447 mctx->__gregs[_REG_RIP] = lsigctx->rip; 448 mctx->__gregs[_REG_RFL] = lsigctx->eflags; 449 mctx->__gregs[_REG_CS] = lsigctx->cs; 450 mctx->__gregs[_REG_GS] = lsigctx->gs; 451 mctx->__gregs[_REG_FS] = lsigctx->fs; 452 mctx->__gregs[_REG_ERR] = lsigctx->err; 453 mctx->__gregs[_REG_TRAPNO] = lsigctx->trapno; 454 mctx->__gregs[_REG_ES] = tf->tf_es; 455 mctx->__gregs[_REG_DS] = tf->tf_ds; 456 mctx->__gregs[_REG_URSP] = lsigctx->rsp; /* XXX */ 457 mctx->__gregs[_REG_SS] = tf->tf_ss; 458 459 /* 460 * FPU state 461 */ 462 if (lsigctx->fpstate != NULL) { 463 error = copyin(lsigctx->fpstate, &fpstate, sizeof(fpstate)); 464 if (error != 0) { 465 sigexit(l, SIGILL); 466 return error; 467 } 468 469 fxsave->fx_fcw = fpstate.cwd; 470 fxsave->fx_fsw = fpstate.swd; 471 fxsave->fx_ftw = fpstate.twd; 472 fxsave->fx_fop = fpstate.fop; 473 fxsave->fx_rip = fpstate.rip; 474 fxsave->fx_rdp = fpstate.rdp; 475 fxsave->fx_mxcsr = fpstate.mxcsr; 476 fxsave->fx_mxcsr_mask = fpstate.mxcsr_mask; 477 memcpy(&fxsave->fx_st, &fpstate.st_space, 478 sizeof(fxsave->fx_st)); 479 memcpy(&fxsave->fx_xmm, &fpstate.xmm_space, 480 sizeof(fxsave->fx_xmm)); 481 } 482 483 /* 484 * And the stack 485 */ 486 uctx.uc_stack.ss_flags = 0; 487 if (luctx->luc_stack.ss_flags & LINUX_SS_ONSTACK); 488 uctx.uc_stack.ss_flags = SS_ONSTACK; 489 490 if (luctx->luc_stack.ss_flags & LINUX_SS_DISABLE); 491 uctx.uc_stack.ss_flags = SS_DISABLE; 492 493 uctx.uc_stack.ss_sp = luctx->luc_stack.ss_sp; 494 uctx.uc_stack.ss_size = luctx->luc_stack.ss_size; 495 496 /* 497 * And let setucontext deal with that. 498 */ 499 error = setucontext(l, &uctx); 500 if (error) 501 return error; 502 503 return EJUSTRETURN; 504 } 505 506 int 507 linux_sys_arch_prctl(l, v, retval) 508 struct lwp *l; 509 void *v; 510 register_t *retval; 511 { 512 struct linux_sys_arch_prctl_args /* { 513 syscallarg(int) code; 514 syscallarg(unsigned long) addr; 515 } */ *uap = v; 516 struct pcb *pcb = &l->l_addr->u_pcb; 517 struct trapframe *tf = l->l_md.md_regs; 518 int error; 519 uint64_t taddr; 520 521 switch(SCARG(uap, code)) { 522 case LINUX_ARCH_SET_GS: 523 taddr = SCARG(uap, addr); 524 if (taddr >= VM_MAXUSER_ADDRESS) 525 return EINVAL; 526 pcb->pcb_gs = taddr; 527 pcb->pcb_flags |= PCB_GS64; 528 if (l == curlwp) 529 wrmsr(MSR_KERNELGSBASE, taddr); 530 break; 531 532 case LINUX_ARCH_GET_GS: 533 if (pcb->pcb_flags & PCB_GS64) 534 taddr = pcb->pcb_gs; 535 else { 536 error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr); 537 if (error != 0) 538 return error; 539 } 540 error = copyout(&taddr, (char *)SCARG(uap, addr), 8); 541 if (error != 0) 542 return error; 543 break; 544 545 case LINUX_ARCH_SET_FS: 546 taddr = SCARG(uap, addr); 547 if (taddr >= VM_MAXUSER_ADDRESS) 548 return EINVAL; 549 pcb->pcb_fs = taddr; 550 pcb->pcb_flags |= PCB_FS64; 551 if (l == curlwp) 552 wrmsr(MSR_FSBASE, taddr); 553 break; 554 555 case LINUX_ARCH_GET_FS: 556 if (pcb->pcb_flags & PCB_FS64) 557 taddr = pcb->pcb_fs; 558 else { 559 error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr); 560 if (error != 0) 561 return error; 562 } 563 error = copyout(&taddr, (char *)SCARG(uap, addr), 8); 564 if (error != 0) 565 return error; 566 break; 567 568 default: 569 #ifdef DEBUG_LINUX 570 printf("linux_sys_arch_prctl: unexpected code %d\n", 571 SCARG(uap, code)); 572 #endif 573 return EINVAL; 574 } 575 576 return 0; 577 } 578 579 const int linux_vsyscall_to_syscall[] = { 580 LINUX_SYS_gettimeofday, 581 LINUX_SYS_time, 582 LINUX_SYS_nosys, 583 LINUX_SYS_nosys, 584 }; 585 586 int 587 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg) 588 { 589 struct trapframe *tf = arg; 590 uint64_t retaddr; 591 int vsyscallnr; 592 593 /* 594 * Check for a vsyscall. %rip must be the fault address, 595 * and the address must be in the Linux vsyscall area. 596 * Also, vsyscalls are only done at 1024-byte boundaries. 597 */ 598 599 if (__predict_true(trapaddr < LINUX_VSYSCALL_START)) 600 return 0; 601 602 if (trapaddr != tf->tf_rip) 603 return 0; 604 605 if ((tf->tf_rip & (LINUX_VSYSCALL_SIZE - 1)) != 0) 606 return 0; 607 608 vsyscallnr = (tf->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SIZE; 609 610 if (vsyscallnr > LINUX_VSYSCALL_MAXNR) 611 return 0; 612 613 /* 614 * Get the return address from the top of the stack, 615 * and fix up the return address. 616 * This assumes the faulting instruction was callq *reg, 617 * which is the only way that vsyscalls are ever entered. 618 */ 619 if (copyin((void *)tf->tf_rsp, &retaddr, sizeof retaddr) != 0) 620 return 0; 621 tf->tf_rip = retaddr; 622 tf->tf_rax = linux_vsyscall_to_syscall[vsyscallnr]; 623 tf->tf_rsp += 8; /* "pop" the return address */ 624 625 #if 0 626 printf("usertrap: rip %p rsp %p retaddr %p vsys %d sys %d\n", 627 (void *)tf->tf_rip, (void *)tf->tf_rsp, (void *)retaddr, 628 vsyscallnr, (int)tf->tf_rax); 629 #endif 630 631 (*l->l_proc->p_md.md_syscall)(tf); 632 633 return 1; 634 } 635 636 static void 637 linux_buildcontext(struct lwp *l, void *catcher, void *f) 638 { 639 struct trapframe *tf = l->l_md.md_regs; 640 641 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 642 tf->tf_rip = (u_int64_t)catcher; 643 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 644 tf->tf_rflags &= ~(PSL_T|PSL_VM|PSL_AC); 645 tf->tf_rsp = (u_int64_t)f; 646 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 647 } 648 649 void * 650 linux_get_newtls(l) 651 struct lwp *l; 652 { 653 struct trapframe *tf = l->l_md.md_regs; 654 655 return (void *)tf->tf_r8; 656 } 657 658 int 659 linux_set_newtls(l, tls) 660 struct lwp *l; 661 void *tls; 662 { 663 struct linux_sys_arch_prctl_args cup; 664 register_t retval; 665 666 SCARG(&cup, code) = LINUX_ARCH_SET_FS; 667 SCARG(&cup, addr) = (unsigned long)tls; 668 669 return linux_sys_arch_prctl(l, &cup, &retval); 670 } 671
Indexes created Tue Sep 30 20:09:53 GMT 2025