linux_machdep.c revision 1.14.2.2
1 /* $NetBSD: linux_machdep.c,v 1.14.2.2 2007/02/06 18:59:08 ad 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 1.14.2.2 2007/02/06 18:59:08 ad 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 49 #include <machine/reg.h> 50 #include <machine/pcb.h> 51 #include <machine/fpu.h> 52 #include <machine/mcontext.h> 53 #include <machine/specialreg.h> 54 #include <machine/vmparam.h> 55 56 #include <compat/linux/common/linux_signal.h> 57 #include <compat/linux/common/linux_errno.h> 58 #include <compat/linux/common/linux_exec.h> 59 #include <compat/linux/common/linux_ioctl.h> 60 #include <compat/linux/common/linux_prctl.h> 61 #include <compat/linux/common/linux_machdep.h> 62 #include <compat/linux/linux_syscall.h> 63 #include <compat/linux/linux_syscallargs.h> 64 65 static void linux_buildcontext(struct lwp *, void *, void *); 66 67 void 68 linux_setregs(l, epp, stack) 69 struct lwp *l; 70 struct exec_package *epp; 71 u_long stack; 72 { 73 struct pcb *pcb = &l->l_addr->u_pcb; 74 struct trapframe *tf; 75 76 /* If we were using the FPU, forget about it. */ 77 if (l->l_addr->u_pcb.pcb_fpcpu != NULL) 78 fpusave_lwp(l, 0); 79 80 l->l_md.md_flags &= ~MDP_USEDFPU; 81 pcb->pcb_flags = 0; 82 pcb->pcb_savefpu.fp_fxsave.fx_fcw = __NetBSD_NPXCW__; 83 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr = __INITIAL_MXCSR__; 84 pcb->pcb_savefpu.fp_fxsave.fx_mxcsr_mask = __INITIAL_MXCSR_MASK__; 85 pcb->pcb_fs = 0; 86 pcb->pcb_gs = 0; 87 88 l->l_proc->p_flag &= ~P_32; 89 90 tf = l->l_md.md_regs; 91 tf->tf_rax = 0; 92 tf->tf_rbx = 0; 93 tf->tf_rcx = epp->ep_entry; 94 tf->tf_rdx = 0; 95 tf->tf_rsi = 0; 96 tf->tf_rdi = 0; 97 tf->tf_rbp = 0; 98 tf->tf_rsp = stack; 99 tf->tf_r8 = 0; 100 tf->tf_r9 = 0; 101 tf->tf_r10 = 0; 102 tf->tf_r11 = 0; 103 tf->tf_r12 = 0; 104 tf->tf_r13 = 0; 105 tf->tf_r14 = 0; 106 tf->tf_r15 = 0; 107 tf->tf_rip = epp->ep_entry; 108 tf->tf_rflags = PSL_USERSET; 109 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 110 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 111 tf->tf_ds = 0; 112 tf->tf_es = 0; 113 tf->tf_fs = 0; 114 tf->tf_gs = 0; 115 116 return; 117 } 118 119 void 120 linux_sendsig(ksi, mask) 121 const ksiginfo_t *ksi; 122 const sigset_t *mask; 123 { 124 struct lwp *l = curlwp; 125 struct proc *p = l->l_proc; 126 struct sigacts *ps = p->p_sigacts; 127 int onstack, error; 128 int sig = ksi->ksi_signo; 129 struct linux_rt_sigframe *sfp, sigframe; 130 struct linux__fpstate *fpsp, fpstate; 131 struct fpreg fpregs; 132 struct trapframe *tf = l->l_md.md_regs; 133 sig_t catcher = SIGACTION(p, sig).sa_handler; 134 linux_sigset_t lmask; 135 char *sp; 136 int error; 137 sigset_t tmask; 138 139 tmask = *mask; 140 141 /* Do we need to jump onto the signal stack? */ 142 onstack = 143 (p->p_sigctx.ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 144 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 145 146 /* Allocate space for the signal handler context. */ 147 if (onstack) 148 sp = ((caddr_t)p->p_sigctx.ps_sigstk.ss_sp + 149 p->p_sigctx.ps_sigstk.ss_size); 150 else 151 sp = (caddr_t)tf->tf_rsp - 128; 152 153 154 /* 155 * Save FPU state, if any 156 */ 157 if (l->l_md.md_flags & MDP_USEDFPU) { 158 sp = (char *) 159 (((long)sp - sizeof(struct linux__fpstate)) & ~0xfUL); 160 fpsp = (struct linux__fpstate *)sp; 161 162 mutex_exit(&p->p_smutex); 163 (void)process_read_fpregs(l, &fpregs); 164 bzero(&fpstate, sizeof(fpstate)); 165 fpstate.cwd = fpregs.fp_fcw; 166 fpstate.swd = fpregs.fp_fsw; 167 fpstate.twd = fpregs.fp_ftw; 168 fpstate.fop = fpregs.fp_fop; 169 fpstate.rip = fpregs.fp_rip; 170 fpstate.rdp = fpregs.fp_rdp; 171 fpstate.mxcsr = fpregs.fp_mxcsr; 172 fpstate.mxcsr_mask = fpregs.fp_mxcsr_mask; 173 memcpy(&fpstate.st_space, &fpregs.fp_st, 174 sizeof(fpstate.st_space)); 175 memcpy(&fpstate.xmm_space, &fpregs.fp_xmm, 176 sizeof(fpstate.xmm_space)); 177 error = copyout(&fpstate, fpsp, sizeof(fpstate)); 178 mutex_enter(&p->p_smutex) 179 180 if (error != 0) { 181 sigexit(l, SIGILL); 182 return; 183 } 184 } else { 185 fpsp = NULL; 186 } 187 188 /* 189 * Populate the rt_sigframe 190 */ 191 sp = (char *) 192 ((((long)sp - sizeof(struct linux_rt_sigframe)) & ~0xfUL) - 8); 193 sfp = (struct linux_rt_sigframe *)sp; 194 195 bzero(&sigframe, sizeof(sigframe)); 196 if (ps->sa_sigdesc[sig].sd_vers != 0) 197 sigframe.pretcode = 198 (char *)(u_long)ps->sa_sigdesc[sig].sd_tramp; 199 else 200 sigframe.pretcode = NULL; 201 202 /* 203 * The user context 204 */ 205 sigframe.uc.luc_flags = 0; 206 sigframe.uc.luc_link = NULL; 207 208 /* This is used regardless of SA_ONSTACK in Linux */ 209 sigframe.uc.luc_stack.ss_sp = p->p_sigctx.ps_sigstk.ss_sp; 210 sigframe.uc.luc_stack.ss_size = p->p_sigctx.ps_sigstk.ss_size; 211 sigframe.uc.luc_stack.ss_flags = 0; 212 if (p->p_sigctx.ps_sigstk.ss_flags & SS_ONSTACK) 213 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_ONSTACK; 214 if (p->p_sigctx.ps_sigstk.ss_flags & SS_DISABLE) 215 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_DISABLE; 216 217 sigframe.uc.luc_mcontext.r8 = tf->tf_r8; 218 sigframe.uc.luc_mcontext.r9 = tf->tf_r9; 219 sigframe.uc.luc_mcontext.r10 = tf->tf_r10; 220 sigframe.uc.luc_mcontext.r11 = tf->tf_r11; 221 sigframe.uc.luc_mcontext.r12 = tf->tf_r12; 222 sigframe.uc.luc_mcontext.r13 = tf->tf_r13; 223 sigframe.uc.luc_mcontext.r14 = tf->tf_r14; 224 sigframe.uc.luc_mcontext.r15 = tf->tf_r15; 225 sigframe.uc.luc_mcontext.rdi = tf->tf_rdi; 226 sigframe.uc.luc_mcontext.rsi = tf->tf_rsi; 227 sigframe.uc.luc_mcontext.rbp = tf->tf_rbp; 228 sigframe.uc.luc_mcontext.rbx = tf->tf_rbx; 229 sigframe.uc.luc_mcontext.rdx = tf->tf_rdx; 230 sigframe.uc.luc_mcontext.rcx = tf->tf_rcx; 231 sigframe.uc.luc_mcontext.rsp = tf->tf_rsp; 232 sigframe.uc.luc_mcontext.rip = tf->tf_rip; 233 sigframe.uc.luc_mcontext.eflags = tf->tf_rflags; 234 sigframe.uc.luc_mcontext.cs = tf->tf_cs; 235 sigframe.uc.luc_mcontext.gs = tf->tf_gs; 236 sigframe.uc.luc_mcontext.fs = tf->tf_fs; 237 sigframe.uc.luc_mcontext.err = tf->tf_err; 238 sigframe.uc.luc_mcontext.trapno = tf->tf_trapno; 239 native_to_linux_sigset(&lmask, mask); 240 sigframe.uc.luc_mcontext.oldmask = lmask.sig[0]; 241 sigframe.uc.luc_mcontext.cr2 = (long)l->l_addr->u_pcb.pcb_onfault; 242 sigframe.uc.luc_mcontext.fpstate = fpsp; 243 native_to_linux_sigset(&sigframe.uc.luc_sigmask, mask); 244 245 /* 246 * the siginfo structure 247 */ 248 sigframe.info.lsi_signo = native_to_linux_signo[sig]; 249 sigframe.info.lsi_errno = native_to_linux_errno[ksi->ksi_errno]; 250 sigframe.info.lsi_code = ksi->ksi_code; 251 252 /* XXX This is a rought conversion, taken from i386 code */ 253 switch (sigframe.info.lsi_signo) { 254 case LINUX_SIGILL: 255 case LINUX_SIGFPE: 256 case LINUX_SIGSEGV: 257 case LINUX_SIGBUS: 258 case LINUX_SIGTRAP: 259 sigframe.info._sifields._sigfault._addr = ksi->ksi_addr; 260 break; 261 case LINUX_SIGCHLD: 262 sigframe.info._sifields._sigchld._pid = ksi->ksi_pid; 263 sigframe.info._sifields._sigchld._uid = ksi->ksi_uid; 264 sigframe.info._sifields._sigchld._utime = ksi->ksi_utime; 265 sigframe.info._sifields._sigchld._stime = ksi->ksi_stime; 266 267 if (WCOREDUMP(ksi->ksi_status)) { 268 sigframe.info.lsi_code = LINUX_CLD_DUMPED; 269 sigframe.info._sifields._sigchld._status = 270 _WSTATUS(ksi->ksi_status); 271 } else if (_WSTATUS(ksi->ksi_status)) { 272 sigframe.info.lsi_code = LINUX_CLD_KILLED; 273 sigframe.info._sifields._sigchld._status = 274 _WSTATUS(ksi->ksi_status); 275 } else { 276 sigframe.info.lsi_code = LINUX_CLD_EXITED; 277 sigframe.info._sifields._sigchld._status = 278 ((ksi->ksi_status & 0xff00U) >> 8); 279 } 280 break; 281 case LINUX_SIGIO: 282 sigframe.info._sifields._sigpoll._band = ksi->ksi_band; 283 sigframe.info._sifields._sigpoll._fd = ksi->ksi_fd; 284 break; 285 default: 286 sigframe.info._sifields._sigchld._pid = ksi->ksi_pid; 287 sigframe.info._sifields._sigchld._uid = ksi->ksi_uid; 288 if ((sigframe.info.lsi_signo == LINUX_SIGALRM) || 289 (sigframe.info.lsi_signo >= LINUX_SIGRTMIN)) 290 sigframe.info._sifields._timer._sigval.sival_ptr = 291 ksi->ksi_sigval.sival_ptr; 292 break; 293 } 294 295 sendsig_reset(l, sig); 296 mutex_exit(&p->p_smutex); 297 error = copyout(&sigframe, sp, sizeof(sigframe)); 298 mutex_enter(&p->p_smutex); 299 300 if (error != 0) { 301 sigexit(l, SIGILL); 302 return; 303 } 304 305 linux_buildcontext(l, catcher, sp); 306 tf->tf_rdi = sigframe.info.lsi_signo; 307 tf->tf_rax = 0; 308 tf->tf_rsi = (long)&sfp->info; 309 tf->tf_rdx = (long)&sfp->uc; 310 311 /* 312 * Remember we use signal stack 313 */ 314 if (onstack) 315 p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK; 316 return; 317 } 318 319 int 320 linux_sys_modify_ldt(l, v, retval) 321 struct lwp *l; 322 void *v; 323 register_t *retval; 324 { 325 printf("linux_sys_modify_ldt\n"); 326 return 0; 327 } 328 329 int 330 linux_sys_iopl(l, v, retval) 331 struct lwp *l; 332 void *v; 333 register_t *retval; 334 { 335 return 0; 336 } 337 338 int 339 linux_sys_ioperm(l, v, retval) 340 struct lwp *l; 341 void *v; 342 register_t *retval; 343 { 344 return 0; 345 } 346 347 dev_t 348 linux_fakedev(dev, raw) 349 dev_t dev; 350 int raw; 351 { 352 return ((minor(dev) & 0xff) | ((major(dev) & 0xfff) << 8) 353 | (((unsigned long long int) (minor(dev) & ~0xff)) << 12) 354 | (((unsigned long long int) (major(dev) & ~0xfff)) << 32)); 355 } 356 357 int 358 linux_machdepioctl(l, v, retval) 359 struct lwp *l; 360 void *v; 361 register_t *retval; 362 { 363 return 0; 364 } 365 366 int 367 linux_sys_rt_sigreturn(l, v, retval) 368 struct lwp *l; 369 void *v; 370 register_t *retval; 371 { 372 struct linux_ucontext *luctx; 373 struct trapframe *tf = l->l_md.md_regs; 374 struct linux_sigcontext *lsigctx; 375 struct linux__fpstate fpstate; 376 struct linux_rt_sigframe frame, *fp; 377 ucontext_t uctx; 378 mcontext_t *mctx; 379 struct fxsave64 *fxsave; 380 int error; 381 382 fp = (struct linux_rt_sigframe *)(tf->tf_rsp - 8); 383 if ((error = copyin(fp, &frame, sizeof(frame))) != 0) { 384 mutex_enter(&l->l_proc->p_smutex); 385 sigexit(l, SIGILL); 386 return error; 387 } 388 luctx = &frame.uc; 389 lsigctx = &luctx->luc_mcontext; 390 391 bzero(&uctx, sizeof(uctx)); 392 mctx = (mcontext_t *)&uctx.uc_mcontext; 393 fxsave = (struct fxsave64 *)&mctx->__fpregs; 394 395 /* 396 * Set the flags. Linux always have CPU, stack and signal state, 397 * FPU is optional. uc_flags is not used to tell what we have. 398 */ 399 uctx.uc_flags = (_UC_SIGMASK|_UC_CPU|_UC_STACK|_UC_CLRSTACK); 400 if (lsigctx->fpstate != NULL) 401 uctx.uc_flags |= _UC_FPU; 402 uctx.uc_link = NULL; 403 404 /* 405 * Signal set 406 */ 407 linux_to_native_sigset(&uctx.uc_sigmask, &luctx->luc_sigmask); 408 409 /* 410 * CPU state 411 */ 412 mctx->__gregs[_REG_R8] = lsigctx->r8; 413 mctx->__gregs[_REG_R9] = lsigctx->r9; 414 mctx->__gregs[_REG_R10] = lsigctx->r10; 415 mctx->__gregs[_REG_R11] = lsigctx->r11; 416 mctx->__gregs[_REG_R12] = lsigctx->r12; 417 mctx->__gregs[_REG_R13] = lsigctx->r13; 418 mctx->__gregs[_REG_R14] = lsigctx->r14; 419 mctx->__gregs[_REG_R15] = lsigctx->r15; 420 mctx->__gregs[_REG_RDI] = lsigctx->rdi; 421 mctx->__gregs[_REG_RSI] = lsigctx->rsi; 422 mctx->__gregs[_REG_RBP] = lsigctx->rbp; 423 mctx->__gregs[_REG_RBX] = lsigctx->rbx; 424 mctx->__gregs[_REG_RAX] = tf->tf_rax; 425 mctx->__gregs[_REG_RDX] = lsigctx->rdx; 426 mctx->__gregs[_REG_RCX] = lsigctx->rcx; 427 mctx->__gregs[_REG_RIP] = lsigctx->rip; 428 mctx->__gregs[_REG_RFL] = lsigctx->eflags; 429 mctx->__gregs[_REG_CS] = lsigctx->cs; 430 mctx->__gregs[_REG_GS] = lsigctx->gs; 431 mctx->__gregs[_REG_FS] = lsigctx->fs; 432 mctx->__gregs[_REG_ERR] = lsigctx->err; 433 mctx->__gregs[_REG_TRAPNO] = lsigctx->trapno; 434 mctx->__gregs[_REG_ES] = tf->tf_es; 435 mctx->__gregs[_REG_DS] = tf->tf_ds; 436 mctx->__gregs[_REG_URSP] = lsigctx->rsp; /* XXX */ 437 mctx->__gregs[_REG_SS] = tf->tf_ss; 438 439 /* 440 * FPU state 441 */ 442 if (lsigctx->fpstate != NULL) { 443 error = copyin(lsigctx->fpstate, &fpstate, sizeof(fpstate)); 444 if (error != 0) { 445 mutex_enter(&l->l_proc->p_smutex); 446 sigexit(l, SIGILL); 447 return error; 448 } 449 450 fxsave->fx_fcw = fpstate.cwd; 451 fxsave->fx_fsw = fpstate.swd; 452 fxsave->fx_ftw = fpstate.twd; 453 fxsave->fx_fop = fpstate.fop; 454 fxsave->fx_rip = fpstate.rip; 455 fxsave->fx_rdp = fpstate.rdp; 456 fxsave->fx_mxcsr = fpstate.mxcsr; 457 fxsave->fx_mxcsr_mask = fpstate.mxcsr_mask; 458 memcpy(&fxsave->fx_st, &fpstate.st_space, 459 sizeof(fxsave->fx_st)); 460 memcpy(&fxsave->fx_xmm, &fpstate.xmm_space, 461 sizeof(fxsave->fx_xmm)); 462 } 463 464 /* 465 * And the stack 466 */ 467 uctx.uc_stack.ss_flags = 0; 468 if (luctx->luc_stack.ss_flags & LINUX_SS_ONSTACK); 469 uctx.uc_stack.ss_flags = SS_ONSTACK; 470 471 if (luctx->luc_stack.ss_flags & LINUX_SS_DISABLE); 472 uctx.uc_stack.ss_flags = SS_DISABLE; 473 474 uctx.uc_stack.ss_sp = luctx->luc_stack.ss_sp; 475 uctx.uc_stack.ss_size = luctx->luc_stack.ss_size; 476 477 /* 478 * And let setucontext deal with that. 479 */ 480 mutex_enter(&l->l_proc->p_smutex); 481 error = setucontext(l, &uctx); 482 mutex_exit(&l->l_proc->p_smutex); 483 484 return error; 485 } 486 487 int 488 linux_sys_arch_prctl(l, v, retval) 489 struct lwp *l; 490 void *v; 491 register_t *retval; 492 { 493 struct linux_sys_arch_prctl_args /* { 494 syscallarg(int) code; 495 syscallarg(unsigned long) addr; 496 } */ *uap = v; 497 struct pcb *pcb = &l->l_addr->u_pcb; 498 struct trapframe *tf = l->l_md.md_regs; 499 int error; 500 uint64_t taddr; 501 502 switch(SCARG(uap, code)) { 503 case LINUX_ARCH_SET_GS: 504 taddr = SCARG(uap, addr); 505 if (taddr >= VM_MAXUSER_ADDRESS) 506 return EINVAL; 507 pcb->pcb_gs = taddr; 508 pcb->pcb_flags |= PCB_GS64; 509 if (l == curlwp) 510 wrmsr(MSR_KERNELGSBASE, taddr); 511 break; 512 513 case LINUX_ARCH_GET_GS: 514 if (pcb->pcb_flags & PCB_GS64) 515 taddr = pcb->pcb_gs; 516 else { 517 error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr); 518 if (error != 0) 519 return error; 520 } 521 error = copyout(&taddr, (char *)SCARG(uap, addr), 8); 522 if (error != 0) 523 return error; 524 break; 525 526 case LINUX_ARCH_SET_FS: 527 taddr = SCARG(uap, addr); 528 if (taddr >= VM_MAXUSER_ADDRESS) 529 return EINVAL; 530 pcb->pcb_fs = taddr; 531 pcb->pcb_flags |= PCB_FS64; 532 if (l == curlwp) 533 wrmsr(MSR_FSBASE, taddr); 534 break; 535 536 case LINUX_ARCH_GET_FS: 537 if (pcb->pcb_flags & PCB_FS64) 538 taddr = pcb->pcb_fs; 539 else { 540 error = memseg_baseaddr(l, tf->tf_fs, NULL, 0, &taddr); 541 if (error != 0) 542 return error; 543 } 544 error = copyout(&taddr, (char *)SCARG(uap, addr), 8); 545 if (error != 0) 546 return error; 547 break; 548 549 default: 550 #ifdef DEBUG_LINUX 551 printf("linux_sys_arch_prctl: unexpected code %d\n", 552 SCARG(uap, code)); 553 #endif 554 return EINVAL; 555 } 556 557 return 0; 558 } 559 560 const int linux_vsyscall_to_syscall[] = { 561 LINUX_SYS_gettimeofday, 562 LINUX_SYS_time, 563 LINUX_SYS_nosys, 564 LINUX_SYS_nosys, 565 }; 566 567 int 568 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg) 569 { 570 struct trapframe *tf = arg; 571 uint64_t retaddr; 572 int vsyscallnr; 573 574 /* 575 * Check for a vsyscall. %rip must be the fault address, 576 * and the address must be in the Linux vsyscall area. 577 * Also, vsyscalls are only done at 1024-byte boundaries. 578 */ 579 580 if (__predict_true(trapaddr < LINUX_VSYSCALL_START)) 581 return 0; 582 583 if (trapaddr != tf->tf_rip) 584 return 0; 585 586 if ((tf->tf_rip & (LINUX_VSYSCALL_SIZE - 1)) != 0) 587 return 0; 588 589 vsyscallnr = (tf->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SIZE; 590 591 if (vsyscallnr > LINUX_VSYSCALL_MAXNR) 592 return 0; 593 594 /* 595 * Get the return address from the top of the stack, 596 * and fix up the return address. 597 * This assumes the faulting instruction was callq *reg, 598 * which is the only way that vsyscalls are ever entered. 599 */ 600 if (copyin((void *)tf->tf_rsp, &retaddr, sizeof retaddr) != 0) 601 return 0; 602 tf->tf_rip = retaddr; 603 tf->tf_rax = linux_vsyscall_to_syscall[vsyscallnr]; 604 tf->tf_rsp += 8; /* "pop" the return address */ 605 606 #if 0 607 printf("usertrap: rip %p rsp %p retaddr %p vsys %d sys %d\n", 608 (void *)tf->tf_rip, (void *)tf->tf_rsp, (void *)retaddr, 609 vsyscallnr, (int)tf->tf_rax); 610 #endif 611 612 (*l->l_proc->p_md.md_syscall)(tf); 613 614 return 1; 615 } 616 617 static void 618 linux_buildcontext(struct lwp *l, void *catcher, void *f) 619 { 620 struct trapframe *tf = l->l_md.md_regs; 621 622 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 623 tf->tf_rip = (u_int64_t)catcher; 624 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 625 tf->tf_rflags &= ~(PSL_T|PSL_VM|PSL_AC); 626 tf->tf_rsp = (u_int64_t)f; 627 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 628 } 629 630 void * 631 linux_get_newtls(l) 632 struct lwp *l; 633 { 634 struct trapframe *tf = l->l_md.md_regs; 635 636 return (void *)tf->tf_r8; 637 } 638 639 int 640 linux_set_newtls(l, tls) 641 struct lwp *l; 642 void *tls; 643 { 644 struct linux_sys_arch_prctl_args cup; 645 register_t retval; 646 647 SCARG(&cup, code) = LINUX_ARCH_SET_FS; 648 SCARG(&cup, addr) = (unsigned long)tls; 649 650 return linux_sys_arch_prctl(l, &cup, &retval); 651 } 652
Indexes created Tue Sep 30 20:09:53 GMT 2025