linux_machdep.c revision 1.46
1 /* $NetBSD: linux_machdep.c,v 1.46 2014/02/11 20:17:16 dsl 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 1.46 2014/02/11 20:17:16 dsl 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/ucontext.h> 46 #include <sys/conf.h> 47 48 #include <machine/reg.h> 49 #include <machine/pcb.h> 50 #include <machine/mcontext.h> 51 #include <machine/specialreg.h> 52 #include <machine/vmparam.h> 53 #include <machine/cpufunc.h> 54 55 /* 56 * To see whether wscons is configured (for virtual console ioctl calls). 57 */ 58 #if defined(_KERNEL_OPT) 59 #include "wsdisplay.h" 60 #endif 61 #if (NWSDISPLAY > 0) 62 #include <dev/wscons/wsconsio.h> 63 #include <dev/wscons/wsdisplay_usl_io.h> 64 #endif 65 66 67 #include <compat/linux/common/linux_signal.h> 68 #include <compat/linux/common/linux_errno.h> 69 #include <compat/linux/common/linux_exec.h> 70 #include <compat/linux/common/linux_ioctl.h> 71 #include <compat/linux/common/linux_prctl.h> 72 #include <compat/linux/common/linux_machdep.h> 73 #include <compat/linux/common/linux_ipc.h> 74 #include <compat/linux/common/linux_sem.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(struct lwp *l, struct exec_package *epp, vaddr_t stack) 82 { 83 struct pcb *pcb = lwp_getpcb(l); 84 struct trapframe *tf; 85 86 /* If we were using the FPU, forget about it. */ 87 if (pcb->pcb_fpcpu != NULL) 88 fpusave_lwp(l, 0); 89 90 l->l_md.md_flags &= ~MDL_USEDFPU; 91 pcb->pcb_flags = 0; 92 pcb->pcb_savefpu.sv_xmm.fx_cw = __NetBSD_NPXCW__; 93 pcb->pcb_savefpu.sv_xmm.fx_mxcsr = __INITIAL_MXCSR__; 94 pcb->pcb_savefpu.sv_xmm.fx_mxcsr_mask = __INITIAL_MXCSR_MASK__; 95 96 l->l_proc->p_flag &= ~PK_32; 97 98 tf = l->l_md.md_regs; 99 tf->tf_rax = 0; 100 tf->tf_rbx = 0; 101 tf->tf_rcx = epp->ep_entry; 102 tf->tf_rdx = 0; 103 tf->tf_rsi = 0; 104 tf->tf_rdi = 0; 105 tf->tf_rbp = 0; 106 tf->tf_rsp = stack; 107 tf->tf_r8 = 0; 108 tf->tf_r9 = 0; 109 tf->tf_r10 = 0; 110 tf->tf_r11 = 0; 111 tf->tf_r12 = 0; 112 tf->tf_r13 = 0; 113 tf->tf_r14 = 0; 114 tf->tf_r15 = 0; 115 tf->tf_rip = epp->ep_entry; 116 tf->tf_rflags = PSL_USERSET; 117 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 118 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 119 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 120 tf->tf_es = 0; 121 cpu_fsgs_zero(l); 122 123 return; 124 } 125 126 void 127 linux_sendsig(const ksiginfo_t *ksi, const sigset_t *mask) 128 { 129 struct lwp *l = curlwp; 130 struct proc *p = l->l_proc; 131 struct pcb *pcb = lwp_getpcb(l); 132 struct sigacts *ps = p->p_sigacts; 133 int onstack, error; 134 int sig = ksi->ksi_signo; 135 struct linux_rt_sigframe *sfp, sigframe; 136 struct linux__fpstate *fpsp; 137 struct fpreg fpregs; 138 struct trapframe *tf = l->l_md.md_regs; 139 sig_t catcher = SIGACTION(p, sig).sa_handler; 140 linux_sigset_t lmask; 141 char *sp; 142 143 /* Do we need to jump onto the signal stack? */ 144 onstack = 145 (l->l_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0 && 146 (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0; 147 148 /* Allocate space for the signal handler context. */ 149 if (onstack) 150 sp = ((char *)l->l_sigstk.ss_sp + 151 l->l_sigstk.ss_size); 152 else 153 sp = (char *)tf->tf_rsp - 128; 154 155 /* 156 * Save FPU state, if any 157 */ 158 if (l->l_md.md_flags & MDL_USEDFPU) { 159 sp = (char *) 160 (((long)sp - sizeof (*fpsp)) & ~0xfUL); 161 fpsp = (struct linux__fpstate *)sp; 162 } else 163 fpsp = NULL; 164 165 /* 166 * Populate the rt_sigframe 167 */ 168 sp = (char *) 169 ((((long)sp - sizeof(struct linux_rt_sigframe)) & ~0xfUL) - 8); 170 sfp = (struct linux_rt_sigframe *)sp; 171 172 memset(&sigframe, 0, sizeof(sigframe)); 173 if (ps->sa_sigdesc[sig].sd_vers != 0) 174 sigframe.pretcode = 175 (char *)(u_long)ps->sa_sigdesc[sig].sd_tramp; 176 else 177 sigframe.pretcode = NULL; 178 179 /* 180 * The user context 181 */ 182 sigframe.uc.luc_flags = 0; 183 sigframe.uc.luc_link = NULL; 184 185 /* This is used regardless of SA_ONSTACK in Linux */ 186 sigframe.uc.luc_stack.ss_sp = l->l_sigstk.ss_sp; 187 sigframe.uc.luc_stack.ss_size = l->l_sigstk.ss_size; 188 sigframe.uc.luc_stack.ss_flags = 0; 189 if (l->l_sigstk.ss_flags & SS_ONSTACK) 190 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_ONSTACK; 191 if (l->l_sigstk.ss_flags & SS_DISABLE) 192 sigframe.uc.luc_stack.ss_flags |= LINUX_SS_DISABLE; 193 194 sigframe.uc.luc_mcontext.r8 = tf->tf_r8; 195 sigframe.uc.luc_mcontext.r9 = tf->tf_r9; 196 sigframe.uc.luc_mcontext.r10 = tf->tf_r10; 197 sigframe.uc.luc_mcontext.r11 = tf->tf_r11; 198 sigframe.uc.luc_mcontext.r12 = tf->tf_r12; 199 sigframe.uc.luc_mcontext.r13 = tf->tf_r13; 200 sigframe.uc.luc_mcontext.r14 = tf->tf_r14; 201 sigframe.uc.luc_mcontext.r15 = tf->tf_r15; 202 sigframe.uc.luc_mcontext.rdi = tf->tf_rdi; 203 sigframe.uc.luc_mcontext.rsi = tf->tf_rsi; 204 sigframe.uc.luc_mcontext.rbp = tf->tf_rbp; 205 sigframe.uc.luc_mcontext.rbx = tf->tf_rbx; 206 sigframe.uc.luc_mcontext.rdx = tf->tf_rdx; 207 sigframe.uc.luc_mcontext.rax = tf->tf_rax; 208 sigframe.uc.luc_mcontext.rcx = tf->tf_rcx; 209 sigframe.uc.luc_mcontext.rsp = tf->tf_rsp; 210 sigframe.uc.luc_mcontext.rip = tf->tf_rip; 211 sigframe.uc.luc_mcontext.eflags = tf->tf_rflags; 212 sigframe.uc.luc_mcontext.cs = tf->tf_cs; 213 sigframe.uc.luc_mcontext.gs = tf->tf_gs; 214 sigframe.uc.luc_mcontext.fs = tf->tf_fs; 215 sigframe.uc.luc_mcontext.err = tf->tf_err; 216 sigframe.uc.luc_mcontext.trapno = tf->tf_trapno; 217 native_to_linux_sigset(&lmask, mask); 218 sigframe.uc.luc_mcontext.oldmask = lmask.sig[0]; 219 sigframe.uc.luc_mcontext.cr2 = (long)pcb->pcb_onfault; 220 sigframe.uc.luc_mcontext.fpstate = fpsp; 221 native_to_linux_sigset(&sigframe.uc.luc_sigmask, mask); 222 native_to_linux_siginfo(&sigframe.info, &ksi->ksi_info); 223 sendsig_reset(l, sig); 224 mutex_exit(p->p_lock); 225 error = 0; 226 227 /* 228 * Save FPU state, if any 229 */ 230 if (fpsp != NULL) { 231 size_t fp_size = sizeof fpregs; 232 /* The netbsd and linux structures both match the fxsave data */ 233 (void)process_read_fpregs(l, &fpregs, &fp_size); 234 error = copyout(&fpregs, fpsp, sizeof(*fpsp)); 235 } 236 237 if (error == 0) 238 error = copyout(&sigframe, sp, sizeof(sigframe)); 239 240 mutex_enter(p->p_lock); 241 242 if (error != 0) { 243 sigexit(l, SIGILL); 244 return; 245 } 246 247 linux_buildcontext(l, catcher, sp); 248 tf->tf_rdi = sigframe.info.lsi_signo; 249 tf->tf_rax = 0; 250 tf->tf_rsi = (long)&sfp->info; 251 tf->tf_rdx = (long)&sfp->uc; 252 253 /* 254 * Remember we use signal stack 255 */ 256 if (onstack) 257 l->l_sigstk.ss_flags |= SS_ONSTACK; 258 return; 259 } 260 261 int 262 linux_sys_modify_ldt(struct lwp *l, const struct linux_sys_modify_ldt_args *v, register_t *retval) 263 { 264 printf("linux_sys_modify_ldt\n"); 265 return 0; 266 } 267 268 int 269 linux_sys_iopl(struct lwp *l, const struct linux_sys_iopl_args *v, register_t *retval) 270 { 271 return 0; 272 } 273 274 int 275 linux_sys_ioperm(struct lwp *l, const struct linux_sys_ioperm_args *v, register_t *retval) 276 { 277 return 0; 278 } 279 280 dev_t 281 linux_fakedev(dev_t dev, int raw) 282 { 283 284 extern const struct cdevsw ptc_cdevsw, pts_cdevsw; 285 const struct cdevsw *cd = cdevsw_lookup(dev); 286 287 if (raw) { 288 #if (NWSDISPLAY > 0) 289 extern const struct cdevsw wsdisplay_cdevsw; 290 if (cd == &wsdisplay_cdevsw) 291 return makedev(LINUX_CONS_MAJOR, (minor(dev) + 1)); 292 #endif 293 } 294 295 if (cd == &ptc_cdevsw) 296 return makedev(LINUX_PTC_MAJOR, minor(dev)); 297 if (cd == &pts_cdevsw) 298 return makedev(LINUX_PTS_MAJOR, minor(dev)); 299 300 return ((minor(dev) & 0xff) | ((major(dev) & 0xfff) << 8) 301 | (((unsigned long long int) (minor(dev) & ~0xff)) << 12) 302 | (((unsigned long long int) (major(dev) & ~0xfff)) << 32)); 303 } 304 305 int 306 linux_machdepioctl(struct lwp *l, const struct linux_sys_ioctl_args *v, register_t *retval) 307 { 308 return 0; 309 } 310 311 int 312 linux_sys_rt_sigreturn(struct lwp *l, const void *v, register_t *retval) 313 { 314 struct linux_ucontext *luctx; 315 struct trapframe *tf = l->l_md.md_regs; 316 struct linux_sigcontext *lsigctx; 317 struct linux_rt_sigframe frame, *fp; 318 ucontext_t uctx; 319 mcontext_t *mctx; 320 struct fxsave *fxarea; 321 int error; 322 323 fp = (struct linux_rt_sigframe *)(tf->tf_rsp - 8); 324 if ((error = copyin(fp, &frame, sizeof(frame))) != 0) { 325 mutex_enter(l->l_proc->p_lock); 326 sigexit(l, SIGILL); 327 return error; 328 } 329 luctx = &frame.uc; 330 lsigctx = &luctx->luc_mcontext; 331 332 memset(&uctx, 0, sizeof(uctx)); 333 mctx = (mcontext_t *)&uctx.uc_mcontext; 334 fxarea = (struct fxsave *)&mctx->__fpregs; 335 336 /* 337 * Set the flags. Linux always have CPU, stack and signal state, 338 * FPU is optional. uc_flags is not used to tell what we have. 339 */ 340 uctx.uc_flags = (_UC_SIGMASK|_UC_CPU|_UC_STACK|_UC_CLRSTACK); 341 if (lsigctx->fpstate != NULL) 342 uctx.uc_flags |= _UC_FPU; 343 uctx.uc_link = NULL; 344 345 /* 346 * Signal set 347 */ 348 linux_to_native_sigset(&uctx.uc_sigmask, &luctx->luc_sigmask); 349 350 /* 351 * CPU state 352 */ 353 mctx->__gregs[_REG_R8] = lsigctx->r8; 354 mctx->__gregs[_REG_R9] = lsigctx->r9; 355 mctx->__gregs[_REG_R10] = lsigctx->r10; 356 mctx->__gregs[_REG_R11] = lsigctx->r11; 357 mctx->__gregs[_REG_R12] = lsigctx->r12; 358 mctx->__gregs[_REG_R13] = lsigctx->r13; 359 mctx->__gregs[_REG_R14] = lsigctx->r14; 360 mctx->__gregs[_REG_R15] = lsigctx->r15; 361 mctx->__gregs[_REG_RDI] = lsigctx->rdi; 362 mctx->__gregs[_REG_RSI] = lsigctx->rsi; 363 mctx->__gregs[_REG_RBP] = lsigctx->rbp; 364 mctx->__gregs[_REG_RBX] = lsigctx->rbx; 365 mctx->__gregs[_REG_RAX] = lsigctx->rax; 366 mctx->__gregs[_REG_RDX] = lsigctx->rdx; 367 mctx->__gregs[_REG_RCX] = lsigctx->rcx; 368 mctx->__gregs[_REG_RIP] = lsigctx->rip; 369 mctx->__gregs[_REG_RFLAGS] = lsigctx->eflags; 370 mctx->__gregs[_REG_CS] = lsigctx->cs; 371 mctx->__gregs[_REG_GS] = lsigctx->gs; 372 mctx->__gregs[_REG_FS] = lsigctx->fs; 373 mctx->__gregs[_REG_ERR] = lsigctx->err; 374 mctx->__gregs[_REG_TRAPNO] = lsigctx->trapno; 375 mctx->__gregs[_REG_ES] = tf->tf_es; 376 mctx->__gregs[_REG_DS] = tf->tf_ds; 377 mctx->__gregs[_REG_RSP] = lsigctx->rsp; /* XXX */ 378 mctx->__gregs[_REG_SS] = tf->tf_ss; 379 380 /* 381 * FPU state 382 */ 383 if (lsigctx->fpstate != NULL) { 384 /* Both structures match the fxstate data */ 385 error = copyin(lsigctx->fpstate, fxarea, sizeof(*fxarea)); 386 if (error != 0) { 387 mutex_enter(l->l_proc->p_lock); 388 sigexit(l, SIGILL); 389 return error; 390 } 391 } 392 393 /* 394 * And the stack 395 */ 396 uctx.uc_stack.ss_flags = 0; 397 if (luctx->luc_stack.ss_flags & LINUX_SS_ONSTACK) 398 uctx.uc_stack.ss_flags |= SS_ONSTACK; 399 400 if (luctx->luc_stack.ss_flags & LINUX_SS_DISABLE) 401 uctx.uc_stack.ss_flags |= SS_DISABLE; 402 403 uctx.uc_stack.ss_sp = luctx->luc_stack.ss_sp; 404 uctx.uc_stack.ss_size = luctx->luc_stack.ss_size; 405 406 /* 407 * And let setucontext deal with that. 408 */ 409 mutex_enter(l->l_proc->p_lock); 410 error = setucontext(l, &uctx); 411 mutex_exit(l->l_proc->p_lock); 412 if (error) 413 return error; 414 415 return EJUSTRETURN; 416 } 417 418 int 419 linux_sys_arch_prctl(struct lwp *l, 420 const struct linux_sys_arch_prctl_args *uap, register_t *retval) 421 { 422 /* { 423 syscallarg(int) code; 424 syscallarg(unsigned long) addr; 425 } */ 426 void *addr = (void *)SCARG(uap, addr); 427 428 switch(SCARG(uap, code)) { 429 case LINUX_ARCH_SET_GS: 430 return x86_set_sdbase(addr, 'g', l, true); 431 432 case LINUX_ARCH_GET_GS: 433 return x86_get_sdbase(addr, 'g'); 434 435 case LINUX_ARCH_SET_FS: 436 return x86_set_sdbase(addr, 'f', l, true); 437 438 case LINUX_ARCH_GET_FS: 439 return x86_get_sdbase(addr, 'f'); 440 441 default: 442 #ifdef DEBUG_LINUX 443 printf("linux_sys_arch_prctl: unexpected code %d\n", 444 SCARG(uap, code)); 445 #endif 446 return EINVAL; 447 } 448 /* NOTREACHED */ 449 } 450 451 const int linux_vsyscall_to_syscall[] = { 452 LINUX_SYS_gettimeofday, 453 LINUX_SYS_time, 454 LINUX_SYS_nosys, /* nosys */ 455 LINUX_SYS_nosys, /* nosys */ 456 }; 457 458 int 459 linux_usertrap(struct lwp *l, vaddr_t trapaddr, void *arg) 460 { 461 struct trapframe *tf = arg; 462 uint64_t retaddr; 463 int vsyscallnr; 464 465 /* 466 * Check for a vsyscall. %rip must be the fault address, 467 * and the address must be in the Linux vsyscall area. 468 * Also, vsyscalls are only done at 1024-byte boundaries. 469 */ 470 471 if (__predict_true(trapaddr < LINUX_VSYSCALL_START)) 472 return 0; 473 474 if (trapaddr != tf->tf_rip) 475 return 0; 476 477 if ((tf->tf_rip & (LINUX_VSYSCALL_SIZE - 1)) != 0) 478 return 0; 479 480 vsyscallnr = (tf->tf_rip - LINUX_VSYSCALL_START) / LINUX_VSYSCALL_SIZE; 481 482 if (vsyscallnr > LINUX_VSYSCALL_MAXNR) 483 return 0; 484 485 /* 486 * Get the return address from the top of the stack, 487 * and fix up the return address. 488 * This assumes the faulting instruction was callq *reg, 489 * which is the only way that vsyscalls are ever entered. 490 */ 491 if (copyin((void *)tf->tf_rsp, &retaddr, sizeof retaddr) != 0) 492 return 0; 493 tf->tf_rip = retaddr; 494 tf->tf_rax = linux_vsyscall_to_syscall[vsyscallnr]; 495 tf->tf_rsp += 8; /* "pop" the return address */ 496 497 #if 0 498 printf("usertrap: rip %p rsp %p retaddr %p vsys %d sys %d\n", 499 (void *)tf->tf_rip, (void *)tf->tf_rsp, (void *)retaddr, 500 vsyscallnr, (int)tf->tf_rax); 501 #endif 502 503 (*l->l_proc->p_md.md_syscall)(tf); 504 505 return 1; 506 } 507 508 static void 509 linux_buildcontext(struct lwp *l, void *catcher, void *f) 510 { 511 struct trapframe *tf = l->l_md.md_regs; 512 513 tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL); 514 tf->tf_rip = (u_int64_t)catcher; 515 tf->tf_cs = GSEL(GUCODE_SEL, SEL_UPL); 516 tf->tf_rflags &= ~PSL_CLEARSIG; 517 tf->tf_rsp = (u_int64_t)f; 518 tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL); 519 } 520
Indexes created Tue Sep 30 20:09:53 GMT 2025