1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 1994-1996 Søren Schmidt 5 * Copyright (c) 2018 Turing Robotic Industries Inc. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include <sys/param.h> 33 #include <sys/systm.h> 34 #include <sys/cdefs.h> 35 #include <sys/elf.h> 36 #include <sys/exec.h> 37 #include <sys/imgact.h> 38 #include <sys/imgact_elf.h> 39 #include <sys/kernel.h> 40 #include <sys/ktr.h> 41 #include <sys/lock.h> 42 #include <sys/module.h> 43 #include <sys/mutex.h> 44 #include <sys/proc.h> 45 #include <sys/stddef.h> 46 #include <sys/signalvar.h> 47 #include <sys/syscallsubr.h> 48 #include <sys/sysctl.h> 49 #include <sys/sysent.h> 50 51 #include <vm/vm.h> 52 #include <vm/pmap.h> 53 #include <vm/vm_map.h> 54 #include <vm/vm_extern.h> 55 #include <vm/vm_object.h> 56 #include <vm/vm_page.h> 57 #include <vm/vm_param.h> 58 59 #include <arm64/linux/linux.h> 60 #include <arm64/linux/linux_proto.h> 61 #include <compat/linux/linux_dtrace.h> 62 #include <compat/linux/linux_emul.h> 63 #include <compat/linux/linux_fork.h> 64 #include <compat/linux/linux_ioctl.h> 65 #include <compat/linux/linux_mib.h> 66 #include <compat/linux/linux_misc.h> 67 #include <compat/linux/linux_signal.h> 68 #include <compat/linux/linux_util.h> 69 #include <compat/linux/linux_vdso.h> 70 71 #include <machine/md_var.h> 72 73 #ifdef VFP 74 #include <machine/vfp.h> 75 #endif 76 77 MODULE_VERSION(linux64elf, 1); 78 79 #define LINUX_VDSOPAGE_SIZE PAGE_SIZE * 2 80 #define LINUX_VDSOPAGE (VM_MAXUSER_ADDRESS - \ 81 LINUX_VDSOPAGE_SIZE) 82 #define LINUX_SHAREDPAGE (LINUX_VDSOPAGE - PAGE_SIZE) 83 /* 84 * PAGE_SIZE - the size 85 * of the native SHAREDPAGE 86 */ 87 #define LINUX_USRSTACK LINUX_SHAREDPAGE 88 #define LINUX_PS_STRINGS (LINUX_USRSTACK - \ 89 sizeof(struct ps_strings)) 90 91 static int linux_szsigcode; 92 static vm_object_t linux_vdso_obj; 93 static char *linux_vdso_mapping; 94 extern char _binary_linux_vdso_so_o_start; 95 extern char _binary_linux_vdso_so_o_end; 96 static vm_offset_t linux_vdso_base; 97 98 extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; 99 100 SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); 101 102 static int linux_copyout_strings(struct image_params *imgp, 103 uintptr_t *stack_base); 104 static int linux_elf_fixup(uintptr_t *stack_base, 105 struct image_params *iparams); 106 static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel); 107 static void linux_vdso_install(const void *param); 108 static void linux_vdso_deinstall(const void *param); 109 static void linux_vdso_reloc(char *mapping, Elf_Addr offset); 110 static void linux_set_syscall_retval(struct thread *td, int error); 111 static int linux_fetch_syscall_args(struct thread *td); 112 static void linux_exec_setregs(struct thread *td, struct image_params *imgp, 113 uintptr_t stack); 114 static void linux_exec_sysvec_init(void *param); 115 static int linux_on_exec_vmspace(struct proc *p, 116 struct image_params *imgp); 117 118 /* DTrace init */ 119 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE); 120 121 /* DTrace probes */ 122 LIN_SDT_PROBE_DEFINE2(sysvec, linux_translate_traps, todo, "int", "int"); 123 LIN_SDT_PROBE_DEFINE0(sysvec, linux_exec_setregs, todo); 124 LIN_SDT_PROBE_DEFINE0(sysvec, linux_copyout_auxargs, todo); 125 LIN_SDT_PROBE_DEFINE0(sysvec, linux_elf_fixup, todo); 126 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sigreturn, todo); 127 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sendsig, todo); 128 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_install, todo); 129 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_deinstall, todo); 130 131 LINUX_VDSO_SYM_CHAR(linux_platform); 132 LINUX_VDSO_SYM_INTPTR(kern_timekeep_base); 133 LINUX_VDSO_SYM_INTPTR(__kernel_rt_sigreturn); 134 135 /* LINUXTODO: do we have traps to translate? */ 136 static int 137 linux_translate_traps(int signal, int trap_code) 138 { 139 140 LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code); 141 return (signal); 142 } 143 144 static int 145 linux_fetch_syscall_args(struct thread *td) 146 { 147 struct proc *p; 148 struct syscall_args *sa; 149 register_t *ap; 150 151 p = td->td_proc; 152 ap = td->td_frame->tf_x; 153 sa = &td->td_sa; 154 155 sa->code = td->td_frame->tf_x[8]; 156 sa->original_code = sa->code; 157 /* LINUXTODO: generic syscall? */ 158 if (sa->code >= p->p_sysent->sv_size) 159 sa->callp = &p->p_sysent->sv_table[0]; 160 else 161 sa->callp = &p->p_sysent->sv_table[sa->code]; 162 163 if (sa->callp->sy_narg > nitems(sa->args)) 164 panic("ARM64TODO: Could we have more than %zu args?", 165 nitems(sa->args)); 166 memcpy(sa->args, ap, nitems(sa->args) * sizeof(register_t)); 167 168 td->td_retval[0] = 0; 169 return (0); 170 } 171 172 static void 173 linux_set_syscall_retval(struct thread *td, int error) 174 { 175 176 td->td_retval[1] = td->td_frame->tf_x[1]; 177 cpu_set_syscall_retval(td, error); 178 179 if (__predict_false(error != 0)) { 180 if (error != ERESTART && error != EJUSTRETURN) 181 td->td_frame->tf_x[0] = bsd_to_linux_errno(error); 182 } 183 } 184 185 static int 186 linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) 187 { 188 Elf_Auxargs *args; 189 Elf_Auxinfo *argarray, *pos; 190 struct proc *p; 191 int error, issetugid; 192 193 LIN_SDT_PROBE0(sysvec, linux_copyout_auxargs, todo); 194 p = imgp->proc; 195 196 args = (Elf64_Auxargs *)imgp->auxargs; 197 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, 198 M_WAITOK | M_ZERO); 199 200 issetugid = p->p_flag & P_SUGID ? 1 : 0; 201 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, linux_vdso_base); 202 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, *imgp->sysent->sv_hwcap); 203 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 204 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); 205 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); 206 AUXARGS_ENTRY(pos, AT_PHENT, args->phent); 207 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); 208 AUXARGS_ENTRY(pos, AT_BASE, args->base); 209 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 210 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 211 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 212 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 213 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 214 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 215 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); 216 AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary); 217 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, *imgp->sysent->sv_hwcap2); 218 if (imgp->execpathp != 0) 219 AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp); 220 if (args->execfd != -1) 221 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 222 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); 223 AUXARGS_ENTRY(pos, AT_NULL, 0); 224 225 free(imgp->auxargs, M_TEMP); 226 imgp->auxargs = NULL; 227 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); 228 229 error = copyout(argarray, (void *)base, 230 sizeof(*argarray) * LINUX_AT_COUNT); 231 free(argarray, M_TEMP); 232 return (error); 233 } 234 235 static int 236 linux_elf_fixup(uintptr_t *stack_base, struct image_params *imgp) 237 { 238 239 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo); 240 241 return (0); 242 } 243 244 /* 245 * Copy strings out to the new process address space, constructing new arg 246 * and env vector tables. Return a pointer to the base so that it can be used 247 * as the initial stack pointer. 248 * LINUXTODO: deduplicate against other linuxulator archs 249 */ 250 static int 251 linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) 252 { 253 char **vectp; 254 char *stringp; 255 uintptr_t destp, ustringp; 256 struct ps_strings *arginfo; 257 char canary[LINUX_AT_RANDOM_LEN]; 258 size_t execpath_len; 259 struct proc *p; 260 int argc, envc, error; 261 262 /* Calculate string base and vector table pointers. */ 263 if (imgp->execpath != NULL && imgp->auxargs != NULL) 264 execpath_len = strlen(imgp->execpath) + 1; 265 else 266 execpath_len = 0; 267 268 p = imgp->proc; 269 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; 270 destp = (uintptr_t)arginfo; 271 272 if (execpath_len != 0) { 273 destp -= execpath_len; 274 destp = rounddown2(destp, sizeof(void *)); 275 imgp->execpathp = (void *)destp; 276 error = copyout(imgp->execpath, imgp->execpathp, execpath_len); 277 if (error != 0) 278 return (error); 279 } 280 281 /* Prepare the canary for SSP. */ 282 arc4rand(canary, sizeof(canary), 0); 283 destp -= roundup(sizeof(canary), sizeof(void *)); 284 imgp->canary = (void *)destp; 285 error = copyout(canary, imgp->canary, sizeof(canary)); 286 if (error != 0) 287 return (error); 288 289 /* Allocate room for the argument and environment strings. */ 290 destp -= ARG_MAX - imgp->args->stringspace; 291 destp = rounddown2(destp, sizeof(void *)); 292 ustringp = destp; 293 294 if (imgp->auxargs) { 295 /* 296 * Allocate room on the stack for the ELF auxargs 297 * array. It has up to LINUX_AT_COUNT entries. 298 */ 299 destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo); 300 destp = rounddown2(destp, sizeof(void *)); 301 } 302 303 vectp = (char **)destp; 304 305 /* 306 * Allocate room for argc and the argv[] and env vectors including the 307 * terminating NULL pointers. 308 */ 309 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1; 310 vectp = (char **)STACKALIGN(vectp); 311 312 /* vectp also becomes our initial stack base. */ 313 *stack_base = (uintptr_t)vectp; 314 315 stringp = imgp->args->begin_argv; 316 argc = imgp->args->argc; 317 envc = imgp->args->envc; 318 319 /* Copy out strings - arguments and environment. */ 320 error = copyout(stringp, (void *)ustringp, 321 ARG_MAX - imgp->args->stringspace); 322 if (error != 0) 323 return (error); 324 325 /* Fill in "ps_strings" struct for ps, w, etc. */ 326 if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || 327 suword(&arginfo->ps_nargvstr, argc) != 0) 328 return (EFAULT); 329 330 if (suword(vectp++, argc) != 0) 331 return (EFAULT); 332 333 /* Fill in argument portion of vector table. */ 334 for (; argc > 0; --argc) { 335 if (suword(vectp++, ustringp) != 0) 336 return (EFAULT); 337 while (*stringp++ != 0) 338 ustringp++; 339 ustringp++; 340 } 341 342 /* A null vector table pointer separates the argp's from the envp's. */ 343 if (suword(vectp++, 0) != 0) 344 return (EFAULT); 345 346 if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || 347 suword(&arginfo->ps_nenvstr, envc) != 0) 348 return (EFAULT); 349 350 /* Fill in environment portion of vector table. */ 351 for (; envc > 0; --envc) { 352 if (suword(vectp++, ustringp) != 0) 353 return (EFAULT); 354 while (*stringp++ != 0) 355 ustringp++; 356 ustringp++; 357 } 358 359 /* The end of the vector table is a null pointer. */ 360 if (suword(vectp, 0) != 0) 361 return (EFAULT); 362 363 if (imgp->auxargs) { 364 vectp++; 365 error = imgp->sysent->sv_copyout_auxargs(imgp, 366 (uintptr_t)vectp); 367 if (error != 0) 368 return (error); 369 } 370 371 return (0); 372 } 373 374 /* 375 * Reset registers to default values on exec. 376 */ 377 static void 378 linux_exec_setregs(struct thread *td, struct image_params *imgp, 379 uintptr_t stack) 380 { 381 struct trapframe *regs = td->td_frame; 382 struct pcb *pcb = td->td_pcb; 383 384 /* LINUXTODO: validate */ 385 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo); 386 387 memset(regs, 0, sizeof(*regs)); 388 /* glibc start.S registers function pointer in x0 with atexit. */ 389 regs->tf_sp = stack; 390 #if 0 /* LINUXTODO: See if this is used. */ 391 regs->tf_lr = imgp->entry_addr; 392 #else 393 regs->tf_lr = 0xffffffffffffffff; 394 #endif 395 regs->tf_elr = imgp->entry_addr; 396 397 pcb->pcb_tpidr_el0 = 0; 398 pcb->pcb_tpidrro_el0 = 0; 399 WRITE_SPECIALREG(tpidrro_el0, 0); 400 WRITE_SPECIALREG(tpidr_el0, 0); 401 402 #ifdef VFP 403 vfp_reset_state(td, pcb); 404 #endif 405 406 /* 407 * Clear debug register state. It is not applicable to the new process. 408 */ 409 bzero(&pcb->pcb_dbg_regs, sizeof(pcb->pcb_dbg_regs)); 410 } 411 412 int 413 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 414 { 415 struct l_sigframe frame; 416 struct trapframe *tf; 417 int error; 418 419 tf = td->td_frame; 420 421 if (copyin((void *)tf->tf_sp, &frame, sizeof(frame))) 422 return (EFAULT); 423 424 error = set_mcontext(td, &frame.sf_uc.uc_mcontext); 425 if (error != 0) 426 return (error); 427 428 /* Restore signal mask. */ 429 kern_sigprocmask(td, SIG_SETMASK, &frame.sf_uc.uc_sigmask, NULL, 0); 430 431 return (EJUSTRETURN); 432 } 433 434 static void 435 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 436 { 437 struct thread *td; 438 struct proc *p; 439 struct trapframe *tf; 440 struct l_sigframe *fp, frame; 441 struct sigacts *psp; 442 int onstack, sig; 443 444 td = curthread; 445 p = td->td_proc; 446 PROC_LOCK_ASSERT(p, MA_OWNED); 447 448 sig = ksi->ksi_signo; 449 psp = p->p_sigacts; 450 mtx_assert(&psp->ps_mtx, MA_OWNED); 451 452 tf = td->td_frame; 453 onstack = sigonstack(tf->tf_sp); 454 455 CTR4(KTR_SIG, "sendsig: td=%p (%s) catcher=%p sig=%d", td, p->p_comm, 456 catcher, sig); 457 458 /* Allocate and validate space for the signal handler context. */ 459 if ((td->td_pflags & TDP_ALTSTACK) != 0 && !onstack && 460 SIGISMEMBER(psp->ps_sigonstack, sig)) { 461 fp = (struct l_sigframe *)((uintptr_t)td->td_sigstk.ss_sp + 462 td->td_sigstk.ss_size); 463 #if defined(COMPAT_43) 464 td->td_sigstk.ss_flags |= SS_ONSTACK; 465 #endif 466 } else { 467 fp = (struct l_sigframe *)td->td_frame->tf_sp; 468 } 469 470 /* Make room, keeping the stack aligned */ 471 fp--; 472 fp = (struct l_sigframe *)STACKALIGN(fp); 473 474 /* Fill in the frame to copy out */ 475 bzero(&frame, sizeof(frame)); 476 get_mcontext(td, &frame.sf_uc.uc_mcontext, 0); 477 478 /* Translate the signal. */ 479 sig = bsd_to_linux_signal(sig); 480 481 siginfo_to_lsiginfo(&ksi->ksi_info, &frame.sf_si, sig); 482 frame.sf_uc.uc_sigmask = *mask; 483 frame.sf_uc.uc_stack = td->td_sigstk; 484 frame.sf_uc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) != 0 ? 485 (onstack ? SS_ONSTACK : 0) : SS_DISABLE; 486 mtx_unlock(&psp->ps_mtx); 487 PROC_UNLOCK(td->td_proc); 488 489 /* Copy the sigframe out to the user's stack. */ 490 if (copyout(&frame, fp, sizeof(*fp)) != 0) { 491 /* Process has trashed its stack. Kill it. */ 492 CTR2(KTR_SIG, "sendsig: sigexit td=%p fp=%p", td, fp); 493 PROC_LOCK(p); 494 sigexit(td, SIGILL); 495 } 496 497 tf->tf_x[0]= sig; 498 tf->tf_x[1] = (register_t)&fp->sf_si; 499 tf->tf_x[2] = (register_t)&fp->sf_uc; 500 501 tf->tf_elr = (register_t)catcher; 502 tf->tf_sp = (register_t)fp; 503 tf->tf_lr = (register_t)__kernel_rt_sigreturn; 504 505 CTR3(KTR_SIG, "sendsig: return td=%p pc=%#x sp=%#x", td, tf->tf_elr, 506 tf->tf_sp); 507 508 PROC_LOCK(p); 509 mtx_lock(&psp->ps_mtx); 510 } 511 512 struct sysentvec elf_linux_sysvec = { 513 .sv_size = LINUX_SYS_MAXSYSCALL, 514 .sv_table = linux_sysent, 515 .sv_transtrap = linux_translate_traps, 516 .sv_fixup = linux_elf_fixup, 517 .sv_sendsig = linux_rt_sendsig, 518 .sv_sigcode = &_binary_linux_vdso_so_o_start, 519 .sv_szsigcode = &linux_szsigcode, 520 .sv_name = "Linux ELF64", 521 .sv_coredump = elf64_coredump, 522 .sv_elf_core_osabi = ELFOSABI_NONE, 523 .sv_elf_core_abi_vendor = LINUX_ABI_VENDOR, 524 .sv_elf_core_prepare_notes = linux64_prepare_notes, 525 .sv_imgact_try = linux_exec_imgact_try, 526 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 527 .sv_minuser = VM_MIN_ADDRESS, 528 .sv_maxuser = VM_MAXUSER_ADDRESS, 529 .sv_usrstack = LINUX_USRSTACK, 530 .sv_psstrings = LINUX_PS_STRINGS, 531 .sv_stackprot = VM_PROT_READ | VM_PROT_WRITE, 532 .sv_copyout_auxargs = linux_copyout_auxargs, 533 .sv_copyout_strings = linux_copyout_strings, 534 .sv_setregs = linux_exec_setregs, 535 .sv_fixlimit = NULL, 536 .sv_maxssiz = NULL, 537 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP | SV_SIG_DISCIGN | 538 SV_SIG_WAITNDQ | SV_TIMEKEEP, 539 .sv_set_syscall_retval = linux_set_syscall_retval, 540 .sv_fetch_syscall_args = linux_fetch_syscall_args, 541 .sv_syscallnames = NULL, 542 .sv_shared_page_base = LINUX_SHAREDPAGE, 543 .sv_shared_page_len = PAGE_SIZE, 544 .sv_schedtail = linux_schedtail, 545 .sv_thread_detach = linux_thread_detach, 546 .sv_trap = NULL, 547 .sv_hwcap = &elf_hwcap, 548 .sv_hwcap2 = &elf_hwcap2, 549 .sv_onexec = linux_on_exec_vmspace, 550 .sv_onexit = linux_on_exit, 551 .sv_ontdexit = linux_thread_dtor, 552 .sv_setid_allowed = &linux_setid_allowed_query, 553 }; 554 555 static int 556 linux_on_exec_vmspace(struct proc *p, struct image_params *imgp) 557 { 558 int error; 559 560 error = linux_map_vdso(p, linux_vdso_obj, linux_vdso_base, 561 LINUX_VDSOPAGE_SIZE, imgp); 562 if (error == 0) 563 linux_on_exec(p, imgp); 564 return (error); 565 } 566 567 /* 568 * linux_vdso_install() and linux_exec_sysvec_init() must be called 569 * after exec_sysvec_init() which is SI_SUB_EXEC (SI_ORDER_ANY). 570 */ 571 static void 572 linux_exec_sysvec_init(void *param) 573 { 574 l_uintptr_t *ktimekeep_base; 575 struct sysentvec *sv; 576 ptrdiff_t tkoff; 577 578 sv = param; 579 /* Fill timekeep_base */ 580 exec_sysvec_init(sv); 581 582 tkoff = kern_timekeep_base - linux_vdso_base; 583 ktimekeep_base = (l_uintptr_t *)(linux_vdso_mapping + tkoff); 584 *ktimekeep_base = sv->sv_timekeep_base; 585 } 586 SYSINIT(elf_linux_exec_sysvec_init, SI_SUB_EXEC + 1, SI_ORDER_ANY, 587 linux_exec_sysvec_init, &elf_linux_sysvec); 588 589 static void 590 linux_vdso_install(const void *param) 591 { 592 char *vdso_start = &_binary_linux_vdso_so_o_start; 593 char *vdso_end = &_binary_linux_vdso_so_o_end; 594 595 linux_szsigcode = vdso_end - vdso_start; 596 MPASS(linux_szsigcode <= LINUX_VDSOPAGE_SIZE); 597 598 linux_vdso_base = LINUX_VDSOPAGE; 599 600 __elfN(linux_vdso_fixup)(vdso_start, linux_vdso_base); 601 602 linux_vdso_obj = __elfN(linux_shared_page_init) 603 (&linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); 604 bcopy(vdso_start, linux_vdso_mapping, linux_szsigcode); 605 606 linux_vdso_reloc(linux_vdso_mapping, linux_vdso_base); 607 } 608 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC + 1, SI_ORDER_FIRST, 609 linux_vdso_install, NULL); 610 611 static void 612 linux_vdso_deinstall(const void *param) 613 { 614 615 __elfN(linux_shared_page_fini)(linux_vdso_obj, 616 linux_vdso_mapping, LINUX_VDSOPAGE_SIZE); 617 } 618 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, 619 linux_vdso_deinstall, NULL); 620 621 static void 622 linux_vdso_reloc(char *mapping, Elf_Addr offset) 623 { 624 Elf_Size rtype, symidx; 625 const Elf_Rela *rela; 626 const Elf_Shdr *shdr; 627 const Elf_Ehdr *ehdr; 628 Elf_Addr *where; 629 Elf_Addr addr, addend; 630 int i, relacnt; 631 632 MPASS(offset != 0); 633 634 relacnt = 0; 635 ehdr = (const Elf_Ehdr *)mapping; 636 shdr = (const Elf_Shdr *)(mapping + ehdr->e_shoff); 637 for (i = 0; i < ehdr->e_shnum; i++) 638 { 639 switch (shdr[i].sh_type) { 640 case SHT_REL: 641 printf("Linux Aarch64 vDSO: unexpected Rel section\n"); 642 break; 643 case SHT_RELA: 644 rela = (const Elf_Rela *)(mapping + shdr[i].sh_offset); 645 relacnt = shdr[i].sh_size / sizeof(*rela); 646 } 647 } 648 649 for (i = 0; i < relacnt; i++, rela++) { 650 where = (Elf_Addr *)(mapping + rela->r_offset); 651 addend = rela->r_addend; 652 rtype = ELF_R_TYPE(rela->r_info); 653 symidx = ELF_R_SYM(rela->r_info); 654 655 switch (rtype) { 656 case R_AARCH64_NONE: /* none */ 657 break; 658 659 case R_AARCH64_RELATIVE: /* B + A */ 660 addr = (Elf_Addr)(mapping + addend); 661 if (*where != addr) 662 *where = addr; 663 break; 664 default: 665 printf("Linux Aarch64 vDSO: unexpected relocation type %ld, " 666 "symbol index %ld\n", rtype, symidx); 667 } 668 } 669 } 670 671 static char GNU_ABI_VENDOR[] = "GNU"; 672 static int GNU_ABI_LINUX = 0; 673 674 /* LINUXTODO: deduplicate */ 675 static bool 676 linux_trans_osrel(const Elf_Note *note, int32_t *osrel) 677 { 678 const Elf32_Word *desc; 679 uintptr_t p; 680 681 p = (uintptr_t)(note + 1); 682 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 683 684 desc = (const Elf32_Word *)p; 685 if (desc[0] != GNU_ABI_LINUX) 686 return (false); 687 688 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); 689 return (true); 690 } 691 692 static Elf_Brandnote linux64_brandnote = { 693 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 694 .hdr.n_descsz = 16, 695 .hdr.n_type = 1, 696 .vendor = GNU_ABI_VENDOR, 697 .flags = BN_TRANSLATE_OSREL, 698 .trans_osrel = linux_trans_osrel 699 }; 700 701 static Elf64_Brandinfo linux_glibc2brand = { 702 .brand = ELFOSABI_LINUX, 703 .machine = EM_AARCH64, 704 .compat_3_brand = "Linux", 705 .emul_path = linux_emul_path, 706 .interp_path = "/lib64/ld-linux-x86-64.so.2", 707 .sysvec = &elf_linux_sysvec, 708 .interp_newpath = NULL, 709 .brand_note = &linux64_brandnote, 710 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 711 }; 712 713 Elf64_Brandinfo *linux_brandlist[] = { 714 &linux_glibc2brand, 715 NULL 716 }; 717 718 static int 719 linux64_elf_modevent(module_t mod, int type, void *data) 720 { 721 Elf64_Brandinfo **brandinfo; 722 struct linux_ioctl_handler**lihp; 723 int error; 724 725 error = 0; 726 switch(type) { 727 case MOD_LOAD: 728 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 729 ++brandinfo) 730 if (elf64_insert_brand_entry(*brandinfo) < 0) 731 error = EINVAL; 732 if (error == 0) { 733 SET_FOREACH(lihp, linux_ioctl_handler_set) 734 linux_ioctl_register_handler(*lihp); 735 stclohz = (stathz ? stathz : hz); 736 if (bootverbose) 737 printf("Linux arm64 ELF exec handler installed\n"); 738 } 739 break; 740 case MOD_UNLOAD: 741 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 742 ++brandinfo) 743 if (elf64_brand_inuse(*brandinfo)) 744 error = EBUSY; 745 if (error == 0) { 746 for (brandinfo = &linux_brandlist[0]; 747 *brandinfo != NULL; ++brandinfo) 748 if (elf64_remove_brand_entry(*brandinfo) < 0) 749 error = EINVAL; 750 } 751 if (error == 0) { 752 SET_FOREACH(lihp, linux_ioctl_handler_set) 753 linux_ioctl_unregister_handler(*lihp); 754 if (bootverbose) 755 printf("Linux arm64 ELF exec handler removed\n"); 756 } else 757 printf("Could not deinstall Linux arm64 ELF interpreter entry\n"); 758 break; 759 default: 760 return (EOPNOTSUPP); 761 } 762 return (error); 763 } 764 765 static moduledata_t linux64_elf_mod = { 766 "linux64elf", 767 linux64_elf_modevent, 768 0 769 }; 770 771 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 772 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); 773 FEATURE(linux64, "AArch64 Linux 64bit support"); 774