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/lock.h> 41 #include <sys/module.h> 42 #include <sys/mutex.h> 43 #include <sys/proc.h> 44 #include <sys/signalvar.h> 45 #include <sys/sysctl.h> 46 #include <sys/sysent.h> 47 48 #include <vm/vm_param.h> 49 50 #include <arm64/linux/linux.h> 51 #include <arm64/linux/linux_proto.h> 52 #include <compat/linux/linux_dtrace.h> 53 #include <compat/linux/linux_emul.h> 54 #include <compat/linux/linux_ioctl.h> 55 #include <compat/linux/linux_mib.h> 56 #include <compat/linux/linux_misc.h> 57 #include <compat/linux/linux_util.h> 58 #include <compat/linux/linux_vdso.h> 59 60 #ifdef VFP 61 #include <machine/vfp.h> 62 #endif 63 64 MODULE_VERSION(linux64elf, 1); 65 66 const char *linux_kplatform; 67 static int linux_szsigcode; 68 static vm_object_t linux_shared_page_obj; 69 static char *linux_shared_page_mapping; 70 extern char _binary_linux_locore_o_start; 71 extern char _binary_linux_locore_o_end; 72 73 extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; 74 75 SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); 76 77 static int linux_copyout_strings(struct image_params *imgp, 78 uintptr_t *stack_base); 79 static int linux_elf_fixup(uintptr_t *stack_base, 80 struct image_params *iparams); 81 static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel); 82 static void linux_vdso_install(const void *param); 83 static void linux_vdso_deinstall(const void *param); 84 static void linux_set_syscall_retval(struct thread *td, int error); 85 static int linux_fetch_syscall_args(struct thread *td); 86 static void linux_exec_setregs(struct thread *td, struct image_params *imgp, 87 uintptr_t stack); 88 static int linux_vsyscall(struct thread *td); 89 90 /* DTrace init */ 91 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE); 92 93 /* DTrace probes */ 94 LIN_SDT_PROBE_DEFINE2(sysvec, linux_translate_traps, todo, "int", "int"); 95 LIN_SDT_PROBE_DEFINE0(sysvec, linux_exec_setregs, todo); 96 LIN_SDT_PROBE_DEFINE0(sysvec, linux_copyout_auxargs, todo); 97 LIN_SDT_PROBE_DEFINE0(sysvec, linux_elf_fixup, todo); 98 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sigreturn, todo); 99 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sendsig, todo); 100 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vsyscall, todo); 101 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_install, todo); 102 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_deinstall, todo); 103 104 /* LINUXTODO: do we have traps to translate? */ 105 static int 106 linux_translate_traps(int signal, int trap_code) 107 { 108 109 LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code); 110 return (signal); 111 } 112 113 LINUX_VDSO_SYM_CHAR(linux_platform); 114 115 static int 116 linux_fetch_syscall_args(struct thread *td) 117 { 118 struct proc *p; 119 struct syscall_args *sa; 120 register_t *ap; 121 122 p = td->td_proc; 123 ap = td->td_frame->tf_x; 124 sa = &td->td_sa; 125 126 sa->code = td->td_frame->tf_x[8]; 127 /* LINUXTODO: generic syscall? */ 128 if (sa->code >= p->p_sysent->sv_size) 129 sa->callp = &p->p_sysent->sv_table[0]; 130 else 131 sa->callp = &p->p_sysent->sv_table[sa->code]; 132 133 if (sa->callp->sy_narg > MAXARGS) 134 panic("ARM64TODO: Could we have more than %d args?", MAXARGS); 135 memcpy(sa->args, ap, MAXARGS * sizeof(register_t)); 136 137 td->td_retval[0] = 0; 138 return (0); 139 } 140 141 static void 142 linux_set_syscall_retval(struct thread *td, int error) 143 { 144 145 td->td_retval[1] = td->td_frame->tf_x[1]; 146 cpu_set_syscall_retval(td, error); 147 148 if (__predict_false(error != 0)) { 149 if (error != ERESTART && error != EJUSTRETURN) 150 td->td_frame->tf_x[0] = bsd_to_linux_errno(error); 151 } 152 } 153 154 static int 155 linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) 156 { 157 Elf_Auxargs *args; 158 Elf_Auxinfo *argarray, *pos; 159 struct proc *p; 160 int error, issetugid; 161 162 LIN_SDT_PROBE0(sysvec, linux_copyout_auxargs, todo); 163 p = imgp->proc; 164 165 args = (Elf64_Auxargs *)imgp->auxargs; 166 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, 167 M_WAITOK | M_ZERO); 168 169 issetugid = p->p_flag & P_SUGID ? 1 : 0; 170 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, 171 imgp->proc->p_sysent->sv_shared_page_base); 172 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_HWCAP */ 173 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); 174 #endif 175 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); 176 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); 177 AUXARGS_ENTRY(pos, AT_PHENT, args->phent); 178 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); 179 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 180 AUXARGS_ENTRY(pos, AT_BASE, args->base); 181 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 182 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 183 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 184 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 185 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 186 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 187 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); 188 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_PLATFORM */ 189 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); 190 #endif 191 AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary); 192 if (imgp->execpathp != 0) 193 AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp); 194 if (args->execfd != -1) 195 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 196 AUXARGS_ENTRY(pos, AT_NULL, 0); 197 free(imgp->auxargs, M_TEMP); 198 imgp->auxargs = NULL; 199 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); 200 201 error = copyout(argarray, (void *)base, 202 sizeof(*argarray) * LINUX_AT_COUNT); 203 free(argarray, M_TEMP); 204 return (error); 205 } 206 207 static int 208 linux_elf_fixup(uintptr_t *stack_base, struct image_params *imgp) 209 { 210 211 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo); 212 213 return (0); 214 } 215 216 /* 217 * Copy strings out to the new process address space, constructing new arg 218 * and env vector tables. Return a pointer to the base so that it can be used 219 * as the initial stack pointer. 220 * LINUXTODO: deduplicate against other linuxulator archs 221 */ 222 static int 223 linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) 224 { 225 char **vectp; 226 char *stringp; 227 uintptr_t destp, ustringp; 228 struct ps_strings *arginfo; 229 char canary[LINUX_AT_RANDOM_LEN]; 230 size_t execpath_len; 231 struct proc *p; 232 int argc, envc, error; 233 234 /* Calculate string base and vector table pointers. */ 235 if (imgp->execpath != NULL && imgp->auxargs != NULL) 236 execpath_len = strlen(imgp->execpath) + 1; 237 else 238 execpath_len = 0; 239 240 p = imgp->proc; 241 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; 242 destp = (uintptr_t)arginfo; 243 244 if (execpath_len != 0) { 245 destp -= execpath_len; 246 destp = rounddown2(destp, sizeof(void *)); 247 imgp->execpathp = (void *)destp; 248 error = copyout(imgp->execpath, imgp->execpathp, execpath_len); 249 if (error != 0) 250 return (error); 251 } 252 253 /* Prepare the canary for SSP. */ 254 arc4rand(canary, sizeof(canary), 0); 255 destp -= roundup(sizeof(canary), sizeof(void *)); 256 imgp->canary = (void *)destp; 257 error = copyout(canary, imgp->canary, sizeof(canary)); 258 if (error != 0) 259 return (error); 260 261 /* Allocate room for the argument and environment strings. */ 262 destp -= ARG_MAX - imgp->args->stringspace; 263 destp = rounddown2(destp, sizeof(void *)); 264 ustringp = destp; 265 266 if (imgp->auxargs) { 267 /* 268 * Allocate room on the stack for the ELF auxargs 269 * array. It has up to LINUX_AT_COUNT entries. 270 */ 271 destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo); 272 destp = rounddown2(destp, sizeof(void *)); 273 } 274 275 vectp = (char **)destp; 276 277 /* 278 * Allocate room for argc and the argv[] and env vectors including the 279 * terminating NULL pointers. 280 */ 281 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1; 282 vectp = (char **)STACKALIGN(vectp); 283 284 /* vectp also becomes our initial stack base. */ 285 *stack_base = (uintptr_t)vectp; 286 287 stringp = imgp->args->begin_argv; 288 argc = imgp->args->argc; 289 envc = imgp->args->envc; 290 291 /* Copy out strings - arguments and environment. */ 292 error = copyout(stringp, (void *)ustringp, 293 ARG_MAX - imgp->args->stringspace); 294 if (error != 0) 295 return (error); 296 297 /* Fill in "ps_strings" struct for ps, w, etc. */ 298 if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || 299 suword(&arginfo->ps_nargvstr, argc) != 0) 300 return (EFAULT); 301 302 if (suword(vectp++, argc) != 0) 303 return (EFAULT); 304 305 /* Fill in argument portion of vector table. */ 306 for (; argc > 0; --argc) { 307 if (suword(vectp++, ustringp) != 0) 308 return (EFAULT); 309 while (*stringp++ != 0) 310 ustringp++; 311 ustringp++; 312 } 313 314 /* A null vector table pointer separates the argp's from the envp's. */ 315 if (suword(vectp++, 0) != 0) 316 return (EFAULT); 317 318 if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || 319 suword(&arginfo->ps_nenvstr, envc) != 0) 320 return (EFAULT); 321 322 /* Fill in environment portion of vector table. */ 323 for (; envc > 0; --envc) { 324 if (suword(vectp++, ustringp) != 0) 325 return (EFAULT); 326 while (*stringp++ != 0) 327 ustringp++; 328 ustringp++; 329 } 330 331 /* The end of the vector table is a null pointer. */ 332 if (suword(vectp, 0) != 0) 333 return (EFAULT); 334 335 if (imgp->auxargs) { 336 vectp++; 337 error = imgp->sysent->sv_copyout_auxargs(imgp, 338 (uintptr_t)vectp); 339 if (error != 0) 340 return (error); 341 } 342 343 return (0); 344 } 345 346 /* 347 * Reset registers to default values on exec. 348 */ 349 static void 350 linux_exec_setregs(struct thread *td, struct image_params *imgp, 351 uintptr_t stack) 352 { 353 struct trapframe *regs = td->td_frame; 354 355 /* LINUXTODO: validate */ 356 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo); 357 358 memset(regs, 0, sizeof(*regs)); 359 /* glibc start.S registers function pointer in x0 with atexit. */ 360 regs->tf_sp = stack; 361 #if 0 /* LINUXTODO: See if this is used. */ 362 regs->tf_lr = imgp->entry_addr; 363 #else 364 regs->tf_lr = 0xffffffffffffffff; 365 #endif 366 regs->tf_elr = imgp->entry_addr; 367 368 #ifdef VFP 369 vfp_reset_state(td, td->td_pcb); 370 #endif 371 } 372 373 int 374 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 375 { 376 377 /* LINUXTODO: implement */ 378 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo); 379 return (EDOOFUS); 380 } 381 382 static void 383 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 384 { 385 386 /* LINUXTODO: implement */ 387 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo); 388 } 389 390 static int 391 linux_vsyscall(struct thread *td) 392 { 393 394 /* LINUXTODO: implement */ 395 LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo); 396 return (EDOOFUS); 397 } 398 399 struct sysentvec elf_linux_sysvec = { 400 .sv_size = LINUX_SYS_MAXSYSCALL, 401 .sv_table = linux_sysent, 402 .sv_transtrap = linux_translate_traps, 403 .sv_fixup = linux_elf_fixup, 404 .sv_sendsig = linux_rt_sendsig, 405 .sv_sigcode = &_binary_linux_locore_o_start, 406 .sv_szsigcode = &linux_szsigcode, 407 .sv_name = "Linux ELF64", 408 .sv_coredump = elf64_coredump, 409 .sv_imgact_try = linux_exec_imgact_try, 410 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 411 .sv_minuser = VM_MIN_ADDRESS, 412 .sv_maxuser = VM_MAXUSER_ADDRESS, 413 .sv_usrstack = USRSTACK, 414 .sv_psstrings = PS_STRINGS, /* XXX */ 415 .sv_stackprot = VM_PROT_READ | VM_PROT_WRITE, 416 .sv_copyout_auxargs = linux_copyout_auxargs, 417 .sv_copyout_strings = linux_copyout_strings, 418 .sv_setregs = linux_exec_setregs, 419 .sv_fixlimit = NULL, 420 .sv_maxssiz = NULL, 421 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP, 422 .sv_set_syscall_retval = linux_set_syscall_retval, 423 .sv_fetch_syscall_args = linux_fetch_syscall_args, 424 .sv_syscallnames = NULL, 425 .sv_shared_page_base = SHAREDPAGE, 426 .sv_shared_page_len = PAGE_SIZE, 427 .sv_schedtail = linux_schedtail, 428 .sv_thread_detach = linux_thread_detach, 429 .sv_trap = linux_vsyscall, 430 .sv_onexec = linux_on_exec, 431 .sv_onexit = linux_on_exit, 432 .sv_ontdexit = linux_thread_dtor, 433 }; 434 435 static void 436 linux_vdso_install(const void *param) 437 { 438 439 linux_szsigcode = (&_binary_linux_locore_o_end - 440 &_binary_linux_locore_o_start); 441 442 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len) 443 panic("invalid Linux VDSO size\n"); 444 445 __elfN(linux_vdso_fixup)(&elf_linux_sysvec); 446 447 linux_shared_page_obj = __elfN(linux_shared_page_init) 448 (&linux_shared_page_mapping); 449 450 __elfN(linux_vdso_reloc)(&elf_linux_sysvec); 451 452 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode, 453 linux_szsigcode); 454 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj; 455 456 printf("LINUXTODO: %s: fix linux_kplatform\n", __func__); 457 #if 0 458 linux_kplatform = linux_shared_page_mapping + 459 (linux_platform - (caddr_t)elf_linux_sysvec.sv_shared_page_base); 460 #else 461 linux_kplatform = "arm64"; 462 #endif 463 } 464 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY, 465 linux_vdso_install, NULL); 466 467 static void 468 linux_vdso_deinstall(const void *param) 469 { 470 471 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo); 472 __elfN(linux_shared_page_fini)(linux_shared_page_obj, 473 linux_shared_page_mapping); 474 } 475 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, 476 linux_vdso_deinstall, NULL); 477 478 static char GNU_ABI_VENDOR[] = "GNU"; 479 static int GNU_ABI_LINUX = 0; 480 481 /* LINUXTODO: deduplicate */ 482 static bool 483 linux_trans_osrel(const Elf_Note *note, int32_t *osrel) 484 { 485 const Elf32_Word *desc; 486 uintptr_t p; 487 488 p = (uintptr_t)(note + 1); 489 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 490 491 desc = (const Elf32_Word *)p; 492 if (desc[0] != GNU_ABI_LINUX) 493 return (false); 494 495 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); 496 return (true); 497 } 498 499 static Elf_Brandnote linux64_brandnote = { 500 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 501 .hdr.n_descsz = 16, 502 .hdr.n_type = 1, 503 .vendor = GNU_ABI_VENDOR, 504 .flags = BN_TRANSLATE_OSREL, 505 .trans_osrel = linux_trans_osrel 506 }; 507 508 static Elf64_Brandinfo linux_glibc2brand = { 509 .brand = ELFOSABI_LINUX, 510 .machine = EM_AARCH64, 511 .compat_3_brand = "Linux", 512 .emul_path = linux_emul_path, 513 .interp_path = "/lib64/ld-linux-x86-64.so.2", 514 .sysvec = &elf_linux_sysvec, 515 .interp_newpath = NULL, 516 .brand_note = &linux64_brandnote, 517 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 518 }; 519 520 Elf64_Brandinfo *linux_brandlist[] = { 521 &linux_glibc2brand, 522 NULL 523 }; 524 525 static int 526 linux64_elf_modevent(module_t mod, int type, void *data) 527 { 528 Elf64_Brandinfo **brandinfo; 529 struct linux_ioctl_handler**lihp; 530 int error; 531 532 error = 0; 533 switch(type) { 534 case MOD_LOAD: 535 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 536 ++brandinfo) 537 if (elf64_insert_brand_entry(*brandinfo) < 0) 538 error = EINVAL; 539 if (error == 0) { 540 SET_FOREACH(lihp, linux_ioctl_handler_set) 541 linux_ioctl_register_handler(*lihp); 542 stclohz = (stathz ? stathz : hz); 543 if (bootverbose) 544 printf("Linux arm64 ELF exec handler installed\n"); 545 } 546 break; 547 case MOD_UNLOAD: 548 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 549 ++brandinfo) 550 if (elf64_brand_inuse(*brandinfo)) 551 error = EBUSY; 552 if (error == 0) { 553 for (brandinfo = &linux_brandlist[0]; 554 *brandinfo != NULL; ++brandinfo) 555 if (elf64_remove_brand_entry(*brandinfo) < 0) 556 error = EINVAL; 557 } 558 if (error == 0) { 559 SET_FOREACH(lihp, linux_ioctl_handler_set) 560 linux_ioctl_unregister_handler(*lihp); 561 if (bootverbose) 562 printf("Linux ELF exec handler removed\n"); 563 } else 564 printf("Could not deinstall ELF interpreter entry\n"); 565 break; 566 default: 567 return (EOPNOTSUPP); 568 } 569 return (error); 570 } 571 572 static moduledata_t linux64_elf_mod = { 573 "linux64elf", 574 linux64_elf_modevent, 575 0 576 }; 577 578 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 579 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); 580 FEATURE(linux64, "AArch64 Linux 64bit support"); 581