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