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