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