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