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_futex.h> 55 #include <compat/linux/linux_ioctl.h> 56 #include <compat/linux/linux_mib.h> 57 #include <compat/linux/linux_misc.h> 58 #include <compat/linux/linux_vdso.h> 59 60 MODULE_VERSION(linux64elf, 1); 61 62 #if defined(DEBUG) 63 SYSCTL_PROC(_compat_linux, OID_AUTO, debug, CTLTYPE_STRING | CTLFLAG_RW, 0, 0, 64 linux_sysctl_debug, "A", "64-bit Linux debugging control"); 65 #endif 66 67 const char *linux_kplatform; 68 static int linux_szsigcode; 69 static vm_object_t linux_shared_page_obj; 70 static char *linux_shared_page_mapping; 71 extern char _binary_linux_locore_o_start; 72 extern char _binary_linux_locore_o_end; 73 74 extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL]; 75 76 SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler); 77 78 static register_t *linux_copyout_strings(struct image_params *imgp); 79 static int linux_elf_fixup(register_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 u_long 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_elf_fixup, todo); 97 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sigreturn, todo); 98 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sendsig, todo); 99 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vsyscall, todo); 100 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_install, todo); 101 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_deinstall, todo); 102 103 /* LINUXTODO: do we have traps to translate? */ 104 static int 105 linux_translate_traps(int signal, int trap_code) 106 { 107 108 LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code); 109 return (signal); 110 } 111 112 LINUX_VDSO_SYM_CHAR(linux_platform); 113 114 static int 115 linux_fetch_syscall_args(struct thread *td) 116 { 117 struct proc *p; 118 struct syscall_args *sa; 119 register_t *ap; 120 121 p = td->td_proc; 122 ap = td->td_frame->tf_x; 123 sa = &td->td_sa; 124 125 sa->code = td->td_frame->tf_x[8]; 126 /* LINUXTODO: generic syscall? */ 127 if (sa->code >= p->p_sysent->sv_size) 128 sa->callp = &p->p_sysent->sv_table[0]; 129 else 130 sa->callp = &p->p_sysent->sv_table[sa->code]; 131 132 sa->narg = sa->callp->sy_narg; 133 if (sa->narg > 8) 134 panic("ARM64TODO: Could we have more than 8 args?"); 135 memcpy(sa->args, ap, 8 * 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 } 148 149 static int 150 linux_elf_fixup(register_t **stack_base, struct image_params *imgp) 151 { 152 Elf_Auxargs *args; 153 Elf_Auxinfo *argarray, *pos; 154 Elf_Addr *auxbase, *base; 155 struct ps_strings *arginfo; 156 struct proc *p; 157 int error, issetugid; 158 159 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo); 160 p = imgp->proc; 161 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; 162 163 KASSERT(curthread->td_proc == imgp->proc, 164 ("unsafe linux_elf_fixup(), should be curproc")); 165 base = (Elf64_Addr *)*stack_base; 166 args = (Elf64_Auxargs *)imgp->auxargs; 167 /* Auxargs after argc, and NULL-terminated argv and envv lists. */ 168 auxbase = base + 1 + imgp->args->argc + 1 + imgp->args->envc + 1; 169 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, 170 M_WAITOK | M_ZERO); 171 172 issetugid = p->p_flag & P_SUGID ? 1 : 0; 173 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, 174 imgp->proc->p_sysent->sv_shared_page_base); 175 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_HWCAP */ 176 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature); 177 #endif 178 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); 179 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); 180 AUXARGS_ENTRY(pos, AT_PHENT, args->phent); 181 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); 182 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 183 AUXARGS_ENTRY(pos, AT_BASE, args->base); 184 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 185 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 186 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 187 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 188 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 189 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 190 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); 191 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_PLATFORM */ 192 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); 193 #endif 194 AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, imgp->canary); 195 if (imgp->execpathp != 0) 196 AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, imgp->execpathp); 197 if (args->execfd != -1) 198 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 199 AUXARGS_ENTRY(pos, AT_NULL, 0); 200 free(imgp->auxargs, M_TEMP); 201 imgp->auxargs = NULL; 202 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); 203 204 error = copyout(argarray, auxbase, sizeof(*argarray) * LINUX_AT_COUNT); 205 free(argarray, M_TEMP); 206 if (error != 0) 207 return (error); 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 register_t * 219 linux_copyout_strings(struct image_params *imgp) 220 { 221 char **vectp; 222 char *stringp, *destp; 223 register_t *stack_base; 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; 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 = (caddr_t)arginfo - SPARE_USRSPACE - 239 roundup(sizeof(canary), sizeof(char *)) - 240 roundup(execpath_len, sizeof(char *)) - 241 roundup(ARG_MAX - imgp->args->stringspace, sizeof(char *)); 242 243 if (execpath_len != 0) { 244 imgp->execpathp = (uintptr_t)arginfo - execpath_len; 245 copyout(imgp->execpath, (void *)imgp->execpathp, execpath_len); 246 } 247 248 /* Prepare the canary for SSP. */ 249 arc4rand(canary, sizeof(canary), 0); 250 imgp->canary = (uintptr_t)arginfo - 251 roundup(execpath_len, sizeof(char *)) - 252 roundup(sizeof(canary), sizeof(char *)); 253 copyout(canary, (void *)imgp->canary, sizeof(canary)); 254 255 vectp = (char **)destp; 256 if (imgp->auxargs) { 257 /* 258 * Allocate room on the stack for the ELF auxargs 259 * array. It has up to LINUX_AT_COUNT entries. 260 */ 261 vectp -= howmany(LINUX_AT_COUNT * sizeof(Elf64_Auxinfo), 262 sizeof(*vectp)); 263 } 264 265 /* 266 * Allocate room for argc and the argv[] and env vectors including the 267 * terminating NULL pointers. 268 */ 269 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1; 270 vectp = (char **)STACKALIGN(vectp); 271 272 /* vectp also becomes our initial stack base. */ 273 stack_base = (register_t *)vectp; 274 275 stringp = imgp->args->begin_argv; 276 argc = imgp->args->argc; 277 envc = imgp->args->envc; 278 279 /* Copy out strings - arguments and environment. */ 280 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace); 281 282 /* Fill in "ps_strings" struct for ps, w, etc. */ 283 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp); 284 suword(&arginfo->ps_nargvstr, argc); 285 286 suword(vectp++, argc); 287 /* Fill in argument portion of vector table. */ 288 for (; argc > 0; --argc) { 289 suword(vectp++, (long)(intptr_t)destp); 290 while (*stringp++ != 0) 291 destp++; 292 destp++; 293 } 294 295 /* A null vector table pointer separates the argp's from the envp's. */ 296 suword(vectp++, 0); 297 298 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp); 299 suword(&arginfo->ps_nenvstr, envc); 300 301 /* Fill in environment portion of vector table. */ 302 for (; envc > 0; --envc) { 303 suword(vectp++, (long)(intptr_t)destp); 304 while (*stringp++ != 0) 305 destp++; 306 destp++; 307 } 308 309 /* The end of the vector table is a null pointer. */ 310 suword(vectp, 0); 311 return (stack_base); 312 } 313 314 /* 315 * Reset registers to default values on exec. 316 */ 317 static void 318 linux_exec_setregs(struct thread *td, struct image_params *imgp, u_long stack) 319 { 320 struct trapframe *regs = td->td_frame; 321 322 /* LINUXTODO: validate */ 323 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo); 324 325 memset(regs, 0, sizeof(*regs)); 326 /* glibc start.S registers function pointer in x0 with atexit. */ 327 regs->tf_sp = stack; 328 #if 0 /* LINUXTODO: See if this is used. */ 329 regs->tf_lr = imgp->entry_addr; 330 #else 331 regs->tf_lr = 0xffffffffffffffff; 332 #endif 333 regs->tf_elr = imgp->entry_addr; 334 } 335 336 int 337 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 338 { 339 340 /* LINUXTODO: implement */ 341 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo); 342 return (EDOOFUS); 343 } 344 345 static void 346 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 347 { 348 349 /* LINUXTODO: implement */ 350 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo); 351 } 352 353 static int 354 linux_vsyscall(struct thread *td) 355 { 356 357 /* LINUXTODO: implement */ 358 LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo); 359 return (EDOOFUS); 360 } 361 362 struct sysentvec elf_linux_sysvec = { 363 .sv_size = LINUX_SYS_MAXSYSCALL, 364 .sv_table = linux_sysent, 365 .sv_errsize = ELAST + 1, 366 .sv_errtbl = linux_errtbl, 367 .sv_transtrap = linux_translate_traps, 368 .sv_fixup = linux_elf_fixup, 369 .sv_sendsig = linux_rt_sendsig, 370 .sv_sigcode = &_binary_linux_locore_o_start, 371 .sv_szsigcode = &linux_szsigcode, 372 .sv_name = "Linux ELF64", 373 .sv_coredump = elf64_coredump, 374 .sv_imgact_try = linux_exec_imgact_try, 375 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 376 .sv_minuser = VM_MIN_ADDRESS, 377 .sv_maxuser = VM_MAXUSER_ADDRESS, 378 .sv_usrstack = USRSTACK, 379 .sv_psstrings = PS_STRINGS, /* XXX */ 380 .sv_stackprot = VM_PROT_ALL, /* XXX */ 381 .sv_copyout_strings = linux_copyout_strings, 382 .sv_setregs = linux_exec_setregs, 383 .sv_fixlimit = NULL, 384 .sv_maxssiz = NULL, 385 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP, 386 .sv_set_syscall_retval = linux_set_syscall_retval, 387 .sv_fetch_syscall_args = linux_fetch_syscall_args, 388 .sv_syscallnames = NULL, 389 .sv_shared_page_base = SHAREDPAGE, 390 .sv_shared_page_len = PAGE_SIZE, 391 .sv_schedtail = linux_schedtail, 392 .sv_thread_detach = linux_thread_detach, 393 .sv_trap = linux_vsyscall, 394 }; 395 396 static void 397 linux_vdso_install(const void *param) 398 { 399 400 linux_szsigcode = (&_binary_linux_locore_o_end - 401 &_binary_linux_locore_o_start); 402 403 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len) 404 panic("invalid Linux VDSO size\n"); 405 406 __elfN(linux_vdso_fixup)(&elf_linux_sysvec); 407 408 linux_shared_page_obj = __elfN(linux_shared_page_init) 409 (&linux_shared_page_mapping); 410 411 __elfN(linux_vdso_reloc)(&elf_linux_sysvec); 412 413 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode, 414 linux_szsigcode); 415 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj; 416 417 printf("LINUXTODO: %s: fix linux_kplatform\n", __func__); 418 #if 0 419 linux_kplatform = linux_shared_page_mapping + 420 (linux_platform - (caddr_t)elf_linux_sysvec.sv_shared_page_base); 421 #else 422 linux_kplatform = "arm64"; 423 #endif 424 } 425 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY, 426 linux_vdso_install, NULL); 427 428 static void 429 linux_vdso_deinstall(const void *param) 430 { 431 432 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo); 433 __elfN(linux_shared_page_fini)(linux_shared_page_obj); 434 } 435 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, 436 linux_vdso_deinstall, NULL); 437 438 static char GNU_ABI_VENDOR[] = "GNU"; 439 static int GNU_ABI_LINUX = 0; 440 441 /* LINUXTODO: deduplicate */ 442 static bool 443 linux_trans_osrel(const Elf_Note *note, int32_t *osrel) 444 { 445 const Elf32_Word *desc; 446 uintptr_t p; 447 448 p = (uintptr_t)(note + 1); 449 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 450 451 desc = (const Elf32_Word *)p; 452 if (desc[0] != GNU_ABI_LINUX) 453 return (false); 454 455 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); 456 return (true); 457 } 458 459 static Elf_Brandnote linux64_brandnote = { 460 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 461 .hdr.n_descsz = 16, 462 .hdr.n_type = 1, 463 .vendor = GNU_ABI_VENDOR, 464 .flags = BN_TRANSLATE_OSREL, 465 .trans_osrel = linux_trans_osrel 466 }; 467 468 static Elf64_Brandinfo linux_glibc2brand = { 469 .brand = ELFOSABI_LINUX, 470 .machine = EM_AARCH64, 471 .compat_3_brand = "Linux", 472 .emul_path = "/compat/linux", 473 .interp_path = "/lib64/ld-linux-x86-64.so.2", 474 .sysvec = &elf_linux_sysvec, 475 .interp_newpath = NULL, 476 .brand_note = &linux64_brandnote, 477 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 478 }; 479 480 Elf64_Brandinfo *linux_brandlist[] = { 481 &linux_glibc2brand, 482 NULL 483 }; 484 485 static int 486 linux64_elf_modevent(module_t mod, int type, void *data) 487 { 488 Elf64_Brandinfo **brandinfo; 489 struct linux_ioctl_handler**lihp; 490 int error; 491 492 error = 0; 493 switch(type) { 494 case MOD_LOAD: 495 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 496 ++brandinfo) 497 if (elf64_insert_brand_entry(*brandinfo) < 0) 498 error = EINVAL; 499 if (error == 0) { 500 SET_FOREACH(lihp, linux_ioctl_handler_set) 501 linux_ioctl_register_handler(*lihp); 502 LIST_INIT(&futex_list); 503 mtx_init(&futex_mtx, "ftllk64", NULL, MTX_DEF); 504 stclohz = (stathz ? stathz : hz); 505 if (bootverbose) 506 printf("Linux arm64 ELF exec handler installed\n"); 507 } 508 break; 509 case MOD_UNLOAD: 510 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 511 ++brandinfo) 512 if (elf64_brand_inuse(*brandinfo)) 513 error = EBUSY; 514 if (error == 0) { 515 for (brandinfo = &linux_brandlist[0]; 516 *brandinfo != NULL; ++brandinfo) 517 if (elf64_remove_brand_entry(*brandinfo) < 0) 518 error = EINVAL; 519 } 520 if (error == 0) { 521 SET_FOREACH(lihp, linux_ioctl_handler_set) 522 linux_ioctl_unregister_handler(*lihp); 523 mtx_destroy(&futex_mtx); 524 if (bootverbose) 525 printf("Linux ELF exec handler removed\n"); 526 } else 527 printf("Could not deinstall ELF interpreter entry\n"); 528 break; 529 default: 530 return (EOPNOTSUPP); 531 } 532 return (error); 533 } 534 535 static moduledata_t linux64_elf_mod = { 536 "linux64elf", 537 linux64_elf_modevent, 538 0 539 }; 540 541 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 542 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); 543 FEATURE(linux64, "AArch64 Linux 64bit support"); 544