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 register_t **stack_base); 74 static int linux_elf_fixup(register_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 u_long 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, u_long *base) 147 { 148 Elf_Auxargs *args; 149 Elf_Auxinfo *argarray, *pos; 150 u_long auxlen; 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(pos, LINUX_AT_RANDOM, imgp->canary); 184 if (imgp->execpathp != 0) 185 AUXARGS_ENTRY(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 auxlen = sizeof(*argarray) * (pos - argarray); 194 *base -= auxlen; 195 error = copyout(argarray, (void *)*base, auxlen); 196 free(argarray, M_TEMP); 197 return (error); 198 } 199 200 static int 201 linux_elf_fixup(register_t **stack_base, struct image_params *imgp) 202 { 203 204 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo); 205 206 return (0); 207 } 208 209 /* 210 * Copy strings out to the new process address space, constructing new arg 211 * and env vector tables. Return a pointer to the base so that it can be used 212 * as the initial stack pointer. 213 * LINUXTODO: deduplicate against other linuxulator archs 214 */ 215 static int 216 linux_copyout_strings(struct image_params *imgp, register_t **stack_base) 217 { 218 char **vectp; 219 char *stringp, *destp; 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 = (caddr_t)arginfo - SPARE_USRSPACE - 235 roundup(sizeof(canary), sizeof(char *)) - 236 roundup(execpath_len, sizeof(char *)) - 237 roundup(ARG_MAX - imgp->args->stringspace, sizeof(char *)); 238 239 if (execpath_len != 0) { 240 imgp->execpathp = (uintptr_t)arginfo - execpath_len; 241 error = copyout(imgp->execpath, (void *)imgp->execpathp, 242 execpath_len); 243 if (error != 0) 244 return (error); 245 } 246 247 /* Prepare the canary for SSP. */ 248 arc4rand(canary, sizeof(canary), 0); 249 imgp->canary = (uintptr_t)arginfo - 250 roundup(execpath_len, sizeof(char *)) - 251 roundup(sizeof(canary), sizeof(char *)); 252 error = copyout(canary, (void *)imgp->canary, sizeof(canary)); 253 if (error != 0) 254 return (error); 255 256 vectp = (char **)destp; 257 if (imgp->auxargs) { 258 error = imgp->sysent->sv_copyout_auxargs(imgp, 259 (u_long *)&vectp); 260 if (error != 0) 261 return (error); 262 } 263 264 /* 265 * Allocate room for argc and the argv[] and env vectors including the 266 * terminating NULL pointers. 267 */ 268 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1; 269 vectp = (char **)STACKALIGN(vectp); 270 271 /* vectp also becomes our initial stack base. */ 272 *stack_base = (register_t *)vectp; 273 274 stringp = imgp->args->begin_argv; 275 argc = imgp->args->argc; 276 envc = imgp->args->envc; 277 278 /* Copy out strings - arguments and environment. */ 279 error = copyout(stringp, destp, ARG_MAX - imgp->args->stringspace); 280 if (error != 0) 281 return (error); 282 283 /* Fill in "ps_strings" struct for ps, w, etc. */ 284 if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || 285 suword(&arginfo->ps_nargvstr, argc) != 0) 286 return (EFAULT); 287 288 if (suword(vectp++, argc) != 0) 289 return (EFAULT); 290 291 /* Fill in argument portion of vector table. */ 292 for (; argc > 0; --argc) { 293 if (suword(vectp++, (long)(intptr_t)destp) != 0) 294 return (EFAULT); 295 while (*stringp++ != 0) 296 destp++; 297 destp++; 298 } 299 300 /* A null vector table pointer separates the argp's from the envp's. */ 301 if (suword(vectp++, 0) != 0) 302 return (EFAULT); 303 304 if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || 305 suword(&arginfo->ps_nenvstr, envc) != 0) 306 return (EFAULT); 307 308 /* Fill in environment portion of vector table. */ 309 for (; envc > 0; --envc) { 310 if (suword(vectp++, (long)(intptr_t)destp) != 0) 311 return (EFAULT); 312 while (*stringp++ != 0) 313 destp++; 314 destp++; 315 } 316 317 /* The end of the vector table is a null pointer. */ 318 if (suword(vectp, 0) != 0) 319 return (EFAULT); 320 321 return (0); 322 } 323 324 /* 325 * Reset registers to default values on exec. 326 */ 327 static void 328 linux_exec_setregs(struct thread *td, struct image_params *imgp, u_long stack) 329 { 330 struct trapframe *regs = td->td_frame; 331 332 /* LINUXTODO: validate */ 333 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo); 334 335 memset(regs, 0, sizeof(*regs)); 336 /* glibc start.S registers function pointer in x0 with atexit. */ 337 regs->tf_sp = stack; 338 #if 0 /* LINUXTODO: See if this is used. */ 339 regs->tf_lr = imgp->entry_addr; 340 #else 341 regs->tf_lr = 0xffffffffffffffff; 342 #endif 343 regs->tf_elr = imgp->entry_addr; 344 } 345 346 int 347 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 348 { 349 350 /* LINUXTODO: implement */ 351 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo); 352 return (EDOOFUS); 353 } 354 355 static void 356 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 357 { 358 359 /* LINUXTODO: implement */ 360 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo); 361 } 362 363 static int 364 linux_vsyscall(struct thread *td) 365 { 366 367 /* LINUXTODO: implement */ 368 LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo); 369 return (EDOOFUS); 370 } 371 372 struct sysentvec elf_linux_sysvec = { 373 .sv_size = LINUX_SYS_MAXSYSCALL, 374 .sv_table = linux_sysent, 375 .sv_errsize = ELAST + 1, 376 .sv_errtbl = linux_errtbl, 377 .sv_transtrap = linux_translate_traps, 378 .sv_fixup = linux_elf_fixup, 379 .sv_sendsig = linux_rt_sendsig, 380 .sv_sigcode = &_binary_linux_locore_o_start, 381 .sv_szsigcode = &linux_szsigcode, 382 .sv_name = "Linux ELF64", 383 .sv_coredump = elf64_coredump, 384 .sv_imgact_try = linux_exec_imgact_try, 385 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 386 .sv_minuser = VM_MIN_ADDRESS, 387 .sv_maxuser = VM_MAXUSER_ADDRESS, 388 .sv_usrstack = USRSTACK, 389 .sv_psstrings = PS_STRINGS, /* XXX */ 390 .sv_stackprot = VM_PROT_READ | VM_PROT_WRITE, 391 .sv_copyout_auxargs = linux_copyout_auxargs, 392 .sv_copyout_strings = linux_copyout_strings, 393 .sv_setregs = linux_exec_setregs, 394 .sv_fixlimit = NULL, 395 .sv_maxssiz = NULL, 396 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP, 397 .sv_set_syscall_retval = linux_set_syscall_retval, 398 .sv_fetch_syscall_args = linux_fetch_syscall_args, 399 .sv_syscallnames = NULL, 400 .sv_shared_page_base = SHAREDPAGE, 401 .sv_shared_page_len = PAGE_SIZE, 402 .sv_schedtail = linux_schedtail, 403 .sv_thread_detach = linux_thread_detach, 404 .sv_trap = linux_vsyscall, 405 }; 406 407 static void 408 linux_vdso_install(const void *param) 409 { 410 411 linux_szsigcode = (&_binary_linux_locore_o_end - 412 &_binary_linux_locore_o_start); 413 414 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len) 415 panic("invalid Linux VDSO size\n"); 416 417 __elfN(linux_vdso_fixup)(&elf_linux_sysvec); 418 419 linux_shared_page_obj = __elfN(linux_shared_page_init) 420 (&linux_shared_page_mapping); 421 422 __elfN(linux_vdso_reloc)(&elf_linux_sysvec); 423 424 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode, 425 linux_szsigcode); 426 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj; 427 428 printf("LINUXTODO: %s: fix linux_kplatform\n", __func__); 429 #if 0 430 linux_kplatform = linux_shared_page_mapping + 431 (linux_platform - (caddr_t)elf_linux_sysvec.sv_shared_page_base); 432 #else 433 linux_kplatform = "arm64"; 434 #endif 435 } 436 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY, 437 linux_vdso_install, NULL); 438 439 static void 440 linux_vdso_deinstall(const void *param) 441 { 442 443 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo); 444 __elfN(linux_shared_page_fini)(linux_shared_page_obj); 445 } 446 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, 447 linux_vdso_deinstall, NULL); 448 449 static char GNU_ABI_VENDOR[] = "GNU"; 450 static int GNU_ABI_LINUX = 0; 451 452 /* LINUXTODO: deduplicate */ 453 static bool 454 linux_trans_osrel(const Elf_Note *note, int32_t *osrel) 455 { 456 const Elf32_Word *desc; 457 uintptr_t p; 458 459 p = (uintptr_t)(note + 1); 460 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 461 462 desc = (const Elf32_Word *)p; 463 if (desc[0] != GNU_ABI_LINUX) 464 return (false); 465 466 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); 467 return (true); 468 } 469 470 static Elf_Brandnote linux64_brandnote = { 471 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 472 .hdr.n_descsz = 16, 473 .hdr.n_type = 1, 474 .vendor = GNU_ABI_VENDOR, 475 .flags = BN_TRANSLATE_OSREL, 476 .trans_osrel = linux_trans_osrel 477 }; 478 479 static Elf64_Brandinfo linux_glibc2brand = { 480 .brand = ELFOSABI_LINUX, 481 .machine = EM_AARCH64, 482 .compat_3_brand = "Linux", 483 .emul_path = "/compat/linux", 484 .interp_path = "/lib64/ld-linux-x86-64.so.2", 485 .sysvec = &elf_linux_sysvec, 486 .interp_newpath = NULL, 487 .brand_note = &linux64_brandnote, 488 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 489 }; 490 491 Elf64_Brandinfo *linux_brandlist[] = { 492 &linux_glibc2brand, 493 NULL 494 }; 495 496 static int 497 linux64_elf_modevent(module_t mod, int type, void *data) 498 { 499 Elf64_Brandinfo **brandinfo; 500 struct linux_ioctl_handler**lihp; 501 int error; 502 503 error = 0; 504 switch(type) { 505 case MOD_LOAD: 506 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 507 ++brandinfo) 508 if (elf64_insert_brand_entry(*brandinfo) < 0) 509 error = EINVAL; 510 if (error == 0) { 511 SET_FOREACH(lihp, linux_ioctl_handler_set) 512 linux_ioctl_register_handler(*lihp); 513 stclohz = (stathz ? stathz : hz); 514 if (bootverbose) 515 printf("Linux arm64 ELF exec handler installed\n"); 516 } 517 break; 518 case MOD_UNLOAD: 519 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 520 ++brandinfo) 521 if (elf64_brand_inuse(*brandinfo)) 522 error = EBUSY; 523 if (error == 0) { 524 for (brandinfo = &linux_brandlist[0]; 525 *brandinfo != NULL; ++brandinfo) 526 if (elf64_remove_brand_entry(*brandinfo) < 0) 527 error = EINVAL; 528 } 529 if (error == 0) { 530 SET_FOREACH(lihp, linux_ioctl_handler_set) 531 linux_ioctl_unregister_handler(*lihp); 532 if (bootverbose) 533 printf("Linux ELF exec handler removed\n"); 534 } else 535 printf("Could not deinstall ELF interpreter entry\n"); 536 break; 537 default: 538 return (EOPNOTSUPP); 539 } 540 return (error); 541 } 542 543 static moduledata_t linux64_elf_mod = { 544 "linux64elf", 545 linux64_elf_modevent, 546 0 547 }; 548 549 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 550 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); 551 FEATURE(linux64, "AArch64 Linux 64bit support"); 552