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 #include <machine/md_var.h> 61 62 #ifdef VFP 63 #include <machine/vfp.h> 64 #endif 65 66 MODULE_VERSION(linux64elf, 1); 67 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 int linux_copyout_strings(struct image_params *imgp, 79 uintptr_t *stack_base); 80 static int linux_elf_fixup(uintptr_t *stack_base, 81 struct image_params *iparams); 82 static bool linux_trans_osrel(const Elf_Note *note, int32_t *osrel); 83 static void linux_vdso_install(const void *param); 84 static void linux_vdso_deinstall(const void *param); 85 static void linux_set_syscall_retval(struct thread *td, int error); 86 static int linux_fetch_syscall_args(struct thread *td); 87 static void linux_exec_setregs(struct thread *td, struct image_params *imgp, 88 uintptr_t stack); 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_copyout_auxargs, todo); 97 LIN_SDT_PROBE_DEFINE0(sysvec, linux_elf_fixup, todo); 98 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sigreturn, todo); 99 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sendsig, 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 sa->original_code = sa->code; 127 /* LINUXTODO: generic syscall? */ 128 if (sa->code >= p->p_sysent->sv_size) 129 sa->callp = &p->p_sysent->sv_table[0]; 130 else 131 sa->callp = &p->p_sysent->sv_table[sa->code]; 132 133 if (sa->callp->sy_narg > MAXARGS) 134 panic("ARM64TODO: Could we have more than %d args?", MAXARGS); 135 memcpy(sa->args, ap, MAXARGS * 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 if (__predict_false(error != 0)) { 149 if (error != ERESTART && error != EJUSTRETURN) 150 td->td_frame->tf_x[0] = bsd_to_linux_errno(error); 151 } 152 } 153 154 static int 155 linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) 156 { 157 Elf_Auxargs *args; 158 Elf_Auxinfo *argarray, *pos; 159 struct proc *p; 160 int error, issetugid; 161 162 LIN_SDT_PROBE0(sysvec, linux_copyout_auxargs, todo); 163 p = imgp->proc; 164 165 args = (Elf64_Auxargs *)imgp->auxargs; 166 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, 167 M_WAITOK | M_ZERO); 168 169 issetugid = p->p_flag & P_SUGID ? 1 : 0; 170 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, 171 imgp->proc->p_sysent->sv_shared_page_base); 172 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, *imgp->sysent->sv_hwcap); 173 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 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_BASE, args->base); 179 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 180 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 181 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 182 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 183 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 184 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 185 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); 186 AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary); 187 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, *imgp->sysent->sv_hwcap2); 188 if (imgp->execpathp != 0) 189 AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp); 190 if (args->execfd != -1) 191 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 192 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); 193 AUXARGS_ENTRY(pos, AT_NULL, 0); 194 195 free(imgp->auxargs, M_TEMP); 196 imgp->auxargs = NULL; 197 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); 198 199 error = copyout(argarray, (void *)base, 200 sizeof(*argarray) * LINUX_AT_COUNT); 201 free(argarray, M_TEMP); 202 return (error); 203 } 204 205 static int 206 linux_elf_fixup(uintptr_t *stack_base, struct image_params *imgp) 207 { 208 209 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo); 210 211 return (0); 212 } 213 214 /* 215 * Copy strings out to the new process address space, constructing new arg 216 * and env vector tables. Return a pointer to the base so that it can be used 217 * as the initial stack pointer. 218 * LINUXTODO: deduplicate against other linuxulator archs 219 */ 220 static int 221 linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) 222 { 223 char **vectp; 224 char *stringp; 225 uintptr_t destp, ustringp; 226 struct ps_strings *arginfo; 227 char canary[LINUX_AT_RANDOM_LEN]; 228 size_t execpath_len; 229 struct proc *p; 230 int argc, envc, error; 231 232 /* Calculate string base and vector table pointers. */ 233 if (imgp->execpath != NULL && imgp->auxargs != NULL) 234 execpath_len = strlen(imgp->execpath) + 1; 235 else 236 execpath_len = 0; 237 238 p = imgp->proc; 239 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; 240 destp = (uintptr_t)arginfo; 241 242 if (execpath_len != 0) { 243 destp -= execpath_len; 244 destp = rounddown2(destp, sizeof(void *)); 245 imgp->execpathp = (void *)destp; 246 error = copyout(imgp->execpath, imgp->execpathp, execpath_len); 247 if (error != 0) 248 return (error); 249 } 250 251 /* Prepare the canary for SSP. */ 252 arc4rand(canary, sizeof(canary), 0); 253 destp -= roundup(sizeof(canary), sizeof(void *)); 254 imgp->canary = (void *)destp; 255 error = copyout(canary, imgp->canary, sizeof(canary)); 256 if (error != 0) 257 return (error); 258 259 /* Allocate room for the argument and environment strings. */ 260 destp -= ARG_MAX - imgp->args->stringspace; 261 destp = rounddown2(destp, sizeof(void *)); 262 ustringp = destp; 263 264 if (imgp->auxargs) { 265 /* 266 * Allocate room on the stack for the ELF auxargs 267 * array. It has up to LINUX_AT_COUNT entries. 268 */ 269 destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo); 270 destp = rounddown2(destp, sizeof(void *)); 271 } 272 273 vectp = (char **)destp; 274 275 /* 276 * Allocate room for argc and the argv[] and env vectors including the 277 * terminating NULL pointers. 278 */ 279 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1; 280 vectp = (char **)STACKALIGN(vectp); 281 282 /* vectp also becomes our initial stack base. */ 283 *stack_base = (uintptr_t)vectp; 284 285 stringp = imgp->args->begin_argv; 286 argc = imgp->args->argc; 287 envc = imgp->args->envc; 288 289 /* Copy out strings - arguments and environment. */ 290 error = copyout(stringp, (void *)ustringp, 291 ARG_MAX - imgp->args->stringspace); 292 if (error != 0) 293 return (error); 294 295 /* Fill in "ps_strings" struct for ps, w, etc. */ 296 if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || 297 suword(&arginfo->ps_nargvstr, argc) != 0) 298 return (EFAULT); 299 300 if (suword(vectp++, argc) != 0) 301 return (EFAULT); 302 303 /* Fill in argument portion of vector table. */ 304 for (; argc > 0; --argc) { 305 if (suword(vectp++, ustringp) != 0) 306 return (EFAULT); 307 while (*stringp++ != 0) 308 ustringp++; 309 ustringp++; 310 } 311 312 /* A null vector table pointer separates the argp's from the envp's. */ 313 if (suword(vectp++, 0) != 0) 314 return (EFAULT); 315 316 if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || 317 suword(&arginfo->ps_nenvstr, envc) != 0) 318 return (EFAULT); 319 320 /* Fill in environment portion of vector table. */ 321 for (; envc > 0; --envc) { 322 if (suword(vectp++, ustringp) != 0) 323 return (EFAULT); 324 while (*stringp++ != 0) 325 ustringp++; 326 ustringp++; 327 } 328 329 /* The end of the vector table is a null pointer. */ 330 if (suword(vectp, 0) != 0) 331 return (EFAULT); 332 333 if (imgp->auxargs) { 334 vectp++; 335 error = imgp->sysent->sv_copyout_auxargs(imgp, 336 (uintptr_t)vectp); 337 if (error != 0) 338 return (error); 339 } 340 341 return (0); 342 } 343 344 /* 345 * Reset registers to default values on exec. 346 */ 347 static void 348 linux_exec_setregs(struct thread *td, struct image_params *imgp, 349 uintptr_t stack) 350 { 351 struct trapframe *regs = td->td_frame; 352 struct pcb *pcb = td->td_pcb; 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 pcb->pcb_tpidr_el0 = 0; 368 pcb->pcb_tpidrro_el0 = 0; 369 WRITE_SPECIALREG(tpidrro_el0, 0); 370 WRITE_SPECIALREG(tpidr_el0, 0); 371 372 #ifdef VFP 373 vfp_reset_state(td, pcb); 374 #endif 375 376 /* 377 * Clear debug register state. It is not applicable to the new process. 378 */ 379 bzero(&pcb->pcb_dbg_regs, sizeof(pcb->pcb_dbg_regs)); 380 } 381 382 int 383 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 384 { 385 386 /* LINUXTODO: implement */ 387 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo); 388 return (EDOOFUS); 389 } 390 391 static void 392 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 393 { 394 395 /* LINUXTODO: implement */ 396 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo); 397 } 398 399 struct sysentvec elf_linux_sysvec = { 400 .sv_size = LINUX_SYS_MAXSYSCALL, 401 .sv_table = linux_sysent, 402 .sv_transtrap = linux_translate_traps, 403 .sv_fixup = linux_elf_fixup, 404 .sv_sendsig = linux_rt_sendsig, 405 .sv_sigcode = &_binary_linux_locore_o_start, 406 .sv_szsigcode = &linux_szsigcode, 407 .sv_name = "Linux ELF64", 408 .sv_coredump = elf64_coredump, 409 .sv_elf_core_osabi = ELFOSABI_NONE, 410 .sv_elf_core_abi_vendor = LINUX_ABI_VENDOR, 411 .sv_elf_core_prepare_notes = linux64_prepare_notes, 412 .sv_imgact_try = linux_exec_imgact_try, 413 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 414 .sv_minuser = VM_MIN_ADDRESS, 415 .sv_maxuser = VM_MAXUSER_ADDRESS, 416 .sv_usrstack = USRSTACK, 417 .sv_psstrings = PS_STRINGS, /* XXX */ 418 .sv_stackprot = VM_PROT_READ | VM_PROT_WRITE, 419 .sv_copyout_auxargs = linux_copyout_auxargs, 420 .sv_copyout_strings = linux_copyout_strings, 421 .sv_setregs = linux_exec_setregs, 422 .sv_fixlimit = NULL, 423 .sv_maxssiz = NULL, 424 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP | SV_SIG_DISCIGN | 425 SV_SIG_WAITNDQ, 426 .sv_set_syscall_retval = linux_set_syscall_retval, 427 .sv_fetch_syscall_args = linux_fetch_syscall_args, 428 .sv_syscallnames = NULL, 429 .sv_shared_page_base = SHAREDPAGE, 430 .sv_shared_page_len = PAGE_SIZE, 431 .sv_schedtail = linux_schedtail, 432 .sv_thread_detach = linux_thread_detach, 433 .sv_trap = NULL, 434 .sv_hwcap = &elf_hwcap, 435 .sv_hwcap2 = &elf_hwcap2, 436 .sv_onexec = linux_on_exec, 437 .sv_onexit = linux_on_exit, 438 .sv_ontdexit = linux_thread_dtor, 439 .sv_setid_allowed = &linux_setid_allowed_query, 440 }; 441 442 static void 443 linux_vdso_install(const void *param) 444 { 445 446 linux_szsigcode = (&_binary_linux_locore_o_end - 447 &_binary_linux_locore_o_start); 448 449 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len) 450 panic("invalid Linux VDSO size\n"); 451 452 __elfN(linux_vdso_fixup)(&elf_linux_sysvec); 453 454 linux_shared_page_obj = __elfN(linux_shared_page_init) 455 (&linux_shared_page_mapping); 456 457 __elfN(linux_vdso_reloc)(&elf_linux_sysvec); 458 459 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode, 460 linux_szsigcode); 461 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj; 462 } 463 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY, 464 linux_vdso_install, NULL); 465 466 static void 467 linux_vdso_deinstall(const void *param) 468 { 469 470 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo); 471 __elfN(linux_shared_page_fini)(linux_shared_page_obj, 472 linux_shared_page_mapping); 473 } 474 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, 475 linux_vdso_deinstall, NULL); 476 477 static char GNU_ABI_VENDOR[] = "GNU"; 478 static int GNU_ABI_LINUX = 0; 479 480 /* LINUXTODO: deduplicate */ 481 static bool 482 linux_trans_osrel(const Elf_Note *note, int32_t *osrel) 483 { 484 const Elf32_Word *desc; 485 uintptr_t p; 486 487 p = (uintptr_t)(note + 1); 488 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 489 490 desc = (const Elf32_Word *)p; 491 if (desc[0] != GNU_ABI_LINUX) 492 return (false); 493 494 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); 495 return (true); 496 } 497 498 static Elf_Brandnote linux64_brandnote = { 499 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 500 .hdr.n_descsz = 16, 501 .hdr.n_type = 1, 502 .vendor = GNU_ABI_VENDOR, 503 .flags = BN_TRANSLATE_OSREL, 504 .trans_osrel = linux_trans_osrel 505 }; 506 507 static Elf64_Brandinfo linux_glibc2brand = { 508 .brand = ELFOSABI_LINUX, 509 .machine = EM_AARCH64, 510 .compat_3_brand = "Linux", 511 .emul_path = linux_emul_path, 512 .interp_path = "/lib64/ld-linux-x86-64.so.2", 513 .sysvec = &elf_linux_sysvec, 514 .interp_newpath = NULL, 515 .brand_note = &linux64_brandnote, 516 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 517 }; 518 519 Elf64_Brandinfo *linux_brandlist[] = { 520 &linux_glibc2brand, 521 NULL 522 }; 523 524 static int 525 linux64_elf_modevent(module_t mod, int type, void *data) 526 { 527 Elf64_Brandinfo **brandinfo; 528 struct linux_ioctl_handler**lihp; 529 int error; 530 531 error = 0; 532 switch(type) { 533 case MOD_LOAD: 534 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 535 ++brandinfo) 536 if (elf64_insert_brand_entry(*brandinfo) < 0) 537 error = EINVAL; 538 if (error == 0) { 539 SET_FOREACH(lihp, linux_ioctl_handler_set) 540 linux_ioctl_register_handler(*lihp); 541 stclohz = (stathz ? stathz : hz); 542 if (bootverbose) 543 printf("Linux arm64 ELF exec handler installed\n"); 544 } 545 break; 546 case MOD_UNLOAD: 547 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 548 ++brandinfo) 549 if (elf64_brand_inuse(*brandinfo)) 550 error = EBUSY; 551 if (error == 0) { 552 for (brandinfo = &linux_brandlist[0]; 553 *brandinfo != NULL; ++brandinfo) 554 if (elf64_remove_brand_entry(*brandinfo) < 0) 555 error = EINVAL; 556 } 557 if (error == 0) { 558 SET_FOREACH(lihp, linux_ioctl_handler_set) 559 linux_ioctl_unregister_handler(*lihp); 560 if (bootverbose) 561 printf("Linux ELF exec handler removed\n"); 562 } else 563 printf("Could not deinstall ELF interpreter entry\n"); 564 break; 565 default: 566 return (EOPNOTSUPP); 567 } 568 return (error); 569 } 570 571 static moduledata_t linux64_elf_mod = { 572 "linux64elf", 573 linux64_elf_modevent, 574 0 575 }; 576 577 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 578 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); 579 FEATURE(linux64, "AArch64 Linux 64bit support"); 580