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 static int linux_vsyscall(struct thread *td); 90 91 /* DTrace init */ 92 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE); 93 94 /* DTrace probes */ 95 LIN_SDT_PROBE_DEFINE2(sysvec, linux_translate_traps, todo, "int", "int"); 96 LIN_SDT_PROBE_DEFINE0(sysvec, linux_exec_setregs, todo); 97 LIN_SDT_PROBE_DEFINE0(sysvec, linux_copyout_auxargs, todo); 98 LIN_SDT_PROBE_DEFINE0(sysvec, linux_elf_fixup, todo); 99 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sigreturn, todo); 100 LIN_SDT_PROBE_DEFINE0(sysvec, linux_rt_sendsig, todo); 101 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vsyscall, todo); 102 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_install, todo); 103 LIN_SDT_PROBE_DEFINE0(sysvec, linux_vdso_deinstall, todo); 104 105 /* LINUXTODO: do we have traps to translate? */ 106 static int 107 linux_translate_traps(int signal, int trap_code) 108 { 109 110 LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code); 111 return (signal); 112 } 113 114 LINUX_VDSO_SYM_CHAR(linux_platform); 115 116 static int 117 linux_fetch_syscall_args(struct thread *td) 118 { 119 struct proc *p; 120 struct syscall_args *sa; 121 register_t *ap; 122 123 p = td->td_proc; 124 ap = td->td_frame->tf_x; 125 sa = &td->td_sa; 126 127 sa->code = td->td_frame->tf_x[8]; 128 /* LINUXTODO: generic syscall? */ 129 if (sa->code >= p->p_sysent->sv_size) 130 sa->callp = &p->p_sysent->sv_table[0]; 131 else 132 sa->callp = &p->p_sysent->sv_table[sa->code]; 133 134 if (sa->callp->sy_narg > MAXARGS) 135 panic("ARM64TODO: Could we have more than %d args?", MAXARGS); 136 memcpy(sa->args, ap, MAXARGS * sizeof(register_t)); 137 138 td->td_retval[0] = 0; 139 return (0); 140 } 141 142 static void 143 linux_set_syscall_retval(struct thread *td, int error) 144 { 145 146 td->td_retval[1] = td->td_frame->tf_x[1]; 147 cpu_set_syscall_retval(td, error); 148 149 if (__predict_false(error != 0)) { 150 if (error != ERESTART && error != EJUSTRETURN) 151 td->td_frame->tf_x[0] = bsd_to_linux_errno(error); 152 } 153 } 154 155 static int 156 linux_copyout_auxargs(struct image_params *imgp, uintptr_t base) 157 { 158 Elf_Auxargs *args; 159 Elf_Auxinfo *argarray, *pos; 160 struct proc *p; 161 int error, issetugid; 162 163 LIN_SDT_PROBE0(sysvec, linux_copyout_auxargs, todo); 164 p = imgp->proc; 165 166 args = (Elf64_Auxargs *)imgp->auxargs; 167 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP, 168 M_WAITOK | M_ZERO); 169 170 issetugid = p->p_flag & P_SUGID ? 1 : 0; 171 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR, 172 imgp->proc->p_sysent->sv_shared_page_base); 173 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, *imgp->sysent->sv_hwcap); 174 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 175 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz); 176 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); 177 AUXARGS_ENTRY(pos, AT_PHENT, args->phent); 178 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); 179 AUXARGS_ENTRY(pos, AT_BASE, args->base); 180 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 181 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 182 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid); 183 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid); 184 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid); 185 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid); 186 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid); 187 AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary); 188 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP2, *imgp->sysent->sv_hwcap2); 189 if (imgp->execpathp != 0) 190 AUXARGS_ENTRY_PTR(pos, LINUX_AT_EXECFN, imgp->execpathp); 191 if (args->execfd != -1) 192 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 193 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform)); 194 AUXARGS_ENTRY(pos, AT_NULL, 0); 195 196 free(imgp->auxargs, M_TEMP); 197 imgp->auxargs = NULL; 198 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs")); 199 200 error = copyout(argarray, (void *)base, 201 sizeof(*argarray) * LINUX_AT_COUNT); 202 free(argarray, M_TEMP); 203 return (error); 204 } 205 206 static int 207 linux_elf_fixup(uintptr_t *stack_base, struct image_params *imgp) 208 { 209 210 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo); 211 212 return (0); 213 } 214 215 /* 216 * Copy strings out to the new process address space, constructing new arg 217 * and env vector tables. Return a pointer to the base so that it can be used 218 * as the initial stack pointer. 219 * LINUXTODO: deduplicate against other linuxulator archs 220 */ 221 static int 222 linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base) 223 { 224 char **vectp; 225 char *stringp; 226 uintptr_t destp, ustringp; 227 struct ps_strings *arginfo; 228 char canary[LINUX_AT_RANDOM_LEN]; 229 size_t execpath_len; 230 struct proc *p; 231 int argc, envc, error; 232 233 /* Calculate string base and vector table pointers. */ 234 if (imgp->execpath != NULL && imgp->auxargs != NULL) 235 execpath_len = strlen(imgp->execpath) + 1; 236 else 237 execpath_len = 0; 238 239 p = imgp->proc; 240 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings; 241 destp = (uintptr_t)arginfo; 242 243 if (execpath_len != 0) { 244 destp -= execpath_len; 245 destp = rounddown2(destp, sizeof(void *)); 246 imgp->execpathp = (void *)destp; 247 error = copyout(imgp->execpath, imgp->execpathp, execpath_len); 248 if (error != 0) 249 return (error); 250 } 251 252 /* Prepare the canary for SSP. */ 253 arc4rand(canary, sizeof(canary), 0); 254 destp -= roundup(sizeof(canary), sizeof(void *)); 255 imgp->canary = (void *)destp; 256 error = copyout(canary, imgp->canary, sizeof(canary)); 257 if (error != 0) 258 return (error); 259 260 /* Allocate room for the argument and environment strings. */ 261 destp -= ARG_MAX - imgp->args->stringspace; 262 destp = rounddown2(destp, sizeof(void *)); 263 ustringp = destp; 264 265 if (imgp->auxargs) { 266 /* 267 * Allocate room on the stack for the ELF auxargs 268 * array. It has up to LINUX_AT_COUNT entries. 269 */ 270 destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo); 271 destp = rounddown2(destp, sizeof(void *)); 272 } 273 274 vectp = (char **)destp; 275 276 /* 277 * Allocate room for argc and the argv[] and env vectors including the 278 * terminating NULL pointers. 279 */ 280 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1; 281 vectp = (char **)STACKALIGN(vectp); 282 283 /* vectp also becomes our initial stack base. */ 284 *stack_base = (uintptr_t)vectp; 285 286 stringp = imgp->args->begin_argv; 287 argc = imgp->args->argc; 288 envc = imgp->args->envc; 289 290 /* Copy out strings - arguments and environment. */ 291 error = copyout(stringp, (void *)ustringp, 292 ARG_MAX - imgp->args->stringspace); 293 if (error != 0) 294 return (error); 295 296 /* Fill in "ps_strings" struct for ps, w, etc. */ 297 if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 || 298 suword(&arginfo->ps_nargvstr, argc) != 0) 299 return (EFAULT); 300 301 if (suword(vectp++, argc) != 0) 302 return (EFAULT); 303 304 /* Fill in argument portion of vector table. */ 305 for (; argc > 0; --argc) { 306 if (suword(vectp++, ustringp) != 0) 307 return (EFAULT); 308 while (*stringp++ != 0) 309 ustringp++; 310 ustringp++; 311 } 312 313 /* A null vector table pointer separates the argp's from the envp's. */ 314 if (suword(vectp++, 0) != 0) 315 return (EFAULT); 316 317 if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 || 318 suword(&arginfo->ps_nenvstr, envc) != 0) 319 return (EFAULT); 320 321 /* Fill in environment portion of vector table. */ 322 for (; envc > 0; --envc) { 323 if (suword(vectp++, ustringp) != 0) 324 return (EFAULT); 325 while (*stringp++ != 0) 326 ustringp++; 327 ustringp++; 328 } 329 330 /* The end of the vector table is a null pointer. */ 331 if (suword(vectp, 0) != 0) 332 return (EFAULT); 333 334 if (imgp->auxargs) { 335 vectp++; 336 error = imgp->sysent->sv_copyout_auxargs(imgp, 337 (uintptr_t)vectp); 338 if (error != 0) 339 return (error); 340 } 341 342 return (0); 343 } 344 345 /* 346 * Reset registers to default values on exec. 347 */ 348 static void 349 linux_exec_setregs(struct thread *td, struct image_params *imgp, 350 uintptr_t stack) 351 { 352 struct trapframe *regs = td->td_frame; 353 struct pcb *pcb = td->td_pcb; 354 355 /* LINUXTODO: validate */ 356 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo); 357 358 memset(regs, 0, sizeof(*regs)); 359 /* glibc start.S registers function pointer in x0 with atexit. */ 360 regs->tf_sp = stack; 361 #if 0 /* LINUXTODO: See if this is used. */ 362 regs->tf_lr = imgp->entry_addr; 363 #else 364 regs->tf_lr = 0xffffffffffffffff; 365 #endif 366 regs->tf_elr = imgp->entry_addr; 367 368 pcb->pcb_tpidr_el0 = 0; 369 pcb->pcb_tpidrro_el0 = 0; 370 WRITE_SPECIALREG(tpidrro_el0, 0); 371 WRITE_SPECIALREG(tpidr_el0, 0); 372 373 #ifdef VFP 374 vfp_reset_state(td, pcb); 375 #endif 376 377 /* 378 * Clear debug register state. It is not applicable to the new process. 379 */ 380 bzero(&pcb->pcb_dbg_regs, sizeof(pcb->pcb_dbg_regs)); 381 } 382 383 int 384 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args) 385 { 386 387 /* LINUXTODO: implement */ 388 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo); 389 return (EDOOFUS); 390 } 391 392 static void 393 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) 394 { 395 396 /* LINUXTODO: implement */ 397 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo); 398 } 399 400 static int 401 linux_vsyscall(struct thread *td) 402 { 403 404 /* LINUXTODO: implement */ 405 LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo); 406 return (EDOOFUS); 407 } 408 409 struct sysentvec elf_linux_sysvec = { 410 .sv_size = LINUX_SYS_MAXSYSCALL, 411 .sv_table = linux_sysent, 412 .sv_transtrap = linux_translate_traps, 413 .sv_fixup = linux_elf_fixup, 414 .sv_sendsig = linux_rt_sendsig, 415 .sv_sigcode = &_binary_linux_locore_o_start, 416 .sv_szsigcode = &linux_szsigcode, 417 .sv_name = "Linux ELF64", 418 .sv_coredump = elf64_coredump, 419 .sv_imgact_try = linux_exec_imgact_try, 420 .sv_minsigstksz = LINUX_MINSIGSTKSZ, 421 .sv_minuser = VM_MIN_ADDRESS, 422 .sv_maxuser = VM_MAXUSER_ADDRESS, 423 .sv_usrstack = USRSTACK, 424 .sv_psstrings = PS_STRINGS, /* XXX */ 425 .sv_stackprot = VM_PROT_READ | VM_PROT_WRITE, 426 .sv_copyout_auxargs = linux_copyout_auxargs, 427 .sv_copyout_strings = linux_copyout_strings, 428 .sv_setregs = linux_exec_setregs, 429 .sv_fixlimit = NULL, 430 .sv_maxssiz = NULL, 431 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP | SV_SIG_DISCIGN | 432 SV_SIG_WAITNDQ, 433 .sv_set_syscall_retval = linux_set_syscall_retval, 434 .sv_fetch_syscall_args = linux_fetch_syscall_args, 435 .sv_syscallnames = NULL, 436 .sv_shared_page_base = SHAREDPAGE, 437 .sv_shared_page_len = PAGE_SIZE, 438 .sv_schedtail = linux_schedtail, 439 .sv_thread_detach = linux_thread_detach, 440 .sv_trap = linux_vsyscall, 441 .sv_hwcap = &elf_hwcap, 442 .sv_hwcap2 = &elf_hwcap2, 443 .sv_onexec = linux_on_exec, 444 .sv_onexit = linux_on_exit, 445 .sv_ontdexit = linux_thread_dtor, 446 .sv_setid_allowed = &linux_setid_allowed_query, 447 }; 448 449 static void 450 linux_vdso_install(const void *param) 451 { 452 453 linux_szsigcode = (&_binary_linux_locore_o_end - 454 &_binary_linux_locore_o_start); 455 456 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len) 457 panic("invalid Linux VDSO size\n"); 458 459 __elfN(linux_vdso_fixup)(&elf_linux_sysvec); 460 461 linux_shared_page_obj = __elfN(linux_shared_page_init) 462 (&linux_shared_page_mapping); 463 464 __elfN(linux_vdso_reloc)(&elf_linux_sysvec); 465 466 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode, 467 linux_szsigcode); 468 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj; 469 } 470 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY, 471 linux_vdso_install, NULL); 472 473 static void 474 linux_vdso_deinstall(const void *param) 475 { 476 477 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo); 478 __elfN(linux_shared_page_fini)(linux_shared_page_obj, 479 linux_shared_page_mapping); 480 } 481 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST, 482 linux_vdso_deinstall, NULL); 483 484 static char GNU_ABI_VENDOR[] = "GNU"; 485 static int GNU_ABI_LINUX = 0; 486 487 /* LINUXTODO: deduplicate */ 488 static bool 489 linux_trans_osrel(const Elf_Note *note, int32_t *osrel) 490 { 491 const Elf32_Word *desc; 492 uintptr_t p; 493 494 p = (uintptr_t)(note + 1); 495 p += roundup2(note->n_namesz, sizeof(Elf32_Addr)); 496 497 desc = (const Elf32_Word *)p; 498 if (desc[0] != GNU_ABI_LINUX) 499 return (false); 500 501 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]); 502 return (true); 503 } 504 505 static Elf_Brandnote linux64_brandnote = { 506 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR), 507 .hdr.n_descsz = 16, 508 .hdr.n_type = 1, 509 .vendor = GNU_ABI_VENDOR, 510 .flags = BN_TRANSLATE_OSREL, 511 .trans_osrel = linux_trans_osrel 512 }; 513 514 static Elf64_Brandinfo linux_glibc2brand = { 515 .brand = ELFOSABI_LINUX, 516 .machine = EM_AARCH64, 517 .compat_3_brand = "Linux", 518 .emul_path = linux_emul_path, 519 .interp_path = "/lib64/ld-linux-x86-64.so.2", 520 .sysvec = &elf_linux_sysvec, 521 .interp_newpath = NULL, 522 .brand_note = &linux64_brandnote, 523 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE 524 }; 525 526 Elf64_Brandinfo *linux_brandlist[] = { 527 &linux_glibc2brand, 528 NULL 529 }; 530 531 static int 532 linux64_elf_modevent(module_t mod, int type, void *data) 533 { 534 Elf64_Brandinfo **brandinfo; 535 struct linux_ioctl_handler**lihp; 536 int error; 537 538 error = 0; 539 switch(type) { 540 case MOD_LOAD: 541 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 542 ++brandinfo) 543 if (elf64_insert_brand_entry(*brandinfo) < 0) 544 error = EINVAL; 545 if (error == 0) { 546 SET_FOREACH(lihp, linux_ioctl_handler_set) 547 linux_ioctl_register_handler(*lihp); 548 stclohz = (stathz ? stathz : hz); 549 if (bootverbose) 550 printf("Linux arm64 ELF exec handler installed\n"); 551 } 552 break; 553 case MOD_UNLOAD: 554 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL; 555 ++brandinfo) 556 if (elf64_brand_inuse(*brandinfo)) 557 error = EBUSY; 558 if (error == 0) { 559 for (brandinfo = &linux_brandlist[0]; 560 *brandinfo != NULL; ++brandinfo) 561 if (elf64_remove_brand_entry(*brandinfo) < 0) 562 error = EINVAL; 563 } 564 if (error == 0) { 565 SET_FOREACH(lihp, linux_ioctl_handler_set) 566 linux_ioctl_unregister_handler(*lihp); 567 if (bootverbose) 568 printf("Linux ELF exec handler removed\n"); 569 } else 570 printf("Could not deinstall ELF interpreter entry\n"); 571 break; 572 default: 573 return (EOPNOTSUPP); 574 } 575 return (error); 576 } 577 578 static moduledata_t linux64_elf_mod = { 579 "linux64elf", 580 linux64_elf_modevent, 581 0 582 }; 583 584 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY); 585 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1); 586 FEATURE(linux64, "AArch64 Linux 64bit support"); 587