2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 1994-1996 Søren Schmidt
5 * Copyright (c) 2018 Turing Robotic Industries Inc.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/cdefs.h>
37 #include <sys/imgact.h>
38 #include <sys/imgact_elf.h>
39 #include <sys/kernel.h>
41 #include <sys/module.h>
42 #include <sys/mutex.h>
44 #include <sys/signalvar.h>
45 #include <sys/sysctl.h>
46 #include <sys/sysent.h>
48 #include <vm/vm_param.h>
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>
60 MODULE_VERSION(linux64elf, 1);
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;
69 extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];
71 SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
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,
84 static int linux_vsyscall(struct thread *td);
87 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
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);
100 /* LINUXTODO: do we have traps to translate? */
102 linux_translate_traps(int signal, int trap_code)
105 LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code);
109 LINUX_VDSO_SYM_CHAR(linux_platform);
112 linux_fetch_syscall_args(struct thread *td)
115 struct syscall_args *sa;
119 ap = td->td_frame->tf_x;
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];
127 sa->callp = &p->p_sysent->sv_table[sa->code];
129 if (sa->callp->sy_narg > MAXARGS)
130 panic("ARM64TODO: Could we have more than %d args?", MAXARGS);
131 memcpy(sa->args, ap, MAXARGS * sizeof(register_t));
133 td->td_retval[0] = 0;
138 linux_set_syscall_retval(struct thread *td, int error)
141 td->td_retval[1] = td->td_frame->tf_x[1];
142 cpu_set_syscall_retval(td, error);
144 if (__predict_false(error != 0)) {
145 if (error != ERESTART && error != EJUSTRETURN)
146 td->td_frame->tf_x[0] = bsd_to_linux_errno(error);
151 linux_copyout_auxargs(struct image_params *imgp, uintptr_t base)
154 Elf_Auxinfo *argarray, *pos;
156 int error, issetugid;
158 LIN_SDT_PROBE0(sysvec, linux_copyout_auxargs, todo);
161 args = (Elf64_Auxargs *)imgp->auxargs;
162 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP,
165 issetugid = p->p_flag & P_SUGID ? 1 : 0;
166 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR,
167 imgp->proc->p_sysent->sv_shared_page_base);
168 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_HWCAP */
169 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
171 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
172 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
173 AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
174 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
175 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
176 AUXARGS_ENTRY(pos, AT_BASE, args->base);
177 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
178 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
179 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
180 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
181 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
182 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
183 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid);
184 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_PLATFORM */
185 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform));
187 AUXARGS_ENTRY_PTR(pos, LINUX_AT_RANDOM, imgp->canary);
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, AT_NULL, 0);
193 free(imgp->auxargs, M_TEMP);
194 imgp->auxargs = NULL;
195 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs"));
197 error = copyout(argarray, (void *)base,
198 sizeof(*argarray) * LINUX_AT_COUNT);
199 free(argarray, M_TEMP);
204 linux_elf_fixup(uintptr_t *stack_base, struct image_params *imgp)
207 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo);
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
219 linux_copyout_strings(struct image_params *imgp, uintptr_t *stack_base)
223 uintptr_t destp, ustringp;
224 struct ps_strings *arginfo;
225 char canary[LINUX_AT_RANDOM_LEN];
228 int argc, envc, error;
230 /* Calculate string base and vector table pointers. */
231 if (imgp->execpath != NULL && imgp->auxargs != NULL)
232 execpath_len = strlen(imgp->execpath) + 1;
237 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
238 destp = (uintptr_t)arginfo;
240 if (execpath_len != 0) {
241 destp -= execpath_len;
242 destp = rounddown2(destp, sizeof(void *));
243 imgp->execpathp = (void *)destp;
244 error = copyout(imgp->execpath, imgp->execpathp, execpath_len);
249 /* Prepare the canary for SSP. */
250 arc4rand(canary, sizeof(canary), 0);
251 destp -= roundup(sizeof(canary), sizeof(void *));
252 imgp->canary = (void *)destp;
253 error = copyout(canary, imgp->canary, sizeof(canary));
257 /* Allocate room for the argument and environment strings. */
258 destp -= ARG_MAX - imgp->args->stringspace;
259 destp = rounddown2(destp, sizeof(void *));
264 * Allocate room on the stack for the ELF auxargs
265 * array. It has up to LINUX_AT_COUNT entries.
267 destp -= LINUX_AT_COUNT * sizeof(Elf64_Auxinfo);
268 destp = rounddown2(destp, sizeof(void *));
271 vectp = (char **)destp;
274 * Allocate room for argc and the argv[] and env vectors including the
275 * terminating NULL pointers.
277 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1;
278 vectp = (char **)STACKALIGN(vectp);
280 /* vectp also becomes our initial stack base. */
281 *stack_base = (uintptr_t)vectp;
283 stringp = imgp->args->begin_argv;
284 argc = imgp->args->argc;
285 envc = imgp->args->envc;
287 /* Copy out strings - arguments and environment. */
288 error = copyout(stringp, (void *)ustringp,
289 ARG_MAX - imgp->args->stringspace);
293 /* Fill in "ps_strings" struct for ps, w, etc. */
294 if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 ||
295 suword(&arginfo->ps_nargvstr, argc) != 0)
298 if (suword(vectp++, argc) != 0)
301 /* Fill in argument portion of vector table. */
302 for (; argc > 0; --argc) {
303 if (suword(vectp++, ustringp) != 0)
305 while (*stringp++ != 0)
310 /* A null vector table pointer separates the argp's from the envp's. */
311 if (suword(vectp++, 0) != 0)
314 if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 ||
315 suword(&arginfo->ps_nenvstr, envc) != 0)
318 /* Fill in environment portion of vector table. */
319 for (; envc > 0; --envc) {
320 if (suword(vectp++, ustringp) != 0)
322 while (*stringp++ != 0)
327 /* The end of the vector table is a null pointer. */
328 if (suword(vectp, 0) != 0)
333 error = imgp->sysent->sv_copyout_auxargs(imgp,
343 * Reset registers to default values on exec.
346 linux_exec_setregs(struct thread *td, struct image_params *imgp,
349 struct trapframe *regs = td->td_frame;
351 /* LINUXTODO: validate */
352 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo);
354 memset(regs, 0, sizeof(*regs));
355 /* glibc start.S registers function pointer in x0 with atexit. */
357 #if 0 /* LINUXTODO: See if this is used. */
358 regs->tf_lr = imgp->entry_addr;
360 regs->tf_lr = 0xffffffffffffffff;
362 regs->tf_elr = imgp->entry_addr;
366 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
369 /* LINUXTODO: implement */
370 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo);
375 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
378 /* LINUXTODO: implement */
379 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo);
383 linux_vsyscall(struct thread *td)
386 /* LINUXTODO: implement */
387 LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo);
391 struct sysentvec elf_linux_sysvec = {
392 .sv_size = LINUX_SYS_MAXSYSCALL,
393 .sv_table = linux_sysent,
394 .sv_transtrap = linux_translate_traps,
395 .sv_fixup = linux_elf_fixup,
396 .sv_sendsig = linux_rt_sendsig,
397 .sv_sigcode = &_binary_linux_locore_o_start,
398 .sv_szsigcode = &linux_szsigcode,
399 .sv_name = "Linux ELF64",
400 .sv_coredump = elf64_coredump,
401 .sv_imgact_try = linux_exec_imgact_try,
402 .sv_minsigstksz = LINUX_MINSIGSTKSZ,
403 .sv_minuser = VM_MIN_ADDRESS,
404 .sv_maxuser = VM_MAXUSER_ADDRESS,
405 .sv_usrstack = USRSTACK,
406 .sv_psstrings = PS_STRINGS, /* XXX */
407 .sv_stackprot = VM_PROT_READ | VM_PROT_WRITE,
408 .sv_copyout_auxargs = linux_copyout_auxargs,
409 .sv_copyout_strings = linux_copyout_strings,
410 .sv_setregs = linux_exec_setregs,
413 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP,
414 .sv_set_syscall_retval = linux_set_syscall_retval,
415 .sv_fetch_syscall_args = linux_fetch_syscall_args,
416 .sv_syscallnames = NULL,
417 .sv_shared_page_base = SHAREDPAGE,
418 .sv_shared_page_len = PAGE_SIZE,
419 .sv_schedtail = linux_schedtail,
420 .sv_thread_detach = linux_thread_detach,
421 .sv_trap = linux_vsyscall,
422 .sv_onexec = linux_on_exec,
423 .sv_onexit = linux_on_exit,
424 .sv_ontdexit = linux_thread_dtor,
428 linux_vdso_install(const void *param)
431 linux_szsigcode = (&_binary_linux_locore_o_end -
432 &_binary_linux_locore_o_start);
434 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
435 panic("invalid Linux VDSO size\n");
437 __elfN(linux_vdso_fixup)(&elf_linux_sysvec);
439 linux_shared_page_obj = __elfN(linux_shared_page_init)
440 (&linux_shared_page_mapping);
442 __elfN(linux_vdso_reloc)(&elf_linux_sysvec);
444 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode,
446 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
448 printf("LINUXTODO: %s: fix linux_kplatform\n", __func__);
450 linux_kplatform = linux_shared_page_mapping +
451 (linux_platform - (caddr_t)elf_linux_sysvec.sv_shared_page_base);
453 linux_kplatform = "arm64";
456 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
457 linux_vdso_install, NULL);
460 linux_vdso_deinstall(const void *param)
463 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo);
464 __elfN(linux_shared_page_fini)(linux_shared_page_obj);
466 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
467 linux_vdso_deinstall, NULL);
469 static char GNU_ABI_VENDOR[] = "GNU";
470 static int GNU_ABI_LINUX = 0;
472 /* LINUXTODO: deduplicate */
474 linux_trans_osrel(const Elf_Note *note, int32_t *osrel)
476 const Elf32_Word *desc;
479 p = (uintptr_t)(note + 1);
480 p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
482 desc = (const Elf32_Word *)p;
483 if (desc[0] != GNU_ABI_LINUX)
486 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]);
490 static Elf_Brandnote linux64_brandnote = {
491 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
494 .vendor = GNU_ABI_VENDOR,
495 .flags = BN_TRANSLATE_OSREL,
496 .trans_osrel = linux_trans_osrel
499 static Elf64_Brandinfo linux_glibc2brand = {
500 .brand = ELFOSABI_LINUX,
501 .machine = EM_AARCH64,
502 .compat_3_brand = "Linux",
503 .emul_path = linux_emul_path,
504 .interp_path = "/lib64/ld-linux-x86-64.so.2",
505 .sysvec = &elf_linux_sysvec,
506 .interp_newpath = NULL,
507 .brand_note = &linux64_brandnote,
508 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
511 Elf64_Brandinfo *linux_brandlist[] = {
517 linux64_elf_modevent(module_t mod, int type, void *data)
519 Elf64_Brandinfo **brandinfo;
520 struct linux_ioctl_handler**lihp;
526 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
528 if (elf64_insert_brand_entry(*brandinfo) < 0)
531 SET_FOREACH(lihp, linux_ioctl_handler_set)
532 linux_ioctl_register_handler(*lihp);
533 stclohz = (stathz ? stathz : hz);
535 printf("Linux arm64 ELF exec handler installed\n");
539 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
541 if (elf64_brand_inuse(*brandinfo))
544 for (brandinfo = &linux_brandlist[0];
545 *brandinfo != NULL; ++brandinfo)
546 if (elf64_remove_brand_entry(*brandinfo) < 0)
550 SET_FOREACH(lihp, linux_ioctl_handler_set)
551 linux_ioctl_unregister_handler(*lihp);
553 printf("Linux ELF exec handler removed\n");
555 printf("Could not deinstall ELF interpreter entry\n");
563 static moduledata_t linux64_elf_mod = {
565 linux64_elf_modevent,
569 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
570 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1);
571 FEATURE(linux64, "AArch64 Linux 64bit support");