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_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>
60 MODULE_VERSION(linux64elf, 1);
63 SYSCTL_PROC(_compat_linux, OID_AUTO, debug, CTLTYPE_STRING | CTLFLAG_RW, 0, 0,
64 linux_sysctl_debug, "A", "64-bit Linux debugging control");
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;
74 extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];
76 SET_DECLARE(linux_ioctl_handler_set, struct linux_ioctl_handler);
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,
88 static int linux_vsyscall(struct thread *td);
91 LIN_SDT_PROVIDER_DECLARE(LINUX_DTRACE);
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);
103 /* LINUXTODO: do we have traps to translate? */
105 linux_translate_traps(int signal, int trap_code)
108 LIN_SDT_PROBE2(sysvec, linux_translate_traps, todo, signal, trap_code);
112 LINUX_VDSO_SYM_CHAR(linux_platform);
115 linux_fetch_syscall_args(struct thread *td)
118 struct syscall_args *sa;
122 ap = td->td_frame->tf_x;
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];
130 sa->callp = &p->p_sysent->sv_table[sa->code];
132 sa->narg = sa->callp->sy_narg;
134 panic("ARM64TODO: Could we have more than 8 args?");
135 memcpy(sa->args, ap, 8 * sizeof(register_t));
137 td->td_retval[0] = 0;
142 linux_set_syscall_retval(struct thread *td, int error)
145 td->td_retval[1] = td->td_frame->tf_x[1];
146 cpu_set_syscall_retval(td, error);
150 linux_elf_fixup(register_t **stack_base, struct image_params *imgp)
153 Elf_Auxinfo *argarray, *pos;
154 Elf_Addr *auxbase, *base;
155 struct ps_strings *arginfo;
157 int error, issetugid;
159 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo);
161 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
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,
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);
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));
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"));
204 error = copyout(argarray, auxbase, sizeof(*argarray) * LINUX_AT_COUNT);
205 free(argarray, M_TEMP);
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)
222 char *stringp, *destp;
223 register_t *stack_base;
224 struct ps_strings *arginfo;
225 char canary[LINUX_AT_RANDOM_LEN];
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 = (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 *));
243 if (execpath_len != 0) {
244 imgp->execpathp = (uintptr_t)arginfo - execpath_len;
245 copyout(imgp->execpath, (void *)imgp->execpathp, execpath_len);
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));
255 vectp = (char **)destp;
258 * Allocate room on the stack for the ELF auxargs
259 * array. It has up to LINUX_AT_COUNT entries.
261 vectp -= howmany(LINUX_AT_COUNT * sizeof(Elf64_Auxinfo),
266 * Allocate room for argc and the argv[] and env vectors including the
267 * terminating NULL pointers.
269 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1;
270 vectp = (char **)STACKALIGN(vectp);
272 /* vectp also becomes our initial stack base. */
273 stack_base = (register_t *)vectp;
275 stringp = imgp->args->begin_argv;
276 argc = imgp->args->argc;
277 envc = imgp->args->envc;
279 /* Copy out strings - arguments and environment. */
280 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
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);
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)
295 /* A null vector table pointer separates the argp's from the envp's. */
298 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
299 suword(&arginfo->ps_nenvstr, envc);
301 /* Fill in environment portion of vector table. */
302 for (; envc > 0; --envc) {
303 suword(vectp++, (long)(intptr_t)destp);
304 while (*stringp++ != 0)
309 /* The end of the vector table is a null pointer. */
315 * Reset registers to default values on exec.
318 linux_exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
320 struct trapframe *regs = td->td_frame;
322 /* LINUXTODO: validate */
323 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo);
325 memset(regs, 0, sizeof(*regs));
326 /* glibc start.S registers function pointer in x0 with atexit. */
328 #if 0 /* LINUXTODO: See if this is used. */
329 regs->tf_lr = imgp->entry_addr;
331 regs->tf_lr = 0xffffffffffffffff;
333 regs->tf_elr = imgp->entry_addr;
337 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
340 /* LINUXTODO: implement */
341 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo);
346 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
349 /* LINUXTODO: implement */
350 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo);
354 linux_vsyscall(struct thread *td)
357 /* LINUXTODO: implement */
358 LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo);
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_pagesize = PAGE_SIZE,
377 .sv_minuser = VM_MIN_ADDRESS,
378 .sv_maxuser = VM_MAXUSER_ADDRESS,
379 .sv_usrstack = USRSTACK,
380 .sv_psstrings = PS_STRINGS, /* XXX */
381 .sv_stackprot = VM_PROT_ALL, /* XXX */
382 .sv_copyout_strings = linux_copyout_strings,
383 .sv_setregs = linux_exec_setregs,
386 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP,
387 .sv_set_syscall_retval = linux_set_syscall_retval,
388 .sv_fetch_syscall_args = linux_fetch_syscall_args,
389 .sv_syscallnames = NULL,
390 .sv_shared_page_base = SHAREDPAGE,
391 .sv_shared_page_len = PAGE_SIZE,
392 .sv_schedtail = linux_schedtail,
393 .sv_thread_detach = linux_thread_detach,
394 .sv_trap = linux_vsyscall,
398 linux_vdso_install(const void *param)
401 linux_szsigcode = (&_binary_linux_locore_o_end -
402 &_binary_linux_locore_o_start);
404 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
405 panic("invalid Linux VDSO size\n");
407 __elfN(linux_vdso_fixup)(&elf_linux_sysvec);
409 linux_shared_page_obj = __elfN(linux_shared_page_init)
410 (&linux_shared_page_mapping);
412 __elfN(linux_vdso_reloc)(&elf_linux_sysvec);
414 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode,
416 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
418 printf("LINUXTODO: %s: fix linux_kplatform\n", __func__);
420 linux_kplatform = linux_shared_page_mapping +
421 (linux_platform - (caddr_t)elf_linux_sysvec.sv_shared_page_base);
423 linux_kplatform = "arm64";
426 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
427 linux_vdso_install, NULL);
430 linux_vdso_deinstall(const void *param)
433 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo);
434 __elfN(linux_shared_page_fini)(linux_shared_page_obj);
436 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
437 linux_vdso_deinstall, NULL);
439 static char GNU_ABI_VENDOR[] = "GNU";
440 static int GNU_ABI_LINUX = 0;
442 /* LINUXTODO: deduplicate */
444 linux_trans_osrel(const Elf_Note *note, int32_t *osrel)
446 const Elf32_Word *desc;
449 p = (uintptr_t)(note + 1);
450 p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
452 desc = (const Elf32_Word *)p;
453 if (desc[0] != GNU_ABI_LINUX)
456 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]);
460 static Elf_Brandnote linux64_brandnote = {
461 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
464 .vendor = GNU_ABI_VENDOR,
465 .flags = BN_TRANSLATE_OSREL,
466 .trans_osrel = linux_trans_osrel
469 static Elf64_Brandinfo linux_glibc2brand = {
470 .brand = ELFOSABI_LINUX,
471 .machine = EM_AARCH64,
472 .compat_3_brand = "Linux",
473 .emul_path = "/compat/linux",
474 .interp_path = "/lib64/ld-linux-x86-64.so.2",
475 .sysvec = &elf_linux_sysvec,
476 .interp_newpath = NULL,
477 .brand_note = &linux64_brandnote,
478 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
481 Elf64_Brandinfo *linux_brandlist[] = {
487 linux64_elf_modevent(module_t mod, int type, void *data)
489 Elf64_Brandinfo **brandinfo;
490 struct linux_ioctl_handler**lihp;
496 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
498 if (elf64_insert_brand_entry(*brandinfo) < 0)
501 SET_FOREACH(lihp, linux_ioctl_handler_set)
502 linux_ioctl_register_handler(*lihp);
503 LIST_INIT(&futex_list);
504 mtx_init(&futex_mtx, "ftllk64", NULL, MTX_DEF);
505 stclohz = (stathz ? stathz : hz);
507 printf("Linux arm64 ELF exec handler installed\n");
511 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
513 if (elf64_brand_inuse(*brandinfo))
516 for (brandinfo = &linux_brandlist[0];
517 *brandinfo != NULL; ++brandinfo)
518 if (elf64_remove_brand_entry(*brandinfo) < 0)
522 SET_FOREACH(lihp, linux_ioctl_handler_set)
523 linux_ioctl_unregister_handler(*lihp);
524 mtx_destroy(&futex_mtx);
526 printf("Linux ELF exec handler removed\n");
528 printf("Could not deinstall ELF interpreter entry\n");
536 static moduledata_t linux64_elf_mod = {
538 linux64_elf_modevent,
542 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
543 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1);
544 FEATURE(linux64, "AArch64 Linux 64bit support");