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_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>
59
60 MODULE_VERSION(linux64elf, 1);
61
62 #if defined(DEBUG)
63 SYSCTL_PROC(_compat_linux, OID_AUTO, debug, CTLTYPE_STRING | CTLFLAG_RW, 0, 0,
64 linux_sysctl_debug, "A", "64-bit Linux debugging control");
65 #endif
66
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;
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 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,
87 u_long stack);
88 static int linux_vsyscall(struct thread *td);
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_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);
102
103 /* LINUXTODO: do we have traps to translate? */
104 static int
linux_translate_traps(int signal,int trap_code)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
linux_fetch_syscall_args(struct thread * td)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 /* LINUXTODO: generic syscall? */
127 if (p->p_sysent->sv_mask)
128 sa->code &= p->p_sysent->sv_mask;
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 sa->narg = sa->callp->sy_narg;
135 if (sa->narg > 8)
136 panic("ARM64TODO: Could we have more than 8 args?");
137 memcpy(sa->args, ap, 8 * sizeof(register_t));
138
139 td->td_retval[0] = 0;
140 return (0);
141 }
142
143 static void
linux_set_syscall_retval(struct thread * td,int error)144 linux_set_syscall_retval(struct thread *td, int error)
145 {
146
147 td->td_retval[1] = td->td_frame->tf_x[1];
148 cpu_set_syscall_retval(td, error);
149 }
150
151 static int
linux_elf_fixup(register_t ** stack_base,struct image_params * imgp)152 linux_elf_fixup(register_t **stack_base, struct image_params *imgp)
153 {
154 Elf_Auxargs *args;
155 Elf_Auxinfo *argarray, *pos;
156 Elf_Addr *auxbase, *base;
157 struct ps_strings *arginfo;
158 struct proc *p;
159 int error, issetugid;
160
161 LIN_SDT_PROBE0(sysvec, linux_elf_fixup, todo);
162 p = imgp->proc;
163 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
164
165 KASSERT(curthread->td_proc == imgp->proc,
166 ("unsafe linux_elf_fixup(), should be curproc"));
167 base = (Elf64_Addr *)*stack_base;
168 args = (Elf64_Auxargs *)imgp->auxargs;
169 /* Auxargs after argc, and NULL-terminated argv and envv lists. */
170 auxbase = base + 1 + imgp->args->argc + 1 + imgp->args->envc + 1;
171 argarray = pos = malloc(LINUX_AT_COUNT * sizeof(*pos), M_TEMP,
172 M_WAITOK | M_ZERO);
173
174 issetugid = p->p_flag & P_SUGID ? 1 : 0;
175 AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR,
176 imgp->proc->p_sysent->sv_shared_page_base);
177 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_HWCAP */
178 AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
179 #endif
180 AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
181 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
182 AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
183 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
184 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
185 AUXARGS_ENTRY(pos, AT_BASE, args->base);
186 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
187 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
188 AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
189 AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
190 AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
191 AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
192 AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid);
193 #if 0 /* LINUXTODO: implement arm64 LINUX_AT_PLATFORM */
194 AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform));
195 #endif
196 AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, imgp->canary);
197 if (imgp->execpathp != 0)
198 AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, imgp->execpathp);
199 if (args->execfd != -1)
200 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
201 AUXARGS_ENTRY(pos, AT_NULL, 0);
202 free(imgp->auxargs, M_TEMP);
203 imgp->auxargs = NULL;
204 KASSERT(pos - argarray <= LINUX_AT_COUNT, ("Too many auxargs"));
205
206 error = copyout(argarray, auxbase, sizeof(*argarray) * LINUX_AT_COUNT);
207 free(argarray, M_TEMP);
208 if (error != 0)
209 return (error);
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 register_t *
linux_copyout_strings(struct image_params * imgp)221 linux_copyout_strings(struct image_params *imgp)
222 {
223 char **vectp;
224 char *stringp, *destp;
225 register_t *stack_base;
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;
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 = (caddr_t)arginfo - SPARE_USRSPACE -
241 roundup(sizeof(canary), sizeof(char *)) -
242 roundup(execpath_len, sizeof(char *)) -
243 roundup(ARG_MAX - imgp->args->stringspace, sizeof(char *));
244
245 if (execpath_len != 0) {
246 imgp->execpathp = (uintptr_t)arginfo - execpath_len;
247 copyout(imgp->execpath, (void *)imgp->execpathp, execpath_len);
248 }
249
250 /* Prepare the canary for SSP. */
251 arc4rand(canary, sizeof(canary), 0);
252 imgp->canary = (uintptr_t)arginfo -
253 roundup(execpath_len, sizeof(char *)) -
254 roundup(sizeof(canary), sizeof(char *));
255 copyout(canary, (void *)imgp->canary, sizeof(canary));
256
257 vectp = (char **)destp;
258 if (imgp->auxargs) {
259 /*
260 * Allocate room on the stack for the ELF auxargs
261 * array. It has up to LINUX_AT_COUNT entries.
262 */
263 vectp -= howmany(LINUX_AT_COUNT * sizeof(Elf64_Auxinfo),
264 sizeof(*vectp));
265 }
266
267 /*
268 * Allocate room for argc and the argv[] and env vectors including the
269 * terminating NULL pointers.
270 */
271 vectp -= 1 + imgp->args->argc + 1 + imgp->args->envc + 1;
272 vectp = (char **)STACKALIGN(vectp);
273
274 /* vectp also becomes our initial stack base. */
275 stack_base = (register_t *)vectp;
276
277 stringp = imgp->args->begin_argv;
278 argc = imgp->args->argc;
279 envc = imgp->args->envc;
280
281 /* Copy out strings - arguments and environment. */
282 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
283
284 /* Fill in "ps_strings" struct for ps, w, etc. */
285 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
286 suword(&arginfo->ps_nargvstr, argc);
287
288 suword(vectp++, argc);
289 /* Fill in argument portion of vector table. */
290 for (; argc > 0; --argc) {
291 suword(vectp++, (long)(intptr_t)destp);
292 while (*stringp++ != 0)
293 destp++;
294 destp++;
295 }
296
297 /* A null vector table pointer separates the argp's from the envp's. */
298 suword(vectp++, 0);
299
300 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
301 suword(&arginfo->ps_nenvstr, envc);
302
303 /* Fill in environment portion of vector table. */
304 for (; envc > 0; --envc) {
305 suword(vectp++, (long)(intptr_t)destp);
306 while (*stringp++ != 0)
307 destp++;
308 destp++;
309 }
310
311 /* The end of the vector table is a null pointer. */
312 suword(vectp, 0);
313 return (stack_base);
314 }
315
316 /*
317 * Reset registers to default values on exec.
318 */
319 static void
linux_exec_setregs(struct thread * td,struct image_params * imgp,u_long stack)320 linux_exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
321 {
322 struct trapframe *regs = td->td_frame;
323
324 /* LINUXTODO: validate */
325 LIN_SDT_PROBE0(sysvec, linux_exec_setregs, todo);
326
327 memset(regs, 0, sizeof(*regs));
328 /* glibc start.S registers function pointer in x0 with atexit. */
329 regs->tf_sp = stack;
330 #if 0 /* LINUXTODO: See if this is used. */
331 regs->tf_lr = imgp->entry_addr;
332 #else
333 regs->tf_lr = 0xffffffffffffffff;
334 #endif
335 regs->tf_elr = imgp->entry_addr;
336 }
337
338 int
linux_rt_sigreturn(struct thread * td,struct linux_rt_sigreturn_args * args)339 linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
340 {
341
342 /* LINUXTODO: implement */
343 LIN_SDT_PROBE0(sysvec, linux_rt_sigreturn, todo);
344 return (EDOOFUS);
345 }
346
347 static void
linux_rt_sendsig(sig_t catcher,ksiginfo_t * ksi,sigset_t * mask)348 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
349 {
350
351 /* LINUXTODO: implement */
352 LIN_SDT_PROBE0(sysvec, linux_rt_sendsig, todo);
353 }
354
355 static int
linux_vsyscall(struct thread * td)356 linux_vsyscall(struct thread *td)
357 {
358
359 /* LINUXTODO: implement */
360 LIN_SDT_PROBE0(sysvec, linux_vsyscall, todo);
361 return (EDOOFUS);
362 }
363
364 struct sysentvec elf_linux_sysvec = {
365 .sv_size = LINUX_SYS_MAXSYSCALL,
366 .sv_table = linux_sysent,
367 .sv_mask = 0,
368 .sv_errsize = ELAST + 1,
369 .sv_errtbl = linux_errtbl,
370 .sv_transtrap = linux_translate_traps,
371 .sv_fixup = linux_elf_fixup,
372 .sv_sendsig = linux_rt_sendsig,
373 .sv_sigcode = &_binary_linux_locore_o_start,
374 .sv_szsigcode = &linux_szsigcode,
375 .sv_name = "Linux ELF64",
376 .sv_coredump = elf64_coredump,
377 .sv_imgact_try = linux_exec_imgact_try,
378 .sv_minsigstksz = LINUX_MINSIGSTKSZ,
379 .sv_minuser = VM_MIN_ADDRESS,
380 .sv_maxuser = VM_MAXUSER_ADDRESS,
381 .sv_usrstack = USRSTACK,
382 .sv_psstrings = PS_STRINGS, /* XXX */
383 .sv_stackprot = VM_PROT_ALL, /* XXX */
384 .sv_copyout_strings = linux_copyout_strings,
385 .sv_setregs = linux_exec_setregs,
386 .sv_fixlimit = NULL,
387 .sv_maxssiz = NULL,
388 .sv_flags = SV_ABI_LINUX | SV_LP64 | SV_SHP,
389 .sv_set_syscall_retval = linux_set_syscall_retval,
390 .sv_fetch_syscall_args = linux_fetch_syscall_args,
391 .sv_syscallnames = NULL,
392 .sv_shared_page_base = SHAREDPAGE,
393 .sv_shared_page_len = PAGE_SIZE,
394 .sv_schedtail = linux_schedtail,
395 .sv_thread_detach = linux_thread_detach,
396 .sv_trap = linux_vsyscall,
397 };
398
399 static void
linux_vdso_install(const void * param)400 linux_vdso_install(const void *param)
401 {
402
403 linux_szsigcode = (&_binary_linux_locore_o_end -
404 &_binary_linux_locore_o_start);
405
406 if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
407 panic("invalid Linux VDSO size\n");
408
409 __elfN(linux_vdso_fixup)(&elf_linux_sysvec);
410
411 linux_shared_page_obj = __elfN(linux_shared_page_init)
412 (&linux_shared_page_mapping);
413
414 __elfN(linux_vdso_reloc)(&elf_linux_sysvec);
415
416 memcpy(linux_shared_page_mapping, elf_linux_sysvec.sv_sigcode,
417 linux_szsigcode);
418 elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
419
420 printf("LINUXTODO: %s: fix linux_kplatform\n", __func__);
421 #if 0
422 linux_kplatform = linux_shared_page_mapping +
423 (linux_platform - (caddr_t)elf_linux_sysvec.sv_shared_page_base);
424 #else
425 linux_kplatform = "arm64";
426 #endif
427 }
428 SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
429 linux_vdso_install, NULL);
430
431 static void
linux_vdso_deinstall(const void * param)432 linux_vdso_deinstall(const void *param)
433 {
434
435 LIN_SDT_PROBE0(sysvec, linux_vdso_deinstall, todo);
436 __elfN(linux_shared_page_fini)(linux_shared_page_obj);
437 }
438 SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
439 linux_vdso_deinstall, NULL);
440
441 static char GNU_ABI_VENDOR[] = "GNU";
442 static int GNU_ABI_LINUX = 0;
443
444 /* LINUXTODO: deduplicate */
445 static bool
linux_trans_osrel(const Elf_Note * note,int32_t * osrel)446 linux_trans_osrel(const Elf_Note *note, int32_t *osrel)
447 {
448 const Elf32_Word *desc;
449 uintptr_t p;
450
451 p = (uintptr_t)(note + 1);
452 p += roundup2(note->n_namesz, sizeof(Elf32_Addr));
453
454 desc = (const Elf32_Word *)p;
455 if (desc[0] != GNU_ABI_LINUX)
456 return (false);
457
458 *osrel = LINUX_KERNVER(desc[1], desc[2], desc[3]);
459 return (true);
460 }
461
462 static Elf_Brandnote linux64_brandnote = {
463 .hdr.n_namesz = sizeof(GNU_ABI_VENDOR),
464 .hdr.n_descsz = 16,
465 .hdr.n_type = 1,
466 .vendor = GNU_ABI_VENDOR,
467 .flags = BN_TRANSLATE_OSREL,
468 .trans_osrel = linux_trans_osrel
469 };
470
471 static Elf64_Brandinfo linux_glibc2brand = {
472 .brand = ELFOSABI_LINUX,
473 .machine = EM_AARCH64,
474 .compat_3_brand = "Linux",
475 .emul_path = "/compat/linux",
476 .interp_path = "/lib64/ld-linux-x86-64.so.2",
477 .sysvec = &elf_linux_sysvec,
478 .interp_newpath = NULL,
479 .brand_note = &linux64_brandnote,
480 .flags = BI_CAN_EXEC_DYN | BI_BRAND_NOTE
481 };
482
483 Elf64_Brandinfo *linux_brandlist[] = {
484 &linux_glibc2brand,
485 NULL
486 };
487
488 static int
linux64_elf_modevent(module_t mod,int type,void * data)489 linux64_elf_modevent(module_t mod, int type, void *data)
490 {
491 Elf64_Brandinfo **brandinfo;
492 struct linux_ioctl_handler**lihp;
493 int error;
494
495 error = 0;
496 switch(type) {
497 case MOD_LOAD:
498 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
499 ++brandinfo)
500 if (elf64_insert_brand_entry(*brandinfo) < 0)
501 error = EINVAL;
502 if (error == 0) {
503 SET_FOREACH(lihp, linux_ioctl_handler_set)
504 linux_ioctl_register_handler(*lihp);
505 LIST_INIT(&futex_list);
506 mtx_init(&futex_mtx, "ftllk64", NULL, MTX_DEF);
507 stclohz = (stathz ? stathz : hz);
508 if (bootverbose)
509 printf("Linux arm64 ELF exec handler installed\n");
510 }
511 break;
512 case MOD_UNLOAD:
513 for (brandinfo = &linux_brandlist[0]; *brandinfo != NULL;
514 ++brandinfo)
515 if (elf64_brand_inuse(*brandinfo))
516 error = EBUSY;
517 if (error == 0) {
518 for (brandinfo = &linux_brandlist[0];
519 *brandinfo != NULL; ++brandinfo)
520 if (elf64_remove_brand_entry(*brandinfo) < 0)
521 error = EINVAL;
522 }
523 if (error == 0) {
524 SET_FOREACH(lihp, linux_ioctl_handler_set)
525 linux_ioctl_unregister_handler(*lihp);
526 mtx_destroy(&futex_mtx);
527 if (bootverbose)
528 printf("Linux ELF exec handler removed\n");
529 } else
530 printf("Could not deinstall ELF interpreter entry\n");
531 break;
532 default:
533 return (EOPNOTSUPP);
534 }
535 return (error);
536 }
537
538 static moduledata_t linux64_elf_mod = {
539 "linux64elf",
540 linux64_elf_modevent,
541 0
542 };
543
544 DECLARE_MODULE_TIED(linux64elf, linux64_elf_mod, SI_SUB_EXEC, SI_ORDER_ANY);
545 MODULE_DEPEND(linux64elf, linux_common, 1, 1, 1);
546 FEATURE(linux64, "AArch64 Linux 64bit support");
547