1 /* $NetBSD: linux_exec.c,v 1.86 2006/09/02 06:52:13 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1994, 1995, 1998, 2000 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Christos Zoulas, Frank van der Linden, Eric Haszlakiewicz and 9 * Thor Lancelot Simon. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the NetBSD 22 * Foundation, Inc. and its contributors. 23 * 4. Neither the name of The NetBSD Foundation nor the names of its 24 * contributors may be used to endorse or promote products derived 25 * from this software without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 28 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 37 * POSSIBILITY OF SUCH DAMAGE. 38 */ 39 40 #include <sys/cdefs.h> 41 __KERNEL_RCSID(0, "$NetBSD: linux_exec.c,v 1.86 2006/09/02 06:52:13 christos Exp $"); 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/kernel.h> 46 #include <sys/proc.h> 47 #include <sys/malloc.h> 48 #include <sys/namei.h> 49 #include <sys/vnode.h> 50 #include <sys/mount.h> 51 #include <sys/exec.h> 52 #include <sys/exec_elf.h> 53 54 #include <sys/mman.h> 55 #include <sys/sa.h> 56 #include <sys/syscallargs.h> 57 58 #include <sys/ptrace.h> /* For proc_reparent() */ 59 60 #include <uvm/uvm_extern.h> 61 62 #include <machine/cpu.h> 63 #include <machine/reg.h> 64 65 #include <compat/linux/common/linux_types.h> 66 #include <compat/linux/common/linux_signal.h> 67 #include <compat/linux/common/linux_util.h> 68 #include <compat/linux/common/linux_sched.h> 69 #include <compat/linux/common/linux_machdep.h> 70 #include <compat/linux/common/linux_exec.h> 71 #include <compat/linux/common/linux_futex.h> 72 73 #include <compat/linux/linux_syscallargs.h> 74 #include <compat/linux/linux_syscall.h> 75 #include <compat/linux/common/linux_misc.h> 76 #include <compat/linux/common/linux_errno.h> 77 #include <compat/linux/common/linux_emuldata.h> 78 79 extern struct sysent linux_sysent[]; 80 extern const char * const linux_syscallnames[]; 81 extern char linux_sigcode[], linux_esigcode[]; 82 83 static void linux_e_proc_exec __P((struct proc *, struct exec_package *)); 84 static void linux_e_proc_fork __P((struct proc *, struct proc *, int)); 85 static void linux_e_proc_exit __P((struct proc *)); 86 static void linux_e_proc_init __P((struct proc *, struct proc *, int)); 87 88 #ifdef LINUX_NPTL 89 void linux_userret __P((struct lwp *, void *)); 90 #endif 91 92 /* 93 * Execve(2). Just check the alternate emulation path, and pass it on 94 * to the NetBSD execve(). 95 */ 96 int 97 linux_sys_execve(l, v, retval) 98 struct lwp *l; 99 void *v; 100 register_t *retval; 101 { 102 struct linux_sys_execve_args /* { 103 syscallarg(const char *) path; 104 syscallarg(char **) argv; 105 syscallarg(char **) envp; 106 } */ *uap = v; 107 struct proc *p = l->l_proc; 108 struct sys_execve_args ap; 109 caddr_t sg; 110 111 sg = stackgap_init(p, 0); 112 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path)); 113 114 SCARG(&ap, path) = SCARG(uap, path); 115 SCARG(&ap, argp) = SCARG(uap, argp); 116 SCARG(&ap, envp) = SCARG(uap, envp); 117 118 return sys_execve(l, &ap, retval); 119 } 120 121 /* 122 * Emulation switch. 123 */ 124 125 struct uvm_object *emul_linux_object; 126 127 const struct emul emul_linux = { 128 "linux", 129 "/emul/linux", 130 #ifndef __HAVE_MINIMAL_EMUL 131 0, 132 (const int *)native_to_linux_errno, 133 LINUX_SYS_syscall, 134 LINUX_SYS_NSYSENT, 135 #endif 136 linux_sysent, 137 linux_syscallnames, 138 linux_sendsig, 139 linux_trapsignal, 140 NULL, 141 linux_sigcode, 142 linux_esigcode, 143 &emul_linux_object, 144 linux_setregs, 145 linux_e_proc_exec, 146 linux_e_proc_fork, 147 linux_e_proc_exit, 148 NULL, 149 NULL, 150 #ifdef __HAVE_SYSCALL_INTERN 151 linux_syscall_intern, 152 #else 153 #error Implement __HAVE_SYSCALL_INTERN for this platform 154 #endif 155 NULL, 156 NULL, 157 158 uvm_default_mapaddr, 159 160 linux_usertrap, 161 NULL, 162 }; 163 164 static void 165 linux_e_proc_init(p, parent, forkflags) 166 struct proc *p, *parent; 167 int forkflags; 168 { 169 struct linux_emuldata *e = p->p_emuldata; 170 struct linux_emuldata_shared *s; 171 struct linux_emuldata *ep = NULL; 172 173 if (!e) { 174 /* allocate new Linux emuldata */ 175 MALLOC(e, void *, sizeof(struct linux_emuldata), 176 M_EMULDATA, M_WAITOK); 177 } else { 178 e->s->refs--; 179 if (e->s->refs == 0) 180 FREE(e->s, M_EMULDATA); 181 } 182 183 memset(e, '\0', sizeof(struct linux_emuldata)); 184 185 e->proc = p; 186 187 if (parent) 188 ep = parent->p_emuldata; 189 190 if (forkflags & FORK_SHAREVM) { 191 #ifdef DIAGNOSTIC 192 if (ep == NULL) { 193 killproc(p, "FORK_SHAREVM while emuldata is NULL\n"); 194 return; 195 } 196 #endif 197 s = ep->s; 198 s->refs++; 199 } else { 200 struct vmspace *vm; 201 202 MALLOC(s, void *, sizeof(struct linux_emuldata_shared), 203 M_EMULDATA, M_WAITOK); 204 s->refs = 1; 205 206 /* 207 * Set the process idea of the break to the real value. 208 * For fork, we use parent's vmspace since our's 209 * is not setup at the time of this call and is going 210 * to be copy of parent's anyway. For exec, just 211 * use our own vmspace. 212 */ 213 vm = (parent) ? parent->p_vmspace : p->p_vmspace; 214 s->p_break = vm->vm_daddr + ctob(vm->vm_dsize); 215 216 /* 217 * Linux threads are emulated as NetBSD processes (not lwp) 218 * We use native PID for Linux TID. The Linux TID is the 219 * PID of the first process in the group. It is stored 220 * here 221 */ 222 s->group_pid = p->p_pid; 223 224 /* 225 * Initialize the list of threads in the group 226 */ 227 LIST_INIT(&s->threads); 228 229 s->xstat = 0; 230 s->flags = 0; 231 } 232 233 e->s = s; 234 235 /* 236 * Add this thread in the group thread list 237 */ 238 LIST_INSERT_HEAD(&s->threads, e, threads); 239 240 #ifdef LINUX_NPTL 241 if (ep != NULL) { 242 e->parent_tidptr = ep->parent_tidptr; 243 e->child_tidptr = ep->child_tidptr; 244 e->clone_flags = ep->clone_flags; 245 } 246 #endif /* LINUX_NPTL */ 247 248 p->p_emuldata = e; 249 } 250 251 /* 252 * Allocate new per-process structures. Called when executing Linux 253 * process. We can reuse the old emuldata - if it's not null, 254 * the executed process is of same emulation as original forked one. 255 */ 256 static void 257 linux_e_proc_exec(p, epp) 258 struct proc *p; 259 struct exec_package *epp; 260 { 261 /* exec, use our vmspace */ 262 linux_e_proc_init(p, NULL, 0); 263 } 264 265 /* 266 * Emulation per-process exit hook. 267 */ 268 static void 269 linux_e_proc_exit(p) 270 struct proc *p; 271 { 272 struct linux_emuldata *e = p->p_emuldata; 273 274 #ifdef LINUX_NPTL 275 linux_nptl_proc_exit(p); 276 #endif 277 /* Remove the thread for the group thread list */ 278 LIST_REMOVE(e, threads); 279 280 /* free Linux emuldata and set the pointer to null */ 281 e->s->refs--; 282 if (e->s->refs == 0) 283 FREE(e->s, M_EMULDATA); 284 FREE(e, M_EMULDATA); 285 p->p_emuldata = NULL; 286 } 287 288 /* 289 * Emulation fork hook. 290 */ 291 static void 292 linux_e_proc_fork(p, parent, forkflags) 293 struct proc *p, *parent; 294 int forkflags; 295 { 296 /* 297 * The new process might share some vmspace-related stuff 298 * with parent, depending on fork flags (CLONE_VM et.al). 299 * Force allocation of new base emuldata, and share the 300 * VM-related parts only if necessary. 301 */ 302 p->p_emuldata = NULL; 303 linux_e_proc_init(p, parent, forkflags); 304 305 #ifdef LINUX_NPTL 306 linux_nptl_proc_fork(p, parent, (*linux_userret)); 307 #endif 308 309 return; 310 } 311 312 #ifdef LINUX_NPTL 313 void 314 linux_userret(l, arg) 315 struct lwp *l; 316 void *arg; 317 { 318 struct proc *p = l->l_proc; 319 struct linux_emuldata *led = p->p_emuldata; 320 int error; 321 322 p->p_userret = NULL; 323 324 /* LINUX_CLONE_CHILD_SETTID: copy child's TID to child's memory */ 325 if (led->clone_flags & LINUX_CLONE_CHILD_SETTID) { 326 if ((error = copyout(&l->l_proc->p_pid, 327 led->child_tidptr, sizeof(l->l_proc->p_pid))) != 0) 328 printf("linux_userret: LINUX_CLONE_CHILD_SETTID " 329 "failed (led->child_tidptr = %p, p->p_pid = %d)\n", 330 led->child_tidptr, p->p_pid); 331 } 332 333 /* LINUX_CLONE_SETTLS: allocate a new TLS */ 334 if (led->clone_flags & LINUX_CLONE_SETTLS) { 335 if (linux_set_newtls(l, linux_get_newtls(l)) != 0) 336 printf("linux_userret: linux_set_tls failed"); 337 } 338 339 return; 340 } 341 342 void 343 linux_nptl_proc_exit(p) 344 struct proc *p; 345 { 346 struct linux_emuldata *e = p->p_emuldata; 347 int s; 348 349 /* 350 * Check if we are a thread group leader victim of another 351 * thread doing exit_group(). If we are, change the exit code. 352 */ 353 if ((e->s->group_pid == p->p_pid) && 354 (e->s->flags & LINUX_LES_INEXITGROUP)) { 355 p->p_xstat = e->s->xstat; 356 } 357 358 /* 359 * Members of the thread groups others than the leader should 360 * exit quietely: no zombie stage, no signal. We do that by 361 * reparenting to init. init will collect us and nobody will 362 * notice what happened. 363 */ 364 #ifdef DEBUG_LINUX 365 printf("%s:%d e->s->group_pid = %d, p->p_pid = %d, flags = 0x%x\n", 366 __func__, __LINE__, e->s->group_pid, p->p_pid, e->s->flags); 367 #endif 368 if (e->s->group_pid != p->p_pid) { 369 wakeup(initproc); 370 SCHED_LOCK(s); 371 proc_reparent(p, initproc); 372 SCHED_UNLOCK(s); 373 } 374 375 /* Emulate LINUX_CLONE_CHILD_CLEARTID */ 376 if (e->clear_tid != NULL) { 377 int error; 378 int null = 0; 379 struct linux_sys_futex_args cup; 380 register_t retval; 381 struct lwp *l; 382 383 error = copyout(&null, e->clear_tid, sizeof(null)); 384 #ifdef DEBUG_LINUX 385 if (error != 0) 386 printf("%s: cannot clear TID\n", __func__); 387 #endif 388 389 l = proc_representative_lwp(p); 390 SCARG(&cup, uaddr) = e->clear_tid; 391 SCARG(&cup, op) = LINUX_FUTEX_WAKE; 392 SCARG(&cup, val) = 0x7fffffff; /* Awake everyone */ 393 SCARG(&cup, timeout) = NULL; 394 SCARG(&cup, uaddr2) = NULL; 395 SCARG(&cup, val3) = 0; 396 if ((error = linux_sys_futex(l, &cup, &retval)) != 0) 397 printf("%s: linux_sys_futex failed\n", __func__); 398 } 399 400 return; 401 } 402 403 void 404 linux_nptl_proc_fork(p, parent, luserret) 405 struct proc *p; 406 struct proc *parent; 407 void (luserret)(struct lwp *, void *); 408 { 409 #ifdef LINUX_NPTL 410 struct linux_emuldata *e; 411 #endif 412 e = p->p_emuldata; 413 414 /* LINUX_CLONE_CHILD_CLEARTID: clear TID in child's memory on exit() */ 415 if (e->clone_flags & LINUX_CLONE_CHILD_CLEARTID) 416 e->clear_tid = e->child_tidptr; 417 418 /* LINUX_CLONE_PARENT_SETTID: set child's TID in parent's memory */ 419 if (e->clone_flags & LINUX_CLONE_PARENT_SETTID) { 420 if (copyout_proc(parent, &p->p_pid, 421 e->parent_tidptr, sizeof(p->p_pid)) != 0) 422 printf("%s: LINUX_CLONE_PARENT_SETTID " 423 "failed (e->parent_tidptr = %p, " 424 "parent->p_pid = %d, p->p_pid = %d)\n", 425 __func__, e->parent_tidptr, 426 parent->p_pid, p->p_pid); 427 } 428 429 /* 430 * CLONE_CHILD_SETTID and LINUX_CLONE_SETTLS require child's VM 431 * setup to be completed, we postpone them until userret time. 432 */ 433 if (e->clone_flags & 434 (LINUX_CLONE_CHILD_CLEARTID | LINUX_CLONE_SETTLS)) 435 p->p_userret = (*luserret); 436 437 return; 438 } 439 440 void 441 linux_nptl_proc_init(p, parent) 442 struct proc *p; 443 struct proc *parent; 444 { 445 struct linux_emuldata *e = p->p_emuldata; 446 struct linux_emuldata *ep; 447 448 if ((parent != NULL) && (parent->p_emuldata != NULL)) { 449 ep = parent->p_emuldata; 450 451 e->parent_tidptr = ep->parent_tidptr; 452 e->child_tidptr = ep->child_tidptr; 453 e->clone_flags = ep->clone_flags; 454 } 455 456 return; 457 } 458 459 460 #endif /* LINUX_NPTL */ 461