com_android_internal_os_Zygote.cpp revision 8f4eab2fccbc333604f1d248dac9a736db96bc6b
1/* 2 * Copyright (C) 2008 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 */ 16 17#define LOG_TAG "Zygote" 18 19// sys/mount.h has to come before linux/fs.h due to redefinition of MS_RDONLY, MS_BIND, etc 20#include <sys/mount.h> 21#include <linux/fs.h> 22 23#include <list> 24#include <sstream> 25#include <string> 26 27#include <fcntl.h> 28#include <grp.h> 29#include <inttypes.h> 30#include <malloc.h> 31#include <mntent.h> 32#include <paths.h> 33#include <signal.h> 34#include <stdlib.h> 35#include <sys/capability.h> 36#include <sys/cdefs.h> 37#include <sys/personality.h> 38#include <sys/prctl.h> 39#include <sys/resource.h> 40#include <sys/stat.h> 41#include <sys/time.h> 42#include <sys/types.h> 43#include <sys/utsname.h> 44#include <sys/wait.h> 45#include <unistd.h> 46 47#include "android-base/logging.h" 48#include <android-base/file.h> 49#include <android-base/stringprintf.h> 50#include <cutils/fs.h> 51#include <cutils/multiuser.h> 52#include <cutils/sched_policy.h> 53#include <private/android_filesystem_config.h> 54#include <utils/String8.h> 55#include <selinux/android.h> 56#include <processgroup/processgroup.h> 57 58#include "core_jni_helpers.h" 59#include <nativehelper/JNIHelp.h> 60#include <nativehelper/ScopedLocalRef.h> 61#include <nativehelper/ScopedPrimitiveArray.h> 62#include <nativehelper/ScopedUtfChars.h> 63#include "fd_utils.h" 64 65#include "nativebridge/native_bridge.h" 66 67namespace { 68 69using android::String8; 70using android::base::StringPrintf; 71using android::base::WriteStringToFile; 72 73static pid_t gSystemServerPid = 0; 74 75static const char kZygoteClassName[] = "com/android/internal/os/Zygote"; 76static jclass gZygoteClass; 77static jmethodID gCallPostForkChildHooks; 78 79// Must match values in com.android.internal.os.Zygote. 80enum MountExternalKind { 81 MOUNT_EXTERNAL_NONE = 0, 82 MOUNT_EXTERNAL_DEFAULT = 1, 83 MOUNT_EXTERNAL_READ = 2, 84 MOUNT_EXTERNAL_WRITE = 3, 85}; 86 87static void RuntimeAbort(JNIEnv* env, int line, const char* msg) { 88 std::ostringstream oss; 89 oss << __FILE__ << ":" << line << ": " << msg; 90 env->FatalError(oss.str().c_str()); 91} 92 93// This signal handler is for zygote mode, since the zygote must reap its children 94static void SigChldHandler(int /*signal_number*/) { 95 pid_t pid; 96 int status; 97 98 // It's necessary to save and restore the errno during this function. 99 // Since errno is stored per thread, changing it here modifies the errno 100 // on the thread on which this signal handler executes. If a signal occurs 101 // between a call and an errno check, it's possible to get the errno set 102 // here. 103 // See b/23572286 for extra information. 104 int saved_errno = errno; 105 106 while ((pid = waitpid(-1, &status, WNOHANG)) > 0) { 107 // Log process-death status that we care about. In general it is 108 // not safe to call LOG(...) from a signal handler because of 109 // possible reentrancy. However, we know a priori that the 110 // current implementation of LOG() is safe to call from a SIGCHLD 111 // handler in the zygote process. If the LOG() implementation 112 // changes its locking strategy or its use of syscalls within the 113 // lazy-init critical section, its use here may become unsafe. 114 if (WIFEXITED(status)) { 115 ALOGI("Process %d exited cleanly (%d)", pid, WEXITSTATUS(status)); 116 } else if (WIFSIGNALED(status)) { 117 ALOGI("Process %d exited due to signal (%d)", pid, WTERMSIG(status)); 118 if (WCOREDUMP(status)) { 119 ALOGI("Process %d dumped core.", pid); 120 } 121 } 122 123 // If the just-crashed process is the system_server, bring down zygote 124 // so that it is restarted by init and system server will be restarted 125 // from there. 126 if (pid == gSystemServerPid) { 127 ALOGE("Exit zygote because system server (%d) has terminated", pid); 128 kill(getpid(), SIGKILL); 129 } 130 } 131 132 // Note that we shouldn't consider ECHILD an error because 133 // the secondary zygote might have no children left to wait for. 134 if (pid < 0 && errno != ECHILD) { 135 ALOGW("Zygote SIGCHLD error in waitpid: %s", strerror(errno)); 136 } 137 138 errno = saved_errno; 139} 140 141// Configures the SIGCHLD handler for the zygote process. This is configured 142// very late, because earlier in the runtime we may fork() and exec() 143// other processes, and we want to waitpid() for those rather than 144// have them be harvested immediately. 145// 146// This ends up being called repeatedly before each fork(), but there's 147// no real harm in that. 148static void SetSigChldHandler() { 149 struct sigaction sa; 150 memset(&sa, 0, sizeof(sa)); 151 sa.sa_handler = SigChldHandler; 152 153 int err = sigaction(SIGCHLD, &sa, NULL); 154 if (err < 0) { 155 ALOGW("Error setting SIGCHLD handler: %s", strerror(errno)); 156 } 157} 158 159// Sets the SIGCHLD handler back to default behavior in zygote children. 160static void UnsetSigChldHandler() { 161 struct sigaction sa; 162 memset(&sa, 0, sizeof(sa)); 163 sa.sa_handler = SIG_DFL; 164 165 int err = sigaction(SIGCHLD, &sa, NULL); 166 if (err < 0) { 167 ALOGW("Error unsetting SIGCHLD handler: %s", strerror(errno)); 168 } 169} 170 171// Calls POSIX setgroups() using the int[] object as an argument. 172// A NULL argument is tolerated. 173static void SetGids(JNIEnv* env, jintArray javaGids) { 174 if (javaGids == NULL) { 175 return; 176 } 177 178 ScopedIntArrayRO gids(env, javaGids); 179 if (gids.get() == NULL) { 180 RuntimeAbort(env, __LINE__, "Getting gids int array failed"); 181 } 182 int rc = setgroups(gids.size(), reinterpret_cast<const gid_t*>(&gids[0])); 183 if (rc == -1) { 184 std::ostringstream oss; 185 oss << "setgroups failed: " << strerror(errno) << ", gids.size=" << gids.size(); 186 RuntimeAbort(env, __LINE__, oss.str().c_str()); 187 } 188} 189 190// Sets the resource limits via setrlimit(2) for the values in the 191// two-dimensional array of integers that's passed in. The second dimension 192// contains a tuple of length 3: (resource, rlim_cur, rlim_max). NULL is 193// treated as an empty array. 194static void SetRLimits(JNIEnv* env, jobjectArray javaRlimits) { 195 if (javaRlimits == NULL) { 196 return; 197 } 198 199 rlimit rlim; 200 memset(&rlim, 0, sizeof(rlim)); 201 202 for (int i = 0; i < env->GetArrayLength(javaRlimits); ++i) { 203 ScopedLocalRef<jobject> javaRlimitObject(env, env->GetObjectArrayElement(javaRlimits, i)); 204 ScopedIntArrayRO javaRlimit(env, reinterpret_cast<jintArray>(javaRlimitObject.get())); 205 if (javaRlimit.size() != 3) { 206 RuntimeAbort(env, __LINE__, "rlimits array must have a second dimension of size 3"); 207 } 208 209 rlim.rlim_cur = javaRlimit[1]; 210 rlim.rlim_max = javaRlimit[2]; 211 212 int rc = setrlimit(javaRlimit[0], &rlim); 213 if (rc == -1) { 214 ALOGE("setrlimit(%d, {%ld, %ld}) failed", javaRlimit[0], rlim.rlim_cur, 215 rlim.rlim_max); 216 RuntimeAbort(env, __LINE__, "setrlimit failed"); 217 } 218 } 219} 220 221// The debug malloc library needs to know whether it's the zygote or a child. 222extern "C" int gMallocLeakZygoteChild; 223 224static void PreApplicationInit() { 225 // The child process sets this to indicate it's not the zygote. 226 gMallocLeakZygoteChild = 1; 227 228 // Set the jemalloc decay time to 1. 229 mallopt(M_DECAY_TIME, 1); 230} 231 232static void EnableKeepCapabilities(JNIEnv* env) { 233 int rc = prctl(PR_SET_KEEPCAPS, 1, 0, 0, 0); 234 if (rc == -1) { 235 RuntimeAbort(env, __LINE__, "prctl(PR_SET_KEEPCAPS) failed"); 236 } 237} 238 239static void DropCapabilitiesBoundingSet(JNIEnv* env) { 240 for (int i = 0; prctl(PR_CAPBSET_READ, i, 0, 0, 0) >= 0; i++) { 241 int rc = prctl(PR_CAPBSET_DROP, i, 0, 0, 0); 242 if (rc == -1) { 243 if (errno == EINVAL) { 244 ALOGE("prctl(PR_CAPBSET_DROP) failed with EINVAL. Please verify " 245 "your kernel is compiled with file capabilities support"); 246 } else { 247 ALOGE("prctl(PR_CAPBSET_DROP, %d) failed: %s", i, strerror(errno)); 248 RuntimeAbort(env, __LINE__, "prctl(PR_CAPBSET_DROP) failed"); 249 } 250 } 251 } 252} 253 254static void SetInheritable(JNIEnv* env, uint64_t inheritable) { 255 __user_cap_header_struct capheader; 256 memset(&capheader, 0, sizeof(capheader)); 257 capheader.version = _LINUX_CAPABILITY_VERSION_3; 258 capheader.pid = 0; 259 260 __user_cap_data_struct capdata[2]; 261 if (capget(&capheader, &capdata[0]) == -1) { 262 ALOGE("capget failed: %s", strerror(errno)); 263 RuntimeAbort(env, __LINE__, "capget failed"); 264 } 265 266 capdata[0].inheritable = inheritable; 267 capdata[1].inheritable = inheritable >> 32; 268 269 if (capset(&capheader, &capdata[0]) == -1) { 270 ALOGE("capset(inh=%" PRIx64 ") failed: %s", inheritable, strerror(errno)); 271 RuntimeAbort(env, __LINE__, "capset failed"); 272 } 273} 274 275static void SetCapabilities(JNIEnv* env, uint64_t permitted, uint64_t effective, 276 uint64_t inheritable) { 277 __user_cap_header_struct capheader; 278 memset(&capheader, 0, sizeof(capheader)); 279 capheader.version = _LINUX_CAPABILITY_VERSION_3; 280 capheader.pid = 0; 281 282 __user_cap_data_struct capdata[2]; 283 memset(&capdata, 0, sizeof(capdata)); 284 capdata[0].effective = effective; 285 capdata[1].effective = effective >> 32; 286 capdata[0].permitted = permitted; 287 capdata[1].permitted = permitted >> 32; 288 capdata[0].inheritable = inheritable; 289 capdata[1].inheritable = inheritable >> 32; 290 291 if (capset(&capheader, &capdata[0]) == -1) { 292 ALOGE("capset(perm=%" PRIx64 ", eff=%" PRIx64 ", inh=%" PRIx64 ") failed: %s", permitted, 293 effective, inheritable, strerror(errno)); 294 RuntimeAbort(env, __LINE__, "capset failed"); 295 } 296} 297 298static void SetSchedulerPolicy(JNIEnv* env) { 299 errno = -set_sched_policy(0, SP_DEFAULT); 300 if (errno != 0) { 301 ALOGE("set_sched_policy(0, SP_DEFAULT) failed"); 302 RuntimeAbort(env, __LINE__, "set_sched_policy(0, SP_DEFAULT) failed"); 303 } 304} 305 306static int UnmountTree(const char* path) { 307 size_t path_len = strlen(path); 308 309 FILE* fp = setmntent("/proc/mounts", "r"); 310 if (fp == NULL) { 311 ALOGE("Error opening /proc/mounts: %s", strerror(errno)); 312 return -errno; 313 } 314 315 // Some volumes can be stacked on each other, so force unmount in 316 // reverse order to give us the best chance of success. 317 std::list<std::string> toUnmount; 318 mntent* mentry; 319 while ((mentry = getmntent(fp)) != NULL) { 320 if (strncmp(mentry->mnt_dir, path, path_len) == 0) { 321 toUnmount.push_front(std::string(mentry->mnt_dir)); 322 } 323 } 324 endmntent(fp); 325 326 for (auto path : toUnmount) { 327 if (umount2(path.c_str(), MNT_DETACH)) { 328 ALOGW("Failed to unmount %s: %s", path.c_str(), strerror(errno)); 329 } 330 } 331 return 0; 332} 333 334// Create a private mount namespace and bind mount appropriate emulated 335// storage for the given user. 336static bool MountEmulatedStorage(uid_t uid, jint mount_mode, 337 bool force_mount_namespace) { 338 // See storage config details at http://source.android.com/tech/storage/ 339 340 String8 storageSource; 341 if (mount_mode == MOUNT_EXTERNAL_DEFAULT) { 342 storageSource = "/mnt/runtime/default"; 343 } else if (mount_mode == MOUNT_EXTERNAL_READ) { 344 storageSource = "/mnt/runtime/read"; 345 } else if (mount_mode == MOUNT_EXTERNAL_WRITE) { 346 storageSource = "/mnt/runtime/write"; 347 } else if (!force_mount_namespace) { 348 // Sane default of no storage visible 349 return true; 350 } 351 352 // Create a second private mount namespace for our process 353 if (unshare(CLONE_NEWNS) == -1) { 354 ALOGW("Failed to unshare(): %s", strerror(errno)); 355 return false; 356 } 357 358 // Handle force_mount_namespace with MOUNT_EXTERNAL_NONE. 359 if (mount_mode == MOUNT_EXTERNAL_NONE) { 360 return true; 361 } 362 363 if (TEMP_FAILURE_RETRY(mount(storageSource.string(), "/storage", 364 NULL, MS_BIND | MS_REC | MS_SLAVE, NULL)) == -1) { 365 ALOGW("Failed to mount %s to /storage: %s", storageSource.string(), strerror(errno)); 366 return false; 367 } 368 369 // Mount user-specific symlink helper into place 370 userid_t user_id = multiuser_get_user_id(uid); 371 const String8 userSource(String8::format("/mnt/user/%d", user_id)); 372 if (fs_prepare_dir(userSource.string(), 0751, 0, 0) == -1) { 373 return false; 374 } 375 if (TEMP_FAILURE_RETRY(mount(userSource.string(), "/storage/self", 376 NULL, MS_BIND, NULL)) == -1) { 377 ALOGW("Failed to mount %s to /storage/self: %s", userSource.string(), strerror(errno)); 378 return false; 379 } 380 381 return true; 382} 383 384static bool NeedsNoRandomizeWorkaround() { 385#if !defined(__arm__) 386 return false; 387#else 388 int major; 389 int minor; 390 struct utsname uts; 391 if (uname(&uts) == -1) { 392 return false; 393 } 394 395 if (sscanf(uts.release, "%d.%d", &major, &minor) != 2) { 396 return false; 397 } 398 399 // Kernels before 3.4.* need the workaround. 400 return (major < 3) || ((major == 3) && (minor < 4)); 401#endif 402} 403 404// Utility to close down the Zygote socket file descriptors while 405// the child is still running as root with Zygote's privileges. Each 406// descriptor (if any) is closed via dup2(), replacing it with a valid 407// (open) descriptor to /dev/null. 408 409static void DetachDescriptors(JNIEnv* env, jintArray fdsToClose) { 410 if (!fdsToClose) { 411 return; 412 } 413 jsize count = env->GetArrayLength(fdsToClose); 414 ScopedIntArrayRO ar(env, fdsToClose); 415 if (ar.get() == NULL) { 416 RuntimeAbort(env, __LINE__, "Bad fd array"); 417 } 418 jsize i; 419 int devnull; 420 for (i = 0; i < count; i++) { 421 devnull = open("/dev/null", O_RDWR); 422 if (devnull < 0) { 423 ALOGE("Failed to open /dev/null: %s", strerror(errno)); 424 RuntimeAbort(env, __LINE__, "Failed to open /dev/null"); 425 continue; 426 } 427 ALOGV("Switching descriptor %d to /dev/null: %s", ar[i], strerror(errno)); 428 if (dup2(devnull, ar[i]) < 0) { 429 ALOGE("Failed dup2() on descriptor %d: %s", ar[i], strerror(errno)); 430 RuntimeAbort(env, __LINE__, "Failed dup2()"); 431 } 432 close(devnull); 433 } 434} 435 436void SetThreadName(const char* thread_name) { 437 bool hasAt = false; 438 bool hasDot = false; 439 const char* s = thread_name; 440 while (*s) { 441 if (*s == '.') { 442 hasDot = true; 443 } else if (*s == '@') { 444 hasAt = true; 445 } 446 s++; 447 } 448 const int len = s - thread_name; 449 if (len < 15 || hasAt || !hasDot) { 450 s = thread_name; 451 } else { 452 s = thread_name + len - 15; 453 } 454 // pthread_setname_np fails rather than truncating long strings. 455 char buf[16]; // MAX_TASK_COMM_LEN=16 is hard-coded into bionic 456 strlcpy(buf, s, sizeof(buf)-1); 457 errno = pthread_setname_np(pthread_self(), buf); 458 if (errno != 0) { 459 ALOGW("Unable to set the name of current thread to '%s': %s", buf, strerror(errno)); 460 } 461} 462 463// The list of open zygote file descriptors. 464static FileDescriptorTable* gOpenFdTable = NULL; 465 466static void FillFileDescriptorVector(JNIEnv* env, 467 jintArray java_fds, 468 std::vector<int>* fds) { 469 CHECK(fds != nullptr); 470 if (java_fds != nullptr) { 471 ScopedIntArrayRO ar(env, java_fds); 472 if (ar.get() == nullptr) { 473 RuntimeAbort(env, __LINE__, "Bad fd array"); 474 } 475 fds->reserve(ar.size()); 476 for (size_t i = 0; i < ar.size(); ++i) { 477 fds->push_back(ar[i]); 478 } 479 } 480} 481 482// Utility routine to fork zygote and specialize the child process. 483static pid_t ForkAndSpecializeCommon(JNIEnv* env, uid_t uid, gid_t gid, jintArray javaGids, 484 jint runtime_flags, jobjectArray javaRlimits, 485 jlong permittedCapabilities, jlong effectiveCapabilities, 486 jint mount_external, 487 jstring java_se_info, jstring java_se_name, 488 bool is_system_server, jintArray fdsToClose, 489 jintArray fdsToIgnore, 490 jstring instructionSet, jstring dataDir) { 491 SetSigChldHandler(); 492 493 sigset_t sigchld; 494 sigemptyset(&sigchld); 495 sigaddset(&sigchld, SIGCHLD); 496 497 // Temporarily block SIGCHLD during forks. The SIGCHLD handler might 498 // log, which would result in the logging FDs we close being reopened. 499 // This would cause failures because the FDs are not whitelisted. 500 // 501 // Note that the zygote process is single threaded at this point. 502 if (sigprocmask(SIG_BLOCK, &sigchld, nullptr) == -1) { 503 ALOGE("sigprocmask(SIG_SETMASK, { SIGCHLD }) failed: %s", strerror(errno)); 504 RuntimeAbort(env, __LINE__, "Call to sigprocmask(SIG_BLOCK, { SIGCHLD }) failed."); 505 } 506 507 // Close any logging related FDs before we start evaluating the list of 508 // file descriptors. 509 __android_log_close(); 510 511 // If this is the first fork for this zygote, create the open FD table. 512 // If it isn't, we just need to check whether the list of open files has 513 // changed (and it shouldn't in the normal case). 514 std::vector<int> fds_to_ignore; 515 FillFileDescriptorVector(env, fdsToIgnore, &fds_to_ignore); 516 if (gOpenFdTable == NULL) { 517 gOpenFdTable = FileDescriptorTable::Create(fds_to_ignore); 518 if (gOpenFdTable == NULL) { 519 RuntimeAbort(env, __LINE__, "Unable to construct file descriptor table."); 520 } 521 } else if (!gOpenFdTable->Restat(fds_to_ignore)) { 522 RuntimeAbort(env, __LINE__, "Unable to restat file descriptor table."); 523 } 524 525 pid_t pid = fork(); 526 527 if (pid == 0) { 528 PreApplicationInit(); 529 530 // Clean up any descriptors which must be closed immediately 531 DetachDescriptors(env, fdsToClose); 532 533 // Re-open all remaining open file descriptors so that they aren't shared 534 // with the zygote across a fork. 535 if (!gOpenFdTable->ReopenOrDetach()) { 536 RuntimeAbort(env, __LINE__, "Unable to reopen whitelisted descriptors."); 537 } 538 539 if (sigprocmask(SIG_UNBLOCK, &sigchld, nullptr) == -1) { 540 ALOGE("sigprocmask(SIG_SETMASK, { SIGCHLD }) failed: %s", strerror(errno)); 541 RuntimeAbort(env, __LINE__, "Call to sigprocmask(SIG_UNBLOCK, { SIGCHLD }) failed."); 542 } 543 544 // Keep capabilities across UID change, unless we're staying root. 545 if (uid != 0) { 546 EnableKeepCapabilities(env); 547 } 548 549 SetInheritable(env, permittedCapabilities); 550 DropCapabilitiesBoundingSet(env); 551 552 bool use_native_bridge = !is_system_server && (instructionSet != NULL) 553 && android::NativeBridgeAvailable(); 554 if (use_native_bridge) { 555 ScopedUtfChars isa_string(env, instructionSet); 556 use_native_bridge = android::NeedsNativeBridge(isa_string.c_str()); 557 } 558 if (use_native_bridge && dataDir == NULL) { 559 // dataDir should never be null if we need to use a native bridge. 560 // In general, dataDir will never be null for normal applications. It can only happen in 561 // special cases (for isolated processes which are not associated with any app). These are 562 // launched by the framework and should not be emulated anyway. 563 use_native_bridge = false; 564 ALOGW("Native bridge will not be used because dataDir == NULL."); 565 } 566 567 if (!MountEmulatedStorage(uid, mount_external, use_native_bridge)) { 568 ALOGW("Failed to mount emulated storage: %s", strerror(errno)); 569 if (errno == ENOTCONN || errno == EROFS) { 570 // When device is actively encrypting, we get ENOTCONN here 571 // since FUSE was mounted before the framework restarted. 572 // When encrypted device is booting, we get EROFS since 573 // FUSE hasn't been created yet by init. 574 // In either case, continue without external storage. 575 } else { 576 RuntimeAbort(env, __LINE__, "Cannot continue without emulated storage"); 577 } 578 } 579 580 if (!is_system_server) { 581 int rc = createProcessGroup(uid, getpid()); 582 if (rc != 0) { 583 if (rc == -EROFS) { 584 ALOGW("createProcessGroup failed, kernel missing CONFIG_CGROUP_CPUACCT?"); 585 } else { 586 ALOGE("createProcessGroup(%d, %d) failed: %s", uid, pid, strerror(-rc)); 587 } 588 } 589 } 590 591 SetGids(env, javaGids); 592 593 SetRLimits(env, javaRlimits); 594 595 if (use_native_bridge) { 596 ScopedUtfChars isa_string(env, instructionSet); 597 ScopedUtfChars data_dir(env, dataDir); 598 android::PreInitializeNativeBridge(data_dir.c_str(), isa_string.c_str()); 599 } 600 601 int rc = setresgid(gid, gid, gid); 602 if (rc == -1) { 603 ALOGE("setresgid(%d) failed: %s", gid, strerror(errno)); 604 RuntimeAbort(env, __LINE__, "setresgid failed"); 605 } 606 607 rc = setresuid(uid, uid, uid); 608 if (rc == -1) { 609 ALOGE("setresuid(%d) failed: %s", uid, strerror(errno)); 610 RuntimeAbort(env, __LINE__, "setresuid failed"); 611 } 612 613 if (NeedsNoRandomizeWorkaround()) { 614 // Work around ARM kernel ASLR lossage (http://b/5817320). 615 int old_personality = personality(0xffffffff); 616 int new_personality = personality(old_personality | ADDR_NO_RANDOMIZE); 617 if (new_personality == -1) { 618 ALOGW("personality(%d) failed: %s", new_personality, strerror(errno)); 619 } 620 } 621 622 SetCapabilities(env, permittedCapabilities, effectiveCapabilities, permittedCapabilities); 623 624 SetSchedulerPolicy(env); 625 626 const char* se_info_c_str = NULL; 627 ScopedUtfChars* se_info = NULL; 628 if (java_se_info != NULL) { 629 se_info = new ScopedUtfChars(env, java_se_info); 630 se_info_c_str = se_info->c_str(); 631 if (se_info_c_str == NULL) { 632 RuntimeAbort(env, __LINE__, "se_info_c_str == NULL"); 633 } 634 } 635 const char* se_name_c_str = NULL; 636 ScopedUtfChars* se_name = NULL; 637 if (java_se_name != NULL) { 638 se_name = new ScopedUtfChars(env, java_se_name); 639 se_name_c_str = se_name->c_str(); 640 if (se_name_c_str == NULL) { 641 RuntimeAbort(env, __LINE__, "se_name_c_str == NULL"); 642 } 643 } 644 rc = selinux_android_setcontext(uid, is_system_server, se_info_c_str, se_name_c_str); 645 if (rc == -1) { 646 ALOGE("selinux_android_setcontext(%d, %d, \"%s\", \"%s\") failed", uid, 647 is_system_server, se_info_c_str, se_name_c_str); 648 RuntimeAbort(env, __LINE__, "selinux_android_setcontext failed"); 649 } 650 651 // Make it easier to debug audit logs by setting the main thread's name to the 652 // nice name rather than "app_process". 653 if (se_info_c_str == NULL && is_system_server) { 654 se_name_c_str = "system_server"; 655 } 656 if (se_info_c_str != NULL) { 657 SetThreadName(se_name_c_str); 658 } 659 660 delete se_info; 661 delete se_name; 662 663 UnsetSigChldHandler(); 664 665 env->CallStaticVoidMethod(gZygoteClass, gCallPostForkChildHooks, runtime_flags, 666 is_system_server, instructionSet); 667 if (env->ExceptionCheck()) { 668 RuntimeAbort(env, __LINE__, "Error calling post fork hooks."); 669 } 670 } else if (pid > 0) { 671 // the parent process 672 673 // We blocked SIGCHLD prior to a fork, we unblock it here. 674 if (sigprocmask(SIG_UNBLOCK, &sigchld, nullptr) == -1) { 675 ALOGE("sigprocmask(SIG_SETMASK, { SIGCHLD }) failed: %s", strerror(errno)); 676 RuntimeAbort(env, __LINE__, "Call to sigprocmask(SIG_UNBLOCK, { SIGCHLD }) failed."); 677 } 678 } 679 return pid; 680} 681 682static uint64_t GetEffectiveCapabilityMask(JNIEnv* env) { 683 __user_cap_header_struct capheader; 684 memset(&capheader, 0, sizeof(capheader)); 685 capheader.version = _LINUX_CAPABILITY_VERSION_3; 686 capheader.pid = 0; 687 688 __user_cap_data_struct capdata[2]; 689 if (capget(&capheader, &capdata[0]) == -1) { 690 ALOGE("capget failed: %s", strerror(errno)); 691 RuntimeAbort(env, __LINE__, "capget failed"); 692 } 693 694 return capdata[0].effective | 695 (static_cast<uint64_t>(capdata[1].effective) << 32); 696} 697} // anonymous namespace 698 699namespace android { 700 701static void com_android_internal_os_Zygote_nativePreApplicationInit(JNIEnv*, jclass) { 702 PreApplicationInit(); 703} 704 705static jint com_android_internal_os_Zygote_nativeForkAndSpecialize( 706 JNIEnv* env, jclass, jint uid, jint gid, jintArray gids, 707 jint runtime_flags, jobjectArray rlimits, 708 jint mount_external, jstring se_info, jstring se_name, 709 jintArray fdsToClose, 710 jintArray fdsToIgnore, 711 jstring instructionSet, jstring appDataDir) { 712 jlong capabilities = 0; 713 714 // Grant CAP_WAKE_ALARM to the Bluetooth process. 715 // Additionally, allow bluetooth to open packet sockets so it can start the DHCP client. 716 // Grant CAP_SYS_NICE to allow Bluetooth to set RT priority for 717 // audio-related threads. 718 // TODO: consider making such functionality an RPC to netd. 719 if (multiuser_get_app_id(uid) == AID_BLUETOOTH) { 720 capabilities |= (1LL << CAP_WAKE_ALARM); 721 capabilities |= (1LL << CAP_NET_RAW); 722 capabilities |= (1LL << CAP_NET_BIND_SERVICE); 723 capabilities |= (1LL << CAP_SYS_NICE); 724 } 725 726 // Grant CAP_BLOCK_SUSPEND to processes that belong to GID "wakelock" 727 bool gid_wakelock_found = false; 728 if (gid == AID_WAKELOCK) { 729 gid_wakelock_found = true; 730 } else if (gids != NULL) { 731 jsize gids_num = env->GetArrayLength(gids); 732 ScopedIntArrayRO ar(env, gids); 733 if (ar.get() == NULL) { 734 RuntimeAbort(env, __LINE__, "Bad gids array"); 735 } 736 for (int i = 0; i < gids_num; i++) { 737 if (ar[i] == AID_WAKELOCK) { 738 gid_wakelock_found = true; 739 break; 740 } 741 } 742 } 743 if (gid_wakelock_found) { 744 capabilities |= (1LL << CAP_BLOCK_SUSPEND); 745 } 746 747 // Containers run without some capabilities, so drop any caps that are not 748 // available. 749 capabilities &= GetEffectiveCapabilityMask(env); 750 751 return ForkAndSpecializeCommon(env, uid, gid, gids, runtime_flags, 752 rlimits, capabilities, capabilities, mount_external, se_info, 753 se_name, false, fdsToClose, fdsToIgnore, instructionSet, appDataDir); 754} 755 756static jint com_android_internal_os_Zygote_nativeForkSystemServer( 757 JNIEnv* env, jclass, uid_t uid, gid_t gid, jintArray gids, 758 jint runtime_flags, jobjectArray rlimits, jlong permittedCapabilities, 759 jlong effectiveCapabilities) { 760 pid_t pid = ForkAndSpecializeCommon(env, uid, gid, gids, 761 runtime_flags, rlimits, 762 permittedCapabilities, effectiveCapabilities, 763 MOUNT_EXTERNAL_DEFAULT, NULL, NULL, true, NULL, 764 NULL, NULL, NULL); 765 if (pid > 0) { 766 // The zygote process checks whether the child process has died or not. 767 ALOGI("System server process %d has been created", pid); 768 gSystemServerPid = pid; 769 // There is a slight window that the system server process has crashed 770 // but it went unnoticed because we haven't published its pid yet. So 771 // we recheck here just to make sure that all is well. 772 int status; 773 if (waitpid(pid, &status, WNOHANG) == pid) { 774 ALOGE("System server process %d has died. Restarting Zygote!", pid); 775 RuntimeAbort(env, __LINE__, "System server process has died. Restarting Zygote!"); 776 } 777 778 // Assign system_server to the correct memory cgroup. 779 if (!WriteStringToFile(StringPrintf("%d", pid), "/dev/memcg/system/tasks")) { 780 ALOGE("couldn't write %d to /dev/memcg/system/tasks", pid); 781 } 782 } 783 return pid; 784} 785 786static void com_android_internal_os_Zygote_nativeAllowFileAcrossFork( 787 JNIEnv* env, jclass, jstring path) { 788 ScopedUtfChars path_native(env, path); 789 const char* path_cstr = path_native.c_str(); 790 if (!path_cstr) { 791 RuntimeAbort(env, __LINE__, "path_cstr == NULL"); 792 } 793 FileDescriptorWhitelist::Get()->Allow(path_cstr); 794} 795 796static void com_android_internal_os_Zygote_nativeUnmountStorageOnInit(JNIEnv* env, jclass) { 797 // Zygote process unmount root storage space initially before every child processes are forked. 798 // Every forked child processes (include SystemServer) only mount their own root storage space 799 // and no need unmount storage operation in MountEmulatedStorage method. 800 // Zygote process does not utilize root storage spaces and unshares its mount namespace below. 801 802 // See storage config details at http://source.android.com/tech/storage/ 803 // Create private mount namespace shared by all children 804 if (unshare(CLONE_NEWNS) == -1) { 805 RuntimeAbort(env, __LINE__, "Failed to unshare()"); 806 return; 807 } 808 809 // Mark rootfs as being a slave so that changes from default 810 // namespace only flow into our children. 811 if (mount("rootfs", "/", nullptr, (MS_SLAVE | MS_REC), nullptr) == -1) { 812 RuntimeAbort(env, __LINE__, "Failed to mount() rootfs as MS_SLAVE"); 813 return; 814 } 815 816 // Create a staging tmpfs that is shared by our children; they will 817 // bind mount storage into their respective private namespaces, which 818 // are isolated from each other. 819 const char* target_base = getenv("EMULATED_STORAGE_TARGET"); 820 if (target_base != nullptr) { 821#define STRINGIFY_UID(x) __STRING(x) 822 if (mount("tmpfs", target_base, "tmpfs", MS_NOSUID | MS_NODEV, 823 "uid=0,gid=" STRINGIFY_UID(AID_SDCARD_R) ",mode=0751") == -1) { 824 ALOGE("Failed to mount tmpfs to %s", target_base); 825 RuntimeAbort(env, __LINE__, "Failed to mount tmpfs"); 826 return; 827 } 828#undef STRINGIFY_UID 829 } 830 831 UnmountTree("/storage"); 832} 833 834static const JNINativeMethod gMethods[] = { 835 { "nativeForkAndSpecialize", 836 "(II[II[[IILjava/lang/String;Ljava/lang/String;[I[ILjava/lang/String;Ljava/lang/String;)I", 837 (void *) com_android_internal_os_Zygote_nativeForkAndSpecialize }, 838 { "nativeForkSystemServer", "(II[II[[IJJ)I", 839 (void *) com_android_internal_os_Zygote_nativeForkSystemServer }, 840 { "nativeAllowFileAcrossFork", "(Ljava/lang/String;)V", 841 (void *) com_android_internal_os_Zygote_nativeAllowFileAcrossFork }, 842 { "nativeUnmountStorageOnInit", "()V", 843 (void *) com_android_internal_os_Zygote_nativeUnmountStorageOnInit }, 844 { "nativePreApplicationInit", "()V", 845 (void *) com_android_internal_os_Zygote_nativePreApplicationInit } 846}; 847 848int register_com_android_internal_os_Zygote(JNIEnv* env) { 849 gZygoteClass = MakeGlobalRefOrDie(env, FindClassOrDie(env, kZygoteClassName)); 850 gCallPostForkChildHooks = GetStaticMethodIDOrDie(env, gZygoteClass, "callPostForkChildHooks", 851 "(IZLjava/lang/String;)V"); 852 853 return RegisterMethodsOrDie(env, "com/android/internal/os/Zygote", gMethods, NELEM(gMethods)); 854} 855} // namespace android 856