utils.cc revision 700a402244a1a423da4f3ba8032459f4b65fa18f
1/* 2 * Copyright (C) 2011 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#include "utils.h" 18 19#include <inttypes.h> 20#include <pthread.h> 21#include <sys/stat.h> 22#include <sys/syscall.h> 23#include <sys/types.h> 24#include <sys/wait.h> 25#include <unistd.h> 26#include <memory> 27 28#include "base/stl_util.h" 29#include "base/unix_file/fd_file.h" 30#include "dex_file-inl.h" 31#include "mirror/art_field-inl.h" 32#include "mirror/art_method-inl.h" 33#include "mirror/class-inl.h" 34#include "mirror/class_loader.h" 35#include "mirror/object-inl.h" 36#include "mirror/object_array-inl.h" 37#include "mirror/string.h" 38#include "object_utils.h" 39#include "os.h" 40#include "scoped_thread_state_change.h" 41#include "utf-inl.h" 42 43#if !defined(HAVE_POSIX_CLOCKS) 44#include <sys/time.h> 45#endif 46 47#if defined(HAVE_PRCTL) 48#include <sys/prctl.h> 49#endif 50 51#if defined(__APPLE__) 52#include "AvailabilityMacros.h" // For MAC_OS_X_VERSION_MAX_ALLOWED 53#include <sys/syscall.h> 54#endif 55 56#include <backtrace/Backtrace.h> // For DumpNativeStack. 57 58#if defined(__linux__) 59#include <linux/unistd.h> 60#endif 61 62namespace art { 63 64pid_t GetTid() { 65#if defined(__APPLE__) 66 uint64_t owner; 67 CHECK_PTHREAD_CALL(pthread_threadid_np, (NULL, &owner), __FUNCTION__); // Requires Mac OS 10.6 68 return owner; 69#else 70 // Neither bionic nor glibc exposes gettid(2). 71 return syscall(__NR_gettid); 72#endif 73} 74 75std::string GetThreadName(pid_t tid) { 76 std::string result; 77 if (ReadFileToString(StringPrintf("/proc/self/task/%d/comm", tid), &result)) { 78 result.resize(result.size() - 1); // Lose the trailing '\n'. 79 } else { 80 result = "<unknown>"; 81 } 82 return result; 83} 84 85void GetThreadStack(pthread_t thread, void** stack_base, size_t* stack_size) { 86#if defined(__APPLE__) 87 *stack_size = pthread_get_stacksize_np(thread); 88 void* stack_addr = pthread_get_stackaddr_np(thread); 89 90 // Check whether stack_addr is the base or end of the stack. 91 // (On Mac OS 10.7, it's the end.) 92 int stack_variable; 93 if (stack_addr > &stack_variable) { 94 *stack_base = reinterpret_cast<byte*>(stack_addr) - *stack_size; 95 } else { 96 *stack_base = stack_addr; 97 } 98#else 99 pthread_attr_t attributes; 100 CHECK_PTHREAD_CALL(pthread_getattr_np, (thread, &attributes), __FUNCTION__); 101 CHECK_PTHREAD_CALL(pthread_attr_getstack, (&attributes, stack_base, stack_size), __FUNCTION__); 102 CHECK_PTHREAD_CALL(pthread_attr_destroy, (&attributes), __FUNCTION__); 103#endif 104} 105 106bool ReadFileToString(const std::string& file_name, std::string* result) { 107 std::unique_ptr<File> file(new File); 108 if (!file->Open(file_name, O_RDONLY)) { 109 return false; 110 } 111 112 std::vector<char> buf(8 * KB); 113 while (true) { 114 int64_t n = TEMP_FAILURE_RETRY(read(file->Fd(), &buf[0], buf.size())); 115 if (n == -1) { 116 return false; 117 } 118 if (n == 0) { 119 return true; 120 } 121 result->append(&buf[0], n); 122 } 123} 124 125std::string GetIsoDate() { 126 time_t now = time(NULL); 127 tm tmbuf; 128 tm* ptm = localtime_r(&now, &tmbuf); 129 return StringPrintf("%04d-%02d-%02d %02d:%02d:%02d", 130 ptm->tm_year + 1900, ptm->tm_mon+1, ptm->tm_mday, 131 ptm->tm_hour, ptm->tm_min, ptm->tm_sec); 132} 133 134uint64_t MilliTime() { 135#if defined(HAVE_POSIX_CLOCKS) 136 timespec now; 137 clock_gettime(CLOCK_MONOTONIC, &now); 138 return static_cast<uint64_t>(now.tv_sec) * UINT64_C(1000) + now.tv_nsec / UINT64_C(1000000); 139#else 140 timeval now; 141 gettimeofday(&now, NULL); 142 return static_cast<uint64_t>(now.tv_sec) * UINT64_C(1000) + now.tv_usec / UINT64_C(1000); 143#endif 144} 145 146uint64_t MicroTime() { 147#if defined(HAVE_POSIX_CLOCKS) 148 timespec now; 149 clock_gettime(CLOCK_MONOTONIC, &now); 150 return static_cast<uint64_t>(now.tv_sec) * UINT64_C(1000000) + now.tv_nsec / UINT64_C(1000); 151#else 152 timeval now; 153 gettimeofday(&now, NULL); 154 return static_cast<uint64_t>(now.tv_sec) * UINT64_C(1000000) + now.tv_usec; 155#endif 156} 157 158uint64_t NanoTime() { 159#if defined(HAVE_POSIX_CLOCKS) 160 timespec now; 161 clock_gettime(CLOCK_MONOTONIC, &now); 162 return static_cast<uint64_t>(now.tv_sec) * UINT64_C(1000000000) + now.tv_nsec; 163#else 164 timeval now; 165 gettimeofday(&now, NULL); 166 return static_cast<uint64_t>(now.tv_sec) * UINT64_C(1000000000) + now.tv_usec * UINT64_C(1000); 167#endif 168} 169 170uint64_t ThreadCpuNanoTime() { 171#if defined(HAVE_POSIX_CLOCKS) 172 timespec now; 173 clock_gettime(CLOCK_THREAD_CPUTIME_ID, &now); 174 return static_cast<uint64_t>(now.tv_sec) * UINT64_C(1000000000) + now.tv_nsec; 175#else 176 UNIMPLEMENTED(WARNING); 177 return -1; 178#endif 179} 180 181void NanoSleep(uint64_t ns) { 182 timespec tm; 183 tm.tv_sec = 0; 184 tm.tv_nsec = ns; 185 nanosleep(&tm, NULL); 186} 187 188void InitTimeSpec(bool absolute, int clock, int64_t ms, int32_t ns, timespec* ts) { 189 int64_t endSec; 190 191 if (absolute) { 192#if !defined(__APPLE__) 193 clock_gettime(clock, ts); 194#else 195 UNUSED(clock); 196 timeval tv; 197 gettimeofday(&tv, NULL); 198 ts->tv_sec = tv.tv_sec; 199 ts->tv_nsec = tv.tv_usec * 1000; 200#endif 201 } else { 202 ts->tv_sec = 0; 203 ts->tv_nsec = 0; 204 } 205 endSec = ts->tv_sec + ms / 1000; 206 if (UNLIKELY(endSec >= 0x7fffffff)) { 207 std::ostringstream ss; 208 LOG(INFO) << "Note: end time exceeds epoch: " << ss.str(); 209 endSec = 0x7ffffffe; 210 } 211 ts->tv_sec = endSec; 212 ts->tv_nsec = (ts->tv_nsec + (ms % 1000) * 1000000) + ns; 213 214 // Catch rollover. 215 if (ts->tv_nsec >= 1000000000L) { 216 ts->tv_sec++; 217 ts->tv_nsec -= 1000000000L; 218 } 219} 220 221std::string PrettyDescriptor(mirror::String* java_descriptor) { 222 if (java_descriptor == NULL) { 223 return "null"; 224 } 225 return PrettyDescriptor(java_descriptor->ToModifiedUtf8()); 226} 227 228std::string PrettyDescriptor(mirror::Class* klass) { 229 if (klass == NULL) { 230 return "null"; 231 } 232 return PrettyDescriptor(klass->GetDescriptor()); 233} 234 235std::string PrettyDescriptor(const std::string& descriptor) { 236 // Count the number of '['s to get the dimensionality. 237 const char* c = descriptor.c_str(); 238 size_t dim = 0; 239 while (*c == '[') { 240 dim++; 241 c++; 242 } 243 244 // Reference or primitive? 245 if (*c == 'L') { 246 // "[[La/b/C;" -> "a.b.C[][]". 247 c++; // Skip the 'L'. 248 } else { 249 // "[[B" -> "byte[][]". 250 // To make life easier, we make primitives look like unqualified 251 // reference types. 252 switch (*c) { 253 case 'B': c = "byte;"; break; 254 case 'C': c = "char;"; break; 255 case 'D': c = "double;"; break; 256 case 'F': c = "float;"; break; 257 case 'I': c = "int;"; break; 258 case 'J': c = "long;"; break; 259 case 'S': c = "short;"; break; 260 case 'Z': c = "boolean;"; break; 261 case 'V': c = "void;"; break; // Used when decoding return types. 262 default: return descriptor; 263 } 264 } 265 266 // At this point, 'c' is a string of the form "fully/qualified/Type;" 267 // or "primitive;". Rewrite the type with '.' instead of '/': 268 std::string result; 269 const char* p = c; 270 while (*p != ';') { 271 char ch = *p++; 272 if (ch == '/') { 273 ch = '.'; 274 } 275 result.push_back(ch); 276 } 277 // ...and replace the semicolon with 'dim' "[]" pairs: 278 while (dim--) { 279 result += "[]"; 280 } 281 return result; 282} 283 284std::string PrettyDescriptor(Primitive::Type type) { 285 std::string descriptor_string(Primitive::Descriptor(type)); 286 return PrettyDescriptor(descriptor_string); 287} 288 289std::string PrettyField(mirror::ArtField* f, bool with_type) { 290 if (f == NULL) { 291 return "null"; 292 } 293 FieldHelper fh(f); 294 std::string result; 295 if (with_type) { 296 result += PrettyDescriptor(fh.GetTypeDescriptor()); 297 result += ' '; 298 } 299 result += PrettyDescriptor(fh.GetDeclaringClassDescriptor()); 300 result += '.'; 301 result += fh.GetName(); 302 return result; 303} 304 305std::string PrettyField(uint32_t field_idx, const DexFile& dex_file, bool with_type) { 306 if (field_idx >= dex_file.NumFieldIds()) { 307 return StringPrintf("<<invalid-field-idx-%d>>", field_idx); 308 } 309 const DexFile::FieldId& field_id = dex_file.GetFieldId(field_idx); 310 std::string result; 311 if (with_type) { 312 result += dex_file.GetFieldTypeDescriptor(field_id); 313 result += ' '; 314 } 315 result += PrettyDescriptor(dex_file.GetFieldDeclaringClassDescriptor(field_id)); 316 result += '.'; 317 result += dex_file.GetFieldName(field_id); 318 return result; 319} 320 321std::string PrettyType(uint32_t type_idx, const DexFile& dex_file) { 322 if (type_idx >= dex_file.NumTypeIds()) { 323 return StringPrintf("<<invalid-type-idx-%d>>", type_idx); 324 } 325 const DexFile::TypeId& type_id = dex_file.GetTypeId(type_idx); 326 return PrettyDescriptor(dex_file.GetTypeDescriptor(type_id)); 327} 328 329std::string PrettyArguments(const char* signature) { 330 std::string result; 331 result += '('; 332 CHECK_EQ(*signature, '('); 333 ++signature; // Skip the '('. 334 while (*signature != ')') { 335 size_t argument_length = 0; 336 while (signature[argument_length] == '[') { 337 ++argument_length; 338 } 339 if (signature[argument_length] == 'L') { 340 argument_length = (strchr(signature, ';') - signature + 1); 341 } else { 342 ++argument_length; 343 } 344 std::string argument_descriptor(signature, argument_length); 345 result += PrettyDescriptor(argument_descriptor); 346 if (signature[argument_length] != ')') { 347 result += ", "; 348 } 349 signature += argument_length; 350 } 351 CHECK_EQ(*signature, ')'); 352 ++signature; // Skip the ')'. 353 result += ')'; 354 return result; 355} 356 357std::string PrettyReturnType(const char* signature) { 358 const char* return_type = strchr(signature, ')'); 359 CHECK(return_type != NULL); 360 ++return_type; // Skip ')'. 361 return PrettyDescriptor(return_type); 362} 363 364std::string PrettyMethod(mirror::ArtMethod* m, bool with_signature) { 365 if (m == nullptr) { 366 return "null"; 367 } 368 MethodHelper mh(m); 369 std::string result(PrettyDescriptor(mh.GetDeclaringClassDescriptor())); 370 result += '.'; 371 result += mh.GetName(); 372 if (UNLIKELY(m->IsFastNative())) { 373 result += "!"; 374 } 375 if (with_signature) { 376 const Signature signature = mh.GetSignature(); 377 std::string sig_as_string(signature.ToString()); 378 if (signature == Signature::NoSignature()) { 379 return result + sig_as_string; 380 } 381 result = PrettyReturnType(sig_as_string.c_str()) + " " + result + 382 PrettyArguments(sig_as_string.c_str()); 383 } 384 return result; 385} 386 387std::string PrettyMethod(uint32_t method_idx, const DexFile& dex_file, bool with_signature) { 388 if (method_idx >= dex_file.NumMethodIds()) { 389 return StringPrintf("<<invalid-method-idx-%d>>", method_idx); 390 } 391 const DexFile::MethodId& method_id = dex_file.GetMethodId(method_idx); 392 std::string result(PrettyDescriptor(dex_file.GetMethodDeclaringClassDescriptor(method_id))); 393 result += '.'; 394 result += dex_file.GetMethodName(method_id); 395 if (with_signature) { 396 const Signature signature = dex_file.GetMethodSignature(method_id); 397 std::string sig_as_string(signature.ToString()); 398 if (signature == Signature::NoSignature()) { 399 return result + sig_as_string; 400 } 401 result = PrettyReturnType(sig_as_string.c_str()) + " " + result + 402 PrettyArguments(sig_as_string.c_str()); 403 } 404 return result; 405} 406 407std::string PrettyTypeOf(mirror::Object* obj) { 408 if (obj == NULL) { 409 return "null"; 410 } 411 if (obj->GetClass() == NULL) { 412 return "(raw)"; 413 } 414 std::string result(PrettyDescriptor(obj->GetClass()->GetDescriptor())); 415 if (obj->IsClass()) { 416 result += "<" + PrettyDescriptor(obj->AsClass()->GetDescriptor()) + ">"; 417 } 418 return result; 419} 420 421std::string PrettyClass(mirror::Class* c) { 422 if (c == NULL) { 423 return "null"; 424 } 425 std::string result; 426 result += "java.lang.Class<"; 427 result += PrettyDescriptor(c); 428 result += ">"; 429 return result; 430} 431 432std::string PrettyClassAndClassLoader(mirror::Class* c) { 433 if (c == NULL) { 434 return "null"; 435 } 436 std::string result; 437 result += "java.lang.Class<"; 438 result += PrettyDescriptor(c); 439 result += ","; 440 result += PrettyTypeOf(c->GetClassLoader()); 441 // TODO: add an identifying hash value for the loader 442 result += ">"; 443 return result; 444} 445 446std::string PrettySize(int64_t byte_count) { 447 // The byte thresholds at which we display amounts. A byte count is displayed 448 // in unit U when kUnitThresholds[U] <= bytes < kUnitThresholds[U+1]. 449 static const int64_t kUnitThresholds[] = { 450 0, // B up to... 451 3*1024, // KB up to... 452 2*1024*1024, // MB up to... 453 1024*1024*1024 // GB from here. 454 }; 455 static const int64_t kBytesPerUnit[] = { 1, KB, MB, GB }; 456 static const char* const kUnitStrings[] = { "B", "KB", "MB", "GB" }; 457 const char* negative_str = ""; 458 if (byte_count < 0) { 459 negative_str = "-"; 460 byte_count = -byte_count; 461 } 462 int i = arraysize(kUnitThresholds); 463 while (--i > 0) { 464 if (byte_count >= kUnitThresholds[i]) { 465 break; 466 } 467 } 468 return StringPrintf("%s%" PRId64 "%s", 469 negative_str, byte_count / kBytesPerUnit[i], kUnitStrings[i]); 470} 471 472std::string PrettyDuration(uint64_t nano_duration) { 473 if (nano_duration == 0) { 474 return "0"; 475 } else { 476 return FormatDuration(nano_duration, GetAppropriateTimeUnit(nano_duration)); 477 } 478} 479 480TimeUnit GetAppropriateTimeUnit(uint64_t nano_duration) { 481 const uint64_t one_sec = 1000 * 1000 * 1000; 482 const uint64_t one_ms = 1000 * 1000; 483 const uint64_t one_us = 1000; 484 if (nano_duration >= one_sec) { 485 return kTimeUnitSecond; 486 } else if (nano_duration >= one_ms) { 487 return kTimeUnitMillisecond; 488 } else if (nano_duration >= one_us) { 489 return kTimeUnitMicrosecond; 490 } else { 491 return kTimeUnitNanosecond; 492 } 493} 494 495uint64_t GetNsToTimeUnitDivisor(TimeUnit time_unit) { 496 const uint64_t one_sec = 1000 * 1000 * 1000; 497 const uint64_t one_ms = 1000 * 1000; 498 const uint64_t one_us = 1000; 499 500 switch (time_unit) { 501 case kTimeUnitSecond: 502 return one_sec; 503 case kTimeUnitMillisecond: 504 return one_ms; 505 case kTimeUnitMicrosecond: 506 return one_us; 507 case kTimeUnitNanosecond: 508 return 1; 509 } 510 return 0; 511} 512 513std::string FormatDuration(uint64_t nano_duration, TimeUnit time_unit) { 514 const char* unit = NULL; 515 uint64_t divisor = GetNsToTimeUnitDivisor(time_unit); 516 uint32_t zero_fill = 1; 517 switch (time_unit) { 518 case kTimeUnitSecond: 519 unit = "s"; 520 zero_fill = 9; 521 break; 522 case kTimeUnitMillisecond: 523 unit = "ms"; 524 zero_fill = 6; 525 break; 526 case kTimeUnitMicrosecond: 527 unit = "us"; 528 zero_fill = 3; 529 break; 530 case kTimeUnitNanosecond: 531 unit = "ns"; 532 zero_fill = 0; 533 break; 534 } 535 536 uint64_t whole_part = nano_duration / divisor; 537 uint64_t fractional_part = nano_duration % divisor; 538 if (fractional_part == 0) { 539 return StringPrintf("%" PRIu64 "%s", whole_part, unit); 540 } else { 541 while ((fractional_part % 1000) == 0) { 542 zero_fill -= 3; 543 fractional_part /= 1000; 544 } 545 if (zero_fill == 3) { 546 return StringPrintf("%" PRIu64 ".%03" PRIu64 "%s", whole_part, fractional_part, unit); 547 } else if (zero_fill == 6) { 548 return StringPrintf("%" PRIu64 ".%06" PRIu64 "%s", whole_part, fractional_part, unit); 549 } else { 550 return StringPrintf("%" PRIu64 ".%09" PRIu64 "%s", whole_part, fractional_part, unit); 551 } 552 } 553} 554 555std::string PrintableString(const std::string& utf) { 556 std::string result; 557 result += '"'; 558 const char* p = utf.c_str(); 559 size_t char_count = CountModifiedUtf8Chars(p); 560 for (size_t i = 0; i < char_count; ++i) { 561 uint16_t ch = GetUtf16FromUtf8(&p); 562 if (ch == '\\') { 563 result += "\\\\"; 564 } else if (ch == '\n') { 565 result += "\\n"; 566 } else if (ch == '\r') { 567 result += "\\r"; 568 } else if (ch == '\t') { 569 result += "\\t"; 570 } else if (NeedsEscaping(ch)) { 571 StringAppendF(&result, "\\u%04x", ch); 572 } else { 573 result += ch; 574 } 575 } 576 result += '"'; 577 return result; 578} 579 580// See http://java.sun.com/j2se/1.5.0/docs/guide/jni/spec/design.html#wp615 for the full rules. 581std::string MangleForJni(const std::string& s) { 582 std::string result; 583 size_t char_count = CountModifiedUtf8Chars(s.c_str()); 584 const char* cp = &s[0]; 585 for (size_t i = 0; i < char_count; ++i) { 586 uint16_t ch = GetUtf16FromUtf8(&cp); 587 if ((ch >= 'A' && ch <= 'Z') || (ch >= 'a' && ch <= 'z') || (ch >= '0' && ch <= '9')) { 588 result.push_back(ch); 589 } else if (ch == '.' || ch == '/') { 590 result += "_"; 591 } else if (ch == '_') { 592 result += "_1"; 593 } else if (ch == ';') { 594 result += "_2"; 595 } else if (ch == '[') { 596 result += "_3"; 597 } else { 598 StringAppendF(&result, "_0%04x", ch); 599 } 600 } 601 return result; 602} 603 604std::string DotToDescriptor(const char* class_name) { 605 std::string descriptor(class_name); 606 std::replace(descriptor.begin(), descriptor.end(), '.', '/'); 607 if (descriptor.length() > 0 && descriptor[0] != '[') { 608 descriptor = "L" + descriptor + ";"; 609 } 610 return descriptor; 611} 612 613std::string DescriptorToDot(const char* descriptor) { 614 size_t length = strlen(descriptor); 615 if (descriptor[0] == 'L' && descriptor[length - 1] == ';') { 616 std::string result(descriptor + 1, length - 2); 617 std::replace(result.begin(), result.end(), '/', '.'); 618 return result; 619 } 620 return descriptor; 621} 622 623std::string DescriptorToName(const char* descriptor) { 624 size_t length = strlen(descriptor); 625 if (descriptor[0] == 'L' && descriptor[length - 1] == ';') { 626 std::string result(descriptor + 1, length - 2); 627 return result; 628 } 629 return descriptor; 630} 631 632std::string JniShortName(mirror::ArtMethod* m) { 633 MethodHelper mh(m); 634 std::string class_name(mh.GetDeclaringClassDescriptor()); 635 // Remove the leading 'L' and trailing ';'... 636 CHECK_EQ(class_name[0], 'L') << class_name; 637 CHECK_EQ(class_name[class_name.size() - 1], ';') << class_name; 638 class_name.erase(0, 1); 639 class_name.erase(class_name.size() - 1, 1); 640 641 std::string method_name(mh.GetName()); 642 643 std::string short_name; 644 short_name += "Java_"; 645 short_name += MangleForJni(class_name); 646 short_name += "_"; 647 short_name += MangleForJni(method_name); 648 return short_name; 649} 650 651std::string JniLongName(mirror::ArtMethod* m) { 652 std::string long_name; 653 long_name += JniShortName(m); 654 long_name += "__"; 655 656 std::string signature(MethodHelper(m).GetSignature().ToString()); 657 signature.erase(0, 1); 658 signature.erase(signature.begin() + signature.find(')'), signature.end()); 659 660 long_name += MangleForJni(signature); 661 662 return long_name; 663} 664 665// Helper for IsValidPartOfMemberNameUtf8(), a bit vector indicating valid low ascii. 666uint32_t DEX_MEMBER_VALID_LOW_ASCII[4] = { 667 0x00000000, // 00..1f low control characters; nothing valid 668 0x03ff2010, // 20..3f digits and symbols; valid: '0'..'9', '$', '-' 669 0x87fffffe, // 40..5f uppercase etc.; valid: 'A'..'Z', '_' 670 0x07fffffe // 60..7f lowercase etc.; valid: 'a'..'z' 671}; 672 673// Helper for IsValidPartOfMemberNameUtf8(); do not call directly. 674bool IsValidPartOfMemberNameUtf8Slow(const char** pUtf8Ptr) { 675 /* 676 * It's a multibyte encoded character. Decode it and analyze. We 677 * accept anything that isn't (a) an improperly encoded low value, 678 * (b) an improper surrogate pair, (c) an encoded '\0', (d) a high 679 * control character, or (e) a high space, layout, or special 680 * character (U+00a0, U+2000..U+200f, U+2028..U+202f, 681 * U+fff0..U+ffff). This is all specified in the dex format 682 * document. 683 */ 684 685 uint16_t utf16 = GetUtf16FromUtf8(pUtf8Ptr); 686 687 // Perform follow-up tests based on the high 8 bits. 688 switch (utf16 >> 8) { 689 case 0x00: 690 // It's only valid if it's above the ISO-8859-1 high space (0xa0). 691 return (utf16 > 0x00a0); 692 case 0xd8: 693 case 0xd9: 694 case 0xda: 695 case 0xdb: 696 // It's a leading surrogate. Check to see that a trailing 697 // surrogate follows. 698 utf16 = GetUtf16FromUtf8(pUtf8Ptr); 699 return (utf16 >= 0xdc00) && (utf16 <= 0xdfff); 700 case 0xdc: 701 case 0xdd: 702 case 0xde: 703 case 0xdf: 704 // It's a trailing surrogate, which is not valid at this point. 705 return false; 706 case 0x20: 707 case 0xff: 708 // It's in the range that has spaces, controls, and specials. 709 switch (utf16 & 0xfff8) { 710 case 0x2000: 711 case 0x2008: 712 case 0x2028: 713 case 0xfff0: 714 case 0xfff8: 715 return false; 716 } 717 break; 718 } 719 return true; 720} 721 722/* Return whether the pointed-at modified-UTF-8 encoded character is 723 * valid as part of a member name, updating the pointer to point past 724 * the consumed character. This will consume two encoded UTF-16 code 725 * points if the character is encoded as a surrogate pair. Also, if 726 * this function returns false, then the given pointer may only have 727 * been partially advanced. 728 */ 729static bool IsValidPartOfMemberNameUtf8(const char** pUtf8Ptr) { 730 uint8_t c = (uint8_t) **pUtf8Ptr; 731 if (LIKELY(c <= 0x7f)) { 732 // It's low-ascii, so check the table. 733 uint32_t wordIdx = c >> 5; 734 uint32_t bitIdx = c & 0x1f; 735 (*pUtf8Ptr)++; 736 return (DEX_MEMBER_VALID_LOW_ASCII[wordIdx] & (1 << bitIdx)) != 0; 737 } 738 739 // It's a multibyte encoded character. Call a non-inline function 740 // for the heavy lifting. 741 return IsValidPartOfMemberNameUtf8Slow(pUtf8Ptr); 742} 743 744bool IsValidMemberName(const char* s) { 745 bool angle_name = false; 746 747 switch (*s) { 748 case '\0': 749 // The empty string is not a valid name. 750 return false; 751 case '<': 752 angle_name = true; 753 s++; 754 break; 755 } 756 757 while (true) { 758 switch (*s) { 759 case '\0': 760 return !angle_name; 761 case '>': 762 return angle_name && s[1] == '\0'; 763 } 764 765 if (!IsValidPartOfMemberNameUtf8(&s)) { 766 return false; 767 } 768 } 769} 770 771enum ClassNameType { kName, kDescriptor }; 772static bool IsValidClassName(const char* s, ClassNameType type, char separator) { 773 int arrayCount = 0; 774 while (*s == '[') { 775 arrayCount++; 776 s++; 777 } 778 779 if (arrayCount > 255) { 780 // Arrays may have no more than 255 dimensions. 781 return false; 782 } 783 784 if (arrayCount != 0) { 785 /* 786 * If we're looking at an array of some sort, then it doesn't 787 * matter if what is being asked for is a class name; the 788 * format looks the same as a type descriptor in that case, so 789 * treat it as such. 790 */ 791 type = kDescriptor; 792 } 793 794 if (type == kDescriptor) { 795 /* 796 * We are looking for a descriptor. Either validate it as a 797 * single-character primitive type, or continue on to check the 798 * embedded class name (bracketed by "L" and ";"). 799 */ 800 switch (*(s++)) { 801 case 'B': 802 case 'C': 803 case 'D': 804 case 'F': 805 case 'I': 806 case 'J': 807 case 'S': 808 case 'Z': 809 // These are all single-character descriptors for primitive types. 810 return (*s == '\0'); 811 case 'V': 812 // Non-array void is valid, but you can't have an array of void. 813 return (arrayCount == 0) && (*s == '\0'); 814 case 'L': 815 // Class name: Break out and continue below. 816 break; 817 default: 818 // Oddball descriptor character. 819 return false; 820 } 821 } 822 823 /* 824 * We just consumed the 'L' that introduces a class name as part 825 * of a type descriptor, or we are looking for an unadorned class 826 * name. 827 */ 828 829 bool sepOrFirst = true; // first character or just encountered a separator. 830 for (;;) { 831 uint8_t c = (uint8_t) *s; 832 switch (c) { 833 case '\0': 834 /* 835 * Premature end for a type descriptor, but valid for 836 * a class name as long as we haven't encountered an 837 * empty component (including the degenerate case of 838 * the empty string ""). 839 */ 840 return (type == kName) && !sepOrFirst; 841 case ';': 842 /* 843 * Invalid character for a class name, but the 844 * legitimate end of a type descriptor. In the latter 845 * case, make sure that this is the end of the string 846 * and that it doesn't end with an empty component 847 * (including the degenerate case of "L;"). 848 */ 849 return (type == kDescriptor) && !sepOrFirst && (s[1] == '\0'); 850 case '/': 851 case '.': 852 if (c != separator) { 853 // The wrong separator character. 854 return false; 855 } 856 if (sepOrFirst) { 857 // Separator at start or two separators in a row. 858 return false; 859 } 860 sepOrFirst = true; 861 s++; 862 break; 863 default: 864 if (!IsValidPartOfMemberNameUtf8(&s)) { 865 return false; 866 } 867 sepOrFirst = false; 868 break; 869 } 870 } 871} 872 873bool IsValidBinaryClassName(const char* s) { 874 return IsValidClassName(s, kName, '.'); 875} 876 877bool IsValidJniClassName(const char* s) { 878 return IsValidClassName(s, kName, '/'); 879} 880 881bool IsValidDescriptor(const char* s) { 882 return IsValidClassName(s, kDescriptor, '/'); 883} 884 885void Split(const std::string& s, char separator, std::vector<std::string>& result) { 886 const char* p = s.data(); 887 const char* end = p + s.size(); 888 while (p != end) { 889 if (*p == separator) { 890 ++p; 891 } else { 892 const char* start = p; 893 while (++p != end && *p != separator) { 894 // Skip to the next occurrence of the separator. 895 } 896 result.push_back(std::string(start, p - start)); 897 } 898 } 899} 900 901std::string Trim(std::string s) { 902 std::string result; 903 unsigned int start_index = 0; 904 unsigned int end_index = s.size() - 1; 905 906 // Skip initial whitespace. 907 while (start_index < s.size()) { 908 if (!isspace(s[start_index])) { 909 break; 910 } 911 start_index++; 912 } 913 914 // Skip terminating whitespace. 915 while (end_index >= start_index) { 916 if (!isspace(s[end_index])) { 917 break; 918 } 919 end_index--; 920 } 921 922 // All spaces, no beef. 923 if (end_index < start_index) { 924 return ""; 925 } 926 // Start_index is the first non-space, end_index is the last one. 927 return s.substr(start_index, end_index - start_index + 1); 928} 929 930template <typename StringT> 931std::string Join(std::vector<StringT>& strings, char separator) { 932 if (strings.empty()) { 933 return ""; 934 } 935 936 std::string result(strings[0]); 937 for (size_t i = 1; i < strings.size(); ++i) { 938 result += separator; 939 result += strings[i]; 940 } 941 return result; 942} 943 944// Explicit instantiations. 945template std::string Join<std::string>(std::vector<std::string>& strings, char separator); 946template std::string Join<const char*>(std::vector<const char*>& strings, char separator); 947template std::string Join<char*>(std::vector<char*>& strings, char separator); 948 949bool StartsWith(const std::string& s, const char* prefix) { 950 return s.compare(0, strlen(prefix), prefix) == 0; 951} 952 953bool EndsWith(const std::string& s, const char* suffix) { 954 size_t suffix_length = strlen(suffix); 955 size_t string_length = s.size(); 956 if (suffix_length > string_length) { 957 return false; 958 } 959 size_t offset = string_length - suffix_length; 960 return s.compare(offset, suffix_length, suffix) == 0; 961} 962 963void SetThreadName(const char* thread_name) { 964 int hasAt = 0; 965 int hasDot = 0; 966 const char* s = thread_name; 967 while (*s) { 968 if (*s == '.') { 969 hasDot = 1; 970 } else if (*s == '@') { 971 hasAt = 1; 972 } 973 s++; 974 } 975 int len = s - thread_name; 976 if (len < 15 || hasAt || !hasDot) { 977 s = thread_name; 978 } else { 979 s = thread_name + len - 15; 980 } 981#if defined(HAVE_ANDROID_PTHREAD_SETNAME_NP) 982 // pthread_setname_np fails rather than truncating long strings. 983 char buf[16]; // MAX_TASK_COMM_LEN=16 is hard-coded into bionic 984 strncpy(buf, s, sizeof(buf)-1); 985 buf[sizeof(buf)-1] = '\0'; 986 errno = pthread_setname_np(pthread_self(), buf); 987 if (errno != 0) { 988 PLOG(WARNING) << "Unable to set the name of current thread to '" << buf << "'"; 989 } 990#elif defined(__APPLE__) && MAC_OS_X_VERSION_MAX_ALLOWED >= 1060 991 pthread_setname_np(thread_name); 992#elif defined(HAVE_PRCTL) 993 prctl(PR_SET_NAME, (unsigned long) s, 0, 0, 0); // NOLINT (unsigned long) 994#else 995 UNIMPLEMENTED(WARNING) << thread_name; 996#endif 997} 998 999void GetTaskStats(pid_t tid, char* state, int* utime, int* stime, int* task_cpu) { 1000 *utime = *stime = *task_cpu = 0; 1001 std::string stats; 1002 if (!ReadFileToString(StringPrintf("/proc/self/task/%d/stat", tid), &stats)) { 1003 return; 1004 } 1005 // Skip the command, which may contain spaces. 1006 stats = stats.substr(stats.find(')') + 2); 1007 // Extract the three fields we care about. 1008 std::vector<std::string> fields; 1009 Split(stats, ' ', fields); 1010 *state = fields[0][0]; 1011 *utime = strtoull(fields[11].c_str(), NULL, 10); 1012 *stime = strtoull(fields[12].c_str(), NULL, 10); 1013 *task_cpu = strtoull(fields[36].c_str(), NULL, 10); 1014} 1015 1016std::string GetSchedulerGroupName(pid_t tid) { 1017 // /proc/<pid>/cgroup looks like this: 1018 // 2:devices:/ 1019 // 1:cpuacct,cpu:/ 1020 // We want the third field from the line whose second field contains the "cpu" token. 1021 std::string cgroup_file; 1022 if (!ReadFileToString(StringPrintf("/proc/self/task/%d/cgroup", tid), &cgroup_file)) { 1023 return ""; 1024 } 1025 std::vector<std::string> cgroup_lines; 1026 Split(cgroup_file, '\n', cgroup_lines); 1027 for (size_t i = 0; i < cgroup_lines.size(); ++i) { 1028 std::vector<std::string> cgroup_fields; 1029 Split(cgroup_lines[i], ':', cgroup_fields); 1030 std::vector<std::string> cgroups; 1031 Split(cgroup_fields[1], ',', cgroups); 1032 for (size_t i = 0; i < cgroups.size(); ++i) { 1033 if (cgroups[i] == "cpu") { 1034 return cgroup_fields[2].substr(1); // Skip the leading slash. 1035 } 1036 } 1037 } 1038 return ""; 1039} 1040 1041void DumpNativeStack(std::ostream& os, pid_t tid, const char* prefix, 1042 mirror::ArtMethod* current_method) { 1043 // We may be called from contexts where current_method is not null, so we must assert this. 1044 if (current_method != nullptr) { 1045 Locks::mutator_lock_->AssertSharedHeld(Thread::Current()); 1046 } 1047 std::unique_ptr<Backtrace> backtrace(Backtrace::Create(BACKTRACE_CURRENT_PROCESS, tid)); 1048 if (!backtrace->Unwind(0)) { 1049 os << prefix << "(backtrace::Unwind failed for thread " << tid << ")\n"; 1050 return; 1051 } else if (backtrace->NumFrames() == 0) { 1052 os << prefix << "(no native stack frames for thread " << tid << ")\n"; 1053 return; 1054 } 1055 1056 for (Backtrace::const_iterator it = backtrace->begin(); 1057 it != backtrace->end(); ++it) { 1058 // We produce output like this: 1059 // ] #00 pc 000075bb8 /system/lib/libc.so (unwind_backtrace_thread+536) 1060 // In order for parsing tools to continue to function, the stack dump 1061 // format must at least adhere to this format: 1062 // #XX pc <RELATIVE_ADDR> <FULL_PATH_TO_SHARED_LIBRARY> ... 1063 // The parsers require a single space before and after pc, and two spaces 1064 // after the <RELATIVE_ADDR>. There can be any prefix data before the 1065 // #XX. <RELATIVE_ADDR> has to be a hex number but with no 0x prefix. 1066 os << prefix << StringPrintf("#%02zu pc ", it->num); 1067 if (!it->map) { 1068 os << StringPrintf("%08" PRIxPTR " ???", it->pc); 1069 } else { 1070 os << StringPrintf("%08" PRIxPTR " ", it->pc - it->map->start) 1071 << it->map->name << " ("; 1072 if (!it->func_name.empty()) { 1073 os << it->func_name; 1074 if (it->func_offset != 0) { 1075 os << "+" << it->func_offset; 1076 } 1077 } else if (current_method != nullptr && current_method->IsWithinQuickCode(it->pc)) { 1078 const void* start_of_code = current_method->GetEntryPointFromQuickCompiledCode(); 1079 os << JniLongName(current_method) << "+" 1080 << (it->pc - reinterpret_cast<uintptr_t>(start_of_code)); 1081 } else { 1082 os << "???"; 1083 } 1084 os << ")"; 1085 } 1086 os << "\n"; 1087 } 1088} 1089 1090#if defined(__APPLE__) 1091 1092// TODO: is there any way to get the kernel stack on Mac OS? 1093void DumpKernelStack(std::ostream&, pid_t, const char*, bool) {} 1094 1095#else 1096 1097void DumpKernelStack(std::ostream& os, pid_t tid, const char* prefix, bool include_count) { 1098 if (tid == GetTid()) { 1099 // There's no point showing that we're reading our stack out of /proc! 1100 return; 1101 } 1102 1103 std::string kernel_stack_filename(StringPrintf("/proc/self/task/%d/stack", tid)); 1104 std::string kernel_stack; 1105 if (!ReadFileToString(kernel_stack_filename, &kernel_stack)) { 1106 os << prefix << "(couldn't read " << kernel_stack_filename << ")\n"; 1107 return; 1108 } 1109 1110 std::vector<std::string> kernel_stack_frames; 1111 Split(kernel_stack, '\n', kernel_stack_frames); 1112 // We skip the last stack frame because it's always equivalent to "[<ffffffff>] 0xffffffff", 1113 // which looking at the source appears to be the kernel's way of saying "that's all, folks!". 1114 kernel_stack_frames.pop_back(); 1115 for (size_t i = 0; i < kernel_stack_frames.size(); ++i) { 1116 // Turn "[<ffffffff8109156d>] futex_wait_queue_me+0xcd/0x110" 1117 // into "futex_wait_queue_me+0xcd/0x110". 1118 const char* text = kernel_stack_frames[i].c_str(); 1119 const char* close_bracket = strchr(text, ']'); 1120 if (close_bracket != NULL) { 1121 text = close_bracket + 2; 1122 } 1123 os << prefix; 1124 if (include_count) { 1125 os << StringPrintf("#%02zd ", i); 1126 } 1127 os << text << "\n"; 1128 } 1129} 1130 1131#endif 1132 1133const char* GetAndroidRoot() { 1134 const char* android_root = getenv("ANDROID_ROOT"); 1135 if (android_root == NULL) { 1136 if (OS::DirectoryExists("/system")) { 1137 android_root = "/system"; 1138 } else { 1139 LOG(FATAL) << "ANDROID_ROOT not set and /system does not exist"; 1140 return ""; 1141 } 1142 } 1143 if (!OS::DirectoryExists(android_root)) { 1144 LOG(FATAL) << "Failed to find ANDROID_ROOT directory " << android_root; 1145 return ""; 1146 } 1147 return android_root; 1148} 1149 1150const char* GetAndroidData() { 1151 const char* android_data = getenv("ANDROID_DATA"); 1152 if (android_data == NULL) { 1153 if (OS::DirectoryExists("/data")) { 1154 android_data = "/data"; 1155 } else { 1156 LOG(FATAL) << "ANDROID_DATA not set and /data does not exist"; 1157 return ""; 1158 } 1159 } 1160 if (!OS::DirectoryExists(android_data)) { 1161 LOG(FATAL) << "Failed to find ANDROID_DATA directory " << android_data; 1162 return ""; 1163 } 1164 return android_data; 1165} 1166 1167std::string GetDalvikCacheOrDie(const char* subdir, const bool create_if_absent) { 1168 CHECK(subdir != nullptr); 1169 const char* android_data = GetAndroidData(); 1170 const std::string dalvik_cache_root(StringPrintf("%s/dalvik-cache/", android_data)); 1171 const std::string dalvik_cache = dalvik_cache_root + subdir; 1172 if (create_if_absent && !OS::DirectoryExists(dalvik_cache.c_str())) { 1173 // Don't create the system's /data/dalvik-cache/... because it needs special permissions. 1174 if (strcmp(android_data, "/data") != 0) { 1175 int result = mkdir(dalvik_cache_root.c_str(), 0700); 1176 if (result != 0 && errno != EEXIST) { 1177 PLOG(FATAL) << "Failed to create dalvik-cache directory " << dalvik_cache_root; 1178 return ""; 1179 } 1180 result = mkdir(dalvik_cache.c_str(), 0700); 1181 if (result != 0) { 1182 PLOG(FATAL) << "Failed to create dalvik-cache directory " << dalvik_cache; 1183 return ""; 1184 } 1185 } else { 1186 LOG(FATAL) << "Failed to find dalvik-cache directory " << dalvik_cache; 1187 return ""; 1188 } 1189 } 1190 return dalvik_cache; 1191} 1192 1193std::string GetDalvikCacheFilenameOrDie(const char* location, const char* cache_location) { 1194 if (location[0] != '/') { 1195 LOG(FATAL) << "Expected path in location to be absolute: "<< location; 1196 } 1197 std::string cache_file(&location[1]); // skip leading slash 1198 if (!EndsWith(location, ".dex") && !EndsWith(location, ".art")) { 1199 cache_file += "/"; 1200 cache_file += DexFile::kClassesDex; 1201 } 1202 std::replace(cache_file.begin(), cache_file.end(), '/', '@'); 1203 return StringPrintf("%s/%s", cache_location, cache_file.c_str()); 1204} 1205 1206static void InsertIsaDirectory(const InstructionSet isa, std::string* filename) { 1207 // in = /foo/bar/baz 1208 // out = /foo/bar/<isa>/baz 1209 size_t pos = filename->rfind('/'); 1210 CHECK_NE(pos, std::string::npos) << *filename << " " << isa; 1211 filename->insert(pos, "/", 1); 1212 filename->insert(pos + 1, GetInstructionSetString(isa)); 1213} 1214 1215std::string GetSystemImageFilename(const char* location, const InstructionSet isa) { 1216 // location = /system/framework/boot.art 1217 // filename = /system/framework/<isa>/boot.art 1218 std::string filename(location); 1219 InsertIsaDirectory(isa, &filename); 1220 return filename; 1221} 1222 1223std::string DexFilenameToOdexFilename(const std::string& location, const InstructionSet isa) { 1224 // location = /foo/bar/baz.jar 1225 // odex_location = /foo/bar/<isa>/baz.odex 1226 CHECK_GE(location.size(), 4U) << location; // must be at least .123 1227 std::string odex_location(location); 1228 InsertIsaDirectory(isa, &odex_location); 1229 size_t dot_index = odex_location.size() - 3 - 1; // 3=dex or zip or apk 1230 CHECK_EQ('.', odex_location[dot_index]) << location; 1231 odex_location.resize(dot_index + 1); 1232 CHECK_EQ('.', odex_location[odex_location.size()-1]) << location << " " << odex_location; 1233 odex_location += "odex"; 1234 return odex_location; 1235} 1236 1237bool IsZipMagic(uint32_t magic) { 1238 return (('P' == ((magic >> 0) & 0xff)) && 1239 ('K' == ((magic >> 8) & 0xff))); 1240} 1241 1242bool IsDexMagic(uint32_t magic) { 1243 return DexFile::IsMagicValid(reinterpret_cast<const byte*>(&magic)); 1244} 1245 1246bool IsOatMagic(uint32_t magic) { 1247 return (memcmp(reinterpret_cast<const byte*>(magic), 1248 OatHeader::kOatMagic, 1249 sizeof(OatHeader::kOatMagic)) == 0); 1250} 1251 1252bool Exec(std::vector<std::string>& arg_vector, std::string* error_msg) { 1253 const std::string command_line(Join(arg_vector, ' ')); 1254 1255 CHECK_GE(arg_vector.size(), 1U) << command_line; 1256 1257 // Convert the args to char pointers. 1258 const char* program = arg_vector[0].c_str(); 1259 std::vector<char*> args; 1260 for (size_t i = 0; i < arg_vector.size(); ++i) { 1261 const std::string& arg = arg_vector[i]; 1262 char* arg_str = const_cast<char*>(arg.c_str()); 1263 CHECK(arg_str != nullptr) << i; 1264 args.push_back(arg_str); 1265 } 1266 args.push_back(NULL); 1267 1268 // fork and exec 1269 pid_t pid = fork(); 1270 if (pid == 0) { 1271 // no allocation allowed between fork and exec 1272 1273 // change process groups, so we don't get reaped by ProcessManager 1274 setpgid(0, 0); 1275 1276 execv(program, &args[0]); 1277 1278 PLOG(ERROR) << "Failed to execv(" << command_line << ")"; 1279 exit(1); 1280 } else { 1281 if (pid == -1) { 1282 *error_msg = StringPrintf("Failed to execv(%s) because fork failed: %s", 1283 command_line.c_str(), strerror(errno)); 1284 return false; 1285 } 1286 1287 // wait for subprocess to finish 1288 int status; 1289 pid_t got_pid = TEMP_FAILURE_RETRY(waitpid(pid, &status, 0)); 1290 if (got_pid != pid) { 1291 *error_msg = StringPrintf("Failed after fork for execv(%s) because waitpid failed: " 1292 "wanted %d, got %d: %s", 1293 command_line.c_str(), pid, got_pid, strerror(errno)); 1294 return false; 1295 } 1296 if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) { 1297 *error_msg = StringPrintf("Failed execv(%s) because non-0 exit status", 1298 command_line.c_str()); 1299 return false; 1300 } 1301 } 1302 return true; 1303} 1304 1305} // namespace art 1306