utils.cc revision 943af7dab1454517c5bd11a31ab99f260afb22d1
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 <unistd.h> 25 26#include "UniquePtr.h" 27#include "base/unix_file/fd_file.h" 28#include "dex_file-inl.h" 29#include "mirror/art_field-inl.h" 30#include "mirror/art_method-inl.h" 31#include "mirror/class-inl.h" 32#include "mirror/class_loader.h" 33#include "mirror/object-inl.h" 34#include "mirror/object_array-inl.h" 35#include "mirror/string.h" 36#include "object_utils.h" 37#include "os.h" 38#include "utf-inl.h" 39 40#if !defined(HAVE_POSIX_CLOCKS) 41#include <sys/time.h> 42#endif 43 44#if defined(HAVE_PRCTL) 45#include <sys/prctl.h> 46#endif 47 48#if defined(__APPLE__) 49#include "AvailabilityMacros.h" // For MAC_OS_X_VERSION_MAX_ALLOWED 50#include <sys/syscall.h> 51#endif 52 53#include <backtrace/Backtrace.h> // For DumpNativeStack. 54 55#if defined(__linux__) 56#include <linux/unistd.h> 57#endif 58 59namespace art { 60 61pid_t GetTid() { 62#if defined(__APPLE__) 63 uint64_t owner; 64 CHECK_PTHREAD_CALL(pthread_threadid_np, (NULL, &owner), __FUNCTION__); // Requires Mac OS 10.6 65 return owner; 66#else 67 // Neither bionic nor glibc exposes gettid(2). 68 return syscall(__NR_gettid); 69#endif 70} 71 72std::string GetThreadName(pid_t tid) { 73 std::string result; 74 if (ReadFileToString(StringPrintf("/proc/self/task/%d/comm", tid), &result)) { 75 result.resize(result.size() - 1); // Lose the trailing '\n'. 76 } else { 77 result = "<unknown>"; 78 } 79 return result; 80} 81 82void GetThreadStack(pthread_t thread, void** stack_base, size_t* stack_size) { 83#if defined(__APPLE__) 84 *stack_size = pthread_get_stacksize_np(thread); 85 void* stack_addr = pthread_get_stackaddr_np(thread); 86 87 // Check whether stack_addr is the base or end of the stack. 88 // (On Mac OS 10.7, it's the end.) 89 int stack_variable; 90 if (stack_addr > &stack_variable) { 91 *stack_base = reinterpret_cast<byte*>(stack_addr) - *stack_size; 92 } else { 93 *stack_base = stack_addr; 94 } 95#else 96 pthread_attr_t attributes; 97 CHECK_PTHREAD_CALL(pthread_getattr_np, (thread, &attributes), __FUNCTION__); 98 CHECK_PTHREAD_CALL(pthread_attr_getstack, (&attributes, stack_base, stack_size), __FUNCTION__); 99 CHECK_PTHREAD_CALL(pthread_attr_destroy, (&attributes), __FUNCTION__); 100#endif 101} 102 103bool ReadFileToString(const std::string& file_name, std::string* result) { 104 UniquePtr<File> file(new File); 105 if (!file->Open(file_name, O_RDONLY)) { 106 return false; 107 } 108 109 std::vector<char> buf(8 * KB); 110 while (true) { 111 int64_t n = TEMP_FAILURE_RETRY(read(file->Fd(), &buf[0], buf.size())); 112 if (n == -1) { 113 return false; 114 } 115 if (n == 0) { 116 return true; 117 } 118 result->append(&buf[0], n); 119 } 120} 121 122std::string GetIsoDate() { 123 time_t now = time(NULL); 124 tm tmbuf; 125 tm* ptm = localtime_r(&now, &tmbuf); 126 return StringPrintf("%04d-%02d-%02d %02d:%02d:%02d", 127 ptm->tm_year + 1900, ptm->tm_mon+1, ptm->tm_mday, 128 ptm->tm_hour, ptm->tm_min, ptm->tm_sec); 129} 130 131uint64_t MilliTime() { 132#if defined(HAVE_POSIX_CLOCKS) 133 timespec now; 134 clock_gettime(CLOCK_MONOTONIC, &now); 135 return static_cast<uint64_t>(now.tv_sec) * 1000LL + now.tv_nsec / 1000000LL; 136#else 137 timeval now; 138 gettimeofday(&now, NULL); 139 return static_cast<uint64_t>(now.tv_sec) * 1000LL + now.tv_usec / 1000LL; 140#endif 141} 142 143uint64_t MicroTime() { 144#if defined(HAVE_POSIX_CLOCKS) 145 timespec now; 146 clock_gettime(CLOCK_MONOTONIC, &now); 147 return static_cast<uint64_t>(now.tv_sec) * 1000000LL + now.tv_nsec / 1000LL; 148#else 149 timeval now; 150 gettimeofday(&now, NULL); 151 return static_cast<uint64_t>(now.tv_sec) * 1000000LL + now.tv_usec; 152#endif 153} 154 155uint64_t NanoTime() { 156#if defined(HAVE_POSIX_CLOCKS) 157 timespec now; 158 clock_gettime(CLOCK_MONOTONIC, &now); 159 return static_cast<uint64_t>(now.tv_sec) * 1000000000LL + now.tv_nsec; 160#else 161 timeval now; 162 gettimeofday(&now, NULL); 163 return static_cast<uint64_t>(now.tv_sec) * 1000000000LL + now.tv_usec * 1000LL; 164#endif 165} 166 167uint64_t ThreadCpuNanoTime() { 168#if defined(HAVE_POSIX_CLOCKS) 169 timespec now; 170 clock_gettime(CLOCK_THREAD_CPUTIME_ID, &now); 171 return static_cast<uint64_t>(now.tv_sec) * 1000000000LL + now.tv_nsec; 172#else 173 UNIMPLEMENTED(WARNING); 174 return -1; 175#endif 176} 177 178void NanoSleep(uint64_t ns) { 179 timespec tm; 180 tm.tv_sec = 0; 181 tm.tv_nsec = ns; 182 nanosleep(&tm, NULL); 183} 184 185void InitTimeSpec(bool absolute, int clock, int64_t ms, int32_t ns, timespec* ts) { 186 int64_t endSec; 187 188 if (absolute) { 189#if !defined(__APPLE__) 190 clock_gettime(clock, ts); 191#else 192 UNUSED(clock); 193 timeval tv; 194 gettimeofday(&tv, NULL); 195 ts->tv_sec = tv.tv_sec; 196 ts->tv_nsec = tv.tv_usec * 1000; 197#endif 198 } else { 199 ts->tv_sec = 0; 200 ts->tv_nsec = 0; 201 } 202 endSec = ts->tv_sec + ms / 1000; 203 if (UNLIKELY(endSec >= 0x7fffffff)) { 204 std::ostringstream ss; 205 LOG(INFO) << "Note: end time exceeds epoch: " << ss.str(); 206 endSec = 0x7ffffffe; 207 } 208 ts->tv_sec = endSec; 209 ts->tv_nsec = (ts->tv_nsec + (ms % 1000) * 1000000) + ns; 210 211 // Catch rollover. 212 if (ts->tv_nsec >= 1000000000L) { 213 ts->tv_sec++; 214 ts->tv_nsec -= 1000000000L; 215 } 216} 217 218std::string PrettyDescriptor(const mirror::String* java_descriptor) { 219 if (java_descriptor == NULL) { 220 return "null"; 221 } 222 return PrettyDescriptor(java_descriptor->ToModifiedUtf8()); 223} 224 225std::string PrettyDescriptor(const mirror::Class* klass) { 226 if (klass == NULL) { 227 return "null"; 228 } 229 return PrettyDescriptor(ClassHelper(klass).GetDescriptor()); 230} 231 232std::string PrettyDescriptor(const std::string& descriptor) { 233 // Count the number of '['s to get the dimensionality. 234 const char* c = descriptor.c_str(); 235 size_t dim = 0; 236 while (*c == '[') { 237 dim++; 238 c++; 239 } 240 241 // Reference or primitive? 242 if (*c == 'L') { 243 // "[[La/b/C;" -> "a.b.C[][]". 244 c++; // Skip the 'L'. 245 } else { 246 // "[[B" -> "byte[][]". 247 // To make life easier, we make primitives look like unqualified 248 // reference types. 249 switch (*c) { 250 case 'B': c = "byte;"; break; 251 case 'C': c = "char;"; break; 252 case 'D': c = "double;"; break; 253 case 'F': c = "float;"; break; 254 case 'I': c = "int;"; break; 255 case 'J': c = "long;"; break; 256 case 'S': c = "short;"; break; 257 case 'Z': c = "boolean;"; break; 258 case 'V': c = "void;"; break; // Used when decoding return types. 259 default: return descriptor; 260 } 261 } 262 263 // At this point, 'c' is a string of the form "fully/qualified/Type;" 264 // or "primitive;". Rewrite the type with '.' instead of '/': 265 std::string result; 266 const char* p = c; 267 while (*p != ';') { 268 char ch = *p++; 269 if (ch == '/') { 270 ch = '.'; 271 } 272 result.push_back(ch); 273 } 274 // ...and replace the semicolon with 'dim' "[]" pairs: 275 while (dim--) { 276 result += "[]"; 277 } 278 return result; 279} 280 281std::string PrettyDescriptor(Primitive::Type type) { 282 std::string descriptor_string(Primitive::Descriptor(type)); 283 return PrettyDescriptor(descriptor_string); 284} 285 286std::string PrettyField(const mirror::ArtField* f, bool with_type) { 287 if (f == NULL) { 288 return "null"; 289 } 290 FieldHelper fh(f); 291 std::string result; 292 if (with_type) { 293 result += PrettyDescriptor(fh.GetTypeDescriptor()); 294 result += ' '; 295 } 296 result += PrettyDescriptor(fh.GetDeclaringClassDescriptor()); 297 result += '.'; 298 result += fh.GetName(); 299 return result; 300} 301 302std::string PrettyField(uint32_t field_idx, const DexFile& dex_file, bool with_type) { 303 if (field_idx >= dex_file.NumFieldIds()) { 304 return StringPrintf("<<invalid-field-idx-%d>>", field_idx); 305 } 306 const DexFile::FieldId& field_id = dex_file.GetFieldId(field_idx); 307 std::string result; 308 if (with_type) { 309 result += dex_file.GetFieldTypeDescriptor(field_id); 310 result += ' '; 311 } 312 result += PrettyDescriptor(dex_file.GetFieldDeclaringClassDescriptor(field_id)); 313 result += '.'; 314 result += dex_file.GetFieldName(field_id); 315 return result; 316} 317 318std::string PrettyType(uint32_t type_idx, const DexFile& dex_file) { 319 if (type_idx >= dex_file.NumTypeIds()) { 320 return StringPrintf("<<invalid-type-idx-%d>>", type_idx); 321 } 322 const DexFile::TypeId& type_id = dex_file.GetTypeId(type_idx); 323 return PrettyDescriptor(dex_file.GetTypeDescriptor(type_id)); 324} 325 326std::string PrettyArguments(const char* signature) { 327 std::string result; 328 result += '('; 329 CHECK_EQ(*signature, '('); 330 ++signature; // Skip the '('. 331 while (*signature != ')') { 332 size_t argument_length = 0; 333 while (signature[argument_length] == '[') { 334 ++argument_length; 335 } 336 if (signature[argument_length] == 'L') { 337 argument_length = (strchr(signature, ';') - signature + 1); 338 } else { 339 ++argument_length; 340 } 341 std::string argument_descriptor(signature, argument_length); 342 result += PrettyDescriptor(argument_descriptor); 343 if (signature[argument_length] != ')') { 344 result += ", "; 345 } 346 signature += argument_length; 347 } 348 CHECK_EQ(*signature, ')'); 349 ++signature; // Skip the ')'. 350 result += ')'; 351 return result; 352} 353 354std::string PrettyReturnType(const char* signature) { 355 const char* return_type = strchr(signature, ')'); 356 CHECK(return_type != NULL); 357 ++return_type; // Skip ')'. 358 return PrettyDescriptor(return_type); 359} 360 361std::string PrettyMethod(const mirror::ArtMethod* m, bool with_signature) { 362 if (m == nullptr) { 363 return "null"; 364 } 365 MethodHelper mh(m); 366 std::string result(PrettyDescriptor(mh.GetDeclaringClassDescriptor())); 367 result += '.'; 368 result += mh.GetName(); 369 if (UNLIKELY(m->IsFastNative())) { 370 result += "!"; 371 } 372 if (with_signature) { 373 const Signature signature = mh.GetSignature(); 374 std::string sig_as_string(signature.ToString()); 375 if (signature == Signature::NoSignature()) { 376 return result + sig_as_string; 377 } 378 result = PrettyReturnType(sig_as_string.c_str()) + " " + result + 379 PrettyArguments(sig_as_string.c_str()); 380 } 381 return result; 382} 383 384std::string PrettyMethod(uint32_t method_idx, const DexFile& dex_file, bool with_signature) { 385 if (method_idx >= dex_file.NumMethodIds()) { 386 return StringPrintf("<<invalid-method-idx-%d>>", method_idx); 387 } 388 const DexFile::MethodId& method_id = dex_file.GetMethodId(method_idx); 389 std::string result(PrettyDescriptor(dex_file.GetMethodDeclaringClassDescriptor(method_id))); 390 result += '.'; 391 result += dex_file.GetMethodName(method_id); 392 if (with_signature) { 393 const Signature signature = dex_file.GetMethodSignature(method_id); 394 std::string sig_as_string(signature.ToString()); 395 if (signature == Signature::NoSignature()) { 396 return result + sig_as_string; 397 } 398 result = PrettyReturnType(sig_as_string.c_str()) + " " + result + 399 PrettyArguments(sig_as_string.c_str()); 400 } 401 return result; 402} 403 404std::string PrettyTypeOf(const mirror::Object* obj) { 405 if (obj == NULL) { 406 return "null"; 407 } 408 if (obj->GetClass() == NULL) { 409 return "(raw)"; 410 } 411 ClassHelper kh(obj->GetClass()); 412 std::string result(PrettyDescriptor(kh.GetDescriptor())); 413 if (obj->IsClass()) { 414 kh.ChangeClass(obj->AsClass()); 415 result += "<" + PrettyDescriptor(kh.GetDescriptor()) + ">"; 416 } 417 return result; 418} 419 420std::string PrettyClass(const mirror::Class* c) { 421 if (c == NULL) { 422 return "null"; 423 } 424 std::string result; 425 result += "java.lang.Class<"; 426 result += PrettyDescriptor(c); 427 result += ">"; 428 return result; 429} 430 431std::string PrettyClassAndClassLoader(const mirror::Class* c) { 432 if (c == NULL) { 433 return "null"; 434 } 435 std::string result; 436 result += "java.lang.Class<"; 437 result += PrettyDescriptor(c); 438 result += ","; 439 result += PrettyTypeOf(c->GetClassLoader()); 440 // TODO: add an identifying hash value for the loader 441 result += ">"; 442 return result; 443} 444 445std::string PrettySize(int64_t byte_count) { 446 // The byte thresholds at which we display amounts. A byte count is displayed 447 // in unit U when kUnitThresholds[U] <= bytes < kUnitThresholds[U+1]. 448 static const size_t kUnitThresholds[] = { 449 0, // B up to... 450 3*1024, // KB up to... 451 2*1024*1024, // MB up to... 452 1024*1024*1024 // GB from here. 453 }; 454 static const int64_t kBytesPerUnit[] = { 1, KB, MB, GB }; 455 static const char* const kUnitStrings[] = { "B", "KB", "MB", "GB" }; 456 const char* negative_str = ""; 457 if (byte_count < 0) { 458 negative_str = "-"; 459 byte_count = -byte_count; 460 } 461 int i = arraysize(kUnitThresholds); 462 while (--i > 0) { 463 if (byte_count >= kUnitThresholds[i]) { 464 break; 465 } 466 } 467 return StringPrintf("%s%lld%s", negative_str, byte_count / kBytesPerUnit[i], kUnitStrings[i]); 468} 469 470std::string PrettyDuration(uint64_t nano_duration) { 471 if (nano_duration == 0) { 472 return "0"; 473 } else { 474 return FormatDuration(nano_duration, GetAppropriateTimeUnit(nano_duration)); 475 } 476} 477 478TimeUnit GetAppropriateTimeUnit(uint64_t nano_duration) { 479 const uint64_t one_sec = 1000 * 1000 * 1000; 480 const uint64_t one_ms = 1000 * 1000; 481 const uint64_t one_us = 1000; 482 if (nano_duration >= one_sec) { 483 return kTimeUnitSecond; 484 } else if (nano_duration >= one_ms) { 485 return kTimeUnitMillisecond; 486 } else if (nano_duration >= one_us) { 487 return kTimeUnitMicrosecond; 488 } else { 489 return kTimeUnitNanosecond; 490 } 491} 492 493uint64_t GetNsToTimeUnitDivisor(TimeUnit time_unit) { 494 const uint64_t one_sec = 1000 * 1000 * 1000; 495 const uint64_t one_ms = 1000 * 1000; 496 const uint64_t one_us = 1000; 497 498 switch (time_unit) { 499 case kTimeUnitSecond: 500 return one_sec; 501 case kTimeUnitMillisecond: 502 return one_ms; 503 case kTimeUnitMicrosecond: 504 return one_us; 505 case kTimeUnitNanosecond: 506 return 1; 507 } 508 return 0; 509} 510 511std::string FormatDuration(uint64_t nano_duration, TimeUnit time_unit) { 512 const char* unit = NULL; 513 uint64_t divisor = GetNsToTimeUnitDivisor(time_unit); 514 uint32_t zero_fill = 1; 515 switch (time_unit) { 516 case kTimeUnitSecond: 517 unit = "s"; 518 zero_fill = 9; 519 break; 520 case kTimeUnitMillisecond: 521 unit = "ms"; 522 zero_fill = 6; 523 break; 524 case kTimeUnitMicrosecond: 525 unit = "us"; 526 zero_fill = 3; 527 break; 528 case kTimeUnitNanosecond: 529 unit = "ns"; 530 zero_fill = 0; 531 break; 532 } 533 534 uint64_t whole_part = nano_duration / divisor; 535 uint64_t fractional_part = nano_duration % divisor; 536 if (fractional_part == 0) { 537 return StringPrintf("%llu%s", whole_part, unit); 538 } else { 539 while ((fractional_part % 1000) == 0) { 540 zero_fill -= 3; 541 fractional_part /= 1000; 542 } 543 if (zero_fill == 3) { 544 return StringPrintf("%llu.%03llu%s", whole_part, fractional_part, unit); 545 } else if (zero_fill == 6) { 546 return StringPrintf("%llu.%06llu%s", whole_part, fractional_part, unit); 547 } else { 548 return StringPrintf("%llu.%09llu%s", whole_part, fractional_part, unit); 549 } 550 } 551} 552 553std::string PrintableString(const std::string& utf) { 554 std::string result; 555 result += '"'; 556 const char* p = utf.c_str(); 557 size_t char_count = CountModifiedUtf8Chars(p); 558 for (size_t i = 0; i < char_count; ++i) { 559 uint16_t ch = GetUtf16FromUtf8(&p); 560 if (ch == '\\') { 561 result += "\\\\"; 562 } else if (ch == '\n') { 563 result += "\\n"; 564 } else if (ch == '\r') { 565 result += "\\r"; 566 } else if (ch == '\t') { 567 result += "\\t"; 568 } else if (NeedsEscaping(ch)) { 569 StringAppendF(&result, "\\u%04x", ch); 570 } else { 571 result += ch; 572 } 573 } 574 result += '"'; 575 return result; 576} 577 578// See http://java.sun.com/j2se/1.5.0/docs/guide/jni/spec/design.html#wp615 for the full rules. 579std::string MangleForJni(const std::string& s) { 580 std::string result; 581 size_t char_count = CountModifiedUtf8Chars(s.c_str()); 582 const char* cp = &s[0]; 583 for (size_t i = 0; i < char_count; ++i) { 584 uint16_t ch = GetUtf16FromUtf8(&cp); 585 if ((ch >= 'A' && ch <= 'Z') || (ch >= 'a' && ch <= 'z') || (ch >= '0' && ch <= '9')) { 586 result.push_back(ch); 587 } else if (ch == '.' || ch == '/') { 588 result += "_"; 589 } else if (ch == '_') { 590 result += "_1"; 591 } else if (ch == ';') { 592 result += "_2"; 593 } else if (ch == '[') { 594 result += "_3"; 595 } else { 596 StringAppendF(&result, "_0%04x", ch); 597 } 598 } 599 return result; 600} 601 602std::string DotToDescriptor(const char* class_name) { 603 std::string descriptor(class_name); 604 std::replace(descriptor.begin(), descriptor.end(), '.', '/'); 605 if (descriptor.length() > 0 && descriptor[0] != '[') { 606 descriptor = "L" + descriptor + ";"; 607 } 608 return descriptor; 609} 610 611std::string DescriptorToDot(const char* descriptor) { 612 size_t length = strlen(descriptor); 613 if (descriptor[0] == 'L' && descriptor[length - 1] == ';') { 614 std::string result(descriptor + 1, length - 2); 615 std::replace(result.begin(), result.end(), '/', '.'); 616 return result; 617 } 618 return descriptor; 619} 620 621std::string DescriptorToName(const char* descriptor) { 622 size_t length = strlen(descriptor); 623 if (descriptor[0] == 'L' && descriptor[length - 1] == ';') { 624 std::string result(descriptor + 1, length - 2); 625 return result; 626 } 627 return descriptor; 628} 629 630std::string JniShortName(const mirror::ArtMethod* m) { 631 MethodHelper mh(m); 632 std::string class_name(mh.GetDeclaringClassDescriptor()); 633 // Remove the leading 'L' and trailing ';'... 634 CHECK_EQ(class_name[0], 'L') << class_name; 635 CHECK_EQ(class_name[class_name.size() - 1], ';') << class_name; 636 class_name.erase(0, 1); 637 class_name.erase(class_name.size() - 1, 1); 638 639 std::string method_name(mh.GetName()); 640 641 std::string short_name; 642 short_name += "Java_"; 643 short_name += MangleForJni(class_name); 644 short_name += "_"; 645 short_name += MangleForJni(method_name); 646 return short_name; 647} 648 649std::string JniLongName(const mirror::ArtMethod* m) { 650 std::string long_name; 651 long_name += JniShortName(m); 652 long_name += "__"; 653 654 std::string signature(MethodHelper(m).GetSignature().ToString()); 655 signature.erase(0, 1); 656 signature.erase(signature.begin() + signature.find(')'), signature.end()); 657 658 long_name += MangleForJni(signature); 659 660 return long_name; 661} 662 663// Helper for IsValidPartOfMemberNameUtf8(), a bit vector indicating valid low ascii. 664uint32_t DEX_MEMBER_VALID_LOW_ASCII[4] = { 665 0x00000000, // 00..1f low control characters; nothing valid 666 0x03ff2010, // 20..3f digits and symbols; valid: '0'..'9', '$', '-' 667 0x87fffffe, // 40..5f uppercase etc.; valid: 'A'..'Z', '_' 668 0x07fffffe // 60..7f lowercase etc.; valid: 'a'..'z' 669}; 670 671// Helper for IsValidPartOfMemberNameUtf8(); do not call directly. 672bool IsValidPartOfMemberNameUtf8Slow(const char** pUtf8Ptr) { 673 /* 674 * It's a multibyte encoded character. Decode it and analyze. We 675 * accept anything that isn't (a) an improperly encoded low value, 676 * (b) an improper surrogate pair, (c) an encoded '\0', (d) a high 677 * control character, or (e) a high space, layout, or special 678 * character (U+00a0, U+2000..U+200f, U+2028..U+202f, 679 * U+fff0..U+ffff). This is all specified in the dex format 680 * document. 681 */ 682 683 uint16_t utf16 = GetUtf16FromUtf8(pUtf8Ptr); 684 685 // Perform follow-up tests based on the high 8 bits. 686 switch (utf16 >> 8) { 687 case 0x00: 688 // It's only valid if it's above the ISO-8859-1 high space (0xa0). 689 return (utf16 > 0x00a0); 690 case 0xd8: 691 case 0xd9: 692 case 0xda: 693 case 0xdb: 694 // It's a leading surrogate. Check to see that a trailing 695 // surrogate follows. 696 utf16 = GetUtf16FromUtf8(pUtf8Ptr); 697 return (utf16 >= 0xdc00) && (utf16 <= 0xdfff); 698 case 0xdc: 699 case 0xdd: 700 case 0xde: 701 case 0xdf: 702 // It's a trailing surrogate, which is not valid at this point. 703 return false; 704 case 0x20: 705 case 0xff: 706 // It's in the range that has spaces, controls, and specials. 707 switch (utf16 & 0xfff8) { 708 case 0x2000: 709 case 0x2008: 710 case 0x2028: 711 case 0xfff0: 712 case 0xfff8: 713 return false; 714 } 715 break; 716 } 717 return true; 718} 719 720/* Return whether the pointed-at modified-UTF-8 encoded character is 721 * valid as part of a member name, updating the pointer to point past 722 * the consumed character. This will consume two encoded UTF-16 code 723 * points if the character is encoded as a surrogate pair. Also, if 724 * this function returns false, then the given pointer may only have 725 * been partially advanced. 726 */ 727static bool IsValidPartOfMemberNameUtf8(const char** pUtf8Ptr) { 728 uint8_t c = (uint8_t) **pUtf8Ptr; 729 if (LIKELY(c <= 0x7f)) { 730 // It's low-ascii, so check the table. 731 uint32_t wordIdx = c >> 5; 732 uint32_t bitIdx = c & 0x1f; 733 (*pUtf8Ptr)++; 734 return (DEX_MEMBER_VALID_LOW_ASCII[wordIdx] & (1 << bitIdx)) != 0; 735 } 736 737 // It's a multibyte encoded character. Call a non-inline function 738 // for the heavy lifting. 739 return IsValidPartOfMemberNameUtf8Slow(pUtf8Ptr); 740} 741 742bool IsValidMemberName(const char* s) { 743 bool angle_name = false; 744 745 switch (*s) { 746 case '\0': 747 // The empty string is not a valid name. 748 return false; 749 case '<': 750 angle_name = true; 751 s++; 752 break; 753 } 754 755 while (true) { 756 switch (*s) { 757 case '\0': 758 return !angle_name; 759 case '>': 760 return angle_name && s[1] == '\0'; 761 } 762 763 if (!IsValidPartOfMemberNameUtf8(&s)) { 764 return false; 765 } 766 } 767} 768 769enum ClassNameType { kName, kDescriptor }; 770static bool IsValidClassName(const char* s, ClassNameType type, char separator) { 771 int arrayCount = 0; 772 while (*s == '[') { 773 arrayCount++; 774 s++; 775 } 776 777 if (arrayCount > 255) { 778 // Arrays may have no more than 255 dimensions. 779 return false; 780 } 781 782 if (arrayCount != 0) { 783 /* 784 * If we're looking at an array of some sort, then it doesn't 785 * matter if what is being asked for is a class name; the 786 * format looks the same as a type descriptor in that case, so 787 * treat it as such. 788 */ 789 type = kDescriptor; 790 } 791 792 if (type == kDescriptor) { 793 /* 794 * We are looking for a descriptor. Either validate it as a 795 * single-character primitive type, or continue on to check the 796 * embedded class name (bracketed by "L" and ";"). 797 */ 798 switch (*(s++)) { 799 case 'B': 800 case 'C': 801 case 'D': 802 case 'F': 803 case 'I': 804 case 'J': 805 case 'S': 806 case 'Z': 807 // These are all single-character descriptors for primitive types. 808 return (*s == '\0'); 809 case 'V': 810 // Non-array void is valid, but you can't have an array of void. 811 return (arrayCount == 0) && (*s == '\0'); 812 case 'L': 813 // Class name: Break out and continue below. 814 break; 815 default: 816 // Oddball descriptor character. 817 return false; 818 } 819 } 820 821 /* 822 * We just consumed the 'L' that introduces a class name as part 823 * of a type descriptor, or we are looking for an unadorned class 824 * name. 825 */ 826 827 bool sepOrFirst = true; // first character or just encountered a separator. 828 for (;;) { 829 uint8_t c = (uint8_t) *s; 830 switch (c) { 831 case '\0': 832 /* 833 * Premature end for a type descriptor, but valid for 834 * a class name as long as we haven't encountered an 835 * empty component (including the degenerate case of 836 * the empty string ""). 837 */ 838 return (type == kName) && !sepOrFirst; 839 case ';': 840 /* 841 * Invalid character for a class name, but the 842 * legitimate end of a type descriptor. In the latter 843 * case, make sure that this is the end of the string 844 * and that it doesn't end with an empty component 845 * (including the degenerate case of "L;"). 846 */ 847 return (type == kDescriptor) && !sepOrFirst && (s[1] == '\0'); 848 case '/': 849 case '.': 850 if (c != separator) { 851 // The wrong separator character. 852 return false; 853 } 854 if (sepOrFirst) { 855 // Separator at start or two separators in a row. 856 return false; 857 } 858 sepOrFirst = true; 859 s++; 860 break; 861 default: 862 if (!IsValidPartOfMemberNameUtf8(&s)) { 863 return false; 864 } 865 sepOrFirst = false; 866 break; 867 } 868 } 869} 870 871bool IsValidBinaryClassName(const char* s) { 872 return IsValidClassName(s, kName, '.'); 873} 874 875bool IsValidJniClassName(const char* s) { 876 return IsValidClassName(s, kName, '/'); 877} 878 879bool IsValidDescriptor(const char* s) { 880 return IsValidClassName(s, kDescriptor, '/'); 881} 882 883void Split(const std::string& s, char separator, std::vector<std::string>& result) { 884 const char* p = s.data(); 885 const char* end = p + s.size(); 886 while (p != end) { 887 if (*p == separator) { 888 ++p; 889 } else { 890 const char* start = p; 891 while (++p != end && *p != separator) { 892 // Skip to the next occurrence of the separator. 893 } 894 result.push_back(std::string(start, p - start)); 895 } 896 } 897} 898 899std::string Trim(std::string s) { 900 std::string result; 901 unsigned int start_index = 0; 902 unsigned int end_index = s.size() - 1; 903 904 // Skip initial whitespace. 905 while (start_index < s.size()) { 906 if (!isspace(s[start_index])) { 907 break; 908 } 909 start_index++; 910 } 911 912 // Skip terminating whitespace. 913 while (end_index >= start_index) { 914 if (!isspace(s[end_index])) { 915 break; 916 } 917 end_index--; 918 } 919 920 // All spaces, no beef. 921 if (end_index < start_index) { 922 return ""; 923 } 924 // Start_index is the first non-space, end_index is the last one. 925 return s.substr(start_index, end_index - start_index + 1); 926} 927 928template <typename StringT> 929std::string Join(std::vector<StringT>& strings, char separator) { 930 if (strings.empty()) { 931 return ""; 932 } 933 934 std::string result(strings[0]); 935 for (size_t i = 1; i < strings.size(); ++i) { 936 result += separator; 937 result += strings[i]; 938 } 939 return result; 940} 941 942// Explicit instantiations. 943template std::string Join<std::string>(std::vector<std::string>& strings, char separator); 944template std::string Join<const char*>(std::vector<const char*>& strings, char separator); 945template std::string Join<char*>(std::vector<char*>& strings, char separator); 946 947bool StartsWith(const std::string& s, const char* prefix) { 948 return s.compare(0, strlen(prefix), prefix) == 0; 949} 950 951bool EndsWith(const std::string& s, const char* suffix) { 952 size_t suffix_length = strlen(suffix); 953 size_t string_length = s.size(); 954 if (suffix_length > string_length) { 955 return false; 956 } 957 size_t offset = string_length - suffix_length; 958 return s.compare(offset, suffix_length, suffix) == 0; 959} 960 961void SetThreadName(const char* thread_name) { 962 int hasAt = 0; 963 int hasDot = 0; 964 const char* s = thread_name; 965 while (*s) { 966 if (*s == '.') { 967 hasDot = 1; 968 } else if (*s == '@') { 969 hasAt = 1; 970 } 971 s++; 972 } 973 int len = s - thread_name; 974 if (len < 15 || hasAt || !hasDot) { 975 s = thread_name; 976 } else { 977 s = thread_name + len - 15; 978 } 979#if defined(HAVE_ANDROID_PTHREAD_SETNAME_NP) 980 // pthread_setname_np fails rather than truncating long strings. 981 char buf[16]; // MAX_TASK_COMM_LEN=16 is hard-coded into bionic 982 strncpy(buf, s, sizeof(buf)-1); 983 buf[sizeof(buf)-1] = '\0'; 984 errno = pthread_setname_np(pthread_self(), buf); 985 if (errno != 0) { 986 PLOG(WARNING) << "Unable to set the name of current thread to '" << buf << "'"; 987 } 988#elif defined(__APPLE__) && MAC_OS_X_VERSION_MAX_ALLOWED >= 1060 989 pthread_setname_np(thread_name); 990#elif defined(HAVE_PRCTL) 991 prctl(PR_SET_NAME, (unsigned long) s, 0, 0, 0); // NOLINT (unsigned long) 992#else 993 UNIMPLEMENTED(WARNING) << thread_name; 994#endif 995} 996 997void GetTaskStats(pid_t tid, char* state, int* utime, int* stime, int* task_cpu) { 998 *utime = *stime = *task_cpu = 0; 999 std::string stats; 1000 if (!ReadFileToString(StringPrintf("/proc/self/task/%d/stat", tid), &stats)) { 1001 return; 1002 } 1003 // Skip the command, which may contain spaces. 1004 stats = stats.substr(stats.find(')') + 2); 1005 // Extract the three fields we care about. 1006 std::vector<std::string> fields; 1007 Split(stats, ' ', fields); 1008 *state = fields[0][0]; 1009 *utime = strtoull(fields[11].c_str(), NULL, 10); 1010 *stime = strtoull(fields[12].c_str(), NULL, 10); 1011 *task_cpu = strtoull(fields[36].c_str(), NULL, 10); 1012} 1013 1014std::string GetSchedulerGroupName(pid_t tid) { 1015 // /proc/<pid>/cgroup looks like this: 1016 // 2:devices:/ 1017 // 1:cpuacct,cpu:/ 1018 // We want the third field from the line whose second field contains the "cpu" token. 1019 std::string cgroup_file; 1020 if (!ReadFileToString(StringPrintf("/proc/self/task/%d/cgroup", tid), &cgroup_file)) { 1021 return ""; 1022 } 1023 std::vector<std::string> cgroup_lines; 1024 Split(cgroup_file, '\n', cgroup_lines); 1025 for (size_t i = 0; i < cgroup_lines.size(); ++i) { 1026 std::vector<std::string> cgroup_fields; 1027 Split(cgroup_lines[i], ':', cgroup_fields); 1028 std::vector<std::string> cgroups; 1029 Split(cgroup_fields[1], ',', cgroups); 1030 for (size_t i = 0; i < cgroups.size(); ++i) { 1031 if (cgroups[i] == "cpu") { 1032 return cgroup_fields[2].substr(1); // Skip the leading slash. 1033 } 1034 } 1035 } 1036 return ""; 1037} 1038 1039static std::string CleanMapName(const backtrace_map_t* map) { 1040 if (map == NULL || map->name.empty()) { 1041 return "???"; 1042 } 1043 // Turn "/usr/local/google/home/enh/clean-dalvik-dev/out/host/linux-x86/lib/libartd.so" 1044 // into "libartd.so". 1045 size_t last_slash = map->name.rfind('/'); 1046 if (last_slash == std::string::npos) { 1047 return map->name; 1048 } 1049 return map->name.substr(last_slash); 1050} 1051 1052void DumpNativeStack(std::ostream& os, pid_t tid, const char* prefix, bool include_count) { 1053 UniquePtr<Backtrace> backtrace(Backtrace::Create(BACKTRACE_CURRENT_PROCESS, tid)); 1054 if (!backtrace->Unwind(0)) { 1055 os << prefix << "(backtrace::Unwind failed for thread " << tid << ")\n"; 1056 return; 1057 } else if (backtrace->NumFrames() == 0) { 1058 os << prefix << "(no native stack frames for thread " << tid << ")\n"; 1059 return; 1060 } 1061 1062 for (Backtrace::const_iterator it = backtrace->begin(); 1063 it != backtrace->end(); ++it) { 1064 // We produce output like this: 1065 // ] #00 unwind_backtrace_thread+536 [0x55d75bb8] (libbacktrace.so) 1066 os << prefix; 1067 if (include_count) { 1068 os << StringPrintf("#%02zu ", it->num); 1069 } 1070 if (!it->func_name.empty()) { 1071 os << it->func_name; 1072 } else { 1073 os << "???"; 1074 } 1075 if (it->func_offset != 0) { 1076 os << "+" << it->func_offset; 1077 } 1078 os << StringPrintf(" [%p]", reinterpret_cast<void*>(it->pc)); 1079 os << " (" << CleanMapName(it->map) << ")\n"; 1080 } 1081} 1082 1083#if defined(__APPLE__) 1084 1085// TODO: is there any way to get the kernel stack on Mac OS? 1086void DumpKernelStack(std::ostream&, pid_t, const char*, bool) {} 1087 1088#else 1089 1090void DumpKernelStack(std::ostream& os, pid_t tid, const char* prefix, bool include_count) { 1091 if (tid == GetTid()) { 1092 // There's no point showing that we're reading our stack out of /proc! 1093 return; 1094 } 1095 1096 std::string kernel_stack_filename(StringPrintf("/proc/self/task/%d/stack", tid)); 1097 std::string kernel_stack; 1098 if (!ReadFileToString(kernel_stack_filename, &kernel_stack)) { 1099 os << prefix << "(couldn't read " << kernel_stack_filename << ")\n"; 1100 return; 1101 } 1102 1103 std::vector<std::string> kernel_stack_frames; 1104 Split(kernel_stack, '\n', kernel_stack_frames); 1105 // We skip the last stack frame because it's always equivalent to "[<ffffffff>] 0xffffffff", 1106 // which looking at the source appears to be the kernel's way of saying "that's all, folks!". 1107 kernel_stack_frames.pop_back(); 1108 for (size_t i = 0; i < kernel_stack_frames.size(); ++i) { 1109 // Turn "[<ffffffff8109156d>] futex_wait_queue_me+0xcd/0x110" into "futex_wait_queue_me+0xcd/0x110". 1110 const char* text = kernel_stack_frames[i].c_str(); 1111 const char* close_bracket = strchr(text, ']'); 1112 if (close_bracket != NULL) { 1113 text = close_bracket + 2; 1114 } 1115 os << prefix; 1116 if (include_count) { 1117 os << StringPrintf("#%02zd ", i); 1118 } 1119 os << text << "\n"; 1120 } 1121} 1122 1123#endif 1124 1125const char* GetAndroidRoot() { 1126 const char* android_root = getenv("ANDROID_ROOT"); 1127 if (android_root == NULL) { 1128 if (OS::DirectoryExists("/system")) { 1129 android_root = "/system"; 1130 } else { 1131 LOG(FATAL) << "ANDROID_ROOT not set and /system does not exist"; 1132 return ""; 1133 } 1134 } 1135 if (!OS::DirectoryExists(android_root)) { 1136 LOG(FATAL) << "Failed to find ANDROID_ROOT directory " << android_root; 1137 return ""; 1138 } 1139 return android_root; 1140} 1141 1142const char* GetAndroidData() { 1143 const char* android_data = getenv("ANDROID_DATA"); 1144 if (android_data == NULL) { 1145 if (OS::DirectoryExists("/data")) { 1146 android_data = "/data"; 1147 } else { 1148 LOG(FATAL) << "ANDROID_DATA not set and /data does not exist"; 1149 return ""; 1150 } 1151 } 1152 if (!OS::DirectoryExists(android_data)) { 1153 LOG(FATAL) << "Failed to find ANDROID_DATA directory " << android_data; 1154 return ""; 1155 } 1156 return android_data; 1157} 1158 1159std::string GetDalvikCacheOrDie(const char* android_data) { 1160 std::string dalvik_cache(StringPrintf("%s/dalvik-cache", android_data)); 1161 1162 if (!OS::DirectoryExists(dalvik_cache.c_str())) { 1163 if (StartsWith(dalvik_cache, "/tmp/")) { 1164 int result = mkdir(dalvik_cache.c_str(), 0700); 1165 if (result != 0) { 1166 LOG(FATAL) << "Failed to create dalvik-cache directory " << dalvik_cache; 1167 return ""; 1168 } 1169 } else { 1170 LOG(FATAL) << "Failed to find dalvik-cache directory " << dalvik_cache; 1171 return ""; 1172 } 1173 } 1174 return dalvik_cache; 1175} 1176 1177std::string GetDalvikCacheFilenameOrDie(const char* location) { 1178 std::string dalvik_cache(GetDalvikCacheOrDie(GetAndroidData())); 1179 if (location[0] != '/') { 1180 LOG(FATAL) << "Expected path in location to be absolute: "<< location; 1181 } 1182 std::string cache_file(&location[1]); // skip leading slash 1183 if (!EndsWith(location, ".dex") && !EndsWith(location, ".art")) { 1184 cache_file += "/"; 1185 cache_file += DexFile::kClassesDex; 1186 } 1187 std::replace(cache_file.begin(), cache_file.end(), '/', '@'); 1188 return dalvik_cache + "/" + cache_file; 1189} 1190 1191bool IsZipMagic(uint32_t magic) { 1192 return (('P' == ((magic >> 0) & 0xff)) && 1193 ('K' == ((magic >> 8) & 0xff))); 1194} 1195 1196bool IsDexMagic(uint32_t magic) { 1197 return DexFile::IsMagicValid(reinterpret_cast<const byte*>(&magic)); 1198} 1199 1200bool IsOatMagic(uint32_t magic) { 1201 return (memcmp(reinterpret_cast<const byte*>(magic), 1202 OatHeader::kOatMagic, 1203 sizeof(OatHeader::kOatMagic)) == 0); 1204} 1205 1206} // namespace art 1207