duke@435: /* tschatzl@8766: * Copyright (c) 1997, 2017, Oracle and/or its affiliates. All rights reserved. duke@435: * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. duke@435: * duke@435: * This code is free software; you can redistribute it and/or modify it duke@435: * under the terms of the GNU General Public License version 2 only, as duke@435: * published by the Free Software Foundation. duke@435: * duke@435: * This code is distributed in the hope that it will be useful, but WITHOUT duke@435: * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or duke@435: * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License duke@435: * version 2 for more details (a copy is included in the LICENSE file that duke@435: * accompanied this code). duke@435: * duke@435: * You should have received a copy of the GNU General Public License version duke@435: * 2 along with this work; if not, write to the Free Software Foundation, duke@435: * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. duke@435: * trims@1907: * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA trims@1907: * or visit www.oracle.com if you need additional information or have any trims@1907: * questions. duke@435: * duke@435: */ duke@435: stefank@2314: #ifndef SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP stefank@2314: #define SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP stefank@2314: stefank@2314: #include "memory/allocation.inline.hpp" coleenp@2497: #include "oops/symbol.hpp" stefank@2314: #include "utilities/hashtable.hpp" stefank@2314: coleenp@2497: // The symbol table holds all Symbol*s and corresponding interned strings. coleenp@2497: // Symbol*s and literal strings should be canonicalized. duke@435: // duke@435: // The interned strings are created lazily. duke@435: // duke@435: // It is implemented as an open hash table with a fixed number of buckets. duke@435: // duke@435: // %note: duke@435: // - symbolTableEntrys are allocated in blocks to reduce the space overhead. duke@435: duke@435: class BoolObjectClosure; coleenp@3865: class outputStream; duke@435: duke@435: coleenp@2497: // Class to hold a newly created or referenced Symbol* temporarily in scope. coleenp@2497: // new_symbol() and lookup() will create a Symbol* if not already in the coleenp@2497: // symbol table and add to the symbol's reference count. coleenp@2497: // probe() and lookup_only() will increment the refcount if symbol is found. coleenp@2497: class TempNewSymbol : public StackObj { coleenp@2497: Symbol* _temp; coleenp@2497: coleenp@2497: public: coleenp@2497: TempNewSymbol() : _temp(NULL) {} coleenp@2497: // Creating or looking up a symbol increments the symbol's reference count coleenp@2497: TempNewSymbol(Symbol *s) : _temp(s) {} coleenp@2497: coleenp@2497: // Operator= increments reference count. coleenp@2497: void operator=(const TempNewSymbol &s) { twisti@3969: //clear(); //FIXME coleenp@2497: _temp = s._temp; coleenp@2497: if (_temp !=NULL) _temp->increment_refcount(); coleenp@2497: } coleenp@2497: coleenp@2497: // Decrement reference counter so it can go away if it's unique twisti@3969: void clear() { if (_temp != NULL) _temp->decrement_refcount(); _temp = NULL; } twisti@3969: twisti@3969: ~TempNewSymbol() { clear(); } coleenp@2497: coleenp@2497: // Operators so they can be used like Symbols coleenp@2497: Symbol* operator -> () const { return _temp; } coleenp@2497: bool operator == (Symbol* o) const { return _temp == o; } coleenp@2497: // Sneaky conversion function coleenp@2497: operator Symbol*() { return _temp; } coleenp@2497: }; coleenp@2497: mgerdin@7207: class SymbolTable : public RehashableHashtable { duke@435: friend class VMStructs; coleenp@2497: friend class ClassFileParser; duke@435: duke@435: private: duke@435: // The symbol table duke@435: static SymbolTable* _the_table; duke@435: coleenp@3865: // Set if one bucket is out of balance due to hash algorithm deficiency coleenp@3865: static bool _needs_rehashing; coleenp@3865: coleenp@2497: // For statistics tschatzl@6229: static int _symbols_removed; tschatzl@6229: static int _symbols_counted; coleenp@2497: coleenp@3682: Symbol* allocate_symbol(const u1* name, int len, bool c_heap, TRAPS); // Assumes no characters larger than 0x7F coleenp@2497: duke@435: // Adding elements coleenp@3682: Symbol* basic_add(int index, u1* name, int len, unsigned int hashValue, coleenp@3682: bool c_heap, TRAPS); coleenp@4037: bool basic_add(ClassLoaderData* loader_data, coleenp@4037: constantPoolHandle cp, int names_count, duke@435: const char** names, int* lengths, int* cp_indices, duke@435: unsigned int* hashValues, TRAPS); duke@435: coleenp@4037: static void new_symbols(ClassLoaderData* loader_data, coleenp@4037: constantPoolHandle cp, int names_count, coleenp@2497: const char** name, int* lengths, coleenp@2497: int* cp_indices, unsigned int* hashValues, coleenp@2497: TRAPS) { coleenp@4037: add(loader_data, cp, names_count, name, lengths, cp_indices, hashValues, THREAD); coleenp@2497: } coleenp@2497: coleenp@2497: Symbol* lookup(int index, const char* name, int len, unsigned int hash); duke@435: duke@435: SymbolTable() mgerdin@7207: : RehashableHashtable(SymbolTableSize, sizeof (HashtableEntry)) {} duke@435: zgu@3900: SymbolTable(HashtableBucket* t, int number_of_entries) mgerdin@7207: : RehashableHashtable(SymbolTableSize, sizeof (HashtableEntry), t, duke@435: number_of_entries) {} duke@435: coleenp@3682: // Arena for permanent symbols (null class loader) that are never unloaded coleenp@3682: static Arena* _arena; coleenp@3682: static Arena* arena() { return _arena; } // called for statistics duke@435: coleenp@3682: static void initialize_symbols(int arena_alloc_size = 0); tschatzl@6229: tschatzl@6229: static volatile int _parallel_claimed_idx; tschatzl@6229: tschatzl@8766: typedef SymbolTable::BucketUnlinkContext BucketUnlinkContext; tschatzl@8766: // Release any dead symbols. Unlinked bucket entries are collected in the given tschatzl@8766: // context to be freed later. tschatzl@8766: // This allows multiple threads to work on the table at once. tschatzl@8766: static void buckets_unlink(int start_idx, int end_idx, BucketUnlinkContext* context, size_t* memory_total); duke@435: public: duke@435: enum { coleenp@3682: symbol_alloc_batch_size = 8, coleenp@3682: // Pick initial size based on java -version size measurements coleenp@3682: symbol_alloc_arena_size = 360*K duke@435: }; duke@435: duke@435: // The symbol table duke@435: static SymbolTable* the_table() { return _the_table; } duke@435: kevinw@5850: // Size of one bucket in the string table. Used when checking for rollover. kevinw@5850: static uint bucket_size() { return sizeof(HashtableBucket); } kevinw@5850: duke@435: static void create_table() { duke@435: assert(_the_table == NULL, "One symbol table allowed."); duke@435: _the_table = new SymbolTable(); coleenp@3682: initialize_symbols(symbol_alloc_arena_size); duke@435: } duke@435: zgu@3900: static void create_table(HashtableBucket* t, int length, duke@435: int number_of_entries) { duke@435: assert(_the_table == NULL, "One symbol table allowed."); kevinw@5850: kevinw@5850: // If CDS archive used a different symbol table size, use that size instead kevinw@5850: // which is better than giving an error. kevinw@5850: SymbolTableSize = length/bucket_size(); kevinw@5850: duke@435: _the_table = new SymbolTable(t, number_of_entries); coleenp@3682: // if CDS give symbol table a default arena size since most symbols coleenp@3682: // are already allocated in the shared misc section. coleenp@3682: initialize_symbols(); duke@435: } duke@435: coleenp@3875: static unsigned int hash_symbol(const char* s, int len); coleenp@3865: coleenp@2497: static Symbol* lookup(const char* name, int len, TRAPS); duke@435: // lookup only, won't add. Also calculate hash. coleenp@2497: static Symbol* lookup_only(const char* name, int len, unsigned int& hash); duke@435: // Only copy to C string to be added if lookup failed. coleenp@2497: static Symbol* lookup(const Symbol* sym, int begin, int end, TRAPS); coleenp@2497: coleenp@2497: static void release(Symbol* sym); duke@435: phh@3427: // Look up the address of the literal in the SymbolTable for this Symbol* phh@3427: static Symbol** lookup_symbol_addr(Symbol* sym); phh@3427: jrose@1100: // jchar (utf16) version of lookups coleenp@2497: static Symbol* lookup_unicode(const jchar* name, int len, TRAPS); coleenp@2497: static Symbol* lookup_only_unicode(const jchar* name, int len, unsigned int& hash); jrose@1100: coleenp@4037: static void add(ClassLoaderData* loader_data, coleenp@4037: constantPoolHandle cp, int names_count, duke@435: const char** names, int* lengths, int* cp_indices, duke@435: unsigned int* hashValues, TRAPS); duke@435: coleenp@2497: // Release any dead symbols tschatzl@6229: static void unlink() { tschatzl@6229: int processed = 0; tschatzl@6229: int removed = 0; tschatzl@6229: unlink(&processed, &removed); tschatzl@6229: } tschatzl@6229: static void unlink(int* processed, int* removed); tschatzl@6229: // Release any dead symbols, possibly parallel version tschatzl@6229: static void possibly_parallel_unlink(int* processed, int* removed); coleenp@2497: coleenp@2497: // iterate over symbols coleenp@2497: static void symbols_do(SymbolClosure *cl); coleenp@2497: coleenp@2497: // Symbol creation coleenp@2497: static Symbol* new_symbol(const char* utf8_buffer, int length, TRAPS) { coleenp@2497: assert(utf8_buffer != NULL, "just checking"); coleenp@2497: return lookup(utf8_buffer, length, THREAD); duke@435: } coleenp@2497: static Symbol* new_symbol(const char* name, TRAPS) { coleenp@2497: return new_symbol(name, (int)strlen(name), THREAD); coleenp@2497: } coleenp@2497: static Symbol* new_symbol(const Symbol* sym, int begin, int end, TRAPS) { coleenp@2497: assert(begin <= end && end <= sym->utf8_length(), "just checking"); coleenp@2497: return lookup(sym, begin, end, THREAD); duke@435: } duke@435: coleenp@3682: // Create a symbol in the arena for symbols that are not deleted coleenp@3682: static Symbol* new_permanent_symbol(const char* name, TRAPS); coleenp@3682: duke@435: // Symbol lookup coleenp@2497: static Symbol* lookup(int index, const char* name, int len, TRAPS); duke@435: duke@435: // Needed for preloading classes in signatures when compiling. duke@435: // Returns the symbol is already present in symbol table, otherwise duke@435: // NULL. NO ALLOCATION IS GUARANTEED! coleenp@2497: static Symbol* probe(const char* name, int len) { jrose@1100: unsigned int ignore_hash; jrose@1100: return lookup_only(name, len, ignore_hash); jrose@1100: } coleenp@2497: static Symbol* probe_unicode(const jchar* name, int len) { jrose@1100: unsigned int ignore_hash; jrose@1100: return lookup_only_unicode(name, len, ignore_hash); jrose@1100: } duke@435: duke@435: // Histogram duke@435: static void print_histogram() PRODUCT_RETURN; coleenp@2497: static void print() PRODUCT_RETURN; duke@435: duke@435: // Debugging duke@435: static void verify(); coleenp@3865: static void dump(outputStream* st); duke@435: duke@435: // Sharing duke@435: static void copy_buckets(char** top, char*end) { zgu@3900: the_table()->Hashtable::copy_buckets(top, end); duke@435: } duke@435: static void copy_table(char** top, char*end) { zgu@3900: the_table()->Hashtable::copy_table(top, end); duke@435: } duke@435: static void reverse(void* boundary = NULL) { zgu@3900: the_table()->Hashtable::reverse(boundary); duke@435: } coleenp@3865: coleenp@3865: // Rehash the symbol table if it gets out of balance coleenp@3865: static void rehash_table(); coleenp@3865: static bool needs_rehashing() { return _needs_rehashing; } tschatzl@6229: // Parallel chunked scanning tschatzl@6229: static void clear_parallel_claimed_index() { _parallel_claimed_idx = 0; } tschatzl@6229: static int parallel_claimed_index() { return _parallel_claimed_idx; } duke@435: }; duke@435: mgerdin@7207: class StringTable : public RehashableHashtable { duke@435: friend class VMStructs; duke@435: duke@435: private: duke@435: // The string table duke@435: static StringTable* _the_table; duke@435: coleenp@3865: // Set if one bucket is out of balance due to hash algorithm deficiency coleenp@3865: static bool _needs_rehashing; coleenp@3865: johnc@5277: // Claimed high water mark for parallel chunked scanning johnc@5277: static volatile int _parallel_claimed_idx; johnc@5277: duke@435: static oop intern(Handle string_or_null, jchar* chars, int length, TRAPS); duke@435: oop basic_add(int index, Handle string_or_null, jchar* name, int len, duke@435: unsigned int hashValue, TRAPS); duke@435: duke@435: oop lookup(int index, jchar* chars, int length, unsigned int hashValue); duke@435: johnc@5277: // Apply the give oop closure to the entries to the buckets johnc@5277: // in the range [start_idx, end_idx). tschatzl@6229: static void buckets_oops_do(OopClosure* f, int start_idx, int end_idx); tschatzl@8766: tschatzl@8766: typedef StringTable::BucketUnlinkContext BucketUnlinkContext; tschatzl@6229: // Unlink or apply the give oop closure to the entries to the buckets tschatzl@8766: // in the range [start_idx, end_idx). Unlinked bucket entries are collected in the given tschatzl@8766: // context to be freed later. tschatzl@8766: // This allows multiple threads to work on the table at once. tschatzl@8766: static void buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, BucketUnlinkContext* context); johnc@5277: mgerdin@7207: StringTable() : RehashableHashtable((int)StringTableSize, zgu@3900: sizeof (HashtableEntry)) {} duke@435: zgu@3900: StringTable(HashtableBucket* t, int number_of_entries) mgerdin@7207: : RehashableHashtable((int)StringTableSize, sizeof (HashtableEntry), t, jcoomes@2660: number_of_entries) {} duke@435: public: duke@435: // The string table duke@435: static StringTable* the_table() { return _the_table; } duke@435: hseigel@4277: // Size of one bucket in the string table. Used when checking for rollover. hseigel@4277: static uint bucket_size() { return sizeof(HashtableBucket); } hseigel@4277: duke@435: static void create_table() { duke@435: assert(_the_table == NULL, "One string table allowed."); duke@435: _the_table = new StringTable(); duke@435: } duke@435: duke@435: // GC support duke@435: // Delete pointers to otherwise-unreachable objects. tschatzl@6229: static void unlink_or_oops_do(BoolObjectClosure* cl, OopClosure* f) { tschatzl@6229: int processed = 0; tschatzl@6229: int removed = 0; tschatzl@6229: unlink_or_oops_do(cl, f, &processed, &removed); tschatzl@6229: } stefank@5196: static void unlink(BoolObjectClosure* cl) { tschatzl@6229: int processed = 0; tschatzl@6229: int removed = 0; tschatzl@6229: unlink_or_oops_do(cl, NULL, &processed, &removed); stefank@5196: } tschatzl@6229: static void unlink_or_oops_do(BoolObjectClosure* cl, OopClosure* f, int* processed, int* removed); tschatzl@6229: static void unlink(BoolObjectClosure* cl, int* processed, int* removed) { tschatzl@6229: unlink_or_oops_do(cl, NULL, processed, removed); tschatzl@6229: } johnc@5277: // Serially invoke "f->do_oop" on the locations of all oops in the table. coleenp@2497: static void oops_do(OopClosure* f); duke@435: tschatzl@6229: // Possibly parallel versions of the above tschatzl@6229: static void possibly_parallel_unlink_or_oops_do(BoolObjectClosure* cl, OopClosure* f, int* processed, int* removed); tschatzl@6229: static void possibly_parallel_unlink(BoolObjectClosure* cl, int* processed, int* removed) { tschatzl@6229: possibly_parallel_unlink_or_oops_do(cl, NULL, processed, removed); tschatzl@6229: } johnc@5277: static void possibly_parallel_oops_do(OopClosure* f); johnc@5277: coleenp@3865: // Hashing algorithm, used as the hash value used by the coleenp@3865: // StringTable for bucket selection and comparison (stored in the coleenp@3865: // HashtableEntry structures). This is used in the String.intern() method. coleenp@3875: static unsigned int hash_string(const jchar* s, int len); coleenp@3865: coleenp@3865: // Internal test. coleenp@3865: static void test_alt_hash() PRODUCT_RETURN; coleenp@3865: duke@435: // Probing coleenp@2497: static oop lookup(Symbol* symbol); mgerdin@4850: static oop lookup(jchar* chars, int length); duke@435: duke@435: // Interning coleenp@2497: static oop intern(Symbol* symbol, TRAPS); duke@435: static oop intern(oop string, TRAPS); duke@435: static oop intern(const char *utf8_string, TRAPS); duke@435: duke@435: // Debugging duke@435: static void verify(); coleenp@3865: static void dump(outputStream* st); duke@435: dcubed@5743: enum VerifyMesgModes { dcubed@5743: _verify_quietly = 0, dcubed@5743: _verify_with_mesgs = 1 dcubed@5743: }; dcubed@5743: dcubed@5743: enum VerifyRetTypes { dcubed@5743: _verify_pass = 0, dcubed@5743: _verify_fail_continue = 1, dcubed@5743: _verify_fail_done = 2 dcubed@5743: }; dcubed@5743: dcubed@5743: static VerifyRetTypes compare_entries(int bkt1, int e_cnt1, dcubed@5743: HashtableEntry* e_ptr1, dcubed@5743: int bkt2, int e_cnt2, dcubed@5743: HashtableEntry* e_ptr2); dcubed@5743: static VerifyRetTypes verify_entry(int bkt, int e_cnt, dcubed@5743: HashtableEntry* e_ptr, dcubed@5743: VerifyMesgModes mesg_mode); dcubed@5743: static int verify_and_compare_entries(); dcubed@5743: duke@435: // Sharing duke@435: static void copy_buckets(char** top, char*end) { zgu@3900: the_table()->Hashtable::copy_buckets(top, end); duke@435: } duke@435: static void copy_table(char** top, char*end) { zgu@3900: the_table()->Hashtable::copy_table(top, end); duke@435: } duke@435: static void reverse() { zgu@3900: the_table()->Hashtable::reverse(); duke@435: } coleenp@3865: coleenp@3865: // Rehash the symbol table if it gets out of balance coleenp@3865: static void rehash_table(); coleenp@3865: static bool needs_rehashing() { return _needs_rehashing; } johnc@5277: johnc@5277: // Parallel chunked scanning johnc@5277: static void clear_parallel_claimed_index() { _parallel_claimed_idx = 0; } tschatzl@6229: static int parallel_claimed_index() { return _parallel_claimed_idx; } duke@435: }; stefank@2314: #endif // SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP