11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #ifndef SHARE_VM_CLASSFILE_COMPACTHASHTABLE_HPP
26 #define SHARE_VM_CLASSFILE_COMPACTHASHTABLE_HPP
27
28 #include "classfile/stringTable.hpp"
29 #include "classfile/symbolTable.hpp"
30 #include "memory/allocation.inline.hpp"
31 #include "oops/symbol.hpp"
32 #include "services/diagnosticCommand.hpp"
33 #include "utilities/hashtable.hpp"
34
35 class NumberSeq;
36
37 // Stats for symbol tables in the CDS archive
38 class CompactHashtableStats VALUE_OBJ_CLASS_SPEC {
39 public:
40 int hashentry_count;
41 int hashentry_bytes;
42 int bucket_count;
43 int bucket_bytes;
44 };
45
46 /////////////////////////////////////////////////////////////////////////
47 //
48 // The compact hash table writer. Used at dump time for writing out
49 // the compact table to the shared archive.
50 //
51 // At dump time, the CompactHashtableWriter obtains all entries from the
52 // symbol table and adds them to a new temporary hash table. The hash
53 // table size (number of buckets) is calculated using
54 // '(num_entries + bucket_size - 1) / bucket_size'. The default bucket
55 // size is 4 and can be changed by -XX:SharedSymbolTableBucketSize option.
56 // 4 is chosen because it produces smaller sized bucket on average for
57 // faster lookup. It also has relatively small number of empty buckets and
58 // good distribution of the entries.
59 //
60 // We use a simple hash function (symbol_hash % num_bucket) for the table.
61 // The new table is compacted when written out. Please see comments
62 // above the CompactHashtable class for the table layout detail. The bucket
63 // offsets are written to the archive as part of the compact table. The
64 // bucket offset is encoded in the low 30-bit (0-29) and the bucket type
65 // (regular or compact) are encoded in bit[31, 30]. For buckets with more
66 // than one entry, both symbol hash and symbol offset are written to the
67 // table. For buckets with only one entry, only the symbol offset is written
68 // to the table and the buckets are tagged as compact in their type bits.
69 // Buckets without entry are skipped from the table. Their offsets are
70 // still written out for faster lookup.
71 //
72 class CompactHashtableWriter: public StackObj {
73 public:
74 class Entry: public CHeapObj<mtSymbol> {
75 Entry* _next;
76 unsigned int _hash;
77 void* _literal;
78
79 public:
80 Entry(unsigned int hash, Symbol *symbol) : _next(NULL), _hash(hash), _literal(symbol) {}
81
82 void *value() {
83 return _literal;
84 }
85 Symbol *symbol() {
86 return (Symbol*)_literal;
87 }
88 unsigned int hash() {
89 return _hash;
90 }
91 Entry *next() {return _next;}
92 void set_next(Entry *p) {_next = p;}
93 }; // class CompactHashtableWriter::Entry
94
95 private:
96 static int number_of_buckets(int num_entries);
97
98 const char* _table_name;
99 int _num_entries;
100 int _num_buckets;
101 juint* _bucket_sizes;
102 Entry** _buckets;
103 int _required_bytes;
104 CompactHashtableStats* _stats;
105
106 public:
107 // This is called at dump-time only
108 CompactHashtableWriter(const char* table_name, int num_entries, CompactHashtableStats* stats);
109 ~CompactHashtableWriter();
110
111 int get_required_bytes() {
112 return _required_bytes;
113 }
114
115 void add(unsigned int hash, Symbol* symbol) {
116 add(hash, new Entry(hash, symbol));
117 }
118
119 private:
120 void add(unsigned int hash, Entry* entry);
121 juint* dump_table(juint* p, juint** first_bucket, NumberSeq* summary);
122 juint* dump_buckets(juint* table, juint* p, NumberSeq* summary);
123
124 public:
125 void dump(char** top, char* end);
126 };
127
128 #define REGULAR_BUCKET_TYPE 0
129 #define COMPACT_BUCKET_TYPE 1
130 #define TABLEEND_BUCKET_TYPE 3
131 #define BUCKET_OFFSET_MASK 0x3FFFFFFF
132 #define BUCKET_OFFSET(info) ((info) & BUCKET_OFFSET_MASK)
133 #define BUCKET_TYPE_SHIFT 30
134 #define BUCKET_TYPE(info) (((info) & ~BUCKET_OFFSET_MASK) >> BUCKET_TYPE_SHIFT)
135 #define BUCKET_INFO(offset, type) (((type) << BUCKET_TYPE_SHIFT) | ((offset) & BUCKET_OFFSET_MASK))
136
137 /////////////////////////////////////////////////////////////////////////////
138 //
139 // CompactHashtable is used to stored the CDS archive's symbol table. Used
140 // at runtime only to access the compact table from the archive.
141 //
142 // Because these tables are read-only (no entries can be added/deleted) at run-time
143 // and tend to have large number of entries, we try to minimize the footprint
144 // cost per entry.
145 //
146 // Layout of compact symbol table in the shared archive:
147 //
148 // uintx base_address;
149 // juint num_symbols;
150 // juint num_buckets;
151 // juint bucket_infos[num_buckets+1]; // bit[31,30]: type; bit[29-0]: offset
152 // juint table[]
153 //
154 // -----------------------------------
155 // | base_address | num_symbols |
156 // |---------------------------------|
157 // | num_buckets | bucket_info0 |
158 // |---------------------------------|
159 // | bucket_info1 | bucket_info2 |
160 // | bucket_info3 ... |
161 // | .... | table_end_info |
162 // |---------------------------------|
163 // | entry0 |
164 // | entry1 |
165 // | entry2 |
166 // | |
167 // | ... |
168 // -----------------------------------
169 //
170 // The size of the bucket_info table is 'num_buckets + 1'. Each entry of the
171 // bucket_info table is a 32-bit encoding of the bucket type and bucket offset,
172 // with the type in the left-most 2-bit and offset in the remaining 30-bit.
173 // The last entry is a special type. It contains the offset of the last
174 // bucket end. We use that information when traversing the compact table.
175 //
176 // There are two types of buckets, regular buckets and compact buckets. The
177 // compact buckets have '01' in their highest 2-bit, and regular buckets have
178 // '00' in their highest 2-bit.
179 //
180 // For normal buckets, each symbol's entry is 8 bytes in the table[]:
181 // juint hash; /* symbol hash */
182 // juint offset; /* Symbol* sym = (Symbol*)(base_address + offset) */
183 //
184 // For compact buckets, each entry has only the 4-byte 'offset' in the table[].
185 //
186 // See CompactHashtable::lookup() for how the table is searched at runtime.
187 // See CompactHashtableWriter::dump() for how the table is written at CDS
188 // dump time.
189 //
190 template <class T, class N> class CompactHashtable VALUE_OBJ_CLASS_SPEC {
191 friend class VMStructs;
192 uintx _base_address;
193 juint _entry_count;
194 juint _bucket_count;
195 juint _table_end_offset;
196 juint* _buckets;
197
198 inline bool equals(T entry, const char* name, int len) {
199 if (entry->equals(name, len)) {
200 assert(entry->refcount() == -1, "must be shared");
201 return true;
202 } else {
203 return false;
204 }
205 }
206
207 public:
208 CompactHashtable() {
209 _entry_count = 0;
210 _bucket_count = 0;
211 _table_end_offset = 0;
212 _buckets = 0;
213 }
214 const char* init(const char *buffer);
215
216 // Lookup an entry from the compact table
217 inline T lookup(const N* name, unsigned int hash, int len) {
218 if (_entry_count > 0) {
219 assert(!DumpSharedSpaces, "run-time only");
220 int index = hash % _bucket_count;
221 juint bucket_info = _buckets[index];
222 juint bucket_offset = BUCKET_OFFSET(bucket_info);
223 int bucket_type = BUCKET_TYPE(bucket_info);
224 juint* bucket = _buckets + bucket_offset;
225 juint* bucket_end = _buckets;
226
227 if (bucket_type == COMPACT_BUCKET_TYPE) {
228 // the compact bucket has one entry with symbol offset only
229 T entry = (T)((void*)(_base_address + bucket[0]));
230 if (equals(entry, name, len)) {
231 return entry;
232 }
233 } else {
234 // This is a regular bucket, which has more than one
235 // entries. Each entry is a pair of symbol (hash, offset).
236 // Seek until the end of the bucket.
237 bucket_end += BUCKET_OFFSET(_buckets[index + 1]);
238 while (bucket < bucket_end) {
239 unsigned int h = (unsigned int)(bucket[0]);
240 if (h == hash) {
241 juint offset = bucket[1];
242 T entry = (T)((void*)(_base_address + offset));
243 if (equals(entry, name, len)) {
244 return entry;
245 }
246 }
247 bucket += 2;
248 }
249 }
250 }
251 return NULL;
252 }
253
254 // iterate over symbols
255 void symbols_do(SymbolClosure *cl);
256 };
257
258 ////////////////////////////////////////////////////////////////////////
259 //
260 // Read/Write the contents of a hashtable textual dump (created by
261 // SymbolTable::dump).
262 // Because the dump file may be big (hundred of MB in extreme cases),
263 // we use mmap for fast access when reading it.
264 //
265 class HashtableTextDump VALUE_OBJ_CLASS_SPEC {
266 int _fd;
267 const char* _base;
268 const char* _p;
269 const char* _end;
270 const char* _filename;
271 size_t _size;
272 public:
273 HashtableTextDump(const char* filename);
274 ~HashtableTextDump();
275
276 void quit(const char* err, const char* msg);
277
278 inline int remain() {
279 return (int)(_end - _p);
280 }
281
282 void corrupted(const char *p);
283
284 inline void corrupted_if(bool cond) {
285 if (cond) {
286 corrupted(_p);
287 }
288 }
289
290 bool skip_newline();
291 int skip(char must_be_char);
292 void skip_past(char c);
293 void check_version(const char* ver);
294
295 inline int get_num(char delim) {
296 const char* p = _p;
297 const char* end = _end;
298 int num = 0;
299
300 while (p < end) {
301 char c = *p ++;
302 if ('0' <= c && c <= '9') {
303 num = num * 10 + (c - '0');
304 } else if (c == delim) {
305 _p = p;
306 return num;
307 } else {
308 corrupted(p-1);
309 }
310 }
311 corrupted(_end);
312 ShouldNotReachHere();
313 return 0;
314 }
315
316 int scan_prefix();
317 int scan_prefix2();
318
319 jchar unescape(const char* from, const char* end, int count);
320 void get_utf8(char* utf8_buffer, int utf8_length);
321 static void put_utf8(outputStream* st, const char* utf8_string, int utf8_length);
322 };
323
324 ///////////////////////////////////////////////////////////////////////
325 //
326 // jcmd command support for symbol table and string table dumping:
327 // VM.symboltable -verbose: for dumping the symbol table
328 // VM.stringtable -verbose: for dumping the string table
329 //
330 class VM_DumpHashtable : public VM_Operation {
331 private:
332 outputStream* _out;
333 int _which;
334 bool _verbose;
335 public:
336 enum {
337 DumpSymbols = 1 << 0,
|
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #ifndef SHARE_VM_CLASSFILE_COMPACTHASHTABLE_HPP
26 #define SHARE_VM_CLASSFILE_COMPACTHASHTABLE_HPP
27
28 #include "classfile/stringTable.hpp"
29 #include "classfile/symbolTable.hpp"
30 #include "memory/allocation.inline.hpp"
31 #include "oops/oop.inline.hpp"
32 #include "oops/symbol.hpp"
33 #include "services/diagnosticCommand.hpp"
34 #include "utilities/hashtable.hpp"
35
36 class NumberSeq;
37
38 // Stats for symbol tables in the CDS archive
39 class CompactHashtableStats VALUE_OBJ_CLASS_SPEC {
40 public:
41 int hashentry_count;
42 int hashentry_bytes;
43 int bucket_count;
44 int bucket_bytes;
45 };
46
47 /////////////////////////////////////////////////////////////////////////
48 //
49 // The compact hash table writer. Used at dump time for writing out
50 // the compact table to the shared archive.
51 //
52 // At dump time, the CompactHashtableWriter obtains all entries from the
53 // symbol/string table and adds them to a new temporary hash table. The hash
54 // table size (number of buckets) is calculated using
55 // '(num_entries + bucket_size - 1) / bucket_size'. The default bucket
56 // size is 4 and can be changed by -XX:SharedSymbolTableBucketSize option.
57 // 4 is chosen because it produces smaller sized bucket on average for
58 // faster lookup. It also has relatively small number of empty buckets and
59 // good distribution of the entries.
60 //
61 // We use a simple hash function (hash % num_bucket) for the table.
62 // The new table is compacted when written out. Please see comments
63 // above the CompactHashtable class for the table layout detail. The bucket
64 // offsets are written to the archive as part of the compact table. The
65 // bucket offset is encoded in the low 30-bit (0-29) and the bucket type
66 // (regular or compact) are encoded in bit[31, 30]. For buckets with more
67 // than one entry, both hash and entry offset are written to the
68 // table. For buckets with only one entry, only the entry offset is written
69 // to the table and the buckets are tagged as compact in their type bits.
70 // Buckets without entry are skipped from the table. Their offsets are
71 // still written out for faster lookup.
72 //
73 class CompactHashtableWriter: public StackObj {
74 public:
75 class Entry: public CHeapObj<mtSymbol> {
76 Entry* _next;
77 unsigned int _hash;
78 void* _literal;
79
80 public:
81 Entry(unsigned int hash, Symbol *symbol) : _next(NULL), _hash(hash), _literal(symbol) {}
82 Entry(unsigned int hash, oop string) : _next(NULL), _hash(hash), _literal(string) {}
83
84 void *value() {
85 return _literal;
86 }
87 Symbol *symbol() {
88 return (Symbol*)_literal;
89 }
90 oop string() {
91 return (oop)_literal;
92 }
93 unsigned int hash() {
94 return _hash;
95 }
96 Entry *next() {return _next;}
97 void set_next(Entry *p) {_next = p;}
98 }; // class CompactHashtableWriter::Entry
99
100 private:
101 static int number_of_buckets(int num_entries);
102
103 int _type;
104 int _num_entries;
105 int _num_buckets;
106 juint* _bucket_sizes;
107 Entry** _buckets;
108 int _required_bytes;
109 CompactHashtableStats* _stats;
110
111 public:
112 // This is called at dump-time only
113 CompactHashtableWriter(int table_type, int num_entries, CompactHashtableStats* stats);
114 ~CompactHashtableWriter();
115
116 int get_required_bytes() {
117 return _required_bytes;
118 }
119
120 void add(unsigned int hash, Symbol* symbol) {
121 add(hash, new Entry(hash, symbol));
122 }
123
124 void add(unsigned int hash, oop string) {
125 add(hash, new Entry(hash, string));
126 }
127
128 private:
129 void add(unsigned int hash, Entry* entry);
130 juint* dump_table(juint* p, juint** first_bucket, NumberSeq* summary);
131 juint* dump_buckets(juint* table, juint* p, NumberSeq* summary);
132
133 public:
134 void dump(char** top, char* end);
135 const char* table_name();
136 };
137
138 #define REGULAR_BUCKET_TYPE 0
139 #define COMPACT_BUCKET_TYPE 1
140 #define TABLEEND_BUCKET_TYPE 3
141 #define BUCKET_OFFSET_MASK 0x3FFFFFFF
142 #define BUCKET_OFFSET(info) ((info) & BUCKET_OFFSET_MASK)
143 #define BUCKET_TYPE_SHIFT 30
144 #define BUCKET_TYPE(info) (((info) & ~BUCKET_OFFSET_MASK) >> BUCKET_TYPE_SHIFT)
145 #define BUCKET_INFO(offset, type) (((type) << BUCKET_TYPE_SHIFT) | ((offset) & BUCKET_OFFSET_MASK))
146
147 /////////////////////////////////////////////////////////////////////////////
148 //
149 // CompactHashtable is used to stored the CDS archive's symbol/string table. Used
150 // at runtime only to access the compact table from the archive.
151 //
152 // Because these tables are read-only (no entries can be added/deleted) at run-time
153 // and tend to have large number of entries, we try to minimize the footprint
154 // cost per entry.
155 //
156 // Layout of compact table in the shared archive:
157 //
158 // uintx base_address;
159 // juint num_entries;
160 // juint num_buckets;
161 // juint bucket_infos[num_buckets+1]; // bit[31,30]: type; bit[29-0]: offset
162 // juint table[]
163 //
164 // -----------------------------------
165 // | base_address | num_entries |
166 // |---------------------------------|
167 // | num_buckets | bucket_info0 |
168 // |---------------------------------|
169 // | bucket_info1 | bucket_info2 |
170 // | bucket_info3 ... |
171 // | .... | table_end_info |
172 // |---------------------------------|
173 // | entry0 |
174 // | entry1 |
175 // | entry2 |
176 // | |
177 // | ... |
178 // -----------------------------------
179 //
180 // The size of the bucket_info table is 'num_buckets + 1'. Each entry of the
181 // bucket_info table is a 32-bit encoding of the bucket type and bucket offset,
182 // with the type in the left-most 2-bit and offset in the remaining 30-bit.
183 // The last entry is a special type. It contains the offset of the last
184 // bucket end. We use that information when traversing the compact table.
185 //
186 // There are two types of buckets, regular buckets and compact buckets. The
187 // compact buckets have '01' in their highest 2-bit, and regular buckets have
188 // '00' in their highest 2-bit.
189 //
190 // For normal buckets, each entry is 8 bytes in the table[]:
191 // juint hash; /* symbol/string hash */
192 // union {
193 // juint offset; /* Symbol* sym = (Symbol*)(base_address + offset) */
194 // narrowOop str; /* String narrowOop encoding */
195 // }
196 //
197 //
198 // For compact buckets, each entry has only the 4-byte 'offset' in the table[].
199 //
200 // See CompactHashtable::lookup() for how the table is searched at runtime.
201 // See CompactHashtableWriter::dump() for how the table is written at CDS
202 // dump time.
203 //
204 template <class T, class N> class CompactHashtable VALUE_OBJ_CLASS_SPEC {
205 friend class VMStructs;
206
207 public:
208 enum CompactHashtableType {
209 _symbol_table = 0,
210 _string_table = 1
211 };
212
213 private:
214 CompactHashtableType _type;
215 uintx _base_address;
216 juint _entry_count;
217 juint _bucket_count;
218 juint _table_end_offset;
219 juint* _buckets;
220
221 inline Symbol* lookup_entry(CompactHashtable<Symbol*, char>* const t,
222 juint* addr, const char* name, int len) {
223 Symbol* sym = (Symbol*)((void*)(_base_address + *addr));
224 if (sym->equals(name, len)) {
225 assert(sym->refcount() == -1, "must be shared");
226 return sym;
227 }
228
229 return NULL;
230 }
231
232 inline oop lookup_entry(CompactHashtable<oop, char>* const t,
233 juint* addr, const char* name, int len) {
234 narrowOop obj = (narrowOop)(*addr);
235 oop string = oopDesc::decode_heap_oop(obj);
236 if (java_lang_String::equals(string, (jchar*)name, len)) {
237 return string;
238 }
239
240 return NULL;
241 }
242
243 public:
244 CompactHashtable() {
245 _entry_count = 0;
246 _bucket_count = 0;
247 _table_end_offset = 0;
248 _buckets = 0;
249 }
250 const char* init(CompactHashtableType type, const char *buffer);
251
252 void reset() {
253 _entry_count = 0;
254 _bucket_count = 0;
255 _table_end_offset = 0;
256 _buckets = 0;
257 }
258
259 // Lookup an entry from the compact table
260 inline T lookup(const N* name, unsigned int hash, int len) {
261 if (_entry_count > 0) {
262 assert(!DumpSharedSpaces, "run-time only");
263 int index = hash % _bucket_count;
264 juint bucket_info = _buckets[index];
265 juint bucket_offset = BUCKET_OFFSET(bucket_info);
266 int bucket_type = BUCKET_TYPE(bucket_info);
267 juint* bucket = _buckets + bucket_offset;
268 juint* bucket_end = _buckets;
269
270 if (bucket_type == COMPACT_BUCKET_TYPE) {
271 // the compact bucket has one entry with entry offset only
272 T res = lookup_entry(this, &bucket[0], name, len);
273 if (res != NULL) {
274 return res;
275 }
276 } else {
277 // This is a regular bucket, which has more than one
278 // entries. Each entry is a pair of entry (hash, offset).
279 // Seek until the end of the bucket.
280 bucket_end += BUCKET_OFFSET(_buckets[index + 1]);
281 while (bucket < bucket_end) {
282 unsigned int h = (unsigned int)(bucket[0]);
283 if (h == hash) {
284 T res = lookup_entry(this, &bucket[1], name, len);
285 if (res != NULL) {
286 return res;
287 }
288 }
289 bucket += 2;
290 }
291 }
292 }
293 return NULL;
294 }
295
296 // iterate over symbols
297 void symbols_do(SymbolClosure *cl);
298
299 // iterate over strings
300 void oops_do(OopClosure* f);
301 };
302
303 ////////////////////////////////////////////////////////////////////////
304 //
305 // Read/Write the contents of a hashtable textual dump (created by
306 // SymbolTable::dump and StringTable::dump).
307 // Because the dump file may be big (hundred of MB in extreme cases),
308 // we use mmap for fast access when reading it.
309 //
310 class HashtableTextDump VALUE_OBJ_CLASS_SPEC {
311 int _fd;
312 const char* _base;
313 const char* _p;
314 const char* _end;
315 const char* _filename;
316 size_t _size;
317 int _prefix_type;
318 int _line_no;
319 public:
320 HashtableTextDump(const char* filename);
321 ~HashtableTextDump();
322
323 enum {
324 SymbolPrefix = 1 << 0,
325 StringPrefix = 1 << 1,
326 Unknown = 1 << 2
327 };
328
329 void quit(const char* err, const char* msg);
330
331 inline int remain() {
332 return (int)(_end - _p);
333 }
334
335 void corrupted(const char *p, const char *msg);
336
337 inline void corrupted_if(bool cond) {
338 if (cond) {
339 corrupted(_p, NULL);
340 }
341 }
342
343 bool skip_newline();
344 int skip(char must_be_char);
345 void skip_past(char c);
346 void check_version(const char* ver);
347
348 inline bool get_num(char delim, int *utf8_length) {
349 const char* p = _p;
350 const char* end = _end;
351 int num = 0;
352
353 while (p < end) {
354 char c = *p ++;
355 if ('0' <= c && c <= '9') {
356 num = num * 10 + (c - '0');
357 } else if (c == delim) {
358 _p = p;
359 *utf8_length = num;
360 return true;
361 } else {
362 // Not [0-9], not 'delim'
363 return false;
364 }
365 }
366 corrupted(_end, "Incorrect format");
367 ShouldNotReachHere();
368 return false;
369 }
370
371 void scan_prefix_type();
372 int scan_prefix(int* utf8_length);
373 int scan_string_prefix();
374 int scan_symbol_prefix();
375
376 jchar unescape(const char* from, const char* end, int count);
377 void get_utf8(char* utf8_buffer, int utf8_length);
378 static void put_utf8(outputStream* st, const char* utf8_string, int utf8_length);
379 };
380
381 ///////////////////////////////////////////////////////////////////////
382 //
383 // jcmd command support for symbol table and string table dumping:
384 // VM.symboltable -verbose: for dumping the symbol table
385 // VM.stringtable -verbose: for dumping the string table
386 //
387 class VM_DumpHashtable : public VM_Operation {
388 private:
389 outputStream* _out;
390 int _which;
391 bool _verbose;
392 public:
393 enum {
394 DumpSymbols = 1 << 0,
|