Commit | Line | Data |
---|---|---|
5e28c532 | 1 | /* |
abc490a1 MD |
2 | * rculfhash.c |
3 | * | |
1475579c | 4 | * Userspace RCU library - Lock-Free Resizable RCU Hash Table |
abc490a1 MD |
5 | * |
6 | * Copyright 2010-2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> | |
7 | * | |
8 | * This library is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU Lesser General Public | |
10 | * License as published by the Free Software Foundation; either | |
11 | * version 2.1 of the License, or (at your option) any later version. | |
12 | * | |
13 | * This library is distributed in the hope that it will be useful, | |
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 | * Lesser General Public License for more details. | |
17 | * | |
18 | * You should have received a copy of the GNU Lesser General Public | |
19 | * License along with this library; if not, write to the Free Software | |
20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
5e28c532 MD |
21 | */ |
22 | ||
e753ff5a MD |
23 | /* |
24 | * Based on the following articles: | |
25 | * - Ori Shalev and Nir Shavit. Split-ordered lists: Lock-free | |
26 | * extensible hash tables. J. ACM 53, 3 (May 2006), 379-405. | |
27 | * - Michael, M. M. High performance dynamic lock-free hash tables | |
28 | * and list-based sets. In Proceedings of the fourteenth annual ACM | |
29 | * symposium on Parallel algorithms and architectures, ACM Press, | |
30 | * (2002), 73-82. | |
31 | * | |
1475579c | 32 | * Some specificities of this Lock-Free Resizable RCU Hash Table |
e753ff5a MD |
33 | * implementation: |
34 | * | |
35 | * - RCU read-side critical section allows readers to perform hash | |
36 | * table lookups and use the returned objects safely by delaying | |
37 | * memory reclaim of a grace period. | |
38 | * - Add and remove operations are lock-free, and do not need to | |
39 | * allocate memory. They need to be executed within RCU read-side | |
40 | * critical section to ensure the objects they read are valid and to | |
41 | * deal with the cmpxchg ABA problem. | |
42 | * - add and add_unique operations are supported. add_unique checks if | |
43 | * the node key already exists in the hash table. It ensures no key | |
44 | * duplicata exists. | |
45 | * - The resize operation executes concurrently with add/remove/lookup. | |
46 | * - Hash table nodes are contained within a split-ordered list. This | |
47 | * list is ordered by incrementing reversed-bits-hash value. | |
48 | * - An index of dummy nodes is kept. These dummy nodes are the hash | |
49 | * table "buckets", and they are also chained together in the | |
50 | * split-ordered list, which allows recursive expansion. | |
1475579c MD |
51 | * - The resize operation for small tables only allows expanding the hash table. |
52 | * It is triggered automatically by detecting long chains in the add | |
53 | * operation. | |
54 | * - The resize operation for larger tables (and available through an | |
55 | * API) allows both expanding and shrinking the hash table. | |
56 | * - Per-CPU Split-counters are used to keep track of the number of | |
57 | * nodes within the hash table for automatic resize triggering. | |
e753ff5a MD |
58 | * - Resize operation initiated by long chain detection is executed by a |
59 | * call_rcu thread, which keeps lock-freedom of add and remove. | |
60 | * - Resize operations are protected by a mutex. | |
61 | * - The removal operation is split in two parts: first, a "removed" | |
62 | * flag is set in the next pointer within the node to remove. Then, | |
63 | * a "garbage collection" is performed in the bucket containing the | |
64 | * removed node (from the start of the bucket up to the removed node). | |
65 | * All encountered nodes with "removed" flag set in their next | |
66 | * pointers are removed from the linked-list. If the cmpxchg used for | |
67 | * removal fails (due to concurrent garbage-collection or concurrent | |
68 | * add), we retry from the beginning of the bucket. This ensures that | |
69 | * the node with "removed" flag set is removed from the hash table | |
70 | * (not visible to lookups anymore) before the RCU read-side critical | |
71 | * section held across removal ends. Furthermore, this ensures that | |
72 | * the node with "removed" flag set is removed from the linked-list | |
73 | * before its memory is reclaimed. Only the thread which removal | |
74 | * successfully set the "removed" flag (with a cmpxchg) into a node's | |
75 | * next pointer is considered to have succeeded its removal (and thus | |
76 | * owns the node to reclaim). Because we garbage-collect starting from | |
77 | * an invariant node (the start-of-bucket dummy node) up to the | |
78 | * "removed" node (or find a reverse-hash that is higher), we are sure | |
79 | * that a successful traversal of the chain leads to a chain that is | |
80 | * present in the linked-list (the start node is never removed) and | |
81 | * that is does not contain the "removed" node anymore, even if | |
82 | * concurrent delete/add operations are changing the structure of the | |
83 | * list concurrently. | |
29e669f6 MD |
84 | * - The add operation performs gargage collection of buckets if it |
85 | * encounters nodes with removed flag set in the bucket where it wants | |
86 | * to add its new node. This ensures lock-freedom of add operation by | |
87 | * helping the remover unlink nodes from the list rather than to wait | |
88 | * for it do to so. | |
e753ff5a MD |
89 | * - A RCU "order table" indexed by log2(hash index) is copied and |
90 | * expanded by the resize operation. This order table allows finding | |
91 | * the "dummy node" tables. | |
92 | * - There is one dummy node table per hash index order. The size of | |
93 | * each dummy node table is half the number of hashes contained in | |
94 | * this order. | |
95 | * - call_rcu is used to garbage-collect the old order table. | |
96 | * - The per-order dummy node tables contain a compact version of the | |
97 | * hash table nodes. These tables are invariant after they are | |
98 | * populated into the hash table. | |
1475579c MD |
99 | * |
100 | * A bit of ascii art explanation: | |
101 | * | |
102 | * Order index is the off-by-one compare to the actual power of 2 because | |
103 | * we use index 0 to deal with the 0 special-case. | |
104 | * | |
105 | * This shows the nodes for a small table ordered by reversed bits: | |
106 | * | |
107 | * bits reverse | |
108 | * 0 000 000 | |
109 | * 4 100 001 | |
110 | * 2 010 010 | |
111 | * 6 110 011 | |
112 | * 1 001 100 | |
113 | * 5 101 101 | |
114 | * 3 011 110 | |
115 | * 7 111 111 | |
116 | * | |
117 | * This shows the nodes in order of non-reversed bits, linked by | |
118 | * reversed-bit order. | |
119 | * | |
120 | * order bits reverse | |
121 | * 0 0 000 000 | |
0adc36a8 LJ |
122 | * 1 | 1 001 100 <- |
123 | * 2 | | 2 010 010 <- | | |
f6fdd688 | 124 | * | | | 3 011 110 | <- | |
1475579c MD |
125 | * 3 -> | | | 4 100 001 | | |
126 | * -> | | 5 101 101 | | |
127 | * -> | 6 110 011 | |
128 | * -> 7 111 111 | |
e753ff5a MD |
129 | */ |
130 | ||
2ed95849 MD |
131 | #define _LGPL_SOURCE |
132 | #include <stdlib.h> | |
e0ba718a MD |
133 | #include <errno.h> |
134 | #include <assert.h> | |
135 | #include <stdio.h> | |
abc490a1 | 136 | #include <stdint.h> |
f000907d | 137 | #include <string.h> |
e0ba718a | 138 | |
15cfbec7 | 139 | #include "config.h" |
2ed95849 | 140 | #include <urcu.h> |
abc490a1 | 141 | #include <urcu-call-rcu.h> |
a42cc659 MD |
142 | #include <urcu/arch.h> |
143 | #include <urcu/uatomic.h> | |
a42cc659 | 144 | #include <urcu/compiler.h> |
abc490a1 | 145 | #include <urcu/rculfhash.h> |
5e28c532 | 146 | #include <stdio.h> |
464a1ec9 | 147 | #include <pthread.h> |
44395fb7 | 148 | |
f9830efd | 149 | #ifdef DEBUG |
f0c29ed7 | 150 | #define dbg_printf(fmt, args...) printf("[debug rculfhash] " fmt, ## args) |
f9830efd | 151 | #else |
e753ff5a | 152 | #define dbg_printf(fmt, args...) |
f9830efd MD |
153 | #endif |
154 | ||
f8994aee MD |
155 | /* |
156 | * Per-CPU split-counters lazily update the global counter each 1024 | |
157 | * addition/removal. It automatically keeps track of resize required. | |
158 | * We use the bucket length as indicator for need to expand for small | |
159 | * tables and machines lacking per-cpu data suppport. | |
160 | */ | |
161 | #define COUNT_COMMIT_ORDER 10 | |
6ea6bc67 MD |
162 | #define CHAIN_LEN_TARGET 1 |
163 | #define CHAIN_LEN_RESIZE_THRESHOLD 3 | |
2ed95849 | 164 | |
cd95516d | 165 | /* |
76a73da8 | 166 | * Define the minimum table size. |
cd95516d | 167 | */ |
c9edd44a | 168 | #define MIN_TABLE_SIZE 1 |
cd95516d | 169 | |
4105056a MD |
170 | #if (CAA_BITS_PER_LONG == 32) |
171 | #define MAX_TABLE_ORDER 32 | |
172 | #else | |
173 | #define MAX_TABLE_ORDER 64 | |
174 | #endif | |
175 | ||
b7d619b0 MD |
176 | /* |
177 | * Minimum number of dummy nodes to touch per thread to parallelize grow/shrink. | |
178 | */ | |
6083a889 MD |
179 | #define MIN_PARTITION_PER_THREAD_ORDER 12 |
180 | #define MIN_PARTITION_PER_THREAD (1UL << MIN_PARTITION_PER_THREAD_ORDER) | |
b7d619b0 | 181 | |
4105056a MD |
182 | #ifndef min |
183 | #define min(a, b) ((a) < (b) ? (a) : (b)) | |
184 | #endif | |
185 | ||
abc490a1 MD |
186 | #ifndef max |
187 | #define max(a, b) ((a) > (b) ? (a) : (b)) | |
188 | #endif | |
2ed95849 | 189 | |
d95bd160 MD |
190 | /* |
191 | * The removed flag needs to be updated atomically with the pointer. | |
48ed1c18 | 192 | * It indicates that no node must attach to the node scheduled for |
b198f0fd | 193 | * removal, and that node garbage collection must be performed. |
d95bd160 MD |
194 | * The dummy flag does not require to be updated atomically with the |
195 | * pointer, but it is added as a pointer low bit flag to save space. | |
196 | */ | |
d37166c6 | 197 | #define REMOVED_FLAG (1UL << 0) |
b198f0fd MD |
198 | #define DUMMY_FLAG (1UL << 1) |
199 | #define FLAGS_MASK ((1UL << 2) - 1) | |
d37166c6 | 200 | |
bb7b2f26 | 201 | /* Value of the end pointer. Should not interact with flags. */ |
f9c80341 | 202 | #define END_VALUE NULL |
bb7b2f26 | 203 | |
df44348d | 204 | struct ht_items_count { |
860d07e8 | 205 | unsigned long add, del; |
df44348d MD |
206 | } __attribute__((aligned(CAA_CACHE_LINE_SIZE))); |
207 | ||
1475579c | 208 | struct rcu_level { |
0d14ceb2 | 209 | /* Note: manually update allocation length when adding a field */ |
1475579c MD |
210 | struct _cds_lfht_node nodes[0]; |
211 | }; | |
212 | ||
395270b6 | 213 | struct rcu_table { |
4105056a | 214 | unsigned long size; /* always a power of 2, shared (RCU) */ |
f9830efd | 215 | unsigned long resize_target; |
11519af6 | 216 | int resize_initiated; |
4105056a | 217 | struct rcu_level *tbl[MAX_TABLE_ORDER]; |
395270b6 MD |
218 | }; |
219 | ||
14044b37 | 220 | struct cds_lfht { |
4105056a | 221 | struct rcu_table t; |
14044b37 MD |
222 | cds_lfht_hash_fct hash_fct; |
223 | cds_lfht_compare_fct compare_fct; | |
732ad076 | 224 | unsigned long hash_seed; |
b8af5011 | 225 | int flags; |
5f511391 MD |
226 | /* |
227 | * We need to put the work threads offline (QSBR) when taking this | |
228 | * mutex, because we use synchronize_rcu within this mutex critical | |
229 | * section, which waits on read-side critical sections, and could | |
230 | * therefore cause grace-period deadlock if we hold off RCU G.P. | |
231 | * completion. | |
232 | */ | |
464a1ec9 | 233 | pthread_mutex_t resize_mutex; /* resize mutex: add/del mutex */ |
33c7c748 | 234 | unsigned int in_progress_resize, in_progress_destroy; |
14044b37 | 235 | void (*cds_lfht_call_rcu)(struct rcu_head *head, |
abc490a1 | 236 | void (*func)(struct rcu_head *head)); |
1475579c | 237 | void (*cds_lfht_synchronize_rcu)(void); |
01dbfa62 MD |
238 | void (*cds_lfht_rcu_read_lock)(void); |
239 | void (*cds_lfht_rcu_read_unlock)(void); | |
5f511391 MD |
240 | void (*cds_lfht_rcu_thread_offline)(void); |
241 | void (*cds_lfht_rcu_thread_online)(void); | |
b7d619b0 MD |
242 | void (*cds_lfht_rcu_register_thread)(void); |
243 | void (*cds_lfht_rcu_unregister_thread)(void); | |
244 | pthread_attr_t *resize_attr; /* Resize threads attributes */ | |
7de5ccfd | 245 | long count; /* global approximate item count */ |
df44348d | 246 | struct ht_items_count *percpu_count; /* per-cpu item count */ |
2ed95849 MD |
247 | }; |
248 | ||
abc490a1 MD |
249 | struct rcu_resize_work { |
250 | struct rcu_head head; | |
14044b37 | 251 | struct cds_lfht *ht; |
abc490a1 | 252 | }; |
2ed95849 | 253 | |
b7d619b0 | 254 | struct partition_resize_work { |
1af6e26e | 255 | pthread_t thread_id; |
b7d619b0 MD |
256 | struct cds_lfht *ht; |
257 | unsigned long i, start, len; | |
258 | void (*fct)(struct cds_lfht *ht, unsigned long i, | |
259 | unsigned long start, unsigned long len); | |
260 | }; | |
261 | ||
76a73da8 | 262 | static |
83beee94 MD |
263 | void _cds_lfht_add(struct cds_lfht *ht, |
264 | unsigned long size, | |
265 | struct cds_lfht_node *node, | |
266 | struct cds_lfht_iter *unique_ret, | |
267 | int dummy); | |
48ed1c18 | 268 | |
abc490a1 MD |
269 | /* |
270 | * Algorithm to reverse bits in a word by lookup table, extended to | |
271 | * 64-bit words. | |
f9830efd | 272 | * Source: |
abc490a1 | 273 | * http://graphics.stanford.edu/~seander/bithacks.html#BitReverseTable |
f9830efd | 274 | * Originally from Public Domain. |
abc490a1 MD |
275 | */ |
276 | ||
277 | static const uint8_t BitReverseTable256[256] = | |
2ed95849 | 278 | { |
abc490a1 MD |
279 | #define R2(n) (n), (n) + 2*64, (n) + 1*64, (n) + 3*64 |
280 | #define R4(n) R2(n), R2((n) + 2*16), R2((n) + 1*16), R2((n) + 3*16) | |
281 | #define R6(n) R4(n), R4((n) + 2*4 ), R4((n) + 1*4 ), R4((n) + 3*4 ) | |
282 | R6(0), R6(2), R6(1), R6(3) | |
283 | }; | |
284 | #undef R2 | |
285 | #undef R4 | |
286 | #undef R6 | |
2ed95849 | 287 | |
abc490a1 MD |
288 | static |
289 | uint8_t bit_reverse_u8(uint8_t v) | |
290 | { | |
291 | return BitReverseTable256[v]; | |
292 | } | |
ab7d5fc6 | 293 | |
abc490a1 MD |
294 | static __attribute__((unused)) |
295 | uint32_t bit_reverse_u32(uint32_t v) | |
296 | { | |
297 | return ((uint32_t) bit_reverse_u8(v) << 24) | | |
298 | ((uint32_t) bit_reverse_u8(v >> 8) << 16) | | |
299 | ((uint32_t) bit_reverse_u8(v >> 16) << 8) | | |
300 | ((uint32_t) bit_reverse_u8(v >> 24)); | |
2ed95849 MD |
301 | } |
302 | ||
abc490a1 MD |
303 | static __attribute__((unused)) |
304 | uint64_t bit_reverse_u64(uint64_t v) | |
2ed95849 | 305 | { |
abc490a1 MD |
306 | return ((uint64_t) bit_reverse_u8(v) << 56) | |
307 | ((uint64_t) bit_reverse_u8(v >> 8) << 48) | | |
308 | ((uint64_t) bit_reverse_u8(v >> 16) << 40) | | |
309 | ((uint64_t) bit_reverse_u8(v >> 24) << 32) | | |
310 | ((uint64_t) bit_reverse_u8(v >> 32) << 24) | | |
311 | ((uint64_t) bit_reverse_u8(v >> 40) << 16) | | |
312 | ((uint64_t) bit_reverse_u8(v >> 48) << 8) | | |
313 | ((uint64_t) bit_reverse_u8(v >> 56)); | |
314 | } | |
315 | ||
316 | static | |
317 | unsigned long bit_reverse_ulong(unsigned long v) | |
318 | { | |
319 | #if (CAA_BITS_PER_LONG == 32) | |
320 | return bit_reverse_u32(v); | |
321 | #else | |
322 | return bit_reverse_u64(v); | |
323 | #endif | |
324 | } | |
325 | ||
f9830efd | 326 | /* |
24365af7 MD |
327 | * fls: returns the position of the most significant bit. |
328 | * Returns 0 if no bit is set, else returns the position of the most | |
329 | * significant bit (from 1 to 32 on 32-bit, from 1 to 64 on 64-bit). | |
f9830efd | 330 | */ |
24365af7 MD |
331 | #if defined(__i386) || defined(__x86_64) |
332 | static inline | |
333 | unsigned int fls_u32(uint32_t x) | |
f9830efd | 334 | { |
24365af7 MD |
335 | int r; |
336 | ||
337 | asm("bsrl %1,%0\n\t" | |
338 | "jnz 1f\n\t" | |
339 | "movl $-1,%0\n\t" | |
340 | "1:\n\t" | |
341 | : "=r" (r) : "rm" (x)); | |
342 | return r + 1; | |
343 | } | |
344 | #define HAS_FLS_U32 | |
345 | #endif | |
346 | ||
347 | #if defined(__x86_64) | |
348 | static inline | |
349 | unsigned int fls_u64(uint64_t x) | |
350 | { | |
351 | long r; | |
352 | ||
353 | asm("bsrq %1,%0\n\t" | |
354 | "jnz 1f\n\t" | |
355 | "movq $-1,%0\n\t" | |
356 | "1:\n\t" | |
357 | : "=r" (r) : "rm" (x)); | |
358 | return r + 1; | |
359 | } | |
360 | #define HAS_FLS_U64 | |
361 | #endif | |
362 | ||
363 | #ifndef HAS_FLS_U64 | |
364 | static __attribute__((unused)) | |
365 | unsigned int fls_u64(uint64_t x) | |
366 | { | |
367 | unsigned int r = 64; | |
368 | ||
369 | if (!x) | |
370 | return 0; | |
371 | ||
372 | if (!(x & 0xFFFFFFFF00000000ULL)) { | |
373 | x <<= 32; | |
374 | r -= 32; | |
375 | } | |
376 | if (!(x & 0xFFFF000000000000ULL)) { | |
377 | x <<= 16; | |
378 | r -= 16; | |
379 | } | |
380 | if (!(x & 0xFF00000000000000ULL)) { | |
381 | x <<= 8; | |
382 | r -= 8; | |
383 | } | |
384 | if (!(x & 0xF000000000000000ULL)) { | |
385 | x <<= 4; | |
386 | r -= 4; | |
387 | } | |
388 | if (!(x & 0xC000000000000000ULL)) { | |
389 | x <<= 2; | |
390 | r -= 2; | |
391 | } | |
392 | if (!(x & 0x8000000000000000ULL)) { | |
393 | x <<= 1; | |
394 | r -= 1; | |
395 | } | |
396 | return r; | |
397 | } | |
398 | #endif | |
399 | ||
400 | #ifndef HAS_FLS_U32 | |
401 | static __attribute__((unused)) | |
402 | unsigned int fls_u32(uint32_t x) | |
403 | { | |
404 | unsigned int r = 32; | |
f9830efd | 405 | |
24365af7 MD |
406 | if (!x) |
407 | return 0; | |
408 | if (!(x & 0xFFFF0000U)) { | |
409 | x <<= 16; | |
410 | r -= 16; | |
411 | } | |
412 | if (!(x & 0xFF000000U)) { | |
413 | x <<= 8; | |
414 | r -= 8; | |
415 | } | |
416 | if (!(x & 0xF0000000U)) { | |
417 | x <<= 4; | |
418 | r -= 4; | |
419 | } | |
420 | if (!(x & 0xC0000000U)) { | |
421 | x <<= 2; | |
422 | r -= 2; | |
423 | } | |
424 | if (!(x & 0x80000000U)) { | |
425 | x <<= 1; | |
426 | r -= 1; | |
427 | } | |
428 | return r; | |
429 | } | |
430 | #endif | |
431 | ||
432 | unsigned int fls_ulong(unsigned long x) | |
f9830efd | 433 | { |
24365af7 MD |
434 | #if (CAA_BITS_PER_lONG == 32) |
435 | return fls_u32(x); | |
436 | #else | |
437 | return fls_u64(x); | |
438 | #endif | |
439 | } | |
f9830efd | 440 | |
920f8ef6 LJ |
441 | /* |
442 | * Return the minimum order for which x <= (1UL << order). | |
443 | * Return -1 if x is 0. | |
444 | */ | |
24365af7 MD |
445 | int get_count_order_u32(uint32_t x) |
446 | { | |
920f8ef6 LJ |
447 | if (!x) |
448 | return -1; | |
24365af7 | 449 | |
920f8ef6 | 450 | return fls_u32(x - 1); |
24365af7 MD |
451 | } |
452 | ||
920f8ef6 LJ |
453 | /* |
454 | * Return the minimum order for which x <= (1UL << order). | |
455 | * Return -1 if x is 0. | |
456 | */ | |
24365af7 MD |
457 | int get_count_order_ulong(unsigned long x) |
458 | { | |
920f8ef6 LJ |
459 | if (!x) |
460 | return -1; | |
24365af7 | 461 | |
920f8ef6 | 462 | return fls_ulong(x - 1); |
f9830efd MD |
463 | } |
464 | ||
98808fb1 MD |
465 | #ifdef POISON_FREE |
466 | #define poison_free(ptr) \ | |
467 | do { \ | |
468 | memset(ptr, 0x42, sizeof(*(ptr))); \ | |
469 | free(ptr); \ | |
470 | } while (0) | |
471 | #else | |
472 | #define poison_free(ptr) free(ptr) | |
473 | #endif | |
474 | ||
f9830efd | 475 | static |
4105056a | 476 | void cds_lfht_resize_lazy(struct cds_lfht *ht, unsigned long size, int growth); |
f9830efd | 477 | |
df44348d MD |
478 | /* |
479 | * If the sched_getcpu() and sysconf(_SC_NPROCESSORS_CONF) calls are | |
480 | * available, then we support hash table item accounting. | |
481 | * In the unfortunate event the number of CPUs reported would be | |
482 | * inaccurate, we use modulo arithmetic on the number of CPUs we got. | |
483 | */ | |
df44348d MD |
484 | #if defined(HAVE_SCHED_GETCPU) && defined(HAVE_SYSCONF) |
485 | ||
f8994aee | 486 | static |
4105056a | 487 | void cds_lfht_resize_lazy_count(struct cds_lfht *ht, unsigned long size, |
f8994aee MD |
488 | unsigned long count); |
489 | ||
df44348d MD |
490 | static long nr_cpus_mask = -1; |
491 | ||
492 | static | |
493 | struct ht_items_count *alloc_per_cpu_items_count(void) | |
494 | { | |
495 | struct ht_items_count *count; | |
496 | ||
497 | switch (nr_cpus_mask) { | |
498 | case -2: | |
499 | return NULL; | |
500 | case -1: | |
501 | { | |
502 | long maxcpus; | |
503 | ||
504 | maxcpus = sysconf(_SC_NPROCESSORS_CONF); | |
505 | if (maxcpus <= 0) { | |
506 | nr_cpus_mask = -2; | |
507 | return NULL; | |
508 | } | |
509 | /* | |
510 | * round up number of CPUs to next power of two, so we | |
511 | * can use & for modulo. | |
512 | */ | |
513 | maxcpus = 1UL << get_count_order_ulong(maxcpus); | |
514 | nr_cpus_mask = maxcpus - 1; | |
515 | } | |
516 | /* Fall-through */ | |
517 | default: | |
518 | return calloc(nr_cpus_mask + 1, sizeof(*count)); | |
519 | } | |
520 | } | |
521 | ||
522 | static | |
523 | void free_per_cpu_items_count(struct ht_items_count *count) | |
524 | { | |
98808fb1 | 525 | poison_free(count); |
df44348d MD |
526 | } |
527 | ||
528 | static | |
529 | int ht_get_cpu(void) | |
530 | { | |
531 | int cpu; | |
532 | ||
533 | assert(nr_cpus_mask >= 0); | |
534 | cpu = sched_getcpu(); | |
535 | if (unlikely(cpu < 0)) | |
536 | return cpu; | |
537 | else | |
538 | return cpu & nr_cpus_mask; | |
539 | } | |
540 | ||
541 | static | |
4105056a | 542 | void ht_count_add(struct cds_lfht *ht, unsigned long size) |
df44348d | 543 | { |
3171717f | 544 | unsigned long percpu_count; |
df44348d MD |
545 | int cpu; |
546 | ||
547 | if (unlikely(!ht->percpu_count)) | |
3171717f | 548 | return; |
df44348d MD |
549 | cpu = ht_get_cpu(); |
550 | if (unlikely(cpu < 0)) | |
3171717f MD |
551 | return; |
552 | percpu_count = uatomic_add_return(&ht->percpu_count[cpu].add, 1); | |
df44348d | 553 | if (unlikely(!(percpu_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))) { |
e3ecefd6 | 554 | long count; |
df44348d MD |
555 | |
556 | dbg_printf("add percpu %lu\n", percpu_count); | |
557 | count = uatomic_add_return(&ht->count, | |
558 | 1UL << COUNT_COMMIT_ORDER); | |
559 | /* If power of 2 */ | |
560 | if (!(count & (count - 1))) { | |
4105056a | 561 | if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) < size) |
f8994aee | 562 | return; |
e3ecefd6 | 563 | dbg_printf("add set global %ld\n", count); |
4105056a | 564 | cds_lfht_resize_lazy_count(ht, size, |
6ea6bc67 | 565 | count >> (CHAIN_LEN_TARGET - 1)); |
df44348d MD |
566 | } |
567 | } | |
568 | } | |
569 | ||
570 | static | |
860d07e8 | 571 | void ht_count_del(struct cds_lfht *ht, unsigned long size) |
df44348d MD |
572 | { |
573 | unsigned long percpu_count; | |
3171717f | 574 | int cpu; |
df44348d | 575 | |
3171717f MD |
576 | if (unlikely(!ht->percpu_count)) |
577 | return; | |
578 | cpu = ht_get_cpu(); | |
579 | if (unlikely(cpu < 0)) | |
580 | return; | |
80d90c06 | 581 | percpu_count = uatomic_add_return(&ht->percpu_count[cpu].del, 1); |
df44348d | 582 | if (unlikely(!(percpu_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))) { |
e3ecefd6 | 583 | long count; |
df44348d | 584 | |
860d07e8 | 585 | dbg_printf("del percpu %lu\n", percpu_count); |
df44348d | 586 | count = uatomic_add_return(&ht->count, |
3171717f | 587 | -(1UL << COUNT_COMMIT_ORDER)); |
df44348d MD |
588 | /* If power of 2 */ |
589 | if (!(count & (count - 1))) { | |
4105056a | 590 | if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) >= size) |
f8994aee | 591 | return; |
e3ecefd6 MD |
592 | dbg_printf("del set global %ld\n", count); |
593 | /* | |
c941bb9e | 594 | * Don't shrink table if the number of nodes is below a |
e3ecefd6 MD |
595 | * certain threshold. |
596 | */ | |
c941bb9e | 597 | if (count < (1UL << COUNT_COMMIT_ORDER) * (nr_cpus_mask + 1)) |
e3ecefd6 | 598 | return; |
4105056a | 599 | cds_lfht_resize_lazy_count(ht, size, |
6ea6bc67 | 600 | count >> (CHAIN_LEN_TARGET - 1)); |
df44348d MD |
601 | } |
602 | } | |
603 | } | |
604 | ||
605 | #else /* #if defined(HAVE_SCHED_GETCPU) && defined(HAVE_SYSCONF) */ | |
606 | ||
91452a6a | 607 | static const long nr_cpus_mask = -2; |
df44348d MD |
608 | |
609 | static | |
610 | struct ht_items_count *alloc_per_cpu_items_count(void) | |
611 | { | |
612 | return NULL; | |
613 | } | |
614 | ||
615 | static | |
616 | void free_per_cpu_items_count(struct ht_items_count *count) | |
617 | { | |
618 | } | |
619 | ||
620 | static | |
4105056a | 621 | void ht_count_add(struct cds_lfht *ht, unsigned long size) |
df44348d MD |
622 | { |
623 | } | |
624 | ||
625 | static | |
860d07e8 | 626 | void ht_count_del(struct cds_lfht *ht, unsigned long size) |
df44348d MD |
627 | { |
628 | } | |
629 | ||
630 | #endif /* #else #if defined(HAVE_SCHED_GETCPU) && defined(HAVE_SYSCONF) */ | |
631 | ||
632 | ||
f9830efd | 633 | static |
4105056a | 634 | void check_resize(struct cds_lfht *ht, unsigned long size, uint32_t chain_len) |
f9830efd | 635 | { |
f8994aee MD |
636 | unsigned long count; |
637 | ||
b8af5011 MD |
638 | if (!(ht->flags & CDS_LFHT_AUTO_RESIZE)) |
639 | return; | |
f8994aee MD |
640 | count = uatomic_read(&ht->count); |
641 | /* | |
642 | * Use bucket-local length for small table expand and for | |
643 | * environments lacking per-cpu data support. | |
644 | */ | |
645 | if (count >= (1UL << COUNT_COMMIT_ORDER)) | |
646 | return; | |
24365af7 | 647 | if (chain_len > 100) |
f0c29ed7 | 648 | dbg_printf("WARNING: large chain length: %u.\n", |
24365af7 | 649 | chain_len); |
3390d470 | 650 | if (chain_len >= CHAIN_LEN_RESIZE_THRESHOLD) |
4105056a | 651 | cds_lfht_resize_lazy(ht, size, |
01370f0b | 652 | get_count_order_u32(chain_len - (CHAIN_LEN_TARGET - 1))); |
f9830efd MD |
653 | } |
654 | ||
abc490a1 | 655 | static |
14044b37 | 656 | struct cds_lfht_node *clear_flag(struct cds_lfht_node *node) |
abc490a1 | 657 | { |
14044b37 | 658 | return (struct cds_lfht_node *) (((unsigned long) node) & ~FLAGS_MASK); |
abc490a1 MD |
659 | } |
660 | ||
661 | static | |
14044b37 | 662 | int is_removed(struct cds_lfht_node *node) |
abc490a1 | 663 | { |
d37166c6 | 664 | return ((unsigned long) node) & REMOVED_FLAG; |
abc490a1 MD |
665 | } |
666 | ||
667 | static | |
14044b37 | 668 | struct cds_lfht_node *flag_removed(struct cds_lfht_node *node) |
abc490a1 | 669 | { |
14044b37 | 670 | return (struct cds_lfht_node *) (((unsigned long) node) | REMOVED_FLAG); |
abc490a1 MD |
671 | } |
672 | ||
f5596c94 | 673 | static |
14044b37 | 674 | int is_dummy(struct cds_lfht_node *node) |
f5596c94 MD |
675 | { |
676 | return ((unsigned long) node) & DUMMY_FLAG; | |
677 | } | |
678 | ||
679 | static | |
14044b37 | 680 | struct cds_lfht_node *flag_dummy(struct cds_lfht_node *node) |
f5596c94 | 681 | { |
14044b37 | 682 | return (struct cds_lfht_node *) (((unsigned long) node) | DUMMY_FLAG); |
f5596c94 | 683 | } |
bb7b2f26 MD |
684 | |
685 | static | |
686 | struct cds_lfht_node *get_end(void) | |
687 | { | |
688 | return (struct cds_lfht_node *) END_VALUE; | |
689 | } | |
690 | ||
691 | static | |
692 | int is_end(struct cds_lfht_node *node) | |
693 | { | |
694 | return clear_flag(node) == (struct cds_lfht_node *) END_VALUE; | |
695 | } | |
696 | ||
abc490a1 | 697 | static |
f9830efd | 698 | unsigned long _uatomic_max(unsigned long *ptr, unsigned long v) |
abc490a1 MD |
699 | { |
700 | unsigned long old1, old2; | |
701 | ||
702 | old1 = uatomic_read(ptr); | |
703 | do { | |
704 | old2 = old1; | |
705 | if (old2 >= v) | |
f9830efd | 706 | return old2; |
abc490a1 | 707 | } while ((old1 = uatomic_cmpxchg(ptr, old2, v)) != old2); |
f9830efd | 708 | return v; |
abc490a1 MD |
709 | } |
710 | ||
f4a9cc0b LJ |
711 | static |
712 | struct _cds_lfht_node *lookup_bucket(struct cds_lfht *ht, unsigned long size, | |
713 | unsigned long hash) | |
714 | { | |
715 | unsigned long index, order; | |
716 | ||
717 | assert(size > 0); | |
718 | index = hash & (size - 1); | |
a4ea2223 LJ |
719 | /* |
720 | * equivalent to get_count_order_ulong(index + 1), but optimizes | |
721 | * away the non-existing 0 special-case for | |
722 | * get_count_order_ulong. | |
723 | */ | |
724 | order = fls_ulong(index); | |
f4a9cc0b LJ |
725 | |
726 | dbg_printf("lookup hash %lu index %lu order %lu aridx %lu\n", | |
727 | hash, index, order, index & (!order ? 0 : ((1UL << (order - 1)) - 1))); | |
728 | ||
729 | return &ht->t.tbl[order]->nodes[index & (!order ? 0 : ((1UL << (order - 1)) - 1))]; | |
730 | } | |
731 | ||
273399de MD |
732 | /* |
733 | * Remove all logically deleted nodes from a bucket up to a certain node key. | |
734 | */ | |
735 | static | |
f9c80341 | 736 | void _cds_lfht_gc_bucket(struct cds_lfht_node *dummy, struct cds_lfht_node *node) |
273399de | 737 | { |
14044b37 | 738 | struct cds_lfht_node *iter_prev, *iter, *next, *new_next; |
273399de | 739 | |
c90201ac MD |
740 | assert(!is_dummy(dummy)); |
741 | assert(!is_removed(dummy)); | |
742 | assert(!is_dummy(node)); | |
743 | assert(!is_removed(node)); | |
273399de MD |
744 | for (;;) { |
745 | iter_prev = dummy; | |
746 | /* We can always skip the dummy node initially */ | |
cc4fcb10 | 747 | iter = rcu_dereference(iter_prev->p.next); |
b4cb483f | 748 | assert(!is_removed(iter)); |
cc4fcb10 | 749 | assert(iter_prev->p.reverse_hash <= node->p.reverse_hash); |
bd4db153 MD |
750 | /* |
751 | * We should never be called with dummy (start of chain) | |
752 | * and logically removed node (end of path compression | |
753 | * marker) being the actual same node. This would be a | |
754 | * bug in the algorithm implementation. | |
755 | */ | |
756 | assert(dummy != node); | |
273399de | 757 | for (;;) { |
bb7b2f26 | 758 | if (unlikely(is_end(iter))) |
f9c80341 | 759 | return; |
76412f24 | 760 | if (likely(clear_flag(iter)->p.reverse_hash > node->p.reverse_hash)) |
f9c80341 | 761 | return; |
cc4fcb10 | 762 | next = rcu_dereference(clear_flag(iter)->p.next); |
b198f0fd | 763 | if (likely(is_removed(next))) |
273399de | 764 | break; |
b453eae1 | 765 | iter_prev = clear_flag(iter); |
273399de MD |
766 | iter = next; |
767 | } | |
b198f0fd | 768 | assert(!is_removed(iter)); |
f5596c94 MD |
769 | if (is_dummy(iter)) |
770 | new_next = flag_dummy(clear_flag(next)); | |
771 | else | |
772 | new_next = clear_flag(next); | |
773 | (void) uatomic_cmpxchg(&iter_prev->p.next, iter, new_next); | |
273399de | 774 | } |
f9c80341 | 775 | return; |
273399de MD |
776 | } |
777 | ||
9357c415 MD |
778 | static |
779 | int _cds_lfht_replace(struct cds_lfht *ht, unsigned long size, | |
780 | struct cds_lfht_node *old_node, | |
3fb86f26 | 781 | struct cds_lfht_node *old_next, |
9357c415 MD |
782 | struct cds_lfht_node *new_node) |
783 | { | |
3fb86f26 | 784 | struct cds_lfht_node *dummy, *ret_next; |
9357c415 | 785 | struct _cds_lfht_node *lookup; |
9357c415 MD |
786 | |
787 | if (!old_node) /* Return -ENOENT if asked to replace NULL node */ | |
7801dadd | 788 | return -ENOENT; |
9357c415 MD |
789 | |
790 | assert(!is_removed(old_node)); | |
791 | assert(!is_dummy(old_node)); | |
792 | assert(!is_removed(new_node)); | |
793 | assert(!is_dummy(new_node)); | |
794 | assert(new_node != old_node); | |
3fb86f26 | 795 | for (;;) { |
9357c415 | 796 | /* Insert after node to be replaced */ |
9357c415 MD |
797 | if (is_removed(old_next)) { |
798 | /* | |
799 | * Too late, the old node has been removed under us | |
800 | * between lookup and replace. Fail. | |
801 | */ | |
7801dadd | 802 | return -ENOENT; |
9357c415 MD |
803 | } |
804 | assert(!is_dummy(old_next)); | |
805 | assert(new_node != clear_flag(old_next)); | |
806 | new_node->p.next = clear_flag(old_next); | |
807 | /* | |
808 | * Here is the whole trick for lock-free replace: we add | |
809 | * the replacement node _after_ the node we want to | |
810 | * replace by atomically setting its next pointer at the | |
811 | * same time we set its removal flag. Given that | |
812 | * the lookups/get next use an iterator aware of the | |
813 | * next pointer, they will either skip the old node due | |
814 | * to the removal flag and see the new node, or use | |
815 | * the old node, but will not see the new one. | |
816 | */ | |
817 | ret_next = uatomic_cmpxchg(&old_node->p.next, | |
818 | old_next, flag_removed(new_node)); | |
3fb86f26 | 819 | if (ret_next == old_next) |
7801dadd | 820 | break; /* We performed the replacement. */ |
3fb86f26 LJ |
821 | old_next = ret_next; |
822 | } | |
9357c415 | 823 | |
9357c415 MD |
824 | /* |
825 | * Ensure that the old node is not visible to readers anymore: | |
826 | * lookup for the node, and remove it (along with any other | |
827 | * logically removed node) if found. | |
828 | */ | |
f4a9cc0b | 829 | lookup = lookup_bucket(ht, size, bit_reverse_ulong(old_node->p.reverse_hash)); |
9357c415 MD |
830 | dummy = (struct cds_lfht_node *) lookup; |
831 | _cds_lfht_gc_bucket(dummy, new_node); | |
7801dadd LJ |
832 | |
833 | assert(is_removed(rcu_dereference(old_node->p.next))); | |
834 | return 0; | |
9357c415 MD |
835 | } |
836 | ||
83beee94 MD |
837 | /* |
838 | * A non-NULL unique_ret pointer uses the "add unique" (or uniquify) add | |
839 | * mode. A NULL unique_ret allows creation of duplicate keys. | |
840 | */ | |
abc490a1 | 841 | static |
83beee94 MD |
842 | void _cds_lfht_add(struct cds_lfht *ht, |
843 | unsigned long size, | |
844 | struct cds_lfht_node *node, | |
845 | struct cds_lfht_iter *unique_ret, | |
846 | int dummy) | |
abc490a1 | 847 | { |
14044b37 | 848 | struct cds_lfht_node *iter_prev, *iter, *next, *new_node, *new_next, |
960c9e4f | 849 | *return_node; |
14044b37 | 850 | struct _cds_lfht_node *lookup; |
abc490a1 | 851 | |
c90201ac MD |
852 | assert(!is_dummy(node)); |
853 | assert(!is_removed(node)); | |
4105056a | 854 | if (!size) { |
f5596c94 | 855 | assert(dummy); |
83beee94 | 856 | assert(!unique_ret); |
bb7b2f26 | 857 | node->p.next = flag_dummy(get_end()); |
83beee94 | 858 | return; /* Initial first add (head) */ |
18117871 | 859 | } |
f4a9cc0b | 860 | lookup = lookup_bucket(ht, size, bit_reverse_ulong(node->p.reverse_hash)); |
abc490a1 | 861 | for (;;) { |
adc0de68 | 862 | uint32_t chain_len = 0; |
abc490a1 | 863 | |
11519af6 MD |
864 | /* |
865 | * iter_prev points to the non-removed node prior to the | |
866 | * insert location. | |
11519af6 | 867 | */ |
14044b37 | 868 | iter_prev = (struct cds_lfht_node *) lookup; |
11519af6 | 869 | /* We can always skip the dummy node initially */ |
cc4fcb10 MD |
870 | iter = rcu_dereference(iter_prev->p.next); |
871 | assert(iter_prev->p.reverse_hash <= node->p.reverse_hash); | |
abc490a1 | 872 | for (;;) { |
bb7b2f26 | 873 | if (unlikely(is_end(iter))) |
273399de | 874 | goto insert; |
76412f24 | 875 | if (likely(clear_flag(iter)->p.reverse_hash > node->p.reverse_hash)) |
273399de | 876 | goto insert; |
194fdbd1 LJ |
877 | /* dummy node is the first node of the identical-hash-value chain */ |
878 | if (dummy && clear_flag(iter)->p.reverse_hash == node->p.reverse_hash) | |
879 | goto insert; | |
cc4fcb10 | 880 | next = rcu_dereference(clear_flag(iter)->p.next); |
b198f0fd | 881 | if (unlikely(is_removed(next))) |
9dba85be | 882 | goto gc_node; |
83beee94 | 883 | if (unique_ret |
1b81fe1a | 884 | && !is_dummy(next) |
ae450da3 | 885 | && clear_flag(iter)->p.reverse_hash == node->p.reverse_hash |
e43f23f8 MD |
886 | && !ht->compare_fct(node->key, node->key_len, |
887 | clear_flag(iter)->key, | |
48ed1c18 | 888 | clear_flag(iter)->key_len)) { |
83beee94 MD |
889 | unique_ret->node = clear_flag(iter); |
890 | unique_ret->next = next; | |
891 | return; | |
48ed1c18 | 892 | } |
11519af6 | 893 | /* Only account for identical reverse hash once */ |
24365af7 MD |
894 | if (iter_prev->p.reverse_hash != clear_flag(iter)->p.reverse_hash |
895 | && !is_dummy(next)) | |
4105056a | 896 | check_resize(ht, size, ++chain_len); |
11519af6 | 897 | iter_prev = clear_flag(iter); |
273399de | 898 | iter = next; |
abc490a1 | 899 | } |
48ed1c18 | 900 | |
273399de | 901 | insert: |
7ec59d3b | 902 | assert(node != clear_flag(iter)); |
11519af6 | 903 | assert(!is_removed(iter_prev)); |
c90201ac | 904 | assert(!is_removed(iter)); |
f000907d | 905 | assert(iter_prev != node); |
f9c80341 | 906 | if (!dummy) |
1b81fe1a | 907 | node->p.next = clear_flag(iter); |
f9c80341 MD |
908 | else |
909 | node->p.next = flag_dummy(clear_flag(iter)); | |
f5596c94 MD |
910 | if (is_dummy(iter)) |
911 | new_node = flag_dummy(node); | |
912 | else | |
913 | new_node = node; | |
cc4fcb10 | 914 | if (uatomic_cmpxchg(&iter_prev->p.next, iter, |
48ed1c18 | 915 | new_node) != iter) { |
273399de | 916 | continue; /* retry */ |
48ed1c18 | 917 | } else { |
83beee94 | 918 | return_node = node; |
960c9e4f | 919 | goto end; |
48ed1c18 MD |
920 | } |
921 | ||
9dba85be MD |
922 | gc_node: |
923 | assert(!is_removed(iter)); | |
f5596c94 MD |
924 | if (is_dummy(iter)) |
925 | new_next = flag_dummy(clear_flag(next)); | |
926 | else | |
927 | new_next = clear_flag(next); | |
928 | (void) uatomic_cmpxchg(&iter_prev->p.next, iter, new_next); | |
273399de | 929 | /* retry */ |
464a1ec9 | 930 | } |
9357c415 | 931 | end: |
83beee94 MD |
932 | if (unique_ret) { |
933 | unique_ret->node = return_node; | |
934 | /* unique_ret->next left unset, never used. */ | |
935 | } | |
abc490a1 | 936 | } |
464a1ec9 | 937 | |
abc490a1 | 938 | static |
860d07e8 | 939 | int _cds_lfht_del(struct cds_lfht *ht, unsigned long size, |
4105056a | 940 | struct cds_lfht_node *node, |
b198f0fd | 941 | int dummy_removal) |
abc490a1 | 942 | { |
14044b37 MD |
943 | struct cds_lfht_node *dummy, *next, *old; |
944 | struct _cds_lfht_node *lookup; | |
5e28c532 | 945 | |
9357c415 | 946 | if (!node) /* Return -ENOENT if asked to delete NULL node */ |
743f9143 | 947 | return -ENOENT; |
9357c415 | 948 | |
7ec59d3b | 949 | /* logically delete the node */ |
c90201ac MD |
950 | assert(!is_dummy(node)); |
951 | assert(!is_removed(node)); | |
cc4fcb10 | 952 | old = rcu_dereference(node->p.next); |
7ec59d3b | 953 | do { |
48ed1c18 MD |
954 | struct cds_lfht_node *new_next; |
955 | ||
7ec59d3b | 956 | next = old; |
76412f24 | 957 | if (unlikely(is_removed(next))) |
743f9143 | 958 | return -ENOENT; |
1475579c MD |
959 | if (dummy_removal) |
960 | assert(is_dummy(next)); | |
961 | else | |
962 | assert(!is_dummy(next)); | |
48ed1c18 | 963 | new_next = flag_removed(next); |
48ed1c18 | 964 | old = uatomic_cmpxchg(&node->p.next, next, new_next); |
7ec59d3b | 965 | } while (old != next); |
7ec59d3b | 966 | /* We performed the (logical) deletion. */ |
7ec59d3b MD |
967 | |
968 | /* | |
969 | * Ensure that the node is not visible to readers anymore: lookup for | |
273399de MD |
970 | * the node, and remove it (along with any other logically removed node) |
971 | * if found. | |
11519af6 | 972 | */ |
f4a9cc0b | 973 | lookup = lookup_bucket(ht, size, bit_reverse_ulong(node->p.reverse_hash)); |
14044b37 | 974 | dummy = (struct cds_lfht_node *) lookup; |
f9c80341 | 975 | _cds_lfht_gc_bucket(dummy, node); |
743f9143 LJ |
976 | |
977 | assert(is_removed(rcu_dereference(node->p.next))); | |
978 | return 0; | |
abc490a1 | 979 | } |
2ed95849 | 980 | |
b7d619b0 MD |
981 | static |
982 | void *partition_resize_thread(void *arg) | |
983 | { | |
984 | struct partition_resize_work *work = arg; | |
985 | ||
986 | work->ht->cds_lfht_rcu_register_thread(); | |
987 | work->fct(work->ht, work->i, work->start, work->len); | |
988 | work->ht->cds_lfht_rcu_unregister_thread(); | |
989 | return NULL; | |
990 | } | |
991 | ||
992 | static | |
993 | void partition_resize_helper(struct cds_lfht *ht, unsigned long i, | |
994 | unsigned long len, | |
995 | void (*fct)(struct cds_lfht *ht, unsigned long i, | |
996 | unsigned long start, unsigned long len)) | |
997 | { | |
998 | unsigned long partition_len; | |
999 | struct partition_resize_work *work; | |
6083a889 MD |
1000 | int thread, ret; |
1001 | unsigned long nr_threads; | |
b7d619b0 | 1002 | |
6083a889 MD |
1003 | /* |
1004 | * Note: nr_cpus_mask + 1 is always power of 2. | |
1005 | * We spawn just the number of threads we need to satisfy the minimum | |
1006 | * partition size, up to the number of CPUs in the system. | |
1007 | */ | |
91452a6a MD |
1008 | if (nr_cpus_mask > 0) { |
1009 | nr_threads = min(nr_cpus_mask + 1, | |
1010 | len >> MIN_PARTITION_PER_THREAD_ORDER); | |
1011 | } else { | |
1012 | nr_threads = 1; | |
1013 | } | |
6083a889 MD |
1014 | partition_len = len >> get_count_order_ulong(nr_threads); |
1015 | work = calloc(nr_threads, sizeof(*work)); | |
b7d619b0 | 1016 | assert(work); |
6083a889 MD |
1017 | for (thread = 0; thread < nr_threads; thread++) { |
1018 | work[thread].ht = ht; | |
1019 | work[thread].i = i; | |
1020 | work[thread].len = partition_len; | |
1021 | work[thread].start = thread * partition_len; | |
1022 | work[thread].fct = fct; | |
1af6e26e | 1023 | ret = pthread_create(&(work[thread].thread_id), ht->resize_attr, |
6083a889 | 1024 | partition_resize_thread, &work[thread]); |
b7d619b0 MD |
1025 | assert(!ret); |
1026 | } | |
6083a889 | 1027 | for (thread = 0; thread < nr_threads; thread++) { |
1af6e26e | 1028 | ret = pthread_join(work[thread].thread_id, NULL); |
b7d619b0 MD |
1029 | assert(!ret); |
1030 | } | |
1031 | free(work); | |
b7d619b0 MD |
1032 | } |
1033 | ||
e8de508e MD |
1034 | /* |
1035 | * Holding RCU read lock to protect _cds_lfht_add against memory | |
1036 | * reclaim that could be performed by other call_rcu worker threads (ABA | |
1037 | * problem). | |
9ee0fc9a | 1038 | * |
b7d619b0 | 1039 | * When we reach a certain length, we can split this population phase over |
9ee0fc9a MD |
1040 | * many worker threads, based on the number of CPUs available in the system. |
1041 | * This should therefore take care of not having the expand lagging behind too | |
1042 | * many concurrent insertion threads by using the scheduler's ability to | |
1043 | * schedule dummy node population fairly with insertions. | |
e8de508e | 1044 | */ |
4105056a | 1045 | static |
b7d619b0 MD |
1046 | void init_table_populate_partition(struct cds_lfht *ht, unsigned long i, |
1047 | unsigned long start, unsigned long len) | |
4105056a MD |
1048 | { |
1049 | unsigned long j; | |
1050 | ||
1051 | ht->cds_lfht_rcu_read_lock(); | |
b7d619b0 | 1052 | for (j = start; j < start + len; j++) { |
4105056a MD |
1053 | struct cds_lfht_node *new_node = |
1054 | (struct cds_lfht_node *) &ht->t.tbl[i]->nodes[j]; | |
1055 | ||
dc1da8f6 | 1056 | dbg_printf("init populate: i %lu j %lu hash %lu\n", |
4105056a | 1057 | i, j, !i ? 0 : (1UL << (i - 1)) + j); |
dc1da8f6 MD |
1058 | new_node->p.reverse_hash = |
1059 | bit_reverse_ulong(!i ? 0 : (1UL << (i - 1)) + j); | |
83beee94 MD |
1060 | _cds_lfht_add(ht, !i ? 0 : (1UL << (i - 1)), |
1061 | new_node, NULL, 1); | |
4105056a MD |
1062 | } |
1063 | ht->cds_lfht_rcu_read_unlock(); | |
b7d619b0 MD |
1064 | } |
1065 | ||
1066 | static | |
1067 | void init_table_populate(struct cds_lfht *ht, unsigned long i, | |
1068 | unsigned long len) | |
1069 | { | |
1070 | assert(nr_cpus_mask != -1); | |
6083a889 | 1071 | if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) { |
b7d619b0 MD |
1072 | ht->cds_lfht_rcu_thread_online(); |
1073 | init_table_populate_partition(ht, i, 0, len); | |
1074 | ht->cds_lfht_rcu_thread_offline(); | |
1075 | return; | |
1076 | } | |
1077 | partition_resize_helper(ht, i, len, init_table_populate_partition); | |
4105056a MD |
1078 | } |
1079 | ||
abc490a1 | 1080 | static |
4105056a | 1081 | void init_table(struct cds_lfht *ht, |
24365af7 MD |
1082 | unsigned long first_order, unsigned long len_order) |
1083 | { | |
1084 | unsigned long i, end_order; | |
1085 | ||
f0c29ed7 | 1086 | dbg_printf("init table: first_order %lu end_order %lu\n", |
24365af7 MD |
1087 | first_order, first_order + len_order); |
1088 | end_order = first_order + len_order; | |
24365af7 | 1089 | for (i = first_order; i < end_order; i++) { |
4105056a | 1090 | unsigned long len; |
24365af7 MD |
1091 | |
1092 | len = !i ? 1 : 1UL << (i - 1); | |
f0c29ed7 | 1093 | dbg_printf("init order %lu len: %lu\n", i, len); |
4d676753 MD |
1094 | |
1095 | /* Stop expand if the resize target changes under us */ | |
1096 | if (CMM_LOAD_SHARED(ht->t.resize_target) < (!i ? 1 : (1UL << i))) | |
1097 | break; | |
1098 | ||
0d14ceb2 | 1099 | ht->t.tbl[i] = calloc(1, len * sizeof(struct _cds_lfht_node)); |
b7d619b0 | 1100 | assert(ht->t.tbl[i]); |
4105056a | 1101 | |
4105056a | 1102 | /* |
dc1da8f6 MD |
1103 | * Set all dummy nodes reverse hash values for a level and |
1104 | * link all dummy nodes into the table. | |
4105056a | 1105 | */ |
dc1da8f6 | 1106 | init_table_populate(ht, i, len); |
4105056a | 1107 | |
f9c80341 MD |
1108 | /* |
1109 | * Update table size. | |
1110 | */ | |
1111 | cmm_smp_wmb(); /* populate data before RCU size */ | |
1112 | CMM_STORE_SHARED(ht->t.size, !i ? 1 : (1UL << i)); | |
1113 | ||
4105056a MD |
1114 | dbg_printf("init new size: %lu\n", !i ? 1 : (1UL << i)); |
1115 | if (CMM_LOAD_SHARED(ht->in_progress_destroy)) | |
1116 | break; | |
1117 | } | |
1118 | } | |
1119 | ||
e8de508e MD |
1120 | /* |
1121 | * Holding RCU read lock to protect _cds_lfht_remove against memory | |
1122 | * reclaim that could be performed by other call_rcu worker threads (ABA | |
1123 | * problem). | |
1124 | * For a single level, we logically remove and garbage collect each node. | |
1125 | * | |
1126 | * As a design choice, we perform logical removal and garbage collection on a | |
1127 | * node-per-node basis to simplify this algorithm. We also assume keeping good | |
1128 | * cache locality of the operation would overweight possible performance gain | |
1129 | * that could be achieved by batching garbage collection for multiple levels. | |
1130 | * However, this would have to be justified by benchmarks. | |
1131 | * | |
1132 | * Concurrent removal and add operations are helping us perform garbage | |
1133 | * collection of logically removed nodes. We guarantee that all logically | |
1134 | * removed nodes have been garbage-collected (unlinked) before call_rcu is | |
1135 | * invoked to free a hole level of dummy nodes (after a grace period). | |
1136 | * | |
1137 | * Logical removal and garbage collection can therefore be done in batch or on a | |
1138 | * node-per-node basis, as long as the guarantee above holds. | |
9ee0fc9a | 1139 | * |
b7d619b0 MD |
1140 | * When we reach a certain length, we can split this removal over many worker |
1141 | * threads, based on the number of CPUs available in the system. This should | |
1142 | * take care of not letting resize process lag behind too many concurrent | |
9ee0fc9a | 1143 | * updater threads actively inserting into the hash table. |
e8de508e | 1144 | */ |
4105056a | 1145 | static |
b7d619b0 MD |
1146 | void remove_table_partition(struct cds_lfht *ht, unsigned long i, |
1147 | unsigned long start, unsigned long len) | |
4105056a MD |
1148 | { |
1149 | unsigned long j; | |
1150 | ||
1151 | ht->cds_lfht_rcu_read_lock(); | |
b7d619b0 | 1152 | for (j = start; j < start + len; j++) { |
4105056a MD |
1153 | struct cds_lfht_node *fini_node = |
1154 | (struct cds_lfht_node *) &ht->t.tbl[i]->nodes[j]; | |
1155 | ||
1156 | dbg_printf("remove entry: i %lu j %lu hash %lu\n", | |
1157 | i, j, !i ? 0 : (1UL << (i - 1)) + j); | |
1158 | fini_node->p.reverse_hash = | |
1159 | bit_reverse_ulong(!i ? 0 : (1UL << (i - 1)) + j); | |
860d07e8 | 1160 | (void) _cds_lfht_del(ht, !i ? 0 : (1UL << (i - 1)), |
b198f0fd | 1161 | fini_node, 1); |
abc490a1 | 1162 | } |
4105056a | 1163 | ht->cds_lfht_rcu_read_unlock(); |
b7d619b0 MD |
1164 | } |
1165 | ||
1166 | static | |
1167 | void remove_table(struct cds_lfht *ht, unsigned long i, unsigned long len) | |
1168 | { | |
1169 | ||
1170 | assert(nr_cpus_mask != -1); | |
6083a889 | 1171 | if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) { |
b7d619b0 MD |
1172 | ht->cds_lfht_rcu_thread_online(); |
1173 | remove_table_partition(ht, i, 0, len); | |
1174 | ht->cds_lfht_rcu_thread_offline(); | |
1175 | return; | |
1176 | } | |
1177 | partition_resize_helper(ht, i, len, remove_table_partition); | |
2ed95849 MD |
1178 | } |
1179 | ||
1475579c | 1180 | static |
4105056a | 1181 | void fini_table(struct cds_lfht *ht, |
1475579c MD |
1182 | unsigned long first_order, unsigned long len_order) |
1183 | { | |
1184 | long i, end_order; | |
0d14ceb2 | 1185 | void *free_by_rcu = NULL; |
1475579c MD |
1186 | |
1187 | dbg_printf("fini table: first_order %lu end_order %lu\n", | |
1188 | first_order, first_order + len_order); | |
1189 | end_order = first_order + len_order; | |
1190 | assert(first_order > 0); | |
1475579c | 1191 | for (i = end_order - 1; i >= first_order; i--) { |
4105056a | 1192 | unsigned long len; |
1475579c MD |
1193 | |
1194 | len = !i ? 1 : 1UL << (i - 1); | |
1195 | dbg_printf("fini order %lu len: %lu\n", i, len); | |
4105056a | 1196 | |
4d676753 MD |
1197 | /* Stop shrink if the resize target changes under us */ |
1198 | if (CMM_LOAD_SHARED(ht->t.resize_target) > (1UL << (i - 1))) | |
1199 | break; | |
1200 | ||
1201 | cmm_smp_wmb(); /* populate data before RCU size */ | |
1202 | CMM_STORE_SHARED(ht->t.size, 1UL << (i - 1)); | |
1203 | ||
1204 | /* | |
1205 | * We need to wait for all add operations to reach Q.S. (and | |
1206 | * thus use the new table for lookups) before we can start | |
1207 | * releasing the old dummy nodes. Otherwise their lookup will | |
1208 | * return a logically removed node as insert position. | |
1209 | */ | |
1210 | ht->cds_lfht_synchronize_rcu(); | |
0d14ceb2 LJ |
1211 | if (free_by_rcu) |
1212 | free(free_by_rcu); | |
4d676753 | 1213 | |
21263e21 | 1214 | /* |
4105056a MD |
1215 | * Set "removed" flag in dummy nodes about to be removed. |
1216 | * Unlink all now-logically-removed dummy node pointers. | |
1217 | * Concurrent add/remove operation are helping us doing | |
1218 | * the gc. | |
21263e21 | 1219 | */ |
4105056a MD |
1220 | remove_table(ht, i, len); |
1221 | ||
0d14ceb2 | 1222 | free_by_rcu = ht->t.tbl[i]; |
4105056a MD |
1223 | |
1224 | dbg_printf("fini new size: %lu\n", 1UL << i); | |
1475579c MD |
1225 | if (CMM_LOAD_SHARED(ht->in_progress_destroy)) |
1226 | break; | |
1227 | } | |
0d14ceb2 LJ |
1228 | |
1229 | if (free_by_rcu) { | |
1230 | ht->cds_lfht_synchronize_rcu(); | |
1231 | free(free_by_rcu); | |
1232 | } | |
1475579c MD |
1233 | } |
1234 | ||
7a9dcf9b | 1235 | struct cds_lfht *_cds_lfht_new(cds_lfht_hash_fct hash_fct, |
14044b37 MD |
1236 | cds_lfht_compare_fct compare_fct, |
1237 | unsigned long hash_seed, | |
1238 | unsigned long init_size, | |
b8af5011 | 1239 | int flags, |
14044b37 | 1240 | void (*cds_lfht_call_rcu)(struct rcu_head *head, |
1475579c | 1241 | void (*func)(struct rcu_head *head)), |
01dbfa62 MD |
1242 | void (*cds_lfht_synchronize_rcu)(void), |
1243 | void (*cds_lfht_rcu_read_lock)(void), | |
5f511391 MD |
1244 | void (*cds_lfht_rcu_read_unlock)(void), |
1245 | void (*cds_lfht_rcu_thread_offline)(void), | |
b7d619b0 MD |
1246 | void (*cds_lfht_rcu_thread_online)(void), |
1247 | void (*cds_lfht_rcu_register_thread)(void), | |
1248 | void (*cds_lfht_rcu_unregister_thread)(void), | |
1249 | pthread_attr_t *attr) | |
abc490a1 | 1250 | { |
14044b37 | 1251 | struct cds_lfht *ht; |
24365af7 | 1252 | unsigned long order; |
abc490a1 | 1253 | |
8129be4e | 1254 | /* init_size must be power of two */ |
49619ea0 | 1255 | if (init_size && (init_size & (init_size - 1))) |
8129be4e | 1256 | return NULL; |
14044b37 | 1257 | ht = calloc(1, sizeof(struct cds_lfht)); |
b7d619b0 | 1258 | assert(ht); |
abc490a1 | 1259 | ht->hash_fct = hash_fct; |
732ad076 MD |
1260 | ht->compare_fct = compare_fct; |
1261 | ht->hash_seed = hash_seed; | |
14044b37 | 1262 | ht->cds_lfht_call_rcu = cds_lfht_call_rcu; |
1475579c | 1263 | ht->cds_lfht_synchronize_rcu = cds_lfht_synchronize_rcu; |
01dbfa62 MD |
1264 | ht->cds_lfht_rcu_read_lock = cds_lfht_rcu_read_lock; |
1265 | ht->cds_lfht_rcu_read_unlock = cds_lfht_rcu_read_unlock; | |
5f511391 MD |
1266 | ht->cds_lfht_rcu_thread_offline = cds_lfht_rcu_thread_offline; |
1267 | ht->cds_lfht_rcu_thread_online = cds_lfht_rcu_thread_online; | |
b7d619b0 MD |
1268 | ht->cds_lfht_rcu_register_thread = cds_lfht_rcu_register_thread; |
1269 | ht->cds_lfht_rcu_unregister_thread = cds_lfht_rcu_unregister_thread; | |
1270 | ht->resize_attr = attr; | |
df44348d | 1271 | ht->percpu_count = alloc_per_cpu_items_count(); |
abc490a1 MD |
1272 | /* this mutex should not nest in read-side C.S. */ |
1273 | pthread_mutex_init(&ht->resize_mutex, NULL); | |
cd95516d | 1274 | order = get_count_order_ulong(max(init_size, MIN_TABLE_SIZE)) + 1; |
b8af5011 | 1275 | ht->flags = flags; |
5f511391 | 1276 | ht->cds_lfht_rcu_thread_offline(); |
f000907d | 1277 | pthread_mutex_lock(&ht->resize_mutex); |
4d676753 | 1278 | ht->t.resize_target = 1UL << (order - 1); |
4105056a | 1279 | init_table(ht, 0, order); |
f000907d | 1280 | pthread_mutex_unlock(&ht->resize_mutex); |
5f511391 | 1281 | ht->cds_lfht_rcu_thread_online(); |
abc490a1 MD |
1282 | return ht; |
1283 | } | |
1284 | ||
adc0de68 MD |
1285 | void cds_lfht_lookup(struct cds_lfht *ht, void *key, size_t key_len, |
1286 | struct cds_lfht_iter *iter) | |
2ed95849 | 1287 | { |
bb7b2f26 | 1288 | struct cds_lfht_node *node, *next, *dummy_node; |
14044b37 | 1289 | struct _cds_lfht_node *lookup; |
f4a9cc0b | 1290 | unsigned long hash, reverse_hash, size; |
2ed95849 | 1291 | |
732ad076 | 1292 | hash = ht->hash_fct(key, key_len, ht->hash_seed); |
abc490a1 | 1293 | reverse_hash = bit_reverse_ulong(hash); |
464a1ec9 | 1294 | |
4105056a | 1295 | size = rcu_dereference(ht->t.size); |
f4a9cc0b | 1296 | lookup = lookup_bucket(ht, size, hash); |
bb7b2f26 MD |
1297 | dummy_node = (struct cds_lfht_node *) lookup; |
1298 | /* We can always skip the dummy node initially */ | |
1299 | node = rcu_dereference(dummy_node->p.next); | |
bb7b2f26 | 1300 | node = clear_flag(node); |
2ed95849 | 1301 | for (;;) { |
bb7b2f26 | 1302 | if (unlikely(is_end(node))) { |
96ad1112 | 1303 | node = next = NULL; |
abc490a1 | 1304 | break; |
bb7b2f26 | 1305 | } |
cc4fcb10 | 1306 | if (unlikely(node->p.reverse_hash > reverse_hash)) { |
96ad1112 | 1307 | node = next = NULL; |
abc490a1 | 1308 | break; |
2ed95849 | 1309 | } |
1b81fe1a | 1310 | next = rcu_dereference(node->p.next); |
adc0de68 | 1311 | if (likely(!is_removed(next)) |
1b81fe1a | 1312 | && !is_dummy(next) |
ae450da3 | 1313 | && clear_flag(node)->p.reverse_hash == reverse_hash |
49c2e2d6 | 1314 | && likely(!ht->compare_fct(node->key, node->key_len, key, key_len))) { |
273399de | 1315 | break; |
2ed95849 | 1316 | } |
1b81fe1a | 1317 | node = clear_flag(next); |
2ed95849 | 1318 | } |
1b81fe1a | 1319 | assert(!node || !is_dummy(rcu_dereference(node->p.next))); |
adc0de68 MD |
1320 | iter->node = node; |
1321 | iter->next = next; | |
abc490a1 | 1322 | } |
e0ba718a | 1323 | |
3883c0e5 | 1324 | void cds_lfht_next_duplicate(struct cds_lfht *ht, struct cds_lfht_iter *iter) |
a481e5ff | 1325 | { |
adc0de68 | 1326 | struct cds_lfht_node *node, *next; |
a481e5ff MD |
1327 | unsigned long reverse_hash; |
1328 | void *key; | |
1329 | size_t key_len; | |
1330 | ||
adc0de68 | 1331 | node = iter->node; |
a481e5ff MD |
1332 | reverse_hash = node->p.reverse_hash; |
1333 | key = node->key; | |
1334 | key_len = node->key_len; | |
adc0de68 | 1335 | next = iter->next; |
a481e5ff MD |
1336 | node = clear_flag(next); |
1337 | ||
1338 | for (;;) { | |
bb7b2f26 | 1339 | if (unlikely(is_end(node))) { |
96ad1112 | 1340 | node = next = NULL; |
a481e5ff | 1341 | break; |
bb7b2f26 | 1342 | } |
a481e5ff | 1343 | if (unlikely(node->p.reverse_hash > reverse_hash)) { |
96ad1112 | 1344 | node = next = NULL; |
a481e5ff MD |
1345 | break; |
1346 | } | |
1347 | next = rcu_dereference(node->p.next); | |
adc0de68 | 1348 | if (likely(!is_removed(next)) |
a481e5ff MD |
1349 | && !is_dummy(next) |
1350 | && likely(!ht->compare_fct(node->key, node->key_len, key, key_len))) { | |
1351 | break; | |
1352 | } | |
1353 | node = clear_flag(next); | |
1354 | } | |
1355 | assert(!node || !is_dummy(rcu_dereference(node->p.next))); | |
adc0de68 MD |
1356 | iter->node = node; |
1357 | iter->next = next; | |
a481e5ff MD |
1358 | } |
1359 | ||
4e9b9fbf MD |
1360 | void cds_lfht_next(struct cds_lfht *ht, struct cds_lfht_iter *iter) |
1361 | { | |
1362 | struct cds_lfht_node *node, *next; | |
1363 | ||
853395e1 | 1364 | node = clear_flag(iter->next); |
4e9b9fbf MD |
1365 | for (;;) { |
1366 | if (unlikely(is_end(node))) { | |
1367 | node = next = NULL; | |
1368 | break; | |
1369 | } | |
1370 | next = rcu_dereference(node->p.next); | |
1371 | if (likely(!is_removed(next)) | |
1372 | && !is_dummy(next)) { | |
1373 | break; | |
1374 | } | |
1375 | node = clear_flag(next); | |
1376 | } | |
1377 | assert(!node || !is_dummy(rcu_dereference(node->p.next))); | |
1378 | iter->node = node; | |
1379 | iter->next = next; | |
1380 | } | |
1381 | ||
1382 | void cds_lfht_first(struct cds_lfht *ht, struct cds_lfht_iter *iter) | |
1383 | { | |
1384 | struct _cds_lfht_node *lookup; | |
1385 | ||
1386 | /* | |
1387 | * Get next after first dummy node. The first dummy node is the | |
1388 | * first node of the linked list. | |
1389 | */ | |
1390 | lookup = &ht->t.tbl[0]->nodes[0]; | |
853395e1 | 1391 | iter->next = lookup->next; |
4e9b9fbf MD |
1392 | cds_lfht_next(ht, iter); |
1393 | } | |
1394 | ||
14044b37 | 1395 | void cds_lfht_add(struct cds_lfht *ht, struct cds_lfht_node *node) |
abc490a1 | 1396 | { |
4105056a | 1397 | unsigned long hash, size; |
ab7d5fc6 | 1398 | |
49c2e2d6 | 1399 | hash = ht->hash_fct(node->key, node->key_len, ht->hash_seed); |
cc4fcb10 | 1400 | node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash); |
2ed95849 | 1401 | |
4105056a | 1402 | size = rcu_dereference(ht->t.size); |
83beee94 | 1403 | _cds_lfht_add(ht, size, node, NULL, 0); |
4105056a | 1404 | ht_count_add(ht, size); |
3eca1b8c MD |
1405 | } |
1406 | ||
14044b37 | 1407 | struct cds_lfht_node *cds_lfht_add_unique(struct cds_lfht *ht, |
48ed1c18 | 1408 | struct cds_lfht_node *node) |
3eca1b8c | 1409 | { |
4105056a | 1410 | unsigned long hash, size; |
83beee94 | 1411 | struct cds_lfht_iter iter; |
3eca1b8c | 1412 | |
49c2e2d6 | 1413 | hash = ht->hash_fct(node->key, node->key_len, ht->hash_seed); |
cc4fcb10 | 1414 | node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash); |
3eca1b8c | 1415 | |
4105056a | 1416 | size = rcu_dereference(ht->t.size); |
83beee94 MD |
1417 | _cds_lfht_add(ht, size, node, &iter, 0); |
1418 | if (iter.node == node) | |
4105056a | 1419 | ht_count_add(ht, size); |
83beee94 | 1420 | return iter.node; |
2ed95849 MD |
1421 | } |
1422 | ||
9357c415 | 1423 | struct cds_lfht_node *cds_lfht_add_replace(struct cds_lfht *ht, |
48ed1c18 MD |
1424 | struct cds_lfht_node *node) |
1425 | { | |
1426 | unsigned long hash, size; | |
83beee94 | 1427 | struct cds_lfht_iter iter; |
48ed1c18 MD |
1428 | |
1429 | hash = ht->hash_fct(node->key, node->key_len, ht->hash_seed); | |
1430 | node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash); | |
1431 | ||
1432 | size = rcu_dereference(ht->t.size); | |
83beee94 MD |
1433 | for (;;) { |
1434 | _cds_lfht_add(ht, size, node, &iter, 0); | |
1435 | if (iter.node == node) { | |
1436 | ht_count_add(ht, size); | |
1437 | return NULL; | |
1438 | } | |
1439 | ||
1440 | if (!_cds_lfht_replace(ht, size, iter.node, iter.next, node)) | |
1441 | return iter.node; | |
1442 | } | |
48ed1c18 MD |
1443 | } |
1444 | ||
9357c415 MD |
1445 | int cds_lfht_replace(struct cds_lfht *ht, struct cds_lfht_iter *old_iter, |
1446 | struct cds_lfht_node *new_node) | |
1447 | { | |
1448 | unsigned long size; | |
1449 | ||
1450 | size = rcu_dereference(ht->t.size); | |
1451 | return _cds_lfht_replace(ht, size, old_iter->node, old_iter->next, | |
1452 | new_node); | |
1453 | } | |
1454 | ||
1455 | int cds_lfht_del(struct cds_lfht *ht, struct cds_lfht_iter *iter) | |
2ed95849 | 1456 | { |
4105056a | 1457 | unsigned long size; |
df44348d | 1458 | int ret; |
abc490a1 | 1459 | |
4105056a | 1460 | size = rcu_dereference(ht->t.size); |
9357c415 | 1461 | ret = _cds_lfht_del(ht, size, iter->node, 0); |
df44348d | 1462 | if (!ret) |
860d07e8 | 1463 | ht_count_del(ht, size); |
df44348d | 1464 | return ret; |
2ed95849 | 1465 | } |
ab7d5fc6 | 1466 | |
abc490a1 | 1467 | static |
14044b37 | 1468 | int cds_lfht_delete_dummy(struct cds_lfht *ht) |
674f7a69 | 1469 | { |
14044b37 MD |
1470 | struct cds_lfht_node *node; |
1471 | struct _cds_lfht_node *lookup; | |
4105056a | 1472 | unsigned long order, i, size; |
674f7a69 | 1473 | |
abc490a1 | 1474 | /* Check that the table is empty */ |
4105056a | 1475 | lookup = &ht->t.tbl[0]->nodes[0]; |
14044b37 | 1476 | node = (struct cds_lfht_node *) lookup; |
abc490a1 | 1477 | do { |
1b81fe1a MD |
1478 | node = clear_flag(node)->p.next; |
1479 | if (!is_dummy(node)) | |
abc490a1 | 1480 | return -EPERM; |
273399de | 1481 | assert(!is_removed(node)); |
bb7b2f26 | 1482 | } while (!is_end(node)); |
4105056a MD |
1483 | /* |
1484 | * size accessed without rcu_dereference because hash table is | |
1485 | * being destroyed. | |
1486 | */ | |
1487 | size = ht->t.size; | |
abc490a1 | 1488 | /* Internal sanity check: all nodes left should be dummy */ |
4105056a | 1489 | for (order = 0; order < get_count_order_ulong(size) + 1; order++) { |
24365af7 MD |
1490 | unsigned long len; |
1491 | ||
1492 | len = !order ? 1 : 1UL << (order - 1); | |
1493 | for (i = 0; i < len; i++) { | |
f0c29ed7 | 1494 | dbg_printf("delete order %lu i %lu hash %lu\n", |
24365af7 | 1495 | order, i, |
4105056a MD |
1496 | bit_reverse_ulong(ht->t.tbl[order]->nodes[i].reverse_hash)); |
1497 | assert(is_dummy(ht->t.tbl[order]->nodes[i].next)); | |
24365af7 | 1498 | } |
4105056a | 1499 | poison_free(ht->t.tbl[order]); |
674f7a69 | 1500 | } |
abc490a1 | 1501 | return 0; |
674f7a69 MD |
1502 | } |
1503 | ||
1504 | /* | |
1505 | * Should only be called when no more concurrent readers nor writers can | |
1506 | * possibly access the table. | |
1507 | */ | |
b7d619b0 | 1508 | int cds_lfht_destroy(struct cds_lfht *ht, pthread_attr_t **attr) |
674f7a69 | 1509 | { |
5e28c532 MD |
1510 | int ret; |
1511 | ||
848d4088 | 1512 | /* Wait for in-flight resize operations to complete */ |
24953e08 MD |
1513 | _CMM_STORE_SHARED(ht->in_progress_destroy, 1); |
1514 | cmm_smp_mb(); /* Store destroy before load resize */ | |
848d4088 MD |
1515 | while (uatomic_read(&ht->in_progress_resize)) |
1516 | poll(NULL, 0, 100); /* wait for 100ms */ | |
14044b37 | 1517 | ret = cds_lfht_delete_dummy(ht); |
abc490a1 MD |
1518 | if (ret) |
1519 | return ret; | |
df44348d | 1520 | free_per_cpu_items_count(ht->percpu_count); |
b7d619b0 MD |
1521 | if (attr) |
1522 | *attr = ht->resize_attr; | |
98808fb1 | 1523 | poison_free(ht); |
5e28c532 | 1524 | return ret; |
674f7a69 MD |
1525 | } |
1526 | ||
14044b37 | 1527 | void cds_lfht_count_nodes(struct cds_lfht *ht, |
d933dd0e | 1528 | long *approx_before, |
273399de | 1529 | unsigned long *count, |
973e5e1b | 1530 | unsigned long *removed, |
d933dd0e | 1531 | long *approx_after) |
273399de | 1532 | { |
14044b37 MD |
1533 | struct cds_lfht_node *node, *next; |
1534 | struct _cds_lfht_node *lookup; | |
24365af7 | 1535 | unsigned long nr_dummy = 0; |
273399de | 1536 | |
7ed7682f | 1537 | *approx_before = 0; |
973e5e1b MD |
1538 | if (nr_cpus_mask >= 0) { |
1539 | int i; | |
1540 | ||
1541 | for (i = 0; i < nr_cpus_mask + 1; i++) { | |
1542 | *approx_before += uatomic_read(&ht->percpu_count[i].add); | |
1543 | *approx_before -= uatomic_read(&ht->percpu_count[i].del); | |
1544 | } | |
1545 | } | |
1546 | ||
273399de MD |
1547 | *count = 0; |
1548 | *removed = 0; | |
1549 | ||
24365af7 | 1550 | /* Count non-dummy nodes in the table */ |
4105056a | 1551 | lookup = &ht->t.tbl[0]->nodes[0]; |
14044b37 | 1552 | node = (struct cds_lfht_node *) lookup; |
273399de | 1553 | do { |
cc4fcb10 | 1554 | next = rcu_dereference(node->p.next); |
b198f0fd | 1555 | if (is_removed(next)) { |
973e5e1b MD |
1556 | if (!is_dummy(next)) |
1557 | (*removed)++; | |
1558 | else | |
1559 | (nr_dummy)++; | |
1b81fe1a | 1560 | } else if (!is_dummy(next)) |
273399de | 1561 | (*count)++; |
24365af7 MD |
1562 | else |
1563 | (nr_dummy)++; | |
273399de | 1564 | node = clear_flag(next); |
bb7b2f26 | 1565 | } while (!is_end(node)); |
f0c29ed7 | 1566 | dbg_printf("number of dummy nodes: %lu\n", nr_dummy); |
7ed7682f | 1567 | *approx_after = 0; |
973e5e1b MD |
1568 | if (nr_cpus_mask >= 0) { |
1569 | int i; | |
1570 | ||
1571 | for (i = 0; i < nr_cpus_mask + 1; i++) { | |
1572 | *approx_after += uatomic_read(&ht->percpu_count[i].add); | |
1573 | *approx_after -= uatomic_read(&ht->percpu_count[i].del); | |
1574 | } | |
1575 | } | |
273399de MD |
1576 | } |
1577 | ||
1475579c | 1578 | /* called with resize mutex held */ |
abc490a1 | 1579 | static |
4105056a | 1580 | void _do_cds_lfht_grow(struct cds_lfht *ht, |
1475579c | 1581 | unsigned long old_size, unsigned long new_size) |
abc490a1 | 1582 | { |
1475579c | 1583 | unsigned long old_order, new_order; |
1475579c MD |
1584 | |
1585 | old_order = get_count_order_ulong(old_size) + 1; | |
1586 | new_order = get_count_order_ulong(new_size) + 1; | |
1a401918 LJ |
1587 | dbg_printf("resize from %lu (order %lu) to %lu (order %lu) buckets\n", |
1588 | old_size, old_order, new_size, new_order); | |
1475579c | 1589 | assert(new_size > old_size); |
4105056a | 1590 | init_table(ht, old_order, new_order - old_order); |
abc490a1 MD |
1591 | } |
1592 | ||
1593 | /* called with resize mutex held */ | |
1594 | static | |
4105056a | 1595 | void _do_cds_lfht_shrink(struct cds_lfht *ht, |
1475579c | 1596 | unsigned long old_size, unsigned long new_size) |
464a1ec9 | 1597 | { |
1475579c | 1598 | unsigned long old_order, new_order; |
464a1ec9 | 1599 | |
cd95516d | 1600 | new_size = max(new_size, MIN_TABLE_SIZE); |
24365af7 | 1601 | old_order = get_count_order_ulong(old_size) + 1; |
24365af7 | 1602 | new_order = get_count_order_ulong(new_size) + 1; |
1a401918 LJ |
1603 | dbg_printf("resize from %lu (order %lu) to %lu (order %lu) buckets\n", |
1604 | old_size, old_order, new_size, new_order); | |
1475579c | 1605 | assert(new_size < old_size); |
1475579c | 1606 | |
4105056a MD |
1607 | /* Remove and unlink all dummy nodes to remove. */ |
1608 | fini_table(ht, new_order, old_order - new_order); | |
464a1ec9 MD |
1609 | } |
1610 | ||
1475579c MD |
1611 | |
1612 | /* called with resize mutex held */ | |
1613 | static | |
1614 | void _do_cds_lfht_resize(struct cds_lfht *ht) | |
1615 | { | |
1616 | unsigned long new_size, old_size; | |
4105056a MD |
1617 | |
1618 | /* | |
1619 | * Resize table, re-do if the target size has changed under us. | |
1620 | */ | |
1621 | do { | |
d2be3620 MD |
1622 | assert(uatomic_read(&ht->in_progress_resize)); |
1623 | if (CMM_LOAD_SHARED(ht->in_progress_destroy)) | |
1624 | break; | |
4105056a MD |
1625 | ht->t.resize_initiated = 1; |
1626 | old_size = ht->t.size; | |
1627 | new_size = CMM_LOAD_SHARED(ht->t.resize_target); | |
1628 | if (old_size < new_size) | |
1629 | _do_cds_lfht_grow(ht, old_size, new_size); | |
1630 | else if (old_size > new_size) | |
1631 | _do_cds_lfht_shrink(ht, old_size, new_size); | |
1632 | ht->t.resize_initiated = 0; | |
1633 | /* write resize_initiated before read resize_target */ | |
1634 | cmm_smp_mb(); | |
4d676753 | 1635 | } while (ht->t.size != CMM_LOAD_SHARED(ht->t.resize_target)); |
1475579c MD |
1636 | } |
1637 | ||
abc490a1 | 1638 | static |
4105056a | 1639 | unsigned long resize_target_update(struct cds_lfht *ht, unsigned long size, |
f9830efd | 1640 | int growth_order) |
464a1ec9 | 1641 | { |
4105056a MD |
1642 | return _uatomic_max(&ht->t.resize_target, |
1643 | size << growth_order); | |
464a1ec9 MD |
1644 | } |
1645 | ||
1475579c | 1646 | static |
4105056a | 1647 | void resize_target_update_count(struct cds_lfht *ht, |
b8af5011 | 1648 | unsigned long count) |
1475579c | 1649 | { |
cd95516d | 1650 | count = max(count, MIN_TABLE_SIZE); |
4105056a | 1651 | uatomic_set(&ht->t.resize_target, count); |
1475579c MD |
1652 | } |
1653 | ||
1654 | void cds_lfht_resize(struct cds_lfht *ht, unsigned long new_size) | |
464a1ec9 | 1655 | { |
4105056a MD |
1656 | resize_target_update_count(ht, new_size); |
1657 | CMM_STORE_SHARED(ht->t.resize_initiated, 1); | |
5f511391 | 1658 | ht->cds_lfht_rcu_thread_offline(); |
1475579c MD |
1659 | pthread_mutex_lock(&ht->resize_mutex); |
1660 | _do_cds_lfht_resize(ht); | |
1661 | pthread_mutex_unlock(&ht->resize_mutex); | |
5f511391 | 1662 | ht->cds_lfht_rcu_thread_online(); |
abc490a1 | 1663 | } |
464a1ec9 | 1664 | |
abc490a1 MD |
1665 | static |
1666 | void do_resize_cb(struct rcu_head *head) | |
1667 | { | |
1668 | struct rcu_resize_work *work = | |
1669 | caa_container_of(head, struct rcu_resize_work, head); | |
14044b37 | 1670 | struct cds_lfht *ht = work->ht; |
abc490a1 | 1671 | |
5f511391 | 1672 | ht->cds_lfht_rcu_thread_offline(); |
abc490a1 | 1673 | pthread_mutex_lock(&ht->resize_mutex); |
14044b37 | 1674 | _do_cds_lfht_resize(ht); |
abc490a1 | 1675 | pthread_mutex_unlock(&ht->resize_mutex); |
5f511391 | 1676 | ht->cds_lfht_rcu_thread_online(); |
98808fb1 | 1677 | poison_free(work); |
848d4088 MD |
1678 | cmm_smp_mb(); /* finish resize before decrement */ |
1679 | uatomic_dec(&ht->in_progress_resize); | |
464a1ec9 MD |
1680 | } |
1681 | ||
abc490a1 | 1682 | static |
4105056a | 1683 | void cds_lfht_resize_lazy(struct cds_lfht *ht, unsigned long size, int growth) |
ab7d5fc6 | 1684 | { |
abc490a1 | 1685 | struct rcu_resize_work *work; |
f9830efd | 1686 | unsigned long target_size; |
abc490a1 | 1687 | |
4105056a MD |
1688 | target_size = resize_target_update(ht, size, growth); |
1689 | /* Store resize_target before read resize_initiated */ | |
1690 | cmm_smp_mb(); | |
1691 | if (!CMM_LOAD_SHARED(ht->t.resize_initiated) && size < target_size) { | |
848d4088 | 1692 | uatomic_inc(&ht->in_progress_resize); |
59290e9d | 1693 | cmm_smp_mb(); /* increment resize count before load destroy */ |
ed35e6d8 MD |
1694 | if (CMM_LOAD_SHARED(ht->in_progress_destroy)) { |
1695 | uatomic_dec(&ht->in_progress_resize); | |
59290e9d | 1696 | return; |
ed35e6d8 | 1697 | } |
f9830efd MD |
1698 | work = malloc(sizeof(*work)); |
1699 | work->ht = ht; | |
14044b37 | 1700 | ht->cds_lfht_call_rcu(&work->head, do_resize_cb); |
4105056a | 1701 | CMM_STORE_SHARED(ht->t.resize_initiated, 1); |
f9830efd | 1702 | } |
ab7d5fc6 | 1703 | } |
3171717f | 1704 | |
f8994aee MD |
1705 | #if defined(HAVE_SCHED_GETCPU) && defined(HAVE_SYSCONF) |
1706 | ||
3171717f | 1707 | static |
4105056a | 1708 | void cds_lfht_resize_lazy_count(struct cds_lfht *ht, unsigned long size, |
3171717f MD |
1709 | unsigned long count) |
1710 | { | |
1711 | struct rcu_resize_work *work; | |
3171717f | 1712 | |
b8af5011 MD |
1713 | if (!(ht->flags & CDS_LFHT_AUTO_RESIZE)) |
1714 | return; | |
4105056a MD |
1715 | resize_target_update_count(ht, count); |
1716 | /* Store resize_target before read resize_initiated */ | |
1717 | cmm_smp_mb(); | |
1718 | if (!CMM_LOAD_SHARED(ht->t.resize_initiated)) { | |
3171717f | 1719 | uatomic_inc(&ht->in_progress_resize); |
59290e9d | 1720 | cmm_smp_mb(); /* increment resize count before load destroy */ |
ed35e6d8 MD |
1721 | if (CMM_LOAD_SHARED(ht->in_progress_destroy)) { |
1722 | uatomic_dec(&ht->in_progress_resize); | |
59290e9d | 1723 | return; |
ed35e6d8 | 1724 | } |
3171717f MD |
1725 | work = malloc(sizeof(*work)); |
1726 | work->ht = ht; | |
1727 | ht->cds_lfht_call_rcu(&work->head, do_resize_cb); | |
4105056a | 1728 | CMM_STORE_SHARED(ht->t.resize_initiated, 1); |
3171717f MD |
1729 | } |
1730 | } | |
f8994aee MD |
1731 | |
1732 | #endif |