summaryrefslogtreecommitdiff
path: root/net/netfilter/nft_set_pipapo.c
blob: 15a236bebb46a06523e2473d0acd7fab08135acb (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
// SPDX-License-Identifier: GPL-2.0-only

/* PIPAPO: PIle PAcket POlicies: set for arbitrary concatenations of ranges
 *
 * Copyright (c) 2019-2020 Red Hat GmbH
 *
 * Author: Stefano Brivio <sbrivio@redhat.com>
 */

/**
 * DOC: Theory of Operation
 *
 *
 * Problem
 * -------
 *
 * Match packet bytes against entries composed of ranged or non-ranged packet
 * field specifiers, mapping them to arbitrary references. For example:
 *
 * ::
 *
 *               --- fields --->
 *      |    [net],[port],[net]... => [reference]
 *   entries [net],[port],[net]... => [reference]
 *      |    [net],[port],[net]... => [reference]
 *      V    ...
 *
 * where [net] fields can be IP ranges or netmasks, and [port] fields are port
 * ranges. Arbitrary packet fields can be matched.
 *
 *
 * Algorithm Overview
 * ------------------
 *
 * This algorithm is loosely inspired by [Ligatti 2010], and fundamentally
 * relies on the consideration that every contiguous range in a space of b bits
 * can be converted into b * 2 netmasks, from Theorem 3 in [Rottenstreich 2010],
 * as also illustrated in Section 9 of [Kogan 2014].
 *
 * Classification against a number of entries, that require matching given bits
 * of a packet field, is performed by grouping those bits in sets of arbitrary
 * size, and classifying packet bits one group at a time.
 *
 * Example:
 *   to match the source port (16 bits) of a packet, we can divide those 16 bits
 *   in 4 groups of 4 bits each. Given the entry:
 *      0000 0001 0101 1001
 *   and a packet with source port:
 *      0000 0001 1010 1001
 *   first and second groups match, but the third doesn't. We conclude that the
 *   packet doesn't match the given entry.
 *
 * Translate the set to a sequence of lookup tables, one per field. Each table
 * has two dimensions: bit groups to be matched for a single packet field, and
 * all the possible values of said groups (buckets). Input entries are
 * represented as one or more rules, depending on the number of composing
 * netmasks for the given field specifier, and a group match is indicated as a
 * set bit, with number corresponding to the rule index, in all the buckets
 * whose value matches the entry for a given group.
 *
 * Rules are mapped between fields through an array of x, n pairs, with each
 * item mapping a matched rule to one or more rules. The position of the pair in
 * the array indicates the matched rule to be mapped to the next field, x
 * indicates the first rule index in the next field, and n the amount of
 * next-field rules the current rule maps to.
 *
 * The mapping array for the last field maps to the desired references.
 *
 * To match, we perform table lookups using the values of grouped packet bits,
 * and use a sequence of bitwise operations to progressively evaluate rule
 * matching.
 *
 * A stand-alone, reference implementation, also including notes about possible
 * future optimisations, is available at:
 *    https://pipapo.lameexcu.se/
 *
 * Insertion
 * ---------
 *
 * - For each packet field:
 *
 *   - divide the b packet bits we want to classify into groups of size t,
 *     obtaining ceil(b / t) groups
 *
 *      Example: match on destination IP address, with t = 4: 32 bits, 8 groups
 *      of 4 bits each
 *
 *   - allocate a lookup table with one column ("bucket") for each possible
 *     value of a group, and with one row for each group
 *
 *      Example: 8 groups, 2^4 buckets:
 *
 * ::
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0
 *        1
 *        2
 *        3
 *        4
 *        5
 *        6
 *        7
 *
 *   - map the bits we want to classify for the current field, for a given
 *     entry, to a single rule for non-ranged and netmask set items, and to one
 *     or multiple rules for ranges. Ranges are expanded to composing netmasks
 *     by pipapo_expand().
 *
 *      Example: 2 entries, 10.0.0.5:1024 and 192.168.1.0-192.168.2.1:2048
 *      - rule #0: 10.0.0.5
 *      - rule #1: 192.168.1.0/24
 *      - rule #2: 192.168.2.0/31
 *
 *   - insert references to the rules in the lookup table, selecting buckets
 *     according to bit values of a rule in the given group. This is done by
 *     pipapo_insert().
 *
 *      Example: given:
 *      - rule #0: 10.0.0.5 mapping to buckets
 *        < 0 10  0 0   0 0  0 5 >
 *      - rule #1: 192.168.1.0/24 mapping to buckets
 *        < 12 0  10 8  0 1  < 0..15 > < 0..15 > >
 *      - rule #2: 192.168.2.0/31 mapping to buckets
 *        < 12 0  10 8  0 2  0 < 0..1 > >
 *
 *      these bits are set in the lookup table:
 *
 * ::
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0    0                                              1,2
 *        1   1,2                                      0
 *        2    0                                      1,2
 *        3    0                              1,2
 *        4  0,1,2
 *        5    0   1   2
 *        6  0,1,2 1   1   1   1   1   1   1   1   1   1   1   1   1   1   1
 *        7   1,2 1,2  1   1   1  0,1  1   1   1   1   1   1   1   1   1   1
 *
 *   - if this is not the last field in the set, fill a mapping array that maps
 *     rules from the lookup table to rules belonging to the same entry in
 *     the next lookup table, done by pipapo_map().
 *
 *     Note that as rules map to contiguous ranges of rules, given how netmask
 *     expansion and insertion is performed, &union nft_pipapo_map_bucket stores
 *     this information as pairs of first rule index, rule count.
 *
 *      Example: 2 entries, 10.0.0.5:1024 and 192.168.1.0-192.168.2.1:2048,
 *      given lookup table #0 for field 0 (see example above):
 *
 * ::
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0    0                                              1,2
 *        1   1,2                                      0
 *        2    0                                      1,2
 *        3    0                              1,2
 *        4  0,1,2
 *        5    0   1   2
 *        6  0,1,2 1   1   1   1   1   1   1   1   1   1   1   1   1   1   1
 *        7   1,2 1,2  1   1   1  0,1  1   1   1   1   1   1   1   1   1   1
 *
 *      and lookup table #1 for field 1 with:
 *      - rule #0: 1024 mapping to buckets
 *        < 0  0  4  0 >
 *      - rule #1: 2048 mapping to buckets
 *        < 0  0  5  0 >
 *
 * ::
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0   0,1
 *        1   0,1
 *        2                    0   1
 *        3   0,1
 *
 *      we need to map rules for 10.0.0.5 in lookup table #0 (rule #0) to 1024
 *      in lookup table #1 (rule #0) and rules for 192.168.1.0-192.168.2.1
 *      (rules #1, #2) to 2048 in lookup table #2 (rule #1):
 *
 * ::
 *
 *       rule indices in current field: 0    1    2
 *       map to rules in next field:    0    1    1
 *
 *   - if this is the last field in the set, fill a mapping array that maps
 *     rules from the last lookup table to element pointers, also done by
 *     pipapo_map().
 *
 *     Note that, in this implementation, we have two elements (start, end) for
 *     each entry. The pointer to the end element is stored in this array, and
 *     the pointer to the start element is linked from it.
 *
 *      Example: entry 10.0.0.5:1024 has a corresponding &struct nft_pipapo_elem
 *      pointer, 0x66, and element for 192.168.1.0-192.168.2.1:2048 is at 0x42.
 *      From the rules of lookup table #1 as mapped above:
 *
 * ::
 *
 *       rule indices in last field:    0    1
 *       map to elements:             0x66  0x42
 *
 *
 * Matching
 * --------
 *
 * We use a result bitmap, with the size of a single lookup table bucket, to
 * represent the matching state that applies at every algorithm step. This is
 * done by pipapo_lookup().
 *
 * - For each packet field:
 *
 *   - start with an all-ones result bitmap (res_map in pipapo_lookup())
 *
 *   - perform a lookup into the table corresponding to the current field,
 *     for each group, and at every group, AND the current result bitmap with
 *     the value from the lookup table bucket
 *
 * ::
 *
 *      Example: 192.168.1.5 < 12 0  10 8  0 1  0 5 >, with lookup table from
 *      insertion examples.
 *      Lookup table buckets are at least 3 bits wide, we'll assume 8 bits for
 *      convenience in this example. Initial result bitmap is 0xff, the steps
 *      below show the value of the result bitmap after each group is processed:
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0    0                                              1,2
 *        result bitmap is now: 0xff & 0x6 [bucket 12] = 0x6
 *
 *        1   1,2                                      0
 *        result bitmap is now: 0x6 & 0x6 [bucket 0] = 0x6
 *
 *        2    0                                      1,2
 *        result bitmap is now: 0x6 & 0x6 [bucket 10] = 0x6
 *
 *        3    0                              1,2
 *        result bitmap is now: 0x6 & 0x6 [bucket 8] = 0x6
 *
 *        4  0,1,2
 *        result bitmap is now: 0x6 & 0x7 [bucket 0] = 0x6
 *
 *        5    0   1   2
 *        result bitmap is now: 0x6 & 0x2 [bucket 1] = 0x2
 *
 *        6  0,1,2 1   1   1   1   1   1   1   1   1   1   1   1   1   1   1
 *        result bitmap is now: 0x2 & 0x7 [bucket 0] = 0x2
 *
 *        7   1,2 1,2  1   1   1  0,1  1   1   1   1   1   1   1   1   1   1
 *        final result bitmap for this field is: 0x2 & 0x3 [bucket 5] = 0x2
 *
 *   - at the next field, start with a new, all-zeroes result bitmap. For each
 *     bit set in the previous result bitmap, fill the new result bitmap
 *     (fill_map in pipapo_lookup()) with the rule indices from the
 *     corresponding buckets of the mapping field for this field, done by
 *     pipapo_refill()
 *
 *      Example: with mapping table from insertion examples, with the current
 *      result bitmap from the previous example, 0x02:
 *
 * ::
 *
 *       rule indices in current field: 0    1    2
 *       map to rules in next field:    0    1    1
 *
 *      the new result bitmap will be 0x02: rule 1 was set, and rule 1 will be
 *      set.
 *
 *      We can now extend this example to cover the second iteration of the step
 *      above (lookup and AND bitmap): assuming the port field is
 *      2048 < 0  0  5  0 >, with starting result bitmap 0x2, and lookup table
 *      for "port" field from pre-computation example:
 *
 * ::
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0   0,1
 *        1   0,1
 *        2                    0   1
 *        3   0,1
 *
 *       operations are: 0x2 & 0x3 [bucket 0] & 0x3 [bucket 0] & 0x2 [bucket 5]
 *       & 0x3 [bucket 0], resulting bitmap is 0x2.
 *
 *   - if this is the last field in the set, look up the value from the mapping
 *     array corresponding to the final result bitmap
 *
 *      Example: 0x2 resulting bitmap from 192.168.1.5:2048, mapping array for
 *      last field from insertion example:
 *
 * ::
 *
 *       rule indices in last field:    0    1
 *       map to elements:             0x66  0x42
 *
 *      the matching element is at 0x42.
 *
 *
 * References
 * ----------
 *
 * [Ligatti 2010]
 *      A Packet-classification Algorithm for Arbitrary Bitmask Rules, with
 *      Automatic Time-space Tradeoffs
 *      Jay Ligatti, Josh Kuhn, and Chris Gage.
 *      Proceedings of the IEEE International Conference on Computer
 *      Communication Networks (ICCCN), August 2010.
 *      https://www.cse.usf.edu/~ligatti/papers/grouper-conf.pdf
 *
 * [Rottenstreich 2010]
 *      Worst-Case TCAM Rule Expansion
 *      Ori Rottenstreich and Isaac Keslassy.
 *      2010 Proceedings IEEE INFOCOM, San Diego, CA, 2010.
 *      http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.212.4592&rep=rep1&type=pdf
 *
 * [Kogan 2014]
 *      SAX-PAC (Scalable And eXpressive PAcket Classification)
 *      Kirill Kogan, Sergey Nikolenko, Ori Rottenstreich, William Culhane,
 *      and Patrick Eugster.
 *      Proceedings of the 2014 ACM conference on SIGCOMM, August 2014.
 *      https://www.sigcomm.org/sites/default/files/ccr/papers/2014/August/2619239-2626294.pdf
 */

#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/netlink.h>
#include <linux/netfilter.h>
#include <linux/netfilter/nf_tables.h>
#include <net/netfilter/nf_tables_core.h>
#include <uapi/linux/netfilter/nf_tables.h>
#include <linux/bitmap.h>
#include <linux/bitops.h>

#include "nft_set_pipapo_avx2.h"
#include "nft_set_pipapo.h"

/**
 * pipapo_refill() - For each set bit, set bits from selected mapping table item
 * @map:	Bitmap to be scanned for set bits
 * @len:	Length of bitmap in longs
 * @rules:	Number of rules in field
 * @dst:	Destination bitmap
 * @mt:		Mapping table containing bit set specifiers
 * @match_only:	Find a single bit and return, don't fill
 *
 * Iteration over set bits with __builtin_ctzl(): Daniel Lemire, public domain.
 *
 * For each bit set in map, select the bucket from mapping table with index
 * corresponding to the position of the bit set. Use start bit and amount of
 * bits specified in bucket to fill region in dst.
 *
 * Return: -1 on no match, bit position on 'match_only', 0 otherwise.
 */
int pipapo_refill(unsigned long *map, unsigned int len, unsigned int rules,
		  unsigned long *dst,
		  const union nft_pipapo_map_bucket *mt, bool match_only)
{
	unsigned long bitset;
	unsigned int k;
	int ret = -1;

	for (k = 0; k < len; k++) {
		bitset = map[k];
		while (bitset) {
			unsigned long t = bitset & -bitset;
			int r = __builtin_ctzl(bitset);
			int i = k * BITS_PER_LONG + r;

			if (unlikely(i >= rules)) {
				map[k] = 0;
				return -1;
			}

			if (match_only) {
				bitmap_clear(map, i, 1);
				return i;
			}

			ret = 0;

			bitmap_set(dst, mt[i].to, mt[i].n);

			bitset ^= t;
		}
		map[k] = 0;
	}

	return ret;
}

/**
 * nft_pipapo_lookup() - Lookup function
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @key:	nftables API element representation containing key data
 * @ext:	nftables API extension pointer, filled with matching reference
 *
 * For more details, see DOC: Theory of Operation.
 *
 * Return: true on match, false otherwise.
 */
bool nft_pipapo_lookup(const struct net *net, const struct nft_set *set,
		       const u32 *key, const struct nft_set_ext **ext)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct nft_pipapo_scratch *scratch;
	unsigned long *res_map, *fill_map;
	u8 genmask = nft_genmask_cur(net);
	const struct nft_pipapo_match *m;
	const struct nft_pipapo_field *f;
	const u8 *rp = (const u8 *)key;
	bool map_index;
	int i;

	local_bh_disable();

	m = rcu_dereference(priv->match);

	if (unlikely(!m || !*raw_cpu_ptr(m->scratch)))
		goto out;

	scratch = *raw_cpu_ptr(m->scratch);

	map_index = scratch->map_index;

	res_map  = scratch->map + (map_index ? m->bsize_max : 0);
	fill_map = scratch->map + (map_index ? 0 : m->bsize_max);

	memset(res_map, 0xff, m->bsize_max * sizeof(*res_map));

	nft_pipapo_for_each_field(f, i, m) {
		bool last = i == m->field_count - 1;
		int b;

		/* For each bit group: select lookup table bucket depending on
		 * packet bytes value, then AND bucket value
		 */
		if (likely(f->bb == 8))
			pipapo_and_field_buckets_8bit(f, res_map, rp);
		else
			pipapo_and_field_buckets_4bit(f, res_map, rp);
		NFT_PIPAPO_GROUP_BITS_ARE_8_OR_4;

		rp += f->groups / NFT_PIPAPO_GROUPS_PER_BYTE(f);

		/* Now populate the bitmap for the next field, unless this is
		 * the last field, in which case return the matched 'ext'
		 * pointer if any.
		 *
		 * Now res_map contains the matching bitmap, and fill_map is the
		 * bitmap for the next field.
		 */
next_match:
		b = pipapo_refill(res_map, f->bsize, f->rules, fill_map, f->mt,
				  last);
		if (b < 0) {
			scratch->map_index = map_index;
			local_bh_enable();

			return false;
		}

		if (last) {
			*ext = &f->mt[b].e->ext;
			if (unlikely(nft_set_elem_expired(*ext) ||
				     !nft_set_elem_active(*ext, genmask)))
				goto next_match;

			/* Last field: we're just returning the key without
			 * filling the initial bitmap for the next field, so the
			 * current inactive bitmap is clean and can be reused as
			 * *next* bitmap (not initial) for the next packet.
			 */
			scratch->map_index = map_index;
			local_bh_enable();

			return true;
		}

		/* Swap bitmap indices: res_map is the initial bitmap for the
		 * next field, and fill_map is guaranteed to be all-zeroes at
		 * this point.
		 */
		map_index = !map_index;
		swap(res_map, fill_map);

		rp += NFT_PIPAPO_GROUPS_PADDING(f);
	}

out:
	local_bh_enable();
	return false;
}

/**
 * pipapo_get() - Get matching element reference given key data
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @m:		storage containing active/existing elements
 * @data:	Key data to be matched against existing elements
 * @genmask:	If set, check that element is active in given genmask
 * @tstamp:	timestamp to check for expired elements
 * @gfp:	the type of memory to allocate (see kmalloc).
 *
 * This is essentially the same as the lookup function, except that it matches
 * key data against the uncommitted copy and doesn't use preallocated maps for
 * bitmap results.
 *
 * Return: pointer to &struct nft_pipapo_elem on match, error pointer otherwise.
 */
static struct nft_pipapo_elem *pipapo_get(const struct net *net,
					  const struct nft_set *set,
					  const struct nft_pipapo_match *m,
					  const u8 *data, u8 genmask,
					  u64 tstamp, gfp_t gfp)
{
	struct nft_pipapo_elem *ret = ERR_PTR(-ENOENT);
	unsigned long *res_map, *fill_map = NULL;
	const struct nft_pipapo_field *f;
	int i;

	if (m->bsize_max == 0)
		return ret;

	res_map = kmalloc_array(m->bsize_max, sizeof(*res_map), gfp);
	if (!res_map) {
		ret = ERR_PTR(-ENOMEM);
		goto out;
	}

	fill_map = kcalloc(m->bsize_max, sizeof(*res_map), gfp);
	if (!fill_map) {
		ret = ERR_PTR(-ENOMEM);
		goto out;
	}

	memset(res_map, 0xff, m->bsize_max * sizeof(*res_map));

	nft_pipapo_for_each_field(f, i, m) {
		bool last = i == m->field_count - 1;
		int b;

		/* For each bit group: select lookup table bucket depending on
		 * packet bytes value, then AND bucket value
		 */
		if (f->bb == 8)
			pipapo_and_field_buckets_8bit(f, res_map, data);
		else if (f->bb == 4)
			pipapo_and_field_buckets_4bit(f, res_map, data);
		else
			BUG();

		data += f->groups / NFT_PIPAPO_GROUPS_PER_BYTE(f);

		/* Now populate the bitmap for the next field, unless this is
		 * the last field, in which case return the matched 'ext'
		 * pointer if any.
		 *
		 * Now res_map contains the matching bitmap, and fill_map is the
		 * bitmap for the next field.
		 */
next_match:
		b = pipapo_refill(res_map, f->bsize, f->rules, fill_map, f->mt,
				  last);
		if (b < 0)
			goto out;

		if (last) {
			if (__nft_set_elem_expired(&f->mt[b].e->ext, tstamp))
				goto next_match;
			if ((genmask &&
			     !nft_set_elem_active(&f->mt[b].e->ext, genmask)))
				goto next_match;

			ret = f->mt[b].e;
			goto out;
		}

		data += NFT_PIPAPO_GROUPS_PADDING(f);

		/* Swap bitmap indices: fill_map will be the initial bitmap for
		 * the next field (i.e. the new res_map), and res_map is
		 * guaranteed to be all-zeroes at this point, ready to be filled
		 * according to the next mapping table.
		 */
		swap(res_map, fill_map);
	}

out:
	kfree(fill_map);
	kfree(res_map);
	return ret;
}

/**
 * nft_pipapo_get() - Get matching element reference given key data
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @elem:	nftables API element representation containing key data
 * @flags:	Unused
 */
static struct nft_elem_priv *
nft_pipapo_get(const struct net *net, const struct nft_set *set,
	       const struct nft_set_elem *elem, unsigned int flags)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct nft_pipapo_match *m = rcu_dereference(priv->match);
	struct nft_pipapo_elem *e;

	e = pipapo_get(net, set, m, (const u8 *)elem->key.val.data,
		       nft_genmask_cur(net), get_jiffies_64(),
		       GFP_ATOMIC);
	if (IS_ERR(e))
		return ERR_CAST(e);

	return &e->priv;
}

/**
 * pipapo_realloc_mt() - Reallocate mapping table if needed upon resize
 * @f:		Field containing mapping table
 * @old_rules:	Amount of existing mapped rules
 * @rules:	Amount of new rules to map
 *
 * Return: 0 on success, negative error code on failure.
 */
static int pipapo_realloc_mt(struct nft_pipapo_field *f,
			     unsigned int old_rules, unsigned int rules)
{
	union nft_pipapo_map_bucket *new_mt = NULL, *old_mt = f->mt;
	const unsigned int extra = PAGE_SIZE / sizeof(*new_mt);
	unsigned int rules_alloc = rules;

	might_sleep();

	if (unlikely(rules == 0))
		goto out_free;

	/* growing and enough space left, no action needed */
	if (rules > old_rules && f->rules_alloc > rules)
		return 0;

	/* downsize and extra slack has not grown too large */
	if (rules < old_rules) {
		unsigned int remove = f->rules_alloc - rules;

		if (remove < (2u * extra))
			return 0;
	}

	/* If set needs more than one page of memory for rules then
	 * allocate another extra page to avoid frequent reallocation.
	 */
	if (rules > extra &&
	    check_add_overflow(rules, extra, &rules_alloc))
		return -EOVERFLOW;

	new_mt = kvmalloc_array(rules_alloc, sizeof(*new_mt), GFP_KERNEL);
	if (!new_mt)
		return -ENOMEM;

	if (old_mt)
		memcpy(new_mt, old_mt, min(old_rules, rules) * sizeof(*new_mt));

	if (rules > old_rules) {
		memset(new_mt + old_rules, 0,
		       (rules - old_rules) * sizeof(*new_mt));
	}
out_free:
	f->rules_alloc = rules_alloc;
	f->mt = new_mt;

	kvfree(old_mt);

	return 0;
}

/**
 * pipapo_resize() - Resize lookup or mapping table, or both
 * @f:		Field containing lookup and mapping tables
 * @old_rules:	Previous amount of rules in field
 * @rules:	New amount of rules
 *
 * Increase, decrease or maintain tables size depending on new amount of rules,
 * and copy data over. In case the new size is smaller, throw away data for
 * highest-numbered rules.
 *
 * Return: 0 on success, -ENOMEM on allocation failure.
 */
static int pipapo_resize(struct nft_pipapo_field *f,
			 unsigned int old_rules, unsigned int rules)
{
	long *new_lt = NULL, *new_p, *old_lt = f->lt, *old_p;
	unsigned int new_bucket_size, copy;
	int group, bucket, err;

	if (rules >= NFT_PIPAPO_RULE0_MAX)
		return -ENOSPC;

	new_bucket_size = DIV_ROUND_UP(rules, BITS_PER_LONG);
#ifdef NFT_PIPAPO_ALIGN
	new_bucket_size = roundup(new_bucket_size,
				  NFT_PIPAPO_ALIGN / sizeof(*new_lt));
#endif

	if (new_bucket_size == f->bsize)
		goto mt;

	if (new_bucket_size > f->bsize)
		copy = f->bsize;
	else
		copy = new_bucket_size;

	new_lt = kvzalloc(f->groups * NFT_PIPAPO_BUCKETS(f->bb) *
			  new_bucket_size * sizeof(*new_lt) +
			  NFT_PIPAPO_ALIGN_HEADROOM,
			  GFP_KERNEL);
	if (!new_lt)
		return -ENOMEM;

	new_p = NFT_PIPAPO_LT_ALIGN(new_lt);
	old_p = NFT_PIPAPO_LT_ALIGN(old_lt);

	for (group = 0; group < f->groups; group++) {
		for (bucket = 0; bucket < NFT_PIPAPO_BUCKETS(f->bb); bucket++) {
			memcpy(new_p, old_p, copy * sizeof(*new_p));
			new_p += copy;
			old_p += copy;

			if (new_bucket_size > f->bsize)
				new_p += new_bucket_size - f->bsize;
			else
				old_p += f->bsize - new_bucket_size;
		}
	}

mt:
	err = pipapo_realloc_mt(f, old_rules, rules);
	if (err) {
		kvfree(new_lt);
		return err;
	}

	if (new_lt) {
		f->bsize = new_bucket_size;
		f->lt = new_lt;
		kvfree(old_lt);
	}

	return 0;
}

/**
 * pipapo_bucket_set() - Set rule bit in bucket given group and group value
 * @f:		Field containing lookup table
 * @rule:	Rule index
 * @group:	Group index
 * @v:		Value of bit group
 */
static void pipapo_bucket_set(struct nft_pipapo_field *f, int rule, int group,
			      int v)
{
	unsigned long *pos;

	pos = NFT_PIPAPO_LT_ALIGN(f->lt);
	pos += f->bsize * NFT_PIPAPO_BUCKETS(f->bb) * group;
	pos += f->bsize * v;

	__set_bit(rule, pos);
}

/**
 * pipapo_lt_4b_to_8b() - Switch lookup table group width from 4 bits to 8 bits
 * @old_groups:	Number of current groups
 * @bsize:	Size of one bucket, in longs
 * @old_lt:	Pointer to the current lookup table
 * @new_lt:	Pointer to the new, pre-allocated lookup table
 *
 * Each bucket with index b in the new lookup table, belonging to group g, is
 * filled with the bit intersection between:
 * - bucket with index given by the upper 4 bits of b, from group g, and
 * - bucket with index given by the lower 4 bits of b, from group g + 1
 *
 * That is, given buckets from the new lookup table N(x, y) and the old lookup
 * table O(x, y), with x bucket index, and y group index:
 *
 *	N(b, g) := O(b / 16, g) & O(b % 16, g + 1)
 *
 * This ensures equivalence of the matching results on lookup. Two examples in
 * pictures:
 *
 *              bucket
 *  group  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 ... 254 255
 *    0                ^
 *    1                |                                                 ^
 *   ...             ( & )                                               |
 *                  /     \                                              |
 *                 /       \                                         .-( & )-.
 *                /  bucket \                                        |       |
 *      group  0 / 1   2   3 \ 4   5   6   7   8   9  10  11  12  13 |14  15 |
 *        0     /             \                                      |       |
 *        1                    \                                     |       |
 *        2                                                          |     --'
 *        3                                                          '-
 *       ...
 */
static void pipapo_lt_4b_to_8b(int old_groups, int bsize,
			       unsigned long *old_lt, unsigned long *new_lt)
{
	int g, b, i;

	for (g = 0; g < old_groups / 2; g++) {
		int src_g0 = g * 2, src_g1 = g * 2 + 1;

		for (b = 0; b < NFT_PIPAPO_BUCKETS(8); b++) {
			int src_b0 = b / NFT_PIPAPO_BUCKETS(4);
			int src_b1 = b % NFT_PIPAPO_BUCKETS(4);
			int src_i0 = src_g0 * NFT_PIPAPO_BUCKETS(4) + src_b0;
			int src_i1 = src_g1 * NFT_PIPAPO_BUCKETS(4) + src_b1;

			for (i = 0; i < bsize; i++) {
				*new_lt = old_lt[src_i0 * bsize + i] &
					  old_lt[src_i1 * bsize + i];
				new_lt++;
			}
		}
	}
}

/**
 * pipapo_lt_8b_to_4b() - Switch lookup table group width from 8 bits to 4 bits
 * @old_groups:	Number of current groups
 * @bsize:	Size of one bucket, in longs
 * @old_lt:	Pointer to the current lookup table
 * @new_lt:	Pointer to the new, pre-allocated lookup table
 *
 * Each bucket with index b in the new lookup table, belonging to group g, is
 * filled with the bit union of:
 * - all the buckets with index such that the upper four bits of the lower byte
 *   equal b, from group g, with g odd
 * - all the buckets with index such that the lower four bits equal b, from
 *   group g, with g even
 *
 * That is, given buckets from the new lookup table N(x, y) and the old lookup
 * table O(x, y), with x bucket index, and y group index:
 *
 *	- with g odd:  N(b, g) := U(O(x, g) for each x : x = (b & 0xf0) >> 4)
 *	- with g even: N(b, g) := U(O(x, g) for each x : x = b & 0x0f)
 *
 * where U() denotes the arbitrary union operation (binary OR of n terms). This
 * ensures equivalence of the matching results on lookup.
 */
static void pipapo_lt_8b_to_4b(int old_groups, int bsize,
			       unsigned long *old_lt, unsigned long *new_lt)
{
	int g, b, bsrc, i;

	memset(new_lt, 0, old_groups * 2 * NFT_PIPAPO_BUCKETS(4) * bsize *
			  sizeof(unsigned long));

	for (g = 0; g < old_groups * 2; g += 2) {
		int src_g = g / 2;

		for (b = 0; b < NFT_PIPAPO_BUCKETS(4); b++) {
			for (bsrc = NFT_PIPAPO_BUCKETS(8) * src_g;
			     bsrc < NFT_PIPAPO_BUCKETS(8) * (src_g + 1);
			     bsrc++) {
				if (((bsrc & 0xf0) >> 4) != b)
					continue;

				for (i = 0; i < bsize; i++)
					new_lt[i] |= old_lt[bsrc * bsize + i];
			}

			new_lt += bsize;
		}

		for (b = 0; b < NFT_PIPAPO_BUCKETS(4); b++) {
			for (bsrc = NFT_PIPAPO_BUCKETS(8) * src_g;
			     bsrc < NFT_PIPAPO_BUCKETS(8) * (src_g + 1);
			     bsrc++) {
				if ((bsrc & 0x0f) != b)
					continue;

				for (i = 0; i < bsize; i++)
					new_lt[i] |= old_lt[bsrc * bsize + i];
			}

			new_lt += bsize;
		}
	}
}

/**
 * pipapo_lt_bits_adjust() - Adjust group size for lookup table if needed
 * @f:		Field containing lookup table
 */
static void pipapo_lt_bits_adjust(struct nft_pipapo_field *f)
{
	unsigned int groups, bb;
	unsigned long *new_lt;
	size_t lt_size;

	lt_size = f->groups * NFT_PIPAPO_BUCKETS(f->bb) * f->bsize *
		  sizeof(*f->lt);

	if (f->bb == NFT_PIPAPO_GROUP_BITS_SMALL_SET &&
	    lt_size > NFT_PIPAPO_LT_SIZE_HIGH) {
		groups = f->groups * 2;
		bb = NFT_PIPAPO_GROUP_BITS_LARGE_SET;

		lt_size = groups * NFT_PIPAPO_BUCKETS(bb) * f->bsize *
			  sizeof(*f->lt);
	} else if (f->bb == NFT_PIPAPO_GROUP_BITS_LARGE_SET &&
		   lt_size < NFT_PIPAPO_LT_SIZE_LOW) {
		groups = f->groups / 2;
		bb = NFT_PIPAPO_GROUP_BITS_SMALL_SET;

		lt_size = groups * NFT_PIPAPO_BUCKETS(bb) * f->bsize *
			  sizeof(*f->lt);

		/* Don't increase group width if the resulting lookup table size
		 * would exceed the upper size threshold for a "small" set.
		 */
		if (lt_size > NFT_PIPAPO_LT_SIZE_HIGH)
			return;
	} else {
		return;
	}

	new_lt = kvzalloc(lt_size + NFT_PIPAPO_ALIGN_HEADROOM, GFP_KERNEL);
	if (!new_lt)
		return;

	NFT_PIPAPO_GROUP_BITS_ARE_8_OR_4;
	if (f->bb == 4 && bb == 8) {
		pipapo_lt_4b_to_8b(f->groups, f->bsize,
				   NFT_PIPAPO_LT_ALIGN(f->lt),
				   NFT_PIPAPO_LT_ALIGN(new_lt));
	} else if (f->bb == 8 && bb == 4) {
		pipapo_lt_8b_to_4b(f->groups, f->bsize,
				   NFT_PIPAPO_LT_ALIGN(f->lt),
				   NFT_PIPAPO_LT_ALIGN(new_lt));
	} else {
		BUG();
	}

	f->groups = groups;
	f->bb = bb;
	kvfree(f->lt);
	f->lt = new_lt;
}

/**
 * pipapo_insert() - Insert new rule in field given input key and mask length
 * @f:		Field containing lookup table
 * @k:		Input key for classification, without nftables padding
 * @mask_bits:	Length of mask; matches field length for non-ranged entry
 *
 * Insert a new rule reference in lookup buckets corresponding to k and
 * mask_bits.
 *
 * Return: 1 on success (one rule inserted), negative error code on failure.
 */
static int pipapo_insert(struct nft_pipapo_field *f, const uint8_t *k,
			 int mask_bits)
{
	unsigned int rule = f->rules, group, ret, bit_offset = 0;

	ret = pipapo_resize(f, f->rules, f->rules + 1);
	if (ret)
		return ret;

	f->rules++;

	for (group = 0; group < f->groups; group++) {
		int i, v;
		u8 mask;

		v = k[group / (BITS_PER_BYTE / f->bb)];
		v &= GENMASK(BITS_PER_BYTE - bit_offset - 1, 0);
		v >>= (BITS_PER_BYTE - bit_offset) - f->bb;

		bit_offset += f->bb;
		bit_offset %= BITS_PER_BYTE;

		if (mask_bits >= (group + 1) * f->bb) {
			/* Not masked */
			pipapo_bucket_set(f, rule, group, v);
		} else if (mask_bits <= group * f->bb) {
			/* Completely masked */
			for (i = 0; i < NFT_PIPAPO_BUCKETS(f->bb); i++)
				pipapo_bucket_set(f, rule, group, i);
		} else {
			/* The mask limit falls on this group */
			mask = GENMASK(f->bb - 1, 0);
			mask >>= mask_bits - group * f->bb;
			for (i = 0; i < NFT_PIPAPO_BUCKETS(f->bb); i++) {
				if ((i & ~mask) == (v & ~mask))
					pipapo_bucket_set(f, rule, group, i);
			}
		}
	}

	pipapo_lt_bits_adjust(f);

	return 1;
}

/**
 * pipapo_step_diff() - Check if setting @step bit in netmask would change it
 * @base:	Mask we are expanding
 * @step:	Step bit for given expansion step
 * @len:	Total length of mask space (set and unset bits), bytes
 *
 * Convenience function for mask expansion.
 *
 * Return: true if step bit changes mask (i.e. isn't set), false otherwise.
 */
static bool pipapo_step_diff(u8 *base, int step, int len)
{
	/* Network order, byte-addressed */
#ifdef __BIG_ENDIAN__
	return !(BIT(step % BITS_PER_BYTE) & base[step / BITS_PER_BYTE]);
#else
	return !(BIT(step % BITS_PER_BYTE) &
		 base[len - 1 - step / BITS_PER_BYTE]);
#endif
}

/**
 * pipapo_step_after_end() - Check if mask exceeds range end with given step
 * @base:	Mask we are expanding
 * @end:	End of range
 * @step:	Step bit for given expansion step, highest bit to be set
 * @len:	Total length of mask space (set and unset bits), bytes
 *
 * Convenience function for mask expansion.
 *
 * Return: true if mask exceeds range setting step bits, false otherwise.
 */
static bool pipapo_step_after_end(const u8 *base, const u8 *end, int step,
				  int len)
{
	u8 tmp[NFT_PIPAPO_MAX_BYTES];
	int i;

	memcpy(tmp, base, len);

	/* Network order, byte-addressed */
	for (i = 0; i <= step; i++)
#ifdef __BIG_ENDIAN__
		tmp[i / BITS_PER_BYTE] |= BIT(i % BITS_PER_BYTE);
#else
		tmp[len - 1 - i / BITS_PER_BYTE] |= BIT(i % BITS_PER_BYTE);
#endif

	return memcmp(tmp, end, len) > 0;
}

/**
 * pipapo_base_sum() - Sum step bit to given len-sized netmask base with carry
 * @base:	Netmask base
 * @step:	Step bit to sum
 * @len:	Netmask length, bytes
 */
static void pipapo_base_sum(u8 *base, int step, int len)
{
	bool carry = false;
	int i;

	/* Network order, byte-addressed */
#ifdef __BIG_ENDIAN__
	for (i = step / BITS_PER_BYTE; i < len; i++) {
#else
	for (i = len - 1 - step / BITS_PER_BYTE; i >= 0; i--) {
#endif
		if (carry)
			base[i]++;
		else
			base[i] += 1 << (step % BITS_PER_BYTE);

		if (base[i])
			break;

		carry = true;
	}
}

/**
 * pipapo_expand() - Expand to composing netmasks, insert into lookup table
 * @f:		Field containing lookup table
 * @start:	Start of range
 * @end:	End of range
 * @len:	Length of value in bits
 *
 * Expand range to composing netmasks and insert corresponding rule references
 * in lookup buckets.
 *
 * Return: number of inserted rules on success, negative error code on failure.
 */
static int pipapo_expand(struct nft_pipapo_field *f,
			 const u8 *start, const u8 *end, int len)
{
	int step, masks = 0, bytes = DIV_ROUND_UP(len, BITS_PER_BYTE);
	u8 base[NFT_PIPAPO_MAX_BYTES];

	memcpy(base, start, bytes);
	while (memcmp(base, end, bytes) <= 0) {
		int err;

		step = 0;
		while (pipapo_step_diff(base, step, bytes)) {
			if (pipapo_step_after_end(base, end, step, bytes))
				break;

			step++;
			if (step >= len) {
				if (!masks) {
					err = pipapo_insert(f, base, 0);
					if (err < 0)
						return err;
					masks = 1;
				}
				goto out;
			}
		}

		err = pipapo_insert(f, base, len - step);

		if (err < 0)
			return err;

		masks++;
		pipapo_base_sum(base, step, bytes);
	}
out:
	return masks;
}

/**
 * pipapo_map() - Insert rules in mapping tables, mapping them between fields
 * @m:		Matching data, including mapping table
 * @map:	Table of rule maps: array of first rule and amount of rules
 *		in next field a given rule maps to, for each field
 * @e:		For last field, nft_set_ext pointer matching rules map to
 */
static void pipapo_map(struct nft_pipapo_match *m,
		       union nft_pipapo_map_bucket map[NFT_PIPAPO_MAX_FIELDS],
		       struct nft_pipapo_elem *e)
{
	struct nft_pipapo_field *f;
	int i, j;

	for (i = 0, f = m->f; i < m->field_count - 1; i++, f++) {
		for (j = 0; j < map[i].n; j++) {
			f->mt[map[i].to + j].to = map[i + 1].to;
			f->mt[map[i].to + j].n = map[i + 1].n;
		}
	}

	/* Last field: map to ext instead of mapping to next field */
	for (j = 0; j < map[i].n; j++)
		f->mt[map[i].to + j].e = e;
}

/**
 * pipapo_free_scratch() - Free per-CPU map at original (not aligned) address
 * @m:		Matching data
 * @cpu:	CPU number
 */
static void pipapo_free_scratch(const struct nft_pipapo_match *m, unsigned int cpu)
{
	struct nft_pipapo_scratch *s;
	void *mem;

	s = *per_cpu_ptr(m->scratch, cpu);
	if (!s)
		return;

	mem = s;
	mem -= s->align_off;
	kfree(mem);
}

/**
 * pipapo_realloc_scratch() - Reallocate scratch maps for partial match results
 * @clone:	Copy of matching data with pending insertions and deletions
 * @bsize_max:	Maximum bucket size, scratch maps cover two buckets
 *
 * Return: 0 on success, -ENOMEM on failure.
 */
static int pipapo_realloc_scratch(struct nft_pipapo_match *clone,
				  unsigned long bsize_max)
{
	int i;

	for_each_possible_cpu(i) {
		struct nft_pipapo_scratch *scratch;
#ifdef NFT_PIPAPO_ALIGN
		void *scratch_aligned;
		u32 align_off;
#endif
		scratch = kzalloc_node(struct_size(scratch, map,
						   bsize_max * 2) +
				       NFT_PIPAPO_ALIGN_HEADROOM,
				       GFP_KERNEL, cpu_to_node(i));
		if (!scratch) {
			/* On failure, there's no need to undo previous
			 * allocations: this means that some scratch maps have
			 * a bigger allocated size now (this is only called on
			 * insertion), but the extra space won't be used by any
			 * CPU as new elements are not inserted and m->bsize_max
			 * is not updated.
			 */
			return -ENOMEM;
		}

		pipapo_free_scratch(clone, i);

#ifdef NFT_PIPAPO_ALIGN
		/* Align &scratch->map (not the struct itself): the extra
		 * %NFT_PIPAPO_ALIGN_HEADROOM bytes passed to kzalloc_node()
		 * above guarantee we can waste up to those bytes in order
		 * to align the map field regardless of its offset within
		 * the struct.
		 */
		BUILD_BUG_ON(offsetof(struct nft_pipapo_scratch, map) > NFT_PIPAPO_ALIGN_HEADROOM);

		scratch_aligned = NFT_PIPAPO_LT_ALIGN(&scratch->map);
		scratch_aligned -= offsetof(struct nft_pipapo_scratch, map);
		align_off = scratch_aligned - (void *)scratch;

		scratch = scratch_aligned;
		scratch->align_off = align_off;
#endif
		*per_cpu_ptr(clone->scratch, i) = scratch;
	}

	return 0;
}

static bool nft_pipapo_transaction_mutex_held(const struct nft_set *set)
{
#ifdef CONFIG_PROVE_LOCKING
	const struct net *net = read_pnet(&set->net);

	return lockdep_is_held(&nft_pernet(net)->commit_mutex);
#else
	return true;
#endif
}

static struct nft_pipapo_match *pipapo_clone(struct nft_pipapo_match *old);

/**
 * pipapo_maybe_clone() - Build clone for pending data changes, if not existing
 * @set:	nftables API set representation
 *
 * Return: newly created or existing clone, if any. NULL on allocation failure
 */
static struct nft_pipapo_match *pipapo_maybe_clone(const struct nft_set *set)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct nft_pipapo_match *m;

	if (priv->clone)
		return priv->clone;

	m = rcu_dereference_protected(priv->match,
				      nft_pipapo_transaction_mutex_held(set));
	priv->clone = pipapo_clone(m);

	return priv->clone;
}

/**
 * nft_pipapo_insert() - Validate and insert ranged elements
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @elem:	nftables API element representation containing key data
 * @elem_priv:	Filled with pointer to &struct nft_set_ext in inserted element
 *
 * Return: 0 on success, error pointer on failure.
 */
static int nft_pipapo_insert(const struct net *net, const struct nft_set *set,
			     const struct nft_set_elem *elem,
			     struct nft_elem_priv **elem_priv)
{
	const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
	union nft_pipapo_map_bucket rulemap[NFT_PIPAPO_MAX_FIELDS];
	const u8 *start = (const u8 *)elem->key.val.data, *end;
	struct nft_pipapo_match *m = pipapo_maybe_clone(set);
	u8 genmask = nft_genmask_next(net);
	struct nft_pipapo_elem *e, *dup;
	u64 tstamp = nft_net_tstamp(net);
	struct nft_pipapo_field *f;
	const u8 *start_p, *end_p;
	int i, bsize_max, err = 0;

	if (!m)
		return -ENOMEM;

	if (nft_set_ext_exists(ext, NFT_SET_EXT_KEY_END))
		end = (const u8 *)nft_set_ext_key_end(ext)->data;
	else
		end = start;

	dup = pipapo_get(net, set, m, start, genmask, tstamp, GFP_KERNEL);
	if (!IS_ERR(dup)) {
		/* Check if we already have the same exact entry */
		const struct nft_data *dup_key, *dup_end;

		dup_key = nft_set_ext_key(&dup->ext);
		if (nft_set_ext_exists(&dup->ext, NFT_SET_EXT_KEY_END))
			dup_end = nft_set_ext_key_end(&dup->ext);
		else
			dup_end = dup_key;

		if (!memcmp(start, dup_key->data, sizeof(*dup_key->data)) &&
		    !memcmp(end, dup_end->data, sizeof(*dup_end->data))) {
			*elem_priv = &dup->priv;
			return -EEXIST;
		}

		return -ENOTEMPTY;
	}

	if (PTR_ERR(dup) == -ENOENT) {
		/* Look for partially overlapping entries */
		dup = pipapo_get(net, set, m, end, nft_genmask_next(net), tstamp,
				 GFP_KERNEL);
	}

	if (PTR_ERR(dup) != -ENOENT) {
		if (IS_ERR(dup))
			return PTR_ERR(dup);
		*elem_priv = &dup->priv;
		return -ENOTEMPTY;
	}

	/* Validate */
	start_p = start;
	end_p = end;

	/* some helpers return -1, or 0 >= for valid rule pos,
	 * so we cannot support more than INT_MAX rules at this time.
	 */
	BUILD_BUG_ON(NFT_PIPAPO_RULE0_MAX > INT_MAX);

	nft_pipapo_for_each_field(f, i, m) {
		if (f->rules >= NFT_PIPAPO_RULE0_MAX)
			return -ENOSPC;

		if (memcmp(start_p, end_p,
			   f->groups / NFT_PIPAPO_GROUPS_PER_BYTE(f)) > 0)
			return -EINVAL;

		start_p += NFT_PIPAPO_GROUPS_PADDED_SIZE(f);
		end_p += NFT_PIPAPO_GROUPS_PADDED_SIZE(f);
	}

	/* Insert */
	bsize_max = m->bsize_max;

	nft_pipapo_for_each_field(f, i, m) {
		int ret;

		rulemap[i].to = f->rules;

		ret = memcmp(start, end,
			     f->groups / NFT_PIPAPO_GROUPS_PER_BYTE(f));
		if (!ret)
			ret = pipapo_insert(f, start, f->groups * f->bb);
		else
			ret = pipapo_expand(f, start, end, f->groups * f->bb);

		if (ret < 0)
			return ret;

		if (f->bsize > bsize_max)
			bsize_max = f->bsize;

		rulemap[i].n = ret;

		start += NFT_PIPAPO_GROUPS_PADDED_SIZE(f);
		end += NFT_PIPAPO_GROUPS_PADDED_SIZE(f);
	}

	if (!*get_cpu_ptr(m->scratch) || bsize_max > m->bsize_max) {
		put_cpu_ptr(m->scratch);

		err = pipapo_realloc_scratch(m, bsize_max);
		if (err)
			return err;

		m->bsize_max = bsize_max;
	} else {
		put_cpu_ptr(m->scratch);
	}

	e = nft_elem_priv_cast(elem->priv);
	*elem_priv = &e->priv;

	pipapo_map(m, rulemap, e);

	return 0;
}

/**
 * pipapo_clone() - Clone matching data to create new working copy
 * @old:	Existing matching data
 *
 * Return: copy of matching data passed as 'old' or NULL.
 */
static struct nft_pipapo_match *pipapo_clone(struct nft_pipapo_match *old)
{
	struct nft_pipapo_field *dst, *src;
	struct nft_pipapo_match *new;
	int i;

	new = kmalloc(struct_size(new, f, old->field_count), GFP_KERNEL);
	if (!new)
		return NULL;

	new->field_count = old->field_count;
	new->bsize_max = old->bsize_max;

	new->scratch = alloc_percpu(*new->scratch);
	if (!new->scratch)
		goto out_scratch;

	for_each_possible_cpu(i)
		*per_cpu_ptr(new->scratch, i) = NULL;

	if (pipapo_realloc_scratch(new, old->bsize_max))
		goto out_scratch_realloc;

	rcu_head_init(&new->rcu);

	src = old->f;
	dst = new->f;

	for (i = 0; i < old->field_count; i++) {
		unsigned long *new_lt;

		memcpy(dst, src, offsetof(struct nft_pipapo_field, lt));

		new_lt = kvzalloc(src->groups * NFT_PIPAPO_BUCKETS(src->bb) *
				  src->bsize * sizeof(*dst->lt) +
				  NFT_PIPAPO_ALIGN_HEADROOM,
				  GFP_KERNEL);
		if (!new_lt)
			goto out_lt;

		dst->lt = new_lt;

		memcpy(NFT_PIPAPO_LT_ALIGN(new_lt),
		       NFT_PIPAPO_LT_ALIGN(src->lt),
		       src->bsize * sizeof(*dst->lt) *
		       src->groups * NFT_PIPAPO_BUCKETS(src->bb));

		if (src->rules > 0) {
			dst->mt = kvmalloc_array(src->rules_alloc,
						 sizeof(*src->mt), GFP_KERNEL);
			if (!dst->mt)
				goto out_mt;

			memcpy(dst->mt, src->mt, src->rules * sizeof(*src->mt));
		} else {
			dst->mt = NULL;
			dst->rules_alloc = 0;
		}

		src++;
		dst++;
	}

	return new;

out_mt:
	kvfree(dst->lt);
out_lt:
	for (dst--; i > 0; i--) {
		kvfree(dst->mt);
		kvfree(dst->lt);
		dst--;
	}
out_scratch_realloc:
	for_each_possible_cpu(i)
		pipapo_free_scratch(new, i);
out_scratch:
	free_percpu(new->scratch);
	kfree(new);

	return NULL;
}

/**
 * pipapo_rules_same_key() - Get number of rules originated from the same entry
 * @f:		Field containing mapping table
 * @first:	Index of first rule in set of rules mapping to same entry
 *
 * Using the fact that all rules in a field that originated from the same entry
 * will map to the same set of rules in the next field, or to the same element
 * reference, return the cardinality of the set of rules that originated from
 * the same entry as the rule with index @first, @first rule included.
 *
 * In pictures:
 *				rules
 *	field #0		0    1    2    3    4
 *		map to:		0    1   2-4  2-4  5-9
 *				.    .    .......   . ...
 *				|    |    |    | \   \
 *				|    |    |    |  \   \
 *				|    |    |    |   \   \
 *				'    '    '    '    '   \
 *	in field #1		0    1    2    3    4    5 ...
 *
 * if this is called for rule 2 on field #0, it will return 3, as also rules 2
 * and 3 in field 0 map to the same set of rules (2, 3, 4) in the next field.
 *
 * For the last field in a set, we can rely on associated entries to map to the
 * same element references.
 *
 * Return: Number of rules that originated from the same entry as @first.
 */
static unsigned int pipapo_rules_same_key(struct nft_pipapo_field *f, unsigned int first)
{
	struct nft_pipapo_elem *e = NULL; /* Keep gcc happy */
	unsigned int r;

	for (r = first; r < f->rules; r++) {
		if (r != first && e != f->mt[r].e)
			return r - first;

		e = f->mt[r].e;
	}

	if (r != first)
		return r - first;

	return 0;
}

/**
 * pipapo_unmap() - Remove rules from mapping tables, renumber remaining ones
 * @mt:		Mapping array
 * @rules:	Original amount of rules in mapping table
 * @start:	First rule index to be removed
 * @n:		Amount of rules to be removed
 * @to_offset:	First rule index, in next field, this group of rules maps to
 * @is_last:	If this is the last field, delete reference from mapping array
 *
 * This is used to unmap rules from the mapping table for a single field,
 * maintaining consistency and compactness for the existing ones.
 *
 * In pictures: let's assume that we want to delete rules 2 and 3 from the
 * following mapping array:
 *
 *                 rules
 *               0      1      2      3      4
 *      map to:  4-10   4-10   11-15  11-15  16-18
 *
 * the result will be:
 *
 *                 rules
 *               0      1      2
 *      map to:  4-10   4-10   11-13
 *
 * for fields before the last one. In case this is the mapping table for the
 * last field in a set, and rules map to pointers to &struct nft_pipapo_elem:
 *
 *                      rules
 *                        0      1      2      3      4
 *  element pointers:  0x42   0x42   0x33   0x33   0x44
 *
 * the result will be:
 *
 *                      rules
 *                        0      1      2
 *  element pointers:  0x42   0x42   0x44
 */
static void pipapo_unmap(union nft_pipapo_map_bucket *mt, unsigned int rules,
			 unsigned int start, unsigned int n,
			 unsigned int to_offset, bool is_last)
{
	int i;

	memmove(mt + start, mt + start + n, (rules - start - n) * sizeof(*mt));
	memset(mt + rules - n, 0, n * sizeof(*mt));

	if (is_last)
		return;

	for (i = start; i < rules - n; i++)
		mt[i].to -= to_offset;
}

/**
 * pipapo_drop() - Delete entry from lookup and mapping tables, given rule map
 * @m:		Matching data
 * @rulemap:	Table of rule maps, arrays of first rule and amount of rules
 *		in next field a given entry maps to, for each field
 *
 * For each rule in lookup table buckets mapping to this set of rules, drop
 * all bits set in lookup table mapping. In pictures, assuming we want to drop
 * rules 0 and 1 from this lookup table:
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0    0                                              1,2
 *        1   1,2                                      0
 *        2    0                                      1,2
 *        3    0                              1,2
 *        4  0,1,2
 *        5    0   1   2
 *        6  0,1,2 1   1   1   1   1   1   1   1   1   1   1   1   1   1   1
 *        7   1,2 1,2  1   1   1  0,1  1   1   1   1   1   1   1   1   1   1
 *
 * rule 2 becomes rule 0, and the result will be:
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0                                                    0
 *        1    0
 *        2                                            0
 *        3                                    0
 *        4    0
 *        5            0
 *        6    0
 *        7    0   0
 *
 * once this is done, call unmap() to drop all the corresponding rule references
 * from mapping tables.
 */
static void pipapo_drop(struct nft_pipapo_match *m,
			union nft_pipapo_map_bucket rulemap[])
{
	struct nft_pipapo_field *f;
	int i;

	nft_pipapo_for_each_field(f, i, m) {
		int g;

		for (g = 0; g < f->groups; g++) {
			unsigned long *pos;
			int b;

			pos = NFT_PIPAPO_LT_ALIGN(f->lt) + g *
			      NFT_PIPAPO_BUCKETS(f->bb) * f->bsize;

			for (b = 0; b < NFT_PIPAPO_BUCKETS(f->bb); b++) {
				bitmap_cut(pos, pos, rulemap[i].to,
					   rulemap[i].n,
					   f->bsize * BITS_PER_LONG);

				pos += f->bsize;
			}
		}

		pipapo_unmap(f->mt, f->rules, rulemap[i].to, rulemap[i].n,
			     rulemap[i + 1].n, i == m->field_count - 1);
		if (pipapo_resize(f, f->rules, f->rules - rulemap[i].n)) {
			/* We can ignore this, a failure to shrink tables down
			 * doesn't make tables invalid.
			 */
			;
		}
		f->rules -= rulemap[i].n;

		pipapo_lt_bits_adjust(f);
	}
}

static void nft_pipapo_gc_deactivate(struct net *net, struct nft_set *set,
				     struct nft_pipapo_elem *e)

{
	nft_setelem_data_deactivate(net, set, &e->priv);
}

/**
 * pipapo_gc() - Drop expired entries from set, destroy start and end elements
 * @set:	nftables API set representation
 * @m:		Matching data
 */
static void pipapo_gc(struct nft_set *set, struct nft_pipapo_match *m)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct net *net = read_pnet(&set->net);
	unsigned int rules_f0, first_rule = 0;
	u64 tstamp = nft_net_tstamp(net);
	struct nft_pipapo_elem *e;
	struct nft_trans_gc *gc;

	gc = nft_trans_gc_alloc(set, 0, GFP_KERNEL);
	if (!gc)
		return;

	while ((rules_f0 = pipapo_rules_same_key(m->f, first_rule))) {
		union nft_pipapo_map_bucket rulemap[NFT_PIPAPO_MAX_FIELDS];
		const struct nft_pipapo_field *f;
		unsigned int i, start, rules_fx;

		start = first_rule;
		rules_fx = rules_f0;

		nft_pipapo_for_each_field(f, i, m) {
			rulemap[i].to = start;
			rulemap[i].n = rules_fx;

			if (i < m->field_count - 1) {
				rules_fx = f->mt[start].n;
				start = f->mt[start].to;
			}
		}

		/* Pick the last field, and its last index */
		f--;
		i--;
		e = f->mt[rulemap[i].to].e;

		/* synchronous gc never fails, there is no need to set on
		 * NFT_SET_ELEM_DEAD_BIT.
		 */
		if (__nft_set_elem_expired(&e->ext, tstamp)) {
			gc = nft_trans_gc_queue_sync(gc, GFP_KERNEL);
			if (!gc)
				return;

			nft_pipapo_gc_deactivate(net, set, e);
			pipapo_drop(m, rulemap);
			nft_trans_gc_elem_add(gc, e);

			/* And check again current first rule, which is now the
			 * first we haven't checked.
			 */
		} else {
			first_rule += rules_f0;
		}
	}

	gc = nft_trans_gc_catchall_sync(gc);
	if (gc) {
		nft_trans_gc_queue_sync_done(gc);
		priv->last_gc = jiffies;
	}
}

/**
 * pipapo_free_fields() - Free per-field tables contained in matching data
 * @m:		Matching data
 */
static void pipapo_free_fields(struct nft_pipapo_match *m)
{
	struct nft_pipapo_field *f;
	int i;

	nft_pipapo_for_each_field(f, i, m) {
		kvfree(f->lt);
		kvfree(f->mt);
	}
}

static void pipapo_free_match(struct nft_pipapo_match *m)
{
	int i;

	for_each_possible_cpu(i)
		pipapo_free_scratch(m, i);

	free_percpu(m->scratch);
	pipapo_free_fields(m);

	kfree(m);
}

/**
 * pipapo_reclaim_match - RCU callback to free fields from old matching data
 * @rcu:	RCU head
 */
static void pipapo_reclaim_match(struct rcu_head *rcu)
{
	struct nft_pipapo_match *m;

	m = container_of(rcu, struct nft_pipapo_match, rcu);
	pipapo_free_match(m);
}

/**
 * nft_pipapo_commit() - Replace lookup data with current working copy
 * @set:	nftables API set representation
 *
 * While at it, check if we should perform garbage collection on the working
 * copy before committing it for lookup, and don't replace the table if the
 * working copy doesn't have pending changes.
 *
 * We also need to create a new working copy for subsequent insertions and
 * deletions.
 */
static void nft_pipapo_commit(struct nft_set *set)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct nft_pipapo_match *old;

	if (!priv->clone)
		return;

	if (time_after_eq(jiffies, priv->last_gc + nft_set_gc_interval(set)))
		pipapo_gc(set, priv->clone);

	old = rcu_replace_pointer(priv->match, priv->clone,
				  nft_pipapo_transaction_mutex_held(set));
	priv->clone = NULL;

	if (old)
		call_rcu(&old->rcu, pipapo_reclaim_match);
}

static void nft_pipapo_abort(const struct nft_set *set)
{
	struct nft_pipapo *priv = nft_set_priv(set);

	if (!priv->clone)
		return;
	pipapo_free_match(priv->clone);
	priv->clone = NULL;
}

/**
 * nft_pipapo_activate() - Mark element reference as active given key, commit
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @elem_priv:	nftables API element representation containing key data
 *
 * On insertion, elements are added to a copy of the matching data currently
 * in use for lookups, and not directly inserted into current lookup data. Both
 * nft_pipapo_insert() and nft_pipapo_activate() are called once for each
 * element, hence we can't purpose either one as a real commit operation.
 */
static void nft_pipapo_activate(const struct net *net,
				const struct nft_set *set,
				struct nft_elem_priv *elem_priv)
{
	struct nft_pipapo_elem *e = nft_elem_priv_cast(elem_priv);

	nft_clear(net, &e->ext);
}

/**
 * nft_pipapo_deactivate() - Search for element and make it inactive
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @elem:	nftables API element representation containing key data
 *
 * Return: deactivated element if found, NULL otherwise.
 */
static struct nft_elem_priv *
nft_pipapo_deactivate(const struct net *net, const struct nft_set *set,
		      const struct nft_set_elem *elem)
{
	struct nft_pipapo_match *m = pipapo_maybe_clone(set);
	struct nft_pipapo_elem *e;

	/* removal must occur on priv->clone, if we are low on memory
	 * we have no choice and must fail the removal request.
	 */
	if (!m)
		return NULL;

	e = pipapo_get(net, set, m, (const u8 *)elem->key.val.data,
		       nft_genmask_next(net), nft_net_tstamp(net), GFP_KERNEL);
	if (IS_ERR(e))
		return NULL;

	nft_set_elem_change_active(net, set, &e->ext);

	return &e->priv;
}

/**
 * nft_pipapo_flush() - make element inactive
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @elem_priv:	nftables API element representation containing key data
 *
 * This is functionally the same as nft_pipapo_deactivate(), with a slightly
 * different interface, and it's also called once for each element in a set
 * being flushed, so we can't implement, strictly speaking, a flush operation,
 * which would otherwise be as simple as allocating an empty copy of the
 * matching data.
 *
 * Note that we could in theory do that, mark the set as flushed, and ignore
 * subsequent calls, but we would leak all the elements after the first one,
 * because they wouldn't then be freed as result of API calls.
 *
 * Return: true if element was found and deactivated.
 */
static void nft_pipapo_flush(const struct net *net, const struct nft_set *set,
			     struct nft_elem_priv *elem_priv)
{
	struct nft_pipapo_elem *e = nft_elem_priv_cast(elem_priv);

	nft_set_elem_change_active(net, set, &e->ext);
}

/**
 * pipapo_get_boundaries() - Get byte interval for associated rules
 * @f:		Field including lookup table
 * @first_rule:	First rule (lowest index)
 * @rule_count:	Number of associated rules
 * @left:	Byte expression for left boundary (start of range)
 * @right:	Byte expression for right boundary (end of range)
 *
 * Given the first rule and amount of rules that originated from the same entry,
 * build the original range associated with the entry, and calculate the length
 * of the originating netmask.
 *
 * In pictures:
 *
 *                     bucket
 *      group  0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
 *        0                                                   1,2
 *        1   1,2
 *        2                                           1,2
 *        3                                   1,2
 *        4   1,2
 *        5        1   2
 *        6   1,2  1   1   1   1   1   1   1   1   1   1   1   1   1   1   1
 *        7   1,2 1,2  1   1   1   1   1   1   1   1   1   1   1   1   1   1
 *
 * this is the lookup table corresponding to the IPv4 range
 * 192.168.1.0-192.168.2.1, which was expanded to the two composing netmasks,
 * rule #1: 192.168.1.0/24, and rule #2: 192.168.2.0/31.
 *
 * This function fills @left and @right with the byte values of the leftmost
 * and rightmost bucket indices for the lowest and highest rule indices,
 * respectively. If @first_rule is 1 and @rule_count is 2, we obtain, in
 * nibbles:
 *   left:  < 12, 0, 10, 8, 0, 1, 0, 0 >
 *   right: < 12, 0, 10, 8, 0, 2, 2, 1 >
 * corresponding to bytes:
 *   left:  < 192, 168, 1, 0 >
 *   right: < 192, 168, 2, 1 >
 * with mask length irrelevant here, unused on return, as the range is already
 * defined by its start and end points. The mask length is relevant for a single
 * ranged entry instead: if @first_rule is 1 and @rule_count is 1, we ignore
 * rule 2 above: @left becomes < 192, 168, 1, 0 >, @right becomes
 * < 192, 168, 1, 255 >, and the mask length, calculated from the distances
 * between leftmost and rightmost bucket indices for each group, would be 24.
 *
 * Return: mask length, in bits.
 */
static int pipapo_get_boundaries(struct nft_pipapo_field *f, int first_rule,
				 int rule_count, u8 *left, u8 *right)
{
	int g, mask_len = 0, bit_offset = 0;
	u8 *l = left, *r = right;

	for (g = 0; g < f->groups; g++) {
		int b, x0, x1;

		x0 = -1;
		x1 = -1;
		for (b = 0; b < NFT_PIPAPO_BUCKETS(f->bb); b++) {
			unsigned long *pos;

			pos = NFT_PIPAPO_LT_ALIGN(f->lt) +
			      (g * NFT_PIPAPO_BUCKETS(f->bb) + b) * f->bsize;
			if (test_bit(first_rule, pos) && x0 == -1)
				x0 = b;
			if (test_bit(first_rule + rule_count - 1, pos))
				x1 = b;
		}

		*l |= x0 << (BITS_PER_BYTE - f->bb - bit_offset);
		*r |= x1 << (BITS_PER_BYTE - f->bb - bit_offset);

		bit_offset += f->bb;
		if (bit_offset >= BITS_PER_BYTE) {
			bit_offset %= BITS_PER_BYTE;
			l++;
			r++;
		}

		if (x1 - x0 == 0)
			mask_len += 4;
		else if (x1 - x0 == 1)
			mask_len += 3;
		else if (x1 - x0 == 3)
			mask_len += 2;
		else if (x1 - x0 == 7)
			mask_len += 1;
	}

	return mask_len;
}

/**
 * pipapo_match_field() - Match rules against byte ranges
 * @f:		Field including the lookup table
 * @first_rule:	First of associated rules originating from same entry
 * @rule_count:	Amount of associated rules
 * @start:	Start of range to be matched
 * @end:	End of range to be matched
 *
 * Return: true on match, false otherwise.
 */
static bool pipapo_match_field(struct nft_pipapo_field *f,
			       int first_rule, int rule_count,
			       const u8 *start, const u8 *end)
{
	u8 right[NFT_PIPAPO_MAX_BYTES] = { 0 };
	u8 left[NFT_PIPAPO_MAX_BYTES] = { 0 };

	pipapo_get_boundaries(f, first_rule, rule_count, left, right);

	return !memcmp(start, left,
		       f->groups / NFT_PIPAPO_GROUPS_PER_BYTE(f)) &&
	       !memcmp(end, right, f->groups / NFT_PIPAPO_GROUPS_PER_BYTE(f));
}

/**
 * nft_pipapo_remove() - Remove element given key, commit
 * @net:	Network namespace
 * @set:	nftables API set representation
 * @elem_priv:	nftables API element representation containing key data
 *
 * Similarly to nft_pipapo_activate(), this is used as commit operation by the
 * API, but it's called once per element in the pending transaction, so we can't
 * implement this as a single commit operation. Closest we can get is to remove
 * the matched element here, if any, and commit the updated matching data.
 */
static void nft_pipapo_remove(const struct net *net, const struct nft_set *set,
			      struct nft_elem_priv *elem_priv)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct nft_pipapo_match *m = priv->clone;
	unsigned int rules_f0, first_rule = 0;
	struct nft_pipapo_elem *e;
	const u8 *data;

	e = nft_elem_priv_cast(elem_priv);
	data = (const u8 *)nft_set_ext_key(&e->ext);

	while ((rules_f0 = pipapo_rules_same_key(m->f, first_rule))) {
		union nft_pipapo_map_bucket rulemap[NFT_PIPAPO_MAX_FIELDS];
		const u8 *match_start, *match_end;
		struct nft_pipapo_field *f;
		int i, start, rules_fx;

		match_start = data;

		if (nft_set_ext_exists(&e->ext, NFT_SET_EXT_KEY_END))
			match_end = (const u8 *)nft_set_ext_key_end(&e->ext)->data;
		else
			match_end = data;

		start = first_rule;
		rules_fx = rules_f0;

		nft_pipapo_for_each_field(f, i, m) {
			bool last = i == m->field_count - 1;

			if (!pipapo_match_field(f, start, rules_fx,
						match_start, match_end))
				break;

			rulemap[i].to = start;
			rulemap[i].n = rules_fx;

			rules_fx = f->mt[start].n;
			start = f->mt[start].to;

			match_start += NFT_PIPAPO_GROUPS_PADDED_SIZE(f);
			match_end += NFT_PIPAPO_GROUPS_PADDED_SIZE(f);

			if (last && f->mt[rulemap[i].to].e == e) {
				pipapo_drop(m, rulemap);
				return;
			}
		}

		first_rule += rules_f0;
	}

	WARN_ON_ONCE(1); /* elem_priv not found */
}

/**
 * nft_pipapo_do_walk() - Walk over elements in m
 * @ctx:	nftables API context
 * @set:	nftables API set representation
 * @m:		matching data pointing to key mapping array
 * @iter:	Iterator
 *
 * As elements are referenced in the mapping array for the last field, directly
 * scan that array: there's no need to follow rule mappings from the first
 * field. @m is protected either by RCU read lock or by transaction mutex.
 */
static void nft_pipapo_do_walk(const struct nft_ctx *ctx, struct nft_set *set,
			       const struct nft_pipapo_match *m,
			       struct nft_set_iter *iter)
{
	const struct nft_pipapo_field *f;
	unsigned int i, r;

	for (i = 0, f = m->f; i < m->field_count - 1; i++, f++)
		;

	for (r = 0; r < f->rules; r++) {
		struct nft_pipapo_elem *e;

		if (r < f->rules - 1 && f->mt[r + 1].e == f->mt[r].e)
			continue;

		if (iter->count < iter->skip)
			goto cont;

		e = f->mt[r].e;

		iter->err = iter->fn(ctx, set, iter, &e->priv);
		if (iter->err < 0)
			return;

cont:
		iter->count++;
	}
}

/**
 * nft_pipapo_walk() - Walk over elements
 * @ctx:	nftables API context
 * @set:	nftables API set representation
 * @iter:	Iterator
 *
 * Test if destructive action is needed or not, clone active backend if needed
 * and call the real function to work on the data.
 */
static void nft_pipapo_walk(const struct nft_ctx *ctx, struct nft_set *set,
			    struct nft_set_iter *iter)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	const struct nft_pipapo_match *m;

	switch (iter->type) {
	case NFT_ITER_UPDATE:
		m = pipapo_maybe_clone(set);
		if (!m) {
			iter->err = -ENOMEM;
			return;
		}

		nft_pipapo_do_walk(ctx, set, m, iter);
		break;
	case NFT_ITER_READ:
		rcu_read_lock();
		m = rcu_dereference(priv->match);
		nft_pipapo_do_walk(ctx, set, m, iter);
		rcu_read_unlock();
		break;
	default:
		iter->err = -EINVAL;
		WARN_ON_ONCE(1);
		break;
	}
}

/**
 * nft_pipapo_privsize() - Return the size of private data for the set
 * @nla:	netlink attributes, ignored as size doesn't depend on them
 * @desc:	Set description, ignored as size doesn't depend on it
 *
 * Return: size of private data for this set implementation, in bytes
 */
static u64 nft_pipapo_privsize(const struct nlattr * const nla[],
			       const struct nft_set_desc *desc)
{
	return sizeof(struct nft_pipapo);
}

/**
 * nft_pipapo_estimate() - Set size, space and lookup complexity
 * @desc:	Set description, element count and field description used
 * @features:	Flags: NFT_SET_INTERVAL needs to be there
 * @est:	Storage for estimation data
 *
 * Return: true if set description is compatible, false otherwise
 */
static bool nft_pipapo_estimate(const struct nft_set_desc *desc, u32 features,
				struct nft_set_estimate *est)
{
	if (!(features & NFT_SET_INTERVAL) ||
	    desc->field_count < NFT_PIPAPO_MIN_FIELDS)
		return false;

	est->size = pipapo_estimate_size(desc);
	if (!est->size)
		return false;

	est->lookup = NFT_SET_CLASS_O_LOG_N;

	est->space = NFT_SET_CLASS_O_N;

	return true;
}

/**
 * nft_pipapo_init() - Initialise data for a set instance
 * @set:	nftables API set representation
 * @desc:	Set description
 * @nla:	netlink attributes
 *
 * Validate number and size of fields passed as NFTA_SET_DESC_CONCAT netlink
 * attributes, initialise internal set parameters, current instance of matching
 * data and a copy for subsequent insertions.
 *
 * Return: 0 on success, negative error code on failure.
 */
static int nft_pipapo_init(const struct nft_set *set,
			   const struct nft_set_desc *desc,
			   const struct nlattr * const nla[])
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct nft_pipapo_match *m;
	struct nft_pipapo_field *f;
	int err, i, field_count;

	BUILD_BUG_ON(offsetof(struct nft_pipapo_elem, priv) != 0);

	field_count = desc->field_count ? : 1;

	BUILD_BUG_ON(NFT_PIPAPO_MAX_FIELDS > 255);
	BUILD_BUG_ON(NFT_PIPAPO_MAX_FIELDS != NFT_REG32_COUNT);

	if (field_count > NFT_PIPAPO_MAX_FIELDS)
		return -EINVAL;

	m = kmalloc(struct_size(m, f, field_count), GFP_KERNEL);
	if (!m)
		return -ENOMEM;

	m->field_count = field_count;
	m->bsize_max = 0;

	m->scratch = alloc_percpu(struct nft_pipapo_scratch *);
	if (!m->scratch) {
		err = -ENOMEM;
		goto out_scratch;
	}
	for_each_possible_cpu(i)
		*per_cpu_ptr(m->scratch, i) = NULL;

	rcu_head_init(&m->rcu);

	nft_pipapo_for_each_field(f, i, m) {
		unsigned int len = desc->field_len[i] ? : set->klen;

		/* f->groups is u8 */
		BUILD_BUG_ON((NFT_PIPAPO_MAX_BYTES *
			      BITS_PER_BYTE / NFT_PIPAPO_GROUP_BITS_LARGE_SET) >= 256);

		f->bb = NFT_PIPAPO_GROUP_BITS_INIT;
		f->groups = len * NFT_PIPAPO_GROUPS_PER_BYTE(f);

		priv->width += round_up(len, sizeof(u32));

		f->bsize = 0;
		f->rules = 0;
		f->rules_alloc = 0;
		f->lt = NULL;
		f->mt = NULL;
	}

	rcu_assign_pointer(priv->match, m);

	return 0;

out_scratch:
	kfree(m);

	return err;
}

/**
 * nft_set_pipapo_match_destroy() - Destroy elements from key mapping array
 * @ctx:	context
 * @set:	nftables API set representation
 * @m:		matching data pointing to key mapping array
 */
static void nft_set_pipapo_match_destroy(const struct nft_ctx *ctx,
					 const struct nft_set *set,
					 struct nft_pipapo_match *m)
{
	struct nft_pipapo_field *f;
	unsigned int i, r;

	for (i = 0, f = m->f; i < m->field_count - 1; i++, f++)
		;

	for (r = 0; r < f->rules; r++) {
		struct nft_pipapo_elem *e;

		if (r < f->rules - 1 && f->mt[r + 1].e == f->mt[r].e)
			continue;

		e = f->mt[r].e;

		nf_tables_set_elem_destroy(ctx, set, &e->priv);
	}
}

/**
 * nft_pipapo_destroy() - Free private data for set and all committed elements
 * @ctx:	context
 * @set:	nftables API set representation
 */
static void nft_pipapo_destroy(const struct nft_ctx *ctx,
			       const struct nft_set *set)
{
	struct nft_pipapo *priv = nft_set_priv(set);
	struct nft_pipapo_match *m;

	m = rcu_dereference_protected(priv->match, true);

	if (priv->clone) {
		nft_set_pipapo_match_destroy(ctx, set, priv->clone);
		pipapo_free_match(priv->clone);
		priv->clone = NULL;
	} else {
		nft_set_pipapo_match_destroy(ctx, set, m);
	}

	pipapo_free_match(m);
}

/**
 * nft_pipapo_gc_init() - Initialise garbage collection
 * @set:	nftables API set representation
 *
 * Instead of actually setting up a periodic work for garbage collection, as
 * this operation requires a swap of matching data with the working copy, we'll
 * do that opportunistically with other commit operations if the interval is
 * elapsed, so we just need to set the current jiffies timestamp here.
 */
static void nft_pipapo_gc_init(const struct nft_set *set)
{
	struct nft_pipapo *priv = nft_set_priv(set);

	priv->last_gc = jiffies;
}

const struct nft_set_type nft_set_pipapo_type = {
	.features	= NFT_SET_INTERVAL | NFT_SET_MAP | NFT_SET_OBJECT |
			  NFT_SET_TIMEOUT,
	.ops		= {
		.lookup		= nft_pipapo_lookup,
		.insert		= nft_pipapo_insert,
		.activate	= nft_pipapo_activate,
		.deactivate	= nft_pipapo_deactivate,
		.flush		= nft_pipapo_flush,
		.remove		= nft_pipapo_remove,
		.walk		= nft_pipapo_walk,
		.get		= nft_pipapo_get,
		.privsize	= nft_pipapo_privsize,
		.estimate	= nft_pipapo_estimate,
		.init		= nft_pipapo_init,
		.destroy	= nft_pipapo_destroy,
		.gc_init	= nft_pipapo_gc_init,
		.commit		= nft_pipapo_commit,
		.abort		= nft_pipapo_abort,
		.elemsize	= offsetof(struct nft_pipapo_elem, ext),
	},
};

#if defined(CONFIG_X86_64) && !defined(CONFIG_UML)
const struct nft_set_type nft_set_pipapo_avx2_type = {
	.features	= NFT_SET_INTERVAL | NFT_SET_MAP | NFT_SET_OBJECT |
			  NFT_SET_TIMEOUT,
	.ops		= {
		.lookup		= nft_pipapo_avx2_lookup,
		.insert		= nft_pipapo_insert,
		.activate	= nft_pipapo_activate,
		.deactivate	= nft_pipapo_deactivate,
		.flush		= nft_pipapo_flush,
		.remove		= nft_pipapo_remove,
		.walk		= nft_pipapo_walk,
		.get		= nft_pipapo_get,
		.privsize	= nft_pipapo_privsize,
		.estimate	= nft_pipapo_avx2_estimate,
		.init		= nft_pipapo_init,
		.destroy	= nft_pipapo_destroy,
		.gc_init	= nft_pipapo_gc_init,
		.commit		= nft_pipapo_commit,
		.abort		= nft_pipapo_abort,
		.elemsize	= offsetof(struct nft_pipapo_elem, ext),
	},
};
#endif