summaryrefslogtreecommitdiff
path: root/drivers/md/dm-bufio.c
blob: 098bf526136c942508b4e6bf302ad4cc5c5c1399 (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
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
// SPDX-License-Identifier: GPL-2.0-only
/*
 * Copyright (C) 2009-2011 Red Hat, Inc.
 *
 * Author: Mikulas Patocka <mpatocka@redhat.com>
 *
 * This file is released under the GPL.
 */

#include <linux/dm-bufio.h>

#include <linux/device-mapper.h>
#include <linux/dm-io.h>
#include <linux/slab.h>
#include <linux/sched/mm.h>
#include <linux/jiffies.h>
#include <linux/vmalloc.h>
#include <linux/shrinker.h>
#include <linux/module.h>
#include <linux/rbtree.h>
#include <linux/stacktrace.h>
#include <linux/jump_label.h>

#include "dm.h"

#define DM_MSG_PREFIX "bufio"

/*
 * Memory management policy:
 *	Limit the number of buffers to DM_BUFIO_MEMORY_PERCENT of main memory
 *	or DM_BUFIO_VMALLOC_PERCENT of vmalloc memory (whichever is lower).
 *	Always allocate at least DM_BUFIO_MIN_BUFFERS buffers.
 *	Start background writeback when there are DM_BUFIO_WRITEBACK_PERCENT
 *	dirty buffers.
 */
#define DM_BUFIO_MIN_BUFFERS		8

#define DM_BUFIO_MEMORY_PERCENT		2
#define DM_BUFIO_VMALLOC_PERCENT	25
#define DM_BUFIO_WRITEBACK_RATIO	3
#define DM_BUFIO_LOW_WATERMARK_RATIO	16

/*
 * Check buffer ages in this interval (seconds)
 */
#define DM_BUFIO_WORK_TIMER_SECS	30

/*
 * Free buffers when they are older than this (seconds)
 */
#define DM_BUFIO_DEFAULT_AGE_SECS	300

/*
 * The nr of bytes of cached data to keep around.
 */
#define DM_BUFIO_DEFAULT_RETAIN_BYTES   (256 * 1024)

/*
 * Align buffer writes to this boundary.
 * Tests show that SSDs have the highest IOPS when using 4k writes.
 */
#define DM_BUFIO_WRITE_ALIGN		4096

/*
 * dm_buffer->list_mode
 */
#define LIST_CLEAN	0
#define LIST_DIRTY	1
#define LIST_SIZE	2

/*--------------------------------------------------------------*/

/*
 * Rather than use an LRU list, we use a clock algorithm where entries
 * are held in a circular list.  When an entry is 'hit' a reference bit
 * is set.  The least recently used entry is approximated by running a
 * cursor around the list selecting unreferenced entries. Referenced
 * entries have their reference bit cleared as the cursor passes them.
 */
struct lru_entry {
	struct list_head list;
	atomic_t referenced;
};

struct lru_iter {
	struct lru *lru;
	struct list_head list;
	struct lru_entry *stop;
	struct lru_entry *e;
};

struct lru {
	struct list_head *cursor;
	unsigned long count;

	struct list_head iterators;
};

/*--------------*/

static void lru_init(struct lru *lru)
{
	lru->cursor = NULL;
	lru->count = 0;
	INIT_LIST_HEAD(&lru->iterators);
}

static void lru_destroy(struct lru *lru)
{
	WARN_ON_ONCE(lru->cursor);
	WARN_ON_ONCE(!list_empty(&lru->iterators));
}

/*
 * Insert a new entry into the lru.
 */
static void lru_insert(struct lru *lru, struct lru_entry *le)
{
	/*
	 * Don't be tempted to set to 1, makes the lru aspect
	 * perform poorly.
	 */
	atomic_set(&le->referenced, 0);

	if (lru->cursor) {
		list_add_tail(&le->list, lru->cursor);
	} else {
		INIT_LIST_HEAD(&le->list);
		lru->cursor = &le->list;
	}
	lru->count++;
}

/*--------------*/

/*
 * Convert a list_head pointer to an lru_entry pointer.
 */
static inline struct lru_entry *to_le(struct list_head *l)
{
	return container_of(l, struct lru_entry, list);
}

/*
 * Initialize an lru_iter and add it to the list of cursors in the lru.
 */
static void lru_iter_begin(struct lru *lru, struct lru_iter *it)
{
	it->lru = lru;
	it->stop = lru->cursor ? to_le(lru->cursor->prev) : NULL;
	it->e = lru->cursor ? to_le(lru->cursor) : NULL;
	list_add(&it->list, &lru->iterators);
}

/*
 * Remove an lru_iter from the list of cursors in the lru.
 */
static inline void lru_iter_end(struct lru_iter *it)
{
	list_del(&it->list);
}

/* Predicate function type to be used with lru_iter_next */
typedef bool (*iter_predicate)(struct lru_entry *le, void *context);

/*
 * Advance the cursor to the next entry that passes the
 * predicate, and return that entry.  Returns NULL if the
 * iteration is complete.
 */
static struct lru_entry *lru_iter_next(struct lru_iter *it,
				       iter_predicate pred, void *context)
{
	struct lru_entry *e;

	while (it->e) {
		e = it->e;

		/* advance the cursor */
		if (it->e == it->stop)
			it->e = NULL;
		else
			it->e = to_le(it->e->list.next);

		if (pred(e, context))
			return e;
	}

	return NULL;
}

/*
 * Invalidate a specific lru_entry and update all cursors in
 * the lru accordingly.
 */
static void lru_iter_invalidate(struct lru *lru, struct lru_entry *e)
{
	struct lru_iter *it;

	list_for_each_entry(it, &lru->iterators, list) {
		/* Move c->e forwards if necc. */
		if (it->e == e) {
			it->e = to_le(it->e->list.next);
			if (it->e == e)
				it->e = NULL;
		}

		/* Move it->stop backwards if necc. */
		if (it->stop == e) {
			it->stop = to_le(it->stop->list.prev);
			if (it->stop == e)
				it->stop = NULL;
		}
	}
}

/*--------------*/

/*
 * Remove a specific entry from the lru.
 */
static void lru_remove(struct lru *lru, struct lru_entry *le)
{
	lru_iter_invalidate(lru, le);
	if (lru->count == 1) {
		lru->cursor = NULL;
	} else {
		if (lru->cursor == &le->list)
			lru->cursor = lru->cursor->next;
		list_del(&le->list);
	}
	lru->count--;
}

/*
 * Mark as referenced.
 */
static inline void lru_reference(struct lru_entry *le)
{
	atomic_set(&le->referenced, 1);
}

/*--------------*/

/*
 * Remove the least recently used entry (approx), that passes the predicate.
 * Returns NULL on failure.
 */
enum evict_result {
	ER_EVICT,
	ER_DONT_EVICT,
	ER_STOP, /* stop looking for something to evict */
};

typedef enum evict_result (*le_predicate)(struct lru_entry *le, void *context);

static struct lru_entry *lru_evict(struct lru *lru, le_predicate pred, void *context, bool no_sleep)
{
	unsigned long tested = 0;
	struct list_head *h = lru->cursor;
	struct lru_entry *le;

	if (!h)
		return NULL;
	/*
	 * In the worst case we have to loop around twice. Once to clear
	 * the reference flags, and then again to discover the predicate
	 * fails for all entries.
	 */
	while (tested < lru->count) {
		le = container_of(h, struct lru_entry, list);

		if (atomic_read(&le->referenced)) {
			atomic_set(&le->referenced, 0);
		} else {
			tested++;
			switch (pred(le, context)) {
			case ER_EVICT:
				/*
				 * Adjust the cursor, so we start the next
				 * search from here.
				 */
				lru->cursor = le->list.next;
				lru_remove(lru, le);
				return le;

			case ER_DONT_EVICT:
				break;

			case ER_STOP:
				lru->cursor = le->list.next;
				return NULL;
			}
		}

		h = h->next;

		if (!no_sleep)
			cond_resched();
	}

	return NULL;
}

/*--------------------------------------------------------------*/

/*
 * Buffer state bits.
 */
#define B_READING	0
#define B_WRITING	1
#define B_DIRTY		2

/*
 * Describes how the block was allocated:
 * kmem_cache_alloc(), __get_free_pages() or vmalloc().
 * See the comment at alloc_buffer_data.
 */
enum data_mode {
	DATA_MODE_SLAB = 0,
	DATA_MODE_GET_FREE_PAGES = 1,
	DATA_MODE_VMALLOC = 2,
	DATA_MODE_LIMIT = 3
};

struct dm_buffer {
	/* protected by the locks in dm_buffer_cache */
	struct rb_node node;

	/* immutable, so don't need protecting */
	sector_t block;
	void *data;
	unsigned char data_mode;		/* DATA_MODE_* */

	/*
	 * These two fields are used in isolation, so do not need
	 * a surrounding lock.
	 */
	atomic_t hold_count;
	unsigned long last_accessed;

	/*
	 * Everything else is protected by the mutex in
	 * dm_bufio_client
	 */
	unsigned long state;
	struct lru_entry lru;
	unsigned char list_mode;		/* LIST_* */
	blk_status_t read_error;
	blk_status_t write_error;
	unsigned int dirty_start;
	unsigned int dirty_end;
	unsigned int write_start;
	unsigned int write_end;
	struct list_head write_list;
	struct dm_bufio_client *c;
	void (*end_io)(struct dm_buffer *b, blk_status_t bs);
#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
#define MAX_STACK 10
	unsigned int stack_len;
	unsigned long stack_entries[MAX_STACK];
#endif
};

/*--------------------------------------------------------------*/

/*
 * The buffer cache manages buffers, particularly:
 *  - inc/dec of holder count
 *  - setting the last_accessed field
 *  - maintains clean/dirty state along with lru
 *  - selecting buffers that match predicates
 *
 * It does *not* handle:
 *  - allocation/freeing of buffers.
 *  - IO
 *  - Eviction or cache sizing.
 *
 * cache_get() and cache_put() are threadsafe, you do not need to
 * protect these calls with a surrounding mutex.  All the other
 * methods are not threadsafe; they do use locking primitives, but
 * only enough to ensure get/put are threadsafe.
 */

struct buffer_tree {
	union {
		struct rw_semaphore lock;
		rwlock_t spinlock;
	} u;
	struct rb_root root;
} ____cacheline_aligned_in_smp;

struct dm_buffer_cache {
	struct lru lru[LIST_SIZE];
	/*
	 * We spread entries across multiple trees to reduce contention
	 * on the locks.
	 */
	unsigned int num_locks;
	bool no_sleep;
	struct buffer_tree trees[];
};

static DEFINE_STATIC_KEY_FALSE(no_sleep_enabled);

static inline unsigned int cache_index(sector_t block, unsigned int num_locks)
{
	return dm_hash_locks_index(block, num_locks);
}

static inline void cache_read_lock(struct dm_buffer_cache *bc, sector_t block)
{
	if (static_branch_unlikely(&no_sleep_enabled) && bc->no_sleep)
		read_lock_bh(&bc->trees[cache_index(block, bc->num_locks)].u.spinlock);
	else
		down_read(&bc->trees[cache_index(block, bc->num_locks)].u.lock);
}

static inline void cache_read_unlock(struct dm_buffer_cache *bc, sector_t block)
{
	if (static_branch_unlikely(&no_sleep_enabled) && bc->no_sleep)
		read_unlock_bh(&bc->trees[cache_index(block, bc->num_locks)].u.spinlock);
	else
		up_read(&bc->trees[cache_index(block, bc->num_locks)].u.lock);
}

static inline void cache_write_lock(struct dm_buffer_cache *bc, sector_t block)
{
	if (static_branch_unlikely(&no_sleep_enabled) && bc->no_sleep)
		write_lock_bh(&bc->trees[cache_index(block, bc->num_locks)].u.spinlock);
	else
		down_write(&bc->trees[cache_index(block, bc->num_locks)].u.lock);
}

static inline void cache_write_unlock(struct dm_buffer_cache *bc, sector_t block)
{
	if (static_branch_unlikely(&no_sleep_enabled) && bc->no_sleep)
		write_unlock_bh(&bc->trees[cache_index(block, bc->num_locks)].u.spinlock);
	else
		up_write(&bc->trees[cache_index(block, bc->num_locks)].u.lock);
}

/*
 * Sometimes we want to repeatedly get and drop locks as part of an iteration.
 * This struct helps avoid redundant drop and gets of the same lock.
 */
struct lock_history {
	struct dm_buffer_cache *cache;
	bool write;
	unsigned int previous;
	unsigned int no_previous;
};

static void lh_init(struct lock_history *lh, struct dm_buffer_cache *cache, bool write)
{
	lh->cache = cache;
	lh->write = write;
	lh->no_previous = cache->num_locks;
	lh->previous = lh->no_previous;
}

static void __lh_lock(struct lock_history *lh, unsigned int index)
{
	if (lh->write) {
		if (static_branch_unlikely(&no_sleep_enabled) && lh->cache->no_sleep)
			write_lock_bh(&lh->cache->trees[index].u.spinlock);
		else
			down_write(&lh->cache->trees[index].u.lock);
	} else {
		if (static_branch_unlikely(&no_sleep_enabled) && lh->cache->no_sleep)
			read_lock_bh(&lh->cache->trees[index].u.spinlock);
		else
			down_read(&lh->cache->trees[index].u.lock);
	}
}

static void __lh_unlock(struct lock_history *lh, unsigned int index)
{
	if (lh->write) {
		if (static_branch_unlikely(&no_sleep_enabled) && lh->cache->no_sleep)
			write_unlock_bh(&lh->cache->trees[index].u.spinlock);
		else
			up_write(&lh->cache->trees[index].u.lock);
	} else {
		if (static_branch_unlikely(&no_sleep_enabled) && lh->cache->no_sleep)
			read_unlock_bh(&lh->cache->trees[index].u.spinlock);
		else
			up_read(&lh->cache->trees[index].u.lock);
	}
}

/*
 * Make sure you call this since it will unlock the final lock.
 */
static void lh_exit(struct lock_history *lh)
{
	if (lh->previous != lh->no_previous) {
		__lh_unlock(lh, lh->previous);
		lh->previous = lh->no_previous;
	}
}

/*
 * Named 'next' because there is no corresponding
 * 'up/unlock' call since it's done automatically.
 */
static void lh_next(struct lock_history *lh, sector_t b)
{
	unsigned int index = cache_index(b, lh->no_previous); /* no_previous is num_locks */

	if (lh->previous != lh->no_previous) {
		if (lh->previous != index) {
			__lh_unlock(lh, lh->previous);
			__lh_lock(lh, index);
			lh->previous = index;
		}
	} else {
		__lh_lock(lh, index);
		lh->previous = index;
	}
}

static inline struct dm_buffer *le_to_buffer(struct lru_entry *le)
{
	return container_of(le, struct dm_buffer, lru);
}

static struct dm_buffer *list_to_buffer(struct list_head *l)
{
	struct lru_entry *le = list_entry(l, struct lru_entry, list);

	if (!le)
		return NULL;

	return le_to_buffer(le);
}

static void cache_init(struct dm_buffer_cache *bc, unsigned int num_locks, bool no_sleep)
{
	unsigned int i;

	bc->num_locks = num_locks;
	bc->no_sleep = no_sleep;

	for (i = 0; i < bc->num_locks; i++) {
		if (no_sleep)
			rwlock_init(&bc->trees[i].u.spinlock);
		else
			init_rwsem(&bc->trees[i].u.lock);
		bc->trees[i].root = RB_ROOT;
	}

	lru_init(&bc->lru[LIST_CLEAN]);
	lru_init(&bc->lru[LIST_DIRTY]);
}

static void cache_destroy(struct dm_buffer_cache *bc)
{
	unsigned int i;

	for (i = 0; i < bc->num_locks; i++)
		WARN_ON_ONCE(!RB_EMPTY_ROOT(&bc->trees[i].root));

	lru_destroy(&bc->lru[LIST_CLEAN]);
	lru_destroy(&bc->lru[LIST_DIRTY]);
}

/*--------------*/

/*
 * not threadsafe, or racey depending how you look at it
 */
static inline unsigned long cache_count(struct dm_buffer_cache *bc, int list_mode)
{
	return bc->lru[list_mode].count;
}

static inline unsigned long cache_total(struct dm_buffer_cache *bc)
{
	return cache_count(bc, LIST_CLEAN) + cache_count(bc, LIST_DIRTY);
}

/*--------------*/

/*
 * Gets a specific buffer, indexed by block.
 * If the buffer is found then its holder count will be incremented and
 * lru_reference will be called.
 *
 * threadsafe
 */
static struct dm_buffer *__cache_get(const struct rb_root *root, sector_t block)
{
	struct rb_node *n = root->rb_node;
	struct dm_buffer *b;

	while (n) {
		b = container_of(n, struct dm_buffer, node);

		if (b->block == block)
			return b;

		n = block < b->block ? n->rb_left : n->rb_right;
	}

	return NULL;
}

static void __cache_inc_buffer(struct dm_buffer *b)
{
	atomic_inc(&b->hold_count);
	WRITE_ONCE(b->last_accessed, jiffies);
}

static struct dm_buffer *cache_get(struct dm_buffer_cache *bc, sector_t block)
{
	struct dm_buffer *b;

	cache_read_lock(bc, block);
	b = __cache_get(&bc->trees[cache_index(block, bc->num_locks)].root, block);
	if (b) {
		lru_reference(&b->lru);
		__cache_inc_buffer(b);
	}
	cache_read_unlock(bc, block);

	return b;
}

/*--------------*/

/*
 * Returns true if the hold count hits zero.
 * threadsafe
 */
static bool cache_put(struct dm_buffer_cache *bc, struct dm_buffer *b)
{
	bool r;

	cache_read_lock(bc, b->block);
	BUG_ON(!atomic_read(&b->hold_count));
	r = atomic_dec_and_test(&b->hold_count);
	cache_read_unlock(bc, b->block);

	return r;
}

/*--------------*/

typedef enum evict_result (*b_predicate)(struct dm_buffer *, void *);

/*
 * Evicts a buffer based on a predicate.  The oldest buffer that
 * matches the predicate will be selected.  In addition to the
 * predicate the hold_count of the selected buffer will be zero.
 */
struct evict_wrapper {
	struct lock_history *lh;
	b_predicate pred;
	void *context;
};

/*
 * Wraps the buffer predicate turning it into an lru predicate.  Adds
 * extra test for hold_count.
 */
static enum evict_result __evict_pred(struct lru_entry *le, void *context)
{
	struct evict_wrapper *w = context;
	struct dm_buffer *b = le_to_buffer(le);

	lh_next(w->lh, b->block);

	if (atomic_read(&b->hold_count))
		return ER_DONT_EVICT;

	return w->pred(b, w->context);
}

static struct dm_buffer *__cache_evict(struct dm_buffer_cache *bc, int list_mode,
				       b_predicate pred, void *context,
				       struct lock_history *lh)
{
	struct evict_wrapper w = {.lh = lh, .pred = pred, .context = context};
	struct lru_entry *le;
	struct dm_buffer *b;

	le = lru_evict(&bc->lru[list_mode], __evict_pred, &w, bc->no_sleep);
	if (!le)
		return NULL;

	b = le_to_buffer(le);
	/* __evict_pred will have locked the appropriate tree. */
	rb_erase(&b->node, &bc->trees[cache_index(b->block, bc->num_locks)].root);

	return b;
}

static struct dm_buffer *cache_evict(struct dm_buffer_cache *bc, int list_mode,
				     b_predicate pred, void *context)
{
	struct dm_buffer *b;
	struct lock_history lh;

	lh_init(&lh, bc, true);
	b = __cache_evict(bc, list_mode, pred, context, &lh);
	lh_exit(&lh);

	return b;
}

/*--------------*/

/*
 * Mark a buffer as clean or dirty. Not threadsafe.
 */
static void cache_mark(struct dm_buffer_cache *bc, struct dm_buffer *b, int list_mode)
{
	cache_write_lock(bc, b->block);
	if (list_mode != b->list_mode) {
		lru_remove(&bc->lru[b->list_mode], &b->lru);
		b->list_mode = list_mode;
		lru_insert(&bc->lru[b->list_mode], &b->lru);
	}
	cache_write_unlock(bc, b->block);
}

/*--------------*/

/*
 * Runs through the lru associated with 'old_mode', if the predicate matches then
 * it moves them to 'new_mode'.  Not threadsafe.
 */
static void __cache_mark_many(struct dm_buffer_cache *bc, int old_mode, int new_mode,
			      b_predicate pred, void *context, struct lock_history *lh)
{
	struct lru_entry *le;
	struct dm_buffer *b;
	struct evict_wrapper w = {.lh = lh, .pred = pred, .context = context};

	while (true) {
		le = lru_evict(&bc->lru[old_mode], __evict_pred, &w, bc->no_sleep);
		if (!le)
			break;

		b = le_to_buffer(le);
		b->list_mode = new_mode;
		lru_insert(&bc->lru[b->list_mode], &b->lru);
	}
}

static void cache_mark_many(struct dm_buffer_cache *bc, int old_mode, int new_mode,
			    b_predicate pred, void *context)
{
	struct lock_history lh;

	lh_init(&lh, bc, true);
	__cache_mark_many(bc, old_mode, new_mode, pred, context, &lh);
	lh_exit(&lh);
}

/*--------------*/

/*
 * Iterates through all clean or dirty entries calling a function for each
 * entry.  The callback may terminate the iteration early.  Not threadsafe.
 */

/*
 * Iterator functions should return one of these actions to indicate
 * how the iteration should proceed.
 */
enum it_action {
	IT_NEXT,
	IT_COMPLETE,
};

typedef enum it_action (*iter_fn)(struct dm_buffer *b, void *context);

static void __cache_iterate(struct dm_buffer_cache *bc, int list_mode,
			    iter_fn fn, void *context, struct lock_history *lh)
{
	struct lru *lru = &bc->lru[list_mode];
	struct lru_entry *le, *first;

	if (!lru->cursor)
		return;

	first = le = to_le(lru->cursor);
	do {
		struct dm_buffer *b = le_to_buffer(le);

		lh_next(lh, b->block);

		switch (fn(b, context)) {
		case IT_NEXT:
			break;

		case IT_COMPLETE:
			return;
		}
		cond_resched();

		le = to_le(le->list.next);
	} while (le != first);
}

static void cache_iterate(struct dm_buffer_cache *bc, int list_mode,
			  iter_fn fn, void *context)
{
	struct lock_history lh;

	lh_init(&lh, bc, false);
	__cache_iterate(bc, list_mode, fn, context, &lh);
	lh_exit(&lh);
}

/*--------------*/

/*
 * Passes ownership of the buffer to the cache. Returns false if the
 * buffer was already present (in which case ownership does not pass).
 * eg, a race with another thread.
 *
 * Holder count should be 1 on insertion.
 *
 * Not threadsafe.
 */
static bool __cache_insert(struct rb_root *root, struct dm_buffer *b)
{
	struct rb_node **new = &root->rb_node, *parent = NULL;
	struct dm_buffer *found;

	while (*new) {
		found = container_of(*new, struct dm_buffer, node);

		if (found->block == b->block)
			return false;

		parent = *new;
		new = b->block < found->block ?
			&found->node.rb_left : &found->node.rb_right;
	}

	rb_link_node(&b->node, parent, new);
	rb_insert_color(&b->node, root);

	return true;
}

static bool cache_insert(struct dm_buffer_cache *bc, struct dm_buffer *b)
{
	bool r;

	if (WARN_ON_ONCE(b->list_mode >= LIST_SIZE))
		return false;

	cache_write_lock(bc, b->block);
	BUG_ON(atomic_read(&b->hold_count) != 1);
	r = __cache_insert(&bc->trees[cache_index(b->block, bc->num_locks)].root, b);
	if (r)
		lru_insert(&bc->lru[b->list_mode], &b->lru);
	cache_write_unlock(bc, b->block);

	return r;
}

/*--------------*/

/*
 * Removes buffer from cache, ownership of the buffer passes back to the caller.
 * Fails if the hold_count is not one (ie. the caller holds the only reference).
 *
 * Not threadsafe.
 */
static bool cache_remove(struct dm_buffer_cache *bc, struct dm_buffer *b)
{
	bool r;

	cache_write_lock(bc, b->block);

	if (atomic_read(&b->hold_count) != 1) {
		r = false;
	} else {
		r = true;
		rb_erase(&b->node, &bc->trees[cache_index(b->block, bc->num_locks)].root);
		lru_remove(&bc->lru[b->list_mode], &b->lru);
	}

	cache_write_unlock(bc, b->block);

	return r;
}

/*--------------*/

typedef void (*b_release)(struct dm_buffer *);

static struct dm_buffer *__find_next(struct rb_root *root, sector_t block)
{
	struct rb_node *n = root->rb_node;
	struct dm_buffer *b;
	struct dm_buffer *best = NULL;

	while (n) {
		b = container_of(n, struct dm_buffer, node);

		if (b->block == block)
			return b;

		if (block <= b->block) {
			n = n->rb_left;
			best = b;
		} else {
			n = n->rb_right;
		}
	}

	return best;
}

static void __remove_range(struct dm_buffer_cache *bc,
			   struct rb_root *root,
			   sector_t begin, sector_t end,
			   b_predicate pred, b_release release)
{
	struct dm_buffer *b;

	while (true) {
		cond_resched();

		b = __find_next(root, begin);
		if (!b || (b->block >= end))
			break;

		begin = b->block + 1;

		if (atomic_read(&b->hold_count))
			continue;

		if (pred(b, NULL) == ER_EVICT) {
			rb_erase(&b->node, root);
			lru_remove(&bc->lru[b->list_mode], &b->lru);
			release(b);
		}
	}
}

static void cache_remove_range(struct dm_buffer_cache *bc,
			       sector_t begin, sector_t end,
			       b_predicate pred, b_release release)
{
	unsigned int i;

	BUG_ON(bc->no_sleep);
	for (i = 0; i < bc->num_locks; i++) {
		down_write(&bc->trees[i].u.lock);
		__remove_range(bc, &bc->trees[i].root, begin, end, pred, release);
		up_write(&bc->trees[i].u.lock);
	}
}

/*----------------------------------------------------------------*/

/*
 * Linking of buffers:
 *	All buffers are linked to buffer_cache with their node field.
 *
 *	Clean buffers that are not being written (B_WRITING not set)
 *	are linked to lru[LIST_CLEAN] with their lru_list field.
 *
 *	Dirty and clean buffers that are being written are linked to
 *	lru[LIST_DIRTY] with their lru_list field. When the write
 *	finishes, the buffer cannot be relinked immediately (because we
 *	are in an interrupt context and relinking requires process
 *	context), so some clean-not-writing buffers can be held on
 *	dirty_lru too.  They are later added to lru in the process
 *	context.
 */
struct dm_bufio_client {
	struct block_device *bdev;
	unsigned int block_size;
	s8 sectors_per_block_bits;

	bool no_sleep;
	struct mutex lock;
	spinlock_t spinlock;

	int async_write_error;

	void (*alloc_callback)(struct dm_buffer *buf);
	void (*write_callback)(struct dm_buffer *buf);
	struct kmem_cache *slab_buffer;
	struct kmem_cache *slab_cache;
	struct dm_io_client *dm_io;

	struct list_head reserved_buffers;
	unsigned int need_reserved_buffers;

	unsigned int minimum_buffers;

	sector_t start;

	struct shrinker *shrinker;
	struct work_struct shrink_work;
	atomic_long_t need_shrink;

	wait_queue_head_t free_buffer_wait;

	struct list_head client_list;

	/*
	 * Used by global_cleanup to sort the clients list.
	 */
	unsigned long oldest_buffer;

	struct dm_buffer_cache cache; /* must be last member */
};

/*----------------------------------------------------------------*/

#define dm_bufio_in_request()	(!!current->bio_list)

static void dm_bufio_lock(struct dm_bufio_client *c)
{
	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep)
		spin_lock_bh(&c->spinlock);
	else
		mutex_lock_nested(&c->lock, dm_bufio_in_request());
}

static void dm_bufio_unlock(struct dm_bufio_client *c)
{
	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep)
		spin_unlock_bh(&c->spinlock);
	else
		mutex_unlock(&c->lock);
}

/*----------------------------------------------------------------*/

/*
 * Default cache size: available memory divided by the ratio.
 */
static unsigned long dm_bufio_default_cache_size;

/*
 * Total cache size set by the user.
 */
static unsigned long dm_bufio_cache_size;

/*
 * A copy of dm_bufio_cache_size because dm_bufio_cache_size can change
 * at any time.  If it disagrees, the user has changed cache size.
 */
static unsigned long dm_bufio_cache_size_latch;

static DEFINE_SPINLOCK(global_spinlock);

/*
 * Buffers are freed after this timeout
 */
static unsigned int dm_bufio_max_age = DM_BUFIO_DEFAULT_AGE_SECS;
static unsigned long dm_bufio_retain_bytes = DM_BUFIO_DEFAULT_RETAIN_BYTES;

static unsigned long dm_bufio_peak_allocated;
static unsigned long dm_bufio_allocated_kmem_cache;
static unsigned long dm_bufio_allocated_get_free_pages;
static unsigned long dm_bufio_allocated_vmalloc;
static unsigned long dm_bufio_current_allocated;

/*----------------------------------------------------------------*/

/*
 * The current number of clients.
 */
static int dm_bufio_client_count;

/*
 * The list of all clients.
 */
static LIST_HEAD(dm_bufio_all_clients);

/*
 * This mutex protects dm_bufio_cache_size_latch and dm_bufio_client_count
 */
static DEFINE_MUTEX(dm_bufio_clients_lock);

static struct workqueue_struct *dm_bufio_wq;
static struct delayed_work dm_bufio_cleanup_old_work;
static struct work_struct dm_bufio_replacement_work;


#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
static void buffer_record_stack(struct dm_buffer *b)
{
	b->stack_len = stack_trace_save(b->stack_entries, MAX_STACK, 2);
}
#endif

/*----------------------------------------------------------------*/

static void adjust_total_allocated(struct dm_buffer *b, bool unlink)
{
	unsigned char data_mode;
	long diff;

	static unsigned long * const class_ptr[DATA_MODE_LIMIT] = {
		&dm_bufio_allocated_kmem_cache,
		&dm_bufio_allocated_get_free_pages,
		&dm_bufio_allocated_vmalloc,
	};

	data_mode = b->data_mode;
	diff = (long)b->c->block_size;
	if (unlink)
		diff = -diff;

	spin_lock(&global_spinlock);

	*class_ptr[data_mode] += diff;

	dm_bufio_current_allocated += diff;

	if (dm_bufio_current_allocated > dm_bufio_peak_allocated)
		dm_bufio_peak_allocated = dm_bufio_current_allocated;

	if (!unlink) {
		if (dm_bufio_current_allocated > dm_bufio_cache_size)
			queue_work(dm_bufio_wq, &dm_bufio_replacement_work);
	}

	spin_unlock(&global_spinlock);
}

/*
 * Change the number of clients and recalculate per-client limit.
 */
static void __cache_size_refresh(void)
{
	if (WARN_ON(!mutex_is_locked(&dm_bufio_clients_lock)))
		return;
	if (WARN_ON(dm_bufio_client_count < 0))
		return;

	dm_bufio_cache_size_latch = READ_ONCE(dm_bufio_cache_size);

	/*
	 * Use default if set to 0 and report the actual cache size used.
	 */
	if (!dm_bufio_cache_size_latch) {
		(void)cmpxchg(&dm_bufio_cache_size, 0,
			      dm_bufio_default_cache_size);
		dm_bufio_cache_size_latch = dm_bufio_default_cache_size;
	}
}

/*
 * Allocating buffer data.
 *
 * Small buffers are allocated with kmem_cache, to use space optimally.
 *
 * For large buffers, we choose between get_free_pages and vmalloc.
 * Each has advantages and disadvantages.
 *
 * __get_free_pages can randomly fail if the memory is fragmented.
 * __vmalloc won't randomly fail, but vmalloc space is limited (it may be
 * as low as 128M) so using it for caching is not appropriate.
 *
 * If the allocation may fail we use __get_free_pages. Memory fragmentation
 * won't have a fatal effect here, but it just causes flushes of some other
 * buffers and more I/O will be performed. Don't use __get_free_pages if it
 * always fails (i.e. order > MAX_PAGE_ORDER).
 *
 * If the allocation shouldn't fail we use __vmalloc. This is only for the
 * initial reserve allocation, so there's no risk of wasting all vmalloc
 * space.
 */
static void *alloc_buffer_data(struct dm_bufio_client *c, gfp_t gfp_mask,
			       unsigned char *data_mode)
{
	if (unlikely(c->slab_cache != NULL)) {
		*data_mode = DATA_MODE_SLAB;
		return kmem_cache_alloc(c->slab_cache, gfp_mask);
	}

	if (c->block_size <= KMALLOC_MAX_SIZE &&
	    gfp_mask & __GFP_NORETRY) {
		*data_mode = DATA_MODE_GET_FREE_PAGES;
		return (void *)__get_free_pages(gfp_mask,
						c->sectors_per_block_bits - (PAGE_SHIFT - SECTOR_SHIFT));
	}

	*data_mode = DATA_MODE_VMALLOC;

	return __vmalloc(c->block_size, gfp_mask);
}

/*
 * Free buffer's data.
 */
static void free_buffer_data(struct dm_bufio_client *c,
			     void *data, unsigned char data_mode)
{
	switch (data_mode) {
	case DATA_MODE_SLAB:
		kmem_cache_free(c->slab_cache, data);
		break;

	case DATA_MODE_GET_FREE_PAGES:
		free_pages((unsigned long)data,
			   c->sectors_per_block_bits - (PAGE_SHIFT - SECTOR_SHIFT));
		break;

	case DATA_MODE_VMALLOC:
		vfree(data);
		break;

	default:
		DMCRIT("dm_bufio_free_buffer_data: bad data mode: %d",
		       data_mode);
		BUG();
	}
}

/*
 * Allocate buffer and its data.
 */
static struct dm_buffer *alloc_buffer(struct dm_bufio_client *c, gfp_t gfp_mask)
{
	struct dm_buffer *b = kmem_cache_alloc(c->slab_buffer, gfp_mask);

	if (!b)
		return NULL;

	b->c = c;

	b->data = alloc_buffer_data(c, gfp_mask, &b->data_mode);
	if (!b->data) {
		kmem_cache_free(c->slab_buffer, b);
		return NULL;
	}
	adjust_total_allocated(b, false);

#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
	b->stack_len = 0;
#endif
	return b;
}

/*
 * Free buffer and its data.
 */
static void free_buffer(struct dm_buffer *b)
{
	struct dm_bufio_client *c = b->c;

	adjust_total_allocated(b, true);
	free_buffer_data(c, b->data, b->data_mode);
	kmem_cache_free(c->slab_buffer, b);
}

/*
 *--------------------------------------------------------------------------
 * Submit I/O on the buffer.
 *
 * Bio interface is faster but it has some problems:
 *	the vector list is limited (increasing this limit increases
 *	memory-consumption per buffer, so it is not viable);
 *
 *	the memory must be direct-mapped, not vmalloced;
 *
 * If the buffer is small enough (up to DM_BUFIO_INLINE_VECS pages) and
 * it is not vmalloced, try using the bio interface.
 *
 * If the buffer is big, if it is vmalloced or if the underlying device
 * rejects the bio because it is too large, use dm-io layer to do the I/O.
 * The dm-io layer splits the I/O into multiple requests, avoiding the above
 * shortcomings.
 *--------------------------------------------------------------------------
 */

/*
 * dm-io completion routine. It just calls b->bio.bi_end_io, pretending
 * that the request was handled directly with bio interface.
 */
static void dmio_complete(unsigned long error, void *context)
{
	struct dm_buffer *b = context;

	b->end_io(b, unlikely(error != 0) ? BLK_STS_IOERR : 0);
}

static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
		     unsigned int n_sectors, unsigned int offset,
		     unsigned short ioprio)
{
	int r;
	struct dm_io_request io_req = {
		.bi_opf = op,
		.notify.fn = dmio_complete,
		.notify.context = b,
		.client = b->c->dm_io,
	};
	struct dm_io_region region = {
		.bdev = b->c->bdev,
		.sector = sector,
		.count = n_sectors,
	};

	if (b->data_mode != DATA_MODE_VMALLOC) {
		io_req.mem.type = DM_IO_KMEM;
		io_req.mem.ptr.addr = (char *)b->data + offset;
	} else {
		io_req.mem.type = DM_IO_VMA;
		io_req.mem.ptr.vma = (char *)b->data + offset;
	}

	r = dm_io(&io_req, 1, &region, NULL, ioprio);
	if (unlikely(r))
		b->end_io(b, errno_to_blk_status(r));
}

static void bio_complete(struct bio *bio)
{
	struct dm_buffer *b = bio->bi_private;
	blk_status_t status = bio->bi_status;

	bio_uninit(bio);
	kfree(bio);
	b->end_io(b, status);
}

static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
		    unsigned int n_sectors, unsigned int offset,
		    unsigned short ioprio)
{
	struct bio *bio;
	char *ptr;
	unsigned int len;

	bio = bio_kmalloc(1, GFP_NOWAIT | __GFP_NORETRY | __GFP_NOWARN);
	if (!bio) {
		use_dmio(b, op, sector, n_sectors, offset, ioprio);
		return;
	}
	bio_init(bio, b->c->bdev, bio->bi_inline_vecs, 1, op);
	bio->bi_iter.bi_sector = sector;
	bio->bi_end_io = bio_complete;
	bio->bi_private = b;
	bio->bi_ioprio = ioprio;

	ptr = (char *)b->data + offset;
	len = n_sectors << SECTOR_SHIFT;

	__bio_add_page(bio, virt_to_page(ptr), len, offset_in_page(ptr));

	submit_bio(bio);
}

static inline sector_t block_to_sector(struct dm_bufio_client *c, sector_t block)
{
	sector_t sector;

	if (likely(c->sectors_per_block_bits >= 0))
		sector = block << c->sectors_per_block_bits;
	else
		sector = block * (c->block_size >> SECTOR_SHIFT);
	sector += c->start;

	return sector;
}

static void submit_io(struct dm_buffer *b, enum req_op op, unsigned short ioprio,
		      void (*end_io)(struct dm_buffer *, blk_status_t))
{
	unsigned int n_sectors;
	sector_t sector;
	unsigned int offset, end;

	b->end_io = end_io;

	sector = block_to_sector(b->c, b->block);

	if (op != REQ_OP_WRITE) {
		n_sectors = b->c->block_size >> SECTOR_SHIFT;
		offset = 0;
	} else {
		if (b->c->write_callback)
			b->c->write_callback(b);
		offset = b->write_start;
		end = b->write_end;
		offset &= -DM_BUFIO_WRITE_ALIGN;
		end += DM_BUFIO_WRITE_ALIGN - 1;
		end &= -DM_BUFIO_WRITE_ALIGN;
		if (unlikely(end > b->c->block_size))
			end = b->c->block_size;

		sector += offset >> SECTOR_SHIFT;
		n_sectors = (end - offset) >> SECTOR_SHIFT;
	}

	if (b->data_mode != DATA_MODE_VMALLOC)
		use_bio(b, op, sector, n_sectors, offset, ioprio);
	else
		use_dmio(b, op, sector, n_sectors, offset, ioprio);
}

/*
 *--------------------------------------------------------------
 * Writing dirty buffers
 *--------------------------------------------------------------
 */

/*
 * The endio routine for write.
 *
 * Set the error, clear B_WRITING bit and wake anyone who was waiting on
 * it.
 */
static void write_endio(struct dm_buffer *b, blk_status_t status)
{
	b->write_error = status;
	if (unlikely(status)) {
		struct dm_bufio_client *c = b->c;

		(void)cmpxchg(&c->async_write_error, 0,
				blk_status_to_errno(status));
	}

	BUG_ON(!test_bit(B_WRITING, &b->state));

	smp_mb__before_atomic();
	clear_bit(B_WRITING, &b->state);
	smp_mb__after_atomic();

	wake_up_bit(&b->state, B_WRITING);
}

/*
 * Initiate a write on a dirty buffer, but don't wait for it.
 *
 * - If the buffer is not dirty, exit.
 * - If there some previous write going on, wait for it to finish (we can't
 *   have two writes on the same buffer simultaneously).
 * - Submit our write and don't wait on it. We set B_WRITING indicating
 *   that there is a write in progress.
 */
static void __write_dirty_buffer(struct dm_buffer *b,
				 struct list_head *write_list)
{
	if (!test_bit(B_DIRTY, &b->state))
		return;

	clear_bit(B_DIRTY, &b->state);
	wait_on_bit_lock_io(&b->state, B_WRITING, TASK_UNINTERRUPTIBLE);

	b->write_start = b->dirty_start;
	b->write_end = b->dirty_end;

	if (!write_list)
		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
	else
		list_add_tail(&b->write_list, write_list);
}

static void __flush_write_list(struct list_head *write_list)
{
	struct blk_plug plug;

	blk_start_plug(&plug);
	while (!list_empty(write_list)) {
		struct dm_buffer *b =
			list_entry(write_list->next, struct dm_buffer, write_list);
		list_del(&b->write_list);
		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
		cond_resched();
	}
	blk_finish_plug(&plug);
}

/*
 * Wait until any activity on the buffer finishes.  Possibly write the
 * buffer if it is dirty.  When this function finishes, there is no I/O
 * running on the buffer and the buffer is not dirty.
 */
static void __make_buffer_clean(struct dm_buffer *b)
{
	BUG_ON(atomic_read(&b->hold_count));

	/* smp_load_acquire() pairs with read_endio()'s smp_mb__before_atomic() */
	if (!smp_load_acquire(&b->state))	/* fast case */
		return;

	wait_on_bit_io(&b->state, B_READING, TASK_UNINTERRUPTIBLE);
	__write_dirty_buffer(b, NULL);
	wait_on_bit_io(&b->state, B_WRITING, TASK_UNINTERRUPTIBLE);
}

static enum evict_result is_clean(struct dm_buffer *b, void *context)
{
	struct dm_bufio_client *c = context;

	/* These should never happen */
	if (WARN_ON_ONCE(test_bit(B_WRITING, &b->state)))
		return ER_DONT_EVICT;
	if (WARN_ON_ONCE(test_bit(B_DIRTY, &b->state)))
		return ER_DONT_EVICT;
	if (WARN_ON_ONCE(b->list_mode != LIST_CLEAN))
		return ER_DONT_EVICT;

	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep &&
	    unlikely(test_bit(B_READING, &b->state)))
		return ER_DONT_EVICT;

	return ER_EVICT;
}

static enum evict_result is_dirty(struct dm_buffer *b, void *context)
{
	/* These should never happen */
	if (WARN_ON_ONCE(test_bit(B_READING, &b->state)))
		return ER_DONT_EVICT;
	if (WARN_ON_ONCE(b->list_mode != LIST_DIRTY))
		return ER_DONT_EVICT;

	return ER_EVICT;
}

/*
 * Find some buffer that is not held by anybody, clean it, unlink it and
 * return it.
 */
static struct dm_buffer *__get_unclaimed_buffer(struct dm_bufio_client *c)
{
	struct dm_buffer *b;

	b = cache_evict(&c->cache, LIST_CLEAN, is_clean, c);
	if (b) {
		/* this also waits for pending reads */
		__make_buffer_clean(b);
		return b;
	}

	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep)
		return NULL;

	b = cache_evict(&c->cache, LIST_DIRTY, is_dirty, NULL);
	if (b) {
		__make_buffer_clean(b);
		return b;
	}

	return NULL;
}

/*
 * Wait until some other threads free some buffer or release hold count on
 * some buffer.
 *
 * This function is entered with c->lock held, drops it and regains it
 * before exiting.
 */
static void __wait_for_free_buffer(struct dm_bufio_client *c)
{
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&c->free_buffer_wait, &wait);
	set_current_state(TASK_UNINTERRUPTIBLE);
	dm_bufio_unlock(c);

	/*
	 * It's possible to miss a wake up event since we don't always
	 * hold c->lock when wake_up is called.  So we have a timeout here,
	 * just in case.
	 */
	io_schedule_timeout(5 * HZ);

	remove_wait_queue(&c->free_buffer_wait, &wait);

	dm_bufio_lock(c);
}

enum new_flag {
	NF_FRESH = 0,
	NF_READ = 1,
	NF_GET = 2,
	NF_PREFETCH = 3
};

/*
 * Allocate a new buffer. If the allocation is not possible, wait until
 * some other thread frees a buffer.
 *
 * May drop the lock and regain it.
 */
static struct dm_buffer *__alloc_buffer_wait_no_callback(struct dm_bufio_client *c, enum new_flag nf)
{
	struct dm_buffer *b;
	bool tried_noio_alloc = false;

	/*
	 * dm-bufio is resistant to allocation failures (it just keeps
	 * one buffer reserved in cases all the allocations fail).
	 * So set flags to not try too hard:
	 *	GFP_NOWAIT: don't wait; if we need to sleep we'll release our
	 *		    mutex and wait ourselves.
	 *	__GFP_NORETRY: don't retry and rather return failure
	 *	__GFP_NOMEMALLOC: don't use emergency reserves
	 *	__GFP_NOWARN: don't print a warning in case of failure
	 *
	 * For debugging, if we set the cache size to 1, no new buffers will
	 * be allocated.
	 */
	while (1) {
		if (dm_bufio_cache_size_latch != 1) {
			b = alloc_buffer(c, GFP_NOWAIT | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
			if (b)
				return b;
		}

		if (nf == NF_PREFETCH)
			return NULL;

		if (dm_bufio_cache_size_latch != 1 && !tried_noio_alloc) {
			dm_bufio_unlock(c);
			b = alloc_buffer(c, GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
			dm_bufio_lock(c);
			if (b)
				return b;
			tried_noio_alloc = true;
		}

		if (!list_empty(&c->reserved_buffers)) {
			b = list_to_buffer(c->reserved_buffers.next);
			list_del(&b->lru.list);
			c->need_reserved_buffers++;

			return b;
		}

		b = __get_unclaimed_buffer(c);
		if (b)
			return b;

		__wait_for_free_buffer(c);
	}
}

static struct dm_buffer *__alloc_buffer_wait(struct dm_bufio_client *c, enum new_flag nf)
{
	struct dm_buffer *b = __alloc_buffer_wait_no_callback(c, nf);

	if (!b)
		return NULL;

	if (c->alloc_callback)
		c->alloc_callback(b);

	return b;
}

/*
 * Free a buffer and wake other threads waiting for free buffers.
 */
static void __free_buffer_wake(struct dm_buffer *b)
{
	struct dm_bufio_client *c = b->c;

	b->block = -1;
	if (!c->need_reserved_buffers)
		free_buffer(b);
	else {
		list_add(&b->lru.list, &c->reserved_buffers);
		c->need_reserved_buffers--;
	}

	/*
	 * We hold the bufio lock here, so no one can add entries to the
	 * wait queue anyway.
	 */
	if (unlikely(waitqueue_active(&c->free_buffer_wait)))
		wake_up(&c->free_buffer_wait);
}

static enum evict_result cleaned(struct dm_buffer *b, void *context)
{
	if (WARN_ON_ONCE(test_bit(B_READING, &b->state)))
		return ER_DONT_EVICT; /* should never happen */

	if (test_bit(B_DIRTY, &b->state) || test_bit(B_WRITING, &b->state))
		return ER_DONT_EVICT;
	else
		return ER_EVICT;
}

static void __move_clean_buffers(struct dm_bufio_client *c)
{
	cache_mark_many(&c->cache, LIST_DIRTY, LIST_CLEAN, cleaned, NULL);
}

struct write_context {
	int no_wait;
	struct list_head *write_list;
};

static enum it_action write_one(struct dm_buffer *b, void *context)
{
	struct write_context *wc = context;

	if (wc->no_wait && test_bit(B_WRITING, &b->state))
		return IT_COMPLETE;

	__write_dirty_buffer(b, wc->write_list);
	return IT_NEXT;
}

static void __write_dirty_buffers_async(struct dm_bufio_client *c, int no_wait,
					struct list_head *write_list)
{
	struct write_context wc = {.no_wait = no_wait, .write_list = write_list};

	__move_clean_buffers(c);
	cache_iterate(&c->cache, LIST_DIRTY, write_one, &wc);
}

/*
 * Check if we're over watermark.
 * If we are over threshold_buffers, start freeing buffers.
 * If we're over "limit_buffers", block until we get under the limit.
 */
static void __check_watermark(struct dm_bufio_client *c,
			      struct list_head *write_list)
{
	if (cache_count(&c->cache, LIST_DIRTY) >
	    cache_count(&c->cache, LIST_CLEAN) * DM_BUFIO_WRITEBACK_RATIO)
		__write_dirty_buffers_async(c, 1, write_list);
}

/*
 *--------------------------------------------------------------
 * Getting a buffer
 *--------------------------------------------------------------
 */

static void cache_put_and_wake(struct dm_bufio_client *c, struct dm_buffer *b)
{
	/*
	 * Relying on waitqueue_active() is racey, but we sleep
	 * with schedule_timeout anyway.
	 */
	if (cache_put(&c->cache, b) &&
	    unlikely(waitqueue_active(&c->free_buffer_wait)))
		wake_up(&c->free_buffer_wait);
}

/*
 * This assumes you have already checked the cache to see if the buffer
 * is already present (it will recheck after dropping the lock for allocation).
 */
static struct dm_buffer *__bufio_new(struct dm_bufio_client *c, sector_t block,
				     enum new_flag nf, int *need_submit,
				     struct list_head *write_list)
{
	struct dm_buffer *b, *new_b = NULL;

	*need_submit = 0;

	/* This can't be called with NF_GET */
	if (WARN_ON_ONCE(nf == NF_GET))
		return NULL;

	new_b = __alloc_buffer_wait(c, nf);
	if (!new_b)
		return NULL;

	/*
	 * We've had a period where the mutex was unlocked, so need to
	 * recheck the buffer tree.
	 */
	b = cache_get(&c->cache, block);
	if (b) {
		__free_buffer_wake(new_b);
		goto found_buffer;
	}

	__check_watermark(c, write_list);

	b = new_b;
	atomic_set(&b->hold_count, 1);
	WRITE_ONCE(b->last_accessed, jiffies);
	b->block = block;
	b->read_error = 0;
	b->write_error = 0;
	b->list_mode = LIST_CLEAN;

	if (nf == NF_FRESH)
		b->state = 0;
	else {
		b->state = 1 << B_READING;
		*need_submit = 1;
	}

	/*
	 * We mustn't insert into the cache until the B_READING state
	 * is set.  Otherwise another thread could get it and use
	 * it before it had been read.
	 */
	cache_insert(&c->cache, b);

	return b;

found_buffer:
	if (nf == NF_PREFETCH) {
		cache_put_and_wake(c, b);
		return NULL;
	}

	/*
	 * Note: it is essential that we don't wait for the buffer to be
	 * read if dm_bufio_get function is used. Both dm_bufio_get and
	 * dm_bufio_prefetch can be used in the driver request routine.
	 * If the user called both dm_bufio_prefetch and dm_bufio_get on
	 * the same buffer, it would deadlock if we waited.
	 */
	if (nf == NF_GET && unlikely(test_bit_acquire(B_READING, &b->state))) {
		cache_put_and_wake(c, b);
		return NULL;
	}

	return b;
}

/*
 * The endio routine for reading: set the error, clear the bit and wake up
 * anyone waiting on the buffer.
 */
static void read_endio(struct dm_buffer *b, blk_status_t status)
{
	b->read_error = status;

	BUG_ON(!test_bit(B_READING, &b->state));

	smp_mb__before_atomic();
	clear_bit(B_READING, &b->state);
	smp_mb__after_atomic();

	wake_up_bit(&b->state, B_READING);
}

/*
 * A common routine for dm_bufio_new and dm_bufio_read.  Operation of these
 * functions is similar except that dm_bufio_new doesn't read the
 * buffer from the disk (assuming that the caller overwrites all the data
 * and uses dm_bufio_mark_buffer_dirty to write new data back).
 */
static void *new_read(struct dm_bufio_client *c, sector_t block,
		      enum new_flag nf, struct dm_buffer **bp,
		      unsigned short ioprio)
{
	int need_submit = 0;
	struct dm_buffer *b;

	LIST_HEAD(write_list);

	*bp = NULL;

	/*
	 * Fast path, hopefully the block is already in the cache.  No need
	 * to get the client lock for this.
	 */
	b = cache_get(&c->cache, block);
	if (b) {
		if (nf == NF_PREFETCH) {
			cache_put_and_wake(c, b);
			return NULL;
		}

		/*
		 * Note: it is essential that we don't wait for the buffer to be
		 * read if dm_bufio_get function is used. Both dm_bufio_get and
		 * dm_bufio_prefetch can be used in the driver request routine.
		 * If the user called both dm_bufio_prefetch and dm_bufio_get on
		 * the same buffer, it would deadlock if we waited.
		 */
		if (nf == NF_GET && unlikely(test_bit_acquire(B_READING, &b->state))) {
			cache_put_and_wake(c, b);
			return NULL;
		}
	}

	if (!b) {
		if (nf == NF_GET)
			return NULL;

		dm_bufio_lock(c);
		b = __bufio_new(c, block, nf, &need_submit, &write_list);
		dm_bufio_unlock(c);
	}

#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
	if (b && (atomic_read(&b->hold_count) == 1))
		buffer_record_stack(b);
#endif

	__flush_write_list(&write_list);

	if (!b)
		return NULL;

	if (need_submit)
		submit_io(b, REQ_OP_READ, ioprio, read_endio);

	if (nf != NF_GET)	/* we already tested this condition above */
		wait_on_bit_io(&b->state, B_READING, TASK_UNINTERRUPTIBLE);

	if (b->read_error) {
		int error = blk_status_to_errno(b->read_error);

		dm_bufio_release(b);

		return ERR_PTR(error);
	}

	*bp = b;

	return b->data;
}

void *dm_bufio_get(struct dm_bufio_client *c, sector_t block,
		   struct dm_buffer **bp)
{
	return new_read(c, block, NF_GET, bp, IOPRIO_DEFAULT);
}
EXPORT_SYMBOL_GPL(dm_bufio_get);

static void *__dm_bufio_read(struct dm_bufio_client *c, sector_t block,
			struct dm_buffer **bp, unsigned short ioprio)
{
	if (WARN_ON_ONCE(dm_bufio_in_request()))
		return ERR_PTR(-EINVAL);

	return new_read(c, block, NF_READ, bp, ioprio);
}

void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
		    struct dm_buffer **bp)
{
	return __dm_bufio_read(c, block, bp, IOPRIO_DEFAULT);
}
EXPORT_SYMBOL_GPL(dm_bufio_read);

void *dm_bufio_read_with_ioprio(struct dm_bufio_client *c, sector_t block,
				struct dm_buffer **bp, unsigned short ioprio)
{
	return __dm_bufio_read(c, block, bp, ioprio);
}
EXPORT_SYMBOL_GPL(dm_bufio_read_with_ioprio);

void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
		   struct dm_buffer **bp)
{
	if (WARN_ON_ONCE(dm_bufio_in_request()))
		return ERR_PTR(-EINVAL);

	return new_read(c, block, NF_FRESH, bp, IOPRIO_DEFAULT);
}
EXPORT_SYMBOL_GPL(dm_bufio_new);

static void __dm_bufio_prefetch(struct dm_bufio_client *c,
			sector_t block, unsigned int n_blocks,
			unsigned short ioprio)
{
	struct blk_plug plug;

	LIST_HEAD(write_list);

	if (WARN_ON_ONCE(dm_bufio_in_request()))
		return; /* should never happen */

	blk_start_plug(&plug);

	for (; n_blocks--; block++) {
		int need_submit;
		struct dm_buffer *b;

		b = cache_get(&c->cache, block);
		if (b) {
			/* already in cache */
			cache_put_and_wake(c, b);
			continue;
		}

		dm_bufio_lock(c);
		b = __bufio_new(c, block, NF_PREFETCH, &need_submit,
				&write_list);
		if (unlikely(!list_empty(&write_list))) {
			dm_bufio_unlock(c);
			blk_finish_plug(&plug);
			__flush_write_list(&write_list);
			blk_start_plug(&plug);
			dm_bufio_lock(c);
		}
		if (unlikely(b != NULL)) {
			dm_bufio_unlock(c);

			if (need_submit)
				submit_io(b, REQ_OP_READ, ioprio, read_endio);
			dm_bufio_release(b);

			cond_resched();

			if (!n_blocks)
				goto flush_plug;
			dm_bufio_lock(c);
		}
		dm_bufio_unlock(c);
	}

flush_plug:
	blk_finish_plug(&plug);
}

void dm_bufio_prefetch(struct dm_bufio_client *c, sector_t block, unsigned int n_blocks)
{
	return __dm_bufio_prefetch(c, block, n_blocks, IOPRIO_DEFAULT);
}
EXPORT_SYMBOL_GPL(dm_bufio_prefetch);

void dm_bufio_prefetch_with_ioprio(struct dm_bufio_client *c, sector_t block,
				unsigned int n_blocks, unsigned short ioprio)
{
	return __dm_bufio_prefetch(c, block, n_blocks, ioprio);
}
EXPORT_SYMBOL_GPL(dm_bufio_prefetch_with_ioprio);

void dm_bufio_release(struct dm_buffer *b)
{
	struct dm_bufio_client *c = b->c;

	/*
	 * If there were errors on the buffer, and the buffer is not
	 * to be written, free the buffer. There is no point in caching
	 * invalid buffer.
	 */
	if ((b->read_error || b->write_error) &&
	    !test_bit_acquire(B_READING, &b->state) &&
	    !test_bit(B_WRITING, &b->state) &&
	    !test_bit(B_DIRTY, &b->state)) {
		dm_bufio_lock(c);

		/* cache remove can fail if there are other holders */
		if (cache_remove(&c->cache, b)) {
			__free_buffer_wake(b);
			dm_bufio_unlock(c);
			return;
		}

		dm_bufio_unlock(c);
	}

	cache_put_and_wake(c, b);
}
EXPORT_SYMBOL_GPL(dm_bufio_release);

void dm_bufio_mark_partial_buffer_dirty(struct dm_buffer *b,
					unsigned int start, unsigned int end)
{
	struct dm_bufio_client *c = b->c;

	BUG_ON(start >= end);
	BUG_ON(end > b->c->block_size);

	dm_bufio_lock(c);

	BUG_ON(test_bit(B_READING, &b->state));

	if (!test_and_set_bit(B_DIRTY, &b->state)) {
		b->dirty_start = start;
		b->dirty_end = end;
		cache_mark(&c->cache, b, LIST_DIRTY);
	} else {
		if (start < b->dirty_start)
			b->dirty_start = start;
		if (end > b->dirty_end)
			b->dirty_end = end;
	}

	dm_bufio_unlock(c);
}
EXPORT_SYMBOL_GPL(dm_bufio_mark_partial_buffer_dirty);

void dm_bufio_mark_buffer_dirty(struct dm_buffer *b)
{
	dm_bufio_mark_partial_buffer_dirty(b, 0, b->c->block_size);
}
EXPORT_SYMBOL_GPL(dm_bufio_mark_buffer_dirty);

void dm_bufio_write_dirty_buffers_async(struct dm_bufio_client *c)
{
	LIST_HEAD(write_list);

	if (WARN_ON_ONCE(dm_bufio_in_request()))
		return; /* should never happen */

	dm_bufio_lock(c);
	__write_dirty_buffers_async(c, 0, &write_list);
	dm_bufio_unlock(c);
	__flush_write_list(&write_list);
}
EXPORT_SYMBOL_GPL(dm_bufio_write_dirty_buffers_async);

/*
 * For performance, it is essential that the buffers are written asynchronously
 * and simultaneously (so that the block layer can merge the writes) and then
 * waited upon.
 *
 * Finally, we flush hardware disk cache.
 */
static bool is_writing(struct lru_entry *e, void *context)
{
	struct dm_buffer *b = le_to_buffer(e);

	return test_bit(B_WRITING, &b->state);
}

int dm_bufio_write_dirty_buffers(struct dm_bufio_client *c)
{
	int a, f;
	unsigned long nr_buffers;
	struct lru_entry *e;
	struct lru_iter it;

	LIST_HEAD(write_list);

	dm_bufio_lock(c);
	__write_dirty_buffers_async(c, 0, &write_list);
	dm_bufio_unlock(c);
	__flush_write_list(&write_list);
	dm_bufio_lock(c);

	nr_buffers = cache_count(&c->cache, LIST_DIRTY);
	lru_iter_begin(&c->cache.lru[LIST_DIRTY], &it);
	while ((e = lru_iter_next(&it, is_writing, c))) {
		struct dm_buffer *b = le_to_buffer(e);
		__cache_inc_buffer(b);

		BUG_ON(test_bit(B_READING, &b->state));

		if (nr_buffers) {
			nr_buffers--;
			dm_bufio_unlock(c);
			wait_on_bit_io(&b->state, B_WRITING, TASK_UNINTERRUPTIBLE);
			dm_bufio_lock(c);
		} else {
			wait_on_bit_io(&b->state, B_WRITING, TASK_UNINTERRUPTIBLE);
		}

		if (!test_bit(B_DIRTY, &b->state) && !test_bit(B_WRITING, &b->state))
			cache_mark(&c->cache, b, LIST_CLEAN);

		cache_put_and_wake(c, b);

		cond_resched();
	}
	lru_iter_end(&it);

	wake_up(&c->free_buffer_wait);
	dm_bufio_unlock(c);

	a = xchg(&c->async_write_error, 0);
	f = dm_bufio_issue_flush(c);
	if (a)
		return a;

	return f;
}
EXPORT_SYMBOL_GPL(dm_bufio_write_dirty_buffers);

/*
 * Use dm-io to send an empty barrier to flush the device.
 */
int dm_bufio_issue_flush(struct dm_bufio_client *c)
{
	struct dm_io_request io_req = {
		.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC,
		.mem.type = DM_IO_KMEM,
		.mem.ptr.addr = NULL,
		.client = c->dm_io,
	};
	struct dm_io_region io_reg = {
		.bdev = c->bdev,
		.sector = 0,
		.count = 0,
	};

	if (WARN_ON_ONCE(dm_bufio_in_request()))
		return -EINVAL;

	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
}
EXPORT_SYMBOL_GPL(dm_bufio_issue_flush);

/*
 * Use dm-io to send a discard request to flush the device.
 */
int dm_bufio_issue_discard(struct dm_bufio_client *c, sector_t block, sector_t count)
{
	struct dm_io_request io_req = {
		.bi_opf = REQ_OP_DISCARD | REQ_SYNC,
		.mem.type = DM_IO_KMEM,
		.mem.ptr.addr = NULL,
		.client = c->dm_io,
	};
	struct dm_io_region io_reg = {
		.bdev = c->bdev,
		.sector = block_to_sector(c, block),
		.count = block_to_sector(c, count),
	};

	if (WARN_ON_ONCE(dm_bufio_in_request()))
		return -EINVAL; /* discards are optional */

	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
}
EXPORT_SYMBOL_GPL(dm_bufio_issue_discard);

static bool forget_buffer(struct dm_bufio_client *c, sector_t block)
{
	struct dm_buffer *b;

	b = cache_get(&c->cache, block);
	if (b) {
		if (likely(!smp_load_acquire(&b->state))) {
			if (cache_remove(&c->cache, b))
				__free_buffer_wake(b);
			else
				cache_put_and_wake(c, b);
		} else {
			cache_put_and_wake(c, b);
		}
	}

	return b ? true : false;
}

/*
 * Free the given buffer.
 *
 * This is just a hint, if the buffer is in use or dirty, this function
 * does nothing.
 */
void dm_bufio_forget(struct dm_bufio_client *c, sector_t block)
{
	dm_bufio_lock(c);
	forget_buffer(c, block);
	dm_bufio_unlock(c);
}
EXPORT_SYMBOL_GPL(dm_bufio_forget);

static enum evict_result idle(struct dm_buffer *b, void *context)
{
	return b->state ? ER_DONT_EVICT : ER_EVICT;
}

void dm_bufio_forget_buffers(struct dm_bufio_client *c, sector_t block, sector_t n_blocks)
{
	dm_bufio_lock(c);
	cache_remove_range(&c->cache, block, block + n_blocks, idle, __free_buffer_wake);
	dm_bufio_unlock(c);
}
EXPORT_SYMBOL_GPL(dm_bufio_forget_buffers);

void dm_bufio_set_minimum_buffers(struct dm_bufio_client *c, unsigned int n)
{
	c->minimum_buffers = n;
}
EXPORT_SYMBOL_GPL(dm_bufio_set_minimum_buffers);

unsigned int dm_bufio_get_block_size(struct dm_bufio_client *c)
{
	return c->block_size;
}
EXPORT_SYMBOL_GPL(dm_bufio_get_block_size);

sector_t dm_bufio_get_device_size(struct dm_bufio_client *c)
{
	sector_t s = bdev_nr_sectors(c->bdev);

	if (s >= c->start)
		s -= c->start;
	else
		s = 0;
	if (likely(c->sectors_per_block_bits >= 0))
		s >>= c->sectors_per_block_bits;
	else
		sector_div(s, c->block_size >> SECTOR_SHIFT);
	return s;
}
EXPORT_SYMBOL_GPL(dm_bufio_get_device_size);

struct dm_io_client *dm_bufio_get_dm_io_client(struct dm_bufio_client *c)
{
	return c->dm_io;
}
EXPORT_SYMBOL_GPL(dm_bufio_get_dm_io_client);

sector_t dm_bufio_get_block_number(struct dm_buffer *b)
{
	return b->block;
}
EXPORT_SYMBOL_GPL(dm_bufio_get_block_number);

void *dm_bufio_get_block_data(struct dm_buffer *b)
{
	return b->data;
}
EXPORT_SYMBOL_GPL(dm_bufio_get_block_data);

void *dm_bufio_get_aux_data(struct dm_buffer *b)
{
	return b + 1;
}
EXPORT_SYMBOL_GPL(dm_bufio_get_aux_data);

struct dm_bufio_client *dm_bufio_get_client(struct dm_buffer *b)
{
	return b->c;
}
EXPORT_SYMBOL_GPL(dm_bufio_get_client);

static enum it_action warn_leak(struct dm_buffer *b, void *context)
{
	bool *warned = context;

	WARN_ON(!(*warned));
	*warned = true;
	DMERR("leaked buffer %llx, hold count %u, list %d",
	      (unsigned long long)b->block, atomic_read(&b->hold_count), b->list_mode);
#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
	stack_trace_print(b->stack_entries, b->stack_len, 1);
	/* mark unclaimed to avoid WARN_ON at end of drop_buffers() */
	atomic_set(&b->hold_count, 0);
#endif
	return IT_NEXT;
}

static void drop_buffers(struct dm_bufio_client *c)
{
	int i;
	struct dm_buffer *b;

	if (WARN_ON(dm_bufio_in_request()))
		return; /* should never happen */

	/*
	 * An optimization so that the buffers are not written one-by-one.
	 */
	dm_bufio_write_dirty_buffers_async(c);

	dm_bufio_lock(c);

	while ((b = __get_unclaimed_buffer(c)))
		__free_buffer_wake(b);

	for (i = 0; i < LIST_SIZE; i++) {
		bool warned = false;

		cache_iterate(&c->cache, i, warn_leak, &warned);
	}

#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
	while ((b = __get_unclaimed_buffer(c)))
		__free_buffer_wake(b);
#endif

	for (i = 0; i < LIST_SIZE; i++)
		WARN_ON(cache_count(&c->cache, i));

	dm_bufio_unlock(c);
}

static unsigned long get_retain_buffers(struct dm_bufio_client *c)
{
	unsigned long retain_bytes = READ_ONCE(dm_bufio_retain_bytes);

	if (likely(c->sectors_per_block_bits >= 0))
		retain_bytes >>= c->sectors_per_block_bits + SECTOR_SHIFT;
	else
		retain_bytes /= c->block_size;

	return retain_bytes;
}

static void __scan(struct dm_bufio_client *c)
{
	int l;
	struct dm_buffer *b;
	unsigned long freed = 0;
	unsigned long retain_target = get_retain_buffers(c);
	unsigned long count = cache_total(&c->cache);

	for (l = 0; l < LIST_SIZE; l++) {
		while (true) {
			if (count - freed <= retain_target)
				atomic_long_set(&c->need_shrink, 0);
			if (!atomic_long_read(&c->need_shrink))
				break;

			b = cache_evict(&c->cache, l,
					l == LIST_CLEAN ? is_clean : is_dirty, c);
			if (!b)
				break;

			__make_buffer_clean(b);
			__free_buffer_wake(b);

			atomic_long_dec(&c->need_shrink);
			freed++;
			cond_resched();
		}
	}
}

static void shrink_work(struct work_struct *w)
{
	struct dm_bufio_client *c = container_of(w, struct dm_bufio_client, shrink_work);

	dm_bufio_lock(c);
	__scan(c);
	dm_bufio_unlock(c);
}

static unsigned long dm_bufio_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
{
	struct dm_bufio_client *c;

	c = shrink->private_data;
	atomic_long_add(sc->nr_to_scan, &c->need_shrink);
	queue_work(dm_bufio_wq, &c->shrink_work);

	return sc->nr_to_scan;
}

static unsigned long dm_bufio_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
	struct dm_bufio_client *c = shrink->private_data;
	unsigned long count = cache_total(&c->cache);
	unsigned long retain_target = get_retain_buffers(c);
	unsigned long queued_for_cleanup = atomic_long_read(&c->need_shrink);

	if (unlikely(count < retain_target))
		count = 0;
	else
		count -= retain_target;

	if (unlikely(count < queued_for_cleanup))
		count = 0;
	else
		count -= queued_for_cleanup;

	return count;
}

/*
 * Create the buffering interface
 */
struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsigned int block_size,
					       unsigned int reserved_buffers, unsigned int aux_size,
					       void (*alloc_callback)(struct dm_buffer *),
					       void (*write_callback)(struct dm_buffer *),
					       unsigned int flags)
{
	int r;
	unsigned int num_locks;
	struct dm_bufio_client *c;
	char slab_name[27];

	if (!block_size || block_size & ((1 << SECTOR_SHIFT) - 1)) {
		DMERR("%s: block size not specified or is not multiple of 512b", __func__);
		r = -EINVAL;
		goto bad_client;
	}

	num_locks = dm_num_hash_locks();
	c = kzalloc(sizeof(*c) + (num_locks * sizeof(struct buffer_tree)), GFP_KERNEL);
	if (!c) {
		r = -ENOMEM;
		goto bad_client;
	}
	cache_init(&c->cache, num_locks, (flags & DM_BUFIO_CLIENT_NO_SLEEP) != 0);

	c->bdev = bdev;
	c->block_size = block_size;
	if (is_power_of_2(block_size))
		c->sectors_per_block_bits = __ffs(block_size) - SECTOR_SHIFT;
	else
		c->sectors_per_block_bits = -1;

	c->alloc_callback = alloc_callback;
	c->write_callback = write_callback;

	if (flags & DM_BUFIO_CLIENT_NO_SLEEP) {
		c->no_sleep = true;
		static_branch_inc(&no_sleep_enabled);
	}

	mutex_init(&c->lock);
	spin_lock_init(&c->spinlock);
	INIT_LIST_HEAD(&c->reserved_buffers);
	c->need_reserved_buffers = reserved_buffers;

	dm_bufio_set_minimum_buffers(c, DM_BUFIO_MIN_BUFFERS);

	init_waitqueue_head(&c->free_buffer_wait);
	c->async_write_error = 0;

	c->dm_io = dm_io_client_create();
	if (IS_ERR(c->dm_io)) {
		r = PTR_ERR(c->dm_io);
		goto bad_dm_io;
	}

	if (block_size <= KMALLOC_MAX_SIZE &&
	    (block_size < PAGE_SIZE || !is_power_of_2(block_size))) {
		unsigned int align = min(1U << __ffs(block_size), (unsigned int)PAGE_SIZE);

		snprintf(slab_name, sizeof(slab_name), "dm_bufio_cache-%u", block_size);
		c->slab_cache = kmem_cache_create(slab_name, block_size, align,
						  SLAB_RECLAIM_ACCOUNT, NULL);
		if (!c->slab_cache) {
			r = -ENOMEM;
			goto bad;
		}
	}
	if (aux_size)
		snprintf(slab_name, sizeof(slab_name), "dm_bufio_buffer-%u", aux_size);
	else
		snprintf(slab_name, sizeof(slab_name), "dm_bufio_buffer");
	c->slab_buffer = kmem_cache_create(slab_name, sizeof(struct dm_buffer) + aux_size,
					   0, SLAB_RECLAIM_ACCOUNT, NULL);
	if (!c->slab_buffer) {
		r = -ENOMEM;
		goto bad;
	}

	while (c->need_reserved_buffers) {
		struct dm_buffer *b = alloc_buffer(c, GFP_KERNEL);

		if (!b) {
			r = -ENOMEM;
			goto bad;
		}
		__free_buffer_wake(b);
	}

	INIT_WORK(&c->shrink_work, shrink_work);
	atomic_long_set(&c->need_shrink, 0);

	c->shrinker = shrinker_alloc(0, "dm-bufio:(%u:%u)",
				     MAJOR(bdev->bd_dev), MINOR(bdev->bd_dev));
	if (!c->shrinker) {
		r = -ENOMEM;
		goto bad;
	}

	c->shrinker->count_objects = dm_bufio_shrink_count;
	c->shrinker->scan_objects = dm_bufio_shrink_scan;
	c->shrinker->seeks = 1;
	c->shrinker->batch = 0;
	c->shrinker->private_data = c;

	shrinker_register(c->shrinker);

	mutex_lock(&dm_bufio_clients_lock);
	dm_bufio_client_count++;
	list_add(&c->client_list, &dm_bufio_all_clients);
	__cache_size_refresh();
	mutex_unlock(&dm_bufio_clients_lock);

	return c;

bad:
	while (!list_empty(&c->reserved_buffers)) {
		struct dm_buffer *b = list_to_buffer(c->reserved_buffers.next);

		list_del(&b->lru.list);
		free_buffer(b);
	}
	kmem_cache_destroy(c->slab_cache);
	kmem_cache_destroy(c->slab_buffer);
	dm_io_client_destroy(c->dm_io);
bad_dm_io:
	mutex_destroy(&c->lock);
	if (c->no_sleep)
		static_branch_dec(&no_sleep_enabled);
	kfree(c);
bad_client:
	return ERR_PTR(r);
}
EXPORT_SYMBOL_GPL(dm_bufio_client_create);

/*
 * Free the buffering interface.
 * It is required that there are no references on any buffers.
 */
void dm_bufio_client_destroy(struct dm_bufio_client *c)
{
	unsigned int i;

	drop_buffers(c);

	shrinker_free(c->shrinker);
	flush_work(&c->shrink_work);

	mutex_lock(&dm_bufio_clients_lock);

	list_del(&c->client_list);
	dm_bufio_client_count--;
	__cache_size_refresh();

	mutex_unlock(&dm_bufio_clients_lock);

	WARN_ON(c->need_reserved_buffers);

	while (!list_empty(&c->reserved_buffers)) {
		struct dm_buffer *b = list_to_buffer(c->reserved_buffers.next);

		list_del(&b->lru.list);
		free_buffer(b);
	}

	for (i = 0; i < LIST_SIZE; i++)
		if (cache_count(&c->cache, i))
			DMERR("leaked buffer count %d: %lu", i, cache_count(&c->cache, i));

	for (i = 0; i < LIST_SIZE; i++)
		WARN_ON(cache_count(&c->cache, i));

	cache_destroy(&c->cache);
	kmem_cache_destroy(c->slab_cache);
	kmem_cache_destroy(c->slab_buffer);
	dm_io_client_destroy(c->dm_io);
	mutex_destroy(&c->lock);
	if (c->no_sleep)
		static_branch_dec(&no_sleep_enabled);
	kfree(c);
}
EXPORT_SYMBOL_GPL(dm_bufio_client_destroy);

void dm_bufio_client_reset(struct dm_bufio_client *c)
{
	drop_buffers(c);
	flush_work(&c->shrink_work);
}
EXPORT_SYMBOL_GPL(dm_bufio_client_reset);

void dm_bufio_set_sector_offset(struct dm_bufio_client *c, sector_t start)
{
	c->start = start;
}
EXPORT_SYMBOL_GPL(dm_bufio_set_sector_offset);

/*--------------------------------------------------------------*/

static unsigned int get_max_age_hz(void)
{
	unsigned int max_age = READ_ONCE(dm_bufio_max_age);

	if (max_age > UINT_MAX / HZ)
		max_age = UINT_MAX / HZ;

	return max_age * HZ;
}

static bool older_than(struct dm_buffer *b, unsigned long age_hz)
{
	return time_after_eq(jiffies, READ_ONCE(b->last_accessed) + age_hz);
}

struct evict_params {
	gfp_t gfp;
	unsigned long age_hz;

	/*
	 * This gets updated with the largest last_accessed (ie. most
	 * recently used) of the evicted buffers.  It will not be reinitialised
	 * by __evict_many(), so you can use it across multiple invocations.
	 */
	unsigned long last_accessed;
};

/*
 * We may not be able to evict this buffer if IO pending or the client
 * is still using it.
 *
 * And if GFP_NOFS is used, we must not do any I/O because we hold
 * dm_bufio_clients_lock and we would risk deadlock if the I/O gets
 * rerouted to different bufio client.
 */
static enum evict_result select_for_evict(struct dm_buffer *b, void *context)
{
	struct evict_params *params = context;

	if (!(params->gfp & __GFP_FS) ||
	    (static_branch_unlikely(&no_sleep_enabled) && b->c->no_sleep)) {
		if (test_bit_acquire(B_READING, &b->state) ||
		    test_bit(B_WRITING, &b->state) ||
		    test_bit(B_DIRTY, &b->state))
			return ER_DONT_EVICT;
	}

	return older_than(b, params->age_hz) ? ER_EVICT : ER_STOP;
}

static unsigned long __evict_many(struct dm_bufio_client *c,
				  struct evict_params *params,
				  int list_mode, unsigned long max_count)
{
	unsigned long count;
	unsigned long last_accessed;
	struct dm_buffer *b;

	for (count = 0; count < max_count; count++) {
		b = cache_evict(&c->cache, list_mode, select_for_evict, params);
		if (!b)
			break;

		last_accessed = READ_ONCE(b->last_accessed);
		if (time_after_eq(params->last_accessed, last_accessed))
			params->last_accessed = last_accessed;

		__make_buffer_clean(b);
		__free_buffer_wake(b);

		cond_resched();
	}

	return count;
}

static void evict_old_buffers(struct dm_bufio_client *c, unsigned long age_hz)
{
	struct evict_params params = {.gfp = 0, .age_hz = age_hz, .last_accessed = 0};
	unsigned long retain = get_retain_buffers(c);
	unsigned long count;
	LIST_HEAD(write_list);

	dm_bufio_lock(c);

	__check_watermark(c, &write_list);
	if (unlikely(!list_empty(&write_list))) {
		dm_bufio_unlock(c);
		__flush_write_list(&write_list);
		dm_bufio_lock(c);
	}

	count = cache_total(&c->cache);
	if (count > retain)
		__evict_many(c, &params, LIST_CLEAN, count - retain);

	dm_bufio_unlock(c);
}

static void cleanup_old_buffers(void)
{
	unsigned long max_age_hz = get_max_age_hz();
	struct dm_bufio_client *c;

	mutex_lock(&dm_bufio_clients_lock);

	__cache_size_refresh();

	list_for_each_entry(c, &dm_bufio_all_clients, client_list)
		evict_old_buffers(c, max_age_hz);

	mutex_unlock(&dm_bufio_clients_lock);
}

static void work_fn(struct work_struct *w)
{
	cleanup_old_buffers();

	queue_delayed_work(dm_bufio_wq, &dm_bufio_cleanup_old_work,
			   DM_BUFIO_WORK_TIMER_SECS * HZ);
}

/*--------------------------------------------------------------*/

/*
 * Global cleanup tries to evict the oldest buffers from across _all_
 * the clients.  It does this by repeatedly evicting a few buffers from
 * the client that holds the oldest buffer.  It's approximate, but hopefully
 * good enough.
 */
static struct dm_bufio_client *__pop_client(void)
{
	struct list_head *h;

	if (list_empty(&dm_bufio_all_clients))
		return NULL;

	h = dm_bufio_all_clients.next;
	list_del(h);
	return container_of(h, struct dm_bufio_client, client_list);
}

/*
 * Inserts the client in the global client list based on its
 * 'oldest_buffer' field.
 */
static void __insert_client(struct dm_bufio_client *new_client)
{
	struct dm_bufio_client *c;
	struct list_head *h = dm_bufio_all_clients.next;

	while (h != &dm_bufio_all_clients) {
		c = container_of(h, struct dm_bufio_client, client_list);
		if (time_after_eq(c->oldest_buffer, new_client->oldest_buffer))
			break;
		h = h->next;
	}

	list_add_tail(&new_client->client_list, h);
}

static unsigned long __evict_a_few(unsigned long nr_buffers)
{
	unsigned long count;
	struct dm_bufio_client *c;
	struct evict_params params = {
		.gfp = GFP_KERNEL,
		.age_hz = 0,
		/* set to jiffies in case there are no buffers in this client */
		.last_accessed = jiffies
	};

	c = __pop_client();
	if (!c)
		return 0;

	dm_bufio_lock(c);
	count = __evict_many(c, &params, LIST_CLEAN, nr_buffers);
	dm_bufio_unlock(c);

	if (count)
		c->oldest_buffer = params.last_accessed;
	__insert_client(c);

	return count;
}

static void check_watermarks(void)
{
	LIST_HEAD(write_list);
	struct dm_bufio_client *c;

	mutex_lock(&dm_bufio_clients_lock);
	list_for_each_entry(c, &dm_bufio_all_clients, client_list) {
		dm_bufio_lock(c);
		__check_watermark(c, &write_list);
		dm_bufio_unlock(c);
	}
	mutex_unlock(&dm_bufio_clients_lock);

	__flush_write_list(&write_list);
}

static void evict_old(void)
{
	unsigned long threshold = dm_bufio_cache_size -
		dm_bufio_cache_size / DM_BUFIO_LOW_WATERMARK_RATIO;

	mutex_lock(&dm_bufio_clients_lock);
	while (dm_bufio_current_allocated > threshold) {
		if (!__evict_a_few(64))
			break;
		cond_resched();
	}
	mutex_unlock(&dm_bufio_clients_lock);
}

static void do_global_cleanup(struct work_struct *w)
{
	check_watermarks();
	evict_old();
}

/*
 *--------------------------------------------------------------
 * Module setup
 *--------------------------------------------------------------
 */

/*
 * This is called only once for the whole dm_bufio module.
 * It initializes memory limit.
 */
static int __init dm_bufio_init(void)
{
	__u64 mem;

	dm_bufio_allocated_kmem_cache = 0;
	dm_bufio_allocated_get_free_pages = 0;
	dm_bufio_allocated_vmalloc = 0;
	dm_bufio_current_allocated = 0;

	mem = (__u64)mult_frac(totalram_pages() - totalhigh_pages(),
			       DM_BUFIO_MEMORY_PERCENT, 100) << PAGE_SHIFT;

	if (mem > ULONG_MAX)
		mem = ULONG_MAX;

#ifdef CONFIG_MMU
	if (mem > mult_frac(VMALLOC_TOTAL, DM_BUFIO_VMALLOC_PERCENT, 100))
		mem = mult_frac(VMALLOC_TOTAL, DM_BUFIO_VMALLOC_PERCENT, 100);
#endif

	dm_bufio_default_cache_size = mem;

	mutex_lock(&dm_bufio_clients_lock);
	__cache_size_refresh();
	mutex_unlock(&dm_bufio_clients_lock);

	dm_bufio_wq = alloc_workqueue("dm_bufio_cache", WQ_MEM_RECLAIM, 0);
	if (!dm_bufio_wq)
		return -ENOMEM;

	INIT_DELAYED_WORK(&dm_bufio_cleanup_old_work, work_fn);
	INIT_WORK(&dm_bufio_replacement_work, do_global_cleanup);
	queue_delayed_work(dm_bufio_wq, &dm_bufio_cleanup_old_work,
			   DM_BUFIO_WORK_TIMER_SECS * HZ);

	return 0;
}

/*
 * This is called once when unloading the dm_bufio module.
 */
static void __exit dm_bufio_exit(void)
{
	int bug = 0;

	cancel_delayed_work_sync(&dm_bufio_cleanup_old_work);
	destroy_workqueue(dm_bufio_wq);

	if (dm_bufio_client_count) {
		DMCRIT("%s: dm_bufio_client_count leaked: %d",
			__func__, dm_bufio_client_count);
		bug = 1;
	}

	if (dm_bufio_current_allocated) {
		DMCRIT("%s: dm_bufio_current_allocated leaked: %lu",
			__func__, dm_bufio_current_allocated);
		bug = 1;
	}

	if (dm_bufio_allocated_get_free_pages) {
		DMCRIT("%s: dm_bufio_allocated_get_free_pages leaked: %lu",
		       __func__, dm_bufio_allocated_get_free_pages);
		bug = 1;
	}

	if (dm_bufio_allocated_vmalloc) {
		DMCRIT("%s: dm_bufio_vmalloc leaked: %lu",
		       __func__, dm_bufio_allocated_vmalloc);
		bug = 1;
	}

	WARN_ON(bug); /* leaks are not worth crashing the system */
}

module_init(dm_bufio_init)
module_exit(dm_bufio_exit)

module_param_named(max_cache_size_bytes, dm_bufio_cache_size, ulong, 0644);
MODULE_PARM_DESC(max_cache_size_bytes, "Size of metadata cache");

module_param_named(max_age_seconds, dm_bufio_max_age, uint, 0644);
MODULE_PARM_DESC(max_age_seconds, "Max age of a buffer in seconds");

module_param_named(retain_bytes, dm_bufio_retain_bytes, ulong, 0644);
MODULE_PARM_DESC(retain_bytes, "Try to keep at least this many bytes cached in memory");

module_param_named(peak_allocated_bytes, dm_bufio_peak_allocated, ulong, 0644);
MODULE_PARM_DESC(peak_allocated_bytes, "Tracks the maximum allocated memory");

module_param_named(allocated_kmem_cache_bytes, dm_bufio_allocated_kmem_cache, ulong, 0444);
MODULE_PARM_DESC(allocated_kmem_cache_bytes, "Memory allocated with kmem_cache_alloc");

module_param_named(allocated_get_free_pages_bytes, dm_bufio_allocated_get_free_pages, ulong, 0444);
MODULE_PARM_DESC(allocated_get_free_pages_bytes, "Memory allocated with get_free_pages");

module_param_named(allocated_vmalloc_bytes, dm_bufio_allocated_vmalloc, ulong, 0444);
MODULE_PARM_DESC(allocated_vmalloc_bytes, "Memory allocated with vmalloc");

module_param_named(current_allocated_bytes, dm_bufio_current_allocated, ulong, 0444);
MODULE_PARM_DESC(current_allocated_bytes, "Memory currently used by the cache");

MODULE_AUTHOR("Mikulas Patocka <dm-devel@lists.linux.dev>");
MODULE_DESCRIPTION(DM_NAME " buffered I/O library");
MODULE_LICENSE("GPL");