summaryrefslogtreecommitdiff
path: root/src/mem/cache/cache_impl.hh
blob: fed3eeab544e81dd76c72815f5357d8317fe59ab (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
/*
 * Copyright (c) 2010-2015 ARM Limited
 * All rights reserved.
 *
 * The license below extends only to copyright in the software and shall
 * not be construed as granting a license to any other intellectual
 * property including but not limited to intellectual property relating
 * to a hardware implementation of the functionality of the software
 * licensed hereunder.  You may use the software subject to the license
 * terms below provided that you ensure that this notice is replicated
 * unmodified and in its entirety in all distributions of the software,
 * modified or unmodified, in source code or in binary form.
 *
 * Copyright (c) 2002-2005 The Regents of The University of Michigan
 * Copyright (c) 2010,2015 Advanced Micro Devices, Inc.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are
 * met: redistributions of source code must retain the above copyright
 * notice, this list of conditions and the following disclaimer;
 * redistributions in binary form must reproduce the above copyright
 * notice, this list of conditions and the following disclaimer in the
 * documentation and/or other materials provided with the distribution;
 * neither the name of the copyright holders nor the names of its
 * contributors may be used to endorse or promote products derived from
 * this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Authors: Erik Hallnor
 *          Dave Greene
 *          Nathan Binkert
 *          Steve Reinhardt
 *          Ron Dreslinski
 *          Andreas Sandberg
 */

#ifndef __MEM_CACHE_CACHE_IMPL_HH__
#define __MEM_CACHE_CACHE_IMPL_HH__

/**
 * @file
 * Cache definitions.
 */

#include "base/misc.hh"
#include "base/types.hh"
#include "debug/Cache.hh"
#include "debug/CachePort.hh"
#include "debug/CacheTags.hh"
#include "mem/cache/prefetch/base.hh"
#include "mem/cache/blk.hh"
#include "mem/cache/cache.hh"
#include "mem/cache/mshr.hh"
#include "sim/sim_exit.hh"

template<class TagStore>
Cache<TagStore>::Cache(const Params *p)
    : BaseCache(p),
      tags(dynamic_cast<TagStore*>(p->tags)),
      prefetcher(p->prefetcher),
      doFastWrites(true),
      prefetchOnAccess(p->prefetch_on_access)
{
    tempBlock = new BlkType();
    tempBlock->data = new uint8_t[blkSize];

    cpuSidePort = new CpuSidePort(p->name + ".cpu_side", this,
                                  "CpuSidePort");
    memSidePort = new MemSidePort(p->name + ".mem_side", this,
                                  "MemSidePort");

    tags->setCache(this);
    if (prefetcher)
        prefetcher->setCache(this);
}

template<class TagStore>
Cache<TagStore>::~Cache()
{
    delete [] tempBlock->data;
    delete tempBlock;

    delete cpuSidePort;
    delete memSidePort;
}

template<class TagStore>
void
Cache<TagStore>::regStats()
{
    BaseCache::regStats();
}

template<class TagStore>
void
Cache<TagStore>::cmpAndSwap(BlkType *blk, PacketPtr pkt)
{
    assert(pkt->isRequest());

    uint64_t overwrite_val;
    bool overwrite_mem;
    uint64_t condition_val64;
    uint32_t condition_val32;

    int offset = tags->extractBlkOffset(pkt->getAddr());
    uint8_t *blk_data = blk->data + offset;

    assert(sizeof(uint64_t) >= pkt->getSize());

    overwrite_mem = true;
    // keep a copy of our possible write value, and copy what is at the
    // memory address into the packet
    pkt->writeData((uint8_t *)&overwrite_val);
    pkt->setData(blk_data);

    if (pkt->req->isCondSwap()) {
        if (pkt->getSize() == sizeof(uint64_t)) {
            condition_val64 = pkt->req->getExtraData();
            overwrite_mem = !std::memcmp(&condition_val64, blk_data,
                                         sizeof(uint64_t));
        } else if (pkt->getSize() == sizeof(uint32_t)) {
            condition_val32 = (uint32_t)pkt->req->getExtraData();
            overwrite_mem = !std::memcmp(&condition_val32, blk_data,
                                         sizeof(uint32_t));
        } else
            panic("Invalid size for conditional read/write\n");
    }

    if (overwrite_mem) {
        std::memcpy(blk_data, &overwrite_val, pkt->getSize());
        blk->status |= BlkDirty;
    }
}


template<class TagStore>
void
Cache<TagStore>::satisfyCpuSideRequest(PacketPtr pkt, BlkType *blk,
                                       bool deferred_response,
                                       bool pending_downgrade)
{
    assert(pkt->isRequest());

    assert(blk && blk->isValid());
    // Occasionally this is not true... if we are a lower-level cache
    // satisfying a string of Read and ReadEx requests from
    // upper-level caches, a Read will mark the block as shared but we
    // can satisfy a following ReadEx anyway since we can rely on the
    // Read requester(s) to have buffered the ReadEx snoop and to
    // invalidate their blocks after receiving them.
    // assert(!pkt->needsExclusive() || blk->isWritable());
    assert(pkt->getOffset(blkSize) + pkt->getSize() <= blkSize);

    // Check RMW operations first since both isRead() and
    // isWrite() will be true for them
    if (pkt->cmd == MemCmd::SwapReq) {
        cmpAndSwap(blk, pkt);
    } else if (pkt->isWrite() &&
               (!pkt->isWriteInvalidate() || isTopLevel)) {
        assert(blk->isWritable());
        // Write or WriteInvalidate at the first cache with block in Exclusive
        if (blk->checkWrite(pkt)) {
            pkt->writeDataToBlock(blk->data, blkSize);
        }
        // Always mark the line as dirty even if we are a failed
        // StoreCond so we supply data to any snoops that have
        // appended themselves to this cache before knowing the store
        // will fail.
        blk->status |= BlkDirty;
        DPRINTF(Cache, "%s for %s addr %#llx size %d (write)\n", __func__,
                pkt->cmdString(), pkt->getAddr(), pkt->getSize());
    } else if (pkt->isRead()) {
        if (pkt->isLLSC()) {
            blk->trackLoadLocked(pkt);
        }
        pkt->setDataFromBlock(blk->data, blkSize);
        if (pkt->getSize() == blkSize) {
            // special handling for coherent block requests from
            // upper-level caches
            if (pkt->needsExclusive()) {
                // if we have a dirty copy, make sure the recipient
                // keeps it marked dirty
                if (blk->isDirty()) {
                    pkt->assertMemInhibit();
                }
                // on ReadExReq we give up our copy unconditionally
                if (blk != tempBlock)
                    tags->invalidate(blk);
                blk->invalidate();
            } else if (blk->isWritable() && !pending_downgrade
                      && !pkt->sharedAsserted() && !pkt->req->isInstFetch()) {
                // we can give the requester an exclusive copy (by not
                // asserting shared line) on a read request if:
                // - we have an exclusive copy at this level (& below)
                // - we don't have a pending snoop from below
                //   signaling another read request
                // - no other cache above has a copy (otherwise it
                //   would have asseretd shared line on request)
                // - we are not satisfying an instruction fetch (this
                //   prevents dirty data in the i-cache)

                if (blk->isDirty()) {
                    // special considerations if we're owner:
                    if (!deferred_response && !isTopLevel) {
                        // if we are responding immediately and can
                        // signal that we're transferring ownership
                        // along with exclusivity, do so
                        pkt->assertMemInhibit();
                        blk->status &= ~BlkDirty;
                    } else {
                        // if we're responding after our own miss,
                        // there's a window where the recipient didn't
                        // know it was getting ownership and may not
                        // have responded to snoops correctly, so we
                        // can't pass off ownership *or* exclusivity
                        pkt->assertShared();
                    }
                }
            } else {
                // otherwise only respond with a shared copy
                pkt->assertShared();
            }
        }
    } else {
        // Upgrade or WriteInvalidate at a different cache than received it.
        // Since we have it Exclusively (E or M), we ack then invalidate.
        assert(pkt->isUpgrade() ||
               (pkt->isWriteInvalidate() && !isTopLevel));
        assert(blk != tempBlock);
        tags->invalidate(blk);
        blk->invalidate();
        DPRINTF(Cache, "%s for %s addr %#llx size %d (invalidation)\n",
                __func__, pkt->cmdString(), pkt->getAddr(), pkt->getSize());
    }
}


/////////////////////////////////////////////////////
//
// MSHR helper functions
//
/////////////////////////////////////////////////////


template<class TagStore>
void
Cache<TagStore>::markInService(MSHR *mshr, bool pending_dirty_resp)
{
    markInServiceInternal(mshr, pending_dirty_resp);
#if 0
        if (mshr->originalCmd == MemCmd::HardPFReq) {
            DPRINTF(HWPrefetch, "Marking a HW_PF in service\n");
            //Also clear pending if need be
            if (!prefetcher->havePending())
            {
                deassertMemSideBusRequest(Request_PF);
            }
        }
#endif
}


template<class TagStore>
void
Cache<TagStore>::squash(int threadNum)
{
    bool unblock = false;
    BlockedCause cause = NUM_BLOCKED_CAUSES;

    if (noTargetMSHR && noTargetMSHR->threadNum == threadNum) {
        noTargetMSHR = NULL;
        unblock = true;
        cause = Blocked_NoTargets;
    }
    if (mshrQueue.isFull()) {
        unblock = true;
        cause = Blocked_NoMSHRs;
    }
    mshrQueue.squash(threadNum);
    if (unblock && !mshrQueue.isFull()) {
        clearBlocked(cause);
    }
}

/////////////////////////////////////////////////////
//
// Access path: requests coming in from the CPU side
//
/////////////////////////////////////////////////////

template<class TagStore>
bool
Cache<TagStore>::access(PacketPtr pkt, BlkType *&blk,
                        Cycles &lat, PacketList &writebacks)
{
    // sanity check
    assert(pkt->isRequest());

    DPRINTF(Cache, "%s for %s addr %#llx size %d\n", __func__,
            pkt->cmdString(), pkt->getAddr(), pkt->getSize());
    if (pkt->req->isUncacheable()) {
        uncacheableFlush(pkt);
        blk = NULL;
        // lookupLatency is the latency in case the request is uncacheable.
        lat = lookupLatency;
        return false;
    }

    int id = pkt->req->hasContextId() ? pkt->req->contextId() : -1;
    // Here lat is the value passed as parameter to accessBlock() function
    // that can modify its value.
    blk = tags->accessBlock(pkt->getAddr(), pkt->isSecure(), lat, id);

    DPRINTF(Cache, "%s%s addr %#llx size %d (%s) %s\n", pkt->cmdString(),
            pkt->req->isInstFetch() ? " (ifetch)" : "",
            pkt->getAddr(), pkt->getSize(), pkt->isSecure() ? "s" : "ns",
            blk ? "hit " + blk->print() : "miss");

    // Writeback handling is special case.  We can write the block into
    // the cache without having a writeable copy (or any copy at all).
    if (pkt->cmd == MemCmd::Writeback) {
        assert(blkSize == pkt->getSize());
        if (blk == NULL) {
            // need to do a replacement
            blk = allocateBlock(pkt->getAddr(), pkt->isSecure(), writebacks);
            if (blk == NULL) {
                // no replaceable block available: give up, fwd to next level.
                incMissCount(pkt);
                return false;
            }
            tags->insertBlock(pkt, blk);

            blk->status = (BlkValid | BlkReadable);
            if (pkt->isSecure()) {
                blk->status |= BlkSecure;
            }
        }
        blk->status |= BlkDirty;
        if (pkt->isSupplyExclusive()) {
            blk->status |= BlkWritable;
        }
        // nothing else to do; writeback doesn't expect response
        assert(!pkt->needsResponse());
        std::memcpy(blk->data, pkt->getConstPtr<uint8_t>(), blkSize);
        DPRINTF(Cache, "%s new state is %s\n", __func__, blk->print());
        incHitCount(pkt);
        return true;
    } else if ((blk != NULL) &&
               (pkt->needsExclusive() ? blk->isWritable()
                                      : blk->isReadable())) {
        // OK to satisfy access
        incHitCount(pkt);
        satisfyCpuSideRequest(pkt, blk);
        return true;
    }

    // Can't satisfy access normally... either no block (blk == NULL)
    // or have block but need exclusive & only have shared.

    incMissCount(pkt);

    if (blk == NULL && pkt->isLLSC() && pkt->isWrite()) {
        // complete miss on store conditional... just give up now
        pkt->req->setExtraData(0);
        return true;
    }

    return false;
}


class ForwardResponseRecord : public Packet::SenderState
{
  public:

    ForwardResponseRecord() {}
};

template<class TagStore>
void
Cache<TagStore>::recvTimingSnoopResp(PacketPtr pkt)
{
    DPRINTF(Cache, "%s for %s addr %#llx size %d\n", __func__,
            pkt->cmdString(), pkt->getAddr(), pkt->getSize());

    assert(pkt->isResponse());

    // must be cache-to-cache response from upper to lower level
    ForwardResponseRecord *rec =
        dynamic_cast<ForwardResponseRecord *>(pkt->senderState);
    assert(!system->bypassCaches());

    if (rec == NULL) {
        // @todo What guarantee do we have that this HardPFResp is
        // actually for this cache, and not a cache closer to the
        // memory?
        assert(pkt->cmd == MemCmd::HardPFResp);
        // Check if it's a prefetch response and handle it. We shouldn't
        // get any other kinds of responses without FRRs.
        DPRINTF(Cache, "Got prefetch response from above for addr %#llx (%s)\n",
                pkt->getAddr(), pkt->isSecure() ? "s" : "ns");
        recvTimingResp(pkt);
        return;
    }

    pkt->popSenderState();
    delete rec;
    // forwardLatency is set here because there is a response from an
    // upper level cache.
    // To pay the delay that occurs if the packet comes from the bus,
    // we charge also headerDelay.
    Tick snoop_resp_time = clockEdge(forwardLatency) + pkt->headerDelay;
    // Reset the timing of the packet.
    pkt->headerDelay = pkt->payloadDelay = 0;
    memSidePort->schedTimingSnoopResp(pkt, snoop_resp_time);
}

template<class TagStore>
void
Cache<TagStore>::promoteWholeLineWrites(PacketPtr pkt)
{
    // Cache line clearing instructions
    if (doFastWrites && (pkt->cmd == MemCmd::WriteReq) &&
        (pkt->getSize() == blkSize) && (pkt->getOffset(blkSize) == 0)) {
        pkt->cmd = MemCmd::WriteInvalidateReq;
        DPRINTF(Cache, "packet promoted from Write to WriteInvalidate\n");
        assert(isTopLevel); // should only happen at L1 or I/O cache
    }
}

template<class TagStore>
bool
Cache<TagStore>::recvTimingReq(PacketPtr pkt)
{
    DPRINTF(CacheTags, "%s tags: %s\n", __func__, tags->print());
//@todo Add back in MemDebug Calls
//    MemDebug::cacheAccess(pkt);


    /// @todo temporary hack to deal with memory corruption issue until
    /// 4-phase transactions are complete
    for (int x = 0; x < pendingDelete.size(); x++)
        delete pendingDelete[x];
    pendingDelete.clear();

    assert(pkt->isRequest());

    // Just forward the packet if caches are disabled.
    if (system->bypassCaches()) {
        // @todo This should really enqueue the packet rather
        bool M5_VAR_USED success = memSidePort->sendTimingReq(pkt);
        assert(success);
        return true;
    }

    promoteWholeLineWrites(pkt);

    if (pkt->memInhibitAsserted()) {
        // a cache above us (but not where the packet came from) is
        // responding to the request
        DPRINTF(Cache, "mem inhibited on addr %#llx (%s): not responding\n",
                pkt->getAddr(), pkt->isSecure() ? "s" : "ns");
        assert(!pkt->req->isUncacheable());

        // if the packet needs exclusive, and the cache that has
        // promised to respond (setting the inhibit flag) is not
        // providing exclusive (it is in O vs M state), we know that
        // there may be other shared copies in the system; go out and
        // invalidate them all
        if (pkt->needsExclusive() && !pkt->isSupplyExclusive()) {
            // create a downstream express snoop with cleared packet
            // flags, there is no need to allocate any data as the
            // packet is merely used to co-ordinate state transitions
            Packet *snoop_pkt = new Packet(pkt, true, false);

            // also reset the bus time that the original packet has
            // not yet paid for
            snoop_pkt->headerDelay = snoop_pkt->payloadDelay = 0;

            // make this an instantaneous express snoop, and let the
            // other caches in the system know that the packet is
            // inhibited, because we have found the authorative copy
            // (O) that will supply the right data
            snoop_pkt->setExpressSnoop();
            snoop_pkt->assertMemInhibit();

            // this express snoop travels towards the memory, and at
            // every crossbar it is snooped upwards thus reaching
            // every cache in the system
            bool M5_VAR_USED success = memSidePort->sendTimingReq(snoop_pkt);
            // express snoops always succeed
            assert(success);

            // main memory will delete the packet
        }

        /// @todo nominally we should just delete the packet here,
        /// however, until 4-phase stuff we can't because sending
        /// cache is still relying on it
        pendingDelete.push_back(pkt);

        // no need to take any action in this particular cache as the
        // caches along the path to memory are allowed to keep lines
        // in a shared state, and a cache above us already committed
        // to responding
        return true;
    }

    if (pkt->req->isUncacheable()) {
        uncacheableFlush(pkt);

        // writes go in write buffer, reads use MSHR,
        // prefetches are acknowledged (responded to) and dropped
        if (pkt->cmd.isPrefetch()) {
            // prefetching (cache loading) uncacheable data is nonsensical
            pkt->makeTimingResponse();
            std::memset(pkt->getPtr<uint8_t>(), 0xFF, pkt->getSize());
            // We use lookupLatency here because the request is uncacheable.
            // We pay also for headerDelay that is charged of bus latencies if
            // the packet comes from the bus.
            Tick time = clockEdge(lookupLatency) + pkt->headerDelay;
            // Reset the timing of the packet.
            pkt->headerDelay = pkt->payloadDelay = 0;
            cpuSidePort->schedTimingResp(pkt, time);
            return true;
        } else if (pkt->isWrite() && !pkt->isRead()) {
            // We pay also for headerDelay that is charged of bus latencies if
            // the packet comes from the bus.
            Tick allocate_wr_buffer_time = clockEdge(forwardLatency) +
                                            pkt->headerDelay;
            // Reset the timing of the packet.
            pkt->headerDelay = pkt->payloadDelay = 0;
            allocateWriteBuffer(pkt, allocate_wr_buffer_time, true);
        } else {
            // We use forwardLatency here because there is an uncached
            // memory read, allocateded to MSHR queue (it requires the same
            // time of forwarding to WriteBuffer, in our assumption). It
            // specifies the latency to allocate an internal buffer and to
            // schedule an event to the queued port.
            // We pay also for headerDelay that is charged of bus latencies if
            // the packet comes from the bus.
            Tick allocate_rd_buffer_time = clockEdge(forwardLatency) +
                                            pkt->headerDelay;
            // Reset the timing of the packet.
            pkt->headerDelay = pkt->payloadDelay = 0;
            allocateUncachedReadBuffer(pkt, allocate_rd_buffer_time, true);
        }
        assert(pkt->needsResponse()); // else we should delete it here??
        return true;
    }

    // We use lookupLatency here because it is used to specify the latency
    // to access.
    Cycles lat = lookupLatency;
    BlkType *blk = NULL;
    PacketList writebacks;
    // Note that lat is passed by reference here. The function access() calls
    // accessBlock() which can modify lat value.
    bool satisfied = access(pkt, blk, lat, writebacks);
    // Here we charge the headerDelay that takes into account the latencies
    // of the bus, if the packet comes from it.
    // The latency charged it is just lat that is the value of lookupLatency
    // modified by access() function, or if not just lookupLatency.
    // In case of a hit we are neglecting response latency.
    // In case of a miss we are neglecting forward latency.
    Tick request_time = clockEdge(lat) + pkt->headerDelay;
    // Here we condiser forward_time, paying for just forward latency and
    // also charging the delay provided by the xbar.
    // forward_time is used in allocateWriteBuffer() function, called
    // in case of writeback.
    Tick forward_time = clockEdge(forwardLatency) + pkt->headerDelay;
    // Here we reset the timing of the packet.
    pkt->headerDelay = pkt->payloadDelay = 0;

    // track time of availability of next prefetch, if any
    Tick next_pf_time = MaxTick;

    bool needsResponse = pkt->needsResponse();

    if (satisfied) {
        // hit (for all other request types)

        if (prefetcher && (prefetchOnAccess || (blk && blk->wasPrefetched()))) {
            if (blk)
                blk->status &= ~BlkHWPrefetched;

            // Don't notify on SWPrefetch
            if (!pkt->cmd.isSWPrefetch())
                next_pf_time = prefetcher->notify(pkt);
        }

        if (needsResponse) {
            pkt->makeTimingResponse();
            // @todo: Make someone pay for this
            pkt->headerDelay = pkt->payloadDelay = 0;

            // In this case we are considering request_time that takes
            // into account the delay of the xbar, if any, and just
            // lat, neglecting responseLatency, modelling hit latency
            // just as lookupLatency or or the value of lat overriden
            // by access(), that calls accessBlock() function.
            cpuSidePort->schedTimingResp(pkt, request_time);
        } else {
            /// @todo nominally we should just delete the packet here,
            /// however, until 4-phase stuff we can't because sending
            /// cache is still relying on it
            pendingDelete.push_back(pkt);
        }
    } else {
        // miss

        Addr blk_addr = blockAlign(pkt->getAddr());
        MSHR *mshr = mshrQueue.findMatch(blk_addr, pkt->isSecure());

        // Software prefetch handling:
        // To keep the core from waiting on data it won't look at
        // anyway, send back a response with dummy data. Miss handling
        // will continue asynchronously. Unfortunately, the core will
        // insist upon freeing original Packet/Request, so we have to
        // create a new pair with a different lifecycle. Note that this
        // processing happens before any MSHR munging on the behalf of
        // this request because this new Request will be the one stored
        // into the MSHRs, not the original.
        if (pkt->cmd.isSWPrefetch() && isTopLevel) {
            assert(needsResponse);
            assert(pkt->req->hasPaddr());

            // There's no reason to add a prefetch as an additional target
            // to an existing MSHR. If an outstanding request is already
            // in progress, there is nothing for the prefetch to do.
            // If this is the case, we don't even create a request at all.
            PacketPtr pf = nullptr;

            if (!mshr) {
                // copy the request and create a new SoftPFReq packet
                RequestPtr req = new Request(pkt->req->getPaddr(),
                                             pkt->req->getSize(),
                                             pkt->req->getFlags(),
                                             pkt->req->masterId());
                pf = new Packet(req, pkt->cmd);
                pf->allocate();
                assert(pf->getAddr() == pkt->getAddr());
                assert(pf->getSize() == pkt->getSize());
            }

            pkt->makeTimingResponse();
            // for debugging, set all the bits in the response data
            // (also keeps valgrind from complaining when debugging settings
            //  print out instruction results)
            std::memset(pkt->getPtr<uint8_t>(), 0xFF, pkt->getSize());
            // request_time is used here, taking into account lat and the delay
            // charged if the packet comes from the xbar.
            cpuSidePort->schedTimingResp(pkt, request_time);

            // If an outstanding request is in progress (we found an
            // MSHR) this is set to null
            pkt = pf;
        }

        if (mshr) {
            /// MSHR hit
            /// @note writebacks will be checked in getNextMSHR()
            /// for any conflicting requests to the same block

            //@todo remove hw_pf here

            // Coalesce unless it was a software prefetch (see above).
            if (pkt) {
                DPRINTF(Cache, "%s coalescing MSHR for %s addr %#llx size %d\n",
                        __func__, pkt->cmdString(), pkt->getAddr(),
                        pkt->getSize());

                assert(pkt->req->masterId() < system->maxMasters());
                mshr_hits[pkt->cmdToIndex()][pkt->req->masterId()]++;
                if (mshr->threadNum != 0/*pkt->req->threadId()*/) {
                    mshr->threadNum = -1;
                }
                // We use forward_time here because it is the same
                // considering new targets. We have multiple requests for the
                // same address here. It specifies the latency to allocate an
                // internal buffer and to schedule an event to the queued
                // port and also takes into account the additional delay of
                // the xbar.
                mshr->allocateTarget(pkt, forward_time, order++);
                if (mshr->getNumTargets() == numTarget) {
                    noTargetMSHR = mshr;
                    setBlocked(Blocked_NoTargets);
                    // need to be careful with this... if this mshr isn't
                    // ready yet (i.e. time > curTick()), we don't want to
                    // move it ahead of mshrs that are ready
                    // mshrQueue.moveToFront(mshr);
                }

                // We should call the prefetcher reguardless if the request is
                // satisfied or not, reguardless if the request is in the MSHR or
                // not.  The request could be a ReadReq hit, but still not
                // satisfied (potentially because of a prior write to the same
                // cache line.  So, even when not satisfied, tehre is an MSHR
                // already allocated for this, we need to let the prefetcher know
                // about the request
                if (prefetcher) {
                    // Don't notify on SWPrefetch
                    if (!pkt->cmd.isSWPrefetch())
                        next_pf_time = prefetcher->notify(pkt);
                }
            }
        } else {
            // no MSHR
            assert(pkt->req->masterId() < system->maxMasters());
            mshr_misses[pkt->cmdToIndex()][pkt->req->masterId()]++;
            // always mark as cache fill for now... if we implement
            // no-write-allocate or bypass accesses this will have to
            // be changed.
            if (pkt->cmd == MemCmd::Writeback) {
                // We use forward_time here because there is an
                // uncached memory write, forwarded to WriteBuffer. It
                // specifies the latency to allocate an internal buffer and to
                // schedule an event to the queued port and also takes into
                // account the additional delay of the xbar.
                allocateWriteBuffer(pkt, forward_time, true);
            } else {
                if (blk && blk->isValid()) {
                    // If we have a write miss to a valid block, we
                    // need to mark the block non-readable.  Otherwise
                    // if we allow reads while there's an outstanding
                    // write miss, the read could return stale data
                    // out of the cache block... a more aggressive
                    // system could detect the overlap (if any) and
                    // forward data out of the MSHRs, but we don't do
                    // that yet.  Note that we do need to leave the
                    // block valid so that it stays in the cache, in
                    // case we get an upgrade response (and hence no
                    // new data) when the write miss completes.
                    // As long as CPUs do proper store/load forwarding
                    // internally, and have a sufficiently weak memory
                    // model, this is probably unnecessary, but at some
                    // point it must have seemed like we needed it...
                    assert(pkt->needsExclusive());
                    assert(!blk->isWritable());
                    blk->status &= ~BlkReadable;
                }
                // Here we are using forward_time, modelling the latency of
                // a miss (outbound) just as forwardLatency, neglecting the
                // lookupLatency component. In this case this latency value
                // specifies the latency to allocate an internal buffer and to
                // schedule an event to the queued port, when a cacheable miss
                // is forwarded to MSHR queue.
                // We take also into account the additional delay of the xbar.
                allocateMissBuffer(pkt, forward_time, true);
            }

            if (prefetcher) {
                // Don't notify on SWPrefetch
                if (!pkt->cmd.isSWPrefetch())
                    next_pf_time = prefetcher->notify(pkt);
            }
        }
    }
    // Here we condiser just forward_time.
    if (next_pf_time != MaxTick)
        requestMemSideBus(Request_PF, std::max(clockEdge(forwardLatency),
                                                next_pf_time));
    // copy writebacks to write buffer
    while (!writebacks.empty()) {
        PacketPtr wbPkt = writebacks.front();
        // We use forwardLatency here because we are copying writebacks
        // to write buffer. It specifies the latency to allocate an internal
        // buffer and to schedule an event to the queued port.
        allocateWriteBuffer(wbPkt, forward_time, true);
        writebacks.pop_front();
    }

    return true;
}


// See comment in cache.hh.
template<class TagStore>
PacketPtr
Cache<TagStore>::getBusPacket(PacketPtr cpu_pkt, BlkType *blk,
                              bool needsExclusive) const
{
    bool blkValid = blk && blk->isValid();

    if (cpu_pkt->req->isUncacheable()) {
        //assert(blk == NULL);
        return NULL;
    }

    if (!blkValid &&
        (cpu_pkt->cmd == MemCmd::Writeback || cpu_pkt->isUpgrade())) {
        // Writebacks that weren't allocated in access() and upgrades
        // from upper-level caches that missed completely just go
        // through.
        return NULL;
    }

    assert(cpu_pkt->needsResponse());

    MemCmd cmd;
    // @TODO make useUpgrades a parameter.
    // Note that ownership protocols require upgrade, otherwise a
    // write miss on a shared owned block will generate a ReadExcl,
    // which will clobber the owned copy.
    const bool useUpgrades = true;
    if (blkValid && useUpgrades) {
        // only reason to be here is that blk is shared
        // (read-only) and we need exclusive
        assert(needsExclusive);
        assert(!blk->isWritable());
        cmd = cpu_pkt->isLLSC() ? MemCmd::SCUpgradeReq : MemCmd::UpgradeReq;
    } else if (cpu_pkt->cmd == MemCmd::SCUpgradeFailReq ||
               cpu_pkt->cmd == MemCmd::StoreCondFailReq) {
        // Even though this SC will fail, we still need to send out the
        // request and get the data to supply it to other snoopers in the case
        // where the determination the StoreCond fails is delayed due to
        // all caches not being on the same local bus.
        cmd = MemCmd::SCUpgradeFailReq;
    } else if (cpu_pkt->isWriteInvalidate()) {
        cmd = cpu_pkt->cmd;
    } else {
        // block is invalid
        cmd = needsExclusive ? MemCmd::ReadExReq : MemCmd::ReadReq;
    }
    PacketPtr pkt = new Packet(cpu_pkt->req, cmd, blkSize);

    pkt->allocate();
    DPRINTF(Cache, "%s created %s addr %#llx size %d\n",
            __func__, pkt->cmdString(), pkt->getAddr(), pkt->getSize());
    return pkt;
}


template<class TagStore>
Tick
Cache<TagStore>::recvAtomic(PacketPtr pkt)
{
    // We are in atomic mode so we pay just for lookupLatency here.
    Cycles lat = lookupLatency;
    // @TODO: make this a parameter
    bool last_level_cache = false;

    // Forward the request if the system is in cache bypass mode.
    if (system->bypassCaches())
        return ticksToCycles(memSidePort->sendAtomic(pkt));

    promoteWholeLineWrites(pkt);

    if (pkt->memInhibitAsserted()) {
        assert(!pkt->req->isUncacheable());
        // have to invalidate ourselves and any lower caches even if
        // upper cache will be responding
        if (pkt->isInvalidate()) {
            BlkType *blk = tags->findBlock(pkt->getAddr(), pkt->isSecure());
            if (blk && blk->isValid()) {
                tags->invalidate(blk);
                blk->invalidate();
                DPRINTF(Cache, "rcvd mem-inhibited %s on %#llx (%s):"
                        " invalidating\n",
                        pkt->cmdString(), pkt->getAddr(),
                        pkt->isSecure() ? "s" : "ns");
            }
            if (!last_level_cache) {
                DPRINTF(Cache, "forwarding mem-inhibited %s on %#llx (%s)\n",
                        pkt->cmdString(), pkt->getAddr(),
                        pkt->isSecure() ? "s" : "ns");
                lat += ticksToCycles(memSidePort->sendAtomic(pkt));
            }
        } else {
            DPRINTF(Cache, "rcvd mem-inhibited %s on %#llx: not responding\n",
                    pkt->cmdString(), pkt->getAddr());
        }

        return lat * clockPeriod();
    }

    // should assert here that there are no outstanding MSHRs or
    // writebacks... that would mean that someone used an atomic
    // access in timing mode

    BlkType *blk = NULL;
    PacketList writebacks;

    if (!access(pkt, blk, lat, writebacks)) {
        // MISS

        PacketPtr bus_pkt = getBusPacket(pkt, blk, pkt->needsExclusive());

        bool is_forward = (bus_pkt == NULL);

        if (is_forward) {
            // just forwarding the same request to the next level
            // no local cache operation involved
            bus_pkt = pkt;
        }

        DPRINTF(Cache, "Sending an atomic %s for %#llx (%s)\n",
                bus_pkt->cmdString(), bus_pkt->getAddr(),
                bus_pkt->isSecure() ? "s" : "ns");

#if TRACING_ON
        CacheBlk::State old_state = blk ? blk->status : 0;
#endif

        lat += ticksToCycles(memSidePort->sendAtomic(bus_pkt));

        DPRINTF(Cache, "Receive response: %s for addr %#llx (%s) in state %i\n",
                bus_pkt->cmdString(), bus_pkt->getAddr(),
                bus_pkt->isSecure() ? "s" : "ns",
                old_state);

        // If packet was a forward, the response (if any) is already
        // in place in the bus_pkt == pkt structure, so we don't need
        // to do anything.  Otherwise, use the separate bus_pkt to
        // generate response to pkt and then delete it.
        if (!is_forward) {
            if (pkt->needsResponse()) {
                assert(bus_pkt->isResponse());
                if (bus_pkt->isError()) {
                    pkt->makeAtomicResponse();
                    pkt->copyError(bus_pkt);
                } else if (pkt->isWriteInvalidate()) {
                    // note the use of pkt, not bus_pkt here.
                    if (isTopLevel) {
                        blk = handleFill(pkt, blk, writebacks);
                        satisfyCpuSideRequest(pkt, blk);
                    } else if (blk) {
                        satisfyCpuSideRequest(pkt, blk);
                    }
                } else if (bus_pkt->isRead() ||
                           bus_pkt->cmd == MemCmd::UpgradeResp) {
                    // we're updating cache state to allow us to
                    // satisfy the upstream request from the cache
                    blk = handleFill(bus_pkt, blk, writebacks);
                    satisfyCpuSideRequest(pkt, blk);
                } else {
                    // we're satisfying the upstream request without
                    // modifying cache state, e.g., a write-through
                    pkt->makeAtomicResponse();
                }
            }
            delete bus_pkt;
        }
    }

    // Note that we don't invoke the prefetcher at all in atomic mode.
    // It's not clear how to do it properly, particularly for
    // prefetchers that aggressively generate prefetch candidates and
    // rely on bandwidth contention to throttle them; these will tend
    // to pollute the cache in atomic mode since there is no bandwidth
    // contention.  If we ever do want to enable prefetching in atomic
    // mode, though, this is the place to do it... see timingAccess()
    // for an example (though we'd want to issue the prefetch(es)
    // immediately rather than calling requestMemSideBus() as we do
    // there).

    // Handle writebacks if needed
    while (!writebacks.empty()){
        PacketPtr wbPkt = writebacks.front();
        memSidePort->sendAtomic(wbPkt);
        writebacks.pop_front();
        delete wbPkt;
    }

    if (pkt->needsResponse()) {
        pkt->makeAtomicResponse();
    }

    return lat * clockPeriod();
}


template<class TagStore>
void
Cache<TagStore>::functionalAccess(PacketPtr pkt, bool fromCpuSide)
{
    if (system->bypassCaches()) {
        // Packets from the memory side are snoop request and
        // shouldn't happen in bypass mode.
        assert(fromCpuSide);

        // The cache should be flushed if we are in cache bypass mode,
        // so we don't need to check if we need to update anything.
        memSidePort->sendFunctional(pkt);
        return;
    }

    Addr blk_addr = blockAlign(pkt->getAddr());
    bool is_secure = pkt->isSecure();
    BlkType *blk = tags->findBlock(pkt->getAddr(), is_secure);
    MSHR *mshr = mshrQueue.findMatch(blk_addr, is_secure);

    pkt->pushLabel(name());

    CacheBlkPrintWrapper cbpw(blk);

    // Note that just because an L2/L3 has valid data doesn't mean an
    // L1 doesn't have a more up-to-date modified copy that still
    // needs to be found.  As a result we always update the request if
    // we have it, but only declare it satisfied if we are the owner.

    // see if we have data at all (owned or otherwise)
    bool have_data = blk && blk->isValid()
        && pkt->checkFunctional(&cbpw, blk_addr, is_secure, blkSize,
                                blk->data);

    // data we have is dirty if marked as such or if valid & ownership
    // pending due to outstanding UpgradeReq
    bool have_dirty =
        have_data && (blk->isDirty() ||
                      (mshr && mshr->inService && mshr->isPendingDirty()));

    bool done = have_dirty
        || cpuSidePort->checkFunctional(pkt)
        || mshrQueue.checkFunctional(pkt, blk_addr)
        || writeBuffer.checkFunctional(pkt, blk_addr)
        || memSidePort->checkFunctional(pkt);

    DPRINTF(Cache, "functional %s %#llx (%s) %s%s%s\n",
            pkt->cmdString(), pkt->getAddr(), is_secure ? "s" : "ns",
            (blk && blk->isValid()) ? "valid " : "",
            have_data ? "data " : "", done ? "done " : "");

    // We're leaving the cache, so pop cache->name() label
    pkt->popLabel();

    if (done) {
        pkt->makeResponse();
    } else {
        // if it came as a request from the CPU side then make sure it
        // continues towards the memory side
        if (fromCpuSide) {
            memSidePort->sendFunctional(pkt);
        } else if (forwardSnoops && cpuSidePort->isSnooping()) {
            // if it came from the memory side, it must be a snoop request
            // and we should only forward it if we are forwarding snoops
            cpuSidePort->sendFunctionalSnoop(pkt);
        }
    }
}


/////////////////////////////////////////////////////
//
// Response handling: responses from the memory side
//
/////////////////////////////////////////////////////


template<class TagStore>
void
Cache<TagStore>::recvTimingResp(PacketPtr pkt)
{
    assert(pkt->isResponse());

    MSHR *mshr = dynamic_cast<MSHR*>(pkt->senderState);
    bool is_error = pkt->isError();

    assert(mshr);

    if (is_error) {
        DPRINTF(Cache, "Cache received packet with error for addr %#llx (%s), "
                "cmd: %s\n", pkt->getAddr(), pkt->isSecure() ? "s" : "ns",
                pkt->cmdString());
    }

    DPRINTF(Cache, "Handling response %s for addr %#llx size %d (%s)\n",
            pkt->cmdString(), pkt->getAddr(), pkt->getSize(),
            pkt->isSecure() ? "s" : "ns");

    MSHRQueue *mq = mshr->queue;
    bool wasFull = mq->isFull();

    if (mshr == noTargetMSHR) {
        // we always clear at least one target
        clearBlocked(Blocked_NoTargets);
        noTargetMSHR = NULL;
    }

    // Initial target is used just for stats
    MSHR::Target *initial_tgt = mshr->getTarget();
    BlkType *blk = tags->findBlock(pkt->getAddr(), pkt->isSecure());
    int stats_cmd_idx = initial_tgt->pkt->cmdToIndex();
    Tick miss_latency = curTick() - initial_tgt->recvTime;
    PacketList writebacks;
    // We need forward_time here because we have a call of
    // allocateWriteBuffer() that need this parameter to specify the
    // time to request the bus.  In this case we use forward latency
    // because there is a writeback.  We pay also here for headerDelay
    // that is charged of bus latencies if the packet comes from the
    // bus.
    Tick forward_time = clockEdge(forwardLatency) + pkt->headerDelay;

    if (pkt->req->isUncacheable()) {
        assert(pkt->req->masterId() < system->maxMasters());
        mshr_uncacheable_lat[stats_cmd_idx][pkt->req->masterId()] +=
            miss_latency;
    } else {
        assert(pkt->req->masterId() < system->maxMasters());
        mshr_miss_latency[stats_cmd_idx][pkt->req->masterId()] +=
            miss_latency;
    }

    bool is_fill = !mshr->isForward &&
        (pkt->isRead() || pkt->cmd == MemCmd::UpgradeResp);

    if (is_fill && !is_error) {
        DPRINTF(Cache, "Block for addr %#llx being updated in Cache\n",
                pkt->getAddr());

        // give mshr a chance to do some dirty work
        mshr->handleFill(pkt, blk);

        blk = handleFill(pkt, blk, writebacks);
        assert(blk != NULL);
    }

    // First offset for critical word first calculations
    int initial_offset = 0;

    if (mshr->hasTargets()) {
        initial_offset = mshr->getTarget()->pkt->getOffset(blkSize);
    }

    while (mshr->hasTargets()) {
        MSHR::Target *target = mshr->getTarget();

        switch (target->source) {
          case MSHR::Target::FromCPU:
            Tick completion_time;
            // Here we charge on completion_time the delay of the xbar if the
            // packet comes from it, charged on headerDelay.
            completion_time = pkt->headerDelay;

            // Software prefetch handling for cache closest to core
            if (target->pkt->cmd.isSWPrefetch() && isTopLevel) {
                // a software prefetch would have already been ack'd immediately
                // with dummy data so the core would be able to retire it.
                // this request completes right here, so we deallocate it.
                delete target->pkt->req;
                delete target->pkt;
                break; // skip response
            }

            // unlike the other packet flows, where data is found in other
            // caches or memory and brought back, write invalidates always
            // have the data right away, so the above check for "is fill?"
            // cannot actually be determined until examining the stored MSHR
            // state. We "catch up" with that logic here, which is duplicated
            // from above.
            if (target->pkt->isWriteInvalidate() && isTopLevel) {
                assert(!is_error);

                // NB: we use the original packet here and not the response!
                mshr->handleFill(target->pkt, blk);
                blk = handleFill(target->pkt, blk, writebacks);
                assert(blk != NULL);

                is_fill = true;
            }

            if (is_fill) {
                satisfyCpuSideRequest(target->pkt, blk,
                                      true, mshr->hasPostDowngrade());

                // How many bytes past the first request is this one
                int transfer_offset =
                    target->pkt->getOffset(blkSize) - initial_offset;
                if (transfer_offset < 0) {
                    transfer_offset += blkSize;
                }

                // If not critical word (offset) return payloadDelay.
                // responseLatency is the latency of the return path
                // from lower level caches/memory to an upper level cache or
                // the core.
                completion_time += clockEdge(responseLatency) +
                    (transfer_offset ? pkt->payloadDelay : 0);

                assert(!target->pkt->req->isUncacheable());

                assert(target->pkt->req->masterId() < system->maxMasters());
                missLatency[target->pkt->cmdToIndex()][target->pkt->req->masterId()] +=
                    completion_time - target->recvTime;
            } else if (pkt->cmd == MemCmd::UpgradeFailResp) {
                // failed StoreCond upgrade
                assert(target->pkt->cmd == MemCmd::StoreCondReq ||
                       target->pkt->cmd == MemCmd::StoreCondFailReq ||
                       target->pkt->cmd == MemCmd::SCUpgradeFailReq);
                // responseLatency is the latency of the return path
                // from lower level caches/memory to an upper level cache or
                // the core.
                completion_time += clockEdge(responseLatency) +
                    pkt->payloadDelay;
                target->pkt->req->setExtraData(0);
            } else {
                // not a cache fill, just forwarding response
                // responseLatency is the latency of the return path
                // from lower level cahces/memory to the core.
                completion_time += clockEdge(responseLatency) +
                    pkt->payloadDelay;
                if (pkt->isRead() && !is_error) {
                    target->pkt->setData(pkt->getConstPtr<uint8_t>());
                }
            }
            target->pkt->makeTimingResponse();
            // if this packet is an error copy that to the new packet
            if (is_error)
                target->pkt->copyError(pkt);
            if (target->pkt->cmd == MemCmd::ReadResp &&
                (pkt->isInvalidate() || mshr->hasPostInvalidate())) {
                // If intermediate cache got ReadRespWithInvalidate,
                // propagate that.  Response should not have
                // isInvalidate() set otherwise.
                target->pkt->cmd = MemCmd::ReadRespWithInvalidate;
                DPRINTF(Cache, "%s updated cmd to %s for addr %#llx\n",
                        __func__, target->pkt->cmdString(),
                        target->pkt->getAddr());
            }
            // Reset the bus additional time as it is now accounted for
            target->pkt->headerDelay = target->pkt->payloadDelay = 0;
            cpuSidePort->schedTimingResp(target->pkt, completion_time);
            break;

          case MSHR::Target::FromPrefetcher:
            assert(target->pkt->cmd == MemCmd::HardPFReq);
            if (blk)
                blk->status |= BlkHWPrefetched;
            delete target->pkt->req;
            delete target->pkt;
            break;

          case MSHR::Target::FromSnoop:
            // I don't believe that a snoop can be in an error state
            assert(!is_error);
            // response to snoop request
            DPRINTF(Cache, "processing deferred snoop...\n");
            assert(!(pkt->isInvalidate() && !mshr->hasPostInvalidate()));
            handleSnoop(target->pkt, blk, true, true,
                        mshr->hasPostInvalidate());
            break;

          default:
            panic("Illegal target->source enum %d\n", target->source);
        }

        mshr->popTarget();
    }

    if (blk && blk->isValid()) {
        if ((pkt->isInvalidate() || mshr->hasPostInvalidate()) &&
            (!pkt->isWriteInvalidate() || !isTopLevel)) {
            assert(blk != tempBlock);
            tags->invalidate(blk);
            blk->invalidate();
        } else if (mshr->hasPostDowngrade()) {
            blk->status &= ~BlkWritable;
        }
    }

    if (mshr->promoteDeferredTargets()) {
        // avoid later read getting stale data while write miss is
        // outstanding.. see comment in timingAccess()
        if (blk) {
            blk->status &= ~BlkReadable;
        }
        mq = mshr->queue;
        mq->markPending(mshr);
        requestMemSideBus((RequestCause)mq->index, clockEdge() +
                          pkt->payloadDelay);
    } else {
        mq->deallocate(mshr);
        if (wasFull && !mq->isFull()) {
            clearBlocked((BlockedCause)mq->index);
        }

        // Request the bus for a prefetch if this deallocation freed enough
        // MSHRs for a prefetch to take place
        if (prefetcher && mq == &mshrQueue && mshrQueue.canPrefetch()) {
            Tick next_pf_time = std::max(prefetcher->nextPrefetchReadyTime(),
                                         curTick());
            if (next_pf_time != MaxTick)
                requestMemSideBus(Request_PF, next_pf_time);
        }
    }
    // reset the xbar additional timinig  as it is now accounted for
    pkt->headerDelay = pkt->payloadDelay = 0;

    // copy writebacks to write buffer
    while (!writebacks.empty()) {
        PacketPtr wbPkt = writebacks.front();
        allocateWriteBuffer(wbPkt, clockEdge(forwardLatency), true);
        writebacks.pop_front();
    }
    // if we used temp block, clear it out
    if (blk == tempBlock) {
        if (blk->isDirty()) {
            // We use forwardLatency here because we are copying
            // writebacks to write buffer. It specifies the latency to
            // allocate an internal buffer and to schedule an event to the
            // queued port.
            allocateWriteBuffer(writebackBlk(blk), forward_time, true);
        }
        blk->invalidate();
    }

    DPRINTF(Cache, "Leaving %s with %s for addr %#llx\n", __func__,
            pkt->cmdString(), pkt->getAddr());
    delete pkt;
}




template<class TagStore>
PacketPtr
Cache<TagStore>::writebackBlk(BlkType *blk)
{
    assert(blk && blk->isValid() && blk->isDirty());

    writebacks[Request::wbMasterId]++;

    Request *writebackReq =
        new Request(tags->regenerateBlkAddr(blk->tag, blk->set), blkSize, 0,
                Request::wbMasterId);
    if (blk->isSecure())
        writebackReq->setFlags(Request::SECURE);

    writebackReq->taskId(blk->task_id);
    blk->task_id= ContextSwitchTaskId::Unknown;
    blk->tickInserted = curTick();

    PacketPtr writeback = new Packet(writebackReq, MemCmd::Writeback);
    if (blk->isWritable()) {
        writeback->setSupplyExclusive();
    }
    writeback->allocate();
    std::memcpy(writeback->getPtr<uint8_t>(), blk->data, blkSize);

    blk->status &= ~BlkDirty;
    return writeback;
}

template<class TagStore>
void
Cache<TagStore>::memWriteback()
{
    WrappedBlkVisitor visitor(*this, &Cache<TagStore>::writebackVisitor);
    tags->forEachBlk(visitor);
}

template<class TagStore>
void
Cache<TagStore>::memInvalidate()
{
    WrappedBlkVisitor visitor(*this, &Cache<TagStore>::invalidateVisitor);
    tags->forEachBlk(visitor);
}

template<class TagStore>
bool
Cache<TagStore>::isDirty() const
{
    CacheBlkIsDirtyVisitor<BlkType> visitor;
    tags->forEachBlk(visitor);

    return visitor.isDirty();
}

template<class TagStore>
bool
Cache<TagStore>::writebackVisitor(BlkType &blk)
{
    if (blk.isDirty()) {
        assert(blk.isValid());

        Request request(tags->regenerateBlkAddr(blk.tag, blk.set),
                        blkSize, 0, Request::funcMasterId);
        request.taskId(blk.task_id);

        Packet packet(&request, MemCmd::WriteReq);
        packet.dataStatic(blk.data);

        memSidePort->sendFunctional(&packet);

        blk.status &= ~BlkDirty;
    }

    return true;
}

template<class TagStore>
bool
Cache<TagStore>::invalidateVisitor(BlkType &blk)
{

    if (blk.isDirty())
        warn_once("Invalidating dirty cache lines. Expect things to break.\n");

    if (blk.isValid()) {
        assert(!blk.isDirty());
        tags->invalidate(dynamic_cast< BlkType *>(&blk));
        blk.invalidate();
    }

    return true;
}

template<class TagStore>
void
Cache<TagStore>::uncacheableFlush(PacketPtr pkt)
{
    DPRINTF(Cache, "%s%s addr %#llx uncacheable\n", pkt->cmdString(),
            pkt->req->isInstFetch() ? " (ifetch)" : "",
            pkt->getAddr());

    if (pkt->req->isClearLL())
        tags->clearLocks();

    BlkType *blk(tags->findBlock(pkt->getAddr(), pkt->isSecure()));
    if (blk) {
        writebackVisitor(*blk);
        invalidateVisitor(*blk);
    }
}


template<class TagStore>
typename Cache<TagStore>::BlkType*
Cache<TagStore>::allocateBlock(Addr addr, bool is_secure,
                               PacketList &writebacks)
{
    BlkType *blk = tags->findVictim(addr);

    if (blk->isValid()) {
        Addr repl_addr = tags->regenerateBlkAddr(blk->tag, blk->set);
        MSHR *repl_mshr = mshrQueue.findMatch(repl_addr, blk->isSecure());
        if (repl_mshr) {
            // must be an outstanding upgrade request
            // on a block we're about to replace...
            assert(!blk->isWritable() || blk->isDirty());
            assert(repl_mshr->needsExclusive());
            // too hard to replace block with transient state
            // allocation failed, block not inserted
            return NULL;
        } else {
            DPRINTF(Cache, "replacement: replacing %#llx (%s) with %#llx (%s): %s\n",
                    repl_addr, blk->isSecure() ? "s" : "ns",
                    addr, is_secure ? "s" : "ns",
                    blk->isDirty() ? "writeback" : "clean");

            if (blk->isDirty()) {
                // Save writeback packet for handling by caller
                writebacks.push_back(writebackBlk(blk));
            }
        }
    }

    return blk;
}


// Note that the reason we return a list of writebacks rather than
// inserting them directly in the write buffer is that this function
// is called by both atomic and timing-mode accesses, and in atomic
// mode we don't mess with the write buffer (we just perform the
// writebacks atomically once the original request is complete).
template<class TagStore>
typename Cache<TagStore>::BlkType*
Cache<TagStore>::handleFill(PacketPtr pkt, BlkType *blk,
                            PacketList &writebacks)
{
    assert(pkt->isResponse() || pkt->isWriteInvalidate());
    Addr addr = pkt->getAddr();
    bool is_secure = pkt->isSecure();
#if TRACING_ON
    CacheBlk::State old_state = blk ? blk->status : 0;
#endif

    if (blk == NULL) {
        // better have read new data...
        assert(pkt->hasData());

        // only read responses and (original) write invalidate req's have data;
        // note that we don't write the data here for write invalidate - that
        // happens in the subsequent satisfyCpuSideRequest.
        assert(pkt->isRead() || pkt->isWriteInvalidate());

        // need to do a replacement
        blk = allocateBlock(addr, is_secure, writebacks);
        if (blk == NULL) {
            // No replaceable block... just use temporary storage to
            // complete the current request and then get rid of it
            assert(!tempBlock->isValid());
            blk = tempBlock;
            tempBlock->set = tags->extractSet(addr);
            tempBlock->tag = tags->extractTag(addr);
            // @todo: set security state as well...
            DPRINTF(Cache, "using temp block for %#llx (%s)\n", addr,
                    is_secure ? "s" : "ns");
        } else {
            tags->insertBlock(pkt, blk);
        }

        // we should never be overwriting a valid block
        assert(!blk->isValid());
    } else {
        // existing block... probably an upgrade
        assert(blk->tag == tags->extractTag(addr));
        // either we're getting new data or the block should already be valid
        assert(pkt->hasData() || blk->isValid());
        // don't clear block status... if block is already dirty we
        // don't want to lose that
    }

    if (is_secure)
        blk->status |= BlkSecure;
    blk->status |= BlkValid | BlkReadable;

    if (!pkt->sharedAsserted()) {
        blk->status |= BlkWritable;
        // If we got this via cache-to-cache transfer (i.e., from a
        // cache that was an owner) and took away that owner's copy,
        // then we need to write it back.  Normally this happens
        // anyway as a side effect of getting a copy to write it, but
        // there are cases (such as failed store conditionals or
        // compare-and-swaps) where we'll demand an exclusive copy but
        // end up not writing it.
        if (pkt->memInhibitAsserted())
            blk->status |= BlkDirty;
    }

    DPRINTF(Cache, "Block addr %#llx (%s) moving from state %x to %s\n",
            addr, is_secure ? "s" : "ns", old_state, blk->print());

    // if we got new data, copy it in (checking for a read response
    // and a response that has data is the same in the end)
    if (pkt->isRead()) {
        assert(pkt->hasData());
        std::memcpy(blk->data, pkt->getConstPtr<uint8_t>(), blkSize);
    }
    // We pay for fillLatency here.
    blk->whenReady = clockEdge() + fillLatency * clockPeriod() +
        pkt->payloadDelay;

    return blk;
}


/////////////////////////////////////////////////////
//
// Snoop path: requests coming in from the memory side
//
/////////////////////////////////////////////////////

template<class TagStore>
void
Cache<TagStore>::
doTimingSupplyResponse(PacketPtr req_pkt, const uint8_t *blk_data,
                       bool already_copied, bool pending_inval)
{
    // sanity check
    assert(req_pkt->isRequest());
    assert(req_pkt->needsResponse());

    DPRINTF(Cache, "%s for %s addr %#llx size %d\n", __func__,
            req_pkt->cmdString(), req_pkt->getAddr(), req_pkt->getSize());
    // timing-mode snoop responses require a new packet, unless we
    // already made a copy...
    PacketPtr pkt = req_pkt;
    if (!already_copied)
        // do not clear flags, and allocate space for data if the
        // packet needs it (the only packets that carry data are read
        // responses)
        pkt = new Packet(req_pkt, false, req_pkt->isRead());

    assert(req_pkt->isInvalidate() || pkt->sharedAsserted());
    pkt->makeTimingResponse();
    if (pkt->isRead()) {
        pkt->setDataFromBlock(blk_data, blkSize);
    }
    if (pkt->cmd == MemCmd::ReadResp && pending_inval) {
        // Assume we defer a response to a read from a far-away cache
        // A, then later defer a ReadExcl from a cache B on the same
        // bus as us.  We'll assert MemInhibit in both cases, but in
        // the latter case MemInhibit will keep the invalidation from
        // reaching cache A.  This special response tells cache A that
        // it gets the block to satisfy its read, but must immediately
        // invalidate it.
        pkt->cmd = MemCmd::ReadRespWithInvalidate;
    }
    // Here we consider forward_time, paying for just forward latency and
    // also charging the delay provided by the xbar.
    // forward_time is used as send_time in next allocateWriteBuffer().
    Tick forward_time = clockEdge(forwardLatency) + pkt->headerDelay;
    // Here we reset the timing of the packet.
    pkt->headerDelay = pkt->payloadDelay = 0;
    DPRINTF(Cache, "%s created response: %s addr %#llx size %d tick: %lu\n",
            __func__, pkt->cmdString(), pkt->getAddr(), pkt->getSize(),
            forward_time);
    memSidePort->schedTimingSnoopResp(pkt, forward_time, true);
}

template<class TagStore>
void
Cache<TagStore>::handleSnoop(PacketPtr pkt, BlkType *blk,
                             bool is_timing, bool is_deferred,
                             bool pending_inval)
{
    DPRINTF(Cache, "%s for %s addr %#llx size %d\n", __func__,
            pkt->cmdString(), pkt->getAddr(), pkt->getSize());
    // deferred snoops can only happen in timing mode
    assert(!(is_deferred && !is_timing));
    // pending_inval only makes sense on deferred snoops
    assert(!(pending_inval && !is_deferred));
    assert(pkt->isRequest());

    // the packet may get modified if we or a forwarded snooper
    // responds in atomic mode, so remember a few things about the
    // original packet up front
    bool invalidate = pkt->isInvalidate();
    bool M5_VAR_USED needs_exclusive = pkt->needsExclusive();

    if (forwardSnoops) {
        // first propagate snoop upward to see if anyone above us wants to
        // handle it.  save & restore packet src since it will get
        // rewritten to be relative to cpu-side bus (if any)
        bool alreadyResponded = pkt->memInhibitAsserted();
        if (is_timing) {
            Packet snoopPkt(pkt, true, false);  // clear flags, no allocation
            snoopPkt.setExpressSnoop();
            snoopPkt.pushSenderState(new ForwardResponseRecord());
            // the snoop packet does not need to wait any additional
            // time
            snoopPkt.headerDelay = snoopPkt.payloadDelay = 0;
            cpuSidePort->sendTimingSnoopReq(&snoopPkt);
            if (snoopPkt.memInhibitAsserted()) {
                // cache-to-cache response from some upper cache
                assert(!alreadyResponded);
                pkt->assertMemInhibit();
            } else {
                delete snoopPkt.popSenderState();
            }
            if (snoopPkt.sharedAsserted()) {
                pkt->assertShared();
            }
            // If this request is a prefetch and an
            // upper level squashes the prefetch request,
            // make sure to propogate the squash to the requester.
            if (snoopPkt.prefetchSquashed()) {
                pkt->setPrefetchSquashed();
            }
        } else {
            cpuSidePort->sendAtomicSnoop(pkt);
            if (!alreadyResponded && pkt->memInhibitAsserted()) {
                // cache-to-cache response from some upper cache:
                // forward response to original requester
                assert(pkt->isResponse());
            }
        }
    }

    if (!blk || !blk->isValid()) {
        DPRINTF(Cache, "%s snoop miss for %s addr %#llx size %d\n",
                __func__, pkt->cmdString(), pkt->getAddr(), pkt->getSize());
        return;
    } else {
       DPRINTF(Cache, "%s snoop hit for %s for addr %#llx size %d, "
               "old state is %s\n", __func__, pkt->cmdString(),
               pkt->getAddr(), pkt->getSize(), blk->print());
    }

    // we may end up modifying both the block state and the packet (if
    // we respond in atomic mode), so just figure out what to do now
    // and then do it later.  If we find dirty data while snooping for a
    // WriteInvalidate, we don't care, since no merging needs to take place.
    // We need the eviction to happen as normal, but the data needn't be
    // sent anywhere. nor should the writeback be inhibited at the memory
    // controller for any reason.
    bool respond = blk->isDirty() && pkt->needsResponse()
                                  && !pkt->isWriteInvalidate();
    bool have_exclusive = blk->isWritable();

    // Invalidate any prefetch's from below that would strip write permissions
    // MemCmd::HardPFReq is only observed by upstream caches.  After missing
    // above and in it's own cache, a new MemCmd::ReadReq is created that
    // downstream caches observe.
    if (pkt->cmd == MemCmd::HardPFReq) {
        DPRINTF(Cache, "Squashing prefetch from lower cache %#x\n",
                pkt->getAddr());
        pkt->setPrefetchSquashed();
        return;
    }

    if (pkt->isRead() && !invalidate) {
        assert(!needs_exclusive);
        pkt->assertShared();
        int bits_to_clear = BlkWritable;
        const bool haveOwnershipState = true; // for now
        if (!haveOwnershipState) {
            // if we don't support pure ownership (dirty && !writable),
            // have to clear dirty bit here, assume memory snarfs data
            // on cache-to-cache xfer
            bits_to_clear |= BlkDirty;
        }
        blk->status &= ~bits_to_clear;
    }

    if (respond) {
        // prevent anyone else from responding, cache as well as
        // memory, and also prevent any memory from even seeing the
        // request (with current inhibited semantics), note that this
        // applies both to reads and writes and that for writes it
        // works thanks to the fact that we still have dirty data and
        // will write it back at a later point
        pkt->assertMemInhibit();
        if (have_exclusive) {
            pkt->setSupplyExclusive();
        }
        if (is_timing) {
            doTimingSupplyResponse(pkt, blk->data, is_deferred, pending_inval);
        } else {
            pkt->makeAtomicResponse();
            pkt->setDataFromBlock(blk->data, blkSize);
        }
    } else if (is_timing && is_deferred) {
        // if it's a deferred timing snoop then we've made a copy of
        // the packet, and so if we're not using that copy to respond
        // then we need to delete it here.
        delete pkt;
    }

    // Do this last in case it deallocates block data or something
    // like that
    if (invalidate) {
        if (blk != tempBlock)
            tags->invalidate(blk);
        blk->invalidate();
    }

    DPRINTF(Cache, "new state is %s\n", blk->print());
}


template<class TagStore>
void
Cache<TagStore>::recvTimingSnoopReq(PacketPtr pkt)
{
    DPRINTF(Cache, "%s for %s addr %#llx size %d\n", __func__,
            pkt->cmdString(), pkt->getAddr(), pkt->getSize());

    // Snoops shouldn't happen when bypassing caches
    assert(!system->bypassCaches());

    // check if the packet is for an address range covered by this
    // cache, partly to not waste time looking for it, but also to
    // ensure that we only forward the snoop upwards if it is within
    // our address ranges
    bool in_range = false;
    for (AddrRangeList::const_iterator r = addrRanges.begin();
         r != addrRanges.end(); ++r) {
        if (r->contains(pkt->getAddr())) {
            in_range = true;
            break;
        }
    }

    // Note that some deferred snoops don't have requests, since the
    // original access may have already completed
    if ((pkt->req && pkt->req->isUncacheable()) ||
        pkt->cmd == MemCmd::Writeback || !in_range) {
        //Can't get a hit on an uncacheable address
        //Revisit this for multi level coherence
        return;
    }

    bool is_secure = pkt->isSecure();
    BlkType *blk = tags->findBlock(pkt->getAddr(), is_secure);

    Addr blk_addr = blockAlign(pkt->getAddr());
    MSHR *mshr = mshrQueue.findMatch(blk_addr, is_secure);

    // Squash any prefetch requests from below on MSHR hits
    if (mshr && pkt->cmd == MemCmd::HardPFReq) {
        DPRINTF(Cache, "Squashing prefetch from lower cache on mshr hit %#x\n",
                pkt->getAddr());
        pkt->setPrefetchSquashed();
        return;
    }

    // Let the MSHR itself track the snoop and decide whether we want
    // to go ahead and do the regular cache snoop
    if (mshr && mshr->handleSnoop(pkt, order++)) {
        DPRINTF(Cache, "Deferring snoop on in-service MSHR to blk %#llx (%s)."
                "mshrs: %s\n", blk_addr, is_secure ? "s" : "ns",
                mshr->print());

        if (mshr->getNumTargets() > numTarget)
            warn("allocating bonus target for snoop"); //handle later
        return;
    }

    //We also need to check the writeback buffers and handle those
    std::vector<MSHR *> writebacks;
    if (writeBuffer.findMatches(blk_addr, is_secure, writebacks)) {
        DPRINTF(Cache, "Snoop hit in writeback to addr %#llx (%s)\n",
                pkt->getAddr(), is_secure ? "s" : "ns");

        // Look through writebacks for any cachable writes.
        // We should only ever find a single match
        assert(writebacks.size() == 1);
        MSHR *wb_entry = writebacks[0];
        assert(!wb_entry->isUncacheable());
        assert(wb_entry->getNumTargets() == 1);
        PacketPtr wb_pkt = wb_entry->getTarget()->pkt;
        assert(wb_pkt->cmd == MemCmd::Writeback);

        assert(!pkt->memInhibitAsserted());
        pkt->assertMemInhibit();
        if (!pkt->needsExclusive()) {
            pkt->assertShared();
            // the writeback is no longer the exclusive copy in the system
            wb_pkt->clearSupplyExclusive();
        } else {
            // if we're not asserting the shared line, we need to
            // invalidate our copy.  we'll do that below as long as
            // the packet's invalidate flag is set...
            assert(pkt->isInvalidate());
        }
        doTimingSupplyResponse(pkt, wb_pkt->getConstPtr<uint8_t>(),
                               false, false);

        if (pkt->isInvalidate()) {
            // Invalidation trumps our writeback... discard here
            markInService(wb_entry, false);
            delete wb_pkt;
        }
    }

    // If this was a shared writeback, there may still be
    // other shared copies above that require invalidation.
    // We could be more selective and return here if the
    // request is non-exclusive or if the writeback is
    // exclusive.
    handleSnoop(pkt, blk, true, false, false);
}

template<class TagStore>
bool
Cache<TagStore>::CpuSidePort::recvTimingSnoopResp(PacketPtr pkt)
{
    // Express snoop responses from master to slave, e.g., from L1 to L2
    cache->recvTimingSnoopResp(pkt);
    return true;
}

template<class TagStore>
Tick
Cache<TagStore>::recvAtomicSnoop(PacketPtr pkt)
{
    // Snoops shouldn't happen when bypassing caches
    assert(!system->bypassCaches());

    if (pkt->req->isUncacheable() || pkt->cmd == MemCmd::Writeback) {
        // Can't get a hit on an uncacheable address
        // Revisit this for multi level coherence
        return 0;
    }

    BlkType *blk = tags->findBlock(pkt->getAddr(), pkt->isSecure());
    handleSnoop(pkt, blk, false, false, false);
    // We consider forwardLatency here because a snoop occurs in atomic mode
    return forwardLatency * clockPeriod();
}


template<class TagStore>
MSHR *
Cache<TagStore>::getNextMSHR()
{
    // Check both MSHR queue and write buffer for potential requests,
    // note that null does not mean there is no request, it could
    // simply be that it is not ready
    MSHR *miss_mshr  = mshrQueue.getNextMSHR();
    MSHR *write_mshr = writeBuffer.getNextMSHR();

    // If we got a write buffer request ready, first priority is a
    // full write buffer, otherwhise we favour the miss requests
    if (write_mshr &&
        ((writeBuffer.isFull() && writeBuffer.inServiceEntries == 0) ||
         !miss_mshr)) {
        // need to search MSHR queue for conflicting earlier miss.
        MSHR *conflict_mshr =
            mshrQueue.findPending(write_mshr->addr, write_mshr->size,
                                  write_mshr->isSecure);

        if (conflict_mshr && conflict_mshr->order < write_mshr->order) {
            // Service misses in order until conflict is cleared.
            return conflict_mshr;

            // @todo Note that we ignore the ready time of the conflict here
        }

        // No conflicts; issue write
        return write_mshr;
    } else if (miss_mshr) {
        // need to check for conflicting earlier writeback
        MSHR *conflict_mshr =
            writeBuffer.findPending(miss_mshr->addr, miss_mshr->size,
                                    miss_mshr->isSecure);
        if (conflict_mshr) {
            // not sure why we don't check order here... it was in the
            // original code but commented out.

            // The only way this happens is if we are
            // doing a write and we didn't have permissions
            // then subsequently saw a writeback (owned got evicted)
            // We need to make sure to perform the writeback first
            // To preserve the dirty data, then we can issue the write

            // should we return write_mshr here instead?  I.e. do we
            // have to flush writes in order?  I don't think so... not
            // for Alpha anyway.  Maybe for x86?
            return conflict_mshr;

            // @todo Note that we ignore the ready time of the conflict here
        }

        // No conflicts; issue read
        return miss_mshr;
    }

    // fall through... no pending requests.  Try a prefetch.
    assert(!miss_mshr && !write_mshr);
    if (prefetcher && mshrQueue.canPrefetch()) {
        // If we have a miss queue slot, we can try a prefetch
        PacketPtr pkt = prefetcher->getPacket();
        if (pkt) {
            Addr pf_addr = blockAlign(pkt->getAddr());
            if (!tags->findBlock(pf_addr, pkt->isSecure()) &&
                !mshrQueue.findMatch(pf_addr, pkt->isSecure()) &&
                !writeBuffer.findMatch(pf_addr, pkt->isSecure())) {
                // Update statistic on number of prefetches issued
                // (hwpf_mshr_misses)
                assert(pkt->req->masterId() < system->maxMasters());
                mshr_misses[pkt->cmdToIndex()][pkt->req->masterId()]++;
                // Don't request bus, since we already have it
                return allocateMissBuffer(pkt, curTick(), false);
            } else {
                // free the request and packet
                delete pkt->req;
                delete pkt;
            }
        }
    }

    return NULL;
}


template<class TagStore>
PacketPtr
Cache<TagStore>::getTimingPacket()
{
    MSHR *mshr = getNextMSHR();

    if (mshr == NULL) {
        return NULL;
    }

    // use request from 1st target
    PacketPtr tgt_pkt = mshr->getTarget()->pkt;
    PacketPtr pkt = NULL;

    DPRINTF(CachePort, "%s %s for addr %#llx size %d\n", __func__,
            tgt_pkt->cmdString(), tgt_pkt->getAddr(), tgt_pkt->getSize());

    if (mshr->isForwardNoResponse()) {
        // no response expected, just forward packet as it is
        assert(tags->findBlock(mshr->addr, mshr->isSecure) == NULL);
        pkt = tgt_pkt;
    } else {
        BlkType *blk = tags->findBlock(mshr->addr, mshr->isSecure);

        if (tgt_pkt->cmd == MemCmd::HardPFReq) {
            // We need to check the caches above us to verify that
            // they don't have a copy of this block in the dirty state
            // at the moment. Without this check we could get a stale
            // copy from memory that might get used in place of the
            // dirty one.
            Packet snoop_pkt(tgt_pkt, true, false);
            snoop_pkt.setExpressSnoop();
            snoop_pkt.senderState = mshr;
            cpuSidePort->sendTimingSnoopReq(&snoop_pkt);

            // Check to see if the prefetch was squashed by an upper cache (to
            // prevent us from grabbing the line) or if a Check to see if a
            // writeback arrived between the time the prefetch was placed in
            // the MSHRs and when it was selected to be sent or if the
            // prefetch was squashed by an upper cache.

            // It is important to check msmInhibitAsserted before
            // prefetchSquashed. If another cache has asserted MEM_INGIBIT, it
            // will be sending a response which will arrive at the MSHR
            // allocated ofr this request. Checking the prefetchSquash first
            // may result in the MSHR being prematurely deallocated.

            if (snoop_pkt.memInhibitAsserted()) {
                // If we are getting a non-shared response it is dirty
                bool pending_dirty_resp = !snoop_pkt.sharedAsserted();
                markInService(mshr, pending_dirty_resp);
                DPRINTF(Cache, "Upward snoop of prefetch for addr"
                        " %#x (%s) hit\n",
                        tgt_pkt->getAddr(), tgt_pkt->isSecure()? "s": "ns");
                return NULL;
            }

            if (snoop_pkt.prefetchSquashed() || blk != NULL) {
                DPRINTF(Cache, "Prefetch squashed by cache.  "
                               "Deallocating mshr target %#x.\n", mshr->addr);

                // Deallocate the mshr target
                if (mshr->queue->forceDeallocateTarget(mshr)) {
                    // Clear block if this deallocation resulted freed an
                    // mshr when all had previously been utilized
                    clearBlocked((BlockedCause)(mshr->queue->index));
                }
                return NULL;
            }

        }

        pkt = getBusPacket(tgt_pkt, blk, mshr->needsExclusive());

        mshr->isForward = (pkt == NULL);

        if (mshr->isForward) {
            // not a cache block request, but a response is expected
            // make copy of current packet to forward, keep current
            // copy for response handling
            pkt = new Packet(tgt_pkt, false, true);
            if (pkt->isWrite()) {
                pkt->setData(tgt_pkt->getConstPtr<uint8_t>());
            }
        }
    }

    assert(pkt != NULL);
    pkt->senderState = mshr;
    return pkt;
}


template<class TagStore>
Tick
Cache<TagStore>::nextMSHRReadyTime() const
{
    Tick nextReady = std::min(mshrQueue.nextMSHRReadyTime(),
                              writeBuffer.nextMSHRReadyTime());

    // Don't signal prefetch ready time if no MSHRs available
    // Will signal once enoguh MSHRs are deallocated
    if (prefetcher && mshrQueue.canPrefetch()) {
        nextReady = std::min(nextReady,
                             prefetcher->nextPrefetchReadyTime());
    }

    return nextReady;
}

template<class TagStore>
void
Cache<TagStore>::serialize(std::ostream &os)
{
    bool dirty(isDirty());

    if (dirty) {
        warn("*** The cache still contains dirty data. ***\n");
        warn("    Make sure to drain the system using the correct flags.\n");
        warn("    This checkpoint will not restore correctly and dirty data in "
             "the cache will be lost!\n");
    }

    // Since we don't checkpoint the data in the cache, any dirty data
    // will be lost when restoring from a checkpoint of a system that
    // wasn't drained properly. Flag the checkpoint as invalid if the
    // cache contains dirty data.
    bool bad_checkpoint(dirty);
    SERIALIZE_SCALAR(bad_checkpoint);
}

template<class TagStore>
void
Cache<TagStore>::unserialize(Checkpoint *cp, const std::string &section)
{
    bool bad_checkpoint;
    UNSERIALIZE_SCALAR(bad_checkpoint);
    if (bad_checkpoint) {
        fatal("Restoring from checkpoints with dirty caches is not supported "
              "in the classic memory system. Please remove any caches or "
              " drain them properly before taking checkpoints.\n");
    }
}

///////////////
//
// CpuSidePort
//
///////////////

template<class TagStore>
AddrRangeList
Cache<TagStore>::CpuSidePort::getAddrRanges() const
{
    return cache->getAddrRanges();
}

template<class TagStore>
bool
Cache<TagStore>::CpuSidePort::recvTimingReq(PacketPtr pkt)
{
    assert(!cache->system->bypassCaches());

    bool success = false;

    // always let inhibited requests through, even if blocked,
    // ultimately we should check if this is an express snoop, but at
    // the moment that flag is only set in the cache itself
    if (pkt->memInhibitAsserted()) {
        // do not change the current retry state
        bool M5_VAR_USED bypass_success = cache->recvTimingReq(pkt);
        assert(bypass_success);
        return true;
    } else if (blocked || mustSendRetry) {
        // either already committed to send a retry, or blocked
        success = false;
    } else {
        // pass it on to the cache, and let the cache decide if we
        // have to retry or not
        success = cache->recvTimingReq(pkt);
    }

    // remember if we have to retry
    mustSendRetry = !success;
    return success;
}

template<class TagStore>
Tick
Cache<TagStore>::CpuSidePort::recvAtomic(PacketPtr pkt)
{
    return cache->recvAtomic(pkt);
}

template<class TagStore>
void
Cache<TagStore>::CpuSidePort::recvFunctional(PacketPtr pkt)
{
    // functional request
    cache->functionalAccess(pkt, true);
}

template<class TagStore>
Cache<TagStore>::
CpuSidePort::CpuSidePort(const std::string &_name, Cache<TagStore> *_cache,
                         const std::string &_label)
    : BaseCache::CacheSlavePort(_name, _cache, _label), cache(_cache)
{
}

///////////////
//
// MemSidePort
//
///////////////

template<class TagStore>
bool
Cache<TagStore>::MemSidePort::recvTimingResp(PacketPtr pkt)
{
    cache->recvTimingResp(pkt);
    return true;
}

// Express snooping requests to memside port
template<class TagStore>
void
Cache<TagStore>::MemSidePort::recvTimingSnoopReq(PacketPtr pkt)
{
    // handle snooping requests
    cache->recvTimingSnoopReq(pkt);
}

template<class TagStore>
Tick
Cache<TagStore>::MemSidePort::recvAtomicSnoop(PacketPtr pkt)
{
    return cache->recvAtomicSnoop(pkt);
}

template<class TagStore>
void
Cache<TagStore>::MemSidePort::recvFunctionalSnoop(PacketPtr pkt)
{
    // functional snoop (note that in contrast to atomic we don't have
    // a specific functionalSnoop method, as they have the same
    // behaviour regardless)
    cache->functionalAccess(pkt, false);
}

template<class TagStore>
void
Cache<TagStore>::CacheReqPacketQueue::sendDeferredPacket()
{
    // sanity check
    assert(!waitingOnRetry);

    // there should never be any deferred request packets in the
    // queue, instead we resly on the cache to provide the packets
    // from the MSHR queue or write queue
    assert(deferredPacketReadyTime() == MaxTick);

    // check for request packets (requests & writebacks)
    PacketPtr pkt = cache.getTimingPacket();
    if (pkt == NULL) {
        // can happen if e.g. we attempt a writeback and fail, but
        // before the retry, the writeback is eliminated because
        // we snoop another cache's ReadEx.
    } else {
        MSHR *mshr = dynamic_cast<MSHR*>(pkt->senderState);
        // in most cases getTimingPacket allocates a new packet, and
        // we must delete it unless it is successfully sent
        bool delete_pkt = !mshr->isForwardNoResponse();

        // let our snoop responses go first if there are responses to
        // the same addresses we are about to writeback, note that
        // this creates a dependency between requests and snoop
        // responses, but that should not be a problem since there is
        // a chain already and the key is that the snoop responses can
        // sink unconditionally
        if (snoopRespQueue.hasAddr(pkt->getAddr())) {
            DPRINTF(CachePort, "Waiting for snoop response to be sent\n");
            Tick when = snoopRespQueue.deferredPacketReadyTime();
            schedSendEvent(when);

            if (delete_pkt)
                delete pkt;

            return;
        }


        waitingOnRetry = !masterPort.sendTimingReq(pkt);

        if (waitingOnRetry) {
            DPRINTF(CachePort, "now waiting on a retry\n");
            if (delete_pkt) {
                // we are awaiting a retry, but we
                // delete the packet and will be creating a new packet
                // when we get the opportunity
                delete pkt;
            }
            // note that we have now masked any requestBus and
            // schedSendEvent (we will wait for a retry before
            // doing anything), and this is so even if we do not
            // care about this packet and might override it before
            // it gets retried
        } else {
            // As part of the call to sendTimingReq the packet is
            // forwarded to all neighbouring caches (and any
            // caches above them) as a snoop. The packet is also
            // sent to any potential cache below as the
            // interconnect is not allowed to buffer the
            // packet. Thus at this point we know if any of the
            // neighbouring, or the downstream cache is
            // responding, and if so, if it is with a dirty line
            // or not.
            bool pending_dirty_resp = !pkt->sharedAsserted() &&
                pkt->memInhibitAsserted();

            cache.markInService(mshr, pending_dirty_resp);
        }
    }

    // if we succeeded and are not waiting for a retry, schedule the
    // next send considering when the next MSHR is ready, note that
    // snoop responses have their own packet queue and thus schedule
    // their own events
    if (!waitingOnRetry) {
        schedSendEvent(cache.nextMSHRReadyTime());
    }
}

template<class TagStore>
Cache<TagStore>::
MemSidePort::MemSidePort(const std::string &_name, Cache<TagStore> *_cache,
                         const std::string &_label)
    : BaseCache::CacheMasterPort(_name, _cache, _reqQueue, _snoopRespQueue),
      _reqQueue(*_cache, *this, _snoopRespQueue, _label),
      _snoopRespQueue(*_cache, *this, _label), cache(_cache)
{
}

#endif//__MEM_CACHE_CACHE_IMPL_HH__