summaryrefslogtreecommitdiff
path: root/src/mem/cache/cache.cc
blob: e7dd5efc952ac772520b0ba9341a6f71b9980ca6 (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
/*
 * Copyright (c) 2010-2019 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
 *          Nikos Nikoleris
 */

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

#include "mem/cache/cache.hh"

#include <cassert>

#include "base/compiler.hh"
#include "base/logging.hh"
#include "base/trace.hh"
#include "base/types.hh"
#include "debug/Cache.hh"
#include "debug/CacheTags.hh"
#include "debug/CacheVerbose.hh"
#include "enums/Clusivity.hh"
#include "mem/cache/cache_blk.hh"
#include "mem/cache/mshr.hh"
#include "mem/cache/tags/base.hh"
#include "mem/cache/write_queue_entry.hh"
#include "mem/request.hh"
#include "params/Cache.hh"

Cache::Cache(const CacheParams *p)
    : BaseCache(p, p->system->cacheLineSize()),
      doFastWrites(true)
{
}

void
Cache::satisfyRequest(PacketPtr pkt, CacheBlk *blk,
                      bool deferred_response, bool pending_downgrade)
{
    BaseCache::satisfyRequest(pkt, blk);

    if (pkt->isRead()) {
        // determine if this read is from a (coherent) cache or not
        if (pkt->fromCache()) {
            assert(pkt->getSize() == blkSize);
            // special handling for coherent block requests from
            // upper-level caches
            if (pkt->needsWritable()) {
                // sanity check
                assert(pkt->cmd == MemCmd::ReadExReq ||
                       pkt->cmd == MemCmd::SCUpgradeFailReq);
                assert(!pkt->hasSharers());

                // if we have a dirty copy, make sure the recipient
                // keeps it marked dirty (in the modified state)
                if (blk->isDirty()) {
                    pkt->setCacheResponding();
                    blk->status &= ~BlkDirty;
                }
            } else if (blk->isWritable() && !pending_downgrade &&
                       !pkt->hasSharers() &&
                       pkt->cmd != MemCmd::ReadCleanReq) {
                // we can give the requester a writable copy on a read
                // request if:
                // - we have a writable 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 set hasSharers flag when
                //   snooping the packet)
                // - the read has explicitly asked for a clean
                //   copy of the line
                if (blk->isDirty()) {
                    // special considerations if we're owner:
                    if (!deferred_response) {
                        // respond with the line in Modified state
                        // (cacheResponding set, hasSharers not set)
                        pkt->setCacheResponding();

                        // if this cache is mostly inclusive, we
                        // keep the block in the Exclusive state,
                        // and pass it upwards as Modified
                        // (writable and dirty), hence we have
                        // multiple caches, all on the same path
                        // towards memory, all considering the
                        // same block writable, but only one
                        // considering it Modified

                        // we get away with multiple caches (on
                        // the same path to memory) considering
                        // the block writeable as we always enter
                        // the cache hierarchy through a cache,
                        // and first snoop upwards in all other
                        // branches
                        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
                        // have to respond with a shared line
                        pkt->setHasSharers();
                    }
                }
            } else {
                // otherwise only respond with a shared copy
                pkt->setHasSharers();
            }
        }
    }
}

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

bool
Cache::access(PacketPtr pkt, CacheBlk *&blk, Cycles &lat,
              PacketList &writebacks)
{

    if (pkt->req->isUncacheable()) {
        assert(pkt->isRequest());

        chatty_assert(!(isReadOnly && pkt->isWrite()),
                      "Should never see a write in a read-only cache %s\n",
                      name());

        DPRINTF(Cache, "%s for %s\n", __func__, pkt->print());

        // flush and invalidate any existing block
        CacheBlk *old_blk(tags->findBlock(pkt->getAddr(), pkt->isSecure()));
        if (old_blk && old_blk->isValid()) {
            BaseCache::evictBlock(old_blk, writebacks);
        }

        blk = nullptr;
        // lookupLatency is the latency in case the request is uncacheable.
        lat = lookupLatency;
        return false;
    }

    return BaseCache::access(pkt, blk, lat, writebacks);
}

void
Cache::doWritebacks(PacketList& writebacks, Tick forward_time)
{
    while (!writebacks.empty()) {
        PacketPtr wbPkt = writebacks.front();
        // We use forwardLatency here because we are copying writebacks to
        // write buffer.

        // Call isCachedAbove for Writebacks, CleanEvicts and
        // WriteCleans to discover if the block is cached above.
        if (isCachedAbove(wbPkt)) {
            if (wbPkt->cmd == MemCmd::CleanEvict) {
                // Delete CleanEvict because cached copies exist above. The
                // packet destructor will delete the request object because
                // this is a non-snoop request packet which does not require a
                // response.
                delete wbPkt;
            } else if (wbPkt->cmd == MemCmd::WritebackClean) {
                // clean writeback, do not send since the block is
                // still cached above
                assert(writebackClean);
                delete wbPkt;
            } else {
                assert(wbPkt->cmd == MemCmd::WritebackDirty ||
                       wbPkt->cmd == MemCmd::WriteClean);
                // Set BLOCK_CACHED flag in Writeback and send below, so that
                // the Writeback does not reset the bit corresponding to this
                // address in the snoop filter below.
                wbPkt->setBlockCached();
                allocateWriteBuffer(wbPkt, forward_time);
            }
        } else {
            // If the block is not cached above, send packet below. Both
            // CleanEvict and Writeback with BLOCK_CACHED flag cleared will
            // reset the bit corresponding to this address in the snoop filter
            // below.
            allocateWriteBuffer(wbPkt, forward_time);
        }
        writebacks.pop_front();
    }
}

void
Cache::doWritebacksAtomic(PacketList& writebacks)
{
    while (!writebacks.empty()) {
        PacketPtr wbPkt = writebacks.front();
        // Call isCachedAbove for both Writebacks and CleanEvicts. If
        // isCachedAbove returns true we set BLOCK_CACHED flag in Writebacks
        // and discard CleanEvicts.
        if (isCachedAbove(wbPkt, false)) {
            if (wbPkt->cmd == MemCmd::WritebackDirty ||
                wbPkt->cmd == MemCmd::WriteClean) {
                // Set BLOCK_CACHED flag in Writeback and send below,
                // so that the Writeback does not reset the bit
                // corresponding to this address in the snoop filter
                // below. We can discard CleanEvicts because cached
                // copies exist above. Atomic mode isCachedAbove
                // modifies packet to set BLOCK_CACHED flag
                memSidePort.sendAtomic(wbPkt);
            }
        } else {
            // If the block is not cached above, send packet below. Both
            // CleanEvict and Writeback with BLOCK_CACHED flag cleared will
            // reset the bit corresponding to this address in the snoop filter
            // below.
            memSidePort.sendAtomic(wbPkt);
        }
        writebacks.pop_front();
        // In case of CleanEvicts, the packet destructor will delete the
        // request object because this is a non-snoop request packet which
        // does not require a response.
        delete wbPkt;
    }
}


void
Cache::recvTimingSnoopResp(PacketPtr pkt)
{
    DPRINTF(Cache, "%s for %s\n", __func__, pkt->print());

    // determine if the response is from a snoop request we created
    // (in which case it should be in the outstandingSnoop), or if we
    // merely forwarded someone else's snoop request
    const bool forwardAsSnoop = outstandingSnoop.find(pkt->req) ==
        outstandingSnoop.end();

    if (!forwardAsSnoop) {
        // the packet came from this cache, so sink it here and do not
        // forward it
        assert(pkt->cmd == MemCmd::HardPFResp);

        outstandingSnoop.erase(pkt->req);

        DPRINTF(Cache, "Got prefetch response from above for addr "
                "%#llx (%s)\n", pkt->getAddr(), pkt->isSecure() ? "s" : "ns");
        recvTimingResp(pkt);
        return;
    }

    // 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);
}

void
Cache::promoteWholeLineWrites(PacketPtr pkt)
{
    // Cache line clearing instructions
    if (doFastWrites && (pkt->cmd == MemCmd::WriteReq) &&
        (pkt->getSize() == blkSize) && (pkt->getOffset(blkSize) == 0) &&
        !pkt->isMaskedWrite()) {
        pkt->cmd = MemCmd::WriteLineReq;
        DPRINTF(Cache, "packet promoted from Write to WriteLineReq\n");
    }
}

void
Cache::handleTimingReqHit(PacketPtr pkt, CacheBlk *blk, Tick request_time)
{
    // should never be satisfying an uncacheable access as we
    // flush and invalidate any existing block as part of the
    // lookup
    assert(!pkt->req->isUncacheable());

    BaseCache::handleTimingReqHit(pkt, blk, request_time);
}

void
Cache::handleTimingReqMiss(PacketPtr pkt, CacheBlk *blk, Tick forward_time,
                           Tick request_time)
{
    if (pkt->req->isUncacheable()) {
        // ignore any existing MSHR if we are dealing with an
        // uncacheable request

        // should have flushed and have no valid block
        assert(!blk || !blk->isValid());

        stats.cmdStats(pkt).mshr_uncacheable[pkt->req->masterId()]++;

        if (pkt->isWrite()) {
            allocateWriteBuffer(pkt, forward_time);
        } else {
            assert(pkt->isRead());

            // uncacheable accesses always allocate a new MSHR

            // Here we are using forward_time, modelling the latency of
            // a miss (outbound) just as forwardLatency, neglecting the
            // lookupLatency component.
            allocateMissBuffer(pkt, forward_time);
        }

        return;
    }

    Addr blk_addr = pkt->getBlockAddr(blkSize);

    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()) {
        assert(pkt->needsResponse());
        assert(pkt->req->hasPaddr());
        assert(!pkt->req->isUncacheable());

        // 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 = std::make_shared<Request>(pkt->req->getPaddr(),
                                                       pkt->req->getSize(),
                                                       pkt->req->getFlags(),
                                                       pkt->req->masterId());
            pf = new Packet(req, pkt->cmd);
            pf->allocate();
            assert(pf->matchAddr(pkt));
            assert(pf->getSize() == pkt->getSize());
        }

        pkt->makeTimingResponse();

        // 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;
    }

    BaseCache::handleTimingReqMiss(pkt, mshr, blk, forward_time, request_time);
}

void
Cache::recvTimingReq(PacketPtr pkt)
{
    DPRINTF(CacheTags, "%s tags:\n%s\n", __func__, tags->print());

    promoteWholeLineWrites(pkt);

    if (pkt->cacheResponding()) {
        // a cache above us (but not where the packet came from) is
        // responding to the request, in other words it has the line
        // in Modified or Owned state
        DPRINTF(Cache, "Cache above responding to %s: not responding\n",
                pkt->print());

        // if the packet needs the block to be writable, and the cache
        // that has promised to respond (setting the cache responding
        // flag) is not providing writable (it is in Owned rather than
        // the Modified state), we know that there may be other Shared
        // copies in the system; go out and invalidate them all
        assert(pkt->needsWritable() && !pkt->responderHadWritable());

        // an upstream cache that had the line in Owned state
        // (dirty, but not writable), is responding and thus
        // transferring the dirty line from one branch of the
        // cache hierarchy to another

        // send out an express snoop and invalidate all other
        // copies (snooping a packet that needs writable is the
        // same as an invalidation), thus turning the Owned line
        // into a Modified line, note that we don't invalidate the
        // block in the current cache or any other cache on the
        // path to memory

        // 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 another cache
        // is responding, because we have found the authorative
        // copy (Modified or Owned) that will supply the right
        // data
        snoop_pkt->setExpressSnoop();
        snoop_pkt->setCacheResponding();

        // 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 snoop packet

        // queue for deletion, as opposed to immediate deletion, as
        // the sending cache is still relying on the packet
        pendingDelete.reset(pkt);

        // no need to take any further action in this particular cache
        // as an upstram cache has already committed to responding,
        // and we have already sent out any express snoops in the
        // section above to ensure all other copies in the system are
        // invalidated
        return;
    }

    BaseCache::recvTimingReq(pkt);
}

PacketPtr
Cache::createMissPacket(PacketPtr cpu_pkt, CacheBlk *blk,
                        bool needsWritable,
                        bool is_whole_line_write) const
{
    // should never see evictions here
    assert(!cpu_pkt->isEviction());

    bool blkValid = blk && blk->isValid();

    if (cpu_pkt->req->isUncacheable() ||
        (!blkValid && cpu_pkt->isUpgrade()) ||
        cpu_pkt->cmd == MemCmd::InvalidateReq || cpu_pkt->isClean()) {
        // uncacheable requests and upgrades from upper-level caches
        // that missed completely just go through as is
        return nullptr;
    }

    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;
    assert(cpu_pkt->cmd != MemCmd::WriteLineReq || is_whole_line_write);
    if (is_whole_line_write) {
        assert(!blkValid || !blk->isWritable());
        // forward as invalidate to all other caches, this gives us
        // the line in Exclusive state, and invalidates all other
        // copies
        cmd = MemCmd::InvalidateReq;
    } else if (blkValid && useUpgrades) {
        // only reason to be here is that blk is read only and we need
        // it to be writable
        assert(needsWritable);
        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 {
        // block is invalid

        // If the request does not need a writable there are two cases
        // where we need to ensure the response will not fetch the
        // block in dirty state:
        // * this cache is read only and it does not perform
        //   writebacks,
        // * this cache is mostly exclusive and will not fill (since
        //   it does not fill it will have to writeback the dirty data
        //   immediately which generates uneccesary writebacks).
        bool force_clean_rsp = isReadOnly || clusivity == Enums::mostly_excl;
        cmd = needsWritable ? MemCmd::ReadExReq :
            (force_clean_rsp ? MemCmd::ReadCleanReq : MemCmd::ReadSharedReq);
    }
    PacketPtr pkt = new Packet(cpu_pkt->req, cmd, blkSize);

    // if there are upstream caches that have already marked the
    // packet as having sharers (not passing writable), pass that info
    // downstream
    if (cpu_pkt->hasSharers() && !needsWritable) {
        // note that cpu_pkt may have spent a considerable time in the
        // MSHR queue and that the information could possibly be out
        // of date, however, there is no harm in conservatively
        // assuming the block has sharers
        pkt->setHasSharers();
        DPRINTF(Cache, "%s: passing hasSharers from %s to %s\n",
                __func__, cpu_pkt->print(), pkt->print());
    }

    // the packet should be block aligned
    assert(pkt->getAddr() == pkt->getBlockAddr(blkSize));

    pkt->allocate();
    DPRINTF(Cache, "%s: created %s from %s\n", __func__, pkt->print(),
            cpu_pkt->print());
    return pkt;
}


Cycles
Cache::handleAtomicReqMiss(PacketPtr pkt, CacheBlk *&blk,
                           PacketList &writebacks)
{
    // deal with the packets that go through the write path of
    // the cache, i.e. any evictions and writes
    if (pkt->isEviction() || pkt->cmd == MemCmd::WriteClean ||
        (pkt->req->isUncacheable() && pkt->isWrite())) {
        Cycles latency = ticksToCycles(memSidePort.sendAtomic(pkt));

        // at this point, if the request was an uncacheable write
        // request, it has been satisfied by a memory below and the
        // packet carries the response back
        assert(!(pkt->req->isUncacheable() && pkt->isWrite()) ||
               pkt->isResponse());

        return latency;
    }

    // only misses left

    PacketPtr bus_pkt = createMissPacket(pkt, blk, pkt->needsWritable(),
                                         pkt->isWholeLineWrite(blkSize));

    bool is_forward = (bus_pkt == nullptr);

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

    DPRINTF(Cache, "%s: Sending an atomic %s\n", __func__,
            bus_pkt->print());

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

    Cycles latency = ticksToCycles(memSidePort.sendAtomic(bus_pkt));

    bool is_invalidate = bus_pkt->isInvalidate();

    // We are now dealing with the response handling
    DPRINTF(Cache, "%s: Receive response: %s in state %i\n", __func__,
            bus_pkt->print(), 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->isWholeLineWrite(blkSize)) {
                // note the use of pkt, not bus_pkt here.

                // write-line request to the cache that promoted
                // the write to a whole line
                const bool allocate = allocOnFill(pkt->cmd) &&
                    (!writeAllocator || writeAllocator->allocate());
                blk = handleFill(bus_pkt, blk, writebacks, allocate);
                assert(blk != NULL);
                is_invalidate = false;
                satisfyRequest(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,
                                 allocOnFill(pkt->cmd));
                satisfyRequest(pkt, blk);
                maintainClusivity(pkt->fromCache(), blk);
            } else {
                // we're satisfying the upstream request without
                // modifying cache state, e.g., a write-through
                pkt->makeAtomicResponse();
            }
        }
        delete bus_pkt;
    }

    if (is_invalidate && blk && blk->isValid()) {
        invalidateBlock(blk);
    }

    return latency;
}

Tick
Cache::recvAtomic(PacketPtr pkt)
{
    promoteWholeLineWrites(pkt);

    // follow the same flow as in recvTimingReq, and check if a cache
    // above us is responding
    if (pkt->cacheResponding()) {
        assert(!pkt->req->isCacheInvalidate());
        DPRINTF(Cache, "Cache above responding to %s: not responding\n",
                pkt->print());

        // if a cache is responding, and it had the line in Owned
        // rather than Modified state, we need to invalidate any
        // copies that are not on the same path to memory
        assert(pkt->needsWritable() && !pkt->responderHadWritable());

        return memSidePort.sendAtomic(pkt);
    }

    return BaseCache::recvAtomic(pkt);
}


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


void
Cache::serviceMSHRTargets(MSHR *mshr, const PacketPtr pkt, CacheBlk *blk)
{
    QueueEntry::Target *initial_tgt = mshr->getTarget();
    // First offset for critical word first calculations
    const int initial_offset = initial_tgt->pkt->getOffset(blkSize);

    const bool is_error = pkt->isError();
    // allow invalidation responses originating from write-line
    // requests to be discarded
    bool is_invalidate = pkt->isInvalidate() &&
        !mshr->wasWholeLineWrite;

    MSHR::TargetList targets = mshr->extractServiceableTargets(pkt);
    for (auto &target: targets) {
        Packet *tgt_pkt = target.pkt;
        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 (tgt_pkt->cmd.isSWPrefetch()) {
                if (tgt_pkt->needsWritable()) {
                    // All other copies of the block were invalidated and we
                    // have an exclusive copy.

                    // The coherence protocol assumes that if we fetched an
                    // exclusive copy of the block, we have the intention to
                    // modify it. Therefore the MSHR for the PrefetchExReq has
                    // been the point of ordering and this cache has commited
                    // to respond to snoops for the block.
                    //
                    // In most cases this is true anyway - a PrefetchExReq
                    // will be followed by a WriteReq. However, if that
                    // doesn't happen, the block is not marked as dirty and
                    // the cache doesn't respond to snoops that has committed
                    // to do so.
                    //
                    // To avoid deadlocks in cases where there is a snoop
                    // between the PrefetchExReq and the expected WriteReq, we
                    // proactively mark the block as Dirty.
                    assert(blk);
                    blk->status |= BlkDirty;

                    panic_if(isReadOnly, "Prefetch exclusive requests from "
                            "read-only cache %s\n", name());
                }

                // 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 tgt_pkt;
                break; // skip response
            }

            // unlike the other packet flows, where data is found in other
            // caches or memory and brought back, write-line requests 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 (tgt_pkt->cmd == MemCmd::WriteLineReq) {
                assert(!is_error);
                assert(blk);
                assert(blk->isWritable());
            }

            if (blk && blk->isValid() && !mshr->isForward) {
                satisfyRequest(tgt_pkt, blk, true, mshr->hasPostDowngrade());

                // How many bytes past the first request is this one
                int transfer_offset =
                    tgt_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(!tgt_pkt->req->isUncacheable());

                assert(tgt_pkt->req->masterId() < system->maxMasters());
                stats.cmdStats(tgt_pkt)
                    .missLatency[tgt_pkt->req->masterId()] +=
                    completion_time - target.recvTime;
            } else if (pkt->cmd == MemCmd::UpgradeFailResp) {
                // failed StoreCond upgrade
                assert(tgt_pkt->cmd == MemCmd::StoreCondReq ||
                       tgt_pkt->cmd == MemCmd::StoreCondFailReq ||
                       tgt_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;
                tgt_pkt->req->setExtraData(0);
            } else {
                // We are about to send a response to a cache above
                // that asked for an invalidation; we need to
                // invalidate our copy immediately as the most
                // up-to-date copy of the block will now be in the
                // cache above. It will also prevent this cache from
                // responding (if the block was previously dirty) to
                // snoops as they should snoop the caches above where
                // they will get the response from.
                if (is_invalidate && blk && blk->isValid()) {
                    invalidateBlock(blk);
                }
                // 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) {
                    // sanity check
                    assert(pkt->matchAddr(tgt_pkt));
                    assert(pkt->getSize() >= tgt_pkt->getSize());

                    tgt_pkt->setData(pkt->getConstPtr<uint8_t>());
                }

                // this response did not allocate here and therefore
                // it was not consumed, make sure that any flags are
                // carried over to cache above
                tgt_pkt->copyResponderFlags(pkt);
            }
            tgt_pkt->makeTimingResponse();
            // if this packet is an error copy that to the new packet
            if (is_error)
                tgt_pkt->copyError(pkt);
            if (tgt_pkt->cmd == MemCmd::ReadResp &&
                (is_invalidate || mshr->hasPostInvalidate())) {
                // If intermediate cache got ReadRespWithInvalidate,
                // propagate that.  Response should not have
                // isInvalidate() set otherwise.
                tgt_pkt->cmd = MemCmd::ReadRespWithInvalidate;
                DPRINTF(Cache, "%s: updated cmd to %s\n", __func__,
                        tgt_pkt->print());
            }
            // Reset the bus additional time as it is now accounted for
            tgt_pkt->headerDelay = tgt_pkt->payloadDelay = 0;
            cpuSidePort.schedTimingResp(tgt_pkt, completion_time);
            break;

          case MSHR::Target::FromPrefetcher:
            assert(tgt_pkt->cmd == MemCmd::HardPFReq);
            if (blk)
                blk->status |= BlkHWPrefetched;
            delete tgt_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");
            // If the response is invalidating, a snooping target can
            // be satisfied if it is also invalidating. If the reponse is, not
            // only invalidating, but more specifically an InvalidateResp and
            // the MSHR was created due to an InvalidateReq then a cache above
            // is waiting to satisfy a WriteLineReq. In this case even an
            // non-invalidating snoop is added as a target here since this is
            // the ordering point. When the InvalidateResp reaches this cache,
            // the snooping target will snoop further the cache above with the
            // WriteLineReq.
            assert(!is_invalidate || pkt->cmd == MemCmd::InvalidateResp ||
                   pkt->req->isCacheMaintenance() ||
                   mshr->hasPostInvalidate());
            handleSnoop(tgt_pkt, blk, true, true, mshr->hasPostInvalidate());
            break;

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

    maintainClusivity(targets.hasFromCache, blk);

    if (blk && blk->isValid()) {
        // an invalidate response stemming from a write line request
        // should not invalidate the block, so check if the
        // invalidation should be discarded
        if (is_invalidate || mshr->hasPostInvalidate()) {
            invalidateBlock(blk);
        } else if (mshr->hasPostDowngrade()) {
            blk->status &= ~BlkWritable;
        }
    }
}

PacketPtr
Cache::evictBlock(CacheBlk *blk)
{
    PacketPtr pkt = (blk->isDirty() || writebackClean) ?
        writebackBlk(blk) : cleanEvictBlk(blk);

    invalidateBlock(blk);

    return pkt;
}

PacketPtr
Cache::cleanEvictBlk(CacheBlk *blk)
{
    assert(!writebackClean);
    assert(blk && blk->isValid() && !blk->isDirty());

    // Creating a zero sized write, a message to the snoop filter
    RequestPtr req = std::make_shared<Request>(
        regenerateBlkAddr(blk), blkSize, 0, Request::wbMasterId);

    if (blk->isSecure())
        req->setFlags(Request::SECURE);

    req->taskId(blk->task_id);

    PacketPtr pkt = new Packet(req, MemCmd::CleanEvict);
    pkt->allocate();
    DPRINTF(Cache, "Create CleanEvict %s\n", pkt->print());

    return pkt;
}

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

void
Cache::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\n", __func__, req_pkt->print());
    // 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->req->isUncacheable() || req_pkt->isInvalidate() ||
           pkt->hasSharers());
    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 cacheResponding in both cases, but
        // in the latter case cacheResponding 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(CacheVerbose, "%s: created response: %s tick: %lu\n", __func__,
            pkt->print(), forward_time);
    memSidePort.schedTimingSnoopResp(pkt, forward_time);
}

uint32_t
Cache::handleSnoop(PacketPtr pkt, CacheBlk *blk, bool is_timing,
                   bool is_deferred, bool pending_inval)
{
    DPRINTF(CacheVerbose, "%s: for %s\n", __func__, pkt->print());
    // 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_writable = pkt->needsWritable();

    // at the moment we could get an uncacheable write which does not
    // have the invalidate flag, and we need a suitable way of dealing
    // with this case
    panic_if(invalidate && pkt->req->isUncacheable(),
             "%s got an invalidating uncacheable snoop request %s",
             name(), pkt->print());

    uint32_t snoop_delay = 0;

    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)
        if (is_timing) {
            // copy the packet so that we can clear any flags before
            // forwarding it upwards, we also allocate data (passing
            // the pointer along in case of static data), in case
            // there is a snoop hit in upper levels
            Packet snoopPkt(pkt, true, true);
            snoopPkt.setExpressSnoop();
            // the snoop packet does not need to wait any additional
            // time
            snoopPkt.headerDelay = snoopPkt.payloadDelay = 0;
            cpuSidePort.sendTimingSnoopReq(&snoopPkt);

            // add the header delay (including crossbar and snoop
            // delays) of the upward snoop to the snoop delay for this
            // cache
            snoop_delay += snoopPkt.headerDelay;

            // If this request is a prefetch or clean evict and an upper level
            // signals block present, make sure to propagate the block
            // presence to the requester.
            if (snoopPkt.isBlockCached()) {
                pkt->setBlockCached();
            }
            // If the request was satisfied by snooping the cache
            // above, mark the original packet as satisfied too.
            if (snoopPkt.satisfied()) {
                pkt->setSatisfied();
            }

            // Copy over flags from the snoop response to make sure we
            // inform the final destination
            pkt->copyResponderFlags(&snoopPkt);
        } else {
            bool already_responded = pkt->cacheResponding();
            cpuSidePort.sendAtomicSnoop(pkt);
            if (!already_responded && pkt->cacheResponding()) {
                // cache-to-cache response from some upper cache:
                // forward response to original requester
                assert(pkt->isResponse());
            }
        }
    }

    bool respond = false;
    bool blk_valid = blk && blk->isValid();
    if (pkt->isClean()) {
        if (blk_valid && blk->isDirty()) {
            DPRINTF(CacheVerbose, "%s: packet (snoop) %s found block: %s\n",
                    __func__, pkt->print(), blk->print());
            PacketPtr wb_pkt = writecleanBlk(blk, pkt->req->getDest(), pkt->id);
            PacketList writebacks;
            writebacks.push_back(wb_pkt);

            if (is_timing) {
                // anything that is merely forwarded pays for the forward
                // latency and the delay provided by the crossbar
                Tick forward_time = clockEdge(forwardLatency) +
                    pkt->headerDelay;
                doWritebacks(writebacks, forward_time);
            } else {
                doWritebacksAtomic(writebacks);
            }
            pkt->setSatisfied();
        }
    } else if (!blk_valid) {
        DPRINTF(CacheVerbose, "%s: snoop miss for %s\n", __func__,
                pkt->print());
        if (is_deferred) {
            // we no longer have the block, and will not respond, but a
            // packet was allocated in MSHR::handleSnoop and we have
            // to delete it
            assert(pkt->needsResponse());

            // we have passed the block to a cache upstream, that
            // cache should be responding
            assert(pkt->cacheResponding());

            delete pkt;
        }
        return snoop_delay;
    } else {
        DPRINTF(Cache, "%s: snoop hit for %s, old state is %s\n", __func__,
                pkt->print(), 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. We respond to all snoops that need
        // responses provided we have the block in dirty state. The
        // invalidation itself is taken care of below. We don't respond to
        // cache maintenance operations as this is done by the destination
        // xbar.
        respond = blk->isDirty() && pkt->needsResponse();

        chatty_assert(!(isReadOnly && blk->isDirty()), "Should never have "
                      "a dirty block in a read-only cache %s\n", name());
    }

    // 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->mustCheckAbove()) {
        DPRINTF(Cache, "Found addr %#llx in upper level cache for snoop %s "
                "from lower cache\n", pkt->getAddr(), pkt->print());
        pkt->setBlockCached();
        return snoop_delay;
    }

    if (pkt->isRead() && !invalidate) {
        // reading without requiring the line in a writable state
        assert(!needs_writable);
        pkt->setHasSharers();

        // if the requesting packet is uncacheable, retain the line in
        // the current state, otherwhise unset the writable flag,
        // which means we go from Modified to Owned (and will respond
        // below), remain in Owned (and will respond below), from
        // Exclusive to Shared, or remain in Shared
        if (!pkt->req->isUncacheable())
            blk->status &= ~BlkWritable;
        DPRINTF(Cache, "new state is %s\n", blk->print());
    }

    if (respond) {
        // prevent anyone else from responding, cache as well as
        // memory, and also prevent any memory from even seeing the
        // request
        pkt->setCacheResponding();
        if (!pkt->isClean() && blk->isWritable()) {
            // inform the cache hierarchy that this cache had the line
            // in the Modified state so that we avoid unnecessary
            // invalidations (see Packet::setResponderHadWritable)
            pkt->setResponderHadWritable();

            // in the case of an uncacheable request there is no point
            // in setting the responderHadWritable flag, but since the
            // recipient does not care there is no harm in doing so
        } else {
            // if the packet has needsWritable set we invalidate our
            // copy below and all other copies will be invalidates
            // through express snoops, and if needsWritable is not set
            // we already called setHasSharers above
        }

        // if we are returning a writable and dirty (Modified) line,
        // we should be invalidating the line
        panic_if(!invalidate && !pkt->hasSharers(),
                 "%s is passing a Modified line through %s, "
                 "but keeping the block", name(), pkt->print());

        if (is_timing) {
            doTimingSupplyResponse(pkt, blk->data, is_deferred, pending_inval);
        } else {
            pkt->makeAtomicResponse();
            // packets such as upgrades do not actually have any data
            // payload
            if (pkt->hasData())
                pkt->setDataFromBlock(blk->data, blkSize);
        }

        // When a block is compressed, it must first be decompressed before
        // being read, and this increases the snoop delay.
        if (compressor && pkt->isRead()) {
            snoop_delay += compressor->getDecompressionLatency(blk);
        }
    }

    if (!respond && is_deferred) {
        assert(pkt->needsResponse());
        delete pkt;
    }

    // Do this last in case it deallocates block data or something
    // like that
    if (blk_valid && invalidate) {
        invalidateBlock(blk);
        DPRINTF(Cache, "new state is %s\n", blk->print());
    }

    return snoop_delay;
}


void
Cache::recvTimingSnoopReq(PacketPtr pkt)
{
    DPRINTF(CacheVerbose, "%s: for %s\n", __func__, pkt->print());

    // no need to snoop requests that are not in range
    if (!inRange(pkt->getAddr())) {
        return;
    }

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

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

    // Update the latency cost of the snoop so that the crossbar can
    // account for it. Do not overwrite what other neighbouring caches
    // have already done, rather take the maximum. The update is
    // tentative, for cases where we return before an upward snoop
    // happens below.
    pkt->snoopDelay = std::max<uint32_t>(pkt->snoopDelay,
                                         lookupLatency * clockPeriod());

    // Inform request(Prefetch, CleanEvict or Writeback) from below of
    // MSHR hit, set setBlockCached.
    if (mshr && pkt->mustCheckAbove()) {
        DPRINTF(Cache, "Setting block cached for %s from lower cache on "
                "mshr hit\n", pkt->print());
        pkt->setBlockCached();
        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
    WriteQueueEntry *wb_entry = writeBuffer.findMatch(blk_addr, is_secure);
    if (wb_entry) {
        DPRINTF(Cache, "Snoop hit in writeback to addr %#llx (%s)\n",
                pkt->getAddr(), is_secure ? "s" : "ns");
        // Expect to see only Writebacks and/or CleanEvicts here, both of
        // which should not be generated for uncacheable data.
        assert(!wb_entry->isUncacheable());
        // There should only be a single request responsible for generating
        // Writebacks/CleanEvicts.
        assert(wb_entry->getNumTargets() == 1);
        PacketPtr wb_pkt = wb_entry->getTarget()->pkt;
        assert(wb_pkt->isEviction() || wb_pkt->cmd == MemCmd::WriteClean);

        if (pkt->isEviction()) {
            // if the block is found in the write queue, set the BLOCK_CACHED
            // flag for Writeback/CleanEvict snoop. On return the snoop will
            // propagate the BLOCK_CACHED flag in Writeback packets and prevent
            // any CleanEvicts from travelling down the memory hierarchy.
            pkt->setBlockCached();
            DPRINTF(Cache, "%s: Squashing %s from lower cache on writequeue "
                    "hit\n", __func__, pkt->print());
            return;
        }

        // conceptually writebacks are no different to other blocks in
        // this cache, so the behaviour is modelled after handleSnoop,
        // the difference being that instead of querying the block
        // state to determine if it is dirty and writable, we use the
        // command and fields of the writeback packet
        bool respond = wb_pkt->cmd == MemCmd::WritebackDirty &&
            pkt->needsResponse();
        bool have_writable = !wb_pkt->hasSharers();
        bool invalidate = pkt->isInvalidate();

        if (!pkt->req->isUncacheable() && pkt->isRead() && !invalidate) {
            assert(!pkt->needsWritable());
            pkt->setHasSharers();
            wb_pkt->setHasSharers();
        }

        if (respond) {
            pkt->setCacheResponding();

            if (have_writable) {
                pkt->setResponderHadWritable();
            }

            doTimingSupplyResponse(pkt, wb_pkt->getConstPtr<uint8_t>(),
                                   false, false);
        }

        if (invalidate && wb_pkt->cmd != MemCmd::WriteClean) {
            // Invalidation trumps our writeback... discard here
            // Note: markInService will remove entry from writeback buffer.
            markInService(wb_entry);
            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.
    uint32_t snoop_delay = handleSnoop(pkt, blk, true, false, false);

    // Override what we did when we first saw the snoop, as we now
    // also have the cost of the upwards snoops to account for
    pkt->snoopDelay = std::max<uint32_t>(pkt->snoopDelay, snoop_delay +
                                         lookupLatency * clockPeriod());
}

Tick
Cache::recvAtomicSnoop(PacketPtr pkt)
{
    // no need to snoop requests that are not in range.
    if (!inRange(pkt->getAddr())) {
        return 0;
    }

    CacheBlk *blk = tags->findBlock(pkt->getAddr(), pkt->isSecure());
    uint32_t snoop_delay = handleSnoop(pkt, blk, false, false, false);
    return snoop_delay + lookupLatency * clockPeriod();
}

bool
Cache::isCachedAbove(PacketPtr pkt, bool is_timing)
{
    if (!forwardSnoops)
        return false;
    // Mirroring the flow of HardPFReqs, the cache sends CleanEvict and
    // Writeback snoops into upper level caches to check for copies of the
    // same block. Using the BLOCK_CACHED flag with the Writeback/CleanEvict
    // packet, the cache can inform the crossbar below of presence or absence
    // of the block.
    if (is_timing) {
        Packet snoop_pkt(pkt, true, false);
        snoop_pkt.setExpressSnoop();
        // Assert that packet is either Writeback or CleanEvict and not a
        // prefetch request because prefetch requests need an MSHR and may
        // generate a snoop response.
        assert(pkt->isEviction() || pkt->cmd == MemCmd::WriteClean);
        snoop_pkt.senderState = nullptr;
        cpuSidePort.sendTimingSnoopReq(&snoop_pkt);
        // Writeback/CleanEvict snoops do not generate a snoop response.
        assert(!(snoop_pkt.cacheResponding()));
        return snoop_pkt.isBlockCached();
    } else {
        cpuSidePort.sendAtomicSnoop(pkt);
        return pkt->isBlockCached();
    }
}

bool
Cache::sendMSHRQueuePacket(MSHR* mshr)
{
    assert(mshr);

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

    if (tgt_pkt->cmd == MemCmd::HardPFReq && forwardSnoops) {
        DPRINTF(Cache, "%s: MSHR %s\n", __func__, tgt_pkt->print());

        // we should never have hardware prefetches to allocated
        // blocks
        assert(!tags->findBlock(mshr->blkAddr, mshr->isSecure));

        // 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();
        // We are sending this packet upwards, but if it hits we will
        // get a snoop response that we end up treating just like a
        // normal response, hence it needs the MSHR as its sender
        // state
        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 cacheResponding before
        // prefetchSquashed. If another cache has committed to
        // responding, it will be sending a dirty response which will
        // arrive at the MSHR allocated for this request. Checking the
        // prefetchSquash first may result in the MSHR being
        // prematurely deallocated.
        if (snoop_pkt.cacheResponding()) {
            auto M5_VAR_USED r = outstandingSnoop.insert(snoop_pkt.req);
            assert(r.second);

            // if we are getting a snoop response with no sharers it
            // will be allocated as Modified
            bool pending_modified_resp = !snoop_pkt.hasSharers();
            markInService(mshr, pending_modified_resp);

            DPRINTF(Cache, "Upward snoop of prefetch for addr"
                    " %#x (%s) hit\n",
                    tgt_pkt->getAddr(), tgt_pkt->isSecure()? "s": "ns");
            return false;
        }

        if (snoop_pkt.isBlockCached()) {
            DPRINTF(Cache, "Block present, prefetch squashed by cache.  "
                    "Deallocating mshr target %#x.\n",
                    mshr->blkAddr);

            // Deallocate the mshr target
            if (mshrQueue.forceDeallocateTarget(mshr)) {
                // Clear block if this deallocation resulted freed an
                // mshr when all had previously been utilized
                clearBlocked(Blocked_NoMSHRs);
            }

            // given that no response is expected, delete Request and Packet
            delete tgt_pkt;

            return false;
        }
    }

    return BaseCache::sendMSHRQueuePacket(mshr);
}

Cache*
CacheParams::create()
{
    assert(tags);
    assert(replacement_policy);

    return new Cache(this);
}