summaryrefslogtreecommitdiff
path: root/ext2fs/pager.c
blob: 456b582c1578a0385492d9c8d891296e13f2bf29 (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
/* Pager for ext2fs

   Copyright (C) 1994,95,96,97,98,99,2000,02 Free Software Foundation, Inc.

   Converted for ext2fs by Miles Bader <miles@gnu.org>

   This program is free software; you can redistribute it and/or
   modify it under the terms of the GNU General Public License as
   published by the Free Software Foundation; either version 2, or (at
   your option) any later version.

   This program is distributed in the hope that it will be useful, but
   WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */

#include <unistd.h>
#include <string.h>
#include <errno.h>
#include <error.h>
#include <hurd/store.h>
#include "ext2fs.h"

/* XXX */
#include "../libpager/priv.h"

/* A ports bucket to hold disk pager ports.  */
struct port_bucket *disk_pager_bucket;

/* A ports bucket to hold file pager ports.  */
struct port_bucket *file_pager_bucket;

/* Stores a reference to the requests instance used by the file pager so its
   worker threads can be inhibited and resumed.  */
struct pager_requests *file_pager_requests;

pthread_spinlock_t node_to_page_lock = PTHREAD_SPINLOCK_INITIALIZER;


#ifdef DONT_CACHE_MEMORY_OBJECTS
#define MAY_CACHE 0
#else
#define MAY_CACHE 1
#endif

#define STATS

#ifdef STATS
struct ext2fs_pager_stats
{
  pthread_spinlock_t lock;

  unsigned long disk_pageins;
  unsigned long disk_pageouts;

  unsigned long file_pageins;
  unsigned long file_pagein_reads; /* Device reads done by file pagein */
  unsigned long file_pagein_freed_bufs;	/* Discarded pages */
  unsigned long file_pagein_alloced_bufs; /* Allocated pages */

  unsigned long file_pageouts;

  unsigned long file_page_unlocks;
  unsigned long file_grows;
};

static struct ext2fs_pager_stats ext2s_pager_stats =
  { .lock = PTHREAD_SPINLOCK_INITIALIZER };

#define STAT_INC(field)							      \
do { pthread_spin_lock (&ext2s_pager_stats.lock);			      \
     ext2s_pager_stats.field++;						      \
     pthread_spin_unlock (&ext2s_pager_stats.lock); } while (0)

#else /* !STATS */
#define STAT_INC(field) /* nop */0
#endif /* STATS */

static void
disk_cache_info_free_push (struct disk_cache_info *p);

#define FREE_PAGE_BUFS 24

/* Returns a single page page-aligned buffer.  */
static void *
get_page_buf ()
{
  static pthread_mutex_t free_page_bufs_lock = PTHREAD_MUTEX_INITIALIZER;
  static void *free_page_bufs;
  static int num_free_page_bufs;
  void *buf;

  pthread_mutex_lock (&free_page_bufs_lock);
  if (num_free_page_bufs > 0)
    {
      buf = free_page_bufs;
      num_free_page_bufs --;
      if (num_free_page_bufs > 0)
	free_page_bufs += vm_page_size;
#ifndef NDEBUG
      else
	free_page_bufs = 0;
#endif /* ! NDEBUG */
    }
  else
    {
      assert (free_page_bufs == 0);
      buf = mmap (0, vm_page_size * FREE_PAGE_BUFS,
		  PROT_READ|PROT_WRITE, MAP_ANON, 0, 0);
      if (buf == MAP_FAILED)
	buf = 0;
      else
	{
	  free_page_bufs = buf + vm_page_size;
	  num_free_page_bufs = FREE_PAGE_BUFS - 1;
	}
    }

  pthread_mutex_unlock (&free_page_bufs_lock);
  return buf;
}

/* Frees a block returned by get_page_buf.  */
static inline void
free_page_buf (void *buf)
{
  munmap (buf, vm_page_size);
}

/* Find the location on disk of page OFFSET in NODE.  Return the disk block
   in BLOCK (if unallocated, then return 0).  If *LOCK is 0, then a reader
   lock is acquired on NODE's ALLOC_LOCK before doing anything, and left
   locked after the return -- even if an error is returned.  0 is returned
   on success otherwise an error code.  */
static error_t
find_block (struct node *node, vm_offset_t offset,
	    block_t *block, pthread_rwlock_t **lock)
{
  error_t err;

  if (!*lock)
    {
      *lock = &diskfs_node_disknode (node)->alloc_lock;
      pthread_rwlock_rdlock (*lock);
    }

  if (offset + block_size > node->allocsize)
    return EIO;

  err = ext2_getblk (node, offset >> log2_block_size, 0, block);
  if (err == EINVAL)
    /* Don't barf yet if the node is unallocated.  */
    {
      *block = 0;
      err = 0;
    }

  return err;
}

/* Read one page for the pager backing NODE at offset PAGE, into BUF.  This
   may need to read several filesystem blocks to satisfy one page, and tries
   to consolidate the i/o if possible.  */
static error_t
file_pager_read_page (struct node *node, vm_offset_t page,
		      void **buf, int *writelock)
{
  error_t err;
  int offs = 0;
  int partial = 0;		/* A page truncated by the EOF.  */
  pthread_rwlock_t *lock = NULL;
  int left = vm_page_size;
  block_t pending_blocks = 0;
  int num_pending_blocks = 0;

  ext2_debug ("reading inode %llu page %lu[%u]",
	      node->cache_id, page, vm_page_size);

  /* Read the NUM_PENDING_BLOCKS blocks in PENDING_BLOCKS, into the buffer
     pointed to by BUF (allocating it if necessary) at offset OFFS.  OFFS in
     adjusted by the amount read, and NUM_PENDING_BLOCKS is zeroed.  Any read
     error is returned.  */
  error_t do_pending_reads ()
    {
      if (num_pending_blocks > 0)
	{
	  store_offset_t dev_block = (store_offset_t) pending_blocks
	    << log2_dev_blocks_per_fs_block;
	  size_t amount = num_pending_blocks << log2_block_size;
	  /* The buffer we try to read into; on the first read, we pass in a
	     size of zero, so that the read is guaranteed to allocate a new
	     buffer, otherwise, we try to read directly into the tail of the
	     buffer we've already got.  */
	  void *new_buf = *buf + offs;
	  size_t new_len = offs == 0 ? 0 : vm_page_size - offs;

	  STAT_INC (file_pagein_reads);

	  err = store_read (store, dev_block, amount, &new_buf, &new_len);
	  if (err)
	    return err;
	  else if (amount != new_len)
	    return EIO;

	  if (new_buf != *buf + offs)
	    {
	      /* The read went into a different buffer than the one we
                 passed. */
	      if (offs == 0)
		/* First read, make the returned page be our buffer.  */
		*buf = new_buf;
	      else
		/* We've already got some buffer, so copy into it.  */
		{
		  memcpy (*buf + offs, new_buf, new_len);
		  free_page_buf (new_buf); /* Return NEW_BUF to our pool.  */
		  STAT_INC (file_pagein_freed_bufs);
		}
	    }

	  offs += new_len;
	  num_pending_blocks = 0;
	}

      return 0;
    }

  STAT_INC (file_pageins);

  *writelock = 0;

  if (page >= node->allocsize)
    {
      err = EIO;
      left = 0;
    }
  else if (page + left > node->allocsize)
    {
      left = node->allocsize - page;
      partial = 1;
    }

  while (left > 0)
    {
      block_t block;

      err = find_block (node, page, &block, &lock);
      if (err)
	break;

      if (block != pending_blocks + num_pending_blocks)
	{
	  err = do_pending_reads ();
	  if (err)
	    break;
	  pending_blocks = block;
	}

      if (block == 0)
	/* Reading unallocated block, just make a zero-filled one.  */
	{
	  *writelock = 1;
	  if (offs == 0)
	    /* No page allocated to read into yet.  */
	    {
	      *buf = get_page_buf ();
	      if (! *buf)
		break;
	      STAT_INC (file_pagein_alloced_bufs);
	    }
	  memset (*buf + offs, 0, block_size);
	  offs += block_size;
	}
      else
	num_pending_blocks++;

      page += block_size;
      left -= block_size;
    }

  if (!err && num_pending_blocks > 0)
    err = do_pending_reads();

  if (!err && partial && !*writelock)
    diskfs_node_disknode (node)->last_page_partially_writable = 1;

  if (lock)
    pthread_rwlock_unlock (lock);

  return err;
}

struct pending_blocks
{
  /* The block number of the first of the blocks.  */
  block_t block;
  /* How many blocks we have.  */
  off_t num;
  /* A (page-aligned) buffer pointing to the data we're dealing with.  */
  void *buf;
  /* And an offset into BUF.  */
  int offs;
};

/* Write the any pending blocks in PB.  */
static error_t
pending_blocks_write (struct pending_blocks *pb)
{
  if (pb->num > 0)
    {
      error_t err;
      store_offset_t dev_block = (store_offset_t) pb->block
	<< log2_dev_blocks_per_fs_block;
      size_t length = pb->num << log2_block_size, amount;

      ext2_debug ("writing block %u[%ld]", pb->block, pb->num);

      if (pb->offs > 0)
	/* Put what we're going to write into a page-aligned buffer.  */
	{
	  void *page_buf = get_page_buf ();
	  memcpy ((void *)page_buf, pb->buf + pb->offs, length);
	  err = store_write (store, dev_block, page_buf, length, &amount);
	  free_page_buf (page_buf);
	}
      else
	err = store_write (store, dev_block, pb->buf, length, &amount);
      if (err)
	return err;
      else if (amount != length)
	return EIO;

      pb->offs += length;
      pb->num = 0;
    }

  return 0;
}

static void
pending_blocks_init (struct pending_blocks *pb, void *buf)
{
  pb->buf = buf;
  pb->block = 0;
  pb->num = 0;
  pb->offs = 0;
}

/* Skip writing the next block in PB's buffer (writing out any previous
   blocks if necessary).  */
static error_t
pending_blocks_skip (struct pending_blocks *pb)
{
  error_t err = pending_blocks_write (pb);
  pb->offs += block_size;
  return err;
}

/* Add the disk block BLOCK to the list of destination disk blocks pending in
   PB.  */
static error_t
pending_blocks_add (struct pending_blocks *pb, block_t block)
{
  if (block != pb->block + pb->num)
    {
      error_t err = pending_blocks_write (pb);
      if (err)
	return err;
      pb->block = block;
    }
  pb->num++;
  return 0;
}

/* Write one page for the pager backing NODE, at OFFSET, into BUF.  This
   may need to write several filesystem blocks to satisfy one page, and tries
   to consolidate the i/o if possible.  */
static error_t
file_pager_write_page (struct node *node, vm_offset_t offset, void *buf)
{
  error_t err = 0;
  struct pending_blocks pb;
  pthread_rwlock_t *lock = &diskfs_node_disknode (node)->alloc_lock;
  block_t block;
  int left = vm_page_size;

  pending_blocks_init (&pb, buf);

  /* Holding diskfs_node_disknode (node)->alloc_lock effectively locks NODE->allocsize,
     at least for the cases we care about: pager_unlock_page,
     diskfs_grow and diskfs_truncate.  */
  pthread_rwlock_rdlock (&diskfs_node_disknode (node)->alloc_lock);

  if (offset >= node->allocsize)
    left = 0;
  else if (offset + left > node->allocsize)
    left = node->allocsize - offset;

  ext2_debug ("writing inode %d page %d[%d]", node->cache_id, offset, left);

  STAT_INC (file_pageouts);

  while (left > 0)
    {
      err = find_block (node, offset, &block, &lock);
      if (err)
	break;
      assert (block);
      pending_blocks_add (&pb, block);
      offset += block_size;
      left -= block_size;
    }

  if (!err)
    pending_blocks_write (&pb);

  pthread_rwlock_unlock (&diskfs_node_disknode (node)->alloc_lock);

  return err;
}

static error_t
disk_pager_read_page (vm_offset_t page, void **buf, int *writelock)
{
  error_t err;
  size_t length = vm_page_size, read = 0;
  store_offset_t offset = page, dev_end = store->size;
  int index = offset >> log2_block_size;

  pthread_mutex_lock (&disk_cache_lock);
  offset = ((store_offset_t) disk_cache_info[index].block << log2_block_size)
    + offset % block_size;
  disk_cache_info[index].flags |= DC_INCORE;
  disk_cache_info[index].flags &=~ DC_UNTOUCHED;
#ifdef DEBUG_DISK_CACHE
  disk_cache_info[index].last_read = disk_cache_info[index].block;
  disk_cache_info[index].last_read_xor
    = disk_cache_info[index].block ^ DISK_CACHE_LAST_READ_XOR;
#endif
  pthread_mutex_unlock (&disk_cache_lock);

  ext2_debug ("(%lld)", offset >> log2_block_size);

  if (offset + vm_page_size > dev_end)
    length = dev_end - offset;

  err = store_read (store, offset >> store->log2_block_size, length,
		    buf, &read);
  if (read != length)
    return EIO;
  if (!err && length != vm_page_size)
    memset ((void *)(*buf + length), 0, vm_page_size - length);

  *writelock = 0;

  return err;
}

static error_t
disk_pager_write_page (vm_offset_t page, void *buf)
{
  error_t err = 0;
  size_t length = vm_page_size, amount;
  store_offset_t offset = page, dev_end = store->size;
  int index = offset >> log2_block_size;

  pthread_mutex_lock (&disk_cache_lock);
  assert (disk_cache_info[index].block != DC_NO_BLOCK);
  offset = ((store_offset_t) disk_cache_info[index].block << log2_block_size)
    + offset % block_size;
#ifdef DEBUG_DISK_CACHE			/* Not strictly needed.  */
  assert ((disk_cache_info[index].last_read ^ DISK_CACHE_LAST_READ_XOR)
	  == disk_cache_info[index].last_read_xor);
  assert (disk_cache_info[index].last_read
	  == disk_cache_info[index].block);
#endif
  pthread_mutex_unlock (&disk_cache_lock);

  if (offset + vm_page_size > dev_end)
    length = dev_end - offset;

  ext2_debug ("writing disk page %lld[%zu]", offset, length);

  STAT_INC (disk_pageouts);

  if (modified_global_blocks)
    /* Be picky about which blocks in a page that we write.  */
    {
      struct pending_blocks pb;

      pending_blocks_init (&pb, buf);

      while (length > 0 && !err)
	{
	  block_t block = boffs_block (offset);

	  /* We don't clear the block modified bit here because this paging
	     write request may not be the same one that actually set the bit,
	     and our copy of the page may be out of date; we have to leave
	     the bit on in case a paging write request corresponding to the
	     modification comes along later.  The bit is only actually ever
	     cleared if the block is allocated to a file, so this results in
	     excess writes of blocks from modified pages.  Unfortunately I
	     know of no way to get arount this given the current external
	     paging interface.  XXXX */
	  if (test_bit (block, modified_global_blocks))
	    /* This block may have been modified, so write it out.  */
	    err = pending_blocks_add (&pb, block);
	  else
	    /* Otherwise just skip it.  */
	    err = pending_blocks_skip (&pb);

	  offset += block_size;
	  length -= block_size;
	}

      if (!err)
	err = pending_blocks_write (&pb);
    }
  else
    {
      err = store_write (store, offset >> store->log2_block_size,
			 buf, length, &amount);
      if (!err && length != amount)
	err = EIO;
    }

  return err;
}

static void
disk_pager_notify_evict (vm_offset_t page)
{
  unsigned long index = page >> log2_block_size;

  ext2_debug ("(block %lu)", index);

  pthread_mutex_lock (&disk_cache_lock);
  disk_cache_info[index].flags &= ~DC_INCORE;
  if (disk_cache_info[index].ref_count == 0 &&
      !(disk_cache_info[index].flags & DC_DONT_REUSE))
    disk_cache_info_free_push (&disk_cache_info[index]);
  pthread_mutex_unlock (&disk_cache_lock);
}

/* Satisfy a pager read request for either the disk pager or file pager
   PAGER, to the page at offset PAGE into BUF.  WRITELOCK should be set if
   the pager should make the page writeable.  */
error_t
pager_read_page (struct user_pager_info *pager, vm_offset_t page,
		 vm_address_t *buf, int *writelock)
{
  if (pager->type == DISK)
    return disk_pager_read_page (page, (void **)buf, writelock);
  else
    return file_pager_read_page (pager->node, page, (void **)buf, writelock);
}

/* Satisfy a pager write request for either the disk pager or file pager
   PAGER, from the page at offset PAGE from BUF.  */
error_t
pager_write_page (struct user_pager_info *pager, vm_offset_t page,
		  vm_address_t buf)
{
  if (pager->type == DISK)
    return disk_pager_write_page (page, (void *)buf);
  else
    return file_pager_write_page (pager->node, page, (void *)buf);
}

void
pager_notify_evict (struct user_pager_info *pager, vm_offset_t page)
{
  if (pager->type == DISK)
    disk_pager_notify_evict (page);
}


/* Make page PAGE writable, at least up to ALLOCSIZE.  This function and
   diskfs_grow are the only places that blocks are actually added to the
   file.  */
error_t
pager_unlock_page (struct user_pager_info *pager, vm_offset_t page)
{
  if (pager->type == DISK)
    return 0;
  else
    {
      error_t err;
      volatile int partial_page;
      struct node *node = pager->node;
      struct disknode *dn = diskfs_node_disknode (node);

      pthread_rwlock_wrlock (&dn->alloc_lock);

      partial_page = (page + vm_page_size > node->allocsize);

      err = diskfs_catch_exception ();
      if (!err)
	{
	  block_t block = page >> log2_block_size;
	  int left = (partial_page ? node->allocsize - page : vm_page_size);

	  while (left > 0)
	    {
	      block_t disk_block;
	      err = ext2_getblk (node, block++, 1, &disk_block);
	      if (err)
		break;
	      left -= block_size;
	    }
	}
      diskfs_end_catch_exception ();

      if (partial_page)
	/* If an error occurred, this page still isn't writable; otherwise,
	   since it's at the end of the file, it's now partially writable.  */
	dn->last_page_partially_writable = !err;
      else if (page + vm_page_size == node->allocsize)
	/* This makes the last page writable, which ends exactly at the end
	   of the file.  If any error occurred, the page still isn't
	   writable, and if not, then the whole thing is writable.  */
	dn->last_page_partially_writable = 0;

#ifdef EXT2FS_DEBUG
      if (dn->last_page_partially_writable)
	ext2_debug ("made page %u[%lu] in inode %d partially writable",
		    page, node->allocsize - page, node->cache_id);
      else
	ext2_debug ("made page %u[%u] in inode %d writable",
		    page, vm_page_size, node->cache_id);
#endif

      STAT_INC (file_page_unlocks);

      pthread_rwlock_unlock (&dn->alloc_lock);

      if (err == ENOSPC)
	ext2_warning ("This filesystem is out of space, and will now crash.  Bye!");
      else if (err)
	ext2_warning ("inode=%Ld, page=0x%lx: %s",
		      node->cache_id, page, strerror (err));

      return err;
    }
}

/* Grow the disk allocated to locked node NODE to be at least SIZE bytes, and
   set NODE->allocsize to the actual allocated size.  (If the allocated size
   is already SIZE bytes, do nothing.)  CRED identifies the user responsible
   for the call.  */
error_t
diskfs_grow (struct node *node, off_t size, struct protid *cred)
{
  diskfs_check_readonly ();
  assert (!diskfs_readonly);

  if (size > node->allocsize)
    {
      error_t err = 0;
      off_t old_size;
      volatile off_t new_size;
      volatile block_t end_block;
      block_t new_end_block;
      struct disknode *dn = diskfs_node_disknode (node);

      pthread_rwlock_wrlock (&dn->alloc_lock);

      old_size = node->allocsize;
      new_size = round_block (size);

      /* The first unallocated blocks after the old and new ends of the
	 file, respectively.  */
      end_block = old_size >> log2_block_size;
      new_end_block = new_size >> log2_block_size;

      if (new_end_block > end_block)
	{
	  /* The first block of the first unallocate page after the old end
	     of the file.  If LAST_PAGE_PARTIALLY_WRITABLE is true, any
	     blocks between this and END_BLOCK were unallocated, but are
	     considered `unlocked' -- that is pager_unlock_page has been
	     called on the page they're in.  Since after this grow the pager
	     will expect them to be writable, we'd better allocate them.  */
	  block_t old_page_end_block =
	    round_page (old_size) >> log2_block_size;

	  ext2_debug ("growing inode %d to %lu bytes (from %lu)", node->cache_id,
		      new_size, old_size);

	  if (dn->last_page_partially_writable
	      && old_page_end_block > end_block)
	    {
	      volatile block_t writable_end =
		(old_page_end_block > new_end_block
		 ? new_end_block
		 : old_page_end_block);

	      ext2_debug ("extending writable page %u by %d blocks"
			  "; first new block = %u",
			  trunc_page (old_size),
			  writable_end - end_block,
			  end_block);

	      err = diskfs_catch_exception ();
	      while (!err && end_block < writable_end)
		{
		  block_t disk_block;
		  err = ext2_getblk (node, end_block++, 1, &disk_block);
		}
	      diskfs_end_catch_exception ();

	      if (! err)
		/* Reflect how much we allocated successfully.  */
		new_size = end_block << log2_block_size;
	      else
		/* See if it's still valid to say this.  */
		dn->last_page_partially_writable =
		  (old_page_end_block > end_block);
	    }
	}

      STAT_INC (file_grows);

      ext2_debug ("new size: %ld%s.", new_size,
		  dn->last_page_partially_writable
		  ? " (last page writable)": "");
      if (err)
	ext2_warning ("inode=%Ld, target=%Ld: %s",
		      node->cache_id, new_size, strerror (err));

      node->allocsize = new_size;

      pthread_rwlock_unlock (&dn->alloc_lock);

      return err;
    }
  else
    return 0;
}

/* This syncs a single file (NODE) to disk.  Wait for all I/O to complete
   if WAIT is set.  NODE->lock must be held.  */
void
diskfs_file_update (struct node *node, int wait)
{
  struct pager *pager;

  pthread_spin_lock (&node_to_page_lock);
  pager = diskfs_node_disknode (node)->pager;
  if (pager)
    ports_port_ref (pager);
  pthread_spin_unlock (&node_to_page_lock);

  if (pager)
    {
      pager_sync (pager, wait);
      ports_port_deref (pager);
    }

  pokel_sync (&diskfs_node_disknode (node)->indir_pokel, wait);

  diskfs_node_update (node, wait);
}

/* Invalidate any pager data associated with NODE.  */
void
flush_node_pager (struct node *node)
{
  struct pager *pager;
  struct disknode *dn = diskfs_node_disknode (node);

  pthread_spin_lock (&node_to_page_lock);
  pager = dn->pager;
  if (pager)
    ports_port_ref (pager);
  pthread_spin_unlock (&node_to_page_lock);

  if (pager)
    {
      pager_flush (pager, 1);
      ports_port_deref (pager);
    }
}


/* Return in *OFFSET and *SIZE the minimum valid address the pager will
   accept and the size of the object.  */
inline error_t
pager_report_extent (struct user_pager_info *pager,
		     vm_address_t *offset, vm_size_t *size)
{
  assert (pager->type == DISK || pager->type == FILE_DATA);

  *offset = 0;

  if (pager->type == DISK)
    *size = store->size;
  else
    *size = pager->node->allocsize;

  return 0;
}

/* This is called when a pager is being deallocated after all extant send
   rights have been destroyed.  */
void
pager_clear_user_data (struct user_pager_info *upi)
{
  if (upi->type == FILE_DATA)
    {
      struct pager *pager;

      pthread_spin_lock (&node_to_page_lock);
      pager = diskfs_node_disknode (upi->node)->pager;
      assert (!pager || pager_get_upi (pager) != upi);
      pthread_spin_unlock (&node_to_page_lock);

      diskfs_nrele_light (upi->node);
    }
}

/* This will be called when the ports library wants to drop weak references.
   The pager library creates no weak references itself.  If the user doesn't
   either, then it's OK for this function to do nothing.  */
void
pager_dropweak (struct user_pager_info *upi)
{
  if (upi->type == FILE_DATA)
    {
      struct pager *pager;

      pthread_spin_lock (&node_to_page_lock);
      pager = diskfs_node_disknode (upi->node)->pager;
      if (pager && pager_get_upi (pager) == upi)
	{
	  diskfs_node_disknode (upi->node)->pager = NULL;
	  ports_port_deref_weak (pager);
	}
      pthread_spin_unlock (&node_to_page_lock);
    }
}

/* Cached blocks from disk.  */
void *disk_cache;

/* DISK_CACHE size in bytes and blocks.  */
store_offset_t disk_cache_size;
int disk_cache_blocks;

/* block num --> pointer to in-memory block */
hurd_ihash_t disk_cache_bptr;
/* Cached blocks' info.  */
struct disk_cache_info *disk_cache_info;
/* Lock for these structures.  */
pthread_mutex_t disk_cache_lock;
/* Fired when a re-association is done.  */
pthread_cond_t disk_cache_reassociation;

/* Linked list of potentially unused blocks. */
static struct disk_cache_info *disk_cache_info_free;
static pthread_mutex_t disk_cache_info_free_lock;

/* Get a reusable entry.  Must be called with disk_cache_lock
   held.  */
static struct disk_cache_info *
disk_cache_info_free_pop (void)
{
  struct disk_cache_info *p;

  do
    {
      pthread_mutex_lock (&disk_cache_info_free_lock);
      p = disk_cache_info_free;
      if (p)
	{
	  disk_cache_info_free = p->next;
	  p->next = NULL;
	}
      pthread_mutex_unlock (&disk_cache_info_free_lock);
    }
  while (p && (p->flags & DC_DONT_REUSE || p->ref_count > 0));
  return p;
}

/* Add P to the list of potentially re-usable entries.  */
static void
disk_cache_info_free_push (struct disk_cache_info *p)
{
  pthread_mutex_lock (&disk_cache_info_free_lock);
  if (! p->next)
    {
      p->next = disk_cache_info_free;
      disk_cache_info_free = p;
    }
  pthread_mutex_unlock (&disk_cache_info_free_lock);
}

/* Finish mapping initialization. */
static void
disk_cache_init (void)
{
  if (block_size != vm_page_size)
    ext2_panic ("Block size %u != vm_page_size %u",
		block_size, vm_page_size);

  pthread_mutex_init (&disk_cache_lock, NULL);
  pthread_cond_init (&disk_cache_reassociation, NULL);
  pthread_mutex_init (&disk_cache_info_free_lock, NULL);

  /* Allocate space for block num -> in-memory pointer mapping.  */
  if (hurd_ihash_create (&disk_cache_bptr, HURD_IHASH_NO_LOCP))
    ext2_panic ("Can't allocate memory for disk_pager_bptr");

  /* Allocate space for disk cache blocks' info.  */
  disk_cache_info = malloc ((sizeof *disk_cache_info) * disk_cache_blocks);
  if (!disk_cache_info)
    ext2_panic ("Cannot allocate space for disk cache info");

  /* Initialize disk_cache_info.  Start with the last entry so that
     the first ends up at the front of the free list.  This keeps the
     assertions at the end of this function happy.  */
  for (int i = disk_cache_blocks - 1; i >= 0; i--)
    {
      disk_cache_info[i].block = DC_NO_BLOCK;
      disk_cache_info[i].flags = 0;
      disk_cache_info[i].ref_count = 0;
      disk_cache_info[i].next = NULL;
      disk_cache_info_free_push (&disk_cache_info[i]);
#ifdef DEBUG_DISK_CACHE
      disk_cache_info[i].last_read = DC_NO_BLOCK;
      disk_cache_info[i].last_read_xor
	= DC_NO_BLOCK ^ DISK_CACHE_LAST_READ_XOR;
#endif
    }

  /* Map the superblock and the block group descriptors.  */
  block_t fixed_first = boffs_block (SBLOCK_OFFS);
  block_t fixed_last = fixed_first
    + (round_block ((sizeof *group_desc_image) * groups_count)
       >> log2_block_size);
  ext2_debug ("%u-%u\n", fixed_first, fixed_last);
  assert (fixed_last - fixed_first + 1 <= (block_t)disk_cache_blocks + 3);
  for (block_t i = fixed_first; i <= fixed_last; i++)
    {
      disk_cache_block_ref (i);
      assert (disk_cache_info[i-fixed_first].block == i);
      disk_cache_info[i-fixed_first].flags |= DC_FIXED;
    }
}

static void
disk_cache_return_unused (void)
{
  int index;

  /* XXX: Touch all pages.  It seems that sometimes GNU Mach "forgets"
     to notify us about evicted pages.  Disk cache must be
     unlocked.  */
  for (vm_offset_t i = 0; i < disk_cache_size; i += vm_page_size)
    *(volatile char *)(disk_cache + i);

  /* Release some references to cached blocks.  */
  pokel_sync (&global_pokel, 1);

  /* Return unused pages that are in core.  */
  int pending_begin = -1, pending_end = -1;
  pthread_mutex_lock (&disk_cache_lock);
  for (index = 0; index < disk_cache_blocks; index++)
    if (! (disk_cache_info[index].flags & (DC_DONT_REUSE & ~DC_INCORE))
	&& ! disk_cache_info[index].ref_count)
      {
	ext2_debug ("return %u -> %d",
		    disk_cache_info[index].block, index);
	if (index != pending_end)
	  {
	    /* Return previous region, if there is such, ... */
	    if (pending_end >= 0)
	      {
		pthread_mutex_unlock (&disk_cache_lock);
		pager_return_some (diskfs_disk_pager,
				   pending_begin * vm_page_size,
				   (pending_end - pending_begin)
				   * vm_page_size, 1);
		pthread_mutex_lock (&disk_cache_lock);
	      }
	    /* ... and start new region.  */
	    pending_begin = index;
	  }
	pending_end = index + 1;
      }

  pthread_mutex_unlock (&disk_cache_lock);

  /* Return last region, if there is such.   */
  if (pending_end >= 0)
    pager_return_some (diskfs_disk_pager,
		       pending_begin * vm_page_size,
		       (pending_end - pending_begin) * vm_page_size,
		       1);
  else
    {
      ext2_debug ("ext2fs: disk cache is starving\n");

      /* Give it some time.  This should happen rarely.  */
      sleep (1);
    }
}

/* Map block and return pointer to it.  */
void *
disk_cache_block_ref (block_t block)
{
  struct disk_cache_info *info;
  int index;
  void *bptr;
  hurd_ihash_locp_t slot;

  assert (block < store->size >> log2_block_size);

  ext2_debug ("(%u)", block);

retry_ref:
  pthread_mutex_lock (&disk_cache_lock);

  bptr = hurd_ihash_locp_find (disk_cache_bptr, block, &slot);
  if (bptr)
    /* Already mapped.  */
    {
      index = bptr_index (bptr);

      /* In process of re-associating?  */
      if (disk_cache_info[index].flags & DC_UNTOUCHED)
	{
	  /* Wait re-association to finish.  */
	  pthread_cond_wait (&disk_cache_reassociation, &disk_cache_lock);
	  pthread_mutex_unlock (&disk_cache_lock);

#if 0
	  printf ("Re-association -- wait finished.\n");
#endif

	  goto retry_ref;
	}

      /* Just increment reference and return.  */
      assert (disk_cache_info[index].ref_count + 1
	      > disk_cache_info[index].ref_count);
      disk_cache_info[index].ref_count++;

      ext2_debug ("cached %u -> %d (ref_count = %hu, flags = %#hx, ptr = %p)",
		  disk_cache_info[index].block, index,
		  disk_cache_info[index].ref_count,
		  disk_cache_info[index].flags, bptr);

      pthread_mutex_unlock (&disk_cache_lock);

      return bptr;
    }

  /* Search for a block that is not in core and is not referenced.  */
  info = disk_cache_info_free_pop ();

  /* Is suitable place found?  */
  if (info == NULL)
    /* No place is found.  Try to release some blocks and try
       again.  */
    {
      ext2_debug ("flush %u -> %d", disk_cache_info[index].block, index);

      pthread_mutex_unlock (&disk_cache_lock);

      disk_cache_return_unused ();

      goto retry_ref;
    }

  /* Suitable place is found.  */
  index = info - disk_cache_info;

  /* Calculate pointer to data.  */
  bptr = (char *)disk_cache + (index << log2_block_size);
  ext2_debug ("map %u -> %d (%p)", block, index, bptr);

  /* This pager_return_some is used only to set PM_FORCEREAD for the
     page.  DC_UNTOUCHED is set so that we catch if someone has
     referenced the block while we didn't hold disk_cache_lock.  */
  disk_cache_info[index].flags |= DC_UNTOUCHED;

#if 0 /* XXX: Let's see if this is needed at all.  */

  pthread_mutex_unlock (&disk_cache_lock);
  pager_return_some (diskfs_disk_pager, bptr - disk_cache, vm_page_size, 1);
  pthread_mutex_lock (&disk_cache_lock);

  /* Has someone used our bptr?  Has someone mapped requested block
     while we have unlocked disk_cache_lock?  If so, environment has
     changed and we have to restart operation.  */
  if ((! (disk_cache_info[index].flags & DC_UNTOUCHED))
      || hurd_ihash_find (disk_cache_bptr, block))
    {
      pthread_mutex_unlock (&disk_cache_lock);
      goto retry_ref;
    }

#elif 0

  /* XXX: Use libpager internals.  */

  pthread_mutex_lock (&diskfs_disk_pager->interlock);
  int page = (bptr - disk_cache) / vm_page_size;
  assert (page >= 0);
  int is_incore = (page < diskfs_disk_pager->pagemapsize
		   && (diskfs_disk_pager->pagemap[page] & PM_INCORE));
  pthread_mutex_unlock (&diskfs_disk_pager->interlock);
  if (is_incore)
    {
      pthread_mutex_unlock (&disk_cache_lock);
      printf ("INCORE\n");
      goto retry_ref;
    }

#endif

  /* Re-associate.  */

  /* New association.  */
  if (hurd_ihash_locp_add (disk_cache_bptr, slot, block, bptr))
    ext2_panic ("Couldn't hurd_ihash_locp_add new disk block");
  if (disk_cache_info[index].block != DC_NO_BLOCK)
    /* Remove old association.  */
    hurd_ihash_remove (disk_cache_bptr, disk_cache_info[index].block);
  assert (! (disk_cache_info[index].flags & DC_DONT_REUSE & ~DC_UNTOUCHED));
  disk_cache_info[index].block = block;
  assert (! disk_cache_info[index].ref_count);
  disk_cache_info[index].ref_count = 1;

  /* All data structures are set up.  */
  pthread_mutex_unlock (&disk_cache_lock);

  /* Try to read page.  */
  *(volatile char *) bptr;

  /* Check if it's actually read.  */
  pthread_mutex_lock (&disk_cache_lock);
  if (disk_cache_info[index].flags & DC_UNTOUCHED)
    /* It's not read.  */
    {
      /* Remove newly created association.  */
      hurd_ihash_remove (disk_cache_bptr, block);
      disk_cache_info[index].block = DC_NO_BLOCK;
      disk_cache_info[index].flags &=~ DC_UNTOUCHED;
      disk_cache_info[index].ref_count = 0;
      pthread_mutex_unlock (&disk_cache_lock);

      /* Prepare next time association of this page to succeed.  */
      pager_flush_some (diskfs_disk_pager, bptr - disk_cache,
			vm_page_size, 0);

#if 0
      printf ("Re-association failed.\n");
#endif

      goto retry_ref;
    }

  /* Re-association was successful.  */
  pthread_cond_broadcast (&disk_cache_reassociation);

  pthread_mutex_unlock (&disk_cache_lock);

  ext2_debug ("(%u) = %p", block, bptr);
  return bptr;
}

void
disk_cache_block_ref_ptr (void *ptr)
{
  int index;

  pthread_mutex_lock (&disk_cache_lock);
  index = bptr_index (ptr);
  assert (disk_cache_info[index].ref_count >= 1);
  assert (disk_cache_info[index].ref_count + 1
	  > disk_cache_info[index].ref_count);
  disk_cache_info[index].ref_count++;
  assert (! (disk_cache_info[index].flags & DC_UNTOUCHED));
  ext2_debug ("(%p) (ref_count = %hu, flags = %#hx)",
	      ptr,
	      disk_cache_info[index].ref_count,
	      disk_cache_info[index].flags);
  pthread_mutex_unlock (&disk_cache_lock);
}

void
disk_cache_block_deref (void *ptr)
{
  int index;

  assert (disk_cache <= ptr && ptr <= disk_cache + disk_cache_size);

  pthread_mutex_lock (&disk_cache_lock);
  index = bptr_index (ptr);
  ext2_debug ("(%p) (ref_count = %hu, flags = %#hx)",
	      ptr,
	      disk_cache_info[index].ref_count - 1,
	      disk_cache_info[index].flags);
  assert (! (disk_cache_info[index].flags & DC_UNTOUCHED));
  assert (disk_cache_info[index].ref_count >= 1);
  disk_cache_info[index].ref_count--;
  if (disk_cache_info[index].ref_count == 0 &&
      !(disk_cache_info[index].flags & DC_DONT_REUSE))
    disk_cache_info_free_push (&disk_cache_info[index]);
  pthread_mutex_unlock (&disk_cache_lock);
}

/* Not used.  */
int
disk_cache_block_is_ref (block_t block)
{
  int ref;
  void *ptr;

  pthread_mutex_lock (&disk_cache_lock);
  ptr = hurd_ihash_find (disk_cache_bptr, block);
  if (ptr == NULL)
    ref = 0;
  else				/* XXX: Should check for DC_UNTOUCHED too.  */
    ref = disk_cache_info[bptr_index (ptr)].ref_count;
  pthread_mutex_unlock (&disk_cache_lock);

  return ref;
}

/* Create the disk pager, and the file pager.  */
void
create_disk_pager (void)
{
  error_t err;

  /* The disk pager.  */
  struct user_pager_info *upi = malloc (sizeof (struct user_pager_info));
  if (!upi)
    ext2_panic ("can't create disk pager: %s", strerror (errno));
  upi->type = DISK;
  disk_pager_bucket = ports_create_bucket ();
  get_hypermetadata ();
  disk_cache_blocks = DISK_CACHE_BLOCKS;
  disk_cache_size = disk_cache_blocks << log2_block_size;
  diskfs_start_disk_pager (upi, disk_pager_bucket, MAY_CACHE, 1,
			   disk_cache_size, &disk_cache);
  disk_cache_init ();

  /* The file pager.  */
  file_pager_bucket = ports_create_bucket ();

  /* Start libpagers worker threads.  */
  err = pager_start_workers (file_pager_bucket, &file_pager_requests);
  if (err)
    ext2_panic ("can't create libpager worker threads: %s", strerror (err));
}

error_t
inhibit_ext2_pager (void)
{
  error_t err;

  /* The file pager can rely on the disk pager, so inhibit the file
     pager first.  */

  err = pager_inhibit_workers (file_pager_requests);
  if (err)
    return err;

  err = pager_inhibit_workers (diskfs_disk_pager_requests);
  /* We don't want only one pager disabled.  */
  if (err)
    pager_resume_workers (file_pager_requests);

  return err;
}

void
resume_ext2_pager (void)
{
  pager_resume_workers (diskfs_disk_pager_requests);
  pager_resume_workers (file_pager_requests);
}

/* Call this to create a FILE_DATA pager and return a send right.
   NODE must be locked.  */
mach_port_t
diskfs_get_filemap (struct node *node, vm_prot_t prot)
{
  mach_port_t right;

  assert (S_ISDIR (node->dn_stat.st_mode)
	  || S_ISREG (node->dn_stat.st_mode)
	  || (S_ISLNK (node->dn_stat.st_mode)));

  pthread_spin_lock (&node_to_page_lock);
  do
    {
      struct pager *pager = diskfs_node_disknode (node)->pager;
      if (pager)
	{
	  right = pager_get_port (pager);
	  assert (MACH_PORT_VALID (right));
	  pager_get_upi (pager)->max_prot |= prot;
	}
      else
	{
	  struct user_pager_info *upi;
	  pager = pager_create_alloc (sizeof *upi, file_pager_bucket,
				      MAY_CACHE, MEMORY_OBJECT_COPY_DELAY, 0);
	  if (pager == NULL)
	    {
	      pthread_spin_unlock (&node_to_page_lock);
	      return MACH_PORT_NULL;
	    }

	  upi = pager_get_upi (pager);
	  upi->type = FILE_DATA;
	  upi->node = node;
	  upi->max_prot = prot;
	  diskfs_nref_light (node);
	  diskfs_node_disknode (node)->pager = pager;

	  /* A weak reference for being part of the node.  */
	  ports_port_ref_weak (diskfs_node_disknode (node)->pager);

	  right = pager_get_port (diskfs_node_disknode (node)->pager);
	  ports_port_deref (diskfs_node_disknode (node)->pager);
	}
    }
  while (right == MACH_PORT_NULL);
  pthread_spin_unlock (&node_to_page_lock);

  mach_port_insert_right (mach_task_self (), right, right,
			  MACH_MSG_TYPE_MAKE_SEND);

  return right;
}

/* Call this when we should turn off caching so that unused memory object
   ports get freed.  */
void
drop_pager_softrefs (struct node *node)
{
  struct pager *pager;

  pthread_spin_lock (&node_to_page_lock);
  pager = diskfs_node_disknode (node)->pager;
  if (pager)
    ports_port_ref (pager);
  pthread_spin_unlock (&node_to_page_lock);

  if (MAY_CACHE && pager)
    {
      pager_sync (pager, 0);
      pager_change_attributes (pager, 0, MEMORY_OBJECT_COPY_DELAY, 0);
    }
  if (pager)
    ports_port_deref (pager);
}

/* Call this when we should turn on caching because it's no longer
   important for unused memory object ports to get freed.  */
void
allow_pager_softrefs (struct node *node)
{
  struct pager *pager;

  pthread_spin_lock (&node_to_page_lock);
  pager = diskfs_node_disknode (node)->pager;
  if (pager)
    ports_port_ref (pager);
  pthread_spin_unlock (&node_to_page_lock);

  if (MAY_CACHE && pager)
    pager_change_attributes (pager, 1, MEMORY_OBJECT_COPY_DELAY, 0);
  if (pager)
    ports_port_deref (pager);
}

/* Call this to find out the struct pager * corresponding to the
   FILE_DATA pager of inode IP.  This should be used *only* as a subsequent
   argument to register_memory_fault_area, and will be deleted when
   the kernel interface is fixed.  NODE must be locked.  */
struct pager *
diskfs_get_filemap_pager_struct (struct node *node)
{
  /* This is safe because pager can't be cleared; there must be
     an active mapping for this to be called. */
  return diskfs_node_disknode (node)->pager;
}

/* Shutdown all the pagers (except the disk pager). */
void
diskfs_shutdown_pager ()
{
  error_t shutdown_one (void *v_p)
    {
      struct pager *p = v_p;
      pager_shutdown (p);
      return 0;
    }

  write_all_disknodes ();

  ports_bucket_iterate (file_pager_bucket, shutdown_one);

  /* Sync everything on the the disk pager.  */
  sync_global (1);

  /* Despite the name of this function, we never actually shutdown the disk
     pager, just make sure it's synced. */
}

/* Sync all the pagers. */
void
diskfs_sync_everything (int wait)
{
  error_t sync_one (void *v_p)
    {
      struct pager *p = v_p;
      pager_sync (p, wait);
      return 0;
    }

  write_all_disknodes ();
  ports_bucket_iterate (file_pager_bucket, sync_one);

  /* Do things on the the disk pager.  */
  sync_global (wait);
}

static void
disable_caching ()
{
  error_t block_cache (void *arg)
    {
      struct pager *p = arg;

      pager_change_attributes (p, 0, MEMORY_OBJECT_COPY_DELAY, 1);
      return 0;
    }

  /* Loop through the pagers and turn off caching one by one,
     synchronously.  That should cause termination of each pager. */
  ports_bucket_iterate (disk_pager_bucket, block_cache);
  ports_bucket_iterate (file_pager_bucket, block_cache);
}

static void
enable_caching ()
{
  error_t enable_cache (void *arg)
    {
      struct pager *p = arg;
      struct user_pager_info *upi = pager_get_upi (p);

      pager_change_attributes (p, 1, MEMORY_OBJECT_COPY_DELAY, 0);

      /* It's possible that we didn't have caching on before, because
	 the user here is the only reference to the underlying node
	 (actually, that's quite likely inside this particular
	 routine), and if that node has no links.  So dinkle the node
	 ref counting scheme here, which will cause caching to be
	 turned off, if that's really necessary.  */
      if (upi->type == FILE_DATA)
	{
	  diskfs_nref (upi->node);
	  diskfs_nrele (upi->node);
	}

      return 0;
    }

  ports_bucket_iterate (disk_pager_bucket, enable_cache);
  ports_bucket_iterate (file_pager_bucket, enable_cache);
}

/* Tell diskfs if there are pagers exported, and if none, then
   prevent any new ones from showing up.  */
int
diskfs_pager_users ()
{
  int npagers = ports_count_bucket (file_pager_bucket);

  if (npagers == 0)
    return 0;

  if (MAY_CACHE)
    {
      disable_caching ();

      /* Give it a second; the kernel doesn't actually shutdown
	 immediately.  XXX */
      sleep (1);

      npagers = ports_count_bucket (file_pager_bucket);
      if (npagers == 0)
	return 0;

      /* Darn, there are actual honest users.  Turn caching back on,
	 and return failure. */
      enable_caching ();
    }

  ports_enable_bucket (file_pager_bucket);

  return 1;
}

/* Return the bitwise or of the maximum prot parameter (the second arg to
   diskfs_get_filemap) for all active user pagers. */
vm_prot_t
diskfs_max_user_pager_prot ()
{
  vm_prot_t max_prot = 0;
  int npagers = ports_count_bucket (file_pager_bucket);

  if (npagers > 0)
    {
      error_t add_pager_max_prot (void *v_p)
	{
	  struct pager *p = v_p;
	  struct user_pager_info *upi = pager_get_upi (p);
	  max_prot |= upi->max_prot;
	  /* Stop iterating if MAX_PROT is as filled as it's going to get. */
	  return max_prot == (VM_PROT_READ|VM_PROT_WRITE|VM_PROT_EXECUTE);
	}

      disable_caching ();		/* Make any silly pagers go away. */

      /* Give it a second; the kernel doesn't actually shutdown
	 immediately.  XXX */
      sleep (1);

      ports_bucket_iterate (file_pager_bucket, add_pager_max_prot);

      enable_caching ();
    }

  ports_enable_bucket (file_pager_bucket);

  return max_prot;
}