summaryrefslogtreecommitdiff
path: root/fatfs/pager.c
blob: f892c88fe98d034470f2f19680383b1b8530c194 (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
/* pager.c - Pager for fatfs.
   Copyright (C) 1997, 1999, 2002, 2003 Free Software Foundation, Inc.
   Written by Thomas Bushnell, n/BSG and Marcus Brinkmann.

   This file is part of the GNU Hurd.

   The GNU Hurd 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.

   The GNU Hurd 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., 59 Temple Place - Suite 330, Boston, MA  02111, USA. */

#include <string.h>
#include <hurd/store.h>
#include "fatfs.h"

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

/* Mapped image of the FAT.  */
void *fat_image;

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 STAT_INC(field) (void) 0

#define MAX_FREE_PAGE_BUFS 32

static pthread_spinlock_t free_page_bufs_lock = PTHREAD_SPINLOCK_INITIALIZER;
static void *free_page_bufs = 0;
static int num_free_page_bufs = 0;

/* Returns a single page page-aligned buffer.  */
static void *
get_page_buf ()
{
  void *buf;

  pthread_spin_lock (&free_page_bufs_lock);

  buf = free_page_bufs;
  if (buf == 0)
    {
      pthread_spin_unlock (&free_page_bufs_lock);
      buf = mmap (0, vm_page_size, PROT_READ|PROT_WRITE, MAP_ANON, 0, 0);
      if (buf == (void *) -1)
        buf = 0;
    }
  else
    {
      free_page_bufs = *(void **)buf;
      num_free_page_bufs--;
      pthread_spin_unlock (&free_page_bufs_lock);
    }

  return buf;
}

/* Frees a block returned by get_page_buf.  */
static void
free_page_buf (void *buf)
{
  pthread_spin_lock (&free_page_bufs_lock);
  if (num_free_page_bufs < MAX_FREE_PAGE_BUFS)
    {
      *(void **)buf = free_page_bufs;
      free_page_bufs = buf;
      num_free_page_bufs++;
      pthread_spin_unlock (&free_page_bufs_lock);
    }
  else
    {
      pthread_spin_unlock (&free_page_bufs_lock);
      munmap (buf, vm_page_size);
    }
}

/* Find the location on disk of page OFFSET in NODE.  Return the disk
   cluster in CLUSTER. If *LOCK is 0, then it a reader
   lock is acquired on NODE's ALLOC_LOCK before doing anything, and left
   locked after return -- even if an error is returned.  0 on success or an
   error code otherwise is returned.  */
static error_t
find_cluster (struct node *node, vm_offset_t offset,
	      cluster_t *cluster, pthread_rwlock_t **lock)
{
  error_t err;

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

  if (round_cluster (offset) > node->allocsize)
    return EIO;

  err = fat_getcluster (node, offset >> log2_bytes_per_cluster, 0, cluster);

  return err;
}

/* Read one page for the root dir pager at offset PAGE, into BUF.  This
   may need to select several filesystem sectors to satisfy one page.
   Assumes that fat_type is FAT12 or FAT16, and that vm_page_size is a
   power of two multiple of bytes_per_sector (which happens to be true).
*/
static error_t
root_dir_pager_read_page (vm_offset_t page, void **buf, int *writelock)
{
  error_t err;
  daddr_t addr;
  int overrun = 0;
  size_t read = 0;

  *writelock = 0;

  if (page >= diskfs_root_node->allocsize)
    {
      return EIO;
    }
  
  pthread_rwlock_rdlock (&diskfs_root_node->dn->alloc_lock);

  addr = first_root_dir_byte + page;
  if (page + vm_page_size > diskfs_root_node->allocsize)
    overrun = page + vm_page_size - diskfs_root_node->allocsize;

  err = store_read (store, addr >> store->log2_block_size,
		    vm_page_size, (void **) buf, &read);
  if (!err && read != vm_page_size)
    err = EIO;
  
  pthread_rwlock_unlock (&diskfs_root_node->dn->alloc_lock);

  if (overrun)
    bzero ((void *) *buf + vm_page_size - overrun, overrun);

  return err;
}

/* Read one page for the pager backing NODE at offset PAGE, into BUF.  This
   may need to select only a part of a filesystem block to satisfy one page.
   Assumes that bytes_per_cluster is a power of two multiple of vm_page_size.
*/
static error_t
file_pager_read_small_page (struct node *node, vm_offset_t page,
			    void **buf, int *writelock)
{
  error_t err;
  pthread_rwlock_t *lock = NULL;
  cluster_t cluster;
  size_t read = 0;

  *writelock = 0;

  if (page >= node->allocsize)
    {
      return EIO;
    }

  err = find_cluster (node, page, &cluster, &lock);

  if (!err)
    {
      err = store_read (store,
			FAT_FIRST_CLUSTER_BLOCK(cluster)
			+ ((page % bytes_per_cluster)
			  >> store->log2_block_size),
			vm_page_size, (void **) buf, &read);

      if (read != vm_page_size)
	err = EIO;
    }

  if (lock)
    pthread_rwlock_unlock (lock);

  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.
   Assumes that vm_page_size is a power of two multiple of bytes_per_cluster.
*/
static error_t
file_pager_read_huge_page (struct node *node, vm_offset_t page,
			   void **buf, int *writelock)
{
  error_t err;
  int offs = 0;
  pthread_rwlock_t *lock = NULL;
  int left = vm_page_size;
  cluster_t pending_clusters = 0;
  int num_pending_clusters = 0;

  /* Read the NUM_PENDING_CLUSTERS cluster in PENDING_CLUSTERS, into the buffer
     pointed to by BUF (allocating it if necessary) at offset OFFS.  OFFS in
     adjusted by the amount read, and NUM_PENDING_CLUSTERS is zeroed.  Any read
     error is returned.  */
  error_t do_pending_reads ()
    {
      if (num_pending_clusters > 0)
        {
          size_t dev_block = FAT_FIRST_CLUSTER_BLOCK(pending_clusters);
          size_t amount = num_pending_clusters << log2_bytes_per_cluster;
	  /* 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_clusters = 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;

  while (left > 0)
    {
      cluster_t cluster;

      err = find_cluster (node, page, &cluster, &lock);
      if (err)
        break;

      if (cluster != pending_clusters + num_pending_clusters)
        {
          err = do_pending_reads ();
          if (err)
            break;
          pending_clusters = cluster;
        }

      num_pending_clusters++;
      
      page += bytes_per_cluster;
      left -= bytes_per_cluster;
    }

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

  if (lock)
    pthread_rwlock_unlock (lock);

  return err;
}

struct pending_clusters
  {
    /* The cluster number of the first of the clusters.  */
    cluster_t cluster;
    /* How many clusters we have.  */
    loff_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 clusters in PC.  */
static error_t
pending_clusters_write (struct pending_clusters *pc)
{
  if (pc->num > 0)
    {
      error_t err;
      size_t dev_block = FAT_FIRST_CLUSTER_BLOCK(pc->cluster);

      size_t length = pc->num << log2_bytes_per_cluster, amount;

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

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

  return 0;
}

static void
pending_clusters_init (struct pending_clusters *pc, void *buf)
{
  pc->buf = buf;
  pc->cluster = 0;
  pc->num = 0;
  pc->offs = 0;
}

/* Add the disk cluster CLUSTER to the list of destination disk clusters pending in
   PC.  */
static error_t
pending_clusters_add (struct pending_clusters *pc, cluster_t cluster)
{
  if (cluster != pc->cluster + pc->num)
    {
      error_t err = pending_clusters_write (pc);
      if (err)
        return err;
      pc->cluster = cluster;
    }
  pc->num++;
  return 0;
}

/* Write one page for the pager backing NODE, at offset PAGE, into BUF.  This
   may need to write several filesystem blocks to satisfy one page, and tries
   to consolidate the i/o if possible.
   Assumes that vm_page_size is a power of two multiple of bytes_per_cluster. 
*/
static error_t
file_pager_write_huge_page (struct node *node, vm_offset_t offset, void *buf)
{
  error_t err = 0;
  struct pending_clusters pc;
  pthread_rwlock_t *lock = &node->dn->alloc_lock;
  cluster_t cluster;
  int left = vm_page_size;

  pending_clusters_init (&pc, buf);

  /* Holding NODE->dn->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 (&node->dn->alloc_lock);

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

  STAT_INC (file_pageouts);

  while (left > 0)
    {
      err = find_cluster (node, offset, &cluster, &lock);
      if (err)
        break;
      pending_clusters_add (&pc, cluster);
      offset += bytes_per_cluster;
      left -= bytes_per_cluster;
    }

  if (!err)
    pending_clusters_write (&pc);

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

  return err;
}

/* Write one page for the root dir pager, at offset OFFSET, into BUF.  This
   may need to write several filesystem blocks to satisfy one page, and tries
   to consolidate the i/o if possible.
   Assumes that fat_type is FAT12 or FAT16 and that vm_page_size is a
   power of two multiple of bytes_per_sector.
*/
static error_t
root_dir_pager_write_page (vm_offset_t offset, void *buf)
{
  error_t err;
  daddr_t addr;
  size_t length;
  size_t write = 0;

  if (offset >= diskfs_root_node->allocsize)
    return 0;

  /* Holding NODE->dn->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_root_node->dn->alloc_lock);

  addr = first_root_dir_byte + offset;

  if (offset + vm_page_size > diskfs_root_node->allocsize)
    length = diskfs_root_node->allocsize - offset;
  else
    length = vm_page_size;

  err = store_write (store, addr >> store->log2_block_size, (void **) buf,
		     length, &write);
  if (!err && write != length)
    err = EIO;

  pthread_rwlock_unlock (&diskfs_root_node->dn->alloc_lock);

  return err;
}

/* Write one page for the pager backing NODE, at offset OFFSET, into BUF.  This
   may need to write several filesystem blocks to satisfy one page, and tries
   to consolidate the i/o if possible.
   Assumes that bytes_per_cluster is a power of two multiple of vm_page_size.
*/
static error_t
file_pager_write_small_page (struct node *node, vm_offset_t offset, void *buf)
{
  error_t err;
  pthread_rwlock_t *lock = NULL;
  cluster_t cluster;
  size_t write = 0;

  if (offset >= node->allocsize)
    return 0;

  /* Holding NODE->dn->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 (&node->dn->alloc_lock);

  err = find_cluster (node, offset, &cluster, &lock);

  if (!err)
    {
      err = store_write (store, FAT_FIRST_CLUSTER_BLOCK(cluster)
			+ ((offset % bytes_per_cluster)
			   >> store->log2_block_size),
			(void **) buf, vm_page_size, &write);
      if (write != vm_page_size)
	err = EIO;
    }

  if (lock)
    pthread_rwlock_unlock (lock);

  return err;
}

static error_t
fat_pager_read_page (vm_offset_t page, void **buf, int *writelock)
{
  error_t err;
  size_t length = vm_page_size, read = 0;
  vm_size_t fat_end = bytes_per_sector * sectors_per_fat;

  if (page + vm_page_size > fat_end)
    length = fat_end - page;

  page += first_fat_sector * bytes_per_sector;
  err = store_read (store, page >> 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
fat_pager_write_page (vm_offset_t page, void *buf)
{
  error_t err = 0;
  size_t length = vm_page_size, amount;
  vm_size_t fat_end = bytes_per_sector * sectors_per_fat;

  if (page + vm_page_size > fat_end)
    length = fat_end - page;

  page += first_fat_sector * bytes_per_sector;
  err = store_write (store, page >> store->log2_block_size,
		     buf, length, &amount);
  if (!err && length != amount)
    err = EIO;

  return err;
}

/* 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 == FAT)
    return fat_pager_read_page (page, (void **)buf, writelock);
  else
    {
      if (pager->node == diskfs_root_node
	  && (fat_type == FAT12 || fat_type == FAT16))
	return root_dir_pager_read_page (page, (void **)buf, writelock);
      else
	{
	  if (bytes_per_cluster < vm_page_size)
	    return file_pager_read_huge_page (pager->node, page,
					      (void **)buf, writelock);
	  else
	    return file_pager_read_small_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 == FAT)
    return fat_pager_write_page (page, (void *)buf);
  else
    {
      if (pager->node == diskfs_root_node
	  && (fat_type == FAT12 || fat_type == FAT16))
	return root_dir_pager_write_page (page, (void *)buf);
      else
	{
	  if (bytes_per_cluster < vm_page_size)
	    return file_pager_write_huge_page (pager->node, page,
					       (void *)buf);
	  else
	    return file_pager_write_small_page (pager->node, page,
						(void *)buf);
	}
    }
}

/* Make page PAGE writable, at least up to ALLOCSIZE.  */
error_t
pager_unlock_page (struct user_pager_info *pager,
		   vm_offset_t page)
{
  /* All pages are writeable. The disk pages anyway, and the file
     pages because blocks are directly allocated in diskfs_grow.  */
  return 0;
}

/* 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.  Note that this will
   only be called for real files, so there is no need to be careful
   about the root dir node on FAT12/16.  */
error_t
diskfs_grow (struct node *node, loff_t size, struct protid *cred)
{
  diskfs_check_readonly ();
  assert (!diskfs_readonly);
  
  if (size > node->allocsize)
    {
      error_t err = 0;
      loff_t old_size;
      volatile loff_t new_size;
      volatile cluster_t end_cluster;
      cluster_t new_end_cluster;
      struct disknode *dn = node->dn;

      pthread_rwlock_wrlock (&dn->alloc_lock);

      old_size = node->allocsize;
      new_size = ((size + bytes_per_cluster - 1) >> log2_bytes_per_cluster)
						 << log2_bytes_per_cluster;

      /* The first unallocated clusters after the old and new ends of
         the file, respectively.  */
      end_cluster = old_size >> log2_bytes_per_cluster;
      new_end_cluster = new_size >> log2_bytes_per_cluster;

      if (new_end_cluster > end_cluster)
        {
	  err = diskfs_catch_exception ();
	  while (!err && end_cluster < new_end_cluster)
	    {
	      cluster_t disk_cluster;
	      err = fat_getcluster (node, end_cluster++, 1, &disk_cluster);
	    }
	  diskfs_end_catch_exception ();

	  if (err)
	    /* Reflect how much we allocated successfully.  */
	    new_size = (end_cluster - 1) >> log2_bytes_per_cluster;
	}
      
      STAT_INC (file_grows);

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

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

  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 = node->dn;

  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 == FAT || pager->type == FILE_DATA);

  *offset = 0;

  if (pager->type == FAT)
    *size = bytes_per_sector * sectors_per_fat;
  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 = upi->node->dn->pager;
      if (pager && pager_get_upi (pager) == upi)
	upi->node->dn->pager = 0;
      pthread_spin_unlock (&node_to_page_lock);
      
      diskfs_nrele_light (upi->node);
    }
  
  free (upi);
}

/* 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 *p __attribute__ ((unused)))
{
}

/* Create the disk pager.  */
void
create_fat_pager (void)
{
  struct user_pager_info *upi = malloc (sizeof (struct user_pager_info));
  upi->type = FAT;
  pager_bucket = ports_create_bucket ();
  diskfs_start_disk_pager (upi, pager_bucket, MAY_CACHE,
			   bytes_per_sector * sectors_per_fat,
			   &fat_image);
}

/* 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 = node->dn->pager;
      if (pager)
	{
          /* Because PAGER is not a real reference, this might be
             nearly deallocated.  If that's so, then the port right
             will be null.  In that case, clear here and loop.  The
             deallocation will complete separately. */
          right = pager_get_port (pager);
          if (right == MACH_PORT_NULL)
            node->dn->pager = 0;
          else
            pager_get_upi (pager)->max_prot |= prot;
        }
      else
        {
          struct user_pager_info *upi =
            malloc (sizeof (struct user_pager_info));
          upi->type = FILE_DATA;
          upi->node = node;
          upi->max_prot = prot;
          diskfs_nref_light (node);
          node->dn->pager =
            pager_create (upi, pager_bucket, MAY_CACHE,
                          MEMORY_OBJECT_COPY_DELAY);
          if (node->dn->pager == 0)
            {
              diskfs_nrele_light (node);
              free (upi);
              pthread_spin_unlock (&node_to_page_lock);
              return MACH_PORT_NULL;
            }

          right = pager_get_port (node->dn->pager);
          ports_port_deref (node->dn->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 = node->dn->pager;
  if (pager)
    ports_port_ref (pager);
  pthread_spin_unlock (&node_to_page_lock);

  if (MAY_CACHE && pager)
    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 = node->dn->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 node->dn->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;
      if (p != diskfs_disk_pager)
        pager_shutdown (p);
      return 0;
    }

  write_all_disknodes ();

  ports_bucket_iterate (pager_bucket, shutdown_one);

  pager_sync (diskfs_disk_pager, 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;
      if (p != diskfs_disk_pager)
        pager_sync (p, wait);
      return 0;
    }

  write_all_disknodes ();
  ports_bucket_iterate (pager_bucket, sync_one);
  pager_sync (diskfs_disk_pager, 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 (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 (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 (pager_bucket);

  if (npagers <= 1)
    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 (pager_bucket);
      if (npagers <= 1)
	return 0;

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

  if (npagers > 1)
    /* More than just the disk pager.  */
    {
      error_t add_pager_max_prot (void *v_p)
        {
          struct pager *p = v_p;
          struct user_pager_info *upi = pager_get_upi (p);
          if (upi->type == FILE_DATA)
            max_prot |= upi->max_prot;
          /* Stop iterating if MAX_PROT is as filled as it is going to
	     get.  */
          return (max_prot
		  == (VM_PROT_READ|VM_PROT_WRITE|VM_PROT_EXECUTE)) ? 1 : 0;
        }

      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 (pager_bucket, add_pager_max_prot);

      enable_caching ();
    }

  ports_enable_bucket (pager_bucket);

  return max_prot;
}