summaryrefslogtreecommitdiff
path: root/ufs/pager.c
blob: 139e5723cbe308b034451dbb80cc7777780f6bdd (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
/* Pager for ufs
   Copyright (C) 1994, 1995 Free Software Foundation

   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 "ufs.h"
#include <strings.h>
#include <stdio.h>
#include <unistd.h>

spin_lock_t node2pagelock = SPIN_LOCK_INITIALIZER;

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

struct port_bucket *pager_bucket;

/* Find the location on disk of page OFFSET in pager UPI.  Return the
   disk address (in disk block) in *ADDR.  If *NPLOCK is set on
   return, then release that mutex after I/O on the data has
   completed.  Set DISKSIZE to be the amount of valid data on disk.
   (If this is an unallocated block, then set *ADDR to zero.)  */
static error_t
find_address (struct user_pager_info *upi,
	      vm_address_t offset,
	      daddr_t *addr,
	      int *disksize,
	      struct rwlock **nplock)
{
  error_t err;
  
  assert (upi->type == DISK || upi->type == FILE_DATA);

  if (upi->type == DISK)
    {
      *disksize = __vm_page_size;
      *addr = offset / DEV_BSIZE;
      *nplock = 0;
      return 0;
    }
  else 
    {
      struct iblock_spec indirs[NIADDR + 1];
      struct node *np;
  
      np = upi->np;
      
      rwlock_reader_lock (&np->dn->allocptrlock);
      *nplock = &np->dn->allocptrlock;

      if (offset >= np->allocsize)
	{
	  rwlock_reader_unlock (&np->dn->allocptrlock);
	  return EIO;
	}
      
      if (offset + __vm_page_size > np->allocsize)
	*disksize = np->allocsize - offset;
      else
	*disksize = __vm_page_size;
      
      err = fetch_indir_spec (np, lblkno (sblock, offset), indirs);
      if (err)
	rwlock_reader_unlock (&np->dn->allocptrlock);
      else
	{
	  if (indirs[0].bno)
	    *addr = (fsbtodb (sblock, indirs[0].bno)
		     + blkoff (sblock, offset) / DEV_BSIZE);
	  else
	    *addr = 0;
	}
      
      return err;
    }
}


/* Implement the pager_read_page callback from the pager library.  See 
   <hurd/pager.h> for the interface description. */
error_t
pager_read_page (struct user_pager_info *pager,
		 vm_offset_t page,
		 vm_address_t *buf,
		 int *writelock)
{
  error_t err;
  struct rwlock *nplock;
  daddr_t addr;
  int disksize;
  
  err = find_address (pager, page, &addr, &disksize, &nplock);
  if (err)
    return err;
  
  if (addr)
    {
      err = dev_read_sync (addr, (void *)buf, disksize);
      if (!err && disksize != __vm_page_size)
	bzero ((void *)(*buf + disksize), __vm_page_size - disksize);
      *writelock = 0;
    }
  else
    {
#if 0
      printf ("Write-locked pagein Object %#x\tOffset %#x\n", pager, page);
      fflush (stdout);
#endif
      vm_allocate (mach_task_self (), buf, __vm_page_size, 1);
      *writelock = 1;
    }
      
  if (nplock)
    rwlock_reader_unlock (nplock);
  
  return err;
}

/* Implement the pager_write_page callback from the pager library.  See 
   <hurd/pager.h> for the interface description. */
error_t
pager_write_page (struct user_pager_info *pager,
		  vm_offset_t page,
		  vm_address_t buf)
{
  daddr_t addr;
  int disksize;
  struct rwlock *nplock;
  error_t err;
  
  err = find_address (pager, page, &addr, &disksize, &nplock);
  if (err)
    return err;
  
  if (addr)
    err = dev_write_sync (addr, buf, disksize);
  else
    {
      printf ("Attempt to write unallocated disk\n.");
      printf ("Object %p\tOffset %#x\n", pager, page);
      fflush (stdout);
      err = 0;			/* unallocated disk; 
				   error would be pointless */
    }
    
  if (nplock)
    rwlock_reader_unlock (nplock);
  
  return err;
}

/* Implement the pager_unlock_page callback from the pager library.  See 
   <hurd/pager.h> for the interface description. */
error_t
pager_unlock_page (struct user_pager_info *pager,
		   vm_offset_t address)
{
  struct node *np;
  error_t err;
  struct iblock_spec indirs[NIADDR + 1];
  daddr_t bno;
  struct disknode *dn;
  struct dinode *di;

  /* Zero an sblock->fs_bsize piece of disk starting at BNO, 
     synchronously.  We do this on newly allocated indirect
     blocks before setting the pointer to them to ensure that an
     indirect block absolutely never points to garbage. */
  void zero_disk_block (int bno)
    {
      bzero (indir_block (bno), sblock->fs_bsize);
      sync_disk_blocks (bno, sblock->fs_bsize, 1);
    };

  /* Problem--where to get cred values for allocation here? */

#if 0
  printf ("Unlock page request, Object %#x\tOffset %#x...", pager, address);
  fflush (stdout);
#endif

  if (pager->type == DISK)
    return 0;
  
  np = pager->np;
  dn = np->dn;
  di = dino (dn->number);

  rwlock_writer_lock (&dn->allocptrlock);
  
  /* If this is the last block, we don't let it get unlocked. */
  if (address + __vm_page_size
      > blkroundup (sblock, np->allocsize) - sblock->fs_bsize)
    {
      printf ("attempt to unlock at last block denied\n");
      fflush (stdout);
      rwlock_writer_unlock (&dn->allocptrlock);
      return EIO;
    }
    
  err = fetch_indir_spec (np, lblkno (sblock, address), indirs);
  if (err)
    {
      rwlock_writer_unlock (&dn->allocptrlock);
      return EIO;
    }

  err = diskfs_catch_exception ();
  if (err)
    {
      rwlock_writer_unlock (&dn->allocptrlock);
      return EIO;
    }

  /* See if we need a triple indirect block; fail if we do. */
  assert (indirs[0].offset == -1 
	  || indirs[1].offset == -1 
	  || indirs[2].offset == -1);
  
  /* Check to see if this block is allocated. */
  if (indirs[0].bno == 0)
    {
      if (indirs[0].offset == -1)
	{
	  err = ffs_alloc (np, lblkno (sblock, address),
			   ffs_blkpref (np, lblkno (sblock, address),
					lblkno (sblock, address), di->di_db),
			   sblock->fs_bsize, &bno, 0);
	  if (err)
	    goto out;
	  assert (lblkno (sblock, address) < NDADDR);
	  indirs[0].bno = di->di_db[lblkno (sblock, address)] = bno;
	  record_poke (di, sizeof (struct dinode));
	}
      else
	{
	  daddr_t *siblock;
	  
	  /* We need to set siblock to the single indirect block
	     array; see if the single indirect block is allocated. */
	  if (indirs[1].bno == 0)
	    {
	      if (indirs[1].offset == -1)
		{
		  err = ffs_alloc (np, lblkno (sblock, address),
				   ffs_blkpref (np, lblkno (sblock, address),
						INDIR_SINGLE, di->di_ib),
				   sblock->fs_bsize, &bno, 0);
		  if (err)
		    goto out;
		  zero_disk_block (bno);
		  indirs[1].bno = di->di_ib[INDIR_SINGLE] = bno;
		  record_poke (di, sizeof (struct dinode));
		}
	      else
		{
		  daddr_t *diblock;
	      
		  /* We need to set diblock to the double indirect
		     block array; see if the double indirect block is
		     allocated. */
		  if (indirs[2].bno == 0)
		    {
		      /* This assert because triple indirection is
			 not supported. */
		      assert (indirs[2].offset == -1);
		      
		      err = ffs_alloc (np, lblkno (sblock, address),
				       ffs_blkpref (np, lblkno (sblock,
								address),
						    INDIR_DOUBLE, di->di_ib),
				       sblock->fs_bsize, &bno, 0);
		      if (err)
			goto out;
		      zero_disk_block (bno);
		      indirs[2].bno = di->di_ib[INDIR_DOUBLE] = bno;
		      record_poke (di, sizeof (struct dinode));
		    }

		  diblock = indir_block (indirs[2].bno);
		  mark_indir_dirty (np, indirs[2].bno);
		  
		  /* Now we can allocate the single indirect block */
		  
		  err = ffs_alloc (np, lblkno (sblock, address),
				   ffs_blkpref (np, lblkno (sblock, address),
						indirs[1].offset, diblock),
				   sblock->fs_bsize, &bno, 0);
		  if (err)
		    goto out;
		  zero_disk_block (bno);
		  indirs[1].bno = diblock[indirs[1].offset] = bno;
		  record_poke (diblock, sblock->fs_bsize);
		}
	    }
	  
	  siblock = indir_block (indirs[1].bno);
	  mark_indir_dirty (np, indirs[1].bno);

	  /* Now we can allocate the data block. */

	  err = ffs_alloc (np, lblkno (sblock, address),
			   ffs_blkpref (np, lblkno (sblock, address),
					indirs[0].offset, siblock),
			   sblock->fs_bsize, &bno, 0);
	  if (err)
	    goto out;

	  dev_write_sync (fsbtodb (sblock, bno), zeroblock, sblock->fs_bsize);

	  indirs[0].bno = siblock[indirs[0].offset] = bno;
	  record_poke (siblock, sblock->fs_bsize);
	}
    }
  
 out:
  diskfs_end_catch_exception ();
  rwlock_writer_unlock (&dn->allocptrlock);
  return err;
}

/* Implement the pager_report_extent callback from the pager library.  See 
   <hurd/pager.h> for the interface description. */
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 = diskpagersize;
  else
    *size = pager->np->allocsize;
  
  return 0;
}

/* Implement the pager_clear_user_data callback from the pager library.
   See <hurd/pager.h> for the interface description. */
void
pager_clear_user_data (struct user_pager_info *upi)
{
  /* XXX Do the right thing for the disk pager here too. */
  if (upi->type == FILE_DATA)
    {
      spin_lock (&node2pagelock);
      upi->np->dn->fileinfo = 0;
      spin_unlock (&node2pagelock);
      diskfs_nrele_light (upi->np);
    }
  free (upi);
}

void
pager_dropweak (struct user_pager_info *upi __attribute__ ((unused)))
{
}
    


static void
thread_function (any_t foo __attribute__ ((unused)))
{
  for (;;)
    ports_manage_port_operations_multithread (pager_bucket, pager_demuxer,
					      1000 * 60 * 2, 0,
					      1, MACH_PORT_NULL);
}

/* Create a the DISK pager, initializing DISKPAGER, and DISKPAGERPORT */
void
create_disk_pager ()
{
  pager_bucket = ports_create_bucket ();

  cthread_detach (cthread_fork ((cthread_fn_t) thread_function, (any_t) 0));
  
  diskpager = malloc (sizeof (struct user_pager_info));
  diskpager->type = DISK;
  diskpager->np = 0;
  diskpager->p = pager_create (diskpager, pager_bucket,
			       MAY_CACHE, MEMORY_OBJECT_COPY_NONE);
  diskpagerport = pager_get_port (diskpager->p);
  mach_port_insert_right (mach_task_self (), diskpagerport, diskpagerport,
			  MACH_MSG_TYPE_MAKE_SEND);
}  

/* This syncs a single file (NP) to disk.  Wait for all I/O to complete
   if WAIT is set.  NP->lock must be held.  */
void
diskfs_file_update (struct node *np,
		    int wait)
{
  struct dirty_indir *d, *tmp;
  struct user_pager_info *upi;

  spin_lock (&node2pagelock);
  upi = np->dn->fileinfo;
  if (upi)
    ports_port_ref (upi->p);
  spin_unlock (&node2pagelock);
  
  if (upi)
    {
      pager_sync (upi->p, wait);
      ports_port_deref (upi->p);
    }
  
  for (d = np->dn->dirty; d; d = tmp)
    {
      sync_disk_blocks (d->bno, sblock->fs_bsize, wait);
      tmp = d->next;
      free (d);
    }
  np->dn->dirty = 0;

  diskfs_node_update (np, wait);
}

/* Call this to create a FILE_DATA pager and return a send right.
   NP must be locked.  */
mach_port_t
diskfs_get_filemap (struct node *np)
{
  struct user_pager_info *upi;
  mach_port_t right;

  assert (S_ISDIR (np->dn_stat.st_mode)
	  || S_ISREG (np->dn_stat.st_mode)
	  || (S_ISLNK (np->dn_stat.st_mode)
	      && (!direct_symlink_extension 
		  || np->dn_stat.st_size >= sblock->fs_maxsymlinklen)));

  spin_lock (&node2pagelock);
  if (!np->dn->fileinfo)
    {
      upi = malloc (sizeof (struct user_pager_info));
      upi->type = FILE_DATA;
      upi->np = np;
      diskfs_nref_light (np);
      upi->p = pager_create (upi, pager_bucket,
			     MAY_CACHE, MEMORY_OBJECT_COPY_DELAY);
      np->dn->fileinfo = upi;
      right = pager_get_port (np->dn->fileinfo->p);
      ports_port_deref (np->dn->fileinfo->p);
    }
  else
    /* There is a race condition here.  If there are no references
       to NP->dn->fileinfo->p, then the clean routine might be
       blocked trying to get into node2pagelock, and this call is
       invalid.   XXX */
    right = pager_get_port (np->dn->fileinfo->p);

  spin_unlock (&node2pagelock);
  
  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 *np)
{
  struct user_pager_info *upi;
  
  spin_lock (&node2pagelock);
  upi = np->dn->fileinfo;
  if (upi)
    ports_port_ref (upi->p);
  spin_unlock (&node2pagelock);

  if (MAY_CACHE && upi)
    pager_change_attributes (upi->p, 0, MEMORY_OBJECT_COPY_DELAY, 0);
  if (upi)
    ports_port_deref (upi->p);
}

/* 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 *np)
{
  struct user_pager_info *upi;
  
  spin_lock (&node2pagelock);
  upi = np->dn->fileinfo;
  if (upi)
    ports_port_ref (upi->p);
  spin_unlock (&node2pagelock);
  
  if (MAY_CACHE && upi)
    pager_change_attributes (upi->p, 1, MEMORY_OBJECT_COPY_DELAY, 0);
  if (upi)
    ports_port_deref (upi->p);
}

/* Tell diskfs if there are pagers exported, and if none, then
   prevent any new ones from showing up.  */
int
diskfs_pager_users ()
{
  int npagers;
  
  error_t block_cache (void *arg)
    {
      struct pager *p = arg;
      
      pager_change_attributes (p, 0, MEMORY_OBJECT_COPY_DELAY, 1);
      return 0;
    }
  
  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->np);
	  diskfs_nrele (upi->np);
	}

      return 0;
    }

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

  if (MAY_CACHE == 0)
    {
      ports_enable_bucket (pager_bucket);
      return 1;
    }
  
  /* 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);

  /* Give it a second; the kernel doesn't actually shutdown
     immediately.  XXX */
  sleep (1);
  
  npagers = ports_count_bucket (pager_bucket);
  
  if (npagers == 0)
    return 0;
  
  /* Darn, there are actual honest users.  Turn caching back on,
     and return failure. */
  ports_bucket_iterate (pager_bucket, enable_cache);
  return 1;
}


/* 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.  NP must be locked.  */
struct pager *
diskfs_get_filemap_pager_struct (struct node *np)
{
  /* This is safe because fileinfo can't be cleared; there must be
     an active mapping for this to be called. */
  return np->dn->fileinfo->p;
}

/* Shutdown all the pagers. */
void
diskfs_shutdown_pager ()
{
  error_t shutdown_one (void *arg)
    {
      struct pager *p = arg;
      /* Make sure the disk pager is done last. */
      if (p != diskpager->p)
	pager_shutdown (p);
      return 0;
    }

  copy_sblock ();
  write_all_disknodes ();
  ports_bucket_iterate (pager_bucket, shutdown_one);
  pager_shutdown (diskpager->p);
}

/* Sync all the pagers. */
void
diskfs_sync_everything (int wait)
{
  error_t sync_one (void *arg)
    {
      struct pager *p = arg;
      /* Make sure the disk pager is done last. */
      if (p != diskpager->p)
	pager_sync (p, wait);
      return 0;
    }
  
  copy_sblock ();
  write_all_disknodes ();
  ports_bucket_iterate (pager_bucket, sync_one);
  sync_disk (wait);
}