summaryrefslogtreecommitdiff
path: root/sys/uvm/uvm_bio.c
blob: f91a9f985f47be92c2d0b00066b27a72811a05de (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
/*	$NetBSD: uvm_bio.c,v 1.99 2018/12/09 20:45:37 jdolecek Exp $	*/

/*
 * Copyright (c) 1998 Chuck Silvers.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. The name of the author may not be used to endorse or promote products
 *    derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 */

/*
 * uvm_bio.c: buffered i/o object mapping cache
 */

#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: uvm_bio.c,v 1.99 2018/12/09 20:45:37 jdolecek Exp $");

#include "opt_uvmhist.h"
#include "opt_ubc.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kmem.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/vnode.h>

#include <uvm/uvm.h>

#ifdef PMAP_DIRECT
#  define UBC_USE_PMAP_DIRECT
#endif

/*
 * local functions
 */

static int	ubc_fault(struct uvm_faultinfo *, vaddr_t, struct vm_page **,
			  int, int, vm_prot_t, int);
static struct ubc_map *ubc_find_mapping(struct uvm_object *, voff_t);
#ifdef UBC_USE_PMAP_DIRECT
static int __noinline ubc_uiomove_direct(struct uvm_object *, struct uio *, vsize_t,
			  int, int);
static void __noinline ubc_zerorange_direct(struct uvm_object *, off_t, size_t, int);

bool ubc_direct = false; /* XXX */
#endif

/*
 * local data structues
 */

#define UBC_HASH(uobj, offset) 						\
	(((((u_long)(uobj)) >> 8) + (((u_long)(offset)) >> PAGE_SHIFT)) & \
				ubc_object.hashmask)

#define UBC_QUEUE(offset)						\
	(&ubc_object.inactive[(((u_long)(offset)) >> ubc_winshift) &	\
			     (UBC_NQUEUES - 1)])

#define UBC_UMAP_ADDR(u)						\
	(vaddr_t)(ubc_object.kva + (((u) - ubc_object.umap) << ubc_winshift))


#define UMAP_PAGES_LOCKED	0x0001
#define UMAP_MAPPING_CACHED	0x0002

struct ubc_map {
	struct uvm_object *	uobj;		/* mapped object */
	voff_t			offset;		/* offset into uobj */
	voff_t			writeoff;	/* write offset */
	vsize_t			writelen;	/* write len */
	int			refcount;	/* refcount on mapping */
	int			flags;		/* extra state */
	int			advice;

	LIST_ENTRY(ubc_map)	hash;		/* hash table */
	TAILQ_ENTRY(ubc_map)	inactive;	/* inactive queue */
	LIST_ENTRY(ubc_map)	list;		/* per-object list */
};

TAILQ_HEAD(ubc_inactive_head, ubc_map);
static struct ubc_object {
	struct uvm_object uobj;		/* glue for uvm_map() */
	char *kva;			/* where ubc_object is mapped */
	struct ubc_map *umap;		/* array of ubc_map's */

	LIST_HEAD(, ubc_map) *hash;	/* hashtable for cached ubc_map's */
	u_long hashmask;		/* mask for hashtable */

	struct ubc_inactive_head *inactive;
					/* inactive queues for ubc_map's */
} ubc_object;

const struct uvm_pagerops ubc_pager = {
	.pgo_fault = ubc_fault,
	/* ... rest are NULL */
};

int ubc_nwins = UBC_NWINS;
int ubc_winshift __read_mostly = UBC_WINSHIFT;
int ubc_winsize __read_mostly;
#if defined(PMAP_PREFER)
int ubc_nqueues;
#define UBC_NQUEUES ubc_nqueues
#else
#define UBC_NQUEUES 1
#endif

#if defined(UBC_STATS)

#define	UBC_EVCNT_DEFINE(name) \
struct evcnt ubc_evcnt_##name = \
EVCNT_INITIALIZER(EVCNT_TYPE_MISC, NULL, "ubc", #name); \
EVCNT_ATTACH_STATIC(ubc_evcnt_##name);
#define	UBC_EVCNT_INCR(name) ubc_evcnt_##name.ev_count++

#else /* defined(UBC_STATS) */

#define	UBC_EVCNT_DEFINE(name)	/* nothing */
#define	UBC_EVCNT_INCR(name)	/* nothing */

#endif /* defined(UBC_STATS) */

UBC_EVCNT_DEFINE(wincachehit)
UBC_EVCNT_DEFINE(wincachemiss)
UBC_EVCNT_DEFINE(faultbusy)

/*
 * ubc_init
 *
 * init pager private data structures.
 */

void
ubc_init(void)
{
	/*
	 * Make sure ubc_winshift is sane.
	 */
	if (ubc_winshift < PAGE_SHIFT)
		ubc_winshift = PAGE_SHIFT;
	ubc_winsize = 1 << ubc_winshift;

	/*
	 * init ubc_object.
	 * alloc and init ubc_map's.
	 * init inactive queues.
	 * alloc and init hashtable.
	 * map in ubc_object.
	 */

	uvm_obj_init(&ubc_object.uobj, &ubc_pager, true, UVM_OBJ_KERN);

	ubc_object.umap = kmem_zalloc(ubc_nwins * sizeof(struct ubc_map),
	    KM_SLEEP);
	if (ubc_object.umap == NULL)
		panic("ubc_init: failed to allocate ubc_map");

	vaddr_t va = (vaddr_t)1L;
#ifdef PMAP_PREFER
	PMAP_PREFER(0, &va, 0, 0);	/* kernel is never topdown */
	ubc_nqueues = va >> ubc_winshift;
	if (ubc_nqueues == 0) {
		ubc_nqueues = 1;
	}
#endif
	ubc_object.inactive = kmem_alloc(UBC_NQUEUES *
	    sizeof(struct ubc_inactive_head), KM_SLEEP);
	for (int i = 0; i < UBC_NQUEUES; i++) {
		TAILQ_INIT(&ubc_object.inactive[i]);
	}
	for (int i = 0; i < ubc_nwins; i++) {
		struct ubc_map *umap;
		umap = &ubc_object.umap[i];
		TAILQ_INSERT_TAIL(&ubc_object.inactive[i & (UBC_NQUEUES - 1)],
				  umap, inactive);
	}

	ubc_object.hash = hashinit(ubc_nwins, HASH_LIST, true,
	    &ubc_object.hashmask);
	for (int i = 0; i <= ubc_object.hashmask; i++) {
		LIST_INIT(&ubc_object.hash[i]);
	}

	if (uvm_map(kernel_map, (vaddr_t *)&ubc_object.kva,
		    ubc_nwins << ubc_winshift, &ubc_object.uobj, 0, (vsize_t)va,
		    UVM_MAPFLAG(UVM_PROT_RW, UVM_PROT_RW, UVM_INH_NONE,
				UVM_ADV_RANDOM, UVM_FLAG_NOMERGE)) != 0) {
		panic("ubc_init: failed to map ubc_object");
	}
}

void
ubchist_init(void)
{

	UVMHIST_INIT(ubchist, 300);
}

/*
 * ubc_fault_page: helper of ubc_fault to handle a single page.
 *
 * => Caller has UVM object locked.
 * => Caller will perform pmap_update().
 */

static inline int
ubc_fault_page(const struct uvm_faultinfo *ufi, const struct ubc_map *umap,
    struct vm_page *pg, vm_prot_t prot, vm_prot_t access_type, vaddr_t va)
{
	struct uvm_object *uobj;
	vm_prot_t mask;
	int error;
	bool rdonly;

	uobj = pg->uobject;
	KASSERT(mutex_owned(uobj->vmobjlock));

	if (pg->flags & PG_WANTED) {
		wakeup(pg);
	}
	KASSERT((pg->flags & PG_FAKE) == 0);
	if (pg->flags & PG_RELEASED) {
		mutex_enter(&uvm_pageqlock);
		uvm_pagefree(pg);
		mutex_exit(&uvm_pageqlock);
		return 0;
	}
	if (pg->loan_count != 0) {

		/*
		 * Avoid unneeded loan break, if possible.
		 */

		if ((access_type & VM_PROT_WRITE) == 0) {
			prot &= ~VM_PROT_WRITE;
		}
		if (prot & VM_PROT_WRITE) {
			struct vm_page *newpg;

			newpg = uvm_loanbreak(pg);
			if (newpg == NULL) {
				uvm_page_unbusy(&pg, 1);
				return ENOMEM;
			}
			pg = newpg;
		}
	}

	/*
	 * Note that a page whose backing store is partially allocated
	 * is marked as PG_RDONLY.
	 */

	KASSERT((pg->flags & PG_RDONLY) == 0 ||
	    (access_type & VM_PROT_WRITE) == 0 ||
	    pg->offset < umap->writeoff ||
	    pg->offset + PAGE_SIZE > umap->writeoff + umap->writelen);

	rdonly = ((access_type & VM_PROT_WRITE) == 0 &&
	    (pg->flags & PG_RDONLY) != 0) ||
	    UVM_OBJ_NEEDS_WRITEFAULT(uobj);
	mask = rdonly ? ~VM_PROT_WRITE : VM_PROT_ALL;

	error = pmap_enter(ufi->orig_map->pmap, va, VM_PAGE_TO_PHYS(pg),
	    prot & mask, PMAP_CANFAIL | (access_type & mask));

	mutex_enter(&uvm_pageqlock);
	uvm_pageactivate(pg);
	mutex_exit(&uvm_pageqlock);
	pg->flags &= ~(PG_BUSY|PG_WANTED);
	UVM_PAGE_OWN(pg, NULL);

	return error;
}

/*
 * ubc_fault: fault routine for ubc mapping
 */

static int
ubc_fault(struct uvm_faultinfo *ufi, vaddr_t ign1, struct vm_page **ign2,
    int ign3, int ign4, vm_prot_t access_type, int flags)
{
	struct uvm_object *uobj;
	struct ubc_map *umap;
	vaddr_t va, eva, ubc_offset, slot_offset;
	struct vm_page *pgs[ubc_winsize >> PAGE_SHIFT];
	int i, error, npages;
	vm_prot_t prot;

	UVMHIST_FUNC("ubc_fault"); UVMHIST_CALLED(ubchist);

	/*
	 * no need to try with PGO_LOCKED...
	 * we don't need to have the map locked since we know that
	 * no one will mess with it until our reference is released.
	 */

	if (flags & PGO_LOCKED) {
		uvmfault_unlockall(ufi, NULL, &ubc_object.uobj);
		flags &= ~PGO_LOCKED;
	}

	va = ufi->orig_rvaddr;
	ubc_offset = va - (vaddr_t)ubc_object.kva;
	umap = &ubc_object.umap[ubc_offset >> ubc_winshift];
	KASSERT(umap->refcount != 0);
	KASSERT((umap->flags & UMAP_PAGES_LOCKED) == 0);
	slot_offset = ubc_offset & (ubc_winsize - 1);

	/*
	 * some platforms cannot write to individual bytes atomically, so
	 * software has to do read/modify/write of larger quantities instead.
	 * this means that the access_type for "write" operations
	 * can be VM_PROT_READ, which confuses us mightily.
	 *
	 * deal with this by resetting access_type based on the info
	 * that ubc_alloc() stores for us.
	 */

	access_type = umap->writelen ? VM_PROT_WRITE : VM_PROT_READ;
	UVMHIST_LOG(ubchist, "va 0x%jx ubc_offset 0x%jx access_type %jd",
	    va, ubc_offset, access_type, 0);

	if ((access_type & VM_PROT_WRITE) != 0) {
#ifndef PRIxOFF		/* XXX */
#define PRIxOFF "jx"	/* XXX */
#endif			/* XXX */
		KASSERTMSG((trunc_page(umap->writeoff) <= slot_offset),
		    "out of range write: slot=%#"PRIxVSIZE" off=%#"PRIxOFF,
		    slot_offset, (intmax_t)umap->writeoff);
		KASSERTMSG((slot_offset < umap->writeoff + umap->writelen),
		    "out of range write: slot=%#"PRIxVADDR
		        " off=%#"PRIxOFF" len=%#"PRIxVSIZE,
		    slot_offset, (intmax_t)umap->writeoff, umap->writelen);
	}

	/* no umap locking needed since we have a ref on the umap */
	uobj = umap->uobj;

	if ((access_type & VM_PROT_WRITE) == 0) {
		npages = (ubc_winsize - slot_offset) >> PAGE_SHIFT;
	} else {
		npages = (round_page(umap->offset + umap->writeoff +
		    umap->writelen) - (umap->offset + slot_offset))
		    >> PAGE_SHIFT;
		flags |= PGO_PASTEOF;
	}

again:
	memset(pgs, 0, sizeof (pgs));
	mutex_enter(uobj->vmobjlock);

	UVMHIST_LOG(ubchist, "slot_offset 0x%jx writeoff 0x%jx writelen 0x%jx ",
	    slot_offset, umap->writeoff, umap->writelen, 0);
	UVMHIST_LOG(ubchist, "getpages uobj %#jx offset 0x%jx npages %jd",
	    (uintptr_t)uobj, umap->offset + slot_offset, npages, 0);

	error = (*uobj->pgops->pgo_get)(uobj, umap->offset + slot_offset, pgs,
	    &npages, 0, access_type, umap->advice, flags | PGO_NOBLOCKALLOC |
	    PGO_NOTIMESTAMP);
	UVMHIST_LOG(ubchist, "getpages error %jd npages %jd", error, npages, 0,
	    0);

	if (error == EAGAIN) {
		kpause("ubc_fault", false, hz >> 2, NULL);
		goto again;
	}
	if (error) {
		return error;
	}

	/*
	 * For virtually-indexed, virtually-tagged caches we should avoid
	 * creating writable mappings when we do not absolutely need them,
	 * since the "compatible alias" trick does not work on such caches.
	 * Otherwise, we can always map the pages writable.
	 */

#ifdef PMAP_CACHE_VIVT
	prot = VM_PROT_READ | access_type;
#else
	prot = VM_PROT_READ | VM_PROT_WRITE;
#endif

	va = ufi->orig_rvaddr;
	eva = ufi->orig_rvaddr + (npages << PAGE_SHIFT);

	UVMHIST_LOG(ubchist, "va 0x%jx eva 0x%jx", va, eva, 0, 0);

	/*
	 * Note: normally all returned pages would have the same UVM object.
	 * However, layered file-systems and e.g. tmpfs, may return pages
	 * which belong to underlying UVM object.  In such case, lock is
	 * shared amongst the objects.
	 */
	mutex_enter(uobj->vmobjlock);
	for (i = 0; va < eva; i++, va += PAGE_SIZE) {
		struct vm_page *pg;

		UVMHIST_LOG(ubchist, "pgs[%jd] = %#jx", i, (uintptr_t)pgs[i],
		    0, 0);
		pg = pgs[i];

		if (pg == NULL || pg == PGO_DONTCARE) {
			continue;
		}
		KASSERT(uobj->vmobjlock == pg->uobject->vmobjlock);
		error = ubc_fault_page(ufi, umap, pg, prot, access_type, va);
		if (error) {
			/*
			 * Flush (there might be pages entered), drop the lock,
			 * and perform uvm_wait().  Note: page will re-fault.
			 */
			pmap_update(ufi->orig_map->pmap);
			mutex_exit(uobj->vmobjlock);
			uvm_wait("ubc_fault");
			mutex_enter(uobj->vmobjlock);
		}
	}
	/* Must make VA visible before the unlock. */
	pmap_update(ufi->orig_map->pmap);
	mutex_exit(uobj->vmobjlock);

	return 0;
}

/*
 * local functions
 */

static struct ubc_map *
ubc_find_mapping(struct uvm_object *uobj, voff_t offset)
{
	struct ubc_map *umap;

	LIST_FOREACH(umap, &ubc_object.hash[UBC_HASH(uobj, offset)], hash) {
		if (umap->uobj == uobj && umap->offset == offset) {
			return umap;
		}
	}
	return NULL;
}


/*
 * ubc interface functions
 */

/*
 * ubc_alloc:  allocate a file mapping window
 */

static void * __noinline
ubc_alloc(struct uvm_object *uobj, voff_t offset, vsize_t *lenp, int advice,
    int flags)
{
	vaddr_t slot_offset, va;
	struct ubc_map *umap;
	voff_t umap_offset;
	int error;
	UVMHIST_FUNC("ubc_alloc"); UVMHIST_CALLED(ubchist);

	UVMHIST_LOG(ubchist, "uobj %#jx offset 0x%jx len 0x%jx",
	    (uintptr_t)uobj, offset, *lenp, 0);

	KASSERT(*lenp > 0);
	umap_offset = (offset & ~((voff_t)ubc_winsize - 1));
	slot_offset = (vaddr_t)(offset & ((voff_t)ubc_winsize - 1));
	*lenp = MIN(*lenp, ubc_winsize - slot_offset);

	mutex_enter(ubc_object.uobj.vmobjlock);
again:
	/*
	 * The UVM object is already referenced.
	 * Lock order: UBC object -> ubc_map::uobj.
	 */
	umap = ubc_find_mapping(uobj, umap_offset);
	if (umap == NULL) {
		struct uvm_object *oobj;

		UBC_EVCNT_INCR(wincachemiss);
		umap = TAILQ_FIRST(UBC_QUEUE(offset));
		if (umap == NULL) {
			kpause("ubc_alloc", false, hz >> 2,
			    ubc_object.uobj.vmobjlock);
			goto again;
		}

		va = UBC_UMAP_ADDR(umap);
		oobj = umap->uobj;

		/*
		 * Remove from old hash (if any), add to new hash.
		 */

		if (oobj != NULL) {
			/*
			 * Mapping must be removed before the list entry,
			 * since there is a race with ubc_purge().
			 */
			if (umap->flags & UMAP_MAPPING_CACHED) {
				umap->flags &= ~UMAP_MAPPING_CACHED;
				mutex_enter(oobj->vmobjlock);
				pmap_remove(pmap_kernel(), va,
				    va + ubc_winsize);
				pmap_update(pmap_kernel());
				mutex_exit(oobj->vmobjlock);
			}
			LIST_REMOVE(umap, hash);
			LIST_REMOVE(umap, list);
		} else {
			KASSERT((umap->flags & UMAP_MAPPING_CACHED) == 0);
		}
		umap->uobj = uobj;
		umap->offset = umap_offset;
		LIST_INSERT_HEAD(&ubc_object.hash[UBC_HASH(uobj, umap_offset)],
		    umap, hash);
		LIST_INSERT_HEAD(&uobj->uo_ubc, umap, list);
	} else {
		UBC_EVCNT_INCR(wincachehit);
		va = UBC_UMAP_ADDR(umap);
	}

	if (umap->refcount == 0) {
		TAILQ_REMOVE(UBC_QUEUE(offset), umap, inactive);
	}

	if (flags & UBC_WRITE) {
		KASSERTMSG(umap->writeoff == 0 && umap->writelen == 0,
		    "ubc_alloc: concurrent writes to uobj %p", uobj);
		umap->writeoff = slot_offset;
		umap->writelen = *lenp;
	}

	umap->refcount++;
	umap->advice = advice;
	mutex_exit(ubc_object.uobj.vmobjlock);
	UVMHIST_LOG(ubchist, "umap %#jx refs %jd va %#jx flags 0x%jx",
	    (uintptr_t)umap, umap->refcount, (uintptr_t)va, flags);

	if (flags & UBC_FAULTBUSY) {
		// XXX add offset from slot_offset?
		int npages = (*lenp + PAGE_SIZE - 1) >> PAGE_SHIFT;
		struct vm_page *pgs[npages];
		int gpflags =
		    PGO_SYNCIO|PGO_OVERWRITE|PGO_PASTEOF|PGO_NOBLOCKALLOC|
		    PGO_NOTIMESTAMP;
		int i;
		KDASSERT(flags & UBC_WRITE);
		KASSERT(umap->refcount == 1);

		UBC_EVCNT_INCR(faultbusy);
again_faultbusy:
		mutex_enter(uobj->vmobjlock);
		if (umap->flags & UMAP_MAPPING_CACHED) {
			umap->flags &= ~UMAP_MAPPING_CACHED;
			pmap_remove(pmap_kernel(), va, va + ubc_winsize);
		}
		memset(pgs, 0, sizeof(pgs));

		error = (*uobj->pgops->pgo_get)(uobj, trunc_page(offset), pgs,
		    &npages, 0, VM_PROT_READ | VM_PROT_WRITE, advice, gpflags);
		UVMHIST_LOG(ubchist, "faultbusy getpages %jd", error, 0, 0, 0);
		if (error) {
			/*
			 * Flush: the mapping above might have been removed.
			 */
			pmap_update(pmap_kernel());
			goto out;
		}
		for (i = 0; i < npages; i++) {
			struct vm_page *pg = pgs[i];

			KASSERT(pg->uobject == uobj);
			if (pg->loan_count != 0) {
				mutex_enter(uobj->vmobjlock);
				if (pg->loan_count != 0) {
					pg = uvm_loanbreak(pg);
				}
				if (pg == NULL) {
					pmap_kremove(va, ubc_winsize);
					pmap_update(pmap_kernel());
					uvm_page_unbusy(pgs, npages);
					mutex_exit(uobj->vmobjlock);
					uvm_wait("ubc_alloc");
					goto again_faultbusy;
				}
				mutex_exit(uobj->vmobjlock);
				pgs[i] = pg;
			}
			pmap_kenter_pa(va + slot_offset + (i << PAGE_SHIFT),
			    VM_PAGE_TO_PHYS(pg),
			    VM_PROT_READ | VM_PROT_WRITE, 0);
		}
		pmap_update(pmap_kernel());
		umap->flags |= UMAP_PAGES_LOCKED;
	} else {
		KASSERT((umap->flags & UMAP_PAGES_LOCKED) == 0);
	}

out:
	return (void *)(va + slot_offset);
}

/*
 * ubc_release:  free a file mapping window.
 */

static void __noinline
ubc_release(void *va, int flags)
{
	struct ubc_map *umap;
	struct uvm_object *uobj;
	vaddr_t umapva;
	bool unmapped;
	UVMHIST_FUNC("ubc_release"); UVMHIST_CALLED(ubchist);

	UVMHIST_LOG(ubchist, "va %#jx", (uintptr_t)va, 0, 0, 0);
	umap = &ubc_object.umap[((char *)va - ubc_object.kva) >> ubc_winshift];
	umapva = UBC_UMAP_ADDR(umap);
	uobj = umap->uobj;
	KASSERT(uobj != NULL);

	if (umap->flags & UMAP_PAGES_LOCKED) {
		const voff_t slot_offset = umap->writeoff;
		const voff_t endoff = umap->writeoff + umap->writelen;
		const voff_t zerolen = round_page(endoff) - endoff;
		const u_int npages = (round_page(endoff) -
		    trunc_page(slot_offset)) >> PAGE_SHIFT;
		struct vm_page *pgs[npages];

		KASSERT((umap->flags & UMAP_MAPPING_CACHED) == 0);
		if (zerolen) {
			memset((char *)umapva + endoff, 0, zerolen);
		}
		umap->flags &= ~UMAP_PAGES_LOCKED;
		mutex_enter(uobj->vmobjlock);
		mutex_enter(&uvm_pageqlock);
		for (u_int i = 0; i < npages; i++) {
			paddr_t pa;
			bool rv __diagused;

			rv = pmap_extract(pmap_kernel(),
			    umapva + slot_offset + (i << PAGE_SHIFT), &pa);
			KASSERT(rv);
			pgs[i] = PHYS_TO_VM_PAGE(pa);
			pgs[i]->flags &= ~(PG_FAKE|PG_CLEAN);
			KASSERT(pgs[i]->loan_count == 0);
			uvm_pageactivate(pgs[i]);
		}
		mutex_exit(&uvm_pageqlock);
		pmap_kremove(umapva, ubc_winsize);
		pmap_update(pmap_kernel());
		uvm_page_unbusy(pgs, npages);
		mutex_exit(uobj->vmobjlock);
		unmapped = true;
	} else {
		unmapped = false;
	}

	mutex_enter(ubc_object.uobj.vmobjlock);
	umap->writeoff = 0;
	umap->writelen = 0;
	umap->refcount--;
	if (umap->refcount == 0) {
		if (flags & UBC_UNMAP) {
			/*
			 * Invalidate any cached mappings if requested.
			 * This is typically used to avoid leaving
			 * incompatible cache aliases around indefinitely.
			 */
			mutex_enter(uobj->vmobjlock);
			pmap_remove(pmap_kernel(), umapva,
				    umapva + ubc_winsize);
			pmap_update(pmap_kernel());
			mutex_exit(uobj->vmobjlock);

			umap->flags &= ~UMAP_MAPPING_CACHED;
			LIST_REMOVE(umap, hash);
			LIST_REMOVE(umap, list);
			umap->uobj = NULL;
			TAILQ_INSERT_HEAD(UBC_QUEUE(umap->offset), umap,
			    inactive);
		} else {
			if (!unmapped) {
				umap->flags |= UMAP_MAPPING_CACHED;
			}
			TAILQ_INSERT_TAIL(UBC_QUEUE(umap->offset), umap,
			    inactive);
		}
	}
	UVMHIST_LOG(ubchist, "umap %cw#jxp refs %jd", (uintptr_t)umap,
	    umap->refcount, 0, 0);
	mutex_exit(ubc_object.uobj.vmobjlock);
}

/*
 * ubc_uiomove: move data to/from an object.
 */

int
ubc_uiomove(struct uvm_object *uobj, struct uio *uio, vsize_t todo, int advice,
    int flags)
{
	const bool overwrite = (flags & UBC_FAULTBUSY) != 0;
	voff_t off;
	int error;

	KASSERT(todo <= uio->uio_resid);
	KASSERT(((flags & UBC_WRITE) != 0 && uio->uio_rw == UIO_WRITE) ||
	    ((flags & UBC_READ) != 0 && uio->uio_rw == UIO_READ));

#ifdef UBC_USE_PMAP_DIRECT
	if (ubc_direct) {
		return ubc_uiomove_direct(uobj, uio, todo, advice, flags);
	}
#endif

	off = uio->uio_offset;
	error = 0;
	while (todo > 0) {
		vsize_t bytelen = todo;
		void *win;

		win = ubc_alloc(uobj, off, &bytelen, advice, flags);
		if (error == 0) {
			error = uiomove(win, bytelen, uio);
		}
		if (error != 0 && overwrite) {
			/*
			 * if we haven't initialized the pages yet,
			 * do it now.  it's safe to use memset here
			 * because we just mapped the pages above.
			 */
			printf("%s: error=%d\n", __func__, error);
			memset(win, 0, bytelen);
		}
		ubc_release(win, flags);
		off += bytelen;
		todo -= bytelen;
		if (error != 0 && (flags & UBC_PARTIALOK) != 0) {
			break;
		}
	}

	return error;
}

/*
 * ubc_zerorange: set a range of bytes in an object to zero.
 */

void
ubc_zerorange(struct uvm_object *uobj, off_t off, size_t len, int flags)
{

#ifdef UBC_USE_PMAP_DIRECT
	if (ubc_direct) {
		ubc_zerorange_direct(uobj, off, len, flags);
		return;
	}
#endif

	/*
	 * XXXUBC invent kzero() and use it
	 */

	while (len) {
		void *win;
		vsize_t bytelen = len;

		win = ubc_alloc(uobj, off, &bytelen, UVM_ADV_NORMAL, UBC_WRITE);
		memset(win, 0, bytelen);
		ubc_release(win, flags);

		off += bytelen;
		len -= bytelen;
	}
}

#ifdef UBC_USE_PMAP_DIRECT
/* Copy data using direct map */

/*
 * ubc_alloc_direct:  allocate a file mapping window using direct map
 */
static int __noinline
ubc_alloc_direct(struct uvm_object *uobj, voff_t offset, vsize_t *lenp,
    int advice, int flags, struct vm_page **pgs, int *npages)
{
	voff_t pgoff;
	int error;
	int gpflags = flags | PGO_NOTIMESTAMP | PGO_SYNCIO | PGO_ALLPAGES;
	int access_type = VM_PROT_READ;
	UVMHIST_FUNC("ubc_alloc_direct"); UVMHIST_CALLED(ubchist);

	if (flags & UBC_WRITE) {
		if (flags & UBC_FAULTBUSY)
			gpflags |= PGO_OVERWRITE;
#if 0
		KASSERT(!UVM_OBJ_NEEDS_WRITEFAULT(uobj));
#endif

		/*
		 * Tell genfs_getpages() we already have the journal lock,
		 * allow allocation past current EOF.
		 */
		gpflags |= PGO_JOURNALLOCKED | PGO_PASTEOF;
		access_type |= VM_PROT_WRITE;
	} else {
		/* Don't need the empty blocks allocated, PG_RDONLY is okay */
		gpflags |= PGO_NOBLOCKALLOC;
	}

	pgoff = (offset & PAGE_MASK);
	*lenp = MIN(*lenp, ubc_winsize - pgoff);

again:
	*npages = (*lenp + pgoff + PAGE_SIZE - 1) >> PAGE_SHIFT;
	KASSERT((*npages * PAGE_SIZE) <= ubc_winsize);
	KASSERT(*lenp + pgoff <= ubc_winsize);
	memset(pgs, 0, *npages * sizeof(pgs[0]));

	mutex_enter(uobj->vmobjlock);
	error = (*uobj->pgops->pgo_get)(uobj, trunc_page(offset), pgs,
	    npages, 0, access_type, advice, gpflags);
	UVMHIST_LOG(ubchist, "alloc_direct getpages %jd", error, 0, 0, 0);
	if (error) {
		if (error == EAGAIN) {
			kpause("ubc_alloc_directg", false, hz >> 2, NULL);
			goto again;
		}
		return error;
	}

	mutex_enter(uobj->vmobjlock);
	for (int i = 0; i < *npages; i++) {
		struct vm_page *pg = pgs[i];

		KASSERT(pg != NULL);
		KASSERT(pg != PGO_DONTCARE);
		KASSERT((pg->flags & PG_FAKE) == 0 || (gpflags & PGO_OVERWRITE));
		KASSERT(pg->uobject->vmobjlock == uobj->vmobjlock);

		/* Avoid breaking loan if possible, only do it on write */
		if ((flags & UBC_WRITE) && pg->loan_count != 0) {
			pg = uvm_loanbreak(pg);
			if (pg == NULL) {
				uvm_page_unbusy(pgs, *npages);
				mutex_exit(uobj->vmobjlock);
				uvm_wait("ubc_alloc_directl");
				goto again;
			}
			pgs[i] = pg;
		}

		/* Page must be writable by now */
		KASSERT((pg->flags & PG_RDONLY) == 0 || (flags & UBC_WRITE) == 0);
	}
	mutex_exit(uobj->vmobjlock);

	return 0;
}

static void __noinline
ubc_direct_release(struct uvm_object *uobj,
	int flags, struct vm_page **pgs, int npages)
{
	mutex_enter(uobj->vmobjlock);
	mutex_enter(&uvm_pageqlock);
	for (int i = 0; i < npages; i++) {
		struct vm_page *pg = pgs[i];

		uvm_pageactivate(pg);

		/* Page was changed, no longer fake and neither clean */
		if (flags & UBC_WRITE)
			pg->flags &= ~(PG_FAKE|PG_CLEAN);
	}
	mutex_exit(&uvm_pageqlock);

	uvm_page_unbusy(pgs, npages);
	mutex_exit(uobj->vmobjlock);
}

static int
ubc_uiomove_process(void *win, size_t len, void *arg)
{
	struct uio *uio = (struct uio *)arg;

	return uiomove(win, len, uio);
}

static int
ubc_zerorange_process(void *win, size_t len, void *arg)
{
	memset(win, 0, len);
	return 0;
}

static int __noinline
ubc_uiomove_direct(struct uvm_object *uobj, struct uio *uio, vsize_t todo, int advice,
    int flags)
{
	const bool overwrite = (flags & UBC_FAULTBUSY) != 0;
	voff_t off;
	int error, npages;
	struct vm_page *pgs[ubc_winsize >> PAGE_SHIFT];

	KASSERT(todo <= uio->uio_resid);
	KASSERT(((flags & UBC_WRITE) != 0 && uio->uio_rw == UIO_WRITE) ||
	    ((flags & UBC_READ) != 0 && uio->uio_rw == UIO_READ));

	off = uio->uio_offset;
	error = 0;
	while (todo > 0) {
		vsize_t bytelen = todo;

		error = ubc_alloc_direct(uobj, off, &bytelen, advice, flags,
		    pgs, &npages);
		if (error != 0) {
			/* can't do anything, failed to get the pages */
			break;
		}

		if (error == 0) {
			error = uvm_direct_process(pgs, npages, off, bytelen,
			    ubc_uiomove_process, uio);
		}
		if (error != 0 && overwrite) {
			/*
			 * if we haven't initialized the pages yet,
			 * do it now.  it's safe to use memset here
			 * because we just mapped the pages above.
			 */
			printf("%s: error=%d\n", __func__, error);
			(void) uvm_direct_process(pgs, npages, off, bytelen,
			    ubc_zerorange_process, NULL);
		}

		ubc_direct_release(uobj, flags, pgs, npages);

		off += bytelen;
		todo -= bytelen;

		if (error != 0 && ISSET(flags, UBC_PARTIALOK)) {
			break;
		}
	}

	return error;
}

static void __noinline
ubc_zerorange_direct(struct uvm_object *uobj, off_t off, size_t todo, int flags)
{
	int error, npages;
	struct vm_page *pgs[ubc_winsize >> PAGE_SHIFT];

	flags |= UBC_WRITE;

	error = 0;
	while (todo > 0) {
		vsize_t bytelen = todo;

		error = ubc_alloc_direct(uobj, off, &bytelen, UVM_ADV_NORMAL,
		    flags, pgs, &npages);
		if (error != 0) {
			/* can't do anything, failed to get the pages */
			break;
		}

		error = uvm_direct_process(pgs, npages, off, bytelen,
		    ubc_zerorange_process, NULL);

		ubc_direct_release(uobj, flags, pgs, npages);

		off += bytelen;
		todo -= bytelen;
	}
}

#endif /* UBC_USE_PMAP_DIRECT */

/*
 * ubc_purge: disassociate ubc_map structures from an empty uvm_object.
 */

void
ubc_purge(struct uvm_object *uobj)
{
	struct ubc_map *umap;
	vaddr_t va;

	KASSERT(uobj->uo_npages == 0);

	/*
	 * Safe to check without lock held, as ubc_alloc() removes
	 * the mapping and list entry in the correct order.
	 */
	if (__predict_true(LIST_EMPTY(&uobj->uo_ubc))) {
		return;
	}
	mutex_enter(ubc_object.uobj.vmobjlock);
	while ((umap = LIST_FIRST(&uobj->uo_ubc)) != NULL) {
		KASSERT(umap->refcount == 0);
		for (va = 0; va < ubc_winsize; va += PAGE_SIZE) {
			KASSERT(!pmap_extract(pmap_kernel(),
			    va + UBC_UMAP_ADDR(umap), NULL));
		}
		LIST_REMOVE(umap, list);
		LIST_REMOVE(umap, hash);
		umap->flags &= ~UMAP_MAPPING_CACHED;
		umap->uobj = NULL;
	}
	mutex_exit(ubc_object.uobj.vmobjlock);
}