[mirror_zfs.git] / module / os / freebsd / zfs / dmu_os.c

/*
 * Copyright (c) 2020 iXsystems, Inc.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 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.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include <sys/dmu.h>
#include <sys/dmu_impl.h>
#include <sys/dmu_tx.h>
#include <sys/dbuf.h>
#include <sys/dnode.h>
#include <sys/zfs_context.h>
#include <sys/dmu_objset.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_synctask.h>
#include <sys/dsl_prop.h>
#include <sys/dmu_zfetch.h>
#include <sys/zfs_ioctl.h>
#include <sys/zap.h>
#include <sys/zio_checksum.h>
#include <sys/zio_compress.h>
#include <sys/sa.h>
#include <sys/zfeature.h>
#include <sys/abd.h>
#include <sys/zfs_rlock.h>
#include <sys/racct.h>
#include <sys/vm.h>
#include <sys/zfs_znode.h>
#include <sys/zfs_vnops.h>


#ifndef IDX_TO_OFF
#define	IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
#endif

#if  __FreeBSD_version < 1300051
#define	VM_ALLOC_BUSY_FLAGS VM_ALLOC_NOBUSY
#else
#define	VM_ALLOC_BUSY_FLAGS  VM_ALLOC_SBUSY | VM_ALLOC_IGN_SBUSY
#endif


#if __FreeBSD_version < 1300072
#define	dmu_page_lock(m)	vm_page_lock(m)
#define	dmu_page_unlock(m)	vm_page_unlock(m)
#else
#define	dmu_page_lock(m)
#define	dmu_page_unlock(m)
#endif

static int
dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
    uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
		return (err);

	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
	    numbufsp, dbpp, DMU_READ_PREFETCH);

	dnode_rele(dn, FTAG);

	return (err);
}

int
dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
    vm_page_t *ma, dmu_tx_t *tx)
{
	dmu_buf_t **dbp;
	struct sf_buf *sf;
	int numbufs, i;
	int err;

	if (size == 0)
		return (0);

	err = dmu_buf_hold_array(os, object, offset, size,
	    FALSE, FTAG, &numbufs, &dbp);
	if (err)
		return (err);

	for (i = 0; i < numbufs; i++) {
		int tocpy, copied, thiscpy;
		int bufoff;
		dmu_buf_t *db = dbp[i];
		caddr_t va;

		ASSERT(size > 0);
		ASSERT3U(db->db_size, >=, PAGESIZE);

		bufoff = offset - db->db_offset;
		tocpy = (int)MIN(db->db_size - bufoff, size);

		ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);

		if (tocpy == db->db_size)
			dmu_buf_will_fill(db, tx);
		else
			dmu_buf_will_dirty(db, tx);

		for (copied = 0; copied < tocpy; copied += PAGESIZE) {
			ASSERT3U(ptoa((*ma)->pindex), ==,
			    db->db_offset + bufoff);
			thiscpy = MIN(PAGESIZE, tocpy - copied);
			va = zfs_map_page(*ma, &sf);
			bcopy(va, (char *)db->db_data + bufoff, thiscpy);
			zfs_unmap_page(sf);
			ma += 1;
			bufoff += PAGESIZE;
		}

		if (tocpy == db->db_size)
			dmu_buf_fill_done(db, tx);

		offset += tocpy;
		size -= tocpy;
	}
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	return (err);
}

int
dmu_read_pages(objset_t *os, uint64_t object, vm_page_t *ma, int count,
    int *rbehind, int *rahead, int last_size)
{
	struct sf_buf *sf;
	vm_object_t vmobj;
	vm_page_t m;
	dmu_buf_t **dbp;
	dmu_buf_t *db;
	caddr_t va;
	int numbufs, i;
	int bufoff, pgoff, tocpy;
	int mi, di;
	int err;

	ASSERT3U(ma[0]->pindex + count - 1, ==, ma[count - 1]->pindex);
	ASSERT(last_size <= PAGE_SIZE);

	err = dmu_buf_hold_array(os, object, IDX_TO_OFF(ma[0]->pindex),
	    IDX_TO_OFF(count - 1) + last_size, TRUE, FTAG, &numbufs, &dbp);
	if (err != 0)
		return (err);

#ifdef DEBUG
	IMPLY(last_size < PAGE_SIZE, *rahead == 0);
	if (dbp[0]->db_offset != 0 || numbufs > 1) {
		for (i = 0; i < numbufs; i++) {
			ASSERT(ISP2(dbp[i]->db_size));
			ASSERT((dbp[i]->db_offset % dbp[i]->db_size) == 0);
			ASSERT3U(dbp[i]->db_size, ==, dbp[0]->db_size);
		}
	}
#endif

	vmobj = ma[0]->object;
	zfs_vmobject_wlock_12(vmobj);

	db = dbp[0];
	for (i = 0; i < *rbehind; i++) {
		m = vm_page_grab_unlocked(vmobj, ma[0]->pindex - 1 - i,
		    VM_ALLOC_NORMAL | VM_ALLOC_NOWAIT | VM_ALLOC_BUSY_FLAGS);
		if (m == NULL)
			break;
		if (!vm_page_none_valid(m)) {
			ASSERT3U(m->valid, ==, VM_PAGE_BITS_ALL);
			vm_page_do_sunbusy(m);
			break;
		}
		ASSERT(m->dirty == 0);
		ASSERT(!pmap_page_is_write_mapped(m));

		ASSERT(db->db_size > PAGE_SIZE);
		bufoff = IDX_TO_OFF(m->pindex) % db->db_size;
		va = zfs_map_page(m, &sf);
		bcopy((char *)db->db_data + bufoff, va, PAGESIZE);
		zfs_unmap_page(sf);
		vm_page_valid(m);
		dmu_page_lock(m);
		if ((m->busy_lock & VPB_BIT_WAITERS) != 0)
			vm_page_activate(m);
		else
			vm_page_deactivate(m);
		dmu_page_unlock(m);
		vm_page_do_sunbusy(m);
	}
	*rbehind = i;

	bufoff = IDX_TO_OFF(ma[0]->pindex) % db->db_size;
	pgoff = 0;
	for (mi = 0, di = 0; mi < count && di < numbufs; ) {
		if (pgoff == 0) {
			m = ma[mi];
			if (m != bogus_page) {
				vm_page_assert_xbusied(m);
				ASSERT(vm_page_none_valid(m));
				ASSERT(m->dirty == 0);
				ASSERT(!pmap_page_is_write_mapped(m));
				va = zfs_map_page(m, &sf);
			}
		}
		if (bufoff == 0)
			db = dbp[di];

		if (m != bogus_page) {
			ASSERT3U(IDX_TO_OFF(m->pindex) + pgoff, ==,
			    db->db_offset + bufoff);
		}

		/*
		 * We do not need to clamp the copy size by the file
		 * size as the last block is zero-filled beyond the
		 * end of file anyway.
		 */
		tocpy = MIN(db->db_size - bufoff, PAGESIZE - pgoff);
		if (m != bogus_page)
			bcopy((char *)db->db_data + bufoff, va + pgoff, tocpy);

		pgoff += tocpy;
		ASSERT(pgoff <= PAGESIZE);
		if (pgoff == PAGESIZE) {
			if (m != bogus_page) {
				zfs_unmap_page(sf);
				vm_page_valid(m);
			}
			ASSERT(mi < count);
			mi++;
			pgoff = 0;
		}

		bufoff += tocpy;
		ASSERT(bufoff <= db->db_size);
		if (bufoff == db->db_size) {
			ASSERT(di < numbufs);
			di++;
			bufoff = 0;
		}
	}

#ifdef DEBUG
	/*
	 * Three possibilities:
	 * - last requested page ends at a buffer boundary and , thus,
	 *   all pages and buffers have been iterated;
	 * - all requested pages are filled, but the last buffer
	 *   has not been exhausted;
	 *   the read-ahead is possible only in this case;
	 * - all buffers have been read, but the last page has not been
	 *   fully filled;
	 *   this is only possible if the file has only a single buffer
	 *   with a size that is not a multiple of the page size.
	 */
	if (mi == count) {
		ASSERT(di >= numbufs - 1);
		IMPLY(*rahead != 0, di == numbufs - 1);
		IMPLY(*rahead != 0, bufoff != 0);
		ASSERT(pgoff == 0);
	}
	if (di == numbufs) {
		ASSERT(mi >= count - 1);
		ASSERT(*rahead == 0);
		IMPLY(pgoff == 0, mi == count);
		if (pgoff != 0) {
			ASSERT(mi == count - 1);
			ASSERT((dbp[0]->db_size & PAGE_MASK) != 0);
		}
	}
#endif
	if (pgoff != 0) {
		ASSERT(m != bogus_page);
		bzero(va + pgoff, PAGESIZE - pgoff);
		zfs_unmap_page(sf);
		vm_page_valid(m);
	}

	for (i = 0; i < *rahead; i++) {
		m = vm_page_grab_unlocked(vmobj, ma[count - 1]->pindex + 1 + i,
		    VM_ALLOC_NORMAL | VM_ALLOC_NOWAIT | VM_ALLOC_BUSY_FLAGS);
		if (m == NULL)
			break;
		if (!vm_page_none_valid(m)) {
			ASSERT3U(m->valid, ==, VM_PAGE_BITS_ALL);
			vm_page_do_sunbusy(m);
			break;
		}
		ASSERT(m->dirty == 0);
		ASSERT(!pmap_page_is_mapped(m));

		ASSERT(db->db_size > PAGE_SIZE);
		bufoff = IDX_TO_OFF(m->pindex) % db->db_size;
		tocpy = MIN(db->db_size - bufoff, PAGESIZE);
		va = zfs_map_page(m, &sf);
		bcopy((char *)db->db_data + bufoff, va, tocpy);
		if (tocpy < PAGESIZE) {
			ASSERT(i == *rahead - 1);
			ASSERT((db->db_size & PAGE_MASK) != 0);
			bzero(va + tocpy, PAGESIZE - tocpy);
		}
		zfs_unmap_page(sf);
		vm_page_valid(m);
		dmu_page_lock(m);
		if ((m->busy_lock & VPB_BIT_WAITERS) != 0)
			vm_page_activate(m);
		else
			vm_page_deactivate(m);
		dmu_page_unlock(m);
		vm_page_do_sunbusy(m);
	}
	*rahead = i;
	zfs_vmobject_wunlock_12(vmobj);

	dmu_buf_rele_array(dbp, numbufs, FTAG);
	return (0);
}
Commit	Line	Data
9f0a21e6 MM	1	/*
	2	* Copyright (c) 2020 iXsystems, Inc.
	3	* All rights reserved.
	4	*
	5	* Redistribution and use in source and binary forms, with or without
	6	* modification, are permitted provided that the following conditions
	7	* are met:
	8	* 1. Redistributions of source code must retain the above copyright
	9	* notice, this list of conditions and the following disclaimer.
	10	* 2. Redistributions in binary form must reproduce the above copyright
	11	* notice, this list of conditions and the following disclaimer in the
	12	* documentation and/or other materials provided with the distribution.
	13	*
	14	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	15	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	16	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	17	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	18	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	19	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	20	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	21	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	22	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	23	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	24	* SUCH DAMAGE.
	25	*
	26	*/
	27
	28	#include <sys/cdefs.h>
	29	__FBSDID("$FreeBSD$");
	30
	31	#include <sys/dmu.h>
	32	#include <sys/dmu_impl.h>
	33	#include <sys/dmu_tx.h>
	34	#include <sys/dbuf.h>
	35	#include <sys/dnode.h>
	36	#include <sys/zfs_context.h>
	37	#include <sys/dmu_objset.h>
	38	#include <sys/dmu_traverse.h>
	39	#include <sys/dsl_dataset.h>
	40	#include <sys/dsl_dir.h>
	41	#include <sys/dsl_pool.h>
	42	#include <sys/dsl_synctask.h>
	43	#include <sys/dsl_prop.h>
	44	#include <sys/dmu_zfetch.h>
	45	#include <sys/zfs_ioctl.h>
	46	#include <sys/zap.h>
	47	#include <sys/zio_checksum.h>
	48	#include <sys/zio_compress.h>
	49	#include <sys/sa.h>
	50	#include <sys/zfeature.h>
	51	#include <sys/abd.h>
	52	#include <sys/zfs_rlock.h>
	53	#include <sys/racct.h>
	54	#include <sys/vm.h>
	55	#include <sys/zfs_znode.h>
	56	#include <sys/zfs_vnops.h>
	57
	58
	59	#ifndef IDX_TO_OFF
	60	#define IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
	61	#endif
	62
	63	#if __FreeBSD_version < 1300051
	64	#define VM_ALLOC_BUSY_FLAGS VM_ALLOC_NOBUSY
65	#else
66	#define VM_ALLOC_BUSY_FLAGS VM_ALLOC_SBUSY \| VM_ALLOC_IGN_SBUSY
67	#endif
68
69
70	#if __FreeBSD_version < 1300072
71	#define dmu_page_lock(m) vm_page_lock(m)
72	#define dmu_page_unlock(m) vm_page_unlock(m)
73	#else
74	#define dmu_page_lock(m)
75	#define dmu_page_unlock(m)
76	#endif
77
78	static int
79	dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
80	uint64_t length, int read, void tag, int numbufsp, dmu_buf_t ***dbpp)
81	{
82	dnode_t *dn;
83	int err;
84
85	err = dnode_hold(os, object, FTAG, &dn);
86	if (err)
87	return (err);
88
89	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
90	numbufsp, dbpp, DMU_READ_PREFETCH);
91
92	dnode_rele(dn, FTAG);
93
94	return (err);
95	}
96
97	int
98	dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
99	vm_page_t ma, dmu_tx_t tx)
100	{
101	dmu_buf_t **dbp;
102	struct sf_buf *sf;
103	int numbufs, i;
104	int err;
105
106	if (size == 0)
107	return (0);
108
109	err = dmu_buf_hold_array(os, object, offset, size,
110	FALSE, FTAG, &numbufs, &dbp);
111	if (err)
112	return (err);
113
114	for (i = 0; i < numbufs; i++) {
115	int tocpy, copied, thiscpy;
116	int bufoff;
117	dmu_buf_t *db = dbp[i];
118	caddr_t va;
119
120	ASSERT(size > 0);
121	ASSERT3U(db->db_size, >=, PAGESIZE);
122
123	bufoff = offset - db->db_offset;
124	tocpy = (int)MIN(db->db_size - bufoff, size);
125
126	ASSERT(i == 0 \|\| i == numbufs-1 \|\| tocpy == db->db_size);
127
128	if (tocpy == db->db_size)
129	dmu_buf_will_fill(db, tx);
130	else
131	dmu_buf_will_dirty(db, tx);
132
133	for (copied = 0; copied < tocpy; copied += PAGESIZE) {
134	ASSERT3U(ptoa((*ma)->pindex), ==,
135	db->db_offset + bufoff);
136	thiscpy = MIN(PAGESIZE, tocpy - copied);
137	va = zfs_map_page(*ma, &sf);
138	bcopy(va, (char *)db->db_data + bufoff, thiscpy);
139	zfs_unmap_page(sf);
140	ma += 1;
141	bufoff += PAGESIZE;
142	}
143
144	if (tocpy == db->db_size)
145	dmu_buf_fill_done(db, tx);
146
147	offset += tocpy;
148	size -= tocpy;
149	}
150	dmu_buf_rele_array(dbp, numbufs, FTAG);
151	return (err);
152	}
153
154	int
155	dmu_read_pages(objset_t os, uint64_t object, vm_page_t ma, int count,
156	int rbehind, int rahead, int last_size)
157	{
158	struct sf_buf *sf;
159	vm_object_t vmobj;
160	vm_page_t m;
161	dmu_buf_t **dbp;
162	dmu_buf_t *db;
163	caddr_t va;
164	int numbufs, i;
165	int bufoff, pgoff, tocpy;
166	int mi, di;
167	int err;
168
169	ASSERT3U(ma[0]->pindex + count - 1, ==, ma[count - 1]->pindex);
170	ASSERT(last_size <= PAGE_SIZE);
171
172	err = dmu_buf_hold_array(os, object, IDX_TO_OFF(ma[0]->pindex),
173	IDX_TO_OFF(count - 1) + last_size, TRUE, FTAG, &numbufs, &dbp);
174	if (err != 0)
175	return (err);
176
177	#ifdef DEBUG
178	IMPLY(last_size < PAGE_SIZE, *rahead == 0);
179	if (dbp[0]->db_offset != 0 \|\| numbufs > 1) {
180	for (i = 0; i < numbufs; i++) {
181	ASSERT(ISP2(dbp[i]->db_size));
182	ASSERT((dbp[i]->db_offset % dbp[i]->db_size) == 0);
183	ASSERT3U(dbp[i]->db_size, ==, dbp[0]->db_size);
184	}
185	}
186	#endif
187
188	vmobj = ma[0]->object;
c614fd6e	189	zfs_vmobject_wlock_12(vmobj);
9f0a21e6 MM	190
	191	db = dbp[0];
	192	for (i = 0; i < *rbehind; i++) {
c614fd6e	193	m = vm_page_grab_unlocked(vmobj, ma[0]->pindex - 1 - i,
9f0a21e6 MM	194	VM_ALLOC_NORMAL \| VM_ALLOC_NOWAIT \| VM_ALLOC_BUSY_FLAGS);
	195	if (m == NULL)
	196	break;
	197	if (!vm_page_none_valid(m)) {
	198	ASSERT3U(m->valid, ==, VM_PAGE_BITS_ALL);
	199	vm_page_do_sunbusy(m);
	200	break;
	201	}
	202	ASSERT(m->dirty == 0);
c614fd6e	203	ASSERT(!pmap_page_is_write_mapped(m));
9f0a21e6 MM	204
	205	ASSERT(db->db_size > PAGE_SIZE);
	206	bufoff = IDX_TO_OFF(m->pindex) % db->db_size;
	207	va = zfs_map_page(m, &sf);
	208	bcopy((char *)db->db_data + bufoff, va, PAGESIZE);
	209	zfs_unmap_page(sf);
	210	vm_page_valid(m);
	211	dmu_page_lock(m);
	212	if ((m->busy_lock & VPB_BIT_WAITERS) != 0)
	213	vm_page_activate(m);
	214	else
	215	vm_page_deactivate(m);
	216	dmu_page_unlock(m);
	217	vm_page_do_sunbusy(m);
	218	}
	219	*rbehind = i;
	220
	221	bufoff = IDX_TO_OFF(ma[0]->pindex) % db->db_size;
	222	pgoff = 0;
	223	for (mi = 0, di = 0; mi < count && di < numbufs; ) {
	224	if (pgoff == 0) {
	225	m = ma[mi];
	226	if (m != bogus_page) {
	227	vm_page_assert_xbusied(m);
	228	ASSERT(vm_page_none_valid(m));
	229	ASSERT(m->dirty == 0);
c614fd6e	230	ASSERT(!pmap_page_is_write_mapped(m));
9f0a21e6 MM	231	va = zfs_map_page(m, &sf);
	232	}
	233	}
	234	if (bufoff == 0)
	235	db = dbp[di];
	236
	237	if (m != bogus_page) {
	238	ASSERT3U(IDX_TO_OFF(m->pindex) + pgoff, ==,
	239	db->db_offset + bufoff);
	240	}
	241
	242	/*
	243	* We do not need to clamp the copy size by the file
	244	* size as the last block is zero-filled beyond the
	245	* end of file anyway.
	246	*/
	247	tocpy = MIN(db->db_size - bufoff, PAGESIZE - pgoff);
	248	if (m != bogus_page)
	249	bcopy((char *)db->db_data + bufoff, va + pgoff, tocpy);
	250
	251	pgoff += tocpy;
	252	ASSERT(pgoff <= PAGESIZE);
	253	if (pgoff == PAGESIZE) {
	254	if (m != bogus_page) {
	255	zfs_unmap_page(sf);
	256	vm_page_valid(m);
	257	}
	258	ASSERT(mi < count);
	259	mi++;
	260	pgoff = 0;
	261	}
	262
	263	bufoff += tocpy;
	264	ASSERT(bufoff <= db->db_size);
	265	if (bufoff == db->db_size) {
	266	ASSERT(di < numbufs);
	267	di++;
	268	bufoff = 0;
	269	}
	270	}
	271
	272	#ifdef DEBUG
	273	/*
	274	* Three possibilities:
	275	* - last requested page ends at a buffer boundary and , thus,
	276	* all pages and buffers have been iterated;
	277	* - all requested pages are filled, but the last buffer
	278	* has not been exhausted;
	279	* the read-ahead is possible only in this case;
	280	* - all buffers have been read, but the last page has not been
	281	* fully filled;
	282	* this is only possible if the file has only a single buffer
	283	* with a size that is not a multiple of the page size.
	284	*/
	285	if (mi == count) {
	286	ASSERT(di >= numbufs - 1);
	287	IMPLY(*rahead != 0, di == numbufs - 1);
	288	IMPLY(*rahead != 0, bufoff != 0);
	289	ASSERT(pgoff == 0);
	290	}
	291	if (di == numbufs) {
	292	ASSERT(mi >= count - 1);
	293	ASSERT(*rahead == 0);
	294	IMPLY(pgoff == 0, mi == count);
295	if (pgoff != 0) {
296	ASSERT(mi == count - 1);
297	ASSERT((dbp[0]->db_size & PAGE_MASK) != 0);
298	}
299	}
300	#endif
301	if (pgoff != 0) {
302	ASSERT(m != bogus_page);
303	bzero(va + pgoff, PAGESIZE - pgoff);
304	zfs_unmap_page(sf);
305	vm_page_valid(m);
306	}
307
308	for (i = 0; i < *rahead; i++) {
c614fd6e	309	m = vm_page_grab_unlocked(vmobj, ma[count - 1]->pindex + 1 + i,
9f0a21e6 MM	310	VM_ALLOC_NORMAL \| VM_ALLOC_NOWAIT \| VM_ALLOC_BUSY_FLAGS);
	311	if (m == NULL)
	312	break;
	313	if (!vm_page_none_valid(m)) {
	314	ASSERT3U(m->valid, ==, VM_PAGE_BITS_ALL);
	315	vm_page_do_sunbusy(m);
	316	break;
	317	}
	318	ASSERT(m->dirty == 0);
	319	ASSERT(!pmap_page_is_mapped(m));
	320
	321	ASSERT(db->db_size > PAGE_SIZE);
	322	bufoff = IDX_TO_OFF(m->pindex) % db->db_size;
	323	tocpy = MIN(db->db_size - bufoff, PAGESIZE);
	324	va = zfs_map_page(m, &sf);
	325	bcopy((char *)db->db_data + bufoff, va, tocpy);
	326	if (tocpy < PAGESIZE) {
	327	ASSERT(i == *rahead - 1);
	328	ASSERT((db->db_size & PAGE_MASK) != 0);
	329	bzero(va + tocpy, PAGESIZE - tocpy);
	330	}
	331	zfs_unmap_page(sf);
	332	vm_page_valid(m);
	333	dmu_page_lock(m);
	334	if ((m->busy_lock & VPB_BIT_WAITERS) != 0)
	335	vm_page_activate(m);
	336	else
	337	vm_page_deactivate(m);
	338	dmu_page_unlock(m);
	339	vm_page_do_sunbusy(m);
	340	}
	341	*rahead = i;
c614fd6e	342	zfs_vmobject_wunlock_12(vmobj);
9f0a21e6 MM	343
	344	dmu_buf_rele_array(dbp, numbufs, FTAG);
	345	return (0);
	346	}