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/types.h>
29 #include <sys/param.h>
30 #include <sys/dmu.h>
31 #include <sys/dmu_impl.h>
32 #include <sys/dmu_tx.h>
33 #include <sys/dbuf.h>
34 #include <sys/dnode.h>
35 #include <sys/zfs_context.h>
36 #include <sys/dmu_objset.h>
37 #include <sys/dmu_traverse.h>
38 #include <sys/dsl_dataset.h>
39 #include <sys/dsl_dir.h>
40 #include <sys/dsl_pool.h>
41 #include <sys/dsl_synctask.h>
42 #include <sys/dsl_prop.h>
43 #include <sys/dmu_zfetch.h>
44 #include <sys/zfs_ioctl.h>
45 #include <sys/zap.h>
46 #include <sys/zio_checksum.h>
47 #include <sys/zio_compress.h>
48 #include <sys/sa.h>
49 #include <sys/zfeature.h>
50 #include <sys/abd.h>
51 #include <sys/zfs_rlock.h>
52 #include <sys/racct.h>
53 #include <sys/vm.h>
54 #include <sys/zfs_znode.h>
55 #include <sys/zfs_vnops.h>
56
57 #include <sys/ccompat.h>
58
59 #ifndef IDX_TO_OFF
60 #define IDX_TO_OFF(idx) (((vm_ooffset_t)(idx)) << PAGE_SHIFT)
61 #endif
62
63 #define VM_ALLOC_BUSY_FLAGS VM_ALLOC_SBUSY | VM_ALLOC_IGN_SBUSY
64
65 int
dmu_write_pages(objset_t * os,uint64_t object,uint64_t offset,uint64_t size,vm_page_t * ma,dmu_tx_t * tx)66 dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
67 vm_page_t *ma, dmu_tx_t *tx)
68 {
69 dmu_buf_t **dbp;
70 struct sf_buf *sf;
71 int numbufs, i;
72 int err;
73
74 if (size == 0)
75 return (0);
76
77 err = dmu_buf_hold_array(os, object, offset, size,
78 FALSE, FTAG, &numbufs, &dbp);
79 if (err)
80 return (err);
81
82 for (i = 0; i < numbufs; i++) {
83 int tocpy, copied, thiscpy;
84 int bufoff;
85 dmu_buf_t *db = dbp[i];
86 caddr_t va;
87
88 ASSERT3U(size, >, 0);
89 ASSERT3U(db->db_size, >=, PAGESIZE);
90
91 bufoff = offset - db->db_offset;
92 tocpy = (int)MIN(db->db_size - bufoff, size);
93
94 ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
95
96 if (tocpy == db->db_size)
97 dmu_buf_will_fill(db, tx, B_FALSE);
98 else
99 dmu_buf_will_dirty(db, tx);
100
101 for (copied = 0; copied < tocpy; copied += PAGESIZE) {
102 ASSERT3U(ptoa((*ma)->pindex), ==,
103 db->db_offset + bufoff);
104 thiscpy = MIN(PAGESIZE, tocpy - copied);
105 va = zfs_map_page(*ma, &sf);
106 ASSERT(db->db_data != NULL);
107 memcpy((char *)db->db_data + bufoff, va, thiscpy);
108 zfs_unmap_page(sf);
109 ma += 1;
110 bufoff += PAGESIZE;
111 }
112
113 if (tocpy == db->db_size)
114 dmu_buf_fill_done(db, tx, B_FALSE);
115
116 offset += tocpy;
117 size -= tocpy;
118 }
119 dmu_buf_rele_array(dbp, numbufs, FTAG);
120 return (err);
121 }
122
123 int
dmu_read_pages(objset_t * os,uint64_t object,vm_page_t * ma,int count,int * rbehind,int * rahead,int last_size)124 dmu_read_pages(objset_t *os, uint64_t object, vm_page_t *ma, int count,
125 int *rbehind, int *rahead, int last_size)
126 {
127 struct sf_buf *sf;
128 vm_object_t vmobj;
129 vm_page_t m;
130 dmu_buf_t **dbp;
131 dmu_buf_t *db;
132 caddr_t va;
133 int numbufs, i;
134 int bufoff, pgoff, tocpy;
135 int mi, di;
136 int err;
137
138 ASSERT3U(ma[0]->pindex + count - 1, ==, ma[count - 1]->pindex);
139 ASSERT3S(last_size, <=, PAGE_SIZE);
140
141 err = dmu_buf_hold_array(os, object, IDX_TO_OFF(ma[0]->pindex),
142 IDX_TO_OFF(count - 1) + last_size, TRUE, FTAG, &numbufs, &dbp);
143 if (err != 0)
144 return (err);
145
146 #ifdef ZFS_DEBUG
147 IMPLY(last_size < PAGE_SIZE, *rahead == 0);
148 if (dbp[0]->db_offset != 0 || numbufs > 1) {
149 for (i = 0; i < numbufs; i++) {
150 ASSERT(ISP2(dbp[i]->db_size));
151 ASSERT3U((dbp[i]->db_offset % dbp[i]->db_size), ==, 0);
152 ASSERT3U(dbp[i]->db_size, ==, dbp[0]->db_size);
153 }
154 }
155 #endif
156
157 vmobj = ma[0]->object;
158
159 db = dbp[0];
160 for (i = 0; i < *rbehind; i++) {
161 m = vm_page_grab_unlocked(vmobj, ma[0]->pindex - 1 - i,
162 VM_ALLOC_NORMAL | VM_ALLOC_NOWAIT | VM_ALLOC_BUSY_FLAGS);
163 if (m == NULL)
164 break;
165 if (!vm_page_none_valid(m)) {
166 ASSERT3U(m->valid, ==, VM_PAGE_BITS_ALL);
167 vm_page_sunbusy(m);
168 break;
169 }
170 ASSERT3U(m->dirty, ==, 0);
171 ASSERT(!pmap_page_is_write_mapped(m));
172
173 ASSERT3U(db->db_size, >, PAGE_SIZE);
174 bufoff = IDX_TO_OFF(m->pindex) % db->db_size;
175 va = zfs_map_page(m, &sf);
176 ASSERT(db->db_data != NULL);
177 memcpy(va, (char *)db->db_data + bufoff, PAGESIZE);
178 zfs_unmap_page(sf);
179 vm_page_valid(m);
180 if ((m->busy_lock & VPB_BIT_WAITERS) != 0)
181 vm_page_activate(m);
182 else
183 vm_page_deactivate(m);
184 vm_page_sunbusy(m);
185 }
186 *rbehind = i;
187
188 bufoff = IDX_TO_OFF(ma[0]->pindex) % db->db_size;
189 pgoff = 0;
190 for (mi = 0, di = 0; mi < count && di < numbufs; ) {
191 if (pgoff == 0) {
192 m = ma[mi];
193 if (m != bogus_page) {
194 vm_page_assert_xbusied(m);
195 ASSERT(vm_page_none_valid(m));
196 ASSERT3U(m->dirty, ==, 0);
197 ASSERT(!pmap_page_is_write_mapped(m));
198 va = zfs_map_page(m, &sf);
199 }
200 }
201 if (bufoff == 0)
202 db = dbp[di];
203
204 if (m != bogus_page) {
205 ASSERT3U(IDX_TO_OFF(m->pindex) + pgoff, ==,
206 db->db_offset + bufoff);
207 }
208
209 /*
210 * We do not need to clamp the copy size by the file
211 * size as the last block is zero-filled beyond the
212 * end of file anyway.
213 */
214 tocpy = MIN(db->db_size - bufoff, PAGESIZE - pgoff);
215 ASSERT3S(tocpy, >=, 0);
216 if (m != bogus_page) {
217 ASSERT(db->db_data != NULL);
218 memcpy(va + pgoff, (char *)db->db_data + bufoff, tocpy);
219 }
220
221 pgoff += tocpy;
222 ASSERT3S(pgoff, >=, 0);
223 ASSERT3S(pgoff, <=, PAGESIZE);
224 if (pgoff == PAGESIZE) {
225 if (m != bogus_page) {
226 zfs_unmap_page(sf);
227 vm_page_valid(m);
228 }
229 ASSERT3S(mi, <, count);
230 mi++;
231 pgoff = 0;
232 }
233
234 bufoff += tocpy;
235 ASSERT3S(bufoff, >=, 0);
236 ASSERT3S(bufoff, <=, db->db_size);
237 if (bufoff == db->db_size) {
238 ASSERT3S(di, <, numbufs);
239 di++;
240 bufoff = 0;
241 }
242 }
243
244 #ifdef ZFS_DEBUG
245 /*
246 * Three possibilities:
247 * - last requested page ends at a buffer boundary and , thus,
248 * all pages and buffers have been iterated;
249 * - all requested pages are filled, but the last buffer
250 * has not been exhausted;
251 * the read-ahead is possible only in this case;
252 * - all buffers have been read, but the last page has not been
253 * fully filled;
254 * this is only possible if the file has only a single buffer
255 * with a size that is not a multiple of the page size.
256 */
257 if (mi == count) {
258 ASSERT3S(di, >=, numbufs - 1);
259 IMPLY(*rahead != 0, di == numbufs - 1);
260 IMPLY(*rahead != 0, bufoff != 0);
261 ASSERT0(pgoff);
262 }
263 if (di == numbufs) {
264 ASSERT3S(mi, >=, count - 1);
265 ASSERT0(*rahead);
266 IMPLY(pgoff == 0, mi == count);
267 if (pgoff != 0) {
268 ASSERT3S(mi, ==, count - 1);
269 ASSERT3U((dbp[0]->db_size & PAGE_MASK), !=, 0);
270 }
271 }
272 #endif
273 if (pgoff != 0) {
274 ASSERT3P(m, !=, bogus_page);
275 memset(va + pgoff, 0, PAGESIZE - pgoff);
276 zfs_unmap_page(sf);
277 vm_page_valid(m);
278 }
279
280 for (i = 0; i < *rahead; i++) {
281 m = vm_page_grab_unlocked(vmobj, ma[count - 1]->pindex + 1 + i,
282 VM_ALLOC_NORMAL | VM_ALLOC_NOWAIT | VM_ALLOC_BUSY_FLAGS);
283 if (m == NULL)
284 break;
285 if (!vm_page_none_valid(m)) {
286 ASSERT3U(m->valid, ==, VM_PAGE_BITS_ALL);
287 vm_page_sunbusy(m);
288 break;
289 }
290 ASSERT3U(m->dirty, ==, 0);
291 ASSERT(!pmap_page_is_write_mapped(m));
292
293 ASSERT3U(db->db_size, >, PAGE_SIZE);
294 bufoff = IDX_TO_OFF(m->pindex) % db->db_size;
295 tocpy = MIN(db->db_size - bufoff, PAGESIZE);
296 va = zfs_map_page(m, &sf);
297 ASSERT(db->db_data != NULL);
298 memcpy(va, (char *)db->db_data + bufoff, tocpy);
299 if (tocpy < PAGESIZE) {
300 ASSERT3S(i, ==, *rahead - 1);
301 ASSERT3U((db->db_size & PAGE_MASK), !=, 0);
302 memset(va + tocpy, 0, PAGESIZE - tocpy);
303 }
304 zfs_unmap_page(sf);
305 vm_page_valid(m);
306 if ((m->busy_lock & VPB_BIT_WAITERS) != 0)
307 vm_page_activate(m);
308 else
309 vm_page_deactivate(m);
310 vm_page_sunbusy(m);
311 }
312 *rahead = i;
313
314 dmu_buf_rele_array(dbp, numbufs, FTAG);
315 return (0);
316 }
317