xref: /openbmc/qemu/block/qcow2-cache.c (revision 77a8257e)
1 /*
2  * L2/refcount table cache for the QCOW2 format
3  *
4  * Copyright (c) 2010 Kevin Wolf <kwolf@redhat.com>
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 
25 #include "block/block_int.h"
26 #include "qemu-common.h"
27 #include "qcow2.h"
28 #include "trace.h"
29 
30 typedef struct Qcow2CachedTable {
31     void*   table;
32     int64_t offset;
33     bool    dirty;
34     int     cache_hits;
35     int     ref;
36 } Qcow2CachedTable;
37 
38 struct Qcow2Cache {
39     Qcow2CachedTable*       entries;
40     struct Qcow2Cache*      depends;
41     int                     size;
42     bool                    depends_on_flush;
43 };
44 
45 Qcow2Cache *qcow2_cache_create(BlockDriverState *bs, int num_tables)
46 {
47     BDRVQcowState *s = bs->opaque;
48     Qcow2Cache *c;
49     int i;
50 
51     c = g_new0(Qcow2Cache, 1);
52     c->size = num_tables;
53     c->entries = g_try_new0(Qcow2CachedTable, num_tables);
54     if (!c->entries) {
55         goto fail;
56     }
57 
58     for (i = 0; i < c->size; i++) {
59         c->entries[i].table = qemu_try_blockalign(bs->file, s->cluster_size);
60         if (c->entries[i].table == NULL) {
61             goto fail;
62         }
63     }
64 
65     return c;
66 
67 fail:
68     if (c->entries) {
69         for (i = 0; i < c->size; i++) {
70             qemu_vfree(c->entries[i].table);
71         }
72     }
73     g_free(c->entries);
74     g_free(c);
75     return NULL;
76 }
77 
78 int qcow2_cache_destroy(BlockDriverState* bs, Qcow2Cache *c)
79 {
80     int i;
81 
82     for (i = 0; i < c->size; i++) {
83         assert(c->entries[i].ref == 0);
84         qemu_vfree(c->entries[i].table);
85     }
86 
87     g_free(c->entries);
88     g_free(c);
89 
90     return 0;
91 }
92 
93 static int qcow2_cache_flush_dependency(BlockDriverState *bs, Qcow2Cache *c)
94 {
95     int ret;
96 
97     ret = qcow2_cache_flush(bs, c->depends);
98     if (ret < 0) {
99         return ret;
100     }
101 
102     c->depends = NULL;
103     c->depends_on_flush = false;
104 
105     return 0;
106 }
107 
108 static int qcow2_cache_entry_flush(BlockDriverState *bs, Qcow2Cache *c, int i)
109 {
110     BDRVQcowState *s = bs->opaque;
111     int ret = 0;
112 
113     if (!c->entries[i].dirty || !c->entries[i].offset) {
114         return 0;
115     }
116 
117     trace_qcow2_cache_entry_flush(qemu_coroutine_self(),
118                                   c == s->l2_table_cache, i);
119 
120     if (c->depends) {
121         ret = qcow2_cache_flush_dependency(bs, c);
122     } else if (c->depends_on_flush) {
123         ret = bdrv_flush(bs->file);
124         if (ret >= 0) {
125             c->depends_on_flush = false;
126         }
127     }
128 
129     if (ret < 0) {
130         return ret;
131     }
132 
133     if (c == s->refcount_block_cache) {
134         ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_REFCOUNT_BLOCK,
135                 c->entries[i].offset, s->cluster_size);
136     } else if (c == s->l2_table_cache) {
137         ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_ACTIVE_L2,
138                 c->entries[i].offset, s->cluster_size);
139     } else {
140         ret = qcow2_pre_write_overlap_check(bs, 0,
141                 c->entries[i].offset, s->cluster_size);
142     }
143 
144     if (ret < 0) {
145         return ret;
146     }
147 
148     if (c == s->refcount_block_cache) {
149         BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_UPDATE_PART);
150     } else if (c == s->l2_table_cache) {
151         BLKDBG_EVENT(bs->file, BLKDBG_L2_UPDATE);
152     }
153 
154     ret = bdrv_pwrite(bs->file, c->entries[i].offset, c->entries[i].table,
155         s->cluster_size);
156     if (ret < 0) {
157         return ret;
158     }
159 
160     c->entries[i].dirty = false;
161 
162     return 0;
163 }
164 
165 int qcow2_cache_flush(BlockDriverState *bs, Qcow2Cache *c)
166 {
167     BDRVQcowState *s = bs->opaque;
168     int result = 0;
169     int ret;
170     int i;
171 
172     trace_qcow2_cache_flush(qemu_coroutine_self(), c == s->l2_table_cache);
173 
174     for (i = 0; i < c->size; i++) {
175         ret = qcow2_cache_entry_flush(bs, c, i);
176         if (ret < 0 && result != -ENOSPC) {
177             result = ret;
178         }
179     }
180 
181     if (result == 0) {
182         ret = bdrv_flush(bs->file);
183         if (ret < 0) {
184             result = ret;
185         }
186     }
187 
188     return result;
189 }
190 
191 int qcow2_cache_set_dependency(BlockDriverState *bs, Qcow2Cache *c,
192     Qcow2Cache *dependency)
193 {
194     int ret;
195 
196     if (dependency->depends) {
197         ret = qcow2_cache_flush_dependency(bs, dependency);
198         if (ret < 0) {
199             return ret;
200         }
201     }
202 
203     if (c->depends && (c->depends != dependency)) {
204         ret = qcow2_cache_flush_dependency(bs, c);
205         if (ret < 0) {
206             return ret;
207         }
208     }
209 
210     c->depends = dependency;
211     return 0;
212 }
213 
214 void qcow2_cache_depends_on_flush(Qcow2Cache *c)
215 {
216     c->depends_on_flush = true;
217 }
218 
219 int qcow2_cache_empty(BlockDriverState *bs, Qcow2Cache *c)
220 {
221     int ret, i;
222 
223     ret = qcow2_cache_flush(bs, c);
224     if (ret < 0) {
225         return ret;
226     }
227 
228     for (i = 0; i < c->size; i++) {
229         assert(c->entries[i].ref == 0);
230         c->entries[i].offset = 0;
231         c->entries[i].cache_hits = 0;
232     }
233 
234     return 0;
235 }
236 
237 static int qcow2_cache_find_entry_to_replace(Qcow2Cache *c)
238 {
239     int i;
240     int min_count = INT_MAX;
241     int min_index = -1;
242 
243 
244     for (i = 0; i < c->size; i++) {
245         if (c->entries[i].ref) {
246             continue;
247         }
248 
249         if (c->entries[i].cache_hits < min_count) {
250             min_index = i;
251             min_count = c->entries[i].cache_hits;
252         }
253 
254         /* Give newer hits priority */
255         /* TODO Check how to optimize the replacement strategy */
256         if (c->entries[i].cache_hits > 1) {
257             c->entries[i].cache_hits /= 2;
258         }
259     }
260 
261     if (min_index == -1) {
262         /* This can't happen in current synchronous code, but leave the check
263          * here as a reminder for whoever starts using AIO with the cache */
264         abort();
265     }
266     return min_index;
267 }
268 
269 static int qcow2_cache_do_get(BlockDriverState *bs, Qcow2Cache *c,
270     uint64_t offset, void **table, bool read_from_disk)
271 {
272     BDRVQcowState *s = bs->opaque;
273     int i;
274     int ret;
275 
276     trace_qcow2_cache_get(qemu_coroutine_self(), c == s->l2_table_cache,
277                           offset, read_from_disk);
278 
279     /* Check if the table is already cached */
280     for (i = 0; i < c->size; i++) {
281         if (c->entries[i].offset == offset) {
282             goto found;
283         }
284     }
285 
286     /* If not, write a table back and replace it */
287     i = qcow2_cache_find_entry_to_replace(c);
288     trace_qcow2_cache_get_replace_entry(qemu_coroutine_self(),
289                                         c == s->l2_table_cache, i);
290     if (i < 0) {
291         return i;
292     }
293 
294     ret = qcow2_cache_entry_flush(bs, c, i);
295     if (ret < 0) {
296         return ret;
297     }
298 
299     trace_qcow2_cache_get_read(qemu_coroutine_self(),
300                                c == s->l2_table_cache, i);
301     c->entries[i].offset = 0;
302     if (read_from_disk) {
303         if (c == s->l2_table_cache) {
304             BLKDBG_EVENT(bs->file, BLKDBG_L2_LOAD);
305         }
306 
307         ret = bdrv_pread(bs->file, offset, c->entries[i].table, s->cluster_size);
308         if (ret < 0) {
309             return ret;
310         }
311     }
312 
313     /* Give the table some hits for the start so that it won't be replaced
314      * immediately. The number 32 is completely arbitrary. */
315     c->entries[i].cache_hits = 32;
316     c->entries[i].offset = offset;
317 
318     /* And return the right table */
319 found:
320     c->entries[i].cache_hits++;
321     c->entries[i].ref++;
322     *table = c->entries[i].table;
323 
324     trace_qcow2_cache_get_done(qemu_coroutine_self(),
325                                c == s->l2_table_cache, i);
326 
327     return 0;
328 }
329 
330 int qcow2_cache_get(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset,
331     void **table)
332 {
333     return qcow2_cache_do_get(bs, c, offset, table, true);
334 }
335 
336 int qcow2_cache_get_empty(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset,
337     void **table)
338 {
339     return qcow2_cache_do_get(bs, c, offset, table, false);
340 }
341 
342 int qcow2_cache_put(BlockDriverState *bs, Qcow2Cache *c, void **table)
343 {
344     int i;
345 
346     for (i = 0; i < c->size; i++) {
347         if (c->entries[i].table == *table) {
348             goto found;
349         }
350     }
351     return -ENOENT;
352 
353 found:
354     c->entries[i].ref--;
355     *table = NULL;
356 
357     assert(c->entries[i].ref >= 0);
358     return 0;
359 }
360 
361 void qcow2_cache_entry_mark_dirty(Qcow2Cache *c, void *table)
362 {
363     int i;
364 
365     for (i = 0; i < c->size; i++) {
366         if (c->entries[i].table == table) {
367             goto found;
368         }
369     }
370     abort();
371 
372 found:
373     c->entries[i].dirty = true;
374 }
375