2 * L2/refcount table cache for the QCOW2 format
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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
25 #include "qemu/osdep.h"
29 typedef struct Qcow2CachedTable {
37 Qcow2CachedTable *entries;
38 struct Qcow2Cache *depends;
41 bool depends_on_flush;
44 uint64_t cache_clean_lru_counter;
47 static inline void *qcow2_cache_get_table_addr(Qcow2Cache *c, int table)
49 return (uint8_t *) c->table_array + (size_t) table * c->table_size;
52 static inline int qcow2_cache_get_table_idx(Qcow2Cache *c, void *table)
54 ptrdiff_t table_offset = (uint8_t *) table - (uint8_t *) c->table_array;
55 int idx = table_offset / c->table_size;
56 assert(idx >= 0 && idx < c->size && table_offset % c->table_size == 0);
60 static inline const char *qcow2_cache_get_name(BDRVQcow2State *s, Qcow2Cache *c)
62 if (c == s->refcount_block_cache) {
63 return "refcount block";
64 } else if (c == s->l2_table_cache) {
67 /* Do not abort, because this is not critical */
72 static void qcow2_cache_table_release(Qcow2Cache *c, int i, int num_tables)
74 /* Using MADV_DONTNEED to discard memory is a Linux-specific feature */
76 void *t = qcow2_cache_get_table_addr(c, i);
77 int align = getpagesize();
78 size_t mem_size = (size_t) c->table_size * num_tables;
79 size_t offset = QEMU_ALIGN_UP((uintptr_t) t, align) - (uintptr_t) t;
80 size_t length = QEMU_ALIGN_DOWN(mem_size - offset, align);
81 if (mem_size > offset && length > 0) {
82 madvise((uint8_t *) t + offset, length, MADV_DONTNEED);
87 static inline bool can_clean_entry(Qcow2Cache *c, int i)
89 Qcow2CachedTable *t = &c->entries[i];
90 return t->ref == 0 && !t->dirty && t->offset != 0 &&
91 t->lru_counter <= c->cache_clean_lru_counter;
94 void qcow2_cache_clean_unused(Qcow2Cache *c)
100 /* Skip the entries that we don't need to clean */
101 while (i < c->size && !can_clean_entry(c, i)) {
105 /* And count how many we can clean in a row */
106 while (i < c->size && can_clean_entry(c, i)) {
107 c->entries[i].offset = 0;
108 c->entries[i].lru_counter = 0;
114 qcow2_cache_table_release(c, i - to_clean, to_clean);
118 c->cache_clean_lru_counter = c->lru_counter;
121 Qcow2Cache *qcow2_cache_create(BlockDriverState *bs, int num_tables,
124 BDRVQcow2State *s = bs->opaque;
127 assert(num_tables > 0);
128 assert(is_power_of_2(table_size));
129 assert(table_size >= (1 << MIN_CLUSTER_BITS));
130 assert(table_size <= s->cluster_size);
132 c = g_new0(Qcow2Cache, 1);
133 c->size = num_tables;
134 c->table_size = table_size;
135 c->entries = g_try_new0(Qcow2CachedTable, num_tables);
136 c->table_array = qemu_try_blockalign(bs->file->bs,
137 (size_t) num_tables * c->table_size);
139 if (!c->entries || !c->table_array) {
140 qemu_vfree(c->table_array);
149 int qcow2_cache_destroy(Qcow2Cache *c)
153 for (i = 0; i < c->size; i++) {
154 assert(c->entries[i].ref == 0);
157 qemu_vfree(c->table_array);
164 static int qcow2_cache_flush_dependency(BlockDriverState *bs, Qcow2Cache *c)
168 ret = qcow2_cache_flush(bs, c->depends);
174 c->depends_on_flush = false;
179 static int qcow2_cache_entry_flush(BlockDriverState *bs, Qcow2Cache *c, int i)
181 BDRVQcow2State *s = bs->opaque;
184 if (!c->entries[i].dirty || !c->entries[i].offset) {
188 trace_qcow2_cache_entry_flush(qemu_coroutine_self(),
189 c == s->l2_table_cache, i);
192 ret = qcow2_cache_flush_dependency(bs, c);
193 } else if (c->depends_on_flush) {
194 ret = bdrv_flush(bs->file->bs);
196 c->depends_on_flush = false;
204 if (c == s->refcount_block_cache) {
205 ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_REFCOUNT_BLOCK,
206 c->entries[i].offset, c->table_size, false);
207 } else if (c == s->l2_table_cache) {
208 ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_ACTIVE_L2,
209 c->entries[i].offset, c->table_size, false);
211 ret = qcow2_pre_write_overlap_check(bs, 0,
212 c->entries[i].offset, c->table_size, false);
219 if (c == s->refcount_block_cache) {
220 BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_UPDATE_PART);
221 } else if (c == s->l2_table_cache) {
222 BLKDBG_EVENT(bs->file, BLKDBG_L2_UPDATE);
225 ret = bdrv_pwrite(bs->file, c->entries[i].offset,
226 qcow2_cache_get_table_addr(c, i), c->table_size);
231 c->entries[i].dirty = false;
236 int qcow2_cache_write(BlockDriverState *bs, Qcow2Cache *c)
238 BDRVQcow2State *s = bs->opaque;
243 trace_qcow2_cache_flush(qemu_coroutine_self(), c == s->l2_table_cache);
245 for (i = 0; i < c->size; i++) {
246 ret = qcow2_cache_entry_flush(bs, c, i);
247 if (ret < 0 && result != -ENOSPC) {
255 int qcow2_cache_flush(BlockDriverState *bs, Qcow2Cache *c)
257 int result = qcow2_cache_write(bs, c);
260 int ret = bdrv_flush(bs->file->bs);
269 int qcow2_cache_set_dependency(BlockDriverState *bs, Qcow2Cache *c,
270 Qcow2Cache *dependency)
274 if (dependency->depends) {
275 ret = qcow2_cache_flush_dependency(bs, dependency);
281 if (c->depends && (c->depends != dependency)) {
282 ret = qcow2_cache_flush_dependency(bs, c);
288 c->depends = dependency;
292 void qcow2_cache_depends_on_flush(Qcow2Cache *c)
294 c->depends_on_flush = true;
297 int qcow2_cache_empty(BlockDriverState *bs, Qcow2Cache *c)
301 ret = qcow2_cache_flush(bs, c);
306 for (i = 0; i < c->size; i++) {
307 assert(c->entries[i].ref == 0);
308 c->entries[i].offset = 0;
309 c->entries[i].lru_counter = 0;
312 qcow2_cache_table_release(c, 0, c->size);
319 static int qcow2_cache_do_get(BlockDriverState *bs, Qcow2Cache *c,
320 uint64_t offset, void **table, bool read_from_disk)
322 BDRVQcow2State *s = bs->opaque;
326 uint64_t min_lru_counter = UINT64_MAX;
327 int min_lru_index = -1;
331 trace_qcow2_cache_get(qemu_coroutine_self(), c == s->l2_table_cache,
332 offset, read_from_disk);
334 if (!QEMU_IS_ALIGNED(offset, c->table_size)) {
335 qcow2_signal_corruption(bs, true, -1, -1, "Cannot get entry from %s "
336 "cache: Offset %#" PRIx64 " is unaligned",
337 qcow2_cache_get_name(s, c), offset);
341 /* Check if the table is already cached */
342 i = lookup_index = (offset / c->table_size * 4) % c->size;
344 const Qcow2CachedTable *t = &c->entries[i];
345 if (t->offset == offset) {
348 if (t->ref == 0 && t->lru_counter < min_lru_counter) {
349 min_lru_counter = t->lru_counter;
352 if (++i == c->size) {
355 } while (i != lookup_index);
357 if (min_lru_index == -1) {
358 /* This can't happen in current synchronous code, but leave the check
359 * here as a reminder for whoever starts using AIO with the cache */
363 /* Cache miss: write a table back and replace it */
365 trace_qcow2_cache_get_replace_entry(qemu_coroutine_self(),
366 c == s->l2_table_cache, i);
368 ret = qcow2_cache_entry_flush(bs, c, i);
373 trace_qcow2_cache_get_read(qemu_coroutine_self(),
374 c == s->l2_table_cache, i);
375 c->entries[i].offset = 0;
376 if (read_from_disk) {
377 if (c == s->l2_table_cache) {
378 BLKDBG_EVENT(bs->file, BLKDBG_L2_LOAD);
381 ret = bdrv_pread(bs->file, offset,
382 qcow2_cache_get_table_addr(c, i),
389 c->entries[i].offset = offset;
391 /* And return the right table */
394 *table = qcow2_cache_get_table_addr(c, i);
396 trace_qcow2_cache_get_done(qemu_coroutine_self(),
397 c == s->l2_table_cache, i);
402 int qcow2_cache_get(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset,
405 return qcow2_cache_do_get(bs, c, offset, table, true);
408 int qcow2_cache_get_empty(BlockDriverState *bs, Qcow2Cache *c, uint64_t offset,
411 return qcow2_cache_do_get(bs, c, offset, table, false);
414 void qcow2_cache_put(Qcow2Cache *c, void **table)
416 int i = qcow2_cache_get_table_idx(c, *table);
421 if (c->entries[i].ref == 0) {
422 c->entries[i].lru_counter = ++c->lru_counter;
425 assert(c->entries[i].ref >= 0);
428 void qcow2_cache_entry_mark_dirty(Qcow2Cache *c, void *table)
430 int i = qcow2_cache_get_table_idx(c, table);
431 assert(c->entries[i].offset != 0);
432 c->entries[i].dirty = true;
435 void *qcow2_cache_is_table_offset(Qcow2Cache *c, uint64_t offset)
439 for (i = 0; i < c->size; i++) {
440 if (c->entries[i].offset == offset) {
441 return qcow2_cache_get_table_addr(c, i);
447 void qcow2_cache_discard(Qcow2Cache *c, void *table)
449 int i = qcow2_cache_get_table_idx(c, table);
451 assert(c->entries[i].ref == 0);
453 c->entries[i].offset = 0;
454 c->entries[i].lru_counter = 0;
455 c->entries[i].dirty = false;
457 qcow2_cache_table_release(c, i, 1);