2012-08-06 20:42:50 +02:00
|
|
|
/*
|
|
|
|
* Page cache for QEMU
|
|
|
|
* The cache is base on a hash of the page address
|
|
|
|
*
|
|
|
|
* Copyright 2012 Red Hat, Inc. and/or its affiliates
|
|
|
|
*
|
|
|
|
* Authors:
|
|
|
|
* Orit Wasserman <owasserm@redhat.com>
|
|
|
|
*
|
|
|
|
* This work is licensed under the terms of the GNU GPL, version 2 or later.
|
|
|
|
* See the COPYING file in the top-level directory.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <stdint.h>
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <stdlib.h>
|
|
|
|
#include <string.h>
|
|
|
|
#include <sys/time.h>
|
|
|
|
#include <sys/types.h>
|
|
|
|
#include <stdbool.h>
|
|
|
|
#include <glib.h>
|
|
|
|
|
|
|
|
#include "qemu-common.h"
|
2012-12-17 18:19:50 +01:00
|
|
|
#include "migration/page_cache.h"
|
2012-08-06 20:42:50 +02:00
|
|
|
|
|
|
|
#ifdef DEBUG_CACHE
|
|
|
|
#define DPRINTF(fmt, ...) \
|
|
|
|
do { fprintf(stdout, "cache: " fmt, ## __VA_ARGS__); } while (0)
|
|
|
|
#else
|
|
|
|
#define DPRINTF(fmt, ...) \
|
|
|
|
do { } while (0)
|
|
|
|
#endif
|
|
|
|
|
xbzrle: optimize XBZRLE to decrease the cache misses
Avoid hot pages being replaced by others to remarkably decrease cache
misses
Sample results with the test program which quote from xbzrle.txt ran in
vm:(migrate bandwidth:1GE and xbzrle cache size 8MB)
the test program:
include <stdlib.h>
include <stdio.h>
int main()
{
char *buf = (char *) calloc(4096, 4096);
while (1) {
int i;
for (i = 0; i < 4096 * 4; i++) {
buf[i * 4096 / 4]++;
}
printf(".");
}
}
before this patch:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":1020,"xbzrle-cache":{"bytes":1108284,
"cache-size":8388608,"cache-miss-rate":0.987013,"pages":18297,"overflow":8,
"cache-miss":1228737},"status":"active","setup-time":10,"total-time":52398,
"ram":{"total":12466991104,"remaining":1695744,"mbps":935.559472,
"transferred":5780760580,"dirty-sync-counter":271,"duplicate":2878530,
"dirty-pages-rate":29130,"skipped":0,"normal-bytes":5748592640,
"normal":1403465}},"id":"libvirt-706"}
18k pages sent compressed in 52 seconds.
cache-miss-rate is 98.7%, totally miss.
after optimizing:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":2054,"xbzrle-cache":{"bytes":5066763,
"cache-size":8388608,"cache-miss-rate":0.485924,"pages":194823,"overflow":0,
"cache-miss":210653},"status":"active","setup-time":11,"total-time":18729,
"ram":{"total":12466991104,"remaining":3895296,"mbps":937.663549,
"transferred":1615042219,"dirty-sync-counter":98,"duplicate":2869840,
"dirty-pages-rate":58781,"skipped":0,"normal-bytes":1588404224,
"normal":387794}},"id":"libvirt-266"}
194k pages sent compressed in 18 seconds.
The value of cache-miss-rate decrease to 48.59%.
Signed-off-by: ChenLiang <chenliang88@huawei.com>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
2014-11-24 12:55:47 +01:00
|
|
|
/* the page in cache will not be replaced in two cycles */
|
|
|
|
#define CACHED_PAGE_LIFETIME 2
|
|
|
|
|
2012-08-06 20:42:50 +02:00
|
|
|
typedef struct CacheItem CacheItem;
|
|
|
|
|
|
|
|
struct CacheItem {
|
|
|
|
uint64_t it_addr;
|
|
|
|
uint64_t it_age;
|
|
|
|
uint8_t *it_data;
|
|
|
|
};
|
|
|
|
|
|
|
|
struct PageCache {
|
|
|
|
CacheItem *page_cache;
|
|
|
|
unsigned int page_size;
|
|
|
|
int64_t max_num_items;
|
|
|
|
uint64_t max_item_age;
|
|
|
|
int64_t num_items;
|
|
|
|
};
|
|
|
|
|
|
|
|
PageCache *cache_init(int64_t num_pages, unsigned int page_size)
|
|
|
|
{
|
|
|
|
int64_t i;
|
|
|
|
|
|
|
|
PageCache *cache;
|
|
|
|
|
|
|
|
if (num_pages <= 0) {
|
|
|
|
DPRINTF("invalid number of pages\n");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-01-30 19:08:37 +01:00
|
|
|
/* We prefer not to abort if there is no memory */
|
|
|
|
cache = g_try_malloc(sizeof(*cache));
|
|
|
|
if (!cache) {
|
|
|
|
DPRINTF("Failed to allocate cache\n");
|
|
|
|
return NULL;
|
|
|
|
}
|
2012-08-06 20:42:50 +02:00
|
|
|
/* round down to the nearest power of 2 */
|
|
|
|
if (!is_power_of_2(num_pages)) {
|
|
|
|
num_pages = pow2floor(num_pages);
|
|
|
|
DPRINTF("rounding down to %" PRId64 "\n", num_pages);
|
|
|
|
}
|
|
|
|
cache->page_size = page_size;
|
|
|
|
cache->num_items = 0;
|
|
|
|
cache->max_item_age = 0;
|
|
|
|
cache->max_num_items = num_pages;
|
|
|
|
|
|
|
|
DPRINTF("Setting cache buckets to %" PRId64 "\n", cache->max_num_items);
|
|
|
|
|
2014-01-30 19:08:37 +01:00
|
|
|
/* We prefer not to abort if there is no memory */
|
|
|
|
cache->page_cache = g_try_malloc((cache->max_num_items) *
|
|
|
|
sizeof(*cache->page_cache));
|
|
|
|
if (!cache->page_cache) {
|
|
|
|
DPRINTF("Failed to allocate cache->page_cache\n");
|
|
|
|
g_free(cache);
|
|
|
|
return NULL;
|
|
|
|
}
|
2012-08-06 20:42:50 +02:00
|
|
|
|
|
|
|
for (i = 0; i < cache->max_num_items; i++) {
|
|
|
|
cache->page_cache[i].it_data = NULL;
|
|
|
|
cache->page_cache[i].it_age = 0;
|
|
|
|
cache->page_cache[i].it_addr = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return cache;
|
|
|
|
}
|
|
|
|
|
|
|
|
void cache_fini(PageCache *cache)
|
|
|
|
{
|
|
|
|
int64_t i;
|
|
|
|
|
|
|
|
g_assert(cache);
|
|
|
|
g_assert(cache->page_cache);
|
|
|
|
|
|
|
|
for (i = 0; i < cache->max_num_items; i++) {
|
|
|
|
g_free(cache->page_cache[i].it_data);
|
|
|
|
}
|
|
|
|
|
|
|
|
g_free(cache->page_cache);
|
|
|
|
cache->page_cache = NULL;
|
2014-06-02 14:16:55 +02:00
|
|
|
g_free(cache);
|
2012-08-06 20:42:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static size_t cache_get_cache_pos(const PageCache *cache,
|
|
|
|
uint64_t address)
|
|
|
|
{
|
|
|
|
size_t pos;
|
|
|
|
|
|
|
|
g_assert(cache->max_num_items);
|
|
|
|
pos = (address / cache->page_size) & (cache->max_num_items - 1);
|
|
|
|
return pos;
|
|
|
|
}
|
|
|
|
|
|
|
|
static CacheItem *cache_get_by_addr(const PageCache *cache, uint64_t addr)
|
|
|
|
{
|
|
|
|
size_t pos;
|
|
|
|
|
|
|
|
g_assert(cache);
|
|
|
|
g_assert(cache->page_cache);
|
|
|
|
|
|
|
|
pos = cache_get_cache_pos(cache, addr);
|
|
|
|
|
|
|
|
return &cache->page_cache[pos];
|
|
|
|
}
|
|
|
|
|
|
|
|
uint8_t *get_cached_data(const PageCache *cache, uint64_t addr)
|
|
|
|
{
|
|
|
|
return cache_get_by_addr(cache, addr)->it_data;
|
|
|
|
}
|
|
|
|
|
2014-11-24 12:55:48 +01:00
|
|
|
bool cache_is_cached(const PageCache *cache, uint64_t addr,
|
|
|
|
uint64_t current_age)
|
|
|
|
{
|
|
|
|
CacheItem *it;
|
|
|
|
|
|
|
|
it = cache_get_by_addr(cache, addr);
|
|
|
|
|
|
|
|
if (it->it_addr == addr) {
|
|
|
|
/* update the it_age when the cache hit */
|
|
|
|
it->it_age = current_age;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
xbzrle: optimize XBZRLE to decrease the cache misses
Avoid hot pages being replaced by others to remarkably decrease cache
misses
Sample results with the test program which quote from xbzrle.txt ran in
vm:(migrate bandwidth:1GE and xbzrle cache size 8MB)
the test program:
include <stdlib.h>
include <stdio.h>
int main()
{
char *buf = (char *) calloc(4096, 4096);
while (1) {
int i;
for (i = 0; i < 4096 * 4; i++) {
buf[i * 4096 / 4]++;
}
printf(".");
}
}
before this patch:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":1020,"xbzrle-cache":{"bytes":1108284,
"cache-size":8388608,"cache-miss-rate":0.987013,"pages":18297,"overflow":8,
"cache-miss":1228737},"status":"active","setup-time":10,"total-time":52398,
"ram":{"total":12466991104,"remaining":1695744,"mbps":935.559472,
"transferred":5780760580,"dirty-sync-counter":271,"duplicate":2878530,
"dirty-pages-rate":29130,"skipped":0,"normal-bytes":5748592640,
"normal":1403465}},"id":"libvirt-706"}
18k pages sent compressed in 52 seconds.
cache-miss-rate is 98.7%, totally miss.
after optimizing:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":2054,"xbzrle-cache":{"bytes":5066763,
"cache-size":8388608,"cache-miss-rate":0.485924,"pages":194823,"overflow":0,
"cache-miss":210653},"status":"active","setup-time":11,"total-time":18729,
"ram":{"total":12466991104,"remaining":3895296,"mbps":937.663549,
"transferred":1615042219,"dirty-sync-counter":98,"duplicate":2869840,
"dirty-pages-rate":58781,"skipped":0,"normal-bytes":1588404224,
"normal":387794}},"id":"libvirt-266"}
194k pages sent compressed in 18 seconds.
The value of cache-miss-rate decrease to 48.59%.
Signed-off-by: ChenLiang <chenliang88@huawei.com>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
2014-11-24 12:55:47 +01:00
|
|
|
int cache_insert(PageCache *cache, uint64_t addr, const uint8_t *pdata,
|
|
|
|
uint64_t current_age)
|
2012-08-06 20:42:50 +02:00
|
|
|
{
|
|
|
|
|
2014-11-24 12:55:48 +01:00
|
|
|
CacheItem *it;
|
2012-08-06 20:42:50 +02:00
|
|
|
|
|
|
|
/* actual update of entry */
|
|
|
|
it = cache_get_by_addr(cache, addr);
|
|
|
|
|
xbzrle: optimize XBZRLE to decrease the cache misses
Avoid hot pages being replaced by others to remarkably decrease cache
misses
Sample results with the test program which quote from xbzrle.txt ran in
vm:(migrate bandwidth:1GE and xbzrle cache size 8MB)
the test program:
include <stdlib.h>
include <stdio.h>
int main()
{
char *buf = (char *) calloc(4096, 4096);
while (1) {
int i;
for (i = 0; i < 4096 * 4; i++) {
buf[i * 4096 / 4]++;
}
printf(".");
}
}
before this patch:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":1020,"xbzrle-cache":{"bytes":1108284,
"cache-size":8388608,"cache-miss-rate":0.987013,"pages":18297,"overflow":8,
"cache-miss":1228737},"status":"active","setup-time":10,"total-time":52398,
"ram":{"total":12466991104,"remaining":1695744,"mbps":935.559472,
"transferred":5780760580,"dirty-sync-counter":271,"duplicate":2878530,
"dirty-pages-rate":29130,"skipped":0,"normal-bytes":5748592640,
"normal":1403465}},"id":"libvirt-706"}
18k pages sent compressed in 52 seconds.
cache-miss-rate is 98.7%, totally miss.
after optimizing:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":2054,"xbzrle-cache":{"bytes":5066763,
"cache-size":8388608,"cache-miss-rate":0.485924,"pages":194823,"overflow":0,
"cache-miss":210653},"status":"active","setup-time":11,"total-time":18729,
"ram":{"total":12466991104,"remaining":3895296,"mbps":937.663549,
"transferred":1615042219,"dirty-sync-counter":98,"duplicate":2869840,
"dirty-pages-rate":58781,"skipped":0,"normal-bytes":1588404224,
"normal":387794}},"id":"libvirt-266"}
194k pages sent compressed in 18 seconds.
The value of cache-miss-rate decrease to 48.59%.
Signed-off-by: ChenLiang <chenliang88@huawei.com>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
2014-11-24 12:55:47 +01:00
|
|
|
if (it->it_data && it->it_addr != addr &&
|
|
|
|
it->it_age + CACHED_PAGE_LIFETIME > current_age) {
|
|
|
|
/* the cache page is fresh, don't replace it */
|
|
|
|
return -1;
|
|
|
|
}
|
2014-01-30 19:08:38 +01:00
|
|
|
/* allocate page */
|
2012-08-06 20:42:50 +02:00
|
|
|
if (!it->it_data) {
|
2014-01-30 19:08:38 +01:00
|
|
|
it->it_data = g_try_malloc(cache->page_size);
|
|
|
|
if (!it->it_data) {
|
|
|
|
DPRINTF("Error allocating page\n");
|
|
|
|
return -1;
|
|
|
|
}
|
2012-08-06 20:42:50 +02:00
|
|
|
cache->num_items++;
|
|
|
|
}
|
|
|
|
|
2014-01-30 19:08:38 +01:00
|
|
|
memcpy(it->it_data, pdata, cache->page_size);
|
|
|
|
|
xbzrle: optimize XBZRLE to decrease the cache misses
Avoid hot pages being replaced by others to remarkably decrease cache
misses
Sample results with the test program which quote from xbzrle.txt ran in
vm:(migrate bandwidth:1GE and xbzrle cache size 8MB)
the test program:
include <stdlib.h>
include <stdio.h>
int main()
{
char *buf = (char *) calloc(4096, 4096);
while (1) {
int i;
for (i = 0; i < 4096 * 4; i++) {
buf[i * 4096 / 4]++;
}
printf(".");
}
}
before this patch:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":1020,"xbzrle-cache":{"bytes":1108284,
"cache-size":8388608,"cache-miss-rate":0.987013,"pages":18297,"overflow":8,
"cache-miss":1228737},"status":"active","setup-time":10,"total-time":52398,
"ram":{"total":12466991104,"remaining":1695744,"mbps":935.559472,
"transferred":5780760580,"dirty-sync-counter":271,"duplicate":2878530,
"dirty-pages-rate":29130,"skipped":0,"normal-bytes":5748592640,
"normal":1403465}},"id":"libvirt-706"}
18k pages sent compressed in 52 seconds.
cache-miss-rate is 98.7%, totally miss.
after optimizing:
virsh qemu-monitor-command test_vm '{"execute": "query-migrate"}'
{"return":{"expected-downtime":2054,"xbzrle-cache":{"bytes":5066763,
"cache-size":8388608,"cache-miss-rate":0.485924,"pages":194823,"overflow":0,
"cache-miss":210653},"status":"active","setup-time":11,"total-time":18729,
"ram":{"total":12466991104,"remaining":3895296,"mbps":937.663549,
"transferred":1615042219,"dirty-sync-counter":98,"duplicate":2869840,
"dirty-pages-rate":58781,"skipped":0,"normal-bytes":1588404224,
"normal":387794}},"id":"libvirt-266"}
194k pages sent compressed in 18 seconds.
The value of cache-miss-rate decrease to 48.59%.
Signed-off-by: ChenLiang <chenliang88@huawei.com>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
2014-11-24 12:55:47 +01:00
|
|
|
it->it_age = current_age;
|
2012-08-06 20:42:50 +02:00
|
|
|
it->it_addr = addr;
|
2014-01-30 19:08:38 +01:00
|
|
|
|
|
|
|
return 0;
|
2012-08-06 20:42:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
int64_t cache_resize(PageCache *cache, int64_t new_num_pages)
|
|
|
|
{
|
|
|
|
PageCache *new_cache;
|
|
|
|
int64_t i;
|
|
|
|
|
|
|
|
CacheItem *old_it, *new_it;
|
|
|
|
|
|
|
|
g_assert(cache);
|
|
|
|
|
|
|
|
/* cache was not inited */
|
|
|
|
if (cache->page_cache == NULL) {
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* same size */
|
|
|
|
if (pow2floor(new_num_pages) == cache->max_num_items) {
|
|
|
|
return cache->max_num_items;
|
|
|
|
}
|
|
|
|
|
|
|
|
new_cache = cache_init(new_num_pages, cache->page_size);
|
|
|
|
if (!(new_cache)) {
|
|
|
|
DPRINTF("Error creating new cache\n");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* move all data from old cache */
|
|
|
|
for (i = 0; i < cache->max_num_items; i++) {
|
|
|
|
old_it = &cache->page_cache[i];
|
|
|
|
if (old_it->it_addr != -1) {
|
|
|
|
/* check for collision, if there is, keep MRU page */
|
|
|
|
new_it = cache_get_by_addr(new_cache, old_it->it_addr);
|
2013-02-25 18:12:02 +01:00
|
|
|
if (new_it->it_data && new_it->it_age >= old_it->it_age) {
|
2012-08-06 20:42:50 +02:00
|
|
|
/* keep the MRU page */
|
2013-02-25 18:12:02 +01:00
|
|
|
g_free(old_it->it_data);
|
2012-08-06 20:42:50 +02:00
|
|
|
} else {
|
2013-02-25 18:12:02 +01:00
|
|
|
if (!new_it->it_data) {
|
|
|
|
new_cache->num_items++;
|
|
|
|
}
|
|
|
|
g_free(new_it->it_data);
|
|
|
|
new_it->it_data = old_it->it_data;
|
|
|
|
new_it->it_age = old_it->it_age;
|
|
|
|
new_it->it_addr = old_it->it_addr;
|
2012-08-06 20:42:50 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-02-25 18:12:01 +01:00
|
|
|
g_free(cache->page_cache);
|
2012-08-06 20:42:50 +02:00
|
|
|
cache->page_cache = new_cache->page_cache;
|
|
|
|
cache->max_num_items = new_cache->max_num_items;
|
|
|
|
cache->num_items = new_cache->num_items;
|
|
|
|
|
|
|
|
g_free(new_cache);
|
|
|
|
|
|
|
|
return cache->max_num_items;
|
|
|
|
}
|