Memcached源码分析之slabs.c

最新推荐文章于 2021-05-24 04:46:46 发布

weixin_34221112

最新推荐文章于 2021-05-24 04:46:46 发布

阅读量362

点赞数

文章标签： memcached 数据结构与算法内存管理

原文链接：http://www.cnblogs.com/guolanzhu/p/5850272.html

版权

#include "memcached.h"
#include <sys/stat.h>
#include <sys/socket.h>
#include <sys/signal.h>
#include <sys/resource.h>
#include <fcntl.h>
#include <netinet/in.h>
#include <errno.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <assert.h>
#include <pthread.h>
typedef struct {
unsigned int size; /* sizes of items */ //item或者说chunk的大小
unsigned int perslab; /* how many items per slab */ //每个slab有多少个item，slab又称“页”
/**
当前slabclass的空闲item链表，也是可用item链表，当前slabclass一切可以用的内存空间都在此，
这里是内存分配的入口，分配内存的时候都是在这个链表上挤一个出去。
ps：memcached的新版本才开始把slots作为“所有空闲的item链接”的用途，以前的版本slots链表保存的是“回收的item”的意思，
而旧版本新分配的slab，是用end_page_ptr指针及end_page_free来控制，此版本已不用。
*/
void *slots; /* list of item ptrs */
unsigned int sl_curr; /* total free items in list */ //当前slabclass还剩多少空闲的item，即上面的slots数
unsigned int slabs; /* how many slabs were allocated for this class */ //这个slabclass分配了多少个slab了
/**
slab_list是这个slabclass下的slabs列表，逻辑上是一个数组，每个元素是一个slab指针。
list_size是slab_list的元素个数。
注意这个list_size和上面的slabs的不同：
由于slab_list是一个空间大小固定的数组，是数组！而list_size是这个数组元素的个数，代表slab_list的空间大小。
slabs代表已经分配出去的slabs数，list_size则代表可以有多少个slabs数
所以当slabs等于list_size的时候代表这个slab_list已经满了，得增大空间。
*/
void **slab_list; /* array of slab pointers */
unsigned int list_size; /* size of prev array */
unsigned int killing; /* index+1 of dying slab, or zero if none */
size_t requested; /* The number of requested bytes */
} slabclass_t;
static slabclass_t slabclass[MAX_NUMBER_OF_SLAB_CLASSES];
static size_t mem_limit = 0; //内存上限
static size_t mem_malloced = 0; //已分配的内存
static int power_largest;
static void *mem_base = NULL; //预分配的内存空间
static void *mem_current = NULL;
static size_t mem_avail = 0;
static pthread_mutex_t slabs_lock = PTHREAD_MUTEX_INITIALIZER;
static pthread_mutex_t slabs_rebalance_lock = PTHREAD_MUTEX_INITIALIZER;
static int do_slabs_newslab(const unsigned int id);
static void *memory_allocate(size_t size);
static void do_slabs_free(void *ptr, const size_t size, unsigned int id);
static void slabs_preallocate (const unsigned int maxslabs);
//根据item大小找到合适的slabclass
unsigned int slabs_clsid(const size_t size) {
int res = POWER_SMALLEST;
if (size == 0)
return 0;
while (size > slabclass[res].size)
if (res++ == power_largest) /* won't fit in the biggest slab */
return 0;
return res;
}
/**
初始化slabs，这里会对一些内存管理进行初始化
*/
void slabs_init(const size_t limit, const double factor, const bool prealloc) {
int i = POWER_SMALLEST - 1;
unsigned int size = sizeof(item) + settings.chunk_size;
mem_limit = limit; //这个limit就是启动时候用户设置的-m xx中的xx，最大的内存上限
if (prealloc) {
/**
如果用户开启了预分配，则先把上限的内存先分配出来，放到mem_base全局变量中。
所以这个时候服务就拥有了一大坨内存，以后要分配的内存都是从这一坨里面割下来。
*/
mem_base = malloc(mem_limit);
if (mem_base != NULL) {
mem_current = mem_base;
mem_avail = mem_limit;
} else {
fprintf(stderr, "Warning: Failed to allocate requested memory in"
" one large chunk.\nWill allocate in smaller chunks\n");
}
}
//下面是初始化各个slabclass对象
memset(slabclass, 0, sizeof(slabclass));
while (++i < POWER_LARGEST && size <= settings.item_size_max / factor) {
/* Make sure items are always n-byte aligned */
if (size % CHUNK_ALIGN_BYTES)
size += CHUNK_ALIGN_BYTES - (size % CHUNK_ALIGN_BYTES);
slabclass[i].size = size;
slabclass[i].perslab = settings.item_size_max / slabclass[i].size;
size *= factor;
if (settings.verbose > 1) {
fprintf(stderr, "slab class %3d: chunk size %9u perslab %7u\n",
i, slabclass[i].size, slabclass[i].perslab);
}
}
power_largest = i;
slabclass[power_largest].size = settings.item_size_max;
slabclass[power_largest].perslab = 1;
if (settings.verbose > 1) {
fprintf(stderr, "slab class %3d: chunk size %9u perslab %7u\n",
i, slabclass[i].size, slabclass[i].perslab);
}
{
char *t_initial_malloc = getenv("T_MEMD_INITIAL_MALLOC");
if (t_initial_malloc) {
mem_malloced = (size_t)atol(t_initial_malloc);
}
}
if (prealloc) {
slabs_preallocate(power_largest);
}
}
/**
内存预分配，如果用户开启了预分配，则会调用此方法，先从mem_base为分每个slabclass割一个slab大小下来。
*/
static void slabs_preallocate (const unsigned int maxslabs) {
int i;
unsigned int prealloc = 0;
for (i = POWER_SMALLEST; i <= POWER_LARGEST; i++) {
if (++prealloc > maxslabs)
return;
if (do_slabs_newslab(i) == 0) {
fprintf(stderr, "Error while preallocating slab memory!\n"
"If using -L or other prealloc options, max memory must be "
"at least %d megabytes.\n", power_largest);
exit(1);
}
}
}
static int grow_slab_list (const unsigned int id) {
slabclass_t *p = &slabclass[id];
/**
p->slab_list是一个空间大小固定的数组，是数组！而list_size是这个数组分配的空间。
p->slabs代表已经分配出去的slabs数
而p->list_size代表可以用多少个slabs数
所以当slabs等于list_size的时候代表这个slab_list已经满了，得增大空间。
*/
if (p->slabs == p->list_size) {
size_t new_size = (p->list_size != 0) ? p->list_size * 2 : 16;
void *new_list = realloc(p->slab_list, new_size * sizeof(void *)); //
if (new_list == 0) return 0;
p->list_size = new_size;
p->slab_list = new_list;
}
return 1;
}
/**
把整个slab打散成一个个（也叫chunk）放到相应的slots链表中
*/
static void split_slab_page_into_freelist(char *ptr, const unsigned int id) {
slabclass_t *p = &slabclass[id];
int x;
for (x = 0; x < p->perslab; x++) {
do_slabs_free(ptr, 0, id); //这个函数主要作用是让当前item空间可用，即加到slots链表中。
ptr += p->size;
}
}
/**
为slabclass[id]分配新的slab，仅当当前的slabclass中slots没有空闲的空间才调用
此函数分配新的slab
*/
static int do_slabs_newslab(const unsigned int id) {
slabclass_t *p = &slabclass[id];
int len = settings.slab_reassign ? settings.item_size_max
: p->size * p->perslab; //先判断是否开启了自定义slab大小，如果没有就按默认的，即约1M
char *ptr;
/**
下面if的逻辑是：
如果内存超出了限制，分配失败进入if，返回0
否则调用grow_slab_list检查是否要增大slab_list的大小
如果在grow_slab_list返回失败，则不继续分配空间，进入if，返回0
否则分配空间memory_allocate，如果分配失败，同样进入if，返回0；
*/
if ((mem_limit && mem_malloced + len > mem_limit && p->slabs > 0) ||
(grow_slab_list(id) == 0) ||
((ptr = memory_allocate((size_t)len)) == 0)) {
MEMCACHED_SLABS_SLABCLASS_ALLOCATE_FAILED(id);
return 0;
}
memset(ptr, 0, (size_t)len); //清干净内存空间
split_slab_page_into_freelist(ptr, id); //把新申请的slab放到slots中去
p->slab_list[p->slabs++] = ptr; //把新的slab加到slab_list数组中
mem_malloced += len; //记下已分配的空间大小
MEMCACHED_SLABS_SLABCLASS_ALLOCATE(id);
return 1;
}
/**
根据item大小和slabsclass分配空间
*/
static void *do_slabs_alloc(const size_t size, unsigned int id) {
slabclass_t *p;
void *ret = NULL;
item *it = NULL;
if (id < POWER_SMALLEST || id > power_largest) { //默认最大是200，最小是1
MEMCACHED_SLABS_ALLOCATE_FAILED(size, 0);
return NULL;
}
p = &slabclass[id]; //slabclass是一个全局变量，是各个slabclass对象数组，在这取得当前id对应的slabclass
assert(p->sl_curr == 0 || ((item *)p->slots)->slabs_clsid == 0);
/* fail unless we have space at the end of a recently allocated page,
we have something on our freelist, or we could allocate a new page */
/**
下面这个if的逻辑相当于：
如果p->sl_curr==0，即slots链表中没有空闲的空间，则do_slabs_newslab分配新slab
如果p->sl_curr==0，且do_slabs_newslab分配新slab失败，则进入if，ret = NULL，否则进入下面的elseif
*/
if (! (p->sl_curr != 0 || do_slabs_newslab(id) != 0)) {
/* We don't have more memory available */
ret = NULL;
} else if (p->sl_curr != 0) { //如果进入此分支是因为slots链表中还有空闲的空间
/* return off our freelist */
//把空闲的item分配出去
it = (item *)p->slots;
p->slots = it->next;
if (it->next) it->next->prev = 0;
p->sl_curr--;
ret = (void *)it;
}
if (ret) {
p->requested += size; //分配成功，记下已分配的字节数
MEMCACHED_SLABS_ALLOCATE(size, id, p->size, ret);
} else {
MEMCACHED_SLABS_ALLOCATE_FAILED(size, id);
}
return ret;
}
/**
这个函数的命名虽然叫do_slabs_free，听上去好像是释放空间，其实质是把空间变成可用。
怎样的空间才算可用？就是加到当前slabclass的slots链表中而已。
所以新申请的slab也会调用这个函数，让整个slab变为可用。
ps: 以前的memcached版本slots链表保存的是回收的item空间，而
现在保存的是所有可用的item空间。
*/
static void do_slabs_free(void *ptr, const size_t size, unsigned int id) {
slabclass_t *p;
item *it;
assert(((item *)ptr)->slabs_clsid == 0);
assert(id >= POWER_SMALLEST && id <= power_largest);
if (id < POWER_SMALLEST || id > power_largest)
return;
MEMCACHED_SLABS_FREE(size, id, ptr);
p = &slabclass[id];
it = (item *)ptr;
it->it_flags |= ITEM_SLABBED; //把item标记为slabbed状态
it->prev = 0;
it->next = p->slots; //插入到slots链表中
if (it->next) it->next->prev = it;
p->slots = it;
p->sl_curr++; //空闲item数加1
p->requested -= size;
return;
}
static int nz_strcmp(int nzlength, const char *nz, const char *z) {
int zlength=strlen(z);
return (zlength == nzlength) && (strncmp(nz, z, zlength) == 0) ? 0 : -1;
}
bool get_stats(const char *stat_type, int nkey, ADD_STAT add_stats, void *c) {
bool ret = true;
if (add_stats != NULL) {
if (!stat_type) {
/* prepare general statistics for the engine */
STATS_LOCK();
APPEND_STAT("bytes", "%llu", (unsigned long long)stats.curr_bytes);
APPEND_STAT("curr_items", "%u", stats.curr_items);
APPEND_STAT("total_items", "%u", stats.total_items);
STATS_UNLOCK();
item_stats_totals(add_stats, c);
} else if (nz_strcmp(nkey, stat_type, "items") == 0) {
item_stats(add_stats, c);
} else if (nz_strcmp(nkey, stat_type, "slabs") == 0) {
slabs_stats(add_stats, c);
} else if (nz_strcmp(nkey, stat_type, "sizes") == 0) {
item_stats_sizes(add_stats, c);
} else {
ret = false;
}
} else {
ret = false;
}
return ret;
}
static void do_slabs_stats(ADD_STAT add_stats, void *c) {
int i, total;
/* Get the per-thread stats which contain some interesting aggregates */
struct thread_stats thread_stats;
threadlocal_stats_aggregate(&thread_stats);
total = 0;
for(i = POWER_SMALLEST; i <= power_largest; i++) {
slabclass_t *p = &slabclass[i];
if (p->slabs != 0) {
uint32_t perslab, slabs;
slabs = p->slabs;
perslab = p->perslab;
char key_str[STAT_KEY_LEN];
char val_str[STAT_VAL_LEN];
int klen = 0, vlen = 0;
APPEND_NUM_STAT(i, "chunk_size", "%u", p->size);
APPEND_NUM_STAT(i, "chunks_per_page", "%u", perslab);
APPEND_NUM_STAT(i, "total_pages", "%u", slabs);
APPEND_NUM_STAT(i, "total_chunks", "%u", slabs * perslab);
APPEND_NUM_STAT(i, "used_chunks", "%u",
slabs*perslab - p->sl_curr);
APPEND_NUM_STAT(i, "free_chunks", "%u", p->sl_curr);
/* Stat is dead, but displaying zero instead of removing it. */
APPEND_NUM_STAT(i, "free_chunks_end", "%u", 0);
APPEND_NUM_STAT(i, "mem_requested", "%llu",
(unsigned long long)p->requested);
APPEND_NUM_STAT(i, "get_hits", "%llu",
(unsigned long long)thread_stats.slab_stats[i].get_hits);
APPEND_NUM_STAT(i, "cmd_set", "%llu",
(unsigned long long)thread_stats.slab_stats[i].set_cmds);
APPEND_NUM_STAT(i, "delete_hits", "%llu",
(unsigned long long)thread_stats.slab_stats[i].delete_hits);
APPEND_NUM_STAT(i, "incr_hits", "%llu",
(unsigned long long)thread_stats.slab_stats[i].incr_hits);
APPEND_NUM_STAT(i, "decr_hits", "%llu",
(unsigned long long)thread_stats.slab_stats[i].decr_hits);
APPEND_NUM_STAT(i, "cas_hits", "%llu",
(unsigned long long)thread_stats.slab_stats[i].cas_hits);
APPEND_NUM_STAT(i, "cas_badval", "%llu",
(unsigned long long)thread_stats.slab_stats[i].cas_badval);
APPEND_NUM_STAT(i, "touch_hits", "%llu",
(unsigned long long)thread_stats.slab_stats[i].touch_hits);
total++;
}
}
APPEND_STAT("active_slabs", "%d", total);
APPEND_STAT("total_malloced", "%llu", (unsigned long long)mem_malloced);
add_stats(NULL, 0, NULL, 0, c);
}
/**
分配内存空间
*/
static void *memory_allocate(size_t size) {
void *ret;
/**
有两种分配策略
1）如果是开启了内存预分配策略，则只需要从预分配好的内存块那里割一块出来。即进入下面的else分支
2）如果没有开启预分配，则malloc分配内存
关于预分配详见 slabs_init
*/
if (mem_base == NULL) {
/* We are not using a preallocated large memory chunk */
ret = malloc(size);
} else {
ret = mem_current;
if (size > mem_avail) {
return NULL;
}
/* mem_current pointer _must_ be aligned!!! */
if (size % CHUNK_ALIGN_BYTES) {
size += CHUNK_ALIGN_BYTES - (size % CHUNK_ALIGN_BYTES);
}
mem_current = ((char*)mem_current) + size;
if (size < mem_avail) {
mem_avail -= size;
} else {
mem_avail = 0;
}
}
return ret;
}
void *slabs_alloc(size_t size, unsigned int id) {
void *ret;
pthread_mutex_lock(&slabs_lock);
ret = do_slabs_alloc(size, id);
pthread_mutex_unlock(&slabs_lock);
return ret;
}
void slabs_free(void *ptr, size_t size, unsigned int id) {
pthread_mutex_lock(&slabs_lock);
do_slabs_free(ptr, size, id);
pthread_mutex_unlock(&slabs_lock);
}
void slabs_stats(ADD_STAT add_stats, void *c) {
pthread_mutex_lock(&slabs_lock);
do_slabs_stats(add_stats, c);
pthread_mutex_unlock(&slabs_lock);
}
void slabs_adjust_mem_requested(unsigned int id, size_t old, size_t ntotal)
{
pthread_mutex_lock(&slabs_lock);
slabclass_t *p;
if (id < POWER_SMALLEST || id > power_largest) {
fprintf(stderr, "Internal error! Invalid slab class\n");
abort();
}
p = &slabclass[id];
p->requested = p->requested - old + ntotal;
pthread_mutex_unlock(&slabs_lock);
}
static pthread_cond_t maintenance_cond = PTHREAD_COND_INITIALIZER;
static pthread_cond_t slab_rebalance_cond = PTHREAD_COND_INITIALIZER;
static volatile int do_run_slab_thread = 1;
static volatile int do_run_slab_rebalance_thread = 1;
#define DEFAULT_SLAB_BULK_CHECK 1
int slab_bulk_check = DEFAULT_SLAB_BULK_CHECK;
static int slab_rebalance_start(void) {
slabclass_t *s_cls;
int no_go = 0;
pthread_mutex_lock(&cache_lock);
pthread_mutex_lock(&slabs_lock);
if (slab_rebal.s_clsid < POWER_SMALLEST ||
slab_rebal.s_clsid > power_largest ||
slab_rebal.d_clsid < POWER_SMALLEST ||
slab_rebal.d_clsid > power_largest ||
slab_rebal.s_clsid == slab_rebal.d_clsid)
no_go = -2;
s_cls = &slabclass[slab_rebal.s_clsid];
if (!grow_slab_list(slab_rebal.d_clsid)) {
no_go = -1;
}
if (s_cls->slabs < 2)
no_go = -3;
if (no_go != 0) {
pthread_mutex_unlock(&slabs_lock);
pthread_mutex_unlock(&cache_lock);
return no_go; /* Should use a wrapper function... */
}
s_cls->killing = 1;
slab_rebal.slab_start = s_cls->slab_list[s_cls->killing - 1];
slab_rebal.slab_end = (char *)slab_rebal.slab_start +
(s_cls->size * s_cls->perslab);
slab_rebal.slab_pos = slab_rebal.slab_start;
slab_rebal.done = 0;
/* Also tells do_item_get to search for items in this slab */
slab_rebalance_signal = 2;
if (settings.verbose > 1) {
fprintf(stderr, "Started a slab rebalance\n");
}
pthread_mutex_unlock(&slabs_lock);
pthread_mutex_unlock(&cache_lock);
STATS_LOCK();
stats.slab_reassign_running = true;
STATS_UNLOCK();
return 0;
}
enum move_status {
MOVE_PASS=0, MOVE_DONE, MOVE_BUSY, MOVE_LOCKED
};
static int slab_rebalance_move(void) {
slabclass_t *s_cls;
int x;
int was_busy = 0;
int refcount = 0;
enum move_status status = MOVE_PASS;
pthread_mutex_lock(&cache_lock);
pthread_mutex_lock(&slabs_lock);
s_cls = &slabclass[slab_rebal.s_clsid];
for (x = 0; x < slab_bulk_check; x++) {
item *it = slab_rebal.slab_pos;
status = MOVE_PASS;
if (it->slabs_clsid != 255) {
void *hold_lock = NULL;
uint32_t hv = hash(ITEM_key(it), it->nkey);
if ((hold_lock = item_trylock(hv)) == NULL) {
status = MOVE_LOCKED;
} else {
refcount = refcount_incr(&it->refcount);
if (refcount == 1) { /* item is unlinked, unused */
if (it->it_flags & ITEM_SLABBED) {
/* remove from slab freelist */
if (s_cls->slots == it) {
s_cls->slots = it->next;
}
if (it->next) it->next->prev = it->prev;
if (it->prev) it->prev->next = it->next;
s_cls->sl_curr--;
status = MOVE_DONE;
} else {
status = MOVE_BUSY;
}
} else if (refcount == 2) { /* item is linked but not busy */
if ((it->it_flags & ITEM_LINKED) != 0) {
do_item_unlink_nolock(it, hv);
status = MOVE_DONE;
} else {
/* refcount == 1 + !ITEM_LINKED means the item is being
* uploaded to, or was just unlinked but hasn't been freed
* yet. Let it bleed off on its own and try again later */
status = MOVE_BUSY;
}
} else {
if (settings.verbose > 2) {
fprintf(stderr, "Slab reassign hit a busy item: refcount: %d (%d -> %d)\n",
it->refcount, slab_rebal.s_clsid, slab_rebal.d_clsid);
}
status = MOVE_BUSY;
}
item_trylock_unlock(hold_lock);
}
}
switch (status) {
case MOVE_DONE:
it->refcount = 0;
it->it_flags = 0;
it->slabs_clsid = 255;
break;
case MOVE_BUSY:
refcount_decr(&it->refcount);
case MOVE_LOCKED:
slab_rebal.busy_items++;
was_busy++;
break;
case MOVE_PASS:
break;
}
slab_rebal.slab_pos = (char *)slab_rebal.slab_pos + s_cls->size;
if (slab_rebal.slab_pos >= slab_rebal.slab_end)
break;
}
if (slab_rebal.slab_pos >= slab_rebal.slab_end) {
/* Some items were busy, start again from the top */
if (slab_rebal.busy_items) {
slab_rebal.slab_pos = slab_rebal.slab_start;
slab_rebal.busy_items = 0;
} else {
slab_rebal.done++;
}
}
pthread_mutex_unlock(&slabs_lock);
pthread_mutex_unlock(&cache_lock);
return was_busy;
}
static void slab_rebalance_finish(void) {
slabclass_t *s_cls;
slabclass_t *d_cls;
pthread_mutex_lock(&cache_lock);
pthread_mutex_lock(&slabs_lock);
s_cls = &slabclass[slab_rebal.s_clsid];
d_cls = &slabclass[slab_rebal.d_clsid];
/* At this point the stolen slab is completely clear */
s_cls->slab_list[s_cls->killing - 1] =
s_cls->slab_list[s_cls->slabs - 1];
s_cls->slabs--;
s_cls->killing = 0;
memset(slab_rebal.slab_start, 0, (size_t)settings.item_size_max);
d_cls->slab_list[d_cls->slabs++] = slab_rebal.slab_start;
split_slab_page_into_freelist(slab_rebal.slab_start,
slab_rebal.d_clsid);
slab_rebal.done = 0;
slab_rebal.s_clsid = 0;
slab_rebal.d_clsid = 0;
slab_rebal.slab_start = NULL;
slab_rebal.slab_end = NULL;
slab_rebal.slab_pos = NULL;
slab_rebalance_signal = 0;
pthread_mutex_unlock(&slabs_lock);
pthread_mutex_unlock(&cache_lock);
STATS_LOCK();
stats.slab_reassign_running = false;
stats.slabs_moved++;
STATS_UNLOCK();
if (settings.verbose > 1) {
fprintf(stderr, "finished a slab move\n");
}
}
/*
slab自动重分配时，执行此函数做出重分配方案决定
*/
static int slab_automove_decision(int *src, int *dst) {
static uint64_t evicted_old[POWER_LARGEST];
static unsigned int slab_zeroes[POWER_LARGEST];
static unsigned int slab_winner = 0;
static unsigned int slab_wins = 0;
uint64_t evicted_new[POWER_LARGEST];
uint64_t evicted_diff = 0;
uint64_t evicted_max = 0;
unsigned int highest_slab = 0;
unsigned int total_pages[POWER_LARGEST];
int i;
int source = 0;
int dest = 0;
static rel_time_t next_run;
/* Run less frequently than the slabmove tester. */
if (current_time >= next_run) {
next_run = current_time + 10;
} else {
return 0;
}
item_stats_evictions(evicted_new);
pthread_mutex_lock(&cache_lock);
for (i = POWER_SMALLEST; i < power_largest; i++) {
total_pages[i] = slabclass[i].slabs;
}
pthread_mutex_unlock(&cache_lock);
/* Find a candidate source; something with zero evicts 3+ times */
for (i = POWER_SMALLEST; i < power_largest; i++) {
evicted_diff = evicted_new[i] - evicted_old[i];
if (evicted_diff == 0 && total_pages[i] > 2) {
slab_zeroes[i]++;
if (source == 0 && slab_zeroes[i] >= 3)
source = i;
} else {
slab_zeroes[i] = 0;
if (evicted_diff > evicted_max) {
evicted_max = evicted_diff;
highest_slab = i;
}
}
evicted_old[i] = evicted_new[i];
}
/* Pick a valid destination */
if (slab_winner != 0 && slab_winner == highest_slab) {
slab_wins++;
if (slab_wins >= 3)
dest = slab_winner;
} else {
slab_wins = 1;
slab_winner = highest_slab;
}
if (source && dest) {
*src = source;
*dst = dest;
return 1;
}
return 0;
}
/* Slab rebalancer thread.
* Does not use spinlocks since it is not timing sensitive. Burn less CPU and
* go to sleep if locks are contended
运行slab维护线程，slab维护线程的执行入口
*/
static void *slab_maintenance_thread(void *arg) {
int src, dest;
while (do_run_slab_thread) {
if (settings.slab_automove == 1) {
if (slab_automove_decision(&src, &dest) == 1) {
/* Blind to the return codes. It will retry on its own */
slabs_reassign(src, dest); //移动slab，重分配
}
sleep(1);
} else {
/* Don't wake as often if we're not enabled.
* This is lazier than setting up a condition right now. */
sleep(5);
}
}
return NULL;
}
/* Slab mover thread.
* Sits waiting for a condition to jump off and shovel some memory about
*/
static void *slab_rebalance_thread(void *arg) {
int was_busy = 0;
/* So we first pass into cond_wait with the mutex held */
mutex_lock(&slabs_rebalance_lock);
while (do_run_slab_rebalance_thread) {
if (slab_rebalance_signal == 1) {
if (slab_rebalance_start() < 0) {
/* Handle errors with more specifity as required. */
slab_rebalance_signal = 0;
}
was_busy = 0;
} else if (slab_rebalance_signal && slab_rebal.slab_start != NULL) {
was_busy = slab_rebalance_move();
}
if (slab_rebal.done) {
slab_rebalance_finish();
} else if (was_busy) {
/* Stuck waiting for some items to unlock, so slow down a bit
* to give them a chance to free up */
usleep(50);
}
if (slab_rebalance_signal == 0) {
/* always hold this lock while we're running */
pthread_cond_wait(&slab_rebalance_cond, &slabs_rebalance_lock);
}
}
return NULL;
}
static int slabs_reassign_pick_any(int dst) {
static int cur = POWER_SMALLEST - 1;
int tries = power_largest - POWER_SMALLEST + 1;
for (; tries > 0; tries--) {
cur++;
if (cur > power_largest)
cur = POWER_SMALLEST;
if (cur == dst)
continue;
if (slabclass[cur].slabs > 1) {
return cur;
}
}
return -1;
}
static enum reassign_result_type do_slabs_reassign(int src, int dst) {
if (slab_rebalance_signal != 0)
return REASSIGN_RUNNING;
if (src == dst)
return REASSIGN_SRC_DST_SAME;
/* Special indicator to choose ourselves. */
if (src == -1) {
src = slabs_reassign_pick_any(dst);
/* TODO: If we end up back at -1, return a new error type */
}
if (src < POWER_SMALLEST || src > power_largest ||
dst < POWER_SMALLEST || dst > power_largest)
return REASSIGN_BADCLASS;
if (slabclass[src].slabs < 2)
return REASSIGN_NOSPARE;
slab_rebal.s_clsid = src;
slab_rebal.d_clsid = dst;
slab_rebalance_signal = 1;
pthread_cond_signal(&slab_rebalance_cond);
return REASSIGN_OK;
}
enum reassign_result_type slabs_reassign(int src, int dst) {
enum reassign_result_type ret;
if (pthread_mutex_trylock(&slabs_rebalance_lock) != 0) {
return REASSIGN_RUNNING;
}
ret = do_slabs_reassign(src, dst);
pthread_mutex_unlock(&slabs_rebalance_lock);
return ret;
}
/* If we hold this lock, rebalancer can't wake up or move */
void slabs_rebalancer_pause(void) {
pthread_mutex_lock(&slabs_rebalance_lock);
}
void slabs_rebalancer_resume(void) {
pthread_mutex_unlock(&slabs_rebalance_lock);
}
static pthread_t maintenance_tid;
static pthread_t rebalance_tid;
/**
启动slab维护线程
*/
int start_slab_maintenance_thread(void) {
int ret;
slab_rebalance_signal = 0;
slab_rebal.slab_start = NULL;
char *env = getenv("MEMCACHED_SLAB_BULK_CHECK");
if (env != NULL) {
slab_bulk_check = atoi(env);
if (slab_bulk_check == 0) {
slab_bulk_check = DEFAULT_SLAB_BULK_CHECK;
}
}
if (pthread_cond_init(&slab_rebalance_cond, NULL) != 0) {
fprintf(stderr, "Can't intiialize rebalance condition\n");
return -1;
}
pthread_mutex_init(&slabs_rebalance_lock, NULL);
if ((ret = pthread_create(&maintenance_tid, NULL,
slab_maintenance_thread, NULL)) != 0) {
fprintf(stderr, "Can't create slab maint thread: %s\n", strerror(ret));
return -1;
}
if ((ret = pthread_create(&rebalance_tid, NULL,
slab_rebalance_thread, NULL)) != 0) {
fprintf(stderr, "Can't create rebal thread: %s\n", strerror(ret));
return -1;
}
return 0;
}
/**
停止slab维护线程，逻辑和停止哈希表维护线程一样。
*/
void stop_slab_maintenance_thread(void) {
mutex_lock(&cache_lock);
do_run_slab_thread = 0;
do_run_slab_rebalance_thread = 0;
pthread_cond_signal(&maintenance_cond);
pthread_mutex_unlock(&cache_lock);
/* Wait for the maintenance thread to stop */
pthread_join(maintenance_tid, NULL);
pthread_join(rebalance_tid, NULL);
}