kernel hacker修炼之道之内存管理-SLAB(创建SLAB高速缓存kmem_cache_create())
创建SLAB高速缓存kmem_cache_create() 作者：李万鹏 于北京 borqs
这篇主要讲解专用高速缓存的创建函数kmem_cache_create()的流程。kmem_cache_create()主要是建立的cache描述符，填充了其成员，设置cpu local slab，设置外置slab描述符还是内置slab描述符，并没有分配slab块和对象。最终将cache描述符添加到cache_chain链表上。
首先写了一个简单的测试程序：
#include
dmesg一下看log:
[ 103.617426] slab test module init
[ 103.617427] slab_ctor is called! object 0 has been inited!
[ 103.617429] slab_ctor is called! object 1 has been inited!
[ 103.617430] slab_ctor is called! object 2 has been inited!
............
............
[ 103.617761] slab_ctor is called! object 506 has been inited!
[ 103.617762] slab_ctor is called! object 507 has been inited!
[ 103.617763] object one has been created!
[ 103.617766] object two has been created!
当我没有调用kmem_cache_alloc()，只调用kmem_cache_create()的时候是没有调用对象的构造函数的，可以对照源码，调用kmem_cache_create()并没有分配slab，是在创建对象的时候发现没有空闲对象，调用cache_grow()分配一个slab，然后再分配对象。所以这里调用构造函数初始化的object都是刚分配的slab中的。cat/proc/slabinfo看一下，发现最上边出现我们新创建的cache，slab_test_cachep，其中active_objs为2，num_objs为508，正好与我们的实际信息相符合。
kmem_cache_t *kmem_cache_create (const char *name, size_t size, size_t align,
unsigned long flags, void (*ctor)(void*, kmem_cache_t *, unsigned long),
void (*dtor)(void*, kmem_cache_t *, unsigned long))
{
size_t left_over, slab_size, ralign;
kmem_cache_t *cachep = NULL;
/*高速缓存的名字不能为空，不能在中断上下文(这个函数应该会导致睡眠)，大小不能太大，也不能太小，如果有析构函数就必须有构造函数*/
if ((!name) ||
in_interrupt() ||
(size < BYTES_PER_WORD) ||
(size > (1<
printk(KERN_ERR "%s: Early error in slab %s\n",
__FUNCTION__, name);
BUG();
}
/*缺省的情况下，创建的cache大小要字对齐的*/
if (size & (BYTES_PER_WORD-1)) {
size += (BYTES_PER_WORD-1);
size &= ~(BYTES_PER_WORD-1);
}
/*
*如果想让新创建的slab高速缓存所包含的对象在第一级硬件高速缓存中对齐，则设置SLAB_HWCACHE_ALIGN标志
*这里的处理方式是，如果对象的大小大于cache line大小的一半，就在RAM中根据L1_CACHE_BYTES的倍数对齐对象。
*否则，对象的大小就是L1_CACHE_BYTES的因子取整。这可以保证一个小对象不会横跨两个高速缓存行。
*对齐的时候需要填充空白，即通过人为地增加对象的大小来获得较好的高速缓存性能，是一个以时间换空间的过程。
*/
if (flags & SLAB_HWCACHE_ALIGN) {
ralign = cache_line_size();
while (size <= ralign/2)
ralign /= 2;
} else {
ralign = BYTES_PER_WORD;
}
align = ralign;
/*在上一篇已经提到cache_cache包含其他高速缓存使用的高速缓存描述符，它也是一个cache，专门用来分配cache描述符的*/
/* Get cache's description obj. */
cachep = (kmem_cache_t *) kmem_cache_alloc(&cache_cache, SLAB_KERNEL);
if (!cachep)
goto opps;
/*将这个cache清0*/
memset(cachep, 0, sizeof(kmem_cache_t));
/*如果对象的大小大于1/8的页大小，则使用外部slab描述符*/
/* Determine if the slab management is 'on' or 'off' slab. */
if (size >= (PAGE_SIZE>>3))
/*
* Size is large, assume best to place the slab management obj
* off-slab (should allow better packing of objs).
*/
flags |= CFLGS_OFF_SLAB;
size = ALIGN(size, align);
/*在内碎片和cache所占页数之前取一个平衡值*/
if ((flags & SLAB_RECLAIM_ACCOUNT) && size <= PAGE_SIZE) {
/*
* A VFS-reclaimable slab tends to have most allocations
* as GFP_NOFS and we really don't want to have to be allocating
* higher-order pages when we are unable to shrink dcache.
*/
cachep->gfporder = 0;
cache_estimate(cachep->gfporder, size, align, flags,
&left_over, &cachep->num);
} else {
/*
* Calculate size (in pages) of slabs, and the num of objs per
* slab. This could be made much more intelligent. For now,
* try to avoid using high page-orders for slabs. When the
* gfp() funcs are more friendly towards high-order requests,
* this should be changed.
*/
do {
unsigned int break_flag = 0;
cal_wastage:
/*计算内碎片的大小*/
cache_estimate(cachep->gfporder, size, align, flags,
&left_over, &cachep->num);
if (break_flag)
break;
if (cachep->gfporder >= MAX_GFP_ORDER)
break;
if (!cachep->num)
goto next;
if (flags & CFLGS_OFF_SLAB &&
cachep->num > offslab_limit) {
/* This num of objs will cause problems. */
cachep->gfporder--;
break_flag++;
goto cal_wastage;
}
/*
* Large num of objs is good, but v. large slabs are
* currently bad for the gfp()s.
*/
if (cachep->gfporder >= slab_break_gfp_order)
break;
/*平衡的值*/
if ((left_over*8) <= (PAGE_SIZE<
break; /* Acceptable internal fragmentation. */
next:
cachep->gfporder++;
} while (1);
}
if (!cachep->num) {
printk("kmem_cache_create: couldn't create cache %s.\n", name);
kmem_cache_free(&cache_cache, cachep);
cachep = NULL;
goto opps;
/*按照align对齐*/
slab_size = ALIGN(cachep->num*sizeof(kmem_bufctl_t)
+ sizeof(struct slab), align);
/*如果内碎片的大小比slab管理对象(即slab描述符的大小+所有对象描述符的大小)大，则改成内部slab描述符*/
if (flags & CFLGS_OFF_SLAB && left_over >= slab_size) {
flags &= ~CFLGS_OFF_SLAB;
left_over -= slab_size;
}
/*如果是外置slab描述符，就不用对齐了*/
if (flags & CFLGS_OFF_SLAB) {
/* really off slab. No need for manual alignment */
slab_size = cachep->num*sizeof(kmem_bufctl_t)+sizeof(struct slab);
}
/*着色粒度的单位为CPU一级硬件高速缓存中cache line的大小*/
cachep->colour_off = cache_line_size();
/* Offset must be a multiple of the alignment. */
if (cachep->colour_off < align)
cachep->colour_off = align;
/*计算着色粒度的最大值*/
cachep->colour = left_over/cachep->colour_off;
/*获得单个slab的大小*/
cachep->slab_size = slab_size;
cachep->flags = flags;
cachep->gfpflags = 0;
if (flags & SLAB_CACHE_DMA)
cachep->gfpflags |= GFP_DMA;
spin_lock_init(&cachep->spinlock);
cachep->objsize = size;
/* NUMA */
/*初始化slabs_full，slabs_partial，slabs_free链表头*/
INIT_LIST_HEAD(&cachep->lists.slabs_full);
INIT_LIST_HEAD(&cachep->lists.slabs_partial);
INIT_LIST_HEAD(&cachep->lists.slabs_free);
/*
*如果是外置的slab描述符，则从malloc_sizes指向的26个高速还从中选择合适的，
*这里并没有分配外置的slab管理对象，而是在调用cache_grow()分配新的slab块的时候才分配的
*/
if (flags & CFLGS_OFF_SLAB)
cachep->slabp_cache = kmem_find_general_cachep(slab_size,0);
/*设置cache的名字和构造，析构函数*/
cachep->ctor = ctor;
cachep->dtor = dtor;
cachep->name = name;
/* Don't let CPUs to come and go */
lock_cpu_hotplug();
/*设置CPU local slab*/
if (g_cpucache_up == FULL) {
enable_cpucache(cachep);
} else {
if (g_cpucache_up == NONE) {
/* Note: the first kmem_cache_create must create
* the cache that's used by kmalloc(24), otherwise
* the creation of further caches will BUG().
*/
cachep->array[smp_processor_id()] = &initarray_generic.cache;
g_cpucache_up = PARTIAL;
} else {
cachep->array[smp_processor_id()] = kmalloc(sizeof(struct arraycache_init),GFP_KERNEL);
}
BUG_ON(!ac_data(cachep));
ac_data(cachep)->avail = 0;
ac_data(cachep)->limit = BOOT_CPUCACHE_ENTRIES;
ac_data(cachep)->batchcount = 1;
ac_data(cachep)->touched = 0;
cachep->batchcount = 1;
cachep->limit = BOOT_CPUCACHE_ENTRIES;
cachep->free_limit = (1+num_online_cpus())*cachep->batchcount
+ cachep->num;
}
cachep->lists.next_reap = jiffies + REAPTIMEOUT_LIST3 +
((unsigned long)cachep)%REAPTIMEOUT_LIST3;
/*获得使链表免受被同时访问的cache_chain_sem信号量*/
down(&cache_chain_sem);
{
struct list_head *p;
mm_segment_t old_fs;
old_fs = get_fs();
set_fs(KERNEL_DS);
list_for_each(p, &cache_chain) {
kmem_cache_t *pc = list_entry(p, kmem_cache_t, next);
char tmp;
/* This happens when the module gets unloaded and doesn't
destroy its slab cache and noone else reuses the vmalloc
area of the module. Print a warning. */
if (__get_user(tmp,pc->name)) {
printk("SLAB: cache with size %d has lost its name\n",
pc->objsize);
continue;
}
if (!strcmp(pc->name,name)) {
printk("kmem_cache_create: duplicate cache %s\n",name);
up(&cache_chain_sem);
unlock_cpu_hotplug();
BUG();
}
}
set_fs(old_fs);
}
/*将新建立的cache插入cache_chain链表中*/
list_add(&cachep->next, &cache_chain);
/*释放这个信号量*/
up(&cache_chain_sem);
unlock_cpu_hotplug();
opps:
if (!cachep && (flags & SLAB_PANIC))
panic("kmem_cache_create(): failed to create slab `%s'\n",
name);
return cachep;
}
EXPORT_SYMBOL(kmem_cache_create);