声明:此处代码分析,来源与 nuttx 12.8.0版本。
/****************************************************************************
* Name: mm_extend
*
* Description:
* Extend a heap region by add a block of (virtually) contiguous memory
* to the end of the heap.
*
****************************************************************************/
void mm_extend(FAR struct mm_heap_s *heap, FAR void *mem, size_t size,
int region)
{
FAR struct mm_allocnode_s *oldnode;
FAR struct mm_allocnode_s *newnode;
uintptr_t blockstart;
uintptr_t blockend;
/* Make sure that we were passed valid parameters */
DEBUGASSERT(heap && mem);
#if CONFIG_MM_REGIONS > 1
DEBUGASSERT(size >= MIN_EXTEND && region >= 0 &&
region < heap->mm_nregions);
#else
DEBUGASSERT(size >= MIN_EXTEND && region == 0);
#endif
/* Make sure that the memory region are properly aligned */
blockstart = (uintptr_t)mem;
blockend = blockstart + size;
DEBUGASSERT(MM_ALIGN_UP(blockstart) == blockstart);
DEBUGASSERT(MM_ALIGN_DOWN(blockend) == blockend);
/* Take the memory manager mutex */
DEBUGVERIFY(mm_lock(heap));
/* Get the terminal node in the old heap. The block to extend must
* immediately follow this node.
*/
oldnode = heap->mm_heapend[region];
DEBUGASSERT((uintptr_t)oldnode + MM_SIZEOF_ALLOCNODE == blockstart);
/* The size of the old node now extends to the new terminal node.
* This is the old size (MM_SIZEOF_ALLOCNODE) plus the size of
* the block (size) minus the size of the new terminal node
* (MM_SIZEOF_ALLOCNODE) or simply:
*/
oldnode->size = size | (oldnode->size & MM_MASK_BIT);
/* The old node should already be marked as allocated */
DEBUGASSERT(MM_NODE_IS_ALLOC(oldnode));
/* Get and initialize the new terminal node in the heap */
newnode = (FAR struct mm_allocnode_s *)
(blockend - MM_SIZEOF_ALLOCNODE);
newnode->size = MM_SIZEOF_ALLOCNODE | MM_ALLOC_BIT;
heap->mm_heapend[region] = newnode;
/* Finally, increase the total heap size accordingly */
heap->mm_heapsize += size;
mm_unlock(heap);
/* Finally "free" the new block of memory where the old terminal node was
* located.
*/
mm_free(heap, mem);
}
显然,此函数用于扩展堆空间。首先对函数的参数进行合理性检验自然属于基本操作。对于size的检验标准为size>=MIN_EXTEND,那么,问题来了,为何这样的size就是合理的呢?首先我们先来看看MIN_EXTEND的定义。
#define MIN_EXTEND (2 * MM_SIZEOF_ALLOCNODE)
可以看的出来,size大小的空间至少能容纳得下两个struct mm_allocnode_s。众所周知,堆是由struct mm_heap_s来进行管理的。mm_heap_s的成员mm_heapend成员用来标记堆空间的结束,这就解释了MIN_EXTEND中的一个struct mm_allocnode_s结构的来源。我们的函数功能是扩展堆空间,那么,扩展的空间必然是空闲的。对堆空间的扩展实质上就是增加struct mm_heap_s 的mm_nodelist的成员。struct mm_freenode_s用于空闲空间的管理。且,struct mm_freenode_s结构是存在于每个空闲空间的开头。这就解释了MIN_EXTEND中的第二个struct mm_allocnode_s结构的来源。有人说了,struct mm_freenode_s与struct mm_allocnode_s相差两个struct mm_freenode_s这指针的大小,确实,按照严格意义来说,应该修改MIN_EXTEND的定义如下。
#define MIN_EXTEND (sizeof(struct mm_freenode_s) + MM_SIZEOF_ALLOCNODE)
接下来的一个问题是,为什么会有如下代码。
oldnode->size = size | (oldnode->size & MM_MASK_BIT);
我的意思是,oldnode是属于struct mm_allocnode_s这个类型的,根据mm_addregion我们知道,对于新加入堆的空间struct mm_allocnode_s的size成员变量表示大小的部分,其值一般是MM_SIZEOF_ALLOCNODE,如下。
void mm_addregion(FAR struct mm_heap_s *heap, FAR void *heapstart,
size_t heapsize)
{
......
heap->mm_heapstart[idx]->size = MM_SIZEOF_ALLOCNODE | MM_ALLOC_BIT;
......
heap->mm_heapend[idx]->size = MM_SIZEOF_ALLOCNODE | MM_ALLOC_BIT |
MM_PREVFREE_BIT;
......
}
对于此处oldnode->size,其原因是,此时的oldnode,在mm_free间接调用mm_malloc_size时,mm_malloc_size做了强制类型转换。这也印证了我们之前说的,struct mm_freenode_s结构是存在于每个空闲空间的开头。
size_t mm_malloc_size(FAR struct mm_heap_s *heap, FAR void *mem)
{
......
node = (FAR struct mm_freenode_s *)((FAR char *)mem - MM_SIZEOF_ALLOCNODE);
......
}
那么,struct mm_freenode_s的size成员是表示的同oldnode->size = size 相同的意思吗?即 struct mm_freenode_s的size成员表示的是sizeof(struct mm_freenode_s)+用户可用空间大小吗?显然是的。同样,可以从mm_addregion中就能看出来。
void mm_addregion(FAR struct mm_heap_s *heap, FAR void *heapstart,
size_t heapsize)
{
FAR struct mm_freenode_s *node;
......
node = (FAR struct mm_freenode_s *)
(heapbase + MM_SIZEOF_ALLOCNODE);
DEBUGASSERT((((uintptr_t)node + MM_SIZEOF_ALLOCNODE) % MM_ALIGN) == 0);
node->size = heapsize - 2 * MM_SIZEOF_ALLOCNODE;
......
}
mm_addregion对struct mm_allocnode_s的size成员的赋值有一定的特殊性,那么,此size表示的是那段空间呢?更一般的来说,此size表示的是sizeof(struct mm_allocnode_s)+用户实际使用空间大小,何以见得呢?我们可以通过函数mm_malloc就能看得出来。
FAR void *mm_malloc(FAR struct mm_heap_s *heap, size_t size)
{
FAR struct mm_freenode_s *node;
......
alignsize = MM_ALIGN_UP(size + MM_ALLOCNODE_OVERHEAD);
......
node->size = alignsize | (node->size & MM_MASK_BIT);
......
ret = (FAR void *)((FAR char *)node + MM_SIZEOF_ALLOCNODE);
......
}
564

被折叠的 条评论
为什么被折叠?



