c82f44c4f3
There are many small memory block in NuttX system, eg: struct tcb_s, struct inode, etc, and several disadvantages about them: 1.Their frequent allocate and free cause the system memory fragmentation. 2.Since each memory block has an overhead, the utilization of small memory blocks is relatively low, which will cause memory waste. So we can use mempool to alloc smallo block, to improve alloc speed and utilization, to reduce fragmentation. Signed-off-by: dongjiuzhu1 <dongjiuzhu1@xiaomi.com>
243 lines
7.3 KiB
C
243 lines
7.3 KiB
C
/****************************************************************************
|
|
* mm/mm_heap/mm_memalign.c
|
|
*
|
|
* Licensed to the Apache Software Foundation (ASF) under one or more
|
|
* contributor license agreements. See the NOTICE file distributed with
|
|
* this work for additional information regarding copyright ownership. The
|
|
* ASF licenses this file to you under the Apache License, Version 2.0 (the
|
|
* "License"); you may not use this file except in compliance with the
|
|
* License. You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
|
|
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
|
|
* License for the specific language governing permissions and limitations
|
|
* under the License.
|
|
*
|
|
****************************************************************************/
|
|
|
|
/****************************************************************************
|
|
* Included Files
|
|
****************************************************************************/
|
|
|
|
#include <nuttx/config.h>
|
|
|
|
#include <assert.h>
|
|
|
|
#include <nuttx/mm/mm.h>
|
|
|
|
#include "mm_heap/mm.h"
|
|
#include "kasan/kasan.h"
|
|
|
|
/****************************************************************************
|
|
* Public Functions
|
|
****************************************************************************/
|
|
|
|
/****************************************************************************
|
|
* Name: mm_memalign
|
|
*
|
|
* Description:
|
|
* memalign requests more than enough space from malloc, finds a region
|
|
* within that chunk that meets the alignment request and then frees any
|
|
* leading or trailing space.
|
|
*
|
|
* The alignment argument must be a power of two. 16-byte alignment is
|
|
* guaranteed by normal malloc calls.
|
|
*
|
|
****************************************************************************/
|
|
|
|
FAR void *mm_memalign(FAR struct mm_heap_s *heap, size_t alignment,
|
|
size_t size)
|
|
{
|
|
FAR struct mm_allocnode_s *node;
|
|
uintptr_t rawchunk;
|
|
uintptr_t alignedchunk;
|
|
size_t mask = alignment - 1;
|
|
size_t allocsize;
|
|
size_t newsize;
|
|
|
|
/* Make sure that alignment is less than half max size_t */
|
|
|
|
if (alignment >= (SIZE_MAX / 2))
|
|
{
|
|
return NULL;
|
|
}
|
|
|
|
/* Make sure that alignment is a power of 2 */
|
|
|
|
if ((alignment & -alignment) != alignment)
|
|
{
|
|
return NULL;
|
|
}
|
|
|
|
#if CONFIG_MM_HEAP_MEMPOOL_THRESHOLD != 0
|
|
node = mempool_multiple_memalign(&heap->mm_mpool, alignment, size);
|
|
if (node != NULL)
|
|
{
|
|
return node;
|
|
}
|
|
#endif
|
|
|
|
/* If this requested alinement's less than or equal to the natural
|
|
* alignment of malloc, then just let malloc do the work.
|
|
*/
|
|
|
|
if (alignment <= MM_MIN_CHUNK)
|
|
{
|
|
FAR void *ptr = mm_malloc(heap, size);
|
|
DEBUGASSERT(ptr == NULL || ((uintptr_t)ptr) % alignment == 0);
|
|
return ptr;
|
|
}
|
|
|
|
/* Adjust the size to account for (1) the size of the allocated node, (2)
|
|
* to make sure that it is an even multiple of our granule size, and to
|
|
* include the alignment amount.
|
|
*
|
|
* Notice that we increase the allocation size by twice the requested
|
|
* alignment. We do this so that there will be at least two valid
|
|
* alignment points within the allocated memory.
|
|
*
|
|
* NOTE: These are sizes given to malloc and not chunk sizes. They do
|
|
* not include SIZEOF_MM_ALLOCNODE.
|
|
*/
|
|
|
|
newsize = MM_ALIGN_UP(size); /* Make multiples of our granule size */
|
|
allocsize = newsize + 2 * alignment; /* Add double full alignment size */
|
|
|
|
if (newsize < size || allocsize < newsize)
|
|
{
|
|
/* Integer overflow */
|
|
|
|
return NULL;
|
|
}
|
|
|
|
/* Then malloc that size */
|
|
|
|
rawchunk = (uintptr_t)mm_malloc(heap, allocsize);
|
|
if (rawchunk == 0)
|
|
{
|
|
return NULL;
|
|
}
|
|
|
|
kasan_poison((FAR void *)rawchunk, mm_malloc_size((FAR void *)rawchunk));
|
|
|
|
/* We need to hold the MM mutex while we muck with the chunks and
|
|
* nodelist.
|
|
*/
|
|
|
|
DEBUGVERIFY(mm_lock(heap));
|
|
|
|
/* Get the node associated with the allocation and the next node after
|
|
* the allocation.
|
|
*/
|
|
|
|
node = (FAR struct mm_allocnode_s *)(rawchunk - SIZEOF_MM_ALLOCNODE);
|
|
|
|
/* Find the aligned subregion */
|
|
|
|
alignedchunk = (rawchunk + mask) & ~mask;
|
|
|
|
/* Check if there is free space at the beginning of the aligned chunk */
|
|
|
|
if (alignedchunk != rawchunk)
|
|
{
|
|
FAR struct mm_allocnode_s *newnode;
|
|
FAR struct mm_allocnode_s *next;
|
|
size_t precedingsize;
|
|
|
|
/* Get the node the next node after the allocation. */
|
|
|
|
next = (FAR struct mm_allocnode_s *)((FAR char *)node + node->size);
|
|
|
|
/* Make sure that there is space to convert the preceding
|
|
* mm_allocnode_s into an mm_freenode_s. I think that this should
|
|
* always be true
|
|
*/
|
|
|
|
DEBUGASSERT(alignedchunk >= rawchunk + 8);
|
|
|
|
newnode = (FAR struct mm_allocnode_s *)
|
|
(alignedchunk - SIZEOF_MM_ALLOCNODE);
|
|
|
|
/* Preceding size is full size of the new 'node,' including
|
|
* SIZEOF_MM_ALLOCNODE
|
|
*/
|
|
|
|
precedingsize = (uintptr_t)newnode - (uintptr_t)node;
|
|
|
|
/* If we were unlucky, then the alignedchunk can lie in such a position
|
|
* that precedingsize < SIZEOF_NODE_FREENODE. We can't let that happen
|
|
* because we are going to cast 'node' to struct mm_freenode_s below.
|
|
* This is why we allocated memory large enough to support two
|
|
* alignment points. In this case, we will simply use the second
|
|
* alignment point.
|
|
*/
|
|
|
|
if (precedingsize < SIZEOF_MM_FREENODE)
|
|
{
|
|
alignedchunk += alignment;
|
|
newnode = (FAR struct mm_allocnode_s *)
|
|
(alignedchunk - SIZEOF_MM_ALLOCNODE);
|
|
precedingsize = (uintptr_t)newnode - (uintptr_t)node;
|
|
}
|
|
|
|
/* Set up the size of the new node */
|
|
|
|
newnode->size = (uintptr_t)next - (uintptr_t)newnode;
|
|
newnode->preceding = precedingsize | MM_ALLOC_BIT;
|
|
|
|
/* Reduce the size of the original chunk and mark it not allocated, */
|
|
|
|
node->size = precedingsize;
|
|
node->preceding &= ~MM_MASK_BIT;
|
|
|
|
/* Fix the preceding size of the next node */
|
|
|
|
next->preceding = newnode->size | (next->preceding & MM_ALLOC_BIT);
|
|
|
|
/* Convert the newnode chunk size back into malloc-compatible size by
|
|
* subtracting the header size SIZEOF_MM_ALLOCNODE.
|
|
*/
|
|
|
|
allocsize = newnode->size - SIZEOF_MM_ALLOCNODE;
|
|
|
|
/* Add the original, newly freed node to the free nodelist */
|
|
|
|
mm_addfreechunk(heap, (FAR struct mm_freenode_s *)node);
|
|
|
|
/* Replace the original node with the newlay realloaced,
|
|
* aligned node
|
|
*/
|
|
|
|
node = newnode;
|
|
}
|
|
|
|
/* Check if there is free space at the end of the aligned chunk. Convert
|
|
* malloc-compatible chunk size to include SIZEOF_MM_ALLOCNODE as needed
|
|
* for mm_shrinkchunk.
|
|
*/
|
|
|
|
size = MM_ALIGN_UP(size + SIZEOF_MM_ALLOCNODE);
|
|
|
|
if (allocsize > size)
|
|
{
|
|
/* Shrink the chunk by that much -- remember, mm_shrinkchunk wants
|
|
* internal chunk sizes that include SIZEOF_MM_ALLOCNODE.
|
|
*/
|
|
|
|
mm_shrinkchunk(heap, node, size);
|
|
}
|
|
|
|
mm_unlock(heap);
|
|
|
|
MM_ADD_BACKTRACE(heap, node);
|
|
|
|
kasan_unpoison((FAR void *)alignedchunk,
|
|
mm_malloc_size((FAR void *)alignedchunk));
|
|
|
|
DEBUGASSERT(alignedchunk % alignment == 0);
|
|
return (FAR void *)alignedchunk;
|
|
}
|