WO2007109920A1

WO2007109920A1 - A method for constructing and using a memory pool

Info

Publication number: WO2007109920A1
Application number: PCT/CN2006/000515
Authority: WO
Inventors: Chuan Ye; Xinghua Li; Yuanqing Shi
Original assignee: Zte Corporation
Priority date: 2006-03-27
Filing date: 2006-03-27
Publication date: 2007-10-04

Abstract

A method for constructing and using a memory pool, divides the memory as several memory pools, the memory pools have several memory blocks with same size. Each memory pool has a management structure, each memory block is divided into a memory header and a data body. The memory header records the using information of the memory block. When the system receives a request for the memory, it chooses a proper memory pool based on the total size of the requested memory and the memory header. After the free memory block which can be assigned is chose, it updates the use-state of the memory head and the use-information of the memory block in the manage structure and the address information of the free memory block. Then it returns the data body pointer of the free memory block to the user and ends the application of the memory. When releasing the memory, it moves the pointer of the released memory block a memory header length, and points it to a corresponding memory header. Rewrites the use-state of the said memory header as free, and updates the use-state of the memory block and the address of the free memory block in the manage structure. The present invention can simplify the memory management, prevent the leak of the memory and the confusion of the memory.

Description

A memory pool construction and use method

Technical field

The present invention relates to memory pool construction and methods of use. Background technique

At present, the complexity of the software system of communication equipment is increasing, and the requirements for memory usage and management are also high. The current software system requires less memory copying, less memory loss and leakage, memory management and use to be more versatile, reduce software system development difficulty and development cost, and improve the efficiency and stability of the software system.

The Chinese patent publication CN03148892 proposes a memory management method in which a memory pool is created by an application, a memory pool is divided into a control domain for storing control information, a memory block storage area for placing data, and a usage state of each memory is indicated. A chain of tokens that can be indexed by a memory block. The application requests, returns, and modifies the information in the memory usage token chain from the memory pool based on the control information in the memory pool control domain. This patent has a simple organization of the memory pool, and the structure of the memory block is single, which is likely to cause performance problems when the application uses the memory block.

The Chinese patent with the publication number CN02151120 also proposes a method of memory management. The memory management method proposed by the patent includes the following steps:

a. Applying a block of memory as a memory pool in advance, dividing the memory block and initializing it, the applied memory block is taken from the memory pool header, and the released memory block is stored at the end of the memory pool;

b. setting a free area in the memory block for adding header information, and setting a message area after the free area of the memory block;

c. When it is necessary to increase the header information, move the pointer forward in the free area to increase the length of the header information, and then add the header information;

d. When reading the header information, move the pointer to the corresponding position in the free area to read the required information. Patent A does not mention the organization of the memory pool and how the application uses the memory block. It only describes the header information of the memory block and the construction of the data area. And when the method needs to increase the header information, the pointer is moved forward in the free area to increase the length of the header information, and then the header information is added. This invisibly reduces the available space in the data area, and the application may cause an exception such as memory out of bounds when using the memory block. Summary of the invention

The problem to be solved by the present invention is to propose a construction and use method of a memory pool to simplify memory management and avoid memory leaks and memory clutter.

In order to solve the above problems, the present invention provides a method of constructing and using a memory pool, including the following steps:

(a) Divide memory into several memory pools, each memory pool includes one or more memory blocks of the same size, each memory pool has a management structure, and the memory blocks in the memory pool are saved and used. Situation information and address information of the free memory block, each memory block is divided into a data body and a memory header for marking the usage of the memory block;

(b) After receiving the memory request, the memory management module uses the requested memory size plus the memory header size as the minimum memory block size to be allocated, and selects a memory pool that satisfies the requirements;

(c) judging whether there are any free memory blocks that can be allocated according to the memory block usage in the selected memory pool management structure, if so, selecting a free memory block allocation, marking the free memory block as "used", and Updating memory block usage information and address information of the free memory block in the management structure;

(d) returning the data body pointer of the allocated free memory block to the user, and the application ends; (e) the memory management module receives the memory release request, indicating the memory block pointer to the data body to be returned, and biasing the pointer After shifting the length of a memory header, find the corresponding memory header;

(0) The usage flag in the memory header is changed to idle, and the memory block usage information and the free memory block address information in the memory pool management structure to which the memory block belongs are updated, and the release ends.

Further, the foregoing method may further have the following features: the free memory block in each memory pool adopts a queue structure of a unidirectional linked list or a doubly linked list, and the memory header of the free memory block records the memory of the next free memory block in the queue. a head pointer, and a pointer of the free memory block queue header and the queue tail is retained in each memory pool management structure; in step (c), the free memory in the management structure is The free memory block pointed to by the block queue head pointer is allocated, and after the allocation, the free memory block queue head pointer is changed to the memory head of the next free memory block in the queue; when (f) the memory is released, the The free memory block queue tail pointer in the management structure is changed to point to the memory header of the newly added free memory block, and the last free memory block in the original queue is linked with the newly added free memory block.

Further, the foregoing method may further have the following features: The memory block size in each memory pool is 2 ⁿ bytes, and n is a positive integer.

Further, the above method may further have the following features: The management structure of each memory pool is set in a specified area of the memory.

Further, the foregoing method may further have the following features: in the step (b), selecting a memory in which the memory block is the smallest from a memory pool whose memory block size is larger than the minimum memory size to be allocated according to the minimum memory size to be allocated. If the pool is not selected, the application fails and ends.

Further, the foregoing method may further have the following features: the memory pool management structure further stores information about the number of reserved memory blocks, and the number of free memory blocks in the step (c) is greater than the number of reserved memory blocks. , it means there are free memory blocks that can be allocated. Otherwise, there are no free memory blocks available for allocation. .

Further, the foregoing method may further have the following features: the step (e) further checks the validity of the memory block pointer to be returned, and after the memory pointer is offset by the length of the memory header, if the pointer is in the If the memory block to be returned is within the memory pool range, it is legal. Otherwise, it is illegal, and the error message is returned.

Further, the foregoing method may further have the following features: In the process of applying for or releasing the memory, before the operation of the memory pool management structure or the memory block, the semaphore protection is added, and after the operation, the semaphore protection is released. .

Further, the foregoing method may further have the following features: The memory head is further configured to store information about a process or task that requests or returns a memory block in which the memory head is located.

Further, the above method may further have the following features: In the step (f), first determining whether the usage flag in the memory header indicates that the memory block is "used", and if so, changing it to "Idle" and proceed with subsequent operations, otherwise, an error message is returned, ending. . The method of the present invention details the organization of the memory pool and memory blocks and the steps by which the application uses the memory blocks. It simplifies the memory management requirements in the software system and can prevent memory leaks and memory clutter in the software system. The header information of the memory block and the message data area are separated from each other, and the addition or subtraction of the header information does not affect the memory block message data area. The method of the present invention can be used, for example, in mobile terminals, network side switching devices, and the like, particularly supporting subsystems. BRIEF abstract

FIG. 1 is a schematic structural diagram of a memory pool structure according to an embodiment of the present invention.

2 is a flow chart of a user applying for a memory block according to an embodiment of the present invention.

FIG. 3 is a flowchart of a user releasing a memory block according to an embodiment of the present invention. BEST MODE FOR CARRYING OUT THE INVENTION

The embodiment of the present invention applied to the support subsystem is now proposed, but the present invention is not limited thereto.

Divide the memory into several memory pools. For the structure of the memory pool, please refer to Figure 1. Each memory pool includes multiple memory blocks. The memory block UB size UBBlockSize in the same memory pool TJJBPool is the same. The size of the memory block in each memory pool increases from 16 bytes in multiples of 2, such as 16 bytes, 32 words. Section, 64 bytes, ..., 8192 bytes.

A memory pool management structure is set for each memory pool in the specified area of memory, which includes the following information:

♦ The number of memory blocks reserved by the memory pool (UBReserverCount), a certain number of memory blocks can be reserved to warn in advance when there is a certain number of free memory blocks to prevent system errors. This information is optional.

♦ The total number of memory blocks in the memory pool (UBBlockCount);

♦ Memory block size in the memory pool (UBBlockSize);

♦ Number of free memory blocks in the memory pool (UBFree);

♦ The physical address of the memory pool (UBBlockAddr), which points to the first memory block of the memory pool; ♦ Free memory block queue header pointer (UBFreeHead), which points to the memory header of the first free memory block in the free memory block list (refer to the first address of the memory header, the same below);

♦ Free memory block queue tail pointer (UBFreeTail), the memory header pointing to the last free memory block in the free memory block list (refer to the first address of the memory header, the same below);

♦ The number of memory blocks that have been successfully allocated (UBBlockAllocCount); and the allocated memory failure count (UBBlockErrCount).

Each memory pool contains a number of memory blocks in a contiguous memory space, each memory block consisting of a memory header (UBHead) and an adjacent data body (UserDataPtr), one memory block data area immediately following the next memory block Memory header. The data body is a memory area that can be used by the user for application. The user program obtains the allocated data body pointer (pointing to its first address) by calling the interface of the application memory block (ie, the API function provided by the memory management module), and can be in the memory area. In the middle of the operation.

The memory header (UBHead) of all memory blocks is the same size and contains the following information:

♦ The process information (GetUser) for requesting the memory block and the process information (RetUser) for returning the memory block, which may be the identification number of the task or the process, for monitoring the memory usage during debugging or runtime;

♦ The index value of the memory pool where the memory block is located can be quickly located in the memory pool according to the index value when the memory block is returned, avoiding traversing the memory pool from beginning to end in the memory management module, thereby improving the operation efficiency;

♦ The memory block uses the flag (MEM_BLOCKJJSED), which is used or idle; ..

♦ The next memory head pointer. For a free memory block, the pointer should point to the memory head pointer of the next free memory block to form a singly linked list of free memory blocks. In another embodiment, a singly linked list of memory blocks that are already allocated by the application may be added.

It can be seen that the free memory blocks in the memory pool are organized by means of a singly linked list. When the initial unallocated, all the memory blocks are idle, and the pointers in the memory header of each memory block point downwards. An adjacent memory block, when allocating memory, is allocated from the first free memory block in the free memory block list, and the free memory block queue header pointer in the memory pool management structure is modified to point to the linked list. The memory head of the next free memory block. When the memory is released, it will be released The memory block is added to the end of the free memory block, and the free memory block queue tail pointer in the memory pool management structure and the memory head pointer in the original free memory block memory header in the linked list are modified to point to the newly added memory block. The memory header of a free memory block.

In another embodiment, two memory head pointers may also be set in the memory header of each memory block, so that the two pointers in the free memory block point to the memory block and the next memory block memory of the previous memory block respectively. The header, thus forming a doubly linked list of free memory blocks, is more flexible when allocating free memory blocks.

When the user applies for a memory block, the memory management function provided by the operating system platform is not used. The memory management module implements the following processing flow. As shown in FIG. 2, the following steps are included:

Step 100, the user (referring to the application) makes a memory application request, and the input parameter indicates the size of the memory block to be used;

Step 110: After receiving the request for the memory, the memory management module determines the legality of the input parameter, and adds the requested memory size and the memory header size as the minimum memory size to be allocated, and determines whether the memory size is Greater than the size of the existing maximum memory block (8192 bytes), if yes, perform the next step, otherwise the allocation fails and ends;

Step 120: Select, according to the minimum memory size to be allocated, a memory pool in which the memory block is the smallest from a memory pool whose memory block size is larger than the minimum memory size to be allocated;

Step 130, adding semaphore protection to prevent other applications from simultaneously operating the memory; Step 140, determining whether the number of free memory blocks in the selected memory pool management structure is too larger than the number of memory blocks reserved in the memory pool, if , the next step, otherwise, the allocation fails, the end; the reason for reserving the memory block is: a, for some operating systems without memory boundary checking, such as vxworks, threadx, etc., once the application uses the memory block out of bounds, you can write We reserve the memory block to prevent the system from crashing. b. Although the reserved memory block is not used, the reserved value can be used as the threshold used by the memory pool. Once the threshold is reached, the system can alert and not The system is running abnormally because the memory blocks in the memory pool are exhausted.

Step 150: Find a free memory block queue head pointer from the management structure of the memory pool, and locate a memory head start position of the first free memory block in the free memory block linked list. Step 160: Fill in the information in the memory header, including the process (or task) information (GetUser) of applying for the memory block, and set the memory block usage flag (MEM_BLOCK_USED) to be used; Step 170, update the memory. The information in the pool management structure updates the free memory block queue head pointer in the management structure to the next memory head pointer in the memory block memory header, and decrements the number of free memory blocks, and the number of successfully allocated memory blocks is incremented. ; - Step 180, release semaphore protection;

Step 190: Return the allocated memory block address, which should be the data body pointer of the allocated free memory block, and end.

The processing flow when the user releases the memory is shown in Figure 3, including the following steps:

Step 200, the user proposes a memory release request, indicating a memory block pointer (pointing to the data body) to be returned, and the pointer points to the memory block data body;

Step 210: The memory management module checks the validity of the memory pointer to be returned, offsets the input memory pointer by a length of the memory header, and determines whether the memory head pointer is within the memory pool of the memory block to be returned. If yes, go to the next step, otherwise, return an error message and end;

Step 220: Find a corresponding memory head according to the offset memory head pointer.

Step 230, determining whether the memory block usage flag indicates that the memory block is available, and if yes, returning an error message, ending, otherwise, performing the next step;

, step 240, calling the system semaphore call (VOS_Sema-P) to enter the semaphore protection of the memory management module;

Step 250: Set a memory block usage flag in the memory head pointer to be idle, and fill in the memory head pointer to fill in the process (or task) information of the memory block;

Step 260: Update the free memory block queue tail pointer in the corresponding memory pool management structure to the memory head pointer, add one to the number of free memory blocks in the management structure, and make the last free memory block memory in the original free memory block linked list. The next memory head pointer in the header points to the memory header newly added to the free memory block, that is, the last free memory block in the original queue is linked with the newly added free memory block; Step 270, calling the system semaphore call to release the semaphore protection, and ending.

On the basis of the foregoing embodiments, various transformations are also possible. For example, for the management of the S-free memory block, in addition to the unidirectional linked list and the doubly linked list mentioned above, other methods may be used, for example, free memory may be used. The FIFO queue formed by the block pointer is managed, and the FIFO queue can be placed in the management structure of the memory pool.

Industrial applicability

The method of the present invention describes the organization structure of the memory pool and the memory block and the method of using the memory block by the application, which simplifies the memory management requirements in the software system, and can prevent memory leakage and memory confusion in the software system, and can be used, for example. In the supporting subsystem of the mobile terminal, the network side switching device, and the like.

Claims

Claim

1. A method of constructing and using a memory pool, comprising the following steps:

(c) judging whether there is any free memory block that can be allocated according to the memory block usage in the selected memory pool management structure, if so, selecting a free memory block allocation, marking the free memory block as "used", and Updating memory block usage information and address information of the free memory block in the management structure;

(d) returning the data body pointer of the allocated free memory block to the user, and the application ends;

(e) The memory management module receives the memory release request, indicates the memory block pointer to the data body to be returned, and offsets the pointer by a length of the memory header to find the corresponding memory header;

(f) Change the usage flag in the memory header to idle, and update the memory block usage information and the free memory block address information in the memory pool management structure to which the memory block belongs, and the release ends.

2. The method according to claim 1, wherein the free memory block in each memory pool adopts a queue structure of a unidirectional linked list or a doubly linked list, and the memory header of the free memory block records the next idle in the queue. The memory head pointer of the memory block retains the pointer of the free memory block queue header and the queue tail in each memory pool management structure; in step (c), the pointer of the free memory block queue head pointer in the management structure is pointed The free memory block is allocated, and after the allocation, the free memory block queue head pointer is changed to the memory head of the next free memory block in the queue; when the memory is released in step (f), the free memory in the management structure is also The block queue tail pointer is changed to point to the memory header of the newly added free memory block, and the last free memory block in the original queue is linked with the newly added free memory block.

3. The method according to claim 1, wherein the memory block size in each memory pool is 2 ⁿ bytes, and n is a positive integer.

4. The method according to claim 1, wherein the management structures of the respective memory pools are collectively set in a specified area of the memory.

5. The method according to claim 1, wherein in the step (b), the minimum memory size to be allocated is selected from a memory pool whose memory block size is larger than a minimum memory size to be allocated. The memory pool with the smallest memory block, if not selected, the application fails and ends.

The method of claim 1, wherein the memory pool management structure further stores information about the number of reserved memory blocks, and the number of free memory blocks in the step (c) is greater than the reservation. The number of memory blocks indicates that there are free memory blocks that can be allocated. Otherwise, there are no free memory blocks available for allocation.

The method according to claim 1, wherein in the step (e), the memory block pointer to be returned is checked for validity, and the memory pointer is offset by a length of the memory header, such as The pointer is legal in the range of the memory pool where the memory block to be returned is located. Otherwise, it is illegal, and the error message is returned.

8. The method according to claim 1, wherein in the process of applying for or releasing the memory, before the operation of the memory pool management structure or the memory block, the semaphore protection is added, after the operation, Release semaphore protection. '

9. The method of claim 1, wherein the memory head is further configured to store information of a process or task requesting or returning a memory block in which the memory head is located.

10. The method according to claim 1, wherein in the step (f), determining whether the usage flag in the memory header indicates that the memory block is "used", and if so, Change to "Idle" and proceed with the subsequent operation, otherwise, return an error message and end.