JP2001101080A - Method for managing dead buffer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain the detection of any illegal memory access to a region beyond the range of captured buffers or released buffers or the like even in a normal operation by reducing the increase of processing overhead. SOLUTION: Continuous buffer regions constituted of plural buffers are divided into operating buffers to be preferentially used and waiting buffers to be used when the buffers to be preferentially used are all used so that the operating buffers and the waiting buffers can alternately appear as address. Then, the operating buffer chain is combined with the waiting buffer chain, and memory protection is set in each buffer. When released from the buffer chain accompanying a request for capturing the buffers, the memory protection of the buffers to be captured is released, and when connected accompanying a request for releasing the buffers, memory protection is set in the buffers to be released.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vacant buffer management system in which vacant buffers in a memory are linked and managed in a chain in order to dynamically allocate / release memory in a host computer of an online information processing system. Methods include, among other things, detecting illegal memory access to areas outside the buffer size range of captured buffers, detecting illegal memory access to released buffers, and detecting double release of released buffers. The present invention relates to an easily available empty buffer management method.

[0002]

2. Description of the Related Art Conventionally, in a computer on which a user program including a dynamic memory allocation request, a request for accessing and releasing a allocated memory, and the like is operated,
As a method of automatically detecting access to a dynamically allocated memory due to a failure of a user program, for example, there is a method described in JP-A-7-319735 (Japanese Patent No. 275777). .

In this method, a PMMU (Page Memory) is used.
Using a Management Unit) and an interrupt handler, it is possible to detect an illegal access limited to only access to a dynamically allocated memory as follows. That is,
The page management table is configured in advance so that the virtual address accessed by the user program cannot be accessed. If there is a memory access during execution of the user program, the PMMU knows that access is not possible in the page management table. Then, the PMMU interrupts the CPU and activates the interrupt handler. In the interrupt processing of the interrupt handler, the validity of the address to be accessed is checked, converted to an accessible virtual address, and given to the PMMU again. The PMMU translates this virtual address into a physical address and returns control to the user program.

[0004]

In the above conventional method,
Each time a write or read access to memory occurs, the interrupt handler can check the validity of the access, so memory accesses beyond the range of dynamically allocated memory are detected and dynamically allocated. This makes it possible to detect the execution of a read instruction before the execution of a write instruction to the memory and to detect the access to the released dynamically allocated memory.

However, every time there is a read or write access to the memory, an interrupt handler is activated to check for unauthorized access.
Since the processing overhead is very large, it is effective for application in fields such as software development tools, defect investigation tools, and quality inspection / improvement tools. There is a need to.

Further, in an online information processing system or the like, the above-mentioned conventional method will be applied to a plurality of buffers that are dynamically allocated and used in order for a host computer to temporarily store transmission / reception data under communication control. Then, aside from during the debugging period, application in the operation system is very difficult in terms of processing capacity.

An object of the present invention is to detect an illegal memory access to an area exceeding the buffer size range of a captured buffer or an illegal memory access to a released buffer in a host computer or the like of an online information processing system. Free buffer management that can be performed while minimizing the processing overhead and detecting the double release of released buffers, etc., and can be applied not only during the debugging period but also to the operation system. It is to provide a method.

[0008]

An empty buffer is linked as a chain and managed as a buffer chain. The present invention provides an operation buffer that preferentially uses a continuous buffer area composed of a plurality of buffers having the same or an integral multiple of the buffer size as a memory protection setting / cancellation unit, It is divided into a standby buffer used when it is used, and further divided so that the operation buffer and the standby buffer appear alternately in address in the continuous buffer area, and the operation buffer chain and the standby Two buffer chains of a buffer chain are assembled, and while connected to one of the buffer chains, memory protection is set for the corresponding buffer. Then, the memory protection of the buffer to be captured is released when removing from one of the buffer chains following the buffer capture request, and the memory protection of the buffer to be released is released when connecting to one of the buffer chains following the buffer release request. Set.

[0009] As a result, a standby buffer for which memory protection is set exists before and after the operation buffer, and it is possible to detect an illegal memory write to an area other than the captured buffer. Further, since memory protection is set for the released buffer, it is possible to detect illegal memory writing to the released buffer.

Further, according to the present invention, at the time of connection to a buffer chain associated with a buffer release request, the contents of an arbitrary memory in the buffer to be released are read, and the read contents are written again to the same memory, and then the memory protection is performed. Set. This makes it possible to detect double release of the buffer at the time of rewriting to the memory.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail based on embodiments shown in the drawings. FIG. 1 shows an embodiment of the present invention, which is a diagram showing an overall system configuration when applied to a host computer 101 connected to a terminal device 102 via a communication line 103. Although a large number of terminal devices are generally connected to the host computer 101, for convenience, FIG. 1 shows only one terminal device.

In the figure, a host computer 101
Is composed of a communication control unit 104, a buffer capture unit 105, a buffer release unit 106, processing units of a buffer management information initial setting unit 107, and a main memory 108. The main memory 108 includes a buffer area 112 in which a plurality of buffers are secured, a buffer management table 111 for managing a plurality of buffers secured in the buffer area 112, and a plurality of buffers for operating free buffer chains and standby free buffer chains. It has an operation free buffer chain management table 109 and a standby free buffer chain management table 110, which are roughly managed.

The communication control unit 104 controls communication with the terminal device 102. When the communication control unit 104 transmits and receives data to and from the terminal device 102, the communication control unit 104 temporarily stores transmission / reception data. Buffer area 11 of storage 108
2 is used. The communication control unit 104 issues a buffer capture request to the buffer capture unit 105 when a new buffer is used, and issues a buffer release request to the buffer release unit 106 when releasing a buffer in use.

The buffer capturing unit 105 includes a communication control unit 10
4 receives the buffer capture request from the active / standby free buffer chain management tables 109 and 110 and the buffer management table 111, and executes a buffer capture process from the active / standby free buffer chain. At this time, the memory protection of the captured buffer is released. Upon receiving the buffer release request from the communication control unit 104, the buffer release unit 106 executes a buffer release process on the operation / standby free buffer chain management table corresponding to the buffer chain to which the released buffer is connected. At this time, the memory protection of the buffer to be released is set.

The buffer management information initialization section 107 initializes various types of buffer management information. Although the specific operation will be described later, the buffer management information initial setting unit 107
A continuous buffer area 112 composed of a plurality of buffers having the same or an integral multiple of the unit size of the memory protection setting / cancellation unit is divided into an operation buffer and a standby buffer. Among them, the operation buffer and the standby buffer are divided so that they appear alternately in terms of address, and two buffer chains, an operation buffer chain and a standby buffer chain, are assembled, and the operation / standby empty buffer chain management tables 109 and 110 are formed. Then, the contents of the buffer management table 111 are initialized. Further, at the time of this initial setting, memory protection is set for each buffer connected to the operation standby buffer chain.

FIG. 2 is an explanatory diagram showing a configuration example of the operation free buffer chain management table 109. The operation free buffer chain management table 109 includes a head buffer management table pointer (F_B_MNG_P) 201 indicating buffer information connected to the head of the operation free buffer chain.
And a last buffer management table pointer (L_B_MNG_P) 2 indicating buffer information connected at the end of the chain.
02.

FIG. 3 is an explanatory diagram showing an example of the configuration of the standby free buffer chain management table 110. The standby free buffer chain management table 110 has a head buffer management table pointer (F_B_MNG_P) 301 indicating buffer information connected to the head of the standby free buffer chain.
And a last buffer management table pointer (L_B_MNG_P) 3 indicating buffer information connected to the end of the chain.
02.

FIG. 4 is an explanatory diagram showing an example of the configuration of the buffer management table 111. Here, the buffer area 112 to be managed has a buffer size of 4 as shown in FIG.
It is composed of seven kilobyte buffers, and the head address of each buffer is H'0000 in buffer (0) 500.
Address, buffer (1) 501 is address H'1000, buffer (2)
502 is H'2000, buffer (3) 503 is H'3000, buffer (4) 504 is H'4000, and buffer (5) 5
It is assumed that 05 is H'5000 and the buffer (6) 506 is H'6000.

In FIG. 4, the buffer management table 11
1 includes tables 400 to 406, which correspond to the buffers 500 to 506 in FIG. 5, respectively. In particular,
Buffer management table 0 for managing buffer (0) 500
The (B_MNB0) 400 includes a next link table pointer (N_LNK_P) 407 for forming an operation free buffer chain or a standby free buffer chain, and the buffer (0) 5.
00 buffer address (BUF_
P) 408. Here, H'0000 is set in the buffer pointer 408 as a specific start address. Buffer (1) 501 to Buffer (6) 506
The buffer management table 1 (B_MNB1) 401 to the buffer management table 6 (B_MNB6) 406 have the same configuration.

The buffer from the buffer (0) 500 in FIG.
(6) In step 506, while being connected to either the operation / standby buffer chain, memory protection is set for the buffer. Depending on the processor architecture, there are various types of memory protection functions, such as a key system, a ring system, and an access right system, and the scope of protection depends on the processor architecture and virtual memory system, and is in various forms such as page units and segment units. Exists. The present invention does not depend on the method of realizing the memory protection function, but depends on the memory protection range. For example, when the memory protection range is in units of 1 kilobyte, the buffer size is an integral multiple of 1 kilobyte. (1 kilobyte, 2
If the memory protection range is in units of 2 kilobytes, the buffer size must be an integral multiple of 2 kilobytes (2 kilobytes, 4 kilobytes,...). The specific embodiment of FIG. 5 is an embodiment in which the memory protection range is realized by a processor having an architecture in units of 4 kilobytes, and when the memory protection unit is 8 kilobytes in the processor architecture, the second embodiment of the present invention is used. It is necessary to secure twice the buffer area 112.

Hereinafter, the buffer management information initial setting unit 10
7. Operations of the buffer capturing unit 105 and the buffer releasing unit 106 will be described. It is assumed that the buffer area 112 has the configuration shown in FIG.

FIG. 6 shows a buffer management information initial setting unit 10.
7 shows a processing flow of buffer management information initialization processing executed in Step S7. This buffer management information initial setting process
For example, it is executed when the system is started.

In the buffer management information initial setting process, first, the buffer area 112 having an area size of 7 kilobytes is captured (step 601).
2 is set with memory protection (step 60).
2). Next, the captured buffer area 112 is divided into seven buffers each having a buffer size of 1 kilobyte, that is, the buffer (0) 500 and the buffer (1) 5 shown in FIG.
01, buffer (2) 502, buffer (3) 503, buffer (4) 504, buffer (5) 505, buffer (6) 5
06, and as shown in FIG. 4, each buffer pointer (BUF_P) 408, 410, 412 of each of the management tables 400 to 406 of the buffer management table 111.
The buffers 500 to 506 are stored in 146, 148, and 420.
Is set (step 603). next,
In the present embodiment, the odd-numbered buffer, that is, the buffer
(1) 501, buffer (3) 503, buffer (5) 505
Is connected to the operation free buffer chain management table 109 as the operation buffer (step 604), and the even-numbered buffer, that is, the buffer (0) 500, the buffer
(2) 502, buffer (4) 504, buffer (6) 506
Is connected to the standby free buffer chain management table 110 as a standby buffer (step 605), and the buffer management information initial setting process ends.

FIG. 7 shows a buffer management information initial setting unit 10.
7 immediately after executing the buffer management information initial setting process,
Buffer management table (0) 400, buffer management table (1) 401, buffer management table corresponding to each buffer 500 to 506 in the operation empty buffer chain management table 109, the standby empty buffer chain management table 110, and the buffer management table 111. Table (2) 40
2. Buffer management table (3) 403, buffer management table (4) 404, buffer management table (5) 405,
It shows a specific configuration of the buffer management table (6) 406. At this point, memory protection is set for all the buffers 500 to 506 in the buffer area 112.

FIG. 8 shows a processing flow of the buffer capturing process executed by the buffer capturing section 105. This buffer capture process is executed each time a buffer capture request is received from the communication control unit 104.

In the buffer capturing process, first, a head buffer management table pointer (F_B_MNG_P) 201 is extracted from the operation free buffer chain management table 109 (step 801), and it is determined whether the extracted pointer is NULLP (step 802). NULL in the judgment of step 802
If it is not P, that is, if it is determined that there is a free buffer in the working free buffer chain, the first buffer is removed from the working free buffer chain management table 109 (step 803), and the memory protection of the captured buffer is released (step 803). 804) Assuming that the buffer capture result = success, the pointer of the captured buffer is returned to the capture request source (step 805). That is, in this embodiment, the buffer is returned to the communication control unit 104, and the buffer capture process is terminated. If it is determined in step 802 that there is no free buffer in the NULLP, that is, the operation free buffer chain, the leading buffer management table pointer (F_B_MNG_P) 301 is extracted from the standby free buffer chain management table 110 (step 806). It is determined whether the pointer that has been set is NULLP (step 807). In the judgment of step 807
If it is not NULLP, that is, if it is determined that there is an empty buffer in the standby empty buffer chain, the first buffer is removed from the standby empty buffer chain management table 110 (step 808), and the processing of steps 804 and 805 is executed. If it is determined in step 807 that there is no empty buffer in the NULLP, that is, the standby empty buffer chain, the buffer is returned to the capture request source as the buffer capture result = failure (step 809), and the buffer capture process ends.

FIG. 9 shows the initial state of FIG. 7, in which the buffer capture processing of FIG. 8 is executed, and after the buffer (1) 501 connected to the head of the active empty buffer chain is captured, the active empty buffer chain Buffer management table (0) 400 buffer management table (2) corresponding to each of the buffers 500, 502 to 506 except the buffer (1) 501 in the management table 109, the standby free buffer chain management table 110, and the buffer management table 111 4
02, buffer management table (3) 403, buffer management table (4) 404, buffer management table (5)
405, a specific configuration of the buffer management table (6) 406.

Note that in FIG. 9, only the area of the buffer (1) 501 in the buffer area 112 (operating buffer area) has been released from memory protection. There is an area 502 (standby buffer area). Therefore, when a memory access such as an illegal memory write is performed on an area exceeding the buffer size range of the captured buffer (1) 501, that is, an area of the buffers 500 and 502, a memory protection violation occurs. The illegal memory access can be easily detected.

In the buffer capturing process of FIG. 8, when it is determined in step 802 that there is no empty buffer in the active empty buffer chain, the empty buffer buffer is captured from the standby empty buffer chain. 1) 501, buffer (3) 503, buffer (5) 505, followed by buffer (0) 500 and buffer (2) 5 incorporated in the standby free buffer chain.
02, when the buffer (4) 504 and the buffer (6) 506 are used, the memory protection of the continuous buffer area is released while the buffer (4) 504 and the buffer (6) 506 are used, and the buffer size exceeds the buffer size range of the captured buffer. Unable to detect illegal memory access to the area. The present embodiment allows such a situation to be extremely rare, but this can be avoided as follows.
That is, if it is determined in step 802 in FIG. 8 that there is no empty buffer in the active empty buffer chain, the process of step 809 is executed instead of capturing an empty buffer from the standby empty buffer chain, and the = Return the failure to the supplementary request source. As a result, the buffer 2502, the buffer 4504, and the buffer 6506 incorporated in the standby empty buffer chain
Is always protected, and the ability to detect illegal access of the buffer is improved as compared with the case where the processing flow described in FIG. 8 is taken.

Also, as an input parameter of the buffer capture processing, the buffer size required by the buffer capture request source is notified, and an address obtained by subtracting the capture request buffer size from the last address instead of the head address of the captured buffer is used. The processing may be returned to the capture request source as the start address of the capture buffer. In this case, if the capture requester incorrectly uses the buffer in a range larger than the buffer size that the capture requester considers necessary, a memory protection violation occurs because the captured buffer is protected by memory as a standby buffer. The possibility of detection increases.

FIG. 10 shows a processing flow of the buffer release processing executed by the buffer release unit 106. This buffer release processing is executed every time a buffer release request is received from the communication control unit 104.

In the buffer release processing, first, the contents of the head memory of the buffer requested to be released are read (step 1001), and the read value is written to the head memory of the buffer requested to be released (step 1001). 1002). Next, the memory protection of the buffer for which the release request has been made is set (step 1003). Next, a buffer chain to be connected to the buffer requested to be released is determined (step 1004), and in the case of an operation free buffer chain, the operation free buffer chain management table 109 is used.
Is connected at the end of the request (step 1005).
At the end of the standby free buffer chain management table 110, a buffer that has been requested to be released is connected (step 1).
006), the buffer release processing ends.

In FIG. 10, the purpose of reading and writing the head memory of the buffer requested to be released in steps 1001 and 1002 is to detect double release of the buffer. That is, when the buffer has already been released, the memory protection has been set as shown in step 1003. Therefore, when a further release request is made for the already released buffer,
In the process of writing to the memory in step 1002, a memory protection violation occurs, and it is possible to easily detect a double release bug in the buffer. Step 100
In 1,1002, the read / write is executed for the head memory of the buffer for which the release request has been made, but it is not particularly necessary to be the head memory.

When a memory protection violation is detected in step 1002 in FIG. 10, for example, a process such as a partial initialization of a system or a partial initialization of a related process is performed in an interrupt processing routine by a conventional function, and a double buffer is executed. In addition to guarding the release, it is possible to localize the scope of the bug, such as by reducing the effect on the system of a bug that keeps using the released buffer for a while.

FIG. 11 shows the operation free buffer chain after the buffer release processing of FIG. 10 is executed in the state of FIG. 9 and the buffer (1) 501 captured by the buffer capture processing of FIG. 8 is released. The buffer management table (0) 401, the buffer management table (1) 401, and the buffer management table (2) corresponding to each of the buffers 500 to 506 in the management table 109, the standby free buffer chain management table 110, and the buffer management table 111. 402, buffer management table (3) 403, buffer management table
(4) 404, a buffer management table (5) 405, and a buffer management table (6) 406.

As shown in FIG. 11, the released buffer
Since (1) 501 is set with memory protection, if a memory access such as an illegal memory write to the released buffer (1) 501 is performed, a memory protection violation occurs and the illegal memory access is performed. Can be easily detected. Also, when a further release request is made to the released buffer (1) 501, the above-mentioned step 100 is performed.
Execution of 2 makes it possible to detect double release. If double release of the buffer occurs, the empty buffer chain may be destroyed. Therefore, it is necessary to prevent double release of the buffer as much as possible.

The specific embodiment described above can be applied to either a system having only real storage or a system having virtual storage in addition to real storage. As described in the embodiment, for a system having, both the operation buffer and the standby buffer are generated in the logical space (virtual storage), but only the operation buffer is generated in the physical space (real storage). The same effect can be obtained as in the case. In this case, it is needless to say that the capture from the standby empty buffer chain is not performed when the operation empty buffer runs out. Further, the standby empty buffer chain itself is unnecessary.

[0038]

As described above, according to the present invention, in the free buffer management for managing the free buffers by linking them in a chain, an operation in which a continuous buffer area composed of a plurality of buffers is preferentially used. Buffers and
The operation buffer is divided into a standby buffer to be used when all of the operation buffers are used, and the operation buffer and the standby buffer are divided so that they appear alternately in address,
Combine two buffer chains, the active buffer chain and the standby buffer chain. While connected to one of the buffer chains, set the memory protection for the corresponding buffer and set the buffer protection from either buffer chain accompanying the buffer capture request. At the time of removal, the memory protection of the buffer to be captured is released, and at the time of connection to one of the buffer chains following a buffer release request, the memory protection of the buffer to be released is set. As a result, before and after the operation buffer, there is a standby buffer with memory protection set, and it is possible to detect a memory access such as an illegal memory write to an area other than the captured buffer, and Since memory protection is set for the released buffer, it is possible to detect illegal memory write to the released buffer.

Further, according to the present invention, at the time of connection to a buffer chain in response to a buffer release request, the contents of an arbitrary memory in the buffer to be released are read, and the read value is written again to the same memory. By setting, double release of the buffer can be detected.

Furthermore, in the buffer capture request, when all the operation buffers are in use, the standby buffer is not captured, and a buffer capture failure is returned to the capture request source. Since the buffer being used is always protected by memory, the ability to detect unauthorized access is improved.

[Brief description of the drawings]

FIG. 1 is a diagram showing one embodiment of a system to which the present invention is applied.

FIG. 2 is a diagram showing a configuration example of an operation free buffer chain management table in FIG. 1;

FIG. 3 is a diagram showing a configuration example of a standby free buffer chain management table in FIG. 1;

FIG. 4 is a diagram showing a specific configuration example of a buffer management table in FIG. 1;

FIG. 5 is a diagram showing a specific configuration example of a buffer area in FIG. 1;

FIG. 6 is a diagram showing a processing flow of a buffer management information initial setting unit in FIG. 1;

FIG. 7 is a diagram illustrating a specific configuration example of each management table and a buffer area after a buffer management information initial setting process is executed.

FIG. 8 is a diagram showing a processing flow of a buffer capturing unit in FIG. 1;

FIG. 9 is a diagram illustrating a specific configuration example of each management table and a buffer area after a buffer capturing process is executed.

FIG. 10 is a diagram showing a processing flow of a buffer release unit in FIG. 1;

FIG. 11 is a diagram illustrating a specific configuration example of each management table and a buffer area after a buffer release process is performed.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 101 Host computer 102 Terminal device 104 Communication control unit 105 Buffer capture unit 106 Buffer release unit 107 Buffer management information initial setting unit 109 Operation free buffer chain management table 110 Standby free buffer chain management table 111 Buffer management table 112 Buffer area 400 to 406 Buffer Management table 500-506 buffer

Claims (3)

[Claims] 1. An empty buffer management method for managing empty buffers by linking them in a chain, comprising: an operation buffer that preferentially uses a continuous buffer area composed of a plurality of buffers; The buffer buffer is divided into a standby buffer used when all the buffers are used, and the active buffer and the standby buffer are divided so that they alternately appear in address. And set the memory protection to the buffer connected to either the active buffer chain or the standby buffer chain, and release the memory protection of the buffer to be captured when removing from one of the buffer chains due to the buffer capture request The buffer release request to either of the buffer chains. During Nagikomi, it sets the memory protection buffer to release the free buffer management method characterized by. 2. The free buffer management method according to claim 1, wherein at the time of connection to a buffer chain in response to a buffer release request, the contents of an arbitrary memory in the buffer to be released are read and the contents are written back to the same memory. Then, a free buffer management method characterized by setting memory protection. 3. The free buffer management method according to claim 1, wherein when all of the operation buffers are in use, a buffer capture failure is returned to the capture request source without capturing the standby buffer. Free buffer management method.