JP3726701B2 - Kernel space demand paging swap-out method and method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a kernel space demand paging swap-out method and method, and more particularly to a kernel space demand paging swap-out method and method in a virtual memory operating system in which a kernel is generated as a single load module.
[0002]
[Prior art]
At present, the general kernels that exist in the world are focused on the structural aspect. The traditional monolithic kernel generated as a single load module and the microkernel realized by a set of functional unit threads are divided into two. Can be classified into types.
[0003]
For example, Japanese Patent Application Laid-Open No. 8-16412 discloses a microkernel system including a microkernel, a personality neutral service, and a personality server. However, in the microkernel, individual execution units such as the memory management function, process management function, execution management function, file management function, and driver function are loosely coupled together to realize the entire kernel function. Rather than deploying on the physical memory and occupying the CPU (Central Processing Unit), only necessary functional units are deployed in the physical memory, and by securing the CPU resources individually, memory resources and CPU resources are effectively and smoothly Can be used. On the other hand, the interface between the functional units is abstracted and loosely coupled, so there is a problem that overhead occurs in mutual call of functions and the performance of the entire kernel function is reduced.
[0004]
On the other hand, in the conventional monolithic kernel, since all functions are combined as a single load module, the interface between functions is optimized, and the overhead of mutual call of functions can be minimized. On the other hand, all the functions of the kernel (usually only a small part of the whole are required) are deployed on physical memory, consuming more memory resources than necessary, and once entering kernel mode, CPU resources are monopolized until the user mode is exited across a plurality of independently executable kernel functions, and there is a problem that resources cannot be used effectively. In particular, in the monolithic kernel, it was assumed that the entire kernel space was expanded on physical memory. However, the kernel functions actually required for system operation are limited, and only a small part of the entire kernel. There are many cases. For example, the startup / shutdown sequence is executed at most once, and many drivers are never executed unless the corresponding device is implemented. Moreover, even if it is mounted, it is not always executed. In addition, abnormal processing occupying most of the kernel text is rarely executed. Therefore, in order to effectively use the physical memory on which the kernel is mounted, a mechanism for effectively using these physical memories that are not normally used is required.
[0005]
In order to solve this problem, for example, Japanese Laid-Open Patent Publication No. 2-178747 refers to a database related to swapping when there is a shortage of free storage area in physical memory even if the kernel code cannot be swapped out. Thus, a memory management method for a virtual storage computing system is disclosed in which physical memory can be released according to a reference result.
[0006]
[Problems to be solved by the invention]
However, in the conventional technology, there is a problem that the use efficiency of the physical memory is not so much improved although the division by the functional unit is limited and the dedicated processing is required.
[0007]
An object of the present invention is to apply demand paging and swap-out commonly found in ordinary operating systems to a kernel space in a monolithic kernel with a low overhead at the time of mutual call between functions, and to effectively use memory resources. It is to provide a demand paging swap-out method and method.
[0008]
Further, according to another object of the present invention, a method and a method for demand paging / swap-out of a kernel space that can further increase the use efficiency of memory resources by introducing a method for determining a swap-out priority from a swap-out time. Is to provide.
[0009]
[Means for Solving the Problems]
The demand paging swap-out method of the kernel space according to the present invention is a swapping-in method when a page fault occurs in the kernel space during system operation in a virtual memory computing system employing a monolithic kernel generated as a single load module. Kernel demand paging unit that swaps in the target page from the swap device to the physical memory, and the physical memory by selecting the page with the highest swap-out priority when the user process fails to secure the page of the physical memory during system operation. And a page where the kernel demand paging unit exists by calling the kernel swap out unit immediately after completion of the kernel startup sequence. Except, wherein the pages of the remaining kernel space from the physical memory and a said start-up time of the kernel swap-out unit to be swapped out to the swap device.
[0010]
Further, the kernel space demand paging swap-out method according to the present invention is such that the kernel demand paging unit specifies a fault occurrence page and a fault factor, and a fault occurrence page specified by the page fault handling means. Swapout priority recalculating means for recalculating swapout priority based on the time stamp value and swapout priority, and re-allocating the physical memory for the faulted page specified by the page fault handling means And memory mapping means for performing mapping to kernel space and, when the physical memory is insufficient, calling a memory allocation algorithm to try to allocate a page of the physical memory.
[0011]
Furthermore, the kernel space demand paging swap-out method of the present invention is characterized in that the kernel swap-out unit selects a swap-out page selection means based on the swap-out priority of each page, and the swap-out page selection means. Time stamp means for setting the current time to the time stamp value of the swap-out target page selected by the page selection means, and swap the swap-out target page selected by the swap-out page selection means from the physical memory to the swap device And swap-out means for out.
[0012]
Furthermore, the kernel space demand paging swap-out method of the present invention is such that the startup kernel swap-out unit initializes the time stamp value and swap-out priority of each page of the kernel space. Out priority initialization means, kernel demand paging section swap-out suppression means for setting a swap-out suppression value in the swap-out priority of the page where the kernel demand paging section exists, the number of pages in the kernel space, and the kernel demand paging And a kernel swap-out section calling unit that calls the kernel swap-out section with a difference from the number of pages of the section as a requested page number.
[0013]
Also, in the demand paging / swap-out method of the kernel space according to the present invention, the swap-out priority recalculation unit is configured such that the swap-out priority P when the n-th page fault occurs for the same page._n, Swap-out priority P until the previous time_n-1And swap-out time T_nBased on P_n= F (P_n-1, T_n) As a feature.
[0014]
  Furthermore, in the demand paging swap-out method of the kernel space according to the present invention, the kernel swap-out unit further includes a swap-out page selection unit,
  The swap out priority recalculation means is configured to calculate the swap out priority.P _n Is the average value of the swap-out time TP _n = P _n-1 + (T _n -P _n-1 ) / NAnd recalculate as
  The swap-out page selection means is configured to select the swap-out priority.P _n A page having a large value is selected as a swap-out target page.
[0015]
  On the other hand, the demand paging swap-out method of the kernel space according to the present invention is a swap-in target when a page fault occurs in the kernel space during system operation in a virtual memory operating system in which the kernel is generated as a single load module. Kernel demand paging process to swap pages into physical memory from the swap device;
  A kernel swap-out process that selects a page with the highest swap-out priority when a user process fails to secure a page of the physical memory during system operation, and swaps out the physical memory to the swap device;
  Call the kernel swap-out process immediately after the end of the kernel startup sequence,Kernel demand paging unit for performing the kernel demand paging processAnd a boot-up kernel swap-out process for swapping out the remaining pages of the kernel space from the physical memory to the swap device.
[0016]
The kernel space demand paging swap-out method according to the present invention includes a page fault handling step in which the kernel demand paging process specifies a fault occurrence page and a fault factor, and a fault occurrence page specified in the page fault handling step. A swap-out priority recalculation step for recalculating the swap-out priority based on the time stamp value and the swap-out priority, and the physical memory re-calculation for the faulted page specified in the page fault handling step. It comprises a memory mapping step of performing page reservation and mapping to a kernel space, and, when the physical memory is insufficient, calls a memory reservation algorithm to try to secure a page of the physical memory.
[0017]
Furthermore, the demand paging swap-out method of the kernel space according to the present invention includes the swap-out page selection step in which the kernel swap-out process selects a swap-out target page based on the swap-out priority of each page, and the swap-out process. A time stamp process for setting a current time to a time stamp value of a swap-out target page selected in the page selection process, and a swap-out target page selected in the swap-out page selection process is swapped from the physical memory to the swap device. It is characterized by comprising a swap-out process of out.
[0018]
Furthermore, in the demand paging / swap-out method of the kernel space according to the present invention, the startup kernel swap-out process initializes the time stamp value and swap-out priority of each page of the kernel space. Out priority initialization process and kernel demand pagingPartDemand paging process for setting swap-out to the swap-out priority of a page where there is a swap-out process Swap-out suppression process, the number of pages in the kernel space, and the kernel demand pagingPartAnd a kernel swap-out process calling step for calling the kernel swap-out process using the difference from the number of pages as the requested page number.
[0019]
Also, in the demand paging / swap-out method of the kernel space according to the present invention, the swap-out priority recalculation step includes the swap-out priority P when the n-th page fault occurs for the same page._n, Swap-out priority P until the previous time_n-1And swap-out time T_nBased on P_n= F (P_n-1, T_n) As a feature.
[0020]
  Furthermore, in the demand paging swap-out method for kernel space of the present invention, the kernel swap-out process further includes a swap-out page selection step,
  In the swap-out priority recalculation step, the swap-out priorityP _n Is the average value of the swap-out time TP _n = P _n-1 + (T _n -P _n-1 ) / NAnd recalculate as
  In the swap-out page selection step, the swap-out priorityP _n A page having a large value is selected as a swap-out target page.
[0021]
On the other hand, the program of the present invention includes a kernel demand paging unit that swaps in a swap-in target page from a swap device to a physical memory when a page fault occurs in a kernel space during system operation. A kernel swap-out unit that selects a page having the highest swap-out priority when the physical memory page allocation fails and swaps out the page from the physical memory to the swap device, and the kernel immediately after completion of the kernel startup sequence Call up the swap-out unit to swap out the remaining pages of the kernel space from the physical memory to the swap device, excluding the page where the kernel demand paging unit exists. To function as an out-section.
[0022]
In the program of the present invention, the kernel demand paging unit includes a page fault handling means for specifying a fault occurrence page and a fault factor, a time stamp value of the fault occurrence page specified by the page fault handling means, and swap-out priority. Swap-out priority recalculating means for recalculating swap-out priority based on the degree, and re-allocating the physical memory to the fault occurrence page specified by the page fault handling means and mapping to the kernel space And memory mapping means for calling a memory reservation algorithm when the physical memory is insufficient to try to secure a page of the physical memory.
[0023]
Further, the program of the present invention includes a swap-out page selection unit that selects a page to be swapped out based on the swap-out priority of each page, and the swap selected by the swap-out page selection unit. A time stamp unit that sets a current time to a time stamp value of an out target page; and a swap out unit that swaps out the swap out target page selected by the swap out page selection unit from the physical memory to the swap device. It is characterized by that.
[0024]
Furthermore, in the program of the present invention, the startup kernel swap-out unit initializes the timestamp value and swap-out priority of each page in the kernel space, and a swap-out priority initialization unit; A kernel demand paging unit swap-out suppression means for setting a swap-out suppression value to a swap-out priority of a page in which the kernel demand paging unit exists, a difference between the number of pages in the kernel space and the number of pages in the kernel demand paging unit. Kernel swap-out section calling means for calling the kernel swap-out section as the number of requested pages.
[0025]
Further, according to the program of the present invention, the swap-out priority recalculation unit is configured such that the swap-out priority P when the n-th page fault occurs for the same page._n, Swap-out priority P until the previous time_n-1And swap-out time T_nBased on P_n= F (P_n-1, T_n) As a feature.
[0026]
  Furthermore, in the program of the present invention, the kernel swap-out unit further includes swap-out page selection means,
  The swap out priority recalculation means is configured to calculate the swap out priority.P _n Is the average value of the swap-out time TP _n = P _n-1 + (T _n -P _n-1 ) / NAnd recalculate as
  The swap-out page selection means is configured to select the swap-out priority.P _n A page having a large value is selected as a swap-out target page.
[0027]
A feature of the present invention is a monolithic kernel generated as a single load module, which does not guarantee that all pages in the kernel space are expanded in physical memory, and the pages of the required kernel functions are physical. Triggered by a page fault that occurs when the page is not expanded in memory, demand paging is performed to swap in the page from the swap device to the physical memory, thereby effectively utilizing the memory resources. In addition, memory resources are effectively used by swapping out pages of kernel functions that are not used or are not frequently used from physical memory to a swap device.
[0028]
Here, the operation of the present invention will be described with reference to FIGS.
[0029]
Normally, the monolithic kernel exists as a single file (load module) on a file storage device (not shown), and is expanded from the file storage device to the physical memory 5 by the bootstrap function, and execution starts from a specific point. Is done.
[0030]
In the present invention, all the text and data areas of the kernel are once expanded in the physical memory 5 by the bootstrap function. However, immediately after the start-up sequence is finished, the page where the kernel demand paging unit 2 exists is displayed by the kernel swap-out unit 3. The remaining pages of the kernel space 4a are temporarily swapped out from the physical memory 5 to the swap device 6. Thereafter, whenever a page fault occurs, the kernel demand paging unit 2 is used to swap in the swap-in target page from the swap device 6 to the physical memory 5 and resume execution from the page fault point. Each page holds the swap-out priority 42 as an attribute value. When a memory request by the user process is requested, if the page reservation of the physical memory 5 by a memory reservation algorithm such as LRU (Least Recently Used) fails, the swap-out priority The pages are swapped out from the physical memory 5 to the swap device 6 in ascending order of 42 to try to secure the pages in the physical memory 5.
[0031]
Specifically, the startup kernel swap-out unit 1 first initializes the time stamp value 41 and the swap-out priority 42 (zero clear) by the time stamp value / swap-out priority initialization unit 11. Next, the kernel demand paging unit swap-out suppression unit 12 sets a swap-out suppression value to the swap-out priority 42 of the page where the kernel demand paging unit 2 exists. Since the kernel itself can know the start address and the end address of the text of the kernel demand paging unit 2, it is possible to specify the page where the kernel demand paging unit 2 exists. The swap-out suppression value needs to be an arbitrary value that is not obtained as a result of the recalculation formula of the swap-out priority 42. For example, 0 or a negative value can be used as the swap-out suppression value. Finally, the kernel swap-out unit calling means 13 designates (the number of pages in the kernel space 4a−the number of pages in the kernel demand paging unit 2) as the requested swap-out size (number of requested pages), and the kernel swap-out unit 3 call.
[0032]
The kernel swap-out unit 3 uses the swap-out page selection unit 31 to select a swap-out target page based on the swap-out priority 42. At this time, the kernel demand paging unit 2 in which the swap-out suppression value is set is selected. The page is not selected as a page to be swapped out.
[0033]
As described above, the startup kernel swap-out unit 1 can swap out the remaining pages of the kernel space 4a from the physical memory 5 to the swap device 6 except for the page where the kernel demand paging unit 2 exists.
[0034]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0035]
(1) First embodiment
FIG. 1 is a block diagram showing a configuration of a demand paging swap-out system for a kernel space according to the first embodiment of the present invention. The demand paging / swap-out method of the kernel space according to the present embodiment includes a startup kernel swap-out unit 1, a kernel demand paging unit 2, a kernel swap-out unit 3, a kernel space 4a and a user space 4b (see FIG. 2)), a physical memory 5 that is a main memory, and a swap device 6 that is a magnetic disk device or the like.
[0036]
The startup kernel swap-out unit 1 includes a timestamp value / swap-out priority initialization unit 11 for initializing the timestamp value 41 and swap-out priority 42 of each page of the kernel space 4a, and the kernel demand paging unit 2. Kernel demand paging section swap-out suppression means 12 for setting a swap-out suppression value in the swap-out priority 42 of the page in which there is a request, and the difference between the number of pages in the kernel space 4a and the number of pages in the kernel demand paging section 2 is requested page It comprises kernel swap-out unit calling means 13 for calling the kernel swap-out unit 3 as a number.
[0037]
The kernel demand paging unit 2 recalculates the swap-out priority 42 based on the page fault handling means 21 that identifies the fault occurrence page and the fault cause, the time stamp value 41 of the fault occurrence page, and the swap-out priority 42. The swap-out priority recalculating means 22 performs re-page reservation of the physical memory 5 and mapping to the kernel space 4a. If the number of requested pages is not reached, the memory reservation algorithm is called to try to secure the page of the physical memory 5. And memory mapping means 23.
[0038]
The kernel swap-out unit 3 includes a swap-out page selection unit 31 that selects a swap-out target page based on the swap-out priority 42 of each page, and a time stamp that sets the current time in the time stamp value 41 of the swap-out target page. Means 32 and swap-out means 33 for swapping out pages to be swapped out of the kernel space 4 a from the physical memory 5 to the swap device 6.
[0039]
Each page constituting the virtual space 4 is accompanied by a time stamp value 41 and a swap-out priority 42 as attribute values.
[0040]
FIG. 3 shows an example in which the time stamp value 41 and the swap-out priority 42 are held in the page table 43 used in normal virtual storage. Each entry in the page table 43 includes a valid flag, a time stamp value 41, a swap-out priority 42, and a physical page or disk address. The physical page address is stored when the valid flag is on (“1”), and the disk address is stored when the valid flag is off (“0”).
[0041]
Referring to FIG. 4, the initialization process of the kernel space demand paging / swap-out method according to the first embodiment includes a kernel startup step S101, a timestamp value / swap-out priority initialization step S102, Kernel demand paging part swap-out suppression step S103, kernel swap-out part call step S104, and user process transition step S105.
[0042]
Referring to FIG. 5, the processing of the kernel demand paging unit 2 includes a page fault handling step S201, a time stamp value / swap out priority acquisition step S202, a swap out priority recalculation step S203, and a swap out priority setting. It consists of step S204 and memory mapping step S205.
[0043]
Similarly, referring to FIG. 5, the process of the kernel swap-out unit 3 includes a swap-out priority order page selection step S301, a suppression page determination step S302, a time stamp value setting step S303, a swap-out step S304, a request It consists of page number determination step S305.
[0044]
Next, the operation of the demand paging / swap-out method in the kernel space according to the first embodiment configured as described above will be described with reference to FIGS.
[0045]
First, when the kernel startup sequence ends (step S101), the startup kernel swap-out unit 1 is executed.
[0046]
The startup kernel swap-out unit 1 initializes the time stamp value 41 and the swap-out priority 42 of each page of the kernel space 4a by the time stamp value / swap out priority initialization means 11 (zero clear) (step clear). S102).
[0047]
Next, the startup kernel swap-out unit 1 uses the kernel demand paging unit swap-out suppression unit 12 to set the swap-out priority 42 of the page where the kernel demand paging unit 2 is present to, for example, 0 or a negative value. An out suppression value is set (step S103).
[0048]
Subsequently, the startup kernel swap-out unit 1 uses the kernel swap-out unit calling means 13 to call the kernel swap-out unit 3 with (number of pages in the kernel space 4a−number of pages in the kernel demand paging unit 2) as the requested number of pages. Call (step S104).
[0049]
The kernel swap-out unit 3 uses the swap-out page selection unit 31 to select pages in the kernel space 4a in descending order of the swap-out priority 42 (step S301), and the swap-out priority 42 is a swap-out suppression value (for example, 0). Whether or not the page is the kernel demand paging unit 2 is determined based on whether it is a negative value (step S302). If it is a page in which the kernel demand paging unit 2 exists, control is returned to step S301, and the page with the next highest swap-out priority 42 is selected.
[0050]
If the kernel demand paging unit 2 does not exist, the kernel swap-out unit 3 sets the current time to the time stamp value 41 by the time stamp means 32 (step S303).
[0051]
Subsequently, the kernel swap-out unit 3 uses the swap-out means 33 to swap out the pages on the physical memory 5 to the swap device 2 (step S304), and whether or not the requested number of pages has been reached, that is, the kernel demand paging unit. It is determined whether or not the remaining pages of the kernel space 4a, excluding the page in which 2 exists, have been swapped out from the physical memory 5 to the swap device 6 (step S305). If the requested number of pages has not been reached, control is returned to step S301, and the page with the next highest swap-out priority 42 is selected. If the requested number of pages has been reached, the process shifts to a user process (usually an init process) (step S105).
[0052]
When a page fault occurs in a page in the kernel space 4a during system operation, the kernel demand paging unit 2 is activated.
[0053]
The kernel demand paging unit 2 uses the page fault handling means 21 to identify the page where the fault has occurred (fault occurrence page) and the cause of the fault (fault cause) (step S201).
[0054]
Next, the kernel demand paging unit 2 acquires the time stamp value 41 and the swap-out priority 42 of the fault occurrence page by the swap-out priority re-calculating means 22 (step S202), and re-calculates the swap-out priority 42. (Step S203), this is reset to the swap-out priority 42 (Step S204).
[0055]
Specifically, each page statistically holds the swap-out time T (swap-out time) T from the time of swap-out to the time of page fault occurrence, and uses this as the swap-out priority 42 to determine whether swap-out is possible or not. Use as judgment material. That is, the swap-out time T is given by the difference between the swap-out occurrence time and the page fault occurrence time. Swap-out priority P when the n-th page fault occurs for the same page_nIs the swap-out priority P up to the previous time_n-1And swap-out time T_nAnd use P_n= F (P_n-1, T_n).
[0056]
As an example, F (P_n-1, T_n) = P_n-1+ (T_n-P_n-1) / N, swap-out priority P_nIs expressed as the average value of the swap-out time T, and the swap-out priority P_nIt is possible to select a page as a swap-out target page from a page with a large value of.
[0057]
Finally, the kernel demand paging unit 2 uses the memory mapping unit 23 to reserve the physical memory 5 again and map it to the kernel space 4a (step S205). At this time, if the requested number of pages is not reached, a memory allocation algorithm is called to try to allocate a page in the physical memory 5.
[0058]
When a user process (not shown) fails to secure a page in the physical memory 5 during system operation, first, a page in the user space 4b is swapped out from the physical memory 5 to the swap device 6 according to a memory securing algorithm.
[0059]
When the page of the user space 4b is swapped out from the physical memory 5 to the swap device 6 and the required number of pages is not reached, the user process has conventionally caused an error due to insufficient capacity of the physical memory 5. Then, the kernel swap-out unit 3 is called.
[0060]
The kernel swap-out unit 3 uses the swap-out page selection unit 31 to select pages in descending order of the swap-out priority 42 in the kernel space 4a (step S301), and the swap-out priority 42 has a swap-out suppression value (for example, 0). It is determined whether or not the kernel demand paging unit 2 is a negative page (step S302). If it is a page in which the kernel demand paging unit 2 exists, control is returned to step S301, and the page with the next highest swap-out priority 42 is selected.
[0061]
If the kernel demand paging unit 2 does not exist, the kernel swap-out unit 3 sets the current time to the time stamp value 41 by the time stamp means 32 (step S303).
[0062]
Subsequently, the kernel swap-out unit 3 uses the swap-out means 33 to swap out the pages on the physical memory 5 to the swap device 2 (step S304), and determines whether the requested number of pages has been reached (step S305). ). If the requested number of pages has not been reached, control is returned to step S301, and the page with the next highest swap-out priority 42 is selected. If the requested number of pages has been reached, control is returned to the user process.
[0063]
As described above, according to the first embodiment, among the pages of the physical memory 5 in which the kernel has conventionally resided, the pages of the physical memory 5 that are hardly used are released to the user process or the like. Thus, memory resources can be used effectively.
[0064]
(2) Second embodiment
FIG. 6 is a block diagram showing a configuration of a kernel space demand paging swap-out method according to the second embodiment of the present invention. The demand paging / swap-out method of the kernel space according to the present embodiment is a virtual storage operating system in the computer 100 as compared with the demand paging / swap-out method of the kernel space according to the first embodiment shown in FIG. The difference is that 200 is provided as a program.
[0065]
The virtual storage operating system 200 is read into the computer 100, and the operation of the computer 100 is changed to the startup kernel swap-out unit 1 (time stamp value / swap-out priority initialization unit 11, kernel demand paging unit swap-out suppression unit) 12, kernel swap-out section calling means 13), kernel demand paging section 2 (page fault handling means 21, swap-out priority recalculation means 22, and memory mapping means 23), and kernel swap-out section 3 (swap-out page) The selection means 31, time stamp means 32, and swap-out means 33) are controlled. The operation of the kernel space demand paging / swap-out method under the control of the virtual memory operating system 200 is the same as that of the kernel space demand paging / swap-out method according to the first embodiment. Value / swap out priority initialization means 11, kernel demand paging section swap out suppression means 12, kernel swap out section calling means 13), kernel demand paging section 2 (page fault handling means 21, swap out priority recalculation means) 22 and the memory mapping means 23) and the kernel swap-out unit 3 (swap-out page selection means 31, time stamp means 32, and swap-out means 33). Since, it omitted the detailed description.
[0066]
In each of the above embodiments, the recalculation formula of the swap-out priority 42 is P_n= F (P_n-1, T_n) = P_n-1+ (T_n-P_n-1) / N, the recalculation formula of the swap-out priority 42 is not limited to the above formula._n= F (P_n-1, T_nNeedless to say, the demand paging swap-out method and method for the kernel space of the present invention can be applied to all the expressions that can be expressed as follows.
[0067]
【The invention's effect】
The first effect is that memory resources can be effectively utilized by releasing, to a user process, a page of physical memory that has rarely been used among physical memory pages in which the kernel has been resident. is there.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a demand paging / swap-out method for a kernel space according to a first embodiment of the present invention.
FIG. 2 is a diagram for explaining kernel demand paging and kernel swap-out in the kernel space demand paging / swap-out method according to the first embodiment;
FIG. 3 is a diagram showing an example in which the time stamp value and the swap-out priority in FIG. 1 are stored in a page table.
FIG. 4 is a flowchart showing processing of a startup kernel swap-out unit in FIG. 1;
FIG. 5 is a flowchart showing processing of a kernel demand paging unit and a kernel swap-out unit in FIG. 1;
FIG. 6 is a block diagram showing a configuration of a kernel space demand paging swap-out method according to a second embodiment of the present invention;
[Explanation of symbols]
1 Kernel swap-out part at startup
2 Kernel demand paging section
3 Kernel swap-out section
4 Virtual space
4a kernel space
4b User space
5 Physical memory
6 Swap devices
11 Time stamp value / swap out priority initialization means
12 Kernel demand paging part swap-out suppression means
13 Kernel swap-out part calling means
21 Page fault handling means
22 Swap out priority recalculation means
23 Memory mapping means
31 Swap out page selection means
32 Time stamp means
33 Swap-out means
41 Timestamp value
42 Swap out priority
43 page table
100 computers
200 Virtual memory operating system
S101 Kernel startup step
S102 Time stamp value / swap out priority initialization step
S103 Kernel demand paging part swap-out suppression step
S104 Kernel swap-out part calling step
S105 User process transition step
S201 Page fault handling step
S202 Time stamp value / swap out priority acquisition step
S203 Swap out priority recalculation step
S204 Swap-out priority setting step
S205 Memory mapping step
S301 Swap out priority order page selection step
S302 Suppression page determination step
S303 Time stamp value setting step
S304 Swap out step
S305 Requested page number determination step

Claims (18)

In a virtual memory computing system that employs a monolithic kernel generated as a single load module,
A kernel demand paging unit that swaps in a swap target page from a swap device to physical memory when a page fault occurs in the kernel space during system operation;
A kernel swap-out unit that selects a page with the highest swap-out priority and swaps out the physical memory to the swap device when a user process fails to secure the physical memory page during system operation;
Immediately after completion of the kernel startup sequence, the kernel swap-out unit is called to swap out the remaining pages of the kernel space from the physical memory to the swap device, excluding the page where the kernel demand paging unit exists. A kernel space demand paging swap-out method characterized by including a kernel swap-out section. The kernel demand paging unit has a page fault handling means for identifying a fault occurrence page and a fault factor, and a swap out priority based on a time stamp value and a swap out priority of the fault occurrence page specified by the page fault handling means. Swap out priority recalculating means for performing recalculation, and reallocating the physical memory to the fault occurrence page specified by the page fault handling means and mapping to the kernel space, and the physical memory is insufficient The kernel space demand paging swap-out system according to claim 1, further comprising memory mapping means for calling a memory allocation algorithm to try to allocate a page of the physical memory. The kernel swap-out unit selects a swap-out page selection unit that selects a swap-out target page based on the swap-out priority of each page, and the time stamp value of the swap-out target page selected by the swap-out page selection unit. 2. A time stamp means for setting a current time, and a swap-out means for swapping out a swap-out target page selected by the swap-out page selection means from the physical memory to the swap device. Demand paging swap-out method of kernel space as described. The startup kernel swap-out unit initializes the timestamp value and swap-out priority of each page in the kernel space, and the page where the kernel demand paging unit exists Kernel demand paging part swap-out suppression means for setting a swap-out suppression value in the swap-out priority of the kernel swap-out part with the difference between the number of pages in the kernel space and the number of pages in the kernel demand paging part as the requested page number 2. The kernel space demand paging swap-out method according to claim 1, further comprising: a kernel swap-out unit calling means for calling. The swap-out priority recalculating means determines the swap-out priority P _n when the n-th page fault occurs for the same page based on the previous swap-out priority P _n−1 and the swap-out time T _n . 3. The demand paging swap-out method for a kernel space according to claim 2, wherein the calculation is performed as P _n = F (P _n−1 , T _n ). The kernel swap-out unit further comprises swap-out page selection means,
The swap-out priority recalculation means recalculates the swap-out priority P _n as an average value P _n = P _n−1 + (T _n −P _n−1 ) / n of the swap out time T. And
6. The demand paging swap-out in the kernel space according to claim 5, wherein the swap-out page selection means selects a page having a large value of the swap-out priority P _n as a page to be swapped out. method. In a virtual memory operating system in which the kernel is generated as a single load module, a kernel demand paging process that swaps in a swap-in target page from the swap device to physical memory when a page fault occurs in the kernel space during system operation When,
A kernel swap-out process that selects a page with the highest swap-out priority when a user process fails to secure a page of the physical memory during system operation, and swaps out the physical memory to the swap device;
Immediately after the end of the kernel startup sequence, the kernel swap-out process is called, and the remaining pages in the kernel space are removed from the physical memory except the page where the kernel demand paging unit for performing the kernel demand paging process exists. A kernel space demand paging swap-out method, comprising: The kernel demand paging process is based on the page fault handling process for identifying the faulting page and the fault factor, and the time out value and the swapout priority of the faulting page specified in the page fault handling process. Swap out priority recalculation step for performing recalculation, and reallocating the physical memory to the fault occurrence page specified in the page fault handling step and mapping to the kernel space, and the physical memory is sufficient. 8. The demand paging swap-out method for a kernel space according to claim 7, further comprising a memory mapping step of calling a memory allocation algorithm to try to allocate a page of the physical memory when there is not. The kernel swap-out process selects a swap-out page selection step for selecting a swap-out target page based on the swap-out priority of each page, and the time-stamp value of the swap-out target page selected in the swap-out page selection step. 8. A time stamp process for setting a current time, and a swap-out process for swapping out a swap-out target page selected in the swap-out page selection process from the physical memory to the swap device. Demand paging swap-out method of kernel space as described. The startup kernel swap-out process initializes the timestamp value and swap-out priority of each page in the kernel space, and the page where the kernel demand paging unit exists. Demand-paging process for setting swap-out suppression value to swap-out priority of the swap-out process, and the kernel swap-out process using the difference between the number of pages in the kernel space and the number of pages in the kernel demand paging unit as the requested page number A kernel space demand paging swap-out method according to claim 7, further comprising: a kernel swap-out process calling step for calling. In the swap-out priority recalculation step, the swap-out priority P _n at the time of the n-th page fault occurrence for the same page is determined based on the previous swap-out priority P _n−1 and the swap-out time T _n . 9. The demand paging swap-out method for a kernel space according to claim 8, wherein the calculation is performed as P _n = F (P _n−1 , T _n ). The kernel swap-out process further includes a swap-out page selection step;
In the swap-out priority recalculation step, the swap-out priority P _n is recalculated as an average value P _n = P _n−1 + (T _n −P _n−1 ) / n of the swap-out time T. And
Wherein in the swapped out page selection process, the swap-out priority large pages of the values of P _n and selects a swap-out target page, demand paging swapped out of kernel space according to claim 11, wherein the Method. Kernel demand paging unit that swaps in the swap-in target page from the swap device to the physical memory when a page fault occurs in the kernel space during system operation of the computer. The user process failed to secure the physical memory page during system operation. In this case, the page having the highest swap-out priority is selected and swapped out from the physical memory to the swap device, and the kernel swap-out unit is called immediately after the end of the kernel startup sequence, To function as a kernel swap-out unit at startup that swaps out the remaining pages of the kernel space from the physical memory to the swap device, excluding the page where the kernel demand paging unit exists Of the program. The kernel demand paging unit has a page fault handling means for identifying a fault occurrence page and a fault factor, and a swap out priority based on a time stamp value and a swap out priority of the fault occurrence page specified by the page fault handling means. Swap out priority recalculating means for performing recalculation, and reallocating the physical memory to the fault occurrence page specified by the page fault handling means and mapping to the kernel space, and the physical memory is insufficient 14. The program according to claim 13, further comprising memory mapping means for calling a memory reservation algorithm to try to secure a page of the physical memory. The kernel swap-out unit selects a swap-out page selection unit that selects a swap-out target page based on the swap-out priority of each page, and the time stamp value of the swap-out target page selected by the swap-out page selection unit. 14. A time stamp means for setting a current time, and a swap out means for swapping out a swap out target page selected by the swap out page selection means from the physical memory to the swap device. The listed program. The startup kernel swap-out unit initializes the timestamp value and swap-out priority of each page in the kernel space, and the page where the kernel demand paging unit exists Kernel demand paging part swap-out suppression means for setting a swap-out suppression value in the swap-out priority of the kernel swap-out part with the difference between the number of pages in the kernel space and the number of pages in the kernel demand paging part as the requested page number 14. The program according to claim 13, further comprising a kernel swap-out part calling means for calling. The swap-out priority recalculating means determines the swap-out priority P _n when the n-th page fault occurs for the same page based on the previous swap-out priority P _n−1 and the swap-out time T _n . The program according to claim 14, wherein the calculation is performed as P _n = F (P _n−1 , T _n ). The kernel swap-out unit further comprises swap-out page selection means,
The swap-out priority recalculation means recalculates the swap-out priority P _n as an average value P _n = P _n−1 + (T _n −P _n−1 ) / n of the swap out time T. And
The program according to claim 17, wherein the swap-out page selection unit selects a page having a large value of the swap-out priority P _n as a swap-out target page.

Images