公开(公告)号：EP0239181B1

公开(公告)日：1994-12-07

申请号：EP87300111.9

申请日：1987-01-08

申请人： International Business Machines Corporation

发明人： Chang, Albert ,  Mergen, Mark Frederick ,  O'Quin II, John Thomas ,  O'Quin III, John Claude ,  Rogers, Mark Douglas

IPC分类号： G06F12/10 ,  G06F13/10 ,  G06F9/46

CPC分类号： G06F9/4812 ,  G06F12/10

公开(公告)号：EP0230354A3

公开(公告)日：1990-05-30

申请号：EP87300127.5

申请日：1987-01-08

申请人： International Business Machines Corporation

发明人： Rogers, Mark Douglas ,  O'Quinn II, John Thomas ,  O'Quinn II, John Claude

IPC分类号： G06F12/10

CPC分类号： G06F12/1009

摘要： In a virtual memory data processing system, the use of a pageable External Page Table data structure for recording current status and disk address information for each virtual page in the system and a non-pageable or pinned table of tables, provides improved system performance when a large number of virtual pages are to be operated on in the same manner. As disclosed, each page of External Page Table entries can record a predetermined number of entries (512), depending on the byte capacity of each virtual page (2,048) and the size of each entry (4 bytes). One page of 512 entries corresponds to 1 megabyte of virtual storage (512 × 2,048) and also appears as one entry in a pinned External Page Table. The pinned External Page Table is referred to as the "XPT of the XPT," and has the same format as the pageable XPT. A 256 megabyte segment of virtual memory is representable in the XPT of the XPT by 256, 4 byte entries, or one-half of a page. Where all virtual pages represented by a page of XPT entries are to be processed in the same way (i.e., created, destroyed), one operation on one entry in the XPT of the XPT is made, rather than 512 operations on the 512 entries of the XPT.

公开(公告)号：EP0239181A3

公开(公告)日：1990-06-27

申请号：EP87300111.9

申请日：1987-01-08

申请人： International Business Machines Corporation

发明人： Chang, Albert ,  Mergen, Mark Frederick ,  O'Quin II, John Thomas ,  O'Quin III, John Claude ,  Rogers, Mark Douglas

IPC分类号： G06F12/10 ,  G06F13/10 ,  G06F9/46

CPC分类号： G06F9/4812 ,  G06F12/10

摘要： In a virtual memory priority interrupt data processing system, system events are logically serialised, because each event causes a different predetermined interrupt servicing routine to be executed, which processes virtual pages in a manner that requires data structures that record the current status of virtual pages to be updated to reflect the change of state resulting from processing. The system events to be serialised include the interrupt that is generated as a result of a page fault, the interrupt that is generated as a result of a paging I/O completion operation that resolves a page fault by paging in a virtual page from the secondary storage device, and lastly, interrupts that are generated in response to a supervisory call to a paging service, such as create or destroy segment. These latter supervisory calls cause the status of one or more pages to be changed and, therefore, the data structure which records such changes must also be updated. The page fault interrupt handling process is only interruptable by another page fault interrupt, or an I/O. An I/O interrupt merely sus­pends the page fault handler. At the completion of the routine that services the I/O device interrupt, the page fault handler continues from the point of interruption. A second page fault interrupt, however, is serviced and the first page fault is restarted at the beginning of its critical section after the second page fault has been serviced. I/O page fault completion interrupts are placed in a queue. The page faulting process is executed in a mode which delays processing this queue, so these interrupts are serialised after the page fault handling interrupts. The supervisory calls to paging services routines also are executed in a mode that delays processing of enqueued I/O page fault completion interrupts. These interrupts are, therefore serialised last, relative to the other system events. The paging service routines do, however, honour page fault interrupts which are, therefore, serialised ahead of the supervisory calls to paging services. When the page fault handler completes, the interrupted paging service will be re-started at the beginning of its critical section. All three system events are, therefore, logically serialised to preserve the integrity of all data structures that record current status of virtual pages and page frames in the system.

公开(公告)号：EP0239181A2

公开(公告)日：1987-09-30

申请号：EP87300111.9

申请日：1987-01-08

申请人： International Business Machines Corporation

发明人： Chang, Albert ,  Mergen, Mark Frederick ,  O'Quin II, John Thomas ,  O'Quin III, John Claude ,  Rogers, Mark Douglas

IPC分类号： G06F12/10 ,  G06F13/10 ,  G06F9/46

CPC分类号： G06F9/4812 ,  G06F12/10

摘要： In a virtual memory priority interrupt data processing system, system events are logically serialised, because each event causes a different predetermined interrupt servicing routine to be executed, which processes virtual pages in a manner that requires data structures that record the current status of virtual pages to be updated to reflect the change of state resulting from processing. The system events to be serialised include the interrupt that is generated as a result of a page fault, the interrupt that is generated as a result of a paging I/O completion operation that resolves a page fault by paging in a virtual page from the secondary storage device, and lastly, interrupts that are generated in response to a supervisory call to a paging service, such as create or destroy segment. These latter supervisory calls cause the status of one or more pages to be changed and, therefore, the data structure which records such changes must also be updated.
The page fault interrupt handling process is only interruptable by another page fault interrupt, or an I/O. An I/O interrupt merely sus­pends the page fault handler. At the completion of the routine that services the I/O device interrupt, the page fault handler continues from the point of interruption. A second page fault interrupt, however, is serviced and the first page fault is restarted at the beginning of its critical section after the second page fault has been serviced.
I/O page fault completion interrupts are placed in a queue. The page faulting process is executed in a mode which delays processing this queue, so these interrupts are serialised after the page fault handling interrupts. The supervisory calls to paging services routines also are executed in a mode that delays processing of enqueued I/O page fault completion interrupts. These interrupts are, therefore serialised last, relative to the other system events. The paging service routines do, however, honour page fault interrupts which are, therefore, serialised ahead of the supervisory calls to paging services. When the page fault handler completes, the interrupted paging service will be re-started at the beginning of its critical section. All three system events are, therefore, logically serialised to preserve the integrity of all data structures that record current status of virtual pages and page frames in the system.

摘要翻译： 在虚拟内存优先级中断数据处理系统中，系统事件被逻辑地串行化，因为每个事件导致执行不同的预定中断服务程序，其以需要将虚拟页面的当前状态记录的数据结构的方式处理虚拟页面 更新以反映处理所导致的状态变化。 要串行化的系统事件包括作为页面错误的结果生成的中断，作为通过从次要页面的虚拟页面中的寻呼来解决页面错误的分页I / O完成操作而产生的中断 存储设备，最后是响应于对寻呼服务的监视呼叫而生成的中断，诸如创建或销毁段。 这些后面的监督呼叫导致一个或多个页面的状态被改变，因此记录这些改变的数据结构也必须被更新。 页面错误中断处理过程只能被另一个页面故障中断或I / O中断。 I / O中断只会暂停页面故障处理程序。 在完成对I / O设备中断服务的例程中，页面故障处理程序从中断点继续。 然而，第二页故障中断被服务，并且在第二页故障被维护之后，在其关键部分的开始处重新开始第一页故障。 I / O页面故障完成中断被放置在队列中。 页面故障处理以延迟处理该队列的方式执行，因此在页面故障处理中断之后，这些中断被序列化。 对寻呼服务程序的监督呼叫也以延迟进入的I / O页面故障完成中断处理的模式执行。 因此，相对于其他系统事件，这些中断是最后序列化的。 然而，寻呼服务程序确实会损害页面故障中断，因此在分页服务的监督呼叫之前被序列化。 当页面错误处理程序完成时，中断的寻呼服务将在其关键部分的开头重新启动。 因此，所有三个系统事件都是逻辑序列化的，以保持记录系统中虚拟页面和页面帧的当前状态的所有数据结构的完整性。

公开(公告)号：EP0230354B1

公开(公告)日：1994-08-10

申请号：EP87300127.5

申请日：1987-01-08

申请人： International Business Machines Corporation

发明人： Rogers, Mark Douglas ,  O'Quinn II, John Thomas ,  O'Quinn II, John Claude

IPC分类号： G06F12/10

CPC分类号： G06F12/1009

公开(公告)号：EP0230354A2

公开(公告)日：1987-07-29

申请号：EP87300127.5

申请日：1987-01-08

申请人： International Business Machines Corporation

发明人： Rogers, Mark Douglas ,  O'Quinn II, John Thomas ,  O'Quinn II, John Claude

IPC分类号： G06F12/10

CPC分类号： G06F12/1009

摘要： In a virtual memory data processing system, the use of a pageable External Page Table data structure for recording current status and disk address information for each virtual page in the system and a non-pageable or pinned table of tables, provides improved system performance when a large number of virtual pages are to be operated on in the same manner. As disclosed, each page of External Page Table entries can record a predetermined number of entries (512), depending on the byte capacity of each virtual page (2,048) and the size of each entry (4 bytes). One page of 512 entries corresponds to 1 megabyte of virtual storage (512 × 2,048) and also appears as one entry in a pinned External Page Table. The pinned External Page Table is referred to as the "XPT of the XPT," and has the same format as the pageable XPT. A 256 megabyte segment of virtual memory is representable in the XPT of the XPT by 256, 4 byte entries, or one-half of a page.
Where all virtual pages represented by a page of XPT entries are to be processed in the same way (i.e., created, destroyed), one operation on one entry in the XPT of the XPT is made, rather than 512 operations on the 512 entries of the XPT.

摘要翻译： 在虚拟存储器数据处理系统中，使用用于记录系统中每个虚拟页面的当前状态和磁盘地址信息的可分页外部表表数据结构以及不可页面或固定的表格表提供了改进的系统性能 以相同的方式操作大量虚拟页面。 如所公开的，根据每个虚拟页面的字节容量（2,048）和每个条目的大小（4字节），外部页表项的每个页面可以记录预定数量的条目（512）。 一页512个条目对应于1兆字节的虚拟存储（512 x 2,048），并且在固定外部页表中也显示为一个条目。 固定的外部页表被称为“XPT的XPT”，具有与可分页XPT相同的格式。 一个256 MB的虚拟内存段可以在XPT的XPT中被256,4个字节的条目或页面的一半表示。 ...由于XPT条目页面所表示的所有虚拟页面都以相同的方式进行处理（即，创建，销毁），XPT的XPT中的一个条目将进行一次操作，而不是512个操作 在XPT的512个条目上。