Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Enforced Page Data Integrity for Mapped File Memory Segments

IP.com Disclosure Number: IPCOM000062181D
Original Publication Date: 1986-Oct-01
Included in the Prior Art Database: 2005-Mar-09
Document File: 1 page(s) / 12K

Publishing Venue

IBM

Related People

Duvall, KE: AUTHOR [+2]

Abstract

By using a Virtual Memory Manager (VMM) Supervisor Call (SVC), Purge Page Range, to force modified pages from one mapped memory segment to the corresponding disk blocks, and the use of a new VMM SVC, Sync Page Range, to instruct the VMM to synchronize the pages in the specified page range(s), the user is guaranteed that the segment reflects the current disk block contents. Mapped file support consists of a mapped file user interface and major UNIX* kernel modifications to support this interface. Mapped file support makes extensive use of a Virtual Memory Manager to map pages of a memory segment to the disk blocks of a file so that the VMM can efficiently perform all read disk I/O to and from the corresponding memory segment. A mapped file and its associated disk blocks may concurrently be mapped into multiple memory segments.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 1 of 1

Enforced Page Data Integrity for Mapped File Memory Segments

By using a Virtual Memory Manager (VMM) Supervisor Call (SVC), Purge Page Range, to force modified pages from one mapped memory segment to the corresponding disk blocks, and the use of a new VMM SVC, Sync Page Range, to instruct the VMM to synchronize the pages in the specified page range(s), the user is guaranteed that the segment reflects the current disk block contents. Mapped file support consists of a mapped file user interface and major UNIX* kernel modifications to support this interface. Mapped file support makes extensive use of a Virtual Memory Manager to map pages of a memory segment to the disk blocks of a file so that the VMM can efficiently perform all read disk I/O to and from the corresponding memory segment. A mapped file and its associated disk blocks may concurrently be mapped into multiple memory segments. Therefore, a need existed to be able to make a change to a file's real disk block, resulting from a change to a memory segment page mapped to the block, and to be able to be synchronized into other memory segment pages which are mapped to the same disk block. The solution to this problem, discussed below, guarantees that at all times, for a given mapped file, the pages of a mapped Read/Write segment, the unmodified pages of a mapped Copy-on- Write segment, or the unmodified pages of a mapped shared Copy-on-Write segment, reflect the current contents of the corresponding disk blocks for the file. In the past, this condition was not guaranteed, since changes to a file's disk block, resulting from a write modification of the mapped file Read/ Write segment, or the synchronizing of one of the mapped segments, was not...