allbery@ncoast.UUCP (Brandon Allbery) (03/08/86)
Expires: Quoted from <4479@think.ARPA> ["shared memory: a poll"], by rose@think.ARPA (John Rose)... +--------------- | The "4.2BSD System Manual", in the section on memory management | (/usr/doc/sysman/1.2.t), sketches out a future extension to Unix* | called "mmap()". This feature allows an "object" represented by | a file-descriptor to be mapped into the per-process address space; | there is provision for modifications made to the "object" to be | shared by more than one process. | | I haven't seen much discussion of this kind of memory management | on the net, but I have heard of vendors implementing it. So | there must be people out there who have used it in interesting ways. +--------------- Xenix 3.0 (I don't know about 5.0) does something that sounds like this. There is a special type of file which is a shared-memory file. A process opens the file with a special system call (the file must exist or be created), then any process which opens the file gets it attached to its memory space. The problems: (1) Xenix runs very slowly when a shared memory is open for writing by more than one process (so say the docs); (2) they say NEVER leave the shared memory attached while making a system call, as it can crash Xenix(!). There are system calls (aside from create/open and close of the shared-memory file) which attach and detach the shared memory from the process's address space. Anyone with more Xenix and/or 4.2/4.3 experience know anything to prove or disprove this? --Brandon (P.S. The shared-memory file is an S_IFNAM; there is a special flag in the inode which distinguishes between FIFOs and shared-memory files. That may give some idea what Xenix is up to.) -- In mid-winter, all of us Midwesterners would *love* a taste of California... (r-r-r-rumbleSHA-A-KE!!!) ...but not *that* badly! decvax!cwruecmp!ncoast!allbery ncoast!allbery@Case.CSNET ncoast!tdi2!brandon (ncoast!tdi2!root for business) 6615 Center St. #A1-105, Mentor, OH 44060-4101 Phone: +01 216 974 9210 CIS 74106,1032 MCI MAIL BALLBERY (part-time) >>> Nominate Mark Leeper for a Hugo as Best Fan Writer of '85 <<< (I disagree with many of his reviews, but that's just me. He's better than most of what I see. Let's show the world what the NET is!)