NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/sys/hp/8330

Text File | 1992-07-21 | 2.0 KB | 45 lines

Newsgroups: comp.sys.hp
Path: sparky!uunet!utcsri!torn!cunews!nrcnet0!bnrgate!bmers145!bnr.ca!jasonc
From: [e-mail] (Jason Chen)
Subject: Re: Shared memory in 6.5
Message-ID: <[e-mail]>
Sender: [e-mail]
Reply-To: [e-mail] (Jason Chen)
Organization: Bell-Northern Research Ltd.
References: <[e-mail]> <[e-mail]>
Date: Tue, 21 Jul 92 15:09:22 GMT
Lines: 32

In article <[e-mail]>, [e-mail] (Rob Gardner) writes:
|> I still don't understand your situation clearly. Even if multiple
|> processes need to attach the segment, depending on your application,
|> you can have the last process that needs to attach it call
|> shmctl(IPC_RMID).  Sorry, but that's the best you can do; Shared
|> memory segments are *meant* to stick around after processes die,
|> and you must remove them explicitly.

Rob,

        Thanks again for the reply.  Here is what I tried to do:
Basically, I want to create a pair of client-server processes in where
the server gets a shared memory segment and puts its pid in the segment,
and the client gets the server pid and determines whethere the server is
still running.  The reason I want the segment to be remove when shm_nattck
equals zero is that the client can use this (shmget() failure) to determine
the existence of the server.  As we all know that some signals can not be
caught, and the pid could be reused by another process (it may take a
long time, but ...).  I know that there are many other ways to do this
e.g. using shmctl(IPC_STAT) to get shm_nattck, or using file with flock().
We have decided to used file instead of shared memory.  I did learn some
thing about shared memory.

Jason



--
Jason X.G. Chen      | "A distributed system is one in which the
Ph:  [phone]  | failure of a computer you didn't even know
Fax: [phone]  | existed can render your own computer unusable"
Email: [e-mail] |                   Leslie Lamport