Subject: Re: Increasing SHMMAXPGS
To: Soren S. Jorvang <email@example.com>
From: Jim Wise <firstname.lastname@example.org>
Date: 05/20/2002 10:44:51
-----BEGIN PGP SIGNED MESSAGE-----
On Mon, 20 May 2002, Soren S. Jorvang wrote:
>On Mon, May 20, 2002 at 11:11:03AM +0200, Ignatios Souvatzis wrote:
>> > But we could at least make segments whose creator crashes go
>> > away, as per nasal demons, which I think would go a long way
>> > in practice.
>> The creator is allowed, and sometimes supposed, to go away.
>> shm is a volatile filesystem.
>But we are also allowed to have an implementation-specific method
>for removing self-created segments that just happens to be invoked
>by, say, referencing address 0.
If we want to do this, perhaps we can make this configurable via a new
bit in the shmflag argument to shmget(2)?
Something like IPC_CLOSEONEXIT?
I _don't_ think that we should change the default behavior for these
things -- the current behavior (segment remains until explicitly
removed) is well-known, and depended on in enough cases to make this an
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (NetBSD)
-----END PGP SIGNATURE-----