Re: [9fans] new wrap

Re: [9fans] new wrap

[presotto]

[-- Attachment #1: Type: text/plain, Size: 1007 bytes --]

I was going to call it shmem originally but thought better of it
after the flap over int/void main().

I added it at the request of a piece of Lucent that wants to
be able to have a large in memory database sharable by
unrelated processes.  Any process can see them but segattaching,
reading, and writing uses the file protections.  I may add that
restriction to removing also, haven't decided yet.

In the interest of being neighborly, I added it to the general
distribution.  I made it a separate device mostly because I
don't think its incredibly useful.  It has our normal
shared segment restriction that it must be at the
same virtual address in all processes that use it.
If you have locks in it, there's no mechanism for
backing out the locks held if a process dies.
It's very much a solution to a specific problem...

It's a little nicer than shmem in that
processes on other machines can get at the segment, albeit
with some difficulty, by importing the file system interface.

[-- Attachment #2: Type: message/rfc822, Size: 3244 bytes --]

[-- Attachment #2.1.1: Type: text/plain, Size: 72 bytes --]

I'd assumed it was to assist in the proper implementation of main().


[-- Attachment #2.1.2: Type: message/rfc822, Size: 1474 bytes --]

From: Scott Schwartz <schwartz@bio.cse.psu.edu>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] new wrap
Date: Mon, 07 May 2001 13:41:19 -0400
Message-ID: <20010507174119.21095.qmail@g.bio.cse.psu.edu>

Dave writes:
| This is sort of like '#s' for segments, i.e., segments
| that can be attached by unrelated processes. 

Can you say a few words about what inspired this, and how it's intended to
be used?  Things in #s are always globally visible... ## too?  It reminds
me of SysV shm.

Re: [9fans] new wrap

[nigel]

[-- Attachment #1: Type: text/plain, Size: 72 bytes --]

I'd assumed it was to assist in the proper implementation of main().


[-- Attachment #2: Type: message/rfc822, Size: 1474 bytes --]

From: Scott Schwartz <schwartz@bio.cse.psu.edu>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] new wrap
Date: Mon, 07 May 2001 13:41:19 -0400
Message-ID: <20010507174119.21095.qmail@g.bio.cse.psu.edu>

Dave writes:
| This is sort of like '#s' for segments, i.e., segments
| that can be attached by unrelated processes. 

Can you say a few words about what inspired this, and how it's intended to
be used?  Things in #s are always globally visible... ## too?  It reminds
me of SysV shm.

Re: [9fans] new wrap

[Scott Schwartz]

Dave writes:
| This is sort of like '#s' for segments, i.e., segments
| that can be attached by unrelated processes. 

Can you say a few words about what inspired this, and how it's intended to
be used?  Things in #s are always globally visible... ## too?  It reminds
me of SysV shm.

[9fans] new wrap

[presotto]

We made another wrap updating the last full release.  See the usual web
pages.

http://plan9.bell-labs.com/plan9dist/ureg.html
http://plan9.bell-labs.com/plan9dist/03270507.html

It has a few fixes, like the TCP checksum one, and a new device,
devsegment.c.  This is sort of like '#s' for segments, i.e., segments
that can be attached by unrelated processes.  See

http://plan9.bell-labs.com/magic/man2html/3/segment