From: Oscar Waddell < owaddell@nickel.ucs.indiana.edu>
Subject: do mmap and/or dlopen exist for linux?
Date: Wed, 16 Jun 1993 02:22:34 GMT
Hi Folks,
Is there anything for Linux like mmap but for memory mapped *files*.
Is there anything for Linux analogous to dlopen on the Suns
(an interface to dynamic link-editor)?
Please e-mail with any clues,
Oscar
(owaddell@cs.indiana.edu)
From: entropy@world.std.com (Lawrence Foard)
Subject: Re: do mmap and/or dlopen exist for linux?
Date: Thu, 17 Jun 1993 21:40:25 GMT
In article <C8ozxM.LB8@usenet.ucs.indiana.edu>
Oscar Waddell <owaddell@nickel.ucs.indiana.edu> writes:
>
>Hi Folks,
>
>Is there anything for Linux like mmap but for memory mapped *files*.
I've been playing with memory management some and have a read only
mmap call working. Basically borrowing from the code that maps executables
and libraries. If anyone wants the patchs I'll e-mail them, however I
won't make any promises about writtable MMF's since I don't have much
free time. First I'd like to unify memory mapped files, shared libraries
and executables.
The read only ones do make a huge difference, I can now easily view multiple
multimeg images at once on my 8 meg computer with very little thrashing.
Redisplay just causes sequential disk reads without the nasty grinding noise
you usually get from swapping.
--
====== Forgive me for I have eaten of beans and believed in irrational .
\ / numbers. . .
\ / . . .
\/ . . . .
From: muts@muts.hacktic.nl (Peter Mutsaers)
Subject: Re: do mmap and/or dlopen exist for linux?
Date: Sat, 19 Jun 1993 16:14:54 GMT
On Thu, 17 Jun 1993 21:40:25 GMT, entropy@world.std.com (Lawrence Foard) said:
LF> I've been playing with memory management some and have a read
LF> only mmap call working. Basically borrowing from the code that
LF> maps executables and libraries. If anyone wants the patchs I'll
LF> e-mail them, however I won't make any promises about writtable
LF> MMF's since I don't have much free time. First I'd like to unify
LF> memory mapped files, shared libraries and executables.
In the kernel source there is a file called /usr/src/linux/mm/mmap.c
(pl10) and in libc.a there is
mmap.o:
U _errno
00000000 T _mmap
I didn't try them, but are they not functional then?
--
______________________________________________________________________
Peter Mutsaers | Bunnik (Ut), | Quod licet bovi,
muts@muts.hacktic.nl | the Netherlands | non licet Jovi
From: sct@dcs.ed.ac.uk (Stephen Tweedie)
Subject: Re: do mmap and/or dlopen exist for linux?
Date: 22 Jun 93 02:20:29 GMT
On 19 Jun 93 16:14:54 GMT, muts@muts.hacktic.nl (Peter Mutsaers) said:
> On Thu, 17 Jun 1993 21:40:25 GMT, entropy@world.std.com (Lawrence Foard) said:
> LF> I've been playing with memory management some and have a read
> LF> only mmap call working. Basically borrowing from the code that
> LF> maps executables and libraries. If anyone wants the patchs I'll
> LF> e-mail them, however I won't make any promises about writtable
> LF> MMF's since I don't have much free time. First I'd like to unify
> LF> memory mapped files, shared libraries and executables.
It's been done - thank Eric Youngdale for that!
> In the kernel source there is a file called /usr/src/linux/mm/mmap.c
> (pl10) and in libc.a there is
> mmap.o:
> U _errno
> 00000000 T _mmap
> I didn't try them, but are they not functional then?
Nope, they're not quite. In particular, the standard mmap() only
allows you to map /dev/mem, /dev/kmem and /dev/zero. (XFree86 needs
the /dev/mem mmap() in order to map the video frame buffer into
memory.)
Eric has got a much improved mmap() going, but it's still missing a
lot; in particular, it is pretty much read-only just now. I'm working
on the rest of the mmap() semantics, and should have something
workable soon.
(For all of you waiting for a compressed filesystem, don't panic - I
haven't completely forgotten about it, honest! The new mmap is quite
important because it is intricately tied into both the memory manager
and the filesystem, and the compressed filesystem introduces technical
problems which I really need to get all sorted out before proceeding
with the compressing code.)
Cheers,
Stephen.
---
Stephen Tweedie (Internet: )
Department of Computer Science, Edinburgh University, Scotland.