[uClibc]Uses a bigger busybox more mem for (f.i.) ash?
Erik Andersen
andersen at codepoet.org
Mon Dec 3 19:06:46 UTC 2001
On Mon Dec 03, 2001 at 10:58:22AM -0800, Stefan Soucek wrote:
> (snipp)
> > Linux performs demand paging. That means when an application
> > is loaded, only the 4k-sized pages of the application that
> > are being used will be loaded into memory. All of the pages
> > of memory for the .text segment of busybox (i.e. almost all
> > of it) will be shared between successive invocations of
> > busybox.
>
> I don't think this statement is true for non-MMU systems. The
> current uClinux kernel (for ARM) loads the complete .text
> segment into memory. A subsequent call to busybox loads the
> .text segement again. To my understanding there is no sharing
> of code pages possible on non-MMU systems, since the code is
> relocated each time. In this case the actual size of the
> busybox .text segment (and .data, .bss as well) does matter.
You are correct. I was making the assumption we were talking
about MMU-full Linux.
When using busybox on mmu-less systems, it makes sense to compile
up several smaller instances of busybox, rather then one giant
binary.
-Erik
--
Erik B. Andersen http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--
More information about the uClibc
mailing list