On Fri, Oct 29, 2004 at 03:26:44PM +1000, Rod Butcher wrote: > I read the Con Kolivas page.. it has the following :- > Many machines now have exactly 1Gb ram and the standard memory split on > i386 does not allow you to use more than 896Mb ram without enabling high > memory for at least 4Gb. The problem is that this incurs an overhead > whereas we can simply change this split with this patch to allow i386 > architectures to use up to 1Gb ram without enabling highmem. This is > configurable if HIGHMEM is disabled. > > Now.. I fit this category, having 1 gig ram and having to set > "CONFIG_HIGHMEM4G=y" to compile 2.6.7.1 to allow usage of the full 1 > gig. Anyone know what sort of "overhead" is involved here ? If it's > significant I'll look at implementing the patch.. but it seems to me > that if its such a problem-solver why isn'ty it in the stock kernel ?
IMHO, Con knows his stuff and what he's saying is very probably exactly right. However, it doesn't sound like it deals with your specific problem, which is swap behaviour. I don't know enough about x86 architecture to know exactly what he's referring to, but there's a lot of architectures that have seemingly arbitrary bottle-necks on accessing particular address ranges. And patches generally need a lot of testing before they get into the kernel, but there's all sorts of reasons it might not be in there. It's worth tracking LKML if you're interested in this stuff. HTH, James. -- "Now, there are no problems only opportunities. However, this seemed to be an insurmountable opportunity." - http://www.surfare.net/~toolman/temp/diagram.html -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
