From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from max.fys.ruu.nl (max.fys.ruu.nl [131.211.32.73]) by kvack.org (8.8.7/8.8.7) with ESMTP id IAA26874 for ; Thu, 27 Nov 1997 08:38:11 -0500 Date: Thu, 27 Nov 1997 14:24:19 +0100 (MET) From: Rik van Riel Subject: Re: Linux memory management wish :-) In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-linux-mm@kvack.org To: Andrea Cisternino Cc: linux-mm List-ID: On Wed, 26 Nov 1997, Andrea Cisternino wrote: > Hi Rik, > > > Send Linux memory-management wishes to me: I'm currently looking > > for something to hack... > > OK, here I go... > > I'm a researcher at the italian National Institute of Nuclear Physics > (INFN) and I'm currently developing Linux device drivers for one of our > projects (you can find more info in http://pcape1.pi.infn.it/~acister/dev). > > The hardware/software combination I have to deal with needs big contiguous > memory buffers for DMA operations from/to the PCI bus. > > To get such buffers I'm presently using the bigphysarea patch from Matt > Welch to reserve a memory pool at boot time but I would really like to see > some "standard" kernel support for this kind of memory allocation. We could 'reserve' a large amount of memory for use by DMA/buffer/cache only, so when you need a large amount of memory, the system can kick out buffermem and cache- mem on request... Also, the memory in that area can't be allowed to have lots of dirty pages in it. Kflushd will have to check it every xx jiffies. On a 128M machine, reserving 16MB for buffer/cache mem isn't that bad. I'll look into this (after my crashing HD is replaced by a new one... I can't risk losing even more data). grtz, Rik. ---------- Send Linux memory-management wishes to me: I'm currently looking for something to hack...