From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from dukat.scot.redhat.com (sct@dukat.scot.redhat.com [195.89.149.246]) by kvack.org (8.8.7/8.8.7) with ESMTP id RAA10767 for ; Mon, 10 May 1999 17:19:47 -0400 From: "Stephen C. Tweedie" MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <14135.19821.176947.167396@dukat.scot.redhat.com> Date: Mon, 10 May 1999 22:19:41 +0100 (BST) Subject: Re: mmap operation In-Reply-To: <199905101602.SAA11187@zange.cs.tu-berlin.de> References: <199905101602.SAA11187@zange.cs.tu-berlin.de> Sender: owner-linux-mm@kvack.org To: Gilles Pokam Cc: linux-mm@kvack.org List-ID: Hi, On Mon, 10 May 1999 18:02:10 +0200 (MET DST), Gilles Pokam said: > I have implemented a module for buffer management in the 2.0.34 linux > kernel. Now i'm using the write and read method for transfering data > between the kernel and user space. I have noticed that the overhead of > these 2 operations are quite big, because each time a system call is > invoked. So i decided to improve my module by implementing the mmap > operation. The mmap operation works well when mapping only one page > size. But above this size, for example with an order of at least 1, > the later operation fails to work! I have noticed that above 4096 (one > page) bytes the zero page is mapped instead! Could someone help mee > solve this problem ? (i use the nopage operation in my mmap method and > cluster of 16 pages). The entire VM works by mapping single pages. It would be a large task to change this. --Stephen -- To unsubscribe, send a message with 'unsubscribe linux-mm my@address' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://humbolt.geo.uu.nl/Linux-MM/