From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264515AbUEDQzC (ORCPT ); Tue, 4 May 2004 12:55:02 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264514AbUEDQzC (ORCPT ); Tue, 4 May 2004 12:55:02 -0400 Received: from mailgw.cvut.cz ([147.32.3.235]:18092 "EHLO mailgw.cvut.cz") by vger.kernel.org with ESMTP id S264517AbUEDQy7 (ORCPT ); Tue, 4 May 2004 12:54:59 -0400 From: "Petr Vandrovec" Organization: CC CTU Prague To: linux-kernel@vger.kernel.org Date: Tue, 4 May 2004 18:54:54 +0200 MIME-Version: 1.0 Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Subject: Accessing sysenter page by kernel syscalls X-mailer: Pegasus Mail v3.50 Message-ID: <5737E6105A3@vcnet.vc.cvut.cz> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hi, is there any reason why copy_from_user does not punch hole in kernel space to get it possible to access sysenter page? I do not think that it would add too big overhead to copy_*_user (which is already long), but I cannot think about any other reason why sysenter page is not accessible through write(). Today I wanted to confirm whether my kernel uses sysenter or int 0x80, and I found that simple program below does not work and I had to copy page first to some temporary location, and then write it from this temporary location to stdout. void main(void) { write(1, (void*)0xFFFFE000, 4096); } Thanks, Petr Vandrovec