From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932392AbcAHTlK (ORCPT ); Fri, 8 Jan 2016 14:41:10 -0500 Received: from www.linutronix.de ([62.245.132.108]:39653 "EHLO Galois.linutronix.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753460AbcAHTlI (ORCPT ); Fri, 8 Jan 2016 14:41:08 -0500 Date: Fri, 8 Jan 2016 20:40:14 +0100 (CET) From: Thomas Gleixner To: Dave Hansen cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, x86@kernel.org, dave.hansen@linux.intel.com Subject: Re: [PATCH 29/31] x86, pkeys: allow kernel to modify user pkey rights register In-Reply-To: <20160107000145.96AD9FDD@viggo.jf.intel.com> Message-ID: References: <20160107000104.1A105322@viggo.jf.intel.com> <20160107000145.96AD9FDD@viggo.jf.intel.com> User-Agent: Alpine 2.11 (DEB 23 2013-08-11) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Linutronix-Spam-Score: -1.0 X-Linutronix-Spam-Level: - X-Linutronix-Spam-Status: No , -1.0 points, 5.0 required, ALL_TRUSTED=-1,SHORTCIRCUIT=-0.0001,URIBL_BLOCKED=0.001 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 6 Jan 2016, Dave Hansen wrote: > From: Dave Hansen > > The Protection Key Rights for User memory (PKRU) is a 32-bit > user-accessible register. It contains two bits for each > protection key: one to write-disable (WD) access to memory > covered by the key and another to access-disable (AD). > > Userspace can read/write the register with the RDPKRU and WRPKRU > instructions. But, the register is saved and restored with the > XSAVE family of instructions, which means we have to treat it > like a floating point register. > > The kernel needs to write to the register if it wants to > implement execute-only memory or if it implements a system call > to change PKRU. > > To do this, we need to create a 'pkru_state' buffer, read the old > contents in to it, modify it, and then tell the FPU code that > there is modified data in there so it can (possibly) move the > buffer back in to the registers. > > This uses the fpu__xfeature_set_state() function that we defined > in the previous patch. > > Signed-off-by: Dave Hansen Reviewed-by: Thomas Gleixner