From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 14 Jan 2003 10:33:04 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 14 Jan 2003 10:33:04 -0500 Received: from host194.steeleye.com ([66.206.164.34]:65031 "EHLO pogo.mtv1.steeleye.com") by vger.kernel.org with ESMTP id ; Tue, 14 Jan 2003 10:33:03 -0500 Message-Id: <200301141541.h0EFfpA02417@localhost.localdomain> X-Mailer: exmh version 2.4 06/23/2000 with nmh-1.0.4 To: "David S. Miller" Cc: James.Bottomley@SteelEye.com, linux-kernel@vger.kernel.org Subject: Re: [BK-2.5] Update arm implementation of DMA API to include GFP_ Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Tue, 14 Jan 2003 10:41:51 -0500 From: James Bottomley X-AntiVirus: scanned for viruses by AMaViS 0.2.1 (http://amavis.org/) Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org James.Bottomley@SteelEye.com said: > Is this really safe? Maybe ARM needs to use GFP_ATOMIC all the time > for a specific reason, such as where and how it maps the cpu side > mappings of the memory? According to Russell King, yes. The PA-RISC one should also be correct. Actually, to be honest, there's a longstanding issue in the pa-risc code where we can potentially allocate a page table using GFP_KERNEL in pci_alloc_consistent() when we obtain the mapping resources. I can argue that we never go down this path in practice, but it does make GFP_ATOMIC allocations look unsafe on the platform. James