From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-qg0-f49.google.com (mail-qg0-f49.google.com [209.85.192.49]) by kanga.kvack.org (Postfix) with ESMTP id CC0E06B0255 for ; Mon, 14 Dec 2015 11:44:03 -0500 (EST) Received: by qget30 with SMTP id t30so16315187qge.3 for ; Mon, 14 Dec 2015 08:44:03 -0800 (PST) Received: from mail-qk0-x22f.google.com (mail-qk0-x22f.google.com. [2607:f8b0:400d:c09::22f]) by mx.google.com with ESMTPS id m68si35601821qgm.33.2015.12.14.08.44.02 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Mon, 14 Dec 2015 08:44:03 -0800 (PST) Received: by qkfb125 with SMTP id b125so143570474qkf.2 for ; Mon, 14 Dec 2015 08:44:02 -0800 (PST) MIME-Version: 1.0 In-Reply-To: References: <20151210023708.30368.92962.stgit@dwillia2-desk3.jf.intel.com> Date: Mon, 14 Dec 2015 08:44:02 -0800 Message-ID: Subject: Re: [-mm PATCH v2 00/25] get_user_pages() for dax pte and pmd mappings From: Dan Williams Content-Type: text/plain; charset=UTF-8 Sender: owner-linux-mm@kvack.org List-ID: To: Jeff Moyer Cc: Andrew Morton , Dave Hansen , Toshi Kani , David Airlie , Dave Hansen , Dave Chinner , Linux MM , "H. Peter Anvin" , Christoph Hellwig , Andrea Arcangeli , kbuild test robot , linux-nvdimm , Richard Weinberger , X86 ML , Peter Zijlstra , Ingo Molnar , Mel Gorman , Matthew Wilcox , Ross Zwisler , Jeff Dike , Jens Axboe , Alexander Viro , Thomas Gleixner , Christoffer Dall , Jan Kara , Paolo Bonzini , Logan Gunthorpe , "Kirill A. Shutemov" On Mon, Dec 14, 2015 at 6:52 AM, Jeff Moyer wrote: > Dan Williams writes: > >> In the meantime, I expect some would say DAX is a toy as long as it >> continues to fail at DMA. > > I suppose this is the crux of it. Given that we may be able to migrate > away from the allocation of storage for temporary data structures in the > future, and given that admin tooling could hide the undesirable > configurations, this approach seems workable. Here's my current thoughts on the tooling for namespace creation: ndctl create-namespace That command by default will create a maximally sized pmem namespace and set up the struct page memmap by default i.e. it defaults to the following parameters: ndctl create-namespace --type=pmem --mode=memory The other options for 'mode' are 'safe' and 'raw' where 'safe' establishes a btt, and 'raw' exposes the full un-decorated capacity of the namespace. If you have a pre-existing 'raw' mode pmem namespace, you can convert it to be enabled for dma, and other capabilities typical memory possesses, with the following: ndctl create-namespace --mode=memory -r namespace0.0 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org