From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail106.syd.optusnet.com.au (mail106.syd.optusnet.com.au [211.29.132.42]) by ml01.01.org (Postfix) with ESMTP id D2A2721296B30 for ; Thu, 13 Jun 2019 20:08:11 -0700 (PDT) Date: Fri, 14 Jun 2019 13:07:12 +1000 From: Dave Chinner Subject: Re: [PATCH RFC 00/10] RDMA/FS DAX truncate proposal Message-ID: <20190614030712.GO14363@dread.disaster.area> References: <20190607110426.GB12765@quack2.suse.cz> <20190607182534.GC14559@iweiny-DESK2.sc.intel.com> <20190608001036.GF14308@dread.disaster.area> <20190612123751.GD32656@bombadil.infradead.org> <20190613002555.GH14363@dread.disaster.area> <20190613152755.GI32656@bombadil.infradead.org> <20190613211321.GC32404@iweiny-DESK2.sc.intel.com> <20190613234530.GK22901@ziepe.ca> <20190614020921.GM14363@dread.disaster.area> <20190614023107.GK32656@bombadil.infradead.org> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <20190614023107.GK32656@bombadil.infradead.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Errors-To: linux-nvdimm-bounces@lists.01.org Sender: "Linux-nvdimm" To: Matthew Wilcox Cc: Theodore Ts'o , linux-nvdimm@lists.01.org, linux-rdma@vger.kernel.org, John Hubbard , Jeff Layton , linux-kernel@vger.kernel.org, linux-xfs@vger.kernel.org, Jason Gunthorpe , =?iso-8859-1?B?Suly9G1l?= Glisse , linux-mm@kvack.org, linux-fsdevel@vger.kernel.org, Jan Kara , linux-ext4@vger.kernel.org, Andrew Morton List-ID: On Thu, Jun 13, 2019 at 07:31:07PM -0700, Matthew Wilcox wrote: > On Fri, Jun 14, 2019 at 12:09:21PM +1000, Dave Chinner wrote: > > If the lease holder modifies the mapping in a way that causes it's > > own internal state to screw up, then that's a bug in the lease > > holder application. > > Sounds like the lease semantics aren't the right ones for the longterm > GUP users then. The point of the longterm GUP is so the pages can be > written to, and if the filesystem is going to move the pages around when > they're written to, that just won't work. And now we go full circle back to the constraints we decided on long ago because we can't rely on demand paging RDMA hardware any time soon to do everything we need to transparently support long-term GUP on file-backed mappings. i.e.: RDMA to file backed mappings must first preallocate and write zeros to the range of the file they are mapping so that the filesystem block mapping is complete and static for the life of the RDMA mapping that will pin it. IOWs, the layout lease will tell the RDMA application that the static setup it has already done to work correctly with a file backed mapping may be about to be broken by a third party..... -Dave. -- Dave Chinner david@fromorbit.com _______________________________________________ Linux-nvdimm mailing list Linux-nvdimm@lists.01.org https://lists.01.org/mailman/listinfo/linux-nvdimm