Return-Path: Received: from userp1040.oracle.com ([156.151.31.81]:39379 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750937AbcBVSfL (ORCPT ); Mon, 22 Feb 2016 13:35:11 -0500 Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 (Mac OS X Mail 9.2 \(3112\)) Subject: Re: [PATCH v3 01/10] svcrdma: Do not write xdr_buf::tail in a Write chunk From: Chuck Lever In-Reply-To: <20160222183211.GA22270@fieldses.org> Date: Mon, 22 Feb 2016 13:35:02 -0500 Cc: linux-rdma@vger.kernel.org, Linux NFS Mailing List Message-Id: References: <20160218173840.12825.26769.stgit@klimt.1015granger.net> <20160218174353.12825.14732.stgit@klimt.1015granger.net> <20160222183211.GA22270@fieldses.org> To: "J. Bruce Fields" Sender: linux-nfs-owner@vger.kernel.org List-ID: > On Feb 22, 2016, at 1:32 PM, bfields@fieldses.org wrote: > > On Thu, Feb 18, 2016 at 12:43:53PM -0500, Chuck Lever wrote: >> The fix is to send only the contents of the xdr_buf's page list in >> a Write chunk. > > This isn't quite right either. A large-enough non-readlike reply will > use the page list if that's the only way to fit the reply. > > You might be able to hit such cases, for example, using nfs4_getfacl on > a file with a sufficiently large ACL. Clients provide a write chunk only for READ and READLINK. No way to hit this with GETFACL unless the client is doing something wrong. Anyway, I've got another revision of this work coming, this time it should be a complete fix. Thanks as always for your careful review! -- Chuck Lever