From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jens Axboe Subject: Re: [PATCH] fs/splice: Pull buf->ops->confirm() from splice_from_pipe actors Date: Tue, 14 Dec 2010 21:12:55 +0100 Message-ID: <4D07CFC7.6050209@fusionio.com> References: <17eaa416d82b1c57f70443e56ad22feb3c3a3a6f.1292184347.git.mirq-linux@rere.qmqm.pl> <4D0621CB.2010809@fusionio.com> <20101213150450.GA12865@rere.qmqm.pl> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-2 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Alexander Viro , "linux-fsdevel@vger.kernel.org" , "linux-kernel@vger.kernel.org" , "J. Bruce Fields" , Neil Brown To: =?ISO-8859-2?Q?Micha=B3_Miros=B3aw?= Return-path: In-Reply-To: <20101213150450.GA12865@rere.qmqm.pl> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On 2010-12-13 16:04, Micha=B3 Miros=B3aw wrote: > On Mon, Dec 13, 2010 at 02:38:19PM +0100, Jens Axboe wrote: >> On 2010-12-12 22:23, Micha=B3 Miros=B3aw wrote: >>> This patch pulls calls to buf->ops->confirm() from all actors passe= d >>> (also indirectly) to splice_from_pipe_feed(). >> Why? The point of ->confirm() is to ensure that the contents are >> stable, otherwise the pages in the pipe could merely be in flight. >> It's needed if you need to actually look at the data, rather than ju= st >> reference it. >=20 > I should have put this more clearly in the patch description: > the ->confirm() call is moved to splice_from_pipe_feed(), so that eve= ry > actor has its data guaranteed to be stable before it runs. OK, that makes more sense. I'll queue it up. --=20 Jens Axboe