From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx1.redhat.com ([209.132.183.28]:19914 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751228AbdKUSfb (ORCPT ); Tue, 21 Nov 2017 13:35:31 -0500 Date: Tue, 21 Nov 2017 19:35:29 +0100 From: Andrea Arcangeli To: Al Viro Cc: Linux Kernel Mailing List , linux-fsdevel , Linus Torvalds Subject: Re: [git pull] vfs.git get_user_pages_fast() conversion Message-ID: <20171121183529.GA6349@redhat.com> References: <20171117030205.GL21978@ZenIV.linux.org.uk> <20171117213215.GQ21978@ZenIV.linux.org.uk> <20171118214531.GT21978@ZenIV.linux.org.uk> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20171118214531.GT21978@ZenIV.linux.org.uk> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Sat, Nov 18, 2017 at 09:45:31PM +0000, Al Viro wrote: > in __get_user_pages_locked(), or am I missing something subtle there? Andrea? You're not missing anything as far as the logic is concerned. However see the __always_inline, I added "notify_drop" purely to optimize away such branch at build time so that gcc doesn't need to include in every different copy it does of it. "notify_drop" is known at build time and constant true/false, the other variables are not. Either you drop the __always_inline or "notify_drop" makes sense to me to keep to reduce the size of the inline and run faster at runtime. Thanks, Andrea