From: Andi Kleen Subject: Re: [PATCH] vfs: Turn the unlinked inode counter into a percpu counter Date: Mon, 14 Jul 2014 08:54:07 -0700 Message-ID: <87wqbgszq8.fsf@tassilo.jf.intel.com> References: <1405110219-2969-1-git-send-email-andi@firstfloor.org> <20140714154930.GB5803@tassilo.jf.intel.com> Mime-Version: 1.0 Content-Type: text/plain Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, Al Viro , Miklos Szeredi To: =?utf-8?B?THVrw6HFoQ==?= Czerner Return-path: Received: from mga14.intel.com ([192.55.52.115]:20507 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755783AbaGNPyJ (ORCPT ); Mon, 14 Jul 2014 11:54:09 -0400 In-Reply-To: <20140714154930.GB5803@tassilo.jf.intel.com> (Andi Kleen's message of "Mon, 14 Jul 2014 08:49:30 -0700") Sender: linux-fsdevel-owner@vger.kernel.org List-ID: Andi Kleen writes: > > fsnotify_inode_delete(inode); >> > - if (!inode->i_nlink) { >> > - WARN_ON(atomic_long_read(&inode->i_sb->s_remove_count) == 0); >> >> This is not mentioned in the description, but I guess it makes sense since >> it'd be quite expensive. > > Yes. I guess we could still check the local counter. I can re-add that. Actually it's not possible, because the original count could be on another CPU. -Andi -- ak@linux.intel.com -- Speaking for myself only