From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: [PATCH/RFC 2] kvm: fix module refcount issues with anon_inodegetfd Date: Tue, 25 Nov 2008 15:55:57 +0200 Message-ID: <492C03ED.1090808@redhat.com> References: <200811202010.00125.borntraeger@de.ibm.com> <49295F6D.8010200@redhat.com> <200811250907.54042.borntraeger@de.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: kvm@vger.kernel.org, linux-fsdevel@vger.kernel.org To: Christian Borntraeger Return-path: Received: from mx2.redhat.com ([66.187.237.31]:38138 "EHLO mx2.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751443AbYKYN4E (ORCPT ); Tue, 25 Nov 2008 08:56:04 -0500 In-Reply-To: <200811250907.54042.borntraeger@de.ibm.com> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: Christian Borntraeger wrote: >> Messing with module counts is slightly ugly. >> > > I changed my patch to be more generic. Instead of messing with the > module refcount in the kvm module, I decided to change > anon_inode_getfd instead. If the owner field is set, we will > try an module_get in anon_inode_getfd, since the VFS release > function will do a module_put. This makes the open/release > symmetric. What do you think? > I agree with your analysis, and also that the anon_inodes change is useful. If it's acceptable to the vfs/anon_inode maintainers, I'll apply the patch. -- error compiling committee.c: too many arguments to function