From mboxrd@z Thu Jan 1 00:00:00 1970 From: Li Zhong Subject: Re: [RFC PATCH next]module: Fix mod->mkobj.kobj potentially freed too early Date: Thu, 22 Aug 2013 13:37:58 +0800 Message-ID: <1377149878.2633.40.camel@ThinkPad-T5421> References: <1377078598.2709.25.camel@ThinkPad-T5421> <20130821161819.GA14364@kroah.com> <1377138846.2633.25.camel@ThinkPad-T5421> <20130822040333.GB4821@kroah.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Return-path: Received: from e23smtp04.au.ibm.com ([202.81.31.146]:34183 "EHLO e23smtp04.au.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752527Ab3HVFiI (ORCPT ); Thu, 22 Aug 2013 01:38:08 -0400 Received: from /spool/local by e23smtp04.au.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Thu, 22 Aug 2013 15:21:08 +1000 In-Reply-To: <20130822040333.GB4821@kroah.com> Sender: linux-next-owner@vger.kernel.org List-ID: To: Greg KH Cc: linux-next list , rusty@rustcorp.com.au, LKML , rmk+kernel@arm.linux.org.uk On Wed, 2013-08-21 at 21:03 -0700, Greg KH wrote: > On Thu, Aug 22, 2013 at 10:34:06AM +0800, Li Zhong wrote: > > On Wed, 2013-08-21 at 09:18 -0700, Greg KH wrote: > > > On Wed, Aug 21, 2013 at 05:49:58PM +0800, Li Zhong wrote: > > > > DEBUG_KOBJECT_RELEASE helps to find the issue attached below. > > > > > > > > After some investigation, it seems the reason is: > > > > The mod->mkobj.kobj(ffffffffa01600d0 below) is freed together with mod > > > > itself in free_module(). However, its children still hold references to > > > > it, as the delay caused by DEBUG_KOBJECT_RELEASE. So when the > > > > child(holders below) tries to decrease the reference count to its parent > > > > in kobject_del(), BUG happens as it tries to access already freed memory. > > > > > > Ah, thanks for tracking this down. I had seen this in my local testing, > > > but wasn't able to figure out the offending code. > > > > > > > This patch tries to fix it by waiting for the mod->mkobj.kobj to be > > > > really released in the module removing process (and some error code > > > > paths). > > > > > > Nasty, we should just be freeing the structure in the release function, > > > why doesn't that work? > > > > Hi Greg, > > > > It seems I didn't describe it clearly in the previous mail. I'm trying > > to do it better below: > > > > mod->mkobj.kobj is embedded in module_kobject(not a pointer): > > struct module_kobject { > > struct kobject kobj; > > ... > > > > and allocated with the module memory. So we could see the parent below > > ffffffffa01600d0 is between MODULES_VADDR (ffffffffa0000000) and > > MODULES_END(ffffffffff000000). > > > > It seem to me that the mkobj.kobj is freed by module_free(mod, > > mod->module_core). > > Ick, you are right. If a kobject is being embedded in an object, it > should control the lifespan of the object, not somewhere else like is > happening here. > > The best solution for this is to make the kobject a pointer, not > embedded in the structure, that will fix this issue, right? Yes, I think so. I'll try to write a fix using this way, thanks for your suggestion. Thanks, Zhong > > thanks, > > greg k-h >