Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759945AbYBFUf0 (ORCPT ); Wed, 6 Feb 2008 15:35:26 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757510AbYBFUfJ (ORCPT ); Wed, 6 Feb 2008 15:35:09 -0500 Received: from extu-mxob-2.symantec.com ([216.10.194.135]:53587 "EHLO extu-mxob-2.symantec.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1759564AbYBFUfG (ORCPT ); Wed, 6 Feb 2008 15:35:06 -0500 Date: Wed, 6 Feb 2008 20:33:07 +0000 (GMT) From: Hugh Dickins X-X-Sender: hugh@blonde.site To: Oleg Nesterov cc: Matt Helsley , Miklos Szeredi , Andrew Morton , Peter Zijlstra , William Lee Irwin III , Nick Piggin , Ingo Molnar , linux-kernel@vger.kernel.org, linux-mm Subject: Re: [PATCH] sys_remap_file_pages: fix ->vm_file accounting In-Reply-To: <20080203182135.GA5827@tv-sign.ru> Message-ID: References: <20080130142014.GA2164@tv-sign.ru> <1201712101.31222.22.camel@tucsk.pomaz.szeredi.hu> <20080130172646.GA2355@tv-sign.ru> <1201987065.9062.6.camel@localhost.localdomain> <20080203182135.GA5827@tv-sign.ru> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1469 Lines: 31 On Sun, 3 Feb 2008, Oleg Nesterov wrote: > > So I have to try to find another bug ;) Suppose that ->load_binary() does > a series of do_mmap(MAP_EXECUTABLE). It is possible that mmap_region() can > merge 2 vmas. In that case we "leak" ->num_exe_file_vmas. Unless I missed > something, mmap_region() should do removed_exe_file_vma() when vma_merge() > succeds (near fput(file)). Or there's the complementary case of a VM_EXECUTABLE vma being split in two, for example by an mprotect of a part of it. Sorry, Matt, I don't like your patch at all. It seems to add a fair amount of ugliness and unmaintainablity, all for a peculiar MVFS case (you've tried to argue other advantages, but not always convinced!). And I found it quite hard to see where the crucial difference comes. I guess it's that MVFS changes vma->vm_file in its ->mmap? Well, if MVFS does that, maybe something else does that too, but precisely to rely on the present behaviour of /proc/pid/exe - so in fixing for MVFS, we'd be breaking that hypothetical other? I can understand patches to avoid mmap_sem for /proc/pid/exe, but this one just seems too messy for too special an out-of-tree case. (I've no last word on this, but that's my opinion.) Hugh -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/