From: "Kirill A. Shutemov" Subject: Re: [PATCH v4 21/22] Add support for pmd_faults Date: Mon, 23 Dec 2013 20:54:33 +0200 Message-ID: <20131223185433.GA18067@node.dhcp.inet.fi> References: <20131223134113.GA14806@node.dhcp.inet.fi> <20131223145031.GB11091@parisc-linux.org> <20131223151003.GA15744@node.dhcp.inet.fi> <20131223184222.GE11091@parisc-linux.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Matthew Wilcox , linux-fsdevel@vger.kernel.org, linux-ext4@vger.kernel.org, linux-mm@kvack.org, Andrea Arcangeli To: Matthew Wilcox Return-path: Received: from mta-out.inet.fi ([195.156.147.13]:50546 "EHLO jenni1.inet.fi" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753085Ab3LWSyp (ORCPT ); Mon, 23 Dec 2013 13:54:45 -0500 Content-Disposition: inline In-Reply-To: <20131223184222.GE11091@parisc-linux.org> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Mon, Dec 23, 2013 at 11:42:22AM -0700, Matthew Wilcox wrote: > > Do you know anyone who relay on SIGBUS for correctness? > > Oh, I remember the real reason now. If we install a PMD that hangs off > the end of the file then by reading past i_size, we can read the blocks of > whatever happens to be in storage after the end of the file, which could > be another file's data. This doesn't happen for the PTE case because the > existing code only works for filesystems with a block size == PAGE_SIZE. I see. It's valid reason. Probably, it's better to add comment there. -- Kirill A. Shutemov