From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263618AbTJCCyi (ORCPT ); Thu, 2 Oct 2003 22:54:38 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263619AbTJCCyi (ORCPT ); Thu, 2 Oct 2003 22:54:38 -0400 Received: from mail.jlokier.co.uk ([81.29.64.88]:43143 "EHLO mail.shareable.org") by vger.kernel.org with ESMTP id S263618AbTJCCyh (ORCPT ); Thu, 2 Oct 2003 22:54:37 -0400 Date: Fri, 3 Oct 2003 03:53:38 +0100 From: Jamie Lokier To: Linus Torvalds Cc: Albert Cahalan , Ulrich Drepper , Mikael Pettersson , Kernel Mailing List Subject: Re: Who changed /proc// in 2.6.0-test5-bk9? Message-ID: <20031003025338.GA15089@mail.shareable.org> References: <1065139380.736.109.camel@cube> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.1i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Linus Torvalds wrote: > And it may or may not make sense to not have a "/proc//task//fd" > directory at all if the thread shares file descriptors with the thread > group leader. That would be a fairly easy optimization. If you make /proc/.../fd return the same inode numbers for the identical file descriptor tables, that would be just as good for fuser performance. fuser would stat() each of the /proc//task//fd entries and see that they're the same as the group leader. It is the same number of operations as checking for the non-existence of /proc//task//fd entries, but more flexible. Alternatively you could make the task fd directories be symbolic links to the group leader fd directory. -- Jamie