From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ozlabs.org ([203.10.76.45]:41701 "EHLO ozlabs.org") by vger.kernel.org with ESMTP id S267470AbUIKGp3 (ORCPT ); Sat, 11 Sep 2004 02:45:29 -0400 Date: Sat, 11 Sep 2004 16:42:05 +1000 From: Anton Blanchard Subject: Re: RFC: being more anal about iospace accesses.. Message-ID: <20040911064205.GE6005@krispykreme> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: To: Linus Torvalds Cc: Al Viro , Andrew Morton , Linux Arch list , Alan Cox , "David S. Miller" , Jeff Garzik , Benjamin Herrenschmidt List-ID: > Benh: while doign the ppc64 thing, I also encountered something clearly > broken in arch/ppc64/mm/init.c: iounmap_explicit(). I just commented out > the stupid thing (it was wrong regardless of which way the preceding > branch went: it would either call iounmap() twice on the same area, or it > would get a NULL pointer dereference). Can you check what that is > _supposed_ to be doing? Yuck, that change was made in preparation for hotplug removal of host bridges and is obviously broken. BTW One of our problems is how drivers can tap IO addresses without doing something to map them first. We end up doing the ioremap of IO space during PCI init. The problem comes when we have to remove a host bridge from a partition and must have all IO and memory resources unmapped. Memory resources take care of themselves since drivers iounmap those resources. IO resources do not. Anton