From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:60498) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Uz1HT-0004Zd-Lk for qemu-devel@nongnu.org; Tue, 16 Jul 2013 05:10:52 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1Uz1HR-0001qU-Tb for qemu-devel@nongnu.org; Tue, 16 Jul 2013 05:10:47 -0400 Received: from mail-pd0-f169.google.com ([209.85.192.169]:33368) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Uz1HR-0001lN-GF for qemu-devel@nongnu.org; Tue, 16 Jul 2013 05:10:45 -0400 Received: by mail-pd0-f169.google.com with SMTP id y10so471591pdj.28 for ; Tue, 16 Jul 2013 02:10:41 -0700 (PDT) Message-ID: <51E50E01.5080807@ozlabs.ru> Date: Tue, 16 Jul 2013 19:10:25 +1000 From: Alexey Kardashevskiy MIME-Version: 1.0 References: <1373951995-9866-1-git-send-email-aik@ozlabs.ru> <51E4E636.8010603@redhat.com> <0B82AB0E-3434-4EFE-A3A4-41CAF169B7F7@suse.de> <1373964320.19894.381.camel@pasglop> In-Reply-To: Content-Type: text/plain; charset=KOI8-R Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH] spapr-pci: remove io ports workaround List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Alexander Graf Cc: Anthony Liguori , "qemu-devel@nongnu.org" , "qemu-ppc@nongnu.org" , Paolo Bonzini , Paul Mackerras , David Gibson On 07/16/2013 07:01 PM, Alexander Graf wrote: > > > Am 16.07.2013 um 10:45 schrieb Benjamin Herrenschmidt > : > >> On Tue, 2013-07-16 at 10:32 +0200, Alexander Graf wrote: >>>>> Did I miss anything here? >>>> >>>> No, I don't think so. The patch looks good. >>> >>> ... and will break all Mac targets again, no? Not to speak of >>> non-ppc devices. >> >> Do you know why it breaks ? Can you suggest what's wrong and how to >> fix it maybe ? > > Sure. Get rid of cpu_inX and cpu_outX completely and verify that the > conversion is doing the "right thing" for each conversion. Some of them > are really tricky - like MIPS which accesses the RTC in native endian > (big) in current code. So this patch would definitely break that one > too. > > I'm actually surprised we have to mark the ioport region little endian > for sPAPR. It is not for spapr, it is for devices which expect some endianness and do not do convertions. > We already did have it converted to a direct memory api alias > a while ago and IIRC the only problem back then were the old_portio > callbacks. I wonder what changed in between. ? We have had the IO memory region hack I am trying to remove for a long time. -- Alexey