From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753488AbbHMQmg (ORCPT ); Thu, 13 Aug 2015 12:42:36 -0400 Received: from g2t2353.austin.hp.com ([15.217.128.52]:6825 "EHLO g2t2353.austin.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753403AbbHMQme (ORCPT ); Thu, 13 Aug 2015 12:42:34 -0400 Subject: Re: regression introduced by "block: Add support for DAX reads/writes to block devices" To: Jeff Moyer , Boaz Harrosh References: <100D68C7BA14664A8938383216E40DE04091408C@FMSMSX114.amr.corp.intel.com> <100D68C7BA14664A8938383216E40DE0409144D9@FMSMSX114.amr.corp.intel.com> <55C855D5.1070001@plexistor.com> <55CC2BDA.3080906@plexistor.com> Cc: "Wilcox, Matthew R" , "linux-kernel@vger.kernel.org" , "linux-fsdevel@vger.kernel.org" From: Linda Knippers X-Enigmail-Draft-Status: N1110 Message-ID: <55CCC8F8.6080204@hp.com> Date: Thu, 13 Aug 2015 12:42:32 -0400 User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:38.0) Gecko/20100101 Thunderbird/38.1.0 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 8/13/2015 10:00 AM, Jeff Moyer wrote: > Boaz Harrosh writes: > >> On 08/13/2015 12:11 AM, Jeff Moyer wrote: >>> Boaz Harrosh writes: >>> >>>> On 08/07/2015 11:41 PM, Jeff Moyer wrote: >>>> <> >>>>> >>>>>> We need to cope with the case where the end of a partition isn't on a >>>>>> page boundary though. >>>>> >>>>> Well, that's usually done by falling back to buffered I/O. I gave that >>>>> a try and panicked the box. :) I'll keep looking into it, but probably >>>>> won't have another patch until next week. >>>>> >>>> >>>> lets slow down for a sec, please. >>>> >>>> We have all established that an unaligned partition start is BAD and not supported? >>> >>> No. Unaligned partitions on RAID arrays or 512e devices are bad because >>> they result in suboptimal performance. They are most certainly still >>> supported, though. >>> >> >> What ? >> >> I meant for dax on pmem or brd. I meant that we *do not* support dax access >> on an unaligned partition start. (None dax is perfectly supported) > > Sorry, I thought your statement was broader than that. > >> We did it this way because of the direct_access API that returns a pfn >> with is PAGE_SIZE. We could introduce a pfn+offset but we thought it is >> not worth it, and that dax should be page aligned for code simplicity > > I'd be fine with changing the persistent memory block device to only > support 4k logical, 4k physical block size. That probably makes the > most sense. If that's what we want, the current patch doesn't do that. https://lists.01.org/pipermail/linux-nvdimm/2015-July/001555.html It causes the physical block size to be PAGE_SIZE but the logical block size is still 512. However, the minimum_io_size is now 4096 (same as physical block size, I assume). The optimal_io_size is still 0. What does that mean? Whatever we go with, we should do something because 4.2rc6 is still broken, unable to create a xfs file system on a pmem device, ever since the change to use DAX on block devices with O_DIRECT. -- ljk > > Cheers, > Jeff >