From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-qk0-f195.google.com ([209.85.220.195]:38905 "EHLO mail-qk0-f195.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754780AbdJIPcK (ORCPT ); Mon, 9 Oct 2017 11:32:10 -0400 Received: by mail-qk0-f195.google.com with SMTP id 17so9633967qkq.5 for ; Mon, 09 Oct 2017 08:32:10 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <150174646416.104003.14042713459553361884.stgit@hn> References: <150174646416.104003.14042713459553361884.stgit@hn> From: Miklos Szeredi Date: Mon, 9 Oct 2017 17:32:07 +0200 Message-ID: Subject: Re: [RFC 00/16] NOVA: a new file system for persistent memory To: Steven Swanson Cc: linux-fsdevel , LKML , linux-nvdimm@lists.01.org, Steven Swanson , dan.j.williams@intel.com, Steven Whitehouse Content-Type: text/plain; charset="UTF-8" Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Thu, Aug 3, 2017 at 9:48 AM, Steven Swanson wrote: > This is an RFC patch series that impements NOVA (NOn-Volatile memory > Accelerated file system), a new file system built for PMEM. Hi, Thanks for posting. I read the paper and the design looks nice. Then I looked at the patches, but could not find a place to start, nor something I could actually try out. So let me suggest some ways to make this more reviewer/tester friendly: 1) try starting with something very simple yet working and supporting the final layout - no optimizations (one big lock, no per-cpu data, rcu, numa, etc support) - no support for optional features (checksumming, NFS export, etc) - missing mandatory features (e.g. just readdir and getattr support) - try and get it down to <5k lines, preferably 2-3k 2) pointer to sources and instructions for trying it out without special hardware 3) build on this minimal working version by - adding mandatory features - then adding optimizations 4) each patch should leave the tree in a compiling and working state but should be small and easily reviewed 5) leave optional features and unimportant optimizations for a later submission; try to make the patchset as small as you meaningfully can (i.e. it should be fully working and demonstrate the capabilities and performance, but nothing more). Thanks, Miklos