From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from cn.fujitsu.com ([59.151.112.132]:6113 "EHLO heian.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-FAIL) by vger.kernel.org with ESMTP id S1752611AbbF3Bvq (ORCPT ); Mon, 29 Jun 2015 21:51:46 -0400 Message-ID: <5591F61F.7060009@cn.fujitsu.com> Date: Tue, 30 Jun 2015 09:51:27 +0800 From: wangyf MIME-Version: 1.0 Subject: Re: [PATCH] btrfs: add a test of replace missing dev in diff raid References: <1435296063-10695-1-git-send-email-wangyf-fnst@cn.fujitsu.com> <20150626222409.GA27386@huxley.DHCP.TheFacebook.com> <55910C42.7030707@cn.fujitsu.com> <5591893B.2020309@fb.com> In-Reply-To: <5591893B.2020309@fb.com> Content-Type: text/plain; charset="windows-1252"; format=flowed Content-Transfer-Encoding: 7bit Sender: fstests-owner@vger.kernel.org To: Omar Sandoval Cc: fstests@vger.kernel.org List-ID: On 06/30/2015 02:06 AM, Omar Sandoval wrote: > On 06/29/2015 02:13 AM, wangyf wrote: >> Hello, Omar: >> >> To update some cases to support RAID5/6 is necessary. >> Like btrfs/011 btrfs/071 , and the man page of BTRFS-REPLACE, >> they all should be modified. >> >> But equipment damage and missing device is small probability event when >> we use disks in our daily life. So it shouldn't be so important that we let >> _btrfs_get_profile_confilgs know it. >> Besides, every case tests a different direction of replace is good. e.g. >> 011 071 020 etc. >> >> So I think to test it in a new case is better. What is your opinion? >> >> cheers, >> wangyf > Hi, Yanfeng, > > Sorry, I think I might have been unclear. I definitely agree that the > missing case should be tested in a separate test case. However, I do > think that _btrfs_get_profile_configs should be updated so we can keep > things in one place and make sure that all profiles are thoroughly > tested. With _btrfs_get_profile_configs aware of RAID 5/6 and missing > device replacement, the new test case can just make use of that instead > of hardcoding the profiles to test. Does that sound alright? > > Thanks! Hi, Omar: I think it again, and now I think you are right. To let _btrfs_get_profile_confilgs be aware of replace+missing many redundant codes can be evitable. We can use it in a new test case to test missing device. regards.