From mboxrd@z Thu Jan 1 00:00:00 1970 From: Qu Wenruo Subject: Re: [PATCH] generic/224: Increase filesystem instance size to 1.5 GiB Date: Fri, 11 Sep 2015 09:38:43 +0800 Message-ID: <55F230A3.2070300@cn.fujitsu.com> References: <1440945981-323-1-git-send-email-chandan@linux.vnet.ibm.com> <160121092.hSkPuGCNnN@localhost.localdomain> <20150901021510.GE7642@thunk.org> <2500094.kBUgEuDpMB@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8"; format=flowed Content-Transfer-Encoding: 7bit Cc: , , , To: Chandan Rajendra , "Theodore Ts'o" Return-path: In-Reply-To: <2500094.kBUgEuDpMB@localhost.localdomain> Sender: fstests-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org Chandan Rajendra wrote on 2015/09/01 08:03 +0530: > On Monday 31 Aug 2015 22:15:10 Theodore Ts'o wrote: >> On Tue, Sep 01, 2015 at 05:49:14AM +0530, Chandan Rajendra wrote: >>> mkfs.btrfs when invoked on small filesystems by "not" specifying any block >>> sizes (i.e. mkfs.btrfs -f /dev/sda1) will automatically create filesystem >>> instance with "data block size" == "metadata block size". However in the >>> subpagesize-blocksize scenario, we need to specify both data and metadata >>> block size on the command line (For e.g. mkfs.btrfs -f -s 4096 -n 16384 >>> /dev/sda1). In this case, Since the user is forcing the block sizes and it >>> is impossible to have mixed block groups with differing data and metadata >>> block sizes, mkfs.btrfs will fail. >> >> Ok, so the issue is that for this particular test configuration, btrfs >> has a minimum file system size. What about changing >> _scratch_mkfs_sized so that if MIN_FS_SIZE is set, the file system >> created will be at least MIN_FS_SIZE in size. >> >> This way it sets the minimum file system size for all tests, not just >> generic/224, and any test configuration, whether it be ext4, xfs, or >> btrfs where the data and metadata block size are the same, don't have >> to take extra time -- only the test configuration of btrfs with >> data_block_size != metadata_block_size. >> > I agree with the approach you have suggested. I will write up a patch and send > it across the mailing list. Thanks Ted. > Hi Chandan, Although the reply may be a little late, but I think it's still better to fix the bug in mkfs.btrfs. The core problem is, why mkfs.btrfs is insisting on make the fs into mixed-bg. Auto detection for small device and make it into mixed-bg should be a "OPTIONAL" feature, not a "mandatory" one. So IMHO, the best method is to disable mixed-bg when nodesize/sectorsize is given and differs and user doesn't force mkfs to create mixed-bg. Thanks, Qu