From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thadeu Lima de Souza Cascardo Date: Tue, 11 May 2021 12:28:10 -0300 Subject: [LTP] [PATCH] mkfs: force block size to 1024 for ext3 and ext4 In-Reply-To: References: <20210510134739.37512-1-cascardo@canonical.com> <20210511103622.GC12149@mussarela> Message-ID: <20210511152810.GF12149@mussarela> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ltp@lists.linux.it On Tue, May 11, 2021 at 04:19:19PM +0200, Petr Vorel wrote: > Hi, > > > On Tue, May 11, 2021 at 08:19:01AM +0200, Petr Vorel wrote: > > > Hi, > > > > > ext3 and ext4 filesystems will reserve at least 1024 blocks for the > > > > journal. With a blocksize of 4096, this will be 25% of the filesystem size > > > > without accounting for any other overhead. > > > Is that any actual problem? > > > It causes the test to fail. > > mkfs01 2 TFAIL: 'mkfs -t ext4 /dev/loop0 16000' failed, unexpected size. > > > The filesystem size is very small, 16K 1k blocks, and we test that there are at > > least 80% of that available. As I said, the journal takes at least 1024 blocks, > > and with 4k blocks, that is too much overhead. > > Have you checked if d44387457 ("mkfs: relax size check") fixes the issue? > > https://github.com/linux-test-project/ltp/commit/d4438745751dc4b0faa50063f9a0d9542f0ceaac > > Kind regards, > Petr > It does not. That would have to move to something below 80%. Cascardo. > > > > /etc/mke2fs.conf will use 1024 block size for small filesystems, which are > > > > between 3M and 512M. However, on recent versions of Ubuntu, this > > > > configuration has changed and thet default blocksize is 4096 even for small > > > > filesystems. > > > > > Force the blocksize to 1024 on ext3 and ext4 filesystems, which will lead > > > > to the expected results, as journals will take only 1M. > > > > IMHO it'd be better to keep the default, because that covers what end users > > > actually use. > > > One alternative to forcing the block size is accouting for the journal blocks, > > but, then, that needs to consider the block size. I think my approach is more > > simple. We could restrict it to the smaller 16M filesystem, though. > > > What do you think? > > > Thanks. > > Cascardo.