From mboxrd@z Thu Jan 1 00:00:00 1970 From: Theodore Ts'o Subject: Re: [PATCH] mke2fs: handle flex_bg collision with backup descriptors Date: Sat, 5 Jul 2014 22:11:24 -0400 Message-ID: <20140706021124.GE19036@thunk.org> References: <1393618545-29319-1-git-send-email-adilger@dilger.ca> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-ext4@vger.kernel.org To: Andreas Dilger Return-path: Received: from imap.thunk.org ([74.207.234.97]:47171 "EHLO imap.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751347AbaGFCL2 (ORCPT ); Sat, 5 Jul 2014 22:11:28 -0400 Content-Disposition: inline In-Reply-To: <1393618545-29319-1-git-send-email-adilger@dilger.ca> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Fri, Feb 28, 2014 at 01:15:45PM -0700, Andreas Dilger wrote: > If a large flex_bg factor is specified and the block allocator was > laying out block or inode bitmaps or inode tables, and collides with > previously allocated metadata (for example the backup superblock or > group descriptors) it would reset the allocator back to the beginning > of the flex_bg instead of continuing past the obstruction. > > For example, with "-G 131072" the inode table will hit the backup > descriptors in groups 1, 3, 5, 7, 9 and start interleaving with the > block and inode bitmaps. That results in poorly allocated bitmaps > and inode tables that are interleaved and not contiguous as was > intended for flex_bg: > > Group 0: (Blocks 0-32767) > Primary superblock at 0, Group descriptors at 1-2048 > Block bitmap 2049 (+2049), Inode bitmap at 133121 (bg #4+2049) > Inode table 264193-264200 (bg #8+2049) > : > : > Group 3838: (Blocks 125763584-125796351) [INODE_UNINIT, BLOCK_UNINIT] > Block bitmap 5887 (bg #0+5887), Inode bitmap 136959 (bg #4+5887) > Inode table 294897-294904 (bg #8 + 32753) > Group 3839: (Blocks 125796352-125829119) [INODE_UNINIT, BLOCK_UNINIT] > Block bitmap 5888 (bg #0+5888), Inode bitmap 136960 (bg #4+5888) > Inode table 5889-5896 (bg #0 + 5889) > Group 3840: (Blocks 125829120-125861887) [INODE_UNINIT, BLOCK_UNINIT] > Block bitmap 5897 (bg #0+5897), Inode bitmap 136961 (bg #4+5889) > Inode table 5898-5905 (bg #0 + 5898) > : > : > > Instead, skip the intervening blocks if there aren't too many of them. > That mostly keeps the flex_bg allocations from colliding, though still > not perfect because there is still some overlap with the backups. > This patch addresses the majority of the problem, allowing about 124k > groups to be layed out perfectly, instead of less than 4k groups with > the previous code. > > Signed-off-by: Andreas Dilger Thanks, applied. - Ted