From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1760947Ab2C3Pfu (ORCPT ); Fri, 30 Mar 2012 11:35:50 -0400 Received: from cantor2.suse.de ([195.135.220.15]:43538 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752491Ab2C3Pfm (ORCPT ); Fri, 30 Mar 2012 11:35:42 -0400 Date: Fri, 30 Mar 2012 17:35:30 +0200 From: Jan Kara To: Artem Bityutskiy Cc: Jan Kara , Ted Tso , Ext4 Mailing List , Linux FS Maling List , Linux Kernel Maling List Subject: Re: [PATCH v1 0/9] do not use s_dirt in ext4 Message-ID: <20120330153530.GH5587@quack.suse.cz> References: <1332254489-2300-1-git-send-email-dedekind1@gmail.com> <20120322095342.GC14485@quack.suse.cz> <1332410747.18717.12.camel@sauron.fi.intel.com> <20120322103309.GA14484@quack.suse.cz> <1332854998.31549.40.camel@sauron.fi.intel.com> <20120327201401.GF5020@quack.suse.cz> <1333121035.5440.49.camel@sauron.fi.intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1333121035.5440.49.camel@sauron.fi.intel.com> User-Agent: Mutt/1.5.20 (2009-06-14) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri 30-03-12 18:23:55, Artem Bityutskiy wrote: > On Tue, 2012-03-27 at 22:14 +0200, Jan Kara wrote: > > On Tue 27-03-12 16:29:58, Artem Bityutskiy wrote: > > > On Thu, 2012-03-22 at 11:33 +0100, Jan Kara wrote: > > > > Then we have ext4_mark_super_dirty() call from 4 places - I forgot about > > > > these originally... I kind of miss their purpose. Originally they were used > > > > so that we write total number of free blocks and inodes in the superblock > > > > but when we do not maintain them in the journal mode I don't see a reason > > > > to periodically sync them in no-journal mode. Ted, what is the purpose of > > > > these calls? > > > > > > I do not understand what's the fundamental difference between journal > > > and non-journal mode. Why when we do have the journal we do not mark the > > > super-block as dirty in many places (e.g., in 'ext4_file_open()' - if we > > > do have the journal, when do we make sure we save the mount point path > > > change?). > > We write it at least during ext4_put_super(). > > > > > May be it has something to do with behaving like the ext2 driver when > > > mounting ext2-formatted media with the the ext4 driver? > > I'm not really sure about this... > > > > > Jan, since Ted did not answer, may be you can figure out the reasons > > > from this commit message, which actually introduced the > > > 'ext4_mark_super_dirty()' function? > > Anyway, attached are two patches which you can include in your series > > and which should make your cleanups simpler. > > I amended the second patch: > - ext4_journal_stop(sb); > + ext4_journal_stop(sbi->s_sbh); It should have been: ext4_journal_stop(handle); > Extensive testing with xfstests found a problem with these patches: > > [23500.238579] ------------[ cut here ]------------ > [23500.238720] kernel BUG at fs/buffer.c:2871! > [23500.238842] invalid opcode: 0000 [#1] SMP > [23500.239279] CPU 11 > [23500.239338] Modules linked in: [last unloaded: scsi_wait_scan] > [23500.239442] > [23500.239442] Pid: 15799, comm: fsstress Not tainted 3.3.0+ #43 Bochs Bochs > [23500.239442] RIP: 0010:[] [] submit_bh+0x10d/0x120 > [23500.239442] RSP: 0018:ffff880273a41b58 EFLAGS: 00010202 > [23500.239442] RAX: 000000000004d025 RBX: ffff8802469e5a90 RCX: 0000000000000000 > [23500.239442] RDX: ffff880273a41fd8 RSI: ffff8802469e5a90 RDI: 0000000000000211 > [23500.239442] RBP: ffff880273a41b78 R08: ffff880409645d80 R09: 0000000000000001 > [23500.239442] R10: 0000000000000001 R11: ffff880178439000 R12: 0000000000000211 > [23500.239442] R13: ffff880273a41c34 R14: ffff8804059b7000 R15: ffff880273a41fd8 > [23500.239442] FS: 00007fc8731e7700(0000) GS:ffff88041fd60000(0000) knlGS:0000000000000000 > [23500.239442] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b > [23500.239442] CR2: 00007fc873082008 CR3: 000000012456a000 CR4: 00000000000006e0 > [23500.239442] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 > [23500.239442] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 > [23500.239442] Process fsstress (pid: 15799, threadinfo ffff880273a40000, task ffff880150dd8000) > [23500.239442] Stack: > [23500.239442] ffff8804059b7000 ffff8802469e5a90 0000000000000211 ffff880273a41c34 > [23500.239442] ffff880273a41b98 ffffffff811ab59d 0000000000000002 ffff8804059b7128 > [23500.239442] ffff880273a41bf8 ffffffff8123be1d 0000000091827364 ffff88010e9a7e30 > [23500.239442] Call Trace: > [23500.239442] [] write_dirty_buffer+0x4d/0x80 > [23500.239442] [] __flush_batch+0x4d/0xa0 > [23500.239442] [] jbd2_log_do_checkpoint+0xf5/0x4f0 > [23500.239442] [] __jbd2_log_wait_for_space+0x89/0x190 > [23500.239442] [] start_this_handle+0x3a8/0x4e0 > [23500.239442] [] ? remove_wait_queue+0x50/0x50 > [23500.239442] [] jbd2__journal_start+0xc3/0x100 > [23500.239442] [] jbd2_journal_start+0x13/0x20 > [23500.239442] [] ext4_journal_start_sb+0x7f/0x1d0 > [23500.239442] [] ? ext4_fallocate+0x1a4/0x530 > [23500.239442] [] ? ext4_meta_trans_blocks+0xa5/0xb0 > [23500.239442] [] ext4_fallocate+0x1a4/0x530 > [23500.239442] [] do_fallocate+0xf2/0x160 > [23500.239442] [] sys_fallocate+0x4b/0x70 > [23500.239442] [] system_call_fastpath+0x16/0x1b > [23500.239442] Code: ee 44 89 e7 e8 35 1f 0f 00 49 8b 5d 18 4c 89 ef e8 19 4e 00 00 48 83 c4 08 c1 e3 18 c1 fb 1f 83 e3 a1 89 d8 5b 41 5c 41 5d 5d c3 <0f> 0b 0f 0b 0f 0b 0f 0b 0f 0b 66 0f 1f 84 00 00 00 00 00 55 48 > [23500.239442] RIP [] submit_bh+0x10d/0x120 > [23500.239442] RSP > [23500.261657] ---[ end trace 3db7a7a7ae953551 ]--- Hmm, looks like we tried to checkpoint BH_Unwritten buffer. That looks like a bug in fallocate() support. Not really related but definitely worth reporting. Honza -- Jan Kara SUSE Labs, CR