Latest working kernel version: unknown Earliest failing kernel version: 2.6.29-rc1 Distribution: CentOS 5.2 (x86_64) Hardware Environment: MY20R/B-4(CPU: Intel(R) Core(TM) 2 Duo E4400, MEM: 2560MB Software Environment: e2fsprogs-1.41.3 Problem Description: 2.6.29-rc1 + the most recent ext4 patch queue When mounting ext4 filesystem without a journal, BUG_ON panic occurs by executing filefrag command. The following error messages show in the system logs: Jan 27 15:56:53 bsdB08126 kernel: kernel BUG at fs/ext4/inode.c:2825! Jan 27 15:56:53 bsdB08126 kernel: invalid opcode: 0000 [0000001] SMP Jan 27 15:56:53 bsdB08126 kernel: last sysfs file: /sys/class/drm/card0/dev Jan 27 15:56:53 bsdB08126 kernel: CPU 0 Jan 27 15:56:53 bsdB08126 kernel: Modules linked in: i915 drm i2c_algo_bit autofs4 hidp rfcomm l2cap bluetooth sunrpc iptable_filter ip_tables ip6t_REJECT xt_tcpudp ip6table_filter ip6_tables x_tables dm_mirror dm_region_hash dm_log dm_multipath dm_mod sbs sbshc battery acpi_memhotplug ac ipv6 lp snd_hda_codec_realtek snd_hda_intel sg snd_hda_codec snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss snd_mixer_oss snd_pcm serio_raw sr_mod rtc_cmos cdrom parport_pc snd_timer rtc_core parport rtc_lib snd button i2c_i801 e1000e soundcore pcspkr snd_page_alloc i2c_core ata_piix libata sd_mod scsi_mod uhci_hcd ohci_hcd ehci_hcd [last unloaded: microcode] Jan 27 15:56:53 bsdB08126 kernel: Pid: 3902, comm: filefrag Not tainted 2.6.29-rc1 0000012 Jan 27 15:56:53 bsdB08126 kernel: RIP: 0010:[<ffffffff803043a9>] [<ffffffff803043a9>] ext4_bmap+0x5a/0xc8 Jan 27 15:56:53 bsdB08126 kernel: RSP: 0018:ffff880064857ea8 EFLAGS: 00010202 Jan 27 15:56:53 bsdB08126 kernel: RAX: 0000000000000000 RBX: ffff88006380d490 RCX: 0000000000000015 Jan 27 15:56:53 bsdB08126 kernel: RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff88006380d5a8 Jan 27 15:56:53 bsdB08126 kernel: RBP: ffff88006380d5a0 R08: ffff880064857e68 R09: 0000000000000000 Jan 27 15:56:53 bsdB08126 kernel: R10: 74616d726f662073 R11: ffff8800905d6380 R12: 00000000000063ff Jan 27 15:56:53 bsdB08126 kernel: R13: 00000000000063ff R14: 0000000000000003 R15: 0000000000000000 Jan 27 15:56:53 bsdB08126 kernel: FS: 00007f4b8eb586e0(0000) GS:ffffffff8086c080(0000) knlGS:0000000000000000 Jan 27 15:56:53 bsdB08126 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b Jan 27 15:56:53 bsdB08126 kernel: CR2: 00000031b8e62b80 CR3: 0000000062015000 CR4: 00000000000006e0 Jan 27 15:56:53 bsdB08126 kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 Jan 27 15:56:54 bsdB08126 kernel: DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Jan 27 15:56:54 bsdB08126 kernel: Process filefrag (pid: 3902, threadinfo ffff880064856000, task ffff88007386e970) Jan 27 15:56:54 bsdB08126 kernel: Stack: Jan 27 15:56:54 bsdB08126 kernel: 0000000000000001 ffff88006380d5a0 00007fff96b70fb4 00000000000063ff Jan 27 15:56:55 bsdB08126 kernel: 0000000000000000 ffffffff802aa43f 0000000000000000 0000000000000000 Jan 27 15:56:55 bsdB08126 kernel: 0000000000000000 0000000000000000 0000000000000000 ffffffff804e9068 Jan 27 15:56:56 bsdB08126 kernel: Call Trace: Jan 27 15:56:56 bsdB08126 kernel: [<ffffffff802aa43f>] ? do_vfs_ioctl+0x2bf/0x4a3 Jan 27 15:56:56 bsdB08126 kernel: [<ffffffff804e9068>] ? do_page_fault+0x3a8/0x72f Jan 27 15:56:57 bsdB08126 kernel: [<ffffffff802aa674>] ? sys_ioctl+0x51/0x70 Jan 27 15:56:58 bsdB08126 kernel: [<ffffffff8020bedb>] ? system_call_fastpath+0x16/0x1b Jan 27 15:56:58 bsdB08126 kernel: Code: 08 74 08 48 89 ef e8 28 fe f6 ff 48 8b 83 f8 00 00 00 48 8b 80 90 02 00 00 48 8b 80 90 41 00 00 48 85 c0 75 0f f6 43 a0 01 74 04 <0f> 0b eb fe 48 85 c0 74 42 8b 43 a0 a8 01 74 3b 83 e0 fe 89 43 Jan 27 15:56:58 bsdB08126 kernel: RIP [<ffffffff803043a9>] ext4_bmap+0x5a/0xc8 Jan 27 15:56:58 bsdB08126 kernel: RSP <ffff880064857ea8> Jan 27 15:56:58 bsdB08126 kernel: ---[ end trace 60cf7e0d66d2b1ce ]--- Whenever __ext4_journalled_writepage() or ext4_journalled_write_end() is called, EXT4_STATE_JDATA is set in i_state in spite of non-journal ext4. That's weird. Steps to reproduce: 1. mke2fs -O extents /dev/sdaX 2. mount -t ext4 /dev/sdaX /mnt/mpX 3. dd if=/dev/urandom of=/mnt/mpX/file bs=1K count=102400 4. filefrag /mnt/mpX/file
Here's the patch to fix this issue. - Ted From b9ec63f78b425c0e16cc95605b5d4ff2dc228b97 Mon Sep 17 00:00:00 2001 From: Theodore Ts'o <tytso@mit.edu> Date: Fri, 30 Jan 2009 00:00:24 -0500 Subject: [PATCH] ext4: Remove bogus BUG() check in ext4_bmap() The code to support journal-less ext4 operation added a BUG to ext4_bmap() which fired if there was no journal and the EXT4_STATE_JDATA bit was set in the i_state field. This caused running the filefrag program (which uses the FIMBAP ioctl) to trigger a BUG(). The EXT4_STATE_JDATA bit is only used for ext4_bmap(), and it's harmless for the bit to be set. We could add a check in __ext4_journalled_writepage() and ext4_journalled_write_end() to only set the EXT4_STATE_JDATA bit if the journal is present, but that adds an extra test and jump instruction. It's easier to simply remove the BUG check. http://bugzilla.kernel.org/show_bug.cgi?id=12568 Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Cc: stable@kernel.org --- fs/ext4/inode.c | 3 --- 1 files changed, 0 insertions(+), 3 deletions(-) diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index b4386da..03ba20b 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -2821,9 +2821,6 @@ static sector_t ext4_bmap(struct address_space *mapping, sector_t block) filemap_write_and_wait(mapping); } - BUG_ON(!EXT4_JOURNAL(inode) && - EXT4_I(inode)->i_state & EXT4_STATE_JDATA); - if (EXT4_JOURNAL(inode) && EXT4_I(inode)->i_state & EXT4_STATE_JDATA) { /* * This is a REALLY heavyweight approach, but the use of
The patch fixes this issue for me. Thanks for your help.