Problem 1：In checkpoint process, missed io_end check for non-dirty buffer head may corrupt the filesystem. jbd2_journal_commit_transaction // there are several dirty buffer heads in transaction->t_checkpoint_list P1 wb_workfn jbd2_log_do_checkpoint __block_write_full_page trylock_buffer(bh) test_clear_buffer_dirty(bh) if (!buffer_dirty(bh)) __jbd2_journal_remove_checkpoint(jh) if (buffer_write_io_error(bh)) // false >> bh IO error occurs << jbd2_cleanup_journal_tail __jbd2_update_log_tail jbd2_write_superblock The bh won't be replayed in next mount, image becomes corrupted. 1. Apply diff_1 and compile kernel 2. sh test_1.sh [ 69.605320] assign g_bh, trace dir 12 buffer, add a(14) [ 69.617135] mark buffer dirty [ 69.632237] jbd2_log_do_checkpoint: wait clear dirty [ 70.646949] __block_write_full_page: clear buffer dirty, delay submit [ 70.733846] jbd2_log_do_checkpoint: wait dirty cleared [ 70.849564] __block_write_full_page: delay done [ 70.850340] Buffer I/O error on dev sda, logical block 3510, lost async page write [ 70.856224] Kernel panic - not syncing: DONE [ 70.857436] CPU: 2 PID: 2593 Comm: chmod Not tainted 6.4.0-rc3-00028-g848673e8c63a-dirty #12 [ 70.859879] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.1-2.fc37 04/01/2014 [ 70.862424] Call Trace: [ 70.863141] <TASK> [ 70.863755] dump_stack_lvl+0x86/0xc0 [ 70.864798] dump_stack+0x18/0x30 [ 70.865750] panic+0x4f2/0x530 [ 70.866619] jbd2_log_do_checkpoint+0x76e/0x810 [ 70.867898] __jbd2_log_wait_for_space+0x162/0x3d0 [ 70.869235] add_transaction_credits+0x194/0x480 [ 70.870512] ? kmem_cache_alloc+0x2bd/0x5c0 [ 70.871665] start_this_handle+0x14b/0x950 [ 70.872787] ? kmem_cache_alloc+0x2bd/0x5c0 [ 70.873949] jbd2__journal_start+0x105/0x200 [ 70.874899] __ext4_journal_start_sb+0x2d2/0x320 [ 70.875824] ext4_dirty_inode+0x40/0xa0 [ 70.876586] __mark_inode_dirty+0x93/0x720 [ 70.877397] ext4_setattr+0x1a6/0xf30 [ 70.878170] ? path_lookupat+0xc4/0x1e0 [ 70.878941] notify_change+0x46d/0x770 [ 70.879675] ? chmod_common+0x11e/0x210 [ 70.880431] chmod_common+0x11e/0x210 [ 70.881181] ? putname+0x79/0xa0 [ 70.881831] do_fchmodat+0x6b/0xe0 [ 70.882504] __x64_sys_fchmodat+0x21/0x30 [ 70.883315] do_syscall_64+0x6c/0xf0 [ 70.884045] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 70.884785] RIP: 0033:0x7fb087cffdb9 [ 70.885307] Code: ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 66 90 f7 c1 ff fe ff ff 75 38 80 e5 01 75 4b 48 63 ff 89 d2 b8 0c 01 00 0 [ 70.887957] RSP: 002b:00007ffddfbb7e48 EFLAGS: 00000246 ORIG_RAX: 000000000000010c [ 70.889055] RAX: ffffffffffffffda RBX: 0000000000000001 RCX: 00007fb087cffdb9 [ 70.890076] RDX: 00000000000001ed RSI: 000055850a71a0f0 RDI: ffffffffffffff9c [ 70.891086] RBP: 000055850a71a060 R08: 0000000000000000 R09: 0000000000000000 [ 70.892072] R10: 0000000000000000 R11: 0000000000000246 R12: 000055850900a734 [ 70.893062] R13: 000055850a71a0f0 R14: 000055850a71b320 R15: 0000000000008000 3. reboot 4. fsck.ext4 -fa /dev/sda /dev/sda: recovering journal /dev/sda: Unattached inode 14 /dev/sda: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY. (i.e., without -a or -p options) Problem 2: In forget process, missed io_end check for non-dirty buffer head may corrupt the filesystem. jbd2_journal_commit_transaction // there are several dirty buffer heads in transaction->t_checkpoint_list P1 wb_workfn P2 jbd2_journal_forget __block_write_full_page trylock_buffer(bh) test_clear_buffer_dirty(bh) if (!buffer_dirty(bh)) __jbd2_journal_remove_checkpoint(jh) if (buffer_write_io_error(bh)) // false >> bh IO error occurs << jbd2_log_do_checkpoint jbd2_cleanup_journal_tail jbd2_write_superblock The bh won't be replayed in next mount, image becomes corrupted. 1. Apply diff_2 and compile kernel 2. sh test_2.sh [ 3435.871701] ext4_delete_entry: assign g_bh [ 3435.878143] mark buffer dirty [ 3435.892735] ck wait buffer remove [ 3436.897053] jbd2_journal_forget: wait clear dirty [ 3437.910558] __block_write_full_page: clear buffer dirty, delay submit [ 3437.998090] jbd2_journal_forget: wait dirty cleared [ 3437.998628] wait checkpoint finish [ 3438.003545] ck wait done [ 3438.111919] __block_write_full_page: delay done [ 3438.112992] Buffer I/O error on dev sda, logical block 3511, lost async page write [ 3438.115152] Kernel panic - not syncing: DONE [ 3438.115513] CPU: 3 PID: 2676 Comm: chown Not tainted 6.4.0-rc3-00028-g848673e8c63a-dirty #51 [ 3438.116211] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.1-2.fc37 04/01/2014 [ 3438.116896] Call Trace: [ 3438.117103] <TASK> [ 3438.117294] dump_stack_lvl+0x86/0xc0 [ 3438.117622] dump_stack+0x18/0x30 [ 3438.117900] panic+0x4f2/0x530 [ 3438.118151] jbd2_log_do_checkpoint+0x84f/0x870 [ 3438.118522] ? do_syscall_64+0x6c/0xf0 [ 3438.118824] __jbd2_log_wait_for_space+0x162/0x3d0 [ 3438.119222] add_transaction_credits+0x194/0x480 [ 3438.119595] ? kmem_cache_alloc+0x2bd/0x5c0 [ 3438.119941] start_this_handle+0x14b/0x950 [ 3438.120272] ? kmem_cache_alloc+0x2bd/0x5c0 [ 3438.120613] jbd2__journal_start+0x105/0x200 [ 3438.120967] __ext4_journal_start_sb+0x2d2/0x320 [ 3438.121341] ext4_setattr+0x37a/0xf30 [ 3438.121649] notify_change+0x46d/0x770 [ 3438.121955] ? chown_common+0x1a8/0x310 [ 3438.122271] chown_common+0x1a8/0x310 [ 3438.122566] do_fchownat+0xfe/0x170 [ 3438.122858] __x64_sys_fchownat+0x27/0x40 [ 3438.123190] do_syscall_64+0x6c/0xf0 [ 3438.123488] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 3438.123920] RIP: 0033:0x7f8d29d013ca [ 3438.124224] Code: 48 8b 0d c1 da 2c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 49 89 ca b8 04 01 00 8 [ 3438.125709] RSP: 002b:00007ffd1f18b5a8 EFLAGS: 00000246 ORIG_RAX: 0000000000000104 [ 3438.126314] RAX: ffffffffffffffda RBX: 00007ffd1f18b7f0 RCX: 00007f8d29d013ca [ 3438.126861] RDX: 000000000000004a RSI: 000055c2a2fa5070 RDI: 00000000ffffff9c [ 3438.127421] RBP: 000055c2a2fa3d10 R08: 0000000000000000 R09: 00000000ffffffff [ 3438.127989] R10: 00000000ffffffff R11: 0000000000000246 R12: 000055c2a2fa3d88 [ 3438.128562] R13: 00000000ffffffff R14: 00000000ffffff9c R15: 0000000000000001 [ 3438.129145] </TASK> [ 3438.129406] Kernel Offset: disabled [ 3438.129697] ---[ end Kernel panic - not syncing: DONE ]--- 3. reboot 4. fsck.ext4 -fn /dev/sda e2fsck 1.43.4 (31-Jan-2017) Warning: skipping journal recovery because doing a read-only filesystem check. Pass 1: Checking inodes, blocks, and sizes Pass 2: Checking directory structure Pass 3: Checking directory connectivity Pass 4: Checking reference counts Inode 44 ref count is 1, should be 2. Fix? no Pass 5: Checking group summary information /dev/sda: ********** WARNING: Filesystem still has errors ********** /dev/sda: 45/25688 files (0.0% non-contiguous), 8899/102400 blocks Problem 3: __cp_buffer_busy checks buffer dirty after checking buffer locking state, buffer dirty cleaning could happen between them, which leads bh removed from checkpoint list. jbd2_journal_commit_transaction // there are several dirty buffer heads in transaction->t_checkpoint_list drop_cache wb_workfn P2 journal_shrink_one_cp_list __cp_buffer_busy jh->b_transaction || buffer_locked(bh) // returns false __block_write_full_page trylock_buffer(bh) test_clear_buffer_dirty(bh) buffer_dirty(bh) // returns false __jbd2_journal_remove_checkpoint if (buffer_write_io_error(bh)) // false >> bh IO error occurs << jbd2_log_do_checkpoint jbd2_cleanup_journal_tail __jbd2_update_log_tail jbd2_write_superblock The bh won't be replayed in next mount, image becomes corrupted. 1. Apply diff_3 and compile kernel 2. sh test_3.sh [ 34.007674] add_dirent_to_buf: assign g_bh [ 34.019950] mark buffer dirty [ 34.071456] __cp_buffer_busy: wait clear dirty [ 35.094103] __block_write_full_page: clear buffer dirty, delay submit [ 35.174522] __cp_buffer_busy: wait dirty cleared [ 35.296479] __block_write_full_page: delay done [ 35.298304] Buffer I/O error on dev sda, logical block 3510, lost async page write [ 36.145015] Kernel panic - not syncing: DONE [ 36.146237] CPU: 3 PID: 2814 Comm: chmod Not tainted 6.4.0-rc3-00028-g848673e8c63a-dirty #72 [ 36.148486] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.1-2.fc37 04/01/2014 [ 36.150848] Call Trace: [ 36.151529] <TASK> [ 36.152112] dump_stack_lvl+0x86/0xc0 [ 36.153177] dump_stack+0x18/0x30 [ 36.154136] panic+0x4f2/0x530 [ 36.154990] jbd2_log_do_checkpoint+0x73a/0x760 [ 36.156251] ? do_fchmodat+0x6b/0xe0 [ 36.157245] __jbd2_log_wait_for_space+0x162/0x3d0 [ 36.158481] add_transaction_credits+0x194/0x480 [ 36.159376] ? kmem_cache_alloc+0x2bd/0x5c0 [ 36.160192] start_this_handle+0x14b/0x950 [ 36.160995] ? kmem_cache_alloc+0x2bd/0x5c0 [ 36.161814] jbd2__journal_start+0x105/0x200 [ 36.162654] __ext4_journal_start_sb+0x2d2/0x320 [ 36.163560] ext4_dirty_inode+0x40/0xa0 [ 36.164322] __mark_inode_dirty+0x93/0x720 [ 36.165126] ext4_setattr+0x1a6/0xf30 [ 36.165850] ? path_lookupat+0xc4/0x1e0 [ 36.166621] notify_change+0x46d/0x770 [ 36.167354] ? chmod_common+0x11e/0x210 [ 36.168069] chmod_common+0x11e/0x210 [ 36.168797] ? putname+0x79/0xa0 [ 36.169444] do_fchmodat+0x6b/0xe0 [ 36.170132] __x64_sys_fchmodat+0x21/0x30 [ 36.170924] do_syscall_64+0x6c/0xf0 [ 36.171630] entry_SYSCALL_64_after_hwframe+0x63/0xcd [ 36.172613] RIP: 0033:0x7f2dbdeffdb9 [ 36.173172] Code: ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 66 90 f7 c1 ff fe ff ff 75 38 80 e5 01 75 4b 48 63 ff 89 d2 b8 0c 01 00 0 [ 36.175777] RSP: 002b:00007ffe6917fa18 EFLAGS: 00000246 ORIG_RAX: 000000000000010c [ 36.176853] RAX: ffffffffffffffda RBX: 0000000000000001 RCX: 00007f2dbdeffdb9 [ 36.177854] RDX: 00000000000001ed RSI: 000055e20d17a0f0 RDI: ffffffffffffff9c [ 36.178862] RBP: 000055e20d17a060 R08: 0000000000000000 R09: 0000000000000000 [ 36.179857] R10: 0000000000000000 R11: 0000000000000246 R12: 000055e20b40a734 [ 36.180867] R13: 000055e20d17a0f0 R14: 000055e20d17b320 R15: 0000000000008000 [ 36.181889] </TASK> [ 36.183237] Kernel Offset: disabled [ 36.183645] ---[ end Kernel panic - not syncing: DONE ]--- 3. reboot 4. fsck.ext4 -fa /dev/sda /dev/sda: recovering journal /dev/sda: Unattached inode 213 /dev/sda: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY. (i.e., without -a or -p options)
Created attachment 304327 [details] diff_1
Created attachment 304328 [details] test_1.sh
Created attachment 304329 [details] diff_2
Created attachment 304330 [details] test_2.sh
Created attachment 304331 [details] diff_3
Created attachment 304332 [details] test_3.sh
Can you send your diffs as proper patches to linux-ext4@vger.kernel.org for review? (Hint: see Documentation/process/submitting-patches.rst for how to contribute patches.)
(In reply to Bagas Sanjaya from comment #7) > Can you send your diffs as proper patches to linux-ext4@vger.kernel.org for > review? (Hint: see Documentation/process/submitting-patches.rst for how to > contribute patches.) These diffs are used to reproduce above problems, formal fix patches are under writing and being sent out soon.