On Fri 06-05-22 20:48:44, yebin wrote: > On 2022/5/6 16:50, Jan Kara wrote: > > On Fri 06-05-22 09:27:25, yebin wrote: > > > On 2022/5/5 23:47, Jan Kara wrote: > > > > On Thu 05-05-22 21:57:08, Ye Bin wrote: > > > > > we got issue as follows: > > > > > EXT4-fs error (device loop0): ext4_mb_generate_buddy:1141: group 0, block bitmap and bg descriptor inconsistent: 25 vs 31513 free cls > > > > > ------------[ cut here ]------------ > > > > > kernel BUG at fs/ext4/inode.c:2708! > > > > > invalid opcode: 0000 [#1] PREEMPT SMP KASAN PTI > > > > > CPU: 2 PID: 2147 Comm: rep Not tainted 5.18.0-rc2-next-20220413+ #155 > > > > > RIP: 0010:ext4_writepages+0x1977/0x1c10 > > > > > RSP: 0018:ffff88811d3e7880 EFLAGS: 00010246 > > > > > RAX: 0000000000000000 RBX: 0000000000000001 RCX: ffff88811c098000 > > > > > RDX: 0000000000000000 RSI: ffff88811c098000 RDI: 0000000000000002 > > > > > RBP: ffff888128140f50 R08: ffffffffb1ff6387 R09: 0000000000000000 > > > > > R10: 0000000000000007 R11: ffffed10250281ea R12: 0000000000000001 > > > > > R13: 00000000000000a4 R14: ffff88811d3e7bb8 R15: ffff888128141028 > > > > > FS: 00007f443aed9740(0000) GS:ffff8883aef00000(0000) knlGS:0000000000000000 > > > > > CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > > > > > CR2: 0000000020007200 CR3: 000000011c2a4000 CR4: 00000000000006e0 > > > > > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 > > > > > DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 > > > > > Call Trace: > > > > > <TASK> > > > > > do_writepages+0x130/0x3a0 > > > > > filemap_fdatawrite_wbc+0x83/0xa0 > > > > > filemap_flush+0xab/0xe0 > > > > > ext4_alloc_da_blocks+0x51/0x120 > > > > > __ext4_ioctl+0x1534/0x3210 > > > > > __x64_sys_ioctl+0x12c/0x170 > > > > > do_syscall_64+0x3b/0x90 > > > > > > > > > > It may happen as follows: > > > > > 1. write inline_data inode > > > > > vfs_write > > > > > new_sync_write > > > > > ext4_file_write_iter > > > > > ext4_buffered_write_iter > > > > > generic_perform_write > > > > > ext4_da_write_begin > > > > > ext4_da_write_inline_data_begin -> If inline data size too > > > > > small will allocate block to write, then mapping will has > > > > > dirty page > > > > > ext4_da_convert_inline_data_to_extent ->clear EXT4_STATE_MAY_INLINE_DATA > > > > > 2. fallocate > > > > > do_vfs_ioctl > > > > > ioctl_preallocate > > > > > vfs_fallocate > > > > > ext4_fallocate > > > > > ext4_convert_inline_data > > > > > ext4_convert_inline_data_nolock > > > > > ext4_map_blocks -> fail will goto restore data > > > > > ext4_restore_inline_data > > > > > ext4_create_inline_data > > > > > ext4_write_inline_data > > > > > ext4_set_inode_state -> set inode EXT4_STATE_MAY_INLINE_DATA > > > > > 3. writepages > > > > > __ext4_ioctl > > > > > ext4_alloc_da_blocks > > > > > filemap_flush > > > > > filemap_fdatawrite_wbc > > > > > do_writepages > > > > > ext4_writepages > > > > > if (ext4_has_inline_data(inode)) > > > > > BUG_ON(ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) > > > > > > > > > > To solved this issue, record origin 'EXT4_STATE_MAY_INLINE_DATA' flag, then pass > > > > > value to 'ext4_restore_inline_data', 'ext4_restore_inline_data' will > > > > > decide to if recovery 'EXT4_STATE_MAY_INLINE_DATA' flag according to parameter. > > > > > > > > > > Signed-off-by: Ye Bin <yebin10@xxxxxxxxxx> > > > > I think this will get also fixed by a patch from your colleague I've > > > > reviewed here [1], won't it? > > > > > > > > [1] https://lore.kernel.org/all/20220428165725.mvjh6mx7gr5vekqe@xxxxxxxxxx > > > > > > > The issue I fixed is not the same as the isuue my colleague fixed. > > OK, maybe I've jumped to conclusion too early but the fix I've referenced > > above will protect ext4_convert_inline_data() in ext4_fallocate() with > > inode->i_rwsem so I think the race you describe with ext4_da_write_begin() > > cannot happen. The inline conversion path with be entered either from > > ext4_da_write_begin() or from ext4_fallocate() but not from both. If I'm > > missing something, please explain how you think the problem happens with > > the above fix applied... Thanks! > > > > Honza > It's happen as follows: > step1: > ext4_file_write_iter > ext4_buffered_write_iter > generic_perform_write > ext4_da_write_begin > if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) > ext4_da_write_inline_data_begin > ext4_da_convert_inline_data_to_extent > __block_write_begin > ext4_clear_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA); > ->clear EXT4_STATE_MAY_INLINE_DATA flag, If inline data size > too > small will allocate block to write, then mapping will has dirty page > > ext4_da_write_end > generic_write_end > > step2： > vfs_fallocate > ext4_fallocate > ext4_convert_inline_data > if (ext4_has_inline_data(inode)) -> This condition is satisfied > ext4_convert_inline_data_nolock > ext4_map_blocks -> fail will goto restore data > ext4_restore_inline_data > ext4_create_inline_data > ext4_write_inline_data > ext4_set_inode_state -> set inode EXT4_STATE_MAY_INLINE_DATA > step3: > do_writepages > ext4_writepages > if (ext4_has_inline_data(inode)) > BUG_ON(ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) > > As we call "ext4_destroy_inline_data" to destory inline data when delay > allocation. So，if we call fallocate before writepages while > ext4_map_blocks return failed, will lead to above issue. This issue does > not require concurrency conditions. Aha, I see now. Thanks for explanation and sorry for being a bit dense. I'll have a look at your original patch. Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
