Re: [PATCH] f2fs: fix to set superblock dirty correctly

[Chao Yu <chao@kernel.org>]

On 2016/8/27 9:01, Jaegeuk Kim wrote:
> On Fri, Aug 26, 2016 at 10:20:18PM +0800, Chao Yu wrote:
>> From: Chao Yu <yuchao0@huawei.com>
>>
>> tests/generic/251 of fstest suit complains us with below message:
>>
>> ------------[ cut here ]------------
>> invalid opcode: 0000 [#1] PREEMPT SMP
>> CPU: 2 PID: 7698 Comm: fstrim Tainted: G           O    4.7.0+ #21
>> task: e9f4e000 task.stack: e7262000
>> EIP: 0060:[<f89fcefe>] EFLAGS: 00010202 CPU: 2
>> EIP is at write_checkpoint+0xfde/0x1020 [f2fs]
>> EAX: f33eb300 EBX: eecac310 ECX: 00000001 EDX: ffff0001
>> ESI: eecac000 EDI: eecac5f0 EBP: e7263dec ESP: e7263d18
>>  DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
>> CR0: 80050033 CR2: b76ab01c CR3: 2eb89de0 CR4: 000406f0
>> Stack:
>>  00000001 a220fb7b e9f4e000 00000002 419ff2d3 b3a05151 00000002 e9f4e5d8
>>  e9f4e000 419ff2d3 b3a05151 eecac310 c10b8154 b3a05151 419ff2d3 c10b78bd
>>  e9f4e000 e9f4e000 e9f4e5d8 00000001 e9f4e000 ec409000 eecac2cc eecac288
>> Call Trace:
>>  [<c10b8154>] ? __lock_acquire+0x3c4/0x760
>>  [<c10b78bd>] ? mark_held_locks+0x5d/0x80
>>  [<f8a10632>] f2fs_trim_fs+0x1c2/0x2e0 [f2fs]
>>  [<f89e9f56>] f2fs_ioctl+0x6b6/0x10b0 [f2fs]
>>  [<c13d51df>] ? __this_cpu_preempt_check+0xf/0x20
>>  [<c10b4281>] ? trace_hardirqs_off_caller+0x91/0x120
>>  [<f89e98a0>] ? __exchange_data_block+0xd30/0xd30 [f2fs]
>>  [<c120b2e1>] do_vfs_ioctl+0x81/0x7f0
>>  [<c11d57c5>] ? kmem_cache_free+0x245/0x2e0
>>  [<c1217840>] ? get_unused_fd_flags+0x40/0x40
>>  [<c1206eec>] ? putname+0x4c/0x50
>>  [<c11f631e>] ? do_sys_open+0x16e/0x1d0
>>  [<c1001990>] ? do_fast_syscall_32+0x30/0x1c0
>>  [<c13d51df>] ? __this_cpu_preempt_check+0xf/0x20
>>  [<c120baa8>] SyS_ioctl+0x58/0x80
>>  [<c1001a01>] do_fast_syscall_32+0xa1/0x1c0
>>  [<c178cc54>] sysenter_past_esp+0x45/0x74
>> EIP: [<f89fcefe>] write_checkpoint+0xfde/0x1020 [f2fs] SS:ESP 0068:e7263d18
>> ---[ end trace 4de95d7e6b3aa7c6 ]---
>>
>> The reason is: with below call stack, we will encounter BUG_ON during
>> doing fstrim.
>>
>> Thread A				Thread B
>> - write_checkpoint
>>  - do_checkpoint
>> 					- f2fs_write_inode
>> 					 - update_inode_page
>> 					  - update_inode
>> 					   - set_page_dirty
>> 					    - f2fs_set_node_page_dirty
>> 					     - inc_page_count
>> 					      - percpu_counter_inc
>> 					      - set_sbi_flag(SBI_IS_DIRTY)
>>   - clear_sbi_flag(SBI_IS_DIRTY)
>>
>> Thread C				Thread D
>> - f2fs_write_node_page
>>  - set_node_addr
>>   - __set_nat_cache_dirty
>>    - nm_i->dirty_nat_cnt++
>> 					- do_vfs_ioctl
>> 					 - f2fs_ioctl
>> 					  - f2fs_trim_fs
>> 					   - write_checkpoint
>> 					    - f2fs_bug_on(nm_i->dirty_nat_cnt)
>>
>> Fix it by setting superblock dirty correctly in do_checkpoint and
>> f2fs_write_node_page.
>>
>> Signed-off-by: Chao Yu <yuchao0@huawei.com>
>> ---
>>  fs/f2fs/checkpoint.c | 4 ++++
>>  fs/f2fs/node.c       | 1 +
>>  2 files changed, 5 insertions(+)
>>
>> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
>> index cd0443d..68c723c 100644
>> --- a/fs/f2fs/checkpoint.c
>> +++ b/fs/f2fs/checkpoint.c
>> @@ -1153,6 +1153,10 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
>>  	clear_prefree_segments(sbi, cpc);
>>  	clear_sbi_flag(sbi, SBI_IS_DIRTY);
>>  
>> +	/* redirty superblock if node page is updated by ->write_inode */
>> +	if (get_pages(sbi, F2FS_DIRTY_NODES))
> 
> Need to check F2FS_DIRTY_IMETA and F2FS_DIRTY_DENTS as well?

Need to check F2FS_DIRTY_IMETA additionally? since F2FS_DIRTY_DENTS will not be
updated during checkpoint.

> And, if we have this, I don't think we need to worry about f2fs_lock_op() for
> update_inode_page() as well.

OK, I didn't find any data consistency issue related to this so far.

Thanks,

> 
> Thanks,
> 
>> +		set_sbi_flag(sbi, SBI_IS_DIRTY);
>> +
>>  	return 0;
>>  }
>>  
>> diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
>> index 8a28800..365c6ff 100644
>> --- a/fs/f2fs/node.c
>> +++ b/fs/f2fs/node.c
>> @@ -1597,6 +1597,7 @@ static int f2fs_write_node_page(struct page *page,
>>  	fio.old_blkaddr = ni.blk_addr;
>>  	write_node_page(nid, &fio);
>>  	set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page));
>> +	set_sbi_flag(sbi, SBI_IS_DIRTY);
>>  	dec_page_count(sbi, F2FS_DIRTY_NODES);
>>  	up_read(&sbi->node_write);
>>  
>> -- 
>> 2.7.2