On 2020/8/6 23:03, Liu Song via Linux-f2fs-devel wrote:
> From: Liu Song <[email protected]>
>
> Inode hash has been removed in "make_bad_inode". If inode_unhashed
> is false, it must not be a bad inode.
>
> Signed-off-by: Liu Song <[email protected]>
> ---
> fs/f2fs/super.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 20e56b0fa46a..ee01d15effe3 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -1043,7 +1043,7 @@ static int f2fs_drop_inode(struct inode *inode)
> * - inode_wait_for_writeback(inode)
> */
> if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
Since logic of "bad inode should be removed from hash" is not controlled by
f2fs, so let's add below condition to make sure the rule of vfs is as it is.
f2fs_bug_on(sbi, is_bad_inode(inode));
> - if (!inode->i_nlink && !is_bad_inode(inode)) {
> + if (!inode->i_nlink) {
> /* to avoid evict_inode call simultaneously */
> atomic_inc(&inode->i_count);
> spin_unlock(&inode->i_lock);
>