Hi,
I was having a leave, but now came back to this stuff.
On Sat, 2012-04-14 at 22:12 +0900, OGAWA Hirofumi wrote:
> If fsinfo is inode, the order can be
>
> 1) flush inodes
> 1a) flush fsinfo inode
> 1b) flush normal inodes
> 2) last iput(normal inodes)
> truncate()
> dirty fsinfo inode
3) a bit later fsinfo inode is flushed again.
Sorry, I do not still see where is the issue.
Also, how is it different form the current situation where
you have absolutely the same.
I really still do not see the issue.
--
Best Regards,
Artem Bityutskiy