2001-07-12 09:40:18

by Tigran Aivazian

[permalink] [raw]
Subject: a bug in ext2_put_inode? (was Re: Some question about VFS

On Wed, 11 Jul 2001, yxpeng wrote:
> When I read the source of VFS, I notice that the fs-specific
> function sb->s_op->put_inode() is called before the inode->i_count is
> decremented and then checked. In the ext2 source, we can see that what
> the put_inode() does is to free the preallocated blocks without checking
> the i_count. I want to know why the put_inode() should be taken before
> i_count is checked. I think maybe there are some other processes that
> hold this file open and may use the preallocated blocks. Now that the
> other processes may use the preallocated blocks, why here we should free
> them? I am really confused.

It is possible that you may have found a bug, I am forwarding your
question to the linux-kernel list where someone can answer with more
certainty...

Regards,
Tigran