Re: [PATCH] ext3 [linux-2.6.2.]: accessing already freed inodes when under memory pressure

[Matthew Wilcox <willy@debian.org>]

On Tue, Mar 30, 2004 at 02:39:30PM +0100, David Woodhouse wrote:
> If we ignore the anecdotes about reinstating a libc which has been
> unlinked but is still in use -- is there any reason we're want any file
> system to allow link() to inodes with i_nlink == 0?

To preserve the anecdotes, we could check i_count as well, but this starts
to get hairy.  If iput_final() races with vfs_link() we'd have problems.
Should iput_final() be taking i_sem anyway?  Seems to me we have the race:

CPU0					CPU1

iput()
atomic_dec_and_lock(&inode->i_count, &inode_lock)
					sys_unlink()
					atomic_inc(&inode->i_count);
					vfs_unlink()
					inode->i_nlink-- (reaches 0)
iput_final()
generic_drop_inode()
					
if (!inode->i_nlink)
        generic_delete_inode(inode);
else
        generic_forget_inode(inode);
					iput()
					atomic_dec_and_lock()
					generic_delete_inode()

Calling generic_delete_inode() twice gives us a BUG() in clear_inode()
(if we even get that far).  Have I overlooked something here?	

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain