Re: [TOMOYO #9 (2.6.27-rc7-mm1) 1/6] LSM adapter functions.

From: Kentaro Takeda <takedakn@nttdata.co.jp>
To: "Serge E. Hallyn" <serue@us.ibm.com>
Cc: Valdis.Kletnieks@vt.edu, Casey Schaufler <casey@schaufler-ca.com>,
	linux-security-module@vger.kernel.org,
	linux-kernel@vger.kernel.org, haradats@nttdata.co.jp,
	Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>,
	Al Viro <viro@ZenIV.linux.org.uk>
Subject: Re: [TOMOYO #9 (2.6.27-rc7-mm1) 1/6] LSM adapter functions.
Date: Thu, 02 Oct 2008 14:04:50 +0900	[thread overview]
Message-ID: <48E45672.5030606@nttdata.co.jp> (raw)
In-Reply-To: <20081001211507.GA28377@us.ibm.com>

Serge E. Hallyn wrote:
> I suppose you could do something like define both _path and _inode,
> save away your result from the _path hook but always return 0, there,
> then if you'd saved off an error and you make it to the _inode hook,
> return the error there...
You mean do MAC checks in security_path_*() and return error code of 
security_path_*() in security_inode_*()? Then, method for passing the 
error code to security_inode_*() is a problem.

It was possible to store the error code into current->security->
something. But now, it is impossible to store the error code into 
current->cred->security->something because current->cred is shared by 
multiple processes. To solve this problem, we everytime need to copy 
current->cred in security_path_*() and we need a new hook called just 
after returning from vfs_* (like mnt_drop_write()) for clearing the 
error code. 

Or, another way is to pass the error code as a vfs_*() parameter.

What do you think these approaches?

Regards,