linux-kernel - Re: [PATCH 2/2] debugfs: return error values, not NULL

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20190123130917.GZ4087@dhcp22.suse.cz>
Date:   Wed, 23 Jan 2019 14:09:17 +0100
From:   Michal Hocko <mhocko@...nel.org>
To:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc:     linux-kernel@...r.kernel.org,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Ulf Hansson <ulf.hansson@...aro.org>,
        Gary R Hook <ghook@....com>,
        Heiko Carstens <heiko.carstens@...ibm.com>
Subject: Re: [PATCH 2/2] debugfs: return error values, not NULL

On Wed 23-01-19 14:00:57, Greg KH wrote:
> On Wed, Jan 23, 2019 at 01:40:24PM +0100, Michal Hocko wrote:
> > On Wed 23-01-19 13:26:26, Greg KH wrote:
> > > On Wed, Jan 23, 2019 at 01:13:50PM +0100, Michal Hocko wrote:
> > > > On Wed 23-01-19 12:55:35, Greg KH wrote:
> > > > > On Wed, Jan 23, 2019 at 12:06:28PM +0100, Michal Hocko wrote:
> > > > > > On Wed 23-01-19 11:28:14, Greg KH wrote:
> > > > > > > When an error happens, debugfs should return an error pointer value, not
> > > > > > > NULL.  This will prevent the totally theoretical error where a debugfs
> > > > > > > call fails due to lack of memory, returning NULL, and that dentry value
> > > > > > > is then passed to another debugfs call, which would end up succeeding,
> > > > > > > creating a file at the root of the debugfs tree, but would then be
> > > > > > > impossible to remove (because you can not remove the directory NULL).
> > > > > > > 
> > > > > > > So, to make everyone happy, always return errors, this makes the users
> > > > > > > of debugfs much simpler (they do not have to ever check the return
> > > > > > > value), and everyone can rest easy.
> > > > > > 
> > > > > > How come this is safe at all? Say you are creating a directory by
> > > > > > debugfs_create_dir and then feed the return value to debugfs_create_files
> > > > > > as a parent. In case of error you are giving it an invalid pointer and
> > > > > > likely blow up unless I miss something.
> > > > > 
> > > > > debugfs_create_files checks for invalid parents and will just refuse to
> > > > > create the file.  It's always done that.
> > > > 
> > > > I must be missing something because debugfs_create_files does
> > > > 	d_inode(parent)->i_private = data;
> > > > as the very first thing and that means that it dereferences an invalid
> > > > pointer right there.
> > > 
> > > debugfs_create_file() -> __debugfs_create_file() -> start_creating()
> > > and that function checks if parent is an error, which it aborts on, or
> > > if it is NULL, it sets parent to a valid value:
> > > 
> > >         /* If the parent is not specified, we create it in the root.
> > >          * We need the root dentry to do this, which is in the super
> > >          * block. A pointer to that is in the struct vfsmount that we
> > >          * have around.
> > >          */
> > >         if (!parent)
> > >                 parent = debugfs_mount->mnt_root;
> > > 
> > > I don't see any line that looks like:
> > > >       d_inode(parent)->i_private = data;
> > > in Linus's tree right now, what kernel version are you referring to?
> > 
> > Ohh, my bad. I have looked at debugfs_create_files which is a mq helper
> > around debugfs_create_file. But that is a good example why this patch is
> > dangerous anyway. blk_mq_debugfs_register simply checks for NULL and
> > debugfs_create_files doesn't expect ERR_PTR here. So you would have to
> > check each and every user to make sure you can do that.
> 
> Ah, I already have that patch in my "to add a proper changelog" queue,
> it's below and fixes that problem.

OK, fair enough. I am just wondering how many more gems like that are
lurking there. Do not get me wrong but a broken error handling is rarely
fixed by removing it.

And Cc: stable is completely inappropriate IMNSHO. This is just adding a
risk without a large benefit.

Moreover all these changes should be posted in a single patch thread so
that everybody can see the final outcome.
-- 
Michal Hocko
SUSE Labs