Re: [TOMOYO #8 (2.6.25-mm1) 1/7] Introduce new LSM hooks.

[Stephen Smalley]

On Thu, 2008-05-01 at 08:47 -0700, Chris Wright wrote:
> * Tetsuo Handa (penguin-kernel@xxxxxxxxxxxxxxxxxxx) wrote:
> > The MAY_WRITE flag is not passed to security_inode_permission()
> > if security_inode_permission() is called from __open_namei_create().
> > Since TOMOYO Linux doesn't check MAY_READ/MAY_WRITE permissions for individual
> > read()/write() requests, the permission checks at open() time (i.e. may_open())
> > is the only chance to check MAY_WRITE flag. If I can't check MAY_WRITE flag
> > here, TOMOYO Linux can't control open(O_WRONLY | O_CREATE | O_EXCL).
> > 
> > Also, the O_TRUNC flag is not passed to security_inode_permission() because
> > vfs_permission() receives only MAY_READ/MAY_WRITE/MAY_APPEND flags, but
> > I have to check O_TRUNC flag before do_truncate().
> > 
> > So, I inserted a new hook here so that this hook can check all
> > MAY_READ/MAY_WRITE/O_APPEND/O_TRUNC flags together in a single place.
> 
> The reason I ask is because it doesn't check.  It only checks O_APPEND,
> but that's already passed in (MAY_APPEND).  So AFAICT, it's only O_TRUNC
> that you are trying to special case.  But in that case...all that is
> being asked for is MAY_WRITE permission.  Anything else doesn't make
> sense, especially since that's all you get from the truncate(2) path.
> 
> <snip>
> > This is an inevitable duplication since I want to do conventional checks
> > (DAC checks and inode operation existence checks) before TOMOYO Linux's check.
> > 
> > By the way, Stephen Smalley thinks it is better to copy codes which is needed by
> > pre_vfs_*() (i.e. may_create()/may_delete()/check_sticky()) into
> > security/tomoyo/ directory and leave vfs_*() untouched rather than
> > extract pre_vfs_*() from vfs_*() and call pre_vfs_*() from vfs_*().
> 
> I'm not sure he means literally copy.  Typically we take existing
> functionality and make it externally usable.

I didn't think splitting pre functions out of all of the vfs helpers was
such a good idea.   Making may_create/delete() available might make
sense.  And for complex combinations of DAC logic, perhaps introducing
may_link(), may_rename(), etc. might make sense.  

> Also, all the changes you make that are not in vfs_* helpers won't get
> picked up by NFS.

-- 
Stephen Smalley
National Security Agency

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/