Re: [PATCH 3/3] ima: use fs method to read integrity data (updated patch description)

[Mimi Zohar <zohar@linux.vnet.ibm.com>]

On Sat, 2017-09-16 at 11:20 -0700, Linus Torvalds wrote:
> On Fri, Sep 15, 2017 at 1:25 PM, Mimi Zohar <zohar@linux.vnet.ibm.com> wrote:
> >
> > To resolve this locking problem, this patch defines a new
> > ->integrity_read file operation method, which is equivalent to
> > ->read_iter, except that it will not take the i_rwsem lock, but will
> > be called with the i_rwsem held exclusively.
> >
> > Since taking the i_rwsem exclusively is not required for reading the
> > file in order to calculate the file hash, the code only verifies
> > that the lock has been taken.
> 
> Ok, so I'm onboard with the commit message now, but realized that I'm
> not actually convinced that i_rwsem is even meaningful.
> 
> Sure, generic_file_write_iter() does take that lock exclusively, but
> not everybody uses generic_file_write_iter() at all for writing.

> For example, xfs still uses that i_rwsem, but for block-aligned writes
> it will only get it shared. And I'm not convinced some other
> filesystem might not end up using some other lock entirely.
> 
> So I'm basically not entirely convinced that these i_rwsem games make
> any sense at all.
> 
> The filesystem can do its own locking, and I'm starting to think that
> it would be better to just pass this "this is an integrity read" down
> to the filesystem, and expect the filesystem to do the locking based
> on that.

IMA would still need to take the i_rwsem to write the xattr.  Unless
the i_rwsem was taken before calling the integrity_read, calculating
the file hash would be serialized, but would not prevent the file hash
from being calculated multiple times.

(Introducing a new lock would result in the locks being taken in
reverse order for setxattr, chown, chmod syscalls.)

Mimi