On Sat, Sep 16, 2017 at 11:20:47AM -0700, Linus Torvalds wrote:

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.

Only for direct I/O, and IMA and direct I/O don't work together.

From ima_collect_measurement:

		if (file->f_flags & O_DIRECT) {
			audit_cause = "failed(directio)";
			result = -EACCES;
			goto out;
		}

(and yes, it should be checking for IOCB_DIRECT to avoid racy
f_flags manipulations, but that's another issue)

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.

Well, that's exactly the point of the new ->integrity_read routine
I proposed and prototype.  The important thing is that it is called
with i_rwsem held because code mugh higher in the chain already
acquired it, but except for that it's entirely up to the file system.
--
To unsubscribe from this list: send the line "unsubscribe linux-security-module" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html