Re: [RFC 5/6] iomap: refuse dmabuf iterators on contents-encrypted inodes

[Christoph Hellwig]

I think contents-encrypted should be content-encrypted.

But we really need to find a way to make these checks more
systematic instead of sprinkling them everywhere.  E.g. have a flag
in file->f_mode to allow it, which then only leaves the direct I/O
to buffered I/O fallback as a special case.