On Tue, Oct 09, 2012 at 11:37:06AM -0600, Jason Gunthorpe wrote:

On Mon, Oct 08, 2012 at 11:24:13AM +0100, Dave Martin wrote:

quoted

Partly this came from some side speculation about whether we could do
things like privileged read-only permissions on newer CPUs, for preventing
unintended or undesired writes to the kernel's code or read-only data.

Some other arches page protect the kernel, but that tends to be at
odds with the desire to use huge pages for the kernel mapping, and
independent of the load headers..

This wasn't so much about that headers themselves as about fragmentation
of the page permissions which makes it difficult to map everything using
huge pages / sections.  But as you say, there are conflicting concerns
here, and it seems not to be a priority.

Privileged write-protect is nice to have if non-disruptive, but not
essential.

Cheers
---Dave