On Thu, Aug 19, 2010 at 01:09:15PM +0200, Arnd Bergmann wrote:

On Wednesday 18 August 2010, Stephen Boyd wrote:

quoted

So the only sticking point now is that x86, parisc, and arm use warnings 
and errors but s390 only uses warnings. I guess I'll reword it to be:

        Enabling this option turns a certain set of sanity checks for
        user copy operations into compile time warnings/errors.

        The copy_from_user() etc checks are there to help test if there
        are sufficient security checks on the length argument of the
        copy operation, by having gcc prove that the argument is
        within bounds.

        If unsure, or if you run an older (pre 4.4) gcc where this
        option is a no-op, say N.

or I'll add a patch to make s390 trigger an error when this is enabled?

(Taking Martin and Heiko on Cc for s390)

I'd strongly suggest making the behavior the same for everyone. It should
be fairly easy to make sure none of these warnings ever triggers
on s390, because most of the Linux device driver code does not get build
there anyway.

Please don't do that. An s390 allyesconfig still triggers 45 warnings and
I'm currently not willing to "patch" working code just to get rid of these
warnings which are most likely all false positives.
That's the reason why we currently don't error out and only generate
warnings.