[PATCH 3/5] userns: Don't read extents twice in m_start

[Eric W. Biederman]

Joe Perches <joe at perches.com> writes:

> On Wed, 2017-11-01 at 06:08 -0500, Eric W. Biederman wrote:
>> I won't listen to checkpatch when it is wrong.
>
> Always a good idea.
>
> btw: what is checkpatch wrong about this time?

Well the way I was hearing the conversation was that there was a patch
that fixed a real bug, but it was wrong because checkpatch complained
about it.

So I don't even know if the warning is a problem.  But blocking bug
fixes because there is a warning certainly is.

If someone wants to change coding style in practice so that every
smp_rmb and every smp_wmb has detailed comments that everyone must
include they need to follow the usual rule and update the entire kernel
when making an interface change.  As that did not happen I don't see
any problems with incremental updates in the style the code is already
in.

Not that I will mind a patch that updates the code, but I am not going
to hold up a perfectly good bug fix waiting for one either.

Eric