Re: [PATCH 1/6] staging: sm750fb: Use memset_io instead of memset

[Dan Carpenter <dan.carpenter@oracle.com>]

On Wed, Mar 11, 2015 at 09:11:52AM +0000, Lorenzo Stoakes wrote:
> On 11 March 2015 at 08:54, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> > When I see a patch like this, then I worry, "What if the Sparse
> > annotations are wrong?  The patch description doesn't say anything about
> > that."  After review then I think the annotations are correct so that's
> > fine.
> 
> How do you mean? I was careful to check what sparse was referring to,
> then investigate how memset should be used with pointers with a
> __iomem qualifier. I'd like to be able to improve my patch
> descriptions going forward as best I can :)
> 

Yes.  The patch is correct.  I wasn't asking you to redo it.  From later
patches it's actually clear that you know that this change is a bugfix
and a behavior change.  But we get a lot of patches where people just
randomly change things to please Sparse and it maybe silences a warning
but it's not correct.  I can think of a few recentish examples where
people used standard struct types which hold __iomem or __user pointers
but they used them in non-standard ways so the pointers were actually
normal kernel pointers.

I guess the rule here is that the patch should explain the effect of the
bugfix for the user.  Often you won't know the effect, but it's a
helpful thing to think about.

> > Btw, do you have this hardware?  Are you able to test these changes?
> 
> Unfortunately not, I am trying to keep these changes as simple code
> fixes that ought not to affect actual hardware behaviour as I can
> (though of course you can never be entirely sure that's the case!)

That's fine.  I was just wondering.  It affects how paranoid I am when I
review the code.

regards,
dan carpenter