[RFC] staging:dgap: sparse errors/warnings

Dan Carpenter dan.carpenter at oracle.com
Thu Mar 27 23:52:30 UTC 2014


On Thu, Mar 27, 2014 at 09:22:36AM -0400, Mark Hounschell wrote:
> checkpatch now gives 0 warnings or errors on both dgap.c and dgap.h
> so I'm now looking at sparse errors/warnings. There are many and I'm
> not really clear as to the best way to address them. This is my first 
> exposure to sparse. A quick internet search reveals that using "__iomem"
> in some way is the proper fix for most of what I see?

Looks fine.  You understand the point, right?  __iomem is hardware
addresses.  Kernel memory, hardware memory and user memory are mapped in
different places.  You can't dereference an iomem pointer or even a user
pointer.  They have special functions so you can write to them.  A
common source of bugs is forgetting which is which.

The sparse annotations are compiled out while __CHECKER__ isn't defined.

regards,
dan carpenter



More information about the devel mailing list