Re: CFI violation in drivers/infiniband/core/sysfs.c

[Nathan Chancellor]

On Tue, May 04, 2021 at 05:22:22PM -0300, Jason Gunthorpe wrote:
> On Fri, Apr 02, 2021 at 11:55:59PM -0700, Nathan Chancellor wrote:
> > > So, I think, the solution is below. This hasn't been runtime tested. It
> > > basically removes the ib_port callback prototype and leaves everything
> > > as kobject/attr. The callbacks then do their own container_of() calls.
> > 
> > Well that appear to be okay from a runtime perspective.
> 
> This giant thing should fix it, and some of the other stuff Greg observed:
> 
> https://github.com/jgunthorpe/linux/commits/rmda_sysfs_cleanup
> 
> It needs some testing before it gets posted
> 
> Jason

I have verified that my original test case of running LTP's read_all
test case passes with CFI enabled in enforcing mode with your series and
I still get values when running cat on them. If you have any other
suggestions for test cases, please let me know :)

Thanks for the quick fix!

Cheers,
Nathan