Re: [PATCH] Revert "i2c: dev: switch from register_chrdev to cdev API"

[Dan Carpenter]

On Sat, May 28, 2016 at 06:04:58PM +0200, Wolfram Sang wrote:
> 
> > These kinds of use after frees don't always cause a runtime problem.  I
> > found it using static analysis, but I bet you could detect it if you
> > enabled kasan.  The other option is to enable PAGE_POISONING?
> 
> Thanks, PAGE_POISONING did trigger the issue. So, I now picked up your
> patch instead of the revert because I could verify the problem and the
> proper solution. Thanks again.
> 
> What I still wonder: Which analysis reported the problem to you? I
> always run sparse, smatch, cppcheck, and coccicheck on the patches when
> applying and no-one reported the issue.

It's a Smatch warning but you have to build the cross function db to
detect this.  It takes a while (a few hours) but the command is simple
enough.

	./smatch_scripts/build_kernel_data.sh

regards,
dan carpenter