On Mon, Mar 22, 2021 at 07:58:50AM +0300, Dan Carpenter wrote: > On Fri, Mar 19, 2021 at 10:24:43PM -0400, Alan Stern wrote: > > On Sat, Mar 20, 2021 at 09:31:51AM +0800, Jay Fang wrote: > > > From: Zihao Tang <tangzihao1@xxxxxxxxxxxxx> > > > > > > Fix the following smatch warnings: > > > > > > drivers/usb/host/ohci-hcd.c:1318 ohci_hcd_mod_init() warn: > > > ignoring unreachable code. > > > > > > platform_driver_register(&TMIO_OHCI_DRIVER) is the last > > > platform_driver_register() call in ohci_hcd_mod_init(), so if it > > > failed, there's no need to unregister it, but just goto error_tmio. > > > > > > So remove the unreachable platform_driver_unregister(&TMIO_OHCI_DRIVER). > > > No functionality change. > > > > Doesn't the compiler realize that the call is unreachable, and > > therefore avoid generating any object for it? > > > > This is a static checker warning. For example, Heart Bleed bug was an > ignored unreachable code static checker warning. > > > It's true that the function call is, strictly speaking, unnecessary. > > However, it provides a pleasing symmetry and it acts as a guide in the > > unlikely event that anyone wants to add another platform-specific > > driver in the future. > > Hopefully future programmers can figure out basic stuff like that. > > This code doesn't trigger a Smatch warning on my .config because the > Smatch check doesn't warn if the previous line was an #endif. On the > other hand, the ifdefs are also why I forwarded the email when I saw the > warning from kbuild. Normally kbuild is better at picking the person > to blame but because this is a .config thing that confused it. Anyway, > I glanced at the warning and thought it looked suspicious enough to > warrant a further look. > > When I first wrote the Smatch unreachable code warning there were a > handful of places which used that style of code: > > return 0; > > unreachable_release(); > err_release: > release_something(); > > I just left those as-is because it was obvious to me that it was done > intentionally. However, it seems that other people have removed all of > those behind my back so I can't find an example of this now except for > in ohci_hcd_mod_init(). > > Anyway, I would have put a special case to silence these false positives > but it wasn't common practice in 2014 and no one does it these days. Well, I guess it would be okay to convert the code to a comment. Just so long as it isn't removed entirely. Alan Stern
