On Thu, Jan 12, 2017 at 10:07:33AM -0600, Benjamin Herrenschmidt wrote: > On Thu, 2017-01-12 at 17:00 +0100, Greg KH wrote: > > > How so ? I mean, access_ok followed by __get/__put_user is still a > > > classic, what's wrong with it ? > > > > No "normal" driver should do that, just call copy_to/from_user and be > > done with it. That way all of the proper locking and validation checks > > like this are done correctly for you. Why would a driver ever call the > > "raw" __get/__put_user functions? > > I supposed historically it was considered faster for some things :-) > > Not a huge deal, and yes it's probably cleaner, I was just wondering > what was "racy" about access_ok() that I might have missed... I think, you can change things after access_ok() happens, there used to be bugs in that area a few years ago. I think we fixed them by moving the offending drivers to use copy_*() instead. thanks, greg k-h -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
