On (03/16/18 09:55), Petr Mladek wrote: [..] > I am not sure if it is worth it. I think that we would catch 99% of > problems by checking the first byte. > > This patch was motivated by a code clean up rather than bug reports.
OK. Then I think we really need this "the patch is just good enough" line in the commit message and a big comment in the source code. Another idea (just an idea) - for some pointers we know the address range we are going to access and can check the first and the last byte. E.g. for UUID it's check_access(ptr) and check_access(ptr + len), and so on. Won't work for string() in general case, tho. -ss