On Thu, Apr 01, 2021 at 08:30:55AM +0200, Christoph Hellwig wrote: > On Wed, Mar 31, 2021 at 01:56:00PM +0200, Thomas Bogendoerfer wrote: > > +#define __get_user_nofault(dst, src, type, err_label) \ > > +do { \ > > + int __gu_err; \ > > + \ > > + __get_user_common(*((type *)(dst)), sizeof(type), \ > > + (__force type *)(src)); \ > > + if (unlikely(__gu_err)) \ > > + goto err_label; \ > > +} while (0) > > + > > + > > +static inline int __get_addr(unsigned long *a, unsigned long *p, bool user) > > +{ > > + if (user) > > + __get_user_nofault(a, p, unsigned long, fault); > > + else > > + __get_kernel_nofault(a, p, unsigned long, fault); > > + > > + return 0; > > + > > +fault: > > + return -EFAULT; > > +} > > Why can't these use plain old get_user and get_kernel_nofault? > You "optimize" away the access_ok / get_kernel_nofaul_allowed checks > here, but now use totally non-standard and possibly dangerous APIs. thanks, I was too deep into the macro wormhole... using standard API makes this even look nicer. Posting v3 in a few minutes. Thomas. -- Crap can work. Given enough thrust pigs will fly, but it's not necessarily a good idea. [ RFC1925, 2.3 ]