On Thu, Nov 25 2021 at 16:15, Thomas Gleixner wrote: > On Tue, Aug 03 2021 at 21:32, ira weiny wrote: > Aside of that, the function which set's up the init value is really > bogus. As you explained in the cover letter a kernel user has to: > > 1) Claim an index in the enum > 2) Add a default value to the array in that function > > Seriously? How is that any better than doing: > > #define PKS_KEY0_DEFAULT PKR_RW_ENABLE > #define PKS_KEY1_DEFAULT PKR_ACCESS_DISABLE > ... > #define PKS_KEY15_DEFAULT PKR_ACCESS_DISABLE > > and let the compiler construct pkrs_init_value? > > TBH, it's not and this function has to be ripped out in case that you > need a dynamic allocation of keys some day. So what is this buying us > aside of horrible to read and utterly pointless code? And as Taoyi confirmed its broken. It surely takes a reviewer to spot that and an external engineer to run rdmsr -a to validate that this is not working as expected, right? Sigh...
