On Tue, Jan 14, 2020 at 04:01:53PM +0100, Borislav Petkov wrote: > On Tue, Jan 14, 2020 at 02:08:50PM +0000, Colin Ian King wrote: > > If I understand the question, it seems that get_builtin_microcode() > > tries to load in the appropriate amd microcode binary from the cpio data > > and this can potentially fail if the microcode is not provided for the > > specific processor family, so I believe this is a legitimate fix. > > If the microcode for the specific processor family is not provided, > get_builtin_firmware() will return false and then we'll call > find_microcode_in_initrd() which will definitely return either a proper > pointer or a NULL-initialized cpio_data struct. > > So I still don't see it. It's probably complaining that cp.name[] isn't initialized. UBSan will probably generate a warning at runtime when we do: *ret = cp; But otherwise it's harmless. regards, dan carpenter
