Re: [RESEND PATCH 1/3] x86: Adding structs to reflect cpuid fields

["Radim Krčmář" <rkrcmar@redhat.com>]

2014-09-17 16:06+0200, Borislav Petkov:
> On Wed, Sep 17, 2014 at 04:53:39PM +0300, Nadav Amit wrote:
> > AFAIK backward compatibility is usually maintained in x86. I did not
> > see in Intel SDM anything that says "this CPUID field means something
> > for CPU X and something else for CPU Y". Anyhow, it is not different
> > than bitmasks in this respect.
> 
> You still don't get my point: what are you going to do when
> min_monitor_line_size needs to be 17 bits all of a sudden?
> 
> Currently, you simply do an if-else check before using the respective
> mask and with your defined structs, you need to keep two versions:
> 
> union cpuid5_ebx_before_family_X {
>        struct {
>                unsigned int max_monitor_line_size:16;
>                unsigned int reserved:16;
>        } split;
>        unsigned int full;
> };
> 
> union cpuid5_ebx_after_family_X {
>        struct {
>                unsigned int max_monitor_line_size:17;
>                unsigned int reserved:15;
>        } split;
>        unsigned int full;
> };

New union wouldn't be very convenient if the change touched just a small
part of the register ... probably the best choice is using anonymous
elements like this,

  union cpuid5_ebx {
  	union {
  		struct {
  			unsigned int max_monitor_line_size:16;
  			unsigned int reserved:16;
  		};
  		struct {
  			unsigned int max_monitor_line_size_after_family_X:17;
  			unsigned int reserved_after_family_X:15;
  		};
  	} split;
  	unsigned int full;
  };

which would result in a similar if-else hack

  if (family > X)
  	ebx.split.max_monitor_line_size_after_family_X = 0
  else
  	ebx.split.max_monitor_line_size = 0

other options are
  ebx.split.after_family_X.max_monitor_line_size
or even
  ebx.split.max_monitor_line_size.after_family_X

Flat namespace is more flexible wrt. code.