[U-Boot-Users] [PATCH] 85xx: extended cpu identification

[Kim Phillips]

On Tue, 10 Jun 2008 10:10:02 -0500
Kumar Gala <galak at kernel.crashing.org> wrote:

> 
> On Jun 10, 2008, at 8:48 AM, Kim Phillips wrote:
> 
> > On Tue, 10 Jun 2008 08:23:46 -0500
> > Kumar Gala <galak at kernel.crashing.org> wrote:
> >
> >>
> >> On Jun 10, 2008, at 12:57 AM, Kim Phillips wrote:
> >>
> >>> On Thu, 29 May 2008 03:20:08 -0500 (CDT)
> >>> Kumar Gala <galak at kernel.crashing.org> wrote:
> >>>
> >>>> +struct cpu_type cpu_type_list [] = {
> >>>> +	CPU_TYPE_ENTRY(8533, 8533, 0),
> >>>> +	CPU_TYPE_ENTRY(8533, 8533_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8540, 8540, 0),
> >>>> +	CPU_TYPE_ENTRY(8541, 8541, 0),
> >>>> +	CPU_TYPE_ENTRY(8541, 8541_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8543, 8543, 0),
> >>>> +	CPU_TYPE_ENTRY(8543, 8543_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8544, 8544, 0),
> >>>> +	CPU_TYPE_ENTRY(8544, 8544_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8545, 8545, 0),
> >>>> +	CPU_TYPE_ENTRY(8545, 8545_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8547, 8547_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8548, 8548, 0),
> >>>> +	CPU_TYPE_ENTRY(8548, 8548_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8555, 8555, 0),
> >>>> +	CPU_TYPE_ENTRY(8555, 8555_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8560, 8560, 0),
> >>>> +	CPU_TYPE_ENTRY(8567, 8567, 0),
> >>>> +	CPU_TYPE_ENTRY(8567, 8567_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8568, 8568, 0),
> >>>> +	CPU_TYPE_ENTRY(8568, 8568_E, CPU_FTRS_HAS_CRYPTO),
> >>>> +	CPU_TYPE_ENTRY(8572, 8572, 0),
> >>>> +	CPU_TYPE_ENTRY(8572, 8572_E, CPU_FTRS_HAS_CRYPTO),
> >>>
> >>> this seems like overkill given all we have to do is check one bit  
> >>> (see
> >>> IS_E_PROCESSOR macro in "handle crypto node" patch I just sent out).
> >>
> >> I don't trust our HW guys to keep w/that convention.  Plus we can use
> >> this mechanism for other things if need be.
> >
> > they've been pretty good so far, and until the other thing comes
> > around, we can compact this table instead of expanding it - the 85xx_E
> > entries themselves are unnecessary.
> 
> I disagree and would prefer to keep it as I've done as it provides the  
> most flexibility.

it's easier to do a IS_E_PROCESSOR(get_svr()) from cpu/mpc85xx/fdt.c
instead of getting ver, calling cpu = identify_cpu(ver), and then
checking cpu->features every time.  I don't know what other features
you have in mind (perhaps this patch should wait until then?), but
HAS_CRYPTO is either on or off, and the bit already exists in the h/w..

Kim