[U-Boot] [PATCH 6/8] cpu: Add cpu_print_info function

[Simon Glass]

Hi Mario,

On 26 April 2018 at 00:07, Mario Six <mario.six at gdsys.cc> wrote:
> Hi Simon,
>
> On Tue, Apr 24, 2018 at 11:53 PM, Simon Glass <sjg at chromium.org> wrote:
>> Hi Mario,
>>
>> On 19 April 2018 at 01:50, Mario Six <mario.six at gdsys.cc> wrote:
>>>
>>> Hi Simon,
>>>
>>> On Wed, Apr 18, 2018 at 5:45 PM, Simon Glass <sjg at chromium.org> wrote:
>>> > Hi Mario,
>>> >
>>> > On 18 April 2018 at 02:35, Mario Six <mario.six at gdsys.cc> wrote:
>>> >> Hi Simon,
>>> >>
>>> >> On Thu, Apr 12, 2018 at 6:37 PM, Simon Glass <sjg at chromium.org> wrote:
>>> >>> Hi Mario,
>>> >>>
>>> >>> On 11 April 2018 at 00:39, Mario Six <mario.six at gdsys.cc> wrote:
>>> >>>> Hi Simon,
>>> >>>>
>>> >>>> On Fri, Mar 30, 2018 at 10:41 AM, Simon Glass <sjg at chromium.org> wrote:
>>> >>>>> Hi,
>>> >>>>>
>>> >>>>> On 28 March 2018 at 20:38, Mario Six <mario.six at gdsys.cc> wrote:
>>> >>>>>> Add a cpu_print_info function to the CPU uclass to emulate the behavior
>>> >>>>>> of some current non-DM drivers (e.g. MPC83xx) to print CPU information
>>> >>>>>> during startup.
>>> >>>>>>
>>> >>>>>> Signed-off-by: Mario Six <mario.six at gdsys.cc>
>>> >>>>>> ---
>>> >>>>>>  drivers/cpu/cpu-uclass.c | 10 ++++++++++
>>> >>>>>>  include/cpu.h            | 15 +++++++++++++++
>>> >>>>>>  2 files changed, 25 insertions(+)
>>> >>>>>>
>>> >>>>>
>>> >>>>> I really don't want drivers printing stuff. Can you use the existing
>>> >>>>> get_info() method?
>>> >>>>>
>>> >>>>
>>> >>>> I could, but I'm just emulating the behavior of the legacy code here, which
>>> >>>> prints some information when the CPU is initialized. I think that's pretty
>>> >>>> useful, and I can do that in our board file, but that would mean implementing
>>> >>>> the same routine in every MPC83xx board to get the legacy behavior back.
>>> >>>
>>> >>> Yes, but I don't want the legacy code creeping into the eclass. Can
>>> >>> you convert the board to use the CPU eclass instead?
>>> >>>
>>> >>
>>> >> That's what I did, and I just discovered DISPLAY_CPUINFO, which does exactly
>>> >> what is needed. I'll implement the print_cpuinfo function in the CPU driver, so
>>> >> I can get rid of the print function in the uclass (and still retain the
>>> >> information printing at bootup).
>>> >
>>> > OK I see. Ideally we would have a function (perhaps in board_f) which
>>> > prints out the CPU info after obtaining it from the uclass. So could
>>> > you move your print_cpuinfo() function into board_f? Would it be
>>> > possible to use that if CONFIG_CPU is defined?
>>> >
>>> > At some point print_cpuinfo() could be removed from various board files.
>>> >
>>>
>>> The function prints the following (example for our device):
>>>
>>> --- >8 ---
>>>
>>> Reset Status: External/Internal Soft, External/Internal Hard
>>>
>>> CPU:   e300c3, MPC8308, Rev: 1.1 at 400 MHz, CSB: 133.333 MHz
>>>
>>> --- >8 ---
>>>
>>> So there are some values that are very specific to the platform, such as the
>>> CSB (Coherent System Bus) frequency, or the reset status. While it's probably
>>> possible to put some of that into a generic info printing routine, lots of it
>>> is so MPC83xx-specific that it doesn't make much sense for other CPUs.
>>
>> Well you get to provide a string from get_desc() so you can add
>> whatever you like!
>>
>
> OK, I thought I was supposed to decode the cpu_info and print that in the
> function. If i can use get_desc(), then it's fine. :-)
>
>>>
>>> Any idea how to tackle this? I don't really want to get rid of the Reset Status
>>> printing especially, since it's pretty useful information for debugging devices
>>> in the field.
>>
>> Re the reset status, shouldn't that come from the sysreset driver? I
>> wonder if we should add an API call for that? Or is it something
>> printed by the board?
>>
>
> No, that's indeed printed on every MPC83xx board (see prt_83xx_rsr in
> arch/powerpc/cpu/mpc83xx/cpu_init.c). A sysreset driver would probably be a
> good canonical candidate for printing such a message, true. But since there is
> no real system reset on MPC83xx to speak of, that would (for now at least) be
> the only thing a MPC83xx sysreset driver did. Also, we'd need another function
> in board_f.c, and I don't really know if reset cause printing is very
> wide-spread on other platforms to justify that.

Well I've certainly printed this info myself so I think it is useful,
particularly things like whether it was a watchdog reset or user
reset.

Regards,
Simon