Re: [PATCH 0/2] acpi, x86: Add SPCR table support

[Prarit Bhargava]

On 12/08/2017 10:31 AM, Jeffrey Hugo wrote:
> On 12/8/2017 7:29 AM, Prarit Bhargava wrote:
>>
>>
>> On 12/08/2017 01:29 AM, Ingo Molnar wrote:
>>>
>>> * Prarit Bhargava <prarit@xxxxxxxxxx> wrote:
>>>
>>>> The SPCR (Serial Port Console Redirection) Table provides information
>>>> about the configuration of serial port.  This information can be used
>>>> to configure the early console.
>>>
>>> s/about the configuration of serial port
>>>   /about the configuration of the serial port
>>>
>>>> SPCR support was added for arm64 and is made available across all arches
>>>> in this patchset.  The first patch adds a weak per-arch configuration function
>>>> and moves the SPCR code into ACPI.  The second patch adds support to x86.
>>>>
>>>> The existing behaviour on arm64 is maintained.  If the SPCR exists the
>>>> earlycon and console are automatically configured.
>>>
>>> s/arm64
>>>   /ARM64
>>>
>>> which is easier to read and it's also the prevalent spelling:
>>>
>>>   triton:~/tip> for N in $(git grep -ih arm64 arch/arm64/ | sed
>>> 's/[[:punct:]]/ /g'); do echo $N | grep -iw arm64; done | sort | uniq -c
>>>       412 arm64
>>>         1 Arm64
>>>       854 ARM64
>>>
>>>> The existing default behaviour on x86 is also maintained.  If no console or
>>>> earlycon parameter is defined and the SPCR exists , the serial port is not
>>>> configured.  If the earlycon parameter is used both the early console
>>>> and the console are configured using the data from the SPCR.
>>>
>>> s/exists , the
>>>   /exists, the
>>>
>>> But, the logic to not use the SPCR looks confusing to me.
>>>
>>> The SPCR is only present if the user has explicitly configured a serial console
>>> for that machine, either in the firmware, or remotely via IPMI, correct? I.e.
>>> SPCR
>>> will not be spuriously present by default on systems that have a serial console
>>> but the user never expressed any interest for them, right?
>>
>> If I disable "Serial Port Console Debug" in my BIOS I still see the SPCR
>> configured:
>>
>> [root@prarit-lab ~]# dmesg | grep SPCR
>> [    0.000000] ACPI: SPCR 0x0000000069031000 000050 (v01
>> 00000000      00000000)
>>
>> AFAICT the SPCR is always enabled on some systems.
> 
> "Serial Port Console Debug" sounds like DBG2 to me, although I am unsure of the
> specific platform you are referencing.
> 
>>
>>>
>>> If so then we should pick up that serial console configuration and activate it,
>>> regardless of any kernel boot options!
>>
>> I'm worried about someone who doesn't want a console on ttyS0 suddenly ending up
>> with one.  The SPCR could contain incorrect data, etc.
>>
>> I originally wanted this on by default, but the chances of breaking someone's
>> setup seems significant doesn't it?  Maybe I'm being too cautious.  Anyone else
>> want to weigh in?  I'm not ignoring your idea Ingo, I'm just worried about being
>> yelled at by a user :) because I broke their default console setup.
>>
> 
> SPCR is always present on QDF2400 (ARM64 platform).  Firmware will automatically
> update the SPCR table to point to the correct console (IE Serial-over-LAN or SOL
> if configured via IPMI).  Unless a user specifically chooses to override the
> SPCR via "console=", we expect that a console will be present based on the data
> in SPCR.

Hey Jeffrey -- I think Ingo's & my conversation is x86-specific.  I am *NOT*
changing the behaviour on ARM64.

P.

>