[Ce-android-mainline] ram console proposal update and questions

[Colin Cross]

On Sun, Feb 19, 2012 at 12:12 PM, Shuah Khan <shuahkhan at gmail.com> wrote:
> On Sun, Feb 19, 2012 at 11:53 AM, Colin Cross <ccross at android.com> wrote:
>> On Sun, Feb 19, 2012 at 10:34 AM, Shuah Khan <shuahkhan at gmail.com> wrote:
>>> I started testing android ram console driver on x86.
>>>
>>> The location I chose is drivers/char - please let me know if there is
>>> a better suitable location.
>>> I started out with error leg testing by leaving EARLY_SIZE as at its
>>> default value of 0. As expected it failed.
>>> However, I expected it to fail with an error message that indicates
>>> that the EARLY_SIZE is invalid. I plan to
>>> make necessary changes to indicate what is really wrong.
>>>
>>> [    0.000000] ram_console: buffer           (null), invalid size 0,
>>> datasize 18446744073709551604
>>>
>>> What are some of the values used on Android devices for this config parameter?
>>
>> We tend not to use CONFIG_ANDROID_RAM_CONSOLE_EARLY_INIT at all.  For
>> early crash problems, CONFIG_DEBUG_LL is easier to use.  In fact, in a
>> big rewrite of ram_console for our android-3.3 tree (which should be
>> published soon), I dropped early ramconsole support completely.
>>
>> For size, we use something between 128KB and 2MB.  128KB is plenty to
>> get a few panics without losing context on what was happening before
>> the panic, 2MB is enough to store a couple calls to sysrq-w or sysrq-t
>> when debugging a deadlock.
>
> Colin,
>
> Do you mind sharing the latest source if it is ready so, I can start
> using it as a base for integration into the mainline. Thanks for your
> tips. I will stop testing EARLY_INIT. What about
> RAM_CONSOLE_ERROR_CORRECTION code? Is that staying in the latest
> re-write?

ECC is still there, but I took out all the config options.  In the
future drivers may want to customize the ECC options, but for now I
just defaulted them to sensible values (the same as the current
default values).