On 02/06/2012 08:21 AM, Toshikazu Nakayama wrote:
> Suzuki-San,
>
>>>> + for (; tmp->name!= NULL; tmp++)
>>>> + if (!strcmp(tmp->name, platform))
>>>> + return tmp;
>>>
>>> Is there any reasons to distinct "ppc440gp"?
>>>
>> No specific reasons, apart from just the name difference.
>>
>>> Since I see that there are no "ppc440" prefix platform other than PPC40
>>> in latest cpu_specs[], and all of ppc440gp's values in ppc_boards[] are the
>>> same as ppc440.
>>> Can it be summarized by using STRNEQ("ppc440") or strncmp("ppc440")?
>> The problem is, this is generic code. And there could be a different platforms
>> which start with the same prefix and having different settings for the page
>> bits, (e.g, freescale processors). So generalizing this would be a problem.
>
> I can understand your careful thought, thanks.
>
>> One option I could think of is defining a function rather than a 'name' for each
>> platform to identify the 'running platform' as its variant.
>>
>> i.e,
>>
>> struct platform {
>> int (*check_platform) (char *s);
>> const char *name;
>> ...
>> }
>>
>> Where we could invoke the check_platform() for each platform, instead of
>> doing a check ourselves. The check_platform() could be implemented by each
>> distinct platforms and could use one board definition for multiple platform
>> values. This would also help us to reduce the 'platform' definitions as different
>> non-related platforms may also have the same set of definitions.
>>
>> So the loop would look like :
>>
>> for (; tmp->name!= NULL; tmp++)
>> if (tmpe->check_platform(platform))
>> return tmp;
>>
>> What do you think about this approach ?
>
> I can imagine that ppc_boards[] own least entries "default", "ppc440", "ppc??",
> (or declare as "ibm", "freescale", ..)
> and call check_ppc440_platform("platform"), check_ppc??_platform("platform")
> to check or update page bits.
>
> The check_ppc440_platform() can do with SRTEQ("platform", "ppc440(gp)") easily
> but check_ppc??_platform() has to check more conditions if prefixed name
> has different settings for the page bits.
> Finaly check_platform sometimes update page talbe bit members in struct platform.
>
> This is interesting so I wait for your updates
> even though my imagination is different from yours.
Sounds good. I will spin a new version with these changes.
Thanks
Suzuki
--
Crash-utility mailing list
Crash-utility@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/crash-utility
