Re: [akpm-mm:mm-unstable 62/89] include/linux/kexec.h:41:2: error: #error KEXEC_SOURCE_MEMORY_LIMIT not defined

[Eric DeVolder <eric.devolder@oracle.com>]

On 7/5/23 09:21, Eric DeVolder wrote:
> 
> 
> On 7/3/23 15:17, Andrew Morton wrote:
>> On Mon, 3 Jul 2023 13:34:21 -0500 Eric DeVolder <eric.devolder@oracle.com> wrote:
>>
>>>
>>>
>>> On 7/3/23 11:31, Andrew Morton wrote:
>>>> On Mon, 3 Jul 2023 10:31:34 -0500 Eric DeVolder <eric.devolder@oracle.com> wrote:
>>>>
>>>>> Here are three possible courses of action:
>>>>>     - do nothing; preserve existing behavior as intended by the series (even though problems 
>>>>> like this
>>>>> can occur)
>>>>>     - fix it with CRASH_DUMP selecting KEXEC to eliminate the problem
>>>>
>>>> This one, please.
>>>>
>>>> Or make CRASH_DUMP depend on KEXEC.
>>>>
>>>>>     - do not fix it but document the reason why this problem occurs in the commit message
>>>
>>> ok, great!
>>>
>>> I also have an unrelated fix for riscv, and this correction (for both arm and mips). Is the correct
>>> action to post a new v4 version? Or is there a different method?
>>
>> If the changes are small then I feel that individual fix patches are
>> kinder to people who have already reviewed the previous.
>>
>> But a new version is OK if that's more reliable.
> 
> Andrew, I chose to post a v4 since not all pieces have RB/AB's, in particular the first patch 
> introducing the consolidation.
> 
> Thank you!
> eric

Andrew, just a heads up, but in feedback from arm and s390 folks, I've discovered that relying upon 
olddefconfig alone is insufficient for equivalence. I've enhanced my test setup to include 
allnoconfig and allyesconfig and that has revealed a few more problems. It'll take about 6 hours to 
work through all the configs, and then I'll need to do some cleanup and another verification pass. 
So a v5 is forthcoming. It does seem like this will close a few nuanced holes though...

Thanks,
eric