On 2024/8/7 0:59, Catalin Marinas wrote:
> On Tue, Aug 06, 2024 at 07:33:20PM +0800, Jinjie Ruan wrote:
>> Since commit 282c3a66b724 ("crash: Fix riscv64 crash memory reserve dead
>> loop"), if reservation from the high memory failed on ARM64, the kernel
>> will not falls back to searching the low memory, so remove it in the doc.
>
> This commit doesn't exist in -next. I found it with a different hash but
> don't add it in the commit log here.
>
>> diff --git a/Documentation/arch/arm64/kdump.rst b/Documentation/arch/arm64/kdump.rst
>> index 56a89f45df28..11b9b84bf422 100644
>> --- a/Documentation/arch/arm64/kdump.rst
>> +++ b/Documentation/arch/arm64/kdump.rst
>> @@ -79,10 +79,6 @@ To reserve memory for crashkernel=size,high, searching is first
>> attempted from the high memory region. If the reservation succeeds, the
>> low memory reservation will be done subsequently.
>>
>> -If reservation from the high memory failed, the kernel falls back to
>> -searching the low memory with the specified size in crashkernel=,high.
>> -If it succeeds, no further reservation for low memory is needed.
>
> I recall long discussions over a year ago where the conclusion was that
> for sysadmins it's easier to have crashkernel=,high the default with
> fallback to lowmem. No need to worry about how much low or high memory
> there is on a SoC, just specify a preference for high memory.
>
> Can we not have a different fix for the infinite loop problem while we
> preserve the fallback behaviour?
Of course, thank you!
>
