[PATCH v3 3/5] kdump, documentation: describe craskernel CMA reservation

[David Hildenbrand]

On 14.03.25 04:18, Baoquan He wrote:
> Hi Jiri,
> 
> On 03/12/25 at 10:09pm, Jiri Bohac wrote:
> ......
>> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
>> index fb8752b42ec8..895b974dc3bb 100644
>> --- a/Documentation/admin-guide/kernel-parameters.txt
>> +++ b/Documentation/admin-guide/kernel-parameters.txt
>> @@ -987,6 +987,28 @@
>>   			0: to disable low allocation.
>>   			It will be ignored when crashkernel=X,high is not used
>>   			or memory reserved is below 4G.
>> +	crashkernel=size[KMG],cma
>> +			[KNL, X86] Reserve additional crash kernel memory from
>> +			CMA. This reservation is usable by the first system's
>> +			userspace memory and kernel movable allocations (memory
>> +			balloon, zswap). Pages allocated from this memory range
>> +			will not be included in the vmcore so this should not
>> +			be used if dumping of userspace memory is intended and
>> +			it has to be expected that some movable kernel pages
>> +			may be missing from the dump.
> 
> Since David and Don expressed concern about the missing kernel pages
> allocated from CMA area in v2, and you argued this is still useful for
> VM system, I would like to invite David to help evaluate the whole
> series if it's worth from the VM and MM point of view.

Balloon pages will not be dumped either way (PageOffline), so that is 
not a convern.

Zsmalloc pages ... are probably fine right now. They should likely only 
be storing compressed user data. (not sure if they also store some other 
datastructures, I think no, but might be wrong)

My comment was rather forward-looking: that CMA memory only contains 
user space memory is already not the case (but the existing cases might 
be okay). In the future, as we support other movable allocations (as 
raised, leaf page tables at some point, and there were discussions about 
movable slab pages, although that might be challenging) this can change 
(unless we find ways of not placing these allocations on CMA memory).

So as is, this should be fine, but it's certainly something to be aware 
of in the future.

-- 
Cheers,

David / dhildenb