Re: [PATCH v2] mm/sparse: Remove sparse buffer pre-allocation mechanism

[David Hildenbrand (Arm)]

On 4/14/26 14:11, Mike Rapoport wrote:
> On Mon, Apr 13, 2026 at 09:02:43PM +0200, David Hildenbrand (Arm) wrote:
>>> -
>>> -void * __meminit sparse_buffer_alloc(unsigned long size)
>>> -{
>>> -	void *ptr = NULL;
>>> -
>>> -	if (sparsemap_buf) {
>>> -		ptr = (void *) roundup((unsigned long)sparsemap_buf, size);
>>> -		if (ptr + size > sparsemap_buf_end)
>>> -			ptr = NULL;
>>> -		else {
>>> -			/* Free redundant aligned space */
>>> -			if ((unsigned long)(ptr - sparsemap_buf) > 0)
>>> -				sparse_buffer_free((unsigned long)(ptr - sparsemap_buf));
>>> -			sparsemap_buf = ptr + size;
>>
>> This code implied, that we would usually get contiguous memory sections
>> (on the same node) to have contiguous PFNs in the SPARSE !VMEMMAP case.
>> That will no longer be the case (reliably).
>>
>> The code must be prepared to handle that, so I guess that is fine. And
>> we don't care that much about SPARSE !VMEMMAP.
> 
> I've started to hunt them down ;-P 

Nice :)

-- 
Cheers,

David