Re: [PATCH] mm/vmalloc: fix mischeck pfn valid in vmap_pfns

[Gao Xiang <hsiangkao@linux.alibaba.com>]

On 2025/3/18 16:44, Christoph Hellwig wrote:
> On Tue, Mar 18, 2025 at 04:39:40PM +0800, Huan Yang wrote:
>> A folio may be 2MB or more large 1GB, what if we only need a little, 1M or 512MB, can vmap based on folio can solve it?
> 
> Then you only map part of it by passing a length argument.  Note that
> in general when you have these large folios you also don't have highmem,
> so if you only map one of them, or part of one of them you don't actually
> need vmap at all and can just use folio-address..
> 
>> Normally, can offer 4k-page based array map it. But consider HVO, can't. That's why wanto base on pfn.
> 
> Well, for any large folio using this 4k based page interface is
> actually highly inefficient.  So let's fix that.  And my loop in
> willy as Mr. Folio while you're at it.

The minimum map unit is page size instead of variable-size folio.

For many cases, vmap (to combine many partial folios) is useful
(instead of split all folios to order-0 folios in advance) but
I agree page array may be inefficient.

So I don't think another folio vmap() version is better overall
anyway.

Thanks,
Gao Xiang