Re: [RFC PATCH] mm, vmalloc: use __GFP_HIGHMEM implicitly

[Vlastimil Babka]

On 02/01/2017 03:05 PM, Michal Hocko wrote:
> From: Michal Hocko <mhocko@xxxxxxxx>
>
> __vmalloc* allows users to provide gfp flags for the underlying
> allocation. This API is quite popular
> $ git grep "=[[:space:]]__vmalloc\|return[[:space:]]*__vmalloc" | wc -l
> 77
>
> the only problem is that many people are not aware that they really want
> to give __GFP_HIGHMEM along with other flags because there is really no
> reason to consume precious lowmemory on CONFIG_HIGHMEM systems for pages
> which are mapped to the kernel vmalloc space. About half of users don't
> use this flag, though. This signals that we make the API unnecessarily
> too complex.
>
> This patch simply uses __GFP_HIGHMEM implicitly when allocating pages to
> be mapped to the vmalloc space. Current users which add __GFP_HIGHMEM
> are simplified and drop the flag.
>
> Signed-off-by: Michal Hocko <mhocko@xxxxxxxx>
> ---
> Hi,
> this is based on top of [1]. I believe it was Al who has brought this
> up quite some time ago (or maybe I just misremember). The explicit
> usage of __GFP_HIGHMEM in __vmalloc* seems to be too much to ask from
> users. I believe there is no user which doesn't want vmalloc pages be
> in the highmem but I might be missing something. There is vmalloc_32*
> API but that uses GFP_DMA* explicitly which overrides __GFP_HIGHMEM. So
> all current users _should_ be safe to use __GFP_HIGHMEM unconditionally.
> This patch should simplify things and fix many users which consume
> lowmem for no good reason.
>
> I am sending this as an RFC to get some feedback, I even haven't compile
> tested it yet.
>
> Any comments are welcome.

The idea sounds good. What are the potential dangers? That somebody of the 
current callers without __GFP_HIGHMEM would take a physical address of the page 
and then tried to access it via direct mapping?