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

[Matthew Wilcox]

On Tue, Mar 07, 2017 at 03:08:45PM -0800, Andrew Morton wrote:
> On Tue,  7 Mar 2017 15:10:20 +0100 Michal Hocko <mhocko@xxxxxxxxxx> wrote:
> 
> > __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.
> 
> hm.  What happens if a caller wants only lowmem pages?  Drivers do
> weird stuff...

That's not something drivers actually want ... they might want "only pages
under 4GB", which is why we have vmalloc_32(), but drivers don't really
care where the HIGHMEM / LOWMEM split is.  I suppose we might find some
cases where drivers have mistakenly used vmalloc() and "got away with it".