Re: [RFC PATCH 4/5] EXPERIMENTAL: x86: use __GFP_UNMAPPED for modele_alloc()

[Edgecombe, Rick P]

On Wed, 2023-03-08 at 11:41 +0200, Mike Rapoport wrote:
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index ef910bf349e1..84220ec45ec2 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2892,7 +2892,7 @@ vm_area_alloc_pages(gfp_t gfp, int nid,
>          * to fails, fallback to a single page allocator that is
>          * more permissive.
>          */
> -       if (!order) {
> +       if (!order && !(gfp & __GFP_UNMAPPED)) {
>                 gfp_t bulk_gfp = gfp & ~__GFP_NOFAIL;
>  
>                 while (nr_allocated < nr_pages) {

This is obviously a quick POC patch, but I guess we should skip the
whole vm_remove_mappings() thing since it would reset the direct map to
RW for these unmapped pages. Or rather modules shouldn't set
VM_FLUSH_RESET_PERMS if it uses this.