I like this proposal better. I am still not convinced that we really
need internal __GFP_TRYLOCK though.
If we reduce try_alloc_pages to the gfp usage we are at the following
On Tue 17-12-24 19:07:14, alexei.starovoitov@xxxxxxxxx wrote:
[...]
> +struct page *try_alloc_pages_noprof(int nid, unsigned int order)
> +{
> + gfp_t alloc_gfp = __GFP_NOWARN | __GFP_ZERO |
> + __GFP_NOMEMALLOC | __GFP_TRYLOCK;
> + unsigned int alloc_flags = ALLOC_TRYLOCK;
[...]
> + prepare_alloc_pages(alloc_gfp, order, nid, NULL, &ac,
> + &alloc_gfp, &alloc_flags);
[...]
> + page = get_page_from_freelist(alloc_gfp, order, alloc_flags, &ac);
> +
> + /* Unlike regular alloc_pages() there is no __alloc_pages_slowpath(). */
> +
> + trace_mm_page_alloc(page, order, alloc_gfp & ~__GFP_TRYLOCK, ac.migratetype);
> + kmsan_alloc_page(page, order, alloc_gfp);
[...]
