On Tue, May 15, 2018 at 4:06 PM, Andrey Ryabinin
<aryabinin@xxxxxxxxxxxxx> wrote:
>
> You could avoid 'if (!PageSlab())' check by adding page_kasan_tag_reset() into kasan_poison_slab().
>> @@ -526,6 +526,7 @@ struct page *cma_alloc(struct cma *cma, size_t count, unsigned int align,
>> }
>>
>> trace_cma_alloc(pfn, page, count, align);
>> + page_kasan_tag_reset(page);
>
>
> Why? Comment needed.
> diff --git a/mm/kasan/common.c b/mm/kasan/common.c
> index b8e0a8215021..f9f2181164a2 100644
> --- a/mm/kasan/common.c
> +++ b/mm/kasan/common.c
> @@ -207,18 +207,11 @@ void kasan_unpoison_stack_above_sp_to(const void *watermark)
>
> void kasan_alloc_pages(struct page *page, unsigned int order)
> {
> -#ifdef CONFIG_KASAN_GENERIC
> - if (likely(!PageHighMem(page)))
> - kasan_unpoison_shadow(page_address(page), PAGE_SIZE << order);
> -#else
> - if (!PageSlab(page)) {
> - u8 tag = random_tag();
> + if (unlikely(PageHighMem(page)))
> + return;
>
> - kasan_poison_shadow(page_address(page), PAGE_SIZE << order,
> - tag);
> - page_kasan_tag_set(page, tag);
> - }
> -#endif
> + page_kasan_tag_set(page, random_tag());
> + kasan_unpoison_shadow(page_address(page), PAGE_SIZE << order);
> }
>
> void kasan_free_pages(struct page *page, unsigned int order)
> As already said before no changes needed in kasan_kmalloc_large. kasan_alloc_pages() alredy did tag_set().
Will fix all in v2, thanks!
