diff mbox series

mm: vmalloc: correct use of __GFP_NOWARN mask in __vmalloc_area_node()

Message ID 20221216234659.493794-1-lstoakes@gmail.com (mailing list archive)
State New
Headers show
Series mm: vmalloc: correct use of __GFP_NOWARN mask in __vmalloc_area_node() | expand

Commit Message

Lorenzo Stoakes Dec. 16, 2022, 11:46 p.m. UTC
This function invokes warn_alloc() with __GFP_NOWARN set which is a
no-op. Set this flag _after_ this call so it is actually invoked, and
additionally remove a duplicate application of __GFP_NOWARN afterwards.

Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com>
---
 mm/vmalloc.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

--
2.38.1

Comments

Uladzislau Rezki Dec. 19, 2022, 12:09 p.m. UTC | #1
On Fri, Dec 16, 2022 at 11:46:59PM +0000, Lorenzo Stoakes wrote:
> This function invokes warn_alloc() with __GFP_NOWARN set which is a
> no-op. Set this flag _after_ this call so it is actually invoked, and
> additionally remove a duplicate application of __GFP_NOWARN afterwards.
> 
> Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com>
> ---
>  mm/vmalloc.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index ca71de7c9d77..9e30f0b39203 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -3031,7 +3031,7 @@ static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask,
>  	int ret;
> 
>  	array_size = (unsigned long)nr_small_pages * sizeof(struct page *);
> -	gfp_mask |= __GFP_NOWARN;
> +
>  	if (!(gfp_mask & (GFP_DMA | GFP_DMA32)))
>  		gfp_mask |= __GFP_HIGHMEM;
> 
> @@ -3051,10 +3051,12 @@ static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask,
>  		return NULL;
>  	}
> 
> +	gfp_mask |= __GFP_NOWARN;
> +
>
There are three more warn_alloc()s below. Those are explicitly disabled.
Could you please rework the patch and make it working also?

--
Uladzislau Rezki
Lorenzo Stoakes Dec. 19, 2022, 12:25 p.m. UTC | #2
On Mon, Dec 19, 2022 at 01:09:41PM +0100, Uladzislau Rezki wrote:
> There are three more warn_alloc()s below. Those are explicitly disabled.
> Could you please rework the patch and make it working also?

Ah yes, sorry missed those, I think then the simplest solution then would be to
simply yank the __GFP_NOWARN specifier at the top of the function and restore
the explicit flag being passed to vm_area_alloc_pages() you added in
c3d77172dfc0.

This whole thing does make me wonder if this function is due a more general
refactor.

Will spin up a v2!
diff mbox series

Patch

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index ca71de7c9d77..9e30f0b39203 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -3031,7 +3031,7 @@  static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask,
 	int ret;

 	array_size = (unsigned long)nr_small_pages * sizeof(struct page *);
-	gfp_mask |= __GFP_NOWARN;
+
 	if (!(gfp_mask & (GFP_DMA | GFP_DMA32)))
 		gfp_mask |= __GFP_HIGHMEM;

@@ -3051,10 +3051,12 @@  static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask,
 		return NULL;
 	}

+	gfp_mask |= __GFP_NOWARN;
+
 	set_vm_area_page_order(area, page_shift - PAGE_SHIFT);
 	page_order = vm_area_page_order(area);

-	area->nr_pages = vm_area_alloc_pages(gfp_mask | __GFP_NOWARN,
+	area->nr_pages = vm_area_alloc_pages(gfp_mask,
 		node, page_order, nr_small_pages, area->pages);

 	atomic_long_add(area->nr_pages, &nr_vmalloc_pages);