[PATCH] percpu: plumb gfp flag to pcpu_get_pages

[PATCH] percpu: plumb gfp flag to pcpu_get_pages

[Shakeel Butt]

__alloc_percpu_gfp() can be called from atomic context, so, make
pcpu_get_pages use the gfp provided to the higher layer.

Signed-off-by: Shakeel Butt <shakeelb@google.com>
---
 mm/percpu-vm.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/mm/percpu-vm.c b/mm/percpu-vm.c
index d8078de912de..4f42c4c5c902 100644
--- a/mm/percpu-vm.c
+++ b/mm/percpu-vm.c
@@ -21,6 +21,7 @@ static struct page *pcpu_chunk_page(struct pcpu_chunk *chunk,
 
 /**
  * pcpu_get_pages - get temp pages array
+ * @gfp: allocation flags passed to the underlying allocator
  *
  * Returns pointer to array of pointers to struct page which can be indexed
  * with pcpu_page_idx().  Note that there is only one array and accesses
@@ -29,7 +30,7 @@ static struct page *pcpu_chunk_page(struct pcpu_chunk *chunk,
  * RETURNS:
  * Pointer to temp pages array on success.
  */
-static struct page **pcpu_get_pages(void)
+static struct page **pcpu_get_pages(gfp_t gfp)
 {
 	static struct page **pages;
 	size_t pages_size = pcpu_nr_units * pcpu_unit_pages * sizeof(pages[0]);
@@ -37,7 +38,7 @@ static struct page **pcpu_get_pages(void)
 	lockdep_assert_held(&pcpu_alloc_mutex);
 
 	if (!pages)
-		pages = pcpu_mem_zalloc(pages_size, GFP_KERNEL);
+		pages = pcpu_mem_zalloc(pages_size, gfp);
 	return pages;
 }
 
@@ -278,7 +279,7 @@ static int pcpu_populate_chunk(struct pcpu_chunk *chunk,
 {
 	struct page **pages;
 
-	pages = pcpu_get_pages();
+	pages = pcpu_get_pages(gfp);
 	if (!pages)
 		return -ENOMEM;
 
@@ -316,7 +317,7 @@ static void pcpu_depopulate_chunk(struct pcpu_chunk *chunk,
 	 * successful population attempt so the temp pages array must
 	 * be available now.
 	 */
-	pages = pcpu_get_pages();
+	pages = pcpu_get_pages(GFP_KERNEL);
 	BUG_ON(!pages);
 
 	/* unmap and free */
-- 
2.20.1.415.g653613c723-goog

Re: [PATCH] percpu: plumb gfp flag to pcpu_get_pages

[Andrew Morton]

On Fri, 28 Dec 2018 17:31:47 -0800 Shakeel Butt <shakeelb@google.com> wrote:

> __alloc_percpu_gfp() can be called from atomic context, so, make
> pcpu_get_pages use the gfp provided to the higher layer.

Does this fix any user-visible issues?

Re: [PATCH] percpu: plumb gfp flag to pcpu_get_pages

[Dennis Zhou]

Hi Shakeel,

On Fri, Dec 28, 2018 at 05:31:47PM -0800, Shakeel Butt wrote:
> __alloc_percpu_gfp() can be called from atomic context, so, make
> pcpu_get_pages use the gfp provided to the higher layer.
> 
> Signed-off-by: Shakeel Butt <shakeelb@google.com>
> ---
>  mm/percpu-vm.c | 9 +++++----
>  1 file changed, 5 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/percpu-vm.c b/mm/percpu-vm.c
> index d8078de912de..4f42c4c5c902 100644
> --- a/mm/percpu-vm.c
> +++ b/mm/percpu-vm.c
> @@ -21,6 +21,7 @@ static struct page *pcpu_chunk_page(struct pcpu_chunk *chunk,
>  
>  /**
>   * pcpu_get_pages - get temp pages array
> + * @gfp: allocation flags passed to the underlying allocator
>   *
>   * Returns pointer to array of pointers to struct page which can be indexed
>   * with pcpu_page_idx().  Note that there is only one array and accesses
> @@ -29,7 +30,7 @@ static struct page *pcpu_chunk_page(struct pcpu_chunk *chunk,
>   * RETURNS:
>   * Pointer to temp pages array on success.
>   */
> -static struct page **pcpu_get_pages(void)
> +static struct page **pcpu_get_pages(gfp_t gfp)
>  {
>  	static struct page **pages;
>  	size_t pages_size = pcpu_nr_units * pcpu_unit_pages * sizeof(pages[0]);
> @@ -37,7 +38,7 @@ static struct page **pcpu_get_pages(void)
>  	lockdep_assert_held(&pcpu_alloc_mutex);
>  
>  	if (!pages)
> -		pages = pcpu_mem_zalloc(pages_size, GFP_KERNEL);
> +		pages = pcpu_mem_zalloc(pages_size, gfp);
>  	return pages;
>  }
>  
> @@ -278,7 +279,7 @@ static int pcpu_populate_chunk(struct pcpu_chunk *chunk,
>  {
>  	struct page **pages;
>  
> -	pages = pcpu_get_pages();
> +	pages = pcpu_get_pages(gfp);
>  	if (!pages)
>  		return -ENOMEM;
>  
> @@ -316,7 +317,7 @@ static void pcpu_depopulate_chunk(struct pcpu_chunk *chunk,
>  	 * successful population attempt so the temp pages array must
>  	 * be available now.
>  	 */
> -	pages = pcpu_get_pages();
> +	pages = pcpu_get_pages(GFP_KERNEL);
>  	BUG_ON(!pages);
>  
>  	/* unmap and free */
> -- 
> 2.20.1.415.g653613c723-goog
> 

Sorry, I'm travelling today and was hoping to respond to this later
tonight.

So percpu memory is a little different as it's an intermediary. When you 
call __alloc_percpu_gfp() and it does not contain GFP_KERNEL, it is
considered atomic. So, we only service requests out of already
populated memory. pcpu_get_pages() is only called when we need to
populate/depopulate a chunk and will not be called if we need an atomic
allocation. Also, in all but the first case, it won't make an allocation
as pages is a static variable.

Furthermore, percpu only plumbs through certain gfp as not all make
sense [1].

[1] https://lore.kernel.org/lkml/cover.1518668149.git.dennisszhou@gmail.com/

Thanks,
Dennis

Re: [PATCH] percpu: plumb gfp flag to pcpu_get_pages

[Dennis Zhou]

Hi Andrew,

On Sat, Dec 29, 2018 at 01:03:52PM -0800, Andrew Morton wrote:
> On Fri, 28 Dec 2018 17:31:47 -0800 Shakeel Butt <shakeelb@google.com> wrote:
> 
> > __alloc_percpu_gfp() can be called from atomic context, so, make
> > pcpu_get_pages use the gfp provided to the higher layer.
> 
> Does this fix any user-visible issues?

Sorry for not getting to this earlier. I'm currently traveling. I
respoeded on the patch itself. Do you mind unqueuing? I explain in more
detail on the patch, but __alloc_percpu_gfp() will never call
pcpu_get_pages() when called as not GFP_KERNEL.

Thanks,
Dennis

Re: [PATCH] percpu: plumb gfp flag to pcpu_get_pages

[Shakeel Butt]

Hi Dennis,

On Sat, Dec 29, 2018 at 1:26 PM Dennis Zhou <dennis@kernel.org> wrote:
>
> Hi Andrew,
>
> On Sat, Dec 29, 2018 at 01:03:52PM -0800, Andrew Morton wrote:
> > On Fri, 28 Dec 2018 17:31:47 -0800 Shakeel Butt <shakeelb@google.com> wrote:
> >
> > > __alloc_percpu_gfp() can be called from atomic context, so, make
> > > pcpu_get_pages use the gfp provided to the higher layer.
> >
> > Does this fix any user-visible issues?
>
> Sorry for not getting to this earlier. I'm currently traveling. I
> respoeded on the patch itself. Do you mind unqueuing? I explain in more
> detail on the patch, but __alloc_percpu_gfp() will never call
> pcpu_get_pages() when called as not GFP_KERNEL.
>

Thanks for the explanation. Andrew, please ignore/drop this patch.

thanks,
Shakeel