Re: [PATCH net-next 4/4] bnxt_en: add page_pool support

[Andy Gospodarek]

On Sat, Jul 06, 2019 at 03:36:18AM -0400, Michael Chan wrote:
> From: Andy Gospodarek <go...@broadcom.com>
> 
> This removes contention over page allocation for XDP_REDIRECT actions by
> adding page_pool support per queue for the driver.  The performance for
> XDP_REDIRECT actions scales linearly with the number of cores performing
> redirect actions when using the page pools instead of the standard page
> allocator.
> 
> Signed-off-by: Andy Gospodarek <go...@broadcom.com>
> Signed-off-by: Michael Chan <michael.c...@broadcom.com>
> ---
>  drivers/net/ethernet/broadcom/Kconfig         |  1 +
>  drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 40 
> +++++++++++++++++++++++----
>  drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  3 ++
>  drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |  3 +-
>  4 files changed, 41 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c 
> b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> index d8f0846..b6777e5 100644
> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
[...]
> @@ -2530,12 +2555,17 @@ static int bnxt_alloc_rx_rings(struct bnxt *bp)
>  
>               ring = &rxr->rx_ring_struct;
>  
> +             rc = bnxt_alloc_rx_page_pool(bp, rxr);
> +             if (rc)
> +                     return rc;
> +
>               rc = xdp_rxq_info_reg(&rxr->xdp_rxq, bp->dev, i);
>               if (rc < 0)
>                       return rc;
>  
>               rc = xdp_rxq_info_reg_mem_model(&rxr->xdp_rxq,
> -                                             MEM_TYPE_PAGE_SHARED, NULL);
> +                                             MEM_TYPE_PAGE_POOL,
> +                                             rxr->page_pool);
>               if (rc) {
>                       xdp_rxq_info_unreg(&rxr->xdp_rxq);
>                       return rc;

I think we want to amend and the chunk above to be:

@@ -2530,14 +2557,24 @@ static int bnxt_alloc_rx_rings(struct bnxt *bp)
 
                ring = &rxr->rx_ring_struct;
 
+               rc = bnxt_alloc_rx_page_pool(bp, rxr);
+               if (rc)
+                       return rc;
+
                rc = xdp_rxq_info_reg(&rxr->xdp_rxq, bp->dev, i);
-               if (rc < 0)
+               if (rc < 0) {
+                       page_pool_free(rxr->page_pool);
+                       rxr->page_pool = NULL;
                        return rc;
+               }
 
                rc = xdp_rxq_info_reg_mem_model(&rxr->xdp_rxq,
-                                               MEM_TYPE_PAGE_SHARED, NULL);
+                                               MEM_TYPE_PAGE_POOL,
+                                               rxr->page_pool);
                if (rc) {
                        xdp_rxq_info_unreg(&rxr->xdp_rxq);
+                       page_pool_free(rxr->page_pool);
+                       rxr->page_pool = NULL;
                        return rc;
                }
 

That should take care of the freeing of the page_pool that is allocated
but there is a failure during xdp_rxq_info_reg() or
xdp_rxq_info_reg_mem_model().

I agree that we do not need to call page_pool_free in the normal
shutdown case.