[PATCH net-next] mlxsw: pci: Allocate skbs using GFP_KERNEL during initialization

[PATCH net-next] mlxsw: pci: Allocate skbs using GFP_KERNEL during initialization

[Petr Machata]

From: Ido Schimmel <idosch@nvidia.com>

The driver allocates skbs during initialization and during Rx
processing. Take advantage of the fact that the former happens in
process context and allocate the skbs using GFP_KERNEL to decrease the
probability of allocation failure.

Tested with CONFIG_DEBUG_ATOMIC_SLEEP=y.

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlxsw/pci.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci.c b/drivers/net/ethernet/mellanox/mlxsw/pci.c
index 51eea1f0529c..7fae963b2608 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/pci.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/pci.c
@@ -352,14 +352,15 @@ static void mlxsw_pci_wqe_frag_unmap(struct mlxsw_pci *mlxsw_pci, char *wqe,
 }
 
 static int mlxsw_pci_rdq_skb_alloc(struct mlxsw_pci *mlxsw_pci,
-				   struct mlxsw_pci_queue_elem_info *elem_info)
+				   struct mlxsw_pci_queue_elem_info *elem_info,
+				   gfp_t gfp)
 {
 	size_t buf_len = MLXSW_PORT_MAX_MTU;
 	char *wqe = elem_info->elem;
 	struct sk_buff *skb;
 	int err;
 
-	skb = netdev_alloc_skb_ip_align(NULL, buf_len);
+	skb = __netdev_alloc_skb_ip_align(NULL, buf_len, gfp);
 	if (!skb)
 		return -ENOMEM;
 
@@ -420,7 +421,7 @@ static int mlxsw_pci_rdq_init(struct mlxsw_pci *mlxsw_pci, char *mbox,
 	for (i = 0; i < q->count; i++) {
 		elem_info = mlxsw_pci_queue_elem_info_producer_get(q);
 		BUG_ON(!elem_info);
-		err = mlxsw_pci_rdq_skb_alloc(mlxsw_pci, elem_info);
+		err = mlxsw_pci_rdq_skb_alloc(mlxsw_pci, elem_info, GFP_KERNEL);
 		if (err)
 			goto rollback;
 		/* Everything is set up, ring doorbell to pass elem to HW */
@@ -640,7 +641,7 @@ static void mlxsw_pci_cqe_rdq_handle(struct mlxsw_pci *mlxsw_pci,
 	if (q->consumer_counter++ != consumer_counter_limit)
 		dev_dbg_ratelimited(&pdev->dev, "Consumer counter does not match limit in RDQ\n");
 
-	err = mlxsw_pci_rdq_skb_alloc(mlxsw_pci, elem_info);
+	err = mlxsw_pci_rdq_skb_alloc(mlxsw_pci, elem_info, GFP_ATOMIC);
 	if (err) {
 		dev_err_ratelimited(&pdev->dev, "Failed to alloc skb for RDQ\n");
 		goto out;
-- 
2.41.0

Re: [PATCH net-next] mlxsw: pci: Allocate skbs using GFP_KERNEL during initialization

[Jiri Pirko]

Wed, Oct 11, 2023 at 04:39:12PM CEST, petrm@nvidia.com wrote:
>From: Ido Schimmel <idosch@nvidia.com>
>
>The driver allocates skbs during initialization and during Rx
>processing. Take advantage of the fact that the former happens in
>process context and allocate the skbs using GFP_KERNEL to decrease the
>probability of allocation failure.
>
>Tested with CONFIG_DEBUG_ATOMIC_SLEEP=y.
>
>Signed-off-by: Ido Schimmel <idosch@nvidia.com>
>Reviewed-by: Petr Machata <petrm@nvidia.com>
>Signed-off-by: Petr Machata <petrm@nvidia.com>

Reviewed-by: Jiri Pirko <jiri@nvidia.com>

Re: [PATCH net-next] mlxsw: pci: Allocate skbs using GFP_KERNEL during initialization

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Wed, 11 Oct 2023 16:39:12 +0200 you wrote:
> From: Ido Schimmel <idosch@nvidia.com>
> 
> The driver allocates skbs during initialization and during Rx
> processing. Take advantage of the fact that the former happens in
> process context and allocate the skbs using GFP_KERNEL to decrease the
> probability of allocation failure.
> 
> [...]

Here is the summary with links:
  - [net-next] mlxsw: pci: Allocate skbs using GFP_KERNEL during initialization
    https://git.kernel.org/netdev/net-next/c/958a140d7a0a

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html