[PATCH net-next] net: ps3_gelic_net: Use napi_alloc_skb() and napi_gro_receive()

[PATCH net-next] net: ps3_gelic_net: Use napi_alloc_skb() and napi_gro_receive()

[Florian Fuchs]

Use the napi functions napi_alloc_skb() and napi_gro_receive() instead
of netdev_alloc_skb() and netif_receive_skb() for more efficient packet
receiving. The switch to napi aware functions increases the RX
throughput, reduces the occurrence of retransmissions and improves the
resilience against SKB allocation failures.

Signed-off-by: Florian Fuchs <fuchsfl@gmail.com>
---
Note: This change has been tested on real hardware Sony PS3 (CECHL04 PAL),
the patch was tested for many hours, with continuous system load, high
network transfer load and injected failslab errors.

In my tests, the RX throughput increased up to 100% and reduced the
occurrence of retransmissions drastically, with GRO enabled:

iperf3 before and after the commit, where PS3 (with this driver) is on
the receiving side:
Before: [  5]   0.00-10.00  sec   551 MBytes   462 Mbits/sec receiver
After:  [  5]   0.00-10.00  sec  1.09 GBytes   939 Mbits/sec receiver

stats from the sending client to the PS3:
Before: [  5]   0.00-10.00  sec   552 MBytes   463 Mbits/sec  3151 sender
After:  [  5]   0.00-10.00  sec  1.09 GBytes   940 Mbits/sec   37  sender

 drivers/net/ethernet/toshiba/ps3_gelic_net.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/toshiba/ps3_gelic_net.c b/drivers/net/ethernet/toshiba/ps3_gelic_net.c
index 591866fc9055..d35d1f3c10a1 100644
--- a/drivers/net/ethernet/toshiba/ps3_gelic_net.c
+++ b/drivers/net/ethernet/toshiba/ps3_gelic_net.c
@@ -364,6 +364,7 @@ static int gelic_card_init_chain(struct gelic_card *card,
  * gelic_descr_prepare_rx - reinitializes a rx descriptor
  * @card: card structure
  * @descr: descriptor to re-init
+ * @napi_mode: is it running in napi poll
  *
  * return 0 on success, <0 on failure
  *
@@ -374,7 +375,8 @@ static int gelic_card_init_chain(struct gelic_card *card,
  * must be a multiple of GELIC_NET_RXBUF_ALIGN.
  */
 static int gelic_descr_prepare_rx(struct gelic_card *card,
-				  struct gelic_descr *descr)
+				  struct gelic_descr *descr,
+				  bool napi_mode)
 {
 	static const unsigned int rx_skb_size =
 		ALIGN(GELIC_NET_MAX_FRAME, GELIC_NET_RXBUF_ALIGN) +
@@ -392,7 +394,10 @@ static int gelic_descr_prepare_rx(struct gelic_card *card,
 	descr->hw_regs.payload.dev_addr = 0;
 	descr->hw_regs.payload.size = 0;
 
-	descr->skb = netdev_alloc_skb(*card->netdev, rx_skb_size);
+	if (napi_mode)
+		descr->skb = napi_alloc_skb(&card->napi, rx_skb_size);
+	else
+		descr->skb = netdev_alloc_skb(*card->netdev, rx_skb_size);
 	if (!descr->skb) {
 		descr->hw_regs.payload.dev_addr = 0; /* tell DMAC don't touch memory */
 		return -ENOMEM;
@@ -464,7 +469,7 @@ static int gelic_card_fill_rx_chain(struct gelic_card *card)
 
 	do {
 		if (!descr->skb) {
-			ret = gelic_descr_prepare_rx(card, descr);
+			ret = gelic_descr_prepare_rx(card, descr, false);
 			if (ret)
 				goto rewind;
 		}
@@ -964,7 +969,7 @@ static void gelic_net_pass_skb_up(struct gelic_descr *descr,
 	netdev->stats.rx_bytes += skb->len;
 
 	/* pass skb up to stack */
-	netif_receive_skb(skb);
+	napi_gro_receive(&card->napi, skb);
 }
 
 /**
@@ -1069,7 +1074,7 @@ static int gelic_card_decode_one_descr(struct gelic_card *card)
 	/*
 	 * this call can fail, propagate the error
 	 */
-	prepare_rx_ret = gelic_descr_prepare_rx(card, descr);
+	prepare_rx_ret = gelic_descr_prepare_rx(card, descr, true);
 	if (prepare_rx_ret)
 		return prepare_rx_ret;
 

base-commit: ff736a286116d462a4067ba258fa351bc0b4ed80
-- 
2.43.0

Re: [PATCH net-next] net: ps3_gelic_net: Use napi_alloc_skb() and napi_gro_receive()

[Simon Horman]

On Sun, Nov 30, 2025 at 08:41:55PM +0100, Florian Fuchs wrote:
> Use the napi functions napi_alloc_skb() and napi_gro_receive() instead
> of netdev_alloc_skb() and netif_receive_skb() for more efficient packet
> receiving. The switch to napi aware functions increases the RX
> throughput, reduces the occurrence of retransmissions and improves the
> resilience against SKB allocation failures.
> 
> Signed-off-by: Florian Fuchs <fuchsfl@gmail.com>
> ---
> Note: This change has been tested on real hardware Sony PS3 (CECHL04 PAL),
> the patch was tested for many hours, with continuous system load, high
> network transfer load and injected failslab errors.
> 
> In my tests, the RX throughput increased up to 100% and reduced the
> occurrence of retransmissions drastically, with GRO enabled:
> 
> iperf3 before and after the commit, where PS3 (with this driver) is on
> the receiving side:
> Before: [  5]   0.00-10.00  sec   551 MBytes   462 Mbits/sec receiver
> After:  [  5]   0.00-10.00  sec  1.09 GBytes   939 Mbits/sec receiver
> 
> stats from the sending client to the PS3:
> Before: [  5]   0.00-10.00  sec   552 MBytes   463 Mbits/sec  3151 sender
> After:  [  5]   0.00-10.00  sec  1.09 GBytes   940 Mbits/sec   37  sender

Hi Florian,

Thanks for the rest results and confirming this has
been exercised on real HW.

Thinking out loud:

* I see that the napi_mode argument to gelic_descr_prepare_rx ensures
  that napi_alloc_skb() is only called from softirq context.

* I see that the driver already calls napi_complete_done() in
  it's poll callback, a pre-requisite for using napi_alloc_skb().

So as I understand things the use of the NAPI API by this patch is correct.

And this provides a nice example of the advantages of using this part
of the API.

Reviewed-by: Simon Horman <horms@kernel.org>

Re: [PATCH net-next] net: ps3_gelic_net: Use napi_alloc_skb() and napi_gro_receive()

[Eric Dumazet]

On Sun, Nov 30, 2025 at 11:51 AM Florian Fuchs <fuchsfl@gmail.com> wrote:
>
> Use the napi functions napi_alloc_skb() and napi_gro_receive() instead
> of netdev_alloc_skb() and netif_receive_skb() for more efficient packet
> receiving. The switch to napi aware functions increases the RX
> throughput, reduces the occurrence of retransmissions and improves the
> resilience against SKB allocation failures.
>
> Signed-off-by: Florian Fuchs <fuchsfl@gmail.com>
> ---
> Note: This change has been tested on real hardware Sony PS3 (CECHL04 PAL),
> the patch was tested for many hours, with continuous system load, high
> network transfer load and injected failslab errors.
>
> In my tests, the RX throughput increased up to 100% and reduced the
> occurrence of retransmissions drastically, with GRO enabled:
>
> iperf3 before and after the commit, where PS3 (with this driver) is on
> the receiving side:
> Before: [  5]   0.00-10.00  sec   551 MBytes   462 Mbits/sec receiver
> After:  [  5]   0.00-10.00  sec  1.09 GBytes   939 Mbits/sec receiver
>
> stats from the sending client to the PS3:
> Before: [  5]   0.00-10.00  sec   552 MBytes   463 Mbits/sec  3151 sender
> After:  [  5]   0.00-10.00  sec  1.09 GBytes   940 Mbits/sec   37  sender
>
>  drivers/net/ethernet/toshiba/ps3_gelic_net.c | 15 ++++++++++-----
>  1 file changed, 10 insertions(+), 5 deletions(-)

Patch looks fine to me. My PS3 died years ago, so I can not test it :)

Reviewed-by: Eric Dumazet <edumazet@google.com>

BTW, I think we can cleanup gelic_descr_prepare_rx() a bit :

diff --git a/drivers/net/ethernet/toshiba/ps3_gelic_net.c
b/drivers/net/ethernet/toshiba/ps3_gelic_net.c
index 5ee8e8980393c3491bf9cf91eb8e0dbb2df0f427..f4f34e9ed49c5b7fd1cf4ea3f5bfe7297e97ea23
100644
--- a/drivers/net/ethernet/toshiba/ps3_gelic_net.c
+++ b/drivers/net/ethernet/toshiba/ps3_gelic_net.c
@@ -392,10 +392,8 @@ static int gelic_descr_prepare_rx(struct gelic_card *card,
        descr->hw_regs.payload.size = 0;

        descr->skb = netdev_alloc_skb(*card->netdev, rx_skb_size);
-       if (!descr->skb) {
-               descr->hw_regs.payload.dev_addr = 0; /* tell DMAC
don't touch memory */
+       if (!descr->skb)
                return -ENOMEM;
-       }

        offset = ((unsigned long)descr->skb->data) &
                (GELIC_NET_RXBUF_ALIGN - 1);
@@ -404,13 +402,12 @@ static int gelic_descr_prepare_rx(struct gelic_card *card,
        /* io-mmu-map the skb */
        cpu_addr = dma_map_single(ctodev(card), descr->skb->data,
                                  GELIC_NET_MAX_FRAME, DMA_FROM_DEVICE);
-       descr->hw_regs.payload.dev_addr = cpu_to_be32(cpu_addr);
+
        if (dma_mapping_error(ctodev(card), cpu_addr)) {
                dev_kfree_skb_any(descr->skb);
                descr->skb = NULL;
                dev_info(ctodev(card),
                         "%s:Could not iommu-map rx buffer\n", __func__);
-               gelic_descr_set_status(descr, GELIC_DESCR_DMA_NOT_IN_USE);
                return -ENOMEM;
        }

Re: [PATCH net-next] net: ps3_gelic_net: Use napi_alloc_skb() and napi_gro_receive()

[patchwork-bot+netdevbpf]

Hello:

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

On Sun, 30 Nov 2025 20:41:55 +0100 you wrote:
> Use the napi functions napi_alloc_skb() and napi_gro_receive() instead
> of netdev_alloc_skb() and netif_receive_skb() for more efficient packet
> receiving. The switch to napi aware functions increases the RX
> throughput, reduces the occurrence of retransmissions and improves the
> resilience against SKB allocation failures.
> 
> Signed-off-by: Florian Fuchs <fuchsfl@gmail.com>
> 
> [...]

Here is the summary with links:
  - [net-next] net: ps3_gelic_net: Use napi_alloc_skb() and napi_gro_receive()
    https://git.kernel.org/netdev/net-next/c/d8e08149a5ed

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