[PATCH net v4] net: axienet: cap the TX poll return value at the NAPI budget

[PATCH net v4] net: axienet: cap the TX poll return value at the NAPI budget

[Sagi Maimon]

axienet_tx_poll() reclaims every completed TX descriptor in one pass:
it passes lp->tx_bd_num to axienet_free_tx_chain() as the descriptor
limit, and @budget is only used as the napi_consume_skb() bulk-free
hint.  It then returns the number of packets reclaimed, which is bounded
by the ring size rather than by the budget, so the poll can report more
work than it was given.  This was seen after TX completion interrupts
had not been taken for a while and a full ring was reclaimed at once:

  eth0: NAPI poll function axienet_tx_poll+0x0/0x180 [xilinx_emac]
        returned 96, exceeding its budget of 64.

netpoll is affected as well.  poll_one_napi() polls with a budget of 0
to reclaim the TX path only, and warns once if any work is reported.

Keep reclaiming the whole ring and cap only the value returned.
Documentation/networking/napi.rst allows a poll to process any number
of TX completions; it is the reported work that must stay within the
budget.  Stopping the reclaim at the budget instead would also leave
completed descriptors for a later poll, which does not come when
napi_disable() is pending.  When more than @budget packets were
reclaimed, returning @budget keeps the poll scheduled, and the next poll
completes NAPI and re-enables the interrupt as before.  A budget of 0
now yields 0.

Suggested-by: Paolo Abeni <pabeni@redhat.com>
Fixes: 5a6caa2cfabb ("net: xilinx: axienet: Fix packet counting")
Assisted-by: LLM sparse
Signed-off-by: Sagi Maimon <maimon.sagi@gmail.com>
---

Notes:
    Changes in v4:
    - Different approach, as Paolo Abeni and the review of v3 suggested:
      keep reclaiming the whole ring and cap only the value returned,
      instead of stopping the reclaim at the budget.  napi.rst allows any
      number of TX completions per poll, and stopping early could leave
      completed descriptors behind while napi_disable() is pending.
    - Subject changed to match; v1-v3 were "net: axienet: bound TX
      completion cleanup by the NAPI budget".
    - Fixes: now names 5a6caa2cfabb, which made the return value unbounded
      (review of v3).
    - Only axienet_tx_poll() and its kernel-doc change now.
    - Suggested-by: Paolo Abeni.
    - Build-tested only so far; v3 was tested on hardware, v4 not yet.
    - v3: https://lore.kernel.org/netdev/20260924135052.185129-1-maimon.sagi@gmail.com/
    
    Changes in v3:
    - Report no work for a budget of 0, so netpoll cannot trip the
      WARN_ONCE() in poll_one_napi().
    - Add the Assisted-by: tag that v1 and v2 omitted.
    - v2: https://lore.kernel.org/netdev/20260917115657.20697-1-maimon.sagi@gmail.com/
    
    Changes in v2:
    - Treat a budget of 0 as no limit, so that netpoll still drains the TX
      ring.
    - v1: https://lore.kernel.org/netdev/20260914114821.55503-1-maimon.sagi@gmail.com/

 drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
index 1722b7038f34..243b07fd5be8 100644
--- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
+++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
@@ -984,9 +984,9 @@ axienet_start_xmit_dmaengine(struct sk_buff *skb, struct net_device *ndev)
  * axienet_tx_poll - Invoked once a transmit is completed by the
  * Axi DMA Tx channel.
  * @napi:	Pointer to NAPI structure.
- * @budget:	Max number of TX packets to process.
+ * @budget:	NAPI budget, or 0 when polled by netpoll.
  *
- * Return: Number of TX packets processed.
+ * Return: Number of TX packets processed, capped at @budget.
  *
  * This function is invoked from the NAPI processing to notify the completion
  * of transmit operation. It clears fields in the corresponding Tx BDs and
@@ -1027,7 +1027,12 @@ static int axienet_tx_poll(struct napi_struct *napi, int budget)
 		axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, lp->tx_dma_cr);
 		spin_unlock_irq(&lp->tx_cr_lock);
 	}
-	return packets;
+
+	/* The whole ring was reclaimed above, which may be more than the
+	 * budget, but a poll must not report more work than it was given.
+	 * netpoll polls with a budget of 0 and expects no work reported.
+	 */
+	return min(packets, budget);
 }
 
 /**

base-commit: a7bfaba4823e3c165bb2004c74eff7c096672bc7
-- 
2.47.0

Re: [PATCH net v4] net: axienet: cap the TX poll return value at the NAPI budget

[netdev-bot+sinfo]

Hi!

This is an automated message. This series looks like a fix, but its
commit messages seem to be missing some information:

 - What hardware the change was tested on. For driver fixes please
   mention the device (and if relevant firmware version) used for
   testing, or say that the change was not tested on real hardware.

Please do not repost the series just to address the above. Instead,
reply to this email with the missing information, so that reviewers
can take it into account. If the series needs another revision for
other reasons, please include the information in the commit messages
then.

The evaluation is done by an LLM so it may be wrong, if you think
that is the case please reply and explain.

Re: [PATCH net v4] net: axienet: cap the TX poll return value at the NAPI budget

[Gupta, Suraj]

On 9/30/2026 12:45 PM, Sagi Maimon wrote:
> axienet_tx_poll() reclaims every completed TX descriptor in one pass:
> it passes lp->tx_bd_num to axienet_free_tx_chain() as the descriptor
> limit, and @budget is only used as the napi_consume_skb() bulk-free
> hint.  It then returns the number of packets reclaimed, which is bounded
> by the ring size rather than by the budget, so the poll can report more
> work than it was given.  This was seen after TX completion interrupts
> had not been taken for a while and a full ring was reclaimed at once:
> 
>    eth0: NAPI poll function axienet_tx_poll+0x0/0x180 [xilinx_emac]
>          returned 96, exceeding its budget of 64.
> 
> netpoll is affected as well.  poll_one_napi() polls with a budget of 0
> to reclaim the TX path only, and warns once if any work is reported.
> 
> Keep reclaiming the whole ring and cap only the value returned.
> Documentation/networking/napi.rst allows a poll to process any number
> of TX completions; it is the reported work that must stay within the
> budget.  Stopping the reclaim at the budget instead would also leave
> completed descriptors for a later poll, which does not come when
> napi_disable() is pending.  When more than @budget packets were
> reclaimed, returning @budget keeps the poll scheduled, and the next poll
> completes NAPI and re-enables the interrupt as before.  A budget of 0
> now yields 0.
> 
> Suggested-by: Paolo Abeni <pabeni@redhat.com>
> Fixes: 5a6caa2cfabb ("net: xilinx: axienet: Fix packet counting")
> Assisted-by: LLM sparse
> Signed-off-by: Sagi Maimon <maimon.sagi@gmail.com>
> ---
> 

LGTM

Reviewed-by: Suraj Gupta <suraj.gupta2@amd.com>

Thanks,
Suraj

> Notes:
>      Changes in v4:
>      - Different approach, as Paolo Abeni and the review of v3 suggested:
>        keep reclaiming the whole ring and cap only the value returned,
>        instead of stopping the reclaim at the budget.  napi.rst allows any
>        number of TX completions per poll, and stopping early could leave
>        completed descriptors behind while napi_disable() is pending.
>      - Subject changed to match; v1-v3 were "net: axienet: bound TX
>        completion cleanup by the NAPI budget".
>      - Fixes: now names 5a6caa2cfabb, which made the return value unbounded
>        (review of v3).
>      - Only axienet_tx_poll() and its kernel-doc change now.
>      - Suggested-by: Paolo Abeni.
>      - Build-tested only so far; v3 was tested on hardware, v4 not yet.
>      - v3: https://lore.kernel.org/netdev/20260924135052.185129-1-maimon.sagi@gmail.com/
>      
>      Changes in v3:
>      - Report no work for a budget of 0, so netpoll cannot trip the
>        WARN_ONCE() in poll_one_napi().
>      - Add the Assisted-by: tag that v1 and v2 omitted.
>      - v2: https://lore.kernel.org/netdev/20260917115657.20697-1-maimon.sagi@gmail.com/
>      
>      Changes in v2:
>      - Treat a budget of 0 as no limit, so that netpoll still drains the TX
>        ring.
>      - v1: https://lore.kernel.org/netdev/20260914114821.55503-1-maimon.sagi@gmail.com/
> 
>   drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 11 ++++++++---
>   1 file changed, 8 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> index 1722b7038f34..243b07fd5be8 100644
> --- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c
> @@ -984,9 +984,9 @@ axienet_start_xmit_dmaengine(struct sk_buff *skb, struct net_device *ndev)
>    * axienet_tx_poll - Invoked once a transmit is completed by the
>    * Axi DMA Tx channel.
>    * @napi:	Pointer to NAPI structure.
> - * @budget:	Max number of TX packets to process.
> + * @budget:	NAPI budget, or 0 when polled by netpoll.
>    *
> - * Return: Number of TX packets processed.
> + * Return: Number of TX packets processed, capped at @budget.
>    *
>    * This function is invoked from the NAPI processing to notify the completion
>    * of transmit operation. It clears fields in the corresponding Tx BDs and
> @@ -1027,7 +1027,12 @@ static int axienet_tx_poll(struct napi_struct *napi, int budget)
>   		axienet_dma_out32(lp, XAXIDMA_TX_CR_OFFSET, lp->tx_dma_cr);
>   		spin_unlock_irq(&lp->tx_cr_lock);
>   	}
> -	return packets;
> +
> +	/* The whole ring was reclaimed above, which may be more than the
> +	 * budget, but a poll must not report more work than it was given.
> +	 * netpoll polls with a budget of 0 and expects no work reported.
> +	 */
> +	return min(packets, budget);
>   }
>   
>   /**
> 
> base-commit: a7bfaba4823e3c165bb2004c74eff7c096672bc7